From 4e2dcd827bae864a1d16d05eda6520acf0a39f21 Mon Sep 17 00:00:00 2001
From: "Sterling G. Baird" <sterling.baird@utoronto.ca>
Date: Wed, 1 May 2024 15:20:20 -0400
Subject: [PATCH 01/12] Create mobo-tutorial-thumbnail.jpg

---
 docs/tutorials/mobo-tutorial-thumbnail.jpg | Bin 0 -> 543723 bytes
 1 file changed, 0 insertions(+), 0 deletions(-)
 create mode 100644 docs/tutorials/mobo-tutorial-thumbnail.jpg

diff --git a/docs/tutorials/mobo-tutorial-thumbnail.jpg b/docs/tutorials/mobo-tutorial-thumbnail.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..912d28fcdb75535dfa2b90e7bddb89fa611ad3c3
GIT binary patch
literal 543723
zcmeFZ2UL^Ux;GxjV=oMYfJi$^8>C6EAv36;)De`DAdG}!2q6N}1I!r<N*yVohl2=+
z5NZl7Fw$a#5Tzv4AT0=?g%AjY@{e=wy{Fu=zV+R6zyDfy&BNl&^X~fFzjyEb?(*#S
zVEEt@;A?Bko0fn>LQy}Y4gefXf8~Ai>Q&Dl933og+FloW003Xz4uB#;{%{2VKm<Mv
zcC@niqpKVEkH<a(jsOk;z5$p3Oz%NM9(@1954M2+R5t0tutFKo_>$It8tgyzIOXjV
z0tEmL9T$3=Lmvb`6zVsGx^4Kw2Ve9Ip?(f}|DLx{e=XG2gM|Qv`rsFLufOP@zvzB{
z(RaV-!M7c+3d{DjQ2(Rf-{@X{qeJfpBZOguB?l-X0z!otev>aHIqeg8$4U5mN%$`V
z2mv?(tN<2(KLTz7t^=+D3;@~yW8u{RpaIYUXbSHD|4IBWY0CiY0Z3t7sPOs#5D5qd
zzyN-L`+$c4b)jphFm)f{udmSENB9>2i2TC8FrHBU%4>+PuFh{xhfZ4n07sYy2V5lp
z;8;2Uu=mTsfuQK%VDBXWaCjU5_%-ma@sWQ90E~D-_h0_%rdS97h&%=W2pxZQ^U4AM
zi17fx`H=_rg75v7&tc*9D_>s#V7&|gIO7HYh<yM6PD1{|nb7ZxKcJWa0NjoiJ}N2z
zkdq4lC=0`jxF3uOIr>uP{yXv|k^fHo-{Kzp3J^Vh;_K5t9X=!q_)7H9VbMbeod9`Z
zWjze|>aZ}#mqL7z6DLm{JN}0wM-PAXB|z}kLXV?g0S+Jj>eSb#PaHoa4mkAHVWIc2
z<0nMFIVpBd{G#NAOWMXRp;-+>!xHDE6buZ%zvb!{7M_hGFm&`RobEl0$VB6bbVVb_
zkU#zWqVBD-uKpcouZ()7px~lcC8cHUwSy{GZ$s0HzkTWb=%X-CIU$%se-L8&3!pF8
z<42F3I4R8bH%vjIhX7xF^_39E;XfP`5jpy$6u)5-J@(CUv5P0pX^UU_{+5J-j^n+c
z5J_D<eFGzl;Lty%XVl^FiiTbf!@^&r71h=g-u4eFIo<x5*gl|q^<{C#(Au}x?l?oe
zGlif&o>zG!<$K-I$L~oe=U^0Y`miud(Ziwu5TNbXzry-m`<;Q`8Tg%n-x>Iwf!`VU
zoq^vO_?>~@8TjAK!0`jXGBQElD#`~)SIcW;KnN4dMP}wtO}vTH^9t=arKF9|-l54}
zHak1y&YLHS5s?ipj37J<4O3-y$EiXUV`TX^_&@RJ4dEstWi6#1ht)gxdyGfOjHs4=
zzp;SaaiRty07ZPClbZMFF`2Rq<w_7*tmxmS^ek#QEQSQ2+H|Y$zTY<sD{BsqqkT@w
z-Qh&ix8k5NjSMnUOf>G>WAj@6Wrq>rktuBCow~?q<Cs}>gBAGr@>-~0u4j3g|LD;d
zh<t7dDF1R)(YwLRJLWI_zp*oX<DRk;vuVC={NU3F&tYiKPMy0{kRn<1z%BBmb3G5|
zR2Q-xv7N|ZJIL3fGPYtLdgH8~*SXU=(HBPfI|l#=1>H7`2o{8PfZncr9H>IR;&b=p
z#dgfBW<(2H1UFli^6qvG*+px$=lfXT2BH%Yh6&j92lZ(ny}E&lcR5&AiQ8O!hA*bY
z-$rjg=!vmnk|O#J?*LFoo^pxrdfI}{idl|S73J!1Ngc|s=7|gVa=M<$9!!jXJxJiI
zHmyv@*_BSNMC~merD6l5A}}WPd>5zWsvDJYe^_Nr<P=U7n#bL?`uH{zh1zvnr0PM(
zPAs(RT3$)ZI>u|yuj`=E2DGw7c6LklyG3(hrAg+IrAe_7;}>G`3y)?zd$tj*F?|Wh
z;2j>4wxy+ZPwtvVa6w7@giePkQL(8R2W)I9?Eug_Am7!oCVBucl$b5B8P(uiV_g1F
zo;|y7JULrzR)4#Q_<;4?&wlY6a-WX7-Am}hH~l>zdaeN;QT7;AgSoMN-6jO&E8lLV
zYMa@=c(=HU%GP$zH;h;^UViGgM)-Eg)Sagc=Xd&+g7=%o7B}=`!g<mjUcu`o(zSAC
z-3;-P8^oE7!ba#IkwIpR>F=L-{HL`^ywiJUL+7xGKP`60*Sk7MsZO#Yq629QhJ?E9
z9(FT<VQ7RTjHnhJw>$23dxY~1RccC4>(yh-^BH8MK)xxZTV^}PfEQOA1QBJ8jshp2
z%tS6pm30ap%*`2T4X@mF<xcwnBa%lFI>{5ux+@8j3H-EKn7>OtI|_2HP!!eTL60$I
z@J5>iaWPdlcVbjc&$rnMX4dRME~rqHi2xYl^K(&;J|7g?nwzN>V^z3R8F2uZuDTp~
zb4A_I#Lb;7GjXTfLuYKn&7lD@x9W$<_||L2j8mowih&4#u_{2NuG)9A3k`w;z~O4i
zR(tqBOmYfkz1E4^>PtHk1C*mZ%_Pq&qj*y~3qNADj17VKM7wD=rf$Q&%l}@A{kZwW
z?)A{fCMT;2*iJNzD5dim>^o`QA69|pYA=&nsB6E1C%}^sh{uTghnD)5BJ0Kem43X*
zCNXSip~v}27b!~D!J2e(urv*O88?ry&S0S<m2+>jYd;>TnPc>19Ywia@q3%bufIpN
z@#;y`G_}+5>Pg{1W+*nD^bMTz5Y7%Es(q#D4ju_+qOlA9T^0VlU;?(7ihx*U|0X%l
zpL0bncJKUe?q2^%Wji#8ncRx2l9ifCXn8Sgw@Gs+Hl!`uL4F`xpk3l7>zwOs^o9%6
z?xc~Vd$AD?&df}fb<3v7ri7uvmh@3SF_<&t854PJ?-0*w`V`KoA_~UPUC#-EoN0_j
z!Y9k<JFF^*t>vm|FGO|?Hj~RxT3ej1A}h76*G<nn4A*rGQcW2cH#%ETWEOASY%7if
z>*Is?6cQ<|c7Dzue*NW$N^d7bsWg`#GekfQmz&SYH80Ny#$rbXsh@YR#)Q+&zi!5`
z`rPFbvl_C}!z#T89IQR5S&9%FpBK22NDFO#L7Z*bPhC`tX@ZtkX#EdqHjUi9wS`S&
zl-B2=i2b({p=&$!lWPnK*sSSUwDBU!d^{Sd{Wdh9zR(4@Hv25etk5*jCQvIOFUqwc
zA5AzPKf$0gX(MF@g_6-crrd!4Zh=|-ww0g!)PBWeb-mIh&iuq9$1Bjy(c$$muS*(N
zfM7hB=g+8iE{WnT-o}FQFjl!U9-M9&mFYi-(RFYvxyeczpy~O3=}HGIwUfNwC9B%r
zE;V~tY8J5ohLWwjVAjr;@Fi-l3H=$B*PLrb<t|;iWESD}JO@X`>i#AyUq804F32sX
zu(Ed_#EnG~KM8Bd355m2g!$kmg_smF$3fA3Dc{p#5!@(uZq{vOGuEKF2<RCd?jUjx
zyqMx(-O`xkTkfP#;=OjKD)6G>wy{pwy5(p@J$v#kt*G$p9RUPLuJD>}4s>_2^AOM#
zL_|kxX0gC*1gSn%oEN!tzL*QE=v!Pr&w+9;q?e~n8sN8n;HeQF_VnlbT|}gqXo#gh
zo7@WEY*rT<Cf&vtD;KQtawui#r$?D8ir&w1sk3IryIwn8(_%h;Z|<$kPY1reMRthH
znbPRFh;-aiIAv`vxEMz^RabR$MK<L52M<+)pOC6_VdT-Vai&jbo;u=k<jI}AT?f?H
zQT1yBg&<whQ&P`>d)}>htJbxs5%?pQdCw6ncGRe`H72F889iY8u+_F1W!V>yT?1Pm
z*{ntcgoqXGR4|&Uu!@9%?vNKZ2_6b?CEx8!nOmO9Pdl%xHYe1wd~4mK^{@z3?gm*j
zt$*~XTK+$&HU2M7=2BvpX_rl6iy{9#Q@bd6viVfY@VH%9zK!(u>z)hW&>ZQdJklLp
z;f(eL&C83{8Rb3p{!&1qszvFk!_G__6f690cH42!<Fd`usRZw56g&f{#}=F~CT<xq
zlF-}fU{<vUdyp!x5MHx#508Z5ya%U3b(~(z!qCA%yKyQA;0>Na%4f5OG1@EK<_zL2
zQ+HBZ?vihPbF*GytcIC0S5$PFf={b|bXquEHE#`zNnNVe3+u_;)P&6^k14evA>ub|
z^oPbjlq9~LhTX=RgTO;;gV0lFq9I*2;8zVx<_4J;&dpbZgnLLzo$Bf7Daa#fY*f6A
z(jRE(d%iGvxPq~>CFQ9df7))#_)HHyb7U?uhJ65dV{Gme47)ixG16+o#C65X%+TLH
zI{K{&ZSyRoJ~(na5FVn;T8ZSW5Lq)F3H7Yp9e#1w75Fq)+tzhjP0RpKV&InpkGs1h
z{iXJ7^3OanedDJsZBU*2o}DLmSivv*ZNPc;CrlOnCp|N4N@yx&x(+RV-(zN7!bQ)!
z0oQ`7ww`9<jny2>71w+8NOCqQGmRKf`1839y={A~G&-&L9jE0fL3ZNIo<?Ozyq{Bu
zlW&gxW@!C1sn)L}PnI)Vg0-O!<vgG3C^R4hKrW2bv6p_NxHp7MQ3Bj*eO`5yr%PDd
zA#M6UN?;f6<Av{UZBX(I>bzMnB3jy<I28CPK;nFBOyRhf5jk_!2iJ9P8-rAZf#K4M
z-%^mNoxw?y`szMim%|#U>*jgZ=dcEoL5O<zHixkp>$}m~QdY75JgC_;Hs6I6xY<|m
zbI1zvq3g~hNTb`{#%E+=QJdGwG@aC(yt~7qpvDb*d%A43Zj!vJfkq~iFq*CVpdct{
zJaeys6Fli~jxoI16*N5?XK8xPMQ*jfR(zCmH_=@C%8Nh#FK(Xyl%kn@KYfO2Z@9M4
ze)_QwGOIery_Lr`?7kpt`WoVGeC&meLDrp`hLigDPR!SLK?d**y#XT;j^(HA+fv&a
ziY&L33{0hC;#x24%Qri8X1y`9Dw_yHmK2NcmXDc=*V^9=#M49Ty~u48@3LHmITPaB
zRni5EF5l|2@q@_Od>TqFz($$2gv*|N@oPYcX{!#4YqA@b0U3)~X43}}|6m)rIaBF4
z5~jm1%zT=Ka~@I7Ud<5!g-oi3g=Z0CgIITS+3;CefmTpvueYXhxdt{<wate7sU|Gs
zc7?Kv`>(^4tKxs^7^AKYkNaTp*l)Yf&6th-xPx4v-P*`|wI)}%5vV32A{_Iefk2Ib
z&`K69qA-?r@tt2n|5!_zLani(vFuVUH;lmbp(4fZS3((itVX%6_kkfYgf2ojF}TmY
zQDc7WhHyY1@u1%qDsfsbVft-CTVARx+;!KiA;`Vu&;RnEA|<lIT?G}<f>XJhF7!K}
zCq|f~F6D@DSUAEFOy4qKobye?pGpval~!_cqYPbgIjzSm-A1n_SH`VoYNR8}6Dnc0
zv(1_9C_EbQuqWPJ2A>}p<lB{|VVk`!?22u}E9tx(>`O0Sl|=$x1JX|~`O<FPaz}}k
z+t`Zx9DTg+B`={sdu_g-{AyL)CS{G|Gk=Wh;l(WTi+~^V19R0vz}DgJZ^;i;ijwh@
zCB^c#1ItE`CPGB=eu0r?0d;Ou-n&T8+@qB|t^Y>w{k}-;E@*RC_5e_Hf42J7#QZNb
z+Y1rp@Oulgc}l?u7CSwHD?V+wHekLA!S6yO&cra)6dh)lV+uw(K2Me#tjHkvv7Y*V
zwLwNO!yZ^_oLaGGpvek<C+^LIQCMP_!UB9!qARY@BfD%lMSS_6QsMviCL-TEVBu<~
zH(=rB^H?}AEk17HY9SYt1eMSI`muRLH%RZ|0bux=ZSS>fy}kTrsg=F7DN*@Szm4yx
ziQR>rUmXBQcN>RcaH;22e38dmx~T9D=de&#c_P&bPfB|_MhV{WO*Yvw-G|6p&&Y<J
zP_@%6NE||3{)k>5UB`A0v>iGCoEe=IiA@a3n!`awUNisv<A<fadGmmvyQEXT41&Gn
zn_m7Fw2PrfK&JYg8|4IsAhW5q>G7H;y?!#sZbfg&(F#|#N#iOas3sXZHKb8p7gBt5
z7Gw(^8-YoXk)r!0WbDmnc%(u6>=;;RrtlXzX2h$g$xj^e2$>0too>nJQ)$Tr8?(8L
z7g*;4G+b8X4k6s7x5rwC7WAjl2brsDQqA`*#fG|PHSbwRg-%&lq6I?^b|&OXxx%nQ
zJHgYEFuvwM7i)$)He$1MKVVOXGv1n9Ep~oeP9aW?s8obTbs6Amsnf%I8^n6XHg%ua
z%h{cjP77M*2*jy3gfIEw13*X~L@O0rvowWY^R0~Wle3A)E)=`)XI^vRL(TpNE|Qyv
zlUB`j?dD{i<m+Z&G{*u+i`|5)13Ztu{-Su^+Ln$jAW+~rO}A(CO^n^|&Y6VJ?A`qG
zen8$P87+Cd*m*J*W4Ih)Ze=_eS%6XVA1W!24?m;a_|VwSA#Q%`YSzs9;(D$m?s|0@
z@n?F%D8lvK0pJjIA<SGlg~b;L@WvYc0nq2!c{45BRR|Qu{6!u_Kln$g=tffZs;I8$
z9QFN?rGF*(_uv1Bfv@DuLZXq7a;E%lCd9pmmM~lO0BiNAVIPl;`1o#w*5Jk*oCqP*
zJLW<vU~8jCo>ES!m$Z22N!Lumxfb}#yyZQo@THv%Px;EiZ-CDErJ3O-)!&uueYccB
zP0_+)T9i@ZpCXoqTI^-;%%x~EPud_wBDqS}Gwpc^XNEF8w9_@ROpmg<Xr$9;el<IL
zYQwjFMk$tO)5r4X?=JRfTnLEPVUT-hC^43XJ~vMnhWpKy=PzYB(!G~uKXJPnK?N$0
z>gYPG=?u<$D5p)o@_k}zjjshWRDRTWb2~L7V_||m7-kk|zy#fK=02cU<|}Z-6&}&>
z<d6OXT330i&xR8Arr`^#!Gy~ZdA_B<K9}-!W4OZvco;?viiWsB7i?veE5UL8mDbhq
z<{u5JH3q|wpd7JY&I{PCnr@=!i?^Pj_(c1vZnMx`UxAyldAi!qH5nPd{zshu?)I+?
zygUHlC25(4a{Uj=JUClU4)IfgnoZ*&=e#jG6wQSm<C`iK7SkCD$DJ|cTLXC0X6u1%
zs--(`HshY+WtZf;kIaP~f<3G~>iT4i%}9qRY0@-IPQ;Syn&)tM;NddZyMzVOX_jFa
z_THmf^Al;ZKQr>|VCUlHMN#9+Vde@ym$Xbz8?g~S&kg|n+h!PFLh8WM&=4K(8z)i_
zkt*j$^&APX`te`k0suE8pTwKr4Xy7yRe>r{8bkz%1*eEzA1l^(xZd)x)giY`E`Yc4
zD)&lSOLnDmkEgLkrbGFYP0fB>y;Fr{!A%8Z*H)Pk$%&8Pg_rl8r#8!4{jy}Hem!Oc
zq!s#YwePg{6z<xXEYZ-Y5*2I1V4DacYkR>t8W89G{E#sA-b_`nHZoA_z0t$XGTJji
znZmr4FGBuFv7APtZ>7U##Dey;m$JT`zs|dNf5rT}_U{?cmE@pjZCdR(CM##dSe1TR
zr{0SNP2hgP=nfRayK?u%QdIlP)MHEYl?z?UJ#cC5UndWNdCJ^!$4J(g^e4Z{dy8==
zn9qZT%a<`%%gKH{NMxcnb!=N>|08@fPCBAm!~G$ym)~LgF!dZ3etw<j88x2zI;Kp&
zY@4QFqMr)7Q7!%G?Bi>()u6B(T<ne?Ehop)#7*PVU!j8F@VrFR#OP-0Aw(4x-rq=Q
zC^8#l{~)~L?=y-WZ?yY|!fvC*1~9JeFcq|zjXX$2!rhfS(-UuC;`3#4?9-K2>+elY
ztGCTwb>m8uuyrBMUS1z{#aQO4qoYH=jjHP<urGa;3D3D=V)CxSm%Es}upb+$@;reQ
zTSuPe$SV$FGI++PlJqpDUu`Ho#`zT@hJxI-%Zk<8oHSHR+j8~h;;AQ~$FH&;*`x><
z>_GwJ;o=aM?SUUNmG{I3+fvyc(%IuPFNrxj)M@>f8oIn3?p*qcLU5#<O5&`tmtS;B
z|9DPnyB{1P5vMI*!QC(Xb(W{R@!emU_<yT~M%V0(E*t<HvG?zW!Rt}dZ$mt8D%pma
zrblWy4e5@+RJa?|Ck&lP6XV@Mvy$A%R_ku2rFHP_3NxiL`^b{BiPL+3b~uV^^{qNO
z&?)_Mx#um6n@V*?(`Ze^{qBhFuah-W^($w1i|ZXdi)BR_<%++oCvk1m?|pr<`fG0C
z$*^O$i%4r$nw%`%1sC&`b76gIOnmYI;4UVrE}8<f(tb0$PWZ+p2@{mdEQy5Ql=4MK
z1yj{O{1p7>|F=Z_-{hu~wEAR@ezgVozLkX04_s-!GYm<5KEsWwxHCs;jl$%_HKN6Y
z`+csc{+{DD$1T2oES-ol{nUZ|e5|O6uX?T{{^J3_PTQ+81_dp%oq{{)?)bt*b%m8G
zCugU(wtRdwpr8ZK<g(~XEz4L?d=AFR-;dw+vw5xjW0O}c*7yGgw144LN2oKmgI&yH
z59VUsnsOLTA-6H+<?~xIPH#S-buB?=0~TP%c3rE!``eM!^WEBs4wDJ$dT%+Z>`nED
zSXH^Yoy_}KXcv?feio|8PyHxnkn=3{RU}9LhIeWa82NE0rqtJOQ}-=7XV-76o9(hs
ze(0>^{L|ln_(yejaj3*1)0(&FRBV}Syhz77N5c9`tn+IzCDxfLNp&~vZz?5mX?i|r
zHy`1e{nwA7BBPGmTJ8~SICC?>G2=OYHQLr0kA(}@=SscQT%YIJc3z9*h)}XCZF@Vv
zOqyuVzRa3>_eDkdv?=)nMN`mB{1jtoVcj*UP`96@F)0o7*MI171W%j|NFv@=&^Ou4
zinf#94VUp+!B$;d+d@kL=?qnbqqt&<IqdUL+RdT`tzVCIu2zH&O$>e}F321mf~6J2
z54w*;wc@5J(}UFbXOwFEsKMGC8Kir(yX<hDlLoL}S{u?o5MU}<2^6d9vvo7RKNqgO
z{d`~wepetUUeG5{k#?uUux$n#6ry`(p|)Pm^T)CwhIqF|cSZw_O}ib6(r-z3ICb|?
z4l$sr$7A!xwxF?MVdr{lnX-9g?0uR&nw$A}Kj<-Dah;p;U6j`F_FFG<DySS&yFNC8
z5YKNxUjOuvUm2kx-<*V+2c_o_hWyye_?`WHMrS0pg`WuOLET&)MR<r{;Zv27oAYx-
zoh7?A!q7k!4$9f_GSrFOd+_<}<nl#+UH~yP@WGhXSq|^yi`17LKTO-&EtiNGbzj+X
zS|tid8!;fTXG}DN6`%hx*BmC_zK1>loRo3C;59iNP4=O9%7q!XE)XA?JIJ?H-=%-j
zpP|iww2qj?rk+%QGiW~$Gz_}xyY6@HpBtLyc+~DXHOKTE06HRrhKM-nX}ra>|4_8(
z&n;zo%-3M|wbWxhNJY@B4X{jZbfVjHjTT~d01&({)*7i;3s0AE#N-4UX^BpUc&$$C
z2ZQ5cv<vH49rFVBTb`dDs<GqC?+}|J#|=}Nkoc+TTKPB1P5NyNK0-721E~}w8siw{
zxO~2^EjXs_jodz>ESY;AvaqndO<mE{lcn9NFLHS==QvoA+*GE;$bGzT6EocU-t(!P
z7ZIa?9K@7mk(cArOzl(lNywQ6u6CH_sT^t=Cbk}oxYx3ulo@T}`E?8z(it%GY5V;C
z`>=;P(`{=4L<zCtM&Xcd%E$dK<$UR7M-IWhB*nCL!eb1_@Akiq46BkKtBuBIa2?Lu
z04G@>1Ytlch@JtJziX`**2`lwp<%Ckz}2OXKUxjS*9ym+{w`2>{{g_6?-6)OD<)RH
z`9U-4XhcFM^+7CTs*c-W5aP{QE?Mme8jj(LrVTI5E4pjSPg`A89FPYx4DhkFk|WcE
z+7%}@pBWlfQ6A+ClBx_~5ufFx>Q2|2?|E=!7hHl$aJWzAG5wp#r7R}=c>`A~D5z}x
z&iPwqt6D%4<&mo*wX@&4Mp{BzFcM_d`1JSUs<Zwd%Pq0CK~cmt5}RDa%`<$>%nup2
z5=~Txrcg@~<qe0BhHb=Sn2ru_<}|M^Ao6)f1*x{Y8Hq(y*w8nyY&L1@wQw`Z$R6q7
z=R1@aL)O5Szq5MAjk^H3(^Z>>+to@xqvMs@tz&CzyA}Ck=6}@{{bSeke|#^}&0;b8
znA7X)(@(G^*poy<7Z-i+A|wK7aq|tMp`qFt2D?>cnNfec!ueG{#r~!*CbCFf;r!81
zrjgtbQ4!JnbYf`o6t;UNmCuH}^TA`egoZa^`dgYeT5;tN*%-gKS%&ATRbit_s&iOr
za~opX-2vZ<wGj=A`CFZ%KTwvYtVo6TM~H}cU1DIgM0nunruvgU&mTuNQLR@&HS@v5
zOa*7IN)FX0P;=eQ_Q`}>X+=nMbchFwbfY^8ubiv>4EVKiM|Di*{L1TKXN7E&iJr0f
zQ|tEM+_#{?KZKncm*qu!Mk;}a^Se%pk>zVXX_b~J<^aP%e(Ar?-ETM8Ru3I&8E|?A
z1ARKYta@+v`{1T|2SJYVvFj`9e0KRmO^eR4;U9(e6SZ|-8oo_UwI%T-`ls${Whs$w
zHB#K+uJ1uPh^pkL_QE=vekkNrjQMxO8~?}r|Nj;6bXALWC^*;enPKP^eWhGC#idof
zIjm3Ne6D*erIod5>#cBp?>6*wEO=}HlmUFOp2xQ1E^n=2dtESm{|)&Um7Sw8DxY)R
z5T0@>ehKU*8Rx;y7%$L}i5rR`kmAZIbZV0pkKz`?F=J0r<dm)+pXx$T!|(*PHkDRL
z-fY3|?4kBSsYCtO_Z<I*gu||KFBS)2vO0_Q1&$emaQ{8OPnBo7j7yU0+u7pX7f1!;
z{x<R$algi5=ct-g+NyFMZ64OaUe%I0090Y>xtT2<`E75QMZrc`v6afH@#Pi*O4}{<
zRewXZ^#FNbOmr%3r3kxaUPjkPMHIxgb=<=_^v4YIm;J$heBY0{p!TJm{x^U88%Y14
z{=tEt$ZzANcF<&9+=r{?=RBfFwP^OSw>!S}eHf3vf-aF5Xy&Zh#=6I7(Eiayw{klq
za1Lz)eSfRko_(TnPF<Dh0k(~MD*rRm*<9HG_C6?EZD7yjZEb*AeB6Af?wAdbGrQ?G
z$aW7!JTmm24YKvnVI1wQma2$?NM3xu7yvnYnm*7PRl;b}P&pdyphWe+ESZ^?T3dp<
zim5tCzig7;gHba(-wjG_nubJ2b=nI!Jtby>gII}+3N{XiySNTqdC|$h`#aT5oz-n_
zqetvuK#cPrit3z-Lk&Dp&5tiCfw%Hftke)nlP%ci<SA@vnUB?v^53<d2A<?c(+5V_
zK+j`8u7EeN^;`Rlcu({aVT2>Et0EbY4YjG7%Gq*~lM{hBsejrW2!}^&H3O&SkvvmE
zxOz^|>guFJQcJH>{LFIt(Ilr{e)6-#-HEYSF@w=oa92y;Wv}XlYd$4EGk+5DzP_YY
zUF$-~G4q*YnqEw8DvG}M3wfMgPpWIdmutY`%{I-E<qSe25v!Pa-=)~mdhpFn^>sVN
z{&vkX$uTj#QappTA4ljTUFk==34Q93YXLv@xRnMNugxYxdlHqD2Pr8$yq0joLRtA9
z&o+x3HA{|$Qf#g-M1kQ^9`!5&wnE!~l<+JYC+veGE4QEg9Cc>-^k!b=RztE%oupgr
z#|I>%F*lm07sY8uchT9=@*BI*>eVdfo4m(37!~N>;PD;ymqwu#7Af=T2o_@TiVKBK
z8!q*2#1A22n9=>?V2>AD&B9d@a<~83=BeKKLW3l;th)@3$2Y;?+LZiwNGU-hqV;Yv
zMjG9-T<Li>lU?Y|-s&o+jcGxrp<_4mw-we2c!h<Px6N@?41#%md>9hW$Ewy>d+Hhn
zF2)C$Y)`~$<K|*XNP_9puVAl7=WB+1o|+>2L-b*G&H+c4D;XH$(?^|e+;r4aa0WMw
z<4H!d58Px3m$9Th_hV7y&++9RkHUH<Ob1(=YukghC8=bjqo3d7`3+b^Ww5)5nsql(
zp$vwHs?@>oW$4M!*-+e;YPv_-YUO5?2r+1AmJb|7{Fsw4-v!DcmZ71vcVmp4E%sch
zm_B!24ovRtNLMz!K40Hs`})^^Xlwt3+auqRszKl69{^lTo|kNNB<N-hn>(y{>#ie5
zdao~SMxF^5+J^*$wvIeQlSp90u%Sjz)AX7_Wz3C1{5x27)%6aKo5wvrWJ~=l1hbXf
zO|6ANl-n8>rN#->1pmkukDkhahSN{zDyG!|My88WW}3(|8MPG_Thf_b-Uvi>?bc>y
zx%<e(hms(RSDCOLy~FoYkI=q0HnV9li!iNgybt*>I}(62qUBd=co$hR897Fq5)D}5
zMFbqNbNwkOh*<G_R>oZC235vGI*<%~;{b$%)2}K0@K508|27GkaLR6DSBq*nbJhLr
zyC#?O3k`fr$!LSEx@mosUM5h?L8&rpB(d3Ad==jAR)Vh~=3<v894<Q#s@=(TQ*a8^
z%o-D{YCo!eh^eK8^@aFt)5IAO*cH#H!L93l8u*O|EK2H8khU>F=lYd<)p>Ii>uaMX
zGO4QFoC}PJxp1D*vTa<`@WgNzHbI4)UFoS&Awg=1MX0g13QOR++Q+|H^2D`$T)Esp
zxfkK$Hd{rRUn;Gzerj_pwTx^Q8sd?fpWdG`)OafomV3i_Cm}qWY5E?@9(c@<V1&M=
zTELSMcc7b2yWfSM2Ioz1GZ$OYL1UsevxBn+$Pt0LdN{|QX%s7>GW(2J&&581*Ta@(
zaNWYKSEi1osRj-8{lgW5kt>`FyYV0l#;fs4zn-3k{w{RYG-&Jj&fcZr4W0gvtU(<4
zd3NZWmcLeGo*u(_{7#4asjloDPT}}npQX)x%aNtIB<KYXwn}IX+hopdF6|1B8dN%j
ze)ZvT|3HaW{)@c5t4k>xx(P{c&GC^uN#fE>!fvmL8{fXVa~>D5CEz{a<P%k`dpiYC
zN#mQZY`d_LocnlGR;ORU@Rm#RbcBbYY+FbGesprAY>|^T!sD^xJewU+`_C%MAjfcG
z)Af0n`y-obB9)Fw%-3V=U9I+@^!K7%w5xL;XFqt;?cVvgkkAvQ@q<*Uk|dwxwV7CI
z>LcpOPQMLuG@6(yubaLav#G&*{yc|F-n0k(p|^8CUsn|5`Yii-f)fb4U*cP+H(v`U
zuRaJevr6TM+-DIF05WSq7!Xt4Rgj<$-;8BwS!OWHE9+1f8LL4at}&5n&SZ|S`TjTL
zO{n!|hR`lJzrJHW%wV6o`Y0wpycBdZ1m@imk+efP01Ou9r1%LgJNXfp*n&2fwj=Bv
zHt5leJhrtIinHCkJXh(EQ)(~&XmgCO4C@|HNnJ*s9pY5umM4MW^-=rpKN=m*;OMlw
zdh~>pFEs^rc(|d8^$<Ec%rRw&zE^8po*m?NTRPj)TE{}p1~}SEsgC!6o`r-~bu5B@
zn8+c%%Xg=V_XTCMP@R?tO_!&H{U|yrA_>)vZ#$ni;uv%Q_<D~49ujO8o}hT`T;CfV
zyO1-_iw<TlbzkV+FAT%a@UF$?__JyFGegNMivyy=80p#A!!6Yp9xetn&GCp=K5u@E
z&%jsJIn4w`5mx=|?~|F1gf19F!%d0jgu#7G183gLDlha-t=J2fKup?3lU;c7M?s+#
zUBS8AQ4Sz9$OM+?+#H?Xb_l}}<jsxGH#FG_5bmLgcCe5Jt!GulX8-B=2<-khn;fy)
z>ZtXm(7iB<5r@6vXx{48u(7FZUaPyoH8}LioW@(lMOCu~s!?q(*IV1{&$XIVmE{(N
zrnlc|csy_mKcc@V+^nrx+~WD5^%coWu9ZF;gS&4mm^yaqAG_yYGt+jm7UZ?Jj}2C(
z51WK5!P+JkdOv%2l(*T2h^8BgQ)aeX{4}cF3Imond3jO$t9sk5pK;e$rc5P|4f9&J
z-mW*HFvmHo1pz|@TV2qTuqzHkb399-7fMT%-E}BLK94ukm>K=Ad%QyhhB#2CE+_ms
zQ7PbD0IR->>zv#@UY+z9H=}f7@iwDmW_KB^yB+BOr1$)p6)fp_NAE);sr>woPwowa
zP=SsB8y(&re`f}X(M4>ScZ@~_j*or}$izmfF6D>nxX^MLB9g-!p*4G#R3uJzMrPyw
z9ADMq&Jm7QrKVfQQLS0>l$mF#;yL-E#kn7MSYsEAtC5xaq4Ttv<3`?R{C@g3_=kGT
zPIQfEBIo7grYf&r;H{&|DNS1PU>oFd!N_HVz(Dti^fT(56wHZG76YXEb9Z2nk0p*I
zw~`X;$Rf+A<S~5Iv9i6%mX%34phwZ5K*Wa@(AI#D@K8hkk`xP)p47mz6SCQ?ZKVIq
zD?9Ah8+;E#746<2cjs?&D3oB6DE-ZSZBxkkHEP@K>t0+~4o6hDv?nU2ja|+xC^hLD
zkiTb9P?G!vf6CMp{7t*H0iA#=DMBKxZ=m}7i>>oBX4B;L>Uz{c_e%Cr+k|@7$S?&p
zCM^~0N6S+b)cH_zx#^lVV;jsZJQlX6t9ykqq3Zsy%EVdC+8I{@T@c4Y$H1Mj&xEpn
zEMq$6-<bJ#%E|7CzKGdT9%knLVws~3+Bo*tn-*0nX~xc}?Xxu~4Pck?s|E&s(J5MB
z%TT>lW~LAW95NY>x$tTo!W7O=YnsXMZ7;rr0Hhm*O?}3DL{A1LYuS}41DVwW>!@Xq
ztF12L)5F`E_MCO@{ZQ%Aw#BGvykYbauaPU>gVSYa8A@_7uDk?N?3gUP(x=n|ITqj^
zuOv5~9q}9&K{zd5s7Iw-i(Ns8CEBB$Kc+(`Q;%E-5gd&U=<^{U%Jk7A24?#ft6Bxa
zx+BuPt#JLUcP<##0YC-)epb!h(_2R}OI)QoG%`F0OzMKMgwu54lH<o=h2)ut73tK0
z>4@Yr=4y#Ep*0IZyGYrJM>bcl+CXH3;EM||Hk4eA#m39RQ_r3@;>?g!YOS5|mDjH3
zZ*v1!J+<|Dn%&(fbWfpq0J<kn)24THTgSU)bc?jCD_pQdz;bd5wMBkdHTv~8&Seu8
zx8RYCGBfN`miZy3j!7UR5tFH&T$y><9UdL_jpVr%O&((G&;?xOCJ=e?LU21NdHkaD
zXbUdbwW2c2tB<Sz4*<Gvgbz%*n}lsy@#L!)x%uW3$QXAzHYhk>d3UDO%&hG&<+?^;
z(wzRwfpzYuLSOfgbe9l`K6FfS0CVf^QVZ?QY;AChGRw{`KSVUGRu>UdY3_x<5+C(C
znV;hC^iP7vx^11w=nTqTNi!Xk13$boF42FEH*z0lHc~P?+|%O@ii|$F)nKwE`h)~5
zbZ#m?8EPOarV*~b?m9bLrm=EqF;ZJf5DjTu($tCS*mc=e3XglzH)*G79~?4q3Nt*6
z$(g8`%8%MjF}uap=v?Q+4XUMvaWzy#Z8_Zj%(ym!X{sEcKTIw+C7JQW5?03p=HAXY
zWDS;pR5>nfc~g0@0ReC@%05P&$UR4ynM~lGucs68!zd~@@*95oclG<f?LgTYU7V9C
z70jr-1IMgbk|yF$Jcv8Op<j3(c>uW41@Y5AzrHtd^4!#%dcFIsvBlyh{zkk0Wy_W{
zJ2O~+vErXxe-UE1AHHN3oZ_U=)p}7zBlJbJfuzGE|5(PAh@5#Yp*UkUIXpcSSf+q*
z5aezd<U5stnKXH=JEpd%8t#VGJFTd{FUH~ZYL8_61%{!1Ib>|Czubcq#ZxS?cyiLw
z&JpWB<Zq{*vEY>ERwmrzD@hZsvVU1!_d40mPZT)E$^q3|`Z!z1#y!K~c{RM1psffw
z#Xh%bqd3p^5C?q|db5_%T`1QTzL+up*eBWC;MaBa4E8_;G`*o?oMQM&l`*;sm+vCh
zN^V2U^=5(6<+&P5{$pcsciC6dngoaR87We)j#=hAf&e|C?BV8{H97ZiXUFbB{Q%nk
zJli_N!TQHXeNL;#Lbkg}@D$%=qE90@il<y*mLK2eo<|&`NR=ir6C`q$b9d<UBus3s
z!R8<g@}cYm>|>dIeq9fMo!_`Lo5gwGlhb39k;<Uhyl6r1gFuC}$di+z@`Zh%ZCW+=
zQN5Yd<k^+AJJC_Q+GG1vj$g-}G#q+V-Xu7%I<;KN_woJ;Tk_cmHi$?c=$5;K)j|Yn
z1S%$ZV=Z5$+o@ZG?Bp#6@M}W37nADdmg`faBa246GfB)-*?PT}jSZBPG@b5O4Y4<8
zOo0ws>$v^wcK6oh<jJDrcgj~Q0~Jnczm!y1o7sG|YZ5&aS<RVOnp-u2)7;+EhuL;w
zxgipgG9k4r%~EuM+8KpBEH(&?Uk3Bu{7_$4H>uqcl>dGfPJuq|@bR~Fn?H5AN3Y&S
zy`OrbB=_rFw~{0U_5N<T2f3~ttRw{yjA%UdJZf$GPkVn~NB(E-MC4Xh=<RgY&JK(D
z{{4F<eqwHJ!ar=25JuQwVS_?Kg24U{UqgN<v5&%6+`(8sET`yksp@InS7tln4Y~us
zJN!C7TA|XFPj8=Q_2d}@5EGzXiL#d%lys%#FXuF<SXcsG+4mwB?wpBV_3zKXV&Pyf
zJXTE(A#C;6;jYhfa<ZR2<AUBguZq&{3sM@=T0@>kX~TYIo&6qH9+cd}4Gd0>ZdIad
zs#5h&J%88|bAg`ao}Nj>h$|p-le<z<Ei2JP^c^DnBoA9YN1KXXYF^r-$;*%L|B<^c
ztAh2d0%A2Ek(t@iWf}<|rL-M@^-;cELwFs=7*|h0g8U=J#;2oD{%~zd4njt-S{)_h
z&6+VmTv=}(%7gawu$~j8s_*YWR<ftzh}{?FF^#0bf#(UM>BO74U8bgoMljLKmYkq>
z6OZQZHfIptf5d9wSv@b3ZDS?qq#Jk2lj-_`n;M~eP9hg-w$i3}Ego7@8V)9PqwTM_
zb~2MYhy|-M^It-DW1s^b1?{}E7ci8bM+rj-0jn&sr%za8ppAZguzUAvT}(aceIR0)
zigFww)rvpEQa9z`b%Bt${8>VyMn&Z2Bvd6oqC%Gy#Qs$2a+IJ2a_y>4l`|^TUE=C$
zo^n_DVhewGg+G}QsSt0l>$}$}U`=!j2p8falSXj`k5nt?5oWA3|BUk4T0;Y(F2rTS
zMu}Qtoj%HRe0+cWV}&*BRkG=5RI2l;jOL{K%H8lM%VtM;zRf5QuzNAMB@8KgAG=)9
zoM`ZEPwIq0S1!IPKyr82P4#}FRmjyBlV5fvIC;%r>4>sEV1e1TKG$iD>q<!&A`(y$
zcweD?OUX;<XZ&h>Jgz#MKw|JKI!}pTRY=6@x9SYknRv)ajHsia3uX4J=^HW{#P-?c
zV!e;aEK1Rc>)lbu5^J}OO1wYA%_Jh2pDR$)tNS!7i0=qw+1+h&(PScRA}2<AD(d;8
z{HQD_vxke<>AYrnR;xB3;2t%4a*PIZ%nn5=?Pzr#1&<w*p5R#8hUk+$;+)R8D1okd
zzUc3At+L)f06?uX@LmdirEd7T>utD#NxNCB!(kF<W#FfYhuAV%D~&1yxE0)E{1bI1
z@S=Q3bkh5o3U{}O54k#2MZaP&^19n)$MVlu`$(U5B=Fae0!Yqm)T+<4vF0(%v_Ty#
zI3J`{R+u`AGvL@>UqdV`4M?DjFDYw?Mp-qXv1}H)viH@%_~5L5&C2JcPve6luYBIt
zW6?P^zGP%Oe;u0V$l2hqlu@>}4s>Pt+v~DlURD389J)jE*Q#Vd;*UH?g{_@$tkN<<
z2E60Y_rei2L(b?Ilv{P!Dj%IY3dh^s!$iV5Zncl@_&)yd4!Pd0<DlVIT9h?eXF7y(
zS9u*H=s#-{sQqcx%z)^}&JJc?%9`J)^ea;>uW&H;ga7g=d>zRW+Hqj%EEIk4SxX>4
ziA^7!YKWygk{R7~9@;QOxbAcpUR=(X(s5Y5vf@>BIj)|$tHMe6UU&1^-xQ!mtNsxk
z(g$0`w_Tal7CsPb_SvG$iPaT&{}v(C5Zg^wDfxUp_m?Dv&n=IM&E^xS<>z}&hHktj
z(}q5y0|NG5w^n@mTzEFe#IlOkGfx7uHuz^yuB+ly57OHuf9^zjaGPh}Tir*`r^1Ar
z_W`I4t5i7zWb(6$2T^MR{(hFcGRSUf4A+9YP*uYHbM*CsXp1e;n+CPC&xEqV?u&mO
z0}oF94d8!JzdkZkJY6A|7G$S+Y6|!?^K5^}(F<4`_snE)s3t!;WyP-=8cm|d4+)aX
zjGdF2^4}Mm)luO=SJiNXnEVGuY}?iw=CssDpLyH{&q2+-?n#wbcU7V?sTH+043Y39
zenl$u*xj-fKNV5(fU@igIjXnY-A#|5K>zp-;oD#5Q5TgCyqcUM{fs8QgsM`r3x;Or
zaljIJj$`#Km9hF<uVLIvC>P=0ET+94#Lz?$v$7(kdn%t=uvk0$`z+ENZ+R|?XqXiq
ze#f^uBs4T6NRPLo31?uUh~jJqAb7|xx8!E2@YG()d)F&J(%;6rxSZ99L7-gM;OtLi
zTcKOm4RWG{y3xyQ&J;NXFFetQ6`tf%19x6?{v&TS04+2U@Yrf$ZjnmqB&@ChY1tqL
z|1anHgd^~*Bo^wYzi@gO(?lJycV7hQ!SXrnd9rpPkpo&q;0%?m%$l*Im!@?n*+T}X
zfXw?g+4QheMs|K^GR-mb21PFmZD`<TV<Rnqge<m|#Awk|p7-s9PZ~7^7sM4Xay`fN
zOShz(6REhy6cztLiivx*GzMypp~9oOUJe^*`wdcSv3aBHiAU*d8A6$Ur(eEwZq}An
zp>CqN>W&Y6ge>OTgT)l)WXbf}k`o;VDPWJLpbe+S7Wm}Zh6VRMvN&vnhAx&d*eaui
zDlr{WBSoYvMol#fGjiR8hRP6Y;yJCg#*-O{RtSCle84!GzV7W)WaHHK%+o^P($V(D
zY^zMR6ivI}BG|e+ul95})KtrDjFPqPVpuL_fQ=2+o~2LHt8ClSx9t!s<|Dyl;NZ!N
z1-gWwy@8(wSsTi##?ZOIMZ1?Hdf~P7wum7L-+@J~KpEyT=Hatld1IbA%PCWp32*+P
z?*8|05B<`HH{J|s&f7A?#{FD);w#c8VWO?oHDf#V9OeDfp-y}7;XmHD{QI2PG?>8U
z*8=rVdjqw=^jR~*3qj6_J{&1`rt_;I;U?+H{=uhOZuBdgYvt)9gmGSH;juq4uFx5g
zwgrPwjgxYV1yQ(1>Zwwo`;$1Pic~>-f>HaH7dy^Ysj<1HcKU9#KCo)$J-^Z4&i*N+
zVAJ+`<i)(So&1{S^yDBz(GP3FWA_$2(r8af82--5>EE!aFMN8tthVXdk*-#`R(1~=
zhI#+6+fd?sTW={;!`vam$VT!+`Iyzkj%W#A7D)1^?$PZ#c?kxmMx(kq3AoSxjlc^m
zwo=S5XDp*uhg;7r$Df7C>@hG`e+`Zo3s$~tD_m=56UFWjs#Wq~9?)Rt(AlCFZ<o%)
zDQZxIGs2yWd-=cmV5@4~qEDF${09}!!vCJJ|B1UpTvPPin;>Rc?P@f=>)9Y~!b&8r
zCO}{iyYtlW%+Kx9%Zqotbt2US<qIDOI2^GIXq|c4P9)dUwk9{S1Z`e@<ZTGBFlWYO
z%VWkcvMMBxNuP<psH89`__lY*6`v|ugxy`#9fECRy$=Zvrh^=WYb-kK<z@DLxZlJC
zEz!lvMZ#7fEw$<%Q8U%$hF|w4m8ya1wZfym{5X-1-_ayfr31fP!GzrV)JO8wZHXHk
zJC!gzHu4!|DGi@@TduGK=M=PF*>d5NBr5wYN{?A!{BMzpR_X+q(n$q2r<H4KSKoU6
z3W99Nr%vyE&(pF=UBc8Y&6LuQQ`WZ{;vC@U7y2KkF2(E!VxGQ$RVxg?)H<&dz0qJg
zGg#Vv0MIx=y2Xo*Ca`E74<bYE^K!vKOz1?sU(P0tJ5k-v-(W%eWK}l>6*HRS5XFJB
zq*y6s;nyI4lW#~<iPL22zRXxJS|SwLC|xyX0L~;|T<g1fQ4hT_kG*g^kgyMIYDn4$
zjTvQ*-n{1wQ+g^5BNkan*FtEHlOow~J5m$n9|iu217a-KV{-RA2UpbW`AxJaIz7>w
z1e0jhl&qWT5~uV$kP04+c>ycfSJ1T0bN97oXs%>k3>MDKb+bQ5zp%F3;%g+mK&+@J
zvDC!O)eu-|NH09lxh<&^h0X1R4p&lxd^%pzJWY>uP83fAL3s@h0d?aMf>3f#bke&q
zs|gR+yH2UilppJd7g))C(q8Qz7NsEF_!hKHp7eXCfn2fv+)Qw8XH@1wb?`=NQ%KC=
z*0mA7dC#v)eP>ASZH14ZXWC3%Ud#}@6ePB!+MdGdi%CBt5}`-f_p@=}OpJ`Yad()x
zmP_`u0SsT<=n!j`xdTS0mf9LGfC}QFXIcgf)cANUq0J~qmwH*8{a-bP!%s|Q-{z+J
z8Y!S7PDQ8FT<2osfn}Z(HWoICN!Por`y;%<q)+NAqw}}z{Inh9YrLRlAJ7;FDY)bP
zZ5#2(g}B#dj1u#g4<~Q!@LGZ=OlyL6nFLizK*UziBJ26YjDePGo@JR@F#=;4COnrq
zI7)Th>3KKbfbe+RzK&9!jVo^invG`Ru7QWKY3b+d`|GF8eOq10g_1cODQ$Y()ysiV
zQ$TK&43*+8PH7mP-b&FN3vR;1GF@-;A-N4ZlWS(JI#@SVTF4U`RJf;qC7ikFrnNRA
zeQNQCt?+N&Hh{|RZdL6Re~hYreb>p~K=`edz^RAZ7(+#BY}jRl)8?LRQ$Ib)bT@1$
zBxv_(TPvcPR8d<UnK+F$Xsl8Q|7ouAtn@GB^{t5I%BIE&-sUI_Bz&{cQp05Utel$3
zwN~|vmKcq1ruF^TfBiSAVJe@(`!7srEy$J05~a*R*^#W0Lap^s={MRXzqH@X?*33#
z43pHMoq+#PcxT9$fX%3F$YKsJH&RQJl#{i>=PIm553!elp}eh}6v@D5pu#|t-vZE{
zkCdcO5>n_kVQ1sc(S38&dP79fxq0X3M`}BQb}b!5<HB=zpSOO%lnv^)a#LXMxlZ;@
zFA}YivZ0zr$r`5D$k_!5WWa1mG1VX>fIKy&{o1VQudx3O|Gy}os(Lq1U0<v|v|f16
zSh-~be^z}+i(n)07gu#8Q9UjEon6oPNdn{SM9>D<pShA%UKhbSA5t7t;U@<uHViZ3
zvHn!MW|2y{;!qJg6N1TD-tul=n`CvC5)yU6gey5b8+o^ZynO%&o;DhnMwKr0t0?*}
z`fi8f_iD|x?#=5pO`f8J<Uc9ZkC=eWle|7kQsYXlz-PD{+s($ysh%Ruov7YSuC|;H
zTFWQ;$h(;Gd3&3sW8cnRjB5(^rA8LEmMQZ5YCBwayzAe>jUwT8Kx-on$Pq*eqXx~T
z0>Q4m2wdiyJJ<`Ee!^v^YP%>}@jiy<VAqsFoM_N2@47dG$NKj-OOMGR@?Ojl{ba5q
z-v#fcATSPbN?teulj&9hb3mx2$i3}DhS~f5!rE37Y;K=P^liD;J0NbJ(NUd9%SkdB
zhNm9=HYX=3g1wScsB0$Epe6IKR3FlpJG%eJ&?c)fcV}V~%OCte<j1N9<1v(s6?nJ1
zZ*4(EB=C$#T4@@rq~t1B7u|+UqPbJ`a0`o4yS(v@sw&R#i0ZGaxKmSgnARO#;`=od
zed;2pw1L>Ha+{zYWQron+v-z>n{ePui1hU;3<20`dhXij?xZaF9+q=&bQHfbSxjO^
z#ru(~iMg=$5Oi1ZyX$4$K-2&%S<?baOW2##n3sb`axOG-nO<~qYIo`x53w)bO#bJ}
zQH%Y-`H{XA<Tw+*e-fp5a_cAIhl6)oJCKXC&OE;MkOQy!$3GRF|GQlIXYb5u0;kUE
zGaDbF4SFpTyYpcc!p{QCAZhL>0~HTCUT|ULbgA{qt2&x<#4SP*s1Z*n@mSxS_ID0*
z-jWRc{NeTa9sYR4yLDP3(p!@|!01_+fK2aJLB$y!vK1bB<-_+Dopb3#MWh5y?KZA$
zcA4Ivd@8@>hC>MKjTZ|)_#j-;VAD;eFXG~Ne}N`8^2v0@<sDNj<}smj*=0Mqx-|c7
zjn}gAYO;6k^iGT~*7ZZTBP#`1gVx>E<h~hh&+0K6yQx_G&V-c3f~%<OpiNSc;4=|8
z#N-dNR9&Z85A%^Lf|NuOJR8w$Kc+7jHCC4p+99|KPy`X<XYU|XpmcATlXNH|(l5k1
zg5vxHrV`Is|0(8w-e};fC;LFAy42a8V9oHdU9q|-ljSbKAJn%G-)p~1)o;^XH~4CP
zUsMtKt-}ee27%T4kGW&V5Y{lx*jb8?XFzVLbzje|#=7CcrFp&)eWRi>XrHhX-z6Z~
zEw>>1*zUgjB2p$=F8?2GZywfEmaPv{Rkuqm2Z}-jM4%OcLO^7mlWqqbm`YIQK?#Ep
zW@HK(s!J4+Atfk77*Z4hA_Qa%bCfA!fXI|UfFJ}CAxud~fP{qaRQK(wdVaUNpXc7M
z`}~n?6I^HQefG&&d+m3<?_v~kBI0S522Grc9qR~OH1yGK=#O&eY;wH4c6Y1Y=i_lc
zWs9pyyKVTR0EYC(gjLDV>Q2ONN3-NG@_rI~t1GkkngBow6S02TO@>)b6HVGa@1MN+
z`&2rVwD|+anr3CWJNmq9Br#_w7W>TLgaZBU*h_mKuGzUdjIHf4r-`W1A)~WXe9X;c
zgXWyZc@t{&K-9yQEO<t9D<b8gs*>s<xMMH10zc4uXN)0sS+SV5^j0evP}sbxb}cP%
za>c5Dus3lb5xQGsH5JB4Sqe)-DOcfgpHXJpKsmzHhLOa)@;+_=dnVH^Vb|Mj+jo&X
z`fRbOQfOk#AP(XnV&~cuJ@*Qqy@gv`ffk%N5J`P)8IH>(uAb~4dDI_ju^ulMb7NZ`
z^Ojlt(m%>PG;B$+qBIur^mS`Vd~no_;7TeU=kOvM7MKjF3exK1T4`#&+nnjjAuExh
zET`c?n#6Fw=A)g*E?t6X)G&)FhkJdG^*8tH<i+@PT&Y_utGbtVxq{NwuUxHQ(o{bN
z>gop|RzZFDx~y8FM}Hw?;scc|%K(5Xm1CTjm)m~d>!-hu@IQL5(rfgc$9m>A-rKHy
z^!oPsR<$Ra$Mp+Yf9dxea$-d+4Bpbr$mH8RJOr`skI7gbue3FHGGbN5mkcSr+BVW{
zEbxSObRhD4<?@8CchM;|ymzkBE^B)^_5{aP-HtjhFU^P<K|j%_%PvQ-z!2G2d2;w^
zE_s&6;TXSxT<lQpkw`vcj2(?ZHJsNxGL6c3tFT+Bz%~r4F_p&#E<xRRp)?sofT(Mz
zkzf%fH;xOtn)Bu${MW9$zWP1t5rP$-$R*dEs=_%&QP#g8dn}6SPSlAfi|g)0xQVUT
z=(1C-&bN!3uo=FNpoo5r6cy}3FP8S`!sDQgUSwznsxfGx;lR(5Le>_sg4;w~$$dE!
z=j|I;9udxRUT!xmaLs1b!ee;G=Es~G$-R*$()`GFIg94r@H{TZtO~OO(_ZF_z{U>o
zS=9WGW_J7mhAy13m}rHZ$unBeRWZ0r%`YEbbZl$z&8~$F#pxP<xQl~D{@RQG(fBGc
z=_aDRv&8i9R)NcasLGb)Zcfg5Nu#D~w&lnt+Q+2FiUulezJB<X@v2}pS1r9YT_R}2
zJaRf-*-xk7>}C;9=}$fmh%EUVx#Qc+w-`G1K52=_Y$=aii7vutz4avK55N;!!nIu^
z<k11TLo4=cycnlC$bw;`4<QSd+3a|{6OzBU0MO<o)k8Uxx(ty*lS#7y1dlZl)i}zm
zRS#Jzc3rvio9Dr(kG0cP!}D9RPntUm)K6j6qu^94%>n#vXn)g^;B5&FveeD0Wfg;j
zO%ihHVyu_MV<k~Nt1Avzwy1wQq^{6^vI3p}(|#}_{dj%8hhsswztY*<(!PDG=SI>6
zD5nSw8p5}%!6??xs0)c=A4tYnk9z@x!QWy^G**9rU8m{N+8mV{>av*;9QwN2j~&yq
zNvzRwvz}re*wYu{sL>41$>W4lNepuG>ymEfZ@&MVS_blh3ea$T43f#O-E^M+qiY9X
zSbjpFgmOWioqL=43WZDhucHUw{Qh)b!s(N&Ck_`kz?D0092J<+g`FyI&s<B-#Ef=f
zN3kFGo1ld3$I;(7VccId_PW7~NJ?_g6SZ#=+DTDG6HOj+ukl;ZKOFqK$&-pnr@FMv
zl^c9!I?gAetd*+7vT{XBUDb>A@lhCNO$q3RZP(2O+u$l2U&PTQ?PtO>D>=6qQHiG<
zTz^zA4l&D!QHuDJ-hWfp{N3t*;QfzjL}Kq(cn4vnb6U%WwX;0gI;$jitSPA@c6{7`
z7NjygQhN1A{43*^j=Fsh_W&*pVNPAVx~=DX_w+S5^IQj?pOk<4$GL_AAn3vidwiy+
zT2|JCvCDz>=VNkvnjRSh^Ez{NH;_%#dgrqa))@)p+T8r5*R<Bx)3%vJ;u@m{<SAD^
zD4d{czT#IhwhSFWP;7b|hC)mLUT;6nSbLYZx8gemTVd0zo6}oy1#+D<3a5w>*eLhw
zczhOH8<^XBOq4ht=a3SOj9F+XjCj&cH~UJY=40sN`^nk&q9p(@@#)yaEFdg3QP9`W
zxUBw#4W4x#*<<wKk@IE2H!{COPWV2WMG}TMC>dp2^sW^uc&3LcwxLe+$vp6^>nBJq
zt>zFj8@Q!~<5W^VH@{r}<6c=AF|cEi&W!Or+p>dP-=3M6TrSLvhzkcR<IOx5@vONU
z5t6ErG!K+TY_v}+e_4gSm)^=&nXzbkGOhS%Jnz>&{Da1UC65g4C&ppy(K4S5nDa>O
z#m20OQjZMxz^pQW8!#sr*g!ZtcHJtQq!^TAXltZ)uwVL>&rU5+NEw=_Ps0<laxgx*
z0J{*(11O*b9$79Ed6>b=y}domiL}cUKV8queO3ChCsNU}va(xwc|tqKQHaFjWCA_G
zklFUOF|@OqeZI<NU#hx5cWXCn!|tu;cx}|1>0>t*^xG@1s$RU#%zgq?DFk4p5f|O2
zgYmhB(AGmbQ3fB+pBJd_e6hp&z=x(TbgJ-u^UCp6r(8OMW7p%Kk~|=M*^CLI$lf4y
zvGGd#qnxjX;$yjX0jvvtnT{S)+nP`N9x1>G%OIcjhPo-3pmDA?)R(y)a<_VLvz6<~
zE7wv@QOe=TKH&gSPUFW8=i_&p*uM*6KEDAs>@eDm=#9t^UV)A0wqd11OkGS1dq_>y
zK~)(OQF0aAS$8&(Liey3B_{HrMPt+6*>g4?2Hnj;y6gHUOyxA6%*hety!GZ$rM`Jf
zLB9FxdvxV=WwkPAKP~)VI0GGq-<9?s_Eu@{r*9F|-`V>`drp;<Z2-7Tj~QvlPVkcD
zhVDrk2NACAGEUHbZ*Rq*^cgZQ%x@>uc+YTT6)x-wuNF=rS}KE!G+I<2n2~VsD?e|d
zq}e$%yfSC~*dyuB(n=LWTq||59eqT)&pWoRJ8KpX`$N5%hEveTHY}~&&W^5@^bl%(
zmFCOhp9F;4mKAsTU$k@c6Oltxtw#z|)bS`|qI;vON>_35!2(z3Oa{L)#Wr)a53KEM
z>-%MwnE;Z!H=hu0S<8WWf#qtw!<3I0_dGa-2m@ISahLFq3-fvYO<We~-nKs}i9Y9P
zn|sUiU#xW$@9h-xCdmFG9?j==ftrb<bo_hKiLwf>ixoBWFr`fPr9X{@cwE;BEQ`3f
z?L*Zuh}9Yu=C5zm)SR<->4Z(Z3Y*@{U8aC-L8m&;EOV+(7EHoWibFAZom~c3@2rj&
z6Ng=maWe4y3(;Q1{bgY&TqJve=vHHOzbuNlqGufbcD7wrePcHdH^7>hz+;C*Imu#@
z6I3<%$|iK|Y%kJu0P0v^H@IbdX58(!yG{4XYGceqfBWS5H+<6Rrx-i1W>mRh_JtF-
zS1uZXjSGVdN*)tR&=m^Ak)-F8cxV-AaxD3DRM4yMGJ;E$EWY;pVl94)Ew8y;TTXuo
zg<iq8VMAyug*ot8Y`&Fg$cKS;Ty`K`ZT5lV^N*1q-=5%^@p$r2uCDE>p9P&tth&Q8
z?>eBq=eKCK2Y>XHofyO;GHbN^)V6nH1#3w=;)&Q&^7`F)=4XH^&A<8hzog)g88U0!
z$97~>5<LrBzv6}>Q+nW0?a?_b#94JrSq<(dXrgYFwMTmP(Ag&Kf%HDF{)c+E)LigC
z7uNCZrS7Ip)gb9fw<p6xR?+I-;~n(G_AG(q`XnOWIQ-?LcbIuWN6{$_P+5dl<$M$S
z9A#2R`9;=P2_%K+HaUH>6adjZ`^(#({(BsbjcnyMdZKQ$hf!-LKJoQH-XGVUCbpk6
zpD<ND?CM8Z-6(6Cy-%Cybb>{Ry7J?(;|5ux_eh>%L3_o9mQWmd^!=7S)ngH-X1^F_
zsA+3lu|ccpY+FeLkQm1eXq1V^4Z|Pn@87jLZKUP9u(1|i*?sTgz3VPtr=M0e`(ow$
zfAi(|f6aZC$4XZXUr)-OI|iB4i8l0!T7U2i*Hh@|;PC+$bUuSf-QL-k$_ZsA!%Htk
zPeDUHo6LrEnu0$}-t)M%i#wa#?g#${rk(a-E(M9E#u=9$2v~*XiHOmim8)dy2lZ@d
zsF-MFmzZ9tDu+)07$#%r>CZ3-tW6yXE=&|w0AZY{Lc7&g)ObJIQLg{8BL+dfs&NBO
zL!y;DWwGI4JDY@R2~#FHCoS$**T2BR8YYj{Rw<{>uXhJO*X}Xz^Bk7Hmr-;$71!m}
zJyN^bZc`iM;nr=y5OfBz^>B+N2fGd30>T551Jdsr=@P5x1(EbxVZf}|h!vcfAejd`
znMcl6j+Qvi4WOBvQ(fom5{im}3WKuuA%nd;lVg2`vNHOhIZvhlHfPxvgUOC4=wH7(
zg>j+J6+LPetU^5?o8`lu_3SzEBTr;=P1`X}H8(D;JlZBVt1N$PTkPsH8|&4isu!D0
zjZ3#Lk2s8S_!<*ohfiQnhivBPr~C7Tfx?$pNe)ONb=Rk2GLeoOUYWaEz=jjl-|$Yl
zis==az2VFcb5#7O)vz)S-m<sD0nO=e<=oi98}A;qwinwy&2>%k{F2`6wr6l>s?TdO
zLKVGUM{kyh*u{p8zmDENXh<JW&**2?wH^UMR3SON0oCE0d@mP=(T5UC%}hdx6Gf!@
zV_jQ?3yu4pOT&2rWkHzR-+Fyl>68%?Ggcf7nH-;{eV|e3BU6)Ov!jze_=?ub0*x+8
zIi>3clR8D2><I~YTlYX$EB9EaYq_1fmgne1lOKekVU<`ncf!N@yiQ}YRUX#`Ps~<k
z8+Yp2UZ!ky20}w<8qL*%c6K)kea@*2Q(U??tjQ`WJLdInRr%}n(<$y<=Y5~)JNFxU
z(ft#`w@-Vx`PYplI=WcJ{@C{D#_$VG{8+8Q{c!!Z8dEdhgTBGd#r?-D#|WuD7medD
zP><yvTEP!Avcp|@QFkdD7Q*p-3axhZ=@kF<a1CtGpSc;bNNgglU2$}>J~Uv<WVV46
z2c85@#kFmBfdaLJv|Y0TI)n+y<AEawBJyn$GdzK+CkpZA9#h)1iW%L;Bt3f1o<4R=
z$d<>vVF4()!eLd-01s%=Oq{Dc%7(tJiVF6*{>j_JdwQe=sIVGe8Out9Xd79f)JjO7
zmbry460BxF_vB9VmbnTz=-7Iupd4o7jlkJ=hUQ-gWiyE(z~<>O+l2w!q)MoZ-#g$R
zVer-qp?;(XARvS4WnLNT9_9aHr^x`);j**go|@(|aw6?&>f_v3@m_!WWgdK<2w*h7
zYTam;<I`Q3=iD<7v6cIvcIgGj9D3$krJw2hv_kDKWyG##9Ng2;JRK(VKL4l5u#z7t
z)F-rOlayML3P`r_IJg_41XL(IjLu#bs|Z0Q!n$7|3{h<bVAE>8GWo;{|240i8prAF
z*m_B<a85Co`w;hJ$K1L`cJ6_R?a1;*sTVc8`IznV^chp1M#{bvSlzmPCKC(_?1($^
zbT^Odx~w$C=KAT%zePi*LntD+%p3R>O1K^+A>zRn9P}3=rmVK5{|@JPdGf@%W{zX}
zU|_Tyv+C(Ei`(S?u;14*r?um~|2pI(HTQPl)-JSUMf2tC*aUL={dV!b6nP`gfi$?Z
zTL6Ew#^|@Yoxgk1!eVj(GlNp9v~|!u54JsV)Y|UmYR{9WtiiGQ1RuTHDB7N@bVb_L
zs~%nY^L?rr72Ao?boiVJJbQO4CZMU_A5RhF$9$Cv`{|ca{)fjGDWXFJ@u3~*LUY?W
zIQRqgvAH&Cn}aSt4%TlAg!*mMYYRxTdj)J>Ix<4fIBdDrQ=BKv_+XLJz?8gi$KPh7
z&Z`YB?zIkoiYk7G9@`L6@x*1`{1UHVF5AoplEq_iQS$F162-pv0RRC@G^a;8dx%y1
zf_JMpxT4Z(q&Sh49$004ov{LdCkloQ?Tz3JqmguoVyUXx?%vu5V5<|LPVg9QdW`cs
z1HdH`0Ir0g$B%`VuZ|H%LFs;WyF5csARHi37`-b2VZZ;am(NxIi+;Hv!|FTX^E^z)
zcv#%gIa6Z%GjUYb^YM=7FG7XJ8DD)!zI$jXSHHM_W;6?I+UMy>UQ!tNFYwMbTc0NR
z#yV&;{bshQ;+uPMp(ilA&5A1>`%<!#CI(LM#Oh(RGX9YOEm6xoi#+lC-Bwx>7xl6e
zqKB0uw-(rpuzEeYkBWz%pF#Piw{qxe+Y{9ibb0>T&8c4(&cC_(|Fh+<x8{H-TxW`h
z*uEbty8l$002tJ(S~ZbW?_DQ6P`<S<<smqDQs(>Lo#*oom2slw+$QRQCPZg*o1n?u
zCER5kJNUZw;k5xL=+U0&y5V(~TT97zH;}g(YqsINlnRC2i!@=UTX#hK@?(FyUyU72
zvo6StC$RRVzC8b&72`@!d>utEYS{E~QsEj0HrCI3Zv-%A&yNDi0D*j4qJp<UV*uvs
z?a10g>bxxXpgheBznc2bG!9(bySB^ZRnHz@lQvqY?XsJU_0bw;@Q*yxOwXPS7rhYB
z&Ob#sT)?=OR|Yk&wm|c>28<eRVV<2$9u;;v&5<A0I=qI}at<$RUY?<fs7izJDqskP
zf)SNMzW-PX>{#*Fm>OqVcfvK3<Z$}>ZiF_!5fzOeH)bZ<w(VEFx-9oPgfroxuvW0W
zsR^^qBmS}2&>NR-!eLtNaG2l|u3-cP`tSOFQYXo(W7ieaVX|c%qaJHbN&4Q@y60iE
z9Mq8&<tm591+=ZmUc5G>&^_GaG+TFasgT?Kjs1Co%eBPuvfI^;G|2Lhg=(Lv-;!~Z
zjf&-X=HcWG%y=1OX)Go%e@7*};T~ER3(5Z&pJ>I;FSyLTjUUY=YR!HMGPNT7X_Pt;
z+>DfsfIc1w#tt`Icoe{zIYMoF??E?h{VnX%h%wni?;!ky?|<j;|7^-k%wi9pFN`04
z&)m*WSSZ%aDh<#rHpG}olEOT*Iinv+*F!7k0<~z~kml*b=OVXV@^DnSI>_ovvXcVs
zgdFQ(Z+*-M$277RQaNVf?MVBTZ3{eI6Et^u^*M>e*UcY5A(QTmG>2MRne-<ir?4@9
znPHg3(6zU#3(VMDX7U=nFxFw&*7Ny?bgz_pZFJ5Eh8=CTclF5?ZlA6_#tHO{h)=WD
zRzZ3>1r0iV(dXDL8^8GhdyTdd-QP~Kavvx-7U45taQ}6yvS)0cf7A)1=&`9#;T}JF
z$2ZpcVei?d_X}~Bd2_B!ZN&`flq@Lhw&C_Bic|Xr?&6A_XOAU16if>%6p38;aM?CL
z5qTiSHQrAcg-u<o^48s#Lc~Rh8iLn330)g#laq{DdZT~S-F>M_P;NL3VUc9`9?FHs
zt6@!!hCWTvXr@~(ya>RA-pMDW^^VeMOMP__h^pXbotws~DMs@iKLJvn*8GU?8xZBI
zrZRk%H+jL>i>w>&wq`^08y@X!>GRX<%)xo!@h+3X*z@a%iLc~3`*3<!`@Dv19@?he
zxP8uR6t7YXLzgT|mm|i;hn+KKnMM{t2O5S#qY7g;Av3e-vNxQ+Jk@SebINBgUyElj
zH^;5Bv3uT!;n!%35HZO^le=`iAKqe5H2XM4sNf8lY11lNgnR%Gewjwsb{A02SW`H%
zA-zLCi;5IBk1khy7hadt#e+CE!28gchZ8k(1Cw3|Yi;j$-)%YS(u?{=Dev#<)C5zr
ziY96tg9$S+ZqseI)+48^Zb>R3HRxVdnlT%TONiT1AvE}27upO(-yQ1aTZ;=~I*OYe
z?>nf(G!I+b@9p3+DLn<wkE!!UCa>v}_@;R61;5~}jVl%vF3r|M?;ZV?g7rhzqRpcE
z%MTM%w?jB1<`+gs?xi_v3<-FU#w&IHL@i>Do9#hcozuVYADup!c=79$A)vWia`2+c
zcgY~2pfOh!>kcp;<>&JZ!Rh4XeC(K^FdzZ@;eXy4zkXt&yAY3ZR+^{IlAe559Q-Qw
zLH?@Rq0<V)*`@nl!{-|zOX9)CT1`xX_#U8%ql%Ha=9oMS3xhid|L}w-bDR(TIh2_H
zcyA5VQ;N?75y~TesB;q=E~u@PA})6@J{5J1um<^b53J)vw&QtA)PeN~5b6PWCE&;D
zk<71u^+M{i8As_;4?}&CZPyVO-C31an>%B344fsyP3QJnUrtj?^(6cdxe%umfbi-M
zKknpH%2_?=TFkDlT#cg*!lBV)uC<JFzYBWak}NvuU#16stNP^CCww7d`lXN6ATj78
z^6Ac?qgIvYWz;D<B8POlzj*N9zrg=rCBl)hA)_TX%0ed2mfECePgJ>7bH@NOCv(VO
z!CMNp>gMXn00T<tp@1A2%)Rcf(|sGuxIQf$3bk|Xc$t<gK<tzU1SJ%WFsNOxOWiLV
zQ;&HIs@eK(`9KRm_G4vaRNeFQ_s*zsu-#z!0fda%z7zmKY24Vt9@>{;ks}29Qz#*w
z1r}u!_N7u{yToPtQsp80Qt5vCQeFS=pA%Oh_4`u%&SBKHCTd^m7(!&bFJ(8eFQw48
zF9jE#0g6l|v%z!AAXy-eL{Aa}_NCZKKr#F}z*ZtTB@*OH@~b5JX_DQo%Zz;~aoZSS
z6t1~1g;lM~1XznG+vfXHGe!GSd*<6~dSXM@J&3%R_bHVKa1_N>pkC}a?n{Bd5arx`
zDcpL@zErE_zSK_42F+I(Si94&Ct{7rqW7f)Y~W<qH?a~pvfok<kT#kD0)puHeW^{>
zTq9v?$d{AuyG{OCQx(vHq7+H8=fQhB;`BIQ3BHvqS=`u{nkN&2QOg(tLxHe?z3c$q
zmzu^|iZ#1YyBv|iDnT3(gB1_V0VGRUBtR52W{m3BN6k6cPTMZRzjW*&RSPq}2<%xZ
zeYWm$0GtF(8Fcp@T8~{n&V-)4eLmRT>vp=9t!F^W&&C1C#oJ?w5XGfjQblWTbVsaq
z_Kytnqo?8t_4%eo&9Rm6VvVsI)1Y%M6hZIww5;J;Dpuw602<<JG!)6|$r(`z)L$xN
zs*%x#a*cymu@(8k#K1h>%$vX8Qb(_e(~Ev+4(g778Z{9zQK?Zei}ZQpNwONV@h}Qc
zsQOD>-dO3O(j{H<8WRtGsZOzzXr(DI-8#lm;bGIC{%++z^zIr#3!P)Hbg5XZS3u*9
z&lO@PGp#8bYyQ|Vr4Eh5o^<>x-tF9B6)t_ww2D18q(@$gvTl_4I<t-xAf72lsOobJ
z2YBYM``F{12CE%4%}`hJz}#Y@LnpPaPg8U%|Gs(njWZ|*i=|p@_2o(IWAA0kr#4|z
zV=vcgJ+qUoI_1|>+u9n8rFt-_Z-E(Q{N2&|Z@&5dx9%Q(EYk8yi8bnczmW9gDdv4)
z!h;_<#}O!Jo0Ue>A}FNtWR#H8T6W1LN0zjtcsE!vh!1<QHGg<5m2LJj;^?|rl}z!-
zY}@eK9wH)5brtn|VR?NIvgE6&XDG`e52x0mPw>uxPU3OERPt2}-g~$AobbeoI6AgH
zgh=H`-m7_7iNP^K)Rmse%Qt@K*KZ#F){=`AysV>dzp@S|!Qhg^uc>Yuu50kheo+9q
zLV_PqLGcwMMSpDWEUpulRj`QUE%FH}n!4>U*5k($Z*=WQgck5?5FRDr2%&=&hThXu
zZHkcGDz&|nYyq4kp50AQ<N7|{xW10?!6r?vUmzw*=H82x%QeS<8;pU_7ZZ%4cDg8{
zx$ccxw~j;s%y-%Zj%sY_m@Y<ja;^5I5<P?w*-P>C(g-S<1JFlzuR6<1L{5P}S7-dI
zu6^-xRzf)+4@!&Ibg3&*D$I<jD&#}b@X2Zll=tHc;-AghBDu*B7U-g9@ZF_@MAmMJ
zLhaQaly2G&JIPz7xswCjrWSyV17R^WJpjH&tniuc?VlxFPRIx*8Onl*Th7YUJ^<5n
z6Wgh4HNTBzY!@p|49c9^y(!*%>f(H~SAp`|AWJE4Znjg-H_yprW-U~I0;=tNNF0?R
ztKysJ=jFLA5K|=~!AOnwUELcmmjY4V#X^Ycw~2kqUFXe1t@L^h)F$O~;cWq1Rc%qm
z_&l`v?z(Km14>1Ho^Ww;F$(f{FEC6-B@k{Ov`p!2ZfAYhca3MfYUR^Q6e-(xB*kMl
zoc~ZE8nQe2*x;9I^6x#~`O*qS=Cebh0t(@+WAa~oHUay-u9IP6_nmCimll<hrq!>F
z)c$(X&pruI<CLClY2Cf<`A<>cF&{%wNtq-jwWm%jjom#(Sakm;_^0Nde);3i>wP91
zs^W7$smiQUo~gOmUu_h^r?ln|?oLAtX97qH<i{A$&E&HM+DEQb-&N=tPQsvk2SPf%
zdZ7GSohG#YgPfjqpzOyE(|(_~=d&;6@3=`*+4gC)S-(LRcx-d_qOwl`8QBMez!bwS
zSZM$8m$UwId|8Pk+T<dt5-$0$&aW{p&SfR}4q28PRNqzI3CMWV*$<C*T2P5hRy!2!
zam;3HSO)WTM)h45>mW(0>E7Q={KsCus!X~aA9I~bf8{1DU2b7o53d{TF_w%%8!`6-
zFb06^7(_>p3C!_C(jmsdL|ExTrt~Vj=LBDyqKuYd$=iI~?CU`I8S(@J`m(|^K^|7b
zWejn}hMcOM@KHEnC1F3G#-pjL<Bx%x8@2b!#&G5FpMLAj=Odu^{H2+q*Q-d-^^WTE
zZJdb27)<~rJ;`+_z4$(YCd8O4?ZAP=(*%6Zv_NKG3c7y8WGiBHt>CJ#LcTIT(honj
z`~XrF<8DZtXR#4s!D8R*Jl3bDsH5ABaY*FTt+JMHJ4?mS&5ycOP<PbSfrqM(cWs;B
z_rE#uk6JiXlPY>AGdb$Zi|%k1F>d-=Rxn|4n$D3nyY2)+&DYvcRD3-^DEz^T=8hsY
zlP=fZpg2lC$Gg_XTyz=P(PCL!A*e=xT`ZYca4B)K`ao|WIR3^0_i|z(0_&iGx*gHS
zlL7DDcL5X5$v10V{0J;pBCV|JRn_CCU$iO=EoV`iJMUL|L;%0Z2he&U-*{aj3#@Mi
zR-T23R8STJa9cKFd!9_~bVHjAeIi#*yg&(jCtXcLof$g@E)9&W-gCw1Ed6+&lI+bV
z0_2xLBH#Czu21_?X-idKINDr)^N)J{-=BTPHgron`rv5Jn}B;4S@H$_zVlocUgAa1
zv?I@Pp4#sX;bRxSnZV<muI-W(676~2VE5?{vy+VY##Uvk?6e5<G-Y;zqylFEGGul;
z23G;SIlC5i#0%p5P7BYk9PS<&fR3}8+(=VLA?%)Ge82?vJpZV{QxTZ?F84*rnaL6(
z!#(eJo>-N@&RgPf$n#NwPG&30NNb3tVC%$Cc6sq*Iz-y;12;V&T}^q!!ZCPc!`iem
z-p2rH6(7#Fk`MeKn46QS80Nq_${UkPdlPGPF_*~<W`pYWJ^hF8IlBopj!zc^U@T{P
z8fIQfOx?xgbMJKQh-FNOhN_jsb7j9<r2N;a$b9e5u>(jAi}SPVYOZo`3u6WfOP*n1
z$T&_Fs&=SvGQ^OqTKRA=KhvghfaaV`7X;VEo+Iguud!{s5)l=h)6%zgoL8;%2A_n(
z9?&c*oW`F__q&zj)s@>}95>;xc5rMutre%g_IRdD!*42>$rVcXP5;P0;0T>&sU9tP
z7j14uWbSqyV$@wGZ<L-F8@O_?G$v%({az;Jrp3RymjB{1|MRaD4~v|!L6)tA7#MX&
zYx76#UotU=S42k%Rp4(TJ`!((DV;xjnSAH37645|A!ZUezeLFMvOcCMg!ZHE8xe-1
z%;LZ$ynQLP+`Y6j_C_cEwsm|(HDm5eDr@eD$^_Jn9lUPX=JOmnm|%7@%NR=XP2rh|
z_oWQgI~6VlSGjhLZx)<5S9C5i!=3r=;_(1X?ayu_VjjBq>Z|CTD%sGy3pN6<3=^{M
zu2q-t7~LQCC?5*Y#kMc=VO9>l+3%npsKY){jzB<_`U3SfxSGr{U|;4Vf9*Cc<_)99
z)cV$D{wIR;BJSKIYjOU_RrN3z*Kcf!*1@5yQ^l^BW-I$Wjbx1s=3!mJ?44KL_lCau
zuHW|Jv$30HcHgx8YU@AK`200;oZeu}a%|gev9&Qo2DP8Q>RVi$)gKY%7{agy9!D$8
zWWr7L8*R_E;EW}mkc-c{kS|FGm$0i3mr4)dy2>K66Ka?AY@?%ZUdO5$%z&Y6+P+kQ
z>Xb(lGXkVG*k?V)=(Q`w_aZyysv?toG>})p3L2<ep&SujC<YY=xqu&#AB%73Hp^W4
z{<l7U2E_(ky4BjjQlV{p(;bmaZ7D@J-QXmH%H=I{=AM&i-G+(qF`Jy!#O0Ji>+f>z
z?Oq*Y9kD>X?&#N1!3y=?ctBWO1hJl{z#vnXBH<dvIT12KKfTGwMC^#6Do!1#WQ;=w
zHLRl`57I{{*tKRseL(FV#>KnnFK63-^Wz`2KJ%#!&Q<1goiZa~9kK7k8H^)nY_0&b
z?Y*euShc5uItK%uC)*u_fULRt?4U*X%`iV6W1}lCZZ<--c4~m%NmdmvvqU5%&LQ;N
zhq~!~DPwZf6nLIW2UeCOccF>xMIcph;4TdPbXT=$7D9q`Ug#6ZKjeZ4xaA#{Ym^#<
z52|iyy@NugWcYRo%~yu8Xv5ZJY|;d0XLncZ`Mi&@gNmPD=O^vekf_lTRlTBx2$slR
zoP-*fub#Fyv3v86I`;R^zDxj(Xu5|)UVUuR_pYT%5w|rw2vw+Y?Y~{&aG{~J#e;Mn
z_bzYXoc%$YGQJ%}gBnRlDddjj#x~`+&+8y2tn<4D(YM(vsz9huF*Q-sG+%bqZ5?Y~
zhF&EPbO+`$Erw^SWc9sapTe2*#;m@;u{X@dUYfPw2YMTSc|NJXUt*CIYk8~-<anwV
z=^y4QwkKq2ZqcI`YG(JPR8`@%sbZNgs#A4-*9$$_)sEP#@?7e$|C&u)e~3FMY)1F~
zQH>^8RPx;tPBJxe(0?@~@uCc#9AREqeujIkKcSm-sZ=3}fTcsl0z9nFCw{ypWBGt;
zo&5MGD;1vs@|bcdaW$>EUvYGN5(r&M#8|Pi^^EM6xau`sa~5RXaCm;K)}i76Db8*@
zZ-BH$wRH$n6{{1Iy25WbZp4HKlAer+*1p{X+m}r~zV`9&!~5@g|6hUlrLQ(f%9~YN
z5+j)M(w9;t7!G#!s2;D-K^grCn62-Y_NC6e%eMbR(_g()IpL^?9%R%jzyakNTlwep
z?fr|K?PvZi4Fuxcca_k@pxBEqG%9>(D>U2<xliN2AjOJ<HIf4`7vLUE1D>wxe&Ykz
z0T43};MMmVZ$#$h<)@;@WcBAovH)Be(1%S;MUUw3*lLi!uBixu+0K9>+08B#X0e*%
z^hQLB=ycRZ$R+DO(W#MP&U%O~rd}9cMWN_b>l(_GSqJL~lyv$5Ox?PU_U@(ifu6;W
zi8zS?=U({{N^=vAB^a_JG(-J)e4gQxiL{;kCrh|xW?+3wo2Nd$D#Z$EQhkNN_pfEr
z!D+Ly`a-roxNLO$(ccxRK6kQGw9tIcInGgFS9q5`u6)KAh?@r8V0myVtSD}DX-2`c
zjzFX?4nA!Zrq=Xb76~0mb-wQOtR1xULxsn+kt(Y@N-h9Nz{7L*zjdy#$v7G9RBH1q
z>#xbKyXqGAmj@2{XBB4I9y*)l_KUoedJd4@ltO<0@0u6?#Lnp+V#DKgb7;>OSSHbJ
z2!&ysq&c4VTfVN@fqB?fg!fyV*6Wz(4X8hQ@N7EboZ~_G_@^NqdcFoGI7oVGWLAHR
zoaSiHfchD4Sq7L(a`8!?(~1T$g(fqbGzs{y0}(Pl9|^>Tyjoe6Fd5kpM99M0dszI?
zURx5ELm6ArCjz;f@hPH`sYE(3$G%`t@@fz%t213-`_s3-bGcKqe!6GEIJ|;KF!K4X
z_~G1no^^5UU8B+duRAC^Shtc**KwwBq)Zu&n6BGQR#PmiLgQ9ao&}@<E2buDJ2jV|
zJ2sq}a@N0Rv`qu)>30Sq$aHOuo&s@x7})`SVAc6_dA~qpwml|cPhDSfs-=>`wxa5V
z428Xq9;Z}bSNS_b2?8JRgAJwWTViAs8T+Au@y+r#zw_}QG(SItD22!O@7&qg24TTj
z=C+RwKu8}XI;m+Ol@DZqmmD3jlI`5pdPRY67#yAE(3(a@ckxx$su-KR)%mqn;1|0D
z$Q>lT9~AP?bIWu-f#;PgpS{zu+OH>Z-<Q(1%o%!o4mk5-jl#RVr-N2EbNDcj?IKQI
zZH33(+B-3l*2+(m5OdpjJTY2bo6heNjauel(cq+&&J<4~8iYj2w5Ac!W2`ww521H)
z*Q-J)cXJp1sMrxq3x(jt8L=3*`d_Fx|M_v+l9>A<SDCP^g1S2vnZ2}|R^w{gQCIR^
zeaS|BYf;?8!e{1|4)r<iLA4&C{k35lJpq}GWllWZx@FqB30<WsuREA6h!gInjevd4
z)=dlNYI3jgFP$LRuO=EU?!sVJ#dbVo#c-(IP}HbVVrVlXW-YKYGB^r8r`ePP_e)K?
z!Ulhb4_;B^t=ui_sQm)Uh21Xmi(Y!`k(7A!210|f+s04dB3>(CPENedB(w$nk#_WV
zLP^mZpKTWNCnZ4)hp!l3_jj%5p2e*?lszc@DmG-AcIzEASue)?+mW2$waqq(iS5jK
zoWMZTfmBjPAUq64N~}glK&A2Pgyr}vYseA-?&bJLi<HCdgIxK7(N;rO*HgnsmEAh}
zqBppnE^mn(8V4lCt>F-`%ZmeUSm%ceritsG==qktnCzIRw7#lGXEgkp2O`L%Y+(JM
zqp+~s;JaC_6W_7(WU$GXO+B!TfX*M&`RHTS^spJ*=iwvo*h@A(cRxJZ5E2%Ut5*Du
zWV_YMQ&MqBT9t%!5T+GFpRe^o6eiLfeVqV#aqE=|)EfO~K`6nrVEq&j%9*yib=Ic2
zm~aPIIJ5@TaK?E(hqk@&Omj(WMb(4j2(2U8k#G>!cYbGG{$la0kE!#C8od`+>fvn?
zWN0xcI~U~M1#b}Nc;(`kW6q?&-u&~4@_+m0$U|`@xuIxguC)`ocYNoX+y-Fh4}8R<
zxe08l=_ceAmhJQQi;EZjYFz~#B(@I)M{?%|&#R=!ij5iIwv)arvqDSN4?vmYQ<CJ$
zQw_}FkiWTtGi{8SuS=qV4iMuAkNLDIpJ?`@V}jzb)ILv}E?4Bof$;P7yH)3opG^*-
zFYU~oj`?xs0;Lh6dt=Ay(OR%g<|WOjM@<iLeABAw90r@T3@PBA3N%k2f05WlRAolM
zdXXkwxbmrG)pSO4!s8~Evd_kW&Xc<pzUm!*Wc{`LJ*|SxYVt}N@rj+J<pudMPe&-z
z=l=cQy!*lh9Dwh>u`l%#;(Es_%4B6?6H&3&Wq7pk{d|K-gD1<ZHW?R~Y5OMzyWL|m
zqhIsxREX;yeIw4fo3j3J2%zx{T-D_YuS9(3aqDju|8uXZO7CDMUpzB9%E~YyMPy90
zN;^*756c<8kaeN`O<11i%rU>|_lbd7U>Wo~FDGGTVD3s+0<ss;RX3S<zHG{3Un*pt
zyO-%(&4at%S!fWWf)^9Q_D*GH`VKXE3Ob8IT><%>EdVM$KR!41_T8s-j>vc$8W9q@
zyz0C?Y4UpROAj%eNL5|~Bzn3BDvDa-!`LiCZtj5%qtg-1Q4ZVnpITp%9AI(V@fHFn
zma2wRznq%Fcz>Yo%aQjBRo%ghtwbM>nbG1hi+V?sC7<GWq|$gz>Uh$=lrM>A7S5oy
z)bx}v0j^60pkCV3bbL}E*2b4&c8yJ=1~1RkdJ5#&o=y|#fjlc0@&Qt`7ibC5ZWX@X
zUmVQQmv26jLsZmaafmr}icvIB4iQGv@bmL>sHWff^^!=_>2&5g^I=)pS5#Ni+m)U)
z+Vp5~XWvj?;!=gp!>^2%Fl8Z4uk6(=ZUGEfd`*nIKcK?$HTr~I-fO=BtchJh<UA4n
z3wJ>jqYted7MgG1-}H0K@{K-R5^Y4T+Kxs)6Cm=S32;#N&JF!vcnhE&LDO>8u7dz%
zr}<1*(BKZb>J~v7gm+$Xw1sWFc^<kv#cHHm_RWTGEhb}cziap63nYKW7zd;W);bh3
ziUz0MCyEUn4<Amt;p?ofbvE}q=~UD$92;CCDHmqg3CP5lPHc?)XGwNq3!=0aT+ici
z^-g*m{t)rg($I}w*9}EO-(H%aaS>e_DMPWy{X_SVUw)I3zGt-BzLeusL0E(%c1)(J
zz!+-C7hLf}K9x}(c5C__iu%{V`(5L61?!*EQT;nN<1RABSf?C(lcT@h?EnzYMLTnH
z3P7TCB|0h@^;SVb#Ka;BEWv6FrO_E@5J*%p$0wktms!CxfZ$n*i{{RJaAzcOV`~hH
zM}xKXq=Ba{&q85uUy5rJF;x2FAEl%|Th*5z^W#o-6c$Hfm7dmjWnB<%7LPKEdb@n*
zc@;d1b{Yp?hOY!aqSW+v?Cx0w_P}FUJ76%sRo^EnfS3h7m8g{<vG4#u2#`C=fjNl*
zzvxv30eV$NpagrJ54_~Ns#D2#6>L<SM{oE+4c&nVIWG^}4=Cu#?&*WC2N-CQDXL??
zq^ke<p*WYoPPjL>Rr@-(aG-zk+cEZ4aHyi)a2*#>?N5xtpY{>$HIw?|FO=%&uIf*)
zc4kxloIMtHqo&<h@YVb0!CN9x+q4m)m>>tEcff)ll*Scbnq8?a-wQ8C>uwj0gw>2u
zUc);<+D-EX<e@2ApYPNDILky5EIO+8T~Md?5PP2Zht7a4q4;VcL6<nsq3(9<8nE};
z5{SeN-(l}e5FKoiw+7l1-dw0Nv)&1o-S(v2`ODKk<o((wAH6B^=IFb~;fxsi`Fqr^
z15eGNKpgb8<y2W?z`GEu+t0VW{7qqk?QbqZb;#<r4x?UYxz`<5s8>ok-*}BT&4#b(
z^7llJ$a78Y)H^+!`%)w3;EveL{PzxXFHdEb`&!IUBgeM=C?E9~`y!N1TDyc5$1hO2
zbY@dDNzLb&A0wHh3!MkH01{Byitu^T9YI8No|@{>2kNNM2}U$B&YpColxM(|IlgNs
zbE2jQF%n#}ROxx6?)cL;zjor^)%aiSufwddeB;;R9un$83+$BR7ByY+DHv6_5V?xF
z#FAXq+dXj^b5Hl>Uw!*xFa5NFjZ8AYwsj$0?V$a^Zm`i^g{$mtU{?R(kX8KNaf8{o
zskj<OtfN!gJ3c$Gt8>yEn#mV^ORajFC|cop`J0>qSx~NxXQvu$8^7`za-cs8YY_-w
zl-nTrTe#G0SrzL)f68STFqu2)CIrPC%0MeFvkRrU*tpw#!v#c$9xycAO3i>KR`2u}
z30zKrwH4Xk-f5QutFJ`)1QCa;FJE>EMD0|gI2^us+!51+Gx+Im!|EeDg%Q9S;GGTk
zkY~j6BN6(iOsjqU@GA9@%~Ujo&d!Sw#0L9ETD7*{tqtyOI<M93H4v6P8kswO7OlRd
zMPt=!5z(QzdmC>{|J&r>4FCIHrN3Xyyv<NKMG230@LaT4O*cMlUgE$AniEv3l>DXC
zBRB|Cbf~^nw2F8Z9zr@xXVN@)t~do7k6WyQ_cAYqc*nI)Mo!<``@l2PGQzl6W3r_f
z-@`pMrv|&+7nWtTGm9wXJjo6bafi)_*RZBdkd@^Ku2;NVZCIBx$}w}IdZ)QxFs$sQ
z57O`>m|8D8-ro_rlJmtf-$t^e@1c8KQ`Xm`VCy^ruis<TSj)Af$3i!0##=idqHx_o
z$wd70L+s!~V7J^TB2v+6qFSr}-hj17egBiF`_c`X3l;5m`|tyNN$WB_CnAiqxM7{5
za+I`yPi<nQvN5yhXOscoY8!%DRY^{mimr!~AKQ5l#s;+H<e-eVt?ferwRw$E7A{?8
zjYyg8j2Vo_t~S__{5IPdj+8Z284q8lPYhRu#YbJrgt@;CtVSkyESZLf_p{|u4a*qD
z7e6q(+t6fbJBYe9n#oH)?Q{WH1zvw6&MfQ?HRHyieawcAqAP}h*!&n(H)RK+s{Ssn
zMqYD_y+Bz2038tW=wD`bo=@O#-p9(Y#^AbvRuY4Xmo%TULXM+ghWk<|{w8(Iojcci
ztAJM=MeXj7u(*YM(r(&&U31meiN#Z2U$N(R)p*`;B9cW1;qTdi6f<WG<gK=g7odt{
z7PiGH9N+U4v7w!BPUkq9$O|%NiE~F5_oa5<M<bW9lQk)L@+ep<H0fvXt?BhH4b&rT
z)!s$1A+4>xIPqJq>Wv!QvXEdm5M0Q(!9`sT<Z`nQ6fl$&5l$qYuzV6;fgbPoh)j89
zgtin|gi@LZD2T+KEIQcGwP;JSqs<};JoiwclUP}>qfEaf6alB9j6bcwZJh$OJIBM^
zc>ICN!7D7{O1gQ=%GEqE3d|OkobJ$DJF^qT^{2QbQ5>JQ$ZJL!TNsxwdVk&=@tMQQ
zi-4@JO9~j;Yb(_{ozAv;#HBhnkNai(+iewLfSggc;WnXudf`E3KU|#TTmeg~e4?qY
zRvl*rJC*U8IykuaVGSmRVZqbx*!Jc12oK&uURbXB&revX{OuC1?2Yin#T~Dlyy{^=
zM#G&n$CV-CTr{E>wNvip<_A`~;%8NkdYs5g2I6|}+|c=BZVb8`6{7ns<^4USaOq%t
zeWi!nkb1L|*OjG-Q2VW})Uv9(*%iTvi!ZWAYa<-0v#;yOu6Gq}&)M&q&$k28d8@Og
z*{OsE$NI?jJsUgzi=D!@?DhvItcMbLvi3!uOzMEA%MRD0xq67Uuv?>6@OHacHG5=n
z6JfPDLhETbUVxMvP^b;=LlK7zu0#OUjc$2)(W9BLJHPyhe4Zs4i@&;VeC)f?_MtL}
zW`W0&M?#ko6a1)jC0p?wX8nwjQBl^4(GoA`(Nwt~q;>*lrCtI>l}&Zy+}E4GGGY|Z
z)uX6;@Hr<|mujN8ji^1g)=Fyw4-YI%BOuwcyM@5L^hg(^oBfDos+LuDCP*U;nPfHk
zyr-W&A_F=NgRgKc5Jcxw;<>C{I7)X7U2eUn(Ak!Bf+I9((ub$!b3EppiUT7suU}c}
z-}|lR?(>+v&xDtfXu~V)j{JtZU4|(ZgqYMtaT{|zJ&3r3e|3BPT$D@M!6BV3=#1-B
z)L0mm(o{L#RQtoYL7Y5tJ+!D$8^Tyq=8JqsHvZfSs_2V=XL433y7aQ{`5sjGFxIEb
z>i)Xbc&~i~`UFTX(99b?x#rcfs}kV~k!~}N8PZez8F5jZi-&TR6w3Epy_?Tp4WXdT
zxZziDR^heS?y#`*m?qeueOQXCx-Nt1I{ZWzlKn3$T!0HAW6UVX+;#3Q#k<h4zHL-9
zw%U@&_iW5nrKL_^KIkxYU+T6&7r7JL2EPl_&L!jL-V4Gaz-#n0+WkEg8_17FY*R=S
zE0pP&&xr5DgG-t_=Y7!timPQvrYT1Cd*Bqt{105P_9t4P=LCD#`4%-AL;!<*`xkTo
z6bY(fWxETBYG5%(QFTxR>2(YSX@wZeTjX~jYoE4b<<T2SRtD&@Z07Pxz`oQl^v(c~
z=9U9OMd%9MvL3Mh&kr+jU1!`vCMA;2erq*>8=)B~vy_Z;F4FX~GHvA$Huqe6S7CyS
zMz6@JQ^tGA&kPd0inKI$lHPw~>yt55WLt29zb{oLDc@GT+D`srMfK8BM6bYs(4=&7
z3hrR<J(NTw%`WU(-)qut@8fz<1}Zx%lU$h}2ZD)k_sTRAuu}2ImSIv~7{G1O5gHs0
z5zq;b>|9fM0EgC|-NNMCg!H`n%gC_O-o8pxT!Sk|eshHxaf}|)^}%#nyBg4>2wLCR
zFiXi<3k@o*jXr_5d7+VWBik8+I|X7qEcKvILno@J7`9TpvE{TD_oU?N(Y$wIS2lY%
zmLpvnLKOmp-MYz{EzRh`J!sn+Y(G9Q5L@HI&ns}!JwDL>cEcy%U~{j_>Sh}BXxIZQ
zDA%K%=JDY`*L?@ipQE=&X3Kc@f;u<)qMyI$Ke`4RbWCkbO!rLqd7@^xkxWpon7iI}
zc!88<J?eewzpudO?<5D2N6kT$&^*wn0OHX>jLmf!7$uC}-|GXwi8K7y02iw@`1V)a
z0Vm|SUhmN@mcch9K)a8x#V-uyuYOv+y&47O>~e;xK5fgtgI!s6)P4{&f%J6|0?Q54
zvYh<NWvXJ+6KI5@K5e2la<&I$g{&qI&c|-}c!sp_=F*9ED8wJPO7#LJdnQH4nmGKm
z(5DD%uO<xH6sz1pWRg>h4dod9p~cqiuBlX_$%~*Sy-J%5ZI}`6lT+EnjIW$-mRwBF
z8U=JUwSa6iu)@@-cP?|d9*}kYY|b|WI;VAB17KU@vCfU0?SU)-K&^~}=j#P`ZX^u?
z^z=AjB}oBDGK2FQfwkmdc?J|92lhMj%VBl-GUCh`wPhI@l{0EU+>PbKyT1SD6Zil4
z&Ht|KVk-%}0=+9r$&uWR+%m&1*D8uX060eT-uX#M3K>X3aCK6RK6rljH)m9=XUobj
zj;?e|D0D%;fZkNLR5W`s-Xu3C*N9mX1&xjHaQ`&pIdiiZcKhY+ELdzX+1Mg$=<G-`
zGR&>fK1>FvTEpXh*3gwz(a_Lb*4N+4!@eMnUc;(rXcB)ZfoBjOg5geL%(Qhg05};&
zDpOhzNydrROy6L(imTcs=^Fp=6&o0T-3)5|w9;0on$m((7&<Bq>8PLhBwBe*jiK=~
z${sbjkOXtK4C8vPa)|yyXE~MB=QCm#kL?e-s!lLH4#0!9y4-8@s`olt*m8ZwrmDlD
zGcECfVQw6`$;Z5VXbC_JH!Xr)o?hnT$^%F2Rcsgz7Z;St=ETCv7i(7WozVE0si0V0
z(xIo+<|E#{b?9gKgu$ZD&hcj}q$L-6hgSc?i9;`1nyXwguAj5_k86s%->^uTd0XHK
zk?U-|z#@hwYUztse9j;V>5lXI{M0<`x#E1SoH}?sZW-7Us_y$4vZJATjNSD{O6pWq
zcZcF2B6Be;>2ZsEIzHTcap)D(GxWsspV(8%+15$?`-`}E>tc(C_on<-@49&G+70bd
zS_<u7oby@S2-8`<UHiv-rXM?3b5{Ua#|e?v0t#9k?W@j!LU?VCxAjda)0zsw?qA82
zC5HXT44uW-$_=z1%@{wMa}so-Y2%~^In2d8z4kQC-@m^1;k*`blRURLt~K_H@T}I&
z(i_+RDm?o-tH|R1V)9>wXOpw81Gh<i_ctT}j{ZIPPyffM{4f98XT*-O_j}N8i}Oj@
z=J5%X7WSVhx{&w+T5|`kMolKbQFPJ>As*!RGlksR@@eJthAf6!c1NWRNfjCs@ALf7
zibLsGyDM_lD*o0o0`+`VjLsQSNRX&L8XIbFuS4~3?<PR?(|rG6wM3gHoAF~WxQLX@
z>+TZa&p&a!u&O=F%tdqiA;eK#n5ZFiEuco{7J5X5ZI@J=xHprqygrb{J5IFq@YDEN
znIph7>>f_vR^C$muJl3QfOS8Z_m=CTQM4W{bcYzLKD&R+_?;70Ii*Z-@gW@PiOSWF
zjScoN@V(ap9XM^2|7s^r^lidgC_FxJC!M+RE{xXFjLNUQJ=1*jh^)T6sMdeYgPMVx
zz7ob2+!xM#n(rwtjujmf98PHTS;v0ex)CZ=)^@qTNNom_3Lkyj1(l%lONn$%m2<mj
zBT}bgSO__7dA<X2jyhNfz)e$V?1)oK?+fO%*KR{x+zpk6NzW#=-9A)VX<|q+7ljMB
za(TtGA99ZCyFEE!HXVF6r>GNmF8aET0ehs={F6`q<`!Fmy)S7WOMr<2Y2K(O<BZ^j
ztf$5qz*1(X5{n8SVMd3CFHX`(h9(0TL4JOw7v7k}Yd35i<N_+E<GY;P7>Gt<SrB`*
z&yxmCIfYD5sIrSitI&GQo_ch=2DbiZqUJQXreCk+4Q)KZceZDmdKo(=rqLP^DcaRR
ztEGy7-1fHtkyMxXtP$L&9M80biHy!Zu*L1=@d~56CzBfZ4z;rhVKtpI78{3YX38s~
zxE<u%{$TD*cmhY?(|?TKgOwhvXbVztd*R;~K`vXqml|(mepI3tq_UyZ-c*85W2Xf<
zg1kew#D+52a+L|Gd%C_LT~*u5#E}5WlO)?25-~mxr_?T-NvyI-zvoUOS@;;re7}_b
z2Q7qANo?U8Px$kf5zWtnuRC~Pr5&9oPfI^0w7p+-&AuM&ET{T4A>gXb>W-=9fuF2i
z+y}iC1tu;R3g{c^V~IN1OkN6lv0lt8V7Jx=H&yX;2RRMEOg)BG{6V+4{%|=UBTAF+
ztC%qcHpP}DVPLJ5|Bt#ifodvC+eWpkEUU_?EEPniwkSYEKtyIpx*bp%t3*Kv5G4#k
z2oaDW5J+{YpiC)2nIWYJ6hep$VGxoOV+0HlfrK%U1QH=Y2!SLdB;I57?f!cG_p9#v
z_x*aU|DLs2IV3CR?027?z2EoU&-*;E-aEOSx@qlJN4D_(ljjn|9Lq-4qGYga<e*<D
zA^G9#-m$1U!#vr}hbOgip8{Lm&iRI^_{%P@z0UaFUYx!zan)b8_hU5T8guYQE31V|
zDPa9GkHgguK%7{dd*Q914YRkb58D82Tln-sS@}k`OHG`%8NJkPDYkhJ4emQirSu8~
zY^yYrJ|5A`4<NW48me>|F|ZjFM+H^rd+!*xNg1Sa=45VLnsOAznip7@!Rg<BAj1AI
z@(_(sR&Q?AUG%7bzc~`wJXXlxu}$)EZn%Yo$(2;?%Gtda-kDzOJ!@PZ?bR0h<A961
z?{}2m87GwK>XeQeF+4VGlD|<hMoh)*jCVZV011v_5aHhwoDLluC6G)v44V?qvgMRt
zl-4mT$#<Tw6`-^)HL+I|324Y(sXE2kH;+b*p45m-SUI}~=@L9LNplX!FPbA@>9>ls
zv1#mCTfp!2_z74E${dnJQwtH0!<B)7Z5v{*!quss-e?$5r;VzepO_A)icWL3-$B}t
z=9zCC<88R!-FRWUY4Ut(wuN*`LQD|Y54BxJJ$Hfbd*Sn4mPf2fV$0!@QmwpC^jA*F
zO0w{KT9o@X@_&(UEVKk{7*;MZ<XU~is5OV!h7kj40RI+@szrd#39i+0XSgwQ5(Qt|
zd?lO>c{nCrh=hmqXP1pJP^RTWk*ny(25m%wMVf;bJR|0^7#u&{X~7>lv5o~fyyo<4
zoD34>9};L^_b2`xTMQCpCX6KT0?v-u)MQysDI1gL_GD<(gcnJyE_%e~zMfwRY9>|)
z%EE3=K|G5RrUTeBSIUF4>dT`r$pNgPNmAl%X~c3eraZPRzBR7S1ClBzg=bF9Z<cw4
zkNDXY*<Hiun{et6pxco94-+?DFb$IN85{<anhWwvSf1gwDu*B#&n)gLG&DX5<cNqk
z%%wF14+5g<ua(Bvlnj^2zyeV(A<64;Q(qZKlXPTc3vAa`bu02qVkHdOT85zHTib_!
zU$t33*D+R6;V`q&5$;*x>!<e3z0y2l%@m``ndtD@H+`geoB7CwLxRO%(IwFi+SaAS
zwxPtoyz;cdevw_@b(_4)gv0(xUY8vq;cl#*N1$U~Mb6Q2kLDXLGb=VtW-=2@jua5<
zxv%2?yf8`9tqWh3g|s!}`S3?mwFwIL(Y&TiCZQk$U2+v~xd9OW95%m<)U}#0iIkT=
z|0}TM_t(5n!(}}%@|D$BD_2{p##UNWrF!%8{COY)zbkJ8g;_zX*U%*n9Ur#cI1c(e
zj=Uhc@je6!a#T)zMv$$fs&H5#;6Uqk^Qj9EAeuAZqI=l;qfw$>wjp+|@x~xu!J>m%
zAzx4ZIs0P{{m4x0pxzoKpjFpoaRXi5?Ernaf!dUltuJ;?%YmDOjtSAg^@1z+JwHN;
z-Y^?=1M7emSnYuPOE-09Q*~Ayt6NU3&#l{ONPE}nN5Q`4BrKI-pcIcil=(GY@Ru7u
z{JyEcCtCX_Oq*7zh=kmrSy24QiFmgIK~*Qj5T>|g;9Y}ceQ_{rf{5-tc1F6rfn$ij
zjE?i|LmeN~hh!L6o7T6qFubkzy=7muXO=tCGlb-_n077oKn}s&4=4bL7@0fWFqup2
zS(AGgGUr<HlU#gUlf*dOtvUmCaCWj2k5kW}Z>S6#MTN3#xO;ITWIuljN$XED354nN
zk*{rSX&xqx*sQ$`M7verY_@0R{fv2#|D*K{JHtn`=zY4Ap|<qeTV*`?)Ween)pb_c
zj#PQML!K7#FMXGBb8RQ)ux(q@uwyAf(YJC+d*@;^?`>%9Fv(UOB!ft&WvQcY1pfJ8
z-34ilg`e+cVxMX*kYjH}Z|QBe3)O2$P|P&Gb^~@v6ezGs_$Jw5{~#PbH;4Q}`RDbt
zl7+%k!kbJp3nqGUo!yO{sW0~cJ4Yv&Q<qa-$&<y}q27k~i<K;C=Kjz7*PH(M@fmzF
z&N0_O+%d}Td<W2<G2mv&h^(Gjk(#&diAqmBl$Wns+~K9<$H$rREiJM9oGc=qc$_F0
zEAzcwO>=m2H?g?=rC($v^XRLU@b@>vt`8F1NbNYay2{GtFWp@g_7GbkAn^q_m}I{;
zq-(;04{`a>gMzu<3{!N$>TGW=pi)!kaIOCG#HrwZRQiaH!|ewoA>0)ZqX9ywa{y)-
zf&Ee#8gE(HKGwdSy>;Hz^s{>|t}E3HyWz~X+PUEM>1+OC$6ljZEge(cis>ud;^Vm?
z$Z|`2w?5$)`eYC9SF>0kt+{MDJls7fN3PDa`>QVK$B3q>H3vm@NA`oFB5u5IUG`C^
z*emrTO`)|n1+wPI1ACPXi_!9im=VmG#$a-b^^C|}>rfGoA!yuESWB)g5BeHPN-+1T
z&g6;->y<MvdZPtFZSEEt*<IY)lQML2w&DH<4NZNV`fG#EYKNCX?E3E6Lr$cC`WU6g
z5PMDg@GJOb6O#qcC&)7;k8bt@#eI2)t)`|X*^?72*2MVq%*;5grLC={*r1QWV)VI$
zj89CA)8O;;dYp+cALul)YnvII#UNIOcyX!Ny>@}U&Vjw@kEvK4J4S(sTM^h)VRSA1
z?W5ZBaNiy;;zzYQ)KL4<$0z<jd;I61%0N3~(;7?77WvVZiHX?Y1aFslhh+Tmrg~e4
z-p3IsGj(U0V3s~a1b1gt#&m1}12>TL;1+7|%lLrr@>ZPJx8|_91Gmr~)*18BMK)G~
z{gxF%pSp<eLFld0%otZ$$0iPJrXPq-MyvMr?ZIb3l3oOpZI$K+c+INGTwb0??l^|i
zE$?b^%#?iBdMB{s74K!sgd-(WDePz^4!0p8O>=$+iTn*Gh^97mJF0r=Fgn<#@8b8>
zxm_zLji9`fgxDB|)oT40bIPM{2jM{+l<Oopbb-*8-EOm`lC+72c0a1UAQbT_g2Jr;
zo=$I0{r>tN+ra*!sR<A`H80~F9Xl3I6@BCJ6RCi3z{C!JF?=*ZJ@4ij6p7|$_(*U=
zKzA7eJ*fF^juiUmDknnufMu9xTWi&~s_HN%{>0JC;r5|#>&8+;=P&5C>7?Y|i43#g
zrIf+oYLo}K5uMN;6>#6jw#r>8i!m@2TNfv{_g{U9X|(13nd}dJ!$>T>PipR?Axu^-
z6K*6jLxY{+t=?ImC$*Cy<+;^@JXS61cD=akqRYN300OP+G}e-7n@vI_Mo-q`?Z3|%
zcWgg0!Tt6|k(0-JbX#(xPrt{E0<EA}(=!s*En8w+z7Z30kcWBonA0Q&w+;m+-gi^X
z!R!O4iR>Ttf~d%hY^5*Do<_YU0}X{Yp#7%!Cxfh9vy3rUj(R0QK!zi>Uk|MCzD5D{
z;%MT%BaNGvyw+Izvo4{n!)%pql4J6jiXd}{dUTS<EA)aFd&b{CivAVc3T<6E)|z)p
z@j3*G7znn`8oAy>HzSNi{a8h{_2F9DjI=i2Gn(}?DIR5I#p*tlpGe?Wqjyj{^zhz!
zX#-=HuC8No$xrc326-%DEGx!xt(YvtAWsbE+*3`x;31EFzy9V)K&eJ>)v>t$3_yMR
zTk=MH_Z>7?27vnVHnGN<nfLg;+*V<Bl(pLGG|KE|xznMbu6HJIr<;&#zDbknnri*=
z?l$-sU9VF3FHCZKZ{V6YdsZ;Z9o2oHBgMa!gGykGZkxDv>zmWP`D=z(IA%Fv2$hS8
z1A3-}^}0YdWVVMJ^Jt=?el{~v7AM&V)8Wt9L_X|c=gvnf8lW@SvuCwyq;+RQh~l+d
zUZLVJsAXwZ#yKwYSucKtRL|`%g{C+-wIAZB_aBJV>Mq+fbLsM*xsaW3$G@$;;z^Y9
z9>(7`q^Ys`v;l328K)|=fTRXG7otm4agsN^v&UdB3Y4WFEtR^`Ty}GDxN|+*UWk^c
zw9F?I1fz^S3iY4~DrZEa@v^~bn_63kp|me5!rw1?_xVwh*nq4B@zZt!nx8~Js_9pU
z$*)dlDALq_WMr8f@fZ;sS@5lTYqk{d6YhSwJmO{oOzp^RZM`dhO?!Pp^gp|TxnTz$
z@Y1SbDmh|E{~3IR3C_E{j3mLnIt&jVxq|a&#s1>`1YXoPmItI!R`iReEK+NtE`Qh-
zFu;k<c}uDaC55jWj(9*{rqVq&^a$_fEkg?NfI_2j2J}TNX{nV%oCD+ljMvtJ-%LJZ
zYBvNn1;1fc+6mTtlV0Yc)vE0%2)-Rg&)wzDG@0xH#L{p#LKP{!kv|s+SJe28^J!0;
z#eS`oBWMltp!*#+v9ov%sw+TmD->#3CBI#FV`Qn8oL*x6m|PK4$f%j_^Zhz>LqI>j
zj|R$k*nW4kesfW6)bxaa=Aw;z(aRnqR>jxidcj?xDKh`%gmqiGgY+U#yp+<d7%C%)
zZmt!GQ2h=_!#gv%YaTboI+w|G?R-;L$YGAcFLH=Sw1Dnj1TX@l?z&`=xY%2HX<+d!
z?on$=OGYpg5Sa|^L>^KM<+&)0%7_`4M-Gce4)N;4%+8<h0_Obfhsum~4i~8d*ppq~
z_w+|)jkF1~-v>o&S@&D3iLywVHP3HVCznSafDGeNvE>nCzd+Egk~MD%^HdYwYPbKc
z{+shB3nwtEC{ggZN<C>tYLWpjUgtLkjWLAU5VdQWZ6{}&n9vuu9_FogroD)|(w{%?
ztzR!#D~hM}EzfmYkFvS=mg<1gh4>YN)~BPRye2&ZcjJ^X@ik&OZCygDw{9-$VUQPU
z-GC?{hKt_{>Mpj~8#GIH6G#mZTj1ArsSu|*H%$}BzIs7n`bhu?PO>?ZsP2`2kXI9y
z+cSp`E&r*DS3kB=kIM9?)wyWI#|Foo=}<V`juekMIxm(O#w}>EaMuNncifAoLk%l1
zaSpA^ytb3PUStTuZN_hm+^kAvIM^rJCp3m5X1j4}uCae{ARKyit0a|qXkc-Eggr87
z(|j@*tA;NZ0DrXTz<wWKui-5MYijRpVDFpvpv?R#!RD{Nz|Lmo#@5P2X}91AS<|RK
zqFwEB_ZNTp_4<-&LtzO6W*Fm6*G%t)MO1rtyb40D2;Md~HWJIdI6M;I1zvIx4}QLm
zGIhzGZB?#7^Vy>uS)GL^4D)cVZBL0^qMe;x0UigBiGc^oR^f0skV9(|^GN0#DugA$
z2L=z``3zAzPTUX$0nW`w@xnB1K!g|A{VWi10Hi8gf7#ska=jIxFmtFd?}wBx$onKD
z?a>PU5W0u6=3mQN6*V{a$(I}Q(AvKQy&tsft9j5-afH#L*%$ukO;!7n)gd@!3UXF#
z_(OYjMrw><qaa4KENaZ>>@38&C)-X>jjhKMkWPN>78<b@^1sl+=L)E=Bj2?&T7f9#
z3DR3yAvB|0Q|MOxtJ%uw_u)|%F>iHU;3GtgNlZe0&{BIViCD#(BkZfmsb2wDbHy%^
zA=pY@qAP>=VcX?*O`OdMtBAdQlOH7n{^_yze!`QF5gvbK=3|8Sk~X-hYj*6yhi&J=
z+WcB-fbF#Sccc_dfKC7=;ER2$xs6hf2f{zCZHj~&&}RY^z~mr)nV15mh_^7wPKg;<
z-fa*Ys)LSQ@idf~xE7)T@bm;<@w%}%p}q%b$S2;apMp}n7m(jIun5qB(tvlnu>=*g
zPvle;0|;j0g}@gU!g>XgmT6Ml2El{cvQmbZWAV#Ms9vksrUR2SF`L7}<?=R6@c;{h
zJ=4Y&Cz8=1n91cY-yT2tPfgvX+v|LLair|A&$rdi&vHGh-2qJm{Y!><r#!x0^Z0zw
zp>gmqpv!sYWYK;Ae$!=sXR(=R0(r{-5^+R?Ptjy^?Iu8{E3zbpd=vQbFwoyLS9|c_
zenElA9DMB~EDXG__K)BW5Ixjg;SyWIYE6e^*G2QZ5}9LbM|6%)9qnyfs({+^Tj`?P
z(!+ui2uRUSd50p`BkHOQ$<<K(pQs_o<5#l$G84B;@)Fq#YTA2AGoiXQrDpD}8zMoR
z$?KOhl<d)}bf&K6R1Fpm8Z-9_ok)AJAhwct>NGbqvcF_b1t+CFA$iwNS@G3Z1SNn@
z;d1hcSc)c2pWGO+b$d$NoK_{U8X3gvEHt}t0fms$<D%`WV@uv?Iun(Gt$ELjuV-xy
z{GYdIulTBUdzz&ILcwFVPKD0bWcQEXkT__s*p&?vXi;ZPQp@4Kl}@Jug$+4AGh-Dc
zxKv>}&Nby&LS%f-y1YY`{b8F|nNDf(oh85E+bylN9rLzE+>mzPV*3v0Q9@&+UQVXZ
zX=y`nMeNh5M0p-9H9y^-J)(+#f(U=dGxV_1IV-d%Z>cDY=8V_XzRP4q%~hIYT2nO-
zJB|&e6ALMs@q>q&deG0FrvXnr-mF_#J(|=zMgB&xqfcb5HMw@Rk9}E)thkd3Oy>1(
zn@p;HGF$MO<@e7nOta={V<DNr70wf1o<6Z+X2kZch<BfaJ4?(s9t@W)u-tgY(YKe$
zZJb|hZFQcVHGAcuo&krIEuk+!nB?TeMI=ORkAF5yjg>g!yJ^Agj5tWv49=7-4n1VR
zS|!<LbyLx|jbBZLBZ!AnL=m?~-%hYPybsRxs*|1J5bY4;wF-iIL6cV)Vz=CJ)@`Kj
z&ZOrnky;w4w|~PVq`6=0p+^sHj6P4Z8%-XYY1UCec4>K)X1>e|TbPEeZ@H^M<ol~U
zEFi{S_v}^~7|ti$eE>_}{b}ad3MsuK^m5&`!J1(^QjEwEa;O@UqSgLV+*sc>gqSVB
z*Dr~aip7N78N`t=hNQ1#2B@B|d0V4lLo%Yjq9wQ4A~q}gluu#3HR&Sfz)>pFLXBFT
zu{(gQcVe#K!KJTFYTo_xAKAb4>3^_l>bLgnjIqNht2;26aDdyTqpLTuAIKvSfU=W*
zO#ZXyFWD<rAGS4jf7oW_ja2*I^WQSzUw_GNYASL>1a(bNPh3(z8C2CYc6&<WMre9N
zQRc)<dk7`-CVVoo5gt%sYi7{~lnS_!IvPFn@%Ns!6lRvfT|pLlNr>B|!|Q3Si3^Kn
z&~@gsMW(g+P=v?OG+je4do}@|kXe#%JHVqthW3K9nd01UdqYk8;Ma#osm1Yd$Y;<`
z8YS-5)vN6#eB&HAzWQQ%U=_BMSdXD~Rug;*GI(^%Ap*nLxp!@@h(pcUcwN)(|Ax0$
z8Z$&6Dk0?*scU?h%e_rIHMGcSZlNBzN=$5=U0;sOf}U%D`1Ct4H5@W;e)p2jtAX-_
z^Y><!)&`0EpHuFx<I=G1tyqslFMOyD5l$Qu7rx+B8BJ_1wJ!LtARqU8SVXkpuk}{r
zHO$ho-%#;X0llAz$73VeQ?~AWT7=l4@{%;DkXl{s;`;8*D84c({9UbVYx3M?f9{kz
zy@*LmkOr{qGD*&U|MyeSnqa1Mud%PTC4I*?-9Ndqq5WZ-3nfpWn{;8!tJ@If^>shM
z`>GWYnv)8L4EZK5#n@yJ?5w@UVDF2vX!c3zr}tHp3)kt_?7Xc3h+N9-s@OlKlvx6?
zlZxC*UDw43caO};^eKZhBPvg_baQ1^l)KkAp@@B`fkPp5#&lFc<}Fd#x*_uR$$aT$
z(MF}S$B4mh*_I*(&a2`D_hKq)z)y!EP6a9*?NzSXS1IlAE!tBWGIjFA>{;l1wnNbh
zBon8DY%&QOqx6nuvWl1myGt4`m{f=AIXk1!CW$F>7r)*-m`3)PF(-7EP293fEbhM@
z(5VsMnh6wB=PZ!QP7XP`)0n41h?dO)F74!+Z?0|(uT@6ES5UH9gCTF}6)lZFMP|&{
ztcW_V&o}W}o5Q#}4WF!J1P#yN1EXPuTZlV`xWby^9;XNkoxrxRd>xRa{B0x9$=fwl
zdMb7FPL3sY4b}od-A;iqYK_0+rbKqaZu&(O@t;??JU9v2QxZPWee12M>Eac9OdXU3
z8Ee1?@T39#v!~R(y{*3{%*J!LqsV)zenj2QZzX%N152KUVecKX8Loa%BMz}Q@@y7;
ziSA$s&e#KY^{Z^&{fO?GK-_fnS4Lcm3Tz=MM#>J5-}gBV^x=#Nmk7GAZ%`o-*SA7W
z@(D#sfjgs#MieH}3=n9IlR(<4QMkfSs|tDf1;wpOT=&?kjPkIZ;Nm>*7!9c<96>gZ
zl48zOesZIT!Jci%8P2mr5;JGWW+7D1^=bdWih4q5Qfx!~&(H%pxmy=vU-6WV{OU*-
z@4XHs((mRC_u|3@g-DJ;N4|ze^wx}hKyra?^6p}^`&b9JyUqTv_+9tsNf@UkN5lA)
zj-h9LyYrgDqwQ2n9VyF!p@KAUvt!#gdbAC(w@#mp2p*X3o8NMQsl4#R$05D9YoqFU
zhhmE5)W9K=MC7E0R_)jr$Wg6W4aBNl-zhva)+*-Kgjq;S=i<1tXV<9<dWGbss;Gdk
z$H%*bM*~=IiH#fH3|)7}gNvIAOm%Wqw!p#4tkCV|M92Ib3h~^N9RR+|Sw+)S(3A}k
zIBu9pkGRH*vjRC0PlZ&qG39LQ$@qo7RlgL3vN(aL@;iS^z%3Mso14unyYZ=@df{MV
zFP`&A8UXhp)9)dFB<x$?rJFHtdanBZJZsD!#>*3;iWn7c9+~r$=CahYJ9@n|zB~;d
zAvk_H7ByiaG(wyWd_6&nLf#0i<wV<RIkH}Y59{hUPI17Y$X_;!R@Mm@R+oBa1;a%|
zUXq_+KC@^#?bZaNAMITc4XYd?1dI;mXAaX2uSJB+$6FFxUeG+;NZ)c^x|y#*Mlv2P
z<x|ZO{k20a+Cr?xNl{bB`KKcX00i3{KJL`WjITEosN+D`T>L}F$L{{8$KDF5KzogH
z-o7GZU0(jvVtiojT-uUiF#C_wMxL2qX+>FBNFKQU^wC2+S@fhLzsKx}W*OS4rM9eq
z-}GTyeeX<y%M^XpEie%C;}vQgRh)=K^va`S%i<nx#c?E0!Slfe)0eWY1C2QHR2CkT
za%ENgUP-SRTaj1&IxeH~u=EF9x0pkRoxm-hPoOSH(6iz#72;2-<FUX7jJZH{47(q5
zjdXrOq?%bq*YR;5_h7*6ptO-KoW;|11xn1t-vV|dI-}&W-~8zr;+?DS{;A0TdJ@NC
z9I7H{3>$ygX0{Ah4NN*ACuLJ9JMo_bu|m!2{!Fm^;hHipLe;mWnxWqZfO_;;k};jD
zB;=2}#+a)tbTAzm(686U*vcT~vSJ6A%e>cwd3gMmMSA)V+qQjid-abUGTlR}XAJS$
z#=LDfolXa0kBhf-kHfYo3o@LU*@SrQ47~}wlLgxA7)s^RTRba60mS8oX4TTBO4=`0
z&ePXJSoF*C9#uz2DTce+@fa*E^0M>|`t)BI_iugt{|A>vAiSY~GEyqS4<Jj6_}Rzb
zk-SbZv&uYjq<(66P$0hJMNvgwdv<@o_XA549!3??gwjN+VJ`mk+d-$d&xY@QbBM<m
z0xnY*_&&0(-T{dk*P#UrXZc=P^K<DvB}lC5y_^|b<dR9XT+;H%D`Hr5h3Qke1*n&Y
zI-Yp3*;91=n|Bci&~)hGNk2cAiiF$3;vLT#X}g>i9zh#Q5Qk)gx#8GbXPDYj<kd)1
zj@fzV=L{+ih3<S<t%&mTK*b^qHfCpE70o{q5AGVw@~F3CEX1C<2y6jb63bu@ZXdCZ
z=~fgTXZxH%PE;TJmQE;R#^1HJ2!I5(eG<rbub+d{obQ_2R;3#U$4=(QqMv+mEVLD3
zJqX?tD%AS|^!@k=OGWtQ&9Qcw?<tzYAG)>nO}LXJ8Lm%WB}enfX8b+JTiWsv*bFHG
z6GNTw^UAT*e=+EGhtg4VZ(!tZ<6zAF!M6?llUJ*}?u}I(8GY{g$o5~MUjOM>)0mQp
zWa$A?y~a(QZ{ApaKC{8BD=eubJolvfCB1eRy^0$y)@)xhNf@rSB)i|uf3y;td}`&8
zDI$nb3+L6xOsU`v4PFz3r4<>Bxo-c~KXPGQ#?D=gua?`DP%M&50%nFNjA+^9J3BHs
z8D>!Fa=qCUTH3wbTQptS=cxU1ZY!fI{Ldb~_A?4-V1MsIsPvZKUcZ-<!`UNiE!F}>
z0B=iHSa&g_mq;|@jb$Z5OpOSvhMrFQqcx}hZdoq-3*1*){4{zvab`unDCeiz{sF#6
zvvi~V<`~{x==PFy7hyG}X_Or8H7=8@IsmQiQ=YP#X>`ow%&rV)d;J+ypm(2lcf791
z7(jOzGKs4KjxagWDrsj`NhLJcB(0?TN0{}l`ZBaT$gBa*&9;s+!Q|cvv03pnaG0>^
zu;FnMgJ;CI0de^_38}SI7=p9-ksLZx4d1LAGm^RYW$|J-@+5)>j-d7|B4D_((!wyC
z#*NynbySvl?$4`l>gP1XsUNIcinb9b663wR8Q!-Is-1X40NcKC20Yz@=1e@2bEWF^
zTer`tSKLT$_&xYi?Pc_2@k@vX(c01~3eb(KV>G+Wjx*_L`tV$}2-`5UPZWztGJ|*x
zA2_i4F1H_R1+eW0w3!D3>sP<a`Gs5^=ifAYEP-SLpxN^q9fQcsk(vAh(6h3kKXFOB
z9&hYs>#ju>)(YiL{b82bNuJXu3zTWw^N2+!iw+8-P<HB8kIS8(eTzR+8>X0oKQNm9
zuG*Pod)#FumP{XiiL`z)VTZFFj3syW^0Kn+3umn*2`m?n6Bix;kQvv4Mpy=*scNJy
z38uoi{SYp(#?f(kkEw^ZRlZ&|fmBrw4M8AkZN%^Tp~LFb96$`fgYN|}fcF4OL-!wF
zSClUGh}=aGi?($_PEFKjiM*tYNxw_|kE4^o8dnLIE!<49DXmkUWEN0*0}~IL29Eqh
z@}#X)dtjUI=DZ#BcmjnRMhRr=31$Iyc>5OZ)K;Q!#;zjPri|$qd;h`sT&rs{^PO#n
zSmxH$7>hfEO%^ofA*`|5jkDa$>5}$jm!Hav9V?mH77t5`9e@IhFK<lMLSwuh%Jod*
zRgD~j$0AO_Rx!2MaDTM-(J0WoGb5GVtLU4c7GK@9s~Q<m9Z}?kIf~~SO|UL~@ohV?
zBDZhm5;Fco&S1c!hO2iQ1|5#>UWf(JdZht1jHLXX=n0#EDW)uv);%}TZDYYs=eI5g
zcw)(DwPJTI=ICU%?JZgdw_~!=!gB4wbaplsAb^i5j`CfOnkx3Sx2zz^1ohQr^3A9D
zqW$50esSqjJxNc$L4!2DdwCt8UQ$aqo?t2RxNr5aNwTvTWgkWr0dR`7=CfIXD69M%
zh>X|E@vWua?o*Yo=B)?Yizi;lj9(@N?+%U#&<ofM;Xxn+jCO+~wKv`SbgB7N!zfw!
zlGcUFnVr=lO~>4Z+slz7*tVE<jkxs$v}Os_wyUoW=ii;NBObNeYPJEvotl{=bJJsP
zR7RR5;!f2F7j!GMkO?Cbyv>xeFr93yehMgoX6FfL!*Up-lbeOs2>8$mwX9#LHllZ%
z&5{5(?qF|lLdvItv>h8sXPdb06j4e=ZxNz4sK!&h7WU)}-#AI%=w`QZJ+Cel$WLgr
zz3<mQK0(9kJsM<I=-b624XfNQ>HoU(7xFLidyDKFv`OW~<!L&SA9KFip9~3)4jk#U
zF2Tic)}x$E!S~<Px%jSgWh8&M6Y)g6>wdUz$o18ZoqL$4YpAOwaV?tFJ{j(g=oFJY
z(f(_LI1|!DbByb`b3A&kfzSQ>#jnpL@Wwv<4d(Ut$?U&9xn&j}#Wvo!({LE)L6~>n
z!P^~QHAo*5btKh8648`UI%^Y^LNDSZ{<%?V=0!4{5p!3{p01Y?K<ESoC763^Q@o`^
z&6qIXVztu&oWRhnjykYRZYuza3-t4NnL$4Qw;eU?ve#RyO3ilxsuc*aos(pHx`ceO
zNO%B<NNZLDaH}Yk<+rloF-29ccinhLf>eHIa5(|U$o6T|!R$>HCXJ4J`>Lu^1B9$n
zy0g%-*$rU${8$_)QVK)wW|V*ZFMZuZ1`_<TBRXnim+Y(p63<KW060A(cVgivfYSok
zufl9#etj(*7ZguTI>s`@Thbriw{@gY9=?+JrC>->oG#nR?1As*o5gtL3f}}Olidb9
zwkVg>xM3huyWObl3n^v?=kdrQi{L>(kK-Lw@1yz_ZUE-tHE|xwptG`#Zn0{=BjTgC
zS0_SsV2JG(K4}D`6O8H7eBHP93dakJ3}sT=g6hCoq^?Tr=JZR%-H(0uKYskD=qdl&
z(3Mb6fF{KnE6M^mjq?kx5q!WnpaNvB&1`M`u{T87$Civ^cmvCvf#cWz8`HM;m2|%@
zxyoFzoI+(3qdvLHSuq<nzZzp7DS8=Vaxzh+)>ZO$BnMtR)i!s}rKUaoQAAXMzz04T
zJz)$8z@K_6JQ|NcT(MXcU6ypWStL;CqMi2`pApo<4#ERqW=ZE?$ydO7{hDqXFvlh)
zI+*Gsh)CV}=kDpjF9bfJ@$eI>SwMFe5#S$#6|2Pz+FY*6LuKTl_=&VNjMZAe#-LwD
zJwv)lV?T^5VQ`y*_;*KW?k=^A#5K7*DJkI!ATE2+D~~aReZ0%}^vyvNo1L0JY}+<J
zw4XnO5J3MJXzNV?_Cd5PUya+&T3JVNv#QLi`9^Oy`Y+?NdU@Y0G9pq9#KR^hvUQK{
zs`YZ8w3wVbv4VJhrQ)Q&qAP#utMLf44c1}N_rJa!;+WP<aO#!KT*vUqZj1Z(E>#cT
z6P}6+XosKGE_jg>#wR2ZqvEVJ9@aOPJ7?N(!7;vkBUiH#s>8#dqg5F-(eZ~?im2a2
zv4m;FamzhyR~k&>2o36D4K6nK-)Ow{Qfj3xAv#p%vcAKEpVmIUIePbT^_86O6o1{;
z{cj%LzkF-^*DT9qYV}BD+L{-}Z-AaP5Ap+w35<ImBWm0(=wXK<6BL_qd!7ze9NRjN
zPj+!>8jrD=&dhOhX2Yro5Ju5O_T|iw&|i^H#x@2yde222Wu<X1(_8>bOmVBHl?}O$
zuG_gJplkN<@!WY{mE{nhdpeGYFh_UzZ=Pt%a3XO`@@tao602P2ECh2U=zOCO+j49s
zYGot~1W>mD4s6u*9jdE{nWrW^xIdnG2&W?p`DwBq`|E$1c2jSO7k@<jLTTPUtKDs`
zm47tW+bnw+*j%hRk}vHzUk2c}xVxT<-A?N{w1RWv+>T6^9!be?HLd>fUU)9^W#$O{
zh6I<WduPPV|CV&+fU{*Hz4i#i|IW98RV`5H%czO^>&vZCJ>I>c7@;5Z`XGF2rZ}l!
z&hc@bgZX}I<<8}Jeqw2BRxQOM2_H<Yg5KWjPcgGJNwEtFNue0dFNHC_aq!;TtZQM_
zG|;&OH@p8^+Xt9EdEImDugX>-iODB%&02t1!RN_0&-mevEo~a1J5)`2CewQR#@B%K
zNG?)%ue4xNc60p`Gn@s~Nomc*5pZ=?mpQSa3e_zo#+A<9fW=cr=A~t`Q07D%Q1N(o
zF76aYXm^?7d=8ioQNrMq58GI=>h(4hZ(OIuo~r~C*YY)PN>l_h{66pg0O-b%O(mg-
z#r91$pArFK{aays5(PSw6ncG-t6-x6-o@Gi_8zHtY;y_I!6yNa*hUF(d#qLbT#$S5
z=l^;hzjub-Y0M>ib6Czj@?_}oo0{47BMJ+2WP46Cu&+KML{k>7B-@9s__x<YqC`@6
zYUk1>urZfA_7TZRy^u>4C$P?{bH%6QL^5T{mWDB~o*m^a)&~3l+;9ZkX$Y)m_wqe}
z^^6kx3~@~e27}FVY{94=IU3;R1MM1~SLoo(#>2Bjq8cfncmat@5DEFREQ0PbtBNUg
z0IGn6I>54)cgFE(zc|?jnEZW{Y=Bfow0!|Z`t{TMyMFpT%l`KM@zP~4-$dDg=jtlx
z<TRzjH<1G8-X8R21z=`3!E_Ztrje@9wNNBGZD1jEF)R1iP?20E9Ga&B%MMDUT!+hY
ztIE`C=afRDHV>KnQ40W|xcq0_w;zkqF6rJ;vO5!sZV$EK@(G1eR%IBj>@^;nuLK|X
z2yiz+BW@_DUD&#UEtF&(@tVz<N=(uL_^CSp@?dBp1R45^zdHtMl1>qc=Brd<4@O1i
zgTJ->fAsbL==CgC+@yRqySLyL_^8CCUXz%b*6R?KTAfD^G39>Q(!#W$?Lxb^)WP)2
zASIHChPmgESz^bIOheLOaNROK?94ScN8cL{Lo+d1(7-J}+w>%58UJYCOvtcUy>YcF
zAga*$@yT&#-74QVGx3WI@g+p-TI%I4R5dH4&WM)HG>MG7mH4<0)W<cBE&AO5w|?L5
zQ1(TcU5R+sWWej?ic>TEv0W4?bh&1Y{h2p8%Xj=p{#7FgI~ux=o*dxf81qw<<i$P6
zmS6ZJyyPAX?(#)L-;gS`ow(SECu|op_e<`iB`9zSdI{apW(T5EN$|$7_PD)GV<x9;
z6a{o=B8f)w2@42pv?*S+zLs_8liZf3*VirbY<HUm4h$N?;^7x<(f{^%IQX69XZr(D
z9TG}<B0sym2!Z*FjOCwe0oV6%t!KA|kam~O<)40*8C<jK>$KCEfUSlMRea8Ch<=+>
zYX5-WI_|ynXnjx@x*p#u#}QNO`xDb9L(_)`!#4{ArPexbepbCn=iERd`kI!+v1p_>
z+9~xJi_vUTcML8%Ff<j@7*(>@9@W_eIHL))YdGD-6?b^f@J;M9#p%_B$y_8pCGO(=
z`%SbOt-&J>Km6UyKSU|Z$X)bKmEOy})tHAxbgw0Y`o>j8c&YPn-jHq0RCrQAMnL%f
zBA??y#w$nq%=}*9KX)Q(SV0D^@b78XJP4%<C;b;Hl7M`cFo<VKgn`W$s!V#MrMtYf
z+!Ka7^QNtDP&K}B937)}MNgqChinTfU81g;+?3AtT6Ta^#f{z4858(ikszi>q2+$9
zP~<b1*f&&KNltD=O){uI<|=;4%IHZpxml{_+f$YBTgwL<dp|BAUuQ2sHWQQipvbm;
zW@ctM<9Gwlv5(C4jzA$YpRX`!QSGq;ZLAXBJ6*zo*zO71NMYD)a}`$Tv`?1>s~Vlg
zn2qk8f--3cfU#iL%|SpL1N&{^cRvGBVpGy_YnAwpeOSf@CTy+mFPH@-;Zf68DWwJ=
z-Urr<Yx@;9VoBtm<rXbmpjO{)sfdMv&Y(zsWH3oe-@GiGZBQjd(y`K#vMiu)!smdE
z(c`|eX;&Wqz3+ebWM@_q++9AjGSK9(5CHKe)_lITnq{|MRRN{}avF4X)7=ap1L6pd
zzhmk!rkc{`0B&s_p_pOjav8IWF^1zbxq58nP@73dn#Tl?vAe-FNUw@78$N97uXaW~
zD}#bp-A({nKxgB3qm1%(93B-E0qGJij6d>PlZTW8Q>mT6AMh#icnh~Up!+-!?XgA4
zSWBh=%{}6r@M#d}yVcUg(fZ;sPhHPq^9^w4O47~&;!PJ_@TL`UdG^D$3ndOubdQD8
zCE1J0DX!tz#C?%}(Yi?yMoxE2Og`ly6~p?$DrhfAd7xjoPOn4BX@>b!FpF<%MA$l1
zL=`1t(qUgE$NuSeJLs<;?C>W&*91d7<!sEn5<i}l0R{r8cufScgR<~()u{o>%7fG}
z$l9`B1SnO~LY{*P44qMU?84He<PEOk{etKy)G_o2iV!vBtpe*3!HV?LR$rX`87ND`
zI2hNdcJ(Ax9O~F2C1-MtJ%CkJdGy8PI0^%BGkWANZEmIv*r2a}TzdF~EBD3A>)k}&
zhiz)ym<U$@pEHkAtu?Q+R6~kUgnV-agqyWs0>RW}r>C#8{)Jut-pBtZFFv|+BQ7Tz
z-A=#B^SPl>`BCyF@2j+ukRo$H;pi+s6DB^R>yUCiXk^(~Mh!+^X!rc-UER@%a!4iV
z#$?Q49dQ1Eq|>tO>vQd&dv_s9#xN@(vWPN!=2~`R@0dXmFRQXxRJ<os5^v87A1Ltp
z>m0Fd8$@#gF2WAn#x}H)^J7wm7Z&4hw(L3)^0m}5=}5xLK|Xr4S(Xe(G_?|}ySW>n
zb8QvrBh~{~%boiA))27h8dO%}uS+Lie^&XYfA5h1-x^Hn^~jCfjOht&gJjK6d%Lg>
zD{?0g=YX5T<6CG5X8k0ys^W0pN$K~UVKQO-{QSQ71$hDbP3kmz=Zxn*7a4Tj;!bzQ
z9K^5+y%W$oHQyf|t=%4wgw%)hU_!O9wv9r-k)ks&8dU3QhkI?HO@Ipx8Q)SgJyOxJ
z`JkH`v{m{eOa^ktJ79KVJKubo^Zh^AGGF%TTygdhKIrhgz~}j;sNNDhwIstSJRn8P
z3WLPy1?7Ex23mHLriH8;`TPR&b2;quE|9a?Xe1unAx~z>y}}ve?P=~yGl3$E&N#ub
zYef{ZQYiP*69y!+?}duVWe)VJt~E6dD%<c;?Q2NZTvI%k%H?~&)p`<?7{91>B{*%B
z*j4WJIPpr&aG7rt+%pG%B**E%rOS^m-EesPZISNjpt5U&r|Z*b0E<&tQX&-CYSa(O
zufV&!9p;Swrk@H1^w$Gn>#nCz2h{F2c@#9A>!B_@koN!(1KpZ=`A2(qzv>8!`moJF
z_Efr;J+dGz$$0Pzg+gH#5c|F#NTdk1LV=R2-^!qx_x+kpXOZ_SzZ&GzTWTZW02(h$
zVhVTHQ?T`@(t`DkNF2}sV^Ag~9f$-@6QeQ0ry>)^dmz?`F$?|@h&95K*}VW%YhU2I
zT>#_TGP)vbUV-t>XQtK65zOS7>dN|;ELP|nZMqH!qrM1g!7-)&b&vC@I0^W-VB_(J
zZF{W6_-S)hQRu9h>VU2hbTh?!D<l47-J?JL!_)hn(@%~@x>C^(`?db^U|kM!nUQqF
zmk$mOulfiSNMPPdOlIciFQHJ@K8s1v<i#0y3>`;W)v{vZsC1Uias(NO8un2J)zPWo
z#}rZqz}Vf^h*cTDnJQ|JqZE*3-2Rl_NO&k??@BRf-au_Vq=-%J2UQ-VIadZ`1Tc7@
zj3B24f=05C4@I%lYci3P)i&IWR#8H)y%_<5DOMS{qf~WwPAZS-^9C;8J_rM-=`dDM
zm2XKLIREbNnDzI!y^<#Lnl{#vcZ7h}R54NB`%G_sgKT}@Hm@ZgCrj}FZc)CqyH@#7
zgGBiX1gL1MlHT)8im8H4?WuB@m@5N`fw_NEqz!enPooHvtVkHH2n}o^uN?bfJbCvI
z+qP+stj+-Yre9?PaG!N~dl<9*B6Uk+@71vgy?yc<F^4g4UwqhhuoWa)9T9EEQgJ$&
z^2f#0Rl#i)3`Pg|AnYE9wsgB{68Dfw`Bikm(v~)cyXHIkacC5cD0BV=pz`mA{%@W9
zM{9S0E_uC(kyMIQ=71ZsCf;>OU?^5mw(|8=0XfeFnAR!P)V{3={|=w?zySGQk3Y1P
zjqvXygy%X=OSK{AUSI8viH}4Q<O*tjoVR()0DQ6zwm&NVwPtV0=;HS;%R#zMMRfo&
z@|ssg@;!iw17e2Mm1(JddIkK(eml0Sy3F$cHn)$wZUC3tz{yNdg46yqMyB3I`Nr%<
zXT3*bJ4Yj)V>pZJ)ofnKug>GC_7T7VsIHYYE_<%AIOH$)yGNNVHT{cjyE1CB6lVs)
zr*vR?mPfqCq(>C~kE`ljBSb&A-}+xI`G5QW?u<=}QB(;pJAgLBDB=js2Q*ce7du4a
zBGS-V_@F-%?|@dJo?pSA7GanA!ip<y?{V<MoL}%g9Cmp1{f16Ub%xf3`_v3b+DJE6
z>jL)K<)-u=6rlRXU+mjyo-@%F>1kPsi0H^(aq-185GB!*A)y<u>G|r<`6d6*M$1?4
z{KoqFJ4?*!{m<O=50&Av8@*4KfnvJI-X7g-ue-yWTfBSD;xk|{Q@lCpC(n42k+HFp
zBq|6o+^0F+4oWp5tc2>|tVsk9-OgpIj)voG^b~kc55EN$7in?Uq(cQ-DrzNwFln@^
z!Mua;k&ie346fsl_Z2nag=8r6=)4JBrAx`@Z&q%){9sdc)ZvYiVwZ2x1%mpX%mlX~
z&P(E!l!ABc;hIhHY39o0h@87ZeipBtNR?Knlff4rU>QH?G<nX;3n$8@OpZ`>;4T*l
zUYoq}Revq<SHMqCNit-vRuu`|&FBiGI&+uIMYTJYZ8Oz(GED}$lD8j_b?Ce+Yz9MS
zo~%j%SAB^6+-?NCTmr()162vaOU@pt88X^*y-K0G9kUw81F|cF>fc_UHU5Ww{Jqo9
zoV81VnRddxi<4}gXG4yjFNv>`1j|gMV8p`Ohix;`05_V_U{WEhNGV*|6iX={FGD`g
zvvfg8qz?|<4a{J~%#>d2eB9o6pv=~|U=B#}1kk28fX#%xd;Qe2fjhrRny7xYg-j)p
zT59NX*@^S6@%cCjNHTS0FcH(qU#t2ECr<)YlSY-#6DfL}a#N@dNU^~7m=w!Ul3?EH
zEnq%hH^d4P+|AOq%76)40Nr9*_5+h{>dOhtX5g=`k%WIa((BBKJHmG1_66AD=Re=V
zO_|!Yi{+{UdT+<n4ZiAgK<f-hV+HS8dPG|z>nj``WD06|GH#zpkraPs9EeF@IRMF+
z8!6vZue_f;0H|gxO{xQveijeNdXj%0gasJ^mc}(cRmT`g(<<{-DbO90LmmEI0I@i9
z&)~~sBp5utwvO7Y750Gkb;!31$^iP(g!TB*X&o?>w&2GC>uyC?EaYz{cRvqO@`K+2
z)0Fz7F@WjY|H$nIuitxs|6kt!<R~UQW{Dqf4RE}D*tP@o^^T-pfyFTydZT3{l;gH_
zc(F>(eDS9@nZMm+XZgg3ZJv;Q(+(}-U1?7M-^49}zLJ{1Ib^t(x(>q3MFTv}L6<6j
zoc!VMjrydm)I&AO0{2-0GK&E>0!cwC3=_x_&rt^7UiC*xH8E2_6jyBAl5wp1zrOYN
zuLoBzr^gkE3mx27wUemx{I=0VY?r)}=(%4XniLk=U($v|4A@s*estq-zX{{4Yrn`{
zM^ZwLs_^v5aejF3+LbET<3WGB+8=l+z0@#*jkp6X9Xrt+5PpOJYjsC3p?zNX#fqo_
zhokvb$uRf6yP|cVl!s~`{|##n4(an*sfHIUm<HkXOFJY_Dv$l`a{s%pyR%VJXmz!i
zU>QDk<MeTXAMur3dC`L}xaVrcEu?w?`lV$SXW%yhw+jO%h+EFBb8`q53;eV53!Jcl
z^&vvP7o0m~$$qyUSF<cym&gX!Me8D&`90dZ(#X(!MhlnFFMNjs6Vw=fF)Eh_SIb}j
zLlcj@zm<A~>o=Qnlvy`-&%R@y`uoH1YJKc;A5-(hN`pR<ZOCTFtcmY|GyB&zXZL|s
zm3gJWZ&KZer63rlhK!?hrpBf9zObGtHMV(seRc@=9PQM(*C`JQgMc4FHzLI_qZ6|q
z?GTq|d@f~wMB;e0xo#(~VSdNA>wf~8vAqP@WQ@IsOD8x^CnkfB3mlAo0-T)uBWUYq
zfl8>7wvHJ?d?C5H1pLzP2;J*ZVkrnh79=Qzl#ZCGGO$8Gm%cG$9#3+-#9RKbt%S)j
z19Y$sRAE1Cd;H|QqZ%+mw?`m@t&gs3rnSf@-V**1QOem7fXK1G-HDD|TvOsa2FQR8
z%i@}9A(h>%3z>PF{56m&=Mv+Vk>DDk--Z48RH=&32j~b4Ku6XR=9Apb*rKgCkJTjm
zLm#%y*Wgc{%KwKB{k_vqjtxn<<cq?6(e<DJY_7e}I2x@le{=W5W<0Xx?orjg;~48D
z6j~J}jx=5sHfmD#sZt|RQYUsfxiCUeF5t!7t~+3cn^iJb={OF|vyHpy-%aF*xbu>L
zcwOw+(!5m-SLLD0ioKLD`UU{tZIuo3oUF@qwsONxv!Z$=)Ss5-^S}YVv9iXyYnS`J
z`min6(q)nSy|VDhY4D9Y=;Q?cFi1bK`6jlKU!EN5@H(X!wyczT1p>U51P^n0<pJ+m
zrXi(#?cj%Pv^c~fCUF|MKUbABK+W_gjQ^h5e|w*RZ0ZdafixwS9UUr<C2w^W5PbA+
zylW8vV~PcA7QF2bop8a;qrYE*iIoQyu!zpB6xM~h2pO=FVZf|w?E7|#mwODQB~opq
z?AFQ%u(N{82<kZs+#qgWwRZ#8-kag~NZeIYpz6HUyX>6n_Jw|xNtK=;mg_K#eWZW4
zVI*_J4bWl#db95&1njfw&Ma5eEHWBV6YEUUJN!2eG`_X>ym>E-T3Fb``lf#WMxkg!
z?lLV<-}KKX1Qm)7@@<L%PUxig4L%)5nB~&nF=8u=4Sx8WiT{(Y|1r+s-x_IO2srG^
zZrm*x6d1ugHDiVcUy0c?b=e~6Uj+(7qFR*Ax|<E{2@UB5|4*wvGXR}($d4aPF$Nti
z#!rh!?<(8)d%SfW#EHD(9n_>D2-fmdA1LbXi|n0VeVd+l0g6wQXI|Ko6MA1|efPKP
z9XPrqkM?ICOn5CzSz_eUPHbjJ=4Kh6@gbcoV-{L!mp?Flh8sG0_pKsp5=Al>i;H|<
zCme6(zp}ddw@du*zHZOwXHeAKx>@>;-J>@Unh<pUNuI4*-=l9cA3cE$HpgjIBX(Uw
zC4zKHim|O#Cg8UVN!7oc`MU2>P(A|rV;x(YJqsDu36F7#cuCT`P#XcAS)s7(ypqds
zdC{ANxMKJ%%qcNuh(bOEMCz{G#<9KDYu>+T*LY7Y*SUL}D)m<HR~b1+UzNz@Nr6^d
z=Pt~0r`8J!E`0s{u|NKoG22ZY_lcgac%SF@EVh<bz3=Z*JMi5T&(yXvK<%x5MlUR^
z&nKjXt1KUbUVn4kvqE?@G&CD<elAx;`6ji`T+ol2KWHjTlYPBz&Dq=D(I?!m_)MrU
z&(*i5UuH7Krlu#GT>ISndt71MBy9vHf=_SsPYIEUn0xoaz10i+RE3L7_(Ed_M29{f
zsjHi=x-^GG17r=@rQ2~1L&-+>#Y+_i^7^o1sQ0&Z`NL)!kwwzFGQ(?3EA|sJtAKvz
z$B~ETceQIdrKnV44$r?IwqUKfVKC(mYmB0|%snw@!?H>>2+>7#-#lXFeGWQG{wz6K
zme|ZU8MsAhG`;sDgACdneX#^!5hjO0@Z1+7)>O1VQXg(3C^nTCi}n|Z?syi=G@)GF
z3l}L@R-&*_Xn4X<fM~zN1-*!b#yQGD)(I8tvvt=^3G+~S=!6g1E#hR}Icq;)R@|mB
ztGFQ1-N}h+^+Tc1H9hDH82-F6zrE9s11vG!^J~Z|I#9wmDA7En%(n-FXYKFy^6AqG
z!fl4iEdFQ)k_t>bD3yGdT#B7o)ET3QJD}>hs@+zACe6Sbt5-#XKWy7(a&)N<svwU@
zyww-iWJ46LPzl07=UC?%%gO+#f^vKlfWy>moj9)~OtOR*laGH~a7MH69~T^d*dt(T
z_QSRqz-iJ;(QHUqD!rpFl)0iDFoqNYR0RQY?Q45dMt1+#fcd?%uV`#Dt$>o-IcRb*
ztQdRUc2L_A?pxyBrvX3G?5lFl49SDLpJ8)7j{Y#Rc)8|ViL|xGPSb!5gUoDs)i|?i
zO@m7S$!s=&!Knk0y<JMj-(cO8;K}q5!mZ_(NeAiu;jC|9*1m$2<w%!eydZq8v#r?h
z5FnJxUPEQfR}B<@KjU~~vGh#zqNnxz5KM~}n#FB+8+PYgj|X)UX=Z_Ch=GAgI&pmR
znM?{lA39Bky(l5YkMBkCQf^Z^eFW;${R2tDV*!F@^;qH33-2Tr{n12SR2-9e@E31V
z1q@>7*wGS;c<&MOWr{rfAPn4=AB=Oe^U`%A*}W&138FPN5h>VFJK^Qd{-qmK0`8}7
zW^1Xtu^P)C9p8@)P#heqyVUJkXq>@5otZCD@b-vHD%!~KJRUOjrF0%;!g7)sg^^B9
z#fKX7)-xpciJaUVxLU!37A$C7%&|%KbNG4Nwof0fgL}Q3$vU^!3Ut1vhfhRWjyGQf
zlt&K<U(%Y9S7{N4qGZ#KeSu{cbiMbr-iW?he>gKabL>{TWmYLF0&hsEC(C><pE<F0
z4y~JLAu>mRE>NPT$@$@28_KKc3rImno&R3g*{sSThQ!1;d<noTv&Sk|R{CChV$#Ry
z*>M)$P#BcQx+-<uFl=1E*?%XEj?c%^>+S6wLRLsI@DRF<h|_$n|57h8O}{>n>R}US
z<dg1&bWw!2u+J->l$FP&)E#i*ZNkRL0k7x?m{ThEam7y@1XkZS?AS>Dj(+o&7f=3D
z6YDv4Sxu7#_9r$M;ddwJb(Xak9iZO@Dha`jU<U3|7-c;1Q)&&#*sCqhEcebxb4v_o
zBTS9N|7nq89crlOdRpW&E|_wt^s9F*w%0;xr;?40?BeT^q4n<CARubHH{|&(<Dn)h
za4Q6z%lYP{P!V=GWMo4U$?bIBO!ovOC5K?DuT}+=+Kq-^@&f8|Yn<^$^VRNnyPU4D
z<DO;L`h4a-Dl!=6ndhNkUsPn?P_rhuz=+j}llc{u`4t>^xNoC*_+;aLpeWbCz~Gxe
zSdY@X*}!DKIdA~no;(9X!GQz1nXV}ZWF8!)F81S`@%zW8jtO^VqZ{BSm=j@#PFaYd
z*K7YT_TB`lsVrR^uIlR2s#2jWL{t!35h!MnS*B`KDCQEBAps(UIY0zt2nkSK4u}jX
zL8dUI5g{UkFePCM6her|BvV2X21yu%5FiNz!c_mU?(M$a-|c(<yY8^QwZ3)tS|s}%
z*2>!ZJ?Fe<@ArM5=Xp7(Vk!SlXe2xgUdL6eYcI@GdJW+9R}Iz6&c%%PTC<Itt#e`E
zLPIa$@#@TuL>cTK-R%>|ZYw5Z>faq04JQ!yhW#h{t)3BiDM4N<?Ux{sm^xooZ4iHd
zT1sXXZKN43XnYcj58E5AP71ErpEm6??=<~hpx~Zoe>j3b8}65p4Lh_=_En?yVj>bS
zEW-{TiokA0l0b?R^+*-RX1PypB;vki1-%TeYd5A%wnUw)sI}>Jz?e*B1*;&z7n=B{
zU>FS3aXSXlI@u~0bGxm_FG}WirfJOB27d$kfzTaTp0~UYxiGNp%{ptsKNK+Iwc7?3
zCnk%%jEMJQFOPvAa}0zFRLH=)ceIwA6QQZ`DRJs+1)0+pPW<y6%k{pcZaj6&FvgG5
z6otZr7I3Y4^ZnVP9B8mMEOd8a(4-n}L`^+4@LDf>5dP8p_B1E+AmrHS%Fe1Kp^u?Z
zMJ2_I;J|5!ybjxY^l!cF&IV+X>hlOmw&9~m6W>l`R*;A#ja}AjFLWAAv~cCqF47h;
zJ7;!~TWm%8w)VK--r;m9nP`T!rB!!DGxga6y}n=vMaoVjayBJ+s2|XNm7MK{<lx_g
zAt{3BR$TS@{;DB!(0*o%&efcg=E2(GMTvs%cagC|mY@;GG2ep`Fa=3tYkjF|=wl^2
zi93wgYaQSCH9e82Cz#dcH+^7_&+$c?ur-q7gTM#D`5E!*Ri02|gjz;Wrh<h-xX!{E
zx0Y&2L6u*GLaw(g%3=RVB5h+mVl73nX7^9awZ|YT$WEWJ!FA5o^D)^{W*1d#v~$ee
z95u*Vn0>#`mbKFK2-ckr)k-w3E(vb6(KhXqC>6Kg2!1l%50O7><ru|IR!7GSf1nz6
zlQ)|VSfP+Y{d4zuum0_u`#=4k|JlW>sWBP>_Tz#Opacgk!>N|ERa{G!J8QRnoD0<+
zXrj=%=C%B^R3A>bAl5IqO`Hwb>~&G6+{zDHo`Q#wN+c&TSB*w$6FHgYm>teYy$L?0
zP_#u-MWJ5pB{f1$8i`~!1s_7Ay*8CE`BiG#c-62M-9=pXnt%h##&%#Wlg4(Ezn)JS
zw(A7qZxL3%e)8jBvX$$vpA4kdrYuj*?)0VtXv&gx=IQlAn8%?i<WMR((dpMufDlP2
zfF(Wx&_?gSNnAQC0rK4$At`{8?bFC#KUpNEZ~)Tq_e)DJW&w2rBkZrA{1Viug#Ei8
zMgQ+_{;xg%_g?yc={Z^r-lKspL`_G&XY|{6F2KY&b%qTwc3Pu$3J;yV*uj*HY!?f=
zg*;>ZTNw?*t}jNr-<~WDd982rix4z1Iv>BeY(zASk0VyY2Amd~_2;QxO+D_7p---}
z{E2B;OU;v;x0|mxdz(ZLqfPaoeU?_z#xFpqq~PVnWUSY2a$#)5X!MW&?C8I|uo2fp
zp$H%wQ{{f<<GjA<wT9_&j3=Vk$X$U_H5(8B#Qk&Jlp<oCLk@cyZLI1}d`wgP=bEcL
zE1Or+OCH)k$=tWOg1FpX@!VGs$4;`&FQHhBWf6D_cjHaZl|tN7u;V08Pfv%l<zLZ3
z7PLw`Ww`l-K3{wz5sS{<6pg#s#mDg?4$CL+#;wUifV!9n^5@Jg!LslSF?5NCr^T3i
z%)|xL7K2yMv=8)xsQyQL;Gab#!yc`A)(`m2-y_f>eCs9k<T6D6BZ=B3#Hg0mP?~Zs
zlnuxWNnUkd{(kCjXLz_gUS(YYEI(a7t~`O|r)ee5!gP7?%NJG7z>2eK*g3=+fb(9)
zc$ERmPvxW3W$#>yE@&pdW^}XtMm3-#r|kfw^kI7XGG3z7fH+t|{;Fe7PtSnk$42*u
zeq4S&=sElbEA?pakLSrNxHqlF1|H3P>0?Y}x@i=f6hu_KH6a+?4Sju%e2Gz!gv>32
zIjmIq8N+;jdNBSH%+>Sq2^t$}YF^q&+cUB_vz59&eg8R>wycJnl(a0tejK6Y;Dmeb
zB19>q8rsou!n}~M+41WqpXO)w$@S&KV_b3vx#tNU4qM%gTZ9~3KWG~L7)N@FhHKGO
z@ZayIRJ9ztv7X(8?v`4J>7l~XPdg<Z_%B)L3N%)9QFrgH?PzANE=_&#vKCqNfHSnW
zN^dh%9@JqtFuX5i8WH2U=W^RL(&&T@zz)h#AFpzLAVEYz4ZpnW!6U!!#9K*CJ*?YJ
zozS=aJjdz{w~cVr6cLGi>&Sx2-WBd_a7#`U&uE)8&%huV3)&#I$~~N1Mqb`|hjA&w
z^kr&uq;)3}Zz0*CG}m^j5Kf*nDuS`-8I#M47nxh|dV-84vwuzGx&GWt^M_A9#WI>E
z{5F!66U@*q`EM!VTp&YcrK{`F#%fWMTM->#b#sy|Q}#L_bxftzGA2+9cWnQWu4VGl
z3ND;3*qe=)o*?a~9NqN5u9wO6s92P2p8(<kN6)A17{vpLAt4cQR?qVnK>YGPrD!Og
z<9lLlNtwQ=Xg6lk{bffi*Ka&_fWBSFHAJa;#1!S201R=A&_Ln#=>nFi<0Fl3(<ndd
zm6cTLAk>vU6P6zv@<Vl|3sHG&XlX_I&q=-|%GEIx^3S}E%BXgRdKsg6x)ze+@>5n7
zn`k64TMqRfB<8q*;ZQWdaC=%S=aG?SzLwjG=zxliP3c0eN%Cx`IpIPT&2RdB<I4N~
zHmo8^p)KZ3REysGjFt%RsO+68%s3veD0xBefE%X0)1WW1#H<<N(Bi}Cis`RB+{y05
zPzI@mM7Tr@@dnc?3^uEOz&nSAxmG(P=BqA^0+I^ZAqZYX{%P;C(RYT?stS|QkdMH|
za0H;G7h35E;2SUQ2%+sxaEI$4QW~JSpemC{!t=nR6BKRFe?m1jApi;`5wJ4@nx3Ae
z-@M7}cYpgl{)7H!T*T3zo08Q>Nxru8c*5r|LNLi1T;s+Fq^k9-+{4#g?J{XZ3<42j
z&@{j0WO4QMX5iM8&$N44Xs8Wwf_e_9<%Z$oq`1RTK9E-%`>A(U&1Yt`rPNY6TMZHG
z>->d6v47Wn*&w>WljPW<s)?357chtcZcPerd}d~h8*5}N`N$}i*v!v|218Y%{HR%-
znFOTc^(52{C=Uh#Tr=^xYKE`6;yXDF@zuTHIZJGye(^QNhr!rKZ-pv^o;g!UUJYuo
zEY`hbzHzLB%hyzqHUaz9*Av-ScGgW+TQudpOT^{Db@11^%$6~VE`wvXj&DV!#BHhd
zp66RO7A?Rx*9z-5`yS<Y8In2!mx#eVi!-c@cB8!U1P(DyKa3|bog02RLyq=75GQ2e
zbByJ>B)|<YvF$(KVxI&<r1wxa?+KG-yV2~%d?f}{_3C^FUeo8yZadci#Z@W|Ef@9b
zp&1#_J<D)h&vo0!f|DmX7p=YkGS5J+Wa;4aKWkzC?ryu@eq_t`bEdbng7EelvS_zW
zYg0?d;CQVKn9`V~H|C<%IG<2t_6)z$SO&e6w0X}QKi^tv{f&DWy2$+v#5=eGh`CWy
ze^J9JeT_k(Lspti*8a3I&rt2v>BCi<Bt(uif(ClY(cCS)o#7n+*~0VbF2_Yz#|-)!
z6*@>AmRI4@xKe^B$yoKgl(nQKt7TC=LaW7s4LY@ZeV!W=fE{>Z%UJW$KG$OI6g5O#
z4<AWUY!_>$BHgIxh7C8|m$PO7kX&e)d3ZydZ>wCv)Wh~mk8B~6dLG}5y}1*o<>a+c
z{;WIT!w(b^ChT_7l$d;)$b+S=1Xo(jj!G@`-?RFmS-Y0POh7$7yQRMueTPNwUn}u-
z$ef}Y(dQEeS?<wc@UVcs`&u`a*J`~H^`o#2?`>J8$zU?%5u4Uh-smNs$6;m_?RO*v
z`kba48%tDuuBz5kFk{(Lm4wF?-IDqD@-LrC=y>~?T+r6aSef-hQ=U5>ga{oiw=b?Q
ztpE|#G8)_f%w$k0VhQ6FnCCfY;P-`%?kiV`E<l3}Xp({ehF&k?I?~a-sRAYg=*AR0
zp<fIN^1rAWD1zPaj|hGIIIXY{sNYcXDsnskq+<pK4wXb>F;p(=(>F6St3I#~s&kfB
zXG1O4<enn6f2!8)U8wHo5jz+O;0wCgU$`NI>)%cbBgnaj*6qrcX`tTTnX%A;VEOli
z)oFScfx25IIN!y9!7T2=x6t-2=Nx&<Tr%0HQ*G7|N1HJsC=L`PmU``VkW9s51%&uY
z5Xay-?v*7ufGFBJ%pn@kygJSwc0Ee5QYjyV-%sllpW$t#M;jhAh<BHr)Is43FJl)L
zrVkb-4h|-$TL)WIWIvnzb}JIDv$hsC=)1VGve-UGC&c(Ri6#iXHKK`XcLc69v=XD2
z6;^kr3x8jVL{b3l?(S6GKXUw3rPi6t{x6`^R4J_}?*sv~0tT27H2MsX6{2f?oz9>%
zfEW}!a)u5_@i^hL>4inUfVKn?uWwZNTf^#}ACUa!)D&+J2M0?AkM&>OJ`HHsoT`n`
z;lCOuw~$)MRoWoA(_70<+91OZ6M6Q}UX7K4p1pF=5;}cZ_C<H9j+U1opbM+0#gf$j
z=5W0W8UbF#T1SK)OSMoh;023ZQEf;GngC*8etTe09=<xXqUvK<4ER%<K3&5A79>3b
zPS9a&O+o><=(Q5-lL2T$;xMR9K)$B-r?wq)ImI{+GhZtWKOL?CiGD^g<jr`|SWdUc
zgHSL?czd8>zTWto*CbI?HqB=yd`C+g6?Rb6YIe1l@TybFTX!0vAcOsRbtX4z9Nh}{
zdtBURj=nxogBYKoN+dC{-fmah3W7ozf)&1BSC8<@{n2%BG`Eky*sEm>bfpBHm+Hc#
zM0BTAgr?^A>F))}+dg6Z+=Lo%-(3T%nn9qS_iL|x>+?f-EDV%6NF3CN<*#Y4^no5y
zzZ!LL%?vZF>IKvmZsQDc#;Puj*}2+PV7$t;{LZ{*sG1i<10<Q}1%Xr~v_D-X!m&W;
zItJ$U0q>{$is!H}*2Re?N5|{0-RUDCz<Y@pancYt1_5zIe}9ytW0F5~Sr6blS=~wc
z2k)i-;;H}HBItiKxGSxKH14@6J5KDhHV9uM&JIL0-kC>&V*1WrDjm*$=@pH~gA`NT
z&dI7r{*Y;9J8t)d&9e(yKzhpGDt5fYa$<;s8T1FRhR;>alOtC>B2}$y(d@%gOKjQQ
z<^in!HTT)-Q&23Is-Kmz@Z9iJ+h6g;YEb6OxxK6a0d6_S(#TLpaP`PwyXZ(92NYH1
zAP4U?hc>V79OGfXwVfWTZUe0}zco}|_!54_z0<1r&WXqWXfpqybGFh?ww-b7(%&P3
zQ}2>)=}44hB4B;ye>4oS<K<`HE7T!G74#<(mA{Ich}$<riPJ2m0p|3fM`NJ1gx9Cc
z=e|vrO1ew<vtsAHq4iFzsv@B!U)OzY#Ewi#BwK}p_K*m5h;hHK3{pvH687k+(QwY=
z$}@)uPKd7SZq2q7wol**J3UJU2^GAD(Ryz}sRZ=-pyb?=8l%AqBWwk%(FogBpfZ7<
zcsMwi%O@hu`mb)`b+vU~0)BxA_yt1^@wS1cg9=FQl!l#!MpIayw*Bo(7+?d5fF5Ca
z#`sNA(=CrTi{1}s71jlHXBX-LAN43i6yOs8$52t7=$nU`A*H2Z2K1)GSfr2a6I)I6
zi6$pFn2rL36nB6P3(!O|i)vd|M>XgF)~ZKa7;)OEKMeVkl3dI%6C`ihh2wcL{){X6
zla6ch$`R%o>;OHl@CYP$KZ)P8@nBPjuQ@`nNw7|gRY4Ccrm9l~#j{E+JtFhCR9$NH
znJZ!Ig5t_jU~8g8sm`+~j3pb0>lf(Et~@JaRfU}fXgea^9lEU0tiLodm<|OSMI>by
z*E*VR1av4&Hn{tLB+wXyXKi6+01Ax|Y>j>Y!3lAWq(7DWh`Y9Q2X6Kmt;vF#Yw7$;
ze-<^yw=-C`vRE@P)E%pYkG`!LcE#?}g`6sUI74O_w5LW2inK5iaj+b<)w<2Swxqp=
zUD-G9H)k6IBt@rJP#OHldpZsl6gnn!CYQO8h!@U8=x^H#+ub#KEs~t&ugMkH|6vzG
z*r&+}n~*`rL>NQC_w#r6+ff}~*^Ui)0_eXtH~cSJg_k6yJF+B3F<@%Y9OcfMF4!!m
z!hy0imP@R?P94F)73iu%9OQ{MfImG?-wsM%W}0dnSwybr_q*E51PGe6qra}O=+#*{
z95Pr7eyv;Wo4Skp5N}}-KYm#9cB`aq_EJb>FwLh_u8{f!1%^IcGStSVFX&x&ipVxJ
zY6zkiF%%cIQG)Q>&ko{=ch3{t*4w4k2%T9MJI!RZ4Onzy`LYw<`>9VRGD%f=5&C4d
zhkc#^87id)HcZwW(Nr!TniouZ>zwQ5spSs@5+dp++ue<NA8Z@99M7^AR>(WZ1F8bC
zn$td0^a|^lho&p^Ds9ss%xRi~LrnJ61b2t;RB~{@&$x;KUG==ekBp2652)2T$7-F_
zM6u=13R%s79D~DrU-m?WCd~2rUQ0Bm3#}T?R8N)C0ILGx4Zus6`@)HkqYCH(J*)7K
z__*AzaUAHeu2H<;JNhWKm(g;l`Jwix^xHk3XN)ubhw-ecso~x;=}e2bZJ7|v!r4`!
z`JX^^9APd_;o*$3Tjs=9#-4jlW5;5&y@cBvBVx}uoiRf1w)A6XTO{r-ldDK4<BBcn
zJxFWs3o;6-CtJ)IMg(wW7&RF#k57#+ka+2)1{;4yc@pY5H<Oey%3UwYTyBnvMjCEI
zI}IC8mwv1xI+oP$t6FG`hY-jX{B)0LPogII(M(Y3zGSjCk}r30-V};$Q4W|t*t-y6
zt7L@qZ>ccQ+rh{glk*C+q;!TdE0w(&HEY7N$BML$997EfW_i#0GRO@$_ccypVe{tF
znBx6qB~SWtX>n|%Zq?RcIGLGcmO%|EDVZuDZUoIE&bLeS;nBsVsH?2WMxl9DFI|iR
z?dUB`5-)S`{=>#Y!UJ^jKGwLi{>|EajoVUi0X9g3wOy-VvMfs_{FF!GJDv$e%LZ)|
zLuFu&8?Eh#kIv!J(aIUMr2Y@RuA}B^Z(FYJYMZ5L%qQGW#PA|H#Y8K_y5?BEXM0<P
z8@Qzv^Yu<~Z|=`@{fW6t&t5HPfM18{_YQk^$XFcTxj#O9ejB&KT{WVG!Qvw+XA*qp
zD&TPPd0bxCqs;|9KlBpFTTIrbaf?9X<F$6_RQQC&b$*hucFLt_gqpMVlA3Ir5sMp5
zD=4y^UQZ~2ftS|?tCCP-CuNIEXP>VeDkNVm30d3U(J`F=vSxR0a^L}7e43>*n2+;t
zoNb6z)>gmJTU0S&W`-PixE1<hCrRJRMZVF7(nB&aiHHV!2NWpml<iqi+&7Q3r-t!v
zV#f)NhUKb5zSN#&Q1O!YIH8Dz^!WS{Q+Nl^3_2}oUq2sFtG2)`$&}3T^A&3xk{|Na
zAZs$I=o#?_YHP8nwzSu?#Azx=Qm>Fa{ro8^A1mYIC_qD};Xz828xc=Uk0%qp{#y_H
zulv&1npSt$a_~#r=Q|BE=vz)6C$c9=KB<NXTcyWc@9XHX4QIVc_PZR!w;)gMDxNp(
z<L!<C-FrETK~z3aNgv^ul)7OT5@w_yykjpQvY<$h3{LZ)f*f;t-t@%kCVgydoIyR^
zR|E~VbWEt@R0J_7JWuqM_mh(VMJN^qGqOmE2=~%o4ek%A)G~>3jSlS377Ju6uet-G
zApRFG2BjQ{?YQi#&CY*pb&<h6v}k?1Qt|_Le*w{y=^I+@1rgeS-iFbeB>IU&QFdr;
zq`H&))zX$2t<bQ^w!+0~XH)iSMNeGx&ZV8p?>o-5_FDEc?=FiJ>&4Di3NicbHs-qr
zU<Zq+1@bK(vmK;-+Kj4PxS;0<$!AkfI@l@^sxp6n;I?H2Z-)<MWj4TKOBCoO7Km-v
zJxS<d{hLU}vOdff+XiDV<FjUhEMtYtFPB%EzWZlOXS<xx{U+IW<HI7%iR{_Z$!tyu
z>!h(nYA0{C38@h~8n-RRB^g@zGAJW?R2OKx8sa_dwtcfS!6~)hCFXtyg^1BY-QL&S
zB&4mS7d=|2hY2iZQ|ZjXkv9Bg#MB((n8)lnGf`HjF|t-@3iC$c&cowLs>b)q<t`BJ
zYkACIw-4RQgcE>Y7MtLUM%xKzjRk1f@g&R0Z^?_C>29wg*dkgn$kH7WCqIK%H!&{S
z>93!<`9cJd4i_U+6r+!v&Vu<4u$-1L21}Sk<ASN&7bBS{-<V@5HawY<9$!1>J(;|Q
zcG`?9E(-4YeOTwp(LeH&^S2dK%HosNtd22TXAYgD9Bd5`gx31Amm|U!YfzL9D<o)#
zY;v%wACWX|zGHOPcTz|XO4*pAogWrpP@S@k00Gvj;oyBkt6K-!Xq6NNMBR*bx}|19
zhgMUE8Nd!D<#}`1@n7H0e>wB1({RX9Dp$?YBm9Fg$SCE0FVhZul$m_J%V+Rd`@=1l
zTjYK1AK$QDz5qj4t9nnFURk5pM8elKZjWe;6Nz=K)hiO}p0VXuL|9cf%4Qeevwu*t
zjk~by;WdAk)<<U$(qY>}wjSCk5zOHf<CTIKS3TH!yx9rOA|sk~J?_*&<d*HF1=g}f
z3Y4Ao)C76tP(C@?VYr~L=T4v3{?t9R?0kgRw(r<gH$%KpRo<T+na}^aMJES0{`v_=
zkdpRx`QY=_N~+<^XX>t5NAccinD&gw%U*VK%e1>4T}iQ4xhNfH_>S5+ACD@*nVJ{l
z28k&$8E<$-%?g?aFCd?T_Asc~9(YBu&t2iw+`c!vRdj>P3A8t19=yhxooSTe0RidS
zNsenb!1Yg!XO#Rjs|okW?SJ!D`g+pMKC}+OZuRWJWVE9nZ;wfkOhzZ4*(qj_xv6Sv
z344xv^#KKA=s0uCR(FievC&`tV*3BK@Be8&#D6r<Pp_|sk!u*OSic=(VnO<uZtsPI
z-;w3``KP`;xIsGTjn5UHscz3q!F=63xrxtu+kMtyrQGm@^aTT`)};r^N)#UsJ~ijv
z!ZL&fs7$#T^du=mU4zDpTzLKWwPBtIAo1;Kq`&rkqo&~{|IPt|xX?1^<N=<iZ^eH7
z^4NDTfBfe`eJ)B$%PND+!YxSak?~4rFSEogX)EjUGxYLv&(<5+C&ZuSAl&gyqSxPm
zr(rwAx^f@~P<|~Q7SVAy%djpczGyk|)i3{OhX0XspISpgV!9u7dP_!ekqK>Iy86nl
zT2c`)ZLCT5;07vXkbDTR=wT{ArxrHZei98(JlkCrfy021=e5a7%MFeBpGQFa*BgyF
z=}J|HrHqlt={N4krAk!LaqlA$8N@wWz=x|Mw42T5_Hc0<#XW-16WMbewC}SMfi)Ek
zU&#Bpk>UK%Tw33vJgZx;C#HoT@piPcu)KeEI^VzrtvKcV;D$H~ShEFZ)}$1E-{pnx
zS9kdR-_OicCfyXCvtt(56Ky5E&JI4eo}lrx3Q+o<nie0`gKLf1ofGFwO?p(w*Tha;
z%FDBrLwvQfn%Vh^en?-OK`c8t0^%*>l<(dfCw9j3ijBi~Q<+I^C|lDq!kgtl47FS_
zjKicQbT>YVFru@&gkoYt!XUorv^`xg%SBi^Jes0*QX^Ldd(FW^7A201?!5X%RhZEW
zTHE>k>yY!5NuhfZ8DFh0wH?Q!_P0?cIR^lY5BIcQ7<z6HZxU|8mp`VOudBWov8GB0
zu8)rGP+jznix&4PE`qgebZAFkvc17K-vVL+1ARI+C%Ge#pJu<K{GQ*w9`_*$8Z`Cm
zC)yZMFF)?YUQFY-K{O`=vD1Oei8;lO+TD{`FWzfIDx@0bp9U|_=8_%^y=vqu;>3s3
zw6G{t$Dzp><FRCCbE6lxphYCN7Aa&#u@@F8mZbxp%c<_9nt;rdA|<`ERo>)3|7LXQ
zwGtr4Was`&GzNtB!vGzG&=;k4^B9>v^PIjjPL}{uLnp1#UKt023jl2GK4+qW7v>8v
z7kU995`dTx*zb_i5$QLcf2_DtRqN@I>>mD>?*e*h=odY-IX$cL*X&YhMN3nBids!<
z3T5vZ)5m`OeVsz^<Y3lfsBo7u6_m|Z!i%K*Dc=v~(rYiAbDuoC)|nBFIiOX_=g!!E
z9^+wF<Cn@st}cws6HVdnhl(a9V~clDhqlCBKPu<Ra`dFXc7AZ~@;MV8*DaSSVa6p6
zwcFrS&;uI@u`S1Maa+91mLYhry8#Sd$#7&e49vWrwM5ch(xG(5I3;^%<A$7&Nt)=y
z!{|HYu}H~fRt#@Sm>K4AD%7J40Qnj&+kqY?>i4??5vITgn=7M0aLECH`YLM)Z)Jz{
zo7X=W?Y*3FvFr9S%A>>l83~X|2@9Yl4(d9L29YXwzgfyD0({*GFw;eUKP`X%k<&Fm
z9L~|vrJw*?<LHP;OG^{UD51xAyPY2Wn}7d2{2xbdax80sM(e;*#m3=WNnT=t!tW<n
zmja`Akc5|m!HW+&+&nmlwGgK-gX<Vo`>G?&Y&*j+kF)H|l#{+NFa>TH2)>#UNPFF(
zt#~#0QroaBSg7zUTyU7X967homoie*!Rby_r(a5rq^ZG6dNvS2>L^s4rAxd*VE5aQ
z;1|wY>*L1)hxfkV#0>a#nTX<!H}f|KvCBD-3=yQaK$%%jhUlLj3_pWoN1(=uHeMAr
zf1SbSk1eq4-&AM45>YxmTO15s6+WGe$Ul$nnVR)Z3{My<5Z)y9+%d9Kt9CsVz9Ek-
zS9-;7box>q?CnM;eZ@GP=jl!8g{twDcid^bqqxE((8@VktgD4SheV2xf^kLltX50h
z;LLn;n|rRLP@;1T&?>xJugIIHMtWq@&_=z7#eN+fk9YqRNgiO-3cYYbHM4rOqNeX`
zAx+p~(OegPeR#AWZXZf#ia1S8u%#l`g6O@?Phg2fcm6ty|FiQurTvh7>(G0eRT6tv
zBq*0~2T^<0=jbEt<?$SDogp;TUOryc!5MBfYM|hAfcFj;*@2$1A3@#9fH3B3;(n+T
z_D{NxiuP~p_TkXQV<d>vuEMg|aw~g_JS{nA0~<auVIkm(hiGIxrs#G$l$hY81J%)L
zVpg{M(i4d@(2CgNoT@y{K_H25Yzv5xY`@0_*f5m2TOnLZu)<DL(z*^l!j5JvD6*KI
zp0QqAIXqn9h<C*t$~2!potSSNA0Kb1N{EV$jq(lUc{CG1h!qCG9kFt|%6vk@$v|YI
z#|l1j{OaUg9?XwQAP~|OkDs{={VgC~5yk0FJ^fMg?=J!IrW*J9lPAA81_OzS>oDJK
zA|A;aGa}+Gfekxq5KyN7t%<B4G+K=1Uu=1l4~va~#=w!mQNE+XQHo}@L$&?Hw?3dL
z_IMutSbErI5CE~?)BvJhhE;>w_OXEvgicz**Oxhisi%9jS|tML@O?|Ox#3i!tMgRA
znX?svB;-84D(AF42*^1CfubNF4_T>nn^+w-lShPuZ&Zg3RwvGoL_i3M^HIV@Rc*i}
z4gGG^xa(Svs)b-B*)mU5h)bHZl+Mz>w)iFN7Vh|?J{wrr$e8Js#489yUn{33GGu{g
z4T^SZelJdY$p(AgqV-wgv5MFedK-M6;$W*dlXzgTE+9^}UgzRF_|zGFSJRm)?>v!Y
zN@}rJbJ;aWq@Z@8$HuWGPO~v=82kO+@ah=e)mZ<AWJqaYIl3drGN=a`yjGmilbqRp
zuPG3FQVH4J$RWNMRfYcW|4%mZ7xVsyzyF5;+5c!9zmH9*+I&9TaiG4Ce4I#Z#Qw~$
znV*-Ed*kb5lhHimIrh?L8g?)FN_*1a!`iXXnB0-*?X3$2Bon7<m3XbE)i%!pE0OV^
z`%=3iL)<mZ^gNj(#CBUU{P&>A@@cD<+OWZ`b@*y%o=C|vKH;Ehj72Nrtrm&{A0L?T
z_0|rp!uQ}OcZ2HQa!nr$z)Wj;Tk1IL8-)icnm@k|zTDyfqMMpqLS^GY-D>H5Ufy;p
z)RKdTd<xQ`4J5P!&2Tb7h5>m9)dv@C>>;v3|2_w3UzWJi9;8aDSk?!-ZeeSH1bAAS
z0H{K?>bm><H`h-Gcz=O{WL5K?p2LEI-5v^f?l-~*Kon^;VI56&_L^;~)!}Bws5M6)
zy~y`Gp=>vCxhYv$yPl7<ab+C)@ip{i@0EUwMF({xEYw<NSjEJ}UFKGmbFpga53Ie{
z-On+@sYkExNWU^Mfv%BGsNKdDEI6T+bW60m3&hM0$b3l_IBgym6N|<|+go~gH`}+o
znPF!#k+KFe>q_=s*4E$sqf7N4Irk|QHT`-jVzTzznFP1|*fu9}mPF?Bm>-#!$#)+#
zy1+qK=)^bBwZI23XpARv*Qf<`+LwOCm!p5$$V*eBsTLo4E06Huyr^jxI+ghAC#Yrb
z%1HUf*wSlqD%5qqU262$pgwGsQ}`qz1P+@u%D+&7(Y_z=0JW@5WS<5?*)!0`@0Jn<
zhm@Z&YKOAK@^>bwDZMaDRdPkA0zk5?nJ)!8Ox9|uv>yP~YmK7p+D%R2-PJ|4n@$~>
zGc;<^jUii=Ztq5$(JTJsq_Bqw=!UV#HShE!o(n1WrL)5sl^8F6^>|qh4ds4J{mC0|
z{_Tx=!L`vtwN@#`?$-g`RPmwm=~@C}xkdj5*+j}i)%4d-ZozIWEw05L+V2P#=UT4z
zFjTJwj4>tVS(S?<t>o>)hi~s$`tE{nlakH+nbA?+#qP=Oi|sCJx_XLJ$byR#wL{Xq
zje0psZATnEn(n>Wz5j1q3AuA(lsSkHZG5gXb=zNjcCFfaCOqDYygxm!3<`6<r&{gu
zFeaa&FPV6#3A!E3m4C9O><YRUJfc2c`t`Wp=z9@gW9JHEg4&Pzeiz#7al$e@B9TDf
zyzB@~U+OzWMsc?U9#jhn0VHns<jE2VzTIK8LoloG_J{Z?a$iK`)8XJENB5{L{uF6<
z=LI55*${6TO!W^&6LAAKg6-94(^1r}0(<*A|HcG<J01I3M@dcBCa$Wze{z6o24nM;
zyf9fs4_dKrOQ~<bCk9WX*u1`#Z8zUs6&TwP5Uv>%MBJM4davEka`sY2UW~-IGZkXn
zBJFDt_+h3XWLD=ya@Sj`WN_b5l&p`BDrkuIx&JdB4~EvFynSnjEY1va6_XuwN^4l!
zrS%5nd#lZfXs?Bq)q3!(3#7O5*zlO87-@6DT>qqAXodKSYm8Iyw6|=3wz5biPl`(p
zFFIs5!I62sqC0G|Q>s2O!gTea3o->WNm9t{_0Z34ulmB_%e98H@@?s_MAlw`nf8q~
zUV-^<zh*Bzm`ms&cSl(u?~iu(ZnR6EE$<d6#1WG)yXL|2)o2G>m(PQO+38@gl$Hr6
zC`j{rk^WM@f;<u|&jy>dA>e?louv=HSfVy43Leg8#ssz>76tUdd}q6Lixab&|6=9O
zcJo5#F<$eSPz^hSMhAoXjK<dp=tqgOYfE{d&@TW=iliSh%->52($Z?E5f1MD6F`rE
z2HX^YV#5NTf=#Sag$%c+XGZ|MEl~*@qr9$^*6ot%58^aZ1k)=TMWN>3Hc8FnP2hQ*
zZ<n@^ni|d8uZ1#$>#yijTbfg<003canKczIU~1znpZ6lvS1JU|4?5D{QBqdTt&ZWY
zQZ&gNT<mhfvP)zKR=$5V7^nRD4u;lWVWI;&a+p=y*~*L(=urcQE%{Ga%-?Soks;*G
z1B;Rxtno8$ggHbl6g%9H1wAV^a~E-f;<n(aIu1KehK6S}Ar-0@y}o9~{KXvq>ibiS
zRi<t9npy75ki)>7P=7iKF6BLamn?e|=fMb!c{mpBF>a*&>cJ(<DVS}{l-l6p>N;iG
z>{aX3?xH1cLLpGsc_QW0nMcq~<2l}|59?tJ>B#jnb=yO8*^>utXULLiCsZ*+Ixj*=
zUuj<Fo7Qt@w+`st&aS2w#g?;_*&!<nFh7o;H2kzb{SrHigHKk@tZ~6OYdhxGCOz4~
zh99hl#RTE^Z{U9azM_5q<@MGe{Lg=##FwBA=-Q3Q!gzAJIK0CiKrt`CFbaD8r(WIO
zj=N`wXDIu@A2TD9qPA~UjEYxGPGqeJPvKhgKMyXUv3flNHC~6r3dTM}Slrt|@Y~(<
z@G;_<l61AiWfI<kDU5DELh5yz(Ug0<J%d;~y^lVddVHoC_jH)Hy0D!zTB_sm;-Gfg
zr6r1eHrgp+&3>du7_~%IhQL)c(}PB=zTc}HzcWc}cPc-j4<(5DM3wrtdjI(2UkCr+
z)cL=#Zu#U?i5QJiyK*L69MfY)NTgENKC^JQKX^WVIWOln)^}j9HBP)g0foY~phDF!
zFLhA#^>Oung`mjOZ6)L6Q~E_V4l@2p&dcIT$CCU^su49CC7g_+71DAxGFRxzlIi^$
zaqY=mwT!XFD}h9fTIr{0_<N0DYV{_Ah}*uovtLzs_|~$~MSlxDfUa>@E!EWc3!_4a
z&ua0G@`3>-!R1uPSm?IrV2<vSR@_*v)nrX2sf`NAJx73#lkYT4vI{Tx<_-;~cC>ay
z-PUYTw?4P97>vlKFXqMA4uC!v{ui#)`Q+7Ej~zGi{*C&(`4*w{qDMK{Jn)mt82E!h
z9>AA)dzmA~yBIp#)z+EmTr`#~hh~$u99$p5+9Pdde*NTR%2V2mH`?w1yk45N?W^CJ
z7&{a+XL@)wSawl)DI~f**@4?JD?RXmHIn+2Lr!)#nVjh7z;YDN<HIp~tt_?)%Cei0
zkTphKnv>u`{Lm#q$RW+BBdn*_qQrYdCS3juWo)$%(%XoN#|t2;X2E2R-}|jeE9oS(
zX4AX+zGwnSH`+22Tv*rqVLVGJb0Ng7{ug1X#*i%ZW$^N_Yi{mw;`i6wnVK$eOSuaU
z%Dy}ZH5_0S6|^lAdI)m3o`A`T<-=F+C$^&x<Wj;kCD&BHaw3^3mXWbcg<1J9x?~;a
zRSHR`dFReDI%>_$^X3({j*bq+l!pf}p9Idg9tacY3w7K=i5^1WsE~IRu~!>X_{7Ew
z0Gxz7_HcmRUsu)z86Vg`U88UG=0tPpOzE!=gn3FC8@WVZtscl&d8a6AIJ5mxW)-;q
zT&zw_)YxmIu}FC!VRWfT2{#v67H_b{kJOKEJv?0zKG6i($=zw#B+up7vg3)uQ;TNF
zYZL<gz={)B=i;{*J>c6MXY?GZb+=d{prSmuXtvwLUvGh0M(BgnF{ZE~^zxF`3>hk?
z^oTVTso4XMkw3G(Z4rW&2~SuOC2x@p3&dd@a@$~aIHT)Pur~6-tkUmh=I@cbdl!D}
z9-OR|roE<&Pv!QSz;+7<k@5~hp|$ac#(5-%JM_?-6)NEL0n){#u<?3vgM-I~Mz7md
z$3nH^yUDDRAL47r%DbY$Bg?5NOpISvHQSiTqY1$tf|KdBafOwH^KaQIeD4H<?Wnu^
zLqnk$Sd`Z?qzInUvTMxReHO^>P8V)VXQ__8iJBys)<Fr3X-e6WYEr`EG@_708X{^c
zMu*cQ_|~ln_&@x;&k}ov_j-8ub4bUfoPlg_NmsyX%o%SWQBQf~B9q(eJxZYk2pB**
zt+nxur!P>u6dVh8EkPom+~T>Gg0dY9t|D1iQI&a1@^4mn(F*a(RgLAH<>XYw-pa+&
z8Tt3jaTcThG!+X8hNx^Tb!U@k`RDl#bXaKGJT>3wcyr_XK9p(Qe5JQ(6e~36w0B%O
zw=Zj}0OOng7$~&;t@c*+T+oiqb}zd#>AP`+v{_Q1@s$s`D%{e$z3hG)R>)d2(q2|O
z?(OMQWzL{8l<i}#!-3=-WTFYrk%$ui?zMjRXTDcfo(jr0`1x}*`Vg!}uq)w~$7!$|
zH&4`%`Y@e8$2a=6G=~?-6*)U~v{AhG-62KUdtOZf!G0EvggA$IV^w*6l)|nKTMwcH
z=ZzM;Thya#dSwpWs$qnwD_(5gzB_27$*+Odf`huhFX&eP<Mj&a-<$dGjy`pE35#!F
zM0N78H}AHb8LQ9kR~c^qQ=t6|s6fte?FvTi+!$eP@qtR-+Y+4ev^x>c+ivuQ>XfON
zyeo69yxZ}{#l9|S7T`(%#S@eEWUwA%S>*b}9#oB|AXKb45uW@3t6$1&?Kl^NBQwv$
zZ4BbiUU9E?<5<-}7Vla>q+Cut30T~&lG!2x&c++%lD2!}Vw)H<jJ4d4JM2{r-5hMt
z{+AK|j;1&OC>2GTG`5*3{Vp&%=G}yuCXH!!Od-j>V~w6{&%G+wmRFB32+jz#bB(a^
zy+(D*R(z5lkI;(!0zip;JA#$Nh7DIF?4|s<81j;pnA<my=)sINGBSd>#D#77GIEX{
z?Z5N52Fd?^5m&S3co=(kZi3lPCz2d<J0a4PpQhmO$+;^51rw|1Ox5k3+T9USmSi@g
zU9x-qfNorW-8PLPbhT_eDU<xD#jfUGM*jD`&&rckl2Y4N$TyI>gF8iUpyG(&OyQX%
z4WX>Xw1d-nVq3$zrs~}*uP>Jp9fJlNIKrHpxHC=`@){j6quDYO+28GJCI+WY0I*_+
zed3X|FPR)f<QTqf_>OoF*n(}Xn(oP25V8UBcBc-tC+u-ao=CPMF#*pvta!QpgDE?<
zqi7P5;UPhuW64pg60kB()BSD^&RYliCSoJKd4baNoJIeOL05V*f-oV2AB&kyfJ)}#
z5sa1jn3F+rLa&}M19}EAsKgQink0{(^jrt3n;-96mRj9Peu*AYDi>falbheC{4F;4
zTtPxWoZ@=Z!3mzR!NPj-O{QonCXf(G&bhV4XrD>g74Ln=78+j}Gzcz-=wj(`(d7!O
z;jc`;((g0a492-eT<g$E((hhA+T-$STG|1!Hsx}jr7!fB`hq}Kl<H&}ZtbZZ!$0xA
zNT!wd3R`zAcMnYs3E&$*Syqd}rl_PdU^xky-=9+0h_x;-*41BROpQ=B;~F5F)P^NS
z2V7?N%5_>nLE-vO=9=IA{pI>EAAS0BC(%?K(^j-?j8FLq`{nN_?}6X!Ak?vU8wIKW
zWtrOnZ1N8w{}kZ;Y-IJ<Ppa^#Yl>jW!|r#Sl*;zt?Q0JEOERg@T_r1_Y08_1OJjeG
z`0n2y)~9m;_06k809F0T{JI7}y`JRZxmv2P0<+#zdP;-*Pz6K!r3A*PfMIQB{b9t2
zHGHJ{cv+Nnh5hY-(qfT=Qo6e?*i8kN=-UXO4S*yC(O(<vj-plrxSWrTlZ>AHD5wN%
zYxMMh&CR~5+GfH{nci#619r2ndSe`<oK*>lecxZT-nh5^)m)WFf24xNlEJnI)5++q
zOHwx5Wb~P`E8gK)DBMW<;>`EMQ7)SJk=D0&xZ81?JYywPJK?iOlAS%rdFDHsOpay7
z@DgZJ6A;F6H$Pl7U+7Xocb;-u_J~3a#@x6Z!BQ_;1_5C`@6K1+Ru7|7q}Sk9sc8b4
z+p*=h{_xRKhs$GCpo*&op~LQYuLYcau4zz}YlVG|wU_!>wkHsr9R>($?NlBBcl_(G
zA3(qMP5DKhiC{{O>Q%<;_w@g<bE27X7IL;paP*c@n`w>M*|x31P;=jxF;rky7}&-z
z&_vAHP*UX<Lc!tk6$sQekfg%egsX0x`}SR#HT|JHEX)9p$5*+P!Qs}$&qBcO$}g#p
zmfPf-I5_8O+*-)9zc-)fbt&7^LF<Cng-??I{Bh9#$z~A*JC-^rIBR}{`lL6w&a~}V
zyF2VlOdav9|COEoT^Tk6+K|9rxxKo+s;m9#Ho2{_tsDr!J+xbP2cTM&c@~rt2Wz_V
zZrNPpxSciIYf!1htiwmWfMC5g1(n-RAXbqHD8!cdF2Ee^H<+q&h0_O`$-_Taq2)_y
zXLwx>rlO4$#!2RrF|L3Z7m&e$qgBlSB`)7a!;c~4>jRFC_6<owF)YEOKqtEW;phSS
zJcGX`Ox(^#LF&8Ja^}J(CR}dYl|5)~s(8w5W+hGlr5Del=xsJc`)5tz_2)XidXRHF
z%cJ9b$AqF}G%9q9;FCZhxVDPFpxu{_;4Hbhb%gj4p4iR`7CQ*CfDX00CYT?@G4;+b
zF5E^$Ub{p+$!ksgv`0wlr;L;IHjw#+l$~jGVb0^h&|OJ)n2l4GWl+!z%zl;EgA7fF
zeC{6<7R+*yTASZws~{tWy=7W>3CBS7*{K&eTB@uanu@djt~L(@-(7lpb^0svl|+rc
z^tTXbOdgQh0`K20f!{tGqq!{CGh5ct62@ED<c(V%dp6(bs-07_wrYEV9SfHgWf!*2
zs&7ZSspz@1=1q-t5oz<ph^y?vWsc}<n;KB5<)*?Lyvt79(X>=fkyQ)M>Qh`F#H`<u
zU<5{JnKKO##ju_ph^+Raa)ra@i|+v5&Oh_w9gsKKHB*;MV&5OE=O{Zbf*tn7y0XWU
za~~3nD=*{f)(hV@Q}jPI4=iG+Xm!L55koH+jH^#%)c-WHi{o{^P)v^!b*{v}JddoR
z3Y;U7&0?bC5d8h9HBqE-?FTn_$vjL(k=c>2RQjx`CxB$7r;+~Sicd~cki=YCy6sH@
znIx&_io^U0v6lL4i^YirYP>~doiN9Pq<TwpZKn}<kjpKG#vI}^%HE(hTivFDcwqS@
z_jo2h3u}*1&8s1ShyT!<o0W1WFI08gcrxiK8%!d%mMcY(gS9jUFYWxe^I*@@G|RPU
z8a~G*%>Fpu=iN%*I!0RIt;WRTpNE<Plt7^L)u^cm%82MP96&WWw&?j<1|KV{|DG6t
zcG^j#Gf3xWmRjTy%It)>X5I{@-U{6qlS0)G-c8a?cWvFmmy6;WT~SWeO-VBLTy2NX
z%A)91gb<c9!Ouaib_G*;k@gv3E@~5M5YW`isc+I=<$t@oSX#4DZu9*F-ok!DwG?l+
zR3z<lzfLqwD4a*i1ur5jyN09L#_V}XDKsAXbe=4RbF_<KF)&qc2?0=C@S-fS?HT7U
zHsN94xVgBvq3PWf<@oz9ZoW^^M@S?Mzph75cRUoeV)2@aK|!eN2moQCudiE^u=Z9K
zx%3`9R=bR9Y~f9YCh~{8q#w7TFA4H4zAZiG_m;lpB47GPZdDj`!%S;}H(nK$l;7hq
zA!hHTRaB=m_>5K@^IA&%xEbK3mMPtR-=e*|Wjw}*=RUehfK3f2+h&a8p6u6>nTr4>
zLDoBg=bg9<Zk9yKLlm40o7?5u7t-yfE#y~9$V(CLsS0xXIn(gB3dtHa3KF$}J#y2<
zsP~g1eWMO<cRGtwsisQhSTs6h33i;~mCEAD5=-ZIIQC9y$#^K#v@LU<GPxrrY7qUL
zM|~dEXnz~n6rPzdY>)1Wm|XQaR{dt2{5nAYUX{&CIDg*?6_}_bki8Fas}gH$l)Rk`
z+Dkgx#6Bv=3{7;K#U#9SMqzwXJ(M!?_h`q6<Q%ClYMl-uEw?3l-|VGUCJ`!TB&O%j
zvbYvAdLjiB1Zv|SOjQPGEAm}bD*Y{rJOxz3Vly10JgxWV+-lU{x#lki#K~=ca;quX
zpc|aHuW6WnR{Rr-kxIHMTixq$%!jkGW^C?ITAE+nv;wgs-o`qXgK9>{nrLt@SX!Au
zX#W>b9$#hCxuXVUCY;X69IK_lHuH5N`PBJJlak{0BMN${ErN;8G-52jd(^a_TjSIi
zzQs4c$(^=@D?LAt6I)DeZ8k1oq7I%AIgLHT8L?BugQ=iUw4UM)vv54EumEp-g_UyT
zNhVb0-;>ik66|a2P4vj$cQJE#mMw(U_&z21C*(_Cn?b(yT=Dj+ncGsBE|D&^Obd(H
z;peD;u-Ne#$TVuS*YXF}@Ozm`w-@`Wi8rcDluysznIj`h#V&Bto(I3ZcH6KFt2Sye
z<|_kxGC_l{*1`OChq)%ZTPZoEZsX3hoY^%Rt}#7>s)*_};SvR5i!YKb&H2ztO8KU*
zc>ltwv?rt=|HbC3sV=6`k)QM7-Uy7bUQ-h{62^^a)N_^TEhZXH`7+b-4XoRxCm#fo
zI~m;U`W9h^^2WC8g)h2?2VqH(%TX)3<sC2Yg-}I}akZ(ega{2s<YIg2(h0J&ro}~+
zqz~Vh#DL0S^+bA%Wxcuetw`4d-~L~OGwTrvhmdP9Es0?1h()nmy&ZN0{9f{myeIGA
zb)-nE;BSWbFN6Kq!>bXc`|{5X+coE8hQk3&+S~I#_Dn5=*jeSjJx5Q&MR+;1+07rW
z<yz#cznldY*`Z;~kq3TD=E^#^llv5_((|^~Tf`+9C-X`i#*zj=2yLZzQv|;ylO%J~
zYWN6J04k=p+lp@ur%Z65+dYsA5y|CL!6uo6P>hjnu*FXEmczmqh(T7j!UHBcmjwps
zF^%&^${!_m{?a~0VgrYBV5>=27LzZxx0SARo)Pia+~jQ#`&=}`^fN~~K|6fe?qKjs
zu>~@`^Fc~$b7ylZ_F2adjC6j~O2Q7o=c;aa*H6t+;$i2@VE|^FyKasHFCn~58)s6S
z?FGeHX}sR99*=XNuU@#Z^ac#A67I*Y8<koWk-Vey_t%ezD=F3Aqer}k35_Xp#I;7V
zQ)yrPWqAKx-~Y4REB|<uzx%BHdJ;CdGTv3B-)9az=i|TBp6gJwpT&6bK7wGfm(a(^
z*jkxdN5$AI$pk9k`Qu~bZ7d>qN-Ah&ab-!R&XpCZQf>o|XC|4<N!S6baRD{jxRBU*
zhbDtB9os6LoJhkb0Wi7Z$%rR_bL~5eqoJV=h0KY^w48&Rk1U^pS^k@7@mW>XAmvT>
z5j}2?2Mebgfs!%5o(N>JI$0)WdT+gv8Kq5|5{LVt(?Vyw`|D@Q1`@4OO}Su~gbA$=
zW1IU<^y#YitD)l=qc_`IvLQ%$wSi6F{?~*2eCxf$anWJ*yEIm}A-Z?FLQ}NSV~f@=
zlDbO!&~&JQ1?P@31+svYxMPZ0Kd>F$eI6{*w(0@{Vi~_v<K3)=gYXopMw8~dDvNWQ
zns$a7<1kmumD?5W#ln7ztIw7mjJjMBz7AD=5QzCYL;3RydWQcb=Bj69mxk_DwNiRP
zk(=ND1mW&(+@jg<oqTnacj4y_f<66h@7o#t^uoeK&j!w#MQ1(8)%5muuQF&_zFBc;
zOXC*i8#}F_j6Zxi`GL`x#HpH*pOOM~|0qm~=Y%(HYI+OT1FIl{M;fI%xXK#cI*|53
zZloK#g3`1#(e3kov$b{Mz$HqX-j_ja$=>~;)9Rs1<INGmiK3dL>ioDmesG9DRy+E%
zIjwJ(pIJ*R3*>WE>~wb0)63&A-C@Rb#;IkGMz(&THW^Xhf^#Lw1$41|+<Sk~^8Ls6
z(SP*pXF`H?=HNLx!ywcfJmp$tP5Ekbz8d&!aB{{2s>%)qx$3vFLq9Sy1Z5*sGa9Wg
zXWZGN<)~^s>wmIM&6nlnC_2UwuY6R|{kUvDU9tSJ`l<kc^Xchb-fz0`90nYZ=!Fej
z@9CidyTDzSf?Y?KQRt!P!s6@~!%~(h_@6swTg!UmylS3JXikTF$A&)TpcY`Pj+cW^
zTidygkE(T?cMX--!{`Gy-}J{=4aYlkPG;2S93101a8k!miMt0ebW28kS+y)JJ!6$g
zMCL1@KSy1M`E{|jTGGZDPiUlPqa95gf~W|J{pQatDGs}G#<2Nmo_CUd0Ptvh@K8I8
zfm`jnGPvt_k?s;tKkv#7+;&vRu5x_5_)em9oeeR`_9;^R&r`!^%dJ_A=n;~+inK?u
z-<Np(sThc%`#1aV@`wGiEwLs}XDc&QAG~;PNyk2(G1B8Kw$ds}a%BDMbBrRT>!Oen
z^Ca`$NUJ93yBEV;E6J5vh@$!R1kBNxId(;`D8lEkt2}gH(}`k{0e~N?vDVCB+J}P<
ztJMBSJfy~S4CY0B8Os-jbC$8z@AGgUf!)HmSef*HmHk7RR-=vL-3KDq(yxnFC^86y
z=98p4%a=ETt|Z=ghlk6^Vxp%X<)$YHWDH5(IU#BL{@PvzyU20@!fk{7iv)cho&Wei
zZ#x2D+WyI(2q4eXt$G$Lt})|2R4Q^<>){KvgKKU-V7yki3#^1C@9r6EC8k?T1)8gw
zG4O2h0R3PXG@ie=dJAnRZbn(JL;F)7EqCN>#P-RYVwNg5MHfw)AD*Trl`_#bboAjM
zE%3#mP%(?pjdP-gv4e9~&0=J>UW9+RLrr<#<d#M9MDt_ZQ&P9y1RvYc-3i|CnJ)G4
zSm}7$xTJL;<50MO98AsEaEWrhoN3l~cRw|MG8Sp}Hl`nd=m`?!zZj|3BEIZ*uDNhk
zf4eiQ^U3fJbxUIl2X~fsO0BF~)ogKo+(aMEC$kQBN@f}>dZU>eELB)ygwV|DN{&H0
zD~Ipl?eby<dzW786Z!v<_vT?u<>|g}byruZWhrH$A~LliPzV8$c?z{1K;|k?rT|f*
zAOwhnF@!*ODWFVIkTH<7BBT%k21vpbD8vvMLJ9&2kN`p$WC#$3kU*f$s_t`kRrh}O
zJ@?+d?{ogRJUsAQ8P<B&TEF$1-uL}}XjSBrcAfc+cs3xMVRzJCZKZj;>b^WE>NfGo
zF7wImE8eLmjN4Jvgm86ZjkI^LA06R2jG!73d6a)NEXC9sZLQ)I6G4htl*t_JQ+0KF
z4$GYN!S&xnMy?{QeRrAx`osQ2O#!Q+xtU<V<#t1omklskJMH1kKohbN$R%(tH~El*
zI&+HN2$?|O7b9wh;Ur`_F{k}Jo}D7Wwys^fW}8_9qz|fVZD$vF^lqx;HZ;D6)t{!U
z)U6njK2D4Uj!g{VQ&|;r6{xW9TBuFsHe2wYNp6XzRqm!%xvUGED>f~ADrbw=MPUo9
z>IPDwItEjgMC&9#!7AM@zHjE4VZh4wspVg@cfGdgfUCObNmAhR9_OGQr`f-UU2;JH
zXT_9)09L=N!(^FWf6w7f$d9+4TUm_FB&X=FmECQwUjia+1Fq0lMs2*TDi#ih?`-bA
zMpui>hAs!(zh$74YW4N3iDT8Ba9w*+|1Q{X<JE{}9wHslxHdv?0SkE;9|n18%B5>#
z-Tq*n^&tL0C$L>_jkjQYZD!>b0~+VZT`|J!<Mxca--LdsxfQfWc%c@Lm$lLphnGvs
z=)6)3F-wZcWEYDuPV82v2(adSXhK(x4SyAd0N><_7QWr{9;6dDtw)pf`<8w9t-aKe
zg+I?!;Ox55DKAgy<VM8DydjzyP(~nOQLTi}a&wGB$g~p2jV8S#m5-hdc)#p8aez(w
z7C?JE-!~TR`qxgXo@DjLl6rEiDeL&Pa6e18qd|i-?6J%2o;uCyomw53zMqNgSP>Ta
zjU*>;EHRn=e!)V}B3(mO9$QJFMePOJ^t-iBXwHpue0Gvb35G%WMY5t$@Hx)CO^?6)
zu8mg+NoC!OZR2VNbwd4N#zm?XX>f&H3SnXbw$aqu;yt>z97eEvIWx4Hs*+(W!?-#j
znSH&-@ds%H-50Ih$g@(3^uWP0<Z8UPX|OadMzTHyct1cX^-SDx=NWp^K`iLfY*(Rn
zyT9-H<Y>d49+#$iuV4_cSaW?bku5w;dI?h8ovx|#MpZJ&^r#P$32%r_lZBwD;AeEf
z*!S4dqW;>>y3wiNH)8DT6wCFojKa~^?VfA}REW(voApQ9IVF>%Emy4b{&(AlICLCT
zYc~Y3uq)A**~QX`-5ZdzyDC$uDds8mC40~1<Ln08ByPu}yM49Dt%auJDkZP3!}5eN
zI}uL%Fm)HHZqjIn^6pHZIm6C(mm@B;0SU){cd`Dpc(-A^tR#Zhu#XkIb~kmDOeW6l
zmxSddZMvJGpHlT3Ijl2samoFO?)x(NL39$JYtt|28$8gq&~U^Ud|39Wc`#WPgTJV!
zaib<`0-nXG0o2(~oJqRI{P<Wrmxb>k@3b@6B3bbs&$KrjMd(hBlO*9oq$!p^RcDF6
z{mq)sTWmBNxD!pV$299AmP&SOmWdj^Ed*OCCJ{!AEd9uMaMWgr@dQh`d_luP#vA~F
ztalXMKBdRI9aX{>zLiWe-w7qmj6I80ShO2{bpM2?OI>s8s*W7Ypi1$Ek&_s{e_R3M
zR~*OqXuUgOdJ|EdH=H;4n+gA1^6VP#>~lRF<NcY|jkk{FV_#B6G;U&yUG*(44|Kn}
z`zWd0<SF6Yg@1HBSA67~#<3EM``nNjrKc3wgm6ft@4avOgz#PTikHsT^~x4*5%v`;
zC_;KgPwqTNI8c0Wjr~O4Q82i$IB%yplxMUn9B@~ZUnWMIzpwqBB?gC92=kiSSKD^w
zo86ZMzOfvmE2S_EmuC%iEUI~Qk3MYf@JQ<Z^M>xPWYhMgcCdxF-h8zeiq-3TJ=cSD
zPN#ToAtCQjjUmjR8>Va6LQ~?zCJVNXEl%M+?>MHMba&M!IJtoMX@TZ88r1K_*#~5a
zZ2)qQfmP?st-t)vkpGV2m#O<z?1nPIR5KkeKPkO&Y7N-yXWN*xsS&~x^L5y3!g65e
zr&lih{aEVQkbT#1y9XOn{RZmkiu<NQyOy)i1&dN^jh_JprDuHy*Pm&Z91Z*58+N>T
zQ{g)+;;0p8ViOCIUd`X|bUcfkrwv@<P+fnR^yxdIxsE>WegbL<Gk@vhxZLpmJHp^$
z=cO$<+b}+LT4q=akTRI7elTYL=x}G@g_L7RzUpiZfS`&{+%1(-<1PU_BQ{Po`So22
zY)`B>lhSb<JJ<HPT#H)sh;q5iYD6SN#kx=R)NSaPEZsb}cF|dX{YAv<4Y#x7T=#F7
z9Jld&-3tfvK9~d!F5uH2A*aI$&!W5rv05*q!dGV=QvmO4#?1IDAliEO+BJYt*42tF
zoDCYmwv7h`Yi?F$OgR#GVJ@-*4Y0?Z*J!iJ!@KuJDD#>x{6KSX$2t6S9Q49wy9{nK
zw13}!6roy0_L=O_Sgo3b@QCI+9*k4@Gfoyd6}GlOA62>^wZwxan5`xT&IolGN&oiZ
zzN~dADsO_NijNQM^a)1h>kxPSx9iP$w}^SYz6&KjS3X(!@V~za|2IGJ>0Eq(it{W9
zVqCtVP9&}Q1u{dxcDAo0MqT2Vwe#Vx?G;_&d0lg)Rq;+E#-q>XLKCg(NSL#hqpfj|
z{?8H%^-gwdAk^VaEI#D<vS3Xa-)a+Y^ma`YRf^9^ls)dNrV|T~4;+;i*YMSySk-2v
zHg2#^b#q14d-y%h<OokN%ywiFAn;?x(|wLu!7QVIRw0yve0TR#&fJ=65A5GLyOyHA
zp>bXETJ?L%&1}Ygq~es}hOe7RpEqvA4~@ARQDxvBJfz1hF+GRK>33|r9p-=QMQB9y
z>!6R_p_eDtMUomxZR?JDgZN0-?5e*lIItbO$BW&LwGPBvh@EEVo9>qd>JXSWBytu>
zB}z5hh@&0e)STV+Oi>L@>{`=k-HdOEWF@eQcLEc#`LjLDIhxRPWSme;CK_)`gncR&
z!Q2e)_~`Q1k4m?9`2(1nv(Br!D@tcq{9$dQ{I|O<UdfZsBkD(A7b=;>7`IG@b<-ym
zLU^l@HESSJL(`W94yXK<23EPwO@eX%p$eRaE5>tckE>3u6JSh5{?PDc*@hdPX0cM3
z)dFc`W}3O#Js%uprazrQ>H9=a_u;U!6iU2qI#`w4wb3v`pP91c$?%kpWb;y?#6r~T
zJ$0smYg-(}BKffDO7fDd&w#?l$+*v5iYfsI+T_x>=EK6zgn#+1${bUo<B#8Hooc(3
zP^XB2%otcSB+h}&Q$vYa+ptiFh?a?{=NG%}ARov$N?B^#Q`xI}3Z-yiV>9PXM<~Io
ziw6c-r_Hz9<zs2F2QS7$RK5Awg{jo8pG0?=QkiTo>2pjeb>CXK!6_k8sUu>_O)Bsv
z#5PR~IA85&7fZTXG$Q~<6N*JXuiN`##Nx&u|6O6^Hyi)G^ktTe+ud%AljW(zJB2AF
z9Ne0<Rw+^J>P<8inXNGBVV+mn^Ce|Y(T-;@2nz=Zq<j}eQ0IO7k?(_0zO=Hn+Ca|O
zCK%<ZM=CjMffVhsBOupbCKJlH+&O1#SRicST-2m3m8eB`_TFMtQwpfqk*%Wj8A{OR
zxBqU#C!bR$f|wIQi9BA-tu6fOo^ofKE4D3fZ(^b#^)J+>yan08ON!<6W|!nFY`38y
zJ$l65L$WPe<&AHD+8q9Yw)SRWC8@yQ>tG+`dskRHG};0Hd@{r{Qfcz#LhMLb=xtK@
z9_Q-?tq$~W2mR;L7fE|QXPojzw_taP_J~+S{Ai8VM1u2z{>91fr)|=#vN$r7v&p?T
zz3y&4tnCK{ukJmvRkXLH?5UbB2VftA4S3jOe{j*-G;@JLZ@L^mfx$77fSA#-VG@x!
z6x@;Ca_)4p)_OmO7=Fz?gOBe*Cu7ZX&P!)f(qQVo*qMAw->&Q5{bti25Nr4TIL?{N
z*_gc5Ovyd4m~7?y`m&;~C{aY`Vp+G4g(pe2DJM)xr28|v0)J;WlhBbiYK6%<{#8Bm
z2q8Sqe+ntZbh`Kw{md|~eFU4J9xLgln|ivdi#IUD`5tFxG0|@0nNl2fyDy=J+hC*%
zKC9=B6Clj$I_B1IRbU3!D7J#e*+y@63*1-sJsg;iCca}Ar@k+;a{B2vJO5is|0kU3
zzZ!lhAGY5ba$t)1<&%MDUF3ZQf=snz)_2=|g0~m;%O~t3$yaU-{I{Ip|J=!EhxdO>
zDjS(fk+!ONA@p3Jo?8}voJX+5tev(hmbsFta8q`r;>O#2<#NO4wi6UX+`x&fIK#aQ
zrkAFr2oe7#mQS;Vtw60fLN*Ob+;T~%Qnw@5xUQ)ux33iLtAEunW2i-%Xc9!*;W9~~
zXqXhygXA{;@`>ucrnLHtiTd0(-~M(!f0QQo1z(xzu(zyJ^QsAXac7+Ml0va;Pl*-_
z9J}6-^vS#HL>tK){MT2P?s$ady;Xmm*KM`BU)$0|mjBz9ZHb3oUdZ62raHtyk`lj)
z5zh59r&nu&OZGaHYrGn^NS@Zyy%Ky`pRjn-FD?M*W#`?IHM2iychiN(#`5M<ocd~R
zzWVAngZ{N7o&2|1SA`4{`P!d1B88-nR^WyZ@YUOVGXW##(5Futb1xN6PR?$al0>2Q
zB*#}T5Kg9=eoHfeNJ?TXe1zzRh$e0083I(h9ckN5A%6w)FrOU|X`VJ66KARJ9B=ck
zpYwTTtxYlF<SU4_GXqmZ_L3pTix%NbPjrfN)-a3yUIbqZT^>RZLg1iF9~eHnsCJh;
zKPKROfr%-VdOkS6^T=1LUhKa`PBgC^tT8N`&D=~TPyoy+UQNN$UE?W((-Ves+rzVO
zuj&dnxKv+jKd8<uXpxh9Ffg=q*-?Ll^rO2;sB@Fui#GSWA8b!=75-(_2GdkK^3!hA
zP!pKJoPt9l$J>}EcnPc4PZ<$q`uwxSh@G;e0t#hwp0Qd?x;$HB@7T?Qj}8e$YMIg5
zEEUz{Df8Zl(BKT{$-CUK2^Q{Hxw;=m3pO`fbM0<L#EcCKZZu;x+VcAME-$~rzk!E{
ze->vKG%(6|9G$si8cLye9(8Xjs0+mjISxMw!#{Ew8gmAJ>@zd=nTUvpm@Rj!kf_&I
zmsP;8kXVdk)jAN4r^3|N8ut)>BAjX+VoRy4&SOQTn2I`9`)rIFw*&C(D0m5*u;E~8
z%Ab>}j@6G#YTLIhbF5_bo~La}!4)>#8(W0O$5yRM<cP>&ytBJnXu5`8xl00VU08En
zZ>J18zD?6}dUtK=<Z^HZH{$+Sj~ygAmfuLnhzhc=K5N<RZ6SUW#(DdQMtNY~;9T@m
zfWR}`DS=hz+d&K=FQM*p9)(?1A_tLaoS}_6w?|t#xizEi+tx#d9`v&-%2v#(GP~j>
ze|qvg%HsRwcZ*1$C_i0HtwuhN8fjR}$#n6Gld@Q(#Hj)6yR}Z^&%TkW_jF1y7E^#%
zHPQjPiP=m$xcq@q-VYn&wRyuJhS<8kQD)AjM%k&7ewWQ%f$dNKTFw8lodQhPa#=Q;
z&LGQs0JO>Ej(%>BzRi3|X!+b>=fhWwvS1qDTwTbNoR2W_d&l}!MnpR}I0T5(!UP8=
zKLfx`yH_`Yf!Bimu11)uZ(Dw_yKVd?5qmbaD~)D}x|(l6D2OzpS4R%tYwQtsH)r5(
zj>gL8Uug5ShQA4+!b|Cst?_=YVL6baITerJ``)`O8O!*N=?I|D>KCSYT&QbvEpG{n
z`al|PNkJ|aNs~3w=80y+`*Y@+MM6u^@kjyajeNj5gHiD?uPaATv9%*Xw35hG6@`H8
zZDOO-_F7A9PP8}ngGE_KuT7zH&%nR{w~nTr_N6C@@Ogxv=sLZb3Mv={8PzFA1!Z|x
zy5#q17+k#F|Mv1<e!FH0O+viOEcnlYOY(;dT=xhQ-Q5v&wA5^@G6ZS08n5O#|7e}o
z12X_c-m!1P$3%*Pt4t{qM~JkR3O^tmhUG5Fsp6t(N9f$v6zKGx9dwpF09sWk6K<ox
z*1-fQqJM<XC{*fDYrET3qPO{AHe-&FgwlgJL|=V6H6tM-qFQlQDn+1GMne@&`^#ZM
z*CsNE5-~HlMG3}G>q^G8weN)HJl=DK)RH&MHR73WRULOO?3^}{@w3TA^2p_T?n|N0
z@vR?EYGQd*KP1bXPVZeZUbZ|6cT%)3Ad_dqx<kbZVg<=}>;AN<R>Q6i%fRLiJL8~9
zd|Vm5toD#Ck+q=khq24{Y#oMQkZqBP?Da;=SoGb;9iVQw$N75e%A_k1%W{(XLX~wB
zjDKT3K_0_w_Sto4jdrpvGCQB5GB;9&=2CXy3q5*>b%lEt8jU5Ox(l&;&nfC2!v>h)
z{Q9#4PJX;1bj$GmmJ!b2izVl938}cCa54W<tGcmLKJmgGkBZXg^92dh=;0J++iP2y
zZ|0XKU&X4!e~eC`ODbugYLm{^v;v9S)T#G(sDajngSdl;cqyGuw;tD~u+5%4&TG@T
zJ>9CFk2<dSYy=34nbF=Iv*j@W4>uNe>}4~hAz^nwEh}LIsNiG`Z;Kxf)&Eur{yR$l
z-vb~2Q!5oNGWX)x@-P5<P^S6sF@OzAgw{F<5CV$-TK=fI^gsRf|BE-i_%qrgafE?Q
zuy7CCz<r>q*DjB~j<(k@K~UN~=lpn23lQI2?LFNrzY8;Dn`oC9ZEoOwuHIRz*UM*;
zKORb+0Tb-9x1;j}MOfylocP_MblrtnygLwSj(E3!9)3dIhReCcoG}j^URAYc^Tr6Z
zw;3AMQ*}K^&lHW_W<VEuv$vt2r1|l8wzt>Si?hh}AE6f~IrPT5-2S*iWNS-fd8|E&
zm6}acjxw;;(3-2s*Qth^vRDNk^dr@dZssR7^|$stJRh-Q4!HaDEFcy#!1}ZGU}km2
zqSB#<$HR*dE)Z*FG&FjAjQS;&6k1{w#IAB7=Wu1d=LZp|X^%^w$BP!-@_PaeA$R*m
zC89*e%LpfLtk6if<DtU}=C~9*DZ<7is|THS4q~PJ<#z`EQ}GW+4$gxu_I*FjHq$@7
z)k<ACbZ6y7;DAkN#XM+^*C;4`Q*r~NNDDi;R&(hTDa_gKU6`~)xvb^X-HNN4nVsW-
z9(|s8!k2`Qo$9HtjF}QQ*ym8!{mY}Jz~JFt(rSU@Y%|MI9EvCYG)rw~Yq^Y+bQ?gC
zyEgh<UJr@(KDaZ_jf<u4cJZv@2L`zmRC^qLDcOCv&!@904_MdkwSJ~ld89_R0}sk-
z4GhxMwlQhDmtnT!{?5E{Vq2hDAHDWxLEy4<xtEuSRf2!l)G1ER`{GTN+gCq@4F&i+
zMm;M?b^Lg1%<O2*BnNfn61A1(MIl~E1^pFsSQ@;QKFxWQW5~&&WFFFLc&M%pry)BG
zpNiVHfZ%()Lm^e)i&1T<=l^`BNnoXF|0RJI^mC<ePwx5}bqC)w5hsM<CMoM4n-PK5
z(24XM&Xdg(8>TP0JAvne8_4)EY6O(k9ttGO4*?=^-+%c1b+?41q1Z(&YjhaO@S#jI
za%O+wl!jhF`gi^QGqp>6`&aoFh4nF^t}6Q6v#Vi9=afhN5a`V1le0VKT<??RCeO;z
zY~qCG@MH&N#P-`Ifnmh*-m^V~GHWqN(B;z3_rh5f8=fFsMIVa@<xLIt)wPmBGnZFj
zZP)RIiv7s8)AQh~=~LrEno;rsxE2!B$m@3z_M8CsH<4z}OOae@y~F%u0rursz*iDn
zcx3x^;pG0}&rE=e{&1ADqM$UrG(xV@s3NMYUKQ%;h6VUJ^eXb&`^77Ux+B#dPN)4-
zLhOB)yN;y*sLHzsNp{IA9mj2EMX>#bph;YuXvs5(yxj=jt{uaNrkkLOn04i2$=4U1
z23&J`@SFajrX29OP(*(DM67LreB8mD8C3l0pmi;3Ycnia9Z~2orD!};a`Q0`gm@?K
zVkC)P%`|JKH6QYrWht3o2xf-w21hb2GpfCO@0QTZ61lftyTYC~WNi$+52~A`_6Tcj
zy?By>ov+>9{H@~+p7Hl>7Iv&?sAYfl%@wQDV#>SC5@7H-r)PYhNl-H$B9p7|dgfdK
zj~d`wlYDu3zAtB_ttZepUx+HRfyliib>}Jy?BHX=ESJ$r*m9U_mGMfCyOFX{pkAI@
zhuJo%tt<CtL;LUV|NkmmJrNw=+%7M-`qSRdo#x|Rkkk>CqNVXbvy&V0C-g~(v%;;k
zva0Xv-1ao*yu!?OQ1Gf6(>7b@Ow{IB+by~;+<t3UYhAWf_fF|<PV*wccXSTyDt0hr
zN1z*9*a4jd$vB$@)1)}w$mWf97b_oF$n%L>{d{4g*@nlk$3C%X`=G4cx@~^jbbk=h
z8Y&bQ0y5;{1@+1rKf}CbBgj2k1N7v01J2B<-(@iRT|Cpeet#Ng$vk@uzS^<<b_1QM
zyaj@RJgk%VvPz7H9tj9d<IeD?lu)$2F$?0-nN;?=&#g^=V?%VSMpVxU*H~|rB<*^B
z{8Q9^Ihg|;Cft1Bt<FFz8CF2wmqkaOrxh*vtD{(0DER0eOvC#9{&EpTEZ^6i7BBR@
zO55Mzg1ZFZxU}a}4?h^L5(ShE_9QLJMY{Q3#N)%c;)RK-^tt70kg)=@;QZc)+;30+
zeroW^C#qeY`+uEpr>uFY=Kmze(V`+n*ez#mgl8|8oGq%MXjPjTn`3d4GgN&LYBm?%
zrk=>uN^jeG3u!$&)YWq>9^fOlD7`L2hrHf>GAs?8_-SRYGRCsZDjYMV&ka>cR54l+
z2+hUZ;7dr=xbE6j1!IdiqkdUpo1JCf_I@u?)$kmpdlj~89}%AC9-Aa9MUmO1K<*cF
z;SSrSsBS-B=N*VJXc*dhC|{<6ZC$Ip8ql6~-`A1d27TPz+UboQ?h&`{oE=Z{@>Vjh
z^sR;lo9H&2G%yJ3b7Nm;_;T<ym)PQC68DX>DZf5)Nh=$P;!QKqA~Rxwk^YRS0l`ew
z#??&exFc=)J)zEI`y62IZN?fkkM`5M3<m~W(>;L3X)W=6$IvsDc$Ac6ylcBBz}Su%
z?SpB36`OFl%XJg7mkp!ZqM<<ARccb>uDK2<PE~DqWtI|`yCbZEkzIC=>ul<K&Pp04
ze7s4ZTKj>}xr&sm*3>W4tY&}+Z$G}k*;e;~jK2$w_^2vb^In}m6fc#fV`((R<AsF%
z9P4;hmwr#v2P?u?^IX;`+;YUUtb}jTc8}Lrc&Zwe>yOZ7MxQOBNl6otSKUoe%!aCx
zVzT@6d_$Pa-eqrgrm%yiZGDZqO?r9fea5{h;J%0Sma^U@snl=<GF)DaoBqf*==ZLK
zDM_0)F|*366yG#HE|{+ie?_O`PZao_KAmy`73Gl_^W%3_UaxXlw_=ZHPS}1sM_#v8
zsgf8>_1~5BN6``KvzBEo9$ZPc|7hc-5bBqwii&&4gt-q)zj4`RuY5{5d$u{!Fr7!8
z0=-|1Y8jjdCrbv@!aa@DF%&9%$Pl%El874*K2J6_IvhODzR&MMWP~)=bEwJA3;m=c
zmE~t`y#E&HJ9DAAz^IVpZPycD@>we5m8ODa-J1ujLA=i{X*x>7LdW6Cp5cKmIyEX}
zUW%_;9KhemWbCB_D&ldYl{PY>%E$wWYpHKr@>zMh;PSP1-OxfQ$<A<@{cpnH;gQi<
z%bNV%1Yt~s%lC+B_Is#(LEWB>QHjo;2GWX#2wV8fe!fE8axoa96LOgKu_sc%4|&*}
z^+s0lNa_MCZ=1YOexY%5Fs<{9^WnrhN95Bs7uNnDu|irJdF$NYo^tlId%Ux>u6D@o
zT!zk`F5uPC9B=CDcbGn56Ye5=LQuDHJwg-f{Xi}HdC3>)&E{xN8M>H>cWtJ-m<w3D
zgN{WhCPhinS>%UgGVD<;vCL=WgZW8Sa=mjbfOk)mSS1a^Hj_uPI-qR%7rn{MWZXs&
zc{x2e8_UU)s4(0#G&&z{%m!)#WYBQ;r30m60C1q&v)9JH^>azG#M1ULO^#>*7y&}n
z)>mlnvvY|XgR?n>!Q{BfAJP3ed8|0grOC%K)LVZ8bWUvVgAyqIQr6fnpXkb`55Fzd
z+IGDjr$FXCXXYSdQv#ay+uH5Sy$K)Ds&RJlPs9r;IU>~xx%>2t#gyse;wK%1Q(IdH
zS`S>?5Zn4q%L}r3f}smn0$dZ^)jy%Ib8>gW;O<@cBl^p=V|bsEp}q)#rQ%<sm5d$j
zPq&f(HpOMIVomNGsmOx~ab<f7@X#|dmX;*0{p&2<u7MdysPK#Xom0vEb{WYy<%V^E
zj?cD|^GV2s`9(T55Eqr9$;L%t%9Z>l`*2%s9U9oD9j%KwNUm+~=HqSQ-;l|@5fO;>
zW`q$Gp;Likt`bJbFCqlDh=jK?Qko4W;uPU!a00tP+E8Me8Nqps(r}pFciQB9p`!h;
zaXL<5GNPS(O7lum+Hr|wGJC`lHs{BsIK`W7z7320$f?pFx^h&~Ud-Rs0Tuprl@5Iq
zYXCn)C}v22jDfVtjCZW$dw`PHnafkhD(&DbC;sUh;_!GNy%9%dZj>TGbph>$Delik
zI#%OM;vTHbY|iNL&AZTCkj5Ag7vL*a^dT5r^>mZ|ua5P<^8f!wGpbL^_c`LPmO8Fb
z6j(}N<S|Mm<7qu9zDgWSIhwTgO(dD9JX>%E<mX(CB9S>>MAI*D+FkLFMo5ubsT6Kx
z_K4HTi&JY?z{(h?G-LjkPqY)Gx%CS|(*}-$F<%D7F=N#~2c2p>_<nBTxKCe$e_XAT
zI4M%?0BU9_Yl<;UYrgD}BciX7Vn|Q|3*l9~<_z)9=1Jyf<M!KOlCt|0Oq2W_H_n)U
zoC1<jS*MF?7ItQy%|NLM1xN|G^<KgPR0uN8jzW<TITSf&Vjv_Kcci$3@;+c<6|IFx
zZ|t2jFwn2M^#cCnT@bk`@zzB@kGyk%@T`jY_fv;EzDP2ETjw;_bPUYyiEEFiMBYcz
z3AtlknGR2Kt~x->sH!$fr`&NkcXxH0ok|xzt`!I2k&GwDM}?;=4?8ZI|7Z>V8?uC7
zY;q}TyfoQneD{nl{N8A(B?2<eM=OeI{qw@Eaht?J%nWU~PaCV=@6ym<<WIj<&O@f_
z=CEc?_1*~K!up|aL;~NYadfjz&yq}+M5?DY5_jrnBsQRTG#^m;QB&x}{5LptbKNIE
z|H}63MO74B`Wxo@Rx4-9YgZCkz2xxyv~|iM+(g(Ca-3XffQR$Otdjx{EE(^kRK)Hq
z=KCnLAFgR8SQk$=R$N|h+cD<F>%M<bn!{6}xu38!t!Ke=$>8L`B+u0@N?OP=Fm{<`
zgo=$Fi&AM;S|ZaW6{YtA%oNDXRnySCYm~E)W6YSk8+){BdXadPwSc`3ERfmOvX$?p
zLi%fq!OW28_&~7GgD2jC+1gx^U)_JM%2VE1?OPWOKnO!v967~pFL1t{W}4Ok88}WK
zL)z0+dbrvcU@2%^j1u2w?u8ilgiq>)S_AWe*ywJM`eN0NO^1_%DO1?)3Ske~@=Hv2
z7CP-JWwAOz1(f!ZCS7$eyfD&Th=+0B$Ws~C1J&<az{%K-E>U7|yxN-&brp}!t%VM~
zhrtEk70d&4a@Jg#<Ez&CrUE#+_oN;NUt07~Wb51*lN=piv!mY8ADh$JJu5>C?5vr9
zKRd0BO}JM_qG3-vG72=O9@++t#u>RHr+f0$s!`A6DVi^Om+njo%}r@ro6OeJ1C`;Z
z^+`LI>*Dc|9k1lVSN(nkn~-aM+8S}hiJyO~2EQ-GW;X73FBdvX4``Yao!s>jZS_~Q
z%eNnFhq*o#nCquD=CaLnT!w4zKuq6E?GnpDs1*_HPG)XnVR7|JFxZdwe3khMoW4s~
zuDeV0-eSH+NYus$^pMuWHSh<e>Qsm;EWT_$&bCA~n3fS3)txj*riI6KzNn)#m~;X;
z*KxIns_wJT4Q!l=A-lI+nkn89Ma<TqWPT;Xn>gD8N+w!wg?T!2zTGAnaq_wNYso{j
z(l{_~<J5jV-!j%|{<N7Xm<-yu>^UE?sqD#fN(gTaHfRf79PL|DC#2LYnqR8>SQUS#
zkRPWX(a)_9>6*@4;cj&e9Tx6EUmL$zBK(*D_8QPvNo<JB<i#z|Y{?6|B7~XIxI~#v
zH`p7`umD&u4Ha+2IG=LUmpbTeCVqurjr{;dK&MHo6(sBh2T*HHGGQsZCfcYfM=Ts@
zFQZREI0oEqYjhA@5lE<XHR0H!h(HQ)X!#kI$sqQ%v8!^M@JzSSBiNZPHhUpcOYBo~
z4bb%epFcK07V*$|&BYTMH{_>{4(z`D{rv*`Z+EQ*`R(0XdvSRw%XNAp)d(COoswTe
zF7yakR=G}T94x`xMGhlx)|t{@c0&W)qa94M+OkVGyJy~_A<dxzXSl7`miN#vpFk2+
zV%eT8@h8R`_53@KY+Bl=Wu&Ho&toe4C@XC0klwp+cPcqgJB@po?bmHvGT&v!0RQp{
zRHVA&CCFYqlshWiMAh}zT3@F~;83tGtvc7RuED)VS{oiO>EEj|1>&PWsC^|VNQi!}
znJPS3SvC`>w0Q%oy3KHp7kPc_2a!t%S6%o5S<yXrylIuh@OwU+?K8|b9Zw-n&2A>n
zWz`Sn9q9mO_4l9&OVE@R{13=6%)c#54}uxHA;x*0&aL<FV9iUD;#TbI6vnD@p25*U
zrRF**7Ten~mbOny3!ax5&m4L(o9u-}@ahzr!Q&1()&9kR@0M^vk#JJZDpPLkcOY2%
ziD!5{xX(2Crz3|YB2NtO(q0a>0lrsI@n+29{Rgvbh!K{(EUUpF>|)Uw{WG-&;<2Mn
z?L}{$Go3RUD}lIiMcLa+X*$h#rK07w%C#e-7PSr!=Mw~yk=X=1)FC8$cFQbukB(uf
z^(Vx|HhA}=$-xX~W>GP==7L!*m%7>xBH3oCBsMPs@d3;CDwC9}>{7+|m~cRGam$Zd
zbf7e7!?XD3iU4e_BYpMbRMaqq{m`)>vPG+PX_~ukZp#Z;4%5eX5I?g4s4^}Ao;_!4
zNMfv>Kmd%d>{FF>s=QEl3qoJ8tE(yWhPsLh5+OP|8VrhXb@jJ2GmE1}Pd2@QFMsHQ
z%&`e(5FoqGrz*VIFl;EFQh7f4@?B;|VPC-(GgB=hPury+!&6OH5zu99WB}kI?FhIl
z{Dh*hk;ZlZq#L&716DJ2g-v-cpFa@L-p91-C_i6tp(xi`lzTQm3Z{|3SZr44&Dl13
z<9d~CZBPxjMeX8{U!Wg%LcCq8KigOMc3~`jY%hIwby0j2vs%nWY*4pt!JM8m)>z{V
zOtY@)LI3r|k(hcOUpXdJ8GULoeAL+UFJqN9k{rw)k96F8Aqrz=ghUUnG7F3QpeZUO
z1SvR1ysz`L&v(<b&p6<lqKakfB2n1&%a#v|We~!Sd_gxF=vW$%-hU2{ec=si5TP!P
zsqEG}n}})T#sY*Z)$qg}kmHXt>rK^YeXur<6PitKq~tnXUT}W3<yN-X0TJ$EuPv9l
z-&#%zuq{b)SEzOk2y<b;47%ahAZC-R{oKouh}FC7)@Iy51U+kv6rQIjbA#m0??fa0
zsjJL$bbQj>*jSIHp^+kZ+?zcw>~aSiO*8$17|=7~A<E-)<%_(3UFiO_zU<!Wp}UFk
zEza$+oC32Pi@xD1gc(dNrzLmi)_8=dvczO1PGqO!w&^!kGp#>z;e|F#Io7W`EYBYX
z_nqi-n<;y}4I4Cc#vnd~)b0{EP58kIr`c>ZQmI#Gf0Q!v-F2!(V6@k$WuTr-zZW6`
z)Mhf8(G99Wt+UAEe_Ng)5|&z{T7TNW*(Cdx$OQ=^CGHq0<nIg+3?_rgJ%3r)$YHmt
zawkEj4?Y$)Qo3x)Hw^wdfeD)WVJAcU;+}hfsa{fLNP0uOC!m4pJunlD)1pSrK{7Z!
zI<=t^R$O5v{lM9_Zp;~U<Wd>1?r<x<wtQ!m`OzGdH*C;>oezf)qb*HM(CukwHba@r
z)oWqv`z1|P5D5R&bhEXLSy#>6@SR75I&#2&PQW|N-Is{&Pt!o6SI9(4Phkd-@r+_+
z4!Dwz0Bu6?9=9zyb|a+ohWIpmn7Z$t%w69r$ZS~N!<YZPO^7S=Hp&?_UEzXhgcdp@
zILcfpVI{Q!>T(B!{(GB1ex$uwa4GE@OT1$7a%tt-q0t1q^AGb){lB&eIk11U2~h`Y
zpU>ya>swomRnOawZp>87=&-ShM~+rgNUeahygTgk$=XDF7vTL+!slV`(*UVvT)O7F
zf0|Tjm%47Vz!sQubFBK2JnIeZ=I+GLkALlm!jG=bS7@Po>xc3DQ|U~IXLDb;C4VgR
zmdx(cAKtv5H=HMWK&&r{b5mR$7I%nh3nP*2at_93*@g&TDSg!cL732Hgkq-U`OOI?
z$NIwkk`Nko=C1(SNm_SpPNQD~0jBC_v(=n~c(S~!i~@(Cq1eGL;RM094NWpK_V<b6
zg7U!dckT@U@31dLapjdXA8gGN86zX7&_zGD=)4{hg6yFwF!ium7qFfA`!TF;WyR5b
zch@$8=Ic<&EgNFVxbW@^YOn!m>AU8VKJN~q2iKo3x5Yfics}9{u0X#MF!rFvc7`y?
zXWaO>7^`mVkl{{pNLlAu0oxqkaib5L(bE-rr_aBr99F=iAp3n*p<gBCjBG5%u<i@D
zVH57Xd!tpj6!VWeOG|!8c-T;K``BnL7vcp+1XFL6FkqGrh0UHgDqpl#WT_NAD`IMy
zZ~|7)1I&W4`prJm8*C6=uZ=P?;taW{UKbyVZQMC+7V|>uC?wJ3Zf<EaY{^!SxOH3w
zTDz4)?u-R1c#+Q}w0AHT#eTPn9@osZtJzpo6$TyHj!})1SXn#8Jvh)>Xd@Hv@O-v8
zyLxq_buTGr(vc9?J)eYAPI8e+xMQGD()pC=oOWHG-CrefRbyVO0e!$B9kGt-#?Bu_
zT=c?GWT1$_PtDqzFMkuNqV68pQSDZG(DA4prxtYu@2=)$RN`C{XkRvIwUM>`O~5$!
zg%`=P%3$H=7VjV1v?E)mtm;1HH@Rkv8+LEX1Bp5$=fL)<vbum8U&AU2Fh#abq;^q$
z*sgT3B+phvsAHA-HcGWe%oB31MrV(|VrK24Hg31Qb0^$82zYrJMtD2Aah*H~<espC
zgCZ0M6+)|H-$L)i9N%4v<0rS;HNDxjiK_iN%@EP;ex}cT?${Kz2CVT1@_1rpHiW-_
zK8==gFVz12cv$Nx(G5p4&4Y6Xy#_~9py>J}t7yxFK(`KKn5;}flH-yL1lL@NkUB{K
zq4>QR)#lLfIn6HbIhT#QU>^y~?DTEC(OxV*LQ9jzplc~{>+V!SL6h8+Nx^O<u~=u(
z4;JdU+B=-ub|)`8bBOK2+ex`4E6H`#_S2)iNxJXPGPT49oo=CIG`K3S;u#rFv2au=
z#fU5t4aAn2M_xNS=sfC!j0GEcpUAGY?;e`*t-AV!t?<Q;p(|SZmruT1qazBI%JQ-r
z?WYv`!=?ZSNaMIs_=m>5H>>wou18?puDTmw)cOHYq?KTX5$bu8w_$g5i3Q3)6AgVj
z$kw2&!0%=bg)FXC2p7YBLxE^<RZ69QoSZ*I-L<^#X#0N=4fG#|4FA~+cMU~un9|kx
zebzGc1M%&*bqT;H4Xano4r=6%9S_xR0`dWp!)H(Y<?lP#Lfb9D@9d=)9Nc-ZWkFc1
zQ{Mj>7*Sfjn;tz#XUlV-8~RU7eun%mHt#<>?jPoSe)%L4SmVlu0RGNb4^~pHE*v-*
z%1&cEcVZ95J2u}!pXNvZ<Ea1HKK`G)eyAI%aQ=RhJ8{ACP+q(m;@iP(oMpZ8enVvc
z=<Q2qaE_Ws?s)C!<klhJPNre5T1OMVuR6=~J{1aqaH<1IR?+A(cukza980{9s9H~|
zgRQ96CUAubkxCALd_jD)d$Hv9@CdWebS6LZZcjp8!|97t1`tm*(ETL@Vlk?unAWUD
z!hmcOT{l-m_bOG-RBO1K_vFHTHjPi!B%1-FC`I@+R|g<ZwgR;vi_d5Slb;a{<(g_K
zV`}^(yLW%yeH|@`3jS*eE$hO@W-c6qd#bN#Uii*~Sl)BScV790^zNSCc{`mQg(t2Z
zF$W_Q(jwvy9Wt`iuW2vk?$3R)sWJONJyQ5P2a$J&-Tl(t8=aHwe-2KLR~Zybu>;KH
zlxPKRibHGmAPY*|Q_EgL$%9OUo5<G=9mvnOdvR&;Pc8KxjP6YEgMrriO+;&CFpx6x
zCfM-e6<^!^15d5Mnbn!g9*>^i@3Hma&|bLZ(~s!>T;b*zZO60X1um)TsBw4e4l%X|
zKF#O^;>uE|V`7_D&7F^FM{I0UOrvHxb3lmLwfnCi2F}0!YKcO#(5?K^0&c|WRX8j@
zI%qQwJivu4unT_qq`lpxfCBgs&)cvw-u*g*IwxL>8u_Q{aOi#uZQ^+URu?jDndjy^
zD&5_fp|{EUKl@_i?p7H!|8U=th`0$uUC<<rK>Wg}`bz4^yj3{f%#U(0$gswst@c7=
zVD0XyMTgopjad!PmJp?}2QX%CTZ&em8pZP9tjwl=ZW=Y=*wq`w#q9x5$cMqEW|jGs
ztL@zy7NYOB2l%6j_^C;$x;xCCTQU%Ss;H-ma!Rg8smcY_;cZYE<6EUgGr>`T1-rEh
zfG@`R2h_vah(pbZ+cc=<=L-T~$DV{ENXwj)>_GRA#4pA!-U@ynp^C6&gaMIG&5n}n
zNbd9ey86~+?exk=MUK+6J-z*_i+~F>ISq-zI4%3M2cwosvz9*Y3|RjOMe7@(XhI%~
zOc_r7{Ywstz4z$k<)sI$#Wu5zm5ca}bpf$=NK~ZTxd!3in7x$L-)4_yKi<sMnIbb;
zB=GsmZUginZJ8O_VV+U2O~OG0>u1WU-A6Cs0_*Vg<++{ViIZMkmVfOR)@p4-re=`R
zmK`5?YCbCYb@oUFBka_<%^sWE3~#-68*NIO&0U*Gh;v;Vf;jZjhALE-*udPdvi2c%
zyJA?NwqGYoLc*_oji*+~V%XbnmsjAT4~&Lt!f`J3*D=vqnJ(`0@VH%`wtkZ+38i~q
zZs_b9vgJH!vvcX6>cBrXY=c9LAM>It^(CG8%-ib;aJ?El)VQdf(j64#!@p=O2)ujQ
zS7oVDe<hp*T4V<<tu9e6p8TfA){g?n5wQ&64i4x%06f<;VOx9k>EtM*s-nES4#1#C
z!@*9%PEvLC55>kS;lVv&LPq*A!#pj~sfu9ESHix+(_i`;hxA~Ne|fXrrMZ6c*A==O
zIH<NuVNr#y-@9=x@X-4d7eBtmd*?oWT^MMP7N4;Tf@RF@?-srjB4k~*h(My^viYT~
z;fupV;s<!{UhZYDhPSV{V__U&%#(|Kp_q`U79re=a<6L|yFwKOv~Hvzq>n}gi?+yL
zKG8!!{uVW6#A`-25xbo{%lKj(v^s3Aj(T}<sMUYJDtuffUbntkc{64$xm_%e9}p%+
z7;WIltITAWX>bAyje#3iWwH<*xeW{a0uI(qTQAXV-m;{M#426u56GGQvO{;1a9*T}
z>j<6)yd*0SdwX>67OA4jgY;$a#eo|h-5(wld|&P1A9v_bNX}bfDC?}x-qA4V_g~I!
z8M^H{R7@b8e@w)1E$~F1iKZmS`0+^Xm+^fj708Iykj5$avnlq;8WGZ3CW)7<oxqf3
z(d4t=&&lxKgvq|1XzM-IoK?ic!dOR3P<t@!w3S0v!h@mUDZ}GuVO=>o3?nW$N6QQG
zAFIS4CUVN8>2d6rd{txH<T-u3${%9UTJv4qXoa>h3mP1+{v=FJtDN*M7|0hT-khW1
zZvQmWDKY^~pEdLC^VKxbyuBX@#C3$sC}x#ucqHaj?r5cd_9m*VE~`Y0X-%)$t0X*F
zxCATZb>>K@H-sqFr1TSchJXl<jC5kAZi&9!5F-E>3lC?JL7i~MrdTFl5K==H7Z$<-
z2XC~!LnkLIcQu^aJjwF$FMp{%Xy;|ExncHXbk~gTKiAjvbA6!}zNdR9H4ZM@H<|WS
zrVAYHD+@j5QDNRVwe~?W!^hF~sFeyHx$)|p*)d@B)Su$L4+32OTP-qrz+ysjFlt0d
z0XrnkP^#xAC)(*bV|r}wQ^8F_c<t<{V2pwc>FJD;>TTth8S^4(Y&ppwS&wJwj1E;c
zzPlq<QIgvTCmOwCwK?-4>MNe7H2Hp_`8KJK-++Xz6#{?C4B$gYp^)#|>f0OYoR)?F
zAbcr>v?_|c@9L4tpfi+zO~qGU_*LA14S-@wPUM&ve_&S=%%}sZEqp5(^7bf?og<C@
zz{=Rf8UUou>&T$KoNXW-pZ<3_wregfhWUmr&ud-cNa9H{KqSrd*!~p*AJz|mNCDWc
zHjt<r6zKUx@-M9cc6Ccy7z^v#xrx(dsXT8!wxe~&cCxI?FG<pz?8Mj9KxxtYiyjLG
zU(_yd>v>db1rS4;8F!LIs}VH@mo0rmgO+<V(mPV!z1uG|7jZt8sxHS)1e8qA@3GC5
z8(Zp*sq2Loky_8x-<_3R6jDHLnok(b^u1vH)-&5NB1~J119JNiX)7C1+!xX9DtL|y
z6I(U7w573yry7HHuHEi5d-7y#l&Wv%>{ZZHpn2-YCo^h7h4GZH^wX~T58SJqN&aKO
z+T^s<+M#En?{QnXea*%3fDP@=FXDWUU+wcKaWcJQU_Z(9X=tTYpwdRj-uT^l;T?7w
z;s?VUV3k-(bHK50R>9>18j)*1sy+1l)9MHMq6s^_CaNKPcOt}TyEMO;+e7YK=qiw!
zH)HzM`+cSh6SqC<<Iz)4Ghd;FdrGN481I}^{m#7vl46$uP0>$yvR~iGtEg7$B4vop
zif}HJhJU%}z~&z&+z8S~aed|17mP9JITz~DL%%+7o!w*XKt{KU|Ib^F-g6TMyR@I2
zQGwZ(5#@xp?DX*n68$Dw#&-<f^g%$9&tlF>f0!X&9j+x1-*23G06C;G-t@~S2qk{P
zeL`%N%nr|E?iBNwxtszE(b5{Qa3Q6#)d<YT)h#sC%7is+;Q^5KViA!*O|~;HmB+EC
ze|_ANq3uDo9SL0<R5;YgWq8x@mH@vMXGa(0d*`-(+Ip+sHPm***1@66IVwZgZrdH;
zs?V8eoCrd<ws?K*4zfDAR=>lu1m&3?g@+XjCOs(hv$^rM$E4`3)@k8uPIY3$PoDP1
z-p6FMopg&GYdb(7R`Eph0CF(>m_)lFsem`gWmGfn%LR?XEmNbp9!OMN`y6`*d<Atp
z8X1ia9;_y2PGnwnP^+{N4)b;bbrnzev_tsX-H3<~v+z8hqNP7ds&4=NiXVzkw4FIL
zdhnMGm)3>PMbXZWdc7OdWZf!%;+e})?-TYy+h6kUymbzI6tNMvHD9Hz7xHf8&FGs0
z2}RE(OW$DQv=a5u%%{^0Lu*ftR|7cM<LWvop6a2?bXry|D4E5LLo#`W${iN<_5m)1
z8T{xc6O~Nm))gRJ10w{qy}#%n?&PGrLTX^9sVSj7VVIp9_G){M#EQKdJxwa-=aP-j
zP>o!TXNpTWI`umhTi*kbcj7NU5Wbo`_u{?a``k3mh_{~g4RgnJ`U4@ueCgC}fM9wD
zfzh(OFRvrbE|f=kd-GRFi?9Z2Biy6PLAmq|>o(c1TYf`syqmtxo~Vz_5msBbneA^b
zECuvAf9~|4jF`>1kH!7KEhQl5OyNm4=bHRw5ftV;fGrk9?VYtrn-wjpEu^ja2gFMS
z1U>kd9&K3Vg6%TWmy#!P{<{QOi%c72wRf`y#_}mDVwMy?5U%A~SAYBWEB;^l^M68V
z{8xifUUqzHiqPwqPvncLza`ufZthQ{Q0wIDg|k}zAO1^c_Uo4q1){pBWp|fvW$g&I
z*{Z$_qf~__FR}p7l7V^9`=PD3DF^8z!dP$*u1eWvp&(836vi)9D{G@i`JD0#<i6cj
z$O!h9nO@x@kBI4LlJ#uVN355sOL)qp6tK>kpzTiGsTV$kS5xuPP*nhgJZSvYdUFno
z*RroI=b3hkJKSSn<9r6zRw&-DhD=xkazH$O=BatuR>E~9?L7y)a^h!BjTCxEL32Tc
zve4~*RN=|F@I{sl05rIg&%8Jll}ZWK0`#-=fYD3#jjJQ}6~3DKn=XM>lE5TyiPMmr
zc!ZN^ia2q9KYvF}e<y>@)t0s-cZB!l1lyer%<q93L%7r@kEw0gx<LHL{_b|3PM?cA
zR$kUU`wDe%6d5n9Hqxlo4oW#woPMhTJfd{-qyLqFcoO|luU}&MVP5}aRQKmIU&{}Q
zUqrwfS7ORdk`b1$_#^jZ2en|R?eyT@pD}lOGD7mU-8rr<lT($ps%quW7{<Fn$dw23
zfEjQ{)*XVKI}3XeZvGDIN9wr$!LmPpQ##k4rjoFqJRZJsjlGWUXx~@|5AS?$5V`0%
zX0<jdxc+?dyXxrFbmbMLlWoGosJ;+%>vGJJvjQH_ZAt4nb4A01y!imBaaV`r^JY>2
zt!gh^UBHrz5bkhc6m(JQmZW+}H>{sMq&e%S!zLI>!RzuHv1mU{*bfV2&n>>ftXzVf
zA?=vXqzJDzm3oW5lm;OKMpVA6yBOK#fR0waK&$924XyVXXL}{cx0Zt&d(rKmYJBlv
zGiyiLhLm4=PC+*^xa4x5$~ghJA5e2Ks7~8Ei?RFbYLO9|l)G<7iGnCsJIUlLi8i99
z<IW)yfJR^?K5277CU7XusLtMIbK0*Ay0jHcDWMZGm|Dygj|Ssi9kW-Fx0%8%zP6v|
zey+DPogg(QtmmyHY4ph>`VyQj<;~PdH=}uyULV;~3hb(``_U8>^i=%ex0l%=&nn=t
zZV9ehPM2AS-+r~wvWr+!sN54XcK<qrc6XU-<IgU~MIOzaCN=m0io76tUAp?b@{=dE
z7|48AV!t1)s3^a0M%J!^>#~$4dt&D<F(1>n%+@2WYp0NV&LB`s1ke8&#9{F<aw8-0
zFQ5GJtF~)4F^LUB79keDeB#(bSuwm5o_xMH8O$+0dG2+nucPfv;c@55h{Y+NpxW?d
zXAdPS{1*iw@B;c7-?<vfxKa!<*S#YR+8`WRgC<Xpq2!iylPxxEZlUhpLJZQ~m>mXT
zbVzS4k44mxn=@^-JM@ZX4a>U*tOqbQa=eQ<hw`_z!1~zs7n_YSqQLkvVhbT^xDVPI
zb{7lOlIP}nN|n}LNqdwx_7Qb4=*yfykn;JQiJ!$^h|H6AG+1_a$Wz>_pWZ?HEqe%$
zAy2I*L0>s~DfS#O6?r-1GGY-G<sM<D;%v6L@US-87JqWV095r|x$|v>E6HbV&q-sa
z?X(i3xNnw`)@O{S=g|n=4cT!%n-xqgo|R<=0qkeK7EH^c-THL8!b6?X5_z%>wFlib
z$)7}P#PIi+=5+XU#r(J+uA@~!jkwUCJV-j-=qh)4w-QW^!0|Q_C4(nU4BguI_z#WX
znz^go!8&SCf2&yi(>c}DG*T2=YZaYAE%F*`ssn^``_77cmXzX67GZU@6$MVWsK<hq
zq2spwPhU7&Iu*Vy9o*z;r2g{B^@DRw%VE3O!?9FUl8+8FXI0gqA!Akbrb7|)Y#UeA
ztbL$e85dX@0?N#t0pBBPN%y80FUd<~I}KX`ao(d@q^BQZjn%y(8x_XcVU%Md?q>u}
zp4S1>--IVTe%yCr%RlcWcu?tPD`>+kd(;=SOX;H}&kR6FIK86-%40~y-OydJ>R$uP
z>gMe$-faxmnVDm#3|P%Qd}?+rAK!)>2v?C`#S{mU6L0OTb}qh}-HxxgEh}D7QY2bf
zvrm!KBSaI+KBB@6R+JY!$E;;OfK~RS))i>qRerF!R<nq2*{<!}Xp~JM{O15a3k0Uq
z@np13inX9s>h1V8$_Rz6{yvKUznY8g?c^556(4nm?;Wj1^HunX=5g`V*hRs3z_`G(
zfI??c7|O)*z)8C9Y_&Z_cY)|tI>ZLJi6)w=@Zc^S#4G@l<@+G~c3<A-Ny=jW>2OQ3
zUufZJfA5Sbb^mw(+@`h+@fax?2Y`u5dSR0kItVhfO{z*92Q&!FSrHHiK-BA3J!c~$
zFl<pOJ8nMuAMShR<bjmBZQiwW3Y9S_@!3!?8Lm(5R`{V&rB6@PTeH1*r6PNB8fCI6
zOxXJ@(0H&~HLcU+a$$k-NDD7^KjVLJ_oh)zX4~Gc`>kD8St@J`K~V_X4hj(vkUk`5
zqXp8Y1f@wJDQOKMMEWK)wMzx0i_$|N91DR$2oXp^8z`iS2q8+7Ktdw~f)EIV5Fm8k
z$2#Y{W1n&F{cz8hJI4DA2+bIb5!SQTv(}vRKmR{WT>g6gjDp5dP{Y0*_T!|e-F*e~
zo4fDA2Ixf;SXH)T;SPd@^nsNA`eW3d{)H@^xCDMCq4oYa`Qb)<!5LL_UPxPK^;F<X
z-wxd=LT8xtHXU@{ezRk8ubib%i<J^=PaJ5?lx+=+uxRUn#Vnun(mhGwDj^jFUpxQS
z5-L<oMx(43zv_aYoLWuJ!|8tazWRPN-OI_vtKjc}pW=QPlf!&S@d>3Xw2;U_l{V)&
za&462PUw!&CT_M7rZ5AaOY-wx-hsGZT0Z03{V0$$V6MG7-ren!#&;!4TC66tx1IH9
z6YO8n>ACafmUESUrsC<!tq@86_%<D~(9=6N9)s0E7DsOJS^ACWvZ6Lp$3-9rEs1D%
zqOD_kD}V2CFmC_94f~g0*b!{w;g*E=0|RO(RosYJ4Q4M?HdMRPZ-iVaNpQ)KB}Cuh
zR@;b^eShpEFW10L<B}bev7xcXmKhwVk$pk8)9J*&Lrxqh>^$wNZ{h9EKTLSH!)Tq_
zt8oW>Q895N{A6=Wn-hR9DncX8bv`_<Ew=vWbisSxd&Xge7pQC*hgpiM)U%9>YJS^9
zl&Gb&eX&Jlb=<<&6gW$Tfm1a(%Ge|o7U701JtX1%=OgNM#BIZ~SkniC16&a9t<ChX
zT5Nv{{wiG9m0P2Ve#8IAd{MRUYgM^eXS4!$|LM?AV^OtF#%{mw6LB|}`h>x!_irVT
z5}S7c1ylIb4`p{Pjy4Ul_B5A^J=flaw&qH4#PMA!A~0J}4zifh>s!;vDV&wP<_fW4
za8A3=Q~wdGfaTK01;Ffb|Ir{)2i3e`jy@;ffTaR_Q>D}0BrG~@L?uWRJUud3sDDaa
zRR3?Zsec{%zew@Pg6hz%dg4?xgZNb?dY1#~aKF!s%Ah45M|BCX6*BkEcMBVltF_m3
z(KiRikV1yVLtsl@mN{L!wEVC{!)PKl1o}_WB<#|VG#<cWL5x!8ZUDRzhiCi&h=FKn
zDad!AzP?zxR}xdt8#!ep=y7BsFc^1R6i_7ZJL%*d=j^p-X^u`gsusEbg$6%Xp#y@@
zX}>|A+?hEAm^J8SHVw4j1Z;kEdt>6A_4OyPHHzQz?-c*7IM1`0Xg|o0on#*?KlT8-
zM~A4qFLrnu+WKVbYT?1JUhnty67pjW51qxG-xui;)b_~?ef#EwHPx_NgUCFw*H*Ge
zhOtD}dHj#yA#}P=X&vsQsIhO*XKlLEtrMF5Nw$rU)-SA_H|sQko-Gm>8@cyZZMPwE
zw+<mJ#H|s-7yJ%s%ZA%T>9kxCGaj8V64+UjJ5FcR=qHfIjXb~W_^CQJ``vuaBy!p<
zqoLo~Q{+Z6bji*;msosvKHF<-T|Md7<#tmIZYx`{Z(Z$-62-~cD0%J%ZAWZ+6QLn~
zxV*GH*Zb&e@F33@WVVgO;$s;=%MG7Mj87iIPOX;g5%FhAini!sGbZs9x;uP5uN=(u
z%;Q)pJ2e%sn7O$?>i&PFGJKf@3PZAQ==<HnArYf;(UV4$)=J!g2QT1ZbPl)7Yno>H
zZc}S$36IS@M?L(hD*z2VO5*~u*HoW3F&7O?Rbp$A7F*N9yAz)}Z_Cz3rUOUnwqKkF
z2VS|&dhzz>I<p0fZKFpZd&&OV@%lJ~crq)qVTM^qKgsZF11Ac>yXT2!hNIqy#xRC$
z5l7gZF{2MTwk7OFOCN>1jCiuD8gCbF&$5)Cmc7|oJIzWP9U!QSDy!_<{@FXVAmZeW
z>uEP<UJL29u7n1T)ro)i5^kOps5(D3vC@Ky{i{A7QRv}PY#62U*wt#`Yxm~NUxy#(
zn`2N`KOMwu!kXftaYgJn?4-euTRVe!08%2}tik8~N<vc5gjV`u^IH7O0j!0fE?MlB
z8uD*G>uT#0_G3%IgO4NiMwro{<E|wFtN;{Kf<MQhVTlgmYPQqHMlll5epF#Gs@K=#
z{2cpsntu&VWbt2krGH&_{~y2qKRSH0u}x_AjC(#S_jOh7C*Dio7t~#wjZU5gS&_Q+
z+efn*XYbZ`Oe~TkZz}&btLJNVE_dVnl_|!-!OO9_$Q6@BJbd(W9&o$Mj*_VHcERr9
zuR_o6fb!D8v(sx+#ck)j<MNqFAqk0^2DAMw<FP01AUM29H?`JXc{xZAhR@0v-(WGs
zq)J)VbGzQRD?>BdL3qi-UC>}#jtb#M2b1Wol7Bw-Y5EGhi4o5~dhKxHRf&4{)-P>S
z69esiN(7UZI7TnGo@_mGZ%&8D*`wi)m7jH$C*{9-Pv_V8St+rdY2J`?5WTtF;eLc7
zcmu!G`1BKDqGesO1W^awY<wtPo(btX-A^nO7i#DH!mI~b&E&-S8~V?p$zn3rK0JKf
z)BveDzFn;%4?Vr-PM%`F=Ig1emyNHTPcp7+I)ir)Lqk*83%bfY+CKibk(zCA6af?p
zw4pu*=fkv#Z>NPu=I8f25Ld$C(o4DhQ<JE9+**j%cAj_*7xL;}OKS_+<pJAmjsedL
zg53xpp@kW)Qg>I^n~0<NS#)|nsqX5}W22*8=?HfOO5LY<{bN^hZg0A>4hm}^s6fNN
zROK!_F!b*yUfEdak8cTs<8m||cyfJ=;~;OF%A%kP|K#`X2M%;ap#Wg%(&oWTxPbL*
zckp|Y)iN7iefZual2x_1(K^(XRlWG_E<i&<w5jEl99H`i$8h1!I5TTV(hZQU=^qch
ziJ23#ZCb~BwiGnNI#?ZAC4vY!(kQ3{jE}kzy}lB(NO@W)$x*JE3jinT_0>&JXT=3C
z1+So21Z8Gh#^HWIx;e?)aE@l^l9Wt;@=L4#sz%h_KSlRx$UgY^79mr&vWeE*9~g9p
z+mfS1$Tl3w^h%Wa;!L3E8r<cvkf7nI^%k!_WP2{TiXVEnvbeo)QZ};Ey6HOVh@WIu
z$mv#>Ry}<+gW>H7xc*Hnt4Nv@z($Stp9lYF2-i-_|7g0o8BvW{pqy-p>$60Pt3)&C
z*+7LRD|nPMq=U!TL}w<tt3wrUOjaZijXa8|eL}UCFo1-KQA-RHlp0EoFH_=a_C!Tg
ziwa-6v~S|4cHR!#w+i-`ex4VtGP{h_^~$Z8;zE8*TeidK_5QP!waj(DlSw9p&CS^@
z**#(d)6weZohTuuFD{2;xE#9Yx$*Pt9_kDx%o|23{*kPa6PxDCga<oYAg)PD@<Gso
z-ZV%^2vkLzyn1!FELM~-0uNuu%le7xIte5wI6lkPb#UjP6OJ{yv7Hzk*cHs<G!=7=
z$-+w_%fP7)>L%ePF>wWU#pi|y4_KIUWSH0&`7_a&0rdT4<`OTVMC#vYeBUkZv|bo!
zdFJcX;<+cmVw7?qI0SGxHTtCSs-t7;s*LT}TDfLgpe#wex3F<*NjCkJg6D3qv}8Ia
zQ}-NRCr48kmx#O>W#Ct(UeQ8=+}*-MaiW8p_R+E8n<`|i%3cR8P$(9|JI8~9Mz$s!
zg+97T%H8QxDy#x^Fi%^3?!xMkD0ljdo}}|DTU}1aUpscvzuOwWMybl*Zuf`XQ=M%V
zH&4v(vjw=-{SGack*mD-Pju4T18{!)Yg#!YWl$i{jvl7iK((UE?p%V6!S1|{Czr&=
zszak4Mf`hP(~Vv0=kGNw=#qmw65Wl4MoeIdV9Pi))X7Fq>E8e;<J6<ZeKGWjyfo=a
zXKQ|)TAAE6T#fy7k27sfc%CKM1Bl$@c=)(9koWRUr-V>5gf3ssm;w80l#MOQ1ZNKi
zF5dyXc!4V>mGA+u=>a|#*~4%D6;1SCThCvMBcgjLE#3-kuX`kw2av^{(U1#P+sDGZ
zw%fEcSk@INZ>})p_7q<jf#AI^CVO}Xd4j!5dsjBRgrP0P5r--dUP%73?)xdQ$40$$
zn#;ZPtJu?k6}h*4w@yK)!Nz`spT1F1uHXdf{<ZOR)jMINiFB?@Ddi@{nj;@;_J)p%
zxx=fDUWqHKlIJ+nIrC#3eFQpi+Fk?g3P;-IMcYC1Oa#(rqJAN@m^Bo4MbI{aoZru^
zT#c=^m}N~UhGG;T^ZSfE-9TtgjjG=hdP;w|3lk%s5OJ#Qc}%6!7q779$6q~f>J!P2
zX#)x?mH3+nBuGo#dEl^}^1kVn)i(%Xz(Kam>e}`~%&L8C)8LA?n|@<|sDz-7EAR~`
z^bdxf8%Il{T6ED(?`rFpNR5%sTr-+U*x|nk^;fGVW7TMx@p7#t<KcO*>rpgdePa+<
zS2yGUYXTkb!RG}0Q>XR6M*Ba%&gnf|>^Lt9)bMX{EsZ_zA<#X3Tt9C*De$cW#(8cz
zDDvu0y|XLm0fbiB5YoZ^rNNGxa=xw6bCO;4`1s`AlnVH!qBEZTDxm-<rQ$i}Ars?G
zi(4eqr<@W_Z?Z1jk3Tg|9dXSRtUHM;!GsHC?zV39^x$CMT6t_d)42EHIrCV(z-Jr*
z_4T0$!aAroDJ-f3-zAtE37ekEW3G>AdgXw$vxX~EQp>0*{ri-SHRa8K0sZcHJYp`i
zMIo5=;RUR~1ph$(zQ)y=IbbzLI497n?yKd<Z;$q`DY+C6?+!^F-VRgzh61}aCUB#z
z(esD1lCSG4rY&_LZk#6T_x;&Z`l$o*7H#M708EN6;BG;6F<kyJGUmH!Bg|sgkO&2e
z_z9>{hk?t08g-<h#KH)uQKN3$3L_ftYeu4#f<kvzWn1&ooG{EP16v(KQ^pT0G_>qj
z+M1X4&l`ExGFFEi^a}^3C+EY895YwZVtxG$^W(g{I$*aGU%g8Sj|_*C@o6If9La8x
zWMr$$gTv!PNE8hqo%G-V`sd!7it>s_Ds`2Wb>G=sJI8As*LJkn-3bM{ZtnNm`!el0
zxn#_#Q|b*117(_y6Akk$qV8&ro>vpyu_z1~!0peuBE;c&BSsi_0M-b&c6PV@`~KTh
zP_Uk>`(ZvSbpE7o+D?um8@K<cdA{0*#zJ^CkhZTpYpGr|wiA47W3LjW65(GHoAb_3
z2R<h;0UnU%mIIc@G3mt-?1y4+#K>XOK(xZ*kYox+ei+v@xnQ2?YV|NkbiXQ<oOIlu
z-AIx;o-2+Umtw`i^S$JnN02_R81l3rhIQzRM`g~*jBq$C(%>V9OX<{ffC-?#Q7bO7
z8vNS10B(iuSU}dn*cxsTD%AgdDQo1W>q^q!f+wRx6K(+1XYT3MMhiFgvCEmBoko@&
zM+H}22}LgDOEuAi`gf!K0)LWSSiZ@<fbF33!}>_ZWf}J=K%n!UY9_G>NLtecrZ3$a
zgnzmX*G9c-|K<|GweUSN9%$YV^mT3}x+0t_`{Ha{L{X6P@m<m&CVU^+_k?0}7U(K2
zaP{qpEo&3sr#l6YoSMVg`<E2%_xM%l3e`QlWhVbQ(8mfsrDSP^?cPx{dGTyqbMa82
zZ!aD7BJc-){W{8Pk5*(+;u%8IPJhDdWyo?y_WRep(=HP?_Rb-j`O9JM)#so7%<Qq9
zIQSGuK_v=h&EuOhqbDgv#4Xflq_G6EaXwmVL~M|CMs=#P@N4M)`leqeQKt1tIF=Bb
zRi75Zbd;sgWEucG<@1sJoj5#3Zu-0*k5Vw(#YaRZK_QgX)HGje;MmL9EzvD@ON3rV
zOvmO(hsw^R%G0)Pf_a)SqwLHZuTnQAYEtC=24=CbY?`Fx`gJEfu7d2A5t;FX6Oi{r
z8)xL9rbU@__!XB5{gvQ#sJ?*>vr7Mk#2UUnJ)_kjyBZwkrF@;$?n}ZbzN@;xHHs;A
z)mp#?Tbvnh|M+1(`k>R$zrb1MP`z!&eB=-=1_S(`=0qFxqY&IF`G*=@#K$R0A(c6`
zfw<vfC4YTiy86`iR^*q-oOrM14!s;cH%HUK<Kyn>xKrRSslUH3xA&~7fzv#N0OM<^
z|JulCHF$X7as%tei1ptxP#%)`f$YM17rCJyBXo|L)rg;G2zBgT@}ANt7)Li+ts~I9
zd@;aB!-VRlIb!go*8mMDY5yetyu4-00sNdaHf+}nV#qO(4}ZjnAh%Z8f3wT$V_T^u
zD>|?y!?L>kBb^QM@s^Wk&U%KZBF^^*q*UTD1NO@3zU!HGe)*7>*`Gci`6fw;^qVUR
z>amGxL%@Qx6GCp5zGsfcOzu$TFlFCd;2W(M<IieB>I;qr=TdL=28?Xp0aL2Im6C~E
ze4`}ENNo_MK*EH1E<M;s$AcqH8eBdfflW@+3ikspLh625ogQW<4>cLvJsb(&{=2nZ
z`LV#!n1;T0@b{17Ul^O=0IV(ef(GT=d)c|uD5Qt2&<=EFrDmsU;bJ2L<!mv^1M}Ls
zI5r8rxvy^({TT4`*Lb|A8*|`lAiHxO^*5`pkO2mXh-t}=Gv&PfEx^}yCW^Xvcs51j
z&Q>`XdeyDUKJkZD8N<BtaibSUAXd*j=fd1;S!h~p-9DAA(fD12R@v6ew3nR|>Jlxl
zC<dPJTM_n@(7?4}JElLD@TT6}$^z&wOq%c`CwX<3Yn&<tVNZbr(Wyndyoi{x;G`2h
z;o&%qUeB_R;T<WfDltEcO(S%Tz&UUC(dP+{SDI(c3W96U&M!Y~<d??ns-G0XA)y%;
z&W&BrC7UI9Bc5J<Jo08bN~2L}ZsbvSjtRkS`Jm1~Uc}_$%T79SUd9&D!3pVFs1nkp
z^$M_%dZ3-fDNLbNh&rqAQE3Y^GlG4fU8t=unB|jncPKX`=1{e8izzTk-F?~iXs>=W
zv7c?v|8?(t4>&Kc=9*S=3ngS6P2UZ(L3wF4kKjVCZPfrJKFtl%qN0sQpMCZl$gA)-
z9#u03U(3pe-|_W-hZJlhYpU$F4)4zIoZOkk;vh32knp3|CT%dccBEbiF7?M6qf;ui
ztUqS=qCHlWfZhYK{<D54s_i4ccl&n&D4y@KRmbw(!KPi35c;n0lq>H3eB^)W#9PNB
zEy%OdbVk=|(eGh6i+lA?`i}of4SrJLo$V?9;SD)cEGx)p@(_dn;6AZeei2z%5l-r8
ztU1rXd$L{2k*6y@AL)j5xz~I?@;$M5Z{Jj7Pib?Ds6==sJN`Oou}ayP8C%(>z^GVM
zU<b^H?uEHL{4{e$okHrol(P^cLa9yjqs#8b&Db*IEzLWBkQ%|vVzA)c($pd#))k|M
zDzj9jO5y<(R@@-}Xvc%}nBNrR!7#k*-0?<s<iK9U!+31gTUhMx>y-s=Qf213$;-vD
z*VgnTQupdedT{RVZ1C0KJ*z(J!#s!dRuA;`!q$zs>R_w&we<pE*>$6oy}Y%aWZk3%
zi)<7y9a3sn3(wq>>J~oR-FQ}%o;<e}$HDYS%w#l*mj{kvxfin-_I{u>?&*YO-ylDi
zCkH2jNe9SFE;zzXRF9+E)=M6?vHq0NqOT7whKMw<QH#om%-SUW^zVP}0*VPWL2VF8
z{e9|zzW2~7AK9CPOXJE)1C<g=*#HkAd{?)AJOiTlgmd!}J`|jA8ms$*8wH$**6tN<
z>@$_PJFG(h`nGd;CRt$IAP)0t(rKYEL)#N8&&dZG+u8uG7QOC7B*BzLyaFiD8XrhO
zu`(Ln)M##Ol*^*g%C3a|%;omxe96y2bO<WfB551QksUW$QQ3R!IXbI+#glw9jN2Gj
z$*uDXk(^hcHTY0o7wv1hy{jF?g|@`o*gv5iS{!nyjjqkqwx)9AliN<Zc2|l)KZUAw
zDzPQSYTU2;KZS!$wQEwsGZ$hQtb{|$XCv&}{f{sX(XN`_SaSSmF&`AHs-L!5c7xh;
zkYo}`zabTl@w<rb035hJkUP$dH(y_PxkCpxH~fgpl9g3d1I|y0If~YsC4c`ij9*QQ
z&UFm5t$29bI&q~4B)5JZ`;fOqtMk=2@|+aHcWbHFB+Vtx7EgM&tPhu|!^2H9+6y;e
z57_L#{2_g4p(RA;7lEm~E<ESM>`tP-yUF&IcN@f*8)jlC>}zZegd-FVIl^~&J7}fg
zqK5ejZ4W8t)=eS`;R`gZc{{4oqoNGmRn+6pt`SXwo61fz(^br9Yk0>UGDZnq*5!}X
z+48ZpYLDb~d_K|^M$hZ@I>_BNw?aJ*9~YGjsc=u8_pC}bN}O3gQ_vv7d#}bWv32Xd
zLGkOG(@ZuH^ZHk(=k=C88X)+G9F$$NqHDdZ<EyIhsqT?XSYnm_O!I09+*00j7~vMR
zyXZCl`Xk9TxhsoayD%TCcLo=(X*WRrwRm?N+kacZbXHIw^3V-9deY3MaatM0lQwY(
zy{mIugZqrBl{1}FTR0=7)$T_a&=09~0Q~kr*I!GCON)NH;jN<|ELD5<plS<2W6cF8
zb35tT+b%rvsxYN)@6Kdsq<$nLpW_)&VU?8K;C@EMpjx-Xe)^qQi9hl-@a17soTn(|
zD`f9dO4uA8QmT{2=M~Gb6xKtzA#(&(JZmCK$(w6*>=kHC%R+s|w}za&+Rkbk@&>m+
zHEfx8XNz<)pfA~z7++FQi$Vm@cAaEV5|(UB2jqj@Xn>DOmHXicmM{5uUDY37ro{ig
z{`z6(Voa*{CWdB4j*c0>Iw511u7$#3R}sib8!wxxA~F(#HO-y9EVV9u$?LFQ#;E6u
zJuC7ni}F}*Wa}-da&kWJ+x>)n!;d@v+)N73742h<cFUH89gEOW!;eg*w!u#$>S5AS
zYwRg8ORa<c!}#3mK}T-s$a|(^@s6o!l`e1d5d%>b<>+B*#4=L|+L4)`o1URz{EN1h
z!%(OS$ZGa9EB#iFbEOxjczUP)#|V5rIxydJzSy&xbxkuL?R)HHvE7TsYr`x>GaHXL
z*E24CE4R@ApKEh;3cx%lq&oKedF=r3*IEGk<C9nqwSxfyA%Fz*lwe@6iyD#a{_pYl
zpMLZI2dE2puN$>VO~Id!^pQUwkq?_iL0?D#=a7Q^bV&6c;1rET$!GQ7+PD1iKmD)%
z<L?wV3?9D*7|PZFYW6Gu&wG49+GmiFNWGv)(I6|<-HMf*0>vL(yd+v=&z{n#diy(|
z^(dnrJ88Iw6`9_YatesXx!c9t7ozOnRL~Ahzlzj)#n<{}_HFmf#Jp5RX0=edcG2yL
zSCnCeWl_W)es7{SswCh+;_*aQ%VOjX>fYf+vz2kP5?NQjjN)LTQ-H!e@5Y6AJ3F-B
z+HdC-o9nEyDV)$f6Z%rHD{G$HbzaS8^1ckp`Fx}mg`q?Yj&<|~02;|%CmiAHa#q;=
zxz2O*14_bznKCoUAVN%I91zWxYD$IzwlaoXeI|@lz0-dlLQVO9c{p0H=@t$81I{?1
z_~-8y^hNHBw?Ja0+6rI82g>FlIxs7HnilB+OKl-^KPu%^d39fCoJD+~DpPcS^n;-J
zAlL$KVNtvz0yvnYN;){knD*M296HPGB7m;Nh?L1bK#!QMNibhEEEo`_mq9Rm&EN?N
zkb<3WXHkG8ED*{u$VMi#N)B~!aF8unF6Ey6^RDb28Z7A9%zBHE^Fx|bbcodq4KWdR
z_GII|{z*X5FG{sG9L;t}M=xaJEnAK5(mI(_b}03ndNQ*60wy#N$@ngp8@#NJ>WCD>
zUPVK`$yi~87jpL~SGuL=nwui%^|9M7Si&_&Uv|o6cRl*hdEbfA@k*=A;PFNQGn+AO
znLRUS3RqeNO!sApX$LzdkqtwD!hgr{#~LV{2OiP0jIFbG0!sJ!g=Izs<^_4NYw@hG
zB%`%G+bcN!*QnS$p2;M+&;5#9NzUEkCwWwi(|2cAoHFt>%QL!tAMkttyB(rqWFC#0
z?UJJC@M#J<LTY<4+NzJ3v!Hkh(5oYJz`+L|N2T`qYQ}f!K>|3!9*!Xvhebhz)2JwH
zEns-eE$eX2pKl>gP7-t(jc-F1%gjxp8W4JR1S8{$cy<%$YOI8yVpezU-u`3++2=xE
z8hP=T|Hs8rIGwsP6K(pk^U|_0-7K8_t$@+n^u9cIBYK<mu=yRb$QRAz&*TZ26??A?
ztuwPsYFiE<?>t7UHOsAH4>n&%&uM%-{=O4=c);?V`+TH`o5YhBB?VaBbVYn^y0s=u
zjP@6en3;7%&%_c}ur4vvIB+i}B&b=mAza>MyYB`5B!@TF%P}d;Rm%L3!<99&P-g1Z
zCvjp_0_@agrMa9(B^aG+L~pcg|G4kd>OcL?gAsmb3X%+F#a_7@x#OCP8UGknGw^BP
z+q)GN{)rc{Nl7n^?JzKmEwZD`_eq{c<^aaSGw_M_!+Vl%*8d@@>K@yYJLCR&kRRct
z!l>UKDku0ppIsNpr4CYFlAgI<Q?`zoD1V>Ci0Nl1J+FAT=IA8!*qpf+R~DY(VH_pM
zT4=a+OQM!pof5Duk!6BA5^ZDw3Da-8V%aN%MDKV={FQr(1wZ`DX#Zs&2F|v-A!S|8
zozFI6gC&m}>@L3sJQ-rk&su+P?O|h(5&z<gUvPXGt#;an_LN2jBFuUqP4h2i!O~4a
z&A++@$Hd<{cmZmY;TEP1MUKtR{nFB^9)Y;QV?S4BhO+J-k+~41|Deu^?A)XJC)_fC
zBM4$RWEL&TUS{ev{|vNnc*1h@lc9~U(HZw7t*y>`T1mrfSU8XvhPPSJDEJk4fh1<1
z*u6SD`Y|%fK)w}LvZczJ8>ipcakh0g4(If4cX~DJwMd%}J2zqv14zpVO>hp_Isdj)
zqiMyr%(4g#0gILq6N^)hM2|2x^o@d5jfZ&U$)SIv^jkfh9~2%C^D#x&3kTA&$Cu-x
zAagGpnTnM``PW}O=1AT@u`uK$Lxvp&^Xv_$h5Ck`sv_h)cJxc+(>Nfv3ulfo43|ri
zyxte~hTO!dYxx!}ngBAO^X+R!^=tFyPL!c`L1_u6i;UnO%{uAr!AyRB2Jrj9({aw9
zk66A9m|}rGb!P==*VH(`%JQnou(J<+cWuiGJm_|WM1)_zQLM;XRQ>VKyZB9>zkh5F
zgUa+J>ou64$Q)zjJ$aJHK(r0;2U;8M(5O^e`4xwBVhu43fW=uq&KV=S#R5R{xSTf%
z9esT#frBEw@%|WRth)fx*VhNX-$i+JQcE*fF7+#(>rGd=%GY!((uKR^(a1rZMtz=(
zTpu9%Vy3s!TyksqS2o~2G&0bmLn9lVg7-KM4&JgYdpVKP-8eBIObnAB&&8KU5<lu)
zrL0e!2kBF*Yzj<5Ov2-?4mjxJ@HTVo^w1_i6?j)3Hgc)O_CqMf-My=iR_SxLYNurH
zCkuU`B%NzPc;P{{vJsy>45?#c*}(;$yg)HeYiAtvj?c0AbcrLZNvukXF(dj6AW_fc
zvD5(T2-rVbLIUSEyc;H(olT}Ea|P9>)H=qXXESK+@RAVA*xmhQKmM*$AWE@#0iKRq
zfAL90z(qhPUb2DPZvYOm@tW}-Ba9P|{CtG660JfdUlm8GrK{GN>@gbVN@*`!)at^u
z3ks}o{^-ncfF_dmy`b>d*+hemOFK~4#@(j$F@+>NJ!HRKEg$4+44i<k^a@z<n;+>G
zJvNXtiYL?L;X`$08r2oxC`UQHd9W2Q8}h$w!nm<y>7{J3+N=)tO4Kmz!p03*t(;e>
zb&Fc+6!P6CsBWD`n!!cB#%QltQNd{&<C~VVE8-{$c{_G>hP=6kzu6r%GKqX&&s~|X
zU=?#rP!`W#lzh;A`M52YGt6@4ITmO*Wxw^Stb!j3e&Uj$b~(mn3V>jwseJnPSwXYn
zUCN_jXK8-(y8rlmBk$mN9<PigFG=zlK$8><2NAk7%WFZRpx$b7I&r=c*6|^_HBNaV
z=6Jy5$BPH)PiuJ+N9EK7nFP=h6YV=7+s3<Yh7a}!<_n2@_pcM5qtud3FQj(1_bvqE
z3}PK4S$J@=<pnJK-EQ9o@8xKnuD-6GUqvW3+?v`rLbj%vUmXVg%3#+3_@rMI0^=Ty
zIw~afon$!G0<jJj<Ik+8)IS15>z=OzdNG25^edXWj=9uf*`LSH+JvQe_YQDw2e@$c
zMu#N_4NxlQ#u(UBLA*vKCC}8tLJ8EPAbAA>EF+D{1{U~zwp{=zR`|GK;F@NJgk)sn
z;c)H3**+t{j^@IjhUp86Wukq@UW^HqL#+tFqKkWd>QO+Z8tAfwG#br;ZHsmjza>B}
zUSgM>I(2H1oy(1I{UfJ^(bwF&Y8k=^=L{*gkTQcN|M{*E{G@V#$Yd`S8{dH3pg@f7
zk6&`jw7<8q(ekR5UX90r$AY7Dc%$WL@QIx#dp8q~Z$edOPQBTPVYsdp)@0|35}?mG
ztlkH~+(EqMnINWjhf!vlA6c?E^1e!c@aO(hPFY4rE&GC|riq`}!K-e}dfKm3NvYHx
zK0doUyG^y8%^fPxKr_prg@wiQEDIi|R$hr8#0RXEo?V|4ok<bqsE0p}{MIW(6?0XE
z&(vEr<Sm$MMOQncud*F36=zRB0XrXcNQ+pW2x>j(F!~F|>T2}8VDU!dl)v7?<C1SA
zH<#Hx1yHrBB;n$jV-@xB?cdnXX_-{HebkPjTg<pyPlO@DpVKYjMe-~MP41Oq-C&^4
z#`*g4{elD27m?G?t1R{*ls1g+#kP`4c9YhH76St#Rihs<XJ7aJrrA2KL-<<r;)sJk
z(XR<-u9O+GT@bO))(FebsR;uSUi167#?@jEa8&O<`|l9gT0H{8{B4uQdoQ4Q2t=yu
zlirG0^d#b51=8-SgMS<|XoaCq{IJ5fS+=mthW0IF@YLAfyv2r=WC2I~V{$}XQK)~K
zA7OAH4QnR#1)ehw0$PnR;Dw@5{w_%PVV){$qDp}&kH9H_cMNR>pp{r1pwlXm)|3kH
zP=f{?Z@ucD{(K|=;W%;VIS)z+IlXypa{6$W^4AS#;E8{!V&?Adt48>Py@MD(laoO&
zhX>9@6_Q(>3N!%YF3+G~p3owrVPt^Y-N|bkl0ZX6s8ir8d!WbhjJUXzF`BWgWi{?V
zTea;Q1k+*?r{htS9xyqJ?G2vo$_i+QY13kfwhD{u+tu~v(XGf)bU<0;;*)tplxD@E
zWNS&x7Y1_o!R{)f!M?OrP?WKRIb7Z@<0?cfZ^@$j+CLxZ<>OSuZqKT%zyTjJi>LWI
zX#1+xELSfU!YTAu-owUw((EjpSHHi3@hfYzHWGHD(LwJ4!$M~Jrym+<Y^gJ377zHL
zO#mGwmUnW#_@T|5E2&KO_~M850XEc;qFqznVm}BU*8x<fDa9YnNo{S3P@;g|Jk=^M
zWY!aU!#G;ZK`zUaezL>}di%>EUAxQxG1$}WgaZHa=RSBugU0B^TWspB>TqbSG_w61
zPdHt>se^DY2h)DwzJQku-z|sd+4)yHlg$RK%oo1Iv*`P0KwchF-~JQ`I2OQN0v;lM
zFa4MW>T{U*SOcD}gMguHOw(0Tc}jH((7eBgfYa06Mi6d3W3&L^GjPz+F~J9jZ4G1q
z5_IfmP%8eWdV@&uEzi`-DV*o5-k5pZ+@BvdrrYw!lWDh9_3cM<6^7+I{uLej(Bh_k
z#6n%jocQ$bzO^P!QA3~>gBHN1|3RGTGIyp;vFc9Bu-;mkFtBJgqNd>G2H!*)n`K+;
z+i@b^TZ-s~$hqMUe7a3LhM3Cl0*lC$Xh#{lF>)rSip6TvP)1>)W1a!+m5fV0wr^*s
z$SA<w6si-|LdKnDI%S;R?Hz~<ocnP!Wc%mpnTMuY9^P2$wHJ8O9F~6Cm~@X*k?$GM
zT>SZnYfCHH#O=+nQU8^MNB7#tBa1LF-DBadT$xMG(Ztqx-QAIO$y+{Z6j@Q9{-mGu
zFm!4lSzgivTfiC_WEr=Vc%aV_&41Gkj>j21%2=q(quagDNYpoMtKJq<Cf9p+kT&@|
zvX#7<MG5KDB7-hXt#N3TVZ$9hy7lDv-NQh>lrdwAKego9;T|;|F?3_hsBPxLYMhu*
zfl*NoBdTy|!fV&hZ7HeST+hc3R-&|!yuNpjwTkV-*huO#Qtjra5!qdmU4U#yp>t@f
zE|4Ia3QU>b`^IV@c*HehwmCsP%Qk0bqTP{uv@*XE+@`vL$6#2Y0xBnuu{xJWWg^1E
zT?rg^8>8_N%oUjfbl4BUf9^_AgK^}B^EC8rsR1a!msJVu0Dr4ViDgvQS$+uFDAur{
z;+^3`zyd*s4o*G~Eb0mErNvs^Eq8Sp`GM1oB}O!`&awcRQTwRDrs#b>J~**Et2w(p
zY;k>W>!+>6?iAPh_<!3vabQFkI4hm|#6Uo@)*cLGP3tQQV<!u@n?qDMsj2kJWCdKn
z_=Z#13gE&<PP<ePy7S}%0QS=>twO?#127?oXO&)X<w7YaGV)?57HhkdQADY^T_`;(
z9rn40Tw$jwhPEYSGujrX6KA5Y{kY&UT2+ql_a#f{5D%DYwd-pEE6B^O+&<3G+8(9h
zu7uJJ*S#7(Z+Is9`#laS;fEdUCiH_75LcAMg>ucTCQoniU;B?_xl&iy^ov{LE0eY2
zaCh>mN$1p33jua|TE^jYXC<W0jN?0lR?TsVxZ){(Z>;oO<bcI~qM1~9vDo7C5q&q!
z&fD8V+5-<#x2L5_Mk9F=mw!L<)n+-I{2jICeT~bLIOWS^8=9)H$}JNfZXFkizME~x
zcCIjA04U(s{EIy+9uN1OdVFz*p`BmA%+Bp6T+wwbT=SsmJKW-oGO$X3DzgZH)%x}f
zI>O2n$BSxvYmAh>Tys9YA|)jak1tuPL=EY?nY-Yi_<w87b5%{_wq)<+HScLysRjs;
z4|j%c{$c$t*Z(oyqjo{IdQ#*;uW*>Q!v@x2V<zuuS3Q=67F@5o{451ozyHm?NgT_Y
zJOnEjZHfmP&lIjg%J#EpZ-?-5X<0ad+m%BJo4_Su8x|MLu;hW>1RB3rOs#W=%W&<H
zPnm>obQ58=YCVZCp%s6}>bXE_5Wo<%Q9K9+_8%vhX^oIT_4XJ<&I<@s-FEdrc${E<
z6}|g`B}f16!NcG=$YD{|RxJ(XWUhN-+P*l;;>{T@tuh}02F(ySVeantKKZOSu9${C
zmR)a=oY@ME`|ydRVA~k!LAcj?G}mt&Ala^78^@a}R=`Z@iQ%t$x9D0Zl$ulHhMFE*
zw<bjaOvK9s*Lye(Nfo;j980J52OCOO_=17Cysox<`um9i$VKhyw=qmu!eWoB0>=tq
z4V8mb**J0~Do-ArVWxj`Xe0XiS{z&L!MU-{9$^%@Q5&1q%X@<}`-xuV(*C|0pUtTL
z79lWjJ1uS~Li8_r`qql;mWRv1XP2gQ=(|tGo5fYb7TY>JOAVcZg$L{VCM0jx;0Q=l
zPWB32?W&=m_)$$Jy6LX2U*pO!kZU(O?57f8xq@cwwS59mop@0}0g@M`C3(!O`h!fL
zKv8@_8H<_r9PoBCaEt`J-Ox32(AB@MfipL@zmL$%a?iXG_qwHT!5lDl2|Vs>KY<m>
zej=NKK8pRTwLm^%^j<%ie&S`pd6vELs<@UM;Z^Z+cyuTbFl`Ivwle3%i;^=rFxu94
z`wTM7X=h(bd?oR`)aMvF5qrYbA<BEuEv*k}>Jl2c3v5OT@lz?VYfe6dM42=^39xO`
zGnDYc#mpQc3l(^w9`z<skUZ^ct6o0Hh1{mE08K&e4p#y+YgL0g5I?`05C0<aFeMJ6
zy!{R38DC@`I=THQH<EDs!t=Mtdz9#eB*0_40$<ZsJcgNctG0rVX_3&CS*ROS>durE
z5}G+O_Wr{MZZ}}+@I~N3BGoZ;(vQ;X=->HvQz(y=F$4uQk0k=g;y)A~YKH%*@bLQ=
zg@@r6g-7;w5m10|B4gkqfW{F>tQFKnmHhnY?tRR9eQ!N)6_#p(<Z6}w3`r@`!X})@
zrj#}jj}IhRKh~|mx9zE0^r1qN-dZZRUnV=7wV&RUxVeY9&?ihzMj?uo7oTc;pQPGB
z*f)M499>Z>SSjn6=g8Xws{ohyqe$H4`KmUQZ^?LBViR0oI5_;F>1}V5(JbAu9d*8c
zNUcy58V9Yk%4zqna{_CG&Rx=Kf*Q3rxVv>sNV1EY1^gVmRlH+-yV2RDB;|4TuevJ8
zy%!Lb(tGPWT<z{O8H#0l6*X?n#w?zkenNsh3s2m#yN~xxrRbi8ql?Glo;|+8H7<6G
z2>Ja3l?aV3d&|G6<%(2HWv6#~K(53(upbJcmp2a*%1XQAlK${x*KmN=Z9n$g@hPTh
z$>f?&R-+NPN;jU{n=P+l-WNoM9dH^=D}XMqUC6ilMDjMte&>Mb)wh7o!J8Hm5)9y%
z_|+6N;Zs6PE&O5*NxtDOgso1qkB3yo{PI!68i?~U7Q<-uJ5J$5Dm*F%dAr*t(`>*W
zlhaYmqvZ>JY09{8x%H5X&ZW^`z?aSJn;mx7`(1Z-;&$3(#M!oIOm%${$)VbRLOxmb
zj};Sk=h*hambAXW6=~!O{pRfgVTk0A%&I2p+57Arn93vvycq_2gj7GZ&R%eWJ%7s~
zO(ps|>!c)|Nd>T<opf|la_8fpd3E-jMqG{=4r-`Po9qniY*!DUJRgst9M=kJels0M
z`yrsY=1jEyB@fR5)pFRv;pG}2LWjC6$NCpT;8tQO+S%^Di&*#$J!_--cGGB8XY;GE
z__)cyhciv(2TfPdUd8I&xDzr!JE@$W=q!$rh$Lg(N(LJ8q;F6xjN^!cZW;$?0Hf`y
zPg-owv-@VV`kA)|xu)l{W|AtBY}o=$90?17?mF+~5&E{O%Dy#Z{VatK!9O4Qs(yB(
z@PzQHu!7`zoXDJ-YZ&LZ8f?^Hk}^BXt;K-BLt;>_-VY<ymlM$ykN)1OmYkwfL3t~|
z^WqCMTGCo%rc{%MV)W|$PH{L8+oCkpcE|@=j-I-R#Pn;zU~}5Wd{=$C-idMYje!(n
z{d7G`TLD7Q?zTxyETrD*9&5|Ge&jDJ>*HQS%)HlKnC21^-aNOZDlnsawpMSW$SZT|
zf!u}mn&~l2iH;3-fdYVfz=fHuQS9j}n{rw`0O`X5pU-bcnysD$yRn-XosB83F!z(t
zXreF1XJ`FmFKVU1pMmk%4^S(7xG+B8XlC$bf&W%0@u_I=?~mQrag0cZZ<T#eep+;Z
zdSP~8vK7C_nrLmFO38VBcbC#<sx(aMz~|RTKbzO#o0lslsC{#4xW>2sjrca~*bB|o
z=-)ius?Ls5J6ZwZo)dxX)iGR_GGnhz>x<O0KBR07-iZ?kT0^E4$A?Zx7ANEmttb#v
z)eLJ~SoHDwmU>2XYGp#a!2>E1RQ)L_bj64Oy_9S*qmF)BGFfSvuq!)&UP^yjnG}&|
zuswht{vbr-6OU&{z}__eesH7*F%@c4Ag^42Zz*{$$~zOX*~V{+e()OJj^1j<{EoQu
zdib?Fvbq}fcud!Tx~A*blw%oL()L0!c0E*bQu0rcZBMe;%X@&dpEn{|u?eyK_eg%g
zd(pYFtB0%c{P95DEkI4Hn97Kjm+;-Xio^I>z08mO+i~xvi<X)^md@6Mn7Xd1<&N;n
z3a8QnX8Og1Gj%$6Ul8ehqEd=m;cv3EK1={DL3;UiGo=Lo4(!wi=2tB2hzd5<VcAr#
zRcLhHSQA)1Iq78hjn~De_3j5gdx3Po@7YylBGmFI&7I6iYX{<aBL3;+8eNA`d_nLh
zu(%*_$(vBLJe5M|Y=aekY&l6X#h-TUnK>=w^k+>(N`Lero@?m{pbtUK8t3}cI4?J2
zK}LiHnc=6fVt9LU{sT!40(a7Aj(QM}n8<EaX2i=Y?ROZu@=+k@y;>99G?D#>KM?OK
zJQin*OeNdfV`hrEhFtezSx@4tm%&s$yOt-@s8goD-M&-by%r{173+y?DyG+x8?xaK
z2Nk>7ysmz90+YP|<GSkr#&&>9`F+a4K|ukgmY{Q9J-x=K!~#(-rq1|AQ(zi7D(8ND
zxPE?O)wGa5c?f#FAa9$}2{GP)wMyS?m}BFWAmH8`O{?O8^!2R~m9-%<83$k=Z%dS~
zErJ+Q96Y-1(0H-VfZyC0`C0}uIfLmPYkVTBz3OD|m9Kb3tD`44CTQ$33E?IFU!RX?
zrM6`>+Hel34e~~#rS0m3Z^n=e3Xfs2TB48p7zK7SZN|O}ER7%WXkXqqN0;N<kq0`)
zC6h^4$klDhx~GUJjk0h12Oq`*GPO<(RwzF{WC|?rZp=cj)xuKyQ!eA{gYOp+Sgc>=
z2B^LB11a|g-u*|A`=75zpLP6o>Mfa{m9pe!X<`G&TYiY#m?M0<78!!U53JCk-hG~2
z!D?lDU~!zRyl7Y?2yK!dFV30t+rr}g+DN#XM}WNn)#yDhPG5409XLnBP(~B%dlL`1
z^>bBCC|F<)pG{~8Jv(=?d1)Jj2XQzZ^s8JDMY%sT2BSe-usTp(jaoS{av_>q?GwIv
zFlP!_-N-{fGVMN0G?kf@5vy-7<MV;S&vn2fvl9C?*YCf*|NqGk|3};Ko!*hiB+2be
zZ>-hdmZv9>@5IxX+B1G~OK%1#ll9&6d_4Z}4976P@lD0KY-{wVjS!omTH0kI+a|xg
zqm=&6oW5sv`0Z=#{W-`*EAQEQz@X;(1}Hxg0{c45r{AM^fwLA*uY|J<T>>s1xVl?u
z=4k9+cQ%)H2VZ296RT&`x8ef5CTO~zbWqpZk4^Ls(3VamN_L}A6GEFebn)>?8V!9=
z$os8iaQ5Qw|NZVg5BYt^HJZ@UYBSIe*6MC5umF8^D*LT(6uJ8ny5fFQJITURaMG8c
zC;dRr!LX*rs^A=l8mLV@fl+QuiuUR88hv3;y)R1_D9wI8^1dS}ZFy?D?PKqFi~5u|
zwoc<QhsDN4PC3=sk`Rse1?9TU2`6sQ!JKa5)Ijh@Mz~}@Kzwzcp)*8VqJ*xt;?+`4
z^&)v527Vn@rM0=Ey=^%fBVwKj+jwsX;|Q-0+Tk$cSVqA<RF2kf{OKoclEkUf_pur-
z<5X|Hdq+*ic%)UMV!|)u*SVW7j#bj%&NbZCUDJJKWA~B_n;y=Gk6v}9*u;JstpX1#
z)@?z)qOZ3{`z`Nt&y2??ppb3xfkxdSq787s7HxlLfZS4(qj`e%lw6I&?vGr-I&hlE
zHK#+w`HSi_qC=89Bwxr_1SMcg+`G$83M6%oo)?>H50yZ^BF}W&;z5Ub693ho!)<9k
z1}WV--&&G~>1D(SCP|AW(k2RN(B>3#`JA&R!(${1AH~z@KY{E_&-vhXBjLbv!yG<Z
z`MOKcP-LIKWlT1$g*5F3euCa(@fnUQttDqPPUUE*mn)qI&p_L<p2rB96)vfFZ1;yC
zreH|}<j9DnxRs6b7IAs~_4euHNMtm9Ls<}GrACbHG?~~q6u82atC(#5cvDG?0rROu
zr3kQU+*lvcvrBVIjcuF=S;!1eF)jhf+2MViZ}%rM{=gqv3sx!TPdeNkmyS3AjZ;l$
zR;6q=lhdncCtD?73}eFcCl>dGt9fkM*n=sDUmN;h|Jd`kz-dpLhnS8+o#lFz_jI{N
z0jEu0kp|`(^-4aR?j`7BXMJdy#DI3c$#E99ak2je;Y!5y(Dq&l>_UBlIUxlCL#1)q
z-Qi}Y@{Abp+JhBArZ+ZY2H_O5T3t4nW}6XFx2o+mqEsxG{WhaXD;8!Q6GB?o%t6MB
z5EFiLTTE}Ckr!q!5wv(F6v>~4+2=bbhk_eD7O!(%T}dY9?Yf+sr=DedjMaBES1ggt
zy+4s&Y$mz3Zc&0Z_h=6pZIsm>8$E}|Q@u^;AQTpt;5~S+a9psx-us)$4O0mwL<CPD
zp4(XijT1_mPQF`56qz0R622he4n;>fI=%oBuE(zzy+1fkHh=&xIm!Z&0}oQC<p%pt
z<F;>>>}f&DXzmM_itk2)?A$VgTv=D(_xy^LMyj1(K7YKqJVNqty1db+<N%xw0I|;F
z`;s`<E4l<-peB-Eu@u)^9R~-R{c9c+^XRpM@J1!irvP_UOSA{{{2Tm7V#K(xaJQ$2
zIcorYTt)dU@jVO|SG{5Ke6Zji{>N@Nz&gPTxm`rLnHyOf@jSoeLd}Lh+*~coCX=i|
z%D)s%1p?%9>ts{qq&j{qCr2Inx0wLuZ=}}=pO2J%8n0f`h-FK|hi1G#Fmrz2XmgXI
z_sL`030N#b*g3q%NHSH<HuUSKwUGN4Bw5_5%D7pc#mu{r1*{d}X%)RpKU%l^IhUyc
zlbK#J$d+HM|9qsUNv9sATzAc#OX28w53`l(%A%P{*5E7sQS`RW{G>BQvn-(suk-5t
z`_5&I)?)7{+Ej1H!2Q6>{>38miuou<<)-#3d(kCpA)36<RY1`=Hcj8R6whZJjD`D)
z^4S{^cqXc(z4NCTB5ER8D~Fk);;a*B)!wHEEghSapXlHt6zUxBs0O%4>ff=TMD(Ki
z)t+D&E87cm(N?XK`5LE$GgK$CcbW?r--w<c??M@;`;iy<-yh9<_3_A&2qu+u!ODE8
z^ySFbnNsSispYxvC*cqB?}_Oo+qbSYTCaFy`lA|z&?GI&{Z0mT$$w2kkFgJDT?+}j
ztNkV@dj@v9<6Oc+8T*E9C5a?0S<_I}0)=#!fj}BHB6z-}0EA=?%zv2EFFPGe+oWwZ
zZm874-i`I+l#5~Y+obAI5wyU()Z20TBD%tkK6EhW%L-_Pj}0`QZYoS2h>z%2f|h32
z3w@55FY?Og_Kl~eb;zZD{OwmR_;)v=#7Lf`_3MnQBZSH`CcI<{x^+G*O1XT|SU+PT
zvTfYuaXlGrQ#*2}DqF+(@GWSH@knLnRbuVK(l~%ww6AYJyVa)2Y?#&z^%u(HCy&<E
z^<G*^dfpq>dieQ>0Rduv3(2u9-gzddhmY@mK7!;;rWU307ED5yTn7`Z#>O~x(~h}K
z0?<JwerF$}B&<KPw@vMuGoVd*v0r(v>!0<6sa8<woEX7{lD4L*gPt*pDWfK+JUo2i
z^O2Go-Ao!)+CAi8Z`k+!6^hrM_Tkx>c(a<pK-n4kird9-{{_5FlSYaz05Z;qvDiM<
zJ)`hL+2k-FC*nYI$}ThW+NO4X1yq+!Q`K)Iqec7U-1q&(V}uVvew*bTjO|ik1qS-t
z^Bs$WP{G060J6evDJqLI<SEPmk`f<>+MJwoFKL*$RvFJIP2J0A=oUT>8morpLSDs|
zFP@z&NebBRs<YtGHwh3l8E_o6`iIO!H%VgxetrAeji+UOzUtZ<sEYW6OWjT6j!MTt
zKo}5zn6^Htn0lxmt%}}FIx!$N#yzO%J&!ETod~%nm=c~A<{R}>*Qfu9*6q0>7wF;!
zZ7#t`o`H`q9qRzaDOsN){D#)t$N($&uJrEM%$HwAM;x-al^+Zc4-W+O!jf?4W5IrB
z%aB_D^p;=7^Sw_6;RtTho!2q)_WZ21r-Vwjaycw|9X2UhPJ}+*kw9Phc+2J#gUX5x
z3RK+oC1XQ|lfk|>aw!e*cWA-;(Mk3NtYefPe4>EVcExJ!eOYKDQ763|nPl$Y+L(QP
zlj$tlf<`#idT#7SFE3&~Zh_&uy9{Q^9NvVs3ias92~R#q59rED3Re}t09n>tqBB~m
z1P@L69^7H+($4h~Y%Oi3&kLfWZs1Ps?aoc?zd-Py8i9jM1S~b(b<?TS6P&Pb(WVi7
z|Ce`R5EP{?bGxO&;cIpb-|D6!QW*2bJ{u8s)JVHjuow6JSVqH7Wzc%jc2gMp@~C8k
z{oUw;KX7i4A?VW?hu6Vj7NsbMQIaCk5V@aHWofC#5rd6)hbQ8uIaL}H`q^kDa~)=A
zeEta|#E<lVj)D2Msc6p7Q#P+K47c!ci!ygf(KM^LnT*~|4A}w~QGGg+!p*m?BP9j^
zd84qF<+D*ZD6|+bc=C4KU!rOmWZFDgr?)Y4A?vNDm)&>y+$d8{-c`Ogk6U%)OC8Lu
z;%bA_v6m#Y@3;TnF-@9I=JaJ9*7fC!D-j=gBmAxc1O@@oaSFRX&Sk;<yZ@sUri~c>
zVDR86;Z#)U0+Xn2INR^o+~B59(6i)fS^}{&y+UM!aPt+Z)D8j#F=Q7v;g?T~k@2^7
z>)&VIl-^_`4vu3BF8Cs)E^}lbm5HIBrhM&<FXx*%`5<a=kNTY8C0jBR@3ggJHGD6J
zb*^fW{?$8l^53nF?)b`*=8TH>i<Y<u07SqUuT;P^hm8~vmc34(_|WA;_`?PHUTBuA
zANhZA_oh*4rF+_Us;jF~RT(^%G0u7tqn0shoF~-oq@oqaiY5+#l8EAfNSqa?b1I3(
zI8f0z4;<rEMv)i<oGs%(h$2xG6da=BEErHxaHwaex}WaT@26)y@2B_O%eC09<<jlG
z**Ew7|6kYd(#!7e9bTr|+o<Qf_SOd9neFV}>nUj5k>EGja4iI<E25WTp-Hj#!R8YL
z1fnWO8_0&8ifu&G$v$0!h#ddR>Xx?k+aO8z8Ncqfk9J-DbbrPgtTsrXS;iISV`>Uj
zah-Q+;lECY&fGQ~Yh$&vUgxBBmKNyEm<Nc)3;pYU|Krrz-c#zo9@(*Pwup+2FgY6$
z-!0K?DV5`c68p#8r&;b>&VQ$&aw4uy;blebeqCiV^BsoCDjXk^d8gOg-I%^<o9vZw
zcGstJ`Qyo>iu%HN8?@@Ny(o9)mSM|9>@0GrSs$i;`QQ!bdEo;;kk_yOu@wK;oEXQq
zT(bIK6QxFre%9983LMyxI-Pz<*BfzET}T<c9n?b{oEt`a>_koBYaa*Zb~i$J5vb;?
zoaD~PUtYIKoXY(Vo*s05#@Xc3)EH&C?i2*8{K);^Huw8Gr}tO(_V3dvzeKT5pB`E+
z*-!OY7HtL$KP&0gS>l;T9*<9MyX-Cz<CBFF6K}y03eBHdvCvBfD<-@13#e(~^Jv>N
zndf+DA-=BHdMXrMqVV#Q<WtN>+P^imbef;}_Pp<EiGEPN)~Jfue3TqEw1T-)AM1=Y
z+xA;A#2OV`vc*KXW;dp^LVekdLB*Um;+~Bh{p3_FJ4@TZgI^h>!6=&@rmZzOw8Gt8
zy247v*eEup#i@uRoQKZ!&`Gce8%}2ododD0g-Stog#bj`sKdV2FSv%OKBQ_y+qnbd
zTII~dl()80vmA0|7}ToVWDaPlY>PuK4VoT?$$tMXeYo`0`HHyc{y$FD`wfs9bRM)R
zIL;AetMt2jHl0S7e0a{vF}%lxyLhY9ciZ0zz6V2we0A!x8%yw+4Q5tB?XL;$qvhzd
z8o+{L*M_-NO5SBoMD(*;;a7OH-??^nkfw03ct5Zwf;18^zo29OQ<)bxW(>$#C)HmZ
zvPo$5sK}^^^`G?3jeo+N8D1KUTO_mRYaAEk=U8MlLSO<`arN2~PWEDL_1fPB55+zQ
zUj}uwzuP-(Im#crgWuM3CPWz!A1`O?TR+c+YgGcpu->KY73=pJrmq|G2t@zeI}DE`
ztZrcXkGA%Q{@p(<wgj;D^u`7Fnah<k(X_o>NLt<Y-19~KvS9CMNGo7}2mDp8P^g<q
z3ShF7a>5x#pBCPRgQqR+(tce^mK68(5;F4p<)pem=dG$t%|7GGmcPAHS?V{OIol$C
z(LzMC_E>u~9x`O`6LkTe^d{<Jbp_#@L?GhJkD<EUa#2{S4m0gEOM8|5K&k55CDR}a
z*df$TBmUE3F3@Lc_#MU3Tp2oVh>r%P+8Z6G&lxc4I}0zT!Q^w36N-22fBJ`ootjOq
z3OGjM*4v(IAH6*_H_1RR4WiDrHJ+HD523pD#P~&|{%u}*OMm|GwXs6*xM=R`g^inx
z_W)~6I6nEfMf&N9lZX5A9`A8#P@+bd``gDXawVjTg}98!XPO^R;Hcct<_YFVDh|bm
zO@?MXxBI*F$qHO>bob!oy%i``=Kwxxi8~0;IZ&<mE$MpYUzOH>o0#68YYkx$U3CUL
z-=QI_Xg#-|hSY!i(8_Kb{o|C1f8wRpRV|m<)_e!`pBtA=;orHJ9fIvt+SX|PI?WjT
zl*x~fWe9{2E3kJHW9{q^5zeGdJ{(_vcvUDt$=V?*;r(Ni2k`<)9kHd8-MVRKJC+|G
zZSIqKltSvG@~ENMNvlRV>9|_I-Ft#My!jgX@C(jtC$j3>nufF{p>}5Gc+le@Y&mO!
za*m?gb+5hJBP7ustQ=?Fs!KP0v0U%T1p6@NdZr8*#H?T#6YZGbuW@Ct5&wSQuD_iz
zB9S^kK9!r-!|JXTU}u&pqmddNmr>?~@@T(E`*R(NeCM5tB4k)k0Gcoq8o<QXi;YO6
z{3w*1G+5Az)A&hfe|IA`{_XHfTl<dJ^iNdS9$2Nv?!rnHO{Iz5^{98OGk<t8Eo&<7
zFuO@QNDsN;<5lENVdHfk+uEP|#v#NNi}toNw7YbS>L*ssb>ChweyCy+LYr(;WN-RK
z9H3T*dghhhE;0%F>6O9LpgA|7QJ#D>U7VRkW7}G;SAZwz1D`OHGc`X_FHyGV=CT~c
z2|IkqwM}rEOBnmB;QfZL9Y_yPow5wBzzE-}m5doCFe?xG?=0II9J1}-9eNp8%p7bs
zth+iWfxh;75rT+qn!FnAszr|We+M_Y__`FG4j9ekqt~|^D*iY%p4$%;4z2j*?4MUj
z$E2J&jfg2dVmdkNnE9O96*#sWV;GV-Oci0c7^^zXmCgA9P-XBh3y0?`4o;A{lRT2)
zsu*7$nFtK4hRN4H!XlWdqagZz4~TR!^r=B3>9_<{{pDH4r5K%uxX8iVLWRi$xqM?*
zoXWkL-g_SAV2#Vrw~TKq^W<E=Gxtg1K)+fh)ymu(TB_{9>st{43!0b^sfwM!um@n@
z`pZS7dUdBbY?LO7Et8qMHNjhca0SNv65dEIC1>tmW^&KAzeyPE6LNOTehd(#$v`aa
zBKDm*6-}^;#09Ift#x#`P#lWpPG^{=Bzs=u-S0-iMmdZ`p&5kNY{*MhnIWns4lA2a
zQdLI(rMGQlqpVU!lMs9jkqsAY+@<o?Ana1sif_tZ0be)FcOF95q60gIy(^98Few4G
zalG_sY-`ry^~2+8g2?9tfi>!jQl7zSv}m}eAc>9n2A1?^%q_+*J5Z@4;IpE>kX++%
zax-^+kt|w^!X#O%6zgz{01=mqQp29qwIPqxIzF~nyQobeWO*Qn-yhujas5AQU;E&d
z;g-<p882o-@D7+v|9f{;Uf&j208~{^SM&#>>h`#>MWW(+ar@cWWk5o8&tYS9F$d$B
za?d)Cx*B8UrkAy9$sg4$S!%UP)p_2R@BQnNm91^+P3#<NSbe60TM9|oL56>%X}A<e
zFVb>_yCj4v=TK?S1U@~<D1*SyEorMtC59R1ss%W+9>4<tE#8I=8A5KoO#Aj&?}wNS
z&b=*-vrG7x)S0k+vHQ<0Zz|^%dIAJODb)zRQgtK5cZ@#gjy{J>YR#Tl12batNN>YG
z?6Z}im9*DT&nE@p-qK^SmlQciCdSF!l01|8(!+zBT}L}98No&CDn|KN+>+MrH{V{y
zh(ogG%^q&7%r7Rx2OAm(g>8JR&9sh;fzwt4;~gnDaMIU!T|TulcY?JXd4L_sW|yYS
z$Mu%UmjyPfVMoa>WlDJ2+s}Zl-2qOF959J$OnI+z&NcX%FwvUbVR$->Qe-aG#yk8p
zQ4c7c^LI<Rg<LmQjmQtxF1ZSQTbsm5(9=(njW$rv?m37!aNez6z&)z#^;pt-zkiPI
zbbrh4AJK|NL0h2i)#>m9GyOw^-I*ukFxFH;g!Sa0+2Ain=~SKeXYp?!soNWd54LSQ
zQ=J214`vdzIZ!(i-cLVd$H<w7t>N=;1rL&m0~~F3wrSt9`Ky`*KivHfU&xn5iC^L*
z)Pg;f%1^vee;!cRa9;w&95marEY4AFUw_=)1QMA)9jHclNLSbz--b9iU-cZT*9Tj2
zlR$n}JgUz$_ve?G`MNpWt^oFW=h(A0HkCpbcmFoCRZjV~PdF0+C=G}jureAd47o@r
zn&i*6V0f+LhrT8e{^aQ?M&U`_zQsrshJ#5i((UbTppHkbZK6=Cj0~^*<6PB5bQp6W
zp~eN^XF9=~hO9kEi(28L-~W6!YorXRdcQlG?7DBs&Wy<^du}mXDS}+P*w7FVv164b
z?Va{&(8t#lXx<ndf1%U<E-<3{rbZ629=v}k8WoE6YJD+}5~#td_D-Bh@pfsMn`-fy
zeO8k5@Y0!nAO^zy?Thkuybld$b`$`p%$3aJh+Oq^;a1UsbvE%pzR9o4``3j>{*^-~
zg2C?MSb$G!0AZ}-P)IwDm%WX)b3&c<0I$0d5H(^G(1#Y_Laa!}AYRa8lfn%>GIy6}
zS0r6hsgQxS6&kav47I`w&90Os&>)Pp!cE`by{hUw0U`03BQpP4S1G6P{K*d|6P3u<
zsdv`X&RdE*R?na-8rnC4weX@&1+8bvOWhp=g%;o3-p0b7^}+WdpY2gjiEci`{qNj9
zCG^z}nSWq9(sW)3nArF*+jgT*KE_F{4<YMjxi<sy<=XRzOS6P@WW$O!h1Q<C<3mc7
zIsJrneX2Vi(!aQOoOb#F@>&1=PsN8WZW^-LE!A4xg*FOVud8pofBSZPuKb05IRA$+
z3Z?qWi{F@qw}T25F+8htCy*p`{;~zd%D_1dZVseKTzVJ2GBY>Jo$BH-NEAv-a)~(3
zB#<OR0Df9gAT6|;k92wO^q@Zjb`hj}&;nqm3aA<9l|3(B^I4Y#!>PH?!SGt7EMVdQ
zvn~UN`s75^sV@o13lnp1;yhT8h@_7ZSC=!_xI^O%osQ51P)}bLrHnD@eug;Nnm+h=
zX9qhOU|i^VB$N^=^$I@N2JOMeyO?E6+#Q<?*JhWNGQ0Y(Mn%fdh`FDR^t%U`royrl
zkRvnR(>B?WrLJMb!tCE~@-ACQm~%t=)*7QW9t`8;gN4~Paxh#5V{<Ss|LBK8-Zo;V
z-#M+Lvy(+Sd9ct?<mu`GNi|<qDdVbOh{-#yH@Ghz9_1^?)K%L3#-(|IO_z@iI1tRk
zJYVf#kHb4BSEX~Rii!TtHSy0Do;LCLwvHFiWxv4IIJnF@$2TSs<fVY!JPMp<e~s75
z0Yr4b(k-PE$%O&op7bsT1<ONWyn^>hyvQhZO@`RWNcO->PaPC1XSDTdkJIhc@_QVv
zH1b8Y=~$I^u;&<&q)Stzj_HF7@9s?_U9a{0T$%X8hsVO5>?`?<_X>@U{9pDg)IwE@
zYSiQ+gt9yO)g$8bc4Ug!(<zvu1JrHo^pLz#>NpBL6q?<8uIs|cQ7&zG39w9McW@Hx
zG<ScbYo0DS0*wI1zk`2&BEE;=-w2DGN^+-L-M(bnf;Uj{9D8lT_{MQ#!6|7YNmbX?
z6tr-S=?wNw4H@!XkokeqGiB)c=GNzE=zkwG(+)_5Zb#cbdj4$WWOd#zNf@1gtuM}k
zmmFP&GnhLH{phdM?a~65-=w73EQhTx>}b5Q72*bf4enuYCU?|Gj7~OfRVj1E7x}H6
zhc_Ti0S6fkR2?4_6b*KN^%%(!rt0KWd1UK|$`)1|ZLj9Nz0j;ha5K}Yd@p8@G!_Xh
zfc4`UPwMX296|~>hWxg`Xn9sBb%E2l@k#FFlX<#?)D+kxAp`YqFXY|GIbd*V!?^I%
zdhh879qzj*^ICl)>E_Vl(h(I$AR_|O&Lq7=XoD?pHgpGV`Zck-zWsgX-oVZz`AO}Z
zyYu+#gq#c=v|&*Jh{bRkwYi8Mey`zO^xmHV_$#x=9{)8Rn6X)~d&hU_N-4MY6x!mi
zG7U%j8%B3#(zjI_@{&*DuyV<NoXR(~SCWQ?-6yb8alTN2M$?|47luCbR@fCtPbs7|
zG-O;%E;#m$yWDGUAzY><o+v!HI4J;TSxZNp^=zN*%dv<9?rT-QhN!}d*zMr@oavSi
zg!s(PM&}uNK5HkA$qgojkF<8JkJFKf*ZjfR&r6-W0e6;&4#Pxm6#bU^T&!wa&rF@A
z$DWT?cthbV#?`B<jhOcKm|3}=Z7ISo?*f@q6)EREz0q#sr=(Bd0Buk%@^1^B>J@D#
z(Vy3g`YZ5@qxR<1L}0xLGefkK-W*6BP|3xjft1Ui8WfUJO8kbw>A}BZ)Y|XPeHu@`
zwdeK?uc<cOCW@H9@roW?E&3_}zIUAaldI_u(*cz4Ftf*`+HY=<CtvgQZrt9n{j?({
zJS->7VCDzhzw+<yMHdlZcdYB`h8W+LroPBNeuQ7DsC3S59dO+dJd9h86|K@F<@PH0
z+a$P5^Z|y$1{(J!5%H0Z9e<p9-hTE4mEv=hQkulv>Z`Qt(anl5+Xk~hTdQ5Ks5Djh
zyDr<d70dmC)|AFvN=|BTKO)wZar!x%*d|%p$g$dlzljsGHDn^M*zQ%#<hdi0(56M^
z1&xr2yqf8xjagz-e5#D&n_WR)e_U!W4KC>bQNMZ@CW5BY;RE2->51VY1^S@L79*Y2
zj7olPNd)4p$Tf~?CpKx%xWiSn(5<zI`xJ@+c}}(1#3Sv;QBVQCn~cTl%ksMs`o2K_
zZlT~z&3`T#|9olLCmK#A^_U9;?BDl`=C;9G@k+}f5co=ia+KHn!K=B$ST<Vg@mO}g
ztD_tC3V5hos}Mu<AFGs;{JbA)6e(_&24}QNzv{<iK7!4@m%=!^qM+$#Nqjd7S9Jel
zB1q0ieJyJza&qTZ9fFjtq-Rz_+r}RrU{=vpTyYx&0CX}UB>}BFhj@XxA?26(#+_I@
zAh=i~o4&NUE1l15jtcwRKVR_w&H33U#|;c(#mszrH}dGh!B6trz;F3#FC~3`5|?T5
z3H}PV>EHL|jsNIgsbhnWUBm4kisnZQ)HCFc&6eytlwqru<^P?$@Lap}OxXXyg@770
z_*{EoW7<hft{Qd?cwXND_590>SE+(tuJ=iTv*`O~VB_}x=Ey)RV`Y^(&P|p*T`SQG
zvwgI}X?In5s_Uv8YY_YDZqquQB8p)ZTrw?KUpH4^nMN8<=Hb5;9|=gpxT3cAzSh^v
z=Dz@D(Sn2Ay(HB*eQ3L_+1lNSEj5ntVzYr4n-Q{LU5;oPSfbbx2`%|ip<z728LFsP
z$NXI9hF{n*WIK8sGh6Sh1utJ-)yBzd+$MM3f^S2y<W%6Juv=3BXLiU!rNyFijqjBe
zEHAG(iMBPLceeMwgUztJdwRMFUK;|DU;~)AngxPJhHX+=S?LZP+>HS`vLxTS29MCy
zN+Xo+qp#0F70pNw16VB9AolB?p0hFi0BzW`8RT<4fC~Q87&rt<5DtbqdFlRyNB-q+
zMcpF*X=5<En{Uy*9ZGiQwbDNeJwM~T5Uf8zmN<J~A(Zj^Ki_`UdG-l;2Nx<iqcPa^
z#Zae(wxIo+VNhhpak(~eU}~tTHFzYveYl<f8r<_;&h2$(8-2@4HNxPD(~C7m{CATL
zo|W^`=Nq{CxnK7xD(5#G!xlSmU*hFp-BG?PCW7!=+IwS*K%C|9yQrgdC*WVH8rlF-
z)bVJ{VyTCi(9-uwl@T~c9r{$BO1;#c2RiDK$0DvKJ}vK5BAE3P(uqbF={RjrY{}EI
z#Ny2M)r>S^R)4v!NaKfix1BQ#XchXCM8!atIvnZw#7wUZ6L#C#N{g=oIq`VMb-&gl
zuJmUs)JYhT!IAiI({;`=qR)GP&B>Uv*<L<hyXJ82k5g_Ph3+MdEi-Y!YDwE>za)<X
zvYlTfq>kKXpVmYD+A?&75Sl0ybp_$e+_bPi?kZ$1+PQ0T(HW&YB%kJHq0y0$14+@=
zGzIIWI%=g>!u2CJQc%vw)u+wAL<f6~2br(m3<flX3yZr@wcY79YFJWL@Ph~NP$%ak
zXFWZ=P+cyE1NpgLOZnoiF0=YZcAjVLl|1KfZeD=cKx{f~RLaVf#9}dVd5XiC9?rKL
z8XfxVmY3`-linccMOt_O;Z_I?J&VLQFM08}R`=NpYVHdvRt@Qru<lc<>onli`X7tU
zMLi#YunZg-x09+Ch9~Bvo2`-1KTc6E7{-MMRY3INFJi$WWO7x*c*C}KVWD9Nh0>OO
zQ^f%vM)@Auc`LWSG5MY^LFX+!evP1iw5zlDhy}~=cRQ>s6K9tmxy@n1qh^8qO@wDd
zv;~2zeOT_`(zXf-+pRU?<}b!sjzV~>HqF%Qhb{C=T#SORpjcDGqDo<kdb_Sy3m?Mf
z#lA!Y16l5@c@tHak)d;?!K{53F7L3{>gS&ML$egh!e&<wl?7TkLj<Kcm^Fug@|f{i
zJL=6PX1r>eFxcCgT-NsvEFHxbC>fCGmC&9rlGO@{xfSMB9cqrWl{>tfq98bkY3uc}
zJaHVaq~$L!GTx2q=k|@qMKj0+w2ZtDeLx~>_I_q}OH{ebdWgql_RQR1X!4jsr#nz~
z4qn6;!3n|rQ!xREi9SIcPA~VlWR81(*a+WZhOL2oO{EWqlr-Z(gZB=3mD<Y&DpSY}
zC}rr7t+=(T-SFM#-f6>X?Ad|WwH~IcmC)4Gp|=~gx(HE5PrX~1=UZd?6(PX&P|b_=
zU!JgGC}B(2E9wbxbDCE?LTRx*!Pu2n)OHiN1(j<9hszrpt&ZBv8N%c)Ho3Um`juxn
z>imph-Bp*qsz!v_J&f2yOb+kj<m?U2bVITB?snStciSYM)_*<q<r~mfOu<-OzV1y0
zw%|0c$0G;Q5Z;I|>qbaX?n;rMapvRpS(en<*50rFyL{(sdpqqTv0cw4^^Mo=lPasF
zv+Ei);1cuW+k0}aVl@y-Q*~Iq_nTBR(zDAbV1?GEmV{W5*w6zI#Rn($!CEgtEOOG3
z&b_S^>^1nu1OJ<OJLte=Hk-B|9=Ee3<_`~=J#@q|%gn8*>jlNuu&v*>yo-ua%&r++
zP{yEVj>`KP-F}Mj9fAPrHq)Z=lS|@J28Oa%NLr=o&Kh-#JIq<GSN0U0sStUE^x#>V
zUk|I&<ovmlGYk6-t8+8Gy<gdE$FmQ~=`uq^Vt=fQRXsY6>6=~JsVbOPgIejzQu-nS
zTE_!5JPzU*$VILIlw%ZtRjAZ_*4u0L*?pJ<ebT5XbR0+Kl@{-~HrSH{@YbZl_aB?P
z9~Sw>sHQ1C)KS_fht$BUzsp;QO^zWm0}U@ZIhia+PdhQaf$8)8Zi7eOw(V=4abg)Q
z)bq*Y9_+`;ax-bhMsVsLtYu3)f2R{>#O?fg#8oG)h*REze>5Lnz+58fD()sU+pBFS
zJphs&QLd_TNpAWWSNw|0!x;rxc54ImVOJ_er5?MfNNPC*cmw?Mc9DDM=P-95%{^&(
zI89d2b&z5FTX54y|F?UO{$}G>4UJ?xNNDt2R`Y1ma}7nmo^o7<%1;hjdNkxsl5KSP
z=KAi2%2992jX5ixO26h9S{t?HDAm`S4el~02_Cj;TZ)H{brt<38=yfuC}Mah+;P_G
ztR>%~nP51$<!!Gt5_+o+wQuL?js_w&h0hG-TNAog{T6HtDq5{ZEg;WXI1}({J--+g
zBD69i>zXm}MSn(EMe>BZ`w2L?JE`JB6TqCCOnj$yfx97!qv#C4?k$|Z`EDr^nvq+(
zYUayt<+EzqW`ohS?vqNWmQHzFZ~Ysr9Bboca&m=F(#Uu+=(w>0S#gMk96VXjosc*K
zHqH1GVk4!U6-U^u30k%MZggJ%?cCVjFlCTiJPrV$3D!`zS{<&!A+CQ#Tu+}{@AQ1P
zN$uF_-{w)VNva7Yr5h3M!)(q|Y;ELfPI!yYv!mJU{JqA)V`|x|JgTJPOui`{hu!zX
zte7X#-6Q*WvGhmskuthfZu@ctWKwt4cX<xrR-MXhU3Rr;!J4bBF0V4YJ$97Nw#hhG
zHPumXGa&Xqoj<75b`t!EXV&g!fmT?s=xeVpGCy0%O|1I~e>}A<ht)@7I4!FCG4ToX
zrVeO&-DvL0nIsb=YJv91PA}oeFbntr3AV|F3ZR1kZ<X^><>r6L*G_#oy+HKMiPp<u
zY)*$)p0-WzrqBXRd?1@m2#EZzt&RFf$Wffb5dI7ha1u5?>@_Fgkgo1tb=G%^9yYzC
z7_;AM+U{!GY?IHN;3e?4l+5B|EJ#Asx&Vo&VFGb5&4b4Gr}~Oi>Bb9yJC}AF$YPf>
z0u@IA+8u!774?ArY>$>@pvMIgPnslnA+H9|JaH}~0dKB5OK0R-5|A%@re#L{ta%34
z#dSXD;5~Q7ceul|y>%$MuoHeoE6)r;hth9;OZ)yq{WZiBGS}Q9ptr3-nNqghnFp2~
zoFA~Cj$Dub4VJp)pDlQ>+ei+Ql<<G&UPSe}x7%hVQr?&vIk=wSt@MpFEPxCdlg~(S
z>+KHa^Hgatp?v?GD5ld~FYdf`4<|*S4^1BZ&XLI&rDWTjOr21X%d0e#mBXIPSKI16
zZMc(_0zf*E+ac;X{jjZ-R2%(z-P@D;?YoYQuI$#8`qoe-%MYXQHRotk)DJ5m+}X|c
ztGgSHGR5!43gxRG%lM|Mr}*#66dyg`h<D>9xcCxb3QCLzgn22|jK2@oAJ~!NOt8~5
z%a<lF2gySbri~BlS=*RPd43hch~7-;g>0+J=wyUVHbLxa>v`K75^2H|leT1bSzppp
zsF$pyFDH7#!!`;Ibvgd0H*}N(+k;;xXFguy-xxCaKrFlDh0LC#0T|9#W50)4@70iq
zHg=d5U$_|vp*{)(%YHxE)0IQE=4^<b*sjBBQeR>wIeQp%-7Gxq%}$c~Oi4Ak1$#4#
zSvF`$i-uMvQmY4qQ*{A`lun$yVcrD0e<v<#p+CQU%0}bN<iSxn&Z;-V(F^C~cDoEQ
zMOzxVKiTJQMK%Fzj=%>YE?(OE;w3!(Lh<%?tM9a3zZWBN|K55ahSATKoAntcPR))t
zj3yz>L+DPBEdI;!T8N!ZeXQFBf!5I4q-HsFGRH}7+t&t6{E(iMs%kgl?rdvHrN+U$
zxA7a4cq$YkH?Z}fJH>8+{PHq03-*)&4&Gw*hmx&0Tk^T~h!$wUCa0u9N&9j<+D~$V
ziC)`lqp=Sf$~?C(w7MISuncUhIMl@No7MzeQN6u&{tB@@FaLTo*<CA-&O=8{mp>6h
z^z%9hNqMwo1A(DP=?u3t|6W^XPgh5S*NR}J8(7r)cVkF;_Zm!@bP$uS_$Icov9YT0
zk@HOm%-PxVqHT(wtBWFu6iX6%wGL*$zE);_Hgw5C6NKY!>pss0yfohq1kv*Kw~K%;
z182G@h=%mv@2;oglQ;-gcio>0d>laR?K&MK0+#n*_EeY&^#wAwj3=&sa@XbQhDQ4*
zBj8W>6}>rajW(uxw)_d?Mp2RJ$<6Mb((m7H+tkl(CaHy9$$KGrI`(*CCAYFY`KnY?
zc&0wDr)uvUa%b!jFErjwX1ag>yqz8lp;0uvFo=GyXts!WC_I`w@0nf0LHijdv&O)&
z?mydjyA1m-v|oGPU;(J@#9dp$_9I4F5{VEZG4Ey0mU&4JFFt)L>6x#Yg}+p{J+|W`
z8YAH&Ne11M%`o=Ecq>#HgoiJrs#PKQkzPb-P?S{WuR)sFA>I!(hjO-On4~=t;@8@O
zchm@O@tuMqf`4ZNPd{T5T|F}dI@Wl`>-JxV)?dnUezuGu$hQ!uc1<FJvcVWXEPf+p
z6$#NIg!KEvuoa^k*9HxVxNJsCmcyH+I?Uu`r$H;TV5*O81yw{9!C`3u#k)jr`IKqn
zb##y<VOv(#ydrsC57Aq3p&8>bU1tPfc4~cYF3;(Hk`76zCvrLO;+$jN9(}-0dRF$z
zulY8nRbH2t;Ke!8e&*`-`p@Iq8GS2x!$1-;zT~K4CiRLKpn2iD(as5Kf1GMMzC1~t
zu~AQVuV2sZa)(au90?>=2SVHp)O^yx-%e77c9YY3{)|Bv4%X+4(_wr>r%LPbt^-bB
z{K~3FbbjZ3QhIzE58od+Z;?_ePR{KkquR5MPST8$Kw5=n+oF=Idjf%`mgVgmFNl*l
zHkb?RkPmrIrVY#@Q@?0@tf1;`<T3iyr(|!x%_Fz(HiTUk=c)m8WCwU}$BvG&g(iH0
zXMs^k^iTmU5AYMRXh;tHyg<CFMJh}Bl`}Ov^SLHYBozYNliqDYz7`${q<vvc7D)gN
z_s1XY{#{`Gk3YAc=Wj3it1a6l&R;Kw_k+(fXg1<O*vw)q%2__TfW|KLt$C94Ui6{R
zG;*p@DfmfvC0xskDW64pyq~Bt3<1aLIXms!ZBN)*Z+0FNyTo=nul7bwG<u_pRwzcH
z8%rgbfF^qMwxkSFCg)3StZj5TEK97rzVUImbfQM83X`WrL)sDqI?lnr&N1wakYP*Z
z?b@3A@f2(wC?<11ke-LkQcGdcFf46v*-u;I@EE=x&%kSY^aeGR({N<&tj4B$YI&u|
zoc4#g4k|-)$Ucl4n{hpqN@M5P=FNie*sJm8dm(-+6c<<C#HGiEnGBLm`D*VEG20pX
zsND8`?#jlNv|uERWutH@7ifA&spndhj2>D`kb-g$PS-UTsE2!MtEr!fKO0zbXUz80
z#DoM?6iajE7mIn+zRf^f`vc07Y%lRPp1kcV6!Z~HS%9B<nk)o|k^CumWJla-*;gBi
zss{Kbr1;`N)A{!1#jS_M*9?qD%YWd&TP({_XOL%OpOADPxIL+@=vKZE_*kpn&G!MX
ze_QvxD0x-0b0iv?3wXhYqaarCR_dNFxKs1Ic<~MgU51Q~M{4_3;Ik<uWPl_dNK0~|
z<Bjvc(gU;WK#K`T+<<+h2V{yXu4kNAjm64;CST%>#-H!vLw%YYKki37D*mS;#21sz
z<GYlm?hj0OZJc4|9{pr|?g_?6?H7`gu#8;m)jK(~=`m~M$q7$TS^{6#oB@LawT1%g
zs@)0!s$1wLrBWchwb}dX<Vo~%2Ij#`yIVT|IFIL9asr#@zZhX^j9=e@UYw1NEYC^F
z4^P+$W>s|ibQ5Y@@MYXWz`eBLmqgx3VOB1g?IV3!?=y@n_stM#Zzj1^pjC}Xc5QS%
z>Z-hK%+yLVbGJK4G6(JTKOEVfwPg6kY2@OMkNIvE^kiJ-%FI{gl?}(Ec3@+lO=y8u
zw~}^cSaGOXci~3sZg@*+ga@H{F?lu_l+rL)5sz!zV?YzO)Jx)$afV+dxnbwxtz|ST
z=k&Ix%kx{_??D2)U$*#dMwNSA^^%Asbf<(tAL7jM7b|Uqvd&2pNW~bTW2dTYd2#s@
zQz%{%Zn1fnTUHQ}4n2c)0mXa9E=hGH8z3bi%xb2&h4eAndWiN*Hz9|<$vtnpf2JJh
zkk^-1jN0pE(talQQs;nTNgyFK9__IB;!`%zNihYLcr01DS(GmAN+L9#k1e<L-2B=W
zf8ITZ+t*9o4qL{?=~w8}p(U~8SP?5_iujS19N#p+Vyep?+MSiQzd>#-3=>+?cd+K?
zvg7;9E;=!?JhiVqcX`nrylzGU>@12n=6c2cU;gQR@L#9@5)88H`u@;%UU>?&V-_6Q
zR}t0T;IKlthB|n*6S#ji?UgLv{n?)K<TZkM8>?uy-5k<!K(~IJdf7DE$|WZUsvn{&
zv=?r#8}|3u3OcLG6n?Y&EdVd{tou~fw&eYIS|};xA5H4wPRnc^PqsBFiX|h@EniBl
zOSu>3Yp2#ypT`-jTT-G}GZtTlX1VK7p3R$t;p!}BZ<oS?84D^3Wc{Cxww2xl)-05<
zcYFKyo7Bz#xfRB$UsOSsqNdM4Fe({Z;j$EZRD=YaQI~LUnp}cMy-L2IH!Qr!5h!zI
z8{v%%FS5nAq5j>2;Ep{EU#PXg9`L?@MA>kJ50h%~W!i3XuY|o8H($ka0(W4+asv*~
z7m}UaR4;$`B!g{3Qm!5rnPv`3SifB`(c$Q~j!BDMFm1#dII+yz&z{W8q;Ij{J7jZP
z5-u1w+I-e*5dRE;C`OuuQewT-y#MP~e?Hk@hhltLZmXfp(3Gr26grJWsq!zBxsP{9
zMr=3SBkV;(Hz0KeNv_y0LC70`-lMSO+sH)S2G81}*YCg>{7PrfuhQ*}-YV$o^0tdj
zkuGZ|kd}Ltq2ltK@_V!?bvGNz_^KT`1Ji^RnNzXms8H@?7Do!Po(76)R#Va<e|~9i
z%xW7wWrL-!XhKi*Uuj~nXB*pdkuAR@AC!5OXsX}FDurSao0v=8J@j~+Pe&#y`>t_k
zD&GqEy}jMIoaiy?0}1iVV8e1@b#CPgvU2a0+i0iT8}8;;UEzW4jkaf0{i_NI-!?MJ
z&>1+=P39<nU%h>^NF|HuHk`lwkr1|2jZCVFvf@j6O~NP@m9R#L79Y&&j9YH&x`*H4
z?|>7I;7?m-j!*sP1IYdJ0i>=-655k5U<uD$lFZt_Y5Gx84vS}mmNM0weRqQ3jdvQ+
zUbV^-8+Y&yHM_6I8?Q`kdf?qcXdV?#(QoBE8{9%1z%N!FYv+ZyUK^K4-YNyut{1+b
z<xbLEUf|a;U_<9|p&W$uS>Ibd0KDkSt>_F?Pki3m0S5qhWd*?3jQ~r?z{&L!4fOB-
zJ+ARO{r&lRKzwzQb=JF>ZSB|OzBrkz5JYnzZg$8IW=n?0c9UNqq#p`by@!|i^?NBU
zfp}AGY4IT3PP?Pdj{j+A)7`g6accM-2FZOBvP4Q515woaQBhrHN2w))!=a%;1j^>c
z8rRg8?+7nN7Z(qUyyvp$dKu+QjiqCDI)Z6&<R;${Ck9VPvC@74H?~RxU;Ov={a>E_
z|JZ~0#Sz_9>oFqx!L>JUjY>m?{_^<sW6+Y~kiF@pZ<2-CsUbK2IQ9BzN%7Onk`ke;
zhHnY(?PYrx#@gRYPG{{$Bjb@>pxT~A6M5rg@2a_p#Y&}Gx2~`ycy%bH$ctpI_sxxv
z+oqkYK~J-Gntxc-Z1UH^YS^|?V>pUePgdx&uZr1~`*6ORZ4ndNz&*SO!1XcRErW4O
zucM0}uGA~Fn5eH+b`N<wJ}QqyJ~$jrpBu8w<j{~yj7jzBi1`^S{Tbh29hKU|rG(<l
z9jpLcND2;kE~S`^qIzQnVAF*!C``2g%`fx|xBhA9zc}P@W??g2UtfrKS63IoVSIF)
zmGhsl#PKRL>P9wkk`U=%Bl4(qc#(g%(IYW1&$qtxFOdj3pss*QL`!{-zP4Fml$krh
z2Q$uF16!lvyd5xnFt-38U2ycASO|pP_fk9x>G1-OPXg<KS=ww8-wt>M0RW7FAAyVp
zp{%sf1<={WNSLqfFNxyO%_bFbiz>K`ci<3X$esF8o7%^p3Ts+703)KowS%{B7jJh}
z$CG?vTOx-SyJ<#+69%X^qJZH=Bp<Z9&qTjha~}hN)$8`5FbjwUoD418{o~XxQ&_Vy
z_{=0N%N{ii1imXKWJ`~J90Z`KMNz!LGAH)oOR4P04o>Ge-<;kT87Jm77m+oKxi!ZD
zI_f`>EH}XRD&gaY_P4PeUgYL`OzZi{<}WrJ)u)AN{%s8C63T6UW}>_PDY=x4gunZ_
zR*>q;jdNZXL?3!ZVA5)-x|Q$dXFaRG)km#X1v@<H2K_}^y|Mk<2~iw=Oqu9mzmKa7
zHXNxt6S49vM2KPc_6k6T3Ir%LUZx6?8WR@D;T)jyN5czfl@enP_#wci+f<fcYBR7u
zt%3K=><VZag-R}il4tQ*bmG=Zu{3#Pzx!p6p6|0;NVUI1;@eK0va?B8qJNqH+CrJ2
z6~c+Xw6TxlmagnHEyRs{#h##FN!66Wy?F=UJa%C^OHp0(o~9pmMA;8C&;)S=Z9<|7
zrQX|8o%daS4G6K80Z5TvZ-10!Pm4VX9m!SZTAsAXN!sC?vi4PzGh<`C>k!DV%*i<x
zML^m&`?U3uBAz0SAFQsg3Y)6a@{Z)X4BOqhoxA2eT2tc0HB5JXsxrdDPOr)#vFm8b
zU6$Hu(EhjV-g_FZ%2TDSS40_gFIoy=x{i~nwyF9Z>DCXHj)S*UE594?iWKA3Pqzhj
z!>QQR?V!bWH%A6@6BFTi<v4|nGyX@S%un6cygF1_Q|RU#B6ya%%v}o5PV>@N%T@%X
z{p(^WbuGgi2!tb~r#fs{Tr9Rj_T9na3Uc)d4I072t2_PglUEc>QugdYmsS%s?CoDY
z9+ND@o1RQCp3%dNuojm-w4XV)pR}p4IA`K6WMk*Y*wPc6qUHhUM)~lXO#QUJpjjPu
zHg!iqC$uRo&0LiNq@;_)rK`lOh0Mq8Pm(LK?+>3CS~`9d+xbtU@Lg4V%R3V_lY(QD
zRK}6v<leE-qXydXZ)q1H2R;6T`|C4WM=r@HhP9EtQO9ZSxroGYb)?ijRJbc0Xupm*
zvp%PiNUNvu9gB}fj_L-HNB*&?{4|1iF70lA`;7a7;X?OY<_AJ`%zlQRAJ<ji8Jd3h
zq-(jW0v`(B$m?!)FA%BV_wWPx$s%uZTdhZV+D^2yOi4gz(GYA<Be~fn<q=CDbWwVV
zEP^y|FRYiAG#D*Lu1v0TX3qD7M{794hjRfU$A*@du(BjZ{c=Ezlpt>oR2=so&gau0
zjtgNt-Zllp0j20eex$Ay|8&tpDd#PGOK6##lB*++rPL!I<P33sZh758sy3`EAXQZ0
zt-Zt(y3XG?yX7cHdFJ^f0#r(wQgc3vj}Bj*nVp`RnVF`Ab8R5LZ^Fa35M7cUsALB$
z0a(_@sC+Fx(w&6}f-F(7@_JeFx>;&^@}HY?P9810Pw{$y2uyPwU{%W40*;+ir>sjj
zg522X4k>QMvjyr-5HP`9Zwgzy5wp@C^>vk*4;^%Mqmyi4K7DjgH$qc?GTD2O{X1{d
zquwA`E|2nfLc6Oc_vqiG?Q0e$l3=F#GVP&JnK`Ju=nQgbAF}eYQgeFbp-P%X?o8`q
zW<NHXx5SIpO@cVadtVv)sm{jGZL^{i2tD0bdK7{5Z9exKwxGLFAeFFYsC-7Q?VNsi
z>Uw`3Qp4h+)3J8NEMFXvB=gp2WV`!Ex}VrN2VdncPB||v$Tch{ZR%c9IvzY5+eHRs
zBwEflvY%&9?v*_3RO2V|$ka8qdPyVI0w-f1=3VC}HN~vr!Iml4_`a$B--)j&OmTr3
zGmn9mmz0=Cx%oT48g<FvrM@5CS0Ln)fr?;cVs@nbVbMRJ+cG=xm>x*<QA!HCLT}J0
zS`o0`$vPcAj8ea>GqQP9zTY|+xy}`OE!q5iG)z)n*-3oxrbE#8q`Pfs<)QCP?24wz
z(|4SS9KEJgccnCkIq_T3ToaI5LF%k>>Q>Gha(b~`D*H$`IZcx);#im0%Xin)gq;UQ
z{1B09+^E{D+HCk_*244zm(>GRQI>=nB`3b-_2{tzCR+yvnt;OhpVXKPfEt4ePkgvS
z4F~v!O$Q|7&)k@9AcL*-$AM#Q^@f?q|NKyt&MKtNS@ws$-A+cgg>x>hH_r~I{GDrh
zH0SI$^>9Dxt?}Ve99%9sGzAG7cQv#AwL0?3)v;2{y-Q|ROV^U!`q{wDh+5c}?uRT`
zHV^hiB?QBsFSPgAd*?4%yyk9-#H|z~1e@;DdaGfZYU0c?yq(1!iEG`Dx8#>YI#eLa
zkYas>eqv*p_Spmc;Z|ZKn9_70b<f1s&;i(De_@ga_xyl(8$F0l$z4v#e<6mde>V(y
z>$g&kCBE)|u=U$m?U`^lnQw!#`ru)(yVd#vgkV_Z#*S5fX|tzMWXi9Nwj&eh+PNGl
ze2sFZo!xD?KI8rKV*7bVd%fO_o7+}m3ks1(S~b|d)QgM$mDAiE<#z6@ii+>`Na=G|
zS5=I3qWP*6P?jK8lN?z3qe*#FJo5F0x*ioosTUB@bXmLY^+FxUa%D#r;4N6so4Z>X
zb0pf~o9|b;{`yZVk5i|<usu<${34}TAoC9Toc?{Fbe6q2pWahdJk?^-{~~?_dQ$p^
zdTZR+_ybNU=N;s&#<@1WXd0d}IN9v^WyXyD1Mp)by3hhB?Rl<Ym(SzJ^cyyc8Wio~
zdRz@mqIz4@J1H<z+_a2+qm>?ogwJQt7c6}lq@C+46TJ8~7bf<~Tyy4H{R6%|(Dcjh
zI(<3~wPY}4sG2LV5*qE>f+}^?ZCGg*=rCREU}OH*`mZEU=B;za1jOC6oB204dsUXq
zoZ|1IYNyGE@0^Qx=N1KX>JNKo;tpS>=`JI3LSM~5;JU8Cghib;g8NtLgJ8y|PrqB>
zHU6UBi238x6n?7Pm2Z<wf6W^Pd7t?u*crcfWEH%LfBUc)weM<u)JDY`%r4pMbV3uh
z-)%=A2=tP)Nog8vU3#asF9goYx|J#<rE%ja`miP#I!w4@>|Rplt(tT_pbUKd!Jx0W
zGmLn({D0i>Q(t_TTTMXj$@M5&lgs%-UiU1nnj5|86RqG>Qi2O&X+e3R`pd0Ae28AZ
z3AS~QVMgwy_-tV&3a+~QLro1d^+N+4t6&#ST90f*7wGG&;Zyae^FhHpYna83RRr$6
z3rD|jx-GfAVPCC$PERS(PQhO3J%`k6HCKMeawUsjiQ7wEeTEznPE@qj6J>N8`E1_n
zA5&MkL@;;N0|$#Z_bMDYz(I*p=CpOI=4FD3<Zh@WIN{^_c(s~^PW!Y(v!RNo4X9zv
zke&NIpMLlfMjjmzJ%I`x0(%DSIyY_YEmue8z1x&7m)p*ij2<K$I;SY@MGI`xCR7%8
zv$}#M37Q&Koi5#mh0~O0j=WK3Ug<&G<_e{3sOUjI)mx2#_4dwNHgWG5b8MHYOiNek
zGYc<gvCrx|c2F*!{+=+^)OfyB)`A?;lwucAKpYu>XuDYIN?J~l*1SmsFlQwPl9YuZ
z=?nBXnUw_?`E_?Q>sF*k!!?qwSFML{O@mt^kZ`x0O}rud_(LG3#ApQ=XV8n`EPcHq
zb(eUcF-OG=HUgw(Gl19(%wjIA$>QRY%uHXsA6W=;Cce?+V$<hfkw0@q5^yyG#{p@)
zqjK!;|8t~)?@pV3((tpQw1TppE$05~Qcix}ep6RUj8OTwM!PEaO<}FI8PVo^z(!KI
zIqC1Mm3Un=0Ln|ZkPS5hdw6%{)q*`3{|*>9)Bb%ozbPVK<4%&J!lbqMMG~r$hA5_y
zDv`*gv89+6Zv$W{_-++%XlZ$313^ZcdWos&leK&p?D5C;)WP1YoP`*3t8*j0KGa`&
z;A)@&H>d<<Huzauv`w!(5PFZ?BFGu7Y*wnOkFw9k?(o#T1zIkr?RM09`vY5MV%?rN
z_gwg}UAsF`ueJQ-_&c+9dy`O$qF0*_-n34v-EFY7Rwg?vjt|@0eVzKl2{UzX!+Tl1
zV(m$yhMTn=tgMkV&ChE~+74jjOi8b|HUnvH_0^@l>FOqS5y4tWNZ|5b%L;?)1&uMh
zkx(ha6Q#U*2vGJ@+Ktw3A!Zzah-k$urWO*E8+LMNtJ^4NBG00Qd4i_FIm+e}AzbcQ
zEgFq>s!+c00~5$-zYu!+Tivm0x7$H^6<4F2ue|`W>^`Rn6#72+=Eh$J5iTXhydEf}
z$ECzDyd)~HbxvZKSO#>G>QG1y(;Ps^oBt^k1O6%y(INlz8LI_g%KvyPKoi&~r2vra
z=d!y`0g1<6a?juX$9SKnM<n`uY&|AeeY@lajAzCBVGmhRj^$EIy!XRO+2g0fg5(8_
zDG{dYVtEv3UWQ5!80&s7lt%kX2djh8=f7>DuJ3FMuW#~~CCZ&vM+XvX$s5Tr%utQ6
zdN`-_FGD7-rifCzNNhP=kvbYe*8up4rS>4Wq6A1s0Igv8iOJj+;5si~2Cg(!!(HQ*
z{awd3?AQSVXAgz+QrG_bYGu%b<*@t21D9E)y254^!<_fyr9TbH-Ket@=S)1$yDnKU
z)vo@gfB3y!svB0P-rq5YXL&f_YS9|EyM0k;R=V}EM=8s3Ju{1e`xy0+W*NOeS5ICX
z#$HJ{N*))62#@UDtM4c36xQV3%?okAA{qt(leO!$508PMo=*<4kMECzCrdfNg9Rp<
zwYo2TCZ49H>x`tGo_x`Uv-e4J|NXzO_5bed|4X^+8iRn4kl6qx-PZM}y)t=qveC44
zomubZL$`Ixrz-{J-PS5JaR-e7O!#jA4E$feTRr&O;m`hx$Ev-<Yn)MGKA&nFuoR7q
zez`Zk1VqUtT&tD2`zvv#{TJy;dPdGJ)Ds{4oH)@;=`K5y=zvi|Ay^^(BUinr8R5hM
zY=MDsv7k-J{?|QkUxG(dq|$fF$Eir(R*KS+qDS5ig7<^xQd9PB&F+w#=fb9<t2f*9
zQ)zin?O$)MUQQ4t42YiaUDAgaRi@7dVOlCpJb~(D#$Tpy&N+^v<8v01;|g-Cze$Vy
z&g;#Ma=Wn9;^f~K!-KQtpN(p7Bz3~R_uswS2s#mkM%gy3jL<2K4Dq!$hxonnHz(AW
z;O}YckKagBu0Q^(+Qb4j6gO1#L;zq_WM*6l+6&j}^N=~{l7qt>tBY(sEqlD<r8j|l
z)}TjM@^6mHF|ucbY<bYpD#7%_<NFuCTc0O0QdCA#d*}UYriO-EcSHR|Y4=1FQ$63|
zJ=3~|$o`|M*Zyy%zSSC;9i4BMxmIg5EvL9*y@u6#w|{FlbV~lo+H+B*j@}-tiEC~4
z%qhrGnay{uj`#kO#<^QiSMN9+v*h<np+z+Nw;O_lX#lDX)E)o2Y@j=ltDv7x*=3j8
z{GYz5znK=sY^C_)4=`f;=fa}=X+T<Io@zK2Hdbl6{R6+N*1xhKIR7<n^e=_`zSLz=
z66y!KX^`l5_dPR&sMf>d=%UmQ=AEk5ZLo2eplN@udeoaNR`&1NQdsE7wTYhIGuIp#
z`M4YxL$X93VC@o>PhvIcoBxKXVz@)sW=>9Q=;l!y?rQZ$c7`gTf@-&RUHe{k9klGu
z(c@#CPvw1((3Y8hekz|wWFA1p{R5|M9aL}DQokHBFA_~}&CGUrcJ)V&%FNz<Okout
zW$VB47pQ#I+OkV;-Y9H$=7#N|G|L*EP}#XoFYX^aIcNiX|CHRWegAk7Hk?kdYHNPP
z=08emjZ&9N*bE}xs`1g|r6>l=$}ZTv@@(6Bnc%|wG*y|>EJfGbmUXUTX_)*BEdIDt
zwTB0j(a0D$NeRv8T5uUKbvp-A6qsCkwBDaPZ)|9?u=8<jLhuL@Anrma`#)HHZ!U5;
zEVDQYs(_|n6@>X`WhTt~Wbd~%+rBBA1m8AjY77)}SZw(%5d?ALeWHFSa#N@$q;KjB
zMEH$rpfMQtdL?79nN$YKVfdj&WYgkw`Jg86uUCVnu!{U}8^yhZ@WjX^Mq@uR7*F&M
z*iCb?W;iw5HYN04-HG(c5*@a!TMVYD>r3ombm5w(@F__@KMWaK=UT=bL>l~X|Njw2
zy=EAbY|`;)Uez`Xv^o@*C)jQ3IrHtX^e#)<d-WsG!8GLu)e=9nw$)d@CKiUCD16zr
zBzDC-p47oR$~&zXvkM)hazfwD;QpRH!}Wb61N4>HF7=O7w^!kY`b>!u>E(3Yk(<jD
zpBlbejIDV0sSE_u$<qVFqNZL((&f0ewi~mI3sfVjm&N6t8E7+(0?F%N1bM*`q}JeC
z`<8x+YqoqzN%kG|#b;sh#A!aC>UH8fd*G0hI{t7O*V|RP?KA*{saGI1bY9h@nI<D0
z@cg-m0=(t0TQWH-_O%1a+DznoO}jl*CSE)0L_Kw!aqHOc5DVMhr>-j`GDM#o)RW>$
zFu{66v^<uU@@kjePn|dG;+P(uS2s7}9PGu$OI(<ORu4gHMqDmsC*Crlk~-))n-*86
z=bg{4Ni063YRDxao(&vGI=mPwdf_G4H`!}{dj+8Sz3`}L2YmNnH~59#2Z9^s-kNtU
zD0C9Q_W$H;d$o17F!ZV?fThVxgaaRF4-{`|7K`9?tUA<Jx9B3+TN?=afGp8+_kdQP
z4GTf3-83ViAg(Ci%*^8Af_#ABj|2c$L?b}+<>KNF<j_f}nbObluM_vdcSryHSpDyF
zwf^<(o1M0E6-zq`>`JTNDb!XoYK?ER0-0M#;5t<Mf6q8mL3v1Us4b{~Wn}Mr9oT$p
zoMh+qL|ORu{+oNc?pMh6tnH_`+db;iKTa*rdmK9SDizsXE3wS&tlygF`=uWo4J<6Q
zxg_efRN{>FG7K*fQEoc!TI>5KH@(Ps{JmKJqN&cp)mcmZGpl%gpXu4*)${2U!6S!c
zGF~n>pMSKmK;1Eq8gQre=LI}SfSsTJqUZH)KhMBnS=n4IdrKeN4;$l}H{G&~JjpDm
zUk3XKS8|ffW}DDk2A8g&;?*jxY<~fZTi^IxRerIzId%Ojp_MojL)>f<slk7n4`#}^
zn%hNmA4F)x0kk8x{sy#Z@z(64e|<2#pHd!;F6$E-H4Ke{D3d(n=sQ^TxA%A5Z%Ma3
zB{oOdo|UV*{I2k=#iwIT+AePIe8utJ{t^y=J!R5xvp%$RvlQ0LIK^H6pa1vm{;$`6
z_1UlQ+pnt-ROe4wp!C+zBSm*#xuWRh4{r2-Ntv0$fPQ1NO(mNisGYFPu2i`#qVnGM
zV4Rijyy*}3`RVTd_%I`*H>VeRqHY`(+4aAud-JF!uXcUdwr{nLNRfFuih{@_ATxy4
z0hG~#G9-{x5D-F$fQ$*yS`n2=LB=4d)esRvgph=pB14Q2A_NE!2tptXVGe<W$?s{;
z+u^+Dtn)kHS>IaU`sELDKR|ZweeWmPd*A!IuWO&pmiEjb(yQ}BeV}uO{k`^|nA7Z^
zh9u<4crGk-l5%=tU7H5+Tfw4I$JVzrIU5F!miWlDf(;YUcqdpB-nr6WqFM%{^+wMw
z6e;%$rRBC5dWG6*UEa%*RooJ4`zR`}*EKg<kPvSzY7s9bMIG^+m=6$*qScI^24W^^
z_g%clu9wR@a1B>TUUy03?8OG?2GiXUg%kLiJS*EHfR0G<{dmcSn!Ld&wP*1^l2@0R
zE(@G^vEiy!%!;}eAzEF*yc-ZHON-wu&`9SCnN7j1c4xLbFLqDQeA*K_CzBUaW2<*t
zyfd(k@8HqlH`>)<iR*dO(|s$_e94K(xn5gU;+?Il+#>k3KZE>F<KSnpNbwiz%2e#X
zMWlexWVTBq*Ny|I)#Y@&&C`p}JcBPGo6i<jxMc#oCpYz)PzWZHlFN-ZlZ(?mQcE7Z
zUW`*gY=5`&f?KeuEwL|+58qIC-*IdaMv?qk!<e=1N+1tW9of)<+g?T04S@<`^=0&`
z*qVS=bEbA{s2jO)aR&NO*!W^{)q1S^W31H{5)we;Wpin^DI=mP+JXW=bRWj;_an)b
zI4NyiE68tf^|VZ87Z-;2Qx!zK^yNCW=C7_23g2DDn=JO%1e|tHx6R$K`9;X(M?1ab
z)-}!HcKX(9=1lx=qyQ>e+8?T35I^lHYa+)N2(Uq&92=(vtDV#=VOF8iklw0!qKx|~
z;8^7L$C@-XhMIwkO||6dK)VmelgKqAEg(3W?R<Hkxe<By4@A|LaPgftiJ^SD*T($9
zWO=NFgEAH7`RhusGFdF2a8TEET87T$`Pcpb1i}6ld!Xv6trI)2u5#?dy#U7dtAhp4
z&DIU`G;V|kDL5-gF1$ZWP_XYlD=AKqskcVzbVNuS%RGM<yx2zL$&`^Yt({^wCBn>E
zp2$4f5WRyICt|5LI#UJW#34s_!M3``H?pq=C<G7g#M3o6fEP_~Kkz0k2iSsJ-z;@_
zneDiL3fP1fdQTM|O(R`+e=Dw~{*-}Jb%|3<g;Shi`^wFizwN<K^wS+Lih*xTZy@~!
zaR2psQL3F0CckYR09ruc0n*g8y$DbOZf-YgX)OQMIrVQ#$KmrB7jngLZ*gOH()X=5
z5Hs#6uL&U%pLBBinQK!iOvSltMBS{nM}<K3<#F}F%!00^i3xF{XLe^SEj5E@MqRy0
z?X9MjmgV_vknwA7HGCF7@d-bGNx08ZwiN5gz=N2nhvDk!guGM1(wo~|oVFxom>5vY
z1hkrt-m`xGRl==rU$V3f*8cMAe`kI6xS4FR9$AmsFe4=0#qR&S@gvYs6Y&jaleD6_
zdnn3%2R`_3x9omz{fnM?pvr=fvZw};B)5K;NCGH$H!e7CPs^nEZ>rm_S|q)zI?4JL
z@y+l5{onrBm*c-vAGB+N{IsY0No(bf6jq|rJn1cR9c6EYeA>g@pLqUy_rJ6d<z745
zFTj2yR))_bgK9sgUEnqPU50>JX(t<pHk@<$>G3t`8t2&oWeyv`BY^%A@lc{>w-Jj<
zevT^W@|_TELp}BU6mN^0Q1*?#uAJ>csNA3xgl>Q>BO~3W75!T+GeS%`o+rnjTdz&N
zi_D!)6{8xnLUHC>TN$8sA4oXg=n5fpvD~!0#K;+#(SQEUKMSn~lGL1<gO5cO-sAe0
zMr!cc=LEA6mfgbVp^ADz0j;CwrL*vFrwl#i*tUnG?1C2}<V=D}1G1{hpdGthe)j}<
zf9pl{FPI_CzVn)%j&qq0T`oXBSd<w+nZ7*k2grvGxJocx0`AG;sQ)zO<6BMff|_uR
zqIfRp_Zg2~1ko_EQ%OJ7zUf&C9%K!EzVPdBUj3gRXI?D0M-T?&f3zgOT^GG@8G9^&
zL`^!SJ)K@-dgL9OtJyw`JJ8!HWFXvlxq-y}XA<*l+S8TZ<OVNH6*{%-WbRl2U+u5B
z#F&Pf9f>8=^<#WDnFuq%ZDksg-tp;IsJ%ukd8f2UlkK^Q!Nx$p+CV*}-IKLCH^$?~
zWV}~=7hRD_CohT%GR(mFI&wQvn21QE%o@f`tBv8`uHh}Zc=oSOd=`xTENTCds+cb_
zB5lg%kGN{n<>Ech1eyz#TdK$15598m6y)o<2j&?z-dxZh&mYz!A}9?B5Sy@Ww5HnE
zkUm^dStS_i**-j>yl_m+6#4h)_q>?dC>bt%1~IegJBKo5zcMXc5_<MdSsd|y^h~X>
z6e~3nn>H+oO(+bSGG-@52k9BmSZR!()VRT_Gt?4QRP{{2d;{MJBkG&1nqj3pECM&1
z?NvoQ=&aA8Na|%9ql3?}Uyip=kc4;}WN1`<Ab=b!4Ey3Y<lv#dAqV~b9XZ&0*^Nx|
zt5g`4tJ#i9C!3#lkjsr;yY*d`m!nJ3o0RO&M({fCBa1SY3>gtt+s%6P(14N2I2U)M
zf6|jN4l}q3@r*HAUJy-*HX05bG7hAp^QtA}5;q|X26Nx5hwxE|A}_nuEW-u75*73C
z0c#+k@BfIUs-u1xI4J{CPbli8uD>GwyY;yqit+K&9u{chDt&q*T}>sjE9MTqkb^OQ
zo1W2JGXg#Ddd9>|S^a+Lk7EmsZdY+o1Xew=2=DJ;>V^1ezW`XOG^a4rvH!kQ8BM|#
zGkcb~m3hyjM=&aR`!e^{C5Xn2w})44A`Mb+HknUEhlj5ZHUx2bF*{9ksw%Cw8$ZI+
zOkmj7L{J~DuWsLQ-`h{%&01Ys?>;=db3!xPkaNz>CZ9ARoO04%7b|a=;b6Lp3JZy0
zzb@Z58yjhT8cUT=h<V&HF(YBD`20xgWh7LFT*cXW3&w7izYeWSPFB2`-VLk@E<-wj
zMJ~4bH$hbp?=2$YyICrrryASrXHy>uJ3>89O_E-%8w<#1^DKx3zpqth+l45347x(C
zkI2Y>HnO9-kYLYW@Nbbsrr_(wp2m*Eke7Lx#CnvyP&g;fG;MNl=^$uqKn#PI@86sO
z-~N|i14-QvA@RAarqFy}tIGpI&V5eh*oSubW)tbbyP!x(!?Ugg({+U+K$^bW{&*Y|
zl(;zD%O!YkzfiondXEbYGVWSYjd?USZL#Z_F>VeufN4&j>Rjd2jU*}4<|Z~L2S>Og
z^b}9ez2+;6otRVFFpwq}9XCDDKQubnUWO-P;BXtB+TjGWClhPxn8*k24+1euxTyo~
zV$Oj#0MSdrV6^MwTF<sGG@@RAj~AnJtt#pgp_dWf4#x3tHypjb;tbirvYi}!e71Jr
z&Q=-WZoKSNuHaTbP-&zvGu}B7xp0*iOMI_tHD2#2Z}`shH=r)II~W|BZvBLx*X;f=
z#*nS0mz!;DUK<uEr)FAf*fl<3k>FHA$%6b6{~6XF5WO5bMSjDiYJ&E8>e_W*-u7#_
z&$q~Dm$xNGH{poIaAfz;03G7*b1+qmPmguI+_FNHo*8adnE{VB=-sX>dGVn7?fCr5
z%zh8TL5StG@$iHg&;Ibo??&yKE}V+7fpl(_rM~WmJKt;KjG#OdrX@%EN3EpvhPLE&
z^7u<a74Blf&ipzcPUPaVQO#qwx&pNYw>xYQKC&f;L%rVgJ}t-bYPxM7uv%#=@$s8V
zVA1;cK!3ZFdD&)QLiK|Us8BLw1Q|u=7W~F!(zSX36Z6A=x)b*7`G-bK>#s?8)pwGh
zNY={NQT3Qrb?F$@t+eHJw~+cHZsl5~D^^QNS3FgJWLzUcagfycj^^;!2>Ffci5l9^
z=vQ3iSWIk6X7vTO`USlGUH%4QF7KNGNo<n;Gh&QQZHBR1@u0*vh77Lv$w&?dNUnfM
z%d0j+3vw9bfmbgxm5{4#m^~S#!@FVJiqju0Qnv^jTq{`(2<VCsw7>;vsU2dO7LZpz
zQYefC(0oDF#Y@*(mZa`&L|7`K60)ZBK2IP%Djn&_wZg=5A)}a#vfeU}V}w&opSM7L
z*vp$b3=>PY!7zm`>paifW%h^T@}E0S;Jb-Gt1GIdpf7#T?w_;CYOuXcF-T;8yI}m&
zqj~X?1B&FT`Y3XeH7%+Xlc$634Ra|fpw6Z5d+15o(q=XBLBpHAOW9dHTTErSmCStJ
z&X$sl3`j>vSg7!|&MkPbj__|^q}H9~==qNtmghJo{)fmPVTxQ!MpGB@C$93K^(!5a
z=uWx`*=ZNaT8+lw{KG<=Pa0d5uw!&<iuQuOW)W=;#K=z5r-G=Z0e5Ku5$^6)rNK6S
zqE437x0~%-o~4a19DnSs4j?K&2kO?iD5R|4Sop`-P-Cm2+%86Choc|O*V%HiBJl#;
z)kCkCEDxS5+CYqD*xvonS(|G^!kP&iot>hw(s{^0uHpCi-oVP>jj2M$Ssp#RU2_$r
zy!&}V<Sg}_4~B8IEn1_}19tccwJR|ENoj{l+*7<FoUXd>J36aS(WwG7%~D2;!Jor)
zfJ(H`jOP2SMSp0Hb7Ir!-JT@(orvuheVgiI35IK(w%r4rFJ4e(7aB1xyWzuDS|+8j
zTP4K<2DVl|j90n4g!4N~>Bz{~7pj7@oASRjNL`ayXBw>5E=Ct~*!n(+u?s3sCj1w{
zwI>n-!H^9r%y$+i;6C|=A&^lS(4k18*MErkZx>%^AK6$5U>}IFuio~J3{pO0nP(jl
zReT~XVC{TDpfHp^WnxwP8%QYkf>(3fSWjN(T(A%8o`wQ?1%SMaQOB2>?!|x0sSHZn
z;5S;$#cd#`M;E4U-ySYRcPmm0gb85~SPWD+EkJt+)shZ{Y=BHh>=E-tQKj?gQ23LQ
z(<>8NP>zKW9ICNq=Zu=5OK+a&If6GhQ}CCo`{K~X=9Hy^hHo~Leg+Stqjb7ayoFTb
z;Mp|phR27<YruRZol=4<&Gr~DHU^TWh2E!fgG=rO-1|PWS$4E}-z9OVh*}lE*g$<;
zo32_Zv}h2oN9<%?(1MI>YIP+9<xN$?rDbG{Z#~%T(ouJf6-12k%7Q(bjSLtw^8rNO
zG{o^75|OB027&2Hwl!!4n<iEvU%wOo<>LR+dIuXW+OimBY+(kPTU_t`2>orR1M?h2
z?0Ac_llEGc?vm$^p40ff_0MM3lllf)H>C8o-;U3O+p?g>ThDR3kq`e`zkeP4Gv(+1
zvnyY`uiPCaecGc00H&rdj}PvqFt;;ctGb)`MZpd^Y5HFC@iWu^v6&-H>v+<(1q;+1
zkrEguy!E0qJm_nICq#|*B(uKkadYYZ+2I=^Li}4?h`P=K#3N){y<$!r5*6J4emWvr
z&aDr3fOY+`VLxz+1es>a>A#<f%X}KPWRhVxAh>z9L8@il+hHS~dMsEr>{MFTd9U-y
zl%V**$>;E-rm!5uY3_8@Il??b+dm4ePyM0wp7#&GyZ6PnnE_0iz>s|NPSPRh%;NQ`
zf=7MYf*m{lvob9c_;wNE)?0<erm!K@Ch2-(vp6L3>ck<&LR>}IxE)scS{Zl|apL_O
z*uOXSA3OY?uRgoX3?W*?gb%8f3f%kCLc^^sYJ>D%br_~;kIq{WtP>Vs2Ij5Z1XaJc
zUbN0`(1coe=2L6Z>^dl19KMMV$m@WUUYl9bzNAeyvlwqyiVm;ec3&1s@}?+S0+4Me
zRt?nTC*PI}hZG6GLfeAWM;wONr#%7KFzvYAZFsM3!|Uv`4+j6{7;rp1Up6l@BC4+j
zmnT4%4l*{mEhja;+B^<oJRkB8vGdwNUJK|ejUSgPK6<Y>33}-Bgg4fR@Ob>Jx}mh%
z&f<^n!zoyNJ*PcdFa34g4(LQ2W-IAFg11((uQe==`q`kUsmU->Nx~Dw4fa-8tZai!
zUNDzq!O0I!KK`2@|1{4xqtFK}gGN5>$u(2idjMJ|j6N?fA3m93(C)?gC>(O&mhNgL
z@g9bF!RvMV%b0Ee-<S???7mQNumD$@K=O6NAB@AN0NOm#0_<~G_(zHU`2^<~r7;7k
zc?e}O^yZ}bf@N&EpCI8EIElXU9RM{;g}6o(H6JEC(ySZtP#%tGf4P{LE^-;Nh+e+2
z{ZrS?4ajPW>Z2zqzt5TrQs-JK`7Be4kH^4pV~HkEskdI5&UO8&VrmUEtx-6s_ccQJ
z03l!;bV2ZxVmNB)r>a>J2!re<23s%=EXF$9<$zI8%T{^t=F5S6pY9NF;%?~dwk<k%
zQ*lK)6>@<P4b-3sa-Rq{r4W>6`9Zxr&h)8RR5xU0C-aaKR75>YB2hD;m3c|_=Mk$D
zS}{_NTR~0~C(f{Yeft1weo8EEO0W%O@^|<)l^BOld+s@fPLx<QY#9bHc%z0t<d#MM
zOV~e$THBx%wtUA@(>z9lA7btU_?K|#6{1C1Li;EkUP!vPoK@WrT;}Z7s`XAj$=&UW
z!9s;$eDaw-eLZA!J$%#1YarS7>rZ=Bp&4ikzmf6CF#C9shMJY#krcM+GQ>0^?qN~n
zaP5PN4m_k)L?96J)W%HWZ?<FhN#%jp2WfZW&0g;0)p6t);jS`o!G6fjMP9Y1p0yZD
z-(;PmHwLd5$`dL>+6MZY7)a?W4I7uX4eMS2Vk`tYycO(J0KM4sCYKq^Pw#bfVxO8%
zti35N3-P`M`Hs@RY%+J{WRfu$SE3=~0ha5pdRkR*#(lC3J_SfX`wP_DGPOfTr;90A
zZ}w2pw(Z4_2<(N|;fX`o?A&3Pl_nJOga!fIo)IciW3<gWogL>wxOLU}qh_pba<9ow
zg9}g#-)Mueri@X|j#o}^TuW;jW<&lltg+)t@bSWPyy=QY_FV-B>~InP67+vwN`1ZA
zLDLS1Nv97DPa3k<`{#-C1`9KzL(rj7x8h_5Zj{iLbJ~aF^IdVg$&3fy=1d>HuxOQW
zl6Bs?ZcACIV0j@Vh*3r+<N+cr&(L|TSQtph&dGZ@kQOr8o+sxpAYOIrZ#3{C2Wp;x
z$>WUN2CLj^4Y?%8UQw#s8jvr84EVuAhMW8V97{e{M(q?-e`Cw?na@13(`pPHre|5-
z=?UFk4F$=onk?-Pr3yD+qZ;l%L6}&jE8)79@#lT+r+BarXX2FD*UuHTj{p;+F@WoF
zY4lxdUbI2=JH<8n%;qOTzZxAIPE*}_G(dc`C92)7L?_&>DKn#6jy@`&Qjo`5p(y(=
z{Z6!R*q*nXwgFI!J7sJOD-_ZE4G*)TERQ9o&JOquM<S)Ynx3*{H5hnTy(#CWk7Cq1
z?S-=L?PottbX6SOABD0hZls%=yH<uLa?Ul?&TjuyJM>-hJ!;MjtB|j7s4qX^mthVI
za{Mmo-iM#ecORl1HqBP{w*yMqC5_Flk0;N5tVnb%i;#gDQ<i%`gUwS%4*GrC6I-OP
z3N}o1`#{-{&NC>4M_6n-Xs;J%4%LOx=+ovS`L~BzKFZ=GB#)A$uC5jKWPYQ|zJ4+=
z4RVe+AP%1DB}?4C`D^^YXlPOz1r$kG8%aNy?<+LIXli7{X#{lnN9HNz^gp{F!RINz
z3PtUeI*d$+3rce782g^-XPB?NRp&v1LcAs`&I~e<F{hLZ4TTdg`ijS9_x%id8K2p6
zPTa`t>8~NER@lm$q=1;}sd8+y_LYYF>%M43@<L@^I*%GHQ4N)6C`8PM@JiZlwua^-
zy*86tRvNQwZWn3k<@w9q-|g^;Bb0@(ObD5tAV5_DHL&6^%}7;K=IWT?jKIMr)t6p4
zR~!@oBvDb=jOA_j7L_PMc<xX4^19hyR(PW_AJrE>vy56DYDjMrt#XdUuY<4g9y6zj
zRTf3}Rvm2?L~*#>v;F(}4BV`kuSc;lHJd3z5-59cV|sn@!c(SD>xcrGkm-8CQ#E{c
zr6{_A+e4s*4{z34_E)p_HklVZeCK)iZMY5Ug1lnuE%?f&S5kNiG+Rv&`=F?{%88?o
zX0NClgHV&5*=^2CL1$m3{HNsho;{!c1Z8RY=-1Hd3z;A<Z-ypm2x`?nItq1k>~XM8
zZja7qX!(<X#a}Ru_3A=KmD;LdE)JEJJpH;Tp(~_skTw5OWG@xECY8H3?QadqT_chZ
z<E?j*i_P=!nn2EUO#{GL#>})Ho5w;p(jhz=;ivZI&_`~^w!dAL<JBiNn>fS{%quA@
zz{M--i9C3wtKSpe>{{Pe^7gKAXx?UmH6vW{qSw{GZW`~_p}T_QfsNtCtMj=I-G*#)
zsmwF*SgX?6_#3vzBih2kO<7=%VwkgQ_FAZ&!2Xm=$}_Mu(=U)<V;N&a4}UmD#`mk(
z#Po1+vtbpn8SZ^y45<6#R*0lfRQzm}YSa<pfo3n+{1kJyiBJ_=a+yc<S4E&bDB1^?
zmwtuBkRq&i9TI!q-g0SDc1l{?&{}L)8^Y3t@Tp-&7Zc#Z%D5G|2be*VS*JIDIsL!0
z{x3>sD<*7#)pR^}N*6*$k4_*SH2_m*pUtE7&d0orhP3d8^r17?&hRLym?7<JX?b3|
zYGfjcHVD8D-OGY0>`8dRi<scZ6v8)W<o+2d>uspJ)?GR;GS(?)SKTC~ne=|bNL<#L
zmgp**O&f8qfAfw<TsZSxePwfH{R=yO_v$cNCl&tan&3m}sVgUoKdV$HZB3-Qw#z>U
zm#oxI-nurQwBKhC0G#;iK36&?m|E}hIKyqKe}4?2jcSfbpR$QRMk4cZh~OKDT5#B)
zN`S)9_rceTv=shFW9PPS_|@BytMe(IrrIye-cFg@2fXdMcSCfv=2v7)9@{V%1hekJ
zSAoyR#%+JNhlMbwne4%x=1eUdw_@dqw6J<>L$*1~FUc}|)h6ah)9HL6uML*USis^7
zFmYIXuln|t8|1j%+DcGDZEQlC;b!BY84?w{tFA07+?n$9PPAM8%g@hZTm#(Y!ySpw
z${0+$U_^up9q%{8P6Pdd>rlU3xX4e_*fPF~FRCuSmDr|#GIRco?0U<>L~|)LbLNu3
zf?Q{2LCsd5Ezw&{JUan=fGg=?wkH0fZk@!Jjba4t*2gV<q&g<i(_O_?K+(HGF57Hq
z3w*h2J9NH34RkX9FGv2cq}2Gs!>hv+-K*T>QQYVO3Oky5u9Qu!qY_<0t}iT}zMdO$
z{nY^O+}d1k>B-*fS-~%hy*d@%TIW~&Mhk;hG`hh^Ib}4?c(hK-lk<sA<pDzBng&i!
z8!nZ`cY!GnexKA$2?&Tsf2v0hfF_VOAqQGJBV^V9H7qCBUjJF!OGzls{Hn-|)U3Ot
zvj1QrA@a*<0A-<+tLa#wK7`~8OBEv<qRta?^g?i5t8_a>n*)OV@Zcr*B#}(GznXMW
zazY%p<zy4EhWQPq>rJ0qy3(?`Y(GWd+*lm(AOopJ)nOxd8wuEtK#1#~_lgk)uKV}1
zr((Md7YR=m6O%>7?ZTK?SESFB!D4f|&v#)u{wH^17KUXD?HYy?vBnOAaF}VMNXVHf
zX^ylut&izZnQ{ch@(v0RAo6_Z8z1tkjwWV1M1|$5h!_tbbHLx-Ke4;SvK|=T<pflO
zlC@dI&F?-?)&?*Z(By6Rc%2m~8rv41K!O>VW$t56S_YXYh%?69+w6--*pyDp$wk4q
zP&zMboU~eW9TCFYTALDHe&kE~IXa`D5Y7l^SCo$prDXRP4DzBuO*->K1pg`MSo-Zy
z*50-@r){kCRL`*o-~I5%DF60su~d-&oo*VAP|cH%T1qzsIdCd=`XR3s;K9KYCff-i
z#e%74B)_Io_>cX!pUx%bfgrX|XhY6mBXCcd$09*VR*zP?OW2H>VR{cPaB=HR2CtX%
z6V_U@QMzL>SwlF0{xy8OrZJvB8JBThv^~S(MNuNzDiBi#_rpdviR52JTsR5op7Bw<
zlW`$rSF^tzrtrosh}J#CjH^%$HMG6bi9Mw~-R>92lEu6dN24r@l-XmCYcEhY6hN)Q
zFR87Qx@(GhWgZ`Q&TM*Jv91>Qt$M6)ps`OV#WLs3FwW^2#3{FPldLd(gdR!eV-k=E
zwiNB_!@jq1?6F3tlq6^wFb0@|c!2fxN9Om`LPq6c$Rs_bG-T@GLy5IrLI#ZC$LjWL
z$LOayhCQ6ImulHAwy)+$S-tlFh{L|~E>?G0nLxV$1f>$?0sjrixf1Gt|6(o~o^!Jk
z4u@NX$VEr5>FF+<cB7EV6r2l}JHX`@RlkCKc<$mIeD%5A(<*nILs1sp0RZ=2xuJXf
zE5A;CpW4+)CO~#DM7r7JGgQSL0;IpQSm*+v_J{#6Q3KfC^#TCZ`<vJ|Q3MAMkO_X|
z@=I`V+a<VHCIIdI1Jk={-!o}G_RZf%{O`8qcN6wesa4B6i-bj({EY%i_1vdDW+R-z
z$$7%6Y504fEc>px<LlhKJ7wP$h9(AmM9y4KxAkp=$HyEC%pz!O<!*25-QWz0v)0sM
zj*&t3`bdwuy}B4BjAB$|$Fq7fWo0oOb9iO)(frH~4|3`Rccf;OBRkR8SavQmiDG#t
zgM+S_%_R$p+6D@=AuN8!{iN(*!Y_V-1qpfOVxjj-u%_k-Z-^-{3*7me<1xrqcfZ>H
zhVylfxnxoN^n%vD@z)aP+02&EhKEE!+*jv8iLmQ8a%)pcKvxT;(@1tO&mvv<c2C9B
zeDN*OlB>bP^7u<*8?$_C8grtKdFnmfX_u&M8H?|V%U|K2K*wI+#ZFVGadV1rG1Gs1
znYoTp9yD1^S}6$amwlyi^56dE%pYb-rix4cwMBzVX0PKr&Y&}OZ(n!V?&+u|++uhh
z&%r`yzVx~|k6!nCVfy%?>i4y<?8@pJ=Y|fKDV96V7M{a7j{#XJHh5Eg<sh5T8RqF(
zvYG6YAVY?kOcV}>zhAgfysxkm-sRT{)p}Nf<x*Ci?5I-IPS^a@_x+o#eR%Sn(RXZ`
zg#}0{D33Q$$(MkZGJ0yuyfV8R=0jXyc^M$;5RIt*iz3Ms^Uc{myx#NGdBe2hU)h1e
zfwJV|3Jk>;Lg17i)Zj0a0CmX{{&vBmwyv&=7oED8z1H@ET?3Ye)2BUqck^9)q3Pk2
z>^D6B%qf>ELcSUl4=oDj4lXL^+Nvt+f%)F!N>9(sS`sq8@N6RI;v8Xll|v25)DlId
zV#0S#=8aKAV$xAxJeFtrA_y-+?tX3QTsbyMtbBVL1%bs*tvCk_Wi@OVR?;IjB_%nX
z5-a!xhyk{a-SO>?_MCg4W%T_Lbfk;p<4=3uH@pN}Lotv$wE2N**8viq;G`tc;lqWq
zg`(X!?e8wi(*Nx4|JL}w2YhW$eMF1Yi~nL=IIQ327!!=B4xd)KSzhx#eex-s)jYM(
z+epc7k0uRWzc;2H<QhMCHb#;ftlgJZqoa%KzY;*D5Zu0L%QPwMoR3SRv}5UGfuy1N
zXS+dW`R{!_n&?wi635rnA>CrT^@;dA`55*_j=<M<)U(u$+};i*hcM3x844-oe+1ea
zGUU0WH|=KaR((W|85BedH;dFg{;P6Adi-u0vgg1e4vpO#q=8iE)C=vZUpw`|FekA1
zD228PjlOloEhOvx;afla(bj+W{H0@K*^#rQN$t<<`0?igll9z_Cd1CsJ#(ega<nE@
zWA>A81Ya+-J*WBHKye!X{Hr*-vo*CsQJmv!SM#8V11E5HZq3vh$Rzp`M2SOX89|Iv
zD3!jd0@`)3=NV?9V5&YO(aP#dqFPYbO>s7_WE<%@^^5IuDpfq6yCY*Ycy@l`;f^$K
zH8GAjzEze80zG{bl>GJIeBXNy%1BZKR7{mkA=js<wVOtfF<KF-$=*|GTEw(=M`F0I
zb^|~fvEepp&e2kGYn{`WD=x=MW=k;OB(g1eeBSmLM+J1K38?Zi`*zMZDDa7)35`m%
z#v&UodZ1C$km@<q^aczAc3nJiB7E~R8pri#_rU%YiD&*~OsTP$qh|1=1E89h@8Ye-
zW{wx^Xz%(bpuYJpqPh<^7{y2mYnh5lCw_T<NYt>9gAPMN2g7r{i`4v8BXxsE1_KN&
zl;lkn59&PBP}J88TM%3=)AiT8qM66fj7gi6=*vH5P1Ph`h<P{_yDeU?wg4F;{VzQ3
z8Eyy7Lo88+wT7XBAR4m5dN9KMl$HQ;v4-@qg`A?<)U5(hl9ZYYzh`C?+@7K#ai!UE
zmm3Wn0isR-V48TA5Q9p<=zqpX=EiWJO%vh;@2G0K>xq_~t;h4mRd(ig>x%jBa}<%$
zP?O?`hLysQ8jEPXu<mqAk9>xXx{I-A-Wgk<J~igmj{Krc<iM_#S?Q?1d}oU))+MUC
z9uSHm!elq58>jcl4!@Aqkv=x5*AY>$1fS+iQxH7W7oYY-tnIru)fBdLodMmK$Q5*F
zz!IVAHVgbf;D~iTd{^MgyKgM$tvp7<U*4X<#e82QpH<h=g_-_bc;p20%Xiz6In?&h
z{E<YM^0PNtxLu=Gi?sTx_Y-0I<rND?UOpvOrx1!ESEB6FM{*?xA-c)RWtNH`*Vuhw
zWpiwMLDMg5=_A{jo(oEGX66I}L9fCpu%FEt;;k=FvqosCAeG_zO^ju&^}t{R>~NJJ
zkV!y$FMS=?E&o{1Kv_EYexgEk=THu<GNT}z|I&GdWoqTFuLI7laI!F*Bo63F#}WjN
z?WIs<PuJ6byE8&Q8`pZd0xBIB^gA`n-cl+~)fBz8Mv{@n0`q#5PCEW^g^5oTWxOiX
zLh4L`3KzO%i+oZDZy--iG04qzv=5~Y-2>cYJ%>c&+TE!xwBu$gZUm)eIEY}^Zv>BT
zZb#I|8Ba`DcI*4Y&Gxmg)Qw@vp5urell>1`@F{zBYZJCCbkLHjn@txR4Ou*sIx|RT
z$BZD@Zd!^qzv$9rj-$ljA*#_7<H^lQMy^DVl^{XDJ*?eG#|VBWf0V!exBb@Ny7jv`
zYw*y;blPBCFxshO8BPdSE04tWJ$)`@Sl8^D4MxZYWF5WdZ*sYpj2)$aWfwfXMmJ!d
zlV2K#A&=y>QGZ4$F3h*95jRnf*jP0kE1MT-122aI{Z3S(`!~RWC&<F!&GU~%U#PCG
zg^fAbV-h<gn2U)uCHGOo8X_izG-2x`zVE5m?8W_?U#j$FLAx)%ey7(bp{OYezmNjc
z56sLf8|CmMkn8WD9zY>=)B!nQ>EtYB9{0BB#<Y~Vot*JRWaN>Q))XSj8{c8MX))(<
zIum3gF^x|$AlxlZx;V+isGdk{vKhOKZMnYY<iQxNOV$<@%#WxxKdO=Fh2~+G+$h*3
z6VKo^`riltPsh5MG-PYWR*7ofy9k<^faEP|ep6eXOT+imARP*XgGqH81vZ`OI&agm
zI6OUGlnJ-<<#4>-LI_+w>dhcMtfbGVAIMv;$DyynHwCjUROk7TqCk-&%UKm_fGHX9
zdd6)ZSl)_?1@&hjgX!d~vmI(W#f(LIZV^QZiReDfuNK5=3(U?pyUoi_uH8n(9*Hz{
z+RV;qCI^Rcx}7UPTu?r5)O3(Vd^8ZXV7`LFd+N5yRvjaK^G6>J?iKHejqr`GuZ$jj
zILU~>F=3X1!m3XCp=4s2%sZdfpyeSwcUQb#@C|#7wEh=v>`2b5&Xr>)bMI5{(ue!1
zhR@g)n}B)^9;tvbHgDvy)Jorb%9ld%w}N`B1%c~}qKa%Re=bsGv?N<-5-oX95q2=S
zVY@uHs0e?&w6IqEsGXSdi_L-2ykerwJ2;~Tw9mwd)<rCWj1X*UO``nt)QtQGguach
zw&%p@XbYWlL%DYmq1iwD(SiT<^Zz~RsV_BcC>br#IUC2hePBbjJEW5VXdbNIj|a;p
znCQPjOTXeRuZuhB$EWAs)H8)3Ej?Y*+oPok72rO+LdWpSe^aFX8Ktcww%llt;aGKs
z18b?I12KZ43LxEttXxBp|GJH<aRBx$%yFjhN~1@ubpjzaJ}3)mnRKp~_WtbQkLJj%
zaBD`J_4O_VMVkSsoSHGa$s!Z;>&db1ZA;(|gIlyKOSpSg18d$>r#qJda=ostiryZ*
zhyL+@w%&XQOg3SS*XD!nkKMQ(JA#%RE=xZ<;4}W}ibGnMzV>8n-Gl|>u3YZwc@sGI
zrui<S^Dx{lbGBb+ax%Aw9GG^xFaOh?J-a2Iy`T0xf;n3pCgcxHSb3}`{>T7tq{R}-
zIF<J59o?an{1R)W(F}8&d3H13Qw>+by1c!~arLaxp01O!NKvybAkil9hGo)ff=gln
zdVcracA(^3;>M#&b?>4l(sDap{KVn*%%A^^>7Ka!`Gj@I@f(jiz&^0asu;@Pc0xK%
zTQCY8gAU*k!NKtJMI=}71b+f&c`o>hZcr&e@RKWDc=jhEnMA}9NcB{V8l|G*P`#Cv
zB2i)aBb!Y_*OCJ0!H;;fR8<6uJMd_LTu_iHRp%x_I>id$lk|81Fen3uo#F!=Kb0a3
zm?bT(4smL`;Mr8L4BQkVN8-pA2DOD7sWl|`0g`Cp#Yx0c^l<F3C#KlpUhs~6Eqx&`
zt|<BR`C#6D{eXM@XBc1gVO(5&b}lxTK`Z04=d8DQ`@$BsnL0H8PPw!ib2tLFZ(mv7
zC@3)yfA}YR@TWZ(Sa&-}`|`?VdXW3Le}$+*>fR<wyl4i2S+i-BEttLmHOo3N4P(ZL
z-94BP7&JHkzW>b8zryx6OEFp8&h>@G7HyoFSbP~WfbWbkjc`^Uq|gLn>x6V$qNvmx
z`~fG06<iU<UEtpzYW_leR<NxVubla%-&3knd$IqDFfS-;(lpv&(>4Y`&3)SQ56)WU
z_#A<dbal&zi|&euE>D6xDQI8b73p~GuAb{9Ab<GFgS`d2`#rA{t#vn(z>Qa4Gc})7
z4|iIxhyb}l>t_w#kluSqeqht}wO4e54I}ql*kDNN%F_LOkbw=N#Iju{tfeh5zdZI~
z6YZKnvdW<enw+SO_$?hUV$=Jpwfn3j7t3mZh$(oUEf9<G8nq6zm5UxHZ0~;k%LARq
ze|hn*mibcN8W{b&ZRB}JPKZ0s@rn6TR8L4hK{xP)OEHIVb6KXZOc&Uux?G8nE`0gs
zp8vTw!0wescfV878ipIG%Y);PC{Z!v*L5Yg?1>HP!S{0!*9o>4p)7*5_m0fjnTWdZ
zcPmAJM31!cuySFfvG`)BV6w&7wng1|?p>=^dQ<CUl|tGxGDJ65e|<^$35g(e3FRnI
z5&<jR*PT+f&S2TuKAaoOT;h82hLGyeH_Ybt4~+JAI&Ib$icLi%f9yF1u`;9f&D6`?
z!ke@ZZuXi`=Fs<(G_<!t>Z!m2L+;?J+gQ4RBXEM5s16?1)Dqtqx03qX$T%W@%v2nh
zTsr1L97dJ$X|n|cQnpb)F6Fgc^&K1Kn~aIq%&$5X;^7O&ZkvDF<Jq?v*pM6REIcor
z{AtfEGdSK|d|3oHYtkWi&n%C+x~ecUpCz^rnl~OE_CW6-G9NT+kkqdT6*Nie-uboM
zjBw$*%~W>q#F&A*Df4))pY@i%qpQEEHnv9^k>BEK`(kW!*qiKm^Q(UytlzGq#>l?k
zPZUqw_7JjnZ?-YBSR3&DH?uZbBfZ^oLtENST3=8F^5hwIp6VNcWPdQ%+p{nS0x!H>
z(>}RC3|BmqoZQb?L-kjT&!oKaP3kMotoMZ?e;mR<V75)$Nd_I$#P^f2P{>{@K;6hJ
zdjRwOAWgo+$3p^%FNoTRB2x@wZSpEs8q4mFe3baOXM0AfKWMbJLxis7Qw|eQU9Ke1
zp<;t8`GvzL23I}A#H`#dPGh4oU;js82P(N1+JUN9V1(!JD!#2E5s~Gqfq6D<ee<SH
zryR09a+62R?Hp}VKeJ(=7%?WNPzOX6gctr#nul5Ayor4;(e|4$w3sGLrH`_-LiLs{
zn2A+`dcKaV%+RZBUmAFFg??YS$ZTIVOjSGLY8O7wE77g2BoH$zrdhSnY)0UOg(ahu
zA8`xmsai&S7vkLD?th4hMZB1GPBVr>_Ii;u2Z*Foo8LFhUH)h2RhO4VGy4q0_aiH;
z9+`8wCCb;@Iyw#BZ)4*EJf?kl2_Rfs-*j``YH?bDUaH<=SSVT5FQ;9(Vcp>wI7i3p
zhGA`2_Zcg+<M!n{*!xMYruCV@bl*k)Xye&Ci@2X`3?mZon2OWnsFk>iEB(B4jbf6}
zhrDvLl@>CbQ)2UMr^41gj5IT<^>oR)8)ST*lmd|_8EC!ZW{0XImUG&;K#eIz5e+V^
z>$yciKQT6TG8YE6mq_{4ZFH!5j3f5_ScCdf^NS0-5j_<b)3h5dCH6XT)~pe}srAm(
znQJ$m3{srP@UDT#<4V_XwePs>!t{gi*sE0B>zLPSxb-p5t!aOeeEq82WW9XASQEdW
zH|pzaEiIkun96&@m1sfICaPys>@x>8xZxomY4Y?u-N{b$-DPWa)d6w3zWk%s)jQR<
zJ9NWATAGebpq7<!>%qqKAl;NvxO+k`YdxO;pQvQaeff7TyZ`F0{~!F$NPZm;>W+y_
z<co;Q?&JUF*#NvqzJFag$s_UPSWCw1Bm(Lmhrjyu_om`)up*N;JFT5bXKUhBT-VrX
zyEDN_MH9E@lP*l}+8apZE*9R>d-Xqf+tu<UYUrmu9RwhG*J~ba`&{^Gk4D~hDyVd;
z4qy<1NyZJDZtv72asF}2hu5u6mD>7#Uy*H?`ueX@=7}S=3qPp+{?|R$<6E~s?V;?6
zYh4HMDxrIXDGRN;gP%=HaZ+~PeDVLy3l+&*VUX*L>F~{fmy9Uub?}DwI^uOQwygWv
zK?3b5DK@pfAYhk|ssrogoXJj#(#tVVx_!Xq&nSEWH~0skWIB8~eR;OacZ3-O9V<m!
z5$woU8r`=DI@d`o#H_O!I}Ne0e2%cq7Z#qyiM~tYhCgulEB*OrWB=K(MwLO}0tiNj
z6JnMdlPe<VFX+Ew5fM>vXc18SH3H;2`pT{n9fD{30S1=RKD>1=Q#YXSETD;y2LlKN
zjV4>MdmRE>F|rGp`noj)dEhL}!3he?0OUpkZ{XbPH^vo5et#;Un9yDdRL}zFVhD(R
zsz*N-r~xyUkx|nGWp(<ukzN%*pKKW9i!CXj3AXEVY|V&y;VQMExES_uT~JnDTmVk)
zcRkJZXa#rh+`2|Or%p!OL>fE5N^yHfVF#M3XE(QIVD_TVG3Hgm0`qwvt%w%aP>6l)
z?V8(j!kX0^ME-cmIxLTIYoeea!}w`W0~1q8pWEk2@3RWhX=Uj7I!sg8ig7~2ytH9*
zRfrQ9$Tb~K7_KFV6L^3s^2eEoOQ(PRGk!9W9~D`To|P^|<7z^^YL{Z`%X__qrsH6O
zBK}R|@Cx-q_9Iat*~Yk6oOXZgd@;i&y)F9#{rfwN`@OLjxGkT7dWb&Sv-?_^CM_$N
zs1{D6N70V_J)D8;AbXq=wKLkBgF_Dh=sCzedf1tI1%jeUoLWpe0|uh|e%f=%^5weA
zbQu+^!bunOxOj#HY6d41#jJD)oSUVh)_*b8q&!T^H!<!_Qn=MUu;2QdKmGo@@mT>g
zeYX^Zop{D6y1B}jZRS*uzD{GsZHBmy)KD*RtS2!5_eYMg$!Xej+Nt8FYdlee-O*v@
z$AkGqv*qL&N`IiW;P%M0RBUDk<zs!!jqzqnTO~%>3!f*((+nhJm*1m{-fSteY<jVt
z7CF7nAo%8E*~;kXVrFmB-H5KFf3(qNE}iis{QV=(`n2b*ln2IjnK7HI8S))Yxk>)*
z!r3gNO@GCh)cIV&DX;TqQy5U+yWn0L;N3EC@S(x)E8_vFf8+#}?fk9*dUm66S7Q|5
zCK*kFdHzTSVeu*}q$`;nnj{&~{fmuy16xFVqpN(T+M<>mQ#`ebkO107OySPj<6kD4
z{vM!zu>)0kU$9>{hOEfWcRaD{g(D(wal>jOoeIB>*9(DLlTMmFiXJ$cc3ffTqp>2o
zqf|FgGVb{q7<T=#<ALmjf@R0K0Gn%KzcoCTUTCyhWDjfYl?-A~RqUKuN^x^&y)oa<
zcol33wr*>4opygh1L;h2d}8BqEv~SogcKJYla|WbFjqHpY`I`HeB&L#cJTYoJpJmV
zBaJTv;uqiQ$o&niEXP{+vXp?_v#hj-RZ2j}pI9QNeFtXQIi+@h#>z9=ar@h&CDp}e
z)1EFZ*AO6AsRZlF<ls<BZK>mw1HYVfBYOcYI#c%y(%XE+#8hKih^A?bSu`ALuycyH
zR-?Gc{E*foBS&dkA2VP~(!3wOoF1PL^2Mcwn!Czk3wraQm7GIjU<E}@0|14?L+j%-
zkYT~L_h3i#62M(F_7K@{x*gA8d7W^)`Y%!b{c@i>P!zp*V9E1vfH>%xNo|DJFn4q7
z+VDy*2aq_5BW*o=HH@%~hLR7gym@z3m~7Q)*|%tDQLi4C${2+41lOJ`(`SMzxX;lx
z8_)Dbvm4r$^<F!L+cYV%MZ|If=Aq;Xdc@WO=aj#@Ldb6890`qXuYF#*!1*x59-{d?
zaf<H7wZ`ac8OQ)Ud|i=?O_Xi}MD&`e4M`S|bu^)pP9~h_%ctn}2{TO$G6Kt&w=Zt%
zY^jeA)i7Mq9nKBQ1MpKeB@ep`n=I#iT`lG|a`et_>6NoaJ$lVfbFNs6$PHu$9QCfJ
zp^Ok{s-vuf8HN~b8*alkeOIjBUaDIhiP{xDFY_5qWi3MB$Ap$@Hr33=_P}yK8gn^3
zw+9>%A6uIDkDDzno{p)5gY9>hVU3$e=9+VD`bcDL8=r2{%a1BW)&~lWP1-AA-IU@H
zw!qwI7#^3}-Q=@|BT5(M2+;o?`5c2IQv0E_*pvmi0vnVnEF<;{#T<NatuVJUgLp_c
zi!+exlOgjhjj0@8&mBE=+Vx~;sLpq9#|0=NLmw%nj8XD@^J&jSCeIAmq2GsUf$&=7
z%<6c6jMAdUz_o@@l$1oS>J9Qts$B3eawp!-ZKm{K0}ISHN|&~Cp4Tf-4DL>@sXN6j
zibziR;hYQiXt_Gnu26q$_5^7wP*pe5<k{$}OvCw?9V+g;CB;fqsr(04>9dO;yPq!a
z$|TpwyS=D@dks3q6m*W-<?z`Jq3$2>0A>&|O3L~`Eh}VBYB+c~y90{H{s?W(#Mx+t
za=n}kQ;zqTcVmcXAG>&=dCcY#KO&;2f?Z^(<mD%hO>FM)Jhen@y={%2q&pw6f{cN5
zmgppM;ZRK6;b)A6^z?EPraUT9`PGV7Y=*qYSvYS6U5l95vckwn%MWuH3*~%flC)?A
z?ElhJPRYO3!JinIi@W;6e|4|_zx&<)KlKFv;^XHJn+9C!!x2Nf<qO$Yiy=N^pU*2T
zD8K3y$yS&oo_{}C@ygL`o{4ZpR+qZL7H>qk*p6y!l$)1WglU*%mbh4Hp2o7xqmgl$
zOK9<@J;|O?ofif3?rZFGkgb827n7cx`I+$fFMqx%BIsjYna?a&x=Ga>RUf>dGZLkf
zV(Rsw)Sx8RZPdh3?>)D?cPGw299fJDpHkO;zLfa-D5m@9{!4#0@!uIo5Ars~*)bz(
zREQ9%Fph7YE++FLy#<F5$<#(VavK8mHc~EmX7@}Hr1@!&@jbCQqAt0-vZl;~Klc5v
zyioi*M8q}xlO@tm6OmHCKud4V*pT8)b}U7%@CzA3!-Jjk{epyg6%OPse+kZf&+ZwM
z0OGW9<)qazC+}FUJ=`X$2IaQyIKX`v*8i(#KCjYQQqi1g8sd*bPjY#I8k$^!pRJ|j
z%~qnQ4Axyhj{HcwLMO{8&QrkO{`yDH_ECQQXl&&YdAJ-NTYewqH&|7(Ew^vq!)iiO
zP8Xjfe@)jgwj7aMO=35`yNRqg4fY=XfVY9CKI)rw%ME_*sDpdAYH4k+-rO01DIAGW
zY@eN9Rms@p$HQl5W@g6eBx)_W5>OV|OQAIB7LfS#TP4=l$V3xDpij+h{tzP-29p&A
z4R&T8X*zTrGufG^H<7;rD21B>@{auk9j^j%$q9|F>h9&IGuQJ{KJEFA47Ff@y_~x+
z3~!5rMG@=SJ#`!P6&}4gRtOLFaaZ<=IYLV6R0%aeNfu8Votz0WJ*a(}_I*v3<qLEW
z`qDi5@+PY*Izfda-BY)tH_&L?%karhXgE>~=NK4bz`G!q4Zl%Jn7j2&ns1S1f%&eP
z4N)b`%1CN9&k=ulZgW_%%=LYJw(>{^SWJD^MSf60Mye$YR##p!nO!Krt$LcqUvxp8
zca=8zgWvRzQ2+D!v-7sWV-JE@R?MGQak!U#mFS+9x*b&KJ62lN=)y<>h7nR5@5Y>(
zEj6hqGfkP3@H{Si=E^wg7S9D!^tUwY32|d7-E@{PFm>^c>(#h3-?uCZ&10dRw3$dH
z&%)hvX(MgXw(72kCf^cGgxQ1Sq@&xgQLEg%Qx0NU=WVg(#>y6fpfW_RqG4I{1J4Y`
zx0m{PjIbdGufHw*Q9g1qI;6nmVeWzV6i2f+0pVgNQ_Er4_4Zv)y*Rs8^?>s*^Y}7f
z-;t}&9bo2hOzStopK$M%X2UI&M~E85g>@=n#mwW4<Xp7h0E^X{?H>fXvdIV^Z^;C9
zmjDiD*eCz?%|S-$UL_Z&p;h-Ix>1|M5+ks7MouOh4%SgmDZujJ4_g}&{M(A`+l#KJ
zZ6s-~<f+`+PRQJF+YCk@DJbWebh<|A-Y_V^)^xDe9A{IWHeFp!$;)=^UK)zF!rocf
zX+@50RIrCFb~{N+ShFpfRVQM#pi#(2bNC3zC@82)ve~KC;E`_LWRT7?6id+2_lq-Z
z)l*3BFttYOq<vJC*a*wKm_I))|1y-orR<e;Y#quye4;{U@jmVWAgI!o_f$;GDI10`
z+#0qcDL2cY!4{pHRh?4~t=G8~OK%>h`76HmS+!~KH>jhe^#*Fp=Zflz)Ne1U-4j9p
zjj`H~ROIZSaPko^gV`Fw+Gk4>^acsnEixKYV2J?RUrk+y*Tape2l@RgGW|jDbRH_K
zLYoq1w9>H}T^@(637y2u@*8&3;yJlZ>i7XfGtAbZtNoPs^2KldbYGotKz6aLhfIFq
zF)=qmu#{V0({=^)ct!YR%X7DS@cUn1FY|cWOLr;CV%S{a->Z6SY<Dz#lAd|Qx`1@E
zMAxhIH@sD2IKc3%8aQNhlO9X}G<ivIE=!i+7Z`X#om@WbM}9~q7y|idfM5{|&^?O^
z3N|l#3XHdHGk$J5HfmSeTams@I}#XPy|SI0kV<&hK-PVeHvxKRHCEdhSji@zT*X1B
z-D&8|7mnc$(s(G#gHo#7Zv$OTZ~&d%2__P4kJ<<tZY{uq2=GPs@;mebYUj=)(O6>c
z{ec4<;mQO*I&o((NVeHZEexg?s6nF`Y0VNni(jy6igirQX^rUUu}R*nX~u<^4xgGO
zlkFi&g>4RQF%wrt@D_s5U8x#h|CX5e&yUZ9m}qS2>Rxgo%Z%XvS#5~2YI-QNrEI|*
zb!fdAL-V|M98A+mc@G)&UzN*2KUbXbh4kYTak)s>*WUh8uE~5!G4=gSWFsnfvIc}q
z9V{7B?p#r>w~jgWoSZ4)b`&obJ6Tx{mGRLxTeET>6l8)lZ?urh2=zpadn9Zxx1wDy
z6-4Pi6V!rHZB`0s&!XT2wypq9)HdA%(0DdhNos-x)=IrVHD(?QT-@0ag-K7Ik@#R$
zMxm~XulsTM#iuu(4pq-Z`G56xOsbH!5L+|qk(*9&gtYbGkr#maF)3Kc^i4tCrfSf{
zz8?k`!3v4cMym~QNiqIb>s@QNJib_`8!y`pE=~&N_6&_vHI7)p^zl0Nr~tZ(RbDGM
zV?l^qTX-K{K;#SCkPv<6gAjqSQhm5$w25D0(};I4DRUGlOD8_6^m^?YU}?;OKA#E;
zAHDZ6eJiv4m2by;4x+B@T4(q=nA)&$XX@>$+}_Y)OinSe&sDFnI0QGPeWo|LoqQns
zt({+^j`*iF{{ezE1X>uM!nV>!=lg@8%10LN)ayU9X%cjQ+5?{ScQ0ZQv_@s}K*_hE
z!Ex<*r~F6K(SR26{0>a7QdnfVn0|u$egSf9s;3i_h}A+#b@g=`uVE3!&J&t8{pHLQ
zclHyfANK6AN*MtC7_+a&TSo`F=VOB4PV1=e`2mUuA`fcp7^LHhgY`A(t7@$(fjFew
zuy@F{;<v>f6mUcJYxgSJ=*W4njJK2r=%mfpYxbTXq?^q}!HC?ViaSOfN~5Y^LpTo~
z6or=!)V$;HBiOnCkuRfqWtN8-;3%7i$I)dyeP?a(<eRuykbERkmbgHf=M||WT0C4q
z7Bpg~<_6#staE9kT(py09`ym_i(fztDcP0uMUBT>R4Px>99%6%b%v?+@^ac&OpVry
zZg?CWq9_;qKkU7ESXybCHlFHqRTe9;Q^uIYsGUvAC>rCwQk_NP-ld5~a49uv6cN$5
zAShLp#4IjU+;=-MY8geM2xtVX#Dy3{q9`aTMsbPYhKiyn`JHsn^fy;e&o|vY^ZjQ2
zXf7{391rlkFVA`2=Y5v@HY#@bt+V|zU6z4WY3h?7ijv~+F=VuX)uc&5wCuk8y8M3R
zt5$CE?g_J`J<7#1@0nwpL!6OCXwv$_XKdB3$x`}hyqgn?IzJv<G!+DL;iF{u691|E
zxpHMHLL(A7))rAhXC+ezsx_8KWWkbFfYct1p_MMFOeQ~e^|cEc>~>9$UvBz>R0m7)
z-+icNZcCwZZIC2a!g%e5awT3CHV0LXh3m8eSRBZsZ;uT98H*!t?UqO0Z}J?^-VOZX
zEXp66Gm8QV!iuLTo0$Pw!_1a9D<xGkT|k_#jZ(zHj6jQ9eJ;!|M20l604C|IO~de2
zn9DOaI-OqiC4}F6UJU!L5*Y|&+*~Zq3+oaA=J5t@6T^woPr?J$qpwdc`J?|zP58gJ
z5dX=U|EY?(zYy#C9{w*AFj_IbG-S3`9aVejAqsX&xZM%%C@LFK?y(vybIZFplXy|H
z*c!t-A~b4b*#v)}_CmYQfZfaVz#>IPqXgr9`o`*p>nMMtU^}R|x%O6*;$`QWIG}Hm
zhdU(Ay~2x;$D5sJb7q3^LK(2LVMt`%#yit0-|qaGkxBW5a7-;|`heU)=xX&gGH>in
zaVqJlXlO3kzb=(Qo~e6r)LJ+BD>nGKSm!6w6J?id)~u%}q{!~~4-mWCOW+SD>Q7D5
zfBOr2DH`L+I1W}1jF8vdLif03)CEb?Z!2bQus~XVWd`X2kxx}#dHK^rKwj+5<6r(_
zHvan4!`CGrmhaQwYqWz$>nRb}&qZsCvud)Z$IJtW(%(<eIC=?teIv~*#NpugzA#F_
z=*7Cmch$?sQyBha<6g#|TwYxySw(K8Fpb-Bn6MbXl<YB68}MzS21K-0pOD28>Lk8z
zT4pnaW@e526l}x<sez>(16{HiSIaEMiZU;`XAiLt&~B~0+}7r%@P2{0EVrU`baZP|
z+uU>zpf+n;BQDq+$EQ&hk!aIcK2i~SqCgkY(fAhkuLJtk2L&5w8ND)-;SO0U4<Wtf
zqT_aQHV?K)x?*}wYQ>3lrFCl|(_{Hsac9nX^iI!5opby{aFlNMiv6icI2mtMXp>2U
z?)wqklqU4t6G+ipUYPUNb=fu`cfgX$`4-!XZA|@4)h+58T`%kaEnnC~M)r?o!&LF=
zt<4pwm9;gs-JU5vTc_ue*Az=*)?ZLL5o;UER>x{5hm3C4b;J}OZbGFGhlgjJh<~to
z18?oAlC8b36P|ap|KC<N|Lbr*{q+v!UlZH1&#$3BXWjVHmnS#(F&D%06!_yQ3K8zu
zQX@o!Zfer)R!%+!S{+zBv3RkL{ye_9?%+ky<kfF0Y&*X!PR^y*3g<N2Ry35eMlz!G
z1~XgoAMvW3<>$vq(vgsv#C_eU)vIJTt1(<A4^JbrTHc5Q8R2v%VRw+-Z>^*){^j@c
zU!{#V@F=hb)PvPq!i}Jpt;ieO7;l?fGgL-Fy}qU4ElGgUH=%E?y8IvE_x-R85`$(#
zZ0s%k(P_$$IL&5@DJf&Hfy(3L^|j^|Q7^p^ZlZm^xz_t+w^~th6PKiYcM$omdz9%$
zvbJgx?z4M`yS|*{z1%LSO$(b^E&IcJLe=&`tMQTivvRW&Lz|<df}O@Du4u@T+EDa(
zvN+&&r+Fo@zwh>rx|lbwu?lb%jn%Y6P&B3s3Lk7gQ0gQrSnrt{r{hD3U|;NoK9%``
zaJ#{W;BWqq)c@`Xv*#Q#FQ0V+PLt@In)o^LxZ!D03(Y#B!xwc?t-$Zc*L15Dh-!sC
z2jbsf?QbgC7)Um^kc+%{WFn{eiPJNZmit1<(&Ys|bc-}&z622{+U~k<*}f=THNJz*
z7-kTEjPcMuZKgD?^>&!oRsyx!U=oVRomtW{+*w9M%xKE2TH7PzIfOWLV7L=8zdZA-
zUjUMEB`S-R!qpm}P|ql$qj6}-(0y}Hg>BB1COzf=Evb)3nCYIgzC55^5kg0Dy=XmK
znZnGX8|cEy?i+n&JP;h=(bSxr)o%JjEd*V1J&`PyB}*xI%<q!CxilxSkCAY@8h^ax
z<E}%;PHk>2g>e{<sJuO-+4VQ2era8PHZ6LNW0UJ9sN)_+aA9zMvlPBqI%M8#`+-DX
z!WLFHm`Ys7z64qc7KA&R7d*6EU_J$%4K72Wp+obcb5h;tf1T#fQ)y*@6Q<PrF5Z5;
zar@5wvXld2JM6K-hI`YLdPdSrn$OqMqysVOdd`bcb+_5q^CXR3HwFO`VbP7+)(A3M
zBO=GBcs+O9<^5WI|0q32isv(awyQ<9CRwZ8G0e!M^%=XaZPGn;8bf7d%W-uL?4#Ag
z8$yb4%YBfsBvG8TxuwwuA8OfJ*O)i9G52S?3~caN=9UpMW-UQ8xU;5av$$B(M%END
zNzyJ>F@1Lo|2FFTe--~P{@R?#G}kD2LG28w({C)G-lIWnQq51pE^fWv>Yu#IA;8dK
zv2J)8ixd%S@6@E{Syt+VOeZo;8q!Z}wL-0|EzC{Gv2zRJkg8-j-gnq^J#JOl1O=Ez
zB#=U62;<hnDwlY?_|`+9J3{N2kr3eGW;)$uB(V^<vq7>YN8Azo+$RuLl|KRq$|}Ic
zHpgvDEho{%wA$zVgZQGBqKGyFiQ(IlOX#yblGU074<^n~y8VUyWL>byU|7#YqsL&4
z$>4KGTkN@?a!Zwj#go;X=Jq@HrFR-uwk)3NZd$kOa9)Z>R8w->eRqnXSyOQho^yvY
zG!}d+gLIwk!K~tB=GM1!A9tCWk=f9GE;GDvbM)r)<kFGyE{?0x`&U_<tuO=qve3{_
zo1JfmU&;+Nj`aGv##OCQZmt33U$?qmTbunroYZyJvS5N}`qdimdYny8Yw1=s_xdGO
z-aWZ5q<GMzKlTSXlz!4;8<|!zr^o|BmY3}rQ)P^fMZtJg%(y4<W`99>-_ho6JMA#B
z(PYC#__o#UQEqX>O7b0d_PSPL-w;EK-{rc0`s1#QqTa9bhp&sN20O9w6-B3m0ZfU0
z5yXaq4fwcgB>2o)3IG!FJ{eoSz5i{>kfCRaJK6TPA07jE-KAw=FWj`2H9xcofh)CM
zZ5tYY+{L2&xPjWqiV?5{J3RL>0DTl_Ulg8buaFy#9sI@DF|K(f;3j2rs_2kRDT8u_
z`(p46fY%87_BC><@Mo3@W<pd^{$#K7(e^&kO6lczWfEcXbXI5N39V~uM0{s0SXAO`
z9yo1Y1HS8e0kICdb@_5inxsWL)Mj(NV^B1=ZM8-{k_L1SDwc2x9cxIEEKDXmm>&9Z
z*Bu*&<ukCmR4p4b*?u&XX87C}QQE_I1^JjA{Fd@&W!+0H#$Lq{jnH(W4@+dVFOheN
z9yNDgbar1v1$x}Gx_<gtg_?Y$YrNC&XGyWLwOB1&xE6b*Cd&NhY3>{ZtKB!r-wb{(
zmRMUCz*UTrR@BwwyM5NDNaD#;v`Jh24>mevektsd8AMjV?)KWQ3b>u3MwnKb_tdCZ
zb2}dVpkVcoiq!EezOu@W%kS#q3f{i$y{8AYwsty<$HPHCcZ1FyS?8{WWM%9>9L@w;
zT?=;vSphr<S<d?c_nzv-Y?5O`hK5-Ac3qf!R=FJ}pP6swoA)WyX5dI#xe4$fpamQl
zBouBJ_Lh!^IrZlNd-=;=pL(X8064m)I)G!KVe%C4u={Dn7x9O`xVQhIr(axD49pJq
zE+R$+B*lKrnf)8z1O3}u^mc@Z-BBzv8g!QuzWd?HpZZmm9f*o5vQ{$+F{7q@>7k-X
zQa*OTaJ9ru1NQ;gz`X&K_0*ee*Uw3RzbG1d5*5)HB{s=t$gUM*ONvxTo{al`xE@ti
zZC{#yRy6oC9Fq`hJ9w439M6e5>U*W$jY{OC51~u$HLcobBsIFfU-#Fx>{Wg&w7EPr
z#A_1w$R5;lb<_DoV{i+u)X9sZH3Si)WG#e8Ci7rlU<u@W92Q3?iQgVAcjwp*7vd%!
z+?>tpyt!=YTc%<fh<XSZ8@1*-vf)XD2j^r)x3E3}5-=11;pRDUDR~LkB1T5v)t({U
z;#J#PPUoRQ@1FX64wFE)$t4*&L-H%1!Xf|~6A~QYn5Z?H3`%J*n#6wI`7rG=M5o8)
zatDB$YBZw_&eWtW@BGVbeV#|lb<6NyAL%@OY1(CdKL)3Ao1t|ud4y2=PAUw36ED@9
zlo@}1TypnaMPa7=<F2_q$DG1XE2U|guwJWhib{x=P0Rkang2{|Y!PYCxc?R)<kizE
z{bd%(8^Q?G3`5{q9n%s^y$CYuM5wz4%8gXHZl&lY!Jtg*=hxwjaM8dMpsZlF?8bUi
z0ws|)X6KwS%6~VNxPwm~JFug_8JA!}iMznDs~1%DDh0C$M0~l1@ptUhHRgSgUbaSU
zL<d7N(t>Ps^HsT~&5&_0U4JO1zr_fMn*h0^m0+qX!ub<)H!yiU+&-goma=Bq7YkHX
zRKxn|%HuP!=!tWK2xIH=FNHN@vk56@s?Dh{Ti=LsLN+#C8JYMe;`H?u=Y%GaS!1_{
ze42VzeV(iHo#Bs5O|W}Xr9EnM`2^o;Ene$_?j468m!MwljAf1fzJ<;^(1LETq5GZr
z&9IRLW_Wu?tWOn;a9n+Z*Ef2<ta?k2*bW6-4@8HKO=`pQ5f2$^td`D>O;RJwVfN<8
zI|7e|wGQ*51IY#JF9QtDwb^U>N2l9RsS`(mBYziC$L1D;tFxRFul0vNem|>5L-VNt
zCxah$3<*Lm6go2_Q_R&fNd(kSZvq5fS^VCvw{Q7Q6t#r>1XbPNlgX574^q<LnMx<e
z=dSt<;srP?92OZBHUHfJXW&(S7|_6JU~`<$A2J#-=<4bMn5XnD&QIbSLzsC3j89aQ
zx9VKzbI#K=qoB%_5THa=)5;5h7ObjP>N<7s0sfUwyp(XeQ>Q|`b(fQu_qGhgK&o%G
zj^QEKdWW~QUv(u%F-Q8TmG@|O2NoO94n|P+5=d_vAG|M0LPZ|ad)LL2MU3RE-n=y?
z;#!!>_;LrMSO}Hz46A&iJ_$LjNKohRc=R(dYU8hrBB2<EtIU9?$RreX<ly+Nhm3rf
z7OQ&CAw0&0j)KmGJ&hk^;)mpAv@lcKLYZPWC`gj}ahG%O&@{|0JTi9+s-<NV(J#=6
z3YEYkp!j<^f11<3qWBoZsap_=?gwkiF~rfgK6vAwNVfMK`4Us=6Zg~g3rCBo2BB|n
zZH2Dc3$@TlC#S5}#+;(?2<K^&7ljKzUo}&Dknzhpy^WH5FaaNXBFy%aPrNiZCClis
z-m;uL6+E^=jt-`B!#ba86W0pk%$rCpAW$zj_C&~lfXl8XKEi%kV1|Sqgrt;fiv0U#
z*HEil$Gj8tYFjC;*1B)UVg*yR0cRN_8t6!as0bA(5NV6Q=y(B=G9Ii`GB5#uA*Lj5
z--8?$d2L!L&{WrsWh!H|Tl;-V>Q>7yNwaeSZc;D-QXoYes^JV2cW<T_CPu|V;I+Nc
z#VW$F)x=)+gT15+{!MZ&^}EqdpLSv@w`&$AE-rO{5#xJssj$1E!L+k@WTBp0bv6BE
z@ikl>I`%!u(DY$>KKf>7_=d2_;YNPA+xLH-+jrn$?wo2AS$eR|HY%}YeiKZ}4y5ci
zF{rV3rfU5rutw0H5qT1H)%Cb(d=ko~ego-s&+!bcV^*4metBUmP*DvvY?|ebrqpcl
zaGZLjsFtn#ib27IVfKNYA=E)J4Wq&*+i-&n-L&~d$!OIJUT{JSBWC!#AbLrA<B7(_
z#Djz=7~%xy@qBh#gx;;~c^m()+orZ0%3sHyHUJnM!#X8lT>Mke!@wh^YH$85_g^_U
zq%yfVkw!>Ni@vTKN63!0&6upk)I;x^wYDX|VTJctmU<zHoCsQaLy;dMyo@U_v)unU
zo!`~&#5Qss%*2;7LaNGNIRMei<#v(k!rLA3<F&Tt5KjNn2|KM6Up`oSy{bH7^8964
zndklb$Oa<*#)Ii*=4^jkMiFwvD!z5l(EWTu?{T}OmF3t`c+B)989qs3Ej}nINnvec
zFU_-H^WDa?P#aHu6vI+7&Dx0TCoh_hn)&U<@ETv1%~gl4Sk3g+&4SL|m|pgwUS~Pm
zWSXB0hx`VB`zCcBn{!t#b9|yt>^60Zz4&aze%WS2iJ$0Z*ci9y?q2vF#X8{c`Ea^<
zdtj)4%3`PIIvYSaDk=+-HBQoiH4CC0N>rt~2K{<50)7h~86NeXiEpVXHlX*?BUqmx
z6F4pfOP`juTYwu2P6x=K>3;K8zx>NQFCwm3{jSGWHj78!v#8d9+$)iS8cQCw`T7jv
z>ycHDL4#I4t`(LZK64&;|F2v<Mn1cfClag%^{^1>q_XsEKYyPg!UIIWmP31fwQoLV
z(b{4q3?CzkD{DzJHCibpl$V&%t5q1>J~Zb<kGJ93H*WgP7ObP)l%f{U(de+}TL|}Q
zG{N-FgGmW1+}O8RPJTABN2~hCgKlCGhN~?An}2b%DZtLB{#=g(UCIfoqeV^9pDkxT
z`~_ovCx73wTLm@x_?g<pYG&BVpipv*RvS#KbV|Qqt;xtL6YAzu@147ryn{_`AQKr-
zpJAxzj;QUqUmF+5awOX?+1zV6h;4|1+BtE?{S3_APhjZa`;*K5n>yZUlTtmLr}M7`
z8=HxWs4ic*PwXu7CpzsHBrW)U|Aa*eSjXM@um{mj#{ff=vPcby4>pU{k7*CbjCVM^
z6!5(Y#7^F2d2<VYMWg?_PyG+ohW^Dcc7qY&TJ29a2!4;YWhIC^_=7;&JW4c+%Io5S
zcUQ!coJ<XgYl}10I!^`69MFfcs6_tY^J_J+#CL+PS<BpkhHEefL(eT%v%n37_l@1q
z^gPqcAh#70-ngI}FUz*|D2FWPjHM+FWp<AbKdfs84YBSl4r|**?R`)!rXK~8WXK+Q
znA9tL`&<tw_+Q%JKAShPD5=8grN8=4((+|5!t|%cj)`Siv2}W;r(VvRQ4WzIL^jK<
z;_8<g)p&6~^M{A|eYMCqq*v+w!?U$EjU8s%%bzW`R!qdb!@FM|MXs;O)mh$VzOROu
zbAk5wNU!nyPUyQ|kh!$6uLm{?di8n9y=`6)Z|@s&{Oj4<Lq@qFM0U5+HzmLQi~0JG
zo__XgyOWAvl?E=x&AjR!4fM$OmG90i5}rg8VMm{P+a%CJL`k*oNcxAo)#1L%oHxGq
zi1d*-Xu6v}zO%E-8yrIRb9NJ09+ZX1uh1ZCdVACdT0;y)*eX6U4I1!pdc!b{7`Jdv
zR*8)3ZO>|1O16zmv2#%B2^#GOvXpDG`E9df<f}wmabHyH2&nwAu*yO+*6`)V5SMY}
z-6+#86y93t^UU5Qube&1CF>JmQ0=71kq+I$QDr5~`L#^p$>biYZ5~}HKR0-`qUQv9
z{|H5GbmTQQ!D(?duxrFQuIBa3EmsgePntH^n7d=_YX3SZ#gkR4d|6>g`?YlA2>D>f
z*qT&sfQci!We?Lzl;NG(Ba8GCC!)AM?LECm3rdE!fO5DRT#uJrl~UB<zF=3pu6Poe
zKb=@d=>s6{a{E?W7!W$<%2K&TGG8#as#>*Z8<|(I<;fV-+-LISn_av1UK*U%e<8I;
z6#eEtTeQ0M?k+fF(^&Y4>)WBQa7SzV<F0K5W@jIK4)-G=37w68^j6r+m*10B$wz&C
zW7x(+H>LWvU1oJ5<&$xsh#9VWl-Ra&@AYC0WNHT!&ugv*eVNL9lvhW3$KVCvwC=ZE
zijE3jh%Zgq7>%3N)=xHnIxmW~9wOJM1mv<}3W)9pe&K1%<nlXdEjCp)@g+s0UQ_+Z
z`UKCl?3VaefH@A=-65f@^GC};E=A4wOl;#IkCY;9F{^WDlJx@;iZ6R>cYR&ZIH##u
zc3|fk5Go)y@}q~kThL{6%?kmm@tbLlo1SX65M(71!GMgC3>&CuT&-KgKs6yaE{h+v
zuxJq3i1xaqWvgce5JTt7CZ+c$9Sl-VW23CRs{<~82=9_+1c?J9c*CX6eBuhxFPY9H
zI$tsLMIR+4rdt{Xv1{7Jk)5kkd8O+8Wu$&R*Ak1_oEaM35Wq34?qL{fg8U9(%^4Mr
z4-1MthRxtdNEaf<wH|pv-gU=*{~z8CKozdm$6W{Z?=($V+2&5_Yd9-j#mW8X#it)I
zpsawi-PGT~y~IX^+y(rc^Ko9Wb|Dh^m>*}Iq#3ugut>&yPQxqQ4ewRMvcbfJNupMT
z?Ybg2!R_KpMR8Jgx44VE&b`XO4_cFb6K^-6Tv`l`aC_8xCmFRJv6dGu2XmQe52s>6
zxC8y;WT>h*p_)6teFeQoEfvulooHxj6ig^7!KwPteP2rJF0<B6L6OB)gE+i6xx^H|
zGTpAq-f2Ad)t?8at@jGo9&0#W|5g}do%fZ<He#YD(>Qd-rqk!#w++KEP8p1aeuUgO
zpPub|(_Fn6vHgd8m5%2o)<l-*p)sGS`el=)KYZN9aPKrIZ^u=lAyiF;)=Z7SG+tH)
zB>lbQC%mkAQ!&Ymu*32SP0F4$Lxs!cAk7@rQ*^`5E!o#5L%K+KwzJ1De*WkRe4=e0
z`=}cioXa#aa!tD|Uq&`j`r*m7-s>NC-RzhgNT(Ego?S+-Wi~Z6G5C<78<3IV8^h4P
zrpCsm3O9a#e?Pw@2fhG@159x0XmrvAN2h!HJsjO~EdhotpPPBVNyVd9fr_Q#;9xVq
zjEwXa<1lvd*_9bqF=%EpW}1m20lOSDWb=YQ2Q3KY0;H1ekGM@hEQOm>34?f$9sn6c
z2~Z9(sRz(YetI}O*h<m7l(GKj`TT1KzZG_}UlD&N+0CnsDOY3tC@kZN3%Ws);_HLD
zbyE7Z&Ktj9^AY*tY;*Sy&I$P6W<lYwxJr0JR(oHbi<?ub^PaQ0Hg?~nDA6sgwb~In
zxS!*PrqB)*n5wZx__B;}ti|m~OGEZpd67uU&MH{AW4aJYCG5sDn<fy<MhClV-?*Wp
zhR3hj>h>+lK;1~r{Yylc@R4x132nNme;apraJ=X;-|&XSFZm6ACFnuAk6yNNqfwZY
zWpn!NPTW*!_Jd8r6*hAGq$mMN>Jg(qCp0%_si3+>P6i_Fleou9wC!_F#y#krG+25u
zjJBK;1S6vG)me2Tr*GTsaA<C#(SX|@lJ95E?}Qb!uQs9Fj7_FG!*Pf5y@QWjt?J5D
z0NotnW5Nw2Y$-hQ%JKqif(7+DsRinY!5mRA)R=R#?*v>5#f+XLhil=;YfUa313kuU
z5bF`IGcQ|>f%Xatdo-BF`qShm%>IFXVk}jnG{v_H`|3MUkO{_hjEd4KSml~Wzmv(;
z0WKR4HSgQgl?JilFDb=?KXQx`wDdwETtS^3Lv^ZOmL;g4NI6Z{*)f*g5}_=6XWw`y
zl#jj2UpMTNEu<R{wHgct!F<|NIpIQCx@GXlo7mPw#ccJBlEjMGpSo)q3tlvxC#$sC
zpEBGGzntWBCIKK(xoqcTw!;Zo;ReLE4L~#~=f@e1%uE+<K&L0_U53`6AI9x1tp9)k
z;U)w}ItFWOJ0X^St0^o`ayfD;<N0A*mhU<Pg|IIH3R|;``R`xEXLYCj5?rAUv+^9}
zy$J{qACl~r>=Ethz+#3Mca~yu>#YLZ>!oArA%s1H&8Y6BYK_kKOf+Qp0$;EA@su%|
zya_7UY+3TXAtgy?v2Jrzu;ZkW#Me!?u&K*$Y_!2=WPDgM85>3lI;D4G)z<<h)cg`#
ze6<*=?Yv<Wtoq$w)Aax5b6?)usbLdKf!=5uZ2Pe6s_GK(0iR2gQ^yC(ADS-*KQ9;i
zzuJBAPbF^vGt!Q0Bsh9yyvIQEzRHfN5d89Zz^p}ywbr{&RhLY~mB-peN)i8nj1!sd
z(aLJ;xKztMKUik_R}zuqXWko}_Yv=_IBU)HsVZ_V35``k*`6EMo`)=Y37D3I>2OY2
zy}40f80rkM3->hYzD}1YhHq^Xzy387Ps&725hiYJYZvMw2owXo&X^!-a(EIK<Bn5d
z-Z68<G_f8DMue1?#4Ygkr$ON3d>d}xR#51%JF@J=m{_g*du(o6FjB}bhb&gXUUI#5
zDg}naBM2)7HBL<@4WHk#$J6zYzjQ+cY~)~AOHJGZYR$!UE_wkA`myKcWce+z^;qyv
z<f<O1HrbK1mT-!0G(YiBvF&LFHu4>!ZdOtuwnIyq>x8C+21H<b&h`l70W&jfwFV)F
zfivJTmt>vJr$MJ%T&@4`$JsovXD1)L%cZOIr@n1BbY&H<b9Qco!<K5@i?8z?KrFv&
zt=ub=?{1wmoOAYK>)MhWYmo<!tl!N)x)nDymDE%dz;L1xmFB<yo4*^yInQeUQYX`z
zZm(P)?P{-F#}FXl6w5poR_)>V<k%;sKj7Cbd!NemK$FvZyOf{v2S|uNMpp-Dp#ZKf
zafS?Jnt)*8Wq=gAgHcd!*!`f%1qiSH@mT>t5Z$3LGLkVSz&{$xmETJzIKyehRJxpt
zxAy1P&(wO=3p<SikoGC0F$5U>Q@3-4@5GB@)oleboo8*YZV6Do5*>QOL(<O%RAJTI
zROmUA?UlUbYqORHT3>qPyqBlLHZ-jYkUQb{TH8BEfc5PdAwcuA+9BGL8#p5^$6Wd>
zP5(n_Dwo;Hlb(+p>FFI;Hy0YSswH&ymXRYGY3vEJ;r1d7_V6k9tesqXb7<5eOrjHJ
zi367Uq^c*u)QBzY*WWgkAuP6ZuXnh2ZkU2<y3XG_m!SH5^!3$?st7&zqXX%^qBhva
zUC*s7kim;7*=s4xtg=&b4Q&3*R_>~oU745@Zyj}fx+xhEnK<_dZWsPzuH**U2G^dj
zyNF9s8*1@8&gSa2;A`J>hDk3^mRWIbY@Y%dBeXJRQtE4_c5=#GE0GXi^O!Yfxld?1
zNE~8;{;tFKdiJ_D9v)5)oqJaVCZ5Sp_|5(N^;pz4FU@n|p(wx8QLtwGLL6FvD|B^R
z>RJ?5wz+s3M72JNe5@jKeKA?MGp}1$e!w>!608>4Ps1o8rzVM$O^ju3Ys*s|*(uI8
z!Z=|4TMF~x#OCwGswelWMYxhU#cz7%v%B2E#czFqtN=i^^yK3%+EHgM!+9z_<)HII
znz*?Z{CH>23<W-E<8XF-E%BwiEj2Ew+R*sW+Rh>4Lt8dGcFP#kC_Z-3mg0RFwH+P3
z8Lz{T+?+C{w~cRMm6}VS^Q1T(zasuD0MJruX(qZmFk2H|+s8>{?nax?%l)it&oyT(
zS%dbMd7>Och{L0fQaf*>6<q003m<pED0SN<A)Wm@R%;tI6WvGf#xcBE>#)-VmD!O<
z%ehuw;%2w|GZedHmVqFYfCDN$yS6Q)>6#-kB|PId{5GE|Qk=!M+72?p4Kykke_z!J
zlBFaFX9Bmx6A$0@*7CZhEnxT)u@Cu1o*wvMTI7&C^LOKTL)}ZFr3QKCW?aMU3yNyk
zYT2^_^&Y<_qQh7p8KCpgb=;fmCL_uoNwJz-0!j*Pijt^5PKu0M)Cp6Ahznmp!T=b^
zJv<~56Tg1WyEEJ-^8l522*5Pi)aH=}-zVG5B}WadAZ@Xij`Zr{Yg!vJM@C3gY#0s^
zKfD@9?Ln1iW8!eHNp*?$lMf9;{%<y*QZ18l=ZNQJ*2373D70LHLFQ)6bh81r<YAU6
zTWk9HwX#{m?w6^W+>&7LE8VW9&O6KlA#TQV`F$_h{NHCjiNk6?DMJuB!dlN4AEL{C
zOk%RO%h=VvC%miu7&&a!&S8@|$Yj5$M6eD`tp0BBK|N&;t*r0F;@<YIneI)*5n9X7
zII-cgVUtyO{35Y-@g38AXr`H!q=?V+S&gKoR-I$;+;w=`f^*4nhDX@Io||x-WE1Lm
z?3X`{>$9=4)Z5@9%r`CV$J)1DaYYOjn7VKEnH77d-br$8S5mrkXAmQZsI;tcMrF6&
z&0UBq4He8}UJCAJn(j9R8F5=8$roO2U%_0y4NevOKpAVzW5t`t+QaSegQQp?qbHso
z(Yzy*?d)yxkB$f|_A~^&o$AbnrVSKyCx!Bv)zuX5+w^?qMoO@jT9@Z=CVjM6zNr_^
z6-q+|pkjRnF2y;qNG#^|C8y`Ecl75n^ELBq9YpzG-KP9#(z|P-#MSu#^g5-CehHKe
z+PAL3zD3W=8D3z+Xix0#QZ~1zJ9TvkHcfMU;ry-O<>pqjzJZ~7F#LHNUb7(q?6-4i
zHu>rjlr%KGnV!qQR}(CCyOuHbRDs2!(*0N2+KxP7;<k5Fa2<b`Y<3}XX<?d81KJPr
zf9n^3M<zSp8vsIG{w#Pz&lD3IIXg1k_K-~>CUu}!XQ-$q*(2dI&x2N+BR01QA$j<}
z+F<r?^lR>ZY|<XCrtHIb!Qw9;jtoa;+V#(*R=gSd`rYdl|N9W(pvj;VUhf<st>Lr$
z&$qPjR@OJ+P#yo$b1l9aYiqYyZd+Sw5-kzCo$}pw%N*q<tFo3@e8de_yK*QS`#k<!
zumqrsSYBu1G>G;VB{y0?`b8Lem9_s%OezhL1#*n<&E}!2%Ev5O$EUlU_@)WVa%9YL
zFAA@h+~dW4ByY1Cy3pkpnfIWi0AX6!soTn!mnJCJU5}C9-{&*9;BaW<8X?w6{JV}(
zkLi+wRw}MGjgdA1PL=DIj7^W727NvL!5a#0yLYZdY2N+|!TBfdHs#8IubuGJu8!w?
zcWHQTZqIF|ZzBGF!(woo%Tzq|vTL(oBuOm3b!9;%kvE^%T);4|BGZ>n=h=lkbwI9}
zlj4S%r1pNaj+zhYDc=HrqUv~x(J|Qv(o%j$vh9}F&_-fOe&>&E#lO+b?z$O0tTgWy
ztnn#W5e4(PX;i4^v~H=S#6$_5Ci~ZWMWuSh&?42dMWqpHho@?tP{C(boK3XvoZ`1a
zTDSP}m)2JDm*~Wrawx*v43q&cIR;Z>lz)LS1G0aOUO@xhk2ucUKGQ<nSB=9)Q=RBZ
zplqZ&6NoUcc!ekV6DJ30phLh~zg7e+^F1Am{JZm?x;(Y6{49bR?2uu{4f5~3i7ZYP
zznF-#b@%K1ebKCL`Ea~VE?MjI)UFZcQI`M5UAj3`QUmdc;o$Wag4<QhDn<}s%0l|Z
z-kX*yTesDAMwcibYwKCHDVw?4S=}Y^^{@9^yo5%8sP?vSTbpC?#)6cb&dl}lXO_~q
zcyn@G6Y2bTFbrX*;RnZjyVzYG9DEiT*q+%o9H8yY&B1lSZ{Y?=gp0O?w6@@yyHio}
z`PJ|n7Vhw$s((3w%5Q2LX}n#To*Nj_7W4T%XQ~+y;dKFBJZhioX`S2B*jzlikksXo
zelz<ZP~TJ@F|O3f`w0L&J&19M9lcL6AknVQ17H_2Kj$7-9k8z$dbpkTcIp{)oY2bl
z#&J5!4`X`O^9HD{&j?O344-Y0Ag^bSo!`U@@AfgF3pXE>+p)$(t_?6QCXo(uWakYu
z$%eVifXY{i)ejTdcW&Ea(=F~eaAiy13@)5)d=>Gs{=&H0qsHo;+=x}r3PUROjXyVi
z!$%)4h1QjMZQ;?|xzmHwNrSVm-J=(i2?GZ^yW<55LzZsrVN1+v`A;*J8PsBQ@sVxa
zOheTAdoHg!-Kr=_+F4Uga7o%~j(*Q@iOZoP)ELWKH!Kzn-2*N;K<GWk`9nfYD~Gwl
z2V3K3XIYIQ)Ddmxp7FDIRI&Rh-TZ%PfBZL=;Qztl|C0LPU3>hk9R^dB7Xes&j5>SE
zuWY&^@&e!qKR-|E6(1W(x?UrUXMruxWRHnQ>`s}ASi#v%=^?cjNdn{32yE(`n@C<+
z7^ve`Y?9?|OMO-Loiixh>kC2MlgTa7Hzl)2ASeLRV-1C$93a)&Y{`$2Vt5AXEQFRj
z?+A_qb(n}4Sz}0C5(^%t%P_7s`{kzZM)TU=K#o`&vr7XgMf8ijY!dZalv%Qps;CGR
zxrw;WxQOpnpW2Gfu#(fy34~7c>a$9%t))Pr&q;Sba;8Q#N8wF?YWLC*ZzM)DeUwLK
zF=Ea}YXNKqDG#2roKyL6+_A6|X8(2Fnq_kGn@9e^2N@G93wC;_ox}yO<Ep&79FrVm
z+ow5q=gP07-vi%$OqlXATYr7eW7GOG?-drKEy39dO2f!4ZpIVXCcwVH^tIP~PJGuj
z`N7vatG72@r?vYg>~$tsE!c^suorQ=1hk~Ur{EuE8!Pc~9KFi*){eY(|0%oZb{=u4
zd800GP|^xMD)EQh4QXM_T3*sYn9itYn^XM~J@J8rAkmmaKuW%7PBNqv&q;U?nP1L-
zr$S;?b-TSB=dNct<ve#ZwaO7W>#tn<#D%`%e&$^|(GjT8BLanJ`<@(qeLns4s)D;-
zS(^kIq`{BiYmEU2^KzgVMN=C<Dh>+5bjS*J5Zw<73JM-R#HpyrT+aDTHq5WwK$u%A
ze`mGlUe|YKBhPt085|&_qY`$XOIClxaYv$ZC#^@`l80CUdR}#$LBi9idbnv?O<0)~
zYq8B1Pu>kG#N8PONR|ncLR)hgPB1c+=xfqm@hYiug}4+$4Lx@S`JS^_B-71D<Z~)d
zH8F<bkfGBrSivjx2c23PSFDiv1B6SuEs@)^j-9wK2CIhSWy*0zu-T>_fN8Q-0K-A`
zdwP0v7hGYNd&J|v?r*-p36AMsS6K67ivalL+SxTc)2FTW(Gbb6_7L`mq4R}$Sa%xn
z?ZnD$5TPzA{B?L}hI(33ti9t@jsEE~D=f2^L<98Q`E3WSA**r2uGMWVJ-lnLalyn!
zT#KeT*85CKXtM2?OJUr)5hN&h9C!99NX4gs6naV*E`r~RoZ*ad*)<dbict;?Q(X>g
z&uYs6ZYot8Nb7l@W^jk%tIg^<@h9EfweQjL$nwqk{0Cn+sI~e-#-hieQ`9`msPBd)
zarDz+<R=i5C8N>n_1+oy<SvD=qZhv_K8akEdX8pYW3X!O1GS7ckH${s3#<FbQy2W`
zyT6FE6%OjQBU)}}>cnK@9(HGIwSK{io|1G$VzisC)Z|fbdzzcKWXi%0W@-Rv$T`tt
zZa>U{!DCCfs6mPf4`8__$Fy3^z#Aa!G@&pW0)w}1KCGK5tg6?=KOKf=wP|mCCiSkt
zsBsRSe!a~2ES&tZsy2OXZ2E#~@Y$KZmay3hw+sdE<!pt0Mi(~ljlAZh*Gyeg<UIT)
zr86)Rc>{VP#f%3nzmXg&RpR;2wF!0{W{qP{bd8r-6%C|)H<@T>Mj=UMDSI+E#?>^s
zVCdQ5M|pI3I+xkf)<9`%0Q$9~;_4gP;|vUyRbEGUL!nx;H@oc&iIlRSQ*vw6_7!4S
z_oNL!NY|+o8CYdwQf0O73cLQF-gLieC>u~df}9#N6=p&!o<<x-S@9aJ3<v>_Z;2Bh
zK}m3jlU<_}h}frvA3n@1B79HCVgw?>I=Hl&KuG~p>A~;s!6G~z^9&n3DeJeS?55kT
z>jqIf&W4q=`*{qm;oGS3@2gGm{G^ccE6vSmp`~A9@@)O^J4uVeSAMw16k=wK>jeQ(
zGv!p`*fCKFR`)oh;SHyp!y|8h*QuUen|#m%POvr#D2_bRjXN-_N-)_7uB6dLj7Pd=
zM=k;G@SaF3)r4h@nYnu#<@vAT+w1dFu%zYDSJhqsj!s423=w-MF=cgGE+)6(z~H_%
z&lj>CtogIm90?HPKj)v12t$ezc4kB0!vUDxoWXMW@n#e8%|C-7|ECl9U%B9a3g7YH
z$^4nZ{Y>teqEaB{<T#z4-EH%tsAO&CF2GD!S#;=bKWlH7it|zXLwAGswlimNJq!Tu
zMny$s#wHn~sM*oF&jEPWzujxgUdL{IKGNEgH+G-1v^DnDXD_e0c~oYWr?>luo+-Vg
z>U+JMuKalO<5Bj;qt|NVexm;3BSIg6Ox6uvus)FZI^diNP)}DM3q-{ODQXmiu_e$?
zA8M1(w7E4sNDBL$+S9{FdLM-LDB^vR_|)FTrYoU+<tMPg#Vq`xt=jA!h~4AKhHz5w
z+I47D$~r{OWKfPNE$$zeVP6dGHy+qD>#~HE=}al{W4dl}GB5|nB@o*ZPWQgRty%B8
z=eo|PpRL||E#t*h^4VEw6<4@^yG0*qyg;DeUI%Ga^x7z*9m>~7cT&nLn4|4OriPQo
zo(4)Og_2{sT3dq>)W=<W_A<t{)v3R4RxG0RH=HaR6(sOA?eeRyg)IjCcDN$DG=;@m
zc)Tf4AKH>(UATkwcfV#Ar(6DPodt7^iH5Eh>{uA_BH9FGJkbkFyRgK`5iZH#LAZq(
z<i}n4s!0c6uZ=s65%~g`K!f(KjgsN>fthR$Z3$wkdsJG|RL@pszlq=$Y`v&9FCns=
zW>9B~4gq1IW5Ltz3v&w}LdoRC`mVChAY!DtMa*|h2>;YzlaL7!oa&o|Av8fMGdk@p
zI0G9}+9d6Q2c`p(m&V1U*nxZsw%q$u+8npsc^{UPzmyvZOtx7GxV5<bt7*~R#BNY>
zapC6a^xNt#zQpb2+Ml>lX3WJa=|9~s3Cq`dHqNb{J{DG~FUj+H$tL~y5;gB~tr~?J
zvKEh^Bi3_5D?FCcwg?ZJ0qlg$qzIN<+xeK(U-2tx%GXZAW5Yr^j0ljN)q$qo;eou8
z_bqD9HkL1;6XHsSlgNG7+zC&&mEL=0ofL@AAZx{c4s`5U?Y<3v$Dwmm0LgoKr=vO(
z6LrbhPE8F`=@>oK1E*BoeKPike|HD`mrmRcXt6tCdu1tJ<$a4hZ9Ycbuk!BD_MUU1
z!82S-cb^Jnl<EG&g~r}{jw@bAa`n6j2*QFH`d|a7Ga}zOP={bny$ZgR@-;17h8o{g
zl*BJ&M5&wgM6URO^YJZFPu`BN2A?H2M{er*`IQV-8ck0%r;mn#EaqIkZS##=|5?C$
z_p&XveSStJ3JA?v|89y@Wx5GU+>$+t<Pt57xS?=gbyGIsNHKIuUKQe@;G-{;RtS$5
z!g00gUBn^U#q_~Nkm;9$FayhJ0L9PpxhJXG0ff>}4k_!ud-^PFOWlr={Ft!MZnt+<
z#5+o5A=W-+aOtW%#6;HSz4dVyXbxwZ*qLT-{=_x2j}sfZ<uQy;*4Jt@I0vv~L5eml
z$~*(SPmC;`pjDihriRP|EhNdf%#xNZ!AzTNt%_NFeR$P<9f-o{Mmj~F*TJ>RPMU{}
zpQWNrpN5V6W={_-a$d{-Lffg>>isJTUyZXMNk`s|-=->B!XB3eTuV+&FLUZjH8e(;
zoGpq-xKYd;fiyC025Q~)8`3ZQu8#&laSX7)ZCmJZtNd7AIx=5h+6&MT8H>2d4GUuf
zgG1eviac$YT6@c#FCttz5l%fF`uQN<X0MWA_e^4@LprsmGeurg?q6t=Vw|S!<gFQ@
zs12ft!B0b8M{G{*+C3E1VCk}G&pvdwYs&ao<UJIq&M2|gIz82XYvviIW@LC6;)#!q
z^?#O9<#a9F-2+uIt>2JV+tOOi5X}N)b%T8zE+ewl8eJ$6N4)R6u^L&pZVV}SA*@S=
zWAYzy+k6UE<cu_f$ehmm8Y|P?w}1xFxyZr`5bXZ`T%esD{8UfHc<UrcBXrq9a&(?1
zFc9JPPFX;m+lq3e8}$GvT4v+|JObqo>eRTn!kx{E$r}AudYxFC!HF`GPAu*$UgYOS
z#8rH`<1PP&e$@6{drRsL(6SNxQj#PT3rhM+%grJOR((2C5MM<4dAlJ6rwSztu?6c$
z8I|J=n!_MtlQt@7t1qA$wH*Nu2HGB$_QJ-VK)g~KB18i6c(;l>UPrUCVPwuYtKn8|
zr4@Ky9H)c#ZpMqt3?t%ev+GVH5@6w!z^?;k*xf%>AHBYjcOq7Tx(ZkQnQgLmIso-{
z*2vt$(SALZ@Q{}0^Q^<20EB$lb&BWjN5z8D_k@Za=ir$@R`I>@ZrWIj1{zgO3~s`d
zI!@{)KAPI}!@A?f+F=1-xCJEqx(8wpAaP$zVg%4Zy%DR!sN-lj1Rj{~>!rltu}o}-
z?N*KKV2h*7V%5o0Qf_BrVcHiE2t>F5uZx7^=j$TzANW4Q{BNs26&}Og0Kir5AKM(1
zk`CX@$~|uKEUWZfjPsc+rv8h&MucKBc*$_Q*@bC#L%D&==IU&BWD7Br8BX_Yi4nj7
zxB`DJ-6fxdM?>J@KyIF{iVuhM=i+spxJ{L?L!a{UBz$m(7wN$tAGkwn+cx*r|8uOv
z&rv&7X!i%p!Va&MRoTHe$c*vwg(PX&l?K2mT3{P)`54N-Ud0d|X}m=#9;~&J0~KI|
z9@oqUyP^yh++aU&lD|BnE0JenGbs`jZ@ju(0cLCr%gK%xe+#%=L0idy%S9GVZ<JTR
zZ&iEkokHNtQk=OkU=8g1unJX%zSt4=;k@IvQJac_gn2t~083OtvcY0UH(ruxrRDVj
z8zmcu5He~i6`)4oUaul*fAe3mz<=wD|F`5{KgqjKRa`ChdnB(YRA8?kP5!;2mi%!S
z|0Q^4&D%z6tyOJ(6Fa9rYf4Lc|F@3rZ@$hp#arc3Z}1}0hZ+htN~2XD_bwR4my$1O
zCD<F?4R)L9T`YwqtqG4_G|IwTPNxU%?b0%CsA*|o%ukNK;}7?bzCxrNz`Jb5k)y|#
zH-gxy1I$c;jUo=DvTd}Lav_LOtkDq4aA_uJ8sg6CVG?kI5fE)M?tMYNv*qq~>-MD_
z1JRVZU5RvY<1A8~!Sz(_Dw?UuZ7oaOWO=RX*7&?%TGCT==#7qE$LqIwg04DGe=ZXT
zsFr_~>~O%e+0fV4i8t6OoQ{Z!ulFqAw5#F{xg0O)O}PX;2<G9AoC_}8VuhwWcR+I%
zac}qt%3z}P6lJk7NVp6OX=2aOT_J(A>N$BWBouu4<wo55zzO$+SPv25+qk)NgShA6
zaFkY29sUVW6_pSc)_qNX?$XCy@wxlXe%$q|=}GzT=;xOw>yIaK-a>M4hb)uYF0u2l
zWOjM&O*TrQKGpPbmr^^WqO6F|2^Yyug-ISQt*j9?YTPc8a7$iV;m;GzWPx+pOwU7{
z6kTeO5AFA<!v~5n)kilsH(i}kL-FT4nSh6AlkGgcnuMW;-RZ#~w|!PgMT=X)(da7`
z9SVmso%#_5PRm)UFi$6i?c~B$X2r8i3VzPLI;q^(&oiS9x1+P9-nJ4I##Y((SW-T|
zoS+oV!gZJYuEQp{GjS<yzp)Rp_Z~nSPV~5+pOf9kmE7RRot)J_t%cI>D>^TAojtl)
zw<kino1TEj(Ke#QBi1QwdJIOv9%@NJ>-%FiNr<;ma0pRat~m@CCg8`*2CLj5G1<ff
zs4<Yu+vwz?;}Cvs!H(x~nFJs)!LIH2K6~c>;BU~<U+uPiR_p$7xwn`5$$r5m(QRP9
zbZn;1=Ho7T0#lJ^PiNYU>rGNJtIM?#KO{4iCGzN7^NAmKO=)7Ojb&|~V{_}_K!4xL
zcF4zF6XXc&60tv$dgq&6yAFnkOUHkSSNYM9x=mLuOr$Td*Xddbueg8QF3ti=!#(LG
zWkIsBR3(5>Ufzd}($XT#ET6asSg)}O(X!ckY>DZ|T_`4!`*BxVODc9A7SH)S`^Dcd
zs{i^W|64lL{y36fKaz5!^_|ImJs)?0pQqdXjJ-aBUGURdP0}k<-ttS6W7^j8Vv7W{
z=>Ph_|G{_o><j<C5JE?yUrYwV5d6cG|B08t*PJaL2k0WPJ1szup7U?pLbVaO668Sm
zEJI+$NNnFu5wubofF{wOP>Pk_q@{p_YMUIzD)td&+qxO(m8JJK5R3K!0$J{USg+Wf
zYtbubbC!L7sPfxX@Y*92`2y@((qUd~`$kzgUka>%JFwDrp>$hd2-CtTECoW){8Y1C
zv5NK>m&6Kzc*C+uUC7>laDac`%l_BI8vwCCJ4L&_18WGl9SdfQDwLmV4aSely&bj%
zk7n;E@`DuPhC63VW+(*Tzi){DomXFMYtjmrXdWKBho7EZw%)VPZ*vFhPAI;uqy=8a
zMkQgz*n2j1EZ=cGp$f+dCE3U{Z~Np({cgTk6TG>nd<mpeZ@Q5L>t!z)OOAZpb-$pm
zS|PWH@7sGA5OX7GaQv6ZM5QIL`>Rk1;6iPJ4<uitl<#Qb%rE7&Z)6%~@g%de5_th8
z1*p)^Dg(DI#)wCT%RdLK4*;+<DIfwQd*69w^;CZU-#P04@V)+b^g#SkyffbQ<@k8L
z#munHooj*8OI!6LA9qE30QesIe3!BP^8q8kt^JHx!j=9h;QSws>Hp<F_F^Q<Bw5?O
zdLDBxie;a_-;b%a304{^JA?N)|8dvOg==@QuZ~uY53`G>wnLK4!^<m|#?P~tx0hVd
zRaT5`v=^32DkJg6=JG=odnZ9ae6nnE20OLFde|iT43Hc=WPhPc<)YEuKrIXVLo4a0
zmA`&{9=Gt??%S<#1weHMgy(^@hQtUQ@C>N@THA65LY_TlhhMbZo4%JjWkaoB0M!Rh
zOR2SeS#>sTucpC0{nlZJ-w&~W#GbHhoGceI&>q;C5OWRMTII)G$xb5*lgogIx-x3`
z|I2FkPiyseY3CK?Oc0?f(JOuhmtS6Sz0Y8~t-}(FGX!>6qHS>b&c2Lw?98UtQO!pH
zU3h}gQ1r)L>9^}y1R>BpZO+)>a=6FkG_!?6TcoB>nXa6^Hig)j%XLZqBn6Lwi&B7p
z*SAF7R+zhU>$gAdLQF=JuLEyBC*5p%I^UeKFeLw?4X1Uwcsg3P6QG9;Dg)CNLe%XA
zI~vviCbhT!xa37Yb|D$v5&!+)HL!p1;(z)|f8WG?c2r?(R<1tr!)Sz^e@dny1z^^D
zZ?*vBsM!i!D&RIf(NFop7=ud-xU@9OXeCdzD~w;Syi|bORx#LQgzXJN8>gP9h0C-e
z12MrDd|umj*|^cRe&<MV!xSLOZeRMoR+-{A6Rk_$wZg<&oWYKNN($lO08*6-s8bl5
zH#;i_`&G96d?RIHytV6-E;Yyd+JU{_&QQXIask)98H$#r0G;J+!rMT$M&0``V2!4)
zYp#9&_l@))yZWE{^!Lr%XBuT7GN4a2W5@0M2@apbDJcN~GcU3PTSh>3dRc*ezv9Tg
z^X2&I$?A8-o03JTQfm9`PDA9sYShJ~(g1Ar4%r4JX^$LH=#Ujy+QRD2jK3@8s0a|{
zGralrD7tm+4uZbAD(n4xLQVjW$TDqtpQ2D4@oIRX_zSJpEs`knlO_%LFWhH{HvRIm
zt8q(cA5*c{hP3xrfGSN`)MC5(UV+zJ=Az1m{`wHX4`;J-vIlE{8bD57xaeBhEypui
zj(b@EZ>mY4ej`)=^sy=+Mso2`dhp&$GY5+n1*cOqN5j!j^*6MIxvq!Vl;#>(fO>O0
z;$9pzL%qn81>g!fe@R~mPdJ<8$uuE_yV|SjcKJN<@JTkhyZb#riqQI+xbZegb-P}s
zwDWP-?lmY?4~(gP67Y8E`?m3A@IU|m&-wW~^kYvcP6AzSeapL7Z{2}?vXO6$vWF&V
zynKI5%${AnG}gcM+awV-^($xd{1TgO>cq^6zM2H%>JceR38X@z!Nv>?g7wMVqI@f4
z3L_S2BDs4e@7|Yo9F$jGXnNqjKsUz<6F~KQjlQ=&=~ihqFom~&mK%OO3zGAkQv}Ho
zd6wO9J+tDj|EWMnKhT8ed=!Aux*G^o=R5%aoxbEh&G28hQUB4|&-%0utGj0`e^?yT
za<A><`hVBDva+gp*S|s$5VqQsMXNhFL-5d5?A;Q;6r<l+e=^0v&jME!L$kM~+gR+3
zOtF<N{fe%*KO&UIEqaYhOjgDLF4H_1;)uofz{kQq>sN2X%_aQ)kIrH3_U?pPIaa09
zsuTfMJWEcA4sRDP0)}`&?DfeI4-&pz{^TDvubKeck*-E4XJJX?2T=93iA&D536Vv>
zt`^p$<n;3|e|1Ecl+zP30{k#1Wkv0$UqV^kw6rN-HxJ=Gi`_~;Wiigac-@sad~Gx#
zz31yOW+jqte%6GXf9yFq9~CK;>eac`JbXJEyqa(E4YRGS%vkfJ{~2pdkJT5Ws|`Mq
zh0%?#f@JFHhzm)3_Iy{T#}?;x9k(+)>!I!!0lp5h**H0spiwb)t&DH^2H;4Y+WUXU
zcKWQ>=K8Rpd)YNY&F1uwTlB<88sr5Y4o$NRzz9W6=a}a8s<+rRvl3&^nth7e!83tc
z-X)lAMYGEil=idxF@u-(uipwAy)?e(M%s>7-<VIu{WulyXbp7{ktt|s6j3RtrJHxp
z)GW0exrOPgaUbIrB@sV={1UlPzpP+G)$s|0TB!LE0XIV(UFIR?Lj1qsRQMBe&IzUy
zd5D(R3tIwh!TF+y$FCPgaXB^m{Y(dSoBl52&zr2Ulc0VBxmaqZ`E1=e1C?c#7ZWfo
z#h5|rwI0VYt+G9{p#L9x?;X};w(pBO+dbo08FdipTTxJH(p!Kz=z#PQ6eNK}Iw6Dz
z2q7fE>`_E%QF;x-ATS6aQbK3}sgW8XkkAsO3!#TVAb|IE?!M>jeV*Cpc+T(M=XdY(
z%=-ttWhHB^?^@qCYkkY-lT#F@9qT810in%c7Aqff#PoV~3`|ui6d|Ei9$9Y<e*N9y
z`zJRKUx@S#TAD}ete}9I@TJk2FR{Sze-RFIrSQF{toP0@7VWZLfA=o#x6dERANhF5
z9wIUhqVSGK=K-c=>rnd2leC>ag;TUmqsW;+AUV#%yzr;dAOA^x|H<t?eIK^4%W`Y{
zFc+CSWGZ{d_|<aLRvCVWDoQ`HuhPEX2-~&P>qS9p{*&+dKlpfo@|0NAA&5lBaK7PN
z>)ggl+nGl+<v=3-GOlv9MRqdY!mf4i^v*$@U{7sM*wlvqSsW*L$}bw&9gV!Jj`lw*
zJue&rc5kxp3k=xdY`Gf8kB8V15s^(&4nz3|_dI_XmyZ7#XukwefM(-R-_;oh%`ava
z-PWHt_F}&dz-zWCfP6eO*9nrq;;dbP3L;5@hR~SZp~bXn<6%fWFO3FhH@GzEAp~1z
zYfD)mkltVpbf?G{BWI%n9RB6^Bhz%<%d^*i4o3eK6J4HN91fAC;S^{P@5{?W%kXaS
zL7A*wE8CMhz3#xUwGqq6E@ck5tyO<Ra9Z7<q=r14URCTmVQxSl?{i(<hU+27QtXVn
zx5)8~3h0FU^-2!2zRgF%h1%{LhL&8cxSN_>o!OW2D85=!`cg97p{dxEtCbwjL@6dU
z)}4C!BhdN(@YwB@4CjQlBAZvo5w*)WEEOR(xDm#j&tK9Dw@0uxH1jm8R@bsleYMf5
z8q-ju#IU^F8_zyA(5*Sp5(*e?OgIq|GOTSs-9HJT(-FX4%Xf<&4Yx)l6nC9B@YAjx
zW?RZCq^QOb18iq<bW)4K5MZ6@SvkNA=}E!7f_c<RY?K;F)c6=MYy^CRw!^W=T5NGc
z@2ibQ;!z}geqAt@v7}=tc@=HpTX7$A_*R!Wyb4~Mdo4KMWK1uPT#%S&p?5;7!&1zF
zv->(^?LX}6KR)zOUA76&UYXnP<-*<<wR}ek1)i&xENLH<?wL0|lD-2Vn{RzBar(#K
zKDLQ9UPQPGqf-xDA8hL8E|A3c-vQ)0dTo=#2hsGs1lWesGtGC9AO5k9&%U}vaxn%k
zIkKhbyxm`07o+(ul!Kvuz@+g@EwGL*7;d5fzBF@hJtCL&+28rUy`VH?^>Ptyv#3TJ
z*b2ojPTWU=nov`^S;&0APH9W(L>tNS!Y5exn_D2kj!an?DvJ*6Rl1)v<*s7yOazdc
ze%*QnOc5e41%SZpK$`8B{ydhTRtU87uzlnSgQmDO?t(PR*fGCUN&`O6|9GfM(@zVZ
zwhxO2VjUiJcm4`%C&<27H{dRVLbt=noXW_!5b0&S@x)GJWaHcIqLA%Pn>^!^vh}+k
z4?Qc2$rYAu=8snVt{5C7N4aw+SYo02D-wN@myNWg*6G=*tvi@vYl-012UTSuD$WfZ
z10_lWgGHBT_H0o$TTXJbLGe+`DRA%VI?|=)&xr&-T<!u{&UFiO?2*_u2#!vvF1t9x
z_J8?0FPlknRW5`{N^Bz7h%B#Ti<a5vCSQ3JQSvtm%PXcU(m{{mw+V9{VO3Y|6iNgM
z<b{^b$EVD8(zp}uR}gH>N=`fV_zC|rIw{Jb`2_WEi@TLA_H0wTT^ws$kSzNa-=WYF
z*T(wpp;U~*pho@U3XkQrzxq7*T~Yk*%|Ddmcg6A_+>|;zDzwEfvOJY=RG9c15Dh*H
ztRHDuX<60X@CZx%hdEUJZv6t3mJ4j-Bk;eeMsDmU)qpB^k(!8G#{kuMJX-=-gl@PQ
zFMHIDx5xbq)F<&&V9DM?J)u}4Fgv@tlGas#5$hXDi94WkX04dWcY}|nHAL74*9_4`
zc$wAPnDGhN!AQ6DZ))K#7&*7u1=uZZcawK`(@}#(dltQ*NbK0Ouv5g;^Y;<JOLfSs
z)}6B3Yr*5%e%ji&WL?=@U0AyM#z3l;;)tv7_>!|3Cb7W6$*b7TQg6qNQ?N8}qial3
za%_}gjzmB)H?I5#a{AXtGA>d3Tk5-Xg}8=#nnyV+WLvNo71fTg<Vtca2AsiAC-xZK
zW{SDUoXXMPVrgD82gk!N729CEQ=NX6(7+l*49x8D!t*wD+BE$0%rgBOZ3hKsD{8r~
zLbk@6(Yr3GV(~k^TU8$Uh%O4Oz{xkO9lkIY=t8S)^*f#SHrQ{&U=1quug&?-PX5@j
z{bv&T$LC+Pjc>!(_r!X&Wq==xDSy){;3eKA$T8M1>psYZC01<C>#(UNAF^a{xP)1z
z1_szoW%d8258gVk$dlcS?LDvk$Z20sywFKX$JG6kbNQ<r!bx<b<YW{hj6-tFkD2>T
z8Ad}CWo`FiPk;*E4USExiM$n5@jD10Y(36v{a|=XxA%YjMxYDMZSSIo6!-y^FaX-?
zTVS<l{>v4WZ~n3A|MTkb69e%#dy%6mU(TJX-6)9n!%OqLrIFbP$`9mXWy#%p;S`bJ
zr=-Z3+Dwb*6+KghwMkc_njR<Fx~5cCHWZ0qOnz`P=R{=iM83nC18?TJB`v10|FmJ^
z7EL^p&_ynkiyT-{QM|*bmAgwQ9rCPaqMfCbD|HuUhz{f^3)Vg7m&!o^eL(!L!Iyro
zegE#>kp&APSQO68w=0ks^fh-#0e`a3U8|aVyt!y!e96^NKY9MKJdj;<^kwEZnfF?*
zf0YI8iN>c)Jk)BiHNl2gn{dh5k;FT>(HI>)%zWf9@i4eNlei*l7#yHOxRGAvtW}B%
zbSy761YRe<j=C(1C3ENb%M|65<MIElDgXTP59jMYm(A}z5s90@+Ty2sH)!4T>4&|)
zFuJ)rd#0y$@o8y-@CaLU39u>h;?{psA@tMUc|NY}2J|3Oz_O2M+&?)QVYF|ZDadUJ
zA4bW-1WreFAZ{OkTE0r>9NC76Pr3|V<3vQnn}v{pWjhA(T&bbygCJ6MSE2wHF}Wiv
zc0gv(@MxVka~}_#U!bDg>DvdW+|`PcANHvSS$xbsxE@%?J?lc*17;M(TcNez{az0a
zU5M^5XtpmSs7hJ8Ca_iHr&Ze5nF?1?E!jY%JgqZ4k>wf=LS2IkJ0wJju665;mY0<%
z=<{GoNQU1@AQ`o~IJMo`LnC$MbvLbvaDsCIDp`Ndj{bW5WI1$w`Q^5#8jjbYjWGmP
zh`jXOv%8lY_r|8yt3Do54=%y+7;iRY3~WXX2b0EWN7e|j87)e0ez%$bcr%<Xz5x}s
zI!C|Upr2cI3YUE$nD%M$&or9&u~uN!QCrESY^sO)QKL{+8z$78W$LxYMY4W1f9CO^
ziVu@g;*dO9zWNg^hLAJW%1{nUpOjbpqHjDBOvhs{(}owOZ4zqVGON!x=2yI5;@TRr
zl_g)4Z<u)xDHLn<8XT+i;)`$Sfd~c3_1W^DUVEQ){IO*FEHV9EA*mz7Wvy!p&Y?nu
z>X5^!4<oJ&t@;IOJEYY=31mnIWz<Byd`H5w9;{`SA-ucHZAvR?XUnE5FjtFvkG`zR
zwAC%sDKyFmi(QQN9Gnc-S)e|u*%;Q=3Yu$f3D~JF_nF%1-`K)fFU0r+=qH6)N9=Si
z`}(%$kNNehp3RvLwK-xY&(iB$Sg&IhS~v)uN)kqj#&!l-?fl;Iq_&-%P_>!PDO@{F
z9pdlN@o8guyhZSr(8|rU8;fFJ@<q}x4mN+8?K8>>Z5yngiNgs_84gcZZ6qwzOkJDA
z9J63$a`bafnD)k1P@gOW7vFv2kTQE*0=-rBfir&p@bvb32CJ`?3;PvHe%P4nzdRtk
z^gyzzd8iukb_=San~>l9AggME>ri;g-gieQl-RJ#Yj&N3ZyUVtEJxy)osXoNHzNl*
zc?Cl?{-mxzSmKVg_;|a{%vIfoAx6Bn%ga;Q2K_qC35v1q(#nz{L%A6chZkY(D8#K1
zj?;|Nvif;pJz_7MPd(nP$kpF0D_vDqfp<ek7nyU98Sh9aXxzr#JZKzmpu`5FkccVB
zyhl5YS@RyEXF<GAxgc|Jwb%U+^E8~^npZo9mUha|=^}tSP753T8_zlM7kT+J+G%AL
z%uN7bh)F3}@FA;0ZdxNPZ}xyI&z?Q^49JyD&)C?=w*C3<=ET39%C>Eo|CXH#SpF?F
zQ?AmLfMzHF*<|LkpVE=MQq>+0be#Up{0}7OMq5)NPr*d|t*FGr8R)akX<VnX^PM_U
z;<$cy%b2_=dB|ywKyJ6(>qx<%4Z|vXK|{X1V6AJW@O%&>EPqb+%iS|ZqPbM!JJQUv
zH?WmMvUzfoPQp?4v4dOLUmm>w;*ZCFh~=M82SaUGGovL^$C3>V_1{d0BcylFKjj*q
z=JIj*fFZ@^&0_i}hcueS3Ck0n{!t&>BhxegQ<EA5hI6!jFdm#<Uu0(){wR7!Bsp|(
zUfd8yg3_R^{;+h5luP(zcxTdJp!NpF<eR+N`L7>;6~Y@%9rQ1)5a&D#;Tk^a>QoM^
zc1Ddl8>R-G2Ex7clRwzcCaAn|-Ox4gJsn+>Z!`a_8X&cuPejDUk+0D0X0CQ~@W)lQ
z1OZDgTDF>g)MiV55&S$$niQCNEya)$UA@@9oit`%w&6h4+N!j_fJR#ARB$nt4GwH?
zgl45@@Lu>WTZ<p?KYmiyKQjf#Lx1taDGMhGHdI=ZsA*-{_1Xq82FgDbX!%wTe8})i
z(VfDaiHb6ICRdK%&L-OWR+d<Bs!dm&6WAs0^OU{8@w-3z{T2WA-X{v3n))%TR1F0<
z@;S}JPvr{d-o!BUcH&&$R@VwAEOpx<&4GWFRb7*37pC=LFp+q5E)nka6m$PmQ$FMP
zzk)0Jf41`lv@U`?Osdn%o#)?<9uirsN9~v7AE3SuY6b#|%Og=~2Z?<_5LVe|bnE{}
zk;6P9IE8wR%pEE+i8>a<B!a6p7zL&&dhUi6WM5lxkD^Z$svsU+6Np_X&-7{eCJuI|
zU;~=IJ^;|BlHnG10BP!0`X`i7c_u)YDq4E;6Kbzj_Z%~(ZIp5P^jQrJfFi-DQX3#o
z{RHN_XFOkG_6%PI%MTFs`*`TYrh~4E-AooZO}AlN-{V*Flqt6UVm1RZe!JgKPP3|3
zP7i4Zeo>TF0;&@&fGd}b(n6|9K?p=zWRz;9$ZIq>fX$scyMe#GI8qWDl&~puytJ@&
z5Oi`%-L1)W_-Pj3-&onJ9C^$EdoB#30xEwqON-U^nOzhiN#LC4r@~O3Gnix6<z91v
zH$A3<k!Q84;M040Cx+5<#S7gnDn-c=J5FSTH54&V@LUSnPMqm`IFmz3rev48D|bZt
zbur&EDb;bjp`y69QA}sTn@OZsF-Yk_s3j;wro}wpqN5{9*W{MBVqO@!pQ^&GquMh#
z#MWRLYLQGpK!bLU7CcF{S}~kB&R7|pLGv8Sw+Uv?hetRgRo`!E%v1?qn5E6tLa}It
z3q2c>&Ixn2Z12%-J$W)u<NUhg0{)c5POg`fR-sbBW`W|_$+noGxli-tPuBVuraH$O
zAI$|+<FIvX6zE@U=FWPBDn3V5h{u1m$RxvDW_H`;N!mBQ>%H@sF;tK`<U0ysHEe4I
z4VUg0Tp7L_BYRr*D^ARwsn4?a()cNvCN?p~-E7c^LL%mmT%V(vmmS++oS%kO_}I6@
z!;RcBD%_iSS84tMBc$z^ag~fkZ=VIH0G3(cShh1{iz6Af-JS0X#&LqDkU9gs#!hi{
z3~=wB4Gyy)tHPqKZLXVgQBTWq9eDA9!Brf`H}@EI9Kj7FwGzn0^x)g&L)4;5%-E2X
zu!sJO<|a?*(fKyo49Cj6ayRmL%5tMNGNkeH!D*_Uz<qAv5!u&1dtp4GeT~2J@leZq
zI!H_bB}Ubxlte~Ak7w^_Ra<4-Y+2&o7UHtK<`qgF$q!l`#4qZ6!Qp$2N0<-JtPLpF
zw=P4Qu+6f!vTF)2DpE?c(_@z0o^A<?#n<0=hE1DaQ!ZKi@2DgW{g;5bsBQ@_8ciy{
zz&AUVCX7&Tl=K}yhewK4EoRBcPcv56M;gBJXKsrnRThhTW@@S``!e{Qmv31fok{ii
z^3Nc@{UqC-GK~kzvQY_=mtVBHBS|5K+lBnH4ZWV(0%Aku_)JAkSoZ|8vD<snCRw2&
z_gN*&5mUU+W=o6cB$FUV^6ZX@xq!!q$HzyYW30-CN|FO}WneMh5=a3BkeKaAr~rTm
z5?CBYIVlu?<DD)S+q|qa2a^5owR|D?)a{cnYVyQs@%pSudhGk<2_304t<8k_qpcaM
zv82^NCxR@sqM_zJl|Xvs(<|YSxuu}i@gmOMFf2uPVpg(~ItR{XL=k@cJOTa@-uroS
z`c&$RNso$qOWOJ$52>X$RkXVWI5;{f72018Hlaz!)9$@14#q}CVuS5c$T=xF<ZBc$
zM@LckJ4&8Sf$z|zqDzb$xsHwqECOO_2{D?Ro5ShXWoFg{`XP}>KaVo$a3m6^)#$v#
z16XlfTVql?)09gr>{wMd_dGiB=L}}M!`#~wQah96Ex%WrrrpYdexk{-uq)U8^waP(
zBjah*Hy|$@O54BBtGcWr*vz|$x!aE~3HXVP9{=7lI|NHm0oE=FCQ-+0uqES;-<rbR
zQQk2{YunhIukSVu%W2DOj2%I}-hR{cM)vH25&rZdtrTZNxw=@^s}?9F;=(RdE;C>>
zI1&R99-U9Z1=_;6n;lld)Sa48B|>k6PRG`Xh9~#j7ilFz>1b^_lnW7KaR;+p$JjIt
zypEPnMl4f#v(ZS~n-?&RM^<y^gtBC44FX(!hxxZ5I5NIYl1J_7Q4yf5_N)l(jfu^r
z9H|a6nd`_eAHm(-7rE2;O6voyi%4x3UmS%EJ9k29m))WBvd^pQDr!k4VOVmo4TYn>
ztvF;5d7KuaLh1AM7?=(9cLZmFA(`9jM&s$LGVmnPsao=unvG$+OI&Dma0Q@dG8Wsq
z2RE=WY}QODyL1#%mmvo*i%{ofoE`B+=~cGO>Svjp)Mtej;New>9g|pfWi5O?au8&q
zxtf=Dxm6TEW-@K=#8e`-II$qHEBBYELsWoT`j;OFp&cCZwUjF~zrjf@iyDL2OtheL
z<#pQG8=^$j4py^{&e~94MAt59LT|#%;K&6(DO-iKz6i78K=J*ZfZ)dVMDo*4VEdOK
znXwKIJYNoh`PX1h+%^bG8lc}HV>?JQD<*2bAub@b@Ho+RD^P+&D3S88L}LeFG~6b_
zXeWtcG@Vv>oPETPmT&$h=l|NjEBnjFp<hhyMA}b)%y;31kZDYkZ98Y#<>x77n7Fcp
z6hK8%Ob^bQQ-+ASS~y*vJ?YkAv17f_5`X2!^WZVd(tXvvnzr$N&W+#Vl)iA6^{lVE
zy{Ho_c4l#8iPAZa9JBKLc*tpMH+V(va&ngZB-%51N-9oyJ`*ge9u!EBbQ)4FTeQAx
zspK%7vu)q2+zu9Pb^csu;2&nC|Jrcd2aU?F+cwm^od+4|${K1l>UTkOx4Z+|Tz2ks
zx(p>Im~=l0V5Rc*zmLk%BdR;cYl+VhHny<J+Ej~g{j_^L4+0u{@Q<a=^MO3jXRzD4
zi?gG2y`~JaqxI3f!4C#d#Bxsc>lxC#-+YzV5chOg`-A&sG`)mc)yh^R_j_X(>Md%8
z*N1TxOqo8`*u&r+k>81V+G!)7R=R)nR;I_qx)BmnSWsAQU5dH(=iWr+AOF_BQzlap
z4l^<*XX16zbW6ND;kf9Zz5TY79X;A@mC{0&Qj+_llQ$Q;?Km4PTaUNTi95S2OU8x3
zMA4YoPB8l()cIYB6<Q^SN&RkdeBN?J%9Lcl#LCKM#mx$+UEtx6U=7+dkX@g4s_#S!
zt+J?=lqADUos*KxtPSvc-=0Wa(MaS>Zu9kQbiCyYw^DuQnZxZy%2vty0Y->0aCN39
zEwma`vy8(i`htgYG;!*}^`g)yJ{~-~bVqI2-?X_Sp`aJ6oC9y*tC&<&6&SJ+ieY?&
zOY7ZHA5)iTySicrE&!%f_R38Kg_wJgk@-;TdqJ5XXx>P|&}lZzVV;Aoqg@c!BSa4e
z4S<4N5bj#sCm9>l1=kPmO{w2pZ|IY-7uzj4sS|LYmfw``mvMA+yKWIps3&!~?SxX+
zguxi(HQP-UzT#PUftS}r$Vh2NIKpa-r>slFbVn;W+_;_R{J*6AUzt>gcQT#4K3Sa3
z0%&kb0FKco_-s(V(mQ<bc~e<USJ}+#JNdlBm+A1<mdx(fqWYvj^ch6*11e;>-6i(9
zU`#TmDs|!{4fM!0$QzQPTgHrPsDQYa>+TT{vegUPcYYli9`&s^Jn1cJ+K?Gy((otp
zr)F-~HTLEVR3p)WLltOUmn^%!DQ#qsnz!bqhPL?!ddG2Pata8L>v{7a*XLD~x;#zS
z^Og?D<v`-ja~<enk-n?YkZD8R(m1?wxs6wJd_7kQ`~xL-$fm<!F6Zfpzgm!qhP9zV
zoHf(5QGJC|8qOyV6(GXaqQLeoU(kR4JOTaTQ2IPM{Z{JhjrH$W%8c{adNXBkN^QmN
zMUN}yc@O%oZGi6<bOoj|kQmFu0fsrDra8+AC7#G+cICNXxl3v3IsO|hqMjtQ$ysy!
z<H&!hj<^{9V38`KqV(}lc|0z4$%Tvh^Mur5u#B9F<c4kuAvcFq-aX%4{AR(MR5z@n
zXC-+S;@+Cnetc2KzsPd2AzBq;zq+F+;cTiX$jx}W#O!~<FyDn{4whLrROOz=fwzm&
z$%{QA#)RHwr*UR~#Vm?fFVmU&cAhuVKpcXM7rJYuHxOTR%~m+3%n_*U1C!k;Dtorq
zH|Z`3O12=&XJV!`kSYcV?<NKNr{rQ7tZ6}`s;y)Ebj?0Uv$-fgui%kP8G%qO72q5i
zOeU{Sw?>OzRJ3|XR6H#jQ_jvcLT2v(Fvl?M@K0PMy{QabMv!O2guGQ+RWA|u8`O!{
zC#(~mU^g1m_OD}*D)JR)$MHjSK(IC5v-@u4<f(g^^thdBW`X%6zWj~Hs^`w(@Dj5=
z-v@QmJ<GmMnXJHfMC?gu<JhzERB%S|^m*iZ=h`xUO1A%$d)!pjLT=OMFnzS{obNzo
zW%=91`;=LkS}VduUUX#uzcK!LpT&IwO!j1LxrU2G%UqZ}TcKc!h>VY`w2@8P3fQB1
zd5_?~vOMUoR})~Xf~-8rzsU1Q$hbRrJue9;qxiDZB>>y^nqBcF^3}m2&B{(Y^V!D(
zwwE?K)O#-t0BC6d0;N1);WoXFr@x9X_5}}8kq{;W;?Vzz26Y`k@SSq|bh!7}rsod6
zm9hd~{O!sr+|4cW=(9be*9Q~$cmFD#|H6rhLX}p3iOhDWQ;3``?S4V`2I=_yG6Tt$
zA)j~KBNkZF+37$%T%U;Or3&F#EOi=xZ6(Z*m{tbXs=(|ahi;jY;~NIP()|vo@-PJE
zc{XchC$>55@dJ$FNQJo3S2H%QlaeK`$cD3_V0ZMf@n^&l2Rt*;*(SeJ;^1Om@&yr7
z{g!TepIRL6%ja)Cuibwvls->Re=GH^;#RtWi+w5<^D3D1#f!|cJSNg6If%LGYc^R{
z>zL@bo^>wp<$94$ZGGDDBKv{$kpAMMi__{U)!196$@~>~Z@J@$?T6(s)GyieSJ|nh
z_L+Q0Mnk}e>b;$su;{yVN=E>n)Wll@wh?;%Xw)SKgf-D3K(A2w{5tr8cbk|e!M2JU
zyr&0R<S+Z|3_Z(5)~cAKZ)sQwb|JIA9fj9+c)A-l?KoQ9J7U(=?mz!q?@lW0oy58a
zj{*o_1eL~N(p=+&z=BHq6G}&PQ8-pFj6W+GQ~ayfihQxAKZ4mEaNu}1_ri11V&=y~
z_w%xgSFhbkB;`AZREhLA7{%{dgs>wtpjx{V;WYdplo~5-m>ygnqD<`KfzMBspD$-v
zR=|0pOeSqgOE8`t9-cxho{J1K7$J^?8B61~vsMGn`iliJbF~{5+-dr?M2#J0cAcJ9
z;f^xDF>G$Pr>*eIqiMJPt@WROCfg3a!(|kklJyjC6EAJP8UV<VuMQR2-RraiI6Yd_
z>7QN{^iNQ=mcJp?o<Dpd@*9~i@`fVTx2DTNG~LZkwz@o>r5<+zyk_@?7;D%vw+ZX|
zO*-GMW&7@q%*L4xjWAc}nHJ3C*S6P#KY6+(46)ygCj}%gvO+^6e9JQ#-_Nh;&t!gU
zcsuUucpyc3oGYf~s>*zfu%9e3mphN+VV{~955EzM_`F{IVXgT*IRT|)PT5{|f6_4%
zByNxy5;}f2o@03d4K~VjsMglztCX$y_6ooK02wHD2t8=4=uqh>EbG`}R;Ly!dXJ3{
zx0VopdNKFu7u4~T)Ezv-FGCgVU~JW6r0!8Dg;AfWvX{9A@D<rkf&m*nIPxa|`EAcC
zxn8tKJfb;hLY<d@2ACgb%dFU#KT)ahNf7hf>i8!BL?2*pId2O0tZs91flMu5?L5!O
zN5D<vb%k|-&9#(5Lpe(Z%%)DKXb)WxtnYp2eHaInOsdvaZ41cU#cV5|J*|<3c0Xs`
zuw8|%@an~m8ZtB7<#J}#dD$%G9xonG&*%m!Y$$`b6ra7_?R<N;Bqp^?zb#x5eA8?|
zO0$gccWR`3d#Rik#W|i^FV2eXp0JU8E})$5x)%%P%dO0e=rY_ZS){A?EXDU;)^**l
zJyy`WbSZ5p08u(d2fd5Rex4v}#3I?Q1<GWHxi0#jj5J_-Q~dcS$A`H)jP{T+Rbz#p
zglA9lTZ`WYXHGS1=bq;Uj$FLxZJ1~|*_vK0gCe{qMUGMgOU|j+=#u~~R@nOndp<;7
zdI1RL33mzGkJJXSCx3kT30D8Vew+sqBf!MM{OM?pqQP-E(r;q~`psx@_a#GhX5DDr
zNNep<J1cFkK5l=@fp`xH|8aoN?KJHD{EZS&*YP#s$qaf>MQ94w-WI^Gco1&YXIFH7
zAN7JH6mx}QT!k{EG~y)4mA|jKcTn}k&!3rp@(1YI=MUt9<I@9PFirZ)6{AVl07<~1
zSq7pYW9osJ1!hJZCoAR6lnT7#baV;=mukl(U%~8{zD#WK0pK^{1p*iDFxIvWowp~`
zXPi?)S&X*qu8pQKv=gHBir%hFQ3fJ4<jV?uq3+yot9n9nUnQ>Ar%vzMBE|}+8ER{w
zoB-G+O9g3FqXss9QmgLZE^reE>l||bdxMkZ>W_2Fuv_ata8In&5;(2C^)vAbiWW}!
zmnc;sHdolo>mfr%K4ArE9z~DpQ(-HwMj4Og&hLDGw6S02FaTTgXtCE@G|+ENP}gJ(
zfUrJ%-`Tz`b6B1OJ}802QZD@DrLbVd%-FwP2aWT$$t{)+31@))#v*MmMI31Bvb&~B
z0hC*g+{vbd*8}k-e~SEk!Tw#z6Y=3Cw@hSosJ{R-4mm7c+Me3pDh>(IcwD+5vipV0
zH$YkF*l@@`8g!#6me3GsUh6Dz;qSiSKRW%xPy8QA=kK3=O<jLUC?$x6x5bOk7lA_O
zMK3YTWv_IkSg+a@L2q@)jt}JlgxRfI9t0n=?q21_KO-!0&S4kCO2Or$I?AVgp5ep8
zwl(}h4n{LBQc4eEZ`w{z`cuIfc$<31XX&|}f`jtiA~aZ8wq0#bL2x<axSzb?-!<UR
z&qB=yGt-{IEDHhc@|t%s&F!7vj|Zm!blq87a!NcCpMBI78hy^xIBv1MQIs0;Y~9EZ
zG>7D`3kGUdWhb7MT7?_S%7{)BYD^EFDgYq6O7zPC?{kmsUgDr$vE;>{|E{*rK0Pdl
z@?|V>wqga+EaPTge6!1^kNz<qKU;wR-GZkc3ku*ol#$dv<4vb2BBrX2+gY!COubut
zzBu3|f`TPBRL!KG&I~Ijkr>-{mPd;Sq(Y|);`v`+g0-mXNMnKQ<pfh!+{30~wtY-z
zqW2isZGHy?I^{$zS^%3GS#X3_?QV@;UgdLcF??Ca)f|!P_fNO29)J?s5xZWap~{qd
zPxm!enq`cmX<W}&W8{SI{;e-xuyaot1x?#jp|NG-+#dCLIhquz!J*GDMXSLR)aN#k
zV;!j#*WsbO9;~ud*pQ#lGcjEuek_DPPkmxEx`GQX*xyXu($N-~ie{{^7D+!F6@_63
zyzvSBqoKzF_U>Y=>wP}Vkul0AiA2Bp;AX=0nbCgMD3e)j;rQpJ*=CpQ(P4nML)7J6
zEy0+yb#OdmHUd4asiUCX#+{C<MVx)44MLacRT~Ef%r+Wu6paADaEP@cwjGYt(UmN3
zkQ~`R)tx9flEgQ%W2hRACwXec#Zko0*ViRag3nW9PeLJnobt>f8+AJDNV$7C#`lc1
z(opG4D%@S#cK`d$ra@VA?74{IMPs|UFucJdwhv>L>@8#1pjE|K#d&RA?WS#vo@=lU
zSY6M8ggGgnJ<l6;b5j>>;f?FK4HR8y3Dh4?E!Z>bVRo)m8(3j>Fzb-9EQqEVb6A5R
z1ExL8ylJIJ{1@tv|0eR!Kl4B?gTu2p{gH?3<^@IpHHtq2pFKW?JvC7_{}jC*#KJG=
zC_cLRm(Tmg|66!&KDG9YBFmOpFS>1}v3Kz5`flF7BY=EEei+&Np)AyRAw+d~<Kv+L
zr#?g&|IO!h?tgQsU}1%H!F!ebb@u#~t+Pp59GyWTby@LR2I5(XS2ZZ=8g%!(izTEy
zao&*ZwP+vRkS6w5+JuXJ-qDp6k34tNsi~Z|o0I|B+xtFC!}|Eb9Q&kY5EJIfW7tZ=
z^V_-519}tBxv;%FVB9DQxl>7b>&Km}w)H>Z&`A)i!&>Ks3P*=FQEzz-Pfi&-G`c0F
zGW4p|Z(?@z>o#M(MtarwJE3sw_c%^J_~z(UjVasZqd_|G0HODZ;8?L2?+J(LP99|T
zqx+4SeNCzx(54r8;I4yi1uVjDRr2Ti>eazTl)7ewqvJ?@=ILUKZB#M7sjk(WFcF2X
zwEboG`^wgBl0yieYQmRHpci+pH~2Sr(+0vwmx+m45+nTUo`uq7ot^OcmmckJ18c~e
zY928#*yF;!Mrq+q=*L4`yz+?Ja)2RL#zyjSuXDP_d{gy!(S=8sz0wLTnMl0lSKuC|
zAb1(aB&9bF4s5rIc6gq2{qFC^P-^H6JAPi}H7dk8<&AT0bwJ3x=dnQRWTv|P4|&N|
zMIf?!?C=_Ywx$KC3r=xWwh$P`N+-8D5}dq9o_280%x}ehZA{yZT%2&B4~3g8;&lAq
zR|@B}^pLl@lkBxhG=&r5)C54OCHXH$Yj#d)AnegnRvm^hZ9TRmK(}cw>`%IF;Ox%^
z-JD@|YQ>SC@iur<l=@R0;<?1v^R%Rsg%`b&)Xtv!@BvlTU9fB{7chY%>pMs0<c{C7
zn~%`%fX!&e?l#>vg%GE2P<C$IA6ldzb+ubnE$*#wTR|z(L3MMLarxpeTUc$jBj)q~
z*@&fyQ{%lt1?kgCUI&DbaB8@JB(w?AC(|@su|F0+yKq*+S;Dy|01IHK6TOpXLn>}s
z<t{{ilbL0#U{>ebkqX}$&?BBeH|ESmya?~TCgdS{+AxhSPV@w3Mmmc{leM%4z$q?%
z0Fi!mr5vcSwIaS7&{n($y}#9KpqDUA-Xk`+i+d5e6=Q1ZDv4V9&ofiP0*AJNGzLNx
z<|?xYlsF|;5u89{)Dl-}$8=R7MBbhky!q{e@p)NtM<5vE&K0E>)yAH}1^BGv(3x2E
z>C71#!fvSQ?Y{KBiml_hRcQ*5T104EGSJT`pKnfVD$7s<59XfTsm7~mec~#@i&TdC
z5C7dv_{`HkmN);ugk^%TC0PhXL~M6==DF06<ZEmfwu?iN$tZI)N^g->Q%eHFuimEQ
zTbM5nmDs_708OVN0A)(a{gjn%;Cwg!s}mTFd4LK1^<=xHM|YYE5Q#I@p727aGwJf2
zqJ@Jp0^AP<_yCQNz5~_NWq{W`3(QO>BEU4@f9~nu0yfo7pPsv0vVLwWHq`dV2<Q|i
zXY4)p%kJul(+Q0;;2|z+^5nGx*j>G%%D3(%W9+~L8xGucfx=QP-_4e@%kfzrImV=<
zJ5}VbwQsqhS8vZJEORI`GvQhbP7w|OM*$5l1FSjdh@SReYWdaWGp&Qc>TUah8<|2-
zPdoC~WN>lXsZFzk9d6U<7Tk)?E0%~}ms!2^_MBH}xB+dzi&9V$VRktZOA14nrB^LO
z5q=sL2dyFQIrWgYm{aykBA1u4)9PTtE`T4BVrQAqGHmVOyWvX4iRKviog@B~ZwG&M
zyA>f=3emS9Ut`5L`1>-6_8^&zUHgUke5yC(*D&NzXk$q;0*k^)MJU=Z4|?os=CNrr
zK@e5Bo<O~;r=7FLQ~Q`b164GhyQtd03&orz5Z#m0bh$53QtAnNuMMWvglIL-Rf~pv
zd9fz=bz4u5TQgq9SVMbDSjYgKt3jRrM=cB)D`w9xi<yV4_M%)cjs}H{`nvp}a`@2V
z4K-T*<ElJm-GR}n(zK@LnZ;hD5;IwS00LP$UnlBob~3$#l_^?KaZs^m6lTVqKfWV5
zxROibrMAWhYWjktBc?|}&6~2m7ertkN8b9{srPtw7WmbMvKx|1QE;UDvzLSW?loIR
zdE{-vktBNYxhjYev1s0B83@>o&jAsZ^c2SC%2O4c1yf!RW@)HC=jgn~{t0Kj)fwcG
z>hoNg%D|LC<kHJ=`MSEs_r0mRkyeB)abazVXDBC(FK1{QHC{wz`CTzTTBb8fUhGDm
zkjCBIPb2)g8??Pb`!L6(y<*=G>7o0dxjxC_ABka+vakyAt=}@1MQQh9_Et{S?GZZ6
z(`hVtLQ<+H%#sfnHX0q$yl6Jmw_{_s#<Io~PL@Z?;^~y^Y+TRI9=o>s*v*KLrBSyM
zsi}&ZhB+PC%oYznjq^BS4)s7)Ii`W~);;i^u+Gl0iu$^-vKY+V=v#ce;XtXSl-R}l
z+IQL|^DIGfQc!>R^12YBZ?Ic0VK9oR;35}R?gI?wCFM_ec<si%Mqh^Z2|<O!U}vnd
zVv_U}!nduhW#7FUuMNUHm??C2LO`aw6Znk=1M72?0<H#x*{`NW_5$2@ZeB9YtiVHm
zQZ+zNpMBSuX`9%xx<^Gc25M{jRw35SPBJQIq&b|GIImP<f-lxNics|Hgk9zt(}INA
zDg=vXPXMQGQO-0gQ$+<A0CwU&E%fe$c`qxhanMoTXquy(OI#Cw0Bd8oo79pkc_0pT
z%6qK5c^#BWMLLx^4;T2?o4v|WePv^1FkXO0ulKaAXWy~M23F%=GL)E`%%O$|4Qpz)
z*t!t{V$vYBzib<H&TDR7mI<x-5GJp3_D;T;=bTcX_jB#zuQV_%IkSzmuUr;%olnj*
z2JS#yJhzX-^xVY@OO8%n24#i?Y?(Zam%3L?y~j5z50P#XZM0F$|L}p|sB?0fR{xca
zeR?Y$wV|ufb}$xg`M7t=?P67C&uBX@^7LLAz+MJ@|FEijCp|vFQN?9x3*Ybio>^Te
zr2{}3`_<q1c*xTzQd>z+gl5(*`@(-k(RnLc<JWsbe3JxOye7wpp{!I)Ki{Fbyt=Mm
zl_d&OH7+JbW>ISsWqwF)xbXGAsMgU(R@dU0*QX0r%{4K^n9!`1E&rejUTx*5TpPd<
zKd<VqPQZo3NqLXHr85I|QmQ$8{2)>YMh^3Q2XKP+{Nns#@XdpmiE~xnxw5~YGPm#(
z`fTi!%artUlEYLJTIhJ$f?*0TD4&lo?|ApPe?P{dW;tPC^l_6g`AX*{V>g&qf9qQD
zS}zeVqZIMuuCea`=;r(;uLS9lj%VP;Kv2J01+K6(Ls?_jZ%uVkB&?ygs3zoAMy|&4
zT<wYA13sKUetPir)968jjAG-Z%;W*<n3baYWz#mOJ58Ap-*x#9bX>>h4t3Ykb5lzJ
zyGN3p(`=lP55Ulc0g;;NEiA3Gl7vUQ$mj_fg$-NC;CnA)GU1r+!po0`PI}23Ln?`j
zjuAua8}(>Clx#EyU9l3`x?M}37aq9GOu>A%+eitEBjz>BQck@BE~dB>I`$%zF7lL0
zLMS9x+9zFY9}3;7qCi*Yim(`q0~Gcs!o{v(U)|z`uy#KUc8pEPI@`Q%*z_~fBNie9
zEg#)RKXC6wX+R+M6#I(lYFYHknrmYN7Q{HY7J)3ew@#q${xx4So=7lAFYvlJG%;Q7
zMlIQOQxB-jH(HgAS{BPSlJ7^AS`F|^MxS!a)-efhw$M*>peO2Uw(lse>^j`|K+L~g
zGOu)PVpi>;TB!8&Nq4~K@v^#g5{<2uf1EsRc+XlQ@y*9WcOyDw?RO8P31$QS{5QLg
z9?3Sz8yzMv_oKI@lGin?!wnm<aCjJJ_}SA~Pxi$7>!b2W(U-Ie{|T5N_B~qF$}slX
z*dz%a(>GUW+XW6VilB5RsJ$O=x0QwYD;cEXTYC-^zkT5C9;#5O|2S*ni@=T--4c{n
z@rtDhIXP9G%r>?|q85C5OUk*@JdL?gY4_OI>}>Aqj+vK5v!JcS?(?R)Z}~?6l$;;-
zug)#AsAdZQRFm}_2rM9D`V6!SbRJAs*^PnxXWFZ3U38n~Oy?!b$;AjvL6B8j%kJ7Z
zKtM4Z?2djsv$<o4dI{dGElC_E`tZJ`$1Y^SRPm%Px)laU-!Rks)$H?EL+aXFM5EqS
zpQgXEgZ5-~$3;KICw#d~jYhG8+6)A>T-s(<O7kzDthb+W{6FrLFwQ7Ub@IxD=P6aR
zXjp6ZKfauL??#`7_4&Pbk1y~#)OPYj2*%g-xfHa>uX~Z`AlJqaAjZO)pUkjts2r=U
zkK@<Zo6jC}S!Ut-=$>E>JhCpeG|^PH@=&NgEjSEWKMom%hm*M~>|ivgil|IZz)bb*
zxA&S)&t5hN(<Q!IDS*%8dzF%BGvWC>lha2$eH`<HO;dWwzK-Yk3!UJoFwsjTrsQjh
zExy6<yG5D<cZxvepF&jn0gKMp-RE130z?KsW1Rj!6!}6%x_JHWhcp|1)B{{6U4I5$
za^-Gnk+@HCS?%%XS6%FO4vs568Y-R`7+mqa(Wfa_H-T^%$`ct_k(*bNPcp*q$>MH=
zO~xL%&`&j(l_U9C%LCz2FcB~p-5yy)XUynfxv=zh^L25#zp`f42fTUZW~}z(BZlS+
zIS#DWF5W#Op{*JyA{l}WQ<fd}Y$b`sBWyni*?fSz>#g`4-IwG-u4|+PGhXT8dS4m_
zPFHihI7F@8N>iD@POZL21@2q2camsbk-yGZXkAb;JY96Oh;RNvC2GqkFyjGtX*ZWE
zlbV1!ZQ!h9%H65G)@J)Gy?uCRpP<oRB7@jryI=MK+>=Hia5rG3Ide}DG5tP_=}g|;
zqY@y2H3|axQE)PSb5CKR>Flm?loX_>s(#A2@=i@c(~TcP7vs$fx6k;FyC-%6Ik@L(
z{UTT@QusLKRkv!|tCyP>Y%qIsQAaPuUG2V7qgs(aiGEh7+NEb`wzP9qG%j^k=v7vr
zKC!Tp>kXaMg}A?@=Zcn@lM$QQsa4p*zm5rhS{*x!dLbqM*f!S>5ip`|?Y<d~)~^1=
z3vf1vX6&DqH9JKcJ-hA@_byXvpV<Ai>FLcIj|JWg%?q-+mgnjDrjb|n*6_(2y|etn
z#}T29r?=oZnKmoiBB@e3US?sH%Iyg^s+^+!MCQqW+Ip-od+xLWT+`g_qHV5mDv((j
zyJ2>AdXG-uD|2Rq7vRrW`%<Y^o1tRQKiFk{L%IQ<5;maQ&H06?zIEEHBD%#(Py8$M
z1<RwC31@n>Boa-01Af_|S5!`Vr_ca;*RWV-Ahc3<BS4yRir$&QQ&Y2PZRMp<-Q3$t
zFd3OWZP4oUs<5!!Zw!Udp=@5u>dkPZ3qgN)zke6km)b{qwr?+cE4N>5^H$L@;obCI
zN=T}7>|!`OCyw(Jh6E{7Z+AK~@YxsaCm$M4b&FY)f*R%%eH-E})OWBoEhxu^R5dic
zoadFU;kWC~m5~yjUfGc{#&Sor0>B_E#@v=lQ?CDJ?n{S$Xt_^D?nA54Eo)30&fE?O
z4q|(4(l)o3)+Jvg_V4c3L6T`_cc=UIB1ojEoU+o;x`$^e<fRAW?tQ~jU-2Wgu4(GB
zl4tNzC^4~=1}gi>MTTp!*~w~X@7KGHvX*^*N#<YMito!jYbc*z+-^p-p?^u=(y8|@
z%ErsF$Xr~8f#VOkBWAGnp9tY&lC^@a8P4QuGrX)$&bz58hE`0Quw-t7-*iXc7M|%k
zU7M||qpBu4y>n$Rw-#ZG|F+$V=M2%=-A5E>foLrHz~#%g$^bt!p2G`G=UcV{K<`->
zdLa#I!WEMiY_IK;%@LvoE~_dm4^+)MGnpT3GU5D&Sz`0Dd}OPn!RSor*!bA6i--Ed
zh3OJ=+31NV)q4?_Fh0x=xG%}ObB*l}f(++#;zp%ypYAEB7@%<38G1G*$z<;hTajpN
zM0wIV>p1<<>gwJr5z~tz;01|;0~l(eJ~1iWU(tRn>-or@TlEffk@b~$db+TqcilA?
zSz-yhWUBYe#8X(@Ho|}xlrpFiSh+@rAy@>6%DH~^w8XKwrGjE^o{l`yj#1u;+9o_G
z0tt4R!$gxP-*YVOFk-by+%|vf8-ce#bQ3$-eJ)ok)`rR*%ns%12`5|`&3c*|z2UWT
zuO9G_@zC06ZWzm3#-$M<56TGep55Sb>I9RtFur@mHpFA7+I%WhuS%#JuT;G?N6PIb
z-0-@Xp+l54&kiXw1do!*S|y2wyl;xj$Z!7rtN2f*`rl9G0CrIiB2p0fd~sfPKqsKK
z`vK`MdtSXAPUjs;Bq%9rg5p(W(&k>yAkUQ80ki77k1FF@TFyEFJDh1a2*%%@Jkk%W
z>YPWg6sIED(P@#cfz=$}>j8!&AWF>Z*fNrJ%EcdLf70zL-ZE!s=7B3sN0{yFS>;7U
z4;q>+8!BE#w4-eN4JKB)+fdkQ-6eu|Z(%x>-&kg8%M4ufXvKo%=4>oz)&9}%R%HkT
zw8#un5!X4uwbhvr#+pAQnDvZCxNSM-z^tC<|FYp}p*yZ(N^Hk0{7~7l#H=f$VKCrx
z_i#|(Z>NLp|NQaLX=85~k@GZMNZ{6`@$G*`?Da6{{O}Wizax}tfhFoJ8QP_dP7D<)
z1vgp=_Z*<^am_y^<)Tq$><xsnK8$DxA>(ok$6v5^;s~Srj7Vjph4Mt)O-I}3>9p8O
zqh1%&LllXGtz0i93=Ug(Q;21k)o!BVuLVZB>eRcSWP^{_NtBWBsa9qDrdi}wa|fJu
zQmH-Bf~y*=gc$9_M1qN_M~*AA$$ib~Z6kM67tZg0yLoo-<DsWAL(rmyBA4E-rKxN^
zy#$ySvWQhTN5GE0Xrc!)X*m6R-Xp2Q!3)lN344>hvGM)v;g?g6;YE}+Ljyd)ESp_M
zRbDBpCSO$d_n$4jxkoe%O|OPT1iNi#6rM0z95ZshlN%V@_!lTBr(#8!)11Fv=<xAS
z8R!d)VQ;*&hG`cwkGfp*jCzOa9Q{^1@_0&3+67bJ0S;H9Vc5_MVVOQvv8nAdW|>?w
zY%ob#THZvw+XJEQFNbAve%*cXlbRnaXGtP-BbXx@0)o+;xK0Nl1Oo*0s;-_tRPPj6
z!%aAsYI9A!uTVFD4cPP9uSPZnOSQVE&hUIq3kl|BE6l5S*~R5WY~ZM>jN9%mFB%F*
zrKQ%(Ok`a5daK^|X@Vb4sX^rE#gZf&jT3bZ5-F!t*z1jU?}jy(hBy}&sZC`TDQdO)
zcdp@NyhVbvbS=Z9dLXbV)Iuv8I^tL$UfdDCYDCz?+lQVc5EqZ!oispyLFRXM@G{S%
zA53>IXR;7>JBqwbbint8!P}dG+!X|@cg>p?8f$2zkTDyo+=W^b3Y|mdjK}8(d;N4W
zcb;WqE-H*<i8h2bN|76tUTgw{=HIvNX&dJE+9s+5H%hGLoG_MIp(Kq?sYyq*jP-W}
zbhM)F87P#DMKdF=N+7&|c4Y6_(4R)RT*Su4u=Gy&$3LsetP+?Aii9dV1Y&IRs`052
zG*;*EPqh?uxaLG>5~?a7{pw0*sy^w0PgaPvpXN5MQ2{+dXUXWhF-MmqYkx^ekv`0#
zLQO{ZY#+vf4bn2FTKil}-QNvY9UvpGxZ(>FEa~`=4q5#Ot`#RUv}6tmgQF48$<>YJ
zwG?VOm6DJ4#=yZjy+k&#R}TRhA9KE)8K$+&?I&1NA?yfY!En0LhP7s;iCQbuPmf&b
z$r^4?`>Nwa?RD?wT!7uLHA}QGBfzsgl(;_oH}DX$vePH<kc;LHLsTH~1h9@k?D$pI
zNQhbHC}6kt9hC<TH#Z!OvA!XNdj0R;XaD3{ToUV$>$P14d$Uq|Ad>BW>;q8nPg|T;
z3b%Gjde50Zy=&d|zW-AY<YyfJZ==zDA=!8Vs+1$O*2eHz%K3G63}iZ#R#Ad2e9)I=
znlRQYIje7fH&Ha0Eit+=N+igK4f>AKkYlR+9F?e<=jHVy%Bui_@2hXESB<N!+0gq`
zx-V63I$ZIUR5;jotEP7q{<>phLLUll(XcCezMa7ZH#im3u_iu4#HaZM$nBa}>xt(K
zQhEkm1MVn_3>$531CWO-G4_jPRS?gLBC1fVqWE0z29l1@HsZ|_5@hXayUZM$M@`k@
zB95PY`noC=UYPmq1x*Gf`p%mp?z(c6{nEuPyHxiwm*;Z_B8>$O;L7mhwZ;JhORI(&
z!nggM)7PD1mAs&RrTzG(@-8QilXDBafP`{zDtv6xwSQr&rNN};%<2(hNGqh@Acb~r
z$Oqy9L|z`3QJmvCKYS4-?$xm}>aT9Cs%AIv*d2_~F4OOv)42O9qX<nN9wb5!E?`AY
zHc2!-4$YssfMgMDa}Rf?-V@Ad8&LyuBu^S=TpzsBpGsWbGX7G<QZ9yb-Qjwi9QO4G
zabm7mTxvm8M%{31eN4#x)E_=$^8I&<95FYjoI1{g{`5g@d}ltm2$|VA<~%e|Lvs=M
zQ7)slHRUvuZnLyT%82=v(iT;(^TI3U$^yWB>ZOCrtLmF16{Tf=)cnu6CnHMs%tCD6
zLUSHycjuL)iDFhpI``N{B`h_Mb+rilq&TN=p~FidUqCzo$*yK!9JvmMeWkQhZYSN;
z5WJF9u-%C8SIoMM)XxqP@wd4VHnbbs{1Xy(erm#U$+OxJ7o;ZEeU6OqHP~*Y;cKXE
zC25RJTyF&VLHszMIMZK4*TR~ihSe%N)LRJKoyqX&pjs!y0HHv#N4wt4fvKwesvD8q
z>gk=YJhSuhkRViO5H7N5lIc8gzrTI4D6$poL1-S`)Y^EE?LT<aYOIWJl{z|IH80JD
zyywiUc5e&^4g@Zu;7s06uZ6X{{iDJ=vqJ;D{!MLQFF9L$Z|Er9*7z2CM)7>*Ebpuu
z?w%Y3%Fp$;a+1IJ_WMkDQKkl6lR*ddXTl5P`Ux5xKGvUzJ<fN8cBsX<p9lWnrg{4q
z{%O%thjM_Y_bJg$Q8cC<2bk0We%@W+@4AfKg1~7f*T~D8o~xk&_-xYFgr7#;D(!pR
z18m+vmq6J;2q)6BWfO{#NLUQURsr*(u5Ie7skuCPdVd##3USd7JGX(&_A)9R+1Kwl
zo$2GN<!ilq;j4oS5r@K$@+V|3^dQzYH|Lv{9aWseTYvSvKh#=VADZQ;<H1eImU7f{
z&QO>jSKgYgr9N9f+cQwUoJC}W-(SwLc)*2uVajA>t1CM;oK&%H?XvL?(a4qI!F1Bx
z^vk_vaO8(djoD!TlF<51G#O-EzQL4BBd;iD;a@+j#1UU^tia9|`R!p4S{4qKP>#vQ
z_6$4vFzf=9Lz8^X70a+x9P3{4zt&K#C+vJsaUaDL`Kjrth?%fSoS_rus4$jau^~3S
zi#8nCr0Dg@ReHt1H19WWTo32Q%>;1m<vS)MyYDmTkG^VT&gl(LKY&CGOb~bT81)LG
zxk>F^@24ZCpJ)VvafQy{EPc_87$I?`shj=;rBw*prF2_z1ncxl!|4;>**+yz6(n`z
z>f2>x({}}aeiNtTJ40>+wwgG%FY8h7EILfFB~?w}4+gzczlm?`Z*X$x>a{6=BOfj)
zvyWS~jYtXHgo^x!9}g*P%RV#AL?O+(M~61+A8|4Untj&87?hBcN-qpp0028w?ngS_
zM%e-|abUl)8oljlFi+Z1-#gIA9|&;YTf*q=m%^$!M&%kE@Y&LiE9v`veX|M*&nOf}
zQLa^`y>e^dSnYA|-j%UAdM;b%Wa>e+z3XU59Q6Quzjvl&2P<4EbH55*JUCd4w7Ou3
z$mA<*b}K!P`($iVAD2DS!Y3%)_>tS-DB1iV+kvVn=#zaqz?`n(7&}VQyIG8ZL0T*f
zCW`A}rk4*idLfs$tqa{%m>lL-2eOwPC&WU;vNN_q0w;`cBg)2*+1^AZbs5D%SRH(`
zQ&jqU+<Aw&)s5f%5BA<Xtf?$r7q{9|T574~p_r#q3Xoam8M+;ym`hNGgh2>X2q6$;
z4y3viQKl%!4529uh!7$}2tx=eLyRFrrUVFs5C$QHDI^dkf2pdjuI}eK=iI*McmBBd
z+_V13UfIuP?{9r;WwX||-tT>9jN22rS~Ib%IQ7Jo6Rg`KHC&)Vh<~mpL$h|$dz@3e
z-3Fo7M2=}VWOTd`jAe#vHEgR-GHYz6y=q+yrd7`~_SSs}swu~Rm~AzWMr22TiU{X>
zKW^qPw1w|(2GHNZk?_|K;@aePU(S*XqC*x}=S8!5N=`uDLvHwb#9H4MldK6=ef?<;
z#D>x5y@|s*07jUvrD3qV2Q8ggZ_mT#x`|=&M@5JXR?HuNed&4vU(fWBBbT#bCtvww
z^@qeoJYT5)2-=iZ=lLwQIR8sqTkR?3caU3Mvy$Uh=EWG7gqO<_?`wh5Iz&3SW#g?w
z?kldz``zi?)3cU$-_KjxZg<(DK3E0}cjt!0dC2*{pX;>gO-C<}EWq}CN!7vWOjcjC
zF!?E~940>ySOy-MkIz_G_N(nkb}>?rL{m$mu|dsl;!9H{pj3QyPavZ%f#js9zhqU0
z7%<_LxOVC3-5KwC(AMz5_$=~dOIT59Fu5*#EHgIAXY2~0MW|^7J5|t(L1s_`wj}b@
zS-$KoiQ6Bi06wgd8k^43BP+vbaPWP=+*YZd@*B8!+S&e4=q7?7-qA#cBO;EzZ10$m
z-SfiX2-O5^aWfUyge_NaG<`XbfuWSO9Nbk@#6W(a6O1p-0pn4tkX4bQ^9h7;y`$I5
zg*rb4@3=Y|%wcFyl*gtU*6-+esIdG#`d2^yFF*e$jGXx4uTH8iN1>WWv3Jw=!&Vfp
zHv-5P1p{<;B%<UU(*SDZY{e<t`6(%>R6h0oCN_QAG*{6TTp9g<lsYXEf*Q(UMl{nG
zpD!4mIYx7MD$4t6>w0Wd@4k3;J#2o!ergNPON!WRUtAZ+7g*=EZr$JUlB>7JC%u?X
zhYu54DmnSn14DUd%<XS!h2E_E2JwqYd-4v@iDs>m&)Yw;jF66#_;74~wATdRyYHIc
zxZx_pteIdUQcX5})YeRyYm7DidL9Q(21Q?8oUY3m!MNMaesS-*%V*YM_60_Z25Tj|
z1>w6EC)EP0$PjPqi42P7hRpi6fg(Z$w7Jt_5%?<WOx?N%U!@ZE=Nar}c1&44aSIe{
z?yXH-)7y{qb^KorrT_E=UlHb5bHcyG$AY(;lhtR=zAeK{>AFp=m?@9;^RE<ZE6f3!
zp*z#I`F&XP_V~Axb_Rxr!CAfeE|As=zo}_ZlsnVQEV|=E_lL?xgcC1rRZifu*0e17
zd7HNdGoUxUd1lqTwW@&1G$SryUb0{d6RJ52xD2c6^C%VTSb1!JYm62+*_m%^`%_pS
z8WXuLup~w&`)B|M<Tc$EH9ES!E+m3>t{-H@<MqNPS~d~!wc$_DWJ4ZIBUUFzknOjK
zWf{J;ctn`C**9)j(xAE$H_}EX#z5RqiVgLAYTwOmui1_G5{QM(rM|2)7SJ+0R{bt!
zJf!08xw2){?YTskYaa~Em|nA2Y)EO+WOnh8>L#tJshF3XHFen0vp6+xthY2ZH71In
zcx?rDVz0mo7EQI29*-@UCik;g`yM5WV>GK^WRku6CEdC1Y?4~WIrU))_3I$AJ>!zx
ztvcV(f2wn|Y*+caHS_V^aj&**(Xp_k-jYRoO!cwVw%X_9kI1*Cz~_wI>=!<hE(;E>
zGq~E@54x`EUv;$FjJ6q?&SbEBMiCL;SPiY;JM<lY1-Jg{s4U8AV1^k3DH+Muxb~&(
zYAzN$n+|MZo-^y2DdrUKE?eZgC~&8{(uIxlnM5nOB)IMS8CSh1R$*d@VZ&0^6vbfH
zNFQK5Vm_|5XrRI<*RAbHD1Rch-P;z50kN9&v)Aj6EAReOjXU-~SgXJ8F8-nVXxJnB
z#!MRSYha9OootY-)jTi14X|$5JXdSyro==mQQoY)shY8=%XK<)&S5HSsJ$<WYeJQ+
zJ7LArCEm_TovJT43e=-_bacG+nev;N^7ftaopkT$?Cik0F0QSuEk5JdYuDFnxL86m
zGSY|X?wyHEbXTn1VvVb4Ua$Kw{==_*HUt?D*c_n>fS+Y=u(=Gb5fE*l6`NQVHjmBT
zGt{@}CMPn?62}#c#`TP2lst4Pa&Zru%9ZcG8>#V7YW?UQ$6L5)1#x;CY;>o6<DVM_
z|NB?{!GP#WykSs&F1y`t>+{Kc={3KpBUeVe@t-=-p_G^0W~>@LuDpM+!nJIqJvi^`
z+OS$9v(oZZ{fb8qB!4GA;p~s^7N5S@y}k=tBB^$|FgWqhCh}eM@iijhxjgD@fL325
zffhE)ZieNwbX%-jB21fi>*DIa>-J6}q8+9Aty7uCgz3tiPUo~&Gl_$a2lY$+%-L<I
zKHI9_U1gki-4VD}oYx6ob#w~mnQXX^kP-p<-z}VTnQ5zh9D8(oL91G;ee1+dX~&pR
zS-EkPY~}+u)dUMmx4tX)o3U7Eje|X1KD)WvRxzWGH35oZw6yL!KoWts`Pvx9o+@BA
z5qWdYU{`^_Y`zjjv=t+>g@B2~!b^zj$-%<tpsp^f0Y*~bIBSAZ0)ZGoGlg(JMIuV6
zg9?Mfy|V{=I=Pb!dQ(F&Vn!Qyq;F+$1LFCx%FJB@rC=rsp$+9>K_5nCpZ!C}oa1j7
zbm!vb_gJK>AIi=(PgGg$<)NpjK*xwkgn!kWP&-RIy$(5OgeoQIS*3ekSYjU@TrysJ
zvsQUT*LCb_=YkZEv($U?nc_Kb<Aw9z?~7~nuX1Lr&kUQO^4`WyN3)EuS~b0W;W4Ab
zrBCr%RCrsnNnKps?Ed^J?Mh%>d)n32GFHg^#)O;M2+;$QJHF%3Za#XVFCA;yA}k6}
zQ8DCmN<=858Fe+W_<Ze^B%`S(ALNfqzx9cpw6>HdDSdXeO9@<v4KpKL8V`DYGgD}R
zPOIzID6Ve5qa>}8C1;4X@fmswvPsRC_20>>1zbb^>1d?OA=!4dBpEVtBqn^aFg4!@
z;>fKnX~=bX(lI$lQ%=sk?*`DRyV~}b1|_HISc|iBI=NOmvfL^Qrhd!b=k$$Y&z)9z
z+_-!cgb~Nr<s8nW6hRc=fkRAT5eRm2m>1=yI-|scIGP?zjAFTmsRH#qo^)PgN8fw{
z^Y$uSLQ=r5Td&iBs?)Bn46F^!*~f6}v9)%VGJ>72=W9hT1Q7>+6IJTk<-L8g{o-At
zZNU|2*y<d)0o#NvP)JVBlwOv;vLuQ|Vo}<1tz7TIEfBsY5ugPGsHlX?FC#oi7EehQ
zh1@Vdd5%G+j#nWb2D}bv!TuC5TKMP@PfWpAsch@#k#Mdzrn08<KKIkHZvdM(-}vsp
zd7pF)5SFLi(9%$WU1>4Ctz_@rx6nMbpLwqKVZPqTeVry_LF42|?yP1|2dTz}fqlvM
zVE8#%ig6M;JZj1a863GFje@npmL?l#w}P?W?}#5a5H^ZMZ+Ws!64j+LGQw(cD6d8J
zTqit8gr?i0n@+^lk~h3#`veKa2UHL)D_O*@Xyemak`&1aK#2*3Fu%LxmM4&Yci3mM
z{MB84E_N^I2-Mu_<I;3MX@!h#jI$6&myC(GE5_%a@4=<ktl!Lkyh$PsIILtT{$vHn
z%T9UP9Liwm7RT5Zh~&S1uzuvyTsw-o)iwoH{BB2xvP(2RMPW~~8>3iVaclA!AUD-A
zXg#H?HfL|wL|B+{)Vs}X=$nQEN?T@(0h|HitB*%%{I%n5D!$}2j|O;$r&fkx^xV#7
z`z?x}BpH+O?Tz(f>+|oVHBc7ukC^0IeI)YgK6{liShGdHxP2bBBX(4t?1n2XFDeex
zM-021QLvbwygz)aTCvz(!qfLUx=y7yEN?yGMCRk_@F&P~Ghv6E9fhV%`8Nmej+i(<
z*S-8qo9_-Z%;ig7K(4zVS(~f$eLCieRk@^wItl2k65jH-yPNn($eL1xu&m1xJ_zNe
zLHu3=(=)=tL-8tkVY&VybNb=tZ6_fzwJqd_xKe^lw1+&D2<n`z`gAM}0?d|KKmT|~
z%*C}#cHjC@wU!+ElE9$4T)wDxl-M)>{HmmcB*8!>|5>!^P|}hkUkSG?s81W;^&Mvt
zeo9$?w~;g!7efkkZy~iH^AadE9Y1P`PYklE&eFWg@vC=wb5t6nh0dzEv&MQ+2c&GK
zd`CLc+7BZBavH>$+upIwEX1(~ELs<L8*P_d1-Q{Zb^Ovlyt$6^D}3h|b6FznM+?(v
z6zoB0t#B8S?0QsM;MLqK-Aji=bMrjAAat3Xib&Q-M1~{{(!9hLar=cQh={)(TN{UV
zx8^AJGYXDzb&TCD*FnQh9iS*|f1NSf+|$!@h=hRadTaH8?g67Bpd&=rN<>`35fkG-
z!ZB2Lk!LlD(n)Ncq{QgjrP#V=!sJYkZX;VEtQt_K@-|36o+_<W592?}^tae<JN}t{
zzWdh7*Uwo__z7s?zNuxO_J!;hpN`2Zo*Ulaanf|ik$cg&_vUpW+_H0H%evT_3PMBZ
zg4v>ZynrzUd(yeR;a@VN89CkxEfok$qQCrM&t*!fjqen`Z<>5SmRw~ELrW4!j;+gL
z_Yy&yh11>Q6i9sf;bMu&vPhB_e9iEeU8aBV-@p0ge|8<FJ0pMb%t=5})`XS$vWrIo
z74KaWbvn`Ii%kJ6=B<}QwEfU}z&pTDKHXzD>NbaJD6L;Y8cj$YxpJlUe2f*u7^Lj^
zrgWHgtS1fMS+d}}5UW$LZ%u1^zGSLXM6k%8Y4E6)dH=F+xOFGC)d_UcrV}t4jxW3R
z8903upLuum1z8a{yV>F^cov#Oxre*+Yfl}2+**(qcXVdqoYi84`Z!!BPB|)K7Pnq(
zG?;<*(uqv6YRDx}=X91GESd&!%Cb~|-vw~AGq#M|4Dp)&9tZKoOLs%)UutiM`6_iP
zd`Ovqe7(6lS^mw+kQdyh7KO6}BwA}h#H(MyZ7`bn6_+IEH{BtPo$XBZ`HRl$sx>no
z^xfQ2HOthh&K>@ngXjPM$>WURCKGJj`TQ~caajwi^Pgvt`WNBjGTYp!U81`l1G{Fc
zlpF&9gMnV-UhlEdF;VVd@L+{=VlGXyvF{2u^y3;lsT~26uPu*zJRghSzIzwnz%m+`
zUYriQ!x&|De}~^%8l#Ca3fRw_a5;QzKyZLuLUh3E3>Cz<XfY=jNBcR;0F06_qiqFt
zR`<9+eddG(i*gA>%+J#Ic7KY)1+ITO7H;edz(yy)`T&QJGR6Bb#k|{<E+_k1A^YXI
zaPigZtsU7D>?lhpF0lBSI-~M2Ww_fR+Cw(A&+Z!Z{C#9vz0PdcAnAA&sP~>u_<Y0M
zs{DRF`Rh*MxzR3M#yVh3pmD&<`mpI<YAp$%>CQ7pTejz_7)Mi$ppFsWrG}|}?HN_0
z&iDf{sFk8GVn1?o7Y8fI87)`P=K?}PK=hIuTZre)?C$JKW;qF0|Djv<ro^sJwOK=7
zhk8fVgeNh$O*3{`OuVJbd1U!mwDD+AF<7f%tS;zG2K7bGa6ZN!y^=X3@-Q}WFel3_
zx4U=MGUJC8Z=H){`?iL?qW^#onJr9JTD1AZWSndz&hBB050*wO-0;V>L+tn0#Sr;b
zW=V~J2Fo%_Es|}>x%>z9$y2n9J`0(NF<iM7kj`X@oIN2Z7kOA|`GUUw1u>1N_`Cat
zPf%~YadT%MZ>1f`Dra+oi;9Kp!&JM=TP?!WHer{DLcsk!?m(0kxZ)l0{?oAsAI5xk
z(fhZotK^MmaX>nc>w~)`q@8DDal?CSM>^dbrHHNo#;0Qg9+Jlo*vYDUsv^`55A8w>
zJX~``oT&1fE5s>xJV3G`V85NFWct(`5Ow|i-E5tamw@wfVP=buo^d1$HroZ&HU6tt
zWR7I|N`o_7##Y_+Qt@S26S@IE%%)5_7z@DI^MlcN`&?S;boZF>h(&|Tqne|=tNoXR
z&<rRXLD%nWdb<T=Np6Z2ghazsyoFHl^dZuIziVE-*$ps&@vR^~zg)gH&EcBIiTo<O
zlEPUZ?_00$i7Roq<czP~^H3rhawQJ9#HFaHfN9|YmiKwTfm=o<Y1PBp3Np1ey)!IN
zg-%4P%8fJ+3vGs{hEvtQA`RSpNhCw_1tb14?(v80vrX*PXT=G1T2KwAb$?aM)4N@K
z)+|R`D<z8skBBLrFIoIJg>GvjRhCSFNiI*w^JA}@x`U`lI<4syEXv%JE1s1(UZX&%
zFb}CZbZB#DF0_6mcj{j7^|%VDeaq~L$8{OFicnCXW&4PLF^m$u5nuiSXIU!n+t+yJ
zGZSove>Y)k!uft@6l>@+mk{6m<fDUBUzpUMq=&d3Kq3~boc2?TJI{YOu+~-2N##<W
z30)>Z1SR7_wrk6~ahR1I=cPmI1*Be4TXvThx6|+2p6h|*lr-@e5rw2g5zom&F2vlR
z4$v}93B|RW2^7;JURNhAa1i<BIygQ)hM>9?d*C84z+!m@O^&uhqRoKnq61xb6~t9z
zo!z*&%0Jh`Lbs}KnUcMm$xapo;OF)CN+t}JK`!*sHA0rnjAgZtgQjxkQLXm^7^J%V
zI5RL)#ic%X(|AL;qOB>pEMfsbz6+j6A5bth2ECs%rDVhl_IJ-gbYutoi+J4464bL^
zqDPNty8a$sCfOF*(zT$C@8>WT*P58NMt#qa5{uaAFp=7<he12cJqwtKzwCi=O-(fX
zg~M8<S~Rs$*`_;VOc(MKW4;wBj3lVsSIG~x@v8XZH3FO%h=H>^;h@fd-F1f>qJ@D4
zW2D(pG_i><p7<Gy(6gJTrmL$jCMvqRKd091IXb$`>RM-78?!d~GH6sgtj*rjQu6x=
zW5>~i3T2X_D|3sb;MzQl7@{~RrZp2gY77R26{jxu=^j3`$PRfKX&LQ2ndZN^n|;z&
zE`id~zXkC2#ayzMCJn41OduEb7LTk`#~H({DT>rxerznyWi|tN!ds}iTa!*`{&ehB
z?zxj2E}I%ltPalrv*DJgdWo)%ViJ8o-*wJACQGJE*pc1}t(}d7?73aNQheq0N<!ig
zWf>jjN-=1%QQLd8$=Sdz_sI=+tW=+$*xB5lE48iOOj*J=B-$>4s&>xv;^mWi4z0J+
z^f7jX;@-tUUCEOVwlhY~Mev0_$PiSwl_g7#oUJc^33<j@t2|<Xy?r!{%||37;tM@B
zY80q+$*{ZzmT|Oc%1)l1J8uZ<mTB@JJinr%FEQ$V7+^83F7FI<DWC!O1&w!S6A_bq
z6|6=(voh2NmSXbDxlrz8uTebinwfK#%P2*H?8};}^e9>NWnJHgf8=w$ixNK_lP;Ir
zEw^|u6T+m3g_a?r=R~~cR`T&&h#RN{PMB}BeS|j85Ei$mfHv!WCiChS6Ge<U{^ZK|
z^3_GF41<lClba`fo4rG%Bh%zoy#0TgySNmBU6INNcPiZxbE|f4u~l-<Z0P6@hF!F`
zYn1Nq>=pPayZeb)w4(vKHFDqYWPds~mF%!;o{Z}fWQVi}<i`)Yo-Xdd*Fx$|@K`5k
z*mR0>cD{6HEw%!a;pgC<yT!+Ef*#g==c}4Fx2t7=K%TF_lEDFoeP_LIlkEr(A5jue
zbZY*Q(qb?bvOv*Y+-W}ae(_A<@wxU`x0<q8?pl@yDoLc9-lfdh$e%N57be+Z1(v~U
zbKkku<c!1;V;n4VZ&RE!8>#N=pFe$B8*T6Rpkp_~Yvp_NQlbaPxMn&%_H-=FF$twJ
znxJFChAHHSQv2YoQ2?lqFXX{QeTq)u=%W&+NVr*vcyEHRHN`!Qsf<u~%`nj(t~Rq+
zm&LxlbFfS{g&5Lt5ZgzT+NNx912zX?TQW<rYqo7UTIpwig)@r5<%RKG3)z2KR-wcU
z$XK3#NUH5g3PJ>H6l}yc3P+0{bQa`0M&RDw)vB>`PY6vk^vMUOD0IuGjpJ6Fd=V1}
z^W+z0sx60PY@)C?;G_J#Z$ZrJyR)0zh`Z=(2D{G!=eMh3+ftfxwXmM0Pa;6+kAx+%
zK%#pmNIrmhcQ<7re!Z-7#7L_jOfI2I;=|+H91T$Vy7zQ46Ip|vJS*!FmYBjjH;E)+
zGebXYL{KIf(7z(vLO>4-Xo-S)W9vGoN#=v&ciseEPPmHC_B)$LXnqUAso8`kQ6U|P
zw=wtqrJvnbI^nsS67q1>BDlc#6`UAn+zpRc!;|5YHfsdszI9F@Z*^>VdYB?oRXTg%
zLeQ2YkVSDOVo&A3K+6Cc(cV!k<i|$(MlCYOdvis5XQ)0AS_%sX5mSn%8+7;3e_j^B
zX8*1@qng+(hit;mk~_P<b9*~}r#beiNwhW8($5Jm&%pQZ9Na8U)w%T&i_{)A^f6<N
zT32P|^Pgfh^VF%%?v_VGIGd=%%8ezeAXzJ(GM=nN&m~5QFuSI(SHt2l9oeWZmcCQM
z+$G|)pEwV3h;+{l8-i+3KE{O*KAVbZj`oFAjxGCYSOz1FI&#5XvIXQ}+7W-rm4u9T
z>d&1EzW7s;=l1H~S^OVV|7Z2|+v@3ai9>w-qR-0Ywb7%85NY+69SiKs1@n{dMxY#;
zJ;ulQk7)KSwxbr;G#J%8*CJ0<#fPl8)@U{kQlt75h`BBX4fp?C=J}hi&oKAD_UUhb
zwbd2cF5WO7Mn3u0tg|3+m&E9VIbpKd7?19clVzx8jCAEX92i_%G}uTg`ZoXT)%dxi
zx?vmrtH%ca8BNUJ9{0a<6z2=FI->n>hf0`k8Ims|_Hd$<Kai%Fij|QZF{$;b1Q#rv
zdEq9S+!*LJ`a;aKuioEE)NYVPyLyG>Bz_-Qx!Seur&yyvVs*mzZclK~kE6!+zecPg
z_5D;M=hsUdYW#_g@tMWZ_wrU{#xG4rMQi|a4GY`Jd%Ov)H!)w##WW6BPE*i#yNr?v
zIN2XH3Z68(tiyZp+v{GafnGuZN9@Zz;N@@rVe{tH(~b2FXRa+meRd~f|1?oVN!Rvq
z(tTlSMl42_R;?&qv93IKXqK2SefC}0V^p3dH>uRB<lCUO9H(@S64*$+S$em4m%n{)
zcDiH=2@F90Wv}Q#n=WV>abi*Gbx-(iIkW?23JB*!MtSYcpvW0XrG!#<w9dJf_(=@o
zBk4)3^Rfv+$8NS&!_jf8q!7D4R=s>sQlsx!5^a4y82@r5h0~Yh+ocY(9I`&AUAryj
zDm?#Vv%O#OWNMa%Wk<fCD|G<=fGwRo$^VwJiaZ5X(Yat5c8lZ2Fs!vu5i$CjYLo2I
zM`6r7Jde+tOB)s8kk=qvsOLWEpN>7jx9F6lE+ec5*n0&h#nwcD`QC%oCPD`QFhbCJ
zw>Yx^a{x-tXkeu_TtWPEsf;~m9A&n-a@|x-d}3aZ+}0-FrJw>*>0xx~eBFP&kJK5-
z>8+gFsyc#MuG!`Vi-ynfc_Uw|_XJaw>Q=DK1Ew=LS<J4H;~Bo1YqzlQvDh1vH_}i*
zUECINsuj={<^{X;)UjQ)j7X8lrExbvbxLBc&1cOjU<TE$>SiVVNYn;J4-b&718h&j
z#Un(RjPUAORtxC?@~ODkyl=6P*K&ps_oqX2XKa_`^4IEaXip*4B2VqZLa`kN^;E>B
z`34L+Z*pTjkZ(Og9PQLkOq+%eZ5np!#ZDuYVO}UPNo?DxjO*kfQKTKTd6)Al@^)eI
z2GJ>4wcjb8Yf@Ow6q5q=h*+jfLGqb>b6RSi0Z{^z4w<jEDuc3{D}ul_3g4&YX+Xm6
zJ|9jK(c!9nd*7ns{KU-;^CVy!wxO-5fzW`QTAHBcT0xG=2dMeX&aN(SV*Dn%*Rx99
zO<rlE%8ZeRn!^TMvhF-@`sx?-*h$;{C$pm?Bmtx;AmP)oOkJU64&z;tR?^d$=R*vk
z%kBItXaoN=78knW(a}JtaZTKUu;fW`?ln=yU2U(KwWmVzKOIYy7b(CSchgm6NhN}*
zGG7g&o^W`wZC=J!*|oy*XTEXqj^hy2xjuVFE0+?L5WnWz%3he8*BuIOvK5dXG0gMN
z=g~xaXAJ<SD{!+TL!nmaUOxi+f)PAY*WZ6<tR4z(7Pd&2|7%qUq7&`Rh{;$e<KaFX
z6Q6LFZ`#eEEU%0eVcSsdWV@AG>1bjy)O97fDhq2eTq)O*=BE`GCle<v%k}VS%zfwm
zQ8Y=YvO*;`?I?7t*uLG)AchFzH6r}rdfO3x!P6Q`sMR*e1EP&1Bg1bkIlVak%Kjz<
z!jnTMB?qM4-PUj9ua|Ya-tSNpcZsPoY)DG*36EMt1P|6t#0E6XSw%cuz#jmW`7dw!
zWn?4@VmK2%@3D{=2msqj!}qFvbZ^1gB-Ah}e6#EVMKrGJIX^1`Kh95#rW!QaX%6*$
znH%s+Lio!-a>=>yyVMMA^FGIfsOd{16qIuCh$uOHFl@R>H?+7P5D?&QHHwFdpfW)B
z;9s^K@F6|DKczJ2TH8I^>p4_iIq*ik`u#=w*|rRNuyn|%3=+f8c`+w&x^|}<DkE@C
zqxs2~z6ykHjg_qE9_@!9Lts%Zg6e9G{>6B-dHI~F@5y{-)r>f8v*f^5wI)|M%~X85
zvM@Zw=-U5sVtk8D*q3!E_XsgU50Ius`XYVx2%w3cEgNO@Wgazs^T4nZ5_+V9Ke$L~
zJp#{WD=o@7KCNM!0yDRrZC_$+;zuwx^w`l&pw{Vsb&;YOysvOk)l#fQ)nCZDQod3l
zQF!?2*o*LOEB=(34HpL`l}+rQ_ZTPH_v+*te@2CFdg%cg<}Sg3uCVQc83`0@xLZ}u
z`l2)N5!=)|>H;c<y-nEcaVboXfCP-y*>2jnDB_m+@bsL-Z+>^pe^7hNnrPk^F$x}?
zjG=EJ*pVK86kRMSnzncO(7#%rTf!+0MR=Te`19F+qWs@Xj<07dcp+@rDMW;JoV|}e
zGTya1HQX9~pC7+>2nTq9OxDl7l~ul<BkcQ{B>7k<8I$wu<rWXSu9|sKdhYbn55IY*
zuq$Mz1QsvWr(a<X#U3z>SRp0>tw0UU;XWW_GR_jDS8BRZkYl4aTx&1M!e=d<Pg>IY
zG8L0I|2P#G7HuzU^tW66cFX^o<8e#B(Eai-IP+mPFgv_{u@)D@%-%jSt*|NGEFMZO
zw+>ObG12&Sg%tVS`e=(kpSnw3E|?D6uGs?_UZ%SRbv1H^)3@mle5?i=9;i|F%EpBx
zdX4}kUE8xcURf8p$J|sLKt!^LRfKYV!v!u(S2CS(c5jD9SDkMB=sIj-B7fHBi-+a?
zWtsf_qk->GGsqWFkeQ7GZYXMS@s9G7(cHglsQizdklVt20^(6f?^Gh&GVO_={b&w*
zpnF%j;vOSv^hI9*6&OoNwj`Z&e?zfQ$z6~eJuExTpVcaR76E(l9$yQ!bmJq3Q8y9?
zQl?h?Ane4h^Qc8Pzx5(j)r;@Apl5z-7OIjsZz4*kD~tx+@j08ES-9E1NRxW$Dg>`Z
z%EL;yvfJ!dZELF|8qw*<2jR~o>vM%Hpk(nZqg!lSScJ2ub?K2-^RdC4X?ZL9$>TBY
zQrh2hQ@tfYrd7<@Mv25Mpc#JG_WE19Q>yvEa0S;?+3fEq6C-?K5aEn$_`fJwnsN=O
zNgOlj#6qdvvVAAZ8+?bYG}mo-DXL0d%$PPxr@2{1?W}cUVKKq*Etw*njM}i^I&ztn
z7)VFSN=v$JvRK#%cJ;n%eYMs150P7z#gV*8d@`Y`ZYU&eh?opwr6~tYJ^=L7xgB@w
z&v&zGBr$MW_QydC%s%h**EQc)Tz&d9bg=T<=H;6hXHPrlJazb0FSViSZyPD!<~W=R
z%Ks|g^Xgakr8+v-b9N{>0ugZzHfdTQ(y|iC5+qFwi?po3Mfj@VUy8>HKE21iLByML
z;evul=kUpe3)>GLqApz6Mbh_3et-PC{ylaq!<ip%z0D#S2{i?qm}}VyIg6Q}j-8?r
zL08GgTbUjA(YI74!&4LxlHUNXf-k#bg_s6h83tr;<mr=k>J%H`6}Pdo*~BU6ZgKK8
z$MI$wQ?k<e!l{!jfn)RW7M{V=aNCm&JkVoK0uiXZd1OCECeH+HmLoIgElHG?L+a3U
z&6JW#GR>tG&Uhr)I=Jh0ZWPPrrR=zb&racn8el}%=o?DoX;z9>UPn8hj$!sy|Fo~$
z5<?A{+K_t`{pa2fv2wvo+_+yr&3vS};ZwLtQDY%cSg-CM4QC;Qaz0<?2M<hxYiN+d
zPK>Mg#YUZOWc{RH4-8{r{6fy<GfGURclL}@A9Q0fzprHhpP-;>0g14wIwFGVkM1v^
zWLR;dgY=wBg9Wy}+ky(!xD}wh#!GIbxon{^H%&oqw<Qupx=cw6Kp*%SpFJE-Gj6fA
zmWKyp5M+WDn<KC3m-_ch7K4rrqv;3{QOFtX>iPAE;5O20NJ&eU-?Y;TdV1#C!daGU
zH_*nqK}#!1?BR~yr(@AD!^^fy_W{oj|B{SU1nv^~xgIAj_u}5dOV9h0xhQc;ZbOXr
z$)ahhVt0x|bljrLGDWidjo0z*xL6jcG^VtG(7qcP0S5L9i<u5jCK=s+>vwuHdQ#H+
zQ=T<eyjc9<AI8JK^X0!U7HqBS)K+#{?MJBtt0aE{Iq>=P{K^Qv2K=PfpqVp$tJXf0
z4h;c3O@=$2s#tK<O)I@9b=6w$P25`LSX&AO-$9NlGQz_lE3R#|CGJm*$Uim{li*7j
zHY>DnkJ&>nLa8@Bn!n56ZkVH?)pA-W`d(Fbn-awSWQZN=4bqPTrw>xoy96ZW!A#MC
zccpdFPSmGkD=M(fGQ|V8p#vNlxqMqB2qi*0{JGp#Tm93qiG4p_O2}r^0lfqD>DWzu
zWah_D$L3J)4&sjT{C10wyVBe6Iou&mL<nxYh?G?9-*5eNj4x)qJAgX6{Qva*8j3cb
zO@yxbYN9WeoW+Cdq1qjOR6(q<n4P-|64ebgZUvrhqyW6ssrDGO)pVjr<Dg;ZX6}kV
z?@#vJu%)4KdJR51a+MQA`f?X**ev2v8U`|;c6b#?XkihyN_E?@yDNK}Y9C03#prnA
zwfHE*LY^4^VR>Cj>sF?RkKnf-m-)59gU)1Rd-9*?z^l(9BFid1eA!7<yQ*6&vpelE
zf&*VIRo;6(B1+GH^PK~)52mA2DqH(c{Z4;-!asjJemR>5*J5hL8RNZC(XNL6ErR3u
zjrmtHH)~Ey!Dyu5nd3pq-w6|!$NWP<t2MR)FpH|VL6F=<aHQHpDpF7}h}-G(v&#g>
zk+FjrSpRhPaa$cFG(@qD)F%*yt5JjDxnF{JwD#f2gvRv=#3842LC=k|)TcuVu9vxf
zR9Gc5^u)RfpLV+-)#>*(9-oHj%?6Lh2xILu==EgdM+Mf(SlENMk~6aVp+P|vQ6Fp9
zG-#mpS$jf9`P8k<rahJIJ<t+-xXV`~mmnD)MLBYrLYs|89ZdQmv%;IIO?8uMdAfaM
zY6*MCmPCTAq4YzHx+wd7UDKCBP7d_)=9;+8vpdYi8&JN0-ZV9baCx+Mh_q?+N#cZj
z&iBi1AOe|dMvPXb-OKxV!7pD;&}1iRt7>*O2nMyRqb8<wqKWYJvmAS2=RGA6=><kv
zUEe1u{Z`HJFleOf*5Oq?vWY8LcTIB9m}uh`;gvv7s2A>C^Mx5_DQIMOdwU}<klCHs
z@I<+Y`g)8G#P`Ypby`3{ea_sCz*nHMKI|T*Si7QCzz@30i;u=0(HhXh;Ve_^g$v>1
zTY(ha3kvJu?n<TkhpWFk@z+r>pfe$c^{r0tvZlB{8y>E}Zj@<&RZyjBH&JIk;uot$
zY)_t(?ya>_NNTMGGN@T+9OD?XYq>o)f`NpWRwt8y*W*j3&6)mWiRPoCOwY|>`a0P2
z@C$Ly(}`sJ#kr7vo30Dus8K(PnH&4aU;;@>HQ)jM`0K-e{`~AG>!w^l7+OjW`R%Zi
z!DTq=%FUb$nV?h3+2iiSp!^F*eV%kc!l6jb%z)>_U(E2b#-DiNm*Gqa7418WBsy_*
zFNlG&WeN}akBS6Cc%GGNZ-Dg|_c_}#h`qSV138qM7@x;Jysf{Rf_f(0<R2i8$}gB}
zCg3jrZp433`|q36@mr&*)_u879fJthj_OoW^;)eTe8K5mDXEaU(lqY`@tZ$$nshUb
zz8)S6NqkbQXfW&|UBSEYuN?ztM7o@D@|8WT*l6;{gQ-$a0mOQy-m`p`x3pb<QXGa<
ztkvDq-AB%e$UTK&8T`xJzdi9}!oihojic{36+imk%dT9n4WD|=Da#_4K0(_{wIlE~
zQ;WDT%6W8{)PZFu|FB2&z8fUeEG!v6HooxNJ^qW2C#+@V`3x8Tfq7hYmc!(O5Vhnu
zS8HF)=oB|8;}Ls_C#1VS;~|H61n}I3|0vsJ$|qiMhw*;o^^XmQ*>W<wn{3<>Rs+xg
zMhhtAMh^6n8EuCNqz^U$U%k)M+aB0|_;(gVXuZt5Ex8b~$BtnzHahvI72^W&YE&t|
z*877l|HUZUy2(%kE-)B8EbP+--(<Z6F}TU)KyeL}M3?{KOAy&}69DR^gtIXkia+78
z0S$^Tq_44|#_VPtFb}dEHW9+7y>tymcIoY7EY^PbyC?r#r5jJ!j;Jt*jslC35%#yH
zCm+NYz;Z(0Slz}nk{9mkB<Peg@2d`}RPqqhFK(tW?*p$S=MzWZU8gOAGXYwbu+BIu
z2@riv)oP$TKtNyN$sPD{_H@K{9lxRD+M94I5{PMw<o!o%qo0s|jYW5f<elwBYvP|J
zY$R^|ohe|&4;N}~k6Ng=k6ZQ<8O#nRx)zk6=c4)(czMAzBZ<)Kop6}2(Ee&HpkD3P
z)>f8!^FgQUn+w@kztD9<-w^i6QL{OUbYzrNJM2vQD4|^gYW0NK7y;T#SsZJV&+7NU
zeeBd)>Ji@G+jen}=``Z-!f3Fj2$!%yYZu*uys3H(r1(oW@ycBlYx}_OqhfyeJLCS{
z>Ss>~c(4GLbmQP0r~HGk??e61f2IRnHBae4U!3Ws`&vJntUvqJ&u9OM@_&>-*kA2E
zfwUSS{DHwf+S*()_3|9S{J1K4E&$WnRb2YMI*zn?V}1f1dq2;jb8_Ys#y9F2Nauy2
za#lj6#mk(qsAAK-CbBcZ8rucftyadk#1NBJLCz^R*z{R_8D8!70f>$oscS!irC>Jt
z4<m!hFYMLiTZ%&W4>q58-L6e~^;aYQKvkeD0klnbdPRIngbZRbq|>8zGJztuiLNnJ
zBmF_z>&i>ErOX7LUOni%mp%Ko5Q|cTp3Y&mKfuSidZ_~@d=sw${)ry=PhAptgW8Vw
zk<AmD+v84QaMfE1oLcuB<m#-4@3@|4@yim9?dU14(kW6Xf;*Dvq2v5|xV^v<)x%yg
zBOmW~9X(JX7Kqoh*`dh&PSMxNj1;uR(F2_}C-Bg`P~JRYCk)2r0O{<d6Mc?ui`IU{
zDFg-exIg}Gs)Dao-;!!_P<A865YnGIh0nrBrzE_XoMPmpo_2lnRc7*S_0b$VM1@pB
zpZ3|ZOhECztk=5K2ILM^lnJG1AJA<_$>H%I7Gg9YE)3P4Yb#G%9GLP-v@Jk$T}lTa
z0FPSl?=d`JUTN8z?))<oYL(=8PF6CtMM&)Wu-hRcVUPkFJJi14WxQ%oI={08FD+1_
zEFLAlU5BnO9J#?jDxz-NUCnvtOj7^wI_-00bee?4?IWhV#&@5N<>25S+y1UGzckD3
z+5~p?-H1_kRqW|^eTn%(ixd0ZB`VH<lv`ZLoko*GrhP-L`&XzJNkSv7qt4S&nU+`j
zrkuX8Wg1tTc7bN%$+?gC1;gR{;`L6>w)H#NI<7F`c+Fc#_=bg`S8ZuohQT(QwkUX5
zu8&Vw0kISU(t66pb(<Io?_7i1u0(7Cr7e;*!cEoMs_n|VX>2Qgts;sWn|H0K_Qr?b
z9iQ?vbQI?<ip#w4!1+G=B`!0n&Zu<_{WUHFUb$6c@jl;9`Mzi_^FH6xM)$s=WT&pH
zP|vq;la0aUDs#m#YMv#49sDmk<N6TCp7=*ag_{fe7mNx;NX6iQOP4JA`wuT&QsU_A
zZ$CtlkDws4xE}G=<1Z{DXMxnj*l}mwS!bi+rtM`LPl^>iZDccc{|2+P%T0RSNDUsB
z+viu4$zKxDWxlSO!5&VapC$ILq#?R<#$(Jw5L}BHSx5rUS2eDPMd=CHwD!!D3tYKV
z=fcp)zJxsCXoi!tPz8L5&v4hs&7Lzk^d3~K21&G-t%^WOeinfu49}O;#xveKuYPyQ
zzhC=*;9dDIvPpkR*6-w&|Ivtc&!v~FL^%wj+*&&8n%emVgA&9q)m9hKbMdv`QBHAa
zo5jgVS;f<R6Iqbb`e^6Z?B`TY>;0g=SlZ5N^~dxJBZsF5y}+#pQV)5MS&Wts@g|#I
zZlVJ34N!sQo-n&^TUdbYhT60&1n}4j@j~~*o8ZFI*z!5TqVVuSLiu}|zhvI~-@0QY
zdi?W`3fJlm&rF#6ko?)%^S<10DQp3-G7c(}@Fnbc)T+J|+5{c%em{(r63#11rpICo
ziGeNDm2t~Ud-1o6v!sU3MErKU|Kj8ED<~=LOC_lm_8l=bb`V2MHY86+XfFPRTW0>K
z?m-udymkL&qp-pm0(2;AS(pFzvT&Gy3jz#$B@%~RrnDemtKQ8PZeTy+48sRYK-Zo_
zVk{{oNsY{qXRE7xH4(#x7crQY@u0g2Oc9PpEmgs=aFJXh@&&?g9`(JsO|v`1aWYRR
z7sc)kF0HaapcYdeVWIeNqzfiE*Y6o4K*>wp{owaT%TI2M&zT)8&7?frEHjsGG)hQ`
z9@IDGn1c7|$<M*2iNJxSQyAgwX?Pf5It*jCcz<X($L@ub<`7jHbcdw2UuEq=k~<9H
z#*Sr$@A;O;$}5;#V!7$j-bZD7JD<-XrSWZOHyeDNqZ2l?&e5@_*YCYaN3<ogE^5<`
z;z%6G_uhWUTk>$zjDQ|>JqzP37L~ubq!Mg#?Uv17G9Ul&yQRO>{-BBoH|t86z@0k^
zV;CO5-j&Qqy??o+bz%q*g^dR4_YF-=FRg}$4(A&}=6M!LN$qzUVn-}qcn5Cf->o<`
zYXGL}9{u4E_*xV`>S116Xw&Ly<FAe9h72E(iy7$w<jRSVbLY%K0fNFROD6>??loMt
z2qtXh8|qI7lPpJDQtD%<0^bZO=MI@?Y_c+3G*o9>0@P8IBqh39I1YHX8?(C88~81b
zPf@sXvEnlj39DCsYr|)%N9U5g+qr2Eb&iAcR;xOO@1A{-X1H~I_~s7d>Grfv=gR@A
z|Kr?ux>@<d%}2p^Uks*I>2(LCe&?i9Ir>HJX7|dXj;)o<;q@imRU3uq#;77QMfZnY
zR9ZxI%ZD-c5(qHkBE6ffw1qQwA=Bq;P~pOjR#ILJH8x`N%?bd0=p@Hq*f}5<_f^X$
z1`B+x6PlWKpmv-5PseU0h)5rsXTuWT{tdYDH@pAM%75GKyT7*fIIq?|FY)$ek7{SU
z%*YZ3kmaRZV~<tO>9c?O!VN8EXK6ZL{Ieoxu@|OTkM_kI-FsUBT4G`f0V%nnCBt_#
zNwQeCt?GNL_X|drVt&|bb>7Rl@3@F-kDDqD#yZV4+(pDO=x;k}G)eA;d|*dSeT=;!
zTk+}LjJjK)Q-1RevgcplAnlLhm<f;tCYhZ+uY;IyuS*F&BisoZDtdS@-VI6U7Td2A
za}k4pDBSwpvfW8$`t;07rmi?uY<JrF^a6^x_V7SN#C&jgNrmKZc-8W+j7MU!A5v$5
z?Ynhxwr+Fdn{7a=&M5q}NTfrs1762!ZCFIXkOpAz*ks+~`}zA_m%YcmS0>72^hZcd
zm2zuS%4H1w@2OY+IBtA<6i<`&WEx=2Ph~P<<gJgViT2H5XF_Zh)0li3QW5s(jlZsu
z#1;h>l-Uc?(hAqqx@@N`3jrqH0#h|}9KJaDpA&8TH%H_@ocvrnxdz?rJ-4$)ik&%?
zQ8HhK`waP-CAd0W@Y6I4#HiPkP5Com5A44n;@h+S@!gucL9L4v{fw60^mOQ++WUOz
ziT#V6w|fc9WczIpGNLPaR;UI0-fXeOK}adcT!?~1mgN-li;Akc<lllffRbdtXG=?Z
z^ACC9xw`su<~X3j`pNMPc6naYj|WzKj)I2wo`8a{G(L=@r)f`JLMjGhAiLMq?fbWM
ze>eOOgJ|lE*SfLxTlhLSTI$m=cl?`n&AalhTb-Awr->PqJmu7dKC|YaVnE&%GX+(1
z(9=rIAatH3Q_LcGn~~WRV9bT`f#*M8j<es+uni$=Oz)pxMHRlzJ_;@ay>JgXVObZY
zYlu@CBg7wOnbX?|gyU_Vs2B@|4Z@7i;qng3K~xvM^#zQXNR8qkPIw9xQv(||x!Vv@
z<+<l*Gp)v(LpUwS98`VtyV1Yatcx)BH;RAO)OPwp1i82>64Y+G_6{7JZxxxT^hMu=
z`~C}3h{2Q|TCkQ^VY8#Ci1O5gj~=1w99<YHRt?sk(gnM3wu5;;mC>eSzBDqz5YBA{
z0S9i*I=s6JQqGdMz4O+a{$v;|3Q$)n&nSjeBw-H>!srQBryq0_b>uv4d-tzoNcoS}
z4=ohGBBk<0^yXo0NnN+96e?#T0Y|ZV(cTd)>YV6&Z1qgIWdTYkhZ{)u_&njuQgpT~
z;I5t{XlB%^ETx2HY#0S9oNGHpQ10;~FWS^2@=I2(_rzy}E@p6&5VWJ~cc?H$v*!&C
z3xvP<;IkBDh-BsM#KY0m@(^R9fxx1C;#9Z}f5L;ob0xis%$LD>@lW5KuT>=yN)T^H
zJEkI3F~+SF4li`HIv>D@HhjDP{n0lyD=xK(=v+Q`Wi>{}PBp|YmqhX;bt%on)!6gm
z@HJK<3e1?Av_F%9<Sw@y+#Rx4amanfwq&)-DW+bmR==CGZNt9aoeXC0%P6U;5cf+e
z7R>cFmHpVl@(8Ddq4Em`o(mZZdw(<gKdk&W2890B?#~!0(=j?}=J;3p1%9U8CyjvL
zPdvAa^3QF`QkRt|Y91TLZk^!P1gZ~Xzl<RN(28@#V7+P>O{6BQh{O^QcqPFRZMIHX
zY=vqkp=e|nd&i&3$8xr*_Y4DObU>e1HHlNRLzeZ$fbP)j%`o3lf>q0akK+g}9RX<K
zixd=)K@KVHtc~(vMLU$(1x>QWpAIY%mb$BpVeuUWQHFW|5`c7dCtU~<5AsEqQT95q
z34ysv8CZi+OkcRk-fgZL7m&I@bF6Ah_C=Knzcf6OG|f++%EiJhLlP%Sno9zR6B8v*
z>-x?TZ(6sD7^Q>ldwPOjgG=U`kBl%4^PNeUj<lq&1ADEs6qHod%8kh2L$Te2IuQo)
zCkr6?V_{)-s@l@8NriFIfALlH$=_62mFQ$<H!bHykM9eg2O7(FD{IlJ(t3--wQE@O
z1oYK0weN8b=YM>e>T(0$X%=USy{9CLkRW7HgV&*&8Yf<(rKV*LPR&9_4<Ys57ekDk
z_YBp-OB>#qI+z);{B+<>w7kxGP~?DUN2*6tGcYIBBBHhqI63LLvd71x&GP!LU6ezE
zF8}K{zjfn(_0i02PIja&Brfq@e{y15R$fbY(}#-%Y+~=W#q$pd8Xh>6IQzG^TJJaA
zZfb0!jt_N*XuQu4+hoGWGZKPov=VCQDOS>*!L02wi~-b2@MU^I)?9*4TP##7ftUMa
zh1g;qbs)e;_4n#8^<vQb2PvAlQtaH}A*#_cb_V0|x(`SZwlazP92OM4dPmqU4Ns9u
zI190XI_ARqnp*%inhYFsK8WfkwpLi$+)@VeA2NbwII^N7CyqvILi8pEsDGvmW|3^q
zl)v_4_AcrgZ>tDDit@uH2oeRiDy90$!rZ8e)|YRHnC_{YBGjf>FOR6b)}`Fd^G+$y
z>V=yTGTT`PFeNMsW*hCNoKo;wl(#PtZhS<Q=U8L1oyx-ZlBD@Y4SpvE>ciXjrQo&%
z6}eqlctVtK*H61!?xW2qj~=D{l<54`Z#{D|4-}M?+u78|hHAs5zf7I#%t}qoJ5BH?
zAP`~#D<C0Y=v6z;q_>E1Oj$-^I|z*p8>Tv!ilEdlQ78J6lO1JcfnE)eFwPxHLsLV!
z@|6$F*AMb3C05~WTHf&ub7Y{Oig!L%&7Su2DuX<#%Tnf`<h2k5c(A>ziijaC^k-$I
zB0N@Kgb&fz*S&Bc^FzG*c3;|!L}2fv{u4r_%M-n3^(|{%pPN~xvwpcd8bBFg#}^)z
zQbTPOHSF+>cFVE_Yc-v^m=efGs+L2|hY{S|r(;ZqgR6<_tPVzt1FLQ&u-ao;3MYx9
zt{@bAG81L3bySyTDgE5w=~Y=NEk1K>p^+mjo~vtAT_(6?myEQosqi08Q^dO7O;|Vr
zpRll6#c0rSD1+>1cDFfYt$4JH4gRiCSL5YxZv5@|ndFhCuYxG&=3$&Ce9tmZF4nrl
zWPd@P`lC^8>u`luKI@E~a@PpFVEC%o`%%KaIY7$0A7l{f8F=}QNZ3aE1?Q<&7}ez-
z@=><KY{_Vu_Une^GIZ*A3itJ1TK9*Y`Lb-8Xh2LG)+xR`!r8Arn3kJlW7(x20#sG%
zbyJbqO3-(mN-P?TjH~Y`aDI|qN<#IsW@BM<mL<MpJ1g_VLkPALz&z)NBgk{^1way*
z*boExj+^7#^3?%D#*XLQ+*x={hOPHFzkjL>@}-EM>S0H1P;T#hT`ff!0-<WW6A$&k
zSiCIun$NU~Pp~ofj3hF}#&)xH%0DE?TY*v+n`7FXRiCnX!yyij**lR;kXy8el{ug&
zGURnm0XdPyGHgx}+Xk&eJU|X1R_Hakb{3$m@y-~NW>@7c-2jN1C+FmgK*U;NK+%5e
zk?KIF?Qh;aR#<)lch+$Sxi5)&2OUd#;S(~uTCZ_u(PBZQBs3AuU#81W>U{0L7I;5*
zGAMTW<5k&z5#-g}dbRS^+j(`Li69-r3F*t$o3u?gGQxNd)KvW@Vy^4O)Qn4X*2jZn
zm7?e<DXyYYQtNsEp}m!F;z_a2g%lUylxWWN_{hcK*Od9ucJ$LdWC<~Pgn~~WWbW3r
zHf@bp(;y9@<tYlr;-bwd?GdIxSl2T<DlAeNr69y9&w}!@^0nX5WwV;_WOjp~Su<Qy
z6Z&z&%X1uXGba^WR)#&L_Pzz*QsG>7CMi+R&OS@++iGXKz?`ohU^`@`1p<98G=$cK
zZz@sg8yM&VF47yy4Y|2S2w@{&TyK$@!=Q-rivdC%Fdl1E^i+sZ%YI7E&J%I8G>bxi
zSPlCgcRUty0l>gNOlMsBr6?^gwEEron_vH<%70(5&EH!6!RlR1+CB5QUatYERYgVr
zw)59$LXXsZ`_gT=r)H<R;YY9V;(9nE%oG|DM7E3a1!^)B))-r5GCM{wAz1%}1W42p
zywo|(usYW^har-rbiuXEILr~R_rhl>3OX5D7IcELiq>4PDh(Qn4wse2H`;g`7yDW}
zridw4wLIDbH0=d|$RZ`ey6evq<_uR?1W7NLMFZD4S}jS<7Uz|dN+_ByrCO(=V>7b&
zF76J=<@UYb^CR+OZ-JaM_iY|Qe6i&AwzmuXvw%0HO*BPw<Em1^h42!K+i>_KvZ%0L
zJ&C~Q?Dhn`{*H&%P;@E5N$}>N2$L7J+iji657Wb)B7+YsCJ+I{qgLXfn*KhEKD?0u
zxXozCw>7r4HW3Lpa1z)Tiyy_()SqfY>qBo9=OrbjiwEUnWwyhphGxw#9JK6Xlz(1l
zywAJ{g?iP|pugsU&IN#$bwkTVF{b#!^6n#KgeqOc6|=+<Q}nY!o_cg><!e5F?3m(6
zV_)bS&#NK<82?j2BAs=b2z2_@FZwSi+85WBGLicigRD<~Iu@+csql@*?z!uWnhGJ+
zhyJa#A<0p1<;t<X6L8<1G_;;aarxI#zRo87G`?nP-F8*ORO~LNJrEkP@b1nU;dFk?
zOZ+{@frs%)d|)>+RntW01rZj6;Y`OCbi@uP8*1V%!DGo#n2Z>g_=G@*UEKlEMtAa#
zBr?^;KSWB5F7<|uc}qyxOHEvy`gQ*Gw{i4eSoykHfI=8RR`#`!Us=h$9;vD05(<gN
zppAGIxYEU!rA~!DWC-ocPV3p&o0K&T2Gw)%7qz6d1aJMf_`$>u7Cdj2*Cy~exXoT@
ztwWwIV?V|wL`a@k+`&vfV0dQ05tgdj>uMvg9G+P6fZ|lcyU5WZ;`}V>5c6Gk+Jj+o
zv2x<h@-`>jC4TN@KZollp^ZFYc6cSc-$Ua6WA97g(md0B&2*Zj*~Zw!7#AirY7@7*
z?=X|38pWm6#088rYDC2)YTOsvY$k3b8gZ9eqt+r4RKT4vio_xkH*m!$3ULpj2&m-J
zo^xjA-kCdh&YX7U{>~}CpYZy9@AodxyME96JpVsMOA+8C*G@RAddz*)2GiBsStsU(
zxL3m!AOD90_639BUfr(ig%YQkcC~Jto5s1hBkdcva1Z?0!w5?BnsrlP4jW<L-Wgx~
zPT>?Q*W%Q+92kJZf2`f>2#<$wKWC2hJu+-sp2Hq@ShBKHSX8TXh*@jBMot+YV(=||
z7n4mh9gY@@`Jx;!*?kaYaHwg4%Hl@E%vJ*tIb%giZyaj27ZNXQAM=7BwFHSlmc=*m
zytR#|m^Z~#O8ng%j5%)3j>M5ff%}(u&-`NX-voU5AEp1(Z#^KJeb?FMQC9oPHsq~%
z$AGbXI>BSg>;vP^o|E~h@rvR4HfP+YT+DPKY6V{rDh{h1ySQQOJ|1;-HMYHM8j$zg
zY}NSw+rt9JZjbrfB=A!8+5pl~lDbGf#PEv_o8J`XP1DWC+UM$E=C0V6D}iM>4$xmi
zcKsF_qgxA&?U|Y=vwK^}A=J31jhyx*<LQdExIKKS$KG0KEym70IVJV9I%0}%+z%FK
zXMKJ>r2tiM2XV2#Q&She5S^eWGC6yH;6}7^VQH_bK0GGpd;jRd<qUM<kq9s#-Zgvr
z!NuOSY$$?m^POqEe*Jkj`wpuMj;7DY8o}{yxg#U<ma84%G7u8FpO+&-r$8ae#U++l
z3>4u3e_y_0v&OoUZgyjif#>Kso1Hc;;Yy#h8R#_|2X>Xn;ZF&EV~scH;DEkNEqyaN
zXw<r5s2}<f4jW3v2Fp2D)9d3(zxX#n+pN#7&n_}L&$csf^zMeBnqMqhW{AAnv2-7`
zSZXh}S2`V<XP>lfLO1Kss~>(^Wk2_z_S;u&m8m5iHFlw+QDb^(aEr_w(1{r(^~1B6
zZ&u=RMjJDcm}wg3fZJm6q_-eOz4B?5zb@*ew(*SH!(ozhqE$IUTXp%(7scEv^TeZN
zM~}9$D{n`@I-*CoVkz`kxvKjvMkghXM0(KP&n!t(EP!kIkFr>}ZzOmn6U(>5=Dtav
z#X_+uzaLKf&9GSM-gCtli^XLfPWkm~H8nAq36biU2ga&7Jn_Qq;1p~Fygq~u5%E3@
zhD^Mw33=2r5%9clD#)DaY3Oa|yVlvfEsGq9*n25={3zo^vzp(|;^ZvNAeFn*exQ4^
zTj42~m^tx^XK8{_@<tFNS3<Th30s#te+OYY5cN^19_b^qmRTTO56?!E<OEV`>9)X%
zM2f!bz!+l$3zHKvSOjP9TGTPygvGJ^hCd#8|EZt<)zFdb4y;>a@`T~*c;9yl2L>$O
zzb%iqFFEqY<kpnnn9Lwau&2FnEBM}DeflphzIPTG8neQWw^)mWyi<4|c;fvZ3bHR1
zc7wbHpL5wgy;LJ*KmXmUtRJ5|P&xii!2^1H%9boVoL(XK5_3slu01G_4CoxAZ32<=
z(Q=o&K*#Dk6JP#CPXE~pn(HLLQ)NnP&)5=R-0P!mn(eTmO~&_-to%)&>-~`PJ6J>s
zqx)KIcIO!PrVb<ICK>$PKjhMYH@6yXgf?XxjY!`q*qD>g(r%EbGE8ze9~Pml+2hi9
zqZn#P-t@9y)UdaKpkr&~vt;btC%h(ikhqKj0=yOD+NmkBq8bwNzCtN`r6`JI&+FK}
zzJq8DHRpVnH^`EtU+Rt|l(6|K=PyMJN86^_;VyWjI%@(kxmYvHZniBPZ8AzumZ3s9
zK~r!O%#lRQX>#l$VQ;<cj#M^=n(YoL+yf#30m=b=0B7k4c}LHy7%3o0%8}elUrz!W
zN@k#kv8aBgSuE%Vt<N;=O!cL(g|g)HQ99d3wWxyhM)MD~OGGyxL^L(*ay%G(z=e3j
zXV)Jzf^?G8>9D~yMoWfpH+Cm~8Gx0{x8<)uVs`4FW|-IYHkXk|p1jor^<7JTRPrNM
zgt@Yfvd!&Guioomdt$=yyhiu{Pk(iZt|x0v*ojbR>YciVcBBYB;JAgXSDQOymzL0N
z`udsL&Yl8uS~9nZ#?l^XVx>6uxS`CXZV%;5lECNN-}n!nLcQ?U24w_%@0f6S?9Pb3
zAud|Iay_CC(<{ptw4(M2>`mM|B$a&VY>U!k;bhgp7s08YT3+KGw0AR1Hid6peA*s4
zik5!*+F>^Jl})Ju?L`@!6=(2y*GNSyoZ-uKkAN(whE-9Xo)3ENbG<r0<y3n9y~mp_
zF})MKqH{woo{CkcS-3M_8Gikr?RS5F{P*}IhMNy(D?8n3>a91f(tYxiasBupwn^Mo
zW6KoYaib6j?R-Yx*i@hU{ptLzGAcs8H+xnrKM-*_*v%Iv`C6EKzqJO<<~JK$ihLwj
zi>>;(Wz27y%(JGslB0SfV<S@|V^p)gbYN%7ZTXC?tq=9oR#UZ}*lCq)nZKPIAN|4c
zzPB^f{zug6VQsyqaxN0^P!Oq3?th1z9q8_T?*%B~t!zF8bymPJ(QS&=O&YkG^MrV5
zX1-oLqw}OZX;aKjZOPKXU+j3s5k06Oa9qre)||A@%CD~;!!@9D!?<t3f(o|mTx%!U
zXJseJ*P(7N8HHbvX{7npYYf+7Ptqbr*-@8SMMgYSkSijyOEK@^R0ppxAd6mG{K{?A
z&^g}P;cV*ak;J&9US7yfIg=kq2IR`8RJF9#&L3O`1HA0`y#tmgbIBIn``+dj<hTFc
z1mN`h2kv`mCS+|qGmMU2XsT~|;{%0#Q={!(v4)+;(m(LJ#l`)qYV-`c<tW=Q#Lj9x
zQqy%=y=u<jJm18$J|ZG^_f+@cG$Yw3NtEsbpGnT-)i9H=V@<-e9_y-&<V9>wTF~)v
zi-2lXwmJFzs$%`qpL|LPk7;Yh)85RSQu8f1+dCzn;q1(3*4r`tbd^|Z8QMwaG4J20
zpBq5!oUJl{RPWMGEZl}t2}}M>L2_b<=Qo&@Y_+I?Cq0l;mAz`=IBx9G<Cp#rK+JZ7
z^n9MJmZo_)lbky-z<R1(oJ3qo(%LTYD!7K=%uJ8qlar&O<T@du%k>?HVEzu?pVt-{
zvS8j)PaSK~uZI>b#7C}=PtQ!z!&-zeIb0RKmDBHS<e_mi#f)|rreSL-;MR#=y&9g|
zBqpqZU>fyk^pPa#gt(Gwbb~5?;{;^&i&5qs)se|A_+*Hv2t!IR82c-|{kO|h-7b>{
zHtTX?c@2x&`?KZ4?xZab&Z757&9~8ktdR<>dc*vvgzrxpgmNyHS?R5ir(B4sd8*+b
zT#fcDu{Z8kebo5!zzZE3IY}JDtQkpmiQi#Av!I)i{W_`PjMyvmL{fMshD^>a40#Za
zfR8P=6-5b>eNrW6F;S{eRcKM&`nR$tQFg2rEB!;%9+4$}IQqO1_GNMaHEP3K&@;gj
zC_*B*IFb>Y8toOJZZubQ%;wV(gYM1YCkB>uf*PXaRg5hD<gJ#&8PS}jwg<)&bxktJ
zuzAVHqc2tpYddattU9NTY+Ejk-Uuil(gJJQ4OIldvoxag<}${II<aTG3Q<Sp2r^2f
z&8wDHIZ6n9?ZJuxfy!RVuHSre<VyX(X)4jlwHF<ZgL0V(0b^2%&8WpHP1v3mt|#x?
zVO~wia*psGv81)cRG1zBVA&4XxDfhxjM*Ajh?+RWcQt?bLG-^CBp}3!ahk1OmVQn=
zCzUlhtr!lB7?0EAXe0a_=EG7r8Jy*8Vpzu^ZBj1`+UxB;V-60vb;2>mN%&sQR4R2r
zuFl%Mnp(jrN>a>3ezzG`8ef}M>)sN}&*W}vT6it(m~97ai^^eOtx5w<r<+My8+D=F
ziA7CO%V8f7egf}H+yiaG%T`I<FQbW!uD42N6ec5cl0rc(0xkE#2pUuIb$b?Zx~D(V
zSMEZ^OeT3s%5E>mAK7xqh~+51&el}^=1uX!$Gv6wSx>M}`WJLPs<dRZ#vt9K4b@|P
z`Ni2*8&;)dN!zg%uJ?&n-;P5jAg|c)k@c&|?7Hi;r|W4I!kyG~LUe&3W#?1|B6ul)
z$cJXmH4(%tt|iX01B<P!MQNzTgcU`qn47>Exxu$=4=Ayk8EHEAhiO$+cz~0;?L0l#
z|09XWm|4Of>zmWj4nxZ+2RyaULs;$e<n_S6(#!warTzAt(VFp&Rpqxf^~k$?)jO>}
z#OfeFUzM&Uq$8E>5@dPN&8Gi5%-UZ|=*MgO9#e=`qL{XlZ3`Uo4(8CWI$q1KcjBK<
zN*e{*vLmq`dx){$2HXD_`MtrFcM7$*^fe8r<bL<7Ib?M^eESl3m#>r_AO(OntdK7o
z&JF)Q{C7Xb{;z%gy4@eS>1nc-S|AL)@_E~y3RyPJe5U}8Wzgiu-=h3?3fgkaoYXJ>
zYq|XGr-3$)EqAUgr0T%AIafyCw$(psnKO6}wZ{^?2Vc$#y|I0%Csz&ib?K&g7nLqT
zMs)jB4D=0ZKW}To*)QEcGqTr79#52&F{A7%vC%ub$8WqQ-`cQsFru!&ypk`){8}QX
za|nx30yq+l1ITTs!n``w<b3B^ixJhL9E?sap@M=-j|4}ae7>>_e-=nBImzu|ZDDG-
zbBN&GrHPM53!5%e)oefS!1#T>ydXmswZhWJB<Xi*ir4jvxT=xJvW18X<a<FlKeY9b
zn;$sWblEZBw9gE;_WGK4Cg?)qN}3<1cL&itrZlU%s<Q;FLPSj~vBNor&ywzYcUM(R
z8t-<3YT}PzoD)`w#hhtB+yJ}~g~2zb_-d>n&AFu84J(MZ;#vN;ob(M}b>YSy_*9zL
z15BK6du$4CUi|%DcreZTQ6HriRznicO(xXNnFP=wf#bwrIV41uaJcrIoFSt)L?<^f
zK+Exh<LN%rJi6^!+eSG<yxtJwpxjmMwEb_gRJ?H5B-UmVg!GUYu5=2A4bHsW!%A&b
zg4H=EqLg^HJ(Zh1JwGaq=KPUon2V24>~{*xEjisQ!0it|H1R%v*x`5bZhdn#mPP27
zlRpo<E%3cC(nKG7z4pc}b0N)_$~c60RfC_af744Usa-h#Ldm5W9zIW@&_*hu*Ivx)
z8@^dNy&JF#*{e)d9qhLy+i#Xw0A@NquZmd=i=6FKO|_{Gr<Eq^_eI=j_bBDX>MJR}
zh#J?&;Rw*-LJo_8PqPS>N@L&D=#PivzZYGA&x#KfFI1~u=FYd+A|30#pe6`wak;=o
zVlg3k$Ul}4Sbv2Ho{IANFp6>@N<^UgI#j=II%YR))Fgm#u=V3~cs32<XgWMn4vZ@A
zSlC<W*}_M+K&9+NXOevf3w%C45g8j5<>lwfooaLYR=ohdraoLYZ=(~$kHfWmAEQ>D
z$RrYr2T#f|@q81^-umM{2@P2e=&r58W}d3AK4LA7ntRsIwPYLvI6$-Kwmw;u&w5$j
zX3sswbMjGj%thjQvh$2KbV6c0PizR36Ouy`kGj~r{z2aV=jWI{PeXA|qbtNp!;;6b
z9F|<V^8H1lR<?zn7CN-Itmiz_6@3X{HhR83$|~6>lwtM&eiB00M`2HU5!%VPL!lSz
z7u9!KrQysaQO?~7tIjIsV!yqH+ua#vB_S+ri6fmf$`EA6gH2fkLS+jISkCZ|TKR(R
zSlrI>N}rQ3Y>U-8AK*1>@|MF?6A<VpkxGiW<!`qA>E1=6o3zlI&zQx2ja!~NWI3#w
z+RE0(j+b*{#(K8#Nggg1PF15ATnZjs9I54>i97B0``*u6rp<Q44}$fYo-t<=51jQ3
z;bn&rt+aBzZ{`g~65ARDmYdDU>(=%fqqv+VYDx2E5o^xo{aQE0@a#1rImK65t1&kS
zuGPFMBjc_-%u0$QRu5~~fV!(6P#adZfS$XltLDUn0gsu+8YN(g$1#y5iE|`s^Pxl>
zzYg*#%~%(%-z|8!eI4G`JSnAigf|}!smUsV{bsO&UT94kvi~G|FA1XMJAgb4KllPz
zEArJi@hn&3=KeI-<=k)}dl;1ixIg28O$QuWV!~z<^$3c#ptw%If|nj<?al};<wiA1
zm4fM^We9v9?1cjn;EBQDOl5}j+Dw8<2dQ<9$NE4bXr%T;7X}f4cEU3r1?FeoDIDun
zGXZ4vB;JeQy|rQ>w4xr2y&b=fGzzi!W?iX-=I7tw*_a}KHIE8F+TR~l=?vZ-Jr4#H
zgDm==7ONEizF*>U=C(H_`g((F-7OWwyD?s-Y3=eeJ7fPoKst+DD5Z`5j@MlR(;~p-
z9EcFN%m7A?948rhtu_JoO@P>6qiQ-zfu~`x%ml`yj-X4;UdT_N*-ET!jdC0AbYoO*
z8SJW7$(-&N^3&7sXElIwA~fS^?Ps}4Zji)t=#Yv_qIC~Dt2d{^OWL<NCBp<yj0rPn
zrQPBH!h=?rl_1fdPF^6&t%`3`p)=?D<>1sI4>4XjuUqQ<2VHd@&P&c0iNvo4$0r|R
zUZf7+JvsLTFz!<Q5rG%vuXKq;fxkydVZxMi7qH<vbrZ2uo1vWR>qRBu*kp-yO+i^o
z(A{U#fwjFdZ-*rUG=f>>g+9SthgX5VaeR&W4ijqFjrx>Et{pZO%CN#h|IM!S`ZqUc
zrP~|?U)0e_W8N5TLNMP<%{G_EtM$_w$a1{XxTJRqUzbgyNw%eBvDTuhYJEfmuYQ2v
zfONIf2jAR%ry$<~5o&|?OaM-duca=peTI&e%=lYC%YwEsEy>Dj5&A6$3iSmozH!HO
zoQ#=*2{_;t`2iNss`7J+KfQu>%Aq&;BODK{n1}k+?s?~Jt*;)G+a0piQ#pqL`gezO
z^2tJ@o6FZVwOQrtv$YAudoyJ#MaE2{z9I2yN;0ICJlCB4J#>=LGeb?eQeXLI#%-y9
zN)AsJrqqK@)o-8oO*ESA1HjU%vk}{2N=k}hiN^N5z2{N=V<@1K`JLszI7R(GC!Rn4
z|KHy??#HTXah2XFd_%y*tWZkaFJ{lYVNgT^NGY^F0>n>vr_h9#d?Gs5_;<p`{ZV-T
z$$;Yf9LoF03gaJ;BBLoR-CJafg#gG>8f<P7YMaV|pLh4}RZbRR+&|pH&2UCHnaIaD
z2PN|L;asN^;iRUD`Qmp9$6*@ft-GM%OLtidwH)bwOykicw@5k$ek`X7-ZQ&gIvq5H
z-?vgz$Y#+}2b7QEtirp4+9eRq!lJVKno>!qRM-<ONtxWJQH5|xzIgy}#1b<*m1!~;
znHz9w^qY?e#4A_%L_eku(t2BNEoekL!vi!*wc3TuOGZ7h7d)olK)|P8JROfPWH2w?
zD3{jV6pHmP@RO^Ye5$g+ivcQCdB%RIy=vGoV6_L#p(#3<lN+a&kBbh8hj>oycOF)>
z$Tpa)@GZR=yQn1h$s23kS`BXU^7H>F$V3FYWJ5k+L9f7P5<<NCd;JW8EIz_i@&x16
zEj%H2`wV*Op->}IS^<iCWzlKIoDMC`t_pA6KNtYeosNuSn^;$;4FGHBH(I;-VJpVw
za!l598!}9!p4_?rILIpCpqt-;hS;@H+6|sH1=JnqstFS2czn5FnwVo_nvj`DZKJ`*
zSoF<8%)|zBXb@=Y>@>LgW=$Yj#ZhXBRgG@bh}Go<785oo5jVOQ=>p}ODw+W3xN7~i
zg7hfNV<J9f;Q>KTE4zQDVd26~X;24tw!vzbXHVWPIpV6#S(B~F@uu3y<XVte1Fqly
zcJ}~)K#!TitGtMU<V#9?G5QoNElN7FC+l6~`JNd=$mrTmw6`r=L{hwo2h0NoWc4}<
zQ2R-GMZ!SLsIxZ~F@d~8EOK7=w8PB&CWp!i*ibG13c)J-fIG!%ov>=aIlIzSL#po~
z+<T)Qfab_W8@CsGeHI}r;=N0%en#0=jaNe*-F?0wW8W!+0J`0(Pn*X!8_-Hd{b-pp
z&bAI0tKu!2>53qO$y73YeoMOKCot=b(190E&=l8p?k|V-Pv?-|pc46b@am!$ei7O;
z2`?z35>_$2MuL&;2-hp9I<EkYdMhE9f4K*jJZSi<f#TaFlJ4#P*_stMGcLa}{!*@$
zn<5IE&TjZHKyPR|*f;jj#u*vWFCTKdK5Szgnq7mS`SzKLSJ6Y+j<|(xRFB1k;xeXQ
zylttk{$`U_vBDBZ$3|4PHM(%zAOa)HHf4D6a&$_)%@CW40eb)xyI1`S>)9uZx3U_w
zTH%$c9<{Fe?PC;!4KYvO@6r>hBlV6z3<X~ra6TjSc}O&#Ck+hPIFu%T_^7zL=+|Ch
z|F>~|pl>zlSY{{Gqk&`E;TF?ad1%&Y7{UHz=Oppts~g*oYa$C8+pBc&!^dIOSCZN*
z`22ZMLK<x>X$=Zr-d#88f>Tgw5rC&0uJO#|4kpe;Piz<&StQ`WX_!iEV^vj~cQvoP
zep-{65w<+X^cIT6^$>_Ab5WX#1b-PR)frN&8rdEPmTY-xC42B|6OrtC&Gxo(U>edZ
zD(c&-I1A=!+ENYbsTTk?r_7#qHp15~qw3DIZtV>mMY``)A2Hhllw<<TAP(DDu`PIW
z0D~S*Df7Suw#zIi?^8L$yK$tt{&xy5q}Kz!o&J1C_aYy>tn5cT{Lp3jek$yK^UOYo
zd}5{F_jM(P*lJ5C4O?;mxcdvyoTelyb0{jNvo5KC;a@1QltgS#B%X0jtF7ki7B>RA
zxZRC-mNlnzhiJWh02>KRo!E7Ka7ZYH59lW`F|#qNI2<v*L)At%=@dF{WT7$=fw0d%
z9Qt9%ixYkJ*&nz&I_00&O0V`ghxWOd&Nz5c!_Sjz9yX~G4QCd}G2+85=p}N44!Qqs
zihLjL7cR%p*(!kq1tC?Eui--E`m~dB1Uq8gr*as(gFis$$Ti=O<)qJF|KD2rFR%X(
zFd6-^(kBFu-YE=vGh0M3M(tVq_06D}9XM!DpKSpUwj(FJb_<ga`1}$1;xFa&(_9hb
zxY-H(JOc9UhUZt$HclyRGG*x;Gv8C=nS16Z=w9y>DupbS!*0Vj(LSMRd0V&8-Y)!u
zzh94)4UbF9`3uKzW0E!E%d|*o8-8Oq{b-9M&%|xZO`U)%5o!k^cHq4Jej3W4oDZmY
z$(7hX$SD^A@T<_`u{eS(p1b3h5|3y1B+7@K+v_)R%daJ~`IM=t>+mE*H|ttNP}b<p
zVWaUhiweCU`LWgU5L#k3gf+|@FuJE7Ja|tNVEOxxECJb+jJ1}`QwBC{gZ1)4|Bvzp
zRzu>l@)X4>Q`05j0!E2^gUAs|M}lUm%FEX#Sd0<zcP=&kd&3&n4+?ir4$^TDIdDvm
zC1P{mt`^sJm@#U_T`o3BIy;d1Sneju<uB+pMDG&N-l@Ttn7nkxjuT%FCSxNgI+;PM
z`&{mZ7%v2FlEQ%?XGW`Yj~r=1u$2hwcN;b|F=0^vgAzFAA!<(W{iXb-_5GA1`lphV
zX04;B1wyFHxy{)Yx26rfJePAfPN#edyf`LGig#x$Y(-D?K_HLsy{bJld-{cGyp^C&
z03So-w#k}BdkJ(UsM`TG2k)j&zwuAI-pW?#nLI+nFiJff>-G>vAg`+0MTeRszwz!-
z%gLRTgdH{~p1UYc%wE}1jGWE9EZGBI-{Q)&=Q_1}Q!=S3A++6)^i8@>pDnU_xj4dP
zP)BqkWq(U{?}YT+M*26f*agF<wiUmQT`8lJxAiF$k!*C_#&5HKa=jz3bis;2iKWCw
zwnvvheL7gO?^9*)O{|NCNjBGLx7-3X$Q7gkbR9hQSA!JA4vqV5W2LE}OOHRCZ2xJx
zuN-JpGct5|g(J!otFu8}!tAc=Kcwt8ITp7UZ+nT7S>BFxZ1FvF_V3K^C2XQ}$7-6f
zQ^KNjx!wE9@2&hnnAceNo{u<BZdOt&Bd+|$RP)Ew_Pr4yM_Aq4$Mtt#%)W^N4f~nw
zR8+}g9(5D;j$<ThlzE>MAzzce{yVe&ThV^yX9SY!s%!)465_d!r)61_auN=CH>_&q
zab@wS9jA8+dB?w4{xd(Zzm?kmC-;6)EdSFK%RhCC|Btb^P*6COAU6$oQwN{gG31Rr
zu8fzOMD8_WWa#A*l!OCX)s;Eh-L3l(uJ@-F^3RsuFI%jC<`(NGZN^{JO8-2y(ofoO
z3JQVYtB}+Z&LHx=R=yY*)vWuqJcGlN$;W2u`48g-F>KVLh3?(Uzy15NRXM!lfq~n!
zP-2qU;`47G>vDcbZE5Ey1q(V6`I>eO#Ok=S|K1lrre>$XlopSr%`Q+1eeI6vND;lr
z+oqI(D9bodh47t%X2EXeM=qwPM*hC6|8^{Jug*J#PXmDbaisaP`>K?|669*>JB3ZM
z?-X)CAlHV#Hg2q_!~dF%#otc&UwQYxyQBKQ$+jTNO_tG{%50E@JAPn+tduMICavBn
zEJn1*kub}+vLVWR_n0;F?^lBVE1T@!^Zu9OtZWrwoyg)8V)k(dN=Gkj;`1%quuew{
z5o?dBk33(gSKczagnF7%!3TUSu}M;?#dTB>@|g54GQLEIWIH*YX=W(*h(npI)`?L@
zytMKgwBp7TpAeYFGPfguIm;U@M1i_^*N^Y6Og-5X`o2@J_DwUqg5eQHDT&`BGyLTM
z72mGERDGv_KOhDpr$9Cq=a4CE&_DqollkY7FpuTHx_-ysqTeYrr^Q8FkxA;2Wd<N+
z3?l%WCVcu%;X^rO(Ie1rBW3M?S^akkxBj;bLB%xBJ^p#w`OJ3;C(?d5vzGvR0hP^4
z##=OZcrCKq&AWa`vmD&cfXA+fvh3XMR}|@6<UR6Ux3DqVVVe=NyC^3&lp)li+od()
zYQtSe$$QAH*K#oB$enb4&^&6rm@gUP3KDl&*5q|7H-R%uoVT(6ox*aL?8MCW@c~93
zV8)CotfI{cC3UfpoM_BkAw{4we^mbBe|J}lvIZzC$pv*HFnlnI$`^Vs&9pCos7A;+
zw_il_i)eo7HNOm~U*?)$rr}>!s9zTEU-mV>Z21&=e%THGQ(qLglv5wV$%F=S?P9jm
zDfA{hiy$muJ7wj%ESB(u1*3W?l?%r$6H{Jt2}uiMOmt^zi^<K7ls%ov<cUEZuv4dJ
z?^@YM2&c<3BZyp8t9JQvJNfXNpWxLKFduN}#_8Tq?wvi@Ym}8z@q=r3!0El<ivT|m
z1mOEnIQNmQM0(kQ&6L9pg4~`A8~6)OyrJ#$#0Svru(Nhhqj1xW>xkRrEKeF~f}w*Q
z6<9yOLaiDn!At0nbHyZWN1a-|kwu2-Mz1#vh<e*6cz+3Ko2+gzxVA%Iy!nu%;l78A
z$((JVQsfkDF{i~copU|T&a4tN5n+V5?CBqrw~##JGOHgNom%jaMfcfVO_-u$LOVgJ
zSez9}^KGHfvT`d8sD5IiqhjW;OY~Z@Z@fuF^0RZodu#~)*oH5<yd_{k<y>vk@<h*Q
z8q)G2?o8~*ZU>W|;Y^>PgI}frRv6;8g-?3J>odD<0oRpIx8$sk!YxPv{g9D4D<I_j
z<}M=HOAEdjDnE?;n1Q)&XW11}4}?ZW(uSvT)M!N*sLRG$DoaUge)MQh$uNr`deUxU
zuohGS3*?2}{Mg}pW%3mGNo}|(XaTHkq4T(XBVD%jOndKJ`{VHeD@1X9cZPAmT&wvs
zPiNq08XxOR*j2KD6pv{mGfio4Tc2)i4^K`Y!gHV>rpb1L9P8dGBx^X!IX@meU$M|g
zSP(;_?VLQvWFyN`>Xu`L9xEliX2amFYgPc@e1Mlx<C6DYv@vbDH`J%NK@O_^ED#E0
zh|*H7$0p4fFe`U4`grf6tUY+ivU(0%Af}oolq29zg@E+TH}2tksFt-iQtINsE;UNz
zd}g~yu*w;o8GN|sVHFn}^;)oNgicAr9QrLc!hg;+=tXzH+_BHN10H{{|B$Q3)`&=2
z<7XCd=%@Si!K?9FPe@g^ukOuMjnwMB&1wXz*tz%qCbJA%xed&u*ur9$>`K0kHZp(s
zYGq1Urz@P<xENAT(ZOQTTPP#RC=NS^Fam08Bi2CqbCNJn!kk211AJguw5r~&njlA)
z=8261>)hTt_q3b^x4r%ehQ|LhiGQKO?EhJ|e^vy5QOjB@Qax0Hn?X**uIjk3Rzl}R
zN3n?hfVXEv$^d)g7ganjqx@##a{?J1GGD-%@~Y7bb<wg4P~oF}IwB5##kX-VtoTE0
zCm@gCGT`N3TpD@2`_0ORZ@G-IyK^fvB!tT#5YYbGsfC0Ie<HC)=pSS~nVi&mW}Iw6
zx)mCV7>b6EkJFt`e=-D8mAeCtR_SgDcq|nDA<E=JXFOsee6H+Qosss74uCxsa~sDs
zy_*ZNOI3MnF=>K#rPnQ!wFOQp++)U(?!78@j4<{}QZEg4)Mu2AmmFFg!J#R?2lmOh
z<3?}d0Df&cE2^0c-(;nsB*vDN!F0jVcIf;@X)k2ut?zyxe5R|NuU)1cB7|CQ8mAjN
zWD_Tc9`duWA8%^SDOk9k!KhWM-VE&SWeskYEH8V8kXyVl&!AdOwLH5~Ni(l8v{?l2
z!MCfol}jf}ZH{*E%`zq5U98bqW{+&9S>A6SPRoykIt(w~xTOMnP;>fKWpa5V{idd?
ziG9aj5)pB&wgBV=JejJn>DektI#8XOd<o6Y*hx0j!AEpW=$v!lFVuh-Gs#azBqUT(
z-U0Ozu)DtE=`F#~(<)8Ojji{Z<4~!IIq@;(L|)39*X<#16827(*?D)T8<K!eQby=`
zG>k=2QPG>$iyyyI*tgk=<~IkcEUVXCo;1I%3#X3ck}Lv78e3c<?NB&5+#-Zzm0~tp
zesLfXp6m0h_TpTrT)+W`+Gh{8m&5PO+UTegdg2UJAMwRyt&<uv%8V#!N{xDBj{hhz
z&XoLo1rb!o3exv2#3`LBEEE{8gs!dz^Q#BY1HI(~Rzg&Om$^%dm@g1IH{+c1iwaS=
za3yJ4En~^6$%QjOxjy%Nssl|GVRFJXojA)%FJ<z-r+VkI2$=btYu(;1`7JLQryWL-
z9scnxHR!~}i%y?!U;CK5%tAJ5&hke}RZFII+#-iG2Fnc#j`6*=*S4}sPCXL!uDfx@
zvTY#*shL#7_sd}TbCzkHWI2-?J7S@0=9Gqq0kE$Ec9JAlBAzDCVQLnly=$$)^A`Ct
zEtxU>&a{V8XDIm_ph214`lxD{vU>RFrU6WxFiM#PI+p<MOk?e{Pmi=Uk6;JeTVAXu
zOaL>s0=%FUIfB~Q(EYXe_wS;K)=@7gJ)o{g+{`fx#!FF-GQfWkWHFAPNqQExq4xeq
z+oIRBes1Rx=w5QDY~{8nx^=9namDXe)KvOeQ{p$D-DY03#>Zv6el<ZV&|MX#X+ADZ
zh}Y>!Boc!R!=TMvYRy2$D!K`kmSUOGqxJrYf)9WG3FTj^#s2xL_xI&+EHi$x2k_;t
zT>`r1T+A^@SPrcSy~fOL#=jqfP@l{-K@T~eK{sPVhwXji4yc+XC$(vhBi-ifP7;$$
zckzu@M}W$@XMyU#GadF~^jhrFo~El=Gee~t(<G8zIu-mni2^{gGI&$;oR8vK)W4;n
zOTCWcBRiv2qUOnEsp_e#drlUu+d0aWQ8AS77fOpZc4}#Y=5Uxi23C>RybN+qh>|=S
z8*fjkGb=<f=_%gPaO@042}l>;Uq5N$iz6`c0@4L3F7^1mEd-q&0UFWY<>yW;&9Opr
zuTF?saJ*vZUbmfA*nOw`wP63?FP=9}&W6W9#8KkUH*01q@xzyDqGIQFTzMTvB8{E?
zFS<$^ViIjVqi1O)lc8^SO#>g6r77hQ&9D&vzP7x_Gg)opU(dvC9ey5P*2n1Mm!3I|
zH>v7D_-zf2;#s-~&mbK*d=i8q7VDhrpJ=W;28cL~I#>MJ*FQAKdvsc0v}M+Pw3kw9
za_08|2CM6WalKeZvhHr|UX_o5eq<P4zosfGBtG2FDHgXDe6_~&J|y~Tjcp!X??zW6
z7Tbm+)U@F7<XUpN0f|*Vpq98rA`0EdHeQc)p~=*eW4@jw+vWz(kv=elB%jUz2^+AJ
ziv=gwQQ9`1;*Ef&q0NpUeelp$Au$qq*(KcP(W8fjh55wx^Qjeey#3qPmf1*?e3Rm7
zD%kl{<v4bX>0yIsSX55}0{WllI#!l-+XI(45%yJwViI1`^*IY*tH`Lw<589Y;+`E7
zvCQFs(GsjXUA~=vO7opcw^+1q_iRUMDe?^lX!b=-4(c5MOs+MRcia3M`0)d2AOn3r
z-DnpnV!1rbg7vzM6vDuV)T{e-C+@W!m_5BnzwA&t2493xXjx-?A&F}io)>VkpD?bL
z1_K13duDtVjC{a0VL29BIV}~*sF&RJV8<AEYCvMR9W=SkrY)&ydD8b)t$TKo**UCJ
z;(WQ1%U$)d@G||fY&%AZw4}IbV>vu@Wv=fo917nL_GjV=clARGj-^<Iq$EZK;-Ulk
zpzd2ZtkDUIhSSX|<GBvGBa_oSEl(Yr6j-%<s%kN*pEW`Zk^?3L$w8)-xB{W?@t^-N
z?_YTC{rAl)KYy&n0}l@*Ct$jl@siHa`C4v4zwg!BFS$UQhSr+M()n=S9E;csI&jlH
zL^lQVu&Kr=EQk4gl#GwO)BPC^;-oBrX1QwWAUjn%HIfT@YHF?3kAKHrgLq^8z(i4)
zIJ^M0te&h9mLN(zvXvRIGq8s6GfOBpZZTSzw`g$B%Y(U{;;rET(>oZIL9z1`<t)fT
zzxr0xglUtPxon}<y!wo@MC6Z06*ZP4%(^qQ{mJw&?}i}MM8uF#Cn~Wrl<yifp&JHG
zqsN${(NB5JRrt^|TW3a>Zfs72PC_~oZ?*QGC`(-Q@foglpPTD|*44Z7&lbO&Z}2L)
zCcU(H++h>+%`F)w3E30Cwe-{d&^3jX<uxzq1Zx=$$CvRfbWv*S8<Qd$l+O%hpVEu%
zuBpw0UV?gg?FZTD1TM6}^c(wys;8JlLP$oyqrxui9etBb4==TefMcDfo2ciNOhQTm
z=?I>j-?eq6W&Pm4H|eRHZ%qVGD9b@qe5`6?vK@8ZFRrfY2uvk1v8`{};V#J;*VBX(
zSRJN+muEkd5mP|Cd4EKw{8X{IdsB?@#mV;*%djCCpeE(lG0;(<N$yt7P@2xE;~ti&
z&LydMeDD|bmRs*-4pR{e)1zG(1Qi<;;7*=pHL-O1uIrOZlY+-xv|-b`oAINFtzEG-
z<;FO%S-<oNDrbrjQZy?vhW7Y(;7nGzF@8v!eyZt)GY=J^$}8sZd@4+}isfHFB4BIE
zQDE#f<N(;mi$9SAS=s|kZ@g0&XXp6)2We|->-6;OX={T{7w*a7`IOA|e)i7%FE#!9
z*=zUrWpLoxc%1XGvU<K~6>E6oOO{!ffk(iYum$``->A^{*oodwup#>8^tG(>liu7q
z{Ro4#<&ms&l^3&}^6Y<uHrPH%-LU1Kr`>7?qG?+6=pAV(OqmukM@sFGIUwa;J?L`M
znah&~y#p-Ygj?J?7kutkUFywl0^zR5{#0CPQx_hWshyfNu78r2e(VfX+qXp@<TR{d
zJt1lmKdf_Z{7B#`cyBW0;*Q<(4SjIivcYeMb?1p12Z!tJ7C!K{JCVIE!*o8+;$%Q0
zwmc^uAmPmz&eZ2wQN=4{uhWc3<qx1=?EZGw{MO7-jl8H1|H4?24cP|eRlKECN<G<^
zD(g-~E=F|rJGAavz^Z*q<CStc`+YRmQ_7xmCl})O#G6}y_VD<6Dxt%?!x9G$9GxZV
z^E8i@!X$f({GRX#1_I-sM#7gS*s*AL0T0<u`G1mor(t8qoXg{t$P+VL#|uRt&$ncL
z_4ZUjN^HsZzV>0&J1<0eP=DOY8sb`c$cD<ANh=QZ;`$WhOzb)_rJM*ctzS1vT*-Hd
z2M)vTXGel}%E`FvfUF(7!<JGmBfarWsK20wZNmwsw-f*`xD&2N*bwKc2V`oWpQ6b@
zccVu#XS>L>&>?uz$(eO#RKe0zJ+kQgx6QiYN5=Jcg4Pv%>aB-{L2V2gBx;seL&f9r
z{8IeH!ub+?!qvyW`%^uBp3uM0{_p3h(0@ABli8wzWW3iy>DITg#~ta+khZuBO)<N7
zRjq}ew^85rVxQ#6_IZZJc;GL(yi<rGk$Z>BONFmGx|T{Fd?b3xk{rAk_IxiNc!%98
zccDreWrlcJ>|4CAnX?>)QG@NDNm|&#$!ck*G6VV|TDlhK#Lx*s9_%sC(q&3HtC-_&
zyZbxQIM>q74_}K$v<xtr9FZL9NMJp#Db(juzp)@K`$id=D0kzC8G_Yz$v2rdZkzqF
zlKaN|z7?6oEW!Mn^PWr9LtinY9?WJlJ!u{-<t#J}&oE`FdRC=E+^nj4SBpnkUVfCU
zvntO>FGsN8VMv)(6)BA%g>@w{7Ev*qbb>o3uPySB9IAJmbcRP!?z?PHJKn!MG*e>Z
zYoVL&mdAQjn0a#)XUIdhyS1?rS33#JsFby*{Xtw3fma$8QU~wa$;LtZkpjE2<eOc{
zQQthHt%qpr6toYLZX89rPq*Ih?mUp!XZ?WE7TRKEXWT1%zhQ1QdTeod2%RX?SOX(!
z&vPc+EIJ$7a^=ACH*xmQ8Wxu)Tk#fyZEpUH9eM5(wsHmvUW>tdg(KvmEVcBL<N|PL
z^ST+H(QS2WLTut`SM+wGh|tgN2#R*;-8j=u22}Me24+o^KBzy+2z#=JR?mv1=LFd@
zNrTYKetE!<gkmPEVQ19(FUTAJDCVCi@XKIRc_i)<LHJFav#IIRi+(P#--VQ1XC4)7
z=8nqQhMHu$%`Ky~egXWfAUVzU%H;c`u2x6A!7VWKChj2JRpaXXq|R&9P@Ui7UDG5u
zd;QKQLbdgy>Fpw@i*^y7v3<n5CHeI@{^6ssx#p?%$oE4~&xSLfNhqjT0(S}X1vzlt
zc5NOKj;kd`=-Asz>Ouu1QWFfYb2Mv<Nmi7KbQS@3jE6cRfA_}*-<LKMglgEbihNe0
z^L|jIPRp2-+BtP2YU66{8R89^S^>~4YNQ71nb2D|{Mq=o8>h56PgAe1MkUnU>`<9?
zbqpS4E3JBNR+kyF2MkB_)lO1-lN#AZ>njYzhtEVgwbxyhig4$>MVDb2eJpY`#};bN
z<8gfROD4vRU6ARVE?dlJW3vsMZ~~i*%XnhK3%K;Sv;@)_ltw8bsE%O)_oL=$-WbI>
zy`bI`yI=mX+&@p?e|!f%@V*H76h3>56!T8Oc(B$fpMCNcMJM`X)@T(vQTyf689#(g
zPURY3THH7V^PYTF<G{}j&DW|rXm9dDu^E`V^})8B)@sk<4ZCW?e@%V8b_@6LS-*3r
z(046eort`h8seL@qY@3%5q*fmn<&C+vd3D^&&qujE$Ekf<RCVQ>Pv|zBP+95lOUP~
z$a!M3?s8fY<_J2%K=(QuA0O+7V~jFF$;!$n`Ng=TEv%avWOVsjtx9z{4)UZpDgT0t
z9B-%C#O<{F05aZ$np0@R?M_#hBiH2=5&~$DVGX^6<qxZ^-8(;Os`|n2)>GFzqbK!r
zpBzLT!uo-#03r9y4G+PXn3%|EhJ+?2(lkn@AR6?i?zS(G5~NgMu(>9Pw)k)quX3xs
zudAgscj5)o`NFDie0g`+X=ouGP>k<J6ijvi+zRTwr(Yxn-OIg-PVq^t+cP|F@%Ehp
z%O=~XXMGd-S%P#RG7<w<92c*ySs*`jcH^b2dBE^9{me%a;P@jV)(F|DE&+#btILeT
zcieQld6-CyM%--^f93-YursxFdkO_W@g>4JZEYzoR_V&0O8T>e{uzo@xvqPqSe$vA
z+`TO)=iR;~nc4T^RB6qZ%Z%!;pf+Dl$K<XQoQZQN!iF{BDyo-l)I`JnwJXh5-_g$M
zKO<H=t~nfn?afbcKU|(_eEbe<P0U^2e6NIpRR9(p=rY_+iQ1O#W6X?x32?sqx?rc%
za%WG0x4!g;<JG@^3*MViOO<^Y6TkJvf#KhZwPfo#z;_D2vucw*e{b<|%87Rh@iYMP
zYqjY;?|qvufrpT7TomQiu?Sh6f;15DYs$2MHuL6y1=3UDm#lvg_%Ge;m%;VRl=9zo
zA_XN!2J^dsYjfE|gIz&;zAp2FXqT9(82$YAMwhKh$eJ+wTgYVJPAq~|<Lb6n^HiO`
zOnykY)P{JRqm1ilTi!MbyLC^y<g?*K8?TP(g&GyRd0oacub0s9hDm#GmxI?`9^7w-
z*FZ`y^j4i|c&X0(FRbXF7G&p8y}aHM*N!=5C*t@mgJvCZp}8(^vPQ_q9EXV3o0_Ln
z6C;9ymY{Bi+tJsuo=B=urS0MmxaaA+*QHw6lvuj&^~|U9b0n0J@~k!4uAke--5!e$
z_R@#%sJiQH>eH|-eCTk_aIQ|zBp7x80|b`%KB#V@$(n+S#e2Y%R3Ef?d%&Ojv4N3a
zxzgfs;_qTF&P~(>=IIvAhevN*k-R9QRwt*o98fuP?d(DN(x$H$xur@d0Y_cDCqM^-
z7fxRn-H%F0gxy{D7Tm09ok+5SJ?#n>tkX(z*3}UahCYQD#ju2pxxP)IWkaBxokZfh
zZZDjKjg&WrG9UJpAm8V1LR!SQl{6hOE|dcr_iU7&@g{l){)Zm?zaZ$p`RG4m_w+Xl
z@Q>UCB_m2obW4)yHW{J<bpE36#5D9;?M2=M@4jQ`t7bCZDdfT+-o7q(q{_2&Xr#v7
z2c8@3a?pNX#PR(zp~!~opH2|dK3uOKZHa?_$IkQqB7D1j)O)3L^@XQw!|9ZyJ&hI*
zKubK_8fr{$6pe`bAAsSFwBZUEcwq2>!F4(JZd8;^4&4w<%KkPQt{Q&&`Xp+5HRd=u
zq$I5+)j{X0@z~P6>?7r>W{)W3;!q1utHXzjEzGw9QB18X7ej8AD}fcHKfqG_W1q|E
zIRn;VHTMDGt5s+dd$ek%u@U+=SpkP`?JoDKgw$-LW1gNv_nI&OxI;Iy0)&@%Meh{g
z)%~F<{!&A9knZ}->I4VA$cs<4TWyU3&_b9eQEU@MfY1HsTLpmr&EZCtCNs-7!VS<D
z;~VFLS^reqr^dWCK>TOLb&9_eWIzg_8rF!OYMM3<a!6s%0(3;5SKj5iOq#{$rBO!g
z9{}qb_aeW|nt?8VbDT!v*csk)gT@3tES_JUa($LeIoL;;5Fb_Z*lW5u{*1l%^deXK
zX-2dZHHw7+17ZSnhPQ5Tw@u{K?AgyGvi-#_pZgs8vl9F)!GFR@=g&~?R|m$Jg7eJz
z`WA4va-Oh2VjHA;sE8lY<U79(Z_ax!B>V|&GMcZ<osIHLe8XK^ix$fqSpv_iixU&q
z!F}VZk1D-1Fa8cT^$(X#U0<Na{cZ~Xs{D0u96Ap+->9CjU{j}+Z)LXt?#P6VY(TYJ
z!d<UiG|G08OxXj=Ea?XsdP1=2^Q>2Qc2W#QDXA87hJ!#VMovQ%EG-%_id=yVLSA#N
zG~iyD3&s=h9krCSwb50j&ElS({>!xnI~BGVr&a4~bu-$2hjZnJEr})kE?i!K9|2Gw
z<4;rE)vJSkWz_s<HPWdW@8murqm=Bt;M<B=w{&qP+UkQO!o8r}dlKhQF(-%1gJ}ca
zE~ZmOXKv2|&R42bnm!ihnt*dKL_m`RO#C^|5>@meC4Q~L0_R!oub<nRJl=MHSw3(c
z=F}>Mpo}y=(oTVV6zTAF3X-#Tku@fg<QQhoZ?}u3o{<@)VQzBY>4oEp?-X`}BUrK4
ze^#`gCHT{0M9ga*SL5&!Uwm>estl1OU!D`vVqKNeOF<Fx!%ex8I6^Uf<=@c2fBFg?
zN_~5H;Rs$)+?$ZPQdoMw*SbFWx%Ij-hj}sb<Lh%RxilTBie74>=D|Ksj&wM>_F^I&
z4SZCc+hN%U$-a=gpZUc<T&P{dtIF7mMBL~HUZ>cmp`XcD7+0qz#2`+NMLmdfN^sF~
zx)#DP%|BP}^Q_#l8Xat}xPju>7JV{LD>=EIX`sv2)@xo5g=siJ<hC%Qa$6YiWeDlf
zBiuLIjY`K$JW2mfR74E=$bZmi&i@DQaDd|@sdK7U9wo{Wpu;j4j+xmvzxv&zLslXv
zlx?K3!(KOxb&NKPZ0heJ)Jg+``LU|~+MyFk<=3>ylq$0;7ZYgAXuzet%OhLxA%JVD
z{RrKeFlk+Iyl-vCu5&ugKQ;tE@P<iBrI_UKo=2Zi@h=ZiiN&A9C>{xpt5?>!zLLT|
zl-mgFo5~qD;+G<eBlnjq-9&+45#i7qbQlO0&ZXn9-~QPuwr|3<Vq3Gk)ryYxrL$M-
z%PEAun;Qm5#k|Hm!|x(msR%lFc^Vr2Vx-ndS)57FR1JNO4T(YZ2h^W;r@UB*DmrsD
zVsQch**m()pL?&0|AzDIt!Tcvg}J?E5$|Ew)RJ=;U43LiY?Y>(Mr>M+cughb<~fg%
zZpTIWNhhVrN$gJ5tEo_;@H*&ejegfk?@<)IkC~~ZX+Yw&a|$|1EY*aPxsF;NAy^Ir
zJgD4VpgyOs%zeui)GVhX^#l+Ne`3uCaUQcbbGIMGIpfX~+OioRd{(}EmTIHbPii7%
zC^_nFUo$KD&NKOe!PA1M91Goy=a?^A*oJeJuP==P(vR$vWjbi2Nfu#@62vx8*L7=R
z)+mrdK1Pn3Be;f@+CwHryIs9bHoh&1C!Nm5*)fU*&DHoC8Sa+dMB!$mi)9XSa&9-M
z_lhJr?x7GNXgek9?RUN`w>>^7hm}l}BpSKL^!I=L{m%@?s_;{r2<S4yF5MNjN6mFb
ze!p!FEJHlsOXVJIJf2^3?o~xp@vD%)X8etw(OT1n7hz}vx(920EiiBNT;YmHBbceO
z<mOs<?c;xV`2MX2_gSP3zz*;eCwsk~jg@|i>KTioot=q!FqmkY=YmM@x+QBVm=1<r
zw|5LG#!y$9r<32i0jf5p85Q;Rxrg57m)Ctoc=NY1_}llM%@FquFLwmwCy-lPd##OF
z5)0*<Jt>tCPWp&X^!}=V`1of*hTpx=h#8r<ota#B5@7kFti3e(qGRO;1tV1@cE~_H
zt~oIRGr4zQffTS&<1xvcw$8CF0@(!Mgf^!?iIG}w3g?>Utah6d)yi!r_1g@<my=`H
zjTC_&B@u}&1{P#Q0H$rja*?`W0@*WKN4T?w>4|$LjWy64s6p>nC=m3;w_&}NpZ!@y
zHtN<5jmYRCwb*IKgNrYC_gAyG@{>s9xcE3?ARZAP5_>UGchYy#D_2f%Fjx#}3Z?5i
z#W>nN&P)d$QS~-i4*!DMh-=K$KAN2(UM!KDztx5%-hxweNz{T<zMiFtM)R5On@vN4
z(1{y6;2HJhJars+XfVkST|8gy)DMx*mqupt$r2&wb}N+AJ8osVC)`Z7S1nMH!Nt%{
zvP%7@ug<3!=pGoe(G4B<a~KI<*qnS-@I-(fV#dBK1c#=rh3;r&TWwE<-mcf3fks*s
zT&)Oscs%2sLgV4PtH`HqGu_qt8ia8}A=zTjY>e&##4{G;90hr!qK$t4W!neP<Vt`t
zA$YkZJmo5z&VK9P#Lx9BmkP+V-5i}(sDbQha?~vUyiXqm2YaE3R7ct*9}VakplYrY
zGbSRh(>Znpw%qzFuM*?IY>9H_wt9TppgHz~M7#K~(P&apsRm@=Sx~N-eq`Nu*DB4N
zP6pOS#)sPmwluU=L59|{4}W(uxbIFqb94s)<|i5uCKG^e^>GFO|A|t}(f$zLt9er)
zyJsf`vEp8k!O8P&j%Mj+0#&)pI})S7s#U!vi_I1nXVu*}*3oJGx~M!D9I6#POLyEe
z>(CD9cb3~}3PR;e6*b)F;ST)H>X(1k_CI~_r*$;Z&0PINe3ftOC<Qr$&UW~xW;RWG
zz6?rKePndX7_8R2S95bUEI68SGN1xrITa|<a|Il23Kh9K+HvI8@Xw1^OGO)DUitni
znp2a$K2yFv94MREM#OtgEU|b`Hx2abKXx;_Alm-$0$m?s_GT0v8XIyI?X$?#8*<ar
zzIQ`m_jf<Nh!hm|7x}%U-IMi=t{t1Qkuc9>547~FV_1TTnGaqIJq2V&XxXh|u6hHb
z!FOT=UDj(YEQ-^$H4o(std%0b)r-<JzX_Z*^wWj9is;w>Z~^zR4Cc2YKV6H5CW7lz
zzB-=Xv$6kI$tkXJWrq-)C7U>7qHT2rnLD2~!Yk+djp^$dE=|xOSw-Lv;+vDd!@U2b
zH8+{?CLY^y_Z8)8YkKWoq-c78*ym-y2ZZIJ4w$@Hb(f~h941DFf79Fov@jnanz$H4
zJ^aTkp0n4_jNu^h#4QGJ{14H-$JR5roGvpeW^#>G&$7xYZo7&yD<U1)s^SrXv!2R%
zc6N**I(oA3S)3d@#h3ijSLb{0(88lTS)#1++J=CdErz{C;ZkR@)sfi5MR)OvLC`>Q
zSnpvS*DKJ4=WLdh92%P3T4fsd_Ki9pCoLM1veTASN9U9Hh&eN>9K%pMLP8hO&B}~a
z3x&jPZ^rb7Yg}??mli|+7kh6W)nuBjk6Nd?w8{a>LO_tA6@@}3k$FhyRtpN5R5T1>
zC_q365CSq!UFCo>DoB_~QwR}52vZWKfD9o<2$3m)gh3#H5E4j22!Vh%R`>dyI`{t8
zy63J_=ia-{YX1267AyI9-~CSC-uro;&Dc|J^KXls)lf+$$q{;@N~i5EWJGd#d(zJu
zii<H_`O5}5V<$I=@Xzb=T2hoh;j)_`Tt*ipYy7)``Qpz2S>DBUQThOzKX)Sh=DT|d
zkX<t?>WOe{J>NQ}%@yu>ALTVZK=vt5i3^0&3M^xjA_yRop~E}}-6XdZIXdmft%CO3
z_2OT4f7dC?IuP^?ApA$^UHXLC_)x!zb%|<Ccyc@n8x~6<^IYNExQH|PH3J0)L0;_>
zW3y#Rlc?0fwIhBXnQ2x}r?N^cpx_9Yl@a7)IB8UKPG3{&FbQ`~8~aNL+!}-zxBfbb
zPR*6|mG5edPlw#UchBP0y|R%Ul0yQ1(|Iv1#3gdzN>-jE))mwaSJ>|0*4koFW97X!
zk#T`nl00x>n)sNc@L*t@jWnTkB(+Xr6on^GZl&D?C_+LbBP%L5OQz7h;S`Qd-1ySw
z&=Q57P&%qXRqf83jJ@n;3<FDQbd{oMko(YE?Hh^ojMbpMLOsV+PW_osVRl`rhJX+8
z0bplnQg$IP*-)`MD8Dh?hmO{C8UaMU`JL7O@Y_Gu2|FbJ4G+H!HlHJaZ7B@>=SwZz
zKXGJ;G0&(`&Y2CPoutS)i(vh#TilHRl&@^?v-ZWD>x14-0I#kz^(SxH;2hgJgKb?O
zEsLa*X@jdk8G3wZ9QCC|7~HHy6|@Z0B`sng?y2ET$~jk!h+@m}P|9_~H~bpBW<ZW$
z)MXX^5^*66AKhEvX*8aPAmp8t$@#+#cIc4RRk~V!tN`EE9xu}v(_Ad>-G-FJ;?0s1
zNP7xNzI{~CCAepY+SHW+Wn5$MT8T$+_hj?w4rN%0`<*tayS}Vgw;8@|SJ#RNnb=cZ
z$aHm{!#SR(Db1MGCeG{w5Vn26-&6G{*cEe*>E>g_L`()oH$O!-CWN1;w?t8nUcqJe
zo$%T4E;G34TyAN+I#T6){dTRrp5^V@K|Ra+t7Gu*$j)VXy&iY->;oS*+3TLwZR(X4
zkN*10|B8mBTL3AkH<sPXIH6D8lxeeFX;$%gSPQM78@Aj<rckV5xp`Rn+xsE*x<H%M
z@#Sm?`pUflaNY7{>_M$DG@|nL`v!9N)FSVB`g<-Z%>iJyWRN*pG1u+Y8K;F!z;$2a
zoH25=d2!ii>;TCw8~Wjr0Hzkn+q+L+K6lAa(wHCAze`yy?*P4sm67`sxB=E0)F$MS
zGWolw7sC0Flj?{Lln{s6_GqLCk#o?FAHG+DL17Rd0(wH3MmL>!4W0(mkpN|y@FUg4
zavXA__B{4`TJ<d{G4A@yChB=UMR#=2xbZ>2bb{fsR8Csu#bBns<r1+_6_r=4ok*<B
zSG!KB%`1FKL`f+Q6GcVEPoG|afF`i0P)fK*2?O^u6EuO4Q+T@j87-~FmfJ$0{sRg^
zTGacmuft1ipEmiK>1hQFhgwBopW_DEOjCS8S8woU0@!|ksoqWwK0wEHk!Hc;4>hZ$
z?Jw}q*w1<|$Q)WWkI-n~fhFnyB2Pj;@osHtl(kdg=Hi={%`!2fGfTnI4iGi(lrkmA
zvO%W4)#!-2`|ec)I5^lLKxVdLAlw_BTrl_`z*lc`r4OFJ+-4X^GU>9Sr8b5#G;OZc
zpO`P;s50($8%=93obY*UdH{XY&z>pnU#zam?yR^3m!@ofq8rBY$B49&_$%d38y?Mt
zO6mz$YWgt#uile6@{##Gzipr32}$)_sxptopyif!kat}>nk`HlK!B;sb1(U&37X9K
z1sh93QUS#Z#v%I{<uS?mu1{aqY>DWv?}}CI;)_TL+Y3FDu*D~&1c{DH{qZ|bNH!0h
zxdfmuO)KE=MY`q?GykiKo^W$75V<Z0W~I<kL1HXL7E2Ut#|$11Eo{6Z!vgY%fri3*
z-6#b`xbQp%k{~VCP9Q3=P_{4mPP#+-QPy#1`=MNz7cOT-iOgeKw1{P_o=s4yyQ|tT
zN&b@))8kV$fb8%xC7rN@h9K#5u%d+|()|6j8dmz6(-j7~yb?Bps-xy9kA>1{qSNnm
zIe=`BHE^IwxsyRsqO^Yy5EHiA_Lu*L-RGZ={Dbo%hYqPg^XG@pQG7t;pm<C+f9lwi
zUDO<FE5-kS-;$bEft{e~qZ6Nt7u79RN^GqE`XS|uRx^$fQ9X%pm1nO0mxvR4jy${B
zjT;i@Bk#{VNFB4jDe4%vk^6b;@Q1k;>Zug2P`3YHU;39J{xu{1wKV><NBqCDS=uD|
zd^?|XCc`Y&>nd?O4kYMBA1n3#@;1J;@|N#<=uNbP+=c7WgL`sh<L0Z|=VY9a&gI(Y
ztFPN@yCQX)T#?S7Um-7~zi6^Q-|(<$Q1`5~PQ~Xp$)&Fo{-XYSSi=AFbFS`47V>BB
zTz^zJ0J+LlAjwkQPPPq4`F;ncLbBcnE~I@=BTShjuw9^jKeHGIt{^fKR2%+i7Geuw
zX0&o%S1)ao;fS9xEm{F%$}qKOyEEO5gjJe3ktFn4&1;JltL5fWn~x|C(hA7UZ-v&W
zQqEJ@{lRYfsyv<KRIk=*vh#|w)?Iaf1sU8F<7G^wfM-gdE$Wz%(&6~Ul<Wk=slK;N
z9vSQu$#vnt%aV+e)^Dx^(r|iaHpLUB>oNgu43p`CS#>~tnET+gY4kC81KNbJZKRX&
z_9niIG{*qtC7PKQ-q&`k{QQ#X>`{5z=+hx$RHv`iz(;|#fkxFJ&$^;sm2385(X;^y
z+uH@7$|~7QOe&eW%wA_NlS1bdl9dc{FG$I>PB-=fe1Y}J45oV&JM05F%wsc{l<d*4
zR9%B7l`8y*=FR^0{+W4=^fditqRh9yACK>_^CJtqDU7W3M1`#HTk1QNuM(bR50iU3
zVx5;AFMRI>65++pz#)n=pPjpXaYV{+!t$i036EDN=DSUrOcN+p4y8ZF@Bo{4R_*$`
z9!Ffwo50)>svmCgUj+PhJ|=Q_|G*p2zFS^stM@?m?xJD;trk*q;Sr76oaedOOU8|0
zQ(d}F<!;qvqG_mKPhNuBcT)hSiYd(@`IX3}lr>6O*dCdb;)l)IVM^h^^<8oJDXal$
zRQB81CehxN$8$a47y0Y%D2GI5xBU8ZERhd-5bZnWGFJY^J{oqc)9Jckuk=LK5!ee_
zlL-Js6WhG_-p8H6E@q4RY+iplbYee~Z}7@X6*UmPOm_aV73OOe{d(KlUBaZQlEDhv
zeQjQ_jmd?C<DSX&*yKQRtthO>)X2#PG21SR9o?g!z{??2Ay#D(B)sC;YDoNUP>%w>
z<J8bj&G4B(IvEJ@s@R<^*ig1X{RgxEehB{S_kXaP{$Fp<AOD0qOh(v;iq=Wt1NU_w
zm)ywh@@z66S3qnhot<As$d?VyLnE2MG@UHwy(x+_GmPZ^lhC2WytHnDFmccgyG47f
zb2d06E-o+yzR~nx+~#&Hz;lUkX?(q7YTsadby^g$)K1%Si_RIH>*?<2wI_L*2=<yb
znP8P%!%WR(z2ik#vC4f)el!@EY#wW}i0n5w@y4~DEtoQfxmJ+LYcR&%WDP86@bP4<
z(Gm%JisxP8D$_Ec6NJY664aQnRgHESscg8V`q?yJZnRP}+Kh^~B~W#q<|8~Gv{8&0
z%bsY-PSb8$(GPq$>Hz9&UB82!NCd^swgq=~p)XijT?EE#=zsq{FR`I!&o^?dT<#Hb
zy?m^3v)t`(qq$aI-8SS9p6BTr|1rnt=C(o^tDr$n$vnLpg8)bZ`f|jY+JIS`Y?eiM
zU7~dmoPM?qo*Y0M>&Xu=sT#V*oXXI<SDh^GvVT&o(a#u>TMAK^{d6cOGiHhlv(BFD
zWbNwXof8$|okuEz)-ibN@n>6i;E%-w|JSqgSjf@m54F7(V$PJ&61fbleVTD%&6S5|
zP0t;iPn}3T@|4&yai9{vv#wHIrFnd|Di&}eJN20DoTsyM2OS^sAgOQ?Lz~1W)Oo7J
zs<hORl!hR81>c#yOa^k+!3v)a#ZoNe&-n}Ijhk=nB?PPjXJ)2dlP$*)y_4ARLKBqH
zNra~mru7ul+)XY!srsO}-PP`=WoPxuGj;WlC$D>xG5S2t3&0(AYVd_hmpOiHbMV(r
zdfK`luO^U~wv2*dv*5hPc7Ay+J2&>6ERnC$@)Hd(W{Zhtv1I=JEi1HUdx}oe3gYyt
z*f|QkBm}xy4>@rV=Y{)ajqn}ZPB9rZz#I}Dd*5~SkywJinI*x<`^I`nyu^d=?$NA=
zhQF>$V8yJeKLlw+6_gMg*RS2B%IBc2S063-o`wWmY(Z8Oc;2~NQhxo8Hbp5IdCR`?
zn_8;#JzEb}$+uDxwo=eTjl=FCi7YhK0klPcNb^XUAYscjX?<vyr*-{ebk&fjl!xtq
zbNL^HY)yZEe~6*sR#6e6*FWuank@dqZ_AP8ti9Cb^noJsU6T3ptf{jl>x3s)u}BXE
zBD)O)>aL{)cs>-2=C*f|d`5)~7SC*}o|gc=SWp)QDg?!_Z&Va=0Y<MAAJHxJt7qCj
z9a3ai&KW$8j8d^NK;=SR0cpy2>wEw^rVYLW;Z79D2j<wM)Jyi-iC(O^F4`G(eLjzm
zjIstT4?Nz+{<;$z)8uUZ0<n%VQg<xcP=n+v{D`w|DXzKh2`{10B$o-lI-4&~LL^2r
zYm+i(mtM~u%jk_Z`6gtcwJ@z?H{(<wW2piP8?d^-3eXkSb1&Lm>2BS5J-!iRkdX6+
z_J&SOo62Wc$}Z^(Kc2AAXO9_5W`?i+$|{Wz0r8KmS>3FJy5!j@;(V^5qxtIIIXaPF
z8=e}>Y<_zSyqm<XPdDV7SInE6cR8oDZt_PK6%{+6y+FZ=AfA(%q8T2t2X{!SF+i=T
zDK#%CkJo=c+FA)JNHkfs_k5ISGziKLH3@vUwKiINe$4(>fxRX?K`ZaxpmpQJ(Sb`@
z=lZd$$!=ZNFL2gHoY1?ZM2dmKXb`@deNP%BYH<peDvd~q_2>M~>D+7i0cB`tD3!`s
z{v0&=jH-AlkW<f<%%PUHejC*Pj>{iQ+WPNk;eYv=$Y5;XyRI0^pL-SJY$nNM394ag
z_n^BzJox5Veaf`Po?7-=VSRjB%xTx$gQ(-P`?;z%f|Rfgabto2;DF?{?RUSn@PU*P
zxd=L!P2VQ1383pI!3(Z=yTS-HxJ2VQt9fkaU;|YP6`KVdKu3}JX7?JT&0bf$9si{P
zqyFt?w|7Ir(-<b#dv4P+#ycl0Y^oLF-m#aPP&TF6=h{W2z#n%%tb8_Pr*ypLGI-;c
z9gil{%M;lXIw5TJT1%hYUGVI_6U#XG1%1^!`q_#OabWtMiw+*0%Ovq&gz%Ksu?mMY
zP27v{l(Ep5gEUHgegEjCxoNhSrVp9QvzAe5#H6~$$<(boTTqmDHlWOA$*V4#Qhp?n
zFX=AMom=0Qyh{-MFd45kmG^N@)oYPIO={zderk%%cP-hFeLm#D>w1)C{OsX~bYpJ6
z%)MV-l2#EK$qIe3j9hX~&`hrn+0<K8XY}yCoZ^LtY^5HnP7JRt(etumv|H@C_|q~|
zY|PTiJ2%~mF!Ld6!WPe<k$aNKy0-Z)wf59UyrrdBv2*l=Ak9e(yezdql9JjQqp<;7
zPP+!(;cZD$n=Lza=+1}PkEIp=KBWG6aai*W^@36a$JiSz_cOYn;-Q9~P8s*gJBDLL
zPukZ`W~j5Joo{__3$jAJD5nQ>6Gc?l{cFOjFC@)tTCwXh)aRK#k<05UT^kxsmaL-E
zg+*cW5)OF=1JoF(e|#c5afj!4bH~8o<eSaT!?GQboj$orkNpATV>IF~QBDE<VAJcM
z%fa1ja*5c$re<{c(NW>w{#y7n_h*c662--eED1U8_NUm;<?<x5D9qXP@~ZXo=MNhz
zuB%j|7g=x<D0k)IJ!;oaugJ~sMRRh+IOZ!45C4~M4SEY1N7mf>IIRcO3}p1U%XLGr
z+R@E39X<&M81%AESh-!nR149hfi76A=PTmn@~8MMQJZZ>E7RCA;gJ!pZ0kUTU2d6|
zfnST$O0<HFa*{~mZ=shAQ)4p%NOO_>VQ$n8y;#(B|BOG!mcA<gWwr3aEX>W%T%_|d
z?DnEhm!ifk-|UGv-5-?SSL^pwhtWnUnUk9nDfjnd>e$IXrUaLK0VsY5eWvv28QH|E
zY`2=(_83Q1(wOrglFEaN<3&P<jYiQ25!Aku*SR4F4}fOQ)`4aj!ALTL<lFgig*n!x
zSN3eMDJC_#man*&Fr(6*us9QDhG3fDc-Q)q_W>DV65^CN*T`P9*dpm1C&Bfu3(+yR
z|AFZJx5!T8<j9U=BiAQ^#;k9ek~uU_`R_dreloH!CqWlGlY<^&_2W@@yGx`Mb+JhG
z6#eYtTDRLB+1juC*Fp>9M8&m@(!846<`=7oemx5rI<3lWjCzMp*Mv@~<-D}}W@_7C
z6O!2q^PH$#p_I%O10LA6EMFp4Y0e~MDNh>vZziU&Wi7_S({pFVWmKO!V(L<<)-LJ_
z;vqA8vKtWyQFkAu7@^bTyrNN%xyw7=x4DWwNyq_sSv9@l%0NVsb|v&G;7Qi|2If{B
zHt{A(<uC^|?ZU4hz>(aw-WU2Il-zgkt4*&SNh=`UsZ0ZVe>&uza#CpYK4+rzRbnv>
zx1)bWOW)Zlu75qUa~p0&40!$_AuMJOsfC2b00bEmQ{#k1@H5)@bd6J@;{le@3S^S#
zc^~kNoOuajnJA^;TjO$VG@^aR1eXGHF}e4hE?-O9oYS;GuQX>mzxvC6dB6SNKj06K
zQvdJ&&HtBwJsb}lp<diSY3+#H_bi}xx6C{1EO0MZ>*goEvWKNSX<xDu-`ygOD35yc
zZ*SvW`-T%j83T-$DWJ_fr{b=`S;XGyom=w>Neg0+j+IExOugV9FK-eniWHs*!50Zm
z;nYxGlm2O@Atje-)`7)EdTjcIGqZo^Kz`_?KUs`F09vP>ge>4Mq^Va8fuEHy($r11
zsoc~C%eQ$$Qp0FfP4z5bepP%L+lae^6K!{D^-oulU8&X;9f@%sx`9af$=#~ci5jq@
zo%tpcb<YJAatx<vTQ!siGGEmu)F5&{OL8%4+T^mNlOSq~(MNmIlJ9n@dUx*k1@zmk
z*jgY#Ssv)Ex`BJdw3>UFKDmE2pX*M1fNnM^#+)38vVJ&3S|&>I3};+0^IXB(B$S*`
zVBcc5EY6lu9!CB}>MH$hPo1WAl!I4VGS6>NNqE2l7-_7buRM0n8jq4jer9f0UA1`~
z%+ahYrWtML-L3#9<COWs%f_LgW(*+dSSiYrt~j~@O1ZKXj9nGkICYqvr6tVxo|0yk
z&ilILjf-evb8bL3Maj4=!psEaK}HfD08f)WgGr~i4gCgP<Phvmv*1NOtZezQD#sDw
zUjJj!-~+APlcivQW?PJars;oqo)Eqe1TVcsL%%{V1b;9ZCK%_W!CTK+3@5V#k;J=D
z==%Ekwj~*Zfw3$hQZoC&(3I~w{FAO;<r=yRVtRl5`_amnxTc@5;Lunh##C<S=H3qg
zfRM_EItu0ZX-Hu0irBM#SB%rke6>A;N`le1t;i99*wzEPVBH-OJt1s+ikw{30836c
zESWO=@c5uI(O^jEf*@~pTa~hfqPv9EAG<G*0&-0eT_1?<S(Aoz5x8BL1(U~aUs|K-
znEuW7M}>b-oxd2KiCHI^3vKdv*7!FnPs<n4O!?Jnz^rB`*08PfF2vrm>&tsRKK8yR
zn%Sc0Q6ymaj9Fk(0KoHh#LH@DK4W$vyF+L-ig4KvSSu|z7|&@W#ikyKrm!Exsf<Cc
zMi?^HfDTVsm%wTwIkF#`;{|4%Ic=)r!0LlA4|^41);GN$6tS&yfG!W`D3xtFDb2~Q
zD{5h>{^9AtCw`s8UGjv;g?ZE_IugTADx$*`f8Bh?<9y&QB>&u>^0WBJ-}cVNZMHJM
z(Z>Y0EY!P_$G1kRPiR>cwAKRg4Rg~8_I9>YnAg)@G&Cdxy|N8v2*c{IydBJ*CEzbx
zSNVnax0w9gSF`YWaR=Cxh*9cDzcG#}W%55AQb*=sm*Qu_V5a0r^YTQ0Vy;@)G)9&9
z6UkA$Y=i8=M((3nu;dl;yv1G;Y*a4|s`QqHc$q>4Iykad2wl&SjR~cWqwV5tgOF<7
zZwKD}o{{p0Zc=|NChZSx+8^|=FSR2iQ7sl79jdRFEaQ9?bD83meEm(6{ZzhOC=Kjv
zce2*?;Q<UDe}e{_l)BOGS3D|RhV294m*8ifBwK8KIy8CY?_IgVV+&#{YXfA8>k*Ay
zCWkk;4pZ*vq!7B!K`bo1%ZKvimLXipYVp(J0elRrTQ`<FNdAE^&RA|cHje<NjgL=}
zGd-;j5XxIbb=0aF5RrRJvC>i6M}6(}8lFpm{+u9X3BMg<-j{)8Q(b7HsC8!SV)%xJ
z?Dyim83jy@yJQw&`7BElx%5ySr{0>-{cFmXhkdId%%pU!2h>Gwh0dc7{Jpf(y!CZs
zflV$#p0~iejGV`0@5Wa6Pe_7u79+u=wd`#4Yw^KB+0Lg!>0A1zdoQ~FP1*)&UA?cJ
zzLwG*{7xs>o)piYwqy5R+MToh8lk4CMar#nOtFraXliZ6%Xjfiwo}X>G(VrT$gOQ+
zcgj=EY~%|DR~zs-R_SdvXKh|%6S%OQU3m>shtvL#0}0Ko$1b)WFQ2QgiS{Wx`<yWK
zDo}m?D(8FdZ2p1M$Ac!S)|w(RzRS6y(83}B(jn9f-KjArHe>U;n|w4pyGvElE>x7k
z>rbQO51eDLkuVEg^;bUWJn!@k6^uugikS&C)~mf@u8<gIp*6$^cZ@-_uFs}+Tb+=~
zPTFxM&xs8hJ%Q1=Uq;J<EzVf<Pp^=eipu=M1lO@=(yh__MlFbuux1}#&_DOSqRR;I
zOSkCG#Q>unhv~;>v~1mJ4PC9(vaU82h~3OlfGb;Bb#RP3I1pQpC)0=V7Y2gG1vQ>D
z+@MrDG|!O3o6f<vW0{L_n-FNIkAJDuS=zdK5b$Uz@!%lBsCa}*+V}}U?Wt&#Gbu(y
zIJ(s_)xylShZd=iMZ`R7ioreDIc&mD((#a%M?4dMckeHq<caPfX*KcKn(*}Uox=>%
z$xaXzFn(F3hq$u}Fn;BbMP}7&J9LL`pPK(z2%f~eS~jG0t>t_fP?<~eWEYxuNp~yL
zb~6;>#a?$6Zy+&l&DOWPt*xIfJJaqG&S+tX_3*W*3<2FDrIgy9;A7;gnL~r7i1m#$
zhle9v<;7bf+5xyeIY*}{`2g||gI5C-<A;{w*I^6i3|ZuRdt8FXkZl)nQQvYW>x?`p
zr+vHA<twRz$%*aYoJ(cc9_vT-eCMALc{WGN>D)z-DL%@z`qs+H5|WqwRUJ(5`dQDi
zHoY{fz?`7~WQB);?kgwbhEm^8hi>efZ2(_vXCD-=se7_^F?Scydui%qzl`K*Q|xMH
z+*OEbQwTpfmee$cBE!Vn_hV6D@Zsd=atnTu(<^e5_b>62O}AyP+F7Rtg((O)E!UhH
zgGl%4y_9N~H8hWp*LX!s(Q(Obi3zMcdt^n(dF1)$6$vX@7o>TrhXKZ8l+(T6=6%;|
zm=&<}^_7!Z0jysk9-Typ9K_Qfi}nwqwNRftd9ph+w0};6y|SwAR%>)p&dBiO$>pH5
zD`%`y7MtHk)OfRE5|d^&fKkpslUH6L)=I~HCPFg0kSo^}n403~{Du);^k^GADJ(n3
z#-wbL7o$B6-|%_#qrLM>)|Siq)=svtPzGa=+VNnl0@cjt2dwP>wY*CmxJ>A_jN+xl
zDb!Pcn$BD(S(=!f8d=;Syqj%K&o9|SgIyUZfves@QzlhWlDV_FLtvN3;v{2I{BBaL
zS$JMayaY^%x>Xz<Ofm$nt0iBfgAoMe?EHpV@iDb87S+E0kP=lFoUi_D$a9w8{bz9R
z$&&BfN)1YK<m?3%z89ZkQKUkjW#o8rAkiW~x!7WBaUH&8hxtI86sqGy>NmvKlY)c6
zcBb9VY!91ny1pNMzxU~ojktOFc?qFZwyV3GZ|Io}>W@x4sfvoCBae|_o%-JhPIY<C
z<<DZ)iU&(stI{ctyC2R$Als$jqn9ifdFppEH_<V^5H8dj?2uqA@b(NjQ$44+7z}?I
zz^;<2WRhPk)-s*G9&l%jVe{rjgh2M#%AbgAn<F;IQf!Mt`gRppXTd+A17sr7e(7Vz
zqKx;4?U|g|WgHsbEELlz7AbSSX^7X1jorezHtm-6?cvZaF(T&vHir~|u*piUw={tf
z(<X`j-Bo>c#cJ=ijSP&(0^9aV`1vlzK_y|i7ol;@PNzx3h9zbeYmFt+A{>jS#gYTx
zmJa3Ff?dCA`D`0!ONBC!^)<--1Hv!oY;N1?TUJ=7=Bm4VaGkO`0w@%3m<_THT=QnS
z#U9mLi5sva*sD90zPCAA{=(O|F$cgmo7Pig_o9&!m-T@2VW!}Chp+9H`Fws>_s!F`
z?gO$0`CB$4siDP&0^svRzjtH0PEa4a${6pronUIao+w903t_m(UZHjfx=*JAWyYq2
zL0aHO0Yl&+@aP%jpYC)_l~bc+bhPy!SOB%k>|#Oy$e?ECwmx&@=#IX}n@~wQ<QdZW
zQkM42a&-Jt(b3dbgTgzFdnS*DHnaQy@x6w>^u!r73+9d5{U}}sFJLp#P=j_DA7GZZ
zv^yLQ0aI7}M&8e0zJaQxnU;Ei2EH%KJ-yi5un6!M%4K1L2<fI=%LDTSYvsL(PP=)9
z!O~M4yG@Fzc8Da%)rN)%O<XfGjZd8szN60vm%J#nTJdc%S-p)*r<@;~SPcJk$S9h3
zYrs$>tk-KhHyau?Xbs<K7f{aU^|;&tzIC+G0g8U@dKDzbW-=AMg<5W^Y<K|W*A@;S
z+D`Ue5s%MX9GqzSp#~CGTC|`S(g8?Y0M4%Yb5u^GG%x6N^stIb4swl1+$DU6kb-&(
z^0(fDfrqYnxV=shtp&$Krs)Wo#WC4q8c>MdY`~ALXjflWcr0Rk5R}Y)T80d0WK4}>
zjKiykn3FRf40P;m$WstT<f9)K@HB&T7mt*~Ea1{wob6JpURWDaCf6pTfVE4~hSrUm
zAM}x!0hxm(I}4YGbDXlQbB=lfg~(sVOrCy0>%h2kCR3qb4MI39bjs9yj?sg(Kc!d?
zhwF|wWsGX3bX&IIjxn^BmO4|sdp$1P`t5Z2<1T+Nf9j8G*dO;KnXfsD+T@;n$jZUp
zl1`k^yyJ0#K}t)-z21|hXHg+jS;MBYWf}q(ANJXh@NlTJ8Vu!wwT$XDa+0-rw5oms
z|4Qw3B6MU?WeF3p;q6u*RF~p!(v})+X9u|;NnKG>4V#Ku!p}S2Cm>gS4HO&szRy3a
za=jC8Q&s!<XRwF^G8$E=Z%JZfF`l%>-Ec$o>b#*bC+m@lgqn(k<*YF$<De{Fh0&FJ
z6yt!e2aKcwTmhd|xB^zx5o~VoN%=ZZqvvSB6STjMd<HT3+BLh+XP5NKQ`1=83D69t
zO7*r@hK7d3f+Y#?jg*6tK09nCpCThMiaAk(#<zAZ*SjvMIiijEww^-0^~>AS;M|h5
z+Zxv2k2f!d#(|nhYg$z?$4Vlz^aG4{2lroj_ia1zz!cu8LBXXhVC8sX^*Nj%r!CgF
zX`54o$)%JPP#NLr#s_N#=}SJz3mLTgyT=ChNn;%EVhnD&`@&nk+_JSJtr#*e8oaer
z0^@}qJk3n168Qwh7j5`NejeK|#(@i4!DzzTVo6!dB;Y$bK(E@W77Jj?E1TC}d4JjL
zmV@q;lUWFSmM{!Vc<ovV638VuoUIS4!QrbbdYAM*j#cTcZCzbBHerz7HR3~E#rO}B
zoFRS~qc>bzr(x`r3CuId#A5hd)FQFjy}HTu@bWQsREfHo-VSHq5szJlbG)rEBE4_k
zuNb2$v=z$3JADkHA$ReM`L01arZ0#3op_!G56p0lQeefCM@OygJ5MKRygw$X-|&OR
zP*kBMT*CqWL1PA5Zx#iGfs8qZ5>)+-3T{_r($V~}mTCmjQ*h%Q4IOgfI3WkL$2#Ve
z2~<e6+}GQAMjvDEDP7v~y^Z=s15zEuELxE@C{zVZ;S$rd^DZqeH#k9(>5EWDmDvi*
zASqQtIq}n>7tL2@r3I4BY(@?8qYChPZ=9-@_$8AydMFQsJc$gUK`rqDLwBlNiWjcr
zho~{Vm9PBB=I~d)Eu=4)&VLsrHO?rp5|i`kgl&OWn^Xm~V(isj0N3>|bB(y>fe_f4
z8WbSWv=|(3e`}$j;`(foYWgvro7w5(!<yVW`Y}Ux@ZW)jlE20z38A-kmt|ZOE-vFw
zQrJ!=R$5*Riu==<4{aHXUE$XGq0*~k!PKHzm$Pm%JZD+i_JYeqN4m`ze({+!I*frh
z1q8ckm9<;7vVp0f*%pR&p0y(}3KW8nZ6Cvw$|I*aE6JVrI<jZkkZdcf24<w_IH>aa
zBl=00_p*cH)5)r%QN0WGq6W9NQk%9>w2$2F{9|b_=xgScg6iv@qIbqxWG@ewbm-3p
zR}a*p1D1CV%naf--1)wm`fTRrS~wSmD2En069xi^D$uWz<Ck^9ImrnLS<PV+-4{e&
z1rwy(aW0@Fx9GaH6{?TU)1oJM#Q~<KwZ{RX+^UTeLdDBL>--2Pw0nl*PvU|y_>-+s
z@Nb4r;`oM6ZDps+;#ZT@;%+jqdkhN|1xj5YX46BWe`OZ<T5+;$K!1Vu>5yuYEzs4q
zy`{C<)5WH?LJgMC*)7{2D`OP*Xr;Bwret2Mt2n`muK>G%pD*HtQ6woUs017@fP{VM
zOwr3UwOYjIqJ#S_gt*xIf(6F-Bn8#{3ol=0pbI=t%{-stzr$a|M%^0O)dOE6v3Jps
zc|7MVr+wmp(69FAZ%f0*Unc`~lqRQTF4~lyNK}iufv+X+4xgJnWjgJo4Y<@kB}5$y
zuVIv*o@Pbvr3hz|n8{X>Tl;W{fewZ&f<E4Wo*1GpbHjIi-*VGf&d$kTO-~^Z9Cw~I
zkB8nv<oEMCl;X2+o&FOu^D&!?J$cvf%levPHy&hg4`O42nYrqKbj4lNk$L$FJ74+a
zaQ`Pf?j*fB7IKnH&J5iyrLo#U93&rD`}1teiK;K#Ly|zIbB1V+=KzRm_VchmcS(4o
zYe0(cMlrmZ<NOl$+f|@@4%YdjrDM13mkJL{L|;;b5y8*dGx<+h@%nM;406@GNC~bY
zMaC|d*$2D~;jD!%k`RYIrS;wN3IrZ#)H1Ga2tCf6n4|S{R5hkUU4vDlq!=t#O_)+6
zQ)EWUS78*^9kdpI#`gRB31u(*JNE61(}WRT-sa@TuMX_I)~O*|fPiN_TniTh)34PB
z=e7Bi(Txe5N5^xX{ykt@mxfh&QQ_)78@?5@?(OM%Wg+>Im9ORI*KSjU=1+%?hN*7!
z&hMZ)N()ZgLJ#tKV0PcSPuhMfu2c$SbbL|IT2zuqmL5({Syu~$w(eqwZ^am-$7c6F
zHAT2L9mD3AeYr|YHu!WXdM+hZq?#x}9@>pT9Q$-=S~TZNAo+A_zX6s9_kKDg0*L4v
zFgoU4r0l=G|1bOfYySLE>*q-GKvGDTf+ChMA^Ik+L-u@tz-xvF{gW57S1PMga6B$h
zpIdo4bsX0$nQJ~Rd9$Kw#t>~rGW0)!m$!%eU=-zMhogmig<l0i(R!6;^b#qRzI1k@
z2JXVG3=f7Gr?tW(e}tag+1{}fggaPTx;-)Q<zY7#&%?(p>?{gDFviAfmp)bjfMLD7
zt#Oo#vETU@p}p7h$Fo{k@X&k*-7BtJ`%Fic^;bDle-CF22vB!fqN15X7wQ5oJWa#p
zo%`OP%CG`g%J<fKF-n?%M&I()u2i?FIJEoQ+uLtv-R~~{^I?hK_AKtdo4fMq&@paw
z3%+f~3$GoxBgkkp+9ep)@()%rDC?&03x`kE@;rU;&CCIh0O95Fl9%sGu17m4(YK-Z
zrG+%0L_Bh;$zR-_a@C<Z?gV2fVf(=OgUHx?x+}HOw_upx>Fi#-Z<0aE^wiWdLw6Rk
zHk{isXtUF0aBms2Q~f&v!8UK1>N_u5gRFxBgpSmYK@x_^zkMEv(24gI7jwV9k9<At
z24c=%R!;kpR}N21ubL5uo0ZgC>ELdQ(v0tBr<8m;q~mj$vSQs>d)h^*Mo?lIm;8fB
zc1p@V1~$YU4HPq`=U$a#Y*h1SxPtzlmZWtaLiegxp@seI)kT4t4v=c{B=@9}_No_@
zm0CsHnZn>TF~_9EN8))0wOu&t4x`1bgvl<C-Th?h1T(Pc_J`l*iLC6|+=fASpM@qI
zC_y*q`sj-KbJAVCSF`RcUFi2zXk1`YFzkXu63XGV{KcDXIJ9%3$ww-{GSBJ!2;7>%
zv{FtwZ7MWco?uQ;G6TTDV7<*SNSGfXYM%PQRL4zIWikr%naY+ZRbc>+j@)~i?DJyK
zVPQ~Olqe5Z;}}=xQ&_zK3H8A}m1@`w5kG^ijHGBQ;p=oMny~bB)G1S$(2pivT32T;
zCWG9p=#ahF=7cXpT1$+)+z?q&^n&xa_t9qZfT)p)sDhUx^;#&~$EJJsP0b?3;zv?A
zp^G>ND>kZ>*s^BUy}n8XAyn65aVn)L`A>+x5xOy)5_d=#A=gem*;CS0tv6{%D<oqv
z)D^1Wk}dTH;V^AhEa|qd*V7}5?oxVuMl$XE=fIxv>p01bE)G#0)8rW9IAm+JI$FUT
zGk9h$N?a%>{j@mjFdnNan>Z0AZ<wTp#QI|{6|b8J%^i&I`7B~;Z(<zi*ArtnYXkPL
zyuo*>V#04>388m_hSS08Gij}x$<Uy+GP=@ayT}mjJ#0`sFS-QfF&U!by1PBin54z-
zrjbyoM&M<?L5KN4&%o+ccLbrTZGf{p!ZMK-1~kE#Z=2v?V3{zt+;wOUPa)UPFqE3B
zP)9@=c|1krB-Q`J_xNr24;}g@W3z^Le<@9dv%tJf9e>AHcQc))Tx71C9&gKr%zy|p
zp7$|v!<S3e>RuRvO|YkvBsDUj@EySDY?f|R{^Z>yVp#k=8(#XR$5@x<&hbvFsyDEC
zZRJE|zPdN3fVfUy;(b749@KBQ&sR4^z&xT7h_!Ejh>B8?@yc(k^)9z(E{N)pHC;%4
zid<BC8&Xr`HARu@AUSo;d4?M*&Mel9afprXyR5$D)>yW0kW0{yM1(>{=bd+DNfb%U
zt0?LzPCS~KYc`};p9Y$0n5Y5`UdU;G_+GqJ;Tfk8hl^pM{ql)(H*sYD%dz<Qp>lNn
zmC=P+cw<T*a@*(UDREcvbbQM2k>;2gNhj2xf5UaE5ruv1pQcP{T<JlmHm-+;v*+~d
zxiw*D65E;~K}5KqBb(KdkbT0(C2NX^uj+W*Cd3)Kl-wmjKOhvU*@njC9G+KE6KIkT
z!dIhIeyN%H$7+-M*NEU2C(b2k{LJFBcDZ>9f%-Zi2AoN_T9M>hz4UsN(PRa{UdNTH
zD{sORdlRD}?lZ9q_B<;$&4v1ahP%xzm_|$QSy`ng2tURz55+1uR}G$Z4F+N`mx~?C
zmxixi$a_0YBfPx}-OzD~PDn{fhz5W&I|e&43UsEYr*GGK5<>GQP}Ty9@rdl#|C|E<
zxnC-TzL{TEDhOfUGIr!j#K8}@<6W*N;AAvePWW6Sc@0~`Y;(}tP><vvCK4_PvH&xo
z&0|^3PUlAl>}wugA<md=<E?L5PPr;QSnlk~oE`?Ylw>RKaDDqmXl)ngR2{~YUir28
zUd0l9Xx%7d%`9=mU+ETVEAL+N_qw(3?ftzqIQ`Qhtjm1W^j;k>{~cJywYsw}zvK}b
z0I7&FZ!;(9<)KesF=bHAC$G?NJyHs)xNiJ9>O#8G?3@>6b7wky{h(-A;Z8$$<2&~$
z7OxQArV?RGo841?*4SDlMWqMEUBAAKhf7&F<CW2n=quHhG|{P0m+M$<R;_cw%aW5%
z{+UDi59a!xpXjZ{V8QRYwP0GVtydmV%VKA*%~?2Tg`D3MRBNe0;yEtId=^}6%pIc5
zG=uMAv;0mwnUv4H@Ka6$1_DW4h%<mOn)80XV(wc)K)vvUBQ1A^8nhJKRxXK0ZWa?&
zU2}2JUVo!`iT~V>6rL72xyz`IchY7Ayw?f8PWrf)(*^Kyg>t&T56U>`4(z^Fpvyn^
z7bz;Gyd6)>kIoI0hG6<4D@}-MH<u$!-g8-b?iAft4ieTk&o7iuv9Z?Uz(jFoZYDsi
zPNGSu(`g4?(8k2<oEl=+ZOtWKTtG~%8cA~V?}0ell~-syn4KLbx!B}*G0kEAL|-|x
z!9>uxAb&Ng*J7Hkl-Zua<L21G7h5pu`?<kT1r<1A+oT!|ars#1cyJPEf`ohAg@0$S
z`<ajh_qYIexBup>!3!xf%eqpxLCWyI@|otJaCLc?5#&gz2@fgt95aGAN<)OwP+{@s
zv{z6t+eip^`t{szKi1dmKBhrQAOAVyH0WJjN(6u2H)E;xOZx#VpKJY#YmASQvDzCg
z4J?3dnq@)w`kD{bHuMY-bF2ib)iJ)Ut<P(OFK`#-foBXsaQ|6{WJ?KXw|R4Hcwh9A
zKbOX~Dm_J^ZaG6VQ*xRqb{S)up==s2|LIcK9f+4TRVmC0X3q1UuHVi!sD?l0b|GSB
z1+{!>4cmM|M!q)?4AODpc<fRjW<|Va47a#6bR~09!e%Ox2MysZ3cRm2(d!%OIaIfg
znGYn4%>x$=lbX`ul9=onZylqJ;0jZBN=DYfF8c>&===27DbZH9j3ZDtJ0+AQoWrq;
z;%{P2cOW}&wn`U*pT3OAxyqM~9j9eYuOK^@2b7~3hCk*a@B=>G2C{zhwQ1D}&cF(Z
zahLkL5-*>D<V4ydz~Wd0aU|AeVkcEZyh~mmUycb08K<lajj6>WVp_Jt)pL#u7G|Xa
z`pi==XO!*T@o)ZV>fe6pp8W8`LSgh+j$b9cK5!D^Km&$QJw3@udF>^p);44hrH_*V
zaSrhl6~?+TP8n&B`k<q!`L?5FMTDl%@w+{DKL$2pw$qc(E(l34?+Oj7U_~nvIq9Sz
zN|<sh<c#Xv>ovOO@Z48<+UiCDMT<HDeEU+jjV1&f{W#+;{%mdjb<wWT8lWAi<8us7
zy~m!$CDIJ?#?}|wj_02u;njkPCe2j5!n#YwsD>J~WI&~0@b8_{T5%_(c!SU7;%oxZ
ziILs!|Iwu{4y%p0q%|L?Rh&-7ck{R)bDzNF<O-wuGMmDxXp^q1-cxPrN@cB+amT&J
zJK3XytCG0CO*b=ytOk)O_)dK{XoY?qxX=)QIX!T1Af!qf2^UDUVFwIz=~)4LKYjYW
z;r>(iUtCgCJ}@trZ1xN<thWZdh!Ev`Tb-omB?2e-u(hx-P1{j4G(lQCyBE!|#+(V_
zo;4O$4}Y{!Q4yInd5%xk;dE2J;l=dXQ;!<lGkEEvO`9aheNl*dL#Qh0HQ^JYSVP#X
z%f;>6(x~cP{rJo%INU~>kLvKeT_2&<(%90wp%AN3m|#8dJd1cU|E#;#|J8#!Y&$!F
zoFZ@(H?8W^9;}@CrTufy2>1rhE;l|sO%V`(kRR1*eC0pY-~V5J=&OGt)O3L)jGe!+
zlj9)KO6%s3U+#Cx&y;Bi9j*pWM|CKOpSNaprQTlihO|+DWuuA7ix;XdKF{*AKC4ga
z4wQ9~JldRj7aP_}=ull52PQr1K`<AhU|wnnC3;L_|BhsBe#!hGVg#5LQzGO^v(<9R
zSNOh3+}&Lhd<o3jwC0Q(l^bi?v@%16ekg@3Pvok<ZnUYfcDtF6eIhlP(~-X&kx!A6
z+=w^j67a-@)@IqaWhi+52pf)gYZ>8s1T#GV{k|8`%PrM!gz56qsU{X7gG#{_9@W33
zg~&$kKb4%{{W)FDH@}T3YNxu4;t_Jtch<|uvt)nXR;hEuv5cigO7i}>u@HfLQEBtX
z!ZTvKaUW%YZ>+wRg6IO=I)TTJiSp}#e%R)$LzvXfcuDZ!kB93+xc_TF#e{C>aq$uN
z;SlgF_UGxA{ZduflSym}LN1`}O|W5H)UG~E?B#s@5hYt#9g<uLI94LBJ#YyQHMrf7
z1#p%tT~3_SBMD}b3Rv+#=u-x<V<1Z6yi8Ij!}j8XXEK9b;@mX#vXduQL>7j3rEDV-
zLhcfq6B@b!Sst6A_ux^DIg`!oF#q4;p{t>R#5}1?XljN3$R%W&)Q!r{?>V?<f$3X&
zM1#m0^e^FLS^}fw^`nazQM+ufi3Qb6EuYXSw#WJX%N4mz-js6qDU|D*;AXs;IVb8d
z6nF|auIGG}552PUVldq<HWgHOX05&g4TJ505E0h5xt&2Z=15gIO~TxkG58Zquu<Kn
zZ<R~TguEcj?_b$c)Ko_dZe=h{4Tc917*;3K)O?pQMhqq!S)fktgsgi;r)XXZqCf)Z
zX!~Y&$6<WYJ2q26L+fwUc>^vlp-v&Rmn~n|e|KHSB|IhjU4X&r4MIa#b1U-JEO(fj
zt&^EJzFg(`B7e}hVQBgBLArJok_;<tFDpFr7BbJ2O4*r4azmh@VFC5tDQG7`604{A
z0no}#UuEaAuYzN2{dJ#-q89;_bki=cB2&Y=J%|&WOjDI{obF`Ptsp!pIWA+iSo>ZH
z1O5d28!L#dN;>_&#$Ng+a!vLSaMw3h^y<^0tNMQ(vb#W04JbWQ8N3m})c|W@d6b!g
zkoYlZXGL9#aQF&X7vyq-GEFg|5GS@9?JS<RWzY6~bGELV@z+f@<|7iV-ZYGrrsSTk
z^+&1ZX7{8kHoK;CPRD8kP!D@NQ-kzEsBN{BkN!`HnF`%{`MDFs)A8h{`@0ur3>cOq
z7H7Rn3M`V8OGDX&Dbx79TLHO@PAYBcToFHX8NcF5r)N!R%^~_-zEouo|DzZG_ON35
z@k~B;Tw-?L@@9MCiL%KK2fT0Jgf;FCKe68E=S!1M27a`gp|xYN)6uD}P6qXif^|P<
zL-^XuKP&NmXNv#iEuccqteDnpj?VNNWNn@~eB?r4QO`<bp~$^%_>4hTTXj;-yM-vY
ziPT$ea^HJr98rg<V~@0B+p3)jM_bm&y6=x|#VhZ*DT)$Q4Qea2lYYQnkISEgs6eNO
z9)@^Ot}NUeYw4d)QvJrynFdK5vn_v7MuA-D3`ah~(l2<>kZyptu~HPjLkD;d3}K%s
zO=9XWrDVf6ctP#7=L}=o6^<oPutX(fwa+X;TPjO~)m}k@%J|gtV-Cg|-}3*@{rsPQ
zP8<dwTuer7-Qy{6rJEvkz|*cYmSCnBr@L8EPLs>hsvI!aDUrKkD8RM$t!|s3D+RWv
zw2)u-x{G_yL?^3DqimB}eTu&VVkf(7?|rnkvA{ounBY_oro76VpA++J1g)YrLPj8B
zEDwh;&8=H39nW##<C8hyMrcfM;nX+@YUALySYy6!n!9aEMA>85b-gyD_Sd8%p;x$G
z%X53Df@SJMG2u(%yhXyXOvA$H!Ig~~mR|k^T+qiGwM*)+afMl%Ml%kw-wz&Rx>Q`M
z0J?|l`B|mUv<FvtA5n8h3<63UxA!|uiUmKit$i@k_{D`ZkO#*rJCaJDF*l0madyeM
zkB7~cqf$zXuxXX|@3&6J6-k1V0yKx+>!_=Bu$`1SCVlJrTAfVwn}44o=OF#io==DB
z4$itK&14$*xny*F1@YsWa3jYXW5@1dg7B3`LFU7@Jr1CE^wc*a@N#A2s-&2}CV%wD
zvC0RNjkKHXjh)3d(mc@_AS&57;=`gbE_@nbLTMIQ0Dq>0FJ(KPEIdCK073EPyu^ts
z8VNXuIhG%l$h}Kwic$%p#rHM-G8IR4NE9UyQ?goe%*xP!R*P(}n6koc60S(J*+1Gb
z>3>c~ZO=8})HOGm0n4P?L>W<VAG;awr@2pu43h=bw{g9!`udPnuX0SXvs-rDo~K${
z6&O5OgTp<t&Q=EA!C>nqx(tE_ZdH^gO#{wYWm=)oS^Iz5d9j6Du&TGSTAc$pcI>t`
z`c|xYA>K(L2~d6eC)-hFcY_8uCc0dP=u7vbI6SM^%z&Pb<Q;(@@#tQf-MXq)61R-@
zk}l)mX)IPvOLewM#O_kCf<4Fr&lfEexG4;nOQ~WR3^?AtkoV)iTm8^C_ck3l-t;@x
zSK<S-QVL2{{~mq#y7O^XhKWuR>S$5w)sB`1T2ox}c-B475xUu~XHDYr!OLeQbBcil
zl}B6d`FIUK_EkH`^%JJgBMEDeRXU{Uezu432x_(^bKNzmZl2QoZtX(`X=pJ21v_rp
zT;e%_P>L0-vbKFSkKbtPdLHW#Q!nUeb@>)nzhmW`N!3ig1%rCN=*bh<Tpu_wNuxlh
zT-YL6%^%k?TzKM+(wF~e{TGfK;a2rZ6<0rYds}Hdss!;7;I=W1dd}_I(X7$Mmf5&B
zQ~noM>UG3+GdFOv#!g4RI10;ZK*O5BB<h5%N6_(q?(4rZ#{d287cfMyR54>JF|A=0
zps{gyHK8F!(IhUoj_0Nu78iqw+7Eo|JprFnd^*TN4WKEa!ywG9a&!{7cFkqrvWupJ
z+%0%DB;_gggh?GHj1P+aEbfaw03Oi(A@1u2io?a-e)$h^-^3UC;Su+4=V6a-1y>Y=
zx{_~|3G6>_tbh5B4gH@_Pf9+b18wDAT;SR-9Vh*II>+pNoT^A$Ie)J`k`%Jbweus0
z^2L24T73bp2SOzCtA)eBNWS29idB7hoXpHbJ3O~zgMe(@Oo?`#Qe!*Q&B$xKk^{yL
zBBm>1!rJbs0I1R5?gi%p&2m^4D&4~^#yEVGSOWB^KVH1C^68M9N|PaF`LdNvfF5GV
zi<$%<W^FVpIlOCp80yW6in{}a)`!|{2V<gr&>m8K{*cAx8J(+kZdoIvi}xGd4i~D|
zVybF?>ApTcG2Z8td>W-rAZ4m*`+j(Q$taMF(D#xiIR;BF@L8P)rsTY_Bu3g}T5<7#
zSzYY(Pf{@UUee`GS8>MX{-juDnem-(w3$^z8{6S6d9|bqk6aLPWYwYHbngdRd1row
zABU%*j}{iZ<nKNiu>6pMn(5<<`U@3lus{)H%hx4d&C2v^d=crLASoH)KQ(B$pDlou
z<sS3E<f0=N@rdDEx#0PYtUct9le5@lU-)vY05oe8gF$5;hFg9*bleT7S}WM)818CO
zGKvO#f9DK-G0@r8*Xd35JlsPpBmc@S%O>spP!}GZ*fo+<{fy+*s|Kr#xni^%VL0WW
z+~sE8Ic!R(0Lv#`&{+@UR4*ytVZBcDjpNLgZ#$dew<7x9#iS1QFJ&=Qx=%Kzc-Q%)
zN_GBuD{8|vMC%3S2!mAW68#khgMVOpyij`Vy;x}FLhNkLyjz8Ad|dymBN+5NId-2@
zwWph)ZBw4BZ&B~%_CzWaIHHiX(-|f)+@bcB7VI#(zuge)iqxa4yLcsZDJ>xg37DN2
z;D;)DPUv(f@2T22spfliqngQfUW873%kX<-W@po&OHnJn)GVb5^Tv*t3)w-e&LGue
z*@>NTfm?f8zVBIk2J)fv^LtKho!02h^?(J=ikAF4$k<byv&-S*#jWc&W)H%--JC!0
z_-nx02C^*eDyvnzAZiL4S%$Qg9?ltZ!rce5!{f19{?=~5W5dLNuQoiF(XiR#U_F!)
z=4<j+%kQ>;{d;ccaL;TCHWVjP5QfkzeID7bf+{^<@W-n?vpu670G2H)QtjmNypZ6$
z1V7`(hkjpoT@D}>syTYV#x!+5^ZTU{cv8M9+XQD_(BkL=fBIPp5{;7Pn3nza<V=dB
za)6;jPT_LWyc5vhKY*<!W%;YSBg7_8SLw#C>u$GqH-^<uUZAXKZf-xq1}!6%0`yM&
z`st8&<Kw9S<G|b+6T=P@Yn&y%Mh%eyu|Y2LLps>$*YGvKhTYQlz&njUt)<L0ZtkAE
zS5#K!+^yAUZQ6frKc&>msRb9Z!d|EN+JGE(k$Nf{o*&wB?|Vve<*S=ZNx3zYP1lXv
zyD8R-lUXeY^_r&jJle&7nC8fr+iCS>A<POww5*3XU7k-m6Ael)f4`lSaG{z9zkhkV
z5#9$#NBNX{=x*x;==~q;y=hoe*|sihZFiMrOIfIZbXyT9q?5jfy|o0TEk)@<QxFA0
zfJjeBLshAObWxBtK*~Zugb-<x03k^!O~eoo0tpZrArOSn1qhJPc$a%WXP;g7`*ZGj
zz90AQ=d2%Lt?(pkX3nwZ9CM6!ysvBf9ZrMoa#&%{zP93%3K<0S!jJNei5ZFLbwUaT
zl0YM@(+myXL5KMH{p1W_<wgbMHS^VCVl)VEi}QR$)rUnb>83;YHS<pg(wbX2o7zgf
z^|{q+Wh%=DKK}kM!}VW1)o)yUvAu9d)WW**W|FM<e5?5wx%JHkFJ}|6&5HVk$&}HK
zY?+HD$L!-$nAZ>U+v8x+%Oq`WS8GU!$1+p@x4_lFU;p>gt$%yYgs-bb#97!g)f73X
z;YK8?*8#?kLb_*H=<KH#;I7(K9V!>#s~Np2=q!*Ct_`n=E7|1a%|6h`4Acph`Q5>7
zuNdi*WK)3*F!FpbXJ(Z>B(UZlTD+1rWA#-E2&KuO6!i07-At}$VwL*Rq@5&jLe6f%
z{N!Ec8Bs3OT~#sLdK<fcbELg0zh1fv4Q9bHY#)%g(Ig_0@!1HDUqfmq)_l~T8e)%s
zi-7$?DR>a)!q5&;usN+8Q&Q?k*O<fCFn8om8Cjvk+J)`CC!2Gg2O7W6R8zjXufLPJ
zxH6r<S?&Z#bdGH;humkgmbX03R0mfu)uNH8KJI31v9$(O<iOsfc=mk1r>16$jl$(v
zYI+P5??gz{KrmfoVzFQ*2!3p+2`z1jCxFl`j}rD2hnoMzD*yCwe^5^xQ@Y}CI&`2<
zH8)m|EWABip+Nniz=RA#?Cm=?HSC4mt4U`Fo7&1{ZUvJv`!nxmJ}wBtG&m80`1`6>
zsdgq9YP{VoxNl4%CcHQGD7STjRA)iPnHEQeuO+OjoL4jUWE3{_=xwKulOkukfT9bL
zjov~3ewF0yI?kqe&21IV!2CIJH`%OjDl`u|@`KmEwD5~^SmO}YttAn3uhNis{pDDc
zfXhtM4QrHi!Jp?o**s3_Dgmn&8}foSd5Vtyg_pnWCs*kp`8sR|&&ehV!Snuh>k2lu
z)?F@mN*WNzaNJJA<A=5<<bwKv5(T}<<MqY+Xc7(NHf3;l_G9GCk~FF#sAxSxxy;=)
z7}>MEvyXb|fHT-BLENfM+|~BNbee{KOZhy$z+w62wA$3O!A}Qdmt_fW#-H^I-gO6>
zjBlw%`x9;Q)$+`7KT-gOI^X4=iN3>k0s~D3zXLfH{eCw*dbEyV&l$*n{Y&@dcU{mQ
z`rO+5->Rn3zX)j@sjC{F-(yav8{MY1J3Z|Au+1!5jCx<3lsh`Nq0BP?*WHDMS0`!R
ztf`2U>?|%0xJ1F~f`lHqa?a=ULsrou??D*)kipdJ@z%UHLH_A`Ho>!d4Q~tzgl~K+
zYpMnHjp#8Nwplv$HB0m5H<>@G8c9#1zf+~sCj(|s!vPD6GVg0df8AT!mjcCl%K-My
z^GEc<fb>Hvur01cHC|3iIk<6g_Xj&fGL8Or=M{ZuSoN9b^v^rIJ@LN7_{06W-}jEt
zPsvP7t`OF27xl&Qhd$ok{uRCX=>RJE$yQ);%`3ahgwBIo<Q?HIhH`Yz@8|!y|Nr#=
z|JdI@pFfrRCR6xn1AO;ctlVFB@IO}_M%}0`UZ7hB%{-v5m78ony6ok&FVA?Zd~3|u
zhL4GE{6JnZg7&XJ-Sb~R-VANY3ek4!H&(}SvsvCFe%&97FXboSmQM=KJg53yhw68K
zpScdQx@H>Wkq=V>P9`JLnL&hzXs@0!(*<hw73IhE1a(&4-ovG)jk@ra@MXx&W@0P3
z#+6D<ioRW8op&KW^ec)=Z$btV`GYjXN$wq-0xBC7ljL0vl|uS#j+pTT7Z(PaT2)Pc
zXed9d)vYH6XVDo;hP>;JS)?7pV%|q@@TC#oFPo+UFa5nS|9ttkN9TPRK&?jK-tCKX
zAne+~VwUwS>`L_GdBujTS)1iU8L^95#HFS+_?}-|OYap9OcPU3LR`{CC&IgA?-CZ%
z=0)97#)3mUGeZ!#;CJxQ@$~fo6YdNm!3<yURVaDWLK7pjCLLkm>a0g-x%?i7fV5?&
znYcRH3_O)Q1J79UvEGBFW({_OxC|}}Y`Raa68jMMl6@|j9>4PVz=1=PBK!MKW+QL%
z{q;JM_C01wKOJy=ABA2H@N;{ba6Ch~ZUoU{T`t5;N8h3vN*spFDo;K@eyH3j01wTN
zUm?aXO@>3nk>uWZ8+sg-xI;MH6%h^J<46+;?}q4Dl(j>V6y>*U<fN(I9km;;Z)f&+
zVsL<Qa(kCmyS?hpSlwxGNNE`kx3>ytpS~xb5f-0VqG8B5nb}yi?oMQYBOu=1AotX*
z)-!{Un*{(1pO-Oda(rZFi)^7hM(S&M8?hL<G)G@<3||t^kdlafriO1p8?cvUY7R%a
zf2HGP;K|E4VHw2m+SPWBOTSq!MdzhqR&1~@X8vh|0I|4suX6T>$fkwH<+d$mlB%rm
zUOu_qe;IK+MT4*H@!(ZiMHxN4+9RCC$a(PO!NByYse)40HHQs~_t0e3eu4DS0>Qxl
z5WB2km{Oo+<c7Wf3*XetV4i(}1;OQv;nkcnh5$1Dv6*<2IOsG=@^Z~08}U#mWs_Oa
z#lFF9@Xbk>jR8o9+rhd7E|O3btdm97r@C{Juih<L$)rb^7qlu1#^Aa4*A=qvXF=v$
zO^beF3GR80jype>_8Qg}DGHTbeS9p^<=;sZDMgEdcf&zzr`0prTn4DUUjf)t$ZtSf
zq^DPR-o(OwFpf=}7>Z1IwcdtNEq@DL%NrzUlbynZi_4H+7aeS(_}+9~B720#sol;*
zSlC(d4(@p_j)g=lTOkQD1$t;(6Jzlu0W(Fqo*yUbxkEL$E6Q(P%GoSNjs|rb2qLoS
znVmj?y08AFgWoHT<X1=8rD}xvthwk~+T_eE#K$-YZ;`VJjMJ560ei!KL#<Uis<$sp
znk_Ow$;9}=8qt+<^CekiS-#k6E^QZ@ZxPv_eLgvLhJBo#-Y;dxOf5&)95r@!{yt_m
z^JcOidg+-hSd5MxwaZoytTf`j|CihUmrqsID5Cw;?qac^qWi3C`CH3Md#~=toCFl?
zI8n$Qus@k^=pPuv@!aNBgLS|<cL@>JYF~G$_}|F<;?K(Mn?Y@AhzQ7ITG#3+ODfHm
zU9(DIc)%jeyGAFPP`9<5qPpC{jY@xufr?H(;0DM%<FAQ7+9Q$_YV<ARLLaIlol$w&
z(hZn$eRRQBU6|O!L&rsCTSa9C+y?pOkikn$fj5axd(4!n03+06#&G`wTYV!2oI2oW
zx)Ir^&=%a7Rb@T+%PZM#6N{=GcyJC}BQv+#T1^KK6aXKtDD!%A^0TZWfQ-0&DHFg*
z9N7OW&@q4hHPLoEGh$y}d0{%TJ1El7%s0+MTK03_XjW!O_3^$3@ijX~5Lr+2qgfa9
z)|19KB*orm7lyQ)l*yiN`dez&;t&g`+aWA)Fn7xj>OV%b?o4Taxzud!%Br0cd4B6R
zy<WVZ^kz~!r3V~Y-DSKSZkQ;fKzUZ_QJHyoA-fvr$M=WY{y0Sc|7yto@w@-8xH@%n
z^Eupe-HA7GY}yKeO9pqg$tkBVWNNPG^nqu_>H10!n;<}yj0Y;&aj<R<%UqcE1GVp@
zyusZGaN8~TSoOk&t*2j5&!b+>);}(w>EbdLkQxQ!4V`oswVb(@5)%OUntI{WfjcSm
zlWhePR&Fv~9uo^uZDxH<4x5Tf*&Il>_rAjG>YC)~CUgZ@;f@2a8vMB?@q*|U9M-Y^
zvM_xUsOeLD5or;5yBiP=uXgSG=>2jo0jbNn2DiVI$#YfxBE&vfzq*C;%xrKk@k$@2
zY-;D&wzD`jz6kBlHF*Px-SfFhCH3v-3(WTqCZA50u{F<{aI$s$bl`5rSJ$NbveDFY
z{N#&Ck2v){w8bO}9IlgV_s~v57(CVv6P0*rt1jYIyVzPU-T(6Ye-2tYjhEOPEA!%;
z?y1d*9~)<SuXgP$xQh~wT<&`u|I)JwIa?iA-0c%EI-gWJd$ugi*J}yns;Z@KvA1*#
z)mLd)d8l0VPm?G2NqcO>!+8s7mFMWDzf}UVWDlB3J65JaIeWH6+!)i=dlX1{k<8I*
z2{$7Gf?Z1dgFJYR()xiUa!)()9sBVlL!g;GeV)QINZL1M<`mRb5~q>y%HCd`C$fSj
z@}q}8T171BZSo4xw>Z-Nmr%c8cBIV`RX5#$IAiNgAr>3m$O<TkNgTDzrifzIR&2V|
zZnQF|>oh+d_-QHb;*F{}r6#Gt=B*@y0owdKDZXcoR<mBXSES#`=oQ+Sod`{9b6-QJ
z3`y$TW@a$qcg6*;J}!d^((+WiA?TPxWS{8E5s6z7fh#6TH}RE2nV$||>D?1_ul4mt
zaF(yi!X7ech~yi;*VWT=wUUa==-$OJcRa4ZXHi>GGg(Y+vl_Lg6)TMZNZ<GySr-MZ
z!FPhX6_~~|M<dpeQ|CNR>CR#(<*p@<A3W|(@LE8Dp07&B5mj*$GA~z?qLzlHfrzqP
zMMu<dNXPf1#Yyt_e#|UL@0jT}mK15OH~4VYwUy~_SJtp*V>9IVmg8+@w!yuEF2CBX
z-Qe>hMm}32735Xq4s`qb?`;)+X#MMd8o+_$ayF$@J)!$*>ogKSq#w8Ow@C{`EIHp$
zdeNR!w)K?#-B7g-`lYS(Y%KIf@ufY((TXEE5`aJ6!7I6mSK8i#^i9yUHrv|oYF%;s
z(9q-7T3<Y#;8sf2HrUwhV@5eKHZ7nzjp9dXrTY)VKp=LVk$O2}k_>lg-LrNhnMH?v
z-;*pej^re}1Z%EJt4?CXQ~%Pu|AW&nJX94|f(w2@UKUvu1X<Q+a<(-oKaeOazrMSU
z*)r)nhacy7*9<YDT@8CDOj|d%a*Jh|f$zMqNeOb~A-)s3v5$AVu?PCBAo<ej%6H<O
z+%2@6052EF_jl)KiYJH^h~n+c?cIrk29S?sVREOy3iF;Mz_SO6MdWfD@NpSE({+mt
z!&v4R5T+mhk^ckG_%Esn4f@kZV@6>zyHy@rwH?t)75%$3HHWTSl;$kZF8xo-H}4wy
zy2KcVb$Jc$GRA}Ad(Eq0P}eYdJk*nYF3NM-@aCHzN%DW_h=XddF!tiaX2!NM*Y5Rt
z)@#4d|Muq%$L}RrU!iA|wIyF~@jZL)4<Gt}>Y@kNh0Q}}OlYE-diz_l5XyiXGu?)=
z2oVo7rG6_+^tXD)>;H5h_wc52j1k-?00<?&SzcBHIA9*U8r}MErE32TO%|oVUkj2T
zl<2;!G1G{!9T;{q<Q;4?!nw#6SGvTva8WviQ?Yeb3h)*QqJ#^R>Pl5FP&i}IxC?4Q
z)C7o)@=aDh-mOL3WuuU!!aXtG&h7P?`FQeQfK=1p5IVcJCC`=;euV!*ZG(<RVGa#1
zm=0;=_BvUwXhoU}YqX0r*Q6PbUX>&?dUwm%y?mAsx9&3gHS>`jk>Bj5@NuPbTvH{W
z082{P%ZBO(P3Yn<3rZ>+rENf=Gs+%FuuM0O2H6%}nvyb#9z6*-VGZkBY@5>#>tJV3
z&DVGrO5M%WiP?iqZXa>E&+4!3#S!T@*I7vRKTRd7KzrTp+k3863Dk8vb%+wEY9#3y
zC}?8S*R?UTtm#5AJbX!eHWL3V`3e5(BwYj2rRKN(w=<g}B5H0IU)Olqns9X){Vc{j
zY3u3&unUM3Y+AdR_fk6}qbX6>n>=C!jH*=qadEw59SaSLg8um2W&Li2Hq@cCdu=S>
zI6k5m^q9>jLToi{m%DV4=#?%l&|M-*qPe3KyXDWG*7hz+^YkRq*vUSL+*upwOQq!c
zc)8H>75vA@9b^6NxpaI@d#_`H$J1z2_Z5os(=etPtpWSR3m?CM;!==u`tw}rY&W-D
z_0HByYRz&9p%#^!xzU%_y|y;<!MvKgxW~}55J3Z3sBjV2JmXG9`AEG|ff?@j^hRGN
zXDAITJ8F9lKzpmMn@Bm?=!9Tv4h5^Scfznys(-E9dwgWGkAMAj4anh9CEVi>iKgeb
zy6hXBSmr7yA=NWR0H$v{S<&};=35x>-GDxBHq&c5a$zohN2bik$th{rG>u*MEz1?I
zcHvclu5IS79MVdvu;tQQ{eGh*l3BAzL|~#yPz4QLNVE;b625f_7*<oeud(8Nv$Txw
zd;CdT&OPyq`&C=ETnmFWYmy&IsqAB!QRY_YN0#n3zuP&Xgm-K21_c+w_*iKF+im>z
zA0F^;QfuCM2QFkz8lRKF4b6$4eT`<Wg)oYW($X+SMp8S;rvs8*iS#UK!|s7FuCEdm
zJOvt+^%2DCDfWvfHis)V?8|21spOvJR}ac=vM%IC6zTK~T(>K|ey7+w%*A-MICP#T
zIynhhNUlTIZ1pIB^)TQS>wT_lj^_CChgH~}{tt93;hDlIacmG<6>IvaxWlu<-mx|0
zq_vE@aO|pZiCNJJ;SUoqDyA#xE%E!f-)z@&b%1IzOcxM2-ed*mUHZc>`mbMTf4TIR
zL?z+=1#7vdi^&Io0<Ceu)bSu<FZI~%BOY9vpZQ}tc`JnGT$_*iUI7stV+{&zI+NM~
z_9zPuYmHb6^FpQPI|Ljn*IUx7)vF7K;cG@FM!0wgJS4%sf;v<&M7{oE4CKq><U2O$
zeRn%2>Gq4E9G|Bzos4stI%{3u$S47gJ)^W4u!FU`VeMn?ezfSFOCSUFl0;qD6uMS~
zzTwVBUZyhqL@pcOxyggNHeEq)<_oXG!t5PM<`JnjM@($wOkQ{%Px#}H{&x@Xd@;YN
zvuWhPUB}vnO@o|GTT;8{8n`j=q==}(jEj<^JCEA<Unhl*@`h>v-D#DZdx@w7L?)jk
zK|5ilV+~wU%kBvG?16UGY(MIFX42?zd55#48>_3Oaz0tLeZ#y_z`)-(?SdkrpwqzH
zbOh$K#uoYGuC_Ps{oBSIawbk=&GgfOZj0tVrM|x`^E^ZanPG_T@?4$~St8tK-86U*
zQfB5U1}l_*om4LNPqvDs)o{g@Ri469(B<>SN!%~m#e4d!O~$x+q*XvR2NRAuPuKJ>
zJFhvvjZV>*9B(U0*<ap=%~5y=t0FWlIedhL<VyE(y;K?@9kKlDfdhXMPcV~D?FYBs
z@3n4;iCET`IF%+Fk%+;$@1QJLCh?eY^UkBN<Tyu{)L0%~Jy&?$gVQwW{~Sr2-|^i^
zoecol*v%(4?6T11^DDE~gBkElrK)hjR#k`+cf)@y<lq|tHu<h|Gk@pQJAla_y)x92
z?K&hvI~7Sz<5VeJzBuZt?%YO42yc{t={Lb~8gpf<kNv|A|72kWnSR~xU&zeqisqUI
zM)Ni;Pv61FimF;BCFjkaK9k=qebOFS2hJX>dwKa}mey8V7aroteCPCnWt_v(+*8;#
z-+n0O#KpIUT%H^dV%?n_qL<M7<RRKsf!&ZLg*QwS_e>EAGq{A%HA%HJ>ih3|e(=K$
z(y)X3!M1J%?F9)+&Ta~f?THy1(O!zP7k*6~VOPG=5?5pAc1Fr5{(X}-8oy|F4q1(l
zQ_Fg#bQnLh#o651pI1*kAD!1JFL{$E&c3)XYK?lyDZ5^-3q9KAx?`g?Q{}-maSiSA
z@S)1@Xi$xW0kN~F&?W1F7i|?bMbUA+U-4{0UhwGyFhV^!Lw~WLup&YwYH?~zNF)Xi
zGM|NtzFQgDKPMSqDI##pca=2SB=`i<9(F9RTsXba8Mepp3AQ^6Qd#U^6bl~4SR|LU
zWV}+0=n}KWp0%yjO-8M#XD1|$L}O3S5|8BW^!?nmTOQ+|t*PJ|`sU1Vo1eCI9Nywi
zMFiCGX^wk==C~$%88c5ljv>UVRB&69k7n=;7~*AZx6IdAbue!$i_kj7K(Q#__ssHm
zG*m4I8=wcB;1Fxgw0fe)CWgpxv_nkJC>h5in;OnhgF^ZmRePGcbfWgMb)tOZFeiOD
z=g)P9bqap^<NR}>{Eb<SiB;Sr?Mps%_s)o}8fbWajjZ`5%m5G8c-S@cUAP%Z!OQ7z
zxq1q>xcry?%4z@}8P&vlBg@W2MRz7=>3i0EB)4N;P^$tWTLGs+yOUQHr<M5x6EvRD
z#MM-|4|eNeP{670>F0*E18QCM)O1)8Czpu7Z?xCBg)Wcjr8vjrx&;JT=sB#T2D9{R
z!8Y>+Vt1&?z4@N~onX4xt`E{g|IT0k*zW)O)ff3(V)ax;)`Cp*tMRsJRPqe%+vU88
zcHBs<!RD)~G&V&ptvGAzoHu*Ru&XqWpy3@bx@Kyc*Lm$WZRFybu}<|sS5W8=zn<0s
z?s=#mvg41q!iz)A-G1xQu_z7hd@w5R!M7o6_%Wl)t~~v;w>L10w;SJj^TymUi*Bw5
z#wm45{XZ`z_-`A%@}W~wD7{Ko7)r#4RgP8E{7L1vJ}GKYQd4p0-mQxx?00VlsFsf(
z`rQ5qz*oF4c3ZzX5&DOp`ftBpsMzY1;5QUjKP;;(srTqC*Q|E6KjYU0+EHp?`LpRH
zq-Q!FW>=ziv{>Qzy9r|30HfD&aHi}V%x_u-t5Ei*10eyY&}pki8FN#e6WhI=#B``>
zrHd`rcbp!ha&1@hj7y}9o^`$5^$L7I7xKFRcesAH%^_kW5Q+g*q8#(rr=hvI#Zk>m
z-g{yg*_*IX_-OAK1$56~02qgZE|@RNE=s@I4if7h^x=&KM4bJy?_$7UvD;Np^V$TX
zc*fT+Y0}u{MsC3H>Sk>4+WNK=DAC>^)8@v|K|ZA!&7$d0gJ&QG0%(*?uG|ltTDRiB
z?4r((aanD#dp9DTGRFH7KON``X4<*gs-5%C5z%w@tcCKyG~Qw;FQ;u=WV%~jk4;vt
z+I|%tRGVCnzxg<gnfcj5WHKjP<o-I14^X45#fz$=Ca|ehE727HmHHr)hy+#5r~MBi
z*A*yC9aLM$gGNNlMPDbI>(AzMyakf7IbWf2Zgcc#+adQ47G<p7NZv?w`>Im#ihlPZ
zujl*SEahWXiWcV^4VH`fh0oov&JII^<C6pYxQ57R#o|`W!rwGCdId<IVFL-KimnH}
znYS)IA0Y^tKO)`aj^3W%Iu1<UVy6Fym0c^(9#6GqWkwpG&}eGS?R~m_dvv{dbVma*
zwa|51i%EcN$%o_5#vt|jVXhV0g)`;-m<BRuVtvcieQ@6?W24|_<{~pfIjwVeCm(4l
zN=(1E8_%)T&Ms+M2Z3@0A{b<*3AwDMUW0T}9ohejC0TaeWLp`P-PsaglG(HsDtg+4
zvZYwm^T1{#biKSu$eC-H(<6k!hF3Nu5q<^O6xq7JnQMvUKQf;?xAKD3Rib=!>EI!s
zceTMe6=g6xSaJe2d~?g#=G0295zj^NZQEMk&1{tcpWK^wtNJlN^`(3|aFsSsXT<jS
z^)*2e`pSgC!`ww=iuFbsF-Ei->4&sIJYr^Own}%b6HeYwRt_e1Q_CQ7g-auQzRP>%
zf}p;4pv?E(0!_@+p42NiSD>j9E~ajoYKQe)oyP`mM;HK9vw(n5qyYduVcQ$I7z3=S
zF7jQF%FoP~U3S)B`CU;aI+3y2MuX@2bOxP2ukF)u6?$^|yxOW1eRA3s=l1Eqg{zIp
zm|s6z&c7aV^PAC@+6ZoTL$H<pTQ75di&f)>$Xhx1;nLULPkz6ba#40{!BBE5=d{iO
zY(%k!+ADB00tDUZdRb9E-}?FVIp#NiIuIt)p(o>x@*3Ev!o&82)l#PjmcOS4?HY-o
zKDlMn+)};ngfQ}~TbO?RI*4>FLm2HoY{16`vOp8iq~mP;*Kv!~RD?|9lf@eu8dR8y
z{{2*jWR#u>uU(!Yl9o;y$-O@7Zma8^R!@-={`WO@QtUzxRpwL2S1Xpxl)d?ofR!<5
z6?aizb}&l;>d#w2YDd6HeSwc246qJIgYpX;?(yHrb{Ss&Hu5x(t+>+r{8`rDcjnx=
zk)LK4+FCwT52j5{bZwb*-I$F$$8=NJyFsp~i0+-$tG>J5Wpw?h%OSNF{rT4zf_yLg
zYnMv~oDSet|F!r3&B4rH?7A2?I4R5(dtU%bu#MMUbS^x2T+j=tEVlh2-|I8iFF^hK
zz?#2~f(<`v`UmB&w$~$|?pbv5q^frj?)mqqVD#hNqO?avg=uN)Mt0BS{E9?{t1AmV
zTf<Za?jC4)HqgZJcC5<j9RC|5cUsHk-%zvm1b>>gL{pP5UwUwHcnRe3mEdINxg+il
zLti^#9BscsYoPq=Xp@KL8Rxi(Hih{4;hpiS7b4Hj#7B&J{QVq;(xfIK$F^yBYCGHE
zR0kZ}k+>d7k$Jf%hi8eF))N!r=LMP$kJ4u1&-ZT+Y_-E$KeVBX!xyK>+|}LN2_mp)
zl5I8lyM<oU>g3^Nsp&K9m>5-HL0$Wbb-x1342rYD*5;Mbl@)!-tYIIdf22GT@0+>P
zxk0M$K<X6kJgS>YfaI^-hc&jBZYPM<Jcs;6lZhu@vqJCr(q#c3k=(I%<~@5{5EV4k
z*NuJ)cP>`6pt(LI$7us1C5KsClUOVB@?%phWgBNP*1h#!k>%Fuj!4LYfX5kTwNUC)
zjP5e<&^whfEkdrY)-r3Y!PSixz+0o%J0Yk;c*I5R&o;VscJq)XF<f33d>Cdj5pt&o
ze91h+8?1V}*UbQRMKus6<E()xg;LUur{tZC(KSK`p+kB1nq`Td;@jvQ3ECASS5AJr
zEWov&sLZFi;7b|>LDudkv`3LW4r~!d&sRP=e6F34LR+LLL``)DOx|H8R%lC;6@An;
zh-|)}nnD*c?pG1*%r_~Dm?pgEwT?I5=UlzOCeEi*zIyX5yFifh4e?dQ>;+p$;H-{q
zD00+$vo0A%+(a`*6Lx#I?##7hCnvC|>sCm{x%#(@@TP67vtWddb>rK4?|T20zabJy
z&-%zGmqNmzkZEb+9szXx5_?{1XZmDgPllMNHw>)*zwOnEH0EK>x^?q0bc=;%zlMoa
z>a(t-S;&@(TU*)lvNaA)pb2`oA0TVLwWAnb$8bA*zZhmd_^v<4doKTPNmi?kLV8`x
zy-7CRO_H^*sNoju%6)gUP1XJNX);;23OB0>a*E!zi^IBQkUkwacalF+1uy1=PB15>
z1@8FrB0*ezILDfyN-!e7ot&4{Z@DC+eQzYcX^B@mp=w|o?eM6u)*%A*0_`@k@6JgI
zY2ThO=l}#XHpiYZ*M^rULyqp3FIAw@-iwGrh~>Cg>zPiVMvU&;{M*lqL*4E=CcZJ8
z#R6-t1CWo7ou$Su^ZXj)X+1?r1=mZn6$u_9lP%%$rfzNQoWupeC@kV{NV}4J&O^(@
zkG34L`$TeY7m+M<RVSEV?^euxG8=>i(KJ_Sx;h@EHV6g~1MQ|?$8<f$<E-OjUX|iT
zHO3HM^eE2hgjAFfoi0+)=;a;7hxu)RHV72#iwlEsd;V|W&lV(Rew}<Bb?lHa>3uPz
zT&Yr;FmF~_mLod(S_&ITT5;i5HVey2^p#Tspy((`opl>QEwsBkns@q<XsgZ*1`7^4
zuc_1V{neiZ6d8+Q5?b6FO&KeL85x(>Zqcl7z`9zcA_D^9I&NR{zKSJ$pTYd4;5h$Z
zC+B~hYrtM7x9e+9e=&U4>ML<(8(FA-{7!vQod?rJ*XNX(b*<*ZWtm5>w(dPpySW~;
zb+2`_X4FmBSLGRy7aZfSjg&@0o3%wFqJXN^@p+P~86n~WbAb*Y8ht-6y%BbY`5MyW
zN<TNYbx(qeRMDR_PYj3G7|&R=bYAOhtSWzJuUJ2CSJi@`TPSQ9Oi&Szx-n*&`9VPx
zIC5uIREXawq9v;=71c`GYnF?MhCt3aR1y*b!88#W{H2K6aGi)3-Xq@L?<*ogAR!Ts
zK4&;2(x7_I!=quRF687LIrTgz@9fR%7jm3*8kooZ8s}JzSkvC#aEhX$LUhRRa9@rj
zCWMv0o~s8i=eji9W#(@Z*X~g9hza&*j*2RxmTz2zme0V`+M7?GUxhky0MEnY{COh4
zdi#$fHD?<2|LxTO;ycy|iW5+diB<euiRqay#sk4T$aVkG1<rKIV5IKycY_ZGg*OO}
z@cE&NzYF$407{m8gP$%HEdjLV?!vxsip}q_p&T-K%Jxl+ne6CIg0c7_GlWr+Q4ioP
zTs*3VJRUmvx}44N%mg0+fEbYAV(%&AD|NpD4P;}6Uyetv+bWi(4Co12{w|s=pxgVV
z-h%aQ=U`rYn}z?G*xCK*OvB-D>V{o#*cSKX`r6uqKTUuB@IQ}#^EA&R)sw37%lJgp
z`?f(NPhv|9Zthi%P~lCn{bi2RqHkpA8|p)@T7)dSjS(u(dl}PD04m<=Ji{oB&7`;B
zIOt-8*W}Hj$z!;Kw7Ol#JHQ(EW0lj!1jQ&7Mq8JT;0a`T!s2lI&~`EJEAE}q&ag@4
zwIEhiM<5HMN8;)eh}XEp`W19dTrUQ+Fx$3PGR7Xqq_L@}?Y(|M{^W2W%|<Ey(WCUy
zmiP#e$F`V|9W1&wv9L``fVRn5WZLd@CnU*yI}^suNj(}B0cE*6U=?pdp1CqY@DVYl
zuW4PzQ;uOkqWBoeQ~79%MoU*vy`hUP)mkEzS#|rU`m1?ULhH({_Til}ya7)Wu{8L)
zXY=dPS_eA#5u@N#6g8ofD(;lXI=Mq<oC8P4#+pFt({hnwC-T<OyPcG`0f4H6(dlXP
z5fTZ<ERwnh4UzlCrHNRSP|p{=jRkk6Yw}K_^M$dW?de2+Y(doYgcV@h-YrM5hMwkV
zt@m(Bb^>Zr9h%<O>jq63jL(d3w-$@!7d#b?J<R0F?<9F^BaT8TBO>-VQoVG%LyLF2
z$)9$wl+u`_xj^^%=+jI_V89=3%76M`621H*pRc!N(y=Id;)|eoGVu6%c~W$Mr(Tz@
zxi--^z{*xBCcNo|G=;19Y&2<qOFw70d0ZvnN=(d!AR44j*Dn(;&^%*KavF%puFq-0
z6O6G4C^WHlA(w%~Jc}0O)vejdM@aX9oLlpPw_gl&s3E>3Q^UWjHn?D7)PM9=wWCpg
zjk*(*0^zF<<&B@+On4gV_a1NW@>hhN=ORkKY(`2CpR&-S3^zw$=hsTi^sBKB$xcCK
znj_c%6KG<cqfg7dSpww@(u{%oa&4;HiFd(LQR-V+NdriHiT4wnph|bSBb;vE=4gNn
z0h-jqC@d@<ZJZu#^C<vI`PG*EWt$Ja58rXEfBqpF;W}6UAx;;3G{Qs>0v=i0-aT<h
zHbnF3MZcLf2`IxE^|0wl<~bWgP5Q0p>!~k8;nWkFX|pZM4(@3mnrT;xXUS__$5r$n
z4UEjdmB%*Mm)ff<sxnM_yFiIJ`PFIaT*r3p_Pxb$;X+_{BN2C^j&_l?HaE<4Hjlb#
zr+9~%GgP6iTFGhp9^G_2o;PavaV}uro!cCSoC_oG*r6_$s@PTQ(Ys8!O>I+0;3_|`
zygi>KN7$9*-k|lzjZeI3QZs%+DYtICtRTAqUVOd%vsi7p1P3Rme*AO*#9CYP*~`y;
zyV=vM&tE4d<)<LwCJo&68Wg&{dsWex{3z2x+d#nFO!l`-KD2%*${pJf$sYHr%iPVj
zBxCw>pRlgiIra@eQM#yi&dDJvw|`n9V^q@)K>~AMgDp-J_%{o7CVxmNc>RPnwC3q?
zTg^D<W^c4Qh=Sib=YelygJ!-18zmn5>at_rIZ!~AD+F>qA7^jdJDIX_4L$!hG7c*(
z7&X}|(kvB!Wg<cOdJCu7XuDeXuk<K*$C>mQ4s9Y;tsCkBU7W#(5|@-Ic6jB<24s)?
zR{8z+^j6#X_N(GGADmYji*8C=5<qPY?RADP-I4Zu00`NQ$NhXKeaHMK#yr~EC<Kp%
z<>dvUQwFx?rC77Ze_B9=RMr5rXpJs6@nN~Fn-$jq_V;q@{>%_tC3B((%zd4qa!fvo
zed3sT6jw#Zyq=?lIimxVSg2UXrE{LvUWIyzywW>#ui~LTTq-_1ziE`yG|DC74NQr6
z+u*++$R0`29_;ly$v@YM_>ekflO?%GqsgM>*41orw|L&S8}g(lus-=v)dykU;oy81
z47_KsYkqVF0JDWu>A@_(HV`D=-sg!VzkEp5cTa5tl)YQEkboABJ1L7mrghnylYIio
zvAC`8c^N{f32)BgT80M{@}9q`?Ub0oN<Mw%_c+(iQ6psbp|tUtVBOO+U-P}z!U$Vy
z`!0buqnaSl>_ILL7Gc`oA;k2%eT@m$n#3gv%si@0v*g88;FJDs#367lOIT$6Q0sYf
z^)I6>wwtswyu*2|O*z!OrtMO&%`fHoWuws2U(DZnwB;&|RgKQh+p6re-)l4b>k@*7
zo0;4p{J6ZE(p9DUE)42VV)#3xE_i9t3zFT>Kp6G;ls2;R+dBhwUY%Wg%TXVQ3*7CP
zMbl~1_;`4>m7IZZfRS5zyK@n7&0j;;JRYv-dUmQiOi~eYhJ-X>nrsD2(6G~WGi4@|
zi3x~NbLLYMorx^^d=WHLqyOci$88<M;mj=ZQ?TRo!)r-lH}H;0o*rf-Hn;xmFqin2
zzZ@A}7qP7QF4benFQzSSS&bB@7jorQTyKJ#+_{bzrQoDR$nwJM`bi{Q+ZGB`6(X(V
z4Lw!5>bk63dFJ&Yow>^4ht4sUMIefGf?5sMEx}#))gm<fW|W{ahyGi?EwZDdvpT7V
ztzqS{L97lNMW6e@^2dMJ#07ABZhJt9^;Jnwe#J9y4#%AAO@CFxv}WWs)VTW&)SJxZ
z*KY58?J$1*)ry6cci4Ui5#+dbo8Znom*3p^;|u@U2%s*pQ>N45uiJt1WC7v~5=oAn
zC2nf5lT7Q9JfF^y4LYk8LAExV{G|O#NY$)ufT=|#fK;m95Jg5?T!PFJEaj?p8ORwv
z$T6S8k6XDuBk_G&#ASCH2_SB7jF#^ba8K}{??;gVy7hq>Gj%yOP+{`7aYiXk6NjgC
z!?GKHgT#MBCg;Es0(48?c|K^I-7Lkbt6LfgOvCqPG-=G4m))+ioH7u?31J;a%*!L;
z{a+?kjN9B@v0izB+RCHr4!4XyV&+elVw^C83_+`}gKUjix1Wk;*6Stiz)%I=q#{Hw
zylArlm^y8NlASNQ;>tALA5NR_1DssTMmx4iz9YI~W@CNMc}=}8goNgsds4t&s(@rS
zRlvt#DsTV!0RA7J7{rLCUo#1b9f_%40K?m5egh~w`clEWZ!&o{R>yPT<+FQ(>$8<*
z$s>Sr1Q4W4JO5dR?5ru`8}Be;D{o@#6mqZ2s>HlT*R9vieI>@{Onfi-sU58U=Vx=O
z-@)72JN`c$?fOHr{>@cDUWR0nvv=FBBoO1=p7Yd+lNY@?RIVhTG^+j?9K%PvG;)LH
zvJ)3-*F2iKSDG$5d9ZDo-uMhJ={+Y#4JT*tDw^8{V`Ns104B=#_I7EfkSmuf<JRY_
zRN%U?1r_`_vSdy;&8+T?Px63Z-Rov**=EWe@X9_L@5<h-?x_%!6JaJ&g3vlDenjYZ
zC5SXr4gD}9DM9o2{3SpT-Tm^U&&<T5b>|&hO_dS}BV&@dOsmi`vMA~5%<rDtJXYsq
zRmut&pN+AoW$kg?bPL|y(e8^$&Utr-5a!uNqqnux+?-xsTcAR*#lTwQ^IWV^3~6MQ
zmzQ_WhefqTmQ|@TLZLsh```C-+pDXps`1$d(1nFCAzS4*DqOzWH}PH|!`Ta1aU41I
zK$J|+w`Fan52xWg(&#@DEnp(<FbmX|Oli(V`bLO}%_1V<?~H;*Muy@#-ro}k2fk28
zE787F{&`h6W2s=Hsg=-jj}xP5&(BFTGs`7t)HH(nj@;5~*gia$v*2f!SPpe8(J?LY
ztL<+t=JYZH8+rpjwkxI6PpKZZ#UfqfN{ZVyD^+Bup<w~49^!*n+?89<n(D!lOY_kt
z_*Z>qzu#dYTp1@@iq%ft?ZagP%4`8E+CRVFyE<p7A&!R3kS1lZ7u?RZ&vl1#hHLk9
z3;w)K9#4RJVG8QA>7`3}Bev;xoBFpiM5U{<t-@vEa0wkuR0ao(gbVG<e$K10*)rtE
zkZyMCd3wGT1yhbq@QzjLEmUaxNRC2}mIf28GIa2jN&GJ#*T3_Ha>=XOqQ9jMmVwyH
zs!lC5P|kGYfbNKimb|4YSG6p{5=kS&X6CM>l)AK;kiwHjNKw&)Bx@<WdC@n8)KzH#
zuq$OhjFbr|WA-fwKMYs2h0YYwbulJ(Wzc}E*X!}LU@K=2m~T9{kZ8!Yc{NiB`kQ_P
z^~EA_@?8>8dg++2k%<x$Pk7Fmv#D=cI>>qZl|x;Y97ppV%LX{o)p93r%y47D$>zH|
zVoUey9;4wY0O-|zerl&mSI9uq1lJ_+=-5d&>bA*m;7-2GIi7`|*GQjJvgIh3Z|-8{
z0fi3=U}q}58f|dXM!NgK^qbRar$boK?sZ%JT(Qe5(WpdIS;I~#p^umkEbXjNk%<SQ
zevi4N3iutxCpu#y0l%Z#caXxET32wp&t8q26FV78C$dtive?97)`j)!*2BKanc}-^
z4<RoddUa~0ejh`gqcfr=b5RB#Z4lK1+t5@f!pUuJy(GnQyO^OhY=$TIMwjV&5y8oi
zhw++y%ydS-Dtw<iOtFXI`l5H*GCL~=I}0?)?#HY8Hf+nj`@7%Kv0GwXgkdmiYhG3T
z^dEDe%2hm(wZx+4l~NHDX6pt%LN_9GkPs6S<Ny5W3lQ&pxy}&O*Q3U`Y5r!F&d|kM
zK<qi+i1oEYhMupN0bqZ6)|T5*lcwP#1mDE^AO<Kv!?#7>z&WvgK!Z!I1o9@M?9-15
zhM-WNCLmWo+7HM80X|$)j|Z_94g{uwsK&a{w3=AMv)c9`?JqVJ5+kczeyd*+AG;Fc
zZ$+?pvvd3H*l^#hwfG(~4xr9Py&bBC852cA4T;WQ(VyfeMOD0VuQDEjyvBL&_p5$(
zcDEQ|@la{Oh<I=N(}BN8QCn!R=b26<r)kE@W<5yQI#_oNsJ$JWoPq0l<Egt5_>)e4
z==zV0#6_I8sf+xF4c(J=2mwkuqU*-!RhGw-qn#OCrJ45B6W0nwWoiq6OvO!V$wc<5
zOgC1>$h|kd2Hm)2GKZ(L3j+9x&{2Q?ID_55hi&*P=OXNex_uZ)!{l%j1gx_*^1jR)
zWu7m_E!l1VBvxHPhFm58efQ9}!tjZsxT*6DKhSZkktNEu!3>9~iZvb<0`Qc^>+9Zq
zIgMH;IW}L-k7+Xd-gQNGQ{2uJO{BeMc!;DB1|Y}+0E)NV6Kn-mXxq1iZhC~5!n3Z2
z6lx$zf`d0Wmq-I6Q(_ia2J`!eDAQ<rgB-R(S|?}}yr}y;6Rhx^4{xACzXWA?g_Ye4
zDx8(<JHQI(;I0mz_guXIgeV^<iGnNlZ0>O)sN+2?q<F*8^VT!0{4JcSw|^EY0%a{P
zCYr2<38!k^&(LT(iI;+&qfaw3PD8WvFu<6rZJ%42h$)pTQ!6{=7b`k51deI;-jiuz
z9!f86x1ZAue)G^TRoU@r7^m^~W%=QuCh~;wyxvQe=L`~vZ#c=F<;I!R3Xb46FGIT{
zR;i+_h2(ZQ1Nmm>PX`(T+**#m9a&VtMe{7QyCVEW&Txr}Y?u~SGhj|PC9i6d7<G4u
zmmWyaW~>==Idw+?X%OQN2mzTAhKmy1<8MM@EVrj)_-X4V?L^GD<8s(eWytIuVl;Ri
z!eynvawq>*wvq^qr~Hbw-tN8k9?BVjMW5Co`u{k1E~adHk1$tlLbdYy1>|FL37)<&
zf2;a9IIQvYqCrj;G+kbpaTqee8Rov%Pscgfsx=ocZJc0W<ZSNyv`Dg_uzFRjNUNU?
zv}BT4UA^H?XN>DdhX}Kgs^G``^V~{ntof$Ku+|Yua@8L8`J$^W*pXd@V41Y|1;j^M
zg@`sgV>*NG%{E=FTHY~fub*xV<CE;<${t(@OaJ_U9}-{49V&B-ohhzx%yqhub6qF+
zv*^}^{OjsDCyPI)jf(C9Y&TI40NqGA7T(8$OZ|;l>w<?A0{g6v(cobAQd&B|m)ntu
zF`7QNbJEDbr=zE58xIy58Et>>ApCToGzfRRFNBRlSR~i2B|{b)_YfbBJR!Zi`THb~
zi4maf87zg1@5yhVEsA06&_s|F-Q3r(v>AP;@|j}Nwhc%(1m&h5k_k%>Xewu8u8SqM
z74u)A%?9Z=wbtYG-wJ0+%QA*GgLIvX5+pXh&HQ?pkk{dvF%`q}Xi?L%y;8tt5h8T9
zc)v<jS(}O<W)SK|tH=vG(e{em*|nVwk;>@Q>zu|dYflUV9Ek&DZ%QmB+KEi3p4E_p
zA~~PtKEJ8`6aANv35;Um*8`oxZA)3EuA?|m2WHp;5~k;^d5>+nK7UNe?+cDxX^=SK
zEVsCM3vid?DC*zDvaQnUQHkx+!lA$!E3@F)HN>Vw?}~abG`kFKUsguD$RFZw@A)|M
z+5Dz^!uhbEJ;GPV;-nh-$KQxh$(Yxj_oQC2#&P^EYk6%0SxDN}EED0<5X{N}_xef*
z$QE5ObxF_n2JACv{=a*G^_RXv74?1N#&1=9Pj&ug0{Eiir%|bNjz0a<fv&zo0HNa6
zd;fnpGXHUtf9L9`FMRY>RnH*y31LLb3~of(8Reg5tmQh^JZ4i{Fz-q%7$exwHynms
z+~q`<^ck0lf!d)c@(<WqWlxXHn9ZizHTbx@0W=TRi2cA07@`Hynxxo{UPk43$m0?}
z&-)Xd`{gUj1y1@UJjKKY4=}7&yt|Pe5m>lG&}VM*qbF?%$~~b142_tCMYrH>0)8Gz
zkHHS3#!@D<(IcIqvjhQK`M2b-$+71WfSJe$1lp8#8iow9d{kWkkb*Jb-G=45C{z{$
zf_ZE}+vM|^<g;KNz!i)s0nf?r^Mbxk0|F)|Pk$E5yBbul0%Q+>^zUczLdRu!dcphu
z80G)`JLh^xsQDwsuF0uVypv|u>(%FEqoDuYg1jPgH~y^CqjFuv(p)?1r}xfwKD_=+
z9$8U;*Pu`moT*;CAZ!0kR-0IMDA@(2KCM|I)}+KuU>VN-!-g@){NpIR5S)gcp5FdQ
za&^idNdrXO{EggU!Qh@Qt+XP9wZ8wH)tSC=b6f3%+&+PgsOA?gBVCh`u^57rH<_PX
zM+YMd=jS>Qv>J7><_LUNxqOC-Z7F%3#Xok{EhZ|abN7l#_BgfA5#}5=JiFDCNA=SL
z(JJKKzGmI;j8@Nss9m_zR&E{gps+!VU5Z?=cAMBHT(Pw0hL02IDtf%}a_~dpP#Ziw
z91=I~`}A{0u7leVt%w2#<qC9v=|3s>bRZ6s>go_OyBJUNR8>w#`iv#WEsl-@i*7-J
zdWf69NV>|Oe9qXs0pJnc+QAj2mxxU@uay*DPLY^NwhLs9@Yvw77Taa`%!cn^%2ESR
z#i0AB9I=<j(FWlT-7kZ=Q95q|6LNv9|Epbh>58z<cTg-dr~*c$Fmop{FQXC&Uk<K*
z4;oLWKTd4s42;uc+RsdC`R!j;>zRz+Bv#(B^ZBUkRLyB@zc)Hef}WIp8}1P=QQWg_
z++txqXEwWriE}_^Q^kOhmbT((tNj;XVfb}4&Ltp`+c26h#D9q2^Rx?bS1#qmXWH7?
z5d^gtyH3`C^V;u5EQ>}I9iQZT4U#A%ynP~5CpgF8fi+b{t#B3y45(MI?par<y@^qZ
zPg)f15>8SYyViWHjMtZj*%NGLD%8^i*+<qp*3<^aosl-1nh?pCfR&irPQ;hxVr*$$
z7lU78m4M>54D|?%ddLQ6s1P?u(1f9i_kW8)3EU!tr0&uM4j-{nr}=wOIJhlJd3;8c
z<y~1-HEWo-C}<R!Bs*?E5q`9TiuR%Q?YA?vE<bcGlxl5ONr&EF3cEK^De_IIY15!L
zo`x&4FvFh?#J2xJ(>hHytFCEetkEyxaA#)Ck!{(WYlXuCwMh$Yu$9}d$^nHmm%5{1
zGhZGZ>#7vHGbe}Wkx0X4x-q8o)({4uBlCeFD=D6I3+KAyP&k=yBTqHI9Ty@Ybuoha
zyT<F?xG8+2Vh$Y-m9coy6`=;n##1<LWp2g9O1UqKt1h=`RJpD#JcW4H<PuzM2j~yW
zb}*hYTB9v)uJDL7d_>n)nIhEFzlioEM-4ZvR`@~d7t*@<IxPC^neJ>TG_Vtfw)Dz|
zy<h%xpaZ1-Dr(z;8iuGe5?Ohz`9rt4d)FJcP~CK4rH;aHx`Oi@%$dT9naDOk*;=O0
z2dA@|tRI!Rxo!`XNq(<lI@UA-VSOz@-Qt~4RMxCCkjxg#N?ZVhsytKP-bl7K2|G_u
zsI!Xg{d8b}a_4Ru$!>TO@<Gzw<f8%ZOk^$9z(w1fAPu|2X}I4;S5|!|Pa7%WWZp%w
zy+wBD_q;)o9Rqj99MFAge>XL2odI&H5vH5QdvN8*39?Mb2@WUADI;C1vENmC_Ebd~
z-dnKu=t2&pQo4U5>Mko4;0~6o@n5GZ2TNjtyMavw2Z-Z$Bd%(B?HNG<2L$CoqW;!|
z2gCu=IrXMZef;V3);)cz4h50Tx$Q#At}yO0FT=8?7i6QD)_nwCy}e`i)6ku)0J{W{
zRSFA|G8AERZ@t(pqo-f&o)4cApAr<G@oYspCVer#tW$(Yu2sQBSr~LuM-4`rCmagN
zjavcQ_oo|SF$&bi4RUusJ;MdMQ}exxTjvIPvS(DgOH2z1*5z-G^!KIuSnsPWQPZb6
zRk$vj5vahrZ|d8k<)$42RLhJH3GUw=edcGbOA=(Jc;$<9!j*i9ieb0B+s({yaa(Er
zL$EGH?TGOTWThpHBvv!=DNJ9xnO~ZywLlo{{z&$!MA<ndaqoz&a??jItywj%#r6^a
zel<`R+TGJGXt*REgAy-sW2<wXoUYuoXyYl50M43>ayc4OfSVZja1V}&KjuNur1M3E
zo6wocENjMqzG6#oqafr|$M-(HrTVL}A6Ks1Iu0A7O>&>W0~D9FnLWNOt5zwqkjLe-
z#^hpRsgrg@fJ2RaZP**m%B|Pnm>-DMN5fN~>W6I>hfBmfe)9A|ZXYd*dY=fYja?!X
zWhdD=?Jy3%7I5f=B}%lgIVLmy$V?S^E>K``9DkS@fW?>B%#D1jSe<m6-TGj<uZq$2
z>3}AtvpjFs1`CR9OT?jx!Fe2ZA<mX33}j$=wmd=ic7?%7Ibcfr?;l?o?!ONiDG~j7
z-|E@k(ZScpKLFp=?>UUv3j)-?cf(hcH|$Vf{*i_4-<gg0kQ-%+$FeI&`OPeMEH(V_
z);Dap8Qg>lCPH5HwXEsIaKV)egMun!Z6_GwX)&m6oYh3sGiV)3f$HmCY~WYrKzXYx
zAFC`x2N&ktH<%AeF5yVp*iy_cQA>{SB?p)dA{iFUQC->W=nsRB+Mi|Ua8<Zk_0KBI
zY8N|y%Udoj_pFNRbtoN8qY~?PsWxz&ESZjFnV9f9B%70(83H1**Cr{LIUSUHj0|_m
zO+bgtpeFO-nl=fIu#!iOcQ?lg7I3rD@RSe}xX0QuT&sP90<v_!^^j)!X2K`FP<X_Y
zb0#))6x3g?F|K^2TD+ezw0#0Q)Vyfz=_j=IC^+6O+E(i{*@(U5{?*sSs`bW@Hh00|
z;tV!;beQIsxkU?a(GQP__Dz}7rMGmiDG=%@^!fs5hzU^9KFfkrE7ex<=R<hed1;lF
z9!&s{<D}}JUyT3df49-(R}a8|Sc;nQy8Sl*e(y6J75VJz%c`FNsP7-x{vCjtjP*IM
z0f2AL2eXX;7fvMI2k@u#fKLZ>UhAN)0|bUZZ_=E5KA7jsJxZKYTZTGymrN;+ANtbg
z9k~*&f2r_j#!;#;gxvg0VZm$iLmX&Ky<^SEl@>JMlQVQB*?Zr5Uvclu&ICqn+#tcW
zuV&@B*!AZ%E7fx@>F`Id>D@uY?cC;)!nLc=dcxavc{7nm@L+$RNwHpt*9L~Ni7CJy
z0y1n@G#~9g>p^VXV`+ADu1*nkTN*(g!4TwGZaV}cn@<b!KrOg6-e&Kgl^ESwjPGoS
zzaK~1GCO^V{J7<PVsY<6z{Lh9x7)wZ)GW4qI*?i;KTw8Bg%MKT^;hXc5~1*m98$JA
zHDu=&HDbjMF}<hHH>Y1uRHE(Po_uXi(Twg(nSdCf10D}CI-}^OIGh*X+!-+n*EL8j
z=r*qOdSU~f`gEWsDnK7x)>l7F5-v4O(<u&_75|I9GmlFv%l3V$>vkonRAO5uCdTO|
zMy<pt#wkuvoyj;?Y2pAXr9@Fti5dq)q`ESg#DPTPJW!TV%ODa5zzM845eFoSiX%o*
zoDc;RMa4YQ@7^Bn`>SvFAFujT=by8WoCDi^_FiZ2b=La+O2wsJWVOtif(c*&5EwdI
z+mRGY3JAh~7OWOc-*d-v>(eg|Y#Lg#BELkn_~Bo!G!%jT=-sCr57QJB9@N#+-{oVR
zYAS85kubce?QI=-q#FJ76t+&{GSj;nEvv)vnk_7w5lIWiOcAQ3Rksch1YJNNc%x(8
zC&b2}iE|^*0eXrWXA({8P2;D+5)xyJs~?qrCB%YVKQjnBz6Xp>IT~^^PpA>3$HTbh
zjI~&V{ohV>P;B~I9-DJf9$I^-Ed+Y#hSHUbA~|$pZ{UHy_=cjiRAIkI9^xDjq@&yC
zgw~Nipr6Q;G=H+|2)s;BtBWa#%;!$28u@usqb38*Y>r)Nkz)aJKBvU*u-Y)iz}ttZ
zkQwRlXBA<qK4@L#<RJcOuX5Wutb5*EOr@EO&pO7?Gv3p%u|i%J08Mx6f|uqgH~_z?
z!3cnX6hQODdq$DTLnPmXVCMHDF?(vl3{x&Z^JrR^>*>f;Efp=ZG}582n{uH*6&smx
z>X`+F{A1}`CVf$nt8<Nhj?bz<28vWFiK4M6SB^26@0RZYy22gi%8qDuc;pS>F8qBv
zz_yhXP!6nE-s_<cY#5pSZ-C5yJ080r<6rEW6#plVeh*lZkbCAMFy!=wT?I@P1oPoQ
zTCu`@^4gEoza$d>`O<%9VXD%D_0y9_Xjl1uDJ&U6ccFY;Lz#YW+Q4AN`7}}kG+kr5
z7FlmpX|Gc0s&0ohigce{xe{_=fY`#%CRMgi4d$zUhR)o7^4$s`SE2w$Iz!YJexBG>
zQ}AVEs-ZTUV7`;&qgDA^IOVEn-vUTP^Syk=CFshYpHHk1-qDVI9Kk~Gqg?UuIk~g-
zZf+%|5yTWoZ72QsF)iqTq?J-RmJ{01Nii82*mwxZ3$y#~Z*Qi&tX9!squ!S&BZC)P
zw7OK87gpH-LGX9=zHq%74)`0q`_z>!ZBaNW!K*6JvV|M{qI<ZW4sy#*V0+#Rv9}Qo
z$R>u4VV7Eu=Dgk#Ot5$431w57FHj146<y!;W{|cL(;(Eqj|pXVk2@<Wop!y6X65=8
zRJ`{_15GoVC&A0n6cdHn!!1dy3Sy4!iW0z93h)gwhGO}(bNSpVWC5r~Nl!(-tbM|_
zJF}NpE1Ax?sCJOxv1M`c^k;bh;B!ld*;KU2?V8pekdHim7F5@5s&!+ZWk8cl!s(X9
zJgl&I;h(DMy*~nRYGTXRwbyqjTFVxC=}`flW0uSO&Ek@DA@6(*p`ze;tY3hf%nN~-
z<9rD#q0`4?v!2|lM4Ra54+o5(b!+cu%oZj>28SdFk-KJjPB$1W&&|_aN(h|F#TkcP
ztlM&H(QP`<Qi*H?=*-v{LX+riFGM1`4yyX=xyKy02WP%`WNn;BmRh(qRqEHa)i>#!
zyj_o<U$CtT%8zF}L3q<H0nVUfiU=+X({US3ygp7HD*@L^*bIJyG1&?ysH?Rm@LrDu
z1!a!LNmGaR>895_&R*NZIqe|arO3<uTE~1RlXOM6Gh9>>LUWAUo84;U!P*dS>C1WW
zEC{_X{j7-kKkolObNHLn?C)Do18<$*x#H#2MK5eAO>cW|?ES?E^h~5_U<veTC&+{L
z*3TpV52=A=@Q^E{gy#>mPT8LuxZg-$3>P);*g!OY*ah!p^T>~~$zO>xp@?WRv(48~
zEGu!W)?Hm|V|?J4%~;LKOfSxCHup7@7DmCf<qAg-DCW8cDG&Glks~^l=d|G8xRxvJ
zh2kUKjd~_C8xXgbe3*w9YD#M4Nmjz`e7R(bwzZSFIjLNy=i%k9EK*Sxm1^W{<<_Q}
zy8{_$_;A44)^c$lC%i}ZH7pvA2T;>T@n+RsE+$uxk^ARcI-Khe;qG!OgJV*X&#uZP
z2!jZS%w$H5HFDrtyjUGpPUiIZWNy7I<fCgg{wLs}YPD#eTUBU12alSKs>}YqEzkb0
z(LT}kwO>zPOOD!i?ukr3GEb@obl*s`w}fVuM$%i5W<?0s3au1c&-ktW2<2~5IeD)Z
z^Y;f%M_-X+wq36YvoGv0g{AUoG>lWf3q!HyQHI1mLKDI@6`fP(X=}4fB5!VpJ+U+9
zy7b%%|G9?8^-j*=fB>EMF<2*STv#eHUZ878OLtXL!|4hJ1|+V7m!LF|OZY2H{p+wk
zG<rZo8zy+Udr2M6MP$bMu|>x^M}2?FX>1^tu9!fC%oAYG2ildkW%G&Xxd{PF`}5lN
z$!W)8)~aEpvs;m**`{tsyMtKu;Q(CWRS|wI(Y3S)p1Q$$+}f%Pz+|T-F()r^ac0$F
z$H(5NM?WSzIPGp$r1Ql7EgqwmR(;4kr4fL}N795yVr?!-iVLs--P@GysD;bjK&*5*
z*;S635Fk`G^@7DG-x%wi5ZZkasrw(xu}wx_L{>lVY(nRhhgu1q5CUQf?`)RnpUW35
zKe{kJep34RyX5K>ZM162ZqM_40)r7+=V~+BC9qv-N=WYG@om>g(X!?OCeNr_qTv~r
zH9^o}1;0tw;f1xv$ye&9NHJVx^63tD<qK|lUbIP=Qs?^kFl#nm2C(+X<XsGb&$8))
zJsnB)1@?3XrnVasvkC~3^Y(LeD(!yyyAJSA9()lf`;l2^IeU07?&{9T${$pJk&|<N
zBEH_*>D>BoV0-NShXadAJ4Xh-|J&dBcMfLxolT;b^lZfW^+|!>gmy<b+~@A2fzt8G
zUei03SHv$$;9bhMzhWM03yu47N^l&CaPw+I>Y*Cv<dQjFFI(SGJkq&B`$fc&ux|@A
z0CLY+w6w+vS?#25$CLDJc0<}m7ZeU>TpCFn8I=tL=jJ^;vJ5*TwfIp&{h2}zZ|5%!
z4ghqFx>51VPOgiT)==7n-H2`neK-dmp=|Z_-`)f?Q#}%^C6Wwd2O98KCM~t1d!6Ug
zhwH-;#`Ozp+7CpIwhou8w98icLtpL8r=hUXg0CDkeM6SI7fjC!m=jhMhZza#;9b6t
zrpNnwZa>?2)Y4Z&GyM%<!WoWc>}_AX0lpq&%$4Fy&h1B7jZfP}UrnE$HxH@kS_+R+
zMFNXfLRRL-n3y&lc*;=(MscBUf=+Ht?Yq)e<RA&&4ey=S2`kDOC(&w`C>qPN)y)bI
z{x31?BK~vC8UIVSVhrMDf3_R=d<;!qFis)5_w3%Tv^||Y(3K+}@1c}Zb-lGj&oUdJ
z-g2uYb~?Qfw>kwY_wQBYZzX*@J=|*2y;t~LAWiTwG%#q1`EbBpe1fo&ZIt7+jppxF
z^p5gXtz;+<mU5xypMhcaLebs1KLf*T=Epw*!#MSQ&OuD7{d&Nn<~IN*Mra}s*4>hl
z_V9FV6xRHP9K)7xQJ+{|fqB^FI|}+UhV6N!DoO_S=m<Pl0c;{ZZ+!4NPd@F~x1wpu
z!sKgPlc~B*mm1O<J)-noDG)SkM9l01Loo1T{Sk3_ZxVK)hOI`uf@RR~7I(Ol=YKYV
z{&GqepV%_q;)_(@OVY8_fQ(V=PGZ}(BgeCbMyN0*d1rjIHXf(kjZiec@K@Q<D}RtZ
z0QltW{B3{u=gYs@g8$KmzdIrOajz^4^APB75OO8ByI}OP>&Zx!q+t6SII~6X$Idp1
zGxPexHKrl^%tBVLQ4aI?`mR~b`8r|pQaQX2`HqVca4zFmsC$O=vRSPaUO+((W&N6y
z%mxW^^@(z}HCxcrz(g=Rpn2E<#onbr`P+{F=>H1K&4nEsKl8iq|5}}cmNCTYy{z$l
z#5*f#hjMNK=h|H1BRfsvxBc^kb*~VB-J8HzpH80L`RK!QWD@=TdZXh`hFfVgGV%!K
z!D~_rjfjAQI6IZgh}#>{KT2HPP~`AFbr>EV$`>Iouj;ov+N8<eE(t+iW--R7@Ek+t
z8EPdF4A#*<RbkXW#<&mB@EMX=eK_#@x3LjW$NRG@Pj`2>x}6jY4DQa#@W*^p$=w+o
zG^t$8(g<*Nq?Iz^01|^3tyY<Q+THz<FM=48>7d>yPAIOap@|E)Q2gP5J-6w0$syE=
zTJ=JVw!s#x7Pnk#(62m^CRk#s_@5feu&45rWUyVZHN|X;ruR;eS24I@BsLCMMYlCg
zZUvEaU}JR^Z~RF+Yf8nfHw(dIC>K?`JZxAWDp1#JJ7#t>$&U`76N&OlI|=m2I<)N}
z91CkW^^-;H`R#4;Y2g;Of;*cyrg}eN`WI@*@SVX1g~uBJUAC=hk`B{}U1jWP393Jr
z`hsZwMg5C~zOcOWcTxij&g0(RsWm$zzSUIx1!(6_%h;c3uf)m=)Ek1Llm<~TLR2Si
zDemGKSl!7btw%P+wuD3`)EO(CE{r8t6EHRHM|R~Y`$mwFG=NgdAlGcacscaDa_hi+
z`qOEE7q*}#1L)$jC1>V37mohOY4~dw_}7$v^FRDuJD!=QxB4`V?wxN2YG)Z?q8}!I
z8I~mY1GdoM?%o95Q^o3?=s_Y)^@Qq5Ua4lmfI~2{oyo9B>Q8c4I4K|G+GHb<5W$R|
zl}-N?jku@5#JXkfRnNg^u@cWGJKLaE&^j1cPJMb_BRoaqhtcFmM(rqf19uu^lKl(4
z=CQ2c&;F)5I{2|QHXLSiyXnJ$XBiJRD?<~u%)&g~uRw0S_-c5Du;5+hSCv#>+qMO*
z&#*7u#1EAnLf;Dfdb*yNS)sX(UcOO3S8Xk}!CmDVdxX!98M%nrQZToIy<$S_<dN{2
z=~f)r(wEIi$=?-nUarSI&Z{OL1KYOVx+b*F7n*!HK&<h?&g2I3hv#j-{+oLJYl^?E
zcK^N|)o-l<6{n(^&dq=9I5iPJ$)j+N&ZX)!UilLBc4)j5KUckZ6V-6VxcCpFt2NO#
zyb~)@&z!PXfo*i1Or8-p`ERVSjqFk6k(-?*Y}>DwC-%ziDU^wJx)C8c4;MEBvGMb5
zc!Fa*!Sz2O=H|m2Bjb2Y6#=tCdb9bO!drhfRwI!wlZR^4a9HsUcMs5tU+y{I(0uy)
zzgF{i0~);+UrsITeme>X!H1k~h%lK^HBc_<6VV64Yop|tmu~1QHyrDbv)^Sy^DzSN
zfsQykZI07uvuuAV0Jw@2#K%wHM4K+gF!)oe=w0Y*@Clf>R*1fOe`0uyFsqXY-76W7
zuX@h20c}Gk5R7`Kl<m7*16*x%VLhwg5SUz!Yr@ow&XI+4bhCZUOgeuAa|^ufk9<U1
z`g7>nU+elWD1EAe2{z&P$Nzv)f4Bc|`MXCSp%1}$zU_Vu`on>J^LXHnChdof@Ba3u
z{+$y=KQ+K8aq5Pb?Gl=m+G8cr-pdlyuw<J4ZsP_>Nx1Nz_B!)Fzatao^>Z%GL?ZLd
zn#VV0^q=?mBbIJBL=P4h7R2mn?+{$Q_cf@-YnCQmNnoHKd&XLS_opb&x)?n8KCd>g
zB9E+xSX~6ac$3aDIgQ$3r+_E~LOSqUtBuX{`F^3XeL2+vK`PzY8+*qQQYt~r8AK*l
zAI=%jWE%Tvb#!#xQ`frX_19Zsswat(Q`O&09iYpl3;W^W6?LlRkQY#cP`_ot#P!Z@
zSJ!u8%X{y(ve_dIspR=|{NaiR66>^6(q2tV?U9t#uQ$Obf7blc+SW6jq0)eLh*t-c
z3d&^LQIez0kOIE<(&pY|laz9=!S+>r6&w1Xm8)T3lJhR2HA3^H`!j}X1oOLz-nTMn
zC%z(G%hMrS7DJ5;oS*|CrW4rNC!5gc3BUZcw*R8iZ?^S+G~{i&@yI1-=ar}NTg|6$
zoQ9`ilNTu5B%k6QPH2`Oz8%_gyJ5&&b*!@6)IYpjFm|NS*?#5_)9mODXt+F^bz-?J
z7g!s|&fny4$I!oD*utlz2sXRN6MSIZ@^E9aFyRHA5!rBd<gv76;S)ki6lurfQVcUE
zUDtderHlDEqSU%20vo|<Hy@qlwP4Qki3Q;F(Qfl$CPs0$$hmpqIFq|NB?cB?mu`-r
zzdiJq+8=~lT_2`aG|Eb6k2Nh1P7OTTtsiHN*Tud#0~YK4I9p=wi5w5@3ehRi9}N#>
zo$+fcniPH8+RI4m=(RDtbX+^XKVf%yO>xolC3<2=;CFXUmM&BY?+Avq)`{rQNnr(h
z`{^~%B_~8wBdM}3%#5_TQ3{KKi#BVWZc(oi_>z8l=eiYVdso1K!pxw|srs6vvBo-M
zW3#GRcG>Dp%44|*Y6`PSsPrl*vwS5nrhM54QyxHA2R%xwdS5RC9}YQO5|Yd|tMaW@
zJ)A;%AP*+|$_pAsAcQ28=f`(Dcri^`$S$kzqbC>i#-#nQ+UVvy<T1qSsFq}lJ_Ikf
zaH6F`k5YibAmkW`L`GVRG#b)z4MhZ;)mrS<D6giamsh8kUyy5)vD@*4N4UmLA~C(;
zHJ7*~VX5nRiA?}!0mPq?hQIu!P5)Je-<HS!vn3TWIU8u#%zFe%I;BrNol%AW31DvH
zoBS8g@z6YW-S*CQ3(b7l8+vZNEG{tGFO|hAcBto?UT~{)sj780s$n-DG#Qc#RO>c5
zSAi#(F#>nHyqQJ=Fv{E%39x`^aI#0v=aSIdvBuCAk71I*EIeU6Gn5IAa|J&i8h<y{
z?wn@ZrxHWHdk1UExNmEKHMF)HOax`;ij|Z!<QbYBju#<c0v@mVX`=Hd;t|j?bjCo`
z?P_xRX%2kw!-4yPNdy`Na&qnr-B@xjG3@<egZ?RnJD2eyJUTWeg77o*(c2j!B6eUK
zw4J;%{bkaNyHg7(G#b%=xLiZ4Ws@J5+xtswQR1o9=JecE)LgON^C#(c560^`Cc&_B
zHA=OFETPe?QCm(^i?U?+)6tYUR=4@sY@80zBKpc!eL(&ASv}QLiHSv3k&1kVJt-ja
zPA6*Pgx1hsVzEE{+7%{l50Gbes;OW`t08Qf;{vBGz<4uqyD+UKBq)A58J0%&JXkd~
zfIp|ljXl~Uar7mI)f9)o$H8OEm*Kl`eVUTqmGB6_pzG6c49ND6BOCOWA`EJS%W|H2
z6($Is-wKH>UQ^6;eKvoWlPyL0AP)WzeRt2i6K=}L0h&<>Ie_f_@|y+4kMAAXOh3K1
zb^iXmIFY9oja7EoW|VX~u?^Kx9UU$<yKzs$R^5_F1hYuL`rc%14>hRe|DGc9KRerB
zjz39ltyhu2Z2kyLU&LJ5ym&QU{*mcziN=QmUUFW5JT}HgQ>HNe$ApW2xkO(0w`O+z
z2YMU3ofwIETybDJ(qqY1Jr2o!<;S-HcE1Qe6aD<%)`je6B`ZW1uWnE6O8)UBFjpeO
zaC%*}18)rOjHAMMs)wJy&VU?g{j!KMgB?OA&HOU5hl9*5-A@%vxk)ROiO}~e1w2$p
z$d<cj9ydI(*+y5JPa>I2o722{>1$D^n@xd9i+gdX8KbW7zV9aASTH+8ijY`mgT^ZQ
zh*rnkFjEl}{@vEKJ1;#pCA{`zZ4-As%tspc%BPibH_l=?E<b&@V%ilE6AtoT=;B3^
zF5G@(5;PL!?pA7NSp_33?FFeChZr?Am8et>XFr@jf)7=7Pk=0A95$0*uG4<JcMNvW
zjBQDc?@ck#A&tKqC%~-}UG?xMsoVSD^lrp+YFR#dyY?C=GxIYi%UbA%1I6!=E~K9D
zyL(-v&A1qZ841?MW!UFQQKzLkXHDQiRjBp(s(4aT=81X`>HX#}@$dhuJCP3lN%MaA
zf>F?B#ByxG?S}eU=Y)MG@XcfhuD1BWB)+f)@3N-yR#pAsWQH?P)n9O@EQSchv(nVJ
zbR8Ufb8oS#pa^on!PR5ib8iOgEir$To#7et1c}9Dc}mJLb{F46$l8`Yg&bppXC-&X
z1E>{oCy2=zV{`<@cbNKmHeZa=w0M-Ek-Foa<Ze!h;3u&BjDEj_OXi9P1CaVdxn$eb
zryxK}lHN7&XO>vi#<K<ps}~YJ=acVm{qkS`#J_TPs(1H~xS;3`rENj-hLQV!SbQ%h
za9xM^uWhEt!^VVlZ$2D&C)+q#`ggziZw*NOL=$kvPj&C`?D@{ugVGyaRy-N;n`V5M
zP3I4<$;5_WOTFAH$i;+ECc?1&=Z~56s-KlSYbp(Vp5Yx)W%yb;5f%~A=8$N*_w7#F
zI(?#|XxUox#m<6m=22_d%{lW++!<?)>hnUM;t-#@u4qtwfvYNG#h=ziD8I}1=Ws=<
z!C~*F7Jgi(uBGi(Xt_2E-lw>#_so>R;Ev~HL9O0I=5V*MCeTCj^bWSatR|@1W{3%;
z7ONHItnW@&3^$hEszjF08|e)WSfad3kZ8IpqpG;+#XNm}q2$~s`cctgEr8><{oR2@
zuQ!~d9XMNx9&{sAExVoFd9V!fB|565^%_qvxp3%Bx-8|TsPQ6PQcTT8lSt%y_>6((
zF0l4bO)YHA@N^(_>J%Xp5?gRO{y!I@7SZ=h1^pBQg2iw*GR@H4&mi+^R1FV0mJQd`
zv+=N*8)h75xGDt`gwB^BB~On3^vlOCb+D%0zfM&dtokRm&|6;+XMzq->+rujVu7?F
zdupLPz$lv}3stBCb$)Yo9ywPU9s<8XO4zK(F1irSG(DqHzAwp9DI>iVf%981nx*#*
zM+6c{m%6j9t<UkoJPDahHgY<>eGy8i4P<0*(9^RD{00;+&2^?2CN`|U@60q0qg2~?
zMY@e$3hYKc9n2MSt2kSv7CB6$0K`&MyhQnM;Gset0~wBfCDChburi$|6|;b4F^>eG
zICl@CQTc*;CU0y$AB7BY!wn#h8I#lE*|?ciN<mYUdtXS_dVw%alI!0~>`U-ppJ_!U
zF7vpJYaVwQHP_DrBEv)PU2wW#PvuTMLIrufgktk@h%+QYX+QKuaejhI=rp)Uunkv)
z1cd{d5z1nqBec7>&xzfEHjrA;(5%(!`Jc25%c_+S2n6g3XK?VLD%cTpe04RPuw;Mb
zA(g6fuI$Fc&ySa77o9e(?A5lqRoSaA&+x0Y*VoSegTFkFJx_FpQJ^gjVxZ=Ga|~89
zb$s#+END(SEsR=-c6m5zJZPl8@8i}JIzo~s{B^>;<WcmBie8U}C#viQ&uT3xYTm@Y
zL~AzB{O=K=e`i|$?!Y)lav#vQ@K{$krLgE%eh4A|^s}e5_UV<A(;p6SVfQ~ADA`>!
zy76}x{d+4=E2K?tZ(W=By=qC{%d1!cOR#kyS1fZG6??iJ_<ipf^OUmyQG>j2QaMXD
zWq*fo1zCxjNZ44K{HC}l!Oc|xQ61~o&XUrAwJ|P;td&B!ig0nW4Z@^tjW2A;&Q6eG
zlW3fBC-0hFyopm*%I+mDNjOK4XPg#S3(|A6@db9zhS*cA(U)8%f%G(QJ+Nv&h1(7G
z;OLx{2a>{rQ?7M%INc^T1NtDEtHX8HOuI&9HCP_ewv*pt)MMeT&-s(bOJRP+FUtk<
zRJHUe{JRWqf0q&Zotmjib_4!!XwbUkR@sBRm<^@XJN!ZeQ+WUgQ3jQGmZTf6Oi(w~
zteVs_UOaw^=H{+68sXbSQXSZI0W4+oeYjke0A4~ZW|^nkHqUVFLZbxtWO7vkX}Edo
zT(QHLNr;mSHimBo&5g6=`Nm=H5jw0*#?$8;_=O-FD{K6abxMP{)o*JR@(J7x=}!x&
z5k;s)`rW)?T;b-s^({Vsc=eT!TXJiKe@6FcV94NRqVvPgUo5jKvZo!Abg7?h;*~Xd
z$YqGne52ZM71MWejB<F~cs%hs-Zn13I@ut*VQtr!Y!(xHZxy4d2h$FB1X)RrB~LU_
z2~ybosgZVj+-8IgCgEMZ-HdTF-jakj$0|R))9v(Jv#BgdN2~0J<mJ}~4vvJJ{%o_m
zKs_BtKkYa6RyD#65tll=L5MS|=X-x%9M!v6+xEVSSnKJYP@b*$@CcHtAC$}xznK-w
zGKn!RM$m#)wXLHg+rAFF<i-e8=HTpQM(B3l7HoRXWMz34s~CdK%cdo|V%vNPwg&xI
zZO!hPna9OOxuz!U<X+-0wCukMXa~<xxI214kuK_IPCq2Bk+-Fqzi{DHF<!{PL`Fx(
zHQitb&GqhDJ%0c-l<Nt2+lNA{OAa}^P>(LyR!oiGxHOW{wgVX~r(+4g(1Sk$ir$aE
z+xT!mXALha)$HrUD*Rrhp}3M*T(eQcVZB{;j(sf)*j{5d2oT2*N{(SG%!N+g^8Wp}
zW1-vPXvGx5BDj+f-WowalfYc;q|bCKrBA1VEKEY>JYzTEMnT*?&$i5;ZGTr&_G~uo
zLN%)=i=ad9jfQF6X#Z{woToRKX5DL}KCGvz?~;?;@PIXlPa6_8G_I{u8Ty-o2rAXO
z`~>X`z}Ply8#_uPMlDo!+}jCCZ9*TZx<OYRF$Sy4Mx#Zi5H+~yqAB_;xtwi6qWC7#
zb|XhkjpxY6OO|N)n-Q5eTU9EpZC;GPb8@C7u<3e`4JbGKc)6{OhuHY?W({|X-4R5F
zU-Q;J^qTm~-`TW((ZO%7#{ah=KiOmG>&Qdo^>yUJl-_0XAVr-A{`03k#>XFO=cQ_i
zZ)nP1ILX}+eNL{qxjRlyNlHpPI>2k`Sy5}r2d_%U1uNP;zAiIIQ_RpSX%c!!=`0Cd
zE(v{}8dSU%Iu!DrZ2jsPw(w3RMAt1L+SzodJj2BK)O2x{@VZcOiKk{&bali<4vboa
z?<TEv^7yKS6vZF(G|o1}CxuR8RUXwTDuzOb2;qu~!ML#amq?fUEXxA0uGUC=qLjEx
zeY(VDkF~B(!#pnXL6}<;=lYV8uS+1X`)%y+h<-F+vIW}e1TR1Mn~_FrM(Dc|)9EqP
z6Z69TUej8s>{hL-t^GOiM)$cFGy2b-C3=p-gmGP5Iu_P)S8clY@ssYJjU<9o-dUhR
zUW8WKEZ&5>1ot43T4UhOEqwz88V?g9^GC=4y=c#=P05O=Sqzc~`>9$=Y%nMD9?KLp
z?cR=}EC4Fm(FxUNw7{OLOy&Tl*ZV?L`+|5w$~DeiYH5Y;8Uc+2E&Fkl8p+E&PgoUO
zQ*VF-D7GD!nRtlzFw0cx&*B}7f(GW$;~lG^jU%k_{S}Cu`?CSb?AF|f7c%$c={DT8
zPnk!OEGxsKQ<onRYan2T%S(3ReR7oU8TY12my4-&le6ZsW&@(DJ|`E|n-HJiK{|7)
z*Y<FUN{K3DED5-8*nB-|c48Bo!!;9i;!h>8kB(RkzaH5SfGV#YZd;0qG!ye0ji`yk
zJI+zpI^wtL(Ha{ISKwi2IK+0v7$93E$QJAAOJoB=RRb`+=BXaAyz(rU^VKXztKG3W
z(oS{JCy>~ESUcw+y(x)WKCAuks~e!Zk1TxX(5k(;2hhV9d$Am}0P*{ZGS-X@7Ziwl
z;xt&J%@FO7dghriB+apKDz8OK*}-RyjWeYz4VZZU)vjGCx1kTBmB!s@*i+<osjK^(
z&ZXg!SL2Pj6|4=kW>SUv`~SDk{|o-|n<M96(1`!gSNvwY|Dopkr%(RPGxAS2;Xm{R
zzvbBeL(TQip8T5u|7Tn9AHDuJ1OAV;*gtpvZwCCIYrlW=>R*`w|Ad}c)>{Z^J00Qq
zOp1ACcFc6}Vr@b#_{K^x9o~>|LN^<^INlwXLA>Hv_;B)0eo?wPYkJOTWhb0Zz_W`T
zqHj&vmB(^q{mU`MnIqEbVL5Y=>-qS7{A#r;N;?34k`Z>gK?GJNYu*Dx;TCmZh-P7<
zh)Y+lzh`Oao<6gA2_b48Z%HctN>&F$nA@Dfn=Et&Nae`JRc@y-yWf$?8zk3vT#!v+
zLME^=pN|<Bl@`Y6eiit$D`-3dIuF<HI{0;kO-=qy&*2R{RO*d|I=5n_gi1et-1s#z
z`*`byx!=gkI=X@1eRL+eAiuoATKLJ>!p}>0o(864SrSvEMsFzdi8PTC(3=wX!z|Go
zdy>sHlS+H#xm5$>-YaI0S0`v^Hez6M7R|f8(rg{L)HAc(=d>|*NiZ@<s0rs^lO;j|
zc9JIS#8vfKk;wa<+zxEPB@6b1T?8`Dm(<dbU!UuuKMIdj{<1$cNL&iN!b~e03H7Wm
zIj5cB`l8xMfN%4<O>Jyy$S*^Ro(uv$rztlnSJPS*7by;NqW7Jr7BCdy3qi{xct?D9
zO(rm;7}OxXLn&~bid>ncFO1z+R+9@azHX^_WxJcxjp9qv-G7e=$!+giC^kB(#T?A4
z&^#>;WiKg;4N$IF3lhP^q-zK9#Nx5W!^|85!WOVr<a^~3!|aP`V}xpiVf`I!VNK(k
z)x1KE^75tj;X^lgZImH~3Z^FYUD~((d-{py5?g!LyzS>3uSOG{u5a(l<+_?*q}V-4
z^+`TjMqir|Z|uj^xmI1<H&#+VDqq%5ZPX@CC5?@qopxm~dIVv!IL+kdV<SGJ;oNm`
zs4Ht}TZdb*&msbX<8lemhCS3Hp6rU(I|_Bq1@O9aagqpi?f(PedCM;Y#l7j0rLqh!
zM@JP{AYD#LFLT%w$CT)Qb>q{735oMg^yuqV5hpn|T<|A`YhQ12Rjh9s?F~y}FdLLV
zYN^_l#B7t}`Vb@>FtEylRBrn@l6;-BJpG5gku2eggsSG#MI0n^2Fh&#F3)!m;qm#w
z;jWK&pk=18zP^4q`U2U`9kn~uzxv&oycPOIcyojsrytzMd5CcfM`e)mlFC5?{flGR
zhy4TfrGCNV?G}OI=xSefMm#jyVrb^`=k2hn)?<L$orB$I`Vi-N_j*%aH27F@0J|y*
z4bI9Ho9*F7O<@-5nP$R{z}-iQO?%~^gIW^?*!7G6r<mf^5mb^w!JDXD^IDIZuyIe^
z^SAJQjilIONJMa1IAFTSVcJjfL@Fokdq$4f-v0g{HD3S!od2te^8YkV3eLSu{9&UH
zcQ#u0OsHC5>CC*ib!^K=0)%vQdxYo~rxmHDK1{gS((Sw3xHLM;FKc)^rzba};SqXF
zZOd2bxIq-=XB$C<*P;wwuv&CKHW40aqo7{b9pU#pbVRzA0@iSSj!&La9-UxbCdJf9
zduwtuE*H>o-J%x6(aKCrHfNYyFHBm4BFruZJju;_dYx#z){W3OpI*$*GnkdLpNy3}
zUDi-82`-HyTH1ggFkL41z>c*SAl=oUw})hlkzKg1r9I^54XdU8;%p2m<k>_u>9j}A
zJ^v$VP>VqKWf6B#Va$>#!bR~l|3J-@Sd;;cY&JAXNOEz$dEy}JegU5|RjidNH0Co1
zZL>UL!BJjh)VRuuw1!35ySfbDn*tP+ZcWyhSIj3lYB+(ia(7Uh_Y%3I%nh6wZJ6{^
z_G39I`=u-tG1||L)gw0-?ydzO5s3LO_H>DA_wpvT-G6}(83tu1jDU(O(<=h2LUVUA
z9rcIxWB51B#|bx4%R|ea`3adpt}F#R^u-P64@BP$df7VFF<qE$PN2h4QbSjruku;Y
zJf1lbm(4SWD;fKyyT-u{&I9^lt1lA*OvF<K#r)mq=6+FuPkMZw<OY2E%?}2f`EWS&
zDLix#qf-8b*NTahr346Mm-WSeec)5himD6qCI1wa=sa52ls)i6bJ?>u{H89*nLjaS
zh>OzQ0=@_IL}bWwy6<2r>q-%$jB$D}p>(m>t8*c0(oTsf=N|wqyCNr<if?CVlrx=C
z<DkhL^;QB#LxyIk_aRBKcOf09bzWy;Z0-{hjLt5AK%sJ3!XhWB{3DLheJI4x$x?U+
zTLn?}iLG{u!w!Ukbd*h(fb|7qG$pB))2iQRIYIHF@#Fej3JVV-MEjlYfj9%v=0RC1
zW(@I6aGQx^cr8~rwpSG2RelwjEe9p0D){a9YC2f2O(Uo68dZ1P$7=WbJ{)*%<g`b;
z%3&hRB{v?5z%N!>t;DlYLa?<@L^+T#vLaUeL#ypowxGQuk>Jw%gY~hrM9;NvP1*D7
zgEn@!BE?e{=o$@l^Km*(n?{)e<0|&gsoSpU=%gimI^P0t49j57idsDDZ#a8qU3BYq
zIboP&;?nK$aKvS>M=={*#Unz#8!4+i<(_HgfysM_11Fudjo40_qyfKe@{<scTJGP#
zA!g4`(!ZGtVoPSFYP@R8_R0d_6S5&Nhu$TsyIC-CB};hzkHgJtw*{Afnlw6}qQj(2
z<)n=X3EloXNJ^{wx!qb1+`~G!3mOBw$+)z31%Us<|39YjAFIi)Y=s+HVGwO}{4O9T
zW6lpY=L}4}m*`|ejP+$%`~RuyO8ua#t*R=BGQ8yctVeE1U%u|(&NX$<>0MElnCG0}
zWfz84=M6Mte=d{|y%wqRw|s^Be5srtYm>y(qFkPZgZV+_H0v&MxqLbbiw+tknbk<U
zYYH?jlS=#Nq!A_$hq_yP5zE5y_?BIXx!C;}Fpk}+XOUZ&XCR-BLR0QgTg|iT`>qhC
zIFh26FBE!qfY&w@7}YKi3~U22pBQmfsq$UHa^+q`0sSK0zZPaQSA3#M1kU2%T?PfI
z+Ra0C=FweB^Q?E8wTObbr1??6nP%NNb$coD!-4N?t>K>0dEJ_hiRsbxDD!w4uoKT`
z;@^BXzglq2^)iz1#pd4V!X)6#=3Tid%`4sN!xce#)rzfWhlz_@%fZO<E`%{)`OXBn
z-eX{Pr!fQwDw1fC)E}*;FS%p*K!XUivYH+I9?rb|*~ZLW<t-COTe!NrC!JM05@Qo+
zKmS0?jp@5#*pu)&Ei_x~ksreAOD+dIO9)(?{=G}%63%JH3ai3Z3!|$L7BJ6ypVl28
z<yoQ&7Uq#fO(Q`yq;7#A%%e5W=ea}2HP`}5*LsY+5LYxOX*9r^OA_~R_3u;S^UBGj
zv-vv>ya1h#(^q?(l3AYEguU4l|NY0F{pr|Dop({R0hS3L3oAZo`M9~{SqJYfd&{ej
zY`8!2k`#`p8B;x7Z=_E&SaH;LIB7my=~3$DC1<?^Wh=&5)SDumZHuilSwnO8Ebh*Y
zOupl=+bI<kLVlf7)2>zX@!n*>7xe|6I{Nk0OCCYws*xbI+`7OwIm%%wQT~jln>76f
zVi+=#Q^%upW@Mf$TLFT`>y(r>pFDi9<GiXGJ3(wc7c!MvUpAg}sXpz<&3SWG>D@x+
z!`d8{E601SeziKBy!|f2EF$zBrv~Z7@n|_eL67YGCb`-tcxOujNibe^5{Qd)c$Amw
zL<<IU3t-_Fw_~{9A%FP;^d<Ws<8O8w``qegOEyi-j2X2_S0}oqb%{+_qAl)zA*gT@
zS{pemgyhNdPjz!TZwfV*BRa!<i3m9%auR8Bhw@amqMlR}v7aAbdXG+3_Q>8d=pmaY
z-uIZ8p1v!`v>rhNUkJtaE@lyTDC`Kq%J{0e<lbHdwI^8WT%iCvJeRR@=0-i#!ACD)
zVLfey*v{)pq>3x}u$6ngqRm}8xUgmVo3STyRA{Xz{w|c!afgXuv{#IpT`roGr>cX-
zj63`S72G#fE4MDVwYX|KAe_<cSqi$TCEN0`IJo)<2wn%W8yUMN3JW0Srufxu+yt58
zpZ#N2?f;vze@*U{NQ|M_^UJ%(vYyXHz~|7wl#SF|T1Tjv_xeZQ3&oQ+3OCPtc{?LL
z92#v1ljoTRk2F;r#OI&k4}DWp-#{>}=i#k9uL_Dw#p6_H$z~FQEZ&V?P&oOEa98b7
zA2h{-GCG=ioMdLY&iE<TB6cm=9fTu?!RR*&?23A#_8^Dlv(%jFkF(VMm5;Mj*3@j!
zdDp~_xn6|Q@y>LEOzl?l7o_sDp{%L3kO}YhL}KYZM{KJZMYOvMEX~p9B5In3zmnYs
zrVnn9oCvFvVk~tZqP{3zjU)8JW9$adW%C5xotEaEPbJP#6Vs9MN$OON$h^Ma=EH%$
z-SSoVY7_-KC7Ui<HF4UD%}Mc`4+~vCh4f}T+FwnIb2UqPPH#xOy5cWwLYvsFc?0VA
z>bki!+xR``bR}I{p4e)?<*qYNnx$Y*4)xma41fB%=qLV3c$|YA4&~nnyE|{2AIyGO
z6pqze?))xX0DI^#Dg%RRGg#muq5DvBOS%!bGe9WBE1*<^hx0?Yxogh0S{v*E!DmGX
z7NDRl=GX--wS|h41a9>6?x%G=D4SNj;%HG}|D*@H6*X^RVSW~ur-FWd;zoLOjkqW$
zBCe6+3eOnZTf4)*gUfnE#+f*WCtQ!&tm*4;?PYh$;p(*ME}ugUYsde4xF9_&wi=pO
zY_czjR@9%<C*M(Dx7IciLQsK7xEpAM9h>99awmpIBmC?yyd&OtNvgb+0eZ2Jjr2ct
zHw^xMl07;!@kWj-hABIv)5244G4BJu7|IWPgF6PAR`x+gjt_OGZ<4J?Sm0|TL7KOy
z!tq99`QIF`FD1t2bF!<X1bk?pbGWyh!6V<Ub+jc~Ij$bv(v#CN&V{NCN9u+)yy`bK
zfWlkn;iDqj={8)s>;A*}^9HpirO3+OWxSdCt}2<`u;E@?-!rqAl&51UL+zTNk8Xn3
zTG&_V2)o@!n7o387={f>-5M3Z3Fl&do-icSQvDL$F2yjaRW*PyV&2eHx0FmpUuC<n
z+57VN#bAD6xGhN{n9Gmy3ej{ON1h_A<<Pmt&DhY9$4}jLzS`-eHhIrZ3?_*N=>eD9
zGrtP8DMdNBqTAI<*Ap;j_HG4k-lF_6F+y#8BN*(Sx4p?M3#uFb8QgZF)_hdTOH;X9
zwnBe{B|;P^<5R0tcU`lgyMEI~)J-ECV-uI9G}~yYR{tjD*?S!h+>2;iLU-5k&JUd*
zf4tTA)3RShj0X3VFmm!?3{tNH)sSvCfo~NEdGH_VQ2vv`O`KI;be}x5v>MoU?#*%b
zfBg5&{&nqtJ?{n7Jo|iR+GUA@wQ5+#WM)NVL}r>en;S4G&h`RdplCc<cB6CWn<};Y
zAx1F<mSr{V0dhSw#B$s5uquJx4R0;a4Xo=frP61g;<acCE?k}5pMKKWb=N1UYd;}i
z>w0)y-Q|$W_4Pt$xf!Kg`i{k}SXf}DKR3r(Wb-apE4b|Lb?p<?E3jMM60!RU9ga>w
z?kr2to6V7{Twj<rSXgoZB|WW`YtF^}0@cbHNH{}p&3Al@k{i(q184;QqA7RYOm0LQ
z`cH(Lcmr=`_D1@@;22%JNNy|>tZk$gRZW)%uDmm<g=b9c=6tuE%&+NukITVHarYQ4
z)@o_oB}B7$syd{=&b{-eNdB@MTbVlFiHW5EE3<*-mt#L1c$|27KLY3UZnty)&qkDV
zk<L{@1FRX*cXnKxHJp2Ao{rnxVRd`oR(P#mupk(UAT=I&T}W&Ym9uPyOG4u6=6Lw@
zHt?jY0jNO+@vYBh`wknM+<HLLdOT9&)pCcFxZE~!LRYR|b#jUFa=$R)kH4_0cHpcA
zmy!<8`Tj69kC*N+jG@);c~Wl_vcf{QPjEdbM75=w#~EUY0deFS#nVVyJ452!$XP(z
z(i}t4oQbeIV(igyV-x4F6Hnidbed=+dC>GP@5#_KpW_YKQ-3p@l3u`XQ}7P69D>7%
zpKoQ5KHq@lXMxNroezK11HRd7YbCdt-gA}b-sRhu94?d7Mz0GK+zpIU49;h2IDqnu
zIW-A>IZp^n2_PLx8Qui!F<=~=_XMkC^lBV#9NH(3v+C&Bmd9CjbkIk1bZk24awx0K
z{6+zNt6AMms&#A5rjPpLthHuzxH|ZSi1M5x7hd&q+fDP#nlF7e8qUjjQNIWVhf4~`
zXM09o7(_Po-P3vK9l5nKJ=VHtR0Tovtok49*iV@Qg1qK;9u+H$N(94u%+}l`Rwg*D
zTe*Hh$EQUn;$R#A-0w>u<YDh{G)h(w5i1ZqFv^IZl2Y+hHj&Ybp#@Z|#WVs{Oh)9~
zUD9|t=fa3>gODXmN!-A7xxciXdZg<&B>mjQXnr#Ft9RIAQS**RJwvE(Z;3L@n;v<<
zdv<7>Yp;4e;MUI3#?dv-AucD82B;IuTW&K~*~CA!La4th>UYdvai2)=qqLrjOV9C5
zhdzEofS8CI376ksb5=M#&RT#7ce|4+Dul~1NLwa?7O45nCiv$#g{=FYC{f(`%#I|7
zd7b<Tixws}q8f5<B|@5u3ycv>KqoZ#xyH!#tOvF}KmN_#{nZqIHFM8Mph-04=qA^`
zbuJ=uj%uB9uPW}9!<dY20eehtXnc@)t0KK~farEIud1`J*7=COp~m>G0ouJMt0*g>
z=E3u-1fz-~?hpq9!S7&(k^R$5xo1AIMdJ~AvY`g`6M1~1JH1uws^jr-T_^@+NGqmd
zfPln%PwM*S8;f;+7?S7I8NUq`Hs$VT;t*z$9wS1YF}I4@Y7WNur<_1!i#N)2U~PCA
zM^bET4%82xQs6SPtNqvbGyN@%sg3m=azRu%Vcnj4>Ncf&8rf}XXe<q17Aqb>JxyBE
zDw&>OY<i<N4)XJ+Mn4+N$NcHLox6S*9%@h`giTRq<n~flH<sfLk=x6#rA@wStA)C3
zoSF<dht|T}-A?XwEvR6bCdJU?*0KjmGPYaXS`u>L86OT{^Y2WNjg+w`S=7V}SY}c7
zLiK=PSgQ2ehtLpJJ_a%?qtM2qGTF856Q0E<LZPIpqb=m|ef;Azdw;i6;fRw@JG<Q!
zD-8<k7^VY#t`!<t2{B72KgCw<gtHm|xOO+=ejl|PWs!VDAl<tx7U^fFKR+i$k5<W1
zidmDBlcXR$OP)N{0bjsZV?!s&TbAA7BCrWh4_#On8bGM%<FeZ!5$a)nNi`{-*~HJn
z`aFB@vwzS1{_h5nf<!UjyTr6clxkCYs14wLB{)&U(0?bLIc3*>QBF11#~_#sd$}K@
z=08T4(pmIn99wAzy0LG}m4l-VGvoPN_baB)g^m$ksXv|B!!(e0G@|z9!VKEEozM`Z
z<9ZY1T>C7Ukzo3>y$f3tnvDw?u8AFzDJM`ZOP3#fxw*y{ZyX$%ripgcY<8ei<fI@k
z{*<mK*wZdru3gF#Q$3Pgk9K#8oJD+oh@C?3>O7?xXcUZ9ifzJHQQKeb5yF~*ISo}w
zdRrRq+M>YDbii|8m#bp(-T12Okt|(KOwEXmI|z7+GHu@yu>+i3%P@OtO?+pO8wHr1
zSoG|(h~N8tSg`}K{3d7#RYA*fn>-U4d1|n0)Q#2^o-^2XA~ZDi8MgeC>*m^fCH98{
zS7^Jvciv<258fP>=3#==wT>~hu<jnk;g*epBlAqNw#2H)ljCDF+VoMPi7C-IAKLTn
zrh?}VCQJ$vHcC~qx@TK6kzn^P=DGoA^(rYmvsmf~?v$`oxC#*#g_a*a?D<g<^w+li
zN@j^_*BL%rZL+9ABcy7PrHRYzlwmvjoRraS0)JW-7p<RpJ5x2ezS4B;3rRprt=-JU
z21lnyAW3k7J|FLT%JDR-WG9+K{bpfvhh9N-EnM9b(4P=B+k%58X=)zz9<h|@5iM}~
z0Nq3rh{2eXvk&d@Cr+&>85!m98j26fCx}cfeAlwOkU`0Bm>qd6J|%RM=Z6^l1hhDI
zdfMQ_fzRCnZ`0OLuFB@qdW%!XxXc&utJ7ojqQ*EC(8E+((!G749rut&DJptF|6Kz;
zkkZ?SX(Z@2bq8DAD#sQ>WS{Kn#9os1zD~Rri)b25`m?PgFJ%T`YIj?K6EOAl0HN_^
z$-N!HQCwa+N9IR=B-}`wYMnis^>KB5bsAW7lZSMQ1|{Wwv~&RynZ%}<Qt<?<VoLb(
z6~rdix~SQ<*+--c6qV2PG@~&n1GQ{+Ha-sVluUFx0i(lnLob%xAAY@#S3tTRsfO0X
z=bN;dxYcC@GZu0&E;7Rn@u1(qlU^Is73*l=Lpr};d41<zdrq{u)vg|?2XUvcFxiM|
zx;KFkp}++>Rq-^`wJT;~pAQF2MkxlZ43TxBoJieqXRO=2`a++yFy7dXa*S+b%rl11
z%7vF!j$;jvdous?zlM)mN@wBq9-EoQ*==!h!&b!H6xRfp4fGs*ZE28f3@Rj@%OFiT
z`ow%L25||tb0i~CTx3Z){Tp!uwTV)qzo}i;jEPwW;MT!K1<ew_la=eIPvnh#6m<7Y
z8z|$E$u_18QetEQ+~P?!xm+Pnt|+XfOjB#r(X$m;q|71HS`ySe(%EMkbGA;8$6tc=
z8f(~$!_&$6Tw-p)9EJP>+fT)N<wZ}!nvPB$E4bCg808$EyevMEQw9TAhvfT2Fkbph
za=yuKX~>I9lM64|R@xVq9Aaqqe_kErL{M2`nvN9nE_oDBB{&Mz7jvp_j}y!GOV-f^
z(GJoTQpa_2Lor5KUqx3<z=ZHf+k&9UBv-{SpEOgL>p4)3%1`^to)dgu-TF;Sy3N*6
zN(olqEs`&#*Hpm$Xp5a&3@~pvb#u?)QB)VG&ow_@Xow?5W&gKr`cgQEPz9{tqs*lR
zwIg@T?(SPpg3+H_{;+W7^(Z%TNOmdFFo(Q%&75JVELqPlHg98Uqh5h-<EGieh*E`X
z{;ovKCxGu0*0wmFP~!eCyJU7>@9w3gQ=|q)8Uc@_^e;3(a-^z7(ncFF3<jt^1~t=~
z=gx1=qI*k+(7pn7$GH_&w5at2b^_8}NFWyH3h9sYL&H(D&B|3`F5Te{jiiHS5ioBM
z)tT|ad&%;tYG|_tLxUpPzd8v{`mQlf+aF2{ti6rC)HUHeH+qG6yZ;Iy;l*mC#QSY1
z7~r=2^1K(^Y(~N7Zi#$~KE1N_6AsP0Ow*NC-YY`n<w)!XN#?PZ2DTt1uAn~#Xemtx
zgrt)hbvW!Esbe{S*b2>}o2)*u8?uEh$x7$pVPQ#M&l}CQM$KPx%e^#*Q(PUI!&rbV
z3Y<J%Ft&UPvvJ4Sl-d05ploDdv3*&7;b2=xQ1-=q`<M9|QSv9V_(7n2(b9Gi`@&!R
z=F=dZI<tBw&_|U7`IhM9<T|3GyJQUX!o94X{OHD!FH>@JWkyTevj4tkzncB8;(_u>
z`{dN9hSEwzCi3m1uzLq_Vu_q2^5VtrG05g8w@jD1DBTT?hZq2N|Gwc2#Wb?gr-2#0
zQ}pHVFub#-HsT%He2rEq-nODwpdCLxO)%j!Zd>sDcu}H9z55z?*Z9N`*EA&Lac}Ee
ze&~|_Uev-%w4bMOf27)HMO)HpJNLAqid^Z0Hyk@1j0GAUwdr?~1_7g!k6sF-NKCVC
zH&a(_zRjFRtCmd0t?^`l-o@44hVGZ83jE5wBy6W>B!l33`3AT!7Mw^eUCP&3(fx5K
z8wmIg{kvTq1fnrk;c+_i6j#zBchri@!aIq;hx@Ce)?tGokSZ4g0-#w=p>=I_uNm=I
zMJ0FG*fCg32>_V5QVM%Pp9Olhx~7YH7&O*!l}n$=ym9-R1(yzSPEZ`Y`uak@+a%I6
zvT>`{K5HUslUT_Oe>JpK2aVag^}O`a$ZOHC>{9wkU;3=jR3|m41m@ar9CVU3?Suww
zlz!hnVr#i<qv<+hjM5oTG|#q0t<IPI`7Ia7h}yH}j4|KcEAHp;ZVFE_?}YJ?9N23C
z7*jAumKxxZk9$ekh*K*j0{OFpo(kIj3D)QT#wi{zOVjFobYi0WzedQxbCGt(L_{k=
z1IKEHW;1DNjqGK}rQ}CUtBzd-X62{Wb!)>Xo}St{;qRYXpJuPy+B3}vPooc{vQM&O
zH_qk1JnzTlSYav(mD#X0+kywnMHmO*uo}ho?Mmc$Ov}hf`4`^iQEsuZVpLGmPJ%E8
z9<*;Phsb9H)GbCPdTRHra^KC~i)~czYi(c;wI@pD?gr=8<4+{eirj+v7PiJ^F?^i#
zB&)RPD!)@upzwB=nIQdCeB?>KsgYKj+U*Q_;FjtW|E0=xF!t<j%cGj_MV0y?e%$(9
z-&(>{Q7{8e@;7?9lWLbBlhoezflbR(S=-rQ*)svX?~VG00}Y}4rMvMJE0C7EXy_E7
zH_8ugoTzHUi8!4b%@r0TPqQftH~UN0g%sR;M4Z7Qs}nG>Z^w;tHymzttwUfsH`kLi
zjGP@-omUWyvG#bdYITwO3dP;DCqcn?JM){_)r3pmJi2mOEfLnwH(Py~`T{|w;<eni
zU=x>a&E<!jdmDyK%$m2tyd1=9t7umks@l42CJ#+}H%`O?-BCD;iAJC$U{z`tq(>Jv
zh{g@jNt%y`t6P=Ym&XT`$rP2c${!?qyMyck&>i}xBMpW%m8X49s^ID0?M>&2LQD#5
z@fv4$tu1wSRV~i~o+nD9Kw3*mii*eJqcIFrL|dR)5%Kqr#x`4~io#;ykkQCk%PJke
zwpoZXO08U1`wFjZzrO|N4jK_~2&XglSly<8z=K^;iMS(gL!%~vTGS3vVNycaV$}bk
z?#-jpI`_8W^z^i8n!%<<GdP{Zs7;*WJj;19XdG!Z(I^UuI3wb$h-ljk&eAx~oYbf_
zio_^5f=wKVQ6!Ei2u4ASa|8tuHSbP)&a>WUJ^y^`d%yQv=h@4(Hk;)itbO16-q-#6
zU4u^`w?FlKb#o5Drx~A_7|4V}Y%O-j^px`lymlui5~n}hW%cC2f==jWCrx~+*{c}N
z4LT(o4*;QpPLnA>RX%<HN+T@}Th`bRO?<<~-!Ys>>>fC6d$1ug3qeLUMKp#BgX!*^
zTN9x^4#gt}7iw#=iq0iFnjgE}6<9iJ-&MDu6BYShi<P(h@lb7d)H{mzXl*?fc6zW!
z>TVn6150}Q5~d(w7T0J8i{*-7W@KBGr!YrwGh<=#!(^a`R_pHJNh!Xpl^MnGnQbk@
za!KT5b(6)FwTh6IfH4U;Ei;raJ;Ph^wPK9?U^|}>9IswwVU-q5{FeWIDWdaXqgmnB
z`w4bJy}`Kxn~8>ywL(SVwJ)Nv?i0F|rz2lX#$6T;+BigfsQwD9JeGKV)9ZbFuVT%m
zElK+_tg_sw3{fqr96sIlUWV8hRGcG6vCzui6Pp&29h)4_tfxaB@qr%?m7Po>?P6KG
zaZbrzJj4FlPBt!oTtw<$5ZP)QkQ{e<FpTvpYhy5*E`TMY=6QA0;Kq;teAJJCmq1yt
zcuSC}Hf*&0+gkNaO<Ut&dXEpuM$G)pYsx+A1szd0>w-o#2UjlVBkV@mo<mdbo}fKW
zqpcMoBkO@_wv;5f-NP&U@fLcR24;P{OEHB<iFvc~VTOy%M7IPAVtHRewVum}t69@%
ztV-oI{$#{*)Vbu1B%hqBWg=?n<00)YE4PHoP>{l2-T+-3i@13}^A-h9jc%%9Nt1)l
zwo4NKp2FljR8Ah|?-iHVzys%sGy?*c7j<V}J#2{`{vHt5)7q0B!kn9`>V;h`22*FN
zE1!<KJwbJNg;*%nm*%qIVM6!8bItdzF}s_inRxpH+0eF&`F3;-VsEHvU|Uybb2^4p
zC&2Etx|Xd90b4e2so@7N>6(vuy#={>c@q2hyWLd6zWs|m0`!{B_ORQ5w$MVkK$hc^
zw*^pic505K<4=08jk#+OY-eHywn{2272f-PUlon?F6blmUB11x{@v|QMOkf!uk#T<
z5L}Ji>m3Oe)s+<vnx&4pE<i!{-WfY{$pB7Mx#Ws~mgE%)of6IuK-@L0^IIl&gl;2>
z)l0#pEOHtNDKCHWN>^-)w{|E?7joSNxK>P0D=sd^%gd|X0fewAgECvSzyEWD|EHY(
z2aqexgUV$mUvh#*!`n|+N}XKhDV9vGgB}Dq`1W~SZbz-F2|W9EzXeG<HqXe$Jn{X#
zSlO{kZ^1XXp6OdcPg{Iq(DLDDGY5K2ON<{Dtx_$@*hbZ_mi(t18@1WfN<KL=GUGPF
zxmi&ZN(iwmy;Ca{BwAMnRm|;*D_Z{qJjCLxPA|W_+!57e=&{H}y?!rT1MmhfLGEX3
z+*SoP&kCASnBDt$6<J#{K4>d;Va8vx5;Q%Qeek&d+sQ1Tco!L+Gl+^OS<jsr83;D)
zD+d3aI#5cR>6GiW50rp2f)0Wzw8Q;CHx4!!;_0W(2jviNZ2T8V;KgOJMY`Sa+@_q7
zC}tOm!*Se&j7vGz2Y+mq13}a9!A|tPmvM^1ZFJA`B+j%8YvXbIV!oN`tr?;Xzgq_>
zEz80Uh$k-F5eM3=<14H^Iv>@#8I;Z*`SC^nyl&DSp`p%6P4n*V;Tp=}QR-p2J6kKi
z0sG|pFW2yGA3DsplSPH34D7})tWD&wd*Z~d|13Ywv^6@}BAT208+$0ayc^VxJU~2V
z^T`{FVon2jV6}mPFM5P`zGHCxVDZE`cpq#v<eRyP2;q-^J`|q~blJb!Yi3{?&2kbT
zIjtwY=NdL&HBFo>`nh|UPwCaltw6}yb%Muhv-`a5p61Oi2u*Y&6LyVUpu6kb2hFz%
z^VTaiY7(PgBIA-J(>t!!%q^byz_WsHva=5>nMb;WnKBaDh-fqN1aowv5lJcQg)@8d
zG=Q9`1gch^fV}xW(h?YW0c-}be2<z``6tZDN?2K>3xS*|npy?M!?xZbCUrW4CBk1s
z*{t%v3Td0B^ew)%$}sde>+-Cn+ui-0neEMN?X~zL3)YXHV6#WpOwsEE?ZnmAV>bLs
zH$QFW1}X0m??A>pP)%emaDI~(gGvuaO-zl=ErGoIX<tmt?^xJ;JXBq!No87<r8g>%
zix=n2A|OMrKVckF8GvzYHag(D(1c_o4Z~IT2Yev+<DtxVJ=^J-yz#~%ZlTDf!4p^g
ze0OuAk`jB(srDbSy!4g>-O0V*>~>w^0Yo)+R+bqWS!;iu6eii$kMnKTx|rwU>QhN8
zd;@cGyi%os4h9Tx=DNojo>-@CTxGZawo!XDwzHsQK!S<FfR<IQR3SHrgq}^yZGbw*
zDr~t?>z+9_&MdR1M$!dHL)gzNNvFNWvsI5~P4Ix&3bWZW1_$iY3EW9+MJ8Qr8g(73
zuG#NXl(u`LD<Ku6+cR_Mkm{H23Yo)j)Q9PY*hlb_`7OkId3^yoYr-=#+1V2ozj3@^
zQ=BmZB0ePAE@a7*+*a+D9U2YRZ~zPmUIKWA%@dO(EpcSy)>tDra6Qf_f*?HM$A8^v
zURk|IGModLTZGL8S|Ydb)2e+^KdX52)6(8Frl(Utwk{UBwbF76aeplATylCb@IQ*_
zn+Cj1B7k@*Z2u9by@2icQLAnO3hsQ^H`}vpJ{>h<RZ77Rr`S2=T=tpn@hof!s6{%G
z?-%?In4*Bag^h&bwtc;@+}$JMirwiY+)NwTO&fN{GDoc_RS*qvujXu-tjtr_pd&sO
zi-o1v*}9uz8&{|3z=@i#qgdwMmSj+C;>n5jjWW_J=j}#2F4AOzCITFJ@x-iB`A{sI
zk=Ec^=zOM59!TgU@48&WRLr?^Y%^zOVbO;bIf8_FZdZu6$V}O8BZ7DIS@f-kmlX`p
zmI1DDg-^T%Pq7?Qjwo#s)%eT_^!wY+>LUo3((bF&E7-$3_xp&1QL8wW6C-(Qon|*y
zH&gGg6(bcSe85-IM;eRn@N2-lSdHE1dVYCs2+Pw{8XCPI3EN_mx7g4<qDokYLQ5YS
zDHq372^P&aCy}b7Lg|Z3-r;^sO0DE%@NQMbcc=cdUz(d-l<*?nv&v+1hkBXM&MwPB
zYCD)DpW)AD%_Oj@yPwHVJbCn-h(VN$F4h^3F>J8vCPbG=*ogaijeQHaT>@dH^(FBi
zv)?3$<7BDseZAtL@fP?77R-;RsA*4f3CT-=s~+pH-g3JDL`S7u%gee0WFL}AHTr-J
zLtSDCy8~I5naxKw(cIviUZL$8;Rq3Lbf(U%(boarkE|^^+>zXxNgF{y^JLe3%~$aQ
zmZ=uEfQjJVB~fh3L%(r*y{Q?vn{vtKP7XMi^@{zuuEP_X#QgZxKw?11ktHv7kkM(i
z8t_9f^k9hSgVm93m@VxiRt|svQe=ebIH=z4tjVDyvHLgN%RA4uE4gX5t>iT#qog7k
z3a_tLg%cqFywgZ=f1X5u`UMf9JjKlZ^_v6z0bTo`P`+PFjT$Hm$Yl!k9CB-a!Co=i
zz`f=P#foU3+uN1k(~n6{u)r7V!O$t#w69wsiNvW`XxT=t6T-h^Xn3DKbs6kPvRJ4r
z7daTKD@TT{ge`>yjm<C_b*=Sfn$+ROVVz=A6VY4?uv6GEoGghD^yQKP>mzN*=tL>&
zf}fv0)H;9A(Z4UzpO#CaS{JXK2Eb7|p+0eFa^h2zmb^hQ-e7CDi~0HL{}lZ5$M5`~
zK#V^8b#ynpO}oL{g>fNTa%pMC^EZfWDa2pgMbfs)BGAtzi<9CyQ>(w0Qt$@nQaqiG
zHu7GeLxobh-SE@%3h1RppiG<C%$Q$MEss4B-4^X-lZSOxbChl_%~L1^EWQLK+%m5?
z(=C`y$zzR=(NHIowoA~;44a-<julFjR7g~+hZ&qTjp{=JKqyG{*bnRU<|5kUl@;DV
zAzWgN|GGwk0m#yAQrM06DXK}w&1JIhqKSIE5RWdL_!bp(llRr_x{zyct`sh+6wd0S
z;BMt3Re5ib%TA&|C4gfR^d;zz@(*2-naspXE4K)pydJfOwGjLo^-T}}?)#c+nGtuz
z+$O9mt?)%Qwr5^-67U?-x1Uz401sN%n3w$%7LoqwA#@kx&l=ZXZV&~5qQUT11lMj6
z>N<l4$56KiO;Hg;oPf*T5EqZo=d^nbi={rvhvn>^itLz2?>E!t!xnJAj=ssNbZh?X
zEyetOa#6$?pkh|xbbnb{u{%8n^kAf?&PVFk<i+AL?o?FT1Nuf=fbDV~9~5nU5EmwC
z8LEocrwjMQ9Er5j9+k`P#<R&Hh|XxUZ8G%S&|cS>UsIw}zlfZ#o%6nqRlZKB{Fm#y
zg?*a2QxWE~4&Kh<gPqxk1(n%LXY80hV^`v*JD0w@`ZRypU;p6S{4Ik!We(3AOvFK#
zycr6(>N8Ag=My`x+K1;kGKu3B`={wO6-mnnto@JU=~-QPo7af<2(}g(>5D{Fgz{%N
zbcJH($aoEw{s|nt%oTkDU?cT>ff&F>=RRR0@XR})1jNjcrja(*(i?Sdi{z8qHH(+j
zAlxQZzcG~yV!t@+?g1qRrv0!;_nulLdh7UnSV5t^A3=Y7d`E?uTjePEB7KW-aZ<8F
zxiEIGba|54^_<fb?}l|fzmJu9B~bLK+dyz}4oM30vkjCJ+@INwUf#^d6N2DBEjNWJ
zKiVi|8Lq5}q;0SJN-H@zhf&A7hH`+7tDMr=WY@)Lu8agS1ym?)Lj?nsDMw0?LRH<=
z({m;RIM))v1m2iU3-*=~EAFmHZW}+d#xfhV#&(+T6rNH?UxW0&a;}LJu<}EGc#X8u
z;7c&7(*=e3aPR^$PrKl?tot%detn_v=Dopjq6fuc`+jA2z_vaBjk*h4_*oVfzIEhU
zHJt9R%*lE4E`=T-@Gd$b(-nWZwzhs<lv1!>Qj$BP=KF4U@@en~pFpA#ajT38_`j79
z?uh%uar$Oc(}O3sq-AHL{5Q8G3AhtzNGth|rN|mn?*x%S%9K_@PZxUTYkPC1IU}`^
zh<HS#YpACp({#JKvLI1m32bwDJaoXsGG9>YuuyuZ|0KVs%#<1iIVX*}%eb#$beHi+
z(pPpO%rlcWn5m)JO3PDOf$#wOWqoj_ho7H@D0Fh0r0>lw0%(}53xFH7sxHJMsjnX3
z9(_7CK|P;yro@<CUzS{fn0xU#zFxTrv$6i9h?jJm#5gN3l;5p+z?M73h|KBrcCbE3
ze#SDr=3H;$oyVCSD5zU>OiBa5P{HonlGaQcbg@a5xW!2+dKA;?DhTr>I(bvoX=i8P
zio@#&APSN;Wzx>Nq3vTS9wTN>4#jO;>nQS*cKQTp5rpv~R2gla(6-u&CzkK|vsEwm
z9#E`H^8{FCR^C?W{dmt;a-CRZ{H1cftw}v))8@f;sKd~n`_Z+|8`S|ifmy9q#<^aQ
zFmL^un`5~om!6|(ec*Waxuz1Iry7MVQo@!J5B?UEc5XRxAJH<_DACOrCBLWiM?_E_
z$rRH>nHi*JKF{iy%dcV@R6Zp-{0dXBwzM=fY}q%=S&ByiG$~W@gZ0f8k-=ajb)WEE
zEmK6kaLH#kS)JAq_k5nVjt|#q2c5cG{etu3Ki`T+{<74XE{V3(PeC|6K*P^NuGdCy
zTAWbhE)ZKcJzV7{vJrLulPJCS%*jU7Gc*`oG#cgIAGI=au|OF6-QYQQJBAPS+x??E
z_nGwM{ju04vGzTwH)ocZL_OUWib1UQnJz@!rP_Kbel5*9=OZTSKFQ=D{Kg3aeTD8@
z+w$a?nSva(+^(e?6o8Iz!}bXW8wt23#NY}HG`V%KHV<afbsch%HWGf2hq12!FHZTz
ztiQQ)dzq=>)Bo^kv{yZ=Wc@T9%2)t%+3Q<`-)2L8X{+=uGWjlir&CAq@`G{0a(Kz*
zgJ}9g6d)X9c-fd&45LQr8Ht?}kx4@%RZo8IipKyYKvQwRJKZA9ZY$x63C*YcOK$Qt
z)J2F!g04@>&3qesCbgM9tn2#HVI}F|o~J<2wKyLQTaq`K$(Y=&ToQ#e3M}XPU?;86
zT0;#c_lM_9lWfiIO^VJC?tfO&{N!=Z;+2x4&SW)hTzjqplGxO4^)*w`QyhPa_jSG=
z0WtB!<L0ZxXu8h5Gd!zbc#@rpwML_Ykr|$auCmmmHR0eI#I#UcGpy|rwXzl(Hl~VA
z+F6tLDQ<?>IC^X9-YJGzTRdO&qv2tC);ory`kAK61%P3*t|}r4eaW<XB^W5R0!+SF
z(24P=e~tV<C&0h;>EAv%=G}VID)BGd@s7mT*q;`#Mj0ea=z$YrtzO%eD=~^ifoE^^
zub68m_yma?dcTuS6dfh-4BNf)=Zv>j1-V`|4^x&H)yco^wlm->wq-Hv02ne8L-e{&
z)`qniNASB}l=%?P=SD`(*cNq*v@rSX@i9V|$MpQzNn4P~lsUA$4!Itk3+#8{Lmo1V
z3$25pv=SQeQB#=ABFVnE@4l*Yp4y#;S9LZ3iOjI58=zIl@m~As@-|kfIRJ^Qwd0M;
zegOzKYTob7JhK9Fep|>>E8uPzJ|<p$8sZ;gWhva7PTxAZS~C7pqSv!|a+)g3wf&P<
zA-p$^d-d^9+@j2XOm#TLz$h~aOuRcLMQd+Nta@!SCQbg}+_wH9Tv+g2`2>qB**XUX
z8{@paVM}ZhMAdhYFOuhJR%m^A22dPC-G*ZAI4VCpichEIxB8?$^KM_GnZy^hk;vH&
z&jxe)z9)cQ&Rr3UT+D#eB>7C{S@reRRl?G%Zw+N5SnK!P;t0!q_}F^g%Wc)eklwMJ
zcUxCN;PgW6Qxz3m#VcnelBnzUj9KnrQ+rdeB#4naW@@(%DOqUebZ&^0AI)mCrWjZN
z*hse?W)w_&)NJ5vUT<VObEn?<7OpzvB*<0M17-Zr2S&NRCrN2Y<soO)ocF+-9krms
z);JsZgt%1fG(+boCEUs_aL#e8CYiJDH7SnW%0<i!Ln%3$ukyLOV&TI}ft9D?(c6cg
zCI<Z;CI(C$SP8aIeqZjL+$*wreRt$#5_Osq8;i)DwZg>fc)-7WRA`e;FSBp{@>o%E
zYIR@44AH9H<%AlJ#g8|{BJO)07qYYk7oz-2dzp%trRh{3a?NG?Nl63Nqhd*}`DEGa
zy~kaW?q8^XZ^;UVV*j#}r-i_j)_v9QjsLte%0HOnXK+xmQ2OZF=_)TlRV2!t%afF2
zd%NA!z_8>=-{UZV5SyCQEOjzY8d9SlZC?&#z9sU&T@g0!0zb~Q$SbRZjlFOCz;6H+
z$b?PPKo#n{DLwsRc<t!+jiihN<i|tmk!bCtv0ZFgc`A9TY2VLpXRE%9<0Wz#ihJ1D
z_Ibn1;fab1sj9J}jF@%piAz@>FijKt^}Nlb^<{~X)lTZoLF=_0Q>G1FEmh}!7+fv!
zX><)uz9s0)wg@OZ6SJ*8f%<_0U3?LTa_hud%vfX&-|WK|vddQ9`wgJ5+X<Cs$L{eG
zu$Q`SRM}^YkBw(0$K1<w9vTRjCw{0vxf&<f8hK_<;^adS@H8sI!`jXzOI#K-J;+-U
z#Toh!7;Kd#ME^Q#^DjZzXM4A5HN3Q{Jsg0{jf>;U?sYoN18xorKx5Q84{)uxINEOr
zFctv6O86(j0z(F26%qh{#_bkwaswR+;2U6@ngCSn0e~ZtpNDUEYhD;JFaXjuG__Xb
z<#zyX_4b1c@j<^|dx`}xDPqh-+QAr<%B4h0N%LWVdqj27!0m09O{G*t7n_{syu#|l
zf^cH^b}1|HWo~DkQdUl)Xplp<-T$l1&L(LKzU~9E8nA!H(S|I{@7<pfiw#BI-YuJI
z`VZDNeRCjgh@6(b`-in%<FhAvU`0TiP_&-;AMAaFWfi<&b@}c-E~-cZ*+>es^z_to
z*p-pnW{sHxf+UuNNv=sM#K$=^_Y}TZXZ$=H-vJg<2P)sf2fO|Qm)%Dj!K8i1Hw{w-
zDS;<Iee2$<y8f&!T;w$l?L4Nq58Uv5d))^z#MaSJHqZ=vuSJXhhb;v=n91u@j%i$L
zKC^l7s*~&zuJy|_t2JY<QZBE2a;{xEn8%SIu_|wb(+2K~VUd=PZC__rbdEyj#*@SQ
zvHEYipys6R%aM`TEMuA~Ky@XqWmkX{HqEF$usA|3+oilx^9Q|>djEgD`>bA$>z~Y-
zjfppK6cP8@OM4IsdIv_jeX>(yNM#7V61^X9@OlaDm^XEY3q=E-Gwbc9M*|;%eru(F
zpE~(mm9%#y@zUmJ)k(HbV<^Yhz1wW<ijceu<~g%84{hElL^P|Qak&4ODt4-E-K{HC
zVPkK#<LvEJkBy4j+JMZ6@){p+5h@O91=_96W*FHbH^v8y9HFy-)_Yh5uXdAuG%}{1
zb$i9$z#)B17Siam-JIO^@Tfzpb5e;-^CKg(=6=U3^=@JocQNek!R4-lD0WbjwChVl
zme;x=rq42}|H`XHqFMmz8AUbTf?<bthCYk!N-;7?&@GvmlW}t>ekkL{W9<eJoJ>Bi
zp?85tPB?Hv=`dAmVU(H+{*oVdeq7mWfKROMidSS~Yeeqm9=%b^WU&`R*(Q)%019Vy
zG>@LD;)l$OQZySgLz+D%W!bLL*x}8Tkj~@8{&%JvE33?DQUeFyRf+4HH)?})u0kDI
z3U#-u&xzw>mtGMp^Ld~~hbPBi(dBX$lNLzs;TOOv2VFFu=~jO{*G{!UWh}%P$q*@d
zoq(6i%hvj$=|rf(VviGD7l>0Pr<aN9+UvRrnbGzLSiHLpcE^+ilw_H7u#xJj*K^j6
z&~wNcU0nmH7I@$8#^<m8_^-kL_dox;Tdn_vNYy*AjoU{|$3b6g4c8nF5$q?=e>`+#
zV1|<4LS}?2yOcipy0G<JV{KZSq>Vy*{6Ui3wzzI|3Kg?&dN6wK;M(aHKuqMftxVce
z*e@@^-kNXgzP%q!)aZ>-byDm;KfdcxnbZ|;Hq&B=iS`p%WY0k9RQdY#xlO5=YQcx-
zBmp&qPii&KvFwVQHBrI-u-#~R&`*OdN2kJzUn1uCXy=wjEY!#(#gxUGA}p#-9%!Sb
z;=$7s<CJ7OXNX0usg3t-A;VDEZ5=-CYA;B>K4Sq&9HKA&0@$T&AEap8oK*+)OtG2W
zMDH_uF8R%vtZ;QTjYj>`A-Y!9*a{Nia!df@wDWt8)fn}2aQzmv`BkhfrAlo%T<a$u
zsf}X^1nU{W^=!goDk_FVYG#P|&}lM&WSUnO2TR3Wy_Mhsemh~PY>#}vRBta#0l`tG
z&QeWv2DzYYNM_*9$V#YFpyg1ta|+VZa!hWctp7Q0l}WGm<c2sIoYjQI`{s<u1zqO(
zvw_t=jy-QJ2^EvaZWZQBbx8EHwucWtYAr=48<~`LKJT4&P(!0JSIad^Pr)$T@>PlO
z-^1c?-(z4$#?4hw2Fdp2voC9Ldch6cj*bLk$Q0k~MPCp<IQZW5%v1{|;_g658XlZJ
zhfYq3DpfX9YaIyp-|bg%HXRIQ4qfLD2IB&iZ<p-tO%nj*MA*2qOSmzr7O19y7TSqs
zPh68;et_-cCupa=h|!C3|9A-Q{w_Y*wKO{kTbn;hM?UU-U+87dd}4i>7tgjFl{%&A
z!>E%;9}srs<DnD#K^rBYu$OypJ8Np3y-9lrNzUhWeVeO`BKfnityU+nH))A7(~k8-
za7nJpwK<-~fyZL>tn~ISOlWpGmplM^ATyP!DbirvS=9<dbh-!^mrDXLeDhBolxoQ4
zB5bGLb)MpgYp<j{B4pGe<BQs#xElMvUM+dE*~UOx;9}2=;f-GS3roCTpdH+VXpINY
zds=^^*6pS*&)O6<CqACB7@iyAOCRHcie|_@mUmu2W|2_+Ys!ASwrqXCbH1u?c{$W$
zw!+82$S}7{3pST1sE9FlHS^4#z^N_+hAy+Z^{POxDnt_a34mBc^9OR_OQs(L+kX-N
z@2k}}(;WQgt~iPM0`^zt_qao%FYRY%_T|l!X=<tDH~jj<+8x5!-z71zy0x=DV3U5k
z1J>+kWd9&sfp25EeBN7diyXk|zeZi?T>upLrS=Y!(u8imidlFNMz4}!?aBXOlgCL8
zN)yA=`s#^A`Wii<lOUti76KTx6<;gvE^G3Iy)Gy0PwqYGb1SaTyC}uSadBxA)1nZG
zKLv3gago6c`~(YrA|_a(gw&Cc->_;sDQX+*$>~WT_%B1g^P4SI#vczUbVPnU^edTN
zFDw0)9{sqR+&X1qjWkDluoiJc^!R_^gSGyl+O|(iqXQ%~qwuw6;bw9lhT7j)s6wyb
zAr8F5XmymE>5B}#4MJBI2Kz-;9{8cT`ID&DM5`IgTKZHCae=C>R+jwzd|eV5P)ShN
zC~EQ~!hCl{?sg;ACvs=?jgovx=->Z&cMoF{4d!*HbAF6E-Z=e=U9g5V6{$NeV&>>S
zi-n?WGuh2O8>`;}uGq?E=E-MR>>sMKW<KOjhqbkb3?g;c`Y%xJPo}8a{_Wrl<<idM
z6|c5r$;)@N;guEGY3PP>_^~rqFbG4<aEiM0T4p0FM-2~R);>VQjJ|M}TAg=&ojA{U
z@8%`W#`CT0<<(8AcyhJyMr%TYgtD{0d=Lf^%t;bxc=nwvOaiv-%pL|S?h7k^5j;0i
zxe}?5?bs(oGfsKb!yU!d&Z-+v`Qd9o9I*wrX&kv6NYdep-{v}QR%y`5_D2f}9};xL
zlwbM+{aap?imS65iTH1z_v1mSU^g_(wX@K2k@_|Bmv<KXVHOdalqzBvf57(y0@vt8
zYU+3L8?2|)S{Tme@VDcQb<3>p*7S!gj&;VL<s{f7BeF`KLW52SD%K!DZ)cu*M29Yl
zchbdu^?8b5B;52UZ&0C@8Umeb$awa@Pw9W}&;RbywrkAnZcLPo@=_tN^}ag9NLy45
zdwpwuScBR^_!_Mnj$vPJEyMqfm55-^{7_%JVmMj-=r!A}Ws1YJsrB(B6GN2J<)hyz
z%`HOa4>Mh@U`_9t@nZ)M39hp<9|CEn6AOs@Hdj9QBFxIs2Hf&lZDXpqG59Gm!aCjp
zr|N^QQYHk|a~N8#%V=VCVPv#2)q$5VIGN<nESp5OQeeY`ddrRjxT2#}ib0c*!KIjz
z77v2QJi&W=d4dR~IIMUOywT5Y6Fi=dR91VFw6ey=>)p^*HV%tffu$i;#v@sQSrF&Z
z{60rG4fx9`sIiV|@F-pal5g?}e%jt1O((s3um5rU?cxxSkW=!<Au=-BE-Aq`3^67H
zQ(}2r3A@Rt*4KZ|wG*&8+s_1zv~{sh^h;}adsO8T=6$J;CjL?T8*mWP#v#OBqlA*{
z71?&o<gB3Zv{%=fz)C*otKCLo@W6Jc7Ihc*c3<zEkr?;AEa>pKxo%1f;4Ns@o#|%8
z&v4;+Ui+dZPO7q3kS?`UPCgEpkx1Jj{96T#y55LSd=*#FLms(Zw{r)dSD!JhnP0Lg
zGUQsrUuI}6g-sK23Lquu1%MtP*L*w{!pE*ql~X^&d^~iMg8S>HqH6`B?#Y5K_(!gV
zZRl;?->|ouA^wq;DiyO`wZ_`;_?5(8gpbtrlVx5_?i&=$+-yTXCVBe4^hvz0ll!S;
zV;SwU@n5$F^M;%QCT3Q>KFIXtT^ieSo40d)L38tC!d9yq$%{YR8eNag_LJ$u_PvuJ
zOrBzazQDFkQm=jIgS}6ienX-QQ6=KE!qUubpbuF0%x~M-Mafd?RC;QEyhJxu-_7hS
zp$zB){@`iS<R1m}0bhMP*;V(|wLv-6C%gFN)Ah;+P{uL{1rValie?qqf_QzrYK;<m
z*0L+Gz!sV+Uv+wKIXJSkqcmZ0XzYu?KsWHCcM0K?I$u@!>~UT!Vm?T<@w=z|7W=dT
zypF6nEMRl1oPE}Cp(S})C?I{c{ri7iRZi-2@Lvr75)TI&ETDf3_WE=`mPpiYte>v}
zns0_@FH>V}^`n?i5bn3Q>d>xBb_X(}j7KBhEc~mf)t%SijewHmpOfNtKOTCoKhs7v
zYhuJKSyG2!&l^{;m7ciC3F4=XYHs<2z+A2foWdb(#cQS03TMsW>Fi=L{lywLFXn(F
ziZfbDc3OzS5Tx#z83@5ZZW5ytb(kF6YBR<Y!1VOHSWRm+!ywz+!29#Z)OgDRA$6j9
z<-;oD8{Yo@keRDSAIy{8WWEfIzRC-qZyTphH#IvrI<}e>CDM^&Fq*epc2Z#?gTZks
z{dnj<l1FfVrv%E1<QP%zOZKtT(uCTd*D;AZJ1(619bCaH@eaTpCPAR%KovmO_2^?3
zJq4%&Ebh+$orU*A)QXx}ko#ZbRP`#65~`7^6VmT#N&u-}O76EvOD7SCRtcBxNc`zw
z+Y2WzyVp5B-Fh})W7)D;THzS0QR=|2bf~&01Yi8{VASQ}+pK0M<MYPNR{@FT87&S#
z@&&Z!iO{A{R<k0ZK3H<2-bAGMNptx}s7HxtyG34J-T*jkKdCzmfXc}G4+m<fe~zjA
zPw(VrvKXGS*{7SF{k5SP7%Ov6|7iG`v|9uliteCpiphx!EaciHX4TaiPo;MLrR9;%
z_H6IC#NmyW@#?FD86<Y*kj$QPVq_aN$E#f3-QB4o@BRm|o4eyOKv37%v9Spr;3Q<V
z*Qz=fjIrL4P=k9}KR7Kew4Jz6G~f&})lIOGUqdHuM30Y9)}&T1U~(r(nL{!-O3#6K
zx>-Tu(Ojg;;Xybk0~aj4B&`hNeZ}Mv_`lA`|4Ub&^_SNV&Dg(8Bw=Suq((R9DR$hD
zI`H!inXA&RzVWt)7S9F)5V0j)&AF2E-d=$Z$(D@SSzqu{H4^3VbkyNoeeG8G;<B7U
z=8x|IB8d0^p>w0YKXzChQ#{jX(c%#;=&8DI^Bmz8Xd6D9Qbf$4B8xP%<%301JxC^y
zmDWBOW$6=62%<5#(>;V7B#-d|5X0UOg2m=b>y+AFk7lE}FO@$YTD!cZlo7u1q3d?e
zf?i)B(k5ofKV~g4r{c+Yd-d6;bypkd-9Go4#|Ct7Ci+u*mb|*Kz0b1|h-c$)*Tj-9
zwq{oeJK5`JVq0a87(Z<FTr(pxCl@WOL=_orFA_@mwF=D0jmU|X$s{%`zVhuf&DLO{
z$GLKObT~TY8x`$r(q?1}AH7Nvoy>FpYv__o)n@uMhpaQEnsRi0obF^lNiGJAD^J*q
z>trYM>f=vbT;4bq5+?#%>j?O|wQ>CV(pyz$UOCj$d<lIqp&k#Y?snh07nW-ZrM0<g
zw+(i?r#>E9Dy+W3h}n*&M7R5CTksk-mE~^tVG=~4yd1Dv#DF7u7fq_+f?SYWnL*u~
z8pHguWapJ~*XqC^wqz}P`09vS6-TI*jk5*)tj>5}<SDEiWv1(04W@^v9(r}v2>r?#
zrs#ip@GG=*_nz`3uLwW3Ai~}nDI1g#K^5^Pwr_|JD}vMLd!uFtmDX$G{ZMBHj<)|F
zFPH4`{mG9{?x~(y2dAliJd`qXaJ^>_uLStMRkPnn0IpICdtN(@c(umeX9YTTwR}{z
zl0;KlBSQ~h1Y>niUw-+;t=cU0#fw7h^=7GEXzaxY;{uUX|3%7NX1uE7a#ZEB$g`Rt
zBfX-Kx#+$2efd-T-Nt<s4CLf7>xDCb0A6P~VSsO^Na5G0yDKsSwd|-(*=>paa{9~H
zcZ*B*RYCKNck4uXnTL{sb|(ZHy=<zu+n;j1wVkWQN38mCYV+@HgnBQRJ(4p3B!Uy7
z-S2MCY*N!+sn)0O9UzX+P-&H>J5hoy*<{74l7htb1gJe0iygA)>Xm61mCRWj*Ir3D
z7}PM=CB}f$XfQSJ)~ua8h9Uu-LwvM_S6!V6+}~t%HzEk6R2|q>LOP4alZ@?Z1`(32
zla+0LX>abaFVfpF3Zs-3Se~1`+f>-`#W~Vtw8I%U4&g=nx=Eoyy>ZKL-*?We>94C>
zH&q1~t6N{ZulfvIF?$MHIM1NSj`l3e9VaT1T|LK;U7Xx4LK36wgGH?Q!Gyu7bgFZ4
zf(~oyS90d6{TL_C^5da~{iDyl+fpvgKAP@M%Tg9tzV98{*&kkr3Vgd41c)A&FyJ|d
z-JN2cK(JCNZ93q?^igH9wg$pZqLSC|%bw*47305k^5Cw&5l>b_5cQ#3i-pxn8496p
z^%vh(ItYNMr7J}!cf_?IxDy5fcZ&AZb5`;$ONc_Ot_1ak)c9DPG4d#hSfB-$@@csR
z6kT~mfO_t$gEkP*oZDWw>*<hy)H%Sh4xp{033Z<cS*81apW2TN;MB^M_W!fR?%&+V
z{8YI3w{|(3$7^6n!mpIsdatp>ee~;u`mfw`VbPU*svV2ewauS6FNb!YkK6+jMrtpL
zTr}h!)R}zeWYTws@mzoU^T~uY*i9Svb@LB93+P$5h}a4`F1A(>$?<H6R3@rC(l%Nn
zzL{S4sa`WlBZDjMbg>LZwND&ejBf@)`3|2&pqUfz)+jlZ>Puz6%uX{IC{UKHt{*{5
z-?b+vB`<SB-wl>p{P@Q(SRfv>sTIrAUZ0`ryrWvMEUqOFj@3mX8bjNbHp0J~dRu2#
zFbUC)H6Cx;(NQUI%0x$<otY4T<ZzNLI0k!F!9QypmJHO7c6}$svXtL^tJ0Qzqa7z`
zW%+=K?R^EYp@*cWw)W8lnRV*<$~K{EtJ;(M2u#od_X>hv3Zd2ZMsyKD6kPUPiz!-V
zrpd_XT$@0galIF!=LF75h!{n}lEWrx8I|zaMCn6<<+arwx>~MT<&IFTM(rtd|1Sxd
zT3^v*rTEXfqLPB8*JSl!Ia=46argtw+3x>HO##YA!5<HK#%`o0H!m*b9Tr_%@4$n{
zfX0=I)9m6_zppR155I*m-Q~Y>1TyYYaf^_OG0gU-N>QHd0KmI62{F?V--C+{<sM(5
z9p0{TXyG;fb$-n%V`~K#Z#`Avbn@z}q-^@-i0Wg77%k(=3u5Qqt+^uz0r}`&A;A1k
z?S)R@7{_cAe_9So+4jWKr}uBH%`8SN8e!(m(1e*XgZd1yWSFQ@r%Kn;w<GWJ-tuBe
z#!V@kI;R};)b7vi9#qGeL#^DiCy43g-fi^`1Hm{oBU^BzZ&tIK1HDpe$d+jjmfSL$
z{O=Ez-Qn1z^OjCrOSJRs`Aw3_=>j6Z-o`d?zAlWUvQFZ^6%{T978X?WQz%!BoE$5>
z%W_WZ=hlX3^}r2_+iM<%1l0ax_8ga*1cJEZR&1W_r`7_hyVe`)YM|1U@t5-3eSI#X
zdH(Jepv`mqP`2ycN)H}vt!qDNTo%iw=q{`D8wPUUX6w0b8Y#Cbtxj;t7-GoDQ{>|e
z(9IH7PtiqBdL*B=_@+Qr)AodQ`y9iO9sOEZdnPLWWYCLl%@vDM%$G_x`EMq)>MrI&
zxRG`*i}41<b=SB~N`sA{ZVoaG8GzR+lYTucZ&#UcaEfZ)0qflse(>gi$lH1=w59!(
z!jV=b3S-;RgH#65Ji+T<vMbDv!UXZ>R!e72gh>NC%9Df^U6^0Nzu6W<Rh-anO?JxR
zCv;8QLXS1!74hedjJ6tGjcmOJ3lRdF$q_0i1jaVpD?RvTxm^3!pLbkR+MlBSEUZI3
zlG@SMG`Qk5<eG9C8{$!|)M=fgdN8XFk3~ZBM<;;3YAn)`bgpHw@M*!tLNt(U_v(tM
zJ=x5&WWL>29Ro<<^gMGoy!b0Y)lq(odI}-W6MiBrKE2V|%OH?q=(cG>LIRR!=XR2a
zKRNZXIYZ>!A#Dok1D4ovV8I=5@__eFPLTZk;Gd+=VPB2j%m2rJpA`R}f1`RogcVz%
z=0iuyDtjJ`Z`(wx#1>D*A@eQ;JpBFLrpGNFRHRWj=wHvCg0CX$oJ<@go}Q<I-3nFf
z79U*LO2|<!SXHh`J@PWLY>#2|2I)7-OK8*w!&4sf&5XDPa}#4@6SN|D5L{QFjXR+n
zPZ1<}ym&V~%^Z7KhCtb>_#%_y2JvvX#+A9AjK>10;lhAP^`$$ztjU(^_PomlJNIQy
z#j7@99}g*Ref|Ziq`S2)Hmh?rP?Gl0N8{@jL1jKO3H}fNkXK>78d(8#0WVO6TLvSQ
z6<+7778>dy{#7OTN5mJk2#lH%n|jgWSjRc3x*}K2*;2gG{0l&1L!H3{`A>B9XAPli
zlC_Ec10)5Kmhhl32_5Ta7>`do=T|-xfAggeGo5TDVz+h8HS`X-w5{npUeq789)rh`
zo(_R`aT=QKnl;u;#vXHO?aj(B=-K>u=<OLj?CJ|ay!F~T6mv9j;=G&rs%}LnUhA%b
z@uD!Bs&qIcr_cM<w*$x^N3VMqjr8AA%-%zbA`&`gYb)%3TAI<k)G&<ug0`~c8Kj{N
zG(q>ah}?EvwkLzm<r^Wx7fM?A40Op0L*d@6J_E_dl<?y|sf2MAt)<g}1tk*9mC3PH
zpcWx9e?D5gvg7gi#NRwNbbuKIR7kP&E;G=a0Nv#XGrd5am>p;m6r5rDdYo-uZ*q27
z#{SJGM0$CowZ|9ca^xPme0)8Yp4*T+c(J|i(jXtF+M1-3sJ8i8T`u+4ErM<iil0>B
z%6rvk9zOSEjo2*rMz9?sGJF0mz2qfLWZv$oS6I^Wo;9}MwH*$SB|7398Asm))DXL8
z^RlKz&HVzAEpmNgz;?!?TkHjFNSh3{vPO<J?nEZ6Z?|@OT1pg!F52N#vDu|_XJj_l
zKzW7Khp*3iRX;R$=T!^~Ac+wJvy%^+NTXse1L;z|&q)8faPe*nf_v>R9}k^^1>v40
z|1{5+#d}a;ZWR4-X_<wTYUUOx)@!YmH!X|HIeKcLm1DAGbQaCM%(jk!_hB3D<&HY>
zelJovIvM_XEBmm@>l+ES7_C+nlW4vq1Qmd<%|70VfeZZcGs|H&jL<Fc9&BBDM{Kbe
z`u+dWBl>Ud9q`%JjZiz7-a*9b$3wyn?32NdhdxBxgX}*ZqJ_DOZ!my{1TKNNyxrOa
z-V0A(>P=aH65os3M)sw&BlWjh%|0Gl@OYWO{o|ka2>WF{8R@Orlg@6*F1kBL9y&f#
za!ip?A82TK(DSIf`m0w30&S!8_T~Y2<I>fTCmMa#^sh&O4@c&XP*aSG=h~9{`oli}
z5`Tq$QMsA=&#akL`N8+g!lZC96)Yd7RBBb+9J>0hJ52(oXKGp$rx)*9Y?H|`qIuTX
z6urLr(MXcjxoOmWTW#805<1em{Zg=K3y|^bX{*GPv3^)39<!IP77zMXPeX0V62-?H
zoyF&q7WK!2F5c6k);+GZdqkw}nVh}%o;;=VynQdbH;|aiz3if>Z-bF(?`v<su#4Dp
z3;Ihi|3Q(F{pUa)bnBlk_wL@zfx&G?S#A(u63^(S0c|7R#Jfx1bo*Xo9=z<`Be>Fk
zw)*mu%Y95KLby{WP#@x_Wg};iDPJ$vC{Jsx<V6L9hd<fLiv7InPKIaFXM=6d!we04
zf51++U#TxF|M|BL7d6{+t~tHAm3y!|JZyBeqCegTZ$+>pjza|XeGj|656)o2c2YQ0
zH8+6umMRadu<B0a&BxZ=-}8|OOr%|_Rh<(8vB;1$_{3j=f(pb3aV}q91#gE-CpvS!
zYd!MDZ?H*0vcPKwBr)wes_JygZi%LTqUQ%P5=|#AE}gCBe({sXe|s7}Z}fe<*6m}N
zUJItgX&1$+8oUC#c1;b2(f622nlumA9T#tg`?x{=208ngxLjQ)l=aTw?AL`tmyPv%
zb`P2-3Kw@F?`h9$TSSkU6oVxnCUTb0)X()jo<Qr{j2X1Ogf9`0*ND9lyW-;FnsBFh
z?-E_rA@`)bQ~1O9)xE=>6BWal{0ZLkA!lAALPj>#?9MicjRfe_L^ZI<%gl?Dw&cw_
zB^kg1%8Hs56AQM&vw}g_!Y+P1bTa4!V&Pb_kp<OvK+V@28k%_QvOh`J&$j)Z8%q}2
zKC=$cL9d3kDJyJ)ej9522h4v1GIszNvj0t{wpyFqt#7s}R~NRnxgUSLtB-1LG;Pz-
z#n9cthx;b0pFWIJ40U#Y%)bH=Rq03UJmO9R<U2E`^)o?qkhaj6Ru(t114LVN@h6%4
zu~a&l5XxYlP@;_zeX0h1@bu0~cOY#kg)P-jpt~%;UbAY%s5b>726RMV$4SUg0Vbnx
zQjsbQjWcJ1Bk$e2_J;e0#eI<L_6U-Qb6WO2e1B}kMvL%);!b?e>?fZ5eY2x0FvrT-
zqm>?JS-k0~H;fAVl(sZ9Jw33fm(*@ZCmJvx^QEeAl&0hoh4eZ+0M9m!H{vr2M^uwM
zXLBb@N2y8fDk+AKr7aEDubXeXTmJ2j$sjkHMeS&L`teYKn?q9f?$Q&vJ=j?-wsRq_
z|5C`uLqBW6NPF-d-a%YX_JxxFKWc3et@o@B_i`Zfx;^C<t%&Ba9hoDoU!?U*+rFiE
zcfA2SY4-U~GzVk7XozCCF!auz>4jl=A1L>BE^PGupf~CA#u(TIQ&dl3K6tNkG?h*l
zLotC@0500z(7Bxu)%`~M$E+4tT~#C8%hS_b4&5LV_JuB}A~(cIs=R0cl;n(1#S7w)
zx~f_SvUA}sSB$T*Hjkt&1~5twE@diMv&6-F7c}8kxKnbcPlRvEbM&6)*DoZD5$+tz
zpD+$rY8tTnb^(s`9A=>SEOG34Wb?YhJsr{-QZqa31l~2}N9NLPN}2Zapt!Id#~uCm
zYR8fkRb;%ewd7|}-c8V#+&B~XaCt!O+MfO`T4i%l`rIWWb5S5JNS)5@PT-@QPuBKt
ze()RzyGwH<tziZh<_F#PwK~D+HQ>d{X@E>*H;xbQA3rc5GQVs955AO7DZ?Av*JMon
z`Rs%0v?}w3w`#2cqFt%ir*g865Ug|0x_Q9(@uzs`x7$v})E`6-)B*mqEz+Ayu3k|j
z{E7%2x&bH(G@IoWabdm8B?DV{?+)HTf2t%s(<4w*i;h%t065WpDX_i4kR8h%vB1P?
zAJw3;&e4lLhvYcvsGW_|^6w~J)U9RWex0P<Ss0=u&?5^c5cG4mn)ZHrTnEu6c$RV2
zisD%uRAH=-;T~$g;+)TLxArolhxSFLeR+@1UTWADEu%-cCG7&#+BDtKt)?AX^&k(2
z6?bq6k;_pn>D|XuVFk4_Xtc=^jNlcEA~G?RoEOX9A*nfwak}e+bw;)<gs=oDosO+Z
zva?yuvoKj)YDTOu=#w2L=f)c^xe>1tE^i@5b4lmML!CSdLT?8=*jgT~yBL^TeHm~T
z04!hOI>LT=r!%jvWd60adJ_t`2}}ujq&h%}VNB6?C}kB9eg#AqBftjZq{$zw#0M?u
znb*C6gn^Sl<=xLOxyislpsA@f<eUrw6PNd4`fvZ+vi$%2i+}ebj)C}TH*d2ig@3;)
zg$&r3fW<dbAb6lAnoNE9@z4)g%c~n{v&-y^kB6dIu%&RM_vEerd`qb^k5tfC&Sr)8
z5Cl;j;*{g)YwD@PcZH`D%KNp?BaJV6&%&eP6bKt(DBb0&-udB3#3<1b9Z*^>q<>}U
zld3`!vLHUFOI5W4k?g6c(2dBwM`RVcE+?$Xm`)_sH8~fXUh!|BqcTmci#Ox}UtXuY
z6JacA<l^LD-j_+9z-I9|cjv`CC5UsHmYW9|-Npk>zMn=ARDG`z&`@BXzthhJG!*cq
zxV&s#Y~R!A2Ku(OobkP|r(0tZGMi!*I;n*388C6m$Tth~6+~-$pL&0bW6+Y^=o17)
zMfmQTErVv!3@206(rE-;WQ6UphtA-x&oXS0tAoEo^lWmD3fSz7cJt$1%E^@$RL2M_
z&%J4C1)!mzzXH;CUfx_S$0m6$26WwBcs{nIifC6LneI0U#(an3=!M)1ahtSBUCtOv
ze!WEU>#ND4uY^?F9H|>~K1vxxak}>Xf{(v{^*N^7Km(U^yXpno%YBQE0c0swwkMcO
zodRLZip#mEIS5n#LATF=nrYN*_tsk3HcJ4=QrJXSS48cajPQpn<zDqpSNodaN=P)T
z3w*wajew7nF6bMkK1)m=mlbF$gS2wY0;#Lz9*eD>xwLs}c0Xm!pb1EckOC<Y_kffL
z|8)G3&+R6{|Le@RWKTo6+s(gOCb7Hyr15S%>l@y&HGOvih+3U2Q*ZLy2NOsW8!H9C
zC^Uqn_f%A<P<MCJ>@BG<_;yz*_kmVDut*R`z_<3p1iT-&da9#sh;c<UUwXEVLQikx
ztpVFhWhIMd`o!MIGf36IBEDJdK^U$1uK@!lPsjG}w|r26cy5$SF4fcKi(7a#5IQ+d
zlWNTbw%o9wz>!_}32j-55?xzxTYp{BGCQDzd$Vh1w*`6<*A41!-Zc^iJkqv#UDQ&w
zT)SD?nEiE$4ahw!Ftw4p0Vj5Fd?2?A6Xu_rjiTL$XaG2oc+9CgJx}_(_|n=;0bsZ|
zrGIOh%}X3Uy00&^NbXr17fA6ZZD-mNx_nM3dxe-}58Fcnk!JzFEYpMWfUzak5B=}(
zu^{az<G!*)0GyCo_V^Q%d)9r;?5N2Gsq_v?&9PgEx*X+cylw%#x<HX2`4v1fH~9rq
z0@EgUN>mh47$T^?Y9@8c%ouMV+_N5EZTQZ`@y)B$nuir%pSqY^-X6MK0eZIODhDRS
z47b194)1;f3&#WKgw+Lobhfwwro|x{%vXzzT>PfRQ8&b<hUfvjhR_=sy!Kk?ld9WV
zl#<V30i_7%VV8=oX(K`FWT5uO0K(ag8h9y(%5v*2bM$EXW=G-X(m&_H=i=6bsV9J6
zq0_-$qxJ0Y3m_9EMHnE1^@DOii($s2y7?de>y-JwaP^r;2Av?h@96LA2)F18C)I`C
z0n8Sf!@SBmJj}~g4<8MJr~^eAqBdPu8=`~FreNwzw>2NMEV-2i)82fOXbqBk6L$iV
z^{&n4c%x^DAb#7Ns+~2`gU#k%`-RJJLE6~(oS8>`s>|qE-)WlEL}GfcV||pYEO!&7
zVzCI$oS5jzgTrkhHw^CfJSrS0(sN?>dgNCJPPXi%Wehb(Zbom~%p_g67#8Mye!k$<
zc3Jj|)=6uVDh~-c;ti$^438aP2jI%PPm<S8J#nkC<t}@58+vc|9Zw$hLS0@xtHT}b
z{-XOxl3I^`?G$PN5wfvMWW@kM4~)CD&DKk@n$9yF_~T96(j+ta6l7e~Trm)x0q@o*
z4f?{I<MP$f<<#HGuq(wah*>>V3XM-Ouec-vm+(ZouaWLstE<<YN6G23#-ny$Cnqfq
zl&ZTsg#BxnzMxSmHx!d{pew%b<>02UOrqo9yF@y5QD0r@S<C8o{Rt5@fd>LK^@4F&
z9RrrNGBaNIG+!6-ershyG}e@RcV$g`$9Q6-Z(#3cBCIc|Jf0!J-QL|$`&s?CZc_qr
zRb&rQfG^w-v0By=Ha$YZN}9w+;k5@o?3Vg1tMQo<<o%~2w(3#_#!@UsB**ldi=^tN
zPG|`xMBW-pFfj0DyXu0Log*VO<-g7U*RTG6asIbgqknM>Tc^LX7cndS+xyuyd}W#Z
z{$tqWA~$80_&jAC=K4$LyxA^!zI%pTv<oToAQdS5_@B1{UA&ibO)e#%pbrN16Be_d
zneFTq+6qVo`w!O$UI+-(3kq>@oS$4~wdCGz1q4wAzuo?4?uZewxt`ETz=NY<Uveua
zl~z`lSLbO>gr@ttnk5lIFAG}%!K=6T_yA5r*EzqHyiE$cw~f=-gq9cp*b7cV;WVvW
z^qnS?CdVm(0PynDBm+W_-Cik}=CJ>(9&h`FdBwHzXAZ1Jo#sF79&Xd3xaa!8&Xo=J
zS_8<G*viuKW-zo`MRvA@>r2<rwhCK6$5m~Q>|0?@7rak-mVyNg4+#{tHEFFLEFLyq
z!*u3O;Y~L?#|cF+ITghgPZbo?bSLYKACpmp;|>DJBNf+}&_9z$45kxU3B>H7W?Ww1
z&E4dQ*L|q4F||*24`u9LxArE_6~(Qk_)IOwP6i6sx9O<`S+OpQxAUt1K-1h)tr##~
z*|muczSp45otcw3Tyi?He5-9l1UXeWeosewBrMeBoZ&q%Gvz)EnyhAsL`qa^iwc($
zP{B0rZuz2Mm(+yma@Q0OTc2rdQ-Te7v`P0>4Wsy?b7M_+GCxI+)WoVTc*A+<b1OpA
zG)bf~uGuqLcU}{bk%#ePr|ku@(9tq@FZUAYdj5r1@BVYZ4;?zuEDma;1TP7ms4$z#
zJL@*7rlha8BhPdd%t+XA44bdex5(pLLoV8DwE;w(gA(l|Tzcn|pp`<xJ_3>e_J$Ca
z?0;2#w#%W@K3__86kz@rb#ETkRF?jGTV1zGElVkjQb49w6huTs<}ux<P>fMfhJ>L2
z8AFJG5Fn6LtANUo5@a4yiUAQqWC&pjDP@QVAu=UE7-WnA0)#*&Ail@$u3y)E*Sc%n
z?swgH{qFgLoIl9PKEpoy+0Xty-;WE^vVhyxQhnD*?5o*giHa6jt}T+pSvo2#dX;FZ
z--N*{f(;w{{ZNE@C9I{6#-c@x=%RMC*~1dgO$dTiXS53=R6Wvbs0-cIS`27(V-(Y2
zOE=F0L-SPc4iOld?e2I+8MnC2jPEI#Y0{3pigw;s><P2h^BUSqa5lM?F?~JLqJ#&Z
z7LvZncF!LEDp~7EVIko9j$8JYn{3<PwcmfQF#3rkp|U&{o;hGr4V&R2a{6MXr(BRl
z{3R=K^YU#)?d%4}kcC*@(qImG*JjfKV*;)Wvs1d8#M&gDbn=oMqRp<ch6VX<P_b?i
zQZKyU$?fnw*VDy*@rd1$Rj2M-%Q_0L1lt}GzO&lIYUR{J)#8MgdP0ZrZW_7GEzK!}
zM$JqM3-g%Xuppc7*%Q~xZS1xC^3c^IiDbK3!&a8_%A|?M2U97^P>kN6H8LmORST#~
z%4H6Br>QLU=^tm5jW$I{G^2xibIw!@tC=0K!{kl$QF_;Sh_zsx8DJy1z$g+(Pelij
zh`bs!Tu1&GJ_!&59+gYE;qrUU*1rE9<MMU&FhrMThrB|{#@@V2*A_x}$Vqc&EF&`$
zbJamn>1FSe_wfVIN0u^%7==U|R^s?sXI-V)4Dt}ymEhZFkV{Gp2jm>7B}u8%qON@S
zV6Ahcf%T_--dlapKov-fnIfKKp&s!Za!S)MYJtvb2Bdq$>CG5&lO^hedBSzn-2V0f
zmH1XW)w}@7N*c11qB{pm>%TQLzZAH#6VZ)~)3MfCWVxOSE{wkJ#JqF#FTeMnFIHbk
zGMXs)-4Ti^r7{JJ`@Nn^w_7rNreNQeY3TwADO6T(Adyg69utD`Pq4|H!cTqkWQ9HH
znoZIMrl1XuBU6LJ3tME<X*&-;40dL8Y;b0f{TdtS76V2cdj$_)U0EdXIjk{OZ<5)o
zTc3TCP=S=-hKmc``4_%z7WnrjdA(Yw@b>x)bV|Qqah1P7OYDc_<3CSb14bJG%~+>1
z<{(-exFZ>{0%cz!$NjRnr!mQ&F`WywDLq()0BktJ@DsYL3n`$ev>E-Iy2j}Tu9qLu
z(P6iDP!JG`_g5plYiI*&HDf^90pplAReU=<1Wfxd!NB5h%xU&%C=tQwlkZw&gpvEF
zR@!puT@k2tGr@<~_}-uj^JvqTZ^0ElL#6EM<@E&;<{Vf2eo5bht!5JC+Ms;wxl??1
z+oP}s@X;RtkkeQ2u-9Sst!FR%mnG5f*ZJT0{7)?Ndi5r{y4`8wG2!twBCgzmW|i=c
zE$MuxwuC*z728ulWVr{OEXk*RhynFi-+VR-_|s-brz{+znz{~zRGq9RvsvnNM?Ly-
zM)TqMu)4B~d$_Z9x3Z7!Va@P@y`MvS{*4p(U;XRZK%PB@HlO-wU+;cE))TS2HHKS#
zBqLi3WP8c?WY45`Qf7`)N6E)zi;dKw$q7=;i<&<i3j{-@r$Zl^xRa`(-t3!|QHUsn
zd*U(P%3vAfi4I`C7zU_a9BS%EDqZdIWhI+^fF3c_=7QyCcP?Ny9ltr(=NJeS(CcIu
z0`Ereg3Z82X9RQz;4%UWtc`#+QH-1TI}?XF3V3#T6%_RYgow{zoV@_p_CuNan>zrM
z@lQyPmjE5YyD&GuG08D#O_yrW{xq*G=a(o&DcZFsukPLE`?n!a4YM83m>cil^H#7;
zO_jP8<}Bs&{_akwM3E3KJq(zZ&Lq@URU&@eDWmoGNgS)fX3I!z4H#ol^ptrtj6YKf
zgNy@Cm15PBak!wV0q%s-O)$q2_Sy)$6u_x6&iYc>#NN!~UIaDZxgFC;0#aZsX>?tU
zTOntW9@Lu8VPDRup5>==)#J}}!=!$v?0Y83m~U<b&(=c9XgAnvoAPhI|Lg$DK>B%2
z&$~V+Rg#)U8%)M$?p5*=ERUR%M^)V}Utzu8Pwe=4fOCcsTgbCAfX-xFFq|^8Lhg4G
z{AvIP(CT~Yx-1H*$m4>qpPe68B9fSwq;jxH@2mj_Q1F^MU;w%s8eMowV-u`p$ebW?
zc2uEsjvumLB4j~z1*pKOoFk^vE0xc_U-<n#A9*E2wv^uD_TVLTt08*f?J+^VDdA(@
zy4xd?!+oJe_cacF3{P~|EENvE8p*4y$W3hSzlm~Gs?4jfT^g7fg+Lp)pr>Wk(%F95
z<<IkM)pBcKG^bFBsjZXhN0ls{ZY?G7*f73|PjDF`s~A<#5OFoQwYgz|Z(oU0h*+6&
z^H@(3X%n8>1EV=2te4!NFJ?i(z&zVh!k0JYPZaESEs%7U0`RI+!eTQ&l*I2x?(xhm
zHRidU9n)yhDX)MyHSpVO#RIr*SSz+#8_$g@@}wVx1-)EX5DXTT(%Ww%>l1JFbAI&;
z+V8S56V2H}uc|koeQ|oARjnG%<h*%a?T@?bx$k0p4JB)drvfa89ZE3}a{64KN-6~q
z28rYz_&Oq&xLbRsb0iKOa}F_loTd@Kduq%mX<=kUAk{j(cB_CNzFSNis!K}mO?ueJ
zj)P(ro#$Z-J~I{&qs^V*m9pN|gw^|-JqV<3kB4qdEYj$k@9Aqi-eUD3802DJ3gz65
zTiHJbVQXSt>V}$qtjWWGIQlTt^oTi5&oS*oJkLchi&-NUxmY<{dfx7FYEWBk3C_tk
z+8e(#J+5n9z^A@NL>GG@6LjgNd;P<l3j@<h#Y*d-3{-s(P^1I;hQ(;A+p4*etp42#
z)xZ3H`}gfbGe7NP{{Z4tVP<k^#S*69!n(|f3KP@#v|V0m(bf90fBo<Op}ix<OBY<3
zrl5x{I)x@sT*^p1`RW96?@=RQknL4xUU6JtlwiYRDcaz9Q#xjZ2>9at?F$H<Yvl6N
z+u^;ZN5MZXTT(G=J|5*8rX5HFcAMN$mgzfZc(^S`pK|Z=*G?TXgp2+i@$|avB#-0Y
zCS39z#lr!Ct*;|AZQksNN96RBuL(E=X(^EcFc=S@{Ejh?x@z?VTQeaz_MosJ^}5yH
zmZn(smgDbK_;2$u3INQEQ#2WKF%R#M$2|JZhi;W6r+R|w0hS{Mo~eKBA8NNm>YA@1
z;L|Fe{CU!E*!SLVQt10cY-MtYT<-@d=BuZ#KRPjU`x-=+`pTaHn>e^A!K!6!0l8_%
zGG1c#cuA){M>uB59hm%u1i}ugNn+K`_ChxpRrWX1FU||LGd%TC7w3pV)4HL$tu#5E
zw0})d|CC@=hMt*<rBO_+wdo6*PkZ)WHZ{)m(@Bu&bVnDPrmEQFuJPV5+c1BKv=$8%
zt=W>Z4HEq?qgsfD=WaHYpKo5>aoht;u+utcPEOu2D%T3z!d!@)d7Ae`6ijE<vyM=s
z=a&Y(gs|bPZo=552@OX7%2=HQe`z{cAXS=x|Asto85vPF4XSr9i45<T!I9M_RHb4X
zx1B@UT2aFsW<$H^{!iz<p8uZg`A;6K|E!M40~G+_IccRYo`13xjK4558u`Q63x;BX
z&0J|M>~ZW_W*%tDQP;<pK*g;(+k|0SzWSV${sO@WSkHlP`^mcm&sR@myOqDAT%8IF
zJgF$TL8#+e=61IoBeegL;eYUXR*Ug?3@E>3qlj7JuUF18{1W`ze{y>N^LM_utRC(t
z1F4wEel$1P3E@<9ecE@+W4lF!6RoVOZz~KxzaY+(5t@l2UA!!BX_OFj^(Wv6>Z>OW
zljhG`UCUF8Nw~~-=V$7;jOE!J1;9*-m>?y8f$ha4`ZKGJV2vqjjefDfS~KYi4QDW?
znS?HQh$H?@X8rR&+ktVV8#XvvnN{RHZ?FM?5(bWj4Sk9|Uic!l>-YT?fkFfg&H;(7
zdXofTJuwI6y#S@^6(&r74hC2Vey%&vv~)UH%{9R)AY0QmZ3*=eL&D2xt0!${1Yh>G
zSnTl9<TER|`Lm7FF8B3}YgEFnW#<x7@}iZ|4e6xHwHl+BN3pe@$^b%R>Z2UNFL_a1
ztC@UZjSVu;TgBqTVJrN$HV*;Fq+~f}(==J!Vc!AI*iG6fshx(Qgl684B@IY(Y|UP0
zbrK;t6Wh0Y(~7i{ubnMoq^n>8tKQ1u29zE=AZX*(A>tJ^ZK7W{q9Z4M{0^oSa~fv4
zH~&$mUr9X*ene~OPzjonoYQ-AGiTP2LY~v8ii{%qu{6!!^l5~Urd;!mtr<ZToj<1j
zV1*JN1nkh>V>>q4Rv2NzP%JvwZ#@6U?uOo#<!~0g18LQn^kOOgEG<h1PAHMKh?7ik
z`PhQ70+CbsvUTm2jJ-+Xwn#K`dwDfcC4{QGl1t!Wb`@!Dr*@`wI#5+RyVw{@AXSOa
z%zl%9<?r_?7SU)mv?Awh1g0Yc%~0<1eW>(AmLeRzyp&YNBiUk09^+qfO&=lI*653b
zgQ4!NuX=9M#=x%jo5(yvtDDc;(AkMcml?H5j0Ha@a}WLT`uRR5)NY>2^{U^Zj$$Hq
zBw3iaiCwU&q%1nHyMgROC}NI|yspdXF{`vthp$k;&~<NMG%ts`Gdu4xONxsP)WHC_
z=Gl&tj9LtHQ<;C>@8UQ_EveE&QYy-P7u0jN&^cijXW_17AaSm{E-Wa;=0+O}5_*+O
zju4l7xe&{6d&AAd^_K`Gy(?ip<I_K_pV@Nj|FrK{P>_Pd3I-U?QJhcvLXx@xOPw*G
zB($Y_YcJyQ++h0S#oh3j$zAtB`FURo#)CmUlBAM)Pz2Ez0vJmdH`Ee$`!J~2eE2wV
z(P}whY@yyAUAAgI)5PUUu=-1amYfZWUMhc1N8_y}_N_`o(;|8L+Ei1=<*As_`KUs+
ziKP&HYwXg^q<L=Uw3}XREI?Owr48C-z);nR*YYwCY7gakoy<Ez8kNpC&)#X}xPCB2
zH~hvQD?g4s{5oo)#80^zJgj(UtIq$meLJn~%)B~PcX#FtA)&B9K%r84c|lVPNMj_~
z08IVIV7>|dC<^L>`38UvV6AnI&Ye1CX|20%{(Bddme^ujL$>L}+pm|Q7XsY|b@@nC
z6C>lzt@Z6aTeew#{crxT)$w9fc){xF!lWgK&ZFsBwo5bf^(J@D^OAXq*}dw@$X7re
zqIT!)q%?!>uo$bf3)5;}K4dQ##Y!%~^wjaV>={~j=whEUK{7bM9cR)s?{VCd#q7*6
z_Np*<E^kP8L>M0~tub#?ifvrUVxtr{7CH_?min9&0p%ggbfd>xRB;}2X;id83iC5A
zNb9r82aFQG*ntAmlWU*$mEnQqK@T+q#|x7~m0a2*&a2bZ1x+Wyj=zil=Zon7+=tF6
zPdkRkICke7M*3}uQjH+_;i4|WvE!+2WorxJ+47BA<Ww2KJiF)ENt=iPd6&+a;Yua{
zsGt%e91=!Yxrz5ULk+@IDz>F{p(Cf-s+G2^^p`D(R>bUW(Ojkf5+c1e%8(Qyu#O1b
zR9RZ15SF4s=DB%~m9!m)dVpY$&d%f-&c9j^E^!iclO4Oj=`Lk8P*gwwMqS<FNQF)p
zdbzNoLDHaaCN*}n`R-pA0!0tJMd!9%ppaM0PKYyeZVm**Ib(FgEKj%g1aBaP5+`@t
ziO@F}^XA^knA1xc{QLg84z4{Hu>*i`R_}7jT#Ji*E|b>QEl`PDSLh8}s$n0WPCH4+
zcZ-3ZPPQ5otOWNy;>&LnmMlAq{Q`|EdJa>P^1AB?#m9TTCG9cKk0HDZ>OJ;y^&k_V
z)RP1S--HlVhWFu4751Z)gXhgc((kxjDYu!{FDfnBK{sR$cn8W}hH<iM&6k9vltd%V
z%<yBp<w1fv>hwlxp$xx8C%RCsto!Vbj)sO*#I1EQ5jp4MJK|@};I$w%Mgaj94i9+k
zoPi_+6zVS$xN3cuHQUv-utl@+Wqt||Dz#`~h^sK`fE*GLYg8vId`N@Qb>+8fn85n&
zg@EgF|CAj6>jM7o@BL>%M-K|&eSOm|#N1@y5<4OJyLgT4_zL_apkx}bOOyPG7Kuk#
zZltmBM{W~bOB(DrDdCUv)D5TqYWFM9GADAw%GgWQK>q!YGEsJGHp`El;PLTfoo(Yr
zWymJR`lsmKzXrx6yX)uu0Kd3%eD)q=lxVr>`c&mgZmp&`G1mXP@dy9=C-}eo&ug@a
z&(KQR;ZOUrk7UZmCEPmD__4jnWX_&+XaC;SE$*QX4NP^U&R$;_2H!%0d0hv3_P(!D
zl8bm6P7CM_fm+U6NFIq2#@-KZIYF~{$rD=!(}=<6mln&uErLaqDeJ>gSshQ8?>E?#
zc9ZP=E$bNvfP})up0tUl3<1Rt0OghoreKAOiJ$hF^+?%FV^djR%7Sro2iVRi5%EG_
zDb{4G{=)>TcP?0StezRH2{SM%OxfJ}66}U)j&3cUnAj<5kv8oaz~KHgV3k2DXO?QB
zd7ji{LEC^ds3n0nG@)(QbOpbJkj3QT%C;cic7%zVY)vJhj`3t7K1448R9o3xPcRg<
z;nd&EQS2+L{nY*7Z@hq5rA!N4_w>BR`4e06Jojy#isZh(ZIe?gt4dJ{eXp&GGioc3
zPB*^t=tc)8UcG%gIXh~;#=Th>wZWEbLL?$HN=D9HBsEf;{6PXLCXY6S+?^!10+<1Y
znXzdJ8HmVo3~~^TQ(*!F$B0=xFpD9avOIU0?I$<QzLR5Rto5q>udl+z;i4R`S+he5
zF3Rdks|`H{J!x;-2GkT3y@(gi_VF)$*t{9lrb{@8&$2rdWI0%QA!JWcoU;|k8Zd9`
zC_+Y><WJM?8mw;<+Q7#q#4p7UdPQqgiYdALYaPbB(me<~$}9TH9vfqM1rn$b!d1n+
z2r2PiS15-5Ff*QxQgK0p0?-}SE=y;dlqk=vX0%SvkJTh29_4J^=}o>-@cWZh05CtN
zk!cm6)36_qe&sm1c}h3+eM14_lG3tiKU~ilbbm+$#u&3Jcdgnr9F_eYcC@R;=;{g9
zRk4PEI7y3$*@yl)aYZx@*_9JmukCye0jbaZne#+431cjo6n(ilpRyzi_5&cAO&yg6
zbw^i;2B})_xBx^Ga2fwQL^I?O?e7pxKf^$pIykao21g0sfvM@wr!S0F&Wm8W+d~M|
zvR{z<7fPW<B)i)7*N;WVXD$!;y`ASbCpEbc0aNfHiRMw#sIfXK%F4n!Im<m68rD1M
z{IOw_bKffJ4Tgvb!4an)CX4dl&w+}NfG<cP=GK%JGz5lpbT9i1CWQ*8Lo$j2(u2px
zhKYJz;ept=NZ*F7&#bo~?q200EO>nP-n(0Tzvk#d{rT?v(_(C7<w;5F`0hkNGdF*Y
zZ3eg2CTKs)3w1u=d;q|_0otwi85#$lDki+Sm|+dP@c|5L(qP)BeZdEvoScA+8$h1M
zJqOI2<kg=AZoT1we%-hKCdz)QjChB3b9Yo|kiQN8lVf@~OVK%;l(5;x99=6nEL(RU
z<bUtGI^Lq;>!n*g1DOweNE@v>E%8oCX%8lznb}Hz4|d;+T>@!H{fAo6y%Anv>Ax(7
zJ!(xxrHRDkwtS9=aPQH3T)shGk))|SCOEe-Y$f6r-@c+v9kEjS&xjF!yoP<GeIj}-
zLM)93)+`jp=xP{agg53~j=q_m!`{u4z;X_~zddua<MB}G1(lO8SR&!=#pGp{e0?<2
zWdd;ncQl1X@v_pnA_B@0N@k84ujgi~EeTX}-}v+$JWbyZ1dQsoKq?T37!x=l)EG~W
z8e$CTM{n!iqC3^i&FElq_*INJMf|g-2i+=PKal}W16I58OM~gadQLPO@FM_Mg`@38
z0jY&{q=y<Bd?tju)_K=DxFWwmutV&)g~eCp94mj*puB!^y`+k?WYXa@DJ=WTR18br
zXQ_$aXOS~*tb*@ChlK+O43P(trS>4f4S#-n{>%wiM~5f%Zh2lo?>IgyR>;Z?sSony
zZEq0tL;Pw~+U3*_VxEgm#t2OcBjWMtM>;^rhoq(Exp6%M)C5ytpe7=d6Ld@3N>ZnL
zgwrTgLj-TBw+?q&*0u=5i_;ohNWb_2|LB}X*KW*I>|g%T6~7oOYmPOjqqIyuA57$R
z6(i4v1eHyD=SG~#XpEhT&g>tyw-rUOqi>#%O?G|rJg5lbQGCQt)uw$DRU|)3G<?Eo
zQhOn#b1~0$RfOtpsLDel&D>hAQ5v<*%*@(V{7g4V#pH0c_byq#&XW`@Hl5O~DOf?E
zf2=Sit*mXRhLo-DV4^?7Q)=xDosxPx!HO5JcGM-8*5m-G8}P@bq_Mf-En=ZmHwUmi
zGPuID{&Cu3kV<=Oa4E@(^B$ZYt#sZP3kQ{)OH!M9EAK{wqqIioYil*Jj&SP~!l=^I
z%K-ykHw49p%c#jU$?5q!>we;-6--^GGT^`1Fg-=6rjA`=t@z>IY*XV{a^2tpQNY;J
zB|+h(*>;qe{_XbexY4jo?G6d>K8l{&5qk15y()1|1&u2pdm$VVPiv^#l0u3`M+pdd
z9-}B{d!3kxI2wgK5N{|n$P%Nlq$37{8!T_7Z{vvg4^b0BwM_fAgIz<+wklZWQyXYa
z=FBs%^-UxiP<;PFTyCbcG!rdHHM4MEl(NXkPy4bEP8KG})%F?Ks`6O!UrWpU$`*`U
ztR{1@&ZsBDhhyMDy{56Zca~bE1tDrpO(4aM09sa^XU(4Xyy=?yP!{U)qruVnEj~2u
z5iINj?cNVgf|8J8=O|`fL>^y0DNm4dS{COB8NM_y+YtWE&-?a)U>M#bb-@UA{oOo$
zcSUa#XvDu%z|N<}|Ie#{ztY%yS$nnk45ResT&OqwL?ZILHOn-s5-sdAhmyxF!%XfZ
z*yZTfo5buW@2!QR(8SsEIW~T0olVM;(6nj5O35Ypr!}`nr*=89v*z>|=fJ7(*vM}t
z@pawp2}0&nM^nK={?Z_4<=7Ryhq{tp`gX0s)Yj;nP*l-@er&N5MLb@>=k(K*=RkM0
zcDBL6s>;6)TfpbnSQ%se6!@*$@!i8=v9VT5c3BOUIGV?HIC!PZUf(k9)zYiDIzdL@
z={!hN=ieI3wdjjBDQDk&mv=FIQTTZK(Ka=mcs<6yTQL@8jdfd<ZV{46Byu7{>FL|x
zU5%-TU8>Ss4u$IClrJ!_4zk)iI#ast5)YXejwn(}P(9SoD16XOtt&o^xvNdU9X<s@
zdG4vn+uTjB|3f4H&E0Q<_I_nFn9UvBMP8OFRNYnn4VaESBfb(zyG5V&iN`+x4jpLe
zp~0X3c-(K`J7JUZ$o4y56218u4P|Tdf7{V;{Hmn-3<#8SXDx8pK(TK%*P-nC*7dB#
z%23l6y_^F#h`Hp9q<NR`z7F~cIF?ZStuWA?;cuLbIvMp~%&5P^#OmI=ar(gvb%hyM
z7gbUXPQonx4No`MQOHkYY(o1C&s4hrw&<IiU(d_br)qu^rQwr9dFVgE%(Ml5@-Suh
z#5Zp+31N4qU$C817Lp-PWK|TuLhHYDM@)<e5+rmtdBea(&f>ncKTMqdcuxP${(rcz
z_9xnOA8Y);0%OyD0W7S0UE`BXm|>2Nf@W|vpZ1N+T=Q6;Wl38WK$AV6_DR91eh+H{
z!YOKfH7lsH{W{f=CuH(4ra@C|l!Ov<tJ^c=qF3=Z8=J(yCtX2jeF8U+$3B%i*Tjn3
zesLp>Zu$5gwR^L@a{pVdKt=BiwR3Y4(BSfSD-~MmXq)C1rknw*OsgrUg1Fd*A#`Kk
zJ>5(0Xi}qNTA-gF;)oBtMlmi{;2O2W8tZl%i=HRX16AqnBfa)}M!;gSG<R@~(EuM^
zyDjmHSssWadrFRTW~MrC=A}&czG<l*pEvRx%%Zrr$4K?MMmH-DWqO>9{ka<$Lc>Bw
z%MUoJlqDV%;-zH3`hQZa=h3LW%<$~q+>m*e6vWp`p5E=yR@kj{ilW8o<^Cvgu?~dF
zooUIo<gEViP;gv>Ed7d=fGs971z?{Ua_daoBX;0}z67aQB9AD=ui+}&pl4pq6W0tm
z8#g?feU(+aJ~CY?)(LY?q`aqxsQ0@!dn(tB2hP!*t-!H}4R)&a{j;-c>mdTMr5Zb|
zHX`?VUijlTeJ*M*mWl_}A7QUp|2bBlJ>0>?<BBB)#fZun>uKfY(5nDRP!R82l9N1X
z4tN{By*=JyeIc))2H{-UbyYU}-onGBSDzl05LUe|3|t_!uku^k8IN}2>S!52d7Wt!
z>>fa=>bP<FS=f>C-=DqzQF8~^<JLBHDoe*_B;lLLYjc)Kq3^Lrq8A<aD&6a7!&|)@
zt*P#?cgnU8zcCDSH+C;t3Sj$WuT^$0kS1$_L~fTEz8zmvzStI;m|4SWXs&C;`bm$c
z&3$6US$@{^XF${mx!IZXaMgOJTf;ii%}!4%!boD?UWUUZd6$(hlM#Q%2-a?GT7Jd|
z=6=QqhGkizBWuBTBErO2>vt=QwFeq^i1+k9kQw&vm_+A^nQSBL%XH^anKu;(kqdoq
zyc_R(1|YtYFYRQy<t~<ES+!@%YPLQ?Av|HLOMr0sxo%x^`~4u%KQMwA7SYZczzCKe
z*oNU(cKfM$=IFA7RddLQ0Ax@AxbcVNbQ|H}ITs=oOsw$4)SP@60-`nkea?$S_6SPv
z9do+R3i*WEM2%ahA71l{20sw4L|^~mHNrq2!U``cM1%?FX_#&epJ(f!n~CrjL+qXX
z{Y4eIb|(XkWG#f@wk%}A<k=oCO24!$|9;-5eF1c+d-cJy&UxflsYiMeM_vX+=N%W=
zHrAnrz-bK6<me^hn?%@POID(ktsb{%<~X~HV~HIl*rXUs;qD5*NGS{)_v3J1on>lw
zSmJVcL=ZOLh!eQx1A7_X!!qlGg<s{aza}1+u#2xtwLIM9sDVboNmmvL*3Oy%(YNpz
zGgs?%lv5JB7pKEjy)gXrwK+U=5_0(o*&@lMuWuY`|5V9Z<7ld&Hw^MEY^V&cz4&Qg
zUj*gso6I!>P-<)bpk@dFao28DxKVlm!$L<UU*_oVw$IcZIi$%sHq<DvOHU$9t#9jr
z#$-NoWZ=!Sa9F2n=Ia?{N@_AQ!>4mjE`Px~66ZRyn^}c0+QpN-p(+&)jNyUvpr-J!
zgog!WTt^TV&*yX_p1UIxX2@mWiRc~1=|~Ifot9Mv{q_N19*@+jkMIJVwGP-e+3DxE
zv{T#4y!_8++maGyT}rZtD-Yhv^YYKFr7)!;*EvIAs_XtiT-NeBs=32X9mqbSE06Vr
z&0!5Tw;S)&{lx5)w}vBv&FKVVQUx65M4O^d4cLWTSduY&*8VHpt;?IKl|)aB>w>E3
zm0igiA0=9i%Rm34Z+%&}2VXY1g-U4dE8iY|Ae?sYAFi_T#;=`v2ZaYh@KK6=w(Ak>
z&0HYX^&P-C$(B2iy8gOdL$;aEFX+(EXcG`G(T!5CkHr0vm=PKG&9<|$+v(+>rp8Lz
z$vR0<OXB!QdDmhUs$P*L6X5~^2Si=j5e%G<g3+ifDHSvkC*%CaM|o#K`fVNOo_qND
z>WZN-*y&*?1R!5Vj90DU@X1WRe8k!>JNlEzs-(y5$YL~BGbF4hWg#A#aY@PP_}{`@
z)!cT~SRJ-y29Xo~T-;z(Pz{?K$7J28$a5Nq>%o|946leLy2o3H9UJO$lo+8ljlJ~K
zUa61{cez|2QdvkGn3b-#!BXa|>(;t)xqSnReW`qda@wkg;1=VWIAM@k0xhK>H&Z=@
zGn-)5q=GxuAZ$Yjvd4;mqXG7nJ`R9ZBcr9t+WqeQ5v}<P^21Ty30>x2m-n<<9|zl-
zxDVY{oI)Og8xHm9LJP(Nb$ML(j@FFq5z1R{8`Ab3=4SU|ccOsdYMR5N^fe);V`3VR
zzBGDV?Q1F`j-|(lw(u^Kc|L3$R$#Du^TgYHR7I<Fj8Y^O?*_h3S2HC9T#cFxQ3dA8
zgL7M?_&rPgd9|hrmH3M+uG)&TKC{k*j5VROPYWch#{T}6ubkW7!h}Ctcg=)?Xr=}m
zsdG;jcT6xuts%b!tG|BQH;%gv!9uA_MjZyxiK#SYM#7)JSN!jikT#GAi=uDp?b>cM
zHM~K%C==QPv+&FJlgeFchkKP0%C&HGrQ|jD=NgJh)6LFu!~;!sJ7;5I#0@I7s6mal
zeq$$?J_k~dHO(U6k0NG;Vaz17(EmWk$xF^VYbO}EiRL2~DhYdT4b*(@u39uhv|IhI
z_t77fM~60d^jfsnldM|Qm>7>d^(M=QSUj-KhiaYld6VE*g=KDvTF!O6u(7_4Eg8IT
zbxpw1;=urfau?N;?x%*ND6c4+@aGw=rEMKt^c}Qya?umMftdCH`zFa5k>;I)&ZK&M
z+BfG@j8Vb-z;CzplWaQBB7BMOe;*S6{m=i%=vhO)2SteDky9#>VjGsaN-ak3wS_Gg
zKfi!KK~`x0TUQimk#gh>Yxn@eWnQa^tC)0)q}@4sP%qrkP^;3)Atv?gUx225`PZ4i
zFS~4=vZ(#C6P9Vkc?uIeC)`i1+(XM7_eWoODHp3?E`9VAk7k;k-1H6ZIP}Y_v9?f;
zw${|wZ^t38vWFE6Zyn<O{O_Oi|LYrnO0^!z-$k1uc7M5p5Hc9f4`pwOD>3V!b=DYY
zS+DX(AHMP)G|F!c>bR#%uZFyP^c!%5^U14}<X^I!KJ7bQcIDGPn97N6*2L5v+{bNC
zB4zA3Z-0fBsbZ?jF7NdtKJ9xh4iXnFX};KJ73U}Sa7;hpTQ!tI7JW9Fnah`WTl>Ca
zpGof>^KmQ6rRQB%M>&x2i}-HsCA{y%T&9Cj72PoRG{8AJGLR=+S0)qQMzK*4Wobg)
zDtC<=5j4sk8S4z~I4@<J>S2w)6%&sFb>Y6`&nm?$U8b9J-8j{aB^uUPg;aE_J1@5v
z3t(Dw9f$Zv6rdiF@GKyZdy$wQms~RaOsKq=Oe2~cGbi-2(6Y7TCweRv%T55S8;;x5
zOcc+t*rBLI$641R+^EsALg8=*-U?bpQlZeE<lL$T%zby5I49ZSYiVIt2XWgHq$f9R
z1}8R)N)IQBE8G0)S;xa;41O#Q>agyvqEJZj6^cuD>SVDI7{r!(wSZ~@yW6<kVLQEo
zGUD$}17`0lb5euw0WuCBFyW=|Pg8@Nx4^fGQbh)D7_F*`?d`2gkvf##sKrCrn2`Q<
zlyYfjx?WIOjW(2lEF(2MMOAsURjVVzGN+>;J+5mBAw>fNa~UTgW-qt&=0xvt4=fcg
zZbi-#tH<jDG76<8K^fj!k*`B5sk*nL#v2x1(*ryG2IwS~$PDgeRMZn8-UQd5T94S8
z|2zJXtejM2M1B<AYR#)P=Nrs5M2zC2Va}EP9%@UvNqGh2`aEn6Ez$YKwSPWtUo1@d
zPWw4K;I|j)3tPH`nit?$UqaVb#A=lzJSfZ|yn$h4fpSc|^AQT|Rdju7Yp?B^XMZ}+
z^Y-lPd%k+!iuym*I<>?s)Shtr;no-f1ytH@KX=#5H-<try_*C%nvM`rVPP?Nl!)qg
z=E3MN$d!rVut%<^8f#D$N$$B12r#u%qz4Zkq-886{rrzceOX@`3Iwr17Jnmm+OJd|
zg_q<%!MmqeHrXkJoGGiS8E(dMJuD&@bF2NUSB?w%4px>Mhv<e_E!@h~Ifpbknw`O{
z9`%DJ>ufoulmSEiN@uK(#qgO%Qk-A_9hlc-y1VMF_f@5HmS0|<ygsxiwJh<&LlLLx
zyzk2Ap`QE`&%*Bxkkl%>eA>y@yI|!9CMz9%e4O29$Yj;*CAma27CXg1hfJS><AJo+
z<xE3Hwn-L$J!;W871z=({g#gxdVEQ9zv;i|Hfx+YeJRl7FnRdEn`4Anskw9Mp`w_f
z;*(`{xPb~KfVt+0^0N?4Q#U^CV`2PvgF1rc#u?$P;e|Ha#AKMDIqOfl<kuS=NeQ&#
z>|RC)U~kh7i(((+dzk=vt>!&&Zp$PRqaA`(*v+BMTu}W6Je1s~T&54_VGdvPa3M81
zp#-kzCScNje%@DUxXu#Q)vAW?nXvvI*V-N4>i(s2?T6L9Yk9|D_SkMV`&!xB(j^=G
zA}{Qm&onE8r4YN~QI?n!03``!qth+)fxgKd1%rtMnT?D$5LmVl*Wz*+^sunP47fP?
zXKv3uEc(ED&ieXWwEip&s~S*u%2Z~8or{h7$S;CSqRSNbh*0p?R!K&J`Ticun}lGY
zuxnZ%k795j!(oNVo2AH~zc`G3zkjTlw_nWp8vWg*YY8=YeS1ObxSRF@v}3S#==us$
zIxf`Xi4j|z6LflG(U-t8^WP!swV(1rHJYO^vu48R$q1^ONTihCERQ*meARK)?$l<-
zC1F~%Fr3Y{&#baM$vNQG`gBpTC-jX^2D;b9<2a`i6E^=$7H5ie-}y=}ox$<&L)P5c
zt5LE{wsZrq+U^|~<=I~8DNTQmE|)`rYf`0lbTG|lxS$a&p9I}4vtWr=JZT2I$5U=A
z*?IouAKmEl-!Fc(YtXNqoV7IFl$;X^uH(tV%Q;Ot<~f(&DYy$K7^>kh9{Iw>ymgP(
zCIR3hfG`C46UY-Sy>m0-r-0p1wqB&d&{}0rNDbOFdT|j2GEn46D{gO&MFTEEp17f6
zs$9Ho+S8z3{W;?nnVnULXtcH?wUMz4hk1P3chf)BP@(@{t@(HN6#IAnIz-pY;5<;#
zydRX2P;dYAGexEXuH#nZDY4^h7J0r23*0UT%H5L&e?Hj=9RI{wl-y0O^Bdnk$GtB*
zO0>&0)ucFJrW_qOt_u!#La^d<V3@BbCbi@@%3uDc=E#EcycPj)UtWS%9w3ei1r$=*
zp$X3ChiC)xj)oeFwnqVtl-)bf?6)pS{?ZV)UuCD0t5@J<#z~PmwVPLEQ(Z%-Dtr^}
zL4szcswhjBpwshRC2I2s>6O=Vi%!vZDWLp|O3?UTTLK00!(D^lfO|8Bt`MSf%>&0|
zYCO4bg62~f%9N(`+t6R*?O%4P=RLhV?ZqK^^UfX9zS+t|s>ZGByo_L8x{0?E?$>`z
z$|h*9IJ*9YEmywxW)3ag4oO2nJCzPzXdLhw)ipR4WpshCRzKl83}1dVdFODJvFrvn
z0^;5a&t}!X%CiskX4a+JyfLaHH+dM{BLU6%|G54CnNa(A-#$wyTgNi9lnFW+2FLhg
z5Rd_Bk1$22+M>^~DCT-|Q)?4Z*TLlY_D6=*xvg`(0Gbh?GA?8^I3L*jz`%J}tCmKL
z#|mRN=NZJhduKK{uZxcXuxkIlELhHQZzKT$6s$pq29MWtbvZ7LR5(D@9GF{_;C#SP
z0iek7!zY>m?}R}>5(jJn7>BRg8dkTq`WkRLyh*b#x0jQI&tOIjt)Ln;aK8W6xdi&b
zhDi^r0g!TLUc%FZR>HK$!;9I8xq5mL_1o=S#$)hbt%ufqth9FW4^10*FpuEwEzXrj
z(pz%m5QD7eUGej6DNzwTo^v3*!!KZT2~Y>#{)vGH4-oQrvea1XZb4l0O9pgdeU02d
z*bU{6BMdk~nK9fcelCQ|an!7QSV-aKp&#87xxF!HvpP|o-M{H^!hKR)`z|rZbt&t7
z%GwNV6XIy-e`cWDH8-~jWt$O{X&xhs*eGx`C*RaSWKeti^g4=-WcPN%5)bmfa`sUp
zOw?`obOw;pw_Ij6O?`VKhL5gw_qa<o`BQf-y9q~MBi1<;CwiPIaC_XXs{?FqS|pGQ
z(mP3GBQl>)1m?{r??Ci_<V(hv05l`!nP%L^>~pptYDNodpk)6O`PcvHf&8+;=N6Y5
zY>`gB?6iXWj+gWpV?+7IE2G1j>3I0EXEwFsk~h4q0)o#X*18tqUgA$AL=Z@$t7nSJ
z%u}ZNgUXF6w*y(~4$0rmhLmoDS~vE90L3zYXlR@c&qCytMCqG~BM-T>sgSj^b?3-U
zWo6!;ngMLH+w%{5A!)=|m(Gc%4lXeNw-7yi(3h9{D17BA^-u(Fu?PmEVQXx?b4j~S
zfdqp7q42Hr@$9)N^@Ys&cIA<Tm|K-66ryIQ%_#3yCIP*!n%kS<Ta(F(6h_pR-QfD!
z*%#d7hAX|UDdB}|r9HRV9M(DsLl(+?=CJ++kV&lSJtFtN-Z9}R8cM8L4OvMnPE?W#
z8*%r;-XMpaydylE?mF7*z~A{jCYS;NF@Xu_T!dY^%;9~9w@R`WU$_eI4`j3XQrIIy
zuDrZg$q0^6DxS%R=cJLo^-$-#q*w6=9^b{UMV+X~|69*Wys}c@!Q@2;)5)78LSy6w
z&HU*pAFvCoxRqZ-8Vw(<D0@Cu$Wv`-5TOw<^~fWy`-NGUU#L`M0jEW`!|m}{0ghhl
zOz3Z^-))bwo-_>@Og{or<h!Y!()nKfa@PAfQdB6vC~@LMd|;0*+(~hJV|cqikOpB*
zA$phv#lAD8VTFbHQ<nSh0I$jaTO$I{UiA@;w(3XACYIyu9%EyrBX)CEk5&c+_Q?FE
z*K3fV_H&n)?uTE0gUHo5&6n(rXjX2&lmz#!Cae{NwXA?E6E@?=#z&X;hBG?Xqywtw
z+tc<BEC|6Y0pe}!#7_AJurfb?1uD*&<C(TnjT~e_mk|5}-u%vaQYX2}57eoM4C#^)
zE4gmTCAZxE(3?)gh>iHFoI~-!u^Mi@&?I9w6Pi(rxy=4~ra2$0w^}%Goz{yykACS`
z>}|}>cva={{G@A!?wO~S^>mj@oo?7HUv64Ql1Cz=e|b^Tny6Y<3Wfv>MntJ<rO!+A
z`ezFSUfsp0RQ4F5+5kEb*6Y>i*S}yglyzwmTuBLpo=I*o2nB0kD$~J>&KZ+gARw!-
zJUdRae<-Cj>45IKTsOP@+aKEd@9ln(HPhH~H>+(PGjmApO4u7t_}ux`lZ%T7pEky_
z9vJ_~kQ>W5=bTp*+N16pN@Fayzy9hc;1{U-Uvb_ZH)iWSpdOpNasQH6i8`d^%cFLz
z1GX?#z-zC_XzEdA8<t^snn+sSlrBo}+Y&HmCFiXnkZk(ncCCSZ?PJOs46;IeP?4Kj
zG-n=Bw}LeHo`Ciw=1m9#&*V&q5kVc9XKj9n(BC;D+pb!u_wa2Vp-ya-_?Ty3o*XVP
znD~6a<2Qu2+r!tAE~1XcbXm^hsKw9s?$bTrFL;HdOxs`&NB=PLB4ERtmDV$uM?&Ou
zoXD-9_6kkk=~Lb<--&8@cjHovv-+V_>fyi*uSw0p`<<(;Oh1KJE5%3j)JuK~*~k~R
zcpJwY%4;mgJnEaf6lVk*zg3r@qx?uAPeB;F0-2g%;(>B8z&Z!~o$%K`p5lM9|Bd2~
zs~=xHOWpkIAe9XQ1W!Ass00S!I7jvS>C?U!9i=D#=x=9CfSuF%{5y4``fZ_&nZa&$
zg8W|1ZJ2@|47p8;1&cT<ply?pV>{wlS#5%xNS`Xpna!Vr6%_sY`D?q&LwMg@hhyAf
zf-T96>t60||GCudd0jT%{0Zm&5ZEO_Gb5+hP!kCDGCrH|`>Rj*ir2jhc;UWz=|fwE
z1TIih*+i(;HzEa}f`tO5ZCAP^e7)WWfbr;+&pI4h!Dk^(mkjXYr_F-y9@%O5aCFFb
zJc&s97;o0v-PiINQdm1zax>RTNAGd<a`6nW-OUJkI9dwHeR;E`t!-rOPR+E<T1{lM
zSMjHPXNPcqw!WKOUPh4`i7u$uMFSdc2LMEbIVDo<7jha&VRP1ghbK=JTUN9)b8F9=
zCS-Vv)Cs9n9S=J{Mn@BhEY%u8nb60FIzKO_w9X{X>$@6mF5rd%K0(DHNntT$f;mD*
zIk7dNGWlzftKv%{)bq!E5h##4{oUl{tjxrXCACv{Od~8$aT#4D(}{Z17v>l?Na<FO
z<+ECLuj+>9j{4oFtrptjGrj%}DFpv&-2p%f=TN8btno;SeDIHJWY(OSb7^%v03#A{
zjKt{If=ejb`n{|vAjNJdVOzRC2REJKL~h}lW95qL8t{N$32lwM=zBr0j_6=lD1!S7
zdj@b1`0Gs(*MZ4mQ!!9}_Uv!^%s*xm<L=d{UR>ow8#DSb-r_Xa0<<l0D>OK!DQ+mR
zk>U=K72*k`wU)#@^A76y(4O$E&|CFS5bF!s3Ho{Yq&-8e#xg?U3Q+UbZYV%kihR{;
z{Cl$`hVi=+IEfJ>kEI8ffWE{7ILPdRc3aEB*v54tAuJuvd7VDsFye!(5~{ICBH#V#
z<1>IwRI!nc;<nK<jLX21S4ERtB~vHov@c&?k&B8?X-d4j`i_4gWz&{FMW43pZ5$FQ
zJf6E?-QEXl_sMF`A9givJHq02uN9Nq%D(S0<Tq|5-O);c!d{xOwSib5oi#&3wFp6f
zj|IdBK2Bbh0qFr^(Cf-xA+IK831^Wp+s@nY-ruSjpi4r(CK(k4b>?6hYC2lxn9dm`
zJb*4q<t+L<;3IkyP5SO-(GWB?4kn~fsPfzVEJMCT0JRMV7dBwkqR47|&GLuu<H<jc
znhNZHdoNBht^_16MblrZf|s^?SjQeVks2!98#hGx$M!~G-pq{A+H*Im4FxoLN^zsm
zmrIWnT2C<qs>Qg4=p_1DNGaf<>f#ta<A<P6PW2TVRb^?h)77bqbI)xpWAQr29v&{I
zedH%LH#SdM!4F&N^STLDX=)wcMr{drv`3^GCz&xB8%x**76JNrWrvzS_v!=x_pSt-
zO7CeS97BO-2M44}=ib41of%PiyWH}j%F{3poI&@<+U9th1?2lTMi&qb=EsuzzQ8{H
zOtbv7PX~VRq$^_Rr4#sP#?3IUUYDpcsbVRfHER((?FVkznV#;C9po`P@lT75?Qy}n
z$!gu$dkb;VzFMlWTbpzWm7xu{@?H^)wW=EtvvQJaEmRXqO&<sh^0U%|=n%l8vBO6P
z{(*LMyLgc}FLnvZBq&NhU+sTv`+sUJ|1Du)Q0RDHDyiMQY9aGja7MbdQkp8i$-Uit
zsCZn52mK+e$7}C~5g~xB|H|j`ASk!VxY7SKUs}Khd+yoJ?`do`bVzf1ciUih)*5;R
zrU3e<sckY#<v6XA{}^HPXkjrSqSZT6TjVj|^D=>iQ52c84i(o4cJ21xn^IU1ZNMPg
zScCM#>+%gaz`z~0!n%M;))GW9hLVzFAN=+cAJTK^5aFj9Zs4mTp_b~s&cKg0m)^hX
z4v6>&m@R(<%A76@ks8V6wj~;hJ>T0ONZ6&DNYpOJ<Sm#;d@oxGkrJz^+RLbr(~{o0
zmNuN`zyOND?eHGg%O2iYaXcfwILGWv^E<4erlMFk+x_D%U?zB~fUEu*pe$kJ+2$_4
zr&E8wX#d7%kQMK4l=mcN{;MaGPi?}qWAdTR;B3n78EhNTZ@=bX=gpvPNn#)*Q%#$c
z;($iCXoGElIpn$FjMwfz1Ha>yKd!$zMfThEy>nRI=t*TJv(3JH_u+5I+zTr<rT5#;
zKqD|(p><_m<=iYstwugQuRN{DE4bQdiKT3xY`A`|>qPJ8<NRwgXHQ;u6}5h+l{(<#
zv1k#0b0~p|HE6X37}1Fy9uS$W@cxC^(oO}N#-YJtv**0y1@N{nj68Q*6Kc++IQ){4
z?R+3ros#w;>&nW0^XM60vyh|Ka*xG<nDORy#H$PvTFnd|q8*i4g~=p9O0|i%3lndI
zE^e4PL?>B$97^`jCKaJztp@!n#-H1CfxX%szqt@^5Ivk^w3E={-tvLBt`8?}qq5_4
zT9#>P<bK|k_88w7sS#xvT-jqW;JyqwQ92qn?|6m1#vQx4wG2H5CH%!O{a<JJAMX99
z0(vYZpICpi7q^h}7RL*`=h?agr4p)NVyGZs<*cgXSK>@3aXaQMrX};!zAfChRC({N
z(SI}%^Z35zKWp=V%$MN4zj_f+rnfz-m>i!Rx)-2Dd-#w)zap;=Rs$V;xTp3%^hF`R
zHI(mlN&G^)mocTBf=g;L&PQyQ;x1ZGIk)%b{o!2~E=@mg#%=s0WG-Q?-egP)VX|@v
zCZz+TT<H@opnX38%sdC>TB4Mb7y>7*`HY^zs$W~gDvuYoJlu^2e>J|lxD-zeHt*5(
zNiN=Ox2`aK(A_DK5=<hu5S2n=aRU~*si2LC?fA5>+w}=8|L-`&v_~ru361o6KNZ;O
z0&Xhe<spFCIK8noyqH{eB4ltTjXw=E^l2y16W`A3H&Wbo-MQt!*f-@k3GG|o^w?<V
zUyc_i^s3g*UCi)$wIx3f)aA)}&pSRWLrO|T$+-M<;h8DHN)#VZkROv+%KeniN?qy*
zwwELp4Hw%Lj^pGd>grFtdDB?YkxD@vnk#`lqAkrliGS&5(nnMGi<`zZ#gQl^=>oOk
z0tqD7K7ORPgY?rspOh~fr+m5KP>Vv6?3R?%rEK?SCb^{?Q9MQ<Jn#M}ZPZB&N%5&h
z_Knn!Zd^|}SRQ5jPH;xkMOLc4&^_koofzsIY6acq41sOLm_F-DQKSX|0E86YG3{@G
zlgG?3`AZ!G>OT+F(sbX-#*cL;iDThaM3_4pWiD=Cc!aHuRqLJ*qE|lWADDz^Ee*^%
zEEvkb)_x8;I@F)<Jj?G@(mM*ekdJ%DE<Gs>5~H$YcGDKbRyrL8SG?9P{r;mz0vz#3
zpL|IgB92(=1A-1VtF`KUj>N04e6FH+2I3bO6Bi?`?A%YB3FjbJXOcfyef=Tu=>|(u
z018e{|2kiLvWX8a+MTv|Bq09=JR`}$fkN)8^5eNelCaT$)qQLIniw_Lh<4{7*t<lM
z8e8Tx)Zv3RP@P}l(d%}6mpDxUS^^EQ`aE==0j>e4ZsL`{&Sip+KINz8R!6Pn{7g=%
zRd5@=Xi7);cE1<g56<p8`k<+?{;83D`?7o08O0k!_pq0&BD_~cnd#HjQe2%#4Th|X
z8$ji+wu{c>#UKQh>l);D(WnH_70&0N!}`b1COBqR&D^6-v4m5nPG87M%~z^Ucgz}L
z4i0zvMXLTi`4ET|D%QC_!GbdS?TWzwDi5dE5Fa`A_j~`R6MMhP)tOAl@dv)$J=u&m
zL(^wgdbjc=a@R1zkjdla%=PVtRX_H|40ph&xH!!4Yv61D4Jz{G-jPmt{P6NU(IGd{
zh<l?Pk{KR7iihNS1)g!Z8OAvH9Ke(1*g61np`bF)9<Qg0wilCKMcs+F0JE9CtkKWZ
znxQR<CXfaFZ0!NP0@3w)Sknd6AYA#3XgB)oGV@<^APHdF#n%U6SzKwuA$}k?R%F{&
z_0G}<+r@e{w^tz}p1Dn!99c@qWhnE=NDp^Yl=+}Ql*3gDyVPnm)1E;-b`G}b$8b3=
zIRwd#8xR|Bv-fu?{JLjHj#vvE<F(sT%!x^z3dCvsxGd!(Dp2KdewQrBY|kQ0AWp6`
zT>%K%+vI*`A0&`EGJr>yvRl+l)O6gk#@jbluC{~gCzRbuRgZF{A>i{t!(`k*;cF#L
zD{@VL#2K5P|JlKTJX0hiSmF8hC6t1Oh=}Z@A5x7hvwP5Njp~c<(@l~&FrVxew*)Q$
zU8YB|U6-EqIO{DG3>jLAknu9B49W2=Pc@HqFfqQ>RXwqYwAwG@v;v<r5o6c+X{Vj$
z=SeZX6t{chkLcB$7|ang1_nJJNKZAL{xbZrDe@HjthK5%W066fB<q9wSQ#aLawZS4
zX7%o%v%Go1Q4ndCWXNA27EQ!N2I3UDtuKVWzU%pqmYcVYXK`Z@l|@^{>VXNB&hri?
zhFZ<tKh>M$)}tz3JJ>H(T1uzW?>>DpX|VOYTCbH5^XQEa1Oiq}3hHiuy#0nJb^B;n
z-9e6yo7>a-Sh5c5IR7|3=9kge&Mt>CNzZhTzKc_{)5IO@nI6HREuqEYnBW@lTUy@W
ztV^64n`K^mV2slOcd^#>Gu>FA)WbOs_ROW8I{7fmD|@;To^S{ln_R8#SxXGqtLnJ;
z+Q(<MwSnc6|BHr;QRvXL**yzHxxa>ciN-fb;}h;=&EAr}z!lm%yjeX;?J4GT_M2lh
z+6+)dnM?A;os+1LlWAXzs;4WXd=OKL>uBmw{3<04t&OOe8|Sn*MkS&120%3sqxb;?
z1a4qBP70*gfUWqdv=5|>?ScJ)NLoSWfQQkYx&-W-kzrP5#&{H)t8ZqoJVTrR5%=cL
z3701>^34_wRks#V|IDWz_MEg0yVVm#{b5-&1RX3gu=(EgZFGv7U7w@tCEM@H=1TX>
z-{Y1)I@?$mI2tGBkEd*uGB;+4ddxO(!dUkllg+cD<+GX`yRPgI2_wYOMxJM*Q;!Y|
zlH*736<Bw`Lo`-sFHlX5D}cSNv+h<#5=My-UlYHpgSIu<l<@bS_-bvjCb&XacV|}v
z$dv?L`xujZu$pct=sZNa46QsrY^$Gfc8~>+x&c^K;6{3F9HJfN&f1@Tbzgye`2%&<
zBa5I-{0Od;=wzKDmC7_W@fD{)$_tAIk$&f8YVd4kV?VesSFd>F)4pp4{q&s>eJ|kk
zms2N-m&#te_KxzKtGQzLQ3^T_N(Vo5Ig%LTg>)>_>3K5beBMAgv#0ZX;Q9ZDy*H0*
zD%;jYsa?CY6p?KqAku9M1(8mAkgm!GirxYRAwZNKLI{oYO{z)(>6N~x6d_^=p-V!K
zf)J4=LLh+v0TTKEAwYnT5Z|))uDwtF-nsXjd)_<ey>ovze`L)CS!<3t)>?CoIhyZ_
zoj;l7Q|-BcQuYv5tVF7q`v8bD&w-Gc@qYZNPEpQDO3TsLyUeq0DkA$bBC-;`JKU*Y
z@)$;8_Z@&PU<^MIOJf_1oidgO*fz>#lZGgznH<f&AIfPsBD)RR7eAtqye~WLG*rGQ
zEtnsmB6rc)Vxjgjv3jJrLM=zmlp%-LafzvOxqkJKjTkiD^s89bemAQn=7EhXG&vEM
zEfz^{&2(P{>A#p=8%9k>wSlFV89VC!d$C9P>3ACaRM<$Gx{SNJ8Lj%@lYnd7tr3im
z?xo5Pc8wz2oYR_i-;T&mOvxu`g%Zk%egnDyXhVKRTa5jogTppnh!T%o3Oj$-XhsRe
zOwhRky8Ad6k#?~NgO}ad>h`Qr-}N|z`eS1{d!MFXCa84kboht#yNN~@?|r{RiPU69
z-&iZl)Q;w&lF$*u8a(57baY{g$9l=d?eoP3`f@w!M3XQ2ATc}}>zaETI{1f?YRN)Y
z`Bkd@<n6V02dRO<5_Iopn!7ttX>@j04w??J4>jUN?a`aw-KqhNZ?eO5TMD9E+*(4t
zpTBo>2p^sCVmpUz)){ObOQR;sql!Q783fo&Wj5wDv9Q@n_GhGw#-pxSdOU5-1G@wR
zG~PYrn$85Kz3uqvgdui~{`Ro6s{6s#wo&M_egJjTH(XUoby=vfY=W@^MYvAP^kj^e
zrUz)k{JHGmMN*8nV*i`%!!}Z-`()S)Qcqi=aXH`x)$YYuM3o14m6%*1rQXs2;q##m
zUiSZ-hyO7l|4$lUntT#~?x6FHxi4mrc&9#yN!SOY*y)fnS;j@;#x-VgEKJFj>u2&b
zUF{B)$Z6Sq_XY&@tSv;k{p+jF9tTc*{I9_#3V~7zA6%e=B~jVMf?dJJ+*cts+4*@Q
z);gL2BRdC2#co}@{YLtTJfcck6G~g}PBzu@q9^;hi0M_Kp2P2_q`z#Rv6>5<U{zX#
ztW8hiC(>&JTyPq04&zv-2ae)xs}h1KK?&DdTqQ&CJ{T#K`Gu2rHO@g&ay+ypRtZ<b
zU;bf$Ut9+Vm~?09m)LySdm|yNnUMYFBRP_YJGXup{kn2VkPP{WHtgM&l~Y#~-E1+g
z&I9AzY_+2sX|>I?=sPOwK~BZ*6J!4hob$`w3TGh5>?p?ezx^9pC#k}4?DYuJ!q?UY
zbtTKxX>|+%6y{#swm$K)=r2GB{p#cI8)ttui)eKFtMNH$zvc8PP5FpC^$972NT@cC
zcpbu^{y?j#>_foI1`fT8N?PLRU-;FANb{PW+oWBA&c2+QvfrH=JrVVX0sRMd&i)4h
z$G`AkdNI}?hxl+VZYyx_&5=|)Nlb?SS~-(Ig0I0|DeZ4Dj$F27H$C7-3O#D?$m-hP
z{aw62n@0G&wX8TA<AA&de7nA{z5hOjQT~D~JVj>@)>>Y&ABLyQXu}hW?PolK-Oet(
z2QSp;aqkXE$ie)Y=<~Ymk4JXJ=>#&sE!(-+>$fL6N%)OE{z2)KJVB`%iL?b4BjRd1
z9))urGA0ipkhP74d6p+`2!PBRjPWM#F@HFO2XfjZw1AYo?G__ahHWZJd)u?$1Xz*a
zaGO(QP=E$k{@OY~gIoKV2KN|1gNu)71!^eAUo(Q%5r6TRcUzJqf8(K6yBA=iljEWL
zF3=Dx!3-}k99!vkSUcQtD-R~I@^a|l3-wb`+?LP>D+(0F6>*eF$UG;*E=~zRU{O2E
z-dx3@lai+?T~9bJvVp8ePpGNXv0n6}EcA&F&?VdeiLP<5Z{@PVPu=6Acf!N!;@|7F
zM&V}l+9Nvql!KRdum?=t$sXFb(3rW>j@XBp;npequdU0>tF5)N2Mn|O4Lys@=d}Q;
zg8DQ$<*l>}5<YoX)9N2TIMF_T&Uy!df{La5=5PlX3*6mD)B(2J&vdu!<V<ySlPwCs
zs}0b30Y3KvxEf;a^Tji88ZUdcm)XL#T>{m{e&5I>f?P(i!vWi%{^XBraNJ6n=7vF^
zTa1y9ke#zg-&&qAT~5cHpP;|^jt*mVJ8L@4I(EeEW_~P_+wQ79A}2HI^F_B=JnM$3
zi29Hn+lMiFWnGmL9M`pB3$oB&uzz0$*7D}&2E@IvUsMgWCY^~5IGXyIwx=71O<9^q
z=~GdGnSc!p3cZUC5L29Sd;kD5E3#crjZcN`CQTO4$Ln(BA~$Q7WkRPfb5jY=9iD~e
zkq$D&!_;0NdX_@@_lBn8_C-CjUOuZ7k~`yYkJQlrAk-n-G-|5@ND0SLxwf-FJowjX
z@g3ti=FZjlLH%`S7sSJ<USr74!$_t<tgtI52ZQu&+Tmi#z2jUrPK~pijEH<$y)u)<
z!0?%_0m&a}p9HcWZsdd0@UT8evF=7pOsB%TBjeQ+9{@;SCBD%cDxA4%qcX03Ug5l7
zYv<3EkiCyh-Tu*mf)Xgk1TxZWe6vRVTq?v;@Sd%fhh2@|?L2>scM-~|s#=R^cD6LN
zAzWr=FmO+1d^lt#vuvSZIQ~ZAvc6EKefIQ#$>jRx=HgI&p1Yw#ColVK%`K%f_nstw
zRasHl#{k&Q*xF5DR~@I4K=!9qV$UHDt5ZcG7MyZTV<iB)-`G38%Z=q)_xaKTpBvz&
zFZSfQ-}Cj~mOs`#i%^?vJBl&};0-RU3G)ZJzl=rC@Ud~J+<3_&{l3OfsQgcF1O$XB
zvt0Ki9$SsO-Mj<8Hoz+da24}6;OF>yp9D(lh(BQMfnWV6fBtN6Fd7-!({*$oDJ0fC
zYLqzdwB3dfo+g~U0|{tCQtT6_`b3#m`kWtL4#ei^XxD2)2o?#QGYMq**0zRThk?6A
zu-Fr^wuBxy@=jS#W08?2Gk0vR4SR*lw7GJy1KQSw;(vLj)1UqGP<_2M#RIkNH6OwG
zp(n|BQEOEUIbN2;)nPR5Uvu1Mp3#rcL)tzmeNu#{My|FWDM{v3Dx4i8VJN+iX8het
z8ioiLGP~Y+s?|GzB-_31C89zz`#nO$#%q`?j^GvIl$#USY1-U|M%o^w<ywgAjIZ`P
zC-{esA}<0DI_f1G;PDQ_>}%~^-B^Rp^)@0gsJue<URWc&)<yC>hvhuccz+0B>;q6t
zoeR||xXLQ|OJ6W8X&o%D%FnQ;_VNZKzi78??G81|o1YO*A^wza0ay@sc^S#lt=}({
zf#WFk#Wp;GEBu1JpQ^l^XzA6Kx9>OUKQ)Ga^0CP(F8a*mek^a))ZC9=hs|Bgqfq^u
z{FgKx@0G^)MKw4<uH98;*7BR!YTp@YtQ)`F<FUDD(Zrx&sk<7EGEDb%1HEmtGjD>&
z3!0{UVvG$ys`n~aE1su@W=%X98zn@Kbyy;|je1Ina|aZ{;6zuKhdV?w)y!5JX&_nH
z9{Hr|{9p`3QzT!Tbsuk{sq$@@#9aItI=?w+R(cC8M!VSU>*9f=v!T(1HTf2l>9F~7
z`+%1c5n&<TGPq}f*$g#$magx=(C8$|J*?ZuX>!)|*A(K`%U<ng5Y~hPXG;0_DgC%I
za5jN6l9EsGtraq|3dbxTOOi%fNx`L&VnzNdYYc{6T*2%2U4rj?x1=1Bg!*WlYcAnf
zqkyxAF;$8(u>4Ce<Ys{YLTA9FVN`$7e)|zr)#b;`-yGYr(p}qPslCpDJ?~?4aS)1X
z`fHEZi{VC%ejg_1>+S`_$trrI<DPXOO{qBw)!Nt2D)vQ^T3eB(BuXhLWvtqIgWZXl
zFh~LzddiU9+pbA<p^{oe+CiY4t?~W^0K`-C9Gi_<jm_U1Rmr~8QEzpxI$H;1FRB*J
z*2Og)3|Vof6G=iUM*cRg>7^x@o+ME@t`c=eG!rNg!<Xwc^+H7ao)p0jQe({9G6*qs
zEVJ*{=~ZiX)}g8G>!yU1tlHH&;$f@kX5{ScLC}y7f)24fLHDYI>O={>piwaf#Q`#<
z*;s^g0F9>}muDy|@+fT+d}fkfvQ(wMF}I6Po-PV9gaSw-hOQOGqXQBx9j~p(Oexrz
za?lfezI46U?6S6o+cQyvS@cwlIH7O1D30i0NbfReOdY01^5iM!TOeBFcoVR9T#@Gl
z##_bZT8om1w&OI;b>LaJ3wF6{BSxNyTi!%;4s93&!Y<VZ=%;=tyIsaqC>)rkdaBjR
zeSIAq6B_IgY5BGJ*WExJmgbjlgQUdo(V7m7aXrDY#|7~ZYCDt8bg%damw0EKJn}eQ
zn~|q{zV(1kDEC&rn_O7P($v(H49D)_uo@ei>+69~A6%EtFe#u|Vw(zY=r2}5XYG}f
zAy%Ca7_D=_vt`6bVWH=B-lj}FL}i=2g5HyCjW{qWb{zD)DtYYuVSagQN3VCVqgbI-
z+>D7{cPT(AcfKJCPc$^7`l$B=OFv4Ro{!p8%2+yxHxt6NoL0&kAuTk|9K{^$-=7|x
z?kxuS2(_EF0Tfr2?Pdj^>Eyyc)5(oZ5ULx0Vf{uYm-Kf!ISN20cO-^FnshhWCiygF
zzMl?t1k-BwCq4D;Z{}6jY7ok_xfVYzY`9T0&0#k5$hRQ#>x5_O4iD)TG3}k#tl~Fp
zD7iC>gT=!R9ZJ6LH)x^j7>4wVlKh~9NCIli;1pSh1h?TRlaG=^IPNN06Nf-Wf6=5S
zyL50b)$w%R5Vl#+v-8wB1bUU;c_(G1tdqH7b6U4$l$V!ZL`P`<@{D8VTu-YvD1eAo
z@voUBV_kwXLzKiiIyrgn$g<$f%fZW<E3M1&ESp6%^YGYACL)`x)8Qd^L-9#J#a7~F
z`ONg~B3Nh`#LWT=D0DDK<J`cZ_&HuW2D5?jKf85z>RP_{N=H)8T3pYgh1OZm`EY3U
zhaeud*4lc;YWzu43~S*@Ds3oZYzWnuGO3NlY`agA28zL|hIx605pop+esc?vGwqVC
z26xhmxYsyx^2#~oFAtjH0ej0Y+F4HCptqii>44tj_8KyOpIQH1+rKJMeAX(L{dux6
z3Vy%zS?5kSAKx*W+_0cp7#~*Z1xxpiyM5PEp<I^^hFd0S#%4=K`F3kxA*WxHDSU}E
z{33JZb2lxx21md;No9zud$t=Ie1+lVcbBG5sWeOr*L&A4W}m^w{nXzNO|GtXaRF!~
zTmoALNxSL=Q+nNzlzSoaYR4}rUHfq+{1i%n`1K#e{$gvow69f7gEhSWuA~&@5uvKT
z&|!IhB%;(_%|g4r_%yPKwplkheQQ0-`qB?ScH-S!ZWmRa&#49ID;#i0i;inymrnV$
zd&MfZ$aOzz2AQ4ksr=w`01GOXYh(9hZ*F`N_<?%RN8|fHRG*Y=)v9*tiT%iMcwWP7
zUJG$1s9rD4=!HChp`qv5pkI@p0<wKk)m48ZchDs=cn>5TY6^c^UphXC4tBJ+mUwP#
zru24N+E&xY<_y2v(phiXXwdU<oO<=svdHTq)>__zPk;Ku=>FGt{)-cFeiQK9saC46
zs~q!Gi`5<*uXV7K2U|os*m%+W0>V^Am-vyUS6e73&&FmFt}Dv0$+H=aeGuYQ2vzJO
zOBD4+KX*0{ThZ-XMwbikjnP|~zBupI)*TDbi+#n3rAR4fcDYFElYxMP7MqTZ8tMdV
zYl_mUSvG3`GYbW<I+k(KPGbg-J%`$!pEoS!5f%eA6p=W{pzU%1NijgO%m?2Z9_%@=
zkr`;%3}m%?7fPD^^d5Y#J90dsba4pLp4M3~cHqA=!tT3gZBnIaYU_v?*Uc}_FNycX
zV#kN})1h(6HIB>ody1!)3AiOXiWgOoX>3F_qRJ6~n#OYn{a`Pqt#MpFCYIkNqP~pZ
z0xC^==01Cz%m2*$>U8a%89Ej|9oRP;Q1V<Fr{8j$eBf`v){YYtmCtQi!l=P9v+ima
z8aB0=u&W|ozLU568Icwa<R5z%>z&rzS{vs*=wv!^oM0Jexn7nO4Mz8}W#3Q3k7s8w
zb?|PW`an?GT$4UJzlX6o$IE?71RuY2K1y?8Y*Ck)e-^qL7)(wX&4Q1r{cyN!@^&@6
zHkvDyNo|BW9v2ZSg(?RlaQTiijj;~gzHxMbtcH?H(ZLHWCn|V3Q={1-^ZBZj7t&oT
zUZ2P=6|{N-vh;a0uC*?v2_grJW%X&s8XE{DVH@$2#I%tjzsl#=!LH_2)h4f_s*3c?
z)}>R+H)1=P_cx?wRVG@fmdeWu`Wc~5rxx{%rir_pV$f7e==xIA&09cKH4a-ZicWcV
z(<07D1+c-y$`Wfk^QumAuQL4S@#sgQ@R&@oE2P}y@|uI1sPY<4KB?Wu_xH!+ziSey
zfWVz%eD>R_{WxU9{-bJ2dd$Yuhh^;v8+p=tg##n7(#9O=7SdOojLFgt41cX*e!iG`
zvO9;p=YKVW=mj%=YiS~#q*?9*!|R-Gu~GUc>!jZv%{yKaZCn(g>gcYRl+dwp_AN1L
zXquGbPSbE6v1xlB;MROllPzW1caILkn@OHK^4)#GZLFWZHe53cW8|06wO5y>emtWY
zZ)-2=F@$xO8{RyBbNC(nwI8@#qBBlaV*2LqtF#c4Gb~xTovw@iQfBXp@<5RbMhrFB
zV^Ua+_PM)lo$Y7|L>bDzamZ#$HoXp;g}tGc|K0)HJI@LyM_HB5)EbY{Xr(GM1_;Mp
zCNeaR*s|js2d`mo0I9GlADC3XN8n>7XIvsz?_pg_Q9yD()JVl8DtteDdwy)Jr`Siw
zFoMa7b^9cs{oWtg*18TIO=Eh{=oCZBe!-3=u&}gxSYu<=0@KE)O2e`*AuT8)FVzxc
z6HU|Sd%Ewciw3H5BaF7SrMztmTfU*B#h*tlRw^#njbKh}ULI-3jcWVjbnPhPEeHUc
z-{S+RV?yJO@&Y=o`#u|X@a~{)-Q=w6VOtZvvHLKTqyb%f=vJS9F_v<FD6$rRe~FQ?
z|2DgPaW(bSQk4=IoH~Yols?{HHnF$QIl)>*BZ^!FpOo19LqkEVOa&EyWj28NOw-!P
zsUX$S7GUM7nd;rs7}{At8BJ`>=idCz<@2jWAc$$aGb<XFnH-@SxK#Rjz=l2}DPe4;
zp4*nD(A<L3dSE7-&~tA<OI3q=EO0OS+@QS7u+!$YlVp+DnAc93rX}bwc$FN)_GsCP
zJN_bw!x<i$&|x4e5Q*~5EIYZ`#*U79Uyid&VB*OI$}E+j{T78v;lN-+Ps){wmzt>^
zeb5~#?2z`uhpb_%78N!Iv4u|(lG|xm9*Rnf47MNI_H1&fXCL}#(jsWMp?JBNUDE!}
zZ1BDtCIA(ap*@e3<RCD4{y``htQkfY+cCf4O%2?*Vym26l^0-xh6D#ElP?3vvV9=^
zMrOCqTk`b-FymRiTkBE&*szDKjPW=?!udjD5Hy*saDuGvncV-5y{?aR!h8#98(8d3
z4bAC?<yIo_Z<x1_EEl;;>!h7H89|WpR3^<2uAQ8zT}#@irs?RaIF*YS2SXf4c=Vag
zfqX6ds&6Sj{cSqQHW7lIJsq^Qyg+esbza*En>Q-H0Mos^Rc5K+oF{l6o7wu@(Vc{-
zKDnzMx7|l7X_)ZYfG$-&tGu45b|3qmezC%Wayj-+?4@Wp%Vb;)F);u3%|2BDbTLeZ
zi>%JzB}wKPaTNPPFfZu<xA9OKsc@1W>Y2&_%UgHqs@9G-fK|EyCXO*TkLTO!*>2=1
znvNRygQ2Na_PJ2<R>|QGfVUE!`()1t<*jz++zwe@FNJI-)367AVi4AIKFMG!alqDd
zB%JG1g<yIiphe_Hoz9e~I7h$U0q6DuXz^EJy4?#b_LXsj>^0p9Zt=SHrpbi}<nn8N
z&OZ278(DXC7EVL;hRcX(-8#w+XqC##fQ2=+WXJ{jWwcE@%B2{GSB?8w#%Vy-Y2<10
zB&o;naQ^{~YbJsm3gVurM_|B+a?znjDzT=n(2`)C5`%M4PK)~BpzRVhuTzP5sf88j
z_P6l+-!%SjezT(n<>kD)m$Zi-N&6MNSwGjFD5Cr1{f~J|J&I30P#o$2Mi%JWbu9)3
z?Bs}ohF*^}wP^AuT>O|OpIPu=hjxp|+rQ2on2dkrqr&u~u+$Wm)hVukX;-T&v8gsR
zWa~}q$cxVI`B3TlC{yzq)qy#p?F41k2BPGVMxf|t$A9<^c%L8IAswgy1?iQs3b%A~
zfM3Hjs(QTAq2yy(sYTeuv_@jJP1FOx8+y)K`Qzi!zB7;I{xZ5h*|(Rby*bK=HoFxw
z(s{dQuj07rfWv&e2ZgCP<-wJHf~&lQm2{wJI^MlhaigTB+Wb;NNAHmZ_?Cs8GpQmt
ztm|&a^SZBw-uyvHXP`9pd|Yp*Qt9^;U@g;Pok8L#ozU)Ed?SSl_R)?R)>>3DxY887
zudDenD`ucn)3@Rrqj*y+=St6oUs^e)!;jl!gk9fzJd)hRhY<@73y0d1$r<sZAtiAG
zQm%C_j5cpf4q}Ihws2CK^iq$=ilQuqP45BV9x%{^`nVEpXF`d43ahh3`>>hr^yu+t
z_0%F7CdUnN4!pnW*)hHGZZsY005qoDw{fCvP5bg)Y~6xHP%B<~nwZtpaOWU)_RNY)
zO#SFimeYgn!H?Nw<qHnK-urEm`@=~8wVf}ne-iLj=u7Q%6N^_X3PXTeG{_g;0aurG
ztRDxYu4DxhH%e|adDZG4FQ`Fp*S)h0%N$yB;&P<|uaWjWMl=umd@h(treOzWJlB<d
zC%?PM=!u}TtsD=D|DMk8n~o2$dH6n7JM-<fcgXC8w{5C6h^lhFb*M=<jYVO0m5Gqj
zGZkn}Z81*B+kNy#?j`c-_@74Z@e&f9`vw$KI{^MY4Trq_gY^vKrS!%0qx2a?B?{GP
zW{O}(T_Y_eScIWAFW;#){Y)5S6ZNv_%#)~}{<U!efBP@4yZgHoU&@wd$$-S-Widq`
zdiVot8<fcEonBVs;jY#)Uph9Vw-&!^3SVsaB#`e@88Q7=0RbWDj<OZbfjxVlCn>Je
z^GTq`fg{UvZjY-BrqhMT(DU?brvS9FB(ewp9edZ{2pVa!$bs*6uK7RJ4}bIbZvQ*`
zeFX&W?(~+w5d^P`^2rnYe9C-NE#&=BM9Iv7f6ZQ`wRo;ZX!NF9$YJFIpVsrkY(uWm
z!hVzLQpGE%{`m&#tRpETH)OJ;&zUb|JbK!a=o&9O3dE~JVWCHY&R1Ey<588N+tNvh
zH8Y%;kgCU5@!kXzvKC{b_Edp8k9-Dc{uZ5?J62VekrT$&9Z{BA+>7E@x;U3~nq#5E
z^4<jxTa?Q|U{%cWlr_MuygUmU^j}{n=-5DIg)9x28v($<y2Faon(cA@Me+FOuxq{R
z2S<<Wr|R|(r1;-%16aR`U=Ix~=;o`#)uBuNqVkB%THiZB0f~{bsNYzzu{5kPj7skg
zKFWU`S_1Z9@USP>S?vVD4)lB0RI`ev%u!s2OJOj0$`@;xn;OIV%zKsMS&2L&5@QyT
zP8q6r@4xOB2%-9wgCJs75-c~9bDf45RK;LqLAi{I(`P2@ODsaz*l%}=^qW=Otu|RI
z)w}|41RvSdyxN3>>dbU3^5w+wCGhm-va5-<!IN}o@bx|sVipP{Wt1+J*k6=za(62-
zAlF&jGwc1~B1>=;)<baU@?|U3d5n~d4M-`tj&fq+iW{Ibba`t~tnCm!+pxDPXM#G>
zkLq2qv5HvG#rS3AVeH4R!Az$2={Ft|vG;07;W!O${fh7CEYBY1Gv!Ylyg8(9rxD=2
zN4ZMt*DgpuW~dc=rCF^f^G3CN$4RM#;VB#A(k{7BZ;-eJQR!)=1OS<cT*Zovw-^67
zs<P>4hm!K@Sd2{!-%dD{(fhQkj35GeLTaO->Bq;^Qe=9a7M2y9cK4r5b^<UgO;~2J
ztpiIO?*=ZJL|rx0p3tCmtnRlLs562OLcGT5+8AVUsLuNQE|(Z6`P|R;eF`(wK#$hO
z>vIWqX^|Ur4#W6qMvW5zYV2)B3|O0{;nT6yj7+gRBN6-)pzbdi+bu5Mg5p`j-rG9&
z(sa&u?<nPcSg7@#^r$Zw4C}BQ;ALFjrum2O<vD~mL0o30$<>!6qt7Ccc2O7e{GLw|
ziBk*V?8qab0B+mVl%CdZ6FX>Of-*HVIbd~-LYYu^m<$c@8H#rANtK2Ih!Uin8j8CD
zUEV4ZAa*Ohr0(8pxd6ZqAgv-<qm7USfSlObE;kv2@=GDrYigc3MQius-3~4LYYD^A
z<>Pjt!RcC-yv@FJZ^73?p<b5Z1Ec%cC{^bPbt1PlSHo#i&VhWVsjK2rUrV*E1y!t>
z*9q}H7<wLmEdCq7KafbU_4^R6Vr`x2Xi!zvNro*Lx)b8RH02K-mTIO?qFXWJ^bOI0
z!6kA#eNn$?_Tu8?=to-o&68~KnAHruu;jdxhJ9^t6f86S?b29!d;L<OzSY4pzND~X
z#?_ztE#<INX`{+zO6I`RrV{b{y!Q=xUX<pAJ%|K&V>r_p_KKuotA0AmkX$*kPtNk#
zST3Gh*H?3{%hT-3Bfq*{qclB&$tE{4nwVjXWRD0P@e5B9WzWM;y6C>cBP2!XY>c+i
zowcH3P^ln5Q~B0F-`zN#aCb3WUP)}8(}LGyc@7!tmDx<_E8CxwgM-4C4#8`u$+wnP
z^&33<--PI^(-zrsL$utZUVWkH`1De+G2>Qz+qrKKR5--MgQg-9*6P?|?OgbVy=M9Z
zCayv_Z6a)%FSD7>SI~VlI63m3)-W+J6@_>n_26lJ^YaM5P{33oJ$EQImj|k`RGgyR
z)UxI{#%2jiTH(#ZpeNKRy1tEWgb;gQkAK6=h>g?QV{hLIc`!km=mm(%9BaTb7S7lw
z1vy9LiMO~%r%AOgQEY6m2jP*9(k+&EPu{awr(kNMc87-DI0h=Bu3fOTANZ)?clG{z
zwm;sO(%o7qhtg}mWnH1C{dqjp1*~K6$UxNCn51b7$_$sn^x{QeQF^0!77tg}rm*%3
zFFdzJ>27=bD?e9Gbo+;&yBt$D{F_-DPudq3i#V&gp+v&8Vo(+|$`Z9^H&A8svaO1<
z_KHKL7Dmyb!p`Z69n!|nCo9UU%>235g>zq#crr$)PBR8y&5=P)cCDQmj=h_4psa(Q
zj8({k6>GV;Jk26e$yqF2?aO0}NAcaZVGeOcx$M)H^-UEqcH3@0bhgwtH3<FrkWArY
zBZM>7Tfj9^WPJO~uhfC-BFkDvi9{9qGZCSdv;E1GtJK#{8#hdB^2Y&l&qh<Yv~fEP
z7r7>KUE@*RPybS7f9fzcjz4xNo+^AYK2UFo{8d#1<VleDQ=0VQ(%E2=RO&CdBD$Po
zLW|r5<)`V^F-lR+OP!^4uG|l;VuTpy!?cfYys_=^w>4ZB-4qwfgce+To^X=PCL?FF
zAD)^awYB(3#$Q<$s$)#1>h(}37nD^OvTB}I+eAA$<ZB%*Rvx9tK)C1kbsTl}MZe5^
zBOvzYtrJ@g3$miB@7=@GcfKqkM?3Zu^1?leqmH3UXG$|TF&SvcQ|#P^xT020wWm($
z>m1(!43*9x2#MIVpo=<@H#D=y)C=dx<q~Q52f@ueW0`J(X!5%Wiidh+dl*JiIbK)#
z!Szy9dO~_>Q_jYQ^7=-#9S(ziZc}>xtYW`KPFDZLhBt;E|D~G#hwWp|Zj<my-%*qI
z14&&JTisu(e#%!nO_2N~@E7K(sr5k3pIfq+-t$~Yv*UX<J^w;n$JWe-#9AL_+|c;+
z@&&(wkE*q6*5}Fk#=666j^y!0gpPBS1@(t18}Y5JnZ4S>FQ&_%Mt%~g$H|stmukoY
z6``?(CAjxNfdkpqs|X+`z=g2Y+#cS>Tz<^1@aR84sZvS)W9rT8{Qj`Ru<x22OfEdg
zt8e(=&J5VTH<(-5!8ef0Y^jLyfO|w(SnT2QBEomZaIKh@=XRx#m=w_~KcXz6uGp9P
z$v<|Nx&N<^`*|kj(kB7g3}BQOr6rFVcZlxr2BpJkE`iJ56Fy8`9TkrS+$s*!j@&%J
z(6J2lCpEmN1I{HV8>0sHl*=~hS9oSZ!5^y++rz;@j&}gwk@&_>0`_s=gS(QzYs*go
zu1etKtHu)X@-4})|0%^Ocz8in09VQcVw9hBR4DjF*Jpm2Aks@<(3=o^=>RL|b@)vm
zK1hud9xeW29yri<d411}(2e*BU^7?nT6zyZ@TPiy5?~c|#8uYQ>4IbMd0OjPz<f<5
z)A{dqcXf}T&9-}ZHGbZ-wf`~IxqtNi-?)Gc7{j$6r_sWJuOoWyR-(g&(>F9YHBgKt
z6r2@C$zgBW5-Y2<*mA002I6oiw;16VyJ)ud`UV!_(AtuQb1X0@PPyYx>@rT|Sw{D4
zF$clfyX2;&(GkjkOFPbCfe~+8y_4-s>Woqqmv@O|G_Iq9$yxO6m5_t_o3yv@-Fst3
z$C4l`N01hV>FLnxg(37BGU(}B^+(Ufcb4G36qL7jnKx8V(@_V5;*fhFH?8i7xAKP&
zUXePpp0lll<~i<(K0VN>vT(*%MJ;_Rv5h~!^MJ0+Qe(Kvx#Q9%(0thk*HPx^y2(LP
zu%biB*!n|Cn7>_xae&b=uDy{G`%49JvXR3aN7&lHn?H<@PI~sS7c<cclWW7|6-^ks
z#kuG%9e@JXJu5xHF_|-?W0fzsQj66fYJCJQ`pXHu==DJMQM=)?bYfvB=rqJ*bOC_-
zcIt<PX*g?#S4e;B4NKKgZt*ts7)vgAjXj0<jmcB+{=ttiChd18_>#2RiEy9nZSLvJ
z_u+T<j6xIi4ol<0K{bIUy3gKqC+>HWi}at2c2%dLSp}}fJttxyHvXZM&ERUS-PH$e
zbjc05O_}_N4EoA!rIDp&+VbIWY%ExDHb$j>JFgehh-U=a*tC&@Z~Dcsk(kT{SZIt?
z;whj6I{(pLFVs(M74gClX7Y~JQCzpJAG*r43%?qCUS03o3D#1lTHxk7e~Pw~LFG9&
z^+dmWM}`d!-s){GPCCH-q7`6rK4|IZ*R0y4pDH(5(VgXqm>%~Pf>y)zjzxcmro}Sg
zsfy9cBE4T^uDN<?M5vcBE<}We8d)?x__8G;aAr;R{_IBFa{cxZvOdIdKQM$x-Gpwm
z=O#DN-&!tVkm%}PRJ6uNNwq50R&G}r4|-EDFD=S6Q|OUULl?M61q#1<Kl2<L6KsE4
zuaLFb-s>*UQKKoygN7QsaLb+@C{0y!JLY%4J*5tU6iem?^Hk7M*C=0SV92X8Tu>;u
zbP>c=Dj;^g*+QX?T*9PI8BKd3O%V#q0S)KcTC?t?VYHzT?{{vXG2Fzfv1}DtukxP6
zu4txVer*TBLX$mf9Z?^AWCAu;3&+XLYC_093UZtgt$sHIJ^&Kv{*pXn1J0#I1TECs
z=&+l@+1xg8GOaF5)kLi^qNF!~SGo!l?|J92lW{dpUS8qZkR`o|Q&0+xOLMFt8(RGs
zT=$FAKjs*LHnqlEvu)|j8<VXU`X&4}b!=4dp1PZ-J3G+~)H8Z&P3dK5t04wred0Bl
z7|QXNXczt3HaOT3JfRVuqd4Urc3z|BPhfvsPlj>FL^JwHJR0HR0uH2*7-hdg81y|N
zsOiC2*FxZ1e>=b_ywe_DW_|j`#XiG_Ky=?~QYAmMs9!N`i#vmeu=J>cEfxm^H}$vm
zP<Pm!2@Ph{CLVzVtu9yg?s3|84)W>$A%!;n#Cap>X5K$X@Nc)IT!-2<@kYrs8H>4E
zSF(!ECO{*kcK>5Udic%~B8!Z*3X6sEa+C&y<2;_?5MWw;l%-Sktb4;l03}-##e`mV
zc&dJzrMY4z2XL<F<{v(P-#`U_BsQ=HgKL+AAzN5p_PZD%=l!B%`(7(~V)e+Jcl}cF
zJrFND+SG|=8rB*{yE6R%Ie7ZfPoJgL&;NxhLU8q<LiN>Zr}6zRm%5z#RZhl$&jmVd
zXsQSMoNREYPJ8b%hpwr}?U^Rn+;f|62$`XwSG%=f%8bz>zi-97)|RoCXi?pJ1trI?
zF-k5>sl@j=-wE82EJKC%wLg!5ntgD-<F_&G75s!a?>OBvGv4`ul?b~KX4v2|<Lc0y
z-fQ_&VXf}oY%4PhlV#X7a@u#4`CPfK+BTizW46F>LU5iNj^3SQZ_1%A)^nEf%wG{C
z;{D_dXN|rKKTVi&*%~Z++~|32nc4;^0QlG}2sh^h*S`5Az{ZKTNoQ7@9O;aH`Qb3E
zTo3^L-k$|`;N7|u-x!zrbbk`qH`*sNgURR*LBa^9t*#?N9Gg2G#@H_Ta+MUT2E2)9
z<Kp+Fqj?B{|A)ta@qq{1NQ=D5lu(kCM#xe&rEqP{SqZ~F&-6T82i)s(mWqkmiwh;@
z{jtxBkPsVLN3CX__QtvwzkAq^|Et*XHFRWj>db}(Gkh)LfgYh`>)!h*%aUyKiOb{(
z?Ueim)5=nzHk+=j9@v;_;Oc<Z9=6R`P$W1QsQbxPoWIEW>2HbKH^ozAe=#xRIc$aF
z+8M7FnuY_&TjN0#TX(%aOqp=jR`}JTm{dd6w=9SJ0^{$-P-yQWj0zlx@{acsTR+^e
z2RKSwVt5I+E0|tKX`cknP&Edo9T=JNL1sQ-CpNDT@qO56W<twJXN^8!X*A%^ec2+U
zO0vj`X}Of4o1?+@aMSFL@z?UoYnms{_)OknH50u2J+?c1N1Bl4ffr0&mWC#3kG1yR
zMVuR0j<Crn?%HU11seFePd%?nrj(r17!9qEwUR%vwPIWEg9cDRBqLc8K;ZUiFzGM2
zHuRgB(BcO@kja>8@4T7-Y^~?S`c+BSSS`;TY*(M(p*J!gWjKP$m|3r^6n-Mu9khIk
z<>+=WU^8F#%}4EwtF)Ofh0VT1tQ^LxCMtaOxNjtFTs|uSb1oV=Lv<NL!*lwncUt<F
zG$M4PTu_Tk7TarUuR{J2uKk}s|EHQd0LNm($V>RMd|`M~8SE=-0fBE8=O^iKn)G~e
zTzX%Su7>OSv<WogBuAz_WI+wrDcMnBbdS`ENP0WHprU#HtMM803g=~oKF*BA;bD8z
z(!ia5P`{*KH(y4wbA0wYgY0c#**zWlPF85ujrp2*;cU%3&S{~^kT4@4QF|Lw$QKvS
zAjk2x%9;YAOhhZEC_~PPy3*y7s$Dj(#8U7J${HjcCo9*DGE0f9CR`WZGk70(&SC3;
ztLAQ`?cq&obl$!LPV0(23Ko355&%=Ry<`zO1jHk5{Hu6*|3UIlG2=bA&i7{W@GRSJ
z;g}2|cg98_A6r|^=N<X*s}6tZf}idhCtTn!0c9<tu^K$<P?0<T)V+iN^y4;VFEXCg
zQv}$_N(H)J^YE9`w)>hlcw`Z=De#Da?7$KLBip!KR{Rs-2&zbGB+8nctUdNg;MmAe
z&?kYd-c~%2Qh_e_i(t<krt%EU0LT0*-YNJ*Sy|#I0TfP<Cn?o+=okXuu>R$%|84AK
zfrAVy>;YA8K~PQ2By}{r4Ge5ep1cK*Zf}1Q5JoZCR$X#UVF30`QvdQnY$x1tU?B|s
zNuYIgTXcT~X1{+Jqy0%BJ&^?<_+Id4D7ym66(B0TJ-tlhDPMc!NFa<<ZR$|rG#6y&
zBnR+s8g{MtKV$8uF7Uw7leRrG!&`;$9`@<Baz40l@hR*F73S_N(spfZ3eB^z4KL?b
zJKR|O@Y{u!5i^r}@B}}**nrDB;CX9&V8}h|YC~S?s7z1DJ+TQ>Q?0#X>iiJ8fm$6y
zxnayJvq}}^%Y3_?Jj9J_v3;Avon=+0+6-c%SP*~4{mO==i3vhhbKU)(jC!5^d>u=o
z%rysjQpbz>B`;RiZm(5gh+8v@UJ0Zc7NGnrZc@k{h*eRPyEg!zf76Pv<#l(~=HpV^
zv#AW&y1d=y8dmq=jD}WIjM6LHixW0FZ^k&nnRN|QLyMO>K-TuMX~maLK3wT$YJwD=
zAuT|~6jpJNZIbN|yGBC0;9sNts*}WwY6X8Gb=@Z<^;W<!A$NJtCvnRv0Ej|)!Q^y0
zI1a7%(61V4Ijj;Jd~Ry9G*uU}&co~Hk+4P`?a!($<BV-(6XR_%WMT78RTy<1;S6We
zN8Zw`b+2&k<X-jt-*ou-OFCzD-(xSnVh@rSBT_xq3{>);+ng92w&|uK+A`qwt@{O~
z<a^b!E>HI*bt;|?-G4nKv6y}t%w3hws#H|S8;_n^g^MuEDy11j8UN-AFgC~_`dYER
z(q@K1Vb8|&T$H~VmdK>FprEI(KRctI6-LWTMr}B~x`s2UCm9Hf46hA!@WPi2TNIsG
zJ8aal7`2_Vb4Kz!ZXl6(XvFsbfD9dld_1}XuTeX25T}z8aiM<Q?pPzZ3Pj5X(?747
zC&n=ZNG2n@&y*!>B#aBDXqx^dZES5o_HtAcV==OWFPyV*;O6KkpRf2XM6v{Y2~qT|
zraz3umC)AJZg^*P(ErHew3XwNLYBXk#$9xM5)6Ed)M>Fup6p5WIOa?!TIeat?N7br
zvrF0qTyTd~GOUlG|IpvRqj8cQbE8bCxm;AH2Jc0ex|@IqKHN_m3UHAwNQ15;7L-*i
zf;uOg4eocL&fk&q!1j?WXRdJN9CABx9k&~*C3y~C-k(}-Kb5)@X*G-~9ALDGDO~z_
z;;MutPd?&!VSP)V*H%l(-EojqnSxS!N9FY$*(|)<N3>6zM}%qmQ04gfCZ$n?L&^4c
zM)&_9?w9#`y5;(an(6%7P+4%xjWF72+p*1;t#1R1WTN-KD8xR-_}xW62(ue~nG@3N
zlGKuw=m8%1vGeWHdWF4g0Umshgo5e$2d{G?59bW;?3R=t51o<rxxPsjpL?~ad2Uc^
zlpfX@J{quy!?Z7SNS<-_<z<$=qu#ABd-|~08K+Y)$y5$^4T%bpkertPRm}4b(*2t|
z5`HmW0NQfJ4?xPRstiEzVn{t%1RV7zmv#d~81F{ACak`B_}G$VM{&|d?L1hHoZ*tN
zx8D=d<bym<zO!{;P?AvP@m*_MwHCx>;b7LF8Vq*KE1C2!REm;=>VbRye;7LbFGW}X
z#}&}{a{BoqZsYb5DqCNetC4R#aiF-<9Yj1ls5OSqkM$^nfQ*A%0iOi0n($Gl9oStm
zaH#Lp`VOAZ-T2FME(olB<C8%59vujrb{<v9l7C(U9DGN@{K}x9viL+e3Ma&$1gJzt
z+x!~8%bod0zyE!+-vk6?kOmOegjH{Tet9pgL97#uV~OQfN5P4V?Fc>IY{cvcVHy8^
znl!Y`j;$H%Drd&HL?M|ydF5UBtS%N(9e9Pxnr@^y{RC)ZwBk)5gT_;c&UGY=ui$sS
zLIdfC;`gK62?Go~S2hXAsTpBEXBXARz0m8l?K7*{WzlCaqag=;Ffp*SoUqULDQEJ_
zdrgkSaa}~?dK09qBKV+WS;6*bto?%B%I7`=zOG_5ESnz)_gFTxObK#o!lk*+_4B2$
znelYV)YRI$ADq{sI4Tg!bG8;Vf>}Fddu+M#+}-?zmA;Zu+eO1h!8s)&<uILfchyn4
z5iKd@<FXiTKu~19^u%Wh+Y*NcS)?KJWK_M+@>bAc1d2Xs5)EL|*<zt+Wc%P&XZhZ#
zv38>IW~Z?N_#GDt32}lT%aqjFsg5>N`tBYKPPwfJJ8f7Jq`v)G+P{Cao^FXQTTIjE
zi{FkegorGJpO$W!JkT_7h^|L!cNmFWVx5O3jhp(d4tVrM4ara}LaleqZSM7-@UynL
z=TF>AIgGKj<txQ~R|@rCV|C1mi>wJtTM#EDR3^gW!TR;PPf~60{&Wrx+jKJuQ>C)R
z@+)Mo;4}^Kh;~EIetk`r!^C`}6``=k&vW13{=ldt9(wIv{Hx57d5y2HwZ8dX&VSGL
z<aaSPZiGski27-s*){V@0=PJ6PqcBSXx`erTa&zWN=Ab`yq~vgXkDr|1j^Y87vt|g
zBnPaTqUGFGDoT1si{tCsF|&{K?*Fkj>EtzAAAQX#*WCH|(IEbfIeo#$`|ZYz$61xm
zO*M@=q%e6bVJ)iQW17+(a)ePJL#;X>Z?`J1l5|}A88$0Q9(m$d$<!@fb^o}e`O~t?
z9jBWDwQBw8(jvKG7TI?Y4llKjJ_+3GgX+D$yNvkSl66OUJIB|xe>KnFQEP3q%v>SM
z@6zeN!#?~`wEyAWzY<ya{M!>DTkEVY*6FRG&OW-HWTEO_pe(j6AvV)ZKSX?LR){)M
zRs(Uh@0r^S5i3mp)7s|C5mab4J$AD)F3!z4)a$h;Y9YO89xg02>(BgEx}S4AlPej_
z>DMP-*dm!6?=KvPO**#Ta;>Zg&e0zHBp{fIhjj8IZJl)t%Tq=-HoQ?cBr^!>`C!By
z!CL_BnFK^qxk`19a4D_d83eGI>v4;8^K|!gAHGCGqI@D?MwT&1P?4x23D^AGuJZgC
zBCAq=Da+UN!qWeJ<o_SW6@JF~Mn@R6f;k4p@v?hffparS=%Yv(5$?0TPXdMW?z6B%
zHONtYF+Yn?5WU$9lxmxn`(zo%i%!c5a}3~GKvqom@RTlJ<-RfT9KhPw#RSr1x*wVY
zkp|twOs>el$1XMw;831&={mIa(^)yy_U0eszksF)a7E&N{UjhYN9dMJvCm(h;kTEY
z!=JK`1MIPJZ~ct`6jP9K?@^Z!zl848g3bQ!K2Qk10XwR{!H>>A;<!Y0m3r}Q05Yk8
zy{<HX&1HDP?<Gia?jK$JXEshnZ|ig&f7;cUX4v`7_8LV4zh~4NwHgH<Sv{O>n(1*=
zBaD*>%Y@;#NooZ=g(VLZ?~}k;4ggu7s+_y$wtKX=0X}F1W_ypGa9Kc5p$kXlZRjUk
zWejau6O*6c@1z6kD|_<9>DwK^5^RHDa_EeUdsz_g*k|^7pG9Y$r<O1AO<+|sX2;9<
z-_yDMtd)cgDlkVKE!nttZSBMY2K_!!<B{O!2$X=pm6?s(Od{>Dg*2>@^>RBNsz0UD
zGM;6t7ekSXc<!x!ChO&cqg?$!{cou?enVdmXj?pd9Ti<|dCrTl{)mw~#SR@ep7ihS
z`s-m7;;kK&z2~_;lWG~Ra_>mIBo3_d?Fh|1zazdrDmerEjbxWvZ5ZER5KrZlw1v`l
z$v4b&)I@kPo7WrZ?02>+KfbNIBJ-P6fBO<d#wfP&_Ic)0n<_;$RHpqxp!)2}?KEQa
z1Ct971NpWvhm!ikf+2~>4<DRIk_cbNJh9#}bvVYK2*jPAA&%J0p3RKV9OwFg1|pyA
z&`}}njtYs`ay-6_To(L}ueoCwFG2(Hw3*fi!^FP9BkhQTjWs-@bR#ZJ9B><M#(fL#
zSY{<O6hO?M{tL1G?dI8sdyhH_7^(r)?GcZ9Qmrt3p(;_^24G!wk{h<dCwZ|g<ry-r
za5geV*5WKG*swg<5CRT}mTqqfj!ky(pEL>tBD>msx^bsgLYix;jw@!pT~@~(1dz`2
zcSH+%76UfbC#@XprGxDV7t;GZn2IB9^<DLC70!S|W-VPQ6_j%KvLoV>lYHJ6bAOap
zM`q6lyI!_B_w$HVm7A`;D`1b`AM?A@S`_`PI(3A5W}r3vo3mH$q<I&S5+#p6fL#9}
z%KSPAo{0F`u0o-}=HC|v{=H9)(Y*owfax>7sf!VRG^$-x>u>-0b4z4+<m^FGr1z{x
zaEZ`hevGd0lxugV8Ct4EaD$o1^d&fG*H+JOMd%A!_5B<cVdqO~>1TCD*a4n@75#>B
z4To)0N>Eo)A`=tw@$>@k5&kH>nYU1~`t$%I`)FZsvGc~GaF<eBYZG445<H=oj+uoU
zDxf4~Y<vB}_Zl6Zr{jJ0EYZZ*U%ja32%f^E=h=(L_+<5YqEwiyNvXIY$pJ~OWUHMc
zroFKZS610Ejv6|d!vIa+uC`0}x9_(r*)i9W`^(Pn1O0vb)v35@wF6$zEa$H)vm51Q
z#Ue+~z@%j=d=^~?9|PMwZKLTPVrgB|*!)9Ku>&So^e5oZd)&!r?ggnrzPVvCpQNXq
zipI1ZQd$w*9d`L)@!H8F-nCBxj)qCxg4s_3=34_S?Lq6?zjrAh;3?A4Lu?PU<CDhg
zdiaAK3np8EZ^=Fc@En4@HSGz^yp=zvrq8Ynb?ptq4?&%G*RQ<+<`s#KGKqGcvYwsn
zc7&>LJ}_g~5&`d78tM3kW~Qw8%>F^7ak2?tVGD5HxJMp}^RYZLE({<gG!Lj+{lYQ!
z9|HVYoBN%tJFH2TULtXDBiJJ3sJM%qy{fh?1I*#;gk}{p)GWhiq6;_QMJ$2n>WuoH
z0**aG0jq<{aDp9xo2-<nyHnHEKg%nOC$lo?nVcyhtyC|=fAsL5+4$lq?61ze)0{S^
zsHj-^nFeQ0Jh^8H<_ApJD}f!PZ63evjnB^Va!|Vi*aW@gJJ(gt*H0~_aCO61xkIQF
z02^@Nb4J3pn%`%rL*Ie*t(jP}grt)rJ39v|14Sya?R*?MS0|OAwgnH2`XoSH=BnK9
zBpQO%XxvWlCjpbU53RcJ8*5Ug!hCyRzNW9Zc2J91XShS1&RtvEAmEzJT2ua{dhnYT
z5I9UyXs`<U#`<F3)g~0UQCC|bX-~ujM4Y>F=hAi(DJ`cK*(U>^ydHUIQeRpOukR95
z9Acd*^EM7skNz<QQ{JcU)%9w>-xzLX&GJZQnV8o^@%&RwzS%24TOr6(I_o;+05(F<
zGuhcEn67#LSJx!o-uf4J+ZRxRm&_*3=H_NXTEmc}Wx85IAj9@(5GNKjY+3<ThVLT~
z@F<CRSkX7O><+N8IDeu>4@HVQF;SRVnpw&$2*$6LI34iOv5g{!FDM7!!ojUw)e1mG
zitp@36N`f#qp4^fqO7J^_xL<F;t_Ms^~^618~@jCfv-Xhz2izx*G0adhnBS9>$=(G
z*0bwXNRPpD0k(M+AMzdYwh$f(%hoUn8|T92I|!RHWAd}0-3J}_QP2G(2}6nE!XDMb
z63P3SJ#9Fuv{VH+CB~ztO;~yXN$yCj#+h3L;3Y$Ju6=lUsaP9dyzw}DP6r}UeFxXv
z+E5W{X7VU+lS=gNp4SMjeqeRY>JQTEyx$6BB>GgliMw8yi_1#@86h3!P_X;c>X;R7
z$^n#+9jL~!4|+pFEGm0|zGw_=anW{umM7v^?V;`W{Ohr_FXaDrK0V7~t`a%Ek&(4A
zlf3Nq?c9*}%PW1ex&tn$)xk#zySop*Ccx5BO^fwU`>Cetc^2XK4qRqPNTX|7zITvz
z=JH>QuKQpz=U8vpX!X|e%OObs^tAom3Xtp@Dc78~K4NSB@HghFGan+RL5KyEcbZob
z<S10P44a_Wr&ZzH^z7HKA1`JGivNH~A3}e15D#^lux2Ebh34tco_TGW1VA`e4kHmP
z(lW~pT}(F3Tl87s@M;x~|DdA=MRwq{mCWqcZ0+iDs)}Gq+3wYYAfONc2(X<TS1(C^
zW$ES%57Wv)&FijPUe>FnPg%2mL^?#RY~|m8R`;3wXlkRh{Q7AYc;c}SYO;E{2*$CG
z>ONCEtTCW$Sou@DiA^>QwO6>3M6D==D+U>tD)fRjG)IIx%&zh6=Hjl|rr;<(FJz-L
z!g!T$4j-0_!kP#-joDHRj^)q0tHxJm{mP!it?84u^S@jjjlJ7RH2Mp7EEK({dYt(3
z02x;mePRFAGWdgwFv6VWz_r@F!TfQ!E2mlZ|8)C*l>GnhE%jHQcj$NJWrlYQPNc4*
z-NQPS#&LsY=OM9oq3N_!$B4E5jf~b#m><Do6L}9hRmI6IR(IG5>?gLhWb4`tS7Ia;
zFHhXAw!naX%KYi?6Vd;!2H1Z+uJLDu9{axiXJH#on<W<%-Y?A6)U5*yhh>86oC15R
z9mwYEsrWgyl({lP*dyVOv}eEFD+S)C$I{$}d9gg@W^Y+c>yDub?{YvrU`y1<Qv76$
z)ndl@oyT9;eiBGAyaU7-WfFj9MTWwGuD&l2jLdlVNq|zk@TJ?|FDJ113wc8zF1hm~
zJR`_wUZ_`bU=LU%yV<_%dLRAhlR%^K97YBQN0qztXMu2~O>oWhl=-h$fpz;Hjz0n9
z0vcKJd9{YK!wo<J6MWl?3+^+?Fj)@J8db^J`e!HqiLGO|re`2qE+Ye2H#hK!50Y#&
zg29P3Cg)Wy@@3V`vLDv|T+Eaft_+CU*=fxL4J~i&?C3$2IQ&lnZ3E||I^s%jieT!t
z0R})`Xv{NN5B{Y#>J4D!o-rzcl*8wfUf2QVnCjg{E=-MM-0?!U419!J`y^09a@@Gy
zyUkbT3h^R2M<y>9fwi;9wjmH2mzAa3pudN1?lNsmqWVK}ff%O;V1IX(_8B@B{HBRJ
z3`+f2AUZPQmE?L3Z#}Hja$&=ihORRCF|2KkLh}Alj%M7Tj?lTB4t37bzFQE#Px<{(
zrW8p^nob&SgG5d&UPIU~)!P#`>$ar2lq$-ov&8i-PjaTo1eXW*1d=JfDZ+M%j^o@-
zG{H|7!u~Jz-UKSGb8Q<=Pfn9IleLL4#_6O+jWKGCv!b*aw9bjf2^?Z@MkUSzC~Y&C
z#DPZRJW!h`HHyRm5k;(VAO?{*fr>~Jl{kwEPDuZqp6@;1IqUz|de3*(zt(x*cYS-g
z;DN<<@8`MqGu`)fUsnY9C8ub7r%IKV3E)A0t55@d$oe~*=$bA466O(Gw?96)IJEw%
zmGVBtUK#opT<HV55KV;{I2oiM2190F86;gDIkUHFy6gXWiYI6fW>#W6wt12OBb8>H
zvAi}HUnSTwjbD#lR#pTSw!ta7)}Vb5svQRquUwR#jf*(7yw{*wh+J!F7Kk`X(O%Is
z?E8#nw9UM??EM!Ne|JCsOi{^*S(^BMaxf~ZqZ0r@wP_bLw}(_5Ny;=qc8muJ-L!4h
zZFQpK^$D8YQw_~yFPVNab^ei2I}<v*%B_=6>sxV7lyVzYSuE_!2g$96&nE`Oe5!rM
z*{I(d3n6c5MY)WZs3}z2$k8Z62!NI3QvK^M8an;Ww2#MthO8lEs&k&7kLRbFdya!n
z<D2=hRX>Xx2ouE}doo=t;wx+{TQ>xbJjI&))tx+)0F}uaGC{@GWxqK0Y}jCj5<Y-)
z@rrzIWHxmn?(YyEp9ISUrFYOng`Y1+5=W(OBm2t63)1DZu>|ct?z4?M<a^J)_LIR5
zCnArkC#A~k`$wSHt>|yRD6(%=bhJJ_K6#~7v5hAODS&_RM+@F3$YMieI@3DC%(ymB
z$q=Phrs-q&JIN|aK`DNj%3sg6KVu)*7O39dIF#6L_>zfm5z}Kvqd3>}MN+VAP0-Y(
z3;m}8R`rOz(>KS`YglBqc@fSou}KUE_|GEE%OhGg<R&lI1mCX@BFBVx_jt!`s&B;%
zyou^Br|l>LA}OW;x`7!Znn&KO+gynD(_h|k`et_W>ZF-7s{hslZ`R_Q)*xg&Cwk<Z
zLgoD(o%F!=-oVw$yih?#9hp)!5BO_P6#ks*k$>iS_=n%9(>ogghhvvTr}3U*#{0;C
z*0JW)@fEps%f#6?eh8Ea(mJ^gALlG{kNgG_WC*a`=2#Q^oijIDaz&z~P)1%=HYwY9
z=z#=el2T3mK0$Lv1-ukN6Py<~q;+~qR5gNi)|C|NXBLt}0cRoGK%%RsT)Ve3Wf?NM
zo~mt6A{G&kMOuLCeNszr=*VH-cm~#l*B7*MKzRo0p$K%=8~#ejq$h8bJDA;upDpE5
zL^><@$dFq&$mH|`yJv8hGJux4$hQ1U+Fu8Z|ES!*Db?}ZiR)>&pAxFj@Tf7s`5F|t
zdz@|FwX^7xjP^U+#gv$zUGHWA8&qUG*nRTjOr^A?{0%TurAzx~P!b)}WK|+%!w@V{
zUjp2d-8FjS5x`7k(Y_{<rh}yyqn>{B6C}$?NlG4jF@T@oJAu=Y1BCym^M7m2laG}+
zsD7e<Tp|-_aWrs#Q=)3R&LH6E^r5xAl~fFlE(W!!_&WAmYS6bZcKZe&9lSsgt}<|A
zR~qciM+o+$bNYL7DW&&HajRxoPVK^k7}|c;3Gc%n2k>|4V>K<?nB!1$-&FLc{gP@p
zkP|N&Wa**WJN8L}-w!w!Loh@edHQx06RpbJ2hH!GF`P&!;NiUrcnoN3zaKy(;5IRn
zF`!)ZR3`HG1FN{5%-Q9SxBX|c#QWRqIE!CSZJhNvR$&#jBv)$A+Dt@Wp+g+)z2Cg{
z+#@GhL}iys^=xpL-sKs+TIb9q=sDVl)i11%dP~af5E4Vlq2yknY>F;YqqGgY$uX+R
z-sjcL&!tBfHR!LD;W`*1@!qGE8mYd^^&jr0hUVqa`+LBxz9R`(KwHfrV?4K4iDvd)
z5lyZln`*S4xND<En%>@=*4*;^h1|qv1g}iek<*#VAB5V=;77UxHtPL~&+^8N<+WV3
zD(*_lov4bLjeC#YksHfXA|4e@&n;TK8mgyIh{d5}gy@~Ih*iQ;P2`fk2UufhRC8)|
z%41OGAyrL-@Xh<>;#aHlL#|~j5^mjUd>KM1pu1Ru>hp${E~DDRvpS~m{LysWX7f%j
zfq;db=%2eV-e|t!`*XaLuuj1&v$?iRsUgSp$;jhw+5)G$TE13`NGJxmuyK>|dijR#
zTNOGE2|H2m41c=$myeD8*W=)~*r4`@k*1u^TA&-CuY@H~w&;6vz+7tSWy$q~7tKUb
z{GOC6z&>+cQOg-<*s%x|q|G-*zrs5i01NeyMW`6JRg+RZ9lwu&o`?DkmF2LLCzA&!
z7o4@;rmQxt1Z>Q~=Bx>~`Ek*#_GW6y;DCa5;^WQ_hfjjg*D@E`-8;3e3Gog3)2XY3
z_3b6M{Zw8d4yzu7K0}`a8~fjmH};_X$Ibq;?}y(YQ0VgLEPUPo*~E4@N#2%uBeOQf
zd6Gym=2izqz(mXSEz(r_xyNtbVc409bh|-o@WhX;$_?V1pz}z5WcFxdtp(2<+dLCo
z9pkFT)@jO9b-|~Yyfl0c%^+pCFF(IiS8sipZl!0_9ePWU{oR#ZWO9Qe<Te3(C*|=8
z+IwwZzc>%Fw?%I=AjoSc9z78~)U=|?fWAvB33!Do9jur;<{q>)B;$~ue}>`S-)38W
zq%pkewb3WC68us_B1?Eb5<V6ebNFp@HNuae(+wKvGHbqdpPE|0f$trfs*d#VUH>8@
zAdll!1p-lC+B=2NKTES~lKFGHyk%H&eE`aAFDB-!f#W41dmDz}oX-dZwASKW-Cy<f
z`uY_w1Oe&}L21u`FV9$?>xn~YQEpxsF3a^hKDlS>QZkSi3LBEXMh&DbTPr02BD}k6
zMY13D+5p{o+56w_NFVq;IlPM4V(+NN%kQ=x+`K5c0Y;*!?Qiy8wU3OdmU~S9u*caK
zNo}|fq_w~8A89`&Gsy$YYqg{&ia-3pZTsg`F(74ue!At4u6_l+Lg+rYMcP}GU|Nsv
z`~0vgmwG|9t)1sLlQQbJuq5-YTKLWWvh*W4+zJ4C_P9NpWxRGp7(W}HnWg8xwzwQ`
z&R2_4#}iYpO-H1q30;P(Oe!Jyh5gX-7yeN!DjLKdJGuweQr{&n(ltl^)R(GM)`PeZ
z+_0T`MW{b}%<4hqsE-#fHGz73^g?)RKSNz%o{fJpY35|_MT{^}bb0Z#p}pzd3-0AY
ziz6a_CdbHPH-78qcHZ$d+9{DBBX;G?)bWF!)6z?PiRH1=*yr=xZ%s>1pH#LEed(y9
znEbV9!q@jB{5JQk?+E#AP;|aWp*L}M<s<y|)gSO%^KMSrOdRPKir%Z3)lf5gzv+I{
z->w*HO$JswM-`22#Jt1%(r*pMRcpZoCdV5z)!RaA?w(Op<jntiv!%T`o1*|rL53?d
zZPz2+OY2uqd*BnBlZrAolG#Y3IbFWRlee}ub>77fs}w#kj2SIcD;IC@9o((&YcV};
z@r!xS=ING`VEs21x*F@1Rt2{f@DSp{Fu}LNZUi7Z#fOaq7ga!rVU7w5x2jwc86c5n
zSqKxfVQt`O2s0eWfStMb{|YYrGd=ZxBzXAWgBd;&?VxtzI<H)naGV`>AN`@9*U`du
z+ZCmf_B(*p)4=YMPqga5Ui)$zb{9_7S~dS3*lx;OmKN-6Lxl3%CjEhFzaO}tv4j@F
zWyHk*!2vH0%VsB=NJ{;(fDm~5JFKXA8VG??NITMiX?-c>FMl)efA<i0NoziQ&F8n}
zz-Q-X*Gh8)7M@qdRMlt3PptellR?GVy9&jEUG)#*BrMoskm=lzw2wBZijoI(+qD(;
z>#6-iW1)Nf9er%6o?3&d*nG4l42U*;^7G*awK$=Q(w-cELH==xrtvXULvZFF<>^T|
zz_}H8-grmati)NE55&P$o{1cIwB@<IAQ`YEjkcA8m|B=zNh<><jV2udd@|KEAoh}U
z=g?&~z_joQ%XctArZ(mByNl~4cals#{dwx8eT;^TobL<vUOfM&CfucsJF#55*N`0s
z=A6JKgDCyeAQBW|KIj{yr$4$A#vN*ze44DbkXBooUFtWaf_o#jT<EumRxDUNJE8YE
zK5s~t+#h3MfWb_QDDE4WRkKBG*l@{NQ{$$f0v#u@)yQumOGU&6B}@5|>q>?x-fACT
zlZz2RiW)btcK|TW7E;@gc2b^>irq-8(0qhwG#SlepvXOK{<xw=kI%Uh#|Qn<v~za8
zkz66s^446IUvM|8jQ8!UmWmG<+d3L6s25lq$1jqc5y!PxFkr4R#lgve^EML}9@WMJ
zB3J7B9jIyZekR=c+xN!R(*tL{ow(XLbxTw21%z|VsoK(ZIGxZD8a`Fq)s-b$#9EA|
zCaAJuCVta8D-{hcLM7^)*={PQE-R?zR%5DZ?5Y_!4!S&BQTfy-dF17UTNQ+8ms{;#
zWeE&=Oe4Dy_^l#?;VPFG!i<rNw~F+ODr}!C;EI6JZ!$G1%Sp@bPK7=nA2p<`ZRe8v
z#O2wGYwo!<iTF@Yp<GSkldi-^!lNev7TEuF7#9}q_VqFO2H=eLG|t~UtWuoPE0$0>
zt)Y?x9Bd)Tn7a|m-sJkp*)S((nv}8H!4F=Kx~|_J5OHS+_+RY})iKY)u2|h%PvoYR
zP4ALE>5os132Fi9`AnPXCggM4F+l`{u6Y4(a+JifDu}xq*sDJW+CwG9#}bXrjx!vO
zd&lI%6M_#_HeJ|fIqWIG0m`s#M7HRW)VQwg?vy1GFFD*falAVDqQ7ZA#wbE%d<Bxa
z5RUIZdJ++u(?8+5fytdiYw4_n?Si?!e(d76DycVyVn$T<dUu~Mm?Vf!7HoN-$?=hX
zSo%YNwy8|_YK(4#6IiZ&Ln*TKRR~;K(1Fj=2&NVGrhU?N`@N}?_5+sCG?AekVLquw
zV>Btb)jDjbwllN!iU3ZitwwUyw{3U-Ht_t5g=>e?*%??U=)9@lhU*))p@}q^t54#6
zt3vU=+^7?epOa>ULG52Hd3=pauqx~+sGd(z;P&kHQ{c{aR81&N5X0M7yb!?Vwaccm
zwsT{Aem`ImU;H#uTvC$5gNB+L>zbHS4X5$(>l9+eNHhP&c6G70im`3TcyG$n=4y2P
zQXKw5Ks;q7M(YxL=&<$o|5`mhZQd^>R4VQycuS_*wWTquxq*^*QY0oHY}Wo7<;}Y&
zg;8(*et@o;o2tfM#|jWYPBs&>jbM044ziAe*Hm=@UL?t|<mpc3j<)Jf`OZPU)s-E@
zloqOb-Q+EJ|J=F<4cMFING+uD0G?+W?IgO#`MX>Fcew>pozn8vqfde1iJWtQTLAqR
zF02ae(iXhWT;qLM4$|FGoqxEs7Xz1;t)|^N0k|>_UdIOeTBV+rI7*KLcrgFp4@{jj
z0-(HI`#3N=ZlZmHXZ`zuYV*%|ADnk5gxHQ&IvP9oj9<6=KWL7tu-J-N^I2NKWApQ0
z$Fu8LlYYb%v-Ic+0Jd!Nytg^OGYdgMfp%rEOl;`~X2lLg+Fm$lYVpxyP0sXP^b3w&
zZrm3H3#xxVAfL;@>?zy%23`i92I^M*YRsy{ifoO+G6OkfSITFB<4B2ZhsRI9A6R&;
zPcD87CJAIl;iY1d{|@@gQahpQfJ(ZkrU6_F=UQIr0P`Mo$^tH$Sx4tI`@I&hnhH8<
z+@xByLy$3r=Svlxx1{Ok^zXKRe_(U(Q0Mo<uabO<!}{+S;SmS@W1jVgD@aPqR6l>%
z<26)7oV3x-eZwZz(MHDyvcpL7332l`o;!Krs$9JM&$Lb;6nbMo53wgZg?ki&)|;g}
z!~E?@6Avq~R6L37{(514et}P~Yp5&QIMtnY!eXX3^vPg@!533Z?4c&5Vnwq$_~KMW
zeGn6?uJ7NSRW~<wIk4BLLd$b^oJEAo&IP%J3Qvf061TFX-qS0OVEPld7MBP8=6j{$
zbnSwutoA4ww<9C5KFN*Y>!2TgYQj!B!i#$)`IP{j+xU9_!_~*+vSMA{F<xMP;1baD
z_7b|RCfPUMBi5UU`=Duq>HWNDPh5|k>#OzHo$)pVR40YVD6iV;j~-pkKZGhdG<Tui
zECK7+7@L})IHW(6&B4T5Y*-kQW|mIxvV!gZycdlKsM!(E_O6S*8h{R(=tKOSe8U93
z{!2l>AIO_kiJtL!o<o|L;H`1f!`=*2n{fyoyW9(~sQCKVc@~H7H)Muo^Iw_TgHjb0
z>_XnrMxi&?ak_U~RB>%;gHkeXI0d|@WaGM6BrC=7O;aF?(@nPc&>T3zkReVwE)m-G
zv(?(M!N@S7f^q!inZm|yUm#J+Y<Y>2uixJAW@Qe`WHw|YLI{OLe43vu=ET_~Y<}8t
ztu!U)CoXlHQRekY|1ely8d+j(#tZUWPnNEnSiV<AP>P$j&j72ksU(r7fZPDidZ2Z2
ze6Kg?dBO?H;dOb1>R+wB!uaCjX?AbMhqqV_yZYaXe?@(BXWq9kYVy$<=t1asibIp-
z3LWEjzK5p#u9=HXsVe5^qfU+`m7*Y~X!Fp99$|@?3fa5N=AKJAFwdGbo8aPNj`gQ~
zDma}<@atpJ1vEYbwUeGggKXpteB^<5RS#wacP|%hXn6kJ`1vmt&oO&qdSo7=)LN$e
zl18oUxO>On#FM>kd`+#9ABI{vV(qZYu4GPpq0Yr#^b@}FWjv!|awrdH)P~AUZZfBy
zoy&viUzzK>d}4_gY({IQaNERcbx<)Z+syCjW#Z+e{WLxIFDek$e&&fwCk`vx5BL+p
zwvS$L>4CS*X*{UaKAJ%)4~9k9<$_C0|LXAnKd#5eUElj3rj@4yQ!;zkOpbFGLExB#
zBMK^=>obT(C&l{b>Yf8FRcZa+{FM#G{gDb=n6*HmXIrOfw}&!3OLy(zFck~`^f6`h
zt>!o85$iI4i8tnLEN@hTrwCMooUh}zrN>$IvHM2gb@)RoZLO@)wk}~=27(OMZh>+g
ztsmMef6rQJx2*&+M)^zn#-}|gZkd%P3Ym?hS2S3S&aY#mqn_g~*$?ogxfK+EWhNrx
zS67_%<m3ah3j-DBpUwAP%BFvI`0u^s|E9zi#aYv&=mfQ^ViJR@Dh^jTW8yx)A*hN=
z+D}QikKImWxa*ne;m)<IRwn@S;qM2eHXScDzXNs<zYrR#RKZ8fw9}TORrLV?bi1Sx
z*!szfOKmXeWm6K%;O9&rHklj+c*zrzWu+t#Vhs`9E+7X>hiB`=lyClK_5a?xKWR|W
zm#+;R&s_Q9KI}}7)s6e-HhK)K!G`BPyV|dh0m2LG)CPUolLiLUUx5{{1G1`>!J^O9
zR&5$T87ixEXM`%j5=?^0#R@KSJxwH9AsikFVn;vI$h`Q>c=Bdulg-JqtdVk;+#UnR
zI!?XuR_sOo`nvx`8qfA8URrwtk}cy^RX%DF^x0<q^9YaA-DXY<T)tk=jOg+;U$Uuh
z#fA~d$`|4ok&Y-Cor;T|>h&qrMhsf!H7L_rE7h!EW2`b_OkO7|!j?TihkUr*t7@BB
zYtPA%FJcxnS=vPe;HwSkXMz_`ofJfRyp3U+#K@Qi2|bri2N&#IO`W|&m6tVQIJj$S
zwJ)3by!`!ua#mH63vt$7TbqK1B13D#hMUU(U1|3arovo!%)G#HcysaR&8#12E8*+r
zmTJF>`k&+4TfN;TxWkJmYn%+8Nc5J)d}?ekHB{+c_av7FW&H9~R_^|Vxg<~tTmS8b
zLLIELICmx?=~8o)BkuI2Vvbon0;4n;G-L<WsrG7QRMx4W=ug~o7{CK&k9pXhVRIW~
zQb$fX=set*>Nw?79osx9uC*Gg^7fB$4L8|Foj9nT%QB(dA|hVpvlBC6p9h}x%5WOF
z_gwwi$gj8OCnqL+Zau)B`u)J4t%@Z#mqX*7;6;L69BOL&k`FR%iQ*_ke6Ud<w7pq}
zw4g}76GiaV#VPJAX`_2HAu&ag$2_wheYzuE<>G|&hXtI;yQtZLk&&Sp)67!x#Mt0y
zx<6xMHT}e~LL*XC5>=6RntC>oo*sM*X*tq)nus^C`H+^ub3}4nm8;;zmb5Qp+$OXJ
zcaN`^_E^Xk_f0xpqP8TT=&+tkL8EX+&d4>fDZSXI9v&vAhaft{HJ$_Abrj01=`7}V
zGUAiUTp!c8<|+yRY#`Y3>S@->!>rF5_aa8OInxAnc7&W5#m%I6HHXl%TXrQqEU%&p
zoP!AVGQq5}mfGD+^&ODi(avrHc{M;cd-4(+-*YimDUGzSrM)|dlrVP<mv;~0MUQ2^
za8~fGj8)-iy_}7S-Z*S|gH(3fTb;IJVLo=rXJ)u9UN{p`^ri2#skq-NHn=JRZLikk
zMr?kblrg;S9>a{X3q~2EjwjQFiuuk3Lwv!D-IyHpU|4&+qxZJRc<p){S*=>=dp9yZ
zTsSSN^+7J)9j0S6l&QfUmhdNHTI()>mFyM`4`e72jO4Jvvd^<~a+D=W1fCcSR(YV=
z=iDUB34(`EEyJH}ggo~{qFj-|(+5ik&UpzJ7Ew?{iZEI5Sx7G8?nKO^EhEc#6EC>w
zgj?H$_-M!#xpUa@m3pJ)$S@z&1H_N@%{%609jY>_pZR8@jD_2)Ly6$GZSA`4y|Wg;
z$X9uMg-Ko|kYFdu(771DANkSya8QCwnukibTldaYJ+(T_s&SeRb2yd3-GG=`wzmq2
z?d^wUaK@eG=^O(?D#WpJFKKaGvR@oPj4<9Y>a<5SXPy7$#{V8ZKaQhIV7KJ%w<)g$
z5l0CU%g0&vk&~CR-&LhVsfYLz9aGtF`L`$Hp)W7rDx&$a^#@1V+wqr(y1JU#4-FKr
zZ0Mk~>c~yLuaX;vEoC=1>XM&ww^A2hi}vjUYN-}33FIt$!+tA^d(*EbhcmOBbH<C{
zaF?KF;r056h=5KUo>-U8T}(kfUH0zB9PYjYu>TiYHXyG_N>Or>r9I^7N7$g_s8mGK
z?)apdiaAe5I5_?=POqjUX^7e;#QJRaOn$-I-Z*IYny&|$%%wVqkFE%S-FnDGjLTrr
zi#mGv4n2K>IARty{5(*n`%>VuP<~DLhrwYC3*l~If|pi^+g*`Cejr@X45BXGt`L#$
z@o;*LWwc!;a|Iz+FgTqkN#0G2FsyXgc(R97SuV{O(|hV$$Bs=V9Hy;#$Du<8I!-lt
z)p%(1BOQ1!xCut8U01}WhivFO$HH*H!cjRBdiei1g7ncae|H@GbziW8J+hq=ckh{n
zE7J>`(^zKK$#AC~4|ShZJiXKG_Zk`DjEUW|e>?+((Ly-U1J5ya*&+1MtftJB1btuH
z!AEarA`YIR?r%LWL0MZ|f5WX5Q2MrttryWoP2?H~*V~=d%>0(dCQEvwnhFtq^!Fmh
zoww7;(ae>3?%ev8Sp?R=lGRa4_wzaI@Y~<r+dp5l&rb{f@$jRB$GKt?&EF<W5^C1=
z850)*kTUeA)B)4Tw;8!G>xD{HvlvWkA7aZrS3Og)Bv|o%6KC=4hHJi#vmyc%8E_O3
z(%8FDdSIUP2K2)y2hAk)0|_V(soVVy9UT6)38Z&ua~P0#t39(_Q)zK$>ts)|K6OS?
zJCBGnI;(Ig{vfnvU!wq`t%`ZoHR^Ro=x$sv5oA0#*{PiM3P?DC`NKl(>`phuyXURD
zl8Aqh`za!mPt82HewOWuJ6~^q=`Sj`|Dn?V#^kfKBwv*>mziHbYL7Q}N*U>DhZahW
zQKA#oU;=(E2H>5Lf7vZi-`5AwSdq)<p&8VX7qQ<TIB-@AX?}BKt(7mc7H#s!3XJj$
zSb-^u%g{cw2&oET0N6f@YtW;mQ~(l(XDUm1vuI|trWq!e|F^pT*T^td?HjT^-{1@y
z>^+~&FdT9`V}HK>ib{;5;lTR+dirO8R5EN9SWki9?>{PbaIXW%%D*B$`Y87<D=RAk
zoT1ljSi$o$y;zl?;4T~B`z!1yP*Squ@M0geB4a{t@AJ2&1SG=Ge#(>IP~h!rR}z$c
z8mXto^?5g>Kk(ewE^o}jL0LTT>W=nMS+2O<iSX=c>Exr7(so&^`uBr#P1}xWNAo;E
zgN9WC)taf=PO3(nE}8J~=uFeJAj;DW>D9$Qv8PQL6Y+S_jy9X|%b=J5G57PL`gkKf
z1W8J%@CYZhvEmb;r*oFYisHsFm{8xS*V|;2`D9?3`dQ%`UT}NA(szA5&>_p>eq5Q}
zVqTkX*x;xy|L(h)$#0rSRSFe(_HCF%QZ;VkRvxzUOI8CP73g)bCpnubV;(&SfZniy
znw?v>YHOM3*;d3$r83hOOYg@g%W$mv8N%bj&Tf(9*_@EY+7PIu4v=hLTn}p48A2ZJ
zmt09uxe^VV5^b*>{<JdHM>s7hZ-X$WPLEDJ9%<}KEsEL}Fbf|`K-bbVw$l=D1#|cl
z*2|>Z#z!p*pIe6OSlt+Zl{^fnSz~H_pgwP=<KD(xk<C!g1bNrx`Td&rb<UjQ3EUcE
zZEIi=Bi8>wtCU2dqU0*~ck^Z5MAvy@UQe`ZR=2)|XwnMPBl;ZngBjFn*qg1g*|$4;
zz#?UCW02?7sg%{Pj24C~=)=3Dgfq{(9ImA(-|Ac1)1~>!Tb$@DVS2wtc<UON!6g-m
z0CM!A9HLfx`2M6HH<`jtw|O+(OhIdB@J*!_r)W)4A@lGsSdnyVjDS%_p$}tJf2*ut
z3J)f=TA{}90%of&bcTdw&xs2+BvvR2MPr$~XmX1{Y-<f+ul{P&|6uMUIJb*?6H}u0
z*<kalRubOH>TaG@w6h-bGZfF@-cO4Xd7US_62j50Xvhw@?DmW)*K)!OOrkZUsq1Vr
z2S>0wsaVehW6`oadwjGbSvt2JGus^^)pi*vkZ!i7b*E}tom0|Jy0ZA~iVO(q+LZB>
z)U~uX=ke6&rSA&Z^>ZVut5w8Bm7DALHS<g{j$W3R-9+?D=Z>D6usWG~ww0w+(5m{2
zZ|Q84AJ`z~@nTC#a4rQZ+={UbqL}E+Sv7d=trj_q!>0q?qNwFtSsEi$Oz2#-oN&3j
zt~Ceg3qp%)B}<DyJ}Qbpm|rjOZQgU{*1lVWz%10dDRIn#&x5k<OQht_oAZAvza%Y$
zd*5pAf&|4vw{PaD>}Eco=KNShboRMe2r*o+wWO$tY?ksi6Y5-zc@=G~4ZAUK7(t`W
zD=sjysH!oEsLxheiBRdln_&B>{Z7cI8FAvtmIPviH_7>j?Z->w?v+;Ptao#3*Ses_
zpOqOPyY@AP!rkZ<Xw{Y{9?mi?sIHkB2gh*o5Lh8HJ;4QKO~yqBlmPmMUcW+Jrd9kO
z<P|sh&cJ&h(Vh!C7xUT4d5gZ&B+7%_tcF*9f|s13?esnsi;<OzZjV~@LGTx4_t*(J
z0`rPiKc}ILLIgMYnm~{kezti!A|lJry=6P8?63+bl}8DWo`@tVJ2dThnjOcg?EK{8
zI+9w<2|j(?$=XzEula-~|8~#F7?C~sl@>l55fRY}LG5V~%62O2R;Ws;D0MfP{PTYs
zh5w1dCtaXLrjk9YOm{pfznJ9hbdQ$&phUMpVX0B1A#$rW0VGE6)aX>P-UKy}p#{Cc
zk6z!wM*^hTZvpp%4*XW}TR8%bTDk>xLng=$B+DEmkM4m@%I<y;j>pv`R;cmn$mF`b
zQBkAmKIkUE=e<{WcOmRH9*}>rqvhxKZFmR!@tcw&VKJxk9^zBRVgi8>)V>x*S$hQ#
zp48cCel?A+4`}QT^~(Jm<P3@>b6N_O)6x^2w*22y3B#HFmm^YN?>eCWvYuyd;AoBP
zw`Ja0WV+C0pO!07w)Z|o8y|rti*&>m%e562&~(Ri>Pagm+S&5(2;Tyo%H~^M5rx&I
zN1ExBBEv<s&l_arYfs>55&S1hk&Au+Rn;-dLrU~%dKNj6k*d-}#~5mhPO_ZfO*2ed
zlsvODZmbx#b7pmEa;9z{#(NXgZD81)-M+8(aO2@uLk>H}mD*vyb$(T?`-dUGc1ADX
z0s3)Ah!;G{UF?S8H_O7}LPl_gnmG4dTeVA@3$A6Jj_22Vyef)39WVKxxuTaoSQv^*
z(sC$RcNy-@5~aIrHabSwvjzbDQ?CMIkOs)1t7~U%|8Y?tqyZ2|*4MYv(6Fl02UL<D
z?{?9E@H&WK&xcQ&+1DC#!^)Xnc4zDnKKIS6ODQ}XND+lq4|N^Zoc(;Il5U$xEc2><
z$GlN)lX_n>^MPYe<yUr|<9DZ1|70akSjh5oyPO<Xm4h}R-AT@@r)_`;dJ2R*2{$?S
zZcoW_;Ny%W)1^W^Hi%`c_1YtzJx-)!JX)9T$}I{R0kvm$aI+U`q+X@y0Cywwg*}`e
zoxp;Du0_(Y0n1K$ryjj%X(3a|&24SXgoLC72yukV<%TX!aJ?s-jl#C%jTRj(AtQ@!
zWy!e!>LV=oqN9SKD6}#Irhj*;n~;8*ddzbT{+@<8aqLVKTji2zox482J6**TuRl%u
z4xp7{X+yfM-_iDb)YL@8!+pg>0^AV~{2_f=8*s?bT+qiOxQ|(g9h4{$e)@-Q6Ycy(
z)M1Snfm#^FwsSlPkviizr0y(VyLd(s@u-c`Tqava>`u;MI1MSM(>S*7dwPYv-jF+Y
zw>ij+JePIW>v2G`N}G!$-P#mXZzvD7ZnP*PYSti4oSB_S8$Ei*J-Hxl%xQBLi9sYa
z<PEb22S#YcF$r*89UljUrptU9;%fTzjtU$N&6p+{M6NE+WF#7`d2o>&VX*`A37=-y
z#nbi9s|9&3h8ODDxksEe+IHNAgs3Q?Y=E8UXj^AJ=y7@-HlM6ASu&y%US(@!F|HK%
ze7Ly+ARMvl;J$|Bo8-7OwNJztsxt#Wa5zx-wk6U7$YI5%_tS(|x!R7NoGDA4MMb3C
zypNyGFXWiXI{-CkYKOOb)1`^6b3X<(zq-D@xK?DQnKelDCMW4$%`WpT+>NSu*Gw+S
z!N9RRn)BOY(c_}QukHxTamJ5`abI(~#y225xKV1VCmw$9#OL8x<@*1j;s1pXhn%TH
zTj`F7FLY;o)?{Rt4Fb%|hf7%S8zUOslh&0oIpinzMNy#4w++^liQ|EFP4cIzLnz3U
z#Wtd(9~R9q{&FFc`ul;ONT1&M=8wzA(`qecSI~&aDyRrkT_RwxhPaM_P)nKw(J&`E
z7SWf3B+oEbhU?ZB&U_wFHS1L2J&91YF8%Jc?Sj3b{{aWV8KlOfVabCCr^VY-C%&E<
zcYBdnp^yYF-JZnvtvk*tz+9La>l06tLP3F;mtpdaI<=um)^@|8!}kNekGJa_{Y~gY
zDL=wi!VN6BeYcY|&Rgf{4-@oCK+>eTl*C>riU+YOeoRH<1u`Wn34_qY>J)Ozxh=N9
zulHw}7fnD?T}<rNRYHGz&Nr&I5kM~A+=n!tUnyGSd#h{?kjuvykfFAlOC1lSsFR+M
zIK{uHrliJ6iVp1BBjgYylWSJHlEodvPd4)=FY+QIAEk(P*y|w&=XEG<)Cukvdy5-{
znRkc(?`8jQBme*VNPVUt)Lt7EV2(&%-`O1*EzNl%F5S@UDdJt^1T^0RH!daD6s-&t
zkKZ-jg8<N?{PNkrJ{QVxbe!(;i`Tara1oHV$tI!&H@w}1+WSAcG7wwP^;ItWOy%69
zhR5gx_TXLaWJBkc$=&j&OJH5=y=zmg7caYID!wd0&GuWZ>bG+txAF`lqe>7zau79e
zhLblD6hRS^f6|+pYF+vz_FyINFYAY45x*a>;+eywXZQGK28R)t_guK9^N{QB2Q->t
z)Aa?*=@ZThAA0t3)>|vuhKEA}>PC-8E^Ze2ht#B6b)NJOOz-v_heZ`|?2*Uf&0mja
zmI49-6<QY{V`C6#uhS%pHU7XHU|am)nw@#p#n!1LyZ6$o@BcB__;1sE`bq_wmaiQh
z4u=0k`VKu-AMi8RzG6Q)!4=D9C8}=vR?AaxoAjpusvc;il7Og%%2;4nJ*&BACfrzS
z<9nieXKL>N=)P1o;8N(WNf<}&$M3&H`?-ApU}*1;Cps~|{rv!I5hMKo1b7mh(p8VX
z`Cq04$mq`Q|148yGA<+RKPEZXTw#{V;4SuU=I_YdK#PMXPCsX@iW-*uIBUH5exu^z
zP{TvOsUKY-oKs*y6`GezE8b!G&8$it+Tq(YAx*OUu;_Yy+28*EKUFwUzGXnMO#YBB
z$Bo8XMQjQUwO=oAoCk$qnV)z`Cru*k?6BuTTQ;{&i}b4Pji8AE(fM%zJoci;Dcu-y
zcwAeUn`WEU!{kpE{JOw*JBKaJru4`;N>g;k1_x^bt&J-^Mk-3ZAxJWxraPiyPW76P
z7ldN7XY|rFbivMW^~>&!?cpmRc?W((Zw&J06HT6v56G9ZvN~6jv}0bNF;WZ=X!+?>
z7JS`Z9X=Mtx)b!k>YUW5^yO;G9Mt*MTkhRp-<QLLD?^WBy+>v(GILZ7gYvv~Y$D4w
ztbRWbk9P->mK@GEZR3qI0!U?LWT;w%627Oq{TS~PDn?=fxG==AYsvQ(e6%qdk&C0T
zU;C-b6J<~84<~x(jj`Tq*<b5U5NHU@vURd`-ojYDTcEoQZONoDi_-OlA@%LlBS9lL
zd0de&A|pmXo7HLQ)+2{vXfh;~7l}1Jl$n~@MpN9EAL$?t9Nf<f`6!bIyENEy(%!Hx
zL+z5ks;RYXzTm#UYME>rQ?QqcHtFkfaB#TR_3;1-i9&sFhC?E-6n03(;)Y-8rD<XW
zu$EHv_x!R>Ubu4BB3I^N&+Du(sr{rke<}89cBVz9N#t8rUK2g*Rr!2gSt;3%Ma(>k
zbcjC;ehIa8<-ris_m<vowzC{lZ5oq<2|c?$3%<_9M3Vtl(~11Cc>GFR-A0LSsivwN
zu_33AQ}SL^RdhL!-q0IG>4AF8PD&pvA}m>mdsoY-smLD*B-IR3`g)NmfBy9FI6b5^
zBXMy%EMgLFgpYd9toFp3yja8rCb~9(Jbs+>{qv*{Do!gE4dm^<HVO&2I|fbCaz+!v
zN1?qJ0*$!oimO%D8#1}Xrq=9SB>e4V4;pujI|1h|pes_g2B^F@n~0k&VuAOPcLM4B
z;?nhKI3je0$cJW6lh!yIIL)32q~5wo>_%WGd`S0bzh#NS8oy9U=JvFOiDFvblu6Xe
ze9(r<Zc?p-7hHC=qwe8=!rfrnoaJb}dy~+Ja&1sVB#$A={&ye!Py5gb{kcdOKeat7
zb+NCcwUNKsZ^4(S_aK8tFB{Of9LUX##$O@h(4qTmGo>b%bv#w?=%cy{s9FKt*Nd+Y
zXy%Y^U0}M$`eMdYY?moA*7GQ9rrFq@hE&($2|o_5vP7;=qB@M#Ltv+C%PNlqa(LWP
zs-=$(NI@))=69s2FAAb32&x@I4vM7J;OKZ<$(T`x)f5|ID&UYV$NIqDq*{<^)#=`R
zovN%{;<8-HILaC2WER{7FVKxqfM)<XJLd6DyZI-glt?!ol$}&o^{FZL9dR0eAtq;i
z(+{zk47_Z>h`D<7$~5ZGp_+nu?TLW+f^COwr}_3{%9^LM$}D|h!&Wh_zFdBblU9l3
zVP7I6QZ=m!!ZyZ0`5Q^Kl|*>DhS(u)Ch06EY*e`Tjq^>*t2287?Qn7g5RVw?O`L9S
zPk7=_6V-$-bqB%okw%*R4rLMEH$yufrg}RP&V+`LZvTTN_AlJP$3Sx1+Afmh0ABHO
z1ma4GNZSMNo@i^Ad(q`e=z@C{XmAMUR06n8H~t7fe@jCS$?4_m%md*@sD?wK4#ZVX
z%Zs>*9EU^rcsL-ip)dEOuWwrdzN0ZDSL28eeT2WG0qm)|u3bwB?%V%LZPDQO13Hh)
z<(HP?5e<pSN@=4RM4sC=TP?FXt`-(EJ~LM4Gj`D%>Zla4G_E`%R`IbptdsNtALSZJ
zr7fA<3ZdK4q1*Pm3azLh)b$jc%CltWww#*XKwR30bflW-0IFmw_x51yrWY}ntXZEe
zdB&0OjOdeDIu34AapJ^26ABOZY-fDT@G{RS|AVvtl4*5Y|11g0UxyXnrWOBrcGiGV
z4c6VLfKTt$swM~^@dS+JG|1JDlTEHx=B}<$m;KUK7*SXA8W+Y^zHaSh23EzSmLlSw
z@$p%ub~&$KKiFSttZW`?vUuszW3aGi2};?w+eBz|-eT}C*gk8{@jC%fUr%CC-w93Y
zWBa2^5+XXQeX`&`#}wY37pLlqYMlBV!%v3L`lAXWz)pUToBD~NDE&5{<CMH%+LbM5
zr#;Q@Mu|z5_Cb7nc4&297;DVX_^Y6l+_q`Uaq_aPh&}eiH%P}cSl2WoNBOaN`5#t4
zoRVxdMIQH+t{87y$qlUFW78cnEi5|=jbF+a)anm~+J7jUgZ6E84rd|EoHEv(ue!0d
z0*#Ld@zzARwnKR2W#a+n^ug);LDjSqGMeXj7KbiPNC9z+DRUE<oI`?ef!@wYx2frB
zv0twwrQUs{m%d(TWIBP?-4+TM#C1_Ls}20K*lL;N4YDp>#6i2`$v#CqdF3iqTT(l?
zDK$Q%ri%^za{v>eZ(zsY3aI_wN=Yj3gIt>(sJ&#mJ6^j?FSdL@Bz+YVME8f=gpAJO
z&blr5e{CFF*<67OH6(=E$HDQimD+0Wk&4imx<K)<2e}o?%8^n{s;;H|_RmXa#7JGp
znx)?3F?xPYtgbf9Xgf<cVwL72wJ}<JA`w`kPM4S^`IMMeJ;`UJX8`27O$N7njdxXX
zAl0iwMIo&C9mgl`trBXtmZF&?v9R-D-AbmTDFwMm$YKG3gRyl%n|5OM3lHyq=x6`R
z2M^4+pHRJdZEJ0g6QzW%bG*M~5>|WJTcJyCDUzycQtKaW%0tq<bFwnc=6#^ULz}Jb
zPwUK(7wdg)Z<VrbdtF8L)`{O-R{P`PcJdE#_vPa#w-Tx|(fAUy(-6jJsk90F)J%|4
z9Y=^dtjd7c#R(eQN6$A<D5GLe$Oyo?<1kb=wqInZ8-XB1ZA*;o$YG!Sf=)}QZ=Gr%
z+ym)y80M7HQNdA&@d%-*_U>y>kLe62JFkGuF2j|hTV!RY(>JTn=U#B1ftFeKaon>_
z&McK(hBQFlo_2kGd##0~rc78YK>4=+&^dmSHvM&hLrSH#Y&xL!eQJ5Bu9<V>)}^KJ
z<@wjAK5(EnE3BXWbiZ*X&-2Ow@%R5sdmhx$3vXX)3{{;8?Bp$L7I(OQ^Vie<rv?~-
z+Wi}2jE~J8s+2)zwb1aV35n>jhJg26n-|Vy`NP2?QR<)261iWyx3u>SXWBV?Sb>64
zY!w%{-$Rd6`vXuCz3n;G^(2!50QStw?Bhudykr`8K`3ZEW@@4IE41X(hu|fr_HWx+
zi&wYJR6HargQf14`Op4pPT=*w24C8rlM^x9kA-?^HizA3N~a-6VBy^0nDlU3_hMsv
zMQ|Z69-4gF*%zNYklp6sR8F8r!4sOobD$o$EK+mhtH$R^C3JrwX%IS<FohvYoRfdR
zRLAYKpJZ*nd_TF;)@eoD_z>5xtzqw>tmSMvPiCyUVvJQ3(=(b^i8(4MCF~B`JR$=A
zj=R?;ER{(syBe(qL*7B8KjDWa*7__zVQCD73Juad4!1OokBZ0jS;G}~cIu~SDDIx~
zI2FF)r@w^a_3s6y<VP~yiZU~E&x|y&HYJ~)C2IDBnMa;Z9{F%vRO65voIO5RYDY3j
zFw(6bOYh4=sAa$PSyr1r{G5CAT|cHLR>s*h9OZG8=Au_uhYs@fD48MHxdMo$ghh;s
z)74yVk$dzUF*N;z`Sj302fiF%UG?x_yc3-+rh0(&8R48@xZ!q%8Y510_gWu+WVmw<
zc7l4L5V-#cxMTe+)q&lcy&vr(|Knfne~{94MEHpSD^u&&Fhf1ZaHFu+6HcF=DC{nT
zwmu4_6xKpm8AJSfRz2O_J3hKU6aMQWG$Qx2S<&KKQObm#zC0Oa<zj`pE-~Is5NZuJ
zh<q2z9h7P-bxPV?n?t8)Ai=F!1K6Y5{xYv3C>+B~_RsQoINE#GJMUQR7OHgOEVLyT
zW4|I{+suZ(iAoCE-O<RMF`dZAwnU%NM%03c7_gqvR}toWw+M`mP^^l?gPx|5-DbWb
zb1Wh49<e`m>XP8?8Bk-p<`P7-Ijj}Fdx9H*xY7i3;M8rtblX<y*EKPGY3pM9ax}1x
zWNb#UWUQ#E1tvh<G0^=lSAD-v{^9K|k_Y7Qwnm?I-6}K4VQ1t`BUO*)YuJUxP;?pd
z&Xkad9&OeL5nfk8Oov@fIDMS!y&c7*R*yHAO?57AfUFUzO6`T%Pv`0yo)aOg=!FOu
zjg-Q=y<hIY7?=4(gF8fFKm~q0u>98df9rMsS>ebL<hNGp?9!6Cy?&g~DKcbD&rb}e
zTmG`48n@UuGYRWPSsNHVtG3i?1B4uZc6L2i1JAbq>|HCJ24lW%6P3>Wsp26(RuM#d
zT`6O`dq~F1;^BR?p`S})jly<@qG<|w7i%-I1BIM@hUatBa+Y#yNjh&KsDRE2B|k+|
z(Z>6k>bfJAcSVP|X$2%NevI>s@5wtOqg0C|+_|^>Ef^y^NV9t_)RBNEsFIufC_#_g
zbbbu=uDs!X<{#?zrQ!hQn95kf?+0#I*gxESEL4gLMeCF?7*Gl+(aMm$Ke6^;me}Mm
z1?NU2agu`WYp{(hKvbs(cC4Y<iZ1@5<I@TS>R+TfDQCA4d_<~{*|QbGAee2}rv2u0
zT*~f6RvX;Y^)X{$WMpWw1z07V0oLACCu+8~WF_3Cg<ac%z@-$o3LZ^23RE8GEsf4@
zeYI^WQPHp*sr8G%Hx$+J#hGD&y(h!^zoTWP2Kd@{@;cyE+@sGQO@vOLpQHW;8LV&@
z6q1(QUWmT-Sj+x$1f#HEaI3vL<d!<7_EdE2%H_R`;c9QFcZ6Gwwmr<=5Bu0JYt<|D
ztNO^bd~R9VJQX@w&|(6eU_jQ-POcheu^HFnwk%HDV#0YY+nddHlLhP3Q}cDUqm<4M
z(ri&#{+Xudr=lgZcQhuN#25Rf+A)NgWiO&I)P~PL57lKqCPp+EVgnqGoYXCTw=~{E
zY+2mNw2xxHLI&Yqd9u3&M0~!Sf9Y>Q!u;r6-Kd>#E2jr6uo^7tJWL&KPfs4cNEaCl
z6{#A$n@)=K_+h4zTqct<V;bp~IZ7kZqABikG}Yda2!Cw6PN^l=p&&jkUZ1mwF+SGV
zYrD%D&{fV%NuemjHLdxNv6%KSCZX<0;j1}Xh%h`UDga-VorihB8Y#dI9hu!S8LM8{
zGGxD5QT86!+wE@kafU23;fyL+^POY%h5K$A*_^`(Kas`^t5)EQ*|W`;ST=sN8}}Qd
z#skd)?<;(>skeMn6jRP0nX>erj@$Qo|23;1w67-|B%d(c?I)KWIP>*l!1T_-^ota&
zHzDw7sF?<W*grWm>yHcuK+it2%Q^7ca2_EN5;Rn8)Wnm^9uf(?joac4=Z3&pN@HKc
z2#;u0vPW3)+P`&n<<zqC8<-s?l93un&x>JW=>c+8uRE!6-devX)JWUjUd4`>(YSBG
zP+Zm+&q#5#ztWaCRIPCN`8;O)_!@B`{+Jx*2C+p|r<mhiqFHqYKGJZTI1Rc^t{H1G
zuMH`_UJY!BHvHzjdNHchr|?gh$!5G}?aAPW?Ro@1D{k$VLvP%Ls+dg3V9A42tk>2&
z3w(?DSa_<~3fDMqR*X-GVKp7Mx&5l>OTXLkC}W#XiW;!}UvF|@mz!AmO&dvcLrbGy
zZ2^Cd%FWGgebchwM}p0$Xq3Em&xUtomm_gUMHA`qbbzU9`?#Bg%Qrq&C&CzD)PF(y
zv0slbg)Pm`HXUz6?ULuM^wMB}7<j1ut`0^Ap3GcxxJFHxZ$5xz#CVP+D0Nr1{kk~D
zsJ29dTkAR&6N@Etcf`1oa~~Rerk7U;k--}o=jNYVC+_s9+}M$#!Ov!!Yh~<BG#Sw-
zv@8wI=GMK5Zf%)J%{#Im%q@l^G?dOfp9ceaj+@fM#>Z1KYlt1y_l*1D%qRXGPj>1j
z(QZ+EJRwy2*<-sjABhag)<P~J*fRe>;PyR8%a*JAgqau*Xz`7?;?G!@hKgz`@U*XZ
zLuJlxwqgF8CAp@q_v@-G!ANVPuMStWs{*v?z6DcEZ|RHBgh&q=s@Ch|C_C|T$<_-8
zs8V!GVO?A%l7(^~;R-|%Ccw-&*;mGS+L3I@+u^v5y$+lfE%W=?CI&r&Pr>*5sSr-8
zP+zo>?s`6zjM1Navu`2}Hs^GSqFNKeNe=n2roGU4AN%M2=ih`-u49_BzpgssF$nR!
z8p~C?(wOb+M4a^V?DTah!dJT9YI|;o62ccT<HQAQR3-c92lm10dGzObq;AW5sd23`
zabw*fMGVn$@t@bc<@+k|P*_I=9dVoDYK+tWQMlRO<~p_UJzrCW9mMHnM!4VM#dHE>
z+VdeoefcMTtH{)`M+L>R?C!)u01Lzx)>~sM7eFaB{;_x&5B6po*t@PGgDr}H%3ywj
zk2=K3fa3z6bl$-)M{VzUsTmI0Utvw=f^Q8m=S2W3K)NrYWUf_Dr=1zE>z6GcQ(f=u
zq4M?m5aH|IWzv3UmHkY{av!-VJ_hpDSJ6Se$izn&tFOM&TdA$A&7vwh>V%&yYS{!r
zu*5poL&r?{Yn7Kgt069RyFiTDDtFf6A10B1<-<WWnXrRx`KA3r$Jq$`Pm7Arb@^Q1
z+K=SZ?c`71>KPst*)I^2a$wo_a~$DHS*|%a#$sJBl;NUa33nXPD0&e{s!;#B_SMI0
zH>iHE*|0&`u8bHaua-W?^0P8Sph}mbcdE_cuy|_9sL*K2*coHW5?6pPrM6{NoxilT
zauuf&8s&D8QM0h_6E39t%}>45J9ZcMrHGwDR8e@}4;nEgq{lk=fUI!O=L?EaR!3z6
z&kx<IG6Gbe{4GcJMSUW|`WSp6-^;t-X>L_3udVG3cY^D7;>JJM9G~RnH~EPnhM4us
zQ>)?}#`~(DKc4u9{LB}is}u)99og#8RBKys+DhcqRkhJn?@aOI&iGpa)n`rIzTTmi
zq)fulEsa%|LMHg8yic2Wp6sF$a#$iKi68TqkMc8P*FGF}q<%YdqK6|*p~liAMq546
zK__^s9W5*;Y$Z<Wc$d~4sLzpaE;AU&+|ry)ut!$?`mPuD>MWN%Z}@t^Q<Cyf<27CW
zNnK+PU%x3&d}b-`0a4LH@u;Gm%MD^feIqHmv<;6Z0%}wo&O4m?S;H6ErDp2cUj7IE
zKL{!QO?->jmFP0gdh3nt>q(kPY5T=iE~?umP9GWsFJkBSQ#=UH5|DMVyo9vL&XIlq
zco0iwqv~HKy$ARizO*d)b!Yo_)FyalJ?-nO0G;3X5PF*tKpp-}D1?up)2d<%0fI$c
zQ-IOgeE$%jGk{lJ!Dvh40+#eYQ2sFIN0T1-uaTp3;_Q+tNJ#Q*4*rSE&o_v^9ecur
zy%Mw2+I*#CD?f*BWZ~^ABNO3*KjSA5Um|vA9sLCJ020+spSiT0GHdFVX{(KKOtDzD
zs@bX~&hzZwmD$wCm)XlJmp@Dy>G-kXG@bMY5s23_mmm~y>9C(>;w$qqQ*^?2+M2}I
zgL?&LwcuKz#ROZccqT-+ty647%|S8?0MB!^fr?~jeZ1f<Y(qst)^gSt^aXuB87YX)
zNAAfqZ_qS-J<5-jb~^;6Rnn7O1jlJ0bYM)BRU6Jfum}?STqpO)5I5+uA%6K~u=r?S
zW%S&Gi#rNct{nTRt#`HMYr=W#=kq+*j`Z^j)3v-i0-Yb5Nv$>k%yi`>T}0{~L?SS?
z_z#XWIW+PUC(KJyFXhwl@=X@4nDCiatObb?ZCsF%R{}|Xnncw(eh~sYRH^!!s&&$s
zo`?@N?p~j3yk75M<IE=8|6yz3<#W*5OQ9kcWWIMYYeK#HC&t~ZeLs2pMp@dG7b5ky
zi<f<zkNS{44>f#OetgKLXk}IN$0|cDnp!Ya5ARNHB=)l?O^?xMz(*zWU4lvr(D@<g
z)pk)I5{1KY)79Z#L$mmY4aE<54A~9MAD84h(_DY5LGnGUEOp%r8iM#$_jI~8dDP`i
zr}$!DM1C5#Bw5m*8x%aZ*j=ejd`Jzl+QJz(IT)x|1#0S2#&a1awFQtY4j$1`OQM7p
zQc`<=T{&~fiCMj5TC~oE85Bw5dmcK==FukX0b%R))AEM8_=PKgNZL#Knb)4Qw`XxN
z-X_{?vn*G_MvEfI&**iu1;tsfX9ZRQMWNMcO98{&p|Nf-B$zeUopS=RkMhltHBfc`
zHZDGHD~`BjWZ(+(fFIJy4BuW}p%otr^U%1|ND|-bwRa!E4xN<wB`NFfe>ClX_I_P6
zgBDk7etiItl({=%@I`kt5TYKzbx|zcAADzv&KeV^+mkmGBVMRyjA*$H5-l9&QvV<F
z-UO_vbL$_qr(f09fl?|mC>%wg$W-PjkXEfIV+Ca%l*kwc0T~lW>x9aX3KAedQbj<7
z5T=ANp~?^uLWDp9gh9p#gdqtDWFX${`QNAQ_ucQD``!P!&vVXmFZ$pDpS|CAuf6lW
zYp=C_zuzli#FZ$W$j4?Z-PdRQvMpVW?DpF3rL1n?cVliZnTEzc6FB&`Zl?z1S+hW{
z^2LX4CP&hZz9yxthgzuja-coSXla0ptDT$)(AP{K&6%=2Te%Sh4Hm&GHH><=?h2wE
zp*b|Bln_W)sI2xI2L$iD#%jOt0Y~01p+#KA)bK~EL4wmGnDCX$nD7@sF@maGi3wi<
zR25P_ehV*So5Gi}AnKOHdR)*WA2$L{7icE(ddMKu`~dY~)XV!YPM$oe(QSYK{{4G)
zOCP(uLg;oOq1+;DZwSJ7C$00(ZaY@mCDrdg{?p8j$-q|~X-&M4U2N6+-c@Uq0ntSk
z@h!R}e`7<ta=&j-J)ooM&<dLNAfYsl#3)T&kHP6LqHCH8k~T^$=+Uht7K|M{XI<$x
zU(JT1mXdU;@J9<t^{ZDiT?Z$p#)Lj-v`VtjBFT8Cs~TUd$3Sf**J|XkEn;IQ&*#ui
zw4)JF<|rkyfU6)6o#}?f234X50U>24uT{~wq0g?T+~%eb+C<V_uyflWyHS22zCav$
zRJSTC%f%o1bYHAt1xdy4F45aHDy==xm!@)~jVV|uoN_Crl6+wu<mLNq=)C<{1#m=$
z4Ndmriu&PH*zwSeX6opeKm$EWk(`>L7+G$l6S_ic{nJ0)>maF=)9mwv77_Ix_hmT#
zJk@p7Aw3%)%d4v5rKqi<02uNb$<o;mB|)jCn`N)PaE@tCr6Cg!Dj<FIy>Sg|bIo_F
z60!q*h6x<I@=EBKv+ETmCoE_a=y}VmfBEi4VZXPn;uyz8L-`C#9of*F`(<u+c{xet
z>t~{8xnA%AFmJM1mlg->pjq|B@IjrL^?C7+7E}DGIcFcuu&6PK{-xBZj-tfGL}M1e
zA%vPMG6ehj1fL=Z1cGma7$Yw;e0@LDMeZe&PhacGI1cHFo$JXVR9Ji1JpvvO@HtyP
zEgpCf>_S6_w{~aLkGU*Yt=gm;#9Y?uT~j(1kp3#nRZi!LDu{ra>PQ2Sm=>j$;wN(7
zh)4k%rEF{2K=cg|P4`t41*z(1uhH7ep%CHB-ZapdF$LKR1}}p%RuhB?W)?a;p&&t6
zzl(s5cOh-<I+07kxN5;J*xswrPAzT!m^RMSXc+T6VA2yEzB)x?4t`UX<`zD$L<Ma#
z#O1nuK)<6>Zy1B01@Ky_TlVTUmT`@^%Ds%oLjpDfq{Qv^RNF7mZJ7zk$KQX*4AeG)
z126`8c_k?j!M=1^d_ScN8dzMh4G}gCP&-d>9|5#M-o;EGLlZVv%I0QlZoW7FOX`)_
z<Jv2C%TrApZB<(m_z$BE!)d3|+^CvsgH=B37t7IKq7@ti<OOHs`TRTT?K292@*R%@
z+R3C;O0n7n^o$(eVcl{wfGBqI4LoXuJlIVc4vO93Zb=XNtCn9C&#l--5RnRs(g(d0
zT%#SQfC+jGxNAKTc4cS_r`W4R()e+-iF5sgUL7T(O{N`Cf9I6w*&Tf&MO5PEojU%V
zy6yBI`JxUKwvcmd``Vaf22ABimu8Bvr1)bfws!r}OqO(Qa-5ffrQ2=bmoHmSq@pj2
zQd8R|03Z-Qb<F5$C4Eymd8IrZ+bz!ch4a&&7l8jgdpYnvGki_#IOH-Se44WMByYA>
znI*J6pS{)K9Ey0|j0*f~EJ};#gLPMpq@4&cD{=0&1*+QtwB>S-Fh_^0U5$nXZ$f$F
zPQsWINhoRv&!QDh_B;_#J}Pt60cwx1UtR}Hj-pe6+-pHXwVxle+H_PGU;w(fltn-P
zSHW+<tP5N}Lmbej_1tcqpzUXQzJZqLKV$jw)FD=K>h^DH@x95?Y*Nz>NOLOUO60TY
zB<bCYX4z-5y9?x-FhtLY>YToy1Se9{+3RXrb2(P;YE6cc<L?s=8r<&ngVxO5+|_Vl
z;;x0?2i~rHz4i6<qQXT~x0=mdS6C7(!Hpz5&%^b|&MYm7m*(bvjr-4Ow7>7u|5nxJ
zZ;a}HP0mOz3WUVL4|OG_(&Q$mUfr{=leVg^Dcs{c@AEK?esPUoU@7SU(u4`&ZeByJ
zy|Ls2QRI5OnN{2fm&os3`tGBqO~Zq?hl}C2UmwqImP+8u$Q|fL*22JXGWd2vO)1#f
zRf?WPK70cq8peV`%T(80->jpy-7h`KrCc>@#}9|!eOkXGIp};BWq1S<p4|(H(&reB
ztsaOlPc&CFfBftTG&DFoM$-o7>l^X)@=$a06Bh^5@Qa$(aRA8Z)hA2;&4quVBykJm
z?#y=bz8~sjfo;vTB_}&^G(h&ZGw>>)gG??xk<=X{i0VPKMj383o8zFTbw8e_e!D>+
zJY!`d%}r6*t+@d`8VF4i5E5Ls*{fiI#mEuKPJ*m9MQ+=uQ<gVdFb6F}|N6)1ukq+7
zivuMDxrqfTPhA*g<z?agKAv56vkyrh5d&-<U0uudh<=1~3V8kAhG@`J`9<@`HmISa
z;}j(Jp5k3O5vZykSnW3|!aKQ89U-7b<6jB+uqW=g_O9IJm93+@(OaEuewCs(iDSHg
ztG|!@Z$hR1nfh+|;La3J&R}&)^=(|E<Fm~vr;ycwBayXWhp`lQi~DyOt+kWAkAVVB
zPR5JH@afU=4XsLd$b_Lj9(SEQ%30~db~aSbviA8d`(eA~AoK@p2{my?EzfDE6P&%G
zqMDcSc+u#GIejFgK3*m9rK2QmbZ#e0If3sRSU__G-j438Ek~`p?{gXMmzB*`s<CxU
zo9SSU;-l?*4<ut1D^NtQ4hmMGzCL2QJ}uShN?O54%<pdeBmo}+A&qVySC4;>KgkYl
z8|0Em8z(pYBl8C$JlozH;;Z`F!-V;|&U&Hu%IgAIk%dsJke2qUhvWQVH&l+Uccn|t
zUb|6RMiyNC>O1lYYszsyP*QmR?HciI=g3Qb$uKQ4sq!Z_-}9L3gQb=!ct?ORB`X=)
z+RYVqEKDv&LHiLIgX}1(KDoFP9aVgWI6Ygb*)cI;(Q$sMGz0+5ut|wV?tGl3(8L22
z#Jryg`F1kzufJcw|Bdl0v#D$RfylG1jq*v5v4ZEQc<9%0uLa!oYUQ=^g<9uk%DSrA
z0H#n<G;e>Y02v$Kzi10}lpHH3U0lGJeM29kb1JqftP#ITJD*G1hL8mU&xX*kQEJ2w
zakMKdDdx!mx?fAmmOPB(-NzGG!NpD!5|tvTca4YeF8E9TdnpD}_*X%FCOGC+JzX;h
z&g5mB_<CZjDH*F^HY&-nLiRRSI1B*F$$sZADSrK@YeB^IoAS?G)I_JS(O>oJ)uBy#
zlc~zr^y&b<2AWdumv>I=?=?a`wnN~#CP!)}=sv^wMT83Ik%v8SfMX2^NX5W+u3G5w
z&PmV?kYWP#uI;gQ9txXWA>8J~1mcmEMtxq0_0Y=#tQ;aH3f1dqXcA*Nfk8J=d`FCL
zlBTgaq_S01PJcU#j@KjT$0_0LulG)#nez*mI_;JR6r_GHI$NC@sABg*eEVM9YH)vL
zv&m4ZY!p5`lW#?GbZ#Zp9OYWfwH!qCgx5v^O=p=C6w4*i&A_v*vOdcs(MB<I2MUWh
z(vPI*x8}f|(q=qQC67{4Mcn#U`LGAIPCo5QYl{BJ7ykOIp$<1s!_%12#;=ymk1Vs&
z!xEL~Ij%{Q8&T!OL%EYg$@|OH(CBM!siRRisNcm0-4Zcz3>0EO84B0>%r;2C8U+Wy
zj<*|DO*JC0<CMtGyyVx-Q<jGh1nN;rUF7aCh3{i`NCb*gYyBK?o!ws%a5g3eGKC&q
z#U&VYj@&G>zcs(PJUkSL#fp}erdHMLLKvgcC2EX`i4V0*(eNhoFz6=NYt8!Yi|_t)
z5C2K?vomRp29(ix^+SV3t3m^zsqqwsv-dAEX5kyQU(LkM9D+n!NGI!a<?oD!G%u)>
zTng}V-am}`ifTH2<=WW+<K2(rt9}BrR6Ay#u7**aj5=N!_#}%X7ZPOz5)670aOIUn
zbyW@Xnl1C`)}(4seQTM0P7}0W)t7Tl5K63@j`y8fe5QHN1OHrm?0<SO!R2cOyQpjp
zX;uqt_ZdI=bQlmo7#^nmhuoJN3c8$V*PmOWH@s&h#VtR+7OEF$q&>Ycm9KodH+$5{
zE{?69R5-JaJME}qVL52=GOC$jwp3~0gE5~@o_q~x4^Ig%S${&0`6cDkdg6yY*qBnI
z4|@)!aZ8GBH&ZsF33-*vRRDjKpApyFDY0i=J8KJ))+h1@D48QnjlDQ~n-j#&-!^%G
z)vrCMYv`G)i85*#h$<)DvpIe(>il5utKQ+qkM$!tIQ@mn7$`LAOsTcGa7(A}e74T1
zN|%e*QC{DHe*5qI@U@ZfS?4^r3$bv!3_(m!jdT_MfbhivGGgiEmml_AO0zz*aeroo
z^$^(BCIXkEqkI^bJ%AH^0@(;&<=lx26`kT~OT6AO49JX))rb}B291GOiFvNOy}7#U
z$5UU2Aa=F847-^|X5alj{wI$R7G3@56oBvjjj9yPncz`}#rv2~ZvrmZb_8_Wo-`Hy
zwgj%2tlYLwlCpI(N1-1PFk!bjbM~|@1$%%2r_=CbZp|CmZ8N=N2y$YMWS=BwIq-dz
z3-<d^fIM{?yBP<6hS;WHCGKR(E&flx4}PyR>{E`j@U}GmHh6!uj!R%;l9-pKDLXr&
zzFV~Y*>gerZlvU>P--EL&kk9daQ*~1RGrW%?2!|#1a2?JF3A!KL;%%9p5cVdNE-WL
zkIl7;N0F6(4DkC%f7dvCoMULbI+gsc)}QVC!r1=P?#U!#OaX&GdRLie#Qc_-qUvPZ
z#JvOGXiOT3s#g;K2xw5H!D7x|IG?h32MR!=EwoocpZn|5r`7aD_5fR>YVGPG&X!w!
ztU`F)k2S6<Hrq1e1y7AP4S1@N4>mXTK#h5=)4uyY>xwHG9FPt5hH=)X$wAuMQ(Rxo
zSdS2J5}NCG8Hq2Ru2=WI&h3L|5aZY<V?dSU;L{<^0q0qX$63%fqj#>E312oYMmH+^
z7G2Mx!ydX%2Y75N^6aNrZ|_=v-BIO^`f@VkMFK{(T!0~sSg)7fv{b`)#g+gB8?Ie}
z*>g9!VaxNy=j;baH?CC<^u^Hnb7OLWk3)D-Z^j8y<C~c1INw%n^Jy6)jat&k=JPq%
ziGD<`fU5wd5j*cffP_Ux_vw9q=$-8gYr8UjZ03^9m6uZwU-d2_o?x~!``DZ)t*yzB
zrmU>f@3yDXA45!Y&r~3s-h>)m4y{_r&4#sInb#XC|Ex^F9;V9qkkkdA(vR<Sd9MLZ
zcF!KCsl~pJ7dJY~TKjyvok<-@HdU1M00I{}-vti{GCYlZ&KIS2UY`|l0+yxNcBEFS
zTabRSo9ssPJLC}CyuxYhY3Q;-7`9i1a4p`Fggo?M?AxRfW3R|{<1Vv|*DU8%VrS2y
zet5wA29BL|V&zFvctC!tyXcHjrJI-NEVVLGmNRpd)ZQDXWE`=<Wx-Tz+FBvK{!t_+
zCUg8H3Pj5)jE9hIC`H_effSU(jYSGpQEw;45BP0`pSP|6$6criD5#O4Cy1H$n~oFK
zIp6)=hw(rEhqC$hfkIt?10h0kglYzL^~unTH#`xQjQ5-F%wY0-&pV`rs-Kd@W{t$6
zr7Z&qAGh+>rb@OFg`Kl7eSfPYy;y9!j67FiP1nBs9w<bxgp}qbUTVjA96$o-xxCib
z{1Loyfh)g3jZxEBRvKLMd&W;Iz${jzjV@S4?Un>`Zd@Phv))3!qugGc$1~FNku_~|
zZLWWs=P?Qs^Ys`^eGqizw>C`RCqIviVv>IgE7a?yCnSa9=XtP00vlBak%LgtTXi{V
z*5g@&qsw6L`U=Urip3hL@~El%ak?UD+GKdcRg&?%W3WK;HP_bgrv;dtMp<`IS8?w6
z$ovgoh`Fs3%u0~J)^2DtiIgbCLY@IFzWL=^0r-90p=?o(#of{V+lMLIt+yQy`xpBZ
z>g3tZ<(Ht!jRU)#WJk(z;as2Yfv)do<iP)-_IxYA&p?2t46(_ty!y|qL|>WaD*n<F
zQFQTy9^JK9<p+m?ohg?HvlE}=ga;%gOY^6uW#;XB?cS_c7-I+s1$A{yJ^XtX0w2bi
zp<|s*2Az!PV&QrWIbH=-^pHv93u=i>6J=c!Wy>kJA7Y;oC#xfRuvcYmBKBsCy_o4N
zvT3b4#<X%(l$N(Y?9}zuzRw3GmQcB?PEP63FqZjbdQ!8kPB=y3yIYwJe{%LeZ617Q
z+nnw#3MJT{de;}77RP1oiIG;@)@bwZeo@l*`mw<B3g&|S8}C*3cjLhsFVyfq=s@Jq
zF_w2v`~>VXT55;=UF+9p9yGh8vP*mIjVhu_+Ox*3R8`FkQ=)#EH~6;hH4vEwZD^xE
z-dwhsK3M<O-H1BgPFDFBmlvx7aHp?)8W|pZrw;g|3G_>T)xZDDm#8l@9s`77OLtDl
zp})+a?neke75`5!R5=_L{9#XU*#&^x-azfco^a6g7wG>F*Wf<-<zJ&a8$5V}w=^9D
zZ1>o|0z56vhH_HXXW}%wRJjXS=7=>wSku)N?x=LEJ9$y$wd;r47BM^(Pj8n16Mich
ze+9gx8x+V*EY|u1P%60T-(m$M;VTf_JHE4vZhiT~o*iQ@z!`Nj^>eTcX$HRicu@n`
z*BOX4SB2yKZ-0Lt{tXaSx~w(b_{-9m*;V^&j$y@<$Lztt;imp-A2SUHQk17Vz+aSW
z;tA+IIGQq=8hvW3knOo%vLda3($l%r^QjFKO+;|aS$z0TL1*|Fv)0+wCDqe*aIQbK
zhf6P93iYiHgWNAH%wh~JS!I`)V<%lo@!pw?JVlGSPYX&9XZwE<0L?=$CYb4aVWV)k
zYGn()uYjsyP9*57R~UD@4hLg1uaBJg-D3cjEhPDy_^271<j{nSYyJLk7nsAS@i8zu
zYgEGnXUR2B(s%|PYN-Z~j<mMZjjVgm&y~Ac=UpM8O{i`$)_E&P?(*9k;uNm;3RAi2
zp;2tdwh3;k>7?%0<z=m0F_0M2R9aRqcJfM@n)dZP0u|Ler%HnAMftD6&)_5x*0pUN
zPTUS1Vf6Kc@a;mhv#5El``SWR&50HIa1>v_PvNt<MVHSC`u;TMe`LnG+C&}ovELuB
zwG1%@s=1bxzX&s}o|p!2O<grR*6T4HbJ;E>&8%i2ODltBSWt2VQCZ@mhUnZ%KM-U)
zA!XiJ1-Yc|Y^QX+=KJ`hKq$+6Ck9fzax<l4u%~G!-8UNBK02R7)Ddt+V`g!Sc;;3b
zRNdn|ZrJPPpQd#@V@0jcyqKj{7sb1RzBAE$t;IJ8)zNtVV%6CLeJTvc=`$&i{r$^#
zo3Ypz>hB(oiJ3jj(yf@$+OI}xFD&kwY{0M27iRt-`><!U)ju#2fP;xj^3*`CQp2Qb
zP8z*2VG3_&w>HRCRfF$vc`1kB7`P0n#%R<x^EUa@JpOr8*+CM>()8QN@<1GQ>xk&9
zNa(tq(Dv1NyBoyvgqX&4`7%9l>zz?y8&eHCj_wGNy+0BQjQUw;TIS&CC(X%)@fO;F
zKbHHj?za`R&@wUM!Dw9V5SeGFvWoKjy3ABV`Gh&%KAuozigE8f>ejci<?QBt+x2g~
zAODlte|BK^SL|ogqU4HXayT<Bdfvtlp%7$U)~zx<Gj{mmYg!feFdo&spcX{x>^W<w
zUwhVq$mZz9d97AY#h?2vQdE+53}wtVgl5%|%V1yT;SbFOEBt!gT(s{O8`T(*P()YX
zWX>q}723cNdsn@!)6rOfYfk;hV@hPeTr%yv#UwU=N`Jkg?5t+aZ_uQ`XY>O<K)6Ds
z;sbgCr+ewDS-lg!V~~1=;z%SXQZGv9;u*%^o8f7PL?MwlqC+qH2wi#{Hk|8la{+ZR
z_|B&Tf9InAHPRO<{BB#&{J^k`smnVB=CJr*MkALeMOz>CJf*EJPX!@nXHC6cEs;wo
z`sz=uY=HtrO{dKM>*}K)30&W9jCTS{yWJCDX`l83mi8q5zb)-}f(i)k#Yc1vxA_4}
zJJ=K;;!?YXz%Ajo-5&!|Suaw&H~~NX?uUPmEbJ%Ft`YNzXd-2J-Kxwj{Ce-KRdwQP
zw>yw!oH@`~l2hlEns;_J75C()LB~_q_#*!ipvY6yFFBt41wB*1YMib&pO|?MKTMEe
zII&Og<D%melw~d|#lS9<Mjuq!+5M)BtJ}(kRE3=b-c-B#*8{+TgXDY1033ET&%|EV
z@WhGn^!@Kn*|Ud)ZD6)e1*ZlNrv5<fhPu()RT24hm8KBoH9g|G&CtAA2ew|nWBd@}
zMAE6c3~aZ*KUIk`y5A9RF|qpc8eKrA%Mgl&1?lm_yHU=?=aREn#|%R_YR0T(LBGl*
zim2*}Whckl-{&rc1a_gO8U%KsE)2d(hk!Q>cGe-3Rf_6my)n0hNTEjqwdTiD-~4HE
z|G@n8Cs^RMr2G|Tuu-BqYuw<IC7T0=XcNS%4a*HkGf`X!=ckvtxuVrU=}kw~ho(xb
z(`&_^&5gTq)hl58ahez${&Kvfqcy$rPI_>Zpy0uHJ<HISqf{m7DACr<!LkC+1_30V
zsB5{a@ruSj{V|h$1_JRLIMNoNr1FNiLD&7~p*lLR<bjPWsr6denG23`m@umH?EW+A
zAh`6`rz8iO3b{0Ct*FVvOzas!GcV#E$pR_LE;iBLKa{{l*k`sbhp{H;Ih}c0=i~Y1
zd4Yt;wMUA4W}7;$Z~igt{|naj_%&aDW(&8SBT%7>tH#39ZZ;7!w%Zgo_Z<&()@JS&
z7qcE>>w)$r;L~%$?_)p|fEa`KzOAPDOZusDJI+Y?8x`h{;pF5W->kz(qurnndmw`q
zDh#6xj_3@y?vUbl|NJ!nHKAuL?H3+9ARGuiX&Pr`)Gm{~@5v>uSXb*~GIG%<Tc#%W
zLzV$l9N&+)x&2|!U_(ohcgJr4g6E!$3P>3mCwQeZzvWGWC3pxRTQHbbrOv!IfRMYe
zj99YyvPAYXKo7XyyzL1rOXW{fy;cH#`Q1<dCd-mNJ_3{aVGky>#@PAPTVlbj#zwa)
zb&fV3K=I_??armhu2|n(3&9G@tF+e|cZJ|y@RIuq_34mR=!k)}VIZeEhZN%yJAt^x
zAbR0f(Y}Hc+ug`^;aMi6TY}IC;eww=*^H+RsdvdADUw-Uc~tQ1#_vAZv*)O%(ESwd
z@X7T<7c0@`^f<7hdzp)o_<^oF!@R0i8>L=vcJZ{YBM1@l^}=q&Uh)_Wf?0i^A5^al
zsZNvx3`wdX&H7!vrw3$(xGEGx;Jh}MUKO*+R1g^wT#L9i(#Uh^_Y0Zhd@#kx?160m
z-3G77i+|jNKh5O-YVLbJ_AL>nw$Gw=;hJfr$ygRjQT$bET5RJRZ_oPHJDBI^122bp
zRxR^U-}1$qS$3gG+0(Qhp7k+IQ+Fhn)z&WN$JlAuRis7+6>*L1G_<Tmz(3&6GZf*M
zX&L4RM8c4syd%=|ZA~qdOzstZ=fj?{*RZ4!9$Im^`TcmS?KISJ!bFS-DACNqZ0OVm
zM{jQPQnkcYX%>lBg14yr{>4Rg%%M@s!=di=n;`A0Z?Wk}+wRzo!TgG#=`|O3*S+fm
z($JDV`*5mjq1cy;EIZnc9ycWloA~xwZN*lC2FmQb)<I%VGFJ@p$@{Q}CRdQUUiXJ5
zZ0~97=f4b&EgH~r0J7M{uNL%pV{r0m9vQ5g`DVU0d}^NO#gaHYuI0qkM86+uYwRFa
zJIqbY<y()0=GNp=57(|^uo{)k6rM^OYeH4ry0zWZ{4HdhK1rYGfvZ}#`;1q?ZB!&m
zhH$IR;PU##9vGz0ySzd@xnbdDo?4`pVTQRX-9WDeADLowof1YehLE)Mb9UC`BLW%~
zaZk0hv`z*MmVPPBZ^b`6aqq27W$h%Nk#Tm?i@xX_bk9P5phfi}sLIjT<Qiya&G*r=
z?@(^ZFX)vt&+9nlT4QC~0^20hChhK-Zt%G8TM2STqW$Dpg`rEkqH9roG`2r5CaEmW
z3b({Z+J_J(7R4l90HZm}Y{|0i9E3SC*TRvv(mf##?cQh93&*Lk@X=5Q@`Sx6UWpqw
z236A-jBuj+rt8Yfi|Dv|6LYe27K7{EV(2Bg%b1=Izq~!z7DsM;J%+yg-rgjut`$Rx
z9E~>6%(`CB+TxU;U|NJE(_xzaet@QF%douCtk;~h_H1N5^p$6|+I?~PdxWiR(ul0M
zKOv39(w<Bt#(F|BmW+Z5Gb3+wOqVxu{c4pS)GU<RcC?x;4P`>r4ss)}Jf4CPZXB<6
zvq)BR4g94?KCl>hyrLNBLz;EQ`PDmT!_w{SqOBQRg~&t_$OmRXaQNwu=i#5W_MTT8
zMU4Np=DF!!3%*@>^j6z%z&S}7bzCub=MDf4PeaHa^O8+3{L>@t+4C=|c(AWtTV`XT
zfDL8i6p+j4H{1I@z1S7&(socr&L0hxsU7$}%PwA=Fn~y~lH+#_p7g-$0qL(N*v15P
z@Ig|9I^kohf^2#IZ?WzMK?^L_1*Cu8PZ=X5^aubD8uMKxQ1@8x+>DznLGYYDrf^|i
z?*o4Q-DUqKJBu?iV&18(60@9iVz@br;hCHhK{u=|zXN~n55ZfqtsXw{unYJxdsMF?
zq`c#VQZIO40?;#h#*0>htY?;VMT42?jc|;~?9_YsI$0UZE=;_Um=2ED)S%5i6r)Z1
zx9;z@A7#L4_shAuRbjM|<~eeBq#v<-O?_;l;4L+$akt}oznN_K&25vbnLqvRojrdM
z2J{EynQK@0u{%8~&8%5i2!8XihN*@)-NJDmr-(c7z%J&SzGFFb0GN|83hVdzrFiKM
zE``jDPPeeVxh1KEH2IVqEf=#yCqc;+!k`eKfh(-sl8{%s!-IDNCJ9^tuPE@aHA1(F
zZ_}uz9fKjOeAsi?*7=ZC4P_HfEg{IP-%tmBJoW7#llq6&L0U)D&fau#^5aAc`KVAc
zVtZfvkxr`>fySwDdVAyILv^(OLs*K2oD+&+v+LY{=!`>bw$AmQS6`he0XtR_GJdmo
zE)eL>1`h8|w$58<kL(PPv2k0OFx+jhOP|OvtM;NassK{!d3(AKYTp*sY!cqo?K378
z!0uq=Mh+v6K-XTrX5&rpCTb@~7FM^j_+DH)UFT(1kD49EXcuI^)cvE@nTbDNgOx(2
zRZQLvIH8*dmWyG9s%&^!V>$xp6>P&ztIQNF^aW-UW>r1lPBfR9+oun@XUsdo9al|u
zj!FxUjR9>{R@fjMR{;60pQl!KXKHUuRpaB12-mpBmhLrcl%a+cABNN(M_(7&SZXnd
zM8@Zo#_C*%0Xfne>dWlx<-rRMwPqq7PL0oR-=LMitc&m&`j7!9M$|<>c4y~4*#=g>
zP$P-)^$1a{f1lCBLQ$mI6q1NuL7f2hjy{_7+Px<+1!ZzomE6gndx`1u>W@%H=+y1;
z(!VIZ8TkgFqQYe4wqjC;x+vZD<KI^{Zm;)mE)II0HI-5`Q)5_34Z*`5yHVTe%@d7M
zYN@Xiuw4vsurNxV{u+EY)BOlE=M>gy=Wai8F>Kr4wbYgREFPYCy4MT&47S}O{vmc@
z*96`MHV8U-678*Tq8oMt589+bn*5S9H$ycej+K`!NS{5yj>IuUrxlq2>kj!wt<Z}Q
zgR`HZiKu74kY6sX8HqAKU)uI0UOt`jy2~EEwb3CC97{~aKlk@Uuh`yp3yF*EV4$4+
zieh0nhqdsK!qRAadp*WX(t@c;<Z>iKYscK!lb;K{ztn<(xvHP&ic{&x!MIMcNBjNq
zaUJIdK?L#jE_6K@n_;Gd%G)C8g6;L6Q=b3%Dfs7X<8Ou4vcY{3owrJ3rN=Ehjxd1p
zea5lPKr)ft+B=N^k?vMy-pi-o{d2?o?dX3Yg8));NX``;|G1Md04kD`SU63p{wfZ&
zOS9M#BLXk10AZQjr@!`U+w6jt5N#bu(R8R>b<#%?fVW-a&Be4Xi@*#iCTy_?u}1Eu
ze%Lc__N&Bgp3)Pe3C2IY*#~s3rHug?6f4qS01V2c`c$;5EblA0($c@*h7o8O?s%tG
zWYIO~@<R^`#>v<dx1LYFu3hO80lbhK!^}VhDofAZQGqY+>cp$w!{@hG>1nJ--4N(r
z-WozZ<6W=h;5WunvE4mwz0-jq%~i*BF{^E(1X9a~J$hV^(KR1seG?H`G&F0K1~O|#
zru(l9lVc(n3@}rsHi<khky(p6_q~aTo(=-B)>-m-$bPR~BXtWXz#j(Msnsb$%9|()
z&0W9W{NsW9q$257!55d`bsm(}T{VmhX&8*2$lfqymw7x6o1C1CBSTNLT|4dXhW?7G
z3H7(p!9i?IUL85%sP1y)#23Gy4_zXIBx`ABc4`FBM=$G@cVwTUUk)d41n=(DLzX90
zCJ*n%251G1P^l|%?debxZ!)z4TRi8ze+zjZqe@?IzqKU{VU2d`6b~d5aD7XC&G(|z
zpF*qh0@YW!7x?lzZs$IG{l{$op@mSOoyv6v@h@sxoyy|VK-ctgFY0M1S75?`R<2)A
z@<7Y=HvqH)qOQFLwJ5{|MM0fB2sKU<j3TbQyt;Q0_Sv~K_^!r^wCMhWqQd+4`6l2e
z7<QBDRBdM>zzBhMk-HVCpLiPn37*pdYl`>JULe%NAEm(1gudb5pvLaG;f4bJeVUH8
z28N-zbLJ8?_b?ix=f~=w3EfMLHoCiL@~zao_t<CB?Gr0%gXsp-&v<^G0LDNKu1IMe
zt-Jb`S{?0(BCH){EVZlQc9vOlp<f5jn^|Jai;Hw#hfKa5x{YZO%H*1a`XG?LF2>SH
zRmBEGGE~S+-|qK&bUQ$U*6HbZDpUkBvYQl{o48n(nDrEx{-xo&hO7U{aqJvYA`F{a
zM)Ku>16ti9=hkWqT65Rg{^UDGMJ&hEyepMi6A7$|q||lh{kz^}o(tJXAm0nyIL(`Q
z)s%Y~HI-VlY*?vQpTHpnj*y>pB~Am?MdLA}xKvr<j&xl2YvPRjo2x`M*Lzdwb8bn=
zB$6{ev?LxCZeAntI~uekW-KP{*g1ygj3ectS3BSxqYwB>!W#jGw?mI&@o&Z`dFt3f
zp@^%Sbo1=Q^?4j?A=-Asquf3-F#<E!TVN3@18e5bmz0Ky!!O3Sgg4FSjyFw=LO$1s
z&HZg;;s&7>dZh)29Id~ZUIqx<%Bu$(hQ71g9Zy#i6${t3I!%Ih_ccVWEDN(8?2B1P
zKOgh~oSgU=c{6Bq0oolyKcbUU3?ff`*kk?UP1=3iL$v&|NdhEitPrPBFht4Ix_41(
z%SwG@Wp1XGj^Pw)hkiW~LPz=@OU^tW29ssKQNbLA%|~^4&)md49QK_&JYKoPWhs&h
zn$SiCGxIz^va?;aN)zK<HL`tW37zk)UGDMK<=pUV_t(1^l6{w!lEfR~vsTfqwAiA`
zz(=#KWH2;yVXM)exyV_L4(h7ENB7S3oXI)4g`Uofn9w&iafG=(Vsy1r)}ZSvq4>_j
zgSC+<o7eHH6@DHB!@2alIr>~s5Zoy_Exsr5r%#4ln8}Z~@a4Hz8rkGnMjZ@Jl*wkY
z6UO}4OY##wUw(zdljsDrw;txJX|dObJ-;=ygIYkE$**GcO6qcx`%l;&+wm04Gc(=6
zm3CNON5%mwyMO@t8#v~`i&^wUc)|E!9CLJJrXO!4NyjQgpG6#Va$0e1BZp1*uk3SM
zxn#Q3nhm$P6qt}KRa+yUnTAl|E33Uug-`^?IP&~46CjT#)}_%fqh@yxl-x6{Iquie
zNz66r&z`cPL+{=hg~(Yt+R~1NVUL>XxT*tnLqlzM9O%)O>zOPrzd%c^WEUy1`gS!+
zOjud6jbuC=WPY~1EoWnN0jm%(w{|>Zp?l#z|Kfv{#Am+z>O4(I_t;>}-1$fFLh*)w
zF#{wekwZhr*JY|<;jjvgEsE^YgF>-$(4|YS$wS%?75TQb*YL_;xmR!lqZ+CP)^q*_
zHIH>JXCW>SNeeb=-wphTD$!W3^zp;?I_^IMR#28zecMJE&)dvnu4eEue&m#PnSWE6
zwTo@#HxV5fU`+5!Mxym}&c&wPn4E<4bF{*&OMy=(Hd5DC3>H`mJRiTP7!wg16jJ{5
zB%&U-y$m3h1WwntDeoI`%7#iC3N<w{-fAnT(E&ia@Ol}{8^bC6>Xv7x=DsyIo!3Kz
zT<BBQkh47R)wMbm75MYaohfTcY^o|fMokN}=AJjx`~W!<&9hPDt3MsGd(U0cUvym%
zc#eP?m!|M}5BO|h`_&%x&7iIeT&Wt!4(ZQr<~lMAzL}htz}lvr&wQJu6Y0~b5|t=8
zvAC$L`*wV*hM-c1R&NB6Tj9HIuBWCiW#sosKb4-~jWq;~Z+ZUwQu!o&CTmgRDTm67
zvC~#_(Z$gFE<TW+)MkP2Fdzd~o2jG~JHpeEHJbvVEIT_I9MLFE7^C+q0KhlV)byNd
zIKI>R%2ZQHam@{+>Fn(LyBy<G)Y%`#Bl`8yUzmHB>Ka}9xv|<!?YXPjU}<i-TvfL9
zS1I<RSKG1U-)}C`uJuP(^(V<5knEOUZo5Ez>$sPT)8(=Dn4z(S_+hMTWYCfzW~ZmY
z+bb6PC5h!G$T{nqu~?r5ly^)#wJ_Z~eD-*OyV<wHp<7f!oc&5odePHW*QRkg<`O>i
zE=TP_XF{1(ysdS7$h@*BH0KmL%u#z;G!-b9J;`BXYKD8itfD07=`Wv;DDd`)1K7!4
z8eSp4QTMaX0ZC+Np<3$-Uj1QnN0@zcg)sA6O7j{#$*`U9GW6lF_V#gfv2u*OpLJrf
z$5AL=dPd;lFxoX3zAlh53X!FDDYpChTgb$;NK4_W2^fDS+)K<lbx!X<52&XHv|KSS
zI&iP`sPn?EwvNCYX8x!Sk_fsyx{&LTUsF8^iPnA9uMPbo*AbT4ahH7lRTJb4k8w9V
zr1D_Yw%Vh~G+VZLDl=vE7~734uxs7HyzE_zEq;^}6+yQH`Fe2&)BL7czJW3SvB(&!
z%qlcqc-B*~9b1A|9$2Bn9?Xm^Bp9dZ)>mTHdJS)$a~ft)pfF?40T_3Nl?z1QLFqXc
z-Duzar;X}=;|FCW*cCt=<WoZ6OP3be;gWzaR)MqpMX1(AS*6S=NWBFg_K4m4-VOcr
zf8*r;^^kuN==(z2!;H#0zg0si=K8{W#CUt?Unf88p?}U_##iQv#TIUWI^$tGuWmL_
z+NW)MJ_71#1G9uU-cLZcgQCRz5r2~gbj;pHi-Od533rh){jl^A+G@=LIOECIjQBW^
z#;vzX`vi!#IpC>uS80t7ACjN?+wT{DHn8avp_b%(nbk||9gD9k4aeJeQ|y4|$P+7a
zvH{7hk<=*}F)dS)uDWa0+ZZ3E^%HQc+{BC|C_^eRl<->k2<v`nDInrhu!c~QS$^14
zT1US?bpK<7--r6U#y+`3xqAuZ?#&qeO^+$%s?`nxh2?SgQCn^a0XnZawz&;Z=~GMc
z&U-@;cRFw}1r)03+dBme55B02{#nd2K6TO7Sg@~LeD~E(vo_tvP;QKHV}kAkbMIN7
zo#-#a3@LyUXI>%=Py^>&+XL+!cElaE@ei2n<Y@PM6XL52TZuOtd`@bO$<lQpwGrob
z?Lt(kar4-rR1d^i?ONA)T;+W$jr6(FrB{&wi@mIzzNc%JJL0kfPO|EPj#t6qZs?R#
z**;&A8gQF}-7it3j&MfaN+>F1e%G;)dz@9BFgVK;LgMeNfz*$TC1)23yV>(8X?ix|
zqz(_p!|Nr=%gRQQitt?hoEe`9rMC+=&6CMzld>kC&9ml(LQ`S59m2wn=sVl57NRi#
zTQ>3Ls=enLZ56zFq3!g$SNENW)zbZ8kcU^EwX(m4y)^Gc{;;QV-byZ_G8~rX={<xe
z>ZMhTV=IGmn%iCnT<N~58#v=etS8`;0?I!R_Z_Zs3)jpxncTLKJ5HAbo)}4f!}qUt
zt&NAJg=P>VRkjvkdpngIi|@_j>InshiC;wXOc_kFpD6Yn#gW<&(3I(v)H2|c(rN&f
z)4%}XK>I}DHZiEVV|97P&U!86Aq3ap6ML<ATGwHMf!?WcqTM4}D>^z1EL-QeXY(u%
zHme6hy&SVOG+#{n1g?6vJ=<X+e$jnr7<rn}(Aw>qwG%eK>T?BiM@U{bt;ix@3eN`E
zH`mk<=#e$+$sBAjb1{h<FYf3)D4wi69MtZO(19$}MKX$&nxW17biNbNi$dMl?si4F
zm^MTN(?|O#Gkg72S7%E>XhykU1IJtCg?`2z7#<@BHjkSX;hQPQvef{Srr$mMkBtiZ
zoPuoEC3IB#GtyZWm=|#opFmDua*o>;Ty*oIK~ZxgIo)tDt+vmC)G$r-Jv}4(9;Can
z^QQB!j)i#aAaZ<T^SJ1=WniE!6o90aSKofTHB-g|6!M1KCUlEkn}IilQE^-%GK3O1
z*GLE8;*40H<fgRT+Hz+MyQ*D{NG>Hm85JT%xa-U%QRGqCeJn1H+vcqc9@JQMo_grI
z#%Bi}C!Y&05h#X~9{ugVTuvgtDvh&dae>%LwfArJw0Z6u;D4!g>}GK**>4{A%I8LI
z$%rsbTT^AsQOsvuRCVx1f2}73MUfn^8cVUGA=_^ofD|c6{Qc_2A&`9)$tGv1;awLT
z6bio8nI{2%ZeG;n`+zM7Di)8;>zA2r#a*ZMtX;>Y@0JlR2%Rr6U|7hq1f24&6*NI0
zJY}h%u%%B3^WDkg^gi~XH8W&n@I$7Z^?<ZyYzba#F(enp$4v@&=65HgQCRPk^Xyqp
zIg8!a`uau)CC#GZ0^{Dx1eMlkOEk$9=*u|JdVbrV@-|n&Zv&7lLEZT_jVmwo7nd(H
zXC&%n#VNa<E0*Mtlv&*j#pdUr01|vc2NbPTfep*LDP*#5AT;vE`|OMoCfpixgFT%D
z8+d-8u45wvAFvR$Ryke4=<DJHq+`<4Iu_q<yA}!8B%5yaB37fxtVPE*%cWg_&8Uv2
z7{^sqS92dnzMq;k8EG&zQCPncqaV`I3%Tid(1}^lj`Gxx(u+0{`s}JMuS__4o^Xrs
zV!^84d@D!CpHFiZ__ovOWF%db<4ZgbFHAXXEp9cfeo5aO_Oh8UmS_RldsxoZRZog5
zk>(3lo#@UHJ;H{$m${qq{@7cqlb88^cugFh;;b8KHGe9kw6!lV9h7>x$AUcVM38cB
zzC6lzwbFtdk4hbebSq4p;2^J8hbHr@>-q)7_jE6}Hsm0V{*oIIJ6T_kPSM2YPjE*=
z$|%C^n~UW>mT8J}%22EM0LE(4z!(!Vt<2eplo*rrw}>T4A%eJ@B##ZR)l3L4db){L
zDOH<1=M%efLt>gKbZUt)r);lx<=kxShEOwp8XBW0hn!48`lra!<HHhWikU6N&h=6$
z!28y(n*3qU{*vO$SQRq>50+mEHg8-)P4-_zSGT7jV2?@deb+alc~-~nQhOF<XT)=a
zskOC5>cwimfB=6#S2q%s+5-q+1&{zj<!ZO^%BqhFbQG-#u~?kCV`eJYTv>$%_1q5h
z4O9eKnr#NvxGNsCjP+>Y22pbAosyS*=F5Y0FWg7U2Q>tU^dpP(p$nvnZPFT02MY&e
zYZyhvK6De4a6p%?s3<un>0_qjaN_EAs1?xkqKZ=kJ{P*{fTZd$X4)djdeE)>SajT7
z!iCXt<b>I3q2FqIfYuQW<@x*LM&3@5^c0$^tDh#uH)+N7*>thE^JO3|HJ_erH`6^~
zeEt}mSQ;0_6&W+f(=Zx&J(jI@>T%(sdQ#7N_}aPj$PnQGBMqZa$4`riZl6@wZtmrD
zGpC~9c}jebMib!W+;{5tm-AyZ29*73DP1*SC`O4-5;k|G(dvHNEI55#BkEyS+_ZI5
zWJ?Sn=Jmep42EKGW)4SD3c<6{7aRkGe2wy{$yZA?qHG_Z7OAb165Ktlu(Xj*P510b
zJyQRAdmW2YZnkI(7}WqLzav1V;HY_R-y=rCsY79zDHJ>&H#pnFtMXS17#W|cse7;1
z&Yc|HjP-9QOwMtT8O<#k(HbAH`gYERZcju}LOSVuJRFT`@(DF)Mh6vJyH~m3tWFPO
z93H>8sPB-VL)GJ5&P}-n%IgC?%4+Nf)!!MW>Onq#qI6z9OJje%Rm22}VCV}-HZev4
z)h1sLxDT;r2R+fKpr9ZRNHu-?%aBpn1K<n!H@zmQmiDH&O9Jow?!T;){||rs|3wTt
zg#=<)DIbVoLl>10;~k+_Cndwn{AFyVyI8D!3HxEs(H)@?0PkAfFxzD;AZR9**Ze`h
z0G%s$1E`PTQq$NkcWa{;lMzzRK|tHmURVKunnSiFKQDJ~!EeQ-ej;lg>uf)UmF_5T
zC@R7~Zo$6+tZI+Wp*6#;GxOgsra?O3uv-|FPy^}Se5$%O>%*Rxb$(^{>QF4N3B&7U
zdZ&oQ88eu&Q9|{rlBN@NEBUpCwIO|L=EF={=eR3Z9$CM&?mH`K;F)2P%6HkA_t=zv
zNz#dtBx0zcem9i56lxIA-p}!kQ3KyxkQNQNE0Vt%nKNb``)=pfGHPf+Ai0;LEUDDt
zj4YD!rIR4kX;=c$P+o#wUv)E@m_FGF_KAYQsp#&}YlUA&|KY{|*w_K@3`9>I7^|0B
z3?y0M*i|-a9wz=l%MuQ%t4kSFIuP2<GPPEqsx<oudV0J!5~(VWgDYWf&@5TICb`S*
zJUy6#=w<#!`Gk9<hqH~z&OD13(|VTAU1K#jdS-3-pbJSYmeF0gCv`G%dma*Ab|oI@
zp7{)N^!OWJZDoJI$tz1+&7*1Ld`yX_ZX~M|V&`2PgtRi@S{eINg-uUGsdYYdiwQci
zd?^b1AD*Cv+TL3rMCUT7svHW+rl_7jnlO~4_d^BDbArHHu{vfRqmbr0?K|!5KZFhs
zL4mKh;B<D6tVQdL5?ag8FJ=3<K_TPTGq!Zmx^J~GM+5OQ2c)<jZBsC3WeZ!`2#a22
zu?qIWtxf`~0KqUY5!8Zn8pFfAbzm-s^w;bh-2)M4Lk~}5rcW<1#$To`^b{YTr$#xR
z+W~lh)EzyB%$k098f5o!uKl||y!;QW=LJCQ=(&{W-}#Jmh6N|;%))SdAPc)%Fk?q@
z33CO4N8AA*c%)*h^diPKVEXxipo5qXdpvp=qC_|qemGo4L;SSyCht}1_G;iEc4o~g
z4^zAq?Y?hoaA=KMK53(diP%k^7;uN+W4)K<Ur!kEB*1Q}eLlzTykTnJ%FJsC^T!od
zV6!Z})1v=sC;xsSdU5_{90P%+F^b#2#!et?e2<V3)DqKy&>#McyZ%3~N{t6-8IQ*m
zMk=;G9*ZwJ4)ZG3G$;Jr**`OVjUZ|?tMXf1dKt8CaN285>&@hkPhxYk?|vaq#+}di
z#lzf69u*8S<}7ix>t`R2%v<gTxT3OscpW)J6!Flyi)rkP;^R2g<a#C~(Ug0=59wwe
zcRd@PV(1j`W?~4$92@KVD6G!E{u0l(+qfr+iS)9I>}imeBsS62sQ~L0>?B!o&bed2
zR(0scPycJe4cfOWbF}EAdZ>1$+6X(|x4;`;s~>BPS%tZ6m^Ix12|{zHCwz?1W<2FV
z{c^!qGtkGova^?oC<}493S&eUnxf|mzmbNO+SmAWU|-lFxf92ZzL?nINAp34znFbH
zXL}BJay)KAW*5+`&cX{y(_Ll^Jq!E6iA#n~twua_Gk+NGm6*zKIwp6WS`5VpjPz7D
zrx2qsM`=q8jcrdV#lA4}9Wp^st+pG6?tVw&)s2^&@44R-sUmY%FaC`k%dpDQf-0oM
zJn9MNTG{w`ws0y;PK3J=%7u2aL6T~<O@9PIe?4l2jywNsi;W6fC6LP6d!yH5J5Uva
z85)$e$gqv~>+kw;t>x`is2y{HxuDTgEh?=PzO#{Gcd)6#$vYgLgp2aNk5+1!qh%_>
zm34{_N8aquFwkY}@aFow$e_XvZ*)MxGsXxh1R62m<obXobg2dw_J8K<`n5IB-IBwj
zh|jF&cdTQRb!&jL9T<bjspHk}yK+OBz&_%tef!)`|I8r&>)3xyBf4gyo0@+X`+Yjo
zjQ0%8nQnX=owV#H+UlgI?4}*4q)BaIIS|!4gllW(dZ(rgR`lwjU;&u0sg@8r&z>#M
z`mkr8Ra^&Hww4IqA(TtpfeLHD%%&JYuavpC09`rY@9fCfdD-DmAd#~RY`djkryKaN
z-`%wzD*VZ=dnOD_u{7DdH}`l^Vu8SG4+k#Te*a<5IQE7Q(zlaAt4G9iAu25y(NWLd
z0wgk@NPIu+iTn}h&MXID+Ma4LY;Qzfr^t90Kd(+v=Mz=J72hQPKE(gP`1HQj_U+`2
zh7PmWB?@oGQsU8aT2Rln>GWZ>b;s(IXxk%PQO%VehlJpU!BvruqBm#nc^`LL&cJ+4
zOjV!hmgXI3jEn7eoay`|O#O<w%T(!M%BYopx@Qj6Zn+B;E21=(wdN3?vFAnlpyFc)
zjcRjTVpAf~Y|Y3o&uEeFXNSBx*Sr($cyw@h!E6JM>2ys9n(+zlF;U=9Z=Z5EKZwza
zT`V_L3#oZ2nHt8p11crN!Tk%Yl#Q@)=VMIu^%4Ek6SJ+V-*vXORNi748B@)|!$xoi
zQpcq0r+qVvHhquexXau2j1V8=4HDn(Bn#Nj;XjCjyTlnIJkLO_<m|)!zg?&Lv34O1
z&^b!<*?UujAQUOw%80|2G-e)2Icx%PPi)y7gwA&5tg}Z_4Dts#n;mzY=koYXNxI&(
z`n#$VL${$vi?ES2VtBE78$H9;IH=f{b+p;JpwHotd+?Xg4lPiCNV}^mq3qeZ`~Xkx
zN+PF&ux)3X+hJfl@x$%{&%JFXdGKgL6>EHPxqrXMAk>{f=vrfJl{w7Sf{iqDq6a^h
z21~zA2z-FqD+QiE6<a9d@KJ}@O@GuFTM>Je@zU2<&TkEI64i%wpf8mmbp2A2TKt-~
z=PCxXiI&psP`csel*5_LDlmiv47V1{Z!S~ZzCPTSa-X<ye7(t|6~d+BP5gYXAQdcm
z5)=2cWR1l^b9Rp9h$!ZM(m^{B1x7Z=f%uGXCduQ2)56_M+^f&ytaIsYl!Y3Da0X?O
ztE{3}<Y(tHv7!99vFWwXq%WNj({{6#zC07Qlp2~aYtgn5^3KAr{0$d$4_tr^Gz-#q
z93CaqaaK5UMB7sRsmP+HZG!PEYlK=HXs$;rkp}8e_&zEfl}8DgNoV7~4rY#Z#h8+`
zmyL|<^qn3zIUCRQ`7HG0HdAs5q$sb21cR+Voc{N%BR<@IZe`v*Tp|XYpbbSu2a=w|
z-(ITq`7PiK_)M`bI{s4U+a^wa4R?#@KlM5LN!`*~sDQiM#GmJ!c5ENT#a0GOIFg+&
zn^UOWsQ~McoY8OsFKEG9-D#oDDSaS*uIW0H<F{1ng|ziN?9l6Can>x2l&24EtoMGW
z;{;!b7G#dL%t0GR#Jx0(n}DbmnD-K~9Hp;jSYDOexN8I~=2)}P-Lu|kw{($7l*3tP
zQ$F?MUx7BD^NJ4Z4W9=M-TBD)2mB78KU3<pd){jo8#)Kj`_&%-SpOb=B>E$R01lPm
zlP5vv?CF$YknH}?ChP%#y9FTrgNt~w!0&$d(BYe_l|CXo@mTq-S@M4j8B2tMQjE!~
zesvcgFOHT$sn2D6;&H55tV%L9-mkZp@Xbs}9Mhb5^Tm}XP?yx^=0p@zkw5Zb&lT&V
zWoh=0Htfn0j~lcyTiy!?(~6yit%us(gfDwzt(Bfi;FUTA-o|#?aXs(&=_WF?$83|`
zpE>>xXgfC?PtgN;9w;wJN-`A*-A)2(wF>fZiB;6R+Ce2LrPILI<B^F^H+9X2b~<t6
z*EHPwCE6iGo%Bi(!>FJjLEj9hG9Q(r?o)}XM=6X{9#2Z3T(mn)SaeK{ZNq+=i&bs(
zVo^_tK$>?#?$np<9@#-|CCz>M0OLrvn=k8+de`MQiS-4N|9vF<y$5K<%;TNo)t+&M
z(A=SW9XT~ou~xjEGIvCfM*r~IaqWh!UF4%dZV`by&38`YAey-nJJ0|;dmz>l7hVG+
zJG0e#bm!u(RlZwp6IK1;>c9E?Gdu}{Ka#n7G~%}{!#W$M(UuQ;qC4&u_z^osnb;-6
zxXu<#jpd7V@m%sN|IoNroLiJJAgN@TMtt_I;;?7aIOo}`S<0Hh4v4`4fi<{CmYe03
zB-#gVZ>T-LAy$H?`@a)#8F`c6Lg(<z{C8ZpM}rfD!s%Hk2;~c*cC(36)RxIe@p9zs
zGJJYqFlQ9l?gycWGf~rYa9$J)vGsqD_vTSepKIT*-QBIV4wRyxGGr?PrA#98n7y@1
zWlD<*0wGkw90CF|CWN+%$e7AJ^C*Of5F$f>0D&q(3=kqy0tpZyK!iXbgg{6D&(D4K
zv!6cide*zne$HCwyyrcvrQZG#e!txJ{Y$R<x~}i%+dYHdzpex(AtAJSuTIlU3%{Rk
z8UFmw{qo-#{O{lGn-dXuZgKM9_ug%CXZ{nwvtg&=el*b$(_NT|Jb0HFx9`I%qOO9+
z4x;6aKr2=S|89DB{5t?2{KbCB<gPUIHc&?@Pw)c3=7G(Ceq?O_xgu{KA0$n(TL#k2
zMtY!a9niml1j!H~L+isK^N!@(T-e)xF2(;EjlPEtDaz*!=6Joc=h5fl`nysb3qDO`
zRlf;{&&aQ@ubt+%i-XBeTs>>Ht|clhl57`{1OxLm7=f-AMk%L|W=LTe)TswU@C(>p
zmDkibDSTau^6;YU8d7Mkt(Etd<*F)u)Yx!MWhZC%pnk{YXj^muy)<QUIYZk=fA`WH
z9UGL*;f`{i+HkMVcXAcN!HJrm3%}@gHOj%ApV~4<-p1P7GNS~6+ssa{vET+T_tyF`
zb6cy(B>`JlEv|G*NVa%2y^Eg~sM;Hv%*;pKv|D5nI|O5_y34)D(dI_nHLexWpF)Sk
zz&v$fw|!`lY4j;Sajb=u<5NvhmsF0e`^a2VQCs}dzQIVF$kTB#mr!>%_Y`!Y_h^m1
z(bH(yF5q16!S(F6aoGr)Yc-fTo<%vHg{*(8+9b&is*NgcuRyr7IhK1I<?qEvWS5f3
zN%lB>Ua$0X2#_e7-)iwz4u#X5ctkJOEy&ORT>n4a?lQ1-PeAlyA0McGI8@yxB!?$>
zCw4t~9X<O>5$7KNBdL}WxwjJ&(DEwcwj7X%*13kPiEJ?T>};i<O0t~L%5kuVJt~!l
zlT7AQ>(T2HBdFxcuVaI*xEZ})(J)e3F@n`;aCbC1Lb6uIB-I8txD2ZV8SF^k(4b|h
zAR4sC)y{E2rgCm=zd7y~4$X5Rp4>JEb{t;HWc#(pg+LUIf$-wC;B>UKWR8CSYQvxS
z!V9bGCs)uDuWYhS;(0Rz11j-d-PZAgUd*M+yf!J6X<vVV@+!;{RYVUce&zp+vtZ21
zak)01*>l5`{aTi_OpV?uZ5P`q^xQ@HBsXoKlv1Oe0vj#Ger=_Q_4i?8!^aOSP%2qj
z>igacO!muAt~y`*R44feU6hoO#axsob)iaZZtlPbvoWL==c^k_)dv{$Omnewd(_p4
zY#<@IXDdFR`iFY$Kd1YD9Mpd}l(mz&PrPmHDa6M_`?q$k#zuqeu$r5npPw&JM60CG
zykD8XBMVPf*@2%nz=m7C^?IdrElCputMF>MFg!hfCn!JpXQN-7X9?kvvrUO@eG9$I
z$)ILcbzg#TzSTyS(o<L9XTB|h=|xOp3eeHeG2=D<b|lA1jRDGtil%aJ+j$aw`YP)s
z^mI-I9TIr3t#`HRVQ)^e9H#f3bPu1?tmI1--YcR7XKwecJ~f+|nAoS&*VgE$!NCDZ
z!@8lc?q)l=gWO(g=JgMg`JGcM5IZl@bB38CNe7?|mn2jv_ZHoqWA$Zx4w%gU^fJ5X
z?w#K3A`$QxkfU#eSGyO9;5dY%TYmZ|>CvO~*XPO!2(KLq0l^*vtmVxq#E-f9KU&N)
z|6wt&ajQIk2}lhU*pIHrHoHa&Qp(G}Z1XVV={GY>Tb})p6H3x9xE@$>G3WdKuLn&>
zWJOnBR@xtjdN7(xPjnxP<ZGj+bW@cri_>3owY{(EDTSoJW{+qDHiDi=9DTK(<`IY}
z%qd(TvvO>!{%Q^8vB_4NEIDbO%>4_nmSC@wdNUqXNaQs_AHC1(o&jJnTtP~d2q0hH
zdp3e|Rf4z^JeLYS3adl-fbP8DHvRP9M7i3wmYV^CIYo_rxJ`drjtDSW9o1%0)<e=q
z^8nk&k4=#u+aEE^<`BTTi3t#y05%Le@FU!g85p#z=v&X&I53%c%MIeTsn>M;F6^4Q
zlfXQo*jbXYhAt3f)f(xoPS#S)l4h&=cZ}P*`_KWFgIa2{67xTerl_WSLIX{LsxE}p
z@haza)^xUfN7ZZ#KYeR{H<84B!O}jK*mYtes7B=dc&S2&Xqn$FkyoCdMJAn~s1<}2
z8~Z{K5oxrD%52HKo-1+(c&mChTwS)Vm_S&|jZmZNG-5?X%(rbB^E<BrJ#WwXMkQ!b
z<LY-PZQ(xnJ<!q6VNCsjm)#X?-Z4Q~HPAzPhdWjSp`$#f1#g?T*;z2=m1Waa9->2t
zc4wSHy7t0e(G%jFK1q4a@u-b3|C1e3BA+mNWY4gx8Xu;X=F@+TzL7ddLRkAmH>~g}
z$RvA(N4WJ=_kwi4nxrn~)WS~H3Dr#-d2uA`f|HOiCOuL5Waq@uYaNoF*4Wr4=YW|R
z+HKJp^X<U%Q`)xd5MV=|rZ6*s64IClyv<I>bph#XQoVTs-w<&&wlB!;A?haQF5pP*
zGc<KGQ`4;G#_@u-PAd0@%+W!MPVy7C?;-lluf8g|9+*OWM?Q$I+mzV?Z+gkfwP!f8
zSkO$3Rx>|fouR#7^%(Lnui37E+F0+N*M@U5w6qZ>z9vc+cUre37Lv{7bp%2f8dS}r
z6G*8V=rbYUtSG-hZ#T)oQFlzYjwt0P6nFJ)r0+R&^BX&L_OrX&GhSxv&o`JNRPJD8
zAZ$LE?w)wIm44<D$~_I%C8|B^o^ZsRjxh^*5)|a*-andk{n8kKNvw}M(Y&KU5`EIU
zAlx(1)S%cfd|%e$Ia`em=^i>pujM35OdnP~se4bxoFf|rL<KQ0zM&!~2Mm8h-oK8X
z5tcW@^g@$Z6*^J7mqcVF*fR@><(S_inDG0IR4uiWTIvz-w4!GY@Q&I^SoQj;_Vq@!
zTO?8Tp#H=Xb+0W*3#v>-MA+u7ERYPUKyG4hj&T&jw&~^@#IAD+>Fh`fB{47Dzao>|
zu+_-~x!45>wpM40<4%BDBX69p3KE`Kb9Km|7c_^4MnhfcccaSer<vLl%ms;AGwiXI
zF;hx_=Bii;NyCyX+FO47Ooa_xOsTZ4xXUL5o{tJ$bBiE0XKVXuOE1A1!KEP?36BKx
zi(ME0(?>OCn`L1hXKJJ`V6}9%5FMA8D{KPLq=m>x%(?X>G6I&jIq?LR;^|-ONU&P0
zb*ye;TzFgZRYr!kzNg6lWu=DxWVBzm{$!LJ**LH-r`fq^(+@a@!fDq@5l-%i#3AO0
zv5Au0+pB<37lK<h4rMn3e{^P)>sbJIgW93ptuvH>t#j%*tRZvax>g8m+((kNFo^jM
zW*3}(GcYq}+JD;?W}N~x-!;%pQk1%UIP{q+8c6zo<IQ$sf3s#;dN#84HO-d&Z13)#
zHKP|#m?kQ`7H~k9g(n9&)*9c7EhUx|dRgAApgBPdvT29{nK+y>$kj#wbV~~p?vl_;
z^Dh<oagZtEE!NcJM*u&&E6xa%(iBAfzRaeZJ1d;qUTk@NRnyNr(#@{DyQwSHso+1&
z;lKKyW0`i(fu`?Y*u-~u08-^@^jXZ$Ktc<paEB-1iH;oT%YZbQ?!rd!|INhZ-|y=i
z_}cYkD(@k1jyrq}$e8VIO;)a^AGFzv5B8JpA3T-aOW`W^OthUkSZwJToo18jpMA`4
z{)sv}$eQ|Mfw6gg0kYlrvfd|&z-3NMfVq{xSNk~k2aokRmW;dro(3?>GRQ;Uv<AOW
z;9dOV=Tof{!@Yw|<teFa(O7*z>bC>_oUf6al%%&coa~F=H#umCtcP!ZBbv0D+=nuB
zKpWxwyHW3K?*cWr2O9L7=sh#S9XV7q@B(PdJX;SG;cjf#I(#^E^;*Q-);gZ?;n0Oe
z-ofj6X)^E`1WZ>f*=RGS;1E0p^&j`}zd|n=u&m?V^Y;pgcFXhZy<3y&mnu-8%cxTN
z4Mg8KK4Z4!*bM#jkEJ#>VRxUv8q9^k*By9~teZv~XUGO?YputwxsT=*Y|i?AXNQuQ
z%&oV<g&L*Nt-RK#Yx-({6S_B2BBVXZxUi57E}!RXTO;qJwL&^q=V`<@@4XG3yJCek
z2+oBuySit0eFn2DEXu|+r2_Nr)?uqmREcMeEqcxQfa7YDq{(%1mzrEfKmoJv^XDH{
zoICiU{Q3JCVR8)X;?ACk<zyw8<wUjSrV+dip3mP|srY&mQ;Klg(W!F@a8)c2_%~-m
zoNF(9<6c|Z51ScGB~6yjmGT55>yOmbB0d?2K*-`&Fp0dEg?KS1kBS=#AG#WtSZGtw
zAuH<&YD??w<7c_XHj9nNu0(u(W@hkcC#zaL7w^w#yRYnu_@$aHSs)oZFEab29}X?5
zJJL51bj@XC`X2rI-c;G-dUF14nCnE%zwbT3`!hct5nm#9NY&B91jl?^sER$YR@fR9
z#W?QE9dGMJnwxe#LwEjCXG?k}=v?}C)Wqr0<To6;7G3v3U*Ov2>Eb7sAKa4GN+Itn
zPpaE&2AMcRL2+{=F{fsyJSlW>cHV?L453YDou6E%NoIy-b>gyUIR@$5txFGGkdwQG
z?HZM`l))`SC;ke(wVhBpYGtpdwRCZH)V|W@7^CQAXLbd~Jw7etJ3yU4%MimIG>F?7
zDe{zFHp}wvH4c&#;;n8J1IwEga_s7Yp1D=o-*Dq>Y}9#m9}enuf#nrW%EaS1uPkrp
z3PZuUhi~9b9wihRX&Hpta_HQod^&EgYNWu3EZpIBp_)X~J?c3-4S17R5hz{k<N3I{
zY$j`pdWvG?+))}&YLm2*?m}EUC#4k=eF(9d`9y4t2{prElq_lf?z@EvcHdIaiT*P!
zSC>tUjm>D;Qx~}!PTQxDuy_A4-~aXKPhA>j-IwAUiUgjn<QSgWox(L3EA2sX{bCDp
zQ(5f@F@W&I?ikZG?=1lt`DP07?Iv{P?BY<q?mSZO2TrVN{3PS&zE$!i=`pG5w1kyE
zF1a#JS1$)Sv?ooqrh*EW1Yxo)+HzpY!rSqWsm~RFl*Lgo8}K6`$tA^>FpCRe9*;Tm
z^&WK<b@MCnm+q=5hX$;6T0S)iB(`Z;LeC(UO#IWt-Vuw{a#GCNlCe92RMAcA&G7q$
z9M<Xq&ZqIC$+iy>7A0?Bi_K@sNNd9mx&;FZ;org*E_v#w%s8^*c33+xbv{5tk9%EY
zbfh~WbmCjUi=?ViSWr-@im#<c%1YR0&z_w$-MV*g1Mr#{8=L5JKFjXtnACH&ma(NF
z&bKf6;f&||97ppA2*t0e457uF3-nSg@kgJEsKr5quQs&Uz=D>mwPu2t-7Y4rnV6Gm
z0HYGvr%gOy-alEpYpFInR@N)99tP>WSVf(1>pk+z?{i_{Q3;t$xNeN&plW&_yndP4
z+5Kk0DEu4t)D(M!5E&EcUVhe;Uu)Y$#k~Qr3}6stFp9%2i%{tg4O`!|%|7Tia8W<5
z?(D%Jk;$ZrtzII5a9>><#3<APZWm~`As|fD+V3t+(KCE~m7RCR3pHw%7e}WqEe+<3
z0C7LB+GRawr--0?8|Xt)R5ZExR_|r>OUB`m8ModFGWS#_BX%&d_u5GUUJ5O+_?0wG
zXwhn)FcXwS5)d(+3&>D9vU4pqn&FzID_BXKcrE=t@Dc~s+N0j1rk48puu1bW3tw7_
zj;n%dY@wQ1ODo946x#Bl7+bs`@}4SG`J}GoapBK%VrCnq4qLovd~I(zWJ4S`>g%|%
zweJ-TW$Gt%+Rqx{f;hRL<HUlVa>B18ORdv1&uwAYXyAG6n3%OqOt?$KVr=Tmo!<4<
zwlmQ6^LNdaJh+#{?!*`x)CGoD4ZlSc?8>JbT&?%=B$O)M8;ljAg+T<oB6!f~xngEc
z!fnR)2j&MC_RsX;sS@mual7H2xRcb3=R5i_YqsiUA}L@#(MWS4!!+G`Cf~i=bv0~U
z^0P0fH1maTHCM-?{OLWf=9n{sOUO5o1r^}XhzP1SBv+BxoRgf~My&Y3W9VQY|42<&
zw}2Y9GMb}U`OXXz^tPI4QDdAL@=A;z*;Ib};EsZ>PL_t{W{G`_)`M=?iq)6zd&0E8
z&$K?MNi@se7WsB<?yEF<t0jmOlXF?9&s#l3CKBS%8tD5+_41zBW^`L`K(hUwN?_4m
zQpt2{9;9+&eP)d2J71ioFek%&u`^960eyWuuzvI%0%(DM94xCk@@*)(GMDON`WQP-
zx3d$wj@c<){^~(cj^~Qgk*>K560HZ?B~EvJLwjEF4Zd&F#Exe-CTj$iF>Z{rAGIx~
za~s5E>qW)lQ!UJ9vr2P9Ao}X@whcL#v;;r4fg0g2BVi6&+Hfg6?#2!329+h%PF6W{
zWoky`oJF@*oTF=;#Fq{r-iX0x&}QEx57b`hJQ<;Bv^PPx-m`Pdf4_$NMe?^(nq9F+
zDDa`5vNPBRNWR29C^J(L%vq{WKFYvera0_;hq*rXXn$`L;1YwgHT{r4<QYIe>(Hz@
zf2wh2q%d?!woWaQ$^tvuha%*LiL$cCndeMfTdiv{idxC-sg-5AEZWo$C0~^uoH|0e
zSn_~>z*KlxVRW3II-lgSDmPv-hK~x+9_mDd1GAOhI23zat-X;kXHKoHyZ}kdG|>pO
zLxNKq#Eqf*eE6>;)%!@RxL6aI=6;r0%O;a~aD)cXHv>RVug6<VdGefn=-m#Dl&8}c
z)|dLr9XomymZgLf543X!LY^@>tsKhI$^o;5`{Ru{5>$QY8`%>8&+oZd=cq6HW&8J7
z%8+MFlF=6S!8@XH=vbj}dY11;GmfbCr@q>lOgvCWw_pjF2%|TCrP9)QlAb9gxp&VZ
zQo05$187HV4z_x@?NnWcV>g^S8rw=BBbY(U_A_DTBB_@)`t?>!Q16ZE8S@;b;MaJS
zA6_8~HWOUl$koqs+O%%fns%3rjm*S8CqswN<}>JK=vcDv@kb?>uD_=`9npHwz9K=*
z<5@jWLaD+t%u})Vp1F5IHlc5}OyL&J56x?EJ)Bxoku4EnPH&T)Q5;fQTX`LdV3KIf
z#qVEvjxy-I+tD6d2O5!|YUQ3Y0XsP(SBIV3pK_<*FD)A}AQJUa96@6OfU3<8`^_1i
zUscY^<-w4=a;MxaXqWA}8HIX0q_8Px-ctwfaPE#-f@XBu;MA*XF;9H^`F28BOt4&j
zl#}?jp!!?kV{RwDhI0P~>7E9I9%I;x1Ghw#X>TJ9(x~PHBBAm0+Gd++vx}apwQ7-g
z8R=#Z>p#{~T)*2zRO}O%Bv>B!Uh#fBIw+?k)te1M;?mL=@EvxOO=ClJz-{>wuW}eC
z0D6by+*O)8ev@(To>Jt_CZ;k8?3>_VqEEsd3{>k|?&L=B46G}`PgAm0A8S{cHod~3
zgnz3u^>54e$7t9yzw{N_A?9uF@5b&EVcpdd^)n@qIp3~W7@nvUnq)U+ZqYFW(;kB+
z?uR~B9?0Bh%&owe*6)=Q@5Ph&_ZOm*;9GLj1XV`bo}X?;;)o2Hw2VD<>u7;AC|{}+
zHyGJtBAVs&doz`Td(YJiZ+lN`J9;TW+#D^KO+KSJ1rDbY&PGpAmy9P2MG4=2l|oOM
zp?8f=lp2?tQ0}e&*nZr8Y{Uk9lr7udOJFX>+wCXqd;6y|SuA>(1xXTj5hd^`SAYJ~
z^aPEsJbcX4c2pCh`+5^?xa0~h7-93;j3Io%+FZ9<g)Jk4X{30@*q2(C5J_)bAoAU|
z>tDCi-Mcpa7_TT1R(E1Y#SKY26YVU$Tq(?IO&w={VR8+@%q+;;iN7Y$d_7gHn=oh)
zhF^M~SaPedA;Lw^PONKtp`u>*^64$Pliit@E6Kf;M(h@eeO&Ms)~>TsrBvOP-}VVu
z$AcWq{&2{%2EPzbw5#MOBKU%i{2lk^WQ)SLLI&p34ngE@6zZOhIy_^NfbWAx(~127
zkHi{LNlQU)-9fK!#9fdUZ%apX>vKLH694-@AALi6R^P?eNp)GEmtz?(eSQLd8LW3Q
zem{Qia&m@!&xb>f3ASf1{P!>M&sYAx;$%a+(elIQD%B4-HVOuR1J2jUld7eO4)#g9
zTm8V1lVX^hDqu>Pmux_iD~#mW1c-Q5_$5>)6S(WQ(NlN;rq~DGh~|HL<-?&C$>|S=
z=9$E`8(7fEodb|-BKpIj&G3(a86yC-xbi3P{gQ<q+C=^@r#XL!ga214IUV|BkFKx%
zvE5j^Ao!XmV9J%L{ou|b(0IIC^+oSerl$BBXg|*x$iDy`Qbk2FfPJl`B+~|e_b6Cs
z@eBA03hANKG!;CNxIb{_$n)uOuBBtP4YY~Iy;)+WO>7qJdpojS5%YQsH=btFos8dC
z$!90DEBNrtkq39I_7&up56&l@ZXlkjd@(j&fg^bN5uxiW_M*uOf_c6Xo3@%qcSSiY
z!qexOtjXzxc!HO^UPi{FJtR1E3xV5Cpf$Q!z1vkCV#mclqW)TJ1opMDHn~1FC~PbB
zrx?#o(o9T{*a_%~UeqN!Ca<oubN162OGabKYGa+M9yqTy()`0Ax372U7us$c^vk*n
z9hIEU_~+nGd`~o$oC7z^&Q?ARY06lxWWRiG5!{=4_}R1yVyic|s5*$T<Go|p=2?lu
ztDJj<fb$8)_JQNHiRDw(EKaTgWkgx2T$;GXRwhYZy1^AMHInjnANNN<u)Q;>QQ}&S
z<S&XA43_)?{Tk=Kwk@0^aU0Hcb&$B$Nt3oQdf=Pkm!z0WQ*r8n5NG%iicNSS{6(lF
zxiRx?mVVbpiLcdSroXKE?MmfI?g=~A2{lR_lj{;otfT@&(>1im;Q`K&1J>5arm=r>
zOE&af-_}z<Kz)sQ9A<UmmznlXuTF}0S+#r6FcgaEXZ8RU($*N=M&gG<pL?cZ>yy-f
z90>MiemzfWML2PcRm{cL%2DVVauId<hYAqPalB!5`6k}fwmTuH;5Jbi4;OSwMT>9w
z?LO0<@R<y4ylzz^Rx1d&%@ZglfUQHAfL_&XhJ@c~VVT!3_1u%K$rl4~bv%(+x!Tz|
z`RE`^I+$ZlUc}c$Zm%~nOlQ;le-ABQ9TG$M%UiFYF)2fEi!y-8bB@nfFfsyLC}?u5
z+Z2?-eVZf)hK(A_ZCXPjw(%(ue`t7k_!|*VuoOg1+lSPvVGdYZ=GjXCGi|B~AU%x^
z&xztvDj!u|f*{mY@nw8F-0JvTP(G)H({1(=N^`sTfb=X(E#DlUQil-l-x@C*^QJ9Y
zJg?W(pDjJd={CZgv9^nL_({?;w!&v+p!Z!>%6i#kHIMp)5`SJ|OSH}SAdi}9&dSVP
zI<IGc=ccH;EUw#QYRA5ANFKrC6e8TFF{<AkC}+xnBvh~2er$#uE~x=ER17kLc5KQ5
zJ({2|`}cH+w@UI6`*F@GT-~T+Ex0YZSAR?J;LjtCB*B%f63?S{bs<#|rh)CHpcn(I
zV2=kdd=g<cXXAEj%f9}y9lm=rhR`V3EL*3osE0<_`<0%|U}#O6H;(NtN}5G#S2w9`
z>amS3cx{{7muTy4gHHG$W3xl;+L5BtcVj!7HY19X5(xG&K6^}h%`t4XFFD4v?GXWo
zD%dZ975R5?3wg6A--LJcfIY7pdVBEg#0c7IOte+0Q)mvYA-ts9u)yet*0JVOI$1WQ
z<sLVwVYfrMM{=v;itz!_4+pPQ6mLII6x6O{7Y3|%Oe+{I_a~Ga`T_)w-WAckkr5z;
z9e(52#qISr`Xb-7P7!^Zj0p<LoW&21cdM%g^-=IqyWmtIKKJLrs@MM0tfggnc$;x&
zbFS?o-HVZxL_x<;>wKJ>?G2E~(4b!X>HOB)xEG7n7C%|<50?8bCJ|GtdR%ECYWMN8
z{T!|Io&qJ#tk=!VF}S!$Kg;M=LA1}1_bZhmG}D@ZQk8qox#jypNxHO3tPAD(GKR1v
z1a)tni1w%yKxY+7>&Qg=>2?!$JyqM}b9Ko8b_C*JRHn*Sy-95(lr(#faZE{16)avw
zmB-~y&0A$Tm!Q~fqTEwFafGMXl>Sh(Tl^w`$e*WY2NqWE0#TA>z~8k6hYrQ@dik$y
z%Fs5u>8YBLF<Vm<6D#-r!8~c9K@#j?-5a)q<XL5)`sliA*G0L#q2tc27-aV_v-Qg1
z7%$S%^fS(<4&aCLvcX^X9Gk7&bKc4<_ig(L$o##n>Bl^o-jawGmaKbp3LY;>DNL%(
zwChacJl^P45FP9NTd$*kU+#^^%#j5%Js5VDoSiYiO3v<D8mF7$wfX}{RUXNo1qN{<
z6eDthlED~(<VD|L2Wnrs(;a4Iidkkw#oKhDjUFs@ADvtrxU-(5Y-33JE$b+<6T6-E
z%?-kD<8As^zn$P^4^GM7KrWN4^F_s_XVx~QPOA^U6|#jj4c+aDsw;<kcTik_#wLoc
zrQDxs?2GCc$u&AUWtrhmG9x!A#7af|%!1@FmbQ(D#%y+9R7iwC5PoOyVKe9Xb~mV9
zWJ$)Tf`|H&PfjvvYYoioA$J|^hS_gQ(OoU#B?G0fgb=gAhr7mJtrZjcGTcAIk^g$|
zXNCAd>F<K_<D<g%h=)z5%X3>hyH#0Pt(gW<K<goE^Il-VLrJSAV;w0p*L!Yqt@Uv&
zjAmvt@LR4QgTo-cZKj5izmlGoYAxEICr{h)y9D8pn9L2;*_rxAtC-kMu2x-qS$T2_
z!#5BL4OD#%=kJy&6JEF@`y;0`b7+-8>CGB-GW>nqig((4L<Tr$x{t3OsFw>;x_<!^
zzBHjjM^;FR%k2=<3`pj#z)}y2IwM?DwWoH5C9F~J4hO_0)rAZ$L~}Iy%6rJPrRd^%
z!i4v9frsvIowZ&on+Gx>$tk0*L=sr%MYztdh^L=!^_Y*ZF#2U)DpYvD+mdF39*r`2
z?ZYNmR$uKC>idwnhVF)nn%1*pCypnfEd?1T^zRXjfx>nDcb89ut*(ZL1#;N*LO23$
zAYKikqZS5B&AnYdR(i$eK#{i;5S<kli;ce9g8l`VbK5SA{f9$($wBFTo-+$go_d{O
z7#npgiI9-|xTlMtManNP&u__{_VJm4LbQ$|yuQ9kfRV22y5)rCI_bxo$Hr_)#KQ2P
zmvwb@_YHyXfnl$oSWMPSr+`UDV`BjVXtKAg3oDLVAcbV-lh%Xlov9PLXh0^bVdFoT
zGf`uEub!*53G*Z2^ne`HKZ?ymufKC<FlF6G?2omoJG{End1Wf>vl@qHxwg@~>6*hi
z*Q`bSpMJX3`Zxy%a!y+Ovg=7vV|xh~oQFFzS2yWSsGkoV^y*F&Yc^C;snly!g4)BK
z>Wcg*Pg0WB{Puyb1s%xP9h-f+X`S$Nf@Y#Il(*vLeVS*aTAK{Y-!>jAn64fJf$qD9
z$gV8gS@)~?WW4yP!L_duJ5yVibADp(u0Jsi3-V%U7Uc5mzZdU>It%6j5c016&wZo&
zKmT(l3n~Mw^I5L5%cm^mjnEr)QLsOGhHIu13rH0lI5WJ#|3Q9~^~0&G+MuQt@)=yg
z^-!O8O$BC#Ax;}Rm5K;6N2Td?C%Yht&4}wuSK4x6<f4qIX--v*Z)v^;Gb-p!9}5i?
zL(M0mMbQ5Ggcg><x!Ojkdu3_rVnF%=@&2SW^;7mezC#7ijC>4KzR5QTBumzUMHUlm
zswwbr+5_tKC^saiG$IE5dR&2WK-V#f#1Gp3GHOu4D>Z#?FOw0!5?b6Iaxj=91W0a{
zJnSsr;+etzRu$ps)jW8)rKF<I^tZM~D$hJD<$B#sI<~imT{qh4F>TEvcDdT8Fkr@l
zvCg|itviW<2cDwuX*7}822J&*6^L)QI(irL&T~!RL5Or_y<1}hc@G>_mOMPWEH_`2
zXwD?eZ?9ohbRYI)vObSEEopnIIn)<u7A@W8)!1b)itF8HYSy}vo0d%J>mHN$2tBZ_
z+H6KB2d1J)47=@|V)V^;)4*KQ_IB~HB)X-GeqdW?E6&X`upk0b6N26xd?hY3@}F->
znC%<{3>%Kn-Zik)2RG#G>B2KTgk<|^8H$|{jR>1>OMBF<R#?y0w%tt=X~dGuTVgM*
zgbRYk*e?^PmEH(t=QA6#;Ijd}1tePbh@8XJ5@Ua4{c{H#I;0`KJ!O&@P9q{^vzNPd
z;1khB<+;8CZ2Eh0f8a*k_O7k_<=KeJG1$NXGlhPptGfng1qXq3$_GVWYU`(H8)Xfx
zcEd(Ryq)+4Cx~f6Z`+eMh=B2FCJltwNJ38;`(wwvoiT>Uq)c|{X>(GOw{m}LnD^;R
ztmQDz7qetz|M~-2Apr^MJsxIhlGg~Gd)4JIGzpG52G$eRVBbv2?$t<Wm&c-+jh)B!
zZugtGn34?6{U!|^UvZdqt8+}eb;sa(-~&^t<nS}$y%W1L))wd1<?F6%KKJXx*ph9I
zcKFeJ4$r8R`Y5d-gt^;7O1`v3iMU?kr5zR78)dnht(md-Lb?mB1mSzWPZO`z%y!#~
ze0lhx?I8$LTU$r)_2SjPZ2iYW^M|?nzdHQiUi+c9U83Prz8VHbR9Gc~wxdqGnjRYk
zXdf0hv25RF*#rRWdGSuuRx4fm7(o4%N&a|jJ`8n82WIeX&=d|hHnR%%Dv}p1lXt)_
zr5_IUBfPhtQ?cn0VB#3&33y*T?69BKyRpCj*W`nbE@zszeu_2s==$dZ{Mk(YSCC&s
z3y4>CMtUXEF>1fGhn2f*w5b6&$*wZm4IBtuLC|DR+g#6B8N$MRH24_s+I_$uul1W`
zh8tjxtB17!oDpc;y<!YYgyms(Imb;i0&M?zg@0x6iK@nhgEL(Q^A>$d2=6RC<dwo0
zC!2vXmmwhcZ3Bd?j|pEevsEn|uRpsr+2TwZy0D9Y5~E#BmrAUh&4z7^9m}T!EM5VG
zLA2CYXgshHN26~wz|FthF~l}EhoO^A2#}Lm`$vcT@n<IQi4x?gy4X|u6WERjKlUi)
zMc*gfefxQv^)ow3&$3`QyxJ;E@^R;*Ll-LR=LSKgR$H5FO+JD>VhnEeYqf1u4pL>?
z*NY*pWUT+YgSc_QiaeWggX`^%KU-*hF!c-3on#A}!7o)gW~m7qiwaV%*Atg-;9;}#
zSxb6)Az5@@lO@OQNs^`n^)zvAL<$f(mboSw1@Fk#s7s@>L(szGZ+JL^wpZdrV-yz)
z)a??lB_Y_8n||HubX$%Wz^L|o=L;8<_1s)+wZsH@u{+~+07o<A{h-fk;2O*t&-8vI
zx_hbwc3!(R15Se`G%iqGy9^PpKCa2(s1BQ1O!Rvr6CN3r(hR$Y*fY1m@b(COL+l8x
z&@I6=0`zk_{961quW`G@&`Etsebb_ZBWs@F39e<K<TipcqE9w<mv>&rU1cxZtzYSP
zPPPbYi1o3JuPR&VqBP2N--x_b(!|*Z)AttQSDUS&6%_ahx_zJlnl15#2%!we#m3FD
zYZ2A*Na1oosVn{*L0N2RuB#-aGVd-CGuw^qS0HH=qiU6_sJWL(9}dlxp75IMqLSZu
zzG8sh0Mxnmw!KMF%?RwlgwCrl<-97Rp2m(Wy>#|AH3YIyp5CA3$jlv~w9t5wY%=;X
z8t7Wwn3-+04oVm)E1BY(Gd$PM#y4TEtf5_PA3Ls#$VC3)>YW`PUZ-!VjcZBIU}=PM
zKS#O{YDHrx{P7v-^C+~dFLHUUX8WPH$2uDGnp)gj(I|iJ-G~3Y555>w?ab;E<Z9Tj
z=vx#1=llM0PsUD0VBs8}Y7#6vg|?qz5C65-z4mG7sTv##`_@?H$2x?O)_Q9bgs*Ks
z+m(--AjwAXzF~vjHwy=#hG3((%W->3d#A-cO6pIo4yf3;l&v_zx?6chDOKfV^W@H5
zxe4T!fLznaM;O|;C19LDouK3Rf>tqdzg*4AK<7COA9@d|KwkD%W(Calnc(V7RBE#I
z5{Qe(*{}DFnSOBZKTFH&Nks?I&aIhiaWgwniw0vs3$X8|qU70P@=`>r0X~=kyV*0i
zB(^#`=F!#NMXo#jQ4~@+re$?b>uBPov9?y6<81!fOJOrCYyM~BmKvJ%G1+=f_XT^(
z**#}oN2<ksDOe5;UKurh_wxQJ`^R2PTlV&j{~*fS(3=AUk8f|p#YH^kOmexC^fq!^
z$?kS+G%TXf9GWYnteZ%O7;}pAdK!Sh<n?#x3{K2nfjP(I1y8isIi`Gh3&A3~IhqEK
z<b>uUfHKDgykoxAIaA5E;}yZzU9X=z11MLV0Epz1!+q<{H@lq+Yi#3}5eqcbRUn>C
zBN0(vK<J84GH)6B?mOq9*H>1~BaMM}Ng#s#E}^(s_U}EDpbh{~pJZhAF6LdE6S4Iy
zH$v`HH&!>i>*Qz^t~I~6SWnWL%M4gM4MDc7GR+n;Un~OMXWzR7U<o>P*wvL+Dqjz_
zGl@>OS**1~(zTviZ(HJ1O!g)dk1r`+$4$#4KqUyf^@z{*$?@5GaJAQrADDyJ8<I-A
z;7vjJve>pGSoK}cwD&vJq<70iHp;LCVo}hTo5^c-*V%Nnp_Z(!x^X`o`kE3%*EQx}
z%auH0ld52EogFXy@P{4xDK6>d#?DGyqS>=|@*{mk^}=xG0Ir~uW@&l7Ve=vG!Yp}t
z?>>3z$E>}u_sj|l@LvL}heBB%Ulr6``7Gh|<*#})($G$cw2j>X>bYfN#58Z~62Ach
zsf*GOLTaM{IP-@?dGY!?S;>{T>Ts1s^`pU?(#G9}wsZFx`C9Bmu&2-q5FnTFBA4i1
zE*GOtqkme6LGOw-&6hPk9P)RTcV7!0B%tEUiX!GJ!0AQ~FYG}+gdw2Zd3ymRbp3ts
znV_`a$hhFu4=?$~?lvyydIPheMH`#WdW?f^Rx9;Kg0~44pWQ-pqdf6)zIG`&A;c*V
z{2=1d35WR`BQqekgm5Bpd>1r`#`|(cM!B>%k@8ZtYj#%jfyiw}Du7+AVfxe;Emt)r
z)wA1P7OJ<!3RdaCMk^MO-VwPWOvx9s_Yl<aZo03~@N{pfx_OXM^XI_?f^)GkOU<Qe
z)R=Vvrfm!_DtR4Kz4G7>6Myz*yyE7cr|-{&Fpp-WH{W^_8ow{cuFsjmEy#tC1$vK4
zO4X4n$>PIC!!P5n419$b?ED~p@LOgKRDLCKI)<3RaBuXhzTsUwB`D8@`PriQr7LCH
zOOhopbhC9qlfjx<+kvFDJ*qs&6z<E%vO5uwvXb$c(Vk_|4(tj-;~Vh-YZu(dr@nW3
zb;|J3AMWg59{&HLudbVUWsGh3W>5~@9l8iqIey#z#^TaeB3kCK?0?YXng}=?^)lmW
zaGj>)n^3?>7?J<skp1vGv;O#h1Ee*dC6#=$D-FK|P?2A5=7Ht;)qkfV6Xmx6&8|Um
z>KLpFyuZr;LbjJ1^8LR;cA|lh?djNUgRGeUsG5Cp9P{?@Wa-*rDgSfKSJQ4~J%Q_z
z1)0-c>NBw1M0oT^L7Z%A+h|Q1vBf`2r?-@R3M?fUk-U};5|8(+5Af)F0;5d;p`-0f
z+Jj4Z2eG{_=coSYZTc@P{U>jUUx5|ukBd&sak4EIZN9kEx~FBs5gIIJy5xV!iBJpk
zFvUFxH_)y%`(nwgta=OiY|T1raY7#xHJWv{1pzlH+B1wkb-r8GUTgoa(!ll8a>x&p
z<w7M4Teja>BjnRuRa+v9h#|WPPaAWE&<0#oK#wB*PG3?5*lgX2$L4LvwOD-9AR39s
zNcz05<lt%SUX>nI+B^=~{G?(8RPGCyMs!~%z=Hs|T442euX%K5tHz*Drz?uZh1bZH
zX0N&y8yr*%T6PL$-3)9BD(lK`F$s0!xI3*OQ=u)vSI%X2?nLIyv+E0Xoz3=0BC9&R
zWBDXhh<V1A4GRJ}_)CXjDtb$vmqoFiox%8~O_kjyH;@MVt$pDS`KGnG7X-+;oxa8Z
zTvKQS!qtwho|Oe{-(>%JyBj}rPv{N!fV|sHKrG2PredVeTEsnOsWmxJlbbzj>(#87
zdwg%6?wp_(O1)W_<??Y|K5enPB0Oxd1qh05XI^@nN}&hlAz{svs@&x0=Mg1+p<8te
zF~=DSu9PEDleVekg5bo^w>^EnZlu9EBhgf;s;V62An<+a^C@dK<INW6hxU^7NY0O8
zdvTig7w4A}*NvRxYKUep*S$5ak1d!{#2_z!<}|AtX|RtZsO?HNZ~FMQh}EwY^bT%t
zlv3@%L-dCSNJ1yUJdIY@mag4ccmkpbzTV&*m%iAG20mt44h#ha#q6&{C+B1~JE}f3
zVnl~TRn^sfUX8GSG^+|5-DbM9<#KvZt80qxVBntB=i?jRLaGWWXv4WPH#w&&UJT9|
zcXaGI?_C6Jjvm^tVaV(koE$+{{<g`Ma2Zd`1u0I&uexK6=aB`bR)lYT)vYip=74&4
za~8-2PxD$$+qUs-SJ%tHsCix-w(F;vmnVr@eG6UQK@MhBqCGt;Qq+)FOl#fk8n4$(
zNnb_rUEXh|NRR*NWAJEQ+31wL?Z)Cf@qQmwiM;Kd+&8oTW&9{985NP$RuamLjxv-x
zvPpV%>PCujuJr}(ZWE+8;RzX`MYeVbgtyOq8zLQaKl%OvLj<0-$4rOch2=}=L*QnO
z2|%@5>Umzb!kX&S?d5%fYSZ~}yli)=IHuh|pR`~RQfLmdz%CF{Yb&aQx9m+g8H>p6
z<EwGmCRkDzOC3Wh-=D1StF28RF?LL7aOooLTGo&n(32)Pi{b7I_K^$7o&qu)Ik2vl
zN(nTKmb;FW5DgHk+1Geu7hXJQx?E1m&&nrRE1#~vm2xyJTvG)RlF{$w$y+lZJX`y8
ztRv17aLKFNpqSKbu#=MOC2RW7IK|iRc_oMy0bCqXT<X%qpJWsOy$lurn}a>Wp;tyn
zm?JZ%qGmsjLKo~9_mF6tD62S&wkVGIaDRvxUD!6Z@hE*Xzy7K*4~M{NSdVE~GxG^>
z>@Ab0AD7{1o%D3P-W9BFXwvu^6EGQB{+wcCm-?$^oL|%KmTF=XEjn|7B2bCDkYv-B
zppB2iBMS4Lm=|6Pwc5j-4nJkgls){s`)vdkG-bA1`M3R3OUY6%YS>=Hvm|cD<iWE|
zw;*NiEdi$pDG^B5lwS@<6&dl$yL_|8yEgKt2I)@84t84d2Z&hV2!WUAN0x@`*@6G9
z&fWm~+YB{W7VaG1T|=oFu67Qhff7qR+^6)-FZ^e?xk=r2hM~|9{bPb-7Qmn>PgUPo
z#8ty3__oT+eO$U}kNiakg`5MHj}A>SR$upGRbGLAq-bk%>?*qg>55X87D{=x3;23b
zpuJ2_PQUYKN*sx>Vxg%dy>TTSm;A69mo4UfIOMgB1@+%I|C{^WKkVM%8>>?=5Aw#y
z>mLaEqhr+QWPML|e+0qqh~TFvB^RB8RYPLK1?<6Iu344*#$8^Y$fGk??KKDSZ}+Bm
zpMuXW5UkXLGm;esSbgBWi)9r*&oZ~<C!HagXsXUf^QFMM3$o<H9otx58l+PVG5Lo}
z|I5SwFEmOBRUA4gLo<I`TDBT{5HLYBS0w~#R)^hoVH~wbz^3l4wt9b7N_$5PAa(XH
zHjcT|1o4OgD_d<3)tt7vfG_WoD+VZQ<uktl;p<zC4AHr_$mMlRoALe3zClFvbkMoq
z*FCP}&XeithF7k5{MKvr0vQs<EGetmK2}mHLwObJHd^f7VgL!s2|IAx=8R5Jc2H^-
zg5TQeVk?amxrV8pB}dKCbC1s#Pzx%e5c6FDy@eW%bla;n1r&prid@upK31!hw6m#J
zkQ>jt@CiZ#O+!&0E@$4BzIEW1FmGk=gg6R4MuO{0R(|9*mQeDmb#Ot5@}X%=vHj;1
z6u|PkyVbnj#l=gA_3Bp+57Ctcvcd^riS>jyLY3k;6>Ab@@;xb%xgQ--$S6UecjNcf
zl1n`!$L(TPDEAjb#U^$UIX&spX`0IX<}9X?6lKr$I9ypj6wYJeF!>E(1(o%C@J-9T
zZu>#KwC<(>zb5w<qsCqRn2Dp)N&EQ|x(E);6Xp$ZpA|BfGB*LK%;oAWgj<DI9>2sB
zml@gDOlZb6RzjkE?Q3Qm!b;@Uot8~MO|lpd$DG@v-wk>N^p+|+vWt!V2_JPb0O{=!
zrD*epfo|@p<y;L%fM)}5ffeGOYPaG9Cf@EN#%3+feNWNUk)?LfPyfEBj}>;_QGkSW
z>KUy?19+w@2g6q<-%a<9kj2XixEr#4i}zySalJi1Ik3u|0R42Zw%Nm%pEv|f{TODy
zb(*RKvUh7kV8n~*Dun}HFko`nM@VHiZb_^A_UubyKw{y`BRw0t`+w!iJKq3XWm#XX
zze)P@KQ6B@KngV<M?j(OxQK`dM_i~xa5E?W?8%VprV>5(TH8^YYrd6MNn$==N`4{7
z6FBW&R?M_-<k5f*;=k(y6M2CCDsYE}8gkFW=_w&#%pwBfcJasa?$B#{WN^Jpjo(KT
zg)w8r+kj(&2^gRq)(`EU2QU1Zy^i{8x#0?aK$d!X;lm+2kkRda3UMheb8|Z<a#l&p
zX+nq0PVh3lWC<C4PIJpg85j;m)4YA$Dj2%Rw2Fu%?aTY0$j$d-C8<IvD{LJ5z093D
zNZ*B!mIK=Yxf4AZC~0=3*bE3#?F$GZK0kWHiyR3vSqYeVv>V%;)WbZ$F%QCt>NzNo
zE+huy0Lp26l6nD?RpE$0gRhH~e5{cRgr^||?RddS;kr%+B+5plx*wOf^W3`ft7%{F
z8TMVQ+S_(kJm8!7YO`E;siseV$0A{7etTeLcK9GAsnaOGscC>cM!RF&9-1?Thgei9
za6gAV+_T!}=B}tG?lEGh_YKu3JIxDJ+P1Icn(8|aQN=jpFzEN6CVLC*Tb-eX1r=yn
z0<yM;M7d5PoRVW}XNVp0jVSNOgg&6w9FMU#AXe=bX7I|f=H1A_(TSSx%j?Oy&ZNwu
z$a%n|V#rZK0oEGB{BVeK63nkn=H^pNcbUEm>j%y~2q{%W-CKKDa<!@A9~<Id{pUnk
z9I6Vf!p%NSyZhmg*&Vy+8!5vM-*Js^KD#j;&k3%L5C(@?_w9RXF6tkYm^R;bDq^Hf
zhUpFl<vcMRCm14)QvVJ7mHiX6)eZr+{&rU(%y$tHtI<hD7rc{06rmM+WkCi6Cx>EU
zDfde9NsqY9sTTeOO%XlW>{i~3zdOT9qxD5eF^xhRt}|Uv6^$99K4K6O()KWqOOJQW
zGrr$ms2_5lpvx(gqkM?MY+(012HuBuH9W9LIAOJr=|XyKdf+_q-lO2U$!XP+7o~Gi
z?sw}V9DoYkii2K*?*!hFKOebMRMv1Kn45}yM1*u-=rHVwdM~fki}t#Rh|MN;7M>I*
z-6*7GGh4Bm=@s1ax{8oO9hHH(>X6NnEs6+DC{yi7{7sX*)g4qF#o}2(VXJXF{%;qI
z-MRM`%;wS8pymKq_XZsf_Zkylg!vf=ag71A(HObgUp}q}f4F%YQQVO+D5F~5b>+rA
zTeqGF9z27SOwXCy8YJe02AH`tE1u0;6SKQR9Zi@0Bjbz<E>@Hi_ES)5lx&-GLE(1p
zTUAeG6Qc>?zpQ|crj`iy2<OoWncjP_eZ9mLK=!%bG!-xOzC2}+<kR(|@mi}-QYPxG
zx2pBjem7+j$ZHJzT~<PhGviY$XZ(OqK&4#opp$?~v^%)|nse-)wqBxFaR^rH9&&>&
zvYV?^95&d6d)J2G#DP_JLbh(tbq+4;`ERJ0X=-)F`UDbu<9z+zHWfx+hGGKqzk$LH
z^=`Zldgi&HiL~1$_MGZDk|dZzwe1h)HH4#{M0=m<xkKT9PH9|;x?Lw6+||p+*-53t
zRQyBR)%UYNfjUCnpp)uPC!B%95@bm$I!fGlZYj}hd{Ght89?7!SIYeiFUE+7CiBtD
z0d2xduThO`xN<yhzd6C;>SupXW&amDAC&L4g<OWfC|>u1#|liw30?^yZukr5_-F97
zwgEZU?<f#WM;@G$8z;FH(h_-8rv=KAo|juQiXF6(&TKRXoNyndRldVG_xS|^`Uy!f
z;3Tsb4gNm$ch=&t4ym)`6+gFqb5>S{X&C5UCX9CG7-!yDG^04@k#L#tFM0<qUEx2t
zBmWW$auQm-=<hegPelnhJ656%C~kwZ%wI0(Xud8>!W8V;_uR*Y-^j$(6fWFD#P5Y5
zDT!=wc5{w4_+1hVbceP5rJ3QyGeh=Kvw3gv?%kR?0l8JK@x8-e->eIOb>%=#qc;jd
zo|po@%(h8FlN_EA6kQ0SP#3zK$27KU`f+QD_Sy(%sfazIv{n45C^bmA|Cq?+zPbJg
zD080X{1`+vZFz0iTfO+nX^n4S+Iv28eZ9Fsm%ut}d;CsMuGa=%X%Fg8DC)uK2jAkD
zKl6qd%o_cT)u5AuE?;Q=v4ie22>Ik;ovV;SKQQz{U)a;xUQgE$sbSh?;~|jtW3t^&
zYSgZ=<yH(aqS#e6b(hqtK1`#*33au<`bqCyq>~L+T-D-9+iR?6-A_a_N7Hww>|rbu
zvJ)$vXZPU{l&mR6lA70wy6yCpi^>YI%Jk&Z#?CNix9zMIqm1m(WLwuRhL|9W)me`%
z>Iu#baTdQ-eD%P-#>@P6S3p@s2GmtS0q*+H<jBJpIb2YBzZd8B)LY!;%$(j31qj7%
z#lB$cK;g+2Ox61rfgQ@V10DOY1yU?FNy?&hZI=d4FIY5lv}X^t=}lom<l<msC6Jj8
z_++;N{+PdO(BPZTrjdKdIZ=RQD47C~fkCAiC%(NJIoJ1nx*yeR^zo@jsbAMQ01fWI
zG3KFEIbJXE!y)7SnKdC!0^T@YNGvmJoJvTnlKCuY7Q_v?%z!I=_a;6Z0tw>AyJ|8G
zwx{zIAa8E+9za@jcvQ}A#9O1QR+U6qRk}p4ON0WY3SS7fhRNiIT&z&isiXJnheHsK
zB@$YZ{0&SCT!qO=kP{=;&h|itCf`ZRwhr|C7Iq>zf-RY^6HFwZsein;q@r}^$hUv}
zPfyY3N%o^l_!wgR<Lf0v$XG=8T?LiZ?OW%*kcGOvcq2S}EBIOj;cT8T=G(U=zTlfp
z*Ye<BTn>S&=h+Wb1S#ljv>}hhiVGWLk2&$>>ASK~x%IS;-F1<)51CPnPutTTiBA(N
z%`5Evb%k>nv&+QNXXY%|(jkI$UxaINkMcz2dc4O(P73x9H+A-s+|~;>+ct;j_0;Qt
z3B<^b=o3>&{kNGU?UurvXc_4&Nei2IXB}S0aL<G*d|NQ~woV^@VYA=*_x&1Zzs8OE
z^{$n<o}SIapN(3EH7hMOmL#^R4$kIRj;&^Uwe+k<v9xr=b;e56w-f|I7WsOiC0;(u
z)ob<zV_Pl6g6|lWM3zu&46bQblHG{?xvu*UclR$3{~tZ*sL0)JB=09}Ec7&=sC=KB
zyayd&0NP%W_sOKkkw9p<IIAO(O-;fNPNXdleK^GaA4JKk&-DNP{w^G?nbn~TzNhRH
z48K}Lr}&(3BBs5$dm#CelX}X3f6=oJI;h=@in-iMibT0xEZH%tUXY*{xlga_Eq?ps
zgAlr^U6j;$%Vb``+f2UW!o>Tvg7OJ79Xo?_HV^TwkDYIXdw;H$w;1^ov4}OwCs|>@
z3LRruH?q#msh*n>rPg~KrX|Kzj8oq6O`q98Mi$#LO5<6ir!i{@GOl5TOBhR}7}l&f
z7cR_2Ig~^A+#J){f!nuMSXwivjPIA4El4KQ1|^h=ug7)(NsdjZip*b+Vc=AfXMy&~
zK6>hM;ys<2OEG)=(l2Kz`-PaM#K@i5=ut{~V&}xgy&>-1k(aJEl=lc0ol@L>G=N4N
z9=0r*Z*c8tYkU*#pRqvvtPU&o`#}PN$ru)!Br6#Nx?nJ$TWY`uD;0b=lv*OZib#Uo
zd(DPNt6!YQo+|Gukkwn<AD%Nu+gsdWHd_+ze^T81<Z&<mV!~19R*)W+oO8-xy^(D5
zB}>_ZS$LB7Wu74tF;V6I{i_6>lH0G0`{T}0*F`oQXpmkNxj)LwY7-GI+I!fFG+47M
zGv8ihvB#&VSo`j<*t7;`&&V+?ZCkHKaEiFM^hCdYf@G7&xNmS?+I)uwZEqSKSpv{?
zxuLLDEP(keD%#Up^Ktm*9~=9U-&(z$UES(<4{6dqD0ROK+w?bbT$K`kWZe<iJBSg$
zK9<;LTsZaN&~sSt{`6uGV_O{!#XkgI{}wz|4xDxNcceB-17CeOBveiTh*DEYbzqhQ
zanHUu#%L=Qj+HB+=ESmJY5TJa$?r|w72!XkfBru&j!{uj-fq#G@nKaVLy(sNCJG$@
z5WH<`tGY0u=Rx$U9=(CPmanT_>qfYa%PGsDn~q)fgsaTzdz$!DQ&0RKyuD{!Q)%8e
z?(96H<EU^W0wN`=2#nH(E`1jjL`qyyN<tBk8bYL(1emp<QlcO=lu-mk2oXXOAPI~@
zh!H}hC6GWO1cH<Rp@r`MID7Bz{d_+EH}}4H&I@uVfx~sKbFTCI{=UU(npv<$%TzoM
z)p98|g6IJEXfKSPw3!xuwTsIx0wUoap}50XI1Hi#0lom34FqcRatHre@7Utk{wkYD
z{jH&BJ#;$t;u9lKIi^teg&$|i&Af-(FZ9v3uXqD?==s<#)X2d7bn~<c0dA71*1cGk
z;y!x2@=@J(WU=hpp@ot4LEyID8k<UoZas^mq6-UGGQtO#X~P{@grYRs*G?3%6eMN$
zn=n}mcaq>a>#RB{cR#qEg;?*bU1wqHy8SntKDv+l=e;Osq)$*%ts`UncmR=UH@wD`
zeEDT~ZZw#EZgm{3Br_*{NpwEOK;?P5OxAtG_gs9?(q`UP0dfsj-kt3qsU4)Y_AQy7
zz27|NYBpWO^!{60E-cH#PFZ*5Q=u9E?ygeGvE?mS-^IG&x<2XA6ZKO|spP3xn*w7b
z3aYIHyVXfw-)3Rh8u$7?u}~gaNQ_FXH^VTLlv2SXKcAv~K=jq9gjc9Ksx7hmOSZNh
zk}`Q_Q+sF9hm-Lhq}9k9G<or(kIp}QD{l@QlDZ)-t$@ECIFlUm^}uG(k6#ab-l8N|
z3D?>EUk@C0u><1Gz;gwi*fZY$VUOa#0kr@z=UU4}vX623gn9z|EOWl~ifi&c=I9Ov
zUUzx4_8Qhvr6s)h$#Wg%BUVuAx>JD3P@Qlk;bzp^QL<{e;y$n($0npIuH!nk5A$0V
z_hx7GqU@dz7QOor7P?iOl-AiL&@>>tk8J}a+GI)sr>DO6vSeW{v|3GkEZbL{3NO8}
z6gD?zhnS4Lv&*K0yvti%7S%jim|gJDG89Jy7d1~Z!N#BmwYhnb-nDthmU2elCq3GX
ziA~S&HyG}>IK!@{(|HO=Ui-P}UEc@D)S&ctyrBYP_Y&`#mck!MlVQUzz@bA8i*>$1
zI>R-I!LiSO7X-$7AA6amZ2Pn2p{^OHn!vfyy6yx&WA%1*;egpj^we~l;B*0AF082G
z+|&ivo<Yr)@=zJoqy)*pX6z)8=_ECY|IHKm#fN@rI@CE0-&k3Wh%$aocWP+g_=z=P
zC=zJHPPhgId9X1-&9vIc7K2Q)E1OQi-v~knWPCn1BuqyUV#;fns^?m!P24UH8Xfl!
zz!!aNa*$`3*eDg%W3$6<H1&RtVfWVt0DfB7#pqV&Q^Qz<-rPX#$)9o8sMfRmH}Nu|
zFQt50_qXfiiI6*p=$$x{2rw|Y7Yh!|^VwF-De!9O&e^YafkH4Rsw0)prEH~?nwv7#
z_nn~ylL;%~OLP8$Of!pZ7x2L5$^dFmUzPBOpH}14D<;DGaaiqy?<9?;uHl0`>dRrE
z(no#G<?TB(CDsUypDd1~{H+S+nJTPZI?qZh8bCD6T`ySaODVL-Ha{KjC|g!O$DKL!
z7ATT;{dZNQ1`qa`R!M2nQ5&5Z^O&GH>c?NYrr3=1G)O$3<#VDzO&#ff41OV&-O`-Q
zM0uMhZBO=OJ{2a7@@6|~OG8opTM1h!HupSFfKPWLDGpaTw9v$h0^`WGLPF)Z6J~md
z5k@OPKm%CiKL7PDR&bgebMQfT(fJ)mjm2#=Xx{AEeUg=;xM_mmqR<?q8}y@R&kQ2c
za4jUOiFwk#s;hD3evoX59MVS?y5L*)l-{+T_I=4hX+zj*M%L#f>|)SUhI*j2a6q^m
z8OgWqbrcCJ>buekJ!6s1=uUX~#aK^AzYg-q*a!{#u%l+RX@+_Zgqsg@Oh22%O1c#t
z+_gal-AUpZM4ugWzfXw`Z;2WIDPMDSyVzip1Q|2r7j?_s*8wN$$@1!cM9mrx@Bfgd
zXO(6tn%t8UL%hUZr1E~ugV{uE=)8!#*O5PA26(39ue`IcyL3JD%M0^8@3p0z;)b+A
z`gAiGz5at)t0nNH^24S$wLTWqxwHYqx?o2^VIA*RKF&{wTa9y#PoRXdIrKY=lB*mM
znhA1%bSfcN_ma|o-`#z&`d)6U5bU|oxk(!7T#Z?G!&c>eJ>bY)QcK(Fcxc0gS^-b#
zm`$+w>jCt{^yk~_!ofFCp5Mi%J)jOcE$VXm=N=aTof3ka>v!zc6?n(Sh=rx=jdRhp
zvk$DYF&W7r&Q3EWw%36Bs^W{X+zi%UKC~hlp=+C5XD@4zzhi1^u-R*jIaK50Z>+Ft
zMyDm_U?TgywwaYmr<sxs#_kE)su6hNL~(-YUi6y4<|Ro)G)3*p1FZctf!!PUJx^=T
zP4Q$$=k)lZhD81g$;z`bn3me`S&Y$(wa$gY2#I6*WP%UVI^b}ciaR&z7mQ}tTfj`p
zdM(<$=j^KUSxG^1Re{-|wq3(r`;=s~7X8jz?2wF(^RbZ8WXt&RMQ}!^!s?7@)!M;k
z-{3^=rCtZukIXtnLLf{rKY%7}$MiD8yEptmL3-6^TC@{ofcV#R*(f6O?E>`c0Z-Mv
zw2I8u80O&#-4)lvuS)*w8!W3;5FV0Iigz}#o&eNv_XAEke*9(Yb7P*ocL`TZPacTf
z05;oi)KPi)wW5l8@W9QW?X+*VlmD^F=uhuQPU?5eD%)NW^*Ib(=y0Uj|3Vox*Bvic
zd><jnP5vTWNIU;3f29@Ka=pa)sM+~!k*5aazg`4{Mnl5GnI3E955{e-%&e!7?B{j`
z@CB8;8Cj)wVWEPNtmp{#jul8N4-Ez%<?1z3NvKmPL#uf~wz|<~Xsi6`&30L^_UWFb
zISJRRBe9WG&{DftRNP>RtQjxV+jR-*d}2^|c0bAzBDfr~6D+<Mtqz!nKxR3z!O5Wp
z_9<-kSaFXWb)aQi;dWLk`=UbC$c;keXyf{e+0_>%R3T7wWc6pCt19qM&hK#8XQ<CI
zJ>B!yJIJ$ZOx^rypK<r<ZW(0Lj-K2Sdu%2_#|HVBOSfHOed35MHzeu?^5GlwJ|+;l
zFxTgrR!0k-TA`P4o=)v;m)HievWf`j49Sf-X%#Lc;S$KdX26=WQwG>Jxmbv(OsmE2
znHJm+kZ{op?DEZgoC0s;eY)L^f*spG$xZ1ziO&6{z&i$0Zhzd~%N5*jJ;=1ZLbcvJ
zu<5?Gq^<+o4xmeAE~<L8TKHeXi^QX5TJ6?QZG{iL@%0I-^2SoPBhYbUQ&LIS<w&&Y
zyzFc1A+7-{M0lsSmviXYRY@b*4lLJrZ_gKSICOBOl-mJ*bRovloNH5kZXG$ExykY=
zQ@XDPU9Cs;gi$PU;VRe!B`SgS?%VHA8oPo3%MbUyu-#RMb5PDU`o<ijzh=0N?l%v4
z=2aiDzOR%$L3pX7en#Vv8WxvxMq^gt$>yW&OZI}O80!p;bj%0D;?iwksz<>?xgEu;
zqSUPwKYl1+cGJFj(|*>bK-;+JcfSF~t(Z|=eTVz$P90qW-ymGr?lOAT(<462vj%j*
z-Zth&2J1)1`S&u_ix1J7cehJ(3aa-Tz%wPhw#|gM{yQj(**vul<z1NF3Sia6yer9*
z@hOUNoXArU=SX%vUePLd1{S{_kV*P2>F)B`jb))5FV-eI`_(`<@B3ykAba;D-EUmY
zdT#1=rHTkE`1ZgwfA&RY%}?cNv0L5$_2l<N^nKjJj=$CAB0VeBl!Ah6LX1|d9p&7x
zJyWadXJTU2AiL)0a)<Zd$B6mio1R_C4+3Z@nfwdhXvSVeL6y!<sfv|x2iE^{193NU
zp2M2mm_!?|Kkr`-C1{U~gZ(I-t25t8It-@2Od6cKMHF8&&6}i@xlju7HFC&2fD7|I
z30TV45qm?$hvUhvW+&=J<w#ia{J{PE&xNn;ZhKbkvb66d4p#h)40al57adtqH86is
z%Ug$4S=6d{CBX%hJHVh_h3$y3)(ufHi1IV7F`jDFkww-qi3)4}R=YmpfpE)*Knv?6
z)D>#0<i^mnN8zQm%z6u$g*8&>*VcZ{SX|$m3VS!I{~<;ixrIi|xZUAKnPFG&P^BEx
zprxM!?|8kD2kUElvJIo_t`{hI{mug)v3AqHzlv<TQyZHIS<;9{rpD%hom2H@{j%fI
zlm_{K?+TpQUOEi0BVHk6qGD8v&uGUMP7^2d0lu!@hiM#Ze5}Vy3gu<WSX;K-flW37
zYGu2R8&RIHVv!jfPbQ?Bk=QM7g}91*EDqh<t%fq)9rK!>=}IP4Z|Na<cMFewvywUR
zr;)USwm_)yrINOaxSiPAZq-<X$Kb8c4eBHHs23lAIK~2@GQ+CdXzq#w{P=?_?&QJs
z`8K~YZgVrf)zWaVb_5!A629}-c5A!9AyMj~HE1z<)b?tKFx7r>H-yf?PBy#gwbyYr
zmbs_|-F(bv(U!IO>hD5~giG3Og_lxvi`HTXz#rwYJfs6@`>xhMy4;o3)=<#t)|~#m
zAG{@SA4T-~?XD08DDH`8#Qf9yn)grt2sn#azxWaL>Hg{U7eA&VN{$`{Y`u+bQF|ew
z^D``=v0cqaAG=Tg;}#C=o)B)-B}GtLWvVxhf4o1)^S}!(z;3s$B|9SkWX~>fAk6;N
z=@}DF+>-fro)fT$oetI=0`k-9X8V*zr2J?el<n_oW^^f-UeO>eH&ALrP17;ztnXM;
zB`%wfIwNhO8V%(aOE)rIkl~9@I*y}mQujhJuUyzLDlHO9A0XBX$A-`o7!MwPefLZS
zxckOV$f)99{(X=K>&ihsrqLY!vv}AeqZ8T$X85j7rTZwhCijF8K@v5dFtd5jKb~wK
z=Ij{WJ3Z^9Jn>VW^I81O``THb!sjL0{OFYY`K7vX9`tN*|BThltqz3(qSV1p$uUt|
z`TcF_rCt}oqk#be6>pO1!p<{&bQ@DCG5fro?lZaM9|iZ*a5K{p(n&D0pGUYji9?@H
zqr9Crx_&Rjx651M#=j5rf4}$t6nQ#}bhli*&LdmP5pUvg;3Fk-eX_pvDXkh4y@sX|
zx?W75vq7js{iijsU-w|tdym!{-5^uTWEJn4`)}*chKz(RWK69aZ6BG_KIXD1roNby
zo@Ok}Uw#=gF*My4>N-5YO1Z?WLkF23=|C$~pD{@F>Gv+KqFe+&T2R*EwfGiSRVqfc
zX809bgNXq5!xpRsxwJ1I;R!pz)tHydZBIYx;hn#h@{dtWy!YBqL#_SGVqm_&e=E`S
zAzPcW+1qh4@^6^%$+6v4)?wf&dC(|d`)wnIa^FVDVmt<=*t59*QebL-<hyzQ&O<NL
z|Lc@z=bWSSC;s&Oy{Yg7zLCZw@ouCC{$wSze5gJ$^42@!MH)ROA8@2doC*HabH1}K
zv2Sp&iPqxguCJF&A{(ex?74T(6jzu@=WZ8Q86e(|5@uvr=vN^RT2oxbKd;vnw>u45
z1vjwgJ^$<jz*>t$cfrE0PVpbG)^Lo;7Y^6tqz?zv=ZxAb2>mXbR1mSH$1i<~E>PS<
z?&$2@_)?KNLsglYovVS+EQ}q$FK^ogqx#>;n{9Y{aOhnLcKJDWRt72+7oTw%hG@@U
zoA-^<UvGasz}Z6N*Nzgz94QvaIh;2inKa^~9NS_px%EAV46Ze#gmmU~l;y=JS`M3y
zdm=f}Rte9V=67;x%bNin;~{*+XjZ~IoG9}A)=u4+N2q*sV$KK1jTQ9E$pI1JT;-4N
z<-31D7Z_ONs1(kRmCqbuNqh<q^-4yu*bmXx?KvzJM%aQoz2l%LN2F^78Lbo9jI0J$
z344=qY^50}3>te-4{MX90(?iQvk_ZRyd3;$Ux{9*fX!L2=#StAZTj1<`u_UIsXpjJ
zZObTx?byh27tBpM_xqeip#>>=wyCOXCF%avsWRykb7MCei@A1hqG)^L(?IT~iVUS8
z=v%)jOnHGxXLV+10`Cpg?)-Xyx@ZeT?ZRz`JLX>w?}J<{3%9{d^u5Papn3n-+xtr~
zc5A0}elmR3{+Dlm`+Is@8?lktLZ!9>mZ{<P*x~W<;fOdrEc~^5ITo1-C`SM2<1=io
z5+cC5&jF9$ni6l#tN=!a%(r~icW5mjFu`3>^+GKmzno@-Ra<*C4rpx|-KGGD77Z{s
z*6}JX%3uP<4jgLTRS!JiaI64y_P;qDe~V-Q+{=IzhK~o4RPm2%gLML(IaRx4^}&_l
z!Vbcg0orJTJr;2PgZpaUU}H=veB}c9>w$Q?iZ`6gZ+UMVu;+>1N9s+(c1De~zsQc(
z&R5h`#E5o?vZB8Z8t+mBs0FTf%*Oa0EC~?LyhbtM)MP5q&V9RZCx62ab7~~95kRy$
z?yS?_!~=*H4M4OQgX8{%l?4RhTfS|%@roZ$*ztJPcH9yUE-LXG0s8k_i=<4)c&owa
zr?eQfrpM4Gaj{uKoQN{qWIE?m5!c^G+=wMM<Dz1r1aFPQ=bfSPy`kZsS_>jt$dPc1
z7xq3p0L=>hV%N#Hf%jDohxcwQnA8U^rQ)YzeamBZP=31!u+Ov?zhCEv;N1>6qKPi<
z5co8=8P#*Se0&Z8U-V#PKAX~p<}~jk<mnEszOsl4L*+U}qAvq2XKa-mdl|k?_le$W
zJQeOi`uyz-ACS2?sc+S{BGr+g<>%PTZKKInPviP*IpXRPr?WWB7|}X`nptIYHP6Yd
zG><>^+S%qRp5a!iVY157eVY~FlIX{zuQinMkc9>XcO<@GY%H}fw@1#b6ru`J!ZXBt
z0$}GH@p+hGiVb8pZ(;${3$;aqi{?5E$yf+Am8Y7sQJwSC)+1>!kA*&Xp}6AJC4nXi
z;BEj|Cai6s_4%LwIL!ax!=XPVcw$ved1-$h-Mm;9uj1NqdGw3Dv3o`TMFX7ZXZJdN
z<M82_<r<s5Lm}^nur`qi9-Nu`cuo_gGC7ml@wTm{JbxR@PgPo^fYaBc80Px~b#3iO
zRMgbKRC5y~kJJ@=$~|IoDyeLHpqV?!Tq|y{-ZZ#}IN06|l7Ab&sEwB(Rqc%He6co9
z_T{%)?*u<8qT{iJT|AjS<>Jz0oo%r*O~PLz#<Xj-EA947&_4QWMjrQz8dD?P*5AKB
z#t08ERgH?e|FnE&RgZAJa_MInc|(7|yN22|RJ+`o2{FZ6?m_F$8H5ET{JME<llVn&
z!p0%-0`J!6)W1x`zuoU`vxPruXFqf*nauIepzgI=*aFHF{+LQWj@8dh30v`!gc93b
zU#rb2)%f7-r`r9(+6A%%WSYw~%qc6mP8}|V7d<NZJWlgBM!DOL5gzy4MO>i-WnF59
zzWsJaVq_I!NN1MQ&qSbbUF5XTOH(%5dXA>`@CzyxfY~JC_4{zG!jH)j*C>q?Vy$22
zxy<$#@mW1W{U}MXvoj@ioJoiXZHZfKoIB(#N=b~Jh(pSH&#bM4#f&*>W{T$FFn(aa
z9l+34IcWrtjgci%8lj4mc~DdxX>}nQ$nndbl5VDofu1zCr>di-z!<hL(d?;Z)#U;f
zibz%SX4Ni<SU^bz3j%DTH|qOjt&msJ{%qY?9k8}1Q=_)5nf#05$vu3mP@IM;pKTJI
z#n&K87(pvmsVKg=+grw%^Yz8<9{<h(=0p%ABF+r+h~a>}@crz}C}Np6m^{u$G<xE<
zvvbCB`dG)_CHP%!E0tBKvJmGvdWL%5r@tUE7v_RWCPr8cgg;vF<Z{!UO)94fCu=-E
zlQ^{=9Wm>cam4W9^HJWwI}5W7Bib3GscP4{5PF?U!DGU?dnZCV7mKnnNtu__W6X6|
z3JpW&#3%woWk!2pz0!^{VsQx0in1tvaXk!d{ge@DarD7hvk3Fib`J8kM)q95;Q-yH
z_woS!b*=A3|0eg&o^f3~zp5+f3GD)(@Ln>eo`uxldGFwqvDk-jIg!nvQ}Y%e%P;m8
z?}KxvWBXXz9pn!SP*&*<&wHbZRZ&mxWg)p5>+}ja(C?c+T>bLv0hzgVd>6EB&k-B!
zc-Kn)1EiW7Y@O%cZ{%%>Ojl){)%bC^X6skpml;#%@P@OiRUWILOgzI0Ja|f}`TStj
zb>|~<6`iubwKuy5?!Qb4X1C^m1BsalYQD~Q%3cyHY<u<@7od{ezAzd>srp!1?=~J`
znCGvlmjY>+-&-ZcKYn#*Y;lA5lS|S-Pu>zl9#m~BBq$T=7SXQ+>gR7ub}(ICF5X`c
zRM`%;XA*q_Orn;f3ASboAEy1s8q?_Dqt&68v%{%I_EMN|{z<V7-iL5sGfSGd-RB(l
z09wM2Ti?BzZBbM67~$8>Xz1(=n6f(bn_Bf=ou@)fr00u`-I5fwj8Wh^Ndq-aHq;#Z
z7+hgHX1Pnm{jfj-D^8J&xn~BnU+-R=UEs-7+z|@V48)0DTsKF{*UorNy6}Y?1BY>D
zw>TUINi-<>CGeFCt*T`yH)*HPWI_9Mbi!<Xj7)YTw<q~7&t#WvXZLpouiLg*2+QsE
zY$29Z{q(dXdzcbhK&|S^*INw#8}4eAg`wPN8|n5eYUVVXocMiDe8W4Ms8luM81qY{
zgpyNz1#OzeC3{JNu|^(2TB^^yX&8mtLIA~IZ~ic@wx{myDI--Q;ixr@UhQgSK0~Yw
z&GZrA3i1fio{iX&eoKhF)YU02Khp9)Z^?i9fpVP^djVN#9K3n&q4=%!_~>@bO+v@-
z7f*ZOKoZlgA<+ZNv1fTS&zjh$BWG)swQJ}tP@Z*Hb8xNat=@>0o81?6o`LU!@$=-}
zzd95qELkM0-<#FLx^Q=&&wEhwS_3m3$smuRF?u9q3|=P>P4Hb%UaL9FE{!TQh!FRc
z6|95UwmLN^zQN-L^`OQ}u80hQS-((EIBNy?eCXv_yL5FwR2eiN_0IHQXF$xC6(Qu{
z5!=rz3Nu<8D!Hq>8BI4!Cd-Yu`xHG>CKLpDb^hX!nuv{SFYXOXP6*FFD;bmzMX0o`
zbW1fnq#BUP72^A}=P9PMH7{*K54o?g-b9bXGG7I?^L-;@E1rwY8aX_x-5G;M3-c{F
z?=n_sCPGzc<*k^8-}5Va8}x8I10P#flQM?3`Vh(=Y3>ceip3oEt6mmkm|a7XY5#c@
zgn5!L?M1G*uDIfC;(V3v{%gSh#A-j3b|jxH_mpFm${i1^(*Hj+zv~~P?M?!5TQ%mU
z!Mb%RuOoLf^OJ*1tZ(MDRn0Xw*6C_asc`>R1MwytL6#u(HQPVQ=5SMju^&^kzbxrF
zZ)djN-A?TxhJ|E0UK|>s3$^6|sg*GZf2`hIf(a~GWiR`WJLMp>bfRN4`j?zS%i24_
z7TQ+*loW*8%JdK%`s3C)TUcXx!J`UBi&!1zgjV68<@c+wm4T~6VFRSftNaJ&Z~XkJ
zO7ql-pBZ}RO~M8Wj=NV~EzRxGw7fA@05CtD0(5iVAMXLG^n3m~{4E5k*dP+<f^rzS
z8yLQn!o(rDgJuNmANC;_m)_pqk3k-d|A_25b4H`<kIXLHXL53rXY6FR0A#o4%fE)~
zkDc#M_?c1zDw_Ff6d`sp8h@J<WkBbxR{#BKzH)<pWx4RYd9mtnSvKWqHmvkv>|e-Z
z>tm@@&>+lTq*s*-!P}pRaPj|D-~pSJ?Gn`NHi(8Imeok1+BM_3Qc6&#Y9fqr&-0>f
zH`K*lkpvq3W!a~?54s-=M8-3#pNZCq<qh$p3uW>}92!~_6+MJAu98!qM=e?!<e8rH
z+U9A&Rp9f&`qWS@$mZj{UQrU4iT>^1|3RYk{c64ZhLq}ynyBkJj9WjoQ%0Qqu)+k=
zXo7|1cDZm8teH~xqpKdl=5CMDvFmKg3RBmpc;~0I;>zpZK`Q%tdl1~q3R%CZ#8MSG
z95d*1)3OaDuwS_bzu}allEJ)-H?)9v&+y8F%q|o)&qb5XR^_~luzQyC9ZfmY?<D50
z1RlJ2o?tr-F=i5&K!k<wcN=q5VIF+D^3Ydl1+U<jN0tA5!2iMC|I4f^1@T693cP3m
z;o9iKtMc*<HgB&E#*=A%MgwIuCMx4NJfWpd{YEyniu%*D@_FBf9kom5g@p>Qc_VY&
zf<g0q=#!1LQJo!wy_kWu4vt>m16Jh3&^*6sR<f5a!-oaY(EhE<TO>2f`ytFiO;jL2
zKRy?0lYa2%4Mt&Tp(yFm9!|lCtBb(-x*OOG8^)e`-tODiSZF)lIG`8ap-k|(F;(4M
z560ZuIz=v|R_malo*`+(l1@8C#%ufk9$cBv@fX|68fI;2N0)qC@7*4(*VmsWGrvf0
z2D75O`%O-t&9^hL)DKij*&82*C2PR+Lmre>cLv!fPh|Fb-OM`Qb2~L?Gkz_W({8+y
zUQ5-^oH2_ID793(cT77+tm*=88{Wl?#MQ`ZX2@6<Yz<cC*`^r>X~LvbwI`Q^1a~&M
zCS=ZHKk<i;GecEaMP^NS*0u^LqoSgwP;i%BLMP355Lj1=x01u{ZW7yE#fHmZ?Ok+L
zqJeAe@5rk3=x}wHp`WJ{%GE!PGiP(nm92f=eu&~8)nS)kje9+m7vBXPEWj?h2h(`Y
zvw27TsC&=bU#uInuks(p=C7;VrLP)0-ViO-bS1*A4MqKE&!|w(Ul(ElmmgdT^zY+6
z;nm07bP&2*#a2UUf8LC8oQ_DKI!%~~C-EO5x_!%aG~1skGSI716DqqxPF|&`PR_ii
zF{k4}{+cTgAuz4xDMyhIr|XuuA!&SQuNJ3zCK{tp&L?6-QlfI!W#;sQp(f^!J$9`)
zlklN}i1^dGz$ItsJ(3~_nDgmg@)!Gn^f(rLbKMES#8#gb;>pz73VlPx(rV=G!bCB;
z57(i3xy~@6mmLTY3TC5bS1d<J7xt*B<1XvcfYBi!l*?Z)J=J9d8v$JhXM!JZgNN6b
z6nNiG{y|r@_Kd>For^$e4>kiB{Z~gZ`6=4#YoP;9Uk~gAed?IQo#;k{ZzrVDL3QQ-
zfCb3<vG0KQhRH4TI<h*^2KzD}V&d#}Muo4x*;w+#KIr`?Y6cT)<XYC(0}qha=_8m9
z+4Lbre*o{-*hl>Kw)4~x5%3GXr<2Lebgqv;Fb*B{8ilgS6e<Ph#vK@7^Gk;t^rND5
zTs2QAuAOtTfegQ40XO(*AOxwvJLqK0#5+u=nO-W&JHgC&ik@-s=!K<1v&d2SiEp%*
z%eXv!XYRSH-`K5zKmyPs6lAyxj^s91>A?IpnlHm3bikmgLJq?IZ)|`yxcqmS#U1<c
z3-h3l+lfcU=mb@K@DkSc$X|P|;V0`gcgtgd;_YV#!Rdg42j_%`_mwWv#W6R_%b+R?
z<;R6<qecc_K%<rO=sFMp4)A|oyH_2FG)C>ytMcsbOqpUQ2e+(Vce}!~EvSL8Rk9-I
zSj6SpMqF(NdQ~r&oY9c#-mp9FrzL1|yIMZSG>KtTqEzVn$lMKqx<3d_a>D?jCOKQb
z@a9i!FV;NydLWKG9Gjj!o#ukWO0t<rEZF-on4|=)a_xm4IN82D2xJd<B(-_GFwv~L
zI-k(9prBW`Jwta!S6Tn;j6wW$U&iL}W(o~U7;LMy^&|$;6`1@T&krK9hfAJ+lylgE
z;GmxVA@wW8*yUwQpI{^51}7o`Azu*4fv+1Ar8@o~?GI4@96)>bNVvzPpd0~wAY6&)
zOTM%p2aVNF##V=dn(WLp;OCxyn`GZM-P(oEapne0-`u*fW5Vy<zg*^=m)?l0D}6wE
zLywx#&9OXT=&@lm1{r#hd@<*j;`8K(4s#EfM<}*~@pQs#`^hro2WTK(e0lTnz?Zaw
z!G8VkpWzlbL%T@`Rv^1a81!Im$1@@(>h?&fcU|5=xBiY=H+fG*AB$h2BqzywF>Y2t
z?^Y3C5ByeUU|_&vg^#_n%iUCNj)1K1GTQ^!Bu|>9Kzz)X0EU&f9@Jxm2(L6$%=G7>
zy2(y2|M4IE2OqzGY|(S?LQRv8FYyk`G`t17lp}RPZ75S@Je=CYtI!QB?JVjG8>=zf
zMEOKLD?k{m?8ni5{NdV*<ltC`$>(|R@qqAfNPS`!5j8>PT%0ktl!ksL28CXHHrSjj
zF*Ie>lI}_(RmRdp%I)n^5DBecEh4%UnHl=4iPZg#JEJLI52)a~R<?_zab~kCs{stq
z{e2Tgj>Hh<-RKAK;DQJ-ffSE5LWniKL2_dmGmVgm`fE;Lw}snZ<{hRBuotsip%cL+
zLZxUZf2{vgtGcu04Y8q0Sw7bCSz|}dRo=1I5$MgH&$C=#-i)a@Jrs0rDXXDEV#ls@
z4m8>3yctt3GfF5gTdg*X3Vb!FumkcWB6(&@hbr{yq}ecked!PQf)bwx_=z6+3hKH1
zy7r*0^NS@E7sx}Rk_!6iD}DkqTKH<%fMDa2yW+Tx`~u~2G>`;XcoYyy-1|S5t6y?{
z{nwfKo@%2aIle&8*x1<um;vqKge)M_5I+CJEPS!a2hq;EcE2e(<W8;+l!<0uf1w*5
zv{A2silU+O`nb;9EnH5^e9OAz`(?%~xC`yGn>{CcL}7YudAi1L(b^?oSZtH1x69kR
z_yMYTW%Lu#8M%>^c2SX~j7RXWSl2hVxSif{i_(!4hb4YXD$a6C&wEU0X0v5t4g-OC
zcW8AcJzm@Hp~o6$UZq9)MY~FP_F$Q%<9**t6oqz7I~9R*-AWWWG+Z4#ffCvCiM&^S
z?5yV@?ak$ogtwz>WZvs;7kxoaa}_*RcddSfo#e7W@`Ay`_qEy;QXMw~#p$_7H)?BC
zEn|kM=1i4rk47(3_NixL9NHteqv>Z?+ctyx0M-bhWUJ8?Q319ENM096O?|eH=MEhf
zTv(z=q&p;yyZO77T_S2I%x!f|%0_uEohl@)mc6;IgD`Yy?yMMp*rb*9WZ_>!e&E2(
zPWqv4;tY;pre6_n5Kt)-+1gm4)-y(W+?v~*beFrG72xSMxOmPX-1f25BmcA+)bw!W
zu|ILCmWsb%VzH*(6@>=Bf`0Awt8&j=C$Xc9RzEdHy2GmCm}4HV5+mETtU0A3iGWs9
zJHd3ywmFR6S2hw7T_5MjJVTe{q`ggPjgm>Q#Cfwo`C=#f$CxCn{MLCAj^pmpFS7fY
zi=WXCt5`o4q61#{VG_dT<KtsfW22GCXurjkMFI@)^KBdCe2d*Hj~HImhp|gi2zBf_
z>pvuAK)ObWG0SAo^(6h4>V;_7Bn@{@(fZV;p|JDbLDhIVU|4uFD@hVa&jC1=jRSz5
zf&geK<h~UXr^Ns+&G~2W9?XwJpa4wEKadf?v(Q)+y4G<>=Ub}zKeiuoD90(vO={%6
zNh_;Smt!KUy6rx9!5pM)&yt;pl2*`s9m=4|0Vl8HoU?wBHfxhcVumVp6QJ7)gsjV6
zjEgqNOqu|&2(%@=q@*pq;^(hiFrz}J<84Yg#<<Aj<?Yz+#XzaMZZEP?KwSJxjggcj
zW;!YWT5>(|6{4Hf!SU_2i~;0E)(BbY);m;ctiLwd6CyDonOCT_6VUJb>46lN&>nSL
zNm3cg-B5Yy$<Bq{G&?zTBQ3B#Qc^?-&L`Tb<j{G4!@s4Ecwfuq_jPyoam%cb&N|SP
z8(*N%Yc+1!Kbg!9RJnO;7I?v`@OV5K-`EOF7C_Et>syA!mbMx%EG*10B3n1ijHd}J
zA{ItZoTu+oKy)eAgYnsfYWM5UtIxYDp7&JOjmHAY>A>gotCs*Z9}HN_36B8G|Ir`Y
z@~V&QZ~q+H|9%B^0bHQQ+~nYLr|B%blSxze6^1gZx*$Ag=5SS(#y|NvcIKSD$0fZG
zyeT0>i}<*xsHg(q=&V0}yxm1-YtPFv9syr^{1^>5SbIND>WBl>2bI19Q#LvZ85N)|
zYjkq*M{A8l`{%*kJynq;sX{}%Rm9T*li9K8{TPd5*I}0Vq13H2^h1TM1rUom?>F$&
zW!L!)QkT*Twb1f0es=l%@baxWn;jYFE-~1K1`DKT7WaT{=K~i?QLtFKlI8~4_L(V@
zMpxX65*;++MqBE3BAh=&dlF7t;_ss6lF(eCd*B8J=8^s*u>etwT`0!vz6t0-R7dpE
zKKfuk{A*Z_d_tP|f?6h{j*cO-+dt1Vj+4LBZKEsO?+eELgh~G2rE1dHfqVtc+(~}n
z8SzH$_)gn^Lm6Mzbr^4E8jA0EA5pTo_lDmgTugg$r#xewZ%<Y|R6S@-iY$vP7pFEz
zKg{=8TY9JBBa;#A3qFGx0_V$bS5#*VDjToP3>4MVH&bUfgcsoL!{mqIt;|i1kh(5G
zq1=wpum$ZXNMi@VKk+eIsk0VakZ1C-`TOIw$NzQk|H;nLjJ-FiWh!3}sOEg;^>2wf
zC)xzrIFQ{12g>-ZiLqCxj-a{l+qF(6w&~}lO60aCPX0xw5BNhz#&%?g`7_gKxg%S*
zSBXSnUmN#kEx_NUE;4%Ju~XZE@xixkXimmHV%~A%`-{u}$-({K@A+=a14y3FhNi65
zdRF>9UxR(Zo+X5_D?Pkl{Sk~;a`pKeCzAm%p!M*c&Kh9;_}=i|Tw%M;pq3R+-6a7l
znF8ch-{SB53hHTf5!^Gwz--3L85;hwMF{)@eEXJ$2h8nPw3lgl^74~`>7cFG?<QA2
zn|?qUHjB$LXA{cur_^Qo{QVYG{AK1^=33PTAI|duw~AeE`@wEVA<zHR-&*Rf6UuZ=
ztGW?RbjOL61arv7d0|lHTf{psSCft8$aSU!%y(GSu&I9$t}jPZ?2<%gwRP_jOOwv6
z5QK$O33|RjAcjCq$3@C#W?rn|?vdPsWVKt-nG}n4$Ntns?a<KjIY_a2OC+Z~wOj|P
zI>noWC$FZHZ@qHeNZQ0e!cW#&G)wlQu0jv$L|zWiDlqCc0!k^9#w>t!r}1OYH^yC-
z7H}#ZzxmK_?F%ZH1xep7U_by3T*$z-D$=q$S+r_sxK^|W0@o*|{|hZqdpUn7j{Mg_
zc(oP6XDb~hta~Co5oLVgh;yw`io@V3+u{J{hymjhON2<E>@wQ5Q7L_>-(df~bd{A&
z#p9G=PGMBYn#O>?PfPlT_L{YF5Pp*WaC6N<bp&3^lBj<@V7xT3wWl*5vx^Yttr}!N
z5$Z(RJVgANg&udl;hRX@(67Zy%May`ej;EMd+ruwm0;AZ9ka`<8&d(!&6N#WQy^wQ
zE6umo?k3*SRctQt6R$YrkB=XWIx{kd%(Q4t&#qtLflm3c(a-aI7a)<LjK#Lp3-q30
z_ajgwp%#)nf3xagbKXZ6lWpmSn^fleC5rSn-yAEPl7DfDB{-Z_MN6QfqDB1~Su=q^
zTf!UxdXQMli5bX$a(!oXhWScxZS$cp(bl8l&Oo_AAj)&42IN9kO%<=96H>tz<^Y3}
zVodd}FgAQx@t|~*m4AU#yteEAXNlI%>lL?WB{lQiaEM2`a^r-)P`8Qckwx09!<3v@
z6br4JmvmvSS64og-aHup$+W7Z)X<8%ogVR}R3v~brY4rmQT88D$tXLUZgDhn6KDt)
zYeNx=x&`kFd1mh#bM)@n+A04s#mEU9Je{=fjGiV$FGVBUC=biSR?(qgTkB2uW-_ir
z4J((jwYv?a&DvoiXfFe#bJ2X@QoPL>`#Cv5(EN6<p*uFtZefa9*6R2Y(0-&}30>dz
z|D?UIUfb1(;dZ=9qwn!Wc0jMZwc*|$Me*T<k3g@y&_Z_0CU0sao%cr448#dWZ{Qr!
zPmak>tl_^E#Sg7qfTFlU7Ld`uS44bXP|Zr0l0bGF^%uHhwhtVTJ-)HQQVC9c#_;6|
zB^!#vO9r^5sB@6fQE%S$5(0C0d^oxG6Nw1uLDpexto8G7cfj-w$mazr<N5g+9xKDs
zQ^vQfth8krHa|bv9=WwTO$nl9$mm9Y>!4HX07VGY4@k=bh7F`dH3VX`AkFlv%q1jO
z=c-4eh)uv6LfDP+;XruqqvGPffD@@N_+{%A)2UYX=I8`#T&P6xX=P<_xL36P@%@U@
z^5h1$Z0cTh!ikV9LKX2-H19HjL&gRtV&40tVYD1mzuPfRMXW;?eP+^MrPvMPcPae+
zcQvUwGaq0)8YL{=4uh>gOS>R>n8rwnMKrRuG2>!7NCaU>WqmDi{#$LO*82<#%Vb=(
zb)%c{u}M_g_y;7yhi?apDzJIN>lzq+3)IHDi-k{EaaNqfRP;38t@T|r7Wrtnc_Qb+
zCe<+u8}97R=M6F@!~Djkq^CAk-$R}okh0q8i7<#U6I~nR5C-kR_}kmceyfc;arl9@
za{Y^hAld?pzvIXdkwakI@?PTb7~`IL%^|cNilf!xKQOr*QbRNgZ%iu9Q`|w++K^fy
zYtt4g0YWjtWHQc%OGypX(rV08Kty6Z>ggqdSS_RnYPa<j_<nTNPyfW?5AEpUT}_&L
zbVGnh3*bfLn^oU-z>oI+<%U^lVfG;Fp!JG8Lpft(IXQV_V?#MPIZ;SR+A}$jcky=G
zGn;Z%{-07y#QY%WluSH81&f`}cE;#t?+1m@=?Whyf4`6UI0tjPFle-a_Z9vej7{w-
z!0?4R&73gE?XXbO-KcfVs_rwWI23sA<^J;&ovM%VIIww(z<uvpweEJ?wvR_|RNOwP
zL;X&7RDa)WNS0Ny$ttQRWPVJY9R=2c`;@4b3;aq)NP%l>2A9dQVLdo<?*!}jW%W6~
zE#kSAR=oxPwvw=j;Cg}TZ5&aDOv-6F)2d;16xrSrh*@679g(2@E|6R%(o}|`+sswM
zPCiNU4fUb*TsMKk_r$<6dJnnoNj@&7wY1$~a><~KIkl+2GTO^r-|*K-OngcKGP4^H
zD1hNG#2@dIwo91s1ew86b9hHj3?}wUdu02^fX;g|syC~T*{JD7dd0Qu9dq%D+uT!2
zP1UnL8UHqr=7jB%SSS9_Jr`~5(O_e1=X@f7!=cU4U?M;z#V&4lX+^+*q<c+lp&3l@
z&8txyKCO9X?JxiLvi-mK>^u5y4Gopdq#y|rux`W!_&hoh<=QPDI*ez3YNOv?&;eHV
z)s>Q~8-)dhC%y=Rj1o+-O<tw>gle#IVrY6eBKqp9IkgUaoPJNIRHlrHh+pjaJt{~r
znVe>2@R18NUqtv#TBGP4Mh%A})i%1ckxN=*Bj6KEEJj~vEm`0EVqyQNFxkcj+GbT$
zBhyVx*#R^nd+B$Fsc<<|^0|m9S}tG(Jw%0tu^ed~@0OnK+HCBUI|#Rot-c<(VMgg_
zVJAYG_YF+x$|~D~4lIYMT5PS;)%-A~i5`JTgqUqoOT5E*!+D^xtPw6oe^+TYD!$^e
zpOTW-Lf#AZUeH8yrMAxRfZc<(WAohnfQDVbTw}oobyiNk@cBf@mf2=4p=p3;9p^B7
zi&MK_fLMdqA-;6O7xDN44gaX1aSWP~ygP1d4^uyb>gD*uXweIfPkP<$N@EY#E;j#<
zb0f-WPdHML(%Ms!ZhohngRF^>$$MHLRenayiWn6@hVGx6y5qfzTD9jHH_Yf(x`RqO
z`Svi_j^&_x$?=;H&NqLtyG}KoZEuKcNBow^2OY9tw~Apf@!T{zDit+Uk$}^nN3!u4
z$LGRTT0)%pTbB?H>k7rS&ivK(7|4RL-skRFG--ybqt@mL(3)ukYUpg<!HDFI>31dV
z;TVNo_FC~{iLI8&i)EJNxG-I={7-}_Y}A8l93!R5*(SX>savT-KCw!C6FyUGzXLke
z_+WUkxOtG&`XM4YrK#Jr_S%PlhB-&$yFJ${>kiREJGH|%Z>@vOZ%!jH1#_je*cD$T
zA_|?KQ8qCPoKzg)1Ue7bCsAT#dKM2<br?z6{7$d_DU+q-y;&(zxHEIu@1ma$_5;lI
zPRv+h$2i=zxSojs+m)rd7Gawrri->|F@epyV-|sDeNX0p{@1+!(M!@G`Gxq!z%Qs?
zvgjdL?Tv-!c>X|RuL-YCwrAg~IZGsibf)w|@2@x?JxV(66=9-3RZ;+sY9->9TXj9Y
zzeTFC+W+46#LBK?O7WH_K76|3YKlyEg$A#QH;x7cOzDL)X>sDn;SFx+vr>0F^LD3U
zs+*FcPr8+7>@#5ucOa#8?pO-=2Wje2!rs%BHO9T1`Xoa0au+go#k0P*3sn*4?qMJ@
zf&&#E11*!3HEon^T@KTJ+8bh|G#c3r*<RAl|2VZcFPLfq$TjQwxF+Qz^KbH_a0WR2
zH^Z%Da%<gnDlH`+u$-}R5|vo}C1P$aEMh+F%RFrM%k2DTAair*9opa1-P2#~&ip(K
zHi0#~#d-JSIk5cxqi#J18EBOIKIS$9qq#CoU77Ad-UjkEv8ZArzd;yZ+Wf6U83Gi$
zaM-u*Bt<Fs-vz~=(sq?lYIy1tFVaF#W|(X(dws39H{d6EhZ&XJ8JkRn=pTI|ET?;)
zRmQ@rgblS0ykws4oO_*(lTJj+cvmZ<Y0DQ3R6uGKD!b{O(qoOYf*Nv4DY|}a5zRS^
z!eafwEmWWd-`e1Q?Bymtmgj{~26REk%86Cs5wX_Q7N4@!Y(qA-ooD@`gm^btF+ES#
zYltc2ZH=tNT1cR)ggX!|17GU|bd-pxr4^zL^%K~MriVY7`*W5d<wuUy&7Va1k+JhF
z<fB^OfA^q~*d4Ik66uT8(T&nk$ide*40M9f_!kcBzCL!o4ic7fWA=KDriEs~DQ=8U
z;pyHLm9Y*zDv7aDTF$4R{;&ewQBIJOq$DGOEn6gFBmK>~`}$DD#-YptDPa5~MsnfC
z=g9XCs`*?XtV@F4;qIyZ^7-Av{BO`r{F5L$TJ4OEOI$tB66o?Ee8Ucq8BV?bn$G{+
zIYXkcI>`!?D(oYBj)&d4J4-h|He8JVJpOReB<yb$0ol)+@u&1XHgJWVFw9;)foQox
zi5RTR?(>>Q{?MxrU(Szgn3ldnbg^eT&Fdgz?A?XP?Nsh;-S6A?x+kPl)Fc+gb&V`j
z@Kf>@r3T6>)Mp=!cJ_y%N2Zg;j5GE3j3<cv*KsbX^9=$|Sv<&?&Xit2NhJ!U`z+<S
zCVqPCy_Eb0D_*aI^$>@Jh`+<sD+vJz3Tw<vtJmVB$O3Y6#8zoJx$(2HA|W!S#YLpP
zLbTF(%oz1?)}&Qm1f?SU1{S%^xx<yw0QqAY07VWC`*@z~UZsSpG4JefDHU=(U*=*^
z{+d-Km{Z7Ym6m96_+rFDASQ^3!{Z9oBQ3(iu@XM9z85?Q;+@)Bj};A%(((>=w?jVz
zifcGvLKR3E)Y5%U%Jf*Hh7jw!Tp(JSv+DQ$<qdmP_Q9bHkI_lBQERtKKVVT%dq+u*
zeauZ3M3Wvsj(bORD4DQJLew|FF1b-E>&{jb2gTzoQ=i}ms%+LNrMO=dA{Wotb=!5#
z99w42hAf$)&4`@WHW<Yj*;&D%-0<Adt@J!OOxZZ6!r0otrvYU}9t?W|kMzIzIIAco
z<_0roayz`=jLf-3t21r41v!imlgnCG<y;ZzqPf712^A<F8rV4*5<OtFfQ0+VL&MQH
z*rtv1clU<lvs@n@g0k;yGSmIvO=4?DCznTDO!JQ>08_fTQujGLGMcT(-bhGTt12iv
zl43H;&^=r5UlGNO6`u@aMvN+5uND7ch0?Gg&}0B|3q%D1`BcElgT=<eOhpJxP-u|8
z%V-lnWXp)pX=YXjKaD*3%iQNqEbd8B{I^McPsWlKJ=yu>8Cp>{9$f|4IRi)-8Zfa!
zzgtw({->b*pOg82_iBig%uPqAvV`)J$J!uc0~8W{<A_anT*J>`y{o2YC#>C+9-clG
ze>UrOt($+Z>`O>)w-M`b>b0x@(1rV#e+8O+f3*w1L$HqDP2Oy|SqaKu1aDVhy?q=&
z=I!sj6Ew~}uRJRtEp8mWw0?RAROPtw*R*T@yCeMn|7kh+7{I_z%a$Uhv+p5=N78Jj
zv+ARwLPdTW1wAg~IHN@4Zh}j}Fx<Yc9oIhYbMt{k)Ka3ZOpxhvZZ}7*xI5*Ahj;uJ
z&u(d#>~HDO!W1c-F`5!p@sXyc(&m<y_-kkFlMek7imgwvWMw>7(bBa>M&`(~{!d$j
z$_a4^yGX6E1lHB<M8sCB^upYimE=@*IW+4@L=;rB9iJD{=nPxrhIT=Al8i(fbdtJH
zy3>=#iJ_xQiy5B#p1hXaZnr_nWpqdDLuj_d(`*;>h(5Sl!Feljkt!c0OszPtx<W5^
zjMcu+w^%Z$4Xo`pdBwE*ts>n2g&%g0^<WXHxR%ixiK5uWh*ese)@{ZyN%fqpMN=BK
z<}R_K-u><TokS`37nmSMvefJ^=4T}E$*T!TZE1`PA6<B&us}lc(Hm&v89S9lM!}%m
zcH9z4{R}*b2D&8U%k}m0^%)UBJ@Ej_0U227Tg-Mw);ZUu<o&IM*h<3c=7)K{@Nk;-
zsPu4(xL&I}0M^AjpS*a+D>h)XGc(9v-+I$n2uPQg`Nelcoq%_@qK|hpcZiXxsRtcc
zYi6$(5kg>wqFxBG3hT@sqUuF(iI1^$DXdJ7;R<=EM<B3iX<572{To16%a|{WB&!US
zz4x7I7$dhtCnO|)am}r|`@z0}`gVYr6;c^WnDaiL$Zh8YxrJAkl$CcrV(|KFPYzdV
zo*!UbhX&7ORQeTev)+UK*R86S%SdQ!j01M-Sa^+gx4-Tsf4gJ7b3FgTi48qDxyap+
z#Rg}#O|f2z3p<s62p>Bh6k&cr-yN32GzJ;9AZ89G;U;{`e4eGkOVZwi(sCQ^ydD)j
zhDcxW2fHiEw=}LSfz@1JO;t-3V`CA=g)7S0Zp{LBu#mWap%(J>K*+luJqrzp^mL7%
z_$h^qRgPTOy<A#;J)nov(G-diw(qcjXYr(&^&nAb&b_gX_~zC(j_iKi`-CTh&?i$`
za74(lsMtftF5DUHOAAf3p>Un!Wgn&b2z0=$VX1>>H*wQB#_BF@_-&gVh`fc_FF;C>
zK&xsam0i(0O!L}?=sdC@J!<`Wz?r^caCpU`?4zX{_#+gKzS=huYPhy&19&k7q%X(6
z@t_4`1FV|_0V~fVS6a3kvybpB6{^aw1lp0~)aPeAZ18NWC65_Pv`wwCMN%I!XGFN6
zhaQjMg!<16v`9@>T|?rf1cM52NV(B1R&JhRu`+FC*WbJro&HELG!?g!He*{I5%LO*
zHg?|83RQ<}J!;o|=JLFvYTg?ybN#x(r31?5(lhG56tRmPtVvZUyJKM$7UlXSk<$-$
zoLS$7&VD_x=RWDxCr%#*2k$Sff@akUp&X~rz+J=SyN%nGGZW0Wo%`V6EhEt0m*^}2
zs)^sg(b13h&C>m=e6OePn({REyT&$C_J0G+4clZKcR6UU)GXTCcN2msT|*D^VwWm$
zxw%RX0hjvpga24R{_WeQ<|<QUT~81B@hp!A;a$s38D7*Gs=1o|&9Ng{6L7C$VzG^_
zjrZKa2jM{5V`P1uKxp*0?e6Y|bz6Jg37|o5_h9s$J&GhkEbuU>17TL5mLLZe4D$45
z0k2pt;9+?A+*J=PHbETzg#F*1g>uHWrw^#{Hy&;Oo!ybY#y^yWbpOas+aDaR4n8Mr
zxpyd4yYwl>fDuh<uWZ~YuFj<taAmB5-cs-9*Eb~SYO-c?Z)bc0;dR)oo3_(^DUU(D
z=`}N3$sH?;WkfRRG!CICsTu~nrQHk!a*+u=BO6K7EcuAYK%u!L4PIY(XPr|Q4`|41
z-)}BZ@==^Ur-E~|iH?aGgVSPaY*(tDLVtCNKj53N315Gbo_LXF*E&8=q2NWf*T`KG
z8nhxcsM9C;2-Z1s;Z<%u0XyCP9_Tsyi#cZbr<e3%pxvIqA4Z2tlDx*a*Gf(VGizSh
z{b>6DWpnFYMdiA#74>Bi6{9wfchiRvX9%?>J@C}s7Q^{FEqKpWqq+4xTOCi+K<yyK
z$Y&OMisaB*HGHez+H|r?;3rISjeO<eu&`Fn7N*n*<!KGq*+HhpdKz=_Goumo0S&D7
zjX(eQDr47X5-~PhY5(9<{ye8v^OQ*#!xTbzGhFfq3Gr!3ME-dI$#FUhShB1-cn234
zX8n1SI{T!Mk-G*wYw7`9#IP#EfoEofC76o`0l?qb#g70#mZC@7X<e?>fR5!G;)edi
zd-BnCFCbvhkbFH*61RQv^LOL=?A>R?6_4|N4U!+EEp5AlF$rfDqyEASir8{pig!Ez
zATl8#X{g53k|@Pwi;6Sb2z7cQqt$a~ZF)DQ>pSApYa0E6_GPMPOe%Dd3!u+RQw=rb
zhnDPz!>BIzqD6IPxovw7t+w#cp?hUxodtP-3ZaVPM?$|&(c~ckOkv6}jh;l_B=Id&
zJ$Du%F3GFdZoppLHmHk6#MR~uS0q?X?P`7A(_a4jter#c<Z9a?HFrU4#Cu4*wy6AO
z0UgHITY91WTzZP2ydYi|%Xm4Y480y2y0b|kIG?XLoZlN<sS5+&NBixY*`#dVB?>^J
zRyJ)jE~A+1RQZ}dx-4g9o=M>-8Klhg{D17dc|2R|-v8ZwdY|3hiQTQJnog^&t%{oG
zxK9UmP1&V21!)CAL~162y-x>i4Qb6{I<2Bgq=qD>uv;ZnA|)gQQHhu$h?o=Ezt!`*
z&;8uL*K?ld-1q(b{(FAw4_R1}<yzOe)^&ZqpYQwAx2029Z_VjoZgEPZ8S|lwp}S8#
z+&AT&&{)@u#O7vLQcm@gGD!B*@D&dkh{SyI>dvq<_-t$>tly+O!nZY5CD-JJ@BgQ{
z29P>k*h@VhrmFGl$J0#J$YGN7`pQB0=$7tLOqH@_+Q*x?hjVsXu-NmC<|lRCaiDuR
z=&hXhj?Ht~b=hD8!f(;i>DzNVoqfz*zZLni8S{HlzF>pi@Y#`R-Zqh-ayfiyIe-`i
z>vM7dreCAQZlW-s(K6Cp8R1NALj{G9h5{oL*On3CuzCRT?r#Htazk+JCaFI_kVz<f
z0^j8w|I*bG1F_4Qmpj%tutIU72c0n)4c^J9?|C<VDQ`fBEwIj`<~fx-dJ!G%Q<`7h
z;cIW~;sz81ht*y{hkLjq#+AQdc}@3s9`J~^cAJOHum=U^>NQ+lO<i#M=0Q)-lN088
z?|5QgpGnT8Tci{?fh*z?`##;<AZx0kOKi?=x(_4ki;J}lx*2BkUi2J+@k(ELT3<Ft
zmq1402{j03ch9J4sIgyiQO2}^{=gp_EgJv&Z}%GTvMQfe9Z#11B@X5^uLmOT%gE<x
z>Zy6n=?H8z;-ebkdD{!KcfaCYU#&x~>LT<mCgOc$T_r8F9QBO$+9RZ&Da~!zX$6g6
zHtCO79TQi<yB%n@6yE^NUYxqGZ<i85W)mX=vy7it)<S|LWucUf#AHmWU#Bhtb=<aC
zQwwylyCQCXC1)(wGSi!~$sM|%!@x^DldyeZQmKN%ZkJKRiDDn|2DJgJUQN5ON%|`W
z+J^%hM>2ChNs(pE;~V-{nUSgy)n$!15KrqNZG4#0RBLMa;s}#`YjJd2nW=2#)<K<i
zYiXcVTv_1|bxd`H6!-E4N_pAhRCN~Z!<K&2r<Juyx_W~qDQtBC-g#^z-2oTejc}hs
z0DWUETrjQFOIg3$pCx_{vVO+kH=FF*9vdASo@VwC@i@SzG0pTW5M0`04k>Uqny=S{
z5DiX*4LERg=41Ze%xJxepx;8#%<dMXx0AC*L(P}<34JQAt7awq-j3SwhrR9wxs+PI
zW^5>y%AWL_{93@6sWP{YEy(%IO{jM$<tGd(JRYHajGCmMsqBeFpWfVFj##(J8nkuM
zb3VD3)D*rv*OR7*k;Sfh?OB+`T@FLcPmK3DTkW+(0ltojB4@A5ZsSJ$4%t(NIuZe1
zc%1K(a>ka*ZXaiJoR{U^nNC&7>phdwBg@FK?yTc>>BK#j(q76RyaC+rtyOJeE!6t0
z3+fbPN@!WPawQ{owmrY3I8JJUR_~VDqOC`5&3K43^E1`>T`(!C1;JaNj+m`UH!%Om
zX7~hl9^qGktTXF>Dk)4zma~JLV~+(hsoH2^k}bV)U5%3(dyn5WwX>fDE^iZ!T^Jn&
z7J&f<rbV>!iCxk-4JSIGBvStoN<91b_B`;1FzIQjP88^@O~v!c#BaAS%H9cB)6B{~
z<x|p!*2n_%Age;(u{+Cj;pbY>*c(e#%b?bAvxGiJiw;}xqwdjS^JHXX*01GiU8mZ*
zj+Hui-WPc3enb<dOw0(Q4dDWafg)>XIZ@ja)*Ka`-4#FCNlGm7^;B5t;%wK7<7-Zz
z-i<{1(H(6~`#8I5#YwCZKpjHMG6yQf;>?YGw9cDcz6S0(8Cabj9`5YW-vgR-w<JJd
zL)Oj=Q;^jNktTnQ6?r<p==I$?Qz6XzH?PmVGj9`|{W|gsrNLIcZ0{c@*z5^LcB$pL
zg8B6fhJEN9E<|u<P~1U>&n3F{j3wG~5T$^)H4y~KpN%s@qM}qKiO6C&0s*_k-i;3G
zO9K2cW#x*1(dRY6$O!ZEC~J85T{;jm1JpHpJE5mUce}G1d|*_^4__lQ*Bln4@Bx{f
z>$=~we{bY}-%%_U^{xfgvaK(bhogyjNJ>J92FRzRh&tA%;{&F#W%`3B7lPI{Dj(!t
zj4G2GerWAu5@vs6+gI<%sn&~B+w|^<phn_WDkjErST7T7zwgJ70v?VUSA!9w*7{Qf
zoYkT8<l?CuFj{P!U3&EyEqlB%kkHm4x49nSYQZCIH!f1?UKH&zs&i>^J%Y8qbvl)Q
zTrE;j3Iv-lp8BWuuC03^PI%P~xN=TqL>iZCT;#uu-G8`cv?o@{X#!ZVB38VT%CY2q
z3c^WnRN%;%m>TE!_}*}(;HU8D*dKi%tm^Y&im%E_Z=5glu6*=<PjXs<-Z9%Wo?(x&
z-(g{%2#CH7TXK50fu*muV;n$im;)?%YJhcB91g0+9BBrG!KeKTs7bs&Cm4YKtz1OY
zP)@_a0P8yhSW`7S1I8PG$OUu(RHjwB{y!V`AUXNsqJu?zH|}kT@%zFfWYagJq>ib0
zh#;waAno*0j_;Msxlz(==nWOXB%x2w+py(RG*#zqU{Ljw4?nlG*YeNjZqBpne_a!O
z-j=X1v!PFOtHRaf4-#^e`JJN!EX}S%C%O^-qb03TJ^%x@_F7nETT?P^vT)6PU{u}d
zwOE~A(EUMl>H@<UA{25+5`re)Oq5(&_Z;xQ!dBVD?vrh~hFm<}nCIA;z0sbA*8FX6
z{*U)oKndeswDnFNIq=7!g1}i9_6nQ*2o^aeaLWkVsr~5lnMmc;fKBt|kB4Hry{^tI
z^UJDg{X5aZDcPC*Ykb?QtE#3Mk`vqd3J;l-`E-HWx5rweLe@(B47e$-^2Is&gQEhK
z-Q&v>))Shfc^vCc>Q`$b{R|$y>yGrA9|upwkq9TYGxys)8VV2ru*H=wk}g)wnIU(P
zp=f8;2`4-0zxnsR``_95h9zUgDc@W|`B?IrRHq#8jV2gBjZbmzn*e*93=66ny=UHP
zlhOsZfysM-13%qt{4F4)f8KTwd_ENAdZNa|>);VeKa##GNg1UGJvWVt?FEf+v#A_y
z$nOWNpVOBUo$7i6WP}pfUeKKVI+vSjL0x*|Flcy)_6rcr`Pl$=@Tr}lOvR?Xj}_M`
zI5S||I8{6Dhqg6r8#Mvh@sLmzgEyDOHAj;x#3+L|CpMf^c<0U>=GUBhNCDp9f5Y}K
z(c5<KYbz-W#Lmt+IlZRtsOvo)q26M1wO$sOf7G#Yc>C7HSz|M1{3&5)X5I5lYsB&S
z6HQ(|eI@`u3x)8xjK5U2q02kON=+L8Tz9m|^PHuf#_Qsk2)eknvl5q+n&OBE;nT5*
z2c3|VIKQl}_MY(IX3u#(?I)bep$S{FCBILt2A?8FySqXj6ZBpPM`sa+8oK5*ABe^%
zLV?>n+0)=iEe}H2dA)|mC6|JxHyQloZo;O{R&Bb_c7H^Vcqx3WEa3fE-%`y&U*v~v
z^K#tdp$8+v8jsIWMN*P|9Wi1lYz&crK1YKBVI9^>|LXnz2b*U}8>!^r-Gr2QW1?!Q
zR%!r8cV$$l*mHl~**{<$g#~zAZ1f23F2R6OAg<B}7-3he!3V{0EtdP)k)zv?iQ25l
zsFqMouMC_5Vt<}G++gKxJlkhzRvLbL?|Mp9OI_8`^Uu12MdTe!@5L2L7FaQ_xrI90
zK;FuOSHuRT!s1;sM@y&hq1qLZ$VkW1Fw^s2ha~`Z_2@?PwyEWDKcw(D9w}5Gu-Gy+
zj}ddnN7*czS6@1)v-c=eE?gzl5R(|m8C&(n*M>fM;(e@e{caziMV<!6Dw0j%y4y$}
zz%Ed#TntHf8!ap(B7uR43!~3`u{QX7=ZTL0!<Ks)x~uK;EWkD=@p~Xar-t&mzaO~W
zf1B9UKk>ONthqsQL9T;*O1)M=?*?Ud44I>n*K@0eZ|h`P^+4-r>XJcaN-3KB+pNI<
z)yB%vmV4XVgSIJ9*H(bI-RD)gmC$)rjW^qpIbL@imXtohn8c@=wQ_BCw=Dcqaf0?p
z%KIn?CSoKZ9SS+eY(*%{>69O*jKvx5?n&!*)_FyHG><on=FQtE_t{J>$|P!yA<Kbl
zM9nORZ|d>(GcBMIC3k(e=#3ZOT|t!OL+n3J?#&lZ!Az{&uKe&;BA6aG%+G13OKsxH
zi=h+7z{6-|mFUCAPC-_r^2jPH9=S-A?g@Sy#@%AtP9#6wD$31XvOA$zo>x5yP@KZ|
zb#`G-sm9RJ)d^k|!%(tZ6tfN)pl6_O9{qU@6<}fBbow^=>^*Ir(2m;IzLFaFrzP)n
z%tPlOnor7>Z(ORfNode!-G^pbJxTS>+uk7j%<l)vo2tczhnD24{Qrt}rw0i@6|r*h
z)#3z-*bjto&?m-;+&Fz~UL0?_^3x))l68@p!%<F`fFXq*_-v+?qxm+JmGp3yXlV;c
zc+y{+PcDfo6t9c9dqJ#zp1>lSq2VLLT`d~XWd`ymKPJW>Bm&x-KJhENW&ZKXR)ZJO
z^Fsy>fcD07aT$PSR(>!n1y-<c%KG=G0ricwK1sj$H(+RiO2(gI%hT-pfpQks`S+R&
zaAm+nwqJ={jyRUEQ^xH|8_ue+lc@^<(q;fsYV-SnpOZ9K(w_Z(z=Y><vM;b<EIfGy
zqL&)<qz8sXOb13Ks=iUOkS1qV>IkO%{W*SMM>dnmerUPGY9V-(o;Im0tGvO__Ug%0
z_jADuOHTXvjBlHpiD9hKk<soP7#t2W<rcmLa0?JV4G^>Xz<RRsDXWG)9>D?1{TG}Q
zYA~%RXD<<x1!Mqg9o_P<UyaRykj`TOUJ-CsYk>aPbH1H31iop-TkkAp?WWp-b=XOg
zpM1l-Ee%U7?|0@5g8jkIWg6>#)swwp@h9;cYH>-*&$1nvv0Ej216pp**?R}KVCd1M
zj0JJoE3o)aODdg!r6euPbvof|WeR7?l_0(cKC$|kWX)er@{QvVtBMvsptyzjy3~7{
z)3s}M=I-~g<J#w;rrU_XU5tk+l`#kuHX&tV-eO37zRlSBH5YBf@y-UYY<{O=+n5s1
zijxj~^2}FDa+~KGO3kTjts@Cd>X-8F`miCTDeQIbYFJxy&3E+Bys7nKkpFP#k`p``
zc_lHVzu4|d=ZWN_J7`vS!OK`RYuo&mPJepha#K1^C@~I8G8)uyGCP)Ff$^^8wl?n^
z2WdN*TysTYiq4`8Q6zNcgms?KANrhP63khO^tp-jx7K_DSacs@Bb&m&%vN)w@Ss<_
zTbzas2b}ZkJ3x>5x}`M#=-EcFXR|XQ>BUc*7a$X#t8D>$5oX!R#hVq5B^S!@K25Ys
zxj^2ALAM$1ztn)ce4htgy~x@AujV2>PP4$};{wsTS1yDk4nwh0=c@xe;4Y$IZ20{?
zKA52h;48jf)+d15$PN(TsT-gVpY8H4pOL~5uRwHqw8VYge(ICg>LB@pXD4?_MdtXy
zZ*(@C397CN8DUz~PV<`F0BAs;ttySV?b3ipc{mLqohcScMIxSHmF=?N`ARc+2K|U7
z_FHKK1`PRGPHp?6Qs~v-PS`~Mwq^nGMQYGX1V&A+%en!pr*}WB&7(XSWe(oBN{R#L
z5&K~KS&{J~RQbn%=GD1ZveFduDXd8%7LRI81}xH*sg|%s^Fxbv4?G>whEA$DUY12=
zdlWIxD%xYN%<W<Rc-M{i!*`G5ZMKXwl7Cxhv^Tz8sXb>cpVoyBwb5HIx>PsZ+dOt-
zQ@{hW${$+`L*lZAQVv$bi7!j#Uk|3J)<cfbaXfxN>-xIwBhJ};GM^%xLI)FQj8R59
zrX_dEVOUndfN3C)9}%v{Jl>9y7BnH^^fWVx(%}}9E@y`+!+;KEa%)tOXfAM3q2O;!
z*8h2R|2tp5S=>@jI!OA`WYx)d*b|Op_M~^dhp(yhAN8$QHBIZAxYhKL+t@O}zP-FE
z+TO~0GvPNrb+Z=r)s^QB19zk1Qc?$8`{UYV7S`YcDk2g?P$3h#de?nZ2K{V>$A!*B
zdM&w{NWdg+l48YfMfc`=(1C#5hN7qRxNwGkUyiBPCOOqG&!T7TM1P?<&h`NAJv1X=
z@-8y%v}OjA@GlWz1AvD2br4VQOHS#M-d1X3@||veTm)fOe`#8X%RDs|MAG<UYdii_
zud<?cCSsYYJc4)oAz(}QiC|g)dgE34l7fz6!`F)yq{pNf5r=q(m#n*qNCH3EL_v<n
z?)W3l-DU7>ulGDEI~(btrO}B+)>}zPI|GqeXuc?UP)oylFvZ+ed5^FtmqN&~yCZ06
zT#oH$qnc`=alTSLf*mwwfW4bAUL?ZWc1%MD0>+CSnm~?RqsMRm`hQ)NHti<)soGMk
z`)tJ1m0g`$|B(D$Io}zdg4-Ru;|{v{MDi@sYJvkk*ErlQRLMx0dfHr}hbS%-TF27B
z@uBI}hwnLUHK7lsr_WhYKWKkb<C~DA<PuRZt(&I#xD4SAHuG;74BGlYBTzmphOf03
zHM-h(b)CD{QoBkcCl<liL=*UONNUweU{Pl-G>3bE0icgsjhO9wL^1cQVK|>k%xX|T
zB+rL2xGgit-cxQs-~`?ADEjD=d}Ds}sgA&BX(jNFfqIuJAJ}wAd8V|mpVt8!r1Bgx
z{zX?i$Yg`GHBZN$W{)?~%SU;N2B6emsX1n9P<j-PDy<ZN*Q?Z0bLRa|yx;5eo&)KA
z*u(5Yb=H+C?k=wH$p}yThu7El%kAoFVr^|#CuAVrkmS*7xdv#F*@`a*rOIn+JIXLJ
ztX!5WJgsMbZoMPIKodK_8Lh_hwu90%W+g8e9R{`hva!u-g+=iurh0JF4|wa{ilrU=
zn}UMel&<oEg5@_Z;42i=v}jw?e*g9q0HTQ5*d~YX?)!tlCsIK$frk3AQu3P_u{g;b
z+QSp=meAF~n$(qX-~*&Lq_&ZlapuX>OBtd9`l!BLxdm0vMoXS>DX=G;+_d*%v%;Wy
zRHEUZYE(6{<F%7NC0e)rX7lOU)WhQ&^Q2IZ)U>JAo;w8WjF`9Mcc~Pps0a|Rtf&VE
z+nuU5rv??uw)E2J<wHCwZ$s*vlp~4(jNSG+4O*pVJQ<rZ=P^MrQ0BHb{k)E_58X_5
zEO7tQ#3jU2`!;mD81$L#PJNu_+P(xtCjWi_(p`}=Di3Z*+>RU6z@*q21VpF4TdzsA
z<E-;0%d?apMre<U{3s(!d!Vj9oSlAfhqP-}ilwq-U3+DOwTnpbFY@X8bWQCc*RZnH
z?o~jB?etli3Y~7(?;`Icb8J!y9JK2+in-zyp;Z&mP-sjQY6s+zDgFIGQa?7z|Fhl0
z;}0(Er1bHlPZSJll%T#7cRa^kPNN?V{uD_vDoK53{^kb&s~OTgf2*x4nBHqf#}+R}
zlw~-lSo{5cfYF}tWZ(F7a&^Rz;a*1~mZ5>MQn5vCFf_|ib~y&vA#B=Qh3s%2l}C`)
zgZd;6{-^Ra2oE`yP}9Y|=?jr^{Z6dk4=|-oX4?I-u1dG1<bragBn?nJbdjx*s!7@Q
zzbKpLNirgSKhVb?>JzZ4W5mMASux-nMH3~RnBVC{efs@C!TU?n+ZfE=>bQS7Y6`v7
zqv78$+3I=b0h2_#j~o$MbzWd(%tX7NCyje?AVN#b!=!!e7ejUrEo0|R93Z}h+a4k3
zhTX8w`_R4vl33chY!H|n@)ER`wFpsD1~X?Z{Z%t(T1y8Z@#ct<+oLnl%Xj)egd$)}
zu1wt2A262LA4u(6h~x~Y=pd%mTQ;?s^+(^vB02JSs*X#&(*^b9I*CuqW`id--PPrd
z+va$GHmZCa>pxO*6LN=EOydTe9X%4<s{F72J`jBq&=uU?ysc(Al|eWVYyhvWic5EE
zdq45H#&ZY}+J}jU6-+cQKLALqp^Tm7Lvs^8`MKHs9ughgdq#$62OTrltId9~_$^4Z
zpkH2d+Da4=KMX=B$$aw&?Pd`*<D{4t|7iaP0GzoZ&0SNgV>3)xBcG?6JwQmj^7Lo9
zz6f1%;EWLm=H9S0t%xw@5@$_x3J{*>cJ~~*{v`tO!GXDJd>3n1{C!ug@44}RKkxsr
z^Sv^+=1b^@{h3FpxCF;~)wFgOw+Gt=yEhg_$UGUet-}6?VbZLs>Ji1-yIYQLvmI}l
zOKnSgHgCXATkGrI%~n$O!enAJ_y1bj|5?6oO1_hk)V_DsIqw~zMg@`67*k?tFt>|l
z_Vzlu$ER1fvHF<XO#$z`A9pFX_WHybMvQ~qlGS<EX>|V1lb)Wq8+Gj63m4jK@Gn7m
zoXWFsr4Kug>$fhl#XI2vbdCOim*Bg5w84G1S!<d8va-r@2fSrGiJcoRCS|*14SL=1
z2s4_C#q8&9Td3C_@_5zcUQxtaw>%azT3LfqeG(BeS^>UGaNhUaSY>K75GE*QLX0Qk
zi?|1oW2cE5B2t%Kw9YN8OC!(#T`T2EE^gui_IxeA=_D}<Q`3`O&}=-9-Zo~E3BBSH
zptzk3Ha<y%vc+!NR2q|hgQQ-fVE0r1{R_^~AcWcWwGa0kxi)vcHug7{{nbFn_HnCo
zPz91N&q&Rf^`mDcRCiz^;`H24%O00&7Vxkh^!^C*_I5@l%EiZUzzvUh9WY-RttX4Z
zWFE!nNws!f?#ypg%ut=;A9WM)9LmqT@zm;}1mv=Zw5Y*Uj0LL3-}YjBza?4X{7q2K
z9h@mtR&29wZH~T-!dURn+0qaeD2+BuKly_Bz~__RvO&pbQ<{1AfKXaLoe2teF^wGX
zF~CK682STtbypU<{;-kyxS_=SH=o=>HN}o~|7GiMuY@T}@^{}#ib8#bs-%=sBHp<-
zV9pDzbHnvvm99eQk%SH<W5uUrazjP^A-sRfUb+Av<4vrBD#RYMiN_kk)ljN<Uz}9a
zk7p~!_&OY<LiEEH7Pv**{>;AO=xFnH)N}U4WPd?`7+$;Ss$4Bf*r`|WxI9BhpQmRd
zk|~oi8wf1g#-ph~uj{+V_{ac4MnWX7$7k5uvDxt-M0)>2SO2HGH@7D)iqso<GC5xA
z@lMOJ*R(cq=8AT{<RB8;@?iUv>sdzA4^tv$b$<R&yEkjQ)-UM0RQZ5fQF;xNp;ydr
zoyi&xU-Eqv`6u9$4u-7$A*%&EiaM;H8F^cn%T$qY6hEx@jw5|a0z1j<MAn(?xp1cS
z`FVL?)|hUlKy2&gU19U1`v*5-rAf%d@cQAfe|p0IxdXoWT+iJJiQ#pOO+1b!C<F*1
z8k%>U;d+Z4fg9h~M2E5M(NInDg3!(4hI@aattGhh_6M;-Tx>?qclXxawK?bcQd2Rd
zMG6uI$l#@Pl;%Nkwb``uyseZD<`hGxi>JE7XCL#L0AwtEdd67fTlUlAx>bPiWdaS`
zki3aki{6E2P_0v@BS2vHUc&kG4+(X>qq@ep<~_+}p*zg~LU4*_`J0!e<69B!F7Hxx
zBmAKKZo)&V6G@U5L>!64UNyZNKD&8)`?e{XtZ<k77~=jQH>hRA$Zc5#g1yCTvwija
zK^+5qbnDgL)g|T076XswMzj;#NLpu?7ai39DaKYR77{u;D2AR%gh7aeV6lt$=EaD1
zGN$gEz6?^9r2f?}=h1F^a`dG*y(7EiDS^)yzQ_&Kq<}ZuwJsamyAIs{QqullDQr=3
zY%G?Nc5^Vuw6y7uMNBhib$)mX!hb9>e$01H>kP`gijw87ko^LcitTt|KEfonlPi2m
zrrm9_9p(b(Gl0Ao0z1NQiEwbp=e8a*86TS6p;3}$cJYvNiPVOgcHnj;4IX9MF!YbL
z>V)hugO#*E$M#wq0l-{8zTuG=Yj$jBHO)+ywe7VY(5FZ-(5xcWIj0n&;mIHVtv|?l
z&txUc`L{=qeAC}`6tzYM&to_G!FE-1tCzQx&DKZq>mpmjPF}<)Ax=b^G)f~>jghHO
zDA^|fPYnMO*tw9w&BbZ^P3vFjOxr&%l_;aWkTA5CmFTPAuGm-xH-RakvwhI=7wu1*
zB=)i_zvQVTaJw8PPP(wMQwGofLK3a6NgqwCwt`;;kI*n4&52J+63;d*)1C%vO{MFh
zvH-wGjo}U(3@DqJ@6x|2o8W3!gPI<McvDL((lp^&bVl09@Q$!QV~N@PJ_sN_x<o(g
zTQuL3BJm-Tj=1O9t+?-tPk<=<IcgEp(}{p-P}$I~u8nly_XBU`1Wo1etmb{xcFIvA
zLj$?*_4Gpkb7Y>_YEYan6PBfVDh3BQ#A-gq?8rRIqdrS|QgWqDJ;S*)=zCiU215gU
zdg*Puu|{{cm^SY2=c@%?u7kq8`7>MKx9wXmp6^23IhNG?xP7SZ8aWQCgJ{>6l+*4S
z88&{LU<Y238LK8Cqu0nx5-V5BeR6W*GvT~r6u)9PLOEC9pN|e@Spw_hX!TTvtoVJ@
zUUY$6YFiY>)0Y<$zuB2Wa%PXq?3tauP;;o6uZnG_)Z55yw3!;wL%g5#HoCa2V7G0e
zgsM#>6ilEo1FEX6>oXm0n|Q%06+JP59^MplN1yCjSl;uu09e@{%TzT0vFi-&bII!W
zumn~BzYIecyGq))mx&Rt1vZJlz6*{mRgh2b3!ktV=I~JwRh{nxb=_HSPx*)eCbNs{
zw(T!vTz2yqX{UNePFSw+`+*af)Ht@#7w)n#pS4gL{p%E`O09$tJezo2!?PhmxFoW1
zc0F!$s+Q1&cZ%}OmsxZdT!I`@|H#sCye?hPz>D_fE8$#`Xm=NcvNLEWrypaRZrr=u
zkFQMEaLIhK&8JlDL^<Di_b7RkkUsa$3}1EGx`>-PS-&}}UlWbefkGCH{jSWbceknO
zn`-FxNZQ`wkw<V-_^G5!+|s947s}`c?SUlB6rny58Feef{$x;o*(1z(wCdU>{?Y!?
z*8=`>WEvkCpV9*;)&UjiD;~{+Z+y20`=q<9Ns(Up+mgsig+eT8BlpdTBP8(G6hf>2
z&_I-DeJ)T<@N{@POw=L;<91niAu`<!E;d>T9H&7+y4hNbCpNO*sQueLs7WCbd!~Ah
zIyow?0FE=mBqEVq38YkOpNjh&Gmiicw5E=uwYM^I0j>Hu#%Nh+V76E9<p&mx{GK!=
z=iT*y?ld`FKCN6&fK^iS5}5bpTyS#sC|vX+5|Vs$9hu0`$5;b4lSM@M(1eL!!$ieC
z4A2-ppc6pX2A|ah$l_W9pAyj5mv`gMo6FB3rDT{~o#^~(OWQ8&d4sa)uCY0`hHnFB
zS=9UNEaR&w*_XHDlX|wwuY{R~a(FetA>C6hz^w}H94^1Ie^#=3cz8>DbY=1W++9h5
z*HjPoO{BzB3`OZoYHJLZp@J&JZ#VZ-Q0D?G*N{+$a}w@G%<l)TO5m_vh;OMKti*4h
zwA;RekGN`PDFVgrOq=vW4R;_>dqf~&uXQQ;!k4KZ*RmI(*B4;xlR>>3B$<12<vCq3
z0BzEk_?W7&vTusGqdU?_dLObezj>>2Wn%#J1D8@aJoL#waLMVkct(pw$XN9UoiRd&
z#tkx3lwSkuu+AqVovFu}DV~Km_iU=~ZlYYurpQUsDqOj$Vbfq~dO6GkrXFN|bA$SN
zV~kpm#RC9`bq?+Mv&qE}&Z@=*U~c4V6=~-ss!2bN_uiWA1Z%hjK1KY!2=71l{QXBq
zol#(ayIePwvbH#@jyREOS~fja$I-rjxoqy$U4LgebSVe*T$k?X^*(f2zu8$1<=s~E
z-Frt*Aj{aaSK$L_qgdm_MPGD|$QsZAO(oFZqhT-Fm*fCDhR#U378NR&=1o;5<5)2H
zyR!*0&*PXx5=c7$_Q4Ga`LiJPjx_D_Hc=FaLq?A`+#3sS0xfix=@q9iPHf^JPIAa3
z`i#<8bvX#*8tqd43>ZrOzNr;`Q0xCx^2g2qONU416l~`1!GliPsK*ngkX_L(LKib=
zHh&_QoZayjs5VD6kjZAm1!gun+Q~!Vir1~|lJ()_iI)JL=H}kY>djTZBp|Vcdfr*@
z$5Aa{WtQ4z;H<sKK;A)(2Gj>iybObT&_1L(#>D)DS>2u*RIl$|aZ9U)cD$V?<&lx6
z(jakO!-u7B&F%9l+;fJi3nABl@Rm#?;DWV#j_8KybU(UNx7{|UQXm%Z8s}|w7!;JV
zHl&%)-fpDcXYlzF4W~cXu02`ns^-n^9bOEQrf8xUCjrEX)GmcNF5dtBz{MQx*HdDH
zmG;y-Si9X$FkF6L$uU7D#zrOYuDZfUk9EaZL$mWyc9y7ig?_S$l(#>ib9-X0`cjuU
z?GjLy32Hv`-9<lenn4;xUZHXUT!CxSuz?(^^3zr9pq3LlOV0Vj1v`K%%)_C+sib`9
zJcZm8_(U2Y-PQ^D{lH)1s_bS(QaRhct@y?wf@i<&!+A9_L1LG?3s3IaOx>JTEU#Vq
z{XjL%Jq+!#=%0pd*@G|23dq}bF^i?q5UL1MW>#j`u`s)ABSXzk&Knl&x`N|=Kfs3<
zXw2>&O`)MXy*Xx5GJ(!oFYCv3ZSL-;diMB_ZYIj;<*Zw6rEvU0f&>elF`;zYcExaG
zOr;4@8%!xWL9}OMWob^1P6bV}`=wRAKq5FV5tGXRyM@qf7%nD58Ns-a-q5Oyl#bvG
z4DS?m3#Mt&v)mH^picf)X#f<Yp4ho|E#a|f&+eL46DXxo2L0E+_p`shS|$-DiG^QW
zV`gOGzqz=nr2IHb(S8=bI1HraYlyb#nnieQ?J4{3YmnLGrdn)!g?XuNy3?DHi|@l?
zgY!)sYHW?2Mo{4Pt}C+vT~RB3?qz~RVZxJL!vRr})yeVd;o#QTD0uy#Dm9%)e^jJn
z81{i$^FAsu!5{vVzve9E7(W}~)Lq-zPM)we8noPx>GSc)f3oMJXIfZTgg|1unm&zr
z@eMZ@r^W)qg@omZ*R-XI=+H!ar*y<qcMRan_4-!!tDMBEgEwY0$3veq5z4_Zm^)nk
z?efxGKcOM*ToNiBbaX<=Hy%3_7KpD6*;i0FRx-TN?-*G5a-+{@wz6W4sCOcaZTFXe
z0NBpVoCWmG-y7#YydJ!t^bND@TOsI4bT8#TcM6=lu_UJZKe7&Nz$Fb_U*4#=5?)0t
zX{-q?A8on}I8&b+5ER+?yZIQ??IA60Wr-JBG>Xx+OWLn&)0D$DSmbEit0hz)ULUb1
z{QA9!FL>U$u&9z0bTF?qS*7np;t=1|Y*P=-ItQ3Tkg%G?0;xgZFj;djl0BJUS%vgX
zIw{6@7K$uQ#lxZH5Cg2C%Do_J_SxXGLkgE`%~8n((8rWMq2KvA1Vq=7v$s_KLv$Xb
z^;L8EpeYPqzaBwJ?`WyLaybLpP$(?FT}Cu;`H^pa^f4Gn9QfOwl=y=DkGpEJ>eGv1
zW_H1<+hFkGBRw1_H$z(PKD)IRQ2$f;1lg>mETFloxVQQ+cEZ^sOTA)-K7}c^y_%2L
zF-ABj54cD@Hzb#%EDSje%`sk8Z{0uwwup!_4ymVt5(48{cx4rpL|APH{_dk*nmgBA
z58_AoXr=a})4WN6`%0TyoL<zYJoL{G!(tjORdvUM9*nH$g?Lu_WZIOVqeXYu7MDo2
z-iw|1js491mb!p10AX0ud)YmE+UYLg;211}Z)N~IUW*V*g77;<g(RDC!)f)^fW$7a
zbv2X-02?<4T?-p_w~>N7r9y-7<YeZ=@a~ST0+5FXB%6D?bv_rgUb;J(G`U<F?)3uf
zvT@!oP67K=#wmLR%o;$hum^A?n)S~_hK35j1|-G5ZgUqdfO3F0DcnoRI|aa;{wiNT
zym5RrJ37X+-g-DrD1o)E?}UF_fS-#inWDZ4xna=Kb7z@Ac(Y48bwfV>3d*wL@Lh55
z_X#E2fdUHTs)6qK?utTvWI~6O<5b2sNk~5Cue(B~P0kt%1dJZo6lnQ5x66Ak*$?lc
z&bexWSO^tHEJL;}OikC9=g6XfZYNGE<yL2G?Qg0d5O0a$jVgl%Zw1U3(I(Gp8^nY5
z^)<Qq8Cu`5vZt|;%%ZZZ`CE>Wn-I4FftO8kk9TFIncIW6c#n)9H{FFAJJ5~cqRk!I
zM8gu?{Hfka?!NVCwPB6<UaxiFa=UjkdO*OQb}snLPdr`{lkRV7VC?9>unD-yNan3g
zIwPz~l<-wH%h8?yi^i_;>B4Yp{Ju!TwxIwjXYKp}!`^9T(*1O|60G3Nr#mk6qPvOg
z6!%{6mA%qlN#=`mZi&5p#$6ktp^Y~x`dQO}lVfOAHj1_aLp9<}nhAOhUj0Sq9~2gD
z@`T2K&82UWRN1~n<5}$Sw%K<Wzi7mT>BuQ&hYQk=Go7PEO4jMPj?`?RX()g!iof)?
zlAi`6&+K7e__-4X)jJ{6<Cd3FCX%)zS(+;8=aMm89s<EKz}BVKMhV?g%{F_ohBnF}
zNl8ggXr=tzAK7g7=Sma;fpT;_srUNlY{2pR=L%Qv(4h+ZlE%NWaQj?u1my!9+%w<7
zu=>75>w!S!x;U7Y#P2TvfJzs9i}e0B2OVH<Q>p+iVlcBAC^Q}Tv;`>AfogYMpKBh(
zx@#T{C%D8qhhe}w;VTo@#$ia(srj$}VW9f2?;flGo!#G`F*xpDv#uR#*`;M49!Cf=
zGrm5w0G$|LF2VPCTPI-$3e=Wz1x_B7&iN554>s35NAM8~0`zEBbXec#UF}23HZRiW
zBih0Jeq+l825|)?#y<pG>K1MZXcgwFsMHI(N8u|Ayv->mKXNNR(!bx65!-*s6_5A!
zXlWrJ9(?c`YH4k4$;bEnEcgZZAFOL<`bU<#5|e17vKsDdUnY4@AHsZF?3V;i6rHNK
z=<%p&ipo#P4NU~Enyy&9M_RU?IlPFTf16+4Ss~cYe#BE!Qu3-jddc5tm_t?1Y$gE!
zEVf(rLwsposfwOZPwzG3LG!`{kW_ZfJniOmScj4AeI76p`jh;n|9TVtKm0`ufeg~1
z*ssXsFsM5__YXGAhm6zUk7mPd%8Tm3taA^vR?a-05*dB&r^AtMAepmL0ER^LZ*HU|
zKj<hJ{AoJ%y6ik9bmP#odR7%>;yHImFrOirf}*GM&;>du!_`vG!MLbUPmqs8Klx0h
zAIMYv0t}jZOcwQgd0(@3`<R_U33qKd_^HGY#l^O-TWAS{yO2UkYj9rCw!K<Yfj>ji
zVJ217SfMw$jFp|)g~l#v3^_I6)VsRcAKfp`GWVFvu+1%tmK2A=YeK8n{{{K_hRJ4x
zp}AFO)L!$5(9kD{AM)NfIQ=u2=+4j=Z0jFlcAk|L#$D<%IJN}>nYE@oX)ojGuJ)VO
zN1*V%zWGlX8gu&4t57wF*}V0m=O8U*)K1ELLRT?)bPHBCOejw7H{mh)4Q`-L-MP@3
zc43_}tpNfto$rZU^10nQX+r77n)!AXb%nKSJSTtApKlmknut(1k6-pejGxa>6dD!!
zQt)N=@44Rae8Y-Zq+@keXz)fa;QJSq^G@8M9n5WR9c&$l1ad@&KhD%sHT%~l_ByIn
zD1TuX`tB9)c{eRBbBh2Teoy|USV1EI5t9B#VhypPp};*J0s|xdNGT$c8Uo)nJQeL{
zIa?xpa{*}j4kZ8sE}+@Meg9EUNL0!%@ML)~&CFQJ1>}O0$na%h$Y@J@TYF2bo@x>T
zfi|Lh0M+YFz`FIco6FzqgKq@mLW3fMX=zY7zs!u5uVNL<xOk2dIn$k{hy?I!tzY3V
z>FQ@$l+}<*T?L?U<pT6Jo)9*iM_f(%s?TLJDb+myRJDkR2M8qqrUZ0WVRTp%-qdPa
zzN>X_&Ivk5_(1OD3ixt={xInmRTi!3&Tt<kh{6FWw%6Eb&frgesKn!rCueJ1_qsK4
zFXW>^R>{c{r{v=s1*;v}_B1I=QYK%Vwlal?R?>L!q6r#{iSldbFE0s`^tjX71Vps|
zV&&$RS&UIiG71ddFeh#q<Pkryw#<8>J)OZ@x{z$zjXl{=ucRT6s-Ovu6K-Oz^(;CM
zZ501yqxL<jvQ&{o<y#Oke?;26KwP;I(OR7Yl1I)yC7;L{@+T>6;qyqUH!3_%;7Ydk
zTs6NDnaOa87h6v^_IB(zZqD;`S*eem+pS8wiGf_JeBT@kwDzrCsexyeU3^K*^gHL>
z_*Ma9F^oS?ZZ*iunL_#?JW~9E3~p}FYhu_8ZB;4@{v24V765Bi-v@N8f(6Q?UeBW3
zSi#T6%XU<xqJ55<oBOiu%Rjp2|HYqG$`fz*lT|bVm%0xAjJL_hrKt#|<mWHW0saP4
zQ|tPS<8~+9xL}Lj5$a9eu>n2UsVgS4{eD+8Uf<p;5$|Iux*7!g&D0U6VZGvFA;!R&
z*%u~@yJQP{T!xBk5pr$Z5!%S1k|B*`CZl^N3Dls;b;>2Vdg9j_K)SU6H5agxTTDVN
z+t=$fr*uS8gSTX_`%%mqfzgj&VQc^I+bFhMf@+y=(_`S@%jOUr`SEoa_4tA|9LloZ
z*iF>giYltL!SLoyb{);59Y)2wy`WUPVzA~D4FzT2od=uU4_G~?ViQ(9Z2z^p@@WIt
zaDg7mHwqcn+VphYIjg-^C<SHY#$?t;Y^>B0lakf}#Eos^1id^igtsZcrFLcO>2vDd
zy#B{0dFV)#CIl_)f8K(<xTbw9Dv)&(XRfd+g}#bV)OE2d2`;O2Njvnc#`LFNLu|eA
z-i+p(2cL`QJ(J<1cJ6J^{;cvP1N74K^bY|9Um=juKI|I_eNQB)enfcaYNnC2CU}dy
z?vEUSR9<I(z2NAFb<paR>$Z9|(O7D1xyA{`>|<nPNTa`Er<$eDv)b6<tkz+i!j*)p
zs=Yp2r^`#U=D)MgI?L?M4cc$jIv0r3GzrS+GFl3gLZvrDnW5<{g8Qn10@T`%ki#KA
z*-KfCH|ybY>>AUPI`#ts{J#aew8`BNM^yX<kXloY-TeK)w-<-)W|PfYXWC~ZHItty
z?_!$jlvw#g516b18*S=1q&;zl&-ZvZ2sxPSpBIvN^%K!5dlX0y+KoEU`}iOo(7DzU
z@5@md9cW^~w_c(GIep1^%o@KAK-8v#pR=f#=)FjG(~UhadB7p7SML6>DPFmIq)mv@
zY^Ahhsuk49T1dX;h2=A1{)8UPb;P6uV|Y5awm!_F26ic^7jeoa_@L&cz<i^=Q{Dl1
zq}~ga^BF+}pUOwxgMMCr;DN}Wx2#VMj!)#ka_7j}^wLhqvx(U__!I~^51I=fhfmpg
z19|(pefUxhanm$LB*nY<CHvmw_3B|P)&Qqg&7!AybqV{L4#PI&JO8*-{HK3<$g!Qr
zJp>r0%gG&~iRVLFzoojdgg<g100T4+;>5Ya*L(S}tBW+e!@6A)i->IuSIz<3D^@>r
z3OD<9ux^)9p?Ci`AAi2yeJz5rozhCG2)S${t?gSYcI4uXRMZUqWzAn#Co&ZunT1dE
zFWENt^i*IFs~b=f8@=%Qc)w5h%*t?ll9O7n6<x9Z?%TrfgzMjZJ$nkF0X}qedHej<
z(GN~h?`it&f1N?ZG9IiD^rVZ>Rn=>SnFV=UHi6=PJx5Qgsuy6-(GbqfOee%AwI>Z@
zyYG4GK0kc>_|AX2A^!P+OVBX1@o2;%0$McKv$oYh@gNav5Dz9^T)`OpX=$CV^2SLW
zf$jzeHF6i6B2CD*BclQ@-ZfIJVLb9wPQ9XV6SuziXVD*lV9N_N0w3(0)TPI&DKL#%
zU2Dp2-$xbTcIv3QEOKo_Pefozq(u!G8MZ2LW0u-d!pn@V8;)DWLxh0HPOFSd+`d4>
z&9I*i>IqQ{S@mSk@)%WPb+9Bdu{%2e6l_f$a;smK7)>&$cDgGRy{Q^;R;9zVUw=>1
z=%BVIzb>}9Gy2Me{9F4|Q<{mmJWF_u8v;9^@0dD_Hyq2LjBski<_p^F_Ms$)$req2
zi=*gN8k9d%Hp7aHSf4URTpGB>=kRQVadH&pW@cW!<+2q~DxH9|CBS-M+7UZR{0&g>
zwA*AB%)@44Jwn)kh)NJ>LL&fOE^%?i`I_?peR3T<+}wj>Y}U>HNksiqq}fFN=hmw?
z*e&ANcOHNWyr%6=6q@-it5(y_S--(D3E{+cyPdDG{?Plir1q&{KJoC98_t`TS5IW4
zS<y4i66KroJ1Ml^)-7bgSmWQ;tx5>0#GgcS_f|rKXn)yyMJJy>D_iyyW6y^s;40>}
z9&M`-3!I)a#)l@mL6dFm2}X!{s$B(;TtXxO!81%kb8<kic0ZRR{=W~qHpSM~0}uIC
zcIt@e<kaNID}$eU2OmOe``g=b&t9Mqp@=Kq0zj$j|El@*XMmL#Hd5vob;E?=^xYbq
zejg%-EG-#}-P{xf(i_S_jsfq|zP&i;+}ae>XG+B?*~Avleyy_ximyi>7h&JHjqCZi
zk-pA>bbcy^oW~m~;&o3LBYXgJIRGO+q=?lKEgov3O-9h9*cP#0hZX+Ab{SW|B~Q8}
zal8g$^BF{~b!FxEJk8`s+rwAFZnjNFS>X{0w<?q2X|xGh0myPcMdEBLzfr#0p?Reb
zMA*KhX)gasP+Yhc!?6qQ1o|sqZo+~TyG_Y2?CKM=O;U_#C_;JsRqc6?O&}6`)!Os!
zE(Ft&(!0|yL9R^on9qvcQo5Q&oF%Ao;a(C@aT&Io&QJ>5zwo70cdcSE{<>axl#Nrb
zheZCu#OH_BouRXZY~*A?qz5`>g<GYo?f+>)vFhiRY{P2*eur)WY5UgQbFJS&nfa!3
z@4*DkQ=RRYD|2Jam1Kf}%Baz7@!f?|$ov|`T@kp%F_R-zy?%qy7c52LiM#t8@C98-
zwNLD#Vb*zI;==$YKEqPDCP3Lna3{%UgzQ;8V+}ovMi<?5Hxjn8raJv!d~7guj4Y<_
z2YuZy|9n*Vr$>BKX3+d=qu)HnWoy}QK=#H8{=G<peAnQO;dWesmlNfEqD{p<wseb?
z<j#;UUfPHHhmZm#`WAIIsqa&dlAqcH&L@#$_;GM5HZI(HC&7`vw!@m)6wL2qvG2c!
zvMA}noAL<t>VdFGj6cQhFQjf{4R%LjUcUO*py$^M{^|LJY0Vhe<K>U1<oEd`R+mM5
z%RpQ)?lr`ntk+wb{5kssmVQ`gsK)gM;5vJ*@x7+WpGNNUj5hj2HwBcM6JFSxiJzQ3
z7Zotxq1bbDECVsjBnz4c=uB^=jZYV>4!J!M@1GwjC$JX72B_N|&I$WBH-17J>;Cnh
zp4Wf#fJl~(_>jWA;L_)_ps?8$JFe}~69%<9ZR^22LP%S^-I8|A;RvzB?5axo3%5ia
zx~WyH{tGui@9dr*(*03UNu;y&-m3OpoT|23LbAYktGXbhKQfEN&n)XUtleaJbrrB2
zlLR+(>1orJ9{9{2%U2Kcl^8ciqi(&q`7vPk;CZBcTFbHR_J>nh^!SOw#ts2{(0i;E
zlKRm-GapptidM2(srXBN-6o!xtnA+F*360#h7G?GkLD=+O7$LU>!^*+>vKse?CkpJ
zcj)#nKr0+N8>xH}B^7@1O?)r)tcM~eb3PU4xBqovk7)(I&4k0*QTbz->$y!1pGuG1
zHqc+R2jrL|!f<2|J~7@Sk9dR_c)Z2N+6q>=lN{6ji`Da2y@M;$Phudd$E)u)y)Zo-
z4{{h7#lMYm71$8Z=g<8hRPA)P%xfgK;VOM2^eK<d-jHc(Uh%%#_VsN3w{3DwTd-<B
z_=-}sx_Lv%pKGE#tTK;`L|QtRmTW;HnoLtIMhI^r3UMS=528;Z@6u+cL%+|>_6;Z3
zy*%1?G)Dw}5sa7BqO}I7mWyd|b8Ypb>#)w1sw(vIy^dy>NRc|Q2FV?p+4Z@OVV0*1
z(=ec5_I6KB54UXD)|0SR3T=y?^e3OmItx>B&cz`Sv@TTyl8Xx$0B&*8I%0Z?a=Pb2
z4LvTokYhjcqF}1c^5YNdaYugg>?nAYXfILzW%j7gy$7Rbf@)W@8NG`W*JL#>Tsf>D
z(y;CXoAH2XY+2jau@vk8gbuKE(Z-d(N&hvrW(kwo>2Y|~&k2fpE=a_0;G4wiaY*7|
zq@qY;xhSEoLuS)TF!obFrYIewPslXPH&Y_!T{&3}#}di2W%fI7Q68YCW#*I!>j~fV
zXjf$5n4s2Z{}ke1_^xO8`f`(vn8X3!<d(>MQ5sJ3$1n-$pcWx$(G8b4TN&@j(z^11
zl_2Pfx7o6^wqQ}ALl5*X)z|;%qW2TO`gOd2#nPceHREj#K{M7+=+qMlS?{+ftB$g~
z)e{itadFse-zl@VNiW+^7s>nwNjFNoMX)~aH`eRE(DU8G*&%)muo4@9wlgQ+<$N0g
z7@%}`o=2=uQQ;CJ&nYMBJgk{IN_Bkk`G4&~8n-j^w%oQDsu@jfA7@c3A>L!KWj%o(
zhr{u^o&S(hS&2V<(Gv)r;ccx<e!N)?n6kk0O90Zv@Vj>nfSJjC5R-#B&2Aebl*8fh
z3pN&}Yj<DG_7{V~^Xmb}RyWeuytha0D;K=s38TF#>sd-9ud*?a@-nwl_iNdU2w=El
zWAO;*YaYM@6?5&{rxf(n<<;$@jni5HIo4I(gPZ>GbNlAX&KtnLyX0*-!@d7_@6Rm_
ztt_6;Pq$v!GWVY|QJ3b*GKKXA&5@e%AdBP_iHq%tjauM=T1}FLt5XjRbn||htRg${
zwshw@<J;k%#(MgaE+Dd7+UCf*_ia3OjH7&1Qmx43>OEz4(xFf6IT#Q`fb?{%z=*rf
z!1$Tr8{d}yZeE}voDi@(12(&HT{l`ZoiG>o?hZLw>vA-Cs=lmqKa(D+a^SO}__!5t
zM~&@|LoMZ+g@hnv7DKr53ftaFT^K8k=v^sk#3Ud6G~C-U=CnD@JII1Q_TAl-kdd^C
z(Z-<dFAEgd{5^`(9}%f37nc`!^a6Ns|D);Bbi~_;sP<DxGfAiOsX4<Bl}rA`vHHb6
z&~#pH@w2BpU3y)>;}i+9@rmRU5m+Frb0Qq=MO+LKt>Wiw`%c^cG2wr`^8d~64qDlT
z6syF-xH<h~&l4FTbIqXBePg@b{hycxYIoi;-h1sQbn_z3pNrF$doS-l8U)BjAf`I@
zfqy}o^2$b%X5#4B-tPxQh^?%v=&l}M{`_u32B5V6EbrIAE~{A2C(jf(J?R4rVOQmJ
zem}qyO&tpVrEcw?o$W!-i1ZK<&p&Td_ezbui9XSvyhhSG?`qWRMKwTC(cAr|!TEMJ
zO{`C}M4!ae*0NW_vTj$@W7l`;wJfZ6yCO<_S*-DTn=%Etwu@~ITAnPQ3{}GR8ZNTF
zwIDA({U9^ygS#@y>}p~&b+AgzFBT=<T5J2HHVaVr&c1?WGEf)Aql>I>H6Qs;ix^K-
za5cI7A=EK-k9=rzNjm~NwLXSxO*VgE^7JHWC6t4#tnq)SnsK238IRD{mXD1^#KW&5
zr;L2+i%f<1o{Wi;<Pq8FGp@3CNfA;)Vo3{eiEzULP+|(_@U9H&F}>BmNpVpkM!MiP
zxl|HqmUg$EcBwOpNgbZtdJHzPjS5_R4_f(#;Zw$jK7F;RjdN98NY=Sf7Ft<ta8mqX
zm;s<{g0dCB<4>w_KXlKDafZ0d4`8QurC(&AQtcjy(0E9m;$S6n`;u%0eoAJ;UuvSW
zVx^V8?ZuQUqMGT~K$~iUU8i3+7C{%ML(3i~FIlqI3l1(0V{UgcY`>R`L=*V>hZwJR
zYLlzbXSQF@LqGY-SKknMNSrsWuHGE%%5Bkf&Eev=iWbs16{V><_x{hv{+~wv|6iB7
zKg{26f+{)Z*<CN!*r<gTGx)0wYF9tyy#b_?pVsNogAv#_nqvR*iTpn;&R<TtDKoWc
zMI~pwU}I1x??#y{xe?N5WLjLB8(=Rb{k)^zr1b{Ay&U9tPyEnoM@V9J*zCLC!X*ac
zetrlFUG}uwx*Aw;Z&}+sie(`**_Sq($w6jq$<NlAxqzA)&D=7d2!`nTzh{Q@xqjc=
zRQ8Z)sWV*d3TWZn^M9Us`?s<B^Z|*c#t+*{j5(WIt@PVCEagQxZ>5|oHY+QdtMFG&
zkaL|AIGVoN>^tuR{ZW*alXQ4y_4Vyp#GCvkw;Dnq{HWxYe>Uj<*M2Ye&9b=%TFe%j
zRXrR_Imc*q_kBQ={lNvR`rX_;#g)K^RxEZs9cEibtC^jz{?5Hv8_#cuXSj<iO+SRX
zm^^Qh741=$;1wPGB+aU1M#+*|bxCRhS($7X34ksbVTg69YfrSucMy*>4jJW}kT$u~
z<)>bX6g0H0Sr(xF8Y`PsmD)apEN0l4u{CMdoKP$XcMEN(&4w+CP1m_y*y?ub(5<p%
z$>sKM4fx5^FEQBT`GY42iE{G%<%ea~cCMPYJ-@_ooGa}ahjTL+G(D1?O1l?AL?TwK
z8d-Nb&V}@bfA}Bly=hoeY1Z#;Rds7QK~V@OV=Dq80y58`AFC8)EJ2w9iGY9*AObQZ
zVdyFqRECtybJ`M!2q7{g0Yab@AtHpxkOYztWR{sE5RyQscia7(?mpMsef9Ia=X`jt
z>m0tY$xdKracAH6z1I5w{>a0k7Ct#W?jz`QNB3NVP_UUsmwFRZt87QNusj=HT}mlw
zJJ>+6I2``t#9UVlcsHT#c)-m^k64^8O|w}ExzZ|qLrFR%u+m3*S)`v5MaC&8_2jRB
z*ufm?eU7g%wX+C=1KX7UC@F5_;5Ibki%+fF`*nl>L+ICaU%NqvUPdQDx5mgx2{Lol
zZ{gMUV%{C#8_P;uK5=^E0^n&tJCaSl2fc1<RmYX<4&T}(Y^HYqIKfw%O9uV5MY$=0
z$)V<sV!lgKcTJHU-}B$?5Mg3cPOd55hy_}&!9Q`jGEJy!eDXb^Stk5=lRrafk==&7
zQx?`6@mMt93!CXog(1*#y58`M|1d}pC3|V#8K;Ln>GkE^79b1&nfGd<Et|3?ldUoQ
z&i%z|;hN{n^72Bb;QBN8r9x*yy~mMz^HqqN2gqGJTeiOh5PDyXjEtNfXd(~@#hS-c
zqfXY@m@d*WhILtTf9>-GG~FTi4rYUQ=#KO40b~djPzvWU01}*m)n9bp1-M{<FCTV>
z0JtoaJVKkklE0c-**KNxG?z1gvG1tw^+}uAm|V98NU^Ugn=qrEronA&73G@=_|yvo
z!{vaD2@RBXko_3`9L%pC{sN=c)ty?q{{#u%;mMC;emmf*(IY?DvoUSCQa@rXdm_PC
z4vh+gtU^h=MSnnGJUEXrj)ULVGk8`ktVeaEKXbrhfCLht5IdR@O#$R`H$y5a8Y4X(
z^Rv)9o(X$0m?f|OzOGFPB|+%3WcB4z`a_Ishkj^yLQeoRB7e@`y9?kjlp>B;l9SF&
z(HtBT;w@?uXNGuGFCrN`ufsEQ>V;X8+iznn<|}CtA^uUi$TdD<1^d>hFU4VzxG=V&
zG9a|*u=0~|jjuDGOiev8<BMD7#~0$JNsGap#T}c#PBY8wQ%xZy)Nk|akg;RZQ2iK#
z+(tJsp14QcA;F$@4knXYEb9CtnZsK|2YY|<hJvP)J>~cG#BHBj|1bdl7Y@J71dTbW
zu~;a=+mw9giji3x1AIx!z3tiu4i=Ug^7l9m&3XNR$!l$qyv<9*fR+s~Y8J9n75l%A
zx-7bySdoJHb0w=DGHe$Jv^Gq>9C1|TS7!Y<ooJy807D28ImZ&I>`XA>5!xi8B#5u_
z%vRQ&HFrjDUZAdd-gUoI3P#=NpA~D%MZ>Wx!fmO{!1;Z%<G*`9a=&hWmjSm^m}van
zM~;F2&JXiO{=??n2Zr-iX6%pgYDy1}7wdZhQtCX8Jf5mo)LC!LxZc`~_tzRX#5bg_
z9IAyBAaj-vrnGJDzqJhyTC_-uVj7e`9iFv}i}f$eLR8Q5=DL=!KNX=>Xc!H>9gp}1
zzSSa1{%Fq#ku-1Q;o_x7qL+OfsCwFzmy!lBOBO;qsZ9NA7<@*H`MDnmr03DmS1N&4
zYt?(_z{7wz4d`k)+q!hU#@)lUU}*$<q1h|%>Lov&qQ81OySbK{hV|zgfv0o>NPs~Y
zJc1&)O{ry3)&OT`0HThIDE;j2EGcqb0en6Q4FKL77Yw9on=}EJ=gmjxYNt<W^vqBH
zE^PmK{%Z^V9RDY#{aGnix3}>_-zw#og$>%lXDkf^?X#ql0|7+Sp&>gx{b=ITs%wqr
zJgb_Lw*jm^FLd_=(Y3K1=VDQ1r(H~&{_c!>)WUwiYPx<*z{F7J!xlmDdBH(QBOE>{
z+RJe*f*GfJ`bBsanb8g#t^(BP)=wdu;;87O_8@sBSootJU*FjaYtiR2mH&&^{XaiV
zPwiLq>zzb?$Xblu;i^}(TT)EV&Ug-{+5U0(5$0XoqpDWc$y_}~T-+PKAd0}8S*&ah
z`%q=J%x0(qf=7RnHP_p~RPA@4q*tExm4($&Q_>rQwqz=oSr^$Em31qk8?kMMFke=z
zNL-<LQ~^yY@>fpgLM%tv??>1Luks3ym|puj<yYx{d0X^9U;e-TN!?^9FzF+2Fv2Gw
zfu0grE2w=IzY&(NL<(9HD4CEfB6jPVt&dfob}_~+oW*Ln9UJOS&wkJ8*@Xx&LyZ*B
zw$+%tGh#Kce6(t6Xloy?(rnoNX5%)qk$E0GrP^6kP1|Y@)F-iYGj(tzY*BjEEpq=t
zvJn6^;W(AeCIC<qL`g5<$!DkuUwAR}FHn=EQ?H-^)FkUo{UloyWVpPCCvQ3BfmEwg
zWpmb6C2?M3B(&(Jc1TTJsv2TtL{Rh%i5i?#5lwGIP}V(CpP$uMhNk{**H19I9&qTb
zqql@0!ygj<$C4_c(Ra^aCl^%*)rsW{N(Kn+EKw5*rb%-rqg*dpi8eBkbQ*W5B9oX#
z+C#seE5)w>6SqYJ8-yuGD=*j~N`?7}r*Fm}SHc|<PziYI<c~9z15b$WAyOFXc)g-0
zn{EQXgot3ggCX7nE>&N&&IEbh@hUdK%?a8{ml6H$f{174r%SZlDL1xO7+r_7Lbo_=
zo2fk+-OcTZ)~W3I@sa+#mPi=TWqP&u9bc`*x7@whpdL%|-%r?ln=DjAG<MSxCfkP>
zt!Z6WVfMQbEl*$pgvWd<7R?V3UZ#apX|MBjZ7|i)Pt(8uHTmOz{^P$I>I*UTYsksf
zK924h2yITfXCmeTU00NSr_-qk{{EIkZ_C8im4L}fF5vuZDcaJ%f%9${8O5Fzx_#}0
zBa!N#;hM-40=@=Ym#I56w7^S2O4)32gcS`F4uXXMGL7@z6Q*<h$l01R(Gd;b4{T`K
zm&O5pgHXVP5bztMf}y%DfTi$X`5vW!;UHkg1DNr6A6}AgspB;Er95p<l&iMI^%ev3
z_H<(B?YPc?&-1p=<ZSY~igJ3XM>;TXzvOPTt84mbct%>?BiJ)qQx!u(SR1V?!*G}V
z-PcH%-}ZgdM@dOql00b#ltW$hp*e5@%VHPzngk25qLPkv{Anz~w@WSF)&|H_^<bq6
zY@3Zj(vZh0Jg1wuKUtA(R!H-cimA`m@spj@G8!ok07Qlvs!r5zmH+7;_02|Y<%w4W
z{F_d2DB|6>{(F{AB|A$v;<{IVUyX5mysFx9Mh&GA4?}pocj<J0n#H`xx|?@xy=i;p
zjYaJncq2x6zkTo0>H>AjE-^(TW}+*G<h1g%_~Ei@6PUKy=WeAORob31Zn!X>WFzYW
zqmHqMGs87)o_t@uoK~3tDtD$U^o{HeLu1^15@Ce?+tTFLK9sis4on>pxAIqlNsr1y
zO40g<uN21p>(Bgu@64CYnqyiA+&Z_-9-+R-Y0gQ(e-&34AQ#hFFgqXdi(osa(MY;J
zQ=La;=p5(xeuy%htKa_CqXe%S-|uj(8eb&e>@ildvfq*fbtAU)IXGUo&9Tuqme`p#
zHIDS|i>9TVtT)>th~>adt8;`D|KXZkLq-4ICMxDpqvA$oNY<jsNOb<gIqboRN3tP|
zzVj>v2WP02NpCcGj1KwnP0Y4`i4S^U@&A5PUr0ZY+B$&lEq|p(n44Bz4X_ktgvro#
zSa&}xntt*U=m%M`W<cAk5qP<-{S9=V1Rih?Jxi!P<2aqhB({oi(Gjwnk4pgsf_>NG
zsosERWm|jbzFiTRF~1}+^B9NAAnD-I&9>#PwDhL=g-y#{^z{$Io%K<cd1sm<3iyns
zYf**7y4l@*jlq;~opS2?*_}Aa9tcv4lI$DD*a&5d<31~nA#0lwhm2V~cM|@mH?#Qo
zQ`z$|`n;tezI8=3!G)XQrIU~O!mu*K_Raf37@4#3atnuxuDj8O_pAi+Oi+2Pm%QT8
zNycm}vhagxoZYUqdFwrs9ZgE=ByF;j@ETWFR|j#P5G4{5jRA6~ICjb?fYfN8Kv`38
zs9V{S#ongJC>A%W14iIR5wl6KEWit)d~P8B?6k;B!@!S#k9acD1bkW*jTC$~{_yuW
ze@psE9v|x5hIVli1}dLsT<W>r!V*n}_;I77=Q^o;-bTCMb>=F`ux!XL65@g5z)Fu}
zjz4Yd8^NKnFjmneJ4(?eJ8@Uk_Y$Q+%i;Av14kvMM(Vc{T?Mn!>51dw50x9fiW0?q
zb)TMJK!44*S+jVp-U#smKjtxd3Hh4GJ3fVuNDAmJ89oLh5Y$gAsjTk`7WYFfBJPJn
zp^=20*<miH8a4_K?gkGXS(%&Y$s3<xQElprIe;ym-;Ala^1fm={Z?G^Do@@5-TZ#K
z#X5@<m1|&_^$t8re$^jhs+?fp5A}umGi&0*0lPCCg+}pDc)=*E5YRPllJ1eLCuv=F
zPsb)c%$ZxpFKyge|2R<JjwRuW#uF0~2`$W0g2+PXDQ>w1NEv$bil|5?qa^#s;DIdA
z1;nelTEP6nwXJ!E5*G}>+p<FX(K@;?B&0<dS8w*2)*q&`S{eQi5&wJa|H;v@6j=>u
zZUL1(GclpBzZlaT5YJrQnhc&*zrhGA>p33%cDp{<MwL=ylWMvBGQ2%4ddFD3pmo-N
z;}lpv+@iO)*0e9;%isO_Pk!~kedt@CY1nl9I9e&@U69!Bp-IA~G((qYXG{wi&hO&}
zECty7;?q%9*#|TBytY;sniz$%RY@!Z<wv971sv0@$N|Hsm_ur>R;@3Nhg3V9`)DlR
z*!C$kXeGB^m-Oc)@2B;sryXKc4mO1YkN6>j`p%9A$)U8ymuJx8;qJsMBJ(@`e*VL+
zUf)`6@?Y1IfBZ_0{ilC>(f{oySd)syE{{IL{olp1EMvPq!~NYAa)!tq9~;a{q)D2O
z>qfep9?q*4Hu=GPJ<ORl6`^3)S`fYvW$<39CNKFW#h_~k>L!_tp?@HI*o+}$*B<ru
zZ`$WlwOXRmmf6x>tuxzHV3`C{XsRK(O3TfKXwb3N(gq0^uh1(*mYUn79(Q;0&|7~M
zJzX1lxidLybV|`$KrJl`rJDC&!<ZAo&fH7@I)&x2i(NjGQp0yG5jHjy*<7d2fZwqU
zs3*>faYL-}-P5Jv?ul1%1MXL^H&fRw!M90C3C|oBIFDNMhg{pFK=LC#C5hW6|H!>b
zeZL{4Q*o*zj7TiDxGMG8P@nQ9Y_~OD3S;gpPJ9$@j%FUWzco#Y=Lnx;;NoNkO_*dG
z^V<QSJEna9>b!BjZo@t%ta0zM=%V-$-xYOlV#hz-5vCNfbi=4`+p39jjjKCMR(uQ2
zR9V!kg9JoJt%H;jCZ80xh87Wcb;Q|7Xyk5Sjy3JVc?fJGCFh8R_BxF2+}bjl5#nIH
zT|DyJ0YCB8Rn3weh$4~C4@hKFsjD;_km*Foki9+qD9J4inlaw`J?(Cszkj>2@6If4
z$s{7je#dXIqi<|U*a=~2l!b+8w0I32<3sz{WZwlT$6S{5x>zSTEg$fGU3s%;|A)_G
z<%`u=--g;Gy`YD$XX8#a^m(sOX+3#6)39!8W1k~qTsy~}@b&fg!Y3pp#rsF-gV~&s
zItby~9M=6>ZP_<yulRV6+_8zNIZMJl^Lv|f6WmEI9qr=<lqI86+C_=+P|XV`V?A{x
zW}%eWktWO5x8Dz@&h+M6ItTaU0_b0$=GXhAsLQpubvMG<v&oXOW(IgS8=9HTXup3g
zJ1vSQM~tJN?tpp&zN=RcOjtJtxdUPvgLqI}U-E$GiPhwN73HyeVpg7i5hKQGFxS8=
z3#aPixg>$zd2_9)er1Q9`iWi(P5nqQq3*HBiCU#USte8JMm}|>^`@{Un<n&>!NPQ-
z(SX)oRzi|_;*v^_J+H`j$6EqOT>oR&(Lq3ztN;x(oaF9-ajFmz^||DS1TFVr$#0p_
zWPqs~Z<0JlgI05PYST^zT7GJ@Wn;gN4K0iofUmOW+B1<kU7YvrP=pJ2@`_v0`pCoM
zF(V>O0jVU~K&|0c<MXqHMka-@OBT-`Yy|9Ikj04~P7DBy4z8s>E8J>`v7~`Cy-RWQ
z;H0N%*?X%`n8ua90uO%VEfU*fzEkO7P)V-wCmuVE$?TRZEeor(a;n#NQ`>C<CmLBg
z`9SI;C(GLa52ilqGPFmEqt_l+#KSF1zq-P>``^A#{L2^E&T@EXX7>@xI}Wib`ty9-
z^~}!LL{2t+&2J?OZ0}xbcq}lb$Pv-m*;#jGXvj_5I3$R8Eh)*{1?E2<nXp1Ta$_;z
zw*!~$@V9~OL$^B_e|>avZdV1m#x<JK=;4nyFHSCo>n=D1R(0P$cw~-V#oaIJqaKg4
zj)-cnqogFyPTfh?z-A3EbSzr*3jAVWOEIxPGBRqB5Rv88XVZ_4+v)7pS4Bj)l@MwT
z{Cbs1Bd6>j;5Q?J<q;KS)?Gq79)7gmHveX8`$N?huZO)jzd5XfOC%4CxKuqhSB3<n
zSR*bcCEvx_pGr$RW`k7@Fe_KS(5)9ts7>)gUf%bPyo@_RLj$dB!OSPT7KMmMg1(3K
zYlQlB9EDi6TW@ATddy#M5Th+HeR^*Z=wYd*S?Pa*m4UE+8N7F9@0>tW*A8xQbDY%4
zQsR4@2)Z~jC)%^@j=;PNONuTMRNpOYw9Ie@e6T`wY^?(zaFepGqMG@DB=T$aSku<x
z9qxzwQ>y7FJEo4L7<*t<Sgm&Hch|wYqOs|uc6)9j5&cE)tp>PI&CkL@l&H&^;(QqS
zz$*J#J4ODP7KpY_bG0$T*X)-;$26*(<JT8vlX+&!#-)U;-^W9Z7J3)wVIhq!v)D(U
zMG!1u1a<a!7>49MeCt?vGl3z6V2V&?^rU?V1fOqWj{<s&x;=VtwLmbA<HBu88ogX?
zHy!rdf%+ujj+V27Eo(2e8#TGg>2mbJ<iERv*w@mWa{QIjwy)N)n(nY_5464J_dd_s
zYp|)fpDhAQ)uNZm^#@a1`aO(Ao*%g4EDgBh+MCF`({>%357KvxYS7{vLsRS}zGCrU
z<h}GO1eoK4xsT+=AzvB9&|yao02B|%qaB~|sGIX6E3*r6lyacIy@@l!XQl)8>r9}S
zT=6hl6mD;|>for9KHk8>uIypopHzi(?~~`E4=K(s{&t}K6W}0~SH{`(?m8#3->Xb_
zy=T9#b#FN!N7miDvrZ-dcEGq8rPZ_eA6RsPWOWL^5NEAT2bzR0Kn2Nh%4QQEHalMd
zZpWttS74*_ivo;E%oZgL3S|f^_VotnqEt=vCPE+A-MTX~6;AJ8O9g)i(8}D=tc|sO
z<8Bx*`|x>zt7-S82_Bm3jz808e>-6Fe%%Q(EAi<CJTo)@Fq`s4r(P4U6uZ0@Rq2;5
zi*YI2{lec2k^cVKcLg|{t`NOfpWL$|xuh8x4hIx<@7O}6?YOc6%&LX0R;_q1UD>tL
z$>`3$2x=s9{GMV)ouvm7uc1ftix(E@d}e13%#k1PFniN|lkirWrh;YbKmiDq$L+Za
z7N&R)Uz7y=hoOMT7_(Z3>qbMfC{vDhOkS^tBYM)h3-?UCk8j(w<xOoQqsKNvwz5XE
zAr2I<drF#?S;)^c0Fg)Ic4IaqK&0Np_k<`ei6KRIm7q?=#qW_0MKJonzR!Rn6zC7=
zS@|qsE-T}R`^DHvp^%cityz_>)sc8SM_vq?Row^vLK0sZ{=p<|vF>jEUTvjW`Y4Kf
zKWB6KlbrUbn%vU_qXGc}X>fCesu;-a84xwbY}@X9gRVGzlFF0xIohYCSGq;-x$j+#
zk>p^YsbQ38@hrBkazIXh-`*Edk&G2Pnk`8+R#O$LgZZmVvS3}u!&OJK2+w~z;LmP>
zOMoCPYk9ru!aS9}W~F#w2Sx?qM`8(?*nt&fnu>qTXL&jV;v5cV;{B}lWV;N!msN^H
zxYH$JC#FdDQuY9gl9YXnV-^Knzg^)tgxvUjmIS1Y#!75A=j_Af`MN5j0!L;3klyI$
zJ=;}glmiI*dC&G90)aNo&(GBocS-S6z@}|-Vkc!E*tAIk(T>2TZDVnFT?8`7?%i9-
z{XpHTs-igdQ{~C6P0iGihewavfzMx*jzt@O;&^5jGX>Zy*Uo0jo}euIQp@*f*^-ty
z*?v^<K)RSk!2Nc>fb8%2bfyrez3tuP3+yJ(XnH-|+7u`&+klU9A)Z4l?VNr;?)Yt!
zoiY8u3FtWFdKGVdNWkoD`OrpaC)IgA>^(B#;Y>K6KMHt)(aiyy>51{t@z?Ej7)v6v
zF$!$Uv6UESeg;K<#@NTrmi>i#IjU=4K!Ce2i4DEK=<tzVhQ?VsuzXEtcBwwHw+;!>
zCDzoNm_)K$Od<zeU0vT&A8!uY^+KQf<vs+&@#!-wQPFEg`fQ_htNr6N?}z@(36Kq)
z=Tjg4P%N3XXPt`Hl!7UaW=DkA5D?;-DKiI(ojT99Cbb15pZ14)X=h}LE2(57=FuHZ
zZ8;;-^1&S4f`NLrHTUEaCVhSTQMP~TeaKxQ(kz$wI=8(2o<R{uRSjq#miI&^LW<c*
zZOgNrv1Dc9hJ0BHE$&VjrM843G+Q6Qrl-}yLB?iJ+QXV$#}|g%3f}+j?=!CQ!OawY
zBiHtFQO{jR$Eh$<8aA}~n`a@AZ}3HI>+^#B88aW+azV;+VCX%6l2SN%!W{Wc6`^`G
zAbk4>K&Z)oz2-Y1=7^6@JEV?GZYz$asHz`oKq5!;!`1gFAnh@S)F_C_r}$HjR&to2
z+Vej0D&wZfwUI0>HJbmu{iW_69<V=f2w*QsV()fN3S$x~Nc?Mit7LnNjR;}6H;6QV
zwJas}RiwhqqNGiy(!w;##h-TjF#X7k_HO<@?IE`n*R2p<#k9;xIt?#)6Mj*0#FN?D
zO@*@Qp9jL&LMHtJ?`8ZO<jv5$hWvh`whAE8NZW?M3szF|pG<uOgvd#b8UwNFRm4CC
zfT-jT>>-HEu1FRz99q*9H4Xi6OO`eD5r7+IS!YILmngSSgW}L|g9g}9=Cxm3=uH-a
zXDMN@NIGBgSg1G?9JFnG`kzws{vP$e9R*#UL>D`PpwR#q8lX%=={9s|nbvCe*5a@i
zOak6@tlys!qgT7?_kHLKa?O8x-GAfYaKq+LoAzryS&rI%+{k3|{)OfJ_OhCY_kjfy
zu_}tNWQpNRPmK1<*xtej9OzL+jza^0+IZL(F!~n8lQK7A2S0v-1TTet<)``cRydBQ
znezMk?STecjmGllw{#vZK=oo(tvKjEtR<{{`e6W`lG!u>hsQRrO}yZYh|AfnIH0<4
zBkGm^zTNT0V7@i#o3-)(#?}AVpDNaLo$0bzNVU_aa!JpY?enX{<{p%~`6T=rl4!5N
z4K2FkiMPn`<Sjh38@NNaaIsr99qnw!E#V^|If&4@Bj}e&a*SaF1vD^fr((fqmmNhB
zY;q4hWzR^CnV9)?S&xqmWa+#L74^oi@Zk3y>}P3cT^mC9^LMeG>4x?BPWdB@IdcIu
zyNpv}vKmBMF0FNDHz6fLm$#RfK!)JU&)+uzu(A*_J_*pIE-A|#6`~41XXrEY;BjOv
zRS2tmC?aZmvhUkat3Z%n?KU!E;l)NGA(F16dh;$b?~=Ko7f@*QaBYD#e49KJIgi4s
zcohGHB6+{Y_KGM-rzb;cUiZr*@7mA*VvitZxRc8Y8hG10I$RkY2)Lu=aDg(rV`4Jp
zu@K&$@=&lEyF+58wps4o<}!rRmMhypC#1V-nvxm7J$FGljz(jJI^#oHZ1Fm!mj+I;
zOj~EVRy}R)D}H%rj6q6qa9IO8YQ<hKU=pbF2DdNVuDWLG?VTXI5jse!?jlRCtrROr
z!bj!SdJ?0Oo4kyG^<}07!Dj!A@ywad?k~uJY0zABsab9*F+sn+=}zl1z^?6GFikY7
z8Skj2PqwccgkNH|vXoIw)5xe~At-@l;<hc6X-`^Ei{MvFB!$&7jetOg?STUhx!Ur@
z<ECwUnuk=L(iQ|e1t#JKkY?YaPGSF?(My*rOsZCHGuD4yBv@!2@q7kU|7s!`x<!??
z@opzCIRs9w8JT&6|2$`9kt{oY;glud^Gh^|Tv~<&HB1=+F2AVRp8E&Se-D)Z+Vyh_
z9}{rCkBfCL_{GmrGY{IAwuDk5vXsw*b%C9x7#BP86b2EM0v)CRpYG|lZ+JCi$Ro<`
zumH3)q4q8zJllxUJJo-Rn^Kw^Ipj+_9EqxTQ&w7AGp9c)QZjrQu;F*~*n&}O1efak
z{zJA-`6*s^>I=<WBy5LwXk<`}h9XV_`TAYzD#zs+e^<fGBQ2)KS`fCF7C?&|m4`3^
z+ruU0UUgB`RiO&~1@ibrJH3!ZXWLF`z?CI_b*I`>wt#z7N80kJ*Q6%Ql1niQCS|44
z4b#<9w^%&}SO=v(u<R@AdM!zt_$qte29=m|oKzV9O`ZFR7z)^J9;ctM>#}s(xG%_j
z;CMK+t?b%)##=Xkg*tUT={z60lKEDb2VJ<&O-%no%0^|~NSjt^&pnw|v`m~5d(sSo
zxB%qp@4|&-d%VsQ!$_}06LGaOsBUTF#!}mgL@S57Lk}nE>9?;<%r#5U%7FVjuSsB#
zjHdV0xuq!c6|Fm|#c9L+;Q#t<;om;f-eqOwLSW&7haLikriS2*j6eHuKd~(GwzRw#
z-bU{1B)3&n9)*D;Pb~Yiwc$J~RZrXoVEom2ezozimDT9gTMWi)z;Y<u(jhfEI>m>E
z2??=SLCYOG8Oxk7T^In`FtU_w*%H8z4`x}C*}LI-56DX3Ohy2S59)G+|7Ul_zcAYy
zT-l>0pH<BYTFxqBZZd5uk#1GCyyz!7TCdr(QfkeKDKD}lz+9@3^x*Yp=1;Mn&?Im7
z3vVGG9;WC#*nbqAv~_M3aVo6!*-wPl@#i1aK7tQD_gB?yl;ZlP$G1cB8In%<-xHj}
zdY4!lq>Nymlr!m)D4R^lN2iB%ZM<Vo822s@hV{P-qaCF{wu5CD%b0dn3%e2p)g!^-
z0APt0MunT0I9|)OQ<@?6<O1zC$iBX-qCQn7$+hOf$-f0#9y@Z6f`3Ysg`l?r<ects
zcU(mVzKn#y%ZD4A5rg}2qFhI%tfprdf;Noc#s<UB-4)$*Y{D=@h&dpv#qSo|*VXDb
zD@eCj?Q`)8ukYF?E*cKTmA7jH(t$p{*feTACh9hZ8~pmdrVB4vcT|b*&kj3-QY|aD
z+x3_FC&xvttu6N80OR7W)zX%Q?t$LiLK%<(QhK}x>Iw@6;r?UKvcs$Ko?cx=(rfF9
zsF<CTVCvoZ$)fjHk%?7a<2c;;DAtJ2mRHs6sYiKi>J8TU8(8Po1YEj;1<e}b5${=B
zJ#~WDDZnVCHH*@J5Me{)58H%1SPJk)VC<cKm^BN-E8?TlR(zCst3X$YjWctnI<B*5
zHJJ1+`)1>1+e6n`OM&6vWl&ELxvY(V+Fos`ULM_ls8_~26OBIMUB(ugDDI@3qzrGA
zBR2Bv>i_sZV*5LY_?0*T#ye99ql{cw=$l5gpkdIpJ&m@h{v-D;5d9kFYw|{gK6%$H
zdXKxH2qZ#Zm~PEVhMivcpC-au-+AEs(d44;gTEb6?8Nk_yezRhn9QE~6j)TYXZ%84
zx0BW!K2u)?ucv(qB0N`i566cM6vM6<uT|m#xW!7R2+#9~Luv`j;oraJNo8A0)|&#U
zS!$zJB!8xLT=e!fgLda}YvejGDHMg!2l@llG&w_G*R;6uwxYYUNe<7^HeW&BM}mok
zmHjJr-h@&BH)e2uXnOsn-|Z$=t8R9XAq-Y#)_=IV3~6!u0pZS#`YVX@ZUi65s=8j(
z$F?dOQ{vzr{^MfX8U&m<e=qLJ=S3Upc%nt(-P@0(qeceIQRPSdE^c*s1(fKk)b)ib
zmKLeyIr+XWaiKP<Y}Zy&zt#_sFx^Z;I@3DXy7tr(OrLIWaQM9DK*Rrt@qq(}|GZhj
z|8Wl3I~sOPUFK?+%*DK1J9Wp0lPmx~@)^8CA~8ZAF7W#lAxtbhq_B9%%7y~Lp#pR#
z&gKt5YR$|nh>0VI_ipcfyKVe9SH(E!<Nho|c5B8Qt95h&dt!EE)!Jgo?EatDntm$k
zZ^Akbwb~#<Xz%Va3E>)vTOr08WvkNg<znsO#Cj}2m?Y&5K@*v1@?^354ry^cmM3{K
zYAo{GUSRYP6eOc)E0V=Mv+1GGS-D^NjdcTDoAznay6t3atDsCYaKEh#hhv<iS=eQP
zUn+&xAhHKSVybaLC2!`|)=b+{3MoTYdf_lsG~}T}IGK87t*bn-F{CCbS02@vn06?V
zU4c=07xE;uH~A6q!g#YM{65?Trc~GRyt}U0vP#q#0QN5x!se-YqvcQL0pcf`Id6^Y
z1b3(<&*~n}wiJS$rh#3PLKGsFty_9`4hK|kz#;w4ycXX%#<{J(4eY-M|DPNk1f`8&
z+*%wQ`BAF{Tbn+<Je$>AWXuph*E{lhQIA>lfKq*^MD<~{0EzoF_N!ozyne=dBJvWY
z(KJ_601$0lE3@^!R{STw`QJSB#kN#MYa^wu9oCuD=N~ru4ken(f0n_xSE=_!0o#(2
zZy3L*dXlTz{g6O;KW`lQ#;-Py{hmqi?LRA&rhe;Y{-vNwVgH@nL=$AWb8QBbs^Q^N
z<4^@t7z@;MFbpI)>Q&lhq0%Df)Q^ubJE}|@nkCQ(<Ut@E7urE;&#U?_eze&#sV(x^
zpXD|9L-*ezLHNEDa4k$7K`nFF9R1=&gl(9jR%4`;sodvrxzX(S!`>eVuU@s*d`&ag
z2Y<-(d+Fhc^SJ4sPrsP@=YR8F|G%CFeTV#D9`geuPHJ^_iQOfu4PRSmp)Hpe9Xi##
z(hj>rrE}*lO0r<&qjjNW-_rHEp4mr?75Nv;UJwkJJ2X-z?BqQc%*lstF51Mj5AoYW
zEHh1B3+z&S^sq&aEll3<KxX){!L(>f=@~=6GPZ7u39x801UZ3?FsJMA<hrUZE}N4a
zySg%uU+z*T5DcZWwD7mGkY!*elk8C1r8FmLMs+)r&IcTC<DxjF;+AB5`M4NmnBd)z
ze<i>>@mjnZRrAfox`bG7zv!J#`{6vg+7s6H?zM3Iu?FN9B3#l?jcUaqyt;<$@<;xz
zi4W=S$t~-y4QvKLN?DKHsWkekc@TAXrkZ0MIQfW}iW_ptqgzl-gl#FHSbOv!ciAps
znz%^A=5UEod){Ezw}ZiSh4OeOKB3lu+}0fn;5tvBP5dBXn?Le(KCaSSuh7S00yyWS
zFa}|We~=tID3wiCo-fpYBZ$g8nHCw=d(HbqE#9#9H9YgZ*E91%qu8y}_OEh%I*j|M
zZ|E;nEhcADBbHJuWMj1uJ91odhn`NR!-~t$*UYu1w?1)<@;)0{-#l-u>nb5OYObX9
zEoz=(xUmQF>uqA%xRpq0@qAys;@6Si526<Kl6c`KJkTZq{LHgHZ`p-HxnsX>X?2p_
z#+b<84!G@JE5zyC0W)*(UkEDh_%pIU<d33~KzN;+Y#l(lA)I89u<ly+LYJ?v7~Fri
zcV!R6KN;evvr=5`pS*T$AlFoyGX@HaOfE;f&gi7HM7jN<8{8H<N5C%mmTvMl?={Z0
zz+>d<#s!054roXvPs9vMNlU=`8+c5YyMCWm(6fDXqH$ft<Tk&ZN1NLWyfxcKi%v+&
z!J{5cAz&t<{sqN1KFts+;rf8&Yq$-&+|t?3A!c0|I=E0>7`fD*Jlxh?q=00JULduH
zw=Dr*D`yKEd2h1$l>pzwrjL^=m1a&3#!;62?E`UTCG8M*6L4RObsIz*gw}piRAWOs
z*L6XdU`zV5MtxyF+A63&v+)xGhs$O4lKbbY-}-AScXheIKJGXzKsaD?TE?iNZE0%D
z<xbn^>F1+EPiCU)?)oT&)UE&A{ICjKoaC6vmO$+eFI_4eW2uEs)||Yx2A?nT!{7Cf
z&S=4T*Igi_T`sD2xc&y;=%@trU0sR^?S2v16S=2Cmq=3$?n^2h-Dej_<<r>#mg=nC
z{b)pKwtA*b(|bW{11&raw&g5Wx3qI}sjXh!O6@8~go+cCr&%KGd9XE7C6I7&gKLp{
zux_acz%?sH7P#$n^M4mo{PXSQD`CBN;gOpL-k&Y4Zv%Emgz%!u^Ec`SPA%W4^StxM
zPrv^HfNrLpbh+>f@caDcL9ndGIB%Pp`#<#Xkeyb50K{ziiyj{6W^Nt)-U;)~Ug?v!
zgM;4#n1Ro(M}VI-eOf^)dtX8Pf;F2n>*9mK6CqXGjuCYpKk<ui^3^HteGTfW(WY|^
z%YBv1=k?GZ$`svti&V^KJ#_M)IX?dxuYbiQ0y}!n3rC1j04zJd9v#hA_rdX+_)4S~
zcD}LeTaC}(O=me>VbDDijbO#Benkb2$i&~UzCN!D$yDoY?nQ8mvGQ`VpzTNmc{Kq-
zXff#tAvDER6sftUdG}xAM+d@~;lCZw9T}e0y^2v|qLnZm-@DTSC*V$h{QI3aP&gso
zEa{%kh8TBJvRK-o{kt&-)BkMwu(Tz+RL*`>D0+%bc!wB3Jr435h~)wqorzl}e60Iy
z;20clFy#jFxm`wu{A$d;#xKS3L^~iY%Eft5y5}@P705ZcQ=eDqfr09E;+-zd7V3O&
zL}Qx#{YmO6iD3z$Nsz+qI)>y2f{ZL}Y|=kv*7{jj0ANoC{OVldEBajEg;sv5M!+_}
z|6sHts3gLZQCD-m^6PpOkaLe3FS|w1{zm`rX0->tKEN;AguO|64(V8o`8{=-&^WO?
z-f^#fxZabc@Nn_MLCXi9eWES4ua307nv6|sohOba?k>%d>MuE~66pyZhmV0my!sFQ
z+D$b+y(#IR-V%-lR<8RDH$r1rRHTp5Isi4@D`I*i#(*EQ@ZBK#SA?R`;)-dDnF;+o
z*1WjFd#ZBHD6AjZv~)#t{wFT|C|a%?u;5iNF$=IpWS^YvO%NG_4BR>s9yQi777Q#N
zF(+DsBz_5Zj*}zQGF8{ea`)e<4#C|f6E5I*@?i|3YXVcx=Kg^1SdKycu_9YRr9nA>
z@Vun3HJ0~Hm0xw;+3*_sd|LCNFUGCi8E^O$G)~Z=zWCMPa?sQaBFz}4yrK0jBD?`C
zH`K^H3RMp9L1WAOQ;AugJPwob$c0o#ct$99NrGC|@`&J|aprh$l;5lQYB2#=y<V=r
zgBiAvQXivDE=rPKt^9q=e0jao)ZSBq>B-9LGC1nM+x~So(V30a-(;-qjWw8hzVY`K
zu60BshpRG3-65+>{qDJ!Djoks_u~xrtlls>_d_pldd8|Z^NrqR@w4S_gS0l}&Wj48
zM*jpk;eCj46Oas@&sC|U<{IP|td!fVS@z>#sk1q>V%?on`I@ff$ySY4SU^@edGzO<
zpq=%*WIqG@Vnx)Tm)`U>_E<S0Czb7aM*FN&i-SRc-wdPCZe}>CK~_GUJIR+L(6|b1
zerr-tzq6^Ed6l=MGT!1+;VA*KaU!z}vHl7SQbmp}&6y-YAorFz&~YZE7aAf^LG~bV
zOOD7baEE|$Xv|&S@$k4cdG{B?MI*G%vpiNK-lPY98J6{QCaFG`$>j9N3gcFyLSDDy
zTW}dk+v`1VMY+b%l4)W+0IJb+yc`c(c=Wg*N5~rF9PHCC$u3!w>XNFbl$KEO%UsG|
z#YTUR_dh@S4-#7M{(1{wOlf05OYE}pUsE}pcE@tREvEg{JgF}*v^g>H=Lu_4Oh<M;
z?K`)*C&?+72`QD<C|UUJw=3lrnt35j6DhiPOtlw=zL_6tiz}Lcld?i3J1PtCTJy`(
zmNIQP^+w-9lgijUfG2%^cDRbStC0@#%=k{ew2#%<vuNVujR2MaB2xi#{Z;I(e}0+&
z`cb>1te6^6JW1_dhq0fvN@4rxorlifLPEqW=b@;riz5MUry?Lxr2(mxt{{B2@Ii=|
zVH0$QUl95r4435Yi}-0G^`OXX_xk&!l4xo-7mMF6{=O_y#fp2g&re-TFTThbJw^ko
zp_=??PiLaPMHVbyf~EA<`V`cU);r@Uyr%DloIiGwzU$QQI6X~jbW&g#9b~r3igvVg
z?Ly~)G=3-71`V&-R%GQnyldZhGUP=K<X$`4Q8x5tt-J@ZLuRJYIC1IQfpb^?x3_`+
z>!AHFp82wnXR~kNSiVe#J9j)Lrq5}d+OXgu77P=27I+3;WvD3Z$xrQI|0=JDd*Ma6
zs>P;>v23d&%}f?Q;d$M){5Fo+Yf;MHp){)g8>w)`<2X>&R9d;8zqRsQkPM=vYslIY
z?81XaX;7>kk`})i06|Ko_08vqOX>_l>?U82>EXS<YZvHLzbYPSn1Ytg--Sv|hb8US
zYZnSJpPcDs(Tl*U`Xx^BzPgNQ3XZHjHmH5>Z*N*2Xm;K>odgRC`U@u~9^mA>2RJ$9
zWPp>?`~kq(6Q5n!uII9F_64}LrE3nKr+>iIHR9RVn~U_!zjAU2+W=&o4sgT^dhg2}
zyHQaZA2yH|T*42j_LrDKhMhfinI->-$%z~(s0m~C!(1;%IKOpeGl8YR=XJpMPKEC}
z)jEG4v)!3NVCRY=wr&qhc!XBX<aVr{(C0d-BpBDOXtl-5nKra=_G+Td&rJp;FE1;l
zx#JHTSvi}|oufDUeO#PHuPODsUlw%zoMJC5HLR&3t0%>7+vW5}U2o4w9ZyVle@eR6
z=+QBcPSe<@XGH8&q^$-hErN9rt|Ryn%w_z2-R|4#$fi>ftMtCRHY>`KL>bee(MO$%
zF7@L_`74)-JhGyh#)^0_@>OlYt-poxz=5w{@HIDcSCSqT9VEg+S~FGS*Nv5qrPU1C
zc7YxSbaxQ*u)hXbajhsV#3>7;h2Dd6YzTK^(d1j?9IHiKW~BnSCLNkinss6rXSiP}
zQ-`EaK(T5l3}S%-nis`+xl9!?Dl7NNloRygMuEgY6JCB$eJ^c|LYxy%a%Ncr_-NEd
zt@<I<Ynv99Al(YB)#6jitagVgJr)H`_AU@6nUDE1B!XEm+T=dyJV(d$2M)RFR3-5f
zC=Exl<gcM{M`om!{u`zdP^DM=M5=0jAJ{)`agOi|8UTR){A5frY5Ze6fkHzv6IBvD
zuOYmGp?~Z$pqvzE&(;+`0}2NyO~Yh*|5glYj$cX2dcqold8kEAm<Y;z!$PJ2<d`~I
z%OlM;e4?i+{P#hb@U5Metp*_kdWAVckd^H<V#$sg82Wn1BA^B{<*dveWLylxC|cAk
zDIm~WdDs6T*c+tG#YgFO$;olsn%~VdJczeVjKk$%&+(0dMKeXu-i@n8<|D3m6jc!R
zuT1a@UUF|REqB!-+mGomB3ofUyuXESvA)m{FU}W-ZV6|?YoHAxnJ=I_kcQL7P*u5I
z2Est-(}AnP6tf8G&bnUY7}Xv8aGVBo1DMzF=f>tNwg4_At2(I>_KV;trSG<+qrJ9p
zE_smD8o4puKpz>_O!MxuE_d%@7}#87hv}LIbp_a%$1i0jwCPbgt)48nADcLx8tT~g
zinFy>!4&OcG@^t*6cLzBZNd6XoSGdQ>f?}c2{0teUdz_$4}#;PmM=AQP>^e%tX`zc
zJ1oXON8axpBKD?*Q4iCWuef$<h0lbX9aS-Y6Emfu4LT~uC0aE5Xl`gVnbmjW?>nTn
zIM`xhrpXvZbkQ`?VGDDD&FEii<YiFNmMjj?3f0WtE`zj0`PI$ps`2vYISC~<KK?_%
z|Mg*x);GT$_?nlMjEQW0C99q>2*_A_{Fd=HqGGC=Rr5mVo)vB3@?<M|v%H2-GiW-F
z0@stt?^TPIa@ehH(7*twM<XrOYv$I8%~h2qh4)OvKvv_+9XOh(5i=9XFuWM}(+81*
zm0f0$!h>`E0nrt)Mms@GZsuL_-tRm|E$&=9`R1BQ;c0oqH%m1h!K1zI&iMm+7WY66
zd35vUo5h9gysI~g6a+=dE{=ZlRnQxtbk`3Lg0`Ov?~A1_a3wg&>U)rMZ$)DM#a0pT
z{8UWLqttB)OK|6_CBx2}UEIlUc2|V>xetd%z{&Ng?H#9gSc}m9Kg(AW0R%z(z3(>g
zL*4m*)VdZ8AP5#dWlljW({=v@%Hsg!1KO~BSGVoH+ksRa(i0M={1>g%w1mB~1ONBW
z{@)<~pEsoB9ko_{lB)xJ9<<9Pe#)@rX+k<i1aJo)BU@Xx%)|Ic4A=+lUaX6~q6L0$
zB6&?y#WbmLQ3GY)!UJh5p!RU{k-f{Q9)9x<BMSvh*<EcPS_f}!sro3o7*~WA)vUGZ
zWuN|<c2@h;pTGU;E$^*LV3Qg9#-h6@{)+6ZZXRj{qBn!ndNga^bvMUZnZ!v;DwCN!
zu%F9Y#y#NS&)v*vs7hP@!dxRp6q=%joZ7S+>%d10p)Ank?I_t(c6?_GpSSY<qIm6F
zQJ2-D6(sgSc-?y2a&{#0xqJG}lP=CbR2)B)o4oTRAv-O<m#tgg2UNJOj=o*Cm1}x1
zx}AOd>4WAz*R-!7P1NA1>aG5>UF9(6yITVn>^o~7d?C8^ukWw_e><bkpT*94ZBCWT
zJ>5PEbizF7vtqMuwHF6a_L-RscBr`g(t!UBLC#>hTho*Xe7<}04oc>73Ov&lSRY*u
zZ<+dNx3K^H_71a4Mf75~+h7TQW&UEigKQ|;XnuT4=pz{e<}~mOQGYjTuPYp`^?nZ%
z&=k&udds?zOEPG<u-g$@0jzU(9CPCMWqvTnj=cy<)3v>+AqEw+QE7oSv%dQBZ$bO=
zdt2yq1h;E@u3_cw)Je!po1o@xYn(k$ydjUZku}^Er3N%uB|a3g6V*X1GaZP1GGYQY
z@?=LAyoC6%!5;i?uXqG@8Oe@`kqkPzQ^ZxPENiQkQq&sus$irjHq4TGs%;O!U)daw
zd+R#GT@;CN_FWv-!IX($6L!Zkq#|)>9!XcCf;F=_eePh?R6W(xUW|!>c>?}E;XA%t
z{MFm|k{y6?6}QGyPdn@o@+qnfa%t#?XG5uILXFPby?Wxzi;=fWB;n{5tERUs$G^C!
zjNb7e<VqmsG}NHYrwI0fJ^xC)`Y^@N{!q#fdnjoOUom>0@GNg1gA6Pa^=jB<aQjn-
zlKe@+Tv>Nn&Bd^J%D(nlzO8uM+BN><bE7KJ5Yb_GVCdnJJY*?dFWub93@{>}Dj>%8
z#zsOyR@7q}Xe)S&J?$`2Kf$qbt^Y}J#dY(my5v9pEzqA2wok1dJQUe;n?S=HeKfwk
zN~?)V3;C+5Z(@FP0(?Xug-zUULL9MDwqh36NXv+<8|R1XjA{1ymy9+v`s3_DBE0_F
zjdj4#CO0x_Db7BTnQFM^kR89MnHormDj`}3Lohhg>3$bx8llxC%kr3=bu(-=sJH39
zVOZ(iFe)H68kx!g+_05vRvc@Ij~~k}{^&ljYeP|6J-*-)KI2GJxw!+|NcFxG+k4b?
zN`(tyW{2>7QlMq{OS!g%#^>X%_kH5%K7EpR)#Kzi-5X0OKGT(xik<ofRDO#wV~3ri
z`-qPxQ*-rNZ_}P}Vqv38f>)x+Be(krC`s!S>{V>a{yc7t-{kKzVF_+Q9v^Kd&rQti
zr-QlOr6nJ5@v+QKpiq=%XIOu7#_?-vO%wKsHhlqsIlmu|X>pnPqHo9MXq8^)OzxXo
zcW9nbQ5Lw4j7$=jb8Y^l+s~|a)d0EfkR<cdr)mh)7)gJ~*O&ZaYU`}XxAR9g2tL!~
z)@f!h_f?Xyn~hp;NF>5tMK5^TB$PC(%EkCrIpvM_Xr{PFSAB(U^(c2g&vY2j9!zme
zN<Dg;bSllPqDRpz*FyB%C6a-j`Zy)A)`yGCxioRw%AoklmF?-$8HGL}D`R35)OY<^
zgt>q_cQ?h47~>mx5qt7YWT6={(a|xp=s}45$jX~t?v2cXh_P(l2Vq4Ts2TEfeJe|A
z4jSlivCDNB9L$`vOF#a4<97M?uAXJ;3v>;#jlC<XX{d7!&ouNc8@x=O+}i2n?{9Q&
zgOF!hOkf{p6BC=9(4#hlh;kJ)1D;My|9;@d^^VbTpyUF#{;K@Kf^bV3lU$E_rgbA^
z{%gd|U;DMJsS5(>ej4m?76a-cq*iEF-@7m@mh^IMMOXQI_vWJa($c;9NYKT)-7dvq
ztcV*kpT5{5_eVyD&LoA6qZoR;2rQ$&SwNhviOv1k<%$cEN66Pz&FV->GeAr<{cBom
zeEgLMKT8(sE`Sm&2KK%~9{SbGP=fK7otq6m;X0+sdg0(%Znh>|Xl3l->}>E`m>7Fs
zeBS8RlR=rrT3RVbw9tfOkmca_#E-GZ-?^pd!grEuVEgZxkI<tRH>mQvjy%1Jrling
z=p)nC>4y=YgcCNXbjGVhW*6T0Mht-#xf?U}qOYQ9`-&fcd9~gX3@qB~<tfm~snzL+
zE3M=ZjE#!Sq`8ZiJ7ZG~aDu&z0Kgaeil$c~YZ^H(8`ZR@T5U;PW%jPjPM=>uzr-bB
z2WKwwec^zgw0iA2QI)kDor+qEp^9JLwSB!x;#=aALS7|~R>pI&knB6GI%<Uny8c<D
zmp9-mM=VXH>-*>s7n>M0*+GjB`##y6(!S8*f;x$;&i-;6widr0(4~=UHH`5{UVoO$
zPf1vv90;q-DtiLnIn~-`XdAiNnaFh9pJJ~fnOUyxUrU`g6Q(jjX8v3B2}uRFfcm&n
z(@cE{=1QFnAS3j2nNjfz=@qjhovC_L;+}Sm<;}y(P=~v^#$GF$Kh`Gnl(ex_5OUTN
zn~6zzL8PzUPBn%z+S5eU62tqE#D10_Z#21|yO1agTZk~a<2?`@+Qo{qG@D*oQ{OHh
z#a~a9V7yUU4$J7PYv5$P)>!q76ieQ;?wPjwfvpCm9EjDUfjYFX<*sT;YqC4|1X$S>
ziE6?3uiQ)>7zQ=b?zU}rtQTRwc+VHF?HtdxAiZbrMcZ;~7UIGu8m~!x;&qzWV4U7x
z;%2fw_FsyLoObtI>DFDVa2S+^g<P5oG}iXgz_8-YL-Q#5ozFbk(!S+vMpW~7R87Mc
zesiKTu$O0Gl8cA`+&9T(`_y<xXGyhEmqE4LAT{!qdy-RALT-{b*&eR<90k^xT?t_C
zo`j#EV<~k-Bqr|zhngAh1k((RZm@2drJPZY+jG)>9j9jRV^~-0lJ0eOmUiDT#nodL
zJTkmc8haa06t@7ykWgymlm-s`eOr^S{fE!@`^16T%8^A#R=km3X`69KxyIAy>rsg`
z4;_o%p~`n<g9Zcs8{XWe6ZIb+?t{7uMkjbb)WHkiK%}P0;9WJTr;i^#K=$?4&Wz&o
zm@ViJZEb8ucXvp}LPChLQG7$(Xm0Gsd9ed$w%>NwdcTl<Uned?aMFQYF5OjSRYaD*
zOyq^v3dY-G)Pf`ObXzBw{R?a|d;e28EIRlxI@M4-!Z#^CzF-<eF;*~C?*!Rjw$^*=
z^k@mb7np8zH1XHrX2Lw^%pior<Rov{CAh=&VQ;RTOKE>Oe}1^Ft|3O*d#u6Ca_c2$
z5<Jtww8>UB_UV2M&2apo>lCU+Iu*mL`tDD2aZzE*8lx^y9cu`kI@5}%h=<+Y!wg1x
zit7jNmY;&RJTKbn&(dAgXsUS^O?OB0#4+2weLHID`J@beh*Q=8a)_A$bD9R7>9zx(
z?ZrYP3!j8FxJ3sb>Wbw+``SMdKS%oi=r5&9d>a3S^yGE66g(E)to1#)%qLkce&INB
zXUIN2+TU+YU?2N@ZY^9|xj$1d%3dQMukrBCxrA(U4fTyFG>*Bq+9$(6znE&~Rya29
zw{A}Doe%EPbQxWL+W&FCPUsn;i@&qDZRWGW<og?xXUv-UP;d30nBr7#-h7=^kppUy
zk|eg4$YP%ffTgCKb!!0;ptAp`hW!|KM-QEFsqOFr*>A;h*>-T6>WC`mc;-Jv*AJOF
z4Vg$5${g)OPRf$|99tIXHvlU<w1R@>UpO=S>HF!!3QnIvGhc1Ak+Kkn$`SI{XR6LK
zOgY`B;ePc|a=3<|bErK)AfIpwFyuP5fW5>VVm>IuHzN#|nPRTHC3{0Jzk@(!2XIGf
zchli3frlD|4%D`I)s#d2D{LEtCbVd(U$9NExck<rl&{?X8M#{9Ra;9jjgE;9?WwEF
zv>b}^#RCwE$#p2oCwn%bc^w`T+IT!V%3=pVTBN`B_ar<9+8@hidvl2Zp{geO&Iga4
zjbAzS6xa%F2s1upI_v&6UQfr*mls#t-wv~u6(4P6Vb3z6@9~W5id%B_$&Xme%(^3-
zE0)$c^qm10P?&&_Q+P~S{`a`dw;1;Cmlt1F)dnv*TwJTHA2rHh#IN33+Jhc?zsk4U
ztFrUgFRxf5jL>(|Jd*Nif!jb1p2AOYFX+ovjSBC72Nx*be=Y^kX>v!{ojV#-VJGRd
zGc<3U!>a8wL_87(y4%LFrtT{rE`2;f6$n@7(d7{}2AOi2rxxj9A<xx|lTu5jY4{@B
zjr@+BYcGOcO=P8A{tz8G_WzLg?%`}^`~P?Mo;~euJFPKVv`*hvF-oe6I7Q5EXQ=ZG
ztwTbzf*^=Pkr3>e9-O76&Lq24j1s9s;vAz6X-cGoggBKr$0<Ub&F{9K=h^eUuIG97
z@AvoquFGLva$T-<uY29=UTb~cpZ6=(H_a^s-kGC-$CnWXKNWw7<YqXEi!CMqx0+in
zafOE!q_1OL&eL<SQ+t^5riR$+%$OKU9oJ=Cxu^6p@eZZ$#y93*XydJj!bWmDn-1^g
zgJ!}#rD6@AQ#V9#2<gj=iP(l}pQ@uQ-#B`l{R|u}I_fqocO|@WR(-E!e0BuB%)QY%
z`Kh+n{Zf;gQ@-B_oBm*X0^HhEB1LrNQGc6mwBU1FK8c@bc3eTEs;8RWpUu|IcZ=uJ
z0h=ysuxRQqXYZ#asV|8Au53he8!yKGD`(O;zGW<Z{->TC#W?69mk}>Lr_0AX2hq1X
zq8>8PadyiP@VF|wCMslAv|Sf}?xtX-!i_X^38eB~=(AR9H*E*3{61y=y(T2v|CO4}
z4JUdT=<aRZC+ZoOKzCFiYWXS<`i7Op#)dfCb@#oT!zq6GdUd|0vBqGmTW(VR{L*Df
zzAa-4(<F}3)l+%09~5wi@9f3q{2ZSE1cTj3fgyR2xco4eMAGzbe#5-*0BYW{=Mg@V
zB9iLuE*yz%tISU`F4hj{(EHHbwH57%>liw%DRhSuiyhxwnP1E{)L&%jUpZZ2r5)?I
z>t4oZuJ;tnrQMJjN)PClS>7kO3ipOrFq9-}BVwE}@iOlqNS^FR(7OO6BTEd9lufyw
zDu!rE07K$w8XqP}JcO!xrm@^epSQn{uSvifkw)wr(qgpw!dGTb2fZN{hMCOEh+c|$
z<ZH;k4rTX5j)6@ct(0afhjGe1ziWPiKXh_RC>!toZCi(?poko9U)+A3rva(}jQ>n?
z0?DC*y140)9c7smYE9=6xC+X6hGZnx?QZ|z*^M^<q`qRa(yP^iNlJI9SkIg;30grV
zzBx+F8)zfF>9r)t=*Obhjn@T2vpG}K#3>UzwGnAH#vM(7c)wyAhPjl;JJ@ezmxueO
z3(z)u1|?5p0@Ko8pr6-~wKGRO9Mg==ZdwjG)O|6XFoaidk6eqZxBV<R6y-eGTyv3t
zo(-?bz^B@DQ@}ib7A4wN&{I*(p5wBt@@G4+)Yrj0&&`-u5&KA1N@BVLMkT3=XOOw^
zw6%XL#e!NvAdsiyO&T@J;9R_2kg{qg4L<s*?$zyqfu{;^IP3y>g-l+wy8X1Z^^#hd
z#nUtTvl&H~^~<e5CXdQ{v|r9-yZIcOd0k0&Z`7M%7JzN$SozyMIXB>=>b3xM3WLwn
z!4-JZZEX4fm!DT2P#TAh5p~pLZuP}uJ1>Ayrv*Gq$ACKM3zfTj=20F0pO#*Kk|(Mw
z%N2sAjk`TH0w1kBB4P$b^<g5<Tf?qaaIGg~{Q!4n<;p#OJL@XeD{mcCe#x9N)z8B#
zSc~^j$O9WvG2S0+be@{0#O<!PjzYJ(Mh(qB85dQ8SC!;YzQdclXTEp2*oaV3o8Fvm
zR;Hcl3c(;5M)kT#Mj3N@m{$rY)!9GBnoOInDQ!ruHsQNzrj`mljxUUFvh(WD%Up<k
zch(vd^Nh3F<xg{$Etxwyn2#|Vnh022(^<x=_U%38mNlV=%vR1k-x~W2PQ4gSEuWUK
z%=K2w;q%ng&bsR?<nve8vNWAbBGQ<nNU{7uI+%@Wvx;?Jh^v1@)M<;3Vol9qm96Q$
zS)^cPo5UI2f$Y(###Z^(jnFY#R*fUrpvJBoAJsuE&p}zDT+op7%O2r5OJk*MDOSKJ
z9P^jX&cmPt<hfF$a$glj+MU0Px8<neVd@})TIg)4ZG>=ow&mvW3MQNlH`VDKx0)SC
z6*Optws4}zzV)TBLA-EU4>(@&V`mOYD!&&q0fMR6@uxBywWQXZ{!8UQ((B4mSvBuI
zz(z^;4mB6(wVQO8f9xf+PtBWbYu&BmXr#up1WQa12FAfEkypb-W3)61dTI5+!9t2R
zkL|vPlNi+Oo5jBi8%Mt6%^3w%ECr+4tHt7KRF|GiCLDEvT@R+aCZ^_+fw)MkZ=<FS
z?}xUpJTDww5-AF>_qrT>UWr$WQ)v|+A%;=$XL7O;PL8zl`b7g_xAneyLOjmVIt*c@
zJ#sE`WBQqjSBZ5=x5kxxuIhBp)Y(?NP9({92DJ<=42R?V2V?q_aHUFw60ysZ`D#H5
za(W*a6?|^He_F7^Nl3w%&zfyaev?cGhwW%Y+<!H%QfpH%#BiIz7I?B2cbp@s>JuQ6
z^>!{HijYT%agoI)(X{hQbfx9~A1e9()TQE=UXYL$zhAf5O4SiEZ=21okNtikr=}ux
z*52m*7}MEargQSsCXW}f|53(H+ZZsuci9v+Ia$P>d|z3AL2!c?igNl>w&&lae%2<I
zt%MS((e0g+Dy4gFi}jqnDv6fk_;$FbMt&GR`>Pi+`%;4}2mbm5B+@L|UJExRLQ&L|
z%>eC~ks@xd4K|!Kw=od9;r>hkckm*8NMv*q0VTbx;l|Y`0m*ufct%=*W5`dzt<gM#
z3imN?MQCWKb4m1;aF9q2*m}!006azSm+Nx|D;#m9K<>(1K5jl=&rZHuH|611SkW{9
zq0xQ>koU9*%8-a-`!VY-Gyy=2UnudqN+o+~!8(tb%XeaI0fOf-dk<XK3zXfwJW}N@
zsj~ObH-7ejg!v5^C@S|~FYG6%ufo{MB1wJDSX2<h;2|YoulwXc)g4eF%Uj5y)i8!(
zKRw|Ll3|tWPki_C@4b1lRl&ATGhVV0L@KwxkAO;LFSlmC9GEgXFLei*F}oXQX=YH@
zaCpu(60kt6Ex2~vIxRhEWlW5%lM*Ilt5Wi4lh$~S9IK>+Zw6ps{*M8d5Ou%=gbC$4
zW)UiTdw)N1!Lq=Xd)4Q)&f~QwaYa*WjGC}hSK-s(KHl06>|VXc>r_ooL4)92)EPrP
z5iKd|5>gwqct7LiXfOk_P+mK;GE-)OeNer_cS385sG?VWc$-Lc1idyV>e*&Ed2QlA
zN;-3rLggF2V~JeqB#Kb4xS&Ne<6PJ5woB3=6YX8OVQT7<G_;5+*(*}8Fsv7*plgGt
z6Gul9Q#gc<cv+%3Ao^}iO5E@3!rS**cYc3Rh@Ap|E8F~!fAavEs7<n^;!casPM=K&
z&sKWt#@H(tt<1}B(3JZoVi{>Pg?z``hS(**8+W6sH3M)>29X?FDe8viD2vqf@^qH#
z!|>d^bL&$gttFKX%SZ-4+IMndCrya{yi#Ny1GQc+DogE(arnS5@%sHl_zV@D3ECl|
zPnrgaMmD@Fd8j+c7{XYwX}#rCGg~g$Jd?2@3H@P>5$0OqHYyPJTFxW$@(;!?scR`*
zuCGI~T$98EgPU`VLeGf3JALC{&DrKig~zq6522p+-LV=Mg}J+Mj8$FBLeJ=f%^WnH
zv&`dd=8GM)k{Yw0A^(JPgey{uTCEoBJpDZX78?sh+O>)Uw&Q+g;`x>zx$`^cnvsmY
ztzg*Byt=(k-eAef#@Z$`$ZX6h(=wyoZFqp4O=Ze|_6%A)VB5^rgyOrU0@+P_1-Y+-
z%G)TiB8AH%>Z;=^o|i)5-R^QEUzoG6di390-?|>(0maOYgAQilm(0tg8zGX}yiYh8
zYbFPvCV|hie2Z}7WxJL0>M7Q5sDT@<y=iq@mA7^Rvs$@YdZ`Uf-Lu-O%fZ2*CCAp$
zHxUj$?bYiJ?6;jN9^PY{mB&IIl4nBq*XJ2}vWy(1MSWHU;D}W>GB7d3JU&&C1(9uc
zB6X2W&pRozxbrgc@R!rHnnldwOls4<e9-h@9WIId+Uq!1z4tf)W~W9#t)3|<M>{{d
zxt=e2BGCD|)q3>1@!|vlksGi>imhAt#wCp=7hh9d@>fLg!9#=DfUMPJ^$gc>RD|`=
zYIyV;@x9ccTe()>&T9eA3|Z@zqtMI7g{_2N(pK4kR|Tkfd20WfAhmfM0|g8s{63p{
zKVaBm!CoBsh9f|!;}<S{A(`hChk0eiT+DfkkH-yFWzT>7w9FrFSzhRBZK(IZONg6S
zTFWbN+SlfDsWnT7z9@B-OQrsweu(-vqy2xtM_-1u1VMat55N0uc8|#&Tgi5)D*#`=
z5J@iWj+Ue0Pf=#cKE$jQOg_0a*uH0&mmz?4I6v7CW*ht(Wc(I-|DF6npTp6gp(nwu
zfisU{UAG+4+2xCJ7aPS!ajIU82I9RmrvGgIq^HX2Ex}SWTz-X+HSvavGfTx8rJX%O
zsn!-hZ1vbe6v0m-T7UB`wP#cly8BoVMB$CEsA4%%ckPGSlGps5GWG`mDg*#I&-lDL
zZ=?2t<&D|cjoI(5a?!#?WqkCx<#K<m##lRT3pnJc-`=jje|>BMG2uQ`5gGwxS(L;i
zZ)sFG<CYxs`-d4WE-sdExQz|`B#}t`$|OHOzadm83|JBhs<bcjcW!cXk6E$SUR?<n
zwFt5%+?Qm3*Tko|Ti!hxyI~W%w`N*vyqOsseBLdDFTjrVF91a1ECBDqD)sgC1;92^
ze+LGR@*0dQImdX|z62x6Xw|GJfZ-}X^Qg;8S$z(Ag{;}&ayO)|Yh^^1Nqy9a!v4p>
zATkKFc&`G{YNg33g8(Qu+(Er&_c|VRrB`gahj)?g@?YM7kTgD^x+;np3;t8|?uofo
zZj94;dKLXE+^soQ#%3ubf=UkPv}J?o^w^f22<p3;;o;#PTceu~OKN;r1)$ft)kwzO
z0#CIyT^RR;o_}toW%gslS41h;_V;x{Xy9B+q?Xr8mGIf-xWjx<W_L(->&sVv9rHJ#
zjvEcp8u=7IhxPsLQ6I==h2@0MWTkWGjGJO9hm2x4bkA#u&h_~&c$`IJQa4dQPl%t<
zZ!N8CuR1tfJD|tpc3^zn-7UV#+AWHI#&&I09q9?f7QdiXl}Q`LYHZe?vK=hb=@BqU
zwJXyioY=i#C-su5u$=5Ch6hg6u8i^D^g8f7zbyNUH?!6(Fhb|<zkN8D%0f;2+hb>@
zTG(gBG^#pWDr!xKT9l)5Mz%RYWqII7Vp{e{sc%3(1#AfN7jC+|a?I-1(nA{H?ZY}$
zQ<@v<%CTl&Z?Oke5>~~<6Nivllmp;e>X|i7L#ZQVYC_F#)elcKj$ig>C`7xp!~pZ3
zXk}s9!mQ662@jXm9x82ixi{5JkSR!<i#Q6_!qMs7=0TBlkq}!al#@PpQU?@&*4b-#
zgVU!yyx~PHH~xT4z=n#;5<Ede{f*Z@#wHLXm$&N`U3%qHE^itjzUzRqwfoCvA=_X7
zwzAV?;Ac~_B$soaqMJ3JvD7^e`njcRZbRX32YUD)_hLlG9L-^qsxE9_Pq!D3F=k@1
z+6^gWz$bcSq~mqnoj$A=M3q`iH@3)^KMu~KDW$^j>E~L{`CXc5lGE!qHs;J|8gGX4
zB?6qg>&oTP;R^EvfnPf{y;cP~)H*jj=QRk>NxKTJ;orPXIKB<1IE!%FBts#%neJz*
zj&c|`(54b&T=@pePy4hCf9XXEZ4jI*>1}o8_N}d5XENU3%(pry&DL`&*B+eDm%%A+
zfAL<DE|)26B>?7EJ2%ia#|{|~A`R4M{(NBMjetd_lJjYd99iqp&Q3;dF1bQCfQzgx
zuLtz%kE3jm;HrtL0AJondkm750McOTq;KU?gJgEFB+3-j%UAs{P6`1wAw=q85nn8W
zj^ysHMBaEWyVEnNg}0~lNwQ@{`7}@0VyOuK#L05_08xk>_;wGp_9lZK?@*%XeJdB_
z;vS#I{(BYrlSeDC0zTB|P0k-Gj*b)-U^V7e!;Xl)!#&b=H(<Y?AQePad4hxqDf1h+
z9txPzEsjiU9Gn4OnTx23rq*U<Wtk#62+B$)jzt}e!N!LCJ@s;6(5$ZJi%GOtlO6Ij
zh~y0Q?Cl^wm@~-f{I{e0Km5bL{AGu|)A_(e_t(h~&r{x_yd!=v*3S-P<wIj4jZ8g0
zj_!`Gnu5r8ooP*DVX{M4LW`H8>gJ0ViRnW%aP}{m7nAuLhU=Q^cRn2aO>AU#d+Q4)
zOxh;;mx<i`-r1qituJ13nERMhUA~o^8#iR&Ae5YWuW*7Ub}uxS%QikeHafLTs(`+I
zu-H^T)qu|xOJZ9*tq#U=ckq|ApyT9ZcK&F<7AdlnI*nwc6>b&k0??$`=JOrI(b^Gm
zg@5lX7*v4#IHOl^GTtyY{AI6itug#9uTi_oMGfaj9Y4B#v`R%EYn7>5V=fhrWLTzy
zKF88~RWDW0fH1w2Fs=V+hpv^v;R+fi4DXx-*(se<f*5UW>2Ge`bk2>~qP`o*4SCpQ
z4IFM~m1tc|eDYsx5}1o0HFc{&0n`3JY3IPSp93ldURq4t@wD#`<o*}HIC^?>h@#2R
z;{hbD#<rA(Yo@=S3gwO9yAgXFxunkNl^eG%K}sF2gxU{uD5bxvD|23w97&eQ3TPXf
z0E6Hg!Phy(7KK}=Y`3P-nCQZ|Qtr`FA3Zxgn;z84Vzmr!QT4YBsz<@y+`eu!=RBQG
zzwCJfzl*WORu7Em1Ps*bZWEMkM|7tE^Te>(*3jwW^Z875+he-Di8Ba2A*pkIS!qZ1
zPfX0q+QZ8nT&^xM;)R;MTn;|Z0P)j`W!55AB3I!t`!~MeZE&kIXg5A$V|oG(va)L^
zX0#z<eLn7HWJJYME|7O5<z${YZ17=cb7yvRi;+3DS<5+hm}Y+0C~Xp@!Tem<G^$)2
z26KBIx*F&ln-{v}0of7eBnWD_v`Mxto}o&Z<E0M@eCyvJULuzobR*BloVyiiSBMBf
zPiRH2=icIV3d&Y1g71LT(Y3_3K7a5~J0@_RN$k=_QX2mKLwf$)GRoa3oH}MxyG}Ck
zrzuoDtsC77)?R(J8fJtqx^(@HQVrS-7V>oNluyUK3bR0j@B8Se3#ta_X!_B|xz&N`
z&(RN;2-wZ2Mv>VzXCfT+^D^GT)2Mce6jvoFS#bmWWIRl5KhI#vOOBfL#kFuRrnGI=
zZjAi9Iffv7?9{~(EQ_d`H^`_jsrBqDq37v(Cz5iiDcR=cKq?B!Xk);uf2(pISly2d
zV*jyoPpp!la&^DwgSYV)6!<g8X@Y$C#a6=WEUUSv@O&6z--7Epir$}nT4Ivf<FHxj
zJmeGPsaFMLXq{p`Tvsr`UpsjCC_ec-Wu?Pp95YClaZ<J)8Mw8KYemc0y39wfh;4ml
znpV5f1GN~#v1z?oMlq97*kn4ED!nDg33hpf*==i!`i0N*6Ewm&T_{mqKyX*Q;mH*(
z87MSY;t%)b?YV%prGLy#2|Vt3Df5e2A0LF}0v5t*&pUkIZA<cO0Yy13l)#)0!|o?X
zMP5>jv=}H+Xja(QYSOO?Tm!xBZ=g{utOo<E9IbqF0Ked}4@*IQVV9dm`KJ7lEyG8&
zbtHdqY;c4TXbNkFja1-GFOSM&@NVyXN^RXjCkB}|&b}IJvhKNIz}rYo<l7bP)0J~^
zg`f_}U6C!djj9Cfy<k@dTYW(`-@|HU&N8w3*affsv+C7D#HnRjc_{Wtb}ke=+nXxc
zSys_=(5`_G3!8tq8&jaI@hx{FBkgJLl{**d&iU|wZ$^S5K+7~x9Nn1CPHVH%$k~eN
zS2jdF-&_qUc)nt$18a<IJfGrL;5|IkkQ^7+4ZOVG#MA_O<uMz1p`CL~<DKYPcDJpA
zwHdIJwWxPraa&Bt6S+g;H8aozDUj}TU3;y$wbxA9;)VJ-!8nOU@a^U1EgyB(eBV^S
zDq3AEb6*YC9pjA<6+Y=@(;LPDc0GVYjYh|(=F4;GIY|v4R({HsCKM0dP`;TedTM0J
zlecauIy)uc={Sjy`~}3zZ6cE~m){Nbbol#3S35swKQ}!pT((|HMt65YBU)p0m$O?v
zr%Ci|+_FE(W%NkSYgroL_o+3jUbHdnLHuAYITLmCl_!BC+EXPxk|aZPB5dAD8i+s5
z-1ytJijzW-PGWt4Gc~~G(kla#i&5}%h7-GX+7HhRSgmv3-(p+0Yznj&&DvvgCuI-1
z6>(7}1enlMuk)ZPs%hPjWz9}X@h<zKwl4<KxtniC55n7F(rHO=ACGQLa)1;2hDN28
zHRX6yr<AeeUIs8AgY${;zr=@|r`mQTs!59Qrs@>O)QIz+nD$nOx5n+d`4{}|Cmb}Y
zWoDN(J!$(~5o8~K*oDE=V5Y-#xgu7y6&glIU5Kpe5^I{b){U0(F0dteL7ox^t=~`d
zsl3MJ?o*J7V!`;XxlQNqC%(eY%uSf5SDXJf8>_{?xlZSoX1GsY+i+3Xn@gpWd5MNv
z%xk*`uP$eXZk2AI_95_|OhT?m8L=<nOE-#fYqe*mM-#|pcPb(foD~X5Q7ZEDJJ^J|
zd}XBVcvLTAX%9zwgA*&qnmHiMOh$53$DSP}kPdZHMq4aJhuzI|^W{V6HyZE90l1;L
zmwF9ts_)Y}eKGq(cVTvPtCB$a**Ka7yP5s+T|doaC+)q}TRVNto-?^ts+UP6@$*+M
zJ(S;{?}(=$@+|PJf&u&OcK@X=GAX;wb9~}+D6m1n5!<^2<5QrEm3F1*5Ad#d^KGb_
zhe&DiS8=Vi$Rj0bv?sIE31O(~HJVpHr4r5#goo<wSQhGT?Zuf}Vz@uu`}@D2fB#cI
zec~V=FK1`dA^xts`v4V{$Zz(cPX*i`jQB{OzmhIcRGOg%Ep9+XsChGEz!zcYfFYC%
z$qq_Uy^Zeq82%toN3EIU`4`VGTX>x$?=t@tY!o)$qG|5XpSAIFTNo7Ccjqr0wWc;t
z_Aaf`7Dme64TK_7xlH*Ggt1bp%(G9?<x9^fmi<NNp~JKo@Tz8!yUg^?&az=rWtc)1
zz>zfn)Fe5I$8JwRf6!TFpNbfdT(umH^w^IbuGx=VZ1BZMlt{j81oy*ScwS{m=eEnP
z;?tF0&V5Q~DB=x=rSJQsrsJb^7}MU(JI6Qf)X!7VZD+f-`Wm|yL&l^|7XvXzh=bBE
z`Fu#+vr?lDN(K3`;t#sOI;q+<k)XFloeR>_`%<zEj*`#}qrzwTqx>HHzQQGn@$ojA
zg9iBMxB?hAFYi$0HN-Lz-@%JG9!GH!$+le90u1&eIU*tgkp(b)S~Z2-MhXJw|5N=d
zi<h<cfDCi1k}~ToV_yJ`bewCZcUxM(%#{5Kr{~Kg0Q9VtOu*hh6)QqAWC5-Z6PL_U
z(vqdF`6pI>&*w-hds4|{KQbB&`1=B$UI0T2z{Of}eNl_jDsH4Vu`X5X&~-$;&_d61
zrQV<V)=(*-ci8H=;D$*r+=;AlYO3af>Z;hS0#?J)=y%aV!ESZG9O#by=6#Fg>&@jH
zW6e;LO?*5)(q{{f$=E~=D7&cZ*w%qTWd1%h9gb}KRe?WZZfa=>HcpA_*c5YE7&nPT
zsW4(>PlTnS!hH737Y(A1AWEbfbXVipV2X%;2`!@t)y~@U?Qs*U03iZAmQ<B(oF5AY
zW$3z*ldnDb-#uZ9;XUqF-D4{NgXEiy)uN}9SaD?6P-kF1J$iIn`<85$D#$uQlYMP)
zwel8NPJh93x@Koj(r62{pWtYuwA8gx>l;d(Hv1O&v5@5h!|+9Sz~a$kdF=eSA-2jh
zy=Q}FQ0uH-PZs*oKxrPS&P+;UNi6|fx?1JVFTm=d=a|Wd=N(c6yX133y_bVs&954h
zLj%z$N>fhdPHt|lD8QTHmonXxKEymW=u{F$bNh+x6qm<k*LLtLhJ`N}8*@0hw+iZr
z3)Gc~X!DCj0`&4z?0W!-2E-jQ(k^G-%ZG#FL(6WDO4Zn%0S~?DLZNL^$TjTt9L6c<
zotM)$BZ7NO%#d8DmNkaCev!G+aKxdI)1e}EgcUo+o9S9<A8hlOX4^dGZg1PL(#RNy
zTs(Pbzgh8oVIMr%pE`PoLh7U{WJ(8)#jcs>m4Ddxx~kE;e>14lMV66EE|&FA4L|L_
zUiaPct@!<f{~$4S<6aqLaEX)eHbb;3R<pWpCGViUVR!8zq}B#<MZeZY-tP3PzrMej
zssbis;Er1<&iw{}cl{8sr*=z9v7~vSH*-n4(|Cv6oMqe2t_nRFnK^ay%j|_*5+L4S
zZ*8ohG<FK3gt-kux9Pa=RxGuB^g-}b{#Bz7bw`HL2I0BQ*W>2+cJsOw@F{U&LO4j?
zDWgQPl`%}wVp$Y*bp_P_NNnjBmLF&w7;VO}d!laCFP^d+T*kLuDad5CQ5g;(ggM5H
z?&~JpgU~!C%qyo0SY(g5){5?jZxF+xfd*gQE}sv33FA1Ab8&hxHDS0%)OS2YgMY!Q
zxIjs@3mJ8zHPm;qo*->F&a94!@#5yybE7rnyZyTTi!YS|6j=Y#RHQz~0vei%R3Yl7
z#sLy22y!3c33*EPs+N{Mj?z;TMbgqcvF3jjsW_DVej?*)YCPc|Cr((&Ik&*}MT6+d
z`{%*)tt8|?^@|dFMbD4CDEtdXR;4EH&P=ndeS4IPW~G~h{=0l!9iB^xnLqssG?p<)
zKJx-SwDwpf!5kK{>b(Lj3Wa!@)n5YbgBFKFwt`u91-qoy`7uEV$0=vH(zXlv&TnG}
z&>z3+<z~Y6tD=j>n_8=RggHs1f$k20IhIBK;Kx#3XV~Iaj19^+2@4QDU=xknB7xf%
za_U04#14d4!|x|*cM1?y8y<ka+zL2lxn&pe&G1^;CQ~0t_4tuS7ks2HMZdKQYIGt~
z+&*Q0<2->LA1u3#H^SeE&7r^DLnVaQMB6kM4E2@&atp=k6fU|r&_&T(Vn4!%oiHEX
z@*u<WVquF?In$S^7Z7oPMF(bO9P06^+WI_Y30_@@<`15uu;oeZnQ6TlQrV{6>-v5m
z+UF7g`h|MPo#A`cBDECuwaIBj2e}yZi|(1Aptb7VvkJ+E*)Cj{M{fY64=Ms<W-SNi
z>0fsuV?r%TtPRQ?K-S?7*9N+c+J||=gTw6d`p4iZF74QsYgC?+fb@ha^g)L8rkEF*
zDPZj^8YFkaxt<Zt!njg><)(8=t)0|OxjSpt6FDWX2dN<G06wjf6$GF^hU#oPfwGg6
zJr(=+<EQF>PC1_^k?AyiGwwp%V;99}cI`RGD4Z^iDRu1Y2$;Dn{e=GOZHn?@UF11_
zp~g{nq4={;8Dt`?2ZB8P>g@btS%*0#j>7XaFC%72R%6onq7CCw5vMbvl8EZuH#Ibl
zGyJ)ZC!(T1u!JgZlq7<Fnejw40ban%TULxL4UCK3dr`1vIwVq#Yci<@^yo*@de5$+
z7G-9?RxDnDZlav>mpEzIQ4-E;FBSE)b+buAdA+n9lCArs_Kxbim5a21o5Y(>K0Wog
zA)S49m-SbNc{^}4VFV;P)!VIjSXQ?e%86LJSlA?HZX71$bj(3|V|l65nqA)cM=E)J
zI|Xqv1|olIA)(X3j=Swb5bV<JMf(K@X+dUyXZT+4K6sd5-ci02FV#P&xJ``@IxTQ;
zSs0J5J4A$y&o`g5pTIJ;<8ZzV!46{1%{O0V{EmkHH3Dm)D0*kDnxV}F^&EURU2+f`
zBAUn>1$m*?b4NzjTuZ$)HtkFBM|*GK=AEth9quAWcYpa{**U&lFdejW)K+k~7$;9b
zpVXyT*iQan2ClXHtGl~bv+CLnO{>Gf&A_UiTSX?~I}%NHJVrY%-sIiJ?yez}A;n({
zl1lV6_M(SQk8RS>=@QZycXxg7F&atW1qf$J5*3#@^Q8E`L>33{vaf}lKj51uXHUJ>
z*eO$sPa8e2CwUzWuNfb8-SCtWugTkmUh*l=uI*mjUu{L|<6FZKF5SxwvB?3ByWO`~
z4=X0`7fAf9=D#dliT6&ZI&l7~S;?t6I&>5%msgq{TZO6GfM49(N1imR(uzt=8D((6
z#<9^ZY>ryX^<eX9o$^3L-HZ_(lfI1=XE$s`g*a{P&ntIBj?C+LKw`>Y)EHGq>WAl6
zei$m%dD^VUFkZ=)1l(KECqI*H6QT-wlHWa1ukT(wc#NzJy!7@ZiqKpE{iZ`LNV`A#
z7LRECvWzIZSFlbJ$t<4K9xW0Nr1vvIVWOD1`G&E}hdGqb3{U8-mGD?JGF$Kn9eos`
z)mB@PjqJ5IqGDiUt7iwHIU_pm?v@Wfv8v;xJ!~cP^x3&M;#hR^Adj65;7W}AGpaZD
zc5?HCE-9$w)yGpDAe?zcnCWE*WpA8Rp9|I%d>o=*WqVYY8=#FgT>k6F`U<tK((=@%
z5*SfcZVg1)c4v{@?ZG8<6UIZ<gH=j9p`C}^DOzcZpH+?eWiyAoI!y^&9$wrKZ^T?o
zMx|G9H+&d@ZjEV`U`Eh*Td+smSxW;$aCs&Qnfhr4BIVR-Gq8Y<V)Az3u_-YZ%hLPn
z_RD|CoFlz7d|fMBMKk@rTRRRN+frBPI#*YU9}X{?nXsG06+8z|IL<0n=ro6&7w0t(
zgUQ9lp$diBa7~=PSww(^n|CoL=yqXUplD~`#Cc~K0iEC}q*1Bw>S(l*+DWISqVU&S
zbcJ6dlvQm0F7$}Elv3yrDaILfRFsrI)96CWfa&h_b(2uQ40m>9bc9)eMuUS2N5Upf
zZ9tx_fm*2_HcXA94C`+J4qbidm>M4@rki7amyC8z9>_XALF6f$aI?tBsrxc-fT{V5
z!zZfx504$5>;R*un)krVy;$Yj+g~TJ<-5)GuSgMf4=FmNIqtEklQT7`0EX)P-cqo2
zpxPN5*LLLg%gV$_#F-5dh*mMSIu;N@(DE3dvO^N21Hob45FdX;VU}p@X_MfF|Dl7b
zhhx4?5%Nz_4nwctYzM6uB{x=_xs!Ik#q-aKrp(wnx(hC^57@t-I6aKXNnbnUrLm!=
zzJZa2F=ptWN4qc-*A6W02K+?_Mv0&nR91oe?4P4!tYfCQnYrPOj+|hVe?Re(e38{5
z!YytID*U#Q6Jrd&Jq*8AmfP+pE^$k3ylSXU9i*m6Y5w7p@&EUF`eLsBocaA3CHRB>
z?V_l2w8c?``u-O+LeulmQLQI^Q#JRl!#$B<SmDQrpsJM{ehvmEb!ex2*Z0s6z!Jei
zx$HDctc1=L;EvjuYauO@Qq7NFOg9GWp}@r7PpGt3WHd%dsNi?7ybx4JMs)xpCn<@*
zk+~^c2@qn?D9%`W3fc*EaeWy*&v4HT_hPz{MY9ZSKA)B((!KfWRI24K_B@VadicB8
zP=;?5^sP4r)y$f{qEJZZ{Khk+9=u`93+>h@dXOly=$g^bQ6}R1&7FI{KG;4}i|9~!
zw_G59@bv+H{itb4bd<kWOf7DYZYTfJ@W{|t*FO~cdb*t;tnL<6BHLe$2-7ojUVN_R
z=0-uhC#N*!q2g{)NX3xD-N@){+p+mCY)50!YI7siagiQGSstkr!r7t+ov!N&2^wXI
zaH*zFhPyb4RyQ-h`;X54GYWF+z%S;@PuetvJHx9r_%7}l-iB4~bR}yO?fa<0G{2v?
zuhWUESfWR7ZCxrO?v<8BIR}{<BMid^sqbfN^p<JQU8dVUJ8(P`56zCcx~2vfmB~&p
zZvX0K6A{6Z$WA62Y&MrZYc<~>p6x>a($aU?(wH%D6V_P2xrOfXZH5x}YWp4)Hq_BT
znOVYGY-&p|e7C0@?l3<)Ek?xqwKx>?QOj2IN-=W{VKYVfeoDOaA4A##8%Q2=#8cy;
zqx7U|PHKh+%x*Od3n5>ARQ3T4w@|Kp7*7ez&8zq+M>XNYKuM^*HEsK4oT>^Ag$pH<
z-a@ms+bo)MG(H`E|2eYZd{@_G?-Wb<(Eg6Sq&1=up&Chy*oA6^=}z#Qo9gMNcDEif
zHzXmb_T=DorN>FjW4`C%PsbRIB}B*%6}yc$XA+Zu_^GA{uOQ#j=_z|R<$F7f@)9hk
z-V`L{|Iyw518P5D5>--o_aM;a;1JXFusFN9C0Y%yb3ORySY$+P|Muo9=Yef(&%Jz_
z4&mKU2{Fx0%h!iOY^@`M=1#%-ek?)gmuz>S+8rw6zlo7=RF)!8oE9#yjJD0bTh8B?
zZzwou-q(8|+<TN9B=J}dch4$})+40zSm_~Q5Q<91&a`B-d9RooZ3cRb*5811^2Sr1
z^|yj*WiaHR#%%m_e;3qiA{@t#c4#|E$0mtvZ_HumqiXA2^4u*9pWoQq>N0#z!Jxax
zv@$*j>*b7=s#A%=z?m}!pl!R?tzC5wF`2Zo_@_f_K2Ps*hJw_{&UfeGpW15Yr{?<u
zFUK@{U{00CcAV;A=G1Ms+66nvg7UwIPj*C9@N~ZDM0RCEeXHPTOxUo-sO+4cou;?_
z3bA?{+K+FjUV}dt=&X{lN2=<!vNYp8EPo=YYJx@fWHDXoN@xb@@}OAv<}8dFuod7!
zaV&`*dg}n38rQOpO@s3F2Fo4ah*Gz%4-)-Y5V3`xoqm75U$<^5(#YQ$SrR%;*0gr4
zE75WVes?-k$!LK6MNLfRUe%J82GoqPpnHIT@&7J-UzAW!&pfsF=+?pHv5N6jJqsmn
z1})!$E2DKt%-aJW-@8A<sNu3&pQeNPY3U{(;T3yB!XAG@%`3p+XyT$7i%h=kym_t6
zqA=r5@he8?nDgnu=}rBH6;Be$m(Rc4@Mzt>djh^3geo){WTb}!!Kh3qzSbcwkVm)8
zrrFpPHrCD!xAn3MZs1?QoF)s@HYP{Xt8E``Be?anmx*Mp`B7M>GqBB-`f<(Z@4$og
zr!dT)*<ZR&KP=l0uWtef>^631m2xO82Gwry3+2d#=FYJ0{p99MAa#Q!6(0W&w9Cu3
zz3ZN#)Y)%YnFXgTFFwZ9sR|ftwXD4_7btp1#Yw}{t=4N0U&f|S@%ZDY(k)gm_~>!v
z{?@*V)-RJwJ@`HqM^U-rGdrf+!sHv0)@o~8-O|y>Zh@kPO!eKvDrim@b5&3m4>`~&
zp<oXw;>a4y-d)}FN!xGL{tIaE|E{dpxQ<1ll;RV6UEO>_jp16A!A^1+)K{_TOeE#k
zd$iVfZ8c40&isv!8+<Z6rUIUP|6BJT%N?^|;T)nrl3hWqEv}_BmsFD+!X2PWWV=@P
z5d%p?WW<KcjdN-E9kOkM-sc``pN@iyLO?0U-J}M&=5*U)Hw1L;Iv`G@8Hm%Jsq|E~
zjRKSb(J=M;5+MMU?AP_nOG--mX{{Ui8%4EUCdjOOv8p~McS@cWMVvXh+~r7#&z;&;
zy|fyp0L{bbv^4jKp6YOJ+{!JFl{kD?a(Yek08ymI@^CP3`$EOfg7Rx<XY}8MIhvH>
z$(9F#=p5V_-7M(FieF+Exl`PT8cndFd3J>^1;a<mhu3tUO{x~su7`iUKSZ$D2x-GD
z7dd^ou?~nDNH3<VyC^xy1aeJafF*YXyRbK^ZeC?4Pk0ow<;5Ypob-_H{iDmjF_@xE
z|E-6pcrbJaJE+mbH>yQb9unqms5P|aKhuc#?Z*oJ8t>5l@=GfsQ~IBaVhYlb53*nS
zmN=GlYpH^`GHNesDRT7dP&pG%Aj)wESf9R~ng+722`JQ?C<s<7GyLb78b@`0^G~i-
zb`xK`ph1jwgTWZUR#kbfa%ysLh;1$YN7Sm*s|`_O1$bBx>6^QYxWyBtV7^2xW~wx^
zLW3V9QK{n2SoNTkYw+{7FN?IWdIb%<>I19S;;65eBS$@z!_?1AnpMq`Ti#ThxsNnc
z==HVd>e;y{6u$Qc%)7+H1thx(4P8Kx+LU1}8B%*zf%wO!^3Y;b__EolCJ_!B$Kv9?
zYKhV)$?4@x-mS7#fHk)K0B$%uL{E(i>O*lIjV0BOQD2QJ!+QP;VyMdyANZ1Vx|Ur~
zcVf<kKj`s%vqRFw<Fs}JtY@M?w`#b!+NVUszdDqo{VT8ZkIw#&ubp{`ZQD6uF?IpB
z)5D9yfjVHJdCG5LE=b)WKkvW@^di!cr&b}l_9>g1ZR15vt>w=eE(*d4?_Hk8c)*@c
z_*%nH%IU@bWfFBR<_6{5(qyN`rRv8|*&DbU6^qY?ZlMFkm&F%*%6p=W@)*?>-M&dx
zLH3O*TBzp?qm7O2j$*W^SwtApfooW*vpp8`7GvMWtcNK|jH|5ClS23WG514%{1%DD
zY=$mlebt-#1FJ8sb5wM4J}zH!&yF&G`C(3sD-gDnB6V|G8<{>~jap0Mu}SYRx3hfe
z>KcU-954uADI;Lyf65eLZ?-h@{pKBPY#h=!LxP<ZJCAecl<+3`CMK=`CuH$hrwEGm
z@CdxXNbrQ8yzZi`Xygy*D!7{IfOx)IW@EO8q;5M>;|=Kj{uync+5S_h$jsQzqpro+
z0Z~g1+7@^j0XA%rg%AEUd_aw#Jw8)*Ye+w>GB5__2k9kd*^J-t#?@~4(Fiy74B^F(
zrgPU<9Z2^|^feKSmuF{nK&5Z$!0sjT!f(8SB!N1I|M>lcvdVKIEBY)j$NzpJdrd*n
z^Ve}f%JANL9Vt?&ub3Bg6TaAsP~R~=ayoY)x>#7~Tq4M#&DpEl?-?#?Ow7B~N2Ymh
zY1@H~?6u6yUzbAQa;nYg{aZtXlFh9(Ww5$dbp~{8Ls<RD;`G3On{obi75}fwDY1SH
zTYJDkyB65E^yt^#ma)>$yz|lDHyav~TKxnG9CL4^LjBsjd~0PZdtvH8@$&M*13ELV
zx`t3j#g$ZPu6`Bxle#*@)?v}9z8n2pBtTCxh^=1ednM2L{lv3n2$g3S{1N(SYAXuo
z!A__K*Borb&tvrPp+45SIek=a&K=GIca}B(TO0q1+(FvHWT%SZE@#7NB5x^5i?r?8
z*iqIV72C_*gKx|0n_6@o(KQo@OqtbLU#-3B1qNW-5?oLLg-n`j{wCzC9|UD$lDk|F
z%oXN7y@`d=1_{argPXwB$@l21Y7YQME!v+_Rka_geEa@p!TghlD)~1F?D_q;cyRDP
zuS5Nx*W=}dWas7{cp-Z`83V}8SZMXbBk14%)qk?z0&*uSt*t2HuUqk1k(cX?q~7HT
zwc}0-)FSj7=-@f{!Y9T9y_a(p=1q(#!-n^4C%)=wpH~j5=Uj{<5Fb0FvnxVVJ?ltQ
zD3yx1qRAmTshgwm>qNe2M1oU{Zg?HJ@9;yr7~H4~9V+qKH$9B$i$f(5z|2^;*_{Jd
z3D{7-Jb6D_WpagQ{Hw4q8<#Fq$g+MwvvKJ8G)oaEf7P!C9m7V&8^TeG)3+>WXxPiH
z?4bB(4VrcTx`zMw%AZS!=0u&S{4dRNULE?5pxsvMC60o98K$f0$gjMiP-87OEr(8i
zw*?n_zogn%8_g-eGe-wAjBJ1CZ6cpu^1Wreg#D`bu;BL-U&@=RHBb`c@9$LNb^tTK
z+Qppvfb)hZT%oCvoNk%}s>rXk5k<p|)|8^0KtT%o4Q07Z*kXW)JA>39(|?_`0Gpr?
z!}%){B%Pv}n|v?+aE-G%mW9tE>5@Pow{fQ1VcA9cii~lXZ|a2_RO~l@jI}@cJA8rp
z@@?Tft};r22CLi%DTU5W>U9)45o<RO<<?dL-jj(!nqbxY2`4lntute&!|R6T9`4L)
zVZC97-J4HO&$UUb_^B`=VxjFl(JBi(9oe;Jqnd^7bO8gEl%7S-&$K;|I#-+Qt4dq;
zssjw;FSd@O&&|u8wKU5J3#l>eU$a8x(c1!+`5c6*Qbj?sAl!v93*>)tFuvJ)NA@Mf
zecH~1t%%nkj7g{<K%~FOw6@Z8%&>k`&IHmv7kjiboHDE)YG+tod-tS<sSW(-(Y|u}
z;a_im{p5-mI2%UuCDb)EOj5_^9*f0cc;GjX(R5tIdUL3)(hJ}Jvsfl39$Zx!747|>
zja|+571^DDt@+f=rbD3sQYDvcm{t+@eyLzh{!?B4v$mcXdyayTNlUXH=p!}$h5>V3
zmuc}K#j6XSC3s)cD@El7&xR~kW9yAbzA9Y~<wX=?SqBfznOrrCPY@OyFg#ZB^epUc
zN<$UDM^90tdx333+Y8}t_=>-%MeR^BV^q{y6bz*fPp$pZqJIw%#T=a_3=(G<GhN-Q
zfoopeaY2QHys=pkDRv&^Tt|>M1d_f2_kWtcxr{8e<t&e+*t5*@>VEjgzb;p0#U!&I
zIE<aoFM@^JKr)&7-sNgms;6g(R&P=@vi(s{)U*A8XoCI~c?;>sb3<fd0HQj8uTHD%
zgwe>a(Ym^Rl3u&GOqn@|Lz4KL!=&6?s_*fH&JegA%E;{feq#5C4)fnJjj|WCC@Bf5
zz8k6enm&twmb)Z2#*wXS2Ae+wn?yV_0XN@gkV<iHSDWZ}{!}+H^z^cG&HOF5g0pw+
zY*ges?m46ZU}aJCpM*d(O^COz6K)>$z@iXY)&uAS_~SVVxL8LzIgXkf0-@Ebc0PDB
zv5UInVu>_N31k#LiM~G?G-eX_Owx6Dke$kw`)JE9@E#p3H4kRBofbEe@d}2yiIu+0
z>@2bTKyG46`oH?;KfQ8wyQ`xOj8GZ38fpCE6z-v+K4$2xX#3>U>P&IO(pVUndfhXK
z=ekF4OX~15!ny|<#vMI%kDWkx#Cy{7Up9YvX<+no+f<KvXHMmT18nBzvgP-}mt{U%
zUqqd&c_?aMnj=T-xSQ%r`ar9UR-=Yqa~(Bs$Ra#rhB__oo_RYSO&a$fFzN1=C^+JR
zMv9bGF1ElA?`PB+@~Il^X4&j-X)!Ss0dVaH?eWq?wLDskX&yg3uZZLgD1Ap|yRtu^
z!W(<3<LHAblwM9VIX>&-45iAqJQNrIoZdi*!bKq7Y*pR3J_zKa%o71Iu4pWhAkA11
zK@h}xmG02|=&^Vu#^M78xvcnANrEzsNUWlNdz_iV-w{o<Cn%w9vE_iR1=&k}msJ4x
zUo<OR!W2Ne-ScRAvvtSRIE*1PM{UKQx|SiKQh>A+vYsB)XQ_zy^}mMB*>AB|cXSbv
z6?%_X98eKG1<*R-?zbKb(NM=UIBaaKsN0>Tjh^_a{PgvXWRYB*Z^_GsCu%vNA3jz5
zTua3J(1R;=Jw@9h&g1?=rQc6j9;(lXP+Kk;LxpB3?*hC_-Q$Wx``Fh=<+HqTtCfBJ
z?$y0qdTqPVo1#jjj`qB=YCC!!Z5P~=Bjvs93bJo|d<|~naLw+YysvSXM(2@T7gm%p
z0STJW_|i9&r%y>kp-_ZcDnqamRH>~E`qgcCc;r(_kUz>l$QlObW@iujl--`-PdqjL
z_R}j1U}7%0R=3!5E#vmryl^Of-%T92Jl9iQU7e#Io0ht(lLG~$$j4froNJ>%hlRm-
zOrWoh*T#B!xAHbibO#zlT;Mq8LW$;oKhdd{zE}xN(%3|ilxFAu;RJo+r#61#a^Zds
z(g<C#6Ejl!ew~&QN?)OEmZjFc#T34BCNL#Ybu9;Px$TQaiqPBd>mEu)*gR@0(_x!g
z{6_LlD9QSKcykRw#ym)zTq{aYai6L|RD&rVOZsgS;h)<G28<8k2k8<T3&L(XkZmp7
zHn_u1q{l_>u4+_GxAZBiT8}mxB#C62KlH|EVGs2Z98PKpNc{>?i3dK;5)bJQ{~8u2
zWRVgtQ`IQOVbctv!F%JUvdYM_>A_-GNQCwCe3=c03%3QP4-HC;1&hTB<dfU+<uCG6
z)yPqb7S}JnaxBXFw!ce*-?JF3%v`8<-LiCY8SjZk^KUMMDcmbdqByUb>!c5-auf4f
z9xg{8hBZ}Z_yTOf#*H}|8{6`D2Rpw~%s1TmDi4Ad=gamy%QPok>E5s9jL4WfL}qQ5
zY0w~}vtCK!3k11VbZ6c_{%H1pcI^uzg0lVWy%j=Cy;%(6BRi(c1GBR#-~a0XS$}7L
zOrmfFf;07h1t9Hi2aFVb%<B8**=;^<VP>n<$+amPiuf@|N)oxJ;AwHBYy@@4<W#kV
zY+0@z3g3Ctjnc?*rZKSU!JX9X+|(BY%g1!kLH-CDz7$r-044YD{iwxF=b05Wu6%CK
z#&oYMA8A-$#SiROBTz3!Dgj1rPB7?MePypVkISvOnAzq3$DN4O__cq^lWqQ6SfrsG
zT(><6Qc>!*k(CC?a{s}YdUhE&oVH&&+ME2iY6~`jZOnw%X#;!(#1t8`TuBMMvHI=A
zk2K{v2L8yKUJM^NZ1UTCoKhxk+)MAGYBpo%sGEAIp$;-7r%wSBWN$nqs6mx_#Ktc3
z_b;snK4jo8nXOWGubRr82Da%v9q^1nimLTY;ZNIeU-B*3Hf7#II}WKHFo2GkA}`tu
za8r1$5OsxfRS&C8jvAA@KW`l<-(pDp@_wq4Ts@p>p`s9E%~nY55-czF*H$xJNBWC{
zd{AiXV+G*#M}AHjg&95PjBw{xR-Iy^GG@V&#`8AM`NQKghTFBX{PFSeV|gJwCdPOT
zP#KzD+VJinfzhLY5U_Ov#;xm}sXwm>+o^ugPEb~q59P=z8kM_atGyVmr4a1m-Q<Cn
zl>oO9I95&stme&%f1UlKm;Y_;PuXqBF4|*aOkzTap)k3>q1ezU_;Rp%9V&IbqDu61
zu*IixD04nlEu*TfOc2nc-KPJX_{}ckqjdVX=U<Bq<^HzgyJ=H}WXNw=em7;z;{Om!
z?{lDe)d$Y{V>n=n8f%3cWZsm&yN6q6X#9Fm)EKwLj|ze=98^=wlK~F+<Mg^vzF4K*
z$RM=WazQ-x=~4Pm-~G|#RhNW>>qAq}%`Lr5wbuOiY{Wd>@D;Hs<Bh+adfB_>Y?D60
zHO_m#+u!hp!i|$<dh}z*Hty$JzXLpU{yN;2R7FDUuWt)=?iET`lQK4ibu)r@{Az5y
zUUz7WV*M$<q@+E;?B=5y=l<y-s#XBD?m9h%=^#DU#JIAS?{vz5(H-;sj)+Y1%HRKK
zesM?W(<3u4_SNV-RDbxk)`4dA-<tYIqyGbIe;T^6A^EeC+mf7@@!reLl8>SFa)Egs
zUa+IYQSLZpW$NdCQ1BI}zPqF2tGK|6WNb7lU%$fPwB~&h#~$6WzQ5SA*DCZZMHC|H
zi|je04XsdE=sw-Y=v-j!2Z`b^U|p7%@C#=~4@ojeY}Eo&wdR2QNP@8vrXsv$47=MB
zuS8OBBPnYHZ>5^tuP2q!MsXqBNx}fC)Amq9)3m3#V!xtdRG7b=Q(u2^bxr3+02{Hi
zx;p)4sAA0BrCfhV%Z1_`orS(n#Fdph`j-JQ8u_k<?5~U>-c9j%U4)Zg$yZNxJU-6m
z1s<&nT`9^YUBD`85jOCW?4{|Nn3zcT3NYVc0w~=A;0qA`0^cMsm2B<?Y|j848k_6+
zIBVmNTYnCEaJ5;=r8Z7k2{mI&w8obXCH60dX0L^Y!VESpM$)yL8~Sx_U4p&H_hY>n
zt_;atd0W??uUnxWndR4WZAjJL&iY9IP_OI_fW?Hq-Ac1_9h}-W{j%BfPbo@f`GZsY
zaf%)ZCut=>URkgnG9I1q27AyGy8XuPHGT0zLdn6hL^*}GUE^I?U{ohaAFE>*^syDt
z{`_5y-%l8mloez2%**D#KUBwG%aY~&(e8g=JNaEfRYz2PuxT*wIcZ`ta337=u$4T2
zsdZ%nqZO6WSUzu9h!m2;Ars+u*z>ywqtzi0UuC(+Q`p#meS;mMM!Mjp)4Js0Cr(F2
zGF^31neSB@AG6R!pNI(x3S6GRPa=`v7`BiWn9cM*I^cWiZxx07(IlL+VUE0T6!P?q
zs3Zazfx!hit00QH$ZlHwMCG5J;AhIOW#ao2NIAU63CNS6%1c|}&N>Om^R}1#^`=27
z2B7pUv0(o?AH5f^(y?{uC}BiTWKIuKF>Z&>mAe2%B>LgI^Jlg0m_F&42gOrsjVj}u
zC+7jgv#P(v27$eAUc$|pKO|KRpq_J>49r^SZX0o7HE?nEP=$u}rJw!%glI6^@!f1y
zpi!;MwsAEF_$N%)mU!k(Nmaq~Kd$3Hy7nbDwiIFCGSC9Kx%b0Ub~7n=S4^NPwgM>C
zT6ba7oCbO(S={~g#*@tziJjs7EgN}@uZd*MgO-^m4>NUYGRr^D)77JJhlPlxO^khv
zjzL+rm%45kzI0w{sG1ASG@#k*6wX0xu?Kt@<PdL{vh#QbQ{EryKQN(@HMbcN#@n3q
z^ps^^@jsHU>MpG(P30ci{(b_kdgTuq@QD*GDm9Hv5i6p(7H{U*G87QHrYD4NU3Ow!
zDy7xTdfc#8Gr4#0`-yk24Aw4qYrQX|JHBs4=M#S<R*DKvn4`BH_Y>a6Wo}0cl6kd{
z5H3}St;9;7;l+cM`n4GBA@^?kSVT2zWe%mJ>%yjGCFYu+8zZnjn@U{-i%lRj`sVlS
zJhCz9ZilqWTl(rPv0y#Lu1u5%i%7ipzuNoOXg1Tm@6PU?b~-)njww|Y(`sp}s5+lw
z?;dEyVTKk>kRe40(l|$g{Y<M`r_4ALhiw$2M5>VxglRQWjYFLh5(FVi1QCftLf?Cy
z_gORh{qpSnylcH5-dU4R|Fv?jtn2z+_kHKSumA6W|2NN$#OQk&0PX!X=2`mse3lAp
zZZXRka^flmw~M;nfG@V$MOpk!2DuRibGq>(H4PiI+UVgQsJp4VS~itG`^L#B5$GaH
zly+%=q1_e1!Mt~VE2kG)-^?1K^WC;ESJhP2Ov8|!?~Ky@<~(xVui?w`*;1)Ae5|r5
zEy>j#`{%7!t+k}4$4S2W#kYGilWuQS3PgvX0F}%#Vy0tRa<AR`<DM{xGvJ<n3BqZ0
zx%}I#P9OEmdD%Il-hTh8mR)VX=;~i~@@M<;K+5znP+-t{lJw&q>6UceR5?_4(VueI
zQ_Z7zI7ATLrkU*M&RQtSh1diT+t%mv6@|}=iUW$^qv;I~Z}mk6wia`Ynjkk8DDNUP
zOz1&9zBIWsfcRsQ6I5SCWnMIwli}#1?-Lpc(gn4jbV{tLw_+m{Lr0%s!him4aKG#7
z^E|A_7&6tyZ^-*gN>v``T)T81lycPWr&@F}rx9r#G6Yc=e_dVHng!=316@^WPyeX6
zI&-IqV{kQCh2U8qMYOuN`>tSP3;e3gz@ltG-H_?m$P_W)!|7#T(85l|-hQRBMKi}Z
zk>Q+B3zLPrANP!;%b)CWZ~vdWxqCh(@8s0Y!4mH*n>88Mjt4X=o<`={wC#X4d`CB|
zKL?XWW$|Y5yLh`Y76-6#KCBh$pygCz_kPz6hwBBi#y7Wj6=U5J91t<9^=f$rt>{SM
zX2TC-s@n3AoNYzK2T5VI4D+g?$#H8}OT$dEBOCStLB9PrS6AIWmXvXEbtTIBPFJK`
zX~yAeWi*(2YF?6MU@fVaHn%Ke(aKL?It9^+;;`A%(${|>eRkC-WyApuPpKbqk$iBi
z*r_*EzY>{9Sl^*{KYf}Sac8RW#ae+sLb7!#PrnYFSk|ZPxxRB^c-b2#@o9RA9#T0z
zBHXfxYuhbWY0BG{Z*NLgUWLZ&e%vFlOM4FKs`~q{WB;U|?k*sd!k1QjEnW@|BDrlT
zy=^|#{ip*8Ze1L^mT7bo1(|@kM{Ro9+}zw+Yqt;;R&%G;?Q+(wii0ST+I_}gf{x6*
zN!WXG9q?0aij3|ylb}kvR)M7Q_38zxHx2OLdYr3N2<aER2+-4F8S!95N!oV;<lVNF
z2xUwi8*p@==P@#+cF!r5fUC*QnFeFKn@Ia1+w8A@9p&%5xZTNZ{CSu*Q(wHa`e>jc
zWflbex&MSc7>{WX6$M@`CSMcT-)B7^SqC&Brw>dYHMAfb%B|N-8RH2LW`LxqR(4B}
zHoS8}JhSBz2?gGOR9Hj(AEbaA&awO%_doM87F+G$Cz@zBD%)*K$i;gAXIucqB_%it
z<GwljPuRj-#~eF-ydzQ~V${!fAustRzc~@Ip*23#(zj&Ikd>SOnvWFy2stL{6-re-
zakON~VmrUQ&3jTC%=cgBjY<*;euB-(4pNas`DKUhcVwldw(J?d(?icyn=54`8H4iG
zQrSBmy=4(H{LcMX7~qO7oAxpN?n}x)jpDy6KPd}pEaWUSSG+m+p%(w5i0`wMq=D3E
z46hk1$}Q$q5w7~$a+l6`zWc&{(J6fIH~q6ZfKsNod8{1m*++@bc77Q<7O;3bR_if;
zG_mi^z1bZDv3JwnwXnGMxG7ps_E=mPVJC3laSRQF%w(q}&a0_dENm5sj(ujEI0dAc
z#wNXsVj4Xx6UgK{1E32Vbm$H<@P|!@mgd#P`3`(nj7>i~0IuVhXYt!CR63SK(QDE|
z8+%jR?U_DPif>CWA?QLN(xXzHcrhsA#5=BPerASi*}xC^rW9B5DButql<nRIsDLiG
zUK(9ixFk5f8fZUZS=y9lM=p@rl}xu}W?V>8uVo^`h8iMyzIHF!B_NPRN#V94tY?aZ
zH-DPe|G{B@xMw>uoLedmKj@CkVi|*iElHVClne9g_tJTGanBM9eu-B=$Ob;GdmZ)n
z-^Tm9uKq-CADmZnh+ADKZ{_iv-R2)#vND_W2sP)@gUXXf#quM@t>rc!_dp@Y^e!=J
zxaPzbhSNw%c@9E98q-r;Y>kTkxaa9o@jj;pZy(XnZvU&9PGnrY^RA(;26uI~E<tAu
z9sJ-_E^>2|Nif+gt4`CD?Ct8Czg|CwG!EA-c+aE!XtUW_O50Ri-+j(A5FRMpks-An
z+NpN`^w)nv-tgQ~gCuNnZ=GiPoT4=et$R*tuja~hwYBJWlWH=!8G56W-RA6?;I(S)
zfY4y)qf6m+hL!1Hl!Zttr^hUv!q@s1L={_t16r!D^M!JMePOd1Hr5oUb!K3bV?JTk
zUdj-xPmglShGulMI#V=4HDgQX^x>fqr>L6Vi3q^UQL|QU<IQdnu2-2u`jF<f+8BAP
z=Hx#O@xLqobilS8ROS<d{Bg|p>hP-MU4*OWT9ZvfN*)(a@Aj1_<#xJj?I(GQ2rabh
zUWitkk6ZO;9<;J+RBE}X778_uP?v|q^HK)!Pn^M&o)!D*J2IrJnlq3St1x{K<G7SC
zUyTabW{UG>QRJWf=^aRhFC*W;^~-0Ic|jZyF}!mq0)_N1ChWxO(`>%+l4fliy9Vi|
zcDy>F=ghjE5VmoVZPT+^yi73r>Zji(Vxl6CjK_MCZ1rk;6)sjtgTIPnTs7#gA*sHL
zi=C;_+`ETj)$R^w1taDS0!tzOlf>l<x7U6AW0a;=IdjAy^HK^Rc-$Zk%M3ZGQ6TLQ
zV;RZ8)U0KrRdZ*PYK_-T2rHc^C98mR?YV71zN}Q)hTr(Kc@?!7BX~2luJycs|8RA|
z=ra!X<t&pnGjJ>f-K?zxwjytCZ8c;gA^@U#D9r0ef1R{_4fXB6jr(_Ai3}r~L3}d(
z>m{{MMk=mS=5^5?yn2z;+y2&lXt+_=osPjPceFh&S6^x_dCl6NBP|U+_b7CF-pzGs
zu4OlQ`)T$d%8PC)JDE?vECWx!Uc==e(x6+-H+fYZb~Rqr1gGT(g<FSRd(&LSd`I4f
zv1?49Dm{%DS$y`5GYw>pgzs-<Qd`|KQPo+v9lH>2)v2<(Pn7&FyA$s;bTi-jt*IY!
z@#ve^%whOWAlAGT?@aYqLQzY#{b87(Sd_=DkoV2h=1RBZc++!vl@|y5&#t`=M(h(<
z5uiLSm)C-kgf1*2r@13zud)gs0AgUK`isC<S=pxDcr(2r9h)42+j<d~Y-DsPc$SH~
zYnZ05kJqv0{TG?j2_Sn0iC}Di%vlB~V^ub7bp2L;J0lL?u(_};fIpwq+Ck2QRz8zs
zZ<{z;PfAhH;tP|o@L~cswj-#)^Zlc9vT;$a{D7&K>)hPQ6INfm4xYZGOZa9b-R`#0
zc85fJu{v_7SA7$+WNEbl@6Oy^QjasJet6Vd&}6zx7Bk)n1YxXAsZ#oVxAE~pdDjiP
z>6AWu+x3&K01`Q98Qn^|i&w4ZpT>p@0bhj-%DfE{v$K4X$&uXjH_`jYZv*~sE0FC7
zktn0|=87o?S=R546T`&@KMj$P(P3zpcvErxd5^sKJ4Df7Y5bj^hO^=UCAUd7z{0XW
zHrq)sX!pLm8qoa^d?=?k#Y}^z>TQXJ8m1XsYAoUdSQ1f?!GRFynUR8mr#NmyJU5>h
z2D#22X}nPz!nJ`?>$apSDn>>|Gl-sqyQQ1$n+bC*pK!$wy<U%cGap*IP+>_^Z9VsD
zG@#JAUZqaY{X5G#GL~6TW`udp3(B)n3apl%DDdhZf1SQH`gJ7*|H}MLY5tp(ieeO?
z>`L1zkN!MZ*8g>*KT;j_&YvTx7;Tt&QrCqKV!uk*hF<Wtz+A#LX{Mrl?SybG-<qg&
zy7tR?4g->CKTFJxeaxB(Y#<?bd6me7q_Fqhah>46uuxS_KPprxe{;L)B=1*T&vv}V
zms_s<<;duwO>|`D+iNLNt!bKq61nyLJLRbK_mYhVMJFdtJ^CR_Rk82m9<2SzuFoA!
zbvme(aQ%j_sm)~06{E^8+l>;nEmj_{*l*SMrv*W=Oj|sHv>nTsli%xZ8l^tiEOpyv
zZ{YgOiK5w3*!NV&u6EeWkiak-l<4YV_#gpzp=<339C=T<;&8ksBs>!KrPPaA3Bf4Z
z?%_e>3J$*2?@1xue=L#_=A>>D{OZY@nN^EA--G+{<Tri~nnjmfw2o&qRR-a^{3Yz+
zB|!S6;rPW+1u`Hg*gls+q@0X!11u790IJMO$Y&wg*4C!(@UXR0kNXlPc6N#tt*wN+
zsInjw3aKw$T3QMOENpwrz@hmpf&t{0BwkKqZ&kqeZC(xE0vo!gq8KShto#{int=c3
z%rA|Ss7k;$*bOjGNB-|7?X}UCzQ6v8@bA3%q$H^ElOerHQfz+UY_Y1|bX-Ma?iC#Q
zxex4pdY%W25h5R4V^d64ALZ;?WpWa9ru8qH%r=!jnUJV93fzd4o`5md(u8bie{Isy
zaaWI7b1(|Mp4w7kT+o?)76xX*t1gj;EN|BJk~VO~V0!Z{y4dnc_^Jktc)Z^Go68RD
z<?6)N@fh<}fm}k*jEE>=*Tq$u>x?L?NyanUd=2vq(lL-;TI6y6KxFXWvcF<k{kkH5
za&J4e9pK(lA`~1pH-AFVW!Ru&UGps;$5W+um>$C>A{H#|u>Y>{t<O!*PMW0_*t*TI
zj|Wfmr*sFB#C|45xfN&swmBtZw?9;Z<-pc-OiKx~ak>ezU|Izo3MuATJ~Y!Wg+b2@
z#q&s#<Fl6O7<yi(&xDUAZb&PhsoK1o8*M3`U}g^YrbeJ!R5RUk3rWS1wKehFqK&f7
zE@96!n}TDsdpObb=I8n|Op1Pd`NglJ|5t^T(aiAyu(9M*Q9biDVZ6~uL$fVp$;R`o
ziTE6_MG06{Ad%2aLt}d{d&i~%$X1ARD)+_rwJuIwM}#wurkB>JqjjZLf5VyJuF5+e
z_q#W@D>y+qE9%)}UGiN@=unzFIl67ECV{Te^JeJCZX4U&BJpZ4^8VV12JU)qb$+CK
zVCIUgIez8T5$3U0fohzzEMqcN@$&wP2l>PrbAhU{RSdJhy$R?<UU!b*s5%DvjfT_L
zm>#rP^wctPBwPmi>tKZO*u6UB>ySK44NK*G*V4(|YXj#F^);1{H!z0`eA&m!u|XcU
z+G;%VD4)Z;w}Tep&N6J}gT#MC86b|DvT5Rw`H@lP7U3jKG}G#hI!9clI)mIZ`&HTe
zLDtr$u?JxKC@eI@C&oUU-JZYD)Va3J9+@2KHtD{#J#;Qega_h)pRh<prC0n|Uh>Io
zmGv^y%;TP<ul@`DQ<PW5THO=h&t>R7+Klz2^R`rydfs;;H<9P#&psN0m}19EU6$`-
z(|wN~_OqYYFYcLic44YKCAYry8O}=nJo|jd*R3D-xUV-=Nq08dA~r2t8aO5q;cKw5
z`HA2s1yzTgosQWVM}9v!+HMQBbNDk1EEM9bMuS81>ukXqDYKYeSNA6~lod(`J@4`q
z1K!!sMrEI?mZGOSZ&IDc!i=1}YLHxi%64$_bi;f7y`7!)7U5s9Yq_DY$#9B(MFB<?
z0#Bh$Y}BR&pJ!*(U4$GY_;`9ZrMzv)y!b3&(>wYiP2>4Dy+UKp*!U`?yV61K`jm9r
zTtU*qo2zpgjrdO%H2ufg0}b|)5G<r1edt#4T}-CC?siAf_%qi&F5UD6*Tf|1!Sq|h
zzb;tWs{Or_A?l4>78Z=o#m=<z!oz!UPMrREPuQnP!6OW?ZSzcoFF3}V-S(Y!$iHx~
z`&U=M|4-jL>`z5qdyx1Q{cD~@-p#vS5t(GS1nDCxteItBs{6xbxp@@P$Wo)@e7}Rw
zc&LGaEPXw2qcL=0RzGr;LhE>in9ciIc^Y>kLhEp1!|FsIKW=Q%hNxY?v8;E<SxjTc
z#$4u`38`9O!*>s`7BAMuW1Gfs8|+>tZV5V`+0(ms(GJo%Y1PM32fd51l3KUfGuo~r
z_zVaG?aOyPVOE1g1}Xi2h@Y2pH!mDw#t>om2B--sknzhjO;qSG8RLu!;H=3<F0C~?
z$yMd;*7~X4&^wPPz{9U0QN$ew8fj6BgGF9DfuqHCh737KH+{a@dB@lNv9Qr(%SA71
zkmbFNRmSWd!dk(KXf?*(1!mD}2-ag+nJRIH|1-0ykD6dxyiM?W`Ry3HHye)F2UT40
z>wPT614iJ96uKcGVX}J!Y%XKp+nWEn@@sXWsEB-O!FX6D+u`&NElmwf!BW`b1mMu@
zGm@1|EGZ!#xajRY;XQsk3kX#?pj!)2fZeOlO8Y-^c!)skt2)Gk-F+cKBK>~ha5dx=
zJoxt2<2)b@3OA<j?bYDVfZglPK3D7LMx_v;>_EF<$SmDRP)Wja0aYFuIZOY=Oz9V0
z_Ye@O^0ct<$&;tbsI1kyt80XR&~KB=>HcD(;?rJJG>>n@&Yxh@IOXSv!BM-$$rE;#
zNVkE8{HDl&)Ak{9$+$7WN$we2@K*4=)1xdC*m)?}%#fvkT{@Zwx>XZj5PYx|g9*co
z?ycTiLWTEq6FYxNBkponb-U4A=285N9$}wmGxu44|E>5->-7}qGJ=Td^F=SOk=H3N
z;+}Ppw3hMDx=O9N5pjj2Nph&d5SL`o{pjkAp<luLmxcZQr%lQH!<*lubKA}RY9Tj-
z69(_rPTlD-Q~h!o)u6(Wjl0%{<z;o{hT*Su`#SK)X9Q?_#Zj!O6LsGrn;u$H4~62+
z<6s+Q$?cqMRC=XC?dp8wRO$4fS&V)fJl^OHf@IhkSU|UEXVgBFx53b9I_+{gvhwY8
zZ)OoKO3eSsw^(AUz^7w`CR-Vf@HLbkklZs<D7A~>*2R@%=<!9Nz8~84dgeMqxDlQP
zF+R^1?QEkub6<|z^>5qmR=)JQP4YaNo9D=ZmWm%`H3q}PR$#{GUxv!{6W2fmEHp~r
zx+mdHh`%((qOO-HH{%8V!a}^iF5>u>R8G*%mrjmz)Bru><PgFAsPWG$+#hgl4t8BQ
z*1W*-l9pF7MrYIvt%z9%L<W)Ha#?>?7<vj@^3Ua`3D;MM4YwL+x@W?OG{A}=CQ7=Q
z8q}G@fvYi57E*T*l&=}jZh4OT&t>>z)VDzfFIGH!*BpLoN>lAzYq+Hs3%XXRo~X9B
z*6$wCC0E$r98?QAlt`-=2t00j-$N}zPHr66mo)RRO1k(lcyVl-y!dfX_2kni@v?N$
z&TnMO!r7uYP7kv+jyLQup{$~9bTu;*CEaUSBM8vqitj-%y$P_p(Ty)cmu5DsHoS#c
z*K=uU(<U#&#P^h<P35@NsK38q1ErKtnR<oq!ICgakuzG5Hb=h@cC<}jdC?-fGc7|{
z>tZpXXt2(^l5Ftx+`>{-9(d7su!V5*eX!3^^5oGQ%nA2Ox9aCmOCR#-FbC5YSBt5e
zPC6ZROh$i~TGZx`;$x654s4wTm~kdn4%RlPDmT-nGm{$lhud9GkH$RU%1BSUt2>ba
z?I{}O#RP>iXJLJQDBQRlEj3!(z)ZA7XFB8*#ce0(Z`hu1#I7#54XA+quZZu}1@<2k
zcc%QgzI<@*@Bi|2|J@w$r+ScW;?b=mi4EPcc6C;DhFb(52E@)34VVV>Yros>J2?oA
zX)SFEK)D#-vsCfw;<18$V7TWYdzmmwV`X}>bJ~HRjIk?o@+f7fC?3^N`@yr;hB>iq
zs7o_?uW&d`*nteu!j`0=<7#?k`ZK02OnpjL<l-_Ev48)&fZl`}l6prOGA+}$aT}IR
zsB#(OMAMbjL%x`VL#8|AYz_n7%i~lc`kZ%g(eitYW9Gb9P<uPeRSt-uA<Vl1$c{c^
znKY!6Atu_Nu}(O=yPU4GHGRTCn653|+U1w^zBudTbp03KB&YxUV5BNq=}wq}f^DZn
zd^H<n$9(@^YkBa==ASn0d|^I|39=<Q$@d+xZrzw3B5>NtSvNp<TzOrIH(6@t)tsww
zbbL?*RU-VN9aZ}6DzF=F-8z*Cu+QN4Kkn&bA8N*wrND2|La|@=s#|>AbI00Kvy^6p
zwNb=a>8em355an?bhSdq7#JdQSv0v-D13m(1wGxq8_$Qe#ONd4kfyzy%_W%xoOCjR
zcR~z8Fw0^!A=jPyGx*1j)nb~lpFRDbw8W0I#-N^i9~d{d@)m9W?dHbO_Qia34LY+n
z_AI;gNNs*mPO#f(lsJ=Z0zCi5OnPkxw0C(AB#*R(AYR@kHN=_&OiJ^?B(DVfbq9(F
zh)Kqp0T=%R1v<?AaD}->Pw<VI4FRUz(7XeZD|p$1DM@Ja+_Fp{ST%9R87T;@<k5J}
zL2k1`QY_5WpFm=$mW|j*QGjH|cyY<I%Lg~2wA~bg7PqQywp&B<D(A5~gk9!?gh|iX
zMNQetz0H=pwG+DtJtZMsD2DdIy;qyc)GmcyF2*~iXR*!#!!Ho0%IP=VDt=@+=5QkC
zgQodnhxLw*=^@cBMKrZeZgN7r+KSbhq;BswEX8?5c!nboNwt#z<LW`AF{SHUCOkoL
zXj1mdt8jd0?|d^hV0>|-<6A+J6+fzp#q2@4f!oKJpdOn?aoaa!2oRQwv(LGX71A`!
zNaWzV%BmU~k~RUk|7xP&kHD6P-q_u5_qe@sZspQi0ggzox~3KD+6QXIw#J)2lcqes
ziFx^u5utRo?OKeq(YnBl^maJ?UF*qH)heyTw-O##E*1CJ(5v&o)IN=b1i*!YuKQ-Q
zvCg22+Es-s;>7`|`85uiwRc@Ah@5FlH0e$`hs?^u55YY2wQ4?;rUA}}pNLnjc4LEk
z1e6KHf&5YVA;yUHA(w-DGw3Yg%X|uM+HpuT{Y3>Dn-J}BdB<Rqn+&CBf3_WfUy8zj
zTr+1(VrXDk5+qF!^@4|p{C+k@w%!<h*<FO=!LaH^NYTu*M3I4aZe@}>IseEZg(Z55
z(+P0fsj2>*DVl3@-58}FUh@oxUcK%CG;x1aJ^i7KoISRiM0_}jy%inFHqdRZ3B!7N
zcTmfd`d~U{DCfK4C5qMH9p(oWANOQPFsZ3#*N5S`5j{U}B}eBGBAxUEPT4yMcb2Sa
zCJkQlqTgM%sLlV>(iulHrGsbW6}@`dcE*{(lrinYbQ%IqV#n=_CoV-VK{4733GqSy
zm`EcOTbE?AMQs69O(j@%lhDj)N0;Y(SL&=4g{k95B*s|B6%KUY8ER=5T5etM$ovV7
z_t_L2{c%dT88S9rX{nt=UB*D--Q6FfqPQqqAso@o<UmcdF=&lTJ{N~Zr$ZYGy3dcf
zz1m;9+Z0a_k%GE;w$5VfiIng01Q%;J>B({PPpk7JnD<l7I8d*Fp_^^v)(;QWCe2oE
zj`Bx&T*7=CMU>$A-6Hh3VAH@(wNzc+7uo>YA-3f{#g{aZlPn`=xBKx4^n#hdf?)h9
zj~~V=Yi@t_yZImbHzx9VGqBrwz<+C56tu*2TpcYz6|L0z23e{-K1&Qv9>Ms^8;8dE
zOYsTVLU9AM`t;<TqDt>7@FFpf-0WFnc9qzr;-9f2?9SKv4&Mlm94hv9tpvBB!XxMh
zeGB4t$&*AUg_7yaVIE6^r3)o9^*()wy6I=0psKEuL343V(e)Ig6$zU3FfB9jr)|A7
zARFS_MXv_Y+6IM;k{FqE61yiS!*JW=;CkCIkt^XI6X^~Q88RpM^_x=}XUz{=hY2Ps
zte>Ic`>@XA(dijo+K3S<?#}X2yNB<f&I>f=9NO}v>IU09DsX-K!}!<LKsM;z>0kcd
zGI7l=k&l<g9klvolb)IL7GRjC>WUgmVqIW)#Qgi=UJx4S9fZuXNDYrAn{ZW#XAhJN
z_V+)d!>(dy_J!F_;J5_nmVO$T+GRZr_=6|r0d5dynkp++3mUJtqZeC#{$&C7_A~rn
z7LHfh9emWq7aCh4dEViADyih=w*gl$2f2f=F@%I|mLu}<^_>_8oEW)mZU$5)ay@s^
zWWMx11Vj}ItF>y#9UT;6MMX7<fQ*cc1nY-6J9~dg^;THJ)QLbS%(uXc6DwbT0S~g`
zawtd@z>{CqY3oAD5$icto{X8MY6a7ADl{~Ly5&4`CSjrefaoZ22z*8-Q0>%esys<g
z)U=%+*t$^9yZvsgKllZD9vYDY5ciqmtPeEC(y0N*xM6BDz1-$r8(ClhfhH4!!^&Gd
zYJI&k%TW*eGh7?9wJs2Qw!c{wb<b7h#R~;k^XDxX!xSC1dYV5^V0EtcI4cdVpD=11
zLo(?u%FbWgXN91W%(L!n!#^ayfP22!a5`R`E1w?_jp?ZYLF``AH3=ALP8yU6W_0w7
z;O$P_D<KT}xnB}F?*;&x%+DZzH1ii-Cg$p0zoSKec?1S{Y9^<p*Mxp<Kzh$_9~_{M
zX9otHFBAP!>*n9MG@{8bG;J1FLeOZ2T;;nse$HGT+sp%#fcK#&+5W`9bg6yb{%9r(
z(Tsdw6q4P0VoF-DMR&JHd6xCH_!8nL7UwGzXxc*cPYD%2Ig{;iTyMB8Va-f<UoxY_
z?<H?!s$>@cI)I2%i9K;ZejdhMIe$k~lC8y>j|)2jVI<_R8oukmFyN%2eG|3>Adw_}
z=Yf!E2WRXXA6e>%Vn){_d_SFdTNr-y?S!)j(%VPO*2rnB3)M*vtaRZYFQvW?kFFU^
z_q)W`cpGba(LK-0{#mYVbkJMfYYOO2^c0droNP(Dv9dk-;}3+$eVWVV{b5emZ;10F
zGg%$4202fpTH19p#4uy<ST!}5Jl@gS<UUT|Kb@t8gXD!p{`41X&;sQyXdF|U5<alg
zzEkC|Em9E%TE?ORJYP^%Zf~EG*FL7w0*b)GFy7Y_o4dht(zQn;r@zCWdz~}+jWqpb
zFZ3_?-UAh~!KD9Cd`mYOOWl2E3tF+U=}E*ru;G6)c7`lity?=bVk=P^f4QBQ{3&gf
z*XI_rteO9G3_oC9)LLY0$2^gAK5loN^<)uipsCDi%o>{8Egvp7dH+b(VUFv*v_7}H
zi1~^69~Fz~r;pa;IcD1%W3C$Ifg-OqlDPA$0t>O(0+_-%d%ZVU0s1fmgRMKX+upz*
zn_n<C!B-bbjKQFLN?=Y#g4{(ZC`qe*ZNNwgwmb82Pp`q<p@iJ<H+&yh_gmZQ=`jCy
z_d92nO(xtilgmF{6G}I+HX~rG%IFig-LGpBg65?Ow{4Wu_R_LYwJPZ5OCB-)Z<AZ0
z66L&>Wshu^{`lh_Jid@{b)s~$!=Ws^9i;_0hiq@hcN}gLi@he7ie2b-WNJNCD6UXI
z12aBr+E}}D26thJzP{qvO6iXDG>B9)=TJ7#HOWke!whJK(obv1FCic~-*5c1OVatP
zIn82`rdtl#6(9oKXPH_s`H5^kY-`Dq60ZeW(A*1xM|ma%nzyEy61Cv>MDzSo^G5{F
zUnuFmV??1g;{ph2%0IR$8i_XaF7CMSPO3`6TcDNo<{lsSkoZn;JLH>Pe12;tLL@7z
ze##V8L#%3s+rpesK5@GE=JxpY(@1P_r@wYW^PGP9!a7g5(IyV9Um%p94UpY(&@XMP
zbiq`Y-Z}?3ZWFX7tK(NhN9%ZZxH9WuFpF0KawR;(Pfh^wchz~gEmk?Fc{_L&s;;jx
zGiK7fq(<?lP>9FtZcI(Z=fgxNcHy6P-j@uS%K2uOz&mXeycU5FYck90)p|~ggr33m
zIIw-npm*&5jXS~b=@9K&I#|ueQSwQrv$@fNyC24YrMY^NV|t$q`<AACz$KmB)z`Oi
zS_7GOxb@R88>ymg-<V%P1Etz`8E1o5+c?&d<$|+SddK&!PSB4HY<=J_(&X*ZckAFz
zb=nc)TVLt1GPA??S!qF!XRoj+8xh`SU3pzdM5hKnk;6DCFSpLGwneng6wXEMx@tQm
z_9ntzc_~SUJo{hfcTk79ukIE~9)=VKdr{`q-qGn2=OaQa*k@XF&by1n*mSggHp9BG
zY;Zm(uKxA>ho^p-)#?@YVJS~*hg(K<>FHTr@mcpdu#u9J5eiXl>rLn7**;WxQn2ek
zKy~hrRB`K;?HTcwK=o2_5^9WLiwFuC0Bs7aORc5_$nW%xpu88;qOmpKgn|b_1!>&#
zG5yyRDZa(LQ!yu&aA7*^`fzj2_K$m<(DQHR1v+76{27~`U|vivo=Aef+E<7J!{FU~
zzsS+HOcmpFT1`;an}DTyG|eVq;{0|9m+$54RvsISc0m<Y9!$Nx*sWC*9FX=^3ykZX
zR|Wj6Cat!kqA8Uoi;bdQ7KemRwPC8l5{lKjQ_S?3Y?{&JaASghplYnarRFQ}BcYhf
zLS``V#1RbmPjT%|PUb(MglSdjA2Zta6O~j0B>FY|uy9m;*is#_*4NjE1Bh&cgY7G^
zwY5dkZbx8u_H(y>4|@DaU{=Y++p|SLxGUgtpDuZrX66mto0fuDfZE1lg<uxSskiY$
zBkK4-<h9{_pg;Qz%Apg1_Uv+D>YMgj7<8mk4|$LBBk)MUrW1n;0Pjwz@h>d#W5S>}
z4vS7JF1Y>JIWBA51%Y@kitb2VK{(Ec5`}38dvpTnk0g2bhuLvm_j$`(t<Q&b^X9cY
z^#baUn9q%`2Aie&)@Yj?mvxGdZN{u)_bua(k4m;H$4z%O8>0K3k{>2HfO>Jvrd;t4
z>T!n;-R{2INpZFT()NgP1*K=Cs!v)qJqI!u#!zWGUAfIi%y#v|mZ5o_MZtE}@XW$`
zA<{X?`qW%id;3kLnO{Wv-{bx!GwhUe`F?{hOt%aP7olB`T?t!XuwN_gQ#aNVM&?TK
zX%!B$rFL&TlOd^*?ytLH-?ouS8mHvF#pRCa7p8`0t_k35)YD6}xJm2}%#rxxk@n8M
z|G_p66-m?+eO(g|?L0)C6GXlY7szMYVDZ6SQ~jA{7R!40Ze(v|?9KJoy7g2JxJo19
zR2zqvUBD+AlDz=i0~5)09PyO9a6s^-3wCg_N1<MOTypWV?`XfLThsrux$LdStqg-n
z@|Zq|OX5`c<+<Pw1DjZ<`mvgs6*KQJt@EtqRYkzXFU@c!4UYyM`K7YPwv>~JiC?eL
zc1z<&7CFzh#Dt-wZ>&+S#1(bph1K!C$s>A5-GYMlmBY9~ixScGH%>0Xyd<rP^)^sY
zt!%Q>i)8*<GM>Y>ZDdhuQxzu!GXJsEt}ermdpz@=WX>4|GCNU8ZO!!II!jV=V6CVW
za>QbGETXwk&sFQTwrR)3DNCJ?d&Xa*NGMbeHSY_1lMlW2UbhYerhKDtZ@<toc`DAb
zq%@g-rgW6byuX-TO$+g`akBR<dn*e|Dl{#dFN`*J;-2B8_3$P_Ee-z~zfVR?fJP_*
zJxeNDnbONp*&&^K_<aZ7E<-P?!?iN7qv9uZQh?nfy$O#pA92S;|D`JYg6dYi^>8HT
z_T0{MPM&2(cy8PdsuZFTm$Xa@PvITpOJ<%$B&MzJ^hmU*_uem!b2L+jwii+mkeqgl
z-Gf2hcCRbRII{=xw!S0%-H7TB^<kRcY`lPK(;~5(tn>w|jJZrdKVD(h&6{*)!M2st
zkON)u^&PMYK+@xWa)SrzbuqTrnoydY+bN4x{AHG}@M~o|C0cacuqrKuYqj#lIOaM0
z#P+UlhdKksRA>?Ace15wop6=AI~|P?mfHQ(X?S)g*v^wM-w#SkVdAG^I@1%1w2qYM
zgs5py_*PQQH$ObuFyp11;UWt{1|B6ocQk85)(79k#;ZU+yBZwSKhAlCFbQbRQ-NHf
zMp6@=13JGp<vqsF)9V`*f`S4@{v(k(&3OrJv`$%**SnVvc4v`gX`Y!3`~nE=`n0en
zZY5>%Y0XTme$N_JsuChbao}G9COt^z)G*7dY-rw;xJ^8Oobt_-9jyCO-2QiZ^h~Mw
zTkcS@x>=}84FcObQhC7kFIm!rUP$)Dl`md<^-AIM^*Dj;t&W#9ZpPQNzDfMb7AXta
zN-8p}nP4`G0z2BrJywj!5C2}NRM1DQmLDxXup-YhIq*z|qKG*Sed9f;RxL|xJ+4gq
zSyH5~2hv8UQVg4b`0li2VLBleFP6gA`vq6Tj_)u2Pqsz>pIyLz{O}(L{DHt92>gM-
d9|-(`z#j<wfxsUK{DHt92>f3lp!IR=e*u1v33dPg

literal 0
HcmV?d00001


From 97c8556f377b7da52a3b4adce489d3b27e2bd9eb Mon Sep 17 00:00:00 2001
From: "Sterling G. Baird" <sterling.baird@utoronto.ca>
Date: Wed, 1 May 2024 15:20:28 -0400
Subject: [PATCH 02/12] Create sobo-tutorial-thumbnail.jpg

---
 docs/tutorials/sobo-tutorial-thumbnail.jpg | Bin 0 -> 352109 bytes
 1 file changed, 0 insertions(+), 0 deletions(-)
 create mode 100644 docs/tutorials/sobo-tutorial-thumbnail.jpg

diff --git a/docs/tutorials/sobo-tutorial-thumbnail.jpg b/docs/tutorials/sobo-tutorial-thumbnail.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..9351de700d332fd9abd9b0f888568d9519d490ce
GIT binary patch
literal 352109
zcmeFa2{>Ep+CQAVyLU$&w2Ic$%`U3ud3L%{yRA@-h$+-e5F|oToZZ0?({9y}MA&6Z
zNQfFDk`%i&)m$Q`2-_^i7NID<^t}J?yzh5i-+Nu(^*`sl@BjU--mH~7&sytV&-&f>
zb3gZbrW_6*{s}mD`{u2ifNun({EadIa5(j?_bqdC_q+FOZ{D)D6dV8mzO@hX3PygX
z4gdth!x8tae!Al5<b37W0^obVH-Pg1Lx7Qo7c%t5-MiL+|CF`{1?LKgrSe7Ae|p;g
z`+-y5K1eSB;G5%u!=_%Lh;RY_vw*jb3=jRnzYy>uUj8250=`7RD<cFM6z~II(mnr;
zU--iN{Tu)J3y-kBXD)cQa{~T~-+$me{{!#kj|di=BY1MakHJAGK_31_U!FwB2Mz`a
zK4k@)BmfDx2e1PC1h@jY1+WB|19Slz00Y5Z7oY;r1gHwW0sfQgzsQyZ*Z@KV*Ln%|
zp@1j=0uTW31NZ~N0m_0@lt8+V;L}%-?j!gM0z`emUvQm({vB7Oua@TDlD-l82>|$h
z{P1w&DgbaS69Cx%_3&`-<>BFeApme>3;_58{_od^{00E%^9AX@{(IVw1pvU=CjdZg
z)4!*A<^TY+L;yg1B-8`p@wfXN5$wP9^#uUdi2%T9Cjj8WdjR01%fDeJIQHc@plApH
zu#XX_sv7{nU;zMm!TD!h4><yizS!J9LSKaZBiH}!+QUBp!pBdX6MAvv8)3k=!rvSb
z{^qb5AR{o=BLdSGY~Or&3IUxsdFuGF@4i2J<l8SNz|INMzx(c+Z_fywIPv{=$G<rZ
z`1Z(m-+y!T*zpr5FAJZ)WB_q=dXSS_#~vKIAR_vMp1wmwRxNGdqL?@$Jo0%qRaQgi
z1~T(S4NFc7jH=T7$<E_P?VrOEpH;S9y=VXEr8&s+7q32VpT~lGNq@=1v478rATxq|
ze7VuT<t0$~8^AY5zWw&d_h-+XJpSF0Z_fh0`I4EV$Apic|KWm$L)L)EMNu*FOR_h>
zkQdoh7W;&zuD+e4M>r}or?!7kPFwFMP*}wC>LD$id-k4@ztkX}y)6FG{L$}~g>SEV
zz3G#;Ko(Vf@P7R9a1<bP<QoA@M}z@pfW|+*;PREeBJdS~uLyia;41=O5%`M0R|LKy
z@D+it2z*80D*|5;_=><+1im8h6@mZn2pr+9ViMkX^WsmtD{rOeOL{*3P~3xlOh`OC
zkCqAU&0={Z@~0E6gv`n%k_j;9L^Jh(9eTaftf_Z%V`fL(%(et-z2J>W#r2$Jr$YcR
znoWsa<sr}phk)pkIKv93@JADUTL&Z_Rm^;6pO%kk(K9~;#8*SFEX03ibs9-T&#Z5l
z_Hy-hwwdV@Y|6#~g>cpNJr%})%D9)_w>t!+Xh|0w?ESi1&~f1}d-?&#p1j7){~l6I
z6y9IkrnNvL-$|RsE#xor+%%eZrq`dmH|@E6V^U2%YGJcx%B=oi>l2x>^UeHf@@h}k
zw4`mmnIiDW^5pN*=hyx6Gw#^P?M<wOpc!kwN`Dpd6_c-a^0n-Ibp>DD<JT(VYc={c
zD)<^HeT^RfDG~Mj?&DCy)0)axa>^N&k`MNV&5!;*GhEnq2nbgnuWXY!1Z1D>s0(g(
z_Jon|uQ>;#e`Zsu!$}A9yyHo6t@OgK^)#<Ln7wBup4NweS#Cl6Ez&47bRow49^!Ne
z@DSiPifL0F*qqp_PnedmIt0vQkB@Me5(QY@+Ud60iPceQVqgw`+uAuQp1op2Bn<Jf
z38&*fA(p|104~D4y0GFm`p*d@uFEZVptSzv_rQzR*ivDy#EnTIo^V%=sD8YgxMha<
zZvXX9BjPmq9$?DwP{8gIWCh|oYlg&&T_cnn0?OnRDJjneX;2;30sEq*;LQcErM?p)
zCK#mh|9d)u*Qtn%apgQInH~!MgD0Og(4@dQzdX`p;;0gpUFD!YDJ|@Rs6qAl8d3D$
zhzd`ta$w{PagDdiKk%Z(DG_h~tPs;Lf@*)*NZ_y{M@cgrvXKR6H`_PMvo61^)u95I
z)c;{hDZPxHh4IZLSNRoqnv+-4=)X1hR8s<?gI(z@Wv@A+>(`g(KTB&5&81X#(z2K+
z+aG@p)Vn5=D2?5N<@~_a@p4;5)IYb+3tzJ(mFen}ep{IA9~ejdRQ*Hm+e~7UxJEhE
zIYixR&6<#9r=;lBrmd?sK;@Y;m}x65iPaBK$?gu!{N!yFyw%+lorn$U<C-$~bz}@#
zLn8x&d91+!j;=s1+#X&Wiyd~3ch0+x&8H#^T1@Z1A$4O&CM3$D&Lo&UKQcsvE!)QB
zl8o#kNs?sA9moncR%}$%?p1ibv36wa-GR;?_|`*{gXbt4*A7hNk9&}phk&?zhk&w$
z&AREQ4pP!j3_@xeih8)!W0bW9oh8Ez)3S-Tqb{IhmaE9mOA%C{A83e4Gn?a9{9HM)
zQf<r(?kVaA=~AeMHiIpdolB~VhEK;R@d;%Q1I1(A7gMo~<<ghaI@G6P^A&6|)L`=1
z0Lf&osWgj)|FI+MdGQUgjW_Z|_E%xRx9;Xb(T?(aF(Z?Q0IJZKT+{Ja?f445di2$m
zT@_a83Vx_w)dlUxT#e<GIg%JfJM7U>@JbMj-i0KkPmaykYXh}3v{mD!^^-ALV_KX0
znwSIlQB0b>MchfMcrRTBS>RqUIX^lzw$b9Av{e|RdnqMa;;+O<MTD&Y19LN|_G-Sj
z#vx!dCr!1$GW*)AOi?Lu&&(%fS1r%zzeyEuO4nCwomhzKiwxo`SpR8zdX*iLb8n7H
z)6mly^(`nTsjfP*Dg5cii`__9+aPc+^_Fj#Lfrhu>K@INpwK8zr-cUO#|^L^>u;`q
zmZZlWQ<+nF6F3;MWs72P$RkA~(fZv@8ZaAV{horr_XLiZop2*D9(o=E>iZTp<2RsO
zt%TQxebEibAgQe1QF2+*5y8+?M@l>E*&C|gQ!3N~a}=K%#Jdhw(RYXtwx+@MUc<|r
zm;(Ox9X9e1KyIJ6#e)nD_!kJh1-hzOZyrkyr|a!_2rvf8Tcz5yrJH@4PS-yKtT|}A
z&|qkPDm*1(A>G~%j`c6hO&>-F5M;4lA1o{)pLY;o7_4B+N9tI2b?uK}RxqD64{WXe
z{Kx&#*(&O|ny6tCtc>Q8RY*RQ_!DUa3b_GFG?z{_IWdswW4&Fu{h{eNL*Hv{@-tV+
zLrYy9xdP1lZ0yQ0CgSs6#~A1`V-DW#Uu1ElGJ6bShl-iBAod^?F-t1i8C)Wviw<hK
z?r_!{mbdMm%iqF3-&U&=D$*^G)Oe<)yd+;Z?3J0SAO6FZVtG`0j<}(?M}l~Gy^x{w
z_J-nVm&dMGh88GkkxyN92TPb2#ICU;bcYA7(Jw~se#8*m(Dd#30yE}>jmEa_Oml7?
z@D0TaXu@k_BbU1nd-$Gx<ClXNiB#GJuF009vP&!5C{a)}RWHp6Rg~g&&uyCU`e3He
zZadqBO|z`j;UctJykywIW|&j?oA-f*ZnH}}lYi{LOye4K^-SR1VTk#az6Ezs-_(R^
z>@y3H@QRh9{qRG`l>AxMNd{kOC-U`(_-MvK+ylq<`9pvke3B&%IwowKBa-Yip7;|r
zAydfXEhQp|)v-*A!l|9zbL<}s=<#YF#5#QpHygKk<<QY2<k>XJmCg|l+lz?34Nr~U
zbglSH`;_#%dgHn7dI_W>Tiu_W<Ly>S->eVOU8!5^)`mG;$IO0KQw`He$M>I#okFB`
zXa_qK%-cAeez5H;*T}B;_CSRM%ilIEgUEdN-S0Nt!N8%==c+s{_@to1eRf8D5*puM
z52X3ts+`O)ACddL!+@#k<M`ud!-4u8JbvCoRD4nhbak^{+wsk7)1sbA7;b3m*yYW!
z(fPb-9<vGN6A%tE>n0pf3m37?L%mh<B<+)iujws1$i<c@!sd%sQUK9$bZEuO&+PzL
z;_?o&S8byTPW4ZB8?f=-*n5=~EIz%~B;i<J3Ivr8nfy>c7>>W1=~QKNzoVfkHC9OW
zZH4w|z=nK5e~hGKlhA|So^u`<jUunTa)iQ~5;;qYVd0e{QltFx$I11f%6!Jio|qd8
zmEH)+YYXewf(N)f-*%Su&7Us>ch*B3N(mUcp{;~|6p4S0>_{&jugJM>SbOy|+;wMT
zX=aotcQoH^!bR0|8JUDhc4DPD_t$&AdV>wmyjOLm#HSRma3@!LTi@sNyl>m3U#CUR
zO(QyK5?PWOEoX$ip41UNq@|P{z0Qa?iHdv7EsWqAy%u!>Bv`OjlVTEUa4CS9V=+*p
zynt570?k2fB}6~7E|@_@fF_k3j^wJt(Ne-<t(U@`0d=Zm)NIGW8E`7oq<&#F`#Rmq
zU@vc{{sd6&um5<V|36p4F*&$xoY7xKuS)Xvo(^%%SbKSl_~2U|7bwr%sYe^E?|Mkm
zYq`dq_d*Q%Yc0k+bA)sd%3*dlL`K}YK}yG8B}eN!^-ZKbgf+q^PV0x!_2g9|W749z
z;$g(#vY2P~Ebt{49Ju1xjY>p@kCyM;OR+^XY&C+MJ5#)UbaRyBh9(W^yq1)BX6vls
zyB5(y$p>lZeh8adg@*UWRES1D(~?|Fhw<PkS=V2b7d#1)FnNDH!LKZF<oA29n+$e~
z*Q+<WXHw4%^aZrut?RAlF8EEq$H}64jeHE>{g5CI3)Z*P62BU>@g)A9oO8#9YH{nT
zoL7LZLMdWvaB?MgKu4*BoCOM$3lDT0-2AMZ67Lj|3OwdiY2VddHiGsfE#I8xNu`A_
z97h>fqjcKziDpg}btCf^I?Hb!Rq3E`3*#7Aimw+$Sh{1#VD{qH(1H5vn~^dXQK}JR
zbJ3j!2<D|%mM^riv8I_m3yQ>Wu1;0a*pE1s#hn}J!d^o6HZ14%biGZE`}N?2)IH9$
z*M+=KDLQtlNP6p+x3TqPSnmGqBA4-q)io6x@+M+7C$R5y)bQmv3u{ziy}~iuLd?8w
zNujTAPy($5xKqLBb&eV24)ZrMG<lhqeUb`ewiY)eo$I5M4*{Q3h6r9&z8;sUU=)fE
zzDQ;3T+E^(3HEF-yBA{z0)Y&TeQgRrc7}>`y*OR{!gTMEXDyOp)2VhX>H5<tN<kyl
zX)tzbj6<1YwWV&^a5WqkQSMxrOGt?*=b-&*mLkm{pa@8WWo=<$@zVqw3{fI676tzp
z5{ZN%F$jKmFxY+rgUKHeRIDad2WKJg{)q+S?BGo5As~NqaQ_arX5Pqm>=&%!$Avy$
z>r9N=C5L^h4B&UMmjhc8swP5-ZNx|73{BV%2lnS$-Y-m^(Z8j$PoqUT;3KWLmyf2=
zwQ+|4-e!ILoi~HpOCQ-J1L)g<An$4DGRbRXG{1vju+=47sZ|}mw>k5t&#Z@cAD=Fc
z5euv6<Ma+gtD$1o^24MSOAi5k>PD#PLqK!l4U)=0&I*XSba1bz)%60Rr*8OK^~HxH
z`W`}Bg^$ggyk5a1M?V#N9K0p0VJhsOD>Mo5YIjS*@JbQqKN>B<H(jSxUgLHndOIf@
zs^iPSmn+WlC9u9GVn)76WVj!zG#INfReErK@J!Y((X(_C8d1g;T`c7J8^b3?RUCMY
zr7nXX{d1Gs29;THbYi`IYcQq~uhVVJphDl(TY#5IG}8{H$}>YzK#9Mw&E5=}w0W1U
z)Ry(yE#HHw_0JB#(sM1%q_ZJameHLK9j?5ft>)Dy4{nj)W!a1JG^SUWo1wST3srnR
zA{Xt#s{NzLcPHcOe+vWMq7BhQY)Ir7s|Xg{V2GW18l<}QG-qrfnaFS))2%`T5{%OE
zKrjogMj7Lspi)(f@$nb5LT1<<O7+4H#YE&9a&>r%*GU`99|wDYErm@xKDU^7AMm8<
z{(}uZ$xgi%{5wCu8XR!T4=iKe83<k&7I=QZ$KsQ}*R=$cm0|+xZX^hsn{vY*WjEw|
zU$|-yx{XoQ|6%Z8bMjzkvZvhbX=6nC)W~SPR8;Eewxam}y)`>aJL`$o)K1-i&Q?q=
zRx=D%pAmsArEo_}Nm#-b58IkqKFz}X4s&$Q)?Po5I3+)5Dbm$uTHZ4Y1;J+`Ni@VY
z3MLs{nw*?Sy7t{a)5`ye1?C?Yb?g&H(-jlejgvfIeEX3>ro45!_|?vw@ATUw!`h_v
z&A&Tmo}ghV;y5X!J}Gb^3F4-u1kxe_eT~FeB3&b3Jz)aRkZ;cdYV^QJrMXE-$rwyh
zQk->*hL)DFu;iRym8@K&mdWb7pKSOKwwB`q+@Ds}+I2mHyFgS9dTkSK^&4fKc5|S*
z`h>rqw8c?xD_+;6k}e^OyNfBmcZT{Yw*`@MQd1J#2n*Aa25WzWl(O0mw4zcVgEb|S
z4xR0*kq-|6<GFNE)Wa_49a~hN!1t%QYX;K5Vr6b@rT3d%WSvOsV4F2nkbh+4<=VKB
z)w*H&g*u(vh66hpcj%UmgA^<M#aWnJ@8uz-!T?HMyv_v89zkg_;RCmW+cJ`lF3gXu
zhz8$mgO4I1BD{N=HVL+rj%|Yhjp@|$vlA$p%gLB69!Sbm<L9~?!S#;HVnT8lQG*Bz
zVRNO^?H{}rBR1iaQm?JYRw{UShH?vIY3S4A5E+MN+A~`v64`cizXQ)qhT|&5G)NtP
z3{WKl2)P?QvvxY{)V+e#y@@b3f?ak9(0`>gFPvBed@v-RyqwfnFqp^50&218qHB(o
z?gsOox%qadQGEW;?NMe{Pc3CStfB-5fx5ec46ABuDG$SdcrXNPC)v)|xj<+UX|JBQ
zP7iuF-Ksx@Rt{^)s~i6GDO=fd%4@i_rIl)#`Y<Ta-mx;ncg#|Bd8AtKc14AAg+=V(
z=xc@80JP_|q#5ePA!x@U9l@ZuQcAxV1B^l^QBYto*qRLnnN(|OX_-`g%82}@mz@2%
zyv6Kd{{#P`9)@j+0Sl=0^jN(WW?kC9-p6RgSl>!0?DnH2&+?aLtrqWoOB@N9!iyvi
zd-QkbNSa8B%q@&cfMR%JHBz&+HBv)`-iH8Uf28QMU&?qx@&Tk=N}0*cR?8&1rH<{r
z89On>{81%>9V4!|B*00;hld%7jswd$J$w2rMwGA7qwvDD8hUX(-|iwvi=7XxG+B6+
z-xGD4N$wj>ae#2;-D+07uC%uTl|OeXor=3PC;|OZzv=pk^$G{ZwJ4D8o?zOaae&%R
zT>3@?UPym%-ta*Y#!M<8tZC8VLre^(i~Y&6bS{(}Hny4*%mcGCE0$9`S%Z{d$YV@g
zy`SakkV&c=i+cT&xWk-h>r=$F!ZCLha{mzo4-_a9HcYJxEy~h*CVW$x4A0F?4~yn%
zq-C**G;&;v7@W=D3E#lD-0?HEsr74sopg$pA?pq!2Rd9~ft!iO9tw)E^6KH@7#(}{
z+33a6UwsXuuf-~ZX%DkPyG1Vtc_p}V%Cy2--7fo0Ct8dfhR43EcRK`(V34$u`mxGw
z%EZ!AjG3kqxzyLsp^XZ3hIXOpt)`;&X}>S(MN`2yCHtfT+Uv>9CdAkN1npRx37mXB
zHQLQj8G3s{aUP-!TP)B9S;WSU-CkE_ja)z|R8B01IQ;RCJG}oBB|IL#VjNkEscPTG
zo3*3Pt?Ic2kB<};JVRWZOLw#F&~t)nzYgG^x;xuZ#|cJ-v;M>>RjAG`Dz{+*l)Npi
zqG-dR1xnRI`wJYtv3_S;$^)ZJH>$DP<TwP;7!tZ_e2N_yJ$UsYIfWS*1a?z(gW_uE
z(94Th{EFJ`df1HnoAv=;ZiHw|%*TSDm21q$iS~iJ>;~#Zk4Y#&BMjcAKa0l;Be;xG
zlKl0yO5kTSEWLbGc>_H#G@c;_Tn$0{^mLaR;>@n{e8Q3UAKwfob1+7>b9Pp#P!)NP
zN5Ej4`@QEj>wgSu!mH(U7=YEiOu6Rp1m&yduPNV!EB`5GwHt=4t!2?DtDmF&N#*?4
zMhSGf?7oKNT0j=Wa}Y<HFL1`f-q-A#?ne;mv)(?|sCkj?5D~q&54db3|8k}F=+1h7
z_;`c75X-E)Q5Xu@Fo-~IK$iwA?FMO<4VyRbC_8-k7pVT54VCSaOJI~#KO5>4MCXux
zhq$cj?qW7k13ERC9(W+c3IY^ipq)uk?{64dS|tAp%+z(&x{@vw+9dhIdRXO##@dvx
z*Yxa^j}tv3D=ibo{rnIoR)UKRa<7iSuDDf4ph<<Z%okwGaIiKTXs307Z;2YEq+~Bq
zxSc2vR7{cNPeJo?v9LmtuM&X-OTxmElapsPj9JrF@Vj&CwJfOu#id|tqtbq?aRASk
zHP3mmsNz)Z@{)JI^}2M)4O=H@G)_s<6DN{)2ngv$F4m_*FRD!SD#-ZeP~ax})~iRu
z9P(BI?~mr63z7JUSP`pTRT0~Q$Jz4<a(73;(8`QoERh{a^xF0Cm<R@Yg8395VT6yY
z@DF35R(7Z<WcEU)wZIeE1h*G~s1~+rf1<<*jMc8lUHX6z7~&(st+owvWqRk`+A3%)
ztNsM~Y2$N=R@V0-I+&uLk?A7zc>Ai~^}~Pd!Mo&Wl##NcYHne4?dz)L!4-`@UGhCy
zcXWC3jh9{^YJA#yQY5_>rb#=QNk(u+jd5yn<?n0I^C2)7bH@T7Z96af6w$yp-4yI~
zl6JY5j15~p1;s5o{~BJtyUz3vI})@O6vr#_ZT9K206#09<p;s^xvNNUI@K<Ez`8lA
z=rVK5pbMlY+YoLQUCDO->%YLF{+mS}Nsmly1Xe6Vu6pD}o-WzE6is1Z331?ex~W&{
z7j2?&wtbH;KO|<(*ygxE6gW~>RV>%1HAk{mkaO(`8?m)by2n4(e->U?YN7m!-CTI6
zJvwq3pBGbGz9f;N3;Sb~6PJ=lO2PKfBDjuQ@JQ3h@_<hRvx_4U3v_+MIQu@Z#_iQ|
z9xVbcgXNFyDubaqHo6x}QbrTJSrbGtQ4D%9ou&k<M+8s1kOEeR%A1=@FE(psq%TA!
zenQ%8CImMHGmb!4uqQ_1mB!@{G^<-Een5G($f)(ilee+APwvLyPL>tDcv<TlMkJL|
zC*YSF@-Kl^b~pWY^TvMsGb2~elwTx)erVOU>LdblZoMe|G~xg%NzO25M#?dr8#B`)
za}a}6Y*n@AEIBGVPB$eYu%=mx-&)?nVAh_=(OI~BH5R5Ib9s4Z5))Ey8r)q}3f6Wd
zE>(ub;3?%`=X#`4F8WvP;lj>;k@c9}dZCpKXLa@Z^w7_0%?ah?X>!BK`R?5Y4r}iB
zyL9iQrl)HL218PymMU)_k%VN?k_sT(EUjX+qDG2f7xrQ)S~{7&nYVu`EWi-Xe7Ts*
zpMp7?Rz6KD;bA@v8K9PZJ$yLzY0qF3JCi%LOT;{oy_NH|+@zMyjyLbBbDn0#U$`zk
z<64=S7CoD?1qY2&xEwGDdeMqYD9NJKMXhITh`N40R?ImB)Y#84B~6EDcUR?p9GUgN
z23L9TuHGpqb-Fwx4^;IbWS!A(1Lv;$=$zhy8Yji#lze0G*1Mm>dHr6C4gpwVz67%<
zF1TFgE}1dw9op&EUD1<DnlFbsY*GH~qReuYR*;fqyyz9qE(!v-1v&PIskpSJW)Ih2
zQJdbD(u)}-%XguI`zTY+8QPrmO`DgAa9P}DTZoI?d90eH_SV#lD~U9?-;al~P9xvd
zngIL8cCvNkuFM%o_xg4_-CVizpoSfe7gAtZ8DhQd?8f{FUTG#BL(uPadPX4Mbr?Op
zs$*aJ{C<Djhmn}-L{8$}QQQQ4k>e7^i!{oQ$SC)4d(%}tIIp|Pg0E<ZeXL-8R#o*i
zfrKHJ*}%_%LqS78<cnxHkr=2SfjWI%PsZ$&tl?l)7@QTsb|n}Nbg{Gs=oWht*<>`_
zg(RQyf@N=QRapAu!GE(G`RCuhJVd17t|)c|$CXYDK1o!A9W=1^V12ndN5p5{w8^(9
z;vSM%%snEgs}*-wtG?D)<FSO!cgu+S^u`^7gxN3(&9&+Um$+%Mi^Gx8<EW2E(j2HJ
znN2D9Ha#FG*7*Z1VZAlKyc5i}9Q9n~yw@z9h2W|bHzw}3Z8^hLbxmzlt1bUJWhPOx
zU(el)tpKM%T!`hcgGhZQftR(_V&BX%ab502N;0!9%*E(zxziEMNDbeN%hlxF0?N+r
zA%JAql3y)f1Uu?aLR0uyB(MY3F$)&6lDS`;B{CNJ>_*VbtaOn<rX>S-mBh0AjwIo8
z@v0=|-j#(TzCrlnrfwQxMJj`~Szl4RnWEu$<mY}3#g-PTR3>62oF&DQQ<E*h(WB@(
z<6T{i;HXQGHFpX#%VU*Df*JM%?^}Vp-}UyqBbUc0$j^+?A!C9TFgtkCERg9imEus6
zRLB+S6Sj_eTl$xOzV7~)#4|%9dgo4e1WDLk^}`?ULGsG@eEW~{!J3WMBzi!@OS_+W
z_qS?&I)5E__vqvfB09ifxibI{@2XAm!l+K5vhLV9-2r>(Z}%~u#>z!>$HXT^Uu4pU
z2jQ&W;N6BeeI_wzg4t8bEQV$J&al(K56SKQ)xom?PSwDRGm}Hn+>}As%FFugqaD4e
z-D0YwU<g<XLv+Wk2J8%2`9Wlx!PclLBb6CvT$Cui%MO8sk+3!2Pj<J&!zC@&%Sb^l
zLm%6sI{JbVstp(1{OBO7+=U-9h13Vf*Mpmec|ZESeyREpdbUZo!#O6c>copW-$1E^
z;9#MPu49`83&<Af)aaH-)N2?`?QAt-uJv)Vih}5hp_|2!j*aP2dfH$;G9(b?k{;*T
z$&3_a58jhi0@HSUhcjTl`jqoq$ZLP1Xx$%!!7k|a%^=*19!p{2<uy;&SqtB(=HA`F
za28?=XzE|>8Ca<F;^vR4ivKb_{<jD|^1-l(<Q!RuIi|IkUFLUP^ag7g7mSPOqE~1P
zcK$}?D9*MB)fq-T3uh?>DlN+2Dy-%B&)U1fX_wJ1sM_bG22wuO$Ev0_eSQL$ack5_
z&wL<KdUi;&<ZXXVu$pvPY8R7l!8URB$`UObp#dY;t15;9#m-)4FPAX~!<M;B3(=yI
zKY8L#m$QjEey5I|ALG7je@9WI3JQ^Mk6*2z9n2FKv4EyP>0sRz78nJx?sG?n$UyD5
z)=rt`Tcvin+DJHSid;Q`4A+u(rrv?93{Oj<bFVK35MoSkwu>3WiidXUwjTl#LmWO8
zKGx=5DWp627^*&r{$bz7r0DJO{<>dUC7j`(rcia&c;W)p?xt_wh=HBR7}Y_CT(fQR
zreX?gJnzs+GmJZG4RXjx27h+B+;dYROePJRKC^K%qn2=x^%6WpjZw(s3{YKS7a4qz
z@rMtqR(_ZjHkZ<;bxBV)CCkgs)uB&R<X@;}nd0`5UcG?vfLHUWiI_FIp8jlLn$Enc
zZ(cR3tjfJC(pkR}o!>BoqS-uivIHrqX57-v4;S|O&hCb&{u{}=E!T#$Rm6}ByXg~S
z(KZDX|L6vlF)7f`>{rWvywL=yh1p06wX;gxPO;@7NydO{VabvOSdt)UZU3mb{t6K^
zH#y_h+}TPO3tJCch4}JuFM16FX^3P3Ln(u4-A0pI4J`RUoIUBY?_APp6gvU!=zPvs
zB3{0udw1;Y9V@SQBa2n4&-0&iOGT=FBi4Ecv7EGux0E|?otD4-+6JQu_R+mpk_CJ^
zZ_ZUeAY>O;t1OP)Uf`$TD}r6@V$577Y0g_O>vZ|(=%Enrl|U*Rc|9~h^xFJT{~B;5
z<|sOV(#`c{LmoC|f7Y2!OS@MVdp63Inp^@Npr%j|g+0>2u`gI5rb&yOrgloqh=`?r
zj<2s0QM81hI$9|HD&@Fmg0gIavN~G)Rhv*~icsjnWV-m|npc8;%qOqJ`gECtdNXTl
z>t+G1=jOL3r+>DN0ZET(-a7c}zfkM{<)V*0S_L65Yt*F4bzw&+J3Wc=-a)kM$2IUq
zk_z07@2A4)l^jF^xcZHVPUPqA2kF-B5qn?}{kGKei`QXub{s`AnFm)mr4MQl^h;`&
zKW*whB?le5T#OV9I5El>#XW_Su63}fsiTEKJch$Xy=&y_nx9)Vzz!dmSuv^yrY1c3
zKgTZH46`(KVgr>@t&Jb7^;R+CmYcg@KMb^6UYp_O=F(v4;~ivD0L?P;T2Ev0+QnG^
zDV-9H+ZAZ0?bh~-)y#Qbc3QuyEXG#Q<ADPz82ELdAG{=XVODF?y+h3n_2sBO(T13<
zRHl+7OieFv&Bo;Xu5^%1${y@`YiewJbG_GjZv9Y$nwO=X5}J@gU5JU`4wL3`Xq9Vr
zK#--Djc-L9`!)9-qj3C;bit#C7E=!8jYv9V=*|0&aIj%qCh#FIIX%49E!e&nBBx98
z^(qK1gDK%0<VK+N(MnDUwSW9e6{{llhMCSDx>4D|aT)jZsB^C|7Ya}|3@po0TvUau
z$Yk|B!ai2@yVSTsu3-4cop}zdoqk{P+WXeIrdtVOLdD--RdpN)u`w&1Pi+2qP@n!s
zy_ST{#0Hly#jgD5IjD-SC(nY%@F4TTn0(U$rmC&;oP(Ot1EYi)cC?>_6N%4tk^KZ}
zbzvw^2X~Le&F;$#JF)nInc&V**9g!}Q~Rx<7#rdX5PZIh`Bb`TWU57bdi*>qg6h8F
zz*#}hhtxZ%*^5NFm`t&pfMHZ<dF06`RJfP~>g{vy-P?;N7oGY^0oSkjb`wj9IMwjv
zXPxQdmG4D;rG4=^S;R$gy1v0D1F;Vqq1DRJmd+-D*8=CIx|RiQ_QlA^N%Jp`HhS_!
zpi~69e2joC%}sqnjo|J@tU1x)E4UgEg3ST55L^z5b<qyN4Yy`NFS0biIv_2eeV4!i
zG_(k9{!6v@Un^X);5Bz{A-|>O_FnlYH?}o>$SO5rjGR^8_cIR5Kc(A4#G8)1(9B|l
zk2r$I$m*Grqa=D?N9{;(6|6!mnq-9fap(QG>GvPh;x?a5j7I!1Y#1Ne)hL`*2HVjd
zE4s;XRuK#4bPn?KNQ#u8scAm1up1ZV<`4(&)OU0U@o}fY1MnkE?KE7`_@GV-*CiZD
zOG^uRF*EN2)B(PBbwO$j(%r9NNtl#~;1&p+4HW6B{_Di{Id0UH1UZv>!2$kY6p{5D
zg_6(Z>0ql0jqya$%a~lRRP-gJdgdoDO;?|GA_h|(i!v57nzhE<D{ilKTnO4SvbBqq
z*wy;5R;1fWtJQ@o;810|%5Gd-uVD)VRo_=KO7K+mhFA}Y8@j20$B4K3!dAh_#8JvD
z3X)agJrBE-d!e2-NNu3ruBIAg28s?0+29FIl+tF_GFxNWKY5mmP_+OVTmrMi)WE{R
zqRV}-?uIko_fr1F3KsqL&%M?P!(}wz>wWCQIybFWmg=^+pdGVq1y}3$o<x3X1vK1$
zQ9QZKjx^4YvKU^{(zVQdg1k0`E<6OhxZ3gBE4*s193IYDkofK6_K~|&3THH4)gA&a
zg;x5mEM8l6jJ7>j?&UHYVmnC6FU|wT;p{E(*F%T2SqRxdd^y-n#M3W8jQKRrwQE$Y
zY=Oa!MCx{P@Xo%`Lz>z!3V6A~vn@DMeV3)<?a|x|!9zZJs&3zL(%b8}#Js+E)S+`^
z;>WpDHi7LsRlOlym;5w`17}>9hd5UUu~45{UQpe=HqCrWdR2BfeO=*px}xVJV>7I>
zM$?%PRJj<06Wu`kEpQhO&nRCQ88L*Di(&fP8d^1<KO0dUI|LyN6Q~z7Q5k_Qp~D%T
zo+W$uGEPfJYj<lUCOK}<Mioqs{J5?>1nkr+;}WjZ1RkJ^1r~S!?JnDE$#H4QF@OCR
z>+io#{0FHXxZlqa$%Q*jP(D+$qZ1>iXt(+4joJxhh0}Asolxk=dN#P$%FsG8wL6~l
zc(k_j;f1Sj&i)h|`<9T<XC7L2&o%DC2V(R`a_mB$@vtE{n(*h~8s8JP<nl-V2Nj?e
zzHI4IcpoB9O2LK-iRf%6VM+NBA1lkfeP_J9+@8awnu=Fq6|OrCYRbu+6)gzNf>f!2
zFN{@&m2V_7pT(6dx)jRGEs^P9rac6aH}D|g<2UTpA4XTg*uj=~-S!tjp9V)Zad&vv
z3HHO<6N<o+1aIwo8m%WZSDmsyE^D*tOsy_rCir1pN2ZcJ>6BDei!T)g?86noI~KZ|
zx`OewibZD5Jt>zsoe|@0gu!Bl4nw6|D{jUT%uXrxh`ckX4V6!!j{ea~VUR22?sg7n
zRqO0G&>Bh`s5or?!|ps6XpFNj-VzB;6>=gel-hb*cBOEg13TVfJ;_6|va%_<%Olh`
zK`D0k|D#^b2F_VyG%q10(fZS9@hgvxhoQYea&5;w5+r|U5-&t6RyG~?NE8ZdRStbC
zX=v8F0TzrG5Ofr43;J$#z*@pU?PkFcUABp160z1d%1UwG<cXE3op5h~Vp7t4FK$S4
zbpo}dw<kk45oUkxdorF_8G~m9&Btun;3g7J#6J$AnwSCip{nAC0C)4Vi+0fk=j%{~
zsA;DOj>6Aa)5cyP`{jL9RODPqM=w=-eQS5Dv4R6s_KzO4x59*{ffNd2>H>Bm3!}E_
zkw#$T96ws+(OT8y6aDu=6%kp(mWekd19Yd$bo4C~H#S3CZGahqfI4Ph(quL9;I`o*
z;OEs0{jq6<49QDvxy!kYEmF^5o^dlR(X(2;AFN<LHtpl6ec!$@{(cYTF^S!~8d-UD
z`|gLSsnh9FGVhR)CSUr342{Cs@4^lNJCmYzO_FBYUWry`mkz?e+bmpQOJ(*Sdvplc
zoQmiy*!ObIE!eeg;=C9A^uZ|nL8ARSWPdUA<FOS<Zd*!idGwFbB@CU|55{8q3PHUD
z+rPL>|FP0HB?W4?A)Y-K*VjGvfC(Eav?8SxK_-s8iSLi-+V$T2zQOnmSNaFG@r$pb
zzGC&&7`_&qub$zn7yVjIe63Qy#tdJBqOSqu*9iH4>un<I^W2BIsT1--UYJTjr+Hnx
zS>;L_WZo<B5Kweop-oJ%7?x?`4ZhXwGif>U*NDHL{XVKUro82JmPP!$8_gQ|S#qaQ
z!DH>g3Lejh=luq&7R*>rR_k)D)u~U>E}0+KjbXT0V;sX-3a$}*WNppjro@eD$&Ruj
zrQ@MZFEbiN_2mU4J?c*3!MY#nxNjvpv$pJRs-3U(wM2Do7h;10!QHO=B}oN~=3^rq
z#(*_<l$kze6$kf^#tX{~>3BAqw^XZU!y@iH^6r`@8f;v9qagicVMF(Bm#CtTSKmUo
zyd{#>-*nVO(om+ORVIG?5YSWv<+(HkEK%)j5*fOy)p`cqB0T4DTnXD1F~2N#L4K3H
z9b>R%Ww!WYCV-KyV9@zTcWcsG&rOv*<3<YHg_yJN{@C9C%71*n1~@)`WF?ME=&A<N
zf;mqc99(tF^Vakyy&DFtOsk4+R0F5Tgo1}J6Zgv&YUcPZhDxc^UeAOaotAuidPJSn
z*=@NcTB22qhnrVgrwPxpT2wCwY-!ZJd3nEio;P@n<~bo8o_gK2{7-M&SA}|}-=~Cy
zSPZCIM=_<*+Sa>s-G;t5F|iNKTW)XV1M82t)`Me0RF}6U_}yT3;S85N;#SmXdcWS4
zU$xA-Fh(XJ%B|d)eQ^g7OKA-sLtZkS14~x3$;B5{S~&=seS0}=ESO()rL>6hvW$=(
zY~my3qdfxRk-V)jvV|7!<U7*!<r0*mnmStA-?qPd+uhcfENIJGRdtgGlXG|U>H6DB
z)OZ=$mQjDE?>?>u0$(A&03o;?utG)v36s>F5~;|v6bpt}c0v>hv&`&80z=gfssX;(
z2i4MMTiSPl1x?>Ab{ZPS%tL_ApL_ogr2hwhkw%iEsYJ#_0-XinyK8Z(;Q8LCF1mSj
z8=!ew;sFaP1HVotoS(gwBZRt0yS$h?<FX=3x&{`LXc?rW4+RrSE67A52DDs{qt`uU
z)Tm@t*W{HWk#OKot(e~f9=vGgzU!#h68wM$TYI#vRWPQzD{rim&uMwvR2{FTv{DwM
z2M2vaf;}+QEyy9kqC`KchX%3Z<@ek3%SCxxg>rcC7^PG&&&&l)W+F%`90F;HfaxZq
z-7m!6nldOC9$$<<;UjhI9W`syS*JKt>1C^gPeYK~8zTeX;cEozqB_5^PI>#fRjuMS
zyIUT=Z3!{`3QY5{RlLYWuR;Zj6>?vUC+L^JF6@x7-s!xcv<)V&+9aq9E5~qozX_wT
zfS-fu#e&gX8o_KJb2Qy5d|)G(Lyu^c7X3C%CMh~$pS?$gvcxRkQW)BJUDzSO%A!}P
zYCXD|U7;tNn@sV95n3e6oJphr%Iq-1r(^7|e>avR{n|7ul$XnfrS9#%Zk=A|A{>nI
zbFt6Z<E#oN%92pYG|!~kG`-WxN%uM0L4R6gcd^d9k98;E8uh~LogB%~d(6g%LEK+{
zhYrzN7dad&OX?t+Z_-wF-~ST|0yKfo!naUMBIEB!a#^nfwGjIwxrJQD5;h=r-{|%L
zKNTIiS#G9xubrH8UEFb~in5NG9VDe_L$v#(cFE){kv0QT06`v<LOTHFoRfkiOoX%b
zH+$g`C-biM6kIo&^GCuX9zBl`xBRZdx|I0Ay*wA}p(ZJd3u&Aaj1-cSPQ@l&wfE4p
zHawWJV%scmt4Zz6qav#fiZERQGh*qzHl{E9QoV(sWUY81z(gOxf)56EnMD85VqAec
zoU{Ve>7IQ5=}0|azi_U2KJFJ!3dr8q`=*2Yy%^R_-LsLj!C8*(+RQZDHktw!%pt~K
znyCvmQ}VdwR6aUihJ6$Etl!e~@%4MN-3lF9&%)KWoKW`X-t`(34TrTAJrVEolX$qS
z-c}?M=9OVCBtGE|?oGwEMO41hkFBh|snS#9+|FKI#C1Bh5qGkj-;q`)`eS~NTwL9i
z23{CPd*v7~FWXv~RHv~kAcH8V10%}1FU`gjljhDvm(1dxb;ME#Q2OU<Ke5i7cGc9*
zIqAPX_g?$$K1l9&v{E@x&@xA^?8TjJOyvzQUPgCkf0j*GoApuxije#@+KRJmRfW_a
z3*F+@8$>&Xm@e6rwsv!fqs$)Jo&mDWh#fq6%?`fW|E8}h3MBOm8o9(Y*<A(OcZ0&h
zS4BcMtY^n`Yw^(~2N6YF{w-3>=Q;e%Rd556V_aH@^%_k|npqRdEQ&TKyt<P^Gz1fN
zzI82$zIAWi#;Ws@gKLR;lcairtLSy;ck1Tvw;7J=O=l8J&LrMQ-`)B&&Lok@V|z{E
ztGCJdR8%16vBRJeiSlrVZJ6L@!PYXV*0Jt2omec_&@NZAO-enjcjTm8mE481GM8md
zXDo5y2thU8FTybXdpY>uXJqCSJ<#GW(UO8u3mOTEo=rj?Xvxq8FCj;-zl~ZLp3JCE
zm)<hn*pm13^z@bgLINhLxw$RK`sWYE`&6ll#WF^6G=`>fazIGwRmzDCt?W#1>o!w;
zK|jz|u06L*2X%hlJY)U00@d&f#cfYCIx#{IJVeDL;9)5#F+Y*PD>jN@r-HllL2&%p
zD7cC$S3{;-&^?rfL)mGJfEgv&>jYnHDc0BH$8L@b@*cYA?)&hdRhntdEHBsTA~JW4
zPZ^<<kU8i=&yxJ%{MeFDY?~vlYA741>sZ9nQ9N|<WyW!}Xb66z))r%x8h4=4@3Rm>
z-I10IZF`w2ZrhPL`S@jj)0uN#sroaN-rYi|%GOf{|8xZ~{X?)d&{3l4FYi%|5X4pm
z3`v|`J&~N=z8c*^7fi)nLgc>WtDgF>bPqk>BmM^yVhX7l>*{hUD%cY>s(%pn$zU4X
zxl3<u(XB;*YYu9m!EmmgEv{<CLfjg2Bb#5KFsg-8v3@X_QOO#Zm(0?V@2%~4q#V!B
z)dbJ^s(bWL%PLm3dZ}-Tu^)|PU$4w)J~-#kC~ow!u_0HYUUNnHCFvkA$631mCNo5d
zo-{v3bbC`&Zag((ff5~pEN~vXLfjtHHOs(~bE&Z|qJv-^TPo1$0Wuj}mwXWJ!7T_R
z_fk*9r2c8poC)6?I@LCXz`0bFGf!0ZbYZ`bN-^tcwHNnF7k=|xq0dQ2H{7&I2qfPI
zd=9(j9bW07TPU6S+SiXJWLlD8Ns;HazS#_myuWk_dOMi+(CW>m9$Xb{Ou$n2qAu;(
zBrkP=fO@DU2;9244<6%*v#bbpkN3A9iU=p6QUhbFpA|wk!q?N0J?e{vJw?Wa23@Q~
z?KYZy1Tsm2N?7rrL!2tIJD*}V0!wk0h#GFmV(`#)ly}m{M<^B4X>a;P`z%e*g(ltn
zhZv+!OM<@T8S&}&O}e2Qp?pc#hbv}kVrIuw*<v4&>J_7c7^e~;>tF9B+X7J@7p#{9
z6jK=jUj%V3Xh_vo&62a>5E!4a0;SpaA=slzGz7Y+iWFE_hZfAlF#zgVYDw|Z+lC4S
zKOy#as@eYbjK;<$d@U=&jv?7A(k;^nX|GyHD4rn4t|&IRDr~-RI;A<LRJ34D*k+8p
zdiTne*!5N@C88uDour6}jjYwo|E-}U{st#z`jq4J&<X8aVobpBw$5q6k~i8*Qp_gX
zR@&gZF=q!IY;5+zAC;C69c7pBZkNO}V-wXE10U@7I=wp(3Ue-(U)Aj!i;~<Vvsr#t
z10j%u@_Gct-oqPWzjYr<iE^adufL_+gY0**8y?Ue`3vV>`TIls&t?t*xzkVdz0&n3
z6U03PlV;ORgw&hP3YJM}lbLH0GAj`5ciyExER-uSeV4Q{-&-#%J)!ZJI1^WU4Q>y~
znj@d&YwWjDeRJcxWwl5q&_Tn6p8;BVW>ZLRY_gM8^Qve!Fx6jNaWXscpv^q0&N96F
zRZrbL``M`M@r(l^#6sp)@Jm}!y@MRlPsI)49SXL8bX5yyR&Hgxx-|OPkl62*yaIP;
zEc3eIu%vfyT8S5Kl2QW)sW8n~Icdf}Z`zgGjQ6c=yw@i4BLfFXTbC(CA^3~#OIAKz
z)=o<<&9glz)b1_{isfL;p)~njKmT`aeS1gO<1Hx0Wd;|_<VEBR_(0Vum}6Cex+Pvt
znKkn#9Saw%K;Gm6OVGofI`hw^51O7UV-ikW`0y6|a8X<X9i(5{2FJYqZejnXgXk)^
zmYT-N078K&HOzE6eQwiYom%2Vjq54+ZDX0F;NHtpqKDue2fV?;7GqMyO;X&~zFe-6
z4Ozm5;+0=s()}P;%IzV6tOK@;tTsU89s>Sco<PHv>0n`wC~cG`pX1B0l4r+`*?9Cw
zV0xZkd1hM6f<VATELgVoOg8*nzfWE@G$D|2LpJ`VYnVH-9$dwfcNJ&*j0PZy_q>R1
z>ak!qgOEEIY@|>P_rlZ57dt2}l^Ead^6uBFUhJOyNNTW)OwYsH6wa*u&4S;YJDVw_
zsUBhkL$UGFwc1^?V4$js+fc|XT$16bL}7!$j|nYd94Xg_(Vtm}Yx(V~BPbWTuvDN(
zbN7j@{uSx|G~tfTk?U6B9Jzpqd&NA@tQ^D{hpO2-Hd-E!k86|`SostcWr1C^gbZ~q
zvekP8Q|D7bFoiCRWLOKugj-KbKVVaqIMd;)r@3Y0OP103)9ej?-uUlq9ZOu~#@Ymh
zjgDHUWx)I^VYz4(7XmsUsXjU)i{SrhV3Ow?Cyh_BjcygyDGPKJtU-lU7R-{eKLi*u
z=tEi@`>w9ezJa&*|85NbFk|2s#=JTjw-Q{*p`HD0G^P|;NoakJ(|{^};8eF%6^Ms5
z4F_Qknl#z?vZ_X`VzG0@CZY&^BFy@cY37-yW0UdHpW5D}k`Q%1&A?8o%C_4|q;+#o
z9=<UrGB3t>pH{-k`E7n^BB?^lrXH%HW2plg%nu*0uhH$9r|I&!V#n_0S2*CdRU(Z$
z!}mL#2IU8RazBe!^j+vMXkLFBvK`YN*`asYh!8Mh`f99dc235Zlv#Y4WA}7gg+2Yg
z)mTyUpwMvzv6ijx({!xTj^*oe6o#feT!?MN1ovM-$4UJBCy^a8SQ#=EAQ?R(&L9Z!
zm4{x}o^I5D)y70uKPwGRGuJFPGlgLp+p9e1><rQ1K2yK`bvPJEeZn=7(tz7VD)=wb
zl-*WH+TCuHr3?sdpt8Vs%&odZE=y#_sU1s*&9@Ym@^PT4RCjkS9|Cl}GHGH(BaPOp
zTEw1A7@h0?vaBO2^MsV7uT2N0pQYQVG#8+L<2^=CLS?*G$qhHWILoqWx)i#y_tZ6H
z(pDn;!H~X3{U&rM#gd`w>*+CpyWEq-HL%p>r=8MG4feW4ArZ+Iev^)bZAiH%xduTA
zV}=Gpo2XR}+baT`oMYQbp6t#SmJdcGnHy<2u)x-L1i8jB_>lHC84a611B3dQ{@rrE
z{ghX*s$}*eJKrv74T_Z3;xFCSJdjCoZI2=^3@r}kjCkKrx_pTbnr=19j%JN2US*Aa
zOjm|qdVKAMWqkVO-6dT(!=q=7Ew}KP=R~jg*qF$0M@|lY9BAAp^}h;QZyUkUOlQI1
zXK;2MIxJ#qwo!-Mg7(y?;rfzygS9d;@I{_lW{-@2-Z%st#gofsP}Fiw!OD%9>#FQ!
zzKLPChSY5mDQc|y4KYhku<HAXc+qQ(8qf#?MY)cB=gMFH{v`jS83&0Mms+eo9(zWS
zxb{A^EzOh)LnWSVJL*>h4yQ8Q36KuA)Jw2zLcS^XvEiZ$BSPOY1NbyJQLYCc_6V%w
z>)V;*46ND*!jq(PIgceWjLrM^2u-8WFLqD#BZHcUxy%aN`emw1Eybes<}bB~@{RQN
z%96mXg%%PsTN@8Lvt{(LDOnofF;o-o@|h}Fis2BDg-`)6Z?2f~rE)jf7d}h7#nM*}
zV)G0MGnP6tj!a8YsS~dRzrOl!^bLojHaW?_@A|GG8Q&ALz$0;wYQgd_wq$4O-kBWl
zM(?+-VR6#kcKri4>_Q(|zf-!>85Q?@8w)F)Y9%~xH_dnDt(P5>VgN%Q7g18?3xsJp
zd9k*#Es-^`Ma7XOR*``$YqoV1w}=(Zb*O~VEgeW~o$JGwEpe4&6@DD1whN_V)I|&t
zVqKxNkgEgnL{H`6@5sWciJr(=E$+H6wvaR>uap`iGTu%q@Ug@Y<UN=7&i#3|noG2E
zs)m<4sq7UDj)vSXFTn6m_R7QYvz@8pH?)QkZWyo)rON8`j|)C+whw!xd1?>9?-Hg1
zF*&dBX^fjs7#6X1<9cDw4<b^kk?Rlcyn=7v5#7X7@lXkzeGADqzk$hE?#n2+5S>iI
zVdIkL9gc<3_37b@NGmMdXT0mrd^ShhseNF@u8xY1rm~;mVG75j=i{^G@%BT$o|vS?
z@y-p)RjNt<*zo+=6l$hD%T~!g`F!!aXR@y6z1uUt)11APr7C$PN6EfUJ>djm_KlGG
zjqlDi2|}VL>R$q+$8J=?N)k(dTS<Zr1`k0|K*0{A1BRl&T7pqR_FZg2-SUMx)<M=F
zVW5Wa7m_TLHIkK;tCE#9u&|0b1T^lNy8q?>R`@^T^T<c+FGQy!;B`FbhsC&C9gM{1
z75y5-*R#gF`_#FhdSvF>=Bq*H5D|8`R$-2)iE;9?_EzuYTytG(#GJX1qpj}v{20#2
zscFo`)@@LCu%76;T4&hvv}Bg&oPRZLY`FUk(@S;?EZxcoptI@L-D0vNfkV}LrH(Ro
zsdG-3*NwRPX_4F7Ss5bR@@|W<MPY$2bT(cPEIUyvukf5=e8WKJ(kX*?%*C5;LqvAN
zu#=BsgWVqcxwORk-e?3i5%HQ1H+%F9Xr~6P>v}VY3D~||;<#purS)Rec~!dMh^?sp
zN@tND+kPziVwXyJzuM8&g}yoRgRKY2zfF2Mjzy!SvUwGu{ts%r?l0BVPLuVJ;miky
zRHnTv9eKgElPOu*`!GE2Z6{&%Wfg_>OJ)aYu@A92<aUv_f=M-IW<tmAWJU0U-ASt^
zV4yY&d7i5j9jKcKTjg%>r|(c3Z8mIj=Q$#=CqL7$E1ALku%FT1?Fdx|-lu7)({t@E
zshDS+kU<z}Pp8LP-<Ade-$Gx?U0w40c*G}64}O1j<rjKQ+6YZYd_IljzZ32~lJX3w
z<J!_r>&Q!$+!Kas&jySPhE-Qy&!5lCfXx6kQZCnKko=v7!nt0@6l}VwroA%qV;quK
zUW=YeZ?K@vxWkvb1ctz7O*^<K(C9k#e)eDgmnog){$K39cT|(<+UV^Rolz0Tp-P>>
z5v5D-Ff&p_2#87O!GS=40HOD`#{$wfA|(j{1qqNq5HN%!L`D&!KtKoqLI@0ED4{7;
zs^@WLziWNpAMao1b)WC7Z)atNWD#zj=f1AHUFCNz<lzGv&JLSl3ABLH6=PdVKapps
zz8$dk_^gj+Gbe#Oke*^TS!@I^fPze`voKGmT$NK!_1uhs(x#oj)?f>(=#{1Mq@?8K
zhNyrQhL)5E4_gT88H&l((##|0^vTg<Jn6P32y8KFa3s5J$3TrzkX>k9UmslDI8KQy
zWtm><8)|9N*RWJ-_nMI1TGMXUa8hmHQ(QgJ%X1Q7OHY`D)nH^MA!~t<MZib*Pv^go
zIu~(<7lVa_D&AU3tSC0GF509!)z!eo`>M)+s9tzaqP)cEgnA{I8eT0l*}fk<%XAji
zYa8LJ249$rJeWN$R~6kfKiX!Z^WNSxaCWEW+IS)*j<Y2ypre#<!C{;vqV$g8#Y(K+
zVxmUOV9jFu1{mqP8a_bwofNIU1!dh1Ur+W~WqjFR9b6aS%oM_rAM^%LnttGcMr(Zs
z&a78$CL`O6<;`qF5b&GXO%B;jncn{sgZKY<8=~jmh_&zI9<v&dq8*5^AlE64;WmGk
zPESnK`y$D>;XQ#l<SbaT^Y)o%7OH2Xf`nnzych$cVP}6&{k5l7Vgc-{g*pe($Uj}E
zLl?5?0qMI=aZL7Fqhk$SA8WAt{z0gQg@CY`>zFtQZF@<`E`GACncXDWte>-Nsql6X
zwHz~eB@D;QY~GrUr4$0pS6Wg!E*r8|#<JPex6!{V(00A9jA^`qNhy*Z33GJ-LArEH
z%JD*%xDn9h{zPKk*_(9z0RGQ?aUr9a^`p<tl<KIHjrvilix;B*=_7(cCPegYolrgD
z9vRrM5m`5X@zIq>HHFHe6azfxwSfiqb!+MfJ7Ob+O3~}nP_$)h*9n5fzB<N|4B=t$
z2s#Cc&s#`SUWQJ+?0`*Rw>7<~k#84?&ciyJUP}!7YaQN>s95WXGVJDn3CWud8HKFn
z^#o;>L!2F~CFk)MD^)5>@9wKs{_?N!`IW%=+Q`IW;BxD@3Gq(%G}P(xFJTXBaIT&(
zPgj!@iC0Zb+`n_E07JnRepK;|wcqlh7g~NS(Ub-PR=thSN@}a$=;`SQkd=~Q_2uN`
zWZ!ph%J*eScrj8B7K@xw9IVR%LBK2GeER1=qMu4;2}j&Dk_LD&Xe+KzJ!fliz!?Su
zz<IztC-!O40yugA0ZSV&y)U<>og_I{j~ZQl99;38ssDh60R0-w*LgZ#x2W1+jNev9
zeAkb!=kUl+W6YT<yqS_bu}d#U5(yz39hHCnQo53W3YGW2+2L@YX;C+#km+6B3BaG@
zF?b@1*p}EKq3PCwe)O9c9>M3yYMoT(PlUpiF;&F*3<C~dp%hPe(hj<wy``_7jF}$z
zQq#XHHz0{e&CiE`qh!O_P?zQ~s}~@BVM!nc1$Wlw&6+~Ng}U1O6*x<Oj2e-X@lPF#
zVu7r}FYJ(q#cK+7y?WP|xoIi_HT|Qvw7B!tGvaPnDB}(HRB9I<<QY|AbyOlB=SA-n
zUg*#ATPR>k=5I^Bosgw_3>zvejF3THjY5jBXPAr&Ha&NJxFw9!qde<~fb~*VIN~py
zt)O0P6km@ANEZ{7LwNqIlN}K<9rU5}gIoNDHC=kFwYJHZbA~5Ah~(=HIym(80d_jT
zRNjvScu421|ECANs1M(UwKc;K`Q6l!B8HDO+eDTeb#OU3%yr|f=YbmQZc0bc-cInW
zmkjF}F!nQL<Fy6Vj&EQ;7HZzx{3VJqW02?1o&H8HzUmJg(`IhEcxl*R%Rj~=*Y?Y-
zefGn}(k%iy$2`vF<>H^n_<aHuL~C@<DYK(Omqc<LN-GcaR^E<djx9egI~C{Df<-Ek
zW@6z~!5dev$dxuUB$DC#^d&k=btTT&+Axp`tQ_fPPDU3NyPqwhb_?!ZqAj|N;-O5P
z66c1IeWkwf1dy}ow-=WaTl^`#Tc#nGtmX8s19|d}XB`(Uk4tl&6>G++s&`~q9f+AE
zhOa(CSxAmdD8R>(J@G8P$?+X`vnM!KEJw_+c?*#W0((O^;;cQ)G1#k-@c9*7cT9+b
z=kERKxHCq!4IzXTp)`)|%4=~oiCP9I4{t8{)@>P(O$}y>+A>Qk%X%YbgIhR@Hu}By
z-G247ki7d}lDjf>64<1BD`E*fCXJevGp^O>d;SLYqjEKN;!zJe76d&cr&iMaC-QPX
zRG=eaj+YS0EppN?YjIFpuunIsoSRB}W-3MV+K3sV*}Y5i;Bq{=Tj(1E)Z&EYNEjF-
z5yZnvvifH-lay167+Ns$$WRP2faXa-958Iu`MvCSvsvpIrtAyw6$jW=RjLcNK<`bk
z(eNf->rXF|$*4+Ig=(|5s-Pg>ya*Hx4QK$-w~ZIr#kvMJ`q5!pb0RhI5BiL6=yNsP
z3pJ)1px;4xTCuWxk!)P$GKx@=ry~qRiAz8>qg(!DN-Mv_eH}jYI{TXoGOa~*fgbqk
z%3vdR6I5f>+-y)H9*0H}jpJqVDhi^+3{R`FA|yOscD*7;vKC{kKmhg!y`0|Jw5IMq
zGuG2hw+5&|25Z`@9t2fNVwX^bid?1vreaNJAk_;J>i`v7e)?W3>A$-9|K^_`k7!#1
z##Mkw`2RMp`Z5)hfHQ?CU~k=@I)JrkpApm5?NKsaw-~lcin0t3Fl1WZ%`=_kYq;36
z7iab9@jO0sMmgv$1S-u8?ne{Pt}U#0<9%+oH5Iyp;z$pR;5|k1W;@aXQ*(*&>=Sl7
zd5g*cP8Hyps-MD7-j<A6?yc#z6f-Ib_y#sHJVoL3zK)KkK=~bHW6qh#t_k&!tj$xt
z!vITi(U|gXm9?0NkdOW|PI`N^@_^~0d}`(Corjz_v15V6G5@!lm1V5UEzLC4sBJ_S
zv5bjsaE{G;^wtaSBT<q4^S{J(+GX&RC?}MOjD-xcBwocOAJn3Ip30n5zR}W?4scS_
zGy&>geT6td*8JY6)9s!`*4iT24*BGoU4>^j5r7REmpaq;?#f!;;W@cY$*#km^RVz=
znPs_`!TluP@+o+LtYjB0_+|gFt|rUqgGs79k&CA455i?wy+v$?zKvYP>!W13P_-*B
z=-7%9FK0wo?eZSkiwdlJ9xbTy7(OGC5b9B>Ob5R0@I+)f;+d_L!P%h4!q`eMck@;Z
zxYmE5aezErKHf7P<O;nYqNY9$3MSQ*NXYfXD`lnEm`?;fWEm#JLuM|Ex|6*KOGfL#
zLvuCd0LfJXUAt`F5@+c5E;T2sVJ37UiO+Q9yM#>4!A%qVXC??*>_`G!X=JjCZek+=
z>(;pd?R(Ku+l_iRcRSIdmkf6vz>g5Ah`WH_rlujch6715_3$UWF==r!uB_+X?*@%6
z#+*~&^wcZw{3G~}|M?MeIumx;I{IOKtJi!TN|KkWVSOs)>C<u$Od>|YtMo0@dE`3e
z7aePmhC^oIPZ01xi5z;UD$nbs(#?wxnLNdF=fsq{T{QN_xNf5jZ{dD36C~Dt(pE5D
zfZfvaBw&jhJ*Xf<4HyNg&m&=w_*#O}4T(YP9w6r{D;Oxt;SA(`N;pF$tk|Ds1du)$
zs;dpv6%!NF*48#BfBR<o)mE!Ds*hJNrr?6L3xBjF*tT3Q9?&}b1iDe`BM*AMT^U{K
zlQVvGH#r;)cE(?_Hn`Vn74?&!*+pbfDA#6zHhJgc5lDrm&J&X4Mm}erkqac%>;-4-
zkgcWOuSVfV42~Ui=!$zY>0%&aQnPRN*tQa8TjZ?=qo~2xcUA0zVi@&p@ZJTPAW}JF
zX_<jE9)!u0U6Za<b%3I`I4(Fby-~TDV;2e7c!~<k!fMxjy$<<^RYR!38dzJK?Kr=x
zP)aU+Ut;30pnECZudW$?SzPQqQ$)KrPtGS>M5$9w_8E}>w)iPC)|%%USScX3E`Ev+
zS^-OVQc(h3sQw^K-x}iTz|se6YwLdB<pKD5g9yr)Iy8V)%gn_4CO9}ag*Z6*b-|h}
zGZ_2j`TwORup5OwFYHAG_fe)tIi9P{h_VdpC>A}b&5Mv2wB4R9!Y$gNJ61bZNS+xW
zy()r=wP?bs(@nB6(bs7^c1p1GHOk;;dFf9HX*Nzq<aX}@o~4s;K{orGn9Um(0}YbP
zJ42B{M~hsluyP_MbZj()S9NuGg#}P1>fwKSlZaf5V7yA>!xvB6N3)~+gs#-L)6P+9
zmq`Neo-IXVYSHgv<}EJg!N!s*nS{D-(x$ff_zS$`tDvUBAO7bR{omf=4_fz6cJ$;{
zhhCSLW%o|FEQe0K%^q=X<-8u~hV@<y3RTa9>}F!#vWg^i12v`Wq+}e>B?QZj#2n?x
zMky=^BEj|5EQH?xhgIDn>Ap{o>!1B2r{UJ~q*9qlXjyhj2rM|xn+oH%RC!gl3H{){
zUR?~5Ft+11MejGC-)6^&iY|8vuikegL0b(*aYlwTO!u4Yu~uPaf>#$i=x&Xrva)F)
zH4=N@XN)+;>EGmy5Ef^g)HnEAmqPy4S@N=E5FyxNUB60w-sO&<t>sumpSCU7-5n~|
z<3PP1DUgeN+#Vn@9e2~(${>V-xS8A-%9&A4fx^6Y)y&KD1j$sMMo}?6$ING}?Bp1R
z@?F~<LO1&1LIDUXsm=x+fIT=`;2Wh+Kw>9?DaJH>G!H<aXBU5ByB?xKm%K#0+Ua7;
z<W_Phmk6I9H!z#(eK`k(a&#RnG_sVHR$O3XS(W|fwWW*D9wdQQOGajwkUS~gxn71w
z@cB4VnD+kB!k_H=O`c%_RW5Nrfi#}rO2;gwF6E~W^~EYiOlL{gm#w>P;dz|m1yuyo
z!uUGMaJ-j1E*<>C>{r!V_6k2lKj1%&U!-UTCSADll3{w;*XQkM6lid1bcAwt_fCZ5
z0*4|ALJMb+lm=(X1yVcAq+datmDSSwTll)|m`7a13p{Xm#Hg(Ix=^jG_c1d?U<Ave
zC*oZ#v31fMur_E*X}QKO#Z+E(Bei`NB~|n9*8RN|#)nz>PmqR}w#sU#K{`m(&VbUD
zt%e%4;HNtNT?Uo-Z!5d4k9l*>zU$1>(Tegd2w-tw8WsV0dDCD?|H3Ezw}5%{B;?Oc
zzS^B+pMKUzJQx=H_=$OE6m3?xZ<Zy@((ov~OD(KhmW|TpU%Hr$rfHDkqA_7J9=WRY
zFS@)%XI8F88Xg-b2TO%pLF2SelsD9e2HbbxK;mkf>Y!W>Ll;y}VC31l2)qNc6epKT
za>xmE|A{VS+gQudbO8+U)Lly)D!E85UkxE(Z6nDO%bWKE(AQoU)qVn1*H_gQsa9et
zA+N-$pUIuMoM9z>_JavYRI(0_@D-Y%s<QJd0BnJ=Wb@;e$IVNFX;NBZ%3U$sm)$_D
zL-vT)IEn5W>I&*n13;B!It;{$Ij>lYM@1ZD7mYIz%MzY+VDF00E?#kl0JE{Rv)-sR
zz}JO1I9PQqp2@%Tx8xKGs>aFX8;X!-$>yNkl~~xocqm50n4$)Iv@v*DX)dVlX612-
zdzbRQ4GxW2dyEaE^De-1HHJ7)cNFoazXoI>euzHhzdWfFD$i`9AdbqEPw-6~5xK<a
zQQ%0b>`!Q5^-3vYY816(p5h%Pp>5f1=;9E8I|18TOHh_^sliqX-8_KEtPsS)jzM!1
zi5UX07Mvkqn0Wz=HiM(L3bzbHIWSsf5W%OeyKmdN(6FR6S9t<qyJZrb`cOcH8e$d1
z-c?wt%*pz3i^~M}2{X)Z5!QxUvscw}pITeJy54d8+SMxhm6GMFJ7>CHy2^d9CB4(l
z;`OG>-fD=u5F#w%J{wk^ilWrdCqTgHhrtHE#%zFD2;29<7S|Fwbl#4poe;oyGD|kh
z*7BJbd~BM-=;Ix*O{}@TcgsM7izbY^k`PaJ^6MLWSt$_bXv+`Xtz5DK!!#dw_*;FJ
zA@Xl+)I#6i|NPT?gy|cAIq32)P8Od?*wHlP)vH&7oSo_jgj$^V_kLjZtDD9xHK4q+
zk$^r&bctM=$x|p+_kP6)kX2w!0LGx=CATI7-!<$8Beso#WD#B75LRGR7-Z0zrh*}o
zYVoD$awp7UC11je`pFig?+S|nf{et$e_4XovV!pE%KG}OtfJ6M{_9<q3FM^t$@IEK
zdHGOl8yrvV5}PGkg<JaPFV@vu3qo^om-|t(iA!jEQv+8BtcBHGjL^2Wi|X5q4(}2t
zBwXqoVOl91<<9T?fqU4rI>Vz-*OnY-rwr$X9OwbRf!maHtqIzr97wmx%?~>r2KTPH
zf1_(CWD4U;C-*J3+}&$#5YS^ZJTQc^lQBCPS1S`<IM?`cU<Sci0j-#3B4Ne9C!iIJ
zmT(>@(p$@$b-eiVzs4Nn!R{GCQVFLFUFo4+6x@>caE@M2huve_X4ol1j4q1{_ZTCs
zX6B87kGU892WU66B%16iTRl~f0FUdJuiI#PPG^iVo*02!V)zG7$vG*fLzb6Jn%CF0
z_EZ;%4dU==A2)mw2~}ND0Br|ZexsnqC@G5NV}bG(DTN>@n_<ean#@N*MjSx^VgdkY
zfaUM+0f&wop$1FtmCAxe@){5oizL+vPF;xie>XAIFGu_F+<C#o#TFaV?2f@6z6IH;
zm)ZrxhOM2q&B5wdoPD>4;uS;=$4KZ=ku1qe){4!X$(3u?@wAa{L4I}IRoK+ft+a_z
zvfaxe1Mf_4N98W=U|M5-RS-q(ZlkG*#y~{RgbCST*_+c+shX8o9IlPbkuytl-|OW6
zj$hJHD6^M$z=2gYCOBl{s-gtfg}hz^D$`?9(u@U}8DE;~r;nba@f&`scX$2svkZ+d
zoA;xP^m`^4m_0h~u|GN-rKri%t{+~uY#!>=)ClVxAq4vDYX?3m5LI9iVr7nh*IQHY
zwb5`ulC>G#N`X~+F=S2C(`&aP#W;A{F}C7ass@q+?bHCPs)w4tsz^lU{$4lhjb4`w
zT`V*o2t*niDie!1+F?Jl=q@3fdJ{Q|rES~Y7UDL4l9QrCC*pHsyPmc_yoD&u98veE
zInW@$OU$o!(GskzK<>qL+0Ok3biIj6uLa$pSyq7sge&&eArIKlKgjap>iT!(h@A(O
zD167%3~8J0KK@TzpeTg6QKzIdT-3<qY$Z)|(7HEf5J&}3A@O@u0VKw!g##w>iFE^Y
zc)^DJ#k=aWKZQ_6{{=}dEiDyjYjc4EP?NTKq1a~o!PAru*4ED7lK29d>QKCu@=Q3J
z#5IVa?@79iTv5{(v-^>V8+!SGBS#L(;QOZ+e*6Bpv{FA>iK{s$dd6DhJD0b1dc#9+
zi0w@F!k+MP`1f&=y}b6M*Wq}ECme)pVkooZR{TQcT@8|Fhr<9PF(lCA5aw2>Ee08z
zaMk1!utwUPSDN`r9eo~n)azb)uPijHW#cIWb{}c9RjCsK9^mzMQb7G5k6dg#p@KIV
z4Zux$eCz)?=+>V_CI9<I|FLq7w3Vi;fp!U80W1~v&~0b=QOD)6>jPV+cPi}bR2Hf@
z_WMScfif^sC;gxE3x1lBYGO*BecIfb7tk`U9F3LU#dv+a!JTQyU2X7*OTj+>wsV3D
zj-B$9H3hctu?DGFUF|baH*P%FG;2<itUY%=CuxJ#KAt!N&hzYey}~A5?V=A4Z>#hF
zpm55e<Ln@Og;`_N9G9;#LAfKh<FG@b&6mg46eo(7O6&c7CtuBZYQ(AWd!`UAo{OmM
z?5>pa%^YOw(6Fl)Do%AbGx~|aUW$aC7&2Hb7>Vz}S3jyLGDxx7xhDYXZ4^w4Ug9Ri
z(^8ckV`D)%>T+^P+ype*wPEJXn>PUBc+18SwAW%yl-^Pwgz-j6uGVyy1rkjNpzPo)
zd^)fPCnK|fH!y=`9>rlaa=Ew28d!n!UT}as4Q;>~AtwaNaiEKH_NC&zJq*5ic&{a|
zlDoX=S<TslN-(dZ1_A!JbInd_+T&4=jl8b#TihRtoRE3PCAdR`DFYR$CGIzNxv-wW
zlIGzZvp+VL7s9YZmUs2%li;Sxqu;hV`tOG}WJti8!Xl9_yB~Iv2AsQzM5wE-3bt5J
z%~;QxWokogkc^0mF?Zo7%sls&19;G8{&V9L@3$O{LGP4F$f(PyOaY9c!+JS)Msltw
zA;z=3VZXee-0Pw(%q0F6x$@sD%>UkR@QN4X?l{ypxyMbdOg6Vu&nB+At7`PV<(*Ra
z{w;SkX7jr%_z};Tqw3Q3H*X%en?O69!)tDe1g#d<wj}RBrM+9A-5H4$U_<<=L`IHg
zo}4{LNUfW8BN;(SLSr##!kF~dtYA5}upqXMP>-`<(bi&LOb|#5zgr(j@?&ehvh8~L
zV#OqUq%3h|HT1M>Q8xL-ik!XPTDB{Zpi<>RvDVfW_a^z~9M;>qYz!7@yT-nvTHEj*
z!T)vNr};stj}Qm)!^zo%oi7)4n7Ai124lJ;2(5R1CRct>V`72S%5CxQ6qkm8;b42s
z6EwSmq*MOtQFRwO#nFS$Buwe8t(i{^TBo`6LebpIKl5b_svay#$E@A$s1cTygs;_A
z)@9*2lY&G-y?3Q)b@rh3BEYRxA)M1&+HX*^JPCeIH;B5VQ|~_m%bToTO1jUd=*rDB
zIC+$2)s^=wVCu?W6%%D`Ci1TFCwWgYA`+BTt}l&O+`ky6AKP}J6RhCXH~bz%1`Ov=
z6S+%USKAvffvBw$nGBI1Ytdm_eWV=p(3|Z1a=mvH&p#W*DUBjAA4v7fH4Sfn<Mt9J
z#%mx$q^`4x$}PyuWqf~Q)mtB=LG4^W-`Kq8DU?Sz=SRPBqYJ19Rnq9QmhF+JAr(xK
zh^bLjDvw=jB2=K-Lok|SgfyI9OchS|`*}g5(;xqPI`x14ufM=#$TG$p`ElQcdbfFM
zmQS(y(!}gvqlL$HQ|@KUqJAck@GQf`zz5kfiM6_-X>}}D5Br?bAF~KdgWhvzTiy<`
z`<MjOKg#5)ZsBCMTnxM~ptm)5TjVZJY<g7@Q*|iH7Uq1#6U(}1B8L~SsF8w<BuN@x
zeQppJSBIi+tK_Cdq(yS5aou$=IxbmWnL7AbNg0thcMBFRfC1G<T;o_o<z+)9^CEPM
zoG?q}zOAm^f7wAi_X?|@&lWG_*4(wBjy<TYuBvau^E4qecJ$h;SD>n@MoEev=@IVO
zCF>6cB#?ihQ#kInH#58-meu^$9%p->)D)yEERSwuWS(Wn#HZ9kQOkSM+wqO{K70Yp
zh-=Uwm<|a^8iw|paeM`32fca@F-;_&d5#w$d$<3*vE3i3a_q>^RU87yWzrvY1gyzU
zMwn%e!HST>*6#ga@`i3?wzEdk6j)iYvm+vv^U!XeOXR0(I>yw|q|St%9n2uAL(s?k
z>wKQIYU>pKWU`)uv>TSPhxXsZAHUCO-OH7*+V$wMTbs431p<E{V@fxT9(EgohFdmL
zY`=0HY5@`z!1jIJ!+QM7Cl2~T`sDW2hW*|keEr-B*{wPe%^IEz-%VN_;5QNn&fD7G
zuXXtLgRo;<$%ZNBOTg*0XBe2Z3~Px+19sE<P%%)nt2Kah*SG2dDpJT*W5cA&WwtF8
zJae+Js#oesyr+$}aeccQNfqL|b;a%wjt;zI0r#dk&PgX`<|j~N<<0cwWG*|*-TNss
z3g>LFEHR78u0!3HSJh5*2!E=%s6R9F-q4SI=h15WNoN%6>?Qy>)Z>rd^37{7ct_sV
zZsZjB@+Ig0;JjadufeMly@Z<j!%=lf7AxhAiKrzpPC~F$8ITZ<#A<__4LA1410noQ
z%zTzA0f}wfR-Yi$Q4M|XW+Diq|M**eZE5G}G3%9R5>~5^Vu5k^-hc6FF&NU%tWuLx
z=IP*TxC_2GNpopsxam(f0yXaQHNUeJMN(%}%@CaaTUXMc>|6J(&CF+7i?VxYL}CW^
z=`xV47oq3BC~I%9;e2MPF3a#(ML|*Q1)Ofxo0x?xb!QjwM&ty0z<u0ZO4(OQzrg@g
z(|iHf9%{qUwTvs5UyQ9=jeJGzZLRaNeq;do*L8j_ZTtM~POtnVTT$4;x*vQ<e^p;%
zhi+c6t8ZGEa}t;C!_~R>_|?}GC>1&7uK|G;E%I-&wj&F)L@Q3369g6)t6vr9gwPKL
zqi$1-rkEtzcPn%W@1Yp;)b7+-ff$EEjy6#ni4$>+G|$B}Kj{N1B8{qT>0(Q=QA&gr
zur(z0aKf1!n@4E?dpQ@vMTtgcd+W9SeQ=*hd=fG{Ze=i$2B%Oxser=}q4T1`B-Cnw
zJ=c%5NZu*vuds*)if0)_k2L|+6q}gh&V6Vdpt2%RPEoolTh2EdNb8V;B@J3j)&i%`
zixg{`5>S#Dh)~Inw)$iP00Fx&AZ+pPl8RQg7lUsikZ3|3`qQI43P`fk)#ZJ9ntwcU
zxH`L+`nMoV_tNv2Q5GMMP&^BsT;-oi5w)s9r$>x9jM^T%v{%fQm4|6fbDJ+Pzm9zC
zy)Lq_k>IL6El&M#IfdaoG7M&hR<=15=JHi(KL?r1elTq_+;W)lL&<J9W;8uX+DvQM
zMm$7p<2Jm=Vjx$&106nnx0~2Qj50|*x+u8<E-J_c#ZXSSrrK;xv?W@%k#ba(KiO8#
zeQ3oYXowBAtIAzQDzZjau?z|nk(Gct@G7afmd&x(dQDR9H+NgTa`6zM6t4N-94u;X
z^(NarFXzE~M`TVU!OA(8FNj2Jnb@ZsRnH3g4ya7&qX{twwycBM3ikS}FqaA1NJCr9
z(ta-3b2LVsf5UaSd2t1um|k7`cA~ex?<G}RKA3B42vD}Q&&VU^?psCP9V=GWv=G&_
zkUWpkejlOk(cRsIEN`G2s5=92``Vgue0DV3!FD2J;rk~)e>Odz7CpCjz6ZNIPD*)A
z91+($OCL)a2~rNNFJdR<b52TVcpC#|{IT;aoF+T^CcZ^|HqfsJYl8523QK7?Ch^*t
zRrJpTsra4xjfff1i8qp8zxA7Eu=_(7s@xh&Dw1rkF%bF=9)ScxAkk1X)Q$6`jzqXv
z1O20Og`7}M+;FuC<3G%0v`(UOlj>|-we%Ly`pow*8<svOyKKjyi&y}S-;sAimnDrR
zHjQQ>bUq9ipEC%;N3$?-<x4T#029CDuK1pWMlG*uAJwP5rRq}0x9H?nf{1q3X+%{6
z9;bX!z?4Jdw3@G?bs*Q6Ql7F&HwG!*x2jMnjEa21yvJe5o%;FlVvQ(oi&$(O+_Sf{
zr^WRmx^gr7iR&4tHq3073<gE$b<@_v*P{2^%j9_aGqXN7=(&ZyZzBQwZ<Y`!3gaxd
z7f3?q^+%Xh!pkC4p~jd0H6YY7`ezBPDebr~Mn50}UN*gL-wwkMTz|XwQ-n)UcVLmJ
z8uF-t_`B$ELdD9}^%w%>D*Ee}ztPy&AkRd_9$n3oJv-!jM>3xG$K<(p<Ds7P+F*K|
z^2#J(&PAzven^c%Tqd@}l6Wq29?v>kieP|))GaQu0f@>~%Oq@BOcR4q3^A`osHC=T
zV2utYp>Oo}f&p2R^7^Pns5Aw|K`ARME$cyhaoXC1(DN{jl=J#AryR0+RL6t#qhs#3
z{8j;i+JjZk4sYtM3#`R~puV2G<IRZPyo<{AS<-)I1Ia8>ilg<DOh%HT5(nUxdy!a+
zhJghBqG2dh4+b!OBR;t)X#nLw^)3MeSe2E0J<s6@C@>I*BJqiD0C+7w6^QD`)DcWS
zrN>?<19D?6>{<WU^L=ZiRN>LieT}@W{(Ain)5^u=jW@^&7)XB(wMar6w&sW)bnNyv
z{}x_daN!yXdLO)t!kkz<8vFf&1wT-J{Eh31x9_v_UyzKDZtq9wu>+z8jPV5l)Sv-;
zr*Wk{%VdxT;*}VTXk>5gL=qL9nV1?}k4l$g0WS-|IE~~sHe-+@>kIgR`yAJpPWF7c
zH(&cKP%4s?QZ&|hKj|zc)i9*%!QWacV-kLGtU<EB6<!?W9nduWOT%7n)e~guBM9{6
zmMOVHYvg)=1)eJ!;3Oip;4hdUhwp2}PzG+Ez?kvSQ5qUn&Ykukk#cW~TGuO{`{6UB
zhN=aV+iycerEjptJv}&BT(a5n)h?mN3a-o61xZP4aDsNPODce9c+P?B)uC07EW|Rm
z(cDPCC2*YXq<BHO&$&K$*kd^6IKlD`*_VK*7g3o51|wWyE{2M#&8mI%vIFGx29p4S
ze^nCx_f33`*P|K<pd|6Co}I>XVJ%MT*Vvvd2mT6};C6!@`Qw)YRMn#07G~p#MXGWU
zHxWLMdBKsmm8S@ui-HmVi8?JgWv|s1ywl{|@PmbuX|99Yr)?WpwV(+c)%S35THtPI
zJVE@mqv#AZ`lP<AwgecDG>+9T6G=>}7S;<ByPP>M*z@?w^BZ@#PvCyjt7fw1o(W{<
znnl>vq;s0tY5Su75pp}AJ;}u`y4Pl4R@3{cR>1<g&OR#VUq|gPs-?ctc+I4V%3~~{
zaR8d#G_YL4<SQH;tIJhYh6D-ZX>t{wCPweOlXobmjHp1|-Z=L_TG*&~N4q;@OwgCF
z;CfiAbzEgy%Z{}ia8gEl<J6u(TmA#KwY)uKd}6e@#*;|_X-hT?=F+tFdtKB$1(M<I
zua*MwgrI=!Zw%#DG%~B9vD8N0D~o~XLf!s4;xTNhKM9di1ol4KcNOqL%t0p?c)DUL
zsXitPpF@QI&Ng*Eyy24Gqtxz}3-!MQCb$vP*y@QVQ>qT>Va?7n#d?+27C2AUhOy|e
zPR;bqbdQQspXVY!Zn<wKZ0i1y9=(&-Y9bxJW&|J1zb%WJH8<K#cj0vLO}ccQT=tRl
zZaTd}a|E(B=VBuwd=k{sA5a%|f=@sMQtV;Ljw1oyQ%Qv;YKvS-trz<}{sK-RfUN}h
zN<6nJv!4cF_p3Uf<An0@YPr4#*M9q}SB}H}WHl$pQ5egXJv({qT;`Q%l8Oht-eTY!
zSA)Y1QyU$E`@h<RnOSsNimYR5Jk%xbTl8v@5T01HrZ~>PfXdL6+nApj4nO@5(?Zvw
zb+}aG>bH&lA$fCA#G-rx^22DArDqGLd?D=hVSx>*8h)-0hSB%Xs}G)N=h~W<JkPB%
z?TG(v<_39mXoH&zhR!(wRjX%QZ6kqbvKZt#%3-_-CikX}NmLTZ?cIO!yi?@IPZ>D>
z^_=%fmrfc3UeXMtefG!F3>5nW1_2eu=(&MbL{h@~YTp2b_1N{`GwOicHjPF1<8woK
zt_G_WKioQbtj=#Q{l@d@Tr+*$=7ZhS@HT<hsI{`u!*qY9%50*V4V>~!&dolL#8viM
z!8k<ym?a2L${e3N)`n~TfTbV_F&b43Ge~f*lrl%AxWYUn6mHw$WyLZY?=MU<w0}i7
z*8nP`Emryh*r1=1N`CE{7?-Sjh+wM86l?U#r#kDqF}U6apfi)$Qm(j1hn15b2c<-c
z=)a%}av$2CsZ7LGY-C((h~Tz#V-S0Xb}WN=TC)F3hJSAnsJT@&Xt1|rnQ(U94-Q6=
zN=`p^eeH;EsOdII%#2`;+4S*g3WwuKVM(F&w5TViIF&)P!-67tH&VMiSaJk*mku})
zCn7Ia&PDb#LAD)|voy!3Km|?+y8-6}>K)RT<7=5ra9Kb?fUlotd207EkSUVlUYbXi
z0GT7z@_Ob8n1wm9@7DzmzyB>WkEwf|r1xW+#7NWdwo&U-k`IHnZg7$*b|zuJTG6-4
zCK>%V-o6H^t6a`adDB~=u~`&8ecPE;+4E0_?FWzYt7@DG+nqfzr!t;TUi6)L@ir9K
zTNlE2*`)cRpf#wR)8=~PYs#i@*>5bv{Ki^zdoFg=gUrd^)%O^(o1_`y;_lI&gY?)_
zxlRh0nz62>ll<o!&I<gvP{bhAdtxwCY_dAG_%=HBGb-4>qiN%**l`YsBL6sei#7}@
z(nJ*pX(sd=)J4%8@3mYSua{`i>1q^FPHV{uozk8OCcjL#xP8X4zC^?gC}97FS|&Xf
z7yqz}Bp1HLkr;JcolHu*+)J3UAm{4pUT0Rh^rCCH*3uq+z|H}aa^JHF!VLM<n!1@7
zEl+P))OWC2pdk>@He!it<9L(T3~za9F81S*D2jtqDxuQ66$B)h>yz~*iVEkTBhUdH
z6bg3MQE3sXUg*mTwc24L;NnpMj#xRca$lahZ=O*lP#GGi0v_vxVAUtN)Vq;UtZak%
z76gYDRLKY25cq7D<w1u}?gpSTw1XAkbg;6z>`sP4V}R5>fbR{sZJ`c8S_EVrgAZs$
z01xCcx;$-kU!n1jzl3q@rX2R5{&J<PG`c~m7a@-3MZ1$~Osq|!%g(#nvAwEvC~6A=
zTAHmWMm3RkHwqE)0DU|DnFYV>;>>~%^K=?ki;&y3Gj;I8!WbsiYx1>`ZPoZ|Up~Fi
z+dPTrf$wghHU};XyLbk|!40#l#3pDv?)wAtpv)!K;TOv<2e+byqYax;h6*6}<XK}^
znN;=B+*?)v!wu>k)Ki(Tr<MS5((Jk|GkoK_&!{|KSZjhwBR{fp2NqiABR896lId+;
z4<%v!65?x@o<mG3@ZVnyEuUz>YD)XR!lu21a$U;Pz52{C1QdiZy!TW8J;e7k@!i7Y
z8@R#eF!pnECruPGS#M+p@_HFYHBT7#Y@W5b5#D4C)^~Npwl!^L1Fn5@d%a+3Nqm+u
zTYtLoN+@OaRM+oqOP+U`qY+8zPzW=i+$2WNb8%FgE0%VmJ+SgZYh5?P)mUIgl;)jN
z`1!By_}=`vA<XhZ`8ipa0git8P_x&jW(G*t>{a2VB0EuU4#9(ENFugHN2E*FhOl+@
z^$*Q_=ckoJ@gS>9T{;fZ+EMtD(<!<pmm+SVJev6H?+r*c!l591IeB4t%YV_dMxkGC
z4V}F+ykHnTo)Eb>h>o1-1H-O2#{1Oobafa^hC;VZ4fRjx?W55GE4SeV;+Hg1+B0ON
zN{3}+s5R_E`UA(SSh3G2x_?L05<-Dm@a@Fl&e;lwi>j4HF%yyveoI)Wl2?2$R`U9*
zL5L8vMv#8T{Fae*a?-WOr6=>+Oh$_xy-U+fsW9iIFE?zse(Vg5c6EY}HmGd3kXd3y
zd66RtzJzn`qe>$q>}8Oyn0&$?9CAGoI^Zuc9kU?ta@K|?{v0e;vlP#+B^I%wz%O19
zN|!2heL)^~gw%tVjNEKgZQeetv^+vr<Us!X<J*7r#TQHT*g>z}Hh%1o#<xA+b*1}o
za6k8+RY<Qn?(9?&Qj$ss4~|Z$R(Iw}&xAepgx52?sumtUhaNrZFcI-wT6_LlyW<qb
z_2h>HMto)iNEg4vrw~yh*<deEqGp4ss9-<Tf&(7-cqDVUxw2nW!;DdG10tAi{u=r)
z=&a0EOZ9wP$HJ!BQizGFpkJeF@5vxcSapysIZGV9FkYEe!#L#|y{_cnpz+f9oWmOI
zYss>T8nmsZUvfjpt?GHtwJpe#>uyWK@o->cBt>2x7GG=1{`SX(_ZiZ=((wFWx@Tu6
zOh_veiHE|`NMzs0wgH(Cy!#a9^`;)@<i_f#4e(w@m#$EarGoFynY)#H?a5wx_MFLn
z{CH#ks?vxiV53sdZ(H?bc9X5K5Zk21(7{iT%c*9&Ge3VuUHbBT8*e#yt#*2*K{6Oh
zf0v#(q-y8!zUEzT(GtgvzwzcCtwsB_b5F);M1^ZId%SPP@4=%&$umwguZ!9X%1q7U
z&WM1YQ#E(5MWe^iJ7{kNk2iFXY&KB{O|jR$L(_tf4XuT7Als&~04D=}d00owRj%O;
z+;3}80t)d|&<Y)}!b-W2QF%`hB@V%EW>q-Y=$F_cKW7KqP^kB_MopGDGgS~fcfZfd
zF-Vjs^RInoTB2$XF$J)toU#&L?NQB?h>iiIS7=d8z=Y`534a3Jl7<(VIE?*(R`rwK
z_$^P`H$zUlQ^ebK$-So~NmhwN|F9!Ta_c}WYr}{ms6m63aUdi_ZxkhvfP$j?A_-_9
z0tB4`hPnWZNL5U5Wi0`4SVDo&c)*BKTAuvI7l_<te>`H-;t3gXmrdtc+zc+u@!wM>
z1}XnKAaz{lak_NiYTQ00tf7M!k*;w2yFT*dtd|Dc=hi{FSP=ZVh;PQX5dbgY(o%4l
z?i<RPwv<R!;nMO#TkD~Ij6>hlA8fJ%RwmZfcnj?4ebk5x0eSZXG8{N(bT4ts$2oPQ
zv4`39Os$VcRDw#~i`a^N^wchiNs7g&X&jy&)!Ty8?QL3J49ETBGb+#LZ~Dt3Il}U_
z{U3=*_Jaxf($g#Uj^wYvE8_pO&ELC1KB+U_+)$ak6A#Y^<yy%2CzjYr8l_#}K}#<W
zF1!iL7n(0?>4x;{hS7_6cGjs|lk9E=)Y4rx_-fyXZYYm9o?S?Lr^-0PHSg&g80q}p
z#7=0Q0uBk%cGc4PGXSa27Mog+wQdiFLKWZb@Do)&9_e!BL%38I-M{Z#UH=vSb-ljY
z-J`KYSLZzqY?y}=HVUQ>35V-A)X$XvSXAS3;>tUwXyI{o^Wks(6q&&H^LuhU3*Gb6
zw<pdskF}%{YQAHf*g!G4AGYy1E9X}=WNqmH@d2|En3Nhkv0EfWTxb}sy3UT4^L)Y_
zb;fasI5DXH4U7D7!*bb4##Nf}`0FQq^_bw&IS2pQ$UmAK_Pl1h_UZTgw(+Az%XtwG
zbg{*#&#D9#=1>j&6b+C>s2PD^(vvVuhLFFX#wy`4m0s5aURJ1@Dc$s|8>#~0v?4mi
zwV#P!1ITCaywmoEq!~UrwX0`ostHKG03u@byZ~FKGmL)U)u9Ur?~U$){1uyJ^<q7P
z1z~?mQ~^wxe<i6z0cjU`P%y;F@5PHghuHPJv^#}}Wj&WFlqqhMf1<X?#EYTEsoyLN
zjTm^Y<yYT3lJg99K^<li_<rk3HEeD5PkObP^T#8aDiZf*@*j#j`(*gbot-Lfd-xnB
z7jc3Efx5ioVw1dj@l%=&xlV>4W~0jsfm2h~W3r|G^h8#Yk|3+HJo<&bI~{w4_{)UH
zWG=+P23t4l>xrSRle&nT-R3Xncs<K`bCKwBZq}QC73xCN_m}?Cn`1fq<+;xL@OK?U
zrVhgVP%pg_BVlUEgmRZ6NJkMlyF0NQP=n4m7zh%Sz4xsu{O0oM0~xEkO`USS-RcQx
zyVX3~*B$n-MXWVxCbxY9R84QHy(?s!^Q@S!qe$$-(Kt@);lo54J?#!P(sZy#m-#y;
zANwH=?}4Fl7SM@#IIn>armJduFJ$0QF1c5axvzbk`kTq>XH}EGhs<Yu{mMktSu_oc
zi?0;&OJs78D$yP1f)sv!$iy#HzbexG(Ie`>->g0jEAkV6o!@`<@$|<dh#KxRbH@#S
ztEQr^h;s3D;<=^1LxCh|h2P0Alt1i;dXaElC9O-aMjo1)L1{u{mLHVQX>(uwz8}+H
z_VgUGn8rs&FPUo{%bSdFaFnPYU&yEStc1P<(svh=0xbe|2mbC;E%>YW*L^zPgPZxV
z#rbyyF(U__H=l*nW}QxLJRfwvO6Fv!l~_clNO*?a3^oc#>6m6pCk7{<nGtt}5B1Jf
z2q_WY3{2`R<tf0Yu%$^~`S(|$k*Z5gjbUa}dIC?v-Evr%(MAeA6@ynw@Z&fD0mVce
znn~cKA5VtV0y858%RF+Un9-3hhL&SAIAR}0TtdYZ4^bB?j0KWm$drHn@fihk<cRK?
zKy}_*#{Y$+P3>#vy;Bw8H1K`hB5O#Sd0x5MRN)U5HB!4*e$4pRr#qRh?AarAwQagn
z(I1<z<&Hbc2F8Wmjq8-Vy0l#buW@6k-JN#bNt4M86cETpvdJWIY!=vIz`{(Lb9qQu
zj4uu<hh4p7a)0kcApTOuqhq;qsUK>bE6ZKTEL|y8LZva-*sZ`wYT@o`Ol7}?Mi12~
zWOB|-h%QQ!)pUFPS3evbO;Fi6^k>?{MO7C-qK6{$YJ&P7TzS;3m3SPJTNIWMORD_V
zvXT36NT?F!+gbH`>#Upg>-o;X*Az}VFM1}7bQ1M}+`|YCLgi8-H*7;8%enBcz%9tS
zEm|SHan#Bwt<j5V0gdb}&qcGN?D#Sv9By{)p8;T}iL(9Ru}!Grs7%Zmv!`#Pwu(E0
zp56QNvx?`lH{R=%Wf7l!?L*e;WgWPV%9>#1-1TOLI<T)-SAOkh_6%AOy8Z)GA>Rhl
zBS!^^m**MzEV(_)_!Xcbk|yu+QVj-dG{0VOw(=y9&W471Y|7DEY#iUPLssN4$b2L%
zQs2HQ>d=UgFpxsBIEQeI-RD!~c3uS2()D$>1D|)eNKo^i4Kr&>{uE4xdnyh89hdXf
zHPP?JZ=@H5v$_3{Zd>x4S9I)PE%`dIqgVEU5;>#tt`)cUwL8CkS0R!xFgANxcPFyx
z)|mElx%1N^X7@KZ{HM8zkoTSL*|?EO&21J>5}7$?8^Lw^KsT9++M<_r-x<LNZRyA0
z+vY~*OvR>SadS6LZ;=dLiH3Sa_f1>M2uF=t=j3^Kas7tM98Zp#WcP!wknvfi@dZQ^
z>Rsv+IcQ!RY(`7|F<%*39M4uIlwO**7MCfxKfewVL61#m*ckac3=uxm?V|F&udHw$
zf{W_T%&+JQEl}L9sdnEnvKP7rHg1|0!zDx3?-G%VhEH;`<}-J>B;0hqlyM~BE=dU_
zQrW#SiFONo_7$GT6I+yWHM$1s%Fn$lQ-)DWUDYyG6Rxku5Cj8dbw*S0f1PtA@4Y3>
z-EH7Knjm3k45fc#(T)fAmv;>U2{0`7!7q=wKQJGz`sBR2zNs8%pV1-towcFdrFkDo
zBjg2PN9c#1X~h#12j~6<*}L4t(>Lxr4BEWqdE})yZ~J0|<HHUEiHM7t%NU;mhHQ%o
zHb4BgsHL94P{G<3K0K=4J#(+@jsEd{Gkij9l(+G~;-x=6t291$<6jhPP3ub0PbH6v
z@RrJZS87~0wX~D8Z1G!8oeBNhQ9zZv(@6>yJaa?a-Cc9V#n?NYAl-gPxpyCrSPM7J
z=DDV>4#G$QXT9UsKVAsemIRy?FKXn{s(uFPZ35AAM2EV*?ntcli`rxGG&lRxHW#B3
zzHGhvSEqfmQK%ZSGxW&rrb<hV!%axtHJ?1~KCI5g-%2#U4c~sB<rhCKV&s!=E8PTu
zPuD(BVUqJCK@Sic1BS+oLYS(7cx3>m640!IHwJ*y1!r^)?qcvNchTt5yLa=Vbj>Q2
z&aL!qrF}div%28zP*`K)QzMh)B=Q~mt;3zdHNU(58|Haujc;gq*57+r0syWZet4Mi
zZKA@qeY*TN64(w{V>F6CNJX*0aS<quFc?*TeG!z{T2olz#TlS9jJzyQ-b-k`gMVXl
z4EOu*PMRc1NzL<S^ogc=&lb6kZ&Xbc8Q|K#my6FO#_yba^j^N3?U$ChliCf@Rbp#6
z4T-$}8o(PEy~^q^v}4EPI}hUFIWo88>LWU*77#Oo+fB@4&!6It{fU#UCUM!7)Jj9d
z(B$~nOyj%-X&~qhf3b9HrhGhW$BAMUnHM1@8JusFikzN|IrQ3J8IEp$xi6nw%idug
z=;AF2{<s$W?1{hc?!tDVCxA))GD7WPYtjX~mCj$uo|@4Ng^cDs4>K}0CI8_;%IV0-
z-q#<G?4~FRgVThy3#noA4Fx5f#e`QcG%5&Df`K}I5a)?E<<Ggf4sgYOGK0MQ=h4HN
z7bi5<1+xvZ*PD6|NcA&s*FGLu+W&ZD{&g<P)aWl=`dl6S+1pK)#d%Bag+A|4$oMrY
zoz$p+JskKIWA$omWAad8dkqNeN;{pjJ1E^6a6wI0!Fc`U%NB%j+zy`#I@8<1TcVpP
zN~x{HuDBj{NsJzhBhn5;l4}OAV2c=sol;hXvw`0Z+m~4-$8CXk`@JfFuWkUg`KD|D
z8a~BToZIHNdQz;3vCE$LixgB=D6i5~33Cyzf-lYM?+LnlHy?TKoqm8Wkp8Nx4+87z
zn(usyI{Sb~3V7F1IQZeO&M}V)+es3cYBV{2|NX7wrHB+5JA{$b4it@7kZKqBwQiJl
z0`&DWKH%5P(~F&oi_Ttl!~25ik;4&RQX|e(En4UdN%H>f_4*oZP@Bh{c)I!o6Ig7<
zX$X>@p!Pe~{SGhTuW(}<0GR{j5YA<H4}4ykd?D2yq_bpPb!YOuM$<310T30|1t7|Y
z&u?h=oLyBxyo7%|GNq}eDRzmu+mun+i@NDiwqUOD;m0BgkZxSG|3cXeFE==B(L|4Z
z0JTYGTi*0XwUOa0Ib*)d$0OW7$Z3C4+o+biaS9<XVLtlvf8Pz-(dnSa{~Jm9rYkoQ
zIs#bj=2IRxU6G~ZBAd!l%b`Z333C0P7OPtx8u<W5hFg}9Y{|0n^nEuf)ExP^aY=-D
zBNlU@W00p2W4r18iXrwv+4aUvL*+c>P@!*v^|*d^Ose#Hp)>W&!Cv&oBct!4f}{oH
ziXT32Uw-D+EhXxp@vnnMQc0Z(>3Z&J`q1h8Lus}fan2|I<B{G+^EEt|0gR!=1he8W
zl~KjrUwN2s`|(K5?O|2nZ-~*uZe`F7ehWY&-aUK5)C2jqD83j@l7F=SBLi0PI@%-*
zwUQsUfzw*wnB;PDsxLfzJrCbMu)Ms}H2?9)bk<<>KRRYnUqw)qFO-B&OMS1kzK82Y
zEV?SCkhWbbQk<7|P4+0k&WDpTgBK6CYO)m%iyKMCvKpVaU%vS=!uiqGs;K<ptNG|e
z5EU7E<MPA7!~9-y%z+|MnPi*~Eao(MMWJ7umF!`_g7qJsWjir#51%Y0Ux2ja_8u}u
zOMq#|H}j3Pe99eq-5y5t;f>iPXcVA${!bR`NZ0<l{;!{eYZAM&v-$^X08V-I&aLZp
z=BxW^8&22feLFuMiFSj`q+9O5^VD^BA8aXoJaV~TZ%|tvgW@ng9`Op@BS)SCOq1j>
zYP&44?nOhRW$fWYr|pg0Ujfmbt?n!`_~Vby+fARj^@Lf{zL+pgxNdUbNhO39ZTUax
zcrt#vr|qsh_`tLF@Ep*`VG#vlWaY-6|GE_U(8pT&u{}x4t<zh~LKpa>wDI@MiXH7P
z#o?3~@KQyKdDTJ9YYXTL<nRo&`JbQn^20mCNEGkwq|wbP;rWGs;>0YtlV7B<8nSmg
zZ&t-;ZXE~8AGOdZ2e=Wp6^6|8-xqUr^s7T!-s`7v0b$LspM9pBR>@!GrWHh=9yM>J
zdMO@WH=xY8S11SQbw?JDUHZH?@ZYvR0~(GvJ+}37-keHf3<j=bLz(u=LWf3-Zm9Ud
z&$K~c(1*G`8F`h<yB2lSLOQz5+JFh*+%M-uzRQ-=MjtzKG<sX5S7(}0{b7`6u#T76
zqp9&No0EQBR_zKK3^i<v#>t)Eh&q;v2G>TX_DcP#ZeDU_X5l<M$9Y3}ddcXv3UW?^
zm`M4scz!<h_0|AhyzmYBlwM+f(l8578?*z~=9?GQ)D8Kvtxm*})0eAp7EB3s=V`X>
zWpU{l^)s)^sjkVPQ7K-+Y{|FH%B5^jmU?*WTUh&ijQ>xiz&P8w9w0QZ89qt+t}}xn
z5;0L#k+HsK2zvcO?jTtp|Gufus4F-1Fh-_A`u+UOhtucgpSoq7POI84jU2&vgx@;7
z{{X17t+0Lh!RfxJTk-vIzvxr8B3&zg`_G5}^ud4h;6G#G-#s6EaMiRdgRe>o_do+_
ziEQ^!nlsdsBwpzAs=Nz&8K&f~57X-}RJ|<KG*DpRSG()>FuOWi;kFz6b=R5fqJmpR
zx!BBD;#s$nB>~*Ks>HkB#=Hl`1yb`$7d?SUIrd0qUWD8mc*pxlzEtQaQlJgEm!K%x
z1qM`tt;cZy>^|tyif0$c_EUnt>Fou=2UN9a%~;N>p@Au8eA=?5&BfDK^W;CM;DM>P
zqQX+q<JsC8K>Y8?Ua27#R#$Z70k)Fy)EskFqqslu*iY}1|DR2jlN^gv4^4p`Q_$|E
zjoRs{o0u@*{m%OP&3Z*GEzAAdr@}<4iTmD%)zb$x^PNLg8L20NIu2^)2ACs{liS-r
z9$Bm0s}sIm2w0tIx7@0z`v36dq!wO4RfS_xuxq50SD2d8wZrgNXY-iYwd&oBlW~-g
zgc1#}g)jp(#)Xz^pu%Nh=ulR1YX#RIRCJqrDe0%S>4k^guB7Om1#@aN42}wJ_v0rP
zXErV*rf1g&aTQ{=MGn-UFBrFKeQvBI#z<_-)!T2^!-P=lDC-^?8El4E(l!cs{4TPo
zCBmS;0<LyisozC^*KBaKr64~N+lm_O-(Y%tLO_&9X!oBhyJ}_ptUbW=0fx6{tTt$%
zL#!&3?T(@yZ|G>b#}NDKKOg?n2mjH7|BQwIf6oWsupSCTBdRD)>6I0Ck3H@~pI*NV
z3t<x<<s?dPq!rmmz0$uwm^R0@V@_^2ZU4TAjw?FORMhM%%#$5MCsh|oN;5B)KKUvn
z&JHSl`SqcLaO?>JI`wVD$0LlK?6T7T#ol{IHMMnn!+6AwSm+3-D7{GUMGr<q2#64B
zg3^MN&_hQQ6_gSbgb+fJ79gQW2_+FYg4EDMC<#SGdXp~IZ+o75?|aAn#`niN?sMMv
z`NlnKWbD1O#@=hsHTT+kl{x3{w>rDw6`kGlnZG4xznoPJBSP;d#pW^Y4@8|*GZ3F&
zN%2WtgFEFn2t?}g2BUQPU4^Z4QE+~N3?7V9F^lZ1y)^WrOfz@5wbDApzjvV)%wL7H
zzS?i_`dYR$%m&eexJ#j9nTIA@SOg0iugy?wlDex*+FAqv`_X#mT9%k+6JI{uV9wz-
z=V|(&sQHUKOZcKBgO?vQqTiIKQF-JX9T34d@?*rx>D^S@h;*8sU;jOI>AUr#;=;Nt
z_#nvwtjD^GNE6Fzfi`tM0M}nr9|EwF8Cipj1<O#yxAFLUBul_Zj4#NeM7TvoIj?1*
zOy}5gby?a>99!6H-hfy<K{2wnsH+Em{I5Cf<y$G`HWsZl@SjX+@|AX_Q!E*YQd}&b
zFzaS;^8KN*W!MM0Gkg$s!<<6v3Y8+!T&j%Q(jc&hFk``__;k}Y>hve+lUaX=9sL}c
zWStk}5mxIQcQ);Mxhtf-*V@{>WJI8n3O{3(NHfBr9JS;t_6*(p4ajNbFDrd5Ce+Y!
zaM#1z2`ud)o_Q2y2EsHt;;O}{{*>ZsX+JqP)OU1d)LXH#7^k{{@c}DT-SI{icT4bj
zM)8WDc`2OH)rnN-TLsn+_*dj$3wDiswpdF!fe7vsxW^C0>tnWJ#~&B-BzUSSd#VPZ
za~j<B21fa*?N~5WrYS<UNn3U<7Y$bA;wk+UsrSOd>7iZZ<*7YpV_&aV(wYq(iJ*$n
zy=jB&_l#IdS>@ZBm~O3f3i=TM*9WV(VdBHSZ=0a&+_tk-fBN)(9;PIb8+42y;*`Rp
zu&L-XE6K^SfR#Unlw$dU9TS1M*Di0DEa>*@Q?IHChn9c*?I#n~nrk!i*qv3jMK=GR
zOioCTJ@vxU(ICZ!e>|JS;u8l*{>+f&u5ky_)5f}tahDUkKbZt)FBb<RysNeMa}XZQ
z<yr<^&M{L`A0GX$2g<~Bexk_)B|uRD4~Z7-wuO*fl*n*v14iXtMlq(N(%fk;joGG6
zI9g_U7e8OXS)|qmf%MpeC1l{I-i}CrXsywG=e_KjUp_J?_WLU=ZtAI*$pC@qo5qaI
z3I4sUZX}HZA1&yh&zV0{h<Mu~AeX-`>^q<iN0dSdkI?LL)6liLjBg?K%`C!tIeBq>
zUY}L8I@^DrbhaaR)%JRf-xnD$!mtd--%tz7;)&{gaUX!0#2)*}WFT!O?i>It=;2*`
z3x4fY!rd!nDLNy51|6JyMW2t}_O_{TUJPi<Cc7j>T797|*;eYX-w&2rBa1+97|$j7
zVUC|~Irl|hqgMQRrt<IvNtgYVv-t1(mu`;PFt*3DEPXtKq5TYL++3aQTmhVR-(C*2
zmsql~meX0|$hVS2uuTYTw=AQ*yq!!aM5kD}yehzpF27S57+zLl#9AQ#vx@(+<KI3)
zbyW>Qlq-)Y%Q#pI`=*hW!pcDq9eF(k-bn;$SW+4PB!%U%;yH6+dy}_o=?a?<rh8#a
zNtPI@x~2Hkr1(U9F2H7g1XQ%?ovngf5iHxZYu(Mf;Y8Q<1iZ-Nt|FR|aA;?O6je5g
zM_`tS&l25~tDHNl;niA{_O7_*;wR4r4f|+`)|OKNH#x@Y9(n|(uX=Z=#;+qgl=nIv
z_Fj(dgj7RDOSr6dlUFguFz=eduKc3n!h8%$wz~kI0RR-NHj7RIF&m#RoBm~L`ezh;
z-2|R$84$U6u{fN+WEpKDr)8R%9}!{&^C#(dTWMRe$_E7%_*7J>x7B|m=@EPE`J%8z
z>Ai0Cx1_bwQA_&9(gr)XkcRvr<(mB6Lqi+uuQ?M8dNnBOyg{OydmzMM2P2om$8Po(
z#}f~#Ds@FII^Lm?m7j;E`haZm9$$cE53q7Gm9by6^+{tNgz=0-qzw^0Zf@!dE*wvR
z>sfS^%^);7l0NpL=O@FtW%IL8B$~1DG5)@Au{W(X!wa4`5*_~mg__dUhe$gl>(KuA
zcdZ;r_7L}v)p8rXQ}4touzM;+T*}&p(d?0CON&l_Skn=1^U>ADDvbrZr1&*d(o@Is
zjRNjKx<q<tQFc;k)W2&0Tj^aGH76Z+9bEcS>n79XZ5FRaz@-_m^|1u0*hg;uwmWhq
zfC}%VdzO!fokMg@zut+1Zm2t2btZ3U8<%X0C0a&8e7z|N(wvlZ;|?UUw@Ga^ZKM1&
z!;qcJ$juZcuccG4)zpCI9eEpCZ?ndh<-}0^<HioLG|a-&)Z>EIC*`m$f56C9PI`L6
zCf4)mpEVP5efE<Hru+DzdU@5b^n00=ylwik=K5CBiG_3qwD=rg1h{0qTG*W+pJK17
z@5cS_>Hzjf&V6*R&b3f6TExmNg!1!e%endmk#um@{;k`ArS>{O{lUazFozDxlk7ay
z>G_znyaB6w9lgnp)E4luJ{$-wI_X<J0LPz+dD^0Fj`TNCTru&fD_amo*3V1lyJ*=~
z!&8^i<Tc%p+l0-vDdk~tal8}gg&aUb#(HrMv9+sby0riyfuzcY_4GV)aJY8@-b_sI
z#x8s`MdrOcKDV}2Id=f=mUDKuO}m9g0>0kbrwniad_+13(pwOIip{<`;<@HjjEE_0
z<0q3MNP@xXT3>FBpDdtojZmD($2)d)@xy0e73GG+Y<-_&!miX%v$jCj``$|SnJs+I
zC8`gVtekK7S<OC*&Xl{|)~^`vm7kNr!l8iV2It_>;#mg6o{?GqNn?N2;e^yrCa<e>
z^Rs99@&{lv7m$*97NMaGd+pmLN%rj`W17U&^?YGn^u~(tgt<N>5Fakq{{ia0nXrU0
zm%e4G4@r?f79SBpO4U)2mKedsE_rdZ$jzB04GbdWUC@#)OXUiKMjo3g^8#7J!uIUx
zXmg|*3RXGeomEW*De*4GCSIRf#K^@LOPD>vbs6`zel$2e^z^T)?LVtr_2>3{cl;w8
z6HFZEQB)-ACsQ7g=rbqSStA$H-5BJU-_P!~VM}y5ezUqo?o+ZuFbk5s-(RI*u3RYU
zG(IJ!JfsYIt)c9EpX>8Hu|_w(I}6?eFI_j!qFt?PYbL<%6;s%>d9f6oLjQp8;dd@p
zWGjye>*rTYyT%%7^1@^~*R$#jqHsyi#iun6%#Ve%%anirx2hd|8IZbcQ{@z*<^Qb4
zTEnZog9ORsSze9DeTq@B1%MFz3NRi<BH(68&smduUE2ia_K&u+9Suo5L+uf&)wh^$
zMK(boU<4S1K-}IXwb!h>xXjPvI{v|tWa89sYJ>&i0gcj>RB&LxqX%bzVK;8!lm%#w
zBLNwX%m+yo>_qDl1)I{nX$_&3mFZZSDRL@3U=q!5Xxfc&S$!){#7jOx*{HOC{I=K@
zLNtt!^g_pwT5ID^ewqEm(f*}tx%`6)xnXHF0c0+#6nDPTZ?c1UyGhZep||RIMg|rR
z03u;(qwSW8eTMm;_zb;W4pl3%M=b0@*e>`y=p|}MKn#4M+`mnX0Zx~%mePyb=LqjY
zKR1x2T_yk2(h&*hx!rm7-*-mS>e`S;B15(7`dJkZK|R{gE&7ya*v?)0cq}{wlS<`j
zw|?};e`+)=m4Wu^@5%GyY#sGg&WgzUj{Lc^IB;p>V~m>UoJ_xzLshR!oohX{eQp&k
z?$|A4qfpkVe$GC^&0;F~T9kvQt~ihFgC3wn7unV?qsPWAs$FJ^j&=aCkyZQ!uCeRV
zjs=NqSsBc+inMz?R<F^swkS6*aJZVrf=R)H_irP(O{k3e<L!AZELTyc@2mckXn)b+
zZ)B)f&CMX8hG^0LIrj`b%5HD%$3b`mLCjeavMxMTBvUm~u!>@25X-U#jfF|0Zm5%A
ziYdbvqwtXmGtN&M<zQwu6q?HP>#K>&3;xLJqU8l8hr!Vvmn3^yD7d2(h>56O%R1$^
zz>kB@QS-f<?^8BBO@_)M9uPss+Lz$glnTxTuw>h&qKXN~(%jrcL$&%Zs_Q?o*vre%
zws_nvB~vd7R$2wFkCB#qgpqdOo?yT5iBIL%mXjOAa(^;HW-@|%0>%C149Qb?3Os}0
zAvtvB0!x(2(7i08?0v^twNb@2N7$tHoV0dpPL<_mc-2W4cg#;F!|yFax%yp1t&y%B
zZ5I2eb!`R5)^C&V8t;(L?zW`3`Gs^Qc_<Y&C9R(=4&Qmy!71@lXsj)yY}DD&+-i9%
zxTp>ZI@>kuos}V27G4(>0>H#s_Tj~*I}+tiuAZy>hc1{;dS^)Bi}qQRX~w#0s%*IX
zi1`SWFD;odrd66lkHM8BL+zhTXF(h=kVc1g@u|{BETK<<OPQywvr&k;aEA`BD2H}6
zi^=HAGf^tvZn0ZT#vKC8=PEy$*8M(Jm5j9ZC}?0eJ1K)3p>hZGnR*x{*-RQ*99oA2
z2tE&aTqc!L9(Pc(lBd8t%5%7DbR3rWS@%FtRm;vZ;!hDT694$gRGeLLBW;+mh`!J5
zPu*L(Ir;D#?C54UqS*YzPo{UVML!0d+Ay86skxOb|ESkLwk_K4(qItJNuPSX&4la>
zB^y5~FZZ=L0hCN1L*bm>QiYu7Po{>`LE`s<0%Q*Jg>)hhzdZK@u6@ka>Qqd?xPJ9<
zL-~5$?=xTgN=nb*3wtYGJZVHy3u^0`I56%`aLo+Y2rD`591wQ@2K_(Z_%At@INE+`
zas?e=?{I1l$w=&LOsS$4D6xo(zQ-QppYN<P$neiY@xtzO`3+YaW*2u=>B-uwgUd9#
zOxv9gPWmLd<)kHvEh%X%@>tF+9;9uWs7<Hc?Jpco_8AN7+NPl3T|5*e7{Q|kEjN9J
zXn^#WsGVQ&N;9_hECFU4e8Za@Z1xy-iw!juofk4+F2DUtD)~<?^oZ-25tpdF^7oyt
zKv!o*S*LQyVClzkzwlApx{>U{k2aMxPsp?;lxm#xL}aDJf<pYnLt~CRfo<$PMo}$t
z$UDfkcORf%I?PyRdE<vl1~jPl<#4z10OGV{RdSKp-txoxp@lj3O6#fkB|YIfT0jL%
z6yBX#W3_Cmcv0f9#OzF0ZedQEcviYwo5_{9^Lvl}tl;@G?XMR&JTmq)i8~J;GszwK
zmCKgkau{8C_)PcE#+q1LvL#;?cGdZ^$RrV&u^+6w`Y~gb5VEm(^3MYPJ417nh}ek|
z=lZw-)|-15s7$|~2TD|6`X(1f`uh9;*A9>_jlQ`7boGC2I;k^R)O=Lvdm=s2(7D{8
z3=Zpa*(U?wtxr;+*%4W{DKtfmWRIXUtkl{xr6BIA^Jo0LIOD}D!9)Rv4#k>aF}?iE
z#DL<D!;M#@Olg{+!Lzzuq*?-|eJ@`PlV1}xgO#uJ?I4y8j;(V(BKLz#V4lVK(I;&i
zxhJDf`EK;Dee3P%qrBD8F#>~(1aH6J|7)xJPb-}~E)<MPYtfajG?_>vYb>xI@N^H}
zmE1Gy*XZ<YzR_g%esumP(+2DkX~b37elaLCpNr>5d)oyA%nMZ`O~}16eY9K?_9`~9
zQTNv{1z5fMAKw;_HYs1O)|95BUbs|F5(zT*D7>(2jPz={{7bsFJu3TqDDIBzT_Uh7
z3r_f<FEhS;p?64*mSWH$USb(O<5tl-!4Gju7G$vsNb)*1AT1nR)gRioaLjZU>T}k?
zOjEJd)2ySgY|ii<U3@{XjTCIxEb4eK>I_T@I={Es4ptOajZ+#~KMv2&&BK2*OpsK$
z^5=tLjzB2C*uGG-YVrW)*I1$Xs6T`NlRIUa^4!<~*GhkA!e?yH;%PD;k!9|v@eCq&
zhue2kPRxgIH!46D%|0wMp=55zwvw$8J$Ai@9zX6wJ8+_Ry?Q|V#0)Vrkk<<8|G41V
za(O#`erd=4y79My-cPLBZTpw=As-ClPJU>1IKNT)nSbKZH$_Mj3kH+-I#eTh*TO45
z>jQ!pVW6p`VoPllp7R<>6xNhf*%;Dc_d~@5_zE_k)ckR4FWo-k=5pN3IgW)JCHGZP
zJV!~|sT$^mb+tFARUBAn4?S0H%62!8CQrOk&)#Wl3Zg9Zv;KX_-x~Y7!TkNc{QbH3
z`)m1kXYqGm`S(!a|JNg>P%z`oAQ}$LC%<b64dTUq==9ExRwn59CJJcNB7cwb+IltB
z7^jm-^`!<qoSCzm^{tMSn~xN{?7M|wyVe`KGJ8C%AcDm!QN*NT-B=;Yp!IvBz>5o}
zjxMooD5%PYOk-X{x>ls_VfV+Ah>>pL=P~ZXLNMt>vU$nCex6&nZ)>>?!$$~fo>_!8
zVlnPOMG9I?_snm!dZJm`t~NhOHJOk<`5~M&#aEW{^h&*~V@Kn=8)$=4tfX>bDMk{R
zfY>gqO)f$X=uc(U-O)U?p{x&X2PuZul^JSn&|OG@Cb^R<y2rVn&)xhrT>o=H5U!6E
zcksrwqj<`2e?mNCm@PoeZC;J23%3Bu#25FQSXDqQ7Pcs0Nsx)QmvFiHV%4iL+JdhH
zbX|)JHo9a-gO|f8MO`H)SFAqeBzQ#(#uMT0@~aKFsbmDz;Zp)pu9(kd+MvGu!!#k+
z6kl{RFEj6q`qN($T;WP&8xSAY-U@bItLtHM`0ZDwh%oJd3vuvU-2TTZ*WibXPVR@l
z_h}imxGZS@*z%YR)SfiknVbK$1i-}fzZC=`zhjstoMJE^^}4TMZf?pTO)F&X&AkLc
zgM2}h!yNTm6t_@2npsv?&cREE`MeIhF(`uD9(CDlX@u+@ZI)j=1#nBd)5yM^$dTfj
zsK;4R07or;9&eZ8j1zYG#a)yf*$Dy3DjGB?P}vkA@)y`Y>#AZfi?3b*V1li22iN4G
z0j-`4#{S+*vdVIT;I!l?cZ^z!yLDf+G`-w+>p;W!me#0Nw$ak|9zVYFHn1`uLZ13M
zcuZqf@U50Z!BDV(Dy;GFd97H$TaQZ;Iuo1|RE3cgQSu!EdA@PD`iT?yg3?+MeFUTC
zo{6!Wk_Y?EQ@{I&14vTdG(|>w*|;m8*%0sbK@SP9iShc|o$_(GrcuKcoAiSk!tv-W
z#2_1+LDNFzSMe=H10(yF#rwErzAbvb*zDT$v^;bEv@J7aAXdJu&)HF@x8<PF^C3~I
z%{qqLS50L?8s#t^#`@wT;Dsf}c4ppaBcr_>LfOqY?Mt7_E8^ZEmsr;X)%;|#OQ~4O
z%divQR~*O1{dk8MdqZX&^#XJ6MeG}_Ezqf}g30^xT*OO?YlnE@u%iPTlkapJ&uK$*
ztd>dGj0zIM?0lmz!FYo#f46c^!^`(WYku{&w!ne-$e&Eyc*1laV+sgI?ks$MjEAI!
zGuPLwdF^xKobyuWD5%)SyK8Wi4(=$PxzOcq-8AUZQ{oWXN{j_9M&*@7D+(v`At%NV
z?TFESAKy6unzDRj@l{n<7#7lTmx4ptj6ymUbgQf^D?B{ac&goM1zdY-q10My%Gx!w
zaM{!wc|?LdrP36D`udXnQIBY5YC23<eGX8lhR+G2%D6pSxN~%#pH*%-Tm8II(=hGw
zE9)CRi?#Qv@nmjt0;s#sg+jc&=~DmF#xdHgy!0zvP|kDtVX6EumynM_k&o`6q@iIf
z{?WXjSaQ^|tpmL^8fg?1UFwd<mwSh0liAIGGMyP>zvsflq;;%s&aKRvP{KusJ(fjv
z%T2TPCcz}cmbMl`IW($q{g6R(NJk=APhSu!YmjEF<yQ0HeI)lJ3XqZgj<B%J6EiK=
z<I;EB-C^?pOM$H9NLzbby<dJ7Ej1)qVbE`LA+f{6eUv0oOah{JG5A0&X=oLE)ZYV|
zMI-pqU=*}C?iaCyFhIXMpNva!Mh)W&eg(;kmJSpv^T};!A;HIHd;*=cZ02wviSN=V
z#*LJMn7R0`bHa_sZQGT0_Y=j%jDvf1EK|h1fH<fIGd95pNg2v$WY48s04Fr7>(04T
zn+S5=C{AMi$&YsGR){sOdcj!HIc305sTOL^Hmsx&J(B%yP~r3E&MMlE`o>WKm#pQ9
zmd;v(wPh;^nU2;nX%BqZl=V?%YN;tup<~ZjTBUeO<@;Ptad>=&#IpB{ph{MRL@sNJ
z1IAbGL*q_13PyoQ4Qkj%?|w*h#{i*?i7XP6ahoDRhGQ-G&CZ7}exU6H?b)s&7ibRV
z+KW^(;hwh7uj4$55<uGCj`z&_F4+N;9)?O^YYeuRjq@g&F8qK?yO;t2q3t8xUxJI#
z-p^GY&BK}@c(TR1R?<NKBr|(`EWb*pS+MWi+n-DtY#zfwaSBs=?iar|X7lYe4UFm^
z+&4l=dcO2f_U|8@b0*3XI$DNy&6Zn;9FX81)NJp18G_0`pHC%9?B2;uR5Wr_`tG?_
z5{$IQshbq3qxjVOC8Ck@HM~fqcT~vgAMHeUcxXYbT@03UCm0E<anEoq>V7Abr@t9?
zPqAErI)mWvPmRV~{*F`00gSj|DTSCy%F?T=jXV_z?wo4$`JkX-e!i>unHhW}(&O*V
z&>gwS2vN^QO-7kNs)4rrSbfsfoj<e43MWx=ZzM(QNp#xCG+eCyj#8$*z!HN?)ra_G
zRMC=bR8|Eu?YK8R2iR5!lO~v47w(`WDU)_N8E>~KtCZ||gF!fcE^4wC!qyHTGS~^t
zw%ZCE<uiP6&W0LxxsdPK)-@c{iz`6?@c2lp!%rq;(}_KB;!h?j`O3pDK<;0ej{Ag^
zvWgLh_W;mtma~E#vbWy)Cb>WCkL`jm#O)ev;579#5^UV4VyxKV)%WnM-n?P=DZSJi
zrp6rw*@VuWq&9JUdeZtEq0grkpZ*e5_{PibqqVDE=Jj`L9*-Hm&D%ZUL7sH$y7aEU
zMzOL;i3rZYvU2I0nIs6Cn#bPjtcWKC$WnIhrDUPTtvUF$5cQ5~R&!h|Iv{uckC<iN
z;8SlMI{io06Uk@&s@n51j0x&5D(F7^>BD~E+=5uh8-;jPh~)`Uy+jDJqFK~Jnt*%P
zKwz@%i<lMVb7AK{ynB7oIj0SWOAt>_NjI&zc>OM#x!~gUd(TgK=d>l>IQb}1+Vp%Q
zbB(7C%lTWZ=NluqZ6lZt-~X!0zhpWZS%qozchI=$oqGo|iW;rhGo7B)<>1d#>$lcE
z;({`ZJZaZ#8fav)w&+*5`6%3AC^Lh<E@JOkR-Ksl6bwPzs8!Us^?SN_$eXla#+1-c
zCW~rcI7?H_f%yAG6Ccv7`{RZ|2`|eBQDzqmvC9u|gL=-ClopVY@F{?G!MVyYW*eo_
zXYICQt}>vjR>g#K>>Knc9a&bgrjbb9f#rh``tkraR|r%$<66U*7k*&QC|2O`{zfx4
zH-_k8!RC4<;t-JH?SgnG{cwj1I!{A>4ULYBq^kJS3Px0vUuQE)`*srM=gZD+@fxhW
zvn8OCo*Qq}?XZ(r03VTUgRsPFJm!1sCRr6UnjRi)*RkQp*J)%VSJ$Q+bG1_X^#W(c
zGMDZ~3(Oa{BO*6fT;c#N&94#Ikzl#fJdsq6NRQui!I&5U`~Vz)BWde(Xc2X_0nhME
zG`VHYOu{?&eK}|J>&gEgT(w8XeAQ{v@z__DyB!rXBft8ra3>{7Els;;zZQwh?l8yj
zYqXu8=Ii!~0)_D05Na<-luMV`L>zO3wBx{FVOYf&UbI-IXMLyFBkSBnwfnByQnRK(
zmxfZzKuT58KbiW!FOO4Yq>;`wYVHZKT(a|n$oVC6P%E``&CT*N(VKQ)q<oyVZ`s-*
zb?`Zs&DPGu47x)$Yt3@`)Vm17MQ_`uw)nRd+f{#Vj4DV8uJv0eo-&4TO2P2aY4<L9
zxJIR~Rt<-cbnt2k-V`-Q6Ezj7plf-^Dh(n|X%}GL=8z;5Wr5t<Qy09@z0E2ss_kW)
zmdtYrnUxc+RZYU<B3Xt<HnhKl>85G3`k`n|G=E3D7~w|PvReq)+|R)QVkAMFm>www
z7oFfNhFN^-S^Q#wk3w{rPJL?%RF-M<Ull=ms1Ita^-J43(+svXFCQl-(~;~xGX^js
zN2jI#wMzIF%$yK~vm5uxvKy8C+^)#XbZ*SKvS1#Lf|H*F^#T1_Y6Vr5AvoIuPmRV%
z?^x}DnN5UR_+Gi2xscSV3q&TRPr5h@8=z9MSTs>MfaZLXRbhIbe#s={@s4_M;8N{5
z<YP-O%C);AYBroDd82qSd~%IS&A<)IXuvk6RAK@K_+}OXwHtF>^Hh}5(QC_AHt3>I
z*Os<=6C&UG^KNRE%JcUD;E4*|q6;SS3N-%2+bt^A9uZ-pyJb^;?YknM&zyn2lE9f*
zN|LeVeRnIUt@yZ{1kjHO3g>QsR7C|%-foLGbu)2psX}QvR)Mi)iPnuBl&3mJ<~sG0
zN?%73R+eMyTx(P|M!UNwv=9pzw6tYjn*RZ_%doO|d#xYs0tcNzCw3IZ6q`S$60KI6
zg%kWmR6<)w%dnQ-?RuYA#m?nUURF3M6AmNM>~3g~7{CB!x&^_;f=N<)JhgGO<zaTl
z8pFHF2GP$I3~Gp4a3=NkjUfE>5zwdwWc#?a;Ay=0?rFMZf-^xBKZn!L$Kms^a-iF<
z-+&YOyeA_heKQhYCp=Gh`w|-)c21u4U<CTnfS@e9=H>0R3~@Qsx}!;dOtw3St^g<q
z2^9c0e7krSz%(FyO$?xt^#kG*%|R4%QQQ<cX+COcqzd|r>)yA63ge&x_?IT~9Jc|`
zOa1AOmmm9l4Qp1{3GkyIH1#}q;BxqFg<S?>=^H;75RQryH-@w}J90Yq?{#Y6+c18E
zN++lKtrjmjPs5>Y(yGP}?pD%8Iyw{r>w9m!-##B?2oKzQ8?@Cn0cQo7`D~}d&zi5e
z@^HF?RC&$%K9ONjNCT3v1QA_e?wuxauUH~_Tdo{I#XTfKZ+|V0E0x-bk8-9d?)4)J
z=aQD2*zJu?eX_3^lMT$3q&f^B?oA3^cW!&nH<3-0=d#Sc2ShDTh6Kjft(5SVhEzLZ
zXzBC{=Q~z}A__@A=@D9<=Usf^j$l%m<1q+Gfd7>D%5b>jx`OE~W`VIfS`)LZZO<Sx
zN!HUkdn+;aWn(S&SuDZ2bi~IWMX+lw__*nePcNEZNA3-XwGrL-*YQQ_;Jl-H004bp
zhEX|(fg!JZ+BQi!0_2Y@GTaPUe`qy;jf40pbOgMfF|caH)nwrdRzM5u!Gig(7RCtg
zRGx+P2QX=R+(E8bRW7B9m2XA3p#1v^NA=k=0062s)$j?2G;6`z9m0*b0&uvrV`sOh
za?&x1O2IV6P<%1exE&8z_#Hg$K_10?i&`R&{VQkA`=OqFRIu&^fH$5ER*bmqN<wRl
z%)Dia$EwWkq+txav6e$zUgeSy_Cfdo`5o01tdQnv-cwf4*-PZ>R2IkN{ebFk4ySy@
z%&^<wF!MG<`OSe66Bs@VE|Fyi=ej!atX+gcrvy^0*y=1*Jq(G*XJR$;Bo}DqJW62o
z&cVl%Xt16~|C!sNQk)s%(z27RDsXY$WH}JRcQ`(;$DuvY_XA{hCO5CYrGHE5IN`Ha
z9Bm$(jj-E>j#@{m8rx*nR6s?eipN?RffeiZHNN{+fuH9vocO^-`(Ci$8sLCpgYPo)
zTtN*ex(@9st;U3~ETmam?9z(OJmreJ#htU2`uA8a4T{u0iTc<A^O1#cziYYx%g>BH
z`8`5Kut8L?@nmKc5Lv>V8O<6atHa6v%i;OYpFo64@DIfkvI-plcQ=Gh_Rpk&6Bely
z7Adm}whI=DXYs1Js~HU5fAYcm#1JZ3oB1lxCmy}Ysw1*jEUHa+D>+OX24X(`;>4;o
zzWiOzutCYE`YdB+$N^%d;B*D*F@WWpI~>@yPFrSkUgQmo&*`4s=yP)7+w`LhyYs%1
ziRM*tHy&tVl`Fv6Kegi_q!7}H0P^i6D@Cn6V+`>UWWDZjzrlL4c<tO^2rc!7B6?xU
zPsSD{(~8!bglUPf_mM7&IoOM*+?LTxWIiAHCl2(PgtgEb1?A9hREMQmk3N9YZO<6r
zkbaP}dxLLxkJV!YrgXtpE?OQgj@PR|hh%3fsx-c!_|jE3EV0JUyBw8PSS1szM?L&X
zA6oF6K|=11x|H*Ds-tApn=%1#d9UMsUV>a`(kR-y*}9aGt0NdewJ)VR&mjgVq1W?0
z(Sq@)bW=<~m;ygQjClUw(O<0k|6cJ$=6uqI&Ck75+lSFs6?aAxD|AA(=D8OTq~}I@
zu_eL_A?>H#Y;o-=G?3d{0XBEV2f6K4!^AuB5qSzjnc-1MfVG}d!R{AyrJV}hkT<KO
z5(3wMM-B>(`wp7<0u=2gN#(`*9S^P~cvnE@CeTVJ-=_&W>|RPNo77cHnRBSK!V}!8
z78DW3<|z*!x|HM;rV1GTs**y`x5uj-*aEAo_>U??*3H;EE<M}?V)XAaMfr2{5(+*l
z@KbWg6B5Rs_+ZK(9HjK>9z4vM{Dr+|gqNl2MZow`+_q=xe-(A|*wpRKsH*ORbozct
z1rra`Qy}wH8!-&T2)mj6lj(*OeQa=#hf)8gA}(Q-PU=3?#QqpF`a!D;3i=1CKmV<A
z{`SGepG@8<UNvX#*yMGZ7s)(VgSv5w@2p?9CdeOKF)Ft)qpZU06WB?wirD=LOJ_sJ
zK$628iAr=Uyo*{?5T<PA*5e-w4IiH<eJ}C#B+HwSv%f6UEBIjg#sKqG2bS0U_*5@3
z^q$OIamrFNV*jlAwacfc=A|vl6N5S#Tg#RJ@m>B_`j-+{cb2nHXi?rJ&SS1M@J=5E
zM)AQ2U#cscpPgL+5zOc}_Y+HT#g8fiyq`2#*(tR~!fsqi3O1Ga+-x1vtu`+lS_^qt
zE^SNdY*yWbBpU0yAh|fpOq|su(cE$2AX2B(`$R#8Q$=4d0~oFJE5u*dFZ8?9Mjuh3
zrwFtS2WbgEuk@QfR-}H9>Nz{lD7Wh{p!sdK&G?JRi!YhWm9Kld)+DB3gAQ#V3pHHD
z&|*vce9i7yCK^AFQI=cx@t_3@9<<FULNs<v40Od20C79O>Ru3qpjUyEil|_A>N!P-
za{%y^&fN24`r7T+r?He+d81?-7${utKl6>O^>TWFXd~N%<C)*o=6<}!58X}Pee5mm
zPwEO!YLkXqG$xh5^i92*@Z+)!)TNO0YBxvotjsp<S)uQXiMlk8%Rz`*t&huBW~<HA
zDwaoUpLwJ$HNi*mfrb-3GPBs4O&Z@Oj7^{>WuJkYwCrvjJGrW~j7G<n6Upl~=(HfI
z$HV5XGc$gO`Pv8$2t+g45-lu7yf7SxhI*r>03d%ij^-P$o5~|?o<E?peebdH=OL_>
zaDuOt(i(;F(Nb-h3Br(d6CZx|4!ek~-o&ip9Q5N#Lo_>2v>fDqkFK;jSZFsZ(whII
zn#j~f>Q||GV!gtb?~RV;&ogR?XXIhn-P}NqkrE?%iVaAX%|qFIU30hGdloNFRUljP
zu+t13m(<2U-@lfeUs?onCO6H)28nmcPWt8M;`uF6f`VJwPhKzL2b67I5#8<V4nd~Y
z$yn2Ie9V{i2#-#lpG*qVd@E_?#~6y!`@;bx9Xx2chlIQl4`m9i59bIA?V197ZU8@J
zBvKv@#ws~@fntn(WoC7NNqN87>n!I>9(Sn<XODs#dja>W6PLk#6-DX-W74b{&tAJ=
zRfejow+e94N|N-f6w$2YS5tg5sohG)9335I9L763>itC|gqQr+01gERrK)%Y?*ag3
z&SO&W_<|XEx&8UakwC*Z;%#W;&lz0#h(N(IjevHM?RBTcQ$mUJ$EM49h$hYm-vk1D
zcJz=0EaY@I$B!Es8|QKOOzX;>wsPK)IF;NeA=abdv8k`dYde>uXc(^{M=km~*|}$~
zl2`fN4Jp|hEWODoI<DrYnPM016)e0&u*P+goN*qosuHlq!hkoA<sZw!BeiB)t%DvA
z?`j<0?+{7fp(!wMu4&>2?RjZ1D=~cG>3K|gk$hxVpdG{H=)32G72_=nC$5CPntQId
znDc+f-!Ky`<rDOiDcsv{@JE=Vb(Nkq)|~^JGH_bl`LIQeh5QJ$`eq}1(DFroX>-<i
za0eSva2o<Leo-wn_-&--Aq!h}gMs%^&6~d-f@`|L;87R%t8}6<nZIjFY|~n{XR%2<
zX)wvX4Bl3?3>Z18spjo%Ubi^UPd>yA%blvI(!7(#67XQ$Ra}aP*NmfovxfZriVFe+
z(nVzFR9P2@<x6d_mx<;-ew=eJH#;jkH<3BY>fSxe^ZE%!U?ahk*IVBTpA-Um`-@Yb
zSk6yG99jD%$-i(i%YJ9_$DX8@a*s8tYBtHDim~0Dr3%O8U4iCL<W+SI>?gWzH8t0R
z<iVykEyDOI<x@l=2^LV@!pdntRg>TcBkFI!gPRcN9fp00Ba8m(u}Pcm#pPpen>0T%
z5jUr2>`_Bs_S=2^E^%ui_87EA(G=1Tc>t$E`b^ssjlss^4w$|#80Pz5E;9pqy7=(Z
zN|3zY(a*uI)=7mm29@7pYVzM%>YK}M-Tjb<&8EkSyWi=Lio%D;Hk=+R<Z<6K$4dE@
zlP>;bI_J0tzQcCCGY_qG9}z9=i+}w9aHFSW2!>RP=%op(!0!R0e-&M6w}LnJ>}*0E
z;iC58DuT`2nJp)QP%zoX^I<mu0dC$97b%3)g)MFkPKCZ_YW*b?lOdR}E6vzFc6lez
zePFA8sB3ipWll&2iZ&3ox;Wmv7S>#}PVK+)>(qZshw)C}yfopE*Hj0iw{oZYMA}ed
zCb@=*oEIoqaTUP`RUe@i7THgH7i`%t*uDVb4^(WUzmi1BAirf}P`k9T0<;Py3*%rE
z3#a54rxc{PWZ~fe{YjRc9b8yde@oao`A>F~<m7Y{HvUpsSq2*$gN=XX@GxqWbwQC3
zVD;LSCgs<eh#@Ntud&{xl6c!=JR?k_rS(fq7&2+%w`C>(5Ng|xPq?~N!&NPANteh^
zON_cz3*E~PzfG`hFl1fSI=Pfuo88#S+Z>$@U)Busaj+n=DUNgv5--&xxRVp)VZdCC
z=TpxH-8BVYo->!sy$ON&;cKAVUdynDNGC==G8JD|R%C1Yr56U9cbgc-0sK&7Q6;0x
zHRjW*AWI&AhUXPtP_@Z7{!hbZ=(2go&T}C+v^KRYcA9(xUYHk5gc}f_5RFjrGY<6@
z3ppc;+peJobuYXn5QC?!f=HG<lP|pS*FW)Tve-7Vlg;~lJA?1_Z)m!nxNOq9naWao
z!<1UKL=9%sHtd&Owu4W68zz)ZYZ)poRC-t^&kwqgvVzyPT3YL0zqP-?yW}V5<zdy@
zQtw`Xo5F6n;-(0d^C-G&6%0TrY6-FvS+0H<RX5?ZT=j%BS*^BnajwJ)4bDMBsYXWE
zcwO`?`$n^PT_1IQxBg}y!A+^0z#_OxJ|+8V?w)f6MW{@~=?ZA8{qET}ej;G_a~<#G
z2!88$|EN`jz+3OBu7#l5Gs2JiD^8cW`=nQz&JI^=xu%CqijZ2Iq`~oSqkcscVuhwR
z$qq}8$bc_{6<W+*QL4I{p~eEVdIdUp<R{aY`Rq10P3Hf?42#8HWyk%A-{Oxd4<q9o
z=+~O5kX(G#*>kPQ6zq$Ikiy(d(=`QBI<A|OyD9R^NK*hsIdy{BvkAq)p^Cy2Sgtf=
z=e)T33m%l;zBDa!UTc5bDU+<<_faMRtB6%Mb{twUDS!{$cVsEm7tY#DEAJdgx&2_e
ztKwT~2BHog{ZgYKeuhS+XJ_X|^WtBh|9r`D$paAtDIfhfCorsF*uwgd%Fo-D^bwo%
z_-lhHg!rMik#Dcqwzgj{QSkIWC1nMvEZLnRo!%&|UTjMqSq5lI9ZfAB3H<o?7y4gy
zFRwNg_`Nujku)K<a(-0c!zYivSQ=a{roPCGfDkFQUQ3#Euy&C?oCCP(&fm?;9f+w#
zkl?2=o}c(y*PG3|s-ggE<#dpEpsm3lKpl^iuN;Z(?iy`51xs>C>zm;iRI>ENV|B|r
z4R)%*)t#Lre(#dLD8|b%X_U$#TgN+9!%?i0E|(6^S~+u6aT1Hj1oPD!kARY%+L#(y
z5JCVz-ldZEy_-kg{239EiZ$PWRc)NsURxV2S_&&oBk!EKyQ9rdDE!GZIb0w6?I%;L
z^-Ny=O`wGTk}mL9dqhCjcSBEnw|0Gp$_PVjU(%MI&A?wmet2>|OnA?F)u7~Exs6QY
zR2y*V|8XmrPvweYnGHiJyC$|#cTr_BlXQYWuM&p-utmW@HtFBa9J*%ZG@DvXyqH~Z
zS$ZHQWU{wv3Ql8A_G}hQwtZ!*?p<R(Cat&~T;vF9nDZ^dSC0_pr(3j&Np+-<c6@t#
z!MbdeDvd*t4Q`mBq*97<X#I+k1Wd(VCw^gL9DmLiRU2(gbzEv)J>TUgDh!iGiv_#(
z%<m5KpPVf+8&)cuC>#!FF#oWNa8Qog?bL4X=ZZ2A>cBOBN^=`B!bn$62r~USCH*go
z|JV9^^tnpvV!ib0(V_Hjet98TnHk9bSaotso-*(L)o_+7LXG=F5?H+BP5)$HXKsRP
zw-WifIn)9CmIE1qpi|Y{+5LHw?OAkgD7hBHKEeR%+EL5Lbr!{t0oP94grK8TXz`LO
z9t-lSGB*K7@KkKRgLsRWn6_eA<ga%p@>PFt5cT@wvK63|vzG%e^6pt;DnET(su&yq
zsoA_)xbr)W%<_75lHFU1KK@QR&}BwXhuVZvx!(u#H9x)6CuI8qAGEMRfAygO(ZXkc
zruHcl({D!tjWz9x07N^d%6f?<rsFrS0Hy!tjd<AWH{K%(;4D-8s)&cpek<Rku6^70
zD=z=Z6bC<4PB>KbNKelM&i$|Hw%<e@k5Bj9#^q0l;s<J>2UR0(`{b-A`G|o}KW^<~
zmpq!}>K=3kG@e6W2j2PjC$1(O46<KdQUd%b3>Tk;J^il*{!{-SVcyS;HIKXdL)^n|
zM&CFl8i%B)h30Y&U5(}z#6LGdx(ZH7elmJt5oX5o`(*O`waCjs4y1r7`=3m1+<y`z
zoh?$4K$MwLRm>K{?p(_X58u|!w0x-TB<^;lfv3!8x-{6yT%IUp&>=I27qOOC@~g43
z&EQnoaGt%cpV-%m7Ka`Y{cG#^&rq0b;}rK&nb}al=ap-cjK}Y@DZFWEv&yFuR}-1#
za*T2Z)umY9!#!MEBU!G#HM*7W?=}NZ5NDA=zu@c;Ok}y36@K#f^M3%(*P9h&Lug3~
zh;)~@ix-&`AH|c6Nz<nTVM>}(W{y;Ta4VZsyw~%c(kPbN)bf`G^0M1s8gAdWjofqd
z`a@v|s0;INYgs`})-#9Qh$rTBCYGl+e`S87*Jmx&zXv$H&`C=kt_Lwo<E>V@Z9kS^
zPCuChI`=(9_Fcv`HGc=r{jcfOQe%kROa^sUJDdT{@!i|=s2<+&zkP6@QS|)#^BraS
zc$8yEZpv!JW$C;B=f^X@5kgAO94fh=x3K<ziQ$IVQfG6+MLnbWG_j|H4Kv}XLX(lY
zI~KN_))D^tfRkW^+=plunPxuOKja8NO;4<BWD89e@aE#^{w%X|U%LSxFcF7pr}jo1
z<{BS0xoAK`L?7TnY!IEYhrRIBL2+-`MX~@h5D#;=%AVn($~{n73WUFsrlW_b>eFk?
zP0-iT5rjzzDddMOQ{D=i);yRv%(rc13i<LMdIc4)1oVIZV`voC-cqeO<{VqDLH~*s
zE;o?l2e+6J<xr%Dx+f8XF=(8rDt;x=B~O=CrX7<i<Tx*>z0^ralwg5B;uszgk;-$y
zJiwKWSAx?QY|5JiMZevvpKtbUx4LAXk9lFy_!4o9BO=<;gIK}O`l9WDyQ!r&=?Ql*
z`>=w_<l>tb_g(y@JYIngL*{~vD${c2YAfNg!F-l@27wnHmKS+AQwp6UUb`}EhvL(m
z{hE<DbL;YYqqK#p_h$Q#Nq%G;cfP#g=-L4HB*2x6{(Y(ZOC)QYZf+xA%ai(+p2LzU
z-5A)pz0u1dm(yoFRx=h$vIUsRxJIA;{pW8T{M`=zei#1!9{l}H{JUTHyF2`QNbvs}
z7Cl>YR)vnh3-Ugx`Q<(i4XF-WdO(|%HVnVr-5nJnE}jYi$fVt7+KRza04{7aKBonv
zu;-+@8-4P911^0tS1ry7kC9*rquYZm6BUzM#8a{)HD`spG7H}x*E&=voru(bb+<WY
z?aQ*KY482zOOH(DXD6F&xcIuW)kcl=OpI8$8($2~mPz}!63Ym((JXc-Xw7c4&I|5F
zmTFu_g9e;(YM60CJo5R#Hy5d-;DO5N+VW`vPA*sP!?MXWGlvB?wy$%x*q3Hmj`v2^
zbCdASUQN9X#emd{JqC$An7&V(Y(kVXzd3{)yhW-@(0!UyW(@N3EyaH_8NGx>t2l(T
zICWyIHuX3)S;UJ}!#`PiR*X2Wwq!@J<<HOP|1trbe74*3A0ukG7sZB}B60cr*;1EK
zN$z`T`Rd?Yz$fg=^OLsu#nevoexC2<25~!+`53@ic$fR&C(}17!`ZfxMRVWlv=dOy
zZ|$k^3x`Ti6xFV}H^Ak-cLteMOEBM`;FOtAKaG&Ao6fKYi~x;!jVH_c<w?*W6<<eN
zE}d*v?k~K3zb^luwk~iVWfgmSPD7$X8d|*?e!6IX4x)KmB}}1BW@+0OdH)mFp@k9d
zt^9g!155o=U|u^wW`6$`ISy|xLR$FtS4q*An!|Ab%<V@1W=d0n^NdbU%UDacw1b7O
z>~{Umb6;R({2z$}|23{XcnE#YW#m(CkiGHFBX_fI4)8I*^f`C?s=m1{(8s&Ys`8>l
zW%FFBP2tu;BL8i+9CJr#np;<3`na{e90McjUC~yC|8(o@;ad0)?COcakHhN&HKPiy
zy6(nCZ?FAcxef878btD!xyyc9VxjrGuj%gQjbxtFYKlcD1#~JCu0y<QiEgFzt5LiG
zzg_&}-(~w(*dOU<SdT3)Z5gEs&7>79vzaPr?JX53BG<gl90@pzWt2nA)dthliFjA9
zA|MVaLLeRhnLOKBd4^Ue=}3tZhuWc!%e)*`arf?YCWw14W4u)RqJp9FcDTxxn$c4g
zkj~5ml*%XJOSqib`Z6;!3i^eLu6aP+Tx@fh2Hbv8DO4)lMtiYgjV@G7SFDjzqNFsw
z55mT8`Wwh2lq(B)7*`+W%b$%RpY)789d&m;AKuv<vEYH_BDxP)zhFlPYx=U~OTSY<
zmWkYbs`=>ORC*+BTv{;vcCF3CjBh0COE4U}d@<Y?#-h2yInterWdGqCma!u*O|$^$
z%51jRzn&I)3cQ4WK$CXX9aj5?nJ-Q@pA>gowz~bp^Uyi|16{bb-<J8|j;-2lOmf&h
zKo`CHtmFf5um5&{-@nQGj~%nF`=trA#$m>-eNOt+wY+ggXmdA(Br+c{>UR0X<+i$k
zeG`c($-x+z9q*e=OuS6qJldmPw+^X}VH?Y>Tz|0vgX6a;0<y5;T>VcbeT_7KQ<hJF
z)JKM(8g<%?ODhcwWrm^N@##N+OduWoT*Xm+aGq`x=cE}jVwOGCDc<My+`V}cBk$*k
zj{++c+DD%<RU$ibXpo}Q_GHFfzM|bfIFk7}8V;V+bmvvlr8X1Gr|!nc3r%Xahm9B>
z5ahL<BRETx5!lBjf`t`Oe7^hTXPZ^frP4|E*VOW*LyEbaMq+}y!hu4n0z-UM_>)(5
zhPi~=G#^FJTDS#%I`1En{p)OhyICE)syfESl4kxEuWr>pzi^Bn3FGV7$@Tv+zY?U2
z_oG{UX#8#y(xyVAJ}~RtsxRwftV`jM*{T*oru`ta4+l=|^{T;H!vHNW8oi#gBk{*&
ziO%PDm30#fKR}7s+*NTaU`RlClHtW%bNGQwZIgJ!Sx~DC1RM?lhe{ng{)l3dXJRr<
zIfe&^#!m#F`gby9o^NZDT5Kx5Ax$GC182F8ox2G<Y3)?BVO?c&n7a3&F5}tpU!Udw
zd6pm3Gf71QWj$IynGPh^kfZnIe>n}Du=QN~0jyYi)Sq!1N2p%EgatBzbTq|3kQ{qz
zXkeUp!sk+Km+9~g1BK;54oBo2trp{;4&|sA!9@GE8m(UTIW~`-Mg^!0+U<j7!L)vJ
z7T#Tcex}!2%#|>wBPcDy=jeaAsoo0kCEqi*d{em>Up#oSLfOL%n>sw~Dg-!(Da9K)
zcCtCz3L1L$?-A>y<9DzUViRcHV<0eA@_}WcF$kpcr`7czqW!CEkIbfk309a)-h>2X
z<el|qdn&?+%9s*%XbT({gU--BcAl(*Vr-fqAWK0>*=Fdlac53(R(=HXY}dFk1W`?e
zFy`l-aS??Y8HZYyH=-gk?6XoPly5jltvMHkpYG^I=hsHi&_s$j;|p3a&s1T0lrQwO
zdfIj4doB09H5E)eKhR9|>n%P+OAfZ^pnq$B^bgH+i=cVFV;ZxkrKg^Jcw8+S1Z1Du
z|3*S?R?6=%&}&*FCl!C;ll@OyEq@Osx)YP0Z3B>WP6@^H&yueIN>D!&iEi1m5B*^c
z-*axeGV;re0K65`<>}t-Dad#7hGE^9sNQ$phIc80F~it(vqvRwT}vH5mZOU{xm>5V
z*1hm3nBuWA8X~8lxR{W^cQME0E=JD5LWlKAt9G0@PsnOhq)%e}T4L9>?eJJEY&i*0
z>pXQwQR#=rkWP#*my}aiW%cTFo*IUFd${P5Z(`Ypxa*FmDlF~dp<JrN_=V)(PW&m;
zzt-l+T##y^Ra#Z2No&w>Hc#JnUP>M=Fn$~>cgKf9w4srPR+8G%+k`_30XuFK{@!aZ
za3;xWA%|wQqC~<ozyqktVQX`J6{V|RuuU08)fEVVUi?Mba>t@ib9J1Bm>Sp(z-xp>
zt~n5eaxk?d-{|Q#Ondm_V}DSP8hQGDeuf<3^l1#&Vy4!VH)%>e(cw$-AAf4-@9+Qr
zS3~*V`u@LPk-QXbumI-Ie(4?ofAwt;=3L18jJ!ZhDbK+gJ<d2QwP&z#2j`Ytiehp9
zZtv0Gv80Q-XdgAmn%4Wt#2Dn?VfMx8*YxmTFTHzmF65q1_ZjfAbKTcPANMLv`I7v^
zx5~%f6s-rzICWm?;OI|=t2ryY?MO7zmYZ@ebC$>5!^$C)&QAxQJ$4piN>*%N5zJ9~
zO6|Pa7F7Di+bn;s*}ii6K)WAWfL?FD+iB)JHdHaKk(Yunt@>CWTxB?`6VtAqhObO2
z%i#VJt|8PFdFstA0U#pu$>(1RI+AWAEk7Y2soMc-KX}yZd4B`QWZP-;U085)TCckC
zEwuW)>#s5Lf8P!N*X{lz9l5}W_~F>Co!1j~tg`0r-1gSESNYoV>)p&M>k|_v);x&y
z?Cm@Hijyo_&^Lhd##6osqiyK@u!`Y_2kK?V4*#Q|zvST9%^H_cSE-{bKbiWvMPRg}
zGB<(y_;H=P5x3R6_r0=niyD8c=GR<?>1d+=C$i+T($L(6u0^7psIVh~jk;It(m;#7
z&7QDs=!l8I-nrYRz;U6Gy++w!HBH)#CEE=v$LG*i@{!t^5hEwmfja;BG7On_zly(*
zo?Q}D8j_a0DPiF=2Jbgm2*Ad4T*22*6N<oe(rHzBVenQW9_46M;kMOgyk|vo!DUzW
z=}0Ff8C2`E|B~{=Zum(Y=oeIpR}1I)Tg=?{>|RZrDz+>#fP?oW0Bj6IR{Vl6QJDn%
zy<Z8NyYOYyNnx9+81e#`r4t&CEU%FbrgX&3L0-b0S5r0pY!eY;T)~b4Sp!4seohre
z5u$c)J==yK%9?aVX7lA|hjV{@A$&r_TU_PMXnw(bdHPpB=Y%|?#*@?5kLB-_PUGiQ
zyc*AKA!XZp$(ohoF|u={$D+%Mz6UH<n|bqBTllV=yE%Q=!@7#iHB4u|8(=uSu3W@;
zO{~w{bn2B+jlVRo^qD^1vs*r!%-42y!S5@=V>Z-s!+_NxVw3EpRJEa1xGQ`@PA6!z
z8f(3K*({XpDCTOd`v|z`;2c5g^|OvQV^!|L1Ba2cjN5<weWSls_jhCXmwu0)&#r3=
zRt4@@OTqa%A>`2yrhQWse)^sXSg;pAWR(B*@%#L0!vql?j2TY=7%QDFDl{o4`lvxA
zbM9+{07@((A#a36gvsA4_l`D4YU-RZB>H`WC@qb2wpGP~C~pnGCIAe46d}saiNMw_
zY@^Cm(9%->))Z_yRPMKFN~NS3yF?LE>@%EHqu|#voV0dJwcp~rcdV1bYDfpP%7<1*
zqw97@RDK9g@5;sr6!6!4Ijq|0IG3S)&0uP+O7bU@4|1<y;ii}WxOL=Z=8^F`f1mTG
zeuCVizEAao5d&bMUaatwHnZTaVi&j$^%8h&&L|{m*rjG>gzC~>!F4Hy{hd5M`+X*X
zd=Fn3o@7dYRWM}`al0C4&%)h@U7t5w)qk{zn5irOf&L)>!Oy0D#JXJh3>%l12^#ad
zN+G7x{u&o2fgZ|P(Arua2_wk+k)%NFv)Duv1B1_vcju10ciAn*<P7%^fffSR{nbLI
z0s!!$juEUyKeUG3CIuHIPUbMDy(au>B4(if+->|n?0pAVQ`y#b5ETnWhOQI`6cq%d
z3rHJ9L<op9=_0*%kQxLB6c7Z&P=p{zfCPfn&?6wdNr!~qTR?h;Kk>ig+_|4)8Ry=a
z`+eCD57}q!w$6L@IcJ@<*Ivt7D|4FgP9eOG2?7Q|Hv3yjn_CzNtA<zGD?^%Cmtwih
zv?;@Pk&b?Muzb8VTd%7V!HStGmnvC6NuhaJQy_JEzn}9K!VRj0Ww=(sR#{MYnSutd
z!&D|6RS8>#maswtc@D;V)m}n+ok1oMCSvLK2L30?rH%4*<|7P)k|SjJ$%CabpJz^m
zWLvL`_6ZE?=B6k+*-E0{Ps5k=QpWl!!*T=#UYICEazi<VjFA}2(Ig0CT6`A*%qn^_
z)u7r~U^;t?@*cAu{$mQz77<qd*2W|YTccdyW+A>M+PIs4115M%W*=}wO-3{yo8VTz
zd9b9?AbIW1L`xi}f*|t(1(46F_H294>N(C{>$cO0Suzh_7bW0(Bfb)^V_IQK9vW2~
zxmBJx^b)^;=G=1lbaz)z%cPBd?Y7<N?)Ny_cY7hikM=PgU_+9<Z-@seCF~uyb_i=<
zz?O}gj`vfoUbV@msGt}dwDi6Zx%Aw!n%=YUG|o&OAVLb6FG#?JT+c>rxf&h+H-~UI
z_Q0XDs7s4va37EyuUHH_eAfrZ`W!d2J`48A`ml8@Hsy?;<6RsLaEMaGQoAdpJjLT)
z54l>g4Et~@omNxb%qoq*j<Mo8OQfl;gJOgh{I2gsFQi4INWP7CnIbff^n_fIIW`c%
zh>Jf<Uq|}h2iFyeyWRXSs)+oY*1F9*VD|VC8@24a$wd~2U+V87ZQH%h);q#j!<a2v
zdi~?Wb50)943mejx0t?!W;SZZ6e-k`qRb0aZ>CmASzN8CiRjECXAsgUj=kXEjI@h=
z`6?wTtz@1@EuD5Yvbn`G@*uBb=nXzEJz+8S2#k?Zpf|N(e$+g}y!DxlX&$$c1>l^(
z(t-PLaBTh2L^K&w{9Y+-9{_KhUA37=%h%|A-nQ3nCdn~!+F5DUCTSx?f48*ynqo)H
zKh}qw^WT(-V=gS@V*)VDK+O=G_dd%SX3$s-HAs22Qz_fKaV>NWBDJYO30Dc3`0y$J
zN16NBjuBO~`y6ZPsb#7CsyMxs^l|zo9~7uV{X>o#^hoBW9wJ-)2U9&hA-jaYP@ysT
zeLz+v>4`yyCo@gX6?*iwfe4QK6E}nTHAlnhVmm{as6`?iwN(*~TMIciHQOzhWzJT_
z91qFBE`=D~84aFWJp>Jmi4Zn#e3VteAR20c#7y)I1szkEH9^)wBVueHsJ_E-Ms%L>
zq%d6g?iIOwWI_G*s|G#k{K+ka$LYL=qUX;~n5{vjY2{ivBbV%|={fH<(8gAapa<oJ
z>5Z$r>er(CucVHZ3jxq*`q!aH41F)M`d++*bG?Gwq8dp5da9C|*8sP1?H>+UrVIm6
zK9cw`0fd!_k8|RmGY7tYWr8S#<$R?hrA<(<#A41_rfQD4%w#v*@^j%&>Qwe|Kxyfh
zE=5*9M<?HEoT?&qa%GaupA4?#7P&LZetAKKEh5-q#Q(DIhus$Qn0pw1*!;5HNnrjH
z%GXs<tFoGok(p=5<vs5-fn}7Gb*UW@bPfHg(Snqi?5;aG?R}2iX2qkdR>cy!x&7P`
z$yFTTSBP>zq|B$Z=qg4nHYIJ`g>8~EODb7IcGFKLNepsC@Xa1hQNsCY*nVMazp^~p
zuALA~sy>(Wd>5%A7wOOu$f^3c)A_y*x9f?pOaF|>x90ylo&x~SQ*qfCkNm!i-LYO|
z=vJ4P@O)is4N*CoHB4N7*|mDDS+5FzgbrE@eG;viV-B0<Fb0Mxvs%^hC)aRae)*Qq
zf1JxsH*q+wAXdF#ipkfSxyki{M^R&4SB$wcRqBe0>Z{t2g(#;$YgO+C(vm`l#0t28
zxsw`y0xYClKFXmbOV#7&Sl6PT4SjG3AMScVgXE>1FVL6^(zfL0)pew)xw<ret$B4B
zI?9JYN=~iJG_keNot$H>VF;heWQaYZvv8-J|7Ea)4WFre^#!VAg-2&k;IEZwL__d~
z!f@*{odfEQfDW}dj&^X*L05DO!?1&G*krLecoMHH{=a7{#;t|aYLME-#Zbzr>vooE
z#KXn(b7jNmZdTb2SB_Wi*i#!T91S2T^1>0A!e-3a9%tKLnrT?K>984NtHwlcw{~1n
zKVMB3!C%B{NP5|WhUv`I7O%|1%T8r`mknCq(qFVV>{K#y-Lo0}I4VU(If{SYrZKv0
zTjKmkH1&3~c}dqGw_{>7%hH4lUw)bA{Th~~#OjOpf(C0&cwfT}Ot&?rF(=fk!fAiy
zx?|q+lR;0&D!B5@fIy=td;fmATrLKr=CTZF;rpu4d{`Pnlj6LVtw7YxKpAe4t%l&~
zcN*$-&*;8$CBSV0u|qWjnpqaz)h>itN3I#?qH@s`^#zZZRCcvf_t?Pu09*Sosyy~g
zDLhyJChONcp)U0v*!<zOnh15nS0MCa4P=wIilxKLS!E8ABc8OI@=ikStwqOZVfNDx
zZNB@TEQAIaYTV4KeXLKhi-%g(F^Ac)j%GdXd{L(dbwPGl&iGM#6LIdco9WT-LIT00
z-O|z)jg&WHSw-j4oKa|YMc<fl<yO)g-0GO0V@B<A`cq-nNNqK1Tzp%bj0h%i<3knx
zp7mPw97S#gSA}GNTncI79A_H+8F0~987V7q{ug{C@bgFT;F!V8m|Khg^$c@sXCoE(
zilK9FS4_WLaHAnE@TTcbKuUCWFV4_ZoZ(i-L~Tyrz;nh+VCp+E{G|5LjD0^I$6Evd
zNSEB~AxB?`O5B^*m^~ci8<2n_LYn1M^tBBo;|f7+*f)OqZ*a6f$=3v~Cww{*(kEQW
z?Y1l7x51THFRO&g-+lrCJ1KRN74=9Sk&N)l*nn!(Yi80LERFg6=R?VH9BNRX)ih%P
z?LBuBolj-(#j#BMy1>m7lrKE^bG~i<i*T|g9A5wu@<oxx2nSa(#OXpryo<o9rlV0T
ztW>!<fcZIY+h-A7JS;T}9J&uEt%j*a8H08e4BpT?;aif06^d$!*|zGD#?ocAw#lWW
zH^-Pdc5yZ-lw7KCt{+a<Y~h<ZE2yh;-(%VIM%Kh}ahQU#lRfyV^{bT1O+JHzy|_-k
zBVAe6@IJaKx#||@ozd4RRKLxK)<;{DjO$z@Put7r08uvV1CRx^kI@kt2_J?b&+nnD
zDK-{g!`N*b*1>t(B(e8LTXv#GLT4n3($<uE<6HOIr&u^f1TF0QdT|>e|CCgv@nj%W
zdxb4XZy%65(JIG2Xck#Wx)_%sv}I=J8#8Fx`8XGyX@hKQ2-xIHJ8!m}9;2msVjsXR
zb0y8;<$A&!%6m9I$PR6iMTcAM6uwP4uMkJ~TARn?R^zj08&v`Fb4WB(C3>DexpHS$
zSPpD$rx>(Z#9Z#fQ-SB7zizzRlihA=GpJBM{UrA>W*^+^3naoF=Vcwd=Q6$%iabCD
z!qp=Jl%&RIsc+(<on!PP*L;Xi<8M(S5uwbfV%sTd(D-<<HhLdmDCW9E8@A*vrlGkH
zsL27Hs>3;hM7Aj1!gc$viB-&QkhE}eWs71czxx7{#2#i|$XmBnQdMa7?#}5w;2Lry
zTK|Gci9tX^IIitG;-__cvw;+Av79)k!7fs$H~-q<pL?82$EBTNoZyACKh+$YElMys
zg6p5WVt9OUuV>JkwzmxJa6?!KKPDx-TK)XXVZpb4^!KV6?R!l|f5h3Z<D$^tSt;g|
zrj}Soy$htz^1mp)!yAB$PCrm{;vW8*0ffeek1Dap{wz!u?Qpt6htf*JB9iiIz_Oue
z<~0YP+BMMjgi;)ziR5P0qcgYg4gPTYFA*F8hb3O5#JCBRI3hc_uhGl<ay}n~u?A2~
zR6FZ+tZ*X|uYi`Ll|`*;QOltxq@Lp7;EAoZ3%tm`PPuU@h&lY?%e<+?08;yD`@rK}
zMZ*X0RPOYQl2fr&Fxs!<JO$u<=r11r`7W**L09vssr2=_8lsb)##`-IZiqFw(ySRo
zl|PQc_lJQ$G}VIpqwagR?N#udK6Cj~gM2mZzh=%#>ZCH4g=~T3F`41&eg}wgXGr8(
zb~T`agmNJob9YY|;qxvRf7*o4hfiO8qFZ16`M{<WjM6fFOyXuUQr5^%&&!YR@szV&
z@T{6$_W}i4zvue3g8`o&=4+|{C4xiEY$S1a>!8xKhx5|ScXy@k;?9o;Tx+)>H6p$&
zX2nvy`tBsH4~|HXfN>!FkMz|N@ImuM@T|h(=>@Jnjb4Al!|Zo)eG%IrDB~;;CwGUp
z<+qR2`aB)_9|L>~fh68yp<E(mFOxsBvp5FYt8bz$zYR>-fj@fCDcM;RIwAsETW6pY
zyKRBDn8Q6^x-bVe)~!;?Fg7e!Zh|JFvb&j#Zu@`;M&(`EZZ(OzZGN;Q+KYHx=X!>s
zSnP#w;pH=R>31Cx+_%K0cX@2e<g`eS;N!+0TG}LMIl?GgW@_!XF;GPO(>pwW1Vg`1
z!1xV@eorL$cFp6@FpU=nZeMD=`?e9^opomge}-D6eu<<x$bjL*4~(A};~PlJO}ZQB
zI`#pb6V6*mM^y~xT!T5tJeMWX{)A}C7E*afmS=BZl=YkiP|xQ>x1yN4`pnp8>lo`w
zY$$tr&5eCP-e{=gl3}1^>eft3ucV2G?6y*ZC;U-oT8*Re>2KiWGgXOM5ULdVvR;zO
z*zP(`=?6qUxXNYvKUNMCKU#0Zz6PD@mNL`AcR=`-!K3i0^bOxKDb@NJ{BQkAkoNi}
z48io#KDhqKiVChf!wM|d1F$@EY^(>a;)#DeF}uN02e%+wCCw=Hd5c(MuV`x?tdQRl
zf|W#dAD})X`DlC}fax>3hLwZkpC&jhgtpc3(f;iCZR-NBtg+6hX_mDT`^h~8d)ZYM
zUDfS#b1i1L1<e(TNc)p{m*=fs{c+xY)6#IY{=w|jLdr^M(dJpROQ$d4O6eNLE_<G~
zXWop<rx<mgB0DaHBOZ3UQqc0SnTdthm2=fkru-_NZKcVx?U@-_kmf#A*xHbE<`9lX
z*p2w9-G~PSHX2r;_n*2xpXZ#b<n^P%15Jd~IMSS_LFWB_H+lD=k<<-49Py7Q;ub;I
zA&AO8KqI<)roTzWDl>;Gc^CT5TkWmX3}mKXc99Xa?GgMaE=}-*-VgwowSM5~k*_jn
zjikuG5|hKkw$~kUWgk!%a(*vq76viejmENIB_1px*wj;)nx6Umao+y9C8>AeEpacV
zl|3+0|4hc=gNJd22R1zeC?1lt4f5xg@7U@Ry)?tOAS^TfGj1RA->}Sh9m3$9+wueX
zQh$|Kl*y%gCiqT-W`-HMV8fXMEV-!H!H<rTf9&}WW&S&K{zReO93+`ZWstVD*|Erk
zH$HE0C#%o%bol*)w;mHNJ(5pm69l;M01!4F;ZgWh`i6~1N7_eux}ry<2eb<^b4aU6
zKX`jR+^@qX+NmSBUAYEcqYNYZ@y(}3`Dz-)mCiw5FQr{QQ+>FrATYbEF(14ZD~&Vx
zAcle9&Ma>o_5s*=K311kol_q=CC<n{v^d|Z<dxqi(X(XH=oGrGcNymgll^kiRwx4I
z6|rWT3(DOhsV{h<^}<!?_E&TA`HH+-<5H_DuI*cT{hr~u^e5<X)sK%aF-n}c9O~JO
zAz3VSI{b42j!012;XwEwiJ-LmACP<n;qdn<?Rp)YR>5-lyVbI$?!^eLu<(^od52)_
z2kXkIT92c!Zt~}h7b-E@U2{3cJG@amY630|pkbXV_uO8tz1~zX7M@%m7^tMAtio}`
zPq{fg<!IUT;rbmzCuG|#(d-xO+pkV>E<R}+l#`mNw~L*BYL!_nEbn!e7=EF`wn#b~
zMc>c0@{Ig9&3*!%hd&OW5PxH5Gr%2Y=CbBe4v$HWO?b!1R~A1#k!N%f^TLr_h^uuf
z-lWvPv|&wq+_-sogOf)S(OrZy4pV7dOGaR{wZ&egKw}wy1mhPw0RK7v(7N!j?vhfd
zHLE`5Md7Q3Of;Bz(Fe?JS+yIMhm(x0CD*n6)dC#pO>`XAi?kz)-y?@uoLnd>2HMta
z&Ad<Mdzkj@7P6%r&vrJx3|zk<q%nFbxh(ua=d)qq{?ons!8XsgZjz|0JREfLGxp{!
z(w;ugq^U^d)kYS%Q9W(@D&aHDDB8x#F6!{Ep1M3PE?z6PgBEgV(Olr`$u^O%XRy>;
z8CZ<UQMYcQyQ$5t)8KMkhWu`vwsYwq4^4+lgIa!w0va|mlSZAp1XYYZNdM5+=PM@X
zAF4voXMH<t->1*|b-ZtnNb!%h-)JT9p%wW)U|6ghIGlnrsm%eM&k#90c=}gX{@fhl
z<?=`H{JxjVuOm`LSnLJm?>&&+)6>o?&0OP0f_^k-IX9Z;4B7DzUMztu!aY}3KRl^_
zZG_K#qfPjS#>A)%#ntxv<~Y9*bA$Dqiqgp1-*)=v9TC3J{s5NW_l5TBfd4r^;s1Zk
zg*~cLr-73=Gm35DrNH*5)=R(g@+apATFyTN==W(kKMwhyv<3PwsMtQ>;exKPKLoVb
zw3C=ScFNP|BQt+;h9F#hJD>#N>I;|=gsU$A^vPm^aP>)8J`xGS)ki44K0^?$zK9h;
zxcVYQpQ=C*u0DmzZ)gPJ>NjwF)d4}c`T|S@;pz(jeX^J!TzwLjk3@oS^%07%&k%&G
zFJeUyuD%G-rz#MHt52cw8yZ2l`VAakbwCiVz5o+JxcUM>pDZQ_SD%FCBat9neT3rc
zGX&x4i&znat1m+IsR{()>QkuvhDH#segnr>9T0@8FTg|)uD$@!CyNQf)hA*3NF)eX
zAEEgA3_-a1B31<9>WdJ4sscf{`V=a^p%H|u-@x%z2L$2j3osFct1kfb$zp<V^+{Mh
z5(&c9M<~8NLlCaMh!sJ&`XWT1sz4B~K84C}Xn!Z+iil!I^>XhV>gU6^D#=V5+jAz0
z_0xO$owluGyl_E<(?p)fJ2_6L+D3;MO(i1DyROG+D~T$&74wfCRC}k~H&m~#gK^4N
zMcW;-w4CJ9cIn*;*#)myniM;>d%~l*G=i9nC7=@-3SFe-WJTDk%C^x=f`c6W0yq0v
zi*#VYCenP?{)p>Tmnx5>jrt8%OUiffn2W~8!A4tcCay<A9;e<(+|f{rSuiltuFI1_
zTX)+OBGp``gTiF$Ib+!G#Z1Sm$v)W@7`uZC)#;?x&vR*VR5}m4P<*XbwVj`eQdLHU
zm!FDpdy%&X&NCjG#&T{+cH{W^D`4W*d_U_I;vv+d^_oqe*YPP#(eq`K8|slOBAbRY
zvUNLQdk1!>DQAnF9=)gV-D4YMnIM|6?~R|G2mNfWe--jv;*}?)j%A2vAdI;fc_S3C
z!80PhxObQN|3pV(t+^hL;<L>V8Phh|#-1rqSH^}{MNh&7g?Y3yyGq)X4H0EVSDGxF
zgkr>+#8JW+%Skj0j#XXa{=tk}eXB0lzS%cEJKXb@sytK})hRbxz0>P~GPJ;OKD=L~
zw}Rf<>*&aK!k)U1?%-Y7tx7hYW&pppF<N+$?&0=4YnfZ8I+Kl#y-7GR5Yba=!-3TC
zj(o$F{K_K6CL<JPBNlzq#GCo$z5cm4KDgjOMJ+PIA}GuDCQ=6tD^{(lg^tqz`FVj(
zjc=>c1UR2BW(w7Bad<Ua`Q6CLKsf89xgO8|d4}AfZTsgUUy&XIQA7~W<fc}sk~<ul
zQhk10dvd~(G^Y36Ft=D-*^$Lq9xeH?V!by7;aIxI-eCrCjB*EB0V8Ma-_IsK#@U$W
zBCg!l(A2E3ncKy6t<KTHg7ac0)o4<IT)v)JN<Uw`Q8mUM66r{@6cXa4K-B-{H_wVc
zdhmUYyM7(@Xk8vi1_P2x+6VA(p$EFR88E1)xuFq@=%SWA@z{+4rH#%qclU?Cw)iFY
zjd5Q;G$R_^Ip;oIB7pN{zh*m5lX4m}_Dgqv_V3@y?<H~7I#WipqqT=;H&^B4Q56op
z{GX?p6KCe{CP%Pjb*Q7AX3zI8TRR`R=v^bsV%=SvTO%v?kk6!sNn!JrG>4JB!9nQ&
z^(og2aIrUi<nVcI9eyT7vkvUtW;%ubXwFH+)M$AbY;Wn=Ozt~ls?87@B<HyN6&Pt=
zq*WW#5NO=c1vf^3kC;!Nt@OEn`WL8=Y7CXWB^I?C8=*)*51OF5Fe^AO(=RS+85c03
z$bi>qgWD4e2bQJHng$<E8P$-|tQJZfZR1~4tFV=S7~J`si&B;)k6m3Z{?w;DQ2#Q@
zH~fB|d!}CpI>+9*OJ}`%Aath$Y3ZRqXKJ(eQ=5cRjm50w#+vP@o8+ka=Jd>oUmN_A
zJ1A>&Y8ZTF4<fxoiymi78N_&1w=@);oAK|rVtL3?SJt(o!a07iJNVM3gVaA7k$~RE
z%|hR|<01_AqMoV0@UwmTOLu?v@88PD;gOWlNKvbh6FS#MBltv=tMcL$BZ6<q1oIkl
zB(^`fvUz_r02NGS1dQe|l5Aou7!AM&^RYY3IrI^geL#U>@a17@T>Rq8hPYrq;&XO2
z6caV1pvA*G>K7u;X<lrH)-cxPSgUVFuqy=7Cq?s#$|=h-+8ef3pHB*6ycc-UzJ?M1
zw_d)M@Uh$P$$X#R*N-C~aIn^Ibs2njejm`#;DXHCt9Dr!<6;}zaY5wl142#qgd_F@
zZR6u#er)k;=9<fJ8G~cX^a@M4HR86Y*DkWpYNt5SJJq_Yd_#IC7HeC@S9ZJUekBF?
z%rC86_W?%-ifqXw_5r)Gn=<CLLYu!7{pr7dFHcN?5+b{%d#t*%MPWOlfP2-yp04mv
z!{9qw$xZQ`oo7B?Z+<QKj@<XTQ~Gt_|D5la4`VXz11ek*v6V`weSpcHvX%o`nbfb$
z{JA-r1LJV#^jr$uQ1VyeTvbWSas3nhryUbgT<+j9Xp1}d%C}fW@YLC{n{Mw`IyOvR
z)L7(ns2j8!QvBo_k`GhU_4H!o-IHN_ytwjX5Ht8C>760Q7RQmR+1l3@EITdxeXZ_0
z*-d9eyIvU>+s3&Bi7=_}T-n$MjE82g@b0DClXN))*Mxh~q3aZ?rALTiwpnG(9iRgD
z<S8j86|X02E=jwUojgAQW@6`TbMjc`nbX`)kf;tCdc~#g1MF7i-}#6_w@=MsWw!%h
zdJl;@g?k1~EhPKIZ?BxD#8w5k6`S|Dl9f(PsRzwiOSp7etG-jfMmUh)j)+{&3%9x3
z2(~zHtc`uDo5QV@bBAkDRiTpJF0EMC?0oV*;L1$&eUG7r?zAUmHsv%Zl)LV90h5CD
zvdfW--R-so;F}pk+t}AKwhIGbcN-FwQ=a|4oFMO&;&waHD9pqhrS|rE@<8l^j_N_-
zyMcp)rQ9oh<WQl_;)CjYyVI`FA!lA>Qx7Y0eL=odnD|c|<P@N{Aj-=3d8<WoYPYC}
z<8($Y-#Pj9<1k96h*xd(U_SB~(g<owr>sh9aRFW#a&dCpD<|OU2`}*10ZX1%R=q8z
zyf-`pehK)h%SW`{Wy2cFiIRO-#N=I{&05V3Cbbpn>ja;k+6RmV_f>+i4(|i@0VCB{
ze%uE%GAT#oOx)<;RHf_t`2}CKb9UpjyEZ}%rM0lO7js3_>TH#oB@7nHZF`l}oi8bb
z;i=)nTUF%rl_B&bv)_3y5|lGv1n_U5oFQ7-47k*@6c@M<og;`IG-&8L<)zlO>{g^&
zpM=!r%DsILqNb^-CVKv?To|h=B~IXx2`CM?N*;V?*#AeJ(!oWm=FN&pms*Sw91`vz
z05ML0q*0kIM#L=(U!_lOhBdrbtE6T4sa2HGe(l+gxmTy`;U6r$@(du<4NkB9<?{xx
z-UXclJYe}JGg!Mg8X2!PthS1ML05<DLX89>Kig_1Cc4!ybP3uV%V-bMx#KtM^E<G5
zXP%u*pH30py3+nMM{WDu1gB@M$!$FcY+H7_K2LJckt{IvmPmGDoDx5&sN71<Zg>n-
zomVb3l26rM`jL`<{YK7kw2eAepkG`K)@`325+A@bgFSmDRt<)LnOv#|%R|gAyawCk
zDx}7S(91VIFLLL+n10XqR7MeoRcAZbK+LX+^cFr=B<>G?7hN_ay)_yjd*o#eg$GXF
z9Q+_uuD@S@u=jhX@H=PrTQXsd=C^qK>(Sq0jfN(Bq`TyJynPFehjA^^)7Z!=LL$WE
zm106{2=jE#+tk<a0G>`oFzscfTw_w+<TZEh;7hVGui+K>3XXF^)8W1}|9V>gbgU`-
z@cDc0m@u&A+GHv+*fQt(PdW7k)1_@iHoO`G4l;}m)0_^g*EX|8X>?7LL#%0qTD(Oc
z_qwMrFxQYt=tYa(1>X6m?!G-s_%it-W`Bz>ljRJ(j$^tP+Dc_QihA!TrPR7{@=cVO
z+fSob>hi^8bSq*4$Riker{OT8n@N5zsV_S%znrQfXQF?7jC+8)`_%5hv{!LwQ)aOq
zLa^I1*=fh!d=+^rQUP6$E^Gx2#44+ZT1`jo19a?1A4wGaJ|-kcYowa#R7i=rdfLrc
zJ~bibX7#{^me-@SROiNFg@b<lKt^#MRJ;0kJtmo*dF75la362=(I|tq5xU3tz`LY&
zHMl?W5c=WWyZeBI-yiAAC4{;AMU4Iyb9V&6+Rd<58fm!-6cRrg;gE;ur$fEl2N>LM
zigs^R3n<+OJQ1{-6J_acoibdwcg*H<r*(Y9ytLYDQgMl90M2NiRy0<UVV_o$6e7;B
z+9i52x@X0HeQ~NWmTn0JVVNYsey*!8l@V6Ezl0FrZ?M`;RH<`Q;*1vF>2*EIOymaj
z+~>Ha%WY%vwdUE?9fQEt-?*RtT-`5~5!5zcLJ078P}^L#d@TDmNDnmDtAQoCa3Id-
zGh827i@Ew9h*NXGOCnY+dMK`artTMu2;%+cA^JOr_atQY=QqMETXteOb`!==tGxak
z%X;lMypQIhR$arrGqj1Od_GtAOJxM%<x2<w{uaW^FybuB4_x>-ZN~b!jIWNle2(qo
zXmg5kA@CkzNvRWlSNs!8fuC#r+cJV(@Ne<>A4h)&yI@Aaw@D*1nUYIZ@HM|@4;0`0
z4=#Tt5P1N9++c6R1#cUW+sd;dN&L0w+j9g%jW1yKcQDi-_Ahwmc_VLf4;bnF&f)2W
zp3m{jWBy9=AX<N`xK2_%!)Hb6b9KK|Mo`#)2_eAWKw)p!{X^(u+7euNvg}MW`Rzku
zpJVzERv=&~Rvs4%&F$Pw=WFIipR4?(GJ-|S=MnllSj3Q@98q3m?S^a$sclQwOQqcX
z9LuH9r#n>UpqYg;yIT3gt)Hv<r80un<4Xt;v>tyv_Sdi25{y-FSbRE)rW+SzLZOPy
z9Ags3x({%VU{!oD&6vh#tZO-bGm=B^r`P4&I)!#h6-UmNlhaEdJTd(6Ql&q=0sap#
z)#9uP8E4v_4hk3mtgA)uDG!!K6?hF%e348fUUGsPDGwPJ+qgNdZ?{JA7WsPHJv%Y8
z#|yleGZ$Yqx6JOrlI_7*26;8*w0nk0ST;EzfeQWWCH<=0U-3cZSUqnFOhH$Vrqssr
zUE368f$<wPM=;)o)<^T49k}OfRQS}8)M3>3;-x_Pp`ar#gGg~kg}jG2#XC(v&s-Ov
z+${Hr`_)Z1WxcoV4X^ur8(AeNP)$wEW=C)YmJRjNRp>?A&qdPkLzN)>ui7s9ag)be
z^&~|zpJ?!<|8Y98CfS$Xpz<Z|TKdSo#{v5F;o(1`PcRdV#bfcwfQ^R+gdQ)dVDx8L
z<h&E_JoxaYgft%pZ7YM`EA0b{eIRc>;`mjWzd(UR94G9xc)Jfg!>_uRtoK0fG>%as
zrTFub%yG_p7{#(WhZC2nakQ@wd;IZC2?0THP#iR01kWlgo?hVU)9Cd#Jj{L<M?A1_
z>e@ycHglEY6-xP{)Cc9C(j>9uZYpZy+&-ZHjnWXV);FZT;Ny^}Iks#g%|?Ee0S3CA
zDfQ|pp52rdYL-w%Ts%SR6n&$@BaFvzw6Bll`RfT2-sd<hKK(v-eUj*@y4JlF(59dI
z{Iru4uHw1n2avrE{T-@(!2HlT{Db}*(qFRVH3=N_R7IZM3htCl(@3~|1JAF@WFSRN
ziY_R>Ju_S9A^9O3>Cn<9Im;17*)mgWzm0(+;-B8(h(z&9-2B7*0kVy{^}Bal?QX5$
zTYpT1SbPAYlb*&~?N@GyHMr8O8AO#oj>4l;T-_?c;bFOZb(*I`{)Fgh9Ep;y>gs(l
z1lnJeX7kd#yiWf5{5qN9y6r|(C*vfvY__z)Fr{pww9Vsb2^vJ^4Oo`}?NDr7#jyCH
z$aKBZ=JW<(YquT^q|?zfPqS}cF<XVSpAdbmqapSxc5E@J3#1@5Q*N&_A8s2_9jdv@
zOAJ3$<ya(FrgT!=bmc8Mp&0=Z!Yuik8;3+?F7~m;iGI6&@(6SF?1dl~Q8?7CXdDD>
z5(TAd?ZOyP>=UAv-nSZsV;Ja4klu}*U9m4>#PW)KCxP7A;Z8qrrlwgW50g&+F#7PK
zH~wkxdx-;_TB5m;Xk04Y;^xY;NI8%ZUU|1!GqJ%i8;#^mr=~5IN$mDOyv#>qa|L>f
zbOmY324z@*XPM4kDRj7yHhwdL9rq<kdYNR@o$w$aJmd(^{r`KfLXsTo_CVVonr?Js
zsoSyAgA<F6W!G9x!k=cxJJc0s+85=38W`Z56QnIxd4mpX23dW}YQUle;}UHPm>IVV
zddqFaIoEMgMfE0bU#8s%tq=ZLMZG!Vw>aIckIkK(1OG_;Knf=adX`-cs34(Sh{oLA
z6Gr&F%f(kTjc5qoP#A7qrgK2u5zwI)#}hv$LP%Y}bqIrZZp#nkOZ`<|Q6`t}nc%_U
zUd&YrjK^)QzO^O4`omjt{EE;aP{9;Pq8qesX}S;4J#+9Qh+mlzgZQV^fTTf_BJjCw
z;RB<mKhy5OqnPsdc`(Uz9qj7!2@28ARQ+5L69aDV>Jt}?q+X21D!Tr75x(*<&qe8t
zmD$WSwDX$$(+6)7aI`Z0A1jB6AFVfHUxQ9{OPOimEArQkS9`MCO>G7h>ZhONKE~{W
zr~Z-%ZU_JXHl<*cmg!>>H=B{NMt*u;emppI_tEF7l&nFiUT3!+-y^}Bxg9|DIY~k+
zQQLXA>wySA@C$}t`^(f)sWOU6KrO%MT=R;3VE;k6@H>r(t^x0EXmzr5al33aIFiKu
zPPH$Q$%A|M0oBFUkW3)w7S{MsKfb{siTepV?#|uY)}q^IYo&&6;b{6iPlw+>c<V9Y
z(j)m~HbH<3Uy)&KX;gyhvXf_3UuG8>-60AoJoT4Eh`EeIf-<FAKZF0RUkTD)--IEU
zKH3LIJYHf=HIkB#>W~QEy_Q9cH)A2vpBH6m$s627<1R?I91)K6O^JMRIBjw8o&ILC
z#kNG#n}XWUwN7Mz?a5^kI!h<!WnAdj?DM(0zh6dj-3qtMQa8V)8ydOmlI=T>@8<yI
z$}Mkk(b?VnIIr-r({|@^G=c*DHz0nmQQoH%@R{pO6qD$MJuL2$%)au$Ux{x?B#!Sp
zHWsFMb|wYAASOdPMxSe}Om#<sHcxgnD6TBKuY6MKb9H~ej5tXMw1~i%&bHs&E7p{D
z#dkwc!T&y9|0wxBqnl5u;5D{G0+z8jM-?vlpw%`h{B`)2L|nLG;pw_Ux91pB){<k#
z&G)&+=O3rD9M?+Ps%jq8w7I*S@VUCbUq;j#E4k5aowGWf4AM<cDE`n5K?(o+fc>N7
zd$$sNLJ2P}3%t)Z1*Tu9yfe%EdIlP?rx>N})$Dz7+96VD7$;gRAol^s_fD>>m=-%h
zd{MNtE#+mcYbfBxbreRgp{;$J-W|**YgAewma@eu=S4vr&6cg#$JD+{^QtW1L1S1i
zj}7_sUSOO4)i)D2=D<i2r+t8oInT=>$1^VS7{4M=VlMdIdy#Olv)J-}$yP1-sr!s|
zOyO%o^N^{GvnyA~jY8@?>|B*r)Wf%!NeCkm(EHywBUoaZb|Sstb~vv_VDZ7}y>TRQ
zJim94e3(o2xX!3ed$fqnf-s+2ZS5)_BI079h5s~Zkvn6HnkE}M@AkcFM9g(6-atkE
z(~quv_IuZ<iiC~d#69_{Y#I`XwsF8Xb!gr6T(?JRV`DUE4juBgi8y>WJuxvb=>jk{
zF#LJqFrSPN|AYc0INHvT@ABw$;#&!Q!aYd12md$Tga3!m$#2GX=V{zatWAw;A40uf
zVvL{N_}wXEm+U)yt9uwI-<)ehb8LHKQx^s}VqDCrqET*DpoTWStsx_$%nCbcp2M3G
z%WZR_#>er4%7Y-v{S~4d!JYy@#Z$M2=?b=i2UWMOP}#+Rj61nlTEZ-=8BnYy^s_@b
zgI<l+V9n)7zhku-&iPw&l7~$^i%m*Ywh!sVKT#`Fr1A<^<Uf1;q_LxW!6jEoC&L{*
zFWhpNt;S*U$@ySVP)k|VgQl6G^1j{H&_%cP_7ftdI`m?YBHle7iLHAi!f-zl$Mx%y
zRg;4X(quXb-efzxYZ(=6uA-~R()j7;&u0%`_4~=bo?(GUsi}3~;++YvMzby;%24)D
z3t!)P`<-*o`N?KlBzvdNIbE=!$^y?hz6J{c8XNKR=|5#6!fjs1I6R$=-H;z4rd<%)
z4PWhM;taJYFrODGv)<AuwBj!j3#}0{u-!T8mb=oUw_9tBvfj4DFj16K4^+<SG7Qwx
zaZRk+-JSQIs@W*no_@qQfFNaR>8IZZ&_TR-iDNs<{R+2}SVoo?pr^kUXzv!q%e3#G
zwTb^!>z)bSW@^((PK_~qu~hbQ>|&>NT630;L$gzzYdvFKWX_?3f|>~Ea}EsG66UIX
zas@QrOM-tod014T*Gqu%VD=a9&Och=&;Ph5*w^9g#KNb$yLwtCZS-rm?M`>U$6cKh
z$Lk-@(%3?_#q&1x%>aBK)UE%iRWdSw2q|Q~AYmV1o{idaH9EfYN9W7ewfo*(o}VzG
zqS}1ZvJbG}**m%qShu6v_&P>D*7{2f*r5w$9NOW#*DW9&@OKx^;F-CwTtzb4jzD3|
zyL-8RB01oL(_e3T22eaCXB*_tFW<4%C3<Owuh?=9%<w+CGHHT)-21gz!FhKa?dZ-}
zk@IRk|0V{!B+++75Pua99#~3ruhO2}Yq9p$Km$m9r0~QsG%68`N>JVhfS+S+Eq8;l
zgSA?+bsJb++!7;m(~j-A9g9J6@t<&Y|1~|rR(kSHLfWS!eK9((oVK@Z3wk$QukGH#
zWca=J+Z09*1{!-N$1M7@aNhs`h^5G{p4&#8mqc4-?#1;U*#Fn={^aBDbwBW__1A-Y
zcKrJQbkIJ)7qr`VaPJ#g{wrKnFisaQ1U?@_DaWL&a;zMa!?XNd%dZDOBHVFa*1>x&
z<2#|q17sjvheUvq)c7p*&3!-)Mn7`RhxqgfJXJ&}bE?>OiW)RNUaXB<$u<;oU7`(J
z@)pz3+y~U;fKJuz15P4aly3RpYX3`O6|);8Eu380q8Q5WzQ820hnW}h)@_wk6`H-f
zb9xWBh8&63zhF{g5YP~gYx<7(Y2DszAjMiN=RP3CMGE!iU;6u#55Lnh&ZidHBwvu?
z9NZ`F1BM{`09PVOpKoLNFW68>B5Jcel3d8sVUo6cv-Ci?6prKXTVsPX%ctmT8%o9%
zg4nQc{Pf@8dX(bfIG3(g2AW&z9`{NB)PB$c{ig&*0l*j^sA$A*7h1m$nC^edgMUN+
z8xujw`Zw%iAV^sc96F1-v`BEysFULri(!ZF`rzj995=H*3--zSuyrgp<&2->T^#MJ
z4b=%!7D39w&6j_Rl(qI=stApRJX;v%j68(G*_`|^P>G$XcULl5ldNvZ_MppY1kc6S
z%!O}xM<lg5y?5#-k0pPkWLuP;wG>jhQiWOoc)%K1&d*6IXz#rzA&?1}{MSy15%Mz(
zzLlEUaW2qTrDzVAN$*nG4{6f~Rh7xcYQ}xQv(ccV4#|pvM)!vgUaq1&Q5kaiD*5Hn
z1%r^&si$!_=<lYI>Tna!Jf53Q)jExphQ3B8J;H}Vj?8zQs5xjiUTnpCm6})8Aei?m
z&Xp_m<zZZ`72IEdY$C-^>uS#>lzY6Rkh}O@=Z51<0n<Ss{AP3lnScplLi~ZdfvDwF
zVPmvQlI5w8FxFjTh=fbPOmG^XQU7_9rdU1``I5?mg$_dyYFyBwJHw=xu9EZ0&|b#*
zgBT3~B;Uo~;NQ0>MG%kweDRoOZ}jEPwk~excm`zG=_y_v@=a;p`y_;E_Yf|y+C*l~
zT;K6vpU*Tu#%f7Y(UrQF<%sOkj=D1SnW~>Fax!Vz6U12Ux|{2FTb}3s&<#Ng`8mYC
zT=wsfj|sa#d~hEg1T`dX_Wc{wkOW(Q9IgaC<e>)jSxqw*(B5-5(fL#cUmPgJuM6Bf
zLHWXiKj+)#zX&I5!rg{n?@fhZ%8f&dAZigrE!;E!5JW9J3V&`Qh*}@;{PQ>55=1RL
zu>S^8s~>Z2STn1P9TeQsAa=uYRb%nOJpZCCmPxJyE51X#QM;}&U)c+yV?c3u)9qZC
zKhEg{@%KVuDau>DDb_NPmZh(m`kS<d?9l<<i(!3}8Sh2lI^<%v76$mlUaXdp2wFa<
zv0&P+mUWCrzM06dyNtbxx?el**tNmfi6CK0bzNX!YIYwaj>&s7nn%7n%2wE-J>;lf
z!07UFu*Aq!WY&}1qMJCRLu@@bWK!Jr0bwWtfd7l91@nc1G9~cQ-8i!{x4!s+{C?NE
zlQqd5E)5}aO@gAsnIZ@oC}Pt)&shFNZ!}0~2Dr@K_c}F8AMxzId?6Y(UKf~nGWt>j
zER5CZ^`oH=593e&BNC2Qil8XhmmJs0i(gjn-~nH3jdR$I6`GI_J{8dSgp_7|r_Kgi
zGQeq&n9vzxaq@-*LNJeNI5{aLOkRof{>_vK4uuoaN4)(AcP`=4M|hzBFMUlhd%&;6
zdXsMvC8HlB+|}|hT{<28Om!j<6{qb>!0!IafpJZKsBDh9$(20kXp^#;$od9nU*AQW
zBv@d-3$k0=^2f+0;?`$nuGEDm9Wll~i@wWe&%r$hUC}KJ!w$A#lf~xXNxb>v0U|eL
z@p~D|`vB8?6ekvNSn4#M>R_$k>N5E5JZ{flg9|cmui9l{jEik-#|4qI4+u5c6OPyu
zw2hB{iL3js>6*)M8G~cX^a@M4HR86Y*DkWpYNt5SJJq_Yd_#IC7HeC@S9ZJUekBDM
zXD3UPK=Nb6ZZ|axR!CzXpg8vc<nyl`{;9{`8+YVWo>@Sb8TJ8KnSDUH<lg)NyglW&
zq`x9Y4Y^bQdX>C`erl7-R;2r(Gd?(mzsoTTk>m@oVp)DQNX%hR7W3dR3g4rTo~qIv
zxaU5g7n@nXEk*S72#yH&&pDzsBdRsp2RK)eYy%_r0qw=7^@s2b{QJ_s!}$im`?b>K
z+4jtgEJ$-7Dr{{?I&%nj0Z-FKQgv|*O0IcWE|t9l7#+Z0)N_@*#++r+sDUN!yAd?v
z>+HJoBI?Vmm)RnzW1;%#apsxe;KH~gL)FJ%a<7L;?-<b1gT!+67%lH;hXxc=*_k6*
z=OKu>{w(-V9~5Q@Edk13t7xi0aHM3rciAd88)lo9bkQQ8xE|N+e7*3t-GfS{rMg(*
zZH@YPFg$-YL<PeLgL*8asGHx#v-8_@B&1KclG|-p!f%5suU=LOmB0N20(MgBBrEEX
zJR%w4m9YWUsMpM-H&`0;0RaBH>{f{t5G;Lc!W3zkA9|lB^(+qrEB89CbvaRpXkFm$
zZd_!ambYp|Yey~!2&^-xx`TA(IEsLRMrqCt({PUIAqc5mU0NK^5tD07t{p0UlSaTJ
zyF(om>Iu%&>mt2WKS|yfB<@UQk`Zrqx%Mt}fJ|!Df<#%#F)mDiZQbvdxYmQ;G>V94
z^tjN0jo%ZIq~8o;)!zruLxNh~42;+?_!gt%$GrNyDR<6oxN;dU+*<=Vbx5?~KE*{r
z2JkhI<x05Mi!<N2{`!}=dtQoju0`tJYSq<hx1<4<1+GGzMn%T9fyQWKg;?FQ4ZP!9
z+dvwbYHKDgXC2rDK^lh%H3euA4015aFi`DE&wGCzX+7@5_(?HVGueg13GR8ssVRlc
zwK0_2<*CdWQZpCk9kbaVdcU_If!`7h^msA%_$5op$p<qX<iFcUIzHu8DQzRhZRF=7
zCZ=ZLQ)JzIMl(-N`;Xf-sL>sRpS>>Jf1s_TNe+8qUEVse%;bkuTY)wU@%VnFU_pn;
za-LqJN^Z|LfwypX&R4qx`-V3UiiEj8yx&P*RRHY5aJw$Jl(P@|!sw`GNKv5rrPdzj
z9G<Zt<FiY?*+=a>XqZL1xwXysoy(l+;+xz`Z?y22hOm?`HpaE;cjS#uFmDgj${t>Q
zu>rqtmtEXBeSdv+R=>lAHYHj(cPExgb#<Zo0nFHFs!%{hNrmLV#9&f}ggG?kQch^`
zPxBE4Q++*MZsx_l7lH$)LJbTxvTWh*fr)I95>_?H>uOZ|&Dg2Nc)bb}%Xzd#`G*Gy
zd0RiCV*OAyhSpbqHHS;aB&qLkw2B3{?%HLJOVyRyOxKikmJS=PHKtLvntB$RWi)75
zRayk<oxr?qeD>@PFiu5I`(Z3uc}SS2&}iAqx7K0?Gr5y`!})W?`GOL8>Jx&D?$+1t
z*#LP8J9ULu$W%nD;Xgyit5ifJob_sFS}oSDbgR#Mff3f6qFlTB%vP-#(<_rb<1`>*
zozCV7vOP~|aHEIXXg|uru*WSiP^3v!eWY{qwiKTG%{91M{iY*Q^J4y1L&nIWzN}>~
zIwN-(ZbQczS)HhRmaNP-<Qf$R$E?LniX1c<8pdrjy*we!p-L%BeQES~baYV4h|%Qa
zSWjhnc{M6JIy%xNh`y(wARks578&L}u8PjOR~5spTk0~d+G^7;sHR%Q%eRfR8J~SV
zhfXX;+H?bp)v@h@-45f<G6Kpn0&hFjoI81e0&f*MRo`AyXvkR4$=bE<u}LxR=kCI&
z2FmA~wWmj?3I)aGo953&wk;oui7_$EY3;jPiCz`lJSQyeQMdlcxN5I0v7KX9iBuh)
zhk!yG?WjVxNnR<ocdWZqG!}YL-00hy$27gIPbn}R?SxM1rpPy=o~W+N-`Uy=Qio&=
z1C{5jCo}?F4Ec|9cfug9R=fJfGeQFWdVMM$Y{XQeofyZi93A~Ug~2?T+K>#j;c=Nc
z?UIyyxUx&sysLGm24iMmBpzXsx!hSL$B^{NzBs6;$G!B%-bu*`Z3*Ri^<ziu0%2Hh
zl*SFDw%Pb7^QHO-()#H49)hYH<85s>DvZ@sdsUa9js7x)>)Zi3LKB6>!rU(n<uDlA
zo6TK$aq{dY7l)_OF5->p#j8t?5k+ZvHi$`!e!+oK2lG6OAh0DgI_OP#eaEuBz}zXp
zF%wgx-m2;>S?)Z{Iy}f_D#R<S8I}~M!VBbWhCwepjvAil@$gJUr%;l;!H$H$;>%2)
z)wS;A-7jo2;5#W_ZC?44K=tqoTc@Xr{yM#?;Ph@?x1zSXSW^Bowdr6E$#LlrZi&mo
zZ_`6AExJC^Q9zkY2#VD>S!f~1&(5b0RJ5@#MCU3T?^G64)}13`VId)3t!Fncbf{8m
z?^P9&&vOzL>gsFF<=}vhO&8Ge&iAFfP;os6@2)RZP_k0-U@u0RLp{u)1&<flG%Y%D
z(Ehh=d=r`(X6S+qXAZFBqFx6-I!cb8AK#e%g0Lz7pTqBqyZhD4{x(3khgKVQVy?)~
zR;yTAG>K!<i@WYbyRO-#$V1B)Oyr&YyS2fN-a2Wq2klh!(&<DTEXE8NdE0a@lp;-w
z<lwnFHq_2^vVK=|`};^og!)GG>T46bvJ#9sh3)x7SIxJY!4cPbI#z(UJ84ve?F290
zoXv8C3k*~j_c_{gYc^F?qb#Hg*CyhM1r+)zR1+<#8xuv5=f~-?9L~p=2E%C-6ja=+
zY6GVyn`57?X6OWF=pbTpEC-tFS7YyqPDicC-$`G787(8>WplH+BUkMQzWijPjTFN}
zIl{&kg*`IdPJxe=r8%(H6`nJ5rLO*nI>%{?Y1V4)%*#h>o&>eA1QhA*^)lp7ya;kx
zT-`mex~(DT1<$A|cu`8Hkz;cY?#k6kqY(fW!j8_|S22~Oi;Hcgp2fIYb7`Acik!R-
zw@8iQfR!lJo56VVNcBMF<A!RgOEDH9$_{;Mf<47uc@Cm0EfO+%dEB@Qd!VW0@W$1C
zSK_$rfhA#@=ayVsWKVfiT{GD0HHt1g>#>w@&QX(AyXp;**WhNf6lmbks0Axan99){
zwnx>1js1;{V|t^{&j?DcUMsQ~evgsvynm{aNxfQ$QkebF4%v9dT_Y727jC~e6S(F{
z=i+>esH|eaV9Q7hQdWykKwM~Cb5+Guj7PrI_Q(4?6|+SXdJ?nUS+hwp<)uI@KkX`&
zQhcuqEHJfHpp&^pu^JH-12s$oyGP|v>B8NpoAL@C%-I6(Kd^4#Gt&b_Z_v&@+;9|%
zF)ebx!!eKZ7q32Zv))7EZkQ7F?Ol`=ggqP48hx>+IqYY2r5z$AOhwwkf6y_Z{oM8D
zr(*CVTWW!_AFYc#`vZg~<@&W<u)R4!g+g6wvviKR7_fO@3JZ6C1FfS|$rN-lYL(H-
z1IcA{uUV1(I&k4hT*`tM3ETDN_;FWAjKB`sJP>hrQX2}h)`vz%%|C~7!n$M-_Wf44
z*<z76=~#H(#3K@;mx<-PljGt)>#Yi}FC4s}p%6h!S*}b2;niY`6VinX8Oy)w<ue}S
z%h*zgn!(ap-2bt{h`CN)g*|*+PWvjb6M1YSE&F~;){5R~E3$jdSqYYRR=SzAGOAs`
zv9wcgp6gxq=L>bUyt7ht5Emzg`(P$9MacAS58Bn>*rLt8Qy?|WNrZZ6x}n;W=pdc9
z%a3C<unH{>b=8Hu3VPM0lwEl)QP>ctPU=2r(S`RvSIJ4?CXM!b#BK%iEX^o~?Wn`1
zNp`24SseFjeufB28JG+&=pN-^G+Rx&qzxM{7m~>iVTe*}7rpuT4zNJOfJgkMugoD@
zHAJzw@JyIfXJs&#VG`t+vn2yK#|%y!7@MS6o4EQy=RSPHItXg434_@f55~C1B6eXD
z!B*a712D_l+OCTR<L3_E*LScPwbd;T)h*&u(8X-fO~qIi>T|2kbYr_2NBSwnttDcT
z;_5d|YB<hUBj#4@RffwW3Z7u+O>Rc3Tz$R@QA#Q@(LK?U=P<HXn!>BrRu8tX@AVhq
zgLcY9@TV7X96Pn1XWW#ER9Du_VG<PdoIN8{zw5jrIW2>5(jU=1xhj+`s2fyh{<aOd
z*Po;6s?*QiG&57xxq`Mt1Sun?1odlc+lw<)wd%7#VhY8s%mbd|qXQvl<}L!aUp6y#
zUYE%5>*o{jL?^oEXLbU49Z&AGQ)xu=vj(0o?R*UDchahb#!R$0W(s08+pJ~ZmC7{5
z0O4*wt%eC6Kckz{n<J!&x;+Ziy{f^+9CI{#WU7j_2&1dS2-^mteVu`bMa#70LiEhc
z6vs5o#92JYHNJa<wx<;t-_Z5il^osuK4a2cD<9&{JvydY*sDfUW8CPO?(wWUH2UW-
zn=U<CU3gEli`7_fp~Z^>9V?ByrzFoQbn@C7m>TrQYSOQ#&2-BsW2jh~5dBV#8Wvb8
zwc`94=y=ggBFq9JriNE3k!<<XCkE1=t)y0A4LV}r&xkv<q##B+T#&ZPq>#&_A$VJB
zgVq{me|pf+W^~{^3)sC2kD{?rf%l?99Mx2=B(k!KG&U|hq6-|Ns$ic?$fYfvo7dn{
z({cv$Y0iwB<-Zs_Kbl-<?yXwZs*s3ic6;s*)u9m(Sr$TDSeS7=fBfzyBr-e0OONg{
zZ`t4^W~p0pP}LC%Jx7`Cc=>2Wsd}?Xb=$I)M%nsxh<9vHF;cUAp&i8XQ(HEdnozdT
z{F7LVf$MtKG#a)VUA%M6XO&Z?yOx|4Duz;y&zt3H_hbB+;aYjyYS+UaxW{KSiJe@f
z0=l_Tt1(&dJ7lN{Xy@_L7#eEmJB47q&1w>BE|6rJqe8pLrM4;C4<|-5K0J*B>Ay}H
zcDP@MO|(--aJzC1yha&D^y3@cS$u06_gXM^4<62aQhsOpLJ}NS1LQw65+9=~ex<iP
zyD^IYV!>JIfF-D6wb*s~dhlWT8a|T~<%cO{gD>Bs&iu?^BrSOu&P)w4=a_CFU-qO+
zW>UU(+L_^YMHnJ86)Z0Y<Y3uuKi2^SbC60R&cA*`*O`SL>zT{<I4gV=z2>1?*3I@Z
z;3|ayvQ6$)<h#nP*VAVzaX+Xi7PHTc`o*H}3mp8_PA3kpZOOk^gX+WQqE{Wz6m<P-
zjVlzMT5@)+Snv8v`sq3MqP4;NOhp_X-p;imehioE!=kgV-;D;U@wK={1jrxUR572u
zxbA+(Fke5zRa<9zMa5<xaByS#t#{~l3!TUNQ7yP)OX7trn7JFT<2mU*cPowV%$waS
zXq(;?v15gkYDH6phQh~Is@3Db0bm)rY3RCQd#i8=*zso9;=SGI=uGC_LTGRa=T71E
z>r9)_v^*EF&TCcbi1r<(ZC7icD)1x{^M>TUhelC<j!<xqO>uJT9p1F`q()H>lW57?
z4p?elLq=-1N`=%(^?{j`_WJ29j{$i<Ud_oH&}h3tX)r5<yO$@yJCsE=z?V_Kva-#!
zD=DCYYm?WaYFF4{QpMI{#*K<ev-IVB@yeYyIW}0YJHccw8T>*)VDsuKxN4BIv08vf
zi?24^y00+oRc~>@N>m^mx<?irca=G6O;;g9;xS^?QNb}D`<M<36lii~W3~vgKbV)R
z<}y<c_o@ho4(eZ(a6W5IX7o5WFW1?r{a%E!yc$q~!SR*qjaX6#)4JG$RA&|ZFSGg5
zzpfz_Vz{>(982q9Z>v>TQ;U?DWuWC&cB$GHHV?%1=eTrq+Kj`WDQMV?3#tK~8Wq%d
zc{%xYX)p-5>jhZgG|UVUI2~h|2s4dT?NoRR6jb5m<yF1*c1`JAjfaGk`&!TxInstx
zsJ6C<CfRLPEyt+OSVpKdoNv5(lW$N-^onr<qNMZojp(cQ@+xU5J;c$=8q!uN{)~nY
z4elfNf{vZ@T4cNXzTtYN-Qw~4)>ufAGEEgj(Oz0mN_&LEqd-_qd5-tE$jK0$8jZr?
zq2AhtLVjvYj;;zNKk(eO&eges%GTVzCV6%@5OKi0@#)0qcwLV?6)Krndu)HH8tpVT
z2KGo>+kb(5Ac9L{JkDrv*EoOE=_N$oEf)*t>Q>R^n+r_4JS-Zom@$FA5Rsb0$Y-*`
z3a=}??$mHdL3QbLkxB6`0yb(8st|WU0;Y@AU{UYsKQ(7;z7b8kaCc%+L#?Lxaav!V
z4v>zs*IjVS43*sE<QOD6ktZV>nOdlOCq^O;Y)Ypb{G#W|ffVKvFAa}w8{C&^Y<~Wp
zxKm#i(7xXMi4w<SSnwmLH;RhZq)J7>RRO*RE4*iovO#Lkt!<qGLj6O!E%I91wKMWD
z&H{m)UPZV_LB}jaPo5MtJFC*B+}d6bW{rutmgaEg2KZpwEnM)~N`LTDcDAE~1HYpK
z&-jG#U~wcE9O*J~T~S?8rXDd^TZ?jm%@3g5vCe(PF5}KRg$^({0tSPbIJoo|aVg9E
zKla`OtgR$l7p2E;+ihc<$jLUeF*)aGH>qt97z>FECP#^!Q?~<{+!zaxL@;0k2oMG#
z2_b?3laoP+B!bCNU~<M!?m1`P%y)avJ9oP0p1JRv`aru>d+*Y!RkdrcTD8`{ij9q+
zlxiUYbiCR%H_p#UAE?c2q^~X82asYW1Ej9$Yia3<?{x>Mk4e-`@FD}Iub{jf#31hs
zQT1lGiwemnMxBrP`W8;qiN;gsv8_(T0kUx$redrdCVXklq`IxRVm!Yi#V9`3qV_V1
z25l2hQoFOC*;njOslqemLDw>~$4a4qJaji5pT@^-6u#!SU}sKps|h=HK@-V4VJfM<
zpV64PLSo=0ewix*2@|1Q_a`~SY~6O<P$yY9Q{_YraQrG4EO4IEQG^!3`q=U9s*Sp1
z1NelO$pLXDkiN(;?#qEVED5)~^`zZI?Zaa8O)Ti6Tk5M35e(s?h(nECtl(BlW?!ps
zx0w$5m6rN25&R=8qQkhtE=nmF6XrGW5VffC(@o<mKV^6}zHb=7tr~E6l1}tERCI$E
zW~OEw9EYx5Gx(WUMvlu{0ef0zEyUT)g#_RNW&v_8yGv+Fpw~>0S70E`f`A!mB9^(z
zZlNdgQDzvLp9PK9^*$HsFin@-dghGS%3Wg;%|M2VD9+p{LB5s?Q?VC5#j)E(iSjk7
z@;8d!ZlOQx+)|CIUjyVeaZ9!`dM(<oe_@dK`Os|KY}`%PiAc#(Zs?O`yH%&{>@)QB
z13t7W&Z*8|>)iyDd}9(Vi61_+id$w+?J_L#)I)&Yq@u22N1mF+!BC3(hZ}F(G4NDT
zENZwUh>p?Zs~|Iv?sAo`1HFaQg{;d^`<l_hjab3kyAcRMua?mbk$6>!Pp2shLpKz9
zCPs$ozaF%`JhE+6xXjbLve@qzdMy$>C}DI>MqajtyL023%-Dg}W|m8+yY(3SMqG}o
zQg!BVdh7ycS2t4ih@6~aW5UG=Ul_60?)2N5<38w^9+NpxhI45`M{FzzrC#rrdkRKT
zKx+?9G1oNlg|{Q}3+MIJU*#!+2iu;iJ*cyAvK3c^yzB3qMRP9sPQgVK$y3^1DQ}rz
z*~Tkg8)vS7hq=$S3<Nm!ccIpp^6IvPGXQfKJJ28@WH2VlZkCLzJVq<`DFH)HYm<{e
zH?%`|de#$sg@eN1$E)$Ksh5*5rUN@%z5ptxcHtVr=}u2aQL%9K^j-BG;#|ihIOCy6
z%mRnsj!2xACZl8%{qJjJ=&Mak^vKXxr=oZmZ$~q#tx7gs31yVrd)GA4G~uS>^w%l4
ztE<<etCMeXXP;m1Wx*vtwKr$`_Uq9YJzz4<Q_RR9zb%9c6Ldj2k(ghWBb6_)KFKTG
z<qTVRHC=*au^c0Mqw@>jX3AqCENdN?J70ub8G!ukmjXYW#&eFZkeqf(8?bT~+g4*1
z=gr#M3Umg<wTz&px`0|YcT$dfx!p^wPEnlO4r##1X`<j+(Qr#E)@{<Wj@w0LixuHp
zl%WRXjVZWf(n8Rh9f?_xaynu0DUi5PDq}oM1j!l6fK2y1OsxDt+T+V!z7uO5(7?__
zXR=CG;nJN7CGz4<!9%e<M#}5Rvzp}A%K>iq{vVO)6^tX-tA7;dU53042nLmF8@km^
zIM@o~Bkv!dUut;It2*tvKWeb@VN``IE=Y*9^W@He+%%6LStctU>R27Ub`4F8gY6Ws
z^~4t+;D%ok;RR2|5UUnV3Dm~YsY<9dF{QwAz^~I+4p{8^+9V}z>;%2(k_)eYA;q0r
zFH6lGFj2Dz#Iy>dMBowlPYlbo6|2$yp&#lR?5X&e1uIS1Wd~*x(LO!sG0QI{H|$!N
zkc`dQo<`J7SgOPI+?EJt5Ft45B3x3JCag@%{4Qu@Le%@{+1NEQdO$I=-j4@sIgCCP
zPYx9JRp8Cn(zdKwrB*S|t|$#qtfV}<StZdeI4k45&|OzJTG*-<yC<+*%jze_FQLp=
zl{*iRpPyiffhTPr5Kt3hMK7}|fmXwq!kA^CaqFFdAluL9`ytsA((_eN`E{ouwMKKV
z0;1!Qwi+W)Hce0Bkzb0VrOye^AOh5+>_oC0=1Zg^B?baIRE&=u2hDtHFHb*BJpTBF
zK~Dp%R#P?d3{A&?ez9*hrZ$!3G1feQh0X#Ew#N76{0qz2fc4P<HIPfXow;@yrCbK}
zX_3-nAbSmPTSHrYfiyKb>;_pLN9Am(_36Q4ESAU0-4IGk2*^ZxT&f^Z-y|&s8re5z
zojcB201KeMvvZN1*X3T-m3S-ts1^!1j(F#DemF7xPgj5Xzt3cZ%ik-01>`lfer42V
zSV!w@^#7+A|G!>vX5W|$1e1-8DD5!d0~0=o96)yeHe4VV5muFNkfLlEWY5?RbZliN
z{-PZv*?iWtiIsgS=^k_1xb-t@`Ed+!K@aMKnt|fS@`!Vn;@j+(B$R~Lr6FKTq8|LF
z?sr(pye|x&k5aGdN^^#Gf7_BYBA^XmHjP&KUrG5($t!hwh?gpN^?9E)5A@rm3WL~z
zkIu(UX7XUEHe^J8rnG316YOUFIToDBvWJA)_if%!MFOmeQ;Cw@Ejrg;;O_mUrhg!F
zPQ$X0w=VH8Qjc2p3jFL(XXthO_oJ8nrX|!RqMnjDR?+L;Bn%ae+I7><e+x%B>s<jP
zRc^{p%RW;cQ&x;LR(#}mm|-7xI?E$#y{DP?ll?pTFJFOaYJ+@A^9YaH6H)j&bXII)
zmNgGY$}+m(K3|qTd!(hWQ`gfAUZ+*LO$?s!*P`Bz@ddEWG2pwxI0;9pSfX&e<f5&*
z2mp_G#xZRBu_@>InCnJT4n0HmQ)W21$}n#Q8Hpk*+@!&)7YG2}fliol%^m$NGbG=z
zn@m_ykX|F0-Dp9H+3x@|nRxw8l=Zbo6Rx?97j8vgsJ^Nh#b#scJrrLfzGnRSDCTI=
z&29wkUZCGkV9Iyb#uTC5;q^I|^}f383&s=^6aI6n{%CZtFB;>wTUNQ-Q$LvLe7_%e
z38pJlpZvnEq%hIQ@LgxDP<vhlpW2I2=hs@HkCKko@9w=2TlU1ZQNueWx?-Jf2%P?D
zFdPaK_WVIqs^yBir9abx9-pPDZU4;W?-9^cZto?jrxJ6vuapBR{7W%E>-rL#MXesD
zqg?ZPdE(x=V-mePgx9J_&ewN3in1MVgv?F&?u@&v8nIRJH(6i@3+@r^f_Dn4mPc!}
zPMJ!(R$h~DK26U*T7B{U=tWlMm@9M;@^((8yD^QMuTk^}xual2BTqNwBmqEa*olBH
z4yQcE5KCq0d9(7AL>_Lnq#R}|PET*Ylciu{d715kceYtWOYMytT!Gr7NE_iXeSIHn
z5rlRgK!F$vWjh=AncV0c3D6ZB6Tk9#=1<koQ7ZdTaY60;SgEFZZ2g95cV~T)$qoQx
z^YixyR{)hO4_Bot@aYcRX$2g1S)%;OY0|vl^0`aQ>7So+1nCO*r7c1R54@+uFWlMw
zOrBzUa8`0ANslfW#Bs&>t%Xk(4i&F;x0XQ@%xC?fZpb*30l8Y8>JYLGxMp0B|E>T!
zgaR@V#&r3k5k&p2uAzPNh;v9MsqIx!ea4_H@1z$A&a7RAf{EU4&8&+tbJuuY@8#aa
z{h0g{0|S%RAnln4!Fx*ty;JNFqLk&SAc)enKrB~Sp~+cyn^m~Ik*rGV;vJRFf>Lm1
z(^gjL^>_C1J<q#BWF#9!B@Es#akWOU0h8=)H|)b8$K0m5NKUoPKp7cNYFzkhEeOq^
zW}K-I#V3RkCNIw!mM7V(zOE~;2uQ7$qZHG|MWku%@G^lCy#TB*e7zUxp<Pl1&l;Hv
zQ3(w_!Elg#!uKDKEx#!$pyp4vKbKK^1raMS$B{0Yo$w|qHgL0v-6!g&NN{`8(F=9u
zk@-trjf}QUj9#s`-qjT;=&S0<nGru6a~~j-Ut4z%hyKaa;s^!=i;p&==8X0{`$Dso
z5V=J(qoJVNO%^b1<3$CJ%`9I)sg3C}WnVzt*)l!p_=e%9*_~3wPi2jJdeFz>Bfc)m
zkpY<G=ZaH_xf3#jCDeIsMpNzQz8>Z|h^_D_2h(OcxI>hD&8?;@vrSA_zcAdrsNlE#
zdC8Ns0kw6@a34z#+Ed!vmU>3OLjh7!oy9sim6Z@SBb1VUt77L$sIQD#QP%nI6#Y?S
zC|W}4vw-T*2~QHds843ksV6T5W*1pZ+)SJ>`i{RNSVeGFgooF(ZNx-uwhQMOCFml(
z3oNHf(-NcDt9fK)1CVNBJm1Pn`AIIlOEEmhzFp-Pi99FfPG6Z;;VvEGKMl$O0zvKR
z8JUJ=1JXNcntb-P3R}LOHti&-8>*<e4ir7upogWUh%9|2V?S-*x05PWLM?YnI+ib`
zF9p9a^ruH8i)hhjLB(kEAc%*I7uKy<&fU#iE*-i<=0jS+A;AX7JI<UV$<6WJ+7=lV
zqGs-LAp7{BW>((bAb8X@U0%tlD5=(XlRevQ)Y6TnhhnB98FrKAPONHDI_j{rlJTKR
zJ!UeGX&fi(RDy{vmdEtv>6LSEy8bo-&nI-*=7~qGk6SU2Y7RKUb#(``Zn%Evz&HhO
z6-!+7-gL_fr*GL&k1A#mr-s=V%{!L-Uj?j(;oF~z+9gu&mm<a!7*CUQ4`lc-0~bg4
zZ#Qv=DB2oY@#h>5N8X*VVe~{Bk*~L?fkmdF=Rp3t!af!9tncIFBBRn<sX_Z>ji`CG
zp74VOHT2`U^PdN+r>j1eStN$dk8{?Q$r^sjs`K~%5IKd>C|#?zYUqsuCyhB;`9$?r
z)>YRTNa_j)<mw{io>~@FlrMH>3nYQD=K^w)jm}2WZe{Z!EZco=KXs_eiM=2dHL<pq
z!6UL)Cl?el$wh#wt9gUu`tbG-M&wMiy*r8nB|N$n?$!H@8y|6&j*&oeKp7UTW?+wR
zfmd;KHOSkS&E1tU#ErVE(al++)#?FaUJl$`L=|QYBV1!*=A@)LuFKfJ8{TY2MKz_a
z;xlUP@eiU9RL62^Ax$0`uBtf0EW0hg{GIH!->ciPg1L<#Rk+XRSi$l}`ZHMm?9qV#
zWeWv<BZ@CpPE1pL!M~$-K-$yE$zeSGB@)hIrXW69PhWTokYHcUEqt>cZgwLTcMGve
zPLw$LrDZD4X`Hpx$nZ$kNDHVfdwS9C_HLU(A!BGTfM|^s_I^LfqGkeMCUwu~$Y?Vj
z+b-E=>!$CnybUzmjo77+aU5{DbVK{&sqv^%-KoZI`2`O8a6I?IGMcf}?7=Vlzbv>M
zdb&7qkK?*Z0e#Uq=Lyo`NNaJ*xTw{Y2npL>`BMaM(x%sv)S@jpbSmr<{pNpO`n?v&
ze}p*6!1Sldo{my3EV)g1Rr2E~F6GYS0fRQea=dLpazK8pAX<3S>M-vDeM<d|%;-Tl
z;xSIwwCzdFmmLaoa*Jz#L*MFc)#>(dLJ*!x{OXP?C~$Xh1V&h-_SL!}i=5;4i?)q|
ze2ac|Fe8AGu%NJDv8$6oF6J##027-VF1iG?ztZt{Ieg_w-{lJf<4p_b%j*4OB%`nz
z%FNm2!?tQyR!frm?vIWG)>|xOcK&2{+~e(iKeCf8kGAou<40;yF;tFuz_;<!kttQ@
z;b`4!^U4LGrpx{<PLftKS#C%^4!Hw@tU<E4m19pBk8BY6QG4SBmlAxJfPNYj-<~3_
z1e3#xxTBli-ut`6{Wnhh@uKOO4+NncqK&Ub@DNY+f~oDE5JuTBdmDA6z#-X_82e0Z
zE;P#m?wkLN+qQ*s1+2gh1t0X!q$wk+UuETLwFC3-OJ8jQdMUFMaoG)?fT@FJ!!w=L
z@Vwv^Dc4<JoOXQMtX|C=DSr7oO28=m)!Wmn6;_TNF#mVs`Ff3rb;r^-)ke)cVwUOC
zS8me9Y`L*)tN8iLd)_+KRHLY-gdu%d3;7WOTPX!5U7pyhTX1L?j<XT+An1EPRdl|e
zEtAS&bGtL=8MpA=wk|oC*AdU1Iy@>}{^4&i)7icgFx-9GUQ7y~zVW)mw5MHZLFh4L
z$Yf1cy1a!;=gpjQNdtU)bSHH2e!OXGhEI6QQJT<-cM@mtRaZ(~ROGt>lSfmg%piK@
z=W37P3Gq&s6F-;sqs8vMn@Pjjd+)8_=eADJDC0*~9@`_QBMt=D?x=;;k9x&<rEmI=
zNhvF9_=j*GCsW@$Gm;jlwy27WMC{XsTfar@A6E8zeV_jbn;!-SU!7_Bq#R^+WrHqS
z>7nJ(I1jg9fd0Az{BnV^siQWt@2xb=;51<mIX@p%H(k4=y^L6K+zrMGOA+kVB^=-1
zz6shFb=!&c*LF%;R)5^P+v^akB%eX%E575T<dYkBk`d#ZAHnr*%;{k?w`3+reabG%
zQ<LLW(VZFA>$jOqh@Yz`FINXhzv)XC_xaq)eH)E;K9S25P><f)zE#-0O?hN{5@aur
zux8&5VYg4DL)v|_Q>$OQv3sHHWS@K!aBtahJ?Wix<a$f`q1$<&L|FJ!I+}EZY%BSZ
zGCXu<^_tDm21*_A^FZh1%_p&67$y^>wg<RfYmRf$t3n)(bFv9BkT@Ob#PBX0ho+FV
zh?~pSbGEQnJ@>+zUVOuY0vJbi-7BSl4znH80h9j6ea+8n9Pb-3@8_gl)wMAmvJcM^
z`@%rvL@qw!0a_*O1`$qY-8^&d`2zwJMy4l^l8fqId_$2No^+0jT+=}vGN1py^Pbep
z?;!Q&1Y|DU(A&bLZEHtE2itYr3V`{Rimiw_7O}Y=7)p;<L_K+WN#3_+uB%yZUqxyd
zgCEYtVEdcYg?~f~NoDKGYioVnJ4IRVADuNhJ$6k(z44dcYaQmR-<#17mpfCGR;S*;
z@`a(E(|)Ot=kIm$m0GOVyNR7q)!T*r=3LAr-b>--!2>*MPZ>2fS~=v$dVNZp=-J(}
zdO9RG>b&dX<hR_-K%ggar3Y<J3)mOX%d~jCJ(!xi<bzI0+RS~6M0yTXGK0}C!x!)B
z>ZM6kc%NEc7nVBJI{v}a&;1A4Y|jI8?wq$o>7mz<Yt6#`jgF!+!qB1WcIl0@Z^I9F
z%%pRdf=Z&;^P((Amy14<^%8}cy?|s}#&pf0-#1&ZpZL59IaFCXk$~1cFlW5LNU!eO
z%>$;J7r1rnoFUKlC#ui<@R45lO_rg!%#=*DZkJWZ$&)^>N89wm8wc%vD=Qw(TLg*O
z@T#lszeZIMzR_VYdt?=Oq{+IC_e>MBNj>+dybZq`kkw=aaE|sK;g_I#p(+4)%D{{O
zm8ooY`QpP+wOB}hIAi=I>`wn&HK_#MMSuz2jj~tD7Im_OrLPk;lpK#HfZarEnvHE2
z(O!DAxFk4Q!bS`vBZvubA4e?krOvojQ;Wavd5a+o0e>uhTLhEVI(R&PpIPyI-%m_Y
zqk{uGUl>Bn$*=K8_ypN5&H{9BN5L>{kc<1JEecP~*N1)EB<&R1ZR{g@U?FVwCiCpk
zpkqDwqZzFt_Zd~rOFxs(^5c-qQ?~Q9N>dzoZfUR_tK|yO3KZ3szQ<DxZkO4}tDP*}
zbgTc<tSVBu8u4mI#6b4+QbJ>5LjpvM$NzN2Cajwve92*D49Sx&kp3GT_f1b|<tQd@
z-Puz)8!z^L?xl=%>N4I2?oemcz|Fj;2XKJ*Ls_{gshDdW+530p=0s?B>ZKfuG1r}^
zGy@5b0hxv%!0I;k+08o94BkNPMt8I2iii?3VP9y>JhHpPI%bWt!LzWHe_1TmxShfr
zu=2Le3wh3PAcV_9vK8)gXD|}KRM3tt>)Z*B6Sdb&cmLtedj_`eXBU0^d?%J<=Cjr#
zLk~kWL}_<Ld-HnCP4a^~5+Qevd`w1_`rS4uqbsTC*(P<QrcgoC<p~UC{jw@2{9Cm;
ziW_0v$7*oJ|ATCQ6!+~$cH8I+w$Y5Yqqyn&6}OuhmFZg*?4iT9QM|kCw(GXhUl<BC
zz7D#7V)~uH4VSB{bo|j(?t9TRiB*2SKWYB-kJHNhX0Ga39}k&SNmuW{UX^kBo2oe1
ze8APMT$*1PvZ1;iz9tQ}@BilfKlbI17bh7Q_|_xCxD<oE1=GIm^sT<+QspKmCd-`K
z?bgPUvb|?oSTx(#lJzRJOlj`;`>yU(YFK(+Bh{|=wON@2H7`+V8(ef_y2V8VSQHjb
zV<1_HEV%5R>^!?Yf*NS%sTeQfluD7ap3&B$zo8jpJMm^cAqs|GnM`yg8&<d%8YL0R
z?Mk;u2zjzSerfiR2kEYj3ug4Bx;uRE2*UP2UbQ^MXU(urPi-*lnr-}J=#IC+z*tj4
zg$IwU6nVrmKqc{-Uvg)SUL*UPbk?+y(+~KhcbUr3dU#k)FuzX4hcThtTH|4t0`UCU
zq>0|yM?VssOZLJFtDW^}zA22pw+c$S_NZ@1g`|$a(VXUrcgXK~Jbtrlt~4w|Q<qSc
zvmQ+-^ZH+;p@pu49yy6r_@;bhXFxm^CNUs2^W2m??xE1}a0TLURhLo5(Ox}E@vY%g
z%L6IBOz{g-7qL)ea$l3KIg~Pzb{`vshsdwnn5m<ktUCfn#SF+C)+2E4rQyep;aV^Y
zWTAJIht1&~wR3MDkGd3X#HMDE&lWn^8gSoqnpJYGiK|%}iGJA=SZ!(tXOb4fL$Jdh
zCc%#**Y7LuS(OGSCWdu|pQd_nL}al?V)U)9gbA$JbHkx5B;GDov3qa(J5Oij>?Q1l
z(Ufe0L;!AK*+OmbLVj=Kk}Ww7k|e$qei?GdvdEuRCBIeB+LtHF0WJgC+ZVZwG*c91
z`SXS))7RZfO~$Vlb(&G^ev<Tgfxc7u>#pSlec$p*(9N%Jn}1^ZmsmdkNBe~CcY5!`
zJ*aholHo9?TTZ`*Li<bG<;G@9{S)QsQ>6M>((C!Mr%gZo<=j7>`QuHz{&8CVfII))
zoy13K;XK=?X#|_^OV6D>N9XD93lsGRbjm+o#lO4r-`zP~*CoR)CG&L<{UN%_{}_!z
zww-R3<8Ju-l@{G{__|!O>-e1(#FAgKu95sCq_BDk>~YM&S6A_!H?Xg#>PUw$DvcJ<
zODyJmnDM=>495-SN8P4>GqBx9+TAP1J&v!)`wP$@Ww@SF%oxk)D^TOB+*k+*y+RZ-
z$Kb|_<_7fAxR)=%n3q^yv(Q!ESrxDR(1_p^2S6J4{U`7LrpkZp%RfWH`^PBxWA6GF
z64<$^r12t`wLSa{`p9Z1zMkm`8DQSYB7dn;`d!MmqfI_fmHbBn6&(@8P3KO6Q%PoI
zK7QVk;1#QXU_2D+C<>C@F-{+N+{<Vg6y;%PTIIFYX&jJOD(kIZ=ir`Akj{b@<Yhmw
z8|YRfb!DsESyngOHAPLa6i8d#u}=Ve&msr(U^~>C09W-KOW0L)yyiCeD;EFBd7tBR
zPIs8e9{dT}{e|4<5P?Ueg_6A|m*t{{KypEKqN)`k#EYH2(TsP1&f$+Sc}1Ngq51Fx
zD-4BSW`V;Pw<!&?V*Whbd&i-xpnyX~cxhL&b1)<Sm9)HN#T<7wtVGraGN?~Tm7c3c
z^x2i03x2Q3h){u`M;_>DY6FD1Z>1RWSyp*1{&0o-SA_k;S*gRhdpp~K<U{N6uveAr
zSJ~+`U7FkER==v46xKsMzftRV?!_B=;ThE|OXo&srlfQd?64vCQ_feP``-drr5FFt
zrEl*M-NxO;&d?1g`+M(NQO-!-qZ1Rrc10p-x8>+FzI&Ja>OS$S0SH5ay7Y-_*7rPf
zq}U_Lt|uRU6$K%bMwwDjlTD3=Yosm5)$80}ulxtopGY;i6HEi#Mgd3_g(a)lq|Fv8
ztYQZt`?yY@$V{nDPP0p+tx|K+%!jLgbn57S1l06tOL5O@)K1fQ{HQ$&-{;VEP=R_Q
zEmvi+LX+DWEH1t5Sh3_$QbZUlC0ukJ0PP+g3iXk}kA{o}8iaFSRlLeIZfB7t&vj<<
z`BlvUo?iGA%YA0*r@zpD<;Ol6-vA_>0#NS*@O2PG*NaB;ksfe<en&m68UYjQE#4;>
zm?z=!ede||blkJ&OTz}3szE@qC;m}KQPr56YoV)~@SUh`%Ue~BuDus);j<9ZTx3Jq
z_=9R&5tWtGd?Rx6S(+B5v-YwmY1~ToU^5@o^tz%8E-j7GCiK7z6XY#%<e@>bZ6wLY
zUE~L&|ChGE5c$8KvES+Vd1Z0EV7s>Z;Es#rElGM6Y@M@;VYj?_4>IvrW<nM(U%E#x
zJQu85)%(1M=cZJk$j&qE2F<T}klz9;p%?$prQaK%m0JAV*}OQaJ}9F9v{Cxj_lm!#
zDI6`ExGEX@x53M#Sk|$Mf^yfy!pSAUKwH@A>u@n~9g4-Aezl5vc?T;0*>I81e9=%>
zid#?0beez_E2<}lWNd5#UX~pe%hCPiFJyf4mhF4wv%qP4gQco0=uP1xree2UYcv63
zFE&wlXkPOoZ>S@vXl<14TxZPNt_$~vCMwbn3{c-acJ6caqvKMzb#z>Qm}&R}TfkrH
z{5_d}zs+k1-ZoAg@$B%Q=zRPm)o%>n?DGKoX&Qs$^1UJV(Ia3==JpU!I)Cf?g^aw=
z<OJ<9$+`Ge&ukA8Rrphspg_ZAz#Xk&#{pnhipc|&;pM`GHgt-B-!}=K92i!5u+tZ;
zn^>o4t9R9tmF^0mzG}4A$Hi}_m^r9=#l<K+6n<>w`;HV{*5Edflr0R~&0S=i-0W45
z+vPWSsCUk#b7SDF-q$Ann%>plwRpA@$qX3>&Z`-ggYB_2;+QDB-V$@;Nv5G@p+fqC
zu1U(a@`%~_2i&9ab4J?oBxOJ@{OpSQy*DIH8Wphig#qRaI#-Q93jY@MG>YPulZQ8_
zz4oJP#qg5=f!L>$_q7H7%GskI_-AJiqL0Ia|8(~FDeV1!(DVBa|DFUp16g)$(;H%b
zRsR2>lOTgu8pkOs{Cw<gw=X!x@*U39zd+z9P#seL*nM`p+@aerFDKEm7|@T`FT(vY
zYjrO_jS1Y#>@!E{Ow(1Lgcs2$X{*JyE%)qg(43Chl@>-@J;}>iV3I?^k7P?qSi91U
z;N<tA=Cm|l9m^_1+yn$v@i|s0sQYax3MuhjzkmU%DD625Gj@RtfXPc9-};ZCvUgHK
zLz)C8RjCP4#h=f$!}E-Bk@7~4mRQd(4DRAlao-K=qe=53I}SxM(|E_QB8U8{dCR`Q
zSpAWJjC!}KG5)8Dt{YEpjcP=lHf+a5Vfy^)-9qm3S%G*Ty=kFk%an`?83!gjqHv8N
zT!~a8iH{A;OTSR8S7!B4l^+ezC3`BcQevdYV@fY!c!}@4|L+^Z|9SVn6=uQEf6|hi
z&SDLgP;&gw=^ItuwGJf(zUb-)NP+6=dU3NA5jQU+|LAb5GCDWW;h=o#ZvIpDJUHuD
zx0k8Jugc?)5fUhFu&l4M%!x*-^j-Ha9bzH(e)fGie}GBHAPTc~{3!Zg7(hl+8i#tf
zIagSC>hKj$b0@VK4~5aO4~9b@>+pP6Nw!asVTvw48mId~F-7mF7`J}hoiI>~M~1K_
z$Zv{{)<VNp8?yYYbCIum)u*VdZZXGV)+R8dwtpYk+R!`}jVha3+gpOioJNjEHy?A+
z9Z|O~N_?)+Acux<QzxceCPW$zqI;3P8WWq58iBjL4S4!;K$0Hj`4az$%*XkOy-PY9
z)>pcCAzfb>Lfel|CaD_dsP)H<amS)b+r65|&@nNt<{%BpV1mra`rWO)NRPk%5j7^r
zx>uzq%ctXp`LW);n(W^36YOUQ=?g=bV0~lT7Y6-pB|>duUyaGLt@HmUi69iY1lU!$
z*5MXUb4|=?IA*g;#+6?fl)>$VZ%~<h{^H<neB7K?QGad<vy4MmiYBwnomNIuT}C6#
zc*Uut`#*Z0i9Qrq8V;p;%q$`RH3_YifdVg*d?qqqKxu5aa5e=)9Egpt3&18=#wS4c
zFcpBmOi!OpUp#C9XBRI8Q|4?Ck~mTB>fGWXCxHwPqzO7Ek>H_kq&h>5!VT(YCYSma
ztsMjk&#6H4J0cK>CCBoS#f4S?P@FqNTExZGu`5<vtMFZGo;Xx`zO(hp(W9AX37m@<
zia(X84PK+7(c%^ucy6^wt`njVgUA&|q1|HW|1S|v0s?$?wpi12G#~ocyj}2z54&`~
zIK}kzEN`jFRoYQ@X;}Te|D0id*YTUAA>s37d3D+SyDPJ>H}8LXPp|Ixp`ZOOdVcTC
zzb^px87SsjrW=OyWa>!o=tA31zkK?A^l(T(oA|rQYW+|>1LyYL@8DOg{R0I<ujutR
zF>(ZI%wN(kah!_QWOQg1e)U}ie{|kipWHNV_L;MfW|AC_xyG2GJ=eUjzEtA4(3qw2
zi20-q-C7){MnKIEu>>AiqDT~=4?G=>m_|E~0HruP$A*0i)&LDZOHne!JdK;@CRWQc
zBL;ldb%kQ>>_W-hC>34?qR>~h2wewa7;()aJ#XCPN-b&mtbhNCj5WA@JLEy(s}tmR
zl-CE7dPTF250{J6eh@r~#mbDjtP%$~thklUReK+>>7BbaCC;mN==vG;g${u6U_${y
zw^5c2x3Q?NmG8ECklWmSCeVslVn5QnaBZmk?fhi#N2fX~!Zxz6rzi#;aJ%uOL1%XO
zkk!*#q`jp%Pp>yIY^6ZP0F~jkOq`sv)<%2GZD!9q7^!7UUw_=}|Nd8!fA`0~7f1iR
z#Sy)8oM5do`lzKyX0?w-Q-D?0)ur%`(M(t5b2U+x3heHWodjyLK7H4AaT|_~c#?}`
zJ2}`+v0t1aE29syLn{)fef-Y}^Xq;cAN@{jT<7~_1~Wo-=j)4>_QB|^$euph5@9K^
zr<63H)oO23V^3Z%m=>STE2a?ALb)E{REgT&$mpaHn$cdc10^CR8;W+k9^Yee_19tZ
zZEUd0@qMu)+cC#pNHsD9=Fvy%8WUrd4=wE|{*JW^Ua~En_ojpJ9af-~F;*oaEI*S8
zKpODNWXJSxg0%GyEgUQGP|oH<R3?{n5VncYK!~P#adWT!q7=45#%sf)yQ?<PZq|fD
zxkZkdih^3VQ-=yO120ah8@&M6Ln)*YGq50%_P%sOXM70b!kMs0$cJ*$jNIlrs_`3g
zZhS5+L>aZ$eMtQlo{`m4Rw-dL8bZaG?D;<Je)GEr{;w7MdsX#MTvaiAD|ldR&b7~P
zf^V<txTyX-e&-jgPV&5Rk)+?Ne9P_Ky>o7z-h<k7!v!9z!SU>}%Iz(w7NO+Fj&1q6
z@zP6gZKDqA-d?$?O0B3WZ(>9quR%_l-+Ee!cl-Mnm)-FB1j<`}Jsbl^cudVfy47i;
z{4l9d<JhGbG_8f-FS+k1wTOUQ7Ti@zH{*2DHsVS4J-6*J(;XQ=O5O4V>hE@ulR9ew
z!kL8cTYma=G@L()618o;kkiU;7tPHVCFtGCXiK+>p<A^|9_`ylsh_e>t#3zeEFMkD
zT%)6H-1wTlW;$Pg*qgS0t#9biN*EcnZhCs2<A`@kdeC@OX9?>lSL_83c@-zV6cJfZ
znG!%+dQ|${MiI;~Y08BsNtaCsKf=UQRfXppR5X5!(he5|1tUMdt?mg?j@V`%P;^DF
zaJx*gCIm`EAu?AEvuIR=e`4*#&4pP@B++a*C^fHaQAK#p-@2fx4^&I^f!bTgDVxBB
z%i)$cr@J$KZ*sY>sUjfm>Dxy320isQ!v1lxy5MnSnp>82FcrlIUjYqvql7S7cwN8P
z)d|X(+kfW8uMGJeAODOei|O7#-L0dGeQ~AX(I$b|%2LVt(Pg{5s=4DY3{|zaI~10M
z+oTqCAAlU|<`22v#uf#AVW1WyrhNC&HcGMi9J_6z;`w#I=l~LEjK|nm`8#4}-HN{-
zb2+$G_pNewfpUt;x7;^(m#0zf#8*4YoGu#chOvxy=+<K)qThVH_dA&JyKep0utfh$
z6TV@r3Bi+9yvG!G%+bB~X{(cODERJ(POtkCsB9Kq-GZuq=k`afOxFx{75ygmkQhnw
z{pVg36AQ>^`}2lpiz<Bcmi@iKiEP_%Mh3AJVZFKD*IR6u<jm2<u&VkhaeHeju3Jvu
zXfTufbM6da6a~j6!BV75!01H`d85Fyr}iCAxzuQ=qtBgCI`ctS3E{OwH1P0Ea_sCm
z8Zl8W-=R?RG6J+(JjLI#{|D0MnIp3CfRQn9&}hK$ZSmV|Vbm)O!lRvq8FyVkQ(jwJ
z^}t9|UjIPG_NhZr#Y~g~@s}p<YjiMj`)C*yTsBHfiWn~*^fbp%Ruo)NbELsFpsXdq
zXWU!X1X?ZdXCNMw6Cy87F?-Yv9b1G(pb$%R8MFva&)?0I0~-6Jr>DbX5`jHRtwJyI
z_PsWf<j&}p9z_1A&;;q?Z2Cp;b^+dOr>>p3nr6{yENROa8yq}HHB-*Yeq1{6i><_n
zL=c#|Rf!1pmy;v=mNxpU#)QN#>F$W{PVBW05D2=*r3wyE)fiG*<~y0`ns7LJ79`_C
z3t<AqijR*HO;GO3PXZY;&yjHExFnrt3EG`72WaEWU;s=zK_ZA+kDfBR1|2MS9$8M`
z4$ziL)Aj`@zo@oPQyCAD;Y&xO7sQD#<05I30Y~oRiS)HPexrnV(-;y&ZU9>7+LG=|
z$e8B9`?o8w>IPK}nn-D!iH@Q&Gcl?YQW(qrDUqEoP1@3XNg<|&eLDNxPRO>kfCr?Z
z(r20mF0mQ#S}uMtAdnTd((m78*4Bp$wQeb!svKJ__LimS)4@7HQm+xLi_{||uXLrD
z#g)edN^Eg%^Kg8elqwxe>P`>JK+dXs)pL|!;t}AJ?q@z+@cu8O^&iiixrs5pVxt?b
zkxv_H7rFoQH$Tzqs<XM>G<D|XUKj0MB-hYA7xS<6%e2q6gb8>p9>fuCN;t(Y^3t#T
ztq5-nXMH#1rbF>rEXc5;A*&{<6%qG!5C|UmP_{s{*-%A&dBQ(gJHd&Sn+P!CX5L)^
zR2xXUZCQI(YYVr+7Vnbnjd-l)m!6bQXo?wpZyBlha)ZBq0y5GIS{oI4v$f(L`3+=J
ztBlPf=as{#K&!l!oP22QN8`uy5HRsbkK^hP#hG*IAd>6(jg`$cV8^DL_nXwKudlqr
zS*_u@GA}PT`&Ny3vQ`&?@SQj`9yRVu00Uc}x3l=kZTmV<9)lhvG4jKV4CO6Hyr-T&
z(q&S#eJQ5Y=n2!P2~T&@&&fZk?+VTfQ~^hY%b^z(<OXd+#+VSJ{IML8<38fKRx95R
zEx1|e*MuWP(+c;vGO9b2U0LHv2x*7(@YGvmq^HHRJ{o1QE^F<|J#2G<QDu^`$8{NA
zKsP?jxq+`b)}g(UVhpPU`4p|mht45p49%RnSTEtkSmql--;`m;3k%m|L9&}oBXsuv
zkEGulY~l~W*ZZrMt5?(9@ROw`Dm?R1hy|w_>Iy``9pX^v+&3mh;0%84KOpzmZm~fb
zfl2inxiLqFsnCZ>3-axix1KfNbpmI?V+x<lT7BqlXfKCH!|aHJNG=gh8e93M=vV>F
z*a{mGOV`)Xr#G^GtB2~^C?RCL^Xv3jTuBFote(k~qZ0MWRxDb(Ab@cf+1SwGL#MFV
z7VXs*wsyolYJ9C9U0YmmFgkTtXozBav%-4&bc>O6daur2T2$FA1seRrf^#9J(b)G1
zifC`4%{+|i?_LbJ(YF=HuIl+|@clHDC4hvV{pejh=)KylW_h{&nVc&%#W2=BgRCP!
zh{^Q4Sq5OSYTi=Y&qYW)8tmDm4&gELgR&y!CMM^lt6DIauytGOxyXxqO6%QS7OCj6
z*S+8cYq~t?*k0U)y7y;JezP~?EIS_t3nR>jWx^#!45iKQP66Gzl$<ua-r0!3u1+ya
z_7cWRLCm;q;8Zjn;}W8#RrR@v+pFnjxEQ4}{H`^5(=XFkkL;f*B0o0TO_uZEhUgl0
z;?%I7cdtlOR^4~!?F3v%ZfNH;b;~JQQc5y?Zjaa7EB6sIG^ZSg;znwrWxIv+3q*1w
zFUx72GG{CqawXn9)O?@^saPvhm?wcG^qw0l7df;a76XB2*=?Uo8awIlrLDu8*h4JU
zjveMXWcW^l@>*F1)_H##zfQod5Y3Kx`}{mQYhB|Lva?HRm%!-aY~@nd(9p;4`}jgb
z`R3t1>Fe53SE~@tY*jJ&WB^GeO+=WY(KS$Nw+sSZqk)o<cBdvdnoZ2H^zxCXkaLV2
zBafJ|`I+FjoG9LQpZ<;|qwUCJ%!_Qr(fsZ3)ew!<Lz~Fc-Yq}TU7La8#qwkm6{!#L
z><hicaioovg0!)P%$#{pC*cGWo_T6X5v%(}<@Dv8@nTxR+8EhA<_p7t8JLcVqC4t~
z2uh%b_6eYVYLz`~yMH*K#zyxX+b`I_>YfaSgybl!Z=0Wwt5B_3klo<=!tj(vXh?M3
z`6c9Uu2Xx2+&dII(>%g!{kQe0x$(w*VOT!~?vLc1uJ9HDj!syAo?cCoQ0i9I`NHst
zM3-X_;SRjGEA;q6d#6Ij9)0wXHc7?;C#P}uHUf0hGUl`RBR^Z=_ocekE3TMF(-qT4
zUl^EA<z2ooRJWpUM0{b;eiKCE@9gn4yjH(2P7Pz9&o(<>s@WHL@`WKbl1nyn+@A7X
zePr%ORjcH=1*L+QlgDh3I!0f^l4pZCi}j~p7@9Q_5E^Uc%g-n4`&@X%+cvi{$R9jU
zFX1Q8DHROfUwO1m53&(Co!!!XbI)|=2hYjH$}4OR!NEPQ(+Xc0zNLCT<0h}%+9?c?
zp!;)e7|u8)9hLl>asR((+$5&6rVw&$!Jl@Sjv*C*gBN+lF*-H#J{;&&Yg7#JkZj@b
zH|8YLzA7~pZJ7g#hNqG=d85=Ay<#5}ojl-8H1ne<2dqu+!UEHT`Q&It$b=i66>iv+
zisL@x51a6pUAe)&>%K8Bi+T9=N{e3B7eH!agLk)@2E;708(56c{`{wS^m?R?64ysw
z<N2g{&K1XOpHQzJPS>+buZoI&QRz;xV%in$iOxT8Ls^2sTaXSms(+oak$wCD>k7)f
zszI3Bz9895g<B}S{zGKZrP|7_u$nbiZ(Xf%kOitZ{g9Q`FD>5^-r>3+mEdgJ&@s4u
zT!NU^ct(W9<jsJri~!VB!I-KtKZE2$t3#Uk80R@~Hj6&c+8UXn6OIU`?=^|EkaWaW
z0;VIXk%7`stpGB>cNHu}O{DEoN$~|T%8*`DX=u@i>uuDr7+;f>w2ob8ipZ@*W%Dq~
zYiI@JpjgCspy6zn>k>Ue1yw$m1IVAuIO5gJ%Go%b^zD8@1H>uwQp^#*<tWDxYCRp}
z#=bp9mV4*{kewjrt(d}Q7Y}QT=`wG?JTeoeU4%&|f>=OT>p9OYd{od8om*7Ki03<H
zmHDH(ec3kV>wE!9iUhm(3^(+gnmJttmeVo>3*7FDdy@TElE8m2`^O{whx<>^Jl>f@
z)2QR&Vwi460(cy=r@ZtL_>o$gI_ipp?d+t?+dn5VSH7-r8Zb6U6up%uU}@Bxc(p0<
zRx>qz@_N@&GglZQ6GKUh8|0KrTT5e%#&@N7!PT`{4f<Ai96Cb9JmiBjFNhr=HS|+>
zgL}%mc0;-(2<asg<C!k+MuZ9*1?g{c%bTwhny*OG$blNai099{KR-D)p5SP}3At|t
zl=jCO5;!#P)8gc;DlEI~8z(i<gBBjEX~6~gX-QLzBY*<RwJAe^tT}ZT*XjQrEC2AT
z{>7944Br|a7$61X6qZ`DJfD3(vVjaIUntzJH{;7ISV9df&7CgDL$$khs+yiFYKn+B
zb0t_?6=k#D5Wd^lD_>|Idr?JWGmvXXV1I{GU}>!8o1v|0M8sO$5y2i9u^=qA22r>k
ztH6|!#G>CdRN@kZ_g4W=yW*gw^=VQpHr>0Xef>9G4vNp539>U=2sJ?X(ZDXKV9K+I
z)gD&uu-Xz25^NY<v>>PBSQl1=z5ezn+(YNT=#r4E1or?8WI<zOM^_7XN!B$rqP&%C
zu_bOy1Q5XXe1dbW>f>I?(E#mWEcl4|`WjfISu7l}#COwutkCVPHgk>3$<8}@n;Y^I
z0ezEH=O4$BzK;r04PYg9eJM`j+H%@8CGIJdw)Z`)D-^JJ&a=>)Don(<Dk3XJ3v@&5
zt|>)Ubj3V1DKn*DegwYkU~h>KnUlEzs-D5e;fv?4`E-xE<9xF1ePWBU^kzRln{iJe
zwDa}7L2SGzxDBn2qAGCtjsAF>FiV5J#XZ(FvI^*URUXF?pt=Ci=JOAD(k`XT91Fkh
z=1DYgpwfAhCc1?q+q}^s_8rc9UaJaJBE>x;a`*+K;wT0N#CM)f)~G>sUaY0A#iNgy
zNurKCvP>&b&qM@Uu)^Gz1{Mz}h98SfuZeG<24hMCMa4R`iOY2CSvz-Ti-J<CF_(3-
zPSi`p`qb7kMIRK(=``twO?sk3V5|8r`rx-t{S$U#UqaPJ8-hR9F&&-sj)ctc<gOd#
z;6i-$ykV@;%R*j*8NF$;ZfNBPADhqm?esl%UfNuA%ct8nn77Riw8e7tziDM?`ARA;
zJa;r7E~{c%nP*-Bu^eR_f2*95j8hRAY-`O?PLZ4y?RD>#cZ&(y{un6^YM~hNvfDE&
z+V^iu#W6~K;QIK{<nh>|3&vO`r!cqbG0?<l;C_<MciT9hCxr!>=QM^GM`Ug#az~b3
zcz3?@w^;LkpzL35Im>V?P~eIO42miDasmre%{Zd@-NeDQvtvV!MSJ;UqGnIW)u~=7
zXWK$<2%)`_jkr4=wzD?@wN1_<s?A<4w85k=M!T+M`r5CB*VhE}_;)LKK@#i@hh)Q6
ztZ?>joKQebiQ95+cmfsZGX{)kZlee8Kw!*j0bq=YeuJBGgh27naMZ#G>XbJ^ZdD%%
z3M|G(jDy#-3?mTLIQ>3LA9O6;X2Gs7#*tXhlf4ig4~={*zm;)<wJ=N**LdM0;F8Th
ze8)LKp;kW|n1Y39JYZtp?<sxrRGDjBO%&*0j#a}SIX3#Wr?w~C59#Vg$S?BT+SkKX
z!E*w+*yQv*rJy_l<2_5(ljDkXsO{24f>PIbtf)h`5<8oOHk<TzMyZCn3n~Ncn&9gR
zC{4o#vgYK>?GzXAF8DRJqq9n!T=!6J^w3w>;+|^$q7Xzd*;8PAMG6FjI<^*1<lp=I
zgX(55Y)i+!UFlU*hEKxUQYzLOTw{S{vgiUo9NAr|O+0WZhS|7RkFG4`r2%!V?wc73
zpi)u#Z2}Lo#lQQdIZE-)XkBvt${{!}XC;Kr$UL`wC39#I%6W#`(J&x;Ol6swo0^-k
zLKO#YY>HZ9C{z%cj;3Wymg=zHA5r+cDe5>qK}<?X2pB&O-E~KgQx!@<EL0M-+DVvh
zFoF_B71Aw1fR}}XG&U~hO<%(yom{&psY~WNs2hcwI#nYsbM1LixqB~E+fx~jG@&W|
z4SG%km#k!|@mGILHq_LG3;Jp&J~gRa?RV0j^F<Bmgr0B))uz351^X0F4bljaHdCkJ
zhZq}V5FMP=UyPItv!sAlX}GkIcylz-sw?i$S|_E@fijLQM#Y~_K@ao7Xry2z7V6fD
z0Xjp?uKOB5&Am^*JAPEGRS?zVU+q-&5`_?K%M<9nM#0^X#y8>qcc1e=9)9A&nK|v9
z*ud@slFMcT^|SO>2u}tss_V@LgY?rf9~%nEMeS>NkN<)6ujIz~Yx6JsFcaCNd3m3<
z%(#vM2g~e9<@lNCO=LCjvEKnLvl=6rSc0?lgzs)kHu_t&j()b5*7vtktyU!)CSWXA
z{Fz*9O*FH^8LC0Ox)FCMZ`Nv-X`c&Tnruia$c`b3v3OkU-Kr&(y79WOMQq&o-AmSS
zmgw5IR$Er!%?jCg#*otU?26%wWX&z7Ot-sUivyOVpRZkGake+awvpYk7eKyg{vTEX
zfGLK)F$5jJ@<<=$G3Al*Kp&WFm{*dQtxWeOLF;#fAj}cL$q%y)fqZ_FfMg>+hie^3
z0gtO)FzE|#66yON+&rpgwa$E$;$3lSz1>N*glVnQ9GjRm)+Ze^PBjrk6cXL$Sb$b6
z^k>ydPU}Or4C7V$#s`}IOw!}E7X50$Rqc)nrt1KEg<`&@L~qsUh?#vgn_GW>7=4HQ
zq*i@c`p)}2nM=C`?~(DW3jWIPhGPGWw1m5+cTb?OsO<7qe0q7KmK^K&{p2S(a2tjE
z*r<k(F2(!O?80~N%Y0i`m~>ur!#xkY+TC2P-QC*6={_wV9a_vY=v}F5u|@oBSCzOV
zzlo=G>+xl{DfcB8ep<?O6Dalr0fs-d%ojt~GnFYF^OCrc0<F%1j2*jqs}BA@txJ@y
zSoFj(1AKQ<C2Z!TSPa!}TUN<@`ulD9%`<=e?&z&6Ft$OVNrib@Ugvqw_ibOnXc#;p
zZzI1|wm+<Gj|9`+Y~T7#TK?&U|4L3<e{L*FK~Um(zT>NPgu)(yyPU_shSUuY7^~AP
zrz;vdznySW)CQHZ#$Ggy95wU2X8$|{CX4{zfCPql_w2A6e;oH!3r`8L0otJa<B!Ge
z<|Q%9xk21aje##3OPdQP=M3^*=<Ayg!-)z7Fr|H?)qv7Kk!2@aEf-eQgbr^?Qg+Ok
z*Hl6Pbk-Acp4Y8Bz-jGMBUTp@hoyo`-C{#`F})l6C?zUM-8~D(CM|+Bidss4|6g>|
zQ4KB~I-)BzEMxT)qkL)2w00DGIknp6s6z#pTCLqdH921&y}7g^^tLB@vgzQF?Rx$p
z1|D7N8j_k&WLCg$K&d?LmeJrX+Ew_xwXtS5XI5N67lYW`bDdxBg4B{cLifBqhuh1I
z`$BnlNW1QdCu|-=fKGqiDO9y<DoN?Q?89CTwmS=E*FUB3E6k~q5Z`yC@ygukfJI{a
z#`8O$D!yxC|Mc`?&Fv^vuf=92cI^|Wr~+@Y_2iiH+B14|3u3OjGAh~1AIMCx3W53`
zD;J~F7nmu=!|?Ql4uRnCmXDGZ^njc<q|;x<mso}&?Cxf0+2FK?UuHx3gy(M1THSYI
zqy_&A4*kZdf4@E%cwRJh&YYCj9gAVZD`$U!{tCLnZx~;1?i_RepJ}5`?+$f2N1i*4
zuh5fT`6$dA3@TgKLfecF1mLma(Jy>!+)n`-Mk(D`Drieo0?q=SDgpC#xvw=aOKCHy
z<0U3qYMt+L?u>=ShvL^~W=)A!#e)*&7;*DjCmNBy1BEX}A!uW{X!qP|JG#LzpN(7}
zpo9L!^=r%=k=&}-BD#689sLDm_w}^h*W@H{M^jurj16-Hs-NfShs22)kL|Xni_BCl
zSi)TjGEL0hq#>Ul+hsB{f7F!r2myK(!=pcirm}0FdsLLoWL-TYFqx2w$L2zRR#TCz
zYAA5c;G6C#cE1tF2~|_9K1V_Z#K|khJxlO}RiSA>0P{MdyX+T+jVi1ZnBX?H9OP|n
zg4|fO4P*-f1VU`;t9EzQOyp^Y^EEYkMj((20cz+eY$xChAR5yZ_%<GP6e+s}VHKD^
zm{Li3`Wxf>Ka1S0@LT|^Xoy8{`Ivxyrj66b@jEzhH}Z5B;z+tREF|m%*n9Vd!IgVS
zs`+MZ^k#YML_>=PD^_$_)p*(|LQb|XDf7eS1$dJ7O*8F`%+?}n9DwHwgFS$S%uI(9
z)D6cb)VdXOMQUO#yM~-~$zmollZetO)7+x$CYc*D`_{9BrT70y%zy9r$2%=wJ>L<x
z!n}^$3h0}AM)>4#?kgb*_<oJ2nNla#v-QORA4C3Z6#3ty@DIG|Urn<|hpYRHkv|F_
zowRbR{@7<p3se?HhIdfCQOOpbbWddmeN;&f-S7P=M9qrDO11r!uxmjloG&aU<qqJf
zB3~?b_(p|~SB3YM<fg>5<FpMfB@UJlFc(jE7+@13+4J+BpbJS(C&)^m4~6Ov^cnA<
zKb1x3^tA2{L6UX=rVr3*B7}}hT?4Z8`NU4UqGK1)0(6NIj&S?B$b-{GNSIJi-RFyS
zuFXy^V=c`-`PNU!t`lQ*%L0?4ubhYGMStd1Yz5jx_ILVKIZ@^rVe75(SOH+D+Mdu%
zUt*K+N~(Qqw6JGdjs}H`yr-Rc_l4SRxj1E~rMxQFBj*^iJ+T>A(kr!ul5my4*&$g|
zGH$(;U~fel+di>VOU$$8Exs0M`T;8B^?3JNGp=Z->7hu&EZi}T=>s6}von2vV`)HF
zxom20x$JrtZWTIG&L*``lwcf`)57f?B`QU?D{NuUiDUF?;r43HNo1d5tB+E&U#Hs_
z(l<B1Fz_bTR?*w}RaAaSw;H0Sg&qNw4XGYG`wIKfRa#5IaroaIerhxB)p9mf32iRc
zQ5HqZ1>F<~xNU^5jPpkUUHQb`@Ge>Az?I6NjwbfG$0MS?;tC<7^f(7VqdvpJ)*QwX
zZwrjLX9kqRi}2F&eh1yvo=e3UdO4!Y8etcAFd6|*jx=a0D3byNVs|cjdzI6=wn&kD
z<5h6Plrj|o2wAJoa~WJh7Y*P<Rh6r2=3Q_hC(Je^8J==LG2I%;KT<wcxOmiEL-zsT
z{#UpIK9Cp>i)`$}hh|qb=~_}b*(!&&WC2S>TSnqz$NOiAUYhh!JZBZ@o=LV*OtPNV
zq~wq_?IYMkeFUv@2s4IuBL>h$=SgWI=+OU%y*CeQDofYJttzKdlu~A7Xi-6C!VqRG
z3nQf@AcimpWC{`>%wfhZ2avfCAtYfag9JzzWC$T)umk~_g%G9?5HLW1fXF1HH`TYh
z&$&-`ov!oTbAEmL{%ZfhUaY;IoxQ)c_P3_*eP3ibeSt5Ll{ZR4c-1;;6L5F=vumOv
zA<+%z{g@-n00$AX(~xp<au=$fnwWR|Cn>T1>k(^*zxYVlX*HbPW7kC7;eC?$Q*_ZO
zUQOD&BJe$1n)(8}M&{wKR`Y-Fh5tcpA%A(%Be8lveUR)gn>C9aQhnL|;;u6X$tN16
zgE1g!dCWp<GvZ7SRy3UAP*P;kMa!!qsC(k`H8+kxAikp%AuXLG0L(Hykm`#eSTfH}
z8l5b%7?t|I==QYOiKQ(S-<wJ=Av{)fV97#1*l(}9tH!E?3sh~lNvSXK7^Ao)`auhk
zlAh!+Kvik{nQ-;z%e_3(e1`k=agoq+r)<Jq2E{O=7ZI%LKt!~GLeaf`iAp3Iu1KyI
z1cjjIyXyAKe3l1vPE><+PZI4jAk}aa9ihvv^#i8BuI%}XkWDFneKSp~Rx8&Y8@pN1
zLvgr@?uiBO;L1cCL0{)d5wvSw`O&!ZRJ$awCxz7l87*ASm05C3PH1k__Tol}DK4i!
zDDw{EQi|ud*BL_AQww?{TS$T&kM2_m7gK5}yXUtU<!;bu2xolM-or&Q%tDKDa}pKv
ztir_P5=gxeY+f%HL7z{UgFr>DG5!q?^B3#?+xg4k9IWb(@L-b?FV(d<M>Wl=*c``-
zNS^}e;&ag9uo%;B?c3g3L@!FnhpFU{2&lpk(=&WwYM#Ly*UF61F`6f<h<tqS-M;W5
za98lgn1PsC&#@~N<*4MGQ_Lp6neoJelfGQ)vh&-JoQ}kL^+icG-k`xQf68~}Zul0;
z3X*FkNyaMoFM6@x3|4Jr!Sn3uL|43btcQhqmt%J1avGAL0IB)5*761v)l<h71J~|z
z=m`|_xC_JF`M0<DEvf(d@V_q(MHe?!T5P}6^^>tPj6M48FHiJUl;4F<iVRuiviT0x
z@7S~7@@W4JW&USj#QyYx#}XGO#@DZe7{mk)@M0L?ijq+|4g=iA%$|?QB8=J#7r%dn
z?xH5%<yJ(v?aEnq{CImZXrpa3tT4{w%|)B5hdX|3bq<(I;K$xP#AC$xf_|Y$mbptS
z2aNEJkT#F7kt6x!s9GN7@wMRn3^>*d81D+St+Tlvrp*j$g+alL%R@ENvVrf$8KI3=
zWUX3gtnsSLSQ0y$D91~x>CvsMfC>R1i(;y5g1j4UTHe(vX8EJL)}~W_c;7kEzACMa
z`29u?0|k<;!AbeBf)*JZLRiA=mRsi4%;1aS?V@Lk=hs3?^G#BpSnI(M<{V$UZY<sT
z(7uf^jpW$O+CLBE^-zo3gDY+J$HcLcaX-XHDXv*iYNtz6-1&3}Lh#)&)kseL78y$^
z<Zj9-7K!WLGs_u|ET8oqAL*@RS$j)ncB;%;NmmcH_GTBHlQWF4vaohzf&fbu6BV^S
zUyyN%WZbbF^7^hr&7?jT9a=jhs-qufXr~4|nH;2nm7;|X-Wq<Bsm4W?BpxRa>Kzd<
zmJi4E;f@mRquYIiA#yZo%X3xMw0*o@Fl64b{%YsrBnRSy5_hEC;P^r-Z05a%ap&Ue
zE{e!CmzGXquZ?l=_{u0XNIuIDiMw7x;0!+?-|^&*q7Ue8o=I8CxQK302q^GO5Zg+o
znMWhB16d##bZqU21Lp6y`5*J7q75lgkK<?#>x-AH#(gDy+{lEAA!!G!Ki10ijlQ$q
zG*;7^wx$vHWG1Jg>Um#5r#?zzVx(@UMLNbwJL{z90{3c)^yNs%Qh9R2m9FufjXgiJ
zP$bZQ#y16P0Re7Ex09h@C?u3sl)D5Mv@GY;?+lT(!;E|*ie$Pc27^9md{oYTRH*~s
zD0O`Gn!rgh=?SqVEd-+wy{C=Vu#d4ENOyfdxIZbauM8f&SV4AUkIfHkeHUT|*<?T0
zKV}+|+M=$?oyCtRR&6>_TC^lKWLnhpDLPc;T&G(tnp07Y;|t3Ewyc~bjr`yS+o53#
zKK9>AC{yPA%08o2b+PW#4~;sK6nGCkRGFSvg3;bGgb&cYDehvTufdFzAAU$cww=yT
zxYpAER`HFFG)^Ho9`3J+KJ}G(bN_lhAnm~qZg}y{?F|w~NXIsa7n2dRv6yL#-ss@b
z(DcQ;vAmiBA&tLp{$I=eZ`il}qlvYPT~YKTzBWxo`sG)D;sE-hu{M2An47k+b!p$R
z_QBgf{@U~W_vVXy$svc9)(G+m&CQK+!>k2f@qh1&5htUjE39)X%`9D(#f_u0Kl3xc
z)c(}9K@9&fENfYLsRi)hPj>x+YJ0qwVtXp~!+zoZLyvF258#*lzdwxdcI&br>erbH
z9()Hno<thGjncYszbmhbS)sw`w+25r4k)kbu7mcJR6D(E77t7hp;!{?80`W=iVKBi
zhXJnkzOi!tO_Dy^sME5``z_U!L^&L?kF`_oj79I$k&D-Xo_fP~DY*&O-CJOzuCP_q
z>mFT|VF*-6nHd7V0B_4N_JXtwRu<)KNp(ksaS0_$i!2-$SLT#9J5JShDtgIRBH3RK
z&XBE%0k<`6)$+dWa}4O#s>A}X!z&WeWbeE)>w(yZ38LdI!PGf2N^7QFB9Q5xuU1nN
zREZHW&O)4Rj6-+mK{o8d`zaUBeU<3w?;mIep$fBTB{Qw(3_~}f6!93%^=@!(W2Pwg
zak)wxyQMz1j-=+r7+!vAFWuTd=Zlmc*qrR`eyh&x^{#X=wzN!G`Ir>8Y=!qPGtO|q
zqoCI;D=9|w>aWJQ9evYRpI0lBK3Z5act4ML2V#M6G{!3_^wiwgt@XhBp4V*#^)z7v
zo@kAA4J_(~dp5!qsN!0b3p*{-JrXY%TQUdgnY(VBa?ftO0XT@-@FvTK`R(N-%W7MC
zTC;|L-HRhnN9KCHI|tpt#Ji92HTKwLA3*5`*?Zm@aJ%dg12!&#DZ-^?&DRfZ5=(t|
z0+u27(CpL|qk~#x>fm)`)99ASGaW6iPQiB@!8d22J#E*Z)ysLar4=AeHGE;5fqBZr
zu(^L(x~^<22yzL$6Ztr>SzicoRYtrO=ct=F4w#lEht(>4Q|r7!<VlAYpiq9YHgAXq
z4GT#lb@pLZzD@)~HHVi-c2{t(nia(QW0Fe=j~8{l+?!eB2Zi9gp7x?nTj&UF$>TA8
z3{a=nHpL2N&FsqoZ_T_G`{M7c(Z+1rblMjgx$v;iOP=I|fF50X_T8mXA7%S;{rh~4
zfKqH6z|`EgdQ6eDtg``dJ+0Rwmxc2kdBG2{j-O~%Wp3nN@aK+3wl$KtbouljkmT*L
z@`~IfB6q~bJ3FBflZ!|9su>q!Q^VBZgvWVKoRyT63~cNKP{w$#6wd^S2CkMl@u!r*
z69hKdni)hd8cX!X?wu6zRnJJHK|<&bYs6(j9k1wyMk21th|F+Uq*Ut;m<_axNQ-2M
zmjl@bnNs3lj8$N7mDgl3QUh~XLb4XT=cBlhGN8QhETAHO1RE)Kq(CgOe!z{F7@BGy
zzIzhXXldZKzjt6a8kI-)BEw7JN9!ZeAR388J(9@VI*A$f{BSP7M_1KzIh9LYZwg`R
z@nfEcOFf~+jW<<3ANtz=ONCk`8dfMW;V~y3t~HcY!<$4oXi*?aU39=R37O$^O;N|p
zR87_OZ$LmgF9@$@EwUZ237?zMoXHNpG^D6GLvvoHh=@>3+U2hc{oy&Wuqfl1jQpx;
z^lE%KGHT^<<N;!OJhCj1tr>(&1bkS%OQKiS^eCU0v=svne3+(}wEo;e^^qBN;PjT+
zmQU0Nt81$syTNimnPe2Wjx{&8r@4d9^wv}5<&<Bhqa(&&riYP*f;Kuxbb=gz#N<sh
zZv0l^KkMm!pNbzk-8{uvJs6pc6&-CF3;JFc_=^qC6t%gOX}x~%-->4cYh~TjfAS80
zIEeV_a0m2!;&Bci?)OUS%v3VqDjzqO40t81H}mioKkUOjjxWe3K9#hIBN0eQ{(dk&
zUV}(NxbcZS7D3iMMnon4elVajg0Fb3vG>cG7k|l%12E>s1Cgh@(T_;07TA&~C)R8v
zDL>iDETBiy_(ydyJHFWG;HWLK>u^`D(L7ymDtRH2797|j{-Vv~q1{x<Aj>Iq#<@39
zajd~<+Gc`s?b>x<M&p*es1(C8GV#L_(%NmDU`$f8C#41oF_#`bF!bj6nddB=3pVX0
zMQyzIw$nwKO(juA_yjdmiG+J6F%?b{pYFKKLa^}a6t$v6l-`V}=#5dTs(Cygkg#0n
z+1-WAydb!DKI3qM#zCnDyy0}tzN+@Z-6GjkL;8EGGiUVI@6yYlUHkM6V`WWp#ukE3
z!tRoSwod4U#UgyhaG_ep2Q&J1d?dm5?QbY&_^0?>Sb)fsjy+J3fyUlA9C*)2o8DKK
zDB*IDVk}w?eZwa)@5kIPuHFDedfE|G``dh&P#HEasYQh^qq7)%!IMq9vO&5f65tpY
z+UJ1o0caicmQ`5EvJ0$<SJva(7LI(#;#tJE<h|byG>!>u!Zz(??3hfsEE~=ZSjL$M
z@F5U}54g}`v1Uh#f$Zt1jE&ckGcj5*+Y0M3%`eA;e=qdA3I5+`1-6S-bBAR_wr%t-
z)cPG$V1m?CF2oD8=WMv>UFGX#^v|2A1LqdUX+MVC%lrCAv!nbd&Y9Ev0p1NAi~L-5
zN+hyKJtdNz&8JIRtN3no?BeSn&Jwa8J(~{`>S}~(L9C1`jr#5OWT%?6U$e$WC_$eO
zg?9~poO#J>v+<kHU&zj<O@vKk-aYzg9$x>!BI)h~hHFe$w7Y?a6h#QBc+4a3v^-1I
z8#237dka#Il?s^B4B2@z*O#v=h3Lgfbq6L%1x`qrO^ba~)xc`nYU54lTR9Db*nM@O
zenEaPqnS@9LZuoK0bHU28B{OKA|&ctfwUQC&NzD0<x>hAmmMrJi@dY8gKz;Fm8){a
zn>+i>Y$2!a&F*0xFkjrh&u7@$;qG5u|8GtF-ckR5+fhrIHmaSkw;Qps+%m7AKY*BK
zE6*3fII3!0;n2xYQ$vut_`7(4mzj4+MX8`B#gEfPSgiVnjZ@hT=tFI07)?a$;`jJ_
z;b;2ulh|?_FAeGfYs=qz*A35CmoD>eqqw`~-zHAuh@q-(zHH(w*pZ~xU&tK_-iX@V
zW>*%CQ$Sr$t#+v~evO!<*%W^Roci#!;UwbBhCfn2ht~c<oQ*_Wo)ZGUcOmua=tN1b
zK&@XxzSo6z`mtl5+t}D|(w<bMjtAucdgCyFiDDkEsDxSdfm_&Bg1-V_8NaeG?0X=K
z!1`uptb}R@@7M=bOIy>z+YLjcWOtrQNsSGu%&rDI57j*@?a_usng9GUvj&)F4DBS9
zG5o<@>t?wydO?W=JaXa)d2^fTj6u=~I6zWR$fNPGN3bzgb~ydCu709Fn=nyWJS;t5
zj(kIzJ!b`A0cS)v_}En!(uJHB=VyUplf)7yBr1XxZkSltUD__|No7&o^*A{%kBYY@
z+qFP4t0+S~S;fzii1Y4`q)FG$@WGffaeR(FJukJ<`Ii3SyVGN@-|Agzk#g`2TJYfb
zvIT7hN!gDd<LONm@%g0p5M$^4y{-%qFmn=%rAV;WB9XTIaXy{0m|+{sZnRWBC;Q=J
zTN}+wCT)Sh1bja9)Ch=rM_Gst?8z+~s|WhNsZ8}4rDx{Mf>TvuJ`qy&D#BZZ?PgYH
z_<vIQlL&)%{VmSYjzpa5l5{3WxZMVKd5q4dl|D0(=$(B0@#9Uxg(p@9%T(*`-bzw;
zU#yo^7dC$p+c@_5kZc+oy~HFtpIJ-Gb8*H<8GR5$MBBdId$c__OIdC!nGtxIcSV-Y
z8ztK`T$qzJ%KNt>57*Wlw%n~(<6Qdw{aq~0Q)R-BRhuZ!$9@;FT^3bI3U5)LOKV@Q
zM<baP$G6-26bjn~jlLSq7g@h@-e_;AafE5)idk!)*@D;WAcr4D&&*{E%*5w?vDBBi
zBl}}stdH{&GEbvxduDaFoQ<RSG@Asjb#Mb)`VMV+wZ;g(vAm*Ff!47U_uO0kv_FjM
zaiC1e>TSPYubZ0tW0cl(Q~Y~{eB5xoWKpu_LHJ(mA{9~Q@F6WH`eUS1z}nBo*i{$P
zk3xMtKI{}|o5Ve%*eju|YMZifvtE-3V>P{qcQW0dDxPhB{kuVaH^}cT<o8eJcOUY*
zQ~d|MOvL^rl%R{^>??!o)P;5I0HIkwxDaU*i~%~?g+!jYs(Eq{a8*=h%Uqbhm?D_?
z$)^SX6o#7pc~tjuxx5mgk^Bq>vNHZe!o84}S70S1r2V{~66TnYkm3$YNSG6nOmzZV
zvJ;ls8J}ENBtu_6xMyG$M5Mh>nL^3L88~A`%7$~HJrOU}S_k+8m0fowuahaeDlmSB
zI55Q{Ut7|M7k}6C>~Nr|Nd5S}n1Q~(lw&sT^P$7xp6QT@nCMdJX9e@PoZwVgn6dk{
zRtCjKYJ2VABrr9&vwZY%cFsG0-XhV7I4iR-d>0orw7g)iZ+9@*?s!g#vlsc2M22=X
z^19{2kPStUJ}w6?f(fO4rGyev25CBp*i)4USRpT=cYjJMj0?5Cty0UUKG-<;`WsWm
z8;W_C#O;vn)%J%=Eox4;T$PXLwQMNpnbS%q`ra+SAJX$Cx9Y%rK+I*%GP@|<Iv9^8
z%*I+0;6TzsURif1#Sf7b0@AvUg$3F=FY(HMWF;LX`JtSo{DLLDGVSdKS&UFg!+xEh
zYO`$r*J0Cx?lUkgt=>Je%SnMtbN38gCJ{HbT&azqXHCPm2Hxc3YBo0>;^m1R-qovd
zf_HlCzYD!KMvuJb3rbNnJ{SdZGs{EI&c3?aG_H}MB_7h)emd33ael30yW@3Z<vq*x
zH;q9PX|w=??OoddEB{<L!)&OS3c^#PazHE`8ZI8$bctR}*0pR~WC&%{+eY%(a|yF$
zf+k7Keu$38y9tihtrlFp0Zp&knh7!5W<VB$!z4LY#+@6zr5~3HQ(%Amtur@%x!<E7
zzIss?5q+;#ck)rKe{If_Q$4?!@38kOW)ZylhZWUv+tE1pqw;rtvDN=IL+%$R8t#0m
z^-Pr>ivc~U*@36&qcB!})g(=4>$>hlBIC!Z_kotr({5`uifE5QXZv04<^5z}FF+=#
zMqR<gGCUuIVB<(UGne+AMVtZkMq^ic1<GOLl;A^0ccmnU&FUnuL(jNFApo`R=lA9Q
zm@*Y~Fa;OeNV$#=;|^4+&Gd`TX<ZQ_d+QawUW!S0a9v)&zH2;UbG8)hlQBej<Q6%&
zJ%xp632BFNaez{YwFRitCyyFj9h?_IuOvHUl;h%YI*zu%xocMzWP)j6`m7Tp8#zs`
z8&@iblL#toiTkFD_>B6N(UocHQ{(vjwnLfdLZpvhdFqq<S$Q}Y^Ia{3pR)`v_k!0_
znV=;L@AX}45uXuQy9?lTPO=-!ghP!M>eDmQ(LQgwdpc=RTDnS{@zSoXDipikMqnY5
z_^fo!gMWfAEiKh&pbsEBbmp-27Qs*sr`m6+r0PJ00LvVSc=S9m(VJ+*v;G6W-hTTb
zQ$W}mCPhUImR68=)N?B$?-gSW=#sml9+lRvN(8Xo=R<PDxLTaA&|TDKRHg}_-&qCY
z;Lrt{xF51}GE*9Mkvrh_Tz9BtObRkaZ|a<kJbLs!%V5@M25I2@4CzA+l8YRw0xb8#
zB*g9OsBSsQQhVmLS^-sesp(0v%!SJ|K1kelutwGNob`Z~`klGy;o<6ZL^;fi0ux~X
zwBeMXybyRb8pneZ5J^vA*SakpkIQ{3d%t`k($x{sqrK^rK`YaJR+MzR>QqzoD0;m0
zZlVY2X3UK@H5(a4L*Oo*jbP8)&Sii>TN}Q_@=jDt3><WXq0Uwt?+h834=oRIVb^Na
zBw=?xA3C9X9rR|bK24sj!|}#eLc+><@D?<GhM8h3*&{Ofb2o9wIK}@<c6Uw`unM4~
z0yg4K-8;J)wwwzu<ui3IOPx^6H9f(Gimr_1Su=`VLBTq`#-ZI-=s_ktoQlAy!C+!4
zPW1ql^YB~0y^0R~KY*w@w%w)|kSGpFl)8{45Ev)r)^eQCEE5k&)|g3yq<*s=z~!@`
z2DB8!o5@DY&i_{1{L_(>U*+k#(4^98AoUBi(;Gq(8Ja`co)KF)eoxo#WH@G5WWt0u
zlgl#w7zk5gnyo-Twj78*BDMX(T@Dx1c7l|bg#*t|(Q>4wQ151Jh0tLF=ysc2TlB)v
z2X{4&h*LX5dly(fyRg#+5$@5kO?vLF$6x)}<?9HkvD!$?@-UQ@klgs#fz0q*ifYuA
zoEh%OS{RK15SeU*TaUH1QME?JG&R*6E5Uu9S}qP1?m{Y6TPG_d6KFl&gMQ)6sG6?c
z$iap#cy{?3-Fu{;RX7NGNQU3<lMm^r?C0#!keML27jNA)jV}9%?il!9NRzN@7jrh!
zo|+h)0=^#GGaMbUe#hgH3un05)<IPEM>18&22q$Blz5qVP*^f+d(UbxD6*#aItsE|
z4KWTU>a><G+V3XW_6<>#tHX3}y{V8}YS$+`&UZ((7Gtr5HsfMKPojR`Z6^KloDtDk
zyL{ls-`X1fce$vE;VtQ6micfla9lLYNsiF76P#Uw`dD%y-vIU!Fgi0oFe^|P9d}(1
zoa@>HwBOh?tSOEZa(!NgwKA8&b!e|~98*<4yz~vskB=&fW4)AApRdmcH$H+d`1Lzm
zU~?^%=_pE3&l{ZIg2DBeh8WWmv`wwZR^W0gS`AKhT0t$2?s_>eHhO@x0_89Fb=|Ne
zw+T>1*~_FTjPGbiH4Kevk5w#361_=^*@KgQW83LrX3q>9>!g9jctiOG;%(||loKw;
za|Vtq&Q0z#A9AKgod(>9Tg_6h!5B4$-v<X@yc$AbzLNLoxb;d?%~Y?jCHGac$@%5w
zb0~a><1jF`I^+4QEmA2tx+^5q|IR{qZa9(F%&gfRq34O%DiS$%STvFb66vvSYp*%w
zBN#Y8hNc!Vb=>&DZ5|GZv``+?XujTUPA!lfxB?Ao9ZASe%364gO&z9Q79RiY_3>X%
z!2ac-ZL1i!+_hNLn9&RLp1_|3432J&nSFo~uaL(v8>lhC?|%CO{-cs#z)=^Sh`mw`
zSA2qHayd(V_Vv=uc8Eb1X}!R-9Jo_T&8i3)X+4bQ3&-tbBR?YS@s?1vm{ViJS)92B
zj(bIBiq53MKuBQfH$-RNHoU=FXS0cO^%+K0XLR;j@g_3T5P|hHw21CoQ(dcUm%22u
zA#1HlixV%u{h|H&ecqFxyl?S`HClvA&d#ng+>gsJ8$TcN&v%|1vKY#pL4{dg8U|xE
zLyv|I8{d36TDW`Bu3Z?Xd@9xh?@7IFnyR-ea;oWmRx0?)n+!Q8nGgCF1a2^yymuZJ
zSYA@WrQh?nh7DSja{0Oy5BbvInu=_oIDZ`##8{8in@PJS!y`u3vYnHxU)dMnapR&H
z%&t_4Ckr9N_LOsSefRYu#rZlML2VwC_Bf3xK9C^|vNz%yzxL(bO-g6x%&}T}7ew{*
zA!ij9w!>!rd5fX^@)ZS#>KKTpiS%)WUHJmexMo&qPZ?H0a8tzAXf2eOZY3Zh<hOBa
z#!CLV>P*XUCa)6SYcr;Kbh(Pj@W9~%@Rp_!tpC>Iwb?&9ljUa)XJ1IOD)!yP_g+mU
zhE-Wx#2a&vDn*`!ZO46dj+T1>%pN<+x7@2lkS`T!c}6jdG(GwI)Tyg(d~(ec>FJKB
zn(>QQVg7Un)X+9lSWG!IRC_d3IJ72IsD$JJz8`|E>`E1{6r}#6X7azU|MN*jiYJwV
zgLDnATd`6HOTz^36!hwB?IQ=3Zb0QQCd3CSW=CNR`i+sRqT+qEFg8C}bP8%|DCm~R
z!IW-jxC)Pb@0LkdFFDyf4rco&O1TXJ>I?S;&smJ%ksBc(o2B;)gS{2B3KotKl?Tr;
z@;fHU>@?1CTZkAUx+O70W{!QoAIm9)AM>-PF$>h{cbY;0S|YIPMs2_&LR~LWWU7`L
z47?d;E=o$}^O<X~hTQ(M@ql0QqO%&`jIY_G`Wy_aZwdOE4><MR8SJ@|Zh`;|-sp<V
zcr@?;I9B`<mu_!U5A^P;2`hEwY9yV@iYUh>zxjZDs97Vs0S|XfC|Jomw;F8GwjAnj
ztY9^Uf5M^5sScBP{+b?lY19QyZ)pSX0`%n%LVXx4`oI_fF}{JTS}DkS8LV(~Xp^(E
z(@*pm8PsqqnX|SVG%S;_DNaMObCM&~(f8K0l=J<3Y;s4Hep|cwrPROrq*PF8i03~5
zr*bXOL;E3pqrqV90*lh>wWEQ(qa^BybD%u-1@TK=@e^^Lu$ay%TVNi`v&vC%QC*kS
z2p7rgxe5;{d;UB};!z-xV~CQC-7;eSguhNh^iX|2N+#|ad-{=POxjDwHhj6nf$q}2
zEf(yWh{TnPZ12EC2?uwKPvw1jDv>C#7S;V^p=2pbgY@_pvwn#RzZp0&p>P5`iK^z9
zbNC!~u}w1%mlkWCHexHblpm6CIbKfwICvQLNL#_pwdZ{v`9Q;ztVFW9i;}#ljDMFz
zqJdAyVmPw>h1wr>5z|)G-P2RG?YWm`1zy@O0heYnma`eR=Jr*3Wri?T0dIB^FrL;-
zaMk2pdTbt!0A8q<zeAmo+s-PdMqPqmY7RiZ`{lX*;qlMLDSeS=t0)AtMwrNB^Ll_J
z{f||rQ@Lu;=NVGrt*3+&S+EvkFfmi`^P%ed202_S`((Wk11%(xkMWC=f@|!X`$-OA
z`ps_iQW5yjW#iR`4g9cT(bGwqyVf~>D4KQ<YCZb4Y$Rc`xEf+IsNtT5g?R*?l<Q{d
zTo=tGx?2euKLw%5-8lE<YfKF6-mrD&k#l19cmAQ-{?7tO;<mLUY6-7iJFL3iCwh>a
z)~GSxS6UV}BlK>Na(%lc#DGvUsnS61g<Q*{E+uW?-uNnWj3z<ct@-QffF%2uuwW3S
zc-^ge5J`cA)oux)YoDRk8B0s7E{5*u6g1A~@tiQ6e4D~noEs$3{kq#766MwN&ALgY
z{#~_lC5gcyD!V$+zN!-HJwbVF_U3Hfl&#{*YY2QEd-udF2mjOm`Mm$pfB)jRVUrBa
zS#V>9Cg<}Z-B}i!til14J?RLeuIv@c`cS>ROV_4>#Fk;t!fJ3CUOtuWLvbu2`IL(=
zp@q=fx4*}mM8COYo~ZKG#k^bUQc|m}X7ELAmida9Q+*{*qI=_GNSfkRUcGOR^~+mP
zh+#t~EzaRd*L)LBcoEiC{k&cI?MpLHf-+H>frNr?%sA8Fh4PF4ZSAlBFctiI(xuL3
zIp|fLv`7UO-NhW~AJM^lzYN<DrkUB*tg*r^%DTdeJzGpphHkPMy>;(P+45P0si!NA
zCmvsP@?}yNPt60(#?%&T+@tD42Y7w+r#&&)^)6V@QWBdO65FE2bFb@6H86VhR0|}4
zWqvbOjk?tY9p)@`_-uN=9e|f@XfOt0w?US+Mh4gZqgL>jYyb4Pe641VEXRH{Dxoc)
zzL$uVbB-D?zR+V`?SB9Ad2)uCTS><m>&~{z0nNU3NKH?WJm6KT#*{&pevAyx$<CFb
zK1dt7Jh-DUpKT2%avY-~{UI{EGuGhI@`D1O01#=~@GhC=Q@aAvVre@Q%bm!HMMpku
zrReRMcGYWlUq@jI-!rtp?ar60ZDbwfW9f1YH%k`HybKV=CEw1t`KbNHPQUg1qDVeW
zsrTioRGsq_ksqUP#u_3bzGi%QAYKSH3uyF+E>MujFpyR`)7zrigHaN2Yrg!U8Kp6m
ztkM2<D2?ebq{?NBt-g@wXgiHdnUCT8L}!Z=%lQbEIvsMZgZ>%`XS6^d&0&-fcjo|F
zCprCVlKv|!3;~<n@|JRP%~mZCJl+_eB;7RFhm*g+%AVt0`}h{nL)QJ{Z+(M9|LUtm
zR$E!ptx7Y`ci<VHyzXusD5*@9*BTg`ji(F^+a}W&O;T@9DBS3X3``!oOOK2g%fzZs
zhI4u>g%^;;L2<S}QJ9ROk?Qq5fWY*(e>R%vr}ycrH|zYO(EUqSpevtV{lt3J)RFmP
zWgT-@`smg19DIndy&Ey<>-3N(B^&qIS3yO#2Hv-4Eh~DSj`;`|bG=JYf!1+9{rPtP
z@#&x5W8Xr-j<VG<H&ZCfRbBkG+L-&OJ8$_ikuGWiZn<VkD?lSIakYngzFl1P6YR~P
zf%k?MA?_M(@liQWDkg90FkqRwa5eCXkgQ>1j!a+Ya)e~2q}Mm-2_0x>(#DjPq$oGh
zfN{pgGJsZ#Vo^}?obXSh7-Z^Pik5|1$d7MPwm$yLjy5}kcK96=zd)(k$Mss6qVA8p
zM1%D?hK^-r;v&`3gz_eV*b?D;`#-kj-(H2gJn5mD=eaOcTT!XIsLc~GX5A9OH$EP*
z{v>yw5R-rYl}mJ=$a*%PfX9O3v@-fYZEPz4uw6?^-{tX=)s~{7q-W16NNcb6AeQJ4
z?GGbkZ9{038a6ZnWe!|OEF0?`hq-(2-jm!?+>{8%q46cNFjVo}EX3#R+z>m?aXsKI
z%8s$YYmc0?$=Ut1gE$41BKl5&!m5>vmRYpe{iHy^C>wx+P8RxR;UzTx_62?|=U;r_
zzZ3oHUbUy=kri!(a*9rYI;S4g`T0;xy^YyY2)bS;FEO-@8%V;&iO(F-5jmy3gI!$*
za6Qda)#8j#|50Ok`h0y(dlQd`vP4Uy#zfosoGzX{ISZqX6G56}HEwHD2vAz1GjQEK
zRXnDXyHTV%a$(aX%c-MwJLAKBaboKq^?v@dVZYdth|-wU%(_!K+fa$Fcroi#9bSu&
zYK);+_Ymmz{#nRTF=faY%U<rX!fjg?nOm}WdBLMqphM)WXAedg@anYQ(RWTjYM|dF
z`gLt%z2xzd7iB}zik0x{7*#2a7Bip0df`xgGnmTLA6wq?1JqR_(=lrh;uDoSnB-0e
zeXYPu<)TeJQ|cdot)~B7{y%@EU=}{Zi<e(f@H42UO0DTiDqB4FES~q<q6M8VF6ojW
zjlv*N9owX=YgSj2x%^EjR$17Mr`eE<N6lGz1QyWutg`$xeNS+@m-h37(o*kb-j3V>
zYKZCmrsHSQ`!Nc6e%l81&L`4%F<m6|^5A+-Y4=Tkh?r4VB(IconZ!vvZ;5KGPoQre
z6#ay-ygl?RP1~_ixE1E^!6M;DXxgto&H*)5Sv2jnB03aeLo6oJJh5dvI{BMxWZWZy
zLm>itr*F&)yKHZ(=}6$Nrl-K0{J(=VvrY7aUX)i&+$|tB*p&ypYtG{$%Ae7o?!t*B
zR_?>8KTkUPC*ak6S9s5IEefc*&KXIi-Vp%p);HA|W(KQOWej+%SMRHrka1Phvn1lO
z4BF(n>Ijj8o7QP9_0}H~Ey;b3v?2DQ@-u?Z$xZ6xBG)bT%GuB}!xVG^k%v#0cFZCZ
zI3{hy5pFNczx{Vr{*RMi@Xm$3g`18a)Ot7F_l8%TtBoA&?g3%^&Q8bq*OUz)W<)?N
z4QKf`7@((WysG_s+sPH?i!MCR97L}oNCeAa1nP=S<X6RfKVV!@c^CWz-n6iGp*LTW
z5KJMnu+H+wN#T*GQzA@|Zm6^^a2AL!a__MY^BpIP&XU|{B>6GXq})16^eHh2l^v`j
zkf`F2FgHjQ>LpW`yRY>C19?#RF7snvAvu)U84;9gfv%?dX4J47xhCw8t*<jQo3cD_
znjcPjbgd*vS$)_(!)ts^?QY$Tj8sj}@oEdtih6`{GEHyyFo~7J4bq4MpPg#?v0OdS
z&}Byb>{gl(Rhx0WqJ9HEmf7tS$_#4DjfaS?YL{J*q;K0gtECsZPkDz^Z1*nN092X5
zxxv=G?VA%1;%h^UwZ3~bXJdj-fh3|}XCPtAJus&?9W>!xaau+?oV6uqoNOE9W+*3f
zK>d2;A6LU)O8jfbcz-FS|Lxd{3)T5_)Q7&>#L2tP74)k|(;SM-vBLBQ`Nav}E}SA9
zcwY|?=SyEO1V+=wpIc@mYk3<NN_(8SnsUYRxVy?<OmXOxVT2fc#U;Nk#0L@2CdW{(
ze5<P%o{M?zE~7W&+hI`sbV-5`7yosjyO{ami0R>C%w>e>K#Qi~!#9x@dEme{v6Fwj
z*RLo3>AQJ{jw{UQG@oU>{_4`eLR0FSN>14jcb}wW2=R*!DnsWwbv~`L$X7p)z~`Gk
zB%Nal%+EqPBydz&<lDaDl1dl(j;isC?4af5os;Xs-Sh17iBZ<{>@(7bupk6zk0ri7
zjKNlIBr!nrEt&L$UY|Fw-Vkt?hu874Egp^(pIeOiiYt<{GDV2+%H~DK$+NkZH<R>s
zF-Hgck)&DGF(!b&*y!pp&go1V1ozlQi0n!#k;r01mGPunt@%M|qdrNi++ejFqz$a;
zee_mFIYkuQm#KBBDW+<}lg1Auu7cWF4YoWCSbqyrBjiO(8ykD7jEX$XC>pTLFz~u*
z=o9R>&-sb#Sb$uVjD5$V&q|l>E7%WI`ElNPO~Vrw;FYn1Jz4;3TNIDRLANf3skji>
zi+uWk<619*wYwXVPnk+?%AhWdmA-fPu!7YO@yA0%y3%7#Uk(i1ZouF1D(9=m5xiDA
zroZ|1X7IOHzSw0026(1JIs2LG)SH^2TP6#qgJ~VroSYKO<$`iLT$#AuKlAYH6xWl#
zSn%e)AH0)e56ojDsSKu5d1%N}VTy=mN#kgJZEp(;*R(ugFp*D14?Rk7Mw3gCSmBY^
zE$szla3~2&s;5>sk&^W4YA@GkVl(*QT2=h0Zka>5t_7o-5<MPc2ySh}=VF5JcArMp
zAz>(6Kws5`Dn8y*z8<55^J5Wk<xV7=BB}W8%LheY8#T6cS^8+r7EZbKM3oFidGK|P
zvw7?jyB4_|Y<dS7LmX9W3MmQf%@lDe5y3X^TdL&=w1gHF6c#o65jy;e$+Pi%#bq$h
z@_E~AB^6z|AbI83W)nOt#8_Q9&jc~c^iIk`vJIAXM1~<Et+>TXg7>fw%?2Nj-e1fG
zvrUzlIkQu?0=~XZl53%-XEJD7yjJ}P(f!;~sfcWn_TJ4|IfOyK2*-t}n6#xxz(w#^
zCP>8oAC&2TTMNIM^DpixOZQ$xzcT1~wLFE&)b#iU(bID`&ZQ*oI3Z4A!29`}5mSIR
zV!C3cej>SA4)sm@UOpmIN#zPg;817uUri!5^By=kceC5*W>5oE9wb-WLpixuV%I9F
z>1jb3M#-?PG>^x8`UBSL$V8Sz<y736o;aM%nUIki@r9?3eEZki{CeV_KVm#|e0|bP
z0A3t;s4Jb}qjocnS^N~ZclTp$&HG|+6&wbmfq7^*aJ!d212Ms9+g${wB1~qh;g=?u
z@)8?bWkI|!#kWq+-&4(cvy9y5*(7L~yAK0ftEuUBP*JJFZOo^H-XOWD!wUrItj6a^
z-OI%6wPCZyM_4Bo<vKhfy9Sm7Kh;ndQA=$yVvya!jCL+BxFbh$9S!2HsKwlrU$G$t
zsa)gFcRy@v!>Q6p;<zIX5TtaDo(jp^3PIMEbAN)-Y(ube5*czu&959$4EFcW3Lgl#
zbC*Ze^_)HJNjr9tH*N=X(z*anO0yxWvBwJ@2%K*Z&5_+HY4~yys4}%G(^lE68tY9^
zi<og&4&wPHF@sXeW$FochI74gHpu0nDJQ{g)s{B-zzK3rWmr(au`sR<Gx%D?xxlG4
zY*8J||CkA1!)>iTAv02i$>U>p;k%iZ`peeMlA$NP@Gl0UG9l#KKxR2jzC5EKd0gvs
zrk78QN=HCzvdPHv3cP3YB-~6?qjf2?Mh7OG5$P0oKibaqZM#0OPfAa>q0$fGhf?~m
z=3E)|)M^g#;jN}WABkic+?aCx9xia@8qCsBR<4J<4sLfC4}MhAJ64Ye(_oFJ{)>8{
z(7yaIMi7qykYS6-_XU?@H^~H2Nj}nZic!_%nXaj?RmIx`a|7MgJW6VD+t5qxeCeb+
z*ux>8KmvB|@uqxPVS#}5iFUy`rxQ9U@^r=W6sSe{FQNf-tw0iy<khI%4bRTbN4g;t
z##WxMrDTA+EsQ^9FJ<Fx<?fN(yziw2W8)8wMC_>5?;o4Yqr1@UWChx{+vl(!ul&@w
zP5OMuV@p5#-aFfT!s5%fMjHeB13!O4T<MkZdUJ4pYnHs9F+S`1`Opu}u}S5h4}H5f
zY`$gw)l~N9LwXVU7j{lg;wm;R|LqsP>VvDZRRrVQ_oil})8^G<kuOJ$dl9mqMCRt{
zE9-*BAumH-m`J3p#sB{EzZ?JeruX}Y_q*Hq-Rb`4yBwDR3!84H(4BNkjC%DPB<c92
zj*Po?5WJ?*J)HQZSC+zot*O$dA8NGxf?l2dRAD!u+7Ii(mE%!~atQijZyPaD&TtI|
zgFk+phC;X>q>Z$7U>_$(t!Wt%<d_}UeA*d4Q}w&}2JfNnEq^jb$;Srghlf`e!fnly
z&y3Vm7{G4b$dvs~N=X?t`NK~X$KoAd*JdTwBR`#se7^BB7A@k>=2<;`?G`7O8XRf(
zio`vmx20p)grZyQ9g!T|LTd1Msq1!AX*Zd1MoSb>3U34m0sNm@D)t^_jPDe7`f+!o
z9J)y5>!XaI1;T}VHA<KvCWvae2OEX}c%5FG<POqAYz4EQy(JRfiFfV@@4-mz7^c9f
zw?lzVI2BjyL!0rkrxgGWIYf&yB~?6&P+9b)I}+Q~Jh~h0A8K8kx}W2^Lvs3h6zD${
z(0<HrdZKZVgb!OS<@m)ErJ)T5H-acJUPp6nmB^vpwh#lv15853H5kMo7mM?O=TiB^
zBj^Q+RT#WJg$mGr%>;Fp%`NCKCO5@Ls11GoIcV@?7d0q>qvh5#;>hxl0@jwiFe-ag
zWLkBRMjM#As+I)wIf~jW+k&p4#iOKx3vFfvyuCU0z6`IMnYsd%ne$5#VP98y!XUMS
zP8xW);Z37hE~HH+edi@U{D(T9g74bikPcF+LY#R;JXO({Ol@0t$nvmHUoU~L^&RTx
zeDJKKM`Re8Ob%yUH<kcZZyp^R)noXWwIp=Wu0b<No}ql4Iea+~eozKL40c~@Gtm+C
zz4jUgT%h&RRuBf{TfaVU_?s*L;{0_sf|)GL`EDsGEp~{IVrA|Vojy~o<RE#-eaXiq
z3JC1zSx=BrPT=N^?3<)M?I2>ifbqpTq(-yei#EOUX%iD_5SG4#=n}j__N*#1BNC^C
zw~^y@`Z+99UWeYrV|@6Zn^!K>INhiS^LiMhW5wra6d9irhH9OG#n|DB$kN*jufcs3
z*$vO5A7cLa7wc*Zyw`nC;@Xs7G)S_746xM1PL;hUgIr8yHz`Arei2QxN~lzg^(!OL
z$)JYufM%7zc9wX^TRwKg{Tr7*2O|9CUVr!eKSyRnaK<D$(O!szEV@}lOjpxiTTtk}
z55}JLyqX-WV1d`X?AeJ?J06&-pWZ5<VHyVjrHZ?zs@J2JooOJl&0;;86Vi`l(~;$z
z&c$p#k2N71M=mE7Wok2obil8#fnO?-6cb1#d?<@Nf<)w#TI+XrcYowNKQD}e=z_vo
z)d${?HX@%O<Q|o!vyen=ruSHuV3E5K&r<me<cc~&IgN01UBBf#-=jhY88`EjjgP;;
zB=gaQx-8#yNA9_XK)*S^iLX}`43B<3q&Afwj6~AWr$Jh0jH9D_EM)~H_nRiyv&x52
z3ONv`Y1i-PR(Ki2aZ)de`<CZiu(-r1cAu!f)=q(wMcCzuFQ-V^<PS!g{*nix0|+xE
z<9F{>-0rKmCP~<bS<Q8|OI^tq=kHnvUjp+<jgA+#3-jH)jb6F5^p`vEaUzcUE^GQR
zNzP4qaI=w4P}nD_`3w$&_s5V#r8Lxeg*>$n)ADX%7GioUA1m{%?+tjr&dy&<$lg<*
zV^&Xjvwlc_iPq728$Vopp>VDt>)mSY;QqVVNdJJD=A0`JkS5ndeLp#%-+B16Q^4zg
z*f<>0u=QkI1~``AaE!Qh+DNhG6R_aDmrdWJBGl3A`y+#W-wEC_Gw+K~eKq4(vitMz
zH~QU1zqg~`Kat<v$nPHVAN3;sT;-H|C%qY$Iv<4VpFB<k@u`7y)sfl>;qGZ`?o!-s
zH7D$?&xdBDM&3OTTueIF8pH1a9SNj`cd0Yk1bM_|eh=6=dtr>oZj2-?=F=d;-F+0n
zk?EDw@@^MGrh@af3?<gv_YtMvKR%#kr!{ZI7)%?!D6^wMN2^Y+p!cI1Z9?j-?By2_
z_Ei()f+VT)&G1tMe)JQ6JH%psCjL4|M`iTw)L!vV&9X86x>rCG>5jUhv*?+3MQ20y
zPD-Oz@~<<P)gJ)UhKPn_wyvzYXsdKChEJ=Skk^UCK3)7OcB#>V0}P;EPl(*OXO9RK
zh|(qJ^9MB>ESl572|lap>AnmD$h8uf@LHi|x`~sJM{kA{|4*i-)CCOC`VAoQ>nAhu
zQgXit$^1zprCkcpB9cHu)x(BS91#``MJlH8ML!eg%drbgNC@)`OvefUfN7lp@Q+>*
zxDLol1j<he2?_PeE7Z!%8}d03a=>Aq5AD7rbN+L(AOH1w=-<jJsh2{V{CiktS*Is|
z_)+|qf$Rk9`-2l9?Hf@~4$4?p=fC~CgWLaf=IP);_Rsb~)LNBom$9v-SEtXI<E<Ws
zch@)&T~|O4x2p&(k_#o_yCVY(aI;h0i&V)(Z1ANSOX<F`9AF=FdIDX+4q<CX*n()O
z<!Ddo4C&s)>F7RW-HLM~m}wn^N0W%fw6n&h4wRH3N?PZ}!3eH?v#gKW>%O><w&|T`
zbp>t>4n>>}Cvs7)?4dSYf1=LXqWU&jcYesH&NG3}7CJ(zPQ96NH2Td@r#F9NQn6$~
z3+-49eI=h|@y(k!U#pG`;c!<0+cY~?wzi)!6vP0i&N$?pWZF}Zc`PhYF*!-ciG=Xs
zkCzvLToSL7L^D1J?v(Rxi0-P+cbDJqB$Lz_A~5THKM_lt;>Vm$oEwhB0bjco<%CtK
z)Cu^jQ6*n|{0=w+CG1_|jTNAuV=cqkz`6zJG6o2$E&;pcDOIgGT7X4iBmzwZ!f#uU
z8PTt@%hu#wQ+2dysgfmD?dB%=-#9q^^?tve_#ZP#`I0tnu78!!7ko0$tKYB2QPp6h
zBFmlvOky8|ItAXH#ZSCnA6U3tzNHMe#_eromq9-?gIZKMr25dR`{b|JZGBm0VN_c?
zmhn7!Luac`d{sMltJSHYS2|$M1aDs_8CtL1_KB@0vVo-BJ1L~r6V8ePcT<fS=U`0V
z_nh~m%V8hZeLmUS?c8r!{VG+`VvLsloVSanl%cwr$c=51*G<NXnWr_IhAOTA?$b7P
zW@d&_Y?097`b-cjbQI@{N_#B&AuKZ}*S5|$xYxe3N5n!$*#7L3asdi)Gy7SR_;YvE
z8~q@7g>s7(M7=QM$WS@UHY#7uNqlI0VJ?K>Heq>u2l&DKhwK{rsS3dmV?hH|5Pmmq
zBas4UV1>?z88PcNZdPqXP|)RJN-foyUSG0duyKG&f9K;?)%M0Mi$$7_53bD_UQTsL
zRzfC6pM*DXm&5H=ea@AEgC5tz=oPRt(ovb<U_Ju+(%+%J{@Z5qRrO}|YxVGYt-9~?
zsl4D`35(Vv*F{9IU(UdedNX{8fcF04&1KW>C(98?IK*1qUCwWeZF)5(Q|&ULfX;w4
zfBmCfs(Q%XclV?L<&M*qf;A8?E7IUtB@N&Qdye9F3>^PDl&V@O<aWIck`t6{RGOS?
zLNCbNGpvDF+uHDY*66jlfd(?wm?VVKBJ5F6T?b<sy+*Ehlr9FCCKZhIutKaN<-AIQ
zupB_}X-TJ_R7Kf643)c0(`D$!!{Z_go;t*iK#Zg8^n=?g!7t+A#-i|grv=2grn5WU
z+FckQYu&v88_Z+^ZnKs~YgZ!bp!QS!n41+P^)K~Q6UZw~;<KKs=i=&w+1-7uq*jxZ
zTH+C@1zLK^)U-pvIXK@3e5_Em8}M_VXs;Fv3a)Wdk&2=*GRIA3(CKtFcdo%keWW}M
z?M*?vw>1exiWNPDf|968xK&LxX0Pgu2MjLKXOm=aiyDR$K6dDG`Txtp&=>3WNtz=c
z3#&pWJ|7B9M?K5vZy8kprD`dF=iQSv6+Ny#gpx=Xwsvluy_1U2GflQLefUkQsL&~`
zn}a$zbd8#EN36Zr75ud*NyWJ{y6R5k%z9wSoU{){iTD`%c&^t?q3r;a;}eYD@AT`J
z8Q+@q8Y5=&NF|$uY%J#F0;<VHE524$B5ZU?BFN}MrdDCvakb$acDwyiLD9vP%~?1P
zlXYwDYCyAIVCo56Vzykflf>6Gp8|{KbGvk{mm{-Eq)0^hyjoG|K$dhHJLEOH9=zEG
zT5u-=AU0)VBm?Gc$++q$<uYWuAy6U5avdWVWt`JK#JI8qHEx(yA{A#po@?#CamX{j
z?SdDq{_<oS#k=)V(Fa7%oDb8YV*Y)QXOsvXk%K(dHWfsuVZGqG&VSz#HX62+APl%y
z(`Hniycd`XyM|>1Do4&|-gQj6B8razJ8miHo_rcq&^@XHgKJxTK6EN6mmy1~eo_mb
zOu+w8$hmGKO`44uz80q3*w`2;cOmCMvIOVz`OpGr=PCHlKH!UGM+9qvu-p>hU|8F8
zsrEp45aKf^MJ=K^g99LmPo7k3I9<9u=v>aY_)62j@ihFWHvSH7w$JJV2{cQ(YAPT%
zS<{;c`XdTNO%d#H^m^-ej@6qpunU!&&%(YhQ|}y@#vG<63`J*P=2e6#%!&3D3%b+P
zv$mN~QsU*=At%0g+FUo@AfzKbzUpuN(%<I%vqSm+@<bw1Gc#>+ymWH#={8^m-bYoU
z%KD7smJlf-sMb?8);SYu`Ou0hZ&g3gF0kV`xD)Xf+_y!J=0eNPx?`StlLBbH-TYy7
ze~g9mLp{d_vS9A9)_u`7R0`ZGqeb-@$(<2(4qDFSCWa9yk0P<usH1FFu*LYR4nlnd
zRjv`qD`_pbQIAPBLQWTs8>}qxsG0E6;!Vlr!ZfG;pw}=<6ebCsH-=(Dtn{hH*?6li
zQC$7=_fS+eWD%hKm$&!_XTH=yJ3n$j%0PTdRD6mLK7VKz^@-1|eh;V@wJ&h*VE@?r
z*DxWWKMjJ7D?4LXHg*Hl3V$|D2C<+DBWOc<G|N7$@57DM4F9pBY_dFTTr1_&DXCz>
zz|s_&l&HU9Phj7$1jgQM8t%ax0&d;JXOT%)PFsz9=$4e;DKI`Buvyq>{`}Z+Z69&R
zAaeS>1VdABB_szfctL*MF@;TV9eYgmy3yH)W;aGzv=Q);2uKLBZnI(KWpM;6Yg}Zy
zkbRwOXc*)%90pWen2*fdnI(NJCSb9QpmL@b%pN#TM80lx>YgMh$k6L4qH_DyIHbqh
ze_)n>fB9b+`zI(k*ci}76Z(3yQ_#I>_HIC;fa*v2DV}*qZ+xT35K+KwQ*Y|?Aqr2z
zSjQ<w;5`13iqIeO>dryqck-t35fR2QhBvSYYmm+>ah*l^zOJKWjDtesY*J1sqZg6N
zgXuWl)Y^y=kLIyB_bxOJ5@S+?J|CJ>N_~_6NpaJKugP%s-O^}IuxKc)mA^UkHKqFV
zq0Cc8Iafye-*zObrBUWfT-!&u9M`ukL~cNv2Y+qeV>^R5BitRNaQ*Fi&oNPr^==0h
z$2p;(_q)p5BCpLpV8x0?*%h@B!5q*TW5*DyYaoD`2gjw0rmH@Vo1>DY7w#q;_zAar
z0w#E=qj_U|!MN>*?9;!*mNHwA1u^vwEk@ocafW&C(JM+Tc<qm=_btvitu5aoXUd_X
z7<%U)Y=>rUN!~o+`|9b%kc@>3-JABJH)WAsJz|QhjtA#?5#y4UIDqQL1_Y?wm?GR>
z{iYHil%DYUkO&L}KzF0N*#uQa7oDl4+rQS(kv>SXRTkTZx|nQ;+|!(U$B#{~_~ezd
z@2lm5SMtDST<pbYT8~06WCE>6<oL;sKzD(lSw2Gm=x_iiPNV&>ovSc_FMHq`rw6$2
z=8gSDr!)^LGztJrCZy5JU3xzKxuw2kJ^8BHWa`P%u!SOSBU6F3vfbQ|?l@;!HNgd>
z+nPRE4k~ik_{y;mY7syB+&$+x$m++B(oW&Gd-5~S_YU8^Xlt*9cA{MqYyS65|JTz0
zQ|1~M))Ed@3zr-bk`?7y+fUDU&3h|qUVn1S={!ag)DEqF80@zl>ua46*bI*`aeRJ%
zF}EB&Kl6aK8C@Av$k&j-pH@kq$NaoN|1NOYD_~1L=P|1~a`R=w$3P-U&E3!VG)^;A
zNp3@?OLDd(vefvPki1-l)*SK^^?$MV-eFB;?b<l+IMzWOq$}WvN(ZHP7zYrQ1dK`O
z0ck;6XaOX&84E}YgVZDx0SP3bNDU++qm<ABLI?<<47~=FUPOP}dCPascU|Wk-|wII
z%=LTMwP6=l!rISTd+oK>b3gYDfN6;^L%tAu@6jUClE!S3MvI!$Li*B^%yNL!nCgtg
zEth3soM+uSBeKJ+2W97mJ0}k!8j$fbbE-()!l3o5v+|{_2ayd3Pvn68sxYO+b?I^6
zf`3V_K;YsM5O~YKB)8~pDlo*F1K}i;O$8t>0|SJjy9;L{kE9e%x-9G=Ghz~pJl=_<
zlPY2#|MJi|!5gIKB{WH}RB*ATqL>62RKTtVg_n6kl)Q+@C`Yt6gk6-J2`L|^2W;*S
z%HdZK^x*okna(tJ_g@uGDL5RNgv%L<xTQY*U)B5nb^Xvv<$!CW>j>j6bI5WPk+E)I
z*<iaaVc%%Gh`fVPpe|hwk59|zqR44jepyQC;0aF*&1P-#jkfw3H^wP_HLc5gWf?_Y
zc!z@gB@yf4^pp}nvkgPc&Oi~~;r1dU^}Fmx^a$$BtdR54XSp2VQZ_$0c!`XsOF6vT
z75;GLU3)7Xk&xguQ2wp~MIz1+w!Grg%k9YeY3iIZyeE+y=3E@>sn`U)FTj}@s!jId
zUROI4nB6@DNvpZVM!gY*ZDoTyR2#oV;sO>f#2Ma$j7?>j5xo2PKm>8chsw5YT{sma
zpGnfiQMvn<Lkd>KLW)i)hUU*cj?M708Ws&zRkvpjW7v%v3xaTrvtDnM1(7nW5o`=K
z(X3ItqC)l{=0MH~O?+A?y3Yk)v7?YgLloKZBs*${Y)$cGPeA(<qG4|@#yN`u3&iC!
z_aUnK+xL|xZ2vDi|97W<+NGpJOQ^vxx3~xS3-`;2R6Ek$R-HH+Df3p^k@MS5O=1x{
zv!+RFz<TB}A`_ROGILGL-{_RG^?e&mG`Iu=Ow4RF$%~vSWNaC&h?4^u0;@56g4oI?
z0?lkV1O$4yGD=0pl_AOqXb49gu2-<v$lHSf+?hSD#$5sa=EN1rneeTaj7u(%d(|ea
zj@;1r0y(bV!@LBoeJHQ;LMggpqL*<jFP`(EDM8spT*9FJ;<?T09hdl`ZM}m?#o1Yg
z9QQ6m?tKM$l@YX6bp`sVi-F4zZH9)u8;OJ79$0Z$vTa_nt>FkFRuN^oEcO<cWf-zF
zx5DC>Wje=;RJ6(fTRuZ1KMwFmi&QWquv$}CgHHhSPa{@2Y-j2-Pwx6C+2hGy6H!vz
z*f@l)+uNb5Yj^+=SpWFyXC9uz3Uzz$LG)WM-Nm=KM`d86_fVUHySe_uhUxPaxsG0S
zU2ojnrmO=@h);RlE$!rI!_*`5-`y(&Q>a#5#ocOm+=mydj%etLHiKjB?HlXV?D!Gt
zOsgP0dDuE$E+@!}kYpUpRJYKn^y%eSW2<5$?~&*3ka}yZGYd{jl|jWw6~UzC%vGpo
z{T5NTAFkTrk|w@`t<e;^2?dG=B!%tvmEVtN?__v_3I4kAoz#!HFSv#c6A(v)!CAUu
zPk`vtPP;xZD?G+3T1DP%nr1s3no3K_^{FsAd5HwK(Ui9`W2On8t2IAq6Y$Bg!cF&N
zio9Exdb!d}R?WTInUD00<y5r+x_VrTdqE1SdWb7DEE{l-JJ9KNSY2Q6@X1R!Sbu8e
z<r~RL$M5_e{r&!5i}gp@{_9`=3-I256z-c0F|to!u46IXf#ddSO67z?Rm!?~^=z0c
zu42l<%D-DUA;8kCxFqe7b?2Khrb)_rnLvctRe};hxxWP~a;rFv<?FR*_<nCd<K5)A
zz?wgaauGhTkr?xQX3*G#qeYVv@Y>^21iMV2?mC~S&zPU==gBJ>o)^<BDNmYTJ=I~t
z=X^%LiBA!5nVeqS)tHP(v{G5=mbwgQgq3?HfV#`60^{T`wlVW2A;;y9HRu1@sQ=+F
z)R@b5_H)qL?0z5Ogg;LF@^DjXO_S`fOSaG%S-Fy+Prt7p1nDI>xAG!Ko;z*~n;5rU
zh&FVtVoj)uOuRYm(pSX(Wr9n`n;z4mB*ud^Uu=qd#q4#vsTmJqy7Lm}<9((^>-qQv
zoJUw8;7&AK)xuF@nZ?HC)ZX6k8SpHU4)EDq0kHzQl`7&<?v?lwK-zsNS$w6Yt+A*D
zIzEHcJ;*A%+nXZ)+kYtZ|8V%HrHKqRDDd2}VYMr>dT@^iubN&V5-T9gheOX=Wia=o
zM%<kq`EeE|=YjcQ*8?}uJLflY7>^fDVt$r(j;@OwJPpMu>caS_2hfYh^D!wMq44YK
zjV{`nF11p!cpr#z$IP&l2a$vfCM7-Q2lb#EJF_)Ba^CHy1h`~DK->EfslY;tewEni
z?9J1va{vt}%8}w9mjm;6G4diV6x!Xp5ML1UlZoMVB|Pft)LR_fK(<Ep>L(C10xR-E
z%*EPILw&(xntrU3sP7QUqM{k0Rv9VE2a02mjwhLOK&N9ucCVFL0>i=QGP08i_G}Em
z!>DhD-m=S=Ze$lxsVi(%^&YAo{d^4U*pFZ8_%D_yxtCeX@pWO*&V!}DPNZ*S*sKID
zU9q9xK&cuY@-TmH?{1<(ESpM~X0;Kd!RwFTpp^F&^=MIkgS573COJ8E5sC0+52S?n
zb)JWkSMr{1E9foGdNn+@dePJYD9V#ZnnBJE6l#_+pkiAOY5@YD(qKBlR757R+tHR=
zyqv#ms+($BBN+IZ=O!p(D5pXBQ-6?0oY7L`&t8_p>)alY`W&b#I?bI93cO0{+<nz+
zz|kS*9It49FymgrIL57i&<+=?`|zysofpc|E}`UV?lniKv<T9~sp^HQS{I=%bZ5{%
zpwvQcd9(-^dvJ-xIm2Fn5@G%fFpcqFCXroInO2^gl&u9il7@m<VaV=Tks`S$Dg8P?
zOKmwbHuRg5UwA?f3_$HRose~S{6&uyXxGrHE;9ZR!Q8e(RHFsmwOep?)YG4FBFyLM
zYq!Eu+NjsuRJ1|MO(Qc1L^Utz<`s$PA6|wWb-3{g%?tsR*D1-ts#H)nlDYC@cEgt+
z>|#YZ>Mhu-{dcBHcge3I2_<PxaTXN}L&+KW>W_1diPzQYqxYZ6m9lHiGVkf91Rx=-
zlA`oAp=+H&Vi`|Wf}V2H*K!~o1LSi;_MJA7J!W#zY`2P`(n{IVtxg0D3nF{!$Z4PP
zSeQ0@U{CUEiWWJv@$5?z_@B}@&+KbZHR-=>f96?mPbpjcCG!X1Bf8(kWH9z!B6jSM
z`yTB-jQkPqdOcS(vbyKYskPCTh)m>FXXKaf)>xbn9UxDXkDae>e39_v_^B@!{~wa*
z)Y`A*f|>?WLuGN{su$8S&`~!PHtsWGeiBV(EpOI8(3<|tvwP=#$t!jzBkwa0_bwq?
zA6W4b21VT&KQbU~a#G@eM#69EuQsTbx9IPoI7|l18K7XFB<Muo3qGAFzT0?Si5>VP
zR1E|&Nt8FMu`op8ORXZ2hDU8TAqW04!rpMV-@dlwhkw>qx55+L&){ukzX!!xI=_v8
zS6P}I5f^b|-N;QVf6vg}rJeauQgNrV>3mdyhPHK-V?*gUx$vRamHdXKZ8@SmODnc&
zy~G5uS2jYWEL4FTcj(wW4Cyn^1N<Ph;9$7)!#8+90`gNl$(E8)16ECpl!`iP%TzuH
zJb6s4zbfHti^{|EXJyX}_w48}i%{4b-opu?(Up_g5I9uhxWL-wYpX3P-b2Yo-_~7d
z+MSG3xQcmK5hk@s7uhOvYz!{WXA@xX0B*<PD%3>4m<@+=jg32J$)tQ7psI=Yzy`r#
zEo#lCkx2K-iz{{8&=7HzkXi+NgutD`C?)W9B;ksPJE&<qT0voB-D$o(MofQMo_u!;
z6|FU8VId{s);|@tmHQp3pnh(}I&Nu8eYo!)C0n2;Qd)=0-%&p|&Zw~pT%HNwwvj0i
zv&FN^J+|G#uE8xf{Nh#Ojq{|O;m{~54CoM%nK^K6iSXS`D6yYyL4ZIETm#*Ea#gfh
zMNFKKvZ;pmP-cr*dvXg_*;LHm1(-R18CYwTgXzJ`y(SbG#FI^Pjt}Otz(U+!kTDc2
z0Ob>4fdT-`qoky+ws8PdRTajBgP6enOVZy3CdIg|*)YfUtv_1NX|7~W1Z)@9U+E6(
z@4@ks0NXNj!WSx99$hs{le1R_I6<{4-iX|NAm$t|=GVll%n*|Z8lfUbLwxP2z<)>5
zGf;3a9SuxV*l=LZ0+aw0h$oYQ_4}@-Ej$u{U&fLF^rb9O$`hB<_T&i{m(vY7f3Q-=
z(yudB9o}xZ79;8xCFNsalct#`c0a7wNN|z}Lx|p<dOJU5WOlxC|NR66az#w~&XgO~
zUBIJ=cGm|NN8W-V*pz~7wJi|n?r<64z%ZN=CM}xE5-cgP`5Um|$j3dDFwHE}$gKN0
zL;KkS#WS6sc@FCpHFOtq3-==!-m6|=bQTExFktvJ5#L@EFw+&y$yiUf$uEglGAGC%
z=d%xpm^2m`Tuuv(HrBe87T#fHrZ{W-D&dS*VN-&&i>t3gLrU<>$9U?MF*<B8m^-{!
zA21UX%!CgD7Ka;*bajz29>h3jyev=~+EB2v0E<t2{t31%sg7*GtrrT@?~mJ0cOLpQ
z8D!*LlV;srobu~BH^=;?2U~4Kxm8wZNiisBUCL7(AFX%wjr}m^lW)&vWd(!ty;<%f
zgMp;>jbFHS(?0&zZB<KB3j^N6-BivGIMs7#Q5R^Hl6e<gZN~(AX0p6y(cq_!Ex8lL
z697?m5s^u?4UNzcak7)pm-rcYYhJd)UW$9?ahZEX$?^tJ#jSJTx`Ub~lOrNg8p>uA
zcAOoskp6K8oxby7yW!(iMqc!V6}(5={?ySN&%0>^#W3$)e7H-@@ia_qX`O>pw1Ymc
zOWbiC)3hK1blZDiNj~~mUK%fC;%%I3o$?k4V05LnC9}fx-H7`y8)lH<t8%bAMWbqS
zzS>gp7dcZnd9jOePZe+X)QFbN`BPzx=riO93ULTI#r2kQGIKP8Xdq0k+R3|#DQieK
zwKi$=RA<*u`&_Lb>@zZn=;^&#%EXXx&!Z$|E3fQj6BbuPhiRi4`MjPJFJ1#QD>7s$
zP~4DvUWU+__op_$t^@tolL`E_oF`7MWpAv^q*IrY1DDaG-zgD&wq=;lOW%n3J@8qQ
zL55G<efhQptA&=#HT7XA_II_4`NtgBl)E3rb>c5DB&pVa2y(y@Vo)Yyq5IHSTeffs
zsF{8J2=)0q9(&5z*ussz3nAMAYOjMWbo~{O+`{Q!0?%WdW|@Y+EuNK9=yj~otO!sk
zaVXfEr_N%E-rE=2YxG)v|8-Uba@jziZWdW7G`&)U_uI?vc?o54Cl7qdlm-9z<JvKL
z?jW40Yhjbv)rGrpWxa09lo#<LG*o)D%!OzL$v7ExyG^?g10}+5C$np%RuzrV{;KBf
zHO^8O%Zc2Bqlm`bkQKI2VBiKuJ7A_qQVtjUB)07owodrU>PZVjN_Y2>$Cy$0K5V?J
zvpJt>I0&a`s{-6R^0V4Q$e+LZ7I^-o5$bvX(E-w$HEihi5+0`BimuDgtL$_6wlmSe
z;M4bv1d*1$zP^ojE{}qOv|dY}N|$j9Q=9io5xhy&AObYKn^(Ozm5a%;ZQTum(0CV4
z=28%5&^HJvrC=LJ;S#-8f@e|2M^Jf54~?0_^&^>cCgf0XP{^*um01?Q*es2Hk(!}B
zbLw!vBD`xa7LFAb{nT64flELK%V|wT$5|MOVQ$tayY#g<T0PR*-{gP4aF-!+P%4E#
zHksajvZ|^&u>#<C^ks^&q3p)uaU!nKQELk}gnh@T0_w1Xs_$8l5nLZkb}l?eD4uqQ
zf4qz^gXqmyfqL#3OtU+e7N2|yf(&ukan7L@ICAq}Yx~f)_IdMlgJ)LzcUnx7YTQhD
z)dxMZB-4DTO~OVJ3A;N+!WOzB{tM3vlX-`6nNjk`9T<-SFBDOa%=HBiXa@KD0zPbn
zs@9Gr5x=&E)HQraRX4R9v2e300}GLI+9{=KEal8hZPA)dZ`QJNMD}of3r)Y#%}RDo
zl?&xlw}?tL)yD3!N-uPt4V65;(f5MpIF-@YKkjweXn6<ycH?52ZwglVwQo^;fsja(
z9ME+&C7D5?IwE(YdHqMv4O)&2N{tTB@q>wF!PQHG?2)<Db>d6>LBjMX7NQJCvrw_v
z;8Hsr*|X#Yx+`2{4isX>5+l=L!CTvtPjVt|h3WRH!T_a1((=q=hxR|gG{(d8FK3PX
z+Y?*LkbUTxhH0!8!_{Tgo2pQhxnUY`*tlz32CIuuxR<TN$XM66sOX3hWzE_Inv`F(
z2UaH&@ZTwEV`)g0!q*xu9)@>v6EuP?&Z&2cEM_cJ6@Ku&UgK_2HiNLeH%h;iXmqE~
zCy0_sMP(gaRP}RS!@{TCcL$S99mvIncWKvr$eW1L*s^y-&qk^WnY;lNn?-Z4d%tU=
zqz1D!0(_iCfMH&|{#gACpGux;{+7^J_ch<`ta)jWtn%qsyD*!X^Rsu5iLBX<u+3V~
zf<udjd8S9dRTlNBWVW1EM7n5(-6%tE_Jl`+*rfsMl`T2kRdhv|!kFJDa*~MmYPZfp
zmQLR)H0;L2k;Fi7N4-e|MhV62F+W8rZxMv*-z++P=$!8Y_1=s+z(ig19h}{rQz%;T
z+sPV-<m%NW8v(GWZi*CPYtkjPmGE%X+i5+P8%xK`y6w3yc~%88A>uD*Tewc;zkT)1
z{rNgl?b9J)+e2ebDlpOQ;Qa@zQNc>~t!$M}->wktnxP(b7mK8y!b@;P^^y41ob{|-
zkq!MSdHI>nYS^z+%W09X<#~~J1yx2v6d(HHetNZ#QR`Fv@iR{;-;GuuMX3TWbI-g!
z1&T^Ouk(VE<;yfDs(T41sv6XEOg|31xKNZ~{UT1!{Z8MMl8E%TVxC0`hVf+)*@hz^
zUANw|AfyS`TIC4Va`x=dovZQZU80HQmR0EmELgwOQKRGQ)12ZNP|~r2G&74+T4s;5
zDH5=4u`aKJ`sg*q+pNz#Z(eX52K4hj9-QA!zE`Vh@v-rFBjGcTVYs+>;ma3gS3*yP
zXI7<MA6Bm=9RGC+B7eKuAtE~8HOska=!ay;a~nEn=0<iO+Al@N?J}(1f1`p~R2;+W
z*32JOV_>HgJsY=g={V-lTh;w^LRS>V9!|=ur^Ue5mIe8hV~&-n?;cgV?V8Z(Qhk#l
zGMTHy!Ub{%MExcN)dgMK4x+!WwS21;y<8PJq*9>oqw0S)v_y#Rb4d{QfazJ+Q#n_X
zC#B~XEsA5ASDC1%=+7L7USx=(H?`xh*pcJ`h;*^1oSRVsgRRC!z<N#8tq?(gtTPA<
zhr^*A0Nai>Fj!aR<Kt6RRpvY0p7{iao8c3ib#LI?cH7<~)D7E3)hg5<kyYQUL-!TU
z%Kqx4F>WT@$ZKso&FGDOh1S}@*pgmM8J*H{5-65`mPCB;!N)+@p<~5n|EYW{BL~i5
z$&l~0H;C`7Bm6muD3?ZL?7H!KW?2b_Gy~Tz_Nwj$w4*v``_8^SvQ&%DJWbfKmw<em
z!5jVJGU=*>GU2oPH9jV-b2fmN!Aez>tsWHUl52qO!kq8Gx^_9oO8{yKyg*>w_peWh
z7y%n_8sT5z|6<v;rMbfEoCH(ZgrXrs9Hq(br?M<$upa@fS|pN$BZrb~m&#^>tb)mu
zGSVFU%Oue-8tKmYRMyRC3VZeAmq+>NL9|FgoQPA5$Yhe(g=i5!z{L|UhOxg8a}1Dm
z_%at2aXO_hk<<`lD(091g938Ae0=<0Ov%UBjYNh4lTcuatvgk$)`_B?8?0HinK}^<
zCM!Mlj=!WeCB=0-dQhTYkJUd`ivF<w&gF`jexnK?=u=q0I3Ac^9aa^)^G=*~$d~zY
zq-_ZjXb@`kILQhV_9&zh@H}-C=c+>Win2fRxEaHAB4ZFMr6aFuC=n?_^g(&-XP!#H
zNu|Hy=s3PyREa3{tmY`71-3d(PO8_An;pF=`7|A-Qg|(9eM<PJ^O4%gQ@wHC?>_wT
z<JvZrB<9@uW)?ZKC@LJV9-uViI))?9EU}qEY=i#5oNy9R7l<|lglQ!49&yeW-~fjy
zQ-D)rq6Ps45b-XlrlX(gb3S&hf0=A&yvxh^`HL;THl6q-fK3~)Jv#oGr!dEyPCjz?
zXW*mDN)nFexry5oJ_K&jrH6rcfop#`8}7Ep(z~JC86vpll=(U~6S&C3p*(W8r9?O0
zkL>&OM&*Y~T3;^yKcq-^?dj$6Vw|a(s{+^cSy#QT&dh!*oN0qukUze)H0IFw)}m%i
z5nhSWauHXII1JkJc-thPj82hk6?v9)%=wK+o)^A$XrMZg@yvbN!*=#D$;s>N^}<-k
zz>eSgtCtC(_iZvaw>D8z)5Fx@*v!l&C_vjB7#q{s%;;%pf74lY{kjq&{@sv<lBsS;
ziDNz`#K%fvRJ|7t>!EhIrzkr}5x(ik(x)oos$Z>}pv(G}88+xH{5gPWX8*F3e;dE&
zo~P~2&Az+`xni8uhAIoEZZ+}CYP(j)WLM7=&kS9u)xp22#K>$UaT+ey!3$9%=XNsp
z>1I{2g&&ThhRZ2o#v`fLPZ}ae<7QFU6HB(`Y1r75ZZ7Qux&AR?4<HEMw+xzUR!kvO
zP>)>C4vtS-7+k7PE|_JDEp)ubA0(<<@d@lwgiFC$WNoN&faEk($k&Ekyy`mq_$%w<
zFPHR`b|U5?FW6;t;)9H?p$qzW3++6+_EvL9POT<NHf@!Q`B*pn(+LTY_J^A8-X3z^
zqUKNS)zxoXU;!0KabAvr!%JpSOIiBCrHxaGVzO*agX`BXA8<>pyZY{jQp>vR=@it7
zWfoX_u%aL>=pK#Np{96AlT5Jmq?#KLz<a7e3TKt9Ms(io66X9|e(yEuP<v_1o%XUx
zWytxzpDYXX#P9c};n?I>MdblXsrooxt|lqigY{6(&6}QF`O3cC<|n3aXFXKr9ar03
z!CLXog~D1J1tGcPLUeuNIQZzAD}+<pU%ipqm$zNmp2n+%_IusFQbkz|duYj4=lbeO
zLkEINjWjLUH42?6Ay*9V^s6Yj%^{~pDS*o~i!kVmD2t@c1X<L+bI?#v&<hjEtSzoc
zj7JU_=*Od5G?w30nP_lrsi0t}2~8GwJcpk=iP*E>{>mx<vi;ct-NQYHhjo3#5v@~J
z;A_6mh`DwMgq;CgQCv`5p?HoBXR|G^_Hh_bk26!*2I2F+r%A#1x^QAwjj&pW1+==>
zsHK>wfUbT22}i5D-J&@U+#&#}#l3AWV&in@<?t*QBB}7LcMTxC%F3q5%3Y1;RlleX
zv6_y}WOXqVW?oKsms4l)cb{D^^OB$}%|1hlTMmqxcH_%Q+Le_$%PraKpIiX^aA#01
zu%3BDP}8mf47U{Mf7`OGrgo<w!{i#6mv-wc2IFyO`Zo!0!2YL)J!T6Qk35~-S9<0V
z9K_TK%~6?wsi!M&Qm#-H>N9=dS3mQN=hJOHWxXy<s7NFLTnmI{_!(y`Xh-T6R!ez{
z(bgD?88BC0%G?$5isL#|AuLxoWjUKLD^LrpXCihrFm^eEiG~E=i_7p{`NcDEb%6E(
zvMnWDX(c}4+hG~K>v;F=C&?PiPJ|l_7MxH~zpq~gu4EgK^_-F=&8ucO_N3r_!zw1@
zm}Jk8!-KeXGge!?@(jmKfRf`Gy~Q<<*}h5rrJjgKVybP~rM>+`;O(ed3gH8bedZZc
zV1fCDMtz_lt5A^VSz#GMo&BsXvLrMokPqPUAm<Ey=6MYR2+68oTX*UTUrk&KewIfY
zrWCIP)3e3Z-*5t`{q3dC>TX-MKOkah(Uo}Z)J=2{di`<_k(8|B8oP$PrUCPREvD>(
zSs>#V047BuO`a%6BqM<XIV_qDhIGq)!Q*`a<aJADMB*|taj`&}QPxl47vIU;syDD(
z*!l9vobGp1AY@k6bf0r@pmvpHe|$ogp6bmoF_KciNc14!SP6!O{HaF>C6Nszv-P}7
zvYhvbUq(W%%^;e2=uDux9^M<$;XGQmWg{xcIzRtmd3^~{p2g5kaxw<N18DcQZX|fy
z?S|-P$4S*MFIiyX{SCXz$~Kih^OUU3fIwRop&rC1ZX+?g@o6Cy<AE9?MnalrW39t1
z$5q7)(+(5yddgced5m2x%kBm8R1>d%2G?53^a7Ss9<>-2hjO0U+&GX!daNtbKJ!Gk
z7i6tvbylw}H8-!X2n1A#Cp!j{GwVvw2a>Jl<<r(0lE+ulOe?5{0c=$u9)Mxc5%z%{
zn;91HGx)bJ#`x;=zhw=PNBdBi-KAdfz4<hjNl+!rKnLVL7200k!F&`eO0`h+#)tNA
z*bo#f`)_8;iN5W~R!N`pE=kiF3ae#i8#!7CiL_^1V0Ml&m<PRskrS)T#{|?Blg6t{
zc!`jp`ur82v-E2Fw!R?mupmy^Jv^5c#;?>8y!KwDn3|jzplweLNFUudx-mx6AH8T*
zqwf$g$gebLsq#W4A4g0`z|2Z%ui=O@v`L>as;Wsx7|T5P+`&Y0Jyv3Ov6e;JMV700
zxcT(CDH#pwRe6^XepQW77|U@xsUKu)FuZ+%Aeqd2-^z7=)0nm!xImE_BKr>5z_w|x
zvUIf2FD5K*`ZOfrsyFS<I;=99iCYsBTa*mv?DqVz0*blGNEkfiiowRcdbQnLJ60(>
z2~Mw->8>Hp4F-PGeS8%qfjXIpLzJ3IyiC-fr<e;>C80=fI*6f6fkCUvbtM=JST=2h
zSqVbzhsg(-#7)R>JNwwzHIPeDG2`EjcKV;_5fC^Vm?5u{lMJ!{o!Em}#gzsHtSMO+
zrX(nj$Z+{D#-)fY5%N+J&Mc^8M-z3PczVhZ9VwMuJetEaP~1pa@Oe{~c=m|-^*Oe}
zlJ3B#%T~eLr`1a6>L_vB%{IJ$`78^pIrU>NvJPLqq?1VY0aHR2SQVRki!!h2j^2^t
zsFUkB6A>Z#mcIFVf!QT(p+x|*?~*zQCgqe9R7h9^$WA?o#)fDvR(K-0I#S+rx8chb
zihNb~qMr;Ev4$6lT0)pM<kdv|>puOw9H=B%pP@cWPt=PG4HDcuvoPJ)+snFHR6A6f
zx=~9<DC#Tfec#fRBwojD#2EQ5b3gMmR+Wx+(NR~c6VIPevhA%MOZ1%S%p5jIN(`9k
zD0ShKaXSZBN9zOSb(OfwFX@9<5V(Y;VLP)MLwJ+Y^QJXw9Aa3RId}~2Iy=qg6P%tN
z<O9JG5D*ls)}q1vB5%7u#5;?7R>@_3UG4mlOXMKn8D>v+;ou7D)lFO8W4yrC!)v>z
z&L0yqedX@+aA%Bm=%+78{J-Q_D5*}U#+Y^?Iu7pk`rZB#P0PRAj$T^wc3-DnnU1LY
z-s?+GdhF1Q$!1&#bC-T^?KS7U$|2wy&pD3A`Pb%|pUMLq+4Ot*U)g@XeDXh~zkG6z
zKjs%FlmAA<sp$(U&#&p2U(=~SP<dOxbgmK5^LqzS94_+qgx>s^7eBkGibB6L?@(=V
zEAqnKn)2`1n;;K4ObDuR8M9_V#^?-t(p~!Pu4r$JxEnb|1M9tL^S*B}JDh)s=wkKs
zlFd}(VU^$hD8gTmpANtDD%Z7Qiu!uc3z$I}h>BUlV?}~RRq0xp4vrJmC44sPGW2G)
z9Yg?nx6iD`e&_B>2nd&`ag&_H+vF3S&KA0x3=IlCvR8hC?q$j@hdZPN7Z;p8XkRiX
z1kmV`p18SLo!N+N{Ut$*AfhC8LGVUUgq(^jMOq>=exsS$A*mhW>1;3^weB*UbKt-1
zPNgVEEd*Zf&QL<*eg&s#97EnBD9Iqra7wOPCQ7b#6Rk2X53~R!7Uvi-r?@XsB>qi8
zCdSRpARqy|^*QGCM$VUtL@VL$39f4H+kVJA$a$S{aRJ%+Kt{)`rq5DqBEg_PSwqb4
zIT3fd@v#p~WJ*k{Nl;zgxb6V|of};%xoEzDb<}LdHl4($#)iQM-ZYbSrRi42&Es^?
z_OcCG5;rG|;*fRH%r_2%wk#~Uy95pC)aHPh;OVF)eu-A;0~-B9hX=8k8=~RR8LUU~
zN%831S`Zo*)#r`C8f#67EoAdvh>HO2?0K-&`6i;p^>){#!125nnnY+foSJyU`!^9r
z)}7^=cRJ{4$$?en0TGD;>$v#fQWhVINe=C?9Hn^x0ra7PI2>i!WvG=ZRANUQ<>La5
z@(3C^XPBIe^2FgH79IZTJpNBj?!P_x-?CiHbDBB(F*~*9sKQvKWmu;W?_A3I<!_f9
zoH~Yxz1Jc;2kAc7MVxe^m)vPk_HBDvDfHD=SwXdRHy44cEqxnro3Q*~e*1yofzn^f
zeYkda-qti+ksNctQ5Hn&bx?&xGj_O@P?6BO429__hoY=z`<j5bT@b4$wd=xKg>qoU
zsxxxxU1v*MKE>)or4J;sWhp2kNZ7}oNC{#YjKD!}UJ}UDy~|5B@^b4E<Ekj1;W0M~
zP9=OvorXseJsu5J>3F0_Ca0+it_<Q7)}^}R^LXjkF-4V-8}Xtr8_aP(vy4++_w_g7
zbrTgO)_pr^1@}e}eI)=-xN2n?t47mD;LD;{eaS+mD8?a3%?>+U6}&m}F+N#vynm5Q
zUL*kbVd#Ufq^+sxm6TGGVN9s!sA{nXOSK#6(b9++GXY@r_N7QD^2u_6grwy6%;Ky}
z#VxnaM37Dt=^Um%UP5FlQ4luLLs)a$Lu&zQ8%}m^bk`0JT&xR#4>(2Vs@Y%O-<c42
zhY>*s=6Xk1<|OG56-gs>yI6wV?HouX^Q?J%na9AaORcy2gjd~e#zO2IzouH2CtcDx
zc5cXHeNWd{URZAZR9>E^P!tX}?dI$NaI<3BBe%FL)uADCaNEZ#C$ihSCm$=hw)=V1
z;w_>@D&C0p)x=yjQbLMEjXylT`PAL%lwnUa|7ElI&paZ2_BsZnZDTx%NT!-EQC0a=
zO~NiZ;7M~_xW@8GK2pD)NGn(YJ!kzEDcRh0J-3L>H^hIoOj`D%@<5q1zzZ82Wq?W&
zFmKiIpNPL0`g(IIXXD2Ea`eX=gCDbZ7x<ce17GNQJzkH|G>wyX5!ZU(lKk>_%v{46
zJ^4|?GwPptr1d4W5?6wcCi=bdo1On)yzEjWAyTQCb=rPA-}G@<(d6~U>AB~od=1^d
zeJB&|78{k2$xc8|H3?l`|K&gYudIzHlbZGKB-G#RHW3P1_$#nk_nq6u{PY9IVkHDu
za%0ba=1Iv(7Q4Q!d#JU}Q7aIAKR;NRmBedvuKQ?NSntI*4Xh6r9oGW;E<PEXyL*4|
zySY>U;l6&`?7>{RZ$*o|<n_rMW9)x!Wcl>jiSP+#?_}&dhPj@ytFM+s{i<=&iEWGK
z#HjY38`z6s*}qKLU&+#@=|OZw78ubAiF1br_2-i(92Rzx%XhW^ZJd7_=ie>o-;dS5
z`<$=rkk9v7NNdLN4pq)Dh<hFu-}*c8Lg&1MKyk2Q|Ia*I$xRkk5FHu6PmL}qDuy2~
zw0v-68_uVI1>xrNqZv|?NX6bt^|rG5{ZwYB#(Zq9S2;p?8+pF^{rM+V%KSwO8_hjR
zHMV8aG9OjhWu%z{i0Y2SdXas`!3I({O;0C=M8L4TB=i;FX|UZCEqCwN(#I;APNiep
zz1s(t!1jJlEEgd`QS$gv(_?-QyzjdRUF9_NY})P_c-*H-dvnb%uVnwE-4d^fZaDoT
zs|;k^qr1C{nYDDw#4TfIp1bxa1*)++B;BpW#i?ogsiAbp+l_t09&%PfN#}<hk1F3Y
zqvWFZ=3u^^$mt#*ixg-TLYVTr3-g%=SA|@tG7)QsJ|A9H>s@%2d6OWYr()lZQ&k!(
zvs!xDCXIdvN{Q&vJSXU#bs-IYw^a<;q<Z*%(<W?Z?$z>G-;VF|`H>5KSD*-CmG>wX
zP1C1Cehv&{l`Y;Ep#=-7LpmuA+0+j~-w8{Rxt<VZQ_E#;S<*iACOLj@?7HPtd%m+X
zd)WPuc`O#nKWiq8a-=Kk?=%nsnEoAd72|VG_N3IcTI0@O@7$*Wqwz9sdiB>N*%@oE
ztgGKz2OMCBz^8A5eSx5#09l`{J?-!J`xX{tXKp?`m1K5AT+Ft2Kt0>hZxJ^<)(9ZI
z$EwJ#_FeroK6a`XA^kPH^s4Sza})~ES+A|RV&MaTGt25{I99G#(ls<RB=k${L1AB%
zNQk2e;bAbF3goDv9l)WvK#WJTxBxVfZ~J_nnZxh@?HK+)|M5>IHYazX<IUu_-%vM)
zYj@%8OSVC0UBjST$6~@}cqMzv0t0|iyGW%Kj&cj+%PqW15&Kq3sUe^wmy+a1fg}}n
z(QU6w%v!|nfS_x%!*74H#CID3-aDxSEBvNqVM+tMEjn`){Gq$afqhF20|lZ~-nqQ$
z^<wTF?{(>}R++^Wnw1}!I&?M8+Yg+Zva;(8*26z35yp=MTztHrUut%u;X94#K?0Ez
z_d&@@ZxN6S2lJU~bMYVK$b^WOL~kokN$L8LH1V?ETjDFyJaa5pE##BM7k$HI=XNBg
z19RH;=E=y2)Sr64d+{sJp_LNR;7z>J9Uq}{K_(rqBK5^CwLRo@%D7Sx&1XR(2wiO_
z7CjX6+q$_vl_7TuD`El++oMG~Hx;3WZnV@Bif3$FTB$wRA<{!uR<biFcHJC;QxlL=
zN=LI1_C;yH2w9@m8_`xTRMM*{pX;e#yk7%ai}OeVBwr*Amg?0yeYfmM+^|;nt%m)p
zO?}`rX%YRWy`F^AYk6BQNiQb2i#Y$V9XQ)8`ndlDFRZ_hc<m7YFb(#XHndr()3FF*
zRkzQXgh)DPjdZwYS)j|}in_MQ*aHF_3^&!P$Iv_pX!c%Yyw|z35p3zw2M*tO4Jzte
zkf}rQXCBYu8Ga5}S(({u=A84No9$Pe{V$;G@H?__KY!WI9j-2f@EKgD7-f4*#@REq
z+TUAw9(mT5%d5rsHonZdePbK*vLa#LDSmiUyH@@%?|fdXXb{BYZH!p*gs^esuUURI
zzPfkcPg%FUMHvGxfPvXo4#!Hds~Oy-Z)Zn;>~6Uds6488uY%-i7aL3Nf)3|vp69DY
zvB?zMSJcGo2)#)^Wu4ujes*UTG~~Pxp~fXBUq8_X@g|Pm>!@vn$_pr}eeGV?-{+Ce
z*`yoXzQGW=G^z?;y1Xzgm~tUrOc`r(`6M!e{NSm<BjuX}_(^^H+a{leqmA#o9Makv
z?0+~p5UfbyRK?jH^>rQWpk!jMpQ6{3D?n?R#F^#zbHPk25;tz4N>iZ)vaF&6T~063
zeI7kp6F^2E9J1=l*KW3NZ-{*;7_c>omuFHasvLe8)Jn;IJ{D_n3fIw8>!hOPx_q6c
zZ4E%7r*QS2p-{<(1s}6q(oQ|-jt1UbyA#pgT^AU@QvtxuDSHM5k@hUmVBh((_}ws>
zg2z^XlLYwcxhLtf#dBPyK#%*tK14d_>DF$KBF>&+(4PGo1Y(S4CJH%AzYok^ioynq
zpS~C5n9!el`%0q?{o-nD)t*fG$F@ieQRcl;d4lqWXm?W{_Y~y=K4tOSO+ux?4#6-#
z?lVu<L$<Q%FG1l>Nzc0f&v!_dj3#e44;=_teR7o`#p!)Vnc|H&68K_1%!vFE`F!se
zedKb|E2*CO{KoYkHbIcz5so4*^Qze~M#JJ(Q|i+F$Cor+BR*(EnFs=J$fIQEq;D2?
zn$BO#TGo+xyVxj7GI(b3A4;hY6}T^K)he>bJLmLvD+a#9yt(-ehB<UT9Gu&05;c2m
zF|*(|+ABb**j@KjOtnsVtEWXV{{X6Of71I?ocw1VmD8{miNov<(FPB--=X#M*paGt
zB7V%N3N{_*ci(_G?*963JN(-Y|85WeelGsqKm7lyp9DC4(_*|@T}3S3rG?k)Ye;m8
zm3BhAao?UF6gWe?r}hMC@|ow6a(7y^;h@CnxKF@)7y|3<J9m~msoe_>m#3p*b3NtC
zawP5X7gxF<_KcB?WpoC$TZf}kmxOZhgQ}iq^XIepKl9kc&=&^f0=+cpiTV;n+(CPl
z!(yM(k7w9d1h!*hEKF5oUNniOq1jfviHJlo5wtT-Ck|g*o)@z(cf4-utsO91y>P$b
zm-v9KzRBR<Ml!hT<Sc!FN`j+#g(Y^7NEiUXUWOrLXIWzWwztikL677UuK5ccQ!af_
zwg~dJCgH4f6>VyyxrZ3LU&Dgm584ER$L|5_#|{0tp2*;u<{p>*HiFbpp3Z)1xy>uz
zh8<d0-M}>1I8N*C!feyEl=a8zcy3e*fYc2`HY(g~7=GR&aFu~znVb^&y-^q=^8rvM
z>$l2iHW2gDI;$xDDt4o|qAs`@s>;p}P;WT15-dSL49$&H)OfyES$cv?h=i!n&jmiQ
zGYsmo>#ccdHw696vmz=UuJ<WochPwU>dD<H+GS6%Fe9Mmpj~=P(uuCG7$yHB%byDz
zb(B4RB{-qVjXtaQI-sEHowpGE@yd2cokfP(HP32WlC;%G8Jw-E3=ndPpA^?y#N*bL
z-yFZ=W~u^UjK%zqNQPU4RN;1_eJTFBrCsXq<l&m{^hac?hkLD-pB{8BoU&%{))x+W
zp0J+W(zzAKi}R^5Z`l8~{>+LfzHB>`AbEnGZLWOu(9eIfpAN+}ERU(HGHM>TVk!qF
zwMG)=bWSWcc@dKe87i(`>3l;~vzCNY%CD7V0#8Be92B2kB=kl<t;|n<1;oOcN@+>$
zohl78V!sbEQ`P(6%Yt`Y{LDj0vU<^0qu=<kQ{Boqa22j>6?x%wMv}RuK1VRh02k+e
zfS;<3YuxjElG`LU)u8CP*6mX1E>@qh9&A$&bqE;qsI;w;aHRJ#mQyqAF0VZ>>1G8K
zCgsIgfDzZ=7H~-ez_$pXp5HbVJXH2UP$SutEl&VeMK$5T89wTqdXSX~#4v#D8JqN4
zZ(x#)!~-nInPeQ^Gj_GPc`Y+{F{kIX;lMt%-~$)qk#Cb^&>}Sjf7B^q(+Wg-%|YbP
zds_if(=*pRi18x+xxMq~S0lK--cAOwRfLQ*i_1NGG-#E31!V&0vS%AvKFLflgEInK
zDXD7|jpe|V@Wy8a*?ajeS+@CgbbjxXQlEU~7Uz(RjWkoAs9;6bIOQ{sz*@dO3D61*
zNx#!>6L`tasb*#=oMr<-*Y)=^-VzXw8UUsC&=NVn>3GW2_GGJwVTz<VpUYuAoXTgO
z<Y=8G+bvR^%G?!bJ9*Q;ThSJacA0{})>rOtIU@<^Eo0C=03V7g8$>vCc!}AYi3CrM
z#z|(bvSC|;_6Ng@t6iT?b+|MDL_aQ@9y*u)@icwy!N1<2x}Ipy9#wrCq07{%DiQ5w
z6zzpP<@SySN_cb$I<TMR6-CRq-AqvuxqyTF)m!y<rCjr<;*b_ssPSnginW8LHjZ!$
z3zg!bF!`Vl{DBMSG+pOyqo+rPSKwNBhh*Z=O1-u)Qi`@jCgByPoTqGwFB7^(DCvIh
zD*#R2u10qWHmCYGPYf});I8W;hKR@#<24HmyeM^mu$3X|0>E=X7bXY<L9uC`0P{7B
zLp2FvgH?^kF55o?V@9~0Y95w@N}9|=oc=&hAM0^$N<qz%1kp~1??yH>s!J`1kJ&DC
zE>#5W=9WN#f&*}-Dn1&{T~@ZXL$z_{k7@D|;8)^)1pXUiVu%*jPGWP%y&Bvc3c-h~
z*ebIRs7DQC6US@7XFnChTgjAL4k#j8hzDv2mRgrttqij<I_t-&;F`KJR}Xr?4440>
zIru9s3OPn8wp#^TmG{EB#R=Ivj8-y!HWu$YP6q=hTP_GEHk8cF77LW>ijz$N;#g&q
zyRcdnaaxl&A*~hxh$$9ejJiLSVEtvo>WkTF5i|2;^MUB&H~^(C`-J!<+5tepFl7G5
z!vySR#wwtye<ZzG_gbWTqpHEU=5B=H*S13--(L<~EwaHnV6fUk^`^kOTTilqaodb}
zvJT9zD@_M9<ZmQaUkNNO&2&BZg0k@Y17D#WZcjyV>FH&h{259r95q8vWg+P4OGFRy
zXoqKP+&o#zgN)DjAOk?eoWRrXa71S^5;%(`7JtDd&hhh$S9F9qyVL^SSF;Q6duNV$
z1&^-a9B*od1vYr4KXB2&_sSbsxoB`p@b1^J-?tT&JXqBT7SxGP=k5SvJ+z+80y6W?
zy^4hrtKCZO2F{aic_8<RNPInsk|=+<BOCim)2E9jnt@)Nx&x4YB5h3<g{ZI;6abf2
zkX;!_0~AgOjal9DJnJJkJ;wH5ojYAORdUZtxw@e@V5=g3o!E~Y>|UW1^$rOP3L~9M
zK88i?aUJoM&Pf}Zvx>b*h`H(>C>j{91V~u*A1E-l^tOw3y9cG#(^Vbp%QF0`M~kIh
z1Vk88XpxOIPg(Ni?)<vj@N1H<hWv`iRC`&H7s*q0CUAzfl{5X06K^?e*J#|Yu`>O(
zO8zyw+6OA}(tAV!FDbV;h$`#gj#3fA0N?-~c<#zsZb_CMwPMwlZt3E&hdovLIH`D{
z-ZIcHfZH+>P7rJfF6-5s$^r$qz`DSKy)Y<(wgA)#ptns*8kOC+_fNEpKNs2P!xhM@
zDROYn>&{T^r!2TihRR~FSq*nvX<wkrl}V7OM}CL;k?GK4Am)b!4TJUQ?z*X&GoN{?
zw`@gzcqn?m<<XE6_Q!viM{Q|(e{plp|D}q?gozTu4dC8?^pcUMc@}fug+y4!1cv6`
zOub~*#OsU|h0*-&QtlE2#-7?Ixe$`BpSXV^Huc;&yVecihZ(skUJ?TY@Q6bEHYHB6
z*Id0lo^K%MzyV2aQR@|3B)@{O#AYx2w3u15I~P1kmMf#Mz{9dcZMJHF_Ml$DCxb@q
zoHp#x%(zW1sCL1(dcp$J<JU$TdB1>ctGFh_?*oGXqzj4Y`%zMJkkbv1?)U-1$2-ak
z6Y(eB`9E+K0v2Trr8NQ8g@yW<SiSlBVmAMEq;W`5*T$ClWitJP;|F{JMR_)K38I>4
zXp}!_x#9{9n2RJBB2WO6C^l?R?lVu!(nadhzLXAt26-U%!vla$;+S8w$c1RJAKvg@
zm^b3RI`Q_62mlYc`nN4Frs3m29MThDaRj&|ke*~DkV3n<dc41ETU}cnd4KD}X80sn
z2aV^S{lo5Gd+;x(y*av;tpnG?)$q@DBug0=&spT^XaR3bDVVb97iyCh-ZL|14@3&E
zT9a=?F1!KQ^?0u`gxLaC!8AoqJ|#m5{s-sBAR=fGAl-m~mDPcWwl5&ZW+qgbk6&e%
zTuhWA0lH|;fcTlSBtYs7(0?TJiC3u(0hn?kdc^B~q2u;wY1*Dsk#K!A$FjGp-Tzq|
zT1`;7$&|E<9~U$!qg$5WZd&$Zx0ODSqk$gYggmmgS!*y3dUvakHY?v>q7ytX+RJY~
z|KUZ&HfsL2{8i!Y*0lKKx&Rd)sODCBT8dERhb@P1du+W&D`3_Qb*pxnSzL!vt%I<4
z7>ARA5A5wH^$+GKgT&t*@WEp#$pHyKlrx5gH4)<uk^sClN`4FG1cgN%PJ#_~5OBhI
zM@xTJc>lFE{MUTa^j{bP`(vHzcy`{|(#W{rfgpk^<n8Drjoorm>Q{Hl%K*h%?)6#_
zK}h9Z<pFyBp3m^(n+bP30J@NC7P#M&IH8elmpS(1u<!Cu9hJ~fa_7?sCJ5~`RG<mF
zB9IuQ)YcZ~opPYOwAT#F*d%0A0V91;-pAodeIPRp(dS9@UWAP`HQ0+!ns5s#cD+IJ
zBU@voDPZBSLI;K!lg_648jSR~3CU;Xj|GXRZp#!&NtZj{qvt(Rb`tu&h1VJo#s0#Z
zi~WPty0-}!mW}}|*W2$Ir^Cby55nI95XEL7t!3u4$6`N%!k`i(s<=--f4!CdT;xCk
zOEZAz#ggVpW`iqDC(HF_H@&^1CgY~x6qawmu9BGKB?wZh*gxR)n4pdNz$+%+&hI;8
z(-kC}cqQNb(Dfyr<ge-Moq)@MQiA)JM_cdUl<?X(7c|;x*yF>khi8Y>c6)9Gzy<c~
z<&>L5U-X*lUTT6U;YIRN9)sYjG<zf%EB$hM+FV^LFmida7xOaCf4>mXO4ydk9}USL
zHz-1r<IxcH<{(;iv+>&O&LU)xkbNucIZFT>UQnuL+|{DD|1lrwo)MPy=`dU3^ozfj
z>7UsD%Vz*T#H3z+DVJTj>^xT#dS_jOMIt<-TlCQ{4a+5Xte~~ONyo!G;eapRNPut3
zLZeA%G?_e^kyLeqVF>NNx!{trT)8E_skU>`EAHTj51JZP53eKpC#1w1CW;}EBZ%g<
zkr32ieANq7b29^70HLPT0ZyR)`f~wX@-EB17ml04E>e8Gmol}e#DToJGB$AtFSj^S
za0Kr>TNIq@OVN;{#=sbr#6S<SHYhM*6`_6=_D>l2OV+;><o*Rb0b8Lu)&JCqUsWyJ
zE2M@+>!@eT$&;OnY4*#(w3~36Pv&9)L7*e{{h;g6MrdTj8?yIseD=2NI7O<3#fZ<y
zCCk4)7cRcQU+O-;qvqrzAvJ~3dFwMZ9B-Y8k9J#JMrAAYLj;ENOCkbB4{jxRsX@H>
zEe6m_0;ZOk=@>sdG`l|7K{1QK3c2%^vQAMVFD}{FBS)1Ta+6_QOsL=_92{Fl!k0M!
z0SH^fnqMe&Uw@x}E=T?er?sJFesb{f8iiio`s5o`s82af8s1ta38X+(sRYl&@JuO_
zm^DAFtF~>h%AW6x%cJ`2E9=I*Z-4u{N&fQOR8sy*H*xs-W)<uX(a9oo1Zmc}R^-BK
zeG{jv=#hQ7H6O<35_2=5{Wh}W-fR;<?tHa7B_VR-Hu|TPzkka=v;UV5H4dNLy=_}>
zLkOW1+TH}jkQMOP@@hHJXI$b7c(se}REbY6gyqWd>S!4$1v|W$CTX=`zwe6kQRpTd
zWT|{Wn{Hm#KqQ{~nbU-UgAIJgMt<yQT}frgW?%GNxUYqYpe+IdCSyg#OM{XZvz1L2
zL&IdtM4bAhVHEig51Kt_Tl|VC)lqh<GKd#o%YyT5OSL52*soISQHqFT)sh*%|1-t}
z@*d9HRoJ82WNcg1=?em+)<TKqy*gr>kMt!HUIv~@o%=4RcJ$~?tk%Xw0w8kSt||KK
z#Oj0@FW&@KT<Pkf;o?hg`q_+QT3(H^L>zP2Zu_DyXfq<qgV1E$-qD^@ZR(@u?gLiJ
z%#Wb6;Hu&k+QPjIXj$v@pjs?0-Znb<y1P<D_BgnMiW&*nukyOx21sY$T33n~iu?Ne
z_>(<#vuk9uWTV}E)g^XcugT8!4|$@jTQy_YbkX|^-4nc2{Pgrkj-9mFRmSnC2Pzt2
zOyM|EvP-nY+h>%AyqJ3vG2gfSA0GCz?Q6HxBpR>Z*IpR#q(Q=dqWfdoDSLB|a8EyE
zzy&LHvne-9;OZhK(IOie@=8x{j}|L8i9Ex3*Fj)LZ@0qz4&#3OdQ1N!=f9lipLagg
z_=Wwm$qdQ|%c))>E*TAw2CaYQks6K0zq`!*6!n>h(@Asrm>UL!{;$#8EzUzFOKfS9
z-o=chiF+e<i&-n%DA-E&b(-P6ptV1TGqkOJ)eu_|pp`^sM%ig$K*OJTb|OQpfoa!n
zMbMYA@*feK|FFRSeDF>9uEq;{mq{n*xEL`%Vp=}FLVW7fBkQR|`dK@*dh4)v8&{jY
zd)eurFNOj5Yqwv_Pep5ar7RNR@fIffKdD^Y@!VeNE;`yUlqOt7CWpCX0dq;U*<e#&
z?;e8^5P0c#Ui+LMhVTZpYKi8kt*EuV+I&OGjt!Obna9Eh;0urn`1yY><==Dssp#Sl
zb!@UsD<db+R#s}~Zj)R}+weden7Xx|s8-?>9o|)wTgmQ}AM-E7$~eD)yVAaUzMC5O
zzUdfg@mZPyuOHT4>vs|76{BnG1A{gDHX3szH013ZlfOR;pu@sDs7SS?!2LU{Xr0$J
z?ZQ=Q%HV2;Y1Yk`HlDx7xyJ3A1PS`64fSuHAu{&A$ZY_kCLgD=qUOY3=)wQqMh^8?
z8<|#hPXwiysY!!Mp;s`EOLTrot>#E^F2B~$ldBR6cm=Wci5!Xp)mOCTr$7;hP2QA?
zD5n9TzxkDsSKE_Sw2T51tLLv<46lr#O;Ot49DC7DJ6DJkHdEPHsGLZHI{LPY28`d$
z2rhpkQn0WR@?_qHHYyE(Hq(U9cHq{f0J2G$$<wI6*Y7{G|CbMkWzst0diSzOFBL@x
zZmJNC+}Q_b9=uj@^_xoI54222+CA2vdPhWGjd}i1$z3e%lnPCR@okqBs8uOF{OYy0
zJ^e{zTQ%5YtEGtnG|%p#b6cgi@E&`M=!l6z7sTr&me^dWB`MEqU2qEu*SD3-sj0`m
zU_Ez8v(#AZ3-WY!9w0prIgVR3Hzy<`G(#;$62e~^^mvB$XwUHprusxA{rJyV6Uh4=
zIKa9(l6X+B$nbi;09CGCs7)F|&*EhOjG6Mdx|KT1;dg?ZL9|4t4whB8EvtIB378)#
zEh)(YEFgRL@sIY(!^Q@(-;4?SZH=N^Ef>ay_VYKDt#p%%mJiOZ<f_3feD92Y9OcCS
zf9$<^Jez6XHs1T0PTOf!M=7f4l$N3>v2VR+D%Dm((};bkEeK*w)IQV25*4FrNm7h0
zNQhb^NinV3w-P}_sH$2+?Mmr;b>8>)x!>RY+|PT@{k(tNzxjRUvV4-$iR)ahb34Ar
zaRlSYO2mbi6rzG#QtD}${-&FhrKXAnkmwk|)b9QIN(c*OXEmN!ywoFJv+$(bEJDpp
zPf&Ch`6J=M?=RCIYr;KocIdbznavWveA`SJi)3~Y)cVXM%w@$sd_KS;V5pfXW}gpS
zD@u}mkR{igX9FMTXcqGP9xhGfc?eWkxwmE_ROp4kuE367kPQlsVLmn2jy1>1XJgEE
z&Nwa*7C0#4bObri;&_f#w*|O8s}{1k5nal#imkS+mEKEtENB?Y>QzjaXx<Uvy{lHv
zw2N~&fHB@N@!TbAsc;QQNm2L7%DCZTW<fAXSfou8yp%L*A0AVdD^1YC40R_eKt$)&
zF(DVMjbNiSZml{E#kToFs^VpLyzFWLLy??mh+T7&*eg@8DEeXxB~p$+PZ1H-?taDk
zp^vz;O{WrQzBB=4o<OK5U!(y>A`R%MB{P$#kHcH%pqwnLU;bCSH=LEPcv#-7kFxfB
zn?%J*dwP_Ewr-Hm+$m(hre1dwQ$q6y9)sD?yq8gzCp9^#Rf1^j4^=~eNW$YLEu<q`
zh_IgNm?RDQ3ViGL_oTn&{I{<tj#x}_t0VYaKTFW~DhBUAgQyb(wb=$cX&=8*?S$sW
z+<WXVO!g#RbdA$rdq(RNRYcwpAH5NRS{fe9yF+Kc5aYgX7LVAo$k7{lGko1&@hiu{
zi@$Qmpe@O%70&I9u@4_ME|Jzf{GIGSb+gf9QwvrO152LHG?Jcu487zXpK>uDsVX6b
zEs7)|-#NEb>ta-ozMHRq(r;-m7>iIMY6}52B^(<UaW6Si=qnC(Q<bQM$C!eC#vJ<f
z*S}Td?|OA8?b8XYze9CH{5!*#fXbdKRBw&7<&YDkke0fr>)>?Fqo}C3gW=2smives
zy6PSgDzXyCz?SAR6(C2@d6Ao$CKxkl*I~<jefO27!fS8a`!PpBo5<$s05^f*_Ld=8
z*0Vu*V}ba-JVk4*IwIN91+Zl!=?vRzu0QoxE1^-bg4Mq96aF_hX<aW1l)HiX6gAJO
zJ*rZi&ck@Xp?X;xYnxU*j|?tt2oPNi565;qu#64K5*Jm=`duabEz=)LVDv<Xm*2Ec
zJ0qs*^8swi$CH7Z$_CfnUY_2Pn)|sW+4z)+<a%tTu4E9-Bl^;<_SA!;2_H7~VTuOA
z@q~Jep%nU*N$0zb3g;{WJt+m!hG}PYWme`Z$PvR|jWH17bDM~lXVjI4WN^z)Iqi$&
ztVIg5c)`2)Lm&N04{*4>n0QUcw@1QnI-1rkK^X6{vD1}y%@o#%{Q82HjI`41+?^!p
zXk1Q?@Pm_+vr}TW>uh9;^jw+(I3vO&^<m!J$PW6rjJY__Bm30mA{JLknah9^tuGxV
z0EQ&_bTP0NK#d4k4Oa7XqHS<z5FuPyj#%4JKO_=>T=Gqp{rdEY<R~gEbnX27_|={r
z_tMj79%)rPx@*{xN~-F4Ygx^!us+jszEkDlaZ^e6gYCu{kdexCn4AZZdE-N8X+f?W
z8B+TCn#Yt~cUx0Z8Y*ezqxeE~1Sbdt?ZUhQcAtVagGi2<5;-~@)GO!Si(7J_kuv6q
zD*1p;O+gc&t*Xe&ZPpmiS?^OqUTH~tI8+!oGHj3pxlIc%4luYM#2C}M3ZF`Q)UGUe
z2Emb$rahdKp5H^kxvS+`9k+eoy!=Lto52@))a9Kfhk!;?dT_jEpK1T-AH1_Do1`=Z
z=8os;A_f0xYCLON>>W2Tfnq2p(2(#>NTlFYdJJ#TvyqBbq;*s9LewsMpMMsMD{c_5
zO<JQ%v+{3$lki)Q{tx%Nmp{Y~S*Y^T3fhNGhe5LG(T23@@i2+<TLZDP;pQrXR$v|L
z+nsCk9^Ls26c$rlRF>D7QSzoF-9P<nw9j)kq>793Kpbk0P$`h|r2AiB^N;sGE%0rR
za3~^Jv(q$9;MJYDqbkO@KIm~{t>;Jx&J&8kK<+|lEGO!W)!4GGJ%v|0M59ePAP`!{
zfBpTB<KJ@r+a+AS+INo=`GNxgW^eIVr8NKuKqxj>C~!X@_CFLel@h1B>J$w1gO?$e
zwe4O#ZH5oEr;q@>d(1h-wxD$M{WD|)bIiv-G;2~GT)nXxIzL~Q9+)P7!mY4lm{LMn
z(AsZ2a(>|Ujh9#p$~`cY)^C$#0Ncqm*0_9ksycAnwP8cbKPbdwPR9?t$Xe0W9Uo&0
zc$Hqxt+aPr<~#3?Csex!yF<pBVj9T|6g8u2&2O&WxYb08KT6EzN8YT~_UFUpC_8nt
zn+6vC^JM0SC9v@mH$DYbacJfuIcp}#3T!Pv#5Ad<mHqg)TK_$-F8TcG^-y&pbSLi)
zKci;Ey@-cReV?c+RElEtPl=WL3`Q{!LOX_@s@kR|Ep|UwMGu+cp7UIEz8Tu(3}slG
zrz$h0Lo9TlF3~WPtSqB*qXsF5>SPO`<4OTgh4(tugoiP54T-fIh9LC5(?VTwg?o_i
zUCdZzfqOsV`BNc7&H`PfEo(sD<kJV2IC_|DfqMvQD1vz?GfhIH+^L}2y06m49$;Eo
z1y>vsUxYg&LPXIUo$e>IDn1`bT+&?(`MD?X_muO0<@qC-^PD|29h6^l3&B@IXqL=g
zRC<p<4EDAaWhy{Qf^d2(`%yuFpuFDq2lc?x?Uixg1h$CXYL;kJQ@e%+KY}<QCOSG8
zfPAHQ0&#)i_zDgQ%QuWf0-$Er3<56psyoV&mIU}bF!KQ^MiixV84khrUHN>#3fmW9
z2%}P%sSYLFID@9mnoYwPkkTS{du1*fmn->fZbA86GobWQN0)RZ7)nO?T`-SBXo<85
z<$9zj*I|d+Mt3~yrolZ|6)rMLJlN7ZTW*Tf6ckb}@*zG1+_>com<?L75z#ljRn1JE
zFRZ!v>v+#ViHf7eJW9z-SA~n0Lv@v$k|1Hm#F;wqK~|+6%9>r8*CJ_q<5?a_7BTFy
z$s2;m?3|f{w=4^&nGbAVyIj&Wi1Hk}q>t}2QTOLo)KPE_=}fY;0=zH&2>E;ILq+6C
z)JWJf%rP+!cV|4J-^PP&746@JX;Uue=SO-oOxXz0acbj6b;Oc!Cz?z1rU6khd(qoj
z%LvF8nXpJPTwL2O2vi%>d8roGbAyd?SYiphPIC}7I}bf;%@Cx}0X`e7<J4qRro+&1
z4Jgh}-)|QBe89K3`J!WGx>|IrP;p@WZCgn`OXpA#oytg0me4*fholrJ@THgC*gOpj
zf2RD1Y1s3A)3BqA5NbUKp~E5PQb)QI=#DK<@$K`F-D7Dc0DrToH<kbzPVp_5?YvVI
z!2Yf8+os*mRfJ<c-BQOMx9HQ-6)^lpj}q;>$hAmcvANSJw*^0Uzf{`HJ`BwAx1%|h
zWpR?F@_icvlkDAtcR3bu6`*ndG|luosMwn1Ji(TG#HqBpnL2;tTA$u5kIty*itoWE
z>eN2QZNStZx=|BPE2^KhP7y&5_dOCr06mvNE(ZY!m)(#bts7N7D)ra7|JH`!{{ynv
zpK>DF2_Se?j5~YpYUifW!ebDILg^q>(LCsIr?Rjvgooq{7&wqVWKNQcD)}%AJ^?*p
zso-~VrUUOd7ySOJ9&b0^aWrB)lcCdFUgB-PpG5GwHGoN0S6GtGD}HZFt?-pUqtIgI
zNOuDD5~hu)`IPc8TnUYWlmHzQg|)Q5=xwXcj88{X^(DFDfaTH#Wh&juF~9yx@aGl-
z(9{XeO1y2BXsKw4sChh1)Q)wpq02Z9T%SQW&DWTpu06$M{Oscvr?fU3oCC@yfUNw)
zY$tWjL>nbOn?FeOo-N-VU5mD&@Yea;>oz;Zk%r-K=gz1Ue!b|Q`Ft*a=v4Nem_uc?
zuFPx}JwV=WDF|$6cDqRw@{Z3i6)>0Bu?m{&4%#E^l2<0*1<Jz*mOSQ3CL|&`6NDJ8
z!DDPFuJ8$>29pgcwJqRc<Oq*u$=>v_Dhh77$8A~1q1A~AJf!#7A(hSo1l${4=iBo&
zqH4jvyg2D}$n28%4)?b5?EE{kxXTvBrD<xR_0L)*-^8B{uY*W0%2=K_17nA1XiE!q
zX?eKL+<9l$)!fy_646<n@))53mY-;NCq>Pk##Bz6fx;);dw9BjPPCp~n@i)hR{dDc
zs4^YP4>F?UvoNoK2wRO#9|{m2_sOJIeC=PDKMwOoH1|Fq&>r_^gsIr6eN4MNjw*UR
zdP(P)zyJF+OnG+yrm~@b6<>kOWKsf}7DLwu6fIKn02=Z|8ME8$P(Pt^OPj&VH4UKu
z=&?J^#22Nf+`k{6Vm3U{+10`DJuBNIlkH1K!@$N*=K))uSOvlJC|})Rbv{LPwAzht
zgXYsB2Oa2WB9+G2EF}{Nn|VIMQbtet*9HyAHbc<u0fbK@v4a2&|AqGnS30<pnO$`C
zbqTqsZ=*)^&L_taWu|6jMKv4&?W@()pw@9ZI<%j4(%N9z%KN4J!#q7=W9zISqIZsu
zV22%6Ow_&9AA<7M?lTNS19yT`UWb+1Z3d-eV|yA{sF|RyO6R5&gzZY|bCWkFF_><D
zn+UjdZt*KNHd$Dih!JW}@r3Gg`ZeAkJB7sv;8rsTOJ_bT-bH!Xrc?i_iidQiYloep
z8n-Ir7U0fhDC^jzkLBO+t{hq1{CJuDnx~5GJbhc~;%uqM3n7N)`J4BSoNFF$2I6|A
z@*2GCIa{*jB@3hb#kpk*_Klu<SFJc0BPy2}mq$tqM!)k!dcLy>-UI_Kw-jn7+6D;m
z+1fpqF;ZVYV!NqvdC{(=rL`9PG%8Yjwiej~Sj~9T(R>lHelVcG4A8ozbMwtb{)Mh_
z;E!~qu6NcM<HdK1>Mp%gIy=qsG}|Y^6^tUx`K_i-!nsHC+VV8{U}@u{j(|F-7YrtZ
z$Q67Shu70Ij5h@J`bA)}Zw{-%)rLQaQ;WHG+{RMMvvNj*E7+X@A^2Kqh<J1|!iDVJ
z;{{)X#vrS048`X+F=i2h_8DFuk8ZgZU0T7#u9DqJ!#Cs(960;dgAr2HrtJ6a7kROw
z?o3Td_e@R4z$u|XvxBzoz;|IiKpYuxzrKps11#y?J^|Kqi3xz<W>(frU)GQ1A6Bv0
z?^#Dly`k)*1`Vb2{GrgD(MK=H;*^_w{pVv|EssK!L&Q@D5e}EC-V&r?Jp_<EKlsuP
zk&vdm1V_X3la7>+H#IvQwk0mVg<AvWfIaXK@dT<kvpBbk^-vpz)?fX6U@KzfJu-ZF
z=*g6*Ls#oOJ2;AsGBltornjdspyTi>6S%kXRdT0(8VgXEN|eP}k15wIE4^#z^gJ5i
zk!u+-P{#GGLteV*I@geL25;Jze%g5}x%bk!c&;kL8VL|2kXIKGwUVxyO4tJUWZcXP
z3Z=4agma0_lh9CygU2gA*?ts7zT*JK89B>90B@;a`bs734L>0%-Nvng5lkS4?;oAT
z6=Y6QOK=SqBSRZnHU;<!fkcRbyQ1)eTh?AZ#|9Uv0VSDdubzg_so6ZOH2dDc+(7n0
zC?_;s^f&H4R}^WEm#m}@U!-v`qYOyD6LTJL0>nbK`$15v?npQgEwwNC>6=#&1n{v&
zc2iI=1WZcay2T1%o1a#85#pp4RfB9R^c+|AeX%;sMU~2}9oY>9P2<W>`&+wBGh0pB
zANE3>GZw=l3?@yWCdvM1L!_wEb3b>+85_NBu;^L;;Qq+=>Z6ePMLRcUMg~MimWTmE
z<}}ZL{{XHC$yAff|5afZ6_rL)-A$m?TB9toJ^YJ3P-B3wEYc~scQfKq+ruX5JDYyq
zT0D#58KX(HWg)$m2T9feMIE=v-S;KE#=NIngl6weJ~#MyG0F(*W#cw@JUr8-p$AoC
zNv<ejp&IayI|_7EvbBH^Y*^;H$>XVr8ot{-6THYi2qc_;8?5<_cE&CypL`_ff$Cu>
z)Y8n-(#*_KF5_{ADFtP3?{n*VF}XPPpk^VQ-JUL_S=gSYEa};qt|oaW^Wfzd2lW7%
z`8#Jv{*{sCZ+ZXiP|bk@!j@zBmdzv<#V439`ip&RDQ5M2q;(p|hF8oDX*sU;?Ub+>
z_S0U5av5RhUed%%b)VZVg#+%I--NiOKI*vrHg4quz3{BmT@@^I+x?Y)M{UrO;}U(n
zEQ%V=i&a?fhD>zuh_z4m4lBx6XA;$YIO8}Z6VU&53exIrs>U|XlI5NIFV`NO!<Xxy
z>PI~7RGRLhUPEhRgJ%Jw!T_+*Tg6Ph#H&kr!EW0zfO;mhOVLZUWiC{d8F1&>2&ngV
zw|;P1!6Vzr#miof53@pydgB}7_tS{Qp?BhiQ&v~-Mj-)yd)5|@?R8YkoI<Wn#w}4M
zfW6GTL#!QJ+mN3s+Z!s=LiW)LSq8!!3wVN@K6egT6&YrhoSGI)&0<7sW{ZwjyMx5{
zaIqB44;AkKX98=;bAx~jE@;Vaypmzz(Ch_walNj4XuI|_Bw&94h?pID-)ycR`L_AQ
z4DR&p4@t6kV+9%6dv9PHckUL5u}^(Sy2Eia1tVaTC1oYRgpR1N5dxf5h&sW~VsD4f
z@_t#%p-pk}jZ2>T>1XXeeJ_`l<H8ZyK08G$r(2BrqxLxM6SIvCchwXlUoT#fMxW)y
zM36$&ztIl3H~Ey~&&0BQZI)=)J|8gMcvGLfeD&F<p!!(94V_u%lc<TWoAoKYPyu<U
zg728TpSbWM$Z$N~b$cpspF8jW`uTZI<FU99MBN+@TJFMRoY@l{zV|k)LAZIPb#kwI
zke7G&1~p1T=sne+T}QHS!UXH?C{}#^YB@qCek!?n*Q=EN!-q8MZMD-q*1TeWo!XfB
zcLD2`&c_nR*9}j!or&M<gQFi#DwKb41pV)&$$2gz+Axbfi`P9N$RYEOy@WcvsrS-$
zKxfyCa(1byP5s9ITXBt{Oo<0X`isC`R2|>Yr<S^9BOjX|@xjqzD{mx5Z|q9lpY`!)
zef-%!{-5FD&*$aO@#D|&^gqon02lvpMcp>?#iPCch0qQ?a2&|9vrAtXGV*4>{tijz
z{Pu;Glnpa6LcQx=pxbuE{I4t5yeX<5QtrSVUu`H?AjWJiCT@}`N=<0r6qZPEV(w<%
zuFRQG*&3)riVZQkzWB^hLhd|)T#~ZXw_(h)U}#DkuO@fqOg{~=c&qLFjgxw{ZHP{E
zmwuQU!u^4*dy1v-5A9qBQ0E*|vRUF8vfkj^sZUQYr{qbEqhl$}12*<8tfag-X-zK2
zrmwVQ&LIpjxMA)%^#p?&i@a|$fSTrN4EN7(I_(D;rIiV%7@iTI*413a@Nddz8tR{}
z-@Js_3wjuzoGQJ`CunfT1p>&mu{vn&@xogcc|7v{x&6#uO4>th%uYWFKlPLrS|8e=
zs>(A)9!+WtWet4)$XH7(Q)Lff#7Kn@)Eq$y-FTyOZ)0Wa+ghWN!nU(tK|5I9)h?Je
zc6EqS>+2P@K9jI#b(^GsV?VrXu(GtKju0`VbhubQ0x;-p(S?6MddABShL93Yx?c>1
zacCRVkPx+L6OL|304GdMpfQ0%yHK(S1c0*BS1O(i`|)pU=YRg{4@A2ADqZ($>Xmyw
zGw&Q(h}Vpvn{%MALlZt9NDnB;J`CN~v;vrD+_JKCaY+t1&R8l!??YysWKZc_>GFx=
zSQiqSH)c?ja~JEo;8VY)wAj%Ec6`nWCM=IsIzsI5j<j(yD%aIFeKf>=kd>|NIzrvP
zS+I?I8&j2tkM0G;w5JwWe0#6Vm<sn-=KF<TD)}!1qlXJekP+PQ=F9L<ZUhei6PimE
zuYOAEX*G@AEVy^cKvZ=Qba}S7S@sc=SQ1EXE6!A=-5y*y6aJB%Vx36lPnS+R0G8tw
z4pi<PcSrn&wwb@aomlsVelVqnK&s?WSlY#-I;7L5t|4|}f27-d2<?jy1Z}zV=%#yN
zSKJYRY8uWrK3s!i_s^U9zsm9Ni#Pwv%KZGCw-U;iO7WR3@dj+ghO802nNaf^=9X$o
z96$|SH2)q17k=^=yx$z#Jq|FcZ;=iHY|EzNi%7av0rqWewZZ$~-6wXHZat@*8X`7E
z^OLO0?emps)4X)f7seYDVS@nJZXk^lfT#ctMPruc<CG}C<wo?$2XG+-!U%Bq25j62
z2?_X+?eC7=eV*TwE*PL~rLCh7Qy8ffUlJTFtuM^XX(<x3k%|h+lqT}w^|y!HpWl3G
zbO(n(UhU%ca$s?yp6=M9Y<20&+{0h(lNDw4a=pHCn9!aKyzl!ShS~$wS1zH>wC=06
z<4P;edb>NdG9Egzhj3+RRm-d0wIFa%JQkPdL3fea%(#zObW%`Tg6zVbL?sa|**Zy2
zHlf<{+H6NB{Cdk>YfS-1`mGxK@{iFj(;6#Z^&#NjvCXd~c}J5^r-E-hynn%>_hbA>
z<?Q-*_56)P(=k<uwi9N;4?2a4y_?7Pt?HS2b%s@qMe5gj8jwUf61_X|F%1Hq8?0T3
zI7Ar6JI=Sf<k>Q@(KwV~#=@eSWZaMr@6_iS!wY4%dni%MwS}{rj6Qh1Yp>_NOYrpA
zLT1^p#+G+$?Y3*DgskD1jT%P8s)wIFK$4EoyQ;u1GC1X5d`~t75cy06lTY4BD%$sx
zCoHD9HSZ6$uWSvcNzC63HF#AkUh-z>!U*1Rp>y2mhfWs4k%dIN@Y6ZcyL<KqPfYjx
zHDqw8nbY;@SX#}JZ_WO(2y`+!E9HflPygACoRB_VlrSyeR1B}I?acys&mliLfEM4l
zI_lCIy|p{Aw*L05<-^lw|1uEu<NW&-8v(OBM!3d>!&rz8HQve?*as~)ji<4b5>3QB
zq7L6!{fbIzaj9F-*f!{owVo;8c)d05V7}eg`1wGDrG?u0ZCL6D8O8Sh*FVgm4_}=K
z_g9}X+1;Bn)~ipOyk_=mXzq6gjOU15boJT$<q{j{hzX|?(bF|ek9L>03je&PKkMnw
z{`BW__vcvf=a~8D-16rf{QskM!QtKXHgjAzq3Zn$7Ik04Jn6_UacH9VgO7Ko;@3h~
zz2=^FfB;Xa)^8q9Wu|PrNP62;y8L=+`V`FO>-AaOTOUFJW_bbAb_GooEHQoaW8Sf?
zAd$EC@ER-1gv8|j*r%2z2%jA=tnWjMeblSAH(6PqPK2>@V8f!1(u0c2dQQ)nYe(R1
z%!PI>Vg?`;K1HSR1@TZZgp}PO(%1!yR$2VFV6}>GZnfY=z;Z~*(0X1^|AxAOqVEXZ
z&*w~R*dldfjTFubi9mg%ms={~+*Fr#=-0GEO;?HeRO9f3X$hs79)n+}hQD1=gF9Ow
zvnjB%PQJS40Kg9G%N+Kp{3%#*31ZW!*d#g*+2`oria481nO%Lqq!qW9k6C+fp1S>t
zUAn<c`q_1k>Ecn=;Q8D~sd~tnoS6a@I$8E)=O-9Q(iL~<e6v`z!qK4k`5P-={k1yQ
z>P_PYo4#bsL@^A#LU@&Rx@Wyi${!uSt?tAd%^qbb`1UdC1ngcn!@^e>TJPxGse<<i
zXmVfC_<#x_VGa8=VrDX$OWC|){gfxx7I3lJ0t4rArFSni6-CsN?I%;qA~Fk4FRRR(
z{Lme`KT?e>D+numKC1koIq$*F-QodAij>1c>E2}H)?vG^jXSw^2S*JbUeLBGt-8Z8
zS?01T?x2rfMFsdbQ2nD@iug-*qb9WvBM<9z_vz=q1H3zpc~W~C{Gi}PXtyI31pVYY
z>^BS!`Pl$nT)ZHO#T-`$ow;%h;41;LLpA(b8-8&%$k={1*FBgY1Oz#ovYo^@q9U^>
zAKCwEJNu8)|3If4E9LGSsc56h-LJ)hi;+n+rH=(#BkPruH`FKYd5)3+D&^%g%M`Pt
zK0%ZCQwUpE^&{?S$`mC;Z|kUTm`NUAo$VxbO~#5t!0NO4)J4i8Lqyna(a5wes<^I)
zKqR{{7h7(Z!*3NI>a$KvY-fmymo9I(H|Q#V68n5WM{$9&C`h0_2^t(6w6?OOQgb!Q
z6iR3(5trECq8a*etpMl2tww_Jfa;7#5U?2s<Mmp^^`=r9I}NSNZEI02Nui_FNSwcY
zEm#8#?7ty!uozrAUx}QB^GU$NL^=({R1r|PHUb31rf{qP(`$;?)9&sqKu*jncfaDz
zdd!#h+kh&v?fKW5n&tG6>=D)eo4HXH@xpPJJ6kUWWquEp)pHltu5CDftE){GfAYdr
z9kQO=rrrf;0X<`(vUZ)zu)Mugkba8CULl9C(d6XHl4{!gY%4*cRw0QD$9@z6-++K=
zume^48oKI8Iv=3`6ZWu4r;=cLxPqP;`6PrA*)MuQeCMd4(<#+DzNPZxxHwY^fOB77
zuPVHrcJQ)!=ApCOz&S){P@6R1ss0ECK=@O5z4ElQSdiSNHFR|OorfcBV_{e^J50^y
zOHuc@gk`gf35$RZ(Zg`Cm_t4(6gc4nYXJIc5psrd=hkG^pZ%kr`}@+Yj%Ab&RlPR(
zd|>YqaFBTmbP)Kx!>)m=fOwf-K+`e534i_bXGmZE*b(-W`Vism_u~%r71ctmLXcm9
z#6x^Xc>|84_HOfL=B7yMU(S6={9V@x9nPK;yZjLQ(p;HTHBprdvRL15Lu7%jxiKNF
zQW45h=!_%h+tbvK%&RPg9W8yb$ulO$A;7IGSZaPP&xhhmnR^GJU@f}FK7siW;_Y1K
zHER}wu8`*i9nk8&>4tV`XY~T&e4S*H?MVHQ4<nsKNOgMdc*#&=b(J>SRvkR8xd*rU
zwLkPzgAKfcp7_&4B=w`jQlGMSe$F~vWRc1UE+zA+R$v=7J|zh|M`6y9LpX0UTf^+P
zwz8HO$Ifxy%bXi+X(an==R#*`LH!fjCrwNFr}gjKC<`!pY(OMSBU6nQf1#t4(#geM
z^tjsJ$e%NQJs3*w7yX6ya(0UBx@6qq1s#9x+))c26R=a0B-Zz;okAmxbv;MAYb7q1
zkK!nmsTGd3UbcfD$8zjrPAwby-<c5EuB0yeC1u$nTn0AzOrV5fj`PmF2{~OmMb`7v
zSQIcDCPESJwwqVxXl(xcLxl)c$=}s1f1gQU{5W7Ch;?Le)cE_?i<35TeQ-M{$jIm6
zdrDn}1xB2&lBeb=?A^6Gf8qM(y4ZG`GKb*5KiF-pH-%I2Z?z(DtT;MnQnID*x(tAP
zCU}m2?Y}hEQwuPmVzGBdE4@PfA5lJXEUu-MBG9a~6vPPmd5?-B%ZK;?E|Q+))OiUQ
z%KeWED*VcScqBNhTvCe&?L}rsQ09N_qsydS11C`AE4rRKv=>lg0N>&Qh6uunQ*h!$
z&*uZPZQlyMC3CVb5e(X5tNV(p=CN7)`8i66$<z0xkL7Y79`_5roPK=h1@@lRPIoEM
zXRdTkeO<G-S(C_%pu1ek$Z?Ihv5uZ_I!)i;6IgTf5+BCU(u}~E`w%H8R}A%DA4>{J
zbt9!YPR`q65a`~@LPr*MGds<RS<1@%DgCDs<V;Na_>RTSfYRVPm!vB|K(oy(SO{&8
zJ09I3^EO(mOBrgf5Y{??nwCTf5Q)jETehR~A4OQ-`TJJ*KfM3rMAChV3GdA)i_fGi
zZ`tnnU6=%HxNo&Q04-^+olBC6(3g?cXcKa8m3GetNxHoR8;N?g#OayOxn&Ckhe1Io
zZ0Mri^w4Iw#Crw*w^MKpiGZ<$={{<&65ULXl!?k8gFqeMN&qH2GPnvy!88{vVnks$
zZk43$A-$e#_#^_Nt<xGQgf-P+-3gV7Ey8whIgJDy4a-QMXWxkbsHvn>G=(>5fkM@V
zE_B}i67v$L(T&ru>?VlY`50Ks#urb`)$8nJZ4UxwLP|72M-FD3Wfi>O#IiA<0y1eK
zVr;Q~!1}eRy`7aF!5T&_*f@SZFp#zHuhmuy)##Kn+#l-(l;vjFx?ku@CkCdaOR>FS
z^Ou^C>FhyH8y@5Ze&GB9eB+RwJKO;wrWY%hlXHZr5j#$j)qHpJr9Qjpg><p~P+*OW
zq5SQ>3M#TYI6on4i6Ai#h%hg*uQ&z2Yx4sB6QXmc2Id44Nlp$?+Y?tcGnmie=9cGY
z?aa?-0nCNi87?G*kU-GyApF37thp!dQe)osjn?b+4M31h_|&P5w9E0lQ?+`N{j)vU
zVtS$HW~YP*w{8|^zD`#@68Neo%%OWsDt*sfr@{*xQCAKJ=~R}%L1UEy)2)(xh48+M
z#0BCKVsU{${Cq(4<d6)}Zxr7y07w&n9xVnFBh~m03zsg%x5v*vesKN&(69g5kN$1;
zDPqFYU$306N4vQA)>RZSh*1$=(Rcaa3e72B3ZO#KaDB)(G-P(|_*=4H*0Z=N&-M`c
zlFo3n<6Sdo^7u7m<3M*=5yM`Q9U8Ww%~W*{q{d5ra^del#a7NH>WQzemvmPHn;U=s
z<a8)&2YHI(n@E+o5(F2o&Av@Rk}B_q++@@G-8R+yS@o7}$M-@47chbbUAd@t{pyGS
zc?Rc`Y)m;CmA^MiR!r~GgAcW~C2oSSn(A<EcJLtE30S>J|FgRN$3H(D68K?4{FlWO
z!5`HfG?NjX)ynQRH|}t{lg(id#EAFF8@0i)C+xhgXG?1}#d&)+Z+j_?DY^^mJZ}SE
z)$s^=4XMA4n3}nfld|hu7<&}Me^_qcQe6g48qMe$(=CVEY>v-WBAX1WXBl8^=SR|j
z6^}u$wup%zTT}BQS*fEyYJ=G3wk#w_tJlzuvZvW350iU~<DoFI8J~?|nbGc7tHsRz
zvoEI{bDa9*aLT!^+HC7zMng)%4Xvw$A%lCyjj;EQDJ7?oJ*3s_ow!RuWadb~`bjnG
z(e`sGN01$OPi@mCFLLfdCedyiLM}~L&bR6aT(xBC3^A&8Wmz?Jwd~lNV`l;oz5SSM
zK>Js1a^3*_Wjk^T0%1GCu^gvf0fWL>2m$n|htW(_Htvp>8|tnt5r2GM^YU7yz|(D_
zAlF)uv<?Pnmw@9h2MPm0VW;3W2B-vzg8XLU$xHvZ?*CEhA4X5_FCWx04UVZcRxN~<
zrke%)QcQr`zzt;|ob>2wE)X}_l$;WSxeqj}IqHEm{o3<@$khUzWKhDZSM9wWmGOxv
z)<=C^rD`Ncq?fWl3pb5!Pgv#vysUl*6cC~m)mRCh7B3$To`wM%25p4ZE=;CjIbg}u
z%&+fa0^x}dDx<&#rQ=F)I$Oo&mIRS3-RWlHXLpPhG+fzMmVns$eW}s0J8}MyyY3P<
z;!c+4;q^m5fbj^y@rsn%_0FIvYM7ziMOKNYDR!(u#HieXkdW0@c?v3Cn&ad{T<X!{
z1ef3r!+I+WJ`Eeq)tVXj^-s*R_Xl^KB0idOCZ@R%NQ*{aHwX2l4G(t|6j16B>|3|k
zJ(TjYTxDVg(NwZK+azlC!#%6gz^ltGVnHnr^nmjX{FUKPeJWdvQ+6ysu`)^Lk^@b!
zKv(`P@*af1aICiS0nN?W)o3^r!GZo(^cF2PFt!tby}~XqhKxUk%CVK544zFMH`OeD
zZ(exv{W$f`Ecn6mz-Dr$e<9)1{$OG2r9hmKf$V6>+`QcE_lqary=~p#XVOgyKqy~&
z(OZO*q`2Zp4-C}#oI3dV05tYVFrfN#)t8Dk&J3kZhmn2O-*7h_DNP|y_%!`Kbq1R7
z`G8w4I?{mM>s9^4th}VA`ych)-<kGr<H8C{zC|!blXDhh1Bn|b(!nxrMzEu`(NE%f
zFCN|AIBe`#5`4{^_7W1iGf2Ad`M_lu^4LQy<0E9As`f-@rjQ2Y{&!53K;p!idH~tC
zkU_$<XLo1~QgKrW3#N#=x=pofO|+?=&e+$(xy3#VDYP+nM$GfQ`z8V;20ioYc#AcM
zyYs#l749+hHcVP=eYLlnaZUC@v|ZK$n6x$96r2lqXYKMg?LuY-5mpwZrP(@UGTD#2
zaZ2=kn?cn7+-<YTU-Y|?Jp?gWHxS8hCLQE1X4~?}_U%F^pJnaOt3~m}0mz$#EqBN#
zssh608ow#Mg0)~^tuH+Z41AfInqqx%6fz)>OQ28y@*{-|sO9E_fIz%2pEiDdlR#vz
zV!?;s*`R-)P2$D7_N5F!O|YPZ<ll`5;yCN{KfOq)s^P9h_@w&Om02{o_}6jo0wUz>
zl}axa9qi_fEcZpN3|<RvLd}s7e0~y?y-a!WQaU=X)wsGvRjo%S@NRj&OEtA0jkju@
z=hpa)#yBJ35HKsK#M^H4gf#&_7prm%)<K~FTu(NkL6$BPkulo$TAicq%;Q=il0nAR
z#q$`K2J~1+4kk>9W%qQZ0n8fm26o^xo|}}@hyK10|MmMnj2g_9_j;2+bOc_{HSWvV
z#MwwdBKhM=+u8S9&*b!16Yx>Ha`N<ClXcU`sJ6c>G?q<~mP9<0jzesxlbp=<Q(?;L
z@0^!3gIcA%T*YPAyD;aYCA}xNru#?_5cTUHc5P1Eooj9&NGWKcMw3g{cmNy5nwsKS
z{B|rMi{2UO-Ps^RFRr47LToodeCE1yWobBNhOh__IGTA>E<$F3E?-^FoygDG5E-XL
z=LS?x_a^<4`Xf;BUok_!x)I$saHE%jBjrU^r1n?C`tA$ou)jAs5kv71iDlK>+-w^M
z&LBJO^(xaP%yr25R%PLDudT23ZbS$ZB6D(+LozVCm`Ld{5+*i}T(Yz17coWuNUvZI
zClFNul3Wv*9ht0=8IwCnE#6jExYFNm6&<>kQbD`If~@9@lI4qs(r<mam;d(l?^?@#
z?HLaHxHIL>B`10ul`1JMUE6#!yCji#%N;K=%+qwf)|37%th>43H7?^!!xGy?Q#MmN
zy85b_xX9H98dqh9V+xUDErWV9O<@Y)oP`i{;LyI#af~Tc?_<Q~WPdp#CZ{(jBJdd~
zXf7v;l>y$)$_dh%8ooZp%fOOT?}On`ede@;THAWY%H>rQPH~AX5n#{I*yo5@*5sS#
z)k*z#EA*Ijup4NjSuJ3}=v|g_Wn5ng>29ypMyQ5+N@X8w+Hd%=`VemHa<jAnXtw0L
zZn4#sq$7DR!L1bqC!hmuhbl~{YGbV&13kfB3{^v)=bp%v>CzY382V5feb$9t?FZ?$
z{1CRP%tjB3bVia@bH-sBh@IILQgG{LcQcM*-}ky|j+TO=)^qhT1vZ}#;Bayq8UhR&
z!R@P!iPwh={r-aeVQg=tv+IrRKgq}yoSW{<-7_KE+{jueH2XwRwtE&Dd$0p99X(HB
z1i2>r*}A_@1r(+}ALt~(bVs++;z;T}2;P`|QOQ$Z#752Y$S0dF4PYlirc%xxN!v_C
z<jR~nT-Wd--7J*6(O_PB%GH+2<w#cypY0*L?}=INB20JnH%m8-tF1Qf=EaKrrh#J+
z@YhfhU&f<sCd-HEd%Q9B8r#g^#5<>@lXA1}CLSWLuL<b9FD{{N!jCCcXI+ydr>~c$
z%M`u6wzgK)5+P0q%P2?-$nXTlg-g=Dsrz_hel*?!EXoydt8JX+R2WDm>bL}SY2!<H
z-B0-yZ)PcmYQvUvjd9||oT%LugmYU16lOfymVW2=xAi~BdGx1X4s*!<Frd^}lw-l7
z9J=!(kZVy**5#}qd27<2Bq%<AKjaRO_7|xr&mFeWb;GW(Q=6_?FA{&N2Ojl-mpwJE
z>2A!x{!7j8&g%g=f7gYu1gLV^FnSN=y>}4dp>cs7=pha*>Av4}nwxaKIWYB6;8cqF
zqweoIIg;i{(gv4bh%KIL6Ov0Rs8=u9ErA6jM%@`8D=iU7XG{mZ2?+}=QphH~VL2je
z1=3p)P$9wvi9t4BY0wMP3Dm1RJDCOt`#J<+_v!NF#?;g3aodgHnQ_4`1x#Oj_3iJv
z<NuuLFD9QzXT*3-Cl{bEok5rHtIWnwCs3E~OFoM~jX1fwZx{bfz3VdgaP@m7y~zhI
z9h;69rT|x0NvX-5($R;KPNzT`6dbbLCE<*rCB0V;sY16drb5_T%1aTR?F2lRj7nT2
zIQJgllAeEG>liYfTreL-nI50EvTYN5lJQ;td|){3hfJ-qze$q*=X?CSlt0SjemKcj
zr}R2|L9teK?Lv2tu8+nGztj|S(CA5x*)9UM6cCORer#bZd2zbCSrUbM_W3}yCIt8N
zT!oI>YX+lww+^657P{I%?ZV`<O}rUggq}484<DlGBO)kX*;G(pw)R*pNQr}yd|AuM
zcMDc~wRLfMLmohzQW!D~*K<0ln`EU@>FSwYO0EI0^6kXo5~~&a0BZv>G~2-MZZ}EC
zul;SS-g^lCVgX=Y(+c<+NC&k^R(gKT!Y=aiITR&2iX+AYSj@kzdPM+h15-0V9wq@m
zn}C4M-u7D@gf@UP^&kMi2@dK-mVR0EP7rdqIk7uUY5m(k!+kZRl?IvL5C<IP7$eZs
zO#jrb>_{45e?0d~o9rVN=F;8C*PGK`=HfECO%JsmDOiEyG_EQ@=0OqYE<+a!;w(`O
z6&^Yq@RckL-dVC(p|GLx`9OPRZ%*$XFAZ9f19u9A)bLHp7RFdUl>xv8>&ho7C15rI
z<XjGA5gXq@9>4Ct?chCEzAm;}da<PNCb@{5D*Ww@R5^&`_sUlFfwQ6K?nz&67EWnO
zSp(K~Q}|0Z?d=WeF-sNz(B#q-%%aoA2ozSP3I#=Bk|WnZy)iRm)8AHH`~ot8#agsr
zA^^<f0fztp)P6}6J#F^yy6eC8_TNUxe3)~EdBTQfq=gBt>-QsIx((!`IwNMhiqq7&
zist#9+Ueqs%PtQ9|AX7_Mr(LlUMejR+sOXw^fn<U;S{KFSTuBB&Tew+W<#dCuG&*$
zQ0yIFBa>cMaAWzYcfQuDGqaLwGO37gQt|cIAIeR)XdY~h@oAmRWTvWp+lktn0tZs&
zdgRDx55-Nt8<vEy4pI3e;%nK_#j<LvHT`_RLC*#T`Q~4%%)jLp|MZSJYwQ@7J+OqH
z_9Ljp5xGy+)fJuP)Au`Rlt!BLZb{8aWXtPWGjzH?`<u1A`{cO~41fgT=-K?h9=h|)
z$h6wx#F4v}NtsZarRfEu?}sn@RB-KTQ;^uDkq(Aw*p$NaavnM{$H}gx;)$h0wUU9(
zY%P@ER*!bb$|v0&)w#gMch1vFdiwGSyEWhj14_Q(O0fP`=6Ubw^dJ9rr~mBL9|Uc`
zB9RMfpnf+!PZjiDlE=$FO2P<fObYF$ztnSU&B?IV+x^gtzYNwS2bhbWRC+_wm2__v
zy3l6KBOXG&8(f^*PSs*xdFkl^1?j8=l~Mo}Y~z+bJ_t&Unqu-PMEzF~4nzUz^jzhX
zGlO@<0W4zD$((6`#GF8vemPBAB}|K+KAow-4{=&*+zT&h&2MbPf)w22&Az#PVKUxa
z;c<>oP}@jpX_2cCL32lM>cO{@kAUqz%uL*N&fZxzxp>N_US~JQL|X)EJVz5m?`G0C
zB8#+f8gjamVTz$lL*T&2t<w@q@A$NhQ6^^?1lp1_?1yRy;E>=1z~e{e`OMSCM9sTj
z+IH`nMW>CQAU3~iGdx-wY~^J1ab;3mYh5dN1ppE_u&bXBAi+i&uFJ8bBpv<R7ed+O
zu6s(?80D_sQvQ3E1~ltMb|;^SG^Bxd@*^{hvvT&r6cp(VU=c*;M;I6`B3VDIxeRj~
zj8rbJL+Zo9A{!hIa3`M+l&z8tk>?ckuPTiOL+>T2bEiX}4E{VNHn%c)qD}?%-u27(
zf_Nc=@U^LUp&&rgRN3(hl=96<&lkc9z7f61rcw`-9cP0!o@3>JkM5K00crTkc3Ted
z=%Pjq%y}Yh806GRp}cXMa(Gn<PJ%6Z3kW%Ti0(+``7h(}r+j_5OzO9L2DiUJI{z2f
zzinjPA?@M%h<-L@)mi0(^hZru0^KvGR{Nb)z8SPJHy+1h8fVaDX%Y=O8BAP$w6<wU
z!|XH^FBH89BVr4(3xTy#9U(pP@a=&$4UEhcUD?tC`4fgm_fG2(tX1%%>h2S@XaS|5
zjyOwpf4<s)9D<VSw*+i#%WqOMjv;_4TMbk~by#$Q%v2_#QqT&n85JU`Sx7-Vx*Ap^
zak-OH+EtTY`94~udkxSccs^M~-yt#x7F!3~^Io1#AFL|0njUL8*hoYucvp_koo_QS
zC4goy-2OB?Q~f&rgg;(xF5Rw{;@p<37&fOfpGdvRAH`6$7W7WRMXlH=5v+__V@*oF
zN<*R78)pVoLWRbcdN<)<sp7fhxF`MzfxxM(T_dGhVLg55^OT?p)9zs6EDlu}BVYBD
zq+60mbW_eb>+v5c|Gmcl%Z^`WTFAA0<fn?CHNlUYG-sp*S(iTAC{Hgq=6W%l*G}y>
z=5W1gVK-YZ2I9%gMhaM~#0_`EAUq2!9rz+n!}a1Yrbpgy7FZ1oudS??pZJ~~Qk(1!
z@2+4>-SHu!wa1>1b#1Itd}>otP}8Fy;o2CVEHZ)OrXVjWs(3}^%rW7f@YCU80(ZBm
ze(8mVn7z<ZIyzw)vp}Qx;G+uYb#78We^?$>t#)VqFd*|B-kVp^ZHMYH*eJ>DiV|_P
z%+i+ELO>d3jy)h$HXq!f95qjN)=?4LNerIX=*oyRDy=F4tBm~e&b@G0DtPCwK@|}z
znu61kqeyD@HDNA$6_v;!3;g9c1~G}rPGa4>+SEw#Zj{tR76K8_2U(n-rPWtdShr%;
zhUwLtjhT6HEyt!56nbVn-#B^is$6M9dayz`0wFdAY_=&SIMzcdOhcMs=2xa%@%x+P
z53=fx#*8imRPub$TlR!d>)oh?dDOhZ)ay9MlFR#LmuG>Xs+LEZ(>L!6A)zF(lSR#9
zLCwM%dUwMO;t-E68U7}62c_ziEOVS(u#tnn;V#Ds1A(IxD9ratWTc7OSbh(4X^a9`
zIU*^PTohpN+UV6FBK;oeCSMmWe-*xQa<YVh^X$ZmDwS*HbU@wfS|rSl-ZXM+F3>@L
zJM~D!ddU4C2pD%CISB=3S{@AQC{BOf$uFwYsB0uxsLD7nGI!`0p%OfA_#+l(+KnlY
zZPU+h0d-ZO7+Ni1zPAkGFyX`R->s4PUx1Wdt-ctqC5}x@>3#{mJ)BAT>z5Y&Xr-kc
zX3*&n{k733bCkPUTIy5S{3yh$FI)Co^UQ;iu|_a3VE@)X>w2OIuQwSd`F5qfy)rQ+
zQbj=^9<(ADj@aO(iS+~OMSzhR+=Qdmw$T?MhYU9<$)~2n(SUx&LKXn5=y!K-nwrMW
zZJ)F(WO|tu>&Jcrznl)sLf^XiWsNUPI^h9KDVYI9DbfZ3KK<+cSB0E6wLvQcxa_x+
z(u~_d4&h6z-GbZ1Iig@LtJ<mIifr_w4WyfLM}s9`Tk+AfJFMIRrwZ@&p)nVFbS;6X
zI;CmE!c+1Wl3oKCwBK@NUu=yMsUVkY1MB$t_xI(0l=I(5)B{_J3!f$^CX7F9@hQad
zGBBSHSSZml{lx~4mS|*Jsx^`w8FZl};>?Xq&3v?ZPyYb?1|?U`ai@%dN&7yld{y~P
zQ9XBA|NiT#B5Xr{YyylFp;?PQvy@by6}jclr6@D3wJGxjX98!}9Vw(x-%(ZF3L7LJ
z;ZRL+9Zn^Cj8*Ffg#`N3MfD;Ro3wmt%|hCSFDo&^uJ=;(O>L}9haaX{nR4_9fC8`8
zE2y<vi^ME0pOitH`KnAxzjWuT5^li_?UV}>TAvTVH2q&1qTRYAb^?5dZd)`TL0&H{
zCaEb6h8ZGSbCeK6Nt~+uD)wEvCAofdLFFRK!wNF5?y;O)^eVcdaM?J8Hn*Dw_(;`E
z`w;6$kxXub=^7cB01;v?uew`V(C(-)X%-d~SXIt%?W*>Y<8+1IxkfFVH&BSNAW+^q
zzvz2K^PgyIvb1J;H31Ss5Oy%X6RA>*<sj@?XbReaMo|VZbB>$yKk5VjH0?^QvC&TH
z`g%>3q7}!gH`5Z3>rvKcQxtpgr-MDN0~Wi$)>_BoR#&F7qXqyb0d9&w0Oy@G7M4#6
zQ{ywx9Fp|{3o96%qoX;j&>qAX!wCRs(Kxp+n@b>$lTu2NJ6pp>AOK(qz-hlgOb`wN
zVR<njLGT5%#B+1n1<txZoj80ZWBeUYH<kIcK$Az4uve;|9VfUHUVVstJc51_#I~A|
z?!O8mx2C|>zHTIAWEy*5LfUo;Eim};UdQMC*pgqmr1zt5?LXCfZeF4O)7aS|pRN(f
z%-Q5M=K5~h`{<dHw=QHs`H36tc&7>9If734L}za;$k%01)#2$)>cC}z2Z|g2BIlIj
z>$l?zV_DU{whX9q=-hBAzLDTIEef@>)3P)nNAD;Ty=#pquc{GF14`{I#B<+kwbJds
z{?nTNTZTV`#OmRXxJ&1nB`e#GxB^2S?ov>*(79&7#0fZ2*d1|=lik;wT6T^9e4tbW
z1Q<esKtQMy5*gVqBIw?|CBN&0ahIM6PExa_Xl$vYO4^EZ0jy|z>11WA5Nf6EM1+Bq
zcO~wqWQ5)eLs?>-8mYMW<+YoEYsl7zNmArIw9!^~t&9W46!OO1rn974<cdQo)}YqI
zGA3aj;Cs?~I#p~3pL0U$?I=-wC>j^PF*;RUzU^IHcX6|Lfe{^`7UB5nx6xCE|BF3z
z35~Z7>egyd-Jg;#Ko#X4k~H(8_uZF$H1bQ5@FP}oXnj*=oSv&@#$)w6ZITL>l=L{m
zm9ya=7h)PtKv;dy7RKiTEfaK`%H9t86+1}Vu-g<zwvA4WSP#37E1@G9Wug9{de~{3
zxqT^S?EU-_K`mrAWJ`s1WdUlOorZwPXw}SpIXe2kx&Ccu!UqmUE)U0HN9yA|r5r4(
zJv&L)qPqr9i4i1hHA6!;f%wZOdD&tiZ}M;yzr=OBv`Z~E(jha`Q7z^!=s5iijCEGo
z7Aoas)hfK8_-5~=Q12ny`$ebQL+G)=Dplt(LwN`Ck~MD#IR<7{$@NI6P7KuARzn=<
zRG-YCRU`@PkGzJJsnA4R4RDqf@Hdj64GywYySj^%r>ZvMKJ@IQUZfkorp3y`y$24m
zHs0|q;35!>(^%S^Sl+nqfrc#snzj|)VW%t8?E^hHnm8EVXI3(=re6i73Ay1Qi#<hp
zZLp{Q8za}_^f@D*1iOKjfP2mwmfgl`m+9<ZKOdmz_D?0%1!@T*9r0qzLPS+q?A7rt
zf!pk`=KM6YQ>%PRB*3(zlCMD2m_bl~7}P}3l<4F&Tv5U<^&Y7YJshzqtC%mNlSn8~
zznM=Zz|e;5RB{Ma4foNsC&ud$OUpEbvU6%_)3?=yf#6nbQqvzV2ap~mJ)|sukMLu_
zEaUSY-A2v&dfuRrNy3toZ1Zg8BOhib`(NZp+HKqCs6G|w>WW@DrjWo$COk~_QWxK2
zEIU@Gy0J3Q{l?oA+&novZ0NCoRHueO?NNb5BWf)cS3Z^>Gi_xff&gtR$hp`jh5z_3
z_s>I;?!SS@Q~hm60E1nP)=O>x>iR!X<cV1zoMm&iNx|)l*hE4iFxcZ^Td`DLsyqum
zsoNwoiRkYc*gMv3Elq-H3Mr!EN;^u~<%Z;9!s}V<+x&`}$4lY)kTSRX+;<*HIzII`
zUg`xAz(O)wwr<Ok5uTygH|eB7^!Pa+BA$gJz8@bLJvBPEOB8hKuT@5#uxPMWNONi@
zyYYKo<QBTyXm!I<wK1;F0@~?iC<k%MXH<JvG93~!h`C7p>OY8HAJ_Q}5XX*$#TiL#
zl`lRBWXc|81RSRw+zlJqKs$4i&8TS=q}1AG<+^y?an<dCcdCX=@L@B_^{I=4;45S%
z%nUe3^m6Qy@!vY~uB+bGJNQ8-j;L4TH9(qwxT0J5b+RQ=5UHyd^0J#zaaP{+MaL2^
zDZ~xzn4jHWKAIHuY`H}P!9RSH+?vawmHC^=+voGCE9tvm-M1I-qqP?2IvET#1*C_1
zI?*{Ve`Rzv{gh;566QUlVSzLqj(iI4(sJwOKzgg4yW0PicKkz}zJ2vr_ZzA+J013O
z@U-QAx`}X_=J55DiQ-hvv$H;+b8RO!H@<a`)0KSImZzL^Nl9a_AR8u>)ZYvx@E2+s
z+U8e&<xx-rl}qsbjA>9;4~K0PTR_(^K_>3(rt+uKwd^U=pz?VqTk6pk=gvq?6|#F3
z0ndY(aNwUlA7IF~HEEj$1%Fby^`&Al!aw?`PtKR>^EBYD$&c^FUh4vX0U3SYnZ0hM
z`H95xegQDXAC5#wzChTWxFN3BS(K|Bo&VNieT~@@aVRq~BL>`Pw<X{#K3iFW_g;co
zMWX?Xaj|%zY=l~Fsl3Fa@vr93WsjD0HCcYJLiBP(WiowJW+0GgPtoyB3PnGxG%IJu
z{q4W7)&0jE;@{vfbLDKzpwUn1?<0kgefn3_pa7N4t?|9DYQ7ir^MQBknMX(OG6Oy_
zWyF>(d4_2=#T7;VdjW7dX92S=P`RR$0~c&mRxX*TH}@QiG(097?tye0$z?>!aY_MX
znK_p^(<CdU!i_b5{+;I)Lxq^k-L+q{`6~f-tBUctQkG0&SaVQg=%>5h6`p%)AAWlB
zX)(fOx$i&1!de=<+EYxlaxq%VaZ^!q+aLO!yLOx=zf$@*vfeb2b8u-nDWWe`Hp<6j
z-e=-?>qqv=WPU^DI_u@3|H0mSfHjro`{V46V;Qk96hQ$;7?B!!Z{td2R02W>krEhM
zAc2J5LYuMBK@ky>P&7ya34~rk1V#{|R3QX}5D@8Ix>Wzi*}eDf|95xqGxzR&_SxU_
z^E{Cg-<+JB^S<wS+xJtTYh|<1jRwbfqaqu*QU)J;Bm3A5=aSbb<IUi3<=K$C{FG_e
z-cNi{p@Pf?rFkQhu$5=Kg_H>GbJG^An1oHI-sPiW7JzE~!QO{UArbcd5piqd*S4VM
zAF<k9x%W>9FEr+loL0-tzJFtX`<?mH$C{#Tf$7$?C35zMVV`W{zq((^|F~W0?&M26
zi2hjZs9uzT2E6<rBlL&1s9;0ot?6I%neRs`CQScV^J5@m&PZReUoT9Ez)TsavU;4s
z?@T85MFi#<K-tYHs;)_D-%mw1J)8jEoAxw0sEq9uG_JJREQXFZu$cav?I}u&#|J3c
zT&AP;<55PiYla-dF>UHPEpyTW163b9@2`8KV*ePrl5I|z39%yQ7E_G7I@(YGR7gys
zSNb5uh2h<JXTLx1EVcNnw^@a$fjAYe|K|NcpkCx2EMCkT(s=NRZ%55{CR*!wMIy_Q
zs1r=;%#jX^p~r2mRv9j5haA+bMf604AMjp7YqT<sV`2R?lfg1$e_k?(V?12dm-mV9
zMP&4Z?h9_I(W_5<O(@E!rBT62;=#wP$i0L4?O4{6CE?w~N@E7)U~v1SZ}7sYa^nvK
z)E1lb2X^$R;Qhv2UbG7C<d{<}^U5wEdNKFG^bzc@yTF3WF>-FC;Y|tc{S}?v?frqZ
zBVW$+C7-_3f-g_Ym$u+boB5?j_|kKJ85jJG@lyj>hNWZd3cd6V;*Wz8U~}~gn_}m0
zueVkVx*il5($vU-jD7Lw{Vc1z80GocGwBBpB(zAJuEm#8MnhGcoceApELY-sWF50g
ztAIMHyh*A<8qLpT1E#W-L4lB`Fo#`f28t7q%is{mmv$ItQAZG{sD7<<vZe+j@oMmF
zP0}3%NZhb6Q-gaga9jxvM4iUYK;+jR7YPMf0*ni1ai|J4L6&sjVCj0Mp`2TjxQFCD
zE<q9J<A=?LLpM6MBDDjT4ST5L-EQo*=Mr$sMx=Cxo(sN0KxkJqE~UkSf~hZ~e5Of1
znYjgH^*gcGJNR)2%R0>aE(wga%^kM!y`IFMa{u=1?(e1k-#{Jy>Sa|P?EBe$;*iK-
z`4p1&JZ`L}{lF~xP9bSFpA}s28EAM{z*(qHEI?(}qpw-$R<jyJ(?oQDd&*;g5wOQj
zhgi>i)RxPlGEr-=x05iPU;t>ike6;N3^iCB%#0&=)nlLK+4Q-7Ghj8$B4Ana(iuk6
z7DiD2f<1{c6N@PN2vmr3bS7Ib$YhoyEQqwoHF67}m=ndJm&~)Z6kR_f4Od!a{~Bp1
z`Wx6##G(f*Qt%tp5X%E38|nbp6cS*X!UhYmCG1e2g)PWZQc}96Zn9x`DiUDUeBQ!<
zz#~8t_Wu0Z`oy=}NL>64Yd8=XYlf~8*^jMuCQ2*Wji+m^*`B(}k6ibg&F9J?qqjMZ
zqONS#sRNn?yF2l`TWgNP+@)|V*WSrIXHvs~6Cg)TItTbqq&T)dXZ8Vh@L-P$n%O5l
zGLusy*$RT0bl+glb=5C0S`q^UaRr<EwKZm2JDlBsy`jjA=tpL$R#sNrS-AqwXV>%d
zo~3>!u9=-~Pm?mqRLKCah8LQTIs=X~O`^Wdr^5L~tD-g*hy(u0!}N{aEd?fSv5ib!
za$r~k5oeBJ1t4qD(Xs1Kp8;@&KyLgQX9!qb5?srrq=({CSg*s3e*Sy4_IIZL)2U;|
z0tG|ql%2~D15Q3eKlX2D9J%0rU6gK{s5QkbzDQu9o;B%w|3U&5<o4RClnAxdf?O0<
zeC|puDbBS3gQX%70CMHht`m9*vRE=pwD+9B<xC9Qb(y%#6148eh8y4Ptk>WmtOr?T
zkf04EdlGYC874HyV%3C5?E{cjhJU8P{*@y)ni~tyy1^=?Wd+2H)p9Hw-AUD5jR!*f
z(%Ar8608bvD5(TlK$|6wkm+;((d<W~#jsc5C)vA*k(dmzbsd1*!vz`j)3e!NklsQf
z`9XIz?8S{kI+u+`L%U)AQ{8wWtNGRM3$31{Q2h7LH#{39_<jgE^J8{<#=X=K#)Ro9
zEu@Wd5G(HyOmsIk2=bh%9$bj8NxM^1#DUbI63VuOg9cz)Woe=7v2%#}d*vUB5u}{J
zu#LhY28SNbi!e0Kt^?TW^C1D~cxt#pQye85Mi~Ho7d|ljX!ovN52Mq$ON;yLHCukQ
zTC8a@qg+O`yHyc+McQJuvR&r-vpeEan>y~jx7?cV7F|e>kz}1bJyvCOKG6G}@^Pi{
zt1rilUiYZ201#Kb=QEjgIqwfnO34w4zzUNwK;_q&hrolrnexJy<k}&Tzj~b;4e{E5
z>s60&{ELV2gfc+*xAlxqbm5(3EtQ#+J%3WCVHQdM0*f}1|HJRBn1AcA^A)d#QVR&q
zVd<IcY@=z2L5G^^Z8`0QWP%U=`r0K?DPX}nEwB{MozEXXspBkU`2C)qfOCva`0UBC
zA+BriPXlCZG2_V+Mr&G!LBXDow&?fuijOCUlV$Smr`@+UGkdvKWAHxA!bSPi&2Bcm
zW#3p^`S4yyiSxur3-dh3za<cUe*IA!#^RCi%j4Mez~JGnY}p#!2TxKQg~I~u40CEU
z+bGTXO%XO2i|R3H?dv7k{3@6E_c-dhT-1($DX24PutElW$H94ruzxNz8+l;EdsFg(
z=m!HMSkSlG-RG~;r+osF7W3128g?rMcq+Okuji?|!=?r&yAQxy%rDY)d?F81m8K?p
zYD-D3#jZB<@N|GBqR&8U?Jz7IR#RYZU5L8yFuQ4Wg?Qin_hVI{?>6Ud08C021;>*g
zm2ltsQw0xkQOo<zH@2#fTDTkcQKL&W`s<jCx6|FNb*zPu1o>k(`=<;T8Kt@M)^>lI
zz4)dv330Z(#A8W9FT1nU>ou_A{!)qQvd=qPSnn^}${J8b&rdGaUtQo}<5+!qZ?#ty
zsI9=Y2Vm=F+mcgX?pHM^u{}<%znd4pdNt-DVGvzQ4kf#71!SraR*jleAE;0bPLNV)
zbYbb-%`21e`}uE`{@+Razk%O<<h|xs215YT0n|4N(izhVX_OlDi7~j4?O$ukZ+VT9
z(F4dw)Zzd&?d-d%&Qdu?$D2f_!o>^I_G@a^v}Gt>7A^6l`|J2Hb@9dDR6RhPEp(0_
zm=^IKi=Yz1*cNtm&gub5W4EXqU4zT^HQkSf0arKHFhBt};;0&K15jAqL=|2DWa>94
zEP2c~J87^iP_@aS*0(u7-NNFqPRoLI-)d=$Sv;Tse+_ihmnBM?X-T;?i$hA@|0S97
z6L;y>+y3_n(z)gI!knrQEJr{hyv7U9@Ncl}rxs2Mh?A7|cem6)1Gq&5dY+cmT_xEE
z75EH=B?FyRq|{HVKmq<?=Xplj(%`g;<>ao(`9-wU=5<2BSc06$j#y~(V!~V&0LH)g
zIz~|?<|Oebr7Hnu@Z686<tX~PYSHJE&uq<x;!yp)g1Mn8h~*hIe}(v^&`nbK#(+k*
z9s`Jg3iD`hEv;Xg%P0*q``%Vb0GJPW6<B<{7~ynVmo<jZj9bjS)@s#OorQM>aJVQV
z)(z7mS9QyrV+7~r^P2h5zu{mFA4Zyl2KI~tfZU@hw@nIuN|Se}RaLj`SJ&j3N?Uc6
z6yx?CTjQpO7wjlmc4bKX4mY@P0mx}<jz5E1H&4YL)-o)z*RZV;FC^=GZ~;*d4xC2~
z3qCk$IT9C@t`Pp;sosCH*8j6(+hOF=xr3}3NxTQdg;Uhy^O_Js+0>&Qcyz#ZR${;s
z!bYbrgTMxHDk<c#KnARXQf*r-q=p7*CsrYZ={yJ&0k@EHhm~vDOuyG!cvF3~qgO46
z*U8X4u7^z{yFdj2mGO}vm_IQ70aU0mt5J-MY~r6)sI>n>g$fS)wMS#&&ni@R?);`g
zW#O&?ZIdFn7^a6{VHyBeJB~WZtpZ?cnsAs#m(TN?ItBB}0vNNRnDX-Jht-Se%CG<p
zifQ3oWXxl*L6L`5uZQCHiSK70Z2@sBKT?V;xer2wV%mD;sosM=$NW7IAnlwy+{M5j
z-tt7N^An2Rz8t8{F8b*fC3HFbaorY;>0n<UeT7?rOTeO-KIfz|$%%oMUJ7t}XVIb+
znx16lAb#}yrEfJc+XXOTc!1#L)_Ut%0!-MSESD14lR@O*$p=AgGGQH^)~-;5jo-G(
zFaxI`cDwr5zgh6VJMq*Tce6qKstmGp+8{l%MAt&wJ+WAL8HDs^2GA--6+;puX3(<O
zU|Rm6S>Z>Tl@;$|yb}mvy;C<*#8qPyzwPw6+N&*+Rxo3?iN8}cQ&%mqcLI97Z$)ux
zb|*#z!l+4&$H%9F+jT{h?Wvv&KuHT~Z6TN96XBheq?o81nz1zQ?usC78i5njhw71w
zjCo6cPMeGZsSX`SK$BfuoTGD&?fo>n)tOFugNdpeQVK76EmAu-3;OmM+fG)qQI<Q=
zyqMJE`6O+@T7xn28g$;R3m?pr2P?rqI>@4Wu>8PB|G~#$jkEJ_9!{I`9ChNGlgx%2
zPA>7+R)n9eTbi2iKH(+9Si)TmhIG~M4jqn!3eD(|VIj(P^@4eG@cH(&!;Zy72kG}&
zKH{hzP3Gg`B?tWgn6{K$)zz{@taNc2Iw8!`+&yzzH(04fgJYpc6l!T$^pX5^5u+UX
zyGw$P?_Vuq{M83sGL_(*H;AR^qDS&343*l15a+ST;;UU^;orO0nLieHP2_5Kboy1>
zO$2@yBfGL`SD**5BW~@!b&ZqSvI3YJnBKbDnF{06$$X5i456DXl39+*gqCY5DuFoc
z!^R(rbl=AyCAU^xGAxAvww8uyOV7+#Ei+-}4kxghsc+cJ<=vmnSZ~YDSDp`i+-Bg%
z+Ah02=HK(p^<zNTX<WQ6^GWm0`(Gs@^Y&ar;a)A^6NmB3TV5eI>U-T7S{?P7njVU?
z2cEUv#1zpR<R@)Iv9{>}!tU0$fL5EzwwJMXULH?+uF@>|^N;%y?iX7h5*kKf5X2RG
zDHu!wBfbsONDpMcYjPtbpyLHhWg_Bx-^aKW_U5HHypn2?m1-6DWu4ys?yt)DyWKb^
zS^JpSlLT6WdEW^Hv1o{g!W%uL{i%8z#4fNM>hw=dvYP7g2=9Q0J6>VW0opjPn;LqE
z>WD_I8~MaVW|S@C#bFf};Uhd_C3W23LG>bP2s^4_?;~at2zBEIh&(Y3>*txLgH95T
zojIl{=;2ncUt^IQ`zI2H&XhtHYNU>z&^V*_)*TGJd74r<{NzC0uF6s^Pu-8Np9xZB
z358x`i>n1EvD=y@flSG`^y6yvwb*BO_;IJ#_YFeVRT6Z@@jdrXe-JZp{8;pgo-lso
zTg-ZkrHGDe5`#<aA0mjW0^gKSD$mpFtZt4q-Tn}h#;@8u!s7Sc`lYW9(xx+hEw7@`
zuWXrOsg4nfe_(1fNYnK(*9?X<nW;JpS##hL;Ma^qmFwx+{Cc;h?MpO-OPf=Sn&ArK
zxbM<=(vuO`^QxoXt2Iw+hVK;w+v%Z1h6A+jad;YQDqPgl)}vH<*~G$NP$$Df6`>#%
zD>{doSq>9<=pEX&sav_e-dk8!yQ>~)JG80L%zS3wh};-tq~h=6FjE7DWu}gxjYUT?
zM_t3N3KF^w6@*3XY&cLETKV~~YzGI56ctU*DHfS0%k+gXRtaH%Iwmm5BP-x@fKEBC
z+@}2_Wg#QV`cdVOFs*+u?7&?ECS0eR`y}<@26FUn;jM0JletX^m3T>=ZF#NvSk-`c
zazLkzo`H1CH^gTQo<^-tPAs!h;)TH3s36r;xNeqsj*Yt9q@a2mz>PfP0_cjZ%-5rH
zwj3{&U7tj4hXxo9^=;^`CRT>)2YliKh(%$$mJU$s{g<&+x6=S=a>{;jqsDIT2BwaH
zaSsXt<T0<KR07$TEV4xa0QX_hK$Xu}!_fpqz~@uXGcOSsu!&z-_4oU*vg|&IAhjA9
z)-5|S{llODgu5RYu>rOU`ceQ~yG2z{{qzngVF?bYlLI_|3&;f5MoqJqCXu5)I}OXS
z*UTTGibb*}KcdQ-VBno`{=%*HBvI!Sfi6G;DB}u5=$7-D%7VJkv9z9Meh)}OkeTm`
zq_^%jS`d?U<=R*CwOvcR%v%67_eZv|J=SB;4UrP@yeFClfvrx9%p@qXHNk=i*b?!V
zt0&?BEep;nbt5_^ybo%4kZ)LaJ~)gs*d=8M`H}qB58q!shyRVY|Lj4M<-nu-MrKUj
z{D}%zvBT49Hb3;>63x9hgX;l1-&WXAc?+xa3&CgOo`HpgMT&mOQ2v$zNd>6kc%t%v
zobdXC{z?15k9vGN={#}2;D%T2vqx%I<VyT~m|+C1wwkDcik{6Vk)T^cG5Kha3Cu^`
z9~}ypB~L*6DOcHLbXP}eYe_n`$Z+-#^|WpdW?@(%&0Hi6R+MdpqE%e7Vj>kPT*Ay+
zY)bu>UUmYGmhu~~=qU}Fb);gY=c+IN8~OS-50y;K8>-4#KW|+wQjSVq3PyAfVF*gL
zH42Dl0O((48dZ3kbWHf6O3G0$Owej6(TB;s<SMMx%YoTqX}8^#tY`hKn4?I{k`zXQ
zRaS%i`fScHPtP2gNE;hylV`*ek+dKLi=n~MQd^-B)O^be;+XMxWETWaUL?rJ$D|id
z&D1>>DW*6Q#p<FUMMYj^T%qAAYnaf0g&Q1OJB?ket54<z!LbDdbSB!XeDTdBWNI<-
z$C?X|<i=_&=Zx%b(^s6ENN4eQC+#8hoVRVaA|!h+bK2RpW0RYI;-bT#m5hT@@66K8
zSLvW?2hbIr#U_aBFIl!D&q4k4F4%qvwKGe?l+j5IiWJo9EWYw5P60saGWgKzU#6yP
zM`Au2cjH8Vi-jW+Cd+vFwPiJqyxxsHJC_9YmG`u291qpnR%1Qb-b!Gu4f`t%lL%xv
z;v?`Gy)xnEg8tP1j_uUH*oyUi`gaC5s}arW=1;qKi#bf}ty-Qeb<W@DvHOfW67o|Y
zV?A>mB=4R2V3z7O0P7ut-4%j`5!UoxKV0BIXD3J3G%3ecNlEJa?~k?fjhc_qu-nE+
zk>gq&l~lJNjfd)r_~1zJ!1EWE&s$s?oL>V39`xnhlH@0xzc#58&q&{Y^Ol;22W8x(
z`?jZBeM6t?ulAB@K+bjS)AfoAc3N^k0v#ze+GxOSJZxFD&s7UxtBNk`62=2cHbF#w
zJzJITCbQZLw~zT=Idr=DC;<EXs<JxwrBej)-1Aq@F%|(kHF@Rv*3K|vg|`)o^EufF
z$`$oN-}@@4DEHf26x;ja(ya#0t-I+<qV1cuBdjFJPHoEZCm#kTGvDwO%11K%bbH$r
z4hsvgxl+8tu|u?}A-NR*n0K6<KhvS=Ko49k1Lmg6XZF=<9sF7YLvK`}?im`1j>{^m
z=M8l~<KAXxRnY3r7Yypd)>*nY$NbvGrZ0>ZZ>NC}v!y9=(5IAIu19y;*Y3h*j7qqE
z;GWyC68xC3)_l5>Qjo?@nDgw<|MpB9#5xA8nFhajh95;#ZQb-=%0Ovl+mc7sg9AG@
zLVLNzeGkC=1ROd{?MXHw_&HBO=Ev$=MIs4y^OFZFaIi-?cYZ0Tva;fex{>Ph!s0d`
zn`!*|6}K&ev8v9(mShZpsmk?oE}tK%<dR*5@Ef+dYIgE6eW+?^?T^iGpG`z$1jS4u
zz*=-ipH6MUxg1?=udq505&~X{qNGG|mju}|U|}K=@I3-p6}uk+Ez-}p`f-4|-*D>l
z0>i1YN=MLO)I6r)EhyvM#iee<jetr=jBu4!dogwLirzN_g?rWGU@eIIiJ*%njDT5c
zp~v-H3q1!xce5Cs3yG)d0BU2PLk+@ao>V~P1;R82z}DS*XfhQk5B$@832d9qt1j~K
z2_zXAT{E+Q`1oZY$t>DIDX*#KQDWBW#VE%&4HvY%9!I<iQZFbL^YK@)8n?|UcT*h)
zJvOO_kw|Ixu7<Zie7dPKr+!a7o-s9gw=RMBGHF)Aqhe_JWY7D^JzOzr_hDVR>J*C9
zU)bg$^)lGxDnf&0kZ8F$LcS!GGsxg=uaUC)0`<HSN}NqB9fCQo+vy=w$l>mO*p_Ur
zv(W$?utJ;92oQ1V+&m6Kw`DNR-P^@d1}=xUZ^tzIK4aP{KlaMksB6dcvjd+HjS|MN
zCC{?u$tmRz1BY;U!S9GNZoXjtTce3^j8hGEgZKymK%lX93~Vi(2gowQAc1hRPEDw_
z84yS6FtgBENSA{F|7BZ4G@<IkP}MdU*K!w^aNxDFADJ7l<aqyAz7X<llrjS+7RFS8
zI>)#$pzE0Ril+yN$2_C+(+NjQXWDD~+gbb`<M(&3qdiOW0jGF1cZcVG-ErLR?udj+
zgx0d5^0(7~WGZfWSA35auxMsc-om2l+8<!JW!m-W8n>tlO6MSJngVHn5Y>`tDLH%-
z(-H1xIG@fM2$>>;K?l_x^eoZrcmG!7|L6OEG8@_W2-WTv_ufdgJ`A+O&-t(;+bo_~
zvA<)0yrOW_dr)6_SY5bF5!n)>(mv~1%wCy&-z@&!C%&o_(Xrc4hxC2)!?2xYl^V5#
zDkV;BC8CDH7U~Sp9h%BdCcBtg=|UWqC+#;?RW(CM;WzCb2W6|9clVWFN>4UEuAsHP
zHn&=p=)&C6k1TP<hDVaG<A?=#8a(Kn5t<%a3l8qGOU+W34C~Yut@AcyM0aZAn4JE=
z(=9@wZE3<7@8zFT@7{VLb@khc7b&N2#PaKF4TZa!4%$0mz4sfFMR6kiJsS0m&s<o_
z?*bR4ziB-zAq-J9gFD<RGYw>@3Og|R8E`!+-b?ys_iK*&T96_)EkxN^L0VmA;uD|I
z0i>-2L6xp~1PSW4(eGo)rdXcd(|daP=l^-x{`s57AN#|#kNK9AVtu?ah*{8r@_nRD
z|8OjZs4kW!=k8>fAv0%My8FGQseH^g%&6?{bPI+3yTSwKlFe>58<l$pRJw0C)uIxU
z3qIr~*6O7CNJGnL3lgk!KtoWQ#(tk1;1yzC1lF8@BqeuZnN<L=(Z&d*RDiE>N$_l2
zN*<aFdONJ|zt|OctK=NKsYab^<~SXn%bDg1fvtZsgG#n4Fc1zT>J$(qxSmA%{kQus
z#^klhxNjk=4+*t(Pv|__qiZ7p9o6z}@0~B&b}Uy^SjEQNTK-n5bs$&M2BEkefL<0Z
zcWG*rLA@+oA2`l2E=BS;TcX<47ma4`C?>zaUb5eU({w8wwgd|qG@A`4eVcLn%!<c8
zwXXKQ1x?~EO%HjOEkMUa=7n!_WE5o=!mur1O#9}!2!__YW;NhMKnVS+^|!qIwxX#y
zNpdmtCc4uI+oRZNtbsb|{M-_T-UL?C+o(!feayf{7YeN~{gPNv=*07F`P<gt7UAw?
z%=$8Kc2M%3J!c~gs!jyk)gG>P&<{<(G{#pCuO5tRjb={u+fk9tZrx$uZYOJkB6^e7
zq#lDI-QRm8LJW4|<gglVEYe{ZMz48c`ILgmj(z9DI(kCh(I+Pt|5D25WBWR9yt31z
z;JG%MH#BWmjj~X!@Olq}l`QqItu=zZ8IRRA4((>YU-m6v9WDF3NNX@y`$GO$+BEB~
zn%uLl0zAmt`N5X>ffkO|LUX)id5fw%c*+B3iz`Weio2pm)k9_AAMWSGAn9}N5*jmw
zAiG$+>^S~jLyQi!cqcrATGlQiS0!8(MpK=7a{Dirb!+Vx53w8l9Tl7m@9=vaL2yIL
z^zH)nhSJiqty-XcWM>Gq=(}bWF2&mPOerdVsv$}+pfg4E3Ax-FlAWV;`V}=+?ah=I
zu~fSrCkZ}llOIg2j3A~8InWDp^?l&*9Xn*$Y74{q6W@x+u2ej}(Nu1!l9xKS<K}Lu
zD}su9MFO}p_t*PgMtXnZV+lw$MK%`YbQQ6mM#Qh8-Ch@xv*k$yk~L%X(kuL-LBR>g
zonsFv*NDoZPlzCIfh*#S2VfzA7l3kKijHfH;=|v1c^Mfg7nd-glYjY=6(yt39?98B
zjE2w5MK%Jw1_EAeS``|I(SX*43k%Jt&SGJ`(lb$~iHB|lN<Ae3#0C?iJx$FAfP=-Z
zu!gU-?ZHJK3zZ>8O(nB&53{3-)?Var6Yu-yL?Hnl97SqhuxnxyLen%ctzZh<7gQp&
z>o{T=vj^YA9~_Rqzn;!Eb~5<Hhc{t)?7~c=)76DTj-TiF{{Bb*?;a`tb-T^Ht>^jZ
zgU4ZdO9f(|OZ2PK+OUY|YVFZ`)jri(4^MRd3?v>xtz^ubt$eqvGz=e%aXl>a)8`|x
zbb)L%{p|kO?v34}uYP{;4oLhjwyPdu;+^VpzAr~0{nj%RU|7DYdt4r{k_Z+mn#cQZ
znrBOkK4tOy-6lbH+~OaM?`2MQCLI0HdNi+Dym&@8A6m3t@h(^*5b9t%qDp-U6R-x`
zolGzEdIZoPGw1CY6sSKwWe81{C%Y8HCx*9p$G41)%rgWvHY&T-*WxD!Gg4i@MIBH{
z>ulA<mmT5WFSFgN-v1XD{5wC&U4kIp@h<OSrrx&${l~fpf>Z|D`*cf^O84!D$Xi`H
z*YW&@{L;6@7*-B@7ZUSg1iybFDH{03Ipr*GhpGqDiv@da2U$X(PyJ4L%vHjyO*-pJ
zpft~H95Nh5_3W1?A@y-p9J6kDrjHMwnhfi-Y?Yv@7)~qyzG_$GBdE@H1ZC+HpKt=B
z+e0UQdDYWoDnIL|-+A%CVgFKV)qfPKo%O@EMaXE~bIMuF99jlUXBT{Q&tt?+j@=a&
zstPO5i7^SOH|cnUFi`iHz^LZea+d}QPdlF)Yv(m9x}W!`X~s)4?e9n~n)fYU?oYnb
zg{z=$H|*waT)Md?%+~Vop2pM2!zGOSs*Y$RY#1AI_I2&LD^V}719#183?O&Qh*?*b
zyIH+Xli}Nw{c|>6l)f7xl<YcYTDMn0XB;NaXKJ+W8}T03e4W=ztH0e%9r8-H4(0TY
zNkzQ!L<LmCo5<8{3jQJT+iBa9TC=BAqQ2s&x+ZzyoV=$@9V%w3($%cy52Wp!44qjd
zePUiK^tD(*+R^+mgnPYJv!y%?^L^JnaQO;0dtsBgNT@!heeU91LPL*#s)ZBSQtB4&
z-Jh>Zy>l|g&!MDYYW~tU`=TAt+43dJSS9q;1-e$<t0%WV@evee<pLx#QC~Y)uU%*j
zs!+0<!_=h{JSk(A*AP)mH^mFBghn}dTc(iipU=q0=Twshdx@;@)sqexOrbhXI{*tP
z!DqvBhzKm~R#!=9ov<)Qll+h?6yTZ?1jeQ#XqZ7CvD9$!=W6NmGCk1icr~_vVWN}{
zf<f)`Rd}-og4hz=YfD7x$Vl2ww4H21I1-6|NvC2077o5)3IReFg#>-Xi^Ma9Z8hS>
zqqKBCfkE(y!&LMkhd4sx3e<@gXP;(T*+8*W&UtNnXi(RMqvp_FZ*8a{+Q+*8C*5GG
zZKi9C(c8l2lS$hTTw^pI)sAb6J|(IV#HtQ|<n>p#=Ju7rH0K5%am<DBa~zSvouLY4
zsNuW~Ir<K{;Na|rs)<PtWT@b7@A^(fNw{n#t438Dy&O~z0KI_GG#oT|znGG>HI<1!
zpPx^DN~YxH0VuDBkIm}P=&;jQKxXmZ{?z19sFwdC=JX4Y!|CQjDge&W8F<C6sbV+g
z8E3G+Ny?QtP#3v1@&v#>XA%kv8n9j?P^pD@KsgT&w`M~C_?xQGZ{pGdZ2<Vq<g=Xn
z!u{IqQ`&F-t!(~*{Xe&@H?<JiBf@ia(93EqSFTB5AX;%#1_LJyY7HN{S*`L98Ejdq
z9W~C;DV$v$9*T30C8sF5{}y(N5%&*y8pZIyt{1*@%uB@)Md-_m%2$b1-8Pveem&9H
zAu=`Z5%dgp?vWin$C~Ao4HbIXcK3boMce1byvMR6FKV{5v;s|_E6f=pR%K>dqrtwf
z89CxNIMmMR=OTXi2eR<bUp;=pAL)-ly1Z8i>F}SGfjoV6E~3wC5#e#;p!H$o$kxO1
zWbp#B>t5C+&Xs6dlUae^=Cy~<U}iDm?(D7v3FJPk)4vtplX(0FeM86%Ykm<EY1Ek#
z)wS_l@kF>?qcv2b0^2v+kK5>$N6{w<NZf!dnKs?C_APC9Y4IZ&nWS{eXd}pff$Zb>
zvV|-YSx94Y+Vs~1tXC(6%{=+1CW%je|2_Uowf!GCRh7X*dd%cjI^p){Wr$IAU0rnq
z)lN@d&(^ZLkVY%a75Wx1BbvC8^4T`+wl)4HV3~IF#YvraqObf|@2`Zv_j!5BCWvr*
z!`7B0COHOw#>G^a3JqF|>`S$2%fA-cdo6$4;6iK8d?DO($?gs=g?w~=!mP3P7P8Om
zNvZw!>_f&s|0VO3&%M6zzB9UaDErD-OTpTH&m?VnEcqv{TVnouc}4FtL!L4=b4xix
z(kTxc+=dKZT)ta;t;!Dd?is1ZW>8aPQBs|z&X1;`FnfshLKxYx&%?YPmayy!6YWV8
z6CTPWJ`OVn$CGQT`k5A>N{Gl(z@)19J5Bz!58f?1-XUn00&?yAhF&ThSD2F@@g6;C
zfzvf(Md!NDL@RuA|3P_l==hq{Bo1Q#A*NB7y?*<~%){v6@wK(JOOwu)t0ET5;!$pf
zGQ}USI3>*P{IGt_c*$>jdpm8lv|#)ArL7cWHOVxay$dBRf&1U#zR%s#+1uNDy?x6%
z;$GS=>pcF{HCoi^f%kP$4-CJx*v*QHigMd4k~jL(>;3Wj9kI_ijYqv!5XL`B#$Vd+
zGB!3=i?R_uf5Giwg*A1h7d?7c^CLa?*(sB>gX#0Rxp?D$zu*hAnm$cu_uv-2M@ZK`
z@de!P?(WD<zi0Z&*7&*K_Jego&De=31<LGKJ>LVIV%GHZl%r;CinMJs_N)MHXd`PC
zVxa5w<NBKBG&d_Nz{H9??AsoUquo;?Jk%))9TRTdMaU33z0+d^OpV465UHbnDQ-Hh
zUlsaN&fQ?IX&&E=nf7aHWu`fV4Vppx02HMXTfHDMvW+C0S02W(HkHFqY*dsCL~SrN
zoLP2@fj#G&J67i{KmmKFaMD>Fo53b!vHgp#-Q1XdrJDm7Xg#Du?>=pNO=C_g_&KE2
zK&7}NtYHlvUI&KA>r6RcyDUR4*9*rSDkvq_*iW!;TD<@fa;$5;d!L2<+xY#fijQGS
zIkOcfCTB#(3PGBcT8j8wC5@q2nS^8M0=&QqRr6-$IWX<4tm>>6#iH%ZxuRV+d?};}
zyhIpqk~xL3WU~&9|In}Osn;M<h?HTpGP@QFXtXmc^YJxN9o=$qOOL5kIYJ>-wuI&+
zM{quok!<d0$Q2F_n%&g&pXSi@J*m=JA=J2O@$syHlnAnY_{4E6yxMJ{QCX(X3O-3~
zM+o<Ff=RhD^qlQ$0pxUhsjaWwjdWfY-JX>RtYZYJm9?6^7QE<e(aUa@KN?Y$Y_D$L
zr>3bmpB=s}9FX9e8&HAHX%ALps~g(e6iz}^yO18)L7i}UDz%}kZ1Mddsk2}bBEU7`
z1z8Nht#^Z9h82sZZYCyR)$`>yB;Wo1=Irkt`A^n)I`RftROQ(f=A`Df*L>qd&>gK?
z%3|nk@cApx03;5EBf`v9yanJ@1Ji|I&`EGlthk=Qcu%vq;RMT+6A-9mf1`iBoqfO1
zb0|fGMj55#>i-tudVZs)%_VH85y|rd_Uu8`(a?$Up>|edVpL9j5>eZ4+0t-s(UC+8
zk;~n)h5;xH8;E`Co8f_K435+J$FAP@&1xv4@D4G(a*CMhkLkhBs%<)Ax-&f|J6NJm
ztEcQ0gZEx|B`D%N7i;8gGXbBkK#4~mYU1>ag=+)KKPErn2<9(K)RX5QD}nv|ixdWf
z<+6Bc`+=KbwCi1zOOTGZ1&V^mh9ujTaFCo<Q&kK#>4;yuiYG>YOJxMPENac?Oh(1!
zM8BA@BZe$)ZgLrByFP;O*FXH-a{Ij>Fqbd3W6mhzpXyE4IlKTbh|W8rE8O^hkub?}
z&Q_O2d56SRNK7S!>}M`+J)e%wbtRtES^of^5aR)ZwEP6@FHy7%;j*BS3m*W<Vmqv9
zVDi1ZX)NS5q;8u_vE%i<Yt6V5m9np^7Zpq9_Q}%bJ<j>{2&fyp&uW~Ru32*N9BLhJ
zG=+CZ5O3+ITl-V}0%w_1iIuQLZhc0ScSkr}g>q1xhyMEEcfa?J{>A2HzjRr}EYDYm
zG^zUyJZJcQUYUCjp?PhK(2sp&cc<GHS<$U+he!#5qRmE-w%8vR^^*OntgTwU24Cz-
zNufKuc6^ca>MO}7|91P5+f~Wo9pMhc*aYku`Efn{)>O2XZC9q{HR6KK2*##8sU*<W
zKd=h`r$~CVvq5-V<5MBg7H~j8?9q2Daj7QPXLDx)vn`HTRdko!OGg7${ZfWwf<nqH
z+GC$sy|W28j>Rw*7egm(DN-$5;Sf5Pni9Rx!78H~pV0KD8uMR4bTFeUk<S}lI<>wT
zge#nMuiz?g-bhNqp@cURTZ9sm>#EJ12IQ$JWRvaCPkau7DlPb6dJm4rte#6K=GB0Q
z<Wn}R9mw*ef<1sTGlrp(J*g<FE<|0U!=g&nfNZ?B*)YE(2?Rla4Swelw-|<VZ$=0I
zP~F`MLpIJE<;-?Qd`k`&zAEOW+#JfWFE&vHP#5DC!rkAw@4_5TyT+EkW{%5X1%4!I
zi58a#2T;y<b!n;_VodeVoh^u;G;AAiQe*7aJ2q%i0f|tBNy9X`kvm>>9zoOfNKO#z
zK%QYA_X!Yvpd2L4XZ9BV@cBYMZ|Agq`GzI1Q#ytV_LiUxRhUze9yli13E4$l;G`*3
z9#<Q!v>{ZFsf)X$xCg!ONma@iTqCy#&E0Q$(ASwLmPTTo&V2IQss6z~|Kx}YUxBF{
zJT?e3mBW5{<hnOZIQ=nI|LV2jpmHumjpS{E*7ddb*DCDFkQdEsg5;QJ44&7t9#0Cr
zRz033|0LDo8m8W6tZuZ>z8BkiVz?`r$+}tN((1m)3%@rQ1}~^G@o=IS)xW4z;vg)|
z2D1IJ^E-wGaYVdhlr5kpUS8_08u0A!OZuYLRs$YMovQ&CL85qdVcq^DsS%NURe|pG
zg1NB?4!3O-FCUkH87YP!SKmJS2lJTk<l`Zc{;Er1#Cz=QdQ_y)Y{u|1E>&he9GI}c
zD!!kTzxKeqQ*7iDUs6uSF)i;J^d@?+l%-|Xe$S(Fp-;7=&S6GwYA8p`FT~4NJtjOa
zKD<VXrx*3hFaFOBjhXfp=_T{<LL%<!FBEXZ)KoAnV|ZnT;K?x8jBm_Nh3I7B&nWes
zbHaKo)dzfag<XKJ$Kx`Y;f0#^zy!v{KR|hPvaR2c9@MQHPp9Wob%z;X^+N1z&&OA|
z!AY~#_fW}o`pB!j3lwx)@G(JB3<hJC7(q`)^v@a{z4!ta*5H7&aV&vlURM65y{f64
z_V(Q-Hshp8XmVrC8KIi~x%ZL_uPF+?*hLRz+diy)Jh_F2*IWhs)38)3@s?hcl9q~!
zA_Nw~Z{Q_@;{@q;u{03U$A9Pz+&^bQ!i>$n?O*XW+2mMTT9h;WT{G+!=hG1LoYj>2
zp?O*balqQn^Y<6fzm)IMHv;}+K`|tvg73kbZ{%ll`kp&&ux|w@)f~$lQ{alBd6aEs
z=>fB}7@aXSkFu%yKT*=O6jfTyjJo+nTkdXW%X<8WjM51U9nHDGMZ{Qi1wR)Gmk5F!
zeqf231e?5F!RzVPfy3)jMmpvxpd`P}Sbo3<jHND)G@A+k-VdA<+9G1_oy6!IB1p{A
ziZ#Zd$$*J<PAFXA<zg<gjpbuB>XXZ>8Om>%%uTldS!yEz>#+UQGsb&?1o=Nd8*&3}
z?>Yd^*A<sR^J6Px*TOo8cRFAiwXU&c9fj>So#(52EK<%#Z~MC=+qV77SUZ%i&)WZQ
z$|Kwu*0ygH*EThIeRQ+P!K$mm8h~0GJeV%MGYQ4iZrTP#wt2I*$P1tNK2j}<?5O`<
z)ab}pBb_kY*t<aoJ(f`MZQC;;gwAx&eA6>m&&u0%5TD^k^iZMEyxV%i<8a#~<AYVY
zo6}YuU)BG{Y7XB&8|fB_SZep^sP?KxbZF5cH(x|m57?GGzELk>*vbj5M93lE3>Sy!
zN4ha8tNM3#bk_>KWar$?QlS=FZus8u4Dp-e!oTEZ=X*HM;gCr4+(Oh~==M-ll<;mX
zBh$7v!}C>3`^(oAurp<=ekEMeM&c&zvj8X*Aq<77s!rlMixT19pD_uE{qq&L&LKGl
zdXe@BP{JD~6%f1tRx*lCAfc#u0#fGUWl0qSfq?#QA3VjE=vOhk*2Tewd_$}6Z@(;0
zpC$yJ;--<Z)dr)}dkTu4qdE;<t)#nlWsYk_0*#+d(qvK{UdT34bv@`!2Bm*w9Gm;h
zmz|*dmKdERn%VDLy0+?75&O*VT&;bATpODfSLJ#nfJe@0c>2`D(noK70RMDjVYrcu
z%Cu}TI*nYkAJm)r1wDlvVj$qyL6STpjkGbH#ch;qW7Z0FtX5=jVbn0Y_BjsZ%&J1u
zAgSf&f3INwqc?^uEZ3($vVV?_=*d3vYRh3OXH9(RE<yfSREtjq_a^|1S8>w)xy3AI
zFa%^bhCkGvJT+^SlX6w{dg`G7^c+f#!d<%Lcf;jFh{s&`d>rx4DIW|Ebbm~F0yglv
ztBp>eoHy$&#cY&U8A3rQf?uhh(r~@M6tfBnPK49BS)O`_iy2i4XU&si1Jt#p!evQP
zN<%e*65-otJkpt>oA1L0f+qko#st^wAgI|)SyRcqk?aq=UcEYgFD1sZQ_vs5!sf+u
zNd(!6YuWdVM9uSFG1Si1suVwY)i1VUs&;hl0;Lw2Sgo$SMZoDeR}3nOzg+)y){jwV
z<#R(<z*WJ=cN{a^=!L|b!wU`5<C$HFyLS*oO>-mju|5n~g9BkOlb<<n1TLJbnF^>n
zVX9w?Y0f%vhYUA$9KqIPLxj<|WDPTujd0KAm-D-NDhr%07jy#FjW->&2x#4YjeJQf
z-S+=)6w|*{<oz(lphmv#?N(Za-qCl7$0M}{11`ENcHLifDJtgZuM<A1H1Ciyq(-p`
ziqR#Xf%)&gG5_~_{pvc_909RD5X2pqCy6_!gNaoQgjjLEnK21PTuauxbi(;Pm8NQX
z`AI<ZH2u)6U7m=gZyPc=DpzoXd=LTMcf>RTre7RQS%lBNd-x~56S*G>%&i)({{YnW
zKbSYb4<iZ++s=t}KKSwe*&AAJ!;*+XYBRrK3c}pI+H=1x!Nf0ov^`rvZlEJQ#+V}&
ztP^MmOf<(Y){(n@3C}1aE#(7)49JiQ6%ZrXO%6EEktbrhE7~bbY-BW3+ZAvRwn}#4
z+ys8@x9uzF($!5EwS8E+WDA$zIdQ$){&~MQ2dt%oJKf*gGr{}*wLo7x>tOw~&BR=t
zfy*bptw(7W0JnsmujxPk<sARe&ieS{;{NhCMnHe2xEjQ}seTAJ-4SQs)sLU}ynPyv
zdyistzduL({2NEv<I4T`_m@ZPYfCElTlvj?1riVQ9HjL)4w}2on;Dw|seiceIq|>R
zPOWHJ7p{`8WrL3OdR!JSEGNI<e;wf75i9DNp?0*iRdkYdQm0v5z&T0vN{R|#Gwzyl
zYDj^~vmOA$ED^%sK)6Y#rX@^W5Ms>{1paiGd`8EybQYt`z?Q12f}c0>_b%ZwGM5?!
zB|a`&ggF(jtcGl+)*UaNz8PpY1_jKT>vlB}-p}zaPnjr*gK5CQxTClZRZ}r!%-7EW
zfC^-e8*$318jvnu(!>WZ0%SmM-owD?9dd-nn2hAfq0!L(+WJc0-E5+;S6$3ik#%;N
zTpWogN~AHSw#QvM4bd}IC8(VohZe0BZ@^=?vInenCG>-`H&&r0pfX{Oi%TD=g0)^G
zfM!FMEW8H-EavT&G&u$&baTDSZ~%4E1~iPv7n?KuaiQrm>$~@^+Or*K^g;gwx0(D=
zjaYqK>b!>~pr`OEJLP7V<;32o9@V=cW6-l$#H&lqBe3^fuBE&vT6&0bs-Z9hj9K<M
zJ3_0gN|8io%#!5UTl72G9+HV`W+B(5?cU7xuiGj1I2n<zVuSe|YTD|GxpX8?2e9z)
z)s8!BJxn`q(c15Asp(&zKV?T{Q0mRPKo0LGAuv-=plwCSBfwQWZ_wiF(OQ2$zIJG2
z!g6_JU)urv+QK^FL}87cfeWCNkai+6PYG@b7SVA|Q5idNRi`CJwB?j3ziLaWUdZ*6
z$wS2w-jasCw;7ZDOCcZQt5lX>Mh>z%Iex2BQb^sk8VN@~!S*cKcJA@;@oojCslVx1
z9=H$Gv%h<5qvsRfn6csZk>$inc2EOor+x;9nv0wH#J4>CVzd|Awu3n}qaGT1i4g&&
zhROZ)abf?z!@SqUo*M`bH%M^|9_`)PY8gV+jh=~gTsPP-35?!ad$&BR?H;Lx;~$Yc
zw&Wy$&bc0SdH4OI2G1w4p#&$1Bk39MZhr__Czdx54G;UAI<%!6S7mY0_`Cg;U;?(L
zWfWe2%Ft5c{QllY{T;}l6C~%3cI@r&jF)zlhq(`;J~AQN_V1TzBe6LJDfbT!A|&=@
z7D~w7mx;|U4BOu;L0?<EGLAryiNz}+oaHOs)~2SP_@a$Y9;5}s*&a_i$KK^ku80&)
zP$tedfY^}!kIcEPr_)sFyia_MxksX}F%@Pj-XK4Gu<-wwXuS$uJvDj;MV^m9Y=pmA
z;|KtGINE)u@fk0Deeg?WeaWmZHS5b`_@&MI(!hV|4Zrk;UwXqYz2QID8<xx}INML>
zsvT3)a$<_Hk>K9;bOkJ-QT%nP>)xB@L%RB+Wr@JVy%`v-&PTv|mGdU+(2RhT;d3Nl
zkp?#Kaz72C!Is!2Q0(L4_S8S|%|*4gjcA&gV3a!54O0c4%%!P+l#)J^^@$J4Tc5u(
zDyU?^fO5Uh>GmK0K1c$j!<kbeCtfeETrEqG%)x-8@*UKQ4Mf{x#a&xX1Xys%)8mH<
z1(EBeuGlSONLPaW4Ur|sJl2V{0;o^1fQL77!iS$$_0qy7F?1Jx?iRxk6-Nd^GzQ5?
z8RjEsS3&kqz7=tH$$>qt=(%cgSh(VDAnFnHd@yz6cxIV6b4_TCwggpwi{GdEJ8b9y
zi@T;#BU)9rR=<uBE1&9Q^F&(32eRf-E6u0eC?(xux4DpXwH9ktXAx1bcj@$?p!<65
zuBJigIKPfu>v`&UmBzgG`qB|^gMC&3Fc-c$HJYktbnm{u3@*d3`hb`oa*I=J=07cl
zO9A2-2%mRLP=T%mOhXDTd8e?kt}0wdE9_<0$<+2spt{I9uZ~7+mN%Ld_K7dAjA+IJ
zl;w(lsjAOeWFzgOFF~po$-xHq+*8hmWwbwww}ej(kNZ+GgiiC;9(&VUBbQZ&_0%u9
zN}t`SomniOVu3yKYV60-OqI<I5aS_&A7>dLY|bTY{T09ImpqEaONO=U%#e&`8>z7A
z&ANAJ4P`d8r;h{*%jwi!rICo+1FPj!`hJ*ljTZe*<p)=bH<i0OHXbx>I#eQ}i1vsD
zn$Jjg=v+wH^`H$<J!=Ft?8v-^Vga^-RlyJc)vCk)@!7wWe3RIM|JZsjooL{@)i0-F
zsui=AuK9@%0#a9r-Om<wv}4Yupein_8O1hAbeVfMf@Q>2S^O6&yID_9nV!s)b@=yt
zJMn{#pqqTQ{usDY0qm`M1s>iO)^60%DvNV)CHEmLyS7KzXRbVrUG*(@!jxV<)5v~t
zNOe6rec9dT)XMZZPtQu2U_S$SU;-!*dR&s<-_ai#2w<3kA}b`QN8lwT1rr_7T2#-2
z$vV&6eHW{DMZLcT5<bV3NmV!^5Lb?_e$|RW{_JIxGp8BhpTDbgRo6@9N@y>w*Z=^e
zs*YKl=r*3*Eq3sAzF&5&(p3+u@+w_R7E>bZ>!IM)1yMD`=<A=8DU6#oY+H6ZS^GJz
z4Ebz#Cp&GJ3dEH|7}G(Q>T#Sv(*yZ1hj<{aoRD`=^q=C&+)3=<M<A|jzO>$BHCi$R
zjwv5?H&0*Cd^lgVr8ZhcB_xa~i(g1V1m*PPEIyW7MK<%#MZfT6Lt>9TbZO@M*N7{Z
z!1*O*c1_t!=feN^DlAtQdEMdl!%g>UZ(1~&@@kM!4mO-tp6Q2FIZRxmQJ$uN%?e?U
zDxPFm%IM^q$B4_^NEDT7J{@8VBf2|*DlG0S3{Jh%E8@y;2N5F9BfGW#z(Eee+l$!m
zcD2H4Ptvs78W70bBZC#4zXsV>!&c`tJ{ooAV85bvmX~DMQeJF&ulY*YVA3P%I*m?M
zco2V{tMH~x_M<*T5)ForeEE?t+4ZG%eR)p4G%H^k$S?iLm!9$enSq4b;N-XH)*AyS
zRk<vVk^!B{a9d!*o~qi;dto+$>+m0Kt6!70qZ$W274qAtsY%RiYPGdwJ4hUP<My#P
zf$kl*75^j*jNtze(QqZU>4?BGp8L~*HvQ)}m7j<89%=1KAI-*k*3S4h-qLUoG6d2N
z9rz>kcDQ1Di|yOzv4_qc0BL{Z+qR5!$zF@X4jGgN@A3cqE0B12bI9}qlx$2J&e*^X
zo%$1y^q;pQ%_SwIs-C`KtfrN00qUZu5tUAMQ+xJf`e6ytJiM9#p!8)-<X4$lcS#h#
z5X<k9$gjH9+WM-a(d&#zXGc=vMhjC51Jwi!Wf60LZaN;nIXwi}2xi9NBj5;;87ea!
z3KnANQK?}jraXMQAp_2^GUR-it({8Bc4?(Z6wSrJIXl<)<jN9u{o6Hr>1RW-Ng0Q3
zuCG77?iGb4b!x8jj(UW{<(r*sg+$rTH{_@0B6OK~i;CHR6+1jATw8T{qzqj^Z3TVL
z8D6%HPiY@OEmb1wgD(}D*b*>7g$YZ%)VKYp*L|{OSY20+a1SKNRG!9YFNPo*?aelj
z1k_|do+?8hvF^!y&!N2gOD27_u@jwI0isNIq>~wHJhkO4%_CZq!>Q>Qt2(3C{vBb|
z6;qe;o0s1-Df%YRC!HYD{Jyc`zO5dnPZPQ4B$ph)N%lgJkX0W*tu=0WAb2B?TVwHw
z4+b-d=ByzCxg=(Kx4b8rI+x3UsHy{t8jm!$da_l^X2=gv$Q1WX-&oMBdA3`zPFdi^
zF$HqvL{<%^Jy1~U&dHS1EUm{}kXv`OxF239v_;4g@OPYwIeGCpzoOYW<C)TxA{Xw$
z<bonFIaY5U!Y8;Ob&K=4X~B!_Rgpn$^YmzX9T~9P>UVsU5b^nLziQn&MPfMl?FiC_
zj%R<@EDk99gb`Lx^hTFmF~Zb|vplIay(0zh5=DKRCS8iNAz{^!CRvwM->2m4L%uAK
zXz9;=rQUP6C~S80z6e%x0XLmt=ebk~QAHRoyJHK<b(e;ynYpNWmqi;u2-9M~j!qc=
zu$mJ=`2dqh`zT|3cCIU9T1k*|;&RWvNtIc-N48p*Zmcwt>Chp?`o@0Bw$6G8EBYf|
zNW~dysuigSol&?rwxHPuMa|2-Pchl{Yl+#;%=n-N%xpc0VWNNz=m;s2_LxSHpETmB
zt9B9c$^j{z_{_}oS2g~`>KUfV8fj8!!0g}~l8o_s&Opx7+@JnZ?8^;jh@i%mL!&ud
ztzyq7bZR+Tk~D<YRBO01c(Em;GF+i!3qobwmao;b4TR1B{-kU2J>uJyA8e3kiH20O
z1qF~s>D6<=LjZzYn9g-Ae`94=wCNQoL1lYANd-NQ;d$3U+w-10A(7Mg)e2{2XWiN;
z<WzZ4_t$A4K<tTMW$b68U;lD3|IqFqH4a3Kd*%uoRvkQD&zQRyCvo~8x|sj<FUzg9
z92sZR8@Z@~LC~L${PAa?72J-9J$3h_^r(9Ei#;F2AD;e(itWCXA@%{^oHqrUj4Ji4
zmBj&<a%^v+E@-1#eVd!HYBYRQa}P*6^oLc_PK*B39<MU&r^SzAe*n(NcVe;f{yU0U
zG@PM4$E%dQ3ncxg?W-AekVm=uIS6EK-yW0@lY0^XH|!O11k0Mm-FthlHE#zAtK&q+
zK-SMySx@gmrCX0#O_|+x%}43(^ka%n=t~wQC02Al3bwB@F$2R(9CBUJoX`5v&`|=4
zEKjGRa`}@Ib*iTp=Kv+_g9__iYqrD=uqq<GGo7xgE-*>eV}$piJKUTX9(Pnhv_)qW
zKbWz^ku4{6#=ds%(ovm7d#M5#soVT|8tcB}Neof2q&oW2M1RlVt#ghq9X54w#dErD
z9x#i36P8|)3)DKO_*VP^wPZbmk`9-VO8^WFG*4V#3W{K2-jc$&Rf2#~eAVmwbU<C&
zw7+MBQj-k}h<b)Pu;C;dcH1Fxl-}@Gk$Dq)?WyYSbOYa}b1C+n)#&yUOMseIjKO!U
zWUdobUUcWj`V`sH`+yy-xNxwIW>$xHl34<7Q>_TWsyF$>r^lI-ukp{_VOux-F=4I-
zaYi-=W=|)~<}zUF!ocU#r(b)865YDOb%$F^7nz2a0Rl(H^znN+x97ET<WXU{@_l1p
zcgsxe=-yiQy?tME(vAz>F?lYPc-GTv<ry(WSZK|?ITkmP9&G~fot~!zI9}W-W;@Bn
zF_DqkiG8dp$o&7u-kX3mk*xc}%($T<iYx-MC@A~B@5lfml87uJ?8puQ!Va=Kh=70~
zvW5sKNPq-_>^qDi1i~hSu!Tj0um@z9)qgW{=AQZf-}61+J$H`h-gBR+r=N7??XK$T
z>guklx8C>ngXnp2k1ly-4X3oWw&x3S5cd>^_@AZ&5Z%}<pWZ1{L5AbX`s{2(;qR};
zdDV4GIj_0KA^6@}-JSJkkVzexr@FeHZ!nq97^UkricQ{!PpiLhlW-Wz@4s7=z&|q8
zIF~U`*+3CZ;F~5~0webgCKf&;OqR{|c$n&4^oAszK9v3p#|#e;9l16qJJE`nCAK!*
z3n-*G#J`T{8y0~Y&@$!0>W>@6A!{m4TB~IR>&3y|jC#68lcCtrMHCM#Jtja))mGY@
zJR(dVM#oyRe%rslic>)phvMl5D4&0ZS>Dc5njL5!&)=vD^*GRgHFvN~-W?8}543Te
z^egzvlsDt}ax`tf;>fSJhco&abeCIk!R%t@MGzRdLgf5$pdoGpssD}2FS}=Cw`GkE
zaC^Do7$RuB@p|EDdB|w)C)2dVEN__A3SpO|Di_KnZ8a0On&+zGqbx4tUT+wDUVmM(
z$4%#OG)&QP?&62rn1jU3&Elw@sK1vlpt~@vUU$y*VCri$G3<q8Q_9=jkDDJxnkaf-
zgmv0`<ts&ZoDV43FIE43e;nyZ*>SYWslou3Ihnj8ifU&0M$~6+$kCmiJ)@R6bppM^
zex=qgm7?7Lar2&cI1HQ53hNN|^Qxb!Xty)=d!-yH)#vZ0hLfJ1^1;yBCh=sv%Qwxm
z5*adp%4CeBJG~)o9(|+oD2U96jNesz@oFI-IvW%HrJPM2Ps+f)c7cA}Iv*fg<`ioh
zsbotw$!bK~Amt6a4}uJdL)COJia6AiXT0unM%@liMw(VNBEg|37$d3403tdxHP+P&
zWciw<qv}xd5M`KS8vmT3_j0^XsaeIRzkP&Q;a*w~^@b>6@S$#C>EyOWHf*qXkc>9F
z{bB?=U`!ZWzfF2wqi<P&UF*ygvCx!mF%|XOH*($P`+^i@OH3=f5OQNdQq&)I=f~+s
z|E~|*e??GmMcAy^{#6J-<7qOpRQJk(8pwtA{S<laXm~P|X{2e?>yE6<an5x4ZC1sS
zA09VUds;cL#4#ft0mAbQZuFn?tC|N?UL0zap2#^q@qbP@rSkCAL7?x@zMI1S&GuWL
z?gK@$5`8}P<>wDUvwp*>&z4?E2_F}iWoCU^g<j4lxOJ>#@WPc-7jFS0eASm}_nNBO
z!W6xZv`9anZhi!u_(;Xj$~ZrG5?PfX#NK@3niwGap|L9QfMFqFap&ftS=Hl@$LIau
zfit{!G%9U#o$&RiRn-yLoI0CKr$+Rr?Jc|yv-JVOW4iA>7+^7oV$JAyzRwWtId2`&
zv4&5S$sr*f^~XjMpFMUjbQmMZ2UK$f_r}6Q>SpDASBP}V_*TiGT5-u`H-ZKJ+J<-0
z1JcrXmaneL94xy88+K+4rJ8BDtyS3*U?4t}Retnf;5>lsNYwfX)L<Jq?v}`K|0Buy
zGcI@8XMGK~>uuXY?OA|E5OfP-m;;?&r=2X+Uk`G3Ym{}fzRh#0$#;mM#|R+*_L9JA
zg!qo;DDnXq7lKmj)o_6<J!y!A@BI-(+6H__hv=szmow(QsK@EWz0xZ(nqTGEJVHnJ
z^~tS>A57!*IM3HdE?0!c`qte{Yd-6;^2i>)D_<fyx*(2xNEE*M2BGI)Tzwy(L}LJ#
z&({?|jcw4Hze<U!!fwaLJC+mFJnFU5gujkX+@vT}#tHbwowsn)x>spvS*fYC)cNjd
z#v&>`eiS<m8uA(!XNS#WDRhG+wfb4}84<P?FN26)<$<;s!=2*sRgJEF!y&dpbWoo-
zXj{wFWJA2w<k4@BZYQM7t=d>Nq(s5GYWx(wQoa*X=`Ptv4lZ;^B6Q}Z$oNPKz@hcy
zt`c)<PMz|wKCo4~x8`~djk9ZQ=!j<>#OH>7+l5zOZCy`*zB()u?#kq%)(k9Q5_wIE
zD=0^VkKFx17`%9v--jr_q&(CS!(Vwh&s^|LM<sdmTwP3EEEBk#TvYqa9)wXU?28Kf
ztyCwz>>*Hy^CmtUiq0y9_GT&%3D30fJ4J{riO>gmk!9WK>uuJrvRxZYMHs$L(~93I
z$|*n!GLyALf}__WL!=g@FO$0;NpMcTekTcecz5Z;;?f1_eHGVL!&S(}rX)09A7{Y+
z!nMnBNb8LP!UW92ssn#b#>?^8!m;tJcQ5?ZIgse!!hm3`Kww134}SjgTd~Y8CWZVW
zDCWMeFLCrmRwsGM?dlBzTL(|QsQnV9$xX@Zb<ap>HT4(#U@hZ)wI}xyjYSNH?+%!^
zo|NR<U^INX=Wi3AQ>8@m!v9bhqO7dUt`u`u`?9mM^YKs{m7j7lBQOwvO%VWKQ|#4g
zYW3|pBhPx}=|nPL3uk^1&UU?pM(pD9Z$<f&zkk+AroOFDMIgxKP<=skK<ARewLe*$
ze{ynzxtNQ%e2cIuLP?o<1nnsx`%~+?w4BszT1}2_kH1dTo%rb^Ap9XqS4473GD@q<
z0)P0R&*T0kAf(^z@L5`Nb=;^Hoe8PA<Z{fUMty>;xfSEzy<6|J{*LrT;shY0x=M29
ze0MMXsKm>PSns5K+@bv6;o)rI(kbM<1)JQYWhkgmPY{(`(ZT^Rt<XFS<0@NXW1eD&
z309sb-)UinxhT)u2eG|$fZWB(gtqsL?c^NxH1Q<DYGQYjbjup0ddo_q7_|h`LeMK9
zFou0^kGSOq6&&!@F_zl0P3els9VX3<=&H11mvOEJJNyau&`P{Kk!~s7n|)5&0_F}~
zT&UbRSJ0d$;h;=?Q>O(0FxLeOa75GvIvd-@Uw(5plEuS1UJh6e)nqW9X+z|Fn*Zwe
zR%$s=e68q%7b<HM12AjV`qgej;X|F7{(=a!W3qn^e+<{*e(e>3#=u%E*>Y@O58$}*
zFGMM$ZIjg(=+8W-qRMk;YdR?#XjbJt5!t@(KXqO@RJu0XT<V12M@`1IJ`doF;NN(=
zpCrxJ;8{`uwT<hHbZH;ll`|R@RIUu{G~7zfsKeZ|EDhR^A`&bPz12zwtD&abp_#C3
zdrgZ<^NtVIx<h)<Axm72!ho&<MWatrvo2qoIGw^=>+6+H24l8#?Qnref1V%$`=Sei
zERGi-iw~k-8c)T^O#^oP!1fVG6SC}qyExavrkXv`YDZK&)?QIjDZBqv!D;L-jbRZ~
z_o3JI<|$ZgdOdMtcs*4IKVUp)z5+?|Xp<+*f#UfMjWC*3sb`0-l8;s)YPN4p`CUY>
zV4j>Ri$`PcxSajP>SvYAFX{amFLw({$>UlRO@$-Q3zdrc!s}d^vV{o?U`p%$IZDSx
z&`ulkSk%C%JrRQRR;?@$+!$Yc@gfJEk=o(dxs9iK+y1vkcB0*PPEE?+dsI7flh;aW
zUt}>d2_~H7n;>h~)vNssTsN&P!ROFkn4c*iFFlCt#+I4KX2yJ@azejuG8Sb>hQ<oN
zkDSxef$&L6O|x{^kD>5bM7G(su1n9=H={3A#v7sb^==HZ#esb;J`FaRDru7T^_of<
znc8^2UFLnEm#Z%*dgk*^d62H|UDIxL*I0@1_b5Ge260WA`_sS%I7%=94P4whL7myA
zg)@xC0|(G2gdB+Pl<vGcx&^#`+f{-WHl!?z<=urwu(h$`R#+*&SjO;Hcv1aC3DhNS
z*QaXV0!4IB(zEca8G?XQz*?GGT)tq#-4`?MDQxXiQJ3mhPfch?h7BN7jmDxFVsDbb
zK7%hpPHpg01kQCiNps)79?lO*CRC!P-Ls$FLz3mc?6cJsJyp}=)mupiqu(?eIv@C`
z3^qhY??MIa<6=|2j&wRq{5q1oLJi8zyjGz25!jHXZRd#JJI6OEYR+XQLW3g=gZmcf
z5p^!`%~MS=SzzItQiWc2cf$dEt4ATnpNoG&_-nj+T#4U#y!u0TMl*H*0X_G!HhH!K
z+trLuGO>!%R*1n}QD@_dI@+wOf63YXL`7LD#`qi6R#TQ8oxY~ew(dny7kU@N!GhXn
zdETgc=8eHz?)&kc#B3NwE+xocQN%L;@S;&h&gLCyWt*2y=r$>WrIvgQV3m;58<zbE
zJ@<)%LhP-0vztsQ;PLg(Aez1*pglakwzRhpKs{qy*q8(1Q1NynrPB<aI40HJxML=l
zs0|5@RzTh2Ha$IBZPS{;Uk%VS+Qj-%2qR!AJ&>l<2dzrFZ&WV}&!N$EXh4&SMkCN@
zbe6ZI@-89^V02pi+WV_l^0{?SxtU!Gz=7er8*5y_(L}7az*&eD){F}{Q9{^n3@mkZ
z*<?y!l*kzKI6;Z+2epP_OS{wjc!Zvc&t84jgHD#RIR)0IenY+<)Axp8^~ikUZA2ll
z2fc`jeyLP9SDmGa=FY@a>bgOfaq>%Ge_=~(z=x$1Ce5)7qxJNmA9Mw59KV`*?(m_5
z1fu6gG_At<(do<jC;WqzWPdKL-}K9Db8554`<Yax_l;Cu9^s|w$lHA(WD|Rqc|?+e
zBpETI7xSzyfMgU?z||7{9f&hig2FVzr72G$2mK~0Uzuk}D_yN}TJp-}N?)l-3p%LJ
z(rogZbx^b!HvkVm-8M8(jw!h+dsRe!xE^fK`La)?jjTGPu@11G%&{%(CyDUiNqj(c
z;!vGd+?GbHB~DK>{!$Aqvu(V%ZM-rdpiw^X#Bm1Ei0!Q50ChDJ!&FH};b+kgO!dOE
z2Q7($etFe8MLtB|j6ST)-F?l^t9Hnbh-a1=&cYy)rC%>SY#}E+r`HFYmLhSp2V{&B
zR8nWUgimMH@o4>YSPo~X^VLh)UQ5ci^yQYg-=05E9NPQVY-O~vfbJbz;hNc^D8qnZ
za;+op9dwF-TQxU1(9AwY!*G+hqwQbRLaE<+aQDIAKS7UIt_Npjv9f8RvAekYQZWv3
zOKzj~n)XI@lok7YUGK_SvMLH8QLQ-gxqRG<=hXmFcbm;+Ww%_wT+$rxET?a_Azenj
zmOxc-dtsL;Xs}(n+NFvhdp3s@i7&H3s*uurPoM933JV_5+q6w|PtujR=R@dB-enG(
z-+=p$ED3aA;+bI&s`S)l1YjOdl)*DQYsh|B^g_Ap#Y0%b^hr-O8j3_LuZyr!m)QJt
zL_4oDhjY)z&BT_d+nja_+%35gbni~~&OG2a!y9piv2d%;?9)_blN0PZjz%1z=%PRX
ztFKPSb9U4Y*mW2Uf1^T9wy`z<l(JzQ0Ht!YLV|X!+?ca;q_2>Sax9EPADAheq+bav
zNBB(Nc|)3MWx9leK;Q!)a+>$a5&gHoiwCajv@&A=n1EumYxajDmMNl%$|V8uQ?8Bf
zj=q8ZD0%n8XAnP4(l~atP=2BeWIVN8Radl->1HC!Z?6+k;~C55LfO7>=hb^TfG2@|
z3K6%U=6R2Ej9D(gW8ifzhUE9e)B)y-zOw*l-6>2YES|2m5X<Hf2<)<(Q%qra%JNg{
zlx~~^Rt79zVb@S=6>$lKE(d25z|l8v{r$QAsn}mS_<w;8{u>n??&Kv+Hcz#aHP_nY
zrW4q%19ZnN0B?g_DUL?$@i9x0C+;<Xh8Zt6{V|jP8WIPKOG6A>r3fGp5n`Z}!^K_Y
z;8@0`7_hD7vhKje<6y%DfUWJ{lO-O8#e`Lrwf%%?JFD86U+Z8(XG`PVoW)WxXgsy^
zDdpgcmUjyqH)3^Pi}~97bG2!`c3&ohci7E5hnJwsaGM1v7Y{OYT-JL5u?^Wl=hG(>
zbR9Dkh+re9ye(yyY%|Vt*q*Jf=<OqlO#Vucfj8;M(NSRT`Py-bf7^27f2EvI?TrU~
zGOaWNKXY*>dqqc+>DcE1WSxCr+b)bc2z_+iP7i5~@k0$${`qCbe7Qx%*$H4Kl>>&&
zGBF$t55!0pNAn;bmD1OekGRqtXWSxUF3Af@6Onf)E^^-EVC*{7;X*R;0i=?EJH3qq
zQbu?q20LDoyyYCfsX;U+Xi7skzJA^D!V(<1=^?&hWRO-UI=>MtPU$+Lj`tR)pOlyU
z$xHLGu(_*AOt)i?=d7R)3GVn<%yfRGDmJeUOPh-BeukRnLA|ig@MC8d;78kaw>7^e
zK&BX&?<VrH$ig=UWN&NfW2;2cV6`t@Sp}Bo*ghk7AsQbno5htgGkdx~1h{;+?^yBX
zp+wIGrF{`Av^><$hC4aPM~|~Re`Vj-YYtjTcgZY!=A(z-I3r|q$77Vj6Xkfm!5h`n
zUGLJB<_6%+x*o+#^7PGI^?_dCNRi>sJUiTlKoN4=+`&K>MAPvWcN=xbeIQgrg@{c_
z#Jw?_9HZ5G)y7vEj?~-SzgKw=OD#UX`MlLE)nR@){c{vY6qt$OlFwJ`kMEfklrKH(
zzH%KSq6U>vOt`#14mmAu9ajev&b@E8`04T9RD4zF3|8MkGhs{fLX#15*e>&9>}FUZ
zv}H*^_Z4=~qHQG&y67~kgV2L@Zq<3`1(+`MPe+yADalDp^3HNj(~+d{JOY<1n8BuY
z>P&lsIMR7>P7POU?3ZCx;rrDWUM)C4d#WaF;uG}CA?&?|l?pK^tDpE&{4#G1?5DbS
zv-sJAy2Oxr+(KhT1Hcrb=GvPy9*w8^;ow`xE$DlMKZ~&B|12liaC>Mks^6+0^_tJE
z?@`A8dY<{c44et~>b_CA<y+a+QnQ@{vN}&*XH&T(mk``TI<wa3bNb0EKzKqa%iQt3
zBW*%@eahauoKLUk0U__ki-n=V;odp8Rr|ijwV`7T*rhZ0*d>Nl<c41#`Em6pyVHP-
zO17$BH9S-N&~m03<M%P}(y{E1;Vbm&vr{>dsA=z_L^U-n0sbwBujKU<9#BmnV8{_K
z<k13rNTz_*<-QoX&ym#EWB6W2aR!^4$Bo4r*QAQr*-c`FN6KLcEEzwS&4)@qV4Fqf
zCmVpk05dtg_?iV?xDV2qqRF7g4*SiPBP(oIt6Fk|9ea*A?uqS5)O~B;K1V9UGEa*H
zcLh2_s>j4|uj3%b0G^luX>H!e{{Cj5bB%2TekDW58n9q)OlesTIVZem`o;`D$isy(
z^om*AT<O^=fvtda?E~mXH1^p-_2>gBR|^-n$Xi9jOJ=w^al;)<<2S0<!weNN`|j!x
z#XwgH>viYXick&9-}4R|(7`guAS~ahbMcASPVl~Ck}}#Ax9vMJlJjZmg|f+MQUc}m
ztR(>8hQ3xF4M|5lYDZ+A;Mc}B<1jQXA=0Y%^fA0(Wmo!x*YeZ3u_6Pd`r^IMPC2|O
z;uiRVYI8W#g@dI`9X%pSwcjFro9T1dW=8&Qg_i5iJ{OOTri9E-ox48*;0!q3kMT;^
z;+v)wsco>ehU&C(LDHZxv6j~-m6_ir>WWg^836wKI*)<rhv)R^Ph)|pGu;@;`}NDt
z8HH`F)soC#t5CMXra2BF$?Ne8URZyTE{!38c`d^w4y4FzRG*m<<d{6<RZ>z56)|7T
zhN+Onmz}qBUv~9U^lupdv0?nP<9`;&Qe>?{HzsTJ8`b?iovq6sXruqx1^jdQG3<VR
zytXi8MX6X?PP=~M=UagO<X>)^b!N_6#ktIGUsH4O8<p89Mus@?NH+0#O+Bb8n|Par
zFk77XdxaQ8BTb;ThmV4>N4)EBUPqdQE4?zW)_`;u=^?9IOvz*|NphmnVGQJ3VpY3U
zRn*rz0)e`AwK8Zz1yC^!5)!VTN0HT-7{2J`?L~A5+Bv@&Vz8r(Jv5-B7w1sO#>jTx
z#<rP1?5ZM<I@h$q%Q_M9V$*TTY}bowv@ZU{iXMKA+f`;$j!b1PO#@%N%}>am8SEd)
z2puXN%*UjmEpoVdRnRs(Tmjyt&p;bU8LomX@1qx+1V)Q2J<WQb+ph~PdrFe`&olHY
z`sIW)P1<{fWm=7`8h~^KY?~pL_x|Qz^272dB)9sRdSQ4SX2dPPRAV9)y%5K&u23cK
z>BFNp`U(+ozdMs)9amear^YVLqb#etOnCi1L*e2>_c-=)SW%GVZHH2?ct1}rtdDK0
zo;diR0~OGxwNX$`f5qJ^-lFFzvZC9EBC_JEXJJm6*4P(JVy|3ZCBS)`RR)M8u*#%-
zd{MjRW6_cIMs!A}&Fd63_cu{_X*@>En|(MYW}l6@?pPE;PfwOjEXG%37%-a?4dpAH
z2k_Skg)o!4Ci}5Nb*_e&NJ0-LrPp2xH%O`mdF70wxdQZH(J}n-oBqVgSf)}60oF~Y
z$*_~7xS=u*(FF!qBy>Pm)54Eif38XY|D^umzKChHY3D6qD7jhQG30xz>$r1af6~%9
zeOTIbAY2|&G4mYmdYlWlZd|je<I^(%@Xt+MlFZ_g$HFTQM|oRy-NFh&Ztgv`Iiqnb
z`=1le1+S@n-CQc!9`)VUdH(Y+uYjU8SzT|LIuX6oMR^-`b?BCp>2Yx-JIHOpJRbAA
zv9(r3tRm+)9sw3bVe%rGA#QO;e&mCM>K`v0FPr^96<p9%dT+N`){VnhG4%AU2Y~QI
zkQ?_`egY?PMQ5+t?aXUH^vAGqs5TJXoNrX*f+d7x)dkZ{xt0@fw2akf-|kPgw*X+G
znnE?DEhNUI>uu&m;_sQQGgajEc?^**E{kT_l3ibY{FQ>dLF74eR8XWe{Z0VDk%x{C
z!MD;6VuN5A!ma*%Klk*OKXxzi3M3MHYe%2PH+mmhb?2yNnr~~_X_*#Ar!V%hY=}AT
zH~5>8b#GHEIPnShd&aO{byL#w0eH!c<fab_>MrdoYjoLU>oJH;ODI_bidMVrNm3o0
z^gjQhZsjsM2G)VErITlY+Q@_bG1049@+*FY(baGsoAjpukwA23S2NPP&Jxqu!v`$p
zR<rF7;KsLS3+EsfiOTcQ2R6aK*-x>HzhR|=Al6V_h$hgYP<?`qLJTQEm#b>$Vj}}R
zkB@ky{mnAfAT}QB{sAMksN31YZ*Zp!{j!5RX?v2!eHv-WEUejX4RKAKsZzVE5HQ<k
z=f*T$_fkrwO<{WN5LTcPAWZ=Eq1zWH@lSJbp|K&Q7W1ngdIO!RRsGA*xk~{Wt={7d
z$;lNA{e1I<GyVGd_;5{SX{DMQ?TI|*a_Yc3b>0-s<xk;kV!)$_0yI&(>AQhj%dyp>
zll5L(OGKg<g;Ytj=u9ChEn@2y0sS8epp<Zcgh&xIrTPG(EvXL%U@06Qio1-9OOcC<
z^ZLx*AKSw}KmKPG4bI&y(-AzSu+SDivqG6heS(f}i7EwpOERx%(jzHD9Q!J#YY&=I
zVpvg`P8gXbOhbU@byk|W9VuZX`5<<9I)*mFIO>7#<;s?_?PP$~9(mAqMo-XOQPf#?
zho!me-}K1(g`g=j^nO@BprqZ6()Zckap>@=8by;VGK}mZ=b)#0c3eyEMy3iZL6k-4
zbh3R?6dWY;)uP^Z0+@qwAwqHy=D$B{<!X2xd&rny3E6_J6B-K+-$34+oYl8*e?F+`
zifR!hcQ)A<zDKd1Wt4t6t&-8Jppk8p#JL|(R+^BPM$mku-TUufo&VdK$6xpJ@A_i@
z_2m9~a{s>F(*NYvJ8LocWnmx$<=k4XU+x#&!ICxP)5A1|6qroAIr*w2&CI*P3cC)Q
zwe_y*H`q5aN1oCsOjnm>Y3fwb)j2H|bmM+Rz@xu)nhM#@Nd4~m+Fto6F1v%yG_S@`
z3%qx^8yIdMPG)D@lfM!R&Jr+THeC?poid19YCzQ3N|;MD#u!e-g~*7tMEifE`rq7Z
z_N>Os#8_@GwetGZ$)>MtW44xf2Ar0GmBSm-9;1tE&2HbQ?s_Y$R$7-Nf}xR@`Q7!M
z^Mi!a6{IgaOrHUZzKOr~^dnH$|7s&v4pDR+&XF+cH1Dh&zLb%VUnq1-*Xm1BKiQ=q
zvA9y&8B3jaciuL_tMGaZYes{Dx%D;+ovndL{m4G2qPImp*N#t1A7u!?ZBAHT?Ur0F
zvIeI079mMse-sZ9G1psX9$jgO(I*N&AV?VyYm*Fvt_JV(&9SGLlSy=xR^=23`-<Lj
z;#yeOmEd0A!iGt~eCU|K=lQNoS<XOez}fA^4V{+PB3T(m%+oZ8%df)?YeMF&TXGY$
zZdQ!Tp4V=v{5gNR@7Xg4c|Caa-eP{U7umZ9ljf2>XMvwZF9LLG%X+dU7VWi=C1STP
zQMnvo&s<=@1!mZeq5xvGZ4JLB1jjLRIMtcGaOzo{h<U9i2r0q$AWh_$Fvy!5M{ONW
zZQa5QaFCb&It2k%gFIRQ%yQ1KS#)7e1{zy{%CG@6u_%62hVxSdV9r1W0lZz3Bb`-2
zoTc=Vbaax2y?oKDUA=)^Dof=Rk8fRjvyFA^O@0ePA8G8v&9Z6f@TeoFyn%@axLlgi
zgxMI2dXBh4j&135P-dLrz_wtJCl_X%q)n<u#j4;Zj4*JAjFo+e={ta5f~kI!n8Jps
z{YKTk*hl9_F+e$$>O#3CvV8r(4S~63S(~j;TM}l)EUWz}7}RT&RSlR_{%r%iW2_H#
zgbK5djR3!A__I?YH{yHAVn~ZsgQ|0wHOr;TY=0>1iO7e^2X}nuj&xI0K9)uEkN$q_
zc_@<=ha1b`RsANs_gmqA0=eh!Isbp{qbH<P8I@k|ZQ{f9eXlVF{C*xN`_#p_d-+Fs
zSWZi?J3{#1<;Q!P(AxXTA?%w+Y1LLYicg<BUSarGdQVZWjNzEDto16{Ue$2Uz`zTM
z-t1YfBeLetG+!2O%;uS)I&jQ9Iz<F_`4pt@xhx?bb+!gEWa{gM*>@6#>{1<`Ez}xD
zuhyF$%G=>bDiGdMaBibFItzO`S=mk&Z6tNoJltzqiM++)eGt|-h<D@N^5CxW#aTqZ
z?dpk6;o$Nt`k`~_Gmr)B_OjROgg5WY7T%a@D`eQI9&J;*W_+W1h<ShUAU7}E@f#K4
zjH?rl{~J|!TN#l=F^1!n1)Iif!_UezI`&g^+ml~#CA^MQjgU+ZFe2y6%-@^_+%qNW
z<?p0LH8Y5}o)p)51P*=Sk{YwL4Nc436ogjbOxEp&Kc$JvINuA#k3LD`LP}WDKS=uO
z!fyc|oOiSp5#ruvMvEaMRF1?RT$CdoKFsTsgV;A@D}dqFE%}LN-ciGIeI{HxQ}KpH
z;tP4iu}?c++AQTX!CIV+dduan<1?-nmp$|iGFnkURcm&INDhzTTC@CC7wPEf8KL5M
z40B3YCA(R5VU}NAGFa9@t#C<ddaKB$3mR6HvEn~puDlSvjj83_LOosG$nwF_K!`-B
zVQ+D9K_(Omk&)k$mZs;<R0&>4v}NchcK`r}W<l{no^5<Nv232L&#PQVaFy^jWwg*p
zxuy)ov=S_3R;?+HNBSIFPZ7Z4bjP9N4g>-rN<$5S1$apa1k}~l8|vT)tY3@=`(tK~
z8!4y4h-;0d%Ys`fT6=5f+{oLGyf>)FEJ^vG*uB?L7iS~c6vVp_Ok$lWBFrdI42wrA
zgV;#G!(q7{uk}?kTwAHeSYsf?!jvyZxfZcK8(;tyU&f$qGPq#rs89+CO932m!H^;I
zbySA!%dW#7tG%wB94?2T>yxZnIZOvhCy^@;!ZkakTWRS!TVB(Q3Ubi3^tD+q$xZe5
z6t~cLM*4=S(<V)(XhmE;eBpafZuQ<ZyQM92Ld8gJY_7(Z%W7X@0a&s}uo)jf!p*VG
zx6LWT@vGesU$@bsif*on8wP<pf(S{rX;MzFDlDPiYP@80RSz%UZE>=yWB9lRzGwKe
zmgyu+grg}YviV_X|IvGiYd7TnP}nmxYy4lIFZCW|3xvG3zhv;MFFSDjhieByO-iGZ
zdop<;$#W0PKKvog_lf>M*WeT?5TL%bkbLxJ@s_}^eeUt45Gv<5nKN|{b<9;7W}cfQ
z|MBw7hhDD&_Y@ZnMd3}_rZm(vz`=i=UHcDKBEoFB7g9W$Vh-~LAbN0hKSRTs7E!f|
z^)(%^dk17=W?z^gu(v>EsMd6yhEAvL?3#U@02?lQI#D4!2X64Xd~9L$`RRxI2jo+0
za-Yvz`B%<m`qj_(JU3S^*X90cd&4Y%!1jb48!Ferc-jSEFD~Dx%4t!!ciFQjF{>QB
z-c70XG#o9eislKjyc99^yCDjIvOgWKI(QMVyxn?sbtn@~YVyb*xNwp<ezz}f4Sf2t
zy+fmD&qxBdRSs(5ON;_*Y^rGu2?nEV9DWrjCVKPUC*tGfrKL9PLGjd01!%MY;1`}^
z2w^l(PQs%dU+VJD(c^8iT|uhH)aSM@K_J~N-B8uV^3?YM_f6kDUn&YX5z1m}`PMGZ
zGf|DBVnz^B)ruaJ2XOHUL|Bl5`mJsqyDRZoY}ekS1T|IeH^1XP7qeSlu8d>mLyK&W
zf3AW{8?9BJO7CtXGXzv~#0Zl~@$E6#k@&oqF*b_4g-md-Im7PJS_7#Sho50J=mv%Q
zl#nB`c^5E}X_fXH(m3o1ahe-hwyhHR`IT8@pGBW@W~{N)c|D1heCG-4NFhG=-B6UB
z<xo)yO%1wer{2OGZPsSd{4q$V25u61s0ME!eMX3|;#pKW&VYj}4t35LeaMy7?Wj(A
z=EaNg8tYYnC17*N(i%`0S!uc$PL?vyed${m?~%7^QlH?3ffk_0eRdWwRRKW;0LXno
zHVo}x6SO9=ZRgz%w9zvQ5?F(#ImL!p;qVxRy4pr+lvcv+b1;QquJ<e{zU6z&ExKSq
zR0}GBO*~$%G}17gr#IfbfNa`Uo<DC4nOlT@?#@u;M=7U}c6fN4%WWp*kkv5q2pyh3
zI%5eP>oa5+G;M3yS<h6|Ck;{FB`-TC!0Q%|+{AYro0St8K_70Yf+})w`3%ZZFR$0k
zo+~s~_iO~GMM!eBa75_o-+#v>ZmX?M^Uyj$n^vV0SY_xcj8}PKSj*f|FB|{5P9gv*
zZR=d+f@u9E2kV0q@w`$d5FJlR1pLuZ3aL2`v4(`fS_pnxm*DzIZ)ABD!uj4Jos#|4
zDPgvc(>q%szp8LmBM_Ycy7kAm`Fq6w?2_44UyryYh{w?C#Zjx_+ZrdM{#eMVtedKr
z*#SJpB(iRWP|i8OM}IqdgY!um4tYE5_CQJ5%jNSAf0zF;@jo}tCHw4Rt1e$?qeS*V
z=ABcPxBiwrJ?}g9X?g%cqR=I3=;e#^5|93H-Tz*KsdG2hpUQ)v!h#Ks<4-*1dLA0U
zZlc}b21AGa<cwW!9y`If2HU(Q=lB@&1@3ri$<~t(!Bvjzck+>O0Rm;`I+^abu}y>Z
z(!B~ClRRt2M0U<CRgTt`E%??MLCoR)iq8!9NKe|P%v2OIMi&=H(`N^WgOGu4gdPN<
zCt<hubb}iyBR8wXU>w!*Yn}KVeW-b~!W-3u?Sn-|Iz0V2x+K6ivq+mhApo1Ry8tdV
z=t@9UIs_>dzI}JWSE;a)^HI4p?Pd27zntKa)Wt4-7?werKxwYo60SzxvphAj2oYzI
zOX)UvmICLkimmV_4@M(a{GG{OV_n_Sz7BP;h_3ggNSNCMN`5^nuRVH+lwA)x?5zjw
zvo_YOf}CyNDjztdT{@($T)LNFgj2HsOMsF3QWF#iSW(6~)#LqbdYF)GJOADt7-h@D
znjLQ1<+E;o-$BtxYyh94<&j@EDvZ#o_GdaxCd~J|$6yRWX82ZY0dlx{*Bc4>0P;p*
zM~f|@7sqqtKNuK7H5o)4*vGCCV9VZ5Db-#oRb*9))>zg71hh?1_2WTgebW8-i0Z`0
z%?)tLUaZe{9u`_zJh1letNw-xb!qOav@6a1Qy&*X%1tCV;wAXVtr6bhN9OHs?!1@(
zXmAzj>s)VpvEBhX4%j6^5n^scx-hPS<vGfZAVV);V4^_)1+m(YM^G`jZ{Lc}nJ<#!
ziy3;qq`?xbHNw6Udn+~kfN>{xrXCl9mupypG3Puu2Ki>8?#>cs&YH$|NFW+;S8~K;
zMu=^W)5cl0a?xB*(xuT;=K)qwOJkXR58oBSl}5{@uQtLhV4iF9bH|1|0Xu>+vku(Z
zd`n}P%+7K;SpiJ4SMi$IvbAZfpYwi_g*V9p-$XG07Tf`kuWT5Xe6!69u3!G}4t<~K
zUo+zWnKR;ZX<oTs`^Ku+UvCD<lCm6;u<=l`M5gaO>+Y(%->6(HD<su1`9u@J4e2IS
zmYuX;*M32z;gD;2qDx`j-Of{Wq>K#<X$W@99q%h`ojRZgnCSt|g{=(6(?BF%?l{s1
z*stOy0h?mLmx42lE#x5NOW5<@nwdY2mi-XsPCJgb4{>YX5QOjEM%%?TVn#hr*%@w+
zYZ371bne<w)H^oI!Nhboz(wb4+r386W9H&3i_RRqPc*a2*4%0g&1^mG-HP=GLzRdY
zi1cb?>>&BJkyLMaPDnw$*3d9Ev~0$2BL_W?zNkn$;TBU2#S0|Z!{$UtzPP}!3z8F9
z<vPdm<PRX2b#7hd^U66LiPPoIA=|g@2^0{1Fm5VERO*mAq}r*WUk9Beg4O~v;IC#f
zc!FNln+YhG)<gO;%Erw^AQnOk_w7+k4l1voCS*SK5}CF)$=O0?`^9_)2Aewz=&Vr}
zlw+}`Cq!BU1QOmN`Hs|D07P!mB!D@UcQifQW#z-p=ODr{i;(4W03T_)n7<Fx^lCuO
zvv}LZ$S_Hkjd#p5iW(+D*BP4^|AI)<?ip!k!6u%jP$kEOLd8<QZaMe#i|xHv{_%V7
zNk-^a+@gca6}~xlCf>{jW3)UWNCNGcDUP?nJlyQEsAoetjMwFSc0nyby<saU8<o%K
zmoa2un?#N2+VxgSclkQ+r+}3W9}gE9y{vZ0D>vb@BT)vP`@C?TnL7A7y^51@^#ms3
z8?&tkwZIP>;La4Mhhs3_G$t0n<m<2&%amj5Bg~dJmg3Dmzc<Q<k~e^l(=9tyYZz$K
zu@$1Oo0*imaE!|o^)Z;C<TbR8$|WZ=@|&e~<q@mB0o(Ju1>fg;^v7KPtXZMV;uOmu
zda7iCv!eNJT=ZGmKV<$xi_+cD8wvodF)u#b$wY&J^83L%s^>)$Wa5sV6<XQL4>PU(
zF8_Xe4R=<~n5_GUf%d9}3eV3qJo?+Ecyz;~+IWF3L4W9CiTn*F`#%=pzZ0+e2QuqY
zG0S<lsW5tU7Y()grOrFoGz94@sxUSc&QiXxJ_{~0nuL3NB-pYwxWIiG2K4$zDeIrN
zgPyhd3SUs6(=AA&y&ipUXUDTpad{kG1FwfH&UJTZ;IiO|cnDUZHl?CgOYj{7({jpl
z3O%aj<ip{08NJ+cd$Zf(YBVDiMhFO_va;dshQk?w-z!DLv%_Ih*}Q#=_W@#kL%~-m
z!zH!UiHEWLbNo5E>)AXP4BEF399k1f@tHE{$!s?!ypweb&4KTa3$nQq`D~s|Y&VDI
zzEQ=HYOnA(M`L8~E$Q0%;WE}1=6aJs-SP~2x{j(+Lz7wFIx94HjxI_>Q|6dTyjFX|
zYKK&J!Ss^fs7?myyc7*j^I3WBBz@av8*AUA@2EE{zQFs@lxtyT(DK1fPPng@dLBaC
zHGwu#Y*WEB12~#z%ag_LJ!fh``Cw-1=+|E%m~1r=lG5;UnC#U?vxCkV3XL7{*~xE)
zwOXXKc9ufw1TIM41|W5y+yXpE_7t;f5~Ia|w3ZzK9uRsC6K}wln_@7QJTS27v+m4)
zLpv`Ib&PHR3-HiR<N_6VgrC*7*NLRQ*$J33MN&VAqW;yIiAKzpLaKHjs}0=3uUghy
zKt;xDArJt<&-)m4g9xbOh?2kQ<Tg?)z*2o6gI`UW0OcK^xC7iy;`QiuxEqdEWsld6
zzIXTs6jkg87OoqjV}f$eN%lU&J+ZcC`Nc^mRku++l};$1I(!l>=Mckg^8JpMi%V@P
zQCb)0PIG+%%74N|<g7_mg2I>BsosFsh2dyTroK%_ek~cAK(Tiw&n2Fn)~F%aCVZnx
z)AH!_HB1saDty7!vR8L4`N1FVo=x&Y{%7?P)%D=Q?I+1<_kOROlaIfYecf>CogYov
zHpu+h{n8)Oe4psa+iS1C7LqRxp$ztt3aK#1%7NqXqup1&U>jf6En$znjTdST_kR=r
z_G|kXuhKmbKzf#XBq}8>2Dy_n0D78ii#&a7@Y~KRhzg*c2FI%&S+KVV6ZN`Uj+VW+
zqV|5Keg!0N*1C@FX6qcfe9)2rrm0i3>?`X3Ms<Kl97FHiv`~Bv@H*>TSKe7@{_;Lp
zwYP8QvD)!9zXL49sk;F-_2ngfw5+x2?k4ah3JtEFVpCynHiPgkO@tb}88zTC%5l_%
zb)_*#zI_k(Ny#$`il_;^EKszANntb+*G9K#eChN#F5YoC18}~2LX}bQeRsvkDq>e|
z5-KAz4UR})vFsoi*_ox>mbsu_@r|nd$Vw-mz1p<}hGX_fU{<arOo=#%O`SA&cESLo
zHCQoz7m*%kUJn^3PhvO<R|2HMwx)-?5{}-;*hhD?HQR*krq1#oFyhK`!WPU8R5#sj
z=IqgH3uKMIERSH=+JM;+#)MK`xzY^Oc|h~mvUNO4;%RJ?pbB<1uaICJT5*DQB%gzB
zo6oXG3fSAf`mz9iW%S;7J%o7zKWLP%(Jd>%T_so<rJQ0`s=Qdnozhu4;!@}`GXBCj
zP)d*3-i6|6wuRKwlNf?LA;YWSFt7Mg=OcfGdj6#+8HSI3t%~pb^AGg@4~^zG2mQ(^
zNKaE`u;5uDYsqWVT2#flC*Z<5vb%K6fu=;jN_}rHAnlIze3+F6vnqg;(8?gL{u`wv
zjj?US{nkij8td3&l!O+}AU5%xF92r3<e6RICQs+HC4>x-Dyt(;g-*D@J$icTRf$DI
z$@ykH;P>`+#j-?99swdDJ?o$>BPsT#sH@sdr!)ByXx^6Vo?Z+H`$VZ$>?fvQ{!m^2
z?!%8CzIJO3(B!`v4a(ZmRT-Z#mXOJ?8|%&9$lS(oJds8U*e%`k16Gerj*d8&-L`{#
z?ah0`jy}9pd*U}&`F^K1vwdmyb2%Hm-J~?ZL|V{j#i!i%y`3Q=>*JU1EcsVmnOou|
zrV?Gl7fv;07#4*@f|aETB|IWQm((9UGzRKX>dB18LZ)WEbN%oj-^}y{nUFD++XQF&
zwMZf5RCv-t14~8@4Hs@AhDB^iD9&R(Ov=Efg)LBSQ|x^0BOpD^dVpv4*_R3`HY?Dj
z)3M=EuK;0ok50aonXPJOZ5TJ@m&aDW0J2-00-{9*P43+!tyJ2{pPw@WGO7vEouhL;
zxUHZUN5v$y`CG>_DoH505>~N4yCV~uQ=^NNhyrgn`wTeQZT8vf>Ut9?`rRy{_0-pK
z9I32ND_G_25vt4{^R2wBAg)-BG@9$@ppSmd2KYJQ9?FnWtZMev(N)k;(6Jj`@9+%e
zWM~qJcGc6A$d?+~TyII+fZP_>iWhH<SjNHXz~OvrF?_FL0RL8CO2_uoty{n6^<DS>
z08=cL*ve<O>d|71=h+E|M}~TuYMS|1pAyge87DC?cecSb>ShC%Tl0O7aPgP+*<?Xa
zA9hyrJ!nCNQ^USQQqQUZ3S-*ix<5JX6<YI3?fp(zfR2hIGR5#&?Gl7hB_7U|MdR3=
zUv96{J(!v=vgjkF{}vgyM-l0Iu~1Se-kyi9?dhmA8SHRQdgiuH8||v@R6pu{3N;k=
zc8s!vVP6^7^e?w`lTO~b3K=DMp;AoXwImUSBwl@}+0y)+HafQ`4=Iy{&T_3?chC}%
zv-zO%ho!22``1~g2b#d>q7v;kij3S&oYj`BM0znW?a$VOE1>Gn9I~8lUU(1j<uR+(
z)Mb-&at~|`c$K8}gvNji!IU|&p@b=BU|?x#H85n!J8p0n{C9;(A()RU7p{6@>VapS
zE<LeRKnia#Xgw{~%E%MYEiL1kXhzVCkn4(^Lz(Duk9|tWo3}n?C-JFG?q-XG*l$of
z|9qU*6GAE`oYi|+7H?AaUh~Y)xmMxL*?`~KjvHp8<KSp#Gk;2SUj?A2%_IC+@Zk7f
zehni@N#m5|hhkq2+5k9^BI}&;#)Mq}mV`~R&Heq*0f6_o=~$_{{*9_-V*bbf$29Ht
z9RB(Hf8^zpWFV(eXv?*$4^dYu%1K;lNjr{js`l-V=0u>kskgk<Z;-s~7r~$rkA1e6
z0zUZqyPnn7dQ9|7JZ7uL9oyQ5^vcovzplWGbavp#*GV$<&UnSMlY=hk$L<vyh0B2_
zYOXdc5Yy-F5Np$?M39S06RB?Ilxy1`x~}0zaxXe$=dFPbF8YMSUruX6^WWue#qZj6
zZJXV*>-3K`vhJ+jcVnDw4J#ev7a24oX3c>`%1FKy<9xh^_s5E>&&l`03NQt8J$V=u
zQY83k`Asbv<EYU`|52&5ET0Y5t`BAJsu=G7guj;W^tU5Fs(m#H4ybV4bi2FH-}6g#
z^naxIUXdTUAAUSAT#s5J@E>XqJc9gMX~4yPkO*c_5m(6GORWhB=(twS_xA$)NF_Jo
zNg++1Q&w!S3of@xs|*Vq$?D|%ysBP+6(jcBMPm&UHC+dk-H`C?C-+S7491`&k1G2(
z$3BN{wS-|l*ZU-)a`-Mz^t8_C64tc1*3D`_IADlv!GA#6m?+d?ma);~-q@AK?lwv>
z>sO<Zm}w6Ov)2v5bCYynr6X<?o2juEQX?nrEA;ft`1twlqLLl~z0hFAmD}E^Y3ceq
z73c3cHB&owI}N-{$=kh+WaiQ0y?XSpOXi@~EfX5oz|{tBdV!icWzmKfNrkYVjEO1i
z>rCRY^N3tD70i>j`bo-LdsF>}c1$*p(`?3!g%Dn!JCd3^B^4MZ3z@EG#+$d>Pes%?
zsJmA5Pf1dXz30eA%?JIm`O9&|{quq1PQx0iG6k}2Cyx51^Uq4!e}vs0f-CsaOr#3k
z(5-W3DxAd}`rOxyXBUE8rtG!Yr>U@U&A0HL19?VrAZcvQ83VQRTR7ov$-oLx)c{=W
zxj`;=puBqfOq7$6e0VCHPMuxGsd0fED>j?qYMnsC9m&EHL-9^IDQ-<;taegPlSXdS
z`j;EWh6LX~`~zD4A5R6`G5Y~KIY)EBv~;Fk2>Q&CL^RCAEnK*HBhJx$^@`BXF7<bN
z<y9hC?oUHa^pH}63rN1@?EvQK3j{fWy%LX3yuDELE1KY2$M;3A+aj7^b&2*$qMp4l
zUvEW5?kh2)Ae(x4UB_2l$py*{`bF_fE5++Z?KOI{7813F(+b#JHm_%cX(EH;0;=a6
zluFX`bs{aV`zhzIw}16mq%@7Vk8R%`6waMvsks|TF;QqWDBgqWO2}D9@=<n6?ooze
z?1pln_i`vTlQ%M7?eJ|-Z3D=avLI#<?Y(K-yGdm|1L%lclv`b-rV+EPfTd>-z0YD5
zM}qn{DwC?k3(Z32PHQ`d2LcOrwo%P}Ewqu!RGS<{&1bU~PIm^|RJ8PqDU==X1iiT`
zycR>ac;}bJ&M%0S=dRUlf*@t?6ByC=j&*8flg!0pn{l-N(qwG@%`3u6V&xMbE3^4$
z0H1iu6={c{))L{nmBvD>Tlum&jcxWS4?7tvFX`Rao17gO(Bg^DKmptid987mZ*(KB
z{6x(aONBW@MpiV9TwwYVo2vBrzyA;I)D-T+8QZ#`1~6au$dX@u5T3<yNH=?`)`6xK
zy(`n__&TN;T<%*J!-{$0HVT%CgDovCI4yiZNO0O079{b|#(7_Fxm0n*$SR`O=dpd0
zxT*C5GrLJBy>kDB)UU11QEwjj+J6P@+>tSxW<+r${!C1pk{rJ_;vQ#|9(eoAU&;Qe
zlD}HXUoXyIABDfZlm9Eaj&QNBWuD-vN_3Dbe%BvdJyZ$W2}KPmmHQWzlfj;ZB{E(U
zi{zUdy&D@bsu;2z{FQHp2-P#_E{LJIsiuI|sL6TH|JM2v)ONjl)XFt7_Mv7chv$j^
z_CF_!JQ6-DzO~>Vve0N;v|RTad0w=0fv33g{@4cmtz9hR1q-~=R5I+-t9*3L1V8y6
zCMNX-I`>?Fr>!K5)P{pkqY#}fwXx`b;}GsYJzzXDIQDhW<9S7UTMj0AqIb3vus&aK
z%a~4FH8@-amNEN^hD`TfdIA~=nwvhJM@8ai&MnOeL_Ow-A}w*3tz4B{I<kKA-d&yF
zNhM{@Wy>C<KZSeq9x3e@<adkU{I(~7_67B;UOL;0`|VSamSVU|5S?guAp<`EqzBgI
z+=i?8{MbpJX^sa=GC*HJ=5f4oIqV7dqr2Od(`xptDe1;S@FLoz-2UEsqDzlS>WCf<
zGSafw4_)hlqc&KQnl0+f@@C(Vn=|hS?O~~xG+z=8NRoCywzh~lfLa1-dSXyh{!LaQ
zL`Hwb%AUF$xw%2;5YZ4~*b^^bj>1Xs<$O?yIG$dZD<+cOvX1*s^mt5cdz7`xg8ahL
z7AYnISgaOUGjOOITsAaZtnciRdD3ue*v?HE(`^z?j%a!YesLH?Sc>Dov1=-Iow)bv
zIJ46MN;SO^J?xy~{(@VNHQ%X!1cbVCHwh)hrP)1W&57He)Ji@9vTGMpt_<=|%Fa@9
zJI}cP(*Elt&fV=z*eM0uy~d6(Bg(7IYruI_RE*r@IEqZO*+bWv#t-)O&?%ah3SqlS
z2eZvxuhLA2_M6d4uG?9TLeBP`k^IB^91ml)SA5r6`jWs-6=N>ww~^o*25%zGAN-mb
z@S~_31t8GvlO|OLx67S!GA($aTT%SO8@iQ;Nh890zHXQuZM5t}T#<4TwZzKH5pVFe
zqjsFQ?Wua^>j(gLj^l?Kz*R$;9RR>*dn#3t{Nkp8em?;p4pvrHdg_5s<kCQRw1^lY
zBtAu&7wS(%{>nbp5+P^(NVVqoGJMzle|~8ffG9remGIDP$YX4>$+><exS8?vu)mr`
zy08CFu!6$InC?^;K@o{XU+z2ho6psFQsJ2nf)$7=0hZ7Xnnsl_6G7%h$DtPQ@Rx$~
z-}RRMtC=5Wy<C{}BX)hGdf`(JE*tNaTV%?etv`LIgGIlamvqGym$3ovPwliWBh5@R
zT*5)UY6lJUwYTq(@saX1S~7OE9<6NGD^|~naLU_8-uqDD=)#mv84@BpRp09gBqD|1
z+D61J6q~eJNQVYCYG*PO(xZ!wI{KZw{0y@VxWEo9l7jZN($9lnp!uA$zbUN#`Rs!~
z%+zh=AkKR!@M2hhKpS<rv-%e81tEP1@}y5OED%YrZO<|z;}uCK0j_WdDd=5o#!@3w
z_~og#kh(ctZ42Ev$pv!jBb$qj&rmR<Hd#pVzCJNVzF#sjZVrPrZX>hb7!?2wrHRM0
z9w4%o*Bvz(9vYVq>KB#Gc8ZAO;@aJY0s|oIH$k=ipJ+{fe^ge4#Ke}Ea#BY12a<jf
z{fwTqw*N<q3QZ4Gym{SjQvMx8(O#^}w$p<)*BDV3GpT-@|C<;$rBTFQwqs`h0BP+#
z)yGS}XZC&f|9K=u0ITV0Y}5r|xzBm!N;_3uQwono9Wx$I_3SYgLJUg{!2~zU(mtb(
z&LPBC?QjLeE^J5F#3IGJxK4wfXVBh-*yeWf@~V$t;OMnBZ_7<yhzzc1H1QSdcrsl6
z3Zx=i+Puuq55FVl0Z`kGk|b#(7EPZ%Yz2Flk2}rkk1t=TJDWuDUt$cY+X9?atA_dq
zx-W8n-@5s)<<3gGu3aNCNFI^F3XqcVYx-j)O!b-d4!Z)oS`2j=iZ>@Ra{AK!Pg!W`
z(bC`43a!#D>0~uo{_-wj>4JlGPg&ty|M<G{tg+Rv_*mbgSD`YDO%Z(OMQ+NZ&z(pe
zEBipgF8Yy@U7RH?@0WhGH3%FMzEItZZZsOE3{<RtM%Ot~qzek`9h4C2x)G4tm~PK*
zdDAS-(H%wJkAz6`n!mF##4}IDvjc1a7l8%v7S355O;|R-%&`0Ox4%{84>!&EA3C{Q
zWH=<vbk{Bvp_h1clO(dGTq4odYU~!-+MC!&tl^c07D$}2H5ihP&6HshdX_c|PGkQ@
zC1uoiCtD)E8yr|~+TwWOhmXY)B#xMc&`BQ`5d*TaNIy-QZ977sRkt6<D4o%&C6a0Z
z4!L(pO@bU)cQ4X+@bF>bdx0h5O4XEK2|}i-D#z7&>6ic7GyX?Y{pUFW0KYJ+08R5#
zMGau7LBHv$1?S4)>drU24))SzT3~ILsuf=D(QBCyLC}V>&$69<-<*rNxS=h|q;5*V
zLWq8@tD`mtb<eV-a#G@fy>5m?lU)Z(l9=5XIc-(mK!&W--7De{T$)ZvNi~0$=6x!)
z%(rgMu5n~V!g-0}?_v?6v?utasY)XQtxzawr<)?;TJt*Mx3SVcde7hYe1L^bLFtF`
zGjxr&D&g7QBVLpT!-L3K8Xi58%T1!0V72T^SKrLDBizEh^V$VzD4Dy9!OoR}vVB@K
z5CXZUUEUINex|D<pyN84F{jVDa~o+9-r`z4BIKvz%b3&YP@!jFR%{dtuF8Ng){VGU
z_u-SY)ZAa?Ch@$8iAe>aVW!UdGj~!>7J2~85z)Wi>vu2;p!zY2Ee8&w5cJ&TcS6VH
zy`a_J#iPD{Zeyp3>y>rPFz+XLxHxgrS5-sU@V+|D!H5SUj)d|gD43}9W?*uQn`Xv%
z9bkncumhH5=>0*Gpo?&|fC46@Z4#oVZlz|SEAyf%vg(Da&z7NYa^)qRY%m0@V%{Om
zu#xvp1~kl_CP8f*$>(t~=GQxXp2emlQLWc{#=@;mv}zflf(n0;6P{}z?Xa(70Y<T^
zrJ_-|2>*lCbUQ+KiU=wg!5BKXUU{a<TLy;C%Ldz38|vhhpmQ$xJhb}qfz*4hvFiV{
z_uWxVZQI^>ZK$Y6r5jLDP&y<uL86F&2uMpP(nKKCgbva@prC-^K!6Y;y@ZaGP=bJv
z0|+7X7D6%f(2*uZ{lb0k-FLq+-aYqwcieOD`Nm-Vv9rhC*=w!&TW#*W=KRekQYV#f
zS$!}s0@)Yu*5(i9oqpuBP{6fv+R0kGontPyB<WQJ2_9^{5UAD0+7D$P$+d~#`UzrM
zEIqjJZSKbN-?x|jCHMc$ZYdMf>BhB*sd>8dB|p(|mzwA3c#kmu3Jp`a7XO^Mv-y0>
z7lSG4EQx@UJlAgh_GT)UE@-Wu+Lo9i6rWzIB0_Mg#Ac4V?MLu<>9ACLZae!@@P&rt
zM4l=$DYCCd-VR?4H%n{G`YVFanYbIe9?!ChGm=>J3%B#xdek&btlGUEk=_P>Lty!4
zVbz8v^EV*v=P&<7l0V7`$C=M+t*`sYG&3M#-g_+`B!F51LLgiy0ejIFD#cNMn@i(+
zN71N!UNXSG_(@Qvtem`hKUc7q%s%zqk}mPwBNV)R)Xu$M!agDfsQS29GbP?*PumS2
za>LpbZCx>4kb(MAm>u2}|7F0UtcR42r#Qv?IVJj@fmZ~_XcBy%FamgGw#FTkP&&c+
z^nBlc;PSgLD*x$&|C1d#tk{H)ti*h6%wq-n(Za^$+vCXnpJCZ~C9~96j8TJnXc>K(
zjh%|{K;H7bc%2-dw^*H$hCa^9MnSGS;CqJBS=PL_KXSamoqx1fOVX3#UeDWALu`>B
zBt55o<C)aopq{Y-%QdkPK}0zIA$izS$dr>5>OsCw;&KR<Pgp}aRhjV<Uw&=mxW~lg
zn;bRgRknOdUWeZC(jJ_#AQo7?A(`GfY>5J$^JzzAO#-=GA>WxvUVCv)gxk7m=cJ{}
zGD!??c>oOZnBg9u@loswQ6YM;O3|QdnyfXUjGbOu(JX0BoaGInPtSB?JbOV8=NjbA
zDnT>8>fVvBvZdJEGSI+uq0=AM6IbJ2LGy|_V~djb8)K!;4=?x0tduTYAra12@Cdf;
zDrZp~nX`4L7x7vOT-(C2bhoSl7)xhB)#JkYnUQX{&<81vK3I_HawRGBqO}s#0%-c?
zCyjIem(Szx+0XxXvjNj}+c)P=pG{oT=)G-x$vrV(k&;bRGrsv)I$xkg&2r$ZWNv=5
zAc~K~m2Jc3kw+ET`AP#%qRJGH@7)}kC5lz#wZ83%51v%zDq2&U>UO4v1u#f>z*Cx)
zuA@ambI)>%qVoxAXyN%Xm~I)V`POQSfk!qHJ}x<}X@5gW#?uS-o=o;RmmFhC$d8AW
z^+Ra(zp!=wp7{I&d3zPLRe7&me{GQZsF$2IXYSh1-6|d1R~;qYt4p)%b&3$m;N)wr
z1UraM2#I9HO)F^upnA88I37C78oNzKKf{kYrpKvENg31FS|L>4mW9%6!pfg$Pe+ln
zB5S(lP6M~Ms-xj6>MI*x8^9{&h9r(Wd+07IVM$s}1sEbXJIt*B7n!qLHn{Wi(>khs
zvF`EG7yl{7K`0~1xp~`jk#~z(BC0iguk5X#h_no2v_a_I{h$8{pLq@nQeB>ryT^#L
zbzG%wx{3!ir|R5}ep6;UT1Ji%3KaF^n+1F5ssTM)^foH;`Df?pb*-%K*TVB9JDN{L
zw<CGu1yPc(`Gj?RuH4~WPx5H{P>mSri0N&K3=l2T#@9jHLbbJ`Q~~_${ifJlw1^9O
zzspaw5g|wWzRtO}-7q~+r!sgt{Pb%abpHFUxLbNuc4*9WJ~zKWGILPM&}w*R<ryuA
zQ4p9%Mp~|^YtHW*Kr-T|B{=6)mCKXv;hKPfz(GV}wO>Wcu{OWtviat$nR)e+hOW6u
zm)5PtR&@F9!m)OjxfX%hSRwikrW9s=B}Odi*IIQ*JN8#g<06(Z9%ocNiXW5)ULaxE
zje6jB+Y)^&vGSVr&(ONy{@5jhM4`ljt1SS@yc7YO2WC>wTVxK%L6iqR{%0%w?mtKL
z#*DKfR_jmIf~!0dnnDf6vVXB~Z8<*^Lgh9o=CaLh`N71XQ<Ahi`TgrEaknA<6QfG<
z=k1M==Gvp?6n7_jm;R$RYDjlvi%h!pbB*8!H3AvaHSgRrFh~YL?8UUKW^3I^B&YIA
zxh5pzi<zwFUqm`~)p;95CSH0CwV9Q3I2U1+F--Oe4^T0$gwJG;XIBr<T~Z?T1x7&d
z#52%AZTHat&6Q`(73<h9HsiqdXyMZiCY!8P1i9n)yaPVfjwNHEBgKBV2StTOn)QPP
z%W~Y1{^5w%4!=sYMC$fZ>8n+JFdkr?iuMU!B6!U#xzwCZ-&eh-08%=m8TXitGIEo7
zSpaLCTM$>Iz&m0Nt?aVmW%J4F#(4)QeY%*Aj=F!#r7u7>MAn)sVREJE#Kp23&)d<V
zApVV>)#j^eac2Bx$055y^H0Ks+Kfv+<!0tb0~528Jw7np-Y{QoDj%aT_+MEp8D>4#
zkO0OP-}gk`*I>E7pz-fMIUmqHn!6n0(qBIC?HDz=dO^#^w4MK9T$ZnH^7^(GyChBQ
zRt6?+(NLWc1~lTz79%DYnlBl^hxBiCQD!wdwy+cu6VkNq635;)vXX^4?`V9#hD$%A
zU4AD=Cfu)%Rc?|+e*%Tu^$ZQ_lp4HswfUC)Qgt#ZE-eYn`835?jZefO?F@7=BD(Ya
zTjre7nWf>`{cj&4SCm*2t0_-BKpr6vt`>xF+BO^;$%-r5X4y_~NzWmk3b@>yNh&e7
zpJo`^+@kTN!&$F6)=Ksi0=qS%pY>|-_Kw75CHlk*x}=k;uV}k9VQ)?=@QZP7xT2ii
z8{L`{jdf|^P_-Vh)>mSsm#D}%>9#YBh0l-wr8PQk#&0KN^(_Zk8Ec$cRgq~;CB!9U
zWH;WC#0Z(&^$DG^&Rx@(+vsZ0W!Fg(V_qY&2&iZQIbB=0TrFSdZZPs;7nsU5Dydm(
z7=9b19%*FbUtG1YEVrRJCy`RgqfatYWBAKtC0K^AF!m*C**Sym-s7e-9{Us1j&8uS
zr1e0=@SXQw`z$X$GA^Fj8dh7?&%T>Kg5E?7pTEzz^zTRp)VQc$gw}D#;m>RJ)WZYi
zEpvTOrp#YOhiL&=QFwiE`u@(4QYdFrB=A~iGFkqX%MT_S?PK)rNtRr9$&WxCMx{^g
zp@G(zn3*;VAh$Fl)G4=gGW=A`j1v0=C7`Lx<Vzj3V>*mrjmpRijHrzVF)?4rx(q+Y
zn^a!IzpX-?=rt50ggq&jeYB}+y7QbEw&%1KBZ3JpO7{3^BOo{rSxN)Pl*Wr(!)O@t
zc}fjk#R$nsoO|_GVEMq(>|k%^r#LnG19*GLkI^jkmZ4|w_bz57T@BS%PFplm#)xv<
zazS@(6Er30zE}FrS=YK$(9oFc&G5Z2M03%4NlSKfI<0*4L~bsfpa^hn%tqV~PX`>d
zV;3e&D^!u0AZLYS+TYd_ex>bh@r+lHz&3Y_=V1|>B01po^Xxgp)KpKtD#*6<*0l+2
zntAZgec?3yk-?#`aKtO$(<cufJA~x0mmCh9!%5+Ao;+MS4wuD4P2pdxJb|@to<guo
zfoQo|H2izI`{(V?81^11b|)Q`6^3uDrNTAMM&{Kyziwja1=`AnB!l?)fntWhA_^w&
z2^FZ(u6SDkcw?r;dO?4{dh}>dK?ziw^@Fdyt9nkU%_FX!^pGoo%&AC#Wy^S2>N#wj
zkSWG(suj(Ms*-wOXu>#DyL}F0J|H`GY2>_mKuy4DN&GKrf{#}g|683s5-ToTX_Dh<
z8!6+|S~0&bY`?XVIxL$3HF^|<%fi(R>REWE&+<VSQP6Yd1t<FHtOo`)7@?wgn`8Sf
z7?sFsOQVCm`6u<!y$sFXo8!?)r+}ueo;|F86Mfr2doHL9_0@bEM{w5`=UgLZZQ1vN
z`iw|%p>mo?)5JHw&>}CwHjYAmVPx_D?i~8x)&CcrF;gecXI}E8zKZ|B<k1lyF8sD<
z0sq##-e%ALC%+qQi@m>;W_hLB$lsRmWB3Woy9*fN)uTC-cP_z{czO4){#tNQPlwu8
zjxka82{}`f4L-U{XkhAMrw-@@dW4SCZiF}D9=ns9={dU=0ZHdLCl{qSEuWLCPxp&+
zK5aQ2ARm>%V!qM(vwiBnNaBCEn3GdwTRv^WmeHW1tXXWWbjS_*j9$CXtei#OL^HRB
zm_eJDFX{W!ntqkA`m>6JUg!LRBqZ<mi&ITV-melZV=@yfSi-2+#I|C>ly-XcLS3oD
zIh>_KitMSiAyG^*&{WI|P$q`Bp!U*(V%{WAMbjnelXH3|H)xB|lVRFFxj$u8T|EDj
zU(czy+nqtXI}Cobiv<)b<?;~m5x$TeX%I+y5IlQ_rXpawsiJ&N|I_;XQ^t$etHyZo
zpsekX`x1=j|G-`Ge3(mR*JzG#p`rXE-#vH-GJD)5EkEwq?vr6|iexXe2hz$2*Y+5f
zi^Bjl>AJ+oOZ+XY7c7ySlP@Hy#=egUSHykL>R3uAKs7%|-8J15e++leBlyEc+{R4F
z-tr<ized57rR&i_q@SLMdz8UE1hle>QNTMA;UX>{Z`(1zVZS)l=gNnpE&S>&pliF|
zB-T<{*rVTN_qY^d>X4RQYD0Q8FwL?g+!qnqTB4Dag+_tOcVwvei3%rzhL|dOXN&-o
zn>@z(nR}g)+aEg#NwzIasAyMcp-j9?aMj^ZvGNrK?g%e<c3OVu<q9KephbSL$chE@
zSk5F?O<ZrKv~bs5>8xpjpfklmR0EfdUBXI@{Q_J*5R?1U7mcm%uP<q*aFj2T&dm`Z
z1R5fJ)2t}4Q#h+Av`K7QE6YdVfjeV4cw1K{LdWwu;M+n)tck)>9l1p2o6x?eXFxCh
z*k5>#pm8F$usH7Pzupx<lr}jvKYlm{PSi0l%ePT)KJDY%klrVw8HS}xhMLN?qF%w2
zF<IyY0kz2L!OKbT?}5M>HEmt+Q^>;uyd0ui{U=48#=VA@uu6k0oi9yls3){5se%jQ
z(nDI+3xThVU9(E33%+T{r{9H-l))ZF%k;)d^Ko@U4GhRFVzb7Zc+cB;OWLqAMhE%*
zfz#1<vaSR&Cq-ZsujW4d4P!I?R<h?^4S45y$_XIMtk`q?dN8sf)~$(!-q$ti4nx#I
zbZkcOj*b2ajtmA(yPJJ_Jv3eDRY{Kt(YiwY3B{Tvv(fkh6&^N{j=1?i8<)J9$MBvx
z*1izMoC6d~rr4^9J*c9)EybLYGj*KO?V=@sc%bW2P~Fxl3t^YnatHRnycy}-N=dTE
z3kbSt^6pAs_3t_gq)(A0TFW_UprLt?fDglGwG!5tF6g4hnkXmCh`Kxg@u+Ih(l|E%
zruLNEL<H05CB*K{*sZ}-U(w2fH{zT?#A|!l{hoJtR<~4?e0JwArA=ib-G<PHW4bi9
zfJSKL>4ldE;T;1!RUYh=!N?fEr;wqK48PTXASU3ythtNp1tTouYUXJ0yt+w|XmigH
zCwz~$V75nj?AH#0S*Eq4Sr1z83!bAynaxJin=DC~38#9-$DisRE3=6Z>LK#UzZT4X
z#1g^~l<8zSeA0!`gjclh9c{GZeO_ZfR%jpi6xZ&fENwia%rFD2*3rb}rxs@>v9u{P
zg>EBKjBVNt&b$lHSNYyz;Zio5D^$3NzbJ5m;Lk2MY$wh>%8}KWSVYE|lzoypEtoi+
zTIMJjrz_x5V6D?PFNUF+fo)(mJ$wEBtws$pr|@Y1XqTdFl5wdBth2-dbW3aQUk2nq
zQqa-XO_G^Z1YBM_X_I6Q`fl3u?N)ET$ljJqM|F1PX$5Fru%^x<ln}FJDBP@V!7U%F
z4KFoXg7jDBZ%!12lWkIq2?jlSk699`zN|I%nTyKziwjRZUq5l5f<T$iPj=S=i!e*V
z@opbTXz;j#5ndh{??3PGj^>q-qOTQS5dAnmt#`lD7~BiT%CC5nz4KQMR}Ta+Ax5vf
zA6^O^FR~sJb1ZF9(3EkE%fs2s4-@>YQfnM5a8HUbR)Qu|$_gxk3(wjxVhB$%-u;G^
ze0qa_$aQEtT0kN0xayIddk49_07EQ{aQ@@iiYO7D`lDA5w3H*Y718^w^r-o5{%<A~
zAq<k&e@_ZZ_#AA_*r;TYYyikL04@*@zQG$=F(`o;OhAlds2q(|j>a&Cj8X>3bA|xG
zIycYG7hMLZ;5GPcBI|*@soZ2;!$PW8E)LghmX5kUvlj=Uj@?T~n#z4b$!fTJaYQC*
zBar^oNG=1kp!vdTFrYyr^5aXL0}$qiuN6J1tsv{S5~!)WVVI3#=*IVj<FmJBBhy@w
zqdw&WmSVRrO>YcY0u$?~t{F+X@LFSp$7Y4&Qg%^kw@>AuiLg-a%@a}YuIw259<5e@
zLAvVr@Dz8%<&c))cnz;gkAx>)+j*XHmSS3pUY5mJ8S6q0GIVI;S@olOO`?{A5WjX&
zDw)fcy0xG2kmGV{aT=kk+f3B694!oQj!CW?cTLSj87o^T8w;Ecf7B*Kcs`@w*VF$H
zBNm{j`37~P69xFQlBWC<_>a6!F{ezJHTzisM7|X4?7iso@&3lJt8B7Bc`e{B-E!xB
zKa*mEW|!F|ki;tm=f21o<s&)Q!eng|7tSVTLlvF!0p6tY*P<YzScizeqqln#PUmTU
zzh61GcdC#1I(Rw@>(uJao%2ONFN*nQkDdj?ptBX?=^K$_5rQLU<#xyp8(*|r&9CmA
zzRa#yKC%GUzkd)xmgyvgVUOtF`n-I4vEH$`cD#5{{)}i0*&#E<!lPz%8LyWEuc(2a
z6}jZ#=(w<8+nSl4i9-Vev>s60Y?J_>XkkAM3kTPqHwj`{=1sg=`J|0o@JYT)WlNsI
z&jj5`J}+6gQ)hX(bahd@(6CcL@T6LdoqYiTZe7*Z$n3btB{paV_#h-?+r;^3=wSH$
zj{~(ZKIuz()wLmyA57PcmTT&3^i3Bj3L?V@$Pm26<I*awxYF1zXW@pWyed*FTHzG*
zB1~h;7zBHS>Dl-)qsE|lc^%l2r1Y`3l6z%>La@+ha?g?D_L?-?H<8tte|paI<k80?
zj`a~RdVuDUj?~Zg)VJPiPNU+vFK+h%&MLSTeHX=MZAzS!5BV8~kRER;^Qg;an$>ZU
zT57KOT7f|KeFV&(JCRg17{pHQ*-bclSkxhahuz|ETpi99|Nl?7cXi`-9*>*{nc0>L
z%GM*w`zBfCl#q5^3^MX?x%UxiVOn4HHZ#h-a$RUqwH#%GkHW76idfR`#-2iSogG+0
zd-^y7Mk@*fdw4uaJyWe%V~ecBpX1##4N_p+skpjsqh%<=1j6j4#o`RL*rS~7ivTSW
zv(Ym{jO097f3QIR?G&;M-WOhar!OOk`2vIBb(*RJSQ{7eOdRU88Xtrwei&4DCbG0-
zZL+Pq#x2Sy8kg!eJVl<`gbWPd8C8oZ6KxZWr8skej`Bp{vlpHcMivdJ@PNc^n<|50
zXLi3r9+f(}J%Q=~>mFg4`Jb+KbI5U=bC~MLD#*x<6G~4^hqQvS8i34cAHg3EDtSa?
z|E}D~8Yit79Nb??cCRv*%8vH}2}zwa<cH3SMf5eWJ|`y$QQ{q!rO>8tc_51}NhTKA
zm=Jz5;mKfT$Avs=8f#AQgrDkkA-qOL$E3h-h{~R2w`JQhB$&6}@yhVYrlGdOtW_vy
zCt=VglH}4#nH1EaF&JIhNj9q41GF9@yEIAy!wS3~xM3C5c~Z+WGc~78VZSR9FJH$j
z&%R_?Yg*1r(*TQ4wWvddjSGF)H7`wT&b85}I+HZOA`(bscDAP5Of%2q(=mfCcWYoq
z8#SNmWfSZ#(ZqMBzUYaiPpF3^EmT59NZ89RHxzb(p_emSYVxI#RW~u?#a+*@yImb*
zZfzRJ&7IHZTLwbC?0SNjtM7I)1%k^#WvZyJDy`q2A6?r_OG=bVX%y35|9B;;kY%wu
z^ZNz2pk$`~06SaQ+ncKqsQuNGlMBiUQucRWnx@RYxO?RZ7bQb|pKt01)14!Sr5qx5
z*vAgX%i#=jI9ndBFNbU8p?2}d(Uh*}dW{TF3xKuBdsT`*m42eTMo@}|i+@?_A(-Ax
z7Pl^_Zw<*J?yk*kNPK_qE`VFJqlIO-3260ceP7~k1w%O7!@E_AxjixN0>IU|A57)w
zIL5B%Hb$T!N}1C#9GWL*#o)sMxJHZ}lSNp?tu*?n@Q*C*D!B~62Vo9&H(S#}G&Yti
z1*{X6_1;G?TbqGGi^-qxJ@BFSu%U^X>4IGBv`VN%M+UASnxphjxzy!pH&?)@Xx(;<
z%%)15j9PZ3GYr>^({yxnM^yDuah8A!Jy1u{-I!6aRuGFsoQ%rgV>i`%?GRL2)G2RD
z49rTLDL>GreFrTMJJamcaXWMe3wmr>oJ$yA7h}-OjF%n%LO0|EIV>ASkbBbeCf$p3
zF-VY*u&{$|-(w7t5hRxp|LHY2^3N|39)S;wPrb%(n+bpY#Svx{9vHRM_=AZD?L0tj
zVh4A3`u&X3779@H%#PDPrk7tpkVNE18!JPP)mujBA+}D}zymp0-uPs^lq&i2i_`yT
zity(<zsuoh--fNJbOeI%=4pT7$-c$4OnG$t%{|2^1^SFN!N#dc`K7MXTNN>mGb&nA
z!&tZngIL}j51j;4JTsf+72^1zHz+YyXRlX-?_}A@cKWwbg~xeR&|-+j>~R^a20lJK
zc@XUuI&qVF!%u3Y+P|cm;fgwa))K2M6yV7wij2n!+7<{pP8q{@VBf?B(JM4pm){hW
zYxYaY*Bg^@h@Ft|z(F@8cxL@<;{(4>pL_x%cw?F{VxVc=F2qHOb5=-J=pZKlrv5PA
z>hyE|CUwkB6Zh!{iB<0BJZpv<LqGp!F@Knwr+~EEM3BMP+~pO4rCx1`!uL~on$|O;
z?)?w#N^&dg&10zx_ETdk(R(+*1NrUSuV+Z49d;s6M6wp<u&WKXQh%0KH(F>}x;*N&
zlUbwJ^$kcZdNnP72bH<29%lzcYLWVaNX;t8uV~9;)7<hSSZ<CwZ!g5|QkSjk46J~(
zz;+JdxS2_LPfyPWL&V{G2gv+%vMpkPbp^Zb)jb_j`Pup)-1+aBjxH5&?9D*yDDJ98
zA56b&F?g>XJ%2EXTP())%hZsOJGKQV*YYt@4tc>CN6|qAn*c>c_;Jr0bf{FE9zWzE
zmO<oj{iDpA&-uUp3CwRQUoPzC-RaoY&n&t*Y*WPQ(e(+<XYk&>PfvuBy}*}<{r2`a
zuSZsRca8zclm`GeC>T{nAa6}(h06sS-1{v-uV^HEnX%}&?YkFSW)stoajUQ^GySq`
zLqabEhz_>|y&ucLQJvXM$VCLmu5ert-8nZWtx5;lnxtW@4u;>^(dXAha)t<g*FAp(
z&`K27%a`S<G?AW^Tr`A$0D4O$%5Tkvrz`JF`=kI`<!#>yXIp`baQvpYCO|E)1I<;^
zks`$DW^R{<#pEn`7tFE!{9CpC!`wTm>qNDEvRN-moH|<yC*JHN@1jzcd0UR@OD1Wd
z-IC2S!%e;SH<MO?QxHb?mKE;S+}fO9kkO=9ZP+ix6Eu>xrXBExCx~}$RvFJ}pZZqv
z#6$cMzeYP|MMA~7-Irc~-`aurpbR!oKzV@zrnaO;n+GHrZPIFIUWK(tl2BecO6-Y1
zpLR}8^^8V`O*_YXRM(%;`RP<!c<8MvHfJoeWwVG&2wc}gg2?dT;gk(Zy`aLO%p$p+
z7RK@{;>t`TK(Ynk%lVAu#f#q=*CXl9*B0~;-dgMvCo?bQ=UJ6m2k3MKJghXa3Ky}X
zBYbgj(w9Zd9W#}YJU67JHwcQ)q7B(OH@Y8HbX&6XFrsMuRQuZ-kLMfby%Oll<@U?2
zJ#UR#tAljPsX8JYfWjo8fkQ>_=$G%-@6D`Lw9Vkv)OkxJgbuQ4>Q&<9R0b#dU(Cye
zspp~uJdVV#@Gd_#$XFe#g`FiWA;vMD+E(5Xv28VxoJ43qr-DOX6~yC)&HfK2dDBZm
zE56Qvo%g!aT3Xn1AVI$H6j9qD`bq+2o!#?v$?py6KcmGIG%RcWG}3ZmDUd{cPCciZ
zf>~E-6O^XlKqC^lX`g(pV&CVW`6iP+F9S*0c_{Rk=^Ex;BZ*Kg=){i2Tg3>G;bP<9
zy5K<wH+A*x=k<z?i}KfyYCZFIpyu+~C&Jd>Tu07sF`^I1%!^JS0k%r^czsKm;*rpI
z6rq{UZOND%VIlrrViZkbsqtIFys$I$n&{{MkShWq<9jJ!B<0k%Z91=E#y&p=k$4>A
z?q!ZAiV2o2Z`TUuRXQ}~b*hM6+4B~z^?Dc$$J5v)8osLLp()RT&n1c+W&39!^qYCu
zt9*Lbg<aOBG%tCd^vu{wDl^#}7qWILW2Lw%f$0w2ENN!o(iH)bAe0P+?FSRonf-~~
zYD;$GwABYvI6wl;qWh}<#cvjIWL2yp-)w8ptz=xZoC7O56I|NwU2bMaUzVvUETU-x
zW9q1GsrhL67(y^1#$(u_O+HuOe%^kj9VKt-Hi-n>4;yj3%|Eb6;&m7qJWb+19{vp|
zMA~{z`LH|eh%m3w#6)_s4f52DAv`bMa_fhI<b5MO^1c~g7b6E-?{IOiKSitAwzARg
ziL6Dt7hzzFMBHvvth<a-r|~1YzBSnxbla%N$5v{Z)C4dSg%<5S{m`pxCTe*4c?t1U
zd0b!V&4ymfNF7Utm*eZe18aysb~^s@IoQp+(zS_w+p^D}g!llb+Pd_s=YDXkgO_xm
zJv#?>K$ctN3}s+p%P^`8?`d)024LJ1^xl3m*qT5=Ef41RR{*kH&#HP!5*gH)`?t#t
zM$=3ny$W6`q>F3W1?1b|I4*BZM%Zj)q{?0V%*@h5^15|f70mCO>IJPWfPk_VtfYtB
z+#N5<F3cLPVJ7GGU)tq&cbi8!aWW5f^kyF>iQNWeb_8?9WVd0?Nob;hrq5N5wYp!8
z;Is-Bkqo`%+RXXBg~gS~)KU|Bq0wF1rD3euxkYIDQJz`W!sza|lr1X8ZmI>?Vq0mb
z9p~0Qd>abpbPWyO&Z<<1m%-Ych0yzD??ixv#@BL^pl0s4%9no~P4o9?MERTcdxisr
z3y3pXGUUp3obEem4M_19r;10Mt{4G|Lq;%XUa?O;$n+@+d|6Ddt>TR|dXeOlBxKFL
zr*&3yjvu1E#ckh!Z~oOQ2&ce2tQraz)_59<FzWt*QnAddZPGf<4|5^+m*K|B#{8;3
z1>d5@xpz=Hrv2MDV^=kiTCAj$azV*ZWPGyek%7;DM?~i03GIe@D^oqUNK)WLc?Yx%
z0^j;nxFhb8==18F@qkm(%b(h}l4Q7|PIKp(j-mx&T3FWGckfWwc}?TqGDw>$LMdKY
zG5S?d-Gpqb<%DRdadn8oTz|-y-Y!QcsC10GS0u$Th<iZzK~bbjdk5i4<X*CEqkpao
z8KZBA^CY~n&B{cv1q{@Yk3H{IKiviGf$BgZU4!}%T|-^S<IIALd=xM-FD5V1TF*f+
z&wf_<7@h6nO(j6?7g)owR{g8t%rr(n31j;D_eA|WNc`TnZ)iU=6_lO+HAqqe57)ov
zZs1J1=(w6%KR^22oJ&E~E%Vd$VD1m%iKw8)_y+m9aEMV~2g%^nU^`8hk8`48dU$TN
z(Nrm@j_pItXfB)=CqBQmE#`H0Vxg%{$RzH>!xZ~mRK<r^9&s-ve=yBa9<K|Tkqq;5
zv-bop77f2qb|ziQ(nJH~?bX~eqdk5VC912dJ7;F4re@>P)6#J$bu~400W)cS9)77R
z>)xJT{np@9QX3mzo(4lO7gK{xb6u3hm@OF&E+4^vHW@yE>zwhJDrsxFK7g5g*<<uX
zvI+jY*1DL4!9%o+h+Ww^-S)oNO~_W{QnWkk>mX(ur^#oHrTJ$(CxY)N#>D5<b$Z~c
zjSy%Ps0*y**RUAZdNR&p#kj_PAd$Q{cc*Q5I6|%Jksxd0cARKJi;O3Ky_`J^l!nC$
zO+9ReveXQqwcB;a9IVMz(CIrFLg^|5rh@D=dzn0Yp5-!;`8v$}iG&#iPfZU)miioH
zkmOX&=vnLH?bx{Zf>wCNU@wI9gQ=d+^UkFMr-Tzs>#M(G4>(B~7O8yKU?xPdEFW3%
z7W}?}4M-2pn>JGzC!XJe#Jb_MF;Qv&iSG|S1nj#oNIYaNb5E_TX^u;OZ~A%$QH@jb
zAk0Q4CVsOUiA&$ix(!Bl0`w-DC#_1%3lnTd*57I-r0;WtK5naY^n~nR3$0jkOmBua
z-m?cao(X<D_*|7<``}%O_*z!g$gC|2-k<ad=BTJx51bLQu*DT&V@shar}JZL!M$%}
zPd{<^<ko+@t&Zkfkdayiqb@(q$5Tr1FFXD|KFox%A515}m*3rgRy8Jjy@KH_)o1cD
zA$8x~`i_uCuMvM|qeV$?|KMyS7C2~vbxjM)c*Rx^RwsL@^GF4pF*Q<Q=X|AF`Q}q%
zN5w8S-f{@G=;RYaL4s^`?L+6_o?L-@8Up5WEw2>8h4%Pxx^3NncF2`y*(kHDYSkFi
zp{rZ{)S+25`2^vB>c>oX%qvQkc5ClZ1oCG_s+A&J6K0mqXICtcLv2`!qO?77(3$w1
z(!FPu#Fvt4K!jbXk*z+7$c5Ltq0!7)!W`Lt+fmZyGTtmidPv;o*`5-J3p!v*bSNqv
z&B-Pr<*dYtsBY0(##$f-@g&RBBmri|QsvTej`0>rzxVgC`ZxXKq{Gr|*okl37Wyi}
zh`oCw0(m-vRt%;T8S`G2Zt?5tjZQ|&LJFa&qjsVc-hMHJ;uR~bN4u0;MkD81@Ew=n
zJC<jTFF8in|5{_1SBO5_Jj=s6YyeG(OeiWxJGF`%Z^|PfLut(-cEET{zxF5Sx{U^X
z4I@-(=(MIRJ2Xw~rLhi89WSUbY*nV2;ho@IQ4lwn?U-t&$r>&X<2N|J1Cm-+X*SeY
ztQszN6j5?)?T^QdimP2#%}~iz6p&=l3inzOIVIxR)k$0?B2({b?b?h%Vnd}!D5>uv
z+>51*p>gb;FfXABrqKlp2mD0yl4GdwG+f9IQ1BK#v$r*D2RKa(`*6PYuUg>9THYL8
znBhNVsMkBt(-D@Hq-Dql)w(n1D`(Y1M#(czx13jFk(^UGC+FcpyfS~X%P*`o=mxO9
zi#>&KWcdpJ{x5AmnC?9ZoI6?NkN%=8{OySSUMer`i$yC-@y2o7@{7a&AL4V^9}dUM
z;r#H&nGeAOqc&v}8un~@uM?iA0N_hHO9+k+<tkAQyU&?T)DkkdBpG(zkLB6iWGL?z
z)1Jg%tr&m%^dzlpEWtiIhbmd>fJnB6tCU%>xLDbB`aEjD7A7=+#xUokxPr%BGMhP9
zw==?q2p257=<k2M@GZGYWEI`#P~A^S3G?c)EXYL)#<AVDQUwMoC+W(7k&+!LyegH`
z?Y7zs;<hHBfBQ2m5GL1-ykpgpeG)u=TXz2v=$xdL#rpt7T=N}Yx$j3ihpF)r#Bs*f
zpa7sUcOc;@<6ZuiMA-THIgcYdflgsZ5#7S#VP;*s%SW--K2pmck!XV+;F=dRpIl^?
zE|z}v96ew3`F1JbY|Vp$Iv#K3g{pIh4;=z=*dq>y#^D@sI9VR98i&j0p-OP58~vt|
zc#@nt)w!xzsoj{KV!89o`jamYv*Tuz8;9|guN$se)eAf}QWoE1-}&ApM}0VNPI`B=
zO3ubU21b?hU7F7myr9g=3A<D8*>Q2LnX%RH@s8?K+m^%Nz-|joKlQ%p2a_L*%Sl%{
zxLqb7b!?#JflO{C&lQYX&!pB}&2kMbH&IB{=Q*pF<M8QsczAAjprmDQi!NXY*N*AN
z09@6OlFxWV{<bcEc;^s-Lj(>HI7HwOfkOli5jaHP5P?Gk4iPv+;1GdB1pd$noc}@j
EKbLI-$^ZZW

literal 0
HcmV?d00001


From b59abcac0d6ff7d5c97b743eddfb48972f4ab0ac Mon Sep 17 00:00:00 2001
From: "Sterling G. Baird" <sterling.baird@utoronto.ca>
Date: Wed, 1 May 2024 15:20:31 -0400
Subject: [PATCH 03/12] Create batch-bo-tutorial.ipynb

---
 docs/tutorials/batch-bo-tutorial.ipynb | 1544 ++++++++++++++++++++++++
 1 file changed, 1544 insertions(+)
 create mode 100644 docs/tutorials/batch-bo-tutorial.ipynb

diff --git a/docs/tutorials/batch-bo-tutorial.ipynb b/docs/tutorials/batch-bo-tutorial.ipynb
new file mode 100644
index 00000000..bb6f3d12
--- /dev/null
+++ b/docs/tutorials/batch-bo-tutorial.ipynb
@@ -0,0 +1,1544 @@
+{
+ "cells": [
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "### Noisy Batch Optimization\n",
+    "\n",
+    "Many high performing materials suffer from poor corrosion resistance. Anti-corrosion coatings are a common solution that can be adapted to protect a given material in a number of harsh environments. You have the task of designing an anti-corrosion coating for a new material that needs to withstand a relatively high temperature corrosive environment. You decided to simulate the corrosive environment in the lab and test a number of different coatings to see which one performs the best. Based on the lab space, you see that you can specify up to nine coatings in a single test, allowing three replicates of three coating designs to be tested at once.\n",
+    "\n",
+    "You believe Bayesian optimization can help you in this task and decide to put together an optimization script using Honegumi to help solve this problem.\n",
+    "\n",
+    "Looking through the literature you identify the following tunable parameters for this problem:\n",
+    "\n",
+    "| | **Parameter Name** | **Bounds**  |\n",
+    "|------|--------------------|-------------|\n",
+    "| x1   | Resin Fraction              | [0, 1]          |\n",
+    "| x2   | Inhibitor Fraction          | [0, 1]          |\n",
+    "| x3   | Insulator Fraction          | [0, 1]          |\n",
+    "| x4   | Stabilizer Fraction         | [0, 0.1]        |\n",
+    "| x5   | Coating Thickness           | [0.1, 10]       |\n",
+    "\n",
+    "Additionally, you identify several constraints on the relative fractions of the different components based on prior work in the literature. Notably, you observe that the best materials keep `Resin Fraction` > `Inhibitor Fraction` > `Insulator Fraction` > `Stabilizer Fraction`.\n",
+    "\n",
+    "A dummy objective function that emulates the results of the corrosion experiment has been constructed in the code cell below. To simulate sample variability, random noise is added to the output of the function on call. Although we can easily find optimal values using the equations, we will pretend that the objective function is unknown and use a Bayesian optimization approach to find the optimal set of input parameters instead."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 317,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import numpy as np\n",
+    "\n",
+    "def simulate_corrosion(x1, x2, x3, x4, x5):\n",
+    "    \"\"\"\n",
+    "    Calculate the corrosion damage based on the input parameters.\n",
+    "\n",
+    "    Args:\n",
+    "        x1 (float): the fraction of resin used in the coating formulation\n",
+    "        x2 (float): the fraction of inhibitor used in the coating formulation\n",
+    "        x3 (float): the fraction of insulator used in the coating formulation\n",
+    "        x4 (float): the fraction of stabilizer used in the coating formulation\n",
+    "        x5 (float): the coating thickness\n",
+    "\n",
+    "    Returns:\n",
+    "        dict: the measured corrosion damage and  uncertainty\n",
+    "    \"\"\"\n",
+    "    score = float(\n",
+    "        1/(np.exp(-40*(x1-0.42)**2) + np.exp(-24*(x1-0.75)**2) + 0.1) +\n",
+    "        1/(np.exp(-30*(x2-0.22)**2) + np.exp(-1000*(x2-0.22)**2) + np.exp(-800*(x2-0.42)**2) + 0.15) +\n",
+    "        1/(np.exp(-25*(x3-0.27)**2) + np.exp(-1000*(x3-0.27)**2) + np.exp(-150*(x3-0.37)**2) +1) +\n",
+    "        1/(np.exp(-400*(x4-0.08)**2) +1) + 0.5*x4 +\n",
+    "        1/(np.exp(-(x5-8)**2) + 0.05*x5 + 0.3)\n",
+    "    ) - 3.18\n",
+    "    \n",
+    "    return (abs(score), np.random.uniform(0.01, 0.1))"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Applying Honegumi\n",
+    "\n",
+    "We will now use the [Honegumi website](https://honegumi.readthedocs.io/en/latest/) to generate a script that will help us optimize the coating parameters. From the description, we observe that our problem is a **single objective** optimization problem with a **constraint on the fractional sum of coating components** and an **ordering constraint** on the relative fractions of each component. As there is room for several samples to be tested in parallel, **batch optimization** could make the approach more efficient. Lastly, it is expected that the reults will be noisy, so you decide to use a **Fully Bayesian** model to make the optimization process more robust.\n",
+    "\n",
+    "![Honegumi Selections for The Problem Statement](Selection-1.jpg)\n",
+    "\n",
+    "The Honegumi generated optimization script will provide a framework for our optimization campaign that we can modify to suit our specific problem needs. In the code sections below, we will make several modifications to this generated script to make it compatible with our problem."
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Modifying the Code for Our Problem\n",
+    "\n",
+    "We can modify this code to suit our problem with a few simple modifications. Wherever a modification has been made to the code, a comment starting with `# CHANGE:` has been added along with a brief description of the change."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 337,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "[INFO 05-01 08:47:30] ax.service.ax_client: Starting optimization with verbose logging. To disable logging, set the `verbose_logging` argument to `False`. Note that float values in the logs are rounded to 6 decimal points.\n",
+      "[WARNING 05-01 08:47:30] ax.service.ax_client: Random seed set to 42. Note that this setting only affects the Sobol quasi-random generator and BoTorch-powered Bayesian optimization models. For the latter models, setting random seed to the same number for two optimizations will make the generated trials similar, but not exactly the same, and over time the trials will diverge more.\n",
+      "[INFO 05-01 08:47:30] ax.service.utils.instantiation: Inferred value type of ParameterType.FLOAT for parameter x1. If that is not the expected value type, you can explicitly specify 'value_type' ('int', 'float', 'bool' or 'str') in parameter dict.\n",
+      "[INFO 05-01 08:47:30] ax.service.utils.instantiation: Inferred value type of ParameterType.FLOAT for parameter x2. If that is not the expected value type, you can explicitly specify 'value_type' ('int', 'float', 'bool' or 'str') in parameter dict.\n",
+      "[INFO 05-01 08:47:30] ax.service.utils.instantiation: Inferred value type of ParameterType.FLOAT for parameter x3. If that is not the expected value type, you can explicitly specify 'value_type' ('int', 'float', 'bool' or 'str') in parameter dict.\n",
+      "[INFO 05-01 08:47:30] ax.service.utils.instantiation: Inferred value type of ParameterType.FLOAT for parameter x5. If that is not the expected value type, you can explicitly specify 'value_type' ('int', 'float', 'bool' or 'str') in parameter dict.\n",
+      "[INFO 05-01 08:47:30] ax.service.utils.instantiation: Created search space: SearchSpace(parameters=[RangeParameter(name='x1', parameter_type=FLOAT, range=[0.0, 1.0]), RangeParameter(name='x2', parameter_type=FLOAT, range=[0.0, 1.0]), RangeParameter(name='x3', parameter_type=FLOAT, range=[0.0, 1.0]), RangeParameter(name='x5', parameter_type=FLOAT, range=[0.1, 10.0])], parameter_constraints=[ParameterConstraint(1.0*x1 + 1.0*x2 + 1.0*x3 <= 1.0), OrderConstraint(x2 <= x1), OrderConstraint(x3 <= x1)]).\n",
+      "[INFO 05-01 08:47:30] ax.service.ax_client: Generated new trial 0 with parameters {'x1': 0.462035, 'x2': 0.140554, 'x3': 0.120071, 'x5': 7.491386} using model Sobol.\n",
+      "[INFO 05-01 08:47:30] ax.service.ax_client: Generated new trial 1 with parameters {'x1': 0.681195, 'x2': 0.083172, 'x3': 0.147011, 'x5': 2.201755} using model Sobol.\n",
+      "[INFO 05-01 08:47:30] ax.service.ax_client: Completed trial 0 with data: {'corrosion_score': (1.163775, 0.049476)}.\n",
+      "[INFO 05-01 08:47:30] ax.service.ax_client: Completed trial 1 with data: {'corrosion_score': (2.737126, 0.048125)}.\n",
+      "[INFO 05-01 08:47:30] ax.service.ax_client: Generated new trial 2 with parameters {'x1': 0.540494, 'x2': 0.28639, 'x3': 0.106513, 'x5': 3.150732} using model Sobol.\n",
+      "/Users/andrewf/miniconda3/envs/ax_env/lib/python3.9/site-packages/ax/core/data.py:284: FutureWarning:\n",
+      "\n",
+      "The behavior of DataFrame concatenation with empty or all-NA entries is deprecated. In a future version, this will no longer exclude empty or all-NA columns when determining the result dtypes. To retain the old behavior, exclude the relevant entries before the concat operation.\n",
+      "\n",
+      "[INFO 05-01 08:47:30] ax.service.ax_client: Generated new trial 3 with parameters {'x1': 0.17495, 'x2': 0.00773, 'x3': 0.057119, 'x5': 9.218317} using model Sobol.\n",
+      "[INFO 05-01 08:47:30] ax.service.ax_client: Completed trial 2 with data: {'corrosion_score': (2.172886, 0.019982)}.\n",
+      "[INFO 05-01 08:47:30] ax.service.ax_client: Completed trial 3 with data: {'corrosion_score': (7.653779, 0.056008)}.\n",
+      "[INFO 05-01 08:47:30] ax.service.ax_client: Generated new trial 4 with parameters {'x1': 0.282052, 'x2': 0.122099, 'x3': 0.234689, 'x5': 3.671898} using model Sobol.\n",
+      "/Users/andrewf/miniconda3/envs/ax_env/lib/python3.9/site-packages/ax/core/data.py:284: FutureWarning:\n",
+      "\n",
+      "The behavior of DataFrame concatenation with empty or all-NA entries is deprecated. In a future version, this will no longer exclude empty or all-NA columns when determining the result dtypes. To retain the old behavior, exclude the relevant entries before the concat operation.\n",
+      "\n",
+      "[INFO 05-01 08:47:30] ax.service.ax_client: Generated new trial 5 with parameters {'x1': 0.369966, 'x2': 0.357531, 'x3': 0.268685, 'x5': 7.204235} using model Sobol.\n",
+      "[INFO 05-01 08:47:30] ax.service.ax_client: Completed trial 4 with data: {'corrosion_score': (3.35743, 0.092051)}.\n",
+      "[INFO 05-01 08:47:30] ax.service.ax_client: Completed trial 5 with data: {'corrosion_score': (1.162666, 0.012073)}.\n",
+      "Sample: 100%|██████████| 2048/2048 [00:29, 69.31it/s, step size=3.53e-01, acc. prob=0.912]\n",
+      "[INFO 05-01 08:48:03] ax.service.ax_client: Generated new trial 6 with parameters {'x1': 1.0, 'x2': 0.0, 'x3': 0.0, 'x5': 6.657872} using model FullyBayesian.\n",
+      "/Users/andrewf/miniconda3/envs/ax_env/lib/python3.9/site-packages/ax/core/data.py:284: FutureWarning:\n",
+      "\n",
+      "The behavior of DataFrame concatenation with empty or all-NA entries is deprecated. In a future version, this will no longer exclude empty or all-NA columns when determining the result dtypes. To retain the old behavior, exclude the relevant entries before the concat operation.\n",
+      "\n",
+      "[INFO 05-01 08:48:03] ax.modelbridge.torch: The observations are identical to the last set of observations used to fit the model. Skipping model fitting.\n",
+      "Warmup:   9%|▊         | 178/2048 [03:56,  1.33s/it, step size=4.18e-01, acc. prob=0.778]\n",
+      "Warmup:   5%|▌         | 108/2048 [01:52,  1.05s/it, step size=8.54e-01, acc. prob=0.781]\n",
+      "[INFO 05-01 08:48:07] ax.service.ax_client: Generated new trial 7 with parameters {'x1': 1.0, 'x2': 0.0, 'x3': 0.0, 'x5': 6.658105} using model FullyBayesian.\n",
+      "[INFO 05-01 08:48:07] ax.service.ax_client: Completed trial 6 with data: {'corrosion_score': (5.56047, 0.016152)}.\n",
+      "[INFO 05-01 08:48:07] ax.service.ax_client: Completed trial 7 with data: {'corrosion_score': (5.56029, 0.082192)}.\n",
+      "Sample: 100%|██████████| 2048/2048 [00:30, 66.32it/s, step size=3.76e-01, acc. prob=0.868]\n",
+      "[INFO 05-01 08:48:41] ax.service.ax_client: Generated new trial 8 with parameters {'x1': 0.5, 'x2': 0.0, 'x3': 0.5, 'x5': 10.0} using model FullyBayesian.\n",
+      "/Users/andrewf/miniconda3/envs/ax_env/lib/python3.9/site-packages/ax/core/data.py:284: FutureWarning:\n",
+      "\n",
+      "The behavior of DataFrame concatenation with empty or all-NA entries is deprecated. In a future version, this will no longer exclude empty or all-NA columns when determining the result dtypes. To retain the old behavior, exclude the relevant entries before the concat operation.\n",
+      "\n",
+      "[INFO 05-01 08:48:41] ax.modelbridge.torch: The observations are identical to the last set of observations used to fit the model. Skipping model fitting.\n",
+      "[INFO 05-01 08:48:45] ax.service.ax_client: Generated new trial 9 with parameters {'x1': 0.5, 'x2': 0.0, 'x3': 0.5, 'x5': 7.195933} using model FullyBayesian.\n",
+      "[INFO 05-01 08:48:45] ax.service.ax_client: Completed trial 8 with data: {'corrosion_score': (3.228185, 0.015832)}.\n",
+      "[INFO 05-01 08:48:45] ax.service.ax_client: Completed trial 9 with data: {'corrosion_score': (2.850999, 0.079142)}.\n",
+      "Sample: 100%|██████████| 2048/2048 [00:26, 76.71it/s, step size=3.86e-01, acc. prob=0.895]\n",
+      "[INFO 05-01 08:49:16] ax.service.ax_client: Generated new trial 10 with parameters {'x1': 0.424484, 'x2': 0.424484, 'x3': 0.0, 'x5': 7.339948} using model FullyBayesian.\n",
+      "/Users/andrewf/miniconda3/envs/ax_env/lib/python3.9/site-packages/ax/core/data.py:284: FutureWarning:\n",
+      "\n",
+      "The behavior of DataFrame concatenation with empty or all-NA entries is deprecated. In a future version, this will no longer exclude empty or all-NA columns when determining the result dtypes. To retain the old behavior, exclude the relevant entries before the concat operation.\n",
+      "\n",
+      "[INFO 05-01 08:49:16] ax.modelbridge.torch: The observations are identical to the last set of observations used to fit the model. Skipping model fitting.\n",
+      "[INFO 05-01 08:49:20] ax.service.ax_client: Generated new trial 11 with parameters {'x1': 0.424514, 'x2': 0.424514, 'x3': 0.0, 'x5': 7.345266} using model FullyBayesian.\n",
+      "[INFO 05-01 08:49:20] ax.service.ax_client: Completed trial 10 with data: {'corrosion_score': (0.953817, 0.019769)}.\n",
+      "[INFO 05-01 08:49:20] ax.service.ax_client: Completed trial 11 with data: {'corrosion_score': (0.950801, 0.022215)}.\n",
+      "Sample: 100%|██████████| 2048/2048 [00:22, 89.76it/s, step size=5.13e-01, acc. prob=0.835] \n",
+      "[INFO 05-01 08:49:46] ax.service.ax_client: Generated new trial 12 with parameters {'x1': 0.426213, 'x2': 0.0, 'x3': 0.0, 'x5': 0.1} using model FullyBayesian.\n",
+      "/Users/andrewf/miniconda3/envs/ax_env/lib/python3.9/site-packages/ax/core/data.py:284: FutureWarning:\n",
+      "\n",
+      "The behavior of DataFrame concatenation with empty or all-NA entries is deprecated. In a future version, this will no longer exclude empty or all-NA columns when determining the result dtypes. To retain the old behavior, exclude the relevant entries before the concat operation.\n",
+      "\n",
+      "[INFO 05-01 08:49:46] ax.modelbridge.torch: The observations are identical to the last set of observations used to fit the model. Skipping model fitting.\n",
+      "[INFO 05-01 08:49:52] ax.service.ax_client: Generated new trial 13 with parameters {'x1': 0.5, 'x2': 0.5, 'x3': 0.0, 'x5': 10.0} using model FullyBayesian.\n",
+      "[INFO 05-01 08:49:52] ax.service.ax_client: Completed trial 12 with data: {'corrosion_score': (5.697937, 0.069679)}.\n",
+      "[INFO 05-01 08:49:52] ax.service.ax_client: Completed trial 13 with data: {'corrosion_score': (4.724092, 0.051947)}.\n",
+      "Sample: 100%|██████████| 2048/2048 [00:32, 62.65it/s, step size=3.01e-01, acc. prob=0.918]\n",
+      "[INFO 05-01 08:50:29] ax.service.ax_client: Generated new trial 14 with parameters {'x1': 0.420853, 'x2': 0.344086, 'x3': 0.0, 'x5': 5.92896} using model FullyBayesian.\n",
+      "/Users/andrewf/miniconda3/envs/ax_env/lib/python3.9/site-packages/ax/core/data.py:284: FutureWarning:\n",
+      "\n",
+      "The behavior of DataFrame concatenation with empty or all-NA entries is deprecated. In a future version, this will no longer exclude empty or all-NA columns when determining the result dtypes. To retain the old behavior, exclude the relevant entries before the concat operation.\n",
+      "\n",
+      "[INFO 05-01 08:50:29] ax.modelbridge.torch: The observations are identical to the last set of observations used to fit the model. Skipping model fitting.\n",
+      "[INFO 05-01 08:50:36] ax.service.ax_client: Generated new trial 15 with parameters {'x1': 0.619151, 'x2': 0.230517, 'x3': 0.150332, 'x5': 7.898638} using model FullyBayesian.\n",
+      "[INFO 05-01 08:50:36] ax.service.ax_client: Completed trial 14 with data: {'corrosion_score': (2.554639, 0.05664)}.\n",
+      "[INFO 05-01 08:50:36] ax.service.ax_client: Completed trial 15 with data: {'corrosion_score': (0.45323, 0.083762)}.\n",
+      "Sample: 100%|██████████| 2048/2048 [00:40, 50.65it/s, step size=2.53e-01, acc. prob=0.903]\n",
+      "[INFO 05-01 08:51:21] ax.service.ax_client: Generated new trial 16 with parameters {'x1': 0.81943, 'x2': 0.18057, 'x3': 0.0, 'x5': 8.956439} using model FullyBayesian.\n",
+      "/Users/andrewf/miniconda3/envs/ax_env/lib/python3.9/site-packages/ax/core/data.py:284: FutureWarning:\n",
+      "\n",
+      "The behavior of DataFrame concatenation with empty or all-NA entries is deprecated. In a future version, this will no longer exclude empty or all-NA columns when determining the result dtypes. To retain the old behavior, exclude the relevant entries before the concat operation.\n",
+      "\n",
+      "[INFO 05-01 08:51:21] ax.modelbridge.torch: The observations are identical to the last set of observations used to fit the model. Skipping model fitting.\n",
+      "[INFO 05-01 08:51:26] ax.service.ax_client: Generated new trial 17 with parameters {'x1': 0.819333, 'x2': 0.180667, 'x3': 0.0, 'x5': 8.951648} using model FullyBayesian.\n",
+      "[INFO 05-01 08:51:26] ax.service.ax_client: Completed trial 16 with data: {'corrosion_score': (1.247548, 0.044192)}.\n",
+      "[INFO 05-01 08:51:26] ax.service.ax_client: Completed trial 17 with data: {'corrosion_score': (1.243592, 0.038841)}.\n",
+      "Sample: 100%|██████████| 2048/2048 [00:33, 61.85it/s, step size=3.20e-01, acc. prob=0.874]\n",
+      "[INFO 05-01 08:52:04] ax.service.ax_client: Generated new trial 18 with parameters {'x1': 0.553224, 'x2': 0.184695, 'x3': 0.262081, 'x5': 8.289913} using model FullyBayesian.\n",
+      "/Users/andrewf/miniconda3/envs/ax_env/lib/python3.9/site-packages/ax/core/data.py:284: FutureWarning:\n",
+      "\n",
+      "The behavior of DataFrame concatenation with empty or all-NA entries is deprecated. In a future version, this will no longer exclude empty or all-NA columns when determining the result dtypes. To retain the old behavior, exclude the relevant entries before the concat operation.\n",
+      "\n",
+      "[INFO 05-01 08:52:04] ax.modelbridge.torch: The observations are identical to the last set of observations used to fit the model. Skipping model fitting.\n",
+      "[INFO 05-01 08:52:09] ax.service.ax_client: Generated new trial 19 with parameters {'x1': 0.553114, 'x2': 0.184659, 'x3': 0.262227, 'x5': 8.29011} using model FullyBayesian.\n",
+      "[INFO 05-01 08:52:09] ax.service.ax_client: Completed trial 18 with data: {'corrosion_score': (0.409177, 0.034298)}.\n",
+      "[INFO 05-01 08:52:09] ax.service.ax_client: Completed trial 19 with data: {'corrosion_score': (0.409134, 0.061954)}.\n",
+      "Sample: 100%|██████████| 2048/2048 [00:39, 51.86it/s, step size=2.33e-01, acc. prob=0.921]\n"
+     ]
+    }
+   ],
+   "source": [
+    "import numpy as np\n",
+    "from ax.service.ax_client import AxClient, ObjectiveProperties\n",
+    "\n",
+    "from ax.modelbridge.factory import Models\n",
+    "from ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n",
+    "\n",
+    "\n",
+    "obj1_name = \"corrosion_score\" # CHANGE: update objective name\n",
+    "\n",
+    "# CHANGE: remove the branin dummy objective function, we will use the above function\n",
+    "\n",
+    "total = 1.0 # CHANGE: update total component fraction\n",
+    "\n",
+    "\n",
+    "gs = GenerationStrategy(\n",
+    "    steps=[\n",
+    "        GenerationStep(\n",
+    "            model=Models.SOBOL,\n",
+    "            num_trials=6,\n",
+    "            min_trials_observed=3,\n",
+    "            max_parallelism=5,\n",
+    "            model_kwargs={\"seed\": 999},\n",
+    "            model_gen_kwargs={},\n",
+    "        ),\n",
+    "        GenerationStep(\n",
+    "            model=Models.FULLYBAYESIAN,\n",
+    "            num_trials=-1,\n",
+    "            max_parallelism=3,\n",
+    "            should_deduplicate=True, # CHANGE: reduce duplicate suggestions\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024}, # CHANGE: increase\n",
+    "        ),\n",
+    "    ]\n",
+    ")\n",
+    "\n",
+    "ax_client = AxClient(generation_strategy=gs, random_seed=42) # CHANGE: add random seed for reproducibility\n",
+    "\n",
+    "ax_client.create_experiment(\n",
+    "    parameters=[\n",
+    "        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]}, # CHANGE: update parameter\n",
+    "        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]}, # CHANGE: update parameter\n",
+    "        {\"name\": \"x3\", \"type\": \"range\", \"bounds\": [0.0, total]}, # CHANGE: add new parameter\n",
+    "        {\"name\": \"x5\", \"type\": \"range\", \"bounds\": [0.1, 10.0]}, # CHANGE: add new parameter\n",
+    "    ],\n",
+    "    objectives={\n",
+    "        obj1_name: ObjectiveProperties(minimize=True),\n",
+    "    },\n",
+    "    parameter_constraints=[\n",
+    "        f\"x1 + x2 + x3 <= {total}\", # CHANGE: update composition constraint\n",
+    "        \"x1 >= x2\", # CHANGE: update order constraint\n",
+    "        \"x1 >= x3\", # CHANGE: add order constraint\n",
+    "    ],\n",
+    ")\n",
+    "\n",
+    "batch_size = 2\n",
+    "\n",
+    "for _ in range(10): # CHANGE: decrease number of iterations\n",
+    "\n",
+    "    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n",
+    "    for trial_index, parameterization in list(parameterizations.items()):\n",
+    "        \n",
+    "        # CHANGE: pull all added parameters from the parameterization\n",
+    "        x1 = parameterization[\"x1\"]\n",
+    "        x2 = parameterization[\"x2\"]\n",
+    "        x3 = parameterization[\"x3\"]\n",
+    "        x4 = total - (x1 + x2 + x3) # CHANGE: update composition constraint\n",
+    "        x5 = parameterization[\"x5\"]\n",
+    "\n",
+    "        results = simulate_corrosion(x1, x2, x3, x4, x5)\n",
+    "        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n",
+    "\n",
+    "best_parameters, metrics = ax_client.get_best_parameters()"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "### Show the Best Parameters\n",
+    "\n",
+    "After our optimization loop has completed, we can use the model to find the best parameters and their corresponding strength value. These will be our optimial set of parameters that we use in the 3D printer going forward."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 338,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "Sample: 100%|██████████| 2048/2048 [00:42, 47.73it/s, step size=2.50e-01, acc. prob=0.877]\n"
+     ]
+    },
+    {
+     "data": {
+      "text/plain": [
+       "(19,\n",
+       " {'x1': 0.5531141417704178,\n",
+       "  'x2': 0.1846591406143514,\n",
+       "  'x3': 0.26222671761513805,\n",
+       "  'x5': 8.2901097602077},\n",
+       " ({'corrosion_score': 0.41040928033269086},\n",
+       "  {'corrosion_score': {'corrosion_score': 0.0008997393689987544}}))"
+      ]
+     },
+     "execution_count": 338,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "ax_client.get_best_trial()"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "### Plotting Optimization Performance\n",
+    "\n",
+    "We can plot the performance of our optmization loop to see how the optimization task progressed as a function of iteration count."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 339,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "[WARNING 05-01 08:58:34] ax.service.utils.report_utils: Column reason missing for all trials. Not appending column.\n"
+     ]
+    },
+    {
+     "data": {
+      "text/plain": [
+       "(0.0, 2.0)"
+      ]
+     },
+     "execution_count": 339,
+     "metadata": {},
+     "output_type": "execute_result"
+    },
+    {
+     "data": {
+      "image/png": "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",
+      "text/plain": [
+       "<Figure size 720x480 with 1 Axes>"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    }
+   ],
+   "source": [
+    "import matplotlib.pyplot as plt\n",
+    "h_blue = '#0033FF'\n",
+    "\n",
+    "df = ax_client.get_trials_data_frame()\n",
+    "fig, ax = plt.subplots(figsize=(6,4), dpi=120)\n",
+    "ax.plot(df.corrosion_score.values, ls='None', marker='o', mfc='none', mec='black')\n",
+    "ax.plot(np.minimum.accumulate(df.corrosion_score.values), color=h_blue)\n",
+    "plt.xlabel('Trial')\n",
+    "plt.ylabel('Corrosion Score [lower is better]')\n",
+    "plt.ylim(0, 2)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 340,
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "application/vnd.plotly.v1+json": {
+       "config": {
+        "linkText": "Export to plot.ly",
+        "plotlyServerURL": "https://plot.ly",
+        "showLink": false
+       },
+       "data": [
+        {
+         "hoverinfo": "none",
+         "line": {
+          "color": "black",
+          "dash": "dot",
+          "width": 2
+         },
+         "mode": "lines",
+         "showlegend": false,
+         "type": "scatter",
+         "visible": true,
+         "x": [
+          -3.952154604781554,
+          10.903313786058435
+         ],
+         "y": [
+          -3.952154604781554,
+          10.903313786058435
+         ]
+        },
+        {
+         "error_x": {
+          "array": [
+           0.09697307193758328,
+           0.09432580615956097,
+           0.03916477554641934,
+           0.10977525367163038,
+           0.18041947262022728,
+           0.023662859070073566,
+           0.0316572302731216,
+           0.16109585493320808,
+           0.03103124681089616,
+           0.15511805238701318,
+           0.03874663888694734,
+           0.04354174301098419,
+           0.13657119941037701,
+           0.1018157587494321,
+           0.1110139250248606,
+           0.1641737784542783,
+           0.08661683840337824,
+           0.07612889196206886
+          ],
+          "color": "rgba(128,177,211,0.4)",
+          "type": "data"
+         },
+         "error_y": {
+          "array": [
+           2.7256679784174875,
+           4.199633838796122,
+           3.599793582557131,
+           5.2253166045607005,
+           4.516845300842414,
+           3.0953335998152287,
+           0.16104538196927914,
+           0.03166602726203759,
+           4.657165685333809,
+           3.574701887771811,
+           0.045083000885533506,
+           0.040628188869374526,
+           6.458009708196848,
+           5.374882199652275,
+           3.9081294574205843,
+           2.977922279320509,
+           0.07742037877804649,
+           0.08754464854448725
+          ],
+          "color": "rgba(128,177,211,0.4)",
+          "type": "data"
+         },
+         "hoverinfo": "text",
+         "marker": {
+          "color": "rgba(128,177,211,1)"
+         },
+         "mode": "markers",
+         "name": "In-sample",
+         "showlegend": true,
+         "text": [
+          "<b>Arm 0_0</b><br><br>Actual Outcome: 1.164 [1.067, 1.261]<br>Predicted Outcome: 2.102 [-0.624, 4.828]<br><br><em>Parameterization:</em><br>x1: 0.46203490160405636<br>x2: 0.1405540555715561<br>x3: 0.12007084302604198<br>x5: 7.49138588104397",
+          "<b>Arm 1_0</b><br><br>Actual Outcome: 2.737 [2.643, 2.831]<br>Predicted Outcome: 3.082 [-1.118, 7.281]<br><br><em>Parameterization:</em><br>x1: 0.6811950104311109<br>x2: 0.08317248616367579<br>x3: 0.14701138995587826<br>x5: 2.201754858437926",
+          "<b>Arm 2_0</b><br><br>Actual Outcome: 2.173 [2.134, 2.212]<br>Predicted Outcome: 3.007 [-0.593, 6.607]<br><br><em>Parameterization:</em><br>x1: 0.5404942771419883<br>x2: 0.28639009688049555<br>x3: 0.10651267971843481<br>x5: 3.1507322339341046",
+          "<b>Arm 3_0</b><br><br>Actual Outcome: 7.654 [7.544, 7.764]<br>Predicted Outcome: 3.416 [-1.809, 8.641]<br><br><em>Parameterization:</em><br>x1: 0.17495027277618647<br>x2: 0.007729803211987019<br>x3: 0.0571193927899003<br>x5: 9.218316878192127",
+          "<b>Arm 4_0</b><br><br>Actual Outcome: 3.357 [3.177, 3.538]<br>Predicted Outcome: 2.838 [-1.679, 7.354]<br><br><em>Parameterization:</em><br>x1: 0.2820521341636777<br>x2: 0.12209852878004313<br>x3: 0.23468931019306183<br>x5: 3.671897889301181",
+          "<b>Arm 5_0</b><br><br>Actual Outcome: 1.163 [1.139, 1.186]<br>Predicted Outcome: 1.707 [-1.388, 4.803]<br><br><em>Parameterization:</em><br>x1: 0.36996550764888525<br>x2: 0.35753058083355427<br>x3: 0.2686848407611251<br>x5: 7.204235409293323",
+          "<b>Arm 6_0</b><br><br>Actual Outcome: 5.56 [5.529, 5.592]<br>Predicted Outcome: 5.557 [5.396, 5.718]<br><br><em>Parameterization:</em><br>x1: 0.9999999999999886<br>x2: 0.0<br>x3: 1.2886717519318773e-14<br>x5: 6.657871762887141",
+          "<b>Arm 7_0</b><br><br>Actual Outcome: 5.56 [5.399, 5.721]<br>Predicted Outcome: 5.56 [5.529, 5.592]<br><br><em>Parameterization:</em><br>x1: 1.0<br>x2: 5.879023101716506e-16<br>x3: 8.221970846891384e-16<br>x5: 6.658104954751326",
+          "<b>Arm 8_0</b><br><br>Actual Outcome: 3.228 [3.197, 3.259]<br>Predicted Outcome: 4.24 [-0.417, 8.897]<br><br><em>Parameterization:</em><br>x1: 0.5000000000004156<br>x2: 0.0<br>x3: 0.5000000000021749<br>x5: 9.999999999999234",
+          "<b>Arm 9_0</b><br><br>Actual Outcome: 2.851 [2.696, 3.006]<br>Predicted Outcome: 2.316 [-1.259, 5.891]<br><br><em>Parameterization:</em><br>x1: 0.49999999999999983<br>x2: 2.9442422767142952e-15<br>x3: 0.5000000000000115<br>x5: 7.195932963944851",
+          "<b>Arm 10_0</b><br><br>Actual Outcome: 0.954 [0.915, 0.993]<br>Predicted Outcome: 0.952 [0.907, 0.997]<br><br><em>Parameterization:</em><br>x1: 0.4244838630578679<br>x2: 0.42448386306381597<br>x3: 2.1289729987226923e-11<br>x5: 7.339947666832862",
+          "<b>Arm 11_0</b><br><br>Actual Outcome: 0.951 [0.907, 0.994]<br>Predicted Outcome: 0.953 [0.912, 0.994]<br><br><em>Parameterization:</em><br>x1: 0.4245136618061755<br>x2: 0.42451366180615974<br>x3: 0.0<br>x5: 7.345265540626473",
+          "<b>Arm 12_0</b><br><br>Actual Outcome: 5.698 [5.561, 5.835]<br>Predicted Outcome: 3.77 [-2.688, 10.228]<br><br><em>Parameterization:</em><br>x1: 0.4262126487368945<br>x2: 8.147112918011196e-16<br>x3: 4.376823862696568e-15<br>x5: 0.1000000000000202",
+          "<b>Arm 13_0</b><br><br>Actual Outcome: 4.724 [4.622, 4.826]<br>Predicted Outcome: 2.098 [-3.277, 7.473]<br><br><em>Parameterization:</em><br>x1: 0.5000000000012181<br>x2: 0.5000000000106196<br>x3: 0.0<br>x5: 10.0",
+          "<b>Arm 14_0</b><br><br>Actual Outcome: 2.555 [2.444, 2.666]<br>Predicted Outcome: 1.663 [-2.245, 5.571]<br><br><em>Parameterization:</em><br>x1: 0.42085294074668167<br>x2: 0.3440857934253744<br>x3: 0.0<br>x5: 5.928959908522031",
+          "<b>Arm 15_0</b><br><br>Actual Outcome: 0.453 [0.289, 0.617]<br>Predicted Outcome: 1.004 [-1.974, 3.981]<br><br><em>Parameterization:</em><br>x1: 0.6191508786933483<br>x2: 0.23051694149204408<br>x3: 0.1503321798146068<br>x5: 7.898638178540372",
+          "<b>Arm 16_0</b><br><br>Actual Outcome: 1.248 [1.161, 1.334]<br>Predicted Outcome: 1.246 [1.169, 1.324]<br><br><em>Parameterization:</em><br>x1: 0.8194301997175271<br>x2: 0.18056980028249606<br>x3: 2.299616162607646e-15<br>x5: 8.956439436443388",
+          "<b>Arm 17_0</b><br><br>Actual Outcome: 1.244 [1.167, 1.320]<br>Predicted Outcome: 1.245 [1.158, 1.333]<br><br><em>Parameterization:</em><br>x1: 0.8193333945709272<br>x2: 0.18066660542907836<br>x3: 0.0<br>x5: 8.951647627618732"
+         ],
+         "type": "scatter",
+         "visible": true,
+         "x": [
+          1.1637748514190176,
+          2.7371261343507274,
+          2.1728861476173438,
+          7.653779219276586,
+          3.3574303343959673,
+          1.1626663489625728,
+          5.560470369357521,
+          5.5602897702587555,
+          3.2281845802895037,
+          2.8509992024517383,
+          0.953817334433674,
+          0.9508013533028197,
+          5.697936838066944,
+          4.7240923300902775,
+          2.55463885701626,
+          0.45323031180355544,
+          1.247548031754366,
+          1.2435919914412081
+         ],
+         "y": [
+          2.102091278956899,
+          3.0818544551702387,
+          3.0070574018667036,
+          3.4158654322816258,
+          2.8375954991177417,
+          1.7072225760669248,
+          5.557351121990914,
+          5.560187764542006,
+          4.240162969938776,
+          2.3160229549868583,
+          0.9522162444816504,
+          0.9529349604944826,
+          3.770055514641586,
+          2.09797615809072,
+          1.662826192337171,
+          1.0035465684178826,
+          1.2464475640682549,
+          1.2451972626406453
+         ]
+        }
+       ],
+       "layout": {
+        "annotations": [
+         {
+          "showarrow": false,
+          "text": "Show CI",
+          "x": 1.125,
+          "xanchor": "left",
+          "xref": "paper",
+          "y": 0.9,
+          "yanchor": "middle",
+          "yref": "paper"
+         }
+        ],
+        "height": 500,
+        "hovermode": "closest",
+        "margin": {
+         "b": 90
+        },
+        "showlegend": false,
+        "template": {
+         "data": {
+          "bar": [
+           {
+            "error_x": {
+             "color": "#2a3f5f"
+            },
+            "error_y": {
+             "color": "#2a3f5f"
+            },
+            "marker": {
+             "line": {
+              "color": "#E5ECF6",
+              "width": 0.5
+             },
+             "pattern": {
+              "fillmode": "overlay",
+              "size": 10,
+              "solidity": 0.2
+             }
+            },
+            "type": "bar"
+           }
+          ],
+          "barpolar": [
+           {
+            "marker": {
+             "line": {
+              "color": "#E5ECF6",
+              "width": 0.5
+             },
+             "pattern": {
+              "fillmode": "overlay",
+              "size": 10,
+              "solidity": 0.2
+             }
+            },
+            "type": "barpolar"
+           }
+          ],
+          "carpet": [
+           {
+            "aaxis": {
+             "endlinecolor": "#2a3f5f",
+             "gridcolor": "white",
+             "linecolor": "white",
+             "minorgridcolor": "white",
+             "startlinecolor": "#2a3f5f"
+            },
+            "baxis": {
+             "endlinecolor": "#2a3f5f",
+             "gridcolor": "white",
+             "linecolor": "white",
+             "minorgridcolor": "white",
+             "startlinecolor": "#2a3f5f"
+            },
+            "type": "carpet"
+           }
+          ],
+          "choropleth": [
+           {
+            "colorbar": {
+             "outlinewidth": 0,
+             "ticks": ""
+            },
+            "type": "choropleth"
+           }
+          ],
+          "contour": [
+           {
+            "colorbar": {
+             "outlinewidth": 0,
+             "ticks": ""
+            },
+            "colorscale": [
+             [
+              0,
+              "#0d0887"
+             ],
+             [
+              0.1111111111111111,
+              "#46039f"
+             ],
+             [
+              0.2222222222222222,
+              "#7201a8"
+             ],
+             [
+              0.3333333333333333,
+              "#9c179e"
+             ],
+             [
+              0.4444444444444444,
+              "#bd3786"
+             ],
+             [
+              0.5555555555555556,
+              "#d8576b"
+             ],
+             [
+              0.6666666666666666,
+              "#ed7953"
+             ],
+             [
+              0.7777777777777778,
+              "#fb9f3a"
+             ],
+             [
+              0.8888888888888888,
+              "#fdca26"
+             ],
+             [
+              1,
+              "#f0f921"
+             ]
+            ],
+            "type": "contour"
+           }
+          ],
+          "contourcarpet": [
+           {
+            "colorbar": {
+             "outlinewidth": 0,
+             "ticks": ""
+            },
+            "type": "contourcarpet"
+           }
+          ],
+          "heatmap": [
+           {
+            "colorbar": {
+             "outlinewidth": 0,
+             "ticks": ""
+            },
+            "colorscale": [
+             [
+              0,
+              "#0d0887"
+             ],
+             [
+              0.1111111111111111,
+              "#46039f"
+             ],
+             [
+              0.2222222222222222,
+              "#7201a8"
+             ],
+             [
+              0.3333333333333333,
+              "#9c179e"
+             ],
+             [
+              0.4444444444444444,
+              "#bd3786"
+             ],
+             [
+              0.5555555555555556,
+              "#d8576b"
+             ],
+             [
+              0.6666666666666666,
+              "#ed7953"
+             ],
+             [
+              0.7777777777777778,
+              "#fb9f3a"
+             ],
+             [
+              0.8888888888888888,
+              "#fdca26"
+             ],
+             [
+              1,
+              "#f0f921"
+             ]
+            ],
+            "type": "heatmap"
+           }
+          ],
+          "heatmapgl": [
+           {
+            "colorbar": {
+             "outlinewidth": 0,
+             "ticks": ""
+            },
+            "colorscale": [
+             [
+              0,
+              "#0d0887"
+             ],
+             [
+              0.1111111111111111,
+              "#46039f"
+             ],
+             [
+              0.2222222222222222,
+              "#7201a8"
+             ],
+             [
+              0.3333333333333333,
+              "#9c179e"
+             ],
+             [
+              0.4444444444444444,
+              "#bd3786"
+             ],
+             [
+              0.5555555555555556,
+              "#d8576b"
+             ],
+             [
+              0.6666666666666666,
+              "#ed7953"
+             ],
+             [
+              0.7777777777777778,
+              "#fb9f3a"
+             ],
+             [
+              0.8888888888888888,
+              "#fdca26"
+             ],
+             [
+              1,
+              "#f0f921"
+             ]
+            ],
+            "type": "heatmapgl"
+           }
+          ],
+          "histogram": [
+           {
+            "marker": {
+             "pattern": {
+              "fillmode": "overlay",
+              "size": 10,
+              "solidity": 0.2
+             }
+            },
+            "type": "histogram"
+           }
+          ],
+          "histogram2d": [
+           {
+            "colorbar": {
+             "outlinewidth": 0,
+             "ticks": ""
+            },
+            "colorscale": [
+             [
+              0,
+              "#0d0887"
+             ],
+             [
+              0.1111111111111111,
+              "#46039f"
+             ],
+             [
+              0.2222222222222222,
+              "#7201a8"
+             ],
+             [
+              0.3333333333333333,
+              "#9c179e"
+             ],
+             [
+              0.4444444444444444,
+              "#bd3786"
+             ],
+             [
+              0.5555555555555556,
+              "#d8576b"
+             ],
+             [
+              0.6666666666666666,
+              "#ed7953"
+             ],
+             [
+              0.7777777777777778,
+              "#fb9f3a"
+             ],
+             [
+              0.8888888888888888,
+              "#fdca26"
+             ],
+             [
+              1,
+              "#f0f921"
+             ]
+            ],
+            "type": "histogram2d"
+           }
+          ],
+          "histogram2dcontour": [
+           {
+            "colorbar": {
+             "outlinewidth": 0,
+             "ticks": ""
+            },
+            "colorscale": [
+             [
+              0,
+              "#0d0887"
+             ],
+             [
+              0.1111111111111111,
+              "#46039f"
+             ],
+             [
+              0.2222222222222222,
+              "#7201a8"
+             ],
+             [
+              0.3333333333333333,
+              "#9c179e"
+             ],
+             [
+              0.4444444444444444,
+              "#bd3786"
+             ],
+             [
+              0.5555555555555556,
+              "#d8576b"
+             ],
+             [
+              0.6666666666666666,
+              "#ed7953"
+             ],
+             [
+              0.7777777777777778,
+              "#fb9f3a"
+             ],
+             [
+              0.8888888888888888,
+              "#fdca26"
+             ],
+             [
+              1,
+              "#f0f921"
+             ]
+            ],
+            "type": "histogram2dcontour"
+           }
+          ],
+          "mesh3d": [
+           {
+            "colorbar": {
+             "outlinewidth": 0,
+             "ticks": ""
+            },
+            "type": "mesh3d"
+           }
+          ],
+          "parcoords": [
+           {
+            "line": {
+             "colorbar": {
+              "outlinewidth": 0,
+              "ticks": ""
+             }
+            },
+            "type": "parcoords"
+           }
+          ],
+          "pie": [
+           {
+            "automargin": true,
+            "type": "pie"
+           }
+          ],
+          "scatter": [
+           {
+            "fillpattern": {
+             "fillmode": "overlay",
+             "size": 10,
+             "solidity": 0.2
+            },
+            "type": "scatter"
+           }
+          ],
+          "scatter3d": [
+           {
+            "line": {
+             "colorbar": {
+              "outlinewidth": 0,
+              "ticks": ""
+             }
+            },
+            "marker": {
+             "colorbar": {
+              "outlinewidth": 0,
+              "ticks": ""
+             }
+            },
+            "type": "scatter3d"
+           }
+          ],
+          "scattercarpet": [
+           {
+            "marker": {
+             "colorbar": {
+              "outlinewidth": 0,
+              "ticks": ""
+             }
+            },
+            "type": "scattercarpet"
+           }
+          ],
+          "scattergeo": [
+           {
+            "marker": {
+             "colorbar": {
+              "outlinewidth": 0,
+              "ticks": ""
+             }
+            },
+            "type": "scattergeo"
+           }
+          ],
+          "scattergl": [
+           {
+            "marker": {
+             "colorbar": {
+              "outlinewidth": 0,
+              "ticks": ""
+             }
+            },
+            "type": "scattergl"
+           }
+          ],
+          "scattermapbox": [
+           {
+            "marker": {
+             "colorbar": {
+              "outlinewidth": 0,
+              "ticks": ""
+             }
+            },
+            "type": "scattermapbox"
+           }
+          ],
+          "scatterpolar": [
+           {
+            "marker": {
+             "colorbar": {
+              "outlinewidth": 0,
+              "ticks": ""
+             }
+            },
+            "type": "scatterpolar"
+           }
+          ],
+          "scatterpolargl": [
+           {
+            "marker": {
+             "colorbar": {
+              "outlinewidth": 0,
+              "ticks": ""
+             }
+            },
+            "type": "scatterpolargl"
+           }
+          ],
+          "scatterternary": [
+           {
+            "marker": {
+             "colorbar": {
+              "outlinewidth": 0,
+              "ticks": ""
+             }
+            },
+            "type": "scatterternary"
+           }
+          ],
+          "surface": [
+           {
+            "colorbar": {
+             "outlinewidth": 0,
+             "ticks": ""
+            },
+            "colorscale": [
+             [
+              0,
+              "#0d0887"
+             ],
+             [
+              0.1111111111111111,
+              "#46039f"
+             ],
+             [
+              0.2222222222222222,
+              "#7201a8"
+             ],
+             [
+              0.3333333333333333,
+              "#9c179e"
+             ],
+             [
+              0.4444444444444444,
+              "#bd3786"
+             ],
+             [
+              0.5555555555555556,
+              "#d8576b"
+             ],
+             [
+              0.6666666666666666,
+              "#ed7953"
+             ],
+             [
+              0.7777777777777778,
+              "#fb9f3a"
+             ],
+             [
+              0.8888888888888888,
+              "#fdca26"
+             ],
+             [
+              1,
+              "#f0f921"
+             ]
+            ],
+            "type": "surface"
+           }
+          ],
+          "table": [
+           {
+            "cells": {
+             "fill": {
+              "color": "#EBF0F8"
+             },
+             "line": {
+              "color": "white"
+             }
+            },
+            "header": {
+             "fill": {
+              "color": "#C8D4E3"
+             },
+             "line": {
+              "color": "white"
+             }
+            },
+            "type": "table"
+           }
+          ]
+         },
+         "layout": {
+          "annotationdefaults": {
+           "arrowcolor": "#2a3f5f",
+           "arrowhead": 0,
+           "arrowwidth": 1
+          },
+          "autotypenumbers": "strict",
+          "coloraxis": {
+           "colorbar": {
+            "outlinewidth": 0,
+            "ticks": ""
+           }
+          },
+          "colorscale": {
+           "diverging": [
+            [
+             0,
+             "#8e0152"
+            ],
+            [
+             0.1,
+             "#c51b7d"
+            ],
+            [
+             0.2,
+             "#de77ae"
+            ],
+            [
+             0.3,
+             "#f1b6da"
+            ],
+            [
+             0.4,
+             "#fde0ef"
+            ],
+            [
+             0.5,
+             "#f7f7f7"
+            ],
+            [
+             0.6,
+             "#e6f5d0"
+            ],
+            [
+             0.7,
+             "#b8e186"
+            ],
+            [
+             0.8,
+             "#7fbc41"
+            ],
+            [
+             0.9,
+             "#4d9221"
+            ],
+            [
+             1,
+             "#276419"
+            ]
+           ],
+           "sequential": [
+            [
+             0,
+             "#0d0887"
+            ],
+            [
+             0.1111111111111111,
+             "#46039f"
+            ],
+            [
+             0.2222222222222222,
+             "#7201a8"
+            ],
+            [
+             0.3333333333333333,
+             "#9c179e"
+            ],
+            [
+             0.4444444444444444,
+             "#bd3786"
+            ],
+            [
+             0.5555555555555556,
+             "#d8576b"
+            ],
+            [
+             0.6666666666666666,
+             "#ed7953"
+            ],
+            [
+             0.7777777777777778,
+             "#fb9f3a"
+            ],
+            [
+             0.8888888888888888,
+             "#fdca26"
+            ],
+            [
+             1,
+             "#f0f921"
+            ]
+           ],
+           "sequentialminus": [
+            [
+             0,
+             "#0d0887"
+            ],
+            [
+             0.1111111111111111,
+             "#46039f"
+            ],
+            [
+             0.2222222222222222,
+             "#7201a8"
+            ],
+            [
+             0.3333333333333333,
+             "#9c179e"
+            ],
+            [
+             0.4444444444444444,
+             "#bd3786"
+            ],
+            [
+             0.5555555555555556,
+             "#d8576b"
+            ],
+            [
+             0.6666666666666666,
+             "#ed7953"
+            ],
+            [
+             0.7777777777777778,
+             "#fb9f3a"
+            ],
+            [
+             0.8888888888888888,
+             "#fdca26"
+            ],
+            [
+             1,
+             "#f0f921"
+            ]
+           ]
+          },
+          "colorway": [
+           "#636efa",
+           "#EF553B",
+           "#00cc96",
+           "#ab63fa",
+           "#FFA15A",
+           "#19d3f3",
+           "#FF6692",
+           "#B6E880",
+           "#FF97FF",
+           "#FECB52"
+          ],
+          "font": {
+           "color": "#2a3f5f"
+          },
+          "geo": {
+           "bgcolor": "white",
+           "lakecolor": "white",
+           "landcolor": "#E5ECF6",
+           "showlakes": true,
+           "showland": true,
+           "subunitcolor": "white"
+          },
+          "hoverlabel": {
+           "align": "left"
+          },
+          "hovermode": "closest",
+          "mapbox": {
+           "style": "light"
+          },
+          "paper_bgcolor": "white",
+          "plot_bgcolor": "#E5ECF6",
+          "polar": {
+           "angularaxis": {
+            "gridcolor": "white",
+            "linecolor": "white",
+            "ticks": ""
+           },
+           "bgcolor": "#E5ECF6",
+           "radialaxis": {
+            "gridcolor": "white",
+            "linecolor": "white",
+            "ticks": ""
+           }
+          },
+          "scene": {
+           "xaxis": {
+            "backgroundcolor": "#E5ECF6",
+            "gridcolor": "white",
+            "gridwidth": 2,
+            "linecolor": "white",
+            "showbackground": true,
+            "ticks": "",
+            "zerolinecolor": "white"
+           },
+           "yaxis": {
+            "backgroundcolor": "#E5ECF6",
+            "gridcolor": "white",
+            "gridwidth": 2,
+            "linecolor": "white",
+            "showbackground": true,
+            "ticks": "",
+            "zerolinecolor": "white"
+           },
+           "zaxis": {
+            "backgroundcolor": "#E5ECF6",
+            "gridcolor": "white",
+            "gridwidth": 2,
+            "linecolor": "white",
+            "showbackground": true,
+            "ticks": "",
+            "zerolinecolor": "white"
+           }
+          },
+          "shapedefaults": {
+           "line": {
+            "color": "#2a3f5f"
+           }
+          },
+          "ternary": {
+           "aaxis": {
+            "gridcolor": "white",
+            "linecolor": "white",
+            "ticks": ""
+           },
+           "baxis": {
+            "gridcolor": "white",
+            "linecolor": "white",
+            "ticks": ""
+           },
+           "bgcolor": "#E5ECF6",
+           "caxis": {
+            "gridcolor": "white",
+            "linecolor": "white",
+            "ticks": ""
+           }
+          },
+          "title": {
+           "x": 0.05
+          },
+          "xaxis": {
+           "automargin": true,
+           "gridcolor": "white",
+           "linecolor": "white",
+           "ticks": "",
+           "title": {
+            "standoff": 15
+           },
+           "zerolinecolor": "white",
+           "zerolinewidth": 2
+          },
+          "yaxis": {
+           "automargin": true,
+           "gridcolor": "white",
+           "linecolor": "white",
+           "ticks": "",
+           "title": {
+            "standoff": 15
+           },
+           "zerolinecolor": "white",
+           "zerolinewidth": 2
+          }
+         }
+        },
+        "title": {
+         "text": "Cross-validation"
+        },
+        "updatemenus": [
+         {
+          "buttons": [
+           {
+            "args": [
+             {
+              "visible": [
+               true,
+               true
+              ]
+             },
+             {
+              "xaxis.range": [
+               -3.952154604781554,
+               10.903313786058435
+              ],
+              "yaxis.range": [
+               -3.952154604781554,
+               10.903313786058435
+              ]
+             }
+            ],
+            "label": "corrosion_score",
+            "method": "update"
+           }
+          ],
+          "x": 0,
+          "xanchor": "left",
+          "y": 1.125,
+          "yanchor": "top"
+         },
+         {
+          "buttons": [
+           {
+            "args": [
+             {
+              "error_x.thickness": 2,
+              "error_x.width": 4,
+              "error_y.thickness": 2,
+              "error_y.width": 4
+             }
+            ],
+            "label": "Yes",
+            "method": "restyle"
+           },
+           {
+            "args": [
+             {
+              "error_x.thickness": 0,
+              "error_x.width": 0,
+              "error_y.thickness": 0,
+              "error_y.width": 0
+             }
+            ],
+            "label": "No",
+            "method": "restyle"
+           }
+          ],
+          "x": 1.125,
+          "xanchor": "left",
+          "y": 0.8,
+          "yanchor": "middle"
+         }
+        ],
+        "width": 530,
+        "xaxis": {
+         "linecolor": "black",
+         "linewidth": 0.5,
+         "mirror": true,
+         "range": [
+          -3.952154604781554,
+          10.903313786058435
+         ],
+         "title": {
+          "text": "Actual Outcome"
+         },
+         "zeroline": false
+        },
+        "yaxis": {
+         "linecolor": "black",
+         "linewidth": 0.5,
+         "mirror": true,
+         "range": [
+          -3.952154604781554,
+          10.903313786058435
+         ],
+         "title": {
+          "text": "Predicted Outcome"
+         },
+         "zeroline": false
+        }
+       }
+      }
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    }
+   ],
+   "source": [
+    "from ax.modelbridge.cross_validation import cross_validate\n",
+    "from ax.plot.diagnostic import interact_cross_validation\n",
+    "\n",
+    "model = ax_client.generation_strategy.model\n",
+    "cv_results = cross_validate(model)\n",
+    "render(interact_cross_validation(cv_results))"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 325,
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "[CVResult(observed=<ax.core.observation.Observation object at 0x2a5c793d0>, predicted=ObservationData(metric_names=['corrosion_score'], means=[3.33038581], covariance=[[0.0086991]])),\n",
+       " CVResult(observed=<ax.core.observation.Observation object at 0x2a5c79df0>, predicted=ObservationData(metric_names=['corrosion_score'], means=[3.38045204], covariance=[[1.14749778]])),\n",
+       " CVResult(observed=<ax.core.observation.Observation object at 0x2a9d89430>, predicted=ObservationData(metric_names=['corrosion_score'], means=[1.57425067], covariance=[[0.00012101]])),\n",
+       " CVResult(observed=<ax.core.observation.Observation object at 0x2a5c79cd0>, predicted=ObservationData(metric_names=['corrosion_score'], means=[3.74291177], covariance=[[0.00775312]])),\n",
+       " CVResult(observed=<ax.core.observation.Observation object at 0x2a5c797f0>, predicted=ObservationData(metric_names=['corrosion_score'], means=[3.35092202], covariance=[[0.00104975]])),\n",
+       " CVResult(observed=<ax.core.observation.Observation object at 0x2a2ab1a90>, predicted=ObservationData(metric_names=['corrosion_score'], means=[0.75377368], covariance=[[0.35887005]])),\n",
+       " CVResult(observed=<ax.core.observation.Observation object at 0x2a7a8c8b0>, predicted=ObservationData(metric_names=['corrosion_score'], means=[0.03287673], covariance=[[0.01573419]])),\n",
+       " CVResult(observed=<ax.core.observation.Observation object at 0x2a2e09e20>, predicted=ObservationData(metric_names=['corrosion_score'], means=[0.28427335], covariance=[[0.02445275]])),\n",
+       " CVResult(observed=<ax.core.observation.Observation object at 0x2a7a8c730>, predicted=ObservationData(metric_names=['corrosion_score'], means=[5.1717073], covariance=[[3.27016158]])),\n",
+       " CVResult(observed=<ax.core.observation.Observation object at 0x2a4643a00>, predicted=ObservationData(metric_names=['corrosion_score'], means=[0.23590281], covariance=[[0.05287956]])),\n",
+       " CVResult(observed=<ax.core.observation.Observation object at 0x2a5b707f0>, predicted=ObservationData(metric_names=['corrosion_score'], means=[3.954885], covariance=[[4.13191409]])),\n",
+       " CVResult(observed=<ax.core.observation.Observation object at 0x2a5c79fd0>, predicted=ObservationData(metric_names=['corrosion_score'], means=[3.79301819], covariance=[[1.57803025]])),\n",
+       " CVResult(observed=<ax.core.observation.Observation object at 0x2a9d89550>, predicted=ObservationData(metric_names=['corrosion_score'], means=[1.57698842], covariance=[[0.00856682]])),\n",
+       " CVResult(observed=<ax.core.observation.Observation object at 0x2a5b32460>, predicted=ObservationData(metric_names=['corrosion_score'], means=[4.39157658], covariance=[[4.529502]])),\n",
+       " CVResult(observed=<ax.core.observation.Observation object at 0x2a5c79970>, predicted=ObservationData(metric_names=['corrosion_score'], means=[5.52922756], covariance=[[0.00146855]])),\n",
+       " CVResult(observed=<ax.core.observation.Observation object at 0x2a5b70460>, predicted=ObservationData(metric_names=['corrosion_score'], means=[0.97394723], covariance=[[0.36594216]])),\n",
+       " CVResult(observed=<ax.core.observation.Observation object at 0x2a5c79ac0>, predicted=ObservationData(metric_names=['corrosion_score'], means=[0.54739115], covariance=[[0.2633537]])),\n",
+       " CVResult(observed=<ax.core.observation.Observation object at 0x2a5b70280>, predicted=ObservationData(metric_names=['corrosion_score'], means=[1.38298294], covariance=[[0.0981112]])),\n",
+       " CVResult(observed=<ax.core.observation.Observation object at 0x2ab3ab3d0>, predicted=ObservationData(metric_names=['corrosion_score'], means=[5.12117681], covariance=[[5.41822595]])),\n",
+       " CVResult(observed=<ax.core.observation.Observation object at 0x2a48c7820>, predicted=ObservationData(metric_names=['corrosion_score'], means=[4.41398979], covariance=[[2.74877288]])),\n",
+       " CVResult(observed=<ax.core.observation.Observation object at 0x2a5b62730>, predicted=ObservationData(metric_names=['corrosion_score'], means=[0.0970607], covariance=[[0.00400847]])),\n",
+       " CVResult(observed=<ax.core.observation.Observation object at 0x2a5c79f40>, predicted=ObservationData(metric_names=['corrosion_score'], means=[3.74686509], covariance=[[0.00016273]])),\n",
+       " CVResult(observed=<ax.core.observation.Observation object at 0x2a9d85d60>, predicted=ObservationData(metric_names=['corrosion_score'], means=[0.12282408], covariance=[[0.01368547]])),\n",
+       " CVResult(observed=<ax.core.observation.Observation object at 0x2a5b62580>, predicted=ObservationData(metric_names=['corrosion_score'], means=[0.01725345], covariance=[[0.00132096]])),\n",
+       " CVResult(observed=<ax.core.observation.Observation object at 0x2a5c79e20>, predicted=ObservationData(metric_names=['corrosion_score'], means=[0.64291407], covariance=[[0.42318268]])),\n",
+       " CVResult(observed=<ax.core.observation.Observation object at 0x2a7a76e50>, predicted=ObservationData(metric_names=['corrosion_score'], means=[3.75959764], covariance=[[4.92338383]])),\n",
+       " CVResult(observed=<ax.core.observation.Observation object at 0x2a5c79bb0>, predicted=ObservationData(metric_names=['corrosion_score'], means=[5.90193752], covariance=[[14.30188832]])),\n",
+       " CVResult(observed=<ax.core.observation.Observation object at 0x2a9d89ca0>, predicted=ObservationData(metric_names=['corrosion_score'], means=[1.25197049], covariance=[[0.43754259]])),\n",
+       " CVResult(observed=<ax.core.observation.Observation object at 0x2a5b62880>, predicted=ObservationData(metric_names=['corrosion_score'], means=[2.22755944], covariance=[[1.44631305]])),\n",
+       " CVResult(observed=<ax.core.observation.Observation object at 0x2a5b32280>, predicted=ObservationData(metric_names=['corrosion_score'], means=[1.10185481], covariance=[[0.3645476]])),\n",
+       " CVResult(observed=<ax.core.observation.Observation object at 0x2a5b32a60>, predicted=ObservationData(metric_names=['corrosion_score'], means=[0.52667369], covariance=[[0.06076658]])),\n",
+       " CVResult(observed=<ax.core.observation.Observation object at 0x2a5c79160>, predicted=ObservationData(metric_names=['corrosion_score'], means=[0.81682378], covariance=[[0.00202599]])),\n",
+       " CVResult(observed=<ax.core.observation.Observation object at 0x2a06d0ac0>, predicted=ObservationData(metric_names=['corrosion_score'], means=[0.30203122], covariance=[[0.06494224]])),\n",
+       " CVResult(observed=<ax.core.observation.Observation object at 0x2a48c7490>, predicted=ObservationData(metric_names=['corrosion_score'], means=[2.92546491], covariance=[[2.91704295]])),\n",
+       " CVResult(observed=<ax.core.observation.Observation object at 0x2a5c790a0>, predicted=ObservationData(metric_names=['corrosion_score'], means=[5.5097873], covariance=[[7.68646695]])),\n",
+       " CVResult(observed=<ax.core.observation.Observation object at 0x2a5c79c10>, predicted=ObservationData(metric_names=['corrosion_score'], means=[0.81779914], covariance=[[0.00204636]])),\n",
+       " CVResult(observed=<ax.core.observation.Observation object at 0x2a5c79be0>, predicted=ObservationData(metric_names=['corrosion_score'], means=[5.52898111], covariance=[[0.00526972]])),\n",
+       " CVResult(observed=<ax.core.observation.Observation object at 0x28ecdf670>, predicted=ObservationData(metric_names=['corrosion_score'], means=[1.73056953], covariance=[[0.32395525]]))]"
+      ]
+     },
+     "execution_count": 325,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "cv_results"
+   ]
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "ax_env",
+   "language": "python",
+   "name": "python3"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.9.19"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 2
+}

From bb8183d39e6116ec599b233b3cd2dd102c4e88c5 Mon Sep 17 00:00:00 2001
From: "Sterling G. Baird" <sterling.baird@utoronto.ca>
Date: Wed, 1 May 2024 15:23:04 -0400
Subject: [PATCH 04/12] mv files

---
 .../mobo-tutorial-thumbnail.jpg                     | Bin
 .../sobo-tutorial-thumbnail.jpg                     | Bin
 2 files changed, 0 insertions(+), 0 deletions(-)
 rename docs/{tutorials => _static}/mobo-tutorial-thumbnail.jpg (100%)
 rename docs/{tutorials => _static}/sobo-tutorial-thumbnail.jpg (100%)

diff --git a/docs/tutorials/mobo-tutorial-thumbnail.jpg b/docs/_static/mobo-tutorial-thumbnail.jpg
similarity index 100%
rename from docs/tutorials/mobo-tutorial-thumbnail.jpg
rename to docs/_static/mobo-tutorial-thumbnail.jpg
diff --git a/docs/tutorials/sobo-tutorial-thumbnail.jpg b/docs/_static/sobo-tutorial-thumbnail.jpg
similarity index 100%
rename from docs/tutorials/sobo-tutorial-thumbnail.jpg
rename to docs/_static/sobo-tutorial-thumbnail.jpg

From 35096488fd61406c6db23d3a204022eee0f5bc75 Mon Sep 17 00:00:00 2001
From: "Sterling G. Baird" <sterling.baird@utoronto.ca>
Date: Wed, 1 May 2024 16:50:00 -0400
Subject: [PATCH 05/12] add thumbnails

---
 docs/conf.py | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/docs/conf.py b/docs/conf.py
index 82847e41..51697011 100644
--- a/docs/conf.py
+++ b/docs/conf.py
@@ -217,6 +217,11 @@
 # so a file named "default.css" will overwrite the builtin "default.css".
 html_static_path = ["_static"]
 
+nbsphinx_thumbnails = {
+    "tutorials/sobo-tutorial": "_static/sobo-tutorial-thumbnail.jpg",
+    "tutorials/mobo-tutorial": "_static/mobo-tutorial-thumbnail.jpg",
+}
+
 # If not '', a 'Last updated on:' timestamp is inserted at every page bottom,
 # using the given strftime format.
 # html_last_updated_fmt = '%b %d, %Y'

From 2bf724747417af15f648e618af572336de3437d4 Mon Sep 17 00:00:00 2001
From: "Sterling G. Baird" <sterling.baird@utoronto.ca>
Date: Wed, 1 May 2024 16:51:27 -0400
Subject: [PATCH 06/12] tutorial content

---
 docs/tutorials/batch-honegumi-selection.jpg   |  Bin 0 -> 1004721 bytes
 docs/tutorials/batch-tutorial.ipynb           | 1546 +++++++++++++++++
 .../tutorials/freq-vs-bayes-hg-selection.jpeg |  Bin 0 -> 1260172 bytes
 docs/tutorials/sobo-vs-mobo-hg-selection.jpeg |  Bin 0 -> 496939 bytes
 4 files changed, 1546 insertions(+)
 create mode 100644 docs/tutorials/batch-honegumi-selection.jpg
 create mode 100644 docs/tutorials/batch-tutorial.ipynb
 create mode 100644 docs/tutorials/freq-vs-bayes-hg-selection.jpeg
 create mode 100644 docs/tutorials/sobo-vs-mobo-hg-selection.jpeg

diff --git a/docs/tutorials/batch-honegumi-selection.jpg b/docs/tutorials/batch-honegumi-selection.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..31e012601bfcdb97b67434a4027f0ab2c93aee9b
GIT binary patch
literal 1004721
zcmeFZbyQnjw=Wz@TS}1@FJ9b3A-EPNXmJTriUtS}AUK7#xVuA&yF+j)xVu~NqQzZ4
zp7-2uob%oBoqNZ*_n-5=<Cz&*duGiwSLXWdIoH}*b7$XA-~R%<RFDPB0?;0a9gQ6T
zxL<i>29}aCey6S~3s#bO5C8y=G;K{`PLJOJ05CgeM|C;L*SdQ8uQ7fD9s?)=1OQUN
zYZFr^2Z?v@lmP!JolOEBdIA8(IR5JEKW+B^FTgT0cQOS4(4IYr#Y`O>ogaAl2VTk5
z+2Jof?tv#XwK6e#;0qpjcE<+;ANcXV>OcG&|NAfA^56KIzj#MYb*YE3y?o$bTmA?B
z!++pStsG$wZ5~DrpoiHa9&Gpr{WTJ7b2}}Kha>I7n*!hjPzT5XBmu7hV1Nuj3cv^8
z0th~od;ksrHvstX3Gkoj|J64IKm}m`plkY2Isn`OjsPgY5?}>z2CzTWA|84-e>hq^
z)SExNYys|n;s2obK>v=blLd(TA601Bk^sPy#rykX1^@sf767=7y1&29yuZKA0szqG
z0f6sz|E_QU6#x)8eW;K6_d0qc0Pw;O06?|<d)<db0H7`q0C+XyVB%=<kMBW$C?8o^
z002k%0053206;Jd06d5M8)gr(zvO`IDF8sz^Wjtt007A;0084d`xgfHa}OB(<#Ycx
z^j9bUH~Rn3z5fotdxo)#tAvh*2Y7^shK`4J-wvRDa8|S@0QA4S|9=oB+Ea{Y&mTX*
zLVxto!2iK*o}xW^g#P3O)^ma<XlOJ5v`6TVpFFg|#KR|`Atd3FP<u{9{7O^X1Rfn1
zpE)*uK&zqkDJC|da%6OZj$1%b#})Nsa+wsw!^^L(`@zW>;o??RT~k|kNJh^mY0e<z
zm{>o>n8YNdXZksdnMK+xuWd!f!ZIj3M_%DeZvM~d_6IYl9t=c#{2$GHFyp}n^e32)
zpFhYw@E+vo=#L&fefsptW6bA|*#3u^c=$9JT+aw3)JzTt;h&;2M=*(KD}OAL&~bk_
zB-V}_rw8#1NUA$J*D&z#@;M=5s%ueWuSf-@G__n@W8)L5>P9CRWi)g?XH7Cm%j%k%
zXXmuZnYsD>T>0W3m{{LFbw3NhMn`)96deyB4*2uy|3&fd<KGDUjlkas{Efih2>gw}
z-w6DTz~2b`jlkas{Efih2>gw}-w6DTz`q;<7{5-E4ttnr&XO2NW0a>}&M^1=1aBof
zt@nprx(O!6VJvT^+ZQiiuOo*5$Afhf2e(H<o4)RwknM|)mzKxheV*49No2BB`Nj)m
zD?&9<`1yp=);wRsfATWN$Y#?YbJKzjNZwE{WguG11OFuWh@cM%E$OtK_lakq{jYl#
zIW9f*68EaU{6*zHdDTkYs&{knLS)C=_{9?Yog2pA9sbK=NW><z%1w*J?jMRKajx#~
zzj<t|vy8Tm2)~6>J0pA1Cizd>y+?hRF?5%v@Xo<EUnxy}4CEG=j(q=a@;5GjJHy{W
z^7qW}_jLH*dzoNWRA|D_Nxpp6eMkF-`XM#@&rew5z`J;i8@i=!)T6G4{QjeP+gvuV
ziIwbc<a{SQSAJ9^Vo_NvcGZRWlw;cvnZE7fT!i&}wwNq_(>Xo_azf}nn!TJXzI}0<
zR&+8=l=iy)W%EeJ^htp1B3!g9YvOlf*HhEoo&6_hE<z5cJTAI2Dqzz{S?T78m<_Yj
z2IV9ZMRV6zAEl_IaH=hhP6UA%5-HF^fxJ=^-81iizK`0Xaa!#Sg$C1opWYX34a6QS
zChEQJHTR6H$Pq~z)Bm7hr4QcFHwHf@`SpKtGG2@zw@Q8d<gVJ$8UIi8{~PKYeYgiG
z>D(Ro?9gLm{5Q1kziiT@p}udq68C^-=U1h^mEQpny^rQ0&^tk!#ElsoS++2>b`S5C
zz1@q`+W5m)l4*!7?sb7Vj$9Tn(z*fpiQdA?*Kw|?6_gU&OmU_%=%xp)T`V$M5ZmLE
zD1BGyalO?!S8M=g8Ex#30o9|#DKSQs6Ur^OtpkZm_nT743}R*}o^eYG@RxsYOf_x9
z%b8o8r%c<Eo>au~8DrLu%}#L@<gO^-5k&O&T<R(Hl+P&VWmy}X$6X&gy0ee0Dl~ws
zQ>pzm$9Z|k5p^mggnDI?CC19G<;6|jgJn#$30p8ecd)mfT*5p3x5A8YJ&vusuh1v0
z;%Tdg79jk|h@a-i0Rb95lBxQa{jJ~a*MU856IPE0tB~+ZyUNvJ`c#C~*aGb5k$X@9
zn3OMh7S@DAr=h|4x`A4Y+6X@ujE_0d(6fRS^KnC<+UnG%u)kv0hMslQQ7JfOp%w=I
zVHAzWrdIgE8h45p!$HVVwFX(pq!u=n2?YzG$Q8PM4Fotz7K~&<PlZye{S8xz_7s#m
zmrm<&`^n=r>G*qeJ4@b<gjGFCp--8@)2dpl8)f{arLS%Fe2Isv(mHh)Nj5nfZa<ff
zqM^n1Jf1KVh)YRXt#Y4y-2%J^5GKc{cZG9oG!_iC2;#J^hG*uo$aOzXhShndYAoz}
zl9Jm`QM}Mr$q>db&048rS5DWX0w0<847=xPniU$W4NS&Je=FD=s(3o?`(N~{ThhCf
zsS9bJlDFLN|BI^sO`E@9<UHx{`Gav+_168@e^VR(OJ=d(G+k*@A4K{TO=D3!{bFAp
zbT@n9^J~ydzb%een5U{dnd$kct&6mt-l91zNI~5*1wA5{h_HQ&0puv`FH`t=N!i-H
z>!nhGK*gxxtQ}S)?g<NOy=kF$b<SXTM<kW^IA7Ue3tm*PA=jSwR@zG}Z*bD-P1SrX
zUEq79FlX76nLqbz&g6HveV4jaYfIX$uq}jO?02JgouNQchp8)w!@z7PlTR?<-8B^G
z#7z+)sb%thCv;;|J70scS}atVhF)5NL=5`zj*+EJ&7J48u)z5`bXbR8BqNe9nv=~D
zjd{RY5-cjn!ceFHW?;131)K<XPwkT~{NibyddKgX)u>m_lLGOZ{xGO+esH7|NRz4-
zIYs`eV*?b>UC~$>wHpWIunl4$Vb4rMZHy>bNEvEm=6FbKeC3(K9xLuQr5mMuKUo|`
zW+tSdtX?VNqr|0XmE&J=E#x0-@khI-AXSk8MaRa(Zd}kB3Z!p{pK$6`pu&zD!UGCl
z|0pf`HfuEO>?*@EMH3KO;wsqc=u)gVKQd&_WC(2BzO$pub2%KZbx!kWv{8vs&xb(I
zf-xOLL;}BT61J?d=hzE-8U<`~n3NLNPvY+~u1I-&kk2iO$jAQLHd{gmQ;}mPp1M?$
z1fMl-Ic?y{bXk76i#eE@785LJgcZ%8%<T(=wsIRa(reQvOpl8##%F|uX@gBPm1OK-
zv_%Xn*nCMN<}AUL`GZPvDGOVKaEjx-)yN9@{O7xgQYc7^SGIk_9cDfmq_nA}&7psZ
zx~XG1o`}!X5*`1ORon{Vi1?c@yt%XV9w5%PHH`p&iIQpjESrOP%0O|xnsjt@c6{lu
z{T2V;JT`1$9&S3D(Ob(W7PG@*t$x%)_=cM&fiC!MW2uaHaJ^b^j}tdpeB77qaT^%d
zWL$LOc7SE~(~^6@Eba2~X{qx)V98bS28HJGOT-gJS8%yVlAm<Mn@MuHp%-BiXH)Ur
z+EgcRmAO7pPPcHBtZz-IojX<VOUyBR38Hi|nP8v}evX+@g`_HK^v<c)lPUFQ^@@dw
z^2WK@3ev1f3MPWh=B$6`Cp$+*I|19!o@r-T$-O5ccHvb|oqD2Bk`8PISLfI4SS|U}
zbPtg$YFqfTRKs)0aO5T9GGw7(B<%}~yAhHX$Woj9yX!{d^$MKPnMeX0lzkJ;KxH@%
z`z>)2k&M>Dij&9(BCO|M?17rE@}Eq*qHc$`je|A~0$+AXu)@F^Y2#qbr~jK6M_U<0
z*h(f{TlRE*=tf}lAsJAkld<coYi5SN0t6F4P?{9it;)Fd^K^Sh4nr!}#BuFTc9(b=
zs@A&KmHwEY)w5FKl;zA(J+c;-BO8`ONw|6SZ|vlB>Qt-iuQE#;IJ5dEgep@^KYA7s
z>auFdDA^(|@W%V4`9@ONDq?#UBbDX#7-Y?2)(CX}H<tdBh|zmOZ@G-Gf7}B~qNv>d
zNsRtqnu$d&l#Jil(r|{K{qqYB{#t)B$Hr<$C}}{{>QJ}$1WleE6jOa3-ofw;OcrM*
zGfh!!-#38yOY`z}a7@5UzOy#g!Cn^md`WY2;L~ENl6DWdV~N5~+k)B|(R4jTWGWL&
zkySC24aw4ZSyd6?UA<xeI?>RkoDalr83eDOH`|PbrcYUAU>HW(sPI0T%F~!>tkaZd
zD<3!b9GYbbk)wR+eEgDLC`AOi6~y+METu)R)77M7$i>4yCr!BU94B|SGGAL~x-8T>
z#RN);qTaMfflkQme+3fi+px`}a|nCL0d00^B>fR%)nCl|{TlB9ynde@mV-BZ5yG&`
z&eXjfXXjTyg6x#c^!JLTgIr!Znzh$TcB~&dQuT8BEoQgdk_P31)#Z@*><|0*L-Sq0
zPIL0d^V&bDcsXCq@0aY~15{B{9oc%dS3n#Ha)oHNngA~RxtsTCV^W!;07rslUeZSv
zbi6TJj^KCea!6ca32=dC^y_+WLcs~=$8utef)}_uXULyyes>urMJ*%BSfip^d3`kF
z@hB6!B9V=3cA*Sa?Ez$@X7vl@<SG+u#hs7EFN$OGgX^g2;HUv^_B$SOmtp6FnaNJw
zkp<o$T^KtSh5WKfO%zFlu3TKd3!|o}F)>9D5Y=;P$o!{v)aI1eIm(+Lc#lB*g(Y(7
z>^8kK#Npd~#yZNodLYY;tncZNmP>fitXlYjhJ}o1pI5B(ro7*h5Vc5-TsY8$S3FwT
zKXfszkWhm%9UZU7mg~5vgR4*j#5E++@3~sjg<Pg=I&<#H*N^nu?qtwP+d7jnjy2`S
zNm<dIYF1&u&5jnwg~;{%)>tXIYIv7}8V?}pPO}XL^AbrPFH&?|maliIVv7pwqZVEG
z^9f`LuU<K%xGQBviu0?Fg+SL(E7AQ?wLHr-260w8K-pKGAraHu>0Ktx997;gOWTXj
zF{Elz*{?|k_Dq#RV{|&kZW(@n_~<_*YPmQc?NmSSVbQ!XQflbZ*BCA!w(TxS=N8U<
zeS_^fmTK>0?$*+ZS{rqwY=@pC3+xy4)nt`+a@f<A*|v35KPoJ;U@#rFFG`|koGhaw
zI@+aCkyKBdb@NsF>;8NI0N2MY|G=rg?nhDYR2YdKSDNHv_47|}H&_zh%@);m{!Z~8
zjSJmIO!qQ$ONKRm$)!h`85fmACS%~*fHQ3*&Qb_-k`tj$;ACn}2MnosHUXgxh+okN
z<*MqWh$TUr6@o9Dy|m-QyV-@cvBfciUW(1H${Kw|MV}{fLb6Q?l~mGDuRO~P%u6Z3
zfi{Wp{s=|5l0Aj<oYu5<+GhzNTf&P-8UEmE4YTDMm6M*<6ikbZNU+Fdlmond)sG5V
zna1F8byvPh=I^)Om{0y9Qg?!(lvo3QjHSG{%Bs8t&(cJ_0599Lnyz()V^E^?#MDkG
zDqnM23WX6gFm&uC^`tbGm@>jVR_()q`QVNMg<j$$yqK;#!Itm%w!g0`YIQ+TMR+3D
zP@lLA{APd0pZ~c?Y+(Fd&W5{7>B#nz{p?j|-Q$z#^~V>yyib%(xWBJg+gs8U#)h9N
z@`2;WrRf<cP3p6C-hQ_B{RiIvdEtZ07YKo<(8tXMY1l|y5c=ng<{&GeAi6VvEbr3O
z2(o0N0h9zB#aftfj&j=PuyF03)V-ZBw`7^A&N;<H5C!$89_0xA(V|R9vd<lc=FQRP
zd_nM?)`^nLZeUR>I+H&O&%dDMv^2x~8g3u8_<EmXkd>W@wT-#12?g)L3?_X#yIGhJ
z)r;Q;%TIm<`~IJs`p+c!m;2rGCgGR!CgsE}SD$k=P^n|(%0FyP!2il7x9O2jy3Ryj
zS&G9(MNSKuq@K4mbYomWvOBUa@in!OXto3H$b$Qdcl#I9H;ATp6-mPOZB`s_D2j++
zii7iQ-DWF_7GkEP%qBBy?Y4f&ai?ZQU7O!X8s`y=*QLHoEp+1?dk$%f3m7;H!gEv$
z=j%)_$Ml=CpCRRZO_TU4E_*nss~e=0H$cLPX?0`)yP>?Z;-cdWU}^Uk`8n_3*d@d1
zM#HeewT}@yh*ulN)xb-L(;&}Gi7<a>71JMwQl&3hDH@VDq=$u8X=&}4vDVS%kw!$4
zY@3t)W!2V@VTNihIF-{961e#1kdpcm>6Hg`%}zxP{!(QVx><6W(bj-`umQPrHcmbN
zxe#o2q!}j=KL`^Y^)|KAOAWAy*~Nct^YJA--1=BX&Dlx4P%8mC2}y^YKW^^GmG#yd
zW!-2m5Y{Tww1_&vU_>SD)xB9qDL|e<zc3NzlqIhg30@1H--vzOUyvQO<9wfzTv0l)
zJ*d?d>oof#M~_Nas%j_AQn6~c$U-frF2)tqNRcYN2|a1Hv(`~bVbR!vvslGjKDXJ8
z#h=mGvQtR6$niE7cUG+%+KWGJ=MQly8r_={oG!ZuoGp81UWbhvsOvk3+ykhiiWf*H
ziH2)&)kK4aeEW{bYO14>mHG>$%0NonWGq9S76x?DMDCtqODy+*7vhy1xyu7h-g4n>
zPL@KVVGfJ)`^$&PWK0nsl9hK0$V>vZR$CN^(9Tn|y_33X8nfKR&jZ+}vuWWGs&7X7
zqb%vbaBR9D+35>^Yeyb*!hXJcK(Odnl9bxWy{a}L+4Tyq(Z*tSYw~<kFxH^rENNwO
z!o0`TytY<H6W(&y7lXB&+@!MTE`qr^(YrrG=R-3un#}|vZ1S%7OwIFB{U&fkM@y|o
zaK=quR;Q;>g|F+IBry{7Z&V%{s<UUSbIdo!zng$|d6?VEzYF0qdq>2Z5u)q;Q+>g*
z?dJ}+P2eM58;|RTi7-=jK5&71tt^M|yR4O$Z=}BuW2>CiQMKop)Q+DW{04)_N>cG<
z!ZP;C{o3#;)SL%Q3Yy%$P8yM1Jb<)ME06(YH0rO#kQ&QL`KA~>5Y1b_=OP$ZELQ>X
z3}$0wS)J~A*eK;54NF*`{M9z_RX9I#Fs#$^Nu5;PE<a}%FFJl_b(-WT?;=N{+N7D=
zZ8JaqF64Qr8%*U3OKzruDwAomKbiYUR9OtMbPnlDgRg&0nWD<~fYXqjn^@Dkje{3)
z8ux&)#e2Z>&8uy+lc&EFuOW?ccMjVx4)Uxp{m}qaP^_~jCe&N}`S*aMa~%M@7WHig
zFVnd7EbShERrJ;y43HJa|HyC!Yy0!4+@PMrqOmKvE1?8kRgFf8dzDL>qr2Ntj~!|F
z=ri3Q-v~pQ`Yt_N!%C`IqOq00-lPh*hy?HFBlDEn*=tp6B;q(c4BAMyN_a@ySgoO^
zHNOWm7>kHd&eCka1?%CJb`jQ+`gSFW6W-nfwr<lDSz&|QN~0l%B8cQCsXF6Fkx0c%
zzQ&Hg<d_kp7_2-+73qtiLct+0uWgslmOHn&X0ugMYTcyp;k~=RBrY*sA+b^yyvxeG
z18!|k(YL|ZyI5#~#Tt6v*jsPdVTXu6M@y*Rw(~J?sjJ)L=&fNBss7H5b*u0(VY0Ur
zZ-Q3K#yL_!NsA&Noj$vqR_{bUtUoC160jb5^`==?ywMbN9r7pHcQL&gzCV@DGE8Y0
zT<Dr0G*(WE3s$36$TL?$>E2OFYvJR;?Jeg^D>nmmM3u^j)V1ybZZXmg3QDr$bD5Na
zn}oM_+r$!_uj}R~v$Y#&Kjd*~h0oMU$mmzENl(BGJvrn;DMcD$Ss3V&3oc!fFrQ9y
zQilKBQpq;ejo1QVHa29tqTX%WRk59=cyj+tNxFvj(S+N!vBX+JbQPI)S(jkz4LEg)
zFYZp_Xuj_qVNavpW$FI#Zg`U~8_7){5T^*e7}Zzo7nu!BX2IpDDP=SakbTR85#;l-
z<<0UvAeVhTS;R+1KQunAp3Ej+&5Y2Ue-dvvHrf)#KwqnEity$E+O~fB#xW-S3{#`e
z{o+%nY~5kX(p`-iLCNf7Q#f~yhjjdp_LCm$Rm2+;p_RCHCL>8Iq!}E-4JBae@qqZn
zLdps)tkJ?NfUIe6dG;qca)BGQ_#IpZOe)0;8Iw|q%sqy{u-Q3AN()Ks8x~`e=wwnC
z-|aDf>Q~n3dy3j2I~<051of=jUAxxQg~m>%LWvNAQ7zHgf*2#nX#)=*eiP;?cHwXJ
z544)~l)dw&K_)_q`VfVLT83wm$~OA_F$tqFAoYZ7Zu(q$Sqlp@X)s9b2(60*Z&6eq
z>0Us}7rWbd1*FRt7{7FmL42lnu`qKV@3yEPW3RL{%j|@5m2E2wCOC1Vhv!(NRwBB0
z`=yB=QA01%?e}9&n?o44dClht#rpQ+=6yG$6UXZlF@qhD=4!~W<@~S5j57?K1M)iK
zNtoypMq-{><DyN!biHw!OL?c7?LquUdva}A--A`tO`zE+Z5%x%t|hFVIr9y~(s^IG
zHofv73^i)RvvX0Tb$)VqdmH=cA3eai`c0)XRM1am2%(6*PI0q&;CjLyU;0`dqKtVq
zVlLPK!g^8l5~}CKcf~Zo%KZlARtRTXL<QIn8J!d>oDMOzzoC%*NJ35BRh;@|zIIO!
z+DYTe6J^&43E|%@$&=SlQ$vDRX?JN~lc8gTk>Ketw!u6T`s4h2Bm8ql412&D1qv}f
z=c<o_`ol2XJfjDy3^=Vl8j!2nc9Z5wcadhKlrq&U5-ra+fI@5C#}I`W*lD;tErK5U
zJ0_?}&;{_T@Bb5H@q^fDQm$X>_x47h3DrBYGHE9GCx1liK86gu4B_6jHu>rDIK#FJ
z_R|1SU#-e;Bo=)VbPcdWQm3!jfvhoV>3`$q&Toxa`j<WL9n9p4bJ+}Y&1Mz*Er{LJ
z{ng`j-F7e~mG|tXB9cBCLx_6Be}(IIj1)Pq<JJFkXW8K6Im?$lwHr*A7nYu}4dd*E
z=Llm^;B#|3x_rgF4sYD9!Xg;UhwVF=(JH@^-DGL9B(?04l!}i(YPJ2G4G%?@7EVXY
zH}<o&A;<+>dGj~+gM<PBD+At|y6|afmm~=!<LvK3W-z{d10EnLgTl)T<fXo$l5QVL
z?*#l25!0zHTC}da+ThJmtL2qkS;RS85U|rg?-hKeVMyf^J;wLMFR4tNZg5Q`?rDnM
zQiR39=^P04M=oX7iZO|ccjK_x2Nxc6Ngpzvk_x>^q}@)7<uj{sbiXJef0xbPQ@<tl
zslH;0l740CTPbbNS2Ln(zy7M}Xfv8Ew@w5LDNxXANpB319A$yB2G%mx5cD_BmBb7v
zv#`=~<Nm65`||~VDPgJT^g{}kCaJm7kd%I85bCm<-IXnzG(Vkrh;4gH$Eq#3qDbRg
zahwjguA#JA|0jXlo_#rj95QbgXW~IntkdWnzGyW^%25haph<PIOYg4==@@WVt=JxK
zT`3Fdm5&1}KsICsWFNh_2V`Tln_lOE#Oup0lW<!EnD()B_&KTtt|Cj_!vmaBn6!^=
zJ%{G#tCAH8O2Q+=pVBr4%t^*l_8fUU=Q(KvIxF?ZNK5d)|7)@Q=O>Y*Z(~IFfM)8e
z9<)v0fBx|R0M$o-w%+;t6um2cFR(j<TH^uPE?PB(F>00RhpOhkS}03^hC2%vcb`S*
zl))uOp^-c%A={Td0RgkHj>1cCDay3Q!9;rBidrVGCiymzF$V2lAe^7j7TKU|SQ5-z
zJDcdr!qjQ6tftl0db{9gM~!*0GLaiDYxg`=kby>q5|%l(-OsC`dg1GE7T@_LD57pZ
zmDmY_nkniwC1{8Dm)oPrS~T2&f!L!8y7^2h<y@&y?5fr2rhSP-Q~5cIu{UcqRZKJb
z*{egx$E)YJt$%<o^lXp^6+^tZ`^NzheHUdgd*dTEMI}qpeixWlDl}f}bhIQEsLUJh
zmG+rz#q`L)Pf0Y0cI6A1pLf6dZ*91KggQ?|rGWyElurtOefS{kn-Pe^y8kp0Z;<E9
z)A1X-fJhzQguBnjl!X;vwI+CREX4W)tUgHby&wJ=sYi6=4po@q7}K?9Ty^WQ%esbp
z1Ts03n|_<&GjTPqeQv8yl?^>T?cneVudq}0IQ-IJ8e?Gi;bJeXNpU_NXT5So0m6Hc
z>pZ6&_c`n;M$D(HuKMt3Lu8Th&h8@7u2qFDerFM0r<6}M`Vj7jhwH|klb>yP794qL
zD`)j<1_5;HhwPDJ$}ajhp&=bV%w4cc3pUPmSBRx0nd*F;RS%HAL<c*ZNkSNr(7P_4
zB|FH<xFOMK__e8PV`}K+qBR9f&OeWLd|i=mAx($8r$RW83zerbP@E^M7y%iVNLsVO
zjE%Xe8|YA=%eXl_SiL}1sXp_*uoO{*I@du2e%Qs6`X|i8Kz`N<5#|l8>Fd#z^`r)@
zJ-W{3s|{ixTv3BQ(r2}fbX+4?rRG%{SP-Z#1I0W<fc5v}!a<AP(55wU?Y=3Tp^>uX
zm%>nWUjz3sdvB=D8|-uDUQ07|+s%2!Uha?E?WqQA#aLhT!3{XAstCV??3uIckKZzo
z72wBv;tL7gP#4kidPcgmePQeUJv^7;6Hns$ie5)U&Y}qzVJ87{ZuKQ69_?gaMy_hn
za*1OZ%pcv6UJRvMqPSo0$j!=2jyCa)JH6;tJfnm%h;=NS7>Ul)mMHf8NJL>YR2R|^
z_=2WDG4vx=Sw+7k1QK2Zx@91Ao&Q}hO34V*rhCcPQ&%6IVv_J`1jelhla+#1d=?Q(
zmsO_8QiQ646~yFeRY79X=wknvLr(ec0h4UyN6L}cso%xaR3A2TJ#W~MI|YwFTQLz1
z9PA}3Tl0j<^7^M~0J36Dfo0MC1giTJq}O_ise(psMytepDzQ}t5WmFS`h?392z^da
z5Qyo-u|DM<;7o6SdpfIn6M|Nr8n>d@3|n~r(z(qTp_MlGytJ3Cni+w!>?S6XSH)Vi
z<52)=Y9jBl&M!&+;%znLB*>y>upd#`sky3Di011ubxq5|cD^`<Tt>*hC?p$~mrU>u
zmekCpu885KZhtFg)hty<?$iid-f=0K)-6MDAG3Mspxb7^rz;-PVR%T^L}F`v3SL)X
z?U$>pXBr$xP3?r^d5IGPG)^yEHv=f~Z?2LiM@V#MM*o4nSAYs>{`BJpZLK)h$=bs#
za6_-1nP&edu}&PbV3wk?66w5O-%O}FwB_Uao`qFJmobZw^Yy{&*w3}jmU2(^Rf|%L
zr9hn0*lbOv!yB5;mEqSOo@5xCf?BvzL<8Lw;g%3G`Z5&v82fte;wKCuO;_#ui3t@c
zIUz&C38FM|k5p!Es$jm>UX-HavSKK!(A-Cnq)|QCq>*{M&{z;dCRhzG%A4b?xInk;
z_{=*?A6X#K%8}eTuJp9bLQOe(Sj%Xnr^n`MTO{7Ppf;2nU+?2Ss$;{nG@d|7D4=^h
zX6uzhL&xTmpEYwqD!(cgMB3qL1bBs2?GeSqZNGQ8nq8`!9F(lRs>*Wn!*Nm1>6PPY
zH3~vWb2!e^9mdA=<4kHOUb+xu%?(eKN;XW%vU5{NfmsVnNx!cu#6k??r{@(sgfUyH
z%43NKh@!w54Gv$$DSOJu$UiEW5|lI(thR&v_y)KsOSe1KavC{8w}1uENgVgY<xAYB
z<FkH)@!>jx7Dd9_dAUVpvqXLG2Fb@5BwZ%yA_(iN@yl8Y(S9<&$W_Je?{smm;K*Hb
zu?H%DKGDF<OJtcM-Wgr$Sf&E572?)gLt@N|kY!OT;j(q}aJl6*^q_rC^6<KOyUW^-
zJv$@}E3*j4yQJlJJ)1hjINGe+nbRx@#I-!_!HM9E-3jhEf_)k+J+kipr;854%hy>=
zMZLA;y8L7lu#J>0+r}^->L0fI3VGjv101GJrZm@Oc}jXQTku}gs?eU3I$fQZcQfu5
z{5RY}G72B9(<UHexMHAuai83-K93}KCL>YzonX}HvUYg2T>OWr#FvNXF?)|J6iu2!
zuGti3@bcc;F{#iDDE*8w`+=fp*c;R2#vz%4F-%YxdDhnHYI}z*$R<oda|_NYLm)PC
z+=`{5c&at4Do(I=_R*ST7r-B;(7ZK<Ehd~j<v}>OY+}KB?<s%2TYtL%tUF{dziZ2Y
zC6HX*z)o)&w#L!v9iKUU6gQqQu3+Ztb!PKRu?G3H3LPs}sMEmeWj3F|d^FH>pewA+
zd=*kI5HM`S%N&H^c@6s2Q*FpKEW%AyZiV71_Q;Mk<#m}^RL$Vx8@M<_Rm`Q_Dq|D%
z+A#{_7pzh6B^tC5tFYt>0-XafS>>IKJF`t1^{Q~(fA+~NBpc?Z{E%>s(J}#<4OEFR
znm?YM=HsmZvYYi)Ru&`_Gty68t}`n88HMi3$1|`%$6m;Nk3Acxp}Pq!MjoYdqm8;!
zg#*U~uZhN$!Rh%)(k;^Y-+ruWZdP$AFjXr|kJ`;d((+JbIuN-N^90=k3dAXsyUizk
zG~l&#6&M(&OH<FpSEGIAzG6%y#E7`NwMgsPf>?P%SU1dMqeiJs&M3;wae0n`G|G*0
z!61b>Jr-7QG7F<zHwJ#IRkjk(oKhVm#}?X(G!gQlrXQX@zr^-pBF^qGT{LPa5NziM
zmQH7Vd6G(Q`9UU>bdJ=@#k54$B%ey16vl#WzTquyYfO5(jS#-0HbQR{SyJIS*063~
z84r7bJx-=NA$}%iA3Cy*FfNJ#FWn-&pL(g-;lEV8mD}!NVQ~5#{nBf^ZYtcg-sRAh
zF7$&YLST?rrdZG%!6+_eh@Y2`)e^kT_<B-X0vhB%QxW^Sl9-L-O|KvgYBYxSb+X;#
zc>$O8CJ}qB+Byek_1fo*n8{M#D_ewPg?jT}vDjpncITSaakH?q?QyI{tA4;H%kBA4
z8pq2rv#al_!6AhcyXxNtvb_?nV<rz8#t|qSE%`#&$HKLnl}IF5vU>T1=*X;MysKy_
zmX@CMG$y<D9^l~@bUQikPZ(Ff0&c@qw*HkKpN_CLV2n2rqsK=`TPs=urECcTWM*@|
z&To;DK%*UBQ(o_V`NG(Kax7J;K(QlRS^A-$^Dt!&`<v(hfz@qkxPK0@xmXI{Of~dX
zJ9D~wm;WXwTv`34I(?6$<mFd^%N6h1W=qS|kqK{-AMtMvo@E3&Hj@@(%*XTL3MRdK
z+hmt{X_u(SUa*^^r{0yYo%T^v|BN0CS2u8b{OvNiWV@5G#7<Zxkzrlaz(gQVjS8=^
zb<<+7Y}N}?1CJUhq52J9Q=7seD1M@1);-w`y#z;Nl#*#8gf#vtH&4j3@D1IMB(>$u
zdJDlM*~o!Jl7Rqjvbm=-Vu=$mDGJz$J<?`byDyTc%6|Ck6q6zKAo!x*2$#%_iuV!$
zeee|t{}=Stk;L@6dkpNAR3IEwA2uouH72j{GDbl@NZ9LGSiJUZ=+92UxBXYXxIu6K
z6@DH70Pb8qd2A1RXyT_a=tKVsLqfA$UmE>-5AgH6)&B12)TZB9v@^gAk1nbAe?Izn
zfb2VIw4OP!$g~ETg7YU<>7bZ*$DB2WIbirxS>rMZ>i4)owQPY@ZMC0|fXapB;9;+9
zn$LYoKN}MBo$+jGX##~dR7664M@^2EG+h<jGsSj0#f(;|;rXtU1fa$$1id^3YPOzY
zMy*}J0--k3M+awIGemW&lSBhzq(Ysn^ya6RSbnpXty?wHHT6H=&6^dZn4m@}fh%*;
z3u1nwMw)x82%bMb8i(wiW)Nf}xtcQ&jS}uYOO7fY<)+RJQ|+z&#6^S}F70X-*9pG7
zI>89q(|PSb=gQ%SoWCwPwTgIBIh#r+L)~>{aHD!!I`g*)dMdl(#iBQ=uQA6jZhm9Z
z4+M)@sPtpD$IH28er%Bt^!U=0PT0*>dMLVR+vhw@q~&BE=eG?Ub%b?6?*Zwkmjl6d
z%G<<KulJ=?;oF02@JZV|vt%kxp%%aTCV}B3T8FmeGFngj<<Nw+lppcCq1qK;M7o3D
zPm%97P3GVe-MqT{a<Qf}Wf+)Ys%$@snR=NyzC|h`O>81rtEzrNWdviBLBV6TAUW*t
zm_xDgy+((Mc?+%vjtbSWbkY&qfFUJ#S{t}(-0W5Q(`<Fq?c)5#bh+3rS&Li8t@kzR
zjlGY%q}3f}lI%G!F|}hW?^yWeD#{k5TNYpeku>F_&_<O71mmKkV01`=_bN%GfU<%0
z;~a`Fv6R&3{-lM>8-DrG$FGCSdGzu<&TBc4+IF6N^8_iSNyc`?c7GHygMEjWr{~EH
z#2-(jOsi*tH0x|BVcnC}dfR47Vv&%^LTet9*GZ+=Gba>tW)STO8)e%mg6eiY!~C(z
zMzZD(A|5W>G3m;w97%!GgXE|i{^uY(k9>O0G&6)je_WYKUq<RKKl+3!sHc#;t<!OS
zOcCL#{Kor7$B<6QZ1-_2#KC|`u(mPW-||#d#1<T|7sgyp(HYE5Dy2dr8lw7E7&yJR
z+5j9xb-7*X!sa+$j0MKcf8eQzh~P9Bp~>Q9@4-|wlR1^!yZ8<h_$$O;X!zWIc(_M}
ze)=BJ(2TxLcn>%^|Ms)tuF)5<{xj6yqW7G)s#Ij+)?`ePB2xHULqF}zY;~d;0@ggB
zb&~WhD(M{)7g1Re$V!ffPc|DE(~#^NeGm9e;xKpHvG?u8(LG?)mzFME6~{0Oh6QzN
zEwGZ!Q0Z`D<8W*{=t)p_4BbhX@7i<NlR~9pXXy<@Q?j}sa7=Tq-l=>|;$e_ZB$lPv
zfR&#~Tj~61$rt~*9Yw9?39K93!OhN|awROw&mYR1JjF$Liy5|1A%Tg&MT75N$b}uD
zE}{KI$e1x?(rN27oo?&o^6aG1;L70I{dfcY+y0~5AGbK7R00OUZN0q4O?@fKgS#B^
zJwk?-v~)liLUvSZR_WBc*;j$8))m9OMgMs0YG!dr<hbj{t;BErsm2>o<oK;+V0GU#
z`g$^a-2D%9{@eJX+u**>0p;kQheX8o)oZaYA}@R44hzE(dz-_&7EI_!@oj?MJr9b8
zi^jX(2QE$<=rIwzODISK=TQXSJpc%QgI~B|=5uKD+h<^x>5hR1Llu&E8%ef&Jz?qd
zqv<6z$1iHxGQ5pthl$z3DnlkkWlM5tBYshWHSb5)*6TNK#e5*4g3srvnpjOp6-CX)
zkK^Q)Suj$K5bYnOn1rOwFPZ=bWHQ;pW#iUt<law4!`m#A`i0G>Ao@^ZtU^*zLP)!G
zC_^_f$f37kVi!XDI7yv^Sg|lEDY640&uD0_fSoFCbt7F$D{5?s65Z1M5d~HX(l{|T
zu#onOs%ebz;0c3v8qy2+3B>3Vg5j8Yj_UH=qLd#y_*;Vyv^@nD)G-O@m?ukHtXIJF
ziBT5WYAcVy^hX=kAX-{FvG+pJs~u!3)d9GRMfq*<PkyI-%I31mR5qC&dp+}Gw7)Q7
zEt&Ep^WrGUvx&?*Fe*_WQJdMG3*|=Zo?O)TJ!`RQ(pKdARj&1hzFUte2P&5pGH9&P
z-*BwQz$p6>UHtWnLMYxph=lfGhn@BSk0dk-SEPB&=S+9D?lQ2@_;6n7vG)P}Z8hxK
zJ%Dz5^Cqv~7eUbFdRC8s+^m21@niS>h2M0wjB(@0dF>3is$C5nqRLdG<`%N9rhD-8
zdEG<bRk{JryFL!gdw`r9@|aB=tz67<?5yJuxtOo<ZTa+<@Fd!?&<gdnK%L_C(T4cp
z6TsG7|G-@*<<F1z0IqNA`#^N3`t8Jf03=eoC3K$t9&lGb<MRp^{ZRErBitu(>mIPe
z<?F5Hz<Gn^e-B7;+q|KAg-hBz+~afFZrdax#Fh8-<IwmLQ{6o6we(w%WpdeVyne^^
zF#pZ#F!Y)x)g4&mMcB7<SAbX5Y4tV8dh<_u#zP*n)YoNo!^OvD54kj@<}=R)pnU23
zuW$|U@Vw1`=|9b^zX;Hj3f+jksP5-1R?bMB>e@OP@QP?D#fq`oFf$Cl`~{|uPJSJk
zW#}hsGW8x;N<cjqV<Y@xhNV?O!F*-Snr^elQYY}FFB#7z)~_bnEiA0b>$T7^TJI*g
ze<_850OrYynsc;t7nifUY-z*h=CuABqt}#bz8iJnT8zBHTD(uq6ND8_ZOHv)4;Uuq
zlu7~_f0k()jH#$6zp)VNT?)GgygS}ym^6?IhHk!#$2ZV><+%i6;`ZPpG7Sd~P8#BJ
zz_V~xwGtk3fvdx^y)N#koBcB!Nz3l2qN}YLJ2fBoO)qnPoM4x#2oN@36^v%20GH5f
zJR-+BA`^D*qFXzesoRq8fzi~Gv05*i@sPcXG{iN6XX)J;JGlz|F4l?a&(MJIwAl<1
zN<x!yGvNt2d^vs>FmX$My@U?bP{N|!7y?=pTVu$c)JY_P)cT-%qHPA{W&2&WE)&sF
z`H@hQt$sC6dWu4mhcD0oZZK{#^l|QND|F61X<%YIt>llM9;)jNY*ryK{O1IA1l6<S
z@k^W)J>>{G%8ecNSMd`J+Dh`0b1Ah#b~yvquPMb2x)PG-#rWJ=*cfv6(8t#CgD#_w
za!n{M9~R%&{5gMx-Bawz^p7JP6q$8@a*OI2pY6nrrP(opV;*ij6&S0eR<x_tU%hh^
z7z??I6A0>+oam3q=3YZz?>~%5M@V~h&+A1e`gV_7$end#AVjbp_Ku}9cbyb#s$7bG
z#m|OgF!bw32k4o_*UmhmLiOYTOB0GW#VKbvry0&Xqm$Zq$_v|RYB&^V(L~tv$tCCE
zP#9`7x70@MA+0Mvo?EKGt)Lzkof2muEp71dOn~=J${T@*a~Y6m=8C>$kaY%5yk$il
zX26nSN9BA3YGaZ^)PkTe{)}MBy5aH}G$<Cvw$f??XO$YAk;qR*zFtdwTUu<zJ0!;m
zw5WursTRGWOwaYpJ%xXr`q8sP-=9e175TN_g&aPm)qTY*8QG4$HIT30m*dfew$*4{
z_{zu!hiiB@wxf!yFVIvjdQ?iKFBD%YCX4SdW^{I~YJ>x<u5JP4etkJ`NHm8SoaFZ}
zO)(X!>?Xvu&=E8O9O**d<-d0~yCl)wbg0kC<)kg6|0b|rSYD`f)gVtxALH&<>marg
zo7-g2RW(p&2Ru^Y{A6gxT`ej;M4!3kT)&VrU8&pBX-8(l=7DPLnD%!v7R3?h%21sJ
zaVVE5aqG$t5iEj)ZOZ2o5V@(=<0-cq;{)oM=|Oeypwe>|E;7XcT;74SYf^0Pz?^SJ
zYatg0uNVYs6(FG+R`$t(B(L_x;R(j0tFLY_wd-bxN}lU`Oe;0enq3PEDogp}gK!vB
zns0(R`Q~7mtooBxv~-{~^etHJE;`<f-7t@i((x{myqyL*h^eL*J#RGswy;1}%}jM+
zjlIr9^o-PN@aShxy`LgK(maRkrJ3O)T*8a`%?s`P!zc5^Hu(*_vc|y>{SQIK-ORbP
zFoSiS_ZJJcqv0&y3Y@ctf*C;}sywG9rSuju{XbQ22KmaiKBjKY`Q{HD@2X-|^)Vk3
zbJfrA_OsfH83miLsFV5ap6i-Skuklrwl8I%en`zG`);*)=uhN$nAF|s1{MfK2doe{
z;&AqpW|g;mEB7~dXgG0svz!j|j6AoUT<fKa7#o>}$BW1=>5bCSzDTxu@pTG=(s`tK
zFBF!qRD=-5OfGQSCjl01VyZ|PEUG^3Af-IXmC^$JsT15BA)Q>}=`!C}W-9^xjtTn`
zJqYxKHq05F*SfUKc70e8iu43&ym4QfOYak_X7zU3g-C6K`PK?tfvR?<utGzHgjvmp
z+;y9S$%<j<u8q*Z0gj3Cr{8%c-TCQ!i;b-L;GfmcKf4gxtV_epc20?T9n<`sgZ7%}
z6Q37JNhhqzD}I_JEMqqi&KgZ$sng0+sT@!338ZqKDy85_1ew%hewkvGa7onajzpSJ
zhdr(TjX@q>W2cHp+L&|Rh-9<QXN`oBeNjKs&Q{l2FpFBHXuPr_9I7vt?H$4H2`AkO
zGO_F5x0r#pi`cw1PP63!&KbISZMQ3ZB$(8hHRX%cwXw<0wfU$h@NhFsRAJ8Ra1|gG
z6Ff?nKw3egrDp;aFO-89I24mI>;0rfYY);zHki<hQ|cU-2#OSI)X<=5mRaS?;hys|
z(!8yp+}Yy~kLc$WmxZ%Y8cob9(~kERJ$nff@#NDi=#51f6D5tYvogtr-FPxjMxYAA
zVsLO#_T)esHVNO2rf|NSlQu81J%88qU@;Yp>~a43dF)OPBEn;q^D!Gs-NEt_MAb(I
zC8Eh%GJ+#bRb4V<loBsRcKwBi(PJbEmPx&4XmAgRn~B1$6KLByO~^htoaZo9a%EJ_
z?)tU9CvBc9a^tt^0?!u<fJl&qL*J)4_Xk}u!>q-AzM&QhNYhZ^+x9V-9bR$22iW@K
zha%@*Qy#z|;kZ5A-VM<8*1{bQs<drB(STt!iwir+{<FQrXm!o@PnO)#dD*S0LUSzG
zthOZj&XdZmN}@ey2CpgF4<i=ulK;dyPM?|PIB?AU<k?i69&6m`OHd%cRE(}N2)L6O
zw_g8P{^-$!i&|DKbFX4>guTF=*@`;)tOMWpfcU8Q&B}~QuK9_stiVQw^%HKpBf&lG
zN%t-2b1VlMAY>x};-NmLufIbWmz;B1;OqLuyHsBybmdL;ZlJ~is~774uApArU~i#b
zb`4`pZljh9Yx4Y=3SS(D3hxv9#?O)i^1zS=jDc*g85QBwKj`oVI|ew$>&_}1oj&$L
zyJ91zUwYp^dJH7Psc2~tf%0L)RX_#c=q}o*T=MtHbh<X4hlTVQIPN7(GVtd#g>rzA
zH_G^JK{cwx6{~9}HPLANJV_}>P>ek&NJ>$o$UB#Y8N|IoUflIXm!}H1>y%D$-5QsE
zF&koNvahqu()4*>=^9?D#;~3)rP9FVM=MC)S%||UyEH-H@9U5q-a2j@@>y}8A&+iB
ztM#~8Odgh(|7JCU`Q~vP4P!2KP-xPN{JgN8ys(W~8L7mTl>ARdri20+;+v;y6j_;w
z6SI(R+v1`8DHhD4jV%Op|88@OGe%=TRvJABD0G#c-&pz0rX*1r^c;JYIym>rTZ(p*
zyuxCzl7uLzUhv?#nUe-koG^!LGS)_4u|$=D+Fly;BKc@y;mxS>$hLHg^59_@KBH6~
zOF?d8oW5K^u!)%td{MwwAt7%ywtb2p6DaWccQjJ$J7uahV~bS=ayh8?SXp;$*Pfy|
zKb0?HWr!`$0%whHn%1C5PT@oIS;QCj7|28iLrrRFOq}eBdtkq%_U`++ZB*WzHrN6N
z%gD5F;XQt(qa5sDvF4)w0c`G_&WJRbT!0z07GlxuZlG5oK~rV=!iv=iJXM<-4n^KF
z+BF-KILoM+CkWW_!Jf&=z+?*<_SCeanzDiGTu9QbO$0|S12ZoP9d7#De~s(?fAMY1
z&^E~*JycLo$bv^tl>nJbe=Osr6OT&hS6dz%G0Opqlp|geZf?EuWD+14x6#*;;b%~a
zP73F$UBTAxn&jy$P3bH>M2B8b>^3j`xU`K|np)0KPAo_vC{9+I(#3aOHG(eGGE@~!
zM{QZ<my;tWOulPu&kg%KLKi)~3{<O9{AS#yT2)}cp=pspzgc`Nqlf28-_r+j%i(`5
z#{HTqvcD(!N&zH|w3j@F<;cy=%2lN2v30Fnl^31?oIj)c+w+)&rkBmvS3pt=P24i}
z_@2a};(5ndpsrGm29||lW3(W{o_ZDL$Y`!%paQoVh8hJQH^X`w+gZFzHE)sZXoZ`s
zcb?^@EN;k$gJk4qG(4AUX_;<cC-0i6g~W|_=iMWCh3^iPciX^tE+?jI{DHGZUGn{x
zZp{FuK7aIpKR7K3>^DqxH(1Eo5-=dj_b<1n$cBGiE^qOk-fw-HYxOiN<Eshw)}pX<
z)8}zR){dTHT_(CFTKBn>O@Z~yQ;VA9=6NDnC*-(DHXZG=kh@PqgpvD3I(K*B&G&$V
z8MI`+tsjxW+F13X&j*gu>xz)C+G=B!9MdAEPkkgd;?19mr>a8?rYVF5-X)u}XX;4D
zzFwJXOXQQ_pO6l_-ht%?BHT4uv~^p-kSDDunE#V@W|;Hvlpd>=a{yi4q)-mu(!-<S
zGEze&@rH788R#t}z6a?2AP>Gx(2AM<IY*l#g^V&72&VhSx0JfGs#qCw{lb}g!eUWX
ze=NvYVq*>=3RIzCqXR$rc8i|DLOtJHetS#)unsO)J#wih-o<hy7aHGl>vpBi#mFa|
zNgxOwH+57M9<yy4Qe>TL0>|nxUjR*GyMyNsbGWNLek6r8Ovb!|Pvo|Lr7sVf!ZBo9
z!Tcl{h~TLR*K1FT{f(}^BEeFwL-<88a&$>TaU|Y8QW=h^G2)q-zaj#T$C+GEi1`J-
zEV;qI2!=lC?{+wm)Gm3)b3HJ%o}%oe8rN23UJX*rjV!WdA8b*u&&rmG5F~wCORgkN
zBd_M$c|r2*j!r1hk_uI9?GY<CB{=N(u!gB*Xv;{j#YT1~rs$30)yy~<OLEi8BN%!?
zFPKMTanmjg4-~u3LZh{9P`~z>5hR8-sZ5ou3bJeRKb?CO`#F*cG8IzBU8@YzqISZo
zoll-|Sr2C`8%RaPR(kQ-hfU6$enZT?<=`BqU_YM!UCZ>cRSVP1Edyv>P_lowpSYP|
z=^3Fx?R2_FTHzwYu$Z{bUJuoIy|#3PA)+qc;BI_<#dmxUXfwk=YW+YDVYgA3!QBg1
zs-yDaPj~$$wxn7lrA`jFbF|kQUunby7so=?hZ5;?Cc48I3Bvhtl{f8%ASc3{b9a$c
zUsv5eE*{=c)28VM=GNDxH)v8caaJcA&by=R!}}9QKG_(YyS@!)MNhZXOjbZ7lE0O*
zJi&F#@o_x*c76ieyaxny<7_RL{<IGw_Wffld1|j+gQFXeu{=Z3bIB;_k(*88Z;aFH
z`K-V=UAHr9!LWaa+C8tap0_GNwl!unVYD_!2(py-!#VKdm9`u3Yi#|Up4@^4&$U|7
zD$ngCjoq=CY%X_gRjNMNTFR<6*ZT|-H$s;E4flg?!Z5s}+yLw1v(?1<7lF<s*(r93
zWFTWTaI~o?cxWIbmqjhYjua+T&Gku8=}54ov#_Q#^mRfT2}o_c+lzOTa#^EYgC%1C
zW--DWz!SKj$m~R$6B_tkVcaWdv9`~0RIAe}@^q*>Y3vzrqCRF14FthF7Qea&gjyS6
zlw3!T3$-w?)819H6#Y2aU*S_sZ?V*1z|ZQc=~*VIuO)bTEWZ=4$j{`WX)e@j<hv3)
z9%xu)Dwj-o5i@PAq^HzGrUb>w+@P29ehxjOhMMIqKqtrZ=3!8EVt*D^I&ktU_*6Tm
zFLCksraNxFZA?}j?{{v}s?wa_V|7|_g0cjR3h}f`O6o>Ze60_gaO6zN6K_im(-c!W
z2U0GgOZB&>C}{a&{pIisKHoN5<BlLci-wkG37IX7%5)as*CJ^(l8t1<%1NM~56To1
zF(W<$ZQ%4~??sI+_!yFRWfrK~5sHjCH4~ze?fW41V$#k<!IKNk$GYvGWYl$7rNq4S
zMTcv&N(%b(Dhsh0d2NnYYh3W#%<EssMAHRF1UL36!s`Y0OZyUP61hJJ)qG<ONYq&a
z_a>5N_{=qbK0sQz?s$T=V%6tX>ne1sb6ZQ(RqYK~9<M<sP-^Lce00CX+N`Br)1c<J
zMUjbgduZ+ZR~H!js>-1G&^NOQJn7I;Ea(`A&0$fKwa2aGPt(@rRs?=QS>7ZQ;}=r9
zIO_(1wlkN=s>thgb{^iTM!eh42XQML1tcZIlY1H(%m$7MrY+EP3I9)O(sJBP45Ic4
z%7q~(7BLpJ6i<}gN5Od@&PFTA&r-a9`Y-J4(MC#1TxD6Zb}~@<a~qBb6fd7iz=v0z
zY`Q1dng>N`h&|~ZIQ@L}BZ!~^JBMk`rmlQaU&LE~%%Jcbp7Z}XZXcKLn+DjOpP~m?
zbn;yXTm4wD-j7&s47Dt{@bL;~{_M6(zC|3NA-=hnhrl2~?LeYNJ&J^j`Lhu%BoN!+
zHhCG0|2YV;&J`|L4*9xxL8?bs;kB%-MX}tSd$#$mtHVs!5!hI6F}dJvlkK!%yHjkQ
zI-&Z0k^hIiw~mVI+15qLk0b;S5Zr<^1P{`<gy7N;+#z@xXk3EDac|s%HQIOsjU~a|
zx@n-X;10oq=CaQl@11k*k$1-1W4wFr9ee#%YgMh1IcBYzt7^{gyS!vi$uJ$}dOlbA
z?3!EbxCy@Nv<}}~|H31Tn#;ygzFK`yMkgjeA0DPb!v2Z7qWsn7{_?4=sRu)mVYS8@
z$_*Am9?iD$Z^o9tM;%Iaanfl!do@9{s5q%mk}(<c8gXeK>o+)|LDBD^XLo8S3ZX=Z
z23jkD(DKojmxEW5u*FRuO?Y=H*DIx||9NcxX0Z4|3sZz`XHE;4-=15T+&7W6(w#3w
z%z#~|rb!54)v_CWJ1WfY*0Sy6+`wFnfkozc`8ut7&xX0YGjf<z&1LR&_M_o{v+`y(
zO#R4=(86nCq#gEzH*x3EcO%%y+u9HZ-|{Ge4$9%kC=QqQ)$&RF2BClydeh4^4|X-L
zR3(^U0qgV2NB?J7!2gGDdH?*h7>o&eem%xa76Ta;;S@tNxal9Zx>4qFpD!&Ztae&(
zWNCrcf-7H!^6;?&yoK_skN_^{proM0yEOnmzVjOWZqgI~@lUhOlnOoLA@4Lr)rPk|
z<3%eT#DIsZQlI3#vpl%g9Te|;*#2T@ay}FqCuUu~>6Y@vhD16A%cks%=;&EQtH#9m
zA6V--QfByBeGKTrcVp%Gf>vRKfSPt3#vA*UMicNfd}QBnhWdOdsdm_I2Pp}e?R7f=
z8*rf_OKLv*>);R{9Y%I|KSP>4ZK@b)>gz{1D}b)WHyO!YY@g4Xz_<R+Nk2_-%PwG!
zrC!d5(jQ=YWAHK+GQMz9dBuq9vCb%wZjI0s6iTNEZ{nWlbX*1<*0M5R_k@!OP5e^P
zC@Z$qNP9)n1HU)JNWLp?4zmD9ZHmJt95k7CE`F207Rd~ET|xAEn5vXzEXl)~@jq^@
zF-;*F)m+DO_Z=UkLZ^F1O)Jy91hLUlqn*iMGRpfY+=a~pt8Fax-V;#W4*`{rrZUUN
z;qkAHE-j$DW;~Uc3O&`j6pxA>ivvN;K`KMGJUX-dvC>MOJlhmdqa@^A+Dl|gNNM|8
zE|`NHzo($#_@ih&-C~zbpiAlSUs&H4OSznh;jZc(Rzkl{()V{|zLclB*(6y-fA<2f
ze-s&fYkb=w{%L!m`pwzBsHNt;(iZ(b#>ALyZ}S{iR#VpjcUYSFUsd6IALmA^V=o!(
zCc<M%gRObKTYod7S%l^7ZG_~EZdaYqeZppFEtKpC&O|#`W_R^_KHKlgW|I0<omyDM
zZXP!5pc;ziv`|X`m`7hcRt)X-(Kf4XCzAR&Jv-}L`~8{Etk}G3cCjUpS+rNMLWM3j
zYtFRk`b}D9Y)jUfXp)<xJ0rZHDE=6w`-NH_K{v)|?!JNA34S(>nL5E<RY(Qhg=toJ
zI>&Y=;OjEL?R&_c<!dN-7Ub)xDTZR>u}liC>{;qY!|e31|M+`^o4UZt06`SI$=n#I
zcLUW#Q;4G&WqN*Lcbh-mN>5`Zj_IAgDN)5@>`sqT=N>u1&LVLW7Rs|y>^qW_RZTH!
zgrGQ)`Pgj2!9>JP;p&ZvpR}*=8=gF|X3n9Uv{iOvKzk}HfR!M5eBA*m`<-EK|J<iJ
zz4gYU-)B^R3O}<v#TtzoH=yVSMJ}HzZaWEX=f^iPnUOVL2(9(IPn2R3D>zSePyoSn
zl90Py;Ze`<bmhR-ppF~^2iaB|hm0+g(P)^~-VXTmxVAOpo@5gI)snCQdF+#(3j{!~
zl3p#t^d*|n^~1lGz(0_^{<S>*8{Yj3WG^B%D7Z3vU6+D~?}rCNhOQY$BI^}nDIzFC
zTNZR$XCuAwqFJS1hF@tyc4UkulXYf)(i1}>s@4c$XnZ|6A-V=W6R~sJ7%duZKH<rg
zT?o23-n&;hLyb>DkK}b01-SgdQFgXy@?Na2?)3=u7&IX#W{=%8NETWe<hl=KPwd(f
zS5#EYzYC_jNpL4wstMSkVBLE6K5Dwdl(8y1V#l&<oS!HZdSo1&!fFp7(|YdKv|@MS
zr-ONhE-ZOS-!DDVKIU56>L*$;3GcH-^0WFqh@KWpTTW-reh|pmRM($U$1A`BS<&NT
z;jL7YT)`W{nraSxH^VUCvb_xMb)VmMh6FmYY$Q;y7WJ-mc?c*aGGL)2<omf~CzeU2
zU{Bf(_d*r~c~t%jUg@2WzV-7z{atYhJ*P=bjGzYEpE~O1#PnRqNbH&hz*GW;t90eL
zy6UF-_7kxNn!TF<E=c|FN4!JKpK@x!Iq6FQeo`uDa^;Q@w!E*9NgV9a<d?(Knt2T_
zlw{dV!6iqFUa-qyv2u3lpoybLP&*nq_9aeaMkDd|UjE7m9kG*G{f5_B+T!+r{Ea>b
z&rvI@o(mvom`^e%myK1eu!SIpGgl&4P9nGW+>6;yrYH4aFx2%7hzCZlXS_BuoX=he
zc4njVZTQlN56Rd<g_k!)UiWT{*_)2W8;p9Sf%<&g1QY@k{bL(eeNJXy4g?-EUIo5}
z$Pv7IlDXIMEBkTFx2~G`d^=-4sY5rf`IjKs((R$bA%%)>cvL^8xM!(LATUFfhYq0E
z*Khcljpy%wR(UagxH0<cX2Eh~@x8}{!`$2VdsC2{x3eq1cZXJ92^{Quwx;R<Ssn8$
z8WuVFHp(W>CIkl{l;U@Yr&<!c+lOTXkv$A1NpInWL$S)cMeI=a!U!%kGICu-HCFzP
zzd4EUk0@G&KX~yUPqUXPN5APHfxadUX63*ODz8E;A_J@1)^{Yywbi#Nz;BoL$kPHQ
z@a_&lO@?axGUh=DK9?GRPkoQ=QB%XQAnl+pF)Yf@xS|=65i_;t9f`gFdu!vAOYu$m
zUbMdx>&CDnH<=f|-IwECV&rUCdiN|QbHQje<Gf6QqouDh7ztYb^$d&J<<=-M5f9z7
zu5pXEoT&|Ahr-k^8BEl+I5d^&+Y3ME?K;x*Ru2$lnNbwMdp!o8X&UKGpcNg6A>K57
z9Mdu6yxuG@QwTmQdl6u?BWP&od*J_sfXiGUhcdmx>5MxKfB2v9*8azW`tO%>Z?*e2
zt?K6IT|ka1sr29ahnL@1o8|bp(xBZWLC5CwnTbe?X5>OU6v<C#SPRi}DO8S5&Y(B#
z=QvQE2q}qzYWDG6iy3%-(`i0N&8vQ|tBowmS6f1c$|U(oVs-M+vfe4e&B>y*B7FSb
zAObe^@VM$vH~*d>?rpi%N(WL^Su|(|gJKOD@)c5yF!4@2V|D5nRr5i-6&3ZEN2U?n
z?$Hw=&;HgfZ-t&&y{&YPkMZ4kYpCNJE%AQLa{3upf;MM5fF8~u83l&o_e5}u(QjX>
zvX`J06@xx=e)#vA{_pqf^f&{&P?8O4l<*#k=mG>GPyCOz$d0aF$h`MdJQaE0+IaDm
zv7Z7xetWdI{=tu3TTkU`@O{R;xda&y^8w%O8@zuM`-edO9Sy<JVNG?iRbdmjmq>i(
zPkQ=Q`hSxd@&7hQnKZ4oBLCnlklxsRIR1WD+ADoMx@lzoSKe$aW6WltcU_}n)5Z~Z
zHgBpA9b=ne>hjx8gYMMr?n)WHQbGQWss85Df{thVCLlnLyrcyp$_Qwj-7e+A40KTd
ziRT`MV>Dx@=nCd?&V96mD^2Y#h`(I5R%Q}9t@V@kI*bDP92LLUkK}mf%Q1aQ9wc(g
z26b=`(A-ChF4t3IR(|t(Sk5>Q3!9uyIZ7fvycX|-%*;EH2snmq1C$vl%Ei<-I2tE3
z8-=AMYvs_5tlKv}2EPiMhD?NJlCF50$Vb1s4b!+fgC$R4^j`A~750~lb#S%*H`TNh
z2>!EPz9HW$p<5S8Q_ZQzgRTtK{Y>BAtV)r!7Hi*y`fKAcXbx(M659`{STBU3ck#Ll
zEE@>B_6ObRMjdr(+kTH0=T?|T$=3F?xJrFWHWO_yboO9<GEKf>U8$7#P|=9)9dGzP
zztlKp7S0DM<$ivp@&D+|MXY~)=3v>#vw#Oo4a*a4A-imvIcCo*09|0Ma`%?}m)UwQ
z(%2`<@z}@CLO`x(FaIe%equ2lKU=i-^Qm+uKD1c%q~AF}@&d|~bS}@`E%|K6%5-X6
zTX?g$s1KRT%I|<%@_#kQ@WnqrSHxB>tUIeMeYf3G#;s+RK-JeYz0}u(=SWhfNq?a;
z&RX7#f}ti(D3f(E`csY`^*GNF72uVP@%2EUS%AyO>rV{azb^KlHidd9O<rupE^Lbp
zRyd%`XAYNzOR%%=<MfBJP(qw#>~@wa9Uh|geKEuc(P0ij!(_#ejDj?Z4&<~dCQUCq
zUd=M`?w4#g^fm&%EYs{c<_~Iam*ds-msSrnpIx;L?!wORdp)(qvw;))X_g|{BW)}6
z>`HlJmkHxla|yIX>e~64k(C1BJIX#8`<>u{G&Dnjh7TKy7u)EsOMem;QuD2c1$M6L
zecZg@c#CmDytc;F;<E}4keRz+`Wdmkf&8iA7_t9E_Sl>D%zZqC+RNPob)YR()!?s+
z{2G->t@4~?Kp%=tVm#5rM@OR6dxn_dY`Q+Lt*WL75ku8;zhp0Lm|{3?VIFg6*J(Rt
zZu<T8nMc8X=Hm(uCA&Gpd}m$#I!IWiJY4Oinuf?pG!!bls)$=LDg=?V4|{M!lPGJR
zUTaV;c^#sfa#d@kt5ce?^-_Ol&@5C<jf_3U9BN6Th+FszhvCabHHdW16$z08*SEM$
z5W8R)LQ~z(&(evt!nbaHO2mL7BU=cxvWx5M*FK`hUkxQBdJ4q&y+BIlh=u`kv81Fl
zr{c7w*~%Snw&#miSIJ}Wj?FojOuAWH<3Q{hOfsQ3+Kv>Bb#^+S>&Y4|*_uW8*0c~5
zndVol^3!n|m6?d93&vQ_|KHo1Bn-b#{YxxfW0jk>Z#x^bi^;0@Oi&V!1>2P8Vhfc@
z`dU{JEj>;`hBM=PvcPFw^<;tvdjX^ZXgO&Xn3KzEL~)$x5dFDAaSqvKO!oFUQB2f!
zN{M|Vay@_hA-2AgZ`eyZ=f{=#XO4aW!QZI8O1=vAQa819QL_=rWwc$=F~_Vr-OI=E
zYYM|rkWSx#opLkU%!amyb6Pc1w#`298c$>ZV-YsF<W&i5+726ZYVH!M{6NLE+pcs-
zDQ)+mOG;<0X&k^~qF!XHP^@w2(`L@niIiPSdha=NVa9mD1*m;%5t(0znIjZOU4I(+
zaFVLULJ0@C`Wfihie~NFfWr0iYIoT(>Ga<<R@tN3B`5DbmjBJJdoHVq><X@|1r_nm
zW`&htbz}nIq0*^hP%G!pdtWoZEaSmuT}D`5OAK@5Va>IBe%cfDca<EPOh!~yS0gIz
zQ<<6upA|jn`lbzVZAf#r=i>`ueTgZ4JZ1-PcI~8^TM!ORer<4Jp!7~IaPytBgpb7&
z{K{d;-zhRKhFg^f)8>}>wyIk#EUC@trkvuhYNns9gRW{WkKP?kt>gHg+n%o8tW^!+
z9Pjt;^j{5HrQO?@kDE$7Px<y7ceI7Mit~W)-D|vDUpY(d1=H*Aq_Vs3T60?w7ay2{
zr*Hh_{GE?+r2kRfKV<hGD++MM;xthJ)!+=9G)^Qlc<7xDZ!|OE+W>_Xh%OrH(QOv?
zWgA(GW3_X^$E^3tnX>14cau|pt_vJ}_(%F5a`=Z1{6h!+D>@*A5v-!CNl`6~v4~f&
zAG}@O7tRE|!qV@rkFCf|xES?|c<U}vWb0b4+S%o#dkPbwP;+8c3<?W;e{eW(Bm+ZF
zSHc6kzpZ}9({^8I^&oxQc%6o`wNF14%W^>V2PYASQKmECuuEon@#r;fag&>+cv!31
zttO85Iq^>K<MXKZ{U1njmfGGfGab5JJiqsk%Kssje`w49hLecYr-=5<3_6ZgZ?ASR
z@AI1|hwi+x?v)qnQswtnBT}*F=0|MkYv@WksH-tslRdU){{Pb?ZmJo?*8|9g%>jyt
zyrGI;I1@6%`GY2)8`FRNy8H3DcU*kzMkk-%HI*H|v!UlEpzX=XHhS_Q*OhHVN5|o<
zuU+)8=0H|xtTazhZK9<5PsOB5rn<0@8JP#gIyyEDpRIz1RhY9&evMDaC}0iu+^@^Y
zJH2k~>LxGwzW;1mNsj{RRk>eFQPlJz1Sm@Y4g&gy!e129;}&cTW*~899zwN-QdGR&
zCKXbfV<apLQhqzJpdJxUui1d9CW#pK%fPnh$|f&co2QtYKb_D^wx+5a9Y_g>mhJJZ
zs)Yzqp0RiNCrokh42wg9Ye~#4kI4cZ6XMOMbIwomi5Uhnx#~}Vl4haRrvT0LQ8}|$
zzGo(UEe-q_1+(YL4!g<>84V=5E6<CG3&UKD;fizbse3kED%V#VM3O{Ksc0FaAi~!1
zAb4dQfx+Ba+Kqf^N>%OZ1*q;4${Y-K+kW&R2n)0tPCS(Q;p3+3&K*NA?Hw5;rBlc7
zvKEw*wvRu{x#O{$F_1xVyb#k&<6)7tYYdU#n*O!iYaSvFD|kLrR2Vjat6Qr!?VIg5
z<72DCT-)tlQOsUt6Tsw4h$`7Ep6PP3^;nEz>5Em3#ImzQ2rig$)Yo6}h&07tthrnb
z#~v8zp3CPoacsIXMu8pXYlB}Fz%kiMUs3aK%i21<x3sx-b1pM{+Lt{C&fu>x^~cG3
zZP9kzi2`L&OgSAIyWQ@5H>dt}*&}PFVS%x`xN7wW>LLMULE%XOVHMMJS;vOQ&vxms
z4u?ZN<(p~zgn#*|E;=*3d+6*iVH$lB9TOtO4gY#9D!e}G#}~J)#TVm&Gkzn022tni
z_#in7HsA4iggQTbUP=hYA7Jd=SQmXu1^(+E`G~Kuk<Mph>A{h~&)&zsW?<E%`L|q2
zE65SP7>DVQA)+Yqp=Vr2Syn9N!9vag9&?|Xfe65`cT%uoIF%1{1r&ZMTQq&pupId0
zg+%>I>UwN_V+g3z%zu1{JJm`o%2A2nc9gogJ`)clVhLTB^RdMQ<Z>jt%v-FA9fHca
zgT?b}^6UMW;(Jv5ptkY69!YjT`eKYS374<%gxxr*-@BxGKL83@ppXH4-2B{4ZT?Mb
zfxZUmS<+iD{qb{2X3N(n4P!~-#aVu^#=lH}$HQq3!VOXRm&u4>rg3ve??JSbi?Bxk
zPZedr$6Q;=v@7vC(X{paEy@Mr5>o@oQJOwivzn@m?&I2h5Ab+w2BvNL=$U6$62~8$
zq0Pg9xngRIr*kRqk-TgQ)tT9h=qzE99@{tXfH^BZiRD7PHa_SX>WYnw)Kc9dEhl}$
z*~q;hY}00}vqFuP`^2_F;-YBP6`1bc?S~B8KR7?gnqM#qUBoS7-`?R-tARZ3!pkY{
zHD<kO+JhW_VPaH}2ftWHxJK!eE;gfr@k^SO2wxqdRg=xLU!q&%^Gtc3bVN=tlx?T5
zO=<TD)2j;alq}Uftb6+HD+#l(l_*8`0bsz+($=$Ks@PjSsgZT*?S3@>LS9-f+<`Yr
z8pYjTYr;?dD3w*iNQZdq;qr_F4}mL>yr?wy^|U^tEqX}f<y+`hPcE2kCOe72kAAK)
zDSJ;j%g_R(%W5cCPNS^hmN?SG!@JG}bJ#dB5k7e`u_kge!Y3aCw0*D7hyp0qs*mOC
zK2a<z<xp3L=C3CMl!l`JcEu&6f^lv?*zCDa1_jF38yTGr@@tfJi#Tf3-5B*x{5T3W
zKFePAoe7>Wk&V2N^r=5mtJ^GxUIeSRK4d1E+Yy-zUUJo7AMq@wKH77fOQ{}gtV<VC
z2=%>KFR>4W^eV)bqN48aRnmU`OGxfjWtFS6>TtfaZ-#1c-fzgSy)iA?tJmMpj`NM-
z3_OyT+(xIiIxCaiaVgGJg}{4UeO_7ZbLpTM=$eU)g!rCa?|fL~7(bPa){yCMNWJwp
zEzz)74cRXlAYELI^d9W>;voq|Z%m&kC$p#t-wjhphc+Q0;lqXd1w4l~L#fsRvw61U
zdl9a)wUFw18zqq(zqP<;MORGS?GvU6Sl$}qJsdFoeOT^+<z0>w@7|{`_%vbDnD}DI
zv0L$NU0TRv!10jw{vD11?*!lUZYCGfd|;2IPms}LLY3UGvG7IjDc4|hpx$;Km&O_$
zw2JF%mz!}Y7XO=UUHJvfZnalZp5&cN-5Iue^aMD*co7BvAYc)8Q5f_-Wa<%!7P~%Q
zzg=hO_wFpX0?DBup!3?YXM7LPTMLqJsx3%x9|pzB1(ukACOy+qAAV*+78jL;aUsto
zgf>(R1$hPE-COSxZ(N><4&40#M{K)3_aa~2r8-u%TR{wdC{VYGGVu>x(>HlS%jJ?%
zqUShmA=;Iw09`WfeVNZIHTLtSN7${^vR-*n3KcLD0I4u_-?Q*-*i<^L56r9}_Za|i
zXB@?zruoFI{iaeayLzHy2cI-h=kmc&rix4o7l_WqPcDjRcb9v#veo0g0O~=08%C_p
zkd*B_XvnpDe%it?a-{yod+d3ObWFO)_%9h%6nK1hQj{JfezHSPQ%tGF@;v^q=p^+1
zv3RSSpIP@g>pVhv&HKvDSQ_%6Xxy1WCV)QDNm=A3@iInAFMqy(-7vUZGHtFWzcIvs
zQP&AuY(|*y+H*=gJ2ov3$e`2|xi=WfKPS-0(zMebjj9n8-zWUMGTfV<<&+aB^rRml
zu1s$M_&q+C%Po!X<pSYl?}B|Ti4ekIor#`*k!F@<?7XPW0W?RX&8UwtkiPKUD=ta-
zgOj&SU4x%dmr3d0G7+0qEI3$UJ7%V>p$LLs5)nR$>UZW}HSH&T?4s2lgk{<gblHym
z{4RqH)(kJAqDYJK$$b+8jLmf#;iJ+RF1o6>#h6gG<7M?$y-BIJO<4XZ%zlE|$hYT%
z<x4Uv5R+i`c;-?P!22zmvGxWfk~(E?hu>LzcDVBZdx<Qpk<XwiIhu))ZSG(QU94<^
z)kTpkKo=OWQEy*WEU8EKsccZm@?SYEG3NuZsqhE38QY@1r`eWKo0dF6BAT6gu1=<r
zYZb}U>JIvvdIY`d!2w4VBR%Xr+$dQh?tzw_D}n)uA3qH1{A>2PCDwH$-zBXCx{YNi
z7FLb8cou6_0RP%5xqp;pUGe$Z$PppUxhVU1lE3iQVTB<@7t|v%+?BSZ)1kp{eWKAU
z(vi|2aA6GeY;v9_<NBE|QZ&NI9Gp^S>jGa^+`;fv)=KB%4F~2(oWgA%5aOCQLhU!p
zSf8ZyNzpClQaJ5BTcRkt1~nM?6kU#+85N7meg#v)EO1$yQA3fU+NHnW;X6`uz<O(?
zIk{u3Rig%*XPq8cj#;cA?e$K@YGzVRYz?iCSQm2ke3*j+^CtJOPHX%WvD>2c3sZIy
zApR`O0^0gH2gWw3sgv0D6e+5fy3ifj^hAfzpGpEyUY53+`uNxdUa}iz#_*tZT=w*c
z&wFX3W7#!z4-Eg-YsvTZ%@FI7WNnPKv>bI8oTR<q1h;wa_K82wg17R1e@JnExfkpY
zjt+>qaM6$=cRfaY-4*qP|LdOHlUaux$gEf;W>-qjlL5Zx(S_h)&$JRFZ0T}#!uDo=
zM|go~BB&IokEvy-bl}-O>jm6Whp)oS-ilfcT^$<Fol`TyTnSBpEmUI=8Fa8D##x5<
z$9R$PyF8;M07Ly0*+4m;XESe$K^^^USXx`(BG<0I{o8F0{az;qug}k??7pt;yt<yI
zrn11D3K~^99dv1-{peSKE%z^`tBchEV}A0unWTTV*Y(()s0_uTJTdO9#(V}1yc5_F
zSnY=}BdWkjl+AJMW-#puN=J~-0u_yG_?*ztts11&Q3H9|vA>e;Lrs(8sDj7V3!z<g
z;;#%=N|I3JPdL&jG#`y8@oybBqTf?q!v6%p<hqGh#&2X?RGJtX*s~@YZwnK}QRYo(
z(v^l!`R=1%BiFO%d)t}^k2EbwfRYB6J4~Oo)9N<NXEeoFn|Kd|i@O)YNvZAkg>}R@
zVe}5DJ^XR>a+)O<q(Gdbd?lU4tlWgG#UE^59<*VKV9gtF5hjTu0i8NOao^3(&Rq7i
z#Wqq&Z$dw_cr+g?V%a);H6di*J!k?yEKr^jIiG+#Qw8*1WAzkV2PPE~wug5g+D&u&
zbf}5yoBAv3RPBRb88%Fb2ZLY|U|$C3dH-sr=a_WiscL@T*;suf%T#8^wun4Ze1(|L
zu3<Uf{#MT~6`rDd$Gx<#izHCX$PRYigtLt2;>iIM!rXWae$7|8m1|YE;bQm{5uang
zDV1AugD`Jfi6Mxc%15;?Rx2|~$+z;w5+w2xs|vkvtAQhP*CVWkfo^v9Y9*5D<4oFr
zqm!F|!L^N>zR$~CtH*G&DHyk8=!xsmr2GfR8V<TS*7(h%_?PkbzpvZilmoen1i3y;
z?!l`U%1{5k%KnqLi4Vr#?^I?68gE=xye1d=PfGrMS-1q7ey5*V|KLbfe48tO;qdjJ
ztE09A+?eoO8-K%oeQ<gA7*|iMr8ef?GI1lrhVvD)$Ii~wE0dL>5LBAT)X3lI84wn$
zC71Pjs+MIGL}DTj(BgiwrjSlx&Lt|6YH-$cgFtBhyqG(18}f2HQ0Dj4jB@mUm+QKy
znUTn$B@;8)4xh;OjxoVr+L(<ykQ!~lhti06kU@v}Xz^pW9BWUTWgx%0P3gU1lApfT
zt;BhA93-p0vz4aAz=%?(2e5oF#v}rzeFxNn4P+dEO$o;>G-!Ptk=`Z*A7YkCOXG{4
zUfdu3m2s?=IX_L3zhoTS;H2e&+{>4-Hqgib!-_04hmM1&pGff4u%0U&OMmjL0z(d`
z)y*cGl!^v_6Au}h7a2f^`k&W2y}In$I`fZkf<5@+<csS=0r&x=C+j(G6=3~P#7%t0
znX0GICpeqtqytco0P9$jI6tf}QbuiVvB#x94&rL0zG6qrp8rDM1$^%PBt_=5ytgK{
zyV*vN{$_K>U7{{-`Z6p(S{?MT4dw98BX!023w1<65H=k;*HIL`^jyPbK|}^fupQAe
zJYqqT;eQe?J8NdMlvm4Gw09N<2{3Bp<1ozW=(X0hkF}?H$eFmXorNiQKAEFYW_S*y
z063ga)tc=0A6QiC8+W4(CzYyTAI&Tt%W6n_`F6vuB$vOn4!46JiZ3w3#8ZN!SB)*X
z7`u^HF}mA1>8~$qhaX7rHf!qWwPeXyDB!ER?`*BS{NyI>5i=Qv7W*J-b7uA<HEiuP
zD<n0$8JCbo?`VG4;bHEo2@@G<;ZCHG;+x6o*++IUilMfmSk?G1<f^jDv*DAxI{{aN
zlW?QDnuPcbv!tnO=n+vwnS0DkKv<?M252~F7gEM(Z!Sg9?9@aAoHU&54Eu`>-$oe3
zF5A;Ww`5-z6R}rCG~1K)CM9hS_7GsAS9-Co@M-1T#G%gKsp?B~mB0wxb=lO|<dUs1
zdiVG5>xj5lE7b-hp`RWE1MO&B@J2O;t=-4QWy0L>RDH}`je|uqz*>91=B8NFVa8~{
z)bejw=@w$1!R68rV2W)$$m{Wx9NHnQkBOQTrmKzpc|cIfOBa!hsD9}l)Wy8(xTt-w
zRkZk>(4CBBeew?ugzbgg7=`Bh->hP)xG*cht1In0tiLU;O1T}I*Ro!728}vNZVKU{
z`(lKNru}tssm_LAy)Qr(jm?QIR|*Il;l%bsT3z%ke(CqkH?|tvx)U5GNms!7bc$Rc
z-h_5oWx++xoeUJTym86q<46d3l3v&AoMr<1PHkTb2;G6b+oN9?g5PJDnVOBJgXsqL
zka%V<uZy=lE>phhqLpmg2_HTEi5kWS^Zz(me2+S^WROhFPv2iSO$8m!gQ|P$DTKOf
z=Vd0cb&~xPSJEr-l}ET_gLqwQoyAX?0tuObHRNB9)qviw$`5~2y0eCjyA3V6X0p!5
zJt8h&Ln7)L(-%@Pb}jD*`PgN5OA|>|+}3Sr>@sO}saDf{YrTKa@Rs}6W_xPxh^-C#
zcUQ`nba>+^_tH|%Hg<SL5R}T5liJfKC@e7VC^Q=+HpaheY`pAf5wN{-D_zH3rRbEu
zb`T>Q{@6d%7e^M?27=2alE>U%%z`+F618Zxz$_?YU-hrV=XC1bU)C)w>9=F{7{g(Z
z@oz+mV@3OG!l4PH0`DzOML~3zYW%Tc)Y?0A0}10oTe(^z4s?}JeKl>PemjSHx|jUY
zT}!69WW-~oiGOgWLu#)&s`=IT5Qa^gHK<%U%$XwS!1_k#_%f&C6_MEbpd6%}=nsy9
z1>+x_wC3r*%!B#_2>F`VGQEd?f7yvmU%D(WMT(ZJN`mt>H`1=9Gg3kB*W?epX!-Tx
z&~pfm%<*?tiL=wFZ$W$BnUZuaI^juF27P_C#;Cv4rBPu7^9?%(BsU)OOk73A{DeqQ
z)k^jm7m)Kw&sW<i69-ZiYy5zSFAfvW3c60T4R4>VWv}6$M;mMO9;t^rm8YASK@#%2
z#M_Sf=q}>l#L>1j=zMQ}SM_A7Iap`x`j&vYKMGquEYl}6r`&e(?D8QDp+WTBQK*(5
z!AS-Y&F7p`?<{^Vx^_j};CP#8!g*cMey{kQb<15G?cp749?-Yy4-WOzPqM}Jjz*ot
z?K?(#W#6AIYJiV>^lI*=1Ve9=YCm8CD~u;OD%WWRpQ6DAA?*V3Tix{g?9eaW-PU=J
zG{RWi?jtYKXPF3ElxLU#l`YY8@A^y5;T7jI-R(}M^Q8e=J%w9Tj0zd%HfhTDl#G(i
z7_LPh+`*#Y=~T=|ku)!ySz6Jyfc473f*9ux+3QJui4I47t`luvv-_Am*eZ4dk~@e3
z`Tg*n&)I5g%^IYw>xS17S?PP4OX1bHr&V{m2!h3nDlH>drCg$07%N*dt`H~Sr-@a9
zV$Cjbf@6FO5Ipt<Uwf_a4P&aRtt^wcrmHb7Ae2K}J6jh6aMrb+Fj&O~>%~72h0a<A
zZ}<s^r<`06+#6iXFr4StLGC}7Dt5(A=x1rW`9FcwZw(8V>-6@Pdgmop4opYKCKP9n
zlnjn~8s%&zMoNXQlnj?t0PZc|r7tfD)bAnP*Jicq?J_uF&0$RzLtC<pRTBH98^#u*
z22iE4Nl9-9V`b{-Z;46>FS5LzYUfMtOuFl!owPmy<Jxuv<b18yiN<M)`alR#O#gdM
zD|^QmC6rA+Q*NkFjR*{>C`Uz9q1L7<j1ASggoE_%buslEqV36-2K%3G2<0Gpb==)O
z${#%0G1@QY1vrM{;kn4Y&SV>EV1Wk|YqSSb4%L-jwERw~-pmoi!+|UHW>M^UD2tq3
zv7*b$c}=eCn4&TjtB{%EU+r~w;c0(x+B7)nNYf6O;6kVv`sKa0HsJ|gT!M*t@h$`f
z&F9FUbbe8~)^9C=1X?LSeaTs0+5BFm!>Yd>9Zl!=sxbNyi-NAHLhCa}#SwTDavyKI
zeKqjH-jif^znr<9TH1dQ=sFQJ8L&nzFH$R8RV7%{H7!L14@kmrNB5hs?{7o(1V~nQ
z3zNe7xA-Z_x9~)AT$!(gfBnAM1s?x4f_-I|8q&D^ZZUO79CjUdAz1YN>C|wTi%~1d
z&9A-rWs&**;@RQOz$!oh+r#izK7;Bks1@2zR4Ej5zuG94blr)Pi=-x1TfP}O+l~iM
zA6q`%>K%$dFhdD<QZa?t`Q9v+%<>L9fJJ%=<))0k718Pp4V`?NIoh~PqVdQbI9bom
ziIk+gbor#iJ6d7ve!d<rRv9?A{G*A!j9#w+1^O$kXWF}Z+Iul{O*v7!J|x}rhX;be
zKfzPRxMuW#hnKCmKqwL-q}EUTd~lGo(CVSgSN&gIOA7e`ff<I}al&v^R7~eSOYndy
z!cFHmAXmjC29l_@jbiQn>#IX9DBV3mW}NWRCx;>~bsy^^iOG4Zf%CRj%|R1jrqE}3
zsH>SpZeN{w&#g0PY)QulR{&Df_fG#p&}M4cn35O6^22&dhi6-`S$zEQs;UmzB;F*a
z8{N{n`yF>j24fAleO256S(Mz=rRm1<G0DmIgFTgGe8cjSK&cy04KkbWdf^YwQ-AHD
z#_RykiIrMPIvzpinJ5zRpy}TOVA1gtsqhp8dY4}R4)VaaGnj`C9m7*U;j&OH-d)}3
ziGAxk!-T5?ePv=IG)rS>XXO4SPwM$@Sq*wzIc4PlF<+~&LwLvV?M<CIu#{YT(_`<;
z6y)H)5T%RRNaa9?x4`95NKAlju7#`<kj`#)UEF>?8Wy9@!9L&v)s3hX#%KFyllgyj
z>R&-_d&zWvK2x{NZF?wq%{}tMRqWB?F0k}lJNW&BHy<=;{=U8v3(aqwLOCz!P&eC;
z9(DbFmHj7g-)p-RxbZ=`IWzph@rnLR`Ne-y_V3HGwjjB!wzJysNxe#bLxle4%KqBv
z`c@8jA4PqokMr$s?Yr}=e|fv*;jq-(MJ+b#8|m0=afhIR$pQD2Y_7Fj0wY&vg0l3U
z6<F|Hr8FQjJh{v9er~%8<QgjaXS~^?459%^cNDU$RdBWp?RXewcOBR!V@$hCPG(|%
zo!H!-Q_e|?pVI7eI$Bdrv9;8rcyDFdq75c=I==Q*cmK$jNJ`dmW*De3PmhI34WByU
zPY($aDadtK{i^tbL(?m)$-`Nm3QDr^4+gp_MXf%mGu6{8J9x8it67GqNE%3*E74p2
zM!7yTw&SENbng(boF2PZMZOGZ<#lsB0vGPJ=bC#x3b1^OPI0jDaHfZTMpkrks01j2
z(?H6v$_Fa=&DF?R)w}c1a^wn!bKuFaJM;E~l!O@r{oB9-A&rPe!w>p0MA<t5mLYbQ
zi>g1KSa^_MB9=&22Axy4Ep!vmv^A4yq;U<)e{k$lTHf8#<zCCQ-a8kba@%F8$Za#s
z`-5}e6(>mPcW_6H?Kg{rsbR1{XKqVd!^@P8!8_Jd9uxNlX0ZK=MhxE@owxnUxcdRP
zt9t8Y(4@nz`8=p<%-e-obO9?XkQP#|h8i|+<s4Vn8@6bBQW&+Z`XLxsdK4t08TT=#
z_i$@{OTK-4UICWjA`Z)WmXc9kT5Xrf&)yGe)@-kXbft3<uI38~-8+)|6kX*)9~Ziy
zVkGoUVz#GIMd_5iX(vpU%BT>Dr2rdo^X6+LI4C#uE{31@<Eu~m6F`;$07cw=zx<5e
z;>?f<KG7)|KwsWk2Dn_@QYP10;yx$6nW6ZthG}=!1L;uAhNpJy>`gz|{3z1EK9PI}
z)_`uR68Ug3>yNNNgV&4qqnzk1Iz1nyB4uaX8MOvZ<;$jQ`(t8On|GwNDK#RE`@oY;
z7$rhl6y0@buyag7Q`u3oSd4Rk^PV#P<KxHgYjT6-zQI!8<M4e~49hgSsgJ?=gYyC;
z_y^}_`216n^q}5@J?T4E)k0X)9~^kZhl^;}Dv;ym$mTh1p^+f0r+Q-V?Ic{TItIvJ
zCqZg`ae9i|_TKyHyvB$hp5~kn0g$f0698>dXd?z5)0(kP!lk~SYTsXsZCj<$DqNbz
z_=j5>SzlTSJ!ltBK`FEs)Q0g3N--(=BkERE!RzK8jgObO2!w>5(`}TtJIH+&jdVJ0
z#;cgK*DI9l83xK+)?ev+%HW`OKN^_alPy!3apbQnwo$-9dVDtHjf4C5RYVJ)xaKSR
zF_Iv~rCsop#oZxzt+F3?TIYMfh}#nT%VMKa^2M%oVOB#OlW5dP89XyYHCaihog%H-
z%2h$2>98*o(5RjG<|%nTE{(HqvMqL}fUYH<)-&KOtQH&^T;wk#S>H0ziy(Z`(B+yq
z`D`Lp(g1U$QLY=FI+}k6!QCfI+YE<gAnM?K>z5|!J${cdF-_a^Ffr#zPB_RC0=p2x
z#yK0VrjXBv*fvlm>}~E!CMzJy1BgH0-&F&ZTAwK`fx4O9u654_!<=qY?bD7Wi%nu*
z^-RsvC`G9ao+Q4?ia?wH$W^}zaX4dyI8${P;-><VC+(67KWT=#{+ve}d`X?2F+xpE
zo6U}f7#J9l7{p{}N5h^@O&R^L_H@wB6;;Q^kT9qYOR(yfk~DPr(wkMnAH7VHKPoxa
z&(ilU2zTB09#t3JN;Y!D;?<k`3H0s&nG(=yy*{J{K1{wcoVxrqQdE73exVcJJD(3(
z#_q8<iDHsXlzQY?eI=X?ob4!z*13HB<E>4Zn=SU%l6ncKn(UeyqQOu4L8VB&8W4BD
z5lUC!DQ!Xi4dsIKdzDzxOLO7l11<iQ()x2Db9a`o;gol?hP1b#%~E%W+&<NK8kV+E
z?#90Pzkc+Ka5E0sj-DM<nXCqd(V0~wN4sd~U4=<@)#@uNM0#)S^R2CWj~6zDqJ52r
z%Q-$lU<_{oVsYyY%XE}K&V$cqd|C-cRWc}(-=qlAIV$xY2;DeE26$-b8Bymq+H14+
z0Zgk(AnH}{@@izTW@kkrApWB0A`1k3uiu2;RZeKg(WBYVTf-&;*DvkDE5HuPju|10
z{e&tf4Y{?wPhY52^fy|i4f%NO^hHa~9=lpq&pN!{0z%&p4BnZq>-66azmaPn8#(>K
zIShVxkLcEE2jt?;FG?TP$Dxuo#Ax~b1W&glD=AD3X)u1`_827recHVBet{)L33gu7
z3@L>Ob4!=J5%J@ypH`iE<R(m8|Bh3NXo7KeVE(phSt7+kPO8U;8K>)TUKUyPy-u1_
zw6%9q`S=*xLd9I)V{0psW!xY`tD>E3FtC~BugNE0ml<}<&G1Y#zT!!veD`Ob*(=jj
zYd51$!;>RNzZo4=BtM3eyJAC}6XgCvJSwx9KzOaHr!t9HNwcs_L6N*g-?MTa1Tmkv
zbvE*2vTd=V=VUG0aPtn2AfX#<)E;;L;ehz5*qyV3OTp|`c2PMehf}PJZvx*k81w3A
zdW#EjM4C;?um^0N$Pt)R^oaf>T;piw8mEik7IvdU7e70NqT2U?fQL(Dr@Du8%at06
z=3}Ar>h!`dbzjStnm_jGrI@3}5B|75ojeqn^a_%yA(F`Hk;wiHPkNl!v_;J;U~9aE
zA&M(2k)+zcUOW`VK@{#Qs9Pdz)BMvo%TRm{qVWY-`5EoXLo?~J{;rirG|@=A061-b
zs}}w0=SHlcFrbg?Xm{$7{gX3M30|k|qB;jp&TkgJuco18rJ%3EVI)l4^g7FVA2$L*
z5Ngda{HY-REy33D?3bun7jBTj>5X@}v4IY(eJxZfP%xZK3zTVmvsO9@<gf8%9Enmt
zRU=KA$`G^1h)*4Ck$fF1ax$bjByZ)Ii(x?d%~;j1@UwJX=;l9X*0y{kM3>yb9HtRn
zUif8CPL1u^ADoL_O;2eR)E}I&_Rff_7ZQIRyZ^zN$qnm0u&&DOXanA3S+KD>`)mK(
zjlk;r{uusyd^vAAZQ65gsU#1hZBK3kwnx5nX|0|F^QwsL+Sz58Sbzn%dY!Ql85F;z
zEzGMf$e|KhOdcUY8ABBI_$W{4Gil}@oCjLmMp=A5oId_85?&}mGtrRp-w$~xL+2B$
z2(7lbnETHzl1+N-c}!=@hS;=wviw!ebexaP#Im<D0eyBd8|gbUL@|GGJPgM)HMlAR
zpQ1qQjS0wL*MI;cN$FiZ4VQ9=(>6co$V4kD+sXWzlcdT19C{L&6^hl>UtZzkCap@T
zv<8$0=#*mJ-M9sv9_Fm4sr9L&C(?MU=xg)Epv(ST2;lsUwRj@nDQbd^?I|<9GzG<1
zKcioB^hUgV?z|SmYP}vZTJ`%5X)PU_)QO$N!@N?^{`;MxO=})As{7!*M9^T^q`Hbc
zKRe#o0Cie2mN-c~1vtAlb#Z#UTq8=fY3!P&9wy2Fi@!GU06b5nNu<&sVOeLkcLuMg
zgwdPE=L-`{^al-7KhWecu5D{~lm<Khl7VypFh;AF);adGZ##quX3j+>k;E=f&})XL
zq}u5mDGSZ-y$sK7hY(y`@uArY@UK#uOA&R$64T1P_1aDuDrJd?D$vOc?e(>`?ykX$
z#v^0XsL~5q@t_Z>&HgvVYhy!Xc1V!e&b&usez(2rvjs7thgTo|*^vETo%$CPs1iBL
zKdu0RqYIyDZYdp~iqU^@JEKD{E{c5nRR01Heb>|bZ})d196s%K8hX2__!zbTQxl!N
zc2&9-d&+2PDy=<*2!=fH9Il`%Ws>SX752<`-Vga&Jk_Yaj+e6@{0+0E8{<Cs9@5t`
zl!Zy=H#y4Y-`~Gx`zE%WSO=G$58)rG4LTd)sGBz(U+GFIVgD9^Sw-@A7%9b?*}d`3
zjDXm9Es2B<@YJLQn1R%xKN9C(ss58BNqZE-4{3Zgi)UhJLovrk?2@6z*04UzN^I*g
zbQaP81@bxT>V9Gq7Ghn;{Z(08*%&$;1Ny?Fe!&Io$+qB%s9t><J>g`d@5pG4hW7QL
zvFIGT!eh3lbX;w&>@BaSg)al~tg1Z4Q!-L+;_|Q+M$BGMb0V=}EOlT%y9v8)Mad>N
z?Z{D6XS31Pt-MeMj(I)nW@AI^*liNF7)SBsDBs^b7HSspV>7kt8eUFtn?@mqFL`zv
zE||emplB^1&wMPOP3;5npGNav)`<Ua$6lmsPU_^vRCY;u97WHr@zpib)35F;4Sjct
zg&r8a0M_Yw@&Uek`Zv(Nlh}x8nk#F}H!>5`J*=@gWmprhA9dpH8+H<%2C{9>0*dP&
zPXAMHyD!Jf#TP=sIeKO3UZSKcO-XN{6&dCBLXL1`%JUt%XS;AnQTLOGii~hfjw*ns
zyAQ8yOvb&dkpiZVE&n12re|zT^&p`)A{(a)uqdog*5RGECy`7l(zTI<(DGh_PV|GK
zFK~vw{#Ruq_D$&zPGvJ>&o+0z=7VbAfAzh8Dn|@h6uqwZdrjvT>10{=V_;)!?c$Sr
zSx*f&reZyR()-Lv5lPg>Dr6Qjm+VRIN{)8e%99}ncs-rj<3)3W%u$SzlP+Ki+Fqd8
zgD#(kAIl#?yUt3zxCyQ!;mDypS3vrk&o&))^&u6RZ&N<$h9lD~fQ|Z7oV;ax<$PN0
z>UrASGlz~0T;Z&`%Nox{nmr2hxYd&x_dk~c$TA*9e{ja^HaZuW@lEg$gQo}s>4uak
zqX~RziYhJI=HYFbk&2R+1FMp~mPeO_9^HKF8Q?-LXA@uc&KZc8H=SB5BSWvxs1o1$
z?9U5&%;RY(^>c3cPDPnL^5IA&2HD;0;iv$efG(5oz)J}rm@G4A^=Q%m;>N^7%CUo;
zhf248Chw7kABJDx-~drxW{BC85$}D=Qdd30Lt@U(0+_r~Ds-m4J7l$Fm||qg)*GIg
zazP9=(|uB6@VKaTf-2RDBYl18zHraVv7FIr>;90q%Yn1^+hYD#h=;N)xEeKtY$oZ#
ztgv@ZumZF1Z=)()_3_L3OVxBIW^$i0qempL^F<L3gWe2m=bZ)9fm(TD5|t8WaZX(X
zZ@1ZW1`>3>ASFAIGxFZK!M!Hdg=bB7Zma4{!*8Q@uij11p3j0Y+(bP+2m(W;Puf;l
zl5vr6LXI3~){cs#$u9Qli*u=3tOF>(&g(UdYeM9)>)*VF39HC&E0S+{t@d91^*20=
zdn36S6hA*QM_ESM_`~8&Y^P&?dDAhrnA*D~i2F=06fOB4AY>2J>gdxB3)RZms%=MG
z32x5pzxUSCQvKqnAt)C>Hdm0Qx0^J+<B%3aM9hV5fc%W=8O=Jh%*YVbG$GkP%UlW9
zVH|C-PB~P|f^&(qH5}$~tiDo-X0{D*kzJo}T=n8Jcx{tRywbXUJdsib*u=KB@oibQ
zBqcY7YSh^&^D{-F0rn6m1?as-tjq)dC@(6oMs;NdLZY0VVz{-iAYQEspH$eW&EmIH
z3S-JXr-l!ObBjVb^0RHO!)2<1+&#3*6?3dw#=clVnwE}bkAkf*n_KAz%vY0%n$J24
z$wCXfyD3<Bh2fBF)wC~5^wt?$<WFM*UIsVf4#5Fw1&UrL?0&!ikt1D26i7kC{0D8~
zbEoYkrO#yf!W=8rSc!CYye>TVK#w_Ic=t|FNi%81giSvNuAac8L1&fO#`ycdQA|td
zN}y30HgAtXr_%cn+`){JPZX#ne7JhE^I1S`9AXtmTg)7Po4T=E-N{!vmE%9BHA5_N
zM_ib)Qpu;L04d_uMdz#QI@l>QQ{OUYzMZ;kyV!+}j4A*ln;TmE0$sALt)(>UjN_la
zh^T0?i}2#mNKl<6v+1^OWKLi?e>KRP=h|G)LZE{;tBBXXF|6<=wr<_r>0B><nNF$n
z!s7m%m+NL&V9ZYmi#2IP*DNfhD=DqP$44knwqznLK;%pP<#0N&<lvW7dR<ki?PD35
zn=`&2xi-eP+G=jGR)$+WmcucCQ@eqvfOGT7VTE8~5WLz!y04Ni%+&u;a1s|teD+J?
z%LxaD5&yyi;p{5Y@+yg{#};2+djVgcs_E>XlQv!(q)_HgZRkAhgr}Bf*uKEhdWlvJ
z&4dnzI!9u-4ISx)Jm0D4;jvZr^BQ+4qzzY(thD75Yt71{mOO+fnr0{yH<Jr4E#mg+
z$)c>hTrLSZmSv?LHWLMLV|16Hl|pHtv7tLwi?yOMn;s8PSnYB`hnf4FsQ2eLbmcA4
zHC~NetvhW!>$JBxYdaEOZe-!8+asZ^U>b*#Wa(On`T!q7fv0x3@MT?7Nn{g3aM|2s
z`r;5^Jw=)p5`HO)Y?m0&em3EAu4N<gZ7(;h7`hOjiziU(8f5z`cjx-S%n%#Ta59*x
z!tokZPq%U(MXWIF#O9&`a<;?T+?|v_jZ|tK-KDaU_5zPfcQ}-toONJ|@vAC~8k=`Z
ze#(sKmHI31sCBjK>z0R0^g2oz%QH;OzS%v_<_Nttk(9SLWj1|=Jab1{%~)PFd&b*O
zm#HyO$ik;)0wcwX`)dBhq{2&;*hRPmHJMD*Px4dUDIlyurf*Z3@j+wVxsbF?zKk@;
z2$dC_N`C8J@##}@Ws5+k$EKSDXS|VEB6@0#-)?FnE)1AyvXa}q5z2wj=}e<sJa3cs
zj$w7P{!*=pn5&|nuf9@qvUKLLN6E#S6m2Mv<y^9akSu`JrpDOIGL>Ie8XbYkf--ST
zi@P0atz<U`T9|(8^$}n*EO$t7LDap?_!g^`kK%QuRf%CT@5AtfC}<~WeEk~ru#R$K
z9%!xc@j)|zu}@PEH`c`&cSoj<uVgEw*B6-|s5|j!(E0wjIaAA%kFp9x%C-IQOKlT-
zfB38pQknO&_`t@z{nY%3TFge<P)Bs=N+Rg1s5=fG7t<Sop-sa>0caZrV(1I}a3CA`
ziBeQsA*ccD9^7qv5)A{2Xh!a(7;L)fJ>z+t5$z_FgU%Stc#~`#tH0)B;UlUL8#Xei
z({S1q7D`gwUq5_DqwReqgpPD9dewzCeX!B&5V4oJi)6`?Gx`+~f4Ezlp&1@6nLZtA
z<^cju!<*2>$&4HyoQxjF0;G_Y5rrA4E%3RWHsG$|Gx~`gMicaILoSyJIxj8pm$B&R
z_kmqzAvJzbGAKI<;VG8uqcU{A_ZRG#e|BL31BqI{g8ev-$$?l(V~XukGeXoWtjy|~
z`$ho0Z@gzUg~a>%xU8C7X2H=1G6fYqPvw*RyyNS_tS`Es$rd*A6wui>LO4+K#N$~n
z%W)@$dApt5=a|W1U{FVKesOFHTZ5~3G5bh1m)+`)Gx%;<Kq8Vwv6wru_}ao|W>G6q
z+yX2z$HS;||5q?C+GQlY?JM{R9$$?n8%JL`nM8~{#xeH`@{}@>6=ue#8oN>xGvVK`
zXz%<|cbizgfX<}zAR3PZ-n#cK64S$sDQliLK~*5LwJg)DQJmW(QG-Qj^E%4fuG7z|
zkg}@el&z3!UZF&>%Km*1LQTV3=iT1^*W)8)J(abO$NA8Q#LGV2b%Nhdjx7;V27UtX
z)$IP@1j6r59b`)3A29L&=%?$;7|!vbm$AePx+UAbbgt{Kuy#7jYRP{|uUo4H2z9o~
z2@`)$ef2w7f6Z)oehpWAVdK=1RGLluq-KkZmsjNOk8|UDQ|o))0PfYPUUbEgJ%~#U
zUIo357N-8-UAMup0T_4&Rv@@&6nktYITNK^pBOdQuC+`yx{VI4Z`k7Jjp_iu%z2Y!
z|5J=yW$7v1`*r&88@fYwBeu;HK7<X>Q&!QbVd_|o_V**Snoa^U7^*q=KbU*ZaJJk3
ze^{?hZB^}6wDwjcwz^uYMyL@xT(*QzBh>7)_a;_r#7xX2MpaAgphg5SY7?V2wR-#h
zzR&L8lk5IJx&P;J<ntss^Eq;kch1lI{TkJ=*M9yk3Z>3cWfEuj2GsS~90E3@n>xP4
zMH*Raztz5D;k$Ldnt}mzIvd#W@ZHv(tT2yAT-!`rOm!HlTv<%TC5!Mnc{&2-_V&qn
z`^UI#vC-hcTkU<OvK@A*<Lq^DEF>DqK>e#3<D=pYuH6Om7U$BG*=qTn%m{APZ+val
z*g|!4f}=5%g$;Z2S6SmMiznb}L#*@s$bF*t@Jc;VD|YTcp6ezp^Dw7y7yp1-U8?u`
z-0<4K6)QW(bf(vt&Mby5*4hac^eYdM;L#oXh!NViuc-rV=sBL~r%hWMIR~DX<A)A7
zURq2unpu{oq@-CfjlSZNk+5X1$^WoMe2w^8oUM|XU=6WyVtWQ#)8$cgy~1z-O?jjF
z`a6?v(C9?J@}1ltKg~Y`%NYYcO;ii#k;HC9tx4B3E1PzH8%GF@o=Gv3kQkL3Ype)0
zyqJ;*7j`n^r;={951UtK1(`hVE9g6*kf1Wf&3aLC3{P&>n!qRMnWSP7<modC5h0_4
zt)gk|GH;}C2+`VhcCpoDu=B}p4F*I$PW}qXM7{I!SV*}{hHG|3ug6oc`WKvF2nOSv
z*h0G&ekZC%K<B4MTIN7R0yH$&ZksF1%>@$UaN!43bU2c9$lOdDodD_3mg8@2amcIQ
z?oa!fZQW7Dkm>X!z?{NW>#`R8B5ME}BoYSZ5NyDN!*Tn1fweS13E^nXblqxC)-D)}
z>-r%^Kjg<xw|<B1#VONlFUHaS^&x1UOZsM?H$c(T5B<DLXc4mTO7^2&laE@7g#>MS
zPnmkgPGfSxKZ$!HS)`qxrqgZRd>zHrH5m!BQkJZ|Z(SHA1Sr!hw2-WM%W3pP`iK|j
zSLm2zNZ~O`hWW7kz3!e^KQbrI!s(n@Qcy-ed*S()Gz8BRt)%8Cdu$@a3snJ|Lt0Y5
z-h-XkmLbC`t!}=i8u`2~Z(j!j60_y1&$J%)AglcJjY~vhM^}X1mJ4{3@Ap;M(bHfR
z({qd+3!cTk`&+qF_Il3R#rGp!t30~^X9fp3zp8-<+YE;c)~Sh?30l6r3VLzO%MS!+
zV@S1oDnCZt+|ba@lMva_<a_O@sc4aB!)c$OMy6qWaUwb~$;n0e(ncnlj{j-F33YAU
z93Ij+5(pn>Yb`nP!bwOWdt6ANOr0E7RKtDyj8>qfD|>`SS9>Y*LgB9v4eL^h^7Yd@
zoJ#2#%;fc~sP160tgRDC4NEKayM1d9ZY4t$CtaRN;n5V_x=gXr0=7{_xi|J*ibnlA
zCp67dB?z1gSgbVR{0TJRzN}^KXL;#i)&`F<V%_5ZsCaf|@7<o(g^P<<T2MTZ^Y^6-
zoFuSN8;GI!eR2~>f)<Nru*Vf!roh}=vKIGWf$UU6a_j;&7gBr_s_d)9C8UDL)?v)p
z`#Z0;#qZxI;%5PJok~vZy4<kw^hQEl!cTBSnB9I;Pq8@;T54%{o5MBiDpx^$|Ef)Q
zc1(ON`K8pEkCeU!IR2|M5C}F=EcJT$n9Gya=FVr0hORB~z{1jc_TuohCj~3aS*WI8
z92P=w`|o(~-VU-d#*b|^wt}C6^7j-b&5_I;LPoSLTzC`~ahV57cLJ^)Gz`JHN6x>j
zF=wrMq`jz#UBe~)93RO0XW#HL*U>P>Z^)cpR+6_`Y_yCKQ&+M3XXil$+dxmQco$zc
zqT>eW^2z1@lJ@_lYEIj8G#=eo>-1#w)rKtIHst7jpBP{gq>CU~13uw?{3rcm$S?e#
z^v@Np#@EbGIOS!ZSL!~T#w4LO%XUTivEX3;?e&uH0bWZQ_4S4JZMd58LEM7$?cj#R
zky4`vfHHx<<+yc$5jqF0OffmG7oY6e(<1eZ$O(ya{=|vvM$_*764Ku5wjrols5;au
zgGS<!I`q=$^{aRG^842=t@rYGi(Ssf+o*i6mRsw~a(rU53iK&g${Gpw2Y&FfvwQ(`
z;GsA@z_V)!NS@8dI1)Lcw3DYm6-hLxv4{t&pP(`{^~3#;;d9USpDrhA`Swi-VH4Sr
z?~~__UU<T6z0%4`CWK#d1=M%P&zq15@=!zxeUbc!(%3g@9yObH!b~;3|GGQcuSpap
zT6Lu)%_(U3h2wKH!{VJLif+BAd^{E3#bWy8d;xcp3rF+z8M?Q^75yOA({dEF6&01p
z5a;Oyerlme+3lrK(QtG@jZa$i+B4pg;{BxGDq!6x3J+(6)|Yw6Q-)KEVhP>X{IutP
zw>t|MeSW$*1vSfy`?84W^XPRLr$FG2lij<hf2ufElM|982ld}I3iW7en>|bvGmz}?
zAp};4x7CVRSFwB!C!7So=dO){8#>*7M&0SymQPJ8eD3sJC^Zc?AKCr;O;;efA>=+e
zRKx<)L<#389v?|aeLiM^m{Ken%yhMi3NvrtVyqCZP;)cldFu504qq03-IK5E6Gdl+
zepk_1Vh@z`kF<&pvp|*e=+aNMcpvt<#=HW<0&!)%WEEpvg1ojNE>k#oim!BNnCHsd
zhAxj79p3~;tE8`jFcgPcg+gb__|Uw$+62h)!OX<xMxA131plKxj~QLZ;brA0DRU#E
z$b^QpB<Z^6u%nys-Axv~hS#onY~`m#Lt=sorOxLptTS%q@>p3d8;_|T#D~Eu91iAl
zV?jiRESPN|JE|Gq&U+=%A8qw5B~pD;J_$pv8-mI;`3v~E+YjEF(qu@m)TwGox<Y>7
z(RFcZPdMpg18<LMt2<ZDH)SYqREK*z9(;Rb)|@ke8d`vIg)dpQl$;oK<cY?Up!y1l
zHvXc14Q(d<{2T{|(*5pz@q4B4ysad0&AoED{8#<Xbu?YP{B%N>gu%w7q&;5VOUZ{&
zH3_YsXMY1*!ABvuxO;=un_|Qer?$tr8lyKwE!$s`vrR6&&m`uD0BIVx8|#}NY1p|J
zyrpw{Mt>F$$H4#<iH;$rV=AIQ2%3IDmF#b$p`D!8+FPF^B%8u?Ey*!YK(!IepNCW)
z5xv!<CIq<<&d=dkq3$=Z8#!K<b1UV)^^_&=OzUS5dfY47$}7;=x|F!`$xfY~(jVoH
zj)o%h6RP|~{)W{Ld&_&#e7r@4JXan{=tUB=1e*nvURv52<z(Ws^-SlB>kvkaze|z2
z)UI*uzN63iSPY7^H?P=357BFWm73Trlww3U?3$(0b)t25`>tYTSb4L+@BN>H5bHAG
z7-8jvY)N*O@!**1ua1xZWwc~>Y93tYGuQ=|oQ~e7i~Lu;{)(2?W>38=B*%^xO7rmW
zEIT`ggmpLT4OYj@(}4T@Td@zg&Nm$R#x_Lwrz~_Lq#r;BEEuV^H>_k9O=~0o{5Y4M
zw&Xp?)eW<UY_rs(4`@xi4tK?FigxXZ1=^bqyYHOJ<6e%wHi=-fs)4-BMyQ&k)vr7q
zWn&Y*G)w{9aaW|ReZcLd7?JE64!h8*0B<cOZq*3ZYKqPM%taJd*RmoImEg8C&Rf;+
zZV9I1u<`|Z#msfM@Rep=k_wNT(X#8_$=f)s(FUd!H_>}3_JpADj{;>CL~RN;fc{6c
zOOk8tv(%!ef_$Y>s2roKv=+bGUDMdEW)e32zh&3i9dw4k2faZKv^dQBxK<y|2$+Sj
z+j3Y+3d0X06P-n<;eEIYs0DoPE{fD;H6_X#k~9A5ZPGQG(~I5J2C1^^-nKFT)U&Z$
zKZ}jMAxMODxW-2@dK1NUI=0v7f0Xc;O^iR><iovNF5xA13087F^;sJBP>dX`PRP8Y
zbz|nqIFOB==~5A!TB<uzH<@Rk0ha6o7dsL5{W;A@?MyOArukm$szXRDrjL!h!7_jS
z5`yi~i5(!Zw3KzbDX$kMOz)2@Sg5T}5^4r&@>U}k&<OON1X2;gw<qEFw~F9xgK^L$
zN9@v<@#4YEo5>K<sV^T~-&bl9#q~ROgm<&o3yGi{gDK@f8HvWGNbMARR6tUOrD0Q@
z+30@bn{z;gw<QZ2qr$C=P-W~}>9mt}B$|!mVj_3OLKr`vs~6aJp!%&xhOO;+4B5DO
zAF_S^SF4Wd#?fl)0SVMn>2p5zn40b1b-MBa+vjz{i9_BNyYv8%WEnp5=kLz!5ZaRN
zS;M@yOY8}g?m*&fEkT?=8h|5CY+}Qyb!@~kb?wKjl9uu*?w+0*VDqg?4=w$giuFvU
z!|HOj+Q7r_<(ruvAg9#sdb*97s<%&M`g>v|ydA3dol;gUktZM7{PZ_uNwk{w9$cmk
z&^mTOJJ#Z7qj>M8RQu$}F}S}HvkPCAcc3l5CL~aAg~tfY=FZRqxNnDMp0}-mv^mq&
zUoyY_1xqZRKvh;8#vF93le9fw?g_6oOp95qrd8lA9>zgM=9Ar?nH13{r#!w4`--io
zCZ=~}iVen@@mPTRnOF3zz?Vd?iiZ!nhS=q7cK9_@0*>xaKDo4_?0#dWec%H+7`$FG
z+Rxv&wpYpUQ1~V&#qf!@o7io>2%aAKU+l&s6i2{<?n&}uQI_BCm3YK^e<k>j2kp1N
z0u#(1Z$2R)n8qgBYBr*~KaD%;E7{SP6_M&)4dp$|2V!V$6lx2vB)*e*68w((^Qh{^
z(;uGu;SLuz&VTc(svTV|A_vS(WA8SD5!|;&x;yGaSOViV`gqH`$PIN>L%}cyF!4)3
zPrNZo5%=SRX2O^vR$Q2m`}v!vo>aL@@KZf6^h__Qr5%dNv2dPT;1K9?fH{8}b{LTm
z<s*UQT)_C(PMi~l_xGZpU4_Zng0~V2SGwQmVzz+!A36^K@7gthUH(C#`B8ecZxPJS
zDplqx{O7=a`_WQAuq;1R7&Ysj<8{Z`Q6VuO4pWcBBnc+M2RiF*;fl1l4eslc;O}>{
zHMC9f05*82F*i`tLUF9)XiGA`XRho8ny)ucA2<K;Q>39$@1yoOc7R`|yX?llTde=9
zN&(1^6!`7i9D$mUfpo%rj4;8gTY(Q~uUEWG3Myld^V64#nqzEzv72~aznDZw&)fyu
zS`r!~W2Ba|A|CQwwSORTaMO=1*MMGMDm>S9%EpJYVv}{UQ{z@z0f)gi{vL3+<g)=9
z;fcrMb?G1Fzc<EB0@fL&7r!jkO6~#mx+kgS=A%BxaB&3D5p#Rj#?r@J`~wm6@9>Om
zpfcdAY=&RrpGG>XyU*s;jCmZ=#F`ePK%^X(qNd2S)r>3B@$UbO4qQ_<1M<c=PhqW$
zfH9h0kjH(1X5*wceL*)N8*xtv>y<ml{@Y6?-+LV=CrA^cC^ameI50i^*s+|<T-3SO
zgQ$MLJJ0#8ik*0xGTqts%VLuDe!t>YW&RB!jj}zLYAmHCN7LlD9eeAu_UFZoZ<E=T
z*TLr>gdu9)V;yHoO&ZPA)dOez$ML<hU*erI4>5LCQ086SNRpuTky{~yL-Um>w-@U{
zD0c79p?9&3;+0Br5XaNo4Yk%GCrUTOrB(zcYpm`aPVs-BqPk9sdTpxnhl*M9u=Gmk
zoBx^AeQd+Bir!fxn9+o2RMLbm(MDH<-B9n!kujQ1-#2CVR;zeIY}if5iehzSu5A6c
z6BSkPSNXJ^tOa!YCB}ahci1xO3?8SE#Vj|P#tu=OlU-48vqV=3+f-`^R2<E@3_>0H
z?Lqq(135)^p;hE9j$St;b-Y%6X>}N@fS~<_d!d^yIsK7ED?!O}ek+J8&gqUq_r8x_
z^(jPt^j`UxeQA}{csz`Tkq|frreB|IMP-BU6~|=m@r{m4T*KKvyJvIu!+<T=E%x5e
zeaxxO{))gzF7=NGTKt{$lqaEx$Wgt#S^2uMT0(6OvX#5BrOTDKS5qN&%=m>V*}8=-
zI!jh&Yq8mV+Uc?Xon^g4ME0gbSX?Ny%$LEdvK4VsmALg6E^iVLDR&>xX(JbHrwH@A
zKda{4XTfB{ulfRN&0|!w@ViAX3@tGC_^U7Fl0W!0%gn;Q0W)s)cg`Y9w4R`VS_1#e
zMZtqw9`xZ@=`X?W>zZL1QGYA{r^rRc-TNcsevvq}C3A|-k6M0yZ9{J(laxyp_in=F
zK{2+Ow)$p8<;Q}etptTK7)%M>9%z<}`0>+Uw^s(J#M{W`Qst%T?2Hf8k&%&k@{daO
z<yX?)P;qkCHd^kede82D@W^MWE3#eM8C{=zzs#Ehey_*qzIv+gQsr<mUs4jwL)IAY
zK#~@&rFs2Qy~VuPE@4$ao?=>UZRT%hRzA=S+9v0G0b3dCD-SFsITN3K)CzdKgqG}f
zC&%!IFf(%tNyu9Vd(TKCnm8hOJ+K^pBQ5mxz0$9U{AJQu1OA@f&lp@m+3&hvgLg(K
z6bdpDSk;BDjEQd%CW&3j<+#@8yz){jRDJxrD2r1zdFKjA^!A$jeAD}USLRFySx*hG
z(i9H>D#FCuy(Q8+U45!pBcUSmt&W-iX2m#;Aza5dIdN<BW{m7L4xxdznh5$}xsjnz
zzVVIT*xlDyM8uB;eZg_dyKh1XwFJLxdnHhHyO6G4Z7=-@NcP?8Cmi|drP(~adg>=(
zzM{zgw41A!6IfkQiG$v?B>i1$hktSEKCYf1g_?n?qh{2+IlOKmuyn%rilST%>60Cu
z#<{%c{-?3<@a#~PQqkU~K_IMj60H7W^0|ZP8Y8e)sylgpF0`r=ZrH?yDN+by3*nmB
z(*DZ{TfW-=pCxcv+Wh&vqB28*?>D%bZ}1JwgsrMQe!oG-5+o|0Yz;Ti0`lwV#HD>4
zO%qDUVoUt5jlh3<{{Mfyo}ckr=luoUuLG+qgd=9!3E}wAeb%n+3WX~TQ@pRM$50o)
zi_e$}u894k5WV+28#5ZGIx@{RnVwD_EX=L)sT8jZNUbaS+Em+FA*hZj?qCn)jZ;Qe
zMpLZo@UA)9`mmZv3Ho_e@e(ySZtbg?-K1y1_`stIq&B1?Fq`My?Yh|A-T8yE+dW0?
zPJTU|6X=?^%wJH93yAU}CC;O`@E`kW;$6(ix8NgzIWg~Fvk~ri9^V7`*x{yoZGVnW
zEU?IWw$K6C@m}04h4}OycKfvDo0-MwlZ}+1a>xTVFD?rp5b7>52uMlj@AD6-Az0#w
z6b`ZUf>2dke@^`Z4P|of(+<&s$u4UI?u4r<RM&gbSOQ0}b`qHUQL7Uvz+liow|!nZ
zEOhJ~gpvlzDA$()9h<rm7g^Q@Dr>2CVpnX`5<8OrX@6P#c_T7T<+iP92-m807+ux}
zh3$l_EO)CIzb_kicLVU2POuAO+uIM;Bsut{QD0ic+dPor&W4kp)8ZyI-oHE*(tZN<
zP3pfkk!YQb{zEk_0snPYRbadO`yWYK0I2D1c$aDYol7ARs{KAt?ZL29U{?xW+UG9b
zI3RFZSy0j(!>=kGj%;COksy`f4W;#b>B&Y$d6AB>y834WH&lwG_G`OThX~Q$WmfN@
zrRga^nt)Mq=(21^=yV0JvWDKx6K&987wH=QPVKp$()Tp>_pIAAW5@D0HjgxFr*4`A
zt!?gJ`nz<-&ItDmXjv-;$yx+D*%`oOYLF?$b~M?tI%3ZGHPIXOIQ3c?8kZlr%K2S)
zqC@eHO)H5r-DFfG6FN+j3F5`ybbMS4-;+nNL>u#F)tN7wG#D%@fQw9rM80*W7>}`o
zX{6o-0gc~s-&4`-qELi<5$F7Mr5X23tV=+(O1)%~4sVo2oz%MaSfz5t^VASG_u5kf
zb}o3E>~JD`S=W0f4)2&+>r3<=|2vM8zG8Ea#5D|O4xzCj3(7!WJ^&Lgc$Vqy+(b-S
z`is(^;i9iL!mS|UwURzZA1P%<J;T_{JZ00dA7lm$()hF#aHvn5$5e5)e2Hv~pr6rm
zzPr$#NAUG!gA1L6j|^tW@IGc|>$0*}vYn{hEYKx?$?p1UJFsVslYM~`7Z(fObxzul
zn*3f`>7*^toaIMPh{L=sot8FxqS4LQ9Dg`6Xes|{&AE9Yt&0~x4p)h=F<0U&G-*`y
zPaGZ@TYSDe!GYmHc7RRSc%NXzq;aVjbxlWljkfpxQh29oG%wLTe7dAgOycdP0<qDE
zTPN2|a-=l&@9O57mxGiCJwnC?nV+2@%x7m->Mo<DU;jQNtS@@WEkMi3Ma}U`sQAry
z-+Zb^d$pTw{N&e%z2YQehrz=MW56f;DCZ1wLtz<M#bSf?BLCk&vVY)ak=4TdUglc(
zHljw+qVBw5snPT*eb)LNRM}Rg@0fK{%oth2T6@*wRHC<VLL>C&0wSq&p(}=M6sb2u
zQ$Hh+yG=C;!L1x$L{m&-q3LLcOz?DG<`Qcbh1$Hkw~SKUk1A<~L40vrqMk1GqtvTj
z)TxkRgRT&)bowYq#5XIGJ<ea--$p_AM{_!d&fedDE&1a(>)mW1ct2R;h$D`dIn=kr
zdL7kUx^*hjnZA%K2dkLV@`72K&cB!6PMA|jdO1-y)<-Bb{{*w%G}e?c`|C@19)e#5
ztSKo#(SxR#=L$Sqg6M(U$X0RttnuwCH%>Gugm>=$Hd6neK+tH6W;xTn4C?!RYDZF#
zv9(WFSA3MLNK79BGu+oRd!5=%yyb=j1VlDR(;+S?Xz^sP+s~49A4ceN0fL|G_RA=H
zS}fS>yD_dSY+5jX8cYGx9p7t6H7(?<Hfcb>FZsgS-<2LUmZ!s3eV!IY3M6=tAww<r
zx(wB#`x`u8N2#~rE2-Id%){bR_s_^f5k;-hy6Z(cAt3P^>zLxvl;j6+T+|i%DJ4~R
zT7T{p=CaoW@(Wgf#H2IYaUQFrcjr-UN8xhxH({6QQF+`3?`Ufg@D4t{E!*o$Mjja6
zhenhzjVcrH?1>SqR&kYl8Q=yNI3(_VpGZycb;bz@^CS3^Yyc=XhBPoaYN{%&gcJ4C
z`SjewYJ;EtM<$%*X$K<YD$kWDd4S8&k)l*N^0k}P>8&O=<TD>`_yMw)@owH1w_m23
z7FkGp)G*Swb8=gD@u-R0s6hJ$c9T~BrOMDR18)|ND5!#mTI}=;mqMf?y^Xp*Z+nls
zh^1#E^_ZCekjd8-EyIS15l5aDe;36p+ff?I;LT6><M%Y+`@P{+4Bd6Q&FqBtcS}NC
zW84hrz)tL&kF5hpqGM5#jpijt$~e;d$tO;CK@+=HrHbLZEE@9Ouo>Pr4$nJ8@;`QN
zm4F$u=*MER#R{4HI`mulFRR;%iw7=!1<9-4d>O8H<Qx@~1E02?_v@*FT6yoHiR;~7
zQ)U|Uf@*!<OwaxyoO$$6k~J1&R*%1GUG%5bx2V<iWATB=ER95$qls(D<sIG=0aue2
zG*?IpF2W6BiUtp-Eo#i}7c;shvGXAWFO4RXK3j1Xt))LjAy;ZYUxXK`cGY(GPtU*l
z5YsuQ{U+T}57Z7f*Lr5YxOWrc#9AZ*SRMJ1V#n8$++Mh;*sH?vkGzdt54EX5jG>KU
zcy|x((qSV`LhU*4<|^P6Bbyh>mw4yj!3zJ+5-NlhWb}FX3H3z^uywc2=&K_r<)!h2
z1iXGzl>o%kHjJr&h?~l#-M@egg5AFiUuzz0>gq&xJ}w=jc4jl|u%i2@1S~si&aE>J
zxt8L1U#&gGiQsWF^dzpj#i75pGT%eVYP8?8yX2=hS@o|%p6-`9mHVq3Ozet1QjpKy
znCgv;2Pe?z401He^mIU;BdRMUf_Yjt0L3oZs8TsEGL&7an0u85i|&M@i8SP!unlQ<
zH~uQ`Hs2m19wFn^1KVn-9oCg|5Ep`W_gKEuIgY2m2kN)Eiiw}&+Jg8-<L;4XT#}X3
zhljbkciE>bd(X$RFR>UU@I{c-rDp^KOq-6*k7~l^x-B*hKe%n-xm`fGru_S(C;95!
z^aZ%^4^_6jIZejii~Q=-r>evBs%OoFzdho%MtoU(&P4slf|lE^6R~mP6I?g5{`u9S
z*v62Wbr;k;+CW$yxEoHr6+Hpf99dj&c9no`@7)sl$=!YgvuYoKx6f=b=sB#UXPOqi
z7;^Bo#p0?IPkMYM^pZG*F&S(Uh~!&2Kt&xx@5L0JWlauvr%V{G+pnkQQWbLFN^{=s
z_vRB&#M7$J&Re^%@fa89hCNG)z(A;C)N*99Gm6lc0m!0C&$~T~&A=w?DGnZkeE%~3
z``?Az|2G-(YBqdf0sYG@K70^Yky5{u7rQb0yb|w~>mbd=Zvug-V3DHxJ$*vznAmzX
zZyr>a+Nu$q09KS|747(O?O|lHdTJ8CFwcLVB^v4ISnRwnE#5YGj^Aq^C#<W}G;g^0
zC{DPPu*O%Jtqo+2FVgRis6<Z@l%(w%t9dZlkD@mM;4L;g+Uv|=TE0#2hds*OvG4U7
zZhPK%AJk0aW>J&dwDyEVv3KM8_;X_|JeGvFir{~BrfhGO^;f1k`$aCc9HYoE<CWGh
zZ`ZY^_(fft8g-@r%bHiGpH)|hV|Hd+7DSfgy8&IFY*iI8@qy661at9x8qph&&!2xc
zPOKRrpg?@)4SV64qpVu7hDcfD_%?hwnEwlmc~+*LY?OM5^3yC9*8Go?owKvcxVpxq
zbI1o<dLCzc9sy^U$jDzZGD~z^6Zwv}Jg02AY#u0icbC}h0Zo6IOC!tZ3#`GeHu5>k
zoYtlB91<Eq61h(CKU4S)x&y+x%6i>xBG*OL_tg#YW1igM+=ckzm<}2lZmXKQYFYiH
zERX(;2ZQ&q3S`pm#tcdnb<o1zDNxejK6!vJ^v=#gY64kGF~}N93nxUr#Ac_^&Q*bF
zo>XQ-^oE1Ma&Xu2KU8`sqKN?m$O1f769Iu3N&zN&8UX}7GseReuGK1}&>Xgsjnh=h
zEgfbP+20Ucpiw(kDR5mtfy-#HpwF*Sf+NMnH`QG+xN%~-zp6AgzarO;PNiUJHouU2
zINK2#fW&ddVf_zcIeY>SW!p)+G6%Zp(0I2hS{)Jq-I8BwU7%d^Qfk-7W6+J2RJ;v+
zqU26v%rgcgpw*t2Z$bc5#Ol?$1SW%2NcZ<{LY*@b^_na}(kORnL5y&Yya)LhB#nX>
zSr4%AO`4@(C74yO#SmOXoKoq4m-sS(>1C*DtFFprS}tnOrdui0j$ikDJWsYX3A~#7
zKDT&zX{{-b%QUxi?YG=^-Tv7|(0EpZMJ7q4E+r}>CP}0DctZZD6hk~^K@a&6dqwcx
z(QYVvnd=o}wc=K-_M6RPl7<>&xk?hK&kSyjx_p1SSD;+N5F*2L`2rC2zJDj^P9&uD
z{&y$-5I=g^I(|g&Zdu4Ny&P@`6Wf4fxz7bGQjZ;nqwIu`&WqtvpO2Vx!HF^^`*|NG
z%==?2#PmCTq?RUM4EQPO{QWn^U|F)k6f!O&onq==7v3S*V<ZAS(0%UyPa}6eP=<3V
z<Eg~Pxc}GmMI{GeAp_kKP9mH5l}06-J&E9lxK`v3%7YN*z5}vo$D&rg2Cr8DdGc+-
z_xi{^&#t<xw<g4S%b<`)#r2v~!6}Rll~5NF@CrSL)AIXII3zrr?3DiF5|{cPn5!<l
zo($!aCBr{P3H>K`@Q6=a=lOqf2X9{gBlV<dU6T$Ec{s}Xax`P3;3v{0UZ>YPi-~E=
z%A(Gg8dQAHhIVrGc3AW%Ea-F7Nt<*GdAGBDWbZ-CfX7UZbC}MEPj@sOC@E?txycFS
zDC_J=b{KnP{|0{_vY%xH_xF}%#!t7z*I5v&_r*n(%tu8U^_lY%Z1h2<rP+8M;sXST
z(O3#doRN>rG`G^94o~#d)e=RJ)Rt!rzKUahxjyRdUY+eiGdaZTZE*l3YSrSK!jpJw
zGC<A-yp+hu=IdPxlJA2^U*mSt=6*#o12*7Y5pPRiW@u<&Sc?I@6jqn}`FEuErm+T-
znd~V^-MF90FDux}#RnRM*TGk=9?VqT-drCOs^B4vH?E7B4f$jv^a$$UCFP?8f!<it
zhX1YyRt6t((g<h4s-I^A#4*F?4ru#oXAE!J(-#K^bvDtvl)F5IjUt&|B~A<(jr<Ot
z`^*q^rxt$eAHB%|ENz~5{9WRnJam1+?j9sBcci&^?xyI3)|H*f4YQ@dSqw>gA%0H1
zZUuw-xy4IF&+@E;3~0fq7w~%%sH19)Y^&8rLclOXZ{P2~Y6Zs_2X2H2iVSdKvV934
z>zQsmlmEQ%a@Da=(jTh&h;~7eY4snfUq$%e+tak?8m+~zFTNdWU2J9AQZ^jUpS%kC
zLsg|%bt&!thicnXf(5BC7a_i1$E!%R92?9~F!qfQ@cOWKF##;G6Rc|sVWp@F$}FqB
z=iRqCv6#RE$z-ysMCmRzp~V;JVL(){)x^%u%JD9co0^mZx3|jXEBke(UUydMc3+Bz
z{0c#YQY(<}9pJE^o9>7uz%Bb?V8NCenxd2j>iVMIovn3VdCcrwmPUwxg#o=q@z~G7
ze4X#qTV1K6Y(^|>yPn3umfIX1vsZy)5XlZuT*ao+#Ezfu#^<Q^!Jz|5V^xxIiCdGV
zX=q$_AOQq)bXahi>gg)jEZ-HIL~S-!JldR|#vM7uWO2MU?Hq;HNwI7Wt&1&=Q!WUz
z-+S;bB1E8`o*|?UFl0SqVhA<{S;K|CyTx94>#v3vSD7B0ML59OtG9FpvfZBW%25-D
zk&p_`Zjze)!ab0E?UK~e5l6DoH?3e2vzp|0#Bo*ax}^O3dyYGo9ND*RxI!`elM(Py
zMcNC^OhQ|m1VT&OZUh9U<)|{Xm|Q^8X{1WjEdSP$RNSJ7Pr4NqG#D=%@F~6`qsaS8
zP(`J1*Lzd@vs}ZuykA9$wy}2OWP?=K#l4CY#m>&Q;HV6?8&AEYx+*<ubx%2zPS@oY
zEvzj-$kVDynU^&5)1Gv3^&5kRfY-X%rV+`vY3r$B9bdD$yc7biu}KshmT($c^v9oq
z#!3c7ONG}I(=G~1xX#_2wW`0|J9&Iuv49$%Avgas@0!If7jamywNPa;=Dya*yjo#E
z7@hYLgWw?+KKs^pqt*(b=MAU?#ZS&2kl*;jywJ1~*pYN;&lqg2EjAcdHDq$KZmL>|
zK^9zkL|6pI?+rq;4(00IhWDtb95!DNLjXg{BD02AAH|OSl^X@mSfD9Or2A_y2lNO)
zBQlDC)3d44MX6T>VGZc6sZB@C{~aTs8&}_)0d=1ywjwsfYibV<s~V4Pd@|`2NGX`J
z9UwG<9)SH0Og9Y8$~w9VPo^rH_dHpc3E&`sg9-rAI+#7aq-2dX!{<Pc5PVmOj=_gl
zWK$MxEJHBX1DkvYdpDkE`>j$Mz00D7Q01sci5m5%U;i5h{;#_-heFOpXpgm8iS9RL
zrp#3xQ3|I&9Jt#m7jG=QjJs&J%sySt*fx#4R`X`jp<fz0U>37-sx}P~2b7IA1V^!g
ziwlh0s?eVcyU7DJuBJECYvcVn;j`XoIMP;RO?=NNTj#E#=MQwx>ejq(%DpkTY=6zE
zwVG7dPn+76yo0q2oyAOai(>jg{OX$N)UdWOc5J(d!J|Ew?a)-+b@^&?|Ll^<)YFFD
z&3x)17KU8=N#Rr|(8D(Q#}`7cPZywB#s)=!7cxtYp{ybdIoCyGQ4K3*JkwGR9da2W
zq8?#?sFtI)M5Q)=ZU3QaZOZIN0NIfZxawrghgZNfWNvtWzVx#{RJ$_FqW8W~T6aI)
zR=t0(>L+D6qr#w&B8|98YkE5t@%e3l;UduKd9?HTh(h1>K?Rk8%(cig<ziu{p{HzF
zEDs+fY_HeoAINWdTt&PBP|GSBMz>~=wR&g^UJ}`v`1SMb>c%ZXbh3LG^p<i2`chUS
z`0s>J8!8OlNkO!p7WcROp-S7=y|5L#iC(zW-7tlkp5Zc@(0`8{PvBxEYho7y!|2D#
zflEx^_2hUJOdFQM96M*Dq}5|#18vsNjk08&&_y2!pZOSwDDsmz)rw4J)iPuV4wUWK
zV54iyOsU<)>jA!~^S3AWC~B_vgOTQllr`WZJP2F6?)I!ZHRT}-YGQkU{U9(cZ)8)H
zZn8ta!X8*rFnXijGU6(xudKY%nokMhCCCy@IJD~ys4rdy&AU!NqaM)J_Pz&`mxA%%
z`2Gs%(F!|R2eLBut0g&vIixr0)==iP`yNj<A#j3BYu?(C9o*8LRbtLB2H?Xnjwh6f
zNP~bSHt4+;gz-yu-U8O9I*2vVn%2O}q=id0zfm6q24C^ieVJ{C%S++sqf8kdGB{S1
z@mlO^v%QJUQ#icuZNqr7iAc~(2rzWx)|IFbqs(}?vmMX_Y3Xx1QnjV};i}ZDZhELi
zTrmL0s9&u4p7#o+fsG=2lOsGn3@<6wIj3X`O<eHd$oBb*qvjW+8v>0k#`7Ph8{2xT
z-HQLlH@2Je6ri)xea-iG$?LUdrax4wK-zP6s++pL$EI>+93dqmQ@u6WIl2l%Cj>2l
zvP++Y+E3lpV<PZR3WS7Y$8YJU5DDv%5q?2R<eQi9a1RF)dA)r4%_#D7ECgpV8mk4`
zydPmWbyQyGfyq}1C-JMly(2R09x=p=s(ZGGQ4tfokDpM3N4)ra6PFhR9Ul-(oHg_3
z`shaJprq^LCyjk9&^1mC<HZh(Yh`2c{2X-C2uzY}BRywgD!+jc;bC;fHS)~l@!A&c
zwV9cWp8?DXZUg@LORVs4j03<zfb&Yf?Ux|M@4Kcfi5h7Tc6Rs*sAJKx_&M08UoVQk
z%}VlNdx<DIZo9T~XTkz;F)r%_ih1^KDYY`Z4+OyQE8S;zo$<3unL(@D+tJW9cp!3`
zREn;O!$WO)c3K`Z58z_QJ~${oyrd<>#$V#D=j`jvwI>WU_^b^YUvoN({gq`!UF5Gj
zqM%4Hawqf@3uZ!sf~RN5c$0w_7qeJ@I^wlCpUkQra_7pyt_W8e>GChrK`}gE$fp@;
z@Nz&G7A4a-v*nUQA$80CkKy#ch_N@aJoZj{7|!CL7YEm_+5d|kmmk?zc=9xEew5jN
zA_NQKlCx#=u$r{dNw{4urbQzrwA`Iuq@^YyJX#M$K$yzfw0MDZLc9qUuRJ$St)Z25
zIoeXbV`BrnV?%ooWqfq31t}M9WK~_=HRG+RxB*0K@m!+mbtRsGbQD%?`4inDZdQ_1
zA)Q~K{1}j=<ky!sFHf<8O8gUugLI?HUv7?gy~Qxab>LBro1yZ@PYW@e<N8nfpnmQd
zws%A$TsGfsEu760C4XnbuC@*Sq4H?G6j~CVQ{C0SPdzzmu<<(gb=HLH8U4#!jz0^}
zU8Y)pPrQBIwEPChVYUR-EcYaVTi9Uqelgr^p&*E}Lju^GDNiG5EX7@9p5h=2-VrSx
zug5BTCzeKK&`Bkhm>g;U;3&q&wQRiODlvvhST&dT_fSAqKk`1hN%Xr8-4>5^O8x|b
zh-_?Zit%-N-P&T^!*EwQX&dFdJ1rWC4PWu%^1|aln0vMJoX}q7fX*qn_N91CWjiMT
zeMcK<nnssYtS!UKyR6lHqi|FAN`BdyfahaaWS7MJ$$Rs{5l!tR4qU%e*4^lR78xEH
z1(2+IUXpJ@^7GT@HyGy-8rV3sZ{83I;<@B783O6>))nTfJZ0|UUo@>BPEVF~=*C4i
zR;7udraznkg-G`ajk2j?v2D6Cxg2u`DEDV+`D^KuQUdPIGJoB!lCox}Mf8QJhi5v`
zs`1Mg3-x8u>qRyqV<Tg~dVm2skxanEQr@Q#TFlopMR>*leL#XKu8o1ZR9uU%dP==Z
zpjXjC8vybRtX<u`vhn6$Ek4iCm_JkyU-D#^>a+t@QsTerM0LsME3F#${FuKcWA%qh
z6exCWYQyc2bF=U$>+7XLkRg?iS{wJ_&he#gLq13Oe>8*Aulk;iL~Pt2N!U;NuKK1a
z`s(*7oRUj&K+d#+Jvyx0uL{O~B}AC!lZWtQ5h1TE;|F{=&ERa;IL+WjIu@CYD|yFP
zV9VFjwwFh&?OXXu^|eaOEaFwujg%ain%SdnzLpLaT<WxnLX;i?hs|@DMkL-LeZQ2Z
zrtv?b`TX*r@6vKMN`l}8k?${)5e*UP-9CA6z$mTLCZHPNS$14Bj-?l4<$vneL08bQ
z(NMFfyuo9vAuKLD*7LOG0>*I2?J9WkOzRI-MnELB-O#1}(rvWpC3m^&!QANw9NCrE
zXbihKh822dWZkc0bZJ2=F(SkB@P5RyU7eUww%*ZOZLcKb(aAB2EbnVYrrZy%w3?`-
zr*1UOf;z@%I(1>N@dRQfggJv#eXax<@ZtT=L0whj-1|~LtW1H$nE4XBkM#8q)HJ$b
zvb=}AUQtrTkBe(hC2&P({H1n8Q5L&i_YKdV=u?Xz^FLHFm$F_TNPqJK{@qmmFT|0;
z((rG(9a}(3BXYguz7J5wPEkdb#u`h{8Tf=h**g6xb9F)?2}s*qkW-+gM_x~cEfL=M
z0{#+{pSl=5JbC?U-SfCZ=vMC@Pb*ez<>ZB%<QccY8^o=vc^)#gtSvJUSI-qIy{P*t
z<qoc@#JaOd^hyYSa4?WcF$eVcn0aaVKPof9$EjGkQK)qdp8t+edq_nk@mcqz*6g_E
zHTLg2`<K7#ruY+A&%nu=8!K83p71uOOP`r3OIHCH$1yRxT3QLF+$9u5Ju)r)UP2}|
zNB;MP8dt|gaavF8ey_Rjfsx|mgJ@z@Xme=nuPZnY&3i}tg*%#^{L^)FA1J1>{Vit9
zv2ilAFMjMt1JC<DON&hxEZR2W_WK%e`_XGyzv<Ys*r~n{#rUiU?}3Knm1eA>#G2Ke
zzcEV+lSidhij{6D8mBn#T|wHbdIe3+ioye}AcSKV7-;4kV_Lr4sWIu6I32xhG?6z{
zPv(ruT6>beK7(sDqnr9=f>mQe_p=86cFb7n{N`2pDMm+GF}m?Ug30(GZD?2c2M4@w
z9r~-m)y<;ik&5z}?@kE<=3g(q)Tpd5F1sgbh{web6L;%HQ=y}K9{tnb%GX$=fHXcY
zdhv|2ghEM?<*bAxunBbh)!|wc=49dv-KI)a$#|ipUfUQp5Iv&M6kizUDUsExiUrHR
z`pr017vnr_46rJnZt^6ltIsq;jHTFQUQO}(fty>TyLS`<#Y+^yrQV{>Tr8FBP9I*(
z{q_(~;;>$K7aL@c(;0KoB(JKOnvXqitS<T*XlL4N=-!ZVf79UVIUxAewwWI^us`>x
zNiK8(pJor-Olbz5Ygr8;LS(99E8WeBo-V{sFeY+&`3?diB4&$C@gOW2XGgG9R~>Rx
zw9^ze0oIR+P*!#iUvkAN5|tm}!XnE4hl-m>@)<RDmfq+vr1Rkx#6*`O&#fn$0n-J_
zXM-0it?*muvFVDf!sl26&Y?|}+H6UUz2IT&$(aXgtAsyWE8Fhay*VYuD=2))Q<X!y
z-Cg8iOH^?K8&j&n_W-eHNp;*sF~<8oI}D<BuUhAB989w39ZfYf9Qx0^{kUY9QtteZ
zGt?AFbJLn6_OoM(Q`LHC73=?uBvt1vzAj^<nHBSz%R#3_w$k&OyC;NsrUrK;kWpqn
zT<u|IlaiI?Q7*M<gHM9w5cG7ae5YHYQh>%8Q4pjcGf!+xs@X+%N!A}KBHDhq<>oi^
zfusfED&vUWy|7wHo?)i7I;z5{qu7Ou;hv7yJ%f5=kuZ`HeYiMj(Y@MH6(#WO$MENw
ziELi~<EVzmeyC1zU?*kDG7ch-)Qk!XQj%!0hsI=V8N;$lIOxBFnU<Uvi7pKEB)M8o
zODqW72_HRx$Z51Um%=8-CqtPsmzVXylj2rU`5QwFCf`ZkUlifw0+B{<jBo+6JTod*
z{z1?y6Z6H2&g`iZhb;CN{rl%aW6@_Y_4k&E4XgOH67-{YZUr!to__U8uN!Gn3R_7&
zWaa#!p|!BMV&ihsoVb~NKzWN2$Eyv!TvK9cB+>&M=Hg8P4}5tQ!+J2=hr-w|t9(&%
z=nDnDg8T4zEqH0D9LJCV{g$zT<%i!PRX2Dt6OIo$m{%@nywj!x<=PQ~jGqSfPldbo
z+$DvooXHP*cidd%>RAxAw@%8ie#ELvG_Ycv$WwGg`g5;4&=3|5ONJ7~!?zeX+~Eb8
zobi0Lp3`5ur;65i=pa}uPSxmZFaErl?Nxy3rfrQkk9{d>iSfV~3^eWs8p|YC?@0mg
zx+!ET$jV4b3&r51y)Z+89qWhyGmp*2^60g^oqidWdHbO6)%uXno3;N0MSPbn%z$B?
zxPHp?0vt0@dD=z2Qz>qmi`L;VvEV;canCLUucyRBJJOcPn*Er5CThKgO9_-xNdnmZ
zb!IkY!?qqO<=U(&`C@z4!-ZJ9u7n&5O#oFYpLjyk1vw7j)0#7NNw%0V0NG2v32`D)
zA(Dqg2{p@}xT7cV)z~RWU0Rt{9eR1<L{pZ=)EiZSbIv|E8MFmO*SMAxvY#8TOAbjf
zuVnr*Q0#cSlQh>}hMYCSf9CZD95uzgPHJo4Nv(-o*_aQ|v~ZBGuyZ&*^x|c9VW`cL
z4vJOBq!MeuQ_oY_qTTMW<m=Fic}PAT)Zr|vmOj%Zorw>O9Dd0yUUWm&PmSTG0|$}{
z9UKNsHH*tOPXk*52gGAoN2glSV4sEi+OV)q!iGCc2${ac5>3AIVoZGXRp5EW#L=&z
zn1qp-Y}N?{Fp08S;~-t>+Sv#}5R5%an<*F~SPHh}BXxJi`WepE@;%1Vf)}!bJX?Xk
z!aYU0g=t`I%nUdTc5*C3RKQEQahn)bTCQrL^nmMPHwp|VC-w6sI%SxKoSKUHWv`t8
zj#rf2+BauC+gr>8J&(&O)|!~V_O}s+_Apx@E3?i4!pM%4?}m!bMM<1ojqlV(^e-6e
ztIXpuFOR9-X6MYPyxFtH$=CX0zLVITKpEK!;8}}3ksKgf_N}Y45k_23n`yR>VcMT+
zb&f>!ld&X7FBA_l>N@>rzN`i`PXV0skuUJT3>c$4&Qz+VLQM3<<KJcoCTio9%~EcH
zRN!;dA&1(O#C<=$-^<^PwzBTCIri8u?YRB;Lp99tjixH@^V#!p;A*q;`G&RCBu1hI
zchGDZ>ft`3FUwWzsZQMcHFp{B4Ph}E;vw`)69<=E>qZ9Md2$}#Mn5UvChOBWU0N=P
z<x+dxhe3(CVhpVXoUTzYA9HO4Nu3&lfPvj8#qt8-tv-&tS?6I{=sk?-60nbML_sbN
zePSJ|BO$HaIVl=qZ=G-Z!HyVk&sLxeZ)M^yPA*c!6$>|D2<j15a8u6@GYK|;S1Ye2
zNMWJe>sGa6*G-0U%PexlCDkrRUVS7krD{DO&?Cl8qTDT}eh!?^hM0)1GPbWUF|+my
ziV1+nm1#8LZ`8z;5Ca9^<`rYJea(74{eIg!@Y&w-DLs&kpV&!5MYMB6xxJt@KrVKc
z2Hm!Vm1oZ1vU$hs^4bv$Fkw5C%7HeaS3YCqZt#Z;c*YGac%v*EQn6+>(?b(_$@{Ow
z)f9FuqCbC{r~TMkEEgUq>oGr-#{`qx7nfsMYYH#M1L3Vr(6-Kk%nF`E{d;fADeUNM
z)>Hy1BgN*L{Ww;G5M6c7MeM8JSCot4xC-gThP`hP8_5V=w5)TG<CeKPMw01ZBsXUz
zxNwzk8Myh)AE7a2sK`PL6y<aurpr@*)AxZPj<>EdiXdu(zaF=cwJs)m7JHMcyl3yM
zvqhR$MwVT&?6IO=)ET&EyGSeB+QPDoL6KvlH?qV$0MAr&K_AF_qINuoGOl`sn8Q?*
zZT4=N!NwLw_W-&0PmBE(B&lD-5|=HpbLp8XP2j1SnBzbK`-g>DymvpvwNb;QBdOuD
z=0vvX*}R+OvP$fDB(l2FvU{KpPXqynSSsf%T1^2Y4HJ^qaEP9nh9Ou6K+ql9T!}HG
z1r04RHV7nc88hVOIqDbiOE6`aEFrxT$rJKWUElpmmwnbntJ93s0wH25gE&87wc#Z>
zL|oo=Ls@TMdOu@3;rZZ{5~;1lh1+8-6|`|4q|rZw*USow(bFU5=K8}amc^S&HzMOb
zmME9rWE0Dot>8vXWJScY9XSUN<!t#to1dLkvS%H#UeajCpSCUD!D!18*-jO?twGf~
zj}9||FV3#f4SBL=el)IkPG61*`0JOhk=rvN?-0IAD(PDB=3eOdUvl~1tlcD`U-FIn
zU!6wQNM4u}NMv?|(vapgS{eN`Lhox#?8xT&v-??F;0qzmC%eLp`yMet;mCa*%G&yT
zu1<7~%f$l&r*Dmg>bY0JaXlj^(J!|`>nS-%=!u}G1!0chq#od3Bjq#w#9gbsSGKnR
zULKeGru_EdqJ^eqIoPakv}~w_L4+_Uvczl3^;!z}d*Nii=j9yV=|ehcNdxVeFUm7J
zW-l7o9<ebVe^S(pVcTCEpKLgC(23Otgu*w-e*5wX1;0V%A;DeeUNTmd<Mo^De&?AJ
zD%_<Gf}*4Q;8L)`u!_s6;Fl7ajKRu1N?yWoZU%1os0GWJ10J*7GPbIrk=)xzZ_<IT
zxCLccSCh;Q^vc?#6xTP}dYhEuvZdYU44WbBIumg`#s**dJbbwrUtL|`j#mhHmB+Nf
z>E?HS6(zG*a}iPE+8`&oh7@yNKN&F1EwEp*Mpaq3(O-Ph1d~msnqmEZn^WcaRWG+L
zZDPd9jSaw9h!V+%0ykTibb>^#AUEw6tXk5%r)4;S5EOxoDnlX5qst?6Aw!{nGXP>%
zwBlPey8nf#Q$~7JNpaceT`Q##Ke+Sfxw`9_W)nCW?O5S;E$zEwZQh#|Y(_TrkDlrE
z+|}|C|JC}GR;TseesKS{s#k_(POFole9J4#@D^K759IcSp`LX&<|6katnsAYFiO2Y
zz-;a7C2^r#!Zv%){&IVhkp?lL+-?@K+dFbJ*#UAGjSAVinmXU(%ae5n<Irmvt^rkW
z&!$II3yWwuv_R+}%~-G3byV{?R8#J|i({kF!5lCsylnxC(~hnpDKy@7kGWqy@_WhO
zCk}rD@x?-0*UU`<e109RertTc*}u}Iderq+@?y7!pfV*dmg|Y@{&9dp9@{!zbjRkq
z<8ep=3Nn1~xxXhto>j5fg7w%^HZ;_kKgQCpFd{q2->4=`Qx@2}wFrh~voXMabj@a~
z2_UM>F!RP7fvskhq|lsL-B-Yl^@be{e{c5*ykB|Mcrz^8-OSI!j1EM1x!Fz2e|Rm=
zK@yid?A<*O(WK^LDIS~IwLG-pd`iA?uj9$*@UENo*KT!73o_aHHZ?6csZ;pToV;S+
z%0lA@b<btAN-P@ZfK$z(<rSp*b2-FDznXa$o>Aj97G=M6Crg0^7KL=+_$!BiX+Vn9
zfvnZZe#)-fGZ%Ub#tj8i2fibL7x`rqotS_H>}jU&kj9e<%s#m=e#ea36V+JKT`nrQ
zvu;VA<Y=H1hhoY~H0nxm6?tv^Pire{KCExAe9dMXTU5_tT5;&H`>xy<1BV)}h5Oy}
z%gW9`QOpA_2kLD~5UtWxC9W+24`v%+FWS+5X@1er<3d&=rXCDCh#6aLhAjaNApjx+
zUCO2nHVHO;$^xIM3yk)#XmM`xux$2dvM-mgYE&m`_?qc;n2_}Pf$INL^rubhj<8YX
z#)>!nI$wX*ddv^-_4q@zdZ~mR>iIt3Eb{d!<*1$M75QS>Dr0@9b@w8jsDDFAaZ(oj
za_C3pFR)(?UR-ZLeCyzCwtL?TUsb(2Wz0}{|33qwqE<ECwaWIrGj5238XnR&^-FPZ
zKGvc6$Dryyc;VyFL|cwC=tN#x-esLK|MGs(9LAn=|F|Q!>&3CHM>b)F^WcnH*$W%+
zt4ZgKiupP73Q>`Rjhtw-d4cC5q&aYr=)D96VSg?YCr?TuCS^T}ar-n{(wmz`5=+ag
zCw|VLZhX136`n<M(P66D+6oT>7&4iu0E|I^3xYbo{b_g!t#WeXNd;#j92{9|<yFF2
z_#;lXV<R!+KD=~c%0Yfp<FlNih>Y5hi}aowU&G{&d82}nex>1xVF%q_d{)fDQLLP#
zWjV5}@Wl>;FXgxnKxF!Vcze&NrqZ_k+Zjh43p#>;6k$XG=~5Du=BN}Y0V$z{fKmb^
zBuEJ*6h|FE2#82cLK#|s5IP}*fQ-^h5D1|qfOH5=q=*5WfA0T!KfRys=Xuw9vsOOk
zO0xFeYn}Vt=k+^|!w~v`g|^@*jqF@z8Lda8Iaq12Kz1zkyNci94n>Y>)FAH-k}pY4
z&TpkD8ZfZkg=pMYff6J3@rd~NEg_+HnricK6H1s9RWE7@q@(?%L3<`zLu5?!xpM4z
zYFRbt!%WSqMV;`7>)-u9b<0?HI|sSg)t>RPinac7C^qap0WO%csuPdi8rW?;mW-C!
z%oDx+ta3A_FyBANT;f^hMc^qhHnt54Txu;sAq}R=f!p=$rkK*4GmghlqwI*CAV_By
zB^jG}$DWe0Pe8Oim(X$%poOga^4_?nx;HD(p<MDTB6GE^=R*q^%m`hru{+k37-Kjn
zBm|%wz<n9VNa)z5PmsWr&`L?~FhAsuD&?P`jsC|L{iVo@x%<f{J2i7mckmIGvb1e7
z>RBZGn`Dw=SsMhH&ZX(zvB9OBdiEwVw4e~Vf97v$@x6FwRjGM)^cAjsu)c|)X8p#)
z0604W5v41>sOhTDak<8yK=5+dchdq~<t(0b${qsP-F#u`&?ba#e|J>Raa>#q%u$|d
zb8n`L>VwzrIDW_dAT@bo(>V0F<OnHzg~fJYGGtAGq*&myRzULbX<)Q~$o}H^$sMFK
zGajCeIb{qwMtkR`eFq4_<crT4jY_9ESTCX<FNxh5xx2IK7^C`7u+Ze~`s;Vjh|336
zpYLI(&n!Rn{h{cm@9o1GDOg{n-F<F2FibgHqvu9_Rk3>4IW`#=-L*azAuVmNcnI8d
zqK8ZYx11~K$^NJ*YSa2u3w6&AX+{vTaHIN<I(OHNNe*e(1gcE*P>rJI64oeJ9KZl3
z$QCSY=!e|ckW({9zVjRdIPTxQ4*@Ri^Y%Z=f>czD{6gg%C)v)QFnZ$L$%DieMZeYL
zI&tNY?5J;#nQ&_lfqRypSan}^Ta?ad+K%igmtWP1&MBF9V|@C==%pJay67qE!BrQ$
zyga8Xu|V~8Nae;Fg^8LHcCa)2Hdf`-Sr>XkW{)nS{GSiLRa(Ot1p8Hdr3*cfM#P4!
z5amc~m8C?yI}9dsSw4+lq#^N(s_v=U`ck8YPRaMHn>u#;95fQ0CHhN;QQ9d(pX(5)
z{KkR>My%EzN7SYuKtrGsp~4uxXX0J_qprx7JpLtP^BTS3R(Iz=tZ%M;!+?lg2gweA
zFD7=jtM8vy3#*NJc>NPkJ|4y>t=f0&lJ86w*sk>!spt@e6KMr>{!F1PHOa`Doz?1n
zksex#`&uRCZ*36ZaE*OTO&Z<!6mqAT(}lu+h=q4tvx|K?r6gAEmRm1aRLp8mPVvaF
z_@{e&tOVG$`H#_}-6>12OxM>|T9c#Bvvc0B-pe0omwT+_%CF~y!nk4?m2UqeIf&=w
zbM29!E7SC1ZWd+17n9SBse}~DRNT{)Q(vPQi7q9q5#_!OQdM8kYS#J&(JmlMMpFTy
zE;%p?nLvtPEx94JVEtSQkpT0-j+U;y#+SvvUM$SM|M?j?lYEE@dVUdn!x5KN$1Zbp
zTqA40@h)f%H{XQS6K$`qzuL9IU+n*IvB7+$2-^r&PJW^<UaNmwdEM$t*=}HoMu<^A
z#-n|;v*XLx<I5g<1|?T`li!jE_?%W=icaS7UnG3vlJ{|n2uyin$G;6jLH-;{W+&Mz
z<&Efv0`T%xt9OoeI|+6kpAO|mo{w4cjr;r2#Y(b0HO*>LO?^S<c`mlXT}m*h=Ur73
z-*fL?CgirHo96ij1b807dA%#vY{*xl#oKnD^?Xa6x!=;o;Bx6jT>NJ7t3n?<D!Qhe
zi?e?gJf&pb-u-C=eSt$rK}cUZb6#+YlT9xFrbwJ3h^UN(ETWCHYylPz`4$ZxbB-<P
z>Cpo~png;3Om{cUh#-Q5#FOY$6bJ7^MJ2eyNNXeo;tgEZ8*7!QuBsUOdudf7DYOws
zo%WJhMoa!M!~K0n!J{|J_EG02o6p&qJ64ucu!@)TNgV$@cRVQ2ku#;c8_avcbvDxJ
zs5(pPELbFj{--o<O^An7VOYJG!W><-fXS9P%<EiLckaCheZ;D9f{;saP5PF2B(NrR
z7f!w6pboZkzzh&b)13t*@8tcpFtBYPE^nPvw}!L`U0dCJJJq@w;ZHAXcN=KfDR)GX
zJ#H7v31s&>_vaE!^xG?te{`2FQUwnsbFQZ;8ajqjC|_lbWYVvd+$d8mcWO5MlG&d3
zE1*0gqMd=O^qX=%#&x&XklitK@mPQ+2(ZAXP~a^k@T~ncl8S3<$)e{ZC4!)>RIs6G
z=%yn#H~39X#R)G`5??ky<Y3fWQyB!sFXIx_CP)gc$xv${u+}scMJnSv1NvrVUYGI>
z3K173fA8dI#hFtS5=*q$OHox<@a%*RQZIrT>uK3ijAmI_MNi|K3g%38AEtFCD<(kc
z*2C=|*Vu4`iTvcu5cu=V>j+W*gTj{35EiY-$5nQt^^)fP#kvU1M?E>A7R2eTo_<=g
z|9M?`?b9qfxQ-FKfpY#1*oBZGhh%{h=0+<q_fbd*`~Jb;7Tw2**%jxto$JURwg|bI
zA+`S1dI+(pa*_($a2f9*;>oNKh1IJ40;M;VH6g7nj7onUId5IR$vP-0K#K>GwD`b1
ziqQC5(MjxveX<RZ8+a5d?NXRKrBd*=^lVyef=t2#oNZ!iu67vZvotpM)b!x?k;4QI
z@?d|FQtCRSvIuvUD%?GEhK8kivkX_7a*IJ#g=2Z~!p4RZq*kZ}8=C?Ovg}7r%}}+y
z#LycXmPuyW;$&vJmSbNH7v7Ccg~@5utb?Edu!Sx|@MY$qOOWKaMDl-x`^9dpuh*95
zzeK8O#Ea2vpEX`sxZCOB!rZui@IJ!*;FPL$xCM=@GxGWo_0=hxxyE=4!Fi<$rPPmG
zMplok#Om3Yu{Z-!ltAdimXLBFU>(29Xw6?3Lp4=QXY0EKEU$`3qk()jv^k-=Rzrp{
z39kzxt_K^aJa4~7Xi#HeD-Z`^xL~HRLp(}t$%~y(pjbFM|J13c1;%GiNirK(Kz99L
z{>-EA)T+G1`>Rci6xQs)z;fBGT64Mxm-`tmCtWHYblR?D3g~|GYh;^9S=;=TF`Zk9
z@wi_?)8>;j8VFINlX@{N`akI_H>hUfD!A{3ytz8Ar&XMA;S$~uc0`^GgAk0K&SUAw
zRS_zt_O9%JCu)dO1$c$ZMF0zxQ2vHsvHiS_0g;(j1<uuZ%j|ViQHe(`>*o|nWpjD^
z#`XZRGiI5h4&`~{UT4K;#itjkT)ch}$ZtLKl^*;jIX>pHf}+_wSh}pKhE{1}clY-c
zoHQzj-^zitz(V{ZWP$;P9d;~>ZFM`Vu&i9BkYxM!T0091eplVzrLMlHzLosyp}OiV
z-P4R+g8rDIK;P%(Vjb^LXcEjRh9g-e9PjoH(uRq$zxVTLaIy`@Gc?5d#-nHfdjQnm
zz{W8+>nQ^1tCBDpB^Y%hSYVMFX)w~Pz`5KU>U!F;JbxuibT`R;-!d}>Gi?AXPDv@`
zI*mKS^01f<dbk+A(o#epM>FKmEWj3wbnKjJZ$DELO^}GYv<J)Kf5R2b)d(n~MsH|B
z%dc3cy0bXgF$H2-!L&Y|GIm*CT~qL%CUloc%qVfOrn(n7;9xyD?qp0Z!{vO&WWJcs
zu<Ao!(AX--F8Phz1((?yxPJb*Y;BRbCmij9d-z6PX7l;b_G$n{3(>S0`qy{oSSg?i
zr@2oeA^-{PAi1;aj2x5-_l%P(7P;ZnCIx&MwRmosVSCv~k3+(H6aU#xZOeHmmu(J7
zm?BVBZBeZ7M(0>zY^q(m;jz5D$Q0JDz2A_P68K3fXj0B*w2$$~tL%afbN&6W)YE<Y
zOsDWn8*{#O^MkZF@qlw?gY~zD>v0Nlo}L~Shyprz47Df>p`l2nG*lBTVWl!2k0@w@
zzvj2`Bsu<D#y2Y~D?S!PEG5a1U~+OY?(X6Lw==&+l#|cO+Bu$Vr7_kv++e%U)It?1
z9N%OCp0T&And6lom({wjCeDea-;n;Wou$pc%y*J?`ZT@mt$Slmjo}P?E=ZKFdbOTC
zI<@`0>)OK`mS3ve5s-#v#R~R_W=&b=has)WO(r;`g6sUFJmp2f6W10a0H-E7v(&jY
z#i(hiqdW%?;E%dC)v#ZTS3?Q7iCSW7G-0fs%1!VAF^p7xHEb8!+XB*57al_~?5s5}
zL+y41%!Vp~9F4s$)v6Eb`u(=nUZO&l0($z7+IK1d<hr-TyEmxPiVSs<pAk-I%0{f}
zg$_?PZHxog3NwE<=d8zgANA9*QuDo`z|f0zvzF2MMNb_Q6W0#ioY;ruYJwaSaBrm#
z?_zmQg@eCT7O6QqF-m7&78}e8+2A2<0Ub5PVf5Lhs#O0=5&=$isUekEJMVee>i(w1
zrz`EAi}RBm3FWym2}-S?Np2PqFIP6Jbkoz*nE!I=?(5XreuqCv8@R>3H1Wek>D*Y-
zhk3`Ph>H?JUQdFB?^;G=J~Zs)ja#too{Tu1&)UV+@B}?HQ)rJ-?IC&tUV;xV(Q2iH
zcsDQ)57@;MW6io>#=tD##^E<*XO8n5R0=g1vgcS{a2ys+fAyHDFjC=2W0X@?gO@r)
zOKHjN)0Ds|pylkg1PtkyP`y_XBrOe0P^cs;0c(Q-OA`w5G8<$i?WGQIm$0pv5CRXw
z|9hei46ve3ZV6Vr-BO(G_wd&_Ok4_21@uuu&(u`(Jj)(>?2&oAZ>a2jr>ILs)v5B!
zcuzsHub>dz_uGPuZhuoNw&RnLzjUO)hZ6g3&9)@SgTb}l62YVo*z{rq?}{g1=;}j#
z4rYJn5#<E>E5kY|r7rC)(WRAB^q`Q0m6B9c*o%5?=tS2br8%;>k*Y0!aj`j3uQK5|
z6@2x3hMx=lO86K!i$6ebXINXb59|s=y8iLsZR7v{vg&^lo9*7>JEX750cLwTgUDe~
z_5m-vE)RE5gzJnO(&m-c+(EAo`OYf#x7}{Gp7V2nM8(-Ml(ZCl9{+Kz3s*Pf(-0!C
zn!(jROKb!;Q*)|6a+g13p5;r(cM@qg7aryoBo8K*c-PK0{&jQ2II{bfd_RX+%ZZSa
zPCol7Ll%vnYOf|#^#C3!1<)p<D%^_z%3QJVvlf`97_&F+alhYx=#z3i1Nx4dl+jdO
zO=W&0m}u{j#Z^d{8r&kfx;|RD_uFku7#d{*ORwAb+u7^|8fYq<2!!5INTfJaoCmrF
zy^DtmJu+DeU-H+&{!XR7h}k5#-qu%Ws0~Co+q9QbuWsmhXHY7C5m|m@5w6?-ORJ{U
zxC997m?gQqg9(1GZ^=Q{#Z|u5kM;e_Y~7VmG;%cdSc84bdQ$^&Wd|v8_u4M2Gm(ju
zi0!`>54k-D^ySkgFvQ~^ebV>$@)bXPO^nF(g};}6QdkL8(TUO660J@OwU|F}%`mO&
z)y?Ro^<2J5eZ+AhgTgcInKswE{l<Du|9#C3rG#v{+ti}bIIRDy?eo$dWUIWH-YZ*8
z5*hMfNPI5iL#~Yyc}|>RVd`b<@}s!@&MU2}f7H)tdJ%<(VPmVCTi*m;&WZ$zEO2yw
z$vXF^pAKC6szHIFF>H6(GyN`5W2A}%xV-bbU|9`HjP5;t3S8r}2Pn?<erF7m1C=`2
z)qm<yaj&kdX;j|xL8*9!LU}c9u@kbE$2J`6K$#2ZR@ciA*O`uH<z+3{n&n>Lg|I>T
z%I7#SY@g17@*ZHYYPQ$xGe;OMIH^=@@RvnGuaIOOM4&qMq45_r&GdX_FD<IhT4PS|
z*m{uUTQ<ef5{!Bsd~9oOmY_K-k)r5f?RN8ene`KrhLlIqjHx8gQ0RGCMBp&gc2J2?
zx8T_B$TLE}sw*k&B)$H-#_G+3KN)BXjE7%xqk~1iPn=`@<!>5uwMCgDCtg@c^bh-3
zs%N(20%-b3I<x8MjjE}{iI<BZ2AufEx!O;&vCfM^HA{x7FK@6vKBZp~)G<Y$x5oXF
zx_xRkz9jp;$I4$WpKLWj_Ye=s6=dIJu@_eGjS03)429yeF)UKFr;z`oXY}r!xjeLG
z$h$ut{kUr}dY7L<Jv1Hi@QM~Ixc~jOtm=D~i^%T{J7p&P$mz!)?Q5+@-lciw4*Ve_
z5$PhbdnS?)-w`T?gME({ddp;b1c&Qi)O0eEE78SbCt5XH*j;32yM|)bmahd-2VB5i
zCvgd*?s?$PPZUT>HM|6pcv!{#vZBfuTc@*W@0{Y@pkk$O=e6F++~h}rL6E-~y1aYG
z&mnn_l8AuMZ0bT!I;nqu9Jn48)7h~nEr$Lrkb7r9$DekhNaZ~2V_8??6f}60<CASO
z+Kab+BmPnT`gg)L^>jI5V~=KmHm|0jpZ2E^rwB49(VMq>lPo%JR(zZLD2L4-jFvRi
zi}^k2+HXP+OLQgJo5>s48<VKa`AhENZfWylyP@XsDuIC#cb%?1&lz<RD6W60?Wa2B
zdqJ!leapY_8qsE0+{*BD)ic)YJ)(=YAHa3zw-LQ6Yc+tW7Wym9Q~im9ruD-&-F4~E
zpOwwN%FVo+->;R-1nlrJWhS!tmTd4#$N#px{%=I*7re`PMz7W~ehMez<DEAuRia-X
zPl(0)e|$Ug+C~ERvuF4P1x3RwTQQhjkh6=gn%3mVWRh(64Y7gC_12F!r0&u5VTB)x
zB|79UC+Qo6N~qmp1kiCbmHSa8XN@hhUz2QOVQ|&y!M90tB-J9M>#y>2Oy@2~ch9s>
z&5gOMj~d_Cnk#q~kDJ{rVyHAge;p#+tPwlsr1j2*<J@-H$FjRd;MdXP$7ARGm^YfK
ztg{nWzVBu{TXU=+xthG~>GmuT8hUKwYJRE5E(Elr34CYf2H5%OU}`<|B7KSxvN_EN
zG}gc@liVzCa&%bMdQI6z&G&QDD8H`@X3EsZqXfandPz5<oWB0fp-*OPm2Ub>qx}Fu
zU=O>s4MDqk6~}gr_!Lr!mkRL>8-3?CBL4dI;oU!nkEuxUOxV@f>h6EOMjUawG`tdD
zT&~I=GXyHDHyw7|j~5--Ey^~eBCl9Va7K(Lbqs`0B&cM}v%{xiy~~}JwgR1XRov>E
zVEWOT@0@GmjNmmYDyq6arQdaluY213@}))Dn4Ga!f%yh)y!L!sl&<TzeVXsZ`_Er4
ze7Qw2ExPR(t(&}*=&8mAxn#T}$_>>YQ9V9zO__lmdnU-24|v6m-;9b+uWS&EYNJAD
zC+-*N(9mzK%e6~^1rUYYmDZ|`dAE<$zmLtvh~@ywPq)0QmQr8tnEAxtocW!TMDbBk
zoeYkX9u_~()_Yecjf+q(kBF7&%99NpvqiZ+B)8q{GII=s7g7dQ;mG>3$PxHYB+O2m
z(8IlC)%4P<$CV~U6KTsixTjt*DzS3|wP-U)#`(X#-R#WZr4_#2H4)banK_jd!C)e)
zKYfIltk#BwTb6o%->>PZLQ~zxyfkfC;d@GJ5-4Xjgeltv_^R!pus`Xh3jM1%=&XI*
zMU}`$)xUSaZ{AYRiW5}&Rc<0lMAH<e7xFmQFc#<gE3!V*-R7OlmlI$Ay9VZe$Dv-{
zX?Kfzf2kc|`ejOfU}wGDO6d#Zep2aU<n}FLpir?@@|&6Yn4cetthfXwG*0{y$MBH}
z&h;v|u_FHZ^o-9w!C_@Hef3AW-FSZknYnJ6DfG(s30uW7Pdr2DF+NkFalzvHx8b2X
zbK<7Jbg_5lZ+|_Q+&Z_RGr6=4SPN_1FN4vHZL31X(*uIyIIJO~zw<+F>Zn2V(+Iiy
z_Z_=Y4IU+@#fxu^EV#={Ep7Q+ptWm1b@cx2STujgZ(g^jo>v)nk2G>xwdN5ZMC1~P
zL~JD*MUY)&0}_-NZaQB%rV>4Yx~8VPcGY5^%z<k$z5^3qKot`ToUNbQwD8zRva}SV
z#JI1KAp|K8Nyuo8`#X1vOr-JIuA!^%*j8!;eW?VCm*>9T(?k~5Xd=pwuX27gu5_xU
zn4yEV^-zZ3Z?X|@RD8;l{5Y=zm)oDJHnDF}6t4ZERLEu+KP4cWKfU=hB>IQw%Z2dK
zGbt;vuA|=GnMR6z^eLM<BbE_21$WP9BENx<Y1#Jp98E>?*W^cD1<u)!1c8}3r8ef7
zIkjR(*YstlAK;So1N-1n^hLJN3tJDMXTwTG>6>3-QNJ~=BsL&3zBf*wi`E2+-shYB
zE)h_y<6vHlQ6utX#m&E8)Gg6@`q~#Eq=NI(@K*?sv~Lwcfzm>LEi68~1z1?`bdvGR
z4sQoSFaB}&-xHsX7_l!Gbn?@e?nr&61esU&k3wI4bCf*^2o#445T=+K@mLc(CQ~6V
z${b)=i+b?e<hx0mqK3DEPNXBv;5?fUC6nreb+&GY)|eD3zN!E3g^r%u&J<H`InLYp
zZ>DykVkxSS)rZ#v;a6Nq%69n~5YyAW8#j=<x+d&V!p5>#cg5FUePCHM@$|z8DM@pP
z+=8Qdwe9fyKR4TbeFL2dH){=ZUkXMG-IkF0@9rrl&#i>+y^{I&gf6+2Q~n3+?-P=X
z+x$3$GYCAZK3Yu&l>Dj3nj%q#N7F3aGlXg@R2vpcQyo37Hy|5y6e9-v>l8?~TrM6B
zIKJ`l+F<=(yFLTetD~QTt9}_xy4kCwseflAF|bvkz1A%LZ0fnr%8ml(3u(D#Fgpj+
zar@t9%x(U*N|w3OAePov0lfUZb9ujn$~^3XJnvNa!`J~Qd|c*zZ}^qd^Uu7<lO+iL
z=A#+=JR$xIsf_!iua;6@o7S($ogBZ+N!N+|{?48g_BOx6cS5lEQ9y&^Q%ciX{DUh+
zahehey*Ia?&MlWVKP;W^jetQecLb``Z}7Rav&w(4qO5z^5KL8BaE$s{b5;N$z|S}>
zmW{q``<H?*9HTtnc`=7yWT0W0d!AL>?*E>E|0%k{Ke-oGC>_MwPXNtVc8h*D2FTCv
z4rTfX-Habi+qgJhstG?cafA2H31%{O+jnuK-wfQ(K|Y*nc>#4+TZ!e|ZPM?3k1N|a
z-ZL%)mJ~D=T5o$>4+136+lRImE0NZL@_9}$Sl;@Wxy|0|#HvR}x=z+oq)u2mP>0?H
z1a2hu2}w&Ae=B)Cr+Cn%I&AV+QvHb&0ncfE9<p=ax}?T@Xz!f@&iF+*gLWfaP_c~h
z1Q_<@iwDz17B5YG#~EG|1I3Wh*&u&Vnt55rWEDTCHpTd(>|w>MeMK?mN_6u2J@H@x
ziSsu)l105OX>la@oqD%~J%NstQWk~mX)$KR-G#PNo&{3a;*k#EY=&zyzS7pH;=!Ul
zTJxF5LRjIv@KhBpJtV1xs*l!IcH-;xojw-y!-cIUeG|!y2OC$@r{@%@Ng_NLiQEnz
zGBEmR61!<av|2T1>kld7<rSyOliqraBax~_f2_$A$f(G0{{p7w(Vl&{p4r<O{SYvr
z;bflL>ukrB+>y4}9o#J%{swKCHA!M0YFd|QB=+G^C6!*wNWWYxGn#BpH`@zpSp85%
zq6cA<iH-$Rcy5XGNWU%k?&QBGVkJ)h$kXxLK+3Ri6ku7ahu~DjgEzjb*fg$I7ZbJ~
zAhf3HRtYD`Ah=sXKTFEisyv*X>Bw_Euy<M5(b$mbU-k@X{0RbbYH{)hl^h>DizG@0
z*G~nT`;|N@Ql;v@a1v4W9UFwt!0*@n(WlWOw%gAHekvNvFNrhAx~c<862+{9@m!_m
zvvsc+H8(47^!zYH@bnMspibJ=yGg4kjfN#YmogTxyPik~vlI5j*C;WO>aBS-k)Y8d
zU`wjKeP$-~Fv~jR{r={}hhs!ss*#@2m8qU#E?yJKCgKq-Mok4_HJ4pmwnskJPFc<a
zs8P2*)IyYXWk&@R1~(+)-mw(#n~~zmo8rox{X=+65et*NKNm%;yrd@8&?bhxB}+!(
zOB3NN5wl~d$$f17@u?%rY28<59izJysm+mp%xLeDyQ#n+we^-dN~rOvMnc1uwZBZx
zL(_J8OvDI>)aganE-ffevlSXT=!;J;J*dg|%?cM)_vbdjl8dcV5alG64lz2vWEz(M
zFM!3@r9ibkNLi$jF_7(*{=UMbUp~{?m?$ygTX+b;r+<Z)U#O8?m;XsMh{N6N{JPBx
zr>H_bcQfvuJV={h+`Zd}8Ge;)-&NcIlh|GV2-ES%%bdP}3E`O^@HfMeB~78D`i#Q&
z<e~V|4bC+>wOT#+*kC%u0C`p1G8#zmcdAK-UfrGb8Za(S550G8aV?yxHWr97#tcQu
zj!#~6%`9Ezfr7a@prDXWB!|;CrshfjPZDS%m2TzYO>+oB15X)3WKH^zMy#)Ns6zZ2
zx|0^RzjHF!wHH!vcGcJEjk^38bG|WeyYNR!n(uV3=E=vb`2IlMiWI~5z1xYdLN=~z
zX1O1p1^!wyRag3_(ORxNRya(8cPCP5+Z;4pUtu1)F_lbx)*#DpkQz0rxDkYce{Eps
zuo6_Q-L&gBq`|Bz7VJ&@Ku2N{aIaTAC@_3Xaep1vD!zC8^YIZ@9EGaL3AHbgYAP+4
zKY6lHCxC>9-GlAh8vrglY+%Y$jd<=T=lqcYmFf-bv!TM#md`$_@m(J@qVBMB^D9@E
zH6ypK$K4ZZTV)TBaoinc>`f55xG>8rNT?E#pw|PEoqJy8f;$({o#ZxrV_0ao>hXB=
z*#m4VYV(?W@IhXK?>_}eC;fdhw<$UPIu$KgH~ToEZ%}%Cyd^z7NcxN%%CRnxBsWuz
zwdVovNqDW#Tq1K1ON@pUK6d{z7)0D1fO+v$J=`Y~*(sf_#)c4DV1Nn}%8mknFY7~X
z!7l$;odPA#N1E3i{}XU&26C!0Fo$J&=MCWX;-#MFqK%Bnd{TEU#3DAyT+Vy0G5xf8
z+<ewWIe%K>hd0<?ZQxU|cf}1+lDl8|l6LT}9XnTor@_;A$-6|huG>I3H&BWo6n}@0
z{O<TJ9+nwTg!9$Z5{{#26(tBuTdGcEg;lG{N=iYfChY7eRh>OBt=m4jKQN{}`Zak>
z*s|DO)j-t<Y^l%bu7UqB>f4Vj@9d~a!6wHfvF_~Jiryk|ItbQ+-`f3OE)4>9P|k7k
zM)gV-ySgsb*~V+KOubCgZ5^m?6_wt_&B>7v?Dti7UeOv$!--#)%L!TncQ+P=?<=p#
z7AyC~coa)vf8@Le9t6$BPEg(-J(!CAJuIpU;izM|HcZW^%cIpks|G0dDnmzuEsi~u
zZs=Nl8s@gz`+1D_QDgE@Q(-uJ`AWu-Utw#c?BwM1s~r2(LSLRiT|748hGjQXo&4E4
zWDjc#y+gAPs2GCpWxp;^0ihwJ4Lng5QckgRNILqYx($#HJiOJryQW=ZY+9=8^*-s(
zd78lDUuBwV&j`KR8}%DxyxF^5%3Tu~;ZEgv#$IJhvJFdcc>Un(Olt6tROyrZtuY4r
zK8Jb!bF5-ytMKj`J=vBSe1rg!eR)6fO0DB)<-<rrcHNa_cK3%-#jdIXEG9SFTi7}&
zgh6*Z^MX)*?-^1u`$F%JlpJc6ZH)UdmrJ6Lhmu*LKWixF!f|zA(SU-+C0IwfOZymK
z!FxLUIzT#Z57x5D$2s=0&>4$gCBGv|dqVY-MV`JeX;L!p+>?60N`7<dy_yk$Cl9fa
zgTjU&UZnwqBvW6PLy-&Pn6s-Dl$M*d&3C`#&g5^p(&A?=aa4s8ZL=?#w#TN8o1c$`
ztf~+TpT~+9p2B&tFLXeF$+{PRLo1uMrbwI7a$tJan0obc%*3J0G^NqIoe)OVK6Nxf
zaFR&q8m=yDSyMC4H{$vag^CY&4~LQH6KU;6S{*<XghWVjncnhv`D0jnF@80-)X(+s
zzSc?g;=ZdmO}lyx;v3f-udL)}eQ){n>M-FPIndb4Ari0aY=QqJOVjqshKG}RhX1Ip
z+uzYYW6%emRyN3+Eyb&6O2U%njN|eOU2432!eR)eP8=SPvWo?-keWB<lW-*G!LPfk
zBe!>U$yleWJDA&>a}#ykuPZ(M67-MAnihVQHSu;G3brVBQinQcB*7VuVva4sV0BOh
zaIqOvTQRezt-N3j7Va>p#jjNnf{Zt|^(%VH8bO;c^8np4i@0?ANtcQ~jZ$HFQOvuA
z>fdcUTDq?fSMqs@8!nRo#f}7uqm|D70`WQ{&2Mk*pWJwyvt=mJ;V)X#MM8w};Eo=+
zM9FyQtH%7})?!~LF)=>$2mBEQRQCXGUX%?9KzVfnwLHh*{zh2+=t@FsyKcBhq$bC{
zc8i&w>Fp0{CgW8YcnytPR5)R5l>*=p?51eKF2osfrB6aZZc`&SNAa^t&X}+NnQ$G#
zvzLK|mZ)!HIebDb^_Qmg2S%~?0jIZ%J$n>YBFeM{&dYW4pu&RTQlZB(^a!O;jKA|_
z7yxvao@e!?&EFk-1=CZa<V;B-&i7k0^n*kLeX<Hkxs^lxLA}FYAi!zDEYvw^d};L5
zMg(vns0+9ar$sj|w@TNcBDVFy3opHb4Jb60A?)W4Uip_5jyhHY#G1V&K5Q;Bg04^o
zq)72^$Bk*_@<OAtn9u~a`;|tNluhQNMkYOXKxjzuYO*#`|E^xZ!0x^A=&Kk2Mt|(b
zL)V_kJ$MuG^FMX&Io_B2hj0-Vo%B~YroQ<8(&+8I=HnojJ*{PCko4%B8}&JGkrNp@
zYdJ!hGp?%K%2|}IC@VBjUw=AHqIJ-5z|vNzHgZRD?uiGqy)HhU1%%;ws39$9B7TV)
z%@O8X&P<(Tz=L1i4{m+9*AypKTQ^K@HQg2m-E}v%wRSwLZ`^+hi68w)>Q<+J8+R{Y
ztcmwDP{;=M01YSH)wlbx7#&R5!Tjv=*OM1f``e<S{>QIMx#Nw5Hp6M>(dcEG|070U
zLE?th!`@MiT8-T)TT5!-Qi$wTTG&CjgUYcekQJiZ&+dyD<J*x}`QSo=TSP5~_ON%x
zRDE&IFqQ0I%jH-!g1tt?Jfwi{z(iD_1ftlLi%syo3{pbbR+Ry4D=fTo4w#?q^S*GR
z_3eVLOLqX-U%I=5f$?ge0t`HzwrV3GIqy&ghmGy)2)QD1MoE%XoQDm#I8CwGs<$Vv
z_{N88t+Sbl5znJGBF!-<+}tLP#KmXelH8}VB_zuIXPDcR=6*#LBj5&`RkH_GdfZoc
z#qdWRJrAd?m^e=LF8p!{4MWhqHBCR!z=wyZ9|ro^y)m!@g1K|6#Z|Vv&OuW%+10>^
zfJQ&Bw}s7EU@A#}kGVrMI<=V{pTfjVNc*CnR#b(_I^QesgqNc(F18MEK&&*GniX<6
z*{Qb_chL~6=~cRKf5!dekUL9AOa<yP*`uw9ANxD{?GGLJbaHge`?LO_oTL)SwuI}1
zl;rxglE%O<&+2MJ-UF#M^~e6t;9(hcey&uEs{2Ny3tv?Oa-?uYwaCr=k>!r+$*$KX
z2v!o|VfjtLEtZSidKZubgtaHB6pEUpmO5a|T|3%p0?WZe;A`5Nn+|IVSu_z3nW?4?
zR%m5zQrL@UFUK?Zx}#?imLW65N$b+dfTFvj2;z5QmPh(Q?1KTTzvdg9j113Dpp&$+
zNhBWB{nqBtJ1XK9mwb~BRXj!bO}_mrA1kCXJ%%NWw_BfB(~fMWbi4deNo&;lYs`Gq
zl&w63sWgl}wiPvH^f1YbV=gP-@q9-C%7t_Fr0Vrf7yutgUYK*~?rh1`?uwyn%EJ-D
zPkgS5!=m$wmzI3LW6%le{+-5B`uzkuHxkEW80kuyDQgkC98%=~wMCKmd*Myb{Bdtt
zdU<!oEqz}t*rUcj3>`=Cn?*oL5*%zPpW<;f)(?HlIioW5*WRtK^wW>xZw@&rytA@r
zdpe!jt=<rj|6@0Cka`*yZ>(y(|7F>PMd@8haMwiVXgZAUULt}mY^_8UhjynV#A!=a
zPP>jYO%~Wz|6buhaSE=s_RAKBkmie-wluhDb&ut~;4!y^Da}(LYWOd@B!`mcd#?}u
zz4KG<<F&5&{bDNg-xDo94*#C`h4?0E=2+5gi}59hyM5{36H%@Q?o|@cIemf080{Z!
z!<@}8Gjk5@c__jk%qV$w*Tu-0XX55bj^4F3Qx3#oD5h7*f2Miht1f#~tog(0;=IP2
zH(x(?rZ0SqGE3YDobO*$FBx{lW2||KJms(){s!}Ov>eZBMM}aS<RF2X9N7;E+4-{e
zsF#UcJ>>f})UTzl9=-w&RKNfN_e4bv;#*M)bW00p4I^Qr|8vEwxr_gv7z}C1Bl<ut
zX+oIFE(}s<F~VZB7CA$hpl-RoZi$1FF)bkOrc-VL(HAd=pHn0fEeG_5LIr=Nu~3P1
z^-))bjveTEZXgLs7EcA#Rhg9+!o4xFS6%1#%`)Fc5i)RHU3TGOPT*(}zUPraLWO=S
z@Ak2%8gjj2FQ_x=d&_>vX`m%~!^)E@VQc#jE%}8Zhq2N<2;5w5%@g5!^He!8eRv?P
z_TAn7plfgZlO^`E`!FC~w@1SiT^5@HO0pR(bYFMCV30w^uBcN$SXtHz>4-<mm8fcK
z{VR9v>(t!c_E+(_Cf1scH=rWmIN!+gI+8ELy=vedcJ4GO6Yp4y)R@PEPr%BZ-e=pk
z<1;wKv6232oE<}9%=L+zX_dv+87;S86n0&9t%rB~{0NPjg=++=ypT0BfICzLgEB_7
zT9ibgbqLoH^tBEgl#Sn~mOHjBOWy!?s$@6q6e?NK_tj!`Gv(2i$J^(QNqf%zQSzde
ze5P)%6AWCXB&9?zsi55*YhvIsx3|^T=!bUBq(%F=PvtGu20rCs6Acp$XuicM!b?fI
zUt2qM;&W;sX7$q_)Iw%|NE4tT5bs@xcZH+>JrO$O{?)U%9epJ<^yV-jYh8Py^WPJN
zqW>J9?07B-xW;O#@#8wn$B0|X&2<e2)~s}Y9MJ}rq=%mOA3)P<fNNdM=;TuFeg#N#
z4RGOtSfyf((INU<wC*OE(L(X_<wf<9y-<?aEbxN0%TVBe05*84A8TtBR>^lh$i|SC
zbo{31%#I+C^-5Ue`lXoD)4-h<hOAY#-e`+|Z-5TA#kP0vmHe>Y(d<Gdx+l{cJGV<i
zR<syvnkQv12kc9hR=x5FeC_=-==yh?v!%yzMI}YgiQ4h+WjB@Bid{+UOHX{oJ7K2>
zhZ-zXTYwz5;oTz=c@EBNh{ndn<dl1JTTKdWXvQc@V^sl%AJ|u@DS|C{S|GW$S#z*n
zSrK4PaBvvaP_J+usYScwsOk80x1f}0Ny*-d8*In@8s1VnsOc6h<kns$i9~87DmDK7
zgdH(kPK%=Wv*S*O5?NL~j3p))mgq~&AL7hzN?uJ-b~uaIiyrL+%L$iVoxdxhn{L~X
zv5FuXldHSt|EYOvJUYXZMM(=DD@H1A+YT{@vHqyUKECO>7~3`qRZ&;^WG$`J=<uL^
zrs|$eTSr|fg6(EYBf{~<Xj|845SHdRncklbu^me~GHe@cI6ccgWIkD1-RG8T(4Avb
zOO_{Eu$o3%E_fLRHdC=KzSN@=lt{4dASn^2`Ys#w(#=fpQl9{46lM`|OBjp2!mFqG
ztnOP5H;DDlE1$Ir10{H7rUg{clMC8L1b5<|X<6P?+C+~FKQ!zs`d(IHd@M!?Mfyr*
zVy0u7E9P7{1aJ73%e4|?Xt}Vp<<>AdNzH0Rbr$mjWnDTkc#8w@z~H02?l1{%P$85|
zbJ9E!6=D^4m4@-;l$^QuczBbmi6YH;gL*n!F$`_Y9iktLaL3xsZX3WWeVs)Lqa64S
zd^RCP1P?>+6n_LJ?pHkepiXLe&_NZB^*zKhf=Uh0?Fc2AR+c!$(W}z@y<wkt+Xp0!
zX_&B!w*63o%>5k4#@NmOcsrFp7qT-Y#^Z~9wfy=b;_Uj^&RwxOn+qlGpEnx&TU?Vh
zOOTR{9Yel#C14LxRqfMK2I#dBmqbn?m%oq5_nvLz9f^(9tEl}vDm9*CLspJ(8lT*h
zyTBa_YJsY!9!D3VP+p+WInu`Q#f^|MqVJxaa{}hpIH>qpP^ee)15L*8#=j@*6{<I=
zOXGW^dYe;6b1h9Ziji;dl^=Mozx(yS$a;$?rN~sOnkMS)Ar~66*aop8T!-sx>PeBX
z-Oq(x&y6eSvzOu}v~?ldCVz~4LF?#Q9zQMl?0&3kG0@*h<v%vm5XT~JC|OTVgw|I$
zW5b=J2$RLnD+=PVA=}qB$G?z5GQKl7c)1dU=>7hhkoFZ<;@(;=@wnVUks=13N$@Ec
z$VX)-`VvA(B~k{-07I7cWy%uv9&GtS>sRK#CwBZTjhWGjR^LwUR{eV-i*HiW-!fPo
z8T#*uhJR07dcMhezG!iJCcH#R#r9QowxYq6z3j--V{QSDbefwom+DJag%THsw;e?p
zMymPi)<+)!o@GW)Dwm9H&s@AIYh&M~EX$J*2#GZ*a%~B9EfiUrZ~HRu*gc`O1gUX>
zj3~0|A>y6JIpArOHJLr6x;R<48@5+h<cUbN>mZ?2^By7a=@Lp8v&|}PV?rUFg?0p6
z7qk=4030^rmJng7waD)$n*KfEy{hH@pLw%huC(fxtE0#>O#M3_s#WsMKEY<wgv0Hx
zRXM`)MCeXl(;;23V_K$jx?^^-N`r22ZyF<36zO3lIfXomT${TRa&>Mut#CJf-R-ir
z>k3B#+ez0A>7<t~`S6zGFq;c`mjZ&_ie0Keq|S@1z8p2<wnT+G*b<N)hb06N%Q2b5
z{*1m%d5=#oc@5YS--9rSD|xa;8Pd}}T66ok3#WI5NZ5vMJeaBKrE6nI3EQeJ6v_7S
zz2IP`rO$-gnTf+IGq9@78%6*ODHVxNedzzDjm=IAZ5Vw^Lb8kBzY=!xT-!DB9t=VL
z?4KWAi>M+WpD~Nhb(tf08DPk6QpdpU<LgmUAy%=FZ(d%@2PG7Wq>AOY_+Yqrw3L)F
zOxTGRVze?;vpdn^w#GOq)3p{MQd@syHI{)~cB;8^&^W%$MqWCPn4o{M8w*adnaXL+
zgDoa#-L1oH^r}Fy4yD9EL@QI;u)m<H&;d*zGr)rFNmZ4L==?*5-U=#dwZXPNa{0vY
zk%T}$<+&Q?T4&kf)oB3OI-)eB#&sxOog8%jmi0vXbM<)9Of`y3M;8a#XWD0)(i6s_
z-zz}0U)yO4el8yJV_}LO=Q6l&88_P?GB}I|IcREIvG0T;x5^#V>n+T`A?o;oH-D=S
z1$4)HNC7E8eXcZ%VrZ<q`+BH8`$A3?E~ezEyB8Pdyf|ep)2UB)sH<e^fb2NEG$VB|
zZ<Vw+kx*&T51L$(-!I9nsD{@0%1zy7Jc_HY5FJXA=&R-4z0`q_ZvXpfBW3fc6&N)t
z_`{EEQ)XLiXYVx8gi}+$7`&3sRMw5vt<nMAoo|@khJ+vFRc;`DX!^J8y$^<pKBgLR
z_w*6+P}c5-khbQzis<PV;+>uz6O7*Wb|Ml&doC@t4^DI|f)*s`tWNmpw-SBs;A3<>
zOoxNX<Z=K=-9tRl0x@A<0ZqKy&jY+~BC~0M?qx*2+F~kHM*){BCmI&=1=zmpJm!Qh
ze9dr8M5VfDTFmuM*N$%V379aQ*8?E}Y^az0o(?v_<C&kPMc&O?Dzt5Ed?VZ0Q5b--
z0X=0)xaRKjKtJB<dq>T#BjZ;-WWH~f;d}MnXCE{=g<<$(Qjg$8cm~QL=2hV=44LF%
z*WEc-pQ9#vE?WW<Qi2V=C$FuVeJHBX*_IHX+WsX};t`F+Wq+McZQM(2?#o0%TL4_~
zD7r9HRCQsBl9!Zh)Zd?t&4jH_*8~#moT)$r5#_RMd$8b>eD)T^VL(GbIH$hi&Q!>G
z-+0T90V=<?Fv<S6dgsPrc6PL=i=Yxb`g-HObxNWmWX7bfF(1`?D(1Yb;G~{*g*&J}
zukhP*jJMnNbN^NPj&vt-==0KS+i1Yvv3EO2H;CJ0oE$3ga71$#WMayp=dv6EP~dWp
zr@mwAx6$G?fdLeWT>8aD-t8^NhDi5eJ@rE5lH0Hn`>F)t;*v^OSd+^7g~&uaqOT62
zIAD4Dlr^Z=L!3@srG@mn6fEK~D5l{Qk@og;N-0jR^ci*15|rU=+Sv3e9(G4N)Z8HL
zRWVb=W}tpqIDx9C`h6zBeI$JqZCoO0=vlE?F&|)3_&!h9!+g>;(^YfT87NDZluYt|
z$dNXwlY}I0jMaziuZ*fPt<t4UM*)#X38k+Sd`asaSD1#j7b<2iMjIE-4vcsA(!lHy
z_r=*g)^wC*P8sZphppzuN@r)Cp~bhX65Li_n$*t{mzj*OYnG%cHzD8571*@<{C8}V
z1s(DFYA3j?DNbX!T%x)aRqJQa&?KOG9z*{XsgPNjDMkEcY^n%8+|V}uX;^dh675*J
zL2jU=4PT>W^;@I`z04=gU+3AET0qndIFfJ`$Df<#xZ3M)CT1rdx0Q4Pcm<1BuW%SW
zHK^&RFr4G$&}U-*hiRRLLf+P>m(QdyP=kL%#jr;_Jh@^l7mEAh0TrrmXYGA)cg)ZN
zv`S<Nvw-f4P0K-vR_8#O>NQXw7UTCp$&<!=CJ(obgE)czo-o_y9jTX_HDf&8yUR34
zq_A!7*5`q)yxO-xuy!z8q<W1AfLj!Jnofsfe&e}^Z3PVrJza0vm@`<I(KTrofVI#_
zzVtTR-Z4a5-^e=j*s0X#JC3TlxJtXq%V~zV7IdsIH=oY**15`)NgYR#pXN$p>1H{E
zayZ65{#6lI$3P1<-bx%V6jbF$avHoJ0F_&V%6A>yQQ$GCK@*`KLQ)MF1BH*h5TQAH
zW-Dl|DU3iem)nO-DCU<!SPP9u-3JD_69bJUueShV?;U@rX#fC$5;p}Exod@iwn-{o
z>RYQ9zZ?pg?RIc@#^WETNfFW3_2{PCD|qDm6h|UcNcz0a@wHY`j=4ikgsvPMPwVUJ
zq4(QDwQM;Dnq-%rFJmX=4c5mjQ*<(JPG#XLv-Kp>rzjTRwz%r$M8^9&bF3zXop@Z#
zz!k`|;(4KF(5J0DjJ0s-c&zX{o6Zicrs2r>PH0oNTYXM~`;JU4*Ub1-@j;ZH&xQFj
z*Bhz(z-pN5HNvf?B>$H8S?_NxV}3ihKC4!^{6e%4v4+{t%nNmz8qYe|CJntHLimy%
z?!KCX1*4*ij&5F~0k}h`@=27ch*8qLm2}pN(DQ-^zJEJ!+oDk_^K{?rBNy|1-wr0q
z!EI=1k5}oClnmACM?du)ZzwBJ=Znt1sSZK6oGxPsbzbe7FFRROC8}QewLyWCLluH!
z{ixFQ`c!3Rzu7DEzZvvOV78?iAy?M=obcl5hXR=-u<GWiM3REd4fa1#smZcs7~z#q
zt22CEZ%p(pU-6RHKW_Nv9L^4u5q6=gg(#%A&D4a*Z4Ewo&)WTUG>AXJ+EW1@hwR7<
zoHka;dB7-dbQ0M#VuL68=*u5L)J82yg18*EXjOb#zsjk2;q%-$v$S&Wz5L2Vo)N`J
zo9t;V?7Y@u)Jq-%Yx|I7@C9UKDW;76M&wAivIx-$j$NW>GU4{Y$(EWIyk+UVe8&r5
zuUJ<wZ+i68&CZ(dLbWQlJ1UAA`_**UmDX_|eVM8QzMrbm^m4!Te@~n`R&CL{jSLGd
zT*F_A(JeP<S#1giIhZkbBPh_pVEmqbUrbzlF}x{<9HkioKM3*VKllv?)CQms6D+6p
zuEa=ZrL~<sl~1#u{>p@p)&^*5U(ADRwcJyame~iXTf!_HKhOi0jP(=ByO}j8R)<t&
zldAEx%JTHRG4;Z1u@HwgPQ#u_lgfe@J8BfvhKM|RrJQ0~g_{;gQ-3k2`y2>B*xw03
zglh;BeN#!zJLj7TvgN74y{ZM^2X~+hZ^h+%yx_2%RV4#ixWVOvYXt|XHIbU*cP3Fu
zh<j~0Esa`OLY#FWs3ugk8R>tYxjxfASfAB%Z2F>T)hVP}Df|3H=5gdyJLQ#km{om)
zmQP5i$Xf5<9yHXk@(0Ks-Q2JI`eTjuS^>wUy}t!2s_|9FAd@c+opNwjvqrsngH4Cl
z-x=P}_0aJzLPLheGtdPX&8jAO!Yoi@d+NP+7D&(U%{#Un)=Fga)v2}$8;0f6*9#%_
z`x_h&@E1a;q|tonkUUkS7HSo`a${btNMv;ym-`FvA@ca}edL{)rqeB0Z`*Apf5i8(
z;N~NJ#eL+|J!_86l03P82wkCeN%tIC#a8Y(7@qFC=Q~!zrKfvq@||k@);R(iqw5p6
zCWF+9Rk18ID>;8@cCti<!AbCAVg^K122{M<7oe07)p<xE;W9>z>0*s4hsye$4AlG<
zS_gf<sW>fp>+b2ICw@J>1*3UC28^G+nRId&d$~r?ST*_chgCAWJMJtz;F(9$h_>|G
zhAW4;PP|89kmS!*Z(L==JYu4M7!xQ|fj%Am_0oS&c!3MxVOdS1@s=Ry+@?{Byv;{y
z^pTO)0<?g>%ck-*?UvCK;>Qt%a39Rxq))+*=qun<Q0snUUrpeWs>_h{7oTi`C@eu9
zSAa|47ciLjvra+ie+S6@ceK#|$5sCm!2kdL)f0WjMw?ZrN8w#q3>^Di*-uRabt9x|
zPW#uw_lU+Ks;bVcBmG=QY+5G4v`p>eyS~4m!u16|mFud_8r4O;UI03EVxMO2W0HIw
z7jH}^m(qlAFekx+2k}8lt0S+`c6OlgvRgYnJm)(?e$EdAh-MfipQqPv{x46NR$pk|
z=%c&^gw?mw-*5c(KRH>a<JaUbUMojHk2tkk0(I?Q3ED0x-VuL7>2L0N$ud{(s(J+_
zIzFbUcZiiucs<6cl`^cFz9wf~G9qT*1|}2vbiYvBmgX9Ul3JZR68U01Rwz&~a`14(
z+QV*sCu1h2-inmghFIW=n|CAIe|cHoXFZ9G6-Q<`JrMWGc$+cl+c0d3<i{)uKK|=w
zyn>mH>4X17CH(pSWRU(hhoarepAJ+uAvZkoHf#QT^xq`}{Wm}S|6X%JDSu}40`_RL
z_TLkM6+hwsJb{UflPc)*K?V2~<WjG9epRH6WR5u8o*y6KFef?#g_FF}&OQp!>eh7E
zTMxI+L;5xM@trobaBnX@Ap?_Jr_bV(Gr8IVM=*AHq;4HH!lG4wJlZIwImu!W8~@c3
zO&OT3@}z2jlOV+6{DfE!CkE~c^Spe1bIl7;y(gN`D#^;rht<Ot%lEZk+`BuN!1?-u
zV`aDN5g2Ch_iAKY3;^^k>cNg=<oq9+Fxy)zjvhGA@O}ki-ou(~jhiv$`A2q*=blXs
zdGFf7!Vg2T%xVdVwRF?n>q%amQ^)V0rkMQ6O5H|kNL(1`)XflJ{edX-c=w?AHPM=t
z2G*Nxyh_zT#QR^1x7Om>pDfh_96TNGx7j(2OmGR4AaXi2*rnDUZ&!I#r;po+451-3
zRUBtNgtqa6c}+dk@{4=KJM;IXRfCEY^Z3-@I<W5N;VF%-{CUR=fL#pQ`boQK*=mrN
zPH`WlsUd3t3t8)vFPfZou^=G46)jf%-+%V{a^i%jtL(9p(vf59<lm=w{0nw6>5a}o
zG^EZ|7Gg37!VZp)&(>AcuMGta_Q%s`qi<al><`Fga04REX*%WGDI;6|J#Q7QaOJGc
zTltS}{aE1J%J~~__YJvq<{2fqS$B4hsehr~<3d8`Fx&e-_C6ACQ?NjHov(%d{*QVX
z{iGQ?r}D?c6)pt^DTq;cW?~}^V%siI)4EM|sS+&Dk~YK3Q~IY=H*7=2NzE0<n&t^A
zkc37q)=Ah_bI8F8GI_te30<Y*8_V5KT?B;Z<}IhCsMYlGa4L`74NV^%+bE~iQz`{2
z{L)$=Xj^4`bx8vjSo`7ij=RA}y(iObrJ#fVL)@E&C7u8OzkR<mQ!_0(Wol{aH!ZD9
z%?-`<Te(z<W{S9><$?-=DT3zKG)<Y}GLE@{%Y=yvg1I09E|ZB%iiQX%h~~cFo?DuA
z_?=tlI_J*+{BN9d)9VuXTo;#b-tYJ8`FuRkOd2P*xgovfSoHGfzDRcUHS0s=_kyeW
z`(_AW4Dalj$DYLW>f>T-uMt`5AY{sq5xB)%9FgdH{3B@i3Uqhwq31utL$Lw_G3iQL
z6-f&Rn+zo^Ga2-Fz8@I2Y$|2|%`TG)c6^?eKE#3l2#J>+m(-wI&mL@gdIP{Ye%&&F
zRRh|vDBmWg!ueIblXD}L^5Wi7p@C`?p>8w9C;Z7_u^6|qCfrP-G@3NGz#kz`I4a-a
zZdR+JVhwEX$=Da^i3ui{mpCcOf(iyHQQQ<EEQnGimHD{7zSd;as`!?ztX8Y8hA%vO
zU755%hjK{&Q>#xtsYcL*k>ZF2IHKm`iHTz!6>-fscAhiktwRF)BgX#~tJ&Hy{Px{>
z@T~c0M8%byV~*D_Mt4I=IohbE7KCK%mht)2Kb@mK+&PK!yZ<(A)7AOFv_<jjubZ$n
z$9C_IdNOw^bE?dVN5ie}ocLNS@4<Q#wq0k|*@n8?UuJ{|;*w{=cU3|%0N;uOaj0bp
z$skE(S@~_i^RfRRDeFwSGNncZLp?q1&n7lo=*>*=a&kZ_o@&>Xh~pqs269WfHBQ_@
z0bnn2@e8<^u0SO|E9A4IWgl<(#$PdYTiehU{)CZ%FZ%iL$YN&Hixn+J?5T)XCsp*;
zMX8S#{e0i>-{d?yY%yM{sp=BydOPuIjf}T$zxp&GAK{a}Z*$ITZO*`-RspjA5ETW;
z92Z&{3G~OhXvQ6%79Ten%(?^B;_}M0G-m+olmIH;6$+u#n;%OGuJeI78fl9amm82O
zJM0hD8=8*7!l{*+!U!KK-rbn}ftYrAmRXg$6cLvws!-Wgo7Uqzt|X^-xoE)lk8>Bi
z&qci0`@63oeJmHji1*IEoqe(HhEwJkl@odD^5y$eAF?#hn@G7H<F2E{?5Eiij`LaX
zxMudu&E?pHk-?P(940a$A2z0b*`2l7>q!9_1N6alwTlAc5ZmnPC_RCiW?=6J%Trv8
zm45?Ssy<5C*3Jd(MUK;EoWq$vE-(~8ZU>|w52|520eXc<iYu(hjFi*-`Q}&d?7P8+
z@;g_OSeO1HXleS!r&g^1s4@a_;jz1(&(_NC+|B9kPLC-`uQ&OIZM!`}gz7AeX?GbK
zE3kco!reVE+p>1EvAVm0p*1nUfA3|fdL!7dFP?HVs5CKj(O6-tqr*8tGhtmK{Ab<1
z!R1Dy${)I%i>pZvud0ANX}kFTUw2I|t;USB{^h)X$Kg5UTzB`Lt{ZB)OlzAys13EX
z&;QL(!B8<Y?y;hsa|cM{zY@^8p?VY_Xxy<qt0^mv!^S?mEY+Z=7^+dED2-pXc2D*t
z#ua9fplX_tLRY=5pkO?Wm;N|)aN}wlQq<;_`HpgH09e}%d50V=BR_n(;JzW(&#(Mb
z`Rpsb9=+SXzuX*}v-?Wlj2RvFHR~EU9P3blQtvM$#1+?}&{pZBN<ybT-eKaOihI3`
z=p{BQz^<LT4K#z3a0oCAp93;<H7t##{v>e?Lna<OEd+g~9P^-M**6tZXTH2|0U+2W
z9ei*#>$Dgbs0~XaYHEl6q4PLD!<)Tv0uNisG=2C>vlM7LHm=i9=6b<(Q=`(n<WhR{
zm+8U_V|J`1-)mL(xjgfQiu4ptU9u6lb-bJfh3!RV`sya~Gm<GS%&%7MgHfhO88sSP
zV<LfRD&ItkQZPi4Eig3%sJ$^yITWz$Iax)=1fAToSd0^H6RXPYV;qpB6#s4p)DnD#
zF+4m**xa8a507{<QcFrf&aUUU12=rWkk7RKF<(NKYKI5ORh;2DAGy@02X&o}Q1zgl
zDz80pb>sfbhq}BMx)1&dk-PJ@Sl{Xl#Zvngkj%-F!ys!W)elC*syEqfg=Az)*PE~c
zjNv{+mdT44VI;$6FW19Qalo?Z2b_fWYT8ClL#S9!GPP)&0*ey?va4Vt`Il)K6%)ZT
z5K^1|Rgvst(CQ~qK_`)En%d-CA@H;v-@|TWWHvz8o#>l54xRX{g$EO7lAhiyj2Sv+
z>0#2D{;+U69B+9Thr9Tk@`}U+-mLhu39ZPB2(6l5`<tC}37?tFOwQ}8l9XKBvyKaW
zy_e46?a`wGYM=T$XrJSwlHjb^n;*lv2l`twQm}9XTLe^$9btGHMiH?28~P*lrVx03
z8Lf%eU_D2j*%JS=H7o;+Tw+9P7V9XO8R&=A!*DC3BuSAJ6yrmOeE1?;Gf1tMZ~IZ+
zKKf%S!MD<4=GmZ2Tkj<!j|XR6>caQ0XJ)~UHF7UqsV2{j+?%G|lSX4kD{G&hsyW}c
z=9DrDbl)-@*M|9}k2QF&!+E|R<#iEpPnwL^8zic3gQQWMP3LItJboPEmh$9&AG@cE
zJ!DedI}#6!nVkw~$tTYHrW}?S7e*S*Fc^5CsD{+I`h3hl+%LGcBz&UXZ}%{GvB@yE
zJQOIBt#WBA-rY3&=9hR=E|B#lE*2U%DoULk;{96Qs((^6)2?&4ajbP9Hmk6M{noES
z4mz0^c&`1kXqP^6;mT2?Lcoi6OST|gmyHjNreoHqG_LS2DgSyj>FVExMix5*-0z2Y
zj_<!80$RU&{>dO??8rQrmh}D&JonS0(v>oZm|+&PtpBaN5?GYvs4_-xFlqx&hCu!q
zdW+7ubwv<T6dy4=IXzk%UF6tobxmMp@SeqLY+UTEP(7w|N`Hi;qX0;Y40a0~fME1p
z^vfcJesrKzIjWiqB9U>j%?hrnu!x`0J(IxvY=g^fN?dGI^bD^|2<!vpc2Ux&uZ&!m
z)5{;80PHA!EHW*+r2M&-zWT7;DU)`}6zewfc-9v&GvCNrqdTtoZoWbU9B<qPedQd9
z*sUixWTEtz%{D0ExK3<F&BKSm)9xqFp0z;QR7JYBC5IDO!YDW^8jBW!PS0dG1@uVN
ze}<{7Aeefx$b288hbS?&XVy#wk9Nr5R0;$3%F-_UkXs+yWK_n{(bRtA_;ZVG13&p+
zOb+ohqAz~+-rB+BLyzkJI+1V(ov!xT?x3!8-UXzM^fU3=V9>=Y*R72+!|watcRjxs
zzTyll-g9WwUmhUI9O5H_ufGcGSAsdGhw0DCVJiL@NH1-)fFMy?U-DysMsP`B*Xq<%
zqP9iAu{A2p|0qm^E})V}hVkT6921l&KUR~o@3}zKYUY51&N1H)wJ%P!8EvYEkU_Oo
z`vId$&-Ej`zO2YoR~M}o%}$OA?y`+XHI8?}mw*+gp7+}MN^Dq9O^48#V0>=1^QDE{
zL#0b?PL2KisIR_j;*LvQEwRWIdAVgog35DaX_S_inDq?8kB*X&*GUnaWy|PK!6O7J
z&L}>b>V>Zl)kdtvn=g88Ept!~xP=Ncc^|S3%Xr^RN7+nXL0!_XlS(FTpIFo_m_>X>
z;+huAN^(qn@2~5r@T~Sd>w9l+4ny&iXAi@3yn6y}G&Stb1mo|JZk)V~>_!)SfL?&!
zn@)``hF*mhC(Vk%c?q^KspTAcxuuXEr9UTSRrvB^)O1eqtwCe`e52(+uhiAkTx8k`
zyH^M&VLc<b(OaT<nmw9VO4GZgdWTm}vl71=>ENwL&P@A}65QeH5-_|pk#FcKgMP~r
zu!cZGQ)jyk@~o%f@!x^|J6kW>PffXn6xw7>D&8+}b@$c%&)->9*B!in$@?{U5bQ2*
z+Tn+|kh6G(e&IiZ`Ny+{HF*D`EHu`A{nyAD;R35NOysaF_<VJ=q?v4iERThUt61;n
zUzx3{F#7h4;_mVJWc2K5zO}ovGd!{kqqj#{Rr6$O<F8*=@z*1j(U>*E759322wDv&
zSbDQ}7TPQ>^Lm`s(U)z6UiP^!%0>*qDXsMl>koHl{k|rCQMO6GxglS%`TR+<n5+6N
zsP4%Qf5iQhaGw~L_5BcrZ@Y1Td`-99SJ^qX-vt|3JNosp)`ZzUDK7l`A=(en?}z5p
zal;hZ;_$$9V1J-W77D6iNI&a-KUAv|Rl1;*^(FjymeUib)&2V2hcf?Md(Bs=u)vJl
z_Dx{Ax__YY#rEXTrHwnax6xDU`}f7NJHfFA*^7U7cXU|&mT9mvyL){rF6H~7)|x~^
za7Noo2qU>g_4^@%g-H9^L|Yg>(st$m@%>P-=r>FGLD09at_rrRn0K;4^NEIuHOm7|
zjO~7_Z1$Y{**iatE_Y7(W%8JQ)E@#KNGdoH@bK|#pXZm~bmfAw&;Ro0wgxI#6u<S*
zyVlVj#FLM;i?5=BXHHIZmpZ>U@M=vNd*%C#i|-g!^NFsn`)V<I)_hp~Jc4Y>j}|}+
zLGG{&)Dfkst*DwXV@!3Q>Z^EGVczAQLH%)hR%!H1tGJ{1c}i=8?(i6u-vXCQTMZlG
z+qr{S>nA-XB@7MPrgMJ*ExvG(C_y>cxq&d@X@n!cIi9kbwE>2_*Fo2K4K2D9lK=CX
zYk8b8Q=jeVDYscU;`v9g+EBUGOrr`Lo&ME;;)_MTAWof?`exn9U_Ze;xjjb^jI+uE
z_q^gP9RNsyB#O9bxua@Nz-_W#tsK>ZaKGN-C3)DNmeC50qH4ER(!z-pM}s4vdD3h$
zIetuU&GpHwdp7Dlr`JpmrV=w=`^N7>A$Hvy1-fpb_pNFcIccw~+c6u81yH9xpA1z`
zYk7jJ<?Unju8oXUIzSVKyfA38mwo-q@EJkZ+aVPz^~HSV=nVz8>F!V$eYewsM&bFN
z|IFKifK8G#Ue858m_&}NZs&>%Heo8Wk`d-NMMV&@h*cnl-xIDGAK05FOL{$}k~60M
zFVm{%AWvdtX6x7FzY4&}a530T7KrWFM^>a_0<Yj>gvFt9ThywT3q(@czn9$mI`}8S
zuG=BC7P|jvJ-1;lsN><<$9;McsW7N=7BnX30c^eX9lM1R2%FTi&$@`{r6vi<)tWXh
z%5Rn`IbV;cy<*AHI=ebBXt|Iwv9c+#RDI;q*)G5kSyl7i+eQEnUxsqSS5}_pmZ3!v
z8xxJ-${0#gG#L8LS_)ngM)N=Dl3NF@6xM{^gPR9nIa7F(=*7i!mMND6bmPTKOgH=X
zG6Pv?XD&pBrBUR`=t&^wjt6R^Yc-#oYB7(F?3fn8GfU}JDC+CLFl_$J$R8tNU0%O!
zah`3NBY#PXQYSSn>uR-`NCiUluzhAG9&6;{uof`*{m|x%_?uvc-mWYCtL<nm!Y}dI
z9tRu)RJN-PL|YaD)(k1LQdWw~J4Rv7nZj7TsRM1@Eps|`;54sqlU4Tq$ze>yLzjs@
z!L=Ya>ed;qi>D_DZU2%G2?nfRtC0{T>U9$107o?t&lzQW`SqyWJjS3_k21e<vI#5o
zj4?3a-A&TreSM<Ov)Z{gGgt7!_}QR9N2BkG=zWE$ER%I<NzwP_fQ&OgV!=`>N|hL1
z)#~KBo@l<X`N!sq_1HxCx3l5Hu~s*^vHr-7QCuFDDeG_5)6o?0qtFmgRahZfTCgQy
z*H^n`pYc|WUMY`Mdf7?EkCC=zcLAY(qrPTP1V>jM(t<Hbq}T`wp9Sf<at6p|${KQ<
zrrKPjAv_HU&M=7=8Hd-JX~R=1NcdMg4W@K9x+y_=V0oSRyLiqrq4+OBXY=c~|Ev}b
zn2fbscR=TE(NHJ*FL-$~{_QSbf8~M|+?f6M`TSpw8;P|=sv`#Sws*cCdP9Hlg+K4F
zIU!r5POWcNlbOk78BGm~fRzPVAUwW0VK+wVn=An6U6-{Hs2u4Y^*MPO6kSM~L4y^6
zo;LNL0G)zm2Y+m=@vRz7iCOh%d)xIUMINSrip{|mf_ii8-8?`*C8PEEmnD~9R=j*k
z6F(Ph#F<6eebh@0#?7uU!aQ3?!y&rn?|PWI8TF|!>O(}utLqT$%QAQKfqrkTAFi<-
z(Qx$GSw;HkYbv^Tl-w=|t-a~~mp;X*E_BUjP1?Qz<W4NBYs5X9=u}$kukplsDWRQk
z{R1D@cJ#~2=A>uLcQHE{Y<0M|m*Cn4`ke1$FP}{NCw6pfya3ePjQSfQNj|-cB}DXP
zK+(mjzAU0BJ(<i8>inPl;NgBy5u7F|Hq%wrlzdnb*GH3MBYd-bK_6fQ#9WPu_r95V
z$oE5erLs^}e6->Ce8=}gk8JOLKlJ5!+t*%+oV4r7e&pKDw`K4*kAGrLw4IoVV8M1T
zx)`+RRxUmCnUdG()LmD#mhpv@a?GE4a-1Gp1Itdwab}J?oWVZvnabX{lUpp{tyuaM
z-hF;)>T&)2>Zi!eF@u$>9N&M^kI_w6<yRLh8eQsBnGKNoQu;=;w6ErNUk9TSCi^Vo
zXT(gI70PdLDa0PP%yaM<!EOhxM2_KW!V%57UgE}O!_;cGoUBNt+0xw7P$R5|C>;;J
zT%Q#$8>|EOQHI%&)}}&y5=1ekQ+>jUlh?jgqWeYZkMTJciqVki4ilzIbniFtKNpSM
zh9wlVh66xT>uNf`ZhGy0`}?Ek>*x5W0w?Wv#*N7;$*;z5x1`386;7%b(-ZH#d(SZc
z5+gM~Ns9%wmGKv)ofY7Z)!w}M7Du4xie#v9tUMR=A)YG1m(;<#;^%&|H1H%u&eQ8-
zRE3wNbaP?FJ>DGEg23qPrIPvXRvkQGx>RYha1s}{Flh=v6+~+NF@Ada^2B(NG>EzI
zTGV2Bq5}CGg{-cG&ws9Ld>ds&S151J4YGAMI-b0(T9O(jy=dzcX@BxF4)ZhQCyD>i
z?!@a|)rFk-XU$34joa;I?<L<4)kSUfiD)WOYli6K#kV5;-**dQAaG09dTbUeM&jhs
z=DJ96)aNuW>D}CFbK4BBxX=4+_>KQTl6-bIn;7B8@kbYHfND++d2QuWz38C;{Hvgl
zz>yY+tW>pa;AVTf9w}3duoZ+tg}|-mp4saAZ@Pv}%1;t4&BQNn4d9${;Z6>Jn2(s(
z@4dp|A8x(L%D<y$PFT9LQTyc95F2c*Rn4=KFo<!olX};-G!3>cha=(=XEn)r3ndaF
zGpPTeXzwnjA_QbzMciibJ-#>^68ccKXNXn%7Y92(oBdo&LV*3vPNj7V>tiZ_t95Bt
z!qfi5?WY%B)t1xs4W2f3CX%X+?(p3rrXB7>vi>z9-_W2}=k^%#Dd0uUlzh64+HQy^
zL}JsqF0sI5SB83F-E*Fm1ADp!Mz!H>3ee_0gn@?9iXN-h#IJ|IXD$?w0x>^e;HA?7
zk$^T#2kvc-9ve|Yfce~5E#jq4dUV)9CEed1v%mmZIlIY>w^6-08Q8UD!I<eT#)-AO
z7WspxU;7eGZbyl2vh!>H$sLQ+Fs%N9=yq)K;|F_DneR)3K@BG8Ga=7v7)pp&?U{Eo
zmrLvX|18N#IP%fJZ?IWm?T2@Ic$XGzW6VK-*QiK>-(JAcer^dpoW8%SfNe6fsd}QJ
z6fN6sn?cwFiCOJ>1h5C&5D2&74CC6c1q{`SIApO)lzymxXXiUMwXo#srVtasOAtK#
z<VBbn07E&|PB|Vt$cMcrP$U$2rTJ1*tSn_TF+%d*F8!3B7cW625Qp{gi%MmSHaF1O
zxqK?hKIl&(_4nhV<LAGuJcubwsJWu8CULg2s}Q-L6vglY<2B*~_ki)Y_goMsSd3Vu
z($u$)Vh(ssu`VLfeh~1V!MPvgBNxvwBW4z_E19Y#kC|FxE$s*|TPS<@d=d>=$-pqL
zFUv@yyuhW5)GFww34ca4q@`kF;fv@+(mwqF(I-|m*khT1gXG{vy58>A<}Gf{AVap{
z<%geb@8h%EpCGQ5ZGDW`>eOwUP_AJXo3tHxbiOG1_LncnXR~YFli6HnXl2WC{93R!
z6)qEtu3})p13z9#Yc>(DtH<~WE|*f#Ee>rGKTT|3S$3@YfqZ^sscMvdGS6pymr=@E
z%3vQD25fbe)qL$@%O9-1?=;zzZR_D<r8|%#-w!<&xPNAYx9K~SZ<(U(?}z^Vr`5!!
zjjWoW@mt%zdt=Qvw+2q@g(o7VmG6g|EmFQ|zInJN_g*rjFMqxB-|xvGU#G5)Le+aY
zO$c)yL+?im5A&4U-C|!^>hfZ{k;$c0yO3@~>B3pTyx1SldJ;8C)TPrmBadydUV)s!
zz?g8K+^>6;q|{m2Eq|>c$XJ7>(1o#Fb5}w&U+rs8Ehy@vBq(X?q^5S)+7}d*26qJ*
z+6(|1zT`P?6In+kv%in(ne3%xE*`#E(%l(R#`xhs7KNt9*Tgi&sauF1ld1ey*Wk-R
zl*<?rq7N|L$jcF5oN*o8qN9vxt3M^$zayd=`>PtaC@eUL&2VujJ}ax<z?)9<;`tz-
zIS<~}x%48-SAD#~pu-2ABx(_ih=0O;50Eqic?8L%0AqvEM|kaMhP5p&()=&3$PyP+
z!&O<O{zGUBbJbnZ4j47y!u(Rb`$7`q?y)7bMhj%?`nzbW+2zW92P8T@{-|)y*a;+w
zbWC67Z#GU14mzM`bM!@t=5HFRCJm?OQY@8y)-96}^75GF(|k&#-uv?9rK2l6Om!k0
zmj`K%|NPlnW`^NDt?-t0B$GsXLcpWi)v<*Fg6|2+;xJP#y9$22-VjTCINsNCnqvt0
z5aElP<g`{13uUxCA_oTzyuyN<{w(2iDI5$0eg(!|nG<JotAFr(wB!mzOMl6?Alr}#
zsqjcumD*=%)5_?jN!fr!rAxj#x3cP%AbR|_&q`;O>j%kY4%K#D)^|3Y*6^O2uAA%P
z5~Wp2Pd&ON6di}I8Sd&67veIOtaO=FeOV;=Q?QT{?qfZQh}7>NCWUp>M`~jWL?wz!
z(MU-z-sFM7Ni!wem&vC>;d?GJ0B=Tx)C&!U1Ndo3NQL^$8R4pLdW0-Q$EJHGqM{ju
za7GpeMh61+=9JgEK5^d%B_!x^4`LbC?r#Su@M=R2C^ZZA``;bqLlJ@OJ!!Yx(3kH0
zSBu4priG`^$Thdd2>Cf@-Lpd+qhce=Q!32LZGS7VTzw+978Dk^+-Fsda1R|D+D;Gl
z9}w*|nl@8xwbKVG#1(+6{zVjHEzPvcvIKGX7mU6vfip_9Se3WZku6NI%X+##ti8@=
zao$SMqKDT|{N?)6tZ2j-0TGWQ3L%Eb1Etv~qDWurbs{MQzvK80_4_wY49TT_<<!34
zwd!B6i&hUA9GGIua{7MX?{`O@K|qd<sRQ(ezY6+XXnB2SvKZ<F>W6n>i|3hv@7A(-
zbY4xD-UK`_F1#Y*L_*Xkk9NYRI)1xfsTiA4vCv3pmhmQT-4jHu)#?Go1-lq{cN-GV
zV8j&+4OF+?Y8p*ehy&70L^MS7c5Ui&pwxWbFBH-eJtN(6c0<Zy_<RWi3^W}jT_Qs9
zo-2L@3j}x2d-5@{%aMAHM)}CWn0Gd!-@~usKRl-*fsNa#y~N`=Cf!Kth3g7NUF#9Q
zTj^}vq058MeK1Gd{bS1b?<>K-xK-TwxOzvDz=p7=lD7xe=|4E|tnSsv-sYKk$~{Bf
zX{(SqapRQ=pFV@(Me60`1$wmJ0_IxShb@cedrRZQSsOdA?)n6Ef$wku{>d@ef+vQy
zRGo2{6dk#+6FsGk=X0OvABgKsW`;Z4fC|}=5xNlil9ZY@zepg#$(R>ELO%W9r|SRz
z{rvwr74pCRsQ)|owEvwqJF&`rBBkOJyyo_+ADlA!-}w{&ORxEV!g~%KI>WN~c%Z~T
z5`S=aQ0>}~PEYHbQqI~r@RcMz<);nS37`jyzxtmp*VVkUov*e0uZ2ex+Oil(Gz_X>
zo+yj6eg%q|8VU3Xc|Ij@uF2QOV-!7Vp%)Wb7Qs7C`Jp!dHU2+Yb;oK>eXGv>b~Rx6
z-oFoIkNU6Qb_`{-1c@Lk#NsTqfPB{fmt3RgdT*Ff`j_+r71F9c7$yXtj~twSm1SNb
zt@ghA@eO6yc&eXNZ1ZC^cEUAHJoOH-`bKsRTd74kj`~=m?M;C;tRrskT6kxf4mHm~
z9^!$)`emNdMlBL(tN^SoN@)60usZ^X2fn8_P`Ks6^P+x};ECW4=MGnUv03;WKX^*=
z5=RO4E-a+`q0|-$OgJ491L`Xd{mJEu?}Z-Z=me)VKAM-1D(}Teooy7J@uP9><FrBH
zPhC1qHSTO2gK8P|+0XQ=smT<q050g?TBP(W@J5SO%db|6gAIF0m#j66D`F;m%2bL&
zZ?v}yBiJv~m{ST-L5t#mxP_)zeIaBq2KK}&$eUD2LKzvjDZ_-U<Wq6k4Mqko@5ek_
zpJJ*Q<|DDlNwDd_Ibh$F_0OBVJAnnn!k)rfh}*e%=H(KtuI^pi^{jElJX_Z`&Pre|
z(@}-56l+ejd~BwP%OAOa4(WWtxcx21c979%&&U6&M~M_{Df=%KJes+9|9sC3&l-?)
zBuQf(eRb7Q;}9*KvJfj*?WY#x&RYu4)0RYmOnrJJ@!PU(;PZ6vMQovYTzrIE<sMTn
zpbZ)*k$MEI(b<TpXf|b;wT@Vbkzn&{tNp15iCuA|vEtO!=R|O{`2u0N_v^?kjF8p_
zOJh2Z%nmg828qsR&by@C?|z!;VX>7fIbcrq_M(aiFclyXQXuL``Gu^@SO~O=K^toR
zQ(c`J0CYS06DZL<{Y%JHYzAw!>)`vLs@;^y!%58<?f{(OBa@{ok>1?KZ`oO<{n}W<
z!?BiZfBnN&qby7yOIT;Lu^K!F4e0LcX^U?@pkW%b>l?5tfGlX3Y|0+JwO<eP_AWH-
zu<xs8z2e&WwV=|%GcU_L6$=DD=@Jc@rgby!%~}fE?2DO_0Aexp3saTZVob*V@bl11
z$WWSZHyVH(yT_mDC2B`o^uz!LKIEj&#!b_FgRrG~NPa*AKjyZj@;r8dTpe%jB>({C
z#9rV26KBnoJ=V$zN^%6hv|^moQRX^?GVwvRK=boM@3qsR0`hBEUeGa*3U!VNG@QoI
zNzm+F*J%HWSoQ~ypBDRPFk+-J+>}yB5APOxa!x#_x{B6m&#ab^NVOa5(ygY?5O+XL
zhnxM$9aKY>2|hOn`%${QFrOYF7p!59K-=KPsP&g~v$`Mr-5qi^!{zzVVpoZ?-Q6)u
zs9gPd{6+q`>Ri4#-FL)F{;VvokeNktpFCXr$>IAUR#SCw^4kuP_weJ#@z%w6`KW9p
zv%z`%*LpxK7wL}++f(cA<wf`qtUYt>+<}VT%9yUKFXo@i{3T3lbCBC25(c1(={caZ
zGD2lK%g40ERbX#8K*<6pGb^_1bP5YzsVf&<j;VPutC0;%{j-havs%d|7*_7phxkZf
zF3hXXT|!T`FA6pnV&a%Omldk{SNah|w@4K|N<F9c-<c)(_gnN149!x?_31>f7NW1n
zMjeqrry`Q@*j#_VPM|a@3*{q3X^vSJ-vX1>uXJ@_%?oyME{fu&ILsm=rxahBZDp}0
z1P+@s6nV)MC+`7b8fwcjZ*qJq1|)RRzt{>V^_zg%v9c%(N_*L4lw}B1LU{k<TRHK(
z1~<jdk6^8JY#f}g8Prp*8Eh(aFv|{r{<)=|P+;R;=<h>%L7@MuvGby)opYK3da<^5
zacJp9IpMA*;6?EhtvK6-@Cjkcw->FtdzMlQ?X8|M)~ZhI*q|xU*yUplA`xbGIO}!z
zZdB#$4AM%^hwoVvp1;_Wj!@9QM1$Zcz{N!>Hc5YaZPiy`1r&=1RCS%`<qLpKkx@iI
zTfRe5&-<R<me?U5`tw-zN)mq4zRf!~5=W>c|Ig?@?D2eGrMB4ixLyLcL{FSC-sqOU
zaQxsu{JjIG)Qq4Qvymdoo!@6e!e0WbK{NI*egI=TZ=Ja3?7Z@+@9HaFX6L1aSGi20
z_3?GzLEih~t@(h7lrmxtA<{2txvmaFSq6iKVwR8g3%V<A$tqkMQ?4BL=;L^rhzhRq
zDthu}zMvLj0>Rfq!wI_YEfe0iz;CZx`i~W!^?}5DRyn+Okr4_;(6G}kA)`Lads6T<
zN|5_$cG_fXIx9!PWRvYTpipP^85(Np`|4AE4K9I^(DG@GG8*DPO4BuJXk~I@R*jN_
zkDqm~JF8pI3{{D*IN3)E4N?ZcR<^?XBVYJV<qF=6ih9F-CGh$D__B$BinbKX{Pt@3
zA6pE*Uh7@I1)Gn<VGciIY+7QBVvh$cG*+?ysR5N4`+TgA+KM=Ld(R4)@dKotZ}fJr
zc4Mv1$RK1E3jg+LEyefUs50I4g=nkg$h}@x@-|cs0^+6!lWV-=i{dGPGol)m<pJ^D
zVdzD%(#kr7)gWs*c)o3X{9Yjg`c18zcd)lA@4UABQ+4C>hp6=M4p^0fK?(IS0rqgx
zT>He+kHOLTLsSInOQ*Z0>5n_U_yD`yayPf6<#8v2|BgFeG2YQBtN7ir#&2EOK<*nr
zj>1IF`qWmq<sw_jCn-p^cTF@seU32%U_@CyP?ty?RnnG%7Zse@i_G+Kc`LuMdias}
z6^gC0Jsp#_GWE$FMp^|jGIjU#mDClB(kfD^qqD($AT~!)T(6+u923=upw55endtl~
zZv8We{ex@Nm7Z&P*r1Guk*|h*E_ZDl&(g=etSw};bkDSY#e@5`e(gDiWVW7GdHG0}
z@a4%Q9h)%T8Sv{?n?h`3#dS-(cP}j)=#@-YJ$AdZBMq^@gkl3DAvH%_2baWjxZJ2e
zBY}jgtBah5j(7L4N*cFE2OB#xu&={07T2OEd^`3SK0g}HFgB~10rQ+8Ey-0Rat=>s
z_3$O6>ABmV{>*#z$Zy<+vo&{Jka;ShsG~uAV(7AKH=WD5<M$%aUAQ%P5j@##->F+U
zjCsu)yj=_(Gl1uUroI3S{v8@s8y<QcwW_`6`$onRvx%FT5R4C(6pkrKeJlLL0EFn>
zxT+66)*v)^)2BX%%s02_6bb<XOj2{A4=uu4!{GiPEfK5UI;*X16%@XG4e9y^s|yDA
z#?fC`rl`8I`$;fXIZVOB9hM0rr6cvHLMP|E>rF?~;e~*uU>ROomiVcmZY7r*O+<F8
zzBDb-OIufxP>Uulj%1E?v=;v!?&0$otYoZpar5-cf^D5YCga{n@277TfJWp({|YEE
zsp{{e{9TfJ_qopMM_I4o=N6S()1)9S_FXZT5|>f6AZ%DlVmBLvdcR}WVi0^l&0>6i
zOSSaOE(#+dfAK-4l!tRkoz?&&+SPEZ+47^A-KfcuFN}<g)*YL}`q->wJbxm*!H2Yr
z<|(Locy+#KcT!iK`Sz?YkCne=2~EYd(b}Xd_+|Fe2X(jiy>0+&df0BK{88k~vfD8|
z)zqZw#OU%$w+V-5Iv42V%5DRV=o|LY5OzIUsSyLXf5G)H-bQfi!o2bebqoJNYi$`u
z??@^gjYvMoIPlz-pu>iDW6{o{)zP+CO+oJSd_3`!FL9hg=$BLUZi*{tYNP~>kR)W!
zd8k@KI5(#jwVnLj#YRnW{`(<P!)^})gI6Np!iZ^g`XXBLX5S_!7Re^(BmGFE#WWHr
zBxJ2<q_ZO8R;bt5*cN8z#~$OmiDJv13nL5ck{glf*N<Fcd~W~28Q8O)ev=<P61oUI
z|G~2mG1eLIXBJs)mFSvj={l&<bpqVcl3RgPGq~CR5wnsXst^+x-sD5OuHqoim1%H;
z?#LjE=Jk-IC6G%Jn(vs9w>+@aYI}T}SXK|9+B_%ux3}eo8W?Mm3zQ8cT{t6y_z#-6
z4a0RL$4!l#T7ObOrpQp^_A<wQ5shT6AjR=k20QXUAVj@+qmuDCb5CFGq~u>Lc<0fc
zXMapj4$;=j>x{;0Hn^53?L97>_#$VZIeD`>aF~B9^)LbOG0t*9`j%$!%i7mpHfd`E
zI&3{`*@J~W%($=t?)AkIt05007&eG~a?=k2>|0zg`Kh)J?V}=fIuW+LM(NaBis;>`
zGu9tY$Zxmy@1JC`5-o*cA5A03!qwTp5D?@HXtd2Z`oQx_eMYXcsGFr;w<71S_cfda
zt|qc%N=4Emz~5`Usi&>m@R;|BzU;wj;~RnAH*^0S3}5o8fCj^>AJ;=sO?jmcyncdu
zRro~?T#7P|V46aJX_~_~-vs($M4_69w!1a|)o@JPknwQ7A9}jq23c3y*4y_%CeOCy
z+ZXE_&6BBYmHwGg-yB_sw9Djq`f8Yk%v{nJ&dPXVGlS0s2RIc^d_R<~+OH>*v!cVB
zaC&B!NC!nK2;gG{WKi0i@{wal$DH|w(n}S_SMV6By(t%pBWPOailBTGY>4RW%do>~
z9=ECQ*FC4ZbArOAAA4*2W{>C@KxV##ybAj$uItthxOrj+k5#!~?)J%@Z}$}F4GB>0
z3pktjaIDob?ipn_yUiFeEsbY;QheAQgqdOkd!Au9K2AkH3oCPO$TZNZC=Xu(nI0L?
zhSD<A-*!OyyVx%+Hv$Ct4Gj@8r2q|drilvz`7uoR*}2KnTmC&g44{d9?+$FLP&Oh@
zyDYm_50mYOFJqHc{jQ-uhbD}VL1)m$27{`j%29x`)74iheroMolI5xDGW?#{)VR1G
z`OiOwxHu<haYttX$V?cH?9RvoX&PKk6PMnNbaItTj9feR&%n`bVromw^1F#lfTX?#
zx}s)zvx8f=I0%5WuyQOM9PmwB5#mi@vE5kxbzJg~d@MUwQ!>AGGPChWO1ywC*{^>D
zlWh-{tm9`9ad~fRLPmS&XJt;&p(JhWuKBjC#B6Q4M;10QM1`bMv2#i7!)>4EZ{jO{
z<Jlh8>x`<n-g+LS&FMKV=&I|9!r%2W7`8rT(gVB!^APyuO@~E~+r^j@16wC9me0?2
z3)V4$O_6Gr#hCq)UK`WoM5K}6B0&0qg)N5-7k*J%uWZ4vaT~hzfy@XV<^YfO!v(S$
z7YeCHjbtp!(rM@|1!QlAx@K&M07b)y5;G&i6>kC(Xd?8GX($yuc=jNBt978eC~)BS
zae1}y3C?R#EiqpFDm>Xesd$lzBBjQI(Rq*|hbouzSyOW=?635r?b^mcl3b&6i$VV_
zRAmT0^oRPCza5k8VfLJtTxA2<UjZ2Du!YO=pviJeeMKa6M6zpPa~Lr?BG}S%5=@G@
z=;dT~Tx*#TUoAHJDrBo+lz279-t0=3`l%&Iz{B~ylG5tng%PuQv@KCY2DaqH^8u~c
zz%Jm2y0>dmY%zvZa+yGihw6|^0=1pPrgTkoQUwCq5Am{lC{9B-qEqcg@#39@TCO-$
zpp1@t<$6QJcp54wNpR_kS|Uw6OD<J<oACqVQs`D*<SnRUG&|}`_O=1nj4h#mvVC*7
zecd<qffScFCgL)K?zM{xhR?=xR)u(jsO3J_*Heq}d0Ed;{5jaJFghu@EQ_0Jzqc&w
zyozfvJ?jdHIi#eNynFyyxtKosg?C^-y144gu8_&Mln?{H5bRkQOH^PT!#zDxLz?C)
z#_)_GHuUF>JeqQFg3e$IeUT<m{4?~7bqN93NaKlBd;RX;^2QKm&Z1%Px!|~29pvY9
zPYF9P>Q33qyg9S;jmj7QpkG`i4@x_MwNtBEzVF$3rB7?ZSx+7X${M_J3Hhq7?*)2b
z!g7X|{fDBWBm)n@wMycw=TFjt&!P8POoR9^ayD$54@<K$7hm7M6}G@UT%K_18&s`t
zZ$<Ks5x-*UmRp>i8_5@D%l^0H5o1K%v}ztZ%ER5%{iH|te+A_mv*W3&Kb?3mijlt6
zn~6wqYv{74z0lp<iGQYbozWZhay7KciX?oCxXlik7}mznrnBr`axg%ZcEy;pKizN*
z%qh0`u<WvzGhetcSC?q8hKkx%omv@|us#R-(~F-rQV>ges-6so(NU#nitH3y?JYD@
z8Y328!3xPA29Z+%J>4{mOw83z8yo!}-G6JgTgw6o=>R3^$d^4#Hi-u!kMN=XA3BNE
zqZX8Y_177m+DgIwc06a_9^%YtH*ggVg_AN83bXlYv<}Ja;JA2Yy~^-$zW11deL$6T
zkd<)O1bueP`&Lmk-PxcnF`6nmN{%)Mjs;;ET^3QlhF3MFM*cXfot%l#IK8^7J7wB-
zDp7ubzw}^Pl5np_<kozx76CiULg`0Nb!M<)mUTeA@4M3}`UhSr<>nR@z(oe9z2$kf
zl8&k_%;zW7vIx-Y6|y-W!g*isP-cM;=X%T7qy!l!P*`WU3@PJJyD3)#N0Y;|7O6bU
zc!1xbPkz^Sc}wTL<xak(Of6s@{{7eLMVGE?2p~*T53c$)O6ZptM~$ptP<64rw>JE;
zVc+C&;M#0X@N(#2&cLg~Et;eLdC5<|WoHxN-?EdBAeR+7;1)?mN$!sj7OS9?e1Fr^
zT>C7c%Hi3M=nQ|3=dydG+}Nj&b;xjxZn{<C3Ko^^f2w_LmEDpnUEB7T^6DDy)`0wN
z)L);D?~RIvLTd=K4`!x<WNQu|Q*FhHnFg!Zv8?#qv&uKjFCZO#cJywr{I4cF@1pHo
z?tuO`KQmna9I86%dSzu&?-iBN-=o*RDy67(gs+q&a}KQ9-+q46mbWOOu{^x<6&oAd
zUXGQjD=pa`rD+_tKGAG4^woUO6n_?g2F_8FG*o?RJrT?(^f`)|da=`^MT<OI@iXL@
zUJIuN9#u*lA>gj9U1kLx(Uo7}CY~2=CA|wp?G{qK1RGH>f~-q>3eLy?lSoK!dUBgv
z3ba#?%maORdu=@lxHk)9BoN7sD}6GO6ULNy?gnN(;cronzQ1gUSK?+={exBY%R&M3
zNeSP+SdM%F>$g<twsM|&qPW<4=3Hu&Ph)v~v05q(V;>M1P-c^<{--6KThk%2?SPP>
z{pGT(KtoawJ$e<CmBIkbhAbs~!9kWtdI5}Q$wr2T#xVys`LccP++>i1k(H0#EV^1$
z+tP?BZZ**r>AwbfUfFsz{b^QA)MY)xBXRTeY#@V_CgU3VF<@Ib5>bDI0WlFG=g2JN
zQIomglCXfVrfQ&~xL?|!W5ex>t#JLV#Diq%do@4<o&%!1?I(tZv~Z4Ja`O!@Xev5~
zztmNwsy|&B9yRD>dvnJztF-J^XK~>6*LCA}<G*gt?hfstV$X$bE}n^Fww4Dh&j_Pu
zha|gR78zf$G;s8GeRg@cO37>R)(DL~vXE~4w807`+8tU@QJUe_$B@s7+V`ACSSXba
z4B>KyBaju(IqiNDXg(AiJWSK5hiR8A<A?K)00TSB?0h#THEI;asiQ-tSD@$bWECwm
zVJ<!*u|YRl@$ewm^m-fUb+D%0aj@d+j!RKicf&PBH1XxAMG2+j+n-}as#?BlnMrC%
zXPYg1)_U8q>q@502Tt?brY-QpMpMRGF10c~3`r<MXY|n{b-?^BwMaJs8n0mfIA+qD
zQRo0-*IViGqhilR0(U6B!q^iUG%TzO5?4S^3Lp^<!1_L2Y0_Gmtg^GS78|On)Fwr6
zJ}l4yt5n^@YqZrQ)wGdqYPde+jrpt@U3%o=>@~E~zHt?+9`U&Ua?Yu!f?>fA`Ts4b
zt9f0(&86fTB~ZFKe`Vyd&f{Z0uF5VxY8=~+Cn81?2JFU|B~_Cyrhpi$jM|DYZ(Q<4
zkL82VyK|l_(u%dRCfoM*Pz1PNT|03PhvaubOnhsW*L7vr9*^NHUw58S1MAG5T2GCB
z`u-P8jbCaXyL(E#DDf&uw*;%`xRrnP+Ys;Ex-rO-*WvVl8Pwqv$&2x|t;<av1Bngu
z!l`9~cfjJjp9<|VG!-y99*kU$|KJGNxe)XutGo0B>u008cHX(dU$0Oez~{<z-e8tC
zgV(h?_r(Q`=we%gqqAAe##T?e$-osoxh-+*B6}|cU<BvYuSZvV_3_ZUaXX8h-88_e
zS5JF)Uq^Ry>{t(vK9=I2T^wEQD_H4tDJ`{EubWkpDah^ow%j32ZO|DenV!rBag{HT
zNomUx0$+4q)wMBlHm#AKDhz8(S*gEK-*UOL@X9$kjLV8`LO6En(O|*WalkAw{NjsS
z5V;=@Yjr%&(3a|!zkc{jO~u{UM&gt<%ls}4E_QOIKDW*4ziSB{kd|h&&z#$+*B($M
zuQZhmqt7B|M@hw1&Fc-L<S%EH^~c&s{fDzkD@RGuD5!xDWcczCrXq76r8JgZwgOQ$
zcJ9h@f9`KL+!Cv*{xS{Wu230{9;e%%mU+U1_LlGfO$)u`4!i!LwY^WPReNp>-xE8L
zJimz*19V-s;V7CE;_cFLO=_v)-c{VK&S<LsJ9$GyKlgYk_6)Ix9kt|p{113E!Ss(8
zRTsE{@^@T;%N5Oux}BSGo9Hzef$N33+=wg#U;LPh&7w#Pni$^?Raf%D-o5J`qAcus
z`xGl}S^HY*cnxhOp)@a-0ai>&pO<0E+FffkZoJr^5(eVor{S!;N}0c}is7$+@x#(y
zY8y!0xCD}^AQfcJW^P81>Z#O18t?;w<0<>bFgPwPxu~M^`{XX*vEy%xFJ8^mZ}?e2
z{W^B8Nh=@!l%VLXsQ8r`rWBj21~=5LhA+d(lsi?X6-R;YoB`*>(V%(91@EWxHEpqS
zHBEq(*G=1b>)fbKN^~a{rCy9W{0*OjGIky^5nmrM$0bMF$JW)>J=&;yyGIGJ-@l%>
z*jbcF-C^6x!uSVT$28;t{YquT5O8*?C&;w_qKsk;WNk<CVc2{M>a;Jzz8ABy4r)!K
zID>_eDiM*vmRqZa3`$Fst)SWck%}uBmz23c{YCjKXSMElg<@5z{I#kN<+&VcN#Ol^
z2hC0qft{q;_Rj4EvK3P6vu4+QKj})_Z4HeUwI{tbN3EPU?+(kzXZyV&D-anZLse!C
z<~49My!u+6*m1Z^E^_K~0t6`X=(Z9CTlg#kTk#2;O7f=%E_dUS0s`tAlQ$3Cn__HF
zq7xA)Xa2Q_@ZFf8*?^Xu3>eO6IanyAKA7YyV6aU)dJ^i4toW*)Mj{rbE@ow%nRwcn
zl9DtAdA1aQQc)|Mcks?8En3^WF)BFRH#g<*NA+Lh$GfAB;{e8(f~dgv4GBm3B79r2
z4j=nG(L5dY=+hE$Z<DsEu!;MJ8++u9&U2gpgxP3@iToZ&I3l5rwr&?4v%EXFMw1x(
z4k8za6_U54yRVQZ<WEs-Rvhe`6E@v32TNFuJ8M#QF*ZS09=dWu!@!}Z8{Z%rcv+S<
zE4$pcA9dO-JUulwPsHiFsDq;7FA+!!WMW!apHJ$$fH>ce@6kUPO=>mAS&*r}Uh`|E
zBYn_Bw(e<hKQ7eMq`yL=djmQ3u2N^jA$me_eKjrX-zkqs^Ov$8q6_}g@~975&T^Ii
zvH2GvU@T?6=M%<ZK$aZPc{I=Pzl2QU1JI=kW}aj0egaz0;Lg_mGq=tTmdo*J39roy
zsk>{Kn+>XtP?LkP@F?2AS`HdH>O4x>VqF@dVdK0#H;<pNB`e?L$)M?_ZrM59lt5Mk
zAEU>6GW3a?z~^#e>j;-=rvloY!bH+q&TAR*$)DO#f1cIxc}XBXO&^5!)om1ib-bFd
ztf`Gu?|s{>8Iv$SV)eD~7*s8>Nyh{2q8<LhgLszj>MWBwTW>yHCmXQk<=&bl7gk6u
zQ8(HJ*XVhrWeZ;fTND+}vOe+wmmp<+vzAz_<YbhCfU$M5^jk309boP3&G5gm)&-YZ
zTEfL2(PR-~e|Uru8%15nuZ#34^21fNoy0`fUXx;nroq_?r@0_?GT#}@=R3pZe(XPE
zvGYFKX4Xe|uI+X-4Q_RylZ!2R6#$J*{cY^$3@0EXKt15M2NjCZu14g~4dgVg!Rg0y
z7w_#=9`N(6ty}c0u5XsJCV&I-n?Pw<C%bdZ6CJtH<?^xNOE>uupB8#FW_5^_^(bD1
za7fADtm-McBUAnERkjIZ`MxD5<6<dGM`M4o==99b*%P%GRY&sk>sDhNPQy;d3J-LT
z$rE9Fpf)g9su>So#Kl#RT+PUYLeHt~{3giZcT0wQ)oq>)WW9H_;;ST3`Z=T`Z43T(
zqMKRh26`N-!k~M~Snw(dmoG)6i(YOpF2Evx+y8ufjnx@&{HvXCopzO@42@J%ezaJ?
zv@Y_?xc^P5INyGM`|VZeZ1`|~G>K4%^#~3P6p^T!tRY8?orLM&($a{QD9f>w>S)%+
zy2sRKf#6GBgynsAdqsZ|pC!CZSFoj-_%J*ibB(QWUG26Zxbn9duWx)j@u2M=Bfm}n
zVRi(S7mo<f_6xd8=~4Eoys8OxO`8y;T>tBj1*dS2p3KagB#-<(r}>a9pu9RD>rH6J
zWAzIW#b!|pK5CNMzr$WlB8z~=$3W?nD@9G+3{Zv+Sg@|8-WK3kMKUJN_<@A5@spgE
zi5G)Ijh@0XJAULua0suoFxSjvpT07+*PI3BI)igEQc+Y39g;QJ88b<0!Ra=S9anj0
zjQULPc^~lTSB$VJJ#P#xc$_&sF!1!ZX+R^zWM+#Kv*--hHBkcDT)ArW)!=GJT@m;a
zvruzljsz^fmHW2jr#*B8Uj5Y6o^8DQF5(H>^>{<|)x;C5ss+#Wjac=<loUwJ!tixZ
z<1L5=dCSInv9<vc>mZ292)g}GZRBA|^Dxz5q~0v_4-q9Ht!3)1DHHIGm5rHM5}#EB
zJ1FkcgYZIxFjPY)s~qeM&QHzF$jQjYFsU5RUECF+^W2MA70wp|*4b>e+4^cn<7eNf
zQy5^@Lb^&mwXOo@AF_+$V<#+<o^pecZQ900Qa(Q9PEBXTFL=@t2&J{7vVj@jq--AF
zM3e7HS~E+2UXg*z`l?lj_gpGiM#h)^8N9@yLs+|9GJHQj3T6QTEJLXM9$^+%(~C`#
zxa&S0UJPl~e7%+RIx0TC*itqFF;l%s$KW=uh7bARM~q>CB;7rLy4;DPR9qF2WK)>K
z%?ePaOaWsy2kiDBG{}iT>qg$P@#%L`&}&+OT3#`yBjRIWMsUOR<9BK|gms~Ry>J;6
z8fJ$hLjM|8@s3U^u3iXt(w;n(Wn>ncuKu4@*VCU9Lj-1rJ>s6`H`T^kXTW8eOaI);
z1$s4vTRB*XBd0rCpBmFv($39LfS3;#6l_D0_Qs~65KF|o$p0xkBIu#`z%}yd>`!rY
zP2F^?*#fnws=Gu>Ao)O=5f>`1D?oa4`5mYZ)IZ~gk(zkLp}yX06Hct9_!Q^p&q&0J
zT)5pUYJ4%`{UTf&JRJTFK^*y`^6G0tsj+wNu7j&ws}n!`$)c<m5FZMzur1Sgqd)Xf
zR@Ha?$yCn~GoS5m16Uu<BNoLGm|8XZ{Vss%k7h2ZaLpO_Sc4n2IW-GrbqL>H{1X~n
zRr#l`KlWFlGHa@h$_id7W{IlHATDm%35^v*5{&d+A8&(1ELfoDgPN_<WMhcQCp5@F
zyW<i43znPwb6VFXtC&j;@0}ibXSh-Zi?!S%fq(X{8pWD?{112XjQ+6~cj>8W1t=7;
zMRmHr(=l2Va+uJQ)dT0Ha8X=gm~>C$_0BV#0I)g=(Cz9SvlLt*u$jx;=xeAtBhwR+
zkC18@fub~byp}FKxbcRIX@dy)5Gh!*<|uu8Yv|EN^PSNq+CmeEjoIYr2f|5La7k#=
zrFt1|Hx(qhOicT%weVOr99!>j_duU~KGD=#{BW20uV;fdI^T|JsS!KYoiVlbFlLro
z7rAU`aHt*OKqKlUCEi~5UZ5aV5f=os>Y?(-=WF1&);YmeMo-P$G>o~@Ry7B==&Z>1
z0OwS~a3#RezNOJL!ewG)@bnlfehz5D<NL8qc&;Z^l~YqQo?<her9jgYoh_B3qN;sm
z<Wt#xl*q?>*!2wnSuy|9Xt8p9T<AANO8Q^Q8vTFsOaIp_m@7IcT>M0ME)NhuGM6W$
zr(Lx2IkRn(R)5!@N<3G^E3;d1j@7FAy_E`jC5+$d^vK<k98q^iG`whEAJjWB$oZLA
z-?dqxyeTX8ht6?cx!P>#_Ak_}c%F}SjS0@%IaT=m5VP|k_5Wb+yrY`T_ijJuct&Sb
za1;dvfe{syDjliEQ9y;jU<d({fV2P!3DN=tfivSMLMS3NgffC8kdPoHgb)Hsl@gGW
z04X57Hwnck&dquMxa(c({(0|OcfD(!{|I4a<#~A8Z-4jx?DC|7j9q1GO4S42pM5|7
z-)-lO-0EO|(S6N5#Z2g^*>*rze*Yr8mXvb4%IqcT5gwAPw&MKSe-^RWmuIl}-dG9>
zi;#x?{&@gY|N7z7z5dX{Bh!wAoqt+s=(mbbJv9jBd?zlHY8)S|&-Sz$9r@26|7UOf
z=h*lkf2Me6H=W4yqw!juqi@OZ(-(G_u$!MON|--o+e`4E!SYG!wR;IPr^wF{-go=^
z+b?5Pd4GN9?W%h8_BW^R{`2BL``|yv!GF$$|F72rrdV*gwmo?5XV2aw14lv9TGP1Q
zGJPmP+eAiFBR}1LnyHL1xNlx7r8s@Pax3${JyZQ3HLCyZz5kg%{<lZgzjy!7+=Ty^
z-}-M+Gv3^nt(~|-@^{{U{=fX>|6iY_{gowgRo%)+B#-~FB*c6%T1DNQyLd#qBLX70
zdB90{AQVP5UA=Y3LP;rwS5~z~`}yuhwuO^Kc%>E<g{BXM-Cr)hlx;G?eC!Z0PX7je
zKIHHEQ_bq9?sEGcIZwRvee)#@EaH0)>vv_+##6qHYnW_QviE4@-E5zlkN<ZYZAPB$
z`&Xzln9<f`Wiw*cy~uY?58gO^>~FY#Jv7hB^f?x6LKireOR~CF;18TyVXqf2Lmku6
z6PNPeXq7ht{+T+G@Yi<<Y3tKVt8d~}^*=G-`&&tfP)>9t+CsAb8Pv3g&R}l(%mxf4
zz4m-huL_)9!PN6_Y%j5&W`K-lIDHcr+=2@}p~t*RVg$rcm|rZ|uBaRhxP7<dZCjaa
z<e>$y=WXeHA23?#3hb(bCAUh6W(4jX&cL-&5l|W=SJn=hx*x4dQgRX`9Le_v&wO@M
zE^UfP_cs&aKqgsBG|e!?c*F{E{9QFY$;9=7j9L4q%N|YvV~_`#*JrHp2i?mujba(m
zEF)EkgjnE@Vt607RijrV^6@x@?GSjdhhzb6Ckv>0Ofi!thbOz4=y~}~zX>Kl3aLQj
z*Csglr+5-dhw)p5C%Frwl1!Wpn2_?B!AP}%t7P3F5N4H@vO;XkFQ}xE;ePdfS=7Eq
zn3?jD4bw*#CiO}UH){$75M!Wjx88Z);pqzB8QhhDY8pZ(zp{p_$WrX8qAYahiyp7=
zKfidfZI^TO^17}?!bXho^2AaEz4_-YQV81Sbj^+Zqk=7gUy%8pvX^qq6M~y=-D_05
zo-<dU+}AOIH^T${EEIQ{J)Q60cefe=`b=qjG9d@ywy;I$VBZekIQdmae!X8?6zbO3
z!Bpg>h4Oso7X7`tX*weAH6~h<E=}n}OH-DUm&pZ)<mYRKw)I~goi9!gHqnaB9pS=e
zJp+ID+dqfyCcnWn3=`zZv^W2nJBTtY1%#h%9JGhNuY=pYBytw6-VF|pN-S|cO@~>)
zwvWZHVyzB(hK<t}rEWYuHSNi6oCNcVx)|_rGW9XeEnM4Hh~!alrfYmQLUW0Q;6`#C
z(bmZsadLLU`r*L^2Kx2=Cm*MnS2BmN13HxG;MG#{QZ;ej*dBBa1b{{WfO2PMx-CD_
zJ6|rE_ET5IXL|ilKq;4MpXw;?ymHCZys{emvh@9(=v}zQ(C`=CSEIE%C(Vbh5yEnL
zS@#p|mAps3e4scBAaO@5I+`6k8qHO*lJXFbjkguo1if8vyX!L1Xvl<uj<7(_1!i3<
ziV{?(leNhB<Y4Q>C*Q734)n8;w^<+xGuBiGq_}`{1T<}eL2@H*lyg9ioSYOw80&av
z53>{a99fm>)JCu3W%gvu4u51QHEMk*?+vHFUwT*qQuDl@z4CI(lk}4BC_y}W3A^-}
z+}=}zgmv^Yy`6N}o?U=wb1kx3=Y?z$xfOht3*APm)1H?i^V8Mv#hh7|ZImyR;h5;*
zW)*2`*{vLb2!5W7fBtO@4fi$M9F)Ck)c53={K>cn^^$L6x7HrIxh$S9VrTCgPu*5G
zJ^npP{A1l!+7*<0j+&=hcR+WJilv91^<9$!pWR9wiWtwfvOYXPhZV7m*oGZd4*c};
z2D<R%372Wc^IZ4_8lIPvhC}OlO_)^i33(eX&Q(%o7%RdGG-j)BZ1*BK;irRF8I{3f
zn@iEbb$%9Sc2`KyA>=YWR0;^!Tq1$U<kD-Ufy4qh;?r|{u`+&jp%5)+^no4RUb99n
zUij-f?O~K%SJ27@II(6Gf?Km*pw-d|*SS$!6k(AG@HC{z;PLsIv&(lg4GaO)xdhB(
zZC)gmH4~j=@dVm>zCh(^tI4yvDu1r;H&?%4Px!}?x~LPozO$aKS()--+FnfQpwz1s
zdwmx}l>*0%oR-ihLsCV@q?WESUO>RIVy=mCBaILrtFX1{S!Z9x1hFEcptSr|T2m0I
z7(vcY&Yv?yd{sFhEq5(8%vX4zQ=GeN9Ob->k_>vP=aA@Rn0;CIcG^F?X&Y!&Qq|}e
zqy-$Q^=D7PIw~GixcolA#doYfo;S<f980$@Qb{7TzVlYtD<0pLlQRw9NIbFsZBj7q
zdJsKxRK}xVflfC@S={R+nVQ09re{EzC0Q8=gW6AF+hHr!ys7tfh?EUrg%Y?e@l5^t
zP{Z!YL;3lti!zD!oPL%$F9V}9GqYm0|4A1>@hs|N_~MMfU<w1>v^q;n(WLomP-K>Z
z7Ut>R!C#(^KvnjtHTw<IuJ>&>hh9usm$gyfm>i}qeW2vC8k6sV-RAOgb4N^~*pY50
zCFs6@D@eVQ@ye!dg7Ha<7(lb>Gn6@oCxEPmt}1?uiivGXGi&majbcqiue=-Cu8Rnt
z5s?ukE;w8!c5yF=4jq)Ko@`3MQ0p1;1R|gC`7_@MZlYYlMDz>avk+v|h(fiHju3Ky
zs*LdYvm=YfxADqPALY%&l_|A-0Jx0mT*^!GarX)#JWf)_S7D7`c9MPvnS`YSM~$!<
zFCZRev=@D_!J`H5W=(&Lzjf)jr&X!AT+9<YY|9PpcUtAtFuv(lvZCb$g)*m{%N}mB
z4VtS4Nz-09*A>O$p|@{4OMr-VuR0Bb&KBR`)fKWD)o0vVAtt#jRQZT2w>~akf{~B(
z>SLRCNe?S&fp+~afS{POI%b0)`w>OtmCgR{v;9rlGZueZkjBz-R&%q0bJ8$jg$w6m
z3b7~M{;`HUn}c45zw!lVDFy!g{&X5}>>`%mljX`}yf`Ddtf}P?yU$PG&$NE7v=5K*
zSyrCTm297QP<D5Me-y)xgN&kso6@W8jH@`L>flDYQ_<^bg!wb)szFXC{p{?tcY(ST
zZYhnr&G(v^cBEYfjx{*3UN7-9G=Tke$xb@?t$~V2GOsKtVb1m|t|3G)!=dCv!1!u0
z%jt;x<lmgmTWU8-+m7PQZ7Env`QY-m6REdH;k0m;jK{Ec*Hy%3WcQ73%9#zTw8aTf
z|7;O5_~_ryok{n(6MjG7y=zdv#r2X0@y2X@EBUYPV)T^YR6A@pO`7cwyc`4_t{229
z#RbvlT%^wkt>ndign1Wt#OR9(3DDNZ)=xC)pTtGLN2RH35qV4D>7?T3A)76Z_l6Fm
zeV6m=5iXetRER@wiu$)ad`6%o6?I!}Hy*sfC++EgMz~zC6k~UX8|ER<-H5(im&$T5
zN`4e2XHwVq$imFMGpbAD>HREO!lg9xwtL?Gn={5I)i3(%%)eP5(e1sDC0=+jlU%oh
z``{<V^PGD8v-s%;)7BJ4Hv3f2bdW{aDZO7=uMv`YqX2}t?OKRZB1H_a01Yxt!?lR{
zOgm-FdLnurvwTPl+4j4|i&p3K4U11AYIl{G`3k5Bl7{LTxTr7%$+Bi??paB*<?1nn
zOygLfoK>b?o9VOY(?zxDQM#dr(K^SZ9p_kQ88>vbn6&H^p)?5bFw(vC+oFnA!Ga0f
z|8@I|*N!_S%(5(X=lc&uA%#<V&KVARb|n{@zb_(8yX|(5F~FaL8BsxR>9zamPrxfe
zCG+x|c`YS0RFTFpT$I0ia-Fyp7C|9kC@F~Lj^dvVVp}j>(m7gVWYf0MU*B<K8^1MB
zfq=dYZ%lJ6!@ja(@I8y;Vw*@`kp$1glGXk2>a!Q>i{>{<q&KFbrPo0PMY(RsXIgpv
z-7BB8^^RmJK4Tmo^~7U(nSFuNQPp&nsC9o3*cc)1&NF7p+=|VAshUReE*tAoJ<)t-
zo8cvD(M2}weVgdt8iRfai$E`7CqEJw==jFHbd0W%PmsA=4IWL&$HL1ZmQD48x=Xb=
zO`9X(yJTJe>I)cNYZInH3JlWM>>r@hd>5c}CokQLT{>0D_c^_XXt<qmrL<(B0IN1P
zW(9L`>@;=^nmX#G4E+T|YLs#0&{;oN>2@2)C=+lXAs=CqzqsbiiZ*Vi5~|SEKlMcf
z4=F<xD~Ed@7cAABc}_{Mti;}SnVKjW)>yu^v@@IW)WQ%QbBncxR*t*BeL@v7TBC!K
znP8CXUb{5x%YGj#ms7niQj!<j!7*1`%O>8l65w-CoFQ7&yHwMxFdjFEKr1_&h6mH9
zXZ(Y>GviFRauU4L!#L2#E5*u71r6$jqdb{lm@CkJP0WD-0fzODgZ5x-&Umm}>MU0I
zL06?<rr=3bbSw;%`>-Hk9v0ag_UL@UoUm=`2i(IXfBX#>pFYARx~gX*hrt>8BhlV8
zJ>gW0Ld#=Yo~4h_tzjdNhO)8%3~h>}%*!x4yZAKULJF_6Yifv2IM{Kr1VluuaS5PA
zUuihvix#PiTFH!t=akN(sx<4u^^H&Le+rfYFpme-c^L_al@vn_#gxz048S=s$U4H&
z9;5t`^al0?rKcs7$z{$+2b}l@8`6ic2@~dIA>s_$f7pHAdAzpI40Es0!mYw_qQg^E
z%!vq3x$mMpp&2krf6;pH*z4%uR4jtL&Q2fPY^vI@q%tY!<{~T(0*P4f>8z>rS;D;0
zA}@UWV1zJ`LTMZ;Qo@j}uu6s@h7c(%B{zlsqrqvN9~<{<VL=LoW)Lz+L&?S~=P$I>
zvJ?zfSf|g&;(?*k68mDb?P;(j&`{qJYKbzr+OjTGnaKDhtYH_zmPF+Z-5Xut$C8R4
zNwlmqce0wSF8H4}E&hg8F+~GnMT?}OZ_#Ez3mnTY+8ZP>VPATDYD>j3FylhBRT1->
z>tIsi+#*$=loxhrxO*|Qgm-8&xA&A9e@Yrjq8YU$;KiY@v66%gX~E4VNE|a&7wxrP
zRAb?Hw0Vj<MN1%zlstX9(vl;gEC~-y^%Qu!hl+oFr&7&}j7*ra2PuOUz6ZS6%nFVv
zZsj41<jg^a#_zW-jISY~-QSH~g?5!wdoPc-^c_S^KH~cNu>4M+Ktyl{?%(_oB6HV4
z_d2aaoN@5^V|O}#|EWi@EZV^$Ljkp{L{Wk<ktivNNeoSMFucS1&DL%IXmS}iXO%iV
z?+C`(*~&Mm>vGY^QccO@GklAZ;2HD@2;pe67M^Jy-+z~4al(`bsYB`#M&y!ma^xil
z=AE$cpaSMXK3OfM{c%uZ-1c{(XOt!0y;tL5h$3q(660(%#@O>WVnKokR#=revDSu>
z_sO0<L!yq89p*K!+^Z4X+tqrR^aEczS#pzo_3L+AhQ0^MWe4r75R?cA5j7>0myNUP
z_uXa+Q%@*boo7_r9=0;*1bMZ>=LJ@7wgNC>mPUxMKW84Q*kH0H!ZxI3HPo-{gl$W~
zg9b>F%g9!f9yrVN8ZoG!(?we}`LgQ*aP77PK=)93M^VM0&WN~#QK1HeG8>Gse}TNX
zMyBPCXuR9mCk<QuF`XmBHw}L7vGC?nJ#!c6d-g{Y{ebc?@TFg0_N}#i`{__E!8ism
zo00OxFznz~HD72hn=`#&l3W?U|6nX)1B$#R=Y$@+vje)VuY*2=$*GoW7A1G-8PO{7
z{t(<qxTyD9Q?mm9OA-RL?tCKLhx-(tStabPLs#`XPE;#CdPD_!SGD)_FdVT0aPShj
z37M`=D=sJvv@`2`B$&C}gc|m3!j(22-?fmV4rzl=OPi)w+;TsL%}(+bJf9+UD_tkh
zeorTw*J4a_YIA(K4sRwMN9_?2gSPSc0#k_<TQmOR<1<_KP`;PWrgnTQv(EC~SkxL+
ziDeqSahx(#XR<f)uCqsie$qgHY_8VKF|-!?f%EKaRmf;CJVlpJn@^tANctj(+^7QB
zql0r69tu8e>+8Zuz>!BY2GoTWQ?$!U?M@A?zmu{TNYmlSaYxQ^T)6-PV4DR}>8(J0
zIFpF&dX7cq=Mv$uf1AaC8xta~kg54)cP@NO@U5*oUWR_~>iw15JF`uG1iMh%hlRJ8
z#BU`kS_i0E(+p6k-f=nwyT>iUGmI(*>7}cRGKQ58VWTVkbz+$a$Ki-rhLdf2X#5#%
z;mSAKy0p;O$zdxg6iM(4>cThbBRWQHJQn5vWU1+Kaa2U?@TBl&yx-Hmz|MW^hpvz^
z_{5v(zQB~$ld@OhUQAm1kqhjJ83=W`;FO1;H?=A2uEQa^+LhVk!n^XPvu<TAy5}@U
z-Oe5!hinYL8O^ye>9Z_TQDn6vJ}+j8eYc(rErzaT$nCdSC$ZkpMI(4TNjRZ22qjPI
zmQ{K8zGF(-ZTWoK_k)lO*tyOoLoM2<vn~G`O+EMWws(8l=f?eFH094_(0uS%n9&!^
zCXl=wlL_iG8-=i(D|<)vH|+so`u&3ElmUGHM#1EHNTlf5WCP%}`91P5W$WaoLQcg1
zbS76mrol(IMxm)Bf6h21;!+nk2k3O2>-MHf$<d)F=`{0R(k1=Z-B>N_M_9xGuI4dV
z{Brw9LJW%RLJbiix;~WMlX_jT#nLMV7Tzo1JH0whTs^OdF7q2*aleUyhfaIoM$n<2
zySZox%L}!|vgJ1Hy&n>%MW|H4oaNc~!?i(~ewslz2@!5vvbtUEUFQwLHLM=u=#kPG
z|JgBsYsU{^iqrQxM!ZbRX;^8>%2~tgUEPVtZh)hpV_<RgSsa>!`CH0^zNo^upGPeb
zGmT~AyuN3z0cF(?nw=~0#p<cLtD8WppV)@%v(5VA@U1@IgD0(F^=Y#V@nQG?gD4!2
zTXd^dl*TW{qO=i#F2k7OGSb}c#|2+eN==>K$Q>vStd6S(qf?K}*gx~Sgk+7uxb-Nt
z1t&Z{PQD5<L|Jwp3MXI_t`f53!;O_psoNNCZw!rN2zGY8Goo)el-?$`iD)vcTR$?q
z{tKh3eaBJn2SnMWF8ICfS`XKqk>G%zXRPAwx^4mCwZVIZ8_!KmPU6k(u5#`^kejn9
zSKr|mt6^`&kk}2@@#e%~Atbl?@WiiWM=TsjTE`!~S?lbLJam>jVk57<PRW)u7dj}|
z1X9X~zJY#wvUwFtEQ@~*??MdyT-x5@r2ms$$_{#;$^tt|izs&~C3PUfG4y=o{%o<Y
zPd|Jo=1!O$7d5t_@ZeFJB0hWOLRZy-hpuU6CofH*dlm-d3-9I}?M!iqcX(Y-K?8NK
zcC?i^!!&QI-dYOLQU84}YiMlb$<QgkgLBg+gQ9Pjq#h6wD#VIahCDcT5Z_9d29_O_
zt+}bJbAca2*7~)XQj=CC;Q$X_gU%A7k)&kgQSla~BMpo}>o@?0XkKpzxI5u*HoEFw
zd*Jf-LqRa|40ruLR}k;o?`+!_33l8o@O><)<9Fq37&>7(8Om0D;hu$Yns3FbcW?9;
z{C?|!=x*72{dnu{O5Mt-#;WQDY6G^m_hg0B1eibZuDIcx=G8yCl+g~w3J24TkyfHv
zFz!orNPrziApCZ1^RS&!7u?y~M$)^@W5B7E!8cu};p3hC9jU1&3*%~Vd1)D08RsOF
zJ#CRR?Yf@8ZR;CpXc3I;taGZ?WRBRk_0vtSu}nT~N3O+PkY89QZ+J3|yh@wmQIwn^
zZ)I07$WYGm>1y}U0j&$gjjJX;8B3_Aow8suKl$YYh#}2W)tysryWt%X4DsbYUU#h~
zcnOyt;IoC-qqfRB99~PCg)R9L_N~PV*jTJ0YV+h*^AE;@r?>T^sHc3_l~vulDD5Rb
zeIj>rjX15OX@tCnhEl*(*L#8oMh`+b*xz(#kF4szm}oMh$Y!kh%C@N~j8~^0>A=tr
z-$mJLC^quM$oQE39X(hb)WxgP79*i@9mP_l7ErF!wWFg06H3gtX8TFPNMaLwn)V5E
zM~$-;HNVf@pFln}Z%Fya%`gRBW6O$9c`CPc4JuFGd(qeRB#Jco%kWIA`IkYP6QD_@
zYTyXkcHQQIuxTUV`z1zrKdh~vYf!6?wxx`EvJhG|;VGpe0cOKx5D@9&3eayd4<Iy3
z&?OnzVgm0*n6`*SE@(pYyBW5<*vf)RO3HHn3K5Q9*Edua-W~Wz=`30!2c#Iu_P8j&
zzXG`Pa*X}seei|8b1iv2Pq_8X*_CA)=4JkcYSzCs+?QlZ09}VaTR)W-TRt}b>d>A_
zHX^ulT@Dpv(k%UfNdR)iYFeg|vx1tR)HkLvpqalhe4_YNO!-ZakaU}M)=7%!Hl!_;
zOc1hMomHu^-v$%R7AM*nsb5Zv#Z@U7k!bzP@NtN5tg`?4pr^2N;S0_ryu6WrPeQK6
z>CrD6uU@x^pD6~7I+^X<y+^evU{mop5;~q)5eUV(yPNN2;_(>N*nbH@=l@Fxa)mO(
zuhrQ`VOf!2*yBhS0LbyqrtB-k{MVDK@@3@pHA{k9<-+w9W;iF(ML9}pGvfeH_xIn4
z?eB;V3ob2GG^6@@_Pxe~HlB{V4*QuCe)Cr_Nna%<xziyaha8ZHyL0$ItUyHu5szl<
zu=l4so*Fo2H-zZR-=<wE|CZK#SG+4W7<+6n)ufU(W2f#^UA4v;DGzqvp6-4gGm`jV
zG}3a-QORh|KZIhor@jGNb;YjkQ$6U$Q6d&P1wk&lK4f`$+d|HB*q}Djq2A9Ak$>UB
zxdBEQc*u7EM#)1IT;poeS_qL`D<dEX6xM0f-+%i}TK{sw4E-4D%WrRMLJEjqG&o{=
z(Dupj6!Yt3)uxu7M;0rHlw4F5lDxVyS@VXVqJ%MATYdElq$f-Bv7I9H*<asTx12=s
zO}N`hZD~VHgR&E$sMa$L>aEpqClk8>vjUThCL{K3*PcR4i<>uN!~Bj(n@GAkyWhWW
zJ&R?@qebKbO(AM9DW?L(ZvXA+9bx~#husiMhcKA@!h)OAQCq@U+rAtD1C3z8ny?#S
z+p`f|P~?c5e$T*Q<1b<NhSx{er(AXb=LxAZ?j|+C-+(XY9z;=WtILKTJn+xisS24Y
zuN>_oSh&5i_30k_drN6A^!1+;zGw5~b8|NKUn`IlN@e%9Ogq`H>c8IZ95uC`-`bA~
zA|xIyITgKE;^tLdvGm!!@w|93aBa;|`yMMRTy#1!Cnv*q%qlt*QmRQa)(VUP&e-Ki
zq1ELTOJmv}AFi4wxCKn`BIAZaRk{GIJS4%5ZjpzM9C94AWkrpFF9HF;4)do|cF7x|
z?hQ%R5}$uTc>xo}ND|LAf5TYjYobzCgzuMaw^8*_blI<5H=T}#y6eCVWLa6Mf8}-K
z)2hiWsgQ1DOycbRQ^-hMAdrXxuvj;Cb)v1CFN&NBp(os`N5a?L6*0OPj4Kcn*-h^3
zc+axj0WfD`_<+yE6~(SWp`up~B6Q_KQ+(g~T;FwyT=*`?Mc>hj9nPgZ?r=7SGHtCT
zg`!7W)3-XNZbuVdc~UFQ9khOIyRr6c19K~z!;?|&^&PdtKby;OnZo}5#Kph?cZFs0
z*!}gW=DiIgQr0C@eV>!{*qh1R$pY&Xm8%l-gOZwtqleO3)n^}vBZVkAr!t?=BK77%
zoD$A{fPDucy(^E1zj_r~LUBd0H*>F(ukA*1LA^_*)R3*!cGaKit(N~>Y3j<1PzAKV
zZ}kE#bH#<v8MZMVCKdR)jEYHi1c7mll%fOHA9USI&R;r}Oq(=?H0r!q^~wzP?_Jxs
z>~jWuo$2`KK70NTmbY0{#F3sVNfqw0SPuXb8jbSgouUKQR8|p6FZ}jhE46QD&TG6%
zd>*ITR(&s2hOzlvQm$yy_aU)Qb<DhoWB7m<)O8sACfX^D6hT}*Z=PG0wMvKCAMJ=&
zJ-%N_sGnl3ta-99lkX0S$`Gt4e#ip5B;Tqcpx$oiP>7vvsz>#bq@^S$4)@tK#G^F0
zyAIPTVA+G6ZLL7kilt*;wasALR|UB{kCffGRD(F@M=8)~n<swR<0}m>;bA^Ktad7S
zH6M0Vzi2-9Q;tbUuJ0>e@D<Z@PI8+pW#Z^3D_!xlxJB%}9X%w73Q0T}-DSdPHdgjV
z1Uy{k?OBV5v{)G=^vRP#N!hQwhK9V~0=YgD%<v!<9?hz8rqBU>Ilwy~2ah+KPO&1e
z!4ev)9NIY&;<sYt#df2X&jSI`Th*R*DkOzp-?PCWer0+vUEgsKB)u?pv~CCU7dH3x
z>m6in49z^P_?(^gOIxT?7e#mbq}u&egh~zcb9r2LPy5PC;V<!JP!N>pn2wjzFEP57
zwfNU}e+H(sT#UInyQjXh?d-vLg>ry6ga<PWH>7)v<rugq=pY>*{9ptybsNAlO=x#u
zsrTriBuQU(#Ki?Y1(R<KW<xy3Fmx7_SHZ3LI7Q^#G#k${3E^VUT+`a>S$=vxS%Zj<
zlKPaB?G%m;D1NmyVKxy&_PjJ3FZ_(5QG`yX-<%s3Uyc_>qt`w`)!LMO9?ZlR2f{~c
zeMbi4<;ETe?c`3QER6H6)Q)_iNN!naSt&Imw@e-RDsrKg@re<$2T%W8jedg)5l#IG
za>`69Lqi>#Gl;7!<lQ!!&RpQMi<x*Hl(1NpvQ!6?r-AHrx1!tEH|hJZLx-^7oav9?
zGPv&uhUyIHPwru5JqZ`7P~<Gs*8r}bEUO6gHIfuVq)zsY-brow5dEr&-59)Ztn`K-
zBHh`ZTF;EiIp^i@$3wlgSINI4WDNIckN{o`D3Vz4hDCiAHMBO@9~R+q=_j`i@SHu@
zdq^~E^VIs-!8=y84UWG{Wi#kDGEq9mocq{+t`eRR_?*Kby739Kp`<^9&AVWglw*KF
zmYC4Ex4n!5A}FIG7DCkjWY&6IXhA_Ok1K_qsS@@n&ZqYAmis%}3c`Zkj%stgtNPz}
zF~ADd6Qt6D6&kjw6i+N58Y}!RRg`sqhWyF{z<TNJGMkWlS<8i*$B_=OUq%4`L=12X
ziM5?k>)@FSZlMMzo+m3$95aAk3(L(`x}HsP%grfCyxaPEtZY9$K_x`_)Vo80Ul>B7
zyVJv^hr~e~qEHjvBghXsVb|C<6~9dm)qGM5M+?eE-Pu?%04Cj!$z*M5y)j02n_s}I
z$zw*YY}M4(H|cM$Qd~xMu$x^lN|vKb5+Mtw${hj0KqCOF#+3!%3$tlXJ<**kOQm0~
ziw!c1TuYwK<BM`{ouTNxjr<(-(|Yi-+PjzH)V9rIzC8xT$(LT&9JPH?!zl|}aKHWh
z!FJ-(%=Y5DjK4EXQxua@A?Z9(o6uZd)c#>ZBp@V(*}1vNzN+-@X8@?JrK_g;a1mt>
zLBJc0Ga@KNQ4|;8;+!}O9}<44J9ewC;SHI?PRb)h=oUo0*<!<34Iw}TysE#-6CD9+
zDoasklG3GpMd~5xKeA+}o(*eA9}}axT*`G8--t7teJ~R(MJuPhsxQPnIQJ-5F3I1c
zm+VA0<vv@f0=ffP%-WO)*(<-}!V_LP-_1DkIHq$8JM|>4bYQj`ITzZyf7X-#B@Z1h
zD@8;NlI|Cf9xZul!{gN=*>-CB8tNff{O&eF&vUor)bAFC7?c(vH`=7c>n!4Ika#*O
z@Z*A{r;GBSF#Yl=JuR--&bDN4Aibq=X=Ry0>X`L@Q}lU*w~yX2(o?V_cfTV?8JOJP
zb&DXGV_h@+<!(eH2=UF`m){lAZ?2Cz1Cjp{y}U$J8@d$c<ZwN4=}>7hGvWJ~)8z+@
zj9K;NI)>f`^bj8xIuwmWqq@Z`PdDp@vIpdXy*+8|PVx$Juq~OSS%s~}z9;Boq{HrO
z+l+t>jv$kP6qub`7<v*2BpC#c-gv}@d7S&md7jVd>m(F;5jW+~S}E8>MDPl!knk=r
z05ZkL5LTmPnLzz=Oe`ieRq)|0z3+|At#JPL@h2M}l-KE<Vsd83)GU*J>Mi4B{cSPr
z&T5$O7eL2E;_J}v&|{X7lPZcU#j+a~5S-Ho46p7tsu!JMDl1A25mFy?*;z?NUKu4j
zc9!vYDY<mPinX`CZ8Qcmb_80L{7!HVU$-)a)-WHD+t_$%vQfUE?$Hi*9nvjzXr#RP
zRJ$zAHeMz+Et6vx)cjabvw^LRnG;Q-h~eh;J2%R>nkOUf^<7{v$fE#*w-;!QVxLTY
ztZ*Lfoil9+X9|HcJsV4Ju68c9<XY5G19*^r)iTO(^w8$psgsPZsqMsQcL9I0x@FO0
zM4z^<JkbdMmiNth`Mg()L8J_;%FjKceAXYbtFS3xbL1kp*rp{KCA6z3xtZa`E{CJo
z=FLJVN=?67DCPcdH5o_)M$x6H1^6u3TYkeuA7%uyPcUZ^eNFA3cnw7p=&+hV&p7Iv
zndJmfSoKZoK<`_@nbRKP<ov^DCBgAm9R7g%np~iU!2Lb^8a`xs?pDS2)7U*u;QEgD
z!&g^JmVesLY8zDgiV!GOx9MOerv=#Pr+nSX0S^5YG7d}dTu@(|<mWnRPa`F68F&o&
zhB3-uZ64B9t=PzxuaKsUO{skL<MRTGCy4dOhc@B9proQW?_LfaQPm9Xp`2MLDjzRh
zq_)5Bp*sTB;>40Bnr)cTF5oo4mFtQLi=8g-@Qzffep=~iNRrli{mjeuIpBXXX9I?-
z<PCH+*FWJ6k<D|xJC5c<PPXu_%P9q*WAeA|-oKE$T|1*F$oQw8v?#-F@cXYn_Dp=!
z4t0z_773k-OL8y!`$qqy`5~};;bRacOs@-*)DRVNO-d~_DAjgy^)4$#FcHsb$&`x+
z@1{2ngkKuc973g9WAOIt=~)0j3ScUGS&h3^$+Rc;4Ro;#&k7k<Vdx-A{&O;1T@7UB
zyq**thn(26Dd)F5s$VpEN?0`c6oyl(vNy@@IDvD$`<o7Uob@@%p(3{z&|#0B?=CwX
zsp@Oq@@LQ(DKN!L`JjFIdrGMH2|5nDG2APF_#ILGl~K0BKW7XNJ}~y+&=nIH#AT1Y
z1@6Q$A;%Yp-rS>(!K@~{+fYY51rQ5Whp}yvoJ*O7`7Q?0)72-fR$2ogD%rS1wf<e;
z6j7Js`WbX~(BC-BZaTDs6lO8Pl@4g*ZzTTuC?tjAd>Iq~l2pHr32``9A^z(-rPsBn
zs<^9zVGhslKlS@N^eR(5T`fT8>HT!3z6jya@?3HT`%t2L;IK4bi_uFaEUMm_G`CAw
zXG{GYCf2(=N)7UX2s=<t;+m(Y(W9)S_zmjY-fq^}2@VUCPtz9RqOZ|JX5O7PGK{^M
zNj{PKDi~}~-;eW*GSLH4f~AV1o6#CXp={KDMfwYASy1E-i0n%<uUx<<<9XB73=ua-
zzNh~7J>y@ZFh3Ugo;D<I*4*1WKLF7!k5()ym@{6EyVUKVL9E&6rNo&0;HuR#e}NhF
zY`aF?V#s~x#E5F*TJ>4>m1c*jVr4`mQpwc2wLxnrO4EqrU|*(QWD|g5Ynk`~xLc8;
z+e%!7`^{WLYO0nT2E|#8$NXD!jM*Bo=NidmzFJn~S<WJ2A_KcX$*q3sK?FV#V@nI2
z6|vzh6a^~yivn_4dhwPP_g-Htd9LsyO~X^r93RWIG9SL_Q7ng_f3a5cwOGC;XhC0o
zLCJky`}-dFvwfGaKI|yb(SF<{@WI=t`~D(zSH2)>(35?ZZEPj?*i0O|#qtkjAL7c<
z>+C<L+Q*C}d$JNPqO-ReJX#yXFn@-xkJvIGhHXw%S1TDDo*o-A?v&c7kgd<H6%LZA
zn^uXSVIv$~5sii^gz4GUyq@js3rq7A_Itb^t(s(fi3BMd^Go9*PIq(+I>cQWO6w0+
z8E*JOj1|p$UC*CZ3`n2-tYWIeKlu!xQ+0kJ`G!g}5f=vvnmulvLi}fAP3rzCA$uB=
zp~$|nkDiz+wmhhAZh)iAbGbsoAeb9@QF5a$awS|WB-FrgqaRny8a>7{VOEDuZt7Mm
zmSl!A1CFsSDte7lW)l$@c-O0^M8DYu4?;sVhb0DfMsik7b@|MW&s?d-I9OX5-pO)0
z^%ZDr7#W8fh1%E>=kklAsQg!Xejfiog$&)lyPthiJ}?Ri4B;5lNWh8FI-jGEGXmP1
z3L5@L-R#&KzLtEveWS_|Nd#7PZSx<G)%Xxfhoz%3r{S^5gt_c=8>M~lCqr@vufB*y
zdbkq9g@0TVHe2>q!QI*DszPX0DssoUsxD!@sggzoi6+m+WL2Vzx+|6?I(>~d`3n?$
z62_RS(hu!?&tYRb-_xUk49pa)X_-<w-=sy<Ad-E3eRzAK`jJQDx>NOfKlX&Dq&Q~C
z4srB5!;gan74(GCO_jD_O%!6X{x?X#(VlU+8Jp8RTUCJ_`3+okqY1~xB;l4rrzrHU
za8m1_h-vcePQa4IlMbC_>Jr{{cB&(ldf81_kefirJkN9s`&3i9r(#RUl*)|0hxygq
z$vVWmv5S*vy&qbr%>MJ^f{V${4KYejK~p$pJHg2EJK9MQ1Nx%C=(!P0O66*4(-I|}
zj37&AMd>n-k3F5y>Rgml&-HrsuHLerN41APkvplF592z6w1ysS(z^oTb<b*9SCe_$
z7o&gJW}VS~zWfM}HI%&yC=(mMiMO^oX@tVnisQi@g$|0utonupLZv&UJ@K?2r|_$?
z>L~?-(W#o+YL8ZvPdd<2^aLBj9l+D1WAS`)<Op~R*YUN;+qI{ANOvqFvtnVY28t}u
zQFcSmiI|ASQuc#_Whvu6kVp<D7bha}^Ybz7&!;8J9w~n3>I^;IN08fmcY`v`@8ec7
z)USpFZN4G#U&oTNN2rx`VI*nv)fph$m~~ZkVbQy%hCo~O$fryCp=`*5o-XEDa44<w
zYu{MN*id+ZRvB3o$sGz$L6V9!rC^8WqGpRIen%%g+>9Bj*vZQ5cRn|SC_tq9!ZlJ#
z<2w4C3mqHdI~p~&;Z?ocgzE*z0u+@KZabK}<{&5kXvyC$^{{*KWyOaIi5+`60;q)d
zg*46nQTzl&wYv*mj`04sqIjW6o}Pr8kut)n*cvWf-3%3}rWnfe_ikq+IHzHpOn|(E
z26mK|dKv)-jX|g7^vh!Ytt=w-<mWlE>n7`S9yI1*T&eR}m+%5|@nsOcSWP`91rERb
zJUK7KlkppJ15j>GNE-<d^V~ikEndjZ0dZ_8zt4Tq3JLH1$9VuJ$K|-TabM0Q;%QCC
z#+up6wy0%olc(k-oPoB{H%6%4+TAoavHP%?I>9j3x&M{V%9MX}BP!Lhs;|zslQmW?
zrB8sh2yQ4CEBtb)sA-jatw_jS^XeUraty&@KN%RU3FNP>5Ys7RVKf2V85RLB+69C3
zxgggOXUqPMJEOC4=&u)J$R7p&`p_)6I`pGBe?hg%Ga!3P*h*8p6w+y{Zc{J}@tk0T
zkGWQnJdFQ-^4=L?^%Iyiw8JQD_LuCP=RSiU#&E?&-=2LTw3Wy;EKF^y%ISFtRa7<s
za(6mOQmwm2vxmD2V^ls|Zrskvsng#MBkrNTn9N?*Q0uZPmV$hDDUn=gFj2IX*U_*1
z?iG9?h=H2%SvB`^1NBJ7?o2xqYZPv5m4Mm3zX4sQ+kzrpY@@i+nki!`oF4L}ei+rg
z0v|I~AjM#8K`#UFyyq*=E~oizH_u7CY~nI~@{Em*wW0i5q$&!wEwVw4;pc;d$SR7i
zp9#DZa3WsihXWa=kuqU6`VR;rDq!Td2FzeGM9WKff$kd+IFeizY=Ltc7_CLLy{pN?
z=gB`>^J1EfH!?3Ml$J>QgalG)V|{YtUH|AFZFZ#oOGKQ7&26s!jddw-1-{^BcW%{9
zr&B-|(9sAfr>Zkr*TvRy2jtpqc(EoV?~l;4fXgD<RFt8`&_sT*@*fDa#S8z~E9{`5
z)Mv&zt{09nfs<v#QHlU!H*U|>ymCD5g6^JN7}P*bWg2@=f|gtk9sJv4@C3`W<08x<
z3Ju7kEPKj+uc8{tF{xrZsP_Xi;Dv|wK|T*>08tp+gh$L?&<<@~UqohIZW}yZpC)3^
zt(@(de$MZ>`n3j|PIa%naTpygr<ceny1y<cwEwDvStZA3_m#K&?0K$988y+>9PCxd
zAA`p>1g6}kc5h49p#8p}j)q$EdmDCe$!;O0lztB<d;9S5YSUhJZraNjVVF#HSGBFq
zX1(+V{A`OAx5@b>O)F@9J^HdCbg<fakQ>z5S#3XB6U1~oichLQT!79^)!EybC`)E3
zv><7}G8q%tp`+!(lQzef2g?Uue$d}hI;oS-MyjBpB=ydo=j_aI4k59|7RZSXP8UY>
zcQ{)LjSpP6^*>mxEhSK;Y@VKXkD$MJg|F{{ZNQO<I>`@kY5JV7m0hb0zosWy!trbh
zB<RLZ`K}QIhA%Vq=V}x!qy(js5!DgBA#kUS-PD;q)g;J{g8QWBVOpIXjwl$9o3*P}
zquOBCIbf5^F5Ctpfi#PynjSF_OM7`NVK^XsbPPMs0<AKh6<;`b@@HuLDKM{|kk5xJ
z$L{1{qvEyHbvoDeOjX8A^mmo^X;I#6M9DZE><k7$fdJ5*)onXL!rH!EEyrmXrIC5$
z*NDPr0%xi5mcExyQ__ty!v2;fnlPKZ^|Jab>DjN8hxhaA?=`0|@A|m7-^rUqTeSFM
zrS%AJny-q}Cm^dGY_~MIz+f|_hJo@5;V>P5z?nwrR8|Eyo*f{R1xt!w#qk#B{rD+W
z(<0*N*eaSsaa(0N#cVs@=I+b)a*c{ksD#rkar`JZ?U5`$c?7ZRRT@^@xklh=kBq=^
zIbk3^yv00th3t!eRxJ1R#=#2|B}U2;UHL{}k<6eMHOworvOEyQSqnq%MPU!}x!dP%
z;}$PO)lqV{qwdBx1zt!gU2ggEgQQk$KT<cmIF!|;YA=EW#K71!yBt9l)1)iGx+K(p
z(@M5!OZ@dSn-H=4jNV{WbzF^$Um-*-0K+%;P%bc7zF%9YZ1kjH(o8~ZEgwrJLEa74
z!B*%nLtVA?RTls`-bwdWCKgLx=gKb>JYLEFgB*tC7jK)|(0?#Tmg(FXZDtp_w|@fb
z_x63Y`2J-9wkegYEs$gNF8I(9Xf2+LlB*iSl{D~+TfgbcK1!e+`@6@Gt-*JR%C*r+
zDH0L&2vRsscw%2>zL8wDDjujLl)Uols1YllaOBqcqWSH;WCM*e+K{CJG>k}Il2UY2
zm7-xHC8tROqXyA8nC5*-9R<uuOR}-22kH`XBtq5!kCZ}#9A<cMEqBiL0}W3L-c=_Z
z`)QqmPBb6YLA^Uyeo)@^WbPp3#23K5;Wde6^Y7ii8nta)HNReSvE8^fRh+Vh;UPjw
z$Qj>fgPG_0j(X!2Evb!%m%e3BrWc#VXf>FS{(y9KO$!SWQ_OXz2ioX$49p~%my}$;
z-$2O9x!LqhFtoKXVwX@&^eBXQ7|)HF&aBI8(c!y&?yK+8eq&>3OM|1P>I5%&ne`)D
z$GLpEm4$(3AG2aPdWj+bW~UZCVi>4+=v7OJ6r>m9X`1AGN@UEy+ANku{lL*`7yr^=
zUFkhr@Ww-jZR-Yn^6(TGn0qOc`*~(Q-8=Y?wt^|lV46~6=^$R~wZwbFx{{=Vgtpp)
z(^}5ND9xY`gerO)Cm?ZmxO-QYByfPJwHke`m7JKZXct<M3OjO_7arm~q#v$}{le36
zpGjt{#s@^YfL3Rnc*3~qUxwf8O)7gZ-TAG<p+in^^|M16VoL%nBI4gpJ1I(xCbBv^
z27)cniq+u8t-p_*Vt+$fJcw%g>pN)8--nx;SwD1<uLbfKC8!D+fitZo$HdjH)aSEx
zs2VeDwkat`JtRoG<-R&0p8Hy${Z3+cc-8XA_F!NPvOLCurH!_-za=afNT`GJMnx8w
zus5|*irx5_CcDw(TN+YbtJT)EvkEEA1L%*z&W`B23cQYo#7{|G(Gi!!mv^V#``&fc
z;Xt10M?<Svp=#t+;`R3aUh>pgM`w_9UiA4+|DOMb@t2EJcI!cVo2PzRQsjcSkSIXb
z%Q_%_d9S$W^38u(ULccHBlOwtG7-Qr=ZCpNf9t#2ZiaKsw!N->{fnU(eel<JAFF>;
zEWzPb4A2mZ@UDfo6Bx`P7{i+tCOWXo{n3bBP2VDRDH7z1O))-sy`Z+Tn<!nwKCd%X
z=e{qi+hzrB6bby$E)9<m9Lmv+N;iLCk<@Ru{J!k(+KvNoH*M$&!PE~I0Ua6Pz^q(%
zW*#|>w@CIfPlN-IG|S0sW^wN@SpN?QydbW^9-Hbzy$5K^cT-L*%yBO2ZKY@ceaY<v
z5vTA$zRQ(D?=o?w<W|hp*)HcN3`->;FIrj2tLv~V3_z`8nk;@;<)K5ChfDn)En&46
z;?<0(cD`Yk)wHEL;|_AnV>kQYavj5IxJ<g4?<mJ3R)#f>YYwz=b~AB`CBtTgmd(}R
zp_}q?(52Ot!}OMdrjTM|i@!|`tnC#2(D<PrU|xAdrql4g@OPM7EK%|}B{r8Qe)(rN
z*2E|4;id-}sIMBO_s;iS#oSQCPQOmarD%<%8FA*8!dq(H%(IeN`PCYRVYDoZ5n*Hj
zmY!HbU=HpHX~oS@7Z@G7q{wOZ24fBj;HiXMUFgaR!!~lnb!6*cqSh3%HuxbFMYV1g
zn8n~5KCMXGIi*T&AERc4F6RV?>ws?(Q%K|#WJ=5oT-{g0V#|A?cn|Zctbm=e`qVXE
zht;!dAbShofy{Z>^sn_Ipcy?Ne35SRQv+w_7V6Bb#l`79g6#e5Co!d!<>lNNDRM(^
zHt7%Rn+Hj6j0cbH%TPOC2eAw_?sIE(P=|@=F6;6LD5tNqe|A-3otBwth{K2mVZNNz
z)7-Ow_cZnV&9CS9tSm$jze{&s+5@Jc1!QDq*m^X^*1=N3P=bP{txw?<rB(sP_JAlL
z7>$j|$B%j$e7za-bfY><5?j2KeiB=jI<%CgdpRH$Tp^u-)*RnckFGG)8>NwS``%D~
zbfSC5!fvej@$NmFQ7<*nF#8Ig?H(*CV-05v3Euwo-LZYl_;y^3yFicx+}3aztN&^t
zu)%C7(w4^dT1H3K=DOg&F^p3kT--2(66DfYIBE`P0$+E#PRDZlauZX(9K)iieWXCC
zr@XmeOHn!1V(j{8{hDqe*>G2hD_ss6hS5M$lO1A}m{L(Jt#gvD{Y-bBER5`D&_xul
z<(7JVImg=Ng6`aT>9D}^$%%ov)_Z|zK0jUJpW|IMHLOBa8jv-*9gpb{^4m&v#(WG$
zpD8DT{`&5&25$0v%-C!$p7m9IOlL!;CS!WzNvZb$qZH}+z0t}6c3{4@w_761E?>St
z2Q-I#)m&67^sL#yj+f%NVD!9^|EXgplWh7P(XObm2GKrP?b+8$<a)#~v&_cA9-#!)
zycvIgemb%aW3E02R<`6i+Pe04DaaRJu(UVTXR&0Qht_XKyKHrf_kSo0y;{YUxz=`>
z*87TZ*7*-t52w)Qp5<XOd&%KBj`_nf{vkb&UzFLcNRDDXC58v4A7$11ww<xFOAs1w
zU09IQe}bSDi=^8fjqktzhIVLkhcBU>l&<(Aa4tEfs4)F-k+O>%?kjX_Gae0_j&>DD
zV6oVEPmIiSwo@h1SBg3Zy#Ke^Zm^rh(Kq@wMVYBSFm^tZw;`{D5H$!J0{^w_L}edy
zj=(~3Hbp)-cY!ayu)>8_cLM3TnD@d@c0T6}(By1xz$K#Y+|6^kS*IAvbJw$D50~3#
z0+YV!O73h>x2Myoc8q#s<;#fKhbt!HBJiEk4ZEU0ANm9iDJIZK%dvaDJQ$&2fz^mv
zVi`&LKv0I1#^|#}xy>_K=zBX%U&Gc!Y27q`8XdyT%E&PqN;Fs(7_QcY`g>5wBG<vQ
zBD{13{5(JZ)?eSrd7W!^V0_O$4B@V84>K=p{FoQrV6I(zl)5Z?>5O1N`SRNuf)hpg
z{q7O@CYqV4QNuZ!v=`56e9d^kuc`E&uYZX{{E0&YD|t&zaqyPFCnlq1yB@aur?e&6
zh88IxD8xx8LL}FzFC>o<=4gFdWo7V$HS7N9qwvd1pBZdN7)~J`_63ZBylc^j37$Fn
zamsw5`qi?jL1c5|NCGaQ2y(wO<y*$<-~&cuO;&|F+uq5RP@f<oW1F!5)(sHF|7wg0
zGFdh<g?pYWH_r-${IPI=v57tQb0@LC>r*A@Ou&N2Rmw^Bi(gG&x_Vq6H=OCw@PQ4Y
z>Ra+x!x;FmXU*l}hA9i1j@G-?a{B{sM}xQx&pvYNg+)@$)@jtJz#wc5wXS;fFpPxA
z$6{Wkon7#qgDmy5*MNZbRm1}7m~B*uYLcf@`KvYdvHLu2Zbqfg>{OX%|72aJ$!fK!
z`ntYp&BkP11R?qwr9gM=^q^zEqochkVQKnm=Mi^CY&E~+UqN4+StmJBM|(Gwdwh|)
zsk31;WQnh)VgctI@Aa>qyi1)Xwb$w@DX)JXhFA1lJzer<GBtr<oouZdb7b1BZ*uZy
z-(^BI8f9r^3VU1qFqGp|1^ogJ<%V-TB$y8a?|KV@X(g*+;h%JzErH=N#m05<2ap-v
znr73Zr1GVGz2%9fG{He$nftJ|ZJ#mF5$NjCTItCOZFXK0GHhKI+bdwj%d&IzECefN
zBz5HH$}WR7CHqGwAy>KsG}9%vSGnq@fB!zsZy$7nhSY&2T~9g>J&@mXxqAKRs@!Xn
z*G<)O@jo-{CMPFv#xPiW#R2Be#)U1H(A3a>4v<diIg|A6jz*<cRh>+E6tTT+7^b#i
z)fd%e?WSw>1uBNsp&JW=`BsGA2<0e%6g%d&FpwadMw(ghB;;-anhEE5>yuL5i8fEk
z0z~#l37D{Ru_LbC(HctmAJ;n_el@++sJ$;gbbbGzE0oHw&hyKwLM9};>POkW49pud
zmhT@OUrB>Kg}52-`1PZkq|a>c@5Yqfe{y;+Pq{j}`Mh{le3RdDD#nGruljtvF)j&p
zfUJp>i91s*+kC5^qTn&K*ODhaJh%bVYr;ZDa#f(;6#b?d^XB>eg%XJG;VQ@_oa!L_
zCKr=efd;gfJSi2io)Tz?iJ0g_SP+j5*L6y;U5hIwFI3w0!7#|u3PyN}R#~x>TUU{v
z-WCaTj{aM65bvcggNd>NEf^Bb<gFsy#8`>~5)~ZgNfsR6*S^GcO1}AMz6)!<U`sGC
z$}tJm*SziE(DzkxkV#%{+I$|j4;`wubIPpR1dNcEJ<d+B(-g!ad+0#lJNEg{@>O&0
zxJCvQEFbDY$Bu%AFw(YWa=s8Ir}9l5_uAKgWaU#>B-=(+EuD;9Y7IQYmFDG!K(64C
z@Pab3l$0uM2dzv_F4i%eJn`6lBcgD4hH)!|2o?n3&t+c-8Y?=Un*8e`)G<f#{$(4A
z{#1J2M3iiKQrVl;JEf>DPfx&)hl=Z=&YQnbZ9-5SoBs;;j+@J7CgJ}p+&d@e?Ht)&
z^M?=`8f^Qb)p4UBN1?PG5Zo=z72CQPmf!q#>Yyyn!eo~!yY5DtpA#~`t0MVgO0?ep
zPWKwh>AT+P&hBsyMDM&gjEa-X^PEK9uS}mE{<&2DpVhzQU5%oo=l;0!+Tor{lmC=w
zsGNVI8~RK_MXPcCgxTJ%@Nw_TtbCdm4-h2=>7O560Ak1zH9RebA!l{^4%cR4d9f~r
z!hp_z5tsKCRVZQ3`iD%0QAfm}N<4G%`To{)iBJB~LrfXF=%%!@0MPaDT+Bwa%_;7M
zYTujLh4P>iwt@FD@35E~H}jPbl|rvR(E4BOeP>WxU6Uq`H~}vPV=^XZFcQf`+vIF=
z&LBV_qCo}>_%)cEGl*bv29q<`<cyJ!00l5P=bTqFTeaWR?rhb}PEFNp)!bj_UR@pO
z_C0-0pQpQ@c77?<6}bX5hiXf9JCjdbh#z(Ai%hum^2^;4kaEDU-tR%(Hlf0i&U7h(
z)*cXLC?pDuT>w+tthY#F%Zy{xk(-55F)4-W6nq5~m9`Ge(}`+^6wFF(qQ9Xxe|HKR
z)d)gHzeT7Q9Pw%$f1-|DM=l_RraLze92)grY}u!6Hv02TS%D*<+a_5Y6;slMhg_G;
zkUe>UL(B6@m&faELmdN{&!mx`r|%)Wll<Zu&X_6Be-QrTHVYE972Okg%%4c>m7nym
zk+Q9*#$S~EqhBON()4F&T)qm#zzBR?097@@bN8(ogb9vjZOv^U9gq5?obLR0_x~-N
z{~kI2AFI{Fy)~4!XYwK$&n{cWI!B|sI3Mk&xTvOC=#-+2JXV;w4wS+omZue8kgbxU
z3Rhx!*6<JZl)IZfon@>65ic)w2s+>X<NE)<EyHu@QL*$q{Vjp+5881zi|hE*)!lcI
z0Z(dUX`;GBE{VV{yXoe?j9xwO`;R^3J!Q{+cZ@qi67;jWU2*D~3~?2nNvzuNee_uW
zelK6?{B+$&`njGEbm}+zyC^BKurrrI%fbSH&E^v$l6-G=$S0wNn<IFIBSyR{`#BU8
z;=jE~;h9-q-6+*sVAdIBf*z=K)Px_km8+z<?Pv?{8s{EOft%8~!@<#X^j@nNBQrUJ
z^sM>cbz3c*h0VQV$R7d#&cY_Vy<Wa#W86$WrLnCIRF%`5WwUJIks3)hOrToHmybob
z{J11$L@`)Sl0``jSER>oI_D@EJ2NW3|DCI)Ixx`p22b_^OW`$bG%af=D@9`TIZGpw
zcQkbnrmcy(q`EIAbb<J26;fRxSv--4cquqMm#fgK6dYQbUXq$x%JwEKEE-?sZoy$&
zVDIo+?#Y+`n3(^s-~J0z+8U18uCyzj<~3_s!zDZK=qCTJ4|zLz#Rk&ZUh)mwU(#xf
zQD}3=d6g^LuFzMi0#gyAqVO4IjFxr+pp9ZbhR{W~we~aHK9B^gS%^kCV$Y39J}z!=
ze2h4;AE+%wR+zzy5e^6!Yiy%G>f=%$?=U{;L~bUx1)LQ=yM_^|wN|Tbs^i8Ni7cad
z&AV~oH!?GrTasJ;skey9W%R<au#~HQLWHR}9fWteTEs=-*8Z=L^B?@+uGPspzJS|G
zK=RZr!K30k*>C^R;r~G{_K50G?Q|-&{~m6~%o?gUeLP9N-8OHcIXzNR7lIRS`mGo*
zp4@eBf4P5iLD8QZr#-M~-L}cSF1!H$e*WIUOQucDb*%WPQ_e+%x9lP%!F%P?Z05RW
zR`Kc#>a9*a*}>Dmyq;>2?cMFq>)%W3T5EQmoYVc%G?p6D*CnP6ff2PHMeVoTEbefb
zU%6l!Z<Gjp<j-&cZl^E!Z$!2+w*)L-{M^n!A65+-iPZuZ#j@3st<w`?oDfcv>3_$p
z)=#~l8l`vC%@ehu&CL@i9WadXGfdRORg2MG`(eExIcOl4WQ5u<Wa%&GGrSl`BH1Z%
zAQZJTw}~vcRLESbv~;uaTb1kPrlUuB?z2Y%%o}uzPI4i{ym^`n{kC-lM6bIcb@0cW
z@0D0&?`;Nj@7S~weRB%`e6Gc=o>?c^pb8VV-Wf~*m}F9Q&Bk%*q$D=xE5;t6Hg{X<
zTQ<wIqdAo|i@O(jTX1fuVptwF(V*p5sX(6m)TmVqwV>bIVJa&LFHaICNHV2ut4#xf
zS2(c&qgKH2{&`{nOFYw^Q}QC9Sa>ZE2(0`H0(~ACSri0;?9d>)`PcM4|MXr7C2zno
zjGetr96QOoRBcWD&Q{AbP*r0eZq9lNw*+Q4YHRxU2H9pu-l_#>J1yNkm=bnFHwb_J
zw-Po6?h3~0+2}tWLzz6k%OJy>Yz1=yc5De6M_m|gkMgluNnWS2h>%UUa<E!b4+^|t
zF;e@(T|$AB_`c!uIpJyB-h2^KUy{WXy+wD;Iw{VPwuvLO4{@k-#cH}{!0P)yi_uN!
zh)XNMw3m^g4;qcz)n7|U)*BU93eA0nIZj@rRv7gCI?kHY<9uRkTty|=+Rld3v$1aG
zrGYau$P3Gap>^7>{QZ<OZKCcR`5lU7I6|6g)`UGmH$PKzsiC*0_Xp64JL072k-j>w
zN7Bd^7cL6q<WtnhPh1SN4H0&eLiW0JauYhnx-U%4Obw&8LKOa?3jebu#dBer+oq15
zx)zI*R)Q|z#C^I$+l9_k>HowT;J`qm<3TU&DkQniFI6q0Ik#o<kNvDA-E0FRglLQZ
zy|4c&=MS<|0+a)9NOzBo?~I&Mkj!vK`yWyE@0Go|(}ZOX_<<_CIWM?+k$Z>dKlliu
zc8gUF?R{W#Ql4QJMARKr``doV9+5zZc0ka&nP(Clty<Ze+y5iX)Im;0>e<W(b7m`>
zPgN~j`VnP*yj2?tYuQI0p9bfNM~=)eJ`{WV`rUp-m4;YM-}N?)`)+1+4j5H#<0#_1
zINMybkzUAW+N!@jD#C-U9MSZ`_UfqVjPvsEmpM7BY51?&uEPS1Ck42EA?%CT36;i&
z`%6m&912?;+GAqeCy{dIwK{Sp2z`gpY(CuUKC`Hhh3X~2EpO8O7B+Ls6%=BmZPdg@
z3{k-o-+*&)(4s)YRo)N4bD5bL&v&MFjN2<6^&uv?+hr{c1@5?EPdTq0<0`<hj=tQ-
z5rpXrgn+J)HhwY0Lrp!648&<gBoyf->EI&jULkEARbAEIQYI@a-=3b6)U17J`cKV-
zS{HBslg0~!6bFqLmPnu6q*1Kyce`~(7d5W0%v?Nm96^hw=fcruz8hxgZ}DJEP{i|a
zn@1n=S6h{QBSO(ZCMEJkzcH+x_Uuwi(2X5EniNhZiJ+$jIhS73d(Fnk&-Bju8Z*1<
zmQAD6r6jkxGAYK&%nJ1KK-aAed=`B|hI+fkl{l~57PeG4gwS!v*o<#_(s>)_HJeQ{
zF2Yv6nXc|f=@zAzmtgt^n4~9Y5b;7&R|o;ELXHAo$eF0B<Y=CEIS>3N$%WwGw=mv2
z)C{Q8?r^zYg)m%Eiu>RBDUt@cZa0k$_=@enCAdq`L2z;Lg`mgy?@4Yo=HCzq5-j?G
zr?ReJv7Y@yDR+hP`_fgk<tGSA$^6Nn0E=95@$+R=T~Aj8n?zuA<Kwhv%@b~nzZwQ5
zrV7tR-n6wBZg;-M_)l`UN%~`xL+v{>yrZVTm)=B@5Ai-Cp>HNentTYIEk{kayB!;{
zrM(9FV?k+A0QX!ai{zp1HPw6Hh<R?n0p7j@3oo1LzON_GH4LWyK?`_))BSCeh9%PR
z<*Bn{gKsg2B`<fU-sDdpO8R)*!{V4vU+Bmta_omqHl2$d{*78C_8Ycs?Vs+O5>d+1
zTudFatxT7R1k)?L3vC3U1hgGKbw>|;Q^bvat1sn73&sJow+aWCRuO)ndrLHZ0Vn8B
zF#&~Uf23E0_DFt^5Z@A@ZwaRRFX*;YZj|R=NpZH;T_+jidF2&(LAL~{Gp|T00%)@s
zjmzi1)kFflXzbwIEJ$!V@7tc&?KY)I<tvo6=on`@O0lxf<H;bUmuWZ`@Y(7!xRsaM
zG#~Wi@bEe@7xSq^k{GG)PxDcZO%ig`SzB>kY*wOuaFU@0-1zEL3G%U`ZGq;TO(htu
z^CsR(&uN7!StpIZJ=6eg`(mgHOj`Ytyk)0|e$$%q&a@&)R`S8s^v&8)h?jT>T&*2)
zOW;J+5xMSP1-kGkYR<Bw(OF&cj^ZQp=BJV<)X9Kq*{|aw;>OEU+|)m8SRQc}PSssH
zCWd4hK)oBBQVb5&+U%3<noK;Xy6gM-;aA+8NGB2E2p1CS?%Is-uS9+CLXiS~4xX+M
ztHTv>*~7!~rFK4D--Cz}$dZCRFqGI!0tww_aBjR>$Hd$c-1kcUGVeQ5;E!9un|C9)
zzJNWc4N;Oxn(=D7hg%`(mNk1#hoKG+wkyEOD!4mcyD&MdI@~?8VY~Ys7|1L2Pti1+
z7apPYqNDXfCzgfiw^34~>%ice#!XrnKSkxp!A!dMs=u^T=~Ruwr?TYV6O3)jnty>`
zrpEkkf1lD6wYX#5WY-JN2#{N=<JiI}qs!VRdHO~BYL9+>DG>RcE&ZZzPj{d}h3}};
zdPT5^wL@Q%{9Zc*?c6FWtk8(Vhh#1QP<uFT1hrLdYgV$lZAJOI7IjrlZ*_WpMWb^~
zdG3l1mxO;w5jCftw&70NU-KedHm>JKI~t?qyFiZNtr>JQTJWCY5sq!kC#62Dd`jPw
z&&b$K353!O7Ef94c&T0z8Bg#2w`hl=?l{L{5lbzkW9`%H==AeiXO8K5+~+76gj(T7
zvM6Tm(x?C@IrgKq*-iDl<VD&;$r6Q!m4;qq?jUP@O?Xy{;T<HkPV7!Z;Oa}<vxUp=
zP;^dV({%$3?3`w#ZS6CcCvrl%Q}ib#WzTl(dtKo`uKd?N%^+I{5?4euKv$k$9f!}?
zCyF9+3SZpBD;=v8SU-%N#P7ga#6HE?3ayqe$vnQ?)m6jo0|9~hxe+|Uw5nCA?rFVb
zdIn#-^ix@v?_-iTVH~e_rfkgy*_+9O2rWK~KqIs}x1%6sP)|3mjpa=~*p%_|kxpgD
zvAR%+pnIW}PfnyBpYZ{xw3dF7i#9p8+GEz;16j=^%_)Ca3M5io%;}{c4;mZD#In7N
zvTW|CvPI_wIhZ9{HqT(+M4(MfxcXU!>~KB(5C?v<s-mB6P6aFUfesG}-5*n4deFP2
zN%r$Fv40t7*uHT4h?`Z*n>qOcX}mKbDF83qR-DAn-6=r|+PD?8igi-|5(xKBqsTXh
z_tg#nXmn>P$VypH<=;ds4Mt%Pe)lAQ1)VkQM`1)BRlF)574mMSdxy+?%gu2i^8HiB
zS*-iWjhcoB6MS|IZA`{EpUTQyYuV_XV$k)&jbF)`h=@(j$<F9q*KC1ztvW}Vls?<I
zdL0IAWEq;PSy}Z$FMFlAJ;sf+cmfXru(wW%7CCSXnq;M^s6^D%$oT0{uA*rE=9Um_
zn-MZA1oQFdn<xn;+=q)&>742O5Mj*Y9$0X=5whj9qvwfIM~oECakm!q=Zh684+Hw$
zTNNj(-r#`MqKUQuc)oZ|2!J(6vG$L0l-T{X5!G+PG}}sH0G8zVBL_{Jnu1y;!C<`*
z1HXcbA|oRci1Pc;5z+O<rz_R2rtzD^p7oij4)VNBybcJJ0GXDUQlq~;CF*2|hF_(B
zzUkU<aNQKc=6%@RKfOS!=B&6!a#hg`SgV#f>s~cckmeFf5_{KV@kVs$cDYGr3YmRw
zwyIcL@&CxVl&3#A05!}Qyi~j(gbyH(S>#PpM&-#iySPUo3p77bCQqG=7vZWvmy4tl
zFV|`(Lc==r#<7ohg54C*`e(l@=1Q8|=adDJ0-%a%G*pRI9IK`F7SXc&)ga-+J^Lxv
z>&v~B^T6|3$dxU1Rj36}xVw(>IE^A_@hqq$uFNx|NZ{cdOPKOi>dH>V8`^?K`uCak
zSj#@2rF;}TZuB98g0Z+5S{d^c279C{^x`ku88muE*^0>1*6GoXa;Bw}Q&K%OEr*+t
zmu!g^EhKbHc(JmAC`b9#pODZHm{@bSr!*HriZ-39Fb%~JX0BQwKQFZH9W_dyBlHMK
zp<78<F#QdcZG@N*7ieVYKcLM9!uU4oRaIc$;3Z?Lufs30h2?rXI%d3pZrizLzjT5m
zSBa9NR-E)2hIA5i<ny(`NZmof1x2{CB={)kC=$L;#UAS!@N;*RWxfA}$%YxaK0sxD
zjbdZb8r7^WU>IYnIw07u&#`~DmKIOXz=bTVsxB`Lztic$Pt@$S4{%!pR&Kg`7d;tk
zsbgi1)H@X-36UoD>)K^dt~LP*j|+YqWs#!*7W0}(hR#rY-waM)_shWuMVovuDV|5H
zB$NGt($$&iX<C*ln3HEm?zv;SF{7d88z4v(g~Z#V&G$!v3Lo<z=TgO1N)yy0``6k?
zA`^W6kY@0Lk4A_t=px2L!_78s64WeLvsfYoLCN;p*rbTmowd_lp}NG5ckvmB4~9Jk
zYlRPUE?p)@0Z|r?dEZBiYp(Ui*rw`P!H~8hZh7<I39%Blc({!~jO9_Ztk#?ig*&;L
zlA>z=!z!VmzGKhbCuskZqliZn+giw0^oWK+y^2CchqcOq1bTuiMn7<Qy2CMUshNV+
zA%2E6d_-%blYP*_Xm;jTZl<Xc0yOr)EBHUl0{nmN_wM*`X93=Q|LF9h;>(I}1mRX<
zzqHRY>oeVqFIZ>N&S^*OfY0!oCldt8um9I?kXCr(2@&Or8<*d8pXkvQt8Zk<V3*0*
zL9A5s5I%Kj%Q!~<j@BRyDY@8cx>lS)ClZBLlDww{f38J9a92opyXR6K%Gye2h^Al;
z-gqGIo&Rh}V{Xn|A<u;>D+7#rq1q`<_YK1JOa2gF3@myr=ogKHxkhfKtR<_(|N0aE
zwuWGti3n$-q#(-aDb1vx|Ad<KK<Qsg!{BEa-Pa~s8<lhsssng8uD-9|-H#lIRDeIt
zfq_GbYMN*lDyEc*`JeF1lS4dy*(%*t^<2a(#OELQWj^j-;+M)h)6Hbg$1VJ7US(mq
zO@+nta64y6W1$T)ti+j!Ok=n(-LZ^DT<>T1K;b|8IcwwMoVT=1rR!BkC3NW)KoY0#
zLT2KJ0oD?ep$F8E69&s3vEi+jpeDpf3pz1`+U<!l#_fp~#kc42AbEKK<S-P!TSTI{
z>%EYk#&|SSNE1&4#+B}v+rkxE8opM_mR?fIR0<9Y3r&RzB45bjcWVC(Xle@x=%{*-
zk<k^hiT5DT*7K_u;V=Ge!W3fmz-*&IV-~VWO%;wkdrzOTsmsZ86{rt*M0mYPCu8e;
z(#MFk8PhL;ZxG|qeGaJ~F91%IOj@DLZrOt7-G&$q@u`H3Mban-jcg5v&)wxq1@1ye
zRP{b{-{UU|0=6ns6pl-@n*Vw%ivjHmkGEpI%rqL+EPLvOUpN4ux=44Mw?hIu>q3Rn
zxjgT8gg~0IVcasIq1Oencw@9kod*6KJ^3!WF|dc(2}U7Ly|Ms+VO?u`4T@;FO!QV+
z->~bcWpoFO71DF$N7vuaYJxigFX90!7J{WO(~0Io__PIPk=UW;B0VLOnHKF#E*^L3
z8-oK2q*}5)Kb63X<@3M%zO#fl!`d<G+ivIHd^pb-5+wBoGKXTK#8xRJR*Q4iPb0f2
zbIGVJH!`1;yEz4-LG)aSOLW6EHt$spG_V8LvSrNB4e;x+fqq(+_w}%riI!A}L@z}v
z)szMygZi>T6-6fDL2}+tWh&zs!7E>-;s}0MUB1T(+!A;`JG>?E(=$JOOZ8ImY<}--
zib&(<QJ_XW39x!hh@G>zyIK+WagMzrjPK;Y*tKjrPD;n7KNt~D_g<fInY959DxhMC
zd^oUsFEneafuH+XNK<4fJ`vDY(yB}T&8jfFM|`1-ZnW@ZucEFoKa{hfNOM9ZU%qC=
zfqBV71pI9J<b&mOk6|;#XM;m)(fpE!O$TW*TjjrNSietV!ZZtHl{KW4nE<cpKm6N<
z(<5i3Y0vh>6mZgsqRLdzeOTC+YuMhX`#ac+49MkvAk=dP=QQrm!xId(>ahS0N%Ywm
zM;ul%9Qw5CI5`Ks&7Pk11~zWsnmfqOQbOuE5D>=~J5Cj0Ep8{*Vse{;=KliRHfVOf
zE?*yvG){KbbdRW&`-qgJYoBPST#?UobhC#FJ?_*Kw7?+uxJNS5!RcK6kw)%pw**y%
zE#TZTO@;FCT!(b;u;53O6SD*a(m^d~)TX`L6Sq7iv7ibrV`U>Hd-mt@;CCetRg`!l
z@)KBJ0;H^3DzJqSP_9N(a?|iaq^GuyraMQeeJKPfKx(3Gq9&vUl+Yzp-52P>YiU|&
zDVxJ09!UXZS*NZrw|QuMS{)NVm#%*Q_Fs`Dxn{iLr2qZ6v`bl&@qL2mHzs<*5MtHT
zm=$j^^YF`wEG5gALo6-b;SOD_WVDI0byTD&0KoR0vjDk3kW@P$=BMv1LF}E{rr@=s
zK+y@ucK`lLo_vQ&3B3?I&@s;%BAv^Bb!VrtWRkqmr(@kQ=l8qgjPbnjSQ5wYurT=b
z&_Y9VR-)0g@dx-4$s*ZYbxE4~p%<Pz0AgxOx{^aIIw8{#RO5Z$UWiw575+Z-q5GY3
zGLgS?_B_`!8x-3c)XSvBczr!NbE<XS2Gi6dP|v-V=G>m^nk<+KQJ=F0lw`A}yn5;1
zy2Gq0JQw+P_*iSKt2BCOREX5PswB;MqlmXaA{j~loMPcQq(A{h^-*y+Malxe_84Z3
zjtjP?I$`IWnI5xLGL4;3@c~ZI#~C_MJ)dvcGJFI%+u==l?&vVdYOgO?+#LJ#$HO0u
zt=GgJW;l#_HYyAW0zvO?B)zj;g?f4q&Av@WxT<lH9+jEO1l1b8%9JWmRt`OX*fWdi
z?_cNLEd&Bom5pXlHK`TB$KL`c$`;>np8WY4oumlkRTIp$Vt#CVK(|&0CtrXk5PAKU
z{Ie#!?6kWsSjL@T=N%;r>KaQ&WpzdO3hhc`i=?<Io;Bj!3=#|TRNci;Zf0>*`~{Wk
zHYgg2d-_6%PZy6HP2vMdCPxZwB#KyrBc2RV8>w(#^3D4g{gF%I3}2z!(JNAQIQy-H
zBuk)+Vo8PAbO*;?*LOnGIUB)SV&l5JyPV9cn)GRJF_gIcb4O=!QX8qTY6JVZFP9vt
zzeIZy8GA$>A+jp(Lj%p?dILaDHCYZnr_}(L9L<R0K9XG@w2<m`$SuFu@qZ7zQ<n~G
zQM-$MtSBG9VN0xSNmT1)%5Z%GV~!kB4Re1PW7b>VWFlB18<^F@ZjnBoo5m&elj`gt
z{S=WG!nDoOEaq5!eMr?)@3;M8D?NDxRRXUQDdXDS{IBq&f!JM-?i9|kuihf`z_Ic{
zj@Jx*!JD&bJKRK2ad>M{W%bz!d~K=Hgi5H^YDNYuzAM1=tcOX)#%b6Rl_>~wF0lKp
znNFcZ;#($G6u#E*z`Nv4Ge#tQ(IIquveYOn#r=C<ia;nw>JsH6ccpI2bDoUB$L|_R
zzj3JsI*U6ueQBuUcfHVT+A;!}=bux}>7Zu#3P+n?gLJ2-)Kwh3iBzjH5C9td)Rz&(
z$Z6tzKSc{s>Xs5a4f;u)<K1rBsp%+d+S#07WCZ>UQ`9L^l8ZLR0;%DSoblK<+Vg!T
zdRZU<g2_(KQB_zK$*6Xur$nM#ak-7DoUOH8KPj?-dKv<_Utvmlo`F!v=!I&P@bFbn
z83(7MNwxCCIjQ(JFY(G8a!Zi=tCqnmP2rE^Ex`c6DEetTu5(9&zvfq65o2@10JwV4
zTW8=zbU9O;qwt%pwl?F^Ksub-xC90@3;K~n6Rr~B8j%Z+{CYJ-G`bdy<DMT!mhUvu
z_zdUke%!9-!Z6mb9IlH$k1k2`VkdoQO32>U{hG!Vj){-0XJw=A4BOY48PomLOOsak
zE}~Q?zbDEe$qMmDt7$Wu48l3Hmr=tZfYk~e(@k(){1NS}>>0$W{3;gYykV=&&KPJ(
z(U*iitdFuaWlY;!o=PBO1>_8dH&P`IN2VAcB6;mS+2p@bQ4vgy-otf-?dbJ$mck>3
zRUDceKbW^W_XYQ5q=J8)Yvyr=u+;@Au8R{L6~lm9kK#clNh1%3zf3~x=ZJA1pGt)k
zrY9tiRSdJ~>ltN;lIsaq&v!Q^unCg=NFyrf58liiX?|jfB&+B)ex4i;l2JB*1aZbW
zS@!(+dco3_O%)cqROiQ*-<t?$h`t2lVW|)!BieUo@LcIvPxU_>VqH{zL}q`eIp^>0
z6L^xQ4|UIz^K#tM@6k{@r4IeYxr&OH>ccP_UCF-X!TmM0&4X2ES#c<!Vgfv|N6eOG
zrC1GC)>Qnk<TT1PK#f95H1!&iF<YGW8%wVj=xiO|hyDb8pk7UME>dvacIxxG^!5Pt
z725Wjr7QQ9uq%E!M+<IFh+*`8yD=YOY`Zn_J|rqq28URp$bV(k_@9^V2>5=yCHQ$L
zJ?wZ(5Trve{+)n;85&CR0+O@0P@Ef8Y4|N^rQD8(`l#Op?Wnh7Tiz0Su*CNaSlnyD
z5nWjd38$@$2#8ZGDTYa=y+FD<K@4@hgP`!;lo+Wcn7UG<|4Xz;Bc1pUhoTZ5krF)2
zR_jX{|FnqSBQ=;}8zdJ~J|if0>@O4g`jE@84ig{su$Dx)-{Z!<NX}61CU5Uz=4M#{
zyj-xE?pN<<CMQ=M)5TKa(yY~2pi)+qxXH=1)*C2iXK`^$pdIm0Yf|=I9p=TNT>bSL
z`2Kafjo{1*N+d7_&z+-IzVkd;k0CoTm8fK<H%u*?;(@;v;#tn196htE3(^d|HT4$h
zKwf{p^9x=uE&Kc%gXRmpH%J2h-}v*t2v>4%3G9nL&s~n8rO#e_UEN!%%{7o}+>kG_
zHUQz11og%})BjY)5cdkIFbvmZs@HRE`J<*7kA&pD%4R01bBNZ_w9;gYEwpI5oh8?@
zdwNbRH+`8KS0bUQ+L2i(xFzTy^tvS&eEmG_7tkTYV9%NBl5N<tA3`hp`kE~dxim5c
zfap(D3@%<k)il<OnQfMBp>O;}NT{nSb{egPrZ9sb{*R(vVy0FYW|D%N<68ntJkCCm
zVAi>8rgv}AuuJA-^2{-ysBoflb>;58ZS1|pij}01n5dJhk;usQm+No57^L8su{hF{
z26=gTGiGMy?K?+<fAlXk+Qd62LNo$!^mh)@7o{&D5(=FNepiM=>NCP=!jORNfGz85
z;*;T90#?^p267zQf2d;QmLQ*^HekdlmJmY%#aQ%;<kd>Lq8l71feMvIK;OS}hh!vw
zXFy+77pP1Jn+!|7AWQgxZ4si^H<AAqr%@OQhS<n|BkVi1ZJEp78(2`uMK>nv8(Mq&
zPeSR_>CF0mC!KHf@=3-SILtO%$|rY}wK};yepAU)8E(SlqgjmhWPy}RHPh-t3v_GM
zsf_b*SwgWg;$o1OsbkT&pNW_)1F&$Z$kWGg8l&EtJaI~hZL_m<tRyxZXv)G_`+QLq
zuoT?AocEK&wrmdi!nD?*VA-hh<EBQ9tuvVcf-x`UG4`H9e)nK>&3gUwD0CY}SVCOG
zE-hX6_Xz$KkedQXYNiSP6+ru3$)5g;qw{M0`w01l{larl`9_Bl^$JKj(DJVrwsC0r
z{`HkP)6OSM6zB}+hU5;<$ol8QITXNQsn5-nE}8xJcKDAqUx)<|Z=(hsDnCyc7o0ZV
zA2VsbDY~0P0mSYF925_&e2%Ca34Iv1DK*{5S-<J?h$ayeT)$?r**lmZr&ejCUF@UR
zRw$%05k4xUwp>p+i7`;6O&u9eP(H5&u)VhB{MOMO%%GAD1om#)(!eGzGJ_q|=LDis
zHXlYWk{y7f=z4c~Ya$u+s!z;<yn}P`j9`i>fMJOV;ll}WU_Y=`HIR>MhSa(2puDBB
z<9x(|O(Wd=X$8YV@Av7IInP!d&ssItEq#fwk`E;ERc=o_O_n_qQv?P3Q+kJgl=nuU
zjZzy=s(MSqKW^-tbEG$iFtM=qKe9y=KHwvfAgS<VHxNl^GHu`JJ+asZ9N~8?J}-=9
zqK6Q){%L^O%$Mrb61h7oY#J~d3P;FbnwMmB?;y?Jh7{4kw2yY@=zhO_nQl0jeZS(5
z{Rv_8mRZ?f1E>0d;8B}RUJE$xZg#;LE;;2*%(80@7RWr}>^4H-Y(I3sX;42avE<{F
ze4kVIgjmPqZ4AV5HPl?nil6mIL>t}O{B|d!kbbLE&w*iF>;{5#^O?2|SLr5$-&ak>
zLDEH#*J!V&qEO9bN$C4njV6_6w*)Jw69p`))pCI}Y|<SxnkH_EGK@1=D*(>3?Mh9w
zriEFusY?Q$)B_o+>ys4)f!ZOH2|FKie*Ctp5>RT>kQ1Y8;f=CMZntfSUlB68F8r`8
zp%S}7Qu3_7*YjabeO8JvxY`Lp-|uY0@>{k{P2I1j0<;i6Ek6#E!QB#k9$mB78*{r<
z@1(~Nr$}aieoqJF1|J)nB$<M7eu*DB;G2`_*3lNp^&0IL`Chd!y%`z%*o~$8w*=m?
zrb#}gO-?gA(k#V0bnj-91Gq(PPp3E7>+@QijtfZ(UmW~ZcN{!|NtjffZ9Xs*0;C9)
zcoW4d@rluL?_c7QwOe<1U|l2pyanogb+x(xJR?SuJ8IWe^A9x@)b{UQ=bR9|dijWW
z8{ITEC@<lJh%*55h7EClU*1MM|Ig)D*C#X4eRa~*Lr|)2GPbGq_wY6q0j#=X8TEQS
zHsS-s^PW(Dj_I<fTKK_LSD(Lhq$Qv{t7~I#z`nc!#4EOn(UXivn;VQ1OJ2dV=FWW9
z<qup()#iIkVIN06rkL%iJcEfNwi`CQGB@Z+3tCEI%(yy^g^(BCab|CSuGK2TsKr|{
zbXE0G^7T;(WptLM@y<2*;^Bwfn2I5&N*9UhMh&Kj4@g&QXV@`~YpM2LKQzzjwy<vY
zhvhmtw}rf9phIA9310MAG=vz-)K7Om+SEFi$`2V%-X3X%z=o<*s?vQ8OWt&ZIKRV4
znj?r5BBX@qbC;<7N~Au4n!FrdmRbEQt~E^ZdI6P9h9A}^$5hmD7sh|^BlH?wsr)4E
zc=O<;Y@=<1N=2m=sva_dwW4~MvT$+QaI!HQoyt`t_-7goP!bxnY0cCuSo<u7$|18U
z3VUn}A&jW20>-Tbh}gnbT6P<Gk#@cjeQclx?{??JwVU}LK3mrEV15@{to5hQ=}sCH
z*2epweN+axFzRZUuW}$G(42I2bZ}{FDzJ7yQ~NHtqN0~hkW2(U7XyrSrq*25D|W~%
zjrUmes)X<PV|SLe`j*R5<4R5y&9sm%Ld8y-=v4pfGyT1TAB0$a4<Eu9XP8TRZ{xF0
zp7Ga_3dTR4g{ZlYz*Hd{_8v{59fl8v%EKakW)DKxOaEp<rOTqd`BihP6*3Q(a~fg~
zFn-^cp$ct7ozI8@H*X|ciMn@mD`YW+89l8@v#?JSN6C4rg|^Rgk43NhzORz8&QJR*
zC`AK2wsb>HEFqJ#MobO5u<-KA6is_dL}d1J7hp)b)3|xVczD(qJP)iI+ys^I&Kv`p
zvq0Au`c%K9tx-=}n%=3Ju$mqtl2g)1Wh>r+^&Ac=XGhYttG1L;!$kh55lcyoXQ^0*
zyTRs9?ToHM4_G(d7S$_(We&aiD+VC_P@^C9j7mZZ35MPPiH68g5;@Nm6Xo9v@!u~1
z-X;8K?%CZlNbN1bp6S{vlJuzubZLLYZx*2SrWG4qJelyeIkw>%Nrf<NF4@OkXTizq
z9oD5@p9W{%`v_h%&4m9zy*~F!s<ED#Qs{7pI0dr{&@B<7Jak1r>}?_2TTLN8hxXKW
z<O=je3$l1sXo%pEB-8`0B0Su*6?Qy7-3>(djjgFhjS_K@&uYkHB$I50Y{vcF#95o~
z?3?xTI@ehn*jK{5WK~0yg}5pwNP3Lp3i_IVxd*crOpoO_cu54?>Bi3((W*p=CYErs
zvLwDctx$>&O*DHzO64nWGQ9t#7;kI+S33ShLC<?>J3I_E+O*ee0x|P??)Ti9<ncCE
zh1&I@${DxpU4x&F(F;(=#?hjw1wG7{NgqFvvN1juO>3{0N(tCnr(6-Lbv2r}8Q&6B
zFI&K7rKr+-W6iR(G)f}hDJgDaB21-gh3Mq==e|EHJTwn%HB=`@z71)|2Z;_jo0WoO
zeU3}&6T%&)d{Bt-^7rKO@y|w+T>>4>q{84k>}*r24_g&#^x&UnZJ?^{B(Gv?09+TJ
zwvJZDYE*)pBp;`)9N9ICu=1qEK9l+WSx1<&*`!Z6+d;2wQSVLuDIK_suG!g4($aNN
zvQ*B9H!T{AEN+IW)y+4plOEe^o)E+)0GNP4CMG_R-L$OSJW%@OD^g_nl|jS3@6AHN
zRM|WH#PYhDufLmUS`Q~AsFx;<ggK2LW~T7~8Y0i7vYM4`29;dn>q=OtZV8+(dofa*
z-wo9<H$Udlu^-ztP4%^=^DDtVON@PEezNf(U$t}-r^E{blaIVg5a)z&mQ*LxZ$VE|
z0bAQ_y_FSJ(MN-bRYfq5s2g*RdMBGs%4D#CWiq?~ABCt)^Nu<(m+8(+5iIx<#JV}}
zOk%r#aoj*Id5cpRd5l6)4nL<f6yBiih#_@t;qXW&a?pweD>rk%O-5dQRnM`cE?J4G
zGnPPPlqLYX57;>Ev~3KoPIw&I`=HA1o&^v^OJyaV_}on@kMD=dQV|t7reDN~hiji{
z-|5>ve1>5x?sWq{avpE)Y|o@`sBEC&T*uFK?s^#B?`@Bn&6U7QL98lTT9#Cg7ah9g
z47pP$c8u}r(9#gS?%BxMH)*dgi;772hTU+w;xv58tYQCHu)9`5<3n|6WdFl~^_mgA
z3ucd!-8CL&L!?3P$1t^4Y8!o8+s~aw3u3a1>d#@G9BxmP2w5dKq+MQq4p_Q2xS?MB
zBbGEG@-oLLpqO$t352fwx^NMb%9!F-4%mG4_rn`pGe2+WaKwXEt$8+;(WXgjXxYxU
z90$8{z-y|Ef1ffEKErwm&oosiFEFRsq<eh#L1r(a&WKaHi#xhLR=pWocUsZ6s?M91
ztE|h-&3@>VN)lHs6gu9tWv7Vt=J41cInuaz(`LOF0rKJBZh!jeQ^|bCB&vJ|vgm&j
z(+usK{#|dZ=k8fHB~WF8s8`c|$%liTtne8&>!#*zlfBRcYIvobAsa;4iaI@Njz&8x
zHYk%U*6Xwr0`pgw7m>P@bqG_z+5Y}!FgUkrrj)7Dp{%Y)BgMJNnqR(q-{64SOCt0P
zN5=o-(9vPV3{(lAb3i5I<HYOgN}fU-SaWO~G)d!PuTC9Ld1t4D${H#)y^uI#Exfm-
zCASsIJC){B%AMN+a;U_!#^WQy(+crSfoxXBF(ImmkO(!MQ6cRpE4&+q4Iev>Boz>l
zT(I(0)%cVhBViI);`i#`B5NY@g?FuM7)G>Sfn$PdUs~d?&_2m8`l@^x4$YNs>)+?U
z;U$gF2z!h$S?Ay{F=9@@kGA>Zz!WnI6P}%>baPwZr8z$Hw52nUT-xYiW!TIilL;6X
z<q?Fh=0J5fYv<xKeZdjaS8_44LY?FV#c#&K1)-`^e?FFR2Z<0le44i!gAgjiDmdB)
z(ggC%<T_PsW`tFabm^Dd(%cn8lW@+)e99@Ky)6aQk=-T@wZHDUVBXF$(-gvqWG6q7
z3w_#=(|?x_WNT{KxRgkic2)7}=fT@H(F#$rUpH3?DVi%{zt%X{jQQoCmeHu>P;`Jj
zF61XWyH?FN-rW(Ugg$kqH}p!LxAm@*eBr(6Btn}c^_f<;)KPy~xiEAoOv+bV^6p?W
zamxZu8y}Vy)xCbWPD)8^C8C@2`%G+KAJd`M@cHXXiuy4%g0lKZK<pz7`e*+1+=SLn
ztZPMT<Fa%%XDucp+~bG4Y+ru;i`C57BOTEnu?#{i7q<4oaGURuam?C>`f!DLmiAy&
zQ@pfInB2(y)NSKnv!Cv`+U;8c&Ot9&o`@lZT%L~$)Z0y(!eJeijQW(MI{f%xRzwk(
zz^6UdaTqLza0BZ{K15hxMH@H^)BMG=Eqo0owv2&26;@`8Z|L;6>0rt#p~MAe!n(U9
zjyv9Q^vb5u7upDio&tM-CTXlk;7XBn)ur|=0rRT#?vF2?xNiY}uqszGIkyDgWbb(V
zTLu&Up7Xu8n7&S=oBXf`Ozjzp3(H!81y^h~M_vO)Ef+>71}AB2WPMWrkIADZb4$*4
z0hhvD$scgGD4fiYOi<o6{TXagI&fz8%SAN23tvpeWE5$hH-?>9!-%%-IxaassrMDv
zR_JdpK;#)B9VbgCo#q<dN3oQB889eh%DGhsDW}tyU@va&j^6j`%Y!;^HXNPJdN&K%
znZP~Px+B5$wCU@E8Ku>^JqD%WC6%yt|4N6{QmzFnQ$lF~i?)(d67`$%NVrA^f-RTN
zWRztg9?Z02av^c<!T;CQvO&-AV0aPDOJ*jH)bPSySc+9xHV3|dq&}I12JQYhgVRn0
z6k@!HGHFzhGWA4gNbBqLOIz<9qRX~213Ilge_y@UEe+Tbk-jEjz9o>+K8g&u(XhiW
zViUInzvz8+{SvB9w_FlQkVZu}-5$3DzYnC3fa{dE1Xo$(H@Li8g0CLcJ=%uHG>Xvd
zwOXaa)<~@hF_)m6b3tu*69orI4vl!eP(8L*+!FeYy9g`%dqs{ljlYC7UdQ~q%f#uQ
z(~u4xANqj!S<Wp#_KyxppTgQUt79rGkn+wCpw&l4Ttjdn?f?a3ol|K^QQmrsnHjw*
zLUm#j6qatMer1Ji8Jpeb@+7*aJMgoUS~%^6euNhra;>mDEyA0Ey^4Vit9rOm!;8&j
z@6?b`k$|j5;8b_!`$ZTcy@y#Y8pIdoHCa?j46|8I<==%JEg8OBN<5ASS&thfjPoGq
zzmjqC8Ho4qPKcCQT#L9~NH}XuiB4?t-x4TbhPnFq%4DHsYZt^DYt==dLtRaIDZWlS
z6_~L7#=w*ZzwYej%uVdssh=MG;jar`zv64RLgUI6iA%AAqZ`L|$2!?5kLP9Ax*A&x
z=t6PTV3J}=`GUoL_X_Eu;@W3E4(?XyuYE$1TCVX)e1d_NU=yYC=kaN;ZF9on^hcYY
z3C>fi4ESySIL!{BKGB<+gP9wn#cpVC2}}z%Tm0p-6NFbvOY+!y%yZKLd%J2v>Qc_U
z&}u;O!iwZ&!wOesH2l~PXx!77oo|P)k_u%$aTnm%^qG@FbFZbrpq#7W+a}@8nv!LO
zX<Mzmg<Eg`4|QXJ&A~Ho2_AMHs}cffd5@qf$00`bL$%s_p94gziod)2*{FVf4yt`(
zo#yC<jqf7QWy2Xi)t0X!c_%OH_nrMIT(Zt(=QBd%3-#3=#n@8a=0o^4%rTm?D79UR
zA*wdWTos%i&1LyhwpoF4X@E@&Wa@!466-w#>c3<C!?Z*l@k{?i1gx<;=W=!LCxNOW
zu@@i)Q-DWH#P`SXw6G5mZm{d<GrQdonj3~!4-cGbPgu7Rr{8}C{CWM;X^q&+`UZ!u
z^my1@b*^@wQ1<7WAzW6rjW_}BgtfVyFKCn6HfvTX+fJL5PsH?L-!J*{DdO1R^E#a0
z2-LhK$MJjeLOL#jDBt_{a4=bdZPkUFVNT2r(PnT;(Y6)mFf0|05y=NES-DFZbF3fe
z=7&YdS-bg|GH28{T|OGQ&*%hIbC25Ifr;Qym0ddcDntVbfcO?AUUYv|UJhq(HTb9{
z-gcu+ET>6Bl`n*oBvGH0QH#e$%+klL?Iz`Wz~!vNO$Ene-d-=$dKxEgN-zJv9bmia
zefD!HjG{|=tP+>UZl>OgNi-#~DJE`UqYRl6&Ek>s!<D<wp|Jevoxw)GyfnU9r1e{`
zEcfLf`=P?mZQD(O6Hkv&<ZdR(M`1BKOCMl9)bna?OQJn9e~lOZilh8^buW9CHBGW_
zu~9A8_iZ;TuhXx_a<zlJv?i?YRRM(W^SN5y2CW@M&$RomZq1{`_$|1NDf0?LHu2}+
zVW-wsR8+pL`>dSjX!(g7a@brBP$94^&iD0Kjg(q}Au+E2NdkoNiT9L&as-SeG3eY3
zuCr8^Lb-hO2vwJYOjxhvW)lp_H9*evdXb-L85&ZXTIuL#+D>vbb3WB%kn;O`3<Ohe
zo6l+AB$Dn(d3_|Kz0^ogd#-HQ=RsSYn)|)M4Ggf~a`TQpp8@Z)WVl7cHrUb<mF@sr
z+p!hI=p$yv$$G^w;e$Rdhogf!tTMtk(OBM17p(LO&nErLBBQC=lW^d>j2nkk^NQAw
z=vnes=f?CExz-9TnG-{2Pr&q#b>U*xVb8M1&aJV_5v=(w{OA$wYAeyiI5LjdhOozi
zRKXoCUD0`L3!ZtY+t;5R!QvQ6(&DDF^f}|_eKp;TRy@;M#!|9N-U69!DVifKibfHA
zcWv95G4eO7qk?C5s>biceq3~Jp~)8k0(b#%tn}lyY=dRNVWTI7Ib(5JN5u(dRVH7B
z_?t~r?vjU4GqdncSEFix4zyqI*DaFuoO<-6=#+^N@}478j~lef7<zc@-aLSPI{Oxr
znRo<g>Mj>cUy^^n{o)}os`YiYW{G^DH`im=$yr*A(qyFD5Zh5@!h?+97NHro`-9e}
zuj!P1I6eI|l!<axTj`w9EMsX@B~Kl8B1)be7PY8E+B=8#!ndSbQslxH8h$JeWGtA%
zgr=B={Z$7Y-+u`7k=fw|CBqXLCCOQu+0%mfU3G2=`Y};>q}|GVzdZE*)vT%B;>U(_
zqLIQslbU+!O?Bz}M@C`h##A#2EKU9_K)MuK&GY9LZ`kS)SEs(Gmp`!xr4*zkgqfov
z<hO|0T7kugvV<jVm??PU<4wG>;<QxU&^J;QB{rqQ$BXl@O4IX(!N{I!7|u{0R=yW8
zTis=nRBMLTSK*x_lv*n{=2XfL<75k?J&#L$*4Pm<9oXbDJIr`XU=d3@<%FMXJ^Agt
z^=QiA;H`qSj>2KsyZCOK$tCXM#N;R6q+|ii8>xc=Q^_T<Flc;%oaeyM(qvEDfM;U}
zh;lX{;=@Ynw>skyJJ#|D7d6MnUG5SC01P{|8=GmdkvGf*P@_^V5kmFPJD&TmcKA<l
zUgtM232zBP48COjBRK7UpP9*CHfmgv0IwQ4&*a`dyCAqLa8}`V&oN{cqax6=Ke_R+
zbE84cI3MPQpt#Ub8u8TNGJIELyf<THt5}flePU-+5csZjOpw_}+VfQAU1bK(nMM~T
zvZQrr_C@aL)1DaP^dpT@pMIEBWN2M4l&ApWfJzkLte5F?blX5gy<>tGB-6X|$Q0!3
zn(+jeVu_L#jw^k?W1@cwP3a5!Mko$>PEl1BN_hUYzbjYE;aF(W9p#tM-8-80xFbo!
zpfSVhY1~pxh^cD-!Vri%E|ZA;P~7*|+{%GhQOOMIdVf<GAOY3O=&u_WZbh8bp;m=@
zk?+jKa6o~9`pMr6j&v>FA979K6<2{A0j+U8ig89qjot1|Nv;jQ$C%K*l)KMw(wxOL
zK9p75#}+<coD@qae3RsM>UnQ)fJ>Je?JtU~6|hk*)yegu4)a)NsH$P|v#iRX7nv)+
z7(vx++4+80oN$@4Nfv4fE*?^z220ghxKlwWof|c4;8Y-G_b02|W&`>9fb{+Xv@W&>
z{Pg80>h0rB0)ji93+Bl9&fbXYQUdeu__y|oWK`~ELjAn+G_x*X&tm(GkYLuf@#NU+
z-OW=EhO0+LXZJQY2<{n+q*2IjXYVwS%TFGaz?`NOtJ5?#Gvk<54w4Na(sR7M1*}_X
z1ZRzVQv(U1`xMiBdX0}0t;G>&<Yjb;DS1wdpZhN-TCGXHps%u?AJ4Bf6w<hBQ%RaB
zNg}G~CD3f&_d^!QVt-}Q4+%ngZ7IW9<zlggL_WU6RT^`a{DArfaPSRaN#uhv&~|Op
z)+VVTKh{sV>BS9Vmw5Nc5;1qG@a1!1)A$WWgLL<37)yB7*Kz0M&ive^`6b>^lxeTQ
zyZua92zy1mvTMpbq}Lym&mGdR+&Q%=ys1S7Bv{-MwA9^OHeYEJR#>-MaN#%IFOB|#
zfNWZBZH$Om|4^M$K$hz_xzSmd#2m9YkHR=bv+^))1w7d5$)Xmv#C+IW>oLX*b1Y8e
zJ~1T?dsG6SfBt&Kj<I3WKyh+j#;82TLa0a_?}7m3C0M`YMO-1vgL^&1=>~tCESI=v
zTaq-;QF}+ICHdUNnEh#-N}N2_zqrQl{(MXU+O%0NDs?%>!y~*K`Oj4f%u~Ro35j~V
zB`SdI{QXf*`_R!p+dTw?;`HFwlbA;1n~&T0grcwFDsNreSF}2v5Ai<d73W3iGX_su
z!l_sH3A~Fwi9H4e%5P|xWc4$R#n3zqkX=TLn6yQSVaK2`iru*uWr>(%BCKJr)j}te
zn3ihCY0Qe9BxHhZriXBJ@k=|(QKpYOB!A?AHW`%NEj5!j?7f!?U)hrZYl8`}n5mAI
z$V-y;<7m~AQJ&VcX$QY%kY?gWsCeO_KGxuP*pm8XZmTEy#?-D5-P&Snr$;4UW<3sN
zGyU5^op8<PK@ct9=u_1UDCbyZO|V7pF!%7d@)MJQSBIGwnARt);lGz!0}1ca*)D!s
zsdT=lCBn<DD%eOcUbffPD%D<GhaA<PbXrm4;YnU7t%gOW$E6k+zq6BHx~gJs%PPz_
z{fUqCpOOC|`ytLuD&u|wVV7~p!<VHkA`a!=C+#b+1VjbD9lqY-MT=vRirv?ZXw!tg
zqOQHG0OH=%p!4}3gCrWIEp-+4sWz1qi#x3>5)1-tY8RhQKY{WS3hF6kgQ!x})Uv89
z`WE8N$dt*G!{kka92KM=6YQ8r(2p?g6U?u-&{dLiz-!u8)p>Z2>euPC{kqF3xko@~
zlXDf1czm2p+V4pOcEG{jX#Y~>Q-n}8=Sl`iG6?~0U2cA}*V+Z0QryTMxSpknuaLS8
zO~6+glFje(<b9Om-p{VRE|XOMiS2uLH!xx`eo8d1nu^YcMvznS+2c|};zBq7h1vx4
zqGD<}<}*!=oG|&wN=9P2%8M}|$Js0?^&p42VoBYUI;<$h>LOE_32v=1=)EL7gnf1S
z-hL%<KAkK--7?q6(<uEjq2k9&VuS=&g(R`7fC&vlRiC{zuUbxB)241xT>N2v5Nw<*
zy0}H<H3K;xXFAj}|4O;>un46#ER(vkC-N7Tovol}2&Q0JnkLsEV!vx7TYtVAGhS9F
z69QMM7*VOH!1MWj@SJLh5twMiXI$e8gFOv*dps#3u<ZnQx~IcUQ!!y81!rM9bh4^_
z<yBlHR6NL@T@?3bqa>(d#xE^)Nq{w&MR1ejsf&)p5ChTZUz1?SD^!+XYH%z#Ss@xv
z%8!#qQmef{Ts7PhP-oiNsS$6)$MB%QVJE*?&~VI8+l><&vUo>!;do&9>07bXKf~P`
zilXumsIIFC!$$$3L)F7=)f&|KMW)9+ik@n$XTBw61Ha8&*2i-_IoWMhpDTHx#4>M+
zI9PNFExT5-5xXjLYuH%7#CX*UHIM$%gZB9~se2uxYOr^OXNA^{bO8v}5Y5`isKoQB
z5NmW`q3(&do=CEq5>%3hv!Pye#YL(?Qtu5J5cI%LQ1tKdot3s;uM^jYTiYj%%|w4;
zO;u&3nb<@!^ylp5a|~XEsaQR5G9&?0+~cy_G|(xFyd`)UywEIVOT{g0L~J&?J-^q3
z%S`eWHu}4PZVIx=2kw|JRXJg_p<M3y_^V#51W~<7^W<3(MQ5eYQ~006;Ss#snY|(G
zQBz*^S_gwY@83IO!_Zb1bn<EyHzRz>XZ3r!ZSjn$oRha(sUz@N#EXiQly?OnkE#(a
z(PmC`t>RM3oaaTpjGs%Jvu6ZJSah8KAl>Wh%e|`kWebw(6za_DSy$BjL}hDsm*ZxI
zkbHe4^bVW@N{RraRVYr1H#OmSac!((aJ#96*Gonq0M}%V+%N0p4;$%V_c?nha~c}-
z%3<4Y`K$c(VBco(&8rhL#*~el0!O}DthpWUpEDxfK51=7W}047DQ0z6Yyxj@vA`8G
zUBJu!d8RI)Q?r_iL7s(JJ8B&8cXzin<{|E6{#aY*c-R%)m`9^C6(!%dRq5hTSEgn#
z`Al5JJ)(`z6GRl+8y8PjI?SX?CiLcEkxhLqPIM2ydUe-O$J55RK4%+`bRki%R(kIz
zKsggvJ`jrK+VpT>v73P#=LfxN<1&hm-<GnFikStgoP?jY-q7GJx~rEykM4y?10?VA
zbO6!DOfpJt=yvj{!96{H3f4D)tp<(S*ZW=7sr@$x2VeYb2tLz8C4xknO}mBBTEP>Q
zMs-1qEK16MWn5gvC+0`QYsEAD^RZ-_`9*3^^_D={@rqg+`HJD641oVi$G^uB<o;`=
zAR;Sx+oY`}JL_d@)WjETMUAWBbU`xwM{4u`VD2q|+KT#hZ|bEjP@q8Z;_d`1(iRQw
zF2x%lL4#AOxD<zA#R3F^hhT*kw_qVif)sa(mqKZ8-uKLV?svX(znOF9yLaYhCbBch
zOtSadYp?ZR&+~isGkPYGsyMJv-%(DHSv_5``9|Do792pnwsY^PL=b-|1g?Dou<nLn
z9Rys321HufHk9)r6jtKhXp^>Fl`+g$y_Qi)Mo?}2Tq4s%>(i{*Ue`5Ki*9+Jd+_I=
zRY_=_MZDD1z67ZUlDoUw!7_u)$2?TZH?gr_lE=l%f2CQda>)e7m#k2x=a^FA<v)T*
ztpo|ArHri_RE*WTaTUwzlSMMcX=uY2{*ttngjqg#XAs!v87@Em`jO1-pbs-jMrGNs
z$WqIf^S=WyvBCF#_g!xtJl486Z7k+1|1s%k#A}&8p8bh)H4c-Y)83^kMN4r{m;6#+
zsC28UcArD@@QViy?9s=*O;bF4m?2W}%B@bNSt_ADkEuvmBMcu+={l<iMGYk>T5TDP
zNDc{z#<TEQc~>vpYRj#GjJm3GT<b!SuW$JWJ_@QgiJ!JxO4&3^TB+*e`Q+XuVeQTf
zf|t!%ZAM%{2Un_|y54#aaT-+c<-4<qmH8O65cWPjAPy4uc}?1LN++m-?;K?OwYI7_
zVb@u*lnp<Mikn*oKz*{oa11lG!DC4Ks0_bHq~_W22S->*;bX-bH<NunGv_k;5s-C&
z#M&zD=^@vw*o$U<;DljU$^fF@n+gw8<_kOhBGk9w^&~^95eFye9V_E4iWz{kVg7gS
zbEk!C87NwAF!TG6Ne^;&m=K3e^Y^P`{*q`Hz+Wr`!PB)*5#YuFHrVg6*K0N-Kf}<N
z#8rG;upJto`PkI6@|UC!Dw$v^yLOf9dwg{7!oNhfzOqXcc7BFJjTa1rZCO0{nIIWy
z^-d?ETY2a==79~T&=XT&(L{xA_6XW%i@~>Id-9G;fLNdD^UbCjss_l(WQk&Qyxcob
z*Jk5~#?;U6>3`)4`KXeAM~ByYC>uDtseM>Ubh6AYL?6xMGDqWN&U|m&33!5tb?Uaj
z85~=`UVPk~WW+6KQumuW$bbWA-RUK?|A-YG+pcu<x-y*1`+!RMi!kZw)3!x&*4An0
z`XXsnm8aenEs@=(7_3(lqL<nQ-H2W73Axn<t+<=S!R{!|hYY4BH=Z)qOok^@H>i%Y
zvr8zd0QV#$R#k@@Mw7CB>GfX!{NL&zB&)5ZdC)e1e>#7_q2ty!*vgbsgvxxuM&D|!
z5KZYq6720J@iu#_q+>;TF9$aNNiLFfTH&`;EOQD&fBzBNI;4rJ+M~EHuRb5lwS%s-
zn37uPo5QcC`JVxN-Ptinb*+h`T49@7AMW`bzXjNy8MUgnb{~1lQ(y2G3GPZAEw5Lw
z*EW(^8}nIw<82mINgT7WiAhdeY=8r7>k8olS}yxREw<6T`!8~l`oG9UcFvPry?M@4
zlf^RFC___1l;Nia-qaLvaj}Gbt7exq|ElqSl8fet0{)#`L^E!zN~AUzND_ncXozeE
zZI!NOcp1?I<iZaxBW4q2MaL965*a;vVqy#2d018orAht&(gOUo;SE@Ln%BsLq?#m4
zl8N>dj{8UQWGbVFL!+MA$(WsC$K@7J(Zm(#%79gr3f-BIm!1)+IO=F1gRSLy6+!>%
z5=o`6Yk{9G3@e;_VO)h=(+fKeLh(*T$^k7-qQ2T(9xSq4tI=WpSc#91TBlEA+AT!W
zK<e+ST2>OWoFa4Fl>5L^51e|Rix-;lJX$L2Ctj(<@2(S+V59`1wr)q8D5Z>fwO+UK
zOSs&r_An;a<3lY~S#=s)G!-@b*8*E$)+7~2Sz?)e&q=&|O8;;_#|}m}TmzqRZoPA*
z@kBgc8U`ke5KFCVG{S__<V%i%-tuxR19XH>%ShM;>=6i&-G~biE@|joa4mj+Tn=G$
zyK*y`u`jy}8E0Ph&RR^r(H8HE4GH5RyoN2TOiLzW-U_#WyoeLV#p2}!9vq)%)(jTU
z&DM2gNvV=kpE`{Lm1CoJk{lZu!Rv7S^b-lt<_jRN8p4%VQ>!L9b?J09nf-{8lCI$f
z<0hmusAvWzRzm4l{mk#rsP+nIouxN_X0VB%?Bi;W)c>j`Ep)br66Y#pNixh#Xv%(E
zyPiDh%oUNx_4%DkpxauGK^-W##adl*h4HIHS9}CZN(F0fi3ItG^8RW`CfqE<x|-W6
zte;@MRKL-ztl3&HvMfDZobvpxEpavOF9}NNNYVGt<=s0rGiHCtx;rp1)^p=$bpPKK
zm;Z2@-u1$PyEbqAgJ1=4Y2G|d_t-c32f^y=<Q%_?;o7P@kzi#co!HmvGy5-sm2&#O
z5v<0F(=`VVTzeldcbM*C4vM-f6)K2!9rJ2mXAhh4MD*5#$ZQe&NV#c!fL4)iscx&o
zTI*STDYsaaMYHOlFowp+s0iI6g`4ZgA|XQL>jl!=y>X1e?YDwbQgn<|%@1RBRTfWA
zrge->+;~rujCCaR#6;IQ;wk=!kA)w*$`ISs%F3+g7;%-)jfJ?iJ3ySeAKA7|c>?V&
zIZ#ULmeqfh$ViK3BkG8MSxUrXWZ&Y;^8_n<`ORYhlW+P0yOxbZVM^p=o^AApz;6@O
zZNDl6^m0i9lF9}aqT@qOKA|>OeC_S>ff}={s~?CY&O+;XUmdwLStE#o=Pw>xrrx!9
z7qE9X7<A=sQst1D9j4z#tW{0r+D6SBBmUIoMK$Z@4$owVj_U0^ZXTf+FD%ddMf{zC
z6>Y;NU6l&=O`9t85Tjn;@5pUB_WioAW-k%@>)V9w&?CN&a|5kge!ma-U76t0ale0v
zI~FP*jjn(d?_oQjG6z)aKbvcU@(Xn@imvA>Zz<Mw*j%)Dt&8)IH*H?oT`a@~_m!_~
z9ZFxUb%a<3sBF>l1Yk>yD#Ld)CR|QtH<#tY0xJJhiHBbKIqI0t+wDnUyVV>X$9uo|
z4~>Jkp(Nc=^7Ic`_;x0_+L<dNByj>`fU_cN0@;N<CCK;`tOEY{SlQ4xYX#ByNk~*6
zvPeC$Q>X~c(xwd+rQ@8(dry)dL5N2Rh`fo+PvV?ZDLT28>CZXboPYYW+j{tB<bI)Z
zNy0G7Fl<%n>HY?Yez>V>durMXule@Y&fPGh7^C=~lBl*mQxxB;O+Vuj-ng%zfLnGm
z^W{n%VI2n)B-g~A$j%H9h0{W292-NTeHy>$6xA%v9>p~WX<Smw?w)mlnlOL*LM487
z5eu~xpKb(6SzJ<`ZC#GN>^OM*Qp)enQ=)ed0t^2&H~ky_?ndFu$09xq^a&6F4e@#5
zegc@w_Z9?e4x>Vs`#g>|!qL`F7An3^Dxs1KZDq2DG}IZn8*0nC#q6&hQI{_fMqbcS
zm~qxYA&iSZgYF8)j%azMjgIvg`{@h@QIR<_Lu528$A*C96_Pw^hc?$HDwQ;A&PZtj
zdmOcaneef>?23%o>gM)*+!;R@^=>i!Cx>K%F8fm(T6)mkyRIuiS0c~;lKgDLURmnX
zVQfLYj}Gto*Iw~_g;nd&IJj}udzO%6OoCsE>gRHMNN~Jbuyz1xYVC1_aflm|+PsnM
z(5dS^&{>*vk=Ll*(}>LnDAv8^^A6TG`7i-=cwuA&m1?<9E@q!Fsq<=@93tx~S9pI*
zI=?E&n`~XVq;f~P<ZF#~zEWdqCrB(OxQi``2dS=q^2%$PuMoU!W?DQb4sNvO@_PjN
zJYb!ec~@)+8Z)!=dik_IfF2gS!9HVYx>7YLnXk+?WLhqzE9oY|y-?Ph00<67F_2Z1
z_BS(V+}Z0<rl=psK*w$HOBzy6@ZJ{q)n*JQtjfEzwFL=*kpemD;l>x!t51ivIl7${
zI5?AEcuNvw$|l?HP>gQoLv(a)5^;+!pe0n<QoHO@7oXJB2{@UO#@C9VJ$6wBXt%Vx
zAVg24(x&(W{*>dIi>+!zdn%ou4mZ8D;3gSpeL1^6ZuXQcNhD_~bMwiRlhZ@HB2hqo
z6nW{U|A+N@_{IHkhC4-TQg^0HpfgC5LP~dg2MAK7EpN7X(%~VKxKXYfho<-YM3!%z
z<Vdj#hB(WFB8I;2>J>!=$Vek28=9MQfhGr!RbJJ?o(Dd}3Kej=5RIP<sV+qlqU(gH
z<6p%Pc7|{sU3Lj2{m1de)%K!J{P!MbrAfci`W$z(yBKeso?p+4L>d)~Jxnf=rI4Bi
zg2+lwe0vjrvUe*|jkB(OC6y-tB3khvnn#!c!xw$0&I8lsZ57r2Z#OGnK$ac&Xd}h4
z@5{Y6!jVR84@4)(Xj-Iff&aj$C<odg^cGS@ZmJP2d($Exqc<AGtx%}G%90^?EdSaE
zO8%*sKR5JRt?UIqCA3I3hQv>6!F`IjAiiX#k!>SSo(eaW70kgYUDKEL6^Es`){=?_
zRa3wxlgEAx8PA&2>X9L7MNrXoomw9Or^w=IDP8(99wnXASjuVSLq}t;j9eRTU>v@6
zZXzm0ye>pP-!05|GGj6|Q=C+X4|ox2Fc!r0X<6ePy$pZ&%3kkZlH^*E^C^`ZKIhdt
z8Nqu^8lC*>)bzsDdM8p7mR@6*=97w5kQavxvgM?o5>}DocB&M$!d5CSAj4-)iYym*
zY>Hded&Gt#IQ<s3Kl-(8$Rr%;M`|DQg43L})92P#AX5${2G@<De@TFKDyLNX5I#=I
zFZK6>tkf3Iiov~Gr>flm^3E_*rKz$cN+V0@AEE;u^koEPUw7*){g3q2$;CgZRSDz3
z4^2Oc&YtWu{;cZ{pW(du+T*=vrtMmjFNgLqEwC^<W5claatfxEa+vcrHXxy8nKY|J
zeX@0+IVreXT8?TVYAe~Ai&GTdcX}x`8-cd-*zp$Ouck|TU6zSyKwHZpp0?h(yCdWV
zZ`aVujBuEk2}3`~=B;N@PGpxX9*w#FYtA6KsRg_T6gM<pe*iL|r>AEIM<*p^*dGf|
zi0_J@cLx2xyZ!&Ut!+E6bRa)^M5HM#(U3@Nl8}2tV?H8rL50i@v6aZOGxOWs{zx9{
zCx*ZmtI8op0A1Wg8|09f9b6dTtS#d^Mu58o7418J9&4O%VrC{8XW_BjDa&@A;od_1
z09L-7+*Zukf+pxhSy)Vd{qmN4Hul~Orj@^NJd5UxjM~sfvuM!1H&v5EGAoH>f>i?d
zW$&vu$<oio4WEn?3zkAu6+_ZHv=Va<uh(F=W9Ga+7dL_*<lV0?V7MbZrz_0~mz6)1
zNrq~2WBYkT@5nMe-<&0wF;vk!ZSJ;k_;NKkifIH^#`nEg$ul9p_|hr_$a)2*9LsDA
z`-L858;B?=8o^7AHPn+TkZjJ8kldP%4mZ{OOF}DhR(9?68{%$yOCfn`;q3jD5K4oi
zIYtU{@_=(kueWd<!<8oYe>}?%D37H_+XW!(i59ERAN?3W5P{@}MMRRwaL4ywNF}ED
z6m!Bdy{bv5Kbmizde{5%mFUb)V`8-+U|KVU63(z=2ZdX0HRWYju{9i9PN=b3kCCv{
zt>@LVl{S8E{eX!Fta)sQxc%Y-sOMX_(##kwDJDz8veIaNnnO835-%oIoLtmyEU)xv
zeWcw8G~bty{Dz{qJ+v^M<~-#i$>a6V4c_{0n}v2SF5?h*ci?-;#i0BJv%B}~pjY#8
z(c%Q5FIf43PGOo>&?_o~&rHF;$JY01F5)x5t+GRaM19EilKr7XuH3k_X_w8C<7h2Q
zTDh5J*1arPs7cuY!N9_8L7BFQt<tb$b*lNqO3K9ls|pcmF3gnuZbo-``m-Fu6d#WO
zVRPjsZ8TbYwjoFNG<UH^-s?V>m9~o7sXBs%rjn;IZs58XP%;~9^?XGEXR^&5FU1;b
zayIR%CCaWTvJXH}mmY#}emyr+FLNxvM1RyFU+GHw2l=57;Z@2}nkg&5K$6hv^Fh#`
zAf?zll%N`~oE8sRtKRA9<)P^Kxq*1DNGR1QxPwJmG#xuk44Gu>ohlaM)&yp;in7`y
zupBPN)_@axQip1b`|!8%j{OOj?#^edu4<z2dFS;U&L%8UCXM)=ROEdi(Pq!_gq7~g
z^wvm$Qj5)C+X!h^+Ys^Wsf0A_n13O}%VA;BNk}xK_d^|FHjn||eIU#@g^P1I?oHS*
zXedq9fYp_MhSPa<ji&1LcXDtjE}{QQ%34q&SErBvX@`+NIE>XR^SE|qYcyQXUezbk
zvixE{CO#(fTykxjIT^@VS`7N)AO`~Srn2%@@{kRn=p?%RF7KD8g=cOADUdTpk3R<r
zgQv7o@ZH6Q<x%g&xisq<a%q|TvXc0wKdCvX58s)~aI~`S^*5-$Y6k03@4LE7KGhV{
zW(TLL#i#o)!8vad-ETO+m}dcp7{<kk-nes{AiUcYp7M%=^C9>p33<@($$6XH(VNIg
zGh+G^swSM5P+ot#jx!p|qGc5`@zX}Cum?On$Bln6?=|+GQ|hCTb#jV0MVfHJ>eEuf
zNVow4PibEjEL-02z*4nn^kl3)R7A8aKEQ$>wiL4)@%Et*8wzA3=j?Gdll(#RJ->=;
zv`0(BV@RD$+F=b^vru(ml!{N&&b|<Bg(lD98vQorf9r_z^C&v00Hip2_kmv4a;P!7
zIoS~fi%sw*nQyeXo9{{!nS<Nsv8nzHH%#_PQCWD=Ya~alC-Avh{Voh_9qpOVzV$GG
zZrvqx)TPE_+Te>)$ZD^{k|mCjp}Eihs*qx&npo1qbK9G)sp*wl><$Liy9-D##+int
z^E{7|5O+e{d-^*@&2UaIMumS+bT%9i&>3Paef61vr~2|`_b!yxys=zFK02&;X7zej
z=C)$`wt4}Ow$A(UmRIq@Dn8shvb(OT>wDc=!gIZ3=X+(^L;W<yjZ2*yWQt-|2j7>{
zwiCn(L=Wlxe&LF_0Rk_e$M*)1hwyrB-|EU(f8)tnmOI%FK_#JP#aO-L)+b1}2|0V!
z0R(s&sAlrE<7r>F%-&%wVXEx#-dy8-9wq_{PK^E+<OWe6gk~^{!1rc(E1P6%jD<<U
z=GUFeXlbCj8UsGLD#JoZHh^lpm1CMkW%TKxm}G>_@e}&1h{!|p#x99rSAoLFCL}TG
zj_5t!eFb*k9?vmwdMsQ$jkcF^QY#XT4(x@k(;NW%7_GB|P-B^Z4z33+tOf_HPn0CT
z)F}b&X!d+`ZmH+O?3*1OeJ?yG#c<C+a@mfrg5z~q5kTem4XH}p2I)s5#)1uqv<TCB
zm!M`mmXzg~A)Eh4PrHuMf})<$yiCIQ$)RzRQ~86}g<o8!`q7Uup_vf56c>Y)3s<|c
z_CU#TzKgA9VKIQNy0jy7>15$rmozNyI~y|l^Q6<^!&yaA4nx$fxfl5ibFZTvYVfu}
z>F9`O!Q|2vfeHbZS@g)?9+m)8>f|MyMgUiaO~P`$cT7bye;fGCT7@$cG_d@jqzmjn
zP?}}@Dw9Rp)n!$Pa?Unet6fOAxiN{qbFI>atmu5oH4>mW74kDh?0tipOhq%*zWlQG
zrByU7TVhQv*>djso1bqz%PUu8M_RUp1A^3*bBWVV#eJ~?%clt{TB18It({EL3=Y~T
z9`*1|?)E?a$SAUk2D<gysFLfX&g%GdD<WFSmZ}S%axVRh^tcvtfl>WNp?8$yXo(cu
zroI;M+K-ZBI)l2MT{ILAg&}^In_{<i5UaE1O2Ek8;$EyHS4;$p=r5SJ_V=glRgZ$V
ze9K$ZW480k!KIV~-#0E4TEP5>#w^cL4^zvuUOCImO0DXqiQhZ7^QrDBJO;Ye)1dP(
z&I<rv7C-c6xw8EstJM9TOFC?Sph8Z3ovZX%g;kY;$<T=dPA-~n%EF4TA)A$Ev=%HA
zJUl$MQ}H7(DxqzVb{CMmsP(Pv^2JhG_hQ2SZWUk@(04rhc%gDzQfsf%lUlzvYNq{`
z*0r@Rhix?;`u=mAlxvt)VqdgL%oaRb#o8_1DA7#Eow|Fh!tLN<Y11Xqo)A24!#<`X
z?_eByUsZDG)<mdK)_&d9J8!T;FaMz&M{N>|yc*|BRUKR+Weq?p$2n{?nj&<Xdz~V3
zJsH$nfNq)>hX%5^L?j_I9b>%%zbsoRgGMThOUhkuCZ}?7x&i2kJ+q5z|D73$xc$y5
znG~<B>AlKjT9ed^tgss_n1=L`)nr)lwr<X<$*jMUo+Fn#(Dd1`B{KiLZmVKLnZ475
zu&(6fHyNKBt1CqFZATSpD>gFMWI4-a9?b{WXy~hZz-Z$2OVM|&>En_VW<)x93}gH%
zDDPS-fPr&{?h&n9EZ0oxFH`2Ep1SIi);MbH(;%ks-&h)7BJ=1#@Q$WG;pM6Em2H2+
z$~}x^hUY|~nXZeoajq$j3ClJ)hvrCRz>}Jrg5Z?iPFHj1@%X172p3GMC8E#oP;AMr
zV`P%S%EqTwy1v^CUpgRN+PeUQl%9QE6EYUAG8C>QLJi#MosnNw?y{wZXWA~7eT?Tx
zj+C3AfLFs_SLX+Poer)kXVYrR62qW`IW?O`%{5a$K&GFsBE5^cl}FlB)xn<SI&3QG
z`e^L|QouuRiFHlUw%)`zNd9O`_mO23-x$UpYy{>-3M_N=i3JXTQ5qRE=|G4;Eh!*5
zH7uG#l-1;b_1@mdR&RkqvqERv$epYfY|Bc8u;dQ$a1p#w8|ZE2c%-O~z>5i|;&N_~
zp>|wc|JN|A#h=vabIEYTGo@$$b5#6)#Vgv^Hdp)<=jt6ek2?}`=1NX**~>9U*3Qb+
z+bdr2f4VHQFV`~;OrvhrzFCXJ5J#;quPR<&&@Uc9qVA|YeRY+<7=E?dP<l@K;zoM4
z#O0*}({rVPGor}UHjpgp*dV{wmsq*!_?U6PJ$47Vd{q#nhV*(#&;>=|l)xe{9#06V
z55VHoRt~*LEj^(zPU*b%@{=nj{?bi*IX~|<s~|qwi;Za4EJtxBeo#B|^R@MVUVzfC
zk8^32oj@w}H4vn`EBALk&l?SOJ&{(v?XWZu7HRX=)x?fGG)TL^qqNF9gI2ep@Y-G5
zS0XmG@w1Topb(L<`;RSiVJlfWb2$ZZx>j@#$1~j?w7IkPdb?~W8;y?)wzMPgy{pOF
zSs!Vqt`n)kQokhWq+z^cJM&@_WKpiE&_vC$v5Vk&o(!h}z8OFr@788+F~sr`NH+31
zT1Va68H1B5&XC}E!py$c5rDse3<>-S&1S|iaVttZB`vLKHObS_vE!l3W~lu@z^#g{
zX@drvaXzru@^GgDR@3qFA}cq|R3?N2!%ukjr(-os*K#>vpoK71B}`5F`c-54q?U#G
ze)gKE>rj=nEZ@cH02f>^PbG`26#3lmQOdQ@!;!BQC)6>wap>?e7}O?b^R|^8e5}4n
z+V6CF=1dg|pH(Q5ea!_m1F@L6U(LSJ>|IrH_?E$#H|R>uYnMAASF@>nV{HEsKI1KJ
z+l9V>z}D0$R6&n+_{CsTxT}&#wjR9llsZXmd|G}mSd}fhyDN_GJ`a@*%S+*~;>CmY
z_K{Qm+rxZkI0~gs_TJBvK@5B!>@#t5Zsk}E_<1fUVZu}iq_KxIB<oD%w(-n+ScvS}
zNuhLM`hD>Es?Fx`FG<R&<GFj{n@xf#`d33P`Ba;gr)5!P;5uIgeTa#O$%;hkukwf=
z`92Di_+fB~N`*-)rd*U#Xlm*xnB3108!u$v^!2}Yg2_0q)Hd#QZ=3y-2adcsmHVs@
zdB808(Io662;!S&WPz#baTK$sYM+RS38k-lsc;e3%b~bZo4ZQbdhw_Dbu>{VFA`@<
zxx1*khb|yCdm4EeJZwG}(<ZVoj^;imFwj=G^{S!#*A;ePuglAq*T&thnq$);W-h#~
zDm|v_4#6;+6<hpVaNgC(5Y~^(Q@^MBjt)F7hgsmM2n(XD2M*hUpT^P&VT?_f0V+gh
zw|w`Z@O;hxSspmY|2`f#Wk-*zYf3<Ey^hL6oOdHGAIBuWs3+G2X2zyEx>>{db=Hzj
zh30ytzw=`ICs*Q%CsWIY-`$nrJ0#|LFC}QsyguID0DWvEOT)miOhp3k*=fKM)Le4q
zGb}c2lhmmrX4Z_`#dYa;`zp?Lfz%aX3^w78VR|qQ?d}mW`ZU*cyj9R<lHCl7cbTys
zbfrqojtVFh9_Wv;)XFMuwFVxuvzxsnS4J@!H@KV0k1Njmgx%{uv@X3JpAJg9lQ-kQ
z0slBOX#K+PY1z0LO$0~VDlpngoW+?R+-GG_(04yMPC^8h^Ct-<a?m_|@#z_qtSH#G
z6x9*}KTDh8rnhrhpox`>eQf@P)1n{h_sJM-AQUCkuN5O9VK~qUpwG(6dNcAw*5cSG
ze@4kosl74a_rvOI(odgFWA0Vs^9bMdrgFL8%aKQJg#v5fZc?la)<A<AyuXuL8vRbH
z0*f+v0a$*hOgWx7kpyWbJ3X(5T5c3(H0qa^10uB`;N4=>y%t1;zPA7)-!ZEJTV~DY
zZgjA5xrpFIL}mgup{a<ofSWkJ;Q5C|>;Lbf5Jmn?vbLR5vqv7SvJGSN_Q+1y{0L&9
zIa3cSKfxmSZy<N|X5S}C{{Xq?6G86v4&~02Q#StqxeM_tjsJ*VkO)-D-J~c3%Zt$b
zmKvT1T*!klEb*w=_QrO#rE2qflw@cK-bN09H2tE6RKr~cc+7u`vlu`J&--tqXvQUb
zlV4dm`dAWONCg3&?&<$@AxY%=0!NF;#SoJ|?xpNLZhf(o4hU`m_z$}Zm<xYQQ>Gt{
za9v|ifl<Yhj**JfL`k;;uu)4KKt;Ab^-41?YB;4ouD|*G?rYy#D6{f)fzta=CCIbu
zseQzJs#kh^^%NwmEvljrpy<r#iagmoc>>a>CwpGPA5&4U_bNMmvpOO4d^>t2@9Dw8
z6}jW$%kuX7?~iWm{3YQ9{J4tW{!4P1jQLA4o)i9}>o3W)w&ztXQBe6t;om=c?T^0u
z;w2-FSi$GY{!3EC`X4^$Ti<P@R)1x(;O)Qwe2ZVBt-yxz^*vcZm8r<w4b~wzi{rf?
z1tZMeEQ9aWqB8EW@kR@L(p*))xOK()T4Y!6fdWOTe_%zs-xn0zXqo#6?Jn)*1~ALY
zwaE#iZK*k_6!rn=I10nuwo7kl_6!ffT-@Q^*x}yX^Pa|2a7cGPjt~y)AL)>j$^uw+
z)lxzxO-!|e=VCi*9U10^y``(?(CP~C#sI?TT^Y)dNY1$gy^WL}T>lU6586Rdqz3}c
zbxf_FhiBy{TqeuRS00j<CQYdha7jA3zW@FDO0fBtk`7^kSUEX`W}T)h6$@=dyK=}?
z%e_we3|4Gy3S2}{`Z7v*EsSJ0R%{IrW~Loc)Ad&@LI&IQ(s=<@6{zs?hM_bgH~G>c
z3H{~lw6m-1^qMVHcb}xow5GF{!qab(9W1dipmiSANXKzmMiW24?z04RFyD(cO(P&r
zscvAhI(z7}lTLC6j#F_b!cFPP&%P)|Lid+&9zV}3Y!zfLTmUt%XEQ0XVOKb9Vqz2<
ztAmG6(bAM799i$&n~#%sQ*!8u^;0ooWGi`vH~ko?b~~|jd27c166;8dp)H3sMZxxU
z#As|<?U<(h_e|!iLZ)qu=7VAtX8-f$?EkT_78ocU?i&B?FA1#lx46?Cm1knT)!*lj
zCDBS>NxSRTt)k}*)uCOh1S9Sj(1$sGzXVb=GA7Hh>=N7qYAUY%H?JQo%-IM}Kv|8?
z7OTlw9{05CN7~PuvrRrP8%9jPK%bCi{ezPCVQ2OV_qr#v+4_mA+v1snpQxdE`>meA
z&=~C@Vq(XG;1)sEK{tTHDY&Qjm@^B_Km!fUSd-;iqhxEG#4DElDVuo4CgFgS)!q4d
z3WKCw^tsoxy&$TRI6m!ss6tX#^LvQ>mS%Lu`SQSf-~7z;iN7-d7m$_FVd`J~OvsZ6
zuU&j_b+8iNwMv2q=WbVJ;J}WaxJFpw+o_pv6QTTqGt@`FUo*}4SQJG?Iu6FNe?zad
zPxOzi=oZ|65v@|3l$3NEi*?h@T{X(C=sTVhKf_ESdn?M1VzgvO*y=dE!QCHae@=a%
z<FHJgP1a~G>wn|>OKuq+8TSQ+Tq}VBtgTV+(;VH!6+-W>e&KtAfAqT<0$@%yiqWKk
z$Rq=hd>rde*F##5w(jhgm$F8jk8X+5#j_|sPx*E{%Gg%gFA%)J@oP!l`Q8+rpCnM)
ze7}rBvCn!|=yqev2#gd2;3S$)We>Z<A^wZM_Xj)HPQEi!lf0>t&Yw7;_)8LkQNop6
z6Mw8$ww3)@>Pabz`qnF053d~7SLpL|>9Mlti7i;={i5;J&|l~BVQ-!gV*MMo((4vM
z<QZE|=k#ap&)t1|HyqVWLEyPJ+uu!AYPnRn3PWcNi}+R&1{bu{Run8w7QNoMv~%FI
ze2WN*WAjajZ$FRchA2A4hO;XS0z6{cL|97Rhobf0nQ4u>H>gLfSV!gYENl;7)UUFX
z0?%bZBkJ%O)8}~`StYhM0g0yg1yfwT#oR+~y|}qFl>{FKqd_fRZ&HJ%Fb)G&aS;-V
zMWb`+Efjg4i&>^rNVbJ@!^~+~2@N!bCg!7@-`$<msKiJ9$%m(#An<3HVp2;nW#BWX
z1d8C9Gn3*h+h{ut3nw`pFCPI84U3<^<>~772Ee$*$>>OSx@)H;m0HyJ{F~XM9jZ^%
zw?r<I&CXjFcPDS;jXY$nQ}zj8KS^Ag`06KhBM0+iINGL%M|op3jp4y#vlbeDr-4M`
zv{z%;X*~C}N*0CCgPJZP+(z!WLJM5LZ#fC%5T#yQbl2?g!zlii=u-lcoIM0k+-TH3
zh7TXQqo&OIQ8_(AS0)XaxN%a2aj_ZP(W+wR9s}V;t|S#{AERdp*88SC0>;-(4TdfH
zLKBT#1Rd!BQdACVky$c5_aMtP+b~EkTpMjbzs|aOGBK2Z;<{O!jNf8qe<093YAgAI
zcBVd#_O({5vjxOySmrTY!OYRlR}oM_y`lj*b*7xXWgX-6dF`h%0@O$Gxl5(Jt^W>c
ziRdz1ddl`D?qjq!>wp5o8sK29M5i2)ZTd*xM%kU+l9KCz(8)17R{_XUtFTaljN^Ox
zUy?*L;~&lLjunfC_ZfZSG#3*eB66w98<MB)zhhUER-EW4JTx;iEbu)VG9^{>5Zc&H
z$2kSi=9hh1d1X<4clCyc(&zdDyB5>Ju=VU!l2`3#jY6!>0K98ui&mObfRNd`!9DBC
z2nVk12N9?#h<8>cKZCv|_~-ZjbzMq7>lhWcEB&aOtvh*$$(;alW@VhE(F+F`7`?sQ
zZzsVVxgE%=cF15_W%F@x;aW6qB>#_dFMw{cEUt58iS?4a_Qx@nDD8-sxiNQ0HCwDR
zxV7axlPWbc_2BxWXqnwc4Olo^xd)ElRDy5vNSaKSYIPDBilg;L3h~`dV-vo~3eGUB
zzfxcLGfQEZ|5{Ba!pToYqK>;^#ne4Ep{FLArozw<<DO4noa#Dnrm727DFO;)k~jV5
z?)qQv{O9GcfJ6FWPykF5H#jtsN{i;GnY(97^}7LP)tFzB#T8j7Q^RU!-%Kwx9<MDG
zI;qm$wJGq)@7Z-hnYllcXBXjtJD%q~ymtqW**j2z8yEWSBEMPn-azl($YT*Tni&mz
z>!mt&@4fL{!POGp9hVj=4$%4dZc9okYI<q}$!f;HPn`@)F7Y$mQc14c3gEwB8#^E&
zxnE87r!wu2pxvJS^{apV^lRT|O-<*%rC4~{-JWW63^Q)I<c-sP&sXJCRizEoY3bTK
zNlx=)1_QS^6Enm)u5H)mC{cEPN7R$`)DG*S+;>cD;tfspzggvWAo#hg%_|xHl>MnB
z0KyE;&hkb7ZOZ1he)1Ds%1BM1t=R7~14-n>>YVsA#;gb-ux>E-VGN!yhi2`<@)sih
zlCbr-ITzO1sxnOA#61gy5U>0xy0r)5)!*{UXs(2>m-z_<UOvS%zO9WDu-asunksO_
ze?wdp7{q<wGuwhWtJJo4isIdv&?2a&tQ@#fK~7y*XYKa6dC@fBQ)au(FKzhU?PH!C
zC1ikecg%aQUtH|Po9+E8YPi_uWA^evf{uNITW`IpSmqCgU|v*3yGx%j1GL^iO*dDv
zJFncd1~Yci_~u_b56xpg6JIh9(Q%D+OqXRA<3)&*PC}FSx1*gv7N+?AnALO6hOFcm
z!H5<W7xm$|TfHRLZ=FX46~@OL^v^N(;sfYELM*!s+`X8r3|=b1@geV>0;jap({tZL
zhYj%a+j&AyDjf$9HPH%mCX-_pNi^NHikfl)`k|P>uiiq#GRPmUl5_K4$nxmkqx95E
zEd}7|w8*)^Z+?A``Gn2BW(W&BPznCY)h~GTV49sbzLf>Ape2lW)QrN_I0-7|ymqyF
zNgJwgZom15A_vPX+~Dq1o*g|ENX5agkS3*9hQpEe9_&ah==t^5cKU4>k@txROrgZ6
z1)`H>>G`xv%I#&plL+o+=n+lGlKv9E638N*6T6JuqRzD9CTv@2JgSi(gPewsd>(wI
zO2!^|Nf_NJXtipLPTRcpWv>e-@32ZVa|4P+(9|lCIpUM`C#;}ksw5U9++9GbX=w97
z<wifrHJ1Twu0qx)K)Uuf(OC;EX5k{PlsH4BPZAT7<u)cmH6{*=K%i#YSh6*SPhyXK
zmF^Sq@K8KYa>#77Bc*9}0X$DNle+Y=dYfc4ZPN+S;^#uKQF;^s?L`_IN_1M%2fFz<
zi_7AozJ@fyKoQyY5d`OS{fI2*VlLu#L$W>r{5ksWi-#(yY<!xA$(%`)lqCOt_WY|M
z{NM5X-$)p}uF&-UU}Zv=<8eaX623lbo16ZQt~jGdLY{5zQ19|^7XR{1b1KJ~JYd4A
z!8ARY3JCnjpQ9y%%!bD}h4Ufg&;4k#?~m1(-x+uC>)6u7tjwf&WDSCRi_KP`54Q?B
z#5{<i@YLNva5I~B9aimV3fzEy@aZ|rrNokV5hK?W#&udjZWN%zI{;{o7IcDIRJIvh
z$TbZ1=M9+p#g99jq!0{U<l$7l8@j^loJq;K>sucWyYENlr5--Q&WY$k9YrP|NLAb!
zVj0@r*xIcO7D|o&f$J}~4np)F{*GW@4v_y?nLKQ2%VU@Zoo4}w;(tqPdRwalz4snN
zKRTU*$tmw#^Ow@>=GzE)PG0_Kq&9%Lsq$Ic{6TWD4v3PNg)OrT{P6jL^I>*{t*eoX
z#fYP4=2jyZ$rx@k*T;qFJ?Y>qbEgA~<O9!74@7wPil<cpu+(mudV_mn^o^4a9?9>>
z>SopNGJoC;;0n>N(bSIA2bgs8HQ}=nrVo#pU<DMCjU~DF*v+OU4Lf9V1K~Q=NC&@E
z*~u?lV5y0RjMG&YWF;YKSSBJ~KZE*h=J#f8M~BWT5TAhk2~}(QV2hOtQ4H~k$;{?<
zs2O7t{RCHI-QM~`9gB64RH8*nm6aT!3HQ=-Si8B!CnuXB-sI)1jGIp#I91+7Aq6)g
zsqu0YDIi>#>gwb!Zm1k^m_Wg(gWa3h9NCuql2fhY*`UP~8SjV>kJTu2*V|k9R)&VV
zb^a0Dmd&UGFC;>Reoaw&vMVTBNVD6w)tPBX7kLpwI&vqsN~)4~;FM-LOQs0ETlc=*
z(?+y%!=OzWlIag%+{SY+cyeUYALhC#bC*zx6qayyKJbvoa8Hqcn>w8D-1mLlT|xw9
zFCB<`=v<tN)?YMLcJNSJ^-hdqP*otEz==@ILfB`nYL`%@MojGkTP({MlMvM3f`Oe!
z2qVIy0L^mdxjWU=k7`aief5!f+BO(qO}lKW2HmiB;Bd3Ze4vN6rGp7Iewp1~KM?)B
zVg#w5cW`BM_{lWD?eN2a1~eeVM$nP-7Z8IDrIw?=H%48}drBrxq-$|iBu9xwY|BXO
zb@_)h_WbnBT?>vjNHOLh7mR%PIoEJOBID^hMvc#}-BbA2tV}DH)jnST2Jl^B*M6mq
zELKl#reY=yVdqTZj3ss;lyp(CHL-gS|C<rC6Z4*P3G)rp7Bx)12B;sZ2TTzq<l3kL
z3{>lO55%3_ym9FCsFMGO=S$s>3?aAmlwTW>p}2~69vemN(KCY26_(eJCiZr+l{*(H
z(WcS=jKzpY*GWq?m&|{Ow<h%0d+K8)hl#Pi9@a6)1+>mq70z4I>=BMhPJ3?%Ejb)h
z3lIud!5)>vW43!oa9-a3iuGOWfUFwXC~RfH*@@zK(YYyB7cIeoxMM?(eh_*lEb`%w
zNZmHoik|q6n;{dZn@<4T&8K1ch^VEYf83WLG*|D_YWAuMM=-%b)>>0>szU+12qK_q
zlc<9iC@h1y{Zk6h&CPB<uhK)5$rIBJRdoK5^4r}fCMJdtN;nk5J04!ClK$8FVYU6k
zK!e?(5MR0^x89swwE{+`2`JBv%t|{EC@&b?kPQ^GG8R~$(VX<C54!gao-qK@(}?=2
znv&;20j_|G4*6QepU9-q{TcAyCG7($2!|&BD3E-v*&IkQ(j}aM^9y~(x#Kd&DP>ou
zW#0Tc^QB>(f>snEUaY+3qtdvr1qJeX-O(0CyBE<nb*u(%N65`E`)Ci0xP*QUVtUbJ
zyE#hP4VnPBKb_p=ZNQmsmnq0)#PrB6gr=x+jEbl<wX08Tx1^pb49(3o=j)Zu_H!CB
z6(`0!o&@en+rH)^Q-#XA7Q@JuYj+$Z$TIAtK4D#s2+sUo1aJex@CC1C5a;{3H;KY*
z^3n|{QR`f#=C6@1E%6_}Vy$Bqxyn*X4Qn&&d9<1<U8<C6)st6(7zN_6f?U0t0%Ey5
zI%;Uz1mKfc&R-exIh@ga)AQq8T;DYz4Yi!M<FuJRB3`#`$3!H0P`Pb8w&^?RLdfix
zmj06>>CYqOZsLq#a{e)ceJpvn59a<DGtge>&nJpId=O1`-KYx~fZDUMi0dJQ^<f?@
zgFrXLB^QIMyfpSgOX2|ymLIkz+jRMP*q5gAK2@*}czeWaS-F8YM|}|D8u2#bUc+(k
zG`Dx~WtCNkgi>0J)7`XPBo^4xXx|5saPyi-<(xBQ<giFKOfkskEN&$U`ubmMENopS
zXtK{z&FMuialvavQx+A?x>hPvtT&8Ao3peslB#(6lC&3;SoVgT@wIFjnZeCs<W_o;
zD_YJ1FOn^I1it0d9W-89)k_nzw7sQJ3{a%Qm!Y&;w3^~#sm2$#xLCOoOU%4*yj-r_
zUTpdg{76n_Z?oGP9p4v86CI;*!W;pz?*y7clSlgMXP-K%k$thc-Zs$x&oJlSe}y@_
zp~iREiDAxpW$Cvc&;BFsLQ;GPBDc)|CU1WwJ+BPXN@N6WN<k{4QkM#DYCNwTEllPt
zP4ZzIFxkNoP;RYg0hEhCtkiZD0*Q}JgU-boYh$FATQ;iNqY~Wm5Y2hp4lxdH2tGJ}
zbpoQ0oxcJulTdz~$AOX`eYHn1=8gDY?aXj~Js>~u{80+<=u9i=%|v{Shn-GiMW#zu
z9(CFV-5GsPKyIT!UDhKC<(TQbXcVARM~B8}%eVTuH0Ec%p0SG!OFq||`A;Ma-pw|#
zn^K!(v=&HCgy6SSsP_7KN>YV5jkt!P{_t8WtudpirNowatc*`fn(uZzGr*Tjtnhp7
z=K~>&jDT{v>=4VL0k7@g;b4?_Ht?*L-$~OaRm1MBR#a0$#vj$<%V(PKERgMuqNbb&
zJUWN4IV<LBAk?cOi<4h@o=Pi)M=^9U8*n6?a#W1$%Q46*e<r55dDi@Kzk{^mn{UW&
z5i7F16>=IKSb@_X@zL@SdI!Q;;RxzdWWw}WY#(GN*jMX>zcNr>-BM`%vvo24<2UQC
zM>nLwI}y4GdtF9br)HERDDmxmn9v7}(_fF;Z1AO4wJ&G<7FAuutOO@NTp5LtzS7gm
z&`A4!ZTl4aF)wCkp~CqBo#F~B(hJGbd@tPD65+x-ukHF|@@v_1Wz_^Bdoez5*x9^S
zUglxSi0qgodhCTV)nAh5dc<|Cme&o$Fp)cdR7pPN*6y`F#P?>)!evYrfZ^KfKu`+P
zZVA4xa?rUUcyqKSZE^~y%8P+Fmq_NV?ErP<+@a3noDssvnf9zHNP_Op?4U$|M6@Zb
zlvdw-ib^2D$sGGs<B%=Ypm4E5#npHPWKtKGzv>3y*d~E|C6k?x7(DqA5Y?5Ew>q0A
zyD4Ofs50eXkAI)GQLD-`=$S8xzFFwm{16Pi`7Hw&CSf;uqO6zlEx@xI_yT2Ql><_s
zk}`3Rpw@iuBV0M~*iAdBzePT|aMhOlnN-H}z+iBNC%U$7*pV*l)7qG|GOC%|UShNy
zr^m*foMif~VffxJ8{#bUi&_+6(b!3s3ds5VVn?dmugm7NgKp~gCGQA)v^%pT-YlY4
zJgrTstQdA0tJBKtSB1}qVuJKAQhSWj@j{>Dg$j5zAb9<Ma9{W|A11$`N@gcc4v2Tv
z85QD<_-OscxIa}bjgqLVrKk7h;Mf%xSM1^t?&3K3^JIdIZN@)iGvbSr<<Gay^cZNK
z?#Ocn?0xGN)uk$x{yaCQsIml@idBl3M^SI>2hB!9SQ0a-Vfp~F<o15khbBD1Offdf
zkY=cB+)#&SXm&O1jby#YB?A^*<!Z4I_HoVP%U)hMsTY`S^@!K@Em;X6cK~tSt!#Uh
z%lXMgp{zJ3ozaYw`$Y|7<#l<qNLM&eTD4pY_W%o+bYslmbK6WH7ZIR$sVOqynylGk
zpJr@Gn{LuAn^)Z49`0`;D?(?!q}UV$AOAF(8vRg{ziLjFg5RL95x}R1K+7JxIG0%f
z;@6q&KMqz6b2-)_#%&zEm3>aq@rXV|zHD?jYg{MU^xVsp{^Ycpl5$idZF}oM8zG_5
zN&M&>xww_5rLeGBlkrZ$j;g&Am<1~yD9yj_aeL8gw0LJMJ8B2P4yYV$xRno==kO=I
zTh*_#&cjOK==E9ogiKwiE2!00koi*N#`j<|!WeKpXSudc>hA#KJgrxRi#JrEh^SlI
zh{way_v%M*#<dI#4~Gocm>i=b9HmlE!E6!QpC*6`H>osQ<YPaP$q-n2PBz>oPO1D!
zGhNe<Pz;S+;FM<0oG_dToVt-hjJ;x{g<<5whh~T6xV4{_N`~&q_meN|FTnW@c&kSF
zKGD&`w)YalKEKCH>1$dr&PKo#nq~>6n!HRdJso|5TWxxszvaAg1@Y3%SzN;TOxpZ5
z=d=P{*UuYpS?)AfzDyWXQrT*AT9r7a#HruVGUMK0?XL$w=d7gsS1tU|NSJp?OBY6F
zsEY0Rdr`8K?=4}CyE`7D_3_Wp-+U954A*jpTkd#nn)erVxpfLr*S>O?s%2VLwLZhA
z+3;~mIrXJscj7te^}wwmJyLz}hkn&$P5mN#FD<RCr0r9lTAx-V&S@E=3EoiXX_6c~
zz3M$@RJ6sD*h)hj=9$C{8GF-=TI|fe9_!p`Xj}>thEJHoi3-HWig2ZEYfk7IX8xGU
zN8`gS_4|XIX4*#RCNOMjVEXQy6~Y{{`M<hi{PR81SVHNaJ6V6M?0&O<z8yw#?HZGp
z|8Zs7C%+rfF1WId)z*j^79$?9=FaVo3V_Hnk4vZ)<tAI{;-@kcFwG%?qh<%w=i~g&
z%6o9#N?$X5_-A<G<rbkmZ(sd|kCcel+&Yqc=*{vh<dSS)naaHXAg-mBylqcPFW!QB
zOz>q}30J3<#P=`4bsl_m$|g;7N%^E#Cr2Dgf1XeMp>SBp9G!bYnX;~`l|dRHOAPqB
zO*EK&J2}4tWGFx|gcY`gWV$sX)*5DnbH7W{$|~BVtx}DXBhu3ma%<Y@moe*TGh9(c
zJh>c8mYm^C<URRl&$^+vBYpR_cmI+^n_dF)<IYMQvS=9#ppRFt&<8tRKI!+kQga2X
zG8Ab;Dn+KV@_m&uX8Fe>7QwULx%oj!6Mb32EL;|H`hLB{q`sZ*jtZ{uSh`DNl8$G@
z{ps$%9pG3PB05nDD6*@&FX*a=?*|=1XK{|x7V?G7?fdme$B~xDDX!U`0gn=L-o2c4
zEmT>?V)f)pzW*h8AM&%$W<|IAU`4wJhvo`)@vqk9b}4;ZJbOW&yY0xoF>%Uxetu}e
z(mt(IZ82>pfz>_~e(wVTbAYz$E8GM29l!Tnnqz&4<u#h%oq5};d@~o=&T=sSdX-FR
z!tp$V(pglK#h=}*jz7QRb^qv+Rs1EHbPl!+&&1#P^TSH(FG)mP*p=M5_g|7)>(Q(3
zZKaDFGp4@^bZPvls9Ds$IGt=-Cg*L}kAoeIolX)nfa+!+E!tB{JBzlT-bW`TW{8}C
zf?uv!W~cv_ov3$<$~u+F{^6uLyDYn`m{ILO$jawUv{<lC|Lt>tZz<%TsuNNfWj2$H
zHw9gW*>Vl+?d_)Q-I%L-2Z<$<|4ZTttB%-i=bcP`FW2{|L8biBG3NHNdmTdnmSKmf
z&O4cpY}$6(Ovj&}kU9rdhoVi&UvmVSElYn0-UWA1HTpN#qhj{L*n6dFqot7RE8RU3
zDe!edb`wLM5MZ>Tnn!PNpjLGTfiq)Y0~qNlk4xW<Z8}uUE!Yl(<w=^1+XPOtBami&
zid9#L!t5zw?bnkA!SgNadHgXhS<yMmf_HpKNN~CHo&wLMWZ?~XR8wU-mZnpr-VLLG
ztSRgD^l27xracK_QN2yIQ+sZ-Ue}doUV<IUT;YqsuQU$MY1LV19{qT?a62%5>sm{F
zqickmDR*u47M|R8ZLk-EYq(;;F{rT=lx$_uZ4FNTSedD6@dd=)WACBC;LDp$24BI1
z%p?gRMbORcr+TRJomo1unjh;{-_SOdA+rFJ6#o&`3*Kyh-l`@zc*Pdso*Cl+ipp9q
ziyjH~&3BYTmNs#h+5588bfGvN^c?T%hDIg3uj&%Bwl!qQ9t=F>mKyFC0anz;9M`Nh
zOlA&QW`;G$edMe?+a$oMM(0{%LGF#OdZ(otn=AG(t}M7q3#r;u8*2T2EPzq!NYKP4
zE=I_Cc>%3@37B@<lk6FSny4B`%)NA3*$$MkOy@GP6)uR5&yM#2Az~`}2J6*SVTi(}
zm`xZl70v}#$PagMaNy<&N{BxrApsElTb;4<>)?8eMic@oZaFVM&BQW#bI{{pHqB(#
z**Pq!VcFwwU^;b{MqUfHWw3&&emxR{r}i-i8$JPgq|QbP%}i&OE_GQX6E%rVm^G(!
z0##-bD}yKZc#Cv#!%(z$zcgK~q>px%fista)^=Isb6Fvwupx1Yt2+ep{KnY&utpC(
z)h#KavI8tsJP|KaoFP6v64SSf6#U`28=YJ~a5M}%1Lvuyt)}zF(?rurPo=C&U#AJv
zR_!gTM816}<#nqy-MOjiR{Y$$Y_FDbq=U@r9i;e!d>-^$<-`foUUsA7y(%G2)#|E<
z0eWa1S3fBA7|7@3q7s%KU%11VX#92TgiOF->J(|j9;YL%G2vAQVAj$fY?YurtU2%&
zS{NAUu=~wu3ZAyH*6#ME`gK%STjTx1dELppyQi-ZVmC0*s+vs;#$F-z=bKF=4YWht
zs>-557(#-F-xXF}q`T7^PeaGR+a3-Tr3_)tsiyE}+03D_@b0o|$xDG%8}2#ak+9>&
zNaqmN*H?0Phfd75M!Zh6Ud+=78^9<+yXP-1Xj9suGbD|ihBW=?F&>r`vO&MbwNAOm
zrdL}P?p36YN=`%8#B6A=5*x(Ig0c3m?k`cltmA?6e||O|)_U+fo)qeHmYRW!WUO&T
z7-*OE_&linCB0$f7-*0infN@PTf#QLL?@ud4*bB6plqX9+s$Q>|4oS`ROjTw^eGs%
z<e&Jla{GciQEtD$QQ92q=kDLIiRr;!f$2`=<3h~bFHD_d6iiwW%~<1U?6{5*ZSU3s
z11+YjZ)Ca-vf}9MbVY^nudQCinfxU&@%gDe?hvyLSFk~BARPGC*=5)5&pj@~o_Z|0
zaHYFsLohCDlQW7cy&@^Zaod+3j~$P-DKRjqdqw}tudr3paa*r~8OaGEmy70upA4ic
zh;Q`}J^rIdNj=w9ZLH03%;Qxv;k|Zq#TLMR8|GgdofhrNzs`JKDxuxq-&e}Pa{37K
z+ifRyYR|5>)kJ}AZy1Y0r!Ou9&RPWDwh1z=`GHMLqVxXYXpQ!1^_wqrnZMwzdR6s5
zsC)Bpw(|e~yPuiQv@>0&wQDh>Ew$H@*y+^PmQW>@5T=$OBuWsXm`=yOjkdPLGPNWm
z617ASVQMcziC7Y$))HbXs<z&l&;9-5zOL`}{p)-E?(4qKKja*d<eZap-tX7z`FuR{
zi&Ewc;KY=b<fW8G?r^+>^&?sw<G3lv?M(OXxlpNs7Au+l+Oo$Jp9>=eOW*+CuyCNR
z-A!T|M)dICZ&Sai*|n(+`QvTM;ii1cwvSyBw<-6ZUHow8k11&iXOn)njts}BXBjLJ
zb{KYc!&{n9u4_wtlU%L)sJFYjAazp8X|7vMSGkvgoVMVoF{=XE_Cd(P*}vANjGhYO
z)LY!7Tuseu_*)x3Q%%uqcZLz9#v;5+bQp<Qx3LgqgGFt+z$}sVhC({;{B(D2-<S+z
zAtK81hf69A?PacR-nn`8WecKz!h?e3@eL=}r1}CXa<lu*>uW=J4n#Agl1=0N{ii;g
z4xMuULL&;Z<~(v!%aAGuKygN}$R)?W*#cW!aTYE$9TQs5{UW&qMZeJsAP!9JzWV3I
zHHGJ9e*@Y!oxa1TLN6W<6ew#U=fI)s#%u7BW2~CH@U9x0qB5t%KzdP%@|V06v9PN^
z&AW4_PN9ywf(O<F8mT6Y^U<knhTAw#b#%@9!E#uFR#<nBEjwXCh(WW@Ukq;m?zq|(
zPebqZrXWKCzWXw))p*;Tm}oBpK(vW{dDQ-Pf|egR`>aM_+r3)@FsN=)vu`J>gc>Pu
zpEOzOho?T)H#RGmKW&lus|!%>_xYDqukTw6EsujP%%>k~z0}C2PatN=n!n`~;<iq6
zPaps6_i2xv2lcI~@3RGnXg~*LUV15>wuCUbJ4qR78O=?oKC{T(%I1nc_};i%is`36
zK~+D^XO?=VdSuY{rgc-A6;7w4BOvVH<yRk8u})Me@g4Z4+DB$zeM?ndHx3Ms3^Dh>
zDr`U<kP2$!U8PjrZr=1OKWEx@wGWE`&a`wUclLHI7vwR37_Vc|@Gg6=kH$tjTQcsb
z3uBT}=u@Y3u;WRV<16TRhZ>dh?E-SI$L#^^8b}gpBQUJ$eY+%d^m@4PFg0W8E>%+6
zAa%P3$vX;{vmQVoM!Sj2?73<5@A|EOPH+^yixpq_tPbfLgOBTAwJRuNAL<%ZQo`W1
zr4H5U-tz1W|L!~-N_YN%2y3iY-$-xr3x`Z*FM6cM?&`2Db-a8)mB}5|Ci;D;u_dMx
zqhL=F&*HT>cqFp-Z&J(uIE?=P!iInOEzXSmCCsdgK6d&WW!&|A?}2wQ<Fw|H^57k$
z?)NimfXcSkfVl7cf3ueS-@V_9lKX3djwc6Mop##2I~%`*7aI%3z87vM3dO!F5h1Tw
z5Nk@8SIuMb){%^qFrnDD^(D;va&h<sCD-ZTVe8uirQ>U@xA(2;Vk8=-6X{2fHb25w
zBXmu++M=zda5F*6m~xW70edy?GbP}lh>)6#WFDd<RHujo_7TaeCB(`i$bU(FPniED
z^(6_Vz6AwpivLJ`mplI>^-YV;St2S@?Zr?wdqpxWkJA`eFLinM225;UpAWNX?O>7#
z`BB)$-f#lM^(LZuWTZ@uT9y-^Rc>x+Y4&MSD7K4sAER)ZEV$0$hzE=G9Da&miss?V
zuMnmg<l~!F@Loi?aHvw5mzST6w3N}dl~n^c=&)@8DZ(0h9x*@FQCKzqoSZBW<{HEr
zZG~Pe-}&u-x8wiEXPZRt9!8IID7U^KgZ*9~v_a#k^g=KLtx<^2FEfHR-<N)#b3Ze5
z@hHWKZ|}H+#>U{jv%Zy~wTgVbD6~MHWVYCD<+h*SxD#Ure;i4kZGf#*56>f(cMw&G
zPt<L|0X5Lj!SX-UdI(!(6h4aZ{D)e9tGfWJ&ZY32gc%Rl?}w??HK|3?z3u5VYCq&@
zu+dchP*d5*;Y7(@xITcDr4N|z%Bus5Ie>LB$L=gmN3f=*dcv8H-_h^<k00!Rx#7P}
zZ;zCeAdd0nLW@Zm`RsjZ$83q4H}9MH2z%1K|14kR{Ql7I8;=z+%;4BI!>;yz=&y*J
zezEZ7)B4nE#=s}mpX8QRHS9I_Y*+6;ap9NWJk2m8h0RlRLuw$jGJpe6%#m>Gm27RC
zl)$EIiyoskcfH*+EV6D7Kyb)4Z?B&=nG1Wa`P&x}?e|H5H)IKKYo+^u0z+&^z?kDJ
zGWCkh-<Sfqv}v4|UR$)f?9mN0)PLvjzJt={Q#-`_2rfY}+N<9}eW@4kFds%%f8cF_
zY=0U#$ow+_@*rp9;j1#^n7E6_Ur%j;2SX5d=h#^z%dKOY(cRnbZoguTWAwYXlS5>!
z*|w`5em{&k2alHDNd7S|M7`ef(-Ub@_%8EdSZRt^-q;Um-i=N|M|A3(UiM87-8Ap8
zhu&k4-amZQ?$wMh!Z*uHr=M<aJ|7`(meS!LkRWa5fB)u<JL<w?CE;=N*j44orH%8~
zcL&19-<&!c$rZp|Vo$IRCzJ=qKOKvGJ+)CO1P5@UTjFA0wR@O;J=Js^_xU5}c=FQq
zc+*SAfAVuL$q@Oc%Mhk_>I&wSSn(4JQ80VFWI6kI;p-`L^zdQr*hqNb_R>ARq;C=R
zqr0aQ%YVt|Rn`+>G3eg@alNs^(_gvKZryg58%}0MbS?CnJ<`jNOz$%wme3mkUKx+E
zR8TMhiw8k=b5*KKc%m#e6dW?p*LESoa%eO{ojW+hHBf-TY|m8n_Pt|;1bZ@2<I@Dn
zVxIsM&Su$I0c?gYUtSA7?bdFW&=3I9${2?kh<S<!q388gu5`@*N>jy~!A1%bV-obu
zP#3LzViaz47jOsk%;&A|_UHX=po>SBTgDf9KkV2V?^G2|F1d72VW2y`@pd3Cyeq`w
z*bqvvf7=GH%)sSP=Ghi~3IP;_qW(6O4w+}UaJp_pQMv_y1=e;_zbWY)z|YVx*}E}G
zx3K?c!*UnKr=WhQzvk4D;D<d`SWlMeU$2$9I#fOPK_V;^PYlY>i;XMK+BLfv|3dR}
z%;1R`UU!Vj1jNZ-ckrlq9;i*Eq`nyX6*DyNQIoy`J<ts~g3U;)pU!WMSSpMT`QSHM
zBe_=<7>z<taUi^7<wbrBf}tA`FcGDz%p!YGx>}~v3}u8UnySF`2+jn1KYAV!?fsyS
z9`}IRW_w{3;5-g?Y2t(%WT;8Z6;BdR3S$M0scZ)^CQ%*SCA#2j-Pm)@?4BEie4(Br
zI-j9e_WHu>9{_5BkhK5U_#VCTx&Ke{lKRk)KnDzU9JT;}p<oLgmJJ9j6byF>e?e($
zYYW@H14EZ(mu3gLH2-;YZD7bhDN{dxTkw6yXpwe{W__v3%Xr_zhPS-lpC+5^jav+!
zKLp4V|GngdFNVXc2$2a68MXFB*S#DP;r<y1*8=7dHmx8*(b$gIPR+D|F*({|)BA}%
ze!<HDm+sV5Az+a6v(nuyB3A40$e7@JqpP@vC3Oh?_8h5U3A>z9LXUCz^l;0sgDJ$G
z8Igw)b(fofq%<Lyd~cbOKUYCKIjs&#S`qG%7(Y?=fK6xTR#Wr2rkHPd7ZtDTa@VV7
zAvJ03dDvB@{dBrR5&n&X&$h}2d`dyB!q4FC%Qdsc=V#4^NAoY}G&p7#o!gRVufBYA
zV+`Rhdozd`4dc_+Y@O7B)OZhLsfl4v5=qgnpD^|qE;;JH#OvSJ*A=+YMUc&v6ou@`
zlJyHBA#?2f|5QSZ#SZAXn#e8`{NuGQ3Gxt#3t*a3MR_GD_+m~{-umxt0I5Syto=zY
zN7&_14By0v99(akQEZ9ZOF?=becoJ2U|cs9$O%c7lLR4({KhJxWyIsU+pWpL{yn?&
zb&V&IIseMGo;=~IxpquGf3c*W>|`@yV^I-tCB*1Trb6#ezvSV0>-SNsCEc%bJVFwk
zxayUX4p9fUBX{THGNz5;MUwjnGSXOfD7{r_cAPxbJ5q0aX~jtwPlX-@LI&qXb;!Ec
z<+}j(t5x2<9t=2fS$F{>t|T|>Fc0e^@WUJ+K!B1sqd+FpD3+@BS-H9V)sZc))NHR=
z*WNZ{%90|!S->*ywCNl$-)g0YPAg2%LL357G#Pk%I0W#68F<Gi7PaR2zNAZC`CD2Z
zqZ2-mlb?_8UrstMG`B3H%Uauzk=-#DR;yEFzmKGTXobmMO1XN;b8<Xfs!cll*jkpy
z)U_vx2c%(jjy+GW!?FvZ8FW-^noU2OQz9U);;~3Bku-jX=Y~dcn$UOHo4W=<VF!Fw
zXPucEu}{OgH568`{D5<pz0DNW2CT!jv0(ztp$8j{P)V-B=_SJI;9m{__<yq*;pg%J
zeAR2e;i6;P1}7A-6AlZzBZv7XRR<YW&D5~5qQMx&#3Je7YH?jLL+cI(t+1IU6X^;`
zH?S>q0_bbU$Ma9JvqZAvq{l6a-GWI6y`;TTeEvocn;h6|%WE&Rf(%j#vliS4Mfn|f
ztTAG`+Sy8cnL>fDkTNIdX;IrAn5YUDSk<o+`0|ypUoI)XMWyITsVly4R=E2sZy0hr
z|F&w0g4^YVmsj#0xd#<JCSVkxH*OrBv1n=-qt~HN*HF<WxHzjR!?=g<MyQ1<%7{VB
z;tDT!oU0xU&+Ko&FXCP?#W3`2)<6XD@$RpFkNJ_2oQ&@dw4T(d4rirj1ru*EU<={)
zp?O3lQc6kGG0`h#v%@WU+GG-#SEY0{ZS1_?18{{fY5b2Ib2gojWtdj@vqy<UAvnh<
zqkh{!>*+%Hjo8Q<_SJ%dWls;ONV&PTw3=%0Y{EM>Ee(;PbFjO@Qf?>g`bt;%4HgQ^
z2zXzwPg3^fe34)_X3!)RRZZB#G?UTc0$z<?Ry92Dz~>{A|KvDE*Kd5#v~qXYibz9c
zI`5<!?Fw_K8?3H<p7&ntPxidx_+{I>{rOqLktLg)Pa!2{$x>8fdB=aW)jgRquhhn8
z!u=8W_XP2>*9yL=K`xA!!dElKHAzWHASP?Pn9b|c42&Seg5*TD7jq63^zrej5zF{M
zN>fTea()t;4F-v_q2Ue|q7Gp2IM`m41=3`K!Hz-=n2q_sz<{|q%DAQEBXHigeco=H
z_J%9qJw~ZtAA~;5mrM{QYdue|J%ssC4>aH2uFAZO(bRuOSxt>N^0Lpzm2r_|M$`M4
zZy4%mg-pvPLZn)?YdbyD7EFSSRZ}CTE1|?C1Mgkgx$$}ioHt3p9)s4-p)uY`r9|OT
zlJVWA&kl;)%z5x6vw0~?$s~!Z4+ncvglR&9n>AXRa6-{ASeHnc@3iUW`^&BXH=2!N
zW9?~V@4u<YqVhd?o_?aI<(pzN<wxQRE?>S=PR9`?%Imoqo$$HVD~mc@ij{Q|0<L;V
zSUiqZ*te+kUNO137SmIwU5A*WKNG%RUU$zZ0z$+FyR%RwDu0jq*MS{bm=;ys&Loi-
z(4+&h7l~M=XWpHPQ0aY7byu@xSEe8b)%bxFCfJ;*v*22!;estu2uN*M{^!S%ynFi>
zg4+A#B~LsS)zG5tC`2d++lDs-x`_dW(PIL;Qc^zt)5qa{hHr10<AAS6+-P6Iy==`O
zb(Hl3H&561yTv=Uk@s75K;z9u4&xCCwhH~ctEbm54)LOZo22eht+n-nAjI9l!cg3}
zwyLuT0p{AqhIhKBVOH{qc?6P&@y|5b>|DYJ#1ZI0Dz@l!HHc?zN7JpOFq_|~P?+eF
z06aV7IJ~)Hj<}RhQP3-L*nN|}fOr`SasA^1R;EgFu+*31SqAYLjI*$0(?hos?XEos
zy)TLY{OMx<Wd#b9J7$0UVkqM`^m_PF%ZHuRO~bCtbLa?0<I_8n1`XrGXyh(0Tfj7X
zSbdfxO_i(K)^4sTmC%r1uo~8?+4y?O&7y6BZg-^ar48%jMMrzGo@~~~ILMi**pBQh
zGwWcuVJD-56xA*FDVl}$v2XM8aXyKmy|zA6L$RJ5ID;)dPBKIbu~Fh*Pc0@y8pDr&
zj9VX{3qBd)w0){WQn@2$C+vES!>_00z8C^ZrY>qABYx>D`g$s0(&sa`gC@+a1PpBw
z7UjUwgD1#*$5l#EuhLnC`FoD%trSKJ)IR^1Fs^Y!D0S3n{XLEG`|(eZh)bE%snX~)
zEYP@_I&|9T0naVilLj-`@CukXtk&+}^}`XJY|NBTAjEcT9KiyEy9(78W=Obj3@2#j
zOt8n~;mO`)AEmlvJ?4@fvlB~q2tCHcko`_mD8t&8>XxsQ9XqX@%)@#jS0dgUi6FSi
zgC<STlTBS=uI-vZbYFXeHdW1K{DS4ntGf}(+dZ3qz`rXFFTchK$shJ!?!Lw%tr$Xm
zU%391UXH79W_XrnTCrM0HZO^EsiCBk#u-1gculs-2~wToNKLt=ZsUAzWGkN7NWpGx
zxX)k~!yLM#T$-KK+?_9*cbDvszl;l8Kg6ECiP|wPY@ix@n-HT}feKZo1b2RtKohF7
zqL*b7vF_uZ;T0u}_sW;!eC+*tYQ)>Y^cm+@fa1=ZxYjt~pn&5bq-QP(d_1j+56y&s
zFq!yJNO=sXq>MBA{bk2%ME%^JvC}jjXeKUv7X9H1g8p#9-B$iV_BX%3E{Xm4a}gJ=
zIQ`tp^RZG=S(BYs60?yXrqWS;w#9NQ*o_n%iKq(~ERx))Qc{u}pcq>4>-;B8TibG+
zJNGRsgikSsN=bdu8$&{|s>4ucI6k?*dF;b5a7h)miY-#ePX~nAdw?cNNYt;VI%{A>
zr190H4INh+)ZPZ<a2ceniI)!Pj!RG=`mD|F7YR0D!9w=2{SQ*<r~e9|`mBr;nq9^v
z##Q7cXqa!y#RT{gUiz`so;Is2E1mU-;K}}8l;dW4zlpqRa&G-9Hm1Njze&s=lfmWb
zNAB<W6BjwvD_m~~d-rO%8_i~llH9e5FV^Fyl;Ett?JB+mP)+LUn%WP+2=gCGwAPkZ
z_EssCDFG=OALr&>@`#lwUNL?m+8;aH?f$IS`$LcOzG3VSve3qL?~$l^S*5a&d6a#l
zcceu7`t>DCpP1%cFu>k%*bxM9&;~iOMTC=d?anUy{{%AHphT8>votkDgv+(go>x7s
zCg(*OWrhTw;f)w=W(R-zYT15gVo@qoMX64qN&|K~hZlQVJF^4nVfM<U$9L4^IKmL@
zr=Q1PPjs<X(mnTcBw|)0+2eL#IAZ4f9*o{X4Ro|eD-UKRkEatR=wL@kFFRai*OjG{
z7MMK@rbj`FU<$_lHT;z(z`&ywDzodgpyeKOJlqJREnA62uwSB5Hk|E5z><4PDb)*8
zcm-u#<lee_N?afzX_11%%oV<oVGGNjhh{XD#6>Cg%ew0hv!4O_3B?QOp}UV)?5LTJ
ze0ALv>o#GY_)a@4zp}c($s>|`<IUGo&M%Lwckb@|xfy3kZ+d!Xv)XM0xThwrzO)*t
zk5yP;>F_nCO4!X*BW4XT5Gdpi?e?jH2KZZo1O9S}a(+`V>Gk^~t4YILh1>NpZ&P39
zhX&%uZgO2+bT>evOt3|&>;67_rYH})PZb-ovDrEVgnU!vd+zJ0v%$}gX1PMbyspcc
z65%jH_hexITcL~1=-$^;<v}^(I{{x$k!gv?GiF~;b@3@AfVERtzxnOd!uw<AKZ&t|
z&OU3%4bFk--uAD2T>1Hjyc)I^iv06Ro#&(Lcb>^na3+tq>1Pdx8WY&_)9)v7b#bfv
zhHgFro_E-~+5=)``*6L{xr2n5eVWZUX3eFiq`9KTYf4Ccz%eBE#{!6g!yEw%Y+nn>
zAP`n0(WqUUDTL|SPIKP7euf232W8}72Sr^5jGNY3G)FZ-9Ylw1*qYyOmdzvDSwHn%
z>-=^7^Vv)oi(HFBZpB!4Z-qC`9mrKK<zEzWPEsqAGoO0l95e8T3O+dO%1S8=b^ML7
z`VLQ_F!1_YQ0)56FfeL)U%G+K4#x!G6{<P!!<f+;ZusUB1tsc1+ZbDuHO$uOa#Aan
zIMQo-saKCodw?|WGDg=>&p!>wFEaF<>7b=*%5oTC4zf?T20CAZIcyL1IB&$&d5Y!M
zp6cfi?5qWd_bm@wvR4p`f4ZMxd?iRJs{pZIJ8!UhMKa`K?cDqNid{wROB#0l@BSX@
zdiQp)y_b=Si_iC32YM(t8TXMryWjONK=g?(;_AY)cnmoV`UgfWWZ+cfrKQ3J^gY3J
zU1(~<KkJ<l&ccmq<Cno=TuL%Fu;d>mwUB>JR1oH+x~K7-Pxu7uwqB~-(V3`wd%1ye
z{9-iv=<6xnuczW-GsZ5M9dAR>$9>}`@CAcCStVfF|FUrQfA;DB@0F@AsE-}@2<Ac5
z+<)}|L_EU0|J4hBxX=$#WjQ{13R$YXnHtpQsoYN4-abEJPq^)~qr32i`v`Ij&Q_Bt
z74k>U9Jx&!Kp80ieYZJv{;&p=;Wb#(uxGh6u0x{Ltm2}?5qxP_es&_1D<O{%cDw&0
z;~p1fa}3;h?`UCDLK(MAoCRyU0*~bjyZTQ;LX}Dy`6a}v|0d)1fLW{E%UJ!7j5{wT
zS19Aw;u|0gSQRTjcNg4!(J*!QRs3u7viz*9wCGS?zpe^3B1HhDLY;)XbC9F31(<@z
zE>luUluMeGlkr4&5)oS>Y@y2x*nlsCwS~NMbCwW!ZX>*!i^vf2Hbi9brgHBe{$t?z
zKe7B<AKow<c5z&DG)T7EUFY{69kf!vKeqEHxQB%HEnz#YbAMZD_!rrg|69uIf?@nW
zrM%Zw-nL1+sXt5e2#)u-;e~+ftU|h?b5*>NPh-9*RM;hIsq(V}b;Fu<_+hIJ>TYep
zz)){#;oQ$1zc362*xk2W<mRHT+ff^Jtn)v`R{6gz1^->G6{<ueD)*$pgPAJ~i!i1b
zWI6S}sMdc!ze<1ee_}uU^FROHC18C$_3gmdQ)<;Gmmhv>@|$p_)MX+6^B<Ywrc`8a
z@0sma<v-l`S?R9B#4BNfoc#6`>nw?SkBLkFk#7BW=;d#wI^~Cu9Jx09tFNa>FEa0$
zTErdlL)!L=Hsof?Z|7!yJ+&U^dHjVH7jgGG@P4;QOmrM4A|*f^_hfP?9DRy)^h>zp
z62bXl;Q5q++In);5^>Ckc<q~!?Cuvwgn8)?6YqY_ma!PydFOAry#Hm{YI%3gDh>A;
zvtIllhF<$wVPULN$muC^Dy>!XU7qy3+gT%t4&ocW6fwar@o?ypg0w(q1@)e_u$_VV
z3$N#jwugaZS>-E--)cLyzwO#}If$oAlbyE72i)uf4!rJp=x7gsX*3*4jPn~Ra&~Xi
zLAAx)A3b|~cK8?<V5v!#x=>OBX|V=m7jPHSeQKpqmA|LBSr_WTWrW(?(jO_;AmaRo
zxp`6NfX*z}V~L^+oEu_cKc3MNI&IL<$qG>#)7mXOJvh@o!XF`n?W%eUj*Mhj&?5uF
zp1Nl{Uf~uN&kF;~)|z0CFw@Nm;nKits?HPway(QD>Fy_Kt(*IRMY=A-7T3NnS&L#>
ztjzgk1P~RId4po^3ZFI7)^RuHbUT4d-yZ(K$jzRb!^DV>l&p{7lm>ok`y~LoxTjZ^
zyV?67Ubwe=wyvqjWFcJbJiVnIh8%6ns)?Du9p8@8YEfpl{5*0jaV)hNir5bzxDIcI
z7xtqs>3%C>dhh+V8eb19o}5|*p-~nrj=j3o%gKS9@+za2wF0*W1ee(}s)bVLom`Fd
zYt;OF^Z2lpGVDahRN)#Ybf&17#i|0LLf1eGDUWS?Q$~_HRg;>S3FbyrG=Utz#I<2~
zF?i_(O_$iXBt?tKUS4cz#z9A>c`7Cbm3&G)y_+b|tO^!+d+pWp2b{l{T4^ubJ})GC
zUu&pwp+1t3FX@{W&8*vjMB3F+LFmpIS=~Xgf&tU-!`8m%X60r|xkO4+JY>nlK*Iza
z00qM2#Fqs`AXa#e&{V*x1WL^zP*>vYF-1d;cCp-B<SqvS?ju8|(3HU-4h6wuWWv{h
zChDSWnYnHcTMMw~hMl9g1H6*R?ym>8y19IadK2^S;g;w#Lt%=i#-Zj}W@6IGm5_Yf
zG)ns?h3w1E>2VU7{uzrM6_VIommYb0v+CMJmjp3iZ{H_thx%B-+}Bftucyj_leqH^
zQ&&-)#_wj=CutSaq#yy5Q0SDIPFMj}yQiJQFs2?bkWCd+J;YRO`SIO7TeYKRilw@`
zx|rldj)pPq!(QKT7)dg!SzWxY7>qS3u7CqemFbgW_I8v4IMT}YW=i0assk()twa+&
z=+9MWqLrTs4*c6AuWWooa%(t&G{=K)t3o%tmoP@_zxV>H=Dk^b>}a8p?oG9oD>Z>n
z{gjCe*6xg9jILz=2kSSFdnECd8O7$tGU(BhKZmR|M4n1-<)odTxdJpUG+2*kEVDv_
zU>KgYoEC%J)EvY@vI{-WIgM!R9KW+Iv@Oge2@BC76noIX8aMlK;Yyh>fzNRK?i3dm
z3>gMMsvUC6L5JB~Y>+b&aQ+7@9!`>8Hvy3*v}1$EGqWww@-S6S`Sla{S(~U9qn+P<
ztZ-T49)g`Yn}7?qaz>LjmIoOI{miP&7u@jjoBsB4H*cEz#FRRlnwOJuj1_drsiR={
zZt-bd3sM6qLxmPP{bh4K^%dBzK&TeeVavWtPWEig0y}tgkC6W9xAG)~H!^fGo~#JT
z%PS5Onvxi6dJmq2a<9n{17LLIOY1G>->WeCuX+tJH+jz9!_?CkeqRvPEM^HsFFosz
z1U>=a{p<T*xVFVte9191S9@nI8IW;U*wd5z)&djNg~$&2k<n&m1i$ENkj5olg=(7W
za2+M~uTk3RSyuSfhR-I{O1_J>C6udgm@r+wWArp|@5%O)W>n~!=K%8^-B~fgL=3?J
zj^V1-jSae+NUM4-lT=%R&EiBydz)AO-(5X5(YP?SrI?ekZ}>2ni;kiqiE5mZybN{a
zs{s`sCh0LX!8G3|OTW4(Ky&Pf%rI>8_PZ+Ih*v6A)ZQoiovWD-g1==O-CzB_cKL>>
z3D7v$UZK0V5ero{DEw>Zmn~;Q*2}k<;>NO1Bip<~)@39ty9eIu7TFdjdypit>H3iv
z7!6&5RRcFRWKb;us|MMm$nkJhRjyuBU2HRIW68zi2I9~|bUlXAvX?IQ`0=98c(8^|
z$!>4YZUao%EvWwl082u0RHyJHVe<v3?nb2Bn`5wQ4-~P=y(+cSV(&wVi5EBpIB>n)
zPo&5LBOaysdARov9n094%#gDB(&ftgf;DaR;Y!oetDwGNhQ`M7m(d;AilJ%jN>iAX
zk$v{_xHA&+8CxP<x(H{PIYnAzvX7_C2Dfi>dKy&8r)E{INC@m&d8=rOy(t1<UKSW1
zEO-)BdVFnXXx~^g@C^RU;c^Q`SY9Y@7A8_;FCNae9lu7%9d-!6<#S4R6tU!!s@bg*
z*QvxqXv-z5HnWe1R%O*k9qk*(^PSxyJ1<VAg*{4o;7r@y`^DI6?{@7ZW^omil>*$Y
zAJyxa#@)iW1e>NO<vk<ZeyJ=qZ}PH_Ev0!w(a7prbIN?WMIi|F__SA1>doa^fR*r*
zT{UywA#=q)L#y1=(k{@O!lA|K`LIC?9KFr1yC(XgK*tAV?(-qQloD=pZqy#S_Vnml
zP^yds*&Z|Y%N9%VQTOJdLD&Js-ePU)TFiq*BC&)Rw#ps{^}zx*_!~aQ;$0k=7ghkJ
z@K8Ez9X7#T>51*>=IRDDnRe%7L88*s&4f#tUF@ppCzmzzF>Ss?+i?q9|Hyev>M#zN
zj_y*599|j8brw(W+mW)Waa=e*yOP}4c29P2%kOF6&Z>#(@`<$9`h^GU1EWVk2FZ0&
zk8YVpV*S-!=`p_4m39!e;UHa)j7+09$Lj-7C}~YTq*Z!0Qoj-*P`MJ)uxHzZy4Yn<
zm>ybyp;}{KaZ6W<t<t`p;!=jH`2}%*eL~{2t>A&!+1kV455uUsj2qgY!41%R-A`m{
zrr&cOEl7WNd%K@B37C{M!3(yQfwcvCX~ST@%SxY&9`~zU4S14r`4Rd16mGDJ=zjUi
zQ2cyi$+YS76BPl%Ib^?v0N0B;2u|8Fb@(-4uQrVFK&5?j`}W$f1Zh+oc)TV!=R}Z@
zxHxD&a}Wan@rg8v)6Zrtc$ICWDSEW2oUk@m7s?b|sV?_mfThIs=w7K*I}lI}<`AeM
zi*%=hWn_t7w~}7PeIf8#mnm2ALix*n19ol{(oxBo4mWsKk-hmt0Zx@nVp&V>Zm#=<
zv3aWIUOAO``}g->|LC2oy&SQYxtz157#oPIi`tJOt&=?;PF}V9)e^?#O6hBRWS9E-
zu2%&%>27KscncMCC{r!^%Yje&5`HIFdUulE@1%iDt;a&@j{*%SrB$Gowbih&1z$HB
zVG5t3K0YL{&_qsI1@DU<zJBl2)qradv;~-aC!8##|9TLZ%{TdT<z*9lPtEb6YH02`
z>FS1?+!kAnW>F`HrGe`U#q5RVR!?wm%f!y}t(RRISI{~0@6k(Nrmm8#IEUjA_OyC|
zr&ef4Sc)I`T?h6Fd56U>2??ok`Bcmo*voc5nWX#OdA~ED%D8^NgSCy?VVIyayxo?Y
zUYc5!Z0P#s^NU@f;m6@Z&hC4_R+Rp4pB#kx^D{!{)QbGf(VL+&WY3aI$DAXO@Z0R>
zd=h~eqhs#HaT?-<Gv9e(%Mn)3N@+cp<$v+zCUVZ1@Z&32*_QzkQ@798>|sFiffL<g
z8T<@vdqKmaQa$ms&QZ?Q#2c>>y@|lqrWoLq*||ZFM0%plmRuAj<Uj?PmQ<(_wZ0*W
zR!B#-st~)R8yJ-W8FrjBmY<YX@L&UBY;`u(0pzqX@vaZAM8r}o2y?cwmdW@?hq=?=
z$&IgIk<a3@9Z@pj-m?><VFxOCnQQJTfOEPqiWgCZhV=6Cl0g&Pc9wT;8fF@H_utPh
zTO`LO4Uzb%c6LARpgs=gy0Uj-F2>DZjDxMdp3<<(WnU5_{JI{ua`vhEv8VgSt!alh
zZQzq}_jQ*sVNmYZQ$op?H>r56coa1}L?cXa<i;iCD&ss4mXBQz>UP_z1wHR$Q+I~k
zKK2}13BDXterX7JzJTfJ5JFx2t|0FJhW}*zeolHL!^`>wJS03a+ZWXEINOTwXIWO$
zFV}x=>DjuMo@=2wWHT(&fJMC8Ij-L7e-aQaH)8!>ci217$H60hR)IRZR7tj=C=oPV
z`v*FPb)uUZO8f}+H2lgzP7S7ViB!`}q=Uo3?pq|B!`TyZt&#i8LHsq}bt7n?+*MKw
zw~Gc}cLD*4^9*eDbWOTVl*2)*Qc}_%c~SX(KMq~A$^Yn-zzGg*E|||y^Lh32_Z?wf
zfsoPC8|6y(rC*h5-N=%v=uORs=dO6*iO08eh_~=WBB`B0un=X50Kkr`xLPGV)?3I3
zERvVk&lg-a>iFrvc)+sbop6Qj#Zql;ZoOB))WFFOc+JY(IwSK(a?kqn@CIjn%isK~
z_}w+eTA65*h8liV%HdCzF9S&Dl0wQlI*3~Zu#-=~&~$!u;9CBE9n2(tA%f`f;lONT
z<ycNO-9#et{)ct-i&=A)M?L-3uDsman%lL4#T9G>&agk+9sikJ7w0~}4u=}ry;mB&
zHQ2$bC!wocXkY`6o0{nju*_r4jl#GPWMFQn4oIFJl1{K#%@Ht3pEtVm2`n8Zz&<%p
zPIh<gT^*3((M!ax<`tyAC<YB}noeA7S>E0IA>JSO?d$w`(NMR)0U9<+xGu!9mflve
zL4fOZaOL{b!R&Xbh=Sk6&7`}-I9+WYW8x8Y;JrUv7B_QQha+Yk<P8E+ExHrhGk1~K
zQ~@D{B0$_ir<(Z60JJ(cZPfr8<**v2dZ6k6#vm>8(GqfU>bSW3-F+R^iXzv6NR2al
zw$yUZ_Kx8v^+B8<wxmHqAzYI(*^9?AK(3dO3R-JS)Meq>IYptPOel-=8jC)i>jQJ^
zIB>A0;O`(ih~fiU8;8+>oCS(IRYNAh-nI+C{UYbzKAS&(#4Bm9#a0Y?BSk1^9o4LZ
z1P5sfr9{O|;^5)YPm>3+<+E*N>YY7x@%a6==V0j*b#<8*BX{Pbblr2V!=L4dw96pr
z1V$MQN+Ys#SgXM+5uXBaP5DtYQf<q3Q`qc7>1if=l_)r-MJQgR=PE4ryL^gW2ii)2
zp-Y1bpk_bWs3NSEGF1jhhTwICQNIC~GUO!GI)#bpliUS253cuKO~?h))DUl1&&Nrl
zi?t!fw>#Zz&?4$ZbRfuL{jZPP%H<tziv5$RHA|M6Cfe_TmpTi_<KBP!p5CEGn{7}K
z&!3w~$=TcpjLzA+sP1G7tjjOqjG<vvu(Ll-e-d(jY2`2w!5yfJyL$BsNy<6My(u#y
z4QFW1TG>kMJVqP3HAS3Ead)8viPJcm+OUUO6|ts%%_YT;$doD~?nQ(<kxH7^yCPO7
zc@%5~PUhni+sPBsGSYmS7v^_`yFPmxqA#VM3>VL;tvdUz>a$eJagm1G3)6>q4z;?m
zA@hnTw`t(sJ7uSNXE9*S$GcW;uT8!?!0dDDT7DVfZyCk}n%s;GXhaEC+zx{_9?XZ#
zO>vv4tC7tOPPxHA<?2g{XM{1I`}>l++>uWK_?Bw}hIya)%l){zb~SRP|L6H%MndY1
zQ5Tu6(w4Ega(5}(AuGfsKuZeQM#wEdvOr?f6oq4omw;41ZjO>P@%_HZ3->|eUpZ#9
zlOe?u=Gea)+VH@rF{4v?Ur$Z=CClxQU-ZPLo9;Gj3Ws;gar+E#GT@+HQ39zwnZVgC
zwa+CSnoWNCLbbr_v@i-nFB~4Y#fkhlP-dPFbcc^=8{Zjqd-bRZAtn3NDS&NX<6c|n
zxYnbgaZ$4xBaS({T%;TXs<vI&s_tIJTtq66Y?D1I<7{d4`c=+!B62rXJlNO$eZ%xb
zJIN%?kQ#GF1GRK8P_Zi&=2BWUw(c;On5yp<=QLEHR!t>nu?OPHEmAaWaj|LcP55R$
ze#&?0BQI4uw0ITq0Wvou;1DZuEkf6p-<fNlZD0W*0kPaCjTvlXs4Mn^tCLy=-}hR!
zR>Is=*N{W7@&RY?92uZg(zoCPX35;;HdNQ7D)<B#Txd8}iP39VTC1V^p(6>D=~At<
zD#n=*|ASTn6>u&D3if!*p!uC0EKE$!vIyr??-~oE8@oh6E>U_kgeyH*V`)En)(Khc
zQFuVR(}_ng#kV(WL#p87@Vq3vK)<ztG>){_6f^f>$MR4j5^7U;VkJ>q@B39G`ZbQi
z%RLv=0kAFo0C9d?1Z&pL6{|+y$rzK!ADJ7!y;c7zxvC};enH&Ks;Ir<+0V^<Yvri~
zYhifg&jnuu!7GvzZs<<AGe*)o^kE$-&S7SxH-dL2#=p(%z{G;H%R6W4^0tj=N2-A`
z?GSte*lSww(gK8}^02FqrREG4*BOWo$rR?sd<sMKH?`>Z4Wyf$16Av4-0#~Jb_OK#
zg;o&zOc`(&P`hfT4rf7BPK`jiwK-aVXDH>lm7a&hLiMkw<VWK+Wfrfua%K()Jl@w+
zra}bi%uH#k@&AvD^6*%^AoFf`=1-Z~aE~IV#i)NnDt6LbZE9MRR>MPXF!N=lbX^0_
z>c6gZ5|O!C`7(^Td!<09{D*r`473#ESF#J#!FZ>bcM{1}4>;}D^dnH06*`*o6L~7o
zvX#N|7^X~4vsz=q7ogSq73o)xU)~W`ME@5{D7ZWGkLhhRfQ|-yMn-<;LG)^hJ(u)|
z>#SMuYh(P9Ms&*E*w;&V%b`#si0VQA_cqQ0@mlz>&wXYrFJ|I}%cX7F+{`<2pv-?g
zmEXLZdHx7IHHp)Py#FlQRPor?olAcSWl_@&A-%LhBweCmEW2vDyT7ZqDcf7V8wpfv
zq6W(fQa|(Q5xjKQt>^7LUgKxSReId%X`y;OJ{cxxWVUxQ(&NO17cOVp2upw%x5+}s
z*&>2Op;S_^D<w*JBEioOqS7d%&dHeCnbGW>pVRj^Q*scVyx>xoX#b=IGnXhqCcnXh
z$y#K1kZrRKTK=;y;?=U20i{-zF!|`3bGFH!IKnT}CkFeBmWYNhI4(FWGM!EpoU>00
z8n`t$LInkDYoF6t#wM63MCgvF3Ss1v16|$kXlxdSNzVdnzA@DOR^tIDI@o52%GZ9Q
z1J%8x-cv<GvwghhIJAhFsyb*tc*og>B_zj$*<E8fc6K%yPE~(TQE3{S4?ooHtoN30
z?Lu8Zux$uZE$I}9w!Q2}*<NkZD%=gGTt$U6w3Ab%C$q|_P?Ls#_uh-rEWev_Xq_=b
zFS>vUK9CY*es842b$i)qtavE|v2h^+7+*c-zWFf@=H4|{UszrK0iMTRoC~TdaBp7o
z^PZsz?})aR#$WKriM9w{*c9mzO*h^{(Qh5pQTV)M!M&Y1w(H(vm_=URn_E)1jI~(9
z%<-6yczyCD@UWwOXo?D<2wNfS#$~jJxq$#f62nY|?%+=_<?0%q_S;(&{xnSNIkjO|
zk_wJ_b;hp)?sI?8EMI1MHDBDlV&v6iOJ^<zBIEgSs1xA;8aui>gBeT56r^8wo^~sB
z9I3rkhnB@1AY*BTxz8~L(R)=hHK3`$hQL(aDh>m#oT3h)LAA}1CaGR0>gi>3!d9}+
zd_bPK)-7x$#lOw(3Y%Wf3<HD3n7T?t1tOe?ZCt|Q6>fGFq2KnT#<vFNbDfr*Kw7Os
z%Dq89J+N%`PxE;Y51jVNs}+$t_ISL!5|aoCS2gNaug`vo)Z&bkNi2U1JcvBFEF3f-
zKlnwyEQ2uH-`5w|^rL2rlhCkfC*Wyg+}@Iaksiu>S0WH&>8g!G#RkG1gkpW|a;cLm
zk(JSgz}-R-1(I{<Hqwm<;ECtq!lx-jMLZdzoDq;$q>zUdy8ep>d25>>Se!^rb)ZSw
z!g$T(6KvYU`PYH376a3M{yob6(TF3#2B+xAce`I~Gj${PUM&30c8Fkgq6r<C)I}q?
zylV>eCsw3kP)c-YEJX>sBtHd$YTB?(?CqRkThNl`8g(vSzsbos%)w4mjs{%H%Ya2@
z95>}B?S1M<j?wxc=xRuO6n3yZV-^ZIzy{5WQ0t2rm|^|&b_%8SnN|?BGo4iHV{Ps~
z(7au$DQh<L62`S@rrtbzd~?IpcIGa^*$>Iow%NLVpI!+!^z-<ewZ@vQPI6MF7v`5Q
zde6C~jz<7UPVlDK#bz%7H;2r{Hj^kqZcF-4BS$y4?BpJ?HfDyk{OkNouL$@A_hDiR
z!mXDE?ZaaMa3U#1DSrizB@(YE;r)^#9#>;MDPLmaS8Wy3mV%}hqAXv+frh`>w*Ta0
z9Ee6IPK~jh>U{NoeC$j&2+NVD8Tc&Qd#R1YnpX&bw5%_j%{u3i2BD*U`r7Kj?m<YH
z@ugVA1y9#^)v_mG$pfeSMx8aD<+01eSg4j(KwTnix39{3qAhmE0Vr^73MKlJ!T?7>
zY0NW_&dw5~?_z*-4iIn|ssoCW*xZ2Bn7dPmw@TnmLT}u-Fspe;sI$JTeJxSo{YRWw
zq3P86RNcUkWvwdu$M*2~d!cKDpI%a=0pZe_SG#w=m2GNij4i4IBUH(GRaNoI-c;&t
zgEm1|#@_3f+@HrLP&3unVswo!p#6St)es=6=VY@Bn`nyhLSfb&Zezre!AzrW=q}G2
z4(CK<r3dS3`_4U{9gj0r=pu&J*4EV2@G0suy>-I$Plt-D4(&rCVS?d?-Sfv+<;N~w
zY%5)Y@d^X)(Z#nksn?<>brsm;M~nSg7FZP`f=h#jcY<{6!49HB;KQL|L>E7HRQCr$
z-bI_VYQVq_L8GVUY3TJ`y(1i*q)aP)OYvHt<zSj=zDD&8$#nvixi@|+jKi(#_L1Jp
zE6u>McU#<c1g{09wsPo<3?{dc*3Yjtx?<8>Kq}`9cvJnmA3k_3EiJN;OaK_D8KR}(
zo;gv%4!65dJ>4k9p^QvHBJ$zEaH(cnn28Kg*b4v;q&5)oyvvx~{s{5_Tk|rs|CNIb
z9-gtiWc~Q5vR=u`#mz8@M^u#u8k=hpKQgldU-*0q$K9*-uV*iU{bU}^ye=K(a-(%b
zwxxobK8y}VW8cMC{c;u(^LnRzr+>$C>)YYaUB>w}y6d@>>&I^!KAySzcikJ^?p(Br
z7HV_Dq8V_(+V8`!a|4sq2ue%AK=f1d;Z*~?63GR4DZ%6V@?lRfAJdQ(+l@;slnC7b
zR|U9=a;ba@iwz8dJ5#1Ih?U9sRPa!2<eUOjZ0j0#>EkZV7rR)_aR`kX{L42xvfwFN
z?3uS&@&FA(e^{fEPp<pZM_ZGRnNN3y8o+abb5}ggeY*l|dt)lrzlbeAYi)2&D9W0a
z3XZ!BxC*fyc#ooTo2N_TK%;<Em$5ovQ5(UF?NM0%E8q2tChFlpCslNQ))VItsy;3%
zyxR5w4Lj2n9fh%%^m6|O<=)Y4ftN!nEOjr0NFiY|WLC2zw$&%<f>VL(U~MQz-b*2G
zsWJujUF72@Z|8jf;!cdHIkJxdbk?y+k33w(IUkxp39$9PNJT@u^B%HqPDNiG4Z<#=
z(QcD>e&dv|`fp`CFmE~|tAqBb<2&PN<>MvQDY5C?L+^dRRvlg-CQ$0C<<d)n_ZswL
z39ro-#{iyGRxk~6s_W}p&dHKM9sLYPPiCENB@KwM=pVmDMg40ZIy;nPaQgG`E)@Ns
zExml+zS7cyzl+dN#-ZY@2YTw5X$UL%-5B&b2XCO}%+h8%4%>k20Y>1xcXYM4YacaO
z#*#TpB_AVPk!QTmNL10g-!wgX?dW5VpS%y+vkOQaeUlgRkT%xQATYeC1k4NgED9<v
zFnKO}b00mcQKtTkVf2W-=8rSI^b1ci3X@=&MO@7xFT$lQ)`G602+JHKsW23|00vzS
zO*2@gYL5%O$vuv+HgO%}z>uIRhQ^)lY83~1{<BaXRx4yn85a?%)D#gD0XmA-*%=nm
z7Ku=(R`&9%RnW8A<jfh;)HvcLtFhk&;~VXwRR<G?Z{jNhGD0K;f*Zr?-^CR~Qev_>
z;V-f-KFHD@q%39FR202<*F355lwo4593y4)cvCzg@sGnHyK%JVmjGJ0-Cy28<e9&C
z$8Pa{js)Om)-I#E>~R>EB;N@tE<fN31p#z$7a(lT?)#^vGtq(QliDeW?;6N-*8^!v
zZW79UUmYA$U2u*E1@1?<7PhE!S8R1jks(_|bKu2RD#vL{$itno`q%qEdfDf@X>MEY
zln7(+i#Zq3youddu?ngKA{&+K;RVR7lWkuo%9F5l5rO>$QwgsP2JT<7VpR{X&V7uJ
zJs-)fUT>4w9YA~=1d9x$`0GVALBy$zYVKpTpJj&!ySme4&fOnFsf~Mx<0{+_K#}#{
z!{S65^g!K%YB7*KuRb+(wK(wc%3-j0%fKOKNdemLw7vBNuZj4;ccM7!hHh(;*_R`Z
z^Kc=df%!z`6a^s$aCq?ik<orUBJ+uKe%|V|TtImvb3KFE4Ob^q6>YWm6sPIf*|Bss
z%{d7+`Nuwz(pvkpwpiic{>kRx9(|vk1lMR-+`Zzmo>D@3tmqn(@nL=a?UeLh?wq)=
zZ{{_(b}UR2sVj5iMkk(bNvWj@XeIL=cx_aOT8!+N1^1J#{F*B+V%^yy{X=W_G~Mk9
zKZ>4fOBc>OOPy1_7Ma#|%oJfx+1WLF@h$CIih~J}-Pg=7!J5JX_(ChTf8%gaPqg7*
zBMhmY`m_ILI5u|!PV?tZnh6^1JI5Silgr$!6?SK;<9bKQ1aM6VfWIPIS$oxfmz0t4
z&Q|>L@wI8-zV3&dFIy5$TQ8r?2?Q!`91OCsbf%Bnyf;U<{;;OX;U$JGt|KD>VPt~!
zh5^jgy($RN)C)&-KGY1`AE(lQD<fg<h`M-`?sW<Kwm5~<sK!VK3v3OZ2LMIoyVN@i
zOI|DsF!<Uh&C9&RedQqSSgu<i?O%@@&qs7Tcns4!Dema@;eMPr{TH4O24pzy;#9R0
zeJ4|?@ITmp+$2v5?jY^&KadRBz0HU8nwG_lT>abbnvUt&5F<&Rq+6;=Fq~Vb=vgWP
zUxzm7H6CQh#s}IF7cnO@!?*fVX_{J<_yYL0avgsO$wJ~@fnrR5chg1-+{qm%)XL{4
zLRQMk=wl|+?2O@<Gi(X!Typ)At}<eW6*@IAE^6W4=>P~Zie3jsOU@uyhJrrQ`jif8
zmPM(#!y%)7da^}P9m@qf6%~f1!#U0gQClAguU9Fg!5O2LjZaO3tLr9$lt%|`!_@iD
z&tgya2wOlb%cTx(+oJhC`UG^W5x8#rPQSP^uD^O*w)&{}eR$jJ7`fes?D3tp(wOVx
zEWR<ig28V(3Z6Wy3m=^Ldg|sTa_RWn-60mkBUbzp@xYm1U5eTnXt8q}--$lLw7Cus
z8;4jyWt<hiL%i5X4a~#0jU&IoBVREGv{b)AYD>@Pyxsl5Su<cYG(mOd@sXk80Jh3T
zg3y5g*+zW(BIB-8M{s$f=(yd1)lvASL(c_tdsW+Mw5-B>7q{7#-#t$1<hngaS^_O!
z<4T}^m(VdMi6!INx`;U+CScw^%Yu@OUvye(mTx1xRU<S<;1xbln>Hzrrv~1!Ry$Mk
zC!$(wQ-QUW3{AbLDEnYM(k=NPheJ_7a#GRlacOnAqdVwdhzEc_m&Mh5iLV+5i>4er
zCms2+MFEH55ie@3{4{@<a82+_eDEe@B1Ms<e8V-1`=GhJHPM=+ntY(oxPB~WbY-Gd
zK0sCr*&@RruDa3eM#G||+^oZp%C$O&2F5Lc<Nm$xBi5;Yve-~Bf1LA%E8VdN3gnmW
z73!wDDMld+-7Ma54PvGSo<a?F)35@*o}|46zZSNr!<5Qzq>gJ_#S)c>dE3~MXkhMh
z;1I32L9GoAlc^3_SV7uFS2@(ev2H=6OG7ioch|)dz`FiC{LgWDcM?X{R1`aJUGeMA
z$?kHGH?!7~idy^m!2HN(Gqbx*Y!5cKiT3E4GB}+(yADjP^z~Ynp5~Pd)&z$lRry=|
z&hh$~s|vgJ@wi0Pig*oWe!xA|9)omiLZ*`5#?&?Lwem(cY3erzjQS7Q10?Zvm=}!c
z!g?MLS4qvw*LPaZbc632i#daycCuMr_S&UJUXh_&qEoHQTkc;^-F=dZu<Ic0*ok$)
z761?&zERZw%Y|6)3MUjC49fZFl8)Q3d=~Dk<fP_gcwqsAG<1MsR1>_{W|sGMUhf2L
zdcT!H!6O+hw{<QI<<8kQQM_h-gLspT>dqyL0v;hg^`%s^?Bv*O!LSFcu)-}3zFpZ{
zLIS%}y2RHx6hZR5Li^0s;&3nc)nE`K|5>n^aJn5aO&~mGNjZHAze>Bc>l(E`4)fw2
z0>ZnRO1vuYD*5;d`~y8Jfj2rjl=Dbu)vYiZgz_347~I(+Ss%ISy#kA8Ef0SaSDZ)g
zc6LPcIo|)l+|Kxc4a(PY2X?<xdS`|;-(`CK=+`Yid1lt6E|VhApTwc8)7B6Y&hHVi
zwe9pucD|qRFyxlOcHL6<6oO+!*h0lWEusleDafC__>3Z8lu9FKyPA?KPU@I#d{STc
z0BID8?}QLFyk_1}XG$nayM>Nh(|JpmRl3{~g0G~%-H_89WlFN>g%*FWZP}b{y=d=7
zJaCIQ&jyxWO+MXca7!!ZZkJ%fQ6EGxn8Ccx5vBvKUTozgcg)!RQf#eZG4n%=Y{XvQ
zmR)oJM`5#l9Bu35QJh`#IYYR=)L&P5_Wex>zi<`KZ<LT<Hx3&Y7c4nLD~eKIxlt7I
zIz3eOquOZVp9a`E7u5HhaA=Wlc=S<tGwh~z=paCRy{GcssOCCLXsxwjR;0A7<Y%mq
zW)o>LqyE5mJE_)y*msz?3cV?YMd&AynueabAHtL9qifv<m9`6dyaa2I5y4^d?>9ju
zP;0&PK6Cx=WNpx;+Dk$~cXg8Cr7qL^YF|&aKGk8XuDWTs)YRV^0jS2P*bgH-NuWu$
zXzNk0bX{lG4&!c5<>+ReDqD9msEbc#t~!J=`_X~SRtbMj>$6&&7!_!Z6R<Mfqg)>t
z*jtkgz8vSs?5$%S1x|_w@@pb^_N<6Xyz(+hC@Ri-W^&y~O|K{rwnH6q)FJ&}?7in#
zlll6$?TmYLj|D~%Q0j<^(xprPjWiVkVhE9vpmY)vq&I;*TagZfNJ}UqkpKw<DItVV
zMj-Tn7!pbnkX}QP1duk*HTQa+b^iy?x?kM9g1jLs_4=OY=W`r3HQH&XRVnljo>WKK
z=U)Pz+@gu^izwM^vV)FZJ*izw8SSYjVK0J)@UP4Bu03%vuQDA>jI{+q@tp$D*4W5E
zr&`vm;vrL~Xz9<d$lS12!`+Ul=2k9}8bgGSYWmkWeA&$?`#neZtM@D~rM5nhX#tfW
zV$^~6Y1-*o>o;u*NGu1&3WX`0SQX^ONL0CwA`YB%b(w+oUG>4SR(53`!Zr<8sx3Xd
z0`T%CVSE&$)yLrpXz+u^_&+hoN`oJsPq44Q)`z^E6Z3g8UsqWLMtc7`YZhy*@D0!K
z>f{^nenBm1_jUYcin#x_<odCaeg<><^q2E=lZDz4J|f2Y1=?k&LH7R88FUU(<=l&p
zd|(fCY$JNwNWug~e0NY)XdF}61GN^n80uM<)pr=#MgNXRunD!5&%e?O2Xx`&EC{VO
znh(>CrVp@5aSX>{h$OOnV%S&na9~E4ZU&~nnqcOCH6;Gp|DtNxDa1HcvUe5<&-C0T
zRR(O_eE7Q;{Z?BAa}q*UODG$?UjqFR{4dh05oc$q^ONq@fgpQxUiMe$YmRl~{_jAD
zl86p4p?*MLke-s;Y_(ccCnl6n)E9WPtjl5x67wLmQI_pA9pIaVWDubw2Tkx({qls9
zTL^4@5W>(WGb_fP!P*IavB9ib(enp3MnZ@SNewp2hfx416J1hm1Y*3c(={wo6<+%x
z=0D8iTGYVl!8&#=xBcAb*y~-vRdJ7q=noF_`BNpHqkHc*xlXc0sCr@UA7%H$``h}v
zFD!hlbhvHvoIP(a6Okcxl0Y*?Y{!T4UlOBezQOyr*)#2w5VhI)ubhF;<e~{8!q5uR
zqB{5znAJrHjtXNMW*0o>yylp0F725<d~w%u<St1LUPRuVDk(vue4i~#naZIEB!_U+
z!kj-a)Ic#f5H2Oy56JWaA$pZkYBB+!HhEOe>%;H;7UVfV?s2Yb3AEI*NsGA6inZwV
zLrsjqB&&Sbxoa9tb;MdkU3k?eES$Tn_4FNA>5upQNr{h)lWM2#Pg=JZOGXDDNbvtY
zb>YbHbJ+By&_Y<C<v7jy0b7%}H66bI%_OYQ;_W*}RiQcQb-|k76xlE=d$m0?`WVXu
z`nb^xK&u)ka>BQ8^b3p#mw~R4V08~~#AJq*ZqPr2$(ES!d87%_of-h5p+F1+Yg&e{
z5Q-xSz^nX5JxGVi2q<Nyn|){;!u>@q?+AKCd>~?-tk&x!E_kc}N}>jhXUjdAd+6N_
zxt=Uq_w5g~&Q8S>r<$R*ouLkv{+qQ?0VgZQZZtxjR`h!`HbRvdabCl*zn4x0VMkqP
z2$kAko&!)th*0Yij9H3XFr{&L7n&!Cx$chlZ}mNU88BE@5d3L@ay&^TOeLn0;L()7
zil-*e<ridZ++2DQFuS|~*80@?@4mc?1K-7@-?|)^{`{zv>OpzCtrgvt_asNltF2=4
z_P;a#M7>QdHG9}JRb2e~ieKb@&VJ4ZaE?_DQjTDT)A(i4zqecj3jvQwOZXL}W@)m-
z1;Zs!*~#BRI?Tio{iJ!nC=;<PNDy?5Cz+8<)wK`|RS)OIDH`ks4NwC1PO-6Rk~$St
z!#6;+4PD<1mwy?NP=yR&^@_dxYO`+buWQTNx2>UW?Ui`r)4k?08ed}ucA(@CO`YT6
z#+{uF%N7~g80*17n%){PE(#)Fh?$U7<Pp@}s8^gzoAEqa*qEC&=nBNK%Wi8`Vgvry
zP`!^Lk@DgjY7(!ggdw@r24j>`2pIol4E{y|LBV*@!>EZeVkgRVu0?r$al`&PBJ-Az
zC_6eRvb!<Dj~KB5e3r^u1mby>vnn&IqkSRRM4;O%DWYl+Tz0G8JAbof))l%u)|F`&
zf4?=Pv-Plax$VY^fmC-*y}n)Ehq;j%O42>n^r-NnsFB!@41;UZiu6>m5M?1jL3qdx
z;9yE=d8iQ(#45+HSZx1&a<%_8XBTzLtK*aSGgg-IH99d_mZ=8eBsjqj2O%}j0+~dU
zf2zMhbRD4VyL*vZq&UXiL>>OOLsO=ecvPb`A`7K!rrbG+aD33*8!RJoP?}qoij+4(
z5~k|v3Z$f+wAY`Rzn-HjI*wQAP&h}2k=mqXntQkCIeqo{KAb2THGR2JomR8J?d~0<
zq?WE2;cSLk!N|_8+MJv`%d`tIdx-At?%v+^D-5E9h8qE+%!Lgn=TxAoJ72xB;4v7V
zg#orkq1@6@Zhzljz4e(EFe;1|zDOgF*t$`JQ|>Gg2rGESiupBP?yddG71C<>c7*{u
zbN*`gwS+Z^lKBZcOAS|@@B`gj;(8BZhC?F<E=5l$^<=~qD_HD{Um@c*@7_tPTEx<C
zkz-m{ep;znj%aDQxE<)v7yTu)<(S!YWWtM9pY8zCE{Dtw@705h%)Hd!Gp)ko20J^u
z2pgb*P%A4ZxJfEl6UJdLrx4N~O-xBf=HSt-)QVLXQ_I2&vcC5HgES}d2RMF8w-6hl
z0R=kHMI4F(MyUX!#vwH~&0>6F>>p#7Kl6K6hemV{L4qX(JNH5#fXD|*o4dOJ_Iyi1
z$;T;AzNRm5QyyXHO}T7u<*<e`mqxT4Ul(M_C%2`=&=Bip@<+;3MstR2Y(wMjdgo}^
z`r-Js=AtT?DT^rJ(%@%`l8dl8mmr}%gznB>8q}nESgiX?V5`w%)^6NarVbI}J{qEP
zn5(G<4riqRFB5{IQtBAVnT+;Ph)2-gfyID|n#|PJ_RRCHzfW226})#TA-Fily#D$$
z*E~x5M!zpo?ya|`((FKI-J{;k*37wsygytgZcXfXSlT?UT)H<|Vw=$T#wBd=_>G`Z
zKo>mPmLD1GI?kTiJ!0y@=4w`?(0YmG<Ksrt4_mXdJYOt-r&Qn^SbA@YUZv{H%z1G!
z-?Rh3jt+V&1Y*vJrOU+3*qdSnKP>Y%mTr;aiD^M%R$L2>wI<~ROnKoFAdF(b_-GBB
z%HZU#%%(yHZH2ZL7K5Qbbd900Wn-u;Bs|AoytJNg-fs^&7M+Fc=AXy|h({CGzrGov
zi3+_-b3ktI(?bVuucdI_{f)-=qziG&c}hu{-))xN+jlh*rru6TZPr{S8yTARfByT_
zfc%(Cu2o*enqE-5^SUe(x5A4?Xa8Qgdd-yPG<sJREO;q>f+sf2!-YVrwl(nCk==+8
z74M0}*d=;Y6@fK`fg=G9aouW=gLHTMzIOUiDqD4gndu`884FJ$C}ETUg1mqJ|IB!I
z-cE*3?-l%KHLlK|!7hI0xW5~JZdj>~RirvNdp{>A(bb84xA}svgH}w`Bdn?UrN?Vm
zHYwuh?9HamJFa~K3N!Hwa|B7xv_`S&UTREDhLc;uF1jP;zxwZU<fHuZ&~q!CSEuz&
zn?`lhE6bj_q&^Kta|x-`##P!C2F;NR&?F(RjFq=G<N5rR@>R@kPQ0XTeP)GgYjmrt
zZGS0d^EsU2qu8zYJ}VmVNCL1HJwl3)2LYd-oS(O$Chf}Wo^4EhUM{70*qmKxQt=ct
z@Ra4ftMU2~3<hpg{)mB6L2W9IuPtmCQnShvS~mr~1sQvm7A0n_r(Zta8tagkuhq{?
znd~-AY^X+=^p!(PvT=w>e#NME^(;66xh$n#KwJVrl_|mbw5wpmwAh-ugP6>shNO;c
zNHFR=20hhXqv1dVpgSN;Fffx8>fz_P#4&I8Zt^8thzwKTkg&AtPt3TK_~|fF@=`2$
zgRu8FjCrP!DWe6?@NOCmNErh+3>&2hg&8)lhh8~3HWf$YIXBMd=QwGJnS}Yh>F`iX
z^wF8mb&Uy&e;{8bkF!r}JF|kd?=YDwU*nX^cd4Cq>FuCTU@3uI+=I>$in==7VhM@$
zwCW^bqYTC!?Mw_H?o@q+W>u&qF~^lYC9*EN)2I(Swhmg!`?#etQm^Sx3vp!DD-~i3
z?&+(Itk!COLw7MGNW*TEDL3#Qh-897#AX`lGo^sq+O)EG>CdW%EP!<E>K{yeWLSv%
z<*%E+^^|_aymC<B4FohF{!;U9zzM4Rf(4W7wiuJaaNpf`z2s#k4{5iKNHZ0SRGB-&
z=eB7gn>g`k1YBy>MXn?-?@we*3Xg|4S)PpA4XW(?+>q|8%nJLuGMO;mM@tWs%O^t!
zHOyc>5vp}I?GPD+PN$SV<5$@f6g27%T#KmY4AMI=L4JRYs*bsF2m0s~M@tUdA3(Mc
z7b^&NF#ZHjho-w*hqFI7Ui4ydvF<@eCMxgybcSTb#mgE?SEl|X`=Z%k=V$Y0{(Sx&
z|6Kiia-!_YSuM8*S?_Wg@A|+*n@7X#AKbs#j4X9ei|U<8{W`V&iEG%M+43z6<SH-b
zw~<bRr0cbcYF|zjdLcO>mbF>73ox>Ub8WP698)^aNl`$inty_2wd^*Mi3@0y;f9=^
zU7nSxNrbM<g>ll6_as$^ezkxK88Ibc$_dT><rG=BL35++iV3@YealzNmUjPfvSqo-
zSgSBpgCFD=<cgIKlqdK)Je2xZ-iO6&dNTC4-e#B0A9$SKDat@eXX^CJ+#gE?^w^Ra
z@bi(^sV$F(x3NL9sCBjY`<#4?IJcG3q%Y&O@EjqPv(5>1`K$v%EUI$q<LKZJMP7=!
zr30vWjew~vZ`-a%pwn5e{+!_4gKDE^78bb!_^EI?Ng^ejk%F>i738FXh6oizpnOg(
z&+t)DR1C=Y`A~fN4cixO$@}zp+}}K^1$4{m4(;XIkE2?xaKn=i%6v`cn}q@l!LC+6
zM{#Bx-&7tE^DrfN=;nVjgX7+sJ&^zLWe|NUdiuYIC89yl#=y<~JsA&FY+vI^vQ<r>
zZ+DLVZTHVQ4+e!oe@q(qYJH9%9;_X$dawo#ncXt?Me`~nf`^v>q9t4#^}m#`ro9!?
z>NN+cw6ngHT#&PRpm1Cq>uc@q7%B{9F=vCS$KM?5+r(0mW8!+Mbm4lE;L5CTc?9jj
zv-aQz$LoXFC*PC=Y7V@hPZsX3Ryj_taX#8=B{lT9dA|K*CWuR)eqjH@ktoyHt#;0I
zxr0O5lZzW@rBn_~a<YSlVAUG(Ar6&e=+@K!!(QnU>MnhyJ1*GiLR-B+`7{X<sM#1b
zL8FDuqb7H|f8tKwty&FT1htx<J<nprwOp$eyH=22A3{*&&7$Z**0w^BiNg@+E2WOV
zfr_!yZOnEj;{73w5bT^@lR!G4^W5lJHR9LL1Ig;HPqvQoG&9|jzPXtzr|VhVE`HL-
zo(TF`Hz)7F+W5tmZ#REthE)EQJ=?`K3{PkzSIHuF)wyH43s)zoE@f$H2o-XTU$feJ
z0f{aSk<{e7Qv!jC=nBE(cKsF@u@OteuQlrHi4UqwM2(Nt`ctuq*fGSENvIo1xwF3?
zpP`bUjBl<Wyqs0Mk=wM5o!RV?51Cvli}Lw>Rxi3dWA4E2DWB15hg%ss#RwbmxpphB
zh~%{Jy-{8d+1hESO5&%UaYVmb@oa8p7kamRIxuX#6)Srrwl(l_*;I3bE8AYZ%Gn5u
z)sR;4Ll@Cu$0|o9BU{I@28JSBcI`S?&=LUPIE9B$YAHGzP!*Agf!#%!t@^+k&zoOu
zxa*^6nss$ToHN>Rr0n0DHi&H%3)C=@XkKD*OQTmx2{GFMQp;-j1!~gto8Ca9>gMUO
z-1(JocT~+aQ<!Te$7??$R?$b_%)X%wBio^_Da~ze)YekFqkdhHe5j~c*SvSj;HO4K
ztE}6=T<}4B>eQ8e=A6?w&s216obxK+Yb30cf3!XbJL|$>^N~%o`WS@4z4`>wij*E~
za+xWFX{|4~(y8e%{YtWxB{Tp|K?t`zQ5&)~<Q>8BKr&n|3{W3yv4{nc)Ox2|3?rLb
zbQ3K(I|rSQ9p76KhPkE8`eE=%_+|vbD6-`RZ&sgrajbLHMC6m2Q2RXc(cPn{!HjxG
zoq?ayQVE;@g2gKHL;nufj$zm3(IJ+K!H_Lqeki;6D9YCGP2=3>vtO_yC7*_W3=Yps
zjTx~2907^he6cEc?zP^^tKk-@=%2n{pBTR~sW{HQTZ*OB23Mwv<w;gE#312*_yuZ>
z%>Jj3`s@W>1Hvf6wl%N3J3BF5-;brkH3SH&T;OEB8eG3k+PT6l^=XGPp(cca<XOl4
z$Ea7~w5W*_GdzZ+#Je~2-sd}`s>)+%?;4{_(;elPEdGTRlRCMet6u7P6RltrzF~4T
zi4&VEDLid;Y-9}DnxhXtP;-sihZ4uf^fK%wIrr{;-=8!kzIUn7KYfO1z@D>bhca#(
zCX>TG3aD^dy^UeMD>b-yy=F!)p_<Pq%EY5DAhvZ?q9*o`^>8~=m;*vMP8emZyMC7_
z2M*nM^+D!>qs6Yd`xW@r78~+Tv`Z1e&<MZh^6EeCjw|cYX#|>61My<=vtzxIIY)Ln
zC3C(xIgM-XsB%whMlIOC6#r=4v109yPDXDCa0z(d-X`8M{@k%lxKr=8QOwGu`nNCj
zd)|C|yWVlmuq(}njcXPbS=$q;tshKA<9|9pfA#i+;Z+xTgi)vjiHTkS&|?Wl;H|Z;
zi=SG_An0#jY8LzpBpOx&;5a9yrayrw6|S2)BB2!~6EkWFaJOw4PL(Nmpt1>!KZ@W|
zhfzF{#DEeDoq8u!=IK$HC+eY7=E>vEpS|l=p{UA<YGE+EH>i&)uA9uLXsG#k|GgXI
z4r9Q=dbBHl)5>?y-Zu7ooiRx8c3b1+eo*BETMtf)4}g{;?j88xa4vng)84374MRU>
zn47Hr<8Gl#u?VUlElou%oEQ?#bpz-I<fY7uI_c2+6H;>7)#F5I$1a?yRuwC52;#K3
zTUnHa#P06}PSg<S97vDl<ZM2%G?Gw&H&~wip`$)Ut~O8_uB&a^fIrm!>)TP?tqlEl
zzp+6#x&eR^Quf2^`CLDZFs(O}wMuys=4ZNAlhi9gO0!;^AQ26AiKT{ZrbWD{iCl~N
zM9f~CKL)<_Sw|)f6k*>AWE`p^q}5y9pY)Zc2rp{;u3j|&-y2(pu4&e5-AN+cSxc!P
zV8W)m<S!gqu_?vv1b->ZqjY_3_0;^lO14f_G?H?atu{z6On<~!3mUU@V8RHuDVj1t
z+FyqHLT!SYp$*mT{_H>Jle=~T^li6SRZFDu6iTk=*aV(4D0Mu$sZ>lQGz2A-+_L>0
zX{uc=1oge3S*OF_#?%A^mdLe0d50=`Y=EOIvBMIAPMfkNe6UCrP#;a)a|(C*G5r&x
z3*Q_=Pph(xX^4WNa|hRB&CDK78pVKEY7o!I9Zb#b9W_We@1T0Rv>0G$I8(z}efanp
zS1L?wo$!1>eSQC7kkcq^WJATYBQ)fC-`({wSF*V>J@{(#G%hr&Msw|GX!T1mF9iv6
zB-B;<bo{I08<zC3s;1y)4YAA5t)eK`tefOcM+Y3p1pIyKUoRCVu{B+4DrZCZ07cU-
zq3H!FGNN63<N2D46OZ{Z%owrrq7e>PgoTHwP64q!Aj@vDT8YgKR_73~ISA6m7Nv_}
z6S`qyZV!ydzWO3Dc)0jvShWI6Uo%)c5LWBvLL<1+(8^?(t|cvZnUoQ&j4#NzQO%(F
z!&{%n#C>itQW|vr(6?<O%T)EeQ5LJaTQlnZc*Zd&LOjp~itqR-o4k9;%RKm{(oU(Z
ztF?kopK4QL-q(`UW_Bjk7`kl~gj!Dt7F>+!Qd_C^i*nlDoEgc%tzxsE3^BFtboJM#
zXQzVJRd&_Enc0~h$N5CoR6Sxr*TIg*A9O2}ZT)k`XGCx3t^lPW32fd==SQ+LUwN<&
z6c@yymVJ1>c-NR^*a%eCkp(aqwmXS0<(kM|(ZovXMa9JV#!$-Ss{-!mQ*w==(hm~i
z;RUzeEsMD(Xh0Pny-9YE_ZL`ZOXwu6yeNn^QrE*X53Y`z7R4n<nbIest*U2DvppeS
zc8j@8vXOWsYF5P3DeQu*yr{Gm3%VYV$_`{yVWD@hje#(7uqTPvzf4dm(qSDbZd>1W
zVrc~lY*T8#>B?|+mD2;QN|oy)6!`-XOYzkPmIcIuX`%}TIqq~Z#l%F$wq?Yei9xm=
z&Ec_9I^$P`s!CZoB1Ws3%4z{VH}95DrtH{Nmb<$8tf8AYZ?auvTx{RPW2@tD%WK2f
zb=$zojq1{f%T9N5fHVQYH1D-WWQqztWjUpqWlI+y0sajWqBCf#A1IQ-_weA*fT;M~
z)6#icx*OCI5EL3#79{5rZnxNWeeW8)_n)pNz3mWE|2FpLNwI9Jgwxl{3bdY(_blXX
zBWnY2X4i$*$fvSl^9^4lW*Rtl&9mUo_a$&Q<yQjT1v`RY2S1amZ0#?HrWQoG*x@qM
z(Z<@_BHXd8co88{$FhG{?TRZfQ~+OsgpF8g8?#`VVvc@Ow_nF8bNah!A?AH}UP~%~
zwy6nqX)49OV>xb1HS7SS(JqU@?-n-XV^?PSi(WxFxOUZcCRK&TGi4|dR4^Vgk^yRJ
znge(4Uf0)i)%55LcBD)`zx<&;UCyv@4gSY~A2|7l7ocWsKeERjV3Cmx<6i`=6BJ!p
zyJIJ`7ueUngMIf~h&s!IL8H_#1Lau#iwq(Z?UuA$$_s;B8QNHr2=rPp#G9IE!vTn_
z9zDf_;D@P!b<@x+O&A0W($;0!0?7lA4zTRHB-1p*>G?NY^(y?Uj<#lhsOyan-I^vI
zOjYcS=6E-O2P-8-N|pWg%hbqf^D*>eiPzw{{kmIJ_jCVR3Hhb3Ctbv5<9g1?n-(|0
ztU#07b|n%^3ExY?IKkU%_G4OgVCN+=Mf6W@MHzNiHVRc`wdK-Us<YEmQ(^0zT{Uo4
zR+b07fS?W#IeEHJy8{07xtDewTqy}3pVX|t^J30StvZq&1;<U%tnwMB#Zf7N5E3m5
zL{KL9m8Pr_Zs0M9f}~=KzAx_Z^(c&lR{5bL82`<-M|VW+F+IpFS|um+ZJ?I)k=7!q
z_<btY65<3gd2VVOnLa-5Jl64=^2nl~wg44~`y~GZMPhnVPCwGJc@GY`YUGL9&#5Ex
zI)O!?is_VMY8hJ1NFUr+bN;-%2Ji)NFFeqVC@m=o5H7)b>*KV4;*;pTa0#C>a)e?@
zIR`uOeNwJ6IRSyqU1x=;+&CvBUa{nmITSR42GOBMs5@913n8#eVwwYreUbLtwqF>N
zs6&Ua^AEmVu>ZL;Okbw8@7u#@Chq$M7nh_bpx`Qb9G-Qa=xQ}?%~F;*SNRR5-aU6O
z_7zv-p&3Zt{CMZ9M#`bN_vW#N11}~Om`cKbe}(Mi3Gv+i(Qg6td%J$aqk3o&iCYlJ
zLd<OqIfh&q4{t<<dPP2;wacb8?v@e!zDwvFsez?$Xz^W!vf}`03=y-2)eAGCpe?L{
zKi3Gc?*__>e>DV%w4M{)E&jy?@B^J;t@S73;rgkciXNufG-22e9~$%yy2Jb5S5!ao
zu>9kDyW!ONM>dLm#^`%*#^#`<xjU0`a?$3IuvY!u_eh&bsWY^!0@A9a;E3<U)>Df%
zElal?z{C|KR`@Ph%um-^6T3Ux-9Qd*BUhLZpk|Zg0xzo$V8q*eQDtRS`gk}XFEjM3
z=2*e=UgIPW4R@LkqkX)Tx=3rVMjHTWpyCW^8TEUGe&M@1W{yH%TaCLQQ~IqXbuR<&
zH#IA!u9wRsCcQC_$CTD%WJl977JHgE)KtD+zgy{3=zB}SbH2-O^Pw+0(>C5#)KXL|
zR=UMzd`c#;TL0-3hM!(<l?=r8OxAF`zM9j~C<fS2X_O4Z;r;hV0&PDrTD7x^eAMaG
zdNh(2%p(Y=JMAyFn180UB@))6K}RT!%PlMYt?-(%d8P&Y%T5qK#8cj(J$fDz$0vA}
zvSPKN5I3+bORSnMiHaePF>+VK^ZfY#6`nV1Yh+I7?3abL8NC_!=!1Q%NyA!8_iDMy
zAd0WuuX?49Ov(<r_T-O_(jMEyDYQ}X^MlKgsUMA%1sM~r(_MV0?)3W5!iy8W;L%{6
zYI|3*&jyaHn{H%}uQqzX6oT#_@rL`m$snf}Bzh%Dor?{xAT1co(M=?ua-~%qzck21
ze<K^GHitO@I8{#xLS=1*C%j!VUUAn16CE%D1B0L*h@=&~GQmSps_F68$gB;e=<D2M
zPPJidQ2A9=)!PPyoNVXS#=OOmrym<DZKDA=GF^C9sWB)-PQCiFbVX2pk1)`<LlkfF
zi(p+)BXHji898*CXcOOrnYr;Mr1Jun=DvCcq|dyE)to<VNTiP4c+3tbWxuj;A!|Bv
zh+5b@4@1;QFlkV8Mn93ZhYpm91;6lz)`qU$4dFwh7m0XprevT?3JSa$6m}uCUKcWG
z>45MrAF51xR3ceax$*a@VClyR<ro<XSz8AQ**Z!d46eCOf8l(mth%BfXJ2R3LRu0d
zf*ssHzMh;>{JdZqe7Vu1ITz;YqQ884mhwhUWS^f+YjI)M3S@~P9_Wh-Y3oH|PsjK}
z#DbFdQ(d)!i()pc9S>q|trj75Cs{&G0Mb_<(H#LrxprS{twF=e1MN6X<b~Bs#2GIZ
z0!Ze|x?7Wd)v-AT+^RGr@WnI`1VnKOi@@s1!{-L=RiFq-!N{At8&<qFB#tZ_k8ON4
zzn>m<Oe$MxjFt1TF*9SH&wsaE-W5{V*023V<qm2Q&p2TJs+F?mL5c|+Oqa9cocsas
zGN-NnJ|*>1oFnb}i9hqINnh`5ilO?~p-%b)rzX&i{N95P;+K0Z*r92Sp#e4x4i$)I
z?@Br)Cq$9j3YyehC3^@5p00gLPi)<7TOJ<OSI5SPp4t69ClSGC*xLc=y$ysR>@a+m
ztr!mJ_X(Jr2DsnoJ4twdLK846BZl5%tgl787JufcbO}UmR)c$MkDm%GLmVKv&znxO
zM#~cC!`cAHK;HblmrB)pq03pc6xE3*$$@z5;B3&no#>SV!xrDI#7|~o5kswWa_y-U
z0m#+EE~{*Q0d*V^<PZ-KdGF}F9ConNdxb+fODG>G>RArEN7IQWI-qt}U0?|cZ`0wY
z_ce#gs)(V0oYp`OYwor}YE*tc1e_@_#==5EG;T|z%(2}{4V6d)+71)o&NjVp9asGO
zRCsHp6ep@?oSu>6xbW0U8<J_)I4Wy8D#5S%b^1|(`VEJ>M?IQZE=}8o=KWcE#)|Mq
z&5hGR{yGxRj&H<e6=zQA2W(yp9X>;Gl9(9{RU4>ULOKP77cX+ip3fHi!!@@_&=@#p
zQmrbVbSw^**xs0?TJ*p|!M1cTDpIMAlEE4k4jf|bX$hVhXQhVJpt=`;p(MX0g-FUT
z2XcJGkSn(Kx=fI^Y^jgu8_=9<o$L-8lZoEeJ9f~A%s&r5D7c~IgVX!eAXvH9lK4vh
z@!ZCuft6J1lj3IVmBtj|IO$rK%ZF0UhhEcwi_dB^)AZl_vXzn4_-<5RG9>_Qdoj-!
z&B|+xkeCBt5KdN$HBay<p%>4)5HAOel)Tb~<k;!Jg;^3l>yuQ8Lwo7Qj8$H^-h#B}
zxkc1dee$Dvg{e4TB3Z7G_U~G4gM!G~5JCWNR*ey2ZLe!-;xL@}N#7H@IytD;OKwtj
zo_xF*r14Ouiy2t>toe0BMbEX-cRYgO$91AM$i9fsQhECk9a3&Y<J>D#0aTLlIKkV<
zy{{1&2BxBJTkgR#8_If&^zT-4g`nD|M(!+>$ej@SI|S<<pSLI+c9-(OG4Iu8OPJWw
zzt^u-TlO1Ynv||VI8Z%9zVbQ>R~wQ$ga!M~Q+1j+a9W0GVqVT_vuT!RiL*|a(Apjf
zg6Tn}VqXWYg*p$tcJ(`>>q#JwXDAhn&1Wlz9Gf!d!HkBc583Y5mcPGwo@nM(8DR0p
zD*V4)i?}kJIjL?g>HdcO2z^EDtLT5Fed!lOuTh-jriR9ZJX+p+mw`zvHfYFBU81DY
znL$NwyrPhak+o>y;T1*;F;|7EYv}~Q;1HNN4QW027SDO|BCocb#~>#rV*}NR(Qibq
z0Nr3=O0@;EpdeM>u(GsrF$I|tIa`_GaocF?*+iqT@e>Y1OIur4a)HLaMR=)98qxwN
z7e*xmR;B2~Ukywu_up9No4=k^yekMvRQGzQ0`}Rq(#y?*0JG``?dfpQNY1R>U+fkc
z?p0c%^+=<ST^`ZPgU^pZk9a<zl6<^Je5zt&tVROoAwZGTO~_WAB45(9;H9i!JX^P*
zNHnY(hrCL|iY&_6*_ESDOeD26wS8!plF?k2_^h|LgSGG-OICag1f#Cx7rUf)LlfhF
z;WIYO?6qTm>O|?<TqzrovUU2hOMKEW{X^#2>#0Sn7aLM?Mi(Z!uRm8>M4}OkvH%eI
z7p?_+P608Ua(HQHb9qC65d?2RKOJ)#Tb}dwrQ?GT@LLF%jgrKidJ24o8e!Q^-Yx#A
zFeOAsmR#*ISz2{$WIE~z>!W1ZP7wG`N|b2UKyGHbv{(!}eW2VI1&k-r;zc6ex7Kon
zYG&t@`f+9st+0RQlTr)d)tU(+38RclDa|JV(wOqNeN(<!q$_fgyitNIcgGtPwt5Z(
zKi%i!ty6r1;qPm^KPKHpom&%Y9zaSXY@3N^#C`}*%IzpsTk_E1zJbL+;vd{j1+W7H
zbGZdHV+HoMkYAlSv-qlXnnC|8ENVv=d5Su7sm_r*uSotbrrr?O?x7A+uu>NRC^KGh
zh$ise@WAu;i9xI37gc7Jqm5m*wb$5n-zvy(xSlsSJw);huX%P?(F#<NO!Zlhrc%Xw
zYzLW_9BsShB{HPA4<f{AL|N+*;~(dq+>L11`xy^FfmtApPqU;imDqoccPOh8&FHAd
z^X5@+-N=r3<BU8#FS8M-by(}^SQV42M+#eKMSUY{T6OrM;T#?|&w=`Tc3nhVYjBa7
z;o{91SmCiGa&;UG>ZOFFY8L3DwgLFFA+FGUb}5mw6C8LRlRNcg(|d(k4lw40Qh02#
zJ&6n&cD1kUAJru7*s^3;s|SdsBtqo7wyj!@)!Odgr|ON*+lJ+aV4{kd^Ug;cRhRxH
zX~wP2)R25vXSONr$Y=O=#k+!!UfN0*HD4TNy<M@6ZInX=;D$Ht<S!eWY<qsW9TpgP
zUWk-p4~^=i6VrdM@LZ9?CnGlOiYQxm>zC7wo-!K>55Y(+##$IOLWHO9Q$R%!bAceu
zpeexo9h#bmj?1g@WTjS4^o^>tdbAY2Y_2Tb_2xAC1bVvn)<Yr9mH8>?L8chpK}`0}
zZ?pa#JIN$kV2rD->w|`&O;hF|v9<mEdSg4Z0ryKHYDq5Z>(d1aIAcCKGo3pobG|=S
zxyPt3S3Zn(F$7nb^!8~k0FiA=ILT{EflEf~?4xR5KAJW}XSRlFeCrXiv01hH`1%-m
zVetviup7&&^f$5rGrz@#XIDI%E?@pWb_~+@@1aR-a;QF27q^W%<9{I^?~o9aa@p~>
zZ3d2A1+D_b0bEV@Sl*hsU9;_Z!=W}hsa%1vdYi<O1>26<|DyqdoqDnLS(dTfHrSwI
z8)min*(xU4bMKFD#x7wAauD@m$CiPc^%(<Wg3I5|rQS06G4!Fux=J#mF_9^YzP{+@
z-L4d?MrmVg%YRYnI_A#+hYF{O$&+flmtS;3x{=vKa(1R?pdT<U#@Vs%Hmn~#E!UP-
z<c?$|cr@dasUTffQp)PMnMgymy~RYdNM6+hQ6Ajc({2UBY)<qNQq};Rmj!4|5}2Mb
zfahv4mE2{qtt}>YiLv{A-(mX!VE<g({op_Jc6wpl^m&C7^xhCi)J5l&OlpYJQDg-I
zc?F6odWd|=NkRGFsYg6Bef1DCSEjre*Kwc~?HhePK;zU)<q6JQygvrL&f(aL+KqkG
zh>UFuGiqv1piveV2*Qg-zDr*hARyflQ;M{<0(a*Us4qW1JzzTS>V6h+^Wvw<6&EYu
zsITonixCk@4P|?%JCiUTm{JDcj=>3<j4EAdR?gff{n;3Ub4_Hzh^RUHkgE4ryL}Uu
zcYjD*xVK)8s405_XU_fc>jc{@><JA)gFYYdW=op3ed)Ma_AAfBm0zi@(zW@(WPlP;
zk^xuSV7$)@G()28f1$w1lpWOb1^?Y|f^=)>KbaidStDs+b&72oQyaVb{b~Ll-$DSZ
zM>k<S{q@2{!VAcH8zq&lSc82Ry%GS|Ld@NS1S$^cHfuQO2aEwDO28wfNQ+Rx*BB!5
zi+!`J2VbPjQj5@LFAUml;I4_EDCMt(6TP3(*HOVEn8>}SyXHmHu1`3cU{3<8Q9R%&
zdmvqtup<1d*kQgP<#?y2p;3A-=2IrX4l*~X0TjCl(m?(&vF!PV32lbd^_hHRR4lW(
zK9Ju$l|>e`nn8dRfS9+phd0#INbZ<W7Ze<<xyiEN)yMIhSE%qZbo{v%JHUOKXmr-9
zmzIDfAtoXtZn%u_EeG9-%oTP`f)I<9mIDqj!ZJY(xRM+Z`(Gn1>t_c__~@lnKlWSC
zYiAI2(5?~wsrW@6m~sDOc-<6EoKRC&6aG?f(Y$b8DQhZizBl_;_gg$;@n+kXFRSgQ
zZ8s4ZG<8<-nBCwyI=_6-E#MjAv$HI*<D=8eu!|u^8b1o}eV3?m2l~aXY+|bZ`o#a&
zFV=OCPJKT&-8B-L=0@;It3qj<mRCq7U>1SSo0==lcFYpu;KmHblp+AF8WA8`dU&|G
zF15%+Hn}btl#zWB=9wzhT7xOwZ63Ni7B~e*C5qTh`rD-8)g93u^L0Z`<@3dBQEwge
z=0MfTzs048Q_2(H%at_hEuM)_Hngnb%AS4D#UMl9#9fDBqar8cObSyJ|KEbK@SlP)
z7a*3|Lx8zd*U-DA{fDz<>+2kLgDs{SHqb{!SB!<ac%a6jn=6VjWIuud-Xjv@;n@$Y
zEhz@$iuFdxgQP&yG%Crn?z)1ec}9-<lQpM~r?-ccT^_w#$&U4J24(%zFFtmNLk89T
z(=YxWHg*ZvJ0lgF7gO*z<O=)IEH_Sy4>c1iW^&1bWr8a&X3PQ~3v|+`wJ<vqa!ICV
zcah7yG;AEi;uJ^jew&wn*76Hw1>k+!Xji}ZG6_>;7M>H-VF!q<mdwJk!~4IBxCc8p
zf>#z@b0%nd-N87mogHk|@0ZWb_irrp`6mUN2!OcLp#V8$dTKwlw#8^F3Og3M2@_f5
z{`FVaB#mZOn_w4vcx3@$_$GYev%n3ZxgDf%JP=-s9O!S)mIyeOONDTI$rYxuk-q^Y
z!kN<5W_CfK4O7t491;kpR`^fo@~}E%dh5>QS2;OuW5G)n5J)ctof-m(4RHip#yTF@
z6HQyKH(ZaJjt=y%3QC+}qcgdb?=M6j3Vp{ajtAoNPW|5-|L+6xe_ss$2fY!bHSzh5
zg_koR76n0zN^7?j8|S?YUtcdXPZU?XQ*?lHu~RBB{WAX{&(!Og*028?rE<RJ@<~<B
ziK;vQ@sG#A%|D!q^S9Y0tD41wlmNNGExAA9QvN=*H#!&3H<>bPfAm~K{_j)VllH?M
z{K@*8xh=$Tm0Dq(F!e;-%kyk>Fwj+D_Q)S&SN=ZLd*HIRg*zVFEk6e0@yA$x%p9m=
zBQe%R+PC7sH~mvOP3=ZSG{>IwRflT?v+5!K{GDxRAEtI@(pUM@`7G_9vW-Uxw$-Y1
z4;0Mwx2$`+j|7(ll+jqZqu=P0hxLxVw6AOd19v|7tJfIh$fUN?`=I3|Y~hs$i3%XR
zOIhiiNI3;<fj>2RA+3d{Z;Q{!q}PZ91ggm`#+77R8ewh8`Y!?<E`bIQCBBmu3&Jbr
zYDk8}4#JYQ@9h*N5|6mI7}owD-z%v+#`F5!_$FkOC|b;N`N^W}Tlbc^i01d@I;7G%
zH=DNg<-WX<mDwQ~>7Hq5aludLw>I$F>TH}!L;RhStab~SSx0P$bocayt?H2({NX6R
z2>3Xzz*n33qtgzQ2kK)^kA@(IINVYCQMxIw=|B-%!D)TvLAzJbY<f%L?^FMcy8-9k
z%i*KK`4A+=G6P@(4AS}5KMvBHD_h6tkfmqCAW1^Wn#MQ*%c4i9dGU&V+p%~6&)Q@r
ziY#uryxS1X=qt5MDNI#_fzOW|dzX$%zsY5S?&b8CIOOV~Eq_zD9T(2$PQJ)D%?LQY
z1_8I4$j07zd51}?w`wOik0R(P8ZPw(!A1Pqu#1#eFx@ejHmL9=PpSzo1cQi@!BK=}
zpZO5v{+d|XCMcn)#xf+)gQywogCUS@S6s-2Yqeb#U1PASip8X3!V0dz-loPSCb|mH
zil1b3%eLO)g{q{)AiiI;si`N_r;Sb?UezjiGLbD3U|o<cVg3jlcz0Ij{=1w@B`Hzk
z1{gFNUR2@f-Z9Ml2I$2<enXA<-WJpT$1fgKr^-ztu?{7x0jGTP3Z52mhmcaLs3>ju
zk4~W);B=-o2L59jZ(L2Op#Gy8m%Ia9H4y%nt2H?v_*uN2cT*1GYeei7^@)k;aimXu
zn3%t|_@V{zqfzOBuJ<R;NGI98cjZ+!&S5KFFOZn~nY3GTPu_mW>IoR06dk!psn?la
zVRp4X!RAXt#XuZdToF{U22#!&_;l=FgDf1=X6i0C$FH`e5mu9g5@9xAG0pIb@~5#k
zcGTR9%HNOA?!FH`L3_=OT!O5kcn7A6H@BC%A<ONSO__X1Ce&==P2|pld_kd20-K}x
zkfgMd-#sgU%H^)|^x5kK*pKYVF)p~fOnShz_8r*};n0Lw5o8wR_qdl%gUQv=Zjt3x
zb4-bK#MIS5h8_KqH6caKcR&4OT{TnUz(b9$1H&t1<gX;HtYqAafSZmJB|D2wOHmhF
zVq+fGIu7nIoI*h!+(AG|;#2GsRs52f@G<>=ddN0joX0~!Q9+81Zv4YX&e{o|SFgK*
zpObZ*#kJnPTGmxa{I#n7Lw-kGS@H{cue3gy?*ANEp<3<z_V?!GrdjLoTI{P?yKR!f
z->0;LPvpX>*ou{c5eRIJT(LrZTCl4Q>EhOBqp3-lrnFCW^;HO0^MFN92*`=F9PJg?
zlOu`paH)?QyR-mF-Q92|mF(+s2qqcvFGtybf**ZBRV-T~(_Ij9V4`I6?`LIep7)I{
z*;uDP>MXs-eZZhyzX*pdD!p?1bhD6xh2Clc^}+()`hGDRx!AZ<wW)FS$h;rl0wUwq
zd1k12M3lZ#53{YPgfKa#0S4|N7(oymo0%7;5epgjoV57RIAZ96%UTCK>pL64)MKZk
zJr=|5%3}=Eq7{(ss{wPYR9m`E*P!Lvgo9<%4h_~^nnXacm_<!%U&XfME=0%vZX9TM
zQo!M3CF{fXuB>%Mvd+G@hDEWY(#Ih83T!getjel8pS<&lZQ2w>qm5BLD@>}scw1`t
z9$YfzpzW5#O3Ix%$Y{KqgJrW)ibsl44`vbLpOoT{2cVhXHePyvsyW;7QbuXf*N#T1
zCD{&ZpY+rUoC`_<7%xLiC*I|6y`!W?bV9wE*PA%in(wbBJYs+OGW!T}+WzxUz-YGC
z?Q2-q>*ZMO6`fUyFTJ@5mk{N6v^ZDvQHg97fY?;2=`sn_0vyM9pCKHq)m*Lus4HXE
zy@3;T3$gScEpkC};j5YSNsY+(1|p!L&<Svbm6vd8#hpo}O(dR!=6ZhquEFe%<F(cR
zzT@~J_Ty2Q>#V?qf*J3zsc^(2cM4NMek(;`Zyl=f?pfs&|L62S-yV#|-f_U5e<0`)
z{I-1U>7m(|)1grG`g6qOo59bTXok2Y=3DqlL37xJ$vad{CTt*_qW&)~UKvvjwAX>v
zcW>?mLFOO@US8j@pD%3M4Kp6d8nk5wYMqq$jI!i{Y+0|O72;=X?d8Yw?4^5odZT)j
zvXkh*^D6ir9hw8Y_OCqmKb}$T{sb?I+E?T$bc$m$=k9eohPVKT;h<KL{vXWqE>a&l
zS2va{f7ksQ@!C{#pGs>>{+f|lMA+chTB*;brR%@&mk!joz{%?|T2LO{WkkfJvFBJS
z7H(#F<d)p-BXsh~Pg$d|J|5Ntql&2I(F|5LQr>PDn-7!R`W8~{4S^8|HBz*z;}a%9
z!J|-y{+$$Myz)|hO4A#@*N#DpnZA4)WYuZ%8~5;k#0np?7xF)}z!`iCrqQ{o+|!b}
z;Y&%rTK2c^WGmRFE2}D}9m*$28Ld9A09{vopFTG}{o(M7eFxuF1F>I@Lk(ZP9k^-`
ztyk{}?ZO41TLKs=ii;vX@@v)YJ~{=h9zwuhGtJ-tv`osK)SPJ&shQ+a<oZ`^X3ZRD
zRPV$Ioh%s8`&LWrrW#C!g@03HhV!ON7a@WS%rcNkE@h)qfTOWTQW9P%tyQ%@V(gQC
zk`u2HTZE)x_Wm^72`m@CRBK&bXOrfC&nm2>wcI-_b++J2K}f@xad*IY_Otw}T*`{Z
zIO59BJL>uF_iK=pxXfkY3V(TIHZYAA)7k&hVEaaAi`ob_#u+DS^NHo_&f%cb8TX_b
z>+vr2nazzG#N^;1o%z*9wWXCYmG-Q3&mWtk26~#&R^PT#UoN})YC9JpE!d>5E<mq&
z9@dj)%w74E+O$8mN@z9Hms`+<>Ugps*irfwC~Q1GHHNoX@p*<1MrFLcg2KHBd0hUp
zBI-Tbzxy%%b6K#~t$VqJI^`w0x5x#T-r6#2%u3G?JiA53?_f%miJh+Emc}QeOuOYT
z9)wbpTH7&Oh;iH?%U*lBb!p{>*a?XLeXaqS(>2`eds@N6bUgX6NF2p_0G|$3@_XWx
zusoHor2O#h+Yc%~f6%E^RtLP+nFZJ60mYZ~!j5FuGUnyO?Q@t?e?rkVO@ZKZFF${*
ze%8YyrmJf=zj)iEtf=8y@@Ij6^5X4BOuV8=`?`&4uWCfV;%SEcNtY}QFeIeLf9o`s
zBgZ9~?IE`T`H(-#{rd#Pg}-wB?^B3LNx(gwc~pd%mJQ)p<2nC*V}N^%JlYyiTpzHk
zJ2v)iJP01Eh>peQXZR*{ys5k4Dl+Z-g_71AR`+D6aRX_U@SfFJ9lmu)4_x+SG99N4
zu9;blR7I|}2|x@OL@-evumKyZ_A_K1ZfjrR;_SMHiB4oo@FhXjh3Hv-LJA^v*cmWp
zEttK2dH}i-J&0*6&dPapNrojjk~NdmkVv_mjPVf;IAl8Z#6zQ-iamU87u_qg3P8Oa
zUT~U3@NW*Y4^ciy-;aoyCdxPGW|y_~)z>jsKVo&&Rtv6vsEAf)s=c>sPK<t6hcsQ5
zD}oE$>7B<nql%RmpHFs=Uv-sgm#)*+0E46;>C=`p?4u2}HKB#&wOle5y2lG+0jF=r
z=zd~KJmDEee_^2Q-yF^^2?DPTaZ0bYY`$HX0{A3#rO;rD5M39d1R*pZs1hd6WyCbg
zuwl)=P5S<L?8f0*;V9amMw=3-*G89)Ow@NCULN_!>cx6xYb~(TL~HADwlRCAtUwJ{
z9fD&*3pkQpYbRPKf5hg^!d5p}o}v4D6s=H<MwsPbY-@7$^A#FWWF9i5()syr1J`8^
zhz#)|CRcWt*M=dVQT)s^LeRN<yfSn5O;}5Ec?M<&J<Dq^3O1DtQJ<#D#S)ShUrtgM
zU)(Vfe?c3q@@$@}5m=KZ8Roo^(GTtChtrLjgp)PzhgwhInLGf8@bGQ2-@~k|KL*;%
zBpr_JSAxi4fnY29)pk)mb(|n3OAO$Ce~!>w>J#cKkUacotANWWIT>G>>{|Te(*yfN
zmka=JAA7@vRTb8Qs1P<?l{mn0#O$G(+i&nuc^CR=qRms?TrE;*q9a1%qisVib=eWQ
z^0?Oj8d|hSuTl09e07WS{pzR^-|(G*gp=(^s6o>nv>@K5I3cI{>;C!KjLbD%Ssb<;
zEh`ab8(q3l<a1JkwBky~5zr2fBP^1CBiVJIHKGd+X6LvGgFtOW-;rVUDK$RtPp@rX
z2Xf4LWzDr@ywK4Tl9N~<V{5UU8eJ8wf{(-?dNX)S8PsakHGdUVH0gFHv7+L4^)Kw9
z7{0~tFxZbqZ<p+`O2YT~pJy%oCoX##H#7{WRrg4|`!H(k+SS=Vs<r*%MbPHL;^Mq>
zS7%o@1-716RfT;w(bH4w!-c>&V>M7qn$tf@*9AwWt?i)2`td;SI)@s~59N>%DI_;4
z%mF09AF%~AuPkFr9q>QSY#_5{JGd=kax@$=j1;{rcOxdZ=SExA>Fd4*(Z+1&aM@ru
zJ2o=jl#bhzuJi`A@YSW!0w1RiSfXq>CP)?Am2en%@Msz56|Hulu@51@DAL)sHGw1H
z^#0}69Gw7nQ8hjTzDX)Yrc^UaNY5R@o6#W24TxnZE2I=r?(|XwLxvr<NBopTcu~{s
zMGvwa2_!nP5sdDP<djH4z$^^okMZ!sE8{(iBTJSB=bu`B=B;j@o$L<Y5jT9dw&k4c
z?$eh3ZpB5i!sYk0h1%<;OE(9i$V+#tN}iWX2znjd>bxl)zr8!uK6o5N+JE&`A3zBm
zMyD2Fol>L3r~11xfDo<eul#B<VR@{EgZByG5cH0ZT!AU{o!U75J@Oc69qZ_p`2z`z
zL-TeTcNEoD#UPhJv=V*}3+L<Wxd65V9}{pZ@7#bW24I&)TYz*Rnq$St{JH3;MDM6w
zZ{+2~OC6d6EtEU7;u-C8IQbr=7l-TI9|lXxJkxeFS)E9ED+N{1o;Rn%&YH9I^{6;L
zLsVoJSB*TtpGp1j%#*1*%@`Hh5}!W3m+PM)bnESFN3bsO2~8*oINVf3-3|?#449pn
zo9gdtb=$TiUfl&yqzmA6d}=7f@#uQLeSupos1<!&oqa9;VFN?Q@YCJ#ldMbHPub2&
zX~SkGj-L!c`p2^mVlExH*ZQ8<7#S8pkHrP)$p510b3D8|!1P$`ChH?lj{9(=noV0W
zain>vn$n7<Yg)JsjBHq5`3@eltA^UXR9Z^fzkg3h)PobtExE>9=9`5B+1B)&+|mu{
z+g~Q~yKNvT^Uh9>*=4x9AKo11jq?t}m3jrF$sNt-7p%N{CaY}oH`_WBzD#OdI=*iF
zXsiA+^ska@##fn*>`DQwb_SnKTVFkJ@~Of59azRZsTO<OjwsT1WFqKA8j;Wa4`lmE
zFj8vR0)yBHFjmN~N&|SJ?%ST>{G#C)BRO^7HL6rt<adefn)&(e_t(Px7f(#~se*Jp
z{(2MOLC&YJKoGFWlIO`_yhw@LAr>I{<|V{$Rm%Y~k=3WIx7&BcX3S1jF?&Geb)5n1
z(3YP6eQK?H2AG+=x&8O4^Tn^;v@s(s{Mz=ut+p1SPdjV#hnC5vA7-hf^KU_vx^5N1
zv*E-yi-CPjf!4JF9viX{ZvV1Fl~Vn+kKzGMUY`csEzgcegZBqkD%*}ujRYUP*qZj>
zxBpC83}>-*WIM||XFOJ^&IMP;gbWc@l`^HRhWN~Vkh$*dPD?L&Ki-#S7zltr>~-(y
zotfC8J)FAc)JUQbme6X7>^MER?7;@N8<+*Zn_XZ)Js)*<w}OZ)*HR4Md(65T6S<n7
zgn*aDZxBLhN2v~fpZY7}UD`0YU1(3Eox}qVMti%pt5o9Qe8c8pPR$XkH*DSb3F_+q
zo;Lj-_}h=ez<14Wf1mpI-a++$YXAB_@E-pEe3xL6c|xz6Z4cmr?LU(L=;i)Rp8KgD
z<}&Rw?Cocj3hMmTrv~98ns@!^+iZ76;!ss6>0Gq9Au0RUcTZTn^*!xda1^y~2+Ihy
zDWk-_ZZilNwyu$`j!bM<MN*|MN`qhc)d!-#&LN~SnYsp_4s;zd^Gfg&LQO$TaU-=a
zrSiBK8@6NYnR#WmxV&Yz+C@D(<nol;>eVO*Cln_);D9DaAsO<W+o5iZP(crIrTAzt
zCj4w-_kVp3;S(9tp=Y1meEeuAwDV^tldN9)(m+EGub4<f%H_abuO3vs#R<g<)N2ox
zQ`dg>U!~s6twOG-%<BC<JDt&Xb3c<@y{Z>}!XLwdzwi<JyFt3jnz+;FCDU30UKS2W
zkfH~pm+(`d)o7oB)kU8hE#n93KHSM@x#S9nM-{_Jbkqxy4uh(K{vY<<J1VMX+ZRRs
zKK)ccLCHyS&N&*$(17F&lA(bnHaVz><eX_h$#fvo&}1cNr4!IikeorX67=%kcxT-6
z_8a@`bN3ndjd8}^{m)vfW>u|K)wQbTnl*p(7tyvGMyzQMfWeCLmpe^1i-LD|K^mV&
zt4G(4WrX&elAoaVE>K{nHoeNg2I0jm8`ET$mY=sJ$*Ez|ylV}^l9N@Up=*NmOa_S5
z^CLMoy<Vfd(-{YT%dDwH^{8#zH#UVukFAb9uvg5A+!EHSt;gFIo>$OoKtiA{25kA0
z;RRg6^csNEUbRI42Zi7#M_WLJ_vOahCtVmGsMqAXjbQeyx@+1Sd#rnPf7ExJa+&14
z34tF2>7>NHYzIq{wD@vIU~@7*rb_mNN<pwnA)K+kq-h`*mvh48W5DU*so@)+d&FBI
z>?qp%y|rt|6ervi%41^+bKGvkznz`qvUH7q_0I2xCqvI@cMkK+jSv}hA0!)8sov!_
z^81^9`obJYQ>vZ-_Sz?F;Kd(rQ*DI^nS7{mn8Vb!1WXQ(#8(aA>JG&Arz+O0D_QOv
z4npY6)5^D|Px<nZr0%93PELhWP2MdNb4e_gB|gG9YO$)$IBV%`43Z&WjjJQH#dDHo
z*HCh<-qyO^FfGN}r$2wI0`6^D87h19g2>-bQGDSuOCIY^orqE7tJHH`r97CyPunQD
zZ*AM8&=?)4UDF6nvWX3=_bfWL4hW_|b(PEQReTNi&L6a1dSB5iDY#R4{U&)FCbe+E
z67Hh!F78V}3v31b2sy39NIK~U?0v=a^<ekrJFSOXYgXJWBTca-`NHe7$iM`DyACq=
zayEM4O7Ek4z7o8QkYD``mth>gw`8_vKa0^VPonWK)DdO0*H6n!a_E#|g9Yb(Vc_^y
z4<FyN=Q<C{aW^9Syhq+lpw^GG%RJ<;TKuu6eXH6xe*PDFAU~P$ZEvAfuuR0G_sdPF
z0AA0V97B{lOKJ&Zh`{Cb%7_to4`e<iA|CR9yVsUXtZsi(Zlb*qI8WXJ-85%<yhUpf
zz$Fz>95K6{d9B8_)?SlcN3<gBR%{y6Tqv)$r?&Q*W8USL%fByNUAlek^*JRtSISL8
zIflC?_+U$R9w(lGEE-j9Rr2ENMMZsfmX@tct}&Ub5*Pr8g+3`^`{_d+_$z{NamIJx
zPCMh&n4zHn3zlJ2>z<QCQ(GjAwr`*mQ1x+WMb;V4H~<BxSyAU&8R_`1N;O-$U9xVl
zY~HO8XHa-G%dp?a7~Y#vvK7I&EpJFaPrw{V3T$AH&bZJxglWKl^L4D6N=i|Gg`w7f
zg7_N4fNM|td}yn2z+Nr&DX6+jYMwcx)T*Z?E2VlFD2Xc!*-cO?nIJDuGBUJ=(^2Vs
zG0y?o=asOz#jn?Jv2jTsC^1;s%7o*&Sr#WzN%A92k$hF9UBkzw#$_+3hkKE_1IdpS
z=1o||g?e?1FFo7e+tybhYllST?`v2O-skQIts8SKO}R;a(DHhbqN&h^WYJt8ZBa_;
zJWiawlGj{-$l*&9T&xJRZP6F!e^!GA5^4s(odyW{je+srVT@#=hxV_oUulR}W@7~c
z@^@%n1(nzDsusr7d~@6kmO?BHqb21Kv+zd-t<Le{eEqfjkFXO8))gR&R^TtmK)20O
zZu*merIWGlMjP_>_g!<w0T(}t-NV&d-IICU&A}DZ<h+6bxe-rY05P_M2w~eux&b%f
z(jvaWx~0EP;pY3>(zdDfEF5Wpua1tcgSv$I9xr_VqWe9PdvbY_{gv~N$i-icF*$GA
zT9_N1MI}4+9?7}3K$qQ4cemP_1FUzwC~>qFT}!he$!gGy3aj>V!l!|fxx_Yk9q8gU
zY|z(oK7wtV8kuxW)Y{$_KrI>3NZmiGFKJeDVwjTRJSWovgzZUDljRHON92ofo%l%-
z@~Y=1roSsLdrldiE>`W0#w2bTGZ_C0n-ZtHumT-63^1)pW*C=PQ1@n3iQO9^a!m(*
z#11UC|9M8)R|#<SdizTNBz&twmte!Q8zQiP3#ci!(fjHPZ&%}jzR5|7)Wi7~v?0c<
zDp~X~zK8XEt?I@BeC=wgZO*IttU1zZ^F4DG<fR{7#*G}dsO?j%LA;RMv+CKBQzcbY
zZPS$VReK;?-V%Sjb6jqG`Pe{)_wn*sJs*+jF+pufB8PACF7?!e1An1Ij84sfMN95V
z;}GqDWZ<2qhMt{@jut_IgE6&1uEC5zxgNu=@ZoTx^p(}TP5N(=(kL_oKTSl?m&NR(
z$+IhjqFkYc=mEH$&H@=^!rOS>yjI<CcAHD6J<dM1^3fwhe0@`)uD9p`V)DET*Cr~s
zJ)c+K;W^X1jg}f{uW;NrvF?;PvL4ggUG68W)!OPRL1O~eI7ikYIRf@mY4$?1HP90r
zVD^JL>Wlbxry#Ju?#guhj4Eb(;LsPbQjN?J`Es~Zk@Lquom|^{w&ft*LI~}<P3b?G
zKN~m!a=01$;A{hcdu{jbHM8kvfv1CP&owXqn$GWPZS;RY3hpMqB%3_FT=ejK1afUt
zi&(@3Qad{ZB6TAD`p4aw@sH;Q5~L93NSM3qHQ}eOO=;DdRcDI(tpuCzPXo4o{Aeq3
zzBW{VItsp6!d9JE9cOw5dypXB!QM~H)NM7#Q#fN|<CFwYynRA|2l1L_xf5-3DP>+k
z{7FnUZyDFf;!C48uwMz%b+Kg<_o!ofR&MVub81Q1IZQ!|1d5wbQ)M6TYsH!zQ#a$P
z^$=NniUq;~olKX3b4tlJJg)Pcp6*h9Tm8sT?WraKNMlL@>U2JQ^m;jyyv+BsX^pWa
z9Mm^_Kk}UEl(EQWW;PD`8BlH2*Zo1d0*yJViOMM6qQW@~8SAPsTR%RT?U`uQ@o~70
zwhQBJecV5?JBKtjIw#Ty5#x7p#?m~mI|T@O`E)Zvd=oLA`=-a-=1ng2;H|l*CVEe4
z^Vk_mz|JV>+xZfG%!+^AVtLQjfp7P)>!t}|$$<5TN>=yA{Qkj*W0#;agw?^VzRrv#
z=KckXYN^8(K{;KeYG*vWY7(HjIRy9EqHD7Wzc>P%0I&Ijo2r$S@}ZqoI_;MjuX~&A
z4ugutN%Tt5unj_UMvA!ColVK*DHFxGBDami3#kSHjEbIzmu;H!*q}}>EcwKx@S}Is
z>@SD*#L!5_mnSVuxH-EXM`OZfbH3ex;lZWveDu1ry%?ynUOjQS35Sv_h%@z<v3a0z
zW%yHU*%2=*5y#t9+dLQ{72^IwyZDDN^j!9w!-GlB$s;KQ@65b8XR$7Go(+wznp914
z#*-_KE1TfN8(MC2wwuu>x7Thxb|RwfPqUKBEs!#KGS+zf_ZA<gy-<d!K5P2{C~Of2
z+WSzsbYPu8<1sdsU`$w=P*Q{h1p;9w1ps+*pf6WJ{2?D4uOIMC2Ic*$eMubYEa68!
z#`8T%xfe;9VB{@tRSaEEn6@_3U-{9|5z3iocn>|_P=baC)rWq$Hv-ob^?YKCg8lS~
zJN;tWdH;%|PLWpORcn{z_@kForxUHs*ES_o-EscGvgyF*ajZn4%|v;**;sd6bPgo0
zuw;pN9ktn+369F6^T%X3?+{_ak&jMQcYii-qU*&Fi|h6-Gg$8$$NosQYankyA`szd
z3QWlHB(%Eot=crqwP)%3CiqDh>luki-@WScF(h`bJ%76&X;PP@@-@s5z$a432rC_e
z`!eM7hk+(aIJJ|$xc@Qe%WzL!R7EdpF4%gOXhL%j`Sx|X&s6?UR~|FxIm-`7<z{=#
zC#JmG-4qJRBqF5ND{)*IFoF>QGYXnMk~(#gbeKYc6{Q(oR+_^Z*s&Wws73bF)QpFR
zVQwYy=eA@LPbvNRr@o)PuZ$@yS$4zM9m7q$h}{9>v!U4=NKq?0vu@w9<?En*=F+hC
zN2;ieeep*@5)?UKy9Y3@<d*w{pG$lI9JBS3)C|=X+NgoO_lg&Y4kyV5m88QPWGDlm
z=xo~A`Wd)S^$~f$UCp}oM4-CPUC+>|4T??H>re7h^;qI4_(xn4>ab}%)Jb>#bE(HB
zDkR%l_NZo4F2_a{F{h{9;>EJ@!xVawY29jz$mcz|W`HzW@}db)tMBr+7T24sc}47w
zfrU%$tk1V!e+ry)=_K>O@JcG<;el?u5{66)Ura^&hJVd_YgY?>Wg^E2qz#d;gn&y4
zE^lqyO)eVXjBS11ZGebLFxobS;^#vexy}>i@QwPy0NLi{v>A(}rY+MrNX)Ikpp>QC
zJLt`!Qo!s55xmDFS1{^tK3aOqVyf{(K+Ct<vNdE55$|*8*?$qRZ(Wj5Z!9s}3Q#pn
zJr>|6fZ|HsW*(_5d_p%}@*jbqu-W<kzDD~MBFwD<SmNGd3K*9dyJqnK=((;%X<vT%
zRbgkPzXakNSkNw*`6Z1ajj}qhQsO*W&O_J(Rdz6c^ykFOCrG}?eN2b0+H!)9#T^g#
zL$OQ4KPlx?ugvDU%MyU@^RZJMq644jus?{fAg_uap$PZDrd_3^+dHOSEYD}<I_3K5
zJ3|szfbt=dq@xm;3JZuumxm#M*y1%Md7=AV{bk<)eNsq!o5!w4xbLsVq&U^>qZ~ad
z)12re8B7wjoWR;S#P}%9)I<`RU*cR7qau%+ysY)M)R+auysE_vjv;N;8xfCaBU$s<
zx{~0C<ale(&}@wTi~20;n~qA+nh}a&gO(ndW$IuRji+Llx988L-*-s_*)H8)ES1}f
zX>93Qn3Y3zO7$ZOUv!WOziiCP&}CNK{?T0PINlCcg$IfW=o^Qrn(`zc6b6E_GHN`N
zLe+aAMyp#PCeGA~rdzNC(NMN=lihK`M#-wtB$Y9fv7xFT#XN6zk+=JdRGkNlo76dL
zHRd~nbR?jDO?FxMo3x23)=4uwWicz&(uxQa2r=+@#P!x^=+#HwOo^&(l^C6V&W^dw
z(dGG}<HUmx4<dIM!y_4#2dB@O=W6DcKlNjfcfVkef`P(pFrV)yWV+bp@oH@r+-yN?
z4Qy#u(xS5BnxCa7sTj6XT_329+wN`6v8m32P`#RO4fz7?bfBE0xc+RiujztQofuAD
zJ8twFiubwgUHlLS;`LjzyxL=wP`yO_+Kv!I$CzvF>R3gOn7dOSkh59Y*iG|!8l&b<
ze*JXw3iga;S!(vev1_1c?@2<2<6JZ7IOup|CSN^T;#?s_??+^4U-#XEIw0m)QC+v*
zLEaW!x7lgEBPcEW;Hd#dF+wTJdge;C+D)z|uX>kAE#R=IpF>>_>_xVGcz0f;wph4c
zcM>ll*z3DvGun6!O|W;g&bNoO>{=nOjTfrH6o7DH8g}MiLZ09$YeF^HDxnPrwNyS*
zz8D(h*bw@L+*WBSRRb9qoA2&%%hSL!An3|GZNPy47wThUFY^8vs}$kK0Rc3L^HS>6
zr|dp}OtA{wLG7_~5EIPefTq`ryyf(8K}J&7`5IbZ5N;`ifAc{>Cmvzib`pvGVR|5>
zuS1mUr#(NduEbph6o$Yjdls-#-N}$z=ITdQ_@*Rr&CY%V@ds|%5-v6t=?;^)^eUUH
zSOH*61g2tALY_KbiRu+>dtWotnWTO{`kEUeUOoZnZm6$Q$j}E<?+IwMQh!|C9|ZC@
z`_aZ>oNMyi#0968WtJ&9MK#Tm^#F~iJRh*zdTm)sna#;@&H?$a^EvD^rryV=RGY})
zaPwrF;7phloU1vZ%@(p*>gd>kwD1k%z}nfnR$oTp`2zNKW<8(!#vtrYZlN@4rJ5M$
z_A?=?bG3-@Ew_qa0merCl2$O@H9<lPK(!+*1PFsjEq4YAO$D4xW0#(P#^@s>y{Y6k
zr%kGrzbX<8JMZ=UlGw~}^+BxROKyRE#Av0nsPM`9jz^pH3Za}eEtM^Oej_;XYuD<o
z=5wc0T!>JJw`X|9VzXE2LYd82_mXs!Y?JK(2DCg~U$xa9dCkAw(SGi%_MtE<POZ3X
z*=6c{RCA<fq6go6YGmrCQ!8AmTVJ_+b+u>`0j(y?Tk;!Ls?MG|C**9_wCyafSs_7!
zh7k$)>N$w4<V=cw8fX`eYHj{eacQ}B6<Fs0VOkFE;%Um4?@k(W8_4o}w_BWz(BHgw
zAd{Z~%gh1^y|9x&FgyU|ZA=MecqmezvBCzn@l{jl-?e`tMS*iek3*h?+zfd-kCpB&
z{SZy83<^RX^DnG`iaXu#*K{@?KIEuMRb2U?B-<LR+M`r(0SSbNJP2&1&l<C{Cn(P%
zc52L|u=AUQ<}69dzk0jY?Hn|B(iG9ySUw>sd!7aNlk`{rdSH*m1lY5=bw=P7TQ@sR
zn9g?%K>?+21Iax77O?}HuiQD`7wOjG5D;CLZyKM)EDPc`aI`=$SsdAd&;tFSt^iUr
z16%$!(GPu|CNc}Isb&gbIX*eA`hI~9Sbt|O&%)s6u(C7G?eDP1J8NM=o~Jst>s#IE
zi?aaBSkzR~Ao|C-#Xvd>rQ;~rn7)pWu{90LI~$g5d2(qc`s%TJ8#pl$Av-nd>)Bfe
zp)nC(ayd6xk1?f_LT#R;Jw=WUQs7n&RKIPl9fEH$&Pfr$I^#Fysqf!QS5z~Bnl%0B
zvhT3hPH`}-+4pfx7aP+S9{jpxk8^oE>BS<K`;upEa@d~JW2@$=N~F%Y0vR!+@|_Z|
z=e;nE@9N$wKpJuXpvJgD_(mGsqGBGN@}?4aR=q)(h=Urd%#VE)uXQ{91&O?7Zm6<o
zjpSz$tEK4RE1W7!X_fj_F2}XiTQV-y5K?0?IC8FND04F3odj65Up0%1_gu56I6jB*
zG+O7EK)^&y=oQV8j%<ynv}*Oq=(9~8#FH$}>_#-<+p#Pu0V<d&%T0g5fH^(lgK8#c
z$*hh`G3cbFwN;kP1lk>^75ML-o%1v?SqxnEOf$2uZ_&20AVdR%0+_q2HnYVAjiq-(
z=H!iLvo?aWop<^rzO8TP5o{l;($t5m_!z@OHILto4#G9s5w6W1Q|i!GxJ7%6{Z+tL
zay+UtC5kR!Yja=PW68XG-WgIDz66jl{to>V<oPO3A?DD3>wU+Tmrqd{&~0i6{(8>p
zjwxUJ{XTw9Z3r9l)0^)56LpQb2bNAI6?0h`FH3qV$}4cf02YAY@TLyR9};k!N3dbq
zy{4H+CmM)(uC5s8q^U~zBImq>a1iR+Epuu?VtrDab6n5l0UN@~Vt>B7@Izs0nAG>F
zJFn~<jb&L7;A-X4&GUCeh`(wZ7qIjZ^_=mPF*Q$(bK9P&QVZ8Qjgac&N$2sxkTDt8
zJ^Hhxd8$4os-k@E=LhFoKN{g5hCvXlsyxFpq2vCE^})%6)|X3)6w~Tpu8y=uV{=)K
zVd1h8ey)w8tO&ZzU$67eN(C=<rj*!#=H}A%ht-=wRGzsd`*HLUyQO%bAshn%{JQo=
zQR#C7`{YOv_CewP{^m9;?o%DMeG5j`nBOD@(<SllEhefPU-Bwd+AH;RV*4S6{pU*#
z1+l+s^ShB3yt~Z);xGSyHrxNN-R}Rj>k+53xIfBG6oQ(#ok)H7x+==OXl!nP@YvmX
zqTzYL5cU@7Lpf%)FHr$e6$h2~>I-UxyNR)B0eYo9M*RXUOFkcVG;C8N`dJ1W8V1c{
zjFpU|3@qg0FN8LIhCP~^WHI>_ZFe<GOWv9)IXgZ_ef(d0)W6Mry|w;RwM&q?W&q|t
zJ<#u&ux&4}`I4Agrtv=Wq>`@Bzl{Uw4;UJ2&7)xxzft-tH;jmrdisuff$zp<%y_4v
z!;j~lhaZ{$mHe+h`PYE>Kk99eqWQCHSZp)5$&F6%SYm}a06gkU2#sNAC%E;WIJW+w
zy*1v;aop<1Cgga_MSjvOVxu*%YvJd&J0biJ=4gb@u3g6zNcL6jE=pGCFCy1pGi@NA
z`dhUxcwK(K)kzhzmyi>BjdUm?2`s6*n)KZd?>_oOer&?hCS<+e@+;`Ze0=Yj>S*UR
zp!f^PziR(i&-`o5{I`CaZd`1XwW8~ZHT#8PpQnFHOpkX<DWk2q@}sQWo(ry`%1X=1
zg+HoKatf-cj;oJ}|2sFBzXe_Wt%~#C?)rwy>H_-XZ<2SRS0<nQ6#Ju2wH!Wl-WI*&
zn|KD2{%c`J658G67Wt>^Ywds64gX9t$;9T?Ptv)6(qpc+{`#x$pQ++MRpv(Lx$qSj
z<K@oyn*^MAQ|;M*s?<M|$<CVmSDmBnIymPd%as)SKdZR8-PaB0j7*@pFe2&x-*^B2
zdBgwhXXZZxkE@<8GXEwqu4DY-^Xe)+cYXhPV&uKXREDG;=?gNT*M7e3XNxD41OK&W
zxdmtmzH*HFrPlPDB;@T)(f?X)|50hLi0g1D*8W7o;L|w=lC;$SDDQtsMr*3vji96X
zIp_3i)#8@qzpJb5KzM7|cB%gtjzswDzf1pbdrBQ)eZ}|Shk7^8<Bs^O*;B^~*^93Z
zEIo*o+socbzj`eo7i&4+O=AD3fBk>AadLuHetuBuMvJ9q*pPQ<xSCZc!NWO+q6+C8
z^+g`i!ldG*AiQEql-98rG57Ix=7NnMD!rwBd!8Bo@Cp^S@4~RKH_H?FXY!)ddEnex
z4r+X!)yuhm3nnc3lgHV?TX^{s47i6}o45Os%?vhaYeH3&Jx^k-w9o);kdV@4?O7^|
zB-wwKxW$Jp@P(Km#U!A)Qg?(5GD6eT-;Y5l8Q=B1Ec@_0)}WBh$JO9#FYGx_YFLz*
z#dM>-v6Ck&xsamO-E86F%;R}Wk<wO9Y*Vrm#z(_Vg%B#IB_<SJcJonC@By@6`k+-&
zg@|RkJ0vj~gi}&Zx0f~Q_>4$xpa=NkB^|p}Bvuo?!XF`n#o0zu?!p<?#;}*hL)vNQ
zgD<?F{o7!0b)nQ*A$UjwXVjK9zsu04DMfMi`kdCg%Wu-o%HAFEqJ;)nW~Tvn7oPXJ
z;S}B*k`Aw?8&opriHi(D7-utg=XdqgrCD6!`aV}~B{3nJ3Jk~^r|NlU;Uix46XgkR
z5BZsMV=#ccaPB8{p0MZj!#%ErW}ZPLq|a;Uf0kZ=J{MrCy}s}qZ{GzG8r~h=KO2vA
zN1-gMs%sn}%E_I4rUvIkw+eiTTg<#ZXGMX9GM9lBZp9xbOU8&7fO@$<vBH}5+nUpd
zTbHR-nvN%TU+!qdUI;gJUyh@)f<A?c{98-^{*m6P{mV9o-l1gW1@<9D?BCD$?<KdW
z+#}#78xfh#=e1tOa(gXzD~<a`U{hxH{`d^m*eigG6P-wmoL;%roS0_1-r&$5rlJEG
zI!>k42_FgNoL%V|cu~7&hjP~;EJHk@nwvCYYgRvV#GcZTTy>X9S1!z6!I>}0ZU~#%
z047{wm(BfyeBE=0dHF$vHzHqsi(8zipc+X}CQoi3acr{o>S6O^f6_dm5h071<3a+6
z1JUbppKrdE(&vBZT!^n){Swfiu&*e*wV9d2El%%{2<E9XdMh#DS5Xny$dR6;`QScK
zZd=;toI7y~_g<ccDl7G=BYM+ry4|(U{*9nddf9k{05_9#h@v=qa&p`UwzXRhum62-
z{Dbp<a#m!V1)AnRN0Uim1s3$mceSkrSMBJmPx-mfe`bs1nmqVAGmHrEP-zPXI%xXJ
z6*?B2H+Z^j1{~c;p;{~Pr=yI&cUoSfkzdZq8w<ab6dTuOGoaKTNBd)&@_0Ek)=I?0
z*3@pzPW`Otr2U{o%AzIJjht&_SJ@6b?5Dw<X>Yh7HvBQ@4gYg%=JXJ&RzIPzPyS2)
zps9Z&zIAUS@FeOQ#VU{Jo`pYi{zyVXk<i^G$Y8Lz2eZ=C7ih2sOd-&+t(BIZlWpL&
z{O&x3rSV)KpKAiZ3J7z|guWL#3<jq<$`BEsWHAqO;eY)mfz9RQbI-OJrf&#8nEQCW
zAD|~-uH4k$Nv1vFSFHk}e(pb=e?^^VwMb1no9_B``6LIWJmLhDvPhFi1@S0A3AnXf
z|L0APba<?cKZr<OZDG^We|qu(_&#ULXv0y<a0`mWUKVa~DxfvuPzT-+^DhEzq%YFT
zX;h@#9cSI8&f1c+Y=}trKd{z$UN>)kupb0jzIol|uC3`4V*S;O+<r%n0GLVEv`m{(
z0fw$GbFBXSH0z)YyPJL8JvYEhwz%}}{)8(SiS2cN{P+I}T(+MZv3yU7giVvXpz0!q
z0e)v62cAZo50MQlGF-h!N}r7jf2)3Y*Wt!1;CtvMKK@%Pc?Gc9LG{hIpeFIc&<saI
z%P7k~GYl-8y7GUvayho(XU(V%7}At_91y+HbU%e9&~N_~?m%ac7dvK)_;iZnciJCT
zWmcWi_1&YTV3in_tI(H$f5_e*j^osj?6qnYnF}s|a*Yh6&)3k-#!0t+lt-g~o_}bW
zEjIT628_2~8ff#v$3#bs2Lb>Vp@f(@Tkqil0Q6xw>-u&%y_FOv3echRzQvO}L2WM3
zy@u}ntXoy%l?a|ulDc2~0hWL=!c<HQR@D4?r6RQcfgjOa1xB`Vv8MSi7lRs><23!F
z?9bqx)sLnFQUkx?N9k6rr4Db4Cpj1i)JyY>>~qr>TFJRSc*-OsX=#InXq(r{&APrU
zJ+zF^;(F}Hipa;b@74pS$l;1;(47}z7gud7iO-AH3f5&V*@pIDwoNMZ9&b6_%%pru
zM}uAXYW)1{wXAdvN~?0iE15{odateW&8t<GFEG{WMBSSMsjh&H8RhBgnS&JQCF}5Y
z#gxj_P=InaPgAm6slj8aVmS9=|9W3Xd7xD3IU}KcEw0(L@+YpD;>|LlusZs=;6`(+
z(tSN98A2LGzqFhapf#B>!AvwW+BY}1zGR4vn>O5om%K{)jz@u|RAAe6ymUpiPhyF0
z>*pT#dLCT?dK$Fu6t6uti)Ozc;B;xGKbJ*5kHSW)IeAy|sry4bHIo=>lR+`U7DxTk
zUi5JTPrZo4@SFk-;(QY36EyqqAe%Tm8!Xf8VCllv93Aj3W+q}}s~slIt!^-*JORBc
zL22|TxmzN-H@FXa-_HskUG}v}7{HIsD!)e3J<b?SypENSN5_+xm^d_=N;=-_?msJa
zCJeLaw`;g!kGz9*#d(WaYYllwEVVz#@*+P+8wx&p<X}a}AK$*s5@Dn<{=!a6qN54b
zxI-EPRj@ftD3wQIJxhSAbNHfF3%6v|p_-$FAN3ZWSyaZQN^!I&b?YsHeshf(Df9J|
zq*C8dSJ6Z#Q?g2%@Tixroh|wHGn^4NClZVg%suO!%i2CWi*CZqc*&a9O?>SqHR^bE
z>(<#VtsR@K4CJ_$7z~nl@8Z5D1)}FhHf{J31o158F^S(KqOm*ikf1o99EJ#v7&|k(
zS#LN`hON7)V$7o1i_~}3l{1o0=IlH=zNyvhp$NN=fO2WlY9Sq8%uZ~etIBPC#{<>v
zT$>2=jEwQ4U2Ag*xXqj6q8M9x-_o*H4Fg8bCsAZ>Q~5m85gHJgkSqqHV_R0ZPxC_Q
zsOTMpQKN+hmcRn{o(Czbh0WR<(XQ#_TWR6Y8>!kg#1_snVm%sG=rXB2987)#?BHc;
z>n7G<t-KE?$#1)$o_LI9JuZjpRNwVWEAGqDyzO{PfrKQE?BJk)AT=n~?^dHPw{L*d
zmDh+^{tL4dB|2tJUQ8}BVCc#AT)C<~9gV(_EWG<Dy(jfJn_cP{tne_rZ#yv=-!fCD
zqlp#}qxO2e)z-))*#0KQIqMJ2)+o!&MzL#mLCKK__GG-W>#8Qm54=)g6crC8=OI2R
z=z5T+Bp>~~fOy+H$Hv4VXhN~FwYqxsWNd(cYYSciboSrL%W&4T^r#!v5OOZBYSWS^
z<|v2qiePu`jx>m@m|0kW_*;jz9E9k$(=|*|C&je4bZ`a2ztPTCd=W$L`A*_v_L8kH
zsZd>N_~U#sAzF=y{vSEJkk~)1myVsc@a@9=`5;~G%AXz_CMzdjnJp}w@pAMbDwcdX
zET%6Fmb%<Mmn6D+6~4NY2Zwzfq_8H%dob6!gm}7SjA0eoX48eh#$p>;wyIP?%+iph
zTI1eEUyB^aO>4*4u+-e7hLX22!}G3M08V~~S>UNT94$463{xF9Q8aAptE-lO2T`Es
z5D1*4R*W75j#fA8s<QUI*fbp5;>hIWMQ<e%^Z@6qMbec6rp}>U!%l8%4(UW7IrOuT
zNiKOoUy6?~)3utQ5>_&|NotH9HK&aI_NY;y6cHzGr^?w}{{A-MEdTaQ@-mVwN^2M1
z=*c+jJ4(%a0L^QE=(rppg|PuCdODm*PDJi(W_z^}MMjJa5?czCwbN&yTpzP{Kkq%z
zYVIz}Q)_V-(kO7z^P#pALXT6-7qU+6a;qWWZ*1?0@@{4z5YB3G6q)`nE15almI391
zFR2IcflejzLzvrOujhw>z5q=V-%Y18k2_n{u<CCoE^J*BZF-V}VX4a?spi?>7>GV^
zWzMi;>deFccuD@}^krStn}}Y|h~Fe)*Cv#>DH817L$i}Vi|`1J8GX#OhM?xKC3mIC
zi|n)KqI@Ad+m0ej=Xck_GSDklU(#4aD9bmbAD+OIF7KEgy`%@DlApALiN+=0FGkJt
zgq-f)RVmOWi|?CqYvn$X65`^;zcNO-v1_VmeaIfZW^)~*Om(Kcj+lb*Trd?o@8IfT
z)~v*zHPXM)<K_T~V-MFgd<2zaci*Z0CYfW&U5@okZjExBHF#NxL9>etGXh9c90^!Q
zM@R2^PKq2iI%dBl-)l{sMX2|Nm-k2^d7~38Q&wWdlc^~<A4JZh<z=5)8r!r+1R(E$
zjGJdRt<!z*6m^N#!z^Akm1m;!cKuVJOj*V)0gD)U>|e+fXlo8u(zD*InO=;0tq0Km
zxRruIsRw+@di;~_P%D`3{hViaAiZpxRy%K-lb584znfDaY8A`*tPKIu%8Wf^a^nw?
zpL7(wudP?l8itMLboI1;EtoX2pYWo&<C=tMeobe<LA2~C%l!cc=YfORy=Gl;z8S6~
z53E{hv{co)pcB^%CeMw7!dK&HvPevXK&bE7q1L4f-<zDxcM#K5EI**(R)kMlhhy$2
zL`g1gNMOhyT_b(PSk>!;&KbcqbMcRl65C!)#4%KidGKH0N=tpg76u=VCK&?SIg8cp
zI<$ah9xs=gn&Yddb^r0u`Zp=VLdkEEDQl_I`Y&f@pEw-<E&U%H|9r!)dE9Kdua#AS
z`U5J)WQleh>h+oW5Wa5Y?lV;s(5fr9ogi*7Ymtep8rKpR7k{!GW$PoLpCVED)`4`H
z_DBYOban5=!5Y+EU#+RL>__7lPGg?LmydI73<~{|Wiyg_wM`{?#_kx)r*{$O@lEpj
zfwv%T%$)Mn=yPYW`>{>5LVUZU^;2#)>l<p7>&!g$N>om*gfgmDGOLEI*`830SN(Fm
z?})k`=LV_~nfFSc1CGfP=@kWQqOI_@+Sbq0ES4#E#E>=aU!kmkhRR%{JzfLm1}Baq
z04q-rku<@5AjOOJ*GSWL7|pa$`3p>^1{N+TfEX<8gG18#hJa~lHHKe)lguyo`?eH`
zDhdw2KqK>Kt~BD*hd3_OB-Iiuk23%#DKtrP%+fF!A<?2ql_t+-goDwl(SguMQQ79~
zWzSKd#K6~c1|L`LvIeb8tHfSy@SV8k_nQ+xL<-+qzjS%9HQ@hCmcC|-DvWa}Q9c7L
zA8KPzUIXMT=Z&9XJz*1TYcU>*Q?VM8BHEnYExmBWBovz9XXi_A_iS2<2ZTe0d8H`F
z`3$LyzCLsjA`r|gi|I+FX2C$E^f)q35XY182salOCqH%=x)2<BJ4G5*FkxAYkm=jX
z8csKM8($%&e9XM2QF6@D`STp36!zA$nb^Afc&;YT1c-mPmf9$w)^9z|s>A8y>AL%n
zz?h*OY4n?fS28<I&QZ5+lehA(w&~iTSLm%MrPu#pK>eGzr;X|A)lm1uw(uF>SGp>a
zKg<q1mTo<@llQBw$AsK&>>P71vNkWZbj74^bo^3=1{=mkk)u_F@)<MqlM1rA^)z_}
zp;N34=aWo>v(M~nsKa;;cFRlwvqj50?Ux64@b6Q&@#6twOg7J-GB>rZ^3Jm87Sw5t
zG7nbKSsJafbyn4m1G$Zcl@U_kIK(SyI(6nMQm;J$r#vnkQL6)*D0(2lZ;)<~2bZye
z1V5PVb!TfI6m6bKKbh+XTND|Y0s6BKI0reU=}rc?u=Lt%;ImtO4n7R}*jadNG`Z(}
zb_6bm@Av~dwt(~M>3Bn!s?kG^Ilk~e+O~rAjE!w&s<2b!xPFzDZXxk|skjiEmtI~J
z$#jIfeWuz4gW`mf#q;j?rrnyv?Rj1_6l*OsW?$LHE9fr#icc~#eyCYTJx!?GHx5~T
z&zEg!M561KK1z1q%kl`W16m|@ajGSCDW+<|iigE0gE{8gH%amA)D#FLuYrZ^#5f{L
zBlwRsdQT`{dwCr=_h{WrCH@hNgyjEdd-$!?FV`w?^-ng>=JDoO6TFY;eN=Nub+#dH
za4$xfo2q@fNNj6b+TkoL-g((ktEP4@cDI#pTq`(SX&wCfOB?UNofxjyZsgL%=bnw>
zUyWbU?ml{~RTKI`R90w(vABhDfXE#j{5<$4IIK*_diC83YWH2`yw3tupOv*@mRnMs
z*&(s{<0#Q4cMG5Enc^K3@zKAoqx*3v1L7mZ@P+v-cRBuZNKbkjr}33)xtdTz(Ms;5
z*q_g|WB1NxUV2C~?tWak2*zI0#Ai<zIZ#fBAWE4{s#RZW#wXl&oj_8@i1}1xkCgvh
zrs!6dHS*7El|$+?Z_^opp<i)K4<J=auF{cBbP<R?1!?!tSN-*&h#1{c%Zy6SM=SoF
z5Bjy$FbPe#Yi&EUl?g-IysJj{5>_FQro?n}{_f0jin+CR&yfXm^IIPr3X6-LGBNp-
zjMb3*&T92JiTp7Sf|GP0fm+_Pp_h!6tO-p?pv)|bz1`uxhPfmfqi)s$j^~xr8h4IV
z14ZK&_avVbqY&@Uc(5T`%@iWo<YgnMR|{<gr>n29<y3PSiU4@a_CwW&p143R=WzyN
z=>Cc*)$t1@UGj{sfKgKhO#LxpVNc-@T+Uw)8J><&IeNh4kU0r5^DZBeUaoVDp=Nru
zYlCs5@6^wCSHo-_zt~c$tl{v~zFq_v2H-VfZlVksvq-}0*q3Y;onm56W2?8GI2OHF
z$OO7qv;cbZl3soO-mHMOiC+@1HzUfF#Tjw3(*x6yZ8XCc<`AWtwsMO2`|m0to;d$$
zB@pZVr!Y;PsA>i>@=4z}5~G^_vxy~|Vq!f$U5<_;V~O3TUhh(xUi%KO1((*-gomk?
z^gC4@l#b^yE4c2KSD4uqT8t6hvMru-k~Lx6v=}`9(0oCusYyZckn~RL6*(qp;Ah!R
zqXrU~HI|`w=#imjJ>+UAr4t%Rv3UwFKO>zQ&A98S%_CO!*$FzX9vovEUumvAetfRN
zq`9{MB#;$)V9c8$Ombe~ro&BA)@tmHak=9v)FKyQkyH(kB&+42BnsvL-O~CvY9kj1
ze=<k(bY!*N8PS{d{@8$%tYq>`pxbMe+{5LiwmfaWgr=iyOOZWsQn8-L<znbeok8u|
zC9sz~?)dJYoVG2FMudL#v#`|HaEkeWqx#(`je+lq14L))q@@;%(%BQ;p||Fo#~s<t
zZ=E(XMAr3!r<=VVkxQLaoBBX2IR;~fKVG7U3%lDs5W<Yjon|%fU7e{+$Ac{-v-5{R
zL}v<Hwwh+bP`}S*^%l03l3*oe*;NN*-v)!5_D6UX4XO2q?WVJZld((wl9>`n32Z9=
zQWKiLir*e<B)0!tF{Id3*)8`V-)Ao$P4rBo;BU!0eSP*#6kq3(xM`EVm0UTjt({Uy
z4~Dp;6d&r>HT#ePzp#RZDIZ5ev{ZfLdR}}jeGV@v`@sHWrF3lHbDw?|4+lAfGQ6&i
zhPI`d<~;|#jWE5*iEwVTTHQ{B2#gUDSdnVZyU}7}awm`O0q;}m@liU+^xk`>!JRKM
zDactrt=a`2!d%+PrM=!_E>i6m*U0ku=1Ao&Cf{dhjr+NLS9BrA(Bby;GgYfq%{OxZ
z<)YXP$UTAgQO+pc%{3~x!jpjYinbKh2V9^1ti>j(#PaF3!*jF!kC4K0D!F^zR2zq-
zAJj9Fw31lkM2nQY>W*}qqstWHbu_=7^@84-WG?^Io%c^E#WUd*`b*7gf?dn$A(pL@
z?`G(v+(H&9Gka*jO7{6~ieppHD~6k5o)TrSoqw6Kc&R`Xk+0X@17WpvDKtzqb%Q>K
zzDE!3O>`)k6sqbT$3~|qXW3)8sYLbJEIWrZ-FP_O&jnYJ>d`>aSBQDsa;EWAV173w
z_{i&a#UI-8AhUkU==IDsS`c6XM>_SJBpCb!u9S9q`g+5`L(a6qIG^RCELlL&qKI%F
znn^4lhgGmimxLR#eBMmj)cN8}$;qZ%v0bq|>|bDx>Pk<k2<x>1KYfm}@q}0$j(>BP
zQY_97MaYi5ZETr15oh1dT!l&zg{_v&YV`cXnLl<nY;d>l)rUp25%9_lR9u$(fNZZA
zse<6Mr*%vt_UGzZyD5sj4c2}=hr`}`HFa(G>029o{_NGRf5de~<y|}mYM9B?td-FA
zRTdJ!eoiOxzi|0zcIP28OMYOv+;>FuSfEK$jF#{Hb^-f2s$fOWON$ZGqPD$~5l4QR
za-~19CY+p7#?9a(${cpH&p8xlx!z*r!>8({^7kfJHU{Ebv!)H6f~DC|L*fZ$)IU{E
zu8HRvvmDqeO5PWjiC|n)s3pM?)bvhIjdH9F=FLF=ricBV<3HsEaHG=dS+`wtcVr%U
zWZ(Ad=UpT9%xpCI>RirK;`tqZy7zgJZX1tmK7HM){W`{87nYHRHZ$|5CX}SnDFNup
z&YL*S)$Ca572h1#yn+4zjGJL$mWnbdp>$(Cw<smXbw|oMTIa~5^2dB7-&A2;SFVW}
z=f6PBX+#(TA#M>P$AJdIlD|pxff5)hEMpovPo8IN_q^8#TTa&MpsyZi$5v5;w^fPP
z$r3cuJSwp4^oTKK{nf<!n<Snb>1k`WQKwx^KO%Zjs57u%WFMX^Z_4*Lrv4}e<0fpu
z_1N{0g51wFf;VFujN}=w(G<VIYT4hGlkVN$-d%*0m0Q+nsu-}qms;eonLLkC*vf3&
zHONHL<)<`_{mdnGYVHshz$#JE$T*XGF`I@bu8Bek$*$Um^T1WHg<6LSH_dh@N`?5-
z$+0A0x`yU)<}*p@SM6UXJ*$^Y2)s3IK3-8P<(U?GjT31}!7ICI8WyxF0<b$TO%LWP
zw{mV%7WX&2Ma7C+uy*7a^Kf0L?*(i-C9SlT7S}iZ9D%QN`eoUC%gPcdCul`c`BDdV
z1rFL9>{vM6Tf@E8dST>())T)mK|XbMzalM=7sl{@7DQFB)vT`+miMDC%#SWQ%M?9l
zZq^kMG#01iC{+y0uVJ7HYUJUpCztK0pmUn0V04vg55DLm7Ogco??0@bBzu)mC6dgs
zbDyIQ(>4_0XWk+v4VJiir|q)VVky_&RyvwU1T}AWjveM%rV&;sU3wurBMqH}o1_Kf
zQk9W9k68N)0~Yg00TC_qBOsX1k%1GXtpU5k0qh&+lq>}H9(lC#N;>H84Bx*=zJ0dQ
zeKMufiSWC^i`=FA8|mM@yfHxit@}*p$AkT-kKKm}k8gZ+q5Vxl7@j-Z_q_kZA(148
zMZB|)scS7XU`SdAZg!g!7CfO-uwZAFXhldT*Vccx9Mt=e+!ALTaNzjNdGByyM1S-H
zR^c*8lnr@LC1R)E^V7(@8~6RbYt1ueFL`O2OEAR_g(VwZ9*Ig>x<-gd)Zw^qU|UxB
zB%K_C>ZT3%tGyIF8!~)wP<L^I#4)bO7?+)UGcD)h8!g{&#@>-3)?V9N1U&~YYfC>I
zE<GdlTZ@sbFZdA7;71(zSmfYd{DWFPe*I?zYqTl=GP>@&^A^qd8QW`B#VNKa@%X{i
zbN|&kAB(!w3dZfQ8(k6#t}@_d3p)li-cw`LM!q!-lY+j?bI<(;CJI;gmSBvrl-tIc
zjOIF{PM|;+8}x+19C)zJ-AC;sE`nsnFB7sd*F{&33V_<R1F>2}t1mbDkJ@;RIg?v%
zaT&#PRkGgCvYC5FAGfdnLc(k;99-RyN8ZjcZ$|#<@BSr!Blug_o?8Wlx@BH|zg*ij
zzK@kxoyzF;ou;sxy794itf(n!;6}|Q{iongeL{RLMxPTLd-doTcFWg7kgOU{m>5XD
zf_UkP&`|HiqqbeY$&Tm_J;P&+){ZYgnB6x&fmKy;lv*;4klkjL-@Np`FMTX0V-}Ch
zu!mQo4fzZV9O=c^z|y6yWIw(inzVsFyrjhzzEo@9vUyyPuEg^0)o!+A%9Pz9AAiRR
zX2*|u@MnL(d`Pmn@tONopc)`1t|IP+{5RRLC2_6!xgTTS)93Kl3h>YKgH#s=c3c;<
z)L~MQ%hC$2qZl%df2o<Z0n|x0<&DY(@sd|m5iLERu|3%sBO&>7DlXDQ>o*Ct@M+bJ
zh}VD9U;iyH8xQw4Uq<#Q{(`@fEn50~!72J>K8p4imCF-phId}4^w@<3iSD;e&o~qB
z2z9y1gF<3_Vm5@nH!ZbZkbgIQfLYsr7QX_Wz5-mBnW8F;Sq0M$%V0zH!!eGBu^4LA
zPCq!upjC{;-N(cSbV)JWT`9}K19pgM^3-)s^}E6IH;Tq|sFJ8v#?;~$^P@*xXF98_
z(5bLE<1ZQ+Y!nT??CR<6POKRL{r!5b@~GL{rY+-twvIF)VyOGqR<*|TV|3rVwZh{^
z-%$2Twg?|{HB?yFrC@8;8ej@IR{WCOc0Uv4&AX`NZRg1Tgs|K|a;9cWO_M~onwKpn
zDH7F^irP|3afe<!Eg78AyfIteYs2U9WimB2y_}Y7!wT2{D{qmmwI)N!Z`@th%(DYg
z$WTAzI3R_FVwBf$G<*a9mrBsqg!lcHb3P|(;rho%VU&(<w<&}e-%o2JFf)&eB^h5K
zwg1A;KWs~Y%XtMduzo(;Vx-|16fEWC1H_8&71u(;?;0d4Z0J3z^RO2<pY~fGiap@O
zo2W2%zf)>>#Sr?fw_I)#X0opOdUK8oz9P$+PHR+2jA1cuIEAE;^O5EVg>CAoN8RKz
z%F={zzQ3y}EvG_LWZbGt3OPZy2ET`|%ln~f0+83<-}E>-)UUfRtmCL!xd=9t?xEPs
zBerx+A4qA)l6W%ZBia@Id=llsy#0ZU2eF`Igv8ANy{E4J)9<ZSHQk*J@3qYtNUBBe
zK9PUiq0`h+`t==Yg9AA#DvQ1~pc<8pQB-_#44J})8G<5Qqcp%VCJ4p*ZoA56s?|lR
zym9=|<gn$xMh>zOR#N8R{{AEf#q`|f)OEimZ68NX&0O*_iKih%n>opvN_&f22-=AX
z8@EIra;M1+T^~tm$sM<dz=_%nDGn1M`1nRT>u{mL%23@JYcoe1x~y-z7xG1X?m_{T
z{F@{b@pNquc#|+1d1gq>vN|XHo8;S5OdXT$_WtQzeC>SBb>z1<Q%__ojs#hzzXUCG
z3FSn7I(kVVSS>gD(HE5g8u=1$lCeeZG4&zifKiD&e_HipBt>myLoiG>J~sUwym@sw
zu3E--N}|9HH)??YO=6Mm!KvqPasG|B#i~c)iW9SQr0GXj1#Fc|gUQIBp|^8`GSS13
z4bwbE!rClno^~2JG{=b0Dj066!@*=^Zw?YKa*zlvq6;<8Sjv<1u#6{tn{8<btA=wm
z<&=!Y@Ma0NSb*=b?_BSWYI*h*_dz*m{<{Cb=zA<7uR<=V^3h`g8}Rm>oJGp)L%9Kc
z2L3^WtF2{H4!flrFeJ&9{k$x#a^O}H^Bu_Gr=Xzi+pDJu?b`vz=}y^p)Z&3t<A9^&
z4EfI-mM;ldGcBN87Jb{HK+`z9-IDvJrD$<ijy|P>*XtP@`Afic43P_m7UJy%UjsaR
zTmDjBa>7r5GcTjAP+_pXBfev?hOkre*xaC|rV@))w+ht#!|l<sT7W%*v!3)%jIhx)
zpV#HMbC#3EbGeC9_OyiV6oL&#;nH1B!QjYT+|~Tg0!{tV>e%L`sytZ@OXm@0{d9CG
z`wb^=C>lg=)DGyE9QCHdM?1-_+EGtT875dZ!ZKH7PPT0QD~t?ZWuGRRPpfHPTT&f+
ztB?VUTE2$lGTRAS$_7%OVRCm73ZdUg7d%jX9+s-BI%$?M18-8;X=<9Ba~|ecN}V_7
zk%{?yw9rkysLwpo84fOc)L`LsA~R~osswz|@YynwFz{m5)CyXK@5pWoYvNSHJo2pT
z2h=1l4JGkdHS<(cPx1`mW^4R4&bfy;h%R9ODZQf2d0bNz(^QzQ34ATz^4&5R#qtZS
z(*%GpUfVn^Us`rl-HkF=gw~Z0#Kxst9DSNGOpRN^NAt*joKXpH_PQU7UvJRmJx?&b
z&pIFgO1aSk_L2~IdMnN0K7SLuWx$bOrHiPkO^7pTv55C@Xort{_FdKXq~!SQolV2S
zX#H~W(|qA8dZtEFKdnk<)S6MkShvvRdHhV!oH5-qD;|HeU;M8w$uVe3THAY8-9oy%
zuK1l~)H#V95H=?T*%^HgO5A&3tnb54+1tdgy4-5}oI+ekP*Aw4enbm6ts}E6Q0}|q
zqOP3MwMCmaXmLKSlKeJGWXCV`T&LTU(WHJ;!16qQI<V0P-o}_OPO-ZsxX9UJM8Hw4
zFKZ^T#=Jas6BOiH3Q=i>NSwZ3{Rb;3MsPmToGmh~`sw_~*NM;mLGJ%YOb_n6f6Cx~
z_A_<;QWtf*(HndfpUu+ctoLj4<xNw)5A6Tfj@_Y9SnG$_)*L0P%P!$3<POa{_>}Em
z`^b{1&{lsLynlwk)!pUbdDV?bngA>J;nWJuQ>h*RPnvnzyWFBcOd#MzcFTM?FL(pS
zd8cK`IDhv>$5R@NVmtQRVS@#v&)nAb)kD_d$|$qzfHM7|GZMN;$el9zVNIX}R3PmM
zwk2djQzTUp9f8yC?o>+oO)~KP6SH$4QjcsB#6mkzwvHz^S^7cPUolhay_dz?uNdI7
zJP<>5J4+*rb^Rq#Cq30_42vvos&>1N5yFu(G30()b&{ReE2GwDe!bmbJ<N$)ws>Px
zL0GFuqdH?<sU43mS?Tqmuww$H6*z?7Tl+YfOwr`N--|jE$r0k(c=Kcn#$n$4Q?-H9
zN{<gzSHRH(7nV7{mk#t4Gp56zV=ME@q8N5V^;y5~%r)0<-bP7W>5@U($*O`~yob^r
zw0@iTAb)@OYo=7WMrEEk*p718hk}*xb1Og&UHNgy^^X?dgNK48ZhLa=XD6miBUXlL
zD|`wf@xS(b$6CwF)SLtd-`AiwnKE-7)7Q^eX;gZhBr*(r3Y{=v>W1G}%q5eYjopw?
z;Cw3qPCdQ75WnneC;tU*z@Z*2*fBC6Cs?9_GT=}SkC`Qn|Idw*TcO9c=i3$Yc-ddu
zw0|VN{?C;nAz8lfZF^C1dFSl&SKFVe@gy7KiIKaGdlzTT(U+nFB*5CdFIUxxerI#o
z{nO)*ig)r~U`TV67B?o;lTEc6+_fa(e@z^|>vz+Gk@K?U2==a7pa_4>UXxF(J`P83
z8I-BT?!@UG(>hIAK6v`Rk}5nyWjf+LdP}!`k=`Y51E1q?9+_`1>v8tVM(UKzV(zi|
z@pmn`6-SC1e)lE4M5EoY)Eo{W{%X$Cl(L6uG+-H}!=T?Jp52yB6E)}LH=L@1bmbHD
zhFEJz3zBd1#+&O4vnElwu{-~nnd>D>SI#>~{~Rn_RiqrN0jgwS62<D=nKQwok`{x`
z_YcNyn@A@*tJCEefiXUz!-0>h333GL*smb?t`NUj{$@EDz)+Ay1E2mRx0842qPN}9
zf1p->1;nB!637g72CuQ_YZwbzD$C?;>LLP?ruP61A>x}p&9jv{2?j4SWqPLY9ItH|
zWS6XwR|hT@khi<%x_0jDfV?=N-Y3Z3RrGQo4!y!?;(|7-1_=-iR~%<8T~7kYCC67(
zHp;^vrv4{4Z%O_?zUk|s40SG=9cY_!)A{u6lD$Z~X+xAso+d}Vk6YOc$8Qo9QPz&e
z-S9G^h-1%RnI)c>U5=_dD*uDM_l|1nThqs}p`w7GQbO<2rFT%Nq4!>-6G(!TfFOuU
z?+|LJQbP+!Nq_*N^b&d}ln=dwbOFIH^UvHn^PPKV)}8fTzxmDmBx_~8XTSUGbIx<#
zv$FR-`+Xip=GN$e9W7VnX$<B!L6q)?s0u;u3*G*J0>W!C;qOO3?`nULm|r}W$y8iq
z`OX3z&2OO1G^=i$QEUglV(b3+2}B|qHX&M-n6S|ilW(W1-xC$iU-;EG0k_B>=lSBF
z{S{Js_D$k^PJ?mn#Yv-qm~G&oqq4KUM7-6@nqMHDZ271z>($YG$=)I}D3ekWYgZpT
znR3<}@dckElZP2;6E4U4$$AA+f?z)e>Cs;l(Wmr2TH<)|e(p2_TXYm)JXcM44Aupz
zW=F6Oh75n0*&(0mm=c};WC`jBT0h}0sO#6VVy<SfWb{fxC~o@*(0@Wgs@sf9C9`dC
zAd=iWHBr%CNfEjhC<NIm-i^s+$NTk#@>Eb%33U3%t63B5Vb`vt1zy$3bf`51x4J<q
zpX{RVKI_@ZY2nDw4}JX&K|vdSJDi*f_Ah%Uo!+R0P|Wfn%$;_>zBa%{B8}iKtNouV
z5w3=d9HTu7JzIcYoL5ISQ+`#%0Wt67u$fVx_<(9i?HGx9A^k%RgVtup(Q85D=_ka`
zQq5qE)n1FOnhws;auf9IURN~l*thQO!zo08g1u@O?o&lkH{(RkZvCjD>T<W3yda1K
z?<xOE*rvkO)CzfA<13^^w2Gy;3zFOtEYp7&;<u$%<gwy#;zNgOd`CU5XBjt!4jKj6
zaVElzRRJo>C)VFegTI2{5(%1kT1k+gT>}@)lA#D&8fpt4zn;1SHG0$am2Nt-FsT|D
z(9lJy_d2a(5+6IgA}43b!M`7Zp#O-g8F(d(ryP06e6#OJ*Y?r7i~^jV69_LAd7nfD
z-CP%k*C;gTTi$wLVmB>SXj=$hrjqiKmhlR2s>%y+!z$kr!giSW^2sGm%r3u;F)$%|
z)jjc<{q!*gnJ$WkrkENeM`l~%XXc#+>4l6*ebBQJ5IG0^2~Yf+AuO@<k!Fxcf@L9O
z7nrilBkWL^GrekJkJ{C|v?VF-CgV(So)x~&Fa3+?XTS7fB16dB2@RUAWGyYW7<<3K
z@}p3+Ijj>+v^Ug|Z%-da>oo^Cj5w?n!U=LkWF?`Zy3LFJFwBqM{S4Ph2PNKJ5_;XB
znvK<)%Wq<X%;JiBCJc)0)19T1YMlkBjM+RQThnl4A5dL&1z<oT%DA`~6crwmwsutT
zOcZk!QTi+w+u>?+3S(E)vsZL`P?12{z=c|9va+Vc+Q`tGzh4h$TE+gXsAMb8Hty|{
zOyp(CsVi?6c6Ce-m@(w9gh*K?I`|^O%rxw#`a~WGyl)jO?JT!mB427QqINJ~{;FUw
z1%yg+%?6Fcsz5r^-8v9v)_weUFQ>uY&d)`Vnqmv#{KPYums!TZ?!z6c?`*5&{4M|;
zQBU1n9v5Du9@o<ddmsV;5+WHk>#dYOpf^3g@oxY1tlWdyu8kY(9fL<^56AlY?NdhG
ze&9NR!XGo72c7~=YL+LgdbyoS9A^byx4Hzq{bfI;7`L_EEECqpBp65Tjg(5Y4eW2l
zcoe2bju3-K8z<O1pi4O9a2g=0MDwZO_0os%C$-#P8Vj6x-bT%Lgx=>VSB+jD0lbL=
z#~1MiTqx1Nu2zUQ1NLg#no~ZAA#G{3Je%K}IJkwqPLrwtIEvlLMK!0383{=G`C|J;
z64(C76;V6(A)d}oZH<D()*OP_c*TR6Tx8=}`DX`zbWaiPS4`nVPG5oC{iede6G+j5
z#-1O3#Xoft^I|D_@I;NDz4pV}YRR59j7}9Y)4B-&vk)Klzw}eSd+<$IN)BHrR4c8?
zp17IsyFMMQe_3cE#f&4<0Fmq>yy}k{KW?A#4bOZFcTuU6;1MeLVohH1E_%-S`DD~x
zh$MBP=~EL7SXpzEdaxvdTiSs>V_vluw=pK~VJkb{kMS$(?_{j4wJ`|Jm}2+B2$_TD
z*r(hu2y4tInVdI_3$AzhHD5=g6LxdmjL|W;(LSS;FfmR(&c(G@xpY3-FbnIGQMKI&
z4{!qQRzE*5fw<Mx8m|;tHVv!eYI@Y+c4<5eb`YgOA}DyJSlYQO%Ph&+&_v0yG|{tT
z){m`Qpnh$yGzujo+#a(H$6&f&2=3N9iN-<oG%7qJ&A&)$D2?Evc`~!4yb!#f^V?18
zNqR6HU)8yN)xt=q>#bmFvR}s^gcP)<_9W(MgR<t{D!lnQed0l?C!Fm%7JoR6%y=Kk
z-g{7n*JfKPugAeJZ%Yp_V~dM3TJ-Y;TuGW5lGAku4P~B|MRC^cl1a;X8)W0gWXuQO
z!bOOtAoc!<Hq=niN7b&G>%}R{kK7)dN4#_4P=m8t^q?%3Lx%CtDNbVH#5?VegSANO
zOc@+&jqwhxRKw}Lstvt8bti}*^=tLyc+HZM;Sn9{Ww&DryE^v66>6xxQV76*eTRu`
z#!QLkRuNu*P{SC%iK)$gP3BHad?fZVLLj&Y)66E1Y}X{J{JPYo<D7wr1Sprj27Y{#
z9*BXe8!q$N#*&YJE}z0w=Dd3`qB0Gv-Db*QAhKz(Xc!{tpF8O2nRY{Bw;|99r1A6s
zd_gpbaXz`w7{og2fN<SZ7k%>g`>})w=;3+y<6qIBtK-|ZZT^1cfAv<QZxOV?)QG#T
zC)TBl4^PQ@7&9R|%2^b!%-MUhkR`{4jlzkkj@2cN(y~xdsb;<2vh_5$HF$kBbi?a|
zO^h`F80Gu8eYQI3&5C2T3Gu0F&-E=(kJ;%)m4hFTGlfJJt5;bQ#z1NUb)#CjKaBKQ
zPHK*!J2?YH2RvJ(rp!lPmF=$cyYpBoV&i`k^g?X4O<z=G^McM(z}X?C9FKsnRym)&
zWR)rNkJ-^>PuCn)5_X|?xm$xE>==Kn=Q>$m^#!FnMrgNPAL&)KHXnhc-6S^3q|jZ8
zt_c!GGq&=q0N4ttvy$k;?IypQY<w!=dKOD1mGOz=swdZ%=2*w?EYO9Lkxjw(D@!mV
z>Rs`OT|>?q5K7*9{{31T&~r)CbXZ-Dq#_haEF>zHqo%wq=)1NwiPF@xu{l7b2(DE#
zsTmoMP6^ViXTN}}Farru=5IQtc*i=k`PX~jWVnqrv471I8YtmnML7<nWTMZc$QDEc
zFI_}HA3FK$3P)dv-O}<LxiReLOwWL%%H*Sq{?0$jRm{xJ&Gn`$+}&Q%UqWbAsw;Y9
zBs*T<kS}2XrHs3SE={|ihuzl$Bp5v>;+cL(NpbAEZ59vcL3ip6!VGE+;EXd9ux@6x
zcA|Ip-@;>RQU<jcC&0=f{k=bW3;9chO}}<{8tEPxy8A`)_i{~|8ma*^*%bq9b+~hL
zwUiqc)CAX4SXe};L-snN0Z+V`<zsLv@urZlp*-ObmQL$@a)5n^sj3;Ma1bsJ$>DrO
zu8<f{Mn-V??Oy}y?;8JLQWgQg-oJIpNkVpwagX^_MNiKCX?0QA=S{)tJ`LNV%xJy#
z&-ro2e&&5q#Lj_wL_GQhI19*Eh_^z%94A!v%OtPO(51x0jv!9h+i<_7*K<jxj0;>%
zBW2bjh(@W`jU+G(fT2BEC-)2WRrc7<dsW<l8f1^#f;qV!7(vsWhIk7%$}|}9C*;P2
znJt)_#^_iI#bkT-NSCCVl|;1JYDzub7LrX}<UC;<`n;Plo(_}H$4(uwo1D&CRNZ+1
zNW)Ll6bY8>azYt9s!PzQjGm>wYd!6>ks<b$6uS0GcAjK&>yzZdnOAXnEcY@I-?D=n
zXSxm*A!))sL>Z+VZtK3WC6#X`95@gC1_m1>N}QKk*e+k{wm9>W8}qGcSPPw}5`U?@
z<`b=M;58d2_K<nlu_Yq+(uKR&GwnA4pr)ttnQ^qiOLYx9ihXm_tL#U0v3G=4*#!jw
zG|*zw;`_1@+CbvM9szSlvCbVg=}v~i9=FJKt?^mgvpZll-*?93>7x1a*%lHhi#;)i
z&cY~UnyGs1_ks8i5#Z2|ln;2|XFj(xD}_uC_Y%VmQcsGu=w)7=pB0S;>5e+?Gfq4&
zx3d_~9P_x(${IKAQ&C9HnJeXUt!%hR{D&`D!6Y=<ChSeNZv`1L*3yeST!v(~KD)bD
z3eUy@r?mGF^*p|gv}I0CAgd7L-wm(7I3IH7%#S~e{Iz>MXmdy1Nc%wWqLXH}%7r(H
z<kLR!G)x`eG_d_xz+0Pn5mqX!GI~joX7wxK2c>REz8qr~QoWu>DXJ$+bXhZT+B}II
zKmgS26syQhKgah!t<xa;>fkS{cn9Nd?x2PfOV~KEdMcxFMVVs2Ho;JeE&L+CpP1rP
zEfORq`C!=k>&O<d+Y&gqz9rEA9_WFaV@@q0@Rf$Dpaxnpshg8U76=3+`ZbU%IZi4$
zkj1Ym4(I<!d6Sp`Ja-aFsZ_8F826Xmry_|u&New}b@jDP-J-8t6`o|zeC?S~abONV
z&}D85CA{s;DjuSnx8<N2r;+~LWszwkAs<z@ZvvYjPayICsl5b1v<dl6Pjuv&?sNa*
zPzUb^?;M84WZAWk8tXejz{O=NfMPCE)(n@22kzV1prd*&>68P%l84?aW4gB06m&e-
z*#6UJdH*H8#xE_VP9|WRy2ve(z>UHHn;IB&YHsO~d5zC)zi$4i4h5gRqkx0sDhd!^
z8-nsR=>*U#62p5xu`eXf$uR9Th0c^H%e405@itx=WR6v_8CsqH;+U`4?J$#Atl*$3
z#3J~7QhBi>VQ4oMr?T#`{&*X+C!fvdPhZKkez<IY=DyZd|MTA6*fj&I$w~Ud*n`({
z4%Tf^&iYQ3dBuv|s`epiB4l0%La%fkEa_V>UP#CN<>dbp>mSOs_)psWkKF(0blrc{
z*557r_baLIJb~)~^cC`pqPbMVq7s$NqoQKJboGAewg|E843b!$iM^)*8lr=B*CJc)
z8g7VAm7O_fSB}j#tO#Vs*V1{Vk7pD#NybQ8+LB(UPNvON2H#Xrn%A9;rDyuO@d1;u
zIM##`E6hfk9scpX{$1zn2W|d6(TUiWnV8hoN68lIoe43z?O$OL8y1h>{xRDpMvOkP
z3Z>(`v9L5_bqrotyE)l>n{G`&vg6(E=!7Fy$_kw+5bY?C|0FV9!thu`Sd+XjnM`1H
zo2iM7mz>1axJUcexAvdbt^r}Yc{bSg4u0+ZG=I7_s2a=nlLxuuf@4T<zx}9``AO*t
zk91(KT?cD{AY;{XN6L_%y~7eoF{7kn=lZjKRW7hCb#cY#PJ!v7CyDwnGU|^;Iyy#A
ziwwq+25p(@_L*LV_3Gp)S=R{n^5$R$fxt(k*Q=!dsn<mR-MITtoquwK{HGB853cZ6
z93g}(wXM&8ebVKyvPqqt85IocLbxg#*>jTN&0<7b>@*?{i|t7yx5At<*oSLrby%aF
zg-AuS=|6WGd-2IB>nKvlH@(~6*s?Z!et3Dic0zKLoeI+FTT?)1GhA>)p6Icx@It2U
zzX!!bw=59K@cr`BDO?Kma})nWG+kZ4=z$$?K?O=ZB!gu=U9l^&y5=n91EIuG%|cdL
z${wI1jlkFR&48iWX|z!vfef#>V=3E3N9pPZ`s56uG5F<@!Ql&1HBIWx?)Up8XP@sD
zARivvHNPG68@x3<65?BB)#&`y>Ba`ARmOyWPQ*%>;LnEtgyPSr_@DEDydxC7d`UH}
z@3K|5X`44%9Ms}8wb>O;r-NzH`1+A@^hm-}B{GGPzPGWJzDaQYecs^<DV)I82=D2p
zobatb3;zV<&tUi;`5@dm4PF@hQ8Hx^x2l>kQ@lILI~6oxs@Xs<4C6RvWn#~YvMbHE
za~9z=A4p-O{P35q#s6+zNaH{La;5*Q7GI;<d6m)6$^i6>2S@)KK)92g5}+D*MYeZr
zcH@!Mz&!LbB+p>;$3wI0cU)V@KaQ!@5?uTD8g%$N)|gQ4lG3|XA)OwWF)0$3{2{ES
z<Q~ug{*qOwyxUtKaW~+0uRE@Sp>uyD^~BWp$Z2waGQ)9+P6n1@c}=UV$1XK-Tum{*
z=_2oMN_I@g=U|Ox{O>_1BIYr1BEGHK0N5ZEU3Tu}IL}w{PtMe}lUMZ}w(b}$SIArT
z>l#^<9RzUB<DXwmHvJ|5U6VRrO@*Gc?_FqJQ$3&H<S_gJ(|$j0HDjzVE8ks>T=PO-
zR+H{T&Yxso^V{yE27s^6zUG=mIsn{yVRL$kyk>vCaA3ewHt)#3z42$We}eF5B>aE#
zI84qE`InmsJ&6{a8y4j6slXZlJhKvH=6W?HMvu$It+b)kEQPAu%o`rEuSHWDYve4c
zU7OR+7)cHWFW=gKy>eOpAk`~>ypVixRskp2<0YL4G22`Hs&ebks(&KzCj$R75eV>Q
z+zu7M*K{Q?>!x<EcFI>-nDxUTFHx+>eDt%fdPHfidBU)6RAkLJhYx@0UQk-HU7B$J
zGH#y!eCPZx#sB}c{*40A>buCh+lifWz7K7G8f1K!fs!>(y66S?4POxr5hOqV?_K5p
z`d3>23eP`vcH_UI@DCaP*g~p*;81gylvmh|T69bRw&Mm1<3G^Ce{GlN<)Hz_g8YZr
z-pigELg&Bi^v_I(`xg}c+Pd-K^nU#;`E)<82ErxF_3quys)rFI+&ZEzx(u8X$+Q<~
zG`vcX31X$h;!X<CI%Zl!RaSA5y655KxVH6Vi)VJIERRnmy_Rg`P=O$RTQVz!WqC=H
z$sQMWM(uHOtLOSyH;AP5-Zr-{G-rTBA22AfIY(W-EK7%*X#_AyLy48-Znl@K4Vf+&
zPM)o}qj{lX)N?CgiE!YCM%o9p)y&m5Xqp=>hv(}aj4%2Ge0d6KX$a2|gG-j<9aRp(
z9kJ);$uw{c>1^`h+#5aw1Z}=6<yZ^MrAX6QpKTBG!X$}!oyPQ>8e}cq9#0K<8PY>-
z{lI6<D(>J>qG+J31slU?KJL!KM@OoT!apXV`PQOQbJmC}-j``M-s)^lHuE$+T)DG4
zQ-UeFq2!~RQ<eEODdb6T3uaxQdDorEzlQSP+5g}>P8*H4V@}fy`4V$wYQ^slM9EZ3
zL!E8E&N!~I3EFmUn)wfl;*BJ3m%}6D;p&aAnceSFt6<7#)0*lX4Ryku!OJf)W;*V-
z#RJmCY}?(cOo0l{TC(xvVYmjByP20%)C@E2#SC7}sUK3KI9Q}xNjuyNUc4_m;F9c3
zKqX!35Nz9sJX@5*!5y~f9QB)vwZ>S_;P39;8N6k2YEgF5;5Vp#XaoW+SWVgR=t%(z
zTjVRPd!?j5l^|MtPH14`-GyZo2O7MkX?RtwcxuGmz#4yd%#3A-&2)t~@Ge8;*Y`un
zlCAtEK9)N-ZI&4E{wLEgrA57ABjBB`fp{;kRPcW_@#SCOCf*wg*eXwdXS{Y%_L^Mq
z-`#<~wRsD%qj9K!+8EDWGbYXNG~8{vvr<pUzt#o}O2uTknF3L}Kf4d>{ZdnAX5T3e
z8;$ZbXw?w%c=yRJJv3#H-^q5C7vnb5$jzLIXlf9tJ0M@;ExfttZJNCKB&i1OG`jdC
zi8WD4>E?R*ArPhY<>T!W`LgK5sY)A~U;!@C9{CNsao}MMRw*l6Ek4Z$;l8SHtLWR`
zI@$lY{Qt`OAqYfm@24P#{p~FKMdkBSF6QS-_cIPhw7%Sr&bu*e@T?nCxH$*yv6IhN
zhm|FXmh)Rka?)tvb{%PjKR-|68<!q*SkTVMGkF`j{uP%uC5(>-h0GjjQH@f2ZZ}$p
zDAabrD2RU(Sjlo}@Y0sJtWNukqWV!m2*$v^s>x2eS!R1mtWbM3$;1@eg_9U`wS?LE
zG-Nvw1_We=T<+e{ol_X8U{HKmY&19tJkXs3vA&jWy1yh=Vk%LV+CNrT2RWlr7z0<i
zjl2fzDa&{@*pI2v*C3*1Ee)6^njWZUp;<~IL89tPPnIAu(uvNx@;QbK4r)BzQCqH^
zaU-r@F1Xdo8heX405GJsvwU=2FdGC#B#dHL1tJ)pGZYaJ{O_LN{ZE8{h?V9XPCf71
zNE*$XSyoXGGHRvmC~-tA4N9Nk4kJ$%rGz&sOA_AG+M2SUM-VQqvPg!qEnD{fX_o37
zF;=GBz&-&;u3{O-A#WX&gZLZIN-nRL&kWVd2nh>nbGhezC51>IXOUuGH?KZ^rk9lo
z<RDcJH`P{?ZF0LgB=ck{DEi1y%%{Xu^v&}~Rgc!CTEf-cW~m;hebp&4o8^+CUA;)r
z{m{Znx%w<f(OwR3@v!O>YBeXR%yzNp+Qz$X#)@}N1fn({7-^0VEq4;^<~(J)%#uA*
zm+Tf|^6=6ys}LFOdV@`&+3++P1`?LLaz|*!gR8A-lwGe`8j4Ib)yf6dS9Hu53u1kG
zwZ~3uuBfR{(#N%d0#Ff2A$XR$@h)p_n&bnsLMIi=nHN7JKt?2u${HRPCT$*m<te%z
zoW`D3FLv~r1#>11LLpooE)*xQpkH2#Z2o2f7G}GWaAif-T<v?y69dA5)t8)oGi{y|
zw?-j;<tTM6Igbokji*Yqtpio~VVgk{t@pO|H0UOqJhkM}w@*pcKyBmGTH%9fpC`c;
zG5zfV{#6SN@ve+LdlvnxlRAS=D~kGeVJ406OkSUTGY`X+QrOc++)sYNB>axCTlUpV
zu~JWEP3rG|`PSfesFbcWo~hY+)T#ZPSalF&0x5k`cV&`PQuv#IR3mERUZj4%ylf8G
zWNF8^%Z;q3y(c{Jc(*JQT@`qPzu!)hoP~v1@cP5T!oo%te-a;`5){=)c=Iy>vz-4g
zN71B89)u()uB%JL!}vP2U>bx;rBmN94s@*wf7C10_$3=qON)>R&iR=b{j^YdW<VjE
zE1B1B(j%MLVo*!iqdHT&@7a%NqwmMc&|yJNr_fuxeR(4@pkHEkoX>*1Cyw3orbm>L
z(K#GWC*nK8<RU6HRe{8{Y>(6}S85G%q?CrI&RZ*}R~%?U6@@F!U1YbhJ-ik%+o?6;
zNvt0Fzdmic@L;zdea3t(m(o&C-gXjL20AguXr=j1xA2EQ!tB(6-m2<@yq}pFE4<S`
zFtSPSRAY{}tkScCl1U^zuebqFtLb7Na59_3V}2e@0a9wWry72`kqCp4Ptf+*jJ;bm
ziJ^vLn)cCL3ZbQTQlr3UhGbS>a_XK{S>)E$M774UiWr%Tf7@|^;J;_rSU`_}ANo+O
zG&rR5Y^6SZ#w+V>Q9}qlMh9(E7Huf>`IT=;2`j+l5Gts=^{ix}Hy!kp=-y@buaWZ6
z68Fx}wpii{YFWazT3e)qo^jKY@NL1|mjk68jwaYz7P-l7!|h#XhYVOvu^W?YK?j(_
z%%vf3sBgjPN(O~xh?fWmnLi#HpP7s?oE<WMWn^IgGQe=Z_zlD*dv#=8KDOOQhWb%~
zYQ&Zn4~zWUU$7Gtn%EdK7$I$>BXVl!R?mg%aXK1iy{%*G9WJj)FCoQL8Gw{LOKeQ+
z3ysyE6)|BV{Z^GkJ}mTu=0T)AEmRcU*rZo67*dH+AM$sshg#)qOX+N|l#^%(VxIJJ
zo0gcJO9*L?W%_&G_ppeLpRjEjY~@zJgM%k^zIaJVpT!2?tm}=s>k4or)$W;xPG2NS
z6j!_M*Ll{s?>_Gq+NU0L>n3V=533ilrEs5muL1KkGLz>Q-9PFIW1qJ-hn^&1A^JnJ
zsEyUfbzCTU94GtexT%A<Cbdy$HOJkNi>Rtu4;0ri^KM@<1FF5uq{2)XPVsBPy5^ik
zc&mzv)W^P)!>KMS7aFT2Lf;_uZ50B$cxFT)DzMHWRAe~FA4~5Xc_Wx)Z{Yrfp^%RR
z32)F_vJt9`ctJadX_L6O>9?26@-9xsc3utR0)QU=+@T*p?_ZX}5B(N@6VSSc4rYl3
zL=GC)yUi4~I2EwrN}jY`GY4<y6tbH1%LwG!vw!*Kd$<Z0^)9Rr&Z5hIOHsH>wP7zs
z0=0-Ntfu^4*i6@sIa?a9rJI7mo)%!LG$P!((0XR;z|w4M;_>BFs1@lPKmPS1@jEx<
zPiemj4YV)3@}kpoMBDZi6gK-rC}K9cgTq^o%U*HhsVmODl<0KA=zXt4{YIfXDrHTW
zg5&AVIJg&l=qX@UiI*!R(gb?EKDot2b;aVj%3*D~t>*=lNYDtu5O>mnK<auhE?tSb
z7&|5F+J>IF<V>(FWayMQ&3Z~yF{HDq)rqR}`KZ1K^t-)MGB-B4Y;|ISUHXY9aJ)xV
zf{fghh4}{Hd}K8eqg8k`8Hx2#EJ|BST>jZ-sb(LuP`SzUy+yO1+1{6-=fFqEej*uo
zyH9=6$adRG<W9#p0=`r~1bSI1b!$lkJ$vckeuSh#;c6<3jZsOiRv#VrKl4&+z&tpZ
z7jSCK$ONuCiC71fWv!AbhveBe@SR}4`-2N6O$5vVefQmHhGr7f4mH9T1w;srJs3uD
zvaMoOh#1AdKoJccE7|YsWc+mXoF{_I&9;VzcK}H`a=QZw@N`}Bq!7P_h0e!LeH#wA
z>kKuPH%*77!UOW2^orkK|FrrHa}a*}UMi$Fqp#C1x7JX+!I)nL;9Gg5>3%ca2NvC!
z7ZaCoX|#x{zD|>s<-aC373pH!T-C9U(Pf%6E{NKej5&SoA2!)8a-}jIPR4EqOnUms
zofY`Q#fYg{*1_;|rkOR%>;Jf8_%Efxz24d5D%r~}TBk*fERX(n#lQ9tX1d+H44J;X
zvXl5t@VNX&DJ_EwxBZUg_bJ#9d#QPTb}bWTR_~1UMAa<dM%{GM5TAeWK>VZoMu!mz
z+agBahh!B9!*`X#cIUnDzE=;}yiKnB3Bns5>fMVaUUfUYFFizYwP_bNhCaZuf-Jxc
zGCYgV;Cg>3$Va1IM%0)^8U}C1IO+u28)M80l1*-pyA+r0Wdys#Hp!Y5Iot+TFFf&q
z^=rb~gj)xUxX3L`b$?;PfjS}fi?8ia@GV`qU490S%J8+!kf8J$;QDu4P;lAKe$syW
zCVnB)-b$m&956QxcAmyID|%6gKBiZA8<!)W&8$$YYxQOr&!cRO#%`7Bu-#Kl%wy(z
zEkI(@@mS%to#B(8gc;5b1!H5f`D)$HpF2t_-gBFWZxQ|U(AY9)10J$ZxR7K*-9-R2
z;K2gIB}nMob2;oe-uq~0t7mHl#>3ClF6cb`(ajK7fU2wNo~-CEO(%JUHwE~zp3+_a
zL&vqj4uHW(Eb)bt!+4=`#{tO8<o)MfQ0>r1>#%Ui#jip;1(U-EoaC&X!LQm1$A=dM
z*gARH1-IOD##Mqx;2$PX(u-?367*_j(VKkK*-ra-EM9x>;B6&xDeCcRV!Q9Kxn8Wb
z)zd>=m9?k(5kHHC-rLpfYcB`gF8gvpp6a0PXckW|uhC3*VwghhlcKE#8qau{Mhj`<
z<bE}rrwX^r6=GqzasJ`2W94t#|JV!<FY*meW{u)-vQYT29Hl9~Rbv1ZUmSNR;C0d0
z(;d$f65&*G77`Mn&X%?AC~O|q6T8-+lMLB(A=5{_82n^l!>Fj?8{Ws(p7ADz-1J$8
znrXBPi_2!RLxG15;=^XGrzE04N^g)?eMZy-`N;jDx7xF*E?!XOG%ZZbBh`KUD~fl5
z%a^&8WB52w^YFbqa9#AP>V~IE3#SN6-(&C=;Sk-kF+iEJPSy5=;cF|?Z$p!6oy>NV
z(e{Ku*T@F@418{XOGz4A<GBfRX+0!e?SXhkKHqJ_>BPcYoU?8<LuIzM0LQNxkxJ-S
ze00UWd5#6o-=scvE3T*WpVbwFbyc~8i_vkcBW3%baO>rQa5b?>p|GH%J|=EShLX|q
ztvp{dl=v|$T9KmkwRqv@kum(KTvpOpy*-aq-J7y^@e_-uU<dY`VwwKKvg2YS!^NR)
zC4f-TFP)=E4ezZ_*W~5sZz6<Vd23Yyse?k#*kaF#gLVxRc8=&bId;<Ob*W;egg7-`
z=oGAYKuDRSM`;4;c5v%iBf=d^MzDGDvX<x5LFtf&XZ?z(6?j~MVqV8a>V{ThLmFsZ
zsUm13;ztt`SP`;@zegh{JXC9M(w$8Mrin}{j4$*C9F8N}v5UHi{>mnrUa~O3!yPtP
zW}$j#%k>mKC4as!P0MjHP|iEtFm}gkQ2Ki#k|KG}%|t%w>2zZ(B^FX6!4>!{KjAmQ
z+p?O-5oXzl{1>U8c1!9H7kKNriQ#`c6#tFqMzE*U`S<vM%yFw-gU>?R1ow28w3<+c
z_>Lry!?b8&;fkx&JD+6;#gH&2#o`@NX%%0o(CBG*NIoBpc!6GTJpH?}bT`Sz;4~vo
zgB&8_qy<!KfKr{ilZDrfU<H?FV}6MHT_OYKe6~#6TZZVH?W*bTU@-7OzNUsQ4}(vM
zDs<aR#FKFHRT8t71?S)~t(9=7m+wyNg+)!L-GnaZ@oxg=GNh99*n8ju;<~1ya5oY|
z`RJh)Lp^SnYk59yfLjP#e|1^Jp~A1bj-BeQ+Qck}lIb>!oakBPrX_cqWGhP~d|C3#
z%1P4e`sB}gn@R?uHf%3%vbBy=CQ{bRSolf9kJ0@VpB0zGwKaL_N0`5RO%y**(U{($
zd?kj?ML+|gr1w#{H)SOleZx%u5Y;`Z8dxy;0ed@yVg$}vMU;{l{)R(IaJ_Mu{D9tW
zeE1XfPZ^iwT1eg-$+hnR-cwO!Yua8TFAa)l6vyx(%OX;yCrGf}EH{5Iyp{;oe>>Dh
zzeLimKk3mQ_Sa0Ao?XF#%x2$*u30+&smJ9c#?r%vbFflTO+5PaP75!LTk|Nscly^F
zO?gGqn?Ch$%$VUm7*$d_T)eAm*%4F{Y2r3A7&gj|U8za|wuQXIBWa<u_{`R4M@E)k
zBK&E&iO#pZpqZP7ELs8i_sqSf!p7`4*$5R`O_S+WO;9#1<$yfFkiL#QfCQ_M7Tvwn
zeELUe1EhB%6(11CtbRgADRfW;($L2$Wjc;jNTnCVng+das29L~T~-ig3qlWVcENA4
z{@}SRZK}RjmRb8va45uogIL-x0K61hd&ZzCx7PmU;nr%wZvtD)MDzQM?4QJpw}-*i
z5Ywk0!PV;5{JSF$WeFM5LA<x73iL+f9FYPkaBU&3O-sY0aVy`;vjg@nvrvcndk}P?
zw5JFiChdj$$ahNcxt41?zekc-D%#*!b`~lk6D?VSjhM{`6vKQIK#_VT1>xQPB*yv^
zuaw{Pyvsi%DVkS}-qn2MdN(a-*pFT|95P{<qiy(Nbh^*hLu01LTHb-TDVY}*Y&Ruu
zfJvZwMPF%?izkg_MhH=yW;$m(Y5N9EsdEJ8L}%NHN?4A1O~vCIm($h!iHE0>1vA|=
zoA{i6U~Q6k*Npl!WjL2nr-tRudNI#%xQiEHr-uQh61kr9s>l2~6p6)D6xpjLx$v5_
z!TiO^m83;@J_kHG>B%iNHg=`lRmEw)Z|?G|Z#^j~XLf6Mk-pPAH`xm|Nn{XX%GWfJ
zZY*ZP9j;kDO+5!?V><Stn8*HPsY&si%P{t^do^F&xv1lC%Ely>$26~mt&9p5M~>Ch
zo=cy<f9ytdL6)7g^i)j9t=D%ApFm=-P+w^ge_y=|gv1t_tC5J0sKN@;oz_D+GHkLc
zXOUDE1>mfc)pwEx>K!TLCOVxaXzGSp)uv2Hl6l&_!Eka`8@K(m4q6k=HwXt)W=^zv
zq8dr%x~H7-ex_h--jg&?u>5e*(fd#tf@BH(U4CYSjwa0@Qm^T0m&XvhWHJJ(4qCG>
z?xc;YrFC<DzU=5Y8nsQ+cXKm&(zD-lI+JFUhI+K99XMh6Ax-*J3v_Jsuu#?9ZH$x3
zqHo(bVrn35jh5&bvmN}TTWQ^wp3zkKlS^kB!w50U=qAjLiUe%RX_lyySpQrPVnU;7
zQEl+9Dn}fLvF-JJpxz^1@#Q=ZH7b0;t!u6V`-q}>KZFYfn_Pv;A*sL?{4N^T$;yp_
zBvJ;X^!d)jPpZqjvdV=G7mZZP{FCkE<H-uMK&$XYL$y^0+MTmh?ntU9_@0sXTmG9o
zR<dop+|$>*yZ6#%WsX^EhbLrFwG7d*FT@Ok^2eR{;t_1pfiFEd%nq`MY2{BT<92Ib
zxbyU9>mEdXPlM=3oAxGm8u+L7b=f#j1-Gq6RQU>036eK#{!FkntwPd9t;wRml5{E%
zj<>&LPkS>2N&w7PSu<PXGBG_7$O9hH+*q?`mG$9jqGYxM+4m}DkUE??D^Fp8`|#cv
zwo&1>q{kocWr31ol8LYS3ceSWhH?!EcBk00^ZKbA@K6-%I}~X8B*$%6>!d}Eo@WRN
z$vRHImW&#C$WL0<`a?=`Mhj3;=}6X@*cSxgs!K4{8`UxO(ddnRT04j|g>_^I*C}CU
zOrQ2~>rlB_W1@>TzV3)`?ixKAD_K7yyc@KMsc<|=Evuo4bNBJA?bWb1NcRsqHZ5pO
zL3ZfC?Jz^ntM$~}9*~THYPc2lOSXK=TvcjDXv&*9t=jmOY$s(!6e7({<sI-pk3W@^
z+VB}j73mypWtlEfXKR&i9BAJgMU7`YGJl?4_kEuyR-xw8i9)@m<;dc07fz^HP_k9b
z+RU7$_OxThfEqKVaN#NRAU>xk`q>0dqCc~axZZK*omU^qOeC&XTsC(TnykOVH?GqZ
z-WtGzn@Re(wx^6@pX4+0JmE}tIG|B8)GGti`6buRi-c_W6p?O)^j{M=Aha%gztJ?T
zer$F|$*p}2`%TcuNa*p0f2}{w2yu{;Wr1}a8=b>v=Q+kJVD_p@*k~F2+`R(wN3Im6
zBFBBQ)JB;7QM<k);L~)s77Z87*(8X^b&s&aYiIS7S5o|jh60a-Z>G>o7$x_|*s6%M
ziwD0#wLNI>a=cw=Q4^TrNzQu`wX<mqs+sCbx0N)J;`6J`Bwjt#5U$zYp4Wn!7NeZ!
zlrXwqE=%7q*V<ihcJ9^-9xd1CW;kIvFEyaWcZT@-TW;`VNWCc&K4Gciwr#6SKpT!?
zb|efP<@LRY(wf-YvfOP7y*WHH3b1)FUE+SUCa0Q6S`>~@cAVt#JNXH^@x_ML{`IOe
zKwhM&tyr*$6}*t1(%(B&#lcd-(O(rdpawsp7RYT}5`b7MrWKXt+L8*oVtxWaWMpKN
z*4N*-usbjQO@HmLH$cgIa%J&yUE|q%+AowC|JwDpZDPsAI~T<0i(g$HR~+UyNiiA)
z_(Pm$h(<XcZP`rp?QxQ6hgYFGR>|CAN&S_?0k(k%v<J6<xD6{$kduD)e?sZ{JT8-Q
z=skxpHa*yM)2nXW^7SKr7L}e&KxPc8mnf*f-_Mg-R*vo%5H)D0Av+zLViQ?IIOTZs
zj&l&AR5wAa5Lsv(QfTLYzgqCYT95rq=x`!>Ey;pqPeTK|;eS*9WDFL!-1la7Yo!ip
z2WX+I%Yq7lO)q2r7?;#|+c%!KWK;z}c#y*t?q+q)#yEt~qTLEpn2HV`r7V6Pb%r=N
zyXGqSwJ3OhbmM1S2zK+4arS~>(m>K@P)?-+K85sO5)+06+-Zr<rct<o&Dd<A=TvLp
zNp_9mMsD6O>LSjXTJcLhL8_iiP7rrHv?n6sLvY2WZYhPW)D0OHW}!F*2lKNJqyABQ
zuJ@+;#UMG7YX$X$)ZnPJp5acuV72Qpg#&E?rp35x%MubvZEvo@Cjw3_ji;=OA9|(j
z&xU>Mo#(SUSN+h6eT<lthh6`H;Cno8cG#%l7+&K`D4F{c6B+>)rL-M{U?6LWui@g^
z<*SM5*Al(c4o%udP<{ikyQOJ*MSz7=jW6&0x%SQ99vDxK^|dD=pMpd}sr>77nJ2z0
zdI>71C{XnQpM|~3Rn}vwZRnhZQe)`=jo&Z1B{w@9dag%if-dSrswpp>ab25nxrw79
zI0eoxQYFH5xYAO8Rr^jF(@Cdca>dq?38AW~llbcF?@b@!I*MJDDS)mVT%c1)8g0pD
zGu;-87Npq8K~#3k?%pQs&bSaUA7dXnA14O>+fn$}p8r=&^I^?GIh9R!8u1P<v2^m^
zDrFQ$r5j|s1Rnb1IzH911colXZS)<V;R+KYD^Q4aag+5!a!iVLcd3i-vncuGOK_5e
zC>^R!c8G;;Q>r`@;PH*6%nhvv!*n>^dQzVx321EJ(w%*~G-0S5O&katE1u9O)P;%{
z@S2Z0I6UTVtv<N}#$Qq&nf<(>a4Q3Ht-i6Fj9I!e8X7orV<Z$hd?ney!h5ap?s=E5
z80W#u`mHiUO#(a8C=kjzyP|tA6EZE+`{W1WQrDp2=<Ru_O=Iksi48c{u3J^;2D7lA
zeGNCw$Xl$OTB|MWhi41BWSLGQ8S1f9C2XlT;>~<%JaEl1bC|i;T1~wpE2gGbF9vns
zrpbd7QTxzyD6Qv*uPSRhay1c7fSP6-fRYy`zve8aBx(z^w{{!wMLe#+1=6GO_e4hW
zN4~LJ`5zgQ15A_4D;?-yAO8B@{T=(?=h)ws>@8Em`s9bd4@)lPvPj={PsS5{TGXvW
zBb2J`9-CBEz2rvJiiB`mfRMUCqgQ6dt018pLn`+smPpf9B|Tahe5N|GSA6J65lg0G
zAp4k=y6CYOG5Fvl^LRmxCbF?8fxI7ML*v%QtS-URm{^xzV5cf;P50Hl)CB#`;_YjG
z@$o+0Q5mbg2X52V>Ly-=t0p8cc#@*O;lpL3W;yqFG*WU>3le;~i}Jf!Ycba>D20R2
zihMdcf;Gp8K4;I+2>v_K!%MxAxKPH4vSyG^)M{K};r#6}>{E5;C#Mqd3@EZ*P?@nN
zPnX7EW(_Fg_@Yaa{hwTFnt`Tj(YA&+eKRC;eSFa!ALWkMch0VH?u*iVFCT8}S(>_I
zZc<;<bJU$;MKta~ZX+;bV7Nwyp-YEGO!o(k1>N1R>#bAT<eNpV0;H1HWv%Kj%zOjC
zZwud?r^9yHGfgR!>h?#H_^&_Oh`*!730`jM6VB(nkFs|k$h~D>&D*#Y16h4DKk{HY
z@`6~mbZA>lXTi2QPvz*<9PcM*2eD%3BKKPz9+qE1k)s|%<TqH?RN$jd+sa<A@G*aw
z(vtry*xpn3Qj1(y!fw&?=ON}|r&hpEcxJnkDg}AG7I&dQlKz<Q&ojbhZQa3ZFB8;9
zN-?jNs+M%J6h5)H@V<q+2B@D7v;~3?Un8$6+J|?fal)q>mA|g8$|2IcN}0C|rkZN8
zouzB-(SmI126G+Zbfq_;eI4P5gx3?am_<f5hCa2cPDsW3O9{CjfEOgeC-oNqPma^P
zZ@&owMgEZR`NQ?XMfzGav|q#QH$mpvJ!0gc4#St)SIxsA30L~gU+z~<c5VwzPwYJ?
zX#RE{N27mps?Y7M#8;Z+1uwtGo!<TQZOhiQ`z^aP=fwtsXEhG+H`s8f$IGJd0F)53
z>}&C|1xKvp+#{NnnYkhFWJDbWUAmoto7sT$W>vbP2HZv$H59pGaStaUG4nvG{oodS
zcKMl}MO5KBEdn5{TlM~tRu;x>cwr@9==piZJAEG5$ps%4RSJ!Lv7sfAqCxWgNMH?J
zhFsAzc$R89oGo6;0~SPZ%LbW>>D%L&^<tn|&Z4v4Xfz+kB)bIPZ-U!-E?V5~i9$vK
z^};6aZ}~g==<X_|_y)Jy^0uQUi%^_SQK^Q;qhTKNps63m+Ugc}Y32a6lDZwm>N8Ne
z7YjpY8&aAmZda+5?_bpQ{AXa|`OXvhlctY1#MQei+!dwvI7yPyn2WVOGtmSWhz7K&
zMP_0484@D2Iy~P!sjWk3uHaZy^QBigD7A!ZA3K=wH{IDT%p)x>ZS0}EGZkms70PrU
zRv}%Qxc-iJC}20uli`=Ea`TEz90>-^a6#{wz3OnyR(~twR??o3;Dvy@%z1a*7Fanp
zae=?fe~cUeN=B~rUyFpAz_syCs>$eQXKbtvj$I#AW!c(2g0A+gSH+e5i_h(Kh(bDg
z(hS&FOE}+1HpbM4W=5r#Id=+s7V~hS$fugfIj3HA4Ys5kBt7*VPo#Rm9%Vg!1Mapt
zG_p_xZOW@d<0Ujdn=NSc2WWAZxKx%hk1pmWZsgQh8$x<qX~G$C#o5iZVeR+3tvBat
z#7O&Ie6OlNPE$Z;$XRuU9qOe)?lh|5RcY!*CY8jRN5q(!c>0Hdu7%}u;RnEYs{1i`
zjwiDbTmIvrPSEtNCia&l=Yb!U;(=jkjf4(Z_|IL*nFe*=@nIONvokJQ;IpY5d9wKl
zvd;m1+-a}$>t^yJNGr?9xmf*`;rQNy&}+#Di@^wIhwLAhvxe4ry93eL!y5*rV}?(;
z<7NvK!QpSLS16Mz6jk#DWMiKuz+6e>6;y2$*ylDdDq`fc%WoVj7M_~ahM%LcWqU94
zcYzWbTW20-;XOqr4P~FYb6}*_saIwqMD@;Kd&#Zp^<3kr-V7v*%JL!$=JPn+i;3u_
zU4|Gw+@VE<O>m6UjwWiIkS=ygxviJDsL$T^JmP@oV3Wu;k_4!ymcnN6WkoX*;Fma7
z+^v&ac7gajS0Qra^3BxznFr@@0$1Z(-hP8(I|j&Kr$Y><K?JjNr1R*!8veQVfdx&4
zn|H-ASMz+|e-q&P>H-Fu%x^T(gFeL!ciO!5uXz@$5~zs!*_#|IJU#OaqE+xUmU`{{
zz91w?I;LpWq~P}Aw6bhMILpKA*5-)uHe1N`96O~;`RK84<5ltoU$ES#?);voj(S-O
zss6Hb0;Ok=VXl&xw2uDL0zH3wnFt?TUj&9-92i%4i+8J;x~YDh<XAu>+1?nMH))r0
zTIa>@nmwj~l-8qZ9$-@peo4LV4o1{h28|*Hhdg9i^`G39thK4MNjAJ`71phwWO<9v
z{^Xnf-|Q*;V~cD@${DS`;H{c*p>m6`ZzAb|72En>E)N9*PE|jUWBtw;_>}x&<Ean$
z-t+l|;UcqR<9}X|e!G>LiEhrpW;R?x>*kddH2_zbmZ7(3!-CnTvJ?#}6A4UH6UM5T
zcmN7*V8s5|GZ8}$h-4yBg@w>1+|T-8?xzL|aQ6NQ6)L3hDr<v)o%qerU4qdv$PXEs
z80vO&<0MemAA+gLaJw=rQ1RVxu4u?_0{w(>gz&8ZG2n#(-__B%+;0M!R)Y6<z*v;L
z!dssu-}Fp5kti2xi(iQ1vJj8$eAGpCEIpf_pqV1GerU!k8nsx#o(K6`dxAZ*ZcOj&
zq~aHb>iVSOeTpQq;FE?8H`eS&91AQIvyCJ}>g(E^O)fQ+E)!vtO5CH0uX_jBxXY86
z2t%I$<XVL2#9*OrwW&g4iaGscBZCsq(U{zB^vr;JniP@+?Xfh>iSFX)TJe_JDb&S|
zn7**(ZFD9V2Sc|Ej~YyKwSo%~OO<A!sY;~?L5T_FOwrE!y{V9V-l;MF?+{xH-}ZMN
z1>=+Ddbd-uOZ`JX8Nf)x`TD=mx_Os5M?<{@Af8+M8H6;aDI+t2e34vgs1{U-;HnbZ
zVsad=tIGN~CSLgSbQ8lE5uMYBw3L2j7$w?MB2$&OwDiCfJu=(oQsFBU)%U3YI|Z<c
z+4JZPXEuOE>DhEt+4XGb<YTE>|7M8&8^`~v;l`MiQ1*ox)!JS|@%+xl3*Tm8xBMx?
z{2zzXp-;<FT;?K;f1u>Mr$UDFr6?a+$Qp&3n9zEiHDXG)53s<;bP$p!s>?V8V|qB<
z*~8$5R+lKkc90i$?%V!y7YuiS$w|aU0e#lWq=SB>ou$eV*^MV?%v+oa3~terH>=+g
zEK>N)EK^jvHv1#9q=h}7^rLPaEY8;Fm%rCo`A>G<ge7s})f{Wjx3*JuA}^}Div`RL
zhl3k)9`#kJtE7;A_xZVhe}g+B`=T+WUk#dbqsh<zC{XXn%w5WM@g-9DL05@qzolIC
zYIJ+VMPx~Qv_)6rXNzi_=!Sq6jJ#zrM{h=zh_w@;rPyDacGJK!J35wlR%LjzNgg{R
z%TPG(Qsm&E_|Z`Z#pPj2BaH6%H}iDcT9bKS$ja2F6dzT-DLsxwQo&k1;kVefQz9tv
zr#qRO$&{aKrau>*+l&G=S4|8bVQH1Q%TpK^(SjlFj%w+|&Tp|AV=K=Ewva4W<!Oto
zb&0p*KKz?|{kJ#%@6pNlN%0`5aIldO+bfsKoh@)O32(?0XA4w#yH|0KT`gWaR~N{q
zl}_HGw=J5Ko~h13D}1CLbyxAf#1B8Y81Iw`T5R~9;A}kn*fAj3h0vu@{ea7y1sZv8
zc-r7wagxmtdG!6}HwPWu7H<ZFn}N)d8k*8m1nU|9<N{TM4<`Jpoha4>Fao@O!x{4W
z&vF`zA5nF4W4qC=;#ra9qvoydbYrS>KsNp}yftNos<xJAH|H`Sq2L#zxCe;Pf>fdv
z(;i9>G!YSfN?xV*FoieNfQg8$rYP}@cPvw~FBonfv)m;(Z{JqXF{?VP$_Tfyhuh^P
zXe7G37PH^K&kp>g^#Z|u2Do3PJnAVaLmxYOurm3Qy8iU8EX{OM8gAggqSA)53&KNR
zQ_IC|j-)}>WL#vr45;MfHh$D%DDxa$ik^`FgpV{GT{J9sh7*c0Ydv-FY4VEIUzYO!
zZh;!)%z6vDEm>n`8?_)r0;b`h`;Y-0&R<O*7l}i5D}JA><+A1AdPV2IHvWedP29BT
z%&B+m(j;c7<||VFnSh%2UghSNnRE>5Y4U+}$*n_^cQo|{iRdOWRcjTWQ<HTBLR6Jm
z`{7!QZB1R~9Z}wxZ_x5UhSFd)4bz@kw<3|Gq4aPf7O3n}1?r(P2m?o2xQ(c)jH62t
zsV*VW+X*ga??-W`#x>q5#Xq7|N6#qW)wmMkej`~+Mdr;##mjWKjrqKZp^}6*d3Hus
zMc~tCW+8Z>JZthWrRWl-wDxkw`e>45@o2hQuaY)^&6XcoP*dIIj0lYHD6J{98LFq6
zZ4w@<i|8#C&r};WkI3)Q9dF5mgwhZ@&Cc!YYtmq?^J?mnXGB~LD@}qL9|PKj-K*>`
zenQwev>uiA^i6>(leOQalmwG)XkjnPh4Wx*l*2;z&aZRaZ#@1@fS7NR`-Z-Xy3v5A
zS+r*GJ)1KKtVLhGA8fa!+LE|ox!ECGQW9z7YuzTfdD_qFZa#iNuS={dTXc3HJ949@
zA#y5t;uJ()#kWvn`)IV2llr}gNI?Vng4LK3yu9+~7eRmtEX8f1?sRkW6XLSN^{Na^
z6XN{5&6K$1!Ws|mVR1MI*79(r2!1#NmUW}XT$AymD!?lwJ7T+}XQ9L=nhYT#iB7H7
zP7vf|&@Z=cFwx|WcBDGSaHJ0E+f%h4ag;)f_)BxAM@_%Q1<LPpyht4x%Kx&c1&`Zu
z=JSh|j)L2fHE;)#KCDLo+`YZ>f(zXw-Jg<0;ub}?=0!Zi+jT1w#ms@gpDejjZVzK~
zi#GYbkuxCAeLig%vZ#I7C})ZNW71OUNzU++c)Pi(ukY%;J{NYTR!Cvr+N|z;5?hNo
z0&ec5j?9YNis5;#sa$1&vd4$y&Uz{5Kf$C2cEr?di16aUIi)HbTE;sf)`e9fTjGa3
zp~e@L0R1%Y*w5u@eVPRiRKVLm1$Y_z%CdUwzE_5`Kb-8*g{Lg4*}E-Q6^Q1GCt`ae
zUlu=5?Mmw70v-qwL-1d5C%rWr68U{iAVtkI;3!bLa1}akwn7onlw`g<VJ&@T=bgeC
zQ%GBxmY>Ml0ZiXyD^N%zm%qQ|O_cdzmy)Bh4=o0%Gv98RGxEaQd6HV#P*A`}%M*sn
z(mQxIKWn8tJ1a4t(#}M%IqQ{o>6M72I$B`2Pr-Qg;%trDOu9DD?nn>iti$3ZAi^J<
zJdlq!^r6F*a-UV2dZb6Syl;QAR98^f<~53@4&Er2yp@-p@U2X^)7<kaROd}32f+Va
zZQOFbnm2H@ZS!n!F4ERSB2Z|v(5N?>iPc%z*hv>|jk@q!NU7{AS#cT=)_WnK?jd1K
zD)xZCD&8$|-hla)AwCqy&0d98q}jCYv?+neKSwr@P2@IlcF@DngsLNbi4YG}kqW^X
z%dCy3q9Q23hn4F35O0e4X9==9y;Jv-z_q2@t`54&5*7KG-gPc4o(7ZiFWEuQr@Fu7
zOtz+2S(8iue4N)r0_KdWWoNB}o0$o+0AmgeN48~$$MGofV9_OKU5W%Cf*|UF<^yw~
zokPk)6Ell;v&sYsL#^?bjlElyXq;J-dvcU7$%}S7r_2At-g|~Moxc0N&bUTrRCENX
z3d{h4Ql*3*{Hqk{gCT?#7%8EIfOH6gW2FWLsUZ}lB#=OmP7;brF9AacEeVYDBoq+>
z2!bcG);asy``UZ$YwdMjowNS$l3z>mTu+|zxxe@K_IuJHf01`DV?0YpEdhHHrXHo%
z2xqmJHdK)>JTo8ET@1Zoy5>J!*X#QBOa~jYuLfo;<cfl|rce>@T$*HNg;IaYU2C!I
z$mXzP;?#PA`hwybJrtJbbqX*b$diEtk81>L_kxCzSYcGfT{n-!V9wwz!=UEC-~h`I
zZbHQmRQLCvCZ$_ea~F4LuE3gop{PgLLA1T`7W}8Z$gQ|qSyy-5!wT9oW4hwE+?P7S
ze!}kR0kP@|!)Y2qYGZ4s&2$dM#@f~U<mL?@J$L#-OI>(SZl5PsN~ukNglj8~xbse@
zk`+2QM8_63K>s02Ytd8!<fv=bgz}%LRtj}3D%be}YYgbN`cdC0T75+&(d31qT693>
z2W^=S;_(j(&nBTzS9=7Aa8+&cY7Oa3>@ywu{lZ8kn@>Vs@3o}$80YNr^xuV5S`qs-
z)1F8vPPmu>u6Adk2&K&>_ZRi#6k0d5)vx3hWL@0#(P1VF$<juIS->4xgXGGC$Yz$6
zj5kD^Ql%E+=i}ma%_p<MwjJd5lI6na%arb}n96d0wzD5{(MSwrEq6oG`sUOr8sLs+
zUD2qYt6KV^6eHQPvdAINGBa@QnAXR#ZqyCuN)$5yInT8#5i?&vuS}Yu0eN70=nKa`
zSP%aFCDwl`#rFI0ksA*o_d<<^FYDp_e(BM=YvCDf|0^eSy5t-SLBHSksy|h&oR88a
zW(iAq>Im|;NSV(m6@~`b0@*ZDVt|v1$c5gbx~!ljM}w$vSMak&on&DoFJMAA*{<S6
z(g*u2pS$WQ!Z6tVtbYU&32T@j<m5}E{J>MWjxN5!NNINT_uwMcA-RyZ1i31-u<^=p
zYAGp9!o8>bWzJol3WYy+hlBQlwi-8J*>z=8WzK5Zt>J6<%faFIRQ(`TR?oE%Oprso
zqA6Hy<aM9wC|>Ex9iMD9%NqkOG8q9xH*5;+YldiY0Ctc-s4#ZnIN5d^k|AC|hzNOs
z3LDN6n$i?wKG!2U)Th_=WD?3{v7+Tj1TIG2)i=ot5E*k;%{sl%Wj+((v>-B3G3;d>
ztuaMWLo2(ICn^lL1LaqDkib6x^$Ur1RQ2_*2YIEALH%xo*nR#kNjfqjPt51x<oBXo
znrN%oc<To=$$I%kf35GDV$6Sd&S?goz4W%O;Gw;At-ZOLt%Bm~UQ6`F>yyZOp29$Q
z=H>eG8jgl!cq9Y8{apz6MOuJI<G7{;J_c|K?sk#U{u3n$TlKSaB=9uE)5LJ<%@YB{
zx5f#BfB{$kWyQX<_|{`h6Ism7E_Ig<EKL=`lpt2D;vyASetGeJj96d`XmJ7&idHL=
z=xj}ozb)rKZ431TV9Ofm=X98=S~HTqxxeMk1YpPRBCwcSSB08qgf0K}nlDejE+;2<
z<LKY-hX2Ev|FqqFUThr{=M$HlNKt9OT=9klcsK)>K<2KGxTfkVE-)Obu9wSI6+Uiz
zF!??4ia$!YDm&+DY3O*(0x0F&5`G4RoB4=CJ7ouA0pd64tl3t4SYcog=EqP3FU2M{
zIc*1D^+46d(lS+gK1`v1<7I)pl0mrqS@^9eewo{S`sxh$wS<n45q&=@N>8(nJUywa
ziD_tY3w9pbAn;V+qNJ-fI)PTp*Bga@#M~B{x|JKmQwb*E=&Z^2%~g$wisTXEh`b1)
z`ga@O(292@XF$QddFKUX6o-?>q(bwiU74Y0Zm0cWuG4Y0O&hx`gL0plP++HE@jWwB
zd$DN`5!3jXIX%(zra!TD&4zycgoQU$FE=^g8_Bhm-!+P16lQXID>)vGR<tX(P&PE>
zsY?egS&lgsXWnqhrP<&&DqR0+LS?zx6Lfu+Bte-}<*T7~P0rplN!R?~MaFe(LtmIH
za_Hb}Wyqx480!|w7wq@_y-Td5@V{&jt7B!xGpO@g+tvu5hp1oV1pxR--?_q<{NWb~
zj5__#bC1XhJyreT1%c@El9j{h_!cG98I?eQxbE+Oi=_(XFQ4sh3+;~qg~!zXrF-gK
zqn}-<8lXqQ1-PPn)mpeBS<6x(eeF%;vr8Du!CS(jCIeqUVo$7`zEmifO5Xh+I?uoV
z=9K7%P2MT9<7k&hB&R>9|2Xk$R~oSga3S(Q;}72XnZ-JbwwhPVpZE<Sdge_QVcg!l
zGPy&0-?l_a-fQ1a^4SVzFF9th3+{S`LZh+o&lPo@Om5Bz{dr8GKfpF%cPWD1(7Pwn
z()PFqvLc1RjGIJ8+r^F=T&z~~7E#R)9c0253Y?V+o_ggs4B?&QOf>g%9f}@w&JwRa
zUzjxT07aL!TnwmVA<58oc4PT8_^y6a)1R&hJ{-j%_12^<bmYoUr|#a)c=In0VsoI{
z;|QgXfo%}S2UkYzw8B}r#OYk4B5ytYqLm`L@&i-Tk`?gUH4#esy@h_8SfItd&*%t@
zE2`=9Pnom5d%NFGyg^(T_`9IzGbt%SY=w?mK=k{@dOQ;NWCug$^W$&Tl7rTRJh}M-
z8||Jac~+*AuKuaz7wzBC5y(w~R14AFGrA^C<<jIj%gJ%!;9`Z-kAH^koP_gznEAQ-
zm2}}e74`boG$)rSogmY~yVobQT{85pH@^4~aqXYh)Bn~*Pi}`|i{zBt5bxDtrt`ub
z?q7|`rVl$lTo@(EAq#HKx_$j1J0sy+rj5wOXs>DC@xO(RJvV0Rw=QvE{NL~7|IV5J
z9KZc{9`?UG{GXH1|L>mozdrXrnIwF7j<Wt2y@FCY-fx6uE>|v7uFWn3C1|#-MG^|8
zc8<*$h3>}${b#YFb=oP5uNAspW}4vl=PO!kT?cFZM)poR!vPvp<@G9==k+ypYokdh
z>2O;;x5^oK_0Lg<MXtSmS@FTPE|p`SzJ^jDu2#_a6dT0u&s&7ozQ0$T`0N*KPnCm%
z0{SAh6LrUSW1`Zd4V#oaTC~%7mvuC=pI#kBWdx|T=cH6LH^|@D-co_t-qmS$Ua-5m
z`248Z<Bawq4=A?LTCkzGoq#EMl^qomZ$=HchhW}?VJhSdBQ|Y4EW(GfoH-ZYH8&(f
zMVl%}4w`sQrNP;sKHdI#Mwl63_4BirKR;7=Idj=k_+GYPV9YL0`qz0^PEH;AO!U8Y
zi2HxtiT<NUzr(t<9oL(FJ0b7-sQuP=5`X+hZ~Q;LPOnoCBPNm&>)s_Bb${IL|37}f
zfAog`xzfsiyAv9d$M>n-^7?in@DK6T|91HQ!!OR>R6ou+{Z)RB+kEn}HLkLl12iO(
z;RCmH?S`I8Syv$xE3VduE@zkCx*q7DkWp%0At|o?+l84cKZ>q`9{g(hWQIWxZ4E`I
zC5rcF(eJt37Osf6n>D7Ty=dz3i;VOCyOufksDRU>7_}xItT25ea@}O;?vq8+U%UI=
z;GXX2-<G7(P~|&F#V2wr!`Tl>KO}P^-yONc?agF-cq^0VF3ZU4jITU8_uW4a|Fcd0
z*&+T*{Toj9oB!dZCS^-;&uOqwe_RUn5|WlZZA!mqUYwDdr!E+o>|<Eblc%7cvSn$%
zOD}}Zxl2aB|Lu|4AJM43_y@joKixVK`)*cGn*T2E;nmYS&Zqx*`9E9cpPk^p(%+-=
z%~!*gGsG+x7rCHZg^{qhbZ7mpT9FY=8N-)37AQ$+&)Zt(lJDLlTe%brrzQUy_1~>^
z{{LK8?uYdM?uY)TUVd-b64zto7bi~**jaFEaNAjFxs>_&cFzOWz0vD+)vxovR(?<=
zdStNpIRkGJ&Xsq)Sle!h#8PTovxFuTjGAOlA*J>=IcHz_1$SPEIz}7$o&1edKGIYR
zcSh=}!z#Skm}&idv5f;(8>y!y3l_5IfZDNAA#oEjd8?$rv`6Hxwli+pS`#JJ%bsbx
z=79bKT}8EFOzs*49Kj?Z@kj)dpW^_r74fXjKk7(NPQe97l(|PRT_Q%mXB6uFlH!al
zn|Cz88mtU$Dc7_u^%KfwAhAoXI~A1gL&63^$Z`T1eInVma(F2J?d1hOo{7w+^k`kQ
zsKf++{WXyr7w%kk{;RJSttvn=`a#3>z5#53+)+{|thFWTI#vajQYv38o}N?X#WRGR
zs(B6$Pp<NEruJg8k(4ZNMYbNoHo|6@=16l4x7XIxe6JBHl3{UyePjF_&sHmL9kgSK
zL#Nw5^)ZR$v=PqC-PBa6Y80D|9u20djXY)ZKO4RD4zN<_<IY+NJDKX;2{=ZTSk|mG
zHBW5ObdK^oP!!!`6e%L*b2K9DVL@Yiv|TUbhH`(MevWM_-^<q(GXcY6*#r1wG-oLd
zaN+W<I=a#?fCS;+9i(q-y$SO|8Io7i(ySQ^vneY4Oon}RizCR<G}1z(FXD}tRLRrj
z{alTLdp>}UaNqrPMF$5q{k17(#rG%(?s~kIL{e<6r{uapY0|L3`?HupZ5=e~%4DNE
zn?&Tfg@aNZxvuAHDp$$Rb%T3&I?J5cmN#1SjH#s5X0M~v%-}^+vxo>rcX*eI;H!>w
z_fYE?x?x*uOD+Xv*s7+?U`#W)YLgBz3uRo{EuP)E1^x1>LpobrcO10Y6i%rMT2qGc
zrS5C)8WEd8Zx;C)iXUEXL&H_LY&p5mE7@lKy@#B+vI}DvNBXBoV_PHd5%sdf)Do1Q
z-t*LGC0VT=Nik(&#_LhD1mY6VF&wC^ePwqusWxWwjpn9DUDz=}T$N1<Tr|hHcSC}!
zN{Nc<(dqcW`S_|oGYh>O0_R$yf5mOw{Ulu<{@V8k`)F-V#`Soy6bnncNK0+S{x1b2
zrT4~Po05^ch_*9wCFKI~lyFi(D{c}JrAilpWP>I(sqRtfYhB>MfF>|!r4$n$7PAS|
zv|$)qWRJ>H&*!fc{L9CW*^4^PROoI^>`JElgaahn32l>cnNqW&>3M<dqx@m4OPY4I
zgS2$90()SsrDciwc7oAqZc!nVT){<-L>1L7oVBEHM;Z^bw5ht2Bw8gV!|D|otzB3+
zQFWT=U#e(`pEnBY+MhKFx>AjiXs+cpL@V+MVx={@<N!<_0n5G!K#tG=;47nr7Fv0#
z3*a|EllS#s_kQVzfBpDg^Wjqs%65m30sZT0FW#3my?0J=u_9bC`jQf*Kt9m=rOc{M
zmb{AuO{nD+_c9^W-^kMv%Wx$wPBhk>w#FV!HF(wBaUQ9|-3K9edOWb>i;J$Qc`!va
z8Z}&`?_5(6+Ry&|@MwQQne0h%d9ANBdvK_?6eF5Q)iLb@dk|B|-ucT!`D`$!L`Llo
zZf_aqIM*xhpwz{#VtI=A9ZI^naIl83{q2Ni`{Q)_mnE<Df`Zz|E2y-DwSsv^mNVO%
zCvOeA6SbGII=hoMMX`w3M8w%oXZyz)1;gWTF16xO&I5yGR?GvKT}|aADnlZhdbsm(
z-UFlSg@>*8W0Ud=NM2OEUN(vl=$eq17zZkBx7HK4?@29)jBqPV8`|Ak>xrZeq+G+a
zT;~#!ImzgxwJe0kBua~>c4HI}t=sQ?AWX0JnpEp=Ta3Ee-dUtW%TxC&Vx$e;X{g}e
zg@7c_SQU9HWagQ8+wRH?P8wUx#(Cdg$7>RUg_zZUtr`L)Nf(ooof68B82#}@ABI_|
z*EkDrxM2%0)0|8Gm<5OjiI~#0`42fHJDDO>w1y(WK~Eit(V<3C#pb=>grwr@%Lduu
z{)v~5k~U6*KOrRFYVZ{r)NYZKRUdvn_r2Lwz33}0TK#JX&w95T`CKmwb?LPkeHE7n
z_?hV`eAA+~-1B*yY*@^5%8jV_8d6ilm+>Ze{$=b)u&+KXVdSR%e&ak1%$;94jMxEz
zfh0mPduuj%L4A-VMLdtfd5q(!;~L1cnD)m|LU>pJzLQDPQ@oyjz4)u%_JhlqG3iFm
zDZiS7oIlYYC9VXLqI*z^fTKBZNIWFHYvWAP7&F1^>sD*x<khp54YI_Chf-QV=%<D*
zAZ<bHh&g*XXeb5}t-I2W@wRHITn0UU4jvZL)^qCTe~EMm!)Qt%6HIB^W>AP+iheR7
zzccbE?j#~WlDv!q&pdc=y~)GO2|?x6iPz5bmio7dh&UcF$4&Q=la_0}??!#>jrMWF
zO{hZdL<p+={&rS>*tHXhq1m0^HL&>5yEPPq9ijv)OtlV!V<&C(NXdwj2rI6)!hl_Q
zZ{I#;Rh;wPU%785zFN)cyWJa^V(#fskSMBl*z|aK0Ku8@Cc<3<lxz+l6T4ReC1_E|
z1;$LlhK&uVd)g?xhgw9AOoLvqq*Q`S(2}Aq*kd6lgpId1bdbUbgXiN?o!|v=(RjHM
z_yavX3j5-WQ<f{wyn0dC{v58c9GCB$4by@X*ZP%O+&pZ?d-CZc4QzEBS^8?`jY+*r
z=B77fpC+8&yeZu>aUn78)I?ESSfGS!ZUV3!Vp61cK6Izrbz5t>jgY%RHw%TO;9KlX
z`HUV-O<~HO?$X$9ZdSDGgmS03D5hC;QhyhAG}|T^^XJ6o@i1?^H+rQpI&fn`zdeW^
z<61HM^sdJClcA&YZcE17=YYc=Wg}#ZUz+p;@Xop}lE~#aTX5xfkRjsE<>-iWu#(Hb
zAKM~6o%H#3B5(HB`|EL+<M<t?QYXuURa}cb>3n4^crP}zvbV|6HNU5QGMuohoRqFq
z1z!&l)h+(SxasIW8O|F%0Im_WZs6-e7|sQ9ojgTy83B{MWiNeri1~>E{Dk;+!uJ3U
z#?MgybdgaBMLYfeX6PIduF|x<J5zl8SqbWP9<JVWx;2!dF7P1u5OBQg)6e)-(V6ge
z^t+aCC(6w?sZ+;SoX6?`!$O^CQvJ0FfU-`t^Vp~GopwDrycWDEO1V?gu=k3CA65~s
zxb(1(kdJv96}&DB3utbF2DhbXYQS9ImWEc_nT?i18xoE#>X<nWJ8-v%+YlfQtH058
z*?hjP(U!O3B3A5QbL_dADBe9wRRc=R4|Ws2E_uykQSn2DzE7$B8203EWKH3hE_IW@
z(NbFGqK5R*`G@{HVN6%MwJjBqU~i7p1SKT+FGd-BqRBm_NhM!$Cg&2n&Ky|P;=MW(
zWka3DdlRpfWiXcw5f$?uQ?sOLsi?$o-L^t^tA^lSJ89<~WJ-u@owmTR(VfB5a5yxM
z4L9kvW^ewi*0Po%*&W6X>Q~PJVLXQW0^<K=TwXnIB~5v|;St6APQl>3#fDGGEk3#;
z_!w(8{3Z3y+YpNj!bkG_qPW|i=(6Vhqf8{sq_zrF6WAD_Th%o%m}UfaAwe;;sKwd*
zl***!lt*xMlp7_^-%aVVhz7J)In|(r*H@?Cx}E9BY6+QOh=N2xI(-(|d&_YT8k^(0
zCj*%u8%FREx4)fGx{i!JmD2a?PDuK}&wix(>tdmVwvDKuy_E1GUz?6qM7`yk;tDHm
zCF{J{h0W4Osl}*#H@HzZ3wlzfcWq#NODCBYNkKM6dW<3N&kpl<n02rq&?Ucx12Q&{
zhbGF&SA1T-@bm7*9;FC=n}mi^4O7L#5^_?Vbj%Dyp;(q+dWE<TPuJ$p>#vVuMrJjH
zp|wH&JF88-0$6Wy@I#)A=Bm^m4c~t|F|YJtw8R4$eQEjO#<vqaO~9Us9ZK!-9^><d
z@Aj9r6FV+jo?}jWKKtINSBsA_A60t09Yt3&#|&={-kiKiNJ!<WdK_3duDTpka5zT9
z%u0#N6-b-j@$A$bq0`fSsEm&rJ5Jqkcdd^-5U0gIs^;K}(@paHaK*?@_z>2k%a3K+
zJlY}EwmfGVc6L4)m$k|ZlQrK-?v~jod;!eHe>-6`gE3gL3&G=a*ynY9Y*g^KNtowD
zOIv%iydGKeVRqBw`Cyaej7*AVq%l=a2kJ3fW9AfrX^jdyig#L@=8KvQ^JIwV6!Lsv
zem>{pK%?Ho)An^gj}L@FBFbbL*j7^E4Mu4?nTe~k&_o3d7aG{%*CO4{vyN_a*u~Tf
zY^b{<1!J*_QP9K-4-3mxB}rqlQqhNMQRi<bq6QOz*``_SYbH<6T2>*#9%N*GmfrlM
zBPl@>dDnktqx~aRo!$SISx|(MzgGl@wv@AR+VWy)L!p6;C=FX?xK|~bsHjiajoBPp
zWY5a4*JNcSClXoI(p^N#=OA<St1g$Qcy(L+GQNsfTtF^sjH7%Wa5JBK!QQKVvgn#z
zO8TMBd#b~ASKQyF#w{i%uS69zK3hxreYjxz<*(5~EcLX;%0EN+Ydf)xu&77r#~H<P
z2L{0irGWh$@4I~re$AQL`S&<;#x+JLmmbWOJU{@Ton{1w;JvX%Rz3z~Gwg5)GR)tS
z&K;X|Rp-C!7s!Y|!+U#%2xr!oyhnfu9T?q^VUQ`qk!nIb>YQ0QbnmNszUSLR@Wb2?
zOvKkG;9oy{CIlHa7$n8O6xxV{5hFO9>s%C>ZFH(2Fgj`>7Z{;!Z;Op9qMM|0<oXj%
zA;tFDoC^)xdVPr!7(e;j?^u!;xLmvU1J|+oGRL>?=zuFJ{=Ky_YYZ>VriTWldyZyc
zMp#nPD!LdGx&OK;3RVHPoY+`#$csBa5qS}n=w#OE5VdQTq6kmnB-PeRZR#A#V`DeF
z?l(8Iv45Nu%gwFjoVuk`ce1}`D@^shh<<xtady(P#o5TPWy8&|9lyHqyBsMGlQ&7J
zue>kCO#hYs<<n{8$h)mGBWUjjy;+ZIY^BW0%yG#((pJ?enmCT5Ta!$)pV%g}qc~u`
zczf`#nM0h_tdu)k3J}ffDomYfchQCQjJkPU!;~<nV0#<nU`%n`-X0(%+}&Fj(%A0W
zOhdSIj$Y8Is4j0Vh{+Y(u(f8Pe~5F6JhE+c6GKOCEXPWZ(SAWCA_ZH`{$W_{Y2XiE
zky}i!zxGq#JZP?27pzRU&JqkR<>d$lZ6M_Yj)@KRP>Y!i#3r=|SGGW)(1=h+({N3(
zj^!<H|8=12ZYqI2er=>7uxDmL_7q|FKufKM3DuKpO)f4d%U@{R7JAQo!QJtTm;t81
zb8{EZ=U&LCBrptoRr~rv3}YrN5h>HrzQ*Wapr{G*9(BXpb(Ac}*Af{}&lLN9EBt*C
z^RL`*CnmlE=5*hvZp3$*=Y>18U}5yi`yjiz^51ze9ys;XC=NyYY2||L$h8be6uiZl
zDxw{|PVK@47s-maj|H|no$^^UE2y2ph{HnbL%i7B7Hw^N@u0~|dz&_99ut2CpF}0}
zHct}1V3WAE$EC%H$AdMA4`XIe9UUn_5f+d?yEo3p?AkT}WTKhfJf|69+kRgDO}szV
z?bnYXtKUxSyx2;-kJuE}g`KBSnxlbf!zktDGSb>#Fh<Dfd(^RdV8T`B7Uc%byvGUG
zl8Ho0s98L@HC4%TD-nih_tY~T-{9)_{RIh3O9}c5f*c1&^;D-As+;wvlp<zJ{Tf3d
znrk7;NgMK3E#Fm0nAHCzdD8a<S4b-)^xFy3GpUy0ER`&|>#-jlzgD#WaNMk2I2{@Y
zu_z@z%)SFLw)phz#DjfEhn`X*r9REbEhCUCf(@Mbq*bPOzOb?(@N2ILSchiI1wy$7
z#A~=t&Azo-wutk4mXy=^2Au8So2IiSoiZVA7>FW^0hVnEhQZLw==DK^$nf;^_Yok_
zHdUS)(xg+*HW*afI+r>EHO@jfth6h%{A#;RrebsehO+Ks<Ye25Yq3{yO-;QKyRxgC
zotImeVF=V=^BY|RwUZLzE)-``s~tJjJw&h{4+3SPoXmCRA-Wp2WDc(h0M-Ey+YI+T
zj3mU+>vf=@n|DXrtLSK)uSeGacA^A26pHK#Hw!fj4U;1yOJneI1{r>JN>TyUGtmU8
zP}~77#plM+*+3_el)zJfmB#$Y=Hdr1$KSM+9~l#hvT?sV$FjXIQ-0O2Fc7nFeLo%;
zG`uqzvyO1{4qm6%M{<K4h(YUpL<9ns=0?j(HLq#m8`sD?;4aHu$7)RBw1szE+Aq<?
z5j}t&P0<Gf<lxsCgaC3vRWiU%vEq4Zr5AP3P~FbNH#Jrawj`ph1JNbY0U&LlHr067
z%ns2<Pwe;(y)-3#v`dIRe6}RMkD4qxfU5^x0B%uSA|{|c0I_7#)}ra3HkMyLdnxNz
z%C%G|{gpxq-w8n42N$)!+K5-On+62LWtflg*L4Yp9X%NS?}7WefOCemE`^3z#&lnQ
ziynds(&eFX=!INW>ekSG&KGpPq&VNX|NUTRa>6w_0iV>4ilXQt1OH|)P|*kmGk<g4
z8f#RU4|B+%`834}L5fJC0N?dkt2a+&>>n_1dp)#UO|MeK%5KC8@rNsiXTr;$YHLZ!
z5(3b!>V&G*FLkCcj^PU!I!}zi&s|VCskp4R-xMYX^t~9x2sL|X0X7DAhYdOJtuJ{`
z4KxhT^fv?^RpAW-`O%fwr7L~)u!2$;Fzd>%1DiVXT~(=if$klQSspArXK0o4kO5!|
z_3IC=wmKD}IvO)xehJe#f~ep5aQ0E@w-aN<%hHnL9OmWKJ<X~itY6GRPXd<UfK>E9
zkPRd&9UAWzv)=GMjQ1be4m9cUJ;-oO__q@v_yS&;Q+(Vol)t?fZG&2lPh8l^6Q7V2
zvVU+7Jmp5^s{Y>b?L?SUcT`aE>+FH6#(Vb$51L>41Z$K&G4Y#TS;joh#B=qfJQN%Z
zTyulfPwBW5N?D6A&z20BN^SCLX|Ow1S+D`eq32^@DtncEb#Zbln!!DUJfe+-HZ4#v
zI><ADI-?g?w3)0V<C__y7u*-rdfwoT)*;(Gjed?MGJS(?`QT$L=TDd3s@jbKc%zP4
zDm!d?Bl)cS9ia%egQP-#NOR|b_L0)=#Pgb9puOF+4u^aytCHvX+(|D!b&~Ub?9Ha1
zFBRjEC=#{iTo;C5CMOeHQ)1I&y^L>O*_DZcsf(r_l+O(h)TP3d>2NHXk{N|@n#Z^M
zU7}AyjYsQ^D&ia>6YV1?_6_YegIW?MFB&(62DYnHM@YY|1$~|G4p8!B#?E<+XHyKL
zA3z2+?jovGa^wm3Pgw~t47H*;(w0MR9-)5V)Oj>f^iH9gBT}?N%uN7DY<rb=05aY{
ze7T^sC#C*#jTw_}>k!OhxYL2gcHr^wwE$#76fOtQt6r3tQ<SC4;gKZvb5;7=bsUBx
zC%Nkky1ga?NHW{a4xO0Nt>?miD4#+c<HTGlzaYfK)1@bh2JlteD{^l#T%apCm<f2`
zId#sTa}m~|%7W%AbdXf~lXbLxU=XkH)p(j&79VUZgb4{*7rJnd<=PVyN9=WiN(E;n
zhij`>KMV}LR+vhIF~Wn#!8VhSOW{{O_ZmV_Zn7@ppL7aUW{O-2a-nIe{v_OFFl(-#
zWJ?Hz4v;j>fM5%-yN#*+b%lwdcbs5BQc8HW%kn>H9}i;9<mLLUo`(hp8*ijEW1>e6
zuvdTOj3MpHO1lEBvQ7O)-vr7g#$9+2^SXU^GV#zU%^H&$$&G3(lADdPxjzp^BMZ0!
z?^}E+Imm6E?LJ9&w+M-o=NWd$JgxA9OFd)Z&f3*E*_qCi2D4Hfhs{7{t@#DqP3hA6
zvW91S>0XEFjA13yo#k&Q98v+JxMAXJ<BabUQnJXuj88#tvUkT+3t=MhJuXVG=0B9W
zu|cL#r5}68_e&0L0_BY5rZB19h=r$(YKWRzR7M`2MQ<@HT$>46T%PAlV0*9X=W^Q}
z7nfa9s!}-f$@tEFW6k5I3u418WQSNX)?++yWRwJLLx%hg4jg9$Q-}DGEKJs=J~RA&
zu@ZV7ShLE4aT>$n1$y~BVx`Yn<IAUw<3`{2Ff%d*<vq9Tv&Zi9i@I#3+(L#mOtj7%
zb5_-WqZi0m2?Y`kGL8GzA7Q*LLbr71$Kw8`$8izCo)et}HEwJpXCIJ&r2*95VR+jr
zrAxXo^u}s(&9>>m5G{B-6wuwX)!KA)6TbSTV1oeW-nfyF!Q<8ihCnmQmz=wr=)sDL
z#1J316iT11F=K5LOYvWjuSZP<F3BWw%x{&5A^v>ZUiU|=$E6KDYt|IshCn#K+MMIX
zQ#D{cnB9n)cxrIrrR1%b+2(Gi7sOxXWfY$ZvbF9;e0km!gs4$bvF#cnRRmib=K6Rr
z8mc){p&`O)a4^JXajr#NIc0%C({}Ife$dc7nd){YlDn&8E4T65DauTI!wn~YiXet9
zj04P(w8P+dIpj4$QvN2ey{3jO^I1p3-_Y^6E^y=4jZgcN<5LUIlkWH_wQl!rg_K?}
z$1C&}IR=Lo^5!@@F6tR~)3WdU7^C)A^WN5Woa?8<E}o(9QCFPJH~@K#(R4;<*#0#q
zdxoIwr%Z_sNX-<l<VPJC5Mwv5sMpZQIyJ8=z}0YoXW5y;RDJs27-rNhAuu(RJ0DPd
z6Bb(>2j~nmcH0eaYq!Ca&*(iW$aOg>(nwX1@lC1~t=i5l?k%YIgZQbxo5<56hy23V
z{0XS%%45mP`MQ5zNjG_JqK^ayt;O@_)n%sZUM;~}-J%#slaEb!-;uIPuRzJx9R^AO
z5wP7;^e`-?-5uoUVA@R$?IATibeO>mfod&j!7C7Si0}r&AuG`d*y#XU60CPr3b@Gx
z;!(%-g*E|22xE5~?DFf2OkR(-2llI7km}!-h*67vMehAPq;XT}p2u7r8}MnlaWMw?
zE`4fe`MUYDyBjbYvVHl`<(P|tW#YVSIgHEDjeaqqlZJxdWq!Gg(L(uw%B*@(8kZgD
zI)o;-ej;Ovl%T~A(#MY4*wFM;R#Fs|+SrsX^J&+3l(l=UVw!v5;0=VGT5>-zOr_zG
zTs}StUwoY(vu8jkrgw+VHLJ|XUSd~-z|{sn)rfQso#CDDlN=1T*X<b<e!k+i8;W;e
zg)NW93RgtCW62I5OiSVAA4?QI8!`JzOfN_Rujr_Wgzo31wuCma-x1tt5!^;Z)I<%k
zn&%uHj>q!xsxBC(cC&(glC~zG8$_SW12;ajWEwFwRL<TZVW3}){RcZG^>e&wV%T-v
zGAy1zz!S*}yD{mk!QI)@KeHRxp5*=1F|qb!=P$qAT>d4KC)YuFH`x)HzYZ6Grk@++
z+=Al^`+noSD2<KRQAqmW#rkxCh7)+?TdG-vauOsLR?(7iZf8&41f$@PmU1Re_FiWu
zDpgm71luGMgT0|Vtem0T^<ux`xv0%K!}I+O&mRd$47>gII?s7K6=fg-m?}--@vsGz
zoOA(dgCxY};il(=Yic$#*Kd2@+5CJv+-9bycp(0Y_fnG5>%7`u%q)r$b;D<4&@$S7
z>}PWpq#qZgzU<wOjgq;EREeouWrPH>w~Q{|>choCYTRg#?!le-vBaA1iv;BdID*N&
zq^w7?`XjQcxYmsko(^>1?rQk@4<=>4<~HioGBg}(uDTSjB32zNuN8K^fyMwsA0A0<
zLh|e{9nR%`YRwn;Su{M5keYjxb@?DeTn?R%Jl)qW-ZDPFLDvpgdHK=a=amL7{v_$Q
z741KZ+vr185W}|bcJ6J3jDch$d}Dl5U-`Jsi0>4}dqvg+&t`)400h_P!?I;PL*gBo
zB6jE;9;~<oqG@Um4Ar#T+&Ue+hWw<)Ytg<&xQCN#3n`svWQo|RPGdLS>0qD;^)TLe
zdpI&KPwYd?Gekq&)Uof&)!bWA;8rzir08ZD(#K0HjyxvPd@=5Y@xZ9ExAgGf!Q`I-
z*1=)7GSTU{zE1Nwk@(9^(9W_LbVljI@&@9j254--j8Vy`VWd9KGFSW38OFwF!dfX%
zO<QfzNRYtzri(LUG<SP;{|nksbQr(>`)rH1Jja>g4Hj3Y>0Mh*#uF<!*w|v0J?tiX
zQdF&MEl;QSff~GXDE&~iXGpzjM<kV?_ITwM(;4(=cKJ^6puF|Q;K0p2aQWmo-#&ct
z<Im)_(>q36m@%UexKT-$yDcfR!ryh0d*z_12sx3JP;_TipPXn*04%UExg97pEIvrE
z83KU%*2qW9d8esqm)0q%rQsPZt*rz!OtVh8zclE8nH!6*B<5p@6!7t=jJthPUUOCl
zv||%wr_Q%G`+Qcbr&1fP`u(Fa*X1oE`<vX#(6$tZJ|Y@0=~25)QO0wz6{gIWMpaKN
zm9^5=G^E>3Zk?vlqXU>q-KkP`CL}y!*p{Abj;f;J=SQ+IRonx!WKyPp!<c;;{;*w2
z0BY+nv5Wz{*P?0Ep?&l!lWy0J_s>~`Bxhj4?MVi@N>VFhYR<g_c)4VJRRPcgOe`QB
zRTWe%>^fhFbcaSE8c_CrcljybPUsppef7er-}*T9<a0rWp}Mb2SK?Sh^6zxZyh+C)
zUYV33LcYocomC}y=bh|r&nxY}H2YG&e4Le;>U`xo(a<PGVC)IiXlU?X6nS5eSIqyL
z#84|+iZW>7gp_4?x;(D)4*4Ny24e(5$-7cY%zd3F>gP4z;5$Rc=kQCHY*;CrbRq)6
zu$L~&h=Vo1YXhEVk7WO20;<sOD|XpCJksn+{c81s9*})Uq12wITwxQrJjJsiC!!t}
zTQC%olJ3t*ewCW-I2n>wdm;L$Mz?pNo4&Jd$8T3MDY8DV8ELKkNGxEml#PdxT4X02
zE?Jn4UKw0wUdC9nXTa@#7WktWh=xa<&7uoN9O#=uRySp$@m>a>iys9_Ws*#^4zJD>
z*DfM|#piwbvb^8uDj9`K`FO!6yU5PM%=6F5Xtg#svFVqcM%4?^R`;IhDn!;XsJ7_h
z$WKo{81-#l(G|;_x9fgClh-`b66e_Wj)V_(GmJt&bWrvxmc#tkoCSA0%N!=X+M?~w
zXQsf@<%nXri(4Un>rhXnI}9TC(ux<|i_$ptj>yM*EB3Tu1Q>?wRfU0ED`-up56AW)
z#+f6lI*-B&j9&0^(LSe~O(GcdMzbt|cV+B{*ZmusE=O#8MYoJQkmAO0@)K9aq#0KM
z_tL{sj1p7x@df9T-=hs#q3D{n%SI(ZlQ=hqS^%k$E8)n%2S`ZGYE%xDrbf&rzZ<U$
ztWCMTN&zs7t4d)Zw(Ye6hZu{lNgXng@wiAFCqvj>W=y9&#7wmqt`85>dzj*K2I%Et
zB_{Ju<RogMM|X?h(wrJ7dR#8!_lGOMAt7e>q}=B`GP^d>=V@)T*FZC@fkn?>=J{KS
z6zP7E{l{?7%8Qb+g)x;LRBDT{>5PQY?F}YdAH2fR0|?Rl&T3W1#kxpD+J1Gd3kJhF
zp{OWB7@35<PAKMyxNl|ouCKf4nT>}8>Z}Cpd5xI5yd|zz5)rg80fHIys^k!hzgmb4
z+rKo`3HOZXa^GqBf$>%Lm*4AzO#6QEmDuty5G!67p(s-OXJZ(h*Mr~X&@>qGtrgcB
z)&ym4F8r)(Livd0vzVvDX{qKPXOPWKsRY{7EIZg$gQ>^}Pn@iBfxn>eR_z*#gH_(s
z63f8P!wWpl*FDx#em@>6*cpSPl3T(M?&$%7F<INpJ(fPab?pN~#)xuqOvq|<hoI5P
zI62tJ)Ec<q{wG(53$B2awAQqdtv!u`3L30If9(Y+9p=IX3XXdJSOrT@ITkSIKtKL!
z^zP`);X(-KP^<lH4B|*W+@G<uF{D2Y4nWTSfU-VSofgnd4DuvgmK$d`0KmGuf=V<x
z-Feb+tHC=!DmNhq9S2Hl*V9%yt*FM~vsxr2@(6wHmmuPDDp++l{B=V5r)<_~r%%#%
z-#utI?bDKwNo-)*U!h)m_&M+V4;h6dTVtcaj?E{FIR<A!9*lUCOZCg#qYr{Asdu%d
zKxEyGibg%%`57ct*PpT5K_RTOozmj!9?77rO~<u#Sll4bB&>!DPDN_Xm2#=uz4SU!
zt~kc`$~mS~?s>7YRDhjsf}%;V@ua<3xVJJ@e2p3r4ivSeWb;A8JENMTKGwd*kM?8l
ze=)I0cJ~Y%EuL!PcNhdd*coxJERGp-zFUJ~)jRm#&6Fj3QN~SPDzr_SbqmsOL+p|d
zw2rPt9$mgW*|=`x1IF=?uv4{9^;+UnA%R`vbqJWZEfbWe9lSW*X*wy`)H_m*N3YbG
zK+-X1FDNMq^N*Y~yygU`<vJaX&!>1TIT6hO$kCZ}vS0FUpGe<NFYqIpDk)@UOsvq&
zFxR(<-r~giCv-N|(CD5s&(+`*g2XT^#=v0y5FPYxxYTJ(a-99}(A)kLZ~3#z$BL#P
z2=ro{)Ou8K55U4642%F@63APw2qvF(cPyU(saz>tC#yRQhXMg~QP6OlQgXWY)C+vA
z&aYj9S%i$0-X-oq3-=>w$aRf;I*?{~Cp4c2HL|m9nIIQq5~rSRJM1bT&teWVj;`-~
zHEvr&-Sh@!JWtKcL;BC^st{EdT!{Q>Z2tRYa=U;c(btRRXS$l(Vm}@8nNgLUgh)2Y
zjBG$xN#YVKvBiE9QRy&)lBOHeb`|&a6EriIS|o-1iu9X=IxGTM;UBH?RwA!H#2B;>
zNeWwrDeUP$kS~7r!4<7GXURKAd;p4D4o~uHly1}K*)9hYxX0yx(UGl9f#CpEf^!Tl
zYDGkBoyyytCoe0i(W&nC=Q?w;;1@0{sqZkvLW4X8C^~&$hrk{(kyQ}z?Zk6O2b(P!
z5U{;6_ifDHgTwJO-dTEE*dZAB^9l)O@chG<CG6NM$d~PG)}1oK#+MP7s@C5q<Zx-h
zP<T(iZo;@!=T(1*-RhClY!>DYB3Ny@I^a;bNFKiG+C0w)>}5A=X<eg7xSuPsHYAA2
zpGUVNHA9J1f`g;&;Cz9cmv8bX;|Eg(xlv+8s7H>4g<so(RAXxGh$)K4V6Lib3Xh*f
zw_%ZNzz`AomC?HE1r;>~;pK+#5^gi}&u(Lg?VGhM!vW*fOyRM~5o4!PD7&_(smAI~
z!^+=MnB2>OR*Rh<eE()JV?}w5O{k+b-Htxy&H`&P`o&*kx+5ra=zecct>eV@T0pDL
zq7M~7_v&#m5G!5hU@Kcp8Gr9H2Mw`hsPX{P0nuktq99ntF3N@}pjxRC&G61Bfs_YL
zY%NV82P_-nl9r?ljruh17p&l&761Hw?;zhLq0}dGzt=FjAYQS!Y}8uGuyr<O1_*0@
z)x#0z)?cxDIUR)->e@QF6NS7R6Hwoxd(7IuY&bhNgS2yMB1|Tdt9_T~PZRB-oW5S1
zaB^qZ^<!!)9!;1wnE=qf`cE4!S3NEN{a>XVVTkz!GqwpOYI%KnzL=5~=9Ln=O-NoC
zrOHWhT91^Ny)Tz%=#{bG+n;-@Jk!#t8b_*Cr)2Vej7-|@z#5kW0UMm+K1=LGmLj>&
zVCOe)M`N8&n_)3^9nC9=aptqq>;ZB({Owo`SJLsXkBrO^fikGFGY2Y4hh>0JQC<SA
zR$xj%(ni1QM$_|_h(b2}9;WeQz-|>KQ-Hjgb!L!jP10gtlJQ6)h?zfcP~v?jir-u2
zO~Fj4>fMPD7>fTkq(8zY!;(G+iGEGUz}-oA_tie^n77SJ>(Ve>U5Px^D6l#kx3Irz
zoTY}@a{F*$OyphB<JN~(v#mx}0vU%+z$Cz}AZodxq{!9~TZ>z5EiOBs<`(IJ9vQC-
zs7>LlHUU`Z{8FzV+s^r8Dr=@!Cb436JaBXbL-WGyi%vIbXQ1Mg7a?Kr;u3DPdqJEH
z)g81Rj9uLGE<FT|oy&0JgMNwml;#RPXMMZJIpd{A<h1m90uah-y6D@ylO4O@Cr)<E
z$<&msiFY5H&g6AIlMxm~jIZczWPWV!Zo<v7X4pmdj)mGmx<y832uO2u4grQ_4X(4W
ztX>&uPK$^spOxaDj162u2HW;v*P-(AZVY_8-FdZ`9Vou1sxz9Pc+?gL)t&bWJZ@ex
z9tD=6cLDN2SME1MDMgUSWrmiIip6}r?IYtPKCfxAV3<~axopb$`U!Aqb@R6q_k72-
z{Fgr;S16im*QU)r^)QYRs{7n_dv4#&wU^aBuCuPirz(X`k2jd);5{)Cb#`hJ6L>s}
zmBCZt9VKmc*QnLachJ}~HSL!zC?y^Urd};gQUdJU!d(MRhGVjz4kCToq3i}X!ff%=
zpcNZQ2Gy%ysaWCI0hCnsAwnz+^9Z=%H9ed-R-LSgv--6Gq#>|sS#9j7I%OW5nJfLt
z;_W)0^F!N4(9QE}oantDYQ&8^@F>sU%B;JOJaHcH`vwRc0Fz>0lZu`&RNL8hw*?JH
zjOyuf*hupc`gH~$|M^-ML3!ee9U!@%#H*!6k>wfPbRbZdfEEygi<`d#%eQ={cliJr
zp0$pEqG_pO6Mw27T{+O{o=j$Zc?XU|E!2}!C3DX=?tD989EK<2h40xJ$dXI03VT;(
zT#H#>y}uArhNaP4T9PF&aM?QyJsHJ`?pkw}Sd)7dqalLo38@p;T<vSpaV`X2FbwQu
zXxa`U+&E~Gi1ewkEhBh{t@&<FDRiV(L-x$_>oBbcyuOy5ILKBVA)g4xE7rQ?z*Ay2
zMI5&eRof{whloj^DG%*pxW}#F2m_eIsyXh1EhskV+X>)T?i<pI`jzX|M4dZ{Zc4{H
zWjc{;H+NF2xvIu@M>571JLxRLk;zXJy&0MKRaXt(+CgNI6H2ki@DaT(;+UXEXo^yY
z9@zQj^BrvLO&Ojy>o@d*w`sagV!8<=4_k>Zv7UDt^ryBn-Dmj_jZm;A7^t)3TDwcy
zUSIb<dbM?H=EYw9l_5!*%GWC07&SoT`bybD|1Al2R{&sUANRfK^roWs(TA~viQg<i
zh|ex1SI-jqI!zx*059*<bXNb(U8;ZIK-|%=8ROC-hmcK!^>6Ab2RnJ|G_T?5QZ<MT
z-F=u3rW#wE@AThCHMew_!*rjAYf)Nt5p(@%g;sT()M%nD*-$Z%fJc_brn4*(;`hWp
z?We70c>nq4`oYIPeNOe3TlLR-7e8Jd(g-K^%v2!hqYAeR2i{uTJ~$m#Gm?21;$gWM
zR5@Dt=`<~|=cc6%ikA8@S2r;;&fjG$3{W!Z+S4vqpvD7pbw4F!=&8=~`CumjMes6=
zcQ+F;=674rDSOM0>sH^Ah?99{hB>j9yyL~Cs?h~?UQ`HjzMyJ#UW3Wm<`&0~4*5N_
z*nQVw9{JPJg?=qzb2KFtkcHHLtqL@6duJfbHD0^kP`v$?<gFV&h1iM9KFq_7l0}BH
z`ZmqxMB}B12saNUZ+~Zgh%$gspux3;&DM(2Ux6z?mGO#*>4u$HY-FJZScJigB>~$#
zkx>ap%)@!WnMSG4Jd{?bw+hF($ve~_GgB97h$qw@H&GZ%^uwE5#U(LU>~?YWXR8$x
z-D*ESEh%9znIDip_5sB$HmqD8e57u_TLn$qh?6<Pmxc7rL*%iIXyJ45$KN5Z?zeNU
z;x>SwE@NnCW^z-*gBt*siKZc>ZNy@rla^L!X^-esyWT1$W)jK82CU{|^^5O!QNEfo
z<K%WlQBljj;@ayT?HPZ|U*!DjpqJcv1%WuMV3z2PUN+dhy*nc#*o`l>S6eGnbwIz6
zY}KVR8&X+F-70f=y-^FEylYwOyiLL9C9@4n!AJfwps;C6geq){tz?*_7b_g-!^tx2
zEe%=ozVVh*S>dkqDq52(01o(lW|;C645@W%f~mRlci~Abn%X*{5a51Bi~lC{LY$W`
zY&CLm())s9p&LCb!wKjRMR1%=FM#36Rjc@$@M43tp179z_C~dAuF$Chxb3l5thz<w
zoe@I&4_+xlaN~ud=(vCMMLOSk<J+vXmtf>ge^FjQlO@dpB0O)2uK#868i(@8#JD-m
znq2lmw4ucBvcz{Me32uaaH|0n=&~IhVBN^<6L5n`FHJy@0d@y`5j;S$xWX*J5!fAy
z>+w02&UC;KsnOTxBdsJ*^9{GZY;{j#t`7@q706)PVSrW6!k}F2rs`6Y^xCS-=be$@
z>ebm>zdIM6Yo0v&u;H4JSMzK<XcFWy%;otCqDj9-tF6gkT~uE2u!jM1y7tAoTYYb9
z4p;eYp}QJM%@G8tU8m>ITWcK3xQ>DO4Op%u@`V@>>?4R8fF4b9&<@bz;sES50-Bd@
zM6IIepq8)@f`5~Mzl5l7h=yl*g>E1Zb=TO$f5k6IEwwEOa2W<fJz&d92<5Mz1$)w{
z(9RKiP1GIx%+{C}Qp@)CE0J7TcVXLdYq0{yVuwQBltKjBB;)RA`5pCrwN>CkO9n=6
zGD$}E<9gg-)Gn@)qXQub6p1i*SU`>|PFYE;4c=heiW!f9O!mYdTAU4!h#@pKzaThe
zRI^9RbTMLKFwIW_>ca3$!LkneiTNuje)$Uxa9tWD{u28|CXMHj?IV2lOuygk1SA(C
zT^P7Y4&?<nPDs;0if*+x=AyddNUVFqC4R5CvBF6&E^MW~mtOb__I13?U@o_jXOw5v
zOV`#~=fsY4ogMs~-ZfOt+qRL_n5ASvK%F&dZyEv7wn#958o|Q?Fb$S{6FVujkjI%}
zN0dk(SByBaZi4|^(OnMAPbVuSue6!kzr#{e7cA<9h&)B_*RHQ0RazANKAYI~)!6gY
zBRN4Rcpf8KHmqXwWFC+eWTRp5`R};UFO=f@caw<l`{r?{KWe<XSuT$9koA{Q3`&#6
z#lbiJ4h*RW#ubc|6hXL%kbBH89i8Xf+;UtN9LDN7E!1$HY7eEov^A+=Pdt^W2j@u^
zh!LV4a5qn{9(Q0Q0JBXZS*hMyleYatIbzVFTS05mJTYKbuk2Q|N;xKdOjX5AXXO{U
z_YG6pLH^eA*T(hYso&4O&ph6&%2MllZNP4NiNsDcT@T{hWhUTF0e|yvp*@XfZrh<T
z;cbYIt}KeDlB@{01?NjIdTzT`cm--c?mUFWM9gAC8&s)avrEGefWUM6bk5c*UW`%3
z-S-nagG9#Dl3+?J-=jsBNY&kRuY*8I=mI|?<6(5Mck*u8ifhdL+Ra;$5Ca<PvtV(Q
zaf8mBme{x*VfaXHxnxwLKcMg7W~0}DkR6H^DJ{;teXnfU)ZFJ`Gp?(Xw!o2|e=)%H
z_yrLpR1zM%(XKU8Q|>;()_w4q+tmpk^e;?vt%KuxMi2~!K=;)bd^Gj#gv%DjXVs7x
zl~L(2UPYzXGEiA!jRzad)a-7X9&x!vvxvR)t&4nx3`_oyZ6wC#jhahW*pEHQToWY=
z^uqJ79V<eJLA~;B5bx{fYUDlkV{KBo9{b{zd%0aNu;Z|R^>N^KLsRVqz4VnSW_lVN
zc}$y(RHIZoBxXuFHwL=efrpI;Lq7*#1-UdDu0*xjlmJ0sk|>{_+<)MNOReVQOEq7b
z)JLXYO=AU6EC?iBEj4-=YtnR>Drzf`aM%p*41Y>JGFuq!I<#&&ZA3RNK6E@1=FIwL
z{YAP^gbY<4Y#1H3+KBw2rD#HEx`Ol}uA{>u%m#7YH2=tL!1Aiko$%ef)tMdiECmv@
zDJ^0@9l1(r2w7!f?$)YAP4y1eT2D5vAwc#7C7DoVj5xP@ut_h73bv(WfdxtuLAW^s
zbtrB<uDvr%IVkrHm?k+JXqp#6Bh1aO9_H^=m3F#FNAwIungK$k9l8(iwI>F?5GtJ4
zZ-!g^Iq_vE-#a_j_GYx*^YbA^u|N2@1q$+M7PH)4paW{ftLaTr`O(<1G^(PvMdXpV
z5du!0V2#f+409h7xUTB2ZBsLm!+rrSd2hnzsek9wl>R^YG+U61j4tJmOD_L+K8?0!
z-REj*#12EZUJZntg8~l%SBVA!4Mgm-TkJ2dh|;Y`)OMlgyES|9S1X>I&aTiUGx|0L
zr-Qq)#!WD{t%v!;wfIN{(|JP`^{&f2>~WgC6%=CkRv>y^&(J65iYdMpB-FeBivL?e
z4UBA(2o$CM8|q;UhZ~1=N4G^!N$k}+lOBOnbAwq;Ve+;_FO?QY8bn0gs1FPXH5u;*
z7JAu$DZ$Hk>{)4V!v^9n>nq2RFzI1A1Erb6EN;>lT?+wgFuAz0E)h*!rad+l6xIB{
z*n87>w(~yjx7YNV&P>;7mDbuRMeS8v?bE5HEul&hTNqnNNYs)bwyV?8+7-3LI;Dby
zs3nOcmT7IZgc3;*VeBMAm7tWlGuJuiKKJW9JrB<7yk2)6<v||(^7}8p_xH14UFO6#
zn#&meMo(c%!=$6<-zS23yh1z`V2E4a9=#0c7tZ3;_@DQ-M*M>j^mKN&kuCo6fqC`h
zv7KG1KVr}fc|QJyn=fha^B-N=J}TsInM6B0{KFsJ*@|X%(vKwXOx^i--leC)WbeD4
zk)J#7k9<q^s+>P1<kGZ3oKb$G$uM$e`YMJbRPS}LuCp`7M(bYvg7O&#92yd5dr8?g
zmUU@VJ!m%&QzBNzc}{wiJ2zLZO;SPIpD`U7!FuO6%#9^St-<Of<fZ{v)P5NV)CU74
zJZ-Hz@SRZqg;5Iw!Sk%Splf$zg8Hi~;(|HnuD$<g@#J5t<ual!+fOYQN={%@d)uFk
ze$_g;azY_xtHD!0@yyR({`~HL|M-8e#{b?M|G&IaPEK@{pL$0!|0d+Y(CaRkL)Xip
z{INA<b;I!m4fhA}lCN!^-F8xbpr`#%+h^%N|F^04|LJ3v5z>yvgYpL=Pd+x%Z4IGW
zNLOLJZI*hLQgc$I&^f%Yi)vwUzKxEzmS)^uaXZgv$yWJv8~X(t>Zzkm<s<X&EuZNQ
zWH^Og<!vBxoP~Xs^hxv#Q~UjSwbV6zwgw5aT*%YEOB+EYJ!ido-(TkUhA-He>AKLz
zXI{T`Iql1%i!X=1|E-o0T><ZLDRr$FjM}P=%=>rvG8_WfooiOdx6mqA)ExrOLtfpt
zn*XvL!&OOvuZV2o{zUYS#C^1sGS6N@AwMEtJVM3(Se0M1Cu2mV#|r)4uPf@WT~2i<
z^MJMq&fN<&gy#J^iB!|pTWrw^_L&TO^Bq&?*~rz>lk;1zCAJxszi*TZ97?{7fQdBe
zumFYL+3#CP>3qTFmOy1xI>x4Cv7xKU#vrHU^jiEV!~(}Bk_V*8`*#0L4Enxwp}f{T
zrE*%C@~*ZBP`WG=P}BN6*Sfh!Z|LL=YS*?LyvL2KcF}n?+%q8{o?+D^9_i`!H(=!#
zeD;bAijnmya&j>x3W~AEoXp$wdt>d$nGK=AGyCx9!)Kre)tJ)O4ENzrY1?nOr&|2=
z`%tNiUl3LD*zwtM7LQ#rp(CdOUp2f)qlUSn{v}fTzy9ohoiao0&j!tYucd+uh#W<E
zM^4p)VM^9Jp4E8eh+DtX9{KZas)v*3T%FdIAiw2Zx*YVtdKrbf<Yk#<ddlo<u}aUg
z<$7wFp{M)aT<L%&0VTup>a~W>ueNv~DaIE0f4zL!5qK%5`ZUw+WnB~K53)^ic=(q|
z^e;`y=C|iPlkOZmJK@OgP%D1WS~a*KvP+<cwZ45dJ*6?7ro%bXz|2@Zm(<coQQ@N$
zwVe#FYpuAgXkkBQjKxzHwBFz1Z$I5EYG!{&^8VLUwL}}P!0=tf;KOOlQ@r&3ygOv4
z>M#FZxTF7wR6SgX+TXQHni_IBO%Z+jq;|9MfsXc}VPn7g4Oi~M+>Wu5lA^d}UR><p
zb@C^1G}!<A{A<KfCQD!W4>P3z&gz0md~R^Ux}My5mH6bfmW0)w$iGi0^dReXwhrtm
zk5J#(8!JygM(PsAt$IEk98UihGj7t;yq~Ti!=*ekd~$l$hQ1{|w+E)h_YC!38@KOs
zN2OL7OhC<s^-}$!*{0bf?<_xkgYK__D;oUk*-hhu!`Q(u_{THNtwmcoV-o*o9m&as
zuI{K*qf0{<l96BE#wjHmir#+JIXi=1ZM464sn_UUeWWW}CvAZ?TQdI0uT>-u4VEGo
zW#-`J9u5@Y_ecNY{G6e1!$u}^5b?%>Y#Kx%b#gOjH$y|Jr+K{9{5+3oPxrxu_rq-V
z;J;sEeF6eZoHGZCEg~spdAB_aJQM6+c*d+q&5u`}of5Cp$-)GKV;I;b+B4vb)MR8e
zK+Av`owswD+*})esxB;EjwuyRbLF)1o(I3YQ8@$6U2<_{Lu|X>x)?}#s+~zel=+eo
zd+NX0M{6Mqz<c#w8o`H^90G>W?#4P0K58FSbsFU?MAiVIWIo8i&coTxeM;#)v8oTI
zuJ#&5$n#M#JGI1hb#w2${r>*LJOA^a`~UjO|B|uw&6()mkX-LbJ&)Qa!dr}~KHasV
zFzZCclZ72hkyTTS&VfJrb9qrMpokwGZ0dfWeU$P;X676A^BfY5^t}I*kOO$cDK!`l
zmv{Lvn!_DiQMQAqU2f4&QlJl?uH_AfJOTmxq7QM-8>o@n3$q}EsPx}eH!hw3f42fB
zvBRoyGTV%m&}*HNQmwwOW@is&dcr=m<nNa<+Grfz(@^!<oUg6uZ0UF^F}s6_Xi3t~
zU9EozhGFVgD-W#5ot5R!FpfzF)z2}z%R@|3MyzuKh$#}9I`-0S8f(B<B<s%PRih0i
zFcqGLVi~Du6fP?&wFowoz0SgAIT8NB`v=Pk=Lq`~`AQ1A$^sVi4%>H}WiW~<f@&1<
zlT%|Ocz0x^TX=?}5a}_-^#F+`o1Kw=^*o7Ol&E%0xh?p}(tlXlk}Gw&x55Sk>oGp#
zt9Re0^2x{Fa~|Itwv^cGbxbK!3N62=cdHIY_28le^ory149H;NJ{Z+*t7K8<VWd4Y
zO-z$WWHBKT2EfbP79o>=thofaYv)$F(;<f#P<hHzYeDu!$W&N}6I2^`x!YO7+t#Jq
z_TtF3%a;5CY%f>47LKgoha5$6PQ#`$N|z=Vr!c*rz#8y?r7zz;zpK*J-oHOxJAC&y
z(QotJo|BG)F6d&<7rbcpeEZpn*($r9(v&82LLc+%oll!aC!s<_C+SwRo;K3EYH1$9
zQ<b>i+dm+Z>7ieH&>46>jdBE_quUWc3{9;v)$bg9ChIU3%rejd?NaL9+iUsKMM+$A
z?K=^A1G)(5C#J<mvw-|n{36p9=xnYlH#DNB2Y!YawEyI(dj7pH7|@jV!gu0WOQsi9
zpQAF8l{Ma73k_+p&L47l_-fj!^SUk;Y61!K9|Vve+>NQH8LVe8sZmlZgUx}m8esxj
zM|*Bo7RON2AAa!M<~4?qLIp;tFbG&lr(iZ17-6l?00O~!#$6{brJt44Jc9y;&q%3Q
zDANQBM4E2hD71*MmRya;2c-~_la~XNlYhVJuK-O8p8Qs{v_19ri3VK=ozm2b);qg)
zP?)~-ii4|v?y{>*aT$CV&Hbg+eOmuU2K+G<dyw?zIDh1-eujqqSjc|BTs_yst30XN
zor+#w6%`<S{5*OqmWDInY65(*6EOaAkGZp%mF<szQY|v+p{k0<*6tfrfZG+~EWXI^
zB(8|a0Y+RL)*QKLAeQ@&NDRbO1vwNCROdS2qdra-aR|WchUDJQk9L68MmY~l5XC7E
zX)xuVM#mJD9#p@L@$fUG%OsWBq^jYUO1LpC+%k+hk+t7>*-?VUG_bYrb~KA|)cOx`
z91JDs>Gh!TBm?wHYE6+!{PdRDm#~5Fb-neR$!Ebgdp{>191^#^dpB!0Ctj$P{7bj0
z-O)$kUF+E8-doHyA5ov3qgsLNe{*hM$G!5x2|BCFdxVMi*pZr9XTKUXXVt^pGjkI}
z#l_-rfSBNUb<x8)Ud=#-Cf*f1eimZj%(6#0>ki#IuUgr`?H=eItc0)y4ObM0t@VLd
z<qd$oLL)UCQMUpEFY;AQNtutEcYazE3unq(F~6=jF8^E#6t|c(g_r=V;!D$V1%CeH
z&9&a;KTvbtSb4T00G`PoUUiMW+742k3M1x7E?bOD$C#R;5%IjFEI$pMV5C^ZI`6xL
z`Ow?n|52ogaOkzLn!=|w`U<DvS~DWlXt);3T+$gVA<cdKzVb1%c}GCvAZK1D<t2p+
zJ4{w9)q4gdIkql9M~MD(Q7u0u)7LoVcipvXSO(WHn@=v}x1(DHsHNwHe5ornw_sWx
z@1_W?*rY`iRnDER8oc$PnfvtDBe<e{_}LofMfnv;>0PfIcgk(Q{(Pj?p&6a@c8~l{
zOhiBJedjPMYZs0<yFRVHr^<ZtX-_)~TNx1GYklM;Od05FEQW9|Vt8CHa;kZN<S5I2
zPV3sA^6i!I1FuFu4>oaTj3g)jkU?mC#wYUL^^@|fubAl8Hn2F58eB(F1rb|@O__hZ
zO*FL`YY!hix<7gd5^fFf9&~Fr{y-Tg-=3k=W}q%U;gx%`2+ZJ;0(ouO^oo!N7a9Im
z0z1ds3UT{Bou~_E;_gj$D6YA>Vlu0n_Rbf#zI|fwL+4fF7!#<qv{YTdu)(15Mc*YC
zYa??2H_8Nx2~OD<D%#Bs^d2RdUr`itr3~y%_ralVt0MMlr`0GBeXxd4Ar_YT;jx*q
zLl4!-+sko=TuKI<$$;b72<SlF&c=Sj>jlnQmn)7G`Fh~RG;^l<kEc+sytK{AL%Cl0
zwb#nb)mp21;a1cRHv09#h3E9nF7x0n;j=1OY@c1JhS8(34W!z*F2cdOq`@Oi8bkUD
zqjItUdV-MYYFzD6S#)7HJ>t*y?-0R7NP1cw|BJK?)OJkE?L+mM@o7Q@J>JON%Up72
z%X4I%B{5_hS%{4AcYi=-oEd4+79D=IAXuh2Uh;a7F%k~o`NEnWhGdi8`EG#>9JKb<
zijueNem{7q|Kio;^+oc$e#1LUNm1g-uTE2C>Ot;Macd#hPRZ#*7ZVjTkYTD)Z)-Ci
z>Cv({EWaLCgX3j96mk>rA_)P4%g0vYu>-Qz$OUBk$w%BTQJV{XaDCT`rLT!#BVkWE
zLE43(KrUKc4sHbKvFwkBhWZke6=EO%r9lDI2*|R;{o{K6E&}mfFIP`iW+48Dp31kH
z#h2+l*IfEm)iksoJz3z<4Rs;_L3t{sYH=mM@?f#3g^@MsEseg-Gf`KPJYM*$BP8xm
z7wFqj-SacNI!!Vy`G6<@r_MTgm~}g>wLKf>-W<zz!^Z;#OXZDB;OSv4ahE^tCUib)
zy+w)#Ybc#6!6O!yY~<V;)?A}L2?Qe0P3ob(Hy$62w}cn26fVm$>BHBi%;`J{>ZoSw
z$jx`~4P=<0I$YBMt~}irrO*WLFI;cVR$;%Ld__*do?0|$%^$A}Ks4k$Z@^eYFEn6G
z_p!MBQu0o`5`-1WFa<Lau&L=~oAPjWW0IGopg{GclB_D*oe~~QVq(+;$*J=->z$N*
zR`*Pir0_daRaFUvqp!gtuO5*zM!lDQ1(qyo@uc)P%ZX>LTO0YMA)KefGh|p@KwUJu
z22M=$cdcL6SUKfY2UoU@m@)MdG-KA06%~WDN_otsf!GG>xC52l(tJY(tvY7J{QE?=
z<pZ7b)m<4QO=##-oCCc|Fzfg+n4@aCCIgX<bZ_o`T`O)>EY2?QoSx^@Ef-<nnJOxS
zlU^5sxJvFvQBU#wn1bV!#0%p~y_)W;ABs^xu*gu0!?NclSMRoNn)SURzlb$6Z1IkD
z2LRqoTJQQ)=j}`iZ7{55X>I*Mw@dt+O>h-~La+*SzU_~WVf(KIB}jGU+)oP<p;c8D
zFK#7P+P?8x+VPv6%1V~?xchPMd}S>Hau3FPH!T^+ObV57V(CLb%K(Y$0AKF|z?Dcl
zsnt{>qXCRr7vua<eV)fkz^|L_ioIDi0ZVhPjWurLF!~EA==yYXR$dl>Dq?Spj_NOg
zUA7#-K>aPV2#NoYE#F1`j5e*A1XcPTq}9?g#DYif_x>)+nU8*HNRm99vST);F6)2+
zx)G0sN=`X17ELct<b2M^MQLeCdyhph3lcC+A;L3IWP51~dine_Dgqf7*WM0l1_9Rz
z?q1bZHs-72q_cEO+hX@pZtke~?i4_e178+dak&!jGn9PbZkV$2TDUPK;uic*;!KF2
zH4VYa_ERF3J3)Dnc#v9k8^hed5mu(0A2s-%YZb*v{%ij^@YRY;nCy2SR{XOAeQy5>
zDDeHt?*VDw7+=yw`c;?$2*kTRu!4*+Mdwu8@wD$o6UPcr2bvq_Mq0^b(_C(rYGlv^
zO&gNY1F8a~2VbQ&mRe7TvZ#eo<_NIbDuE4#$Sb(?KzO}hCrlFlKGF4O*SKE9lls5>
zrw1rrmtBhp#H9wHK1U#&&MZt>Ht|)(ZklP<r+#|+F+SCUZ_uGu!q<L7tzotDLfCqN
zwSP1}P7epR;UEoWppLvgxqL#@MJ4)`>;KM|F?>ArYL2+HQ@JxRGm~eMFwCSN$EUbR
z_RtVJ;t7FEg0sK{5pIpu6P-3-f!X?iZ<_uZn@^JRo!$=bpyExU_t6)gyL-RbBbiR2
zU)KZzYq`B7w4R@W;<PUdLxaxNG$Uj1p)5%hz^O;KPkv@F{_Bkvjrf(S7x~blXSjBO
zx$NF|e?;*EV|Ly8UZ3Z0Ge=!yJ8A^+-`i|x|FwZEj~d({(qCW0?;lCU>+BU&lz&0V
z+6D_V29LcGb0umLU!mYFM`ykk{C(nPN6L;R3)>oZr24g@|G>6E)G54r+)#NiOa%OW
zVvdZRzre?iM{IvebcwedKRTtm!alNGO=xBx9D*$ycTD5u%`LeELiyyvWC%D(`}z9g
zawpi9eW%%YK1pUozPa!?RFN#f(&+Q$S_c$;?s+{%UI=G~%rDAy-*>;<Wz|p4eNwQ6
zaXN4nOPn5-+6KZ+DIvmkN)&W1_^K;tcL`Gp`YaR}ebSj}&=KWB1DqiEmcuZ(7NB*W
zb}(PNR=7Ojr2w0`HZfiYo1CPM^P|FuypoAV31nDJyR~lUxI0A-cm-cXs|YO2R57i~
z`1}d^5wH!tYU7g^c#w<x{3IQTG?F@XStK`+uj$YNsaRroIYUs_FUf{M-%Cic-3*g6
z0JVuO+MzZV0Hxcj%@m3uZz|Rnwqb@Cty>6oi`qNq-ea`2IsX^)nRQ7bi@J|K0)W;8
z1hz9IJ^^TLUj5f>nbex;q25WON8_lo>YsFxzEhQ_b)l)NR5X$SJUBE`#j8YTVv`pp
z2em>HpF5<6#5&wAz0=cV#&mieVBC^EWle^u%V~W7xV)mW!eq66e%<*Omm7LE<FD%X
zHwFZYZMo5tcK5BHt{JWA(9CaAGLzc&yk<Y5H;~WB2ywt1pdC~R(pvX$$5rEdt=Gnz
zK5xb=hQbRFj@AQ+13G%Yl7J6vohX()=>x-Jyvnf!#lbozbq5<z>q07EpmA$!IW^vz
z9=Q{_zywDZ(eQ*$Ijdjg*PlG(ZtNPsD?v^14_Wf&Z)@JE-!feN#9C-pFHdob%7IXz
z+f)M}<EcaZbl(2(YwG=E`+Mn#w=Y}{rUw-=EXopo!sxEc-|G_AHyfhom?U{-4U&$X
zI&$Y3lRt~(^=!sA{G4Ral_`F!Y+(>t+ld7aMa|H~##a7(actbaZDnY??{H%|9_|XR
zTicC1K=|}(P8Fj=m;9YQ#S5z;C~6|$q6yWwqUg++m7v|=E{e@=aAnG98+<`?L|~`h
zHqiF>iNR&POVf*9Mqb>vv~&b5b!I6YxN0alU5d%zm~cQ=zd4-WN#ELga&Uh?&Ef_U
zAz|~;!1!ZJ`**7{msfyKcM#fbG@;kZqyB*pB95wdEA=R3^iY3?DgTAG3;NhKDRU#J
z(Adp5UrGDen>d)}6xE=p)#D<zPFvGVV4^Vn;}P=}Y<>xH^Qj?kFQ%U2&dr9C>whH|
zxa^BeTQ^jqTWGGr1r`FRcfFuV#=}Do<BX8_{KI4#;(V>ufJbP~8Q=J8miGgbkFyk{
z*UGLU+B2F^JRSRSRJ;wlvYtcieI35hSQ=p>qGvy?onOGv!P)?NyrEcdsYHYmux=iG
z?@A?N1QgLa&jZ-~p)*j10SfcG6vA{cz$2XX4+N4NhLfYgwo9Mo<oumb&^TJ@jj7Ye
zbiH46zfctr&d<$r&1IDb0_8jgPxd#_+v<0QeE%a*fB9(9%k;ES#4r9VDO2_6>6g_I
zFXI08f(mJ2Gc`812hmjp!>BY+>sRwXxA0`b`OL<Jdmkei-*hhwqrR4uh5D_b8)ifW
zgn7|mwk@Ee$oF{4H26y3>Q#@#GfGMt*G|>pg|h=`VVKUUfZ3r2qL!0&=LGS{h1sKf
zsNZKcayv_l@#Gus^JUS91;>=(!`8ImSYn-Tdl_ZqUsuVg2VNCgm~fvc&?_2SJVKit
zrYR1dobkukVzg?{<{~OX%`P9^m{E<sK9|7!)O>eBR>E}*U1s<?<3%l8wDJ9Tc<s1B
zhOH|VWbeZ-+~>J?k0U23{JK#EJg=h$-+`}|LzbW1(PA~gO^&@92@C#kgGr(=J-SO|
zxIAo75|nC=G)HaePvC0#_(XVM=Csv=vKca>N;YN`VLIxLf-D&nMg5ZceayL|XImQx
zH!JGn0l+i;jX&!U+n!ZFG0)4PX*VC_nAlXBt8Ybhmpb=exD=I`Qp7HQTE#E!0}?<C
z=1%{IlCvJnEI$Q=?=X=2MWF%LW02^9rhUQ3Zgt&6FfLp}1YwX#X_#Fbi{dP?hOo)R
z95sLuG~ibVxKtA=(Qg-qyOX@08XvlC%7t^}y~j&`ZDBKl`!rD+x7voFAIB<BV{kaY
zA_q$UhRU@^2c%^ysWA~ri~_}nFRhR?QiRiH)N}8ulRe&N+RP`*<Y8X&`|-6fx(0GB
zHsU+WM-5@U30{T;8KF{a8&6+8TX+6wASji1Pz(}^?6>S~%al2B-*OL*n`;GbGVDg^
zacyy;iLh8ViC@o2Se(X-xrzupYA+@PK)t?_Jv(-`Fwm|g>tW>J!{CZZFYsDs*_?aK
zsSWl(FsR_iPi;twYzEAq6<7FHw)|IyuksM`tn0_o>njq{&)seqmRp$pJR{w4J(MBr
zB*}St%HNAa3j5piKD})Zne$oRfCL)35A<{wi+G)P3iq~yNm&A-C!JqYDSiuyV;Se#
zG7&*jHU*Uv)No;`hQm8%2#?MbAgA8-lG2d$(u)9vCYn|caEA#n7ek<n-Pd0{YGVvf
zh|X9H>#Y@S5>b!0(>s-iltDiSpA7#tvLv66V1Bx8tS6J;JAV%iuk7=zo_$I7xZ5Eh
z7O@jvBCd^*=9fpNiv3_d`-r(=kOD_M=;n0~3^YX8I5-%l;1*!{I*ZozzG+?^qB?gZ
zL=@>1$_9e1^_kr^7sunTiESe_fqWwr1I}aI)6+0qqDTI<<KzWA*yLEtyE4H51JA;0
zJk(cgIV#(}<4T@w;YjuB4*8_kI{PiWGPtZclm`pK-kz*!$nu)T{@@#Iue>)ZR~n;9
z=H8ee2fY4X@j*-mkxwBBLtQzywLBWL>-e-xLIR_#(;2>7SrB`Bi2#WJCJ=d1cw8n8
z-%z21hsI!s_X~ICeHVN^@c5eEzUB-7FpLE>L@<#>_<2PWoLWkmqT0&9kCsY9!w8Hu
zoxRd}F2*j)mFgR!Tx!A3q1?bbHQ7WrwjwU&eN-}#Ik0qr8=$kC>k|qtTn$;<-C6zA
z*Y$uDdOrOo5sA#CR-|3@BSy1n;`x}EQ&yMXOkV#}5@RgHHvlmxVMGxTFqv6$-nfX?
z`uB-?>O$E}+^4dT7eDWC75g@I_wfEb(A~1((n%*F!sRP;ea`ZS@T+!A&j82U1q^ly
zn>|^7!|s>dp{xPkwa4^I{*+e(ly@amOoXJW;H;SjVEBd<ppJmvhy)l!f+2;tj?{lZ
zgo<iP$SFW-cvzDz9*ZkUyAlEfnxUKlT$Bg4x=;*oYYUfEjI*?j@8+LbMxgtCgYF~4
zB)mvt10dNvzeyMP8spddNj^&Qlo!ZuwX~G<M&*kE#^*Hey98q6weC~~u%~a-wBO}f
zR>k%(SmY=v&fq6zS9%Db|26#JJjlTyws`I8{y}efZuM-ZA6p=nJFx2)S&&XQ>qyLp
zOd7npu;kp+RS`x1>$>J4W|;yRl4}bYj08JeutqI-RfzOC%nWfX3h<(2<dCC;nZ4r=
zwJbmbECCODUbz<kyPb@yN!HWltCzV&bmQSkm+m25fgA=<8WFMB|J&l~*$JYH6(B=V
zaTeCn@52Tl2#okN+UH$`y-DuMNm*GVODM9wo%ZQ%v&y<}z><*XE#cbRJ5VgiZNJLd
z^cjVJounc+@x%r&#Go4}-34OS^A8-V-^>!}AQZsT$U#it9P3D-Ml#C=145J)lVg1a
ziZ`h(!c-WiWPZSmdh0clOC?!61J)En{EGj)4-gR&Vff5CJYBh{IL1}@CfFvpPQI?s
zXTs>bUv>&sBT3!tO$$j@$FS*P?rLI+Wz~hAP>DE;nFIxRnS69?HH^+d1j~AxnmRUw
z4Wm`Go1ev=5`Dej&)S$MNds*HAq^1tQb%hhO})A(@a$^BHwfVw*D+B^e6U-4qTO)s
z2noL9gCP{?BAMtS?MOdoZ)<0JYlwl^;|9ExFP}I$^J?RyjlmTgt*>8t1O`{+ax)ko
zZr1sChPiZCONTucOp?*4;G7>rE-LF?{}7myK0aBl?S662GI^@t_XMT&zfW9hi90Lr
z>@r!z$Ot~;I9{!I>-OeE#nR=eff5gLxVcMDKUM@raE5Fp?`b2oF7+;mLm)1wh{U+c
z4BkkbX&e>QGhLW^;w=JjLD|~d3Y~RVCa-z~Z(Guza6zrG<JQN)3XMzbqQn4;0R_|S
z>hoE4LDZf6S834LomZ0V3^Egxc|T+x-~-*(x-^c<2m)<Jj~qPw>bTX=CC{BhQ{NgY
zWLF{A&FnQ9whySSfscABC7j=}ZZ^p18F=jTP>^>NWDOnKDnfB|FM8YeYH}663o`Za
zVv@cB{J{Mmn8;~5yw}R-pg1?YXF4XWtU)1lKNZPxiDb#otHi&}E*)R@zE1)EF*)<w
z?BIvD_>%RlOhJER%3xlM+p72NzGfBEq_hPwTtz?V!#-7|{n0u4Lj9Ks{$f=@F|+fV
zIB#~`w_D)g=tjy;yPi91_f2s5J*EM6Aj3#iv@y-vWr<TJk4_&Of<<>|J?m)}|6nVj
zyMRyOSi5PUg>48PMRcwmo1J(?_R)b(U1x^W^3gRbM(=QCg*Fq?IYrH|7F)b68-bo-
zU<nz?mdt;fqNgk<3q^C8{Eqq+!$Kj_*1gv_uS`QK|M6{;2d}c)mDdyUiKE&*aNjq!
zDo}Eg#ncart)sQ4ry8#Ped7L$6or>%R|N`JHq6tS^vyNghDNtT`6kMk9ljJXXsXPj
z+BeOqymrYi5<<@c(A&{)0*eK<VuB^E9tQ5~zFz0W%6h+Nkv!YdJz@-N8}qJ~;cEID
zBNS>0;}aFFMKG#?bB~i;cf^p9ZFHO2PDJ#So~li_T?pZuy4;*ci>V)UrR(il@9k*h
zkSng2WJs~vM{X|Gls~SJUAu9kDC`tmkG*Ldsu1;E<%<-lxQAR5?W_iPzCHqMRqe<L
zG`t1vB>;*-y|BzC-Eeb6O`{{>n0yn9<>QEh3mAl)xv&EGdL_Vr-nBvFntG@&rnX&j
zfe`_)<akhR0T5@_p&0e+^6;^mf2C0vd3@n>;(`no@1Er6^ZWGNnTjD`=Dk|D7Nh<g
zA^%)^`Z#l<WMbn{)8I|dDr>_xZOA6drLp1>g`9-4nV$~qqvKQNmRBcZjY8L2i;Eqm
zcLG->1*kx99_}xGt5hRD!`5lAaVIuk<iVPrXBn9Yx#DtVqIwJ3@^=%%9=3My9T{9T
zHevbP*B?mOtBt{|S$Y;4zc4;Q@%D5u*S$t!_5dT~K22n7ei|M{Y_4S*;EZjov1{XJ
zgUz0m>4qi^q}(3xZ@qJDo25v$i!r@&%lX^haN8xV5F4c1{S<k%c+J}7>28JevYgoR
z(b(8vj}`c;%=!5d_AU`L-uXg&MO9xKRsl6s!bOYRQX`#9dt<QniV2b<WS|~Hpu|+b
zSMkNU@#}Ht4y%0E+@C|eBugN_%)GXqwsba9r-sU^%G@MyxKw*Fy#l&yT?D&0&u02Q
zVfWq9UsdSc4jD1(A$|9lqyP#`ZA|SaG9VR6r#nogwh#j$2J@m<V+Cr+9=p@_1v9@E
zxepbo{-B|mv1YzT?}%439OiOO8-ft#)HJe)OWW0Um?5|l#Bzd-@<fMX>5H|2%N`<K
z<PorsP?37DmI+_OzIgEVtHv1md;&7zzuZc%QLT$=<L_sE*$KecK{@j^#RLPNoGG#o
z)8ppfowYsPeqk1Hs{Bzv3^WQSgb90=y476mi`E*ZjAKs!Th+C!1gFf@B1zx+apAjj
zXYT}u`SaO1cU8uAAUpNEnbZTakVP(=;uh!ZZZdb;O%2)e1ZNyut6oP{LxnZ;54A!8
z2J-@@aVzzB8U}__ONGL*>y*tGw9w?M2WX4R(&A?h#h4=j(8oFjzvOn=qoRf?5PD`7
z`A!4O%^VFJ&H4%|CJMgi{V<W9@hCeLI+avwTf08DK4ODBvbw>Pl4~ms1G+}n$v8)@
zlsSB{-`LK%e>03;Tx(#--tBnRQ5ee{BeFj<)6wJ!ZdfA`sZl727$CV<8%D{SC>!fy
z4@OLHE9x%XYcw=>By1+^)$xxq%gcX_4#q{tg9xC~n!pu@dIw1(v2J^d-;reOY{eBx
z)kko4Xa@(>V}`j|)XJw>@9*QkpAC3kG!YQf`+BWuhphlSf5!FR&b9}|!R6@WXSqf*
z{I_FVPo+NsbRS--tg==+ZY$MqLVU((9r$fJT6+T4Qd-DF&h)F$kkT`%Cha{S0>Cvd
z$dDB5x-*|pH{6w3JWBAw?-A2-8T5Dwdz7`y=H}U*&L6d5Q3mq&EKfz!Zm*KAI%VmG
z>#6tQsZ_!GbQ!D!gTX`}vplyaQR{(8`N<)ES4VWy!5YJ)>*=2l@CTklQ-C-%gG?5<
z-D^5Xt!IAhEm`4SuK=IV-N;Q*M8ItI%kQn-U2~?JtC89XL}a@d2`wGfRt3WD<fE6|
zGeF@XD@$AkHA?j#Vikr?6*8?A@OZUV@$D$EEk5>4m(cWAU;fo1G#ZJJ$+Rv>%UkYA
z#`8c37P2TvypDbLJ)R@FdJ;>TJ{<~YG2eD*{(a)&$mh~~;|ybL^4!vkKEnIb*TvsN
z`qG+d8()G4G_(V7-X74!gx{m4(nFfYXk{(a?hT1HIu7!CtJIdgQ)5%bmN2B5c~@)a
zfLTd>7pU0*kY3BY)iUDbCqlkX2_DCk$VM3xQ;C=X!4a`c(WHHE4wNfC{+Qc68YgLL
zv_^aXQZXav2ap^UY|IK8IFnScCkqV|>dHCGQv<cKpNx;jXTu!sR^4eo_NyU%IrZ9|
zFTsf^`S4b#Xix6lGD8fIWqPG?|H1D{cUKO=`h2Ok7KdLITywtBh48!ZpyB=zW9@FI
ze(4ddMg9yDE@LzL=j84a@>jm35AXe-Y}65a1~@q$-n4n7PCtC6D2~D#Kwt<`n4r{U
z8h(?k-T!U^V`Gz5y&SMS*)I+z0P=>?bkW6>#<yEgo&}CaGl%H=jJibFvse(%5jm@1
zhwavDh|QRTKYQQ$PnF_}5`BlS*}cd+uh51Jcv$(^o?JLCGS3<Jm?)vDQ^F|4Y5{UD
zyKI>E!LGp!g^c5Juj?3t=$NRk`o<<Q^e-lIL(48~IHBE(;t=W2kDj>739*75<ZD~j
zP;5$(gt#eDDOiJ#uUShb?REYxWB7#)V>gQcj^~f=pl=;H-0)=((<En!MdJRw#_$dr
zPUK`s>bp3l)M*;ke0<Hsv{@)sKk6CIR2$LI2yNKRe!0}_bZKd&VF~BART3x<0Q%@k
zK3Ki==AIq?F1C$CXdxYrUFicPj_GvQ%}N+)XS@k=$076Htbz59&PN1(lV)OLwlG1r
zTA39?tAd+__)Q~dS*-KrzJP^h!!m_|kHR2VV`Jm!-KN{O)!N0q<~yuhZ-~o~TLrJ9
zD3E04JfaYx$9rSi^p^nzGASqn6S3#(_!}34^SXNHKJ;#R|B_StQ?+cL`c|)8>hawN
zu%N0zXgxLjv=!$u_JT?N599~us-bNsj}$v@99Tws|FQ2zL92+JE2Sru-1_^*NqEuf
zcqpqrSo(l*5NdP2IwOM#Jw^?J+95@OEGJt}mN(GZ7y0*zZ(Gb48kUh~|4XihDy34_
zeS2?AR_KNyd`I;UrZIwLJi;0n;c`4aB4>N?(n%68b4`U4sr1dqR5jyM0nShL5{B#x
zRLWX8`2jBjqNpANB|Vo2@Rwo!UhO|*F8=;b&UHv_Ec};J6sBMyD-F|^*`c=4AW1sX
zvmYhJPeweTVsWe6c_wK6HJS4Tf4HT?81314J=5{(jKdDHD3RkyWio96y4h$PuXhS_
zs+{s^bJ2fcZG6!gdF)6?+q1|n$i{-b@wl|a<_JYRY!zqfZ11Qip(pE+zHPxpOe4vB
zDKalurD+~rl=OtrJrt#&X<}{QoX-Ba&OZ8r=SNq!)`8;xyck-xj~{MiWoy2?|1iqv
z-Rzjp{EKH^(!y(JBONd7N)f6;j!c$|lWO3kjjQ9g>pd9*4E(Z-FPyAfvFJHKOilv@
zRaX?nWD<#C`AeVl?oJ<UR5k<z%XC=SRAGuaA%4AblNB{2<hXUGTNWfP45_+{$1Vp>
zPO6z)XpvX{h>&Au$kqaS?L4Zl)+5zUAAjz(liD2>xEp30_7qjW22s@JEs!dg@2_7x
zZhZ4{v_aDB?syz{>q_)~HsP;17_DrZ6JfBve$FIZ^*aeA!&c%_$%B4+tYXNhZjS;@
zIIw@rdOR?7DkCs7hz3o`LJJRNpvrZo;D$&H_mz7}^8Z1ss<}*i+XDXw!B?<jT#H8j
zFb>YjZw*F#`y3}uB#1G|e0}vE$>sCs!G21Ph6bduPkF-l)S2cy1@)DO*;@97N%!M9
z06Wm%C!$dYkK@3f7JglCxX?0kKF(}}Nw#5qMB~n*<70F5B6Ri7>B!pk^$(`Ud$E%H
zJA<YE5d;y??E$DrCxDri&r$H7;@~)e46!lMogFg6LY-~55fWVfIv_;p=}dnh@;ch!
zgw+Z{F2EcG@U)81Q_E%onqZklu-@{u?yDfFPc46+_@U=kVLls#J~!cR-E;E~C)>ud
zX*H)gBg5cF7gZiB+WOx1M!_-%0hfL(H!;Gk49sx*(3@uKadGxM9+e~<!KS1LRIgmo
z@Mhg6d8Ws~+989?Ea)x-y?($g(8kbU0wSJ@Ww?8N+pb~HY@q`=;q2mPDNo@`!{u(_
zGYaNy7!4}(q9Y`dn~MKO9~6jFG`+5Id)t8PGt~68K_|DtUGR+!Fvn{#=cjwY|MI-=
zIO8_ZjMfD9_bR+>-70y7b8~qCi2*392-2R6H}7;J6Y?#ukY{GZ0}7qzWwMKYK{AE~
z6$`VS5fKN6Z2dObtA@lG5r!dkgy5pps1X&ZHwD-=F~U!NhwaERjWE*FF<hqCk4EUJ
z9|I$u9KN$QScDLTE=*R2K_pXLK(P!Z{==%)CUGB2jl?sLYGj$!b3WlGHs!j&!#8H<
zZtA#;*_wgzxvIX$qgwt`mwt0kfBC99`%3;)CIXZHy1a&0kroxPeY+XyK^d`@T=FI!
zMLMzcz}5y}T}PLpt&2lLdx3*e6+LN&YFK=+Vf_>?Fa?_yX?;<z4+&MmF0_mkG=B4H
z1{x9E!WMyq8z;Tq7y|=N^^+9>9a9Q8r|cf*4B}=qUtHp*2EA-Xz}z1}gB}Av46+cm
zJFU@OwXiP-sM)h|29#jS3mT`F2BVB@>nhR-k`s8E38}UkUlfe3#;X`7FRPe3bh_<C
za6vlm?GEik00_yY+XL>%*-zm$aH{F9Wh#Ct^S;rC63tk|EI?25o_e=3$nL4Edhez#
zRbp<+jp(VOqoK?;)mFbTgfV$zgM3`(qthZ#NByAc?tbopr51Ual!(}nKNIHXM#|3p
zLtx@X%JTjDgm+D+m`600!Wo-j7f%sb+ZC5qD)|qNV7dT#<q=^)!(D^#sX!MeY+9le
zZFPZIbXyAy$N!@X!uwK-Vp6ILHzmG*$n@5GN(ShUhn#d0TSi;Cj6d~4z=kiJ{n_91
z;@{MPcE4T6tZlsStz=lK4gYBcvwLwpMa_qWH)WXIcaOg{Q(ooKQ!i}ZgAR>+;&!#!
z%k#)2USn}yN<>7whOC75*yonI^mK1XooDk-nqjzr>WkzOv)lmYkt|2uJ34kd@m!i$
zRW%$<V!)SpMy3)LJbLyssSVHW50q?c$(;pjh$IlHM!@<07st{3QE70DM;!)GTVxlY
zF!meW`IcG3$<i3700Y6pXl#Jol$sQ(-t|L)*WxO`!LzWYSjx)UV(Rq0^DXdS+Ezbp
zVgjSXI?3Q<6R3(dy0Xa#^o;iDNd#`k2O6*{Q)XD3l~HhKPs=+!crPr75<AYLM2U~u
z){(&OUO2mj?mm-+58B)CYz_o6Kkpn;5RnI5^tPsFACVp(2D}Kl{EzYqhmR5OjEl;z
z6#{b+m<x57usW0ptKL{Eza#*Nbx}XlJs0BQalbCA9x0i|a^2BTnQx2FRKMK1nVrx$
zUJeWW`$THuo9!M;$u<+QvzULZ5S@ryD<1iU@iC%#hrcJc;_J8K#NY7lNiTKZHJhn0
z6+<7-%5tHzrX}U;cih8Rcc#Hme4uM*qWvHV3ummp7lAdJ&3gV(2N~=Pec542SoF`S
znPyBT<ZhT+k-oDc7R02m_>!iVw{L4=-mP6R05k@E^YP8k(51oCeG1Mme;i2lH{$Y3
z-Ew~LPg1pslixCR`4rgQUOAtTQVlyG6PiV`v`nc|{IEUj3mEy=(yxrR+HSg^%FVQx
zJdOfhB#B}=>gf-4FU1f3v9J+;#_3~tgI90gcpVceTu;b?AdAFTjzp+<YAd&mRNAdA
zoDbx_Hzd>lR7H9~IB{nntEfcR2-A=s1bY>JZo0m@HS6NM5y|!*1^boN8Un)O=ICZw
zQC~RVSC+co50TUVh=Q;SS!>s~jQYIKM1hk`j^9lb8D`V9)(p1|v=rLknPlq2WOnl!
zCQG6jtZ-c%auEg!nHG_QlN-ZV9L2oUPrmlA%%m34k>wK~{sGy+rW9liTKDgjbE6tx
z#a>0B4%#r<2@GjTpd-NPP|VzhA|!q?ZJO|Ui5$6iyZ!73RsTY35LizZt8}BMItn<g
z?B^!$WiuG&*10`ae?R}oBkV>=vO}%z49dPZR065$Y<6<1(qkuK>)zchN939bh(ZXu
zdN6EgQC#of3FEA>TjGQ=N8svH$A*r;Vrh={R9aMRmWQ=`ybH?NSRzW84q5nPX;qpk
zeYN#V-aH5L)f;S7TjE*lrf+RQPF<a+)muwOhy_<85)$1lF9N@wG2Chv!t2;|5kLb!
zv*!BOov&@o^UHKDVVA?d4>-4`ZJx;ly5(*?hWJNT__{wnRQQECVHi3lyYqW=#@dxR
zV(9bO0>5=}BNc-cSqf^ucdr3ba-ai~mdJ&&F8%|o4(KFDz)T1OVmXP&Za1O>pJB8I
zI9kE2CdAr;?BHuk3dP1@DN@jqBE?jExw9`X-lY!{O~{O!_n~-<%#BofjEwH;Co0nh
zBM)d6K6ZsFhZn>exPS*$FG7E)IUTHwoTpQ7sPOs}+^p2pJkCityjGdh@s5{0`iXL-
z5}X_rXU&ySNNY9F;Kn}Y=00t3K;6+zt6g0CV${&R_T>&$p@Zj+q9L{;SUb}V_@Zi@
z@28nV(=&0a<Ezv!F|KGeMq{y{QJg1MwC>T4EV3k4EXhkcV`>ShG0RFRnO_ACANEy+
zd;TW%iTM4FH~i-nBOmk+xQ<?LhB#SWgtYTqrJmbHT34g4;&ElQ>*lGp$zuwy7WJ}B
z9+&z!;!NwISq~Z<gnyq{yZd&h^cjte9O2RuvgvP)lG1sdAQA{MJ$+8rdlWIx%dmYc
z=2q_O$g^pr9S__(q%jb;=)?hldV4+<*iN{2*fdo0G5Lt~Qj^_cNZp5?G?%IN+!uL9
zY9c8FN1)hdqLv9S!UPkFV&?sfH`AUgq->_0=zemxTA2~Kte{kg(aj1>dYEJwx29|t
zXg7Un^p!0&Ik^;AoMRHUKvE?JR0uGY?vm-|SKC#M94nUBr)4Q{gz#k8GE$y??BGAY
zAw!Z0tMf}}dOv;?pTZ2^(mwe6gvT@FtSFBY%Hnvhi^6ngL)%lHcDAgWn?7O;y(c9V
z?@Vd?^kAuw5hrw?wO0pk0AnI%!C@E8m~xJC-3#Z^KSh3oq%BD=JEgABvTHA_yfSdP
zG`BoF_DhB<3zEjx8u+cwot7u+_{=jbwWnMQDt;gv^zci~X5zyO0KnFo<7*-03>u+T
zOVLCCz^0(Cb_$%W-kuw4d1*#|(9N>0dHbe#B!|?fWI0<yoONA>&?t{H?}Zw_28hEE
z(3Le%PL1DGFV#V;wbFNnFM)mW=GMi|EIl*vt4P?A9I<})N%=?KE_m(D*@jyqEt4`Z
zBX#Vz(~8#q94%I!-p@n%>vrl%k<YWguaS0f{=JazG!Qpy^guJm<eGJ-9b#^kP|$vR
zH-Ccbk<%{BdhHkJ)XHb6RbMJSwcHS;Q0FKX2t^op9U(9}m^HCfmo$2L1PtbMwwW;F
z?oz6_VtvqCzlROCml|H%m@)c?s%47i7iUQkq-!=6u(gFlV8jU0Qa<@p;m7vmM*LY{
zY<^;`HU5GRaRd~qh`&_ZeyZ}Ol|LIJ`Rt*?<<A2*)V-r#<=*YenC=}_JEjzj6^sd@
z4?vK~D^W-e*A!Ww5gO+4r?~t_ou!`sVLDHo&FK&*Fq{66(pP%AuSR4+v?R6*6p1(2
zGM>)$V&_CA^(N>9bd&)0zL+WIRfZkUq{Bj42Wg;A%KyTvMnfx~%{&XO{)s90N-!Nq
z7)3e4Hr3{O-g{LQ^;AaX{t7@_i2@Av9LPHXKi_j2keLYjIL{c@@_*(SxZ1u}-hL#d
zU&?(WG_NK_DVNUHtQBP<cs-y7z>r6gm!X$Iar7u|#fPcyS*cDV9;_0Hc-UHzUtxK9
z!oE*c^waEWf4KN$@(ffAJt}sYv(aaUm!t}(@kq=eShWbs)4gcl9i}&;nbbIZg{Iut
zrlj{?H>BKXJepeO(hF4SGsru2dFsnq7wejvk4DpaajC5Yq`JS_W4D-ND*Ofwkn_0M
zZY*}}1q$;RiJ(yTym;Kh2pvQaI8x3tz8zErvam7d{c`-;sA0IO02{%v9uzq#5NS78
zQ)6)O<-o>?*fWk3cZ?@w5|B_-+d3_L16(YnNYlVE41J5t|Ir5RFeMJ13?lSqpG(i|
zeT(SaOmyC=cN{>;I`8lk44%IGbAdUB^xM`5=zjLnAA4GJk($ez9=2cc>Q_jem(H2~
zCF_#O%XengC0;f=0XzKJnKZ7d`@|s_Dm>Tsw0@(mo|CBe1X#x`2-MNJKf06Jh_ETH
z@pU`UGK>s1f)xo_QS2GIgeP|umst?M4|o&G1787+8<^Xj-qkJ-b*Ny=`5w^fx%~tE
zseI8c638)F0^4utI$maS%b;l0fRa0rfdLp>pEs_IWk?N>ElnB@p-=bv{U|KTivZpO
z2_q1GTF#u3o5GJRhuYfeRkBUqstO)3*v)P+_<wnrH5S+{=lJ8E&5H`ek)1@y8rL^n
z{+qgkx53@a2$JWao1{Z!UP0p86>T`DhT1bh8HLXRE}A<VDfCSidR(!kR{k2W1a+mC
z$6ADnbyxH&I+K0Pz>EAKimjE6xrF1oC<5p?)F%he--?zNu0&5URNa~Yk6Cf2haC*=
zWt3Sa6Qm0qp5Js*K4)h)DQ`~77Q37t8jaOH@0XEa8UAJvF=CM^9z02OJKA15uel#g
zHnbV!w9sXS3H8o3!6ZR;4KNW%WPt1K|I4lLBt9%H&RuYX7wf$!Qu#ZdqO518c<#$+
z0^Vn~LH3$d$h^&s$yD9>8eENttqo>@T|@~S1y3s@TbS&>M~^C36HH(4i!uQvtOat6
ze6g_t_)Z1SK0=9xjoiNN)|X1G*Xl<5R`wbCkV&_vE`KhG6t@qCzWFYLw)sXiWvVY#
zoWU^`X%2RMKl&*YHg#j!p&UMwfm>M1*B}{xk#;kWLijm{Z$kLBqCDW}<<V~I&E{RQ
z3gcsA%UR8m^tQNQs!?rcp47a^IX8hWvO2)ur1boAg1g+^Z51)Jt#z8`ctS`;R(mxw
zFdqz}z@G>qn;$z-3wQcntLdXUl0HpGo&EKX&UmpdvAI}Q0s7t%#Yn8Y?ee=WtJhAG
zLQ44earWk=uit(m1V=kkcBOjQEg7{3+p9%o&0d`{`Hli!A7R5j>rJmzZARLW1E?`p
zVFcf7d#J!5b`VNYtgOAz-!KXr(=yNJhFGEyPjO&!SyQjR4<l7o;*9L6%q$OIdO0an
zrid@r2O;qlcdC<9q`6)Lk;Mqiwdw=;`@&rauv*kC@L7zJf7ZBBf6-!NOc?42%|@{h
ztxR%S=A_l{@6SDVH>h$bP6L(hkId)SFL<YLDLwKL0A@=DI11_uU4~8BFAkTcd8uHZ
zRgK!{FMD<7jsojsqtrzdLb)3xOycrRw|{-(AAUl_cMczQb!Phen22c-{MgFr3ekFS
zFI**9{QR=j1+L*tV9E1J+@Z=d-d{(&xCY|1644bmp{UYbH_hCpU)eu5?BMcT7IVB^
z4d>zVaIb#r%dUg%-P<;K_665Ba<bmc75E@Ge;4o4g&v*nzuuWVmRUVkSU=}zH%T+d
zxN_#vxM13;yui07N&~m+@?WKp&^g>U9*Dm=uap|w=r#9A_2+5KOf}J8VA5-T<fz_8
zAcr_7@ev$PD}dn<)Ev0eCr`s;H<$C}C8v=2LBwnrF&m^8{E`F1NoM-o3PpxgH6lW<
z$kk<l7#V56&@Lh}guI}<_I;9+nG201p-IkspmlmnBW`8Cc>es()&-ZYZYD`fNJV8Z
zN}oM7sy-v3;l^k$O}us~TpQAPXP4-OPxGP)2TL!?qgZ11Rn)z)kX3h!f=WLg9n)Lz
zEaNF$xKZvAKBI+4I6+3h0DTneayN@<?>u5=^@{?@acDtU;36`05?Lfr1@W3kwj;N{
zeny!4MQe~qmp*^}t?zoSo7ehRr2WcIe`0#Ve6>Tcmss*R*7ZFv=S-cC!n6aIIJXH8
zCFnZBvMcd|-V;OUxH$dpJ3T{jKqxjkHtK%Y=vGrxMQ<Z=WGr;Dt3v!T=#1nwc!<S%
zsX=SFTf_C@Bz_7i9_-8t4aTqtQ)REQH4qZ~Ry>4`r5fE1ri2n?@U)xwb*8Ng1M-ir
zEpq&?;xMMB#P?lKlJmrZ-?>(zd1`NCDa=2Gd~u7Pj5^CGOwO4|O?h1WykJp9PqWXC
zsELJ<`)}%qAF-W@C?ukfc6)uYcVN0`qoKinVL{YuVIj6L`opi(O^8U~i18_EHjwVs
zL?f)?%EWglY{}fy7dGWjlb#~Qx2tC<8847aaj5A^8HN-L-%*%~!Q)b)#KPp|SgW4~
z6}R<cy*@OZ;qH<4ynIb#!(dO=UDyBXYAab;CYq{8UGW%z<yj!e)ga2bA0E`h!FraD
zBD2Og3#T&PtYNIP&bArS1MQe^TewEr(>w*Tuud-6mAv7(%7#^FgRRBtzam&AJ}}j<
z@cFm>J1B09@s-c&9Mk~(LpxfmPrNFx&qC7(g_&W{=oE1xc3E_p^Pti%ku%n&<NWG)
zt#<b%SVyr3l5{g%t;Z-Qw*(>jJ$Efz1|YAv2-DxbzJzMy=96SG=}uNWQn5{lM_Ih*
zbj4TiRMiyzs<?$YcDpYjW@AD|+|UOd(PILsq!A=hr7ZIwQ`;dm+^xDza{F8(Ql%P%
z7tR0n_lawxM^=q(czky9i0^SPKJ|Ts<j`QgGYeE~q$F`VpMQPt(i@3(&+vc;cQK-c
zYioD8Gg{GV?QweTb2iOy1jQpkULV6z!Lct+9p}EN4VWNF8-tz*-W>!)>n}Qo6Qq#-
zz-%d8G8K)V$5(rhIsc*3qatCLCDRvIlqBrO{|9?-9n{tv?)%bTU4Wv+DaBobON$nF
z3sx+UK!D&{qs1M9Bv^5R6$vgaE<u9_Deh3bXm9qucV?e^&fe$Fnf=E(cV@4dwX%Fx
zp1gV2`#$gUe7;4PTn4v&Y@&Bp<2HhK!kzE3s6|9m&_Y*5(Gb%`_rBj$YGqx{KJj;u
z>)^v!LioRJ5&YNczgp$L?C^h=j{ma5|4;3pRz5#jC{R40d#rb04utncBH;Qupuor!
zJ`jFcy?iJ0$G>IRTll7&+EpCl-O!MaFaCx2L7Z-k)Lf{21K($K#Xae+dYJfeOZGmQ
zrI%0wSqoLIe_0V}#im>R6x0JCCH55CKVy`%s;Z#bJ;6C(+PaP8_~dwz*!m2?)GP*I
ze^2}UC+bvAQzTM{85|Q12I<ljTAQ*J>5q@u&v-)-qmngU{ovo11J}R`h8{lV>E$WP
zbFqr~*U7gMkK-6s*Qso{pLInH=;kVeYngkjYlJ$U34MkZn1XBaYZ3SbVeOavNAjwi
zsw`n2-FMcLd<8d^kG$i{u6|;h|4AWTuv|Df!x@ZbuE1zZvPv$43Mq)^Kqi#ml*4LY
z*&Fd%8%~mbce*&tTcb!mNvfaY7O_nxE1@U4>r3hCz&ssFHeUg$q4k>jmDZiDv+07C
z_tuZ!-aSb%cG(fsh_FCCx1WF1Uf5qa9?@-($LdoL1zMO26efJXY(3Om6rK|mWh`$R
zk1f!2qMa37==oM~0OI;#)_b+Q!;rZp*=Pz8jOz1}^<<VbJo2+OvRFV>mCCMghRSky
ze5)cBCCfDnYni(j-M5$gVA69@UKE$?ROODk3E(y9Llbl4jnngEubor`nz@YPw~Dm~
zs{JnGv<%*Ay{=`VmpbCj?f9mIzTt3GmRp|%RI^P+K=8P16Qu&g>Wfnm02RQp+U^+b
z|MN=y-c&cFjYsR*q@d`6N5ZR|`;u-CbQ(f*Vg|uMLNi1-pRcWiccx`=u3|EcyZJ4E
zb`GW~756l_;0L&!DY}L|ZxUSl)R1Id<^@R9wF}8`&(Wo)LgOu^#QmDfC`kG2GJ)H9
z83S~~C~FW@R|ky+PZ(t=rLP!Pp>uWJ&4U3gWzf?^BeLvwhleO0J<r~GV0w|wfRGLP
z7NPG4SyXx_M7h+ACBlhk{*GQOSVXH6x;i1tszyU(9{!r7R;y}}on%|G{}Y9~w>9MO
z3i0WWggM9tgSgFHbtAE9T&+0+1f@zaEKxAsX|*poJE_&Y<!$(IdoRfM&T0Gb@inuW
zTU50knwuy4CmxD&FT19{nrDEkI$S+wMSP7%g&jCIH#-PRjeGYdY#_GvYKWCn&F=+(
zM$fvV{P=pd`*G-ou{fyftTw}}3tT8i80|R4leze`s9#HH-r7SjgQus2BB4%!7wlqT
zQLCM@GL`U(yQH#%bG2w}+U5p*Q77e4b&1MYa7H26h-D2xywyQ14$11hg7y7Y${J@P
z8#=kYp>R@F#i|fsY?Nd_t=q`%6lF-5R1A)7!&BI|kv({y(FOQb9S`RzGHms=z2Fs^
zlSlaFBPk7h;0bYqnKFr^nhNNK2NE8X6kWWD#0ps?p}M02U-l)M{nvjx44mynbk2@j
zbS5{i_=m^wc}6=1sA*G56Jg2AIl`kB;M;;06vcMyhs~}NJ}+Qoc9`<bt?wkqw(IR8
zK;*k5v&=@EImcD<@M)|l9_%~|Rc>Ir0eAe4NH_9IuHHa|`J3soQ*@eVB}gf$>X?WO
z(3M|S8G6x8C{*<G!F0~A$(ymDMwr}!q3r1CT~NIX--1e{gK1M^ZvJKP`^G(~9syLE
zs1mnemk<IeR?dW*k&9hA+=Em}!mHygCwV>h6%bpKp*HV0@>x0cl5*7al=C^26;)sH
zf%2)(NJzZXvaDZ#&j|iR!z>q)RoiY~nx)PSPUO~-Qyb;X*<M-@H)g~?2vreC<FlEy
z);9mjks9;W1|G+|7egzUJ7ZTnRTp6`RAyIfeM<35sa-*#9s+%;RzdY+i`8Y=xMui8
z|M}yRWI^hz6#50ze<-#2)hn+Y#xFO2^H6W;3Qj^coA-m0w6k4#TOwaA>rs%i9-Dbe
zy{%&`p!SY(D$lf;&Cj2i*Fc%)KErV3;tWhWk_Mtq2fwYJPE3A}9@;KyF7>AvWQ~OF
zZnL+4k7iJRbDY$7-EH)z%3@%LUNmbv<m*=Lq?4C0n{u234Lo&~nlIN9Y>pJmw;^f^
zT@r@}6R*NOt6oi5!f)1T#&>`h)!#(tBn0)cFJD1z)lZ))*5bY)BuZ#fyQ?8=$zYck
zg)^2V!ZB?T4B4hO<T{%z<OAQ}L5XriLz5y8Sa6L^#X*aX#mTBJq^t$^bYe3(uU$k<
z+npFnV#{AlQSYPey>9)jrF_T!5=cS;T(kK-c1pHfz5PS|jk))KPDlzLlEmBT)4-i~
z&Dbir%yaCO^29&&cEi>e#TbQOdd-VYw}#uMR)LzGr#5IZGS>a>!aLDqkGvN`D*h?A
z0oRS_59P!tJhoDOWUspVoiwYXdhqL>wtK^*R9-;en4g(se-${cxENp<HClT9W(_ys
zQRG<tG_9t-RV5$B+V3sxJ1H9H-VAJcaJ`!)(6cjb5Jw-OoXBA3d(6~pW1y-5VB;wo
z`u69@w)IP6XV^hazQ{4LNM1d*toSYLiP}i6YJ93YV^_eSwy9hHB3sGsq-@W-9R;Ah
zVZj-AT&0elSv$S);8P)0?R)1Ea^(<Ec*SwPDbobm*Z9Hx<q&;M62jw%-T>a2i?YG;
zy&e~wU*Dk&6TDQad`j~XmKWStbC+uT?cT4(f1<TJoT~aUKKfptOc0A%aow>_lp4}D
zk@q!MFJlS@UMfHDYN1VLB12fMFP=uQX%7QMX$w-O?b0yJ_`gn5(z2<H3}o_4Up%(|
zFnzLwC4-*c7Spk56QyDLN;ok)0nW=bC);&CTKE1H(!0w?tX87{q*~a(Qe0E}+@*vr
zZ$KcuJEJJZj!G2#b|rz#I8oQf6QHIjkqn@GUNAK`f-@U6ot#xCe7_vo)ds$1^UMj>
z6{B{Dn_r_oX#%!zjo!$GT7ioYUu`u>ibboIx@)S6OY*xMC-^E$Be@&0816eI!$+0@
zV3=%c!8_Q`tc$w6pF`C!pjjm$Z8s-Y5QH_oEd|ZYBfme!ZnE}K?aLMF+G1ca7Jveg
zWK%DdzPgcwtp{UJxxuQfeNDu~iqz07eR7J8Xw@%;0D}`L(GNP*Jw*CjF*oot<m&ji
z7%4|~;#bLbBdKJIMd8Qcgp)Q?iGyL~p$^o_D{cl*C%$>_WHbCYS0UP;5$=>s298>N
z#!Wv?=O6i`3bP$i!4gh0TJ-xM_os`Hv+e_rIKorjUSjF?<<!7H-~L0@#4fB?{ldnY
zV=&FVb(x&hm>5bjM&0!B6Ln?o<c%3sJN9^qV0_QmU}ebg-gfh<Z!U$kD));a>q>Gq
z;?J%K-S12Db<s@I&OykIiEds%`RMJXw$siZ?oyvwOvy;z<x}OTwM~y9hHYWhOD@-8
z*-&N<dOk1_R(u&2oAR?w)l<RX6?9#8X`&Pu8T5%SK^Y~LuiBHS9?i|ftvFp*=pW%-
z$by{<ee`RJj3wxUm&A1MBly^XC|S;nE8Z#k;qC63m95DlYal!t>=FiT4Yt4y3R!qI
zHDr6dY;69(nwa;e@jD&=!!H{ATzwf`N4}%RvIVu=LKUu@DE0?DOOfkUfb$@;jnS4o
zI~>i=T*I9q_kBSHpMu7$cz<RrN*s-?qHV5b4o_G)-thqLDp2M(wX`b1miT4<d3t0m
znv$yV*jh5({LdDY|BEJ@ux9c1DtgN)%x#t75Mj2zk8CpEYfs(DYfv@e>lSoe(ewin
z$sytK)8w7f)~2Q^8-Kw$%WYmZ1+H`n{=RG6(AL^a{+zJw)raf?_N4qDeWy=#-FE;g
z!m`1tbts3&5upj@joKNnMyi>GqCV@ADg>QLt&fVh8lcH&uo?T%vb`4l)#Ui?#Bv%Z
z<iPNEb8p3SB!~xt$LIB+gb;KtTuBZNak`f4?zv3hAm0^T>O{`$z2(~e#`VVgEsB4P
z(4yaSW{gay)NH(hq661y;ZFh|S0?X*-PV=8ib?EMVBIVND=szY+M1${kh1ZM%APY^
zBVjLey53E9{cE=WX(<hlX=GEb#nT3BFzB0|iCB&U#|@vDrH4eIfu(<W;V{xw3eK@~
zu_0HyU#L(^&pLRu4z^6&elt-JcluEKJ7#W6m%j<C<7X*h7PJ9$IOk#5)BNCb{>uKG
zTx*Lhe19z#yB$Yo;5da8N<kr!V;L<oZWjfePH0Ap-xf4kS*r7q)yyl;@%AohXSF}#
zV#KwWk*4M0y@q^~Eq>+$n^8qWW{UDLw$=eK;$yo;=5L6^cragAjXcrDM1-kYVdInr
zjvpDe#Ts4aaiE+&nW?L1sMRNY`H$bd@xX4SA$r4i1s3qw?q0`yW6S<$64e!=P2u)w
zx13)O0wB*%DxNhV=APcTj%rMv^Hy{Wy*bqY0v&JG0nE%<Z|K*)?W7q{uXSBh5jO8>
z@y2VimCO_U@X>F=__$tCO-BFz1^4nF4hNSnRMEFDox=SDb@jF_#VkTggtIlhfrGO0
zhUf+vy_^=Jy+$cJG>w^i;y<eH;5UcM%N3pXZ`lK;$1_$`$C7Id=M{3W{Yu~!>Z()1
z#)KbkKsrt#buJIGc0=(!wW!kQ+#VaB$9?<q?pFHqe4$$AkFLkyA>r0@(&E~=px;L(
zQ^1Mfz17os;Ps8dA##prY3<}_3@bm16rHzZUu3nQky9ySI~cOrG8;elJAWQT```4?
z{;jX}zh%wz!snz&ZV8Y5H~daF410Hzkh*d3ChEhJ>NwiycCjl$i0c8u>L>6eY4?Bb
z`R-=DXG*tslSy;wri@v146#Vp{=rds9icg|Y^teG+Ev<G9BITXND8$pOm!!v7;jvO
zDdQ_`bW~IS?k~K#bSdSdlIQTX(KUj`INbliOvgA{#z7-jbzFpAMUD$E$O`o}=9_>=
zXV^>_ETUo>IU`{riig6&^Iyg4qN!-Pt=h5|N2-qZO)8_`yYsA@Rx9bkA$(neO5S)8
zkg&|O%<|&}acwkj0v~wWg%+&wYO?lz0yQw)$(}3+%}K}!W=Tkzk$kHc7M5fOQdA7e
z^xdX{zjIjsHDPXE@|2&Qv?Qq$izfh77=xT5Kp-0~1Y1$~I-Wk2_1>Y{9XScL3_}D#
zn|xlS9SBm4QYdO%fh1P`49pP6_x|VX`X67trp~pz^TojoHFVa+sO*lq3ZEU#+=95*
zRUtx4%+hq>7UL4mUYli#`Q3i~r_y1a7m`VmNt_)bX<2yzxhF7}vdUjXDzQGzTDS{;
z7r9V>*D#Md=q62c^LC{dFFUbeu4dG1HL52n(&(78+VG+pKll=k;s{F(dH&<D`mU6(
zZjIxJ2k&A9ur23Cv=j#bMkc$vtPJ9d8l_EZM8Wt3xz{m?w1>24l3)GcumAXq{O685
zj>Hw-K1{ncv$}qj{UGSyZg$6O*=i!8tt;{dMPQYU-!CiQ3b7d*>xt1i%AKyQO)IFt
z!7b#;K{>~zXDzv+UsfPDkhtqBiGw3;TPH8)l2hf$cT(}ktcK!iKKC<Tq$wZKSGP{I
z{4fJ?sHt*A$5$8Ep1@4pdN0Z#onRRFUcS2!^lY?uEa{ji`L#`pS6PpXr;5qe(YZb&
z8G5WmzO0{7Pp_srW~dV%L(-hbsj4<M?^I<3)=N-hNu!Qu6kvQHo?xetLSrsQIbL5y
zok%hHm!Lq5DO7KVh`Hm`5yq((C}*Iwbmw&aSi8Lc!^iH^#q@StO@D?d`ZQ{;tpe#n
zv4Fb#UTf~Z1VhOgfLgV`1Q^wgl{-g=OX%vf=GS7wwJljnZ2=UKWx&r7fw631SW+Gi
zfk5x%MbUBEh}5HSU6%Q&U(-VhTnQpo`>LXY@;1iXMB?vcGZ1p?X7$>R%0+uv(;Do9
ziP?E&TypL*;+wIbUkPP{x#rtGfhJ;E!P`H7*5CZ&EmR-pN?q2dAWkp(NIgUMJ~49S
zh3vDDHM7AimtcWmv7Oui{CM{2R`MrD3pSThSUB-3;n~8Hx)%ute7m{d#-h120!f4K
zeNroaR(4ryZ%L;TH@{jtN!zq;A^v<%%<>F{H_drUe*JCp@C03lP>(0KX0^LJP7^lA
zuI<iAWxsD56ppa{foCvajzW3!Y$z3t$tg^3=u~0vxQ3P$zRPbP>(D(s3(akkZde=|
zJPD;9q03dO3gjcI{w3)@=IDtN#0Fk*e+^b@EB%?Ow!P^=XDiVUBWJ=`pnwgjNjs-=
z#+Z10T{bF@hIL?@tGeCdG7b-t`K2=TkzU{eO~17^{U8#WXq&3Eib4$f)%JS@1A(77
z|0M7U%nKn^Obf;HBY?VMa`erE(DI|S-wUr;=C`_qnPh~Z;Fpn%LJjSqAY`Xj{eDM)
zj@zJor)QR6@z@~O=-m2E1YToG-~+ua`#?2af5yO*<Nj}B<CeA*)AKA{ec9$aw8c{t
z$E<G(4BjR3ixmIN!i!Q@VsL(DB0o=T<pHs;dA?uqlsp>+dy1IE6mQjPrsyyExY1$z
zsRwP*X`-zB>|fIq`#83Lsl_X($LZtH&hB7_d_wAD%Bz6^f%$Q#?(Y5l{9df416GO<
z(q3$@30%_MUPv3vK{b0P|1*q=W(Hsf(zI6HO>d0!L{_Mw-IH^Au<X8D7xpW@-$yt5
zVuaW1W7#Cob%?Kk<08e4)fA?}{VdI0gcdsP`-FX&N^@2)GUN5Imyowcvw!(}`LJpp
zr$aS=X<AnJ`*|j#`GE}Em>j~n2ej#FcTT95ZhkUaUkjl-jJ>A7sWH$9OUVBH*ey)6
z569rH2Z~OZzPL3$eom!**B#`rUWrYe6^?$}N8L6cM785V{LRqJFhoD%oqF~&{)VTt
zG|^2v6KF}wJ+x)yHayx4%AK>4Ebu*GMj&ps?WoKh2<9*jS605e#2Ttpkdb2Zy-MOq
zWUNN-=JKMN%Fc@Z+v<qH+^x&JAbkO&tMJR3v%J<~SyJNGNKF8J!Hj*4$N=xR0tA#L
zWENbLKnty53FnX}RZPlfQA~;_At8BAJVNk5bW&5V+f(eVmof0zu-A!j^hRvROP$x9
zi*&W%Fhj1JS7>0Z8w!!9i$kEoA@)yzkF`kk@>!xV!OLc-Mr8d0SJO#i@~Ve%>i!E~
z#9p)wRg34<T?ndiN2dzX%&sK>+;SDwG_awHlBl7|e}m)%m47Y^wN#5POxFpL5UpMZ
z{!w!@3LD*8FLk`9iP<(surWjUDz}aFGus<`4;a;5P1_>$mS-#|_%gA|DU&@tsol%U
zy*A@HQetVmHz{7TYDDybql*tp#lGyXimb1^$zK?l-CqMA;RXT4eS0x@fe<YZI=1d&
z>PbY4+aW!CNJ`%dzD!vU-P?(^(ke%1XhKD|hiNk6%pRJz+P(PTSH4xtU1a7sL3@!V
z3$6+*B<q5yH>t*A&;benhT0>Sw9&ytj=A^7*-g$;5?v4@ZBEp#>;qP1>@#7L{=A1w
z?;&kxzkQ5vA$vW42|f+0fApkSx2UIe&IO)XX|`$3fGGv*n^%mFr7yEIzR#<FESVSl
z;btM#@BI65g7IgJa8DJcgn0nAo1nPkoWnXt)9;%k0K2>l-)hTnBl%0PUMp?{m9w2f
zn$0m_J=+VFxB%_mdA|_wf<I0~@>smCdjQqGLiC#NJk5evlx6oq)61sfhQb1l)Rjuu
zrK;fs0NcN=`Q4Q`)ni=Bk6c<=B-0{&!x6%JhAr~FkzAE*oEE(xmjQQIKNUaJRPOYf
zj5`cu;AtSt%Rn+@K6T}M`}ya~SHXi(CAuGS7EPl1wERhVOC669YMIFGm~?<o<Cn;;
zgcp~eS^4#VL;B1`%m5Xa$Bm#I)#uRA`wU#S4#NtygZWp13zmJqOWDmKFK0T(=ldW=
zv%jIewvy2f_5n_ul*d(_`s<QS>Li9(rY@~lLl4e`V^tg)UTP<FIK};XTU&LJB}<fU
ziC!plOP4;c%<&g2tcEEM7>ckvk&CmL&+Wz+?>BcFn|;j;h;7Lpvnq^dXx*R)&?G8Q
zAAM=oU5G+*P2k>b_1FfAogi&-%(355QCEhiZA$SLMK#*oCaxS)fs?#=5@&qsk|*{A
z55fMKUeLve^otuRu}1f`6~}+Dox7kxDtWjmtoy8|ivrM@jOp^<gPzYaB#Y%qTEWm|
z660oc7R>`X9$$Jf6<#k)ILcfC|BR<s3${6~P1v01YTczQn-reAcjBwjo5;I$XXGzI
zTIsKj4sD+~$(lMZsR*1pO=N|eSCbwPpD%CZHAQLm_J{BzRHy!?)8p}7S7NJ-Coz5T
z4Y8#gG;(M0m@V$S7Y9h6n{XTpKEFJ0=hjU8F5F6Ztv&a_az{&KG}okbFk6o&|J#$s
z4Wm=B5xY$<Q%jv>R}Pql@sol&<bw&6l`Ovy+g{n;lXA1Zp319;O};7M%;c;Baqjzp
z=Hb5tL|Tkj1ts&DQ@_p6iH0B|ZR}Wa@xj|13x|4VG50=lOMeIW7N&oPr+hG@2vyZR
zzN=Bz?7VptIlb9K8#G;Idrtvob<E`--l^_2+tCapyslG6LZZJq#ns?+Pjs*7Y{|2G
z97|xOX5_hpf@O*}c3l2yz%XZy;kzlXx&x}q?MpX?k_tOKU%3q&OG*pX2-9Jq<I6~T
zTjsKkt#9@-XWK3UX7e`Vk((2P7l&|t{q@Po2771RMaG2V;Da>7pI@!$$CXMyr&iJm
z5hG&B5M`;@gC>}+cz~UK((dV8dCsu1bNmZHI3){hx)g2Q0Ia|El&H2u&^wZJ{BZ+_
zH&&xD<_f!s;w}_+fKe(?R^Q!;{HXw_9Va49m##AFt12y{Kx?voOO%9+?=~ofuE4&Z
z(Z}IyA%h@nv|=Bh0?KUcS>t&+Y>E>7uooAhy@!Ooz}Hdp^B;b{hCC;(@gDzRvyPMV
z<)!~ikmCD`_-o9EzXUPIR>yVBe+i;A@0=Dd-K@L)v4bu`8b5q-qMWMffHn3d{W1_f
zO)&%{GAP^oIUp<yl+)rWgDc$%nYYSiG-cIKO)!CyGQ3zroo{L_b@jc*vdbocd-bQg
z0qVYiFqx_O7RNw!t`>SWfeaNa>Zxi1PiB`Rnw_k)FW*LBk4^ICE1Zu%POUMwXKnUv
zii@X?Q)*q70|?V3@##e5SdQwJzIj*}SGz|ub-9Iz{sf}&t3bXVH-F!K&QxNtkZIs#
ziLwUxR{Koflqsti|HF_e{N~M_ZFT=GxZAf%hDHOusJ?93piD;<FfwR{k$lh%S3Y%e
zL1UHc8^fM+=6gmHl_iWJvykE*b5>z(3#_Q#jpXWE^IP(9kc8EVZma7?AIB2bADV%`
zJak}ihz%wt<kgsdNPKg-eg(EU_BO2ZMzlON(l+v%Z0YtQQ>`OL`n2}8+&6@H5=z2M
z&a*L&i3a<x8r9{b)Q1r`hMkQ0H`8(d*1-SYy!PLa*6*h6^DtKF)7@+e#x5DT--XpH
z|0S?4)4VxW|HG~LkI9e!+@pE`IuC*^1DfYo87fDdyqSW33?DY@(?_eqBFu5><vz~~
zVhyfS*4nG02KEMl{_99+?c~_|e2DH@tl<s0+}oIm6VKnSs=wO|e^0r&I1fshOra%D
zp>h`3?}Fu{r4WjB99m0Mi~cHU!Ul0;;iZWWdhsA3kp`T1^4~ut+-a!t+70p`elT(E
ztWFXeKzh29Gz3UR8|quU02kO4$Jj^)s37w$pG;tN#`NpAt=v&LwP;tS2VbP?1!f`&
zb>P&+oG?wAu|&OSP9bJtAqF#upXbIlFvLqlv(dw&>S(LZpv%y#knLt*RB?iYL|DTZ
zD~$i3>kOz~muaZB8B9w>L4kGnjIyw_oCt7Q*hZbSM+c7hE9O6E*#G$Im9z6NMzGHZ
zQsQ&F8FpPfoWFJ!av%wn%7Ovj&gr49DXWy5#hQ`HTw&oFXq#m71ZP@xGc{9D#k_d`
zz9xK4zdt^A8`WU>kLm9Lw`GVQ1I3msgYr^C1$asT+>iCtAmnOMb1QByd<EN|3N_qz
z_MdLhY>L;I!l%1d(BPvXLx?Wxnov`2`zE<=R+I0)YSI5`5k0WC=#6Bkbs5C#a`hHC
zZEdB6mxSC9v*G@e(Eb0|4g{$!jURsbWv!wA60~y>IE>&=W$d2}iF1O1p^y0>z;RVI
z_5OU0-msQL0mcq(q07h4lGwp4Yu>`=%fz#K&#;FdZ(je){r|^Mg_Hv%=j`BF@*&+1
zF)^AP-FE@`qdedgp}T4^%#DS)jMwsOB8qt@3D~%r3!t{$lnFGwLWVVw!x$#}>#@`^
zwYkDct7tOhfQVkcpGz=9%T|vUJuGi)F)e869~rwQek-S&wPZ3n;8$B{R}anPX5`Z(
zw}iKpvFZCcJ>WrH>+LH(KcKtEa(J^$l7IDR>1X|8i3gfGlT}WWi1kbQ?D`iz)tgzY
z186b^Wcc)&07U1Kotq8|3`jG5ujF0>Fs7ddi12AB4XLU;@eM?Y+wIw0`6F*;y=RNI
zS<RKkqap3H7+eILro^+2zQ&1b!s*<=Jp9?prH5*wIY(0M;Aw}VQL=XHC{>Wc^|;R-
zxhwK7!6<$XCU(sh2OM+y4rg7voc(l|IKhTkH%0U%1vVrGjb_tChUItpMkc1|ktWLh
zGoHwgT74(mKp?C6-YMV_x=VZn(sdd<W)WW8D4cEDLgt!=)J0{#6ZD)}w~mi@k1A=A
zZAi%Bv!733<<1fodbxLisod5QE21Oy&NoyMC69#iHceTG4!&z&RZ&k+o|1Wu$!<?g
zN>CMcSTZeOif_R*)FUV7b&KWzR2Hq^I^9Z5sbAlkYwX5$Xj3vLx4hGrDZGhmvF9_E
zDy1f5yNbp3PHF|F*fAr^b$SI0cAD~TXSU)87!|*V!#f!C`-SAbwyrw<zXU&GD+pLS
z2Ah_W-Fz&nWaRyNVvroTNxzWSV5NRZUICf6QUg9TvgU&uHrq0<toY*NnNn}QfI!yd
z*YttIwXp7|il-{Zzv;)e)?+lr%JdJQeqZdx>2D34PrxS0a*0@bN2G2L5aWLp3gzQs
zR`TYoPZJ;Ch)@Xk;;kaBOYn<`Ht&FBv<~1EZ_g=mL5du)=Bo&YMBNDhU%xHjbkw4<
zdeElBMe*Xu_RW|6Om=&XI1K1{T?X{YV~t;@>+ie*LRzmxBK>@jEn_!kU2cH^dXRM+
zs`8B0n#6<UZ6f6AwCCgJ{A{b<fW7JmL94Vr<V6Q;8YV76{Rt~&^>Hx=Vq`ZdzJBk|
zX+v(wwsZ)896wJTc87)r&ZHO*4Db12H5|&;@R>aFz#wa<iIe6a<6O%W{_7T3x$7TN
z!31}9_#oU>Gy#yOzFTRAex|ZHL$8u**bKdrjIA|e@|jd&obQE)@adNcIjkOPVHZ@y
zUm5njcT)BkJuc5{owv!S@s!u?0{g<MtCk!vjzi>yQ0}N5&xtB;i$&sy-%R}YfL&9j
zmqn>WJZ3tmTHs3ZcgI3`#=&0#+oqto%sX*LI849DIc>>!zoUm)qV?zP)BDmi^KaW-
zZ4p;h<lBL1X(b;Uc6;1+7~tv&?Fw*X!y_xgji$!J?&7j4o`}iDNXG2aEUy4Rm?!|M
z1duT>b}*ftoyNdY;inTLsqgJ(sxt8a^U}tveWrr(9Ae~c$wg%g-fwbVN64=VBcz-P
zg6A5#fIXd_V=9sOwi7R)>ne+`wD+*=b$8yxoI-pNSJ4M)nB=GE^@DMKcJCZ%s^`Bt
zf*(n7&pj|KHv9D}y8hWLR@KGRRliQEBgOOxvPbR8<RR4BRqn<8MhGdbPYieOdLYa;
z+wWmyNaFm9o4y!<Jzmr6`#e`b&rF<{^@Py0{X{rCrW4?xyFM}T&ctY21UM|v(hTDn
zw-rCW=u+}3H?z6KBr>qEk~%eTj}d3-MCAG=<i26)*jt5@hI6TftLl;_<Om7rQhP8n
z^K~%ujS1}v{pt90Aa)=12UxlHdDDYz8w8>C0ZyPM;n9uo1>0%c-)`y3A)W<qFLhDX
zw6#Lwzv)dq9n^o8Cd?dHcs|TKCQ2nzo~@67B#y2LhO2Age$&jT%siQ5E|;<CEnlfy
z+EU9dt?8KyNGmLHVzo`tE;WArvTo4$c7bU7FTtP%p7r-zV#|#<n?J1_6{`7@%sT#M
z-GtB?HOi=*y?VRAa{sQ8%PyU)qDA@e_k;~s*9m;&hT|ibW0z+_Whu_fY6#WVp-p}h
zD3-rOK}aN*4R$tG@tA3%J0?FRS|j^f@vd5PN9>u>ve>OXH1Srd?1yCM@<xZlyKAKr
zqeQBOZUy6fJvWJDqeh1cooXIAZQ*F?G+7~FU8q&Ux-Lj}#->ijfL%yK>xg_7`GR|=
zl!ogHwLRd=)EV<Du)FpUvOSu$TjwF=JO4g)VM@7A06sf2G}hd~(e08`=MbxvojjGB
z!QIQGT^dinIdcbM$ZB6gRB3xZWVxShB~;uDK2v)wZoR$FNap5mSYl>2!N#W6D>48G
zg?Q~v4|iLBnxFEwC^5M|7$W8;M)!WG*of?VX}NRb3n;f#gWFuz^9<sA_B5B}=rVKV
zTGwK#^3@{>2jDz9eUY(!&x;Q-!xO6Woo>0Hf90Phj9Dy=h|dvfs>nSG<b{39UiFCF
z%e!`io$3AhP^W~^u>)Ao_d7Vgm`$-I9!wJlQdZ16pPwToR2BZ6m;9AI&D%2}=Xb(K
zFa32T3+NCbjsF*pqyhlW66o3N=ysHPK&ssf!VXr<ni9kwh#p*#)vDgcT})lqx;==$
z2rYGJJZ%f3w4XRW4ppn+rrz;;Up?w#QDoF&rQS7p`DDu-Umhkr{u1=x-ORbWX4WG$
z_|#R!y{Td{<1tRO9zwb3U7Y8)V4C-*ODkrz_78)8iHnWVQDx-yvBaQY59$af+1x(V
z=p6OonbH?J|2hi)Dpokpz!dLpLC67Kh#h0+qGHPBh*l3WE6yMFYxqKQjat$4ykhu1
zQD$m?Rlld5rkeqpkx2$b*rt6EV-cV4NIE=~ZxBYpYIpOG^|y$ed^>h`R7LM}I>2b(
zjO@lr4Ad+tQ{)5S5)8T1x~A)*$uqLyTdh9H4Pbtno<oRbxvWLff<jp=Q37X2X*3Z_
zsDrBg){(+?^$%DL`X??lLC?-Ez@=ku7q8T{Gf7m`ByE?U#?j1DZWokyceSI7{cNy4
z@-IQ5%YNIRmw-nPXQd!=l^)QnW=edX(Ll4?!;WxNQ3~Q@p7pcj^b4Bih~l!d_r;_g
z!W;>@v&vRR0|VL+ji$f{cMpUV|8-L84QsylTGmk3St10ud|O8gePo_d?ua}hPgr%8
zM)F(=nVY8UBx|Tx=Lr8L=p7^b^_HkNu%CaA8JczdXU_Qga-4up+OK51SQ=XQ3Zfd%
z{|~O}U&~4I*I!`(_AKkdg0F>X9()M~R^`lh7PZrIz3S>2cO{@j$q8LvLAgS|$)a3@
zO-#LHXX&Aia!h$Yw6g!D5%+(@^uONmg2Y^UX$y|rr0<&<*Ie59Qn#i%PI{%-VDmd&
zJRMV#YJ}E}d;Q4@G*uP!%;X1Ch_hQ2)ppnKYKn924+F*E)fb5B@+q&WixQf`6x&Bx
zruqzN!9}QWwOuwtQck4uic|7m0<jO~Il3fgEFabXhao>g(uswaG)lQmJtG;3i`X?$
znv4nq)m{!>u9XfC`uyNmM*2_wak2DPW68g8htbN5&e$sta7A65j2*HB2fY3lt9#g&
z1nT>gg}?&yXEGkd{YcDl_iz@Y1eIarhm_S5mc>qMj-aQKL<XNX0fy##Q5(^ln-(IJ
zJ1L}tXKR80-cZ$S0s;{;eEW1vCb~p<X8bZLXWIz95%tR#V{RiYHXq69K+kWg&FOBN
z@UGk44)25ZOb5Kx4o-1&efuvD@b6AU@5Mu4Nk3}so2Y<NA95t#yUG#d_8f5<dtO)o
zG}I}pE{o1em2FsNXZQtHS-6Y6T*PY{bGBY4SOky8st}Lj8iSand9i^yW6f%~D+gk|
z@zV^U40fFnG~G4QaPtDaWZq<?C2Ef})b&Tn*pqkR5sI0J0*9t22d*?CH}HzZ5qA-3
z%vCLqtyOxYgts#eB?MKP1xIkB4Ru924}~Wd0uy)C?IIEW$dFn6N9eLob|FSNz!^on
zVwi#&P?aTLR4aiz=M^h8bvy~9LnbhU)r9(@9$@@}t9_ua6SXFYPe3honwUW>r=_Jv
z6ojK*`K5+0J1nM^&J~;@jdR$t%v|Ps3j#1oo=Uo8vNJKhw?bzMtM31<x&Hf|0NVG!
z(<ErHBXF-G)!<2!MxlmV*Qx>f{H<4iu^!gm>KPIdLK$i>Y|Pu`E<YCZO52ss&U)~D
zG%RlAe!aOSW<yETBNC$gaxf2N?0Q)GzF*Uy_Zw3R>rXNIdV7li!d`1F_HVu6ooA`f
z&^ILI(sI|ssf(YF+q6Dzb*PoQ&qZnqaPMAbh@U}W;W{&GNvbMFh^Wu@o<4g{;dnWg
zs&GM<D%G7AqnqwqjC}9-f0G<`Nm-sUJDZVBX}XR{y-k`8{-V&4*|S^t8Z^Y?h1>Xp
zezxNMEJmYk(%5tW%+fZ=33h8OX1R-X%bb%f+(+8^+m>E-L5X%``ar9tv(vB3ZOVlQ
zCvj4woUh0|L=hrFpIK=>m<JVDWV3wy@6Gi;*?!37XZ*sMR&w6Rr<P+D0H4AQO-5a*
z>%-d;$7O(Xvr(v{Z{GdoE)Aq9)EnI6N48|E+d9iSF4TFL8|Jyc1i6u3d$9!W(t$0@
zE(`rw*tHZkP$8qm>CJfAhGAsUmbiZzjWcKhX(Vk2%oRcctdO$hs3Js8Sa?_z)29_s
zMn==-tjMNB?ZH%;)o3fy$ReUlX+S~@w*Hp@6{(A-!MH#oBpY+sWBfwKh{dRz3GS3j
zDDRv|-Je()mEIAE{R+|=H!z><?TbUqfo5g;^InPgHoTfMbXMJT7s8o4g>ns^Q;xml
zvKi0IwXPk`nTj^Uwfi2X*g?6k1{{GCXbVQ&9vpn4XkU}-HeO4{kD6M_Khkqs!o(2Q
zm$ODKsnnvyGpaOuT-2F^fRtl~7O%|I=k#j<7Ru4Wdg+0INWg}oNqcWDn~M^;*&w1|
zuqocKp}?^Z)nh&=2y($-l)GwITt#tZ;-NsV$6keIdPcbDGL}5?*)b6O8(5c<OBG>|
z$jmGxSdqdzSMA<p>{vd5Gn{Z)Yf73>8t~yjfy*N{QLj+kIYkIaK~Y38ijD1+wVz^_
zE``3JAwW?RpxD{j*&`?@IO`X4C^&0-c-!^(fAW+6KZoRNp-nN&hSV|ohS7Upy;E<z
zZfdyH`4y5~&VvlV!@~bK0MiuV=NfjR2bwH+yowN?@?8S0Wxp24;A>m8LOf)QOgFFh
zD@7DwHb8Nq=~cI|{>z%Mpb-c_E#g=kjqB6o<ywC&e2;hekl))*IcpX%){$TWU1Jb)
z{H(6-^%$mHozlnhR#;f~j6Fs8H$nQx|L<r1Z_glmQKx-PEPP$t_DAXUvrB@z7r>D*
z%AFE*s3@FwiZVx@R{qgAx(p(urk507tfm{!5{AKqef0QHPG{^R*4*NTsr~RX@?N(W
zO^5cHj8eH7Zk%n6;nS8<ap|;o?$e;!30jH1SJWJ{Y+CA~PgvVbDf*M2oQ_;i{3Hep
zfE~8ApIdSv>&DF-y*6x<Fr5J1!B()PAVtk-ExmC%8DutA?Pu{@&KI#IMb?i1ElUCn
zqEbK3)m;vI8>1_=IGVaT)&)6_dW6$m-CcIhvYKoy^TpW6!;-zHOvcC2FE1PRs>hOk
z9y|9Dx!(bIE#K|r|FQs$wx%=MLiyaSF^}1ca$;C4NEp$C3_pVlo>UqyRuF#d$Y{sr
z<<X@bBE+nOy?TkXf~v}OJS#)j{?ukKr!IH}uTn@@(x#41tWMWzDo8|fnO(o%CldvT
zo*I!Dt@_3ByzB*Om_94hY{Z7)X;K#->wo+^4yZ@d%M~4RD)lHRH#ZJcB%AKQ3{rEy
zp_wp2_f|KI1?iTCa{)v(DM|Xic%OLWVA2#)ZJm%LO}j4rPJ#u7^C|^RDB^_o7xj(@
zbzec>0lF0k*R*%CF#ERw>gRWZxwiHMFb-IC&J#a2#CC?+^R!6sQRQeanq=E?i8Jxs
zsI{*9(w*N}Xk)+n6%2{i^O2tsTeO%t)dzKVa$N@l#-K!l$7%&>YG!7oYZuB$pe0Zv
zVN8{Hu+M&BfYl?e1EUUH<zOD#zj}XdEXIXtQt=w_!zB+=g`1J2y75+<hse<TH`X%C
zm!cvN=FE@sB>4IhU^y#)f}XHwRwIfy{O$4|j_G4|+@5JaO_$w((I%Y8Bpm1)3D$Bf
ze$qM?BbN&L!Oq1S9g;a}QB%jX2&%*8w)^N*%?`oU?M<Ijbmn97qFv)bZz?SzQfCoj
z_{g74{QNVtZiG(aB0dX;#9A%LS>nfA`|g}Z-096i<o@a_x;h`*ed<eV(YXe~#o=NE
zp%EZSs497$y=23;{h_o(K{Aq<&j!fSDNUirDO2LN%!D%Y@Co}11FkQ_eKxi=MoaCn
zgtz<wJ4?iA4x`@3ej4)+3~j<L)4tJ+M$cP;SFZ~hg9L~sJ18T4BJ6x?oy;)~j?T`T
z{2aV_KjgU%q|l9xwL2ZYb?l8=d<%||R1-nE<4+)T0T3kJ82!6BKKrNIg64^ja3N2v
z3mRP$LHs#N(G#s!FoYAxN|pTZnPmAeTXR)$MSX_DlQya@yLF@euhIW~3<%>JR}uR5
zWPEOa&#E=TmL{4uZRGmvpza8DiXL^T%_Z!Ip*J+Ys)FT(uI@wKB7LfFZ<FisqhXa8
zHqE*1OUK0Z9%4jD@A+vb@h7cKuZJVv+`^j2J!zq#JGJwGtH*88G3(EZ=Tn5fmex@l
zU>sO0D)YSi?F;CWS9}chk<ku3IVLl<FDB8GKdb2Un@23qd|TfAkzJ_5&=QPgum!`%
z?fAg&qkatVl%a*?tcj;qhE^x*YR^Se1R;yxgM5crH6JuGW{%;qK#^A_<U$aIvJXO$
z?fkF+$)Ca*zg$|PaN0s;4xQP{v;1|XF;h$>X`kQeJI^-v9uF%1U~3%oBJx0}m1jps
z8=4rSlWbBkW;yj55`KXQ)URlKTlxR(=ej*6ZTqg^-L~bDbs<ssQc8({;7%_;LhV@Q
z{Q5PfuB6%@@p-<LGpeUxd{5Z#Pg3uk3wullh);|wXTkOm@(nV?XvjDu*<~q3Lunr=
ztRLG3Gv$7f9cxE-XOnWTKEf?^(tN)}8U55$G_s_OH!48C=A5}3_|ZM9#CP>aWF}a)
z&~|cqKD~cU@@SjQYI)4QJyI5I*au<Ft4g3g_LcS%M;&Bov2E_R?U&!WB$*2hP%~t0
z?X!N|;-xh2rxvvePirgq(+%)S(I3)HN=qG5g4#!v1E~`xX!&|#p!ZVUsE#FOw_0wj
z+r98SMM(=Jn5ggj#<qt+_aHdmW`=EzRV8Vz-<h8{zaAok^>>9h=NQ>$p;}NoJbr55
zI@r}}>--O15Tk^lBG|Xq1xRQJQk}PImNWzje$2v@Cz`;*RP!?YvHaVFB>W2^tbyP{
z?eOea_VV!iyvul5=U#`VyUg3^ZofDakfC=y)Fb6g${+rL^@>*Gt_~+;#v<Q{Le*w9
z^^cKfJ>N=eXnKB5uZu~)*&2i@*U1v%5Xm=Dqn08Mt}f%Aq;n<9aiDE?kM{df<`JaO
z31&)!PQDIJ9tTmCrbjo1?Im&b2v1)Gj5_t9ZY$lEeu6k03wS>sa;-Eq<D^mrJoYd<
z`$4jm>2>1Y!y5M_#-iRX#zmY9ar^4vU1yQ%hqZ<x+mKh479>-Rf5g|6hO{AJMsH3d
zmZUmDwZ%<$+q6Z|Urnl$>|l&Y<PK%dE7EroMVBp!mG>o#-p;vvwun}LAme4k8P=ZQ
zq7@u!n<QA8r%|K0H5R;HvR?9+!1bZmhxPKg4}a)R>K~3+VV2&Hx%_MkDBZdB+<y1S
z`WX6`;Cq%s)|NV+{>E7~LZiM;_o0fPr0tm7IqPuV<&O15@xnvgm*{#-PuTvnn{=V^
zNutpV3R&(e`&8El4x)LJX`m0Vaz$R!l!3atmD50iPPDx)<vXU{qMXUSy_T(zs?5m(
zz0WfBPJSr*Ew@wLEjRk=h8xS<w`GY?un<jYN}sixPR_Qb+LB@2`~zDvC^wOB?gCo`
zpnBfTuBg@JR`5$uHCwO%{kK1k%d(~$pSSg7nRqJPMdqL%H2<TlZT9W*Ze6;R0LI;n
zc5aQFA-r^#9(Enfi8*>)YBRxuo)dr+gSWGJmi4;hUoH3lkPT)4O?`d;q2r-#vW{0@
z{fHQG16^&K(Ww+gl<tBN!brJiS^wRom9+!K1<Bq4td|Sj&}Y}>QUB5|O{wL+N3y^c
zq6tmpEIoTWiI&Oe-wW{bn(i7bh?{qODizpU718f^wDtXZ<?eX%xnxl|d`<=2f9Y@x
z%rG@GL5dP}doTHhD9$k_cvUdP1@kd5INtpbD0;#_)|Xw^2<nbkA3yh^*rGN{|Jg5W
z9^?ZlbmBvG^MXu-bt0~&PdS62RyvESUu8BHN>c6+b7s?9_|?wnZa`nyCfZ1cld$gJ
z1;DA#lGLDojzoIr?bpXGA{3&_xaA|ikLAb_5IP<c%1!OFeOQgo{!7rh|L)<4#=~x@
z=wTvW=OtVlA*`obT4<Xu24YzRnvJZR*NHT=OJ@}NQT=plMoE(%B){_bV`<O*aL^<G
zVcWwbYM2b~u?yQDG1CRk+?xWoA^8_qHJ^0WfMjPxobZ&J2uM0j7I1A~a6>ND5~(p}
zU^(MBct@GL`Dp1l`jxFnT97?!tMXOM>#lEa66Tq$hpb5E?<~f|Ipq;*EHj1K|7mS?
z!BTm6kS+{ql{M!q)Cu=j0c}6LEG#gFEg9=>*e1lSXUNeM3_s3y4=D%9y`IgP=pin0
zO-q~9a4@OohY}PyV@0uQsN)veK_=?FRb9Jjt!%r<NK>A;{CqBWD4q;@PiV_Cru?YI
zg|&{nHja^ZpHr?D+G%wo<%6%je%>j1Ejx<Sy@^CR-~6iaxwTmND9hk!Ri%tPKPbOt
z5W0InigygIceX(f+fIx}wf}VI5bSzja&E^&XZ{ilM|XV~E_;2a&lO2KJ6X7OPO?ou
z^0u4LZgahDb0K5GU?aYG<C5l6QDYG43R&|H=kmUM<GO@GhWu~+Ir6ZxETpEJQ`o?o
zp<S&4uwKP4yX<05wIyYjo4og=oo{NZPOxy|1u1XMHetax<*^&hDGpQ0BI|hf;=~Cn
z2Gk6uf*(#f(f`mr!hTLXqVzM1FTnXHap}J#Pye4W{ja#UoMiWWz_|h3qsT$G$(>^S
zpu9oY1bkuw7Wli@Ju1TAQWE1qU&c<e-y|`wFt8=L>#LJ1{3<A_D9%=n3IF~{jM07$
zb^OY#D>~mn#x!tZLw?aK9e!cDpwBf_MT=r-Mn-+0M-`wDMK}6mGd7|sG_)nKkhMVt
zdnRjUI<fj6TShlkPg_`8f9{>|!+blXxM?kT+A0m3&49C^X*Odnf`NNb?qkbRv%QNQ
zZChP4=zW&s-dMh;-09X^e_~+P9-;e7-dp+S{yqwETkGYWSzsjSOyYjDMKU#BebB;`
z3mW99HZ!{07<Qb(t0}#rZd#?0yPY!ja^#Xh+AM1_BcO}Jd!s6)inn=HGULr><#D<M
zQCGfgytwfty-*do$*770ZBN*2T*GF>74==+G?oHIG&Qdx?T5~QXg1;M`L`9g<WTUJ
z=%4#qKfMNrx~NmE$r!s*+`g2fD1zYv4yFrMLkiK7^Y?9R%8G_+t_L2qIPo@;q%|?{
z&@Lcr<~^~5o4?O6Ck`}|gKU`Jba;`{X=+15UQ2DbzUtRlK_AUD08?%`%*k{b6Ps8k
zE2#rlEJ$C;r)O%-1XZ84l(v14{!4JvL3ihmOJa;nnR1KCo~yN#XXAWpx$QAeiBO4e
zxj8c#3SATj+vVPJDtf6_*ROkLGRtM?*M`nmm#L7BgW1WV#=B|(J_(xnJeul|En`wb
z#3B?y>BlAVL{q)U`*`HUPiPH5QS2?!K<7Z|sXOQQxC&j;$CFvX=PGA0*Xy_bSyLQ-
z5He+Rr61q$Oqeidd?OKVoupF?<y#v!)clmJs;^`>k>P1Nlw~mdX8rNA;<|99K}Cyf
zrV735QOZJl=Vb$&*lYqin9Y^&TiXOTcp<XlFx`r|^2=O_G%$gIm$I?Fs}Wv|k9^FM
zed^?ssPQ9CQ*tR(IlvZA9yJv`&Xp`Hs)vXii;v4lbqif*J$A+mr6SGxCmq`Kl<csg
zgz4^nECnD@Jt|!!U$K+(oQ}h5hWq&$L9F&)yJ~8=x3|BkRaRzjcA^>RIGoEBi}~ww
zS-usRLS4jjWfhasKCTBFieGFe=k46zkNdqgtJJ+7M96zq2%6Mb6I!0-6Ny=SZ>Y>3
z*<s^hV0X^9USHE1vO12BVV|7Yl>X%m8vfJl6qC2lPCeVY@BOCO%zu2YFX@uHxi~lh
zw=%}4?$Ue`c^QBZH=>$g66&AR#Co6qp4}MoYqw@)eUnrl3yT<Q39~blREYeQJbPN-
zbhcyn6HCMWg41c+8d#qsAMmP|3e}p>c~K<nT@_P>=xzs-O63l19-w*JWy<SU4)2YM
zW40fx`{G+^pePLTSz=sQeEJF**G|4<@?aOv&VbUh7VFRE6obXs!dAIUpzh49L9mCf
zaXRzfajyKzl{YS<PAR2_PG}h-bM6OHF$WAyAz-Cw!c{WhRdV;!!V#-CE`m(eiebYt
z^pkTsZ%8h38n2xUeAsgvLU1nprw=s&Vq91#Mfb_<cK@tP@yv=}em5Jq)w~fP7L_pl
ziDC3S$8Yz5wS|&Cut&AkI_79WhIS1%aR$I+SVLJeQ^AymPBbEq2jUScexy#5c~ZQF
z-VHKNskaKQH2=Wa08W*#?n3cJgN8UW;Nm+;K<Bww?nN?Q+T2(_Q6z)58jBsTP$boz
zXn61`l}|l`!PAbjOBf2g65`r0CMjEo=E~e_)GCY*m79=;`4l#2L~>_T5;3GmR_0R$
zRsB-K2&qc2evQu6$V=CEij4vgWk!rSv1uz-DAg)3)f#+l97_iXeT*YaJCqrRV%}Z;
zi8^qgLFdzq*6q4O=NGc1$1l*-p53;RyL3LZE!BC2nceC#-3EyTaIK5)V{ymAY<bCV
ziQb`iK1gm9(OH~Wxlhu{4?Cjy)oqr?r@N>3_?^iRiwco%K5z*==@Pf{dgDLe9AST~
zDHv_STD1CBZBcDVUZGes2Sz;i0G}i8-!xGcw|-@#9KY3-=^ERHv^$&pee8Go+nKc-
z{#6sZE53wOdi*r#j6QQoWGhQAMlttMea^9(9=|7E9T9}1>h)1_K#5K(Kh2{mSMh_X
zaFbNz=Vss$R!rZW7>}1FjU60>GKnOtC@x+IakEIz!vjH}wr65;&^Ze<_ZYImENhW(
zvD8!fLA?FANw=I*$jS`Nbc0mr$@!XMG*LB=p99I{TFu;xGM3qDy|3b0s_smlI@wsg
zm9L%nz=u+R*_%HgAkI++Jdj$Nu*{ZgSCu#R1AtQO5qiQwE;3@Fe=lWTox+D?DPBCO
zH+@$1^j@`F?{ZLKv+{S1WOF(rBLt*1vxQXjL!o}@sB>5t!sQj~AYpvZC${#qb5nQ8
z38C6&MIO&rjS*IF_<00k@?om|ZtuDOd>~F1aU!YA?w$LYTVb$jWdcZ4aOLeV>04w2
zkDAmwak$DBMYHUZrU3aW?cmgT4_lYPUEqCsV0B%f9#&(!n&+Y6g1eDNLYp7h!Ygz{
zu_f9R4AwTa(uW|^C{7RLKw(eRU53nEmhGWsCygIR(|cmrPSDiI4CR?ABJ)KZ(H;3~
zbgrQ=aBj@zND_5QjX^^wIf(!+;(gzUij(-JSYSM?LV5Y22CJ0v%HozVgP-EjE2dEY
z-#sVQCMW+Fb8j8g*5AK<Qu$Va7I$c&K=ES1p%g0;C=y(XlR$9SwiI`FC>9`r;0Xk4
zi+dp<I23nUyl6M~v(G*=d++bweRg+dXU|MBb3Swa$kDvd=Y74dYonPa`~8zyj&+K(
zqkO+ewC_nuk|V7DQw6UKwGDWp*4AK9zOKFP5%la%Cclu>Dn6*o`405y5wrqA9a>*R
z7pyM#v{zW&NONWzH%;gLEHvY`#0u50fyiJ}3)PSL$hCti1LVsKjv4wuN!@8Va_F#g
zs7djbGu-tESA|+mu%^jC;3~`bnw#MSyu+1(JaI{L{0bPFWT&&BtC?5J#Wc_!>w{5t
zcpLx?bo9(|XILEqm}r$Wqr^CJo{b&^I=<W8Mc8!q?RjnkQys%bdueSc-2PBAdk*^N
zZE%bxqb6vpo`YJ^+3z`;{zJ`mUzkzLS9IfYctL)LTCS=o_{kl)o&ybUKE574FGr!~
zxi{%Dx`#GuVU~0mv)K(^!+`Opu%wg6ItdN_)tdzuir&@51)=_0VHefmEukKEU`$U(
z#Uy-dW<4)7&O6TfT{R#^Z)J7ayalMwRb7OEHAllF2!8(ml04oM{ohF*rT!&(y#ANu
zF&hdbNFKS$qSqxNBiGqRSH{ETj4}T7Y7tE8`yf4ieAW6mSYa@fBoprSP&+$0gL15j
zOIVndpm`KrLzQJ{1{3ln2@yeMVqyaPIbIrsc;x=RX8)IP>xN-L3ymQ+3c(@a|1-g3
z!vCb;F;4T<U!umm)+Y<Ki+@gDk~}#Q#j*-^UMxKsh#j~*w^teSI?XX)L-q#v^JudO
zJU=}B;Q1`-x2g7X3M)Q{tm>N=b)_$_J|`Px71Su+zJo6A!7si&xwr)9sLnFmH$#29
zcbQ-V4^N%94+c!aXLo3BqhLkw&gTPj_8H2GpR1Y6MVe-7_;!P<8p7MaznJsBftBkm
z{f)X5>{3v_bn}K{sP9>ukAU8+@c_tYqNwq`j6(w!`IJd{ezF0EuUo-?_{dlN7Pg-@
zJbNJ!|Dq4i$bBXtOR?7<0muQ`b|sES0p;kGD}E-tI&$Q<DUqBRZ+O{4!l8VJ`uBUB
zsmj|X<)7IF`7@0w;Xe~a894g(%-Q(_xwos<`V$)fpTt;#zRL^KOR4JCy=cODM`?;D
zzrsl}iw7*uRrTlw)Gja?7#Eh%rgf(MTJxi9O&`Jcqv#cO;x7cfS;xlaRc&pKKuNoY
zKD4&L3Wy&><QY3>gv!&(`T(*<24-Ad!#4lCsIcm~LTHAbP3&+^^gxn?CVS>F8bzh&
zk!IH~QpSlJL?|RNn5a9OHVDAvtEBc<fGX%bJIx8ew_+cuvbkB0EJRxmS3ylo+%mx3
zvR*JcN&ttSo*~%9(xotYaKY(gs2lHH>if9bzj95d?^$F7&O!${x^k9jY8W$=j%WGv
zGocU;w#~HC39phc^46t(x;*;5K43MnHiG6ws9pHe<-$pg#J7<(TWlmlf>rWewUs+0
zO<{>veqmJRs&Vs^(os*LU}wvESxlfoqFq`&2hW~(%El1JjEY=O`8EH$|8U>_+c0C$
zf@}0hjoHQqLtaW{W?GmuQvl9PNB?Dyu<LM6V%6d0rT4b(`_&UAE)DeiaD!)4+o>pA
zkt(|?>q(ApaMEyD^EI7&>)raL^yO~cp)}8UpmbcPj~Ukay>1$8Gm{?VpSq+Qp{cks
zUrkDg*o$ee%jSV)A?-Nq%|#aPo-i_<xcGpOOLMi#Kl|{(5t{55{KGBNKJHe%RXJ_{
ztBN+;-_9Bl&z_8CZyjcz>5k_ysCefg;qF1J@3EagRaVsiCf4g9moCa|K1+LK+v#r2
zF<_#wWvpsf`fOaWO8(gU`1OIoxR2I6eL;j#iyRFL*lw#yvcx&aLL19J83L-!NQx6r
z{ke&$5`%Wr`dIDGA&0PfW0Y7^kg%r8Li-EobyUFZo<@6}X7ggm^_<_=p4#BdGUgUf
z)06rD9%gX9GZstPtt~hz%;<x6*=*gj-sP{pd1&1+C|~=R=pt%Sb@4jZ7x<T`MZe=O
z(GUOrzeHG)TDg;1w~&3gCx3~;?T^j9Yfotm2nI>Di)&v(wmE}1Lrw*k#BTPC{}TNr
zlHa(=ghmkpmu3kn7PmhkSHcR+7mY^`=E5-oI<Fp_9XLFB(p!$HOeR*_;on2Jce{h;
zepf<iIY@+NwsR%SMCk-dW+Z6Trv%=}-A0!c_kPm*<JTc1!kFhT=G}qxpKr-YD+ezs
z;P${cYwU6Cyd1D?XW7m9Vn#L&s{v#Z<5~B`k5Zo5WwoAZ>_{rnc^`0Uc6M(K*E<jY
zGO=<2dTj5RcW5bI4EWpQJ!H-!eHnrlPCNF7rxFEut9<@gKh`oSJCEwa^gPT@Vu+hy
z*P*ZN)wdnj<_`C~M|~L8p9Fh>Y4m26@u&IoWb2^|kCfM*zlh9|e{`Gv)r_{a@?1N*
zESd+m1M8lbbJeDx#ER@>N}QF~Q@A|gB(=w@Q2#iP^WL2<!WR@s_(O%6jFpQgQ$n$$
zzO@u)Vl^94SsUah*^ugJ<d`=b%-+C7T2thSQA>D?P}+irMAAh4hO?1G+A>OLa!gCP
zKzMt3vJ6)L0?qO)L5aOrKb@7sJqA`yb`^ZhAS>xBcRcT6zY?O}!oz|$zqW4t?V^nl
zxq|p15C+pX^*oa8(X5c>5cv+I^_X*Z_1XxP@R%l+5Le1@n!kEth)Ekg^w6u04$>_2
z&idMBdMS9qpQB%#wagkjAjl8Yr8J4DU$F5*Gx(wA!CjKra`h_7nPk))6wLm-2cIHR
z^_U<()X^9($s!0M5~Q6?n(fxSkz-~E6aqd!=)(n%N#EDj!<tcz=_%<MC~1pK0153|
zJq-=TPT~KTom`I=cn4Y{CM3_;;_6oMEkHohL>7joa3%%;!X3CjY*OOUuiaW7@TN4{
zaUUSQW4omRII1j3;hkfDURt1f*R+}LLv+@hxe#&ay>HXwRTP?*m14j5IBhfMz!~~e
zr_4iAe!qFfuvMYVBY3ltm$63NVw!DD!~^mYTqj(~p28d+UH3u9lMb2pmIzJ2vg%>?
zd4;Snt?y-;t))?<6vV3nDoqZhJ!v*GrA29_XsVk=Cesuh8-b#HOE!A3ytlY(4lML-
zmie$=>(2ci_kPi&eACR6UlusUdUpmPCIx3%oL4^}aV4%$ll^WfJ4axoNUz-DJ^64K
zm&Y@qv?S8orlnVI5r=0;n6N#~=!^j$HOi6fp_}QxZ7Cl+PqLYn1`Z;hMRxDoet>7T
zOQRE&M7;NIMGCK(H>q$eo}3DrIP<(zQoHyh9M}^}0D|~EHgG*oL+9d4g{7Z86_%;m
z93@pmx$A8XOSg<v-1{ru25^H?(jKr&dOtt`L*DJ$trSAMbq%*ORZowbY1khAGn2pT
zN-||h?jEQnJ-{=gGW{`q=@conmiJF+Ts8rPG@}GoCpvgEbF#xIx8K=Z`Hm)My@kH|
zw?BdZwEFMl{;ko!M2u#^c#W{CR@dH{iR*BIYLoMU5^Fk5N6`5Gy!dRl#Nh;E7Ab1y
ziL&X%N|JF#c2KDdQw0d3?YA3N!k%ECqcRqk06o(3UM!@=1qR&z3a5{5zL%bd8K`^6
zauypZG*Fp{)ts|A=POR;-E(4AQ`KzOUT&l(_N_G}abMnQATA4%x{u#$vc#0B7oXoV
zBai-0ZITAeFbNlL^**Hf4i@J6t+QM1yH}i)qnHI$&U7nnd?eqMm>zeJ-elch+<u#r
zgsBFCq^rnYDa-CngtS+tgXb_RA#$A9O2Oz@2%1GvTuh&g$22S1q%>8nFD|>UcG`;X
zh24C*6W`IpE?kx9*T$IJgPtvVf@)4HTJ;znMn}GGp$86m+CHj|bbTp>>Xw1AQ9Wh`
zdY3#94SQ8`*4Cl$9Ism=A&>q1Mu~^79mVbiz3D&HXmQChm#}O$XUZ}M@s^DuuocN=
z!#EjwZ~rLnpgH627F9kKaMM_Jmb_0Qnv`4EH;I-sn{cxJ52Oas6OfORYBFxLu%BzY
z?h%jSP6Sxp$ATOW8(?dTS1K@*$$Gl3V&u&iMcg!g^+^6tsM6ja?ceR~D4yAT#!LEI
z)K13p>2XwPVZ0RI88Hs+^!SIMn9Wl8u$k0dRUz%;WE<oU2HR4MKrx0Pftsx-TML#g
z7ftc`z?$CVlhsT}0!8UKt7nB9H#~aWHA<b;!oxGB1i64+r{n#nBX!vq>7gJ+JzSd`
z@~VeVDq~dIgA0c2d+8JqZ4*FiHm?n(HwzqDt%)M{G~bNh&=%tL1A4O9mX`23DZD-w
z&_ZVSzv81FcVc6;j3P5??-JQ~(u2&3<6|-{-!s<LeUm#&ZgQ=u@G>yRk40x$FaIl@
z4bi>Q?fq{k!6xGJ`YNw!*u+VW7`qT;$V=VjLN|X+i$m@>-DXunnG5W~EhMdNqv0)h
zsHIxGS(a#@O~Qxe)xGVab`T`-O|VS>lCLj~&-r0Rdo6o+{_eVI93ZY$7Sj%pj`nlQ
z_m_ruO;U0ZV^%clQ22&&zE%?#|9?WwzYVzNJsG=_;;7_eH%YRZ0f((jHm(8KPL3Xt
zv;le5AB6D}H@_=M^&>e`R2As)fens4m(TcW$z(SRm_1{-M`gF<eO+Yy%ftkGgy}Hy
zEW3;Q1G=_XuM`*bfH0t}=r0_%7(D@yEuO5?%*RhuP+WCpZKg_n#|-wDmSeHW<OO;f
zDe9gK>NV2wlBIwZA<Bsx!5Dp$Qmt1zNpnJ3!`gVap^Jj@TaB*;@g61gz=~f6Z*f)p
z*eZo2&AD)h#fWRJBd>o|Ry5ds2I+lM_sumY6-8?rzQR?~4l~v*DTdyY-+dJaz*Y%=
z6hZ)hGxBJ`cRBo)WYG`0%MadC*q`tz?{B$ozW<uQVee5l$TWWB7*GL5VS?_mzo026
zT*bufl`Y`e^p$4{@HyrW4*7&Ux8ti{NIa1Ho0lXpX?-bHTw9p1udLY@kSzuhl>NAY
z@Sdz?l;>oG^H)!24}A=<j;Um;t)4vEQA}KO-ZO!r+1W^36aAd}L_F5yMbqHtl=z~w
zzeL_@3<^xf3~zK8;)#gvs0p{z&q<yrJP+W<<%K)V$_|tQ{d=5MGocwuhdeuU<CLX+
zDfUpuuz@_g)d))Gj`?4G>Z9&{-*qe0`T&l=V}$y#2W`~nE$W<~#s=Z9WnJBv#5xal
zS!h28Q5NT_5nEx9aKYoJpO5TVWlOoVKEAkpeJj$vt;2!e&rg7JL`3DShjh9>o0!z|
zf#X@?zG{vlM!UjJ#TG|@qm)%vUI1+joxocoLAvxzhh(gS{Jl3l2f2<UCDlD{#K8$v
zJ=-R3*{vZA6(3bpjR&h+*zt&M1Z~W124~fvEmBr#?{V~k*92M9GR|_#BS-|Brx8k8
zuk(tS^PQt(DdYCV)go2Vz)u2}ZNQ}aysQ1=`z|9j%U)UuVn9S(T2;MKp_*z$SSXk}
zWh*pY3k=Hyujk9SM^>4%gjyOdyjVlHrd}+G*<c<GoM<YFswJ0~r#izVOJt$sB$JZz
ziC@7tWv*va`i&;QKG9H{OZ5rH@Bxo>Zwb;6v>EGQ^@d^VR{PbKk55zVjw<;o2BMcS
zpN_UUo>Sj;<L6DKQ&9yBqV{fas=x8RETZSu>O5**9yfdW^>_vO!C7SMo6irge*`6n
z=Pe^Nh5Dv<=)ggwDGnmxF1K9P4D)&Ntnp3h&^YV!c&-43ru~ClST_23FOt<O(x(~9
zRkfl~RTGDAm~B&R8WqI@!j2{1swP5$X*xqA@^fwbMhOvyRsoXVh11KJN&<T+XO&}D
z&ZVX9tc_>gKCLE>u{Xa+8}4p?q<#;bzOu7^6wfDLPuHHgg@P%3w5XsD2;C-r)ciQZ
z<i1$frhAju_?%<8yB_W256*gWh1i=Sba5eCZ*m)=U0icNE)>j>2KKL<gk{^7&3Ge<
z!gJEzH`$3ranFlxy!^>Aw8`T*;{QH(I)b>Ap_F#5AK=dv(DthD881{3OCmb4Yo|$t
zkn{_Mgfob)muL_i9Lc{7*Jio1CHF_H{#dQy-7fI^<n<jqX~hRkiFv^#BkDmcP$Knw
zI(!kD+x{b{1e&AtD=&GqS3m^u#<N2U3F5*En=JTSWwvQ78-PpiX7uv`6tSubV6_KA
zc(>wgrHb+QKwKUVHttQ9UT3W%`v$G7V7e+Y_A<)r&u^A?M2$DY6UkJ}jv#*4Sv<80
ztP>d<I%_HE_W@jG8Z9I7t!DRjvgsMr7P}-q%+F)MJ+^sZC1cCl9!>cmLCgCEt(#fq
z3U}hFZ<3L2IL<pcvG=saYSwJ8G;V%jp+RzaeK~N@Bp|?y<7*`!pVRIO3I3qC*mKV(
z>UpxB=)5!`n=-Q`4h5w2QH&SFC>$I`&ikPD%eOdYN?PsdC{jjclQ`-ZhQp<q2gNG3
zFKr`tg+7aP@;xVMm?8hnPH;C4z0c{$V5yOuo!7J}WXFD(nq^x;@2{i#GIvK!Ue@r;
zSSzIiPSMp~lY3ETpx$uL?^d)2(}N$^icQA~<@BPZMiYOD+-vWT4h?DvGUl%|Su;e(
zJsFk4L}3HcCG)*BRCX@X<0eTOUf_|IepbHTY#+6|dt99i7&--;G*t=jJyBzL>jpQb
ztmpj%$*-;!HRC-RZrzdoWK?mXR-aezyr_8N%g)Y`<Fi{6ymp<BaC0nAai)N&Y#t4D
zsh*0uUAw+aBguZlcdJh;xnW+KN#c{9iXI3fa%{h%@*R_g-U#HkV_hys@v-^jJs!0q
zheOjW8bRzZ229G7+L^eM2g!K+y!YVIrLT=<kTFAyAZ$jqC-Kwnt)^%HsW$xoe4O{H
z<wB3;cR~n!iqyWTCb(b2<Lpb1T$oWGNzW3)b$FxL*|5ORs;3XFZ}m=QTl68__k5<E
zKeb8PdULvwn;mDQZ_kuVnYM_;L?BJJw!HY&%BvAOs=ZH3l@laJK9&kGvfnqx1bwvf
zHwj(#_xOZW=>2`g@u`}wm+Om2xb@eHvTz(Nl>FW2<at1r+Nx}wSV^C59gFdTm*YFP
zy{`297N=If)3Azc6WT34Q69b@8aV>U-o7IE*j&N81*}ug!=89xS=FFhx^3<I>7+C3
zbm!Jzs#>&9g4PoR;a8)60Z%*kFyrunCAm1GoJX5-8$tWU%c36sX@|0*ZNpq6Ri+6=
z?P>>PL^~FmucTGp9%E^~{^ov3d&(8Q@(QxWwY}U(&Fh=?``u@EaBArzzZH4RC7jJ<
zl+Rhbluy}?G1gW23u2Ci%4}z+r%3tejK7O>W13&6rJl=uYu&~_3GgumT5)$ne@&D^
zd;7h3zWS#@lDdIJ)n`o92Ra|`^+o`WjA|mNJv0dVn|B*^I23*2tS+K_RkJ#ChFNVq
z>YZIpS5rbs)YPF7_*5@~K`rumM5wo!*GE_Nxl`-7GaZ+M>431hGJo=6isw_Jp_8d5
z4|Y^b-+DJsaQN$|u&MTJ(SV8K+8<>Xwlh8ZCR(rKU0l0OGAx(7k>pl6sJB<I#_PYH
zZ}yMK<3)KsVQ`ubaG#x7b3}<-K^w#zh`9H3HDUpw?xx@+9xB@18OaeTSnStJEE%gb
zunVhxSuz9{WqVNIr_TD=oi2^}p%#v+A)I$Ah)+Z7W0%mRkg%#dF}hod26x5sA^-D|
z@u)6Gj9%>n6SwXo|E&7L8O<=i*PS38;{r<yZYM5+SmL>dd5n<mqvSGR*}RI99l6BU
zOXc4r0!x!$nn~G}NdGy7TaVYrF02Vo^YFA(_yAlJ-PcO*9@Q3he;$05X^flmJRYv&
z&&v9Y`9KsyDsu5L2zCV!<MfG_v7{duWiIcE2U*QQObsQx6^s(5`qhW&O>I$yqkdYF
zVK076YMdu|+^V{<k9hlP=t;!BaK5g=YiTq{XsGBHUn>|jDQl!}$c=<Go-;8w664>z
zPt5FUui)o%K9=m&uYWk<PRwbmlD{jC)QpTUjuGwK<JYrWC8OBuO)Oy;Dm+wE#Q_>h
zzpIk>D=PJgFD|*M8Z`5ZwNy{+0V93HvK`i8i_27JYx7!q=A!8aOO5*4*8Mv=2@+hS
z0q-8ZQ2o3XN>0nM08J~A+?c3*OgF(KG;(|?R2k|LXWLXY&)&d5Z4%6rv|-R^GjGhK
zmYmkBlYFU}<eBZ&AfCehBKAJ_ja>>`LyuQP;0y!3#+IR-lw-W<c%6&*?gMJ7bAIcX
zfn;U-A6)tqHQYMRWaRluM&;+nM21(if*SVESExE>)o17t0mKH}lP!VHX0aR`GAa(I
zt-~MI;;ahZTqc7%6R$k`9g7a>y})m<<HEN@rG08Kfe~T(>-mkJyDDf`j4hJJ`n*c`
zwg)4#zkU41RfDe+wmfrmazZe9)vHb1>#+$Ax{p_;&-%Wh%I72hwMRdcV`HSksExno
zJx*Vf3)g788k=uc<!`3Tp$X%zvas-E2*d$pw~|2@*ay|K_YQm}zTHbddVNOLfQUsX
z&<^qL|6HqC9z}lA<tstqA`bRc3C_9Eqj65CvFF*B^-^!`cKCcm+%5F(l<h?vu*^n0
zmA2a6P_A@>)a{xzIj_(7r7ByI>MBOr!B#X?U#-+4x$?ZY<A+PcNFD?f4Z&C?cYu`=
zhjS`7TUdhm#lVdEUm~^#_w_MH*P&iHXXQAppJ87~LDEBjXz=xd66V^wxj4$PoYrfB
z;iKhgN~WFs3W)BGpHh$FwQ5rsqvPWYM~L;?%uvqR2%;6CW%?{W(g?kojdTYTS?RPs
zut}YfwOSRjL>Y_JFRZl?pcUT}&oG3TKIUK-wtn9GBYyeAkdZFC8b?k9)8z2OS7DB$
zA|&;h04?z8f{2Uh$rAkK@OL|oZ18B_pF_vQ{EhhNOg<PW2b4#y;jMEcWBy41mt$X)
zyiOMPHA2jhdpWHo03)rrqFG+w>I3LE{aS4eXKEL(pv%#B!cKRY`a<HkF-o6P`6YU+
zD@475j2}vWFux=OIL(WJE&eRTh71@!ldcv)JmO3LiIOstm2&?w%H;}E_Qe3-#_#~D
ztQ1IT>;Vh5-Ke^C)A3>foBPIp!lmS6mc>s6<8|=b)~UydJly5m?|VZ*7Uo^bvk=O+
zWI4{lWQhqC941aleM3XBLqnZO=G3lMOM(7k8DT52Xrs7l20VHMDMXI5X<-rm#$E%4
z-c?l<xvpF*<zi*s!726KPg^I<`PvmcDIoY9{Us84^_3}<2p`L7cWI7|v*|zkCv-FY
zaqJGjYA@xSI>bodX?`@xz22-VS=h_XO104_yOJf0UaBrm(+h53{7VC1`XBzd@5@QQ
z`7`jIL1_LnA5+H|Z>M;{2W{(fN(4L*mJVl%F$&TPAlQRaLjxlS_MkNXvIpfO*n_@R
zQ%`!Lbj>KI;=O$w`@02rprCfZ<Fc14qppkh91zJ<QNVfBK_`pXC&S$LjA#k56d>Ss
z`AMoSr6N}2KyN~3g+PkY9sdtfZ0CO?#d!Z4DK-UriJS*{C~8p*9SAQ_Yytl}DP}@4
zLl90w)5_CP+LAmv$>ZtCFDjs=<8M+G744L56jl2CHp;C)9}99qL2mAqysf!vNL>n`
z7<W&bSr(EjD=QN2kX^)?6`9Wwbe=st;9}J1&k8<4?af8ow{{sQg2;sTGB*-d1Z-qg
z{-<i7+oIx~H+b*ga{j#`E4VKb=1bP!J@kAKsPmcTi2Yv@KQ78=LKCfeBdlZkv~Tr`
zYC{&hr_H@qMcpvxwt_E>a-Hi_9&R-uN92V9Tz^6<FBp*imI$=T0j}~8K{QZkqg%5h
zCRG11NhLBv2Q#fCz>`5c9Z<<;U`DsfqS*Z9tMmG<c&_WgR8NtOZNBlivEyE>xRb4(
zq(KB|l|Fsf6EN=hW$kR2%G1Fv{KxkkM020?ft8S9Ai_T-){zv_H0H+m{R~?yX>o^n
zRu%>B6i;{KUP#bwWMl9zN}u^BO}p2>2xBzK9+8(mE&M7IGJJgePLrHws)oX?>&*ys
zlxWbzktU5bDE<?4zy>{Wv0?Le;JCln)S#}keCVr@sl+k0_r@RcmM>o3RpTZQbHQ>?
zFY1IUoN}D&h-MBXf5(IcUPN&!dQc41d8qcXS{fNzQ29=&9q-fDuK93k=o;f};BeMV
z8mYfdBL<Q%HT2+K5fwVM4w`EX(BUbZu{_guT5u>6zBkcrqQGes53^#o8Bl-?=eV)W
z;?EmeL`Ox0fEh)t)P<MC9YFU7m0t(p>}i2nqN}JZuc=cn6n~H2Tm71OT^`@m!M^xI
zU=iu-KPua>gYvWc&21sGDD3u`7hE^bjsu*u+6+}9K|X3dd)~Ki2aMhuQB^*!@Qcls
zfCAGfOZlZl)#Hz9XY=1ClXtfczg|9TBWc;l2^bMKdZ&lC1M2frtI+7a>c?kxH-bvk
zX){11LpXM!s6IuXCC`b$T%$T;1CgepR~ed5C3TVhbe(64A~9~_l~=+#t}2a>Dt7(H
z$KlHZq%u6(hb_5g&AMPGK1Y9&@fk&{yAIg=ePh)7Ue_$I#+o0hk^kXt|F>aMkUNsU
ziNgz%`^!nY!G*87Ij{2kG`);Fgrm&Awz-iGw5E_gQ0;u%mYZk9po(BS(nWytbR0#=
zvM!XG{}QP%@6Iw4t@@q?bj`P<!L0xI!2X0u`QyVXOhyniI^#7b#v)V02%$>fzD8wR
z0Hg!ZW(FD)RU)HORQ{J&t>(!5F_hkuan@%CU%j~qMDTa(7p>*yG#!OZD3w<iRiH(c
zn9Q^(dZjzxi6`#S7@>uez_#gWliHdwQRAyWoT=P`sI#CQV9C-ypTEB-Rld-h6N`hW
zy2~fO9o>#JvncKu@AyXeog{6`-g(R3g-A=A{XBc<`WEFDJ+pqblDEhZlYWohVl^||
zs#|X|1Hy;ZJ2D08YMol#kG)~u2kvEkS@_|!yilywGt!tML1=)%vVYC@AQv?ZSkM$e
z49reS#U+qU{ep6~?ef;=8%EI0`(uJWT_<Lf;1^Rnzwws{qhCsp=GK%QKU`+HY2GTh
zOszd<`2m~%HAT$%d%61P&VR9W!4J;Jkc$mi$i7bRDt6fA$L^(zU-Q;!iBeTWQkvr9
zMq!XmcctdfI=$3x11}X;mrCt!AJ`(l!>Z}k3&7&H5c*Tjgd8EUUlD%B;pX#sCUQzK
zGqAh>^??iTJWUxRk!u=OyA1?PsLIoH-KyO_fy=sTwDAO5SaDO1ubM*V(4M+uoGi*%
z7vUYE-OVSd-rM=n&1EA}0aGwaI+QN!<YZ@xY~bWf3@I@OQ<{)zMmn&bUPS-t%4qvG
z?kyCSSjK2n%B7l>L|mAa6H#Vmv$0L*J4H+D?QFAdx4vrJ-PPUGs4>@;l#~*T;hn05
z%mID)=aD&{x)h=yG77;T#*OYL1H98{_r2M)Zt6XgB18tifoYen5{4^^koTDt3Cp&!
zvu}9w4t4MyZG58A<Z@!PJk~0)UvU^=|FjJ9PpYk_zC%LXNtLKPzN`|gWs__*rQn+Q
zE0@C9fbB@3_t$vH&&BWp;Ez`R`fCB_9of2GtjKQG_bGpGB^f=h#W{22-n=^$Bw|)H
zGT+mWJ>ol=xeki(mWbV`zq&s+^9w#W3I9vPalGAxm{=^vZ%>xzUulO;0LAmPhebhI
zW-X|V==K=z0@;rZa*8#3L|+ldNB2@v)Jv+hF<NVMOzHB1(~sXg-Ki`xDK&aGSn|0F
zlzHN%q-G|@qIkFcXr%e)Xq>F@KEX^;Y^1Q`3?)0)gBRY45w?dy_oU=I8xxegCLt3u
z2T;vogc*7gM7b?_Q;2{x8b|X^=9$IlLuY5>s!<~I#U;~Y$1Pctf=)+zpuC(A{?$oA
zGLT;PK%1tFs2c^qjhB#J=wWIcy$y%g{plL~zkGJw8rg^6%&qg?c!5KH-;2Fl>Gt3{
z-1z#&zUSuY_=kyL(hIK`UFgW@yA^|1m1zE9=buJ%;#f#}HU{uT@OGWlJcWri`O3IW
z)8{O&YPsQ&B#}u)X;~4u+Om7se=0&g`F$WcpE1}qo1Oe6S|;~Rt%mFgG6c<3VzF`&
zaP`#ppSZ5|S^mDW-r^~c@~hmH5e*ND+MeH0M`Yya7Y+4ojY<3z$hS%4Jirm|%pqws
zi&N@nuEy%3I@R!B!p)?tNh^&caWUHURO#AGk$W2=T(&3%c|iEPY1_}KsPMkTR5JG$
zNi_Vg1rwg^J(=?^*mpE+WYazd8w0~3AZYH|L@%Z|PxlcX)NfpSuU!;6k(#gS)jNrG
zgUv!8>pU$fF8V;B@ox{>|FnAB>U4{M;jM?HPyQu(P=2f6HIYYK_*ZM)z;V0U8S|2L
zF*Gn7d@Rir1rb&<9e&r(tMww|Rs=D2JQ*aYQ#PZ@C1)^nQ+^(y_>6Th4wITs$*`%T
z=b-K2g=T*-6x(syE%pIBW8rYu_B<)9YJ`H6tGfzvJi5c>VVqjLc|KZpor|*zY|*O!
zHuGNc<E_#Aw1&h#TW)?vRj0jhJ&c^~tNoF6w9KWNBlgcktXs1YiF%|{zq_@#;ebx+
z%hd1|_#5>iojLXhg`Y%3{H^MQS+YO1A&sve9}{-?UV-zSkQVWoc6+n3oX?YUoa(|R
zfp*l^CJ8pdfj*wCc0D7@=~RQ}J`<YO%{Io?XRngR{rFEBe;wzjNj@@n`n@>T+53d1
zfwyXKsXBuG=GPC`Mw1I>{j$;?%9ZJf+S*qh)jubXb6cn`b_}!ANiLd-95(yW-MTW`
z)F^Zt3N|}feBTuR`*^M;^Vrs;HfhIyySLeW56<7Kgjs<h()q`j&ex;%)<#HfVl#;4
z_WQ=GZwO~2rSlmfYo77A{x8uQQ)}=Z4?(D*nR8+A``6VDHsp%)FVSm@CIQM|CQJg9
zG=^e9_LH2;`-b}4|BUz;M9Bf(;L~TKp4Lr_6)E<nHxx+&^qY04gQ)3@Bl}EpBE}fT
zT^yIpab5{`T9wgZaT;@decxuA{fAqqE27{nlO4Lc<v;}WRup{gVmSv;B+=|1nQ!6D
z5;;pDyGQMYHje?#Y(&3o$#J1}gxj&cEZfqx<3?6wMmxkq-~qZ6d`EA7EI(awd^2lI
z-oW>Wz3?r;vZ$nj1FS<pz||s$VhKy%rPmA`QdIHZaBsV80`AQXgWE?%7X@??{0K7F
zIfwlf2{H82U1w48SBE!UFNkhc88QDx4M*+ZWg2%A4Ba4c!2>hZAxz`v=6)BfD9AYD
z;G>hC?VU-B0X@m{WcnSBFBfEU?Bt*^wK`bV&xGii(Fo&=7-RE3yl|lS*iJvXVt-aJ
z=AxRkRq9l*4%{MYgH0OFIS24fVy&uYnza~vg{g0pw`Zp?6LBp`@D*cxomEtA8E8i)
zsIOVchi_QgBwNlong=d{l+xu(qpfl!TyQc$W>sc$4^C6@@A!7uDI(~5>YM@|5MWjP
z&YrG5q!iOK_%zS96WZ;`1nrsVchIK6W<+FsO!A<RqTzhZJ*rjJGg#}xz&dG0IaiD~
zY6H@|+(8p$h1)R6a(H%eHY6&WZw~Dr*$i2ufliD9y|-Z)?q$2IWw;$UpM!%l>R(^2
zqC3E|2o%%;FY|hKKYQmMU&am~E&XlP8|-$6^wnjz`jh_mmgij4KEOa<i%S-c`j1*X
z&1(dGkbm+b2roMBCL!SYhm8<-GIi<Fl{Zl*8bo2We?DI9v~Su`6kLtxG#RfG9p>3=
zM{unDB|3FZi%Q9H#2cZnfpE+P&vMje!}SEZ5ojQ#fITsFsxkpys9<ZL7o%oz$EMMO
z?lSfu%CT#Gr2#$=7G%+==QiZyRNn3QfpGi?Y-hZeqk+ic8_y&5i8!-C$5&>Oozr2Q
z9E1D;Wg`RLywGRk-@Fan1-Vs}fmNvwfF~N_L7wRsimAt`IR$o3!(+?L>*FZr!7=wU
z(hdB+jq!Mc6AS3LbvCmpi|(A#=wy=5cpcBouXKSy>M~Kom+g?O!6aIGU&m_PMP+ok
z7pU*?iGY$)d=L{eu(bi+F;P!l!jE)cqg0_B^)H?p6WHu`?Y!BI<NiyuUvMrBU?3Pe
zZj=&UkzekuoPjN_gp3NtJ0e0h%n8NUlHA3uYdU_~kolxii)-v(BF@l%-&(FTgp^>(
z-zG-zPz*V~H@Uat)#5wvMRfy2L-l)3j3cN~b`j6&ds+|n{agHAjlJ)pllA@fX~T3^
zjO@9LF?IZ>!R#mt9Yc?<j{B_|zdhbzJ*z3ay%F_|OcsyU3t`KU4&{^PO#SMKI1kpz
zu{t5H6Hf~mLU$4*JVDXrHt)V^lJRB^<+AEvoGx2tPrqx_*PO5;37v;WBD1+qQnjUr
zDX3J)1bZ86Lg`!nW${Q@?pIeeAev(<8cyHwSr*x*LDuj?36es*T)Vb<Cf{@q%)AIw
zo~Td&vWKg(Ak>;vkLx8w!zWt>n--A~>dkAnyt=dq@8Q=l8mRSQ-d(Y`njZe%7vob=
z)Lfetk(7`T<IQ#AM~mKFHHQDoDgw?okM_=QF?iXcXy86K;f!%CLrx4-nH9scGI!A4
z(R*`c`3}uWdLS~8Fj=*pROc)}s~SCP`UTHujCG|Q6t-v|-@Ng=V5@|nFzl^seVjaF
z7Gnt}vM<n*%?R0wV^3AJ_0-iim2WIm#|fmKe~D;B`I0KFe4z+~1m(OI<TcHSxtX7l
zK~3#eBQ`}}OU%9Ymr)tKTMq4v^o>9IF+Fb*z-iKf=7ls(iVpcAUBF|6$3zp_Ii_*g
zbc|99h}lB=A_LTIm3d5`{kJ3d|I@GkJA9eP*!oAnlZt?{v3W^+8);d?fT+~#G83k7
zmba58UWxC~dBZ|B+khF?430(W2Bek8J2pgR%I%=o@Mf8RP>aL{UxLApU-LA?FvI5o
z58+MWZ7U*fTu1a6(|5a$b-SP7Jg%$rIIhVr{CPfyS?<;+fFVUC#omQaa}`Nb&eq?`
zFV(QuuFD(UQcR2OL1jj=*-(VSuY*>^ebAh>g0S7z=cfAO6=Hb$l{cBKwJ@i%tXYO?
zG%ixmTF{oHnYT^0VTiw4gkELqbO}2Zglhb3mQV4kHuRJx<GcALX-iQlPbkwPez2*s
zPSIhhMoDb@G|MwaCC#A|mQOTPGS;|Otd>w^wKM&WN9q*Su=HJ_7*lnTc5}n;Tb#0=
z4Wyu-7J$KbqkYyoy`3ETeCJ%MZFxv~7Zz#L(fw!{i`n+?#{xQytafIl&8bz^qlZJb
zGr5jhB1lZQ>F${8j5D}Jm@BxsLu!XqXXTr--sZsPoTx#o+oA6RM>QFA%UcB+X|$^Q
z7sZeARPl_Gn^_|4hlqYRc(e&~$(-7ci~-(BOtCR#YLlwyi@w+o%E36VBSYW&`%=aY
z6z;|rIcp|XxVcp1LFFPnLbWDiD}xnN_U^!ddGmE+l;7)<AUmVi*Che8q!ao`Azglq
z9s^M_Q0+*Cp3e!MYvJI32N4j5Lo_F@C={j`@sL(HtV(|7im(yB^Q<ixGOUdAR4R05
zO%D>|DCBxkDL^Wl_0PJ=cmI33$;;`q0ENVLe|8_gu$i}ULW2~I^+7@#D|qrkzCL$$
zoSI+XlMW_`ZLg6iAc*QI$<fjS<$U@a`E_NLg+`fh(*!hIyU_1*^@g_iALMU{?iiqp
z(V3Tsnf>}r25jJNO1Uqn6sVL`NDBLB2kHZQQHtuQ(1jWa(&EitWU*x>C)?afb7nAA
z+yA4bIE>P6AAW_CwqWjY-*JqUdWk3w{^nV2ESqNh&ASb3dFjs}_5%?e<DUV`0$dcs
z8K5c7yzJF)eCI&>qh5xj5qmyPn1XVQk@L>*#>}X5wSwRQZCC2ceQO=5N7hep2HA0d
zN6k&eH71{-Uq8OYkn*ywcuaXW{|;P^3TihM1J(gGX8{fBP0BWU+6gJ5MwsQ{{#q3~
zq0X9Zm(6PUlEq53VPPhISK+8_%ay)1iCY{BTe=!cUpO>l?sp49{>-k<ar6eOh%xoR
zp>DPl3O$w1*8Cl#QunjH1Ja*Np2@e6GpP$d(&!QiYaLW}CK{f*b;?r_U@?}bhOcr%
zH1vAeyPZ)1k`DsX$(Zpu<h^r^X>YGv?wZ*+I<`E%4Deauf$NglblGn1$<BVZ{31>M
zU@&HpY~RLqV8<--Qkv!ZU3JFF?&(*c0HfSY&3yHFTbadC0q`rh<3XNf;}8DFwW9th
znssr8sBe2-vg@R`BI?l9b)3_E52AU!o=0)gg^xpmP21NK-BwU5p9kSD0ACId3PjXv
z0y}`aNl0;IF?b*1Rc~oSFO>w|i?@oQdt~X<Utu-Q6wn~_Wn_z6lJs2k9%@b~k6Z}&
z_8Bfj_`2|1Spma@ccEY!-!Y?^?IFa7>X!P(W)2`em1j|#o?Hj9>)l1osleT`oeNXp
zHEFtL&!#_`Bi9wALvJ1E4OO8mdhaWvBXX_8P%|bIV9j?pcwe|Z7^o7He<$O6dFoN@
z>!c)5J?Z!Fd`?l?n29qoSy-{$<cHj#N6T?Vxv!6Zc^fVyEM@0unU!TrWNO)S4Y{rA
zkFD++TKVwGMpuxPs&jD2Cxf8_=Nrt9amOXk<4dB{9xNR3+XU^O6$K__1OX-Z16w50
zm@S8z&(*U$1Zac~KMd!CUG7fgb~qS2I(~Z7)bPEQM9X_p_r0D{u*DnVXxpCNJzo0w
z491tDc&2g3z#l*rlPwd3>dOyCIo2!Mofh(^wJ=OTk)&g`aD^9vPjL67xjUXi{j2II
z{HMr7g7=!nBsyVUs}ZeRx3TKZw+_s2<{$K3NOwlbS1?lZi4w|`803yn5D0ro!aH&}
zC%`-C^6IhjDyy@QAK$Rj5dB8Hu7>pXY;_KmT98H~zCz7NXjHc1ma8)vp;l^`FExvp
z<x$o%i<n6sHw?s?fAiGJ`tvyH{7GVfmPmYxOs|tl(#yUzz85yifum#d29|zWQ@=m#
zq@yq9?_8+~T;BWY{-o3AFOer(Yr>I<sa%+`-JSXK(ji^mo<>JohR@c_b?9%dSrY$1
z*FL`-k)Up0u7BBen7zF<uG?RZH2W52^G^RF8|wJs!hfWI^%gG!?}|Kn`1cM75V;k6
z3-JoN@jlr-hZlCl6g$GD>tl+ucM|mTcs`nWNiK;tcc+TkZ8Q^EsY<sH%e5uFcW=gp
zUSrU0PRV0F(|GTs(%>nq-%E_P_pt8_;C_fX6Irm$b&>ISghtFD<LFOBAx1rsJLVoO
zg6UFqxTHx8)2$Exp$1sDh}yU}{CdIp5L9V+in5$i3J!@|&fJBV<eKl}Z9SU8c>OEL
zZPY?#!p)tnu#-aJvufEYxJq15@kD~+U!rlgMj6H{`f+K2iB`azRE8K3FM6%nW}=Bd
zo?|p=aHJ0#n)R;e?eA{!=BWn;(Qs$~L6)!QbLG>YFSCTwho8(%OP<O$hlN6J08Ej+
z9WlXs1NQVo#@VnIA1;EWds$|@lnQulJpO;4Z$kc0@=g2yCx-6-Px&UABT>1=bC1jO
zjga5Fp7-CVt=rh7yeRLx;IM(P%9@3P*G>Bpk_dj6!KAiBA6!FguU&ssg&sVgSNcmt
zZpu9GA0f2CYW}do;EzZv_6JJ&#>iG^{S;MLF1(%^J5Xdwz3sm3poJu+O@FFp<BJxO
z{Lz=|@FxQ6GN@1iKV2)%j<G~n35@XY)wKop)dpy|u%so<_9Weo56^D5v9ozz7Ddh!
z3`$b*sZ=v|e*R+kC;3K4o$!&I)*+kRO<5JHz&=SrHZ!YWhTksu4gY&-XBz)mH?-`d
zruI5*a5@Na`bQ%WKYp)ZX;{1kvGIi61j6dogry=SR!(0Xd{|1ZQ`~LYEyi>PB(9W}
zOc0UpJQ5k7RX!=B#Azr91;cC@BNPvzk_G9o5w$53p^DkYgnzCepuiLQOz>diM3ob7
zyywg4Es?~DQQ?U~y1**edFQ=nTaC!XP37Xg4As)i*r}cB$SWS=qMqwa(chF<jW*qQ
zP+!y|kw{g^sZqYEcd=7oBGKr=J%^tZU3c__a`*&?L>cVp*SHlOd>WqqfU5P_Od;5M
z=i{W{;=K08PFFkJbd3NZ?v=T3!Jepqyco4<!S8RqIEUHyf@pW#ybnrdI;ye@Av5%)
zoq1n8XpnXuu+-9)00pIn6F^vHvXY9HbD6$ke9Bu3)5P4ihHY<;${8{keQ|}qflHC!
zh-0h@rtcVY>Y=ain>tyLIBq2)NVu-?FlBJeTh$EGe`4O1Dcg;={}bb0SV!0OA8yls
z8@|utlgnB^#Nc<{vthHt%r2Wv+P2M6!9!(D-wYY<(?1Wo?)2iSaR+x3XeO(0#zZGC
zH4}*OJH?{^PBWDfXeOb5(M<U(1ez(3Kr=}ZXeLuE|9jnx|3NcF{X5MBx68_!3=;b$
zqjD$6WPZZ2$$(e8Nk}xSvyA-8Jjds7e^n{6MT{dJW3{R)dVG9DA?>>OZFSI!h=@!O
zBg=dqym<y_y&2_4d~fIhu*6{H1j9itX*0t}_x|Zx2)w!W9>t3&zB+`IM3)n$0n-gi
zaWGb&U{Q9_imTWh+?>_y!g|9>TU%B82_TO_ZB+1{q~_%K?~}*=UX{{$TDr@p&d~i<
z|NWjE*ph}>qsD9p9DK%IWN&dc`B^glV&Mp3j^hPqGtqmKiW-1n?!v$3>&4(L&OOqe
zGiVJZH;u&(pKLoThu$`_SMyBwW&od|0}M)KhBi8V{_!w7k3<F^zmh{8$tRvj3)_DT
zyFMRs?v%6XfiPgTxlhhymB+t&1=JQ9Wu6P|Yeh<Ra-Q6;KA8&Od2^Bt82OeB*{R2t
z7`B*hiydrlY@T1ND<djqqv1ulQ$<$U?)tdjXOP`l&zmx=!-c&Ta9CHkj8ms&$j#uA
z6>i0=ZR$4RNFEzNMVP=|4?r0@CZGP>HSu4qdbGhm+xK|TtPx!N2uwp0{VVe7F2?5F
zL5pVZzeEQtj92Z%pHfHX_!k`;?mb;%w76DDz0PMm6BarR0U-r?lr0<+AIC)V$Jx^N
zHzEM9W;XT(eoG@WWR)~HaL1S21DpCMG-OM!K38bTFjLL!0ZWp^S3t<;`6=P)>GVO3
zBU~7TGJ2Bmt@g4Xc1a$?5?>Oc1TBDS9y5Jxy3dw%Qw^vu?eBrVIg6+|8B%(3sH?j*
z=4d7}kbp~ZOm8L@IQ;L!vrh9hv3l#1Q@dBn;oq7cWUfDd1n;TK4=n8%6SGx^Ja2gp
zMGbN+Cscs>=t<;Ta<xv+DyYRSTre_BA0-!g#wXd(eg+n_YlF@xq)Yw@nnc<0=72*)
zY(UZR!6qkVd7a*pCwZ7>=!|8nl_%Y<GS5X+oV{U>A323XQ_m8XDutYPR=#w(Dqm9>
z%=^|&2B7xsDWgNK8l$|Y#=XJ4`#xEATe}`cqYm*ZWtk*Gna6w?rlXt#*O~%FP42AR
zl!}59B<{&Y$P_Ti1KCfd^vzuItv~;(v-Mw$aR@jU{9(2~hlc!){y=w3bYJ{pA{dVg
z{IT#EVrO^-*^^ncn`GZqlxj9zQHZOUv4ot!1l~5s=vXth_@;0&00mAAD4VK8hn<sX
zJza0}iqGBPPA@`ycU=B-zkD!o#M;rwyK(W(QQ(S2>x0*WJGlf)w1vjze9<djR9vJS
zbN6t~z_3q!6@Tx>%UWfwo!KfZE%oQ|xTh8!{!CaKTuCVw*!FmnA7-WZ)6<!V==M%m
z2ZB8$hUQX_sN*$Zhuna@E{}V>*y8lNM1F}@0v8zae=-<-Kb)~u%lilLXG9m1XH?vd
z(A(~RRapoN(%aJ31O@h`6rZ$ts`9>7a4uW07+XVb&zk&@8fWp?l<@e7+#Bv+K5Z%5
zP|6LA`aaQ+aR1b~u*|kc)(r6nUB)2HJp2okVs$FBkYM0HG3W(S+?I=-%Djk(2@t-|
zzliJ_{x+MLrEdq{JZU<YO3|(NpESdiH>;Row#(qrTgZGr7#tFrna{FG@;mQlmC*6J
zzGC=Dm|a6k6IxTa$P1g0!;JxgoQ9_}<R&z$;DYA&=+#s)K^Kc5^gfAa{91nrH`Y33
zD0)5m1)WNl|8h|uYc)A7%PB<tWqsP2CBQ`;#2pwI!dHPa*=^j!;Rh1tVrqLgu)y-f
zBY?8EpjC&ql-M7t3*hy0k*aHytF#WjVUY~rT!q-0(X6*;gE~m^qsx};*NuCcD;%bC
ztTo5V+w#I_f`Y^`=>V_TS?`^0LJop?n~NATem?&nr@1&5$<vQ9A<DRe+UixS2^T40
z9_s5^z&0|6PQU(_?Jxj&d<-0<nn6zJ(#Tc9va1Yi{4%fc5b8SUhB4r>VBXLs>2xyW
zqg)vtKW>^O+j1K><p);$p2Xgjg9KTj#n~I+HUmK$g|GW^N+X@HIQ`D3#hY}N``-AT
z17R!QG}LdwhL(12@ND$L+V{>zrP#o{?nZ&l_7?LTy&-K)^Xkwh*DCvAgr{~0n+8Ji
z*dQk-&Ht=-^QhUQ(d@DQk0a-~zA~4E@3{<UNcIjotQ?y*7D1&Y45&penZbuZ!Xnzi
zykp%6{jIXn9R$Y7xrlIrSkaYV3SnWaJqz30fohoO?9I4}E#2M9GRZFSSH(hMQmlGq
z*_P=)N}C!=VtdnQY3XKQ&ipwrOWSS9c&|&tCPP}<$$DWAp=uEI?#lA=%9#6&(q-eY
zh;cn+-iIF~2oxC_Qg~l&-k1PcyF7zV)kC6@W^Mi$=nPs(a2Pb#Y>EX^>XHqEK!RjU
z>y{OwbBZVmXsSfG{x2~nVIf5>5?EWF3$~P1F{}!hDz1M&kqR32Nk5H!U{m;ormdpQ
zt|#$m%4Yu-O+<+<a?O`!c{q;09A44iTq7n8g3Jq20k2|RChF7T;zWjkbQ08j){_*M
zG!PT&vxR|{ByrT4-8}Njjft%Cy&FhZn2FM0hY+cB#0aD`0J+I}cDfNNyBvioKGKHm
zrz_5YhAK3Jv8sbHbtr+oR83AU$tqP95)x*kl_m5};N{5da%42@U(FN3htL9mz<dej
zxmJXQgRs!(X@JNGhPi@68a`1`hnWTZzgLC5SuV?%5Pyl1JX`+UFL>f8&G?sSLi11C
zUn1H0#q(~oaZ7W^A_-b<-`M5)vFKz)79lvYFOpy?JDNmB1J26&^C0GX@a^9TlvWbU
ztP&dyiar|!FC$M6X8?KYS8q<)%c#Rta$Fumabl`zb6U8!cxiH_O1iuCP}PR&P#<nH
zwWrH^5zO^=Y$O$r*_-k|9u8mJFIFi5Y<A@-$p2R9=2N+YySVhjqhy!RQArSR^s<|f
z`zzk*AvoNQ8@9mig9K4%U^Pxu07&!m=}a0KP4n#^ybIf#@CjoimBC9yzIclg4IGYB
zJYt-SW);u3Z>V3$+~H5$v1OPB#6e)W+anV6YL@-&SdCfwk&5LZ-M5LE?oPD<Ja66B
zavm$p4;}Rd<{0ZkkM)!VM|P)XCoLhAV5toU=eQiWS)<q<coy+`!u!T+mdqXD5sc@r
ziXzlJ1x7$U|5;Cv3p@+Py)N!!?2RbhKDB34!_|mNQBGh%dvd}oLKC7x9%G;($BdWJ
zO_>Cza?W_p&)nyPX|J)Zo^H`Qz*lWvCErC6Y<-^d&EnJ&6`-cmirKCCiQ~*OIImUv
zI=>kY87hWUL&`qqGd$~eb{45)e76LU`Oi(9++#12v8~6bGyRsO^}_0183B!X(V01a
z*u(%16;8eJ6}xAOT{%JNlk8FB8^ips5r6bR!NP*SCatcc9a^>#v8pTa9rkzZHo$+f
zhjV32*w}7!fA7WdP2rF~7uxg)My=MfEt()hG#CbV9@5@MYEzAw4$HioS|GjHE25L_
zMkSRr;p@Qq3h<_d(3PFOhJl242Mkd8LM3|+rwp&~*XX2V5Ug!0i9WYtqLPCxxwuFk
zRhGkU{RFu9MC>MI>*0#>(S$ePHx>|FvCL7bnT0tasjsr!t5U9PckC{^Z`Su6UAi^`
zq;@e#pgs5DxCS66T;O~Yf{VTem7C`bVou=93uI1+PHoY2&P7n{u1PcAYSA&phdidy
z#^l5UBxqsQw3%HI-INsxD6GGZ08?L{Ni>l>nxiSc{;{3yW!hM*CUpq5v+@;kzc%a@
zy4t*4usTkx57a0Z#tg)U<g3~Wk2vogWn+;IDNO$hbMGD0RO0`8yStWM6<K;yTtQH}
zlmJp!r9)s5LJvp_B#<B_RDpF}KnO*ohqh7zBqT@)fdHb?OF#&rBm|^GARrPzY5U}R
z?>x`k-<|o*JTv#7d;Z8waweIalbmxt=ks~LU+-np)Huy(qaY8{j9*SPcdvALd=2^t
zeB&@uoZKGY51)VY@40M&vMr{aF=PJ34)*=K_{GMfp&;K?mdCz|a38h9lv=OvE>qad
zg5&Xx8XapbNQ#8pQTXCSxm}to@H7Cka>~2mQk+^le{c<eR#)vXfwwsr&WI@NgX1u3
zTJjjLM^9s)ed@3p_<B_uxL;+CyrkZprR3}&{WnbWX3p}r!GJ9RTYYDq3p~TeE<PWL
zE($Wmsb-bpAE199KD0<~oQ#d{D*d$?lZDwluJYVbBEAVwhZNs*nL2n!Du?RSE-zUl
z^q}P2@)v1I#rH+SXn7FTAn<<61m#EvZwII=(ho41P*Yia0;59^?ppTCiwkRu%gaj~
zM`7Ilfg+Q@CZMu}SzNs()vZUM$L*2Q6#!k;#IR0BB6n>HZIiO8na%Y}r{-iXqXQmP
z=O>KBXUL*0e<fVo;~}(UW!mT157?GD-$)3HTgbVUD1)>n(}f~g?l&Tzm%}>@BR3^P
zA~#$gIUFfMU2%^Y;&ay+6Yq*4t0scbqg4kUyo&lpt|GGo#d(G+H6y}ShQMLA_I<A2
zE@NLCgP<++=EBVCrK8ab=z;of5M+8EWM@1NMj_aa>}V_tgDMW+SjS?axa9l<N-;~e
zbR+z}%Dsu<wQJN`i_3Hv!nW9Fq9t9IXX?nPD}=DKIyKoBy!3FOLRaSU({aM{yW`Ku
z#U0HNvfC3|>xFhfn=z%%{TT8VR3gTFbdspcaecX;<4ol+#O_DPmFJ^bWi(qGRhsJX
z{k_mHVYAT<``3r9uDZxdwHM8zv|BhpcU#crK7(X6D$otI9j*T}Y(AEuFnKjc6p@eM
zx_L+3PXB^^_7>g7d#dlS*R-LJ0zXY?h1h)+Bok$L49z8<MYw5jZD8+=)J!%wQM*e%
z8=!zT(uaBe<}*}~c<peStKAIxOtUZ{!^(>0US8<HJI0wQ%SYo#rJ>?08r?JypG}1M
z^Y(+>noLo2_{yg}H7r4CEW_KbyETp-s%PX(6p7y1(=Q98g4OlQT;1<2uAVcRzdlN@
zW=?%4RmF4L!oAC98J&j<kz74PY=qXC*LI@%F6xv*GCi5U^zS)^C$}13nEp&j*xvDS
zl5O6cd*!k!o*7X9GuF>p23k*?cqV=Jp7CLkbM0(UD>T8aqDZUe@nXp38_)V@V-ebr
z@;MvpAfSp_-G$g&q-!6(ti^b)7rNSfYHmg-m&QQZ`YtQ&#Wp5=L&JnnBwR#Ga~a$T
z2K9)j$(POA7?YtGDQUq-+l9Cxri41L<787?>;1!qJ+@ni^_yz%9($(&0<(JG>ANQ0
z6v)Rn%uRF%Ro5Tq)|q~N2`Yzr9`QTd+~n~yQRzS8_hsDkOHUd#c1}NqbGPh~Rp#lX
z$N(-pR`=g?$9Y`K)|Bf)6W5Ji=N<XZF$rFNq42Mz8RAJ`q4Wzam0`90k=A54rYpa#
zf@*tv4$s}csqpKO8LBr-O6q|Ui7@@}-SY9DdTh*YxSCAqfo3(eu2ucBF%^baK1~xa
z8k4?x@eJggr%3ZNqH>P4Fd+)qPS-f*P*XRjC)mo!c&%A8ZhmI^yLMqbSNtHBx6$8U
z)?Qm%LtqOpG6BZUj&&d#K_gS3I*=P#;mA><!e5VyQ1+skJ?o%cF$}`U&-Uve+x5NQ
zQsk9Me#vC_5x)q5-=aw(anD+Idj9H}&9@$UJygH+rmyg^0K(%7swWlb@%tyW6V=@-
zB4B2idn=I>h%ac_-3Rvti8DnS5dJQ)Y|T`>%bb2;?v6+6^8hK7CJwVQQ<hRzb6h<e
z6TNBx@$HB)X-%%8sFr4#R3*3@2u|5kjeO@YCyZV+3xoAX(QE-A0m1HJD?_;8?Btx*
z%^wGc9so#I$T%g1G-8-be8H#X@fwFigEv}${YH;cJLY+l>ON0#Rr80yK#3Y@Rg;r1
z<=vxK8iIrMOr~8Medb0n*e`~Iz$<@@<Nc^Xrwva-LTX+6hWG7MMZCq?GemX1P>kQ-
z8xlzmr`B6TjcFYA<3e@5nDzVNA@Ystr=_mOWENq{@kEjgO=;g)Ibn^g9WehHkx7G3
zF~!H75k^+43Ph2|U$zX!-HdDe;oks`Q@;lKy`gC=I&W*&)vPo(Y2@JL&8;b4@_JPG
zpoN8(y3+RAx2SQ!+dk|x93dbwJjhOv68=?PW9xdVTXZ#74N^BXM-sFH0UR~DyU`_!
zpER78oF+MVwVEnIVboEgveMv2L5X}B?YPOUu0)pWr*F;APmgrfG3~ffw#lq;;Srr3
zNrbsu0(jd*1iO)jyv;*%2z6!iSFB)JrHtFM1?ki}G7CHKvu_m`z%SvFMc=XcQ4jiM
z?b<=*yLP$VOE9Q|JPFA5cA>k>2`3|PL@JCXP&?a|rb1YuS91&TJOB`BXJiVp?^+!{
zL`+R?+j*PK>S(^nbhDzSp{f^|gZUU)EJoF9XIvYTUr<o^%*bf?AUW_;k3w=1;KkD;
z9@1cbUi3yS+PeY}e#I5^nkb^x7yKhJs%&*h=qhviQQYK10g6S2+eF^$zW{RQU#lz3
z`$YOJ-br@Od#h(K&1j5XS+WC0IfVB<uh|SuGPdi{QV*c+*3r92_6L|%V0d)_TgZxx
zcLa}x>4t$Pk#Jp0QV2u|R9e90rqd+D=%}f?Shg)2GYc}jP*=w2ppPuc4B$T%clmGE
zx_x|!`(^r!r&^nGVEJQ*f~BcJ9N=0{#3xT=ie>DSa%=wT0_T@5$0zK=`M*k3B0V9K
zjxXcHcG3!8#GQD1vs_=mVdU_ftV|6`a7G$-zNgB%zjI8$v8FK#+>eoklt|}MDz=MB
zb&jA83Gg^kcfxj4p+2cNBVG#iY<J3CUx;#acZ~#cSiH|3Rl2Yg)xzQtU*PtdQiHBf
z>D;(Cm74Ku*Nyk8EQ%Sf8vyzGddT4^-KxY|!iv&;mE3zi_32mGv2$2(2^@f*)jY2D
zO$PI%an65fvz>Q;Z6F3imnPBE*<W@nC-(;<BrEfXU-$8aRcnh|%+No^;Q<A-&&%sA
zt-||e;|))@SR(#}5Y|U5Q^*mVj~&>JcdxJ5N{f)}ZS=FAtXo{P7D7cQw1chp3(+Mq
zMJe^%gdR15EhwYHwGE7h=(~Yas$Sw!J>DC-hgsWPE5`cQ2WaLR;(?6>Uh{GHZ~0m1
zvcDW!S=*1Re4fNTQ&2QlkU+MlS>LyJ+qiDrTKZJ6RJV<j8X3_EkGB5S^^Ti}1)dM`
zq=i0V5ZgNk(jaN&dXcdWoD_djOWX@_ja{s*LxCW*fMHPZ@#T@RI!tHh0-#tbT#`O+
z=n5sm#>Bg6m{?t5r0Bora5@+p^@Pp1tbfnB4tuO}y@^e`5yt`(6gr$+UM^j?MBWV3
z4*z^|Jm&!9{moltO!9>69I$pf;JV9e%*@<AXd;yb@CdQ}fG<HfxsqS@DYV_@Uo!Lk
z<*k-YT1lojsf;l7CCly0)Vk!w4Mz?GEqli0Gm_$NQnwdb-@oUF#pjCM-)E<edQ|)P
zE&;Wt?LiR}N2q7+q>vgm=M>kd+W!0KY$SP&z>E@{uTn7M7L~L6wxQw_T-H%DGk=sc
z20<bBz%wb)WY%f&%VeKm?wpv+`osXfu?+61x%=?fztOowyove4Uta^{EYzf=xWGDT
zRm=nxgtP-{P(U>99XFM$mbM?Yva_-q<~B&XjY9QC`G$vNhr5l2hmDLv)eUTUG#VFG
z-ns-9Ya=mx2VJ~US~ooVRRY3GkhU`S8sPn0^m)IKv*UebvETlgwn8acu4b!M@@4wz
z*_hlkn`O%)hi!grckZn05d>FBrM+my;3g4v0K0e*@i2PG*WJ~8R#9qzNNtZkiy-CY
zP)zyW5V&lj#GE1dcMelEtm~im2pR2axV4I`GO^#;V<BW%0h>>$x$mp^)rQ>0+dXx1
zU){o4(hh=C-%hg?(60WxdFIK)T;k}!3j^r^Hq6SA>HCrWa5zzPqKRT^GCDL)j5XA<
znE0aNm{wc`bISV?`laD+iL&>~!ip;~Gb1{g4aQBb6AIX%dt^nA`W(od31HJ%SjA3=
z-s!a>638rjh9iZ~(u~7K8?dnaQlHw|9hat1OZ=`?ydA*Ls)^gy(#<8GNM-4a>;{_5
ztCdVt|5}ugGC5D5qV)TcH0cEKC3{uX)fMO8J<aYwOdk~=%uUBEMR*2?=T6KN)`f1g
zMN{r`wb-QQ?K-@tf8;)WpdWKqpMzOgo~9~-ym^)PwL%e2WjLo+EvcHa9(v7pbpTw!
zx{XUeoZy^De9r-jhKH&L*CSS@glvt*(h#SE3Z3?jAxw2(HP#E4ZKq1H?_Okk&2h0-
z-{X4k;M=mt;P(cXWp$H#_IMukVna;Ihh%BOK4g*F%aiAS`5Ac);Hh~#($b$Y{*3W(
zqOi^T8_$wWvI~;p?HTFZsfEaZA?zIL%HYM0S{4L$GFEw}v6|6)By0jCbES{F^*e6;
zn3fdTS&&BeKAEshKf1vRCdPiK*q8x?4STPwdDHt-0+ul3T9606EIBzDLte)ur=Plo
zD)v<ARAF?w(uf66*|N{%H!nMH^{1}rMCE>h<nO(nf#<oWpo&6bbAktbYp!Z@5bK-h
zM-;i<aqBz9nPknxtr>BAwB#N`A_*Gcz_yF-kPh{yrzg}<A!J+`iuMGSKiz5zA+u*O
zsCPKp6iBECU`5pnZxPQe#V+-_7nCMdFVgI(*47t+MQsB1MpjgjTE~!iH)C6M8%I@v
z>wU->Kd7@QSZ9&dd%@V)&2i1yg>ISQ0R~@GfJIvkP%B#|{Oi4&a&%6(c%va;rDZVG
z!m7_QIXSP|*ZcElZy!I%${Z;Jm6QxwwRW&Jrk9}{RCl+iT=hD^QPsH3VI50m<zvtg
z$gO?1Rs@B-J=bc??aX1@@{~b9VfL6+C<r-a-=jgbGUXo2-A1PBdXKE7c5g2nYDqaL
zNJKN-!#8BjWl^q8`@h}@n}i?g#Z=B?a;+Mj#V=ov5%CRc6|oy0DhcjbCvM#Fwo3Cl
z{Bure?71oc7{7dNwJ7CoidRir#KmHr(5;yii2VLxa5b?{eTp7Ql$$@6Y}-q--ezQl
z>z5MlNhOy=ZeTd`C;)BrL>vM1xVT82o$vvG4x5CDr0m==(Na*TjgC?=s7TE0jyJIG
z?cbZbvfZ!fub23}^)^**88uH<6o`Hlo4!wmK&(q^JXfIKzd3GuLq4y0GrNbo`-%2r
z;l0+H#=<*J9k3QO!iC!}y89pbv8*=Ofk5wsF*ilvbB*Eo{|Ts~hA;=^)1d_Bpe&St
zF7E)?sSDg#bnkyg_)jDK|F-2qX<od=i!;8Reb*m`4~2PW<kI6fqJ3X*_kU=am5zTq
zi8=dTc_LKygW{8OszL|z90j>%h1oolrskvdQv32|();?T1Lgk5lpa0K003cdobr?c
zAr)~SVCoy8_rx>_W!Uud(PWFkqv-zR+r|W3g7~}hKB0gx>F$>d)fpRg_1n#K%YuTh
zUlUqiLwidgnt<@{;g=G99VTEh<8+_=jvFg!4+mD4WssqPp9izt%U%tsWOgIV@1Wu1
z25A1WE5oQfDXEpAOg_GlmHrD=d7mRq{&Dy|TD<VB*AGOokl(_lG#v2@HiZc`FU|C1
zQ?140s%qhK9x6Zn89l3O5!Ph`Teba+j!N*HI52;#hzqXuLZfe_2qtJ&OsoDV@pFj@
zJ$oj*AqFkze+vEj!}e~VF5gFm9Gis2>D60*R5A0hweIN+f3-fIznwLna}ld()^}_1
z;V)~Ha^%N{<}icg5Z)!Zk8ihDf3RO|DFNh7^5<4e`EZbVpI%p%|3fTOs&KY?1GBkZ
z12t7Q=M($-19Tz!X1PwVQn={FZ0MxelLj?VqI%*Ir;#EH7BCHe|7UmEgQ~ZmtY7mM
z9E9gz;c`0b-6rv2QMu|FD^2O`L)&)C6T~SR!YrLo6p`l<AZ~Bm{_za|A9`$M>|?<!
z_7AJn$J8a_(;C$y^q{=Us)c^6vEt9foTt*Svm3kwe+HT+3%z#9Zh$7fN`IXlD+z)8
zD0f)bntsQ{v^Ld?VKyTx-J*Xtf~UOMJt{cwQ5Aa9PT7}R`OE)(hD_Ja-7N3Mzhb<L
zK)YFO``#Kc$?dY$*S{Vf|2;Q4=HGLjmq}b_!uwwyPmkac45zTR+x-z~BvvdL28csy
zmc012oSYHG&)|Wj{N<yU9OAuvU*6@_=`e%5(olmJ({dX>&N=^%_<2T=VXo2m_;of*
zxai4Dr^8?H%9Tj`jpN~Fx9_*|wmuI;WuMv}Xf%lg{duXOD{SvaDtVvwMB70w!J&4%
z>bJ%GQ_y+8+>%(Kc8Z%*b=1cVg?&@=0Jb5PeNXaivJ7)@J_R>O-bA-TZ(HDr3F5uy
zv-y*a3VvN_;zYc*Fu(hH2_|LBmnoKct~X3K{!^{XOvLnL>W1%y@mb5cXzFKeBJ!^r
z7THZL!!0>*y^qNj0KSIi(NgJMk>7^?Q5?LXO&xY&P2KK&s;7RvzYcUOB7)=8al@GQ
zan)jb21abE`?Gsf$O7%B{IK%fZpYo@H(hp}4nHyHmnOu(M#J%U%c}<N20k8FlvCNf
z<z{dFh?gjrZH2G*GP-XLvAk*qRcUBIJW|s8>yE_3mxk9vg?sO@FxB<fLz0BcJrU#{
z&HJ_<ruIV_ILvXHqk}1mk6YPdk!C*Gi|=Y6e#7s*bt%p49og1^Tm4~%=7c(NWROs0
zNvT)EDt?U@&3tuC3vjCvBv9?M!oDBut^TAuIq7Kkdoswz<07Hqr@Ltvxj|vSK6#Wr
z69RwvYO#4Za_gVAU#UvBe%ZbKr%S<H0d%r@QXN<$sw`7X>q7?hKF?j+kJw!_Skxii
ze90_wmJ!+~9iiva;Vqc&Z!9!!{&w@7QdQ(NUbAgLqNjF=T!z`Bw1M)zU=}*M#td2R
zIWhR$gI^B{4v^O@)?z$m22!qxzlNCH#N7GwbH0FG{)e}<mRGze2g@_+$xGSOOAh=l
z7Oy6fDb$aN$wAEtcpmb1?L~f~jJ%X>W<r-I*dyp7$hTHB9NCy3qCr3Yxfb|BO}%m~
zT(A1|A4~kA!LrccszTQQ>1FPf<El9dKJ@!jRM)`UQ=#!Kzgt>B4s-l-3p2J*MJ%(h
z9%Dj)e?zHDN!j~0#VJ2){j*>7ZLOxTxt3~DvbU<*rszhpXO77mtoQ9wA$@);C@IYd
zSX8}n`?uu06{Uvts@GAsTQ4QtvDWwsu^&OI{M01_*&TSDQDELvJwKR#ck56u#P3%4
zWjtM0^z^}NqwamHGfaAlLB)lTSRLvX9Tc`TdcG@UqlvXKF*LB7V##gdZSo&lletqC
z#79S$zuU%Bn=@)H97ijR%C2<3{Qs8MM#j@dM(#gKefjl@aviXVm2tbU;PGc|uz?38
znP<iq6p0s#Gk-EmY+A;wxf}gOD{3-#y{i1sBYV-^RY~rnZS#xY4x+53Q`Q8>gJ(y@
z%ZBdChQ(kDrE6Xw9e?{2ds}W=Q_(Dy*j$H&L*{{(k1E8(LVpPQ9Bo-icfIkA46kx>
zxH$Zdn=USsO8q47e4An>hnVb2-70%fO*wxwdpp!1>9?CqgY1o5!+bXuk8r7H)Avau
zO9d`|RiZi}ld+#@5xJ!-&CTSkG$@lXv#YClENoPj#%<nRIr(oc#Fq@+xLg6~*)YlT
zl4-*Yf#J}Yt3Rf{kjg^>v&TVPl#~kAYf{gD*YEe(Rd8a0fQnL|aGvA7nZ;wbq_Vkm
z!}7N9Xyb2S%(v2tZ&{?~N#V+xO>tjwhw?I5)QkgeY%idzHM~7F5a)o!?As&9tZW6W
zY=l7S%mflC^(cIBYG?hvxFiC}5LT@lF;2w^<7r}=Wm|TFjtFcSG<gx?*Z%#^2zo$7
zV_%D>sab-Ef8oPWuvH2y3v;NHsZ_dc{+MA|k#Ltge@pFunel3m`bRhJCqu1^*u@nm
z=P2zQH;gPA0ZNN*Q+KmdSLd@I0~pz;^XXX~?<o~#&o#n0M$}C>qg8M;>LuYNR?3f|
zI(oC0TfL<J{@?7e|MK!BTw8g?01Pwc6Sg10$+bGvPt#{UvvjH15`W$31%JlXDaw_g
zAHTyNxIL(mb2KgBdlh}+%F}%>nr7$eM9@l_kbm@=2141=miDpPyE4iH_Tu?(7mJe2
zf4(EBN$z>gZ}B);reFDYQJmhV4iTDMW-U_MBJA*fePZsIc%NK1x0NCW%@IfYwo(@i
z@8-izV2>_tG-Lt9v?!*-`gDy>AYV%E<zl@LL%&Er?6_1SXB7FNrcp+d_y_g}T8+Og
ze=;m3Kt{x}a-97Gd+Xi_R|lkLFeTilxGR7X|7CGfFefXxE8!QQ^7Tg*Swi$K6kniP
z_1K7t{l{r2@YY58)ffGJyS1VG*o;RhOPqYlb?e4fX#G*OAdM^f6=!ZovsR@a`PDk3
zXIV2gKogjwy81m)@)A5Is$bcqKS)E%99}0L93(p)SF+duQJUhOYh%}76E(`!$H0Lb
zu<v_nwxK`dD{OAEJdN?#7uOod0(UbmUw8(&^)dNI!!m=<tM^f$sc*v;98>u4S8%eq
z$ZwK7XIkki$Pwy_phl~VWv@jv;fTP~6N)t#e_Zw3w5pWa@XHCKN)3n`ZRq2(t{aPt
zxKFi`6LnA764a~i8w#6zaO1z0K;b}@%JB_VO__klq6T?EEM;jD{fE3i|I5$+-#-b#
z=L91^sN*Nr3cb?G3&O4x&__$5u#l>+Fh86`p-nb(;yw&}upn#O6E(Qs{&M-yC0jci
zO)*#9xVCnBJ*F-2+l9;}<_`bCszYAJ!d&6vb@rx<WL1zsEzr`ui{?-DR%@^6OGVj3
z>>zD>K{Sdz(ALPP`(FB`G&CZ7r`(_V2EyocC&ri@`8tQo4!bpGg^CAJT1Y}15U3R^
z$6~$APmhBCycbpY|96{&sVxx6)5Va8T|MYl&Sj{7Ke!ARvU|m|sbP>B*ixl0@0(qT
z$$tC`usQq9MRW7pms;xa_?8#44Rh~=Dg%3uYLEx2PQ@4Pz*3HmvbQlz&x0d#3-Yz*
z3eIcO{^6JJBRLcH`CEdM-W-}es5-dM)&uCl-*EXxSU)pZ^_E(9Sj$4D1WI<mQcc>n
z;M-61Kl^z753<cj3WrTo7sA(zHlOXeE0~4yP(7TgvUfjjDD8RB3I|s=aHOi?3p53h
z-YVxQFs08Q<2G?S69iYroJZ+&S4={I&koB8_O~l03C3cRFUc0m;HqK$h0MEY@=|L5
zS-8H6X&;~1L=$kL>B7l_bU*)tEawpgg&XScF;4sgT?Kap%A)pLOd#Bj&<B7H0cMB|
zt?*mgJ=a<GDA_|S|EW}O0>92L%~IW-)noY&B7YsaTFImN*RISj<{~fujQZ`r{~P_6
zHUGP+>b!rj-NGsFOHF#o2i225-|`@-R*7w(@J;O1Orq8$TA;J9P8tmey^<r0VIGLR
zfVC`FHm9%U0}X=tho0A7SJtlYn?i{k*mPLX;A_dkscJ>ue>62}LEA%=>vA6JRFjh{
z8Y){}Ix7aIw_bHOw6*6X-N=xMh&(K5VdbfbB1#22geTH^p6)Z#Uv6A8T*uh~ZkpqR
zPgJfeT&7oo!OF<6=72HpVC4&LrZ*ANHSmo+ubw+Pw>R0hDL&#iSA)Z&TA1lyw%n5u
z8P>NqAk~r`Foc&=JMP5scEqQZQ^fTiU|xw6a`lnTST5UbIXhlYw}$W!<v?7fIe@2f
z(XL0=C^E{EPktrqsIK+Y_`&$P?95vq{rGBnQr>iC*FX0)05VdOb)w+HOh%Sh%kl|V
z$#Ls(qne*p{?!{FmORm2Z=M2B(S_p;elqjT2l#+JxVxIHWLkw2_v43kjn;4aS~qk*
zgSF$PJG?q*lSJi)hB5)F)E_3Lin6ok&Rv*Fj(4#5_nffGLG=%D|2d8puTU<stYXC%
z4;%3wIkr+~$~pe4Ork9Cs`<}J^4kVHtzhqgx#~34ULP%C<18fX@8KlrcZt_ym22`I
z=lpj1{ZyRcvihAPx7+FKOd(>Ja}x92Y<$4u4jAs>sWO81@_|8?37jD~1;b6&!-lG)
z+U(s(pg}l=FC;l13;3yz=V9@ko*OtU2!`WqkR1}LL6DTPlsdbl7Jyqa5ojl53~)n%
z1!)??Mx*Kie7`cZp3xxadqzQ9UA8kaVy}TrhkyLDeHU({zr0@cU0S|JyZL4h>9(r$
zjZNIPit3TjwZ)}BYu%VyOOI1P$KK|w@6n@n_gUf6hl{Egt9Co*rucxG6S42MJUk1E
zpS6F@Mf{Wl1L)JXYp1<M=R7WgK&93|#4kynWF!>oDrKczAZho?I@zB3TJ|HW)-}oC
z!e-Oyxo77&-b`1I?m9CYGyE0V5$0%p3n#X&@OjnNEZo(8Oic;mW#uwBt25Wux_j>{
zp57T_Ij%7qGm}FNko@<YJHT<bAsTPl%QOFub8^ZH*=i>%P>Alr4zai?qR#H_*J><t
zluk-YMm-N-X~m*L)PkmfEof|+)Us_za_;B4u7Fz$3{omuxM~DjOh8lPLD?)T*e^Z4
zrYJ}6Ppc2Rp1ur+dM2_lV~y|#gRkSN${mct0pfm$qWq>O!)bl0DydSlU_)yWk(Jq0
zzo&Yl&qY>1GKmD1_Fi%|?7K_xvB*9pUYZ}_R(GxUjl1wggVN+^Xn*-7oBFJR00|ZO
zGf8EqFIi7KrqRO2;%VWVlgOCK2IS>H$4y0V*LveHFVC@1Og?u6fS4nk4n~y)xf4(z
zHzEQ>lCq3*PR9rPE#j{vnh2CG)8Wnf{Ps!vm0?+e^bqK`XOWHt+aOQ(Dd)?RaBy*g
zaxs*1WmQqWM(t7P-8eZ+uaQH9?ij;prX;Z>iFB$6I3-mgN`)ghuB5k12HlQ2)|#Y{
zvr>}Avq1Eb%EHpBm$%<`>`-QvTkr7bI5`XzQ5FoJs}l{f>agja=lB1z$Ez#-QxH@@
z9S6j=92rMuQ)6^x$<k=$Da6W<lc<I}rY*c~FH>VT1nUL-nxWp>#lLB_{G`bRl<y_x
zSgic<IWC7@RjujV9fLHo{P|shbHuCB_kr$Og$UPui>`92c-q0|2Y5@CEi5kc2pFsQ
zRtwzGRlUO_;5v3zXnazfm)E2@sr)?Rav?PYuhZ*Zit$G1xkxcV&cp(7=InZgql1Zw
zJ>z9jQs!sJq=>Miw_DKG$W{XawA{Y@@n~xSaHdR3g@)Ka`hqK|PGJ!_E`fVALEE4b
zy5|Rxr_t1ZB;L*@{h}W~geUTJ2t`OYj_EZF`V`1vF&_gJp4jxV#{6V~e_Pqe_aN5Q
zQ%#ap<Td4A=Dl?_FB{129eG#kF1}^MuP5<^MAhwz72yq}b>1D{)!I%=_M+7ck4)7!
zG^p-dk?Y~ky&bjdN4*y-N&9$HS`X7J2>Kmh1wy$oo7#Vy0>A}AHqS&iqZ~ua&5BTH
z6W|~gM|0_FMWpPqdj_$D)WD*ENp;o0Zi+Tjt2%v3O0HY?Emsz*&GtA!RnEWH`{sUj
zEuK~NjH_lUMeLj`G6;1c+*D5qMEq>i{E;vW6iMF@J|Xdv-0U74m?vAbB$uqH6-`-^
zPffZwXCG?$n<cCqJ$u_rrT@u-nr6??b^Wqbty}qT5PBr4;`F8?UI#31to>ua;e>w~
z?PmLR+uQ^cYHKxbXL2EGVJF4)_-!lLp)lxZ+h?;_$2zaZrE5YX)yhIu16R<*Aeva$
zJ0RO~RW=S9lzi$$nmRTJ>rQ0lPrX?7{Z@c=V|40?oBYycCLd6=bPIJL-|8p3GSiwV
zjKN}3{fQvJSTt6is1Y&z^$YDiXFN%8BJCD&GLVh+xAG-tzG7L|EH&<^=oV)+6xy05
zKw6l>_nkt=k63Sv{$&5{e;Fc^Lu{k8%5ASbi_7A*DDENe-|!3AannbUup^zJ#&kMl
zlCD#td9jphm9!M2CwBb9LMPVbiS8NS&)e}CON)?F=s+bT7~4CSua380{qpa*D>n6M
z)zAn);B=$$edoh{9*drcy0E@l0<FkuI7a*6Xc0ma90O2I1!vpBy<k|r;8W#)7M%N%
zg2)eu6tkB4vEq=oRID57*WzsAD=DFC_&FbPl<%J_U*+e0wbbgbqT9y8rEt^z<Qj|J
zo@v6sLB$MXXR@*lsba-DHoxd_up1iiWO1jCx1MAWSAbvI5ZWMmOoY3aRw4qyeP#3^
zQb?_Y&RkZGjxtd^cH{jpw%A(LzlQ{_wNiZG4Xj__-p5Jik<_<m$BC$X;Skk;e$=Tj
zz;Z@Z<X#AtWFJkh4^cW*5Sg+Qfrok_q6bn4?wN)fX-)8-ikq#tt*^HO$_Vk2J@c@U
zPA$c9c_YZvz7U0nVB5u)WNRtb=&1Aj@dG+wM#ROfGS6Y^k+Y$3U7Doy(zkj{<PASC
zH=`(u9=FGg(z(Pcmc)?&bd7fFqR}a#C)Pw`RgPZ#0YE%Kn|m})mtc?T>TZU=L%Gj&
zl3CqYB*U?ZVvyQdvD~AYthF5)us#;C-e@@Kc<M7#Sj^3TqHE8@U3UGEPRM0<!d*Xn
z)K~?Zd?u>=zMxvLIs%ztW(PQgRJ|E)w@`J}OLxEH!F(;ZIctHdk&@T?!#^eP&D865
zyN$Jtn@bmWc1?S1&PQBaZ!BCOOzeuEQm*`7^*dvl^mKD<z0=zUO-uzU_cms+;C?<-
z)tqW~ZVr!pn-I-d0Qg}+a)1$ycuqQWkh~C%q&zal6txcw54QsQ2hgw%J}$AK-9-mW
zM0mkxs;a6|B9lwJv;GcZpYC$&k_&C6XIDBjBpz&V6*$<>9Io>%?{nwbclgBH6BRCb
zeV*?OKlbmrVx1UTK&{8R_xFF;nv*YU_E3)^8k!k{9uFdv9;v{#4pe!JVIOVLd|Yu3
zAn~Fq?$E^lvGf$<p(#Mlq~f$ha*msr<9P&#u<mX%tkrw)<r*40u0%UUz{3yR0hFLX
z$27kKZ0RIdG3hX)tp~DV4u0$}8M6`StEHYwMlL<a*_~RawXI6adLO{;bSnFMMo~H8
z87zQ{>b`3SMvfHza>oE{z3u9z$|Q0}$WFvPc}<#S1wAHfwazGb+=q>!P8#fM8gwQ4
z7;8zB_ueIT=nYlr^y=uCZ-k-xZ7v^NtXix}mXPi$mQ<)RjM|3Zc~G`5h&;XA8xC%Z
zqUQ|8HfyxstYF<&@8&EmGI7aE$<wDFYMmHIvN-(mu4dTq_0-lfzp6<`*enGnIC>Iu
z7_B(AL*UY5M|KJbpNbN9glGGH)wkPJL|zI;Cq<#gd<5w07R5l1=mT)g%yq|%Qfo=i
z>VwqZb91yI%*0Y1uwW*K$(iC*M<!u5jye+z(h5zhMBtGg&0(k9<fGtCbnVjG^2Q_w
zTvf>|Cy8+M>yksIhI>?nyb&&b1trk(3z0vRdvs*9M7%4f46|bl4giJweL{{f6g8)M
z9uwQ->&L{Td@q^0=eJp`KlV)h&fdeqv&mV8lX?+1q^>Za)MH)6>hAU|VLM;T4lef=
z)(;d2!N~^!u3fwHYK5MeV`?>=c-t!@d3?s?FS3HTBJ&-Qslb6oYi-@Q<O4tLtd9OY
zvL-F(=rEk6E$ACvn|ejGfbl{_ss=ma4!JlpFd0S>30Oab)nn$`Hj$|(uC0PBQE~#V
zV$mmZrDOjY#Zg^tWQ4KL-OI%g4ue&;AED&QC@YaC!48n3q_2%%Yc@N9dfv_iv@B&v
zEt89~Q?IX2-7!~QfEQkFVVsd{CpvU()a5CUCalnyuG_V~8z4Yb4k9>vnQ{YMF%D0T
z2ulLRKPL9oXDECBdaVpn_RaFhTK~6$1J3)-#paF|i`E||)cyNr>5`BchK>V7Gbax$
z+!(^q!*^CdLA~yhdbE(<i_aj?)@*91@JgYUnzUxG>8$(tty!kpzC;ZFc2cb~3Gifv
zF*4dubT@86<6VbH1zA~=*p+8?piKk9X)79G2MRXD*7ffh6mCwW(b#LYVF7M%l@Op8
zedGbRXT0jH*muWA%$l#D#H`lrtK4{z+vfFB@weSy0}+$Y2=(7VPj?42`7y2o_1T9v
zH|}zx>2ti11<hKcN#CO5^6F3AsmL340*}UPZ8I?IM`uamZUKfk)h(*EfpQjhDL}`e
zCMnGi!>nyPs;mvryodL8DXaVGF!xPsnBu8|Q)`Sgh!L^n)S8!RKNWw6S7k#0wYiCf
zv=wOYFD`pC2?Q9HDZ1|nih#A+b1v%5LY9HM!>8CZW_)2Z*KB;BGmEfUKmPYzoXB6D
zhf|^PtIF%%hN7LbjIemhm-w&EKNVPux5EG)zOJ@C8oUD2t2TtNCNbrJu--Q`zt$U9
zbuWbk{sroOc^8i#JNYObfJ5#jlK7%-*~q;^7o?VJ+OZBRZT?B5>@M}{`*p+p#xI$F
zIW~Me-gN!)!3bn_d|zUE0_p;<Rd5#a^$mja@7u48!BnU~m)L5Okg;mWR^067RMMFe
z%heR_?Vr!kz2i*umR9EYxl;NYjR*QW!82t_jPZt=WFjEUs)yCORFI-qxbyEhs2{fs
z`kq(NcXk@cHKOg$22`Du)pcc@4aIUvK;*atuFVDz&BahXpX~e&hQx2%aL=lLTNj0|
zVViZ<{&7*MJrvDOy*LGG>iT&?;kuMt-<FJ*ddODzyRX*wrNXd-&gzm?(INlbmpMx5
z^KiYm^h&D6WKgeo*U)eATkN92@pOw04d0j4NJ4Zr*CUTLm6n_*>p5VXJ2phDOJ8GN
zF|4PZrD&IBjEiawbHyy%SsSmvY<1}J*U6#y8I1!1T8N(DiKZ>R&3q&Xg!G!q$uMTs
zvk16wECCJ-HKKuVq4w0_F{^|0i{Cs0=VcZ}T9sny2f&&c#1G9nf$yf=_Azyr!{5mq
zc*mnW<`(Ek(2ww9U{8!d*#Jf5&b@tnY(k*L0l~vK^niuU;^ja$piAjF>thQ+Ns>w%
zx|Sxhaf*Ka`EfS`^*W+c3y;E+Fs!5SQUc=HFiNyF|Gt*$j`OFclbCGJ<t1N_fw~`=
zQ4U{fhKcbO&vfYp5^Ell2iqp)IWI_fU$;p>Ok|EL<_vriFjap~X$;JW@K4#mDenh^
zh9Z=xM&Od<Hcf0*Et~wM`~5dRX!B+RCBp#{or90Yy=a;&wPn4tNzws**Hm}RjOOk0
z@<Cy`ac>q!m!~MM%a=k(zE)|@e0i9`@at+S;(FGchFSNV!m1BUiN2GLE3hs7(iVlf
z5SE0l`IKByv4X=e)ey{zXd;Iai*QFB9e1qjE&-3O+Cq9}aZ`8u8jUyibhX|@MkOp9
zl(vS1$F_Goa-4X(crTC@rA-X;Sz`1bDW#~%PAUm{gEHfk=l2@{jzF?k02W#)aiah2
z@&sgGY@59Mq&tLSGix1F%pB5_|M#53noWX-LsyzWRJdu*h65zX_r9d<eLEkb&@2I{
z555Ah1!Ij4eE}l=r!w(o1jO?~lAaz%d$K9Eav`0`AjjrUHL(jaGBUIBLeQ%nB9>!^
zi3wvBP_(t46A{ePh;g1lrwtEHKxi32?2ePJ#D^H?@J3`Q|85A9CqheeKQt-cjzGRk
zvg#9a7Slm^+}Gm9vS8(kxF^0w;f=yzH<}=ebp4vJgxQQc%JaCGu1nGpsQxxtu}B-g
zm%_9hP_hg7amE2*I;;xtE#T=5HhsP1_=N?AfiBQtqJJ(|3ApmRg+r?DMgI`(5jncl
zdop<<yIYoe^VfrSI3ke?#65GTjlW&vy>nSruC**cW1Wr=eqIfAAr!}JqS-KQfx=K(
zl*S0?SWDTdFrb&0=b(3dTs7j_bRk(Y&frZNqE<IL(RjMec-~nxCyi5h7^=AJJy$l-
z<Dg>9r4Cndttc83>!BuEIcm5BgI_u*AXmq{2?@&t?95-T&Rrbm{H1A^9Ayd1+hki=
z$bGBxT>-D{C!(5bZf9!zwX%vh{-W0FhY?=<a>cJFI<B}Tk`nLd&$U)$9u+>C*q=se
z;i}NG`)LAZ{e`qWRWL)8tB^*jW3t*5NDBnvoe7MJM}4#ZL_^t~@(t&3)jQbyl3n8{
zXANW3f6rY+Zy&!IS-WCzQ*LoT8+Q1s@~!X69yOfsR(85r!jnI~qEZoruV5cY?Y5X3
z%k}4(i}$Ovy;T1BgF9?cf||WX_mr&++xyEy7}OxkQFC*S+_<J=c~H4#P8nS35iepX
zfPApjS@D@w5FK08MqGR{CHno_R2aoBbdw1Eu9Mq=V||O-V<~$X7xG0&DSLjdyQXQ#
zDbK_P#U5s6RnN2)y35K6iHQWW=dd*Hi5D0ifH^a`c-x)iEg&Jz&fmmr(Hw!-!E}W5
z^<OqPb9v`xO)u-Is8VA;ZoEULTUh^)%#tY|W^~8-IgEIwxx?I@Wln;mdbY0pw6_#3
zm67hegle)+>0L2-T(x-9gm+w|m>f#FljxpL-cBMvQ)lYN*b{ZLoEh5LZXY`lec)(G
zTooHEPV2C<qSe343=-dA4WTizHP*LX2eCA|xA@Y036Jx8Vglh?Kyu!5!V}0eZe6?e
zP*!;i5G>|&*!UHOlRV?X-M|L#6y55zbnrd)XUy-t0s)-p4<8p!m$uu?L{fK3O*smt
z!6-3R^<*2`!gnP??Lyf1<;bRKO=nNc6!K6&EwCikYYw)H{6z5TMR>_Ls?Q#*y}5~s
zPVDEjl0HYgD5L{COes5y!iThaxCi3^zOotZAW4}D4QDfE7%xnEnc8C7ni|D-Zpxz-
zs2hWwFQVDuu*MvZ+T&+-jAo?AH9sT*tg?PQWHHuM8<S&OV9)68Z>w3LQ!s_vQH~nN
zO<<g~R9mQOw0{{}qbJs3W1~g;{1z{)KD>&Uwy#>HHr6`xX`ifD{D1>5mRCqU9HONu
ztmlf?rhci2-{lcXiF>WOTvuya;&!t47<1G0$W&kY<VnX8FKR9$BElHmF5B1n7U9CM
z`v<<>!BsU8D9n?QR&FUm9r@f5s)fN@dt>U#b~$D!KFa5k0Y(@bPiim!v1JxSt+jKA
zpsMRY44q)nL8nl$n~yRMZm0$jU+zdY7}lOPW8ryRrpa-u$5-XD_`yI?h>Eb`{^K*j
zf6t8`^qoa+3z0~j^L#Ca+s@p}wK`A4ZqF3^{kB$=_zxcgf}-qCl&kFWyVqDfY2d$8
z_WD=-?BKtkYw`VR(pEXoKRW^JYZo_f8l{)Zg!J5)`Oz&!$8$U`<KJ^UCz@Zn-tX}1
ztF>E>?Gm>rm7&hPGcB$8lV80p8%evn`VdHlwK1+)aa*aUMLm#x!efE63$-X0uCvXA
z%)Ke*#q}x$xlNJWlNU;ovfO3xCPlD)-P~<pr-wT{LYtJO0RkdDJw53-wf*$!)MSDV
zehO5m#QE4h(s{K#!ph6@zDdEV^O?t*t1251GG(`SWJLyJon72Xb(Ote<!AY8&L0z6
z`D=HfpFRzGt3$%K`?jk9RFBg;203)hW=GwIZrvjW`p)<^xO*lt#uDMT-u!fR1>pK|
zV3gYc?&vCT#FeE4`S#Xj#9>$2k(mj`GeqEK@@LbeW{snJU&Bt?x9S{$S^!;gI<gkK
z-E9S=D=y8Gl3+R*NMXDQA^#;t88}l#@aps{xK{TRzSsrq>XIF8NHd}5E>(I{p!TjC
zY;pOyiDNMkJ*H%mDl;)jU52k^AJF|uq$<hz_jSZ`vV^=d9`wCYC7o$O+Z3o-pXPPP
z1I!s{lc`<Q$77cq*N+FqX&w&Xj!IO3LI``fcB8DemZ^+yD^4IEu$_skK2U(8pT!Ya
z&<7T=?H!AYxBpVrJ>)7Pt#!u`9zb(S(=4R*39}{;prA{d%#v6zZnqG&U0ld2qlxhe
z`9!y(6>TNwDP}&0)7I5r4u*G~arl8<5kTFY-Yq^!y~^L;{nOv{aIyp@H3TtomPdK=
z>uzMoB$c=O%g*^uPE;r?p7KRo7uT_Gn#W#osu$BSA)0&~N4miXZjEB%h)PV*i5^1;
z+{jhDdK~~Zn^#wG+!RnDAU6OsSjbeR0%5$RZ|d#4<jDG}>gU+sN07tRIY#%KO0uy>
zu#_4ch78TeU}X!Il9SUt-;pn^Mja&Gv6AxG;a7fx_$gHUh1;AhPCPjX+e&jRjyikT
zB`o%n_D9vWXd{<;N61#9L@Pyn;1082PB?^-+FeZp9IY*CZK&*GC$~DPB5%dOBZC&K
z>4P}4nKth$e(bIOAx5<CP(B?_amPDeF{rgH$j^3dbGVk^o|pC<0r#50vWK)n)q;Q`
z?<sW-#lTaVbvUUd6yIxIRT}Yln8@0zqYBMeNvdwopql8dT!`^(OFw}Pc9c)?!otTy
zbd^4QjT3nmRXmee;wPqb@Ym)~k73tq-k~?mntB`3REleyu5UW=WusA7yM7w@M_Ry7
z4)`oeQz*Irvbx~b^5f6-BzN6yo;%}9oN(Kv=2ga#L-C^W0AUh-pNvdVzzKlHwGalA
zv)bRHM7%<S!+<@NJ|m5xW<`a$^$jd4hCR$c`xc?{H=&-(k)caQY`yEls3T<@&;i3d
zT9sNgur78lPXG_fm4e(IgpK*qtgUQ;j!H*o`9v|mn1t49KGIPU^=hjssKCJa^~Uxz
z(nhYcADMeB8AP_;ZTGwTKET8NC6DGZ%6FsLFh*Ct?oC(`p|ORG%jMh;-3X9;!3_&G
z)*KnDgTB-ImRGS(cUal=^V<*)FQ^Y=Ti1caW22ihTfX9Z+EZAV$?J$>OY6R|uSjtn
zwfcKc)fHVk{ch?ir=pU^uRI>WoQ-{I2x+(%eo&becZQGaMn<pq4SXP}bNSq$8S0jM
zoh(Y7B7&GBf$c$HT&M|aW9@?Sm7}$CE8(;SqS&c%j=JK<$ykFaE>mVuyi}eKsf8-!
z1vGXQv?=F>wV8V+%Vs7UVLW<WEd9hpB|@X4j*#DC-y!1jbhZ%;O3!F3_%ylonHGa?
zkm(L6+!N6_3iS9C3>ZSkZyjOPcaa67t$vu54u@8(GA;abr0IfYeFgQXihmM`%k@G7
z0&fJV-;q+lrcYt8YTI6RF=v7^b;62NM^)|JJ}bBD;>|bPt#1;xdCiG=fRB4TWVaQA
znDqdmRwS*&(ng2OTD{6>>~SwcWQY+a^n#CEaf9ovf(1f{L{-N2C2+B?HhIU|!z|;x
z{_W~7Q&EBNOW;o@HXp_f5*B%e@ghG7Sd8=yzW*>#|6II7JS{pApVf?XQU7vJEeM*6
z=SpKqup^+L_o7ot{N->6Yes+oMK`-kI=qOuGr7JSB+nj30T^xJDRtHwGbnA2pEsp&
zsv2_&a_xZSuVVvR-K_ohHuB&9mr3sb9Z!Vgz^C%O-%LAcftS>0oeX_2_m4k$-~3p+
zd~xY%+F6%N;qhwT7l)VU>;L>e`@*Y^$vYP*59c6Xg}AgX;zB-<OY8E=-*Ry+f^3if
zFSIT|q9npQS*zfY|9@y*brp4&aPG<T6{gT=ok&1MqzCR>Pzy;JQCJ&rR7`QSo@|;`
zbj{|%x}uudcHVcy+0$4G;E2pZugd4I=D%G0pRlg<{{idLB>fLqm&|{Kb#W)=J)zh8
z_yqWN!m2VR2_=cI-%LH;%K-@y33`8{|6qOYg2)#?1X0W_6UXp!wtG@)GU2bjj@k`X
z89ts+uPK953k&n|xB&yQxUl7Aw+=y&Djkwvhk<A(=dX$#O3u#o)k7fig<Q;gL53Xq
zE~X&=o6Yr8H~p4xZtwrk_UeE8?rR1?+Qw_snjH1C5sOL1hp~#>(0HJ&gM9D(Fj_e`
zH2&l-COLaF#H&mY<NEfI!YlZHFiVgB$S(Mm8ybI>Nhxpb_@9_1%u)|c?f(spcN^dS
zUocB!C@yAcwlxXN3hWtm28u|@V<ECGY|OViHBWZ9Sw9`27gzNqUPwC^=>&&JaTpzW
zF;a@r19jzu{5C<#v#|d`Sz6#ymVW<lx8whCg@6dj=7d-Mq`C8(@$XlUxRe(i(&5J`
ze?8M+mecW{A1b>i+yF^@81zQfq_rn~sEW5{#nNYY>*G(b5Z^%Pp#8z_YzkmfZ+4&1
zlQOAzBvAPCDrWlTb#`-;)8tHK?Z!yk#)L2$_E$y>`H}3u=j5wVH(6qJF$O)J%l8uw
z9dF$ABeT^6Wn$0o|M4d8o}_R3c>AXjqWXFaC^8%FhlniB`kO|{rE!$!mh=!EW*Qxm
z4Wo~-^za7DL`a2IRX^^|y|PECPMSW=shUGxZeq<f4cEtqz&CS>5O(1OQW*W5?b4^D
z2K8d=JlO$R6o2o?hnah-qT`>qWKi6o@!qcBQ$G(%qM7F{<^Sx;{~i2(No*Es7!;Gl
z`(4`4<3!}BNPm>W?eNngqiy3ypFREe-29y;`kTe6<HoI<$GKcT?ntrmG;>8r5@=`e
z;x?wH-yiSQad|xVD_MpJ%)bANolW=uOfoipMbA#kf7O+y{o=7h&h6o(r0BRe)zJI-
z^z^*qJPb=T!oM@QxN3Q>Yq+Nq!JWfz1=hyo?o78Ks#B_4VdMZzbxKWsej?4i(NBsN
z9-`#h30YA$9@4^7!)b%^$tg7iW-tgqwYTb~g46{By1P=Z$>>68?pKa+9}W-eAHMBr
zy7EUz_!HQ*g7;aTb7v^I*a7w0rYX6Kl+D17N_2vo9maD@e^OH5Rd;>H1F1x-tI~Po
zvy?dnZVtTl=q1vn@}vsICD@rX953HqOeB9*@{Z3&CuSm0+TYr6W}_X(ps}`t5&Na-
z$!%Mz_(4I8_-bW<EitohujbuvaJJ<~_)t<ubCA1ca&e}hA9Ml~x~>@lL*rVjmS-?t
z^3+g#4|0*JcglYXpnBniY!g`wKixVia1*|CZq^{zVDWC_-7ARsoB78oZVUI3UW2kW
zi5>#_f|!w%9tAjWmQ^F%iH+P6yaBJh<F?D_<GrNi_~)D~!saT8D=eP~Q$)i)PSm=#
z23>L=3)$GGTenpKM<Bs~tIilQet*hc5-jEtXKb636qge22t&XPBmA9owhPsTL)P=#
zylOCrMPJvFq9V(a;8WkiqHa*PF(VYQ&&S%fRqqj2XgG?c1$WK8n?<U)?%{=6xKPL7
zg5De13)YYPZuR>a7Wa-qgQa0U)kYjK@ldF5G`!2D=(cwz6Q&h%tEYa)pni!E^y1%h
z+@aKa4?#`Bdq&&dC3AX-DEu_kY^u7`y#V7xBGtiX*${>L04eyg3sW?o905A=Dky%;
z76R|n?0^)LUb+KGB~_yB_X)r=rWChHSO@X)Sp@RMOOYF)l*Y6kq%B40iCb$-3dHXy
z1tPbeoS#2>42Uc}jM;v%tq^a;!wnFasNOU@O^@S`U+|5uZm2uUJF(fL2FG$e>Kw<W
z=}TAHx$KyD+Xjm6qTp)dwvCna+Vp|zo5REW%Z;wQdiQ3&v6o}8EeE43uX#lvy)qSz
zDDj5@%e1}>{a^-kS$c_+$-(c=_jH;{kTush{^efZoZAp=rj7CZ5u&tZ=|S(2W@`o(
zI(1KN+tYOnA4jSaKD$v04xd=254DAg7ZRt8?FJh34Ct<*F?Lzc-}yOAaaf^&zW#?h
zp0R50+AVgsxd(-opKM*ZT}WY07}efZb0g)*>rR?4GwE!j5BBf5Ww7$u4{KeD-|ll~
zfW^BpXR&Radng2+J#_XB&K;?QP~+E~@taFA-BYg3Y3jz|Jx%MM3T7b-K($5RkS7y!
z*4_2rN$f?@S;YX!_6+%42b)6QD}_H^z%N6Q*}MK*{OFC)TB%XYRO#=A#yl^UfN9?s
zM^p1c6uxwInH6er0-VLraR!Yw-W#-l=*m+^JL>k#NMpsij<Kei+EdPPbI4v0Oq6=x
zA$XUtwB{bC@+|?!j&9<TA)sXdcId&z02bwl>!}BS>**_(>|wzg9ZVorQ^q1HcAz%<
znYyF@hqgD3W_#b?zkAL;r+a9%M_aVUQ}(7cR?R~=Z4Fhz)`$?ot`HSOO+gf=ySpt?
zRMikM>{3BOG>AxIXlrUoC=o#tTC<R-A%?W<JHOxm$-VBU|8=kBAtB$-%GdXOy<gXL
zF-YW6aU{X&O);kQ;O`0*U)#&qd~acIHn??-t!r<X4PE}N+vt_u)n=t5c#%`z9bFx4
z+*N!N=lit^|H~66Cx0%WTY(QRR(!V~Uyz@*vblnLy+iNqjGHKD(vDZYTsb&UUI&}|
z4Zi0j09B7Dm>3YyDuSek9xX)SvW=Y}tkIA2`)15U#wB{kHz)3v_M$mA1K879T!73X
zYRf&)878`A$PgT^R@P}7&DyJNP0BV3oVNUZ{YK7ikLBT?yY$X}N9TH`2X$y&dVGt#
zr(Ikb#(1;vL_Jn^%MjJ94vD30o+A9iwnW@Ywg2mwMOndBL3HuFS7z6<qyGlw2mQRC
z_2kdrM9|0Mo5iP{dij1AO3}D*)Lp(VK7j-5IFwr-x-sRxGqkm<CUS>od$!T8)ca&4
zF*;2pY`?pC2dKOh##c_%pv3DhC4Pl8@$iL2u9=x-6r5j$qgZ!%1BoGeMDpRUD<HN;
ze&tDz+58Jz+%#>o-4mnQL%4$zxcpRv^Jmkg%9W0iJa?n-=38hDt^-Fk96;B5+}8^O
z1KoS}EICJl2|jWNz`gT$zmHsJ=I8%)Z0pBftN?XQt3B=Um0&{BSJ8d>Mrry<W5x!W
z7yj5$lF{0|VQC_%|L(p{M$ASwWZy{e*Re0K)jxOKIj(;l1HWC7&r&9DGY+@-^^g8P
zOYL;~(V__zGG5u20_TbfDx!lg!M(K|FSNjZ-`Tk6W1n>UH`m+lc<;~EA17jd+XDi8
z3^nGx++5rMafjy}OkAIIJ<F#^?MtY%`@v(VXzYZ9E1w0II3Y*1)Mn#emUEH`SH}6z
z0F47klFGD`GfS(ao6FHh^r;r&WHreudch^p%p*SVQ}7C<&RcVmkSTKr`><qPeI~8t
z2gaOtubGap@~0UshRyGTsTU{8m-+KM7#uL@z$Pp<wF)AgwCHnc4{b(kZ;S8w+U3JP
zJV)w%=2+jfn)Y`yYTtqEgpYl@u+zQsd{*suwko3Mi{6RbBh$Dk%cZhFBgc%0DZ7W?
zTe6RE1gTwo&py|j?Md?1Bp;=jK}f_(u=R4rMp@3t=dOXVdm_B7YrQjM_$5NH9VZrR
zORbZl!LH?J%Nx=(4>L)_SdgNHlhn5OqeAO7e%tUp1@dnk|KE4NuXgyaU-#pwtGJY>
z)0cxX<IhCjyr%6e2@i{W;ovfTv&JjF^oOoj9hsXteSvvrJMIZ8j8P|)7QEYOyaU&G
z&u&QI)W=P!nj^{FhNz1mLDL^$pS%hvAtIvYy}yDSTPMY5WB4D@gj0eZ@})8(y%nFC
zD(kf&a$kD4L9R`9Q<#8R;1jD34&2RL|11wiId$?z#+f~ZtcBF4T-bNR3C~>=O+NM?
zuTx+afP$EXzEgf4S^5b!o_5)S+XGEybN^bqQuqvFDJqSFhi=Hj(eYzg=uW&mU$ljg
ze5d7`3;*lbM0dEfZLU_AXkI3E$k8%TQSJNcx7P@Ap**aN*VB{}gNyr|I8v2C6&vJI
zk?j3C)`|+1Sh_Il@6t=d{1Vf&PH=k{pq)16fn<h0xs(t~mS=V>X@_KWqx7<fHf7v{
zQo*$!puFy!k&8dvdQnqH9DDKdmobaWuJ@c?yuy~|xQ=QYhfmrc<c+;-l;^iO<OZGL
zf<cPsdKXzN)~2H3d3zRXH#2t|P(6DgJ5B4$oNhQqaw&6z)kmJy+MSL6t=(dM0Kvb@
zl8zWpr|*eq0eV*kHZj0!e>O7TM!gClJxENn&EA^PG=&?_=uDUpW$HgFQzmn)3qtD1
zUGv5}rrouAxm`BNjW3has`dvyoeF>0)(Z53jPRkkLB~J$Q5!({4<5jO<$P%pO}5=@
z^2Vaf*A6USO9GjBW;rT#G+;tvQpGtjpq{pEQP$YfVT6J<H^BXeDX3kgK?(s&i)_J&
zt$F%P4*{+ds01TiZjyZ<Ac=Zeo0{=vP~H8dbgqk}o82johAESwtn_x$<i@%aZ-c+C
z(Yv=*5=<+f21}wDj!XNF2LXS3)G=HvhD9W8U>h=ehD_7b3%d6NsExxSOUEPBoNeCU
z{yHXD_LTE&qV64j7F#pT?EH7nw9J0F@F;RaaTOc!{bj7t)$F0A=WA*qLE#OO_Ypri
zAa16;kM*6q|C0~;>A}v;RJhE~Brxw(no5*&qBFh(-z-MYWgtjJ2}W3Fq9|{F9F#6K
zA89s*XihX@03eUGfL(f~h_F=Y0TyQ?hq#2gtX(JeIo7v`y!cmuJxN<_DAQ{~V+hL8
z`6bD7)!^{-{3)RAPG}0sJr$faS7<Wi9%J9D(j@uRiTZ}nDp18?e~dDxzIL*w?IeFm
zvKr_<zw7$QFNGW7gG#%hCoO6O=Dp(9tS~$)aSdx1+Sb`V0rY$q`BS&|Bb0auO<oS{
zZ-FKB^{36G7EO%uZONbL$g6tC*_zdwkt^5(f0f30IJC4%x2`HiPWiWZ0Da_DFWy=$
ze4MZ;dIbAKrPCfBk7?>XZOlh!hVQ9R_aAE!kRR$NnXd*=QKnyAe_NgS=krzZnY-FT
z$dotF4`E|5<BB*8#h*C7<a)_$z0-d)ruVO7*X3V1?D%(S_?4-u{o3x`8Rg1?#<kQY
zsb@4x;Rfh!|2XB+d%@SdTLQ)(BFMd~@`u>6=(BEb#x{)>r~CVfeN*rI`*>{npZn!=
zMgKr)c-;?OQFTb<kB7{yZs(&+{3)14Y=O&hWfz4Kib|uP1=Z1j=ShGJ5?x7dGm1YJ
zi~O`r)A(=Hs_Lh5*hFVA@97BSUb(}A{BT;r$&c^RV@;RdZ~SbQ`}P%h=(#$fPcQ#w
z+Yp)YNX_N<TV3@jdLM^4Nqt&wTI%|fXPetJKYUrcJ_X%fe9y65I$j-Aje2U&)^q2G
zA@vL4{R)=#!xN^p+|MiRn$6l>2sjm>mo=lbqNaoFr>GK|h>`vnf~JDXc8niZ2{`0$
zqe8^W`*=7$Q+-XH0+PT>J=P(PG2`fHxouW~9<yDGySPqI+Aw;)YV_;96Vxy(7A_7V
zLQTx0O=s4ZJQ{nJTmpwYk>iw|(__?~2H7P%<Quxyv&z`(T))f0n;aca`iSXQpD<(z
z&1i2U5OntJqK&bs1c7mo8>Rq0Vw-OJl$|?4GT^?9FavmB`Nxo63cif^``5?vd~=^2
zCFm#FTVkt?o#O2EMQS)a^0eXaV(ib=vBg1E?}-KKf?K!O5vl=c#@SDy88rs3;|UwL
z*j~S~X!++cM<1qSxdm}zB@m*+BhXG}>1@qq$$jS4-W)UCWe=`7j2;LcFg-Q!#t?7S
z;XpwMM%S)LVYm9_E*GwfabUH67TIP=H);Wzow7~P`J+qX8awUPDpVVkE(bKt_P5TS
zfSrJ}JzLX*dES`Py#dB@&5l=4>Xm&>x+7y13_{K&RUoZ`jqZwc>|d=qKI&50Fu9P_
zpYwU<4JTT}tPQI5Zu08sF3x+9XC8q;AR2K&E@&8fGd7yv^S0YG?r`uDaf8$X1jUKZ
z4SNgZLttA%LwTfMx38SzwhWOD8u2jgW!T-#MrY;5K8P{c3N?E69LT2+0A?j{%oKrr
z4`Jpb@#<%+PLNftlB|B=1b$WJ;7>xg(02KA%~2t~C@Zg(r4FJ!nOL-I0G#4_Mm$-B
zSZLly<0)heEVQmbOW=4T?=&s=kwc}XYLvou?c^BEm(Qg+ST@zpi!eb=k;=QlkJSo(
zrJyVnw9Fj6I$?hF5~+X|uFovlizKP&Vr|edezctX9uo<NZ$F=+H$*^4@87To-q7Wa
zxeMpzyV}mw<Lr1*T(p0uhx1Tkp9L6onOuS`T4E-8!~6~Yo09*fWI9Gi=g>U>D=wO+
zKE$Pl@mbLBSMbAw^s92P*iz+4cKCeVUcRe~H)@eaY8_;Noe)zQdtS^!dD^K+?;GLm
z<JEnNd^RcmQhSuq=D0m@kCHAY3h%po`s)}yTt2|Ctg0unDCWg%_tl`_O7@-elJqiF
zhf88+eV($;LEEH}Ipji~!hp|Xg#oFyS>turUlQBmZQF(z<Aueu6vLpIj4G|<yLeRJ
zAQeFK{lskWe&i(VOrLo;H=+5Bi<%{5659zA6HMnM(UPsHVIud?1MH6VCFhup*#U=!
z$=dx%ALmt3BS7<Egpm0P&hZ#gWJ4d>du#Fwcm6sS$)sGIQ$wo9ZvNh!S8tC_PK=m2
zIj6h0AJzH##`{H%^=+fr15VQEXNVfjw+gCUh2&k$b&rp-i7|q{#2XMe3H3?;Bs()F
zoSv;78B2<n{Dh=bMHbzMcGgS{6Qbvk$@@naRv9OSeWg_;-JUgM@6_vu*#tbU!~*WB
za=8Rqo1LWd>t=DtZ+AQ(g;~DBT5S@DJ!HWn2eVemI=Jfb5cY+%Z;e4~^I4BRER{o}
z#mLVcLm%98>|#mLw$o#`Zsar>er%QJr|w=ha4YpGFwV|$ah+S_re_U`Xs-`18&uU3
z7+27+1h5AS4D?x<xd^lX4J4N$wChSX<mnvL9n6hh&!*QQ@De=Zm50}h-i4J>Q^z0o
zlM=t!gQC@*m5*mg%F&%i&K$>s_=?k!_b<wYCSbDM8Sf~}IN;Er^me)4befhSfhHY!
zu{@8yZsuyH=H_Kb{dFuE*Hx%vJ8lQn+iCjYHG7qE81W=O)Yt1H-2Zs_Yu2MhrQ610
z?ZK)Xiz|UI-xq5}D`A!M^NoK@nO)f=Vchp&N&cI(w+i1Sd(QH$F*%)|o96Fcm@ew=
zCPb99$`_e<I+=$m*D;<s{tMqWhkrKT^CbZk*Z(pnNMpKT!OUU1e{%ZJnjeseh{(s4
zquEfD25ZRI0%I~uw=~`O1X4=V6+*}$a4B*HZ?#mqE+B|aVLupiPn&^=j2q7YbpnT^
z>f+yUK4)SRxV8`Pq&y7Wd{&s)JEYq8>TcFWPk9Vsv`bJ0q~GeyqxThmT)q|aF{ei6
z)0Qsf+?_sp-t~_aCidq0)ml5NXn4fehi*);x8>3zJFio}6ImLOO(04*^k!7wS+FIA
zNW4sFu&I+dilDHAk;$(#gA=;B2zG;L<IVy|aW<2i&`zkwZDoi5U1-0_jnhWJjqW{|
zyV$Y2msMA0p7i<W?KGEnq^&&)OZHYYmGTHIQPJ!gL|CMxcos#rcm(a8NBg?vE}nU|
z<TR8OhTAs!jql>(7$%FtT-{fK{H_KcYZ{&5i4>ZeLRX5U0D4}G{%{tqc8zy6HVVl|
zKc2uVe2J8$H##?(_>4t^J#jEI7%TOB(iaK09-S_)S(#EJIrfN`Asn;KON00eR5oL)
zF{qay9NgRpqAJFisP5Wx^WnPWnBHCxEa?){!%t+LqCWZN|57Uc|KrF1-$^4Qr_aK4
zxc4{i`B+)Il!!FLejP4?Q#YV<?LltOCa3`UIG?uS=~>r^Q#z{};uBu+wAO9G;^GtK
z-)}VH9`o+e-<m!B-i>;OY1<RmVf+qNb#C%roB%zmjx*MJNkrI@%_B4KRi;}7avG%K
zi_wXVMqQP_pR|9p{=a);e;b+o7x^kCwFNMFZB+2Y+01^r=UDZh?`vSULw5@At@{s&
zCaEFj$()xCUyW}>!)^@4M!J(^ug=`W7_7|8>}WBsf*(!CWUIPnsU=1KCbE@}Ro*p7
zsBta65ZS3^*<nx|`M&=>4M~gvyTZ=(nVNI#!Dd(9@%jB}c43iwJL=_Ut!%~E^Sov4
zelXTo^c3%nzMO6B4fZy7+LXK30P}yaS5I|ERJ5-8mWw1N&v@?!5)Docm|<=vp_Qn6
ziUx7ex`~hT&}%+@`i8p~UKG`sjzN64l%IR&D^_4P#5huNV&@*}9S_;x<Uhs4Lio^{
z<RrxIFDcWOiR<d6n`E{*a7al=0sD$dBXCu4IgE9u$UMhO%YUp@9DK&u#klrSPI;b6
z#4266xe+99o%HbNv)}+@m7F|!z3e>(UPK^IYdUDk!Yh_!u?1nJ&;4~P7Y~dTpQTI@
zCfdMVX>&Bt-RS!pln^g(2bH8Z8Fbwi*$aX#jW2ujOpsPDv+C4Ks^IZcdv1*GK4PmR
z${%hQ#*9Nz>+1Khr9sSEII=V?X&r(@QZNl3BoLYnvVhXfoJPP``!P@Cw=<8CVNTpu
zqJCWKH&&T?>0J(r=M>!A3B!&8V<Ac)AV1OB!WCJawfoK|*2s?|;GOF?(+?yT3`fy4
zYd#R|7{guavvwvVIP{IA9W}TXa!cYLo#Yhm=yPWV?=E{)GNzE~GR!YsYN1hWqetuS
z+LEiQMCbhJDvhA>nx&w(g9Ps69KLO}J4EiR)7EAr6Roqt0zF9n?Ilx1y7iReyl{OM
z*Ms9T-HpPxEfK>zEfOH=9Kc*kUTBre8vF0}mi}y~R#z4wi<3b9p7o_s5s5>aU&##6
z>xHu>1g~pLjHO~Oqq06$L#a3<Gp49qW_qZf2z2}&)H{j}^JC;{8mF7{I$mEQ$8N35
z$(+cl=y+!kYS9dAHiKINJnt>;K+Bytno_>fA~H6P%`YN5+lH-Xl7)%hkh9zw+xiMU
z{=45ok(#HKDPP0m8ZziQZ_$0jIc#Kw1OX@Bk{O*NKQ9wKzmGQ|TEw}bbN)un!t921
zsA^Zff_<nyBiDs4OL@k9r6I_zY|N52aw}6ZBn>3s!LSO9wm?uJl!a8(i#O&cGd+^>
zHcmBE9ObE!hAeK37(SkB&Sf%MyXMSMlD5*`AIe96`2hG~J;~{@Y?&ujuo{bnWtnzE
zyH{8{gwcdXFaP1uwrFcxmr8v75;!LLQy|^b__zhDA-j=}4dey-Qv*3p9zbxcN6K!m
z^^tAV$p-BCiDs$|Gatjndi&xrrHDn*D$NN9f}9%wd$Z*A5MhS?ZB47ge|x-0pSf1Z
zN$Gh6S2>8DE#JMOH?m!=PArZM7yRn~A_iD2xHluYo1@}?QS$JTtmfG$jQ*!cjdX9t
z4(>+aV#o^<37y}2R2@!ILOO5w4O41AgD_HW^C_x4kE@-It}CHAvE$Wm7$O#@IA5B#
zw{+_U-u1m7>IR=<Cpwo!f@hfK91RiiQdNC6ykkUB{72wwx5ls*hE!LgTs((Eyuw@>
zf67sQL+J)2%uJA*M%KHwP651@3(K+qBd}u)hShta%YPm7&QcdXTOPN3$@to{6N>>x
zMb|K_$lb4^)u75Fe8$Ym1Q{;>bd}*Tp0k`dYP3toeWa0dippwFZzmXU&H|&`v$aEz
zA!4h3HkK%|Rt?hhtnm%1!DAm_ftbYUOC%l3!;Vx%SUlt$$O;8gf&l7s5%-^iNdf;I
zD_7>&FZ;#^lDj_^Rlt4eh#Sj>bxF!I37r1G`idm5bD>cAjO>*`efEeb&*V918t%EP
zJH)ehnF$CFiN+`I(T<BJ?j;y$P6Rnym(VN}uJ8EuR~-Rk&HboW`k7?0x2ZheVp)Bm
zF~Fy9)SM*_wk59ia46)ZP%bKq9*4J#C|xTx8aNB{R-7)`xU@AIWR-e?$M$*Gn)J6J
zv%ikH*c`s~o<GW1JuFnR8Mbals8lF#E$o>q+m?`z$a~`nvLC)QJf7UokvRZ#It-@N
z@_LPguf*<Zau=Cr4N3S`=;uF$-b_m?ODCO|ujOH~rNytYqaMAicn2TK5_Vqm15X-g
zq~`g4+<-l9f+<6yR`;R<N|#DWbId}psZ*@S!S<fjK+xWeYPFa#`vIs_#{mA(^Mq<M
zc;hA2<vm$Oik$91>)iuge-T$}GNP7T9tRa8tzZ{}e;s@LdLQ<gu(J6uJ$gf77glDU
z!=`T_CdV2?7mB`IS))H7$O#U`RO1D(b1D4vciKcmT?Hv0*uLRo!p;bM!-u*cXhbmj
z=}FaSNIn=34x%ff4;wJH(Hv8Yi~vk-@>Xv6<X9s!7C*VPl9;M8YGg8G7SG;VYO1K)
z0X-b<zbiYxWEWX5l@vXG0QjI7l*nOOFDyr8!JI<V>$XzX!?EcPW}XH8xULb|TP`I;
zKTT`;6H>z;kBQ*pPkXZnjlsTeNAGp|YLF5HOr>z-`Efx>i4?uWEKv8p(%iOzO|?rv
z?lUwCG8{U-z3;ZX+x;M}kNz|&l@7s><)D>4v?Z|DtK9qCK+0nQ3X+Ubq-tDlod7}x
zS+A$Dx1d=Se@wwI(<RJRHV+~-myR*(mr`qb_iofL6shUNhL2u+(4p}q6_dQ$b+5yI
z^1LkFcVYYrRsCBev?8t1>P$nLiE>8=UC!KFcsnOd$=j)MxsjyZh>EZz6l|zQ4A7NO
z8z7h&1$<i+4~(@4rmtN!b|%Ftua@Alh^ZqlT+z+Z8QA`zq@;S`s`(Rxp%LR9yT})D
zLthS(3th_-hOZ-IjwYH51i}eqrB+?eq3PV30wAYu$Sj$&;yzFNZ-B|S&dsx){A6aT
zwCxmjEunKDix`$>=UuDaflS9DgcdS-t?~wJIJqjMq~x%MLj4As4(J9QdRNV|>A0QL
zDVWW9h^bgcI_nfz>F>izbP~vIa~rdQk*}m#UEyv_tp%p00T=kZuW|b)fu`1<?{5x?
z?|-A#buqJezP`64kN@0Lwh5bHLqtGF0gEJB_;*P6cO5@$emAC;gjmm0*lcBWrhXT*
zgbKuB+G@ky`fJ_aj*RUHvR6KI$_scgSE=_e5m-)Ouu>pd4;$BR6(k;QL*OI9JmX+m
zR_fIW1ts_-Nt^zji8z#L->nJ>sf6U95lxp$GPg_auxn!>Q<ClNr%$|EgAG6AUF<i7
zwAh)kO;gS%W|Xl4xVH`GMp~bHe83B7R~T3L4y}`G<;HizZv1<HZ6#?QQcd8GB?V_;
z(r0_eEE?l-Iv)n*H_an?qH*$X;Ygp+xISzG7sVS*SjXWx0MkHkS9RX0%830!>Rtuf
zrSjzri(Y`m+7!d}PAvfx5%`HH4-PEe6dcqnRpc!9vc6tWLw8(5vzt32&kRaQ1{D2<
zv8^CG_Z9tD+n3Xmum2bAwHF1Y!$L0!Sp_RMQ>7Nn56v?2_0canqeHzTH^p$+$BwJL
zV`df#bb=q3dMhyv+MkbnYF}2@Lj~h$^<<MkJl9F?l+Z2Bh?tZJ*Iq_7M0s?gmw-$g
z+S&lcdXl@qwYo%G)6cH23W><Ayj*D`J`r$N0BkGo#2U_#oVJ+r#kC*ru3tUvGWoR~
z{)^Vd{4mw-;l1cz62j>dV0X#mDOugB$Y)tQQHQy?HtO29;j#{(md~M2J$aJH*h;hq
zY9{2wYe9&MV%-Ui_4nPt?92Fb92c~sGdO-t7nz;WVp4o}|54jsCu<F=AYc}*j{DII
zcFJyTR$YEi6k9TM=@s<S2Emff5KS}2E~&$<t+82$Gd8hKUSTfD4DI0`0cUZwM<;hI
zmU$KFiEe)#o2vcmSRc$QHc>t~SMGI@WQA|H_rG7E&kek)J@9u>`FxjNxezQe;Z>3?
z8q&1nyvB=QvZL>z7H>3o%vSzy4=In>p64b&MavTdN|M7v8ro4L4_W8MHKw=7f=APH
zut8pq*JV~s((<z!@<56aFURR6vSW*h-Ccn*y0bHzHN%%*I#5>a_R3oGX_xWAo*+rt
z?|`R7c@pb%YvEYMEEr<IP!DgmWxQ&lj^`B*G`VD)>r`t&&wQUX3ANO8S?e2b-pL%Z
zz^Zmy8LmQN4VQqj25%=nV(E560tjLj2n2%;adFWa+keKS`tw@V=#Wbsi65=N#+p0t
z0m%mrHufn7js$u@n{au?Ix|$O=svBxOoI;;doO5nM`mM-?d22&lRq0Gz76V?e_U6B
z?x6=s1>Bx|;U$NMrw*ah3#(~vx!Z5@oMJ8jbXud6U|)v5yEeqxmDIxB#@D#g^=NJJ
z1>AWG4u!XD8uby<EcLU9i}ubI=_B|BVV6#Hd0|0T7I|ifmTK9TqO7BKrlS<WRng^w
z!18KsdFjgC-2t@gH2TnTw&R=D#c>Ul!KDho)5VxOZQ-R5WsN(_LrO|ks3gM^Hn4<}
zw4V4A)u!F!FaA0f_(999@WhRQVAom2LyaWk0;;`x?-<1*E%FAf{odWN!~omI>{dSg
zU^?1D&>fN;K}%hQqCJ46)x=BL2Fux>r=kJ#RU01aIj%S8`SEIM#tjD4(0OiNU@(D=
z-##}Q$l`jDx>1-80s=nSlS-N!Z2^`c0jQ!#T}YuaC_2-3X%ldK?sl`R$yyxU6j7Bp
zQTRNyyk5x9^OL?Vl#x94%mx>5KU_HR=6_Xtd0$-HW5nK<Sx-J4bLvz#uC3V^$$z49
z>qCy$_uqSr$6cybl`=>~792Sl@4Qk`N+K>Qs`D!_(Y(3z@hL`Or$Im`KIW=Kqd|0e
z;2k~RbyOnkMr4()mSmEeRM&y_=*+C$w(ldqj(t=RMr0wE^21OS|4vb8yAp(*x5^Tv
zEZ@pR0p%YI-`mVuI>xL-^z)WTXm+na)^on8zw)0}y#EG_p-*QfPos@L=jeS0#Y2Ot
zBYpa0r>cGBT63B{+uVCiw;`Yty}y*1oq}7u7o9fxN^)o-?aG0DEdk|cs1OfJDH0%G
z(Mga8@nP)@!0f@|4as6dz>9g6W?(&%O3p!c`QwVhf&T6>W*;k5<dq{5chPRcCp9aj
ztyJ7H6St6b!%areSeibndy2^TKJEDZ`<op2Z~ISPo3p6}ZPJr$*Y8v%we;8o7XAI$
z)>_X{_NGYyk#U)D$f&QXujJQeS6LTQ==UMhYF@s~sx_7E#$|tMh&W;LEj@kK#hHPc
zo?EnV@?e{(4`|f3pwMEb#=nHepUugHHB$lMe}qY9ExcB0OgHhv6*l?Mxn0<ly8|TA
z$pmnjkYvg{^j4{V{!ldYpEq4b$zW#3x!LL2$9z%$Da#f9PDG-}H$kw5$<;Wwrs{Hv
z_DIxX1OhT!e2-V>fZr1wNpuT-1tLrC!@C#3uf5sFGcT3!9^kCz*bp%gTVOq96^oJv
zkP#7?OhkL`{nbp;U}}VKXDn&5LSPYdbrk42WY+D{Zb+O=>U#Xe?&RYFGWAu%d+oh{
zvMvLLY&4VjPHVnA_-)%IC291fsF+7b|4q-=G+769&N&Q??Y6m033lVJnv<p<uk^fO
zk!McT;A(>LlS$EB818Au$?OP1eFagw^c(R}VN-H<^X^O&t_q1$ebn*oC-G=Lu<Qh_
zkpt!{H(y$%1HF|pAegKQ``VK2#|L-Y7R(<8?X>(jq5i$WlNCG1THxZ}W;@ShFSmnr
zoo;Wxs|qEr{#Tx9eqlHY@FDF~cE?XTs@-$OZv;Wg8)*Zdcg0jJJXQ?5p1=;8I0e4Y
z{vr&7;2vNv_|9GGb|rN>NjZw@tN}{;T}Hohcg>wh+eC=|yKzI=I+D!CLv#gPusRP#
zMr1AF#E$<+Mh>X6AFOl{!fm}u!%DMOTD-1CM!9S2!|Ynx>$j(iK+l#AgptZ#@grG0
zb8=OHW4(QD_UL5Qa$A7P4qA3A8QX#${>q2?j&;)ZP&p6CD|>N!2xUz$+tHii;v&z=
zAyF+RnPA4eyiCo)+rBT*!*r=}-eq7k(O#_QsOMd4<~hiGQUL6<6%WN?*zwNqs}{hD
zFjF57XK!_{6t^J!T6&3yT>c6AGJ6pm^hm06qOU~|GvfadDYpN_y7Xzm{3yluW+9K~
zwaw^vME@s4diS~Qd&yw{!h9_`IywN35{@=gWrk25I(=?pw~ZR3dT3)kA&u^?Us$y*
zJbz&B<eeY#lw~(Q-R<ElzwKLH&dhEPZM1UK0&`wlc-W8|0PZJL9F>VHr3PhDL&{r*
zzyvR`ci=SLgX^(JNXgoA?vET`J}LTV$zqAok$-(f`sWl2iC)65=EKtoS^x^)&1V<T
zXMqFwyf2?pkSrDPM7J-j*yq*xD@^jcs{_U!CWsGD#7EdVj5OU6U@URu)xuffzGV!g
zR<!SHaniPFk~q#qi2bINmllZxgzy6~YvW`9e*W>>H>{zz?>~%J>zvR(`enZs$ON0F
zs*2I5J`3kSum{B1Tgf#dcXO#oo^1Zwu=MwWpBo12I>hhw4Pp$w5~dmE6`$MwVC|QW
zYJ7lMZ%EImt@#IYTbJE_61(Lnaa#qAYChww_gAXQgnGZSIq{cifq^tn-&4-4z)V`4
z9mKVf8VCwR6D#!dEd2($V+=I=f%Q0QK7#u--Y0>@XF6M5!ewn$lvW`C1^mfOt8X8@
zPPrJC6XOlJ4r_?`!nKvv=)qFS<5}y$g;lo|eA*o?y%VoF(x6i}(i@MSdfj5SCxnBl
z)<@*A`DF}5Qqe!aA|zF%3Q-A~CTHbM?d|TGp4{L3yc38A71m4p;!zo*Fo|syxzeG@
zIIaG_%<-qnzn9MJrh{(?&QS}*ggroK<d)s@Hl3EAr6;u^Zp(}!2={6wF*fa-O2o5|
zMpobHfXDBdcif5~ncQaez3Izd%f3|hbZB>Ty)SVD8G$dpoK374A24(9c$+Tmh_q@p
z_TZrUNX29iSF>{gRgvTnoembrV^ee1c;$I-^1;0g`Qs}5Wc84R)kp*LVokrk9n+)n
z^Y!hqqd^OwQ|l$`6~v!|){1xuZt+q})m27!{k5DgJo&U-S)<JVnAq4yy!+KQ$Lc?0
zi5ijK<l!%GJ27P%CZ~4MVZMzZq&_pQELBBU*e<Uy$|Tqgow>6=|GK4l^Ik)!2$YY@
zhbyzdgc$;knIDJuts)ZS{+~#?I@fgC?%)^KU#TRZ`Ox2nE%fvTMQI8V60L&p3#ZEq
zK27<j3}lV5yW6(A=vlD)2g0|1ca6IqF7>}{+omR?{Sa|#oDU^z3+*DqR}%c`jVN}8
z3m*#Tz>d$)3%C|SS2cIeN_T!B(9)1sT!_Teh*<>gS#Z3D`H()+W>OYZSzf=$_L*Cq
zOtG1=@b=&Z!sq;(w0B0GlE`&5R6U_&!)H@*>1n6oeD!OEknys!x9+XCc&90ej3KU%
z=PT~7a%-I31PjV1eR&7-5xQRSydFvMsL)LF(Csn9(?oO|9=F%Mvbt0BWlx?SwcWl`
zULE<LX<N;j_C$O;zU730?hL*rM8IZ|`;C;Vn0pZ|NO~39Oev6*9U>2Z9_xshoF1-y
z@al7@$O3`1L0{Rdr`N)0sF(!-LD){Fld><D5^GCqTU`^dYsPasdDjz|R@`qJ1`k&p
zy;fd@+I(MjRU0NqnY|F*wp=bX4NLMk(9rW9%^l&^(Q(Nqk&aK8XYVCfcA{tAOt(>D
zP)5A^>9GW?Ki%g}2oHym_fg>wAli2+s1RYGfT*jJxz}vmLi)U7>x-t;f#aSUIrRBo
zuz*&f<bRKuB0zaQi=0w2r!&I-vs9_SC&Vs`TW03Gvtc;mWj?&@L4LX8{J;;s(>Z#a
zm^O4c8*Awm_co@T6uP6H%rHpChj!KX<`6cmb~5AvT5gmil}zT)t?Bl!;gR$3MfsY}
z%2P8PrW_YfNmv7gpOL=>O#KHIMcxEB0F<BVG|taf<*A)XtY{x<jLWZsl*=j8&DG8a
zXf~ZJ#21UQC35+|R3wJ<j+FH}rO-0&@f$_~l$9eGS+%^7x}x>t_QR3x>xU!H<-6HU
z_kyuGi_RYRU<1a|U&q9$`x+c>2Iz()3zc500S~ah^o5B|9-TY>;4`_u<xxH($z&~<
zi%wLRN7-m()VIRigW1xl;t&!VlOQ%QW~O;NG)v8)dDG-yQsv*^IW2QQ&`A)v5OaRu
zX?DXjQ94;xMmmpXM1xE%;#0K+;DJIji<BE)Xkou&s9im`6nK8Dp~>8y^Y4dA@ZgoE
zAMn_*JqJgy{!$Oa*zo5)!^AyxS7teN_kHT0gAR?Mmf!WCz80g^acYuLoXOnt;9K*<
zZw+$pZEZv_i%;JkA{>#*!@YX4O^pHwj8&5EVLQ2wgiY|^+4mYA(&mOogcz#a!rfgS
zZ>NpjGvZ_%U&O|H0!lXoON(F+r#^;$NnX^Z=F*rXK85km)*Ryu2sOsQKSXgOVv`5>
zVIb0=eY;=mRdw2h!0`UylO9{PwxBY86?QgwE%T1aaeI|nO+ZXQz;``@G&_UXAdZ<7
zX@o4zX*5w;!9-gU2Z?b_Ah2Sbwp2rTe~G@F_Y0T4=TO+1z@X+Jm=j;>yTdh+6a;Ax
z5OzQ`>J1=-AoK_+m2g-arH_dS=o!bAf~x63ekp+`Z*#6@jOOYXDAwyEu|Huxq`_)l
zJkO<fa#Fspe_Y88ifoE?y#1~py2<?5G+^S&uFUJY@5*VD=sqYU--%v>BsBJ6i3`Iu
zm05L&`B7gBd;$-MqY%P?kd#$I-T%l#kC@c^zfR)^Za2A&nuG_ks+A>SJ_~mO2WTQg
zIgUfm7U8l;<U{i1ZFz+|+)~n$#*drI{m!{wb)=AG7^6KjcSmXFp#i7FUuj`FU}(vX
zz-Wd9Yo5LbpU|)&9lD*d!#$MIe>gw;ru*3a8#~!T+lrAdJQeq$iPoI&b>fx+xKwvr
zPk?QZWi??nDUz0-URqDbT$M)IMzkn$!3=WXlD2@^In`kfD}DgK%`*-S==;E`KF;;q
zmxe-Yl%~4#J%9{d-O{umiw~t8RUtZjzKYl+8}V_krf~5*3LBPkK`?iI+XgUD9T@h_
zR;4`8-<9YZ%W5s)x)%y>m#(_RNUuk}+RU%fH3`-@{MR?cPTa>GV9y-l_=7BhL;>2V
z;@RZJqfjHc;Q>9&a<EmjAI)Q0etCGNarJY^OoC1^q_yKFUyWyCaN>F=F78OvJ}l^|
zTu!LR%CT>6b1n>|1m^jqvy$d<yCwiZ?7#S@1}}eM3Zn+aKOIjAj|SaX(JkB7`${e{
zv?)mK%%;PvjS}S}lSZm2C>Z#o_!>6pO^i$TIl#{r8hwOh_>=0Tu;}~tCl8_J!G4ri
z-YyQ+89G|3QIPcPIVmV!bsYi(tYl_YTBk9Y_wU+i4=Wv_B|v7%UJP`gEDE`Wh@?0x
z2}Ietwr~O?8MnP5!Q{N$G?=c~(;S%`RVT0W2yFPNg|3{Hz@D=ohvYAtX9tcjD9qoX
zFpbxA9qUifb4El(19#?Us|K(^Ic>Yt*|T3-J#4LVFx<|wQU_`ti5u5=yxh3)#;RgS
zwnN@I7s?|Mp(`y0JX~@_4z#<GAr@}>zVwt=s;;k6u)dJ7*vS6Xr}Xz&fJQvv{-Npy
zL~ClU;{I%rq`gFZBednpO;vv3%vl*$<zjKJ?1up%i!X|OQjcrd?i-{7RHZu`|4!1^
z6hE)%1nyp;1pl`pyUQ4Ja~^fS>Ex9JkD&mU-GaP^VTwFK?JR1E>hdz_xar)Xl5C$?
zZe}n6w8zl1wVT8Prp%6Oe7ttY35-<0OhlS5Kyr4Xpj#mkeRy->TB)wBuKZq8Tj>y<
z_e)P}#l?Wona1OHp5G-vLKE8oeQ6rY)yRUL(7nVmz3@KC`Q7K3ct4S9FmyZkHmB?v
znPZ&hm8Y3=q=&IaGLGlWp)~WK;OPn9I=$R$Z*tc~yQIFJ57o+t5V9X3vn3ElE+J%B
zA&41By_S{ffZ1y^lLt6kB^JKubxZ&y|9@4NK&F$R>QeR+W}cuYx3E?G@|$ERyXd1O
z3<pm(ZVgMh9JJ{7A&eUIJ0<JF)6=$|6KPg9@x%qGci3KZROMGvB{2~YFxz<V5$BWF
z%e@@U{M1xG!I5f&?08V&%QeTkcoJ9oEn4@`_K)A@L;%Z%*(xG=S|)#aDaAgK?Gwwk
zXB_h9Jb6?8C6lV+^`Vpt$=&WP=O>e{UdB5x7Aug%tV{x#2_EpOR6Hx*DB6A5RnCYp
z^OPjsaEYV4Vm~ewS<B%gmuF|6OsD*H4C0y_dX<qq<OWOqVlNp)GV0$SI!|eEH5?&I
zZ}03ce=^Ao%H`8MryPC>SxqjL)YD?EJb&rb1GSLg;8#l`Ojkh&V=D^ed@i_*eb%A0
z!6Olb!Y1h=j9oRlIKpYm9ZnMxqrt+KlpN5m7=!ae8Zh(P+^135bEO8j(t51FFG7Dy
zzIXWR*h51_hJ!KQWcoP4;??xqqOh7Gz*BBm=8jgT`w;T}YI{N08Le9rZEgb|U2Q}N
z?A%C0n8t11q{zkdi`U2M0k4}1`1thX#M)?9V|R@I@W)6dQDO))(6>-<7}j(FD&}B>
zV0Im?JkK1qOrV!m*>Q}YB<$C;?xxqOx==GLXF}(pL)o!k$iXyG3^*=ZZvteh5jfpc
zdVfB1uMHIS_5DZKR79Xg$XHaP>l44gz|sC4XpdjxNRqePr2^St>IK*sUn8RU*Mq|w
zE3#JfNx;0`!&Ls8oQ;Cddztdxf$@Cr8;;aqqLqDpqBiS1qa4?b1>2w_2-zf}AReZu
zJ_@L1Eb{H}w}cFEXSgd_$}%W!@*|LTk(4-;AMq0XEUF%tnYC3J3EETZhrqu=xgA*N
zM0W!=)}n9T+r4$i;1ACatF{f<x!wOJO9Dy{07Ji|Cr!E7{1DS-o=bgn7=Sbzu297M
zD32-=zTU}G+||UjT}2O5Q$du6sUX251{DS|FZVKVlb=|zTsL4uuzI?Iblf3ByL6uV
zYwMmCHoy)kI<VDk%5Zr44H^Z(;Llr`@H0P9VzkFO=+Y|vCGNCt8i$pttyWAs<lH4M
zwmvyMAvPZ@vDdLiE>4?$$hEXd`B?S7aWmy`*4-A8c-;|jSy%dxjh+2Yrx)pjt>Yd0
z^`CP^Ut%&)rxlW@Wrk^^ZIK+Ys3_sv5azLl5)$KioSG$yH*&VzaQ3q;G6J(rHB@w&
z@iSRV+kBa%lHU3h5bsVDCtn!FG+>&bMD`T+f(iwd7(gnmn-gR6)eks|kQc4N0sH(Y
zyO`<5(v_J%7jb*lhCi{1czO5c<Uex`O%gl*u7OQ^8+*zt-Yc9=c<r*LH|kz)IU;&y
z8Qv2!da!z7u4{nbby2KffSl#0*8bFSs4GFQzI!_)2Yn25PStm3#R<WaBpy5{Q;r#x
zhUjlqOc>MWl?VCLZ76JgDDqtRZ!oOQxoM6>NK@jXuoO<5{QBkMUmV2Fmh8qogZ*1z
zTo9RF@<~SjBkUgF2az)1d%b~<Zw04{W`~E)spWs$Bo-Ple!I0&EXuDq^ln-Il?(Xm
z7)2@%>qsIjP6kv>Sd{m)N~6kE{HRX~To6Vb>$ey&Ah^E3Sw$?TjS-8ciV((y6F0fB
z7HqTcA#aF`*Hrb?6hCTURkQ$a1usR&uWEMZ&d75ss$FjPY2nY~_qW}z3R=87&#eyM
z$^UeqRjwKv5eZ&iGK$eJ;tlg}ZlzYdJB@yqN=Lj-^C8%twvwK>NaLIB#A7^SgfFRk
zT{B~rIl}LgR#aUUN6BF*fqE3aG>G;P?)`>*WPv7hq;lb*P<#;*M-;|`&)SdmPuJ!I
zsN5WH*=Z1zyOecWndyD6kH;kPU%0p!O|QdN$ffP%60-b;GG|S+d*^iz7|Ka=e)~mu
zZQv>+<Hpp?<=iD7JvjXDT?4~WRaF|o;G{R{-krx|?hV5kFFR!o;j(^pS?i?;^fG_C
zXjH&_60v3ur8Pw7#QXB68<{F0hgeJzwlbfv?2R}yJvT%3^tG6d7810Lz2%xuZ*zJj
z9)A1nyYW~Ak<zkhx*aPb<ErZA9d6iN!vEkKK*lPLklMSg*$WC&Pq#+T*%&)#2P~FW
zF1MMcM(073G)wqHS<5+w2TCoeMWEXly6IPfS*O$EqnAqzV;zytjTH0X(aKe~rj`G3
zF3LyiPR9Vw6^p2!$Q$Y>AjqC^N;Covj^p|g<cFprW}}zozrQpb_xSSQ=E!fK7ZyT-
zKa9`q8hxGQg_K)Lefsz+`qa3~MEyCrv$R5Q5T4~u4SM-mSz0*pajPMtUHtDJ3YHh>
z+qx0Y&J`5dP81rv;+KQ!*VG`$bBIFpvyV>&c9R!y$u^1}p?5-Od%MqpxgJ6NOc<U5
z^qwY$F!bk|qMSUgib*eKs7R^$4vgkJXp^f)V1y9m7|W(9EA?Me$kOl0aRF02pWY!0
zFNRppiJZfWh#oTB^O#cThc_nP#u!;m*j7;b(r(F{Ilr9#Pke@pUXAZ@&7gD_Y>STC
z3DqN*tZoLHme}|7|IiT1Ddf7UY(_+`)JsuO_@ps;<Fj<FNbFr$&xS-+3DsANt87Wc
z%e6R>Se=!$n^H5Y#|1d{#x^`LY#*%qw%UJ5+^ps~-`Z4cGA!~tV@TLD^sPyJ9-3qx
zj>!}s1`O{6xN<K&{^Ksg6VctUYO7ohR0>~%Rva&Q2;}D@x&EcO;0Ajr3P#$E9<^%n
zRuYh?=U808%i`wd=sWYFVKX?ab(=}3uoO<LqlO3}D4_WOhH#mXoGTbXdAt7}yYhLh
zj^3|UJGtavpPm`t(4^UOw_}qNfRz1p%poo{IiWDjA$1H>a^-7VyoS8Te7xF}lQL3w
zEoVV)1FMc4OT8;@>|sdDweALtS@u66p)`D#93*ewce~)LK7}Ivxl<Hi9r;ic8%t(0
zeOTbQc0%S7GNkOZ+OB4G`J-C6T!AU5EX;e9r?g6wdmR>os`~1Pa_JxfU&v7c{~vNA
zwSIcFM&?#~VX1zR4hw%7ZPR<zeqLw0D#{=r*m@I#OY*HiPq-HHCOsvWypjxc9l`RA
z&b6YW#!z!6A{qcsmmD@VTS}QKiQ?I&RkcVLT48!wU}PzQRtfif^S}58e2S!}&9yS-
zo&^$Z$neb2Ly&j>=#JnJVi{R`7TEJ)c#teS8`nP2M}@7mIG@}5pOarI4M`!UUbkv}
zlI!p`kqLU8_Vc@|?CU$QI{%X=;ac0<UWThk`)x`*DQ7vSwI0C!yNSCjw744-#RN}U
ze$7bFJMCtofDT%mWmdaQqtUyZu{dgDh(vSk9cz&G@QD{p)DXg_0+S@pj#D&`UhaC4
z)tEhEN_DCB;VU^1Wq}acC~wBfWfvh0g%Va=&XxcJy}cX~y`Eo7?1`HO(1e%S7RA5D
zwLg0N37iuUjL}3J9gf5S95j%2zgMV}MHPDq!|+!eN1r0ufFtywYVGHyC-{QCv~Ajb
z`)Os0i3G^O1syL4cQk&gerwBUZUF&S*^)SohE#%b&>_UC5M*}5cKx$NcKG)(5jN8k
zHQ>Q3_l%qc*^L2so_03b8!2ZudU*iIWwAPXDXq*rAm&>H+K3Jd>e+DzUS+L~cEyQ@
z%+s@8yX`MtAY(P!J}mDgj;{iSe&|?xg&$X|{cs>>H;(GIj<b{NS5B`ge42SH(ENEL
z+9YyboYvw}8Slfal{^KwUd>Jf27Kq-C$&a5g5q|#<#ltMV2=T3I(gxUeTIVm>ll`A
z7LVMAY8K~L2Bjy=JjKNP25Uj3xd7Hl<q&;hL^ame*du<~C*fcy-b#&`c4~U!c_@@8
z&wFzJsc+HSsdY#H@zuyKns!Iiz#6lt^dD7)zi1}=yGR~YH`}K-Uq5<TqmcXs`!4$P
zJo?+%ee22OeH**VfR_5%w=MPI0i1gx+N%g0v@KLH#8mP|MfCJ^gWF5PCjztY<|1mA
z+xL|5)P$$yF%Yu!{CVnRmDJY=oW20g$IJl8cpnxpP(b(a()4D1Rejl(eg&^5EFxYT
z73vOdpJhcwNKo@KqaNLmabb`1`I?b6qxH}|vlyuh?1sLK#*_XazQF1Cx;CrX@x#;w
zhl|M%2;v49J*zH47QMzmEcA|V%^QF^Elst{0xVa_M18tCB+yHFzSG?+YQM2H2}O!=
z8QoY8WYZB+Sw#0kIrBmbDIqVg2MHm%I)i}50-@%v<=0jPvvA+UhX%<My)9$|Q1fAx
zRT2L{9IHAQ=uilbBClL32l{j|{ErBOvR5Y5@uz@KmA|WPkA0UCI9Z_)vPuv5eI}>8
z2^<L?Z;VK$)UNKmUJvE&kjA-ALc$W--a9fkGK(0Jfq>|)3ZX6|oXcD)af9DO(x3Pl
z9Mb2n0!S1Y^PDgYx!Mvz4Yo*e+T8dpr*0EI)wLy6UUuIJHyHG~RP}B>ero3Frkiyd
zq|73zL;^6VYiTJEY>aI8Tq247)RBh6%j$nT^fY^@yT7~=F`rc3HO8SeWkJz1y&tK-
zzAP%mIKu9U3it!bm(8(AgZ3%S`_r5g{K^1lZHY}Rwz*tzr01e;R86PrieamDGh;h;
zQ?$6vFXV7Rt<I?>3oZHMznUgJdkvROlHz<r<kCy2oHl5lSf_2bv5W6aP9!#72%M`L
z+={uL81y^oglXk+cR5JnxN#4;m`Djp(vXjK{K|T%5M(by{p{EE4;dgifAvCpVh-gq
zLARioxB|6W6UZ#me1{7V<&LY9j0_lww@Sp*(hG2iO0c3*0uX!vGM+_m7@ZBf0}o%L
zT{W6tTueK|nul`>d}gE&lUT612he;6+O%6sGgTD(1-Q*UYRwVNF78icho$^Db?2OP
zuec(IGVpp**b{E?UXl42`0BY6@9WE&#0Cp#wAF_`^rc6m^8;00PlIw*eIGO=;~%`d
z^BD!ZVd60+8nX~Y2Q8-f`HhDK6Qs;@-j>mQ#MRPBQVEePU&r(`O$TX<8^!s!oR>jE
z^rXMDnFopmSwZH`OgWa0z>&j=7<o!X-lYwLCrN=WyR>G%<ev|t2gM(9oC@=u1_1xK
z-DH;3*sn^!a7NC7-IjxLAFZES4CouY#n_6WKFY0q<K^P01o*kx%E9FEd8lQWusYnW
zyP==bFu!_^PiutK1-zA>P0v2O&-29<u4z{Q0zpnAyPtPZhkLYL=kqESYjofGapL=G
zl@&Si!UkcLkxbq>j|3+d=Pun0D(h=y(Z~H~Vt8uQ^2SOo<lbF*q|)QKV3!^D9;LCN
zf@j9n!@biVmO}$6B!@Li$NJOsEV<;;m+pYMlH9a16(0%Rym7+0rzv@~F>urowB{%4
z2&}zEa<BbW<}_KASIstSxPz%9RR(b5YY)jOIWN1>yZbrU6Xete+@s8IQ08frEOhjE
zn$<GOGvkenqNK5h6_=O4{wh^#pG#@&d=dCL0XkJU`(k~od3yTEM^9Tx=s?`<!6&C;
zlDZ)VFaIfF#B1E}N&~%LZPds#-P>1K!OwsH>T!8H)$!!L1Va|A3HRUu2~1imnRZSH
zMat6zOso1-R>(0ROF6im2%47nHE-XOZ2W6?;zQ>nIIOc6we%7x6(slfHtFhOl}4S1
z>#N{|5aQ~b-bgn?rDAOFuVZcHt06egu&OE815U_=6V~j3YUs!{&i(E5HUQBsSK}4A
zEih>DIq#NxMt#jpK`{9iJz&O8W|v&JvPiy*PwHg>_0HZFr~=vHp4B(G4af=7TN*Cu
zIFUR<0h&+Kcd!<{0dli*Np=T83pF-VswwRqWPc>fkKsvyr_sTOfhN^D`6?FG3j)Ri
ze0IQERLvRAddT5qENp6m6)!TfoVed|$*@4X_u{fr-=n$a_@73smtjG_4oL8-bbwQe
za5m3x=d+2_$5oOvBKo<yeC<>4%WipFc0d4VMpJe5YpRL!n8i1F*KfS9;5@_A*_L2g
zPCyplrRAk(i|7bbizwTw9WM_ac+h%D!04QsLrhGi?3xSX<V<tJQ(Ln}P~MYx?W_T^
z^czM_Q-^3b`RByxR$9w>ja@e1Z)SIf_JqJmgt*4hzt8^9p!f>!@WKrjo#<t@a~hzm
zi%cAbBp9p22VE0|6^lHXytYbW*bTLP;U~n=&i$G5U)6e}qRmu-2S0wHDd3^LwUOZd
zaw|uJUwa0~aZwx~1I0cx(>I_Fz|1+$AbZHE4Wxvvl@dV}P5&~qcSXD?d`E1+D#3Yc
z>--iutg9~%ULYKH>2Kk1d7)HZd|7ZCxHZQ+<7Q%vA?w7I9A~e5n-y0DTp{N`h{m0G
zINC;j9$TpQBFT1tQ5se0R4kjj5cxLmx|XIrX3AIYS$e`i@dx?;4|{JK&34}Z|IamZ
zO*_-gw2Gp1sx3uPOYHQTR%lBYjo8D~5`;u05=3!zbU`emYD+9rOOTLMiHHbo?Mo;T
zL4?{##L^0?)$5n*_rLKy|9k&){x|;r8^3dMBgu`NoKHUQyx*_a>-l^<8FcGIO?(S!
zv2*leTOo-DFCv5}$`mj7D6?v5jKzH0*El`l@H}Z`%Cgeiwjs%K24i|{5oCJ(zwH)U
z7Z{lkY<!w$V>}pdQ&%a#Zbs{3EL|LG4<NbOb1|Pb#FRsYrXd52(EPBL9Di!H>eY*T
z1(7Tbo9PV$QIPTbxC{4QpL6V%`-zeNNFa9I5+_c)>7Jf)*O&3ALO{cengDPJ<ZXTR
zeU5To4Jl1TQ8{E%3o=9hD=N4jEn)YNkZjE%dIbd6dLExI5l{~awZ1aB3ol^BR8&}y
zLQPF6jfe@`m}?$h4~Zaj<ga3@g5&x^%pHo<6Y>oH9PSJK&v*NkU28@MGjlDEw3spk
z(mMX&+nT7oh7@B5%=diP{U+9BRblPt1&3z`M^lNaj#HmieNf-%haT~+)B8AVa8eDu
z&LML7VBM;T%nYe=9eusZ;2Decz{vGPPSo)m5B6CN*tYt@m!e7KosbWQ7uL>BzG30U
z60{V%gQ>(u-f0h0lpe@8stauLxg%yr7fmp0F4pV-lsms}p84}rv+=d738$G#7g{!A
zO~=m;?<H7ILU}(9ZPexPn)Y~lB{lO0`DXx(&e>BjW{=}HAM9gVBjzTi1^&7h<RJ}@
z^6e9^D1Ceo1yFSj(VqqbMey#ydWg(>f7+MlP;As@q8Ir?)S9{XOg^N43%yERw(+Rr
z5OaVm#x))fN!=N)rTa$NMs@D3wEY=waQ@`uQ0MC9gWrmm2+6DME}KDtlShjz<AY;d
zRvrhLER@LPF3+Q?x>s8>)1M}RU3XI2y#vD@+8nBvJXhCRuVb|DQgi4A2qB*|#cRYf
zS+wKFby}@tCe0Sw{IpEK!8ArvDHx|g>`G(p$&hYn_TM7rl<mvxYFA>_EaieG(K|uV
zw8nI@8+1kK%s0;mwkFgljk0m0oB(cwJml3nmGPcb7Vx4l(=x_r&dn!|^Vh!u`rC(9
zm`1F!wn+5t$d-%j1Ino#qW_g&yURBdO6y|dTVrgyr=tU>_`8|R8iWKkt_ig89GB|p
z>MW}VD;kBSlpT^Iy{R&<77)Pgay{>ODa<=hzk&1tJ^@D6+Bhh7BLN;)i7}epz_Zf~
znVx6|V#52c{UiRaG2nQXl3a$-@TUyZzK@(cJC0S8PaTsNZLwkWy4=Xpg5oieY9Q@@
zTW92A_`Mr<2ESG6+;IPr^!wb&fvKeO_)#<tfuOfW+gu9u9WjGFDn~-qm+d#+_Vh9<
z!>6#OHC6m1bqW}Y=LH0NktGBYobJm5f;mRr-p?{v|Ce{>_5a;F6BOV_Q{Qp*g>*%C
z4gA+Tvs<2;Wl}Y-(oj{{riEXQE`%5qF%zvEbH-vY#SgXHFtg*`^hd(Kgt_CLxzo?5
zHd2NXt2@}H`O@2l6_&gB;wGsu94UGsD8Kg}fN%hgg*KoJysE4m2ieK_AWGvlND#f~
z$&?}1G;~QC)e^AI{Uz_A&Mr<p0F*u$uj~#Z<KRRO2^c%gF`m&O=sTDUzu2`ao+bP+
zH~bqK0lc3F&x#qHoN(i8B<M2}GR1W-)-7yqt>(N>C}HMGkyRpJ2Vn3-kXZ-lP}w$D
zg<AZG91tonk|hoPZ$-^s-j52rtiphFyfIWhgrGx=(I(WC><k;)-ur)OolzgNzL?g7
z$j=Qt>>X5?pp{2?SFwtlMbD{tD35DqS!N6R*39d(J^eR!YZJ;T|30N^P06UHlcZUl
zmVk)~7_{RpojvGVpk$hFZo!UG!JrzJ^$t)T6(6&fhHGViMNjri4OF1lru~{Jw_nvM
z4+oaJH4LODDfdMn!+JCI=r2}u)|W<l($~B5H4RO~xtw0i^7-fqanZe#@STCq*^y`M
zt=G1Q(Xr*!{d_tqBvcYl%mNE0ES+nh2h@pr<gz<4qne*s6i@eNLES_9iDw0@v$zR{
z6Xb07@%OE<BYQ{YxI5=Ta=({L^IkeD+6J$u$ChbRnKGp4dY3WO5MrnRYO5i{t%*u3
zA`M=);@5xVzeTtGEAHvg%KyCj;r$G7_j~aN=IpQkSdG;9@(;|G@8)0M?U-dJ{JUje
zKc@FPoN8~|8|XtNtHtkZPA>ZYq2G4A_~Uo#Pd^&GQu6<C=<|1Sx8js0j??~JdHC<6
z|L24M&!zZ3cg6p=3*?_n#8`&8-B^uguI6@7MWM2?vYXrK?Cf3HL-|73g+EU0|Bp@g
zdvRww<w_D>>5JQU?f;U!kG|D8`>bTcOra|)IKFuFq(N40XvtrE1U=)of$5>0jVDuP
zCSRhM<4Ji}&MiCSjx8y+VFU6ZO<@ollbs>A<YYh8x-w2Tk%H4YaEdjU%X-<K`At3&
zDvY^3%i`;HSrH@cs*oE9`;40z19OQp0)h#T9EP)n<viPAhy0*>CeX}EJ!*b-)F@yG
zfZc@9+AgjmMyUIRjSR(HQn!0{rQrP~D;@NMYoNln{020t>{?LJWFJ$sO>;#bP#{tc
z^ehO{$yGCu)A9Ldht-q!30Bh^X(HP{Mf5wqe-H{YoO<*N@0fspZC)z_L8P~annW7@
zfib5)_`i2HE!xmcTfNu9___Qht@OlKyJ^{1=g{W$hX#w(I_Ll&m6DTv1X4H91nAyi
zvMz{*06iKd;A6}7#Oj4Ew}Y?OD7!5+goE-Lezfn&t_rw{epUGR9Q>Sgf_1tFDsK&e
zAJ!C*wcE3>1cM1$>M9GF)?8waE`Pfa^wlx;!A2z5gk~|iKS>>=Yl@c7Oa4_g#@65P
z9+-ZxZ1^}_P5-2U<xa>kTUW4J+O-CC(D9g_V+q%4A_n8-QcpQlB@)|g4s;G&V_oTH
ziL*jR^aVB3rc)=^6Bxzxn8>;TuJm>T+JEBnmp_MQ4}pmr*+kUx%W^onnpXw<`cUkS
z^eXpZ<Hmp5qJ{21dc`>PlC&H)wj{B+DVYi%qF09rO*aFj_Ue@IW3|JTBg&>Vq;61_
zt5S<D9TMd>&jtBh=n8zCSh=C=QLnZYQEvOFlGw&g8ACEp?q%OP?pisoAH9ulX5PEE
z1NWULy~@eF{PCQ;VQbk_8=%5szxM(EGN|j@vl4=;M~7N&UF<&=hNgRk-+tb}w>&vJ
z%RpjOzxWN&9aNfr6l%bP(~|bWTw}6R7*i1dy5!0tR)_5xzecSw0^7EAJ5Jt$@(g31
zj&RCBLi4kevT+UGB!q{*u^vCRxJMcfraTP%&v*X7*w3cgbeI$y$}sk0Ol%l)j3yS&
zm}C?aL)q>TQWTtD$NShQF}>E*lymN*d4p#7B1hByUzIQWWEI;p%A}U%iTTq-j(4t=
z<L8oAE4q1go0eZuvvGtUTWba~l5<1Q&Vc?JGO{}#%PBn-5vkKS6%f`_7UEv6$*dx?
zkrHFy;otn&Yd`zhc>5winlnKHd=m94?T7Iw!?TAP1oAkAtMQ-jtod#UDu^CyFy5)t
z4te7z+cSG#REX8Y<Z1ixqhJQ2c74LGZ!<ii)&L%*a(p0`X>%Cwb#l-gJ86GEFDN@e
z0ApI``DA|Dy67>Mv(5;NN=QBg92%Q!-aenuFh_gpeafD@6&Ntx!>(BS?Abh(h@{tq
z?Rzv9n`yF*k9@ON&nXL?f(V*%r*hy+F`<l{BrPKZ-(CMyF-B&8%snjjOJq`)MyfXw
zymhNIq^x`c@2+~B#KLS9zgJD#aQ1Tsng|DaWy*L`+EvCNaDcHj^Px$urB$4n_-NBi
ziEDoiyY$%okEquXpS{}t@v~^JCd#^r+V^v3$kOSCJM3Z0(c_n^&39g~`KsyK-9$`i
z+e6cx%&qvy#H`ZOt;czLkU{^mfUk93F?R*oe<Q7SKZXvEM@O`J6^CEA!+Pl{E!Zzq
z(li3-f*tI@C~YcPn^MC!bzd6X&H@SxrXKpfkGlG<>XOx3feDD&fX_uPCMdQN(7J2X
zrPGJdw6eHmi<CmVAS3BG)PLp$``2mMI4wM!apCWJ{`V|(1|uUg&;Pe$s~@KyY!^Kb
z^uuzOW1HMiB)3by7W!mO7^3yS`tVX}Ch=7AdxO!<p#9S!@v-;kH;VIf=VDU$VZHrh
zuU=6_=rXv+R?<CDX<al+;n2YBCT}MR9B5i++U;JMfj-jd`F4<zi;Zd<R<btdnIXLW
z*bD<71}(IabJmlEO;4)yEZZBES+O*lrP$Z9G+Gf|Nz4dnX1npQV}cZBO8v>9&*iNp
zPwFiFx^|AL4$j+WxrEgzm;9FLF0)*kPGf>r?i}^;ucW_r;enoevV(aygO!Qr7^usX
z%pF-M&Y=>)x!iB1YE-^60eq8<-N0&gzHj7!bxoHHAs&HuGHzYE?=#fVRWh#UOv_Iu
zng(br)5IiHv7i3@wpe5E4`<GZ;m0)+(E}4f+257e4FYf6k-CM;!<L5kF$pR{G1Pyu
zdu?8D2N3yZ)h|#>A(F>Pb;-73b3(#Bqx8_zyuM(|c+&f@*KEt-2eo-b_0ISYK@&OS
zWAaxii>~)>KRR*(K5n=iNY4nsz^f_BdoWz=0bnEPsP!LPx2zn?-`L>MLE(NEVqKFh
z?U#iX%2T$h0>W-?ag7+=)K~{g*Sp*OGAI1^VS09Uc8tz)B}(E=ZOB(Hj0%M(o{3C9
zIncBRp?qRVLwQk{;ts6QgXn{6iyvU~LEiQFo%M7tRshQc0^_)^mJy=@w9_B@B$B!1
zNG-wLJ~Q@fn_z(~$HnDN@#1!U{JKWsExty<$B0Y%<pc8&&))Q0_V>0x^%6f*mL1><
zKnv-PyhRvu(73U*SLUC7;Qw!1+lWYc%(}JkR%DnN24Usj&BbY1N!{DaL@HO-Tf<{C
z2}b&1`RH0vb}9tBVsyPqkvOi)28rXuouK+`*6!1~m-mJA5HpRU_<hXmvR=X^Ih`Sh
zA{F9d3^bePd46lsph3g|i(Fr@2~WxT;ug}#3d4#16hW*QwMNoHa8b`-XLG%hEly?l
z2)Tpzf`s=+0=?!7#|B9U<-=wdYVKDCqBP$IynQ@<J45~0$ztXXZ(2RML@U&a*p>AY
zgH)5hkhsb>ar(+RvrKGz^~$)m%l#E#fKd;~Xb*JYHYDOfwF9}`HOOw@V0cb4QZkrU
zwrsJ*lIuboH8>iHv@uX0=e_CbZwS{R9zX=+<O1^U6+^VDQ7eA@Fo*{)IIw}pC=+5h
ziC`74uO>Jm^!sx=c9ygzbpf60-ZA!3oxU5*H;OcSqc7~0vzc~3T^dl)3QxY;&|mZQ
zV^`REMDB|4N=%gY<F}>L4MLc1sa6HQA~JbK<3QoSGWJ%Z8HD4}srJO<DY&K4vhifk
zBYUE1Oov($QmiJUP*lu^ex3!68`nn#N)r=3Mm1JO4cSP5!VHO>`kJASpU>QxCghc#
zJ20iG?1{cM558raeIwHDFE6080yURGvaTk4X;@q<SY2DCk(g;Af42(cdN6V)2OgYg
zV-nL{%)29BeJv^b4IRK$>mx6X3O%g-y^DQJokDJ~AM_AaPBCgE#c2C!f+i%^a%h&<
zD2cqop?{l9ty%NRLvNw#WRZC?u8Bw3H|g~u<$^rpZKGGHx~^&8!T1TLDft^mpsXmO
z<AnalBz%bRR{q8FCb1Wg<+seA%pW{!{_6Ea!bK`sK6@Uz>!M}(&gfhr=j-uQ$U%rL
z>zDGrIwL2yrQRlIva*lsy%;r;ZKtkDL*T_<igy{KH13yoa=#>~01ypxi8NHH{8!gE
z0!zRBQ3OJg<iuvkmX4~06-8CoXUwD)WZ;Q6eYFzOIVVziM(Z4-`galwnndmP)cu4H
z-xfz0L#9}EhSRF$kJIZJx|+4q)@f)10ZSHQf!&T+^zH2ykype9;)uyrR=`!}X%w|^
zJvKIWig8`gIcLBU$_Wp9ez85vO!(3LB`jaQ4siTt-QxQ|%<h?wSGo!d%g8WSeNE-;
zxzEB@mm5abE|&|l`petnhts`<E{uEQW@58H)eB3GMgHfze{LnHCK#H49L*4+p62_+
z@Xl+x-gz}S_&_IolZnpj^)S=O@D@J?<5lE0+NiWxr@J!zcP`@asK+}z1_mLAw~4j}
zqF29=HukspI$gEvEQWqaC?8&*1;otCV&?RCIZ*S$b4*5<l)wF2w;ocIDxKiR;a#(`
z06tcqZd<j!(`(w)4WDvoB1g1YP2{~44O#qkFVHA0t*`2$fCM-sHHu93U!Xr&SMQ=4
z7gX4NZckbXsUv?aq{~u@1@Sw+vAVRmSYk*QTRcz2;;|Xw5$hbV(-^?VxH^CjTGwWi
z<|9C~jHKUX)Ev;Dqg54zyHiajqKmU+qPoSemU)p`#PU<iPu!<RcV%tk5TzU7`otj{
zKhh$|3>OLIHHb-^XctibE5ps8C;*PUmP!DhemC&!$fb^$e_amS3GX%rL56GRmb|-4
zx-y&T<9pi|scMiT0QFlSP>5XTe7nAl;Y5CkqI|9XW>AA4<<F|RQ3J=>yL#cHIh}Z)
zyp2-yZ<AlpXTD|C?@9!XGyQM5G=I`JSS5Ly4vpOduwhb`8M+OB!@j$x`;Cl>n0pRf
zoSSQHKH<qaYDYOj99t53F|PLU&|#!72KMkQNO3)|%T=LUUr&~fSL*cz!$(O)Wz#~b
z`{H{z{^F(CG}06J)1}=Xn(I!DJd9FCl2#fEBCI{sup+!VuqZ9V>D_B>33)~Hbya!t
z*cf=??v)8I;4RYI>!bM>$Gy?PHgaEYPWhcZaM%x=Kk|jAGlDId9s>*<bXVCQDX-z-
z!;RtLQ!pFX<jQ~*xUzbKnVAaT*;H?;8ej6dnZ#4lCG~7;*^k8>>EkyhXsH{~&kM{O
z13hKFcwm1W{9L{`3!07M5zOiE7TKZ9&^o%413qouldL`kC=aued$R-fuUFj4xLYwO
z)tni)5Pq>6^>l&1s2m}_o8P~1y1A_)cIjpuBO~vcZ(Op6$HgY-mCMvwXPQEjmP|~m
zicj2+s`H>pA`Xv--wzN$)(B16kilXz`<_k?V@ECgsLOEq*rE?Pg}SU~;=4fRK}HPS
zq;2vtbx%aZpxz~)u5_E5Lw+x@ODXMHIcG<bG+B9MX)`0ztAwnnmBKW3$#s>V)ah_3
zy!wx!$RG5Qku(^tj{UGSswL#YKPtzMTYqm~G7w_+#)!*?WnJPq9tY@~D!m9Yg0l#k
zj!tWua|Op@mk8&ye6zH#9V8t5mT>8z`ZhOY|9Hi|e+zZP`Y;F2)I<81nVjq?>n`?t
zSt&UYahtTks#t4kut7+4Ei)g4_u&&sI!qHVYK;G=_57qyewZkpb;*yr_O5~K`;Ky?
zdGKAobR^``J?i4le2_nR7G$AVg&#qN(9!~6gt9{1tlwDZ9)m99m4TM}x7+q#{N<di
zX&2pZ9b?dKp!0G1=H&i#OCKOT$PL30c4jWdfmi(-9;v%|RBGJ9fmfvqiZ4H0NcYj|
zg5Ee%4E|a$W^}`I4dP4muiXLpxD1TYTXkdN*66ut$1x)^cT-4ODni(&I}w{AU112`
zi<pwB2G;I)4Pz5plvL#<`$6n5xb)ii@sIULo=dU&T5ki=0PNHSnIygdLySEPhZ4DA
z%?De8j^pqN_mLe23g_}$(nGdM-vy3g4_wEG{kk|Vct!JOcf}t~yousm>647fH)Fm5
zgc*}BKc~Q)exlVYtCz&~6(5)%YGWq1lcC$CG5a2^bTu>Q41T50d(2&#7KOVVwe!`#
zYK>vq^x<9qcvSRy1k~0D+g(@uB<6z3;50oOCmd*aw9BEX>>knj3`nK#NZyh>+Y_U|
zrViAb)&a%2GfYt=xAYlyMsHwG-*0uen~o)jsT$;~WJ2JuaQK9+ZjpUsmf^U~)N=DF
z0L4?`UXj3g(1d`s?Up+>)$wbEy^DbXNAj_-U^<>+?GB(QJg<N$J8les?)3GC>MRWO
zz#e`(j`KQD!x?OX<O27N|K70i3EsHYTh-;S7E_#YclkqaM>iKO8jV;b98t^;bYnhB
zt%Y_n^I<co72aio%UyLBT>Due;CjE6t<BE5&y`CeLAk;43*dK*q1h?#;MkqH5iDyr
zP@}bWfC*UO;+Ad9G<cBUU<d$5ttBe1QPgXi=P+f2hKg3(=*UQ`V(w1`$jIb{&n5rJ
zyU(S)S{D&Em5DDFQSe0_n08e5`#~zJTR))l6(vW{#q6T|!tIRd{$S@-v&!znN!IWv
z_~Cy3oT;Rp8mjWG@+A_a>f6JmbJ)dIq9&6J{c$KcBK({s>}81V5AJW-@v8c~f-(vu
z5K503?}{GD|7tM`urUsliSg#SHJyX+>hqIx@X)VL10MC_zII3psnv(E&I7K;H*#fa
zT(kl(n&;vof1Nse%I~VRE{6ZRRPIU37AHxtQ9Y0yj(*wlB-ruCDb4$i1&3-jQg=tf
zoZRd{_x(wmqw0fO4@X&7m`_uG`KWT)G*7jD_89xB&6$Mj);0pGkkRGhn2w|~O&P@4
zSleF_wZ31|0LL~b8zOI5xeSgrswK`8hq$-nBQJs9GmbpTsr)cY<{g^-P9QDCPCPeN
zN@-ISTYR25KcvYhg9AW<7F*1K?5uv3%xr*5Ytviru-$FUbjmLn$jOqf#(-}A4V`wr
zQbH-+*~Az6@x1&GGf%Hn5lyC=4E}o49~o+RhwePN_<UY(Kdn$zCB$MZy-948>SY-X
zzc;Wl@2PkH4r-)XUQb!`%At>o%==|<lWqK9!{B>NF=wGP-Lk1Lr+jTHuRg2LUNaYV
zDAQTvTk1IN^qq*4$9DZpZDeZ0v4Vr4xdEcvJS0cvT;YdgxeEOBVqYd5KaO4aLjq!C
zB&HL`4JmXpqwp$nA+;!WaLh<ng(F+m8g|Iruj;zMC_O%^gizIhvbx-jF@81^Uyry9
zf&G}-JCeq^&IJfw45+kdz1o2|c6h{HY`~8Cgg45F-@JBv1QCyezMiNjzylnLZ58)F
zQxGn9Y*=eQw=T#GeOSi0yF?g0Crf+wE4OF32Nz>azF>go^1@}ZG@a|$nf0k4)Aczn
zTBS!xZ@DfhG*D!9$qcrZ6M3x58zc+RCbDDnCW|~7?$>6>7A!Z=faxJ-R=_j{K-ZZ5
zCcNX*n$hSC+aZkVZBquN<wjan&^h)-yz!OZ@}L6)%>Cc|b0N!0+v+{~I-g#^z`t76
z?EX%LL0B1{(3hz>Q>ORFTW7)SlEV0f_qTdZ(%287lRg5k(hKy28%WR5MpldS!JF<e
zRP;lRymPyKO>#a%^nQKOwXq&`Jz{BuyzmDE+ier&XAe;T#pZYD>2|RxrIiv6_-@yK
zzPq$dTjpmc7G-Ekv3iSwezx6$kWr`%8L`ahH-I3cHvCek&GSU`dDzL56Q-iA8OsA9
z`+OGY+@q$i-3`VWnJY`g4`*hIq)nv(FMj_eHvL+eeYiZ0UQ0a3GgiyYOZ_up>EbgL
z4Av#ZHoHwOyuF<NN34A12BpyU>S~;~g3tx&{HS?-B+FiLX&i8(Fa!QAJNzbFv~9Ew
zo$szfg%fj-(-4#_W2kYY$$hDFJKs~t*f|J&ZzNgb2Fe;75D+@%s0!!wj<=D{Kov1h
zT(!FnKHXz%=+rqhv>i3)s+xk-OJ4B~EAUuDu9)06_Na<$9Ea6=W%H~qD%Rd->UG_)
zYOxRBoIq|W_5E$%l%!Ojiy~<+mVE>d21MT6bh2toKWO_P_(HMk6MmE5tf_Zic^@UN
zM;kZ}mI*y$036+;R_Q0c6j~zsUjY)dG&dC3{)*!#e23_k2?Tv#Kdt?6=Y#H!LGH(D
zF5^@V-&`5f6&0y5%~J}K1&oFh#{*>ex?NM?Ss_Jb$paU`Q<DAn_vib4<?0vC|8DYU
zN0^EY+2R^}R|Lm2jbe{S#amiqhl~&#>#NdPwYTS)t3{XwgNt2c=wBqioJg<GPn)(r
zhNW}Es}(}rxczSTc6q!+O=klX$7CIqSA8%Sw!equi9?S6p!ECJC-Ah%eY9AcI#ftm
z9zhaT8=mB+B87_MfV#+0IhXEoM_NRfcP8*GkmAkm-`ez}5(2BnXP6;g^-s)V*9yqr
z00eRYfhu4EoEkPyocg7uqeAd!FGJcc{1=U<HyO0y3+h?>br|&ZygFB-?h}Ep)7E>b
z{(I7}uD$hot@G1lmE=|9lYe8zMOP**TOKcXu8i`zK0%QNC!>cQE;yWI<TgH384Y{e
zGe$p=9>#rbc0{u<e`nq1k~0_S+dJ4fLLOUfs1;deQnFPxnAIa7uO`FLDpQQ#W=H#R
zI@H*=oy};%sAMkp66q5S$mVu?nYSnkz5x<%gA_M{x>6~MoOGORYH9!+51B5c?<5`1
zOnkfN)hfI<X?lmDXPi|E4m~8jyfda93EMw%IofGv(Hm@A7n(%~KDR>z=O~?PiRL;F
zN*kub5Fdm44Daokt$yA~-%5JJE?=0*oqsA~GK?8*>QnTlTETlWb-KzEj+N7l`Q!^;
z%*|1Zu4f{7x9*XhGbVNfWdL$NSfgV@%e!rInK7{PMk5n7TVq6+hcS7GjJ_U~;N;iK
z)nynu`+?G^7;pwImaed~v+W@dF|;|Cb$xPf2-_B+iivAi)B9-#YfMV_l=K{{l{MoF
z4;mhRugcp5-!BXXK0kiC_=P=3Gsfmg(bdIj&z;9DE~>RP-9B-hxcIqYS%Y9oZgp>4
z?oFKsX1aU(mfyg%QZ$>Xa%7=BQfCIsE{DzYEXQqgVs6%p4eO2cqq1wJ2$q)mKmZ?T
z{79xOV4|&qMi?;LQYDiUj<L~;upyX&yaTzTkB+j%`++Pl(bG)3#BLi`>rdx9ow}nI
zPQPo{pY;XljdNO;AqR0S>>uN7ll3GKE5DBi?eB#9+tGRIDWaF4#_o?rfm#TUW`G#<
zc~%jV(dp>C1biG$_Dx#7wU$tJVB;tJa=eoxr-hESnb(s)*gGyKf5CNN6l2w#M1GA<
zf-DCG&cg}zbrpP-#1C9CWz8&V`vENgH&svl5VB{h6C0V^y2`~Q$myi3^_2~OIU^a!
zon+($5KC_}!q*gK5*1r#Q#;(c)ri_zK1!lw8L(tlyM}QWJsj=U&9dXhXQYRA2NweV
zNiN{s>F+Gbv3U1e&!^oPd)?-{#U`Noy6z>D1x$2wpHC2~+Ap-iG)`dr_CMc^c-6vV
z7|wDP>Id%|EkD{8qC*zC8<4y-&1%*G>)5zV6bcg^oCgBDsDzKw6KkXzyh?@wr=@I^
z!L<1`suuhx2DWAq7v(^yx<{4SN>(yu?(|v&YNk{iS3ovrzy_l@?lRhR#Is|k6md&`
zw+P9C_1RNl;KGd_&xIL#WLjEuIPR;4tyi$(fOaFJ>5+=w`J$3u%7+miZyrHW>fC->
z>CSh44J1ZYZ%2}DRIWUm(O6e2Uj6jpNHXzdtXciW+8%ifa3a3h%jeIH41A2z?ap`a
zdRZ~Q(<|nko$V80ehqH$gP}TTQElx*uTJvD%h~;~x$gQh(Lvs)8D;G2JyQxu6u=>@
zu_8|uQC>{<F|m@fTv#o!_^c;AQne_;H<T=F*#P#EhW~YmvyA+oq2K?1o_T$a*{QxU
zr!EWDd8PROk{<lscR&4mw(1C|Ws~nlp_l7%0GB}YAJb7Rg_h#SHd%LS`Y76bZHUn|
z5pAvZJbeBKqs;HM-jpijQws6wZd$_3vB!}P<sMbTX(bsxW(I$R5|hF<gfujye?0nY
zLPw-qLvaNxgbOpSyT2S(xGyUWbD#x%6jd!1=|G%^0klCD*G8YB$Zux-g#ETZCJ`&k
z>?*~j&Ni*#yg5h!#mlq~<VR-Z(I%OOxD;7{3qvN-XbF=A?S`e5$2GdZanLftZzS(q
zP?bc7ADmu24FBm_Qj}?+BEztrT+MTu^sM*MF*!nVvI|JO^<bev`BdGV;0n__ci=Ip
zJ_}}@!@tDPQgC^S`o-0e2u>X`efoq74f5<+PTk7vFgpY9F6%0hyJX~rvN0i*k^?m4
zyYU8Jt3Rvse&2{Nu&>81eX;aArT-&^A5g%O6<CC8BhcP5jT9<5G?hZqD2bcb_(ec&
z3z0`!){s@QYq%J(*P15WHohR_`I_#p_gy^P`UXC#o_}=EvKInw2TVXVCa9MkG;Mb(
zUvc4oeafr}_b6j2?T};3U+vm+CiZQWBhIhu71FW0V;!7<foZp;;CYGUlHBc{{yx%B
zf?`;2Z^JX7U>i<AboGny(OidL-W@Q?cOOwiEhUtA=(rwqOvbaiDPdu^*icp>(+m|}
zHVY}Fm8DTZCQ>IJ_w?Z4N8&?y<U!b<f9xYqtJH|iCy?4%0rl&o2`@9Fx9Qk2wOimy
zhGMuYF-NaC?5~4Mje?&sU6rKwlUJ@RKV9Y1Hi|jT3|k1^bJ8jcD&w`Khs*meEX?wx
zCo*nI@x~}P)3naFk1S+TYp9{Jpmzc@5A~+A+~w`NHw>u9)pwoUon<ZiY2tDfj$#<K
zx++Bmluf1lm;aR_>ACLk*XP1R`l8t$V`D?rXWl9<D&)bUyY2&^q4IzQjpK`TlW0tX
zooYTI=;oE?p?5*>d(6UvoQlM}O+}LoEvHdLtk2w%z}rt|E%Iie4N$;=rmgzAKv&cv
zxVp;_8a<UpP7ACehoqz|q#w;$6*FF=3>G0Yu^TTNClS?<X5>HsD3!c?i70Pv$1^-X
zoKHvyUQH}=B;`1fFCmzKEcf$|Jq4=Ow?-gmXYN9*1>!Sz*RoYWY8o0wCWp5_5-*st
zRGt*u6ST=oMS{3`WZWu2;*a098+7q-;;9*Te(GW5z4d_WHb4stQ>ZcvUEq)gSu?dX
z#@fYp@u&Gj7!grG0%ruD*pT-Lxsfr;^`{>JVUMM7)vD?<0;au=2V~2=Og*0N)IxH9
zc5iA7QKPG2Bz5sS)U(wMKTvOY>A%J55GVrbv|L9c{ptBkO$pIEQ*GTU9|6QY9x>A!
z!G)LUrjp<(%bFHX+vjmsl6C7=-xh!b*0PPd76b;mZ{gr%PCt?h^7TwzBZsAwSZ&*|
z*M`OrjJ1-owT+mnQA&^-aA<3~{A3Q|`viQ0Xkl>&U87$Ac#KK2L-}g1=r{%44Vw_h
zRa{vfpRCbZRvey=Q?<fM5W;6!AqEEU4IBBW`T0doZ%<$Ed#N?lx!cxmib|}9NvFLL
zLLx?1s`1=L+#sFZv16~Q$9C2aEhQKMqxydEoG;TsE!s6VfTh59aD%u&O{NJoqSVX)
zi`2|s_fa|He1_gdUa<i$MP<8|@u{)H^?Kc{XTb0ujt5Xh?n=+c7qdkXrH@?I(bum*
z>+CEt-~H9uw9q=gU@ho;QZ;(xm!!K%Y3di(^)_Dt<W8C@9yzi<rJ558Te-iqp*;Am
zW4gr@nURx^2Yo$Q=jCLDn6k#T*+6%a<ZM?zysF^7+J!eKf3WB6)CCMNYyH@ja8H$D
zitN6}#~}Y$^6oDaPL>Tq&U!$qNo|E`EwI>v68WF0EXPq{^a$H=M4)*sig0+hgSz#(
z@W*SDD7~7AmZysj*K{Wua&_2~QP1a^`!yYKmM?sZ5O<bu4f`M8+l_hBrV%lGy0P^<
z$i;Qi2k$iP>7AJ1QT=YPmuqyjm!6gDZiC(A!j{~CW}3WcKG&G*=7J64H{vb7;mM(e
zN=VNJ^8ML&EZiX23^9_JT1J+1MVWx2<Yj2q9^0s^e+y`ZTY@$ro#BjmP4Sk-=+$#6
zAKOzi{cm-aaRn^}sJRzDKPxCaSSJLQ>mj4(?*z{f&OmL`UO2SbUQf(%Inx@j#f^;(
zu8l?4?-a|9{r=<zfSQ@EAI<8_*y6(C!89DzKn>)fDe456{s_dz_(sF<i91bSkG<a=
zXOhItQ{#2n*`%NZ4IXjleitP-+QUJC!!qHY?m}C04Xrybi=(YHM-zru?f(vyP1Dao
zk$>MJrRr++T(Qy@)tF2I^zRD(3KZV3tncc@6*ePZRA*KiE?PJB8C;=P-L*cw9SGU7
z>e8#4C#6=d9Sqepz6P962N^?Xnlq)j?v>?mZ@qq$ZYMKbQ=1WTkBQ6jKot_3R@0be
z@YH@E3UXnC{e!FEbIfMIg2=?PZ0T!kS)>o%!riEbps@sS{Q~hM6^clKq?(2HpF~q3
z{{Xc~to&ZF<0{5ax)0TEHKDM9^PjCF{1mRF2JPq0T>V(tS4O2ZM$8#v7x#q&+GW1g
zgD<}kk74MNy{yuT@J^Qtt-8~RXT>uY*4fAJPUlBEGA&E8bEv*!hAX%DjY3?fXSuQI
zu1S~pAef?*5Zuz-fNH6pZ%Zpzm0bAI1X|jBJ?4~pJvn&s4{k0<rV{SN0}gf1d{avR
z!+C&j93|0zab;5^jlfKwxmLYLVTxNOH^p`awWt%{w}@8MnCQ!0{!Ze8p};1D7hSF3
z^Y`nzZe#RDr#8kV?XjkEu;b0oSK<6}ex~JYIU}_tGWAP5aFl-N;>4vVn9dMurw16x
zsG0gF=ws4;@mp4ihAtgA#EJs-lE`J2W!3BLjFPdg8G^#$Q8Ssdz3a<9U(}9QT;73X
zV2O>~(^$gl8o7F&8dze@YUBogJryknv+VaC$P$&1L{}aK#57mET&JgFf4oy4D?wA8
z#oH5rCN<tU@$)=aq0P<Im6PLRLCJdQJ6VUnTJ4QGb`);SMy*3%G8)(TjqD`p)70!*
zAxC-uRes@J&n?YQ?5SZX=W@^^e0iXayxO>5-&VZC#KEk%Rn6z#AbZDgKe0+D-IG3&
zZMsbU+yKWHyB#zyw5zR<Em9Dsd;7@M<-O?V`FD(T*Q-E(kFM-|{uDETuJ!c3o$Gol
z=RL#ylZ95v1Nk3!U&mBhilexC2Jh1IgpQ7Jx^GL;+p>wzM({~++g~Nc?Pao#y<?b2
z)p_{{3vD^B)(UcP8v*zfx<RUD>TxJN;m9@bG94;{!;;0h!>Y>yM6*7+#}h+$O&+cf
z*6Nx9RArpB2Wxc56xFMzzlSv4CDN@bBcnuQn6_w{5L=?@eORA)d`iB)?smWQ()qCj
z;21N`{g-dQ9NNN!Y6%a|ZO*#Rs=Ww_s?J3#8S0NP6!dP#^>%mWzd&GW5O)S_bG{V`
zW=FF^LamI^>ugHXs~t5?IC^66eWYj)6&W-&Dz(uuq=Bo0{ppID<5k#oPK(A$Edy!U
zf-6Hvp3T|kzQS0uU48(Mp{dJ%@E$>1`=e?Yz?Rq7f-qbWV#LEAk`}gUEU(~#3{XZV
zwCYFuBzPtsPYEh?zE5)cWAfVibHG!@YWFkiGo|ii`Yl3wp<=cmb5wzE+RSO(Znn$4
zHX_w{vjg2B3Dxmd&!Ulfw5Djaft;4;2&j_o1wU4`cSA;qlBIDCJ0SYYZa9(dp{4-r
zCzn_CTWoBYZM=CiyB~Yb>+@uR$sa+}{R}@fQ#g6aLl0L&YKH=&xTj|<1GEbEZuU;Z
z{G@kk;z@UfZ=lyqe9aKt25IJ{Zwn_7mdU3x`kP-2@I>9`C{cV(ow?52+G%k=Ibkzi
zyuAB>hsJ?%u*eH{>u$i-1XVJRbweZ!CbnqCtSZqR>B}B<p4@f6axH15vmF>jh!&&6
zc|c=hR2#D(Ep{J!^r?5qvpu@iRAPB?v1M_sk4K<U5iVdlfG=}K&eq5<{}7I-j7p77
z$u!^C_1T@%F0=Y%qiszz9g1#eyM|o(=@rfD+TBkhJ%4{&`!qaq%ak-++?Kqsr}UsG
zu{362qlEGD0(qaUz`kx4Q}!aVd5Q%_B9lJueJv<jn^vM*-T1Az?%VH&N~@)m*HH&;
zIp17^aR2!Z%I_Ij---U@@1d2Vt{T>MF|p?Ges#zn&e^gMCy06U0#BS+YL51=m5z%S
zc!$R_GfOPQG)l>IJm}ejsu55M7;a+<$X{+LnJJ%Kz@ipmRsOz`o7=rj+uks?c?Sm6
zGTkZA>pNEPEHC%fh*8Dm=_?QYUd-hA&_*e)lT(>%Q(xjErYD=dyA{>g-|tpp)B6F$
zeWgv09Lr35pc@ZXXDc)<8{^6=>55YesxlFe=9Ww64l|s)ORHfk;2L2!3yapSw-Xe>
z4P2dkq$aKkKvx`$N)0JU$%1#!h8#IOKI<Hq3{%#8Da*dTksJuFujcIsXxD0ZHCT>d
z(+UA`CgFEZ_9D8ZC(Cp*j1bu>e}yl<L?7rMAR=!7i&@Ow4gRFSAGNTN5w7ap-^-_o
z#%Y0mU2m(tCz@|pEqRt9AV3=osF&QvoY%b6@L{%1aV`Eo-<kbjcH`dgTiOqb%X%%>
z*Lui=H6|9xfgFe-!){L7YR0E~)R@7-^aleY#Tjc&9;E$chdj6}1jIa>*dA5LRBU^9
zBK53Ew3d%@Z76w;sv?8t-SbVzzfAn-cc+E)d=#8@sCB4mW>BgA@jattMbM%?HTWLO
zsjCl^`;Ao^qw)<<=K<ygYD&EJHih_4F>Wk)K-&B)g@Fvvs3C_mPsFPpy&qPH^E-m{
z!_>AOG>Dn`AOg}ARr~N^9K)riEDdKFwMw=ugNw-2uTkA?oAH18{ke(u(3Z+DFOoal
zxUqc&+kXmAo#*`XfO)!R(<r=KqLy^|S%`JjZS)nKaUJ8L;k`j~5u&1{>P2hADrar2
zjeV~5#+YnY;&UksX>-F>=>1g8ETt}nD>mpY!7X9<%IU7Mq7y8|5OralBBx)IthqSN
z#>$-FX&QP$>705_n-s4ylt`(b^J}P@<gPWO+5_B|g~ZQZXZdb33@8-s4vFS<?rejs
z9jpf>wgdaIq2C6hgDnABf<;`aUlDarXXjR4>-6zJoU~<kI(aORTu0mR?DESAPT=sX
zh5p|2DSzF0Tax|lM)=Hgze_Hus#&*2O?S&J74!f8MjO$GLElMLpmCU~c3v^=P_~4y
zC^5dCFKZs{bjCo5S|Rfh8S+w1ML*}@l!4v>>7*-MHcXhGhBzqJlw}eUWX>$>9^+v>
z;~BZ{5#P-mX@-gucI)7o{|Jn6$>cIK0_wMG8b^pd@JR~%mXW~+PcQAs(4f);{eGV?
zjD6m{$EPnhnT?2dp6(~FZqDTu&OM!(@Uimans6LE#t?AtA!kOU<q{{*av39z(SA4B
zkagCZy5!~~t5Csq4sL^cUOoeFi&f*q4C%ViAKclVI0PWn28`fM!>Uv`rZok`CJK#Q
zT}5D)Xgp+@^@I@>Mxj-q`U-!KJ>N8L8Pr=ZUj1zEa^-jGx{KC;fl*J65kUTxj^L_`
zka@%mZ5PT{G}etL9TY|3>KA6yaXncxI!J#7x<m`~UJB(z+A(rSVG2vW-0kvP?@ie4
z0TI4~%y><hR5s@W&U&G!22keNC_Au|zrZ14A>yz-nX&cJU}@n4FIUDn2$=Sxt$D?K
z(iP#|xDM<5L~_`C!66~AaOFkxM$*4e%`mK}Q^{0@3Ws72&HEcvP1HTtqY?0YO()+q
zgBzH^I0EDa3Cw=~Vj>xcUh2o($-X$&RPj3Rees(u$DJ3kSs#*~2pr#W<U2SywQKPX
zs=s%4j!AusvFTsQn3I97<48!hy(Xd-2QS+#M{(52LPJsA2mtNjHe@EWJ-SMy_{($;
z?hWJx846;Ox|EJH)Z{Q_dzVT<8Pbc~rb#@@vw^rsoVEpzOAvw(ZwMnno?CpDtn1s-
z?p%HPH~C9#g&KNYtlRlI{uq}mu*Q{e7eb|K!ny3i@Uq9Hw<_SqqmfFjcg9-pZyUtM
z5-DpkLr!HY22PLQn1RTLvF7>kp870NJ-MH{40Laa0K0*vL=-|*`+1{0!TWVYogC<W
zJq_AhPYOI#=ZUxNpiCDC2IHYPPtc%-%dTaBd5YaG^K-GVMVwa;_j~0j&U>{0mg*X_
z2P+qL-oIK?F!^Yp$pZ8r?5&G}J+00d6`M|?BA!)f|9lD4rCXTsajRqJWh`+n=2^T!
zN1bs8BDE$=po<KPk39bSaR2n#<>kP&*zxVUpkQ!DDKIT^S{63A=v_vP<{5LH9AKCD
ztv|lU9EO9(X<{53KSe^?BLt?Q9Yl9=FBJ)q(IaGgrODv78pLUPMf=I98ZGQ;c9?1Y
zn)Nxb!H2b{#dF?qr*_Aqm#=M<&3W=*X6jRY1!orCz!LC)jKPAjtLt{iTSFSgdPcWz
zKHS+p8XLU-q&=ohHqDGOd}X}Xy*bV5usV8*LJznfya5d{^X}{Hq8>tKD5~na2y6rd
z*RsJOYZG%J{>B#h4qU|Qlzo2rBoivObSEa0rdU{Qe16@B`Khk5C0TE(fZBopLUuaA
zJmXHg2T!1OpUhMv$87#J)bQE+unuC>P}HM2c$xPv+B>ymuS=RO#AxY5tt(s`?cTqN
z2-PF`+Wa8ByuYf7e4bWbs9@dNp~hA<?3VDSJWI~)C*`Non>JDzhOb|v_S7tt%?~t*
z8KlYHUefqWP%|4hqi2TfY6(a~Lfl+WyZrLbGOK)MRyK4wwBKOW3YewDTp~)SxYcr%
zrgIo0F3Y6Idw>AGgfLNIiH1eti`4+aQPKp}AY*==Yhh7@dRp?N=5++(v}T~2bVMx3
zy?DYI@%YBY(|!DV1EFF1=4FVVe><&Ipt4aH<2NoXMjcYH24b1$km6SlgA?76B8!*?
zeQ!EokZ<A&T`Nna!er*#n!;tT1$p%%C*O#>I^-6^sPzRd#1{hr5u*hp_(F=PfTSrq
z6otJ0=jFJ<V)|`H;`H8W>Q$dz=gITY;~${++}A7bn%*lfH?wT4vH#Q9wRWI!46aW0
zKGgTS=FOF?jm-H!LnCL@&bgw{W?E1;*MMt4^y%syDacUXrg?>N9&kW&uwg+&H0Sm@
z3<QOA!nLYf5L#sEi=iu7A!DUZSvO5)HF{`KdtpkaDWdSZAy<0po50&X4S#;U(i^Nl
z>~lBnMH|vI|91po^trcNj?oS6orJr~8I&*_n;7lKs)U<c=2ARh)eP+NDNrq0Z<+wp
zP1B{0fql7#9!|sZXC|%@&n_+m1w!1o&Om&V4!J}#j3kjw+ub@pxtMhMb~;536MlYh
z!#JdAUCyEdUBLSHIO~ol%qw##N(|$FkaW~1sWqoQ7*j@ek>HL|mCTcdv%WDt33}#q
zllk+PtJ(I#8<ojjAYN09Ccl5HqvNYDwC``x!4YR?TfYV3{w^Tn#wBNVVyEwz4#TWg
z&`VxwW`f*6(Od;P?%;LJqvhkX+)rB#67dvgpui$_@`WpEMhS>n@GFy=@VPLZH8NR4
zXFeddgnNbv^$@5Rh??@@&iSB6`kCq#A1pF|&=)mD@72~Uz|KiBtq&{XjY&6tp5#e9
zslN8h*LNUo17Ux0Gqgi*nCxn<_fNzxz&4A{`#xNk6t?W{u6Mi}AikKF;Wvf;a-Q}0
z90;MB?Q9ENx770^LPlg70=E$eAj+_ao?tmcs}SKpg4`)OIqY)_M$j~0$b`=z0n_je
zAJK9lH4Vve>vA>WxSsEPr@G@fD3);Tq)ARv919A-Un(WVwV4q+{$?gW?IAO>Gp4%f
z@Yh`jQz_#oZxu9mc0Rk<3%lnNWk1mC4+O53rj}N!c5cc)w3%sBhx@5UfUI0*1R<UR
zW#+zx&!4q})`*ZJyC=UEd83TQ^^h;7n;>ir3y3AP1X2q#??2hO@edQV<6fguipmTv
zQKs?ED1~#TuFvxX1^h4;NR~xw=nn)t0ink7Juf=<9Q{7*CRN7&IRbSUiMR8b&ppV<
zxZAz7=mx&r$Eb#Z)6LPk6Z9N?=HySNx1Segmjv}WjQ&=7@$a@X`x({yz<KAPZQVQE
zEjrW@CN>4s+E}@g56@ZT*Sez8A{O}>(AG2rL`f5d@Bk&0rv^*9$|k09nw2jXXINQ*
z;BK0tM10y#$N-_`0Y3O_h`d|$CpFjW_>6|nr5rY*+;?Zsw?<m--k_>y`?@q(UZuAz
zAS$db-!^kRJ7PE+<aLbGGzA~&yK4S?(yO&2%#~<gp&M%*c6g=aax8x?abt_+y|o#%
zdVt(~*ZFQ>tg*EY*ZTFYA;wzmUfOo`AcvI&fW#aTwv<U)%6WX>PR#Z-5Bk?rBc@I4
zs_^}mAxciRjPZgMrl%QdaQc{8G3o(K2AnXjL2Ph~=3Stl9*i2ijK>I6MiJwyigAC(
zppaYe!L4oNnyi}DEsXJ4^*8Q8o5rWf6i=d}jmdLj$Y?zI68op;i{?EyZ+c-=&>wE3
ztF-^SL|5XkXria(%m+Yg4nD0C{ObHb>>Ik$4FPXgmNHY}@eDQYgeyV>?1E?>sFKJj
zh5{N1C!zc#vWzBHBwRJdK;xUf?aLh(uOt;cjyTDV(>+vPrq6|(=lKClWFWO~U6|-?
zYsY^zJ$v&k@{1kc8tq$6N4qr$uAv$(*M%te2L}F9L%MXSoM``P^z!d~{^ZME1jnsl
zaj@cIWYgoe+(!5vNl1m#+L`U6qe>ko1K&yBzQQ7tN#PbPwg`!%L*MvD$P~?fY+Gt2
z1Kkk|<JH61F)G9u7zJ&JB$Eo$n)7TXXO6E5%>26fThas<gG-9zAaX>|NU9<{ouXtY
zDkQqs&<J5n0JzI9O5WxrOLcHgQlZ*65@u<T6rQ~%d03Q}e8n1nCPBXLz4X{(U3FOC
zMQCnRPp(i$YplBaBkpHWUKV+!t|T+jCcG_E{qn@NE^?LN{9&v%w1tXS!P3!2Gbi0K
z@CYt;hnI}w$igMUS9dsVx~qjNWUbfxGb{)B_|V+dIA)wVR=2;WpEywuj|6&ta^t!v
zg;6sJq#BXLSkOo}3u-+btrT?79!QzktrEGH55Ic5@!-jO*olQL9YO{s{5SYgPUOO0
zlxwGgVy>zLztH)u8+@<SOyyo*<t3W~N8H26_B%AWC~NC!)`c(x=Y~8wj8&3;D0j(|
zMRb>K`{WNk-PEG5tX&)60FFK-CWPX%p_Bvw(AW)BUjOyiQB(bf7_YZH%ryc3dgKRC
zQHx;>U2#*`!-7PaB7A7n!XA=_#X=f+*<!AqwLDi|9=|a9HN^3w(=%P}ZO@^EA8cUa
zlPJKG7k$4QuIH!&ee+QF8Xl)O)1TYu^(fyzle!=E{Mo(H*_&H?<#H9Ig&pY3!OUlY
zYP{Vvb82R+e=7eKUCbP>j|f<TAL!HkAmqj`;|9(^vqm48N3LagmcquuN$@mHebDFn
zt(c8`d}YY;(jQZuV|t>q!cG#$Lx&kDn<MJ+#MRV-vao(~&AeqxjqFtU>B+9{TguHQ
zgCZES+pHS*q~<x=$u<2R5C%*0-p33YMxL3jvO_E5J+AcKMQH>Qf?5{-IXYGrQFOWb
zwL_&GX;->aDNv1!Kqe6ByW^%)e5!b2((>e?xAZH1FZ#>D!-;m4r;snvQ}`H>41P)o
z7n7`Z20JPy{1n#WkKQo5A!|z+Q$qs4&t$E^1upO`7>5@|<A~7^9?#A}-uS%J?h5PT
zX2%m`+9u-ycQEP70NTbd+(q92snP##=*M(?h{64TJzkGn>k~AFk;+Kl`v$xIGLYs&
z(Q<2ie!ET%NV_ZHt=3VH68~^XMmv+bacJepeJu&!qpP}a>#fCEc$rcIor|*rYN$(I
z7%oa*w#Armxu>l^O-63YMgQl!UkrDq5w~s_AIUJ?1&T5ExdCLVGf;7rOgO|SbOH^L
znwGI8usGbwCtH~#C!hB6w9k!47Q-Kj@aBQWu0#bWDop+<?W!-vy+psuwH2$Lrmh`q
zMZ0|Vs&JSWJtMaAznkR*nqBh>37iyKxgaEISUpY;`Uv(=;=H~*v{-ekO4De;wXZiV
z@Pbq!wO)j?AnqlEc$|&;v@UltdRC6igL9?l*VlMq??{_m?RppT!PzB_I&Dl)UeaF4
zSLpJUfkI2n52;CNpSM!}7klp+)>PM~4Wrl<6cCiERHaLouF|AS3B5=QBq8+Bi%Rdk
zgixdf2!swHp((xhnoy*7kgg#5<(}_&kD23n-<fZ|d4A2jcYdwC_Rc!iUS+SfuIoIn
zbNKRy0Vp#@HPhDH3@sF#A{R1A3DA3v)zG^n-N#ybNz#;noHU<+D0J>ktBmm;`DO^$
z*sUHL9(T_(MWi97s31_@TI|OBj0Cv`)cbN*d%^%_Ct3^fJYM=;*3%V{`wKh;ZraAJ
zaZEJFWI@(4|Bx(ZA(qGn`1W#&*541NDw@*dk!N`4?vHK}muc+w?$ms2-q_e!*C}c}
ztFPSBgy-1UGOgJ!a@o+#rnr<S3#xyJT#E=a<v`<6>m`WD>yLZ6rkz8F4{%!r%u~wk
z`EfOQ>6$sm`Nya^nWXcz@hK0*8Nvnd_t<E4O@^w(^*P1e(h|d;DMx&!-I+=(mu|DU
z4bBP_4k*wj93GyvU;X|^K;h0kQ#mS9Ng;`7C6-C`-fGTjb-k@>2WW?|clz@f4la0B
zJ$$_<@@#|{Imct5=w--f>-6X%Z(HsMZIR@bCHEW>TFHW38ftJytvXm?V{yCGNux3`
zea5wA7*AbuX(rL2(~|p#q>$$H*ck5<3W7$%YcYHrOS9&yOp1C<@g?}~uI@|YK%^g2
zKWIjeS3MgL=&&v_A|`DE>(W<~R|2$1agJh6^FoW(x4c8h`g!<sj4`1aBD(rY<qZL4
z<)#ie+<Do6P<y*G1tH6D$|^%t*wG5F9=9nY9*_*yWeC;cttBJO_H>_|#<Lq+8Kj8f
zQfNCHcRm%}6ABj)C^WTjnPBnFe|ZR(yXE-ETjTt#1HTH{MZV)D5qWQl$<Bw~Bg?kE
z`KBS$gM5!88x1+1XvA%e^Di{fit#(#J%u3Cm!?paKoF8Uea&Y#yy`_AO?+bBH=qr0
zrf0ky3^47B)N(<jRyrL3=LCQeoF~p2jMO^)>ZX8+{(~;Mz9zGq=Qo>pt*i{H7`4Zn
zB(gdgl|yM{!>VI7i%q<hQ;v))bbdbM<|f}xDDj~8s*2MZ4ckGe7Z44+=}q*0vSRCb
znxP)P;shUhyFFfsy`}LphAp~z#cGJCAWX_smTpf<=VPj6S20E-+o8l&7<OV(jw%i-
zj5PQ8A)8N}vzd;PfboiQolvVgJWE=UqBtIXZyaRpq$K%dLX0SX(?!9clr8zOI_pWg
zTy+yAdPIU=D)b$s#PRU%v9*`gW}wp%_F8+tJ9qrS$&_9q#ltLgFpHq6wkD#)C^Bto
z=xsJ*?a<pH$|GO(SqZ-6&tOBGUSj@Kq1}`a-}@2B+pY{BeNc(S?oO<sdpfSHeKy^)
ze5d`*;S55P%Er#UhKHv#s*k2hP>|(R*BD9kmztaE_8TE1ULt*T+Fjevj_fA)UXnpC
z-0}M<klM9!F!H8Nid<(-OVrXqWq3`6PP7#)jlic6?`LiStQ`QOu~oN!YOqr>Ip0g5
zRe6w^`Tgc`<d5Ku%SDHei)J4yM1c(j8?<zj>d~qOJyX2vD*!aa5I=D4s1*#iW#X*%
zk!TRb@wN}3o~NIq;P8sGfKNESgIQWcppe0Y0RNg+q=LIZQ9D9dk&idS2ncTgqt?L4
zet^D602s)MxqTbo+3l7MF}nM~PDJQTycT$M@KwJ_9(&S?WZ#=^!M>f!361Et$8>dr
z+-HMdt+19gDv{K|4P$;9^-XUNhB}&zyBw*-%p>q?$%S_5WxLq?*}gI1^g{>GkLr9o
z3A2wtf$e^*2NARtTVT}UI%!f8Ipv@rrS-7g+O!bke44c|c6!R8uv6+Mkr4r1=#F4H
z8OmR?1_DGnaG<<XNH;3N8x*#Q6yNU^_F!J~t2?N_bQ)K8teZrewDZ8oR}pZ7Nm(k&
zWOeOj5!gvE)3r!%Y@J3;{wujotc#gj2FzMzgHqc#=i9=jz`?Vx+4t%~L+E5fjh`5d
ziTY={s<n^TN};`@el_R|xl|+90FQ~Q{AL+<A1W+TZ?h1Y?&I2D8MF3XNN)Cgub|up
zVpJw_sd6G#W>BffFoO}sW5Aq$S*EZt&%-ZKNH6#)yXI@r$b$5{c@oe(K-Baq^|@n&
zo@KlsJvJfF*hIi!q%M5&9-t@OWIQ$OXMLD6+~V1J=cq|%q-V(zP0mtF*hOD^v_9#5
zwQ-NECPX<9bnd5L(%RO7Fwr-S0#`w>{w2C<&T4aIs2{(_`*@(2bc6FFZUa<AvFTIc
zepLhaa<d4_-Ckb#c6}G#b}ed643b4U{#NLuqfj1(*Lq?=W1{$cw>0f|S7WGZ_SA+<
zjX~fX2NH+wY)@@%XH^s>N^iVi$S4{yK`{SRwaMo^%WxadJHFqRJ({`{5K2#K{ll-&
z6H;+<k1W2{jjS*@8<AQM>hJF$oSTWJ#wI7LN1gNN#QRT*h&d{tI5Cg7QO}W_OOHK_
zzTbU4c@nK-*ir^Qbp+!=fI03Hs29kI5_syKwZ6mb6cC(lhgum^n-DUmU>M@e8sn`6
zrNBXTS(}X#n_>dO7dP9#hhb=hcIYZ<qa)tw4_DfV7rD}g%=QetQY%*Puqo>Ho5%@4
z*d=^AiuB&|&D?FoI2;djR))eGFrrP^!|w;cRfB1)#P5<+)Iw($>I5){py6Ac0{73M
z8WJ7SfjATjXe$cxq&Y?lX^igboAwzwkaJvbNR{xj>GSkFJ=iE~PK)r~Le2v8)fA+4
zq~zb+7m^h+teYKr(I<49{?lad`e9?r8M}Woqx-%3IJ2gjWV@Z0CGjdN!Cg@+1uuHD
zox74oDGT*#)06xWyDg)AOJxKCo_wzBdrR9+lpm?W=&>>Fqr)=>vld&Gua$w|lMPIf
zj=X7uLIeamqNYVXAuT}B{=&>LuXBEIpl_huY|^e=h?|uQ_Z$;AreOF8(f|-Xvb06Q
z%Phr-hA7FSIjLjZlfs8J6T-VA3?W;Pr+O@MlcGJa<5v;mx{aG+?^8yXi|taX-e`Z)
zj!KZvl9wtUuZH}Koon^s|1xti3_R3N_Fp|OBC6e$<fi`%`f7-RN=Rrm+ukUWhG!ZH
zEUE_26`_3Bct;+}4>4lX3gOE=BsI-I2vm29T8wOu)r3tPU~^-DV!rgt=-A%tr1AW!
zEkwR*5CJ*D42CKjva1LeCCS@ttreV`_!^AY5Fs+_!C8!WOm=W`uxXzcA_am`1VpBA
zm4%gPr2|o%#Iw_1_7x?8HnW{`@m<0rv{`h35z2lJ<IV#8b|B<XU&i8DIYcGUez1Of
zqyO|L5}WlTd2xSS(xg*A+H6^KTCSqnq@HJfqPN7|TK8?n3amRST*WV`nGjVx&X)OR
z7_9u$93xUfaCEkljHDYVjfy7y%ii$#Y^}c9|1>qzN@4;r<^U?=H7)B9HbD?!?D9iu
z?FT%6ND$E)?lfl9Fn|JW8uoGRloF$^PCU2)0MHZ=AXdA-OF)2F@z(poef<m2Hno#y
znzbxU%@BqDCfR`lwRv8JXn)EHztRlt^Xo^{)tSK>?VXDn?pIlQP4wXjA&f?8FpM%c
zzxrAEron-#kNq5>TF6&m`mv`^gZ(zail6@xG1zHhWHiy*?+y5s;q7QU1$LaSO*JJ5
zxJT>`^nxNbQCXHwdZ`;1nByi*jqzOwu7iMccjxeH@ZeokXh?l=$ZM^o;3sNX7O{nS
zI3YYFKqaN7c{^*;v=roD1!5`$vHeJN-jG0iSgn89w*{{@uR%tqXldm7k76TlyF8VY
zAk|6c;C8iYp?0hZY95QGi}QE(bZrundbx1%tesKvKK;9KQN`_HF(AzisJ>lWe=p%2
z$yXd3nGA>^+NY#1>TB>#q7WpJYFM0GR1yvxa8#{DZ=55e{a5=$MNDP+HHm@yhyWks
z{z@}x4_+d|XTVLz1x3V86{1+v(&R-yd1XmA*zr|^>Z&FoAr&?>bdD*2RcC?idIMiD
zBAhzrq$u$HrKzqlO`cSMc5Pci-->yuGlw+kf&Ou}Cw!?TRI;zlPruDCvS5=i_IhC0
zRfJ5I+esf0Y9$&IvA?VnpXtg}GOpy!vP-Qsl%Yz#`-kMpF>ezZls+$mDKT#;SjG1f
zgO5CR2JgBTrsb}T0MnP?r^o%89bAa`nXIHeIk^)mdES_*2W8_bp2vrWCjQ2UWng2=
z(d#*HKQhhB_fDB=e`8Nns~-}Zz*Vy)^l<Uge0OZ$)%==G<sKV;E`tax+47OqYA9L)
z;WE7*zF(bf2uuwWJp^}PAuQ=EQMm23iC*wFGRro`(xr24n0QRQt3B-EIN#YdUP?qz
zT~Nr+W7Nf<bK$vs3c9XB4y=u8&O*D2A(Nh@yZaEMlMj#qzAA7h6!f7fuIbUE9?PYQ
zP2T%Y+xIJ#bVSwJ?dcVJxgV7cT48O3!O7G?l4BMcbxp$uL&pi8(aV@%pYqmB*S5Lj
z+)d-E`9w#4ozW^7Aw7gwLb0h}`?QcLs-IAQB<3PqqQ~dm2XhAQ_T(Y!ktpYw@$Yi-
zc;i6L*~jUFCCTxHfJD(9YpqZk2R~LSR`N1zEjax`D<0r>Oteg?OZ9iXh(i;0WR=CB
zIkdo_d5xwo+b%t6{auH?X0HcupGqNYB|WqbJU;98+%(~_fLw+^f9?o*k69fqIU-0-
zTk6U|p5d|of$Hl6omV5r13tRL0w<;Re*6aIT=Un@VQ6|u#U{Yk_p+Jph|-xstoM{c
zucgC?y#9bx`-=Kv?SV(k$5hObpyQI&ntr6I#{rQ~oVS4$tpX9(ViBzlfKdc^4SosF
z%K9Y?v6Z?jeACvOV2DnVs6^3yv@(1h_b!j2ce?Z2SmXTbdpQl9J&%9g*=F%)Q7yH2
zo{*AMy|Sg3^|iV={z*P}`~9~=WJ11zvy%61?dTjFO7gcRwj@L~YF#BFaf<RQhPCSA
zGpUuAruahS%7nvT%FzPSlN|lu9hGK`cF`Vp^~R6*@m)P#k>a|R9q>YqFbqIP6s=<p
z5SY(;7UN6!*ks0NB;6#<7@LPB8&ofR$GWBq54V0CIG|Z)67g350isNYmSRRfNtqMl
z+?Ar$`$nHJzivsp@V3#<4Z`Nl@^fL&aF2E<CTHU!@?8T4CXLn0M$Zn^20zAUDP|{n
zuyB28viw6*hB3=@KsbvWtBXucsSRm9N;#XU_gPmcR2b~*X|2?NMlG0z)1VrrEa$e}
zMmn}2dB~zB-JzYM0fF|PwTV_458a+v;j81ogW|>MvBf&&$A<(>#@zn&obr-z8UkB;
zl>*_#sR((QOJT(+N4sb`?4(`S>o94V%Zp-i9(<g6yuTomM1S|~v?RiMR~qv%-{BY8
zqOIhc?n%DYxxTSuXn+DH+$CC$>oTnqYwKifM#M=+HrI8eMG&1Y98=R-l$*wP^?N2?
zh^Q-xo@yX`)KQbBFUJYUxt&s}VyxN>M#0oskMUubh}-d$zQ7O_Ex}Jpz|Gbd?-#mx
zs+_QjCwfHs{T^3BLb%w2wr?}J7-@Sga5ln4<F%+ts!d`t+>PU>1o7U3rIX2)`wD+Z
zl;`?3Fby8$y99dIH?vruP=|m-_+#A3?8VC;KmL%ER={NQ4KrnaOV2cIJK5ECzyR_i
zraEI2?Og#aq8#ME$109zf6X@18yL2z=dKY!K4rK|1p*_rHxV(Fhuj6Y$9RP;d9G)e
zR`QYKXe-qUWqJjgUJady@0@-@eNy8Yd3Z8{N=H?CcTj#Vr;XOTh;7CsjTI+hj5xZZ
zZ>jt&+7D3CREMr_YUb>!-*vO!t*o<$svAq{cQk!ApKzG(ZO9IFBwsBJ8+UOWq1nJV
zaOXlhL||%1LJV@n(=v}hKBWBb`o4@;AC-&?psnT#i69>kMjesp&XqS#$n}C}JwpLs
zhAp}~4h7!`-*0VED>*dj5*@p?nbBWnE#{P0R%zN>RVQ$2?$BQ4j(#`#G(~1KH-p7S
zRa`T1ZIL}!NnxXM)3HICTJWsQY!XlKah+<aJZ7-Ri(KH7QBIq~d|Ok_TOF6C@kC>}
zKicfr?4(JH0=%HaFflaK@=Ic@@inG^FfMfD#Hv!x*Ps)s9)Vn0(}`7gcAXkAk=Ho}
z1GBQ8{bfdB{$bA_G#PF$60L1THz4Hqj#ta)p1WnWqa-Bxt*516l9IY(P6#S|sgkE)
zIdv}oCuUZlA%no)th#bsblfo4oSa%Gy*pOxzBip2>F7KXNbi4I=q_aC0?3(26b~qM
z5#)e(mB>}n0Egr?%Nn*N>T@l#@Bex^)+ke`SExBI<hrbvLECWLYp=7_97s##!JItQ
z8ZsZ<-r8D^iSRs1%~=SzIY74H@Ve0JhHBP2-f}NXXF;bUJQ$hV5eeJ}eo0>m-|AV$
z<HjnJukcoA(e#1kO2o7a64V;Z70T5aS2xi4d%LWVU9xJT66mFp>bT3^_SuQCz5pM{
zbH-p2x~!O)7^^noo`KtH5VI_fC`NLo6=wPST`q>emu@I##%CAXew#qfv;gP+LP|Yi
zWohu>LbaGWVAtRn3QbLl_I?17z0l!lzR}t7-FNC%=b>+l4l59DOsgBoGvKtnrZ3YW
zuktC7tkFiqIU@?WgnX0AT=?XfiMH}!X+lDc-*dm`9<l{|>ja-{*WISOP#Y-v&R$@2
z^yj&8zUl5tk9RFa>K8{v=H1vwxkP9G67PgMake^?+05AU27jz@a;<nFR7DiWwi$z)
zdL_*uaoorOQSawOloXYwI)(I-EgZfc@@Rqoy5G$OrlMtuoCL659w@)x_R7bPLaHIP
z3KqtK!uLYiyLCX!J1OiL;~KEp<c&yZ>d3>iP)21CA9}T%_58&-Y@ZmUDi&=It*oj;
zuCz{`>FI4ZA<7mec5P;Qdq;(#ERDvxn?Ln<2Bm4-jKzx6HxOA>C)4RT%dnVOzaQ%P
z`*X&R^oAxx80hP9eoYbx12-#xLy3EpOO@rSj`ti4Hed+QX;)KH7d~CHsU)y3_T)7@
zAGsi7?kYR<Q~GF;VG7)8>1PC9FVW@$7jb2v<sL6izIy3A@@8zOzJ8ruqt}m5BA-1!
z@4mOHx7^-+<?;uWQgX96@@S06J@snjZo~^=NjS!7yK}6%*5txV1d#m~w#VG1$!M2)
zjQYGjcLIw{gQI|}(*&XWmK#%Uz6_0}1TokB^?}*?YV~O>F{DU3y4?^NA;BV0oR%fN
zhQemO1OpF$&~DHi!jO73)^zA%ypP2*x}45f#`e7BMAU<UJ8U_{;qh@_E33pQBa{&v
zx`{R%K_|SQqfL$$^0GItFU-d<(a(IJa@bMNt`{cIO_rc_ina~uCQ4_TN&vu;I$#Dv
zpQtpTxRR0TA#6zZr2lvC$b2)o-37t8++<*)mPhhrX?bZN!$<(xuM<nHAmZvi0XqRu
zYwCN&%;n%N-b^dT?^4ML43On!2W0Cl*Lkrc3`KQV>&tA)SsUQ)+k&ja;^syvOX31r
zdBtf(MsVmIFIA`Uyvdfg{PeK$x8_tQu@60NL>FvGAghD(EX>B-ybt%@8|EpF3Doq7
z;LAmKcf!j~8o>^xTUY_=30)luwRZmK%-TwKU|{rTqHs`OS7$?_JL0E+17bo${d|9d
zT7jth5dkg(6^6x(FdRwaZ{qf@5Pau=&Fi?I9;HU;K*RN{;D!>KPAkhXPOi(li=IR>
zgHumqS+?D}dsmlHWr9mV!*D8}oIM?DPD*_3Jn8TFKO_{s*S;5me`w}-`c<6_aD)9D
z&-sc=thZB67|N#_5)EcCzvC5|prv*v;t~TpK4xWeQq#2#khPTw-Q_vLx0J|+GxD7J
zD&qP75$n72XQh?Vifl8OV>KRoIth+lRLN3_=lGO9K<5HB(M8eD7HNOMJDOQz2h7@l
zzS%wEnv+3Spd%Bl1+&6k?ZeMfQX{?HyloftgqJmp_=IdIxu1&~Kvk7-1;s{=%bF41
zh0*@c_Vypb&_bk26Jq`jd=6%af{A#u8|<FxK=M?D$0uth4T%5|qc0ej;s%%EljNzc
zhX)i1r_CQ@Yik%KUv;$jJbR*E>oP&qXplF<KH{J2hL`b<sWWuwBhxmKtYNW^&oQ%4
z;9G>|Z?z>_C&)45TAzm%<Mt_f%KVJEp<&8JMcz90qE6cjwI!*RJ8o2LAd?<`8DQQf
zPhZ^pliLBxL*f1uKcLl(B7Z5X5V<1N!Vo$>ro5t6$T4trhPaY#HYIC_!Wz9DCd4>6
zc(v3S^TI5fI=*QH*Pk?LnsiUr+WVYm0lf6>XY$CQ#;AHE2g21fy(HZ#_J@wp5GJT@
zBW2vJLH;;E?f4T;_*cS>bc@GpROG+c?7L4fT9v)^6ZZd$a;7S8>HYsw&UlHGv!~&#
zuYFY)#OFDF1ULE0jaO6(@5=c&K&@QYzP+FDE|rqXw3Rr8?bsVtag|-_@l1K!)rvt-
zD<c$Bx<i6}c&0?in~|C(nhrwh&Zg(fX_|6V;y#$*Iay3p{$UA=V8P|4rQ=xA)_Ykb
znHvC&gM70veO=fyU5@R^YOVh9s0)E1o6*Tvty$>8#58MB@w+Fmy`e&XNCjlB(&TN7
zn%|~sOdmLIJ0^Y7KYT}av~Ve=b^1cc^4kxa3fBsEvvER$=6x}KerK&@cqB$RO8UtY
zZVrU_@k}8r%OImzc?SxFc?9_v!xZb?t@+sV8|A1iZ}gmpL9K`OxDf1a9eN~Kba$5M
z2TwF@a0P^;>~c%0iTnv2X#s}*UZ@jq0|<FD&>~F=T)4|>k(MPHp;V2qu%0E*2;e<o
zWEv^6hG<0x-MU4je2@~Tj&eIvl{cqKMRkC8tbAe4INNzIs)WYb1;f!)vi<Yr^pVMQ
zS5qlioGyBshGS*I#U7812Q;|4pi0s?kwh6c4#Pgpwmi)*e^Jgnc)u-KS&W$aD!@jf
zfsw)!e_5gcI`Zl!@=8KdO1Id%8wYOS4|zVFJ@s+IEl>~$2-Q@UuClK*?fN`x-Rbo^
z#ajwEBwS~03=9K14JU{cs|X35JW;yjCAx<38majP-!W}{lWu6qyXwUK2XG}a75Z)2
zhzQ7X*2p1}f;MZZ0T1Q>1(D_TUMgG3v?(z@fCU|3ge;Amas;T0e)k5Uc47Jg4Abx>
zj^eO2-^Q_ZR=MLP1)<}0c@Xrx3qt<J9*P^!nMOr8vOoJ*#u;Cmrc#BMCXs?<6pdDV
zj&v{@$P?x5ACjL^_|fuE?Py|5L`iK(>R`y<_&T5n)jzZ5i5kamogN6g9qp~hY<$F*
zlSrd>Qw;^8!2$1J<^cW|h1MD!8-m3zR~5y`SAQfh*E&mQxMY!`qfp~-btCF@pp^7P
zZ4J2AIt(ft`^MMhl3j-X2az`3uvoBLQgs><jomeFpQ=7Kf@L%rbE{+GZ8_F5qXV?Y
z9hMfC*8o6-;Y26<_L(!PtOHShVA@;K=A-lF5>pFnG<DCE1B|;kaiP}OGm1Mgu@1jR
z02JOhdjsTZ?{0?<0Oe;RtYe}8&~Xh5gEJjVz_y*A+@Ad%&*WCEJRhpj<iNTpnLYZH
zal>tUa!G~?d&zR;@9e+Z<nICTKYupx;qN)5J__5Rd#qqiy?G~PS^ifsBj;TZQg>a$
zc)#8RZ<BwlU8OXB#Fh}PXtDoq=>DdjsVCOu2Sob{_`{=(e{<u%ZSt-x-|ys`zgQQx
z>aKSEvu|-_Ylokc>)|MG8Q0e2Pfj&tXDb`KR1z&!1W3@#ZkCNMK}DD%L0w*{Fa4GG
zf?~a6*{tp2R(a+JZ;>`X_hs4p0N)+bkuv&i4DAo^lJ)mY(M}7d^+RnF=8`1GkdX7;
zPJbrU&*__aCSp?#Lf=pH2fs+uEjBkeXNRJi-LGs+k)05ten|t4D(?SX_}|U)_h|Wh
z68-&c`TMK*_hRusX$`&aXbv)T=v@IAe5tm;kR`X6^GQv<JJb$FsJaNnF!ND4Fara%
zcoLQ#+VofhnLPh>Q**bR@^@+SZyxjAcULd}HT%DB>Wb~^!szEeBtbrxdS6}<p}f0e
zkoibRGT$?#L?Aa!R{^M&Y$syP?!bpl93u3zcTqXTg|BF*pg^OuEU2e0bhK6ZlQx~o
z<WdT~KwGTQiLCK)pJej{Pw`F}K2MS>Vl!56bdpQNBj+5R*{`mxH>u>*fK=T(i4|U+
zofSUGysGk+?bScmd^fTjTCLO1w`MBlL~Rs{Yn)S+pBv`ba8S76_}J}a<ZY+(;XWj+
z)y?D<r#3_0`oD%r|4#nh4u21V|4EbK#w^H(%zXO&4k53Bu8zxGiV}!yD8%JeVWkGe
zYmnz1yo!EWb)fYVM#>dZ*u}`^0=lmG0p{}QmHAkWOK9!);Fe7i@?&q0ZD@HA?;n!4
zBp!KVEyhq^X2<haq)kkj=LE>&-GEUN#*O3sNrL$;<KJ;4`$WN~I{c3}3&$UBdP_^5
z)J?V05b04Ql7AQfcjNp$a{gybE1JRJ95Soa-Osq^j@(&6&+>uRNv}&)bo%@Ie^OAm
z3JJ;5rrpE;>lWsJ)ARe+TmMZ{|29M@!;tg)%Y?M`{nt?ecdFy)qg%x;$iR;K>E>UJ
zo>F%Hb8kqhhVi<z4*#uK^M}Oy(^Y|guC)K?-;Vx~#Ly;g1bmn5oE1rY+<)}!KXvc1
zzFbQw?W3P_HosSmNJailS-1oKtse8Gj^7g`{KNky|G#aL#^3ak>;4a=mI?ctLbC?4
z)=2S-VXG&tr7Jg<ouYrY86hs#QorkkU90@?&%NUq?|gtEOF%qYcK-8+s8|17VgJFu
z=@*D>v?){3)5@VUgD-5>|G}gG+^rkb!~5Ge6n=Xc24M9*47?(iCM#wJlvMRPGqyqY
z<zq~ZjG?#r@1)eGYbWq?W<PzBWEi(DsQt7zS|FG<!93zZ+fa!AHdR#dTOmHb#w607
zzv^=Zwq!n$(9#iki6Sxo{8!F@_aD!%hI8E~-(v$%)8w}G75+VHu4kV+WrGd7$vPM5
zFGFg@&Ib6up&#6~x^nfGk4J31DZ(x*%m(2(IQ7YF6`u%`^An+_rdby)<CB8-D~gse
ztAomd62&B-%y(rc)1#$7h~9U<$e+!M{a)KC#+qBhH3)0<kK1UL14#~EGov=D`UU>v
zV38pnTglj+5U))P7otzy9iv~P;p#445fgs1^4kfHv1g9}qJXKoPOoQ*=KwvDlDw`u
zCoS==dp2^sZ=iYxjt?9DR*|(AFOxrC7S8mrKrbW2cx4eDg!(^s&sfv(zU0+u;KDA3
zw8(Q$E=7B4C2f$KilJi`4O|+X;>al4VGc!GcZ7^-u?R@CzDcSroe8NC#V3`4$U;Y(
zUZE$VUyMcTAcIBg^0dG6$i3q#DaFM~z1LeaIB{3F@$mN=?8h82d~2eb%AQA-Rk>WI
zPMO$<VemPzK3DH}E;+h-BKW@wvrK{078y*oxqIKQ|2#_5^<B*H+5$S)E_tSxthZK_
z1+Ce;!%3qDyv1N8!%KPk&259F<CMj<ysyUn0ECdJEzxqzl2#bMz?V%&As!2-g>ryt
zd!5=G>&Ium5ANdd9EC|r-ds`I;fX98@ek}Dklf?1M>}@9Pvziq%4wsxxuqZ~2Y2mT
z=c^Fd7<?zM&W*#*lSy599lAYS8#Q00F6=~#uqC}TDZ^(XQTsO?f6#n8xaWODoik+7
zn@S$Z>T<VkSc45?ftwA5n8#DhL+pt7QbSkeROQ<;U+B61&y7z)@?Tt&Ag5a*V$37x
zW^hX3lHI2@g-|O<9ZCUlC;5uDiuKzSZL5OLq;0>OQv+;E_Y1mz9zA`qh?M-cWKsHK
z8xBm}SmTFVMQg_EHGmr(1g$t)x#&#73g-+NY}GsCjui7a24wY2l5B_KtR2;!Jqp-!
zD*nY-`dg1A{T`i}k}O$AA@D)Gak6SoNqoG1T@v$yGSiUJ9aeU-`ojk))`;kYciZ87
zCgI@K{_EqU+L(c2!<4vPS6z|&By|!}7^di0aUDCYW!0A@naLxQJu>SlV<WTM%>8jw
zThS$XW1G(s!lD8vcldQ(p;iPL>TEFM*ZEcVQ3ph}wR3K}SUk<wP6HN8dbKWP=9&cm
zf~xgJ8QX7U|9^agSDMcGFJWVj5QaY_u&Aqwul~L6`G<U$MU>Cr@*fh_Vy_d$Z2Lj>
z!MlcO*6&_ixQ^-eN=kv*OOc$m8QBVGkJo)=@*V^8=;mzAf?-V=c{)wkPE3@UbeYrX
z2rLH|U|Q}@1=_aQbUk-|Hlh|jbKT7c9JUb=1-C!m;kM~a`-*zaS`6bdO6>0r1cT48
zYvE?qWk84SZ@Ki<xnjJKv4_GzohBip=SiDuAdZBTXhjFQ#9<Z@S3Z+3FyDWu_W#Ev
z>75OalR%a~B$A~qvn8*9pGlr1?pM=>=J$9N9V`nk?|1uwKq-bG4vZ;mBK;i!n!{x@
zSxfmKLF9Q48;~?T+Gr`bk(@bw8L`X#2`L`Ie{T#_B{1N(#EF89nU7@9tRtOQmd{0~
zuxwp}RWC!5lT@#MbV1az|E#H{sMKR!uacWq`Np0<CUC(ESmJNzcPE526qiYw^a=OF
zCf&n9KoQzp_G#%m?2H*VozHK~8Wza-tyUXFeHBf5`SsCLrW?#(0xWTnk%|+%L5`8G
zIEVneing=X_0^e;Md{-P(UTCUa>z2~k$5uOahIPw;<+=Ag{EV+F5LNHuHao!jiVol
z-$8xOy&CLQkLTUL?mNJ{AebJcCptpJSZNl)%%dk+R<C<CVF$mGFzm9f$ZdF(pv{DA
zYN&xWT3C$TxSnN^UJ$gD9$#oE%?J9p{_;`}5W1SS^W!whkV(>u3~>h$QF5!qIozpG
z&ae2M6#ks+O<y)UMf`Q*#{U264LxaUtO(-;gXqk~_Z9Q6!X!zh`_5t)T8pZ|d!#gZ
zr9J$|3lKU1S7TMDi-T=j8BbC6$B2|2UG#CCvY*PO8cxw#VJPmtPg|ikZOUPtvN3k2
zjXI~J%5Z4Qiu-eJBA>d--CI~}X%NWUNayLGF$|4mh5Nn9&|-yCkOcct9x33zmexy^
za^J@5HmS_14=OrUG_qb#hOkn)((AG~_x1G*D(H3(wx%fdbiL+~7+qrU*LCkvxw8It
zIkj26bgPUy#`*e8o;T<da4dDgXcP&9jZ1uTzV|j;t*`*Ez0jn+-7jl(MQ5@9dE{n6
ze*KPHJ)?vXcrt5jQ#Tq~(X(whc57N&gv`sLF<VkUtU)`n`*c9OAJ<u+FEKXqE_>pM
z&VVppy`$i%c(*z2<*q~DJHah~dY_kO`JgOn*}<Rbkq8b8map^5+ImN~6AktCih0x7
zi1_>@g5pwd)bU2;=r*uoIy1{TdeCK{qjl!==kXojOzr5;$z_YfjD+)p(g@d5&W(w^
zjN?MAc_qgzSI>)8jq<o94bE=0tVkd|XFDPM()RujoKNU7UO5HAR13MQ2Of*kDC3zv
zrqn2t_jCbRLCJY~v~7}loaq#J<;C40@+tYn`AS-xuPJ|KKaSx!C)Ff1I3ug@-f@)U
z&@7h^FNt*4XSpW6|8!WBASOD{?ZRN;cdM@s^`W%@29>ExY7Qe0uN%WPlEMhU4h(Mr
z80NXB6GgytAiusnYK{0E+Q4FFrMJ?Pt|OW|e=kfE-7>xa<?V}s#4;xri4S%B8d>{n
z`VIFnMu9NlZ;E9H%cYqmzJF|my`FJ5LNVPFY3&hRc|+kNmR!}kkNRX6R}9%Ko??B0
zpL?3ZMD0^IpSAgI+XvAytXQKyOVf43_L|qr%e9xK@W%Kkr239aq}%r8xWNTLS8=6#
zkW#lHsczZMNP>-WiR9|7`pARNX#6DV3;uMrrd6+H&#cpg?q}LBSpQKXW2W7Z@m2I&
z6woxf^HqNF{U$@kI>S&#K0db;N>dgx<M58yh`~^Cx#Yd-P-&c7-5UNO2=iE8QrB$a
zb^1@iEM)V;sPOyhDry-ZQ5g><4<uf1nWzTy$rhvBf~Lpw&1`vam%KYvN#9Ao{(8yV
zcSkhORv$~YYBAH=sbzGLAExuzhD*{=@X_o9^z&>&u81hUmHZ%v-`5Z}8s20!%*3c}
zY=_Y>>2~MmTkE*k0x}pG$KUV`<zn<w0kJJ*m!PX6mU=)D%bP4(L1Ev=mVO5A7qWH>
zn@fIuW|y*3c}tIKPx!FsY(<~&rZ{2TSiguv!}8%WLU((>Yh6o%UMQ~)oPalp2sL?S
z$e>z}opm~%tqTHmlx2^v(&=X}R~<_2?F!l!auo(sj+e|zJXRAes%BDonEGHm_-RTr
zNTh%RNUABg$IqZuy=Ba6PsqN{?IF5g67D`MC{=bHoqTyJbepfS!77+sjnh8y;9cr4
z05I7t4EK=8cZ-cvF2D2p`O8>}U5TCoY$T=-UtOsA%2uq6zLNX-{kE!Sl7TE;1-xV`
z-UaJMx`wxs+8o>9BT!S(Ih&3f*?WYjX7rai>B)Xk3QY3eBmINhG%tj>L8I){>aff=
zMh;uNzK?@O_^7X8$=d2L)!A2XtBK|*IG?cztKN-;&MLnQ5~ydC9p+uJ=jLb|DH@Hz
z6>1$HMA9C^MwRb4>xs$xEoj}z3>mIY*iI$udwuL($Dr)`BI>@dGH#%4&ENpq!b|X5
zBc3@p%MTnXO&uy~oPS6ZO47Gi-@(^YSJ1!DtG!`zFTv&z*K@}YgUWk`T2s}Q^MC1o
zK)TTU+O%els(EqUBi639!j!+2x{Tb~Jw{&Rkxt;`tKvb8J6rLOXX*oQJUkYdlh(fD
z%+3dAOR`woGmc4Y&|JvEKpS+0*K*AgH?tU4gKx+w^LZbsMZSO+CK)=2$9mA=0j_pf
zT$Sk)H~MgSy$rBbi1q03=gYch?dxHRYnUgmRZ#>557pD{gb`~jy_afQ+57hM-rgc^
zgb8xb9&aff6+Sy-lcLcq&@V=&rLHit_823uY}2UzbxY6U<M)jjjn~EXrxdn=$iA2T
z+dBg@RPVdpxi`Jh7zh{YHZWvG=qhwQn-dD&3*(Ka>=z8_`7sKvv%sG3A{_HxBZj5l
zcbkJTbX74}r}u9gy}1)|CJ2BT-n}*TUOFSr^{Jt3a`}Yko%`ybE0{*}A0f8CBx-Ec
zL|8SQq98A;2Www`53i{5N?dFYz4Ef)@~}Wj_Zn|Uk2tlONRMxvI&a&qKHuw9G1kV6
zHU&>8!BHFWD6g)6`#4awmZDMH7Y=<_)+Sli`fVtvn5RIymxs-IRG-m0tFlAMIz?VZ
z|7mFB{aBebF_FR_GTb#}sk7kbrN%{sRVh!GyDQG?lM49vd~*(}7^S8#gW5`<hRz^c
zp|1AgHXL<=1x)~U;G^cr`UOam^qA#^+7c-y?3^MEcv|bZN+Wk~T@ORb-0Xwhpu?3^
zy6D1mP1H=l({;5eE_es~l=I(#Y_*@I3!h|8Io!W8g4Q0s=jiar9mk4zpO#~N5py&G
z9oACr=ARE`Wl+0e=#-i6&=yYN=YwWfu<K16N_|S_*rpT4_q!$$jOQ=F>$J7*>mRq~
z&j;$G;sqrQoOi#eVV%}A)kMMB7L4r<xI`6mZNT_x5%%<2#R0?ev*_=CNM@N*HslAG
zaO1bd%S2S+*7}vZ#-*0IpUzloyI6$OM$>$TjNP;ID-t=lBP{A>bcGW`u-D|Nh;l88
zJ<p#XK87}qt8RJs6`|iM=`+_)#=$e%9P9jwyvSsdaDkWL9_0g)vwvp`EZpajJ^40!
z1)_OY^py3%zw^{TWZ%eizIqvgzx-`2{)dFB>`F-n)B2m(TF8wRwTRY>)fLo+G&3@_
zdr8+tCSo#i#h-Px`-=*mr)1X1*unQ1gW;hH1!pj?(u#_9waAkS`Kvr0WbNG2T_WDw
zju);YcL&tRO<Z2LwjeBMTt$YhPpqq6zFQU<i>IoUrl%c!|0N-mUROV3lj6Am^%F|J
zFPk(uRxe}rk2xE!CUVA&sjX0NB;M{lWs@P?&alU__|jG~BZ9D-V6csF4}DP2;fh5)
zgYv3UX62kEf?`~j#uUiNAvpl&B(L}ljglA(({iWN3uc$MS!BIp>2xyiTzLC^{TeK(
ze~-cEVR-7CvGPXxcvxIv<5I%t-P*2sE%p?8KE>rd(HNW2wb$j-L?6HO5~RPRF%*ih
ziG~%r)j6~&pvP0QlO>DZFqkO3efF(Q{nKy_8v0s1EJj7e#E|ElgTAW<+6R_t6iOWf
z$Aquolv5TdSrw~HPT6T>?RF9ksQ}u@s<!gCHqSS^Rpo!V<=Ba?@WT)9QDgm?k6u9E
z=`Eb@7bd_G4XpZ(npZ=-kC6G<<IHD}IA=<X{WCe$Vs0i#^qr>81xwu^sgeA^2dYc=
zahW%(bKnhRwIz&S_STTBaG4j&P4?F%d*p*^3fEe_<;adROl(&>y^ci}6;&gZ<FB<X
z%}6tWqDO3|2ZDU;eb`ZfFm)ShkTDf^sTGEcdqa!%a8r26+fGImYF#&(gyvGd+58Vf
z=ij;fubxLqoP+E6FUR~zFTQQ?|H|zkfl)VSep%Qx1{p2XOBtn9vvdnLJhc2>m9jWB
zU0UHsz(*7a@lLlF4$p|ZUk-xCmmq^uKlI;}%03OQUDtDKP<7s|+6|US#6q@8F8p*0
zk%0*-o+&r!$dw%p75iU)XTgZNm4H2Ar(@Kj@-Ln4_}H_^?V)d9sY##d2y`dF7O{DV
z@(Fd)PUku35Ui|Co&R{B75ImQDcD@TUkYynU3N0twzur9w4x2UaGux3Z`q*bt5>q(
z)R9xQMq{R;i#ZY_UZ&)^>i{>zmGQeqsC7p7>#z{RXksjuV6v{TEEBu(0b@1vD0^{D
z{<z8cE0!_C`n)Y*`oa6rrVsj{HFnvnLmcl9c!b3DXx{d)(C9}r`7w8o7~6k?clh15
z$wynO+DDCpv>QmB>jRzdjFL0qwdiHW^^hyHqN*$az3lhG+fY%gdcWt|`ME>AAKiIc
z@UZU1Bg%YXRaNx@qBOugW?Xsr9}?QqP)b+RjUVoM$|E*k+<&thHJ&Rfu`;NFQ}pa%
z3C6g%Ugo-jC;BmGa-I2{VS|(jDz`lr``T^QJVQXiWjc}L{jDOTx8Y2A+HUjnZ<s2(
z?6aYYv-J)M#x78vE+apR0^XD?P-l!(XL_6|>22zNXYBlA_LX2a$@8!A0hyELdw|bE
z??}jf>lYz<Wn0u`gqMc9<xeX+V2*xR;Uc>DbX^Ik=9jvKnDRd)2>=u$6pCvtV@8~0
zvenNt;{hglqNS<b9FM@prX!c>b{tvIF*$*>`sS?7IU&Gsn$0coDH?}E@9k$TX-Uj|
zYLU#9I+J>*+Tsl?s+~zE_M&~fA>{|ocnPtV3O^^pSbBqU&Sua1QDcwR-gTCSyJ%PX
zhG2(w`%#?)R?$fjL3#IhV&v`~IA_RDb})$Z#pr_2Seu`ecpVmHFnY#_B*bKvNYr@w
z1y9D93B-ZQgin~n3}b7xjIEfg>sGA!yP(CiRN2MP)77Q?KgHADeW&5>@yP<89lT2C
zQDiWuVG+1i<8CmiA35Ld08`kdZnR~d*4K>Kq+_&{_s~le+^EOn_^RsW-q6IY{UHhY
zB)IpeQBlMvh)E~FFK#zh?;z4|!0DpCyO7H*O7Ek#t;YBzy*x%UKYwJ%Vcyl5zyVjS
zRl%vbuK@W(KWtWzkX)Y%571NjLqfxUT6D$#!#{S^f5+?U!~M;70j<)%OWuoTEq%RU
z7Z{okeEgf*_Nf?skmD)c_`-s4%cq)G>`^!QnZS~#FQL_Ei`^}s?5@-!)-Q7-pB%p$
z3Hatub9L&Q(uLOJKO{f)zO@9y-`)B{a>V1NTRyti#o~>`5;nvnTn+E6c1Q8`vitV_
z3jgVpr>0GI%Bt)YV&a!FnKEarp5CLFutM*+mQjQvM)gqNx0bVUKQ_*6H7d=rQ*!H!
zb}aNR8c33t0*O)4IJ&;^mb<O}Zu>j4M3sn=efNu$A40$CQu}x9Bu^4-5LlT<Icd<@
zIuqwRJkAa+v|47OSh_SFYNxzI?eOME5Ie;Y%Zqkj5tvRCd9?k$@<9{PjyK%+heT=j
z^R$iU@0^(}qQ<ySMwXlnm(SGM_X0@4wY8mD5hNV}p=t69$Rh9v%Uah<_sUy+YL^f-
ze|AEE;;M`3?~O_GGz|Y5dQugc^d4VZ;D^7oHbLd?w&MJ|jLX+ruGYQ-u-8u<JbK<e
zr^?${P>>v|=+iVGGwN;=vG_A(t7_K0kj**uNvlALUEhmwNgq+GF7U*TOvFQwJ5ML{
z2rNZd&nV4}IEtsshB9{Lv0Gdtv@Y9(?`1>qyU26%O26S%&l@(M%(YvJu`;se#!<`G
zLF{XS5WOPW0p?1|TvZ+e{<4OMd@9g0rl>rjx3)t<@wTD?_xwyEc~2TUI9edoHdeOZ
zB5%*}j1p2GEAjt|sKitx_!7Ib$6V$Mo#YzSE&E0{hN{6-w;?~;HtpNKkaGC68Ys0l
zLi$wBVYM;h&{(mKV?f^EU^XuGR!>OyEfsH7k(<^ZNTgJ6-wC~8Un5*NMZ*TaWg%PU
zr6!2c9`%*gUCfU&>;2$7^(B!3HEG6<Se(3kV^>x^D`=*8Dta<=rn@zDc=@^;vcpyp
z>J3uEZ}r{+KzsQoa=W}HjSM;D@;~o@)@T#(@VGwBsxnjJmyNuxOi|I4jVCE_#ahv%
zui_`I`-kPR0h)otBA?Oe&JU)&$v3Yy1;1KZ<x;I&Oq`i7%A=^y8WAmvghXpy51DwL
zpeZr|f3~Ha_+opCu5-g_$32m%V=G6?Ds4{pjtUoel0Nck8LCRi`RUhi##|LTOJR2e
zsD=&erqllu4o`DlXx_qIws_@`<*#|$-YY4E$!fV>H-tiIQ4(w7MdE5uu|D<gonv~P
zGhG_Nl^&2uddl{k*Z^BCPvG3adElL)u=%eLd&^3RYkx@O9;~$`jXsIy<B8VE1N}Ip
zbxI<XHjbtcqR|}51&YZO@gH``d42DkpK^VZi@Z_Cru!!HDmPgrxv(N)3lY`so{*Am
zQP&+868GuNJ5x%&SK=h!oGE3vJ5s~^+24$ES8mXp52;dCNl?xLx$REyzTWj{;ZG{`
z4y*!rXnqbS8}qfG_-7S23upE4^^xCO7EUYnBf$~|XwT7TgVba-jDnt|teCvLz5o{8
z<M4fF|8ZVY)MRw8F)ID1Q#F?k7`w*Y1tW-3agD|^vlGad=W^-M<6z#~lZ>L>qQoFJ
zRT`*ZgfUIkJ)+)&G97t3``B=C*#V(AQZBPcAEr_S7WNP@Z{d7vV^Qu8RyyU{?Yn$h
z5YU%VdhKwh?UV^BcS%>n{afGii};%fUf$M|!~vnoPg`ViV$;<>v-sl&3X_tscUpn>
zF$UsjVbBgQmjl3i&+e&*o0v8dDE}-^LXm)a%2<%?`X<WvJT0C@5ExKuDP_0A3eWzn
zM?A)kU2f682{<fx`gFm_bA5;25gXi`YPTnu;Xoy&Fdmk2BI)*ZE8j|`dvdAgSf!80
z*6IXvXqw=sQn;jbIH}JJwqDpa_96?Fp`V*OncuFDvp3dzLeMMTjlq|^tBJ`B5s~2@
z-m5k!-CDh~JojcB#2G&QV8*Mr>-OEO5k#c@f--HL1&=?NXTVsKkRIEL{&FO1e@qY*
zx>6(g^YU{e;_{LEHZiX9`-evox({KqqA&XDsi`aW1j$q;lpLW}au$fY9^8jmk?+)2
zytkgMo-sV$>4?&NS!IzjO0Z5kC5d(vPQJ=cvwA9<33xFpl~c8)GLghishh60hU=`D
zT%_OxhjUJ#^9)Nqn<Qk(n|x5^XBrb)bVMuqQ&o?S$=V+^`u1Nx_;PVteY*PlZNndu
z=z%NExV$>kW(ZIzeDxfgXXH^dqj={)Hgv(}M=fBkOkz<cTv-B}XodS_os=^2G%ncf
z_DW(<g1+6oZ$ZrY%yq*l1n6x!yEn?u-#m~{%HdRYQPECdQF$Bo*DvzRlYa*VK_)o2
zev<I)CmpfVZry7C&#lT=3Qb<MSk$!yz{vyJw?Dj*>3->ZZtLc)k;_zPW8^o!>ORh{
zq^dJFLi6r%(7HSQyuZFb>Y!(%SbESmyynO#>f-8~<Je$lkXeC6dYEzKK$TEC-?E8U
zmqct)xb5sp)t=#fZ??pVy5?$lNs&jYeDmio_Ds!=eGr0EuYEC}bzh&zYxSCmGJ!&2
z3cf16=;uWLQWM7S92XXo!$_U!+mLMd;w$@D`37fJ1@nHG!|vQHGQ%SNzN2?)p~yH^
zs7{2o8TxSfO@(|B7YH`ujT=`KGlviX2?`BLeYpvOZ~l;&y=FsHq_PCu6;EV;S(x8D
z+Qi~@1~^}5bUC`GJ?Y<@1PrjhnKyHq!c_OYpO;Ji7|@cgvzqL*!}#oByYV%Xq^)(f
zL|N1szlVGbTyciCHKve^442C>N%uM*x7jE|FXAfPrX$RxsGonpZ#&&uEvLi!MedZp
z5NFl}yOAzZ%e6GkdtCs4d67S@WL%_@j@F40{+yPHag<c`P!3wb-pHu6xyjYRosk^y
z$*iVL%!!S`HZ!QsH0o%CXPcwBrz_fmcJ%UHbTN;DC+_nD?h4;*w+FTSh9~r|g^j5u
zXM2t!AMw=nyKtrZ-1DMkuk8rOUg<Mwc;v3KUjS%X53IasKpiy&$rN=kUuP3*GXpvc
z%+e&U(rKmatgGjYTldSqDg9>q!zTr%&8!hE9zKDBqlPNRYq-R#G<XYAwU4^Bb(4!Z
z=yxlk#m9Hq@spXL!Ll(X*|eqo)yfQ%X)S;-0!FT$rHt=3St!d4R<=DCB};P8)9}rc
zdp<J9VFv1?BkTB24#@waNArI>LAud)$`VEU{4NM@em`C&5ti(q+LXB;y#rp*-3D0}
zN{*}R^}Sz#Q<VxfJlYblJ9$yeITNQ|uvTKh5I#oPrEV(jx>8@uDXl_x*JnV=|7+88
zV%#bdtKVkex3|3(XD+VRL4QY`^Zp{qy)r1e{fFdC)!6wD3HQeOulhJrs)G^Cgw#+Z
zA3=EiZYx(U-xNcu@cfj9@B(*btZAgpPu^zZc+FH(E3VFUrqXE24*Z=Thbfo+*Pf@T
zI;w1-dDB!RB@2&U3FOt1%ic3>uBHglVoV!=+_)8DWKlz<UPVxzHDo3ux{WD=;@LS_
zSomHE@ZZYJ*dJ2B4Ajqkb>McWKmGJReN(~5#!55F*4xh~rlY6yP}jWLSV1*%eI~Py
z=gc^vK%Av5x0h_n`+gxZg_2S1Ki5ShB&|5c%lF+agWLRPT*I_QBsV30u-!V3TuT(`
z#qVBcJHBt>4Y_&C0aTI+Igd{;%u?m5Eu?`SrUTyfbm5^~EN`$yKj9xXK3p?6s2Ad+
zy4VmDV&x?#o_&0yT24E}K6GER@rs9Wny}A`JRkQ@C$rhrf@hy(Yc>K&i(W8%KZyR+
zN#8<#OJY$sYpwV5zL7>`5B$L-g40u`*b0=IFn&yDMDt5oCOD7X&M5`|a4T_S<s#g`
zP3r%8CuO+^gVLrW0gQdZGRYGb_0vz8cFEdW(Jx5RI+^xSZ2aN(Cr@Xkk0}a2swe0w
zCO@;2c%h;kA0MCcBse~mb(fWOM^I+;1x?FUVfgp(uCTEF2{9-61QuD;ddjN8w3*EO
z(9O{5PErNm-h(Hfo941vS2yz1?q>51`opZ(`?3SG`5%yBj@15;)NY0b<p0hHAc5K4
zuo*+6JLi?IH8Bq>sHwy1bTIuuAou%qrQ*<(F{;ytcjHG{S7|q+A1dEvWd*^x-LUE6
zkL5MmeiUpyxf7-B-;3GIrAsu_Dok|`EyVf(oEumeP4g$FCEvbHo_O8OsiInKZ-)jy
zYeX(Lueya0o~x6@nZ8)gOy4VX@`e>ONaM-<iqZo>@n1g{c?Tv*UGn()ho2$jFMy1p
zbm|IT6P-q@Xasy$n!jX=rLMO0gHix(#7{hPNA;^S)%r|9-1sG`S{cdKE&}+u8I2_8
zI{751CVe6i$~9XR8NjDukgm!uJV6i*8l}9Do|w2XyB8Y8akn9}dCQFOBVdE31iU8)
z2X(3{#Tm#3Nk&*2!NSCpnJ?fFIc{%>RUy6p5L6C(H@bl>j_}smk7oQ)D*OwmvmOj7
z;bVRaLw7?+HAHh)@?S>NjkKfh@R{@|YtgAS7WB1}nl$z==^6z$lL{AwRnf6MA%^?b
zGF@<eoqgJ|tQqtHR!2f&GX$q~UGI;I!XB-WMyeZk!PY-%QZgL5{+Lf+!Nu3|Lb5$L
zegpm=_TDq9$*pS_*6kK6C|&v%1f}<0Y!rb6HYN0cv;YBu^d_iVkq*)!p(rIlLV}bK
z5}MLWkP<=*flY_dr3ff`@|-{CjOTsdG0u3$_vf5%^XtAB_sB@*J=a=uUTa?SGM2Oy
zwOm%wduI<CEbK6;G0kvk=+Bb*KvhWnEy!ceGk=Y^kjK15`Z0pMlK&amr4rgevm7sQ
z>vii$zQNUqiN5F_P!xPGl9*qrmw)0`n5Ggt<yBE(7vIOyFlzZ;a3GnhCwM2E8Z<0D
zxc=pmw`t?q?UHzpEH6F4+J3c&B~@AQs9pIwqa%L^*cnffv0$Ba0!zo69$ry5Gmri2
z^XPr0JmjtM+%ebh`6ED35B_|Lr2(pMoi0vIy&%^5*JnL2f^{qMZSd8trP}h+2~u2C
zhjh=RyNLo7;{3XSr<fYI;N01u%9iX;9@vr_0jky~zd&Y%`p+e<wqT`ip+E9R=Mgq*
zgdH?Itddez_9}W4#h!~zCM_mpm)qUoOSR*|it_7dh?^75<rPft<Q)(-svkj@dz%1>
z9Rr_-@?LH-J6<?S-8;AV?XRJ&&ON?g=QIw#{fxGIIhT}}8qNA4VzE-gxCztXE%=t=
zO89-vM&Fz!nKRH=>8jIEJe1|&!CI!No%zjG*5XUS#p^{@c*6yAb5^tLIgtPJg`)R%
z<@FezzCC8yQACp6ss%0ure!l)`}@MTy;{-9Im3jJE=e}1h$U`#$vDklnPIrHnJv|k
zD{A}38V?JgDt!mvWZKuHzTShIVsHd8FF_<9xpZtH`#8k<(YEe)*5Sq3P-}<&@<i<=
zl~P?u=Zc|)nX3-{x88Yj&m}-GSBYn&LFw=r2>+#cz7dkSqI&Q#ukL-<bjMz<%?$sC
zC0*mu{PKKI+AREgd`iOhni;W9DeV?Ct<34DtT;zxbu}|9VCng(PNpqtbt@<uUG%zy
z`uEoPF{WL4{zaYX?!R|2%Kk_Dq>?B5Z~$078`ziSrSq3JvvrTZk)f*a;I)939A-~3
zT4vTk#z5)AUe!w`!kT*X11~q2;IKqvMF+)aiDQvGJIllMC)T#^$-k|OOU}=;J7hoG
zu2~{aKd6WQ2_RpMz$Q3nq@*4E)-zYQbZ8o`xjLEhH|F4)>c}9gv@_X3!~hl;Ci}4A
zi%U1<Idpt_7U1#cT)@Fhcvgp7+K3uTRCJ=r^5MN`=-nf+f2C^Ie%h@(P|0|4*{+dk
zY2AnedVuwv0<#&yiQ=uJ!4)U0foo;ILlt+~o{VAJS1wGS6ns^fk3~Y<+>FM+PnGs%
z90o?Z{tC~@B&6ls--;V(*Z)nSnSA5OYf9+Y*l8a8<Vwk+WEoF(?JjlV`Qr=a1@GL_
zD)*kFw3XdLMJFe9JFDqCRkva|i|t3<^6#1Q`P65r)tTzm=Y4&0`ObVs_lC(4twH;b
z>9Q{|(AF2I<MF3Q5(y-(mb`6X&*lz)`1f157S3ry0y<&iU+UbQ8(-nVMpU0nH#XBQ
z#l)45KK20-w~J4Mi*W?x{UVo4p5&J0Q%p%8QIk~I|0)bC7Y7utpP|$^%n({U92fZU
zRbBv57T!$uxn!z(RLZbp%f-_kSS1uwfXUR)18v)Uc5!{@R-VS|6jz=c8`qnZ@2l7Q
zGV=X@9+&XCHw|svG5xAs{(bBwYutbS(7%sfFll@h-QgSk>zo3kW-fl>+&GM-t;W3L
z8>#QtIhw>dqxJoz=ZuYTQ^(;%C?_X%`s>N_bL#Qb+4A{FW!jep8(7Dcrkr=vfO<#e
zT<ugO-Q8`%4`P%SXu)gvN0Em}jA$l4*m0g9(D+6F?iE77P3Iv0%QWo<`a$jW7KPE7
zZj}UQ+dJY7@VqoDZ3jwBSE8TpE0*6uL2)H_i|gK<Xq?>cc*KP~$$6r>806;BUN)=i
z@#|bK;tA&zb<*0#EAP9mE+qPq%+vb<Ssnzz+-J+BU;G_}$$x&m%2g|#)tGFckmtEt
zB=P8k*U$vPI<0y~drW1Y{uuJs&OX8Df&E9d=B=M0w>1RI{#GAka{F`XG49jSu<+>h
zKCx+d0NZk(#XXo*o#Mmd=Q?nE>$?cW75!V^V)WCNV=C+@jpCDTEn1%y@s9DY7Xw6^
zF0wpE1j<k|pQI=anP7>EB>ni?4P^caGQfpyUik2hyA{3q8f|!Gsta}_6uw0@D~iId
zSNR&dp+x)lXZfv{6sXA)ZnsKv${spf#OF5L9N&nu4Ye(MIn|6v61?Q=NO4T?QVA_w
zZ7?(jjn&Rq##Qy%mWfUj6BNq_j4X>~zn5qwI}r6Vg%qwBj*v<ijk7zh^&j3~wklf5
z$*lU)`|Qvduxt*nPbrh)f0M>@xHn8{cW>WV<oGM;*Eu?=ipluj{o_Bs|L>n`#Wr$O
zI(oI8UrR%0zxPx~P2>4$6qLxR%=2}ZmD{W6Wn%p&My8^cbnFg0j!PP9sd~J`VB+R?
z@R+C8b?WLvo`1ibH1zQ$v^sBCl$$xphm;g=N2u&BbaN*WPhD0Veu$Ei2ss`K3cG19
z`}l(!s*7ON&iDg37$|w?%K=`q8yA4)#d?`eF9jbc=UrraY+>0Rj@3P|aQliC(S68Q
z$6o<>vQ$3nd+qZ3$#S7V(P4qA_1W3O4nv&2$w$vtRX!lyTOL)aVGHnyEOSc!9HG^P
z(@0;?j(kRkbZ7B&sIvmchYRb_51=yg6-_bYv+DL*6Upsf*S9!oC{=aLZJ(a4Z|aOS
zC~XtpZm04^1e6B&hrqhA8y*>P&STB3-btUu?=kw@YVrw;Z=U`-r;q>Kz>9W>1=U~d
zU2E&{vJns2Sv)w>duQo5{>VU_OR^$A{<;aqwJHL@gFOURdad1)_SM)oCSJN=`k`wA
z$U||kwyoT-jZbr-D38oyiy{TEm~X;HA1%c}Nu=Q4={6tZa$FB-7w_~jLD={@oL)qH
zZ$)SW`|6^qXr3WSAb!2|6!$2V04aqmtsT8JHp$JW9ou-m)PAEXlp(#oNag1D#NFR(
z|8qzAkE8z~S5jjLZojLk_fz7*OFIqRk+()R(WyRXO#%n)+t6cpNqKf+eFp;;Z3S?k
zZT3Y)v$LaA9Wz)nqwFh1SW?gF#qt$R7tWdP2>AXwcdP6?kE5IQQCa@4b4JYWzs|j0
z-(~r5E-<mR`BlN_-lzGT`h1PYRzP1a9j_K(Kr~ONoxi%2byxK$u_@4JERGE`-dPsR
zc7-MXd67_hI=!p`cFEwuxpOkj#;3K`$2Bi-53U{j$KUjS-^j(HoEJu~f3ofz+Mg$V
z<GL{=o)Ud<t7jLje!g)=<i*Ev<j<3`ADqSKrJiboaGV1BK>2VWC}=>2k5uOiX6r&0
zC>;R15xE1szGTxe_^57)sfXepLiGv6zGo~baL^SHq+g`KC(h^oxy6Revc^^BZVLaD
zuei~l6V}5F_>O_zy)<t{qL94aD755;Wbcg9=js?<4kf;R^I4^?Br1TFSSD@~U_hxf
zQf83l&M%HFj&WQG=dcVqs7NoBcm(ynZB=%tq7SS*=x|RDej<ABowq2jp}yfV=~MK0
z{n)pfjH}zIh8140@p_^+lXQ|>>l@7Fa4BQ2mH4sSJvVc%+^H|?de4xB5V;|7=gB(}
zMV9L}yTt|%l?A2r`BHh_l26B{oHK7D-|7AR=w@cl%eYG=Z5{snnAY;UU3;4`t%inT
z+i>D={5@;K-@#OYMPZ}Xinye>)Skujr^5@4-I78C`TUSn!2~y5|8<G?lv+2(-z=Wx
zx=SoCth0VSe>&$Hc4TorI9!=i|DoOdzQZ-tTK2QC=Sf5ZL04Jj(Ip62*zjK52^-<u
z;qmOX80Qg(nD_;oS!LESxInU=A=q-DLSe}(_-09M4x9U=m+ifzmyr?Ec2LhR?d6qK
zm2daIJc$tXx%|;p<9o$pSn~#}omaV_e*@LGhvT)~@HAVQxVYiXXi(VUMcz`G?3BWS
zfH9*?6tNdXPcK$fu#C}l&Is3qK=kJ;dP$^au?tq;s`Iz^O@q8gnEHcKFOL_vm;iyS
zVX#OSny>gGo8|{qE@11n-^(%!=Za#4^vQS9F#dY+{gKmmajC#qxa6;ME5TbCyA49W
z&TZXdE}QCZ-HROx9*yOl75froYH*+dHqkxY4t~94=HWC>)h*9q988h)4|_EISEMl1
zL?#0R!AP{%p;{$tdPeYVIWEquhP|At<E_m3s`rt)ALBYF2R0Q>8}2l07;P%bq8=Xc
z<h=Z3Z^`Rleq5C!=BlQ1_n|y_U0nhph{C4hSOZr4I^)R5=S}y)e{Y@-e_6Y!G-v9&
ziudyRYyL<@su70M2~jl_b8#>y>d!xk;H9m@TPw!Ab!KN7_%@YS$`hc<%O4`F@1@OE
z(&OH6%S-fcV_&6Al{YzQ;=<=QeB?<H)0ST2^-AM`&OH3;*__WH{c&ElYOj-pjaf84
zXT%!~pzPG8?-ZXFi=5wn5MiS>t7)qO%<^TH4o+J1w~+dS@;P3`uJGAxxVgK5GvNjT
zn@5}{E)2qW8cUQzUhXOYgszcnEq&_Py1zpBAF3dCDkXi>Jq`EoyqGJ|aZA^M7>Wq5
zT=-JbT{Gh%{;+$^tKv56ou+Jft$^E$3Yi3T+?DJ<qd1OO;eg(z%a`i2xTMy09wP*`
z8q@796hw@?@QbbFm_A>vfl7N=L+z$HSC**Wq}x+zK%mLT>)I!5HGsPu-Any;GZ|v3
zLMZgRuQNY$%LQXNb2z+zof{l<0`(G(FJB@YsHcW+t^PWfuhbfS2H(ftD4ZK*kv;6G
z*k;sntNl7hZY$-fJCmh9h<wL&u_1FkW~(Fo;&`F6$^_=#IISy1)>xrWQc*<toA++c
zAL9_cLxv{Vp(u6ox7DoFKTUL*Rz}E!FifYmO+4sbsW4%DP>lMKZQ~li;HKxp5B`0)
zJEIwfIl@aqw12H)(gw{)6fvcvAW+KQTDQhMhj`V%j%>jbemU?xPNeFQA(&HK>xAds
zjcGlQ1987H%l<c@c^*^)edlq?VMMOSdtbR{`Mze!*~#_6zI;)?2eFPA1{uGTgl>oz
z3t!X@x>_M!=?xR`d9v$r8&)yCFd``FtH8)Hlre@6dB-4SJbC-)7p7fv4;Pa(Lmva5
zEjdQ@FPAy+T@ndYFgv`K-O=v>kIRg*Eze8iIm(d}q!}8edzzb<ii>64*459tGmus~
z8ud0qWXX;3_q_SN6ns#D47U%hy_$55rK;uKyvAYYby%TjM)lpi3>3Um>B+|BF%R<H
zE8ueK^o1$fd~S-dLT!g|_dQvU(H0sfyD$9@oBpa7A5YhA45VzNIjrbLiCMm0(T!&N
zgIk<iMpthjM_5>RdEniDvBmxO4q1*3lX+cP^m7E_{R>U?CcpCf9toH#=p-g<WU18e
z3+ose<|v5lo%=Cw{R!hi*m&Pjo#L&GxLKmenPHbiG*-!d?5SKNyn=NYN()q6x-f7a
zMsvA^eX?lcc8kVIO@af<rJ)~8{p$?(O&|6F8cWkn>*G@2-B4G`?0X6Fs;H7^zmsOo
z`P^|{*f7W7=la;;#SsM=r^=&qU%}Ncf0$~{#jvh1y-}Vq>2~ieu!m&$wyxMQXFR%|
z?rHG2IKwEMHQ*1SF&qCs<l;WMbjR|vMqS>sfofh}7g9dpalhxQEwZJPV4Qc9do9}W
zM$8bLk~Cw=m!<vMJTFf;9k(tGAA%>#%E(tGJ1VY{J~?2N$jl$I?R~TD4R1HVAKClX
zb32P(e#mB?F4JyguIng#{-mE#k|dF;A{Cw_i31p=jJ*X7bR%g-!_n-!<tML4t(_;c
zbYqNB-S;HjrNPQiUYq?Uxq^k1Sle|5Nu}oH?Th@uee1t?ga1CvyfgA*1B-~XS^Zx1
zluhB^|NURxUl>(ST7*p_4`-s22Rbo0vE}}XYd9mg58TV}qh(;Ok#A{P1F2BtvDclu
z=}+P=)gYKSGp+xg>`2t%C=Cv99!wdlsO4KRTDvcBpf*s!@I&sKr3_ba=zSRt!hwQn
zESL%#E&}dgde~VS_`~E4d#(?{#I&rYJQ>As$`xsY`e5Yqm*Ia)ao3{(ZeS6S7p8JK
z)l@|#ho;0)Y@dX3(4e0DpUFS-s_Ds5I%#M+VSnp<P}^ezi6BjNONzPAM!I2jO*D~@
z{JxavkilY~ETL<eW9SecdsA;zM{oM6_J=ec9v)T(?E5d=cQ-93HTN1@bhcMP=@CY%
z{@axzxBhrv)0{-}C)qKjV{WzYHbpf@Mf~*{;melJFn4{}$b9XLXObe9ruZFYK}}%o
zfT8Sf_Fna6mkq<xGt;==UE2JwZ2NEH|B=?kI?#3PDTkJ>N|#^h){^8WW$&AfuPW^>
zpwcwDtCaSw_CWlbkYS|WqbiZH!CdN}Fq8B7*(I1JCY}tjR;|zrM#s*3kc<lMOCM)%
zgZimCtXq8UYTd@kjHs>0-?!ZwloL1(g`_*GZ3ImO1Y8}{$KgIyhkm#-7x3#Gn<l`2
zwRx5K`$E)*=~P0BP!nnxBT$~04uzDj>mfDYj2bnfij(~UHm6V0<|S6P9e5*tuEQkb
zprE9+cj@yb)#(Ee*dI!+q|Hrq0N>?#ddp>+&P6_xU_HC7%J53pnt)A374PBM$KW`F
zilWu=D=*T`g32^4)bFwn$!tE&6Md303icf5p~hQtY4c55$qHTWzTG=fut0yr{qTL0
zBe|Q_r|45BS1S^4sCgHIw5Z$rp`o0}SOo0j^$Ln_E=RH7=(niuxE}R+HFg{2o5M+o
zqWInT!%uVW=+N&rRDWJ((9Era44?gLw)4Q^Rt8@=Cu&CJ!lr&RP$SJ?wbZ6Qs9aUB
ztHQ_q_tBT<&izBs|2KdAKfe9{0YP6$^Lf5XNLr^XNA-sp2%@K<cdr3#as{Qflq26_
z_UVq=d~599^Y-R6?YDNB9@3n3kqsDy%*rQk<2`7<vnE=!?<ca%vIS`xFVOs7l-Jn%
za^x0n9E=nIa{wKe`f{(k;i`pC1MfXeA9B*^>Z(Z3pi_m<FHP>AE&5dra<Vk;xpgS5
z-uP%MypE~!jF8W-oMei*J(&gzh+WOsBkFrhbVN&T6+HyD#;~Tn-SBaJBQ~lc8R=R%
ziPP6ZHzo8oRg$U$G$_x0`lIkSzQ|al2;kszde_+E#$s9RGW+eXT6(N%3uU;m<ucy-
z|KkkEZ+lWB(#~VP(H~kq)%R$SDs4tLBWgAryl*u$zN94b2s+<BHZgk5xndz+@pYkj
zk*U_7xz`<%ZXB=3Tsa?Gu6~5@GWh_av+$+Qhgck5E^mL8VPR61CByxJ=y^6F5Wo3x
zvF7v$?16a%6ma!6&`Afx-D&<mo=yMhgG~9bQO&RP_xec#>36ivJm2QJU-j?dr2T1H
zoOvhNh|%?~H?eE;T=LGNJk?1)zp&mDxf}mn{$q!KoZ$bZ$-tLa`UMYie@Sqw{{*{Y
z77AUi$OSkSYhH;Ya;)mY>6saKHJ%g2G8hCWrsD{$_Lg8qZ-i@)PM+Q{$hF2cznJ%Q
zk~+V(fAw%8^@sGe<(Jxm&W^;rU*}>zJpXv9gd{oPcz?HEHv8N`cud4b&rhzHub%`{
z<*LD9zs|J}x7u77S>fEM7CN@$9ys^Ur~hM_e|+Zu>e+NbXw4<PIr0xX4x>k3sTa!(
z70kpXUfX$@@q{okW*(lfCrm#7D;W8Iy1mpj{{Qtu|9e*#TSDU(3BrqnaLcZQg%6zu
z7X4w#nHpXeZwf2N2k6OH6jz5nIV8P|kU5;Zgvq@hD!Y14VB_Nrd%5Usr!VI;C!Pv6
zJUyFf%Rd*2{iN;l`8(l-+PNL@%ZRDh-vWL<J3rfW?y&CC_VSOvFSLZ5-+<hp4V-+s
z6tQ_>oFvqgv;UIipO5{=2>*D(Khxo#HQ}Fy@t@t{|GO>nx0SN@<vp(IEZNtT_X?xT
z9o#dXjpE0szTzDlRtrRZs&wfI5BD8z4;_f}_Wu?;{3HDJzZv5{UY@@Z`E5}XUl0Su
zge1DFsgyt+-)pB>yQ{w2?OfbN$-@V-j3nY3RPYTSY0)cztdT8pe|(C1D_N?SCLHLe
zGvhkjreE~Hi6crh1mYCOtFkcNWZ+{@38?qg6mdc`T7*&d<;3gMCu-ylW+V^nKXnr#
z*T`9E8H^r?=unJZ4>l1*O9Hk&cN$jFDfctzYadhQKk30|L?@}$=v8@#5?es3GILG3
zl!#@e(=Q1ODQXou5{eb8K&qzdsd^zbD5bm}=#MND0<SB>bd(XDA8@b5Kv`PP(^1N3
zP~DRC#~<9w9lGhx-(0jdgBxSS%PN`lBXSlTdj@xhI;A(}S4t3xl^&Y~Lm1!Zw8vKw
z_0B79YlLsRvTloJ(QrQtmqStn8V8*AB5;6M9d!4v$A6$V;Anit<5klHmoMm7iAB)C
zDTX9xiNqjxB|gA259y7Rj6-?lrMTjY?!hRgN|E?v$+0s7G7tQGm36?uqBcoRJyiLq
z(fg+KlJm*M6rHntF0o0Rd~P7E+Ol!b*oz=tOTYQuGblU#z||!>p_(1P(mU0(c&YO&
z{&h@q19p}c7G@i!`Bzny7Pza_nSOjsP`2$r4cgv1s=J4aK!ZRqn3?5>C0SlECfM(Z
zz_lloU~P%**JyAP63Ro28fe&Y?rO@BtA#CE2gIS4KpfVfN>naNxB}^#yR0%PwiGm>
z(JnVLt<^_K+vm3y?6<+>fVHQBH<y%eZcV=ypB!;WT8fLy&KC5K!DJDqGe^f>8>NzH
z<o9#aRD}Ol65^<~He{u#%2S<3MVRR0<}|8&Jy~b#yV3;543a|4J#@fxc3N|Q03cRN
zV-2s9Ao{Iawg+Y)qN2Zi?OJhVJSri;XCwkHk$aYh%EbypGy~atydKItUde4(O<m&7
zl-VNNl@Sd6J^oy3z{&+FiU!v?y&*8)-cTaGowX2K^a>XD7dK<;&D4K9)*^E7B&-s>
zt^vV#cRAhIvKu?tl-8%Ac2)UjaDJ-1;~=gL__LKfYyzxH8K@|)tUIWUymjgBsO*F8
z-nLY$xp4`!;2~8R1l*L7M1UwTXbLQ_b5N^C`*>bQ0k>%4Lhfq>mFBSW)@+ERzy}ok
zJs}cheB$QCY*(On4m9-LQquMng$2!R+Oa=NJWLYSk;|LZFtY%c2yyK4aSbQQU@NH0
zf+Ej4B*fKj9dfS@ydIxhOtFUK?k~W_2ysV&o&Mj4B@@ESFhuuo6XnD@e3aK;M}{we
z55}grZ#_<%C-}Q^5!i}+5z5UCBB)2;;Ok|u%6f6>(s4HfdPCc(XtVjSlC(FtZ2|eI
z%~fwSSitTuQH?J_W)_E>pCb{VV-v9)dQ?Wae1aGuLAWxdGH$uLu`o@8$<a|pRP%*R
zaaC2K+Q9^9QyiJs5`gdrnQ^PHBcb#{%{_r;a!y544jHUT-rV_9)aYYF@?Z0LI<CSf
z&Y%^DZB3H72{EKqWOPxLv(m!;uL_c;Xs2AP!=pkwd(nH%ZByR^MZ&hXF!lZlQVZ@d
zOi!^d$)`EN8!f~w;xM)sG{D2XRtL0}sDrF9mI5+NY-t2&#!&TrtPQoT_kxE+mjonW
z?a8~f`!M@r7fFJ?Xu=LfkKMA16=X?;d*$V(K-3FwU(<NJlH9P5Ce|Q%&8|Jby?qNS
zhYE6**T_n3$*psajRpeJ@%%31?OJ@7kZo59)H{Gol1f45#&rzqv_Iwc-Szu%PQ@La
zLpM}dzRjKdyzAreLm;(YRc3eaQ^)6twE<Em)lLOgml2x?-n8=}r)79tZBl_YfmA99
zV8wt6q?pZpeFnt$XgoMd(xC1GbR0$}X0srmht$bH7=^TnwyAd@aLmAw2@*}5)*PJ_
zrOrK?oW^<+MsAoUFotpd2y8XJ1pN5DxgC5Fz{BdZeXH#468?{e87>_+Z)MzOM);~r
zujCpL_!5ntkir+|Lq31mGwuZE%zr@?#3P}66aHR^z#D&lDLy1=X(gKqCq>%wsO+sX
zbgF$R4j>0RBx)tMFgFk4`@$Oc-jj&LM^grlf|?~y@KV!om#t|j3uT5`9lGQ}X8gJo
zt-#eOj>wSxNA}S(It^aq36!-Ej7#7l+li8VfH9ii#2sWD%i{2bj^M=xC0@$A(`l~3
zYU4kfJ{lz$e{Ud4>**Pfqez(MeBth%s@0y58+{t~WwabLKn~iJ4w(%BSM>FCAKFls
z7DiHmU#b$Ic?1VDGW%TNtwi7BjpA)^C~&yu0BFJcvpnP0(avmBJO~{eWX-6WDkflE
z6U6RA#3I{)-yXIKNKm=nK$(cthpnWIkC!t?mll;+t(g`^RMn+rRH)CE9hA9nSv?`k
zgp~a9X`|c}`b`KOVYn{)mmoX6Q0=(9zFBf47{JVJK_l>{=*X0#%+ew+aA7p{#&GB?
zH?w#>QbO@^fyy#m8_WrCVC+=i<t@7{X!fx}P*2{Jzx#WdvRqAy2ScT7;NmR-vdcDI
z$a{!74qY<n6k1rBGBzY|YycsLee)FDINZ24oZ!n#(c<k<^0*tGR3u!GG-+wgr5I0J
z_7k%-GECnrP){0FznHS>>=v(J=f9mO?(a2oL+HN2-8jK>VIYE*)x`5hQ`6fy4cN8p
zTBVk|H|%jm85%1sNKY=YK@!P3L_AUTP&0J<aDrV0V-4h=Q0m~@y1{!8HQy)~awL{q
zx#0ArTmq$_AhSS$q!zolK|mC)ZD1fpFQKA5u4o#izBLf}?R6)+VMt6dw;eSUn7&Fm
zm{8kN%w17xm&CZKznj9@jlC^%L<$7h#o-OVr7y1gdm3gKnxEXH6d4p<2(Ero#9lvl
zU5VM-n0lIBj@;nQYB9FK<wP=3T7<5y1XF`YUc<Idu!meLi57b70Fb;gfI+*SjjZN&
zU$Z{kBsnAvaLKybAZ}6Yky{`)TB0R|Jp|>skXHy@311wbX?WMMsN1nO54ub9eHaA;
z$WuM>v&+$O3h`lwSxy%ETahLYw|$~5r_RWCHqDU^bi+67Cc%k@@bFE)(%7}1bJ59#
z{HF1IiBVh9UFF>^Wots{dIzw}7r$!Cy#XxcXY;1Fl{Toc11n=-?JyX0(77jh*Db80
zbSK&GP>d_BV&3soP~QXFM@R-?lfv<wW?d+HHLBW{-;`fh_v<ux$N%0abNafdOx4n@
zaodVrY#=@93Q@|e@i5zLi8{RWt+zfz|M83*=&EMez(|1?K1<Yu<m}sk9R8Z|TAN3k
z|J$X~ojBf{SZd%iPpY3La2o!0lTd`U9c+mY<Tj-Q0VP&MIiA`zMJULCC!}b_s7i#t
z9-_vwiy?<T)%p!9*fFx{&C?jJKJG_X38MQfP@svK6Olra!ZLta6g)yW->5P-o}Py)
z)C`z7OWH#YZ|%}etjlj5<<eykMqSTh_dT}0_N;CFF_<!VD8OA%Gu7k95O>EBLS{;d
z`<^KP{a~in8s9QokI~n|t`QM3KD0~W+u#k{Hx<Da0>(G-R)nAm##O8vNRzdoxQcDr
z9v}Gv<8Vj^oYIKp%wdjap@vIkVyJUib=l6DsVw00=>9Iu`EN=Qc%~`$a0JY;WYY~7
zPgJH@J6Y)|2g=KLDMyC6w;b*PMp}|wZpk>ko1ZEytI2oE*lXOCmzhUp*lk{|tFt7)
z+@ODTRB60H+|zZ(j1hXr(hWR&ekMjIY9dBUo^2;LHH50fG$^PvB*$ai<|cfg+?z{N
z{a=!JeI}E2DD?FvVBm!1OkA<&ig5Vmw(8NZbDt+~+3*l1dBP7omL0l^lZ>iHKt#+m
zq7D=ofpYQ6O%X}?@k6t!Aa?K7i1gtMR8L>CoUomzUas6o%fx?OmGHI4>l;WcNCG~a
zsBHZ;<)Bz$H)mPvdkye5KEphHUTo!{{6K;BGl3W~I2Nk2Lyat)-5(r(_7StDHeHag
zyD{S9^=-$Yui#NvE($S5S6Rn#5s_lFFRj8}tMikBV(YVi%#LmT?L}y0N7{8Y1ym6?
zt+_=cn#L?R^x*VdIHb$M+FEEjMQ|Xx*+I$Pd9_S5?eFmM^3bt^z?tYb+_o00ci+Ew
zcs8<{T5SDC@XzoFSyzH)B8J|)lR?vVvgt8hJOCZsrxj+6wB7aI9&Exf7)G*Roz~~7
zYYsJkoiq7$PU3#QY)KURVN@dRzW%RscLTIvVAtuVdPUYlz~=#BuNO>;=6C41L*%-^
zU+3)4bbO=Onm8h=-$&i39*>|`u8v2kq@)yvt)Yw5BDLfE!6V6sWV&{@=9`si$ZY=9
zVpe*|at|1~Mf`fN;VV4iY7K74H+w;>%ri<Mb9%g-suFK#L~0c>dZ9Z=UPD)G(li6N
z3oZoC%!b*0fr$<>b(gp#X()N0MkS89b$@^if~+Gjk7!oNF@UapH99^w(C5a@U9AT-
zm2;oS`*I?Hrex@{dhiq$W&-5iKzbf3gIGDb@)C$ht;JMm&zv<6@aE(f*VoP`EC@PF
zSzErVMeURQy}OmD5-Z%*kzi=7baUF8Eb%+qz#*pladOc0^fI>Z7f06cYsSqj;teZn
zGxi?gzR5MIxWg(i_gx`yH^pzZuO;iXNw8`asC=?TPo=WB+S<|~7mZSnTWo&L_q|J7
z@hRl*r>kChQ-`YyGyTQQ#<>R;FhnS1jO^fu&4V*y`vtYdg#iTLLSPAXp8Onbwzp{^
zOZIHU`iWeLD3J!(8<I;6K9sa*-rI34*AVgNpe_1eDbLsaD?$;i6b$t>>S@gsOroZa
z1crNRDh6z8vj;-FsV?KIHF+PVzSJbU@@}J1UY3iF<vJ8}U9^l>j?AE?g~}OtTKFzX
zB5qrS>$H7t8sb>_IwtG2AG=O7@0R-nl(>P-fBrn40GL9dT;j`x;gLhuN<(#T+x$Y4
zyJI#H*5U`*3_8l;U9%PH-<Br}vi^oj%2_pw_OovFo>d__kHt8a0Vyj=n>miY8A(W_
z+}jz|n_n%m&v>GK`_@*v6FZ$6r|I{63y6PJ4r?*+ns|j-pnt2~c3+={B%&M`hPocv
z;L<e^K|xqrfEexM2d7ZNdBqUgoTJRT{SBMC63Dv7+u)7UsDej`NxXao9@sSTwPs<a
zi@uJ=N_3$rb5jZ>%HvlHRS*z<y`a?a2nD-7y@#1zo9z8N{!Wii9s_Odg2#L(8kZI`
z9^}2#EHt#aljn`yc;_oyWn^f$`q$YVt@9I6d|C!Y%->j=O1Gu;l>}rR&s<CFNOduE
zWxKCyIM)&l8scZe*n+j{-ST1=R0JdvJDAoiui6^VJ1Z!r&zU#S+j*$2Y>(P;)Ut=8
zSH>s?K#Ksil<+OBdUEKGP$fQ3yZO*!)rk9*CtEQBD%#Ih43h!f8z?vYJBlK7ETnjL
zAl@a%<@SQ}p<27WeY~;~^kG_Z@<GIzpyszxkq!aHF6Ij_e3{#j^oW!iB$XVT1cV+1
zH^FtHNZ0U^)z^5p;)@SMR+bx)1eADgd|cs(+s=KSgXElQr{O@QC}V}6+WFoTr*1+b
z5LF^UVPx;ZTR{_ptk#@?{DDKvyVI~AK_$Vb%rP5X%8<vetyWUqHfxgH%$j%cSS=5!
zrY=Ft${@@DCJ3qhWNB?#@X;V^X+4$3J-Wb-7Sst-MIT>Sz%K8WgDD4Fp_WvY17m^)
zGQ&?|oJrT4$dp$-m8b+U^Ai>uOg5(_wghzoslkK%BTA$t<`0A#{M)(DGitdLHBug+
zEa0sZPBRc)Co%5@3zmo>jc9`=QN#$y3rNLkn_LQJDEz&^d8%L&lqrU~R)~gJsa>Cr
zGY^1XyKcv<hr;ajKTl?PoanD2ogwawc)LI}&-FEm)<pYcjv~zQ*FFo4U>5GC2y$)7
zavV6ZhBzi7VVmH`cew!+xx)d%apwAIcS(suxo|VUKKBp=-qKj6YlDMz?FqV|E;CDg
zWtyX!C)L_%tl_6sV&$?U8C|D0a3r%phAz!@*JZH7WE>I}gb{myeb(=mpdR@iwqShJ
ziWh9Y-qZw#s!6OeVJiMp_@|6!F!OpgP8#b_Q81tav{3M#4a?$~S?`MKCSEfPO4cow
zzO(nD18nelhwWSJTf2sm7&KH<)e3DmFjh~OrG6f0GGVkQZc5e%Dmfu%$kFhyCdO6_
zI9hybL5)~082I2t+}2KPJW|$rQEFck;fdNJLWpbY1SP<T;voB3UhabHV&|XUIdD<x
zY^*m|nde!`8|aa-Ca_MDqyaty7T)Hnoj+B_wg<Fd_vYC&b&<+af+@{3*hO7YxO+1z
zP-M>O8ly4KPQ2s2<Ivh4`sC-vlc$iPmXMVkg51H6P3h=nl2()j?<}b5Iec$VvIewa
zIl(cZf<nbJV1>m2R_|GD&j=<Q$`gN}ZFROKtMZeRI*wh?YoaVpsL-)9)&UX;P<VJ8
zN;qsk#o|h-W(8T=VKhpP5@upM+&;FY*^XmqFJxW`kn$>C$)&&y1(DC=<Y(Xfm|m$$
zW3d>fg-BM{;us~s;#co$X+4}ei5B$R+WDyXb#H&aL1$;C>AtEa3!>fyvg}*np{mz6
z&{ce>L@N;zJK_wjvIv30x=`zhnO$abq*IkLKig3yjLRT#bD9z~BVa^9WB?fGg(epR
zz1RjSGN`H+A>t^nl{}Qhdf`n>MYPvl*mPvks@io-tG9H0zK~qs#}lr1n2!z4Q}>g>
zz_O5b5%VfU=9m}uk!9RgpD*FEZ>Q*z=%~5zX!3{Cd(D5&@|k*X&zN$AaEZyRQ;A}W
z-Ccd3XO!JsT{j|ND&-!$XI8s(JZM9uwh|xWOHl#_QItSR-=3ZsjIuu@2lJn4th_?c
z5MCj&V!-{f&U)OEU2Oqfpb<`RLNOdULjeE+JA1TGOhP0ZfolE!&;9N~+TFwgG<YLd
zK0K+a3DDKYY%ER+OKvN&0x)aq?|>Q&mdLUCd&4D$X{oZoTVmU<vECX2-y3#jM{r;R
z!<OhV5LY#rTqo;1)5Iu{#!Am%NfWvZgvmml!m3p_E~cj*mP-gCgE*F5r+SMM8sU}5
zLn2I184$9P%h=?%NFn<V@>7*Y@b`{(#sp_@mEO?4{>3O&HF>2nytIGvg5gxgo2g2%
zS`kyz_YA5TIF{a0G{Uz~wxu)sK~OU61=qIo2!d<%g8yL`=VZC?y`m!_QwIxc(|iMK
z?c?R@tO)(BQ$G+S<fw)M?87N;d~95iSYJJ>E7Wc?e;yzmP0kM$YUvfI&5=e)cx&be
z*VwwnmKN7u+qV|TYGReXJ(2#oVLIQ=-e^p}?oL|n*Ni7*n72l2PRSHO4?)D1#2}}e
z0wfS2SP!@9uHxlhq*!Hf0=5gJtNSlFfnEAdf7`9bDIR5Aiivh2o{lWkKp`naEKq(3
z1`%IS-x`8n*_Bl)!u#y2b;yNLBgUb9JlWPM3sts&hW3JZ$0I#ImJgMq)5&g2u_B!g
zovq*fsrCb2IMGt^4nv_z`|1PJ;}V=o>mrTEQQU{W$`}8BgZHPta*{v?@CSldrzovr
zw#EgW>kpH4#3zHeki_CRYy})CzG8F3^x*Aq@bw@3*CLpzVa=N)9}OAnNEXlHX2h5+
z<2giu1yIoeXePiwDKvKx+Ez5dymhM_s7VDNqW`NH;&1hoNQrD>t6Su?u*PUj*LQ8!
z7f))Y-;xHH&^c_1wKS-tpT_i@KeH%QR`E!X+&|Z}2=VdZN8jRU8VcT`mB%8ofc<;V
zzT&y6V@np7-9{i1O`xPmLMFHch=LM8M2Q4pXs(C0lNr9$O9P1sp@sT-MMT3&0w@He
zi~~L<IaI7E2<TwJIRUIh=Am42QLcr({v-F5K=>9Y<xFu@;*5#%BP65@ay-k<_H%sb
zMwY$FPXljW3|f2-y(yIc<B4~=Y3rCi_+7(rX>E*DZ|aC2`+dEhct@VF$d>22W6glc
z>1?Jxs>YX95_$I{?o?4xzTSs#>RU~XcU!z3+;6s+Gy+58EMOA3aHL2!=Wz4$;X|5-
z@raM79VT%a8Le^=Jpo$)Hi4FCNQ&|VkXwxLRnhR`+`I)%(eh{ZJ)>J~YYinE+6Yy3
z0JD1cvRir$-P5$Q9uJjFYIyc{UmZj{QWj&%n!emjy<@18o|wmcVK<|r;BnUwTS(|z
zxYdW)Qo81VT}|-<R3nCa+u3=xO|O5wCBu)G0n5R1@$&M}1oVn)+&+Z42Qx*53zlQu
zYzb)|fNr$JR-z&R+yqnv1dS+@%s5T0y-+SXo#3sa!3zv>asY9)oG4&XY*!O&17$9V
z@$tz%>fl6JSXOX^<akH;7{!nJI6lUJunO#)v0Nt2)0w)XR>iZjyf5@~;tH=EJ2f?%
zjX&3$vip)(FIN*7a4@KN!vSBEqi<FIx@<&WJKUyWZykjojDfK*Yd{21BJQjbd8^#}
zc)0{t5NR?jU;i(-sVrm}9lAs7L{SKMmu=lae;5OALXya=0Kh(D8ks*N(hk>6C_y*0
zCeZNB*Q%J$H@LU+odlAe$w-}XK*F|pC;SJ9FlkqC=g&u!H!17y=4`n{2J6W08s1@4
zcKJS^+4r<@9#y}Y7Q~c$+95a2!Vka}AbdvX1@1#Z>bNRHCAZYRqj*Knu}#RJCDofU
zqUG2ELmlHeDLjgyFo&oQcpP(AR&u1Qkm>2&s-7NHf}g{wV~7f~HhX|e+*-95oKDX@
zgG1vM!aT)>kM}%fwv~tYhv`5;%5|0w;i}$Ecwc~FBP9l)>zbJXqq!f~2P{uw`g}Vq
zAspnZk712j8m09{X1=kdLTg)l7e+SP&s@6%B42Ety#S2V1^Gx$L(<|gc3)Z?YbqDl
zLtysRs=TvW7LXO7s*E0}bC3uTk4MM3(h!UKsm-`e-wW&pV;^^x?fRM+i$rdg9a7$Z
z+YQ92hgxZcGU5tT)>Zi-$cCR}Ty(n6qF(8S_7<0Eg`9|6z1qm)WQVlXv@~_O?{Ofz
z*ooXmUt9|b@|AfZJWA>CiK56dMkmGbH}Ar9!8?<?6Mo)stJyKXH6j+zn^NbkaKxz+
z1oxqYkp2ZYG4jsP2{PA;_*%+cRK;d%D0aJRN<+YI1tO;J3s<PwgLErftv4yV0WHba
zs!(JKJi?>!4E$YL)$NC0y<`Nj&AHT)*P>D*Sf%YUVK6Hy%jkU_CPZ5qbJ0GlA~n?z
zp;p!l-uJ7g=iOuQcl!Lj?5Lt1Lx!mjeD~J|=eLF8B2yVjO5dv-!U3&Xzo^)SeiV4G
zZ|q~R!}oAS$7pS|U@?ssI4y~Ux42D&%#Z<{AbF2XKkj<WVb#pYv-*O^Ooz>UC;O2S
zE6O1&iQN916kR<O9gVH*MY=#V!jR!dk6lCiI)@Az?w1{ZDPOwP0Irn#>TtyLJ4Hq;
z4i|EB=}S*7AVwv%EiLrAa)(!y?_{D^4(QHoZu{s|eXM1^=Gy1AiN=<D_-<TH!|cLH
z$T%z>r+|0gqy>Bl{$V`X2yaa;p(H{JbSTuCG!P9oI1R<-;^~{T_`*u0*vcGY@tU)j
z4Y47F%PP1E^Jun6ugYk1Bmj{kUmM^;L>^j_M>rNJfKF3ukYx{lK>K&DMdf0?<u8f=
z5kWHe5!v>Z<GpT>@PJeVd{i`1mfBf<&&;+*sNgo?ibL%?*po**bgH>wlyhn7od~m2
zR=@jt1vL$L{gv5B*FaYoIAZ&Ca_R?HXtwO46RzB8Ozi9wca6(;uo9mNYD^4NW`Lsd
zWKxg8$LE%0)t}KJOe{G?uN+xKd9r2(2&rJs5fEMKl;|?ZIJO$}*BXKeM^(kPZcFki
z)&O3Z%6^rSR37U-yt;5yDCJ<#C{c(4^WS6VPg4#T{=0tP+(KJgDHrA6DP0C1O;^gB
z)=AV!&@34(x1ycuPttf<CY$P&wq(r)nyPKa{HOzX)Uvy)9X^#e3PG^|qiQ5jOEU>x
zaA<jX9zDeV*SSuq8#ne2W|B6qH=hPePI38?<H2w-n$3Yd0R%dwSH>q4FX;yDYxTTv
z9o|$-#3t;t>R0si$noz0OpT{H9CMpnf<wxJY!mYy-2#TY!c0|I+J>8Y?yUst&I40t
z&U9YWSSCi%!#g3T;<l3$c%3b3Q?V*#u*dJJ5`UCQ=e!BE{4rWMpv}gnj#FRPyFuF&
zMPD2PZ!E&qPk!9rlQh|5!dT@qYpPy21ge3+`3%r$BO0M@W`Jg(B|Qcsl63lFAu23Y
zb+JuQmMC7s$E;M{%TS=z9-}}Vw6Hz~{@fC%G?OIIDU%@VQ?0>M>BBGkXE-_kD`O32
zPIr^V8i)>s2|wRbn?~!{BbK+4X?T?`zPooHc9TjRd&YJsc~))SR!uHdev~6wJ$V<U
z#G26Y1JWQhihs|V<M`DAq+tGVP-Z*?fhf`_;|Y~E5e>PQXclj!t=tt&M1{kDrkR?W
zY6T7HbWPOgNN6axCOPT!%g-n(k@cTRo}zf<YPbK!rFkx60>>|BF^4UbE0Co`GrT$t
z?HMhIL>>X}Ks0U`Iq89i)~d0-z-xi9s7rxPG`LALCh^6inhG0SqW1-bS%HvRkw5Do
zM65M<5*v>n)Rxy8lC66h_UhFFlh}ux*ROJ3nG~dX2rlz5;3jmIBOcU?imwBiu>veQ
z5B(K&ocNtuqm_7tY6c&6%3Y5I9I~62E>HSHr0f>_MnUZ_KhvhoPv%7PlaCehGsZz!
zQ=1O&r_)ha)(_+Dj7|<55BO&bGZDtmwrRK)mF}r7DmC#(esAZ5KhFMyhertgEckw@
zE!wmuV#fB+cFcZ(kU(5Ry{MCyDUUwJAbkW3u9<up-QUMT=A%>)xjFJ{#R=L`Q~^#s
z`Z@|quTu_W?P2FP)vCkxgynet{B}@uoAUg{t%0CfuKN|)(_veJHgV$&u@=k4u<R$x
zbqeyc+i&7baoLK@4d7sf^dCHu_HN`P@i6>o<Vf4IQ(3<)k*6?lM+wYf%gOVp-!<D)
zf&7hI+5XDR-cBRF3l1eZ6a+a9b~M)(L99gj0?|h<V$g&I&9m5+-EZ4z8C6f`-Tt2L
z9;^1k+4iF-p%MfS?i&&?3l%7;Fuq(N!7IKHi;`GT%k`EyI9~8H890nj7Hl1jG9FW}
zo*&D5<+LzbbK2io6RI%F7fernXB4Bg)D3zx6W;DoVeaIKu*=L!gS6?lJ&-<Jm!7B1
zu{La<PDh5C;u>*4f<`nhzG{|~pRCAd(nm<+{CkmBLx)*ZP?0l0us^7&z>HOXu)0<2
z>1cxr&LkdDN>|6D+G`LSH858T@8Yx23S^4R3=)fikVr%P{+9kjY}Bp%y(^>OdCvkE
z&TQ0e@?uoh&xena0>zxURH_K60{{beb+LFw$-z7$Yn%R?C&V}0cg?dxm?IAocJJCB
z1nw!Sc4&XbtVg~!>lm#suxl|kz~y*fUB);G21wf>lc*uprvsKfz<y*L68e)wl9kSB
z@(;zvc!N8-dOK=DMeYhpxlv%~8Y^XgJrH0TUa?SF=(P~OI9P(fFOIxN0499u9&OhL
zTGpZY%>|X^HxYb@Wa}{#6XRsclCP?u=uB@25XEN|7kARy&W4}z_kqLBIU+UxzO=Ts
zcVYBeJMBUj@KPsp!TWXiPkmgnD}b`6j3yjOj40=C@t#?GcXSg{<)e?7Loq?tid-8=
z$%Tqso5-GSjF5--3F3vU{#fnK{751=bZ>is3~B7|s>=dc0H+sPy^seM*7BVc_8u4;
zJFejCy4BhN_xXv4IPPOmNW_;12M(@(Bvc+9x~DGZOqNvU$>b40*!=Q}{KBDt)_Gwq
zq<v)ZslcOPksi~_hfcpe2hj>SHX^V^7&^jaCCyLLa&&8H`cr>_(}Pjmcg0jcnZ$68
zE;J+B=hk-T*bkxfSFXe(Jw$x9ds3T1eik+Q`&YZzjCu7T!u|z5ZM@JY^>(1KzJr#g
z;IKn$Q%09d2r7Ib{EZjv@gNf;9)@YmRlr5Oi+SsEY`nGxl)ly)N%EU>0_#$(9@L57
zET)<yy9^lW{bi#x;EZSP&Mp>tSUME+obQZhRwQg_K4p0C{*yf+ch6JUE*T58fy2sT
zu`^#f8Eu3mnS7Tf5)&P@K@>*;Q0Fuj72%D%+YFG8UfCn-QF)Kb-Y>1ek-#&RLlqgP
znou@`wHVrJyUS89RJbukMmIFB(8c`2d&K9_v#^l@m2smb2iKdyA{<ry#fwL~rN;FY
zkiioFn;w0blpmK$74-UggtZp0-XE9h^(;@A6<j?wThQ6xowFS3pePIER72Ky;3*fm
zE80nIF%D+nAKu{{Bhd)Z{-Asa+?#PG8Rw2^tD2tC5X-|%)&&ZFDUbJ7pe2%ZX*j*f
zm3vsvykcmoAAl6uQV~Dyq~wan@KIV!Ms26Bpih3MDGt@nTm^5rV}*DdR!r_jY9wzw
zVO^b6RF4asc~fbYmpfo?@7{U)Xqax5e!$mJG~%(~6fY>WS#&dKhdo55=vB)R@>P@E
zN5yBsr?tB^p-ef6DA$E$YgaZ#sAz-B0-`ClD0CoL8OW|glAFaZZpRjT^J0@DA5p3u
zM}w~qYHV$UK3m6!tO)9nBDU=luzF^C!S?XVghn(2v5()(8<d~fhC?BU^`nx`G*?sZ
zKjo54Rg%p#TD|vCCCfo_vbLDqgC}1)BODllqEfLgjZ2||Uf%BBKEoyZuOqgS3euXt
zg^&jOY7lw#p*k%M3j6Yv1QQiUlUJzab?dMG>rEUqWY_n`EHDn(??w366pBOVLTIu&
z$=1eXoUzpg&fEM`#}FPyNI>KDVcN|J_pn}P|0<XHpU}zOl-~QKAEQ{>vzZRbP9bV{
z-Q@iQH+&<Nyg;1?cI9sgl%ntueVAUlkZ-(Xf}cX{aBTYYXwJ>7^tcu^>WgCUF6jY$
zkU&@JLgDA`o{0mnqFs&45z_)<T#DYPqKL6=QgQ*ZhfuiJi<d0{Bg$k&0O{Fh_1Sl)
zEG#exXXFb8r%DTbjGF*%-a`^=wkM#d%6P&M&_b6KI2XU8Ri2*XsI;<Dh@5XSG#l}?
z{jHCchOV`!5)QIj7LDbY(NihsHSUb5ysE^T4#`mZD!&XXi*ZC$tgcT;B9}?mXR0K&
z&Dc2{s=3YvI2nM_cQYV}@q*%0A|UG9lh?Ri&JtDE<B_s<_z>|TwZu$4f5wvHBC!y*
z|3FU_?m<l+>;gTG3>dHYVsHByqZ<o*7Od31Y6O*V^p~Eh>{1ya8r%J8^lUWaPFh88
zEc!HOEosjJNZvo>E0j2n3YcZ$vhp1WVN@$w2PD~Z&M``Ks?<_Ym3SpGqYOiyw8u7q
z7ED7#pM6(x_4DQ!+-TU+vxvMDhsf$;RXp?1(xs5fa};N+5c7@E);Vn&zOQc~l_S%&
zowyo4S1Q6w&{>crwe2SfROU8OR$#XEZBKU=wJf#L_nGd?u$d7*7_j60)=cDh%~$0!
zZw8d$=9oaQb2+LkaAl%w)aWvRENIigL;zeE3irxIiiP<wg@Dx5Y{PM6P`}(>fC9R>
zP#Cq|g9A@jB0caRGu_U>&Oq(OH$v+<a9wI^8C}1~s}tC|A*zIVFt`U0Q*Qq5Ab+fE
zu7IYWiuoWFS-SfC^xiBT^uHUknt8V<Eq(n13~nBs@8>iET!>~A#x;en9Dn$@|LiE`
zUva9VpbE=;`CVDQ2cxxz_Q#Wl;0`2(duN)FvBR+-U?af+7|5|N#a?#$h>>2GmwRQ`
z3pi2+4K&$*qfPHP4{~jgdhw}x!Ck8V6Fk!c>B*BiO$Cktv$fl8fcn5EhniO9668N(
zkYQZ*g~r8IuOyVSNIoVh!Hbl4cT7`ATw*4$t0np&u7mo%6MvS^Ja#MN&KKn0Y#w6M
zdN8FU<aLre9TNSG7QsYrA0i>Q!g6ixUt4yKZo0dhxxlMIsV>_fRX+;|91A2u8JN=o
zPn|tQdJV3FuA0SxH#Dq@aw@=eP_M)s-}y;G8n`KtQIrD?V$?b@fVBAB6$mT)`&Q2v
ze(?P&qIXH~G$%T@bL&YzTK-Ai#jb1-8SIA*)reA)BHo7Q@jC`?KP!DZ-Fq^fsvhPZ
zaCbS4yjn1g5#;x~&!%k3)}T{azS(j~)saO0D2Lk!CuGZc{tx!vG%D$J|Nq|Sw0E5b
zYnM}I>ds1Y4i(L@U5aLcW{NYqolsFQ2OMxZjSh%r&friEs34dFD&nxqDF-w~L{YSy
zL2)QG)9l=w|Nma=cR#<^y4U^S|BMAJhU>cca(&*{`}O|p=1lh1UcG>o;L*j;gJ(#1
z#h7L1@R0qlr6{=&@F-x%#7Tayx*+H*Z;cbG#3pTwoQH*i`l*2#!S>=NjHo45qnESA
zxyU}V-E(dfAJ`K6AQs%yU4u%$`j&!xa8!P(KI0U$0YEN5s}<+r9+n^*kQL=QEZxl2
z+NT~D5gt1GMUBQcM|aib6dulA!tsp6AC6abhM%~wRCBqFPtOdI5-CmQqbacSQdJ~s
zE5G<wO8%f5!K1E8SI*j_zb;CtmjEYl)SXGGsF_siG>v7>0f;KP59B@K)#Q?T-Mxms
zwLhCJYj2%>X204*`t0bkHgyb8ANM3d%e1&=`-jb;8rojb9+Km`eIK6kK|R~?vhKZC
zgXufYW@MYyeg1*okGosS?<rIBP0jQ2LlZmgl_6I2oC>}(W~t^!6?^wc>rdLnZ`#|)
z1(sf|ps99-Kw~k<PAWt=ov17=w3XdKPPga^+p(pB=)B!a^(xJonNd<?y`Q+0OoEfU
z8#Sny1Q3)K=l^8e4TTQe)eq^6SLVW>E#diSXyyuZvU~_3J2|X>nJipUyWX$5-CLNU
zD*S!T4Xlqmukj^*M<rN$Ugo-C`L$hzzt8=)n|M!4qVe3R9XTJ!N3=<zf@^<WwP5cH
z^^<cYXtaSw*v5$Bmb8_~6OV!#f=W}8>9GI$z<}C%TfB=5qdx~{WTH7=*1b4bP$WRm
z4B@V*DNO)%z&t#%+g@~VMy>uiB>Djb%nrA#t(tD3>zhG0c9QJhp4@XV`0YU37!Nn_
zaTs*?J=G`E9ShR*ds3X8C75)*sP~e}ame1omF%P($uqu<j&vE%iANp($u)WkdC!h0
zQfk+?=$;5(=0rJtD2#Pd5y?ywx5+ejw0lY|Ib*$hpld2Yn=i|O(cXCRFQv;JwtP1@
zpQt2KM@kD3hHFG`Hrk14$2AHaA%uLZ?hd)v<XKj?`{Yo50Da+1T5r}l)iW-{Aoa1M
z8MPG3<mJMDzxnf4UDkspUU2c+M70<}lq}Vtf!PV#iQ}|i|6$(#d6>u9!O^Q+r}4!L
zMP{zFg4na6J3~cAmzIs~9{*Qif`w5mcn(n3=y_BN$s3;U;kFRt9gLJJ3erEPtjBa|
zx>-7`;TTBgWTJKlf|i92EEd?Aq-3yCkbad}!;o9>e?ypMr#-z7IMNu8*C%A~mzv%f
zMdezRHrD1yRbK1frm=Mt<T4ErN*#NSLrt^j@*eBoeczzOsW)?$JFPbVJlb4sf7^)r
z&oA_-CiYB=VaMj*Q+sbCjA0(Vu$p!=5NDSh6QwrN*8mR{PECJX=RhPO{fva6ZV14=
zhw}D^+t|0|+N`oqeAgFZ>qx!53g|jHpFK_D71kKLG=p8Jq}Jt{0o9azKEIUeUo69k
z`(e8i<%3B~_WyqNdcMFRm;q2XNo|AN6fFN4$Jfwo+j_NU@yMGlORjNGpd?BC=NH$*
zpjEUlO-e&loO;iQiqqytLOB{;7t)Z^9J%t{GBb}lHPV@pnQ<_30(@YCk2rLq_H4Qo
zfo6k<H&#olR;b?6$B*Ws-A>qajfcMtYoIJIADzNR=Wt6~)pn&rcG>zQVnHQwoYigX
zFw_^A?%e4e=hJ8zFJlvn^=S>SG7Q)|amPMU3FRs@sfo1^276G=oskO}HZhg?_7-c{
zSGnAW66Vh@(O-6$sKlR&$|&tfq5*l#wn&i^f&mOZQkLp{jw3lUuu;b@%=$%zDL<%#
zr&2he55(kV9S5km&iS*If9;nMH-EI7iJt%u|4p{-qh-Nj(34a`HjklKiZ_9XG$wGL
z4!b)P@P2JBLkc22nUhd3>FaD7T3tO(FL1ww^HzpAw4%Lhd^U|2zxGRe<gsqma=~HA
zD3ylP?Y4W@X=AJLl1CTxohslSd)?Er<Pafd&&Kp=7Qv(4$F9=`X?VnQWv7OAI!(1h
zfzCyW>%3ha01!E6d7IWZHGd+Ro0v4Q!zkb;z~c}(%XSEWT7x;a&@hxl4ccYfiAx?S
z=2uphtdE%XCm5ZcXO|vs`Z$#P<SU_x*DE%?a9|q`(zJc6alg#{Nm$GR%GUZf&sErb
z4)4WiYq-y@oGE1^Nx8zQRDNy8o>5n`5kJ-3Y`#g3O<$H+y>*b2{Js*&e#6R?@vKpw
z!dEJxr^Wr*=!gSsdmP%1X{xZFQMv!rhT?d<i17G;_{e+RkK^45+gTkXB-D<xo#YhE
z$z*amhQBc2krY4UwPWo~KVJO6*D$I?;Ddi|8yhO+cx@MC54=}O_@`{$a@jS^v09_+
z_G1gGsd=*ZFS*<bv@1^vT5VW`F^*Itisr~d+~=mf9?F|4J|+@IiIsKe4A)#H<8rbP
z%f|dYEqjnubvKES85T%_lQ1#qeqT35T~K%C&VA9Hi4*D@nNsc#gVxYq9Pi}ji{CV+
zsrVzb!w}OxOvx}NT`|Vj1RH8G%KY-_J251~TP7FZ*QMPAuv`lpdmg0Ovm#J*lUSuL
zq!!6l@y#3GB0spw>;GPH0dqXd$ZRZQG%D;seIuSm`=+`4@tdB_(nr+V<dmUg+MyJn
z#&tY|&+<Mpd`{L^+N`OAejtSEMnKQ&u&1Vi<qjWrZs|0<eH6BHa5IeGO}BtW29a6?
z4WL!(e^n+{Ph2Qq1hBF3{%~=@<qboDea0WGEvGe(w90$+=|Oles?I<c{pd<>$;-u-
zFNirG#mwX0%eotS>$JAP`)R>Uual2>X4N)UJ6<~yvmM#73a)cTL9s|A?RhDxjZz(Z
z+P#NeyW}y@Go}mks3Ai?`OMO!RB05~1_+QXt++&J0I+n2c4EKwP_(=dldU$9IaU6=
z0hUmp%j=<MfLf6={R6#Nv`B9HuNQ~<UoQ^6dOl_ahnv#BAa`Y!`qCL8JZ_M~A6NY`
zyYesSS%+z<NShBt((A^s{tJLFgkMYVfxZUojuoq%3{x8jJ$aW4^qR9uw8SNj61vrI
zpVOD`EZYG8Xs;t~qyoh*Odtyl0W@`?(?2JG4*pzPEN8mbiUC7|OsN43luCh!qF9=o
z9yu2uz(E(P8yCvzHEe!@`issIXYlJTeeI&*he9nub~#|ngcxFsu**Iwr_b!JUI^tv
zg!eZqp#B5@-rlM(`wyX_18Jlyx5GS#CI+ZoGJC5${|%5esArXfXR)i@Gjij0!j2S-
zzK-4Q3X;E$6n<}P+AB(QWk)!(XoV2(J*0jkzJ?XK!+DO2?B+7o_=HeNbe}l;Jl8RO
zcRRys%5VuI56`Sxo-319Pz_md?-gAUGbNe2+mWq7f=Xgc@vY6N)@if3y)4v<Stu{G
zyz0Z2;g8gjYc)Z#{a1;-T*d7{dUkr-50b7m)h%r1MfI8MmE(=N^AGt~-aoW<ZU5uP
zg^76iINJ$hO%4sG<2k+)ry;AiS6S{l(C`$4^9T2huTM8yAQ>^|p@bOSX*$sEyuJ!4
zK=kHLy-c3-^vIoF&Nd+I=DIJhp9V?8Q`nIOy`8XBVl@rw5DImG*mZvghtK6cK{aPk
zA0wSS!+F2bM{y3!^|!Wnk4Ce78n%-q_KH()wuS5Xx6H%9DR53GAQR|$wbiG2;(qdB
z%}yDmf*kbY_|awgE79LRTd=qN&9+a&S6w|s=WZ8UJ;+HE-bl|8hz`Wj68X|Mw^2Uq
zwJ*A{Q&R1)@E*d%e1}@<n~bO^T&wBF8N(O6fl-=q8_pD90=F=kaSfh1@Hnd%Z*o4Y
zg=%=CpP1TOdK-~SG<x((tYOpNT3_Q0K9&l6u=z-Ebcgb`9*6VjcArL+|Jl}9Zm;9#
z092xhno3c^9)x;^o($akSf+fTF_#d{{tU1DRFW#ocrWyE_<I7aQC5)hJg=Ic`$gl^
z5+|?XlOUUevw%u?5>qu+>=3ms5FjK13Xsz|Q$YLn8;<pw&;1F+PEG*ToHVEF5=6F~
z7FxoFaZfl3Q3^*i04>2I0KiCsT~v#fE})s$xn@(g+Ee@jF>8O@Cd|rtrshTQ$b}TA
zN9&GLkCyM?{7p@ss;d_(-oIWy-5qF(7aw}D9c^_0XuG!^-i0L0JLb$=^p8adm&%+c
z-tGv^PJ7S^N${hn8^mGAQ1JbAXk-CRM+ebby?Gh@Z9xL1CAR~iQxO<>izL#b2TPmW
zLr-AN+(){Tl2h<isT-hB@YiY56%9jW1`IIK{s+qh_IQPAaQfj^T8LhXS<rDXSvC5*
zoO_T3*}uQmYhP<HIpx~mMt98QTIeZz>xW`Wm=p8xUgp-|fAO!`|KeW=S>^B`BOjdy
z?}@u@#zRH2hI?Y`g$OMcL=fyjhX;wav?BZqj5Y^@b<RE~yK*9sTCR17C9EM)G`(bf
zcnSlNMsVN}u~{N?D5A$5syoyNm^y1nS<cYguE4yhzT~CzCouI3ywIQi;)JT8l_Tce
zF>9mna&lpWd&e7%)sad1OBhzsgK&db1?Ig>nl>wMtQ<uq9jN|*%rO+;A%zO!drD`e
zcDRJ7NUzU+pEgd_zuk_K&q%8AnO|i4Um!z$eWr#=*#Zg_N~eSD&6(noHj?b4KO!`)
zLfN{6*0U7MMnM&v&(WN2Pm<c0c`tIO!v;g4AE-B|WP2#U9b^aY16&{&pAIub7JZt8
zhsyg3dUe$8<pvfehN#8st;=DTk|Tw8veyHs7Ln!_%SuKYmZ8BK6XV2OB2y{Be}nSR
zpW`!!89_^)w4AJ_U7^pPANuPgSw4zCGpc+8Iki*7MbHNtW*O+D(eoj7KDAU;q%--O
z+}6gK9XnyllXNRH5|`D+qK6<6c3E)jxCme~79F;RiEyNu?(OcVJEoxJ{K%CIU;UOt
zM#Mw`YIJ6bWuSshlRGmIRxir8y^?JLPkP82_-h8KS)qCTl%lJZxvxh#U!l~p$Hf2M
z*e=#obw3e4TwVE>z>(tq_e^*gTjb?;ad4vscH!ZzHy9G4nPUOXw{_0^mc^HT?c?j&
zM-Ew&gl1rbcoK0>T9Qc;Gy^U54d&s+KE*#Cm7Kh^T?ydNwHR4hn*q~ZE=RDZ#7SnM
z&>K_NRj>R}sBP<*1&lDfzYMn$%0HZXPu(aB67gTyyzqec*evl;Nlt<~DEu<|f8k$!
z#^KYs@1c<@4f-v`Dl~med>cs4J(j*u|3_2p^1t<(={VnBT@RZ=L{lygPW|v0JGY2i
z$ctWQ5Ebv`Gfkf6AJt&*f0qhKuVROk)(1OPs<ez1J<0q1OS{<i$PQwk{;4GC(q>AO
zK5B?}V9IF-(((9v^gqA2dkH4F$RX@44)BbdXkUMkp8H794q2a0ykD0v0R+xAgj_9c
zRp~Oa?USRoqT5VP=?ZNj{Uao@b6$M{E=r!U4T7@ZOO7b!Alf|@PJm)s`J@iKD&8Ok
zoRokxKGgm8&aqQRKCHoIj$mubG}(gI4*hNPHE?CBvf~u(d;W?4dD1VN%*qmY@xCf&
z`7gQejbGPUbHBO%7&_LeR~(Oz{Y3UG_;=JYYox(8PBQ50&9gtaF?)fVPbdlSfBc`9
z{yziqe^$u<4}UTq(aW~hIbE3e%N5qC-4$Kvp1#bBqgTYV4>CW#Vp(e|%>}})iGsPW
zS}ADT{T{pZKa{$($Mqu@wI9x%+}FPyi0(etOVE=GHi*>IUih>2vCQ{|N=SUw+s|gH
zV{<v0?v)eM=NBxzXl;j^8e-aGQVZDo7vOh#RxZ@)U4MO;z|=NTn6$`YZxB>^{YQwc
z6F-Iz{LVh11^s57KL5!1)^@QIX%`1Jt1(HGpZ>KEe&yLv&qi2kK^^6nKSjNL^mKY=
z{oc9sQMpKmYt;y2%yH}@)xIjbRQL5<<M^T4Ppsdritg-nc$SXq1oi$O`6vZ><S}_>
zgL<M(3ogLIpaR|8aK*9sQ^ZLZ;0x*WdoN4>T3P1WzH;_6h)_K~5t!)}m9q35nyss1
z<s<fL%<=0To|JOVg3Pi;I(*;pj%}{j)N@H2yHvl6yU*MZmoqMR{JW<@V@ab%T0Y4S
z8sw&`EBHPtA8FiD+jyv{L94Kk{-1&Tf8IK0`D&`&D&1EaFdaB0shisrth=o~6}tMn
zLN+Vpf&`aIqm)++Kl#{jrvzkWYH{c5+vyd;mAMY;+H!IrXoIWw9A;I0&`xZP3H7qK
zuldmU7>b8SwD6vBdxqFkq2gXr*Rw{t9IEPlva|I<`r=FN5$DRB|8<vlLIL+&xQV_b
zT_1N_d#7u|-ZknU<*lcElKFDa(L}X!@m(Xk(Dmcci7LTKO$!rGADpTHOur~^T9c#$
zO*<|KSqr_#8jbiHvU>Kn{p{~G5aD`r{NZbEQwee{NXA0)4Yw@tsZYUiA188%pA_w_
z_nnvHjOIY#{o;yn$7<{Q?0-_lk7rkw4%imtN9NSS9)0)*Y=96^W;aO{rGvo#{DQ>D
zQVdL#bW@lMjo*(~@rwZjclO8TK)r$fR>Yw!{d~}QFCWt|FWmt@@T>0{OD{(LXI!_|
z)&Fq#yC_KHNcMyI8t6$mI$m&j{?CqIm7J7(`K)V{s_iS?>8y<YSBJRzE)5FD@!6I8
zf6Nc2Xc1&S@=gqW(YQJ(mBmISD)z28+fT0*M@5-Fy_-65iX)fsZvlXPRfBC9F<_0;
zQUpyLvjnP3w0U<aRTf>Cd4FvAdtI^S7wRcy>QSZR`|CGG<e!&LUCkC;zmfJTWKs}u
z_%roixErgp*DbP4@wuX<Ss*W;5d2kgB-}U_vX2XZ8<pRzC7Mw*-mRJy0CRuKuzKY?
z66tRU%^V)kQLTw@HZamVlp1W1*|r)}?1suz5}V|PePr~E@IKK$eUd%Kmo64NSvkG(
zk<s`~v@(s8<lCO~6ajg_a%qp}qLm3#60OW_`AFTC4BC|>DOQSI(*HAt^&cUA1(7O`
zee?^c=q#Uo54kkQ`>p+9ZS~6xx!yWKv<88N4ADX^c9E+oF9saet658ySh>b(RO_?M
z{C_>;p19mj2$@ch8U7GipmT*C**Ibr4)YzP?dgUX>@)?}SU!m8`iJ`Mf5Q8T^=C#}
zuL-WQmX?1zULtL+TX!U8J@^JDnG*p1^uMR_tb_5#PI=_uqh}LxjKJCFf$}d9CXR3U
zR42SAgq?4&!#Fj4>P)H0WoO!{uCHk}+E+|umJAQzmbu?M%#m$egV$Y=#v^kQLjQSQ
zW4yzRe9NaxUX?BuCtGH+EY?&SiF!^_Phh|4iM-25`C*Lqj}p~7XjhZWQnpi`eC+S(
z68Ekpi=Z|6IAwRy2l_3g8vlmK)q1jsDI(K1t`H5wlQ{IC>=lD$5Hxi@|0vudThuRC
zI9>()h__EQ(?Eauy>c$JYQ3Vb2ub_ZZT%R|W_-T@<@N}3u5q89;F%;`In!J3h6`a(
z!!P!P6V$U5>&j4_|Lmn)3{lOoJSjPBx@`5zthcz}4<`OV2Fe3<zs$e+Kdb!zt6NeD
zup`xT`<9*T;1dRg55E3blb*4-4NF^|iZt?A&h`nLc?x>vK5ph-sU0_3d5x+OH@U8A
zhN~Y8ivEbw5Oo2_4n$tN)6}N=C5@muW@1^8?dwFp*1}D<D;wx&e%tc0JJC!x9et+L
zOVNy?GogV(-+O4sa?Km%-xzvt_wWCAUsHrIb}K(c6^7`bx6s&Y6KQj&>qb^r0aFp<
z(L%8_)xFSDO{!)6IK^0D$6AA|t@{@_t1Gk;(D6X|6pX>oOtH_a{Lcx5&o6MFhq*KP
zhLM#uhzJC}m_qVenyGbW*6J;hYP_dOL-ugea;j>|GRa$sx~|n8caarE>`e6NpZNNg
zI#-AD4qf;0;WdMOTt$Sm<P1r*fONGWC0#MDLf7!ln}hh$R;Nl=P^yDvCN7XDlm6Gv
zL(QvZB}hfG3<+z0@=%48o?}(`_56dyJlsisYsHOM=DwgZw{J0sS;Y|KR3}gnT0PO6
z{+ZpZM7EKuwa3~d)DdEhlnR4jPzXbp*ACU=<4O6Yib{wQu3!9BaNlS|yV1SwwL+NA
zIoT7JP*6@-@PaFmNvBe;lIX-_d}Z>opUlDYzcZ4Wh$~aQ<x96bN)56QQl76?WMp#=
z3SwdZh@$(APp_fh)uzAl{uStK`!e*+Tk~-^rpZn7$-UmsLeDW1zDYx5$29knJe{Ty
zz_7nFx;L8<bTuHbKw8MhNEKyNd~$2g4D)7utLpA)fDE#Kvbpml!KVEsf&9sJ6WiPR
z4+p!)RR_kkzl7H#${zNY;_J0Hof38(M#Xz1n1eZByYoGO?$`_DWIk2$*?fMl4Csz~
zD%euKqE{tT^P0;mmi-;KOQd$~iq>}wrMuqdzQr&3gp8K^X#X)g+%=IbDD0!G870|M
zc}!~$Hd@aA0hQev%&j$Y7;G=fgE1Tum}IoCCtcUAaxZ}BoiF@-AUeP4Z5`n4I>^pP
zTz*(1pHxk%qIvW5C_PYH3~9C0fD}9Q)pW`n&G|+Ca^5#7PSZt4qG~+5@Gx-D-{Fgt
zj6tq;QNcB?-)|oS8)B9SNM(VW?KS1M*y3-2B0o82tk(MPcdSsL@QsaTNfP!c4ic+l
zO>GSoMRaGtRjK}X0XCAKN-Ey9)iIgGe|wyu;MpE|G-P6dnq`Cu`$cgQia~S-jzhO)
zKnkgP{_5atzQHXIHPyO<q$))E<>a%RTUNC%@?r~C1!?ux<$YF~==-KN((DTzRSb(K
z+`)E#?OC)q*w?kcOfv=V9f%!=VaJD9CdA^F6#kcd;o>08jYx*N7GsoVri+gb;Kc=-
zn+x3J%TfQneKyYXkJ5g`+xGhqqDiPrtZTJl#mf)r1-m-VC1>W+KchqHA6{)j)pu5*
z<*upKS?EtY9x0jTT$;$KUY+9Zjn<xaWPQ~g8bGe|G<gGc;oDU2MSIW(5<GR8$(_gl
z;1{HpwLR*N{&_q)5u53wG$CIm9lGh-i9n2hQB9`9m-KHG5dB4J2GJlz2;jzzF$sZ@
zVpSL6u*ZweKYm;I-ULWYo-IA&{1Hwnxkk!^!1Kdhe2bG*9uPb7zu>jQjEG5Y+7JH~
zFg(Md$<?+*XpEFQp=K$~z#=UVmZk>{a;Uc}ch#`>&2!i^@Mq6j#Nq;%v_$HJF9CF+
zT%r@RyP?W!>634deFg!Sc;Sb=DEeorp9aIW11&KKys9YIXH$IE=)Fec^kb)UMK{LX
zPdJ2L)_E{JursaO)3EM0sByo&ff{n9Yx9pq-NCn%S6LZeC-s1%IjZ7|!2^F4dcD`Z
z^`@23xr4*lgy1S$^&CHO^eTj>qlCPmpf$(Py>tRDGs=pyGgG8y*X*M-b1)Rk{=76x
zWFOo{vBMz)v8*$m#C?@IU5e!~hpz5%l|fc`X!tgw*6srxWN<M`xadr3Rk}q@W}pg4
zDMAMkM3`>{r>w&9K^_JEu2zK$qr9$NaT!|o7Yo6K9uc*!c~$e;_`Al`eygX~rd(61
z!7{hT=qk#Sy5GDvAyHkFg{J1&Nk)dxBcX}UY1nA3--2BpF6aSk(?0PMW4}3(+=(D$
zce4?lQ&c0{oBf7Oc`1~@0<KLwkdsuk0PF52^7@-$j#~D*?x`qy5k?6xA*FAxY*@ZX
zeFUEL_&0(oUW1QG@=sfCO?>2(zt)UNLPt5IEJ>N)b>J4H+-fKg_!%8wy94+;(9c$4
zwcFA=PcZkQE9&(Xk{ArvX&Kp&h))R=_WEZb#OQ?TvKEz=NF0V$@8+uQ=4y<<<n~iC
zRD5{S?|7N75&5pYMc5LH$dXvjf>YFk&PblVp=udk$UX%2LF~D)`Xio^oN~9UG2=mV
zn*L?wE%yzeuk~J)oXu@i6G98|<%Zwc`O_+SCqaI8Vm8%fSJ6+<&*RNVS85WZ_j6j~
z6y{=MyW?{b#ihfAgsHq8^7s8n+8OI8ot+t;vSNtW>`&5;jeG(~w<o+uT8WT3oA2qm
zm-LvfTgVi5C>1%R1W3C~5SivecW~(pQcNc_Ca{fR4dx^Aq2lrn$8fEWkdqq|yNOa9
zF_mr|=kK%f?611Ri!MtYmNwbG`9i9ezo@(OzZ|SANaTX*bW~<xjM8}S_giBYJ<$z(
z9{=Qm$NPyk{W(nU^dR@(h^f6?=3_g3@e&G#4!qQKmiAoGpf2h&Ea4xxPn$`Nb#I0^
z3XbqgaOB*{NEBxRfNoco)tgPY?1=@=j<hGtSlF1%xV1ZykMMLFX3^3V8VV4(dFk=x
ztLf_zr7R0{>?Mqicxa$i2(Sj>wV>Ch{UkOLxXh-7TV=CUUU4CNui~YruObs4`i~sj
zH(5BoyQLH76`l0?r_EWVMb;MW@|mE6^b&vjp0s#-sWxy6`NDsG*+sLVctNUa3AePC
zTiK?`8*JKXpQ_-E(y&pePqPg8ZiI6?3|kcaO>v2i%}y8>k&8pIK8+%q_{`Nk&qguZ
zO-7k>x|=Lu#|VktrCS90XHUeiBNmnYuMRWJRrMn-tu5?o8$?Vke*WV>zc@G<GZ!q0
zeHrqm!6RSJb;I03mXTBJx4`NKBC<_ROPS9_8c6s+cWPpGi>df9>hAOxD!jOM{r%Lm
zh)kAyQ~k4fH1w+m{78O73;JViN&-V>na^Z6dFp=EK^3c}Z};`Qm8DFX&V4JWl*B+O
z`9pl?B2yo?!mbdB#_$XEYFs>4a1R?fP+fixBXg%;y?+v6TEWSYTavahf&=`Y$QwmJ
zUEO6gL^Z#d!lR_xOJ2i~HP}{SK?d1}UE%M~$ESl@L+hiPA^d@{n<U~+0$bNUELj)^
z+Y@kI^}etuftV%ow@~$Ex8YKP^jaTPTf?Q(y8tizhWDO@Oyc8+aOKIaSny|`MzLIf
znX_*KUR=)tefSb*x30&ILAiP5-GN@++to)Y8p}FpB##}o;?HL{4N{b>CaeW=p2m+C
zN{G_9giNClD{iq80wWcfQM#7<?W|b4TFB=>>D6({C;z}kPS{V~-5J^r%uK0(=Csa4
z+j$oKF3_udG41BJ;)}Q>WeTe)ta0gTceL=E5oK*SQ9ScgHEku>YwDvWc5})3d-WE=
zz0*y+bsy(+6{>Sjd26BS9voy*Xt7~&MMcV<%yvGg7$70fTK_d))qlni{!DB4?!lXY
zeR00bd`MYw9d)hv;!2>i>x+3-aG+rkrYRFgo~}J7tE!`6kC<@SnWAcSL-kcQb$^az
z)$tf=cV}BB^YOw$?8Kv5PNsKZ93h`29n>rk;7QH1Rs1h0Lh(5!LMZ;Cc)SVZSLhbt
zp_X_7{^m{@^}~o(KXF-1Q_n<k7!vp=Vv0sQ{Y=={(cT-QvO`eu^sxf)_P6;i^P)Y)
z3;F4H+qdmT!|V2lHWhhgtn$@pyouu`V&d)kGUTAHlCZ<JIBV=i?3h?q#V69V@&cpd
zoAw|>Ypr5wWS~I8xjAbmjX?)MZ|%ZJ9I*S(?rt7^sILSdzI0AM>c_3U<_>XI7bnj3
ziwi%g^nFkKiN%Lo^C6>^Z)$HFZ-1?5j9FmqDBH~xjgHU_wJ}uK*vjpazisbpNwv!+
z4zz4B9*k}!4&EO3tiA0PkdN=9h;m61-4kjKZjVJD+_qffnN943ueAzy6KM}?qk=UU
z?f?13>1Ss{R_w0-*w2)mIC4JkHsnCRc5OFtdSx{i6Z3OKGW4iA=Ft4}S`#GG5x#p#
z8tZPJRAs-q9(Eh_ZLHyhld^n<-=&sM3<Q_`TO%z<s}@;o6>l+?$K_fMF7%Z*rzT{6
zCxqPY{&95G=qkcTWXm?@8|rChK*)xa8SAM`Od-nq2+Dcx9YE^X5nUmUs>6^PxYxl1
zm!=3=1F3*<O=o(RjEufEoV3lSgXlz5qPL<ZXUMVy6_-mC?*AoJj+N0<&Rxq@$sBBX
zzgn@-i0BTho_yx($b1VW@bjdq|Alf}QGo}<zX!hQfK&=jH?(ajk1h0uSehunKDGhI
z8;#{|haw_@NE@8foi6r<Q$YsyVkC(crG%-O#!7TUT4N!eG?O`{^{utw5cfiGJEj-9
zHtDF=S9j}sUT8HIIzz|-IZV&17Sn%@e@kBfa0s64;5E5C^)-wsz1WlH9<DqvJ0!Hr
zoo>&wfz70*;`Jrv(PRWh*XC2j6VUk%h<Nu}H%vcQL*eq#AI6qDDKmJ-H4(`G9%k<#
z&(6-iqgMnvAAnLLrd_^sb0pmBaEH>GM{0KcBozD59^_!x4f=h(i8wN<=W*7e=#6V<
zkgT$1&2;0UuX{f!gbin=<WpBLfxD3!ePGMW-6OOY-(225g&z)%r!pNT+N@Rca>wT-
z^NsZZ;NC%TimEChXu2vOtcmik^!4{enJ%&o60mTJ%8Tsy<FYa*-kk$)UT&A)djbEj
zHSO^6X5$%QqE48vD?4&SijwJAUG$U~I3=yp-~Y3opo1&_%!1@E;}OKxh4%i4X>ylb
zbsJ3<L#1aFL}^c4Z@Z*gM~IS74PEDTVp8GQcw!YjGZJ3C2>{63i9P1-4M$uJF#OBa
ztSoxGYBJPUZ^{6e?X0EJ_;o_qf*f`#ZbGWs_38ed4(?qM^;B)JTpRN4ykv}+^Dx;o
znCnpV3;a$<ji`~k6cUI><FsT~4bd)=-mD$>M8T1tPF56kWskgUutL1@2#KpL7s~1E
zT3}JTH8(ymaz+z&V<Zozx6R#4Ts(!LCK5yTOLF_4Ym9O~9oEAp1?^1@f{iQr{Y+|*
zimJa6&R?p9Y1Q?*7Evz(@u_#c8~Gr=Uk|G_vo^v*6CHeXqHpv27d9Ug6i#=IoZF3O
zq%<|IGc9jl>Md(}1T=091A<9kwU=ucU)O?&<XyHNzj)mK5Wtn{UEHel-)MK-pw;-m
zdBg1}CTXpq2-8C^lb>;GK99<nN=a2TteiGkYL1QI&f<_E_W;OqW1_y4TVRvvZ`Jq@
zi&B=4^0dL}oEr@rWbE;D_j)L0h;6A1X2}B@WI7;?PcAC-lt}7F;oQEtZ7atwdikMn
z?=12x$`Wng(JH;2Y%<C}t>$}>&-U#)64S(tE@8855B&y$=uSc>wPuv?NIxj>Vyk=>
zXgd00ctWrpM(DKBy|@S)V6Y=Qxnx2!6&$`KF3}etEI;`6F5EjdlRY%(x~qR~SYbH4
zADR`ydp)udpenTsy$(0|{j<ZJ``&l0oS<gbxO#INQ=5)XGTO}h$IYDs5#^KQEn-E^
zHhDBOg*Nq0H56o2I1h8V32*XHvyl}Aq(#vBAZslIyzK7_1#@QcO8Bd^oAtZ@`Q=}c
z9*gV-V7tAv;YI@5aqR<j;!Ogxl|=KNFfq~NAK??p4JhRd!&1avGY6!1Sb;xQ5p9zm
zxM9e;<x#>SV&xF)(+ryVul<=SDg-1jE5~4?$-0wl3}3M<EslF-<{Sr_o5uENY_|Nk
z!)vaq*H_|?E8DX-z2>Li_t10nkk0s;dDoxR9(gMB{HW#mA@E_eCq^^l`Rl-FSJNWU
zz|Dw5)3Qt0JXJ;*As*_~<D_~-S|L?$nj9qy*~KS}{y9+u9;OA|>O%ixrpqoYHkffO
zveIr-Rqb+1YwlwI;A5Bb*Dw8QAKxQ;-dPcu9d<Y7yx7~xxpK2Hw=POiQ<D;RVvmOF
zY}^T!J7L+7)7UuEM!+N#iT*GV+S*jxn$Ac<guNt>YDgv*6e~G@bKLUrUUSWjq?DwT
zc)-hYVzpGs7~Ek56mt{~6BQJ<8$db^BcfS-mKWHR5xAGlpEy=AnR{ET%^vc$#_1ou
zvh{vgBdP6m!}Q}u_pzr{&J<!#R#T{{1@@jg7`MfRjRH!~MHX+a6^Pts=v(Y^HeA}N
zn_@kb4IE;_Bo$D%A((#pu%As+%LPz=_L|_oYHvajJqmS*6l}yEQAQc^ci&j!x|r74
z)aq{5qq;|cK)O2zc56RUV_}lmy3&pYkt2GbBRURWb!P2?@LLx@e=dO=pwKTrW76B6
z>h{;xPxrhZU)#UQ3C-l*Z)v#ovcO8`tTu5dZZdUL+>WcOJ9XiK89H*4ue4Ljold+v
zZD_-2r10IMELpl5+B{(i(WBj>q}<E-@2*mzV1>HN4XqRQ4AK(V1EnGjW>9*0|H|9&
zui9VhZ|<I|*9i+*n!}@wQ&k94`s8*|#$Wn43&?Tk0dTK2UN4dV8*s+^Nya7e>lz5$
z#MZ^qz2JB6doS}PZ{|NfQ}S{_&Sz61T9{R<CE$5lA%u1~X*OD&;_om{OA}O)^`Rp}
z;|;3K=5C^F({*fti=-W@WUASm$8cwvM-r7((mxLD?`h00O!m^mrRvtXbqFPRH8wJU
z$U6_T#Sm49|4VeyPCn5#GIv&by@(cmYJ7GL6Sgi@$_A%se6!9d%n1v?mS&@giPd@I
zukO8qJu{16JyMp>$$h)}uFR8rJ|pB(;eC++b5)m4QfDF5yV2o8YzcEqrxKIN)VPyd
zd*rOhN4uJIDhm9w@xy`$s1(K4j7<FyKN<C8-sLx7yBJQ)g~OY$tttfnnLOBccHrvU
z#VoBW=>$!JmO25g;K2Q^VN!kNwFo${$f^2-h|Yo-U{EN{1Pv%la@58`?m%0e8DWdy
zW1IN+>SDGfz%;l$f!RjTuiG$ZqmIbAD05qJUH8%es-jODnB}l-I%4`EqdWTKmhipy
zkdJQZ``s(j(Jw&O+4h<qm2*SMSl81xCfEZx=jZh*EWs@}+3lj`Y2hop%OwY&Sm5TE
zJ>QD1YvPX*;5WwT#2~IsS*bw6b&BgppnVeLN4*W>(=Y`BT3Ai&H<eQ1Z5qg4(th?L
zLHzHdr_b}Sgp8yrR-(A#FtmVd|ACax$cQ8~EtC9dc9NrDrViJqvBLJjPS!8uKaLr&
zz+{DI%Ydrr1jtigBnKeh`Yyr%sm?w9?9SWt{@efbCa+(Ua7mLfk=(tqH$r^&WeiqL
zM?(r)BPX^#Zn)4WLP(<JOsmNZR(H|4nH1ofu2)?-)0yG9?%5EcCB<}7XCLYZq*jq>
z;EJlh76kb+T(~>bQL`*5>uHwAHZ94{$b+}VjzU6bQ;EERQhZsvj!1>kjvzm-J$EEK
zUL)>!vqt}c>tN%GdfCu+r3PB@3%5=>diSXmda%SrMHT1udxd52T=I8W+vYK(->4n#
z!|YtS^_Wk$TnW8NZ*gLeBJ5;Dm{_CM^g<(Z2`O3O5OyfaYuWC9OE5HNhPz#4x@sv^
zHGQiW@A>3V+H(96+Wj*4Ku33-ER#=j9}k&nhWv=l<~0MS$$LSSr2ORq;myn<_(wTS
z4o-u)6MMVQCHBntC6$XC7HoZ}bwKNsT4103@d?@Ms;7^XGDqXHu9wp@qb<t!<2$~c
zcwxjrZXQ*~$|{5kVfQB}WwS#)%>-<SHWSDKIn|*=JLrUDVk^DbGw^GPlKDhLp-GhW
z@aXO-HTv@^XR?A4HcFKSa}?wk{4dqT7H}CVw=nrp<K!nwom?woScc*nl(eGme^G)T
zdF!yb^=Wwg!xSA-jf$rMP5f0=gWD~~hbdxh7TJc#R!>K+#aMJ6|4>vW&GyWg%7tk7
z2QLbXMsOJ^WaoF>)`a8G!CO<4t+=X&Ja{}3U#5Jv#~!M{C90OjLe);2Z9Vlv^4!k3
zvH%T~mA{-|B0NtZy(!WmbEz(563?^20mGp6i4S!T*%n64iggK}T1MR|n@V_rAl?eQ
zsXq7d`)oschx5|=BQZY-dc0@o45zW+(Sg3M<Ur<l6{A$uGj&B^Tl1nh+HO8wxaE_$
zYI=P=I1LPAHHlpCHhGB;-qWh)-Ht*_1%}5%s9l{$&<%<abg(&ud_)yGz$*8uDq=U}
zj@ir{n0h3U?IM>;igpsq?chR3l<jDJ&9W<>VJWolS&t1jPDJG6{Tl~u^`_sg_{KhD
zlOTvOu$eZl&}S<|ja<~PLk`X@FIxmYtgf*-z)FzRHdH10Qj(-Bh;pmO8RR<`M`pD!
z;Rk<4f3Bk`oNeB#ZPwl!Cg%{BmL^$E!^aD9ph7jn;WtUTev>6GAl=1DQP1H%1V5!T
z7F=^g|7<$UZKdHt?)5U8i!wlmj^y?d%u!3FA~O>871cRaKtQu4hQ8ZIZ;DmU?2SaI
zMPXcjW~|8rm26}b_evB)YrHXav9`ul;|HIr^>Qdm_noi4;v2tHxHG@{N;FoP&wgCm
zE{WS0Ye$Cs=a+NZjXzBK{J|q*0qxa+6F_jl`|)a;4O9o4X-W;Es*jF_52vIoFe6aN
zx(bv~lcM0-jM<QMT3=*?u%hU%wP1)nD&9|l){V9WESWC1BHf-%Fe5{&?SY=4I(4ch
z?6yQ>*ENG2xT6a!aLjGeero8hn#S)fA9BO0KFK~Dg9Sk<mdE0y^F}nHr}<?ImE~tg
z`X*<uAsC7JfA(;0Hj9xBJ$UcN>U|#m-80hFM-2-lt&zr^$Lp&}<3D?#P!Ke#RBCTx
zt1EC_c)#7ryl;qB;Gr>NNa`fSDrVSWmI$<Nx7t1LYEn;!>eu9yj39l~pbsHtHuo5a
z*gBWye7_ek^j;~{c~&`i<N9BORfp#6Gn%01_V?oVpL@QXy8b9P{{F1}xq^sEOZCuC
zAsHQt2Syt>+vw=P0-{$CrsHuUa+$}DXp494arLk%BMS16qZwHM5QCLh{+TUZw#P5U
zucmWUqugo(0vIFz>?KgSol(-V+G%SJ3ybc(RP7rL38FCd$U~-St>SvbNMJ$A*JN*D
z`Bod(bU=HkQs?8k{k4>+e4n8OY1P57>hBRCD8NsC>Lm68`!rlVeJypgFgLkSuN5h?
z<KVrs%^BP5O}y8B4m3ThByg3AB)mp611W|$wLiPYG)IfV&u)l8dF?n8C*+a*Zm5*t
zD4*nAdkaml?ZVa7fBDi7p&rN3{|Z=;I&zW83kxLJg%Ro`Sm87vdQ_Y(x4&$v#)@(`
zZ5nBB4c5fM0I@OXeX*kUCeuQTJ)1wiD8YJ?&W18E6^)uf7DG}ebBmXAomj51j^49d
zSI2C76fEV;9tjdI`txrYxRsqdam<nIIJ%8a)%4+WzGwrynKC9dP{<IW0p<`6?C!SB
zHswv%>xv2ePxZD3DGJ+8=TR4WO(8>pM8?jjh?OJ9ucZ)Ey%pmadl=lCN3-1CzS)t*
z02{T~^GxfYi~h<xUYUbK4;P}JY;qtgF>ZRDMoiUyq*2wvmyOa}i<_n^A$fBPoC+;$
z!YjQ!TCD_7Ox6%r;?WRpj#D=89$n_S?i#`j)8-0259ZZy3a8mUi9(pPVdAu^h_!j#
ztOs)04Up0f9<I1JnzNHeE-<l6(O(`hcmLTZo)tyT5Qd8kF>r<4&0E1pr&!h_tAn-F
zIqS?yK-k6n(Rms5A(Mak--s+)^VkGg2dJz)Tzws7IJX?yuO%Lzq3-lazrq&Vf!jYM
zNWVX_w0PIR2nlrB&1W|nL%dP$qXFX@pO`8AC40v9HkCPRiWz(Tq~hAW7K@E7kws$@
zv4skn2n&H$+Gkp5jegl!h+yT2g4Q`whfX?r8rwQY#mX5%O}Vo5V@S8_uZIAXGd;pm
z=!(81C$E37_L|=pX;Z5BTg5^HDJJDZaCTI#jrPd3TOBrPg5)Ioh0gNCl%JWc#K$$!
zMMdTQaAD1vc75yG`=i^9*a+_xU%Q+H!o+qZaW`C4ed7h2;Batcv8Tn@l;PQdzM<Za
z!HX`mA`EXm`9nWFODBvij!Geti}>jU#ok8XZ_Y-IpOK9wf@SB480y7hE6xA>BGG&L
ze3i%bT-$Chr9w)uCPv!}dgaVO4zKB9Mr+~KaFQPPC@4_c@^M1mPawunc;pit?O(m>
zZ0Ja@8D><uu_Mlyc<f@5KGsG2WJG<5rm{QBye2EzqEgra>RI$~U*G@5p(h6J<G(hc
zz3J<{5PxB4VHws-R6m%`Yr!M;NY&J(aG-6%tqdOx;cZ{HxT%C9SYTxGf_UkC-V`N0
z@vm7I#D<%9j2K1UAbrqH&yk{O_Pb59RJHY^m+{siL~Figy>;Qb=jFJV?Q>%ak%5Pt
z?lz{rK5T4nLy6PP_^%dYV`69{RY?|1XmD^i6Uo^$W{4@Nc0|YqMG8lPA{fZ(#t39V
zrN`Kgk<<FTCzD<hR+yIhMXeCq1lV39snKUu2+mF3D(RyDoj#!Tu?@nr;-RRd>KegB
z!@PlS&AT)>`GQ$~lfQrJM9e5eb-?`^iCT9ut=fFIrascHl7c=yCJ{%t*WYpK0I@Ip
zbbOdzXKcWZ{3$g@uUQGAO^xp@pCxA*zIfwvwignblWAJ7!45eguR`m#H)lEYVU|Ww
z8K?Jhx5gKoCh8hAmo0!b_ej0Sb&LrFPvw|nT5tQ4l!S5Y72`g7J;RTyl!R2f|7mqL
zW5Krf{Tfv><SH-=;%JMo-z?8&91G+2SKJTHipEy^Sp@jqevtky>4(pDM(n<@y*}>F
z_6fCPI$?gF0uPXDTvUlaxTg@NJv7-(ECeSC_*6o854sPt(nf{Qni?sU6{BUV6PvE@
zniF;p0y;5_Zzg^RtsCH(+C8FDT-ksL&Soa+gvm=(nUl5&l3i$~lUcX~s48`3;8q+b
z*>?;20@k(#V;Q#Y&$`pfIgTJ}K`_1J`>lhYl%6w7pA{-S%cY+fiCb+gcnJ=}#e)Ny
zX-4~cK)Jfd)qbPlj^p(-_kMaoLOk*VSx5WG0pvMZ8rUpX{k*!RO=)Vn(y@PHH!gUZ
z{@r0xlnooc>D&y+@SZiPBC-`4Y^5mn;(!%^j>!_8kua_K5=ppsQ+RG{J)FJ;mLca>
zxqgthF?;mP)7SAkC@Iq2Ht^xH43{@zue4pzaB(UvV|?n4SnRfKH;=O*f5iD=HT(Ty
zj9Xb4r*eiJ^2y7UwXF#y2Hv`)sa)MvcLdF^JO=40ZK*tr5_R41q`B8>9J|Zfo5)Oh
z!*_6Re*9Fry03JXTkB5LTJ-3r|KCE)1{nO^9srJfk-C=jn_Ln~9QqaJKOSskl&CjS
z!)^WmD;rnCtL6;Be>EJBk_2BPsw*+@$9_d>n`b@B{e1Lrq<Ano<lxWvt8FTihf!V3
zk{V%}h|}6y`+F?d9|<!uQk^v!s!3~|--GdY^GLf>`f%0SHhX|&_TAozAM#U1Nl@kl
zTHSrEVCtLq1}KttdXK_}>8&=){a1pyTAJ#Ue6Yg&<KXt|1HU(1ucv`$Po%#cmaEL)
zx$JWp@d@A32!e)*we%MSYGh>@-nk|^7<%=ckgkBpC<$Q4AGM2wo(s*zi*pR(hRIP6
z1L$59<lLTuvycZ4t+%$;LVct-A{lh3Z+*IKFYwEYKgV=yAm>DB>^D`YX&SfXXElNP
zY9e|WJ`-vV4Xfo+y>EZEjj@=me5CdHJGnKZSU~i7ArLQsm&0Y`!?O&Ws`Y7Ta>oF&
zg6F3BTd;3-^zUrLpp{Fqzk+*xKMtDTyCcS!>MZvh3-S`m)z$0m(P_b}MwWk$MNVLY
zlmqy57rn-I@e>L*e>W$pt_T-B%B1Dl`dUthPKXpL^{aYl^}a?OFy8Vul-WJ5CaUkX
zb?QlFb|0ESE;3F*osj{hB~zsZ)K((ihXy$Rd-mmX_aB+udbh<i?ycLu(Ys}AFV_Xw
zvBh1ATYTuf`}?XOOD}w#!15#hDs>9noW^^#t+8@uqG5$-M0g}{V}xzUaVSoRjXL@+
z)9sOfqwZQM-@~+a9r-ZcKyL8Pn5;vz)evZIG|TnrQt#AQ`!hw4pZnOn>tbztlS1DD
zLbvMpqU$dy*n6tkdkE_Oja1bP0sw7e@|WXALD`SQKC|-{=~6D?&jt4>U$E10scbv6
z<sVoXN2kaO>XoXx@MJ>TyMb5y2Dy%VejV5V&D47#$FT?2@hXjC3RWK*g|Pvnm8*F8
z@vNsHUQq>)ZXs2K4LT=9$U_v9b6DOz=V1e#h}Kd>@mF<d&Zo$~hx$JZb|zIZ*U8`N
zHJOIRAYJD4H<7^Roh$@Hptk2Pn_K#}LqA7TO#IMb)l21F<fj)}uNZlXaqjD06!!jA
zhz3U#KEeme=}YcfI{_5}ac7u=?Ii_yMVY$87AVM%!LJYiL?>2B7)533TJ8lBa@YW{
z?RmNOv}OcP59!)q2@r`;Zly{w`Jx>VKBgpir#swqPS*(fy=3?@PzpWqaf<>`$q6rY
z?$)O{DWXfWPML){uU&G9a1SmDRBB@Hs@dumMCQ-U%#DTeT}qPj9%t4HPxdi;J5v&F
zOyOaY%t8WmiXp)fX97h<dXL>~!MsoU(4C;aOU?z)G)JbH+*hN)q;C|kwUrb}s;Sil
zDMFKl!Ddq{%z*Bi*btx)D^7n+fm+ysLejZJ^;>n<NKG3Zpy0*cY}T)f{%k4h@<HL3
z&^$xUtAbZ%cXyC(I(H^?i=u2uB7ji(;XC%*AO=cvx=hDrfL4Hm4nh-{11g>H84ild
zW{HFMa@_~|L=R-9V3dNFoZB-ZRSHDJTP8}B5RfX;5Okr&;HZ?1RCV!xe>(>;7{1B7
z+xwOYvao!-6usg1C;gx((e#w&c5Jj0+PdJkFM3Md;8pd@$r<Nq*I&75O5cI&Cxs`x
z(z-9kYx;I}d;-m^47Z#)kf$`$+#>Z73NjLw36#GU1L-?1o#wlf3`b4KWHTm%+w9&G
z@r`(qeC(av^Gbg&!Np0;bAwwl&foKQovMAEz@9zrphJ78IG5zTBq6fykkY;54hPOw
zt_~LoHVXY)o$?n(FL&==J@;WNxS+KupIjf7q*9${9h@v>SD+&KRq}Wc9|w4<TxJ*$
z({TNDT44GR58vCxoA!wnuH~w)O<Q`7MQLKAlzI%D_h=At)84YjVLyR$lOJdnjwU3k
zOYI)YUK|7?#`JTGi*RKA)PW;XlATmF@B7!1Jt!R2TKA=f1K-tk7uC%qB*y_Hn)pq*
z&$5_1?~eV|s!+YJNBxu^88_@xSs-zy>~iFS9d$!>G{}9$MaG`A^v15T?|LFQ?Yk^W
z+6&h<*IBMLX5U>@K-ZIB70JX$?F`+*B5CJ_l*XNPV?w@6#EwU4FI42`(1hHcqKnG#
z5%{e!<4cnVN|jH^pS=A?36+a(1HGi>vIL>TY_l;7=q3_il1mKz8_N~kBZU=4&&Sd@
z7C>1!vq_tHXST&%_9a4Tle?b!m8n18xZU-wxD(a-`Yi&TbzRoCX++@Y18vjaJ{`7i
z-=Vje9BcYhQ*Vj;J!gfl|20%&#hd$qj?GMRnC)&qf*qJxKqgazgpy$B3RwGibJ<@T
za%%b(<5hHzv+*qrA}2?VD;+8^Q@Re)S2-dnlXyBh3K#bp?L%fRHfc4*dv?*fthav5
znnpcggzcWDACC`x$$AsjBJ9w*%}`mXc(8ndL|wW5DmtV3WpKZxAno;_Uz`2T-cHhX
zhF8!}*^qc2q#rV#X;Bgat$R${Il8U6B<juel;Vvk9MoA0-dPEpq;XhBj0g@Y5l(lW
zPJFMC*F=jm3Ax`h++6Zhzfzt-|F-Ft(wPw}wE`uHN)am3lYV|Adk88wFI&`dL|W|-
z=w>L6YPj^yWIkECP*uUD;RbeqsLHzOfgdlL=^JWza7(`^!2@cVJCw3_vR8WdtX6UR
zzU6%H_Lglt-}LIbahf`0q;zV0{ja(4N9qFSY&Uk}3`I>(^S^eOMRrm_0$!<huXXvq
zb{G%0jF3P=rPYT~U~ClXFF&X^qtT@uip_Ea9>z$uqU;XsdIlqgwpY#rOe(6U{XU<p
z$5kN8#e7G9Sl-0pCs;G1y-`X`VHl?Las0(nXYm}pdk#+WdiPTDvSE6XyZ3}LgT1$H
z&M6p2P8oP8L_Z}XXPC`-C6xlF8a^ZF7<izNqP;WMO)6xFuc?$0vX~K;h=oQ9g!b9v
zBNszy_p#_@(uz>bH3_Fgz1_o5)RBp_B9ilevG-O{ZFPUUD1GY!EyaqrXtCgKrC5Og
zMS??HJh;2maCbtG76}p(ii7}x(n9f|K@(hpQ``#l<lW<(^Ih!mUH!-2SKr+lS((c+
z*UX&H?~y;EqAUi3!mH%zGXU%%T|kiiK<cK1C_sm$QyY}|wbHC308vmtJ?!$eO&`I>
z08EPo5`df52OBebbtT7j`2pcS-!P(oznAT7IawI6Du`$RIG*8Yj8XEEtf0<3#n&xS
z;ao|D%ewHm42z##Ii&TDQ#7nBX<L_$YYpF|wu=Lg_+tE#3%Cqt08e#TbI_%44q#ZN
zzb_nxqn@oIQ8)q$Y;L$JMgRtm-;|+ciVB}(8qo25Z`<-<ye)!l>k>d3$%6wEZAFs%
zT`TCAA*FV4hUJZAfW~?%&d&uewLS+*cCiAhnB<=w=NHs8YmJFLn7OYtZ2~QNJf4C6
zR+k&=>mnd&JmO+e#m#uUS^FetHdNds-?5@;dm2KV0mL~4*rELFi^2_us`s~6=Gq!R
zcvZ#|8mV>U38!Z?IBE+bTRA4q-nuN&nyqZV!qnXlN=e8+@|bqMG4a5<7j%m|?OdQt
zjnm_bU0jkAq)<@8<!ay<1aw=hNww@<fQEwWPXK9PTEP!t$@v`TGCSlp-T@8}l8U#9
zm~bBVy0Kzi>bxcA@HcF6_0YUJO`aARO2R3v>I?Aev7{4P)^ANUu^hH_;*JQ$J4@EK
zmw<k$Gnyq3T??pBZVdGO$Y9v7VLg8ut6M{RYxvY=Thw;y*1Oa-d!7`Px&iR!T2)4t
z9o7En<_A;po;^*bV59MA$yjB$B#Xn5I%Utj^VDZP0o&`~CGKLktS^pN+)id(ezf>e
z5tKY2k57pNznJ&-cNNiI8Y69h+8bQ`)E#2R_?)5&X-Ej8R(nfrO^v9C4<o&i+o~zr
zhRsW4XZaTaWiegI6PH(>U!HF-JD64N>2?qmHtFB;@cJiY0}5L&M$}OiWZ+kRnh}L~
zXh*JJ>*PJ5kEfBpr%Ds<&3R?_q-V)JrMY@6fP7tiv8pIcjo7=}s`%<B2II0WAr2X<
zbFs7r5Qd#c0dG5^d$*Is#>!XDdx~1RE280l9-T}2k$f)k@r*%7@zBwAl&x{xlQT(S
z7^F8&Y-R%*P9XA|FLL+AEXuYDeA!*cA9qcS-V)=c#YE4RG{#}lg9N+n>~AgZ-LSn2
zJ80g9nogu@an<ptbg`wAq5=Ps8PxObZH@$YgljND&3?x+L?-1~4HTK83w6>87uWWj
z@`4Zmg!)7$|6j5iuEycr7Q>0FF`RelWsP2)AIt0%WtrLs2QI9WcbxQ_SG|pCV^W?e
zv_bL?$5Q8@S)?MWUyd~7u@v}Qi28kP!Q=u?EQKv1EiUfW;EP!0)VZP2BGLZ(g}{6R
z({k54IUD((y9_s{M|<x&s<W9z&CRQqM8zv4%J?nd<b+0jfJ~Od&`$+$1Gu9}PKUpU
z*e2cHo_&9yXn(9DI?ay{mT29W4Y|Tf`MXD9oP({}TB}TYTfF}n+cf!0rrw81pdJot
z;%giuIk%uiLDZh2D_aLv7>!A*X$gZ^J1P4p>tjzI(FYN^`789UxHBJ=S+zM!`Mo&6
z6beJD=%wnIL0`nxKbzLvq$%uny17od_@>yZv2cf#%G;-MTkFLL(h=_)WR#LnY*uuO
zXg9r4j&Wu@cAGv|-5>(S{^$WaSYH3}%f}AY8v%@1BAnU!<FT<`?wxAd*c{ll#Ck22
zwy`XZnfjMcms@bP-&O!<>m<nZlA^>pj5}SkXAi5uLT*Cj&N5j9t$+RG8ywcgVLbNi
zfuc~Q2p=cFx_j1~_gw_5(r2$0Ir78KJ<Kq{c_Ub{UE4sZZCE@s>}myzs@UEVFlR#;
zaO>cFFzDyqJ^2{FQYwvh7=PDRhtGtU>~#6qPAj<t@?fyK+&TNq->aHb07Mxn*iAdJ
zv&2pO#sP=j>&7-k0KhSMA{YB)Kf!3q?Ssplw*9(VoQ){kqUb&bIG`K|N`qLY%n&AU
z<3Kb1Z3e@16{$LpyJeq44RojmIGS|%Z3-AeyymWw&D&W+0#dR{?Ke-GSZHTa!CE+e
zu*op&t(_Bn_Mwi1;*qaq+<7JlWLZF{$SH#b>rwo6cGYHe0HfB;IN{^pcI;;QE1)Nv
z-BlB_X!|j5IzZT{ftU%JQa2q4{9}iN<}c9H^v^h@o&T&fNqoaCnxd+k*svoEHM2en
z^Bab}e99^ATL!fKuF5mpxDSAKg2k`uB29Olao$~3*N#Dm4+}&D`rGkJp01)%;XrBq
zt}iefJGy}~0vY^^zPOG_IIf}P%My_V(&hwcTmM%KN)wgf%qX(sYQ(?P?<ZPiGrUo$
zs^=ze$y$r`)fHV8C?&Yz4h@{lv^5HwG!ihZx4TwnO(w)s`wq2SAkyTOP$*T{Y)6S!
zC`vnX?;Cy;oH$&PdG?G$s32|@@LQ2da-e9D>#FBx-!#5n0veMQuJfZ7W<BAc`Lnik
zq2r04BD}Cd*u@;yRMg_DG9gbiRwEE{?8io29Gm1Z^}C)sZyMd=CMsIJ__{Hm&F24w
zz%&u9xqD^R({xTA{w%ODFBoTPwKy_(*kA0>*!A@<nW9yfHzC9?FD$Y;2&hoiunr%|
zPCgf#y!5j5<**#)5$yFoG9I^44A5I&Stp<G{4o?FHi_@Vci;)mN2MDWhR6xr5fdgH
zDUMKo(|<ITwK_h#G3sl9$vkhxZ-Z1T!tG;?A{Tbk?A=0b1TDs@y^&7zbL!`+{|kyy
zRO_gYass1NI>h!ISY|}Kjw@nFLIQNuOTUF;g7=7psA}B=lhzZ3+qdle6vhsszt^jP
z$H(dP@3}S;LA=gUG6Nc4yW&S96Y8}#7<C2Q_e6iW4ld4~Pb;wf<~xPRLcN!lRyHPf
zJ+{H8XhEB|i|gaC0p@~@O5Br4Qd&ulBFgaJMXzyvMB!W48SaOwxaW;(6gO!IorT-I
zRV3!MRbyCL#lK+9c^?W3JX5FKiu~X#c*FIxb7#2&Lwmu()aDfI%WQXd4L2C|cqbF#
zr~jJZmdj8E2AT+;^A{DibG*_x5W8yV96k*_SGweA7M?P9f9tC4?7Xzy4~LnJ_{2#=
zb#?lC&{6i73_a~Fc<rX(M8B`MS1Z22-O)s*>r*hR<(Q~%f-;=~(d}zkWqi)|hUB&$
z2w%u%5B0g~r^L&u+s3aO#elyjtRwFhpBl;2SE}oHY63+W+{PXdObgRks;ui9MYmQ&
z%0*XT{C+)W?vSQGdEB(2oKHscwwuLUANfZT=VVj2<Ib<PC&UHXD`z^oil`WfB^4_c
zSS+$ryme#Ctl|<+j3a47sP{;4l<LUgv?ZhV&yjCnWW2mAZ6}KW-M>JL4a>ix5PuRA
zW+N7yNLtp*uh7dx{YZ<dFI_$BEFq1@Bt^{z%eyoOX~qmZNl{~k*{a`gy=<DF=wh>e
zAZ}vv%Fxk;Qj{i5nLOp$UB(oY;;WP5V=Sq^IjjW{M^S~Y^4HY~X7ZH@UtzX$s1ipr
zKZ&eTLTie}(Oiofmq3o2Rn;+ag2vc5s3bxB;f-UK(t^BlyG17&($E8uo)Xt9x78z|
zG5+w@pnv%s2B5wSZLuy&peH5%acMLdM)hei;5(hFl2xpnGIMdqx`D{Qk{VFbcxO2h
zSne+3)|;xwCX%(>WWThNn)gQhd1vY-e%x<)LOid%!q|;iTIGaKM3qhaizXQ7RS>NP
zNRf>>;r3%qHG(r0aim|2L`>6lwBrYwn@D9hnCyWhg!Z<qr(!Ee4iDq<q+;K{Iu{VN
z9*N}<nE#jThWgCzKMOs{7KAyI1+z@}eT?)mF1@pSnP0@4zeH9<qo()%mp-n8J_)?~
zXnkTV)nqCo0d|$9tBK&#{ectyS5w0Ujf(pa_y1~YcDMAB?D>!m6-|XRP1Y5wEe6=Z
zKx^mOo2Q{+M}!lhBYI5Gb|0EcTrN5Wxnf{xK*?H5DGuBtVI=FHgYS2|sq`!9Wk4<k
z-P<%$ee)o1$AP^<Bsz~v#&*@vp{MP<(0GgvOnwUu2^uFA&$Ja^p)+(WMa6#MHp?({
zh4uwJrQ_=Ia(i9wae^d+Nwh?1eoUe3VA!RoF7V`FHOajdlM$47-%JY~>h5i{i+DG<
z%&v8SfnGQ~sVOfpSEWUEdi!R0?La&m2bmmUIKFRM0^iT^lKsUz_Nj-MUvne=i>$F>
zR69xckLJjaGv(a2tGwMdFj2P+#vFU%vbN07Y2`NSM|xJ`6~8SD;iF4Iy$JK8{ED0n
zSTy8k5Zwm`i4P4vm%&u#qt2I>1v-8T%!DTg>e@8@Mq9=~TmTR?P8c{%QX{QzWXI$)
z)~@z<O_>XZYo6&5Wo7e2rb9DoP7bRw$PS20TZpXIl90071(K|GE-jaQD=*~E8vgLv
z{a$~=d+q;k_xhoNn?mnY@VRmC`@jRwW23&gk?)4r2cbzd^;TCI`(If8=b-=dO#i1#
z{U3Rwt_hm^ipsIQgFqm&{Mf-d>_A}=1agY~;qc?XRJi|}4@6v&1nVy}KVj$ndj3zb
zPaIiE_Ocb_m_AfzFQP-OHUW({?fn`NS!hNbwbVA2+X#1*<cq8^#5&yUu7<K%>gWOF
zl2-<?e>`MUIwp+0b8A6f&}~J0<VUS8#!;xtyRx;uBUXV{X)32W3`I?D)K@(o!(A>5
zyEC(S5(8|!Me85Vm!-D52wvqi#oO2w4-oQaK8`3s14H6o*=b9=%JPi`%PptFsDFMR
zD4bl~uWhLEGRxETTIR2hY68^b4L|!Rzemb8-#AcUf|lSa$jfZ-2&CEFMiY7(SmJA=
zr7Sa21<g-4T3&M;3`0l73^(k&9DUrqF0e}&lr5Yx8zy-+zWw#b!g}C*T9)d7XOu~)
zF3S#bHRIp<#SB+$k_fR!O%NO!$78cR;~J%CeINT&mTE67`q$VHM?u*D14E*Q8T9-P
z<>+i-4+4=~R@eXYGqWa-hNdA&=$oEWuBn&8$-_o;(kpmJNe6};Q~DLWy(2Tdnj<~`
z2W0_kzllXVtCpD}G?zueN?%TAbrF=CB^@4<DVCDLZM<7P`vP@=cl^XEqV^9>e<t@K
zbYo~<iClBp9_$zK@R&{nnSR3UWVG=1?i6=M8C<ve7LV2}hxMe_fa@&=-QbouC-mJF
ziD9na!j%RCG9P30@8n9f1-`?^ynD=l`nX8vgN!KEGwsMf`qsc0&f*pZLsR~8m+pjT
z8DGa)!Or0E%q&Lz<U&z$&e@5_0+}AG#Xr2{iOv?GPheyB&3l2g&Zk_ZtDK<@3BRiH
zd};W>`81H8Dk(jIr*iJ*qSvC%uRa^U@kA%Tul7-n%c9W!MdDC{By~$LWj~3o2P(e}
z;+-PPi9TloX*Su&mn(+%vquO0@f||JjIwukU7a<N;{1%<0uKENdS?2FJRtj>Q7Exm
zich&eBZk`w#pSBc&BQH!kM_d`yg^e#<em(y1Ym<Dy8mPoSI|{!pEXYza<(g_h7HZ@
zNaVU<TxA-+RG+Aqnytg=6=vqX6r^D)Q)<owvdKZ5FD*dR&9PU#z~|YR?DL$!_(Te|
zipJB!hW%%jsviKd*e}nxU(1aZ+{E6O%!sIMc(5w#bOZf#0DDM^PCR=2`;NUbmqx(i
zw03EEmOmwh57e!Xxe|@Y8j}2i&7g5rm<Pf}{Nbp)WAI1uIB*Uw+F9oxNQtqXO`@y}
zbG@{eXAWeQ<xT9Hm*sszo8Om1A>gu}b>318CDE>?J;l;>C8~DDZ^v)1`jQ?q&9!z!
ziQpf;@EZA;xhWh>kF6mtn6A)P>JO{R0Pi#RdZ`bD7Q@)ms>^4sr5h)-d?O#D^%+og
zR>BwJ=&6*ATYM<Qke-u#XU4p!uk!1Q{U5&G+Pu{QipXyfrZ(vWTeIrYQwg~-m%Evi
z_rPdUW!nbf{l>0AZ~2v@n6h#6=*cYs_I~2ql7>pC!3wBf=8r|%E1xO}>~3@_<;bS-
zLqSlerG)O??;0x(k=@;!w!Set8eTGhqj$U*A(9Wb&NC^Mtv`-tM^Vn?+PsUXG~5Jp
z9))L=qA;(aS}nOkFMRF50!&3c8o3E0=&2#l78(wE62YGNp4S@Z#k;(NzY}g@zX8q&
zr4!mxLs*_JTQy|eV2xCU$Ugm^36BV`Cj)i)R+LDyoA8Oour<n9^43qro{z4@qe{1o
zzB`+L9P{0GVsi!{XD7-vHCDvU!$)C`uQg0LGRsbU4X1>^+S=D03U^c$s%dG`4J&<a
zcUg-5-Xrzirk<W_2u)vlJ}@IHpwTV75hEcceJskLR3FK?eD0bxv2=DNe@l5F-Lc;2
zbXsOTHTf{WtH3ww_4}FE3_L9qT;@~@o7$G{n9J_@AZ4!gJAP4=DXZg_<#c<Q8`%eK
zX+R;F+@jYm<vGdut6AZUjF#lVy_)<kg+9?(HgOhbn~$WrdYQcS*zx9YUr~D13vs43
z%k}5o{S9jfE1>0LHq)bqC5K_X#~PaJPd~n^cmL&J8}h<)3c@mU!dTkWw`wTiBJ#i*
z=F%&#((&$<uka)!6P#&$FY2{$wqUQZmuhuW=fZ?eY6R+*MaEO?t$EwKa1c4(of=#M
zQcW>0Yq|o%rF3aNWO=3Z5cOo10IQsegXek-=N#8RBObQNbr596iXS>27tVE$)uYsP
zXGSJB(|%bDT@^`Y8p7|`fxCr`%E$q>9VQVpgoIHB!>~YsM@=h_d^S6v(E|g=oL}g<
z*qfN0)Kcw8>B!qZY=Y=->>_k}=dd+$HusQ~s2GN@{DnrKoG<>5ymC7|5SwF3?O^N?
zRj3hIMyu2IFggFHIN#mjgySu^i7#MJY9JS^w(Yl6ZuTJ+aI>UB_=5Z9IKSlc-rqFF
z>bC}Vdc?&*2CEe}<y|9++Y&uD=le33_@!8FE>re&C*M%Q1yo9>62GgB$1y+ulrZJ6
zSF}Ekm@iFg>npbC=z3$&1bx5YGDQ=2df9F}Bb!Olu_J0Ys;0b`zU!^qdZCC-)R}c#
zgt>dmV-Pd(32HUomS4J3T0hs)$w__t(*nc68b+HRrDm@@>aGqH1RY-$*S;~ij?n(v
zHB<)GBk)^RY5KY0ZAe3=dNyIYD~a1+<WMzFz6(0i2!Ko2E7-aze&R|P%Qt0?PaiFk
z?S$6u6hC^?)KYn_j=uQk0OZ?e;Gz8tX90%Id>=ru@|3h!JAH3yv*#0WtAZ(ggm@S8
zQ1X2#SpimOFTI|xW;5s)`Sw2kFIoQ^53`^9xA)F3gGRMgo~O-JWd*)>bSvrW!kgbW
zaO<(|;wtIL0F&h^r5S*WrqnY(Y3*ili45(3RPod5WR0SGg!!jS0@b)BDm;3l-D0&4
z&Jhy|J1?2ul0ZAiT7Gi6X(a8RrdsKG3nMG18vn8IAKH=SNr|W&{WjnWRjw_mIgh5v
zof^XSi4UeEswAjYSx+g{HL9gJ=OC-`aVa;_2_I`ZGcRMFVdGAHegaJfb~Oa4co`cD
zl>MfxS4umQja_|c$w0UTb!Rm|%YS|c4IMY5_0neV{mr*OY06SE*2=W<bT?1XpIN8k
z4c1Ljbv@(TH#9$=nQxX$kunLH`I4|La{vmV@x9J%CCukVAC{N1B2GbK6YKeCft)<!
zjIBg8jpAZm9G<E28$zz&lKjprkv){<6GMrbLTjRr`#dd9o_y)|aLN26((cuiEqU5$
z9A`c;w-iXjT<oKTXe`21qyXy2k!^g4o+yO}S&O~ze#!K~U&g4>-65lR-aLtZJYD<J
zM_21K7)+}r$kOXCtLj|fJbI2crX7~KZR~U<s}#XzZl7fa9|Lw`*c_hisMaqxG2QB7
z%y-vcp?$p^C92@i`XT4EVI$u=Z{tcg^uxBOlFQy&;Ki2Z1Mw`UL_41>{;yy^PG<o@
zl`a$eOcA6guz0~t8-^%QH8(ah*f#={DX%dz{yQs+`Pf=oxy7fQ2st(yK`gK2R4K~5
z%zX!B+;S4TxH1Z8@8M*OmZ+hYi0T296zc2SvMSEXSlm4L&V;G@A<PW7X{tt^Q#JB7
ze>_oHCnYB%lt!nmP9CS|GD_mKmKVjeyq{W%LTGLxg#-F+O*7N!TqXvS0Qiq1vshWN
z13t0b7l&0gQ<*a-U0<38|4j#kD%BD!2kOQe6XtYq7PAi;vvPdtXqwZ<T{W)jid2;Q
z3sI%&jSFSVeC`kzO?>%fhj(}5rn9o6d7^*-Y1W8lGK;zM@{wAJfRkRTetL2IXpsmT
zo2V!o;KwIz>{&6(Ls_$47a<q4pHEH#WVHNvixg=@-z%QOL@zuo!SaDmyls;k+-DT8
z5XIgLH?mocr<?LD-?K(=sIjEEUV;5T<*(FKU>V|kuwZ+0UtTep*#4ll{TA->JYM~#
z`pHB8V6zs<_j}E-No9$tgy0&hz&d3WlEdZ80PEfa39-O$#lirldWN*e77pfNVr7H%
zStcpW$LaO??}(XM48ABHG^4&z1OqU>T^C*Q>$YRAoCm+L!ATr^%W;+*YPbGlFK%KN
zq1KNBX9992o~ddWuOpv9%zo+NMy9YkJ-gNuDXJ0_rW(GHmJjSuI|59y{%NH-gK+cU
z>0|Z`aOTXCK6bqMfVT^E{gtPyOhodpraV1`^Hoz%rhgdTYQ!Z@%Pg-Y_QpJHIH$^w
z--<2^VNVh|0A0{jKiCUUauXHpweHMqIy)n>7BZB3e`xbXb`_|`R+nW-zJ9&VQ>@$R
z2w2n-JKAN*pmgmv0a}gtqfqk4>lH`tV40aHXScfy9frr5PjHz+0(%6X;%RLo^JrV&
zLr5a4{e>4#%2PXPn#aT+=0=aE!tP_cJtR^zIa12iR8-XL?L(rYF~CuWNfi|f?BLJS
zHY&6dF-A&hn@_w6r~PWC<Cb1oz<%b~IA4)Trgvq8sW_U+UR68N{Jjm@Ec8wAJ!KuW
znlA-eolWrMH2dBgs+X@GT!&dXVLlENO^*-vPwP0qM7HfEZ;9n;4{Pg`!R$`5#%eHp
z7vZcu>pTnP_WAXl_z>RHXk$wmL;KfD(RWX9+Pb(<`=8Y_IwMSf$j8W~`o1zJ3gm;2
zwRN+t>8Yu|Z-IJCy34q9%WLe)^_EpaJvlL>T8S6NO%H1}9VsM3##brx@@*Mkj^(pZ
z%Z$dD|7=4QDTx3F)y&&a5`3uB>DW8?1P1f7%6Rbv@oKP1r*Z6Tx|C^r&vP$l;S73-
zWV?=7rlGT}nPXRjsIEukpz*wUCyb0d%#>x=_cn0d%YXOsMSqH<6lR?KR?p`1Ubyxi
zu<a;9v#{MuqvIWKbq-7ur8k#5+W6Wel~uQ&Z4QR>odcs~S4I-d9Qg5_im@B^uAUWT
z?Y<Ow*#Vu2@TWROc$R0$iy-Mg9SLyS@0`C9Vp|k&4G#xi)vK~`TQNk`56E3`ONJsE
z0Mo~tm{(>AmG}&<MJ}ocR8eq5Cle+mX(@}|7XxnT7t3PmeCy95ujFp+T-Wq5Dwe)8
zz7!_UBKz~bcu~K!hwD(k6>2ya>MhYIa4a4SPOz5-<gV)!ubNQs_-o69#vd))rDrb<
z?b}Z#6FOEC3na70Dzkl3q;*owUu#Q1oc+?AS(2@rE@6!;w{(k4>z@vU#&58TFjaN&
z3#~3R%GGdlOp9Br_<Fy_`>CM$r#_bn)a+rEZ{a>CN6Z8tXkXCR-R5a%tMoTY(C=Q$
zHJ~M5v$_u5$d~eZ?8I)5e!aGC07Jvn@-_Y9glgjP@yVOpRmG|I4PG^m{=V67+Re9D
zdcvKu4vHvD9hEl26bHq`*{pCnY3g@Gg5Icy$Ln*arZ1LS5yn9hnWH+onxl9oiuctn
zaJm7{Cb_=5J<q)#D09Q30u$^^T3SM{q?1P`zr~0jA7=ULbB>#h>%=d4JsbSMd0Jy?
z#%wqCRi%i+M~Y}OO}2y5eQCKwpJBRmp4eHij99}7G0zw{rhMjOe`k6fNjtml%G3Yq
z`^Iba6sfTkx+69UL5&v@s~<Vl@5#7++fZ?PB$?)Zyh9|BcnY*~iA@cgW<v;hoc&G_
zr}=wCWnWC)eE&Q}3W#JIUc8YgT2U)&_*5i+_jJ%>c#^A9WoVnQW6zc1l)I%+pb!_^
zT4J`8BQULb720p7-I{;y3k&r9hYH}Q==Sra*BCAT34J3MyMR1>*T=Kw#oG1>TD%U`
z_mEN_DUwIyjFxIuaGrWE-~NS{Um#m5arWa#?L!=<+3i-@vcIRmb2mmy8XY>oisxH-
z$7!0eL$2rS>JJW#rm1VT%D3$87Y}mTgk>V%u0ZlXJ8fAQ^-oYcDv5H%(R2$qn-Mj{
zd%D5v?`kqqXG7~sI-bhh#PzHEK)K)P6c^O0<#%{}cp_dF6CO>Rw3`W2k6TW!O&M3I
zEg}8I^L}q*?C9q>bePtV%BcGumw>oSK^`)3LVBH@XFtEg07l52)Jz-6Tiuys@WZeN
z3ubw0m6`x?7(-FKnJhbZw+B9iR1Y4wh+-@KoNm)7fJ;Qa(~OR?7UoiZMyFtB6HUZQ
z`E2TJ?;}>6fuT>njJ62k;=}KJ3%@zKwOl2zR)Zn%nShgE0agxx=Eau~`h&1>xhI2E
zc^3Y8z7)@{9WD%!hmL3uugzX-AivX>k`JaW8_8GwXtzDG9I!7~-s{qIkey>uUl_%X
z1Z;ic$uGp`-zXSuvK=(6AOD7Z`d`q}|D`Sd3(UD8>uyqb#UzD8!?&;=p!dm_PQ?h*
zZt~6r#>+^Q<VnBaJmS$^oB!;Bu3@vNu0+g|YUP)!)IAY&ZK7sd*BLrQo@YryXpUS6
zw`<N#j%(O8lCDKP0hhVK%k5mBk1yUYZY2c@{VsH0`#m+A682zSfqPSd-OaKwuiTm8
z3SyN71It)h2~>!MzHH|62e)SULB~KrIWn6u(XmeXJlm^PVb4wn%UxOq)4nV@W=oG(
z+Ps^Kk2{m5MS`RythWvOY1ebCc_52i+>j(m&B@Eu9Y?Uh445A|URcx9`P#l%tZ`^6
zy*{}lRIdaj1J;1obw>}@1->g^H26^N^hC@U)ay}Ovj=W8Q~_B^P0{h=EcKWRD+;44
ztYuXO9}P~Pa2?CX;V6MKwQkG50#znR$#m*%(n}i)+g)wM|Ll?e_pLoQ{W}I<*XFG|
zFj{96Ke4?hjXx{Pv1tAp_ovI%{DWS$Ee$yyMio!3oO$587U;roo{s3?<Q~@WKB@}1
zO4Nqdhd!xk4ys<%S?wLGzTh2^ds+GdQw%l)kJ*!0Hn-qbdD4cKx!D~kZxOiROl+vD
zp&kv;7uas;Gskcww|(O7IPrDfysOH?))5EM(NUR=^Yi*`(?L>Mighsisy0-BA?rkY
z>4m4tJ!&2Q{eAq0d$<JM3Qk!Z&RoNHW8c#s{@2lNPQ-*8Y5pan6+AAx7WSUB<4h_L
z$$va3v?N<0b$EvCjVm$Z`4T|GN5Lv)G;Xp$zVYc>HbAwf6yVE-|EWqiZ$3#XaA%`v
zyy$YfOZRx;0a2~jWGCVE44}?Fb$sfpF2g3T2f36TDcQwYGr99dL_IDyK|m%;KUHgn
z^EQZdLMh$@s}L&`w>EO`>Iu{zix6s6d7^FqWd|<Q0k9okrZtrIa>@QB#$)hQn#Imi
zyn~7}XXAmP^dR@wmbP{3bH`;n8vkoIm}un&BpUa?ZSKCmImVnW>s@o8&%XWQ@rIb#
z#{Ur;ZAK=~lN6qx7Vmcp-yvTz?9Qy%blDhuCY7?!A+4dOnxcv5PE(<O7~cg+9{)~f
zrm(Xn)@53_<Ls6`j`B52vc7si%-^6*lTE>Q7I$v_=3!rkgequ9o0#__i_<cU@CLxD
z=K9JQJ{q?hpXy({`VC=Px?e7C)(2`MSRLG#s|D8LnPPEHnT-$Djf`YmbF9Ti_JdnE
zKPcEFGP#msiy3l1{${N3JXGxx=@lE*>(H}Dq8~|M`rH1Jy}e?c`(ED^enaf^HoqKw
zgO?j?+;MsDl)XH{s7V*KEdo{al4-Zz3ZxqqGp-kYez^aAJnfEsTOqAM`4s>2VNa}+
z>PoJO$+$F~P<;YN^cIiEixfAxEBoJf7qrk8_kDuss~W$<=OxTCv`WlA#BTU<DL&*b
zM)#DtA*wG5Kd#ayn48L^oCWlk*p&d!%MLd4zZH9lC%~Qa%YHtvkQE~YGHwm-UQwUl
zgr9K<lwfkk%vxR!3I$FWxvX1gk&AT#1P~~7LD<Jo+e2%UP}@FE!<u?%ubbny$to1r
zeimhhJ~M;oiv!MVSeL}@a#{EY`ItQ(Q`U4_>Avvjg}ET^BAm9PmR|`I;5#Ls`KUvC
z41r{oWvKdgsJ<GyO|%{>2~~~f(1{WOs&cTyu_bOc@<RiHq$GA9<Fq31&f0VFhZ;wU
zOu$dPeXtF&1HdQ~#qgTW<0a(mCy5R?BeGi4ynv!&lf5eL870z%U5jG8c|2;HcQfgC
zIeP94@!~qsxY|{4Mq<@Br`6Wnu<6@}baz^9mTbFG>|{@ASJm!?-=accfADLU%>|tr
z2VJ2qPeUhKeXI%YuAZkv$=QT6;K37TiXPEH;X4g7j1orB-{o7MJa?a4ev2{41T%&B
ztBU|*8eE&=-&Stc5U}kuh%m4sWpm3KM7G=xs1^Lk@qBWfLh-nuiG0u_XZFad{2jOV
zhfS`zBZkfEW1*6@SzP3PpzDm?S?c9Yl4Y;08(dBF%oPnpDuT7NHp|g5mZ|y}+T+Z^
z-shodwXQ_R<NGyyp%>=b=?RnwLa=OZ;^vHsv$Gsqm+s^!@4be~@-BKK)np~8r5@-S
zg^U++i%se6tS7PO-QMC1?V5i=XH=+%2!Edcc?2B4U3P>I{u(52?#gQ<Pzo}&{NsD-
zmNl>E!fcOb?F7df+B7zYHXTm(xJ^j(5dk4^XLgE8k8q*OjKSek@1uv|>67ElI;C+6
zR-$o^D!Fj2>do5dQcfH7j7cXq;WXPJ%(r2BX>XJ7QuPd*9Nd%z!>*;uX_v2l`lEj;
zhD%75FOQpvH!(Nbj%A9|C*F7EvXV~IBfN<5)brF8Q8CdwkQW==>#1bxO+BvsN&xj$
zckJVz*Ha>S3Zupekx`TnY@#I}$_vc4Sl;_$Y5vZ5r6S;#%rvBGsR>@H0W~j^LOgI7
zKb#UG62_&$k&>LNrP`gpg;(t=iLU~p6<(TheVg;7o$ClIa~HHU`_UrO^gFg|oZPkB
zy1O1W62i*Sz4^Svw5=>@>{+}4&|^jPHggg9wbLf}`TR@YQ<^P#p5z_xM+eVxIsESz
zx-I`Pnagk95Xa(cwQqe}%qc-zI*fd3_2V_f9ZSt#ht&<}mT?ERta1pvs;GA@f$CT_
z6#HP=w4zEO>q`r@Et?BLXGINGP_X$qPsPZy+FWALS)AT{6RxxlBI!YzLay!d(9eDq
z)b=%C`MCK|sr#GLA9op-MH95~<5?z2o7*Exvf}yIA&fDaOmfO>;$QDeKFY}BZ`apf
z_*!eGkF?EzI*L<L<~(*7etPQUl67pq&3MV}Lv`h}_+Fuq#c}Xc{0>i!K`}n>p+~it
zDkC@Ua@Zl2$9CNtAB}jWW0wfdf00w~!Z6CjQ7vvvn+SM@wcw90;d`dMLaUf(KlVPP
z`WRD~=w|HA=tqD`vk9&G{;~8)Qf27NEeHJxfajDKES>U-s6S7$(SaB1Rh(9i=|h81
zCNQ3Mz9(E~nvM$akoyhfCc6$jC#)&DGY<rQ(S9FVC;48oIWC=>1eC9@L^@0B8~wo-
zlv#yjRNJkC1vo2eKdlB<PG{=JP`0mXceNfbHAIjc(iBwxl07KBcI3Sv&nw14ohu3J
zPyY4F<M?`Ap#DlvU_HBP0;_xW{i^QdKNVWQZ+)-Kk<E)E1lJNRvltG0rTzY;xQ;5&
zDctRk(6>*iZ)!;a?!DW>%(d%se0Nuqf>_}Y-h7GDsn_+czw*kOGSj>^eB*X!Yf4LN
zD%;*+Q!HN*_S<}p!-@PT2hgD)%c$kJTR#HhlnGAiYWz@?otP|Ho?3TUnQd=p>#Ee_
zf}*>dB*|ed3mZ_49x`GRkXz;zl}4lv?VD5h1Z3_`&KTcJz}raCzJo~l$=_)zL5KbV
z+Y*{BD+^(nN1c58n)EK+PMfzzne_40Hr<|`+ngqg;}6HRs~XY8sIRT4hsD1kLyJ}6
z9jrF_!5h=vsc2EO^Ts9b_}Q=xnq0bq!!RTAI{fx65d(@F@%#DW=~`>m_vDoOg<Wba
zU0t;~nzA26J+N}NIbIi!JNYv7@XXWfy40@wyaBS=Lu?1`DGQ%QI+U8f8R~69)U6dW
zIF(7?at+idrAe6$r^T%qZ6?IbKcV2gJa8N6)+u01%OKX)B(7Ch4i*Id^9j^&vimga
z=iIa5X$*<_D&ROI?-kBn{H4CP+}U&M<Zi#G;jaQgx}GQH>DxNWw+23TGM4}hKor@<
zno*&qHc24rdyVL?!%5kh(>uZ0j>iWQdFht&;LncVbxU%pN<3f+7Z9<V@PyU$^EmBm
zx=nTMUCl>7JcHLjo{vWD1#6fJ`(H^Li!hn^tB_q^MiJ{o33BZTIt?~BVhb-I18g<t
z!K=ok;E1TK9;qOI?MLmkwN|$>P;72eh6y*$-*3D$W~1nk>$<v2M>&6>mHaL9?zaAP
zVuORI;qaFkgCpc3-f_<7Fsv*OXepnlp_Q8s^ySs?q5UMZp+ZJ>`=5MgT-<E&{255{
z{&q9bDBG52(o!w3=>ksEm}Va8e_b?5+$8d|BJceaqOr|E^{2Sp)uEBaluX0rNjp&_
z8aS?D>hmb#r#p7mIg3{ApE(bRedP}l%XQwC)+ApX(i+iPKs^7NI@Chb9H~GT)0n(S
zGwf;RU=>!_Ctcg>8@X>X0_OGnvq~c8Y0|x+pA(ft<nq^%7>2k!o0!JD2IFav-p0+o
zY}uUPf^{)$Vt-S-lbia>PW3E9+b0nvl{p)nc4PHNUe?P-V<?=kNwr<I?Y!n~$P>~u
zN?-%aLL2m#EZcNtlEN}!**I9cvD-nbrv8Ul2q05OD`joj^nlq{d*0Z@QuRtJ##rXS
zv)Xyu3&gBUQJtlpVDk7jaf&pLq8R;4hP%gokvaC7$;=~lDsfZl%f?I+#mwU|u#8RN
zRAq184>z`6XA%2_!-lIg-^mUx_vEFq!SR9qow8LYkundU>SpZux+%bDw8tA<q?VOp
zcw4DiuerOEv-UF63ASE8(Kvl#;HBL<k3%sU2r(OUR68}88qp<4U><B7kT-X+er)#g
zlS~p9e3q0r-O-~%6u0euV5R;oh|2wj%o(JA&;jGIcv*ih-fVp>J8|K+1F1-3uy?{C
z7mLke9^kWsv+w|){i-WJz4E_K*IJfZ%rjo_@sOij*acR<=5Ir+L_5RHbr)2VX%l8<
z78_|7#p5Ul5m%NkGEjC(?}<+$3+<-hp4*Qtd3{%?8>l9=Y9abu=}ei<9xMq0u3Cvq
z<Q`I!w|+{!cXlg&i`kp?mNh8ubgcG2#65|EfQe8$v*Bc0svT?m#F8`-Zun;5wbid2
zWQ9epPq}~yFn<44MM>gx5LA7LnZlFnfcXiM{wrriGeDft6ji0aSt;Cs`P$T;xy}r6
zqLmJ}yH}%)EO1iwQ!F+aqizn9BkpVHiA>m!PY-p}d)?gPyK{uf{J@cK|0Xhv_B%uU
zd1K_xB3FS8qsb`NvpE*7DZ8&cD9V4|48Fw&NlIN3%l46jUjJm7zC?i^b11zp9(kEt
z5ibE|3(xJuxA8QO-@Z*hY4xTdOsHa-L+go5b=jYgA1Ua?k<~)2LPKVbUt(E<s!o9$
zHzF%z#@1_oW1_#^;rcwyg>Fjspa}V~%KC}!USsYXe=VqMK1D<4(Cu$Cw|h}7rnv|i
ze*<W>JN=+f*DX1)*aPC}y+WiD>Y0R{i@yjJv)D$YZxz#G+9QROCz(H%ce@T0{?t$X
zYB6*$W5p%p{+^u)ieSC(t&49U2@dXz^RElz;$+i_&hLFSy6;u%z@I{9Dq_9wHWcrv
zIVxz%D3L_FP$zN;jlw_*-M9_g22%i4BH>-4!hoTxlbe7qY~n~vQ03*Suw9Ccsa>Z$
z^Z6`tEW*qerO$3s$e;NM8E=3!a<5s^@C<2Xo6euvvjm|*4>m7Z!z5hosOU;-3~wXm
zUya6|qv0;zW_P@}YT5EJOySf}&+85I4({U`ZrjgB=ve!6_s9nVd>y%dHW;ka<Xt7}
z8yL(tO))Z#ZcF2)vzimr7DgND3&3)(reMa=vZ+HWNFRi#RJzWI;kZXruLk87@4l?l
zM0-vm&6TaU)rq=x?rQT3F9G0G0xe$-Eo;KMg`7xg7fAeCneVqPx0+)?fB@%`ROeIN
zxG;EcQ=;6PPQ?5<<Q^u!kVcM2@@MIEc)*R<9$^(j+j=7j9zCe2SH4iT9uM#%(F?zd
z0`WHQd-t-l#@2%7m^~|iIi(!smYDd487^5sj`m}SC)|EZ|KibF6;|6#RCds>t696=
z7ClZUl{t$N)DqQ}iLg^p8C^We*)uNt9_@)YvgL7ve=$-o@fNY){0jA?_)EssUD@<m
zt2XoSeNqxh!}Z)x@=u$S`bGQae@vS@ShwxnL@#9y2G6dcA2Ds8)g%#IfBhvJ+EiDJ
z!ejp#n8*Ay@R0NLsF=oC`7>sK6Z<9NM-kU?w3`(+wh(G)Os&<}gQ9CA=Nf}QI;+i2
zsl16^TXT{PN&Yfz1dg)uBt8G(3{l$qvPlJyBfujc25p3yapg%+-xr1|$RN%*lxbvC
zZ}pTSWQH?QPZjizE#}bB4kZncRZrDQ^8P1_#nJ*BrouB;0X#ngeC`g65g2r+bv;R`
z?Tag|tRoh%wx3-8h|>yP4ZXkYg$(`2>7R;6XljCxh~vHogqAY-bKO`38)sd5tDQVa
z;CxojLLpk8Qs8i*F*nTtJm%Vy^!GICOhlCo+K7mq?q6t$4?h0WxOnu>S##0A;lc)7
zVmNB)as>WQw$|IiT+?f`xsMw}beFixg0|oPoCaz)!&Sav)djuAtuk6FG<v8qtV~B>
zEsr3h^71knN-|5SUKzaI#c98W)otLCw5&!0EV7#6*5-VK@nPH51gGG0n$LU`-U>Dh
zRxIN>C06Fsf%EZpebW%1mENJE6W*%ILFldDK4Il!UnHge3m^Hvv~{x}D_rx_73J?E
z^J~}>+J{p-iD8FNx_05x*BYkT!@rHguFgs>XbP|Kzx~gyNcw7#>SwJE8q#zs(~+{N
z;~ZhKdg$|HtH}xI728V_EEQo+qqm}-qM)VQyxfT!Z;pDLUcWl}rJsFRQC)E48Rn8%
zP7oI&G<+dN2u{1LJ$jU&lVG4va&VHvpf>4IL<<xs-dubN1{F6RR@IXm19}_#(anB;
z0LDzbX{}n`44jm7pX{ubIu@?9`2TV*W{|wZvP63>SzSlnP;n=o^^FQfUtKiF{1)n=
zRI28z$`gdD7F-iJk|C8D<D@~KLJy&YJX8>W(1dz$F}+b1=W9>2T^VzOKwvYbBo{(4
z1JPM)NEM7L7^?z2vDcDV?1vgz+n1r-;yguHg;j-tzzc^&yRE3s<^Swy{`alDYBP4@
za2Jfe^u$%RO$r-@9g129JgHw_1f}GZa+|M6`t5*ty}fAo3q@8wPAM->$<-4y(^K`b
z4}-n0Qd*W8o3D58k^uThf&-mq_wIrA?y>((gpFa<>(YnXBi{}y#GiB^2Zu|!R+?qp
zLH15ZdX~O=R5(h0&JXkX&`+4PnbA$igvj&ps!ic2BY?aFY~>=Zk;>Yw!`r(s2dOP=
z3<L=?{Zi?3h|1;1|MxfYAMVL+X$qeYELcJPC2Qj)a~L8Wj^Ut!(1_3fJRpN`p0580
z=4=H`z@U0Z{@2<8dCplP=G158m!Y{IMYaKje35q8?xf`V4%#P#n$UNsD5Vl;Bf;v6
ztptGTwI^2B+-Gr<BYWXsIePMtKY@^yj#MA!P!iQEK9@UtLSFgr#b5JXd^48X+7WkI
zLH6bg=>rAOl?$J`o6kX9a83fR^jS|;u-}N>Vt%k|S+)gM3DeJoMtI`Rv-qE8-Zqy}
z25&pv>@Z7LA5Y2l$s!8+!aj*CG*+JTL;b9<b;dq{;>G7I5MJ%DhZO+|RL(Kx<xFRV
z0-bt^8!8gAcM{H}orw0xA6wKnc=kM(qR>_k(dOQz<1LD#%(^L6X|)b><s(x8IEWQp
z<ox#QT3LahwC6QN$OhGfPW4-`!g?LmB<dq&OlaI~ib56NTOA&|i(NH3&l`>{z3W`n
zT6$gUh!XPb`n%2W>wVvvOUdM_jmo=BT4ZC3Z44${eEe^cUl_A@&&yQwbOYNhj7KIz
z(9Of`Q;ZkRC;QDMYH+a)9a^k`Le>>iQc;jwCrd%a_Ia*GlS-Cg`SKgR9_Gi_4u8(y
zzD|XhpJ<;|c;cF;+;rPi;az<sLAS6E81<}n@XpBGC^RBht2<__U?WjV%X4mw`Wt;$
zTrs~@dqfmv^jX%6vir7#oYeEt$Qe)Q9T~+eox&7KG%&|9-g)gPX`Itu1%7KWON5kT
zwy!zg!F$C`o6iJTatf5-o3)>_?vY^fMdjVtR{t82QKIR*k&v-C0-$_Wd#N$G;aN)J
z&6vNJeR|;0I;j|S@*9CH_(`sCEh?nlUi5Cyg{C+6oeZ0MEV#XFhv9SBJJXhiICBwk
zKS+Akh7&MV#Uha>r^y00d>49P@y$kSzNb9t`;-8_-+f49C_RkUwh<DDN8g{4@_RgW
zoF~MT6qwj-Myu4Zp<VagGK?Ih?bPcWSPDs>%4%}|+&HXw5qp3U%2%sm2-5gVhP}Qf
z(3ufg?2)!8Md;Rw)*0|!+1HGNP?21qHni>w{>}-7&1jsae-vT`2#PlT@gb019`F1m
zTgmskzIK~frwBH-2oKf%P9jY?k&%p>k`$HS48M0698E<$bjB0bga(U_=mq`fqj*$D
z03m}9`4OyoJr?O9gI{s%4=cJKCUEC4#wE2css6fFY1_S-RViLm4B@={mu&H)U;aA-
zuaV_XBE;MZ)gFeVjd5FNj~FMtM}DF$kf!u78+IKQXg}Nl5BRk(1-0Mnkr&sgb2NP5
zvQ)rk%)PtVdC!xcQPpT`pIst4^LkdRNXB)lyiS5kZ;y6u&8LW+$6=59%3OcRXNrFf
z+|&ZZ<$Ycz8T!MV`7{_4T4>OE!(yo;tQK!sMXMyb?$bI%bB&iOt#OM$Yw^eXVx09O
zPF)L|3UFHPXC0+S>P!}^`$QQXx=|EO-7_@}u`gVMScBg0!I7DZ?EMBvn_<bu9Jc6L
z&yJGvBuBT8@_{Tebp7ON8Wr{GN@*OthUIQNH#QS_-jRn`%JsOd3&~rJ=D9DlC*d~T
zKCJj1c&(W|$}PMJs0Di;SG6eoX6#zByI0nAbp3us`etbi0K;h%lvB!Xb0CbFI&w(0
ziV~OzkhqlccM>xX#@hG>^Nu~Kst|WQjGX}u@gr7*ihN@8&+7ML>R(Tc<wmj~pu7)Y
zuZl&Mu+MXa*6v5q<&4Y`38IzebhBzN^A=opKpCo5TI|o^#bpyr?5q~`*)`olTgQ=2
zC8mgy5J_88S@rTtTft?w4(mcL^JtiIW_%RR;~>}A;PCko-%Yg5vo;d4oh02}RV498
zuS(bg{qW!~84Egb6qcPzp!KDFM68>Rqz@(kp5Baedz`GadpRYz)sL4v*IDDf80#Pl
zMWMa_Q2WbXC7I#=k_qg;W~KX18(#0(4!x8;j`~aX^LptQg@vaIg44l&$=qb4-<&Hi
zzuE2eVATp`4`xpyIv?(+u@tFUbltFU-7-}jA?79^;#4i!;=b5^x3kYKJ-Veyd}Gjk
zMs{Z7f&36_wkN)@X`Y)Se|KQ^`^4f*{XU}?e{_|z4((@^{H$;T-kwzQ?0RQoHm8<#
zfH%7m{q5-ow<)e%_i6cwcnJX;!LMuMfF`HUycDkWZY%<dWaZ(4YfTR0B{y9yKl>1r
z7~P%^l|D!(ShVYew2#6MH{U#Z<vIHMYaoO9mwn_TEuU;zQrKD{Dv_tN?}$bQg!3x2
zbG`)KF<5urU<$Zl8i+_D6psteetMy}(&0!kFW0+6vap{Iu8X_XBtHzX+F|F(8y39k
z7U*`W7wLL+MyFxrx8xxWxvIM=f~1nE(?RWFktBc1%Zlauy3$9`U!Sz;L*h2x&8{IU
zwR7{s9<R3~P86_hDMX2Gvt45dvnCtFQMQU5D8#INPo^)0?2Ga*J!caYyq2S($wvB$
zUP=J)_HZ)U?cO*2p=;ZnxqI)segARs=1mA65_mFdW*QAvX)le#4GzPf>nP&!tL-T;
zfu7J*5zlNR@kFVQDv?I>K1W26BiJX&g|u?K%uKDI-8pmPbdQlC*8)@1mZE{Mi#m4)
zfO?8P=jn{C(9u8CAbO4oLDrhB<y8Go$oWTE9hgM>=>WwQS)WedDqYk=gXPWM+F%|O
z9>hl<%==SlOTgqpbw%IND^7;tPbP9l@M5d<<Bq!Oj3%rlurZOh(dD^>b5sEhBHi<4
zr^rQ=V@{HPmJTknx?{uYS=u}#j4o~c&zYxrzVze6DqhpBQdt*6I?En*nr;!^RUM?4
zfz$;3@)rf;D@Wh0ge$R-7j)@mWA=G(;r2WQTD43`rBYf?g2F1XXY7yN5;*%dr0GVz
z)K%Vx%AP)a@oO_cc6To+0_vbcBnV9%$P`Q9Lut*Bp|Bve%&X#QexkjKP9@25%X1c*
zcvXRV-yLH6&ia^%yCG2>E5qW=Zuci9dl*qaBeqO~f0r_8eIJb0er}j<R}H3nlt6-J
zW>b4gI!9INqI#A-ios7qEE{}e6L^*<tNfbeX_ac05d3T0c0vnwV13AWtzAG}+~-#=
zf64THBV<UgTN&a9bJJOU_~dnz;$d6NUouXV;SH;^XcsfN7ToVo62{6+dzY>2ME*gM
zOM}J(a@8=rvC5erd*#nUpFW7Uu~A4;mI%eVR0d9^yTRLiCo(9=1L<en<kYA-HCUGS
zwQ<yeI!*DPku368#u0oRWsUTd$swN`gy>7-bSbDba&&Q7=tD>fgq71t{GOpj->+gw
zao-PHd-ch;3QZxJe%DGYfX5N#=D@KyMxYl)P&;&o6*~O&Az}Th$VL$0$C};o{vf`y
zA>oa*jtZtQ_0ZO8sh;*oKo_1NB47FR%)?z4JLTyx)zXoZf#&P)0dFZMAChb1r-QXE
zkdT`h4CzOa<ff*Kpy~;mP6KycQ0IM?k~GO)n*Q{Fq~2GKM(=;mMVCIz&A13xmSk)Q
zZJe?>$XJDCHwGua=wP2rJzwGG9IhOOOaiCC`sqb#O3@mg9GfM>9L)>-vO@>*EMRjz
z|FcRTDAO#7NG4o0MIjsxE;^t149b_bg!@U^acmWFb%=%l?rVMR5>;6QNkX)Jk40B$
z**zb}_U2YM$=Z(5oVq&XTwIpQC0(+MiC!MCU&9J!f4f8pc>E=Mk!SRmtbp+L+9AiH
z`yX?Qmx@E-eysb93fK0=7Tn`A9e%Y(A^Gdi#|0(PsAww6;dQ#aoz2LHtFfc0mtJP?
zA%Mzeuu@k1?dj(?*!F6c-V}$hh=o5y)wa)!UH){YLRX7<t1nFlwqH99*nHjhbkH07
z;n9{=aogURQHD?o0}~G&#Il*V0X9#d;8IRr`p_Y~*K?pAVkQF3rLvRyZV)pf`YhVt
z2P!$-j;!o})^Nvu4Mbub5{YzDOflpldnhoszEIHmK4a+GUXGTwFcZ6fY}NxeUZ^&X
zS$O&{8I^!)N`@^1Ca?d8=Q;m@2U_L?essLGw7(yGH_IpnlDX(vPKdiv*t}x&#q5>z
zZxzQcOy1H0x7AqQWgycBxT`MRB(P?G4lwXHf5`%_eF`B=Y<-=a`+c{Z^?>5hUHb#-
za1^9r@NjcbarxHt3vV>BSJ4$s^_NT;!F0vyEJT5j>$VbLWu2dkB8U!cist>?L>dSd
zT4c;n0(f@QkMbh=L;IrCmP1?hh5(_t(ZUW;@zNN(Hh*1*Ze7X$hqt$kYO@X31*uV?
z&<`(K+@-j~M{x-hCj==LNFYGaq7~dFcyI{>m*Cdot_d#19f}rd>9FU_IzRTFwa=Ng
zW+uOqy!r92`+f4P=f3VsP7Zf>$;+M)QwsOT7d9U%?{bVa<P>|6co(z~LL7%<^OWTu
z$XdCK-7vP~YYXjU4LfO04PR3gfzE9bz~+yH6A~5e<`wgq4ul>E)59V!TRc1N()Hy|
z#Xxw<?5Uss#7mw2*~Elixgqdw1V7)m7XLrLYEO@M-+$_qxhwr3k+t^ymQ`SEIq>xz
zxgCQTO_1{?HFjl1xZ^A4U)HGqa%1=pmh`A`+_2^q8Yfqi(2Ceyl<{=3I<QkC3d~&+
zmtMqJF~D8nG@L1PRyah)_sG+~f1{gA(2V(6&JS;;?vU7L)2LsEmZh$x^!;dZ6wiKA
z_wC|IS~vi08>4{qW&<lmSsr<?zmUFmDAPdi>IiqKh-bXKE~ZsE7SQ7I;qB@y#mlM5
z!I}zgGYr(uQR=04<kSWiQPt;ui^R_@F8Tc1o4ER;0NqjAGp~HBgG8R@M8O~G_@>M6
zE;$jkVQuuHhDS2)(*jH;?;Fc(_zVXC+>^SioOLSD{&n7sMudEQ5=q!3Sef?vr|*;&
z_pITD*CV^O-zoaM!ahT9YMfAPg<A=y#UvMWKm%(Wg^x2)XJ<FM8g<gypO%iJGQvux
zkwx=Kjdt6thbra#m+}c`hs7MA97FAWN}H(5pe=%J-~T-8%IzsH>MyHosv8n=^d)=8
zsu?zP3IeyY4!4BTShm+Fc$CDXP>cCs@iph!8h}n3j?K)Kl)^66rDpn5H`49BA@zD^
z9}I?SAnE};61pI$!o>L6%XzUlEfO+SlVpfUvFLx~ygrtIm!nqdmgfbYwY<nIt~xD%
zYN~i}Qp?zhBK28T2MRG<mpeSr>_{J+mEfbG%`<(H{5jb<d5s}6Nf^*Yj5PE>X29t1
zmD1|bvRTlQk2Nh^q98IQU$vinu+t)Aw|}FH2@E)M|I;1GK32#%(~tYGUjLM{CuHL@
zUXRZvx*B7lT&H5Vr37_^iG9BhcinMrho>b*d7=Q>ZY_))a81#ZKDOyyMsQFuR!xZS
zUlE!RqVpQS?V(Aj3(%ZY<Fo1LNG&a3G2rW{M>!POYpZ0L7MAp(i$rZe4KXsEn8aGg
zyhmYEp-4jgKMVgTO_qJp=7JLJJ75@dPv#5;*UjRKMA+!?Rh&xA&(*)OJ+D5r>5Zp<
zY-eoS=)_%O3qz~lhbz>uzB2qn+MyCKexI}Apmw4FJbB{^YdzF3o;e>%H$>*|rZv_V
zPh1eB8Xir3;`&2bQgI=y<S`|D0?_MrySFZV)$6Ze>*SdElt%IUXq`0;(Qg}zfexol
zJp>q;q$dnQLDtf;n5=yJyxmJ<=lIq_5|*YDhLY>Z>$L=fspBTM6~Mab?~(@8HNd;M
zTsm%oOd_7Cr$c((MUQl`Z|y5e6IbnQ*QQ*H!%FUx)cAm?<~%7S10`bpZqM*6#_Z}N
zz=1C%iYcuHyfhGY8eS0rO^pq5>%(;|#0C%l&+`0tZzhP?B{Q>jMmv8Al~7wCw7Y5#
zgtSfTdj)+kefl@PB#k98x@#zF5^I{Ep`(TyjF--U-G%fu)zi09{7HR*{w5miKz9f4
zW99C(O^Q@R^rlzi2q1nAD@1s$p=nqZ4iVo9A4WLQ1VIIX?iwi~_=`STtrIi99M)Gt
z#E<{t5&WQ&+y#Ik8mN!t%J~aSzhuk0i!H%Myy7WH%z(9#3UtedPC+1Hk5x-!v=9<A
zsIR^SV2I6=kHb@{q}J@>{nqH<q{02YCMuM05SsIWt8_pBRLT=@Q>W`8_cE06J@1}P
zcU6dUxBpcotv4aEdc`B@90`ZKZMb6GN{h*H*_&9*mmU(-oMFwjPC<DpBrgLuybvTk
z`-e4YhV`c?ZGT=-#gOosX+Tam$GuL21pD8W^FLjC6zed`m;NQ=SVeq+ww%5zM^|fR
zL~(L+Q{a@K0Y($3ow73308Qmn=7{(o3`3esb;GsEBx+rS>CE*T2~p3vbL+vAy}ocz
z3b(;M;;VY{LiEoPjYht9DZlQ-TM?7<y_%g434v(nmaQSAjm&4hxB;N2tv_UjS)mc(
zIlTBk8HNk@3`2{FBEgKZ@HUDA+oit9y9)cA{Ssa!J#85Qp?j<Oe^nAj!twAR;G*MF
z6=5UOq`dT^?RG9#4vu37x2!D#lwhT-kw9Ls%P<+=d@f9iES!}1MQ0nJTw4oavY=e(
zBo@*h6u@^i@S{*x@T28Tsk>6N7idC+E=huH7a1d`HxcgXCX8&tO%(27%<7UxK6-3@
zu<IG*tPPDWHf<u-5}NVTUqITqkIv1s$A7_vM<zMw?;x)%4$_*mO9oI_EoJE`1G<Wy
z(rF(sw8bp7C~cw4m$~nB-kyqr+gm5${alqUcSUiT705u#(k$pe?WcJ>`MDdj8q%pW
zCr>ynA*{bn#bTlP4vRbUyxX4eFl(vUZ9w(r<~#TV7^a|C&5K5t4T)f-5~AE|R#tqV
zqLQ?T^>i?sD?PmveQvaa&?D|QVkCcv%`!4T)2BRUs86e)7HCOviEK$xcno{T+h5E@
z{tW-=`Ty6C&vZu4RU#iB3~?gE9TLUGTU)=*D@#2G(#uYC7y7sz8TN@T<-%b|0kSFO
z;Bu(8N<w1()(7E$Uxwmw)(hc*^%C>9PCf`r*@ULA-~j1n-GI9i-fmK}_=yCyCnO*<
z_yySBtsq9(=7ZQniXgbJVt*u}GEPg9r3(GH3_%N|;q<`Yr6!=2O{1}?fNc0P9c=i0
z({lN*H~Qaw^E`5)=<aFS9mwK0Qx3jA-UCN)6N<A)JXtNv&gkM}1znh)v5f?uikCnt
zxiAI?Urdqh^lQL|q{h;POxK}R7A9D7qSe(kB?1haPVm>;Qf%C5q{q&9zy56RJA?^D
zmXCFo9KpUF#b1#uRyITz5_r!hi<>jt%fi>@rnJzhQ}*!RA*a@pWB4&NG*W?bkBR8g
z-=sR!ubo@G#=i7=@r8Op$D_!7#+x2v0UD0y46{!@BwS20E`d4mvNAra?5WxznpIB#
zpu=Bme2}4!aJfB`@iRJix|LlpToTspqbTan?tuwr5~2p`uQM#?G*#i-g2Vl(4zspV
zVz?JJpw~K;Tw7RRAG~2PR3s}N?^}~+BAoKazj-)bLhRxDHOel@;QQ=I<>d=WO}{`F
zDW7hV_8KiUR9fDGX6Zan02S|;WXF;Cp0$b|X)z?z8z`xe?_Ygnq2lGAJO8N>_H@V+
zYyv4hF9S`^_3AL0vJW@6P}9u5kliinFD_!;8*270hT8gF9lt%ZfiS<f%fZ9@FBS`#
z9m4*dOzRPb<6PzEPA|oNf$AaNMDYB#Z}i3Oiq9!sEE|fcrCWe8ebFE>9&mEZHc6(;
z<^KH8NplJM6k6hZSiulF!fYjVvn=+EaLQzlFSexX@mMjN`qUC^0s!bcu$}-LxHoVb
zENAXHF-n*6l{i8Va~e~p|H&*Iss@Q}c&}Iz3-aUvh}z{2d)5cPu}J+PYDxCcDE0`e
z@)!Ka=EoC`bu^EOD&|;<l1Y@5JMR_)m_kIxs7FMF6H(n4y{-+m>r@o6d|0>-uJ+RM
zD5Fl2qn$dGoHCO^H3p>6mYN<*Rkd+XW+*ft^G*<Y);ncLZm45g`dvY#+`jiF%MC&-
z{jbpcX=|&HIG6faKU#i62CT|ra+B=~+@YJ2X^5jg!7##rsFYxGy*YYRCeQsI9*fGa
zb0tlHf<|0vhj$%$x#)szc_hA^S!KL+yCHPxyUK((ar~K1hVt=;?pCaC>UIa19&qak
zysofLRrW>XhD5Dfr6m)RQhrZu0@8AoMG^&CKzPG}PBPT0kjjT|gzqW4F9D1Q%+lzt
zC#>uiR@Tun=-zn?u$yWsp*G-V%fKaTgnjQj-7ot@IZ)*QN-1}3BsPe*7R3O0d;eWS
znSE=CrhM}<v+vc2+P`bW4MR&gj%Vs3l&TnK9TuZsYgm9WLIH^OM#%wS)XVfwl%hXe
z)9YwNSsNycjI(YUlDH0)zKm$A?Tu-#K<`^O5u^-9xRhY&Ocwumk@)NMlyRKXIG1gT
zq_c{yF;|=5%wR1=nPG!9IkBuUhK(_)I;Ixt)g6|AZs+%ly`|8Cd!L<69dxH+>mvW<
ztJOcZ`wm4v6|_Gan_J!k@UP$kfAM$@U;XKG&0%{w;uJ7>S$_9B{n1~%x0ZMt$HTUF
z3w$|$@p8>8@1bnoXAf4gss7^4C$Gueg&;rw#Y++UCJSn}bGzKFe{g%a^cQbP`rzsK
zs}cKv$%@9i8>IhVymTy{tQCl2NekoX<t_<oS$ZC{8Dg6w7PV#gk3}^2PRcISxN({)
zX+15!4hp*xkAWB^HU$AkC5S18=2(QkC62Tt4L(>g2W-uHnk)zRw|8AuH&IE&_`We=
z*hMM&c#x|7*kXTBS2RXvSTRy)Q(M)`W2y%;?wZ_bqC*lVr&Q0&?l>t}eqYXl+-)~j
z9=C}fNs*Z6Iw^RiPxvA%#uNnAOPZ&ITGQgUBw*DA8{S=CR>>=87iakkiW~2EwcvR<
z6BNQ8PiJF|E(uZdg<WoptL?w*C^84;%eCgJBlllNOR+_ZgR`ftBt6nANa-jjA>m8w
zS6gP(Qc76wP2IugY%|zYKR-6?(SG33T`48pfLlXM0po8R>}9^V1@7-VGA~Tdj~3+a
zH?XQX<i)naX4W%OV^xAvKFmE%-Pq?H^B@d$6PH6e&eJr@5e?~zhBnSpUyexY&#ulY
zg`}J8@`k)ijW}vxfU+LhYNz0IIz-wLYDp;1%eTyGuzYSi<|fARH3x4{$&v~+-5+CU
z!7z0f_cRRN)_cCKa;~8@k7<vMvi2yyo(tX`Ww>`-)R<Tm-B@T1S(7*F%KijN+#;PV
zik1cGJlcbGAwxw=T=rKP0T{sRk&{+yPus%lJ3BEw7TmYNw^vd{Sn7ckfL4lyEkWa{
zH1|LKuikXN_R+*=F<otJ1y%@9cj`7*{h<4cCv4Pv3|qkUCbgPbVhO^19+YY-PDht6
zSyDstG1pJbY?3+$UAX%ji5{r+A8@j&y3zD+vfWAPW)u$QR_qDEO%v(7X0OY%p3P_G
z*QTHo>c9gnH>?y&ITF*0Tbq~l$_EA#>@?LXK<@&>t@Xl18`m?vVt*&m^+}T$r^a!=
z6_4u!2cO%Y*h0q?r?a0tc5umKTN-B?c(;Rto~1UaKs!fj7AsIiu~Gr6tpUH>lu^1D
zTjZK*Zps+}#j9_USwG+MAC@YXUE-?fo-ed5ZUW7A3@0$#sue{gz){ViXvxVJQR<Tl
zf>Fa65gRvYFTw!l#!${0th;C@%{zrPT#dyX661WHIj((O8i98k>=45(BspRBp_4nG
zcHlZD1&Ys&&!iv3r9T`>T91yT;`UI5u-#7mGr`%Di%D-PS5u>k^AizWDM`5OYnFL3
zy>II3pLp1|zm>dKn9Q_aQi#dsSGypJ<V%YYKQA8RLirUH)0Tm1ESrl-Xhqtnz?123
zZ*VT@>k|!H{62-{lFqq4Dy8#K6Opf$bC?!hQ*%m^K4xduhHN!kl6nMD|LN>W{`=AB
zm#se$ADxWFz@0J8FUISKWudhHX~1CUM1Gz`#s6H3E_L&dyM_<<9X(Q~pc$PtXd{m)
zu3M;hjz8G!=n-H4TdiuvM>yF~`Cip&uj8)o*R?UyqY#R!`<c0d?~tDofBM4IwS@~&
z46?)dY?m?BrtQX?^L0HL<PfBoqMNQc(93;+<cp@-80~YDL<bF3e)G(m?w0(0PZ&h%
zO-L>g^cDYVX&`j0o>>Hcl`*KW%pn%OJjTZkN$wmd+p%UEAh#P?lOI@n4d_<vbH^v+
zdehSc_e{Y&jo75sp{G<w{64k4dako5JLYGIA|Dl$$JtL<hdqXX3Mf)WA*kVLW4b9p
zhN6N*`m2r=bmxRaapZm`vj&N;dnHUq5+j<tetkJ7JQMKatu?2mp{^ekX%9ZIgD+IK
z=U@*Nde|DU^SgJil=QeL5+Td~b|Mw3Pkq9~n)q}QzJ1DPirE)ViPErbN)$BxC|+>}
zdefj*X3>WK`O{FuINkJhs`mWb*~)7<mruXR=MNqV{$6-`cz-zkBU=6GnzZ5v)7kL9
z0sWcR0jKi{za<7f`&0dk=N7w_&?vtQSDXj1>E(G;qnx@N`8T`+5c*|nisr>uUK96W
z7n7tl2^*e_3(Kg4<lAQ{AO7NH9&QQ)boyN@h>Rc}=5Iu0zMQ;Re)EiQ9JI$5Qv!aI
zf&=*C&G5d$4svtM-ocFIT`Fa&2vvL7k~}u)aCBu5jzGxsLB|#n3}>`b>acf;NIm0B
z;Ib3{8}sabPdTEN5D^`n6)P6@2F4fW*&;maXyS4kV7`i_jdP!7zQ*ymm>yjpbHSQe
zg7DgH-GkoxlV73D?cAW2v~uc3JrM$v3-cM8Z^E})?l<X6H|boz*vJiu3ng_zQ3gx-
z<b~EYYuqIZ2s^lMyeZIhUK7*LcVIX43ca0nvg+a0Ib*3m?^Qy~iClE5wK)|DG$+?%
zM6&Jh9h82+S*1gl7VuGyvr`A=kB#fOzaB(bD0?;~mPjO%yXDUg6B+!T4Sjjf{3Bj`
zB1dW_&_mxiDP!FCc1d<oc-}a6Fewl^&5U>Ary@A85Hq7p!I}A?ZWhIkzp17#E@8-0
zOmW6MIs`AC3N#WYxkeTK9GM7)vAZ_|GxqtG>?`?dWVi+}gD~#-?83o;1B65Ts^fZ-
z|FTq+&6V^%TJ+jLFCgrs1xm2*J3Yg6S+o4u(Vm}3!W_!<hFIod-8|7arszE;P|>Za
z&ebDO76;TfxIt#=jW6^ym3`ZTx3V)16j%G{niYD$P}%~&OHRj_p+c{(dy0#=-EAHu
z{O-4SktvUD8|YcHE=g?*CJM9H=V=a=XIWJAWb)7e3UWbaVZHJS)Ou`4M<IX!G9x}b
zB|3qLTZdgou@K4EM>5E45LEl0rz9TNq1gvB-K4+D$DLQf`FKY{Va^8+R@iUokL>O=
zz~5x${9b>86@|XK0mFXXS&x6dF}%(%cU|+Y%gnUCO<Zlu>WEaHl1`YTNvtTD%#_3a
zPM}CNHNb2-6&Z?TSv5I`=r{X1>;KmLK1(i0NuST>z?R-f{<^R*@WJ31{;`N{={!{_
zb+$^6+&y!}C0(xsW?d@)Qz<U$JrpD)8|0*|j+JP<k`eaPdO7@x_X~KaA=-M!b1h+r
zt~4O;>l{TrB!*_GcYoxui|MlAK%KRU@R=5GgGMC~MI}W$G2FpEo&qP^0R!!mV?=XR
zuyo3ms3G|E$7=L-*}va%=K0w&1_g$Un5~m9JspE6uagM#_L6^mthPU@UB^)R?}}!>
z;l_>tN7krUaZ{RYQ@^MFnaVKmR7TAQtnd%PUhBNa6oG+OiB0acF)rM8(&L_`Th%`!
zGy1-6HOx-ohA_Nx*uxj2ye-lB-h(&bI_gr&H2Zz~t<|28vO+}izCBPWrh&Bs*JJq$
zNrWf<{6uC-melu0{bESf2Z_tv-7R2l+H}+kZa02?3@j3iu$d*YHuC}j6BpMLQ=VB(
zX|Ea$f8k~-)Mc8JvKnq!{v5HJdv5I<=9CSW|6$agIlG%hihyhh#liIaXB~`VtYuMM
z%;oVx5YfD;!dn$I#f3FLNS;INc__6erW5MJOgo^wV*bN#^okL)uxNJW&WL60F7otw
zw@VTh72YdoUny9_d{4?)D9`PF6O^2o<{6K7YlwGUb6xMxj5YbQbtq@fD&O=;t){J(
z0-<})_c2f~D;B3H-s7`KFyM28h)*0ko-8T#Yzt7UHE_9`%xNh}@EEFUzHPY=(t1M|
z^z-+;s7D>)idz4<>lR<xrcY(`Ub0M4ldNy--b?9v1WlzGn*!xSbjM(6Ew$K5=NEjN
zK_mFR)ny$I*L1UzW5~v1-|GS*u^U)V_PusnXhE4Wbs9F~q*t>TIB^6wjPQ%pb#ojb
zCkyavh*3d&C<74{mKAf%PMB)M#+eQDJDN8Y6}s}gW?-cp7T=6#++!ymCU>>Y3s4Dx
z#qvZGYSrszhX*e^1WV8H)Ij%5V~Gzg+t7TuYzusVuuy$<|AhC}|B*ZJ|Ae#psDfY6
z;#PhB&F$V1!5!2BuQTgj4pbmedH!vp<A-b&-uBS2R={k)PRDrX@!d!JZ_iiG?8Rk9
z)jFqCPfH@uhQyd-9}|cNVkl;=JY5?4OU<K?1dy?ra-Ldgn_`^AMJw9_Xb73v&tRba
z<ek<2R#Lu`)Ldj=ZiY|@yQI=7f>7hGx;liY<U%BPW%Z(M$}Myy*edxudY~H<v$ag4
zbWxX;@v?>f!@r@(pyqwer&wg@-6@=&0Vy|Ow)!Ept(`n#AE2e;U=*=37HH?8{BhJ7
z@uT*~oL0FyLmMx}$JQQ69W{WU!`|~gK&^*m<B$IiGx`7I+W)>P{Q+$&Ks>~<Rq~v?
zO&wDlo4QA2;kYGGctwgBe45<==+;MHUTWoXB}O^ZSn6aqxpHYKR07H?ApJ+iOd1cK
zr;FKcg*DiQ861hd!tx4CMv3u8?jWCZx=w~hdEdqWV1m20@A~qsDMzoTY@ckU7zrLZ
za#FyXkx!VL3WZq5(}><sFSXOOapTlXm4u7vCJ8c`l256yhUJ_;rnk;TJkH7mr4s-K
zj%gkS9%OLls$KJHiQ;5lDF)mNuF@r+e-$8b!8Kra1L&+xZZ$vJTGBb+m}Zhu3)?y>
z>|Ua2#HwnQb-H)w?>A$=wIV?028B0Aq`9R<IeB?$t~W6RPTTojtsQsvNlxt|PYOCc
zw&jJmpXgL_c56vaZl>kVIeiOEtJT(%Y~hLq5*Z|NMb4Z5sQbJ)W0_I+E!E{Kydk%F
z*S%Fl#K8;XG`N<5n0yiaZpySEuP<YYv!O2Ewr=3-Up(Uh&4}T-_jcd?qWksc1B3=N
z4mer<k+07@JSkq2$(zNo%{#a_J{o@j#O}qcyKboTIk^2!T9BAT?+MhA^HwVcq|izB
z&+zIRXzc=dUCZqhHD?9hQ?#@927bELZl9*hISK@%<NT+)FnK)~Lq}B|sVuNY`~aAR
z(lM(-JZj?{k(xr`|1$MZ_#j+E+)$=pa}bi#*zmL@z@Fcp+H1EC7d4|*i8+1Q44I30
z8q=T`M?r4pEE1Bgr4h{((bVWMGMq$>>NI=OHZYwL)b9`bZhJla^xqcgWp}f+gMbb<
z@!u5i#RO%Vl)8AmK9T-V`D`nXgqT4dk#;EPCxVsOPKiBQL(`kLW-l4T2_<qB_3r72
zP~L~sw$<4uQ_7=}Uy-T(^<T=plH$1DF0oBfth13-Dpt>hl$@m>cDtRSdVMxz*KAco
z@oW6LRrVAX_CpS?7Js%4xI`;2Da!{sIK|jB5xg(=+jbRM7QQWcm+^_3yC$#isM^79
zt#DUF`+E66EQNNc%rEY@3!S*ZF{N_R;fod@=jCbwRdxtSvUp=A%egZ0xj`SB0YOlf
zP3f#>*eCh`0*Kx~Q_O)NOsgp8QAWAc@(L9Z$2A)~{vfa6$->>2@%fUxQl_EjVgA-5
z#kj@3q!w*&TQ?s~clv58RH4ZGd%hWRL>#L6>XSzIudqHM|E?eNV%^8oJ+=P4SPzA0
z`4Z#6w`o?=`WlfWD~yHxt6iTX*YbQisciJU>XN1fn9*rGjLB?0DGaRJ5&XoeG^F!&
zPPHE{a1ztjXhQ=G_50x>?3M(lVbZdnL?S@n3*=*~PaRXC=xJTnQexWAvKJxHs{RZ$
zRrYCw9O{&|j*KasrhD_Xtb?{<xO#-4NlBC7Ifn9mT0;pZIcHy1ZbfbOlS|rYAm1U~
z=)z&GtUlG_GoHsoyIJK&vzG+LKIKh*4<l86(S`3)lvz|HB)HRr#UGsT=$?AY#HoYH
zTDba8?NmSa1mfUD8W9szqU!2ofCA>wAo>fjqSik-N7;f6g93h2W<U3@us-jO<c&xW
zG~h{d@q?ZC3WRBAxXWUCB6>!DDniYb&y4;9gZcbQxJL)o0c`gbMBVEbaf``szwly$
zDUNE!ofv&=>d?DYF6znY4DtoX^4XBJL&?yWH_KXaT0H0gh$t;Ya5=4x3C#UU)hqug
zt?o@djb7<aN(N-BBH~rN8=u|Wj|bjA71mgr%MZ6(P+h-xokQnZ@&zz|B1BJ24IAh*
zeWj35Hx+;JDgz!6kJ+mCu98{1gdtgY?V0xV|CM=4ti?-x4QOiq6YO%?xp&s)=W5{h
zvX(!IMCP0)VBW?R$^7Pv1tS0_xGCaR1{mNwSTR+{w8C{00hZ@1A}t0r*}_t_WfU`T
zi6e_b$*wQ!&z;$^v~qM%BA^&;PcGF^3xW$h?D|#F#P~gx+nD#&!2*>5afVNS-22u3
z^lw^Xgn$10`27xpFIu{OQMT3;O&-4P6aSU$e9Viw|DQn%+pEzu)?b`_`R9dL8OG<n
zd5ftPoZC#i+mABMnAV?mcgf9^cO6=*yfta_#7VW@`Lo4=2J^y+vQ<uoeNOK>So&)U
zm2Fc6p%L7(r?0E}!M$aHd|dhHrn3shZevqW^vtNVpJdMZ!snUcZOsb}4R#xY@fyht
zuO?So&)nToP}C#{)B1_FHs7f~VZ(l!Tt%ebLPc>yYyiJ%j}l4~+~;M~;LpBD#7pG@
ztrH^s;@8}#i9J**!gaTx;C6omJVZ}WGv(N{!X6bd60oxJBg@ZtZn@p-ky@DDqmIb~
z70W9>>1|ia<LQ~j4YM)oU*i|$1s(urnvx=a)z3&{T)06JF0*APT;uVe`T(YWefOng
ztHTJxj)xR=J~PiG%*(b(fPD$V>VAW44Sz4$&87J)zRX^rm_s%vIgQ?ZMM|WQD&w&R
zkmYgNQEZs?JspgVzwu8eWKq)@=$a1ox%or%C0>kIYmctXR@En7?75{Doh^?|W-vB+
z)_t`o*Tj`b;b8E+f?@pjvb~(2u|8erZkA4=Br__FyK;ltBgTeTsmFW(o|34`#bTtO
zEE7Kme6MLa=~T187|^tkddN1A)!eSHR-^F2ldz`<?jokY^}R8AwNBg#Gg=)`N;Jl(
z9I1pqzp6=FVgmfR=Dr8+9as}RJ<0g`Xt=?Wz3GSe6ap+f|8QqLrCR~+ONiyoY|Ep3
zZ~fj*YtKJ9XWb353ZyV1{pUWINxsZUF*(!oMP_)MsftGYBy^uubhQvXsfzy(BGo%b
zW_7*7VL3)=svHBULpYg1E)zCHUrySk=s#EAG_25dgfED(^w-;@v&j%^^^O0<Gu6<0
zo5xG=po*4Syud+r_bl3I9(91uopySvig0Eq&o^L-F-wOJZNm>5s0X;1(eGC`KDNbU
zzE^O8G*QMGt`hBogqBea=x$%1`G)oZX{jqOh3;a(MIe7e<{sbsuu3lBy;;t*c**lL
z)$|^ExmY%Rj!`*^W{=?^N|FUl&R3#?4{s(@Y5j$t8lBHq>b{wz!$K&u<2uVm92r2<
zhQr-lVCigWzM@&<cfCic0Bg4WI3zmKM>jcSK7iR?Ck?;oIxq6@%4oM;1LzYi99eyv
z`EIJDnH@U4kREBBk|a`ip!)0ahE3~;qc63u(JOQuYTEIgNY3zG4qrQM#JZCSZyi%-
z!Qt&>p8iK2g9b~A#51jn%=wr-jZrSNZp@ET--5XO1gDF(NBfdyIWm>RM|{nHsw-IG
zKVR2?1~K^AnyaQ`lb2A3!2o%Ljii_7$1}hR?^D@FFGkJ!HUa*D@Lq@X)R0ua>10F7
ztdT?U<Gth!%qPQ(J6b;iSeorkg_|jbLaM*|?$7Wz+kwDW<0TgHah<8)VnX4?I;Pcj
z4KoDs1`V3wRAj+r(Xm0ku}yu#A7vj8OVMzjzUlokqJ^EWEe=KeqFFNKE|w^4v*`x)
zVl<ovf<_-{yFE#D?vvt#05NCZE3ne?R~bK7mXs6i`-nMbN71{6Wup}p;b{|WG7;3#
z)(-UO+rgL7VPtM*h}fYtvco;zl3C`EwTy_g7wU!&jXusgbs5(qYD%JQ7w)v*P!IEn
ze0tbB;q=D_WSH6KyyJTN+>s1e3{rHZE&IG?wsSrXIOQ{Ae&A=_l7MUa%|f;GIOEHG
zuJq-Ubum}}U^^F38YK;M5|J8R({^<ysgLtlQ7Dx)E1Z8w<ca9cj-vmI=ic0*T8|4|
z9oZJo3?A64-xa*Mx_Dh+pG=~zEEJ=(Gd);Z(H3pS5TR#K#m`?L7*_ux`UT07_qVem
z_^Y*~E_T;u>(B7#m>v;16%PtW^MgSy8mnFccR|PWK$utbMzsT{otSOpMGu!5U7__8
zHYPI7LM*z8<2m7abi#3k=(5K{&z0M`e8}_-AvD^(x=qinSjKF4h@*a@y4NCNO^*o7
zD_D@~BOS8Qz_T6DaIZCH;tslL9HN*giL`MQrYk2n_eavQ?7f6~x$-M6;a*J;H{K&I
zD@O{SijQ=~q913aC~-27H1$3W4>`ER2bmO2(Gt)!rJ)n2NzQ(!FNb)Hm>$$8x$9(o
zFZPW?hLf0m8OfGm$x4bmi4mB=B&sBWNs;;29!H-FP2d?eS7%pQN-37&qiWt!0}3N@
zN)B$?M1Fe>)Nx9T`QePesP_tpRqYc1?TUMbid3)vuioaVT<g1!qLAe?F^i|hH8PNq
zx23dqF@ba`8XD0PbDb4-|CCylNT;6KsY8C5rq26p$F1omPy3Tmr3leS_(<Hv6Rzy+
zu6iO@17cTu7Zy`ZoQlf!+@a0M)#%mDrAp)BO3}@dGOOA9;L+KVro4z|6ci;<!ek8c
zxe_RHHVe4xK$LI&Sn9B*?p~kmsrFpDr$zAU=^ME=mVp0<f-EJ$kF!)qTif9MLWzHK
z$*f4RcXxu}DMJsG%gjU<oQUqmXy~dAhYj~f^?M%|+v1AtwzdrvL8W3s?0lu}!y}@5
z18ruzd1+Z^WF!6p_d+rzyB5K@1`5adT2hh!$UP0umWbetu(Dg1cH_5^6=V3{Y5N|E
zh<Dvhd;XO3>HD;f&0$g*{^PXZaKl=CYpP%@_NYM8Hc6r^Ewl9FB|O_Y71xzi^W%LU
z&-u*!ta3teY<Y_A)W(Y88WPB=43U+WNk!dUT4!#!Uj<M$!|(@d%0e7Gu*Qc(MQesB
zqL)Mo(ZwFPpNDh^K8>IxcPiVuaz&K^_Ws8T&aYg5fADa2Z$iz~Yv%}JA|HP*$df({
z7(?HbIdcaxCd$w-phpf@o8B$CXH_F2b;6a}2lCRhq03U>;h<hZ#!;`Xp0VJ7GKlr<
zesfn{rJ@MYY_2iL`bfx)PGJ1(MKd_-f^7D%j3*@dQDjvlh^wWssOGD@bd3(ulE2k~
zf8Vv8WFdF}oy934K_pOUWN=SkX3v@1b9yAm9ir0Tj~(h4;wQew`w&1*y%__`|5_*8
zYFj2H7^=x}LCKc%B$UQ>pYx?f6Ln2>(UNR=(%k$&@B@C1Bphu_71Cu=^*w}b*`p;!
zo22#LE5PN3wD?9X=SDgzZ#ti#pynMzZxcefKt!;e(nurTWNx}Mu~0r6=JMNCO6FB9
zT{vM?T|KEp7xL|zntgA9W$V9~?H7fJkl{c|eC)8{Xr}+qxa&qYiX{L~%DhuEjhf9|
zbtXwxr$Pm9!~Rh2G+xx7$rFvG-c2!&Pt~%N#&@!RE(v6(@pJag=JT{l)r-lgiF9Ni
zr)joXD1@)42*#uhVQcMVk0z2yRToCI!@A>o<~u4Eoy4tYtZx<?7XvlDBytMf?v0FN
z75)86m}}v@80-+g1Sc<=#(%U%Myb`4efpQz`yPZQo5D+3y0`X3^*?UoCj3Qb0Ad;F
zEAt|&aX}v=S~Vt3-NYmzOFfwyA+~akL*q;h=%9W<t@T4aHIvD&Iw2PSI}7+;M>p7O
z(8-VYy+J4Q$W^pzi}dKnzNmi*XHRw?1X$wg(XQv^w5eMp#;Q9-oWesf_=as{lM#bf
z+vVuXzs)|6J|-ObL9g<;;kU?wakHUHXu5MKqLJ4@vfL+FJlzQqTBuXp%j(HYU4~5l
zx!Y!z)nNeezyU<yni|hPT(P5m9NqdhN4rUGDN*AN!Eg6^`XYr~FO`PFtM{4O+atgm
zQ{-@*0I{^&JREi0(4~U$^<zRUXqUSgY}L&e^0b-xl4QOm1+YD~#hkLtXzA_wZL_MG
zkj*h51qF>%qpG*_EOO23c0mQl03-F4X7iDJ@Sdb>w$Es8?&coO&Y;ZN&St_B_i;Xi
z3>ZlqY#^d%Aja0zL`tfy*<&^n*3M!5(Ko(ueFH*yA#rbzxQr^5@I1*Iwl)HXqlznf
zmYXW4ySt}^|Kiy#3~9;bD})66xRzUOGunC3@aZPhfAxOu-39w}lveWT=SB2iyz{}`
zl`WL>-KOIwboNAUmz;}V_Rr<gbG%P`bieLpIsAn8ippj%`Y&GBy^cV$BJ&Ok{)>0d
zExI3Eh=n$$AC5abtdp&!M27W9n<MDC!Y?&V+NSj@TxZyu^Ls)ja*qzwSSyMOV#;pR
zbvT-Qf_Gwf{$%e=t@3<yKp&aP%pmS_5oiA5nf!pvoVUnU8sy_g2((9k;^x~n%g!?p
zF}r9Jx#%O)o09%vLF^TrdDd8DS{vx5mFEI?xr_DuBSC%^D+YGBNCef<8thByO%20k
zwvz($L7~$!4;s(b)z6TAj#=KXj3*K3%%~-INlxKK>sqy&v?0+r$2^lYR3|r7r!$Sb
zi~Ysz$)|6xp0#>SGOM1**+M`Oj&l{G*=2y{Dlhuj8JxaWeB)bGFFIJ!ta8RETv_vn
z%_1swxO<?19F$cAJ0|SI)C3@M_R?OP87LW5H|HHuXw#}tiF;YK`qm2fX_}@pT{LFz
zRp`h}`8Q<o-gu{E(0W<ew=AqQ;rh<46&DvwDI5lOipr6XTkg9@DluB!_lczwPkbTl
zEShzn2i*16wqyU|<sm=8<WNVi9(r`_lD=!$=l_eBAh|g#!|!VeEGnrq2u9xZj5qL?
zHfZAX`}TNllK>O~!mSHaIK#1nzj&%L%sLmwz6}(@j!nCHB0TC&E0agKdC5wXyIOvN
zwA8TJOS5z<ww9Dby?NFS`u>yO_38LZ5zG2EiK32V5&I?Fuaj|lsYaGkBz}|yO8&I{
zli-Dt9gb)=n>;~&&bmJ>m4^TeG3O(HN`r;wBczWblv=c*XcOX-MZV{ExcOmgz7}EC
z&c?>H+3bT(t&l)j*?stV=5q1U0hoFpAl;a<6LGMyw%ibJbtv=~@5R!Bfq|0F#eK{*
zT;@8oiB4+wf4PacwUl%;NJM_hG_m;oApe0tODC2mU%h^!G(9+2OUX$u(mHBZPYWKh
zktbZZXsD=*?dpa#D|e^pO#rC_O0$DLD}b|(<pO*Qx&$&F$#X_1+eXP}a_e6(<V{U0
z`zQ`Yvc6ZaQrv8?Fi|a1ROCoiKpe0%utF?(-tLzYep!@lI3|FBIHpYVr3S2K9Fjsx
z<>ceFMY&*;t*)9cOFV3BOC$}RJ0r<2jwH|grq$&O#_ALw>eOE(Ep3|+=Qp9$ScP0)
zRsvjxLyk;L6gUtNXZeTj+mk~!0E||@cMrmzN&VRki)q+nkG+cBIwbX)S`jFqxIe#d
zieSz-P*gvnENp!&Z-vg&Z{_wBvGh!gC0-n%-jpBDD)U+Nso%nQc%7zsOaoo99T!gX
zaAtK){ARkYf&7`UpciRgl{$hmanL)pnz44m)u!Q}IX8L!ROk~=fu7|RT3&#b*Vma@
zwXa?R?t_-!p+5QL`H5jKg3?pCUf1ZV5egA5m=`CmvMKhGY3zi|Icj?gH3$Kct-SxR
z_<@YiCyiXdWW%Bxo?uaID0fyZlsQ88M+f~ts3q{7ZZq*_8tD^UTHynR|E+`0{WAE^
zUXxG7GVBk}cvqe5z^#lf1iS9yqd$1T`%b&%yIXSf1DTJeB!;E<Uz@BIXIi%AeFyXa
zC2kUMOfLhs7g0%}@+P*ZM%ft1C0${<Lo{yfHN}(I>GjtrF#vh7M;#eXXM&o=cw%1t
zy=hzwvf_={8-FL+)9L+4lR&k`a(lythF4X8K7JE}WGz6<gR7MYIELy36Qt+rjj7hB
z;kHH_@%yFwd^U4rYLT#_4@@O+Rp`4($2Uy63Yvan?bDac4Rrb26d;1dCwJRbGHN)a
z1hBqxgDiP=XsYJ7ZgNGP_%KhvI&Du@AOU@OR_Ej?7gmoR<K3efSKw~)d^Y7i3a947
zXf(YeXZ3QpAZ<M;p28H!i%ugh==0`=TpiccVb1t+0nPK+;6gQ>_)tNRN%vUr8>-<x
z+u3x4Q7s>fv8P7m&7<)0U-9=)!^sz+DeSNit?ef5$2y~z;^8k9BGk4<mg&s;V=`tK
zBt8Cb^zL{}k)QlX8{sFcsb$%&CIpY#^Q_1?ZkAA<u9$Vh@Q_(e`|`yNca~JDE|H1x
z5UAxItEk1p+!|p+rLd$%jb|#6Tbv&qgyPt2GJBm}!J@GxT&!S|*FA_x)`aj^E44WV
zbG^dxxxKT6v5bm-3#o%m9ABaJPru!J7yQK|oC_bgJj?yZvF?51ZV{*s4E5kPw5;DF
zEEI^@5Iv!;_&y7Iq3^Bw4I<zyJtu3}C!?s&kX%KNx(tcsEUfRXfkMk7X*<f10RPzK
z8HI*|mg1V24JqnUpHh4q`I_=<;%%#uK!MP_!I=Fh*{Nt=wYOVdZdo>wK_RPx-g6<8
zG5gBXurgBwA53}q>?SOgr-h1-CL+Q3o~J;KG*ck1m5mfmm74Wu@w{f@!6*5bH@e|E
zPK!+PO|l_aGH9#A$;V7TwbpvAc_iEVr|^mHYGsA(z{lwn%k)pSaB^)Da;|*F0X_&j
zW==61#70wN97iW>^*VsPXtrJ^@WaP@|6t3HNzyNU$-Cj|KEHjbdUv4tVnj8`ArUBP
z^}&deooEqia=&L~iV<D*kmR_a1XbpSaHHp$#c*IRrj%?+4GP5BZ*BqNR{tEHam@N<
zOAGds8`LlRv@u2eED~76j^8E~h|$iumFtw-tdc0;evYp`!f276h$K^BhxDIkdc=Po
z81>x*SJytdXm*;mnwHoW%zv+Qy_^-|=DoC0gIH2o1S9O4#>!v*#p~dW*l{M#Y+iwV
zg3Cr}Yr0B@QVf*&KeTzdPDPpZ(ZAG7_>WTUMe<?Yn-*#n4`G*N=em){_1EX0?JHjq
z;k7(l4X>DP318rJ(1Fs8W~}VGk6xKXzv9k2=T9(Or@S;Xu<<iVv#&n-tzOi$nNi1z
zl0w(of=@WB6907XXc091d4@Y|hI(bo99^_%3ziFvw7oKBLj4*awbtQESRaiPV;XUK
z5(?MwZL+OkNv!&%)MeYLt=LH_nzUu}(e9uH?qk%<OQrQAQt#aJPD>^9h6iU)*?62K
zh%Vd7Jew>m6KxtO^CG-A%vM9vPTuQ2w>1ao$e%?1XT81ebzGDb36_t;B<<^VV-G$g
z*mi+=M%tIFyQU<J@@iQ{!6DuLTuRbm5Twj%WP+Pj@|Zni7>PQEyez-}Hz(vX&?z{G
zrOvOy(dnU(^dHwpJLha6dxJ_rZNgIr+Yg(%xQJP|vrT!_{wPeD_Hh?Q1WN*{dg|5U
ztErNaU@2bF1cv)&BtMA0=cB3h<xkJ*!b-)FnlN5NaQayGJcV89wTByDHKT|jm0r~k
z#ll?E)M~blD9gmzb^)$V9tR2iG2^Pmf}jk%@wT&{{c}%=e;Zc5UfK$>_DN%>wF12k
zaHcn?XGa_Bg3puBKKABVAxZ@8Jx0~{w$=J<wULp)5i6<7b5H!iXMtIuPXY#AvaoBd
zl12JUv^^~-Mm0yvIZyOe!xWJ7<&2rexKq6_7%}o35X9}R&KHr9Tu7Y$)#an_PutbA
zE>^+v`mVe(Wo(P&FKC5FgAUK~SY5zi%{!oGbN`${KCf~Q7mrky<RQmT^$f!V+}7)W
z!Di5b;MzmIf0aW2k$ovG-I$V{4ML9Z=!(ckT-6z*Zx?y;f1vKOu)s?W`^RR}gO`sF
z2r?b2tOWwH^=&bA_Hk!PxCt9-bJcg$tvyCB^ZRvVXudArz&p3rQ%|;isL@a_l!k_C
z+j&Pq{A0vOonXlf8ya8PP}QZm>v{g_*>}oN6+)`eMA(pGT6QUcVy{7kPvwVbMci!F
z_S*vqgR46w%NMLp#Q6G@nSqvX9_IHT+=!vjp9i;O^4y|z6Gz-UP&SXu0mwsJ_X@Px
zQ{OB{FJ?JhqfMSd=Ka`$W3nQmi97yO@a<M0;P-&@^FwZ|CV}3G&5^}~Xa+d7-Z#d3
zvn;8WyY)}kF@uc=-3qZd_lqQgu@^gkaCH2F&ec=)nU1(AhTX=eA~ZkFxE2K(ib}l$
znrbp_7CiH|>nm*}5`<r|bnR71r?6Kgp0^CAdlIC7Xi9uB6nVeukb2!<tBWaKNiPzI
zm6|jNQ@3bGaLPl%G?)0nvvP%nQKbrCdmdDlA%l{A@FG+!r<u9;#Ze{WSxqHf=g&n-
z>M$jk_LIw^p^};=+ky`AmUppS8crb(mfoa&UX5)z?1c7id3Z>L$Y@ME)|*4pd?;}d
zH>RJ)(#rZQk0nboyV#rX*^Q_w0twO>CjFVkUS%^q9~<|l?yBF-y^(4-lscOk#}C<c
zdQ>1Tmpv5^@!C-VoS2kJ(VVRfG@cM!LbhduWslRtsu5g9Cf9EVBya@+cl)C7db*zw
z_cp2}rzp#2T#+j)Tc?by9)+GEnPERnGw+yWkh2iC`O)^&y754_t%q;7JcjhtlFPCi
z!5FHLE~9Wpx%gHvBSLAKy!5tf>O~VGBW{aXCwn&pAP*vh>LM(Z`1vPk9MM4{!*Pqq
z0|aY=y&7q$2e>d*?bEAcp{}ksp@twG^bN33?54C_{AT_W-87)n;$N(07iGo-TD7j5
zi$TYl+Ba;uzK|n-BlU>X450{7kEP$Nr}HNYTEqKlc6FVZMQ^I~i!#nB6#k?l6n)K$
z&mBES{~R4x9pyE;D4CnC9<Yu6L|P9=({*=Bp{LSo{ErYpA^kTjB+fNUxe$sj(Ge6#
zYdltoZcjL=tU5z`B6A8zp)o#OTbBL)$nUHP`~>9oYJvy^yYG{DJpS=b7eGYVEV;ZL
zBR-<?g{-9%h=aqO!s_sTmXYc*LJKK<%PM4<op(_90*E!8+Zw+J2aH<zSrKpjZWqT0
zr><|>QENB?_N_BR<cj^EHdq-{LREnqG927Df)Bp5;60Ad>>gV4o{63tsV_gPk@j<q
zT2(zlA18(;u5av{Jtp}Fvin*i2M_OA`pWiWb6Q$8_NJsGG5K79l5MiwhVmN_gdI?4
z=9rtP9A%V(wLdGB<hnYnAJH?3R7&YX+p^)TiD^J##_#<21pr+JQJ+y7_X$Ak*7i3A
z6ZE#Z)n%qG$?7a>9d`0Qb-lWBdLE6~2^Nm1k*289_&7k!w<{Kx_0<NJ8kuH$8TS!g
zA-|;Zo6@&+%tqFj8?vEzdqIcP;?N*PPr;US8&mvja;u4gg2RaqCmtx0{zp?b#z%4~
z<BMxew&b$fQ8pe3v>>e;V_C3=f-k_1uHRFHQiMWNt+KYPD(tKp{0&KWRkPl_{tC0M
zp{ps8zBrafyUJX7Equ`NJb!wm95~CTh)!&NP%J|DA;FQQ(TPvZ@&+9Sb%d&-b<%-_
zom*?}@$mzVam+d43^Hu<zx%ac5WX1V7Lxgw*L&&lJX*(2x#yx%x^HcFZaAc(E@`L-
z)Ae#{H${hV?cK)IM#C?R+X76$ZhA0eBlS2`+qi-=@<-|jK4p-#ySV3@2z{gVt9H&v
z?7um^Cxg_fBb}v%dHE%=BsC~hJq=Iy7qf@LALE&Y7l`L!1*|O8G~IzA6m#N55Oc!(
zg?wHGX*q3=2W{$oLt3u=SJ6dz$|pm^W-?2nO1$$n-_9x-tG3An=5+e^==YmM#@}0M
zQMXcvT6U%|;~a~iJ*BLWvc%MOxzu0%qj_7zErG<@&el5HdIHt+LifZAce9bBBEzP~
zXa4GHgcsn#$H~yq^mU)HLZ6iUxs)vWAQ5ZrGhUIw*4@NZ(X5cOGS4B(wzGQIxwuqM
zdXM?RBMEiO^S*dot7EgOFb4yH)156cczRZ3Mb(QO#kpwV2O!2u>*-ZxD@rz5z?627
zgpHBj+S_f`AN5(sDvxFzf9We~L>0~LifW5OcjMe)Ts**uf=(n=c-+?8b0C~ma4~E$
zuO9}Gqt+lQgvrreCf8@#V*evfDGr0?y5H)5d1ToN1)1*%UeDpQGWQH;jt>F94)YGZ
z1oApvZl90wB<oR69ZbE%Ix^F;g*9VlEXl;{bjP8fI0H@<7W_)cc52tEbHuEObvm^E
z=Y{;sVRpZbMl4LY*Pf+Yp-s3eHA7C7OHkzA+N3y%jRNq%K@AN@hHNXT5vA7mVcO~)
zhcvk{&JO89f%d438SylBg<vd-d4@uXF1566u+pZ&Tq(__hFK-1dl<=G-_(&h^iqVY
zXG3sOmr6w!GlMrRBRA|`Y9R39{-mfS@`%2+N63_)v%+fO%>aBMY$)8;0Aj&P09AzW
zi0A!e?ag43?TVINaLSR;#(olr?p%L6ys_~y0PXZkU107ErSa)OPu(lPadF9dzrVIh
zMQ^#=2<xD0u?{1n9aZ6mQIk!dF)YwWcSU2Y|9R{DJ?a1`fzepqBy}yQA<!D*{-%8w
zpf_Xew!ji+p<AnIN8u{Xtx1tnE(WZg@lc+E-IH`j*C)LGUy0mA@b@Rxe#H9MU8m5$
zrqP=jVO|VLDuS~X-*9KX-=nA8K^T@EarpnUUNa%uGIn3wVm|G_8HVPo{uC;${B$*}
z+t#D}x6}*Ghq;GczBdxyImp7v3}++v_LECTN~M6sWH#Mp#aKa>m-(dqTpRNJaN3G`
zk)W%KYc?@EAe@3{cYK1saeE}@%1wDYwQUtzHd4BE;p@Y@)?L_s`E-R{E#*O9G(<FC
z%+H4&NFc)Ln0%j#;x=7Tt_G)!N&wUaRb)TVqH?L1Nq2loIbPf8ZlMD^d6^<M2uBtX
zn4J1mJ{w@PIhT+bEnQY>>38H=(}-lNmxh?ZSk2|&UJ6P&dO({4R{~II1D;Iy&;FTk
z->~^ZREN>F=SS8%q}jXS3t1wKNOLP7+5K*tG>sXt646o>|0TXe*ZMSGq)@r4EM|Kz
z`XXAK>QT7Z<IKbpgAw;BnI#G6>Iy`e7anGnd3ZLP8pJQj&x>qEZf8JVFU<`nEk!e<
z{rRc@btAPfWq$^$mC!eYy!9@16J;&$JX?X+ADZfmOrW?KKT~v6^Y(;oJ9n7n<ZVKy
zj11T7va{^x1#Gz)c4Xm7yl4JO3`j>2_Am=P!mlTyGpSNhi{Bl9d#s+>*7J66+Ai}u
zdpOz%twqxad+fBTf17Of7<+-W?UqPd$~iqMw&ew5im_EIfUpB+MRAYAA?o?9WrDZ7
z{5|}Fts%v=WnVIoOlMt!tx0jFzW^=db-b;8StQFxhJz%8A(DjDi^Au_yrYb!()GW(
zr{r@Fhg>%^=riTVP7c+r+b-2dx&3z#ugm+nN+ms=v5UC|e+Clfekq!7uu2}gm1U{u
zLyVy*MD$N+s8(%DCF(ISm%6B0>07OW-QZt-F!}Hl{hQZG3F}RuwjO+0A=SodS)a9w
zOe$#Ho*qpz)!xe`)*O}v3t<8kF{J{k>8}oV+2hcdxIKWp?$$?T^{z@okOJ5SBz|v=
zG}@aDglRkDw;z4IM48f*5Df0^Hv@md7!;HHT?Z1bBLiNGNmx}tJSu%&#|YlWNZ-*I
z6;mtOblY-H>M~=^H>jB%oT&+5ZVtJQn+Ka|h)^qKGg6`?QDAKw)vWA1W*sUO5Q$Kj
zgz-AF@YrT_Fd-5*Uc{$tHZ+Z@2zIV1=nTv_)<h+5jc6){zi4MdaY;YZv8rF-AVmEy
z@xlMMfAGL=b7k^gcNXOP$MFAP?meTLO8<UuW}LxM0S9SPRS={jC3LXROF&A11On0t
zkRT;g0cVg-Ksq55X@P_$C4`RBTR=!cO;D<Ir6Zto_}}M$&Wrn;b)NfK>#S!z`$hJ?
zlC112J6HBqe!uVc^WI;hFY4IPxv+y#_CkE@F+%B&Pul=oS0^m;caK*_w^>xwCx_|v
zbjG1NTGf~$TI2El_5+^LwE3poXdV`ua2A-MiRL5D>SeGfO1oZ2?tba}z<UlX%mIG@
zj(m%|tD<fxsII?&z~W@@_Pa7^qnKlEWq}iV0~9HNXhR4*;~sPA8|}vmcM?X=&vyCL
zNU5l%a&Nnjmvq6`WrhaWn8|Bat0o5q-23ia*)R4-vW0~k43pB_Tv$nK9-DQ8ZC8e>
zy&#F}e^1-{>-wpPx?<Ow=}5-=mU}DBX<POw+?PhUOf<~|kchdh;90}N8Iwnux9Np9
z_&<zvdJ2AdM)#mMiCn}$LbZ)kNiH8?89PDXA-GX$itP%=SM~0+O$m{ifL3o_Qk77c
z{uleP135@mB9;<!7q$k!dPm-@&{i8|{nx#0w$aYg_PunMzYY9lD&)82<$h_4a927e
z>^ZoD0H%cv|1k7jO&R`GUTMJgm#+%2CK3L)k))xhlKmCMQr5B!sV5{m_{oib806#p
z)UF$*G@?o)Js+!bEn)c4cB)Fbv2$7#mwH-}r{;}m@iMvtZu-d^tf!nKJ55`j-^x4;
zvS0lMe(~#NXcI@Z%2=YzgCm-dz}%F#pTJ4F=4aRBR|<0DP<Ac;azc-;KKF`6!v(j3
z&R!-K7Q?SdXQ!lALR+#MFN>FcC~PSK>JipV)d=X4khr#*F{iz%A1|y=@P~>vC)my+
zIB7opi`TJBHuuVg0;06D$Y|})^l(^i-5T^^)=;@&lWy@?wPi7qGH4<Y_HL-~>6Gh`
z+epiGYilY`Ynip!A3H9Suh@n@FqZJ(77LpW2DsxU=FFkC4tct{Fq0L(uT|HFW_2f6
z+oZH95rpYSAhTPcdNYp|yhy4PXSpU&S>UlpQeJYUgBZBqg@<>yk0a(@mf#YbX`s59
z{og}IX~V{IzNdfEd_|U-aQD*ccVZxW_ZQg+K=qEnnGXQ}+vF;>huuC^;6PXTz&+Rs
z2wU(&^3u=Cf=0&QyEK6J%4RVRPWy8-D>qR719!lD(y~{#%%A1`fJ&ej@&v9LF2K(Z
zB{jf?KQA|Dh6wUQ;a1qCF=DvR?{PRcfVS}x|GwB;xjq-0XO22<^7qtp7>`~sfB#0E
z&PTrAl9XN8{8;T<JS=Nu%^MLYpFRbfdWzxcGlYqH2ABhi*5%*K>qtwWTHy*M=C@c=
z*6bpkBUT=T(|k@PzcDdHp5vI5F1#)!rJpS*&-sH^?o9D^?dxQpr@vQ`H~{30ym+bG
zs!{%>px`h({m`rGWt&F&%C8?k6L|ptFx)0xtZ<J8|4fmnXAViVLBovWpIt6B1Cn@W
zs9!rsYqlsmM6ZDy%-|`EDV0s^{u|_v^ba)+{rlJF<0jlJ$4d#(uv`AXN7q3wE9AH_
zJhn2z#%&zyBfjPHNO3q^%mB<|Ja|<|z)fZ24bmZG=IExOZl0b~{O0T?By(ysKF#I>
zcEf)oAmiv(e(;nh;EvFm`*_avZ%l$e<0JocUY)a901aN`9p$!txVU(49LqPet-pwF
z_75pl^<7++<KD>D#2vaIH?%7Dl`}4*_DjNdGG5u{{Q~t};AlfC+4=SWI~yCLd=0^6
zO@y{*Z^$m5@^V@phR^1VHlMvtGuKWN{6+39gYxpCkAR}yfO~MZY7~!BNhpz!PYYCU
zR6Iwqh3LTjM0zjh{}k-r%H#X0vuvmav9-zKr_@Q70R}&zwK5S~$tDy?rpo=#UD~d5
z+8e_c6W(9*S=SDEwhd{Bx+XrukOQ-QQ%d$-{^y?o838#W+`uH=*qglgKHtsFKQael
zt*4Fe@J_l_DZe-Ufkp`ZiE(XS_lPg8&k+92K=7BrSb^s4k9+Dfng&t*s81|U8@oeO
zxf_)({pfWY`*bD!?<+w4^4BOHbwl0e7)MdX(MgxI8v`VH?t+?ctAL-)+enWoy;b!>
ztx&t4pC0l%EjpYk-TH??(IaxvjKNQ*UR$t>B!C!}6x{Wgtd|OjFk<?No!=d*)_EPv
z|0LXyvi-#_t3mVSCveNzXI}1JiPvLapYJY=MCPdy?o~yCs!Kz@M<$3*w#NLx9Cag@
z+wnoW%NhwThe3PK%50rg<V|G7+<duMNM&WBH@F@LFYb<L0mg66TeHEtlwxgNrU(HK
zk9ZRP4!q+%P~dNij}G@q)({M5)By>G#j98amr$7o9H(fWMZL!ria{g^)%gom-||_A
zo$Jm33{$NfwU*86&q5&-!Q{3TaN@$O*{_pQ=C9pebUcS*G=Hr4J~}pK_~IO?yea@*
zOE@NKrCI5;V6!uJNZ*m^ZxB5ht^Fl)TH08ZwEGX^-<7CtTgMMhK8s2bD=%r@GGZB^
zBb#`9zhuRrT?wsQiDtsgO>;ma2<2tG=u>@48Fwlxm;Cl!z6;MU{CBi@+EUjQyPMhy
zzh2es7uJvR4F?J8h;VSP*C#U#JuHj@1;JdeOEzLOrbVQ*xnnX_u7{ECPW^g4cx>y^
zyO?0Y`KdYxMdW+nbA)UG^yCH~%lM0o9d~=7`uqZH^#oe;n%Gzy70z{Y9wseFr;3iY
z-e<E|i?%HLR!J7F)-9{GRsa~9Cyi3do`Z2XL-`%~%S{tne*B%$w>Q#5QR$4w_8*4k
z@n2rQ95vu7<+4c0_Q5u&OGiv)tq;ifqtCHdpcj<wz_2N}rnP1A`e9_D-d0Nlj&jCd
z<&WQpgptqcWp-L}u9+@9+)DWtK2=`ei)h`Q#oPG~uzT7jLWQ4Z5GuasW|aKvyl(Ex
zzeK5z>Fszr7j_9$MfQ3zn9NTscYS;?<GK%r`PPS&D(m|)>SnW&4O=A)Qh6-xGIX_F
zTwJak|Bu1VFQc_=yK=el|DF9u=jcB0+<DYrXPh8RoiYjmEB4VkvoxyWCWW5F9WW7@
z`;f-f1e}RbZkEBbBcBPi!yxz_{VvAFLUuqT>xsbJ7=?g}1lAgo1K{#XoKzmoB72Uu
z#!tgth$E1Faa_%1K8_eZ80&Vgw%`)}SFzne`VaUY_&?x(<$^y!=hotR7}-F)TVeoE
zmVI>3lx;zsF1CDZ)Lc|bM3&aCR_nmilZtfzJNys(rp(VgZMN0JFvD&%4h6-k*HiYy
z-ctxnJQDa-W$<ipBMmPoUX$mYm<6k;NS*+hX@g|>i{0bgX<_MM0(~2CSi~&R&q=-5
zt2h@cw8*!2^x}AV(f@y3p8w-s$Rg>{j=Rs5ld7Pf8irT$O_=TOHAQzyRoH9^G!Ib9
z{cq3Zr!S!-*hrAKRpPlh0bjL+{>FpXaQrH@m=xqc3|6{n>kmri*x<z%jDg&5=Wu^-
zasJL`J(06PzpzW^hxQjWINM-E<9#LbW`tApR3!a;p(63et<XYCyI0OTqkHfBA66q;
z=Q$@7FV(EOCw3JgKwNWlLT79H)25{kPtW0B;eyuokl8l#C3f_<c7+aDU58Z>mr54t
zzIi&bRm<Ap`cX<VD2Uhe!{Jq9XU|_u`l`HY+#sv5k%GA1OqZ!uJ7p}7rI59K2{x$&
z$$tgH)XE^@A9tgsYb~-&{U8xJ9eCbEZC*|QRD0Q5TzZ&E8Oj61^hQTVzZceD8CldR
zn(1E+mNHw{b-KvO*)5;k6p_^oP*7aKPv3WF5$qgOGxdjk@seEbDTv`yC|lJPLG_=7
zE4+4CYV%ok{BWeA4Q#;|zOa4axe%d$EXC4NwZ#)z3KLV3cJxXG6)t9(SP!_45SET#
zh_7`PG+AVA#f^|&kA5R9jc~;U^5Shu&R$l_&`h$OYsVV<!s@n|Lg}+iqp=R|#urUO
zGdf1rW^HN11%1&YuNn-H`xwo}o?Lq-h;LW*Qwng>#yiDcAw}nar|>DzQ!TeTv1e#|
z-*7&${lmJwZ%xndB>T8Nc)r({unCawP(=8Xb_@1ymWM!1qPJ0U){;W4>_j+tA&cq<
zR2%o2sT>w#Eq}2|E4&VPt!DKn9px6^_bz7H|E*4V%bzQC_XrBQ^A(yOZxMPPc(z8@
z=jU!@%^xX=_xoL7erp(C<;UKw1T6179%<N4k&S1(LHtvETfpK$#8z6?A>%D{N+(&v
zORDL-sO$0nvCaO^7nLu`vKxi&Y9GfG`yS}MIAw1j9GEE(z!uyb<R;XIijl9IyaeN`
zg4lZ_A+3k%P0R~r(#oY}x*6GzAWA9g%d_Y16Dbk{l>)*DyK3t@-5o-`s)0Y~T$eUc
zwg0w>njI+-^m7)Vjt*B_go+*!%ya-CQ+Z3kJaDhUeC(6eF)CDmxjppc_a8j|V#9-&
z2}D1X7TmSDmYm#LOaUnzklEz<Mx`{qZ>KBM^LO*+mQ-TIk^s?^6~F;xaGq$JMr%fc
zg|h|oUBqDb6>_A77aBj-q)mJj=F%VGI%5uX_wLex-|W?pLGDORv>?9nTR$1^X(86Y
z6d=9C(j`xulZxRVuINO|>Cmb<apGoiYp6#p-YNJ?*PcTUzeO)oHP6rd3fQT_M$Y#v
z1R2B4ru0B$HDLo2QA{0N${|>SsIuy#zHIcj4j*Atk#uKcjjx$MV|Xt;LvRt_F5Lq%
z-Qyk3Bfb7%_(qR^bV%Iw`H`~jaH<-6hWo*o@BU*-p=aT{!}_HZozfMo@ejjrLvQVm
zuFE{CQ$*y35AD<WosV7LkT;Jk!**?bz?`!HF^Fr1{p5;O{M*Vw^}QnFmB-4*!&xyl
ztBwK+$t@)@4KL1WVuj0#OLkVf1d{C^c?hK0^9{uFdTCDu{=}&6;@<wOofhw|Rh3h~
z`%1J$<}oGZMOygDtta|!iFA5Y;&XPWam1RBqq97eb@Z&wZ1ytS9hqq-of*PJs*+%C
zWKA(W-14kt^GAW&fbj^sW3#4nMfRSoH{0#Dk$Q<MtDJ>*s?*(7JcNa%HJhB87VB6G
z>crD!_YwE4&P?^gv1w&l=JfMvkJHE^*IoG6mH;S@_zm{uf%P<j53Y1QVND!Q&-HPk
z5xMqj13)I4!YSa)bpJ#vNqP%N!So4sBQ&`#+vkemhhd7m^!QQmOcuhSLv~b~sxXfY
zg_B2tBbkTu#RZ-<c-Wc+D>3=cSq|chl`zy$IMWAfYoU%^Uh(@Ee^GuIo)?^mt{ax)
z&q{OeB<fO^QywlJS#wOX9ck*F{=;y|yZxDdNbg!dAwC*5UO&nbibPy^SMmQ!bs*~1
zPAHCBDc0f+ICKK#kd>`cZA#Sx)ssl3(2`20<H(XRH$D950=Vl5K?Cr!f?RW{=eG<j
zNUGYEc6L5llPKZI!{f1$?t-Kj7`Wslzv6^3G*8KBkEcrO4xIdE8?vI>qR7UKodrdA
zWDxIzZD*>pATa-`)!{VnM)Lg)1+s@}QpAJch9#M#w%p^dd-BN5um<}UpY`K<|1F}f
zCL<rTB{Yv&!laLwQVvrM=R(libUE`AQK)|%DPng<GN&7IS?CjiW6Y|E=<*qIonTFC
zMFR&gsOFHYy)fA2hyUtgp9{mH+agwxhh+8xvaz<yU&-`;!4m$D+m(De=ts+_?ZxJ0
zB>OY-?Q1qmtmjOG^NVrR#l5m$YA^0?N<7~YZoB<v)(JUQv}!tM_%zZwcSRRpth1FT
z58x3R@P0F*?Lu}dx5{8kQ7wzk?tk5Nmiy2(xNXlo`;iaZs7m0udP;WVl?cT&m+G3G
z^}!*X1s{_{bYQl^67{v~PMNk##84Q6ppf0a+`;v`1`bb7?|!@SAZ{lD8Pdz8Csvo3
zTNSuq0*EU4qqQr6GABOY=U}fw_jl^LPAdP1uGV_MYj;g}a+3)S@O#F*_|>d0_Pks{
z>6>`hKMc`1H_7gphtnkX>7A!jI}Q}}8{+5+qy?tK=T?fQ(7~<<$<rO#^2gqpYIBle
zBhQ)aGhO1%Tns=Aaog~}YIyd)IGv3vX6kvKlkV@Pb?-GzC|#S;O?=w(=G?<JF80kz
zc5Im76>M@{#)xw2u=%rk;Um#CyTsJ1T6YC|P~$tMKIJFLP5&??nZHV^m6JF(AB!ww
ztiN<oKRmtq?3K!w<)=M>uz}CvxmJqNs~>~^CNI56Bf~Q`ZjYIv^dp7-j-||H#hDkB
zRv7oEm6r91s7@aN?-{VfTs8w}aQ2Z)?Vi&=44$vJ*8kUb@rT>;Vm&Z$iLJ-WqN$H~
zeH%dr$yIF`_-s0^qWSEsGLbV8K^AYIV&`a-dJhW`%Dk!v)0cm<AzX^)iF^O}CYt<N
ztXC-Jvi34T%&koBMtFsV9v_}~?B1&5T)sv2ynt!iDg+6)n%peXab4-LQ}>)OH3l4V
z`!Y3^qTAE&?%MsYwfg_{GygBBJNMd>rkUF!?zXa}$QUo+qig20+#Qjx@<$2QQ!WI~
z(T59qvA_agYn<r}-)%OSyXNdbitA<GfuA~f-&Kw^tk}-Hxe87MdMNOMn|-zr)t)RI
z7Hq8j4HYw-B~`UT6g?fdQFhkt1&uJhj4uAT9f0&g8oi<JDU0)d*49Y4qXh4*_%7(!
z`tjo3J@NN@e0rvY`kHZ$?tokkm(Z%t!mGm@HDZ3Pv}lff7ir;u;gLV8K(apfLga4>
zI+MxF;kLSeHk5|C_mcmXuG~bZl<`ImDIp2%$Fn23UmsJZziXE^T$^RG1e-=BKFtqL
zSe5mHYTIV0ScoL|6vs2^2V?H*-dy&b%CkJB%zRcA$j7g$J)0-k>v2gAOC!azMf)T~
zOlq`qyzS;^TtNiIS=lM=`m6*=f4=gNcq;@!&N+2Hq4M?<K-=J4T0nNxLd}lW<2}A?
zh4ciE*L<ez{d)<vOi>aeQWNPG5ZYL*-m~8MYscAECJ_2yJfqsMVPS}5lW|j;K1BUD
zg&y<9#Mm*}*v-y@wKL}|0!}qkW83zbuMtgN$c=rTs&d_y62f<VUf0>!8#Pm_!XY*2
zl}-|a<HR4PbS$m2w5{ZYH_<Db*AhHef2L1b!Tf_%3ysRTQ8fjY(!ViHu%*IsKVTzy
zUo{(Ma>$XThsyLUW3DYdX`dB+U>f&RD&6gziQX?1<4i;&tW!OtODM~yWAYZNCwDst
zXw;Y|zbgP>d#egA>Dpe}z{zxbd^l}CjVvYACoIC6`hnLFsi)WMy4fUA8453iA)d7-
z+Ek?`dn)p$I&y8*WF|J6m17G<YO3li#Te&b_ZBC6t9pPHmr&v>d>7wdP5;=17tj08
z<r6wO4QWjnrI2#U++ul#nfuI|tgB(9x$tupP9>;*Joj$}%i6}B9I^R?<Ldk6y_PeK
zuNy>^5jCrFKSf`{{EX#&J}Q=A;9go{^?#h8LH#98vso&-y)(T{3Cs67hGFY@UFp``
zC&RVe1B<wmRdvHRIXk$as;dUm^H(O-P*Ek%*Im6gXi^gKucl^j_x*3TzbaH;UNhQL
z@m{kt99n6aow8BiAB>~U@NL0~s!P=$2uUSAL|fevqD2wVezdu#UP_^4L@sv*G22F(
zB`BVjK`C>=O`aloXoFO0eT6?ec3BE47xMAqh)w4-eXn^MC}Y*`q4=bcgoh;^?_`27
z77I(0D}d+QZo6(;Rl`D0hvf3=r+#z4J2iBGXXeoGPpOs@YEd8dFzG-4szhkYzc!hr
z90X&OsQ4DJIDA$$a<k5~AVerf_;+t*4nT0<^=Z}-kijPXfPS{I8Zmn>dP+OoMBn^Z
zA;MOI^gj$iPl~@|XCHQQw4TIo#*{jbKmRWMB;D~lrQIYG({IS}`uMZ`bIKbSTFy2x
z>XGmutvWeMYrK8mg8c9kD%o4>*|AS|8~D&!A|^<cRFMm=21`Fla`u3dM_sv13Te~k
zOB}9OYC%*CGnT`VKNKo<d}DL5tlPrv-TQ*e^nUH5Z0+M8wNwbHv^Bdli>O+zrCMBT
z@xrd;N_}J09UkB8wdwc6o||Q(G2{^=iA?_N+Yd}Dl~i3|cPBfAb04)cm+ZOrD#v<6
z{Jo}1kKsjH?cKc9w=dcHTZqZ$DW0%gXB7*ZD)3a!dAg@mm9>Sdh}yF}F<)x|t$m@P
zgc)GhP12N)P~h#|MwhVEIoGNAIz|YbM>}ywDIi4nk|dl#%@0pjO3(04(~1N4WwUJK
zJhw!Izi_`sLlaztZjPRMNVlvXw|-stWUb(ArBGifUH^eA0qm)m?|Q1ul+Wt-!#Z6o
z_;570pk8B6aNymmi)bqqVM%#a-kMx9v1myR&F@V&+M%Uia~)Va@7#`j>G2&Db;6Tc
z24HG)jtB~hw+Kls=Yzu^MJBuE-^v%`kO*9BTHG+tP1?MpmGnEI>ca4}5zz(o<L~ky
zvS>I>j1Fj2N8T7lvMgU4-<&J{D{v%JN!S#Lo^%dglaV%(<9BVd_0lWEJDLKMmhL3S
zOTDiD_V?oGrRO2E%VuINUAzWRVj(5u5}M*@cu?<I8J4D+Q`Wb+sy>PwaVjb=m9|_c
zDjmTHlb-^!z>vRXgYU!7?G+m0tj-#wfD=#~a+Q{7(ctnjw-vHHBv3fnrFM{-6KD)b
zfqt@lT-9bl-BD5LQoBFAp=BD7y(wb!FiTU;M#L9qrw%w^Cg&+hjPIX^p8aOIlw+di
zlev=R<$kv|WM$gSOsfzE0Ln$TE;;K2e4Jw^unLQ~a1LCdtB)Tx#9*TqdGe#&czkm$
z=Qq!@nqxwe=uX@^&Vi7D{8?)i$0u)|)-=72Kbl;cd1{%1!di;Su0BLOJV1yFzh$^8
zZl~MyOB=n_$LZ!R`#YyqRr`}KkKxFxCZZ*6J`q7Tdw|HjEa{126TKvtk32B>B1-j^
zsQgE?*%D7$Jaa6eWoe`@-A;6F^2cwBmAFt?WmY~A854GyBH8Qhz#r~r5a^|sZ?+sv
zQcWwnXCRroW?|q}ZO-HC5~c4W{XICyOX*Lk*RZCWRny9n(3K5(z)O6WP=#@hves-;
z4ld=5^LX=!jJU!j0lF_m<fToblSZCmscy0_dt;T5`_t2SU$@yn^z!_RsRI5>`@0}j
z-(T|XT&G-TdZgCuHdzp#wOQe4v!V==HcD|>y)NmNlj45epKaT*tbA3Oh%(6ut1k3b
z3Ta#Z1bpyk^4(^MR#4(+C61t0|I%sS70ZPVuLN~ky8y2Zc@r8fqLDNssL?Yioyjw`
zFvrX9qNA&n_d-TSLF+AKqiPQKu2SR91PVljZ0k;2YH*Oq;hH1zan0S{^&0%C_U(Ud
zyJ08TNybii()!p;o?8LpH^MzPGrB@7FemY*PM>VtB*ZtKwQkeuj^kyX2{;5lLuw^g
znB97_j1bm`uxpoABxPv3T*<C3_26uCZjU^Xs{GE9;7R99w}*NOHMSP{U-o%!`7|YW
z-4Rd?XpIcZ>91mazQOy#`W9(r<OPa}H$mtwuSzrg!q;fdZ3W3TVKpFma*eRzZx?Ce
z-Dp^UcDz>eSK9qL!L@%F_{zF6e+_HTYIBqGhGpFjdJC%OxIHxRR1>LoW8n)Aox1Z?
zXjl=I3)l1E;rH%q+{Rdn)rccRrbmf5B2m<TAL8xgW^-{ksS|C>#*aPdWNkf6yrysc
z#Dz_)zY-RTYDN9E!sAukXzDUM^-H#+YGSgJEELwpmnI-AFqJl0T(zQ1Hk}iR`T6ox
zmUGGOUFt7?mK?4QI$r;yaFNu-@<2I|o5){i`R$M*alHf3RMR6br9Jj8|M7cpY`I56
z_-YV5*$ircny@#w8r__;95V9j*9eo6wQs4FqDC*ab#!?x00+K$=YB|8&#H!Q=l~Yt
z4caJk;c6prh5gLj1R|SvrS}}O)+s5}%g%tPum*0RZLJrLC1m~maFa`MU^>eY9VkvA
zO3;Vu5--AAc6idZg=tiy8to<$W9F@8aRc=z!+7md*3{JfhkTCPiH{Z{8LL4w-JCCK
zzfMVomQ!fCOvCB}z`ct!2%NtpklxzDKHDL1FWZuC0r^m1zEB3aO{m69`f14)SjMmm
zl*Iq!?lX!E4$Rx?bZ=B2`E_#CS>~YJv{8bizg~SBd{5SwTY8vOf8m#fb<^SzJWH!V
zT5&WE$GpsgS_WoME3@4gFx3aRTPb^NEU=cfI#xi^b01T7xZh#oo-4cQ<+#e;wORyw
ztBIw_UW7LF(<`X^tNke|M?E4->gT{g3r~oCkG#{FYsKE_{ej8pZ+xVD6REN}=S_0C
zZzA>}<5LTxkF{=aZV|_ZU9*>=nKg}{R)&*GXKc9ts%OF7LLcd@bF5Wi(^&^ge!HOy
zeB?J<eJJT~{(_ds*qLeYz(gUd6VS%~T>}YC%uIV=Wz*~A=gj9``Xq~t5c5_xVB%Oy
zZ&ZRD5si)quIMU6S`q@0oFSNzwpc7rWS0Bn#**PcB^V_>r733See^O!<{tdvVWZsq
z*gbu?p}BFKaIIhHjcl(kEHBTWh&tgJgU~zFJg2PAS!rvrJY2iEO#ac7{+k&3>yn0K
z8dyL458La2vPpw7yQm!dJM*}u)H;-8!)(*Fo*oPTq_PgJLjNbQ4A@%H<BGC~WAA{4
z;;~%?<(v{qJv0H@lh*R4uf&na@?<<DO6!%dPZHyraz2bL%qq1JC`($_9^G)A*(|l=
z`aR5YjpBL7B>LG4+u26yG3o=}u|-izUO6|>A$U_*?AFw?HyOWe!kDqXA!QJMe_W>2
zgeO0iEosJ$GFm2zXb{adW8G?ongd}jGw*D7s$b&_99^(!C5C^D%8CAa!p0JD8vVkX
zOkYEEv(&jjrqPRv+&}mk)x@*1xl0;i?Z)nb1pY(lZR~oIm-G5){9doHR5kQc=<RW6
zF)?x&xBFGk^hV>S8;6n1O>3*&WTb|>MLCE<$Tz(qz7yIH4<SlfrzEEMQ59zYCG;-F
z<>Hb(<FYsx7MNi3|3Zbjc!f0o8x`v74gDXe&}k^Z-%?MW&zEhEZ50Sf@C?ljq}%RL
ztv*!4S4-azQyC@9%s1bpE0iIGiy>*cF1l{oZaSkOJC{oD!a__wh!s&qYtL(~Cd-5V
zt@IYpiO<1$<vO{?=Onvn%x00A8{DWlXuOndG0ph<sF|?^+avB}blAOpmH*Ogmi23&
zJIe^+N^I8V3)&6N!2|D0h959LN+)4SAuvS50I9Hxql;#;2TZul&KfFWe7jRvB*u>5
z)!{4&mM%Q*Yhi~amW@HqRx{WdSW{#fS}T{73nrpAHThGYpKMiU{u<3%#alj9iQfh>
zKAlBM8X5aLQWchLeRZQ*f7_^N)rPIPjwZESrkAhTwUl1y+6(OkhA{YQy{#jre87aM
zrF68+2I3H2rbyJroAm?Kqk(;iAW1I`_8I$UHsrO8;3+Umz1`tBvMQaEm_RD|+F$!(
z{;81On{Q#A!e4z&myC*Jk<4MTq=3Yymu}i3l^B%Agt^1Y*%2B-6J9rU8C&ud;~84O
zl_cwLnFs$63lJ934)%X$qWRHi>Lj4V(yx>8E6hp;`+ZFhf3=7=vQkR9q}T6;-M1Yv
zKPk0T9~+z3JSY2C>MEU>2E6ZPh78BWrCFW>)1?En$ls28h|wi{0-|;F`;!B8JCo$z
zky0l}WHkq8jzYzOi;Gc+cwO1h^Y`H`Db~m_%f4ov@RuzT$vE@b1W#}bfxMT#1o+GO
zLHXGwf(GiRfj}tW*Hu{V#B8$NhOb63<FMPdxaw{>{o_&0YZKC;T587c{${!s%sl#0
z_rDp;6f7zfJ`4K2z51h*<OcTbE{}ISdnk+|e``3Vb?F!Z^m25E^o-$A=4B)VpCK53
zFsvR-hevtq`vP-3H9Nh-UgpW35YRqqu~Yf_Q*a*2#<96vxX{4!@vheL=u~~Tt^2+3
zJ)WN^LYBJqcH1M?*8X(4ZgkneUm}u-WwF3t()L*-?I1^MY!F|fCeK+)+4mnf5%%_F
zjSmNII1d5H<a`<glRvrC9Bo#nTSNQ%Wz^GahOq0tozGRp<}Jg6;){5k@A#+hsR2WL
zxRW4gwX4a^Gb7V$7tyTNO{GPQThCZ{6;1oy6vPx!q|{#}ZA|L;-;Le<FExDz73<+<
zozFIkoBpJ+{wc}Ma0;{gI6+6b;dE63fEXhVBX($b%i}e1u#W4<@AdJ&O|+LPW{TY}
z<1w0K9zWAPiCO)2Q|P_d0uQ_$Dk0QEQtO$eDH5v9*NRL;Fr|wdmY%Hf_Ug_i!1>^S
z1^nprCa$Ixm(U!S#rpo6smXi#Az~9ro6laq#NL`kTS>(clw@4&1~F`b(!g+bsvj(&
zH54oOAlbF?3h5ZZz@WR^`Qo66mbJuDsT!S2s_AxiI^AJYPL}HicM{W0R!&@M!ndau
z5-oo5%!p~t0UTh5J!lHRN+fgD57uuI<O91y<Y8oacSlsiyCdc==`BPSVi*!(bOim_
zm=|ufz+4sjFhE@`s~%~*Da6^<NmP4u<0j9xEw)O-go>QF?GPCJD6vB+X_wjAqRri?
z<mye%{72Vw%4G!GL)t+8LE_sWT@5_8xg1NcBzl#jv8xHi=zQ{IHP<BR)>{)mSu;?8
z-^yjyP<5uiQbLd0psYnHNG?-wVe``K;`I<AUv}mGS)-IxnmNRODKg3aK`^O?+I#_A
z2%Xxc6jv7NqVz<>$c-aGo?EqdbSL(3b*1Z<0vlF<Gc3oZ$+Pl!s~ErLnV0=`r~0b5
zlP+g8)@fr(xqJ$ktI`z#OBQUe!lc{l0tZvXztaB|@Rc=X<yLpr3;i#&mcMokC!wIi
zzFIztvMQ`_<o)d-g*C13quy$QVJxsXWgP`@txCF*9o|yumXV2WTA7KUo0gYrk5Rn2
z{{5ct+q+M)c4Tb#XprPj!h=%2x;9lMMKx$2U#Bw}aR-Y(C7#O*dJn{B<@w{1N0NE|
zEW2f#_Wa@38&xv0@0-dkp5!f|MYvz)O-p698rLL`3lO!zEKiI>SgdUOVgXo(%_;BJ
zrdzQF;^xdcM~~$MejIQ4w^X+|JkO~wIg`USSB~y+3b)4Xc}v3}`K-gl(v~6Ci#=Rp
zLn8kd_AE$VbLE@HciEOtxI=HWRj;p`6qtK(G<_u8csRIsd(8PXv+}gbfwlAl><ioP
zn5#>m4zN4a-b-{8{IuRlut;%Q#q;U4_ne)8V&1Nx$k1z5kT22p^CLqDyqwWiyVc)g
z)2#2eFtM(`il`uwwybBlL3kkocR8yx#8wd5`5MhH^r9#HG~Rpq0S#axR&i#y#R`Z7
z=Gju=Dl*MnVUYQXJlgFBj-j06)CY33gz5WYSfE>zT}Wvy_DS{W+Ru}gsO7yEv=cYg
z==3ju2}q6m-RL+s(i38M1~1U)O&?V9)FNx+F6xW8dr_=awq>*s`cMnm1W)&Qq?n%L
zI-fUC@$k~crKIJxBWb$ss!*kaDFz?j{B<xwBhrF&&2xha!R>D}kLa&>F@OCpw3UCo
z<rmqo%g*Ng!K0DO0N-m%hVi!yDvh6XwewNF)wA{oa`KtB#cH+WN{u|v8j~gp-Y!g4
zO~f7MjX{e^*K|tkmx^?%qm5ikLO-k?C@x>x$(Cl|Bn+*hg*_k!2Ij^YnfG)^;+K40
zF>#^GO-@eJ<+<R<|MQYQfySag^*#9ej-Jn{RtkM%fa&EOQJQ8a^sy-z2&`893u-x)
zvm{r}n&&R#c@G;C_6`H?jQh?(zV6d1R%OAYYWTmA><W4%TC-g^$1EgS&Qd@YYiB38
z+VQH%myA=k+f#B>e|@#~a~W+8g8H61eCVO`_fNO9d6}}7UK+-+z17p{=+j&|;|>+}
zT*zqqD$k`PqxSgQ7ubCN$bxMO3#oN~*E$pG@OVB~3d}x@;u*sMs}c(hI=qDL7=9S&
zz@$41uG-jun@4)Tvicq~rhBed1~vdXbiIgJM?1U|T^1oUwYI9m#_@x*m#U+|PSpw8
zod)8i&5WTPp6rL^(WdMy!ZWLG?O(e>TXqwczgH=Ia^le?xl%iN#nnJ3+J2Zas)|6t
za~>yBbZCh|Jhzm+%b1Sn_A19SNgi|7PTL;xjR}bfYm4D;TNBIg!`r$5Qdb27c6>tZ
zM&5UKcU3r&H+O+3ItGh?54A}<yY$Ul*N{RZ(t)!r_NnM}lpV<GK3QFHkFA#}EOhFC
z-ZH0?G^`OQoz1yRZ_~k~lQcr%m8Z(8NWcUc%C2ocm>cRWNNr1<@0*HX=6kFh?8K=S
zGb4RW5PDV$E9FM>FJuzIUJ0K@2fWPDir27xH-$;xm_fT9)ore7EQ7vhH9M$Z^vh$z
z1Mp|%Elo1zk4l0{#|Vjwg{48SFt%_AJ%m$Xl=+jDy+B>L4P_q-cQy&E<FWdy`8cw=
z19ZFjYypFF3r%ccwL<<-W6aAN7Nxd}E}V05uC>(%MHJ<$hI6dglbz@^ucH|vD=4Cz
zYlAySHN(OfIJ_sbJhz37KeGJPG2w|rf8Od{Qey46-V!8N^H&PuiXLBaF*O%IhD-2{
zudE=O@>A3iTsFkF)-JIYUIJrphdfF<YSH|$9AN1(gvUVn)neo)oqrhqeCP3TzaIIB
znb@ZqQ`8_YcUI#L%lu5Q<0+V>${bqg>RM%T=LE^hJI7NLwlQ1?3+_OFu@cE(&=SAe
ziEf)p;Cfy4hK1W$IpPzm=LHnoRr1BAq`-rix>-Z^o8|*|AX;IC)}R`<KVTAKul~1p
z0F?uZ$vxXx^gx)7;#E_aK@$8dfrHR*;b37oj}|DIh3xptQF8Q$ecWqoCpKlbsM8%F
zvlM^{d0!$I9q<F+mRVt+0Z-8_WZI5M;XmKFlOJ1H8Ju3>(DPwZPTpU<+-Pp5IYQdQ
z7ILw{69G{4Xu;rMhpV=uB+Awpo?eCDes0^7Vhg?CFYH8aw!2L#0xPq857iWCZRt9d
zb}R0*Hlp9bdJ=|u1XQeA$Q37g_Cn(EH0!|1t_Jpi+v?Nqe==MZDe$~hoINm+XnI<$
zPT^a=0_vsrW!AqNtv}~fd-|$BmhV=bs?i@0l#BR=-2W`abd33}AT_RpldbC`i{)WE
z@QLi@Z3LV^JmFBb`FVSc-K$kGzp|fi1^Q_(Rk^Oh*(F@S=L{pz2@eSUWZ^Z13~g1O
z^3kX3J^FRw;~7h!8&`F${ZW_pP*=)UsF2?JQ31A9^-OhhWAjTnzD9X1WTeymsy_sk
z;iUF<PNa3E1+Y-HE?rhDe7S(TgZgRQ$3I`|_&r9^0xoQv(;RH#6VF?JYZ;p*>_%!F
z&X_1;nk_8j>PFR;dfDF!rY`PgmTsIY^+GFoKpP(wE}&|LEfvqrbcw~D1tlz`MD|b9
zO+N_}o_Nu<Rn)Ket+?9Un!tiz5quYU&_2RZUQ*&*L1Kl8&LiUv=yrnl0=jQ_V6z1T
zA}OPD$N=rpK7ebl_Fk_ZuxIGFyZPOLK5rKlhDrYrxKYeK=9@SLKQLUNpzJbVt(@3Q
z&UR7e4V9@A_3p3-AADyhHk%KJ0D4k=ifGk1ECg;cC}iFCsrbtWjeO^BQqNN0KMeEd
zk=Da|62WGKKxqCz>F3Qx8o3E8QWPY<o32}wlB&anWHzr39CHC<#W8Ab($;pi1X-T+
z<+TmUChc}8O9~pD;-f`&e?}GW9%@Tb_(Zwn`O#I_+*(Co4Jo%XZ8r;48&2ak@gnYz
zUL9McujNSx?xe1Pg^$5+#v7(yk97_qBxMk8Sb-K#1?6bz$w_V8<aGt*Z5>A$eXuNd
zR*p|G*~1{@=uA5R_2B7Y^PZB?$HNXaU9#++;{D}~IV(=Z{3s`%;K(MK1&ne|R@L$x
zhPZ3j;pag#t!ndY%k&3X$md*e4GoTb`|`9aGvXL{w!1k2-A*Jdv^kj6<?CozO9Mev
zGEjkRDz9$AY8)76%rNO@@vncVO>Rk+!TW%5Gye)Rp>cHdXLKO=7E_sW9NA?#gOiI%
z`%RN^zIKRokCu#B$%9*Sl+ru;iJ3i(0baH#$Q@Ki#L;LN&++0!fY&UiQ*O-+0q<!u
zqa*{tcFo{d@iL?UoyZ1&1*<i}fwLsjisf>)uA?w*FCt(CQUJ-bPg`ju?2>^N{7d4w
zf+EgcY@_!I|8hGaoJ53zTQuya&(A7(<5a)gieewwn+ra)sbNjxypb|1pX~hc8IFw>
zLz|r}ZP7d)+uV1q%`Jxv8FCxhlFpv0|MG2lVm>$+T;Hxl_S<{mr>@z7^WrJMVf>#>
z0e(P&SjrOZ%W*3p&bo~*D@_fI<udFvhbKHiG;W}W0I<v5(~Ao0v0cbTe5L~0j(4FY
zHrReZQkdmzf6u@fM`Ef|A0O3jf`baOfNo{iT3a4Khntfh7>9ftO}XCR&7t%1+)GTV
zx27^$y!&N&(ooL3qK7SVrK&EvF7r$ZM61hvM#W&tyI4(Wzy&SurMeNwVR)p=i<hUb
zQ>1PBD-BD<So+JdfkraPI_jf(06qEZ($+e7i8jBA;=}+D6lt2NV8#XlnRW|`xG%Cu
zh!KS1tnZL|J%4HxpRM{3cCujbBoEU%Yu6AE``*@&#p1H)Ag&Ia5foZkq2h8q<U8MJ
z<lU~C?m9QYN38Kja(1;=PrkB5OdWds;1l59&NmG|jf}dqK;rG<+&M?CIc<E;?gGP(
zeZOu}VIi&knniy#950|!^HF6plt#MR1N$jydt&ICrYqjh8A!pX#(du+LaCB4o<4d@
zW8-`V+0wQpl&)EP1R|rjIejY|g3QN;ib)|HHeI1f#vWL_(`qaBQ|-8wy_>D|JIKj?
z$Mq$N%&E=3Fr$2HxgOJc?_ZhAmR-lh>7%~P;LgFl!0-FQHg&*EBlm0tU570rey5+A
z^vq^I(TVmahfWjqO$cDX^6E?<E!Sjau$R19xP>dxnmTsNFWWrwU64h%dmLYmIHLp6
zUp1)XdL0%l3H6QelXM@8Vv$vR<%hu!#jEQ-2ULPv^DmM-_J7u6zyLjdTo&l_kKQ1p
z!h97;*}<3LUESbOB)eo*mx=r+>d<7=q3y#abg=rxcJg<udDU+D-EJY^*tgMr(h1^$
z^%GyVDjA#3GN?}B24TJ9^Ut8DIj%BdNuZ*1cKu~^YCp(?PL`sk6x))IE`|d}HT;dp
zl{L;jZm`xL^Soiu`Q}EmsFd_npv~!AI7v-4Z(+Qfe%>O5u<K$u=@`jAvhyjmeNXC=
zwFg<h+X0?86TkM!9*$UlTWY7NkEAPk5w-h*WGo`aD>t=6ZqK~`cmK>-;Ztbu7hbz_
z^M%^0?oFKZK1rO6Nh;^B9J0+j`mHT^*h%3&P1aVf^<^BhhVacn)UECt`_Crc#iydi
zTDTJ@4tlB|^wsjTD}MU~9UXr^L5x}No&>0Oh*yOIc$KZ4R_70*{l)uDG};r^2@1At
zVZ4HLeC1IlxQ+b~<o#93O9-(8_qXjOx*c0#{t6Q5Ymck(1_h$Axf*cAEh5xXn@X}I
z`^Do0VF|b|IaHP#=c!Ztv)exW(^3J$;ar{~U|en$?ObHb3--CUrfS6XNjVme$0f2b
zU2jb|Mn_-EP#{v~!(Yy_iL?n?%%+nYy6LH(r%PGN7cW^cdYjOO5;!p5N>K_72G4TU
z*B^SRsI<(Y(b+yOF0N}U)Al-Nn>83;P{bw0Quz?lphH+x?7HL}<~!R=`9);rr&ny6
zK?jKlD*HS!9mH7;A$aC`6jSknqQ!zDZbJB0d<`Ctw&(F1dVYD${|H$E_Uj>Rua<Un
zZ|-;<`TLJst4;w*nrUeT{c3QrDAf^u>!!c=0P&Q@a+@stuP+bOn#cOZhjWe&xWcXm
zX>e*>ReyGvL{az=CUf(F>nTtd#If1-$+AIt8elA;q&(a=I}^f=^im;^bw*fmz~+zd
zq$dLX4lezW%1s9QE;B0o;8$J2UGa!j9_$jv|Mr=c1~g!qjxXHY^$hYTkge7x0YZ}_
zvP02WsX1%{_6Mui8l12NR=QTB{zU!<Y|PKQM{F>=v_&dWDbR{#+R}DmMU@~pCdI**
zJ$;q0)>Yq)36k5q=kz}6I)9iXQGi#f1o01pp+l@KZ_lHfKX{H8dgjlp;!9x}Ux5-l
zYmCwW6>u{o5ZxfPWKTifu-X`;%6TQ!xY(0MT0(bqb#*RsxA3dynV{sCzfKY0{x+Yu
zTRYu!t$^Pr_oa2VvPCYJQEE4lc32V`T_TVWoFGO(<id&*#E7iy6Ri9mH+gs!7@}&F
z?2PM4u`m{nR~}&P)D~0(+SU5+Q9pLCfrpZAxED?|x9a73!SO`j5b6G@t~=dZQi3XA
zIntJqQpucBvOf0Rr%N`xgN_vM<b%+y^YmV^4zKP)7D*>J1TVWoTlqvY!EV@3#?b@d
z8CdHK?Iv$8ExBw$cI}1F{HmPH;uCd74GC%?GpAvFmt*nzCj0EzsFDC`GqQEpwj1wp
zX@^s{6JQY}EeQe!(7Qk*S=&ywEvA(<E3_6%yQI)6+Q;kCjCMMJ8w`n+mu`PtHooB#
zkm&-$DN*`vDScTvbGuRP`GUj)GyIWmYk)97?Ah{YL+`sI^W|E3aR7&j&adj+i0$N(
zvnS!)iet-d^-U?onuF8}p`M{1Ap1J|Lc%)$)ZL^(VHs2fndipnNeEn6yUEb%;;XiN
z3vm;ERJxYD(O_GuQ=;H_6y~ei^?VESJzmfNEy&|cM|mRbSw{*gw8ksRvth?m!CJcy
z$_g5_0~4HJc#LAbC^q4XbFj?b?=*y0<=lJ^@#XPVuNGzoV0)6kEFtq6EmjiSC$;o(
z)rNkQ!A^9@oHN5CHGZmP>G}a<Y!gl2HPi(<QdHn-TxZBLz^TLA?dqrvrS<QqL(?1W
z`gsze*w|RNA3ScT{3=I#N<g@UN#7v8D*w*SXGz{WY8DeUWG`M_>R#QRuSXqs<Kptk
zPc|P^n7k$RSY*F;%IoYwHNGD@SAqGiSXV=Kejso(EGRnE<~{}Kauz}3*Q}(5ckOb2
zfULdNX(18gcmI9!{QrD+<B{~w%&QkdpEjGWpm!zj)LZcVSeH#%wDP;++pPN!gJU)D
z$AQtQg!ZpC@Be4zAm~-Qkx^dx(i*t*jGYsj!;Ukfu_;@`#?m;RgP`1S|DHtn+N2ja
z&0KHNES2@utD0Zc?$EMEr`rx_(u~1RJdww#Mm;yjJ?Bj|N|ql~YQAi?57LjU5bcbI
z{licmpgOjBmK^^3K1<yHJmtR+ybwEk@WHj*8#&7u1B$o5!O_yoJXf2Tpe)ei))H(Y
zn`k*YT@Z*Yc$D!pd_>YQF=4R($Ng?k<Nkc_<<{2*F869D`_^K+z88gFk|#TRE*}>E
zxT3L>_@tTaq`LgqidybM1ZoozXKk9Z7V_G;nMKIVKg@pBWy(pOv74wb+ftYBG&Ja!
zfkik5yy<>JkRphcy{5kU&-MII<u&X0uiP3>XAAZ@N}7)o9MOb`Ur<hHcH7Fl+}#@2
z#FlW$FfePWPSmwyZUk?FBG}f;JYSD0Q(Swq(mwuENgFIw=M+#HTvU0Q+T1jSD68^1
zb0tZbOa#zqwgDbz(ZT=Xl0Oik&f<?0LWehU1;x-y=$s|U_oX{YXUW5~SydLr_~?Ak
zensKa!OIanb+}w)6q<V}=m8n)u&L{@$;<8rl>gVs<m3M)lf!Bd7SWjOc;IlcU}eU#
zi(LZ;m32;}q6F`x5{F*mJ`q@17^n+%62mmXX*nsPx$9V5+^L>=xSdAxsonekbeaBd
z6<IWWlv~|r#}4IDb;A?1r@u??gi(OfR+jyG0Xrox@NyS_7qz*(_v3NmE63DojU*$E
z3~Un&tx8jGA7O%jCTo;5g>%bY8h<X?FAq~?FC->kYN0xUv?g~xgR*fRIXu;&-fL!=
zR@%<<Eof-2k4_gD#JS{&6>;O)!Jf|I+<n@@bUSV`UZ!bA3F+ny`_+pa+BRf|3CNta
zl$6v|5IBH*hX?X}-##s_j5t01zxB2M^XIT}5lt^{%3m~fo$5ShI%Bx<ENHr(F8t|u
z_UMN9*#PS&t?x++hQ))9!VmnT)-qwpZRbm|E#+?s9+i;|T!}3V^ZUjd&Sm|Ka@#+2
zXTPwaAd8Dx%kApQeHo)Ng8dK^6=MFkF}o3FVVc0r-;cEV$vLYM&&fvl+Ev=Y>5@`>
z54NYdUEhfC@2T&^8TTJGdTRbqRI1HoD{YOAPH)}7ru)8hP(2tJr|C-K#y=EZ_nx)c
z1FLUy1j(6i76I-Z-W$v1hNORdNUgb{u6a<vVS6`|2z94~YmC-7dya+S%<H#R&f}Cz
zJ(I93@|+5=TXL<ND#pWccqLlKavhmH?Eq>!a#pYZocYi<+iSuSmGh#z6xF7&@75>J
zA1!gyf=d<rb~tUW_)*@bqGd)$3MOZ-*?*a$A^a?B+djV%7pL3%W`2P^PuWvBXdSp3
zVU{sWg?{uJGhVAkblFvx8Wj5CRI6XKq_4@5UN3{9W~SuD7p4G$*vi%_5gX`=KQBt|
zr?&nF(f14t((OiP_133#k%Y%L4ljSIs~oliwLT#cFd6;B8_J^J84F-KyBW^D9`^}!
z4y)T^SIQD-1&#)e%(~c8`Jl*0_p#09Z!-036m&ui-Q;x(gWFk#D&LP!xOb@QebjKn
zv=EaE+mS}CnYkTSZi<!7CohAoi@q#7#3k%Z|Ak%N2$_-PFoq0mc4+I25Dsv>HF5L^
z&5ckIcdl;%BWG?JisH%e>6B>f&#vJHzZCJ}owI!@`vM{SC-!kqCn!GUsx5+n!r#0t
z;W4j2<j*wX!y}fwQDafa5fuX6K_(l(G$(~KyM0Qn6x;vaoo#J(+u?<sT4pjo#?@#s
z>HWo`;Pz(7t&3elupPEnXgHzph)rp?ytOgB<<PSct|q}{)Z5?N5g(-ruWH4SUp(9#
z8LzyB-&t{Aqo$%WK_l6IKJ+#yh*bS{yQ8*APHz8DO0d5$9W#AcIkoFzSF@@l6M9zN
z`>7*PYBn*I!_19h=1J30n9xz^eR@k7<Eu<}b8nkVzOAkViM|4c+o239&Z;L~?{r8}
zQMddss(j|dLvpw}!*+RQr5gYCi5bP9cP@Fk?QpzpUdRw<+unN7dg15e=+9W6vhp(F
zp#4BaKIr(|`J=pqhYjC+k=G_KqutJ2oN#w9?$F&f_UAi0j7dq*M}ic{2vvOS1Eqo(
z8~5_*-c7&NU5q{d_U&Sn)ym}}5ak~RC&X-r-}oc~+<!^iYC!q0v0+-!{yHOvAgYgm
zlzp9iAhrGT<tTfLsjuAAzR)+|lHFtnsj)jbJYJTewr-2>yXYopDLky-jy0>@#0n3X
zXhVtz1Lle|iB=yBL6&+eU0luNABy4Ek#vVUo(R*YB+T2>aC?yQ@rg7`VuKuEN3-v2
zjZV`2nTUm>)Z|7d^eWrXXAMZVFCZNykx@5$q9^@R?&b8hQ*_DMy5jR`A#Tl!eLnXK
z<E}kUS9o(hxuQ-Nbg?y6xfY5e9HKNYGb74tsTKf`$@B}arJ{<q=0?g6es>Jpz)qiv
z%Ik--_l4OUoJE_VYz3sU2kQq)Dz-fRt7L?<J<#4fmfaP3)J88J!#LTubFy44XMUk~
zcT#?@Q@mt2Fs-?6Y2FBq=ahEr8_hjZ3-eL}_5AQDhV)Ttv|ZicJ3&SmVxwKs_byC|
zJ;%1)1uc=dCvD&IQglz~kw1+RxUfd_;SKQU2%s;~Y${c!G(MHMf&%*<QRhZ`2RBN@
z6J>SqORGv!km6ivolAp(zB_SNJAs(K6VttF?!kt<bFS3VnLH+UxWVx5;>HvbH<z8f
z$#^qX%_r$`h~&HquTpd0SK6zIo@`&iWJyLT@fCGa1~Bzd=MeVIz2D&}Y6$-|U*2bj
zWIuH<<p=QLc%b*bqviN`a;~*tj>fZk8`7lHTG@l${VTe1qj6=LYkLtw@o};19U<t9
z+D@6q&42OVDP(`)dx-gtZl=O=D9$_oFqG2EugIHGB;6gCKtTfWacv4~(Ect3J)I!W
z1|U0~VA(e;4K!K=_tz?P`=KI+gZ~?C?;X@s-v0f2@7}AsVnYNJ1Xom0sz4}Gca?w&
z0TDtENDB}kNC}~HyNU>*h)4}(1xX+wi3lNt5Ll&`fDl4S0Rd@&P$WU5`OE$M=J(HU
z<~#GuJoDtg<V?<-b3W(uIp?~r_w{-gEQGEIVKvl5&(!6vcwEe&61Q;%H@vDV*W9V4
zonU>5;{3<*%sBIrS{HOuO1&-Vu_C02--yK=`y2k0v}`87EVgP;NV*6a=pNSD|AQwV
z83=1;MF+Q?j|y%k9<6(VUopx*XVDdHX&e0o3-0bLUZG&RK<8%6GVin}%j#@=)oUp~
z6EWwG;l(-$6Hf4~FUhE-D}R~w<#o}sQ9i5R$UvEQSVeq^#WXr{#clEZ)U^_anC#^)
zt@kHS_bzKIKr11F?Yb~CLH?ZU@X|BA2Or)9+2dhXyEChEHC9(^3Su2RQLBHc0pK4X
zfb9tGh?(8y+_3ZT>0x9P|Ij?_K*ii`qWM|SZglsC%1_P2S(yCcfl-!ar3a*!L5^OV
zs7pbiN^Q&0xlW}RPinP&j%~DFoMsC0Z?@;}F*$9IwHuQ%a{^?Y>Q_j7TB8tXpuQvo
zrK@)XX<!H2L?qm!-hePHyUvt>RfIZUa1G~E-n?mX-T4zdV_Z`)d!feo`3FC!sPNEB
z;$~muN?f##T8xh02UNi_@y)^o@+rHlDFd&72hJM{+VSqczcIdWZsbdQsgz(trQ`G+
zJ5T~6{=<RQa5-!zpec0eLR@}Y&YNZVNC~5bqLdui7*HUE2rZTRb2O^d?x)%kk4YwJ
zf{LAY7YQMFBL;rwThljq_H0<n*rvg?Qc-nq69USq%d>Q{4dvJ;38cZmTW(3wPW~Av
z*$Q62XLE6dsJS8yqX7m>Gnrowx@^X><XrWa-=z$Ggx3>_p>i~<MgPC{w?2MNbNe&w
zCB`%nui#KRjq<x5Vd5^6hK;YFpMChep!ddC_e<>(#!@3#>y1<|>?N-coT*z&Xtf8s
zPI~4R<54JMjsyr@WQbJ8C1vB6^BkhL^&%l=cxAkU9&;*MX`#)DFtcs@<zw-t=0d0m
z{oxRvYF00$Owbe@7Hcv)L~~wg7<5jD$UVJ6Dc(xW<d%U%Cq7N{f?BYgGu4I$+FX?o
z`(kTE^VaNMdiuNN9?jN#B7^;fX%h{b2U0WG{=U?cbcn`D1^|EKctfdo?#TGqYqhXP
zrq@k$g*rml@sORi78CVhq%F2dm;oWrEdEt&i^eYPxgEZ#_~Oan8j~ElF$}IenL)6#
z`Pe@=`dGJimXl$V5UOIXq8c_69Nj){VmrN)7fLsH>>sp!^F2Hu*tk;pV72r%>XxHJ
zu!LQm2MaWjg_m99kbMNjMDR*5H3!O5kd0PB$X)A9ep+Sfv?82N+|6t`dFE};jamNG
zwOUR8#+RyWIwj?lE~P2jw-k2Or~9jW5$GXkU}H5~uAi??P?Nuxs{?r#aDL~{Hu@=3
zL)}P|i#)SzdEXBjKaNxfQO6rS$?tCrzx%4Q!hH8$i&tqoCIrzNH3zGz9!-+Wx>ATi
zal|w<N{piF7>8v9m#%#oF|paeu?UPOEnDAgv!Mq#gbl<gf;bG<f>fkKx0RisiDs2b
z0()pni%muX%~{j!g}{-EtvcBlcIQ4)vOdE)b`wKM3u0uUrxvLn)yo~rz70L9bu*~R
z(~q!MD7a?uobCH<)Aqxe2{o597sx1S?M16Uzx&RXKXLDF*Sd%x{X}xXyYkM{uy&l{
zUtAhz$G_jRoTa9Ejt+ECz<ta0PSn(zAWCZ{cqKPAH^)<9r5>sHrT%c5R}7_*wUPdM
zU7EkdC9ZONsLsxx0*<3|s@YUW3y^(UMCWeY?hR3cJ%}BY4u80SviqjcDVt?h4+Dh0
z3s%7NY(J?Q;}bl=s0q#|e`_F4Sxqqo5t}ygp<g%jHkp_aoxgK%$}02oL|%LyQ|ET=
zRYy=S-HBT!8+YX}n}=%D+@ZYJHc8y9M|xX=jm!wBl<Fnnpn{XHL)#PV%6Y1s{?o>H
zUx{7HidEIiahRF0_klEh4NHw$ZlapkJ8?s@WQz|ga+%XiV7wtWth5p%!q*^mxn_-O
zqH~^N6Qc{RMdKRG70rqI9A63X6vUBS!MLxc+p|%}GGu?{+L|p2nTZ#ozed^FVJ&iT
zhGw`juK-2j^S%(H!<G6Ex+msL61n!A3rm@yh^Nrkllmq#CwRHqE0SRW?qfT?v*vG9
zFj@7HA0e!z?-e^@35uraio3Pf<aCX;ra#}XQR~ZY2B^Iv^neD>^vMW`BhGwYJ9*6b
z{^%?(M)Cau#Kon^gT0nETmJlfT~ousd|tD#M0FW&-#%KYIj>*KC8is#dHX}5EJG$F
z(ICuwezK=-z;hOz)HvUoxh>n0AfL$bVG7!3Y{+1x@2NGBEx@yT<b5gorBCgyo+`?J
zHuh)NOMjc2904VnI1|FrZpJAc>G+?rrV9>pJdeRoTV=~2wVua`5c3=+ku$_$k4}nY
zzR66CD$mD>hj|nnfkN;o@CD^W6jGDZ&Uz<0As6Pk3*h2SRte*&E!%6SXlar#{3xS@
z>q%^WSHIe+or?l{IIzw!9J%!#OdD=D4Z6y-^Rfy5{w{H8Gmy9aCobOLlj-B$fW-@C
z57+$j*3_JIP$8Lu6C>)7un||MS<;Z$67RPQ8uWpqQmVLf=w;mtgtU8$R%#Nv(;){}
zOnwvUZa9G0iiSJc?*O%lZEpgFfpG@U<qhtB99h~t7RMKIhPwL3G{hbwqkJb}t@=fD
zZh)@RSShSC)_Lm?eb~i}AP4)yr=^g<11nfyaTjo4pe3!pcbu$_rYu^HY+^#Vm2$Wx
zy1r@;EskgEw1hhVcQ%LYk~s7*e}mxnSz0ZIQ6D)-!_#LIoyOBcjzcuou)`ky?C1QI
z8M@B^(DVLB@@b8saP=((Qpx`Ld;tTd_eu?W>oNdO$bR%C#QpP7w%N18*Y5QaA40pA
zNcAnogIp(Fp`fbz^AgP`JtB<LNsR;S8EsDvH?EN>2$PKXSlM^DIxyF@uNyPE=#Yn$
zLsO0}vdwOUXc(<6ewf}ZBL^SF8;Im(BkC6dE8sxgxLTY2TCNQ@9>F-rOP@vOz^USa
zY7Q=3T6u1^S+!mdEp}~EWB2iGrbbEHp_bK{Y*jsg$=sR79J^qpqM`_BD+@FV{oC)s
zm0SA3FRyqQc<fW&-iZyEU#!Z0vYn{d63IK1TmBwv^!!PxQ4*<9H<IW|D^CtviHWX%
zELe2*VWAKUk02x!n_il&X?g^F2Al_H(43brx<jdG9DR`t;6;d<)fXV5iOFq(16V(3
zpo0d+lU>vKL?0I42PJgS8wWT<Yp@(y7JVm01qmF%Y;H@-H|V~e=&VfYl5)mGuVuxW
zZJEbJzQIdA2J@VG*`&GEgt`$FqLtots>o-;?Qbr$)|r;;=0*M^QbI;3;TvbRYqG1f
zbAgf&F0AGl?EkSTpXkHJhs3y+LOWL>W{3W1qmP9>!|&lPT|GmcOswn`1b=tc&pXgS
zCvc<=z!z!R8?`F3C|O&Zm!hEpyVF}s!$Od9b89(?jcbN`R`9}o=6@Wia>wA7$dGkc
z%ks>$VWdv6iqiZ}-MBEi?(zPdRGp2?VmK#p)&b{Ug~sMg4b=;-jp{)o<XWmQSqJ$K
z1(v$yFy-n`P$K0gW8y%q-J6Vj_D@&i6cOCC;=;25zPtB!&xqYbFJ6*pxhHgHD@wX>
z#{)U&$XG$)CxSC}F#tkIt7M|_p^W+lgu>C{ubRWajGBYcD(^3#wlPP0jK@Gv11s7*
zJ>KB>cH5hsCpB=RjT<Fgo~~I0A2k-j1P>IEv(cs16a9r1Ma6_4N9x{d4ok<Sp*Laa
zn1Ym{GZ&LzU?y+~gr3N-3*(zV@|f=P0q1u&`qZt$k3rP%+8fMU;~{qE`)c0aZmIOZ
z`m3sCP5Q36OoIL{*Jhj%HVH#V>F4)d2PPeIi}0bQ4yVRw?+uQI;-mav#n*Om^~g&`
zwQ<)w4?-&!-P*AHbCVULo|p{C{gM!mL@UM4Tfyz6>%NJCl{hS(oLxP~!=DDXlfhtn
zU)2}sHB>P$rmgY|8;2T$*5EPF`7~+-O`2sg1YRFcBi`vKNI>7>=vRKBwjye1DPx%$
z!^=k5uNbs{8s}1G+^UKgE&bKi1aF@Ghym~g%MzDr1cW?;;TH1bTGJQicLP=UgI<-`
zisuDO^-BkHsp%ZA`=9i;nHxhsaKU<T@UqiQgG=dh+w_JgHMt~IHS9d{o!*IE^2uPG
zHUIs1*O>809Xw{u_RZ6$VaAt9jTGz)<hSijHZ9)#imvX1aYUlQeZ096ZNlxSvT{`c
zWEHOuw~uBN)}df%fCr8j#ir~5T<LLmfWFk@eVwLGLAP~1Xz&sD0V3FWt5!<BVKB1f
zsMvQ(7*t?wnZohXO{e5QWebUponi+&!j-wkohe;K3m45lFn<Z2z6<vAwL5X3c5rO}
z0k8S=v%S<|HE2jFy$NAs+D7@Nz@yfaf2%962HVP%dXhYgDjo&P0S8ubbsKB^m^cg}
znzfo9q5J-NSLGGe=~&`U>@4vq)_?7ymL$T|!QfQqzyMb@5@BrWRNGCUpC3+E-PX5p
zqeycxo!o>tXt9wxN8H=Md-&#~S3)zo1Md^}RO==^<puo8$z+Y;fv3jZswKD#8s5Y-
zt<Vg28~^lL;YP?PuRo0aM|BS5{QHYZU0@H@`;+8%bNpkNfXcN|p_4F$;}m6Tc1WHn
zqlN*e<)Bk6-5Ucm+oPJ4h_27<3m0N#h>)I+^mi~yS`Io`b#=KpU!$b|`*)Y`&TC(?
zPf)TdXOA+5@S39Y&1+ZZgPga*1q>WiFyKNu{jh>34Fq`1u8N<JV0Y~J1Qpz@&LQ%&
z2TZp=Aw#7yvvkS5@w@@d8XOTE_JPNr{Zf_Zd^7$rm62{7o|mt=KjPx&YIu6ougC6j
zzuT^9|HyNat8iW3<UD;z8XrRq_($XPrH=^-&4e#8H$Vb4kr}vW^Ukq9&}3ew9AK$v
z9M0w@li#w;@794#J!z;-lI(WtBDRy8NOjuiCp{jAD`zqRUb~>?<OjTgbf2O=^t7l-
z&pW#3#S$)H&%Dm`YM4qy+}206lrrTtZ~Un>hbSqX$=!Qd!TO;xA6TyQx8@fr7V{?k
z8KBNgAq}v6_=}qW=pUN*_Jvf)jT=S9R*oxq_^(6%OiP#;3z@=}vA|({grfA$y#ULu
zsbWm$CoNxx^TlR7=5`maZ}?s0<M5~z!(6I?BKSg4tV>h=S9RZjMQ7iPzMwv=*&6)=
zA;xQ_3os9hY%-8`1uy68g)|43yXK;k&Gn~19A~&Eaq-hT0s|k7r-G%&p`*?RjtuGu
zM0O^uQ$EVg_djcPj!pjrLyH6CO4IH6P~()(5u)RB7yc#<;Jm+_8$j%ok%F>R?(VUo
zPZ$KbzVObrZx4rM9*5NGzrxlp><OENmK(XzaB~U$)HnJcdG5(UD{{|AM@53pfY3*L
zcD`+DQ8H+VDkCb*rRmk~MfuNp;H9+O#zXnC%NWh+`qU&UD*AcO%+L!?w(Uo*TtT{9
zYdN}l&UIAX#~WOVRP4MEUDF0DtA&WmX!W}AFTnnord()RQAIk@LE-D-NrjJ_*M?2E
z1NP03-D*1H8Ttyql9o%4Z*-7II2d8}Qx_xr`&sh6^y{J49a|pu(UJx7w~|2f$}pCv
zo?L=KmVXFZYlPHp2Q1xN+M<|VOT1LOl%%jyP#+$$cKRgNjoGlu9Z+k|<Q;oR4hZCq
zYK;fDej>l<qJemA@Uimpznu$k;=F@}Yaf{%lSo`YxDvatnmFg=L{0<Z^SoD*f%GI6
zzF`up(+TD~PFARO^>?9P3l~`6m0LN%ip9%icm%Aixc#tk_0E30*p74exXzCwOJQOk
zY17MX3scfRj?5)kxF_Xa62$8sdWy4oQ&G+Z)jfx)&c$a(LT&w8=7$B_zEOFu5}5^&
zW71uCB}M5RtGpW!nOCbIc-Omwd-T|^4&i_M{p!XI(nMrz2-4exgNK=>iw!G3j^N~{
zhZn&^1I+9AU2fZ(5I!lG?ay)YEnYx#j)BLv8_R*MYP1{->)bHJQRsm;Gw(6K7{#zO
z$Bj_(P0`2&bLcc-bhgA=$Z{?Kz;<?;Q<e*7i?zCSFK#J3_LCjmA)K7DdpTiM^LMwZ
zI*Fuljf6+J+S#34kud&E?MvdX8*i-Q?(~02l;PNaHM`S!U|`D_)4O<5#iol^R0At&
zYK|^Xg_hwbU8xkooR6;?mICH{6Nk6(GT&nUixAP?Slz0aaRbq>bs<NHR`Tr54Z~Pc
zkUoC7KfvCGrMs5)&G0(>uenI{wK@0BV+N#GT%P|>6vrCyZdGa!WcK^i93Vie4eKAQ
zhpxL(EjOZW`|rkcfp%|bF0k17Dxtxy8tLJOjq>icKZYzye^%x>>FZDN#4$^ksTV6C
z7!l*!8U-;nng?v2^Bf$-PZbXW(K%AuIPWP`l!N{PK^-NFJoRZavcscKb%j;ma9{vh
z$!3#La@)CiNuxGf44GQXf=r<OyfFSX!e&cyB)e$m0C%!6IS<i(n8<he+AYjYZoAZJ
z{=Hmm%eS}ISZTg<>2RU=*ul7q*wfGLspbBV=8}O_Q>g89sV8;CwvykObS$_^edh7j
z0FHSz6GaJr8Eg@M^`V~c<uBuys5rO$D&<?hm%*FmoCgZ6)Wvut>-}pR{q3ezODhF4
z+a2sZ1GVs(u;ZeYJ)~8~$QoZXvy48qrr|1@bEi1j%yEYXh@-q>4A0ehp(u^&te^Y@
zje?9LPA-SJvbIFVG0uRVj!FaN`Q7jefxr!&163<pT0!Yo5X%iIHg>PW4LOJX?=yb(
zji(RZMD2MvYj@~R*A^*j<k}C!wJ!X1i2fJ-T}>0NT;XocsgKQ(C+?@+$g3Lj3OaS!
z;%!br`C;a+W)9W#_k?qt-NVw<+Q7K!ba#1cmXL6}v!&#uHE4(%7p)JKftpU>c>5tG
z@dfHrWDk(6#yD8dqND!kx4zyFlY5oArw#@}S|G>3JzYKZj>FNND=yA>N;EI(Tg80y
ztSY8@UFu?Y@^#^sx!XK`_s5ZU95I5(17V=eGU?2W88rg<-{J|m&*yPQ-1uK|kH4!8
z@r)Jz^IW)0(&=^qHBhw(S98NH<F$70C74Z`$?sash7O(H#-ug?jLQ@RrMIIpX;>HL
z<KR&pvc?D4q4RKfU<LP!SEXX7tY(yGTt?hMXsX?r`T2Q0E;wFi0bW<_4|5GC1h{6D
za9Gncnb3)kICLUnV8GsD+@mA11LR;f>Jnsz-8?>rCAt5VSHbCOo%Cb@+`cvzKgWZr
zrn>tU$G#A~x;wwPTDFHhzF2Ux{N?y8lNe#}K4FZafJt)KE9BWc!@%4)y!i&YoVit7
zs;AuKbkecqDU+D?Ho$(%ft4=8l}QZTnT!J)h4MV7G#HV#++wG+Y>dBn?i=~bIXx|?
zSrjL2=*~K$ET7=6U1()%f*4ua*HFqz4hkhGi||gNL857_GA&lm3#$lsNQ*yqMU}R_
ziKiz;L<lCxdcfLv2ZeA&_!nVvhTU(!1QFi{S#FW766LH_bVwcRLNST8d&{OZG9o}%
z^P}{&#$OU&!x9@3JQuDxHZOcFQPILSVd@<N85Fn3iWV_SAqCC(l*+ubkDJ)*{)bSx
z^u0w8=xQ{GrbeB#{&P5-PtK0$(%auR?ZgJjTOVH_J&o7w0=ZeXRV;c3_yAYP;JowS
zswurq0Ftzc_JJ2}8BKXO<>v1iy&3j-exkpRs7h3$0?|#*?-BBKnQ<tDdq#;3-Z|8x
zIOqjs90<0-o`L9(m(H)EBu7vFd%X5r;JS+KDgtd<k%U$z^hP8YG4F@eu0~~R)%c9d
z*5UHow@xWExW5>}s4tOpl^XJ&Dh-n?>CBnwCuL?U?qMBary3VC&)F_^PH>P9Dvj^=
zjWm=-O2y31`mBSbbQTmznMKffnqu4WZD~#<Qc=__n2#DtAhh1pBTP1EM>c?h4T1+A
zd9dn37`5p-I@+4c1)3fsUkOa491X$L6GAWou^!f@gR@{9ne&fY5DhohlEeWEePbu1
zp9Fap!F7)#A4Wm5^eqIIwPb4v>0V8jhG_f^3^9;YH*|Q->lQ28qOi@8@9BHEX8PyB
zl`@?j^LJ!*4RmYE@!<qZ-X7`WL0l%Gl`PThe}SgcvYP38{{uD65}H$*>#@wn9=K4>
ze=npv$J+xu4#`hi&T))$o$HVeBS@%I8fJNS`zl`!R2;@2*lO5k=0u}&!<1Q)E49gW
zxD`s|cEb0pL4NmQ9aZbj>pAlGB;6dJ=w$@>7Erp?I56MwZY=m4;;R}oSM8zGYs2uj
z%^<?PhYwpV1MQ<PdB26l8fg+R!G;bN&>*?zGT1b|4wi-AXFqmdogP!lhGhY%w{is9
z>t5`ziEWAqaqsJH%S0BDbD#oLj#S4kZ4?Y|RY{wTgO#bIt*<^VrzHNfh8oy_pLPKR
zyEo{xJ2YHzj5xjw{;n=tHTN=NqN$HiG57tj5UM#o`M?IPlg_Nn`#jbI%DAWfd7M#@
z-TVAYTKSt~sRaFvTw&2p@O$Lohs|T+`e<516c>NX$O$Neu-?@ZINe-_xkpAXz4n{s
zJE^;jYTp|Ec2vOu702Rcpkxqg&P!xp5I`d`d`FfDE4Hnxbgo`%&fG)T=UA|^Ey0F|
zjpz&PXh#m!r+KB`Yft6WDzFcRNmnCwI)cDL?Y|WRRi1Y>m8kA9Y<h%JSBhAT#i5To
z{T_^qcxh1wa&r$O2F_k}RPW*BAYVBct|%8sU#Q9@=q)6tU8?jQDR19t%f$*zwK_sI
zyTJ~n@AOb^S$6n!K=P=9+gN53X$n_ZhIf(kL;%L6EFeVX^p(Lll`Q$6U6v}g-{e&I
zfji<#edX|gNKW~-r=F{eBXbS62COM8i}Q(nF%@DLDu0Q^fYwf%9ZOkudJ7Zfv?rk-
zyFl%hs<En-a+XrkyZz}W{)*PR+%rK7?B>R)BMh6aU;pY@fA3_%SOR$S#}WTkk2Wht
zCvjbY@V*Xu(6W9K!BZ2N%yq%>PV~4b&r_9H7ZzbS-ib)Kk46%K+BuFqF%X?}5z<yK
z0=Z3n{n)`2tM?ayPKW`U?6)FDZA=FSj|}l5$<<&Ll{y%JEr1osqvAB<@|v|*<8<X7
zl={i^(Wt`j<X*ChKA|TBAZ~T8@ABu4f9PDPx|8A~3SQN8j2XEYzU>w@n$LZh|LTga
zHtkZy^vCH_CJ(2k)VyOd-|LTUw~sV7KFP!Q!XYZM;QFVzIrw?kwq&^}FkXCDa-tmI
z%}XzetHp-mS?6ipj1Pn*Ij3n_1&qpua5L}d@b(IC$!hay!1H5#3@R7y1H%+!ph^jt
z*eGqRl>cPV!ZBa7J0pvB?b@;R{ylCbNsd&<mtMKozMX?&6%!X_#E9qO*W?@b&7dZq
zcrxm<YoVQ)*)Y`k9rhY2+lSHYgNhEB9M#cali@F3qz6*|TUpn>=wM@E6&PT0oKJn4
z0zrTf<7t@?1lWWH2D??ja5ZIHVj!}vpg=#tmhxv)srH;hdpQuW<%YN<F#@vGsEadU
zmPFoYz<21ZTkMq&ZYH3Hc3nu6<NZUm(f?L;S6mKCWVuTh5{1Q~H?$l~rCP3TbVtV>
z+qDbhG3xYNX>|RTow`HYwM2)z9sOesUMw@IP@cQZ0i-;R=e(Kpq~?p)P^VZS&W|8@
z(5<9dI{%nmUeiF@ZHD;-oi}PtQZ6W$M=m#=|8Af$JD+$Z96C2G+77~9r>@l61JR}S
zrqRpk0fkxe@(bzYEw{#%_XKrDgNox;;?;fv((1X-n5*K?FI{|*Qgyw}c7*7W<*N*0
zY$}hC6JTwBYYFArKm~uAdI$>t<ey<~sg-Jcx?^3@AnEC>;-^mrn9S4}MH+r~IX)ZI
z#kv_TvX<}no|zmX`>LYym-@(Q<>+j<2q0nxs2(&GV&?5lzh&)Ck^UBF7-#`U>j5oz
zbheb-{Jz>ZbqKLD4oZl|>p8=Hpgd=FHZiDZ7Dlo7-XUhdMERneWeLU|$bYFI_4j_~
zl;%kQc0qr68%)l9Z;X^U@9n5VmbfLn682Yro2_sB@q{*09h;Kgd!bI}&PT=hvnC+{
zJZ3BX1pPSDj66dp{qn9QdGn5}J;^wBO(Wix(-3H!e8q_>mq{Q~+92L=d807ChWQo6
zZgI<rvT_lAiJec0G}7689CBcOZ_~K*;t2?xFpLaXGEfJW>YhG(80uQo;OoGL4Ka2#
z28>7tng`OXl?WZ<O#^iw2WY4V9cw6iW=$nQ8>v<sf1IVZ0-1*?w4ZB&KB?62A#L6G
zb(C-$d?Hi5zxsVbN<iV%tiGt|rwddj&Di2Hx=YE?iR+u<6{eZ=n46R_mDZYXCLNM|
zm;MIpy(8cF7{0jDdS&ow)pD~!b0d+RX*sn%j5w(Rdhtnk5G<hu!Y!Oy=;$cMGgs*7
zycC?Rt#n4bvqWM{>de$<R-;p8(%Q-(4O9X;tE;0m&ystNtSL269Rz;g1+Th~t#=wc
z*7SIR8fC(G90gez>9Gi%(k<s&9RF%fpPAN$2qS?F4=@2*mX9xPiTPZhs*j`yHvHN%
z-dxcsgOodEd>2TRQCHO7rA{nJnVlCnY5+F$RzNLBE?x1BSoS>cy{jTv)jCy=PzN0!
z+~vk2gVq#$n!~;@Q?qMVCgjSBaKYo7%jnnhUj0-mLc>e3g9RzMzdn3^D~dmLafjl0
z%Car8!-Uh!!prLLM{d?5PMEksQIQcX9>!x@6CD*_IzY{#T_-bOE3>M!b6~x<SN=17
zM#FB}b~4hidERFXh;NoHboMcbtx)>si3*_ey&P6UTb@=3uA8|N{JTHmTSH@nQoBD|
zALs36f(^mA)CSuvpB09(J+qV4mYiG6J4x1)o|glQ+{Vh^fa6t^#>AS3qSv#*Zb_#s
z#S!8P6Yh|UU7U3jGf!jk7Q8wtu6Nd<Bo;p;u2-*X&<iZ<B6RdX@|sf`t4)}6)&8zF
z7~E%y25b(*XmxPtuzAN}$q0?%h*&DP*Y?rYX5lA=?*6dm?=|j_0a1W0QY$CODKF^g
zYv_A&7*JetgIH2J>&V?6q|}I22|jxv$A#;vVX&A7et)8h^!M}rxA%ns!ib&RnL*ZO
z*zHMl<wt^UOxQ_5OM|?zZsNSh2dwcU&#sQq<=(cZmX@V?c@edJ>u$KF&Vi4GB5elP
z?xMGyoX|AP9PZCAlz)DR7PE7q5I0ON99<vmf}u4kr^6Yt)fJ{54vto}WaJ*C$?C_E
zqCGa8PXaJ+BlYgJ?9(oB{FUFpBAwkqKAIM_UYDcgNOj+{%`ewh6<;k`Sq_f?i<&sy
zyqhoY>oo_xEX_H%`rkJ<>{{EI{p&*;YQWJJharx;ZOiEka8KqHXJ%lVTyzzbM*}Z?
z)B7zW(TE{XEvf`p?$)Mbl!}!?S$;cL8j%_Tb4#v11q#mmX)wAHB&a_(T@uf<BGx%F
z>~m#&i6X}@Hnt|4V97;Ck!9Gdb;nS2b9u{)1g)LmlH-XF6NL}>o}^0=u>yyf`OALP
zla>aNkGOrcMo<6U+56qK)q4?<^#J{>ud34?9Cnxe{zUQ;d8zXA;K+uNxRgMuvKrBC
zOrarQ#?Jm$#0+(YS;>zhEH@~;=4mD*gAw_V>(+)R{~cI<ziTAms91s}4fS4R$%Tct
z?ce6^X024VlTPnyz`%CZqk47|=tdH#{H+%`JFb=HlK;4_88^^}Mttd9!i@pkXi()J
zNB-{F`ExVzpYUJ>Tc)DMZhf`Uv?4uPD~Pq`jw8D~zoRdAZ@c=D|1I0fFAd@rchM$J
zDmO@^A$Q6B#6VKT&tLt4s|{_S<lBke__<Ki$tAD)=&^sOxwd-rA#5PWT-Ti$;#EY6
zwC@=KWRj|2`Zp>}A>!6Eie}Oq@HsBAvxSR_4uyPfn?a@<in}mQ8g~fnj9tVNvsPl#
zEw9#H#N4xoW|Wldqx1$2S^BLxPoV=g9bo<HVjMQC2@w8ZkWw61)W{olWckE5eTsaE
z9`hNOF<MB@9&yFW^oC-5?>(#-UWkD|u;*{3vpa5AyN`n-gJ7Sv7oQzpy|K(#dP-@L
z$gG(ye->L*JoMM05|GLydiIvumQrd0W^GAT-V-dS(sq(q5+D_+^+#Vn?l8puNrA7i
zmh$58dmUo=ag)0YUU@CxWH)R|b7?rP;sC%ANtOCA#Bn;!(|kK*e%4RO8K8t!j51ab
zlVqT2lzW->Vx;bWlq7;L$~5Lt6kz>Gm7MnX<EF^ot$oS^d<0!=ndCC|OYm#I$gk&a
zJ@r0`1V<Uex|Q``6j}bMd!@&xKMQ~S*697jo@HLL?16C$bn}k)Qmaf*!DfQl#@1a<
zoIbm6c$>iy^3kZ%cSknG`Mk#~qTO2bPORe`3(Qji)m1H1LOe1;_7iVwbh7K<oQ;Bz
zK%+*d&pnuhdV*ho-=ckPDM%_i1*vp^{NCQ4g}Xt62WOR_L|?ox)uMcSy<Swuo!FwG
zNr{f~eKO|9k@iHGJYm=A#}O`b*!R%8x$W!Gk77En)zF^t?6CvR&xb|$=556r`Y-;~
z9d_o96FIgrvuL?$<pk$b&`8wD{2S#~?!QjHZRjIoA}~O`8`#mkk&&82pOa_3Em4ZM
z6}(^lxfi=6ZX-O9Z8d_^jphe5tRq>odPrRYYdCo<U^Z}1N>SZYECTJMMMKralEPpm
z!Aj!-Dvx20<IEl6cSH@sy<rgN{G}Qtb=Jk<PIQqG9pX3?RD`Q8rX{h?F+uj)r#ez!
zgw}-rI702{9G(WOG_DfsUXQ9dt~4;5Pu<b81hurx+14w_AeJpAtF&+q;Kvm$_wT)#
z@4C&sW;)`s-Tg6^P8X9Yr_<f^A(3eLJ6eb?Y1s<RLFv+;BbVaW($yLEdhRH_1uwcq
z#}?24hAF~KpX(yPuVe2EpMTN`wW>Mrw=6e+rWGfKx)$DwvtW3{iaMKbl>0HXb*GD>
zkC@no{(n%a|Gy0S|N8fzhWgv8{`_&|*KfOo|189#j;+4#Hjg{uy8|xy4;L=kHWBQQ
zoG@~DeE3NI&HuB5xp)wM{PQ@LGr)VM;M0R3>3j^)-8D&8$g>7zJaoH53^;@L`7dt^
zYwP`A-d67a;%(uJ&%iW?mnfCl%aXjZtTNf4*7HGT@}YwM6bwJE3u}q5UWqF^1!jvf
z*jt)zq)YNL0$fl%IPn}<dVnR#aO~Lj4i$XvDy1X#cDF65=%>*K8NpK7zA~Mhs4hLR
znjMJXVrq&jZWR^^%<gSr2tYkkkTIOF&Bs3)Yy3KL*S+Vdbaf5l?!DrNl(WMc_paz%
z(N)q_S}+lg>hOn#7=ySM6eqZ+EyQWL2!}@aO%a8|(l!<Y4tnCz?HI7qRb1_da$W0T
z12!xf!Z2JpVc|OFMp}8s<^cWW_ly7eKf9>^<G)XV-xDQc6i+snY(>_QQ5xluKf7)T
z>@8If4Qv>`xuIXt{Yce3Qz|as-ZkPwNX4IeV>OMRs(yc6sU%E|Jj`M6s4YMBY@Ge}
z933fd-LtU{+W3k*Nr-fymRtTWOc4NPpW><RI|%%LF-1I4)VK#zeB#!>E+783JYoDX
zlV?jqM;yrU4g!`CJ}#J{T6#{TQb=46Rd2T$Mg??2OdTYLE)FHWQ4G`l-$j}KkLUY8
zZTQJ&d-=zaCkx5><KmmV`d{UL96936h~zA!t$F>_z*LB3s+)sv21i4np5}!k?UeK^
z>c833&!{JBcE9_*QnG2%?0yM3t65+0=vT*gTgJ4+b^+8S)C&<m!v6OKsI^e{tsFx&
zNke+HcI>Yw*65GDO5>J)mIEapy?O3#<#X1mr^f^lTduBR5~pkYdiLiwLC;*xt0Jk5
z)oz!$x$_4(*N$c_<}RH@E7nCAy@IXCd$5GP_wQ#gdtO|thwWWqkM+G9`LI1)8S}i*
zk;-t=#i_B8SLlhB_angD_teX%lvn&0`TY6*a8)ms@Hfcc`qknq@ymExveW$^M`X0N
ztN)esAMr1<<B|QgEEA^ZD_EDuT}->^sqzR4X0763Eh!1c6<e&aR|U{}Q7`+?%q>dl
zsMMCb^;UmEBNoS=oqg>o*a6w>wC`R=Vi?2FT0xj!kbD1W^Rwy7g57%kBvECjjg7H>
zHJ=LEqa|$@g~{*$3}o0iP^PB1DQ%SAsohAw3}DzPD~)O5b1=O6hyY9>r6wryZx1Ti
zp;H@l>-98l-oeDd!iJ@Nk#=A4tp2BeDH?~&K5aM{Tay4s_r5KlmK->a-1=TGJG^=m
z5AfAzxg52>bf%?Gd%v{?d48{HZrtJbdVA#s#8nP-vBUAci10O}F`(1jf^>dgVPiaC
zr;ZLdkhl~WW|R2SLMc+-vjgwcKEN#o^1|#!^faV&N;K|g+D`qFk-ioBXl`+CW;{Dd
z%dZmviu4idlO^@X=EwCk=ors_(!-IJ0gT8%Thr2%<8liyJJnN{2ZK+P#L^+>*f>#i
zFin~X7}hoU{nE3udAXZ2v+gegKyNPa9wPrp$1Q#;NT=o_byFT#dmQ5`xhjB+5yF?#
zA!!d)L%htO9F$(1@p7Hy+rz9MM}QE6FtA=Iw|67Ib^z@Fum$&*om*k&6-Q_=O?59&
zS$>Pt6z4rps79-g8piNrlC0f`N#N$u!w?H7C$m##sc@0i0l#Q04U)D?<_mx{fK4)t
zg+MMBJZkUAXBw{kQyO-D3p`osbPDu+=#}E#DYxhD*6O+W?l;ql#7x0)o$$_p$AvGy
zj<FqAMTV&-#u7(J@Nd%6{OLFAcLumpUd#p(3$a9YcH|NRJ2%I+Q!y66#3hXWHn88(
zK^g?`bpSbrUt(F}XKcihmw8V69Lq0JNyo?7%0NHwmh<=Jt#ZsT$YB6UDlGePq~XwD
zKKny&%SaL<xYmK0tq>p8cH1V+>l@d0T%#{ox@XU<qOVNp`;K+x0PSvyXBIo(GH+Mi
zB(kroOcRnV)xgG<nNREvDx%){cXLpsafaJl9}TL~bxe}k(Q8Od(WkKP{gz;=tN!6i
zeocYT(+=g)rd_d}^2_*{@Kbfy85l?c{aEdD{=<7~G5WG`+xtz79Lclp2zI(i?k-g}
zQXe>RhvPLOD9Aaq+12%7urY(6a!H-_VcntHp00m3_&gNZW?+B?yg=2KGD~X2)b1K`
zDwjF@*245^Jdk415zYi6$u+pbZx#%kadPz8$OY7epg~!qw-ou?<Cp9K;S1;9zRJXc
z%fauk^125;jDo4)w-EX<nev}9V92EoVza5HekOng2Jkq>v;WQZI$MMe&~iW?$*@8y
zoR?r{P6q0(KEJa$rIc`M6zZcKG&F)y8A{s?ZH~>3+SHX=WJArgj04LXPFfv4X&&A`
zKB~PH!}@Uqe{nSVKq=bb(2H|_yKOJc^~aG(E$1IcDgwye12pER&IwOE<PScimBLf@
zmHIsaefw$y&$1{;wPwX>S=MFMWpmZfrTbb@6-BU6buoi`!CFXJgc;D48l#We)mGYm
z+(6x-V>rb)w3z%&Gbn-1J6Iet#O@(oWE2vk1f!(*z!sDW?kp|CI*qj)=vGp3YjgdW
z0}l(ZWUwvM6ih9@g(-%EBgztRG%v-`hHyaYJ@eG0Z|8gBu6bq>S3J5hL*EQ$(xd+s
z>zyrkrvB=UFJC0_C$ob8%<Y)03ba4V#3ZH5Zf(+)TYjcuM7AJL%@8%|4lvkc#vg8K
zF7J?z!xp8hfM^A>hUt|Ru!Z6ZXi;zLWqT+cMVl2@Jv~Ahp4E$Khm9F(tckhuFcM~3
za(#I&Emhb3r=G1QMd-H|Ka<mPF89QC)z38$&}$d;p8C0Xrr%*?J}Z)Q%NVeEs{eZK
z?iUX6VxIadtcU$BdHHgcU^ROrdx>Z1!6AbV+UgJ{{0g47-r7lB$aF!&d7^I`QGjgV
z{lqX(UBKNC*-L2ywdKCQkI)U=I33Su#bGiS?uw%z2hQm$Qg^Ys*$8dmejnya!;rIC
z<iNXFG`l6~*ooSQ{^sDtyeSfYKHnv@@tM}pw&oiq(0b)wRO#coxRS>s6tfzQ8;vNw
z{2xs5;S|;Q?K-dMM>Zd(VOL9w(kc<oV8_AI$fJjdWr`F#J=uK$mIGk2gVR^a**J{#
z#wlL9iD?&cNb&u4O%1nqR!L<X$kuQe=-$RIjn(Z9B^hR4e@21hn*W*$k<Y5>Yi_$E
z<8~VVDMAYI)`T5&_49?F^-&JCUqe5_&rW#(q4#v}Mdw847si<9)jZlE&qPDB^(+af
zS>?VSrjZ(s=TIIT5k#JKkRIN1WH`^J$7VrqEv9|-6ZFHyfDID6cW&t$0P+yE-vO3{
z*2-^2KvfzxEe1Btl6E3{7n}%0Pxj$LV7b<z3ZA@H!LI#|Z(LZhyZ&Zc)}p_n>_MP*
z`-0Vov)O&6t8a&xm&Cp^DNn;X+m0h7B7Yol{MvGB;A4N*mBv#?7e>RDnM2#yl1wW{
zfp$kYr>MB;Ls{lHe2~Z0S2?JdBEtl2=G&YJ{XqK0VJAuva&Xusv866LQzqs<84MZ>
zc98`V)#>X2{Kv#~y)nP-{`#t2{LTSv=oK}o!lP&Jmfxk3b#&6yq%e?KgU<-TdtQc4
zRkjJdnBo%TX?0)lkV#S}-CEVS+ni!zhsFpWlYy>#v-{nm{6$N$&0aM<-j&$22djf8
zb>H!LFxy?!ULJ4L=YP5}v!1{D_RYtI*F8^D-9RrkWy4tA1q!Jg^L?Uz?&fKLspLAO
ze%bu9LNp%71IG~F3qX!T(xBSRu(O?fl|4!bmkMjzW;%9!-yTf5;sA<2jqJ3Swg^3V
zNK`r0J5N!=BsIe@s@WO0m7D&}{Q_uF6h$n`-Ffx({>8=ce8U01T5-bQ??;A$Oq=M}
z7SN}0pU_|Yta{<aFEaK{M<m}}IE5}<roP`mxYg8G!Q_)a?n9hbnNpzh7mQI4y9Q_;
z;*7yLA5cQaKqTm*cKKh(F!n~hBQ0CX30T62Yk~O=hS{fx(Yx9r?wsKDg$*y?^k*h&
z5JF7&=+bCk(RWK}FZ@}(lUxmhDqm-Zc$G6y9i8A`U!_+`623}zd-&Z*v5Aho0Rdc)
z{kyER!1PSm%9Nt_?t_JO^o1t2B|W#E5L&Fa%sfODp|R{SJk2RLGih`IjFhm!e1YK5
zIX)S&>2a{<3LE+5XRmjR!B^GXK+*W#4oVHH+_0=!jSLgH)Y<qsgFMAScv8?(S{gA}
zY#IqcFO+$Oc6oM3H?5q$9BRZQ2A$;~C}#uR;g^xhd6$E9@4u=55R@w(+GGU!pBFFC
zP@dy>nB!&y>b01Z;0!b#DXzv9rD1T;lH<_gn!+?pP+3r6TG}E_6O@!AZP76&pr(mG
zr8+)7@isP~wN33s#c9n|)IVOS3+s*2zSP}=9v{P2y$pW6n`=OtiUOlr=WWH-nPhVx
zUIhxI>GoMil;q%fKl&kxa9x`dG8)?CGkKuY`pWaSUEu7Fb?ywDjORBK2T91;U@$E1
zXk{hKshq&!AaCN)lVWXZPDDoO6;&UU_MTKuG_FFa*$ep>=StKn?wd3{SyukirH7gN
z3+gTdb-A6ZZuLPl1$`VYFs)K~Iu6?=uG$P(-!r`Lc)NVzKTYU~-muynE3KGUD;eJ-
zd>k1QkH}fAkdj?UyN*Z17eB~v5{pCc0@`d=IXlypk;BF^EQr2HLh)DE-mHBt+4lqe
zLmsf1HW0qDIziPH%S&-FTX}D0GI45~v<~AuD?L{df)%z$cBZ})Fuc5gU<VOc^S#JZ
zJ(lfW^T7Xonk4kmg$!4Fu4ZVPqNTB4|3H%`z=%DXBAcn~XA(R)PMlj45Rw=Oce-!*
z5|ZG7iNWzWab<eZE11MQdx*B9UF+p~qf)>)hJOw$3D)~O&jjnvgY{gfQc@d*Whyh8
zk6QYkqD&=u?aGz<!?@z;*$2sl`@9bL!`T-tcx%wB?5?}_Vp0j_7@1<<<QOdv3k10S
zo1U(7V^F2r2QlxNbB=+qN$xA!CQpkYVIs?=ta!_+C)$e3CGOlIj^w3?-u@p)zB%oj
z{{9Q;@Lwf=BX?B}YsSU7l4-dgM}8dny&<)`;PAVL)?TLh_tKY3$mG3i|1|w4v|gDr
zg^ct^UF<jF)}I)_`E->VoF3fV#u@-)5U&^~*yCs!eLZKezw!yxack@I?`ruAG{r8h
zqO#0GGq}0s`Z;0w+3mHXnsqYT5e=cpI|Fp2To;8xR^8-}0(Sc9r9Ikv(>KHGZ|a&l
z0b4@$ZQRE<b=B>WGHX#&dv_$_1^by2`l10iE&pq{j=%wgm|d7ipJl8&J4;4np$jxT
zDbn`fXt)Okm5(omI}CJMa!o!Bowy(97$=qA#cfTQ)wt2zVjz{{iWk1Dadj(2*+H<e
zyii5rgLi&aHyqg>uvF<e1P0o~{xmm542Y@R)jd$JC-D_8FO4VM*gWP&R2vvapk3Wn
z%nlQyI%VtMQ_QCNM)f@_BUD~w2zZi7B1s@pdl^?;Lm}lioW8hkLI#$~E0=z#^y!w_
z)P_-Tu9yKk0Gsjz4Iiw58vZyk=7t0?O*)*lO<1j5Q^Rl8<E7D6Y|pO^dW`0k#(Vy!
z&vz1nef`LVtg&!y^KTvoS3c_MbNd^%XJ_tb3>s1?7p__dZOIYy1-kuz&F44ASs!Mh
z%oJi`OXm!pl>P27i%c!{)HrZ=ZttU2`SY8MlRADJaVIXv9j1VdG#5U6Ooxx-fEgO9
z+lr>eUu%a;fw#`hPNC(yaAX2>aFQ2ivSc&}?9;9za<EEMrd)RAcDT;Xud@;F;JhMC
zNfM2Ejz{%yCx+S#8|a;8c#C%)efiLo8ul&mmkW5I!H*;3WiO|<!jAtqve~Mnk$f+E
zn_eV#;(DCCZlw0uWO#w64AbuxE{4a@PE(2nUiyF2BSubxWNyAYX<-!N^?SsXoLwDs
zl8nE+E&To+7~r22S6>Z3_Cw0#$K;Bwjy-wb?t|f(=cEXp+iBG9(JHJ%T8_5i_&3Fg
z41pI&9JAdbgIh9Btu1{Jfh}x-`z`L;dNxcr4ckpP+Hz2iU6~lxE@=F^I*o3Kl)e6O
zNM!hJAq|^e5!jFYY27i|2Zdm<m0;Pf9d}*`6G%_8PS-Xapzn^FF3uP}Ppe6vmn7=j
zhs%AWJbql5@u7+}{xR(ZO!0TS%No~;oJ6KR55;wZNWRct2%%OUz6||PM^?Yj=a~I_
zWq0ZTRb;OX;;9^6(|3MOS%c76z+1Ou$@F>xsJVH8X6Q=SvvDViOx#zJ9dbV}DB^v-
zSSAVDJ4cj@Rp_^5gNRd93yW}F;5nZ~pDj*0k-cyWccXj`$qBF=HW`%hdh_a0#Zl|5
z<>%=T{?Ca&sbt%4K|U-ed6xqATGD#2xxa!;INU}|<5l}6PMmcM%AbTMoZECG$+g{A
zPMejk6?-d_O?hoM^?48nfaUn71btq!;uMj0X_9=)&FQEaV)yz%QGb6ARTtu{O2jO4
zhROT??i<EcT<@vF2Fcb}a)kTzNZWo!TDeih6p+6<tUJweuP9SeQqo@^*iT{|<(T(1
z&-^%ICQ|oK9swQhC!hOqWFhI&k0a~NY*ghx;w1|3$$YK3`_OGv61eba?Agj`$uUPi
zTjI2phjqY2<(*rWw9B26m(q@I-Lv0m4!W}Ol500I1ahc5kX3t8R8gX^Zo^Wv2gg~Z
z=GxeGRZ~2=+c1d(K}!N{IpRTOUm`S5`_iwyV?IgCg;7^J_)+(#6ymu#dS9%(8SjEk
z{;iRi=1B99;lKa++QEJiQ|u|&0ylcKnXP9k6;ASxvyJ%$*4a$-z;+VSa=?J^oh6|5
zrv<r@GoaddSvbqAFR&;;OHvJ!nMM<L#}t;*@b*&Xn?!FO#UfGsat-M2vVXJS|Kmt-
z*7U(uJ>wl+?3G~i`-^)^kK48jLbdks)QP>}^}|m@@w$9qb(sAhCNJ)DX)BKyy!HmJ
zn{t$nJ&$87fM`WkQaao3T5wm!^mX7@z3ciGYItfx8-05!PKU;qO!DpvECMQ>@506`
zR7hCp`YF0;qGE!_#$0HO_j2~k0DrRhmy!et`L3=sPkQB4rL4@@NqWTkC!M7!`AA4_
zM+G{rlt&9T%iuIXddZ#&_-)~8ytCT|DptS<GxpqIy6>8{G_TM-(VmVzeoO;RaCG>U
zF;1z0@mW-^u1^V@0K!3wRWSlb5PwskB7b<@{Zdbm_NlAxe-tEVsE%lOv>dcpdKUPN
zRX@zAy*5Yn|IIZp3!#72vb;zJdy!O;skh*KVUN8HS!*iLO#WN8!YX?gnnYfr%=r_H
zov>%OA*B>2asV>O1pzc)4K^yvtR7T8<cw)~5_#y!;xgOH=uX;e9vS<EiyfWX3EBBD
zt2xvbAe$TtTc2a}g|g7+$w8B{J^{OZbzkdd*Rs(NKqr_PtcU`PJA!oM(gSc=c>ybp
z=(S*<$*z{5lOyZG^@ok$YsQ=ApKg{s)&4c?Y;SM~UVX+8<{_VMT&RDp;ASTLwYA&#
zoKIUlp_-(=)v(hBb}5Q4(aG)fm8Z6))NN9lyPYjdn=pk{>R>sMtNeY$)g^mLTT7WY
zPjgV^Pm?ie^weCRc}IJ&1JL+ip7q!@1vHuvgoJJ%ee;HIP`3us1twT(0%`ec_NGfT
zs6bYoS6Q@7h2wmnS6(Z`=1osjQMH86#M)hrP5Y%|%Sk_uL~BVTZx?(1I5Jr3ol2K)
z69hj_p6)YWIzL!>EQgWdLk}IxG~u+qUSN;oWzW_f7%`i(&2{<<9SdB~74rvIGJpM2
zfmV1D6s~<EN8_)@Z$2bU1X;iOaRhv!pcKA7VJsUNgKhzwOs#)h%Xcr2gdUQH--RuH
z4$6Q3{s6>%L3WD9=B7RMfl;7~7SlZxS+(>~cRnqGCyi)sB|Lo-L@jQ`M&r+)*FT5`
zvDK&Az%+>CFiWP^K0=THc8nYbdRXb%o$DCzIZ0bH4&nYOyxsl)@E+({8}v*5pU7vB
zv4z55LPyfnvp1&R;Mg>OR!9BqS(mYiZdlhbON$GeS6>-v=18xITga!>;`GAyLn$=D
z{3=iIUuoarQKuB5J$NO?En~V;Ue|6c%A*FOikhIhjH`9^E=*bK?KwCD>`XmVNTGtQ
zO70*NMyXjwYE8A%!?e!&6wv}Ig30l+s?*7nRQS9$L`vEoA|1}F6eB49v3p+DqWbYe
zrc7Ri>$Oz3DnBF1m-jyuWs$c4B_2u+25f)nlLX8679Pq*uf1xDcmMsD<v%-NyFj{X
zi*ZIsFnVq9V5F{D2{0rH;28f!d(G^yBaE+(W8pM1*d~hMOy0}6#7>XH)Zhlk_GLDH
z9B~unG2f2A^>ZHBJ&-wY370b2WTts|GM8VJ?K`B|(F(IJih@q*l1e0+-UY7leE6>q
zZgURj!i3u=-|A*7V^VMaqjyKU01Kun4Bq^LQr#`R%k+3IQJ?j(?BP7)iCf{lBE@Xg
z0>^S|3Xw1~wEU~kaOzOSlWoWe=qQ&E3i@->lE>TzMx#Kz4GqS|<to-6ST$r+E}Ben
zLg!eT_(ItjMn(ya1VR-*&k4|Z3@^wg>=?{K^ZsU5kC(n#8Q%z`+JT&E=e=RKz@P1t
z)&Q&sUhRR|fs|J**>M~qGn>^M;>QYV^@O$U9Tlj38y~&&Ipx84@+Z;ZlA`X9Bdev2
zZC|H+gu>ka3aEV#zb_^ai_>8v0cW=JmMaMwYN0&yc8|~?vmiIwET<ZYdpjd<i!NNs
z$oPNQd()_-^ZtLhxn{1q+Q-~-zs5$(JvS`3X3C{fBtzWMazO>bT)-{uYMgSxB{NgO
zWpY6Q#S9S<$*J5G4G}j)M_j=zbICNF|M&Vmx$kp7z0bMtCqK^N0X+C}_?#~vKJV8W
zZ`@i#nOYaQdB}oeFyN>Raw#9|CM|M>nX2y=7M4XIK1lLpNs>+3ewQ;px9^L?eG1E(
zR7*F|CFdZ7BQq&7sn1q2=2s4i>MHjf)g&s6VI8#D;toyMw7-q_X&YizKT@08aQO7f
zm%8MvX(y*q1A%>O&BmM71j?#nDC4hzaXLIEQ9@c(G}xRU#WZJcIeA;V&<L%VPiq8Z
zIwj|`w_e)cM^a2#o6@whOuidLU6Q`jG|0Vq_@|B9-$$HP&mKCVc0W8SH#*qZpZ)tt
z=t}^}8f-hGXi*t)P$}`>H=B7hSAsS!wz>3g@&2jtB!j2~<MLJ1uchWOx3ssYr;R(q
zLwfGsMcndbeV*&_zS*CSsY&Z+JLDudNxhWYDk0gduTZXV%5XQ`&e7D^z8F1D5zl(3
zmu@OV)=n^-2TRQC2GyOZ4YjU0Z+7yDksW5M9ED_4SOU%lutxyBU@+)LQFJ1k(hX7%
zqoeelvW0cA+BseXx91n!q?uO64fW3}j7G02!yj7sxNe%P;rM&Qxr#N1Cxhn#lhgk`
zqPXV8hRK5u*3<hTFKscbSK_&e?1yA_i1vZJUF6}o+t>cS@7C}!zJRIh#i+A$jKt&0
z;X$o4^`&%ALd(#;HuvH6MHTeI2*1WD%kl|m7W4CqR26Tq-t~{=CU@QP%*fgf4h})V
zeT5r_1WGvtRZb?gXzP|!5CIe<DIS=!t=ik$ot<OT*VTkh#=w9Lyl{4gG+i7}P5t^J
z;AehBgO1@X@Y>WZtvOA6U8Vg_T3RZ@pQMY3=BOi!7D8M$3=tth)5Lp!A5r?|%~|3f
z3INIY+JxfaX07F>x-KYP<(vOCltp}7D`$FtJ9OgO^_(vwyNZ7wp-&w<(9pTJTXnd|
zrXOyJ{yx&_gN!H|IG~Em4?Z*SyC^e+|4`E5J{mxH-_W}MKGF?R82k@#acnhkk~r;g
zSm154tEV*Ain>CWSNaed;VNb~SDO2MzPm?tHHv;AW8!8P`Cjq{cIHD>RW@lTpY%gj
zmqe>tAr-{rzrO}cIJ$RHqWn)RjIr)IlLea-QYaS;skX9Gi$+R4)Ol-i53RUKU$cjr
z4S*?_Ep7}ScIC@^$c`hv^rxT&JBK856i;EXRVbZw*XhpCPHgukpuN(wuLY2+dl70W
zXU`FxXW4de)cH-QG8>wbm;7L^x{UEs<&x9!i{VXuFv1*yG;!fZwtD6Jpi=LzI*VCR
zzP20($JK?ya*of|jg?yNJM-j*p+$^G1y*384X_mc7QYUOb!iD?5~yx(Pp7)DJdGqU
zmd-X&mp0E&<xW)9+S)ic@+&T8U*K?o{ro~>wCuIoOX;O?3y+YAJEiT8X=s39i;`Uk
zxfWz}?$PHF;@?M3Z`UOH{6Az8$n}cFgOZK!4_6-pzeIs3@vI!t?#P^MVgms{6wMmw
zzAe{zGMw8Ko{kP+@@ceQb1wOn2l_Ezzek&OPDl9r&`D`=(oOUasJrp|{(s3&e2G}1
zZuHqRz#i0zapI1^eJPznjR}J`G3R?pItu|rcRHmF4xJ!qQb+-XMOi+gt<Dn3cMKCs
zVY&Iq1baBN-5#2|wW@ict%=N#b8HGp&%T=KAJ}np=*s7*<p^mA!F%G2foGYu*1ek_
z9kn{LE1S1K#tNIK1Azvu$~~~0Wjd<qK|xQjZ23l|(EGIC<nBM=9P*xSYq;R^;!;zr
zyq_i?vE5A}7)_2)cpzJI&|-DQ!Va47pr&xug2fyJ%rD5!PZ^A}H`VI0FgC%SuH^ui
z@r9t%ky+<HH>eI*pb+V@e;+y4_JVy{{Q{^MATIxZd3+Z#KRxsZ(7};-;Qr@zpcy5a
z=6z5n_mxH*8XCLvf#2&K`DG<ax3_>Fc69$z*-w?7g<>^mSXS6+yN!!~R2hU`Fh&0Y
zGD;o!57K37YZlbGZPL=xGO`KM>)oNEUkRJ$9~HwMSazxZ&UjUAMbhh$z6^Sse(r0O
z>-p1P|KEM%@A}(hr%V!_&<*C)a{~7+rXmEpXUsE#9<0&cB(|_#x&#mNR)2O(Hal<h
zy6;Z0>HPu~?_seM@%Ma1y9D}D$)43tS!GCyHza+JnQoPeMo29?+ejW&VJ(Kl5<DVH
z!?x8bHmtk1mw`idmlv|&rgrwG#&Vkw+}qvMS=)*?OpwdtoL4i8-VIsl$$+rbpAi-E
zKXHv21_Eb$a|p0G#&3d0o~|uJ%&Q$5nI6h`bT*S5cSes$dg|+_eXH8~IYEm&!qGK>
zsXILr{5)ncZd*5%3bs!7#Sdu6fovV0`dwJBKWnmCD~-3hh+Ss;9aiMn<7yF;Jag5|
z6L0E#*fLQWM%)}PIgfp}4qxXT1m_s60?solL5c{qt`x~`(Uuo2fU*kBN8ouF3V9Vc
zT=GrTs(-(Z%pq6{F$G9Oxup2u=ty7s1h!;;?B~7&Nt`Kc=lO=}Zyz*r1OG`b=<ma*
zE>cIeO{0pCx0SHRlNv3ZZY8KSJhBhTU~dM6uWmdRd!&weS`8vsS4W?u7UVQ98p+B=
zhMiA%YYTE4=j+k0g7g<$k>4h0?C7Wrk_ZcJA5ao-{%v;p0&IATIHAyh+bOC+gnhvX
zSx*$!eBbx5vfzD59%5RClws~Fkg)+QAdnaQd%}Ev@#2XOo1xSu8on@K>=3H_7M@6R
z`mOU<=(+L2r~-}JY?ch+<x78`Ggo>`u4h%%9nIGuBv$;=#$bZPLM4xtxp_1<*5hN8
zNu;>BA@|YEVH30cQNVu@#!|pcc2VY=Szz-Hd8oh{*6$GYXCJfX-b{#))MBt84h-hm
zKLp3P<{<#7u!l3(#*Oc5@9W%UYz;Y=;+w)XS|w_{Wxq;{NLbXGa7xro4We6o=VYQ<
zSGZ;Vb_m{mSL}^-CT8WO-|jh~>3z$$o8vHa&8`5S@HU@>9C>)>(}JJ`3VA<__Gag>
zIKk}r=8?~#%;;(-Lc$Vh1?mrmLfjy-HdX)moTS;b2D1N_(0Ce2sr6$cC}dVjM>G;h
zm~rJE#uZf;F1D#&eCpm7w&ESHcp^WjX?BDYM3VLScgbC)46?(JTV-BS!1L0_O5{rt
z4O0_C-n@&geC+l~phSfa?i+9VR6e)KlNYrqXtL6>X@!)8>a@k25QE=nE3ODOAE4$a
zw0_6p|7r%**W$8my0tnU4s}kL)Rw7sUReF&`>HK>8v3!p(0THBq?i+VU!BD=dR6>5
zLg%I8PlhcopD|B2);!BMDtDA-bM~~Yy*`@RguuHuu6<yS?HT6RNJm%3Iwz-ER!c0-
z>;U#y6339BIp3=<uvf#{+eAZPuBv_h8#bDt9fY5un(`UCy_c6lq=ZH-v@sJT&x6F2
z<~CElvq09`uD=~3x2e-QY%?HvTY67v;L+tZ1I>G5$>74<wJCnH(cyuCW^_zhicAmO
z4KeS8JXg`>o2qi(u=$B)Bj~nkd@6ls!DU0l_6ePt^K9=IS>f%3g)A@QsK|ylST01x
zxgj@KflON&qsQtJ3j##)zQh=T92Dkjx0vMw$})8s9LDhAeF@15e(1iK$Wi$DCcQSe
z92RW0EdaAfrR21Qz@oIZr%69=I<PSUjyvB#m-aCWRar)ENqHK!rGmfeX(bTInp!_+
zs%C}U8&+9dusZSm$jyA6=1jI$xS7;5tP6bn+RfHq4-{;5cH>_&oBlpB{P5P8(q3}c
zgjHVb;>`WPwJZJoAB5f@3_TB@YO&Ra3%sZ~Add?NFTe1{Gx!#Iq-=8Lg^FE3P>ZsB
zn_n^haT~A5fIA%nGjOYMP>i5|h3eVp%a-#a0P1bFya!*_!x<uD`W|lnAx=5))!wgU
z<+d<|zmNPx<3GH<Y(#W){h^S~$Z?X1R!jMA`JY|C4Of3mx#9G3-&xDSK|EB-TVy`j
zKY6XgNCp&=>g|V9C)TWkKaf@@ZEoDb3+~^alg7&hPPFB_*eDut@mLReU%dM=i!W=&
zciTaL6FumEgtn9dNfGn=mVWZzgiYaG+opMiO9W~8<fs`>9wsBB2I#Gd%h?frp^g|J
zIHSaxhfF#*VvQKQvz}*v_TMeJPrJ!f?49}kwW5A^c<<}T=g#(j@D;3RTQ5_Vwr)B2
zOpCuP^>0|bn?7yvD7|GcjVKG_gK!0wO^6bb=cL@?{CH;Z!w6WGyL$;r-4GiD^{~LW
zLY!gKr9Y*Ern6hr=SR&%CsphO4WG@eb1Eh{$*orz7+S`%lx8qRVP#*M1w2AUzrS2V
zJtV-I;HL%|1E1@)`;f2CzxX^q+bDA$9}z?>eoj;g`?0JkK`R=0fB$8bWmFI-?|$l?
zak2pPB+TPCTRG&^mEDTQZ+aiMZLAywu^5MnoOH4jh$E~P3#vmWvKdjega@mBJp_yw
zALCUan4lz~d$nsaEJ-FWrIT0op}!2r9);t_uk2wJP^>_8bjNb;Bi&8|W4lQT;cCV@
z-)$SFL6?=8PIGJ;-u;4J#}byGF2RmwOBa+-MW%T+Di^w^+JpIj>W9d;^p-i3bZ7|A
z+WD?3oN98p)$c_n%08s0cg1{WsZDZrgiPyqpXiahb<Gw#y`yGz%vd`{9w~IkTxxzQ
zePnyp;@1X{E|<R&Xy?^N7PiHXz)1*hi?|d;2rG_gv_|n_pZ@h;xSL>0PR}iIfw`~G
zji!S__YAeZ_3lX+ILhBgFnD<=L~c+Xrn;fp7th4DYuv)&%VRA^bMFMR7(j@ajmz4#
z%&yck<6lfvw5df&)oXLt>$S`4hOk?#pA2>0n1t#2YbvvrzWjw_No2$b;>=UgNKD-^
zx8bc-e^*A$vKzwVF=i^~i6447R_7i4(FNKnJKCN_8s-bS+&XMVcRq{ah(^~r^$77C
z!n5I#a)-MyViHcNB{m=l<rsmzD%4cg!*o#f6Q4`m=>4cRhmr|hv#8vo{^q5w7oQ)$
z|Ds-z=rOs48Ess#C1p&+tVy2X{sr<F#_DJnj2LK#T^MV-xz79hi1=7VM5?LSVSZZ0
z>OojTeOSUV>A{%+hC}R-`E8SB#v|&rW9nUG=6UR1H?id5C<JUejW-I~{ZO%<gxc0t
z$RppM-qZRv2+#M^pIGVwiVjr0*zhD+LicQ=U8J!=*;Ly(j{3Mm9v|5W!6R4)Ndxh&
zj~Pzs4e+Bq>9{Q~^JSb1=aJzLS~u?32Ja3u345PMIAvFYNJG-xlq4mNnuSIs&?G(l
z&{n8?n9ejmZGl9pyjj3;9I*XXg#;YAT;gFInsrSBQ1TBM!beu%6_iPzZ5(DBCNQ%T
z`79^QgoD*P>1C<;M8x5%5Es&j>I7-dmy?s9tEe5}VQ(YjVh{sR^>7%pJ+i;^hw-z@
z=zmS^R0LaM@nPB3VO#w^Cmxw6j~ZA<#VF+#aw^L$cUJ><UY^S}a1k+nE{?Bs)DCRP
zi>`%+xqEJvZYfRM;@MW$r+lj8S8aNp)MHRqff};EFlv2z8}ehF1m>PS9LsD4b8>@1
zBYT8&iAF7nH8>n}bnD=}{^l{n+^H@g-3udRpPsZ<rZ2OK+|AvWnJ)I3$r(xDUg@&5
z4Gb4+Y8I{cf6$M}$w<x&kCu6FvQm9=&>Tp@&K>B|eBbUveyq(|GS8s9kx|`<(buzQ
zK&4g3vuCsOinDJ1jKd$}CGt=)dr*wpAQRU#_YpMZGQw-X)>8L>o2rQF1ZUPRy(0}=
zldgml|7+D9^{*vmSXxsEY^$vOwojm=3P&-?XAoGWuO7x|UMh@$WWw2gP00FD54)It
zM-6)a=Hf}CfN%G<tdbP=@U9t>1whtCz{0b=oF5tKDy>Xd<n}N#`|g}nYiRJn*u!q%
zAI@u4l8qX{%Ka~Iv66I>PJJl@vIpo4H31vfInQj}G%wcI1s!nQMRC}rCbRu7wBY17
z?VY4Bj!Ol`Ac0H^pvGE?FI1PM-isrW;vOeL>D7|)F;kK^poEXs?V=buz+!7i27AQ(
zewf`j1u2~^ESxP}zyEf?(ECDz>CpS$Pii-Y{q~Zx_ESD4g#9~yaOuLw9BlmTQpS6B
z-R1{gO!C~CE&J7^3A=(SX&8B!oN<`C|6z#vja@=+X_^`b$lt$@^rQ{`edH2aT3}mp
z==SKvHTyjc>7;XE_0sg*n&Xg7EOl{oY&*%<gWn@QmTno5I!y_@!{4L>%wOjt%AT6L
z&bI$A#aCF%Hc7o#x68m46%}PG9)mLNgY>a3`2+cB8{6{9fM(F6YIK<3rIn9yCP!{l
z#E90!qjd`Im0DSLoVJiRH~i`I3)0If+wnK~=iI|~IXX$Qu_5J674*hc0c&S%fz|q&
zkvXXc$YiQ8nYYbQdrA0VU^P2BDJYKnD-6|-qL+rpZudhV*23@GZLYPSR4zHV2I0P|
zc4lWGR}ioO7VD~+AFYTfMbUIJ0{T1Pz7SJcbwS=+Tay;WhRL>nwVRZHhU;$R5@(~M
zZss*;-1v0CgMNaC#itu6oz72Ovz<-5_^f6rvON5&j@{dL(R)S-vJnxt(Hzzng^?*0
zAM`9b38iA++cl}cNG$jqF@KL=0{#}yfBhNVK|u?6ZoU-4-na6wwOxz#3~#?bxqrSN
z<VSB(2Rm5#m{!XM2`K|DB0|B*EvvEuQ`IhK=g6*PFI12<#4I9lBJzb4WRSL~oODuw
zTY3Mk?wb+RMVBI78+k3sm}9I)(S3PiIf2mixp1KTR~xl?m7s}dRe^sWxwY<)#9d;y
zHX0PQ#GZM=GR3aaQPE@lA1@6Xh{roP6lsE6pkJ1whVsgs{i@;WH~}q)5KJ;V;BZ&9
z&JWX9tiCw+2`*oL^9dagB6tjMZK)MRv0u#2&VhTAz*s(ypKEU{F!q3%<7TrToY{_>
zTr%v%A}VgZNiq<q9=@Ocp=LnX(k{sQwJt!5Gj;2xD7tof;!i^K`B;_Y8#GXFg-%iU
zIaO#H_EMj1miptwFGJDy07i8+Yq=4(R#uIvyMgq5H(I-FzA)LxZ{(`DvGVLByIvsM
zoTw2-fJg?`&05wMNy~Kz+Gx?Xcp<OB9Ip&Y&aYXR;#7|h>u}Qj%t2z&@>kC-o|_-b
z9v<m#fQ7ZkY{0A`*6rvs+}+HuGoQ_37<iS=P1SRapNaiG1NtOurKS>|!{jlwqEXjd
zWye@^a$B;OmiOj}sO2A@n{B=;&7C^3e3<q20&M2(+^AX7JF>%*s^#fgGfPwSUk4g)
z^@ezN%jlKB>Ftrlb1ar^eMJUYLP^}fg_lgHa<*=cUT!dXu;Sw9aFN_ggvQpZ8*pKp
z2+ER#I^wAIb=A5Z7dEvPw}##iofyZK$1CI_2hf0LH^<lNS)if#870W#nShSyGcb;g
zIPuP~<3@Ym!dy(1%fGo*I!7H6m6pEjS-0NUO8eC3ym1Y)&MA5omQV`p0V9nf!~oMT
ziQQt0?N=E!grz*r%`t=1CPz{DJU3T2EC6ik*DcP1vgqc*y)%eh@`8Uui1N0_2C@SW
zxM*tFV6zTk<0#kI&Hf=@X0QOUg@z=zxVCUkd@!H*ji~!Umu~R}5irYA^&DFJx4LOQ
zJfNQ9mJzbxp<mTqg$qwEcY0blq<n*O?73K$YHJgBwRZQPV$%GlYJWi^H`-kY+HxxL
zF3oX)DJR>Z2uF*W_3bmF!^e8k={XedhS(ABYU0QmSj`P=4nAm%1%W;Zcdb&+svm0G
zV!hmWf!b)U)D~ax=*(L(1Ll@2!0A#-b<4&nSF__COBWt~JURv9(!36zik9BK4sL2b
z)th$rv+}V|yNx?eO<hP6?aALe7{g8qLwy#fJ7Oxk&<^`&$ZwzyPxK<b$sQ;lYQ@dW
z;P@Mm8ym+0LnqwVI)_an%|z}4ZSJq|E@ibyOY`0+<Oe=LG{UC%G<Z20PO*XY0h_15
zIWUPo`!t<_J#dVe51pX7IF!H%hW=gk=kzCuV<S;vz@R5ZiFMk*rUK{WVTfAut*298
zDyoFP^OCGbGUAjJjJ;}Nbl#@^8he@r5`@!yYn~`aZ)e~7&un`Q#fzf1#^iY^I#oK@
zR~{P0p~!JIV)sv<_@2WLuy9BSk`Z>5@RI0ae?=*l+|`h45lQZ4qBIS*=Ae-6rX?yN
zgsE!sQaGxQfu9ABM%IeD8pyK;ngZgcornp6#cYj&7AV@Z(4qqZ=}Mu!AM<NDxOwSm
zJ<{RXL@0(n2eOPP^t-=_k^Vg6>@K`jNm#iU@(beqXX}OXZm0uL(C`6LrF-kj@6@8x
z8=pRG`NzEy=cE!%+u=Y`!xKPEFp7~7b54GLr=c0i{m=;zrE?Qqy;K!2J0L<xnuA86
zeQDV>rNCzg*d%t<9s6a3VWV)EPMzjdEzI4Aa}cdp7*VYW3nG{>V;u@uQ@e}>&Jgtn
z1?ZneID&L4KkOxClo@-=pWKr46nRQF@zT-lEIaE#)w=FkcTcP1kQ=!0qx!iD&|LKd
zlQ}bSA+{5I@|w-VU&g<_RbLmazs9Q8h9BGrTwCp-Z_;0vZ{1%&pi4q5?0b6K@moHe
z^W%!IX?_igbvzsq##6jRq6L^5=8%0xVzP`p=@>Q$3%3Ary^t$>3w_~02P5B}ooB~2
zg#)>tC`nN@<zjl_%&Cc#h1{Ld>hZOR|MY(}z|aeR5B_TSlWy;+ewk+XbH>!7#=(KK
zN$O5zxjys0&hFAPK9KYOV*&BT@0A6WiH=)$<LX*LrDLMzWbL>_e0mFL2jAQf85Q~E
z=egNHR00N66{w+!4#Q3mchEt6%!-gkjfR_=!OS*Hf5#f#1anAIr9CY7q!%mpdkejz
z^Sy@mI^;+3$-vZW2TRtKP^i%`8`nV~tw`r>2UET}-o-RKsNW{+ba6&T`F7!Ry)xgl
zI-k+5*BH~~pqIU+_s<=#b{l#z@iJdo4j|jUrqd)AF=`QBAJ;{U<iji~QW?Cw6HvoQ
zSSu+=JTotJ^@aNSu;4|s?3MspQFcuR>uFqjceRnCieUz_JhL|OsvO@#&5fTRve3zB
zUa&0JO)J$7EL{1`;v3&^+_Nd6a%Yo9i6koOWo>c)v}72tbTctk7oTnI9W9sSDxGVw
zdFs1=i~?%2i^Uhj>?dT#Rn2fgYxj9Xg++~MY=`_9e#35~3TzYUykYI!Wee@XwFx9j
z%rFkD=%YC;j15|-_e$w9_JDY(x0TwwJ(T_MY)uwoEIrxxrP8YxkzG14GuH3^wxH6&
z>-bmd{d~(?Xv5Z*r<;bw&&LhyP<H{bbiSo6G1^Pe$kF8X>Wc(eRW*=6*x)8)L|bW(
z7$^d9{^ndc51cMdC!I>Vm{t^i2D}FVE+zmGSeyvRcfr<J7|`18TUcF`wh-{&=hZUX
ztG~WIxsV4ElpuC-GP1v5-?q31<_C_|qK@ZXcGUbe@0sue84>itp)KnAY)4dpf6^(?
zY;V{J?PpblyY;RXkflF10+Fj(T`eZxz@IiRDx#Mx-?rst<>at{0%;9L{%3QQk#I2A
zgczBy+KQ~3Am&3IYblnLjQGMo36Clbr-w~Hy*5~~<g$y#<Hlk<;Qcz~8J{Cu0|-({
zn>oM-MVbSYwA~iJpOokJ#g#PSMY7By^Lg0FV=fP@+pLMM@bAE$-tFy|^rMc;e}or?
zw<CiVPz4Hl&r6Xe8rTy_?PmFxT8GDoP^09CBx1vBlZH?9FQrnU5PH(WkVsEYFKT8y
zw9(MGsH@M1OS-x^MfO4-GGmR~qKA(Wc*tDs^B6`;l-<cxKPFsg=~q@#BVwScM|_!2
zvZA%;*-22j?NKlp0D}Np{c2(MQZg^X49YwmCg6gGEV(B~GES=9;zsuC%8MFnwf1_8
zuBugfH-%rM|6|8)F~HvUdHjy}^2!s?lRPjP@%)-ytJ8s^-=*E)nVs#8v(OEVds0hW
zkI=hwk^ASL4b5b3f5(575`Yh%dI8J~Fq&c_LKzk`SBxGZVL5HV)BncC`mdwd5+nPD
za^#K?k~KD_P!F#^gZ5_S;fQXdQI|O8K=iJifG=wuZ#NiJkJ{NZYx^Z8#MPWIrEuf<
zh-B*O#S29@z{h1)(dRQ~4Ji7GE5%uU!exuJA@W^il+-yZxSzFj?e{5FsjXpLF1^)!
z$n;Yq2NI70nS~qNxJx?lyi%~vi)x7ohGt>b>?YEf#<mm@cF@=ua*ET;V5>ym7-sYn
zEu!?p3I#zTdO#?CREHFnPD*D1FcicsQegbAc3}wNpTf6uj5bK)^#XfFs(j<Hiq7`y
z&nk<bB<Aw?$G5_8Ohf}LR`*?JF|r}@qiaRWoA{j<Z?t=m22&dqTVcBee^ozKVRooR
zv>RmAh9%xOD{Em29$G<dNtN7!rE%VHxj<M!wO3b$gk56a%R%>QqItf6_0~?zsh`bG
zZNaWuMTnOsXds6w+&ne*vtBz9L=xdyvgUhifk!(ZNqK}QDlEzg(NPpR>~B>d_s*7}
zI<{@0g~NYUXU6(n&2G;*C+_0e2It(bja9{`z8YapK95oexw%$kKoEB%o!Xj)f~hX8
z|N8XEb7Q#!O`WlQObu(%u50=Gh+mQ{R3PcCcj-8=XPRgAJ(9@Hc(>)*vQ->8IVgd5
zJ^l+CoCFfLRA{Y<dhm_7omb6oHrD3@0WYK7_Df^Zq#tXmbuwJpQXAkdxaM<wI2`7<
z;_3l0Z8tLw^NnlTXTu}9Vs5DO4!qyjndnYt^c1IR#mW`A%Aa4=?_&~@GMv(zPDDZP
z2-H%Lmd^@&Z<m;XZhN6aD$g#hZOHPLKb#KiQ2WpZ@?%#mjknm!uUa_JX|81noDgUy
zUw;%?yK38M?TT0fTNrM`)3*VH*G|xGyf&~=;V|XB1A3fViHs{!a({#2Khkc!MRyJu
zW97Ozd$VB={|DsN*(^+7H5i<_C}X2JK9zNA#5YH+-;^>ouB8B?sQOu7a;M<7qwWkh
zwk2wx`bU6qr|a9iXD&4_OY3w^Oq9o$B!3xlBpPD-veURJR%NJi{(%Z?gglIGF-5(@
zE5krYHX`g#dT#DAreiN)vg2)bKmsLg=3xX}l>R2>_ICfIb8(+saG;o5?8Q(d)uvN)
zs6>XVBiPh=8{qOMl*w<VEGCX|)O=qLS%=)-hJ4ph{~lwC70shM5hGR#r9)5adUwWO
zG0M!Ux}&zM3EqX5BpB27+4Q?at&sX#{=Z}$EOBSnE~X_C6VazTfJj}m6_7IU$M{Gn
zw_5M+I|1Ox7r4Jl$qF3yVOEOD=Z&K<KwvxHnJ+L7U3sLqm9<?f;=PWkb2_ozs`7O+
z(Gb{Q!ED9IRhXSUk*T*#iP9-7y&@*CO5?&EzFZ-OE_848DGyIpoa1gEJ$kF;-8&_6
zdr@u5{d2x6k1lk}1njeZXxYU-6<(R$cqxP!-fGi%qD2ogB%go8%i+~W?cD(out0Ru
zhgYlptFp*>411h#p!CJqC?G((2g>y`ui-a*j&QK|U{DpqfGV_UDg&ooQ%};Kn<ma4
zyB`+ULQ7U~$JTPLPnCgdZ`bmphy})Pq5|gqT3U1nn#JDa(u-6IB{6dFd_>?RVoUGg
z)uxc&#S<uULZaTO4$t18vEZ&8mvo!d)$-vNS8BRySAMq8GQ91yKtA0x==MZLe<Ucc
z`|ySO?>^H!RQM`quq8HIFj>ehfV2OY4GdLa;N#DKRcPGb`+d62GF#n%MF-WwFq$Hm
z$EaI01Y#lqDP|Dd4+jVS{F85fN~Q41i}_K%V9#wpo#j_PXs=G1$6~P_&YRYbGB#gl
zo>_1^Q1(rxzgCZab&)AUUa(Awh=-g3{aa*xjH+7k<R7SduKNv3PPOv0q9t^BMz!y4
zp;z;Z%ApVddri2s)^<C~CpOH=BN~0cNDK{|@T(|@ec&@p^XX+!=H>$?m|1&lMQsX-
z;jRVq-EOcqm51_8tv|e(p86^5P-9jyuKe^UBFQp;e+SJ^g=HX_SGIhBn(#fKu1FW3
zE4Q_2-QNCb%STSUUaqp7OXh9BuMLAOZHz<CBt+5Wc!PRT{T_McCa6BO*A1ZTv#-*R
zT>{*(ZwfSOmMx89to(03y1|`YszwD)i4yD47-TG$+ZAyXw9UCYIpZR%zCqQF`um7i
zKtOxP?x?<f4s?_+HA)yL4V8Y~IhO7nk!Fr;l<Xzrevp{Wz+xr&p45>8P*i!2ht$xc
z%aNximCo+FZ4AcFzAZIs(P%)OHm~n(+;_#wc;}tfDLOv+OU)meRafX+`fn#cch{m`
zNJ(m&x3eRU5C2hN4N*P~j-hUO0a7yCd`}|Z2)V|kH<8D$S}K6BDo;nKKrjts*}HBs
zp&<m^-kV1153?xW@+0ULu_G#|J1t?;huUFpHGGFwk;=H_80m}PQRD+@7s<du%X4c~
z5dfpCMz=B&65<cJz8mA%@Q4f&J|ge@pvgm@6N51unP>JWPko(PE$N^2OBnSD?^u@w
zIWDimnmli4^V6R<O`6wa9XFFHT;g2f$TR)RJ6Vhv0|5(nIN!!Lt3B{Jw=>PG5z_G?
z-<wRS^?AwE%m#W`=+WjpAXE|s2#YH${b?s$<43cDzmMQ+g8#a=aF16V<lV_~33f+P
z<~{1T-IN?CJhI*1RCd(Z`m1sFm)(y`F5RZzoWAQd8-!v{`1z#1ystiZ&Kh^>itB)M
zsGvF6;@S9Uk1wqWFp-CYnk=h{5J;_7W&I<Ev%iWqB)1xwXVxZ!eModGnA6<W)1L`X
z9;x?paezoC)gTgu5SRBoOa&Wvtimphf=TGrD=jV6PN@4TZpsJ78yl-@&cmiO#7q)M
z)Nk29F3+$@5~|53EWdz_3XsTB&A{~RZ=VOOeiq-aU{4IyrobXIp&<V5`NKaZB%8zI
zf#&qupngi1v2#`*2;Mb6cEkB%QSWRAn(1fvJXtQ^pt7;az11J+FP&l8D1KSUeH{xV
zUt(--cuZVTc|U-A?2y!I<(cr5Ucc0|@n0?npAZ$wVSYC9Fyq1MVafea-%<Gvo1ONl
zU7s&Edqyr_{`|Uj7U|JAQ5Vc5vtWH!n&Y8SMSzxWk5z4w5w;LOqv$SNnJqz0g{Z>+
zF<x5VP^o120#I`W9RB1~kiVeY>L-8Sp`{3Yx<;G3-FTcMP=gs>k@lqAzZGAOM!FhR
z5I8EF?*KzXB{Q|)p}o*&I%IX#azs45Gw-RqT7qmH@8KJ7eIv4D+{;pO1BWN8U76r!
zVkw;>N^)zD*+i{BI&YShRR=9!4D)`8mcFh<?5*712xIV)*~3t~j^)Ntf%AA#>13qY
zhAp~z;@FR(3!+2(D9XVqXMWcvIBWahWQ>w^XtaSh?l<oM8?COTxq<(L_ug#@uR3?S
zy^>}5g8CY@Sb6QS_j*Njdi+}@j1B76=(e?i-Yo%YIOewLr9WvS^*L7+&`=j_=EEE>
zWFY|)jnN^r(M?bGAu#>m<yRA<4;rE_ufV&zyYN?`7;1yNd6GQL6iHbUlh+}#gEG@S
zm5KGUJSMk#`K8Az*3lf~R2VAbM(4uK?gnP$m6%QF^T7Ogl}ni&f_A+hbT5F?GQPH{
z9ftyGS6OIfxkQV<6qlqMibLPOUp>|VZfnE2r&gi9Dvy9{iM8XSdc($29=3Q@`M`tn
z{?E27vnkGF_%d7(S;&CPz>NKk71Td%Q0>Pj4iVG+L*~w{nU=;V+M#(|E~2>?7^9tC
z!T$xL%=|Kt3xx+YrB}1H+V=gYUGgz0*Iyf8%gH?(Cu+BeUtY}?{otVG>z;oUmYC5}
z^jufHz;~u6IeonLi&^S%<4UvC(?(kD7w>+H`jqGK>0$?XbjF2E2?IQTjcXFn=WQG9
zVuIN>vS3Ti$e(^?Yr&hICxg;}F#;Cr7|z^f5AD)W0<t)ZVWD8VBHT;<tG%lYpHsun
zg|U@n?WTZ`(fIRNESEy5AW*v~EeS{fkSa~w{Fi9bpKuDur8lSFiE&x9H^=KV@|T~6
zEd<F$T^<vFFg9NQ=&!sUFtlcng4UA#*qf(q<$}$s>CsNPofF-vdZ2Wu9mkz1<!xMZ
z*f<*?u>{olG^D3Dt~$`4dhML3b+T58ndhzzq?@BCt=`Kyu&9;<-E8DyVdi~SYxKvR
zPBuC8Dv(2(mrI%=ar<`kKHx2MDO`4|5MMy9&o;D69=6+*U75KwA=3)KJes7&FA1Qe
zKMTxn$-=t_U#_A3q-|dXRR9yyQm;xEe}0F%Vdq=y>nI8u+{T+I8;k2xCsLoNUx+X|
zy&b+{1Gzg+Si__cYr=TRnX0Xsy5fv=F+UF@*rodp0;pn)M_0+^K9xsa^yV12sZSRJ
z*O3K1HxkUYAf>Z0_=bMhs8B);p+acZ-_>!h<--DgXS$+=PEG&^zp@l0-X|4JdHm+W
zrsn@{>Vr0tYb82r_I__#Uq~3fN_*#)#_bCx|0}rtr-hF*%1$3O&a~OrXZT%MIh`jy
zLlnFG$=^CyXT+QyO-U&UjD!`XrZ=0l!l4?9y5V6=eHVMYmz$6S(xXiRIWpxGvY4bk
z@-6A389305SHG0NS)OozRZuPBAX_gna0_aC8_O(=9!I`IUVuSXQ#5JXSJO(NY9w|!
zKgo$-{JrZdd9E{%h<&`nG&CRjm%+FoH;Qsev0-_E{i7&jt{Dxn3)9P7ZW!4_fqZ=s
z`5MK<(eU8#l6fE@2?x&X7geyfR~w27@QN1{_4IxbL^alC7Kd`ciUA$_WBDSwVrcoC
z52dsvj9Qo}s9F>XVz^dr40~K1((`=L&g!IO`{n9Fps>3<V-FF!Tu#gd13=KW-G=GE
zT>2jll}<cZVe7O90ihVaAvZfrocb|#OZ3ZT4u#ul#VuXI4mZAjAMq;vjJ;2P)!a`*
z=&*}F=I7=ngfOq!P2;T!B0{b3D(&lMHZ~XsDo!IC8|$-cUTp!L<di^50g)f!M41#Z
z9DsfXf-WVURzyGqrr+zIzL~^HAr?IGOQ{|8Q+IbT!1AJxky;woMuC*hg>Mir)UH2T
z#@ZRT_y()P*oR=kq%@VhU^TY)7zq4&HT;QX??rp$UGGS6)lKuJf?FreE`3BAhM5{V
zSks4JQ~B6a2bS|OldQ);NMh8y>i+;+^BGhCL!y>b>bP#(b`V(^W7);Ba5#QPf10FC
zi123O3QIKu3|-~hA!@9&(u>(VTfg>G(0CI3&7_aO&4O7MbHC5$k~#P7%F1_O6V{M-
zMyjv&XL$o~Q}r#>&st7qxzBG7=ALzDXz<j;3&iS0sl`LrPYd(4r>tv*FIQ*RfXLL2
z2Ke4mx$aPTX$zTY2FwL&P}P1%gNJkgd5V}~AWFXXCPl!RNNvo$4cVxga_zXbK^fEj
z4Ur`)d(Ld}bh}Ay%Z-=xS#h&y?dG@hj6qH3&d_BoFZMah>7~o5bDFwE!4QMfS{;>>
zckeg^&u35aA~4_&f2>ZTb^w9KL5LGC+`mnpOH|x;UOo@1pTbANqLHFl32n|54C9&&
zIy-NiIw0n`oB@P9F#`DcIF=HO=coqb=OEb(D8y7o9wz7PXe!g+W%^zg*o^<ZO25lc
z{ElDaeY}og8oe?q|GwNCQjlVG^o=_#@gEsBzmJIh**e|NM5XEHT$QSW3`oN-eSDt2
zC?gU1TW)dGH#pqNRQ!~7TvE}KM%)00wP|b|`9(E2IABTH&z+i^lVuUpYmlUQP<(rJ
zoXFr|3}Fzs8?bV}HMZR}Xrmw#a7C!=BL?u{=!VM%ldR|yXrKfc=s?I=mT=5tTw&dm
z8UW-;7i-5k<6f`vowpKiZ5z4k7qxQg`OdO>1B-VB??B;o<pD3EBYyVyB_XQt_u-zp
z>%MUVPVqI}*mj-8AIEj<7mZSVj4axd0M&J?Ek9}_=HN!+THW`^dy66V6R$sg{&JLz
zubQ0?w$IROyRu_A4~^RLV#0fvU{`y2Si32LLGk99B|*K~kn_uy?qx+aa0ZHGC}O4t
zEZE~#xzTbRSUx6<gan|z?2zAYFF^?E!k286Lh;h7ODf9nYH-as454>oN9?YsNh9xD
zBvd*VcVOp=*1CfB-P39<H>$VOt{qSKUR=HOe&gEvvNN#-X?x5x;aMmjhe#=@M@aT?
zwlmdMMjA(_YiXSZUY1y@OhWT2*aPG0Xp6mZ2Kf1_n;Cql3nOUTrc=k-$nz5)7zx{P
z|Ghgx(tz6|0Z9lH@?~w8Zi8km>Z{n%Jeyxf`Uop|I#j;&slh_v)`N*GG(x!ZK<{`O
zWWq{;!@SG>!L}mrr%DxcZMp;fS9SdE)tT(-S&MYG<A>9VuJ4Ah;F%f`|Mj+-OL4|0
zV~qE^z5!y9nb?PuYX)=Ba&x3G;czTe70^6*PEKrRHM~N4@4)cmATy!a6g!}ur%1E+
zK#ymH5*biCh=t}OWKFhaQjKWh?;{#@t!pXp2R6scc_ses2GL#h<LS9UXT|5p%PuMM
zt-Ymg-P?+$`o22GH6>4b95QD}&QIOe?O!xEuEl2iS46kx{G(N0l&pt(`|t21)U65J
z7PTk=kr(J!eWy3e-*0H^q;ZZrqAWyNd&Gf?-0o)M3b#RP_BOCB{*^9d+;llc^sajK
z8(0O_(LIdp!2@8_2O{V&Cq1{w!xunIl9swC83|!P+D2)>g`$YV6^rf4eaOejdfzuE
zoa@af9om5p<^r21JsS+QsaoCAjZVNOS3TqL%2>V8uDVC^ue*i8?Tt<`AM=LmkF+X|
zaQ^+!_2WZ5;ojSVSJIY4TSi_E8D%y<TJ2eF&x>Xdx2CI4aK9B%wu@C>qGP-%nt)?3
zS}xi+*W&3?$N0oizrXrB+X`}0{S4&CI*F0EwyK~BK)V1etfL%2t5FFRI{^uapN4Gy
z*ccTat}{qX`v%bkL8kgZZ^q#^`&-GRdoUDy=Y~#>gO40;vzj<`H_1oW>F9);@8n(R
zgwIvUS-?qneJkeEs71S|>unRFt!f6@;*#%IR_-;jG)P;k&%H8&fx6?4+FqkF+_3P4
zIT(;#jnJ-2r&|Z{zR=jH>+grK)>dC4KSycqz~bl#oz8MYm~#h};URFtRj19CfhHp?
z(hi@6_uV`%_*S6YPw!paJmG6&7FQehV=Z4>s@B?Btui|F$FEl!_;+S8f0Qi`S5JV1
zcW)s|lD5!WNt3qEr+yNBT1KN<KvTovt0G9`Kr-6_9S6-K1xa>CBgG}xG%j2i^`*Kp
zVJu>DJxP=Vv)kH?k4TzNZD{ny)~+}g3*|#R-)~>ZLlQZ$OPnaeR}Rc9h7Tr`CvNt4
z$t;Z2J>Q^zM6V(lhxviZaEkIH+Z0G0_Mgke4W7eN$z^GjHB>dK;br8$&+&A^=D2S>
zx_S9hrt|gZ`7JKZ?-dD2+6O;aES4X>AF|k4nj$(zO7`Ba{W-q#Am*ecP?P0Fc#5xI
zcK=LssYuB})FSnlloU`IhTv(R<bJ*#4;&4yZ?N_rVy@NaRs0o}kS)K02JKYACR-XP
zOc?vBXB-_<NRI+M%C)_^*<lh%IW44AGy3K3l~<u7-IrAzZF?`TEZ!*(9)Em)V>wcR
z@)!5X#Q`5H+S>DlUqNq8X0fnrW5Mg$oNZbYE%ld|<sDO*>Y0gULqMI+=_&?OA*Jc_
z&JN&fx%AnDv1fj=aCD>1>z<^{k*<sbfxFw5qia~)n2t6Ndw+(Y9g@^$v2f-cutTVR
z({EjlXSH1~pHg!Vqh%b}3lFdf(j`)ugkiQ*R7%dCG|uka?OH;7HOX8*-_baGH)iA8
zxu>)}-+uPTZ@W7N-kw%|lcAwqwYBQ#1$SPvxbjs~)(Oq)FK9s?ZB>e|+_$$pKkm@j
zEzIPR$tDX#T}h?bOSr>8#YSTO3h?P?#cE<zt&!TI7~hOgJPhXH2Ar2;OR4vEE2`GN
zac;;$_|czG{EgM1$$FF81TqUOq$71Ih9_wW<$*=gv}`L!7B+&4w@qL1K1vu1_SF7E
zs%x%|fSr+z`%UK$!^#c!Q1zr}K`w;mL@xB;YG1Xpa7)Sj;(Jred}xWIkaR9Kw#waa
zt9^o4fk9JtTx?VFO5A6d<jHE2lH|2Il)CXFck#S4EFgv`@Z09uIm4Me!M<-2;AUxu
zTYXqgfOBs^xC2@4tt|!)fRKj>O?5GcaR>dQvIE*G0d2?N4stPj^3Y07JYMdz90;B6
zcwlp#U(P@FV4?crWqbWirgewX^;xw?uf}(N=KZP<|C}#E3_343N-dL)wWyZ9I7}n9
z@m$ur=`CMeCMWrC>(;CaO__C4-0TUE$Xo5Ialu}nq5>l7qHlxL2I8+yTCbHXWK(d*
z()ME0%%p%G>T9zGCFF$RfeBSUVY=GEO}L$0KL2QJPb(=2er=*Uz4Tjx!RL+ikD|<J
z$@WoLwRdh!w)C|9eCyRKkMI0~V6<F)ec!Tucn)N~c4g%W<VGgN1b+R_L{Ep7!*knF
zUi8Pdgao%RE4If3n~hqfTHSCR+FI%J`E8^wY5~TDD(Yzq`ojPXPZ%C2X{QlOn?k9;
z1Q!rkY76}Hi<66H4jqU&KsD`G0n$Cn$cK8SHX)4ZVuM?v=xvW)Dk6~PP1UTsR6lpC
zXozzwtY+RvTMHFFM00HLPlNRXhESguDAj#sKkXGZb(rXGNSb7LN(Q%`wA0UkHd>x2
zU3kTFo$!01ju%$&P|;{kLrbjsW0naM=y>+)?D=RQRoK|rSVbysNg9&PRl`wv2^dex
z9Jocd{1T;XLb`C_&6v;r_m6-x$^Sr5DFrTCy{5W!7<Lp!FtU$HAS|b4>xmJFi-lh<
zPko6SS+eckwo>7%RydnqyAu*r__OLN{^2dpN#)&!i$O~2wi$~P-3F+6N?=4_#4VpJ
zljhEL(z$}u!!0j@9ywh56}~OG@y5JvgYktlBrLSlmiA5*d}8c-1I3<n<CrWJ52|jz
zWXubcoC7OCa47*T;`LfO4pCShTD0_K8gJ?TUJw~md|>1{z{#%`4u`SR-OWJ~M)XBJ
z<xA92-*Q?rRiULiO(LO`e%tJ94`-(v=Bzv3TrrPU?liSN+R_GUe(!NBzri2G?Qz?V
z0cCssBvp$3>xColA>ZI9$cx`uO0$uTufRH<i9;T|*~kRbxRmsnsfE!Pct_1Vpt_Id
zmA!H;i8zr7cd~;0$;fg;Dj2Wt?PDbtC|bZYb%~A9p&)eASUAw;`)O};ZpblNy%XHp
z{ax7*Lo|>@;z`d*poqgq;WqNnT)-_U;F?FPMw(I*+h;5OP_Y{EvNV4lW=IVIxi+-u
z=X#6*t@D;u4GHObNL|m?rWET+hqYkYUXW##@^f?@ywYo#$KOiYO_6BKFv`v~O?@42
zIa>6{TM*+NbATCO_{I1SQRrYH-~H$rpec)OAt6R8TJA%99fK=44+D8SE=<2vVx`op
zV5y_4mpI0lwJ)mH)Fp(8S10tOB%M+kPi@~RdMB^_*sWM}`S~4Y;IJ1$-s9PL5Uh4~
zTt76=KwPiZQgC~-zRgJ!ogjR%Vt(@;;TfZ<`dq;+!*kC|!nYoOa`M?q*tq8TiTz+%
z*_}-Z-SxIbuR1wY8)lwtL|Gzis}Xp&MX!e3QN#!sw6;BwMtm`7$F+6_#Je%#b%SFn
z{vSHK+Z*y?wFMv(#N-qyvVJsxOhLoosxX)=AkzXYFpU;5-yZI4GqU0bwv_iB7>UMB
z?`%%R`Z;EE(%P#VwjfVu$Me*>)Wr*Gc}ZO=KOs()N;0Qi+&o4GZ#^*+H_`vh;k^@I
z%kh(2^;a-c1J3Mb;EUKp_MUc8t=A}ZII_(ZV-tw>&m_EV+~4Ide~w8USKz=WNctj%
z{MKZKBr}oy5Y59qof!u0_~lT7gTnsH(G_e^+wm5*1y8a{9PQw2@Iz>Xb_TGr*D!?U
z1#K-&VffVAOAN#gIJWpmP$s^2>+k4QkA3K>zfaelX!y}^RYz`kR9cenH=E6NXZ`ES
zklJ<c_{N%^_uft}4?{oEVV|Jw2f`T~5Tv=+Np3Wayi7MYFf1m-e4~8c>Q}p)u|e)*
zglOuJMzCyHf-nYBTh2-&#1-`<`ulf@EYFnGi^n)wIb?4ZPL!Em9^r$JU7PLq07)s?
z2&$=DjKGFu(^{eE-}>>b$n2OGL)HPkb;B#l$H1z4J}@tW95j<Fty9ks8miXlkNY{4
z@=sn&dtT&hCQ*hE;-ie&hqB3;k7QeYKTU_k<yXexUvF_t<3mTxYi+LGZD`gVXs0Jr
zUqC~2YvGn_)*=$Rzkh-SSA~SFbtK}u&?hmdgwoX9a_P9@RzGC}Ge!4P^=83joVo5|
z_VB3O;I&WX7t2D9`|;4Cu<S;OIuSV^lqfy1dh30|<cPdh?YZOulA*nqI=4`E>OZU>
zOA6#!$Ah=qlg6Z2T~P7et4d4P(i2szJ{bKNTb)WM#b{U5tdwwT(V3Iq{>5=oK)7d>
zB>v;jK|jb9^UaeT#f44aN0wMT<593VE_fA~+i6P^evAM?p90odHGJP4!!^YTT8VBV
z5+0E6$O73;_HbFZXkdr`ctG(BT}QY0kSDT8rdujLh8M#IVhjGtLoZkTwdyg^*KTt1
zRI+`h0?FUhpeDMs-nX!>H<*0s{;<}LFy_VcEysZ5jA?m`wXSjY;`i<KXJ5aLB@GLO
z^pht|H6jQ6wAE*aE^lXs`;zG-U;5SaJJFXI$!HHai){}DtZkB__j}CBqmL_*Ev5Of
z*GCLR0N^3AT@_+y?AZlCe|w-u)Fhzc$<U(=1NJUw2w>ehM{rh*nuso$Igg8wci-kk
z>wBp8R(SN+V|yN)??&Di($7G1Te_F?<XLI=w_nd@l81G`1~<)Wz6nH@M;!!{`#Gq2
ze|lbi)Ap+WFFie7_}rY_+zRPcW$f~Z=DgV?-UAaTh#K}K5yTl}DM@!hsSna$aR*Sn
zGVR@07XtzjuJ#YHJO?nr2x^}M;7?F+q<kb`cLQK3i|&U*JL4P6`@P4SzR;rlDp~Yu
zoRjQpIr>k%J%jIkR;nhS7*<ukc-K4uzwLWl&aUw}>)E!f<H5OP1C+8$0d*YkR1rZ~
zu+>RGcC}`^Kzv!Z80g(wF6yapA|P6H*m9$MQ@~W_JV^X*ZqFV)8IIrG4pkOFa}rhw
z6w<<3c`ilDG<UT4Dn^tBmVS#|hkU<xP+be<V4`YY2+e6;D*yDkPzObKxK)I{2>TiP
z^!buZO^=Ns-xglcOZ9|_0Ko|lZXC6SKJOYuOVDgt??|I1Vq-d>0vp!sc;7t=TW*{<
z36Yy1a%`tGvuQSy8{<8t#Tz8<7n|nZBrV#<R+L&7Hq}-C(lYwzc~F!QMH&$v()OJs
zGVc`CBd9+x*z}z6pcAr9;y>kS-56b0x@Lbk_Nm|<cv%demS>3k{l<yccR0CKu75~I
z%rZrga~<UFPI{iZ^MY%NpS_2TeeY_LW_X8eR2mKWl!-1=r+-mc8rrhN0cuKZGTAlV
z!ERnuuPMOi3*>zNYfy;BT}mTJGWIN4;)J*t&9QI#lAIK$0Y}D_xTWzx&PyMu=i9@a
zuYanoyqUh+@n!4Lz>)RfmwPXb+A0no981(^4mTtnXl(w+YCQOD6@Go`?<2+EP4yXU
z>-#@){Px$G9j^#-!2aH~#UuZ>D}VpLmpyXje-{C`$NqN_{y$j+Eep&NG|p{7Fjnc;
z{6q$Ka2JU@m-N<lXzmx&vH#1rezN}DVQv1QzW?f-e;fdBe|KCJ6$c*_pRO32@~M2{
z>d!*UZ4Mz$u`lwxH0~HX9o?ragAtQtFdc5QfByB_<vt(4OiPh~<UfljkMkzrqw8IN
zgjBt~R^j(W5CdP_nk+lPTnX@;G!qM{hHJtRV+!$#3RUB%CgpQOop`G432jCURf;c^
z&B{Y$_Lp3g{r=*QajY3`;_U6N-k}2+=Urt&@9TxRj+W>KR5rCGDHf>RS(rD2x3h6c
zFzVDB5fqt8EU4>vDr)_Rqp$uS-rhW#&HNAl?o4Mo-<g(9t0;=^XiHJl68kcp3RS`=
z5k#2Ug+%Q{@$Gcf5=&KUNi1ztLLyO1M1-2wzQh)>Bth+KY_%)>-Opdod7gitbDsa?
zkaM4tBlrDzU+?RBT?#}-u$xU?R|Rs|4KRzsV@?o`Zeb6#9@Gb}-xt&By=dg5>=2<p
z3*crl{Hx10wI1Gk$aR{_eaG+RPl+X1(Au_2v{hT1!5R{}W`~;fq3C%oLt9$|iTAMX
zZN8MC*q&Kmw-YFpC!rur*lRBHAJx(+2c81+gT7d2C(ZU}vl*3jYW#u5Klr>#&L));
zxI9ATc$uCwn0cmFKdw=d=B_N{74P{J;QTvhRLBA@2ti|Iefi_4q~XYg6*^6=6M9*c
zuLK!gz?g<a{UTdbGP|1%x3Q|bYAR0pvYq@o*QPCQF-{upqf*&2lxcjN((!^@82D*w
z@NMmrYyDL-Nu`#lv2u~mhnv&OC-?_al3ThTNUuH1o4jN^Mj48xSqU27vqqLmc17hf
zr3@z=cI-SL;xO#1(XU~{QEIpy*{5%iRprPlZdFpbWyFOzY(xhHIz98(Xf}j3Xu~zL
zW@WEzd7Ai{_kw#u<37qH3j6-md8X7Kk#0XFAhIk<0Qo{&MZCa7w;g${SPOp!8*-`3
z#uZ_t&OQ!W8Tk;nE2hSyP^fMP(Q~o<LhGHQow|=ZFD7}mg7HhCQ;G3%h+#F<8Y`!A
z#(lz&hSJ~DX2`C|43AfeeUH7Gv$?k?2T9cqdIZbO1p>Aw57&zmvyA_{xw6+av>g`V
zLZp7IXv5W@I8aGtK(&EH@IC<zm5y#rAU8adaf^n6MXJFWAWkM^wTjhh862B8UD^tH
z;<Q&u(laCx4slBG^^2D&@z!hy+s+bkU&8SG1z3$&1UzwmQG#VaVvE*|6pQt{YS(8<
zU;E`hdNP(vJu@DOrmx-8ULDCsH#gkTeSSQ!Qk6JOqvwOks(l?C^%IStg=I;t!1j;b
zqqTO3*NiW<@}-`>t1_YWN8)@2V9h;n5<MDV{FKCBfAWAIElF2XQdecsJOK<h7CJ@Z
zF7mptAqakSK0^1)?fD~CT~&?O5ifv%y@KdJw8JIx<YRM{mOrGR&nu3|{Kp!5mlU;A
zk5V^vF{0K9?(a(_+WNh)hF<@ISeflAX-A=nH-EnODx&ZHb-!x?@&xPl=Ix>a6sH+5
zkVabQn!67(Obd%?ek%^|2BA<7RwBXI-xZbI&+yD()zL*@7r$@iLLFD6bofOdHk5``
z15dsW1SZ_D>rh$-{;Fifj7!x{N^+_)>vf>|Njx#INSWLR;&Zny>7v+foZ!cnW0P|l
z4Dkv#`>5hf7f(!P1<FT>=p1S1w(D#0U|2?<vu(3=B-JNNld(RRaXnhMcwoG)ceDyC
z1M9os8W}R4Q&>#ue|IslA?mr2{Gj~5R68*j-mzShf}St)-MoEn=~i7;uvb8A<Yb9w
zCLu8YP&#piac*un1Zb#ig3DYMc)RyiI#i{!OpEr7ZmjYullj$cy+J2VS_v>uVCM?x
zn0pc3fz5c`z+!j`602h&Cyvb$?*sMJ3_{AP5WPxG7f*&x`x82OfVD()STp_rV4;vn
zTg|5lAIz4G9&y&E_CKOLj&riVooVzl+y{zx(mU<){Q>Szfdx>+495VDBD4;GqCo)I
ztBZvFuPv?3N0Ny`w;h1!TB)~qGI|gA+tn`hFJLozFh%~ux%6*mdJP|ui%Voq_Tj))
zeVV>}JA*gqeKy;CT(FYxMW}RI7%-0#+k0B}?aY9lE8r-3u5XM4`l~XKW~W(3G626*
zSPJ(zX{`V*nM?V0W>8x&^4NdoBH7sBvD)v$P4@<=;9+Vy()ZnU(QfSvKay8}=6O-`
z+BF>f9}@0XRZ+=G)oBhlIv$Pv+er<ntDJ@2eVyUfTbRNhV((gViU%qFM|SIK40>eW
zAzubIbRLO9(%IL|1rb1WVV#{wK_2E65%+9$)tx*nvMK^Fg<su(YCgwGTNbd`bNpD9
zS@Fov`Cqv!!RgkIrWcV?M^5o?Yv|Vw+B&;G5huEV1hc#}w~u^_HqWv*&>xq$#YCU<
zT*BbPp_!4u<T+j$r#IimQJTsA=jdKEdI?Ch>=}MEBsM!}-%iEVJ$KdAuY2J~%ejCI
z>>0~jSMmWRcd^5s46R<HKC`;~sc`J_UvOZ<-Ed@tkytiOZup8ZKYAGbtSUSWk)4^J
zP1jM=Hd3{(ajJ>EcwG+Y#A5Q|i(?%|G=tn&hN4g3{<V_tBAvJv#M4sqegto4VS??h
zXc=1q<W_|}(M*cBfE&oLc$*u~Wr0ql@vmUIWkl}k@K_tu(LpLdKdNepzy|SfDg+@E
zc#H5}QqE6rpBSm^$>)D?5qEMaJ*-o%V1%#gqo?L@^vmOzr;n^PZGKz8S7c#!ZLdF4
zF!Zw^l@cQF#SMSzjA@#cf7d><b$%#Jdr1BdhM})-CaYG~%j+*+&gv2;oSGjiTF6K*
z^i+p5taj5$eD+8cRX~wjN5_|d$xV=MAAH?N`=`OWn@#z-jvkC`lCKw%LA5iV?Bpu5
z41g5AS02^@Lvd!PkZ5?PAhtdC^1?9^o<I8SjP^3wFt^TR+W+*;A|BfR1#V2S*){q;
z{d>5bbDNL&FFlX#Oi=%wm7w|zztC)3aQ=~IfA8hY<>{99y06xfO58E&vyzE&*WzEM
zm(Yz}HZ5a?RMzY{!9a*k+-NOBR)7TopRL<MJGMSPDJhFGJxW-dL)I?Av&F22!>PzV
z@EW?Fz-2YWpXI}o;G1Ul!@BD$Lp3##lW3pbLNohM#8d*>L5{NQ>|QOry06?_V_`i1
zS<qFJg@x+CbiP0xCwt*@vrjeRU47!J*JXH8G-gf$?RDR@OFRMvBp_?3pJ;9=V*jPI
zl~wV3Uk_Jh!L0H#=cUjH4XM($rJF7v?!OGh)sw1C<eSb^92ue6MjZ(c?0$V>GEg^_
z#B>bgDPohzb3UFvfn$JGinxG0dQKuOJ3`pnD%?Wr`-8P`Ls~qwY`m4Bjl4bj4A0xR
zcvWQ*6Co^pACB_VFmhq&=ey{IED`1#tvL`q;jBwXbC$6O0~)CnT21b#$@HbXjvsYI
z(DmJ}@2MIko%U75-RzO4<&v2Z^BHBxH5{F;@PiHK##88b`omJf^5OoZ<k_9t`XiFo
z%3J60PqmEy<;=SLcKvq~ow(mZPHsF2=I$4z2dlLA4~Bn5lvY!HIEV@?Pc%0xKiiOI
z0kXH5^%$Yw{(AC2bCpn#7G!;p=b{Z2xRmETxid~*OQ!&9Hg*I3Hif%lSn6~@P7U3@
zMYkMWqA7aL>({BdR*l;1xG}*!wdM{X|B3g(emf(+1pw1;3puAZncp=Rm#&_+U^IaH
zpRpZ)Q9p2PIR$QpyM153ooQGyJ7C8RB#2e=Rwp-(G*&G)FE7PVm-0^!r_MC*QG-0b
zPJi^1zhov9ivsR?EK0(_I-rgWRS###h~>rA8u#%3=32P>w7%p=kOJ;_>+_RgI#kjB
zvQ-`=cMkblxTSTX%$6N#toT=Z>Tohy9lpu;8wH2=a-iaz!b&d0kt@aqIcSQB0~V`m
zE%HZZ=Fs=*z-uuY(q7YK#xrQTwY3Ayk!?KQ2UE>WKE@PM7my!Z({J|D{gQ|*{=Lfd
zDDZFPO6Lf-mz@qmJ(wPd%O#Grlb}9ux+S@1;th2eN4mXu7cW`i*)Pv;<WEYUW+$B0
zMW;J`8Q2Sc*PN6WwT=Fa)FsoUq!=N>M}gVF=Zdh0EX(kTj~oy@A9qta!)c_KBq^`Z
zHi^%}<rvT{n4fJ|hqp^?dWQFOYa&yXn|x^>u#Tfp!F}@17YNn*dpu)nnekc|wIZ=F
z9}ctA3~ir9%dnM~*1d?jS9Lpgu!tH+CsH<SZ1P8Wl~3hwKecY4T|&hq0?OM%3qCF0
z=_0?{h<rF3PVXNXw+ndHTW{`dAZ>ebld}U0P8W;xLSK}Rs!YONOyv8iq&tKN26w-%
z?(Y^Foofgdv=&Uj?hkD0KvcDE?Mm@z@zBN!8{|49O`67q@L^7@<?l*;0=Uyc5e=7~
z(JxeXno4Mhn!SbEYLyPSADWr{I!aYx)Fd)wZ4k<O-5N2G>KQ=38?7Sjh{DL$QQSxv
zLki`PNJh>#x!;A58YpFl8Bt1>o36Zldn=2%Yvhr^0x$LETpz!X9PS7{5PPkv8}CgI
zvwL(|oqiO3@b+_MvYg7JWw{@MPO{&uo#ZCw5*OAH#q=GMroh<sYZW+^{$b_L!ObmI
zZY{#FPhDL!JvUdpZMBB%I}M38s2|<%kh|<}RAR~B4~;^Xl6_WOOZid4zgAohV_pY1
zM+$pgYIM9Om{$*BE6mV6DC-GCb46`pW|#5`NDBS-?MyHyHR^PqRe5?|Xd;ul3_Qun
z^E-FHoyq#lI6XA!Y<HJTV0om-y92^ZUxB2vbpA>D?oVGkyT6_Jtlg1e<Z#N196bRZ
z|KHARXK4HE${))Y0=I48*R4+TQ@sFh@DXAn{>Tu>$~MU{PV<6~pKQW`^M)+&hMe%d
z1`O!inXy9P#gp)FXO2L7%^@~@)kX489@i!%7|7-N7`VS0o6!+-&Uf`sJbOKks@ACk
z*D5J%dfA&w)N+d_eiG>~;Ealo4yw@6>bI>GE5u%!{JOW7wXC%CT~pk(+NOzj{`?LZ
z_;8S0)oX5ok~Q6Zof9>gROY+9N1^A3Js7UhtiBUpm=0M{;|8iIV@J0F+l#VWp9l&w
zXGL?HLUq;(3bhv0QkQ8{YxY@=0gRmjmTo0f45@Y7l?k4Lx^c9{fiH$XQ@r(m8+H#P
zR#T%cTri4oGuDd34E45LWq(ms6<%+t>FKpI>fbB~ZU1)W*X}LlJGCo9&CrUhTvdPV
z0{27*BF*RFTKVdA>rU!#9~0n>Kk+7IF~4D4SK}`p*({r0TR#zl4d@s{j4*-t*t0Ht
zey|5uGkhY@as=feIC?TUxYMxQJzTd}S=Mf9qN}ys*4M-GvuSP4Lhns9Jg3iCCPlMy
zJ?B<k(N4O)h#sE1B5KV|ux1mRTNMf{^`BytyE=)evEKHYQMXYf+LZx;tq+2oThQ(S
zu-~`X>Tf?W(o2t)pRNCNwhWu*x7=~(h|*Ay{ol8#>~Ow!dt-M<FIJ|Bku5u(Gd4f}
z(1ld6T9p+As=i4OZkg>k9i)@dLQD5ev#99UdShLI3xkeg7!n!zPxAr7T>5Ih(;eb^
z@jE2WmqPl?V}jF69PD~=Uw~tb&^6w~tql?XZ23FF?DELteRq7t^lF|{c?bK*RE;uH
ztuaPa8;w+M%FaxX+vSzxXk$&mL~fW)+sbSsQWqc9G3(McY995Ph*8Q7P#>x5e?%0!
zxnc@)8<vtXGE~FyR=~H<%RHDjfdmxK$3QFGVgymLjZN;#&)d_z4BNmWN~${O%59GU
zy@8s9Xqkrnb!_8T#1>p8(z%Ad^8~kz3Xs&Be3nH2vaR`e+cRwq!Yo-jh>B2&FtIyp
zM{*w3K(u$Xr8H|IM>(g(GmELx7OR|{v_fq(Y88iy@SBIHs&uAb*IIb%I}xSUJI*Z3
z1`>S^o^7j~^c0AmD%0fBXx5Vv`CW#m{477`+nJ~V`=I#ez_;bgkjyD++T&EO)pBZC
z^K`VvY<~UX$s#xcMB!Ox1g`%NDUssSdD9D1S;{>39eU9*7r=Qbx6e7o%!{k&;QXLQ
z$}E-r7es+)kn^^52m}5HXU?>H5=Tlhy@8^h5V)G;HQ^KqZLDIrkg72;1cJvjgISvW
z`iHMvDUymQspZT}%zVQ#Ih9O|G;$o;+HJaMWZ9j$XG$VYeqtG$vvuZXxq%}NV)F<l
zS>4`*s7lOFNy!k1J{}zs2@ojm?+v!>!(83cwZ<hr1jfJj&1<1Hu3-$W(%WwVwR#hD
zY1e$qkg|hwM(JSdDt@Hu`1<d)P26$i|C~r+9D`9Q)1?2IsvKFZH7wH`zQ%1X{uPvd
z-iZK6`DO5tJM`p%h-)e(xHaTexJM`wLtQCjK_yW_wr(+uixr2K!p}b7?y{#Ddluzl
zPQ|q#s+VI=dtN?~?K~Pn(rW8V@TKTQT)6!N6GK$YXE&R;xy=}IXIqXP_yAT+aaj^L
zG7M$SYAkJhI}>mAO#Vv~5YqcF8?77Gp0KYi0O0i|z%fH@nd0^TV%<jq0RsHV!KpHE
z2>EIT5LqMW@_GO*bUolr!bC*oS9+Pj2}jL)IODr(=8RjDbxU=D`9l6<Cq_rAk2C+y
zqwdZv=lt{>yJtSm5wD;1&g@G%Mnu&tYhP1<vdy~>N37xwtE91KsBsDtc$vWXtSJk-
zU&v!d8@~@ULN-S)+NOJ1a#cH<0<A|h{G%Ierw4Z%-H__Y1AXb|?#1cFZ&&x;YX^DH
z&ahdL8l$lak2f<4$t!X{Z#2b&;;w6J{4uKTexz$_Y!7j{Iy~>z+0xa@bZZCkM^KZr
zJ||fi3WjDCa1oavatwdHlD2G(8T^~07W%|<6yI_DlShl1wn*HltaFkb*#4h@$l`nM
z39&!d*RzW}pf1`sc64xMTBzzPJlAT*azIs=ghbhUQftwV80n>=WOs$21W6NxMhH(2
zw2E0*OquRTIyO9wEWuZ<3wz=VnIsSL9wmja;4?PV>}u-^ZZy$)yt#^v83*{TK@X~G
z&}#e{G%Wc@ox+H8e^zdBZN{x<WH<c*-U;OyP_NU6T2Uy7&3B@5vvv|-U|8@GKa1<z
zX980PM>33h1)QVbrju}g%D$#*x(9xE&naxXvQOSnY20qP^B?yZ2bcDT-Z0mXtH<OI
zvA<XkD(dt;e`SqQbXMJYw{TjqQ!_x9CrmHKS8jgsS~xjx-@(WwEyRhr38GH1g(;DW
zBTjOblwAY;ePw)9#VDd-8Hdb6eE80F_f*bjd^Z<`5e7o(t3ji>8}Qw{JRMArp7IGL
z%72kO5#9>{wL)d5<?*32k`3`9yj`kM!i<kyJtU0tgnwJL@gm}e+mzp4<i(M;TOxDL
zvVA2}ZPcDS0z%%hNB!HRSHCkq3O}1%_6e+8cX0w$$KS%}MBiKPnC@%Yx;|*F*J5{(
zVf(Zs?rcbOZ17JIestcjk9z^GmfiM@zOy>+s>^!k*B;R*``|P?`&s3*w>_pEsvkZx
zKkBt)v>F=iKt=N8lbz^nzq1-jqmKf$F{csy*KiWI?W<8k;t^`z(|ddugs8EFNIqU4
z)EPyi**zg^2iYO%9Q7NnKL=Dbz4RBYLrxL`A?w!!6d8eZQ|{6mPo|IJUny5#i{I>X
z2-cl3=iTRy=<a1LDFb3pPL8Ov*;vq%BCOoA<C?+X)*!q4^76)(V{i>1bztU3?o%(W
z*9t7L@8a&D?wOM(Ys&mSFJE=KgBt5+y~ZY{MA@^4chBqR4bQ^2nO)6ap&Z5@Hp~wt
zA_g1&LJu6vg7%DsJLukO@2+_*hV^9&@bn2XONY;loYz%zu7P$hG&FziUue-qy%bWO
zw-=%0_3+Z}LFsD457q|%EG0VrvPN<ehJK}OXIvS{37KncU-Y&6Q}5BQ9O;(6HW&K%
z?cqv&;6FWdsL^qUB$(7$y1E~<B#stVy?)bnEas=AbbmY=qnJrrk-EBTNu|d^I|i19
z$?Z*SL2zmN%><!|CO*=g)6CdAcW$(ohO3^1^f7V=VNi+f2y0;8s~B{&gE)GP=ISam
zL=zV&yIChRxA%`RGk4*i)&ZQqZX7fH*I{xm$!yf^q0vhhLo(A@&iv@@Uj}DgBEYNk
z4KjhL`4_Ls&tAQgc&C?;E)Fq}gHL|IpfTz`7`FhHS^LH>QK#IVA>FB@hBh9`?a0@s
zu#E^$OYb0#5*C(+z4>W^{HCm=FfrXejWQ;poP|*(3UQ$3PwbQ67W`4ph``YFZLpG_
zOFI#n2}L)m?uI||NTa}rf+Mm5PMs6Vg&CHoHU#A10*khOT8v??1y>ltl2>zHt(c}S
zz8^Kaz`dq=<D-^8`xfqH_@RjnIVJO6jFm^2GZ<;#P0M+Of3c`7=ipWtpa`?*oRpqQ
zI_?PEC}hTK(kDtP17ee`SZ)xf5+keL8oXy3HbiqZq!Q*gJU5=d+>{wQPo4>&J=#Pz
zpw+90i_0#x>}XtuqJ;!lr(YaXfrq-%{3$8ROdBW{UB3Yif!I&})_R(Qlso;}w?IXj
z{)k%|PLJd%#$*=?f?i&a)TxjBbIl1yp|Y!jR;ooUC2hmX!-XRw^U=1CzHgl)*6H}K
z-GAPu>%8A&)EsoAf0~WL)AuD3KTr>1W}PO6yuM%0#WcDJ!>kCq5%cr@hV7F*T^tQ`
zdXt}FV+B>-eNZbT&JlPFN6xcGwQDgTNl4%1<d4Xh0~ZX$wQrQ&xm!EuAgSg#2hjw?
zlv!VXi9*2z9u>@r(a(2uTiFH7sUUlaP~W@gMy`e6>KnAJ5xogj+YPDF3!qg8U4H(Q
z(eSg@y_nllF*mW)-NGCA)HW&m-;^KYK)3i7^w0508&}Xs5O4S^VoV=n4m*|_leV26
zj!-FEDMC^nbk>y4I>7A22Bq6VfO4MEOPAcH^x|&R{t^Us(p{(uQ_9#w&pmRf66<*7
zZI3w4zP7ixRK;(WYi|xB1sk=@47RF%Hv?v&^keVu$Xoi&N@9G#Dl}e(`u##9X5q_#
z9Lc}Un_Sg5s#YpKg>HZ48{)5&1uqbIIpsu{f7#VjtF6V>{vYBloJdE>6Q-WY5GHrF
z#A0Vxc=2n)x91<NNQN^7WBNz#Yok#jC#JafIznshdmSY60o`3Jr);@SrKPWzR^Ljx
z{%~n8C$XUs0#LG!!#>;Q5|*K}lbkg!dL9>g{rlY*>3Bt1J<_OqO`r<1yQi0yR}<e1
zM+z(}mO21+SLe5ybccxyb`>8X(N3*-vi}ENocep%&u^2RB*zt{%;f9dcH=r$@eh4n
z5l|Hfm=F3>Ofl}&FMNIarti4HFNV7sJILtST4U+04u_U03{CDS!}t%rxR4xiVzDzD
zz;j}zmp@u@HNoUS?R7QpXe*4Mi^H@PQC>4ZHrc(avkGeq9_uX97b2{~P&Tm*7AAt;
z2^mvu(&aXjsErwzr?4Qe3T=j-Uk!s&2GLW{s9TIs+%4MNp!FYh{R?^FW_FRg&jJ=$
z{<?G~^(ZjWEBad9o%{7q)?iw7ZIqJP@>-1h-RxOLbOB4}cVATLgDO6OidV3?Z`XHE
z?Q)$)TkxRW%(;avjmPMOi^k?-j%y;^N+c_l&2EW>cW`%Zt9hs_ko0rbmK$tIPb<dT
z3UF(lz80jXI}{13dvh=lYxC^pPrh(96-`QdR<YF3yqaz%CjSfhaan7njQRrT${02q
z@f9ywMP^<BN2x)uA7~9^LJCbm(JR40>!c^JSaMyCUnB5+-AuT&FJ9l&Xg7$uT^H&y
zH#U;_nb(r!Rli1AP=bxxqSpR~kLW?H-=}yeP8^1*O>;)PIR5Zh|2kUAv`5n9Ka(#f
zMWP=ys%;z>T|Mxc@zp(^I0$!1SA|X<CwZ^pib#I8Oh94Cgf{82uhxsZv^Ql3{xX5R
z8ug`mS5-P~@DWl9{>){|(hH4>g%rva#v%qNL*e{v$2qU%XjhuLNopkr!HQ^|0)Tho
zFo-w;M`1xIk*YOfpm?O^-giqIM}K`z*oJ$wC!HoY?{{ouo)F-G9%lMg->3Qh)Unl2
z^IpIwfc9s|L+WZA{~Q@{ij{ft?aZtC<bjR*X8W;Q&EL)-_ftF$7pYp4>~Cl2B7i(l
zJqZ-byxejdL$kh}37@AIZ>#5iJ98#%a3$ec^HF)fWb@)N1K3LM&7MJf?6;u$k8=QX
z@gMAnHRFdUnHN)9nJ0czQHQa?K(Uj+gnXiDw!2Z%aU$FT4$HFdU0MInB7|)@mAVvS
z<GtF-8zaIrbhGsxUP{b4uRMSEw|}&T9GyPx|IkhN)4H}J;J?`!Iy@p=@NXhQTvVSQ
z7;haA;(xcS#{i=pETZ}s=I((3YgfQGL?;oV6YBs=s<jeji5Hm}`>-a3jFb|CfVe>A
zSZMG(0vA9gqj=;1qOXDk@rV-&N-t!F3#_3YJw1apWo=cKvQgVW)qIbH|AMFCI?#?7
z{zX$3YN_3lKly{P!}~(fw?i2+u7#xryS9Wm-c<IVv&<M;ovfN;!Q{gVXL3(~8W*1B
z$hYj|D8AKnPVq$V2Z0*Ou=gq?DQ54?2G4tJowc7deZ{wn7WT5MN`pj{?72Gb1(d_X
zoZ9o6*f4Ey|EIJjZM|RCb9HMJh5UmiIl@<a$#OC_yRzuQK~_@97)MvMgt)-#?;O#(
zJ=VQ2A%TMC!z!>YfM1QZx6a8DcU>mp3wAlZoYqeVQ+?0Sutd@&immdoCnu2VG+<(B
z6o~qFh8}uDjU2-ybQ!+RoG*&1@(Hs`nV0-?Wmi&$J?dhbDchGV_hWB(Yi0S<*x)0B
zXzVnsbd+$aM3BT9k<R)>r_*OIUTi^mXmQC%cGVjWmrhrcU)Ixn92QZi^!=&t7&lxx
zbc-%x7YvQMer_IC9i|1SKsA^p*0YxM!?4yk7HM+0*5;5E1#f3S-CQfR5>^t^ieZo*
zAn>kXA8aPCg&Osf*Ct|XcldO$Fr~L~SZ>J9R*+jNeZBq~!v*ylSq?-RzbPkgnMqQN
z#{0_X#X!^pUfM;Lj-&Iz#<ADv^;L@ejCs_(mnlmJJ%0dZG41vRV&C@Eq>MjHMs&x6
z8>OCG7KnLXY6xYL=Z`jZ(oTMQ&{3j=)d`E;%^@W{!-AlZ1r^;Wh@_FA`98I3=jTeB
zAy6aLX<p32R1{nQ*S)R~O-yde_R~;c(m{4>M_5U;U`83I265x8M#~0ShZKGV6$9NV
z{uNt=&{4%u?|{AEUcs2~Y~A)Zh}KT6wWIduu6->c$#ediA=f13iiY9ms#5e+B0tF9
zdv&q@sQLcIcY4p|=gzh5*l%k5*~+a;NZQvSGt?RHyg9Gmal=r5d2qrJjPx=W(h_g}
z2tDUM!Vo2;BLv9ZKxa)0$QlnM6TK5{QWF7m^%*jf;nnP3l(lnO@f)C#N_Sg9yV5>E
z)sQ7_SJ6rcAQZS6Nh;v8GWC2>9ApsG!>*(&vh~avgP*@>7gtNBm>mW}zMXk~G;BgW
zQB7D6SgQJV=FWu2;n_!^sKGC%eFwm#8~N|IGeqQiQ&K8~a2IwSt-(zCwJBj#<Z5TF
zois<^)hZRzIpgtgK{w_Kh2-h`BIFOO;*}NG_P1|?dJLY=+I@Ie|F<6@z%(CoX*q1k
zo&bz1!PBS59KCf)!^Q^0vpql7wtrwtr$_&a`#|ujSmcO>X*@J)g~-zMRHjYTlX1jF
zP0cIsN0UsIeKrcioQFbn(&9|hHKgZ{G{(D`@vIsPpQ}wEdjuuK+y*w<YqOtW%&PK`
zJC(<E)w{|{O|*}cuLT1iSCE4o#@8sLj-MY1e%4x>nfH)4lP0yhFQr&8|M&)v>viYP
z`9y(D&eE!MKI)ebwD!%DYo{;aGA$6xxM+=;J+yl9bTd}>7Q@eBc!gXjX*0Sp<2I<h
zxr@(|t33ZeaAXRpFlIZQIL`j%KJ?x+7m;RkXb{~LoDRPZ9lgD5gh%n1v2Z@Uchs}h
zeBPslFy}MQN|)!c63!~7(p)<Tit{U>8kl<8?iDyQ>2WdWXU5eAV*%BDpwPce00r5P
z&KV}8a=rYX@)k64ak+)$9*ydy(fU0Dg|Pp)9qdQ<^4CyW+;g{W=#Rw>eKsmjQ#7sw
z2V>UODu3NmjCyV`UVNXUr&yIwwwP%dSRX~tU-9+lMW#pD4euH`@2ilc7tT?<?6@j5
zQj)>`!*doU8I+4ds=t#Sq|zbsPV`Dw#>&R$S$M~hxBxiIlTmY%-Naq2ILrppjh-Ei
zZ-owmr5r2~?)Ycw+^FyBA=82Kl92If?Z>L9dr>!RlRRO+Qpu$9-i8KU(^IGeJd~&A
z@=-Gm^oP!up9;Mhk}v<QD)t>z0R>6z$tRHP(&S{K13qVT?A+cAbg!!4++#~4-2WTs
z4BM>HTqc3LS>>*SUqauLDWXkn6f~JE;GLSP={S^nd(L{g;n4HmUS1*5c6GZ=VG+p5
zTj(IeP4r@)<;oLyErRaW2W<$lg;1hOM%v{}i!VJ*LUiSKGD2BdY~3Z=hw`6ZsAc`%
zW~u+5e{wmD+9@-emIH>n_Y(U5UoZZjjftuAh=Pq^9nvu3@kjNke&uE74zbssQa&G4
zeSGeT96o$ABA|<Os0L-Ze_ei%q<Qd@cTaGXGDIIjmu)KDGy=*LGcBoS&5#FCOr2wR
zu-TtXavzrFUvD^~s8||5fj2OS7N4`Xb-(H2w6$gdr0q3RLK{rNXOE63Ea=mQui89#
zkJiXr>Ld4pRo#G&v;rGS{iwP#flMY$ED_2Xp=w-2oH(1wyacoJFb8Wgj_gY6<<bIj
z&c4foTInSR)ZM(7=%vOhXBp|(&%Pd9S+zH9Td-Ol;#}4?|4RB#&T*mA6?3HK{}>-O
zM;h-=$`n5Vl1lp#Clj=lSgZ>(CDT?N#I|<oVd3%arXVpLVC+BUvRRJ3ZNNya=xH=5
z?S#jc1k_90UV+)l;ZwZMORK~5)WK<i;XJzdsHPag9^m9Uc4(U%fcprKWv36xhR+I-
zzOb~FuzXX3uZwO}7OnYV#b+4pSdCTq!cF~x1VQHUTk@w*ciWT4ao4r{eSXn-lZHkQ
zpWJZI#%l+nnu^~7)uNMFx{ioEDnMdeH$%b#EbX31@(RGjNp1u90gonIMKv;dlB__X
zw?TW&huaM^|27hH=<ge@-r6<j9@f?ht45YQWP-&K)OUNa9PlN8(g##J>Tr5@I<6dF
z?n?T6weDhy=NMGnEqrkLU|vzpCO=(U-4$7N)JX?*y;eqDp6k4oC9O1fJH^guD8$^N
zMEY)JpaDPAZ6r#R1Z~(giFF!|DExz>JHy7+MA-C<k4L!D#WazEspe3B22iMB>>$>v
zMN?F(xNYO6ddbnbwW9{>t3n}{(KFF_rx|kdH1vwS9VphXnSPNG@3)u)IpH?#KG!67
z7f1P-PKMNwoqDUoZkHNqJ;?XDb0nf0Y+Qq?-TTetO70_(ZBDaQf2l<->GN_PuDypO
zY1`8FZ+q$<S|p;~^k%B&S{syo8c0^-6IIBuT;txuD)Ei)b<crU8S7pv*`d4q+`zoy
zb@K=Tjl42rXk-cC)?-Z%ZEDlppF`n0>m&E{AO6?Gby`s;oFxEg=sr7cip<i!a?37n
zw=NIpwDj3P_1yx#%icm>C;hbrLQLn2jXNU}Kita^l4#fWVz$?ML2x1k>QdyHN@X)=
zqKi^U#MtA5y|>NRKA&x2Np$*$Vk*3#VgVU$+{qELMVV`7M*gl?-uvgl@9P7s*RJFV
z%SW8Nb1x0u-MiHrJVHH!m?+b!8V-YB=;`@R5>(PcK>jQK_?YfG6kD7gCZ?b!z<%X3
zb%SAmjA+AwiCb#gXr9x13O_5U*bc3Td#FC^PLChvmLte{Oy7E@Iom{bmV7;Luk~^6
zlSSJQ*?XUAE}Xqi_KDsL!+KwhTlw$XG6yIm+p(v#*2W;c7H$5j%)#T<_6v*pb)2k~
z^ox7o$jPQM6r%0Tdb%RT5y=X39oiLYvGxcjsUQ(~!}gLS=qkv@0mC)W-y_bK=V4!&
z4{Be!T~nCk_vfeHf4fpi@<x)^MZem^AYp!r>_~JQC~Ar9Pb(vqXK-2x#WwYq9|Ph%
z;ywrPzi3h+!M8caQ&)zsJ#uo*_VIE{OA={%nl(iG)U@P1yku2PO$&Y7;4L8j+%Dn$
zIzN4MQvO}+P~a1x8mF3m0{0UPFo52Jy*hfZOu!>AZ^2O1IehlJl`CFzgBWB-vy+_9
zO3)-xkpff1%@~%?grM=CbJ~U?u8BB^$v-4%K-jIrn!XDqp$*rsi=H%M#SmiV;BatH
zI2S5r47Jnj4=-3b#Ri-*mg8gI8E~zbNk|I~RI)d9SgnBq;t1N^tjeEyw<zu?D|Qb<
ztWb(|ue+?&y>(VSgCSuph4Xccg>hzj;A51U*H4JSbzd(|F4p2LX_9yfirS*&5l+Zx
z2QdVhs8BC7VR|=tShtSs^a`&Vs!}f6x3f~RA>@ULt{O}}p`I{j&K3_bAtx7I>UjJ0
zvhH<8lk>;j65FjNi$FaN|JCFyZDK_{k^NAKcstKKj69W`<6CXh9o|5?Vs$ed+K;=B
zDo&nyCo^Q!W|CF^)nuZ92uP56PCqO+&<kWKDT%r6w<l||+R`XdWLg`5YG<wz6@3*v
zy;8iI#6C7<zd_c!_7!^Bll1JGR<~AIWo|6AdN|h-#K$S9WT+N#OcP*nrvf=aEhEg(
z1m}+?4d0Lbq2CMI+A00!6;3sx#x4$;)pE<JDj&W%?BqC*&ehG<yQ<_2yV2n5+B_Yc
z5?k_7c+6<LKOXgX4INh%Yq<A1>d%%EGf`~*kuN$sORq{D3Xvo{{<5u*OfE}BL&W>F
zZT|Q=x??QT^Ul{FSE?8IZ$FVMb!D8XtKRF<9H@|if7Qy?mJH(O=Vfg5#=4r-O^U#8
z@Q772xYc)K>rdKap3`4wKUlXavg=eHO3Rh0YYRFi6CE8jZ*U!=3^a%7lJ+;o<BW#~
ze;u2WI5Cd<Z)4Dy;plVz>O<b4jF(5HHM4A-6WOO&m=ATepgyBM!c_FYkhEckLeeZW
zEY*)DfFYw(=Y53z|JjQi8PSYjEB3`n9fV{_Ae)OrcgxZW0_zzVDnJi(<2AWCOzk1_
zC}RSXKB1Y#e%QOn2IYz?BNwM7b!&6Q;O#*sqVQ|I@v)ImBKMtz#1}}EwUr+4L5%T~
zj_Uj4I~vbE+}W`Ep3x_nj6Zi;oX*tws%CjsdY$w%_+e{WoWtct<NC_WLPvW<gbS^V
zP+sIuNiAtu7w`kECvcDn8E3+7v>vdaMd<Gn+xl3N3iBV-9mI56p_kakQ5kGFkMSQN
zK4{jFwYk)8g74@(5q0xy@^!$myU1Uc8$-|6WG@DFmiaY^L~I@9%c^IaTi?jNnb&&?
z!k`WSc1jOWUeIVh969MUd4hUM#ZXnzp35;Fww$ug66vz;n#hI|H;304ReFbb*tLG2
zD*D&kyjU~Gli(7Uz7oBq{VGhgHrj)+QppwzH9<JGMkK)O#KZW_zMs}deQ=kW5FpA>
zI=YmRo#NmdkuOl(GD@7;11C7xy|8bA*ZrCYr`@|HrIv3b0_N6lmo)gX|M}7u|5Nk^
zP)^dm!Ja&onOly=?Zr>^@(m<GD6J7nx}keyd~Uh-pe+bUspElR?GZ$efGH#3Uzvx5
zM%U}P4-W1Z;R7FxEOjzd{Io_QE~Up;#1`KN5`JxBAdZ-}9i@vUW_it~*RCpYFXdB5
zfYJ57Z4>YS@4^RNt5=?OkDMqyq026g;t$P2>@hQ5EqHN}j@*fz7b03Od<ot~fw^Fg
zR;Q`qk>>$-S+cISuzbcnezIps2N60s{juc!faVT<XT|Bgb}bLdjAwVRF#VQz88(pe
z<P@5JWs!e!5k4R~fWGGxY4P?gkIl$_ouUu(Q_^ZoH37q@9(lyn@?>JAi*L%xVo%;)
zwfn}N7z4FAdo|-atu|^XU=zEXFDo4)vmat*RbA{J5i07IXD@IgimIf&_|*9#>0*WY
zpnvqmWtkbjl8h?a>IbR$<*T^ovjAqT(3a|DRYT!XF3PBAt2y{`S6?|G7ir04N*3F}
zZtY7;)dXMZkIAvIXq>PSpYu3)G|N-7`v0oXaAT~j0D&$=WJd9=qCu09oPK?zFxX2s
zhWp{4Kj$@x$lJe2uf5p*O0aMmn)P~ldv4^F#G(c2PJt3?A)+TZ|3h|w${ggyhB0P`
zwiP>Mc@g#3K_ULF&#cg<&`#riy1aSSJ1AJ_IwD)p)eObIJ~~2{{3OriI9gKj$X>^x
zhmlE1tT1mxAM34a55v6#jEym~>ZT|v7GP-HK!OUfC?$pEp)L2MHC8Tu>mXE^9fw%A
zb1$oQL9AM<-*y;?e{iZaaGO8STwD~re<8}Kk-KD3%ru-tbX#Q)EKaD-e~JEa*=nYv
z#N*KAL~>y%uXvucqA^-b({YRQ7)i)3rX@I?$c(})2an9$SL7O$7e{to7c-&PMjLfS
zC#U2o%^vU`=27Kx0{BQ5>MV%UTbgXHB<8O!F6999#D4LRlk(ZIpf5>%{74Y2&|wKI
zaZf^yFlO)K?UU!N;e)2x8p(C5g?x+*kL(lk=GJ~#y7#1OUgm*IR-N0ahg{RmuMc;!
zdigPvN<K@>7#Rndhpj^@gFCU(U>z4Odd9s6yE9I!Z+0Ag<_Va*A(zC>!?^rx{~*L!
zZ`Zz>H~enE`x@HTo7dGvQS`0ApdEq<NTQ*Fg78Ue%-+75?9}EbUdBOanx)35>xF7?
zpV&POnG&bpUU#)A5bIE>6;sL~Z`wmH7aIU7{;JkP_#DqVOLYI=OzTQ1lWXCqs6RIQ
zjec+C>`RcmoQ9VG)j7^x7?XEUm4u||1-lVYaO{;<C^|2XT*+H3Hp`BR4ylW(vsH9w
z1!~Q&sfy1!7X&qIEn}g;S`h5Wu>mT4OvC17^ELeOGW^rdO`A;HaNU<Hx_YmCDt+sc
zd(Z{lB$aKj30N#OQ#UA5kSSKL{NEkVH|alY8E=?olO>WeKD=<hjsdSDlp>>ZUzhr<
zO6w*C2mdwtF2E@qe`UGu%NM%JEv6?sf)lzwM8B4guhDcI)Uofla<7^PH!>f0Hjs`?
zvH&**RW#I~+ik%v;_Rb}@hWXdgt?8w${IYQ_=(N2Gd!|DX)mG4#JYl;j_&wkWe{N@
z<MMY)*p=SWncdQR_Q7#cBI=``QtJxx9JY<LBVF?KXCFMOQOEwg_@YsCXJD&kaM9m4
zC_xk?!ot2;3_Qf=Doj!tRrH{IuJ)!5+eo27O9Me52lc6ECqZ;Q4FhQS*R5Yd5xT3J
zEyxYXgAodKoXyPj6CDu|%glE-X|_iSkB80qfNg|Qskj09(%ae^D|3+^SG<(d7yiCB
z%0fA4b~>d~xcxF%xz@=#V39iWo1TGtjLoBqQ3k{x;)!OKZw7bRigDWWsg203nz-oW
zCZ&Yvq-^5cy7JIt9T!$<z({SQ$D3uYH!K}c`k@H9Y9*BoL|+UO;JA-J7RFKE0p`G6
zi*)eOoztvXm4MDBy>$$K%f_wP#nF5VnDsKu{2J+jVFmKKF06ET-AL149bt?wSaP>v
zN}rcY>ssi$=~*rtIvN2DQK04^ZdJS~4pCY7@()@yt5~7;RP-n;djGR~`^U{Me<9(1
zGN|QoTt*~_YKQV%aS6!+$e{RK(EJ1o5~WPu?IuoWUxSlUy}F$~%x@O^quH+O^a_{$
zO2b`>1R9!Ih`@$fnAFpQvEdv|u`A|b#`{Gl_@p{vRe&N_C+x~VlM*Sf{m1957bXuz
zEOn~S|5JIhORbt>yHc0)>BfWWbqxh&bG2ETS9uv3{sD@2G_dh1>l@PhdyE;Uqo!wB
z4SGLHBc#)XdcRB-CLs(hvYruu^N5iHRDxI_VkC{vhB&OMgQJw1LRotYlbpJ`Ozhib
zoyn53qM?dxy><3#ZJ#YzKxu-hB;%MY43oq_MiUK|{r5gc-wp~6J^F|X%piH;p6^8!
z^;G1)FuGgZPbh<Pbt0m7?j<@%tK|Ljl2-QIwv;ZVe`t7k;*k1V)ubdQ{>VJq>^)%0
zNs`fa;N|G{jaTJombwn>)+xsONPMB+W;=a#?_oybHf-h>=OEfpO#q|CkS#CuUOm0I
z#_pmltYQk&|7cu062h{e1;;rkpm^N1ql751*R<S?^iP?I?X@`uY>RJ?&J+1hhMve<
z=&ZjTd0%KkgoqkW@H?4w^B=mhZl^}Pd;V}|v|-%#v|`2zA(xu4{eru3iE%o!Wf!|{
z_iL1o0M5^hK#G;G>^~0i|I}Kk62z<0_L2~A!Ko$)+;u{J-eUe@zDI*%X5yOwXG7gW
zNcZE~W785jJVQ}9#^6E|7@?ku<qPE-cQRxVp99|)qPZ9LVKoJT4-vf}d1Xeg7EC2V
zx}97Vr57r|8~my2oJ@QK)|-Gih}Y=oWR_dy|H_A8kIYYQ8XJzY{5(;GF#mD>v9-m&
z13IxTv33_VRMO3Ct87g{dfSvHNYr{;15oN_FNSdIZ4J8+MTu6c+NIhV!TPGoq50)L
z?mb=Pk+-1}%q9#ISOTJN&7}I47b&OA=+6qW+1f*CN><@hcD0mW$@ij*@~5v7i@ij~
zBb;<^bv*9AdUamq_H9}J!eS7JUX|+Dalhczq1tYia?5Wdqn7X${$k{M;c6w!bAi<u
zCC6xppyZ7Nw6`-UXc0SAr{Tbk4x(>CthlewB$<mt;!toLs<roF7Y1FN`EGek+WiL1
zRuP|wCC#*T6B}mbMcXQ3FCoHON8R8I8>nL|1R4hWdc(o@>ksv@$a8+Vi%d~%^7&5f
zCj=nd(&(sWQ=9ZNOiHbSg$J2G7ignzJK#ZIGNsP>KPtIrrTySP==O<|o2J>v9>M-_
zKP&EDgtU`jFudu!4dHa)QxhiL?Ea@J(liIN3o9oJc_CFoa1Y%a(!GpZQ1HQqJrNy?
zIn0j%ss;<@eCMWmnstycWn!@R1Z{Vmm{*b-fLo=c#$~m}+?<ka(QNdPC$uBUGsH?+
zVg-S!d_hy;?(b!Pw>{B{+tqmxGm|M>9Wj4Nx$R|+5BnY%gtAt3<$~0_4E-zhoQAbQ
z*Se{L5<>3l;;b-L1r$GH*j^6}R6~ksLW~{cY^pLnC)G{T0_`;e<=7O7nevr91~L^j
zD5kw=qSI<~45%z3{fEW2*M@FS)P-H;#;;+*a;$Y$UX$VLxi_x8mWJPYk|K-De0lzT
z_O9u7&49hlYOxgBXm}@JL>D$FoOMWnXxBXUt#kmF)m#E$PJL;~Gk*O=e(YqWe-kdQ
zF!lUo99PLqrrECThO7h3ac^mRaeUo|x^O+yTPOF51SmxG!#3?xQH-YSb9ep75V^!#
z`7vtJxbU%<KI*6IIvOvNnJ)%p^~|n67=8@0c~B0_+`Vo6ZMdb>rL;7oz3toVa64o<
zHfI@EUc|N3G!1HrXG=*0JG1&J0BR3ksNIH;yhTo}FZSNy8rxOJz6D}fpJiHl&h=CU
z@sgytH{2hM%gQ;}FmZ(e>I$f;WBt0_4i(iGU6;MCbf)14N4%48%Q_`<B{H*>r!S`-
zsCB(rg-1DZ_fQyg^Sc#8^%8JlaiZ_8vmsG-?OHbqLEPaY=biW0!7N~S6t{uO&emOT
z+0=XV*Zj=FSjX_psT??sqU}@~O7>hg;i37sa-2RnkCvQId@`eWxuc8+Ph80<TEuw0
z9<FQjuzxHl`JwBQ8w56{{!F8!FjJs-VGpPG$Oanm&)-<JYQ5{G{xLrw@BTU@OJYVq
zavSPKwgy&ff?avB&nyHDCcRkAtV|2an-zT?u^90cLsV||Eog)Tjx0xxq<?BzeljHo
zI{Y|f7h+~))1A5~Y2VWIu`5*tr&{oY1)7~qRn6Fe0b<e>uIw@;CG!R%!0fEZ-%IVp
z&xVYDW?f%RdJxA^5`k%~z_LB3*U7AoAm>`D<Iy7FX8zzWj*;vGn$&N5N0744S9%(;
zr&+;P!N1nUFOIP?p*+1>k%_k}@K_gmO6}%WC@p|jrzW`F5VGyx1R86C;^}RyTHq55
zaWvoAY<GqI&62km-UJ2Rj(bqRHd1d#gJ2+;T(!2Zm}M3^kl>#T)c5AC5d9h>txW0u
zP6ey$uNGZFWuWu1=Z7CiU0tx4&@Lm<5kaN3v!&<yj8rgU3q<?DxLVvxw?oNb-;(Iq
zX1$-D%kWVLPwD8HF##OrK&J#qhJ}>%ww7qIK^!n_D-h@3j8<r1G8|j&0`|0YYu-It
z<+(=!?BCD|wLjXN4|RAnlp8+6#->eDIR&S(a6Khicx!LK_LtbJ{K(2jh7-iRg|{^Y
zpT!njy!t$lKPFE6aihQN{ks=o$ak&x#AF<049s8Y6K*u18Z6n>ztv7Ep1!zvzgSXN
zab|NddIj%v%A2@4c7UT-h)`3W<|7J%)({rM&OMAs@)}o4KT;7%R07K474^x!ny1e`
z@AdF;brN)<%dmgnzf43grqWN)gjQi9=Wz`}_9byk3d_yhX0J$TY_h3cSZuWqDRIg8
z2T3<(Y+hR(*e_;#)%^1JwEJ@{vT=K5y>R9|V<sFOax{N`>BiB)yB(j<d)=DUBC&%~
zma)Xv+26Ocy(T|ZHAIfC0;^T}HC4=WS$SGG!ut2V-LWVn0XyLuNy&wLAeV3_aOxe}
z=6@UtT+bCY=saPchZ2R+Qd5?@J7X+ebpK#=g|F)}%f^P&wiX)(yVT1OcMt~TjMp0l
zQ35bcU@8xsv6$wkzjmI-Thww}hy?#>U#*fKUT1#dSYPI|D0*QuC9&K*sA(X=q+CJf
zyV;*eu*^w=Ac%Df^|&0BR!Aq$2QXtIYuR)xaRDM!HS8-%(%lv!$PLOQB{_iltDHy6
z@Gs(Y_wn7uf=<dQV<WX#gB?GwLCi!$<svLz;E;u6Ng6Pqm&CNu4FsjUnb>0)j|Rls
zCE{jn`H9SkYoPK~H~dOa)<{*Xc01L0o8XxEug-YOBXsHCj=7kHpG_pK-#joYziRz4
zL0Yflf2@&_Z)Z05uiLf@oWj}ntO)DnBv-Cxd{VgsvViO-0SGU8inj*`!zehTE!p`9
zbX;XA`b3>7??O0<26@zE>C8@&+xp?2NPO6;a!QdIGKQ2PMDr4UrWOS?JNT}Qq|6>x
zJoylRGcfVKJ1oDqYt?wRzKU3zm$OQ2yLT%mwCs({!}~`lT{~TH*SwTZrOXM)IArJ9
z_`$JEYur7wkN?$%1&JAYuob_a=VSw*g>ld2Od`B_EI<&#N*CEc12b^sC9)sxH#ExQ
zd>xTJyvnVG8w~rxQlfq#S8E4$F;+C(Y9i}X74woQ<hBngi?!w*^n(3}YT556M~eRp
z$f+N4>uTF-?3RmiPJ1vPO@Hd~vE)O?n8TO1cNp@|q2sdo>I%s(80@69E}I*f*P<n7
z_mhh<SR{u@Q|b8tXWU5T8*fxa5AH**pHm0Lx^p<3{ZM$tYm%0Vnk;Bi5-n>I<4P>k
zG8FMI*Y|z<2cv*93uN?GCli~QZxK<}-bs<G?<BB~uh@**Kmq-w6Bra;^|l^grT80b
zL>5<HV5zce)t{-Q$K-pZ&5S<isqQjx1UmwidE-B0C4#&abTZab2Mhj@D&sBd4bGJ;
zVh%QzV`uf=0mc;6l$dtisXLyri`=Ce7GA&GF%)cH7$&+W0!6TV=ffhW%9`B34Fm<1
z8CneSH?B}aWUsHQr<*+>^yN3_XI^FaDXBN6`f~M@S+Y7@8!#whP@XQ=3G|<wPg)n~
zby@#2uu7M>tJrKOR9D-YY1R5E#4<~3`-Z+lU*LdOeR;afug;3_uq>Tw*O$IXP4vO}
zW#yTKlK8jvbh8UJIzcwn!K9~8DbNc4y$FnI2d8q*dlsCk0W)_Y<C%KC4dn}gDal0x
z$5D|T15r%{s?x$nO3tSGWK-r&5Uz@xx-kfrG+e;>2Br{Gqf;_C{mn_l@qST}_iHY~
zKN9Yih5Gj)!<=gjEK4rD{`x*A%_QcR<;D%=LHzoUH(g9f5i)LlRw@P>S$B#O5;6gn
zpc=-C&l_p9V-IXEjdNFFI?jVT*3#yDiaY56>CYqcai+(RLjB<TCK$kpU1n<byVV2Y
z%W;$!w+#A*-@`C0@N9<-g_65g(yTi1nwAL7!HyyHZpaBOVBbu3W|PO;J6!mzYq)jq
z)q^bJVCenmt{ExgG|10>%9#9P_YL!_eOKno!$A{$B!g?AN-5Wl=70J6!wz{XY>*`%
zHgoU@|8~Z8V!sxD)Eg*$Gu@q%s{_X*HcEBiM{nUTK@9&7d+!0&Mz(E>+U;&9Fg9So
zM7M>>APgcWw@nTL6NE?tlY}Lb2_gq~JCSpeNG1soAfm_^n;Zlfgvh}rCxOYCe)7Hl
z-E;o^<DGZLIPd-c?)Y7UQS7R{YfH6iZ`Ims%{5mBv^dOnKc4Cc@O&yA{C?8IfAn_`
zOG2`HiCkKk*vH63V+mDR4|q|N8yGYSk`nRUU7Ve%JqRvStlRLuPb^QrnbRBzO`jT>
znau|_dke)B?mtT}4JmN#kDAC9PZ<zSx91p|I~G3GKq)DlX!xYB-p5hSZQG`r?u$x0
ztseWqlZ~TuO@~AV!&BqcU`I_ZeW>=erTuz1W^E3Uj|s@zqj1J%Ii&><oo|kS8%#$_
zX1gNNFcrwuq=HmW&pv%lF%pQVo_4si(>dAr&9$m&bE&pJ5bwzvkoKCPz4vy$F;7xY
zyESiGNincq+`1)f-re5YIW0$4NE`|w6MEzA`&Ym7k*;+_t1C7YJ4fFF9bE7x;vFBC
zR#FWww1H;le(%4yd9LEf>r{~$e~f%@qiCc~ZvnB;lk{3YbbU8z>$f!Tu%6@rw6n1}
zXCThQq0u@f`!%9&RG+071m^9SfhTp+Z`hybY-B!>(#9V8L9&LE>B|*fOIHlD-?dMs
z>2cBn5VAVly2Ifuj@MBs>gHia@h^v9=B`-ng+KusK)HhF_EwvyM*3vFs})0qXJd)C
z_GE3UIm1#c)@;bFaOfpL(aN}7)n~*JkJ%!jOP0y7)eKw-*=4Gec|KznLAk|lzdBmg
zL>CcGO^#F}NfR3%R2;SpphP$!moRG~DLLf1@sRu|vZ~?ziRU6C-{Z&oclMv8TaKjp
zOU0&?hY!!)sta3+7sCw5uG*Ut|A;Sn&*YSr?KffF@Q_SG@eUbTq&|pmWZWCwCQ+m5
z;L~FnoJe+8M%U>P55gJ++2GA?)N^iKFy7kPjT3i;W!Dm<zPrtZ>vQOf8MR#u*K?A-
zxsYb$(vT<Qw6$GswvmUB=u}vE5VjpEl&uJpAfi0HJ|dT!T|M9jO;siLCYa+t4F+;z
z%cI|nT1;DZxbl@N#0}@q_6T_0bRROm?%M%gh^-h<&+ZAzQ+$2fr+1$BV0+eo;ppw$
zRkifQ+MT^~d2o}zHi@#D5+9B=O}$sHD}pP^#;n#TK&j7mz}KR|!D-cBTPsIxh`(8l
zsbRE~q^r8x>@i|hgj<#t3_*wD<W)N1IWExG@464IidlJU*Jk81n$s8cUszApMJ{`k
zuf29)&&GE;YDpStgxko4RaXco<refSFRZUE<kN=B`KYc=uU`K8u60O3dJK~^)-fr^
zLov9xRif=Qk2Fn~BJT^An2WpooAm`yOzb7O@TTm0En&D*@)PrG&v?d9i%hyl_8n#S
zmlqAA!E411qc+1GlHooF_0A-nb`W!kK4aT1h<IvVS`%5lYCI<4LGz_VI6toA+V2F<
zzAL7*6(4K0y^nc1AlB1KLul<CBBj(pAl~hvPKT6j{u-NRq|Y|AjY6IWQr%uWu8w1-
z>^UCLd<(KDV$2h9gd*py%CPZ@7Pi(vj7&p5liUz>cgc$0fQI<C(MOF37jN?^>Gy9d
z53M$+U8S|m=Sm%gpN~aEq>e-7Hg#0Ued~w3vQfiC_3|vU_0Q5asbS>CS+jj^_F?tP
zz~8En^AFskpE(I2p<<XLREP}3ZB8J?*;|<T`lPd(+BcDa<Kknd(RZ`Cw}(J2k&!hD
z3KMMD;qvN(Y9=2xa51GcG(A~OkysqjlYIf}9Uk>-hQ)f>;IerIQu3o7*sN#JIhC{)
z!Pm_aBh`whpHigVk(HV5<T`w+r&(0En@+pfpi71gUzBeqQf1FoMPm1<XXVd>Q#Z4~
z*7Y7yzqOZFZ*r&-vIUEtId|x+^Nnkh?BewHHR&EERZ_&+@|zDkn&yo6rW#H?EjpTN
zz2PPCC5|jFw+ma_OXbD!MP!rqY-DzOUCMc?C$U|@E9uGLUNQK#_UEBTfJGTUFU5Pc
z!_#;di*V!lFq?*WddUGUl-o4xMBb;O=VqS^O)rY!CL>t+WTd5i+_Eny(J$8C+^pdY
zKL=Z^_zDHlgP*)BBw5J}vV&aYdEdpr6JjRMlTXx|c%_)XOq_DNA{(BLysnbC#qIoJ
zWT!fb%NWdN8Tz}e?&lnC-iAR-)wLf+gGn;gXmE_Tk^L4)(ECM7s<hBytuEyxVF_u9
zmK`^w+QFymx_X)*r$3My?RTjj8Vn+o!r8Q{_ep|un`FUwlT-{Z%(eBQ<Bqa6IpNDf
zlVa7wNUhn0I^%c=U7e(qJKT5Cy7WNhwi+Xqyp*B7fm(vdY@IaS#t5>~vb0{Z&)f}3
zn4bt|78)iLpPnS4jD__<s;QFpcGIa|h3Ifku!IUxng6S`3F+m_stpsISeckuS@~Rj
z{d)Ml>KEn6A2<r5A4V5wH6HFz|F8mo=}ZB{k;&bQcS|>fbQlJ@#ZC$W^C#58ii8?&
zNO?p_*eOe>KVQ`H9JjuTKzs~GY-rwE91PZs&g1<(2O-CtUbHbGan7RiY3SvPhRW|Q
z)A_f-?S%K?;9i1NXq~(8c(zcTSRYD-thgAu(?~NWi<V{eF5lN|N(?#xkLMfIfk9s9
zJ@_dMB1{I$i^~r=@O4kAFSdM#!B=N9S%|-f{KV<>)JxbK#{b2(hV>7(R>#yRMld(F
zl44d@aH{h{$;E!-jM%sliggE3A;Ys6fz|&o-cZ0G7|Knrf)~K*N3M68Ai$&iM9j0N
zIR-!^RjSElpxP&7DX_pEx2x1BySlMn^39eFZz!1GS`p&1mRQtxH$|K+j>=+~wP)d4
zg5Z?6KU|_P<j{2*c)9PKFC)?P$k#`ft)n-UwRUt{bkAu@P$*Sa3U5lKFJw8!;=|VF
z3bu5JI^NdnmRF8~k(7oiTV#_LUf(?mC11&D1Yt5HdU{9M$i8V`Gy#gG%pG%adq7!@
zjv==?WRWSw<!t&RGtjQ$H2sAKITqgi!uguXLu%NxmD)Iv=B?|4D5t(FCPS!cr6SMe
zmHLdUC9fe901D1SVx!13g=|3p(YsGsR#F$+a6r8c0v%=QFT^$imr~<dL02Z)97v8~
z{gA0p3K_XzXylEBWAa0{+77=lnB2Y`bzjao-P$ZXV&3dSMEvDmH{$4hi*xPD^k1ZX
z2yOB`dEhq-;x2Y71HW@DsJ!fEVCAW*d{}(E#k+j|@MC;KgR=`~VKEi{v=m*pr)chq
ztg@xh3b7QC2*tGN(sJR_)JzYi&Bj3R7&VCpt_amf#kzDl>HYpSsFhS%Mxb>H)D?dz
zzP;<V)TFlPsd{EUY-dKzEdvn50`>JEpiUzXuHpQ(q}b;ZpRt(VbB`0XEZPV{x4Iwc
zq*Hw=)!m+~FqL|FyHDlziSk!%KJ*3TXgfuAZLx7aVT+MyfV)iFTDzQ=HXZDHi8xLw
zTMbDABfSONWle^W7~a50^`9CekV`)dFKPaU?tO-Lr?PoNQ$p=I4Rt%ndsF8^;zC7?
z25sG+d%{tD4e_>uWGd)|sS~7XsEYC=)AxbeI?Gunq(5z0`!V5sc+0+;jO0q95&h~{
z1UjZ_ht1iMRc0fyBke|OopU_s@@^{<cdcDwvV&O7>7SSO-k)gtBvu>(Tc+9YLrpa`
z@c6{+b>^}IEEgBVM>fIJ659b`p+Yvg^G~IzHVcPlTIZZ2u?h)A!)wq_<pm^kzMXkb
z2W~UVol@K6;)`{xkx+r^=%p(hYGl+?{=B4f$drrz7=~0;9<XM6ZQQ4(GR?k{S8X9B
z{J^yOSqXPVepbWuGj2?pGcSBx<YQZLN8yp<9M{74oB+x!z0}(&ZT5IaxW&Mb%krFI
zdZ?Nz(IP?)d%@G49X@=LR#bkMEJklZf|kb+`q;0%T$QZ{=7D>@qiFOU0IAZv<Zh}z
z9TpWH%P64-tM{eiG#I%?-a60!anb4Zxi|%0k8qJK5MEAUuhDxZAG1u_DjaG!bD3u<
z+w4J$-D#^z2XE|E=oW;cd5Ld!4K*KBQe%z-sH8yMW!~4`HB1^#jI&#L=;u$8omMBU
z81;zwbdd3<k?lZT`vw9Oj_4=3A@K;TAEn)EncyL2<NW8P>ryV`Baim`fE`(CL9{bP
zTt&x+qGqAZk^aofkh*LD-r1^-p195A1}(2YO`KSY*=$zd0n?G|eM`Thjc}|<eYPG=
zx6KPTTn?H6C7X^W5h6#cLV{h`6N7lZF}{1yhqZ9Y`4&-T@ArSO?`m#y+gN?-oSNWC
z*X_79t#@@MFBIHl1@0ND1T1k*6jDioY~WVdVPa=dodlF~_3d(<)MjdZRYWACF?Qca
z7BWBTmCAyw^g?zty8=eRJvQ(RpgGo|#%D>fM2(t9<GM(SGmY`Dj%rc!<HG?H)tI#~
z5L(Gq`XP>AdRvS0cUdS9!0I+{7hdQ|9<sJF%9yd0JL+BSoF9#v<Q8h3y(7s<h_4af
z-8<LXFA;T~TSa>$=FOxYRq8^hc^km=?!m=5Q9}7uC5n~bbO$HQO4$0N9h4#_y<Eqz
z#Sv>;ll$JQD|LPKg(2vaoebY-SCQ0dmwwO3z8K5?n02s7ZiH+^cD|@{qcPc}FEjIQ
zx$w6Z{c&aM&L{zpE2t#0>&seXo`A(9(8v4pWA?-4oQs{hk7sgSBscG0`q<^4;5LhX
z*ja+1Yb<433#5Caj@cm%(vRFufcqXQs&4dDf`Z<9FMMpHXOCwjGJO9ViON+bKt+Zr
z4p_m%tMnB;+XnQ&YsdLfd{w9SXkMHNc)hI%)Tk%JZg=kTkK13++vgw9^nzG@!HD^h
zn%Wsv5n5<p*gG`F%Op=LH6?zvhEv=`lfh0XNpG7(0(ER@r03;3l`^1iH6&*t{G%B?
z=gfwrE~Mh)bxYo2ktg7=^=^WD{iaETqN@_7nCMA%FY@N~t<h7YBYp35bZSyNDExue
zNREB$p;sp>n=H<s=3ld2Arz900|4sk!r?Y}-;l7J9<0^$QoGl2)0F7FPpJO5P4HEo
z1V7PyU2dNx0ZC~t&oP6&T;f|!7tPQqWA~7G8I`E2N$*U}iU-z5jH@(r>Ny$TMx-CV
zL!0UH7U5TB3o6&idqr{tOx2&4@@>{QrHX%;U8YZTInn#&AVqozhnrDD;AEu)c>QhU
z?|A)qkNb>#t6sM3fMUPwLtLH1zri;0vnj!q6S8xG5rLH#tn5Z(o62ar6DH*(wW<Mu
zq5}6{?8lAL@Nyp@{a!5s$~K<xLHeDTwZ6_&kYUKe>jl+u&!gQ{=XsXgpv0OzS4NL2
z+C~vVf&H2vOit--bgo5?kThF_`L5q$YGU6SZ{OW^2?+gNGNX}i{dvx}k%@6xR<s;b
zaj&;W@nm_dtSrioBt@}EO=uUR`V|!o6^Oo__MMHET;|AG1>vh#8};qnZC9&`qu}Bu
z?^F(W=eT;z^5CNlqgUnM%L=Og66}%JugcNAtC?w$HGFhqRdy~v$3<<q=X)ziwXLC{
zcvVUgJtE=A)+9MO=s4{1yhBmCC|ejTiS)IsKrD}_3=VM850DxTcxPw!o|})BEL!A)
zivW#QId<eoF<E^$aKBH2Jj%!{vdws&!8dbO?a^B5bjs}T4E+Tu94F+Lwkf=6xYEiU
zEG$_{r}i|=RleWT>-1GzS_Y{;6dyR`z%klepeJ`0xc6b(d`NeR+%Oje4^KAk&oRqU
zIoHcE83Ihpotr_--VXu#Q0D+-Iy(&+Y~hDbP4igi3z=Z2aozA47Wm@4XdbES<{-2x
z3$fdkp~BPrcnhMUl;Zu1qA4DFkH<s(oWH#snCLhLmW7{tW%cBaA<kj#Y6?E@47nNY
z?`-k3rpjO|e*%T3tTtbLY%oWROkfzwXuo%Qc6F~}`}JV6?v$CsJJWhPUf@dKrvY+~
z$(QYYzx6Isv7PNaIBaVw+lj=y<I?U)Vq_w=4J62Td%yR0^`B{J#!nFIJbogt%lvss
zt+$AcikS3-df)}HHieji{7^JycBbjzb=i62#%qS2?<F)jO_A#81*b)6&E8Uro;(~m
ztZSc|PP?#t(p@5GA*`Fjlm5f>s!Twnn#e5YIiOK%FLG1@+&%}lj7=_dKIdHu&xhtB
zq$E-$zD^Mv(jn5$ruf<h?^TwflsXWTs)GLF?HYtr(`~!CMe7&FJ5*AZu85++_*i3A
zL$6fNiBK#qTaG&$V~bTRA4p|#S$}Ie=uIe)U`xfzT-fsQcYGi7;j!5kZm|@xG`i!m
zRdJOO@H2nE#92#u*G8oKu-)946s!13S%!D%??GrKDL*-iv!26|V`fdc1kB2+YpSja
zc^b@a+*M@MO{(exQatS}>r{<5&KWgZLn`S>U^n+=g{cg=rtxnXBQuz{k$M({R1YC`
zx(If|NGfLIzIsj2g{m0jE=f$%kC<9X9!O~OmvvCgcj{G^vcejA#;dG~<jgVdtXqs|
z`@$+J{2wcA3KbUd7BYkGDR+gqy?D;3uTg(XMo%VwX)UC@Ir2hwYfD4%TlB{OwQiC)
z;S6@0#$}=|ceT2z_yy^u_&)8FAF_i0^Mv?`uK60mHhppEY;^Qklz&rE)m~&Ob~JY-
z!Opy`F4<eEpsU$fb->DRIu$~+zyWMB!kLfqqX0Gz6C;$N?;+zE-wDQQ<?BGjw5~p$
zXCzw}+3+Qj7_L;FKlZ^L227WmagP-sZVSepH{{u;C^{0;8$|2pdbIr&*i9%?P#pFe
zk(#b7Tw^0J6To%hIQw&u0gh}6(S@v6L8X?Q^#@&}_~(3f%~La5N^%+qT7*b=KgyS2
zQW+^}yEJMuiz{bRt)&KAixE+JpaHK@N-DBmv8oA>GsyV>o7QnFTCb@B@5qz65;m2)
zU_68`jyyGWdT^L3TgfGLCpY%dfNCx}@8P552rl53C`Zjq-W>3}`S<ublm$$h5x9x|
z-4)!1>rZ^`pw}8ECIW%nh3HC#drska{nkVAwrH5SqYEO{P_sOH_3lj0HLk_eq}67>
zqC-G&IKi@^uP2+=#t~l_G9ME94Qn3xwY%aL$x+I-l&vJnVQYDOG;8Z|dZ)+u-rco5
zv9Q5!guqUHX24kJU|y%AA$O7fp-Og;*0IPUz0<Wf7a1R8GH8y1KlnJ5?<<x@6RKya
z4k0fTXL)GB(%{Lm*)ZX37KZksNY1Pcqb%wv`-WV2vN1(nOhspWGQ343wO%r!&l^%+
z0L@I_bG3p*P*EFtgXK=v6RR~~(A{kZt|^F^;#wH)pr-V<gIz4U%Mb^jisw+avJCL@
zPtINn(|ZZili8AQn<A(FNOk0S(<~M^?ZE$~hC%HexH`Q2XIpax-R=7xK6M`tSwbc|
zHaX9G>&4x6_RVYWB|QH&>d$_SMVBBYCPIcVBTl3^Q*cze4$K_x6!r3>1SjJ)=T#Xq
zAR0XJ=p{dXMo}6MUXKO$)`*fvf8aM#sWAlTC>xn5-#7k_(4R%#uWuGPnDrLO%;Yp-
zkUV=4@!amI&o&Ti$@T)bq*gx{N7L)HL=W0?GPCRF@e#XD)rOAzPoADXRCqKl@7j7^
zgmc!G_eLU;s?w<u-Q-qcZ_Rr+ZY?|Y7&3!xI0U#rsLu|aVLMeOnR(Sqw+XlG$oBS!
z%|-3plvUDf7-&`_jSLCJh_o}_RewQYhzsN$B+{tvbzUCQ;mu8LlbB@8ZrM(vBuwXF
z1jHr|goNU$u`&jGtM<(Dc0XcecGIP)u`+DYO@~yjDdz2|({;D0&IT$gi-*CKTpv-#
z&rJ-EG{ZxP#QN>+0(5pet`vrULC_)U>)mZT*ov(xk!DdNJ(5V{$EPXEGjuRZN()iJ
zV;zcjCu;sZR0jMbRK|BRyhQD*V(YFE;hOyLR1xd&;v!o(+U|Hpqn@H&Sy~FafO*UR
zT5EKVwpA)i84~&CbvS*cgxd2jEwo?$U{9v19h$duXL=_l9UWHust0PaDT>WG&Lvzk
zYJ)9ZT{U?tW?oXs`GdiLhC#^b@SKe)mNJpeOQGgJq=y)En7#iIDl-fsC^QD@7MG}l
z_~dzC4CD1xW_z1p|J?x7n+TTZR1|CTXXt#7xDfbbP6ki^#Hi)P_YL=GraYYvbh3z~
z%)Xsj*g{W2b*Q^tf{auA4zo&r`l?Je4}V9@`pJmb!YHbwy^!qEq-{kea5_H2_tjRy
ziz$ZU2VOAa5*4CZjj`%!ylO2vOdtiz7h-RrU~VFlQfmw-D1RSq%S?noA^OO6oU?f-
z&c9DzwRWj+H)AEJarbe6Tm^ek_$h}KM7L#Uq-D;n94A#w4VA$vO6gbhSYymy%m%_I
zH8-3&;$Uy2JHh9CT#0PP>A&aAIkE%Oz|u4<SC}yeUQ8%b^n05Ke=(RxFlH^>Yb6Yq
zv9Ci-oIpSOo%^|+WmC|SRN=ZPtj=$lEqyGiA+ZnT#>tsOO_Hh@gNBy-v8+uOxL2^a
zHC^smj8%)$O>^K*gV4)?Jwy8o2I9^hmg0beAB^CZgjBUaM)e_)MMJbspTgjJK?6%R
zJla{&?L5;PZjGCkG)pxcP#Kh3ZVp1{=MT(wDh}pT@%NMT6a59Gb>7s~C3X8uTjyS3
z_<(F5P&~zFukiZ<w6_AEAE>{U3@;F0P_uD5ickAPxA2IwS2+HGyzks>{`*|-NOsV(
zG4^JURhd&O6I|zd1ZQ(i)j5DVOW6S<urwXwq~{DVLeAaEw8`T>+kql0ql?JG$tg%G
zpPe#rL&|t`cw?Q-g%0d{G-x%5UxHh9=WrDzKwOHFD0<*#a|sAUViYFX$I)lxIohtz
z7vxC}j3_-!aE)hk?{-L+w6YaCUu0kSq_B!RWwqMkpXI&0OhtE4S@nF?jvw8YLw@Xl
zp7uz0zw#wP4#b;h^grrc>Bp3oCtmDC?mXxEkKQirVv9EHgBx=KQm*~g>t7Q1O9KD9
zCE%;07R|&PLIiPM%`we-7c&G_qXs}~BQ)6};^IK!pbfxsS9na&r@sZ@0vokYDxgQ@
z!Gzyv_I{kc_Myv7!5sL0IS3m0%<j8V>BW2u2?|qQr~sIHOC}(M-Sz1iKIW{2gd`5c
zY*h6*KnGpVh1AI?5k9_}Q*NF9o=?rEAY!n2xEHyfH1pW8>R3i}Zr)yhZO3iHZgT6C
zg<{^M9{hN`4y=Fln8TK~^?S-M(04W|BizO?MzEKdD)h;v`iFM${3v}J^^-Wz!;Abe
z^)W#*q0iN_>jd6VjKqvs?DiX%P}j_|j03(G&Eos(P0NkD&|sw4w%E&+eV>%EBPtyC
zm`gF$i~r9{p9o-y3z`I!>xawsV%P~33%AFk$XDO=cSdy*HVryNRQlD{;*jEo$<Cj^
zm|-;nTEHNpYZz~ZR|=%`k?J;A6J8{Hb2WLZW%Ke60g)$f{b2Np3hgpXcDg)biW0l@
zNZHkBPq!Vz{I7;V?uDOQ^92KT5FM2)%LDy;`s;lC!QcGt{h~tmo0-(3@(p%U?DhDP
zRml>sQHuAkskUks*LiF#1g4$-;^wuWFEzL5<`S?Hwid>hse}k!%?NrF#hu|36?o=-
zI$M?{+Fr0dCdp1K-d&okGl96ZckHn^wDPdtqAn8{aO|OE$;dS7wb-v{-?QntOB#hi
zJa;=4P&(TK^aET0>h<*W^o#WOjZ}h`_vXP0ubr2^tkHJ$(LE5-MBf(&QF(%o?<2gf
zW9wH^sH}?F1v^xx0(3Oa#zr$FQzrCeKB;<3;JT}We~I!7tvl%i$1I-fnmhu){bM~F
zgE#=-;QBj*?@~D2cEjg292!@+ZP>bs!J*OUKQ9S+Mo}G+#B;}BFBNL0ra>;Ng^S8R
zFU40pgDqLyvoUT*O(-Op#<HocRbZ{6c3=1=k9(-^xzS1N>N0#;*L|Rtn1wJfel1ZD
zjgHyl-m9Bxa2R52?XAW`k(APFdtnrd8lyq?X-`+@Er{eGjFHI@CaF$^ptKtiRoXh?
z?0X(q!SVWG2t7p7MJEJ<3`t2jhvuoYwH>5nI%((9I^1k6OBs6btu_M`s+_dlmB_-H
zh0BWMeNxz-x^oW~N}r!DDYq6mYLJIXGu^8zQ46v_PHfhd!fJvoMo=*LIYtqdI#Hd2
zE|?~`hS`#-`JMgtJ<aFBxU<6w^TPz9WUw*=wMQwn31nSfppvAd9>i#VoM3~&ZJk0?
zafr~2-5J}A-K{N;sKPx3+5PV#*3DLktDF%xoKwX=3`7vS+)bhDm{_)_9M((Zu6hkO
znkPFHZjP~W*?Zm;Duta@dv~P;)Pb9g(#t}Da%MMpJ43)dg3<O=R%fn#ra{#n!5a54
zpuUkCmF<~H-^d?_El5c@($W4xU63=r?hN8(dT%opU@D{?$89rquKSovIaK;?;>UI!
znaKL8!cMTDo1nNfvt7UlJ}|1ZR+94BYntM3PTp@AL13IMsjliiEeZ}U5Q?29P97Ff
zovc`@pEWcL5xVo9PKt|{6QsrqloFBD0UOPmQ12Vd>w_sIfW<%@101`{O8did?}yIJ
zqB&pdm$eFGjy?OZF|&HfN-*Fb+IhX`o#laK(Bd~Sn750?{D=j9Sx-?BMTY7rcyt;K
zaBOo>`Lq19=;^ui>GxICH-o~oj1wOqw7z6Z6bfnz2x=a%8O1NDslnc<Swwet=`%_Y
zAv#0|BXtuK6I&D$o8eX5rbad#B=3-f%Q9YzNcG5~_spJ}>q)vV1>U9SL+=0`dNZwW
z7U3J>$?mwKqA0l{95O?3iVRW+_>VcYXyIIwYOb3_Coy+-GO{+Igpg&2FkpTgn};yd
z8Okk)6yz2c7gu@$hUinrD8vY=O8svLqVjI*>{53e;2-4MKQ?-?hi%kvM=vJU)^->g
zx*Hk|{9d_zE=wZn1~gbeue6F4$zNmMI~5LQ$;D<$j%sJmS1duFL;|#<_|`1U`$e>Y
z?W{4I1d+iyRGPmG=HuKb%0dT;$hpsAK(5>^-_Te0<FB+80#F)^NQ$f2FhrO`@B3cc
zE7#cS@$=R;wo`bxW1THqA8s8@0YSW0usE!36_L2p*r|3h8^+z~++IIPn>8v2lfE;-
zFn6N0DvGVjP52mBCjjGCLM}K=sMyrwr*N1TxJ8t{9u*1(GXyKB)88e5+T1wW+&FVm
zcGT$y#OSHM$f%&hH9IvmH5EJiuCBAIPv104<&=G9ebg!(7ZIsKdf}8{{}uG`VF<BN
zE$A}L8)z}-Vu4@iOc;eg+y*i^4HZ1TtEi|as08ZMpJu)}IbtqbfKEIOwY3r<fPl0|
zL}DmUVo+TneeQz}OJ`X6Lz%a^XLdrJ-Cb?Xj(iN~8y2aO?OyI;n^a^1ZoZNFYf@iO
z=TdtBQPmgQGKAf=$S(5c&t7CwXVQm>L73F*amdin=r?W61<$OF+mTG3RO+)ay(b|t
zIhD+y>!LrIS-*LJ9kP6w^|R^%%`q!0)Um6l*`N2kd#%Cmd}0e$2d&0!THtVNnUN43
zY#4uyj=oFSi|mv+3_n%?%RPOUL{A;U!e`0adnAd3;f+?^?-qEEo5~S7-EM+3<2tny
znj$5I%!xbc&d#}^{ASFF4<2f2#?q~8O;46St+Z4MtvjTPj&>KsE=CHO4p*H@k}P&1
zk`t}nm8wQ$P@g^o#7NBml+3M5+F1+RDSNr&h_ykKax=HDt)}dx;1-#5bo^x!ztM5X
z9mRftwnbj0l`@ZHxGTjL$?(w2K*K&{Fh$#`qE}PPU{d74^{3b45@PA^tcuPyC;Dyb
zg4gR<SvH|w=ck844SxF*OGy0UWDPI%j`4+<B78k{9KqpmE680#7tD)n>R*@*bqBy8
zu-qY7j12&g5;aJjPF&B<IitDM)l03D>p4yoJv2OW29lMc!F;`$faTuSuDYXNBw%%}
znNatJfNk<1%xJeWSrNpsJ>U{hcZ9X!>jwnYhMsSYJxiVo8T>Mrp-h{#xVu{aF6KQm
zQrw_k`(a$-kUm2+I#ZIKbys{zD78iqD?)a!FUXI=@YZ87K2h9Za5#5J26splz~}Eq
z>cRXA)YZuX<q#+^j^l~Q%>_IAzMD(xU0o4%y9;(|uv~#`T)}^Tb)qrIDBCC!hmIFz
z-z(0Buc}SQA1lj_sD1{nm~{TNZtvLy!y6XRQb1gsCJP7zTI@w<f*5}!Z6?F0yX`&*
zSrS5>H8X{13g^dc0C$y&ihs6@-)WeZQ&>G%>50Cnzo8_jQBOnIQ8lV!@-dD3^FGAU
zKy~nh+z@HwcK)>=czHrPc4?8Cvv?n6Lj2s_6q*PxJsPPHhPqI2A0?oS+xlL8MF(pH
zY9^17q9`GPSD1@AUXiSC0RiiOPf77fIS`7lk<`d;7pv#6iOmynCfN?}?-9U-Em{@^
zLPfS6A#Y_8c!Ug0x%GJ-;_c!d^cAk1d?cWabEs-sR7s|VA=5I_w@&srQI%2BkjZWl
z)(})c9Rw+uW$IpG5gg$ZVD9F?&egA&vK($KVqWPYyE`{skHE?vY;(@6Voj|TubMe0
zcz{-_-c)BG7l|kQ06oxx1$98Q-2VkN>h6b!Ku~U-5ER(3vjKy3h<=>};O+q^6@;7*
zn0u(I$9{G%eSpP%Eb-~qgIbI#SN-uV&M3|>mQVZ)m8q(w$(9OQDO7>Ivr{VORDP_}
zMUCF*Cz5swE;@ASz#BSY+$k9XP0_H1gieXJgQ9*wU~DJnq?GHR)Yb&SZ`#spa}vAA
zxKFZx=o}BE`CDe*C7sb+L2BPqVhs6KtRn$5m%&Pdqbdb7BnxlGOsnv}2-J|akSH87
z%^j)K<2I{1pXgthG7oEjN?&a%ajX$mX%ENr7o``DQnUJoWs^;7w#Y8LR8UPAuWT4~
z5GjYVU!%E+DQszNZwqTC+#5S6R<-aDBQqCd^F=vC1@3OaTJr++45i*h_1FoqvmsMy
zZnm0^>ZZczjC%HE5YlvalgHgwu;21{E3G^F)DrgUc7Rr}NcG3e?PnGi8$>53d}f~Q
zSa&nLLr+gkuaw|Y?>w`*<QHf!=|<dMn#1>{ChEg_jf?QJO89yT31Z~2JP6ST=_$>i
zqXJ8vl+^jJYzMyMDLX?5gLoz0rP-+X-pX3KZ$FITv|mcoN=4=0$$QW-LVZCS#O0z{
zWAM|xNiO<c<Q@u)0FxH6pkm1^YN#5VK8%S9pp+z0(aU6=HflT}BO^y4E(OPn4#8ov
zZE#fPhJuti8ykNe9j}xO()}o@x7lW5<}ML`G-y<{{Ict)9MbUi)m*IemwD}VHaBJ+
zqZ%<G=K=Da3ZI0x=LR~76Tq~{^N$r}IFfq}meNhaVR3?_txZLn116(UJ?f{1dWaV>
zf{iZv^ren4{(26d)naoe8(1i-eyE0RF(b<y8n)DOF@?7F9%ofIYK9EoCB6o}aPZhG
z;njNISL1eWAQ}$#+nGt9R=nv9rCIY5V_Vh|-bzj-yMA`>k(J8j(O9$R{*e1Xa{<8(
z-0WDI@~MZNGqS&Q^qByvLx&{7+3<nsz&|e;=|I-YL}v3DsJbnL2{l2D<1WaQ4M_a@
ze5zR&LR2u@DVWdUZS<G4%AxX`r;=2r@<T&Xs6ZT!Qx=;s>-MZpRzxSexS@+km54J{
z(U%ZVK|YB+h|TSdO@4g)Jr=vz1Hd8h^@vh}5ecFX*6(B_(~B`q)=-%kj+6?iY(j7V
zT)vIy9bJG*veM)^!1xL%#qx{{&1oF7dWIyasRTsyX1<R_`?Q27PG%ILMpcMzeN14G
z*n|%nYvZ#7@HuSLD@zI84avj<g8lcj3=;ALs4EA6mVt_jCIG0C5bH%x??umI&0_6!
zow{K7+m-=`J2Vn|tm1My4RcTaed#pse09|wsM_7Cr&H%44pbkXV^W=rRhbM=Y_HqR
z7Q*po1RP`eu)aqvR1^1{?BVw8R#3~EX6BsE=%KZH_8FPLWTPSWRrlrIyf2y_i^OA(
zWvYt#J#4+feVMCb(OYIYpUTQPpQ$8=^;wQO^WFYC6{9Tc*V{&mOzhPqGYw#)#YvsJ
za6A-t%d4bQvrv_j`KH}s)*l85*$7dF8(R8t8bsnn4~p|?gk05$M<xF!{K9nU)wH7R
z_mJ_B&^luYC^NE7jGobAP|}T_<G>A^3i1*K7kT(`V}X73@CCy^FEIn%$n>+l849Pq
zO$A3&MZ7@F>M%#>)JNvaZ(4h1mEy-$ER4gHRCYcV5hk(E0o|=9gaf=m^cwlRe0QtV
zrDPx$tDn8#qUd)K=^|5;J44>5skf(}B-~?KfQ0J=^(ht-?vF=%DrtG1-4y2&j1blm
z$6btoWkwKPML*WBEPVX|YX<`#p=B#pCYLT<*+wdO-BN3%t#Di<Nnw=XA1#hOBX0nc
ze`v@`76Uy(cuH@YYH#A`T`CS(2#>4XS+WsC&yO!|Mppm;3fbIiA<<LkZ*K(MQj6$G
zz}l7#4O=8(G0p^j>N_;&sRps`U<dU5AW92AT}RIXw7_%sVAgR3(GD-D6elxVHUNz6
zzx}!g%|{A}lXN`~b`F60sNcUOvr61Wy{r(@v59LrU=KD3v@rL5X<0@MBvv1md~oCB
z%|W21`<9rr+9GN1bVBXw+~}jK;?T7&s-Mv5%vtH&if{@oD_+o#^a@7e_a|_{$kKq1
z+9~&>OATYaX&5$D>DyU0ZB6m}RM|-*vmLSW>Mw8#Ia5D$;=i+}{LQlR|4NBD2zD-b
z`$)Z+?DR|2zxs(k_4Rl(`qO!_OS@_!JW_qT7TTZ>8RA-ElI7H*N6_q{3+ms14s(~0
zhnt@QO|8E>MR#4EZMeK||Ne8u<6o}@ZCoQK-M`v(_KNeg?7p!{06X{2CGWqQ{w0CG
zB=DC6{*u6768K93|C=T7({YD5E}))fL;~Nq@nUE6M!QnWd_ktNaAWqv(kc?U%c)y+
z3=X&CzU=nuZ`wlrw&+^7)m-}p`I6|Dzn%V3;`${IX1{13$wT&}33Ci#6r|up9(0x!
z6N<~C7Bu%q9`RJ~nSS9@^&KMtG!bJvw)OL(Qq?25nS<}q$@z<O%U|v7o-cp1Ga<J0
zwKjXRi{2Jxqez!ml)F2TRW|}6{D<|u+D{>aUegl-buKt0mY2T)Ilb;J0;^G-lzxFH
zPeNZ{7aNi`4XKG)7MWB?)I}i(><M86ds4}G+DdLI43&ru$%Up|Bp^Os!AdYQ>ssjM
zl8EwLEjZrE1~y?nXI|rZ)P7m+xI7&#f0VB8+a8M-ZYGA<VXtRVTOO{rCYwM`Yv!Gv
zn~LKF^ERZ{k(qUbCOyl(&X#P-`kwW84_UmTbfP``LiKcbezFdXbD3>&=MUQ)g;)XU
zQu7gBMdet;kr^s%6Fjl;b<U)X02{}K@luSpDU;bn@v3+AC5d1|FxB?iE=C_r<96`s
z|1RPG3P-|c4JStTzefk-#^_vR@)znj)#lw+2-1fTx5V5O(pogkQI7GZudkZZTRE()
z5(&N*Gr^F5i#JE>PCV#g>>cN=8|(g#>Gy||DHLI6E=L1RY(_;YuSW@C#WFU~&_xqP
z&iaT+!2h8WPclgCMWbIO{`#-Ug$>b=S9tf!$Xfv#uw+&ql`$pGLGU($KO!P};t@T1
zniOa%G)s5kbIlvPZ&w;1fBcnmTACc!^nGjjmAB`tvDRU0bH?Srn*Ak^zedQvIv1ql
z)w(jPNO3-)bDrg!W$A#IX{#Mk$n_g`fl9IHy-<~a<AYZ*1M)&aolg#PgX1slR@7{G
z-uF|dVvYVf`IkWc8VvtOPK0~W${OMKyEMXMnPE06s$T({f-@>AS0w-L#r+Qy%fD;&
zU#R$Z=C*5~DU-)`SlQDrw)b0#7p~5~jJ#+yOFdmm{%#!dbNQ=(?XE~7Gr^{S;MkJK
zGap_>{PwSP_7ARRpQZB6jl!f3$_EZjK0UPi2TlLAs;f+2w$`tzUU-`Z*BQU+`|WSN
z{XY<Bf2;bhS^6b2IaurU1^w2(`DNlE^X+kgsGz-jo!^UOFV{|S2bYatFOKuhn6fVm
zKmQ-x;A6>p5?p0I@&ox~$ZRP7@rwle0m`NG|3T~jbQ+Pg^|(0pao0%Z*0bN#e*I6i
z|LE{0TPV*E)oIbln?nV25l=5M<QCNBsbp`n=|nUQE&V3=k1qaCbyxiNtWMUkGb6GW
z>-T<+c=eyE|JNP1)N`Ic?`rN}7d;gG!dh_Y(l2*&^LX?4slWtGLJmMnqe165)0&<I
zfzA}=zx8b7*3m|?Yqh#I-n?!yD0VttZCfA7RfX8(^I5l_9I{DT{x#Hr`}0^5KqhDV
z4vnFXPvXn{5&saA%|urZN7hYEyKj<LSUdJfy29cJm-|W@(NQ%qiKH+n_e$&&rDX1v
z-G%4UrpaS*L-(V0^(wwGr1~b`S=ZJ)y8HeP@@dGKnws77$Iqy@&tl1QZYp$U(me6u
zmy`c>;`slrr=JFToAaot+piy<l)PH<55KmN(D~Sp*7gXJ_G-$7zIn{iFgRf7;^yF`
zq(}euE!pn3rFxCBKb+7FE*)ExRC^NlOed?LbIsK<5L4%soWvhC^bjY#wx(sOy{^SB
zE-An>p`vyLL}54Cl)gbYzdWk&uHu&DW(h)xK67$-L}b*ls8}4y?-iHT;<m$=%=F$)
zpxS7aq3;GT*5h*dum2$#*@}@qhPg>Z`|{7jV5MFVRtnAX_%m;}JF~;o1mw1+jSk+@
zEo$~#)6|sa_`?$~%D2v#eHHXtCa=(-Ft@;UbL*iOKRTn+Zv2AJM;g=AQ=*Nn<=&=Q
zlWQcYFTrM}UeWRusr*?s{9}3kL}vPXX@O4@24unYfd<2~>$F2xO78r}m-$BpHv5Kw
z^@n;plKH2@cj!a@QPaQGU4crMB-y0^M^O19Z7#ge>=Jy!mW(W*B_hP%`HO$8*6TEj
ztI7*B57z@}g=gn%P%hK3zi9V)@A6tNW7tW^!&nkUYdP@2$iK)?r83cS*a7B|-RCz*
z=Fj~Q6F@wU-Cdl!^=>c6P4QxFB<No7-t!x&pVwUx-#0At`O0=@x$)V>Z+%~ANiw}H
z_KB2~v*sU)hk9ygM|nu(4&)KL8ZQB3*q*ex%+0)pPjot~I@)SGGVlt!IpG{M@nsP%
zOSS7j4&Rt*xVECLuFOzh4I|0f(OC^)5YjsQom5}fB^8=qojfw)?d96rb>o!Xcc(bN
zQKJ&Y&zIMSTBY+e+!x@f4P@8=s$2>D?Qc@~pO##MsHq8Io|6-ukEd-f#m4+kFZlNc
z9F1#9EiwzUYsP`Te{8uZ^0K(5*cqFoi!H9(t(A{gSL7OwrYgD1k~uW!9(}GMS*d3Y
zt_me<v^)fH{CO#ShUueO<ZDyc$D;mFpi4QM#>ij=91;#tvyB>EAoR<ML4H1#BC<2<
zbH}UYHYVmN^v1qQVC(0a@*Art4YEA71#*M3<pr*#l(blQ%{se2oRs6zWk}_r|8N{v
zx2M-~rRcYR@<#rzX7`}V&Bd%N$J6!~i)KG`xk^dKXWvVBKD#$*4+kU_-TC~|Po6ej
zup6HzO;?eDDSz9QEN-~TO3&~tDjdw6yZ3&0%7;?BTT=aD>1K}N>?qep(o(A$XlmEm
zLTKHogkNZ%@BWh!mn*1}-!PQGh$(khFuFF1g~zFW$``d`EpYv5+{PrIr5B@h_4qX!
z;EgFUDxoWSeWa4Id!@eZk;$WCP^9k2SS8{0qb^~d-sqpiqptq8QQuwT&Fmhd8Ey!s
zU1;2y3Qm1t9zx^b$`1W{zk^0)$7%gZb$brfgm6`7lRKuT?Im9oRIYW$_<3hp=(Mnn
zV{VXK?x&>ow~nRWfu18LFzV22-Kb`x6tKI@&R!wQmo&%@!?b2HK5gf9SCC=ZNK{3I
zqRvo+NR+#<`>h8(xevN`zJ2+uUCN|&qt;nH`ipy>fZ5U=Pf%6K$V$il9?hjGuh(|l
zJAq0AYl9-Mc^))4`F%^sV)4lUTWT2+Sb>T%qd$$*KVg_eU61K*qEj)weYK1FNdEt<
zKu6371gLKJbDUgbS5oB+a>Uh<E5|-2d`)7e`|YhYoM;&D7&n6#*LB~HFqV(ys*Jug
zj<%NygeZfAwAy=3F5f&6$q#xF?^SEByRddB`f4SQ%l4M6@r~c%#x!kE*Ju+7^su{)
zbqYRZk??HQ=|kfnsV=z>-tVcqoGD2IH}tywrt|yRuo91fL^#yiwOL&(<LCswQHw#i
zMOw{E9-2*@2{zmqkQ8jV&-eYrr<91qRip}po%>7anrI0eQG(NNroW))v0TP*&ZaME
zr`Q^<NZrrBvys=<@%@ZqJmD-e$a?rpKh;8qU6jnedff@5($I^UaPkXyBeO3=Zz}zu
zisfpHPnc`6vh+?Toi1SaIeb+1O|roENac`r57N3g%SvK!sFuwMUc|sX;=au5TPylj
zRRs!3A5Z#=v4S6kn9Js`3ht`XZY^=K$0{b+*ITW!R}DPEmeZPH6byzP#9!>EdjJ<)
zo1*~U`fSrudY~y^`VI>RTb>aPeSzz=>8&}}=Bc!APMfoABO=Y66F18P0aT86t>RV)
zrY5UNRT+o%Q^pvb7cN{2L&f@$?ZVO1VLZMRDo^Olbt!0=)!n`qLy^dq1A+1Yc!TIP
zFq@Z9L>s34$T{|PomZY99$TH=SDEd;qr;q0mFW<4vMzQ~#esfll+PCP#qp|l_QeZF
zrxR~Pt@|`QZQZEYxN)a(QcY+!n7>TWY%gJ5BVpf^H{IZwd72D=T#<<kr3qFPSyf$*
zvH%+yJkx&nNhZ$Pzhk^_WU+#6rS|Pb;_!m@$dpZ<tLFPB)}e-)AIs7;3ODl>$>yXs
z_l`nzMZ)RLb!&~dt!D-^*9`|n+O~b?{=9@icA%;i?OMHJs~fKfc++D8b?S+sQi?xS
zz1I!*t7qTH;q5i~kgjQ)sOgTP1O3u)?`F|^P7nXkPaSBly`w3oKQDQh=yvY*X%41?
z2MALAB_-kmAUO?2x5hA@$tcx7=4EXYeaJ;%Z*#?NpyZ<M?wp>14Phi}B;hL<xOWtS
zyhbMyyPD)tE4M+TIDLWo2`GA-wvAV}^m9q&u)5Jk)Q(&tz0~^<n@5&m4ZjfB9+-?%
z`2gE6yS0%#R5k<APmO~v8;`;QTc9qYU<rd9?&y_dDlb=(p^4F{pltgzL|YdWz8VBE
z3mbRia7H2eTnuf_Y~d;w(dN@MxBmE+*Z$tkxB=7fZMzM658oku!%)1Xl(j#eu9P#;
zJEk+Dy-HXeke?kDVSp^<UtmgNPquYw`#qR;omEF3N<r6}_|ZuOXn5Z!hb=k_0yE)K
zZW8W6LWZep{Ta~FXBw6sday#n=L@J}-I_m6zc`G~eV;9U{h&~PB2>Uh3#9cW?^T<h
zm}v9?g&+#{26`=K`}_BSgo7UzlPbEuPnmJRc+-^>+Ug^uZ0r}j@7&KeP4gW$RE>d2
z%?W3BSjav~d(>~X`tjDlT!8mpGQx;0$J(mqEM)J`O9PQgs`lJ5Sl?}|RC2YV{58j9
zr>Ipz6p?`Ph#fSzaebv49)D|VZlLXB!+Xo9LETRNb&#Sn`39-bbCjtvVgA{JY{u&l
zyfunV%-e1)O#R_g>pR>cZ%s`5-rjPztystB0z9-O%=_3JJGf!B%!g0%CkzfWElq)v
zd3A(x+XL}+c+E7bPyW4aJv82$teHx$gR2KB!`tGtOMEw;fpzPjh?<L&XSgX8N5~H6
z66GXLtr6h|wE22BrR1Dep5h7-u?wY!Lszncu|gUf-K)_zalM=g3l^btN(CwV?sFB3
z1u7_}YZ!r9?oqSIYgDki`dgf5^wZj_@biz!V}~F?eSm`Bgbt(3u&IW2WFFnPgOV2j
zQQ|ISZp0)tq|Ebua8A0h(aH}ErE~dp?R_lrp>4S%eZN_#Q8;hObEuf}Pu$g2&QE@i
zvt<PXNcQS4_7N5pr>pS6m0lN>d!I|VUi^9Kz|skHmezcnZ15AJwrz9LtwCmDLs4ha
zQj}p?Fdbw3Y<hulyK|?Y7WFX6^5w?pvjiYuR9xHn6H@gLj&M$0&&Yh{+#TBUR)0%W
zvWL9%Amzm<QL&NFu;S!2aT-z%6-<(RBRBO&KR%u}((qB=3a!1w_it9xtO_|<px(O2
zn;k<D(h2{BCUh0;vU}3acovDc*#G%wn}6bpA5BR3U#-d}t(uANAJ3saaqqsx&U@b1
zs-Bf*8;&&l{3vyQ?W*$gu!#5vze5$D*{K$5GxSI#Kl@g;hIZrk$QjhnQc)j!&U1xx
zx_aDwy@%+?I;se2YOYHNfk2qh_7-&qj2zq&>fx+4!}9~eaqxs+Bh|Z(@7!XZV)kuw
zG?Hw7IN0~`YkMvxII=M*8schVk0B;8mgG5GNGF>>2khxHC8eAOzmY3<uG6fiR=Ev1
z@2TCgCZd{yC$INEGQZUV-}7t6vcjU>2fTi+*~*d`VxFLI|5jO<-Lb<qId*5H?xm@_
z?1<LGc%(D~a;-&KCVaA`dCaTGa#Vg~bV~pE!OQf;u2E(e+i*>5HTB-uR2M$9Pjm_`
zv;wAAJQ&1rXJ2bheCcxwZfElCQ|Gpd!XbTR!S@HK-lZfh``6pHA{SwOxpMgolQ>PO
z4<j)4#yMD%mNwY71(ixOq;PLbSZEt#kJ^t}(kzw$sF=73x}Al!pL9hR)!FrT<u~vg
zYkH!i1rpd|7-XyXYO<IbJ2|Eky&ec)1nAMNyGQ}D?wGy6%Mtf9FHQjvr)04d_)wyV
zt;ZZyn^(ij&@JH8s!2%pm5-R>{R*t^tgtg8Rfcw_vI~+0dc$_z1J=yxfg9|pK-#fA
znLbErZwcJux|g?f#f0@*J_8YJ-9@8rARfeH4!{?abx!r&Zuyh0AJE$_5=!tEv+_NT
zA<q1%f}LD>A8|Iy?+W%u_M`jWbD5!q_<v3|3;&hZN@HK1Z6r{lWncS_Z>QVYNR$Zi
zZlelXjKj1rbU$t=UIL2($1@5p^V~E?B-771yGUu8y#~Dc=LYA@Im<=;Y7HSOxW3~u
z_pkrlMVBtgv`n#FtTtah|L)0hp>Oj_&R`Bv#Cq{|L)qPrkXLu*2>-$Qo`SiB?XQts
z5A)ZSOL;&1Q8O7W8!E;L3I{5~SjOBq(m*SFArY@r(r?TjveE(k%-^hZkf7p9P%O%;
z0=53qAiY)WnmfCBMAW7(k1@2;wgaDI)66P3=G^fid?o6@$+RM`BIVI^yo5BL|HOV<
zruCW7o+yB<XqO17pS;zk*}kh&{d2Lcn4d1AtzLoGp3%T~x)ZKD3cv{i#XZcb;fWtS
zA#__w?%p^XxAaa#zWY+$U2xo2lw6mD$})o`*-MK`w9-w?&RUb1jYsG}tkt$vbslyK
zC#lxhi6_NF6AW=$8V^KYPR>tXW$jc}KFEq(3pkm+(jzP}B)NGh48vzVCK>!jql%y%
zKU?BH&d2M^c+R{009+`b4S$mx_oOz>{3ugnOp58<w|dja-v(;#7sovuK^#awnn`5%
z<AF$lWtbBV@NjXecq300pO=T|P}r_O{k)dQl>enH`le#_aSm<xH4&8Mro`HNnRKJ#
zB)Zq*rtL`YXX`i9rmyUmSIxkaKXq9GkP>pslu?6;Y2IS@o4eZ6mu5HV1GL@GH}sK~
zv5uRdYd5FqUXA$<7Dz(AkD4<6`j*XO*u6Ah#V|S>GxKqWo9uhbk&gGKgGkNV7#u+6
zdZe@RPmUB?p{&^v+VHc@ia#&8mt4lYy5xcAFxJ&;f=xT+zft1z=FPlJJ{6dJHyts7
zHZU#oSE=f76m+2TeL00#*|2BU)^P}J-xpFDBgvu&eGIKxXSaW8^X#N;-`oAEb8Xu~
zlqRR7V>`k<iT#dXVWA+PhVJL<kLDj#n?!TTYdocf(7@j1P(Y8^RNEh`OO@m08Aprd
zeUH^t|8Sn;lBZ&qJvKB`Wlg1M!`S)bgev?rAQG%b-Mfy<A_;eYp7(avvOhHQYZ~Ss
zenrx~B=7reJnK}M=pyrSo!P%uL~K~Zw8mS8JgtFk>e!KQoMp-{D1}_{IbXVwPg7FR
zv9{dK&%?sDmfPSUcIaGeWzhF3{Tdpn4D5^{xt5)3+b>x@wLmPTYi4z%J0m1yCPNHj
zs`#m4fD#1}5nb2B<pbN;zv*|(0wy*Y^j8iX+izd*maNfwRK;HYVz1QQ%%!`oKxR{n
zQ|)Dz(L(GZz>uko-J0MB>_B{()EASn|5eWY(xuC&IV*Bpb9?wX?vmAqfAXpQAKmQl
z(IyVHWkKCK9)}+~9=|p1y4CTG{rnvjchV>NwdBqX+smDN7!$)dZ{-TVjjEH9N3bhB
z<uOPl4R_<LPFo;|Tl@*?zp?k$VR1BVx;P052_cX`g1Zdv?gVFWcL~k_GdKh%g1Zh3
zmf(Zy;E>=1X9l<6PJ%o6vHSb>I=kolcF%j>v)6TG-}<M!pRTIzdip8ps;BN-|5!(f
z&y7|*P}@r8jR@|CF2TwixpI46pNlces)wSy{PS|=)#IcQgSs;`)(u0INq|qhGP8z?
znjh#A6duFjB%xs-`Uh$D0iTU~dn+^0N$Ur=)8QtxHi(3nSI9lOvZnySozYGfdVLU!
zk95x*>laAFr&PBAy&ki7q>X(}4uT$zAyev{5k~UGrV}QLH9BHk!XLJIuTsG-Ke!8t
zdByj=ycDx>ad<UJ)$|iN!)}fGyu7`8dM&(+`nx!zp6KT96wHxCY7{06%NkZa;S*}8
zJ5WwgYk{b}n=r>Q#P*BfCb1g_@oc=3-AIvmtMlaNKzLyD$OHNeu0V%&YqPD1yuhzJ
zm1Li-u6!G@RaQ3}{ETGlB0frclnu;>cyeS}C588~%#=lbT=^cW^Q`frdfau7i44Tr
zf@IcGv{93lu{YyPyYvODdW+K)aW|xHmsHuv`>>|y<U5|2SV-GjZwy9!{Q6HmUzKmU
z39%PnHMR}cig@^Nl+4ydZBCFTS5%g!-jQmUt1qxR_^Hkj-sA5Xx@#=7%I3z((UZz}
z`sCg3{E+$ZLJ;y3-C0!FO$-Gtl;hQ$^AjeGv>z-fJZ2;K=)nhTL8$nV*^USzSoMIh
z^z4!<Q93fok5~A5pt0gNPpoA<dItq!6%d!LH~?cA_gwF5=;&p-)97t`ahoZ2*jkTi
z*g@WSpGldF+=?g&wpw8-Jd~fY{an_9!KI#p2Xfj<7jh3)>5$IgDS}?n_i6y0d0@S=
zhQ^T4l1HNo!?lZq@G3r^PopZPChb;}Y^@zm7qoi>kJze9uCm7^FHZeMU8ReXl9Yus
zrta5c;)b`WB)Q8e0j=U>J~$=-*RtLS{#+SGdicHLH!u(4Bdak;?O9?lD&RJfj$KpU
zDag^AY8UDW)HOct`GTw(I)KnOiGk>bvPu&umtgqtvi230N)nb-Wqss=<%Nk#nR}i`
zeBtEuZu+l{j<}hl?J8!L;mRW+D6kGL&Pg6A4=NxsPcf2Kt{za~MS|$~kt$yp!H78^
zp86-ca3)2Ca3-4ju!r-ep6xeU6J)f)Nl9XDV)TVkuL(dRBA{!R#~on^-sFGe$abe_
z-d|m(Yv@Y9Ae{{-H8CPuF{a&6QVK<~%gC5df5fguvepq}*OMdC!67q>n%{|~FxjZ{
zXcO>e)q!+?M?kuGTu;|Eq7lQo*avfGDH~1Q@N=fm-zu361fXZkU^U^eG2|)NO0WD*
zf{mhwLM(x>ID9HY$`>mw&h0l^jEn!0=dV&Q#qw<xh?pQ<vMNn?vUn%?dx?Ek0CQ9m
zNQw}rwL`<*r-!;f<3tZSf8>Kh?e~H0P^oon?(lF<rTl;^N&<7zfz*$;jL8|3{ubA=
zH`Yo8#pIMYoIbt?sy9keG=S0@N<wL<NU1ksf1_D%bV~H!^SP#ZP>DRz#H)oF-|r|V
z1OR{lfb1&mec*W8_vA$8J{X5JKaKn<yFx?tN7~dsH>0oG@IIgu9Pk^B`+m+id<rch
zfb5!d?repZ!p0k3>*XngWl*=k;`Fs8pz*_d46*xPR}}yEaR&g<8Px{mbE8?GeyGUF
zPAX!6bQ~ZbJ~VcsiH^6JRG!n8pruO$gcNWp(lJfG?$F$Mno@4ZLvs(=fk8J>PWMWy
zn17ZQp+em>MJ|~aV9WxHYFbnFA?suJc$E-ppqOF&PK{7|{S(yUt1B=^1j1x5uS{DL
zW@FScU!!<vAQFc3qs8b<Oin}%jjJ$0!A8xPRvK&BraxuNH#HNMq2dl|w48XPP1a?@
z_)cv0nv-RW<LH`6Huf5si^PK{+gY0c9&zM@O-E@}yHmH&U{%;!D@fBu=Mv1iNwnp0
z*ncfR(1q}^vJ=Qr1c$5S%Z_QzD5}L1r($z|L>y2z6MJs&KWGn!`&gt^vq59zq;<if
zCYe@nqWll3$pq%Ipz5S`-V-mwzQGxPJEu6OfbaEV5*|r2i?kjQx!4nXH2Ir(k|4Fq
zxwos~iP&k+tC$#}7|KTc;c34pfyF{c;?Pe)G8{PjBJ*<oPzT$0{x-Se${#9*Ib_zo
zWb}AAmItfb`G>h036$c~OVeKug8~Ytgv2@2(rUB96z&3R%6_BSds2G8x4EwGC_gk#
z;Bh{%&+JG+y3y7#g-3r;tumhL+v<xNfEzhTjpyO!1I0&42Ws}8war0-R!GF+t(>_d
zpIg|dQI#>TDnz0ewfC2~Vyw#9Xr!+x7Y}uMs0K8iHi6b+!igMz{1M|sL&N?#7PwoU
z?yJ9fUHTS}`;R5M-xC(a_AT%r{TJoZ6Uo4J0!H-D!(6}7>JpoOF3uAE>Lq$mL}7~&
zlT#lms-(bon5)0E%bugy1VdVBI+}eNdh92w=zdvdO+t#z-RaklxOV}4spLJu%H;%m
zM|%<t0v?IWHaZ$dU3%hXW!dY|@Ej7~OIFtc-r((b!tLD5Mff#eCl`*7<udUA<Dozz
zY!ZS*-iz^CI5WpHZ8`Z<nYTiH3xRjr4iJfT|N9yQ#Y3|g=GWiITW_&A8;h~H%n!&i
zG%+X9!BTN^XkeSsv=m}K+AAF%mU#m6hb!9mm0<P^<)0)qv?PuLUt;Q<Ko2{Aqiw*p
z>+h}#0^gvW&kG6a9f*L^D<?9?9I}=<TMe5h27QJJG$(nYXEX*8?bX#5-*js<cQs1V
zLHX78^f|qbsO5Wn1G#)6HKISL<$o~H9?9FQciKo0(BeLPs=TLXF3doPBVDA3N)<2w
zY%fUQb~EJNv%7mfQ&g`g*)iIF_w<Q08ck>P$iA9W;d#tGLQa%sP*eps%eCfkU_QqE
z0RG#_#S`@)p{-T90iElvAAw0aUswiKT%P6jex|O?%WIEbTIF8o%Sr4>ba}ZZrl*>w
z5}!cEZLM@wFlt4P74%<wy&??4dBdOYf-|aUC-{8U(8Oyz64;0Vn_lYk!53R2W!&dU
zSSF}htzuFqZIKba^x#+R^+*-q8#7F3NS-9%o_PuzW2ZK@J8LaK8Tq1O$S%(i*7~yA
zjuOvNK%yQWCxvo&%mSFjYm&&r<UlNw9wF*!nc1z=EgH#K#(#6o#67tm`K`$(@QbmQ
zS9VqO)?1Oqx(;W*bkoC>TaGJ8w~oo<VJu4>E+I1ejI2%)kK@v=ioFs>c)U>{{ckj;
z_t?7C<%@L9n{MC8o1ZKM;;E9)%JbOqTo~B5l`f#H>yd|kDOCn?9Jp`+)mlx_lXRs*
z?>Zb?&90vCEohxLxv?!3Tsh;kSDKpy%Z+n3_O(B71ze>ve-J5eZBR%zaZ7za8g`HQ
zfTZn49DZDeWg8xGMI8-)i4@+U6xm)7jZn!C-jAS^X?|jDz--L4FONC6YF;~}vT)E>
zElMlTjGtgn0Z3ZCTo-2_D612o3s<a1I%sLD2_iX+nsJXkA6Yj9E4n1hTTM~-7&_BH
zLlM8;wOgRlZDz$hzm}48@WPN}-;P`5QLOU99oU9ya*p3j+@OtQvyNJ%{q}ZE5V;z|
z^2Pdq=W<^xIRpz=m9axt`a4-%i0OZDmQ!6nb7e8S{7}%OYZBTMAZj$90>&l97b4&v
z`k<Jy;*=TbnR6HNZIZmcd?!LUzl^-kJiI-pwdX_6`!bQTWv4i)$tT3JN@-ntFLcGQ
z>Lp4*P<Azfcmixcxhq8>P=G09orQiCBCIMcmF`Tp3x<7c+DD(EFL$6MAp1#l@O|3L
zn|HdwUxm&w7U$bOO-^fz#jq+>s{VS4EcbUe&Z1l$swvaGqA`Z~@9$7+>Kh+J?j|mZ
z=SJ<b+ME|r7UTPmQDj#;>|89m5H6SzwpvAx6Z`JxxxS99;*l*aTgmZ6EzsNJd-l%6
znq7O=eWP$OjZVmEw%?6+g9`&0N@ZRH1SthQzlp#nMNX;R<9K8Pj*NB-WY3bTqOJ7T
zYo|5xFDK0XHJGWzgGHU$G{~nTFcnttYnaP9eW^!VjmEtk4#|z>$YXZJ!$<lTeR$m`
zcEkmM9=?U=leoFUHyG@e0_wJ64fY2tSIqDi{2I7Sb)S5;<zG_p&6y#u*dC(9$!0w3
z)``Gwf>}&O`<BFO2uG?3{#1Nfu)yPBrx<I6Q2Y>>@=mj&ytND%r{8aZPl2<~Pw2w&
zU`qcy*|w`B?9@+@Y@Hw=V0$A~MaI)9wQrY3uAqD0cJ!OT^72L73oZQAi+tX+@--d?
zsc1_uIU)fLmHcs7Ge@!F0S&x<HR7HwQzxOxS<20i(_@lsu5WTCFC8U>W$1UAs&|&G
zp4SCewok^7+|Ag13LzB2(%?;I{0P?$sFJw;9*B#!-tJ6lGcR@ydp%a>lRw6$UN<J3
zb*@-DkxHa<QYKc;Jj766n!V6y$*HIGdPTos{>k`B7YV)Kyc<xsHm}qB-H7vWpVxMN
zLx5Rsf&+^Hc24!?^s;?*MpJ<+jQCb2U0J>CWpKTjO=BgMf%<UbguG1;le~|y;L)0>
zBpJ-St3H}N@&~>lZKm_|P@SF@*DS&%!%Ne~92QLhh;^=3cW~rpB&xG&Vl8YOPraQn
zmrb;+D&(JCG!7xy$;;ypP@NIXb7Ba^*l@oGg4BCwEM|_HV<Zj;2MrRr6Wt<s@KOx-
zZ=bFcG*lzy3EGxHk$4H!!?LYI7NnZ44l;cnlf>9;(;H{+lJ;t(w)kC*nNE!~U2kZ>
zt-1`Ud|U0<Mbe(0+*VZ#fR8U!xrH-6unIY`sjj``Az|dl-(tu}{g&@$@<RF}m^3^%
zmQbbepvF_!oc>788myPJer~82J3YTW8Lss;*w+Ldtsk;8U74R3kRUQf&XBicL~l{5
zTemPhC{rKgNJ-IJ4WTuzm#&_N%GD}q^T|_meJRT*`1z>gXSGj@%kTxh)}^UR(MM)6
z4KliN17?S)yj8?%%N2H+44!olhwwcU;Jzz)PcMsWf7)(ZnU!&q&Ept>qz#YMh?#ej
zw|wTERdFc`8Ol87{A5q8WTX7)TFF^72YJr1=d;}|V}@#dsTTp-SJA%NX0F~c4n848
zPNt;}R5iA*GUJe($fxCP*^xN~CM}k+_^2W82?;#f@2)(OEnok@NP&02D|=EQd(A=Z
zyoieSkGUD4VU##|5%<-w!FO%A!pCOsxAzQFH6-)8$63u7&M1ly;k<sQZo$8o#zFG3
z?i9>lN1~a#-WP?@=3ItsIafx^h)|%E1pwcELW_i+pJ=UXBe(XFpUD{3R&jBb>c?F6
zdZ;v{EYi6X<S{1X@}|868bG<Z_xJbqcxd;Jh&`bQQ#7*e?IE1425$(h?mjbOJASv|
zMNv%4AgP}RfAGSnh70>}f)y*QJeJ{p4v?v$&PCc<yKt?daSCadTSYy3bnYeHUbehl
z;~kCoIDlLb?=Jh%8oT}Ik-piE$zJ2y>Wd}CFdY}LwOol^T%vTcHi&+a!BTHvZFfK9
z?tn+NW-v5l4(2Tzy5@Pg=_|%w!7ALA<=zsvIWcDuJKe$TX<)pXj9HYRM=!)PS)D@8
zmoN6%^@Uzt3y+JBcO}V<?Za+Zx5wlw!eR9>dl`mpd>rY?o^5tn{|wdl%09$Gn-01~
zv8HRmwQId&#LR#Yyz(Wv+rwv;oi$}K)<U{pzCgcp3^3-bO`x>a__N9yhc$F9i64<l
zD>kt!3OmA@uy%M3JePT`K846F57lMrA){ioz0k`&#;OL-b@R%m8xk2~LU*W#s^VS5
ztu|<z4s|7%wLh4{3duZ`-|bOry$*k-smw`H1=eyCQJn{Gtn+)l>2V0RKFaTFK{jY)
zR%YH4!qy32eeghs^u{|xtJH#YUso69;YSOnpm?J3`3?(JUigAr*y)<8h+BA+`4cVk
zFCVQagVLhc0js6Da78+{i|QIi_lpV1cv#f%4F8T4rrfv>DxnWBAotlm*c7^~&X;o#
zOFKby54D_Tyc*~GAqQM&xD&21O~*Z+X+x7Mm(DDz&&4oeqkK;*R%OM5A!s%F2kiYm
z%Aa*%e!}3hP}kbG=>5f!!?c)}L8r7<6$hOAYJI0ml$0h=&9XCp{07oAt2ARzMNUI<
z@0@l+Y%i3Nnlcb2)Q@7Kd-btf>)}wHbZFbwFE*iWj+&@B0P<qAYEO-zB+V<%WYCGJ
zcY_^ZqKuK9CJdB~x?Yg+G<Kc?*#O1EgC-Xr2}>6H&{nbrVob7$4!U>p*lN|`jLqu_
z_wF4=y@zf8obY-*We;BMhM~MabrzCOR40js3@Pw2)OPZ@SkI_z+;}>q+HDXrKUY!A
zY0p8GakGG)hH9BZSB7S!j}yMzkLG+DIur4Z)9QxR2~gl?OuwRUoPQ2f{P6fJv-N;W
zL}r}T%MMS_ZBsbcKd3h`0)#M&vw(fXEQTedT<Y~SrYSv>6N6f!Mnnr7*gjjm*3i<N
zd(J#IJ;;0Ya0)q;<iZx*T>>IaTQJ3s-(gF!OdPds3eNZ;c$&PJwM|B_`;&#{ZT*yo
z3t3LD8qp`z(}j_#%SEJ9^}N9uU_AAg(obSjqBPx$${R}U6AFJY!8)(b=K4yxN~LR>
zkO!sVLip<7{O4%cT%V_l<@73YgH*#P2};|_{fe95661q8nfkl2Z46SgeBtCyX`Tr-
zg`A*na?hN)V0&eSsTeu1#`5&sl!Xp|tq@r<b*-0i3*;z)Y>Lza1qc?*RZx_FUV_6k
zleD0SYNLwm_;n-Csx06>pg?6=>yf6)K$W!(49u6O(QbmVud{DJN;nZM{R;^uW7IC{
zGW8wCZ~mBRqM4iE+*MkcZF)ejlkJ|P(9qM#f1|B2O=lCEBsgW=6zJKRt~)!Jz;;bF
zMm4vqO3OVWU)3VUFUPs<A5rXia1ukAz`VgKsh34_5r?0Jl|Yk=;QBNdoZ20{T~$Pm
z5rn_?Bn~l|$O~ryCfuVb#ovdc{RaeokU)E1e~A{4`WuU+N7ml)X0s`7w~TzyyBCFD
zg+DzQ3aPOwb()H8o8vXYFpj(hEZe@fHk>Ml{(7D)i*9yGkL~U+);S@BrR^;+J+D7~
z)0^`}?wG0W`wym^M4;vT^7uP6G^`4|yVB%4R<nKehkwvF{ST_`e^ayvw(CpdpMRr$
z^1apl>L%9x@l4S&yA^}$hI#l6MBwSt9v{5D4HWp;E=cLW{k3EJE!ZqpwN8RJbv!SV
zhvW5B;0*&Ph>?txaiSI<T%)wuaTY8GbYC`<45jW-?#lw<aNEeVy%qX8Iz+IV)@elK
zCFQP$Iet4}2WBg)-J{M|Emv+zabjJp%}BMi7EkFYU`-d&C2+OR6MCDHKjOb(mFOUi
z9Xyn|=M{1O>IRC{dp_qDWQQ?d<*-Ez2+98Ovx@y{`|SCge_jptOj+?<BUu2i2=G{s
zq$N)nu_Cy9%(*@Cq|@bAxRZ+&{ndwvL9?el+gzfajqQ_?xxFQp+7n&!A9pM}!a8O5
zs2-31&wdPa<$CX1tMEIyd$@a_4-dKiXVd@j_<z!8_m4l|Ut<29SL_6M8;q`SK2a|w
zHD^<+%M1Zz=(&d55+&$PXQ@0*Igo&Ah*oS9pxlzup0lI@&Qva9i1`3tUvky8(fUku
zWHa|5u(}VM_4UPoSwaYv+ulf!N5R#XkWO!9%Zs~0!`i{L*WSS@h6dp@+KZW?avUOZ
zR(0Rkzru!QL7q%q(w4C0oF?fKLG9vbU!->`LJaQC|MFPW5UXY(Se)+M_%rOtMjqG$
z=BVuYq@ZZ5;i`!2!>D)mTmFq^+hJ*iM4T6sw1?P=eJeJYES$FeWuee{ScIRZOy<f>
zqbwR?(VV6(y5iA~x0EpN0XT8pI(cnFlHRh&onn}sIx<6uNs>C3yKwIVlzKZ{WEtXC
ztK{_3Dqp?8W=7R9B`c>YQ)N<!9*Zk0*%Zz-uAmcn^Ci`31S^2-0*Vv;AURYdkU`Vq
zr!Ge(YGH8Kx@j@dzu2e3*%dIW;%Z#i%)1538_6c5jAq)T<W<N7G0BThnx_Jt|7vjl
z#T@^{5u1Bq^E83sk_36X^Ym%gpUB~lC3(<##c}I_baJ?#XZ47BDEH=%CHoVx{`;;K
z4V}y$5xZ^LWG*1=R0{JAv-f-)*X@cw2}mE+U31sPU7!Jcxw;lSBI4OoOjyx`k{W8(
zvN0<vf579G;PJ2Ant2=otRDg{>JDxg5DPJ5OnFheMUKHwO6R4cP@5-)34<8wxoh(z
zbzg%z#0yyMZbKw@iH-$Hf&Clb;v1sJ4sjPUt$=o^XMhT${CEL8*e-~dnnU(0=iswT
zHazWet3^4e&bXn<0o=skQ4hEmbnP_Dj!mh`Ka=~8J$B_erQte+QlX2OfLmKeRka!6
zR^HH7?VF%-O&X3Y6t`0JJ}qhtnpVG8)&JXDhl0VU_hg>uFG@X#hqJN(n`EpHp(MT|
zb5pzYm+KFY=_>=jkT=bQYX@v1!biQzXf>};Cs?25K2NKB0?(-<RtwXzxQg0b$TQDu
z^eX+9Yg~sVoG*I~%&uz6r9;523%KOB%5gbc?mEbgav9RuKVf?RDhax}fQ}i>B)*Be
zX{ftWaSf=vN-pF|o*wsBGc3u<IPBV+GV%LEiRMXb3d1G;iZwKmzs~;B!C!Xpzo9NT
z-evSV5oLxd)NXsYTWc&V44T-tA1GJPEdYt@C;c_?M>jy*gtU7P>N9Vhod6@>_HVY(
z+sm=n8ZX4!Gts<GzqLRmp41<I#<^<{NZ9;wcIW<JX$s@W_=aL>I7sB`$D&g9$GOYV
zCxOPK{cgW}1BKCk?tb)U4|w6U`G)+jEdSEwUq<wwS$hs^0bQ-pOQJFUPh1-hXwzK9
zBYx;IiT2$*=^IEx8~Xa+dg=eC5e6)S{#!x++}RxgQq^^r6ztbR&I>SZEsv+<s71dx
zc=sDk-}a}%%_z=h>}g2AN$e2XU$Vb+@RuF@RTuth5B_Qs|LgS&lsPXy5z<jJ(&~vT
z0O=Z`w&F4}a(-dWel})~PyU%Z@Bi<AAMSRwmjD7I39q%$+TZ-SKKcKwR0fw`Cq)v=
z6u;4QDv1OD6WF|+4D~vx_Vid0h9dZlL%M5?;K#4e%|%xS>$n;`Q*f==St{h?kgj+$
z)%8T)wWX->e(HUfsF4>#7JZo^>#36_7<4gf$vi}@C2f%;db**_FqsIr=8TQJ_A2C*
zosHcCLhX+ekzhl}_`jx!=jKH|ndoYSXoX*m3ktcYkNpaGaErEM5OMzYi&BaGa@(#N
za6040cgjH)KUL_dckWr(Kk)^^9&G1ziOEy`3aL1LBgUH=TC{ria^lYJNkQVvQ?kGg
zlJ+6AlW_}*+}I0gz01!OFlA|!%RTgs{ckj0bR;m$Cq`l!+iWAq2gLP4qxXio<Yx56
zU%CIKnZIo2|LWT0o0|UNy_m+>6M`yOqraJJ4&8=Tv6`U_*h|<Ts>ZjJme#2R;NI7V
z9E=L`UA)$#Y%DPO2{OwK64^Hn*huqg%@2q{`&s954YOW6OTH3BZ=2{n06~3c^|pTC
z{1yE#&HQBq|Anf@1K+nZi9&T&&GZgSWsE88Q*5CNO68={@=;gNAx5X%fhHjND>ZIj
z;^X}_NuPf;4*IW*3iYql@F!0Hf>r$!O8Y+~$^Sa6>K`n~pYoypgL(h!Xc*^P?Wa>J
zt)*_albo2u|N7Q{>iL6Ucaf`a@qw9>W(PW-dH%_R;BVnC559lV#NUoDdSpUq<O8&!
zMzJJ3><V$$1N5`zA?k&u9uuYg_uHq(^ULhH>GJ7qtYjVzcGQWe)=?9%za{QrgVB81
zST!smZmDQw-kb+871(4vSvs&KqhlQn^`Xy^YQ#NJi!-D8+hbUtr|Y;QpSKkDxbB_f
z^;EOIR~C>3dIuUAfgp|C8aToZUzFmIN2rs!93GhtLZm+sEIMiRhU3>ir6>B0MmYiI
z`F^CNEI~X{#d1>|&r!ZgUes2W?WP|#LL(ZiS{_Z^uK^652SZghMipE@QjLs#K~`v#
z-zP&x^UHuzFB1(XYv}GN7>VJE8fSQ#GWR5lePNB+R0QlViVdCM(bLU^X^RZ$0_qJz
zvmAu2<xMKS3FA9CIp$M+e!0a``rgWkt*@NZoslCsejjfBR;qt{cu-)}a#t<$8;#a5
zuvPK_QP97={qJ2#2$+IHCXWvg11eBpuZf`%xb~e;nS%xZu(FX$^I*5-Ht(88=U0?k
z3fl2m;@*&62bZ}1lNzaekDgj47O4b9#Zjv~IK6Q{P=h!=9tt8Rj)w+4IYFP8UqG%V
zS3@{Te;~Cxm2BC2zAJGq%Q2Wpv!z#?X_DAwc=yaomdTF0fPkJw3Fw!3ZTbBUY4yGP
z4hO|+B;(!DRs}Yl-YO+I%L3HFW%)4tQ|7IrOG@*z;|KeHux$Ta=0EAQND(}kOAK-V
z%}0b68fkp#nXj*q8m-L+0PM6sXlf_Y(kg~U-G`t+@klN+<LTdMoqS6|4{jM)x=~h1
zYe~;EqY{z|RvL!>=|hL*9rN(4$;i&yUvw`0dmrN8k^ep1a(mKxY7efob<&2xV}7=1
zWRI(L*_os@UO_<{3y@FnrvEt0h6g(@@^n}*%XZp?mYeG3Iy}HmA~+<|Y`A;(^lU`;
zU>AkRulXYEl)d|0@W@az&c({`$MkKhqwQ4}{88u{B2k0ViG7q{k$NNQozWV<GqbC|
z4oZb8z;`QVe&2eGO$+Jg9ZJqa|J_t2e)Z*6?#Ydb)yko_!lSNj-71i7!wTQ!LCg*X
zHC*317v$Ma*YUID=Z+P%?6?|}&d-*VNf5Vq**Lg^mX?>0lbGth=#Qtu19cVe={Gsf
ziXH?6{X5F|ze-$l^mAK1utV&w<gHNF%C~C(*T~{WqC0$BDgn|@PG=;jr6u0>54CRq
zQU6k`M_AjC%hP7j)%enDS=PEx&+8A{?sbzEjHREH92g5aos~qC2p#*|_KcI(JnI$c
z(lcMnY0|>H`Q;w|tKq67#zB2=S(A77u2r-0SMDC>lM;32UDF9x_g!;E&javUX~UdS
zqN!9~x3o_3sj9pZv%KV5**K5pC0bgm>0Q_Cm+OU%rEeviWSy6(l)I4A-)Mf%NpF`}
z9!&iB+neINdAEBvD}bYRw2^v`>h}zd+@I%`sp8i?^qh9kKYu-;hs)^K9!C53&5?cf
zO?S^%kE@4^1MR;G)b5;20z0jO){Y17_v7meMBnL;Q=YS~JuwdYTju}v@TUd`7+;)L
zw2MT<rqZ1A`a+qan7XUzS%&4~zTX^IuqUeV<_XR=5<`K{XxD6vua$$OVg@|foSKH{
z65*c0`&laqMBl+KE2Y8o+LPK58|F=E^vB0Qe*qkg_|K6En2!fK&>_FkKHt(_7kbbh
z5FY<{_3G1Vz{kVqwyzH^39>@%c9~iZlW(;|hPR?+4+^h!u1apr^xKeE`IL5885p65
zT(_TQr|uZSf1}~Je5S|_zISv-#vTqv-Z5`sKK-}F{|~SJtn;TE8+_3=n+~NSH+(!<
z-&yPq-thL6Ow37PW!j0>P9Fg30xM$j*0A%YLo|g$2Gb=;NZElLXio`HCULeG0<V?5
zLfwp@sPKt%Z;I@(3Qe-=W_wHQjRt>J&E4$+m9fgY4xO|vFauS>iV*|;tj!zxv--gw
zR-=w@RdjDo!Klu30!<jBE~daM^0P($e))LDut|ljc$jj6jhs~g8o6T7e^~6lyZPU@
zkr<EW57ymd?F<Qvd_$+6Cr6p7x-Z7a4s~_I2>HLA6n-(12I4A7l{z^wFu`_;7&U!n
zH~>z9UAjwgqi&%&q;KZW$w__u7Y~UW25N0BJ!6EN5GLWv^_~}VmoY>q!OZFE9^X5-
zx}a7DypQTEajH(>p6c1*Oo-y*7}m$1Y9N+?nI})k(>JazRq9!3$-mZikdu2rXbXuT
zLMHKRQv`{*C{1me18qVIr5M4LJYaQgGF7f@EMV=+`Bl+R47%l-cI-@-wox|Vpcdjg
zY|{DpP|FK3WSM@?Az5FV=Wu3ogpz%8&(55MqloZ19YRS)ar^Z6?V!1Wme>&c8+(Ga
z3!no|SG8~Sma|z~74l|^LVGV=%To(<A98$lrOA@TR}u+&DH0G>dk9OmmIwq@`m;1P
zM?MT=rcd8dcnh7rVf=|M)N^2h#_a*8)ig23dd_|8@$}#m?pB%!#gKGW`A{=Cm*;H_
z+wX?0A6KVZH~WO9dV^)EY`#~cIZgi;cOHLgu&UFQ2$-e_vBZbTc@$O{4p>`i1gjt_
zWNXRfK70vj#kFBB^Sav>o@7DpeqJ3deR=AJz#~uNXT7Mam{NM$H->{LGC$9cCD}mU
zd-!B>NMzD)rCQw5NSdOTXMnxRwIymH*-ELTu;vD=e{yD!kg}$sE+ts|bW5*yX98b6
z4byg(KGdlY>}xTfMcSO%{cTEj>rlC~1)(~a-mJV+9f1RQcanCA<Kl{NDqTosbB1c+
zjTRV=v%d}|pD!9W@2+@ybIe1M4L3~BCoO{B<8o!}JTET2R#~0sALgFBDVKG?bGgav
zXZ`AP)!`dA8@H;$c9sjT6-LZyg=@woQVq6&cr^H)sCdffO&Pjn+6X@GT0okgYJn>=
zV7vG=P~C&XU7217x{;jphgzd#Y0q9mshf6I%asKy;5D8iEN`q#-Y&A|gN;7KBtc$~
zWb>cvl)4MZA6_|TQyI;?Ztxa2mwrZV?Kx0=Fk$$;g2u;yeN3(LWZYZ}c-GWo^7PwG
z1y=O8pTE&0t1_2%B=8pOD^UB=B1+#exJu?j>E#F+H~K)hYz%yvW9%tkhpG=?&xr6{
zKn7Zl-Mi{Z+(Bt3*faN#9aXKFJ1L8GUj780O(h^#lnLdOu`AzDx$TD)4La_hT22Xb
z?78&q)iq!o{z8UfD0_Zgh7H?BbxL^Ux)%6ac0Q)S#d6BO(b9KhLKMr0v)`wllxBWH
z?7)^(WKb(?`ddXU3n}wnc{{C<0ijALT@~|{5>VO3$c5vT3W;=4hBrxWhqQJ2`w9QS
zgHZMJKyL#;dMQn>2KZ>efdG>K?KmeCCz64&-_>AHSKqh+HTPI1p@Sq0RhZDRAf?<v
z%HLdjN8@MXG8pP^c-x9CGrB<Xy`prN_ZqWfss9COWvr~w7LQ)^cD|`zq$OW1z8i^W
z8ZmZ8_zd>i&gBc$cnbuM#m8=x=x(4frq_e~;3ARWK`-fA6~*dU6TXbaPMQ{on2ujq
zf4fCU2~|^T;>UH=#K`5U8A;5-ci)tLu1((GXh589OzqXqlJnCWzPI)%v76rFj4g@6
z<3=#D60AP(VY=;BH6f0lo6UHNF>zlkKkZ&wPr@a2R~p{9U)Up8JHi(S)S?<^=LJ{$
z71!$nCnc`~pS{PDb9Mw7unju+6w4)>JYgMN?HYs|s_f7ADY|ytxj`>qt<8wm@_w-A
z!uY&}vwBfrHFGBYt(jtS4z?gwF&UNnqI$}K6?PQh_3^DBE#dQx+9aXKo|dRX0pb!T
zn$KBXlRry#g=d9BQ{S6o!z%mIS7+$Y8l){R*!SIi{YeVp>Ju5d8XL^D@$nO1N<)U}
zww-a~R|(QQ%_K7{`J<<X!&xg}8u@p56WH}o^0UtJg`C}H-^@9J;N2^S8F-*h?Hus{
znAuX)l~#L6<wM93bHhkU!eX?EPj?jKfnZV-LSHIVii{6HUG<iY^(thSsl#>9iLBqo
zy#74(v`mNJz;s4LgsdhVd+S&EJ$g0Yq=Xr@UBax-<dC9WSW4nP)ekc>{^Uk`Qw|?B
zgkOmdNoJ}81So*vL^;1!qFl@m3S&MePIh~&9ZO5EGhx9@#PGJ)qWAy{2;Rg@s-2w-
zexl)PK7+IASr~r6er@VCRAN*O|ITa6T$KA6<$JTSLcwB5fw1Z=XGtI6AxjIf>W`XT
zW^*x!I;`CZSA~<)5xHJ>m`R`aH#Yi)j4sqT@HD%RrO!9PmQ3qB!?!EltY~N|h<dBY
zApXi10zJBY#@nx!wFdKL8urMHO7bT9UEAI|dY<sZTLZ9kg{>RGWQ(48JHz<4sazUK
z9qoyMYIrIgvj-I-)r0(<j01=BA7ai}@TkeQRfWXqf)NFzpfv#i_2}6V-%77mgSh?K
z@~?CeW5p@{*%QCJ6eMLWR0j;?U%}Q!dR9-?ozY8shnjjk?~A6c*NlDGH~7^JSnV&d
zV%zO2%UEj@>wGDf2Hus{0in~5RFn1cLamF^bWT5>?M=>)1T)R;!QJDs=|4}Xz}zRe
zP>`1V{s1O!Jo(1`fXQ;q^e1DjB`dSA;YvuauIt{mV4JUF_pqc%*mBGwvrKcu>|TQO
zPce|$<`98I<fNL^j+WJaa(Y~$Y_W;okJBsfuk~I&C7-@dviYvC+;37@<F=nXl6mtU
z3|*4%EOXRn8r0Iaim~<JXA{GBqIo@ig8%D>CcLgnf4*;zcP0d~jNh_9GzM`q60~+w
zqt7fd%+2*c@HctDQxfpaJ-mQX{XyONV(V*J(J88tBo4)Xxdd&sO6M<h>l`P#wpkc4
zf=1^@#OHenaR5HZI-7Hy<o#GFcgPfTZgCfvh|)9fG<3n>oWpCvh!n&WsTWUcV@6k6
zq{({JJ)W{HxyP^ZWRZmj^vxE_X06d#5DwL96^@k`sXI7x#3`T&HqHs%BM|{dyP2)l
zO>w>`eWP{3T7%;ZNzSP9CJ2w4W^x*0e!7t6bzBrBFHnH1Z0VrC)cxqWgYw;D3K0Q}
zvecPYyiTsjVx6C>FRi)Nx&<tinjo*e=9svKz=C;+`y}CAA^qzaug`y@HHaG8)bF<A
zK0hK@bNtaBKe@d?n#O%{{i;?A3u+QEK998F{Z*wuMs7&ML=;2e;hepj8R0fG1rQ{r
z3=%)9JJ&Ab+q4r&9$-HnoO)w@+*{pJ&_9ro#z}25y{dOHRJK>wY#WC!Tj!G%Y->%U
zoYxgMujStZbIKv|3M07i@@mPvs0gv?IDI&xXz${ncQ~Yt+Oz5DB<%;RH5{uAjQ|JR
z@=+$1(*Z9(;VUB;p=MyOKr>718CMq5SlnNyhIQYJ{O4SdrHLk<>s>}h$NrXGC5wXx
zE;2^8g6Wrm8LZ{w?$54uPtmDVb%c6WNodEdCIA?><ieNWiY6`G9|?=k=7H5JcUzts
zdL`C|(SBp;%sUw?U|CoDlIp#HbvQ$`4YnR{zU)U3)AwDHn~2wsY92Hbl?{KydnBT8
zI`)w1;?3QH?sZ__Y~gP-E|*H;kCBMWbCKfa#x6nAy>ynUj_T&0(sfsbX_i_AV^;ZI
z$sMu2FGww!D!dynHYC&{lt20=s+R5)C_{$A0DJ|sDv^w$PFvJYmJogXo>K77d(6nb
zsb_|yWFI=8Rofg8pr=k(86Ky+*5xjDy@|=CvbPjN;t_A?yB9aQ>fmqh>rpfN2UYrg
zUpoExZfQN5`W_T__zf5%j3J6!uysul1vE@J7*b;#j_7Z;2r^_h1a0~J_|<dB|8hxt
zQX#VQ3*(W)r4=3p#1LZLAEb+^?jH7ce;4<HKe@#gH1(2W8eFloA)S3F`~#P#vpqog
z0pj*4G|1k%K>7u>FJE(uw9dQr=u*q~!eYKIldcy0i2hD(FQ7`=Nn&4AUs&Ix`0bys
zz-2oShv*+fb(6YxvdJ?w8tSN0#aS{hdT~Ym{%sC!748@XJ-*U@z|1~cr|*ZkfXjO7
zvNQ?_T7AQDwuULuY-x!36y6H3rQD;In0__|=E&mP{0VNHUlXS<mt)@Zx(PAMhqr`H
zGr7+<FYr7gUA2w@Ybq*NqmpGVx1uS(Ugs0@_CeLeKxJ`VFTJsuP8q?=@Qz>z5}uBy
zHo*@riY>r%>3T&Sneyuv6;*)i8Xj*+^cZEaAwOm@z}5{wWLeqBk16FD_YKBEIf=YL
z!e2OKFD%50`F_P)T83&`BlyP1rDM3)BcXtY5A*n^b|Ry?8XemKwJ;TB-W4CaO{9Xo
z?zHDjDd}++y9ZKBz$r_6Rw7|WZ({Z2RsOJRf~=Re%79NX-#);Gk66hd6AYI1ed@{d
zI>#els9%z~S$1^4)6PLCK&(`Fqt&u2M#kc+=n)Ik%wtPTzKr_PvICXfN7N`%aGlm>
zaGpXdx-p0?-)8#hZ2zgPv#6{WVX7S>Vh50wxJ8iF$}tiJrSR-%0x$RpF%^p_!F9%0
zR1vNR-?%4YEv>n+ZJhz!jxbkcTL3q<<LvMnKU3)W)#&Yjtx&WoH8WWh@^yc6L-Epk
zZX;YmejTiH-IJ-8<Vn{W8(%hP^*dye`ag1U@VgRYM$M&E>p`B#)E?{YT4Z1xGs^<g
zP3^an{TGZclMEg`5?Xdi^-4~8F7F`|EJlvV<fWS$qZ$hONA~;oGJ$s4-)QAc;C;*N
zgQ_5fzQ2$0k0dXar;Mx%DLky^*r*{5b(C~p5aafSsDOLcx51G<IdsXrhJs&vg;m)a
zpcb|lA8qycbv&Y&h~3>4LdCHT<e>N`9)ap7-q?aj_;0lDwTvxqgk&aLN5KczLl2hI
zV+$ts=zFE!Zp`wD+rYaK;6ZUwVKJUwX7F0G9dKuS!avI3<ysN)Dt)hv0`Gy4pW?gB
zy;b(<p4F;W%tf3`wTS^odR|8giU*nPSlj_KLTs-FWS<w;b!QEBd_dJPY2@*4y5+CI
zn&@*lH;m-G$SbZ<<9nnKSa%l->FinDR+%hZrH@ZZyM}-})1ax}RjIjex6y~L@MjGM
z*OnG7QfF#bMTh)35i1jxcKFM*2-47&ec!LOgu#fA{az`-4^s{W^%h1SljE!!b>ntC
z&0bNvVIqo9#43{73Dwb_8IUZQE5)O*lMkQLn8LulYO=WsYFUE~xqA+jv#PUpbxQA=
zv<Nlrc}0{R-1=s0oZ&*Eeh3O`&F*WVM`=Abz2zh|GD-T)Oy_3CIUcGFb+9CJERhc@
zuRa<aejQVh?Kro(o>MN%F1U*T7oGA@*~(8xlRk^5q++pen8a3Z7zgW(Rx6$EDpgO6
zze}mvMw^GQlDSloi9G5$X?}Z!d~e<K9?tPSJaMN@Oqj1MAZ-)|3rX(C28er2t!ZV)
zBW*g2IC65(@mNU}tq>ifxQ)Pi>?hToSei$w;*Pop0Un0sgzWVB8{BYeD~?KdfJ^_-
z3xU{V5ppe`g1q^ja><@8!2)JLKK3!@n;&J}1ylNkRz@ldG3AzNs+AT~{k<olnsn1W
zuBtV9J&EkCI#qkA)=5_98e`}qf`oE2LoNBvd#qVL22AIj1uy@Ref|U4-$4m%<c6=m
zO21lxR<)6&#J2ND)-D!wdK)G8jE)JJCDn(jLg+JB>bE>tVs5QJZDM!Jd595nE+L{o
z2BJ1W+fuChqZYlv*de(xVVzt#J?5C8)Wc~hL+h0Ed!*kvm3+1@m^5_xbh6WLcrC;8
zqU)y|wOpp=>c`WBC+g>I(Y>fpDnjKNTweZ0CYBF0T3?K?_&nLEpdNAMZMqhOCUY-0
z^_QhZr=_v-js)o{RT>{#5eO~e+%{X2x{S?5TMQPvlDUc3xKRxsiEdJs^vqWFRa6TZ
zK%B&fxl3$y-xQf&uf`gY&{i!^eFNF3Jz7B9i2031WE(0^nJL=OsiiAYf51LXub}~-
z&QRF0BnxkfX0}eN^L7{913d3-8q#pXJx8@q*vze6-sWcdCSLa7K<~9`@(y}W+kR+q
zkDwYvxb+Xz{C}MYOM?9_`N<9K(hdSW?SPfCT8r#<iz{|n*A?BRLGCx2d6~-XvC;*L
z%tQUJF89P4FrP@bbo!xc?Q%$F_m?=-_Zmrf{o-d75n@JIYaorSKgn&E7XHZiiY>Qu
zoaVu_j%=CrzK-=e-BwD{6yL8hd)6PK^I`maDO@n<Lw6x;5?BCTLj@m=$WYOo2lQnv
zalXu3R&q!4Z9~2ija)>1o&r+zFooDLT2=VCp{|$|jLAO1cE2pg#E#^lVPAYBlyuCB
zr!U5|&?V(TKA!O#O_`G4`wouEl1LVqboT?F3^DN$as!BQk9eqvM}{PcePn7~L<iCi
zkk7f<{R?DsXMf0nIHrhk;dHhu8^$-HUlmdodC^tQB|DH#KGemh%=H~^{`H_FECPM^
z)uE8%cS8=NrMp&n(Hf;JYhN01O=-9l$3dFf!A$fo-7<d}u3%LquQeriHha{2)$%=N
zp%UXx2hqOlFQxjblt^gYi6g>+@trjf&<3~5=wS_yW}J0ZoJ`99nndx8O#YN10H#%>
zJv*mkLa6HwSt&1AIV#4SkpIrSx7y;eISi)*z~)k2b6N?$6$oRu<x_XWd>`Lbrl~il
zyv_wL2PS**+|ZUhpr^1yWL?dLoJrlXL^9USmKEyO?|(+gtcyU5<qT8u;_nLP`!$5V
z7{}Ey(c8!+Y>;w&`$*Tr4Ixg~2*i)a-<lwzYk@2j4Fl}M6GyGQ_-5$571)(5H22dI
z*}x4+aPW6~DPhMpWXaR9W3?Fyu2&8)>d2w^*GJ9%`cP}&>v`aiinaBtu====NEnjb
zJ`p=#xIE%@?FpW}rnU?L^!0a#J+C0uDZA=W-2+*TU@&*nwi9k*zU8YIMK($hSQ{Kt
zy@CxTGXpf$l19*IMn%z+ehS<NgVE3PKi=2lh6fLPWuyZ4nIXzoOdIL9E}FhrRf^w`
z;~JFgoq9kMT6P=1k?c=jf#)zf8gY`DP;T1wViHH|!Nw!;-CB0*HE}uXkv<z!PEyKm
zL6^V2&Jv0*-BGa*)K8BbKY})xWWU6eKzq=8x7&WM<m-HoQfha8(+EQ^8Tt>`g%A`V
z-mr^+E*=^_f{JCbSSZNNRfCd)fqQR#?fMOq|8KO$cH7%WT<sUK3NIWsAS57BA(bw~
zr?Me0RCyIh3J7tx|23|rB4a!;M$4_HFcME3|IQmNN`kjF5^hvC6C<5uGN%Q)lYFn#
zy_BtM6`oX~VK1cmz1%*-kPOTZ*0y$sP@(2N1$RW`Q;l26;i(yYZOU0`dfcjqo4Gwb
zF)I~mX<&{e@i`I44|7m}o@_dSq)1DxCz*V5Ddt#F3sPIl_UxV`0>f&S@UaT}tj!j)
zmkFV>&vx-usGddwswrAFGe$MmF#}2&NaB6%Jqpj69nvN_A6({MTZGrAM&wEsh=8cR
z&Hel#Azg9R>D>sf-eqfBJS{(SSv0Pm<1y`FJ9?-nJfNSjy)k97TC)kSC7a4=<t8oJ
zZfe>jSQ+e;IeFq|41DMAWDDdsX`Ac2Y7VjwF6S}=CJt80NwbVD8hLOK!-qIqx^g-Z
z$#_`@tYNyTld=y*QFRZcP5R05b!J5=<jDi)<d*yIjFFU=%Wu#t3Ig-_?yxC$;&3D`
z7~&Zs?^msWbU^lf)Nlt4&He;akALvYGLM8)?c=HrPe*0in2ifAlQ$_HJfe{_3)o^<
zH7Th_Y)7x{uMLzNb@rRbw05d`9Lh>(==+BH*qPRj<v)p#)zy~el~gx1rV}7?avlvR
z84;I%2~%frSQTP-pK)riLULJjgQnJmmlH@qb!BouT28We-Y~&&h?iKa-v}9?1KH~`
zU6rc~sG&DV3@p;zcd_8vqhlv0w{G#ATa&YMu`v#(VyAd1QLodsbi{u@ZX7grd|Z>m
zox3Uy>eMqDUaw5^^i~Zd3EnQJP7tVOBP&Y`ma@v^X0T$;v9=Pl1D;IgA;9t3G@QHF
zyxyr7o(441Wt^Qv6T2gAN25sA>IvighTU@3@sK$%rsm`d3Mi$3%B8t~&(Ge;?YkI<
z?E@szck3BJyK(=pll#x0u;S$iHvMQfnFlWSl*|H^15!*F)T@b@YMS8|j$K_AY=&tu
zYJze()-^6RpWZ!1<sP0r%FzJFIyH262S~<l<6x<IX#rlbIQ|&xv9MxNIo^^WUxNbc
z@#L%&*~d2m)N#w=ot)f^;wBy2Wf2D9tDN@Eh1OjD@J83Pc;XaGCD(^#CIX4{WJIsl
zYu9|j;vM!50Pe;q3ev3og6|76q)u`@FFcc0E^&UGj;Cud)2}N;6B6W=2G9u)$Wya*
zrlO!E5m5C{cr(D)x>4g+<-l9eqfZBkXT>JB+b4l{0wHM44IUQmLW`YMjslE93k6~c
zy#w(5A0mVC^si@r)*cDV_A3|yWe^#_1enKD4-Cb7RB`2j3loQws0G7~bcrb?;b4+X
zC@C%?n?2$iNHJ;0yx!x{Vo9|DRU!$A<IZYHf~MLuu3$YLaZTvmLET3a8~gH44%(i%
z_NM4MTp`>9t2KBoZK3&^L*bAv)g%E1T~#S^nXT_T>-<)Er}_QskF1@~=aM1Ko3mpf
zbM|LBAb!1QnkIMqJ%UMm<GE=^qj|k0=Akr=C6}W-*uE6WjyE9KKx`5gZO8j5$H=*)
zpshSv!=s73V|HTL5te3R?yJE}RuB$yPLC(w>8{<u?4tI6oHvk{luf&a?(4bq8Aa+a
zt{kXPmEK(XA?JhjS`X{b7=|nkAzv5%4Jm>!u9a?Zff)=jqohm=(?U9<b|&*NJpQ2t
z{+;kot_ufn6G61SaWiSPG>JVz6{5y%=K8c8Xf!<IVcLFC%}O0_sBNwXZ-0&VJ3*>N
zO!!cQV2=G#$X=#{?y)DStAA;4p|T6YsEU3@YOQT0Zv1K7ta4leT|C;xM?L-*h&}n}
zHil|uWyo!GyOH&*hK2y8FKbi$y9NW(Z1pt(Mmg2JU?W6*>1Mx2A>Jjkh`kaUF*PPF
zrt1pNH=egfE;IDYb?r)=&lfAyd4E2S_K4Cqu!oqqVX&Ta3ky#)(p39xnh#2LXHq{x
z?!zDD1N@qRv=8222s?+*upge8sgwzC2V0+~?JY<R9d2$S?FrOkq;#xFli?L0zGN+;
zFeoV#`)MqmaIbfK=}68St|`)Ce9|{F7?aLm{kcWL6m@Go!<pD0lV^n9i@+o)VI^e|
z!A4hRbEAO(9&U>C*lw3bMoCeX%3Y8Kb$5KoZ!|~bdnk?xWEL`C=gaZcEmFBzqt-#K
zdat~rU5AJH;vS?o1M;Z(wD&!p4^4+-8I<cWo-p2^n`G@QkdfvA4(zlO=_p7@l1Vlr
z#}{qIb@GC%3Yj`}aeW!RHc+45&)Ewp!<$q*NjB=%s!4#7@Mi<7mv+EMl1C4~hYZWe
zdhD7*qEYPjm65e8>-y-oxgEH+@9rdNshSIG>-zSKewNbAtz%Z&C5+F1*{TQNbj7F&
z*HP2us$+g~8f@IR-k9?!>vI?=6aLgN7u}1@N31#knXsgg5is50yndjja=(n#{k+Z~
z#_0ZwAdD&gw6cqs7D<%95z4RRAx!tek1Wdr+ge3uhWo~9re%^(L-!Oa#|xchbTjcJ
zKDp}Vb&yj2@Gj`@toI)a{(JZ-^sF*ky53`j=315IlS0*L-C&_GmTWru#;8Op7MnQ$
z@v)(HLDR83y9u59%^P9-2LxpQ1B}KIGZIbdxuze1cqQ@HJ1b1lisfeFHW7>s-S17|
zUGWLS!ls@@OXB_pxD2JV8#3r$-u%Cg4D~-CmB&5oiVZz@+$AlEpYILGaUmNZsXNgq
zFu3;EsthFRQ{-Ox9?TY#&Y{rjN>mW$<vin$#V2204SvyI9QD29C2yjX=omYy@|{S)
z_`m}F$jK&Ulsgf>Mxv>O(DPEw=$wOhTw5D!ct7U4ic8F&GW)bd;3xU?!R;R@M0J>_
zMt^E?M$SSGYM&5Iu!efirS;dXN=&jSq>NY2)F-+=7f@?L6co7WAD=NwJLk>O+Y?om
z^wR3C92%5|-ILtusr-1^V$A+7^Vh+ANRGV(cI3PS^SH?yC=a}wn?=#pp8QcWgYAUH
zCA&z1qWK&q(YJ=wN^arseZ>G;kjwe3))7tIJCXclK)0G7%*e9TD0O1qh<@sdRME~*
zr#`|V!k$f_$Hid$bw-<kLG|vt^#G|ORnjfr(#oouNVL_Ng{Gp@%5==!LC@S~BU4gC
zzIqjc$^88hXP>}=Sr-k5cui%atwh@}q^!jVz9n;kx0<eGWoobXew~}<7<KI{C7Oui
z0nX^{L(y@^C}v1#tI1~`MSl^LszVO_N~ukMMcnKc5xa$SZ!99D?k2Il3ybbcH=uE*
zk}ONYSD^ee95TD*<(wG;i3Dq|8n7X1w{J!~*NX3U+~awFBykVRwml)=(=_T-g#f9I
z5r4Mg4$ER^m?Ejv{L|!xpU(rRo20vp7?$*2^y}B5qOKN~00Cn8`Ok&i31&Pr?cIdQ
zK7`V0saL?+*A%NN|3CKLGpMbkT^IF?$M!e?1DKq~fWa7pO%fT#Ho*j;A&Mlh2?7KX
z1`$P?8H~wbjEN$cC_qA(AV9=1ISC9x<S4R4G|?jX@a?*%_N}|WI{V&z{@gnIQ>jX;
zy1QDd-}SDpexH8&dE=!YX4#6?x36wWzVIq6yq^eb^qRX~1`o29PmBSWN5DLqp`j!4
z9%1~E`re;#ej|@{2Pu{&;I5B>j95UvjiYoY%$-touZrkhq+uZUbHe8U<X+@rb29W~
z`?Igj4NAz*ebP;g=ol9{31K-w$ALQj@<M8cXfdntdofkkg(;&>X?9-PBiP~OUHn4S
z%r8wo<8O>k6cg)b47~WweN~03KmNTu1}qrfw1YsJolXw-lN{nKKbqcGTGfmPz8m2R
zg_nhN_TAmIGLpneNd1)U6+5N#$&<x-Utae&*9~~}v%y{WdUK2MlzNi4+q#%isZ;E2
zMV_})1>Gj7VkRfdP4(;jxxznGXKB{BD^Jf2)b8g`r7?4a`eRUGTtdBz786~y;k=~A
z+cza)3Lg}dIUU_#24UDI1OQA<SZ-zDC(zbTq{X_=alBZKr>S_COwZKKOdFU_1K-5A
zc>aFko#m4T813$bd`E<hj-qo?-e^;wS(~OA@_o>6o-Eu<<*=jOBb)cGwUHEah|8FA
z3r`U9HPT9`>dl_JRmIY%6c^}{44+qxM#Mb=?!%g!NJiSPJh5-b-qJXyj@0Y2nJKUn
z1?vb*dUlZ!$ll70nRdsms%j=?@`G;8A9eQzp0-uSRC`6<{ZOv&F;p1w3)AA2A4uYy
zgWQdE6yT&A;#;kOqtOLs5$M1&P>Oap2YG?sZ=B&bx@A?U&HAMk*qpeIho`sn+=7dW
zSh~}gfj(*IAL@fd$#as&=GNb@c($rdgYJ*Pn6oAvGkuFr_LXQ(OI9Kj*;g{q&ICp7
zyr<Yt2m_K;XI*vGCtFI*w})!3Js#nluQey6myQ%~An$Cm2z9U9j?T355IV@(DdZJ_
z@gIxOqkSloGW&o$$)Z>M??%H7Vi*OY=X!q(*}5FXmZOKKp7ZK{tf@IBVo{NCPWlo4
z@5UT|@57gSme~|JdhE-@cd(JaxD*@Mduk=VyH@<=u*&{=@sSfnK{L#0%OY2QFy88#
zn+ZcU9CBUtOEa5qi?6>pM4&^g9lvcVt-tXel+NP(IoB<dRCSJ4EU?}&HrC5N9wB*m
z{#Mw6@4Y9Szs;#Ecs0`5f<DZ&#$_cs<1DFAqm_`s`rr?!w1~|QWc)qGt!A--M_e&0
zmRo(^&9bB1pOOFwkz_^IlgD`{t4qwF?wyw!^;`tV7qT}~!>K&$5C(7h>ITClPtY~i
zP3aAgWh6LFU%?T$H<n`X#@<;mfN<8p3%xV(CvTaZo;3bK!1l$u2iCnJNkz)m0l^RD
zj_b30@uHRKJZtY~vVifBKT8X6#+f%Vc`cxS;tbA?YxUbV+OS8N@fgl&n!R0I90kaC
zU-K~6`nyaL9f;<^X1f;lWH4L0L&FNyP=F2G6gq8P?33FZt)@}2_fgfYU>gg87&3R3
z%}qq}t;}4(6yK+ColZ69x*tj`FD`Sp4t$FCe6rka_quIt`uI-s0m2?N!Fi&T=^SA0
zQGPEp)gpUC@T#<EJ&+g=Ob&`0*F{vG-ePz_nRNlc*+vV_RN}P(g7jTXP}11kTON@Z
z7v|X^L_zR0U|e0WYp*{<ZMe}^%qZ{`JZ;1nQ~^yid*~f>?Ri;cNNFl`R!w&f{^Sca
z$5*uN=UI~%?RkDZmuX8w@8*eyPKrkTjqGmMj6|#*kveYs+5APxa^x>1t6L{$HbzWZ
zK=1nd$r<}+^)**0r*M1~K-(d}Dv_KfHcx97(FV&^vcDeN8q`07iPgJzN*@EvOCPyj
zth=#4YfOi|^YP6HDOQ<$itVG$>*y!ciEGQfE-B}(e8mWwDqGk4;5Co1#)X>@Bt7R`
z6MeaX34hSW91Amji@)*R?B0*75NPOdb6FmJ{Qb%-&SBc16AYl413#D(*P4BQQdxxy
z^HD}7{6r)3RoFw?(Jxhe+uNTe6;6aCXV{s)n-(-JwGNlQ{kzyohVYUi>)Z6WNFm!t
zp)q_Ro!Zu=kJ;5dPOlw6tx2n_-R(n{qX+DCRQsx?k8UFQ@FO9n_(6xoR07dgsyS=T
zv-*~M<^`=N;+`MJa><iin@EX5Uj&;CnDXF{fHTVzUD)ap0n0pRN80D+)>x`k8fArJ
z=RWB?S$!(U=>_YmqGjb48n*S8<@4SB%IcJ}@(B}eFND|i@?{HWF@dYbj~z<>x(+=$
zjO82-snb*1Jv6e&X0RCL!F+#m7F}8D3*t5CTenG;aB!<{D7(qJ+_aSf?+1!i?oM3%
z9IWbp<I_#PBFxy-W?&|6U#vZ7-uCz9l(d{MvIDKZK-u}Bs5pmMwu#6j_=T3i^~rY-
z)U?S>k9sZcb}fKu$2=!>F9{(8pr}r&zDiDB1Ncezj83Nd@BprPKJ{ebXcOL`VIHb!
zTTJVhBy*}@MRfALD;f)w($H=!H(vbu&(^*FHTH7@V%*3(+jcqux6#7EDL;de>f!<$
za5{BKE~!6RdmjA$VDjN}Rx+h(F>}^fJcAXv)^nYGk^Ip*sZRZR+Mlp^-w->UBe=g`
z)M4!xykmCo`r%gh>y^)5U3##8Y)<<$xGw8B_w-yp<zc-PK<DA{mem-M)x60Q@a>Pk
zRJMwC++7=1N~~Ha!i&mXEG3>To)%6cj_lW%rX@HPhT(MnO3WujTYRr~5@%Nn%0a3W
zlw9hQ1wga(+aH;?GJRRR-A}lr<7Sg!#cgVuIx@Xpv`@74_jhZiCtg&c9KZC9W?m-!
zSsthfffscd+1wO%d)lJ1_AxU)dsCfsts6_9A4UExfAj}-*}6T}&M#mkRrw*QLLFlD
zoHh-s1)CjPQ5xLzejOOgdA05%j!>C@vl)?n%SGBn?+{7E+r%c5`w+{e*%?spscm=O
zabmY^@onO`25Y0(Edd)Gz_Uo~-56uN>E-K~u-{}%i@xov8+>s;Wcvy+!$qVo>rBhq
zX0J6H+<G^nERn(GLl#rZzu#}0c4zeCM0wQakjQk#8gFb_pd^pQCtCP=Qc~YrT5g18
zUy<g^Gw=7s(}u*am|ov+-j201W`$@66|)gGCvH`5eobHNR+}P(i%THRL;QECO$ec-
zGdCnA-Ds-cDcMt{!KTNr)=({)YK!i(iCSabc<<cvJfTjsK>N1J_*E($$BJ&Nifd&a
zi9>;O)Vn~jlV)btn*1O1ytGxKC&#xklSdQ+yUMm};-<YICZy|ZlLhJO+nF10x&B(7
zY@RE$ca$F(w_2U#W%Q#rB)ryu-;7F))M$=*qn~3oyH_k#nIHK-5;M#!WrN&F{b#=X
zAIq)=9oRGI#dBlI7j*8Q;{N!LE&h4Q>Y7Ep`!@FRl4p!33eDXGnLgC9D)MK4JU_6W
z-ejl-0=asfhU-WZuOaIAYW}=e{Clc$%aqAi<=XrwY>OTCTGd<Kn`y77?iGD`IqAZP
zj*tPyJ*}>e23<D~w_>`VK@b~+O-D1`ivM9>+}_$|<Gl$~C;t*kP=$jn{XDB^K+bWM
z%uL%Ov5(Michm3I5x((@(LDw6Du*fAHVGXzcwxyHufj*8<D>7t@%%A3^9<mYu+#2N
z?3%}yC&y1swu0ZkoHUuVEw0R36A0`s#T9nKwMu`vubcEr4>onvlf%4eTo}_ZolqF5
zp&B~vpUFP;_LsqlnuKnB`;n<$!0)?fZH(#LJvQx<k6bi`4Vw^<<dR%w;T64PcflMQ
ze%RP9WCY5tL(P0bl+XREYu6_zDq*}%`#nqz5?X=pl{~!tx*_@9@|T7A=L0#=4RjQN
z15kBq?kiqJ^l!w~VYFUpOQta!A246#M|-7$*I(KLVmE*aGP9N8u({k9*)GN3Mi!23
z9uZ~UOQf$%%Dbie01^lS<6s^GKmKB#^l6sze2+G<PqWJVFBn2mZNoUDQ+)+?fuc7Q
zPPr0g^)aS*PQO2#csB8-aAvkIT2|Mq>I0ykXzp9s1}|p6he&hy>vtuWEf!SfO*s=^
zMHC;dU!g*oa<~2U!IuBN=cizqBLTW?uC=MPLv3+q+Zy}ej7keu2r91iK5|sg;HlKh
zf9^N`q4WtpGR#u;@qR=9d0)+S*yWnlpEIM@hLw7=idj;oY#yoWpx}HtFstA0AC!Co
zKF?~euXRCu5M}f+_{82LRPv2=W7$kh+)&T8KCRRKVxT1oFuh*WN?<dt!*_T$IK5Co
z`dT*R19atc_AS#}HnPHfar>T6_6<z<;cg~|J>UL_3sux4sJG<KsZ3wU=>I6&)Kjr#
zpnUOx_c9sk4f2A(GoG)j&|o4`PENy?cxBc8mwtym^(CTx_5yU!fMCZKpkW>@=q~Uz
zEgou9La{){dvOW9?|t}a<rUoQ>C2*;r6R$0=4vnI4>qp1KM2TQTHs%Y+5g=QMw#ev
z!Rz}q2^t?PVfZ4qt6o*s14&I-gL2D~9<vpgy>9k)Pm<ts_doB;a*Qi{5ixl;6>V=i
z#d?`;)1{sFVh6q!%5QcN^8<lTY4JX&h6!Fa7R6US^A@LWr5C0jfGBeMA|>PI)`Iu`
zp+A{5>W)h3?<#{EZ>wHMv*bOLl|Q<Yw(#$F^PgS)my+53XPy3&`v2;*{7<_2Ul;v{
zb32k_vm9h#r<)Nnzu|K%7>Yucd|<B}$Q7z~&4QZol=G^r@?T`I1uLT1=xe$rqHXE@
zvfP(jfMQoK(>|E_=dA|e_($LYZ3OYaU~!TD%%fDqZ6i0qUP74Lfr*RqJ?;eT7x_hL
zBd*^gNl~>ULXppkPqS$@F~q^fx48Q@BD0=M%#Um=DFQ!#K1LUD)~D9KVs;?%KNQgF
ztdz3qdNqZYPo49>HrSXbf~igvmUDwEa`#bH=m>z`yUG10p8&qGujdzn2s{lOszEA2
z5KNNXCf1lMsJx-Nn6+9N%wwkF2Q{OR9iuU6-j}FBf65`@E;~F&p|zJyS@b>ceXOmn
zKd9?g-;%rSy*>W%AEs9ZV(*yRo+Qu!;>egqp9%7t*A;Fx`b;suPT4J!Up4F-k0vbj
zu#d}nYIl7swb=fNwh3~wmTKQY1VCpF^vmI7$IY^CiN4MEWTH{Q8s`IepqPIe=owpS
zoqsYLg4f`K^@MAzl^yvPhbJS?y)K{Ee{ZVC|Jq^yvs$;*9Zg)4X3tpe&hNkJ{Q93Y
z|1a177xM=HfBwI#x7Pg+!-c<JQN^{+V1I*tymCb-c~eFrGQZckRHjhA)=4*{My-9b
zA^CZ{8u?A}tXs!}FSTPS_w<a5w49*739s1%i0xu#q(++(O^=@DfAz$aA_i;Q=-UUy
zwj0ut#ml{YBUqdak(_09kdq-pWV6%U7*w+Dxt3Nmb2Z?p+qDkWGXqC!bB(F1z3=)b
z56+g%7A2Ow*UM>Adi}Dvz>7p-`PznLw}--3isaRTo6g-{P5!?WmUmS?yk%?8UmP>@
zFfL4>EL=~|{AT|0sLE7Xm&kA79Qm6FH49$1Ym8CWE62ZGn&tm!#bOQ{!h{-W2<D8)
zt1OfGR0Ce51mul2M>d9H%pvX}P-F~D*4iiUF-Nbgabt>3cml$=pN?XB3!S<?Aqsc7
z-6tc#Q=nW`rS`E(ehB(!0K-AF?4`xs=W&g($-&>+6dS&V<S(y*DIvPRr?m^0*8Jc2
zD7brDMy%TN;#=}-M;z?%RS5^_9LPsW97ZpqYJ5v$CX;xhZ><*@JS~a3<g`5DHQ#U}
zoj*?sy)k_eaH9yLc*v}MBINmPK`=FIepH!jA*6Nf50o`B?Du8+Kaaj&$@$@5?f+$;
zfBD0I`}aj|l|5!v=}i+<|I=A!SRS=}<DR@<QiA<!`DvG4U8W06IgOz`kiCvvIkY9F
zdnAS!Rve4<T~PZEuH5+7^}p=!FCX}q5B$II0Twy!1*^ZixAn)WLQKDU>%E_jf5oY#
z8=v-aoXz5co?A+JOQyx7(hJc{v=tZ3FiP&{e_uQOpDh+<x$y5J{qK79|BO1W$L|D1
zi}fpA=HC?rH*L8Z#s3cTC3UHX4_V~E@E^Z@^Llz+Dy2qaWu%Eu32ig;uDC0xqAu}A
z$#YLI&sz5LsNZx!=OsDU>}Ln(zhB|A+i5-d?aKbkKn(smz=WMwgMmeRg<b<qZ_{A~
zX%~mm<Y#%Ey5Fzpd=2@Vud!C}(HY=4B^h;d8oGA?HbJJTR7d_LE*!)*2!FrGlUed)
zV4G7W+3CT{s2GR!G92qQ12;LQ2({ryqQCvy1H4ud2R(tp0&l&ZspIKqVeR#JGEh>9
zooXC<t{h<)aLMeoxeO%l>+ctc;NqR_&Lc!G)~eo&uE!?x6!y4|a{Z&1pLgbOaFG7o
zT_BQY!ce>?**tX={ei!hevCrB!cB#*9U`P2R&i|RwCW%1D8*aN$RgaWX$ZTDP`%#!
z?Xc$_Ew$nIJaVISz9_{kSV1u!^WbD$0DvVe=f;OQ@vEaVSIWLt72|rCv4gY5ZM*2e
z)>C0>t&2bS?=+h}knwExNdIOHU)Dc{bJg}(s*&_*bN**{V}CvTluKSfFlH_i>{1e_
z9m4nRr<1L_k+eOSzb$0d3Neb@ju>_Ht*zB_M0elK6tVh=eKph_8O|x?dg!ZzbqC}<
z86aryFWvd#K~*;`dH!~i<TBiD33*QylCBhA^Lk+u94|94@ctD+)nQ+IM(Efk$zm};
zW815VU9lc5*igYSuI4W>oaO-?FtLeXyAH}S8+$3YX7nyUnz@z}B0Q*7jZl3(x}08k
z0&GXz8@)G{8p-Q*)2oip#3J&Sn)mN&q{Q4z?k3;AuP1l^a?9@PZsgIw4YE5!_I|LQ
z=~`d>^`E$h2G9V<EuCm)QRs(*43om9LODTszu)WZLnwL0na{Gk^qv{^sB1{1P`}N|
zJ<WW@Bf$A8(%Jl%$9usQT(y6O#4q~*D$yBW|L6V|XJ!4U*ax0+(k*>0hEb*t4Wm&*
zD*aK`;ZGGCJ1ZiGm<@BUxBH{(@2-#OmC6GG0UXxlQe)*$_h|_^pBu0|vy(-cUg*E^
z+kxA)-AL=caH+X8>vrZU{nz1=o7gZLed_(<K*3L&YZkGVK5J})7&rX$B+T{Ete>#^
zu&F!92WW#eqmow>=F0^Rv5&gf0+()80S$5j{N+ruGB`Z~KG*;84+C5gnGUu!2XM6e
zwHrM=E~=0ij-IMxd60|2NgDbl|K9wIR=ZGs;`=6}4U3$$N0h*+BZC(S_^rNi=VMR(
zAMXpR5p-`4IW=k3B0C&!5BT1OHne3sweQ}^*WgK=5&wb0q`8l|pKQ*%7c;j&c_jXG
zb&fkWgBR>)@eKWgx{C=>?_EUAyV;ljb}Lt|a5fFh->K*?eSPyALhiO|??Y+XJ3+Tn
zf-JhMf}SR-|EO@jqNB1|Xfr`J?yW70_jF_9`{M3o4c1LReGH0;d_FNhM{zh%3{>C5
zFBW#g9#tS*eA%3i1<bV+agXl%%LAMYlV<irnvRV*Rlh>keLZC9W%_r2$;N27iTpKK
z4&%I8gbHcC{`JnWsoW!b(5Dz7gE1OyaOd&B_QFy+MK{Oan$HFZSxX^(cn~rVcl@zR
z2r|pQ+8Hxl@lpKO|7#HcKlu5-E0<P$^Q&^>`~f;E=3mAe^wzfdCYBTCjq4Ze-psBD
z8J~b6UzdA0=-KM##@<gYEUI=!bcfX7#xC<GY=1ic?Xv>&UF@S9@|{T7x@_ZHk2BD(
zWj(Jt&=f@xqG)evid3rTA#=^~@3fH}Gb;c30cY})#y|e|Ea1J>xgS@%aun9^<MiPC
zhR=3cHBVb*aP0kF<Q;E2V=sSXicTo0#CP-C&X2XVnGNNyS%;NOJVYnTY$9b!N?|I^
z(FfcIH+Pe}K{vvRgxHbCMwb(0y+Zus7bEfl7F~&dVm263^Z?)a-imc)<-oviRv(rr
z15Q?*)v=rpVmanAx4Q@d71V9&R8ZRC^$3?ADNP<~9&uT5O{glwTyqK#EF9}acdbK^
z8cjVfu7MvmQq;Y)wB_V%Kr<~RwCzdx+h$eg3RQJ|zPI(BsI}j#$&XfP$ZT8swfk2q
zOGE4*vs$Ci{}eWi<d5Z%T`o%f@ZtX2&(XT8{)v35y1U0+xf}rLS&Ufr4Oh=|a-BR1
z>3J?Govr#}Om_FvlDK&Od_^ngqhT)T1LJY9IYegY&PEO4QMQZw*0~M9HLKY`m%Q%{
zr3m6@=6k#9{MNMjhqTpwK|r-3*6ExXqUQ`)Nb38QjoJmmxKU*(F9?#bpweABVSX*7
z7F`U--Lka-G!~Ray2$jq_Gi4}{E3qrec!8~^qW*YE!}ktj+wqpI}xVSTA*>Ti*7q#
zDyMn|j`_Pp21z1ehYs6)m3w8Qhg*bVSjK*$<6(%GqoDU!JP8f)Wi%r1`hxfeu?S6m
z#lf_C-z`PX)#L++sx{?_SHr<NzSTH7J4|p06561GJ@i{D!wbW5T%uOS5Q+4bkE<xX
zB4Wf?nw?>N-vRySUU<<sK|}tgAGn_6T=?yYV{y_!chp&_Q6tZXCWf{13y<vA?+rAs
zojo0XT|0L^cj^y}t;D$UwR*2N1ltqoo;sW<xYY$}C%svoF<qJHFk?$3gCN0_4rDt*
zJ{U`!(n&#wVa8Y8>XpF_BE`8jNwKi4!OXy%OcXswB!jAy;6m!Ig|EJ;k1~d4_=@r|
z2{>?QfWpjgX7D;pzvqdycb69Rb5in$qZ4cUA3VqNxhmj36g*S@DwHQhU_4RhQUDS^
z!U6+=zIqb`xb%m5(LpT@QiD3s>l)e{b-J{T^?Zaawm8S7Sg9Bq;MU@wsCp`-DBIbX
zsf(>lNrEPn1H@2Lv-lY0TnPUVP-fuNCC&NBx)a^mxi&>8Qf71R{+RM|8Y@9_mgU9U
zeo(4CTNDPor|RkJzK9}|)P;8w2UErFqo?Ko*Fx-d8_TRZ@WOb>Lq*w*YA}b_6h1oV
z;Zn`<VgndqZY0%T*$q)ToJnv@r^O=8kb>Ke{hMAyjG(Jr6x5(3Kv3C5Q@3zznJ4G5
zo}-{xMo~2AGodN`M;n=3Ah`_s9H&9$@s+ACo-wfi;`;N$@_~yCnGmiTU_pa#7QfUt
zo(7$nu|qZG#zs03Ci_W#Agd{mi11-RR|IP*0zQV_<}xyJ>}oabz`gm<6_^rt9l}im
zE3U}dIT#NRt^lLX8+{Lk@Hr1R%1|!R)S!mlajB3|TNZ7FNaehd(-N+FXey1_%0Pvb
zF+$IqcC62CZuY_cj4e#acv3L%wtZx&^oVW$XB+6gQ-O7u<^_){Zzz3FynU+ztGzK&
zCw(%7x{l#K4l&Z&V01PAiGgaz&ME-0<oG0SDaZzvGEN2??)@?&<u7}bn@3kZ**A20
zxnZ5)9F5N@6Fvgmr;inv1tyfCw#XFzZkYk(mUmjMojUrABGt2XZ0MbKP&~vl3PFry
zafy@{li&PZ_bI=>x`Fq0fP!g~yVUE=qVb6r4tEN*-`9V&bzANZ_yWYvhVh2!%oiAS
z?jk$mcRo@*QNE<?#p2Uh6$PsxJTA-H7t%^N1lky3#f8@IYM%uRbW|(i4u|RRLz~P}
z+Ny%0`^kKZFTB6b2iKTsMunlalQGMk%_<Lw06B#_!R~vAR){(WqmiJk;H9qMr3PYU
z&QS~@&YBIE<LQIgR}G%8Kq(s+>CPDznr>mt@RUcltz){2o~O9|(3cMEe?u<O&wb^+
zRMf#*IQvvchMHJ#)Hkg5Vz04{6hU<BulQ|tdiv=XU)1X=adv4OlP+LRL(9>rlS`~~
z2W)m=8}o1@M?4zNxkr($0b(m?CvpfmpRLD_A9Esx=~A6^Kyic0`l`ytsw#N^xr2<X
z+*U?!j)zRFBi^k=6#J*Y4a^WrWnlQoJ1P|}&r}+Dpd@%K0Klx3G}h$?f!*uT_{%IO
zlHwCPVXs#)ys}PT1phH7b`n{$=bz0N3xjWr9%K@jhAj{r%9=*)-D@J8ui@G>bqGDT
zlK$WvFP=m4m}rsn1b%L;k>dNESZu_Zhvt+O9{kC=Q67jTTyzdWz=r!;{#3{nuj{kh
zRa<121~QI;#0)u37PDH#>KD}aVT)##cC!gXGKQhnICs19a+oP%eA?jEhcb5oTa0Bu
zyt@b5{zGVV3(r;}_xU|FrQPX3<7}kIcHAtYt5hc=O^Ju>dDY!LvqjCa|Ii;g&;`v+
zR2afy%Vxm32)yGfwddo=qe0TR0!bQfn*FCcTT@`G=eTurbBTi^c-*rrr-HTv?hc^h
zGmd`Ue>=}h+D3Q=FwFBC&_W<@LrJOV66R^L12@lz6We&gMweAGn4b?IHl->cvq`(x
z`$6xoV+d>7gmG;e(=y6sBXOO5f#olaM836-=EV5Mt!;Q~5<zO?kk<IV{Q30<V~TIX
z^P!AQ<+f@r=xg6QQ3Q>uoY-pZiM51`&!M~ljZKOc`y)Q;7R`Xyat}tu4jdO4=xU`Y
zgq?W{I;R$&OZOO7@Y?~&Y#C~@leX#jF+g&PJ9ElHPv`!vILyzb(ViKj`Ym~|`&(zI
zZ(EwSXS}I1At!@U^@IGQd?21j?>5r3&JGVA>>{B7J4>q(i(}_?rL)tSZzGA-UmuJh
zF+fsQyrh0}Mzp+hCw5<(6%_+r(ZHiNR&yLl8EPgoQ#m73JgBiGfCymF?pmg}*iZ#w
zI4KF(m^DqqWNhS4=7d3m^2rhQ+@w_m*{8FRL9h>Zw#*HRYZcqa{@iAsIGdgQD8}Md
z@Ak=81phX`RQ>C_Zm%R243C_z0JHuai&5Cq?^mv>3!O1%pD!K(XoPyrex}xZUw6K@
z<sg=Im6UA#rt~bg_DaE87tV3*!PW_C2^E!&ZA;|vx7(WyD?Q9SV0`+1#bVXjD?Wl#
z8a2&f8Ka3iOxXu<JzN!5J=&SPH)ErCetFUvj^&^*3AVp-u+pxrc2qKKRNGE4(l*+<
zUo23<Ws)azUCY`@(P4AOkGSf5x2T+Tsn0dtO3mHVME!@eYZDqt8glMvp$>Oc^XzTV
zF5P-&;$Xeu!$zQ2*-L~#Y4JqnS7FixmNgy+)%HV)lr2x;VTs_5Wlyd;zsjnkK{0H;
zdAtr}%aiLzVt?g9Y?y9iM9Dl}IH635NS7)10;?C4DQ%C;{c>jGAJ`P)H|lZKN@DL$
zt58=>D)g1r)gRkuyB{st%B5fG+@?v}x=5#<{nS>U-dFHY)lHeX+h<<H<6U#&`L7yp
z=kB1CPVB3XUEWTdW{y>q@63cWW?R%%n3pXLqLH5NQz^!<0jlMp|3xy!(aXOqppIjh
zd}`a7yl@<i&JA*(O3lWD!z}>PNp|p(&W<?Ld5ChopD+wN1}16X55uavbJ>TW{a$}i
zRf)iijf8hADsnNSuZ}WO1HBSw`i3eec|;{rb}J7jss*pISrr<(y?FoLtNT5DwK!^+
zQrJ@RDtCB)eSWc`FyV0T>a)FvV|77S5pcLBnCz%UanWz)$xtCE`9sfpDwF7NMn-yZ
zf&{f<YHfnp<T%Gz?HM@;Zf3^mdqUUdx_cOQnj;cuA4k#@N`hWcjA~lT*qSCq*Q<H~
zXQ2n)A->qLm^<z=v44D7NR?-L$z4!S{vv8W;k38OUu}D-p<x3efLUeU)#!?uk-PMf
zI9cse#&dP{V9^8j;o!%EyMu#^)k%#CKyoy8kXUYxH3s;^0ojhkhc~i3!i>#qhg#|4
znaxf>)()RzS}^Y(i<f6yIcluYg2T*h?1{8v2s_vMK8MpxU;diyC5fycR^F!@A)XVs
zvJ=p=r#>s)H-vD`Pv2I5d)y}lIWU;N!4)L7-~GjHBQa~v$6wF*m{m)yY`l>nK$%r1
z-!Am<y-7#44$rPvE`kc10$Mf1CTF->drBvSN;Qq$<q8kJU-=NWQbB)Q<Q1QEkP!C!
z0xEG6g?}98J{CHp8F%{q%4-kLubzh-mOd9X=UxXDMN1*#xp|@3L&1up_4TC<|AsRL
zv7jH*+PZK_>PDi!bRLe4Tg$|#b=h4F`)mA*D&~ZILVhALKfYY$QU{%Jvb$Sn?~Q+0
zAmyF_zlVAMfk>9qHDx?I=TUon;68Tt<nK!pHo6G;x!L5(fzJbRM&%H2O;$7Sp!hI0
zD>%&9&^SA6Ocw}2%7bLc^Xt3rA@u1bcDrUgCE6V2r*jP#mhTuMD4Z1=s9;Nv!nc&t
z2)c$xRORDJac)7$#+u2T5;soCg1Z5DA#nPF#pJaP+hbOLQM7EA;Il2H?5{80H)!j`
zMD>bgIZLUfsOhxS#4&^~&V@9hG+0nCHnYE9sR&4M^o9&gqp??CkX^LKACBO<1dWm|
z`Od8f8E<v?Jj9Houn)R>F~q*Qapt%u;=LIU*6sL0|6DwX$PE(ver2RmBgt&0Kp1X1
z48|1{=cMoOgPN<(?w!0v$OZDUmJlzh92AJ5Pvc=TGa6N)*Pe>-<Km6F1fpj0`RkJk
zX76ompuJ>P^vvFV@S1yt$0Tb}cZW8v9H4f0hPyij@@kp@1l;6^LUTlXJ<@ILP><sM
za1_5S5N2401=A`MX`6$9?08o9;s7GYI7jF_EAm9W>9Bn3Bg{I1XrW-G)X+H~X&UP|
zYD|-8Ywt-YQ=zZ6T7|A{A%&_tye5l+t?OHMxLo;CyX0GS%XXU#xCp~djtVK@?8Vp~
zyg*}3fAv(LDNpu2hI4-NM8J8_Bb^8M`*jl^sP<M(@VY9&KQ$ukZlbgTWXe4&SSydU
zXZ^2s06je>u-fCakv&cG1y?>B<L;QfCAP~1hbO(1eX+(_FN)T5zFR^<X;m;}PuFL&
zc__BXm9StM9i-hkh?o7eB)nG9oKa$lcWAZJz3by>L(NeIneETlMD60{p)H6Z_nVY>
z>o&&~t5-CDeOxx4)j(XRaK9=e27FKF;mPhc-*&CJ^8UN;SA5@R-LKUdQTYirYH#HE
zlb>e6_bUnJ9xN>7wA$J;I8$r6wn|r?%B5x*4XnT`i`kHab}%yHY#^KrO(+Ue<`CrM
zc}vj<+XN_d=OtMlLXC<tVh5#ijom4h#T~=v_U85kQXerV=tw(_9WU2Nnd;j5Ix9F*
zAHXaVX}ZLHsX-PubRBZIFS}E@o+>vR+WE~_TB-jl#3l)?^g=8{{$pyDbkYu29_S<H
zk2%$oYadTyTPy3FN!=y%AoNwSW^pe{@uHDN3f8{8BSyxN`M}iHs?N|BI|hQ-Qe^md
zJ2f7hR5$vi?hFk3#2nCG<ERUE`LeI`&*z%rVA3<IYvq`9+GPm0CD0rly+~}Dh0lx)
z%44hvia%QDWv+}6e!vE0J;Ou9G^Mz=3!N{Iw7ndYv3DKy)x}{Vd5w#avp!w+!My|}
zxX2L5(DN1`_b-hkzMXbpqUY~*wSdEHj9s=47OnEu#~imgd)0Q}a1sd{n2jDB^@h%%
z8bTl$`SWta1FQB-ip&KKwa{cdtQv~j>^tOcRyt}1;$;YGO*Rk!wTux~R;^`zYD%&l
zs(Zt~cXcVS1=#3$!@9=BD5?15f+u{+QRcx$74FKm=c0mJxWn;ClC#rcgnLz%UkI*l
z(mGFu(&d{PEzRykK)J8yGBiU3&m0bdcjr5@mtva`2*_}Fxf3j9qcI1^&qE9+9yW`0
zEKzlsCaPJZ|4an>F!~y{JAIF6)F<GmP#I6D>Lx-ni(!Rf9RKb58pqBv?nXa}#y_Kz
z>>)!o0!aE}a5pR<L1bc`#qktJA^A=Dy`>%bV@!5r&-U*iwi(lW;j(2hJ`^iUZAGkm
zCN^j0#km=D1SssdLe{5@A;H35BS#$D&zl$R6E(z60tpfW1&A$Rw_QfzaBhlnMh%r5
z867Gj^3b(T5s&JpT+9rE5F!brh>3<S#KM#cXLSj%tlIa^gVrPfoPjm<M$8l(N$rCT
zy$t<UZ5I$b3ESsBNFSBMB1_moL&070mP!AfVJQC8sIQOKQk*%ZrQ`({C%Q28BCVWj
z^0gv5;xPOo##P}O(LpTh->)!~oB7Ukb`9D{r$iaWBI}EtNxm?93axEcf%=M40<+|-
zl6B$ZTUTJ9<-Cc9e$CDh`8|-?-T>WhcdsJI6^*eUd4h2Bip0Q(=>fz3px88FQY^(7
zAxWec;PL4RFsgE{Bc{9>A!n+f3!e@doxvEEdwhj!4>LU&{HprUmHg4n(r#r!;QCWf
z75puUqI?dI%OjI4pJSp;Oa&$5WiwdVZgvLbE*ViqY9GhhJ=%=Fo)zOGl})ufoobL$
z_HRi}F-oaBHo_LVjV>?y4AZ}gurLqUHOm{*5x!tD$gBe{I6f+?04LP>+7k1<57k-p
zMrEP1>YQsLa-i<sX2AL?>H#`UGSh7p<wjM`QT9$QKMM=^>t*IJc0&7ZxoEiVTiV;O
z81l0!vRPkcq<NyVABAJsrO&+mRx&{Cg$F_6@JU;42(uC#09zNI&1KTm2`dn!?Jl*S
z50z0pqbZ7ly;GSNMM|f&jY|93I9E&Fw@wq)jFt*>sJE;%3L9dMYEW^dVKEt_BxJYx
zF@}!nx~@(Vli4KzpR;v<G1}&xSssmJaBXwcIt@!12ASLH$_?z&$wq`awO;wtYdyFc
z`~@)nf-epNN@8W7h&h|Kej!)X%=2Mb#jxA=c*aJG9nTIgeIU-MJ)$|Zy+fAqflM~Y
zdpsa=TN_cw4hCK{+T$9z>v>a+q}6PR9lkTM7A5qGC>l1iKEWokc%GPodFc$v3_<8{
zN}V>#bjo=j#`;TcB%JA*n|2_lRFx*ss}U#By1hk|I>&<O7}93c&u-Q)C!13FKH2)z
zV$@DW6!ofv6%JmIoM7uGk?Uzf2X-ha$gqbCII=Y)gd{x%G39Hk$Bcucfg99Z%53Rj
zvYP2o8f(A7cEDllMlXKUc0@(^++yjXxLaMu+jal&B3L)Kc`_7<`$l2rNVofK1Q!uK
zI#K%+e)F3|&ZeL9xLJP&&jypgZKoxvran$Ow1#c^zUb(C4d-v8)7;z){POgaoh?|Y
zA(dFT;MCdBAZ?@p4b@Oz&u<aWC)RnF6ywD&ztM{gR>#=x&!nEZ4%^D;B-ABjL?=KS
zF5`4xUGq!mgTLEF$Xa#vkAU#iOkz>G?<h$)1DFcaV2PX7svv9W<UXuA6~kh4dGCCf
z(J0v$ij^2E&`7m21xYZo8}XuUFj)3j4xlI*c=?|#{(>3NEd?fD0HLd62j5y7v5gLL
zKIL4Bs%)NC=k}m-0@tPuI+amMFUV9W$g%b+2-s-ZC&WwMIwy&bLTQ#tbq(_C#jWN^
zDKOGtX0y)$W!Uy$pkq%B#9XDwS0ArW1~$956gXNtd)g!|sK}Km*J?U?3i?cya0TF5
z1B1)~gS;F0`H*0%UG7j1^WKECB>-JNB;Ps<6t3|LJX;QRLRj=8jZx370b6@qPFvY_
zSvK>!ec5t$GY!y`T<&9r`ytmF`@>DE=3y|FNVW$;)BsD~Ys+py$k(udYAt8CDQRqO
zyrKlXj)yy4@#`Z}pK?Nj@JhpFP^wapzN+lKRx_ujdQ90GbQMfp*>-cTl`63&{I&Rv
z3BsAp;Y*58xOjfOq3yoy(UzDpbI@L)(u~#5(f5HWg}98@J?}#20`?h7=X_0SyWmdX
zah;bUf%oXkgiV2$>mhWztg2FC)F)e9T#nh>3-jZQeKl9Pu_D}ecpM~usH;Y>cRiVG
zJ}e)uJBXXkw57tVTds2&T9c`>y_HGH2ue@>KM*7ziW>OeO!~ne+YoR4Wvi54T)r@c
zR7sQm`;+ROzy2B*jsT1#TzliP1el3cIfi{ylzwRPHQEF3<Si$w@rmfI$*0a&%lRiY
zGNFY#Af(|EJMu)3$kBzc(FLb-7FnF?*bdm%aYdU;j5G)*cZP6y%}4D_Lls*pGd{xA
zgDn3>d!aE4wmseG+H-hGBhah^*dI)$=mxGTN2=T0SPE|m`MkZab>R7XQt|i+0vUp|
z4>FE>Dd=MlOuWt3gWyIq^rCNb2{+92tq&>Hv;~j0BWeo^Q{6%3b}5<#^x?I($Cum4
zY{9ydCyiZm%TrT!*=a@KQUt2{b^ei2YvIOtXi~HiApcNPXQ1^inJz(X$)vhqqQ}En
z&K&20k^|OJVns;9EdI0Dqp!kU*8q4DVE3yhc^ElHA*=h9tOn#brXOJ(qgMA0F>qD6
zF?A2J;kAG{c!Xfiw_aqLg+XI?9z%p;D8yd1-&V<>)fOHDfd?m*Os45H7fs%0D2lux
zab+j(I09|j)%v29`Al~1=<qh$9vPO>id!Gq7Y7aHj!ttS5qBONr*ZB8a}<~)*)HX>
zz+hZ8my{o!7Pn6|S{Mb?_Kgu{A9{3!ve0&nl{uSz(-{`!e@6P_=grmaN0%WR=Ot8p
zX5(HPj}p`?Z?0wiI56CGu2k!9Sq}gsQYBX-m+3QDJNPW~j{X?+Y7#Cz1GWT|kre^y
zKYjK(`xP?)=8;0yEzroc`%Cc06Rw5pYkNWtj)>`?WPOTXi<khab{T8Z7OPU*9FBpa
z2CUkvIVeh#9Gwl#PCj56FnjkM$;HIjphiDWekOK{3ZpE3G|#cLj>cW9MoMBN>Ky+M
zASpd&9aE<C50e&!#X!Fu(%nwYa<M4T8<|+=%KaHdA;dVKI3-Lc^AzKf@8JvXiTHi0
zNX>Zq!H)kcKAn^dKubvGy>-<Ee9zu-x2-)(({FWUkm>Q-S}KK((H#{B4ltk@I#QO^
zJjT$=j80TfF#W+{Slx8f>wde3+wovmxN7ELqSC=Cx#X3j*0R;%Q8VNlkxLUxD>!O}
zs<`CLHz+C@dHp%3v7@NyTh!2O^=2i#MnJT$`uf9N52=8|KIw<vj{vozb>iJguc;pu
zc(>g>yOp915}E~|UzzgtkA=wnUvfXQ=;mJY**adF)7T(XID2lfAZD>vGaR(^T&ypX
zh!=EOTZtU)7`q>4*ty;~E(c_VsLPEHID&4fnatV@`Ph&6mK`uEAB@QEyT#dR8vBn=
z_OubSNTF3NZX2!0%zzy=kW4qz8kOE5<Yl8gy?V8zK{*Oc<q>?2`*30&|2FWcxJYS|
zp5o2H9lJ<Q{`Aoj=x1al<9PbLiLHsFj70t`t&RG*PmC)J&48c{KI$NCa!%BGOnSXR
zHxl)BwH>21x~AV7590IQ3X#54F5vVe`H65@^vB>1bQQ%YA;!a<;I%luZmUTff6(ZG
zer0K2?-*`jHVZ5CqF@vpo|^;)vP1a6beXL^dku(jcj#1U+l^S}t>YAIW_hH$l(cxd
zDb*Uua=#?{52~~yl4{kQY(bGW{3es;Q_9N0X;6MYLbgwD<=$WoN&V_zZXx@Z)6YXX
z=i!z9TyJX?rLqlPd)|rXGS2pdV9hHnILf)?RHCY3US8+s^Y+kabhbHCSN`C5#qC6u
zYP34X=qiuEn}?gnsROdGGO^Lzhp4Hla_gM3X6F#I12A}VPvErLpBo{1ik^JSm9Loj
zICf-_T#YMuQq~-%*)`uQLW(so1<LZ7m+K+k<+rqx?Y*BzdF^`<YLlO-h!#15hPT#-
z7846om-ZwL_M`_JgSgyX(P5hU@hleEVk^qLsi6gl$;upE8oKy=_>%QHCI<D8mIJ*%
z7tr5An2nFfoBu-VC|=D=$#4<b2>5>GMRYz1%Q3x<!6#(Uh-yQjj)~rb3rn3hd``*e
zdJ6@tKsQp<tpoBXh~FsJVaQP*Z)drkCp@v0AYteKO8=2}O&(jHA~GuwnjBQ2fv29|
zyNGGIi&Ah5(DQGXRQ-!&(|7T$t~|79%*7(6^o!Ki!<z3`@;jCs842Q6q(ky)XdT*q
z@-XD$@pZMQGpehrR15^mTta>@g5g?EZ`GwD#f@X<s*mXO#3KcKu5{v5&e_qPQs<uT
ztVKWm;1+YN<H;OvSDradzA2jb&zR=-MD-b%9XH8=HZHos%qt?i?PZ|PE}CmPJU}XN
zjSl{PrJ#TKd7GymM6uJhg8dPKZ@l)c-<R&;?^lfJzHY3S>j5FJ84?316Q4w1Ejt*n
zH#a*#)yF2CLPARR&>d2pl{;IfC=N?Aa;eoQwwGb2I3fn?thj|%XIB^==C>zO)`mQh
zBCms}+Qg96jqR#s)NrinDyNc>OM;2Xi@#;M_ewtHS&ist-$!3}RPf6%;SVX4!kE#o
zza1%LyQ@1Qsd?}QrEuc1KHV6sGTfw4jkj)_nf3>d9x9D1SW)nU7MvJI&W;J9(vdo_
z)f@sb^8%aOnhzMh-aO-*=iNRg6L8tyIDP!W>XK#Ufo)g;nR9yrzY^j#Zf|y|<)zlm
zI$6<VcT&AghfYz7Ej8jz9V{aRO^Z5HrrL%+9m}fnSYT3}6a*-@%+e&xIz%ENJUMFy
z%$N%JIMb3srfF)c?oQr#xLwj5-wtG$3XJdfH~Ui8w64USPbatq9Bh2QQsR^SJhs-6
z^ZONC^FdrcnT{$wPl!Q<xn$G@JM2A1dSrVV{Cx>1VT2y*$Mw(acQCX=8#e+0VbBUI
zU7$K=>3NQLtK5*^{BAt;T73u=-Ut|-b!h+SJgLB0x127@_lP;hNcs0p4(N?y;!l>c
zFT3J>B0bQbdbfm7#(i-2STT0-%jwkC-(Fo)?Z@5kSB8hWqzF}NiUgeX7XpY$9!^Oy
zB9sndQ&jMN8*|p$jPhk*tN{0ZWZpt2CrZ3{OVOQ{Yt><A_Y30O5w(Te0^4ADaZMrq
z;-I<%8zJ4ggCI*@D3Dd&cbpVjC%vP9oXaQ{({#!Bq9JMf>Nuq)<V*WVwg|pWM$LHZ
zamUvp${M%H*V{$bGuBSSV)};P*d%5ifaJg&cVuSXrp~ynu=gu6J(y~Js`*UwYV)sW
zY(4WW_m;M?4d>-M6|Q$F1qj}n0T6}Z#~hm!&mPa6M)lMlI6e1QxzsDpiP}iWaS=SJ
zrWVsCr_16+bSgj?W}hZCVqBJ{J#yTxF&P+EM|VBXL7RDj1R9UGlXJMx)NwdvOC=Ia
zu(|I3pnp^D(b#+KS}xGTx6wUjey*|CQnBd~9wVYpM&j8Z5;2P4L!D`t3Hb^#@^g0D
z!sA3<OeFgH9L60%h0oTNPGZ1A*#?Rm*2Q<{`!cJFDLGiAw<flL(ybLbB-^LKi3l<E
zhx3QHSZ1NB5zsYDycdoUYlrWg$e-tAV5tH7IYE8G^(lLpxv>fVh*DmeqCKBe-gcSH
zrdtqajaPqVubjD5-C-AO{bbfCzG|+-5Cv7eKQk+1hZwr>I>om=RVu$YkUwDBW$6~u
zZO*0|F}h~Q6{YLY4VC-Yk-GcVes`uv6m=>J7a6!0ZOxpkPj)vur!M1ne-(V_5STUc
z47i8%-vX|buA_h`UnQl~8pf86&K<kPL5tk@WF&sAycl*-wH2aX5EGE-;GyL2dE4q(
zq&kFWEc?aZ0zK8XQ_`@l*+Fp|=`^AA2K5&MUg_cxOOh|zSTCl}(zuY#<pmuocW+-_
zg6t;ffwDcZCp7UKMD6jl15<SJJUzBzPZ4uzI+zUQZ^K5V)wM@;DfF+avl-2D*`-b-
z;-Sxhb*9wu`YJLqQe``IZ8B4uSO(0j-d`t`olOvEaHONjKo?;cCWg<Y;cTLJ)F>po
zxi8`^gG#5gFH+O)p^0t`pOILW&$8($X68JA+;3YU*G1~j6=J5UD^Xp<UQkwa760?t
z2m!5;Y2Dv?MT<~*>#Yp+W`Q%yU6ytbtG}Q0*9x!gZ04g-tX(#d*4Bm<jt3jJW7bb<
z7G`7RQCaTB@{B$rwyZSg>)(9GUBI<E|F^{ajdA&&HuBe!>UgeC!dnyhxFAKBOKW1v
zpx6|Ds-{cSk&47*?K;8U`lvdOw-W8vLC>nwBxl}9`<2kW2*efUy1e{M6azX3naiQ0
znL4=u=~~2_PyN_q1fMgsHsQD1HTj#2C-@w)8UPm~bG-Q2LeuP|gQmXMUecWaM-lP-
z@rp)PO#(7Mr@{xI>P#b|wnx7x#C@w%&P+fjrX&MO4*gcQrUJ2-m>n6UVLY(6gOqEq
zQ#)mh#P=hv8p<fv;48Su&Ca7ZNda5~`OQ#$MPZlRZ4|8DqI!l9MNDPQd+wa^K8;g0
z_gzGNF|Puz!4%6+nwJvW1$Gy^E^YDq^)ia()~ma_JJ4-76sJCJ0l06SL#+Fx@Mxo}
zJMMN){=kaU4B7%kKlEd`(sFXF+Z)c{^Sl~sGrUFuq=z!{v+ky!yJ8!H!G4b%R_&@`
zk`?2+5HIb!hvpr=Ai4H3sUK!I3uuCGT#LirZA5gOg)5H$l?EP*n^&p`I8Dc)nTs2$
z(;%#Ot4uk6%BZ$S@%!|=OTA<wM6c<I5(3<*wX@{fuow-|Eh)jr<U)AJd6VCEiV<+$
zK8e>ax}G2)cU|`J!7#m|3!qTa=wF;`We*9F8Db?&2LUdMKDwe3K%f!iLA;!z2!%YX
zThZ569lE2fHZTmr5yAi9C3PStdzs4uKPKE$$|-?qNSZ3Xa4>hY+J2Qdrj0IkO;%9|
z0X%&LxusS|@$Sq0FyKn6ySE}glI8o*sy5uXu=!wOzCdI+{Iy4Rb_I+(h~~@W3ESGT
zSMk`UZCPcJ3|_Pk^r9^;qZxDB7Ku(F-9>`6)l3KW`hO>Q3e>F%mX<nL5epevkmqu|
z<968`=Cn;GTY|2+U4u&5MzIPeYBi{luhTbHv9@!s(F$dnq$)l5HmA)LclT)0TWe}n
zv}s8y<yJ3~)cD=vFZ#y!uj}2<m4=<|xu>W4oO@dS-5paPblZGX=8?89(D#3$Q+_mT
zehjLND&~@rn6Ub4w(LkuigrS8OoVO)>#{1?5`gAjf<3F3Z$QwDqeF|UeN;*<D8by|
z({OkG<UD!2*DYJ-xu>e2^P7MnK!1c^HN3Ny`hKeSmOQRkD<y#Cz!85DJEUE%sWm3I
zw|l&4-!?+H)5vbOi~paLl!h?yB_sp6duaL(Smwz{6G*V`a1-ZlJ~N^sy=+xy&+DOe
zFhc|?cxklO?8$YU<GT(`ZL7}=3>AbPmX7pKq}C7KNL%K3k5SandN6i(;xIl=f}Pu(
zPa;%O))BD$){L&Tq-V|4!PemXjXE)<Be*cvsQGMEczMA+2;;fIGYoAtN4-9jvSDWt
zld`3j)_Hr(5r^#!55VNhP>yMUQiXW6fl7ts<U(RGt4>q&gt8FlxS%vw8m7)%-K^W}
z^0vp8!+)zk-TQG=-EY|ThH!Erd0j5vnjEL`^_m@8_<q(@aN3jNqUvaf6c-zqGdLY-
z4;6jW9b)*x22qrT5}j&3*aIK9G34MZx6fY8W@?Ci!+@twZwDZxq;c7k=4U8h0d`NA
zN)88j(<oQVu98P-tauCGpzbVS1~fxbcq8i0yyvY`oAcc!73DL!hA0O!kD>~UVkFT#
zR<$0G7|BC2*>?KU&yCqHQ4qf^ir|L-{ZfW;ZFU&t4hVQj=Z;q!F4rSfPOH{e<0R<Y
zds)xsR;%DRE(xlzvyauMNtBawUtHQ!Uf1ZUR)KY}{H~^(8{cWy<-P-80V7s^J0Kvo
z+HbJhNZ@itai$czqr<NP2je^vGS9WYb@dWCD}2B-`bI@@0Njs0>g^L!S?_DBoG&Lo
z%Vj#)h5q0>6?;<eUtvuzC{tOd=jU7!l}avw5jj4<Qbns2Y!>jwO~a%7F|lL+T(^4<
z&A<hL)HWM)tsH-yPUfxpj8z`gN~IbkgYrcqem!G)&|MRnq*;I+z7ztx+GIR`>d1~V
zt_P$JQ3%b#xGl1soq+O$Uw)%pQoIu2JfsABYO%VaZv6eqE#CHOHMhB`ZTc7a;L}Xo
zMVd+uK-$py?@Jj3(o{V*e+`e#$E%f51ANR|W+!16f5Js<4kNFSd(D&+ZpL;H^6Pp(
z8rJ58OKiSPB_bxl>R`K>Vk7xWK8-t?Kblh8l9Bm8*n9J^B=h#~f9_eDsc9K=$-PO-
zQqxkzEv=bJG!1kNabF4*0|Z0G-DYV~L2=AX0hb9C^+GTgM8IV#m0ZXa5jU*dayN6S
zEW4kp`~Llo=g;3Czki<NIlk9n*MS@aAI{HpU0$#A{XWmnZj(1U{qc`mgp+;Qt=b7>
zaiz3|NaRKteuXM^?=Cz!TDv@NPYbum&eK?tGIrviRg9LPK$$vHay>w=ja;o1H$bI$
z1k;mQG1;H29~H3K$P(2>`WK7AbdsKp$#QS}>e5g72liWkm>j>owh;j%A1ShX@}PRc
z`>I@Nef;+#>z|=&{-lDiXKo^C)ex}*X&w8Z?1|OG-?+DXZy(@Xq>$T?JLZmGq#jPu
z+K6A%X>ZA=Hk%PsYeR{F1B-GSsQ>M~C?INI3oAU;@Zh!Tj@7OG<+&PDL&*N5t3p*D
zC#3CHk4rRdXtNVZ=C=SJ_S5~TiM7{HUhZJlgN?9y#i@rJ2|?$Eh6Q~~Zz~G6-Z1r#
zB@YbkJXm^r3Ce_@K6l=(7=rx+R^8NVQa<skGycb?YYX3h4d-6}=3*2NmU_2f&8v#5
zV;P7YxOn?dLU3e~);re*2hzsPkqMnB22_WM4|u5bS+5+Zk8pp=WH>A2{K#{te2p4J
zQ)crSNOXyhW860A+Y)BJoFq;SHnOFIe5p~SbDWv-EESo9JLNy$)hCgN<)AsKYBL@;
zzdD*wf;MOhUdx}jx9~Rn=Q4Zl_g{5ee81NQD$4b|^<v?EOEf{<Km6xJ&aPttw&}eN
zsdoF+psJPAxum*fot=5%+^H8XzfXLHYdx~0OHf3S&_pu<+porF9EJ1sqyiXeW+g}C
z+ERy)(fDsZA%U?T*N>%rG+hggTUl0Sm*Umu<yQh9J$(XJj+g2$DY9{<f{wm`5S6|{
z((@O&G}Giaik(tIl0GnRq<%=PyF{+*p%%oBaXtQH`8s=oW0Y@ca`{4r`v@X&p$WM1
zX%T9Dd&1SS&aX-vHFx|4v18-Ln%MA4LRu6`__zOvHYfnvTbJ5n!WOjS;;7kKkX((Z
zb(~p5$fBVC1G3*i;*oNHa%fTBNE|Bl>95c)?B)l;?FyM^Tc208M3{*F{Y5B{J+@Hq
zq8>A95K7x=P`w#-DD`EOYP21Xj@R?n$`1t#H10(FiofOnO_&iM;?pOXp=%MYMOup$
zdR$!F@rEmN73vK<uYxGuo6OyW<?4-(8|JT({j!Hje#=K2$@msuifnJu!Oft!csZ`A
zzC40JnBhEUPdH5a&OIrV9{BQVi^9=*UL3_vjj);m{lX3&N*zV?E3CgVK^+}BmuP>h
zspIjs<C7=+5_V`mB&3>Wl#R-dGpV3$!XV(0IriyPij&}QJwZcua>}{Zr1_V1WBI`j
zGt8@|kC}HwD3AN_Kc5RPciibP(JWHt^%t$&Y+cB@t&9TZs)veFb`H1FCUm!55^`GN
z+A^&998vJCy5qtBWF7J<GBqeVzqyrLv7WTy`-FM#dNw`D(Fbu((EmCAVw0U2AoCxw
zH_9)$FQ>XY2Xi*gJa?09!}QrC%cxIM$Jki80#`CRn;fG*)OsytvPNt+yP#h;#pr`k
zG3rg;S~O+9<R8*;lJMH>NU-Z4#}*Hjnk-FANVo6Y_D8~1iRBlMdq?!2tl$2{Q6=Ta
z4Z>)!9{#Mh?|59jz<f%+A-zPFMsxiKCtqUkMOVY9Wc1WFV{*A$#L3_Q#m1Ai^nliV
z@3Mt9t)<*heBKol{@d@hcd}XNRDJZ!*k(Wtuf{LLE4H1a>!R9QcTTg_D4=A7VE%`D
ze<X&w@3&vMP<?Z!*}p?iO>_Gqv@&kB_)^MTVMoqKTdp2ALH|`{v>lw8Yb0XXm>mE;
zk!>%d%6R3Iyi*zX$csabxA*rw`A;H5wM>qX5<g7|iZ&RfhYb}-tNDahz_wK08(1>R
zct`8$B)qR#!s}e0z2$ujit{1cTVBu4+DM<$$kgV>Iw@S7)^lPweZ@+#gZq95cboGq
zw#m1W{cA;gyOK!9FKT!8d90<?mMM;(#Q&;@^zP8INgb_<$QpWk@E(a~=bYbvTr@)t
z{P!1S=lWwFwK*4->^4h?hzO~7e^c8^#D~;CGvp%RnK{8E_I)S4Q?h+drFwRIZl^6k
z%G=m(#g#tH+G(<p>d?4x)%F!|j%#KJXsS7!-kaqXmDD#@Q}c<>m(9pN1gBK5oNo4t
z52@ExI8|;0r)fTZ@yfR~1fSITTZZ_~lOl7*Xz|hG_@FznY3ld17~6*<nSZ&P5h{mn
zcpM8nS31^j`skFaqi`~~YBDIRBi<B`PyX#Ix1s}qMFvXS?R&dQb8@xRoXuT$JmQo;
z#bmaak_0Z!wDzuR(tL9Gy+|)>f3pc;K~tDtcQ&7@rK&)0GEU6Vg;8vKBC0ojHeq{2
zaPxplYFCu0D>T4jDDV>JbW=NDTOaCq>R*zQ&KFMHWG!ApwAEaHfe4-vVRDgHzInDg
zBkxEn=S~u?v#m_EzvuC9&O1A`)~{|ilEwa@ims(SQK~(`vFE;*^10zJWY$D!=5Hwc
zpszqRip|5%Ruz9d)AwzBqo8FYE5Ig@qrrLSpQqV)otnLoWFYr#T<JqOk~dKbueY;j
zNc&Ke%N;%OAMD1BzN)adL#ClMXU~^<jc{+pC8pWsj6M-FnXm48NU4?=2RWFQYozTh
z9g1h6TbShT{8`Le<3#Mnzcwsi2MpLeBFLMYFG$AUl8X%65$7Xc%2|uDa&#v>tFf__
zgERNdytotpDUUNF^<g(H2&^k9p{4YOkC)fZ#71&ED&`r-609bt=kCSuoV|~R`7EwP
z;50X56^kb(vwgCowi4E76Ia`#&h3_dSouDl877iD20u0kr^EFEy02}We2ufSUA8n{
zAdOphA9eZBB_sM)9}Y3D2>B~|ph7Fje&1bHrwS-l*7HwM+T8k-H?!8kp3m`r7Tu_k
z*2I%lFF=6sr{$Hn57h4suw6`R+j?uOUSF2NX+y3qmpi7w-oDgdmcK0?YUAKhq4NAF
zpPol-Qe+bKI_s&m`lwwg(2A$5j?eOvB`T6IA_+ca*s`x;g?If9)?x1<jrP(PzQmuD
z-h`W`KQ<$7$&d=@rtgKkUte}4Of`~5RxFJ&ObtjW?=_~9_HXZdr#`j%hi~hT@BYXV
zt*q#bRgkdi@-m`H!?^KdnK@cURiSFePntUg*9-+N?zAdPZ$sC7ItU{THsMko7chCW
zr|8`GMuvuaPANC@C#T0t{1&ekmyFO1X%k@2f46^svA1Oje{XIp?FL@I>T*U`z3Jk|
zh1S$g9i$26wo}sHx&hW#Bu>E}xH0$I{f6)^Kg2a<>i5?MMmT@_ADJ(Ix$EicQv+Yc
zydynWUhr<5$SWsg_PRkM=LciZTim~2EZD`wm=pY%cb~e$`g}Y%G>ScTmS<PUe|dX)
zP=M)sIJB8mjVj8?DIrczjc!`A3sc06gtaJucbt{JP#i&@{Z-N)lXdUymaJJC*6(N)
z>b}-{e9vm~D>L(eKlCPG2m1B%E{_nhi#4L2a^gH<142CVRm;DjvA<s{t>9jCP3&`a
zJkav5)WaX~8?mof+m)+X;uT_+^pp3e`kIU5eC#{bZDuNw&~|IV!f?-q-;~Edt--6!
zyriD~=v88C!K2or?^>Fz8#jj<%uHIWPoIu(-&D!Daj3>+prLwiPfL6%K`%rR7C66C
zFj>q<EvyK={6R2e_K0x6X*2iNhAr5q@WN-~!@+^*&2^I#m?yXL!>pF<g8Zf7RFa>Q
z{cvy7P4|?0A3SVvuGP;<PNkZp*o+1Lkn~eR?#s@;QtOGa(NT6$M!9EI_|rIob#&~$
z<DTM;%q{^g3bj$#BSRpG#SHtalSzq%n=d~Ji-gv3Y1{=QE~b7dM3`f31{!Nn&Am*u
zjerFX+ujUhM~6Pha?Dsuu>G0UXUju)cFqgs5|(uGdxr;qOBU+P9M=cle3ODPXc|wp
zTN(K&I<TxtzA+JKJIk^(uNFD<4b?cW_CKF{HJSEDu;#lPB2s*L@(uEGq+^H==c`y$
zTAFwGhx{rYX|44}G}HQ>zyqO2CiV6?jxM%*ygDJ;zocl0{GJhb1v~PVR6Du#b)Z;d
zo=vBTi$%p%JG{h6$CcbhrZDT0e7SZ`ziMV1A4<5ppmFDg@cMZy>>z<`^?~~F`o7Hf
z<ZC*W1>?4-x>AWST1NCGnVZ}0SKo;urlPXX*f_sl@2pXeGg4NVWPN*eRoN$q{eKIy
z+~jo+UQZo2x++*)g)zO7dg{S?s~8`6%wtR(ts;+0NNwX_q{5!ThRnU=biNBmVDrXe
zSt<PNr7>Ov5-RAEXJRBJ=E+%8Q*ArYaU=2kQ>dZqBB}Q)Kk~UffAa*gSSi1ztRp75
z+oaMv<`6+WddcA7?~hf(cvj?d#{@yhecD+K9}-{u`-?fn#?^`9g9$;+{AR?OHy_jp
zM$VM@a1O<xva`IaY6x5is;>}#n6t2{ar~6|#z?)wha-@=pn@Epa#lS4Q<%kHhG@*@
zX^zJ6fy+}cslBOC!vaz>HL?h1l~C7tN=Fn)4@{ouin{ux34<qP7aD3TL<5Um0y|>i
z<@eAA(N;z0ALdU#XmJ`!jP`~NoGn;8oM)^EJ)K{C5ni<~mv=BJ;zeJ2lF|Or=O11X
zu{m*8%C8cL1XT~}TSvmXhXefT@x2#Xz&uhw+(}=-A@f+7A_N*M0i6-f-QuFsc{{;2
z<`af_BQNP{1)DC;p<<SwQhs|OQv~*Y9+!(kGDHFkv(lc;3UBG|-gao;6It;iaI?M0
zk}!X=o!6MZ|I8)5Gxr7;m+QKJSiC)k5$txeI$q+`_VZZi#^oe)+*ni;j-HUbKJR}o
zt?bndqUCwee+^JiK&RBUc<aPwzR5h>2y{RdCtE!~BPrU+X4270siAzh`$@fvH&fM7
zV?_ZL-tlM=?1I_U`kuNGQ3tb@3byrii#E6?{c7twGVC&Y&v|msoqV<Dg-4zH?VU<p
z;KI^{q+pBpLz2h!@qx8}c%e3u8Qf}nmdQ?=XK;`;#-Q~Ks{h>RRA1Y^0p5()q&&`c
zF*AwhLR5MZac|znEJgR#2A@0AIGEKt&nv+5jq<`1cCiABEuf=&qumds{%jxS>Q*hh
zL6kRdZ+4=S4D+flnh2eCl8SA!AKtkss)^b$2J7Mf{ly;>dhhgFKRUf$J(Do6lO2Xt
zU(t!<6Npc1%W(w>`AezizBSdKEy5ef7hW&4cxyOYqe^`85KV2w9R|QI@*dW=UTKLS
z&ud%r@rbTR^61b#Ky59q^GR*o3^a+^$YF1Kp%69(1!7n%N-8YQU{_`o7w%YVwpE&(
z+k5#sO<%pSePbB?>F>&q7JqEpekHp?-Tt%mb?3KpFAgAXpIX?H{aEvG>F<vJ`+t1#
zKQI5E$o$Vi@PGU1u`6EoaYSpzfr(jcUWO-s`(rNbG&{E}z4mBs`@D0elbf;A9oHiv
z)$_4^<xl?H{@=Ce{~}L@$nUguP<MbYLjDGyx5cerJ6t<;<+rx84}SaC+GX9`YrnsE
z&T+UB6NGydTjA5N>S}bg*eoO5MwPrraI7vXM#rqNv68A$t2ypUZ20Jc3yS-t{Gbk!
z*odH4-gQ!?LwYZ)hJT&<u-UA-%W?+`uaZ|fMJzFGv(j9~415~CVm>uYu@d+^O1=iQ
zH*K10`k3<PQ_6mDy_;ON&z6s^IpHU&@oCTG^biCfIraC*rJ$iFw<3^<u0Q9<dN4;f
zecsElFg)h)#RHH7?GebPTh|udHb=LKL0l$wP|IKEUW9cY5VUeX)YC$woBxH>(v{b|
zMBPi@zL>w3`~jQtDr5wmJss!KUEjpmeq8!17z16c_q^*SW20lc8nMA$af=*eLxAh%
zMFQWxwo<53V!gSa6K<`gQMPbz-*FAc^XN<5wcekB>Vv|mw6x=Tg8H1$sDoQ8LwArU
zi$@olC3v3WBg>9gg#mqzVzWrHBFMjxq|lXY^b>6-hcFcye61?HC*y7@REth8*bn>;
zjv!_E<M_F)(3i^}MpnLhhh%?o)UfyB^?!<n|NTYzRLPqN`0sI_U-qSA0&8uLjHsCm
zk?&LBbe-#V4=?_@mUa4{GcD@%P#f=PV(VF*TkZGCnwr)E0~1GfA<mvTID>q7SIcbt
zDrAD|bo}wutD1m6Ci2yHM(;&bBI{e)S0KK1yJ~w6eciqgNuYfl|NO5V^P+j;KHp0W
z3CZ5*&=Hae$8TX2tD*JH$C1WpB5&qfH_<n<rh>;MFQJKNn~6TZSbck>sT3pEPWJJ*
zyRhY&?b}83u*t?~prX*Wch6|jDog!c+v`q_0%vaWefk;+#+M!$H?b0n9Y4GlbPUT=
zw#5e(W^LS|sWH5J`YA&}`9I;^1qILtn->qa-Y;)^<Ti9ZXXKCM@wb`*GR)5v-rm&t
zeq7P+OY0Va37e3g$o2511R~Rx-`m*75s;IT|9-FUVIyAJ_#PScs{UJ*&e9<F*x7it
zlyDT1WpexC>OYU_z6Y0}xWs0~7!!R{Ey2)&b5Qkh;FWsIDsJn&&G}l+zTmpY)-^1~
zA9LCju2$1!XseOlt9J*t%>5Ud-`!CP2aWGw;5NA)@#to2;qZeQ<mBeoz834BQun`e
zye>bmuWdtr?Awop?O*(o%>Z++;-}^eN#o8hW0?DdzVFNXcOX5ldl}Y(@a?=Te?&uD
zH(L4ldRFG43!4TfEBCcuA#}y1jCOtZ7xuhoaHMo#-^98@q1oH79)(`juK!yntEVB^
zvbVmaD=g`;XE;l^ADc5}vBu0gJ&<W)rK2T$S~4#!wW;JkBs{-<5{JE}8e+g7T2lVC
zQBeD3@=yQyy3vnkY445@A5&*N8pPa&S7RxUSaXj1>*O=bi?{tMn|i5Q+G$tV1M`Py
zhF=d9xe-Gxn2)PvS2;E~jYrRa_Y7d<hnbv+JWN--v}js6sFR{lo*1WrV=0CKNnHVz
z2^&(?C+nGa?&+O$Fe=fBgSHpGa*v&iO@_JGBo12a5j{H6j&wJ{1z#HeDc1TwL;B)N
zE!*wY<RIJR*U}-qCxk~tD3U{_n=@bf&V2IBuD^R;qSF_>{V}@p(>J<`9HjLYPpzV)
z9I+ITWc^CjePm#?^Ou)Rx|b90wxkSkd)Jx<i_qiyD`!U^R=szB#?M-8R-A1s`b*hG
z(AvrdtiD^7k~<f%oI8`HgPth<{(B59gMD7#T^l9b{B|G>xki7_W8n04yq28JEXQ@s
zI`JXJQ_`O>shi*2^0L8*^)fDLr;leKco#Eaciw0nt77Z>sp1Y4b;_Eb96GqXTWLK?
zZZjWhp$(&xY-+Y`0HMR)m`j_JnDb}!Wm1^La<MYj^JJskGst8XL2$Z<o9b(<zWH(H
z!64^5&%x<?H|GBvl>a04U-tF1m4j8e-@V=D{{H{5BHaJK0jtP_dNjiEz|FC}(#@Ek
zU&G&?d@^O@mzbQoKR~wh;*tcd=jCUL2|a=QHB`m&;j!J8KdWfbw=(Bz9gR|wUQ38m
z%VswwX>+4vQHKmoO?6xb>MWM8j69KwZQiUR5xvQI<mWA3<2+2&0)4st7nyv^{F$O2
zKS<8{>Z@O$3SnaM^mCR_h|;K<6Tqc5#<UY`0|wf#>&Mszx$O+h``pg~dl<AkU%LSm
z#<28!O*QxTM+T2>>@wU<DCkE^u#LXqWyc!F2q#=sVx2VxA?y6WC8-Cp=Sklnc`x(5
za6}l7Z_CTi&yD81*RngrL-?3aw8V*cfx<v3m@w?^?d4Ie38ABWbcg8VNIjA2%?)6m
z5a5&|G~)z|f;M1a*#^exGy}u*x7F_%F69~rXFN*S_KEf>^9R7*O&Im8u6)bDe6P>V
zn_B$ACB*-FGqkuW;)}q)YcyjwPx{+No7RnvTChkbekC&{GTD1=ZEXp^&F5&G^k<ok
z)??(&BEl#ir?l!PD%E5#HKdwMMdgS0=2-urZh+Tf^QCAJ!1F|}*A@T~KxY6zl>mZ5
z!F%OUcyF&1>G1w}?UW=#FFi*SbJDk(uWxmA`AEUqRNc+Sp1qc%zsleZqnE}I)_3nW
z95Z%vJ?dm877(9{qc<1xPX)I`gTE+woP45E=%t9B>B;K__uaENo`{X(RWSfMJ#|oq
zP?I67I@J1A5LPl3H5uFmj)WuIot&K7VGxO4Y5Hx)co(Z%uXGcaO=a=HWIXss_~BZb
zE=i(>3!s^yFy&kv<kC<*)Y!QcGK(=!vYCA+fLSlEIx4_#i%<id{|g)aFf|GI_At8~
z@B^G+R-fPA$wl|<?g6VRI-@~)76WeAYT1~G=HA}1Sxm&!-;xcOS1;`=x4seij8IkY
zs=aSlkZR<!@O~@p>tzGsPkJ8o{KaPhtV*5<U`i3OD(Ms!8jY^xJT=K!mG?QkO<(%(
zj%GV)J4LfymgMUuk|adU<s!WS+f}H+*?ZW>2Odo7U)j2r5?(=1EI&9(u2AvK0~h3U
z1KWGC{MDSw=vpdN_Uwm=w^xO@l{d<GvvC#&+>QIU#0IjO7<iV*q?dtX0T?X6V4*AN
z8MI@sopZ~P7ASj63Bs!v0d>}}g2O#5EG(i53W_mCXWHMswfwrWmfe}NCo=BX<9m;5
z%iHhDqB`@mwgzQu)2A1_WxWmqgJWFVJnJdkC5>EHx?U}2b18YK&SFJbd%dIkly4hG
zv>aTY1YYRGjXX~)1#w0+&lZ_`dZA#R>PQQVgC77kV;-P@Pnxm;9J`WI1yD-C9evh>
z4G!}2HhBg8%K*ZdOF!5thBs_|b947e#0`6mLs=d4Jz+k-e)J<9v+y5VE9?AECvxuo
zx`iLWxC}V^=sUliNQ^r*IYV6T%cbUGX3x4?JGTK7wK?cb^?(-i4wFs}5M!1_fFN4g
zjbqKIqYz$Fh{ZWQZEreJWF8=Ua8(HNQoy=Tug&u)^92B-3Sg2@P^6cW%nJ@jsaQDF
z;P)(ixN6BUyZpGGTlg5PpBQ$=L~7*}xt0&VVY!~qnLa%UADHs6ZoHdoc?o{2jhdHt
zBPZh2_eP(Lf}!2SsTuTyFgpjmtSUVeIl)IAG7DrS%c4I=U{&=1Z$=z{6_Agf5E8*D
z6hu?NX^eQvs054{tkhZyi?cH&e&zmGE}k{xS{GhMS^rhIc*W2A%Dau-Yx;AHq~Ur4
z1=Qfo%{*hrQ1|;m>RK_=nDZ??69H|%cDCFq5xq4c>cbqhH0epRRk>N$Y3QU=zAbq{
zoQ8E3M`*uSjEY0dfjgjabA-oijQyzNbG7)YFiKUQGFOrcl@U$Y*DWked7}*0ClLa+
zj3Ue7?Tz%jxM!ih3b%{yjUQZd_k?;lr<%3OJgnUtSm(LclZj@#E$6S*uQ&8^{6^z2
z-X%;lj;U4Pcc|8~S{jSO3Hdo_aXeQVx5yips!LGna`-v|sk#|dkJ%A%L-@o^;N2oW
zjzSFJP*{}<;DL>stfZ6{6dM_&yo{7#?=hzuPX<hFr{44ofE{;(mRlR2_Y2iK^<X-_
zE91f67RT-+o#8#0w_b4V)FqX#|1y}ny#6+BQBERWUPvyvw&3f-&S*(1;o|%X1UNx7
zvvGFztTUnqDit?)M$>@ftQj?hhdNUAFto_VLn1_gg<;<2vOdI^ZKK(E1vo7tJd1*(
zRWhoIjqEZCie;v^SN^0P`+dgS<s$Mo@5L<pWr8yDjPHK??yk<jXx-Tmnn2s{;HW<n
zYF&5i$0t1n3mZE~&w*&H6f`R-A(r5bXSSR(|D=Ji3antySUv9)5|^zM@C^lc_@Nwu
zjEvQd;M!2541Nwq{sCzLA#RGy`yn1ln5#l{Bt&)MIBh~w%90g1cDzcQ&wNopJQF@B
z)Lpt-OL5CkK3ywQhKd$nNOA^6SAT9&_&xKxZ&X&|B#g7CEsSq>o_n^L+Es%qD2i2X
zo0nGOw1z)5KDU59_sq+Q8&o)<ypSPq6j<PC(c2V>eWxkk9*;Z;Pi{t>t?Ps`Q5$`_
zk|2>WCu7o%mQZZiH?q9CRv2aar;_NIeJLzY>gv?_cgWyBO@EFzGRgO5Y&1F@?x`GO
zm!y`Ry{lSDcopE5VSX}i4a;cdvZhgaU}}(-n->IBD{|SG=@(%>XwjT7h*5)9@}oJZ
zKt6HCdzMGx2w+H6E2KrwZ#suf(gnHbr?4jQihF8Qkd)3dK$}pAAgcMOqf~r!+u;%i
zcnk5|@0SY-D=N@U-p4Yfxw?}V=-E?uHwH>5MC-NABv`ENQ_(|_%$;fK&$Q7YV2bY?
z4Bx~hGf9iygsH_dHv;6f%YoFPYFrZUdMY!1l2S=+Tn>LOh8;pv7Fnk30^SLpC~*>V
zoYuKt!wea0t#9r~M#@le7E2=lV3#q^;P%6%4()ODse+94H0EEEzC8c-gppuK_VgOd
zzG71H1dAsYWN!D%xpPY+&UePltLo>BI|uR@XUlqf9$dpFP1Lgj#qlutCc<EO+d#R<
z2)-WtP0!Fg2VlnI+r=2pU04WTO#y+UCJc+!pLcY0Q?T8gwYe}=Il^XXtVOL?qH5LA
zjZ3G~#wDtcC6Wi3r&IzfNDk)@b?O%0%YpVL{mIUY`QafKZg6Q4?d)~4t|AcW8asqw
zQms_9<$bwc<My9~2nLiJ<2!Uh0*Gqo<xq$5?MPI#F*VPx!hC0#Z(~3)5VDLJEC#w}
zsU2qf&Nm^~YjQF)6p;dp$wtfRJi}sxDgn%)J;ouXPq$A$YSG})>!r4)9SwBCQ_okw
z#m0ZjGSL>dzFcW76^XO^PShncMac}CShC}Ov2b|NJ_wnrIk34Ce}41dUsOT4AuiiK
zz_!Oi%SDLCwoBz^afM>%;ATnCLMr(=YSF$Wge#OusyX+Z^^HPY25L!fQbE0C8%FgJ
z*>5L-{x)drW~@vu(cO^sIk+62a7>%Dr5uA&(8IH*$7dR*ht?L{y1n25UpLeVFL1$5
zb{gG5Bs1dyzee|ye}!H2u_q(q0~IFq!Tjn4QA@%a*ggy=q6ObaQHTggG!q&=A+}~s
z_d6-n5EfoCxV#TevVB+3REx508h+Ow*sjprB8*UG?Iv+la2OW7Ykd$fDA8+gbXC9Y
z3CTWg`j`YShC6$;`)TcW^o;$8(P*uWlq7ZQ=4)YzhY^Uzvit6-o;jR`==tPa($#w_
zBmLxM{008?CSlT0J`iy^pe34-gq|6-I5onaA;C9fC=wKmzQ=cMH*V#jpy6orv>YBQ
zLw!v40ocWf1#F_HS9{F!^yR+mc8N`&%reoW;2V^wySBZ?2lof2C+xt;Rz5SvjU#ra
zTzJ`96=B9~|Em-JYA+jQSD~@L-Z@o*93rRUHk-2|P=owPsyA3#vt1nLow^u@1~^pj
zm6;%MkOpgptk9STj42#|X9r+#wOLtIcz;%Qs6GW)0H_+=5jN12zQ7>J;D{g#m0pR1
zlgr)nL(k_|{*{Ly_MCY5`s1CBrn%eWA3kF4P2x$~Bt-xv!~1mE*@9)!hn4#`n-E7N
zMgEuENSNP(=^`8RtZZ+i>?XeQxQ3V?!qGv<H}~o$)hG~ZK1m!~p$1nv${iEHe2nNk
zfY1-`?XYSmJ^$xnH_trmB|0pSf==gsf-?CX77z;$RJRFWoB?3K0~r%Rd$YBXu;NI4
zxL%{pVoUHzuZ|`vVSam92VxyukB6+dZ5h@o>@WOmyXqtJJK}iku#+&wp2n$K)H(;3
z^L2NSr&!Dox4{@oNOUSncOzi)y^XnYt`y)g>_ymxVgvi~9&|@0&M_FuwB35oU@NJy
z{-r>|ERPv0^MFVar|6T8eJ=5A5&2GEw8HS8Yn%@*zslT4f+ib4UpC>UYGmVd*R#Wa
zHCT=u)c&Q=s`1aGe`g$ZI-_?kj#zYDyU@%udSxGEzY?WzHY%xJu=O4EiLL@>yoNhL
zrIf+pFgU=844{e*V3M1nu%p*uUE-V@Y1fz9oyJ?TFw0#arWlFlfU0y_zwQ>TUqvr}
zAM}{{nZOR`bDlmhq?*!J9{bF|+sD~vBKrJn`w4`FR&LUHUh910076nklCP6+o|}o=
zx9<ZpfqjGP^-$e0rAjj7oY58rH7p=9y;&>Huy}LD!m=k!oQmp$iHupr#%zPLRO`s8
zF}9rQz2sj}%#0(*h)PE!)JFzWL0Cvp2xPkp#H;sh@s_kPf4*q*CMQ{W<{^=bG2TOu
zAf!L@9d-*q9JwtwitK({c9uO7-uNp~RC*q9zv0a<2Vr9`qxLD;lcNu10{5wZ9@VNN
zLl1l(Ft|pF>>$gd0Yd0-w*WVeiUuj^1j_&q02N?-uH!|-qU$q0QJ1d-ivyh$(JN-u
ze3MCof(4#91{DaB1;hcVqVGAnRL!X_|3d>{i1r}@jkXI={2ar7{!roNZ~*FlQ-TZa
zxvQJc&+I$JcQW6{K=jOmF~Rq9lWoqK9=CxQv^Hlpi{C~^M=9E!z<z)W9C2K0LHYtu
zRFR(ks(t~aNQN;8b{gD%T~Cnh##rfZkuVE2+2QZf{f4wX${S_nexviee?^4by*=js
z;KAeCk$SdkqO%~(E#F8enLaz2IXG2gE6^yNE!U#G)k#`Cc{16yAXyNDMu!Hlo}RU~
zZUKfRwYc<omYiu_R|S5eT02ZFfX%?BPZ}5;J|;Z{7YF&6&kx&}`*_KNdI}0Q6+y@e
z8~|{oD3s$Y$W*DEQMD7~2a`TNPFY{JevLWu^NroNS&wh+H@$#9c_GL5e57-h*6n8?
zOWw5Vxp1CvA4Khd_gaVEtUmAVZc9FX+-S2mGqaGIH_?`JeUlv6a(!np4;8(<C@=wL
zdh$&Gah?K&LV>Wl6YB{Roe&TiCj>AnkoS;k{Rp*oBuIU5n3W7+0pjIn>O<<f;19b0
zg|!;7G=dN8oXNKgSn0|$aaYYcX9Uk5N7g*~>jR%kel<bHkraWK3P~Ntc+bk-r&IN<
z@r{lgZ;Q9I`P%>De8e@}*0LX;Sin+pyytq^)VSq-xF<}_35fuwK}F^96vu5~5daot
z*GzZ-Tg2fU&<8lYs#yR42q0TSAW--^WFqB#*e*ZJ88a7}41W6Y)$rR|zi5Zw^LIVc
z$RK=u*Lv)#+po8u-ElkUbT>yu1?)$>C=QAzDuQ|hrhqu#^fSCK(gD#T$kUw9Tnzwh
zOcfaV@V8E%kA}}*`@g*FqN1WQY73sYqlT{~5vBe=;cYzJ8pg+=4C&*CwKx2lA_kz0
zr`3aaXpASP<ZNp5`*-H)&zZBv6vM=oCoJQ6N+=1p21ly2Bh_{w>9jLWP9#Z0$abmg
zzrPp`F8XvU>)&78w>0HieQ>v2znWO_S7Q-&@K|RNww<9%=p1s+yN`vqJ!4r_8DTxO
zwEp7H@x<i*XeFv2gd-kh4#4sl|0Oay2ulo>5u>W0pdf>tNK@WyN2!8t^Us@Zd;1nh
zvocjZ_VK>Ap~Ktf${tKxm>cC|2mL&QaZY{~uQRb3D$}PML_JFma&%z$Y@nMic4#oK
zv7!8`=ZW+VX?)5mev9YZ)*O#mwBagnn@sTyZy|1(b60xEEJzR>*dGL5=CC{p3kxPh
za2Xi|DQfffLF!19J+TDl)eAD^%c%}8lxi<ZRbvuF{#heK;CS<?F7QV0abCz34XZiT
z(sTPGjWeM}N1xpBM_Qu0Gv9dkrVnWOkEEu34hxRZUbV;a46)<GJHgWQRIoTJH!Dv0
zF@<`_wElX>2WUr+Pqk8DHO`NdcVl_VZj;Y6ZNf&QkG)mE4xIp$;Iou@hAB&g=8(rg
zO8FcxG=R$}06rsjLBUkR$PZS`ize*$I*tr+j$3=<NY%B-AFnh`+dY^$T}Pr@9CP6k
z-o4Qx83%Y+PU;fE+<wV$ayHj3T5MjoSVd!qN@+ed7bn)z9I@Zx{HDLdyeqXa3$Bkh
z7snF<*%-?5_-DMwm}F$6cwv*mfq7w+`$It-1x!+$XqSM=33ehljm%Upx$1<Gk<syH
zxAnSeuJ8{N-el`OZrE&A27ky?v*gRrJ(f_2i7s22eWYCb5@w9c-4*5>K79UudvDJ_
zUG?-Ec{#W8vjYPq-R&(gIz}B&W({VTTR7G-SZ9JZ5d?wvP`|}1qLt!76mD5$KE|^L
zv6?~j3`_=dGYnkz&33Lf*Gy(&`rXhVgZwAb|B-)TS_CRRbaP8Zqj6|Jp&jP7750E{
zHI-$Iu)l%C`|p*SSb}xx2`kGk8e1CoMFRcn`Y;Q`N4PQu+1=5CfFc8tNgW@7<v_M6
zn=xt1s-#r1Lz|si1N<>X=oy>H7gzjRoYrA+wI>6F@IDI&EH;1w;$vuo1Y7XVoJ#SJ
zye)#A{Tb<(=y|GYMI=sFiCiq#jUT-$iT#-&TMJn@TGx9!sWVUB+T&r!n!<X<zO%o5
zJwFeB=AyI*tez%sXZGqchw0v_1kuv?rl02KSWDE9pD<dUst)qmqQJ*W?&TqilNM6J
z1%@%oMbw%cx2!fU082K!$ew{HL=Y4So=Iz94_b#u<K;4hUFVW^qvYCNqd(<0?QciU
zll)D@yMFzE!%jo3zutAUF`wbjAfLFF(K%RlGRwqLk9S`j@6DwKYg7pZ`^#@d117?Z
zO1wP{W7f@;tB(QADVfZB0C<?9MMi)T7AOEfo)2*>I^)qPJVbhZZ7C?dfZ9lk^cSjl
zk-T7WbD&v*CToA2o}S;@EUC0i_}1r6zib*xAq(U|iGF)VzbuKM*R8b|x2Ml8I+yDZ
zbyBT+`&i)v7DMx+?259mPE2m)D{LYO7B^+;-t**97nQuGAfB2E<j0+;p(M>O%yZIb
zkEb~<YFM7w^wS`O2zpTqs{yQKUCCAis|hF$9dU@!msH%B^?s$!;G7KI^?0pET@r2Z
z>e-Z4!@7^Uym6{F&!kl?UaltrZ^tc(I`{xYl`Hg!QPAyjk&ShIwkD%<Qb7*%C(4ov
z2^W0^A+_@@FsKYRTo1DKco3VJCqq1`;4$z#g9HGb5PA9Xl#h+YgIbFf1Oz@_pM{a9
zLM$L~Igx^6V9-XP#YSkO<aWp6pQ%%JVa|Vt+b(Nd{8xr?LHo%+S*^2%md|9>21w6p
z-#b0gmwHwk@}K=lHgi}EV_l>gzH5+%`t(OBa!YXxT25~C;IJ@<9TD~-M2H3|g($|#
z<Y&T0Mk__LN)*^oKx9&|0Q$==zH1usk~w=jW>SnfJjQ<r8~6`ldwRhF;SvkoHLztd
zyjl0RuO8MKAL!{g-HX)9Grp5;;(m@ch`f-id(U0<Kuf@WNVcW>F>_US;|s^i?z+ya
zXl|1oyPEi7?<KtbCq8GSKAJef2UGSVTihT9CPRpu2>Q^_A1H&s6>y}N45GW{aH96c
zEh-9yemOaX@<KpSq55(LP*I$OW5yyOx)KKuV)9O~`}uUVUkr)fy_OPx9CL6yP6u4;
zIb254O*B5wb0&6M-IWAX{qn=)=uu;=o2zr!9fQe%O=Z-1-lp`BuRNCW#+XgroKLzP
zToSZ>f(}rK@{FXxaW;4p%x26sfepXYGw7(C3~HWQr?>T4xF^I4_Q84;261rO0O6W~
zK^xL(c2nYY$+c>mAM9}LYbPTPo%!-|Me&*Qs0%S}r$WZdPl9x9cT98oY^194ET<-!
zGq*{dM*+lr$$?ZAE<UNXR#O&DD)kzyhkhqvD)k#9or4x90OB?$xE?gmU+2+yNyd{F
zuh%T>KT)Y!SqhnABUz(*b7NhmSeUQiQ~<nUkPN11;qbKGk1c1~V@|e{GK;PqeDh<?
zk-q_vm(>O9r@x?0x{v14@}>6>lVQ)?nFwZ)A;F>IBFOk@hpPI?(omO7y1*M|A6&t#
z=y=ap@2G0g%wOsl;j4o%n~Efwaw=$5GdXE#`7%4-rV)sXh10Yz|F=>OjaTA_2yzTW
z_UV#Iy6?|#N@{hHDj0YQQd3fU+sHPpyf*vN&_KQM?7&sRh}lqQX4?2Kd!ME0x!t`<
zFE?^ZNz<?5;r*3yHF*EdD$aN<!h0!D4&B?lz0@Drf%3#PixoHlST~6B>bGj|b?Su!
zwNxrq9ZAg!)n?2Z0g-+Jj7b5_;4@RJ!(Fb@XA=^HyT1B!rp}k1cO{`1VF}Ue(GLEZ
zh8*w8)~OC?dHQ(k^t4E(ho{*MHkN@dRQ%k=-O{tEr=nLJ56XMAt=qWs8s_7hipY5e
zFYX&=eiF8HXk0+OE>|C)V>9d**tmH%BXQC|E#%>&Zh-OT!CEw)ZI1;$IdbGk4~V0c
zJetBbZ0-Ydt&me!e$EW_EwsVcMkZxz*Ce^kw(R<o{f(6k##VOH&TN?9KlNZbUo%TD
zueD4Ez~0umB;9w!_&wP%#CxcwKd*X$LF48TrPgNC3yDIzw{x|yo_yn4*b+xTL}IA<
z#BugGN)2Qtlv)Q$4(STjIjoT(prW5T3IRNb)|K^63JWzU5+TB}aiG*>R7Jy_NS{R?
zS*VF!^r)kEMC86`BD#dZZ8WP-kti1vJG9Ep5H7}XKZKoyTUY2J393c@YLil9!%M;)
z5pjNe8=B3+>SPBG)H;UyR2)JKLHFt=iUX1A%1zr5Ywq^wBn1s@Cjk^}27R*Fwr2jF
z7HYe<@FbjC2s>N%2}<Q}NPN^b16e>5Eh7C@`fTbd*){uF=xJ3{)__;&=2em;C%`pe
z5Bp2~gf0*Jf@_-Ddgk^u04Z$5IAXLEqIdm%)q;O3d=@u)JFbF`QVZam`nBhYv<sk;
zI@Ay!nfXMinhVaDA;Qx7f0(I9cS2S0`cjU@BnGWy#`?32Qf5!EjA;>L4l!rtZ87X5
z(j8yNsK4evsixwOnHj-~o*JLWUf%Y_?meB)mj3=R|5DGs#vUhu7oy>&HSL1i1+l}v
zG0{eKK!Sbnt8sgxdKHg=zKr+3r*9?{f+&P&PDx>US295`nTbKDKz%?qk)q(Ac{=q&
zUQpW~Y1fwnL5iWGXSzjb;|kz2P)`<6Lg~vYDtfF#W-zGpxgc%UZJ_ZETg%?lWZ}mS
zE|22jFr7xR(j-8%{){{SwDqOcKvI<b6fn!Lav6LT@d}sd6T?3_-g4ghW53Nhr0!Js
zOfWNUo=+rBhxtS=cd_ImvwlV3#u$rXJPGQbb9~H@qln<kCrSakIX;eJHU^fbSIx81
zDcA+@!KA}1CX}AX!Im>Z6VD>JLxnHmHFCY(d+Qq7Ef#mb89#cgE*joNUcHbSHu}S#
z0X#FOn@5hO+;Iwes{cm09cNZ(c8*Mh>pCQ1sG0>yvlUrcm<&3A1?K~$!IeQ#C{UL}
z>VYa_jOBacJcGQL7t~VmGDsY_9@Nux&pZ%Fm;$pA<8p{ZrMGvBb|{rhAUkOC$)}H?
zh8~8iy!7$zK&xqz#>^458b`fHv-R#CLTh6ijvO0yJ@v)ho4Pmkb*C_x>1hlmED%_6
z_C_tzwGOSw+vU-lANW3T%iR?%C=eSxu=IQdQ7QI9;=l+6unZVxV{A~EEqDVWE(Oar
zRH~@x&*YzRjRl@YicQTRvPGTUZ-VLeDN7LDeCtS|x$($Bwp)(>)SIJc7X}ymnkSWH
zkQXoA#Y8-x0+$j`%Z^Z0jVGoA2I4;pv$dN$-kpwAjVg+nUp;0h$a>jdfB=PA`DbCa
z9SL`GIzLr&QD*Lg>S4rv(~~EMj`Nxwl0dAhU@ZI>s!_;ZIRpZ6=w0uApK<WyceQ^%
z%+l9AK^n=}uj?F;t)c6?TwbtN>Jokeo4#ZPGPT(eGf<(udCVB(>inAEu9kR1gXQte
zOse2iNL$iOd{l5tT<d3<wg#k0B6z%x3xiNTL1Y9tJH4&lST+UA`ut-I28LY#2BlyU
zypq_04-EdZ+0rJwzu$H&yHk?nRb(ck9}KexqjBzLsCny$mFjWwtD@*ycl$-2@kLTc
z)kXh7oF{0M5^`6$s~Dx&TR>aj<K$|ieB#7Q`5ZV3wGOKQ0nCnsQ1kqZ?VtaF*!sh}
zg@IUrjpM*&3jnAa1Jw>#20MaIrwe*}`x5D^khS$ES&g^<y3P1ShtTj@vOjv$`liaU
z?|uQX`M$A%+od}OuHYM~d>;$|JP^Xc9}$S*03850KqB~{7z6(DpRE`Z68l*m2Gcko
zFdG<l8#oqPbX(ysX#4Bo-5Qhv9X~UH%8k(P{55VIJ95<@*X8Q(<y2N><eY0@X=!-~
z6|kP7KzIdCB-a5Z8En90+`q$1j_{O$I-^}793G>pIvU#0`rUvP$hR`9cLl}h(C@X4
zj*nQJ<upiI{(62T1MaAO(d<IjkqXTmy{%{K2OhlHC-6(G*o_d$xK$TO{PGL_YTyUv
zT59mc`6tH(Nz(4+AbyYyh_5K`l>i~J5>WP#Wma;~iwp;O6=c@W_<TSFpg|@|2VwEE
zXl!JXK?BLGp<RNdh~X&?OPUN=+3RD@Mc=}$R&<A>HS_m}!A*6tO^#Ns9GQC)dI3|r
zzxGJluW5dX<>j4)Zr-<!xf+@>w%$z$t%*}%QTt>dCyMVy4YxEw@r_Y_Vt9X?d<-><
zg7w6o008p<9-S=$<!U3ND{IJdY%B^9i)~W#fq8+$yfXDUJPzOv21<ZzG>D8wMh2mx
z_eEdsD>NoI{yI6V@SttjJv;7#2{f)f+Q=@0<!k(+i*|}1^=BleF$3Hi%a33$!JjPm
zretP1P_+{BSJ5+}L4us@-0V1zW8W_OsE@W_Hi0UT^+C~E433F02N(bt5KRcN0b*F(
ziYMZ8)O?6D7D4JVFHjhkf%l$BFL*n+NwBx_gUaoLJnYTN-I<r6alZyTrF%o0;?C@e
zwcA%>{hVqJjk#pld}H}M*Y;M~2Upljy-(g76L!OIAFRh?W_sJZ1gPrFKyI?v6G<D6
zXTXWC4({g3$H*?YN*<^$jz69fl+S}k@)_=#UGi=t6V^;KJPE|hHvnZ`0E+97QS_Dg
zQ7`$#Pq`>^z?YLe-0rHp#=oY9f4VnfX0952rH3%L`gh?k8W+g?vU@ixsf-IfrH`e(
zvbxCG`S*#ob_JV=ILp_o%FeFZK$#WUuMG_XGk^sE3}DJm!hpz;u)Q+MPf~i<fM^6v
zQI?e~=Dh`4;v`@K!o1yNLP7CbA46U#+h-+N3JrKRKfBMBuy@?{UY=&&H-t=NV!mNB
zsp5PM=HjlU_@XP1v4a~U<j#EQ_m@0`PbYK)$PV$1Dq>bg<(05l*hh!uC}dAZi^d8P
z<ReL2Zk+rn6;x%@c^MR<CxX=B)h_A5BV^Tq?B&e8rZ!h!v1Y0Rz*qMoAhIBhNznN~
zuV|2f%czX{m`T-*T!S~A&wKc!qK+X<VByk~i*9cRA6KT{a@>PI)k%i8jzwAFvCd<a
z=hLSf%6fo*2Fd$OO(x><i|FjxQ3~kh+=cq+z#&CbxZ~7BP5GGEDG)rH7Ag32fC8|2
zSbAE2(tnWp2_LruG%>gn0@xTiwio4L<)aE;vjB;Dqt{825U|0Bsz~VF@SjT1IFvnV
zJ(kJp&ebTbD!v`;0h`US{95DaF#&SZ{KvwhO?T$1Gn@79-50%eugUV(;0Vb(9{Yhg
zUNyqSZ!m@g@W$7PqUx||B_G;<jt@h^<?y!|8Q1(|8y|5y#IyVLs5z=C(NCdNRJ%tZ
zEY=e|&6jBjC3t1<RKgbWYmOyOJCMddB{;!Z#l0AN2)}oy0s14$*W$>ZsSTpZfLP|@
z5BVCdo(JW9H8OG#(&b5E%i&9X19h`XZ(O%5#`1Kukvs;58fCsym<2M)QZ<AZIaY23
zBAQG$CPX6XDwR(U*XKk5rknB0o^iCwR}>Mb$2x*dWgKF3`&t$%tv-C&VRR;xdPVa#
z^zugdqa^Y!C<0+IOtkM=af(L8p6KI2vR!CzhA#~c6kW+zyAg1~YDZN3YFs3zkjO=k
zdAQ-mRb)<-4pLsfzT|nGVjjTW(TJ`yA7gJ?uP6nGCtwzanr}=I;eq3n&RzS>Vq$a-
zsOnG4tWZIj$<}y=5xR<v6TqMd2+FEg25FaCIK=$0rM1hLI#XMamKPeUWo6s<*6XV9
z8%J{Q)i+6ezH+ZcsP;v~Ik|14XlNkcUZ6qChdH~fV;MuWtRUfU&T_t^!Or4!_VP01
zoavX#V?mxFfG|d58CVo35YTNA8@s?Z0oC9CB6WwB@0R5WHV?o);lmcNoD3R(6(T&<
zNf6!knD*Y}U?ayrUuxzK97PwznP#4|Xn>e~OxCX~PePhi4vk&X<)&#xHQu6gzazW3
zd11<OwAOV;*;C{8VCo+aHWA{C)dt0%3;+#VOWP+EUa|5Zz;bnhtt2B=5GYR-h#ot^
zGxYld8aU1&RUK_$(m+)0%f|qd#|tQIrJB@Jb;5ubTEhJh=1V-BZ(v~A>!Z`v|1S@d
zJ(k+`J!Ox*{){X$F^FV!!O3UgmL57sZ^SYp^9PO|&(j8*_+O1l;vVvoHzd75kB6bk
zadwQ9=Tt~E0XE;>f+2PYg@I#~ZB_>Fm4_Zu>Ze&iPd(?>A?c#nF$@i)Vx?Xn8SKf8
zn_rl;o60CQEIDQ|4m!d2_@NWnP-@;E4`O=sSe!EBibK2;6(>39PUGXcW#<NUl#q!3
z2c@Bi-+~m5PK!xH+-i3lZYEX^)s>!~`y`fcuX=8X6)is8P4AWn&iv<AL&CX=+h_HS
zj)(M<NNX(-9wGfp{`HFfK93M^St@X0BT?cjS`~=W@2738)${GNmTC@k)WzRF(s^I}
z*VLvB(TOl~j405G^E}GEX<v4Pc%<|f@3OK%XRM(~LrX`KmHnxJezG#+La70OZ*$JC
znw{SW;iWUK0-Lt^%dqZB3P8c}Oc>~@^n#4LVOOrqARu`eS-DnD9jezmI5IfAy%%Y7
z0>EYzWE6Z(EOblO90Cr~KEjdOlC^@mV&j9Qp&RnS{^=)fyX#W7@S)CxR~ZfEb>#M{
zu$y;WgAN+#`WN%77vl*NLu7)7PYWi4k6EcG$*T<F<8X{wrI;R1l*3fLq)3>mnluK|
z-unjVG`BW`{+FjKfdB9`1DHx@u#{HqeR{&;k8y7jw04!eNDOPNn6Y&?ywiuazfx=S
zW9K|Cyk2F$Xxo5q)Tx%wq~({S4nDY5UM5=S9dva!bJ>_@GNZ3Hns)XJb_vU-0>w5f
zp&(LU|AtCe2bG!{AN(N$xriqf$IUVGkWi_{96N%|K%Z#)<3?VPjU|ZHpw)t9+Ln#k
zB!L1SQazAWl@6w6C}^}Dz1WVP?52y??YLqh(EPhxVp{muH2<IbzzvVx(M9J8;$xop
zVcMfI+p&|4Q=KDC4OXLGrEWjK))xF5In~q7!L?Nj+B}qXRlPgd>OZ$A2f+Xp8Z+2O
z(5eUnaKa$3F^b3-5X~Ag*ch?qyLSzUX1oSzS&9toM@uy-8EgZiDz?%BF1LU?$t4gK
zc=D0?8mlP}%fwgMJ1-J2=P_*N6>pn@7QKe`E5%_w$IVj?O&hZ_+>FL-?DyrAI8n<&
zS7n0WxdDsP1FqwYPx%?@ifD`2<xtN@oz#3I`xRK`XF*-`jwTW%0R=Qn0%=0_+7iDh
z)U=IXA*pTm3=a_%?TIcy-6ov$sfL*L<Rhz9SFf&#wgA#74%oxiSj-%{q;TTYgiAnG
z%dF{|9~y#Blp9ODZa*0(x%u}FVnVrtP_v8P-8D9-KIve+DT_i3U(oVdx8$%XXlkq?
zCntJlv<jrc6BJIt=bQh@g7g@_+6evUfTSMRQAsZlV8Dw|d+$0#%~MUL3Ykcttt*Lv
zSQ<0T+LCCw&+*}^*(}*nfhA=;8yW6diPNS?V%?d+AM>}<dm(zod^xycrR>C9<LJR4
z-u_i3BiO834-&Itc4n6(r1CLX_<Rx`aJF$Ko|jY&f^{5WyGLOjR0DRIS^*Iy2K1?Z
zd>gAgV=&_refo4wPz{YEAVmg(t%)oUjwl5Y0Or{R>9Y|OEG4wzmpPykVAW(HUHn22
zsNE(`_T8199L}V*Kx3g_Tg`bS^O5%=&AzN|bho=1-V(Hvrk*9n5zF)9Z9IFr`=n0*
zleT<jPQ0nfrk4z%f?9mL2re61#bR8|0A)82k3t)q&N98Zz5)ctnmfXGf_s8s2dc6l
z@31Kb#%w!!$w>Mt<eYkWXt{q#%vn@<C-buRP)w1{?C|i*cN_EThkXo?1LfV^IHxqP
z|3%zeN442~`@ZnDg|`9}DHJbWyjXB(ixZ?cBuH@*2oT(9OL2FC7I!DXp~c;UyL)km
z-n@J7d(R!?-gEZe_cz8p>mM>mo~%4;Jjt`>oS*M^N=w^{tWJ7`A}(GY9t)-){p_j^
z^lP}}B4rjQw4)^>HbC<nmw+SLU2OFFvd~2g|4ahO6FnwK*hJ(lmNnL5_`X{9^rA!k
zj_$ICttHHkF}W!H!I6)>@A2QSK4od6+k(LcIHGEArhbME2z^?u_=aJg>AR(D<RbbZ
zikJ%I$eb0uB2irQ%WVe-2S*2P^G!3Ofxg~GgZ;47lqv3Yqvn6`f&Qs7Oy#%M8ncHy
z{3qL?d8#?kzn2xlsixfMr19?oT~^XR@uLtXsK(&em=F0+Y~>O`DmKV47W&j7e|Vf)
zEKR~GaZ@>k?|IjHKY|Nb<e^lkZdeK)W9)IacD`HyEjXhKyGe524jfWaE_s+tSBHl;
zUdl=U1d3gF{!idp?|%V2UmgCpfak#f2k>0+y#`aH9#fxFs+#@uZ@}~4il4s6$Oo1F
zJMese`yaq_xxxPhJRkJOV=!=shFwf$w(y^56I%37T^3W5{--QWqJaUCF*Gu!T1?HO
z%%h5lPKgOghN)a4=wDwiP>r!Hz_)73RcPF|NcYMb{nqd?QS{qM3{Koo4UQ117U0BS
z&l488<cQGp@XHp?QDxH0guSePgS^%1kRDbTzWgm_$@6(J*Cqs$*$Mh*M9@Coq0k%{
z1X*NY(g=J)y5oF&ixw8JQ$D^e$>wOh8VbDZk9gT~+OTB#5i8#9WIU<{W<QJ<vAe^@
z3<i1zjcTds=xb@}ymFRN<3qo4M58hOuj1ER>7FeQ3=u*R7~WYFzR^vi51;!^YI#dH
zYe-o@K9s$h{QQgQOtS<yZe2j{iI(Ze#*{wSqGvaNSXn_yz8@G-2dkNIdN-yu0?T<5
z6EmFLohz9mu{^5>CIcVax$ZUsMdh<U4vw5w!;Cum#$YqqG^tb}#g$sOsAlB`Omz8M
z%cTW#wQFJSZB`~%zSZx}DLq=a7RKS{lzIZ@v;9bDkuA74Ij_D^TE?3Mjn*8H)*$KT
z>5#Usmc=&w*JKX=i=&^0&gDPgA|Aj-cl1ASy#L25|5NeaOXE$WtIYk=0Irh5e)fe6
z$7o}chVpz5NTjhk&prXiP-+Gw+^n86zM9u`H9kRACo?HQjdX7O<+y4gR7z2OB<syS
zn{V>uGLrC_rnfn|UxtY}%NFMgRKr>?+L4iQf44Y=>FqURc=yW)%X38qM_zntuMt|(
zJ8)RS9)Ezv5`VW~)>t@oCkq6X__42;Ubu)NI&O=cL~p8Aou>)vefp{dr05?px~f|v
zN|_=PM^=*CGO>4>nU8-YHqeTX2V9wIuMD|&&W#VXPFO?!W!qr+S~=w9LQA^3YPS_n
z$#a}02rDqU>Tlfxax(;;S;p<glS3WT@IfhuohA7fYMXIaa~58l{GO^cuCPdr+D3y;
zou!$z8S@fQ<Mb>`-ql^5*7!Nz0BDi%rhh(`5f#f$`;*@AV|+xG0hgL1jI>wimL%cV
zuY_<2E%gw$Bk#TJSp9A9f{d}zKzVPqb+ZYwyQYg-y5lX17hb1x{QwW#iFexFopC;_
z4I1eSbN<AvEkKjU0r>>B=eaO0PPGvFU|niKnFbv?K;;&{D=s=|xKp`Rw=1o<Youa5
zL6ldC$dy;6v7)IyXva^!qyko+gu%7naj=^M99Dr5)81@XACEbP-HMJf`<!oP(~@DG
zE9V~dSt60oRVg#bNhulwmn+I34*DW3tVMlAhDK6}#`RR7<Vz!w)PO#ixpbgz;45ib
zHt$`H->Zr0$362>S++wo<@J6ae2MJbplBqNzGg<}n%KF&J5qPDDRM+_{a^zUpw+#_
zoe|s$05(Sa8~|jS)U}e&dC>>hcJ8^wE6(*f)4^uMk^YCqB4T=A)iHK9rLnPmwScr6
z>v&Dq%@ePqL%y$DU(dzg6BKWelrexb+U6{0Hh0b?o+CJRuiqUlu#YEI$Lg6DiBw0+
zRQSGXThrqbtC{$v>Cvd~$J`(ks{;)Fw1jEKimC|3&vWytKF<ywCZ~Bf`w&Jzg~t22
zMRwi!{nEa)92Q_`9K5r^sU#yVit#UB9?#BN#TR(0!es$3sOF{Yao%=NUQ}0$%KagZ
zQwk}kDkJ*6Jc}qi6)Z+R+w<&tPCSqW^-;fgWE!373IlhSmhy1a(*0bYSN%3H!@c6J
z5t;}$AYI^vSbPk$2#{)6;xp<_<X*K4-=s8I0HSRVZ{;dYhv>KBwI-$y4C{BkasC(^
z?xE@0Nh&YO9y*!>hgb<qQ&3FvyPPM!(usdZC)ByGn~CRROMA<6baB6biG9a9j`dQf
zT{LKzNSteq%eOr+fEhboX|lP^n(44~0~DO2d$-@Jvuvzp|3Z+!!4D^tq6$pmNy@gv
z3^o_}u~$%85YCm`08~;}f{?~a;n~sJ-a!DbJVUJ|j+%u)qTVSBg4@=5w&&H1y++l2
ziy?>73m!<h7ZX%z1$r^rJY3T|CFQGT!H1OOYym1dLz?#c`2z*VsHws2g<b8g+Ry?%
zOg4-E7TZ$eOj7H_EO-8t^VjzNUgiAOi=p2m3HPLnrjd)LWOWx(P=qk7)2yjd53jUR
zqs=?&;4>EPaNuC`-r!!nZRf=={1263ZQah2h7PZTon^IaQ7#y%W<Za(A3f0WgOOSq
zl5k9bb+y;xO)q1ghq1+VFsy8)-mvsHgR8qZi6>RDJWn(K=mBdrf*ip)+KU#ja7{))
zKz69=uU;Waf`pkSeH1AP$uF`Qi#37+?N|%3&pu(B?%JlTP;}<&O~yhEMGX=mhq)$t
zICv>07yI<eum|oqbYEU$VN(Bhhz6haro=CV1K6Spj%4a!lBRuD>N)O-*KErTataXe
zj|DVPmx99c_CYE6L@UL3d_!gMfD*hbe{waOXy$lc<>_3inK%Q0Lbl>lx(k`&c@?!X
zDl(wc(ZK=F-KqMO)$nAl!I-XmaYOfM|9w?&zolvsveZ3T>msgRYK+>?^vMNdN8+nO
zo>cSAk(|JOOLHn4!G4s%@Qz{QZr)C<W1SL%vc-C0f(Uo&p`*_4cg-Iw9iR8I^Mr7e
zl`{-*ve9;{vSt8zSGxB&i~|2+IJI8xuZ$!T{fhL+HQEfX`!zOyh`u9p1LbDx*y@2z
zNNy~e0#(lxnb<7K*nVtxC3N?&PWQ9_5Pl^9tm}rJQpGIvwQEa7EzHTrbF@D`I$6u^
zunb_X`WeciaW1M^lTX`Am4gm1-fK~foga}TZTU6rE+9Zm?mlZ^wdsj{v7TF;3KjPp
z{Y}?x%AMT{%?4)1&`OuF%zd3`-m#}fi+9-%S4`=LstL5J){p^4IuACQX1SW*1$%xQ
zi<HxJigxO~<a;a3#q)sx(2e}-$&190$BL<CR8vbuCf1<@F{oClsiJ$ciPxT)*4n!B
zk3BbPi<I^<H}Z?a*E=bA>QawRtF;YXy}RIeKE7|aTzR^8XtmH^cC81lP3|o9mI4VH
z!&Ym9)m@0p9E^c|I`w|D@EzhC!35VP2R4|2IMZ*pYD2zlv1GZ-u^n~yA(=l|aBMmn
z8&em`yt%S9fTe2fLf_tlzouC-OM#0Xo_2&abJk9{^Ld=u?hD-nvQPG@5itTe)0Nr<
zqq#$n%V}dp^L(P2<o60oh(}Xt4)fducs>IHmwM->`n-K<TbphRhH51gp^3W|GT?}=
zk^4tAkspj)ll2YsKITT6!yqPu`tF>+pw98%QPWBoA8LEyyi*m=)y1-4xZhNo*l76D
zVWj-gvl8MD8(Pb=#&vkQKkaX3v>POYre}J;hOHuC${fO^+C;Qz;@d|l4rbkpv2-e8
ztCPkMbvWT#d)IThnox_|Cx!D$CB$|j)ELwII!g-g0=p5*<i-q{E{c+#Y6I8#A*1Nk
z-7k|6OQN&m{l5%d^;0AA80TL2r=q$<4<{aru8TBxs`@8vGOxKy6dn3n&xN)zqK&{K
z;9dNWE1=FASg7KTt3cWO;^*1Kpv9_yxIGs<^Vey+dLE3ZO)A%E8x}vX*eDrU8ON1j
z@cy1Lg<ip|b4mG5qw1j~-5fFL@UBX(xmN5m)EaZ=i3e(?V#J#sK$TsMjHTGkYkPHm
zmg>_D&=#7Izof&xjcUJYwr`yDGB{{67p<Nntyqsv9pHI^6d8?a+SDGjiIb^F!8Goc
zfP05;aBEA_i`jf@07#iKH>bQ+T9mR7Xp-(H$XY_At*#Y{eY!|1czv;A-c~=KK6yf=
zN;F(8^zsKfUmep~uWD60%@T#w+KGIIqbG2u`Yx-Fvg9j%mQBOowNW!+2NOlJN5PQ*
ztp<a^gMOju#|F)Zd<JLq0KE`f@w+04aOmL=h$lJU+2jq^g4E0-Az3Neg?^;+jBXaP
zpQKlxL|0i!1HzlZ^2pY!J7UH85>6?z5b>c7&wdimvGvEiGpE3-#Y(6lq_CC*i4RE-
z73eEU;h5Y#gu$$14LwYj+uf6jV)kN4MjMbN?a4IEM0GbAgyJU*QjBn4aVnL)^BoaT
zIE$<M{r>nD(mXT&;Cw6<n(~jGf$?-7&qQ>0-21n59Ii1v@11tP$PIZH&{*jxRx}>S
zqhHL04W;ZXveSq)Wrz<~DIcy}848nX_Mk`2mdu4;%<hV~o!Br-f2H1gIy&cJRxk=y
z$0&Q=3TWdJwJ={)uV|rv-+E7KyQr_dFDv`nQ2KKqL&B>&nuhVM8pBqEd_0$CAf*Yv
z1tolI5st_2&#9mgIQ3j;Va`jmPHvU5Xo15ftDkdZA#1y1{+6PzjMh?O;JixvNx&N6
zk#%u#t5|+MZ~wXC5SOGQ+-xDW&V_hue!7~k-YlklQ?!^$$!!E4HSd_Dk^wbb<$Dpx
zmKHU;+-Na}m!}c1oD{${HlMU_(scxJ#si^EUlrRjHl3<(&Ck;dqUY9|Hj=eTQEM&&
z-7Eo{bdQXj+esJlmX1~%is;ccKm&t@uaQf061^J2(PpzB3hUXEVDbKb`%-1mF}-;{
z%I&4v?^a7mle(z2uM=oQDAN~ws@5AHFB*FC>2Xnx0~cgQn$7zbU<V5BcK~P0tTezN
z7ucAWFQ@+pDlx(IT4pV@mu)o6a1snM>a5ANqy4+;R7W#aCe&Lru`!Xrop$6<Q@o*8
zhxa2gWz5N44`P^6vHCf;^D>3~;R;guaY*D@`)HtxqojoL8up;X5SweGW3hUc%`>iT
zragr5CAYFQ<QbY|7x7s>O(o^LSDYTiTmI49qT#RFJ*_QOm!kZJo~6TgwCan0584U7
zTCVrFIH#H~<=SAUoIeli1t-N9MkC9fm_Bcb4H$~%)-d2$<bQjVVtfacipWf0pvz|`
z2^fogbG>Lo7PR!TsoFb*y|XIq%8rkzi`QTl*5$|0&3*Zy^paojOi-jVU<y2DFrmA}
zYkdIav|jJ|Y7UkV3v$ft@!-faFh*y7xDXrU<Iqs4xGVm|*YvHJxb2;d!*A{Z-jtbe
zoiS6l7C|uc=EKdZ=dRStVm7u+#S9T}9}vFU?Tkh+F+bX9?Q9)Xv}1Smh)vq*>USpK
zoK2sA+Hk$)>8F-Vx)3N$uGJi<HIW&tx_=+hp<n&jVv|yaw;cD^*4maT{;z3r8I+E2
z0%P0X!|~M;B6y!^lHHTpI}BdO3NXdf6tr~J-q4#;S`Zp07_Pn&I<??&DX!N~&|H74
zmi5$qRhx;L`S@{;z3-{|T<xq&?#d0gC0usERSwC%C!|?>ZRp8s2rdJOQaI(AmJ4JK
z<>p!F6~x62ma3N>g|)yAzBg=&ymrpzm{h^^0FkOxDyAH&w6O!Um+d*|%klik?}O_{
zBx`6CaUTzSYY<0LFKA9`qAav$CxM|7FY@9rB~kDEPZsgfU#=FQf3QfzZ+5z_&p+zl
zAKNM4i~YPbO_5LW*E+P*3n|>^Bb$6K$heB0AlRz3>TB*3VlNC-*)>Izm<v8_vtZ$s
zR|GWKvJ4W!Z!*%YX{OF<?Wc!^pj>4?mT%0n^*WT_jI|Z!CbDPhK@t}?MBGZ-cAq%B
z{>ZCGxW2CNH7X_AipNVW_~p*0mvh;7Qf>wI9aFW(yPO{L7SX#n5t~Hpj^Ws7v0?!7
z=OXgkPSzB~gfeA6jJAu@WH{n9n3W5SH0y8=H&`PR7{}l)KW^sje=Xi~6NZPtwp`I#
zYVbXA^$uz~k*+s!Rpng*qAzPJ+Xa|I3byLbw%!_VItj9O<tVxDXz+9wS;#xv_JG0!
zm;XoT{lDqalZ-@PMZX8a-{)ZLflE??8P*7&v)A26#lqNi^9;TflX(x<xp%L!usJ^e
zYoAc4U=7y7IlFE7k|P6mOx$&-f<u++SIy>7L`Z-Q;Pa$vW0$ECGM=2^qqmidO}5~Y
z0H-uE!$e(bseJTydvKX0cro15)8+T?=culoc?3Ap#l*~-B2PkQzJK~kq!S~Jg%&L;
z7kO~#=5fviWhdo{4q%!bb#b0Io>3pC<Xhc5_+<FmJy)rh1&C|PG*~x=E4yN#gAfeF
zf<XM<x)23)n9HXHuMwL!)Dvx*PE_TztVEaY5V4zzotBXk5|(z8<`(n>)gvwlaGd)A
zmFLZ-G@T@~)ux&g95YB&D8|nq(IeDlXX^p)A3KBLDoS{;me#7t)IX*RkrqX}8Nx-y
zPJ<ef<mnbVfA226|JRoDe|q$zubZx!4Pk!47REW3poJ9OosMwyZUX%U1HGn<rf~Sj
z;i@Dx*2!7p+@UoS4vW_xJ;wByF!dtKf$nY54qsVXEdGoS1Anjx`F}sVKrvq;muVK~
zQI||5UP%*$Q&lc4v54p*8OK<_L%b{V{oJ|M?v2Nx7LStL%W-dFpWiPcUzKdv*5Y^%
zvXw3&{DcH_<O)(nJpyEkcy9htKIUOOA83)oN|@Hmk;9=5|I^d>SDzpzt5v-t;JZV$
zUdg<F@qhED`fIPFwV25_l<Za$tMxr*8hiBSvZQHO+aD|+@6Qj9R(}cumm7a2vaM04
zS?3XkT2D%5;%^7A1e$Ab*d4J`9W?njwXCjY7N})UnO*hmO+C?i=G~p<p>71+`3SZd
zfqqiulG%JCNJt^5O1zXFx-hVvu}e-GW%;_vHrj<rd<q=9ad@U$o!l6TaEmE_{r>mM
zA>#QDSFfOxQM*3MDIbcVvl}zXtn~-=7S~K>ZhA;D;#f#zwS+W>sz^KU&CLv#SRS)h
zwU(zUy|Mb-0w3uOr@hPECPqcE>q|sst>%*5NGVTFx>du6@0YO#LQuQ_72}zY@izE}
z8O&^GIm7AIy9W6m9)1%Q*C0zx9tj&5YN-nJ)y!*}0VaP5rON2ET|ODhKlckpF(HMM
z;kp6vqWwg8GQ=PHs`z(Zy<^szE;X!+FV-T_6{3LxM8CBM?OZBl0sx~cs5zny09ChH
z3-tE;Mf=$-^B4Q+nZryd>62iko_tfSS1;L6D_QpoF$>dXq}985W|%^-0;t5c#9}xh
z)SIgatlZS0+b@MDNb}l3yo)j{y#^jQ%pVItD^rXO(qxZ9=eMYTzJVNXB)IY|Z)(6(
zMBVBW^rGh_pVLQw+MtH+hQv~Kg1)7I$(TA5TnOV|i*lS26PcBN_ACuRo(6OsnJXB#
z9x(Qgdb2}8canFN1)pLucAguC#K(`>uaQf0TyudpFN|qVFX3-Q0H@TnNIb!N_VfHJ
z(Y(*}xP!l-gL~H(ziK!U5j<g&CKRK^0IdU|6+UxMcjthWvsh++OSI4dlaHuRC$yJw
zjkSj}CYO~vCx-kF)`0Z34<)6{WqG|IWVy9~OeE9a%8T;SgKI|1czTGHQXE;}H!x`b
zV`;&>W}MUGaokK#*Urtj-b9shpmAf*EKZPJyKVi$lpvv$DBW0^yoyBecFhHgvL~rF
zcg%!&T!N;Km5J2Z#ixTEx<Wov!6`QSTq(5_qI5;SPcFv4$O2S;^3B5Z=5S^+jgxET
z3}8CXHOd8WH62UpdAN(#H4OOY>eC^vYEK$PdR03UDe|&fC}LVF>js<R>OS=3rvyF>
zc!{@qivRR|=JXS%+dYVyIxjR=^kDvRh#R+=V6uO%h^ShWD0%M8)VN9A!i9CT#;>HD
zl;h8HLj#o%R3!pLkwG7Um|=xGk1GF2ZWZ7R>g=ndwZqY<oc6G#Hm%)n$n8#nDF(A~
zt<_|uI|UtbGUw5XJa0_uSd&d8oewmnXsDfMz+{7aIVX}^_(4)nB6swdBlgZY#uzq+
z*eP5k>Ew7Z>ufM1^>QGxckCz!chU?CcTi^#WikmD?5+_*Pbjk4_(A#LR`{IkZ-LeA
z=bpl-T4`s5Gc;~qe>GZzC{`dd|4njF3~He5LsR?V51rwWg?os)8Z~KMS@AaNhgeI9
z2A7P_598}RVQQT6sYV)-H2ydUJ)e>*OTloK12Bhr3=8YZ3Qjsq2hDgXehM`(mSY=y
zzlgKpC^EF6NKCa?WneyVu*}+063j)E1vFRmMB0^1oQ8lSSY>IoN-I}ULPm&|@i3o-
zLbZL+3NhP|zFr~NVU?S17M&oSq*}8>943s{s3{|K$iyCrCVr-<T&;lH2mIW?5P7KM
zl_Ea3sPA7NWUak<CaA<mI-AsR{k(%uS+Q%IH%BIEEb7cXw0lc8k762RbeL_Hq$Xm8
zeaW7gncG>g0}q4dH3U+1vso7uye*>VG1346)$o#Uii+w8!M6NtH`_eZfzKFS<}l$B
z7~?=x!^B9-Jq(bkRbr8^OM+aKt!PFU($1q6EBA*##lB2|6{jh*Ii#jwSM)^iK|Y$c
z8vAR0R?BC_IG?sm>nK%w+bX4TcH$~Yn1*U>Sn4w#R?OJ7Y`J?+LLYEJ{e5udN;Ti`
zovvbr`f&X@Df|T}=kCO~g{-<~=Ejfk6Y$qI!DQ_FdfPXDuv$Bomb(Ctc8h=QT@<*%
z{$TyMn3i1MeDf&87jSvYg5tYvhZ^0rm1A=lnnYey*HS;E2@l_Wo%?~WeK$#VFmuRq
z*)-RE7u$~Iva|V#zbISfiv8f0#b5PV)e^|@-fdz&S<|hQhj+7$rfehGNW;|Xgh&PG
zRjDl4y(;aD=9Di9e-cOWX@=1+SrxDpCKuc+o@OC0WzO~o%la~De2+0&B#(zvyLW$$
zID~#PhsixB)+TH?3+fn6#<xjCyGhTVoQyLsh{VV-zq{k;l-6;@@<o;LSy##cGM2+P
zi6GfZ>dKRJ!h-6nDk<dtboF*f!~L9wf=`1{4PJBUe0toLp6><>LaB8~zzKcu0zU0A
ze0`7kYG4^BXI9m9qH^6Sld)4-AB2rjZ2h&h-rF@41dr_$KW0g`@R;~ZJVGY+3#<)1
zcolcY5vr>fR(%1sZBqwU|G`S2_3NVXO^N)2)qN)<IdDIM*-_QJ#o2Zb!l)|^ntacG
zWc|U4f8B#Wbh2Md_7kojD>QgQVFSc#-!66VDrfB}F`4*ensh)ywtr~HRca~H!8Th|
zbT!62oW{AC-5JteF1AYfG48{O<LO5+=c+|kP{%{~b}n*ApfdbT&pOQ`*{nn@L9j#L
zsn4_XSm9D9L{v~Htd|3V;xrpG>e^-!wT6(=D0@S2N?2M4>W^^G@Cg21wQGLConUQA
z!3?Z1pzdS8W+y007&9uKy6ej`^%{?#?S>Q8Bxix8b<yBb=%dp5sOi<n)n9Xmg(e<4
zh|l$SD#$gEcl=JuOGbTj=j2@F2*b}M=<$m6YZt@|u`2!|oMo*p<DDwz8Gzby#dsB!
zbKac#7*(`olS9+s){DVU6C7I>6b$}bpx?d<?AU$#E()1Khj>9V+TFF0HVt}u=T%Ny
zDlDcNT-NOSxdxDGy;LBX@l+pul~7nL9afXt9U4^ze=%290b{_Pxk%mtgzesou_={p
z@eew9Yfr>YjGAd)>r+C>Nur#VNV_=6kv~ckEKTJQGGAt?VYNnXi1ZO?BtjvUDMHZG
z7~-Eyk>Y>I#XpfKi8f`iy9wp_TVhc^#7sWo;*<-PNWaz7BWimJ%BZO<m28<@YTa}e
z!j~88>OQD7m)v$Q$!H&>!akdy`n^wB>f6km<NK)Wy%*5U;bB^(Lvozk`VZD=BQEyY
z+NsX$UhH;0F13D9CkG>1dCFU9tod1WML>MqcT;k_v<S<1o!H)+1O_vqhbN}tdKA6#
z(K?O*0>&~BY1FFYKt7C5RZy=pn;^RIjF30g64s!Qa%jjbYBldpav|(P8}r#s0*`)@
zDj-#9vvR&#ZP($7N8L)XjAwSrEOtaMo=D!?`n^|3(}eAW(<zinUqG7_)1Z|{Ov>;b
z2LrsftC%v9CcT((c*D#|Z#@3xTFn^yilOCeTbWocTqXfh=lQNorjS*CA|Y4crq+2Q
z^Q6y<o2Sp&Gw3I~Wva!sMqdNe7u~N_O@+#|&mZWpI~yz%A|p&@V6yDm{bGo4Wf=o%
zMn%YH30={j_X#PO?}dCR=BEhZ{*C~Z-glPn_YviX!lUkbySx@RYxg5rEx1z6;1wZ5
zcdW*D^-7DAxetVm0PjrY%IbvUdN%L4ddEDmvw8xNitBRsDh>)#md;k{*5GFxI!Hk)
zkYLJ;By|A7rOVsLy12m};g||z)kIj8{$&4=yESOXO!8K(SGxTL9UXcJ@eGMD{nbhk
z%run35B8e^6`fr?<l^mxcp__OEmtxeg%^ipZu0#y_}=Y4{#SS1|MKzwpTbd|guFf3
znC&KtMsd)Dm#B($*`~08njlY0d1kn|c!Fo<bkt&n`*%q?l%y6|8Bxf2s+TAi9cA`x
z?lsw~BhI2c*~((1KWP`Ljj|^i-|1<IATw1@6@>um4x+XIyP$z7N|kax|B$RH`)KgL
zeTLZ7lrhtbW6r|{6$48;{Iod34)Zah&WVY&l@^U6LoErS?{>3#HHdXTCi}e(PI&rO
zA?EpRxZEsXS^HV?@o76=V`L2qH$*hU8rGyA*3{%p3v3OSWhzPC8~5cpCt2uf)YtJj
z*KlZ^OWH-OGSrf$msh$3G^ZG5msq`~2ypJ>oGWL(e>K=wfK-*ifbR85@PAX_(<l%G
zlZw5r%#?<3NC#3URqDz~_z|w567(RCEi6YddIeUDUSEOB!W4I@*_HWxKlJ<`p-L^c
zVZHtX8pGBkAL{5($zlAs1XO8e3^dg+j(IYKB%>HnGJkhGMj)ZggYkR_x=nw-s`+1B
zwCaL$V$~DRxh5mZp^E)L=C0J;DlwNxmK;=b_ri{oNO4gRagMH*@c2#o_p5#O+;g$|
z3#_L1|GYcsI?pZ$GQ<Z#Y}IoD=z_-p<5QRh*~T<w>B#(bKEg!i;W*S<w^UUQ5A`!R
zgh^!2in~bdB>@@sgc7k9Pmo?-%Et5qJXp7xCy1Pe2SU1<4o07(`3V=Ct_bbB>Id@#
zX3Hy*T8s~rA|2F~z8L=7tpCjXpN%)&8UQfT%U8phSew;J?pfJ{+f0jTB446hI1gZO
zz1of=udfamlAPCAd97~b4c%?Psns_7#z4<;UU2it<2`hatrZ{1{Yp^Qu)8tS^dKe0
z_TWx$yl4atFEw7<ag;Xc?CC;=nxK<e&(jP@cQy-?Hmqi*!IHF^;n1P4U+uK>xtgwq
zFb(pK5e^$V);(M0`IA}>7+uNE8X~=)<g?cNe|8H0{(Qx;xY~H?p?8`gdCD5OOguLq
z%1;jdtV;_{+)XuPN|*)?BQj%BLe$03+c)5z8+_sZ`5$jgnr@zM?BTB6Iq4jywyH&B
ziPWL#d*~7{6D%=`WY&0>;-e9cs^^mR=RD($r<j<qOpqCcS$sI$g%$lSezTxHErl$s
z4x5<WdreTwOf`*Zyp6Ffpiob7Pb4zSDSgJPhIR4gKc@Q6{xEafW?a!&Uc-EkLcWln
zVPQSVOo!s&<xb~zITw=q2!1n8y3-nL?BnfEBH`D9e=;Z=sIKx6)0Q9J1uM?)G*km&
zxz_6H#?NMIletz(<)U>>8X&RO%<Bhxq<I$QwjkEzU6>g}K|M+>NB2V@S&{l=JX0-E
zdr4z|&e*&h(`yeAJATjhMX{SJky_>Yg+@LJpqk1bEPd2Ly7B80IiszKJgxPDG9tVo
zgJ4_sPzm#1mer>*qj<lz12*0zfg*n-=lbt<5lcq%MH#zBbJ5rA#;CaRZz7DG(F2U+
ztoAn9oeAG~1*&3rv6H@FR>Qz8@!YKZ*aXv|_zT%K<{$VJP4l?>Zqr7;hf8RGdg(sa
zy^i9XKX8S3pqs&I!?w@m`Txr29Xntu>DYI6l<0mZq(>dDt(Tx!?GI$2ulRxvD<I63
z^#Su}u7&a60^zKRV;0mCs%vU(iuJKRa)g6vb>FrI-wDh;s`0F-&v=VYBxa<o?9JZv
z_8<5OZ{94e8?XuW06opR6`fW2p=rG^#8-+;Udq@l<I^+VW+03VU#?#yR|_z`#_P2C
ztC#CUX7|zL1?qKQ-k9dpNOxJRIdj*Lw8+oNavslY_ZFQFU%m4@6#pdurM%0pn6-LX
zqk+li=2&QgbpVh5>Pc8sj?hLbmDIZ)T?p_4<1w)-QsKi}9>J7HQ{EaAH0+K}My!=c
zOC`X4wuiYvZqn+KHwn;Ht+Gz<5SA?c&9031u$*guwi5=6>gQ#gT8?ubikT}gw|;th
z%9~1e1+pe1D5Y7%$+TKbu?>YIw1EZD0yKstL;|>!V~ux4B8|D(<I@d5J7z;%4Q5h?
zN3Z*1&12R`-Ic2mAGx!@$cW+Yz2Y`HvFVGtgG;*kyfIW|Ch-xkqcw$PVaNoPoO0%9
zTUX5%1KP$9-vNi1i&<n&z*(3>(Lf^D4i-rlV6OSAF4qODT>n}52Q%M*;<U7*2^dxM
zvGw=JrA_G5r<(@R95t>gYai9f_F6FN_H_$0X>S^tS-Bp%?!k(wrYXYGDo{b|;i-#i
zxK52%6SuZ?S%ZRC{na)7tfXaVhUDv~EvIyI!6O3%-zR-F3Ou@VlYIqo76S5mde$~?
zzJ{3CX=tq{StKfDqH1g(W}81*+ZxKmff}Dao>w@5?wiWEPb~vsQ4`#N9n^gK)XN#>
z1BkWxKxf6Cx6+9BhxBpjJ(BTL<^n!%e)y?P!z9CWI(NY_P29CBJrh#nWvZ>_1vrW#
zepA#f#tJa+DBfi=XLSCEbZqG1Qphl8yG1W~fhW?GOn6e#z%<<4jikR{pc)AF_EkI-
zG!M?}GWT>jd3&&v<fAn(%eM1j@;5~U7f7oeXx`voC=`&l6JEB(+u$}K3L6&&oD0j8
z0TT`7gZg-{ll;AlwRtO)c+-_ivMx41W~ocSi?+RvPp9bf0)^<fzp^BYEG-cqY?w5v
zp$!@iMhA?;C;=1VKy)cjWZ`S)Jo125f$&a$I0H@V0N+gtB(WYnL>c7Fq4*<cyq<0t
zQj<puZ=&GbT=|1lO-o?6IBZYbmi7))|2F|&JU;Bc=cf9D6)R%)7f5WJ;DE-i{cdKo
z{NX7lCH5~tgwTB>UhBifoImCz?w<)tgRt<my(V4pOKROy{_YmoQ#ihTk*$qVjWTNp
zDsjnUahYARN;fyqdan1nIA~T;I^~5;rI+9*(tvqTe5CO;hg(J74Dj0w?qe^~t5ORu
zhD&b(17p)zXIuroF!;UDR5<Tf@sL}p>?a5brX0~!3#BnF8~vr{O@2hD)E*w7NYj_U
zc@#)VjvRtfM}E6TwPk|(te5QH(3ZNg*QmXaxEtp3l8QkSG?k$2IWd`4cI@P*6D3hK
z+sfOwzenp-x{)dR+D_7WjYW7JyHl$wQ7>z|0tclJr0kNJlDAuaAvI{`ZMU2mUNv^J
zo{K8MOcx;le>qIOOsak$Y?a#pn6&kh!a;hToubpQw#;Nm%(?_M^u+Nc>&fTqd;7j=
zmz-nf#q{t4$46i7_g}O!`8gH~J3CT4@NdcP>}QG>QhYUKJ>#M>*V2%vi<7Sa!wfb_
z-3|eLXo+1072N4z_vr%wUHh8zRk|Q{NQo!e460$RMgSf(PJ>F0m99#jts@+>ut+fe
z71rSrG%%0?W3=#jNlUEUuqQA%fB7WS^Ly8FrDfE%N+l$ra#UNvyQi|mmcLsdXVChy
zLR6pj9^vLdjs@RufPX0}4K(KeHkg?fmR9h)+!{PGLS^S5LQwMBP&uDW_YT}&o8DUd
z)h<6Kp^Xd@0hLjzS5|5}dG=IxpHf~YTuOm6r$1o7Xy#L1FF}%Ea^xy@4)}CcG0^Gx
z+oECqP+JjT<hHr&kG1M?%TUW1Gn+`(!RnIOR~0bUNMosIbDXBn0<B-HYa0p&wt534
zx*~?EtbfhcBDhC94@j1O=j&Wkv!6NE;+9a9mK`(n+aMAF-$rk!1-onwto~qWVmvv0
zm5O0(srI1_JGoOTW)b{(+CQ%E-80!Pi>`*@e}!E;fRo2*eu<~Hs-b?_-|BW>tPNUS
zUtOObf4Q->|2(hlzhGatk`XRDsm!6RWr-{$8waHtjXu0hjhGS|zZWUa>Vvb*GR-lM
z`2eNZzV(uk!08*!4edABlq{n9CZg8m2Adyx55gXfW)uHly(|6+BumP^FFEM;%~x^!
zgZ0K``%!$Wfd8$FXYl=R!0{g}#-pFqC)deFv+t+vuJemC?g&9YAI1L$T;H+)cy2qO
zc6Yx@u-O+)LN97uX&+LrTkpQjRXmp(MEd+Z$i=iy`1J>?w^<&W_Q81WmK*-!{@~*B
zjs^XmW^J(h57ycP744?aw(lB<<dKtDhwnbuqU6mdHQ&B4v&Zt+c@GkM3n;A^`-hoL
z6C}*>amyh!d}7}P5(jIybUQmS$_%0WyNpm}w266<p4t{{t&b<&w^Hj)lCyfIi%wgH
z1R{=SjPU2DG>dD!JGxy(6~_%&$C~16of>UsbIUvol#`&Fu8OA)uX~+EpCEfioz-%t
z2Rdr**q-n9bQjj*0p=&Eel8D&CgF}2ika5|f5d!;Q_ny@#W3cw&dVlYoU?&3>TwDG
zaPgR-AJq8wWAf_~gSElLUrWI9i{?TtciS-AIeRyY^!1y<ym}JrR4~u@-*vWmn}?>(
z2!R)9S%hBvhKMf+{k79ano=ay=OV;`I#fc69bez8Kat=*68_gBbD44Dupg~Br{I{C
z-cY>N_q(Ds5{%!Muk`2-v)Tk;Hz$PmMF8_7Y^}(~(DrKEC*K^1)7**LvdWa?9+j2z
zza8ona<R9Ww+-5UNZBS`bmZ(9TbWhM1;KKsC5XST*WYdxBssSjeQucdXlYs8tflE-
zJ3QnS%bRPMv2aZr&DRxnb=B?|SrZ&6)Rp${#)YcouxFj4IO}$=o(!}Ufj!}9cA5Oz
zM9JHDn@$IMzMGh1!X*4vu}K-vaksbDX9=OZY|h+v+C-khC*}*rozNJsJEO}`!1Iji
z{;o&izPxmWuT|3EaMRHio$ys9i$=>y4&HAn%FYIIG*k30OIW?x>Qg;c(c}p2Dr>l>
zp4xH^Qo%b&Jz?B>H@-(AbvC*{7eDu)U7wdzXov^zefRXZw{DKl>C;|_LxYPg6b?+f
zKyuBDDrY>!<nCz;bl-z_;-CJWz-UGua&S9c1Gx{r?@QhDUbOzf60NYO)y&{8MmA+T
zr_x>uM1CGIL=Ex>ml)ZfIUQ%!^J?XYGt?>%*%4845UrPaIL)e9a@8jdwA{@cXX|8^
z->%esx$eKtIHa4F#FJobur8yh+B<U65$-(k#+ww~jxJ$Q>^KU$QDzGB>J>oR05dEJ
zNCu9*6faI%NLVk3;M;)hLO-shI&xia*wUiK=o;j&TC9uB8fpsqWGyrCbI(0vwqIJG
zl7EXw0NshW;`Mij>iQH-04JNJY(T4sd`;yfuMnZ*W9)Xfm}8bngP%KgPU6{4Q9R9b
zqjL(C`M(e`%OGb8X+Sku@xkk0W8E%D^l(pALs9(*dr2{bi~Y^qFOx;FlXo|I*J-8M
z8;$%cEXV`l2~fQgC^~J@EO{qYO$Da5IXZ;Cy+&vY0wpMel|pVk!X<t2%Ua74WY};%
zP>iu}o5P&(aC_O^XsgDZjCNj?ww`h`V_#}?g(ir;mzkDP(3IA%ci4FHg_c(pP=0H5
z>-z`my#b~?u^hPY8b9f-d+R_<(>JHA&PPlhSN$_)TmHXyaNITCXwn`<`W8*&Th>+h
z^eQ=C{^*kQ*XzOSTA{oTKnh+Caxa&@d}@ltw=pnq?SM%Qc-1>#FC#7W=r-g`+MW;R
zpAQIdiAFNEwm|SH`VVrx_#7&Rjg5_0y@qt2Yy;LxGp(6ur5oO=2P^K7QvjoB`S!U+
zXzK_%X9%bHV}DommLuMzJL4r^J8X?I6*TDACrapFJ958$cD{e!ET=zZG`iy~ZX&W6
zd9h$5ks&v;(LaN|CTa<4-*~n6nxj`|&s3C4)UG%K;$W^q95^#K9eeTK{ikFg6)cCf
zTVD$HBOK%PK=kr^qQ4?2EGCAV91)FmoHac)C_ut`w2>@Is%HM~I)7S?ABpg8c`xjC
zs%|v>$eOlQMfvWTJ?_UwXa)}c%*B3TwTYDCnV^Egvv{VY_3X1pkr#}Mpr>;CWbeoi
zYAM~H!t-D5S#w!o>=3Nf+|_dsxib6F2&@Oq+rQy2N;hNn(h49dbX?r+Hyz|L@yn9f
zv|sMC2&m`SHqEjqCtrdRRuXky&DaPs4S9t%ykKo}*6-=*&x+j|sDDkssi;U~4OjT6
zu0N4W&csU3HO?m?Ep5oh_lAbU`fp;An)Wn&^%Oo9VhSL4A&wyt2n3gn<tC?PxcK7!
zMUr$~&i$El{)}@CB4!>ofw4&EvQPwRLKx_I`8K`3oB6z>{cg>+6AD&iX^UI)Ew|;*
zqXh}Oc|ahYtW?c0*td<$qBwC)w;YAOCBU-GYCT8#n;1|Kde+P!^;W6L;=SZ&+Su}n
zUyMucc?xp9%zHsLaX!+?6Z*sL)9)qKd3hNaHsxkz4Ign6Mt}LQUGA&J;bqr?uuQM_
zCtkGj=S#vrkHze8$5TYu*jPVN_~}WUl85t->6^7;pT}04A)1a0JDGrd{JBii;@6nw
z)M4p=xn*om)z;en!8&TSIZW$~ue-EhR;etU?yE<5QsU}1TM`Xwf%=eg%Aq`@+p9t8
zNkNDr&;K4b!vA0AMv(u1aU+k~eEJ&LgDkYZza17blZQ&9$l(?~e<@PI+0-&ZqiNsK
zRv*L_4D$ynKcNex1P&tJj!vR-h=X_~WFb3Fv2*<hGP^Kbu~d;O5o?@b!S2Gjgh9dQ
zAzN;+M!g9o+kmfk(fjQzvc@KboP!f?7g5G&9hLkKXTbsqdi2c7I}M3H-t<1E#*f`k
zDKh5fx62i0;#GOGpy(-4s7!d)E2$Z~Y2LpZT3hqQt-rd|_Jg!T+jED=ifk_M-%*zP
zy!lVtB_VIbmTA8S^JD>R`=UB5Lfh5@!RA$x=C-V?+)lWsq&C7mf^@)eo+@4j2lfN`
z59C*WPxMX3@EsFoNV2>mH168EC>I1j*sRi?vfyEdGz|wHfVcCnmr9mSb=a`^tIRtB
zmcYE4ZsBesz1)Cyue&rs&A8Ls&+Igsda&3^#_qV$E9!_-JAlvFoO+DUHo<=}ONRSs
z|2NQ?9G7UIUEvVfC{Q^bOj(TPklB}&E`Sy~QuwEDyjvYFtZ4)06N$->PJdKbO#U<`
zKWRBId`7)0i76XoU&L+;XRKm+5*bc)ERi-no|{V!?}c`0LizV*z@`cVJ!txds=TE!
z+YO^;kD_KHjMZ-$YtR9OU6$R#O7to+jr3x9xIR~_@${V~P7x-j8ff>a2PrE_Q=Xyo
zkLeaeaeA_8Q~qZkFK`qVR!}qg_{lJ!w*Gv1j3+s4Ny%)b4%w0H%s%85Qw8%&P*#pt
z_UKow*wvSISv2s*^gUVtL!kV#9h|nb#fza2U%`}PcB4gU2a)vxu*6jr!0-7R35jxy
zz30VYZs1ot<wpPT;%(P5-DYU9vzb0p)+aV2(4dG*%aw&dy)4va&9L@&XalNaNdV%!
zytdmw?csu>?1t7lvza5a^SH5{@9M|&?!egF=M|Nj={3K?sAwA#=!=TdhDNA2tZ3G<
zE4pwKtDm!h8P`*B-!ilEz3x*Ku_N)gG;$G`<6zlMPmCBu1@$R2(QWGU+0<Ni`%2SN
zi_?@ZHMaob_ZP}#mUJRpPy>sTs0bMWol&UW)@-FUi%g?O`Al_i-Mh1rVzi7#B=o#4
zG&Tm67)+_^t|lX*ovmDQoyXkFBAQR8mSw^WP3AxwXqscpk_L6&9j=43g!YO;TFq<P
zb*HBPv!A+08AF7URkTjDt1+#CL|CeG8uSuKz(Cz^p^PcO_<>N($<bMiKSgM5-+Qqy
zG<>IYo!fDR@N9}dSl|VMnqgq)%p|MW!wA8`bE#k^R85^f8Gq?B#jz|&a#m(DU#o$~
zF2{mXUVHB^PW62A>WQzqjG}g_afB(x7}N~nI*lh6!FrUeN$m%W2}M28sUWpUB1Q`J
zEXDW%L9ikhhXXa&xXri44+E&CpZ0s93eb-)3$+1m2Tn6pBmpr2QFguFQ_6Ume=8Kn
z{G-Z!ey*sf4pk3kRUj`^<YmkWtr<pK5E8l|5GN;Jj%QNv6~lQpy1W(c9rV2OH8X6Y
zkmTMKgmhi87aYqOt()wBu(+M&agJcml-zFhM$CJ>69^Z_Uvs4lW#*h7lDIe}<RnU(
zCgH`yCKo>l^#?guA8}%Cw2}}7K85>i?EC-XeI=>6!FSm`V*~jItBnoIas-QFs}XN}
z`=S4g;8QJSnLG9;$VIwFqbXmZB28^`6-aYImOJ7DINW_LYMpqn=%&tMP5FE%<zy-R
zX)v!IzAIp_QoI9R+EXAn*y&*KK!%)SA#<-8qE0FuJQR=k(8k!5s}99!R0`RE`JYb$
zLCPvJ(6I&oeLb~}VNcqw5PbX(+$pV--evaRW@H`9oj};(KY!K~y<##H%@fg!Finb`
znj$@2(-_m>`Mdzl)aVirD&#B8lZ&6PSZ!HTvsBh}i+S|7ec`H3?V&)5@5KIsrxOR>
zz4%~se#6(ih5+5gYz=vtW_BHZlruSnRgvj}j^30gy;$8i{04!JT~RPrQuw}3`SfdH
zW6kpK444K7a*6v%qmvca^0DS%&*@-s?#>C|x?aqOhpTOgP?ku8vb>wunxF1WOQ!g=
zp{37p@Re68bUS<DwIcV6MS2jVBhpr&p*482?-!DRw8HRd`7EakGC?&vxmIOHAlhV-
zv&VwJ&}q0AwliJ2yDF{K%sYEH?TkriL)ib4`uYEswl)<TXL*)3bF4BZUb51tR{FD~
ztR_|siObwSnr;z&)H8_5LJm@EL5tVr{rU4H3hfz*2XZ-6IAchmI>2b$H?$WYuDU%Y
z=UN~BU@0z*yb%wh<vHyl0tNH3qXxBQy&4>m%~M2%z3~>5CM||zO7~wVpJp4H^auH9
z#7$L}&fX)_7=AV8@!E3nh&D^1^?M7fof-@KloWKC@jjT+1w6G#C^_u|%)2Z7NCQml
zh2whM2?$xk#{E%qyB7zwMRVlJ3YB8{>dv&x`tIZ{MO#&$JHCpt^_s~-lA>myjBn9(
z55n60Bv<FsF4?lSGT%eN;CC4Kjds&>IHa$Vv*TcmyVGgS<;E^ZGC(m9Q7a)qSUN4u
z>Hw?Uw|F=~5}TO_YjaDqZC953DC9KJlc_h9jEMtnJ6#H0If!ScTeJ@1L0WX8R&K0=
zr>z8V6g%7y63$_E&#^n+g(W|Sx)aVc_~$i5N($76Q>De^$W*lDu!&GR#yL@a&mD=T
zwm2mhSqQfo$EzOmHtlu}^|?q+0YZviO<t4zNW;w=7fxDKOxb-wKe-w`pTXG2UbV$j
zAGvYO#yT3MO){Bjl5{#cS^No+!6(?Z7NqlR2b?=zI&BI9Q!1QChdmN#huIs*eodU2
z*kz{LK)-7CYPdp=m@(rhha0uUI8uR2+M<K1+Fs5jSPMR4xXwPr5%KE`#Tz})oeKa4
zn%>5YP|O!oVjTSrTUDisb@4TfRiR<<`Q|ufSpo0Bx%bHG&Ulf^KCsLNSybshl*oSn
zH9p6&5pQDG)|&z1@EKUji(k93;yhHtC*kc@z^hn1S|;Y<`lJ0nSR8|k-xc^vZbR{6
zg>k$=nM`LwV~=*#%2Eu+!<P=M8RuyBDnVhlgcA#r#a+(4Ra6L^UNa6UKc<05q6QB}
z;Ip|@Pu74TslavlK>RLDV$wNU)_gOb3U?^blg2bVh$edTGS<b&^Q&#Yg1L(KzIw|N
zZ)Z)YUXk^pH@Z?UVR9?>cmYry4JKETF_ig`Dyw`t*lN73N+DH>uAyko_A#pso6I#r
z>sX2i=7~I#<LykE6qe1~OGv`tJzpo!%x2g!<W8tnn-ORLP~+vym%v<M^pA;lE~+?O
z-#hX;u5F{g@yJ`cD#10-$rh;TDAOKI^NeYUDN6XU$C`SfeVAhGFlBzFfn;|o{8~w?
zXtmAzU>@OG>!{D0SDg8G`KhuP`DJORoEjiL=0GA}xulri@j2RqV^nX`oAQNs;kExJ
zu>?Y3pr<_9aaPlVb&IJZ#hlt~OeubnFbbBBC|-$e;35O-Or<d_Xt@xAlc1!_WChcE
zx?D!Iyyb@cr6O&|F%D+iM=N<8!{TrA^Xo}w(G$d~PS?VtZ<c)cDOJ&f$4C7O%;+;_
z&c}ng=wt&CzaZH@+qv$y2EOEz$|8fM&)Rk-8Ec=gW_Oj6w#BulIptJR1%<DwOT`~P
zlCbm(+9I%JyiskS?o_inqh@bES^($N#|A-D#j?(cPdP<fx(XEq_ypASqsI348JL0J
zPZULQYQy5>*NvB6V-c(<D30ZDvVutCnXJ_1$8rklS=kg$nV6(4WHCdo=T?laGxAj{
zFL5q7de9-<Wo0YfxDh(>BFhwYP5O279AyK$ncW44dbABctM9ZfbKqn?oYGuRtSk^J
zGYF>2*Psb#eNkeQo#gO4<vPAhur9nKdA78`bT$whVM!{X@ZFZ!P9Z<tfF9_OEi-T~
zh~KLl^jjiY*fMgeG$%SWVS0a)QW9mvv(YK2eL}HreV0Zy>q3ecV0cs2dwOF4#xx(%
z%Wq&@y$Xm@?S-fG<EUqEzi*xJ0`TSe`yNOWX6I2~+1IQP*E|q=$njy1u<0R4e{5`j
zzbM^nu^7NTXVGB!s8J>r%%udN2#zqW8S3$>9AFOTKx^k(XP6u;%JELUslrQH%Gjq^
zbiH|N%l`1m0!xDXPPE4K$-Koq!1@P8<}>GrR1CIQqXFcr6L3+7?B|-eFJ^u+)ft!q
zwX2&=sTi2Y#RLn)3UvAY*kSS}RtWsixOqzFC0k>p#mkJ+Q<G~9e^!;epDNNjC_S#s
zg9cGpCa4Nv#CH!E(2+~P9Hw2RWF%3M;`vtiF9xKOlWyZ!;s<?7ymsJ#+hM5i1jYt>
zRZEa!<o;DRN7l@uRUUDgnx#j5p$y((O{F_R`-!RJw`JYV#1A!d>jLDqa&cwYy(I-j
zIT@T`$v-3_-4a;BmfcVSek+;+x@FV>BNHamc|1}$;^xev;cMr|yO6g#ds&%vBW%0H
zCbE8iQT+o_%2HrJ%D)v}!k=#jhz+W2T;?|P-yiLp?-OwMI;$s=ek}p9#I6g>`nr|W
zzUZPealRg+j$J2OGoFMyyl@B!Xt_yP_2J=dImBMz;k_q+^3Cd-fSdSr>ft+2_f;~a
zc;D{B_Z1Lpelk;uZWjzOvq`a$OO6+xK#dQK;+>s$NKxQT16C(`OVS`LX|gy=JfMXR
z^X7%9^J8j#RGY2WIKdOxZ@ALrV5Oy}TEVy|8y@S}z%Wt2{?IuN7bRzL=m3hTgXDcU
zeFJ1L!XBa5aHIVVwy2z}G@OKYatiPQ)bh3c!7{&T&ulCxZ!4~DnsU6BC?2i$GcU0a
z0Cto*<;xxa9no89!pPWNA)3l{!1WOjm$#P1zPS?jJ(mtg%<XRIl5p|Oog&iH3GAFX
zx-+YVSQU{|R$$>q3+XoRZmu*~N0?kH#44R@)-s?%LA?rLD@|EZzW~QX3f@HcMRmL|
zIqeI{c0p^f?v(Dt%X1?IE61Dt767-?qRhvM)KDEHE?+6zZ9-Dm{5o4KV+lsLg?Fy5
zdFIFf2$y|nxtZj~Zs!?dfrJH$%!bP%X8sT6&O559wq5i1>Z>Ax0s>N%E?s&DrAi4%
zuL0>KAwgQ`;QOldBAp}@X@SstfPkp<UP5RIp!AM(r8)V|IWz0b%$o0f^T(_;du1m(
z>)9)7?VY`!XW!3#U)S&DhCzmFQ;v1(;Y-Kw$w+M50H^IU%@7a+-oV4>T}Q;$hBYPj
zHcASS`;>KI)#InfF68~mpkJ&y*2w<-lv`lsS}oAL!0pcg%R%r|wDpF0k|Qm9nq%h+
zoy1rE4_#($-$4al#6^u9M;KEIP=yZ({^*%|^Z;71&X>Gnj=NPmaCB&oTU>^^Sp`3Y
z1_vwzxAw4XaSkWZfL-(thj;G;Cy9JKd+WEG%Xh}dlDFy-AmwTC9{v%wz&-cg0A)V8
z#m8P)pJ`pKbSC3^SXaH(y6QLXflRc&{Y@(B^mvWAi}gJxIPiwfX5651Bog**sf|Vu
zVK=;Uf-)&)e0gNTF}X?lZnYcCO67%+UrEpC|0vkKY%!#GL!{h*v{J!0qh;)rl-zP4
z710xjgfS6IzI$S+J3$FsLe*eWv2jCVl~OnB7fD9RC!#Mkz6f}oX2>`PyiB39viuS%
z5K|~T(Z@TV!Rz}%ZngH1ljK-rIg&gnhJ6dxm@%vePbn-~m~ZXMC4deitA_+dT^3b7
zb1+?g2os-gXxt-L0*_;lZr}P}zSGRUHn>|OJ@S`U%#Vh7x-u`Q)|aDkNm?Hd2^0CZ
z))wWNS30S--)H8fPHkRxYjf0;bxq9?*pT(h4+)<%69Shk%jjz3*!01&2*V{e?4`AS
z{2yQUMXcfzokM=KfyRcyKTCCgl`=!1X|!G*Ka#|gBiEb>w-!dPGL|~;8~GU=f5x}+
z)Hw+AJ_64>Yr_@J#a>Fg#iXwZPi5(T@!Y7#({&vSu36M!tO{F%q#Uuxf>J9}8@bow
z`i9&Hmaai<dOJNKKAwKH10`>T2ZYa#pO|voSmO-72K)uJfY-gAyM{;v(O{lsWOe;j
z;pZc~49c;wa(lCxV!Cfz2x?}6;LEF?2x?{drB8uhk7JQKsA!tapHUtibc^)0J7F`s
zAn$-KSl#y_g6RsTOsMF4bK0LncJ_a{tQgHT#58!a42ho!1^NMICPgP7SV_Nkpq3a`
zyS}PjYUjKcFZw5Pu>1AgAD<h_IAFsbx;tv0rPs7d_KCN}7pxs;6|G7Ul{qeiilMmF
zetnSTo?Tzbl5pDN50TfKY5ojRkOkGKJV2uA=Z4h22>H2*4IX!JcZgr;NoZN!e9JbG
zt<(hu3)BWr6ZSWp>}3%h{XCYE&t*B$j<rDF#>zsSB+&<tmx*pyE_$&tv9Ku!vIn(f
zeP@`tkQh5AqCb#bJ{faP#a9z@&v-uzecAt$vJh%gOrAcl&ezb3FV$Nn^3)^yG#(k4
z<%#mTTN~Gl@V(D-)+L9}62|5!+`K%!WY7<J>Fxdex~eO9F!U?oQe&WDF9IF-b|d8C
zYxmNVe)veKp_K_}JNTU0z;albeXvBJ(zau>{E^?)x6cv^7GC{<uoMPnJH>b|YW5dC
z$?XztJ+$(7p5W^RwK*~F-2JpoZT0C&5vFU;jm-_fME|*)zm6uqK1q`!il%t6?MLd9
z*f091%}IV@_ekxu<?S~Jq@`PRZ<9B!NcHC<tZgDieVZ%N3MpI@7qthU5ZhM(jdWz{
zFr`hRW(OX`kO32`)Ai>lehELizB1LMD&f~PV064^b9|4bTk6`Nk1}CLcX0odZ_6b_
zC$_L(Qn9SRXduQHA~$aKhtJImKM}0~G<rx^>hS5=UnZJx22E4`KAsERX>mKAvchDz
zk~#8qa@9p2@5kP>CDBnf^IwGtH;6wM!`m+Trz*-#KxPTK-_K3I*2k<|tQm7TO!f}C
zH~82b3hZc=ht)z7r}#3IdnP9q{RF)Al9y7Pz2bE#^AA-NDM`0zXgJc&p`?0YgPUlr
zNhn~LEiuBmd?mF=GcEATnl9d>(=tk4_prs#{t<>(vX5;dRONHl<Ns>H<^Op*V;JyK
z@ZL}Dc7n$(Ns^_iGg|(9)alXJ^5vULuj8*etWoEy89&UUZq$%L*F)5MiZ=DSb!swO
z1&D@-RIRB$i9^H3EVZ^Vu{YhVs-qKo(6ONpVqBE?=1_u@2SBGdtGtNEr%DkYV7#M8
zeiwh#y^z!72+vp2lFqAeNW0l`AmclCoHh7}JF{L^1R(+_>dmgse$kL?K))5{?W%1|
zqh=n-m_p^2z(G0{D^^q{drg!b_o(zX*n86MU)%#q>Z@!X-b+j~vj@0H&2%Dz<oCJi
zOKs;SA?9b9yOCCX({Q!%K@J_tuJ2<lnTowLCkYTQJx48oTS86ZChJ|K!pjCK7HDZ}
zHckBuLi?~=4?5rV>KE&4k$gDqLz-Z}>#7r#yd<yndVh)$N!>Q=&fjWmLqswZsNY&q
zDO|r;T2(krnSnmPE9=Cw#HXyfWL%%{czs1X^pks4K$Q3%{f_0_uzYt4^=Ei-Nk+iQ
zN7oj}UyGSVQUe>k9kdR~-TnE))!AAX1%)Oxj*ZP5DO^;W>Z3FMWr2XAy1U0^tv62T
z#q?@{PD$+mhe0xRFLA~tlEbBF(AK=;aV&#59Y|1&rB!pzL_#5h^I2&6!)W%dS~qrD
z*3DiS+~6-yPg8#Yw)HoKVM8uwZpJE-J>0~^Dvi`W$Ao>D-N{gSor#W7<EYPfuhZ6G
zFBy^{KvTHfv{Zqdp#9!d(YZ^EZNJaf?yTrQV9n+rCZrj(9GRsoC3|4(Yf8wjF!_CT
zcXNjGH6W}q!rsg|evX!S;{x^M!QQ=TNlxdp#hat5pWVEWkIVgKX0`9fTa`>8AptTb
zPHu*?TAkbRG~TJpF=wd2B#+pSMJ##1=NT;OAgPk%%@6f|My?_q>?ff%ZVbPqaaJ^j
z*G-^~^Yh}=41>0F+W3hEC+CTqllAIg-%dqdTI0p>uRy->I-bL8Kgrb{x1fYQL(QWt
zGk=#hcJ(j(GoZ<b*<yGRIq9|-M@wgYa6|cX`tNC3EGKKd7}vX)N_B@{*-n+L<5{nh
zQ~yl*s%PgfS-;Dxfa+kiI~F}P^)e0?t>|8LmZjgdDQNGh$sh3|CHxKTsD*M8Tmu{*
z^4Fa5%CiXLZr^cfv~x6Gj;vZ*Xv%gY_S~mNyIPZp$AiSBpFBeSB!y3_`=!D;JFVRw
z?rA3z49T;OD~xx%?q|hp(KHJb-;oHn)p(9*JpbY!D4$-uVYllKXl32~1s+(JR+RW`
zot-pdW@@nII&a5?;wxVwJok#wYqE_rp+#D{bH@zmf-1pxnagH>Q{d*GqK<}T(i(?s
zd+TMP3i58QSye(^tJW`v)2_4E97fD8u6|87)f3V=g%`<i$UbY1Gsl$d>7-|ju}wVp
z-nD;Qxh|txWPwVXl{t}a7a5dlemg8tu6xkmj~n^P@WK=adg|~Qivk96)_+`9v<aqB
zH*u2Lat#F)Y0yi4Weux4J#L6X48NMNJgayEl>ZwHe%!NbDOtIs9<M^Z{BXI5UYJDa
zj~AIV8fxUhtQv=#^uA3UeqcFuKYkBes_26Uj=9eI<X0JV9}GDUt*tPn&%Adwm3FiF
z19cC@*EL`n+s>b@_;8f>xvN^4gR+F787n40Mm9?9-dxX*gr2ADvpWj`LG2lz<8uj1
zzhv{ND63b$e5!x%B2ns+Z39oiNa%-skxmgR7WZN}Myv1Z<qb{9O?wtU@b{ApxDbZ#
zwL|%O>+Z6ubcnOu{Y@cJ{<$`G({k6?zJ{erCE-O$Hd71tQnUU&B~C?7(LPrxb?4L+
zuXzVS63?ke_L4`2p5Gtiz2%Z5JEq_^-Hld*0Rx+u`ax~i0ny|vwcRqp&6NV*p`M3X
z_)<yZ!Ov=8KaUkn{P()~U!2$%TCF-fFG4a)-0@S>VooNFYotjqcl8}#JC_wecflMX
zk7cdmB@k4yP6lDK3b(U#k|O%|EyoRnGtMhUK`73~5yTkS){U{sX*9>o%;@>^v@f{i
zxvq%(?z;5B*r-JFtXt;WC8y%HRdGw82;;ucadwU=H2BqXK=O=t*V84w;G}W)gynIE
zKLtEEwOrFpe^XeZlU-d?Sze4Fe6FqjOG8pZXy3LQNMWP$r>j?43+>xAgsTp7%fJA!
z7f(XD{qgTY0@J{KqRC6gB#AG)^+;peRIutz_;grGnDS2ttk+5U=Y4G-C=?r;l$McL
zu#r*Y-r10n4fkA`N28jskgeV}Q&JU~V1wDYU*8ZKT`#>od(7_i?8}5?C7I?JD7(=t
zE0|AEX52uaSH!!>bO`p=WKKHa0C6X^sBz0^8bxu(>V(jn+&xg?+VR_D7#5imXfilB
zm~1=PP17E<3)q@r+A3YyI)&M_c{g8@jcupMG>J;Ngwm~2aE@|q%Q_hb@(*89vP~os
zlqmQQYfeWMvIpg<(qEk=|H4_iR+ObFOPjo-?!x+R>O!K8(F%S&LR-lpqoC9@0Q*Zt
zT_?iSW*{zwJB<47MRaWQ5OWb#wA}&F*v2^SEh6U;n_TS`KT^`ACG_>er`ooXsb&Dl
ze0^UyyYH~NuVI&7ShwXJ&b`7|H_K)2HKg>L#g`{{pkBSH5S~c2R{6&Grpu3xEwO)Z
zrLT*evd0eeKE@Spe2P~l#wk8ZAa!KTrOmZQr>E2H*r%|TM>TtMTA=?3j&_HsA)l#v
z-U@v-;;ka8^t@&}Qg)!KL{O654~9<O;%%-kE`y0;?Yd&-{5&?QzP^hZOU~TplN902
z+|UzAw$b6a+3LAqdYS?3@sz@|7lEO>s=2F2NdLbaeH#PGx7xE&S_E{7CD9)u__t}s
z6BIlzn%=v_JHzfQe|}1%UO!O6iZ7cwn#BJH^z-7Z7z(m2@^eeYr)f}~zIk|O8UYEs
ziJS#W2cOECO<yiC6$M`dyGliD8oPrMkXCMGE_J%btjC{KgU(WL)EHM^B;H?$e*#D`
zWZkjmVot-m6?>sOn^m!o@FypSig#W$)9oT=z}&WzLR_B?NU0dAmztH7R!w#?HoK%8
z){I>y<*^DZ^x$L_iM@kAxAH4P{Bcv;Db}1;Es>Aj#Em3cv8AVT%8n;Rg*H{sK5wl2
z;6TF@tVc4W6oPX1%B;0kcq$7`o^NE@o^7$^e<{-kpJAlyj0-2x&ZW2~CPFJZq0!}?
zi3&oaSrG?A2oE31kon<6Fiq@hH4;n}W4)6JW94f7*janD7Fh1OXq}Ja9H|zl1`!?j
zAR3m`PoirB&c<zGdFcYZAD><@H0ZI8(NO)ItXdhyWVr!Mnrw%6+(jXWpCXMUS#OR1
zy*67+nF3YmZ$8O=b$f`Yk^oyu=!TTc{;CLmKjLc$l^fO*WBKv}s*)l4Ag~nL@gZi~
zFU~O9M!>8(%LRoZ8mEbG3|BOPK&dPg!ODFbD(gS#!J-L}?o^nnkzW}kFvsOO<;c-*
z3L|0OVGV{Oo>{>Pvys7Edj`bOzOZdcYl%K(O}S8CbmzIT)vJ!GyZk~G!#_xRvS?{`
zEvI-R7lv$yq^6f*TU@i9dTktEy@Yk4qI4lty!@Sito;l390kR-n=FXhwlb$#R0*<9
zFJI0DUeGgi$kv?IqDUVz)0z%lNu99}ur3zJ*$OkwSg?*y(FPEhALDM<%@=H_Z(J+$
zK;^K5RRm!nZAt!WJw#cVIdc6@vDwGPCF_o!@zoeF5bg+K*(hR_uk2=O1j_*<cI$<=
zMY0o1wHll}zu6E#162x=81!0Y92vig175i<<Cz4>%oxM%{lO7;wJEyn8Hg*<TUnrT
zJov2Nz1fV<NM?h6e>1q@MUs_DUKAJ#PjY7SR`*`u<xNqul}=9A0*TsS&*U!|E?ADD
ziSti3IRw_N(`(WirL!B}7fk)tBn@RjJv>XNDGL`xrGgNaM+?YZ#~CC9$6%&mBCg+Q
zrnET1g1uUu2}J%G-=5*dG9M-rN-AlKj8BuEa@>`cS>?NIS&fRq*^6t_DxDs@a|zcJ
z3b^`IGxAwj@95PLS#)cZvL1cqy!+^K9&7QN;%=t)ncx1S-xQ8b;pfr#_M_`(0yaIz
z#|dHA9PJCP|8Kh|ZcKHCU&fFN%y;?85|dGul=8Eg5`{}dPADebOQhDR^e<iS?`EG%
zoQBT`Dkbs$>^{QU(s{;q(l`Q{2gjtN%ab8e`Xl&D_S<h-Bk|GSyHOcM&|ZRM1DRJ!
zeuv;oz*5u2f@|Z)R8t3BF-t6iCZ1LK35a#HLUC4jv%gFg&QE~QfIB&mbnPnlv;udP
z2ZaI<l~DZVHeQph2WU&^F+F4(NnzAZ;$n;7?fh_LC2j}PAgk_d=(s=i_jTJltfpqL
z_+`+gH7-8LxL=<<o^M1OmmSX0ZKr%;i(~Zms)gFtq{N5KF>Tw|1V!iXpiW>WX1K=6
zF*8$JGIFz|G6qfN${-@uHOWX)GB%_%Hr0$_49EbG9y601k~T4uq$iVTG>yc`3qAdg
zPWpU6!4^9Cl$xgCQJ9CoTbL@@vxwe;Rjx8FCnoUoKnsRu)(<2t=doM^b$%h0uPBo?
z#H6G0I^gn5Qi*QgVxwg(vgY#Mp3qh8#Z_KM0ly2FnZd`j8;(D2I;kJAB2%5jpo0~n
z-AZP5nmdv&apw%KbGb{E8^o(JdaK-n6#e3u@R<zPc;g{jBFZ`v&|kGbIEY7ia<$-$
zG|wNLCcHnAwcZa6`z$0)Plp-UDD_Cpj|xnXc1Sefr>vJN(txWmSp?ond6&LZbIxjA
z%%&Ij;A?7U1oIE>?)IBlvA-1ejDGy4@B`xv$+W?}*(x#jT_x)9JJ2DrF@rpOf2+V`
z%qK7(+*|~ig=b)hSTo#j3XLqtiP#)*r=q#qAC}@~iQhkF^9Y{HpD%hJOtYt?aR@72
z^l-6WbSmbiInj*HaX(J-My?k>jeF7b`JeY7Y<V7ghQ~{0#~n6DTUv)^pdlGei8%!X
zQe!nXXwXr{inF`GM2Q%cOxmIZ>wC*}&z^lk#-HdY4a%7K1anRTKRQ52HxprxjJ%~6
z&86#V8@F8n?;Qt>`Fdiys?0b-O5i`Na?Ys@V_gUOOvF+ud1j&G*93^*sgOM%#piEJ
z0ty-!T0UL=IWKj7LOWQ3nh@k0=Fg$gg6eL5vqI+CpygR&xZF~GWGUYoztzIXxoqv$
zJlMwh;L-=UTpH=vKMaWq$e$x;@zPI5-?2sC7jkOUUB@&+9@nPwM5RMs$Gb9C4*SeG
z5e<$cM6$ptUT!&y^!;k?e8izN+2%r3LyohU@ru4(RY3{<FJAZErs#VZFd7F4HPeLT
z4u*a3m~aPWh6M(G)}Frd^CUdi_SaDHNtN>2&4;NLSl-*E3C3m*hZ-lFdi|R}Y)FAc
zb9n{$So0co62K}?ucE;dLwIoZ3Y)Hnwy<5!aAUI%N7&uXhqb7lthk+lBr|)4<`)Tg
za11)Sk(CR?qoKia*fa#1f-dh?!pD4>M6%z$3h2bdtstnK4z}z9A0E84Vwh`7^NBYE
zY6Mm`SXfxGPuf&x2nrd;gHl6ZI<Nh-+Jyhf$f2H(tDLxVaqytzkYQ~J?>IK(SCpJK
z#r0=w&tsfCy~*^44h*+miAqS6QiAA93Bk5>5H$?wJ*=Mkd{T);<Z?S8`pr;7odN{Y
zi&8Fem*$e-QPJ}f;TcTUx=z0?(|tVR(<1xE`K+r7p+yxb9z}SDSK_~8R4N<-le`hd
z$V97a>#j@Y2~T1S_%uGnG-JBQ^Ff^JCmjfg%*g5ZFupG9fT;9Yl2U)-t6ArU&qF8?
zsX=tK`t!MAkUC)mGdYc^sdAZ6vo7pk9?t<4-&LUd=hyLnZ@nfS?E9PI^P%bhSs4G$
zlw#x~g?!nf_Opt!ebCIUdsJzCxp^!i@s@i+uQSIUr<tWpJJk{2qnnO58FgI&AsDo0
zs@n%pXmV{FNJ`}+JHK{9rxtT8GkQ9nye4!!;gK2;{)hshN0WK-8b?#}D7m<FUqr=j
zK6Brr+kUIIPaO->7AK*T>MC0VqCppNZxzOSfUZI_W*$PfZ;&h{=WXoZ)7t4<borJN
zhhp7cIVlE(BlS`PZ(!){0V^3<F(bHgBR{;VNU|YTCa>{}2Nc;`Rq4GQhGW^i$XR&I
zzw)A5{&l&~cQsdy_h_p7@#(BC4Ziw(6St(9cb|^jgQ>E}mRL#vdyW)i`6yGrwY0^3
z1qm;8HHSWpXS5T~z%r-iWlkY{%>4=v(LfK3k54c!1~#Kq7^(Y>f<m-a@2uA1XH9s+
zGb*w;^Pem?OFTzTs67=gAILKNrg$Pr@rd2vo@!%ThtH*)OxJG;xtjZrpdbIq5Bz`A
z^nWp1I2wBOV{<f&^UTZS-R6efOJ})MOe0Hp$#m?8nZA92NoonwD&5>Ws}CaU6^cks
zoADtRUS4qooUMOmNK*destK?$u<PF6Cr?hWJHIL5Vqn+#{aE)GE$B5D8tPzrm%Pkf
zmloxe;ocgoP|*|OyPX`_P=#*s7OdWIJsl5^f+aC0-zfyp6lA=x=_<t7SnQ=>^udwk
zuY?Yp1oih$3R2tN$7k|%#I4yU=}^((7{_Gq$9=9m=~=P`oTF^H9;o2vaLT4uHJR|z
z+66a(9-)JTqfSteBAv8#^;F{LS`3IFa{r_A&*h);zJE3yObI-N*jS0xe+Jki2($H<
zPl=)qEq6D$rZ6@oeGH)-9^C-?rH!Enh^UV`$O*Dt$zsed^yKBzAnOanO+PLU_bJ=i
zK`C;M?7R?A`I{no!|j4!`P!6>;ctqF+D8ZM*z<R9*fM@)7K-@ehM~pO?<+f4)nP>f
zMzsz4jk;{Q499{pgMAt)XDKVa=M<J<Zs1uQ#X`bqeXd-X6hI%Fmxb&E_`N_$53(8y
zqXvZ*1(x%Sc`z1caKoPM4_;*)$pBU(XOZgdY+bb%69qhlt$cM{Tx&D4ZZqFip2<W1
z=Mr-bR{P6%#emBtsf?f36W^>vl8+~|H^lzjf}Py&M;@MrguUblXOk)lvHa=H*_r$S
zx5K78Udyffp&|f7t~Cga(9&{t4vnzw9lT|Gy6-&PBl_{HRfB{JZ>nVi|5p;?zMaea
z;u-Zawb+@*Y*D|m&P%fl18^sJj~rXL-_5>ByLY69V{`K%N9(r;&A!P_mt=iB4WJ?f
zaYq|;8VuxGPd2@d5gA<xn54zB`}F%z)(Gvl#0v~B)ThaJHZa_6j-Rf6*V4dD>sO^f
z)L9PYw#JX0RX&sNJrt?;e=;sIx;$V~;bhK|5^rkJVQlWfHGWf$)HhPQc@8w{tp1DH
zjqEyy87K@@dG)j%@#;54r%mH;3a*Ww^Wz_s&kr@pV#i7RK(G=^RT0{*=L(-=A2s-C
z;W6{Tdk&`OB^Nx63vX6lqiVVjMIH^*l=T{L-qAb8GKabSVQj3Yo18(1?`|{)NrIO1
zIxmf1+>#>-&hcDP;Yt?5X8i&Ul_Ri`ZdNiVwUA|LnmyxcB<9rGisg}I(F02)BB5WC
zbrKSgy1kxx9uk)^bd8t8O%Z1GfTV5XzfszGjLw5GiQ_7IVC7cp!zkmq!Pu+?9^FI~
z*g!kI-`3SGA%ePH!GoJwp!mKH4T>X!Q8BLeB6C}jS1W2q=@8m<Hjp7D@g5`Sr#UVP
zdd`B=tE<07uFU-xmg_N<GN{$L#`j*=`hJ9RO@%C7@7!;FWFWl0Ub6fdM(uoF?kY84
zWt75oBa&`=LwibztRFBQLF({Mhh1DHE@&BEQ!*r1+MZ*>e*zn24VL>!!B~23%Zcq3
z3xq!_l_k#U-wWh5$b=y&VnGa7kW=f3i0Xc9DD3i9i0j6zN_%ecH{w@(rBQq$kQRpA
zT`~do=L=#@4}m7qW(z|zXl_A`ZNrBs5khZ<Bu7yi@9Eum&llusrnT(Xy;1Qnxdss}
z4b8!k>Hz6VtA@mit%oO1&;%3<I?r{eG}r`2D926MU=3A&xaVi0b3#85<8(`_=owp9
z!IiDS(iWz_hRkfpQ0$+BhZ0TN=Ksm)@Be)I7-i5Kjr8T#p_v=5g)i0npX#4;)OH<|
zQZ9d5twT2^G>vC&v6^^l2MSmzGupbbq|w-T48$B7GlM!Kg=ev=p{MlNkiSpOnxL>L
zk*;!XHk4qSd>^dh^<Z%sNG}zsJEzrEC~Fi}40sZKhYt`xuI*KyhqJ-l#5zBbW%{d%
zHvS82Mq%ZnVj)bVu^jqyeZfr@+#&+3IMz6)vZhjqt7UfrOvr~Ol1#bFn%|I|=hSL-
zl<%*Yzo@OHzWuH3+)GLGH^m4b<TphQ_M5VK6*<Byc-nrjn@-j_h?-GssapM}s6HvX
zI%%HzVK;G0eBNa=+H(HaxU*Ku1UomkZY+YCjoOe~C*#9F^+4bBb)zvrB}-`6f)7L1
z;{!{>1*y4)?wZAx8cA9V9(Omv3|YKlIrMyP=5yFdQC0*b8ItW$1}*i-cs0PN^G2hZ
zb$Q|LJU8ii!;-GFYV>M8D1o<-MOhjbB*HRo<)ItYtylYzT<0ITc741IoN(;-(x$Dy
z+?#lw?kLd}s>IW?T=JYJ%OlfOj0%eHP8erSd_>>z&>?;N1r%*P;UV6Lg7QwM%DgKQ
z(1Fvvh{925CXDO{m-h@3HvJftFm^$>zhTxkwBHM-xPpWlm_%|eA?{p~7XCHty`?&E
zBv57I=Q9xY=G8--{{r7EtWN2z7ZLGU`rd;PQTN&`p%=$YG(G#XKMnGtEia?-Hs6@3
zQ;a?GMD&t)2l$ykEUFqolXX7cC|%}oJq61vX?<H!%$jezxLL@hZX_lxNtTCWc?C>f
z1U9<49}aOIsK-KSc-bX%Cz4fkC7$L$#qnp*C-`!1{hApWpC{GAb{X^~8Gf7y_&8+!
zo47s0IkWhZdcskUl>^m>MlP*aBxyhK8PEJ#wO{||3-~{^T>EleabnIuy|e1+IHUqL
zB<#DWwb`IY8%~?VMM?OF-KRI%?sMYZ7rnXOPD`@hAZR!Dd0|rd2qJORGwLdKKJm=(
z<7#-+tpEF9C;H&oAM~Cuuq<zxJh|Mb$^{zAFUf5hQ!EG#<=ZtopDN2WdvHuDb<9lQ
zM4$0p&NFieS&ezZ?w%vfY}zGnz9yp~r?V6!4@8Wd5-mnFWyqwfy%AdWabQzLw)@q(
zuz0Yt!I8<F+&iHs3>=ZLS&U~%Ll{qV=tQK@Dr#$sQ<6}Uy6{|CV%fFQWjV6ow5TW6
zE?yyQ1s@`vU3UbXLKM4!+pCTs$chpBqM81h*dy@zNXN01udCg0>YxR~i2k9zN$0lf
z{*Nu#9%@?lm@G@%nB1G~IpZ*s+s7xFwe0$J&fo|ZG9?>1PEl+12UGXfb$1l|Fu4&%
z?a!&nW9~$0kO|r4?8g#0M!uVP#a=#gDMlnuuHeGYlM60QuijPSTak#LF04oQ!#B;8
z$!>N_0bai;S|YT6Q_%cR`<mS$?vKVXk^967MSQzT=siZ(g!$hTJH#7S!@FCu!YTY$
z2T^1669+L<<_ji6kIaU2*LvF$j#&5GLxP?MJTi>pcx;fQaIt&&B|m&Z|IX2%SRSbH
zXtSn$ny|b=*vj{=^dlX&W6xFV4B3whx}ar69y?PCDv3wAdEKbUw5$F$rrk5*VUW;e
zQ$C3tl~@?k;)(5nVKzb23E!}ECFZFQ`xY&wQq%~!t|ci(S{E+N%sA}_)7}d_(vQ>i
zaYmGz!)@#uL^-81m!YM-;#+S%C5|D`5L~jX`AK)oda>j7X3a4a+zk*HXKZLKY12x>
zIwH~0Z4k4;I!sW^>J=Xf)uYip`lHI<ler<B>5lKEXmfP7jY-V@ep!CNYPn6QWD-h}
z?TC;FE2B5{CAC9_6j=kmmtimO?^3l!;6APTvz7&qTns<(8~vP~pglSIy-nWCNP&iv
zriSl3aZ9f1V|JtiYwojFjs5eZxPZeYL%c8a`pc^r87;%|hO<bwRcE7$wZ*Bypf-R1
zBPTEd=xYWn`*{Xvy9em2vz;Bl9|UY=7tdVX*<r8RtgRFNO_8cl!7d#tX565$IMd8#
zXwH{C1Q*#Sd*@mt$y=<KDpbrD^l@Uy>}FkK9o-knsfAJGwV9Td$>AR_I4zv%)u=)#
z5>3Wx{Ts=d3~(-W4MZSv4keXW!T*3l50Voaw6vYt?okI(CmL$`|HW8;ZDiqg8WWn^
zmvYs1n5}`<N7Pj_OLeffl}~f9IrmL-M(PdJ{;fWN{p=jbyV#6OC}UHt*V(F_1+QLX
zOU?T`RN|xd=3CnpUHUv96U)FtRaGfa5e*jhYN+YkhFx~%^tzQA-kB-?p8j2T0iu^U
zllw*H^stbPJ%{>+HLC`(9|OQW{!st8C6qCBkpX{iO8<a^7dKMqX0$)_;)~2qk%gF!
zQfSO(VY{aTC6=1H{~e?f`zd|aRI7hwsI|aR4P~;=$*RY7tz3?evLwi>OVZE_Xf!dS
z(45%?PuIAZv~^ol6+vV!C3Z!W9Pkc%)lb#uC(QU=GRWvy+qkC@K0viuA8i{J*stv@
zT+qIuYs>4_&PS>^hx8s{4F}^|Z)!(R97w1#KfJB_eMCZrvy;lJDZ-4me#L^Jc~B5T
ze-{<zTRJ#gJO1^CZ&`*F=jJl|UyV6nIf7>s*=R`Xdfs&jE=fr)t{;e8E-{abXI$Lz
z{M%MvRKo`?uIx5HP_NFcZQG2JEjk{O2_4UFCy;t}TGzFIS-fbU^1b!!)qm=Z|H<b6
z>O6Ey+4=?a%i#-L4ftrh`PAml=v4J7LmyjN*@4Xg)+O1<#}B|-e!S_Kp^Z*Jz(3Hu
zf2JA9nme1bV7SLlHkNx>bdA~0*@8SEll2VLF25<>Zv3FhqwpbgrxG)8^S6!}wBD`K
z?9=u@A+{>ICQ=LmgUbs_jZrOKTF_s<o1xC3*7A?jZ6`ml-+4np(XubL*=N7lbw!{M
z8zmpbeT#0B*qg;aMjDAbVc8q7KTCGD7~p6?qmOj%hp#}jA0RoQ`$@K@-b?h+$gEo9
zBDNgF$Hf|gua1Fl+wCKkBe_|-w5Bj#=iSh`rp7j#NBp4aj^gLtcnN`_;@7hzSqXlj
z<M*gK{iWe}SCOE~%u+GityRaIrwcLjyqhYIfTKI#Kuy-~!1Z7T;2F?=9c2<}{=kUw
z5@32b#cby1r?HPW<nT_cUGnT&Ki7Nr>9uocp-Vn0dq+&{_=89B)rWAk+^fMMvgRGP
z_HN<gspW5q@zWU9i}E_vQ`)BXBl<HLlHG5LU&%8)SHA$mSMTm!KE8UddWpc1rP}@@
zl<hxXh3m~f`Au<9OAbp&T8rcfbQ3~HmDky?me7`l(%=uHC=nc3?JzEv-7b4}34s<7
zkuqH`GXt(=2QIr~mkIer{klA?!0uBYW^E%R+sc|@;okdswx8;Yem3e~7w+$e7|c%W
zT=JX5k3^QbTel%5e82;7L#i7n_|A@E6^*bJvO+WLQ=$z=BbQ)t7QROM($m}B2m?8j
z9->n5kq{33V^_sa+Q@an2Ib#o2bEe_T%I!3wDmPK-2SB(9MM|!^nYBAHy=$6!SR4P
z_bb23mX3fEdGwi|Dd7{qld-mY)kJupCejed$P3g5`K$u!nrf@cBJjdTij_`!$w!u$
zFg;mpqzcDwI|H_H{>P~^=m?A|_F1)fU*rHB==xFzgnL!3uGo<|l<;BY*oT>N&GP<R
zm1`E58e8l9oeeDbQ<=@A?`2z&>y|>rM$hNZ4NTsmc-YEpzF@OvneP~kAZ-Ek7bRiA
zQBg~!O`Ay2W~uo6pub79r#;M4bmfOW{5J(d$1z>c%ZEK1-bK%p-j|ZNjU{%JUH$6(
zy!k9TGrZP~{Hdy96~44Ak7ztn;ulmXQuYbh`gYauO)Q2r@@sY%W}~GW+vo+K8!}XU
z5tmCyTY;S0Xp*qvKeh(3PJEYHVpFtlnPqGCOZ&5eYE|MBnH&j^Xz-*=BJTND(!+q$
z{TiQLHxtKU{jseLY-EYcD?D%-Yi4Q!Tqp>d$&cC9{M1xX`3X@9|KKRyO+N7xJ4Y}o
z{@KRyW9;hyn79pM`OH36>r-*J$nY0wES-IOF~$_+X4<$-SzqJ0wL&V>%a%zRJY(l#
zP8mDwfbw_WV}{?0XxXhu%L6H5EpX%~pJ}6bc}tlRe4P*1PUaxG9ENF*)oebV7*dud
z5X;`zwfYUlV;lWKedsyn?X<Gv0q2&dL6^W8)MUkwbWn55&y7k~1-BeW<*b?oYLS^$
z9C4QK&!%j(ub}!vfdSFO(&nayI~8eGlvyU?&E4OX`YTEB&M6sQi1GYZI*8xS!mRI8
zG{ew_*|*8J-yX%=BzA~2RIS0F+*+TBGXZCx3LY9|gta-Te<6sC*}gI{5MvZ`v2b0`
z_Zibv4#!UX3{fS`pxPndA<fh9YH#l<S7H@Q=N!scO9r?d#HZ{#Gz>wL%#=;mlyR>Y
zj4I1UDX#t0jb-$;dW!CLbQXvs2V*+=ST5m`T2>PYZk>b!bNv1S2m)FKs-gTE1gLbA
z3%GB`Eao+_8Ajfm1^aUr5q(Fu%fw>KEbSiMVc&gZW6%`@`{<nh-sffy`};fNA{8OX
z+x_phB<=%Hu|74OjfFd&c|8TY?>9siclK!f8QaQshE(RqCb18|EFxKoGbu?$D;t(4
z`>SzV{vKzj`b{Rm%AladGGE_84Z^VUzGT44HniCQ`#w4m6;z?kL^ZKu21k+eHh~O7
zA+x2(77R+9mTg~La-`;dKoy6mpw*<JQO0*A(MlgS!^aV^ci0LT#^<`7<#tH76m-CT
znsWe;e&Q3Dz#Uy)_2fe#avfLQa!4D+q^h=m{>}M`LtkL<_|kRtNnhfz6Ij8Z3fKA+
zFRgSJ$;H<iLb^*TlefffL`Rnr&SiE@=*f8W&gv7{cn%&{%Z9Yb?*vjS<fseoEndG9
zbQwEgm<F%1#sYDvBlKOSX5vbLZLRH6{GYP<)h&pz-~Y;3l+%|Nk2zZf!mBr8I6IOL
zxQn{W>&oxU7hwc#N}db>{L;}%?lFcvCY^`j47p=P>!+>gHpPy6dO&*Xsj69=0*pE<
zI|=7{d2R~m%Wnm?HT|d)YIrbaQBlPavsI%l+eW^nWVTkydPG*xUg5wWdqPY!F_`01
zaDf?~jMFSZ^NG4T02|2pG)-~aP78|%hlWa>XCACk@Hbv`cq)L#n7oA_JhLdy8u}bx
z-YxY|K8F@0U%XIm7Ryk#y;VGJS3~C<7j#zZ8wCe+pcsxmH&rxZst!}eCC73*6SV-u
z=tkL{th=oMCzccVEa{Aw6c<lrPiEE5wU%@>rzJ+GCFT~B+8pPLid~#$+hffxf*aSk
zsf6xH@=JSE9OrwsDXhYi%(nc+D~%wE*F}*}20EHe0p_bsSm1!a@O%uov}?;>5KZ=5
zmwW)kdJ|Y@r8k29qnqXw(+nusv@70so;2WqTZ7)3e+vTL63+nUKF<zoZRxU{W0Zxk
z(>t^a`1a~Y!WKYFjtcrHmz3v4u64>M3pr6l3s4ZNho?>)aH>fH%ed7yy1fD&j0;uB
zW@3~EsGD}+%wkU}OtMX93k<Qqb}P6exU#8~FS97hH<9+3TDx<|!pRo@W9K{Q$Ysh?
zCT4HgY@}eO2A)|NQ;ISo$fEfK(EKq@iJ(sI7_B0Yr68i(>=iHCQU(~ms}lAn{o_7w
zfvNpH@nx6U>iHZ!pd>>w;>|(+GYkJ7feJa@bd5;)x?|3D!g_2&h}RF+?h(f0im4_k
zpvBj1XiqR;ynZlLVK)~u99st_O@fA-GhA0So7z$yY2l87HZp6@P515TJacIqs4+Nr
z<6Ia1$3e~U*^iR;q$J>(r)6BW!Ba5G2v(9`iR7<TSPX~Mk5JaL&<%X1xV^KO*m}|K
zps3U9PyhD&&v~D4K<0NjyycYz-!QM@*a6B@v9Q-`GjeH=e~J58i!LMRj=ke}KCA3{
z{5NxQDtt)=@jbo05{wEq7#=;oG;h<>Zio(+iEYth8uyq*&qHyId~>tyuFFHpMutJI
zyp93!IPXUD>GsO+i%YBab!m}_j1u5x0H&C4udEb<GSdvfvoE!?{i=)T_gg*mRWurn
z|FWFwXGu7z09~3pfz-9Ov73TL(*j48yuo-9+$o937n_fvspycjkR(5b_T<vk|E8c+
za@@by0Tg;UNbuc3G)qe9D{H^ZA8I8H$W0KDp3Y;507e0vh3z5ZMvS6HXE3mPY_bX_
z(og_FFhZB+Cw@R(Y-?P>DD$nVG9+}!M|w!S2ZljF?{~(={f}pd|6ew)xveaWe<UZ>
zf-lWK`zUw4J<)Z_ZM!LcAvpXDCVP8fTN*?5o(ccc{f*wk+wqIf>*tOkD)wjvTy5K!
z?s!NBfJ^!Q(O1H$(9_l&4GgR=#nVkZf?rZaeVbLfs-w9Q!BFCn_fiq5HW%<!^4Qb)
z;`QmA4lrM_f&iexhteJ{Qq6djQ)<BFuQLo6t=fe(-jGuEeplWy$pUu?W*>cFo4G5a
z^_C_czLX~yHdC(-I1;<Aw-yKbByOMhEX+9Cc&$=Ma}F*sVH>qPl3HA(+#3}QLomI~
zq)~19?5^@>%au41@UcayBVm)qP490!q_AbVvBYc(y2j78A>^F8Bzym<X&(HZ&T?w)
z%ALTYXYkPCY0aC+S4|N)i=&9jo~PyB1<XZn+3V{E!JF>-U*d?+k7R05MY2)yi`c?~
z<!6tvud}32A7Qc_eD#8-$G!oVtDiKf`)^kFEyQN<4u;({;|Nsy2v8r=5v`qnWL!`S
zDc2#pT=V^U!vATLyKZCWW6jvs>ZD=Ri2m7=EoV)?^fT~s=!2t4+mWOLQMcY-6o1p)
zA<1Exst4jbhNir~P{sqY^|+apB>986tgwHWuH_zn5i_0i=W3y2U8e)=CLP5hym#&f
zc3yugTIzb9;u&mnoA5+Iu_}3Iwyk%)SO@+$GE5^glEqce(B}`S%B8dqynI>twcU67
z&m(#ILnnlAyiy-K&2gej?Ja+o;48Xz!*7obiAKwCXPjEw1RQzfuVfDAbyx+lRh8s&
zJ*NX05!SLE&W(ahGj|(vqNp*8TOIbVZN3s=Q@#-O0|k8?*O=0KtanI(*};3f^|IN1
zHG2PRE5d;V{Y7$5*tOPpn@(R_mlWD#oWHDdO|nHK+WMP<eBrdn_in{jG5sG(y5GLH
zcy=h-`Y@IjSIbpls4?q3qO5c3Xrk*9kO4y}$y50F^o+$O^(uacoZe`D#r_RFT=Jps
zOfY<`%Ho#i^c6}O-=nm)X<Vg|R;%;pTHet|Kkw$`jEN6{t$G0-FhxJqkD04Q->Olj
z;34`(d^4hX8KHR0RO0lIAb8O-xUt2GLB=xt8e#58NYjg*>|d{T{adB_?Z+D#JcWs0
zH8jHA-}qoZ!Klxwl3)PE$!W$)%`reA;*~jq{qE^pNzV>hA#0Q2M)Q>|5Fizs47%M~
zQd@<vk=DxS%lG4RW|=8zRc%_>Gp?jbhPA&&+P`=ITK+Np<s>s`s<%Cw!5IPv0yQ<k
zyL39zawJfS<-4Q5)e1GOUus;>tH_0F1K{>~55wMW1m6=)VR4z-hN1HcXP2iAGp1%2
ztSgdvN=-Lzg89Z>NpPJ;mgU*oZ#2Sha4?@IegEZERVTd@aHn9ChfRw^qFjhSoYN*!
zTX1Hk(omfso!zMwt<NgJ9Wn{FVlK`bdft)+glCko5@s5%BHWvuw6mtyEVVQr)(&Jm
z0bT6%toDa>lvjhPW0HSUm>(!a;4=HVe7LRUeXbz`{N1(u8<o@78Q*R?m@xXIe-^!$
z(XzaZ*}>py4UEEJQ?D&6rKJ1anuVQM(n|&rig;~}K6Y*z`-@SU42Sc-7Nr{j_r^+q
z3r#?c9&uZKC4H&JCjA`sq^uA@z=$zf^^+!dh!-K_{(z>oW-&cyI?zaJ&Z*&d?k57+
z9acQx!B=jyiWpkXkB{4;uU=-kwlsohyu-^|%fogaeMx{ierMXmqkDbN%|pK4ZmRt8
zV_|p0A5x%zX{@j^5HFV7VU#yMr<!0L)Ul-Aak=O_LGT0xb)sac5VR-GenUvt7q-5W
z3i%1?sUOzSZ3uD5w;t4!)qzABNr_b?WnBg@`X%uozW!3rjdKEnR|}@cuV*+=9R5o~
zfwt(1jCg&#-&BaC9AWPESINOTp13)eot-5aA##~=^1u0=n+D(BLH(mlxsS7+`SGWc
ziO>SYZ84pZM@nME%yaEm0eQ{vMq*cQ|3J^!(Lht`qE+88PWE8Wh{+2!H5GO=)c>o}
zsnjj%eNovqAg{`Gs$`3Wh9)oDI~Kmx*AD1wHe<xZh~SlF1rna1<r3y~=>E;@{A1^T
zGCLRlH?woLf3H%eB0H3Dt-3yio=<su+V%BT!WMdQ-p5;MUIg!*%~f0)p;N;L=m#(%
ze<<hqjcH1m03K+VXg=tY!gHA7^);nBL%6uOcLT3Szx)?Y({;<!==|BlCLnH;BWXjc
z8n|xmllJu;L8ZWHM+7b%^VW&6tC!t}rbdQMF;(-hvfl-=JtO^xe)795gYv8w&)L|S
z4z=VAp0u$POj6$**e84r%b~&8%~bC~XzGc(uDOcQ4~j&hw|i92>#EdqAE&;n*q~8x
zt)ry+uGyh%tNJusqh;7f=9me1+&EmhU|d^L63UP?gHaTao-lLc8P-Uxaq{<4`8qTX
zg-a9_9P`b@e0?EZ9H~tW?b!*AQrYtt!}R>+-#A$wYYKgut3IQ6GPN0>3HGbBF*QFn
zFjdnGjxNod^5c}`r8f@JFVf@M$PBRlt3%P-il0^AE!?d1^S$Q~;n$dy-~#^8^0C`J
zM+SV?q^X^XUN!=atDKmt!vYJ6sa-h9h=fK5#o1VGMSlGD9!vRV$U*OGQcgm93!9A7
zsb2%nI<3N~j*tvuH=6pml8(_`TSp_!d)eVu2uGuMB}uCO)c+UFl2L@9sFNz=@DroI
zojsQafCf}XIXX=m<KCs!i+x76YWn6Mj1`t-n~@5&yzj(!44ovKy0%VMPSf7Hc{irS
z=J=Bu%q*4P4d{O|PzT*FY<InAbbSUdLX5G%04W}rCqZAPf>Tf$1=q4=y*fNYNvicG
z;-p>V{^N=Am@2&`^~Sm$iMT4UL_S*=!+6?4Gpt{rXJS9nppU1{@-isHu?{S$jW@ba
z$?~OmS2CvO24q$z-l)pXPd6>gSfo*H-^tWUDNa_kr<g-iRKv2)h*HKZ&D#|FFrsn3
z+Qm5?sn`CJwc%?!$ohSKTg<Cq*wl;ZR#FRk#01#KNv=oh72jY8On`hKb;d=He0}4c
ze5{G0%$&{V6lqy1_p#zee~e~3e+xT0K~(;-Yy8nNWgTLiD3zoT`YF>De#^+7mTm?2
zbUpgkAN@I-`K$>gm~%7_=!XrN=cEHcgYQ9@9N55EVCIr0E|g^IK8$q2J{7Sm9fSN5
zhnCw`eK2InOib)RulK%Ae&GO5H+jV*Ve^6d#SRUsq>M>V$IxKq9nnn?ytQQka$W@1
z)Y?9b*EIdLoA*S$&XrZ0%OqOO7)$aZPqz0)DgP@X01ArRViRp?P#cbk9SuS$khTNT
zz@g%mxw5@7Hj`ss@gJu9x&(QTQ=5b3PiKQOQY>>8Jha|iBSDEpOocQRwUq%rB|nRs
z4^EWls2+Kci2kte!`bF<W8)$BSoM&g4)oqwh2wOCYvyrVDA{zsIdfHnO7WJP(`HhI
z74awIv9<TsNZcl<d8<HTrN442Jl9H!%ugsS%>j{J_vayOuQcZ=D5gk}kQVMt@Yt4}
z+xzE(lIf*j(3Wjq_(=u364>vBw-Djy;l^Q+&(Ox5;1Q~a_?2!y>dLlhSKW=Zid*AO
z-JwO_O#`IsPSauIGqTbE6N?GxPWxF#1^Aajn;hBrV`OUE>})*iVM8Jy_v0()SBuWp
zl`G6AbzGp0dhZj_=?VWEyP<k|4JDhY(>*<P=AP%=Oe}GrN~nb!8Q`m0xmgN6o^r=|
z$K>-7%+C+21^#K)<MXkZv_8Rf$jnUAjJ!&k4jxL9Qy#9tgM@jvr8WP5xX5bLvgdR6
zbI;J;iFx*4b(1$JHi~IPzJtQ7FxHsAz%c1IYs0|)l!z~61VewWU|~ajHNwR{1v9%k
zMwP#iJrFm(H0W7nQYy{aZDV;aIAt3fZd2S_vl%}Mc9qmYAK`5>1Q{ij>2$z40-`S>
z^;jNQ?#)!B*a++AoX2-<nYz12LwIkbfhWhSNo(Le=v!uMukx@uc1&593#RtN>+%wI
zFf=svK<Dv_s%T#kpgvHUiI)u7Ih3%8_(Rz@Q6m8DJ}Jbc^>i>p$LSlPvNzTwdG=GR
zz{GJ?O9RUmV^h?JznxQIJrReZC?N?*g|Q*xRy#h79Z4uwtR27f!A`ha=g~-pg>1qB
z;mYg&doG3yZzb>m+mb!Brh*e0$VpwP@29P2P!A}|2MJKOEzD2sVC@U4_dV>!sSVs>
zoNxDeHOwN%tX_TpnsO7WF+-i#!<i59lv?s0Cl(Hj*=#g6PUJr(>{PE?(6$=#<L|`l
z67D*t=ty<lHRn)eabohWDtX+k@i28bg<&k)E1k0uf8V8DKrk*M2@maJddj5oBOrMB
zECcgHmKy6Zl&w7ZFxg=si&&*-^+$Jzj2SzVxJO=IkAs!U7qoKBBS#IrhpPuXjcazB
z<_}9do&^+b-F@*>1`Xlk{`>nJ<O#O5OE{H4`ngjfXmp%*|KpIKmWFEz*py55c62A^
zQ}Qzpm1CLrn3j<S=+n_lrCz5MD+5ZDqrW*bW9aK930PUlkL`g+88W6Pi~1&wVuPRF
zeV0KeqQ{hq6vwrtZ>~a5Z{hF&+VSn{JiCUfHEa^eF$+zP<X_@G-7uAYOopy_pte)8
z`e;x9<N)?#PRJW~l{$|tOP`7>P{i_+SVm>JPPbPQA_*?Un!T1*FFi$$l`_H8mj}(v
zemA^{TOqCqqja_Z)m8lO?SD9)Z2YG9Yc)K5>Nmyhl54roD102cei19;_TD;^tifPs
zz43DS4Wi>*EMZ8gnIQ;>k(^f0@AokxdiBwY6J^GiVx*X*CJfe|ALFfc-xjc0Jn5+m
zaeFjaBNTHH$=HwCs=W(8Lz~cV?h1%3N-H$>e{yzd<}lSsdIb8pSG!0g8<M^-N=kYY
zw=1jKp!_L!UV(TeuKR~XYj4$#p8QoR%K%y9iT}p_9FateK+w|qNUD705~D=?Y^GQ>
z)IEIIYW#DhcQS;s_ZQz-lM(qZMmhzy`qQ||me_opIH8Oy)shss@w(i}QGZI?q3j{7
z7ewP|<YDFeZJ99K!Y=dXf6p<ofdb2ZQ(ORQMh+OihcB{7DGy%KsGbOo-@5u4;P#tB
zYXM+%M7Hs+l<~jp0sf|#Kgw_c{HD<J{7n&Hcvh$+U3Kmk-d}b)00A8&ma6)kKPd8a
zNvTyBwBfGlD0%E+<$OocDsxgC6cW$?R}W>O8X*`P^qH81_|ii8Iin0ozxq7PDn4N%
z?L|abg|_QGLdAqE#`pc&w7Ip%YNhWgvQ$<?7i(3ep+QsLB4dc$#3BRN+_U}@4rL1o
zk#Co(G~T6HMsB{rDJFq$eN|lZ(&w4!Px|`uhx+&TckRxkai^GkR|4`t*=(Z*FuV8!
z^$ERK7L7#5G%Y$TMP@K7Sp+K_mnzy%#8YEucUwVAaB$`4+PLHLesdF{2h}L65-C7y
ztJp2?swW+f9F;Iq$^$N3Uki6CoD@i4Dtd$6GyNdyEsj3|{OVh6X0IMQ1T=r2UoE4P
z@MlcILPje<jDB$t=3vOnQWB))Qa;E42U{7CX?O&T^(4=1k)xc_-GcaGq3L-cC)f_e
zh}e84J=a|Es%ymJeF}=%&zTSpgddi7*=uy50_9R35X^F7(<~s9b7)5tL!Cc;4*!`=
zhao`R1I5#wQL2zp!sKT4*9JS>@X>>srm{*C8W(1>?5+A{&T*&di$L0f)cpLKpz(w5
zO0*<(?M_-Y;O#nUJp*?sh_$OJ+_0`t)C2F7-z$QNY@`*ECdFU-NaU|5XeCA$kL4C!
zs+vqx+Th^Ou<CAXbn)WKDzMqCU))PlG5r3820lauzh5gMzGgHbSrybOB08^04m}!R
zpsXxZ26|W<Q{#xl6qikqP#I-da3-2>g=hn6RJNT^b{o}avOEs|Z@j&CRFhe__U$yC
z87!j+2q=yspwgrx)lrHF0WpLgP&$DGX`#b7jtB%~5NQcz=#W4nC4?kE6zLEULJ~*_
zNQZ>bo1ysSeCK@YTkHIF-uGSW&EGp&&)(0IwfA%1*YCO>8miuvftol01>XM1=sS&_
zYQ29~hvB>>RAzFW6;&!|DDkE+*l77mWLZU9-=o0NL4@e({kr<y!Hx}y=PPkpt&}pV
z>K~)Vz+v*8ZPn3S<Y1+T$6%b4*_l>PFh=!cS47RHj?tR~NUgj0ySmwx;;V*irkSvS
zzXO#J0ns%nqfV09@tJzz1C;WKt~z(VQ$DI8Em?fcApnu1L~RQUHzj99_XSiW0T?!#
zVt^n_1XA|Xo<FOgE%r@{ey6cYJ2Rub>#RwX+y(EnGctGo@`*CCNcs2=@4Q>@qD~U6
zFRs(jNgMj4*x6<Znl8}Vdf(($L}|Uz;>C(jN4gcm(@U2~J9^+F7djZPQ{%zaORR1V
zQ&zBKs6%NXd?{-~XGS<GtW(ln$;b52D%KReKkc(x9rJKp+Shi{N!-7%rN*MCR2m6c
zM$aR(@6>Oqt5xNt_+D@B#JdiGNId*-2*vx~!-Tq7h93CGDJ)}0!6OQLZ@{d&-r(y*
zqWvxTu<;b!%j?IK8z0~2XP)n_S<qf_ysKu`Th}Dx&#Ya=C^~0&b@$IGEKQEidM@rG
zmX>E$>D!5)!&qw$1?Buc&3?Vkn%J20nHfQV-Vwq{N;%KDP!3tuno$nmDA*BA!-ikI
z%*q1SxMVH0X^$1y4<cH%7m6Fy*E-b#WVM6nX`nvdQb`cIr~Ok;XI)4Y(ywQ^r_QQ_
z%5LI!I^0@HWxO@F8vQG%bVMsJ=U!%i(L?%B>F{;F#$ta?kkNyILWROY1x%stOw?kb
z?(9%uOz>jA-qz*NmD!@YLANilpJAF^b!@-0pWvNUF;hb~=d6>x;}Y*iGF0VQ1W~CG
z`zxU287Hh0@(IH_xcSMZ*uAbQ8K*sO({em;y+%%uFiluu&|!gr#up5)$2zf%PVhPR
z>4<$2oM`q_bSqkY8P%MPZ=w1+eGZEg(N~l~)+;U7WuUp5z#}u5ytp@cpYdFqIGbpw
zSIZIg=1&|=p?uIw#28VgnE{o!OthD)33P4AC)@LKOE5>5cI6Xb^!_Gv38@7txpmL7
zl#v>GU-vJ*$P;ZqhW8tWNf|XHSqIX|iXM6PF%9}|WFhj6Xn5Lm%3G6<3t~VGspdiD
zAg3&-t3^|4$Xd}b(Ym-!%jUeH)L{T?(r_86CFy`aBk~acLaNp7*;p!~+1U*Kx;9{6
zs)7i}LuHisIDOt+rJXlZ7^z<y>mW+mq|tos>q>*ta7B%(b5v5`{2V=IX+!5W@)kM6
zS5@M|PQC?NFFp%Wg9#hh3N7JiO;ngh{u?a%`cvp1&*MVeGBM7c3rL)0h0<r!A*m0v
zGdj93XbvGBQQ8q|0`PBq*4mN#OYPI5t=<bX!=EnZX4beRyQM}P-HJ7tkhw~0@gK_x
zyHZd@R)IJ)@naH$2G0`zUBde}ypS4v1xn#!Tz9rI>_NSRZKnp^9l)9s<X}TIj`lvx
z)zEXtmlqwfqo^0IeT$l^QUp%U$&#4a8xEOiPU>9R35FHM`8v3UZPUrDTUE;pp>Pu5
zYN}T^2Pw>Z>ih=zlZ_(Cwo<nfQtu^9{wijQWiZ5wf-i64{sNwP`Oo87(SrMT#1C0p
z^*V!F<Dn?2eEt<?b5je~Hho)nRN)>&L$jKiYC*};t;^9ZY*TB?%;XySqIdS94ZR)W
zGl~REi;<m&c&_Xd+WUqhbN8AZW2-KFhIp&5DaDC<<ZJG3+J@C5-C9gai~l_zR2pdh
z+{Nqj>qfPs$8NJe?Mai5C(BAtTg!xMKmzq^K<evOWox7cg5ue~p#FmPH_rtYIx)qq
zmD|v5qt@j!)3}u5>n<B7Cr2vV4<y!@We#{W6T&z$ds%=5Xf6Vq!;)yES&^lQL75<{
ziJb{6TQ2&&4kcN0g%cX&*H8+o8l-?Pnpr{4CtoUv%C5WaGdk@hI5rHbnPoU157`(B
zN^M#`?=InWzGIukG@-A&hp5C<>V2xzrA~s0dZ4N7?@G7)cS8QM)3Yl7ZCx=%&Z;@$
z>I*4_!l*9I<+}-|FbFxYWvE8P5_LTt0y7qCT1z3NBV0UTu(S-4mg^ab_JMtxZ%9yJ
zEGMVhiBeQbJ?eZ_1@t_bW_g{zR^lgBenk&v4rbf(Zr24wE;sL7R;Hi`E6|C<i8BsA
zNbTBxY0#NH^$5$T_CN>DZ9fAyFyVo=0HX#CnJ9|+MLJ9Iqrr)Y)>_38Lu}wj7diU2
z9}mDo^-Mk^X6;pmcX~TFR4=r<7%C|_W;-dS&}69APg>7w`+i>C(z`KdJo4bn*^bKS
zNaAcw<X$~wa)&;>Luxsl5*D=Cm*V+woNUPexX;IOk2K1G(hY62q$4Lgb7Qkv+d|Yw
z<tfZy4ne!<>os&&*7=$N*kDM4y_Ha;G8&n5<gyKm&(Nq$7?(V5Gaw1Y`j!gk-zcHN
zr-#8n3#3|5wKFG+sS%X8R4B7$i&Fp^>j6TYWGrLk4bX>{mxVOt`=@>JS_PdQW4H#(
z2VMjE7FvRUrB6!%GZAIG_ORk&*LD{9lvy86#C3`^FUFy5Y}L-PdT~}GjXVdXIi6lQ
zSz^V@%ZU&naY{snPtFs(QI6P9_L`%9o$-0~p3iE?0dO8T_7VUS09kPl6nz{`s_il}
z<yNC7wUV(ZOC{A6^K3lL&Z#a0sX~X{3EUUJ)&Q7(pceaYSj#fwrKPJmFu%WkR0^jX
zP}b}HxAl8sU!!%VWWC+|Loc*bp5kxm9<#OYoYsoHzOb!qhI_wYxxEF|(~A}9uU$$z
zEH6%0z>o65_#8p^Z8|yUc6aB;;l0&jj7OJ&3pP~6v7y$rvm#|-2N7;E((Panz|o7N
zFr<cF2^c5S>}bYT=!NjA1|dsXNLuq+77ykj1*LS;a1UCDk6u7-p+~HC0Xgo>WK;Sm
z5B{h3qu%GdUxI)vlcF=Mn)D03(`K(#Xczi#xj$h#Bsg~dt>6Z&EN4=aRj!m9zA7+T
z(!LP8u#x#O;kVe6lV8cFI?`k`YlEaZTUIQ`_Lc@lb_(Q=*t@}61=4yGKx?;x95*e&
zO)N=iZmiX3S01J<4>K~;s-ZVU7^RULG0t7dnRhSYnIOzNrU|>Jv!?_JA*N;eM4=-1
zG_a_e=zmbQ63VVb2uTs+ov7%l>@f+OlV-_`g})3@ndS|#M`B?{_iOHw=Vo|q3aeay
za+(gevgg<LBN&q&ok8S>cUIDd7U6%-sCU086sONRAoOM1Sz07G$E6=ctN>y`mDaV^
zDD8C(jn=T`;ev`mJDXuc2TbZpyrHkVYo`XleszU88^Gr{xaYf9n&qVfzIGslZxtwx
zLoI80=NSF%1ZI{I9?A{aoPBfHnf!?zNevnbaI{{yMC;bXw{*>FR4b`XRNpP(rT^B)
zuX`E0`QvOe+WYI5@HiGAs$ZIDalFVAD!gKHB+3Xcb_Dv7WZXV5G!#h2d|q!<>Sg*3
zFW#?I*qKawjp=NrK}L6?8n*rVQ=@RpOpk5$PJy<d6>Kci#jzD0Lu95?C25{2t6*G-
zwav)F)NlxPHavR2E{)FQlJ<8rw|By}9IgP*UNT3z4hQg61~&sUgCxN|wG8eVw}NHu
zm($UI4*okIU27EUbhd{a-SOVQ>%HLD<zyV%r4yJrK0YqIW9ANi2cj-%EqdQhETN~6
zuvIq6I%mcwEvG@fY0_;`Xn_+zhPC2L16KlLnyT<G%rsY(?ZaCwZx77P(}I?f*7EWy
za{a7CA31WiA~_}4@b^Xp3l1|gdzBdkB%p#$Sq;}|l<F%y!IFF+t)QV|&m(aviKILr
zs6pkZxK^OFSFIc_&ic3ceR2;AxLy(iKU5D7H4WRkRhP)|-g#9BO?B4*$)a#QmDtfV
zrB43_%*kmjdAms4pkVv5%wOdmJ*#)=A=SjEs-D#u%&D^dLMcz|pF*J?c2Vrz3euAQ
z<&we&$ic%f2F#wp5p=;PyU|DRT5u1!k|)OEy-}LL`j^E2dKk&rM(z5B?nj~hFWuD|
z7GHBE=cLCg4@IN#LB@cthCEkWF%aP5WLHY_8c1!R+Y%{TD;3&YpdgfPS*Q+YE;C9$
zclXrAlyznP()$}HcnsH_chb{p0M}4cB;^gRjjcdse&2VwIq2VfZxe&+sRPllNY566
z_{Yg;sI&DVOOQs@aQJ{~!VSZYE)K&vY)NrNY?rl)n-=?SZj^lhnqjScpu-=ag-HKU
z1{ssu&3lDe&Y6|{!H9bI64L`&g{;a$3gg-pldy>F)&BtAxk>*)wwu}xSl$--WBe{}
z%c=<f(d9S;`Rq_;^!#$rQrWi?apDfgv@>j>#JP!mQt-#O6D%>x!BJ(xj0S3>U3Rx&
z*YluPD0^Si=pqUr!j$%gZug_2aMeS5^__L{(NI&-(Nf*F6Rr0DAkfaSZwr|+J!lNh
zPc@(t$Qg0{nC1-(0(wi1!L_o(x1^&l3zpX>qm_C>^56!!tIhrVT9GF<&UZ8#8e3Kj
zZL9|RXgkWridS`hFg|%S;X0u-Vc9s<+27Gu2O{r()!kbv*+J|El<+Z)zz~id-RjZT
zRv(5bCR>$Fi0<Ttvs!)l7boy=dY_MQ0bH>1$R%tZGN7bNxGFx#5)W6$b&*%+&du;!
z^G})i+_wIU?xFV;YzMHiLlPa~x0+9zm0BvH5-R0HD|LVx)wQv6&6SFQSeXZd<<?TI
zUyfuW)_C!+;=hL(?s%Zwt}g_*U7lM2z^^mF1))ysQ_fV*KHg;+Z*OE6)ugvs+lltC
zRFwFYW51MJA^W7|Vo9p|MIE?!rYpUK)yb%3&9q@a9-aY_O^xgf2QAa^PINQD5J)BG
zK$3G83;&EnY=nP?$-wsu3tVT$;q^R5g!_dliQ&W3v!=NzF(NwL+kEwaWvH;r6w|K@
z<W)VT)8--=na`Tv0X2RxHE5aGezvW&Av)rD&S_#_uMJVxqQ@rcyvN7!n@jj?F<(y)
zm%;u|!fjM*ET4yrmp9Y4FU`n7RrrtP?kyu5Fg?lL`D}{8<6*e+K%hwTxN{R81w%En
z>?HtNo#<5%M98%5nA$>6%$r15r#RoeHH4%fs45k7z`XOr3H-{ew>yfDb#($Nu9f~%
z95rpU^?L(d??E}%qKpZR$sSu0waa>J$4;v)cN%7zKOcCT%zc#X&jL$qM+@{|fn<FL
zs33Shu67Nzcbt}9x!jC;V7&YlIAA7TRO+#x<`(b0k409a)wE7y8aqPmHLU;|Pv;~1
zTc1OjRlbqH8jod7TPB#dr=@6AG1WA`f<fMJBAAPciiS{0zWq2>?%6AcX4~dTt63Ai
z0^o%$@!t2<^mnYhit#_pz^jr$LE<K8g4FxJ%I~#xt7M$)SyC;Pm(9#{xm&#xu6^F?
zgRe!ae5=IXUdirSJ3FRpw+OTByldnJALZzoa4Kz>&)?%*8gfh{$InYj!K+`~y+o-A
z<Kf*ZWkTwB-2>FF)&Tj1Cu8J2^C=Q^g0GGm8^4sr##cS{+6Jsx68zDaSO=I7f}yhQ
ziqNTEMjnBK!Q^ln=sbpunl-MF226Ry6>}F*U;AHljr2nw_iAl$?%h8RkRe-zLjQQ3
z{(<KH#~b6gh<}_s9=|93oa$3|V|>ypycZ$Xd&TvasG#=rKLG4m)l&MD@bI}Th@T;o
z-#2CFOjGqk(3RS&_A8&^^dn2zu;bQ>$)Xn(Gd!IgzS$1v<W3F1*s>Qz@F!@^IT3Wu
zA@tzmlVN-)@Q2Z9U<GG1M2k3>v|~41x!`N-k!HM<3F0wVJ%*D#a{DVVIk@4AG_7D$
z=hqDhni9FNsR$`|0_iBn)iV;-2xi$&NByNG#}i?(T;tmY83(F(T-INpVU9hJ0dyQB
zy#ze}l=b7M_hau#?`Zy6$d=$EX(`pejzw0dF4V`uKhncFh!@FOIKOacKqM*NKE6B?
z@w(ElIZ8W1+^zU!4l1Vrk^Uk<|0F3i&Rb&D^QjZIx*-5+b6K5Wv_xyvk9o9eBwrgn
z-(poYDo<vtj1fYFk4+=!6^zP**RrtWMklfrYZ-Y9%eOsmKN&<z^Leu}<@gL*bYcW>
zpCdExC1yJ`!q*KQT-sJw4L~-Pl{N@7q7(=@SuUXd7XjyFx;9R~Sm0J@-R=U!U^@|t
zzY6pOC)d(1!>!Fz$mvmh+|Mny33rRkIP8+5vWZdX@>i|tMee0su|<Df`><F8@?G7C
z9c7?1%@3pA2^YGJ3<KB<Si$nlLKa@jS$kJJub_GFCpR3<`hzgn1&OnjP#PyjD|>-4
z@_cj^8IT(5vL((<!GcVvn)DJUoXpXRU#igkt9|`71FAVv>1FCKsl2mLI%B*1Yrx+F
zk-K>+HpR8GBT7@DNKc1H);ibkQ*S=970ix7MYcotMJPDie<S^utLq}{*GuSJ?#MRZ
zIDn@Uy3lDD0eiOWY?CC$g>NX+Iyij|dgU0q<58-ct<U^Eo!l6$py9eswx@0PqLWe)
zdkq`@4NL<dKq#)1QUZlk^G!*mGU&?sNVrQ2M{+7YaWr5sX`sm<8k|1?JANsfNR78i
z^xm=k(vR=lyp-W>^H<)6hZAGS0}%F=zQqfCN=4j)Qijb7j`jB?+FE4{{-Ca}B{+Uc
zvmTT;s;-1Mg9V#m1))M(Cd2ofJx$fEAw52L&8@Sqn;4*Q^{T>ERS8*{om}*kmQP$c
zS+iiVl*MR_UKONZnd$LersvJf<Uu4q^n#(lA5GF*u9zpypcSzkIIn~R$;A+KVjg}w
z!O>nNA0p6{4=X34GiS|@(Jwy^eLWufcA_7Z@n=)_!E}V`QNIu@D|vP~)EGMS3ima@
z)OcSWd2V>`cyr2<@-^}3+leV#=UC5gCxochS>sif&GGXQ?iNr><NM*3AC&3JkNa+~
zw;F%+mz#5PvkjSHV=JrwehVDABJUDyxmpW>8R$M9%P?y99RBhlP2MZfF|=E`8m}Y%
zbs}2(Gb`H6t(B;AHoQQ?eB_RQMN9t5MDh&64iN5whoyMjCdXVq>UZ2{+$3gX!08AF
z6<|FpvG`T_(|L&V8$gTIWqqwVuQ^VkG7E1KDy-pvKvugIDZ)&iRR_&s7!rd<`sSll
zHN`N_)GY5Tm!GFQvC@U@&kx+@P}OG??r_UZz1v!tvZC#pi>e0Dez<g;KpYp+guP*A
z=WxTb77)&AGQ8J$BrUDsu_Vy;E%Plu_xkt(Pq1(m+s&d6CNm-Ya6lU>dXPfWd`Q35
zP6Uyo)1u?k-OiBfq7lyJ@SCM73qJ`<Ta_5EIrg}Vh~tD~Bq3UHD5-x2?Q!?0CeYwC
zu9P~KQqkeoi@?(~2WC4%wP&Gl<u2Ysf)l1*n^vo5iR6ofDj4u?g;Ac!0qS4B9oKPR
zd^!3Ur!cb*Es%H4U4mOW4mu!4RNB>N<llD9wBP?%BQfrp;dc?0FWD+nDP+;8B_pkU
z+e;f+MPxWWKh{NHt?XoMMyT@g-)FTElZ#3LtSY=4rj$FrJqt$AiV!c;VCn(H-5`X(
zZJvIcEojQg!|$DQWj9p%(!qTpDyA~^fy2=)ftJ}JBo9bxY<`STC$Hu+&*Cm2BB+;-
z88Q(E+~srY`S7BWy`$Zjm~SV%j)dk8Z(U^!6Nr;RN1G$xPAn^TOiLU%>)$z=Q={xQ
znEkd*6P^a$-%t*Wxs-#>SM=$GgCeNv!BF}E?_m$)r~Us~Q|tfcsD{n8<;Mx_hR2wv
zcME4e9>}Q;O}@NzEav;Wj`ZI_yHcZ5Q<{{&8t=&og{!AF>z#_dSjj@U{6FOq58pP(
zE>+P6NG6Zwd|ZXeW-8ue-%IT3v?G3J$Mbh6QLNdT9jPe{vR2_*bQWvC2zFQGf<Xx#
z2rWpNFW`@6%JYEIGMK+cGG*F|6UQE4!vUMWl>GO(e<$J+F#RB7netoQ<C&kZB$`c)
zB-V1NW_wE-`>A5M<73(UkI1+0t=Q5$?`oETrZDpL&W!q~&kTFaC_wJI2K4U{VT<g4
z^qFZ$tN6MXRP1^$M3{<)pVImNQ}J|o8Nb%YO=?Ja*Hrapx?gLDhn4C~YBkqu6^;H!
zD&GFl)ydTO@Aiw9#tC~us%D)6h5bh=9>GW63|GzS$Ed7opTxnt<d7AEU`(925>1$j
zH(_HHkypsSxA8h|_5XwZ3o8E$`uD&D{}=RMisk^R;WipV=096b$ovEUgZZbZ_!wH=
z2J+?tqD@l9h77m^Nr;MB#FMO^Bx2sC1w8%aPQQ2^zSdKw^E`~227KJ$LpL5LKp0@I
z^01YSM(0H!K#l75m>v6nO#T1o+R7*`sVPlc)fRpV`FG4T<5w?VOVpF=eg1(DpS7<n
zymoT_Tp}wwhyC>X<<SM#pTo%}c=KDMlhQ7<UTyX%<*e&oRGk*L2U02D=?Ug^?5Mly
z+cVAGGs>MMP2_7GHD1j6hvtPB1UE{~O%ML!(m#Rk(q;bEcAfYJ<B!*5)KvN>7Gy0J
z>^^bQ3KKmGzt7Ru2V4WoU1Mi{P*x(N@_mFRQ#ZQLCT-&^p93z^1}q<K_shIW`O<8t
z`RH;;O6wKzYabG)Oy%)KIyWBuGmW(x54sg1ZyB|$S_>oLFM0iso}d4B(B<Fot>Do)
z151rFawNvYU^ie6hO-MkRd?&$J-te`DyLQ8uX1|-B>fUYluVXPeEc>x`=S30m=rPZ
z_@#mQHh4RJB<@t`p&6}3iG2w*GxaAbA=G6eyVkp_)T}kDNi%;S?a6J7%71d>1Pq2A
z28g(wTXNk))>;#A)DmJ~-%v6+A-SiCCdB*v_=OIl0CnLzrxA9Jv-F)XAWZ1~$y!W$
zv&^_!q&@rR|9oEmYa{<VlzGyCUT~8}4W$JRJ;Zy8p{d)v$9_s``2-1~C+-E*)~he~
zpT4hB;Ph*4Z{1&OE2a*-P^*xC5SMJF$^9YMKzlYe-nVTtKHsa^x-6*SZ_izEdt?ym
z3%!awAv4HZ9D>KDy5Fw2kF*B7_+Xdw<P~y;*7GC}-@T$1HDvQ~;m6E5>D6hQ&Z|(V
zf`9!v$LG%d#<04(Vy;nQ<`1ZFCa1MGxJC>WKd2JJT9!2@-`4kH)du(fY{s*^U;4<4
zdRs))pJ@C`{IAmByyD2N-C(cfnn0fXsB9hvNvw2A{CA*7y(&w|rnV-na-SiOlK%yH
zv#+-HWh%uZ!0yevWTD&hXa8($P)^i@Z@DA?^{ZX6LO#zwPn5RN8Zje*ZmxZ!O}?x(
zm#;q%bYCsG?`4om`@#>nI?f>R`%RairuLj<2Nl!CKZ^iaD@>Cn2|}O<G+N|7x<*Ij
z#YkOq(P&Sf#y~yf@v3E!PZLygsE}xF^E7MleVlT`S!poU*7ic%8xo0h6Z-VT2@%|d
z<C?5vC3pVg?~jBx^}DVdq!8xpU|;@r;u=ii_*&_MiJ{NJYK2GZ?+I(K6BC5Z1^QKs
z1i`Ft!bJGg(VB&~*ZS2Nh`&XS_~n(Po>ate(--}pdx#P~aHZv=YsV?)HLeN(yDJ+S
zmye6rrzi==2Vl7Em!YIsA;@nvFvh=n<b+MYQgBP>g0;V$Sg7+&NM!$c3{DbS1SCb@
zOxTr%P91&N7m~Q2xe3MTzX@4VAD?e0Dgugl*mt|t<#*d&LuuyQ=}$txMpqUrA4`DW
z%<Cm<gIao{;u=l$LR7pSwdLr|?|F$P%Z7BhCM$dv5YBO<qQJ9f*FPKF+c1s^0@OCG
z@XwHmS`)Q4581lA?`xcf@8CMm2YpX3&>J*5LsAYKx;QhA51_0JUm5>)0_smh8oey{
zk4_AC&W4?pYl+PlaLM2(2fsx$A2fv~(*dsCF<BtHjY-$#jfL?hn_^<yxz7Fo;OtXD
zd6|>j#np!EvZ*F8)&u+S=GAjmug+VW0~%kuXS!EL9ZB5%rB4lUJBxC<<U`L?(K2&s
zWuxC><A-|_Mz1DSJvq$X#GT<CuC}s6H8H+MF?_}w67w>Ux>PbZ?OejCDWDRx&MF^Q
z)9XS#5XS@sD?<8T27atODI>S5$p<jf^BrYXg$v<+r4_kkp><>W>mr(l6&MXcPo9Jp
zu>@v@{Xoqyb`QGcyH8giZn>$mXGhQ8AU6Uo^6z{;%+9<*gG#4AMLc>)1Q?j%p4XUg
z<#gvdo;F{!;4g@8DtX*bD_YFC5}k(<8IA)w$&P%A_A`2ElztF2?E5iqG<Xjt<b&~N
zfZ;)?X1a$jXG8<%P|0JB`Z|}O#9(K-V`N%Ie`P2iB2YoF>F6q)%(oL<SCYWyTv!W5
zZ0`;Hcw$nBx|#Phc-o8z6E~D-8Iau8*9dK4_{E*|YOFez9)i3zNF~!d!{Tm_0#s#`
zVLjW%iE!dzhsxY6QM<JNQ)A+pGKF8NlQ4rJH&Yk>@g(+)vaol4mIUj+Aq>YQDzqF%
zNXAugf=m+3;U7m1?`_wvA6#|*fUjP5aT_GIhTPe_#jCS-o1&>Y6nlY7b5mwd1<(0?
z9UXSiUUls%;q(O%gysj?ExrtO&4hFKl#L*er>s#8Kn3y6My#`iz1;;qZ#3lc$_!MU
zJL92<tj^6%{$uTzLKRo-rtcb4-7V`4d>s#VyzF(H+?u={2T{Z0-M#7-_PffTAEbw>
zYQEG47-=L`Ad#eNU5GQ)P!}hS^TZtSa~4y={}b;1)T1Dab5dGyoN_K%_V6W5?RgmV
zbt#u!`nZ!^zHI+I7&40Ee5@<>a>S!B0q)j=y{!^K?%2*qM}jKI_mmVT-*`GF`QQgH
zs1Rs7Ho^!dkNU^;+N_O!LcN>w0>PW>OmzbKOGvezGXj+T@^VEIR@Jp47t8NMZ^ydp
zHmonroND9&4BOS>Tnr;RLaYLA4NsIR(Nb>vHE#B8lFkNzRIk0tbs9IRXl?i6WYVkB
z<bUKXU78~X3j{q+rgvpt$s`tLPNub~KU;++*pC+tFW}3$B(uGdFz9fVuN$>NFW%bh
zRgSL}rNm=qa*(#ebG#x60^>8&E4?7uft;m2ipRGT=TeO1R}9yhU|*(ZAw}5W5Vg(5
zn$#K*?dk>xEWr#+!VPDE`=D;0(0kf5LjTeo@!RM^=rBl{DH5gk=eHAirFG0xAu+$O
zP}1f;&`r1d+ufqE;{%LF{)l1nczxL;&<fP3F8MjNjs&;N&0Z;bNB$vvCxBu8f()Bl
zw(lIf=&w6=V7<CuHv=kA=cFiDqm=^hZWt}74y%}qr|GTEbx~frU9{bH4g@$dYG`Q}
zT@Za~%mRl}esss2EADaSBg_Nz5y;BzO{<&cB_i*-2M%m~rcb4FGR}*BJ_KE{s6KX8
zBc0HqVHFwW;oN&NTO;~Kb1KfDx4tFu!FhThUq%|`B^8k}JJ4@w@ek+X&HXa+())`u
zpY5ZrG}1Iyv)R7V@$XbG>~O}HHELF3QxEV^%IU^9a@mr=o17#2YuW?X&zhW}8c6sU
za)IOg#+2hRSP&zG6RSH$)>nd$Kp@vOgOi^(H2H`$jIB@ZATuO1L_LVArsFdx?|pbW
z;Jo`4w~CZt48`>FK=c3t-@RLM{a)mY=e*C8vd+P;{TU7Yy&IH!#LOu8UF1`R^aoVl
zZ~QP4c_8{Htm%UclXQ!W?s)0lb*A&d1%;8vgD%#_AhCyxDYGeyso&x*=~xqpyUJ+_
zc4;4r9ZI;!v$JWBj$h<}B<(}Cg=MyLR-<A4dm7v1CLFt+VhtkkXN7g2ZKn9<nyEw6
z{ut8e@I%K(I;N*vAnxkzIMBXx(Z?lO2<9?wlH;iY|AzS6C74D4sTxhz4#-8;($r2H
z2?H8z{`FTqtz5nA4VtugOZV^k(pdTvnld2YUe~v(s0R3f@ksCB^qW3XeTmqsiQMN-
zA$_sYJ-70dt$THsfGO8n6G|3C=GKfYRWAuoWJTr6PN>pP(AnltO^)bQSqbxx{ft=C
zF+rlWb`Byl2bCcGJ^e6cxwD&?lcBBDDPF?2HL>zYxO&HrW?;Z}4kwRGimz)?b<T1Q
zg$k2Q_@s6+dc33sk<Dk7%yX+s<mZE7N!T_B(!<->MeZok#Z1_mQ~vSWi7dx|w_bMR
z_ChyGzny?|CsDpm*J${B{&oVdEcCgmUrqjYB2&$k7FB$X{qr>9*tz`MiGt>*I}Y2B
zgx1QDq2p&&2@7XQLY3uZV?OT_<p5uFSeKM1bC-g*CF8b2ttpFUs56(3uDY+auR9zI
zU<rpE93{-G!U1wM{UEVOcmdlxxDOJ$<+~W+o}I9NLn~7FqEwb}^tJF#LeKi1@WmP4
z7km6m0x_E<*eN=spzkJ)w0?y)NX#k{TjzqjYuL6opRdf?Rd0WPRwvD2e)75VY!ph@
zKF21czH9X@zRJ-phi(I@p0Qk&y_I!J!M_RXQ4tm0OFp{NVn=U18LP0a|KZX0nT9GJ
zD#33tGBX1WAxf$8lnfp26LGK!i2BBRjZF*%yE-udX}Fn3063eX1fXz-m(ev*%ru>7
z=wX7`o;Z&qR4jUmYCOT0*runOIaHC;bFAeHT-n$iRM>EGr3+Tsh98^>mK3J*=GN*n
zjr>&;?M-cc|M<sKYA?n#{ChBns@#>~ksYp0ho8QCAr4v)HnFi3`UN0FwJRRH>#v*Z
z4raLdLu>tw4elhE7L~{6p^~eXNx0X}@ue>^eN|5xpNi&P%4>FtWZr5`&q~hngs~{;
z86NzL;D&f=m@7U6sYn28nOOM*v>8Xn#MApFHg?F2AX_k|3AP-`R0b2k8UXmJ3r-sy
zP(t$YM+=>t(efT19+756d9zF<qlP5L+P0@@O}*4nJ$F!}$~moU?2QRC)cP-mSan!=
zWm3^kp-O#M=3z*0CSFIi>wV-3$mKlna<=X7Xcc;V=GAlf7K+)qANM8?I%(>$4&$UI
z7VDtjd4#vq{8cP~?O+t4?D+NKc;}8F-kv)gb~qrVGV3DH^mapNj-$rU$6Vc)sF`6;
zp(-s#VD6<skjSs&0ykx$&q7kisM66XY%+Gyl&g7&VysfNhH7@gcBtTsR`4=|tI+y8
zx}T5f3mXYBkhSTyo7%L`z@y#CQ{PUk&@;cC_$NG3f5Z1^kl=8%R(hP1`eD*HeHXKB
zez?vk`ijdwjSyH-rV@nb;Whb=Zzl#gk|X;P#|r}EL!QI86Maz+?Y8ca<X-!@yya97
zriMS_do?{?i*&uu+j6nwc2`2E3-K6~bC;`}b|(fe70bV`gtB;)=Hl!BOU`f9{CC7Q
z_wu5I0OZwIg{N1}=s)iHG_uDi#d1lz+J>PpErYnc7frLLL{&(U=!Gki<m^;d9VD+I
zLQ(X>Q5Ze+!l+PDa%DT!s`VogkJ$AiXOxY1bbrD&b5Ua4h=x`uJJ_I8N1pK`3z2cf
zj!g_n%0*<EY=?G<*aTC}r)$Zql{Z*YukPSKbIz~KR`vreLn>uINGerSjNrTf!GElS
z_A#H81bS<JHAShYX9v%<u?yXzFTlXS0D)cPt*r9-d5=L%Ya<n3Wt$d1pAxFs+8Yv<
zI$<R&;UYuSAeFaKJC0;AHj1IdNr{7E!fAyqz>nQS>%zh-yN3Xq^IUe4WcZM7p`l*S
zfZHes<$(^W#C0fD4Qv3Sf$g?vLIW!30u(0iv_4q;>egO?=M{!9KNI@^TiZ4JARerJ
zZ&vMdspk8>8dcw3AmHue>I+w6`)i^g2D5@E>wzB8_FNslt*L~T;p>LbdUQ#OKbn64
z*Jxw~knKzWeY4Uw-FPuN)fPr4$8HZ>nO1)hW35KUc%-275K?WGkO`}SPE7zD2hi;J
zY7XXdXY|8CLe2uXV6fmrFpT%l+-+7&7}2mYe#!dFQ2FR$(OOBLu-E&HoUobAemeEg
z_~1U_+lhq6wL`Ozn@!NduO%c7)-Ux8cG*oXl+e?x73vJ|zpL-FN^JfMoH3zFSt96R
zd!svbgEGEuoz$dVSE(yeNiLk(pC=fbsW>1|-%iMOQfF(Q?q|h3pLb|^zp>aE1Wd05
zh__|Hrnu17tnx5ZBV78wbKttE-GIFLH0_F@H@mwF>1GqOVAxXF^stn-@CtiZNgGj^
z5{SUwyX8&P;!#sdz}_c4A3fWJq6ouOXaD1v3=h_=#JOQ6C=wJEKlR){1XCtwJ6}AJ
zX(5pf(Y&NC=~8QraPq#mXbP;+zhHSb)3Tc+vvG7`QWP?V%7iT=+m4uFsCNtgTX`js
zS<OnUKD-)yV7NR;FAR#y$bigiI;6*_UfA}WOi{6oruWICWAa-yMihcQCq7asX=NT>
z=IB=LW}|@~n&7a)95l<hdw1K`&Z;wvO|$_UbO<fwI|j^ukrCUUIs;YiWw-R8wa&SP
zo0*2CnGbS*f0%6P0(|^BcHzT1t?mIlr8_{*qQF~^;RWd(S1zlVx2b(CQ(yQKJR{my
z;0|fak|tZ=Y-k!yL$6)T7=T(Fp&<n7Hq<vjMHf}BD#O(vm<w7~WbL|*;Ps8wdCz6X
zY)C^|R{W|P%-Y&+z57#a!ny60q}F&wJidw!t2t!bfdE2tArnE3l8w{(&`}q(uzU|E
zL75pLSTuhaK(@2>f&u>!rdGR0gC;BqF3loVHjZWHz{qUbs;IatySJ@Bbjp6&5i-N?
zfYTZ~7sB$Nl;c%?L`rJkfxs$S6SQ6uLkxDwYrh_vZdYw2d`QKa+MJ5j>0rPO$&K_H
zp1!Z?s3zaNRMwH$tp0vaW^<5Pw>xN9>#1H<d7V)jAR()BN@lXaWAU>;Io-V%w?zH{
zy2>)RH$mbIK<)i!MJ0@HcOJt0x+>3dCa$P6u0-w7(_r#`C@t3tx`0b@jUKOP{=_O;
zcpDipi-D)x+}`PTU#S5+SDc=f^0LEg;~pMasJR<HB7DLMlEE4QnDkUfAx<bicp>@G
z(&T~Vwvzw=2^%gVHW(PKPl-y&D##yG$VNqZLc0w<m7Oq$6(?3w;gKl<EM%-{IVHL+
zL-hO*s4DWPE_fmE3!9-u&N}YAIK4bsLGKkx7mRCEW+rEFD!00c4v}w?=cQzP-Uxn5
z+=^g1uHIGswG$Vty@q~Yon#f|SaVX#zN2cRHv<XLSGMfNV8PYlKfh^r3eItX3ZsI?
zp+U>`w0gA`k?p7g8@mP^L;!EFN9Eug(J*I*x=P@oJ431*foF!y!?l!Ze$}{%hE1T}
zO_ZQ9SdZKJF-r&WDSJMfsWD<6p6;E9-8#h?#qN@VM--q3$>UZMx|6&GiLAITFMs{I
zzOsq}Qs<B6tD3h61y}nOrAQs2ru;S)&qsm@D56HMjwiBGU$7_XV$;-#)dzWhH~*d&
zR7?~Bs$3hnk05RXrlisnr%Ek5TKT2k4lio`?DorJ9t;rxTCN|M>Pua;#z1x$Q;dyW
zkgv=`WWrGI9AIcZP~sN4&?7$Rye39!>Wwf;ISZ!hF+cdi8jKY324hVN<Z*3uZf-83
zNqK^xVb$phIxw)gz;~9@GK_^(J-!|A>_N%1^ZwU6UWyw){HUh7%I4^2+5gaE&a`VJ
zCKi{MUB9)b_rxmi-$<PpV}1Em@59sq^4^X}7I>&N=K2ns$P8%RPFY=W6{d#FO<=BH
zKLSl5x#;ca@jP4#DhPnF1Huq+;lCs|C2&|+n+2df3X>`(OiWyoN_Vf{N=m9ET!!cB
z_G<3E3KBWuaNlfekwfh1F^SAeM3dg!-SW9Rc$M83HH-~mQ@`bu_4_l%IaSSvcYFTK
z_h@51@&(=RmX49cWJoFH)mUtwvYcqkInHkTm4Een__<fjQSp-d2{9IRo=;dLA2T`)
zFSUKesO9t(<KWe}Zek!m&HC8TNenWs3$N(%K6--I3M5CsSal%`nVyg9Q=eLgt6>~Z
zcVZBjajOSQmU<=764{ki*>)qxocclJWX-`vhl^&0WV|VSsXMtF57MNAwVqz-iYh+(
zb5Ykiu4MmYUclt6>u8|y?e+ppISomQ^;o4`D=Pr%X*O+%H0M3Y;2@Mcb!m~BG;NDc
zUDlZ!SzhU<6U=<e5~rBE3xWw$!>1yVWwz90I?gt)h@`$19(cMy!}zL>3_oStN`wW_
z@oV+TtsoIj8jEdjgB%RH&2f>-^`UpR!t$>eJ2NYdB@`||Ssw@iQaj|`06bXB48#hN
z7m7kWJno?RLu#jQEdXn7=$FQ{VxEi!UxUE&#_mi#q~6T%O_>WWSIQ8KRzpWWJ$m<c
zp#*vZ+o6zkAb3_+3v7$M*q`FbbZ5ojq~=}CwVXQ`n&c+`yfuxB%)EFHdEaETkNyBD
z-He=NokfCL9aE(e3qoWMm}>R6(2@I{XruYO#`Gf^j7RAqVGhrJI2H5Xy1$#y^u6{~
z_k-+8VU9+-?m|4fNW^`Uk&>_u(7al@b~N9XldyH>K&ZDWY25EVUOv8@z-cTxoG#C4
z8EtBsbByOS@r&OPo&-SfrH?mPa#XQN)7lQ(NDKE=V|N!1|LZRJj7`@`<WkAa-b&~#
zo*dGSA7h44*ft5gqUy4t-!b)D=(@7EPavS7WZ79B#Py-ji3c<V5Fs-QA$p+cJu%XF
zdgYha^Gq{&I5;=ARK5*M8_{>@>HOTNEJNkivgi%<_3E`WgXjZ2?@KfoXDE<eDXUJO
zEV@O6a6%`OX2_sfLVGF1C3d9s2il5*8~5h+*7@qHY!d%=xAWGX+^Tj#UkK(@+||r_
zetpc&7}&8tX2<-7mGhJAJfX*L`XyuYoC=9lH@Tp5=LssfX$GvAh_9JoYtF5mvBuE_
z4p?lOta(?7_QbI7ISZP9*N4uApk05wM_wLw9yEc~W~NR`@oUckhVKO9$--I^)ENt1
zZjxnsg>}r$F^#PkFhq$3GE=c|3l!gzl$1QE9&6DQ+R}g2!h16^4S92(2)*1dzA~}9
zce?6RjrZ)GH*&kO$~xgua-(^>`)2M>ItHNXj^@v`B48CoOOom_SxY)gb+fNuAS03s
zukKS5{CLM=L1|J*B`DPxJit^7<T8)o1vTUOr(0FOXBdf7N)}uj&;U$U-1U~zH?b2S
zQ5UJmTFNeVg;E?K?U81F?zz7cP-|hT1CIbZZih{EqE%fe(dgCbdFP6nc(iAUDmWMO
zrnh%Cx#}m3fmm>yot$?4Xano`U%9}HYvtlzR5LS6|DUhlcd>N}`{!N8wOex*b0%hP
z@$u<=g1!!|+nyyV2GoCI;KB|Cglak1aP{*@?f@aDYe&CoI^*M;hsi7S&W(LGc;q`R
zqg_5=!#lSbs+0TXdVG#JZtMJzUy&2bu;^mybHq{YmE8EvOH&6<rH*l$Le+x9Ebe$n
zfIHrJ7QgR)yvVjZ-kLoYjHqk9i2q+}$Nb+It#n8h*j-&uPe6~I^=*D1-)j}R_ivZ+
zE2o|_UDMJ>iFlnos_K{8-+uA@Kf9XmE~DO!H4X2u!Rr&V4q*zv45?3KYlKG-2+(mf
zH2>F!E1%Ga-)F|U0@*JwCTB;VDK_~_P-?rIy79hMAvSiM0`ugno;!RzL3V-8%z4kx
zE=xfNbpJE1Z^5lXb;b>cJ>pi($1>IEh9%<Jxc9q>tp#r7HGaBxzcQ8C1UJAw`M=Wo
z3Q_N9!n8gU_~PHm9SE7zj_*A*v#&pRGvHVwTl7cn2HR4p@l6Jjq}=rP??j;qs)~=#
z9q<1;Mi~D)i~PL^Ue#L~)Q#7-)q>=hyyx=G)AN1q#z=;q8T-newA9}03nTws4Hxkw
zP36e#F?>D}XyLB^yqz;G3cvcu_~8{7uZi_V+YOsg@09?OxY0PVX?3E#z^L|=gcL)@
zC`Xg_8ZcGr+*t>qs_{J95Mf)cL*d1~Pghpr?6t_!2`48p*5IfOlofbqI4@-0A%Jq3
zgVf^+4-P{s8={8AUZ;N~fB4^-*#G1^q7np~!*d#kO_Y7pd#8nd_djffR}}eoU4;_z
zqJH&ffbYBdGZ>kVgurh`v7fWmNZHM_+ZK6}l9GR>Lw%=pK7~HF$4x>QR81WPV%ErA
z?cfdhy6U%=pO7H&51LCy<A&ud1^AxW{3r<DdG!0Ckz<iVrR|VOx#J(|4bS0-e{C4c
zgxCI9Q`S4TE7#fC7pU^PuK#rvljg*?bLPX#@IT*@q9L)-^;eXC0zj{5h5nr84zxL`
z?Cte4HAAj>{mXs-q&mkx;MXdIv9v1f7AJP^31|J^b6-wLxd{cx-%f}Ih916ZcHd7-
zI~QvFApIz1`-O0yl#%>)Vwga)i<A3yf|phxH05O`?6t~`>c`BCJJ<``4fZFQ#*V^L
zUgQ~t<59|Ct;|E_lF+BS8|;2`<Iwe(?tTo|%V|qENF75*ycwP`-mhjC+6qn8s2ZGa
zC-f5H3zf4H;KJY7|Dsv+)mdoqE!5D=c7Ho@%UHtrn7yWYEUIu=$1+~Do)B8YF613I
z_6tHWEXx`#+G<$3ohHu?eY>KfeuQZq>KdSft*vu8(+7zCIz}zX_X`C{E|gW=c<mB9
z7EVZRfco2x>z8^5DQ9G*y=>LaUSvrUvlSH2oeT5R!@kJMLP*?hxT`xRz{<<9_c!1v
zd;rf51XHmEi8_EZ&a<9EkUkg~ndDabqS_hnjv<Erpn6rELoQIWvjs=)sQ^vmG{nN_
zY7SN%8y!U1neS>P*QCyTJMr{Tr)l^fi}a>6={`9y&$^U$75;=h?y*)u)*qFS`6J)*
z_Gz)|!7r=;Ngw12N}nA4iF{*}{dmm&z;KCEr@i=ZWoi7E?XnFQ0Ny=?y=Z~2TrtEI
z$W$7ONH`Ggy2L<-O>HIWw61&tr|6vIlF#4Ov!DDxNaI3%rn#5Kqq|#oVyDC-nq)q*
zg|;!2axZ#}{Se;J9yZa1p)V|2b!$#Eyb>FfDX=+kGxO$RcxXCXM|oVb$@tyZLKWGu
z<KR~f@7$gXSAX_Aca?i3Lxl_QR*vCk(G^FoXxJToJMmCy#pd>>3QHUKT(j21!1Edh
zqfa-APEkH<Da&r>3RJi*32wNDw4)ob^27*L8)^wXU&JZut2VE|SPv0q0!(Y#!^*32
z<3c*~^5ayWKXs%lNae>u6{UoaNuH;|ht>m6jSK2QfoY!r+H1~8Oe=_NzhDSOql@^9
z{t+tH)^Z?oH?u3;n&?5&reaBCBxPx|b>;a7x>=(dNRaHYYy2{xEvMB)5Gd=TIfgHA
zb5~pOUh0ipshC`56MvbtEh&agIQG489Ww2$eiFO3EkC6l`_p#g_)g?YNqtZ8G|T7v
zQDnbS<v}fusJ=r-|8Sr!dKeDvk~b<MR?!liC&?hqXCC9dy*Ol`m9_7Nr3EG;GNRfn
zeYhwj@L7Of*f<88k8xxw&x7m@XresD))zX*=|Zs$FUuBY+%y9F-sMl0_zWEJ=aP|A
z%91dk%%s$LhOp0YN`0m>Ki+nO+ETrb4ZVlDO?%&iaja;>b*bb(>uo@<Ru_a&tKRdg
z-3kgR4=IVqTKk@}_OlzJW(=LtNH7I9&69#*N+GNrDU}yB%vZiw9od3+GxJ^Ad0C<Q
zq6(l774zb1o9m%fmEsuY`qomNU!-y^9~u)d?MHNE+3=38qnep3-k#Hg_C88r^n$i`
zx>FFtod7_+^PJ2@V({<*YEK2C34-%#t_X7QHPCGwX|jW!VeeK1FhD^d>EDLKXLG&1
z|EfFY`melJ>AQ^UR$82G{Ehqh@KIFAzi+W2zdZN&;N$u*NA+H3TT=WF`_s#&q|F-6
z#t$7r*o%?=Dlfu$DvXO<wO4|~S74%j5M{v3>KRW+bk+)qKS{0gz3zDm(2x-!KaXO*
z_v=|ZH0O|k&-{0}m1duJ&IHKwNx@-x-5mt?DqeC=UoQ%-;^5J#J%ffifG$r`Ko^I_
zfTqhOT;>(c(KCDNe$<UVD&@1EPkBIm%7f7smp`g1HEwHx=Z!!P=tyO~8-$eUp7gig
z8+CoRf*b~labUxvX_`}i<d@!t&`M!lGaO1WPq2nk(BAtQRr@dT@RLUre0ak7@YB&N
zoPOF?V`^)0b0`{bkHZ2us-|Q#MBA^guG))MQ0XshEOj)Pk0^WCZu6w#;alsa>Yc3@
zy3(owq)ep}wGkf8Xp}>(7kl(Q-_$lO`Qn^6Fx`Vl03$+uj+Prjh)d`>u5n(+q)MJQ
z+oWdv=-ARBjpy@nK8kneP7GW8sr$UH5je+b@r?Ut=HRNo=;1=oCTb#Nv+06&f0><B
zFOVjQ$h|39x?Q|~m8|J5ReuPjHP*BOpYA7@^853oUs`wKWUWVZP9jQXpn`3_xq=K#
znD+qBDf1Up6y?MRjt`}#l6AM>3ujF&JwJfymEO>p80THnmN<9QUTVsnHP;HnxNek_
zgU0op6vbsU1(cB~L`=d=K~6|zy1SKIiGr_g--fyM$Ht<|-C-`*KCVT8ioD+`dFuYu
z;?Y-B(LURAH|i`P(^ISIap1L&jh1Q~sgZL%uWqyY7$zxp1Gg2;uX!Z=fiv;nPMdVD
zsHA9eK2}B=6T+Zo3$?^MxcCY}0hWUo52Gqt(Ch2e)o^pqs1e~{n^86;Kt$pKr{!y%
zUg8y<RaLP=VVCTB<3NlQ;835PsUuI--FNU8Sg|@Pu;j0jmv&ceLsC>afN7p+^lfqo
zALPW{(UDIRo25A5oQpqxf+90(b*W%`tSuLt1aWH~@p#wtuF+9xfMT_CRT^c#1ERLh
zXF;kqYZfItD)&H*g9TeyQF)rWlXt<ZBN@rUD|}UJ)r(C{pXl{l+V;f_d&l%KNx@*e
z47{{C&<}+R;O2gP=ks*hf1hYu<#&)^&j>!UzVfC(FU6-BJCEjJ2R4wC_^N&2zYwW5
zL$Y&ufu;GV7DAVSb!vERZ(mfD4kF^=d|vjEwY86tR-6N5;gH^a-p<7G^=ZS;e%ZAv
z5`lY;yOq;V%9p#QGzn+vKdXm4inMps2L3tSI0k>x<=`$K5VCZ=VzbbN-k|K9EYpWJ
zjVse4o(=PrC_m1QzYxUFS(ct@I$7C>=V|MhC%aHwSWe=NExdiy>xUE33-QD<*Wpd5
zBYLnrI_!|G>FxLldo<wM@1&bkU9oG(^6x#qvZ^c#9M_NDNi#nS`S7`M0$`R>uWTE7
zPC?|tpgiENHWz!l&dLpM$41Wq0sY}FL=&nQO{4$yCGgmC{9<GsR2^p)s@L-E#81PP
z9O(~AAvMu1+L5h^#RH6RFDFl3&V-H|w)KSz=vlJ;qlQ1X>Oeuiwri-gjCvWZVcq_Y
zKeIaZOY4~y?|~^j88`NX{+{)eex|hf=!!!!W)_N)7T2UowF19C65nh*%&!@M`l<}7
z07$B4Vuu0vVbch7Nm`ij@axY;ba3hdEcytj*QHY<&tYC;XcbS_L7e^0SGDx$JT!-^
zEOhARI4ILK1uPg)R1KhEsJ_vR&IxVrkJz~O`$SZ}qA1=uCg$2O<c&zw)3$cw<>!iJ
z#G#z)OIt>!1KsZm{~XVL9;X^B+8U2VDl+l;dcK5%_|itg*`hR?{Dc4ndAiKkMS}X6
zsstYt^C^yIdVZiwURkJR%Jh8FPH{bY=Yo#!qI@j^yN&RyrCdU3m!e<3YOQMKbT!$z
zeCf;_iN1r-^Qnipg-cedmF7p71bM(eul*C(ISh8fMqABl6M%0MER6a%V})7q3akAG
z0pQX=NUpw!j>z-D%p$?tRl~G&y!H-jnI50MEpwCcd~KO-j$2Tl$&X4B?w~dnU*a(p
z6YB?m(5ePI@(*K3t(a!p4Burtrah+0hbEfN<dJ8=72EOM2ku(kE8S`$<b*ofS*d(j
zgibm8W9K@HdSL=o5&o?5>zh!3jrZTPVWiiEI9pAGMvmU@k4f*Br4s3rd0yG6a!FxM
z`ddtYZ7sl=cZr0Cj#TLQGH^*`+K9hy>A(yj>?P`=Vx&X-Vq*05CEV$uO@o}Yx>ij^
zT|vAlW^ao}aXZNy*-!xp0nN-HnS4u$P_4@@{||3({*`3@|8e)3W@@G~rz}lP{Y*|;
zmOAFXPg82;Hs*pGI%#gGkP9xjP1E9*I%ejA;wUN#3Yj4SE>pQ=iiQX%pr)p>xI}^r
zirv@e`xo5j+~?dsz)u|Ex{lX%z2A@5bAgA%o&++uzaCH_9j~UuwdwO^(vkO{<2PbB
z<lGQe5{&Vxi-PM0vN6IhNI97flY+X@>ukWtHJx9g5wYwQV8qb3ATsxMmAQER##6+j
z-7iL0+g$A;)Li{lJy+l5->vK)d=dBxQwvt^Ty%c*w4)B>RC;mAFStNG?RPN)zV^hn
zgk+kfBrV|9Al<49NH3%X!wKFvgXny?KQ-4fkvGsdGUnsWD{l`y$pfY7Eh4=@U~AKg
z&?v#S3fzSkdsyAAJ(>+dCaWAWnrENw_u@$tT!-*!0gbcXAieU&X;W__B^BO<l!qR$
z;Q*<$)p#li6?OdYC^g--qUe4#5$=lh5+UCF+?_MT$CgU}AzTlr%2vX+4ZU=%4L5B2
zbus;h;?>H0U(XXAGvzY{Rz>KJ*PrLR*$W*@GgL0ZyC}i)^ZM_YONP!29795ZMaz23
zn$zNd?1{rJtoXLXR)emB<I}WTRmdH)FPFV@i$mhZ9wX-mBrMZT#^&=wWetKju$d{&
z4nO0t-CgW$k;1<0`3MSRR^&sM9ItznnoL$+tc%1Ocm%pzY_wH++#kr^B$(&#azR-N
z;;S8~ziQSyKHGmi5Z+R=_n^gSHs5{?PU&6G(2{h8AKI|VA_mvlFFywIplRtZ?go#C
z?e79!lCK9+&2;y}`J-k3_HLQezaCij@7?!&?|8VvZ@(u_>Fa^Zt0eh)ymoAs%VBwu
z1hAcyZv%uf=QFXeTfLXNH7jQWK}+jt_dmFw(&F%EW@T4Ihd*DLNNwYn?C-1te}5CV
zQnN>WJgRvPm@aCre?2gj<_#EJz)pr0_3~Cgy>UYUFuY{nc@P0Kln{0Wz5B>jMVVXY
z3%OSB>OtTKA<Fi}dHnyTzND=o_87XW_j~}MMR<AEVdA%l0NF+RS%=cE2VT7e<QI$j
z9j|--&f0tg+y8nXYIgg{(%!usd5gS!|4Yi(1G^()d}d|udW#s4eMC$<18oscy3-B5
z9!OXPjtFr%Yu&s49^{Msz&&<jO3-sy4ebEb{<~Rb<x1P0{Vyz=fE%x7FAwSXc!b@S
zyfw})jPj}Mi))6D?hG0J5Wn$zDBasup{kB`wsZ)R8sz8nQhpE$v<b=VG7Dypk`&R;
z&s$JXq`H-iP{dlbx^r7u6Mr)q`XI||Mjb0PoqCTJY{G?!a$<tVNSz$vczV~1Lr|Ay
z>W_e}8J{K7iwjA;+1*gr;WK1^yjdCM(<cyg)YU%;d`1SXaq0J2o?zyhw>P&zA~8Cv
zrmh|d?7+r#AIk@6igSee)Z^IhzxYACdo7U0(BUk_H=jqGetzFD@T8%=h;Hn6q-Wja
z{hIQP{3B$=(YBiNpY(nkr`!J2tC@I9{*(OTR%1mwpJ(^q?OFna{QN>bVup$r`3Dt+
zuTD00J9`ie_T6e~BPUI0Ai|<IOiM+LdD;OR-jffHL3zc{?l*jbcxIy`&y;WCP%VKj
zE&)5Ho-Q;%4MfqO86Ki?D)svTINMy<yikcQ>ovYwLmMMacRj(6P|PZG8&t@964j?q
z8adI*_F=|CvWhcW`<#dB=7|mM70r&!(gIQ_6k)Tij-F5M<Ejt3b5o?(<CH=bLxsK_
z)@kQr1;s|1V>(oX-4YwwwH4$#wmGtDwi%oCWlh*&0Q@741wW6Qj@TEBMQ42i8+|=c
zHpPEoqxJOw({-Et^??0!%?^#E(X;sVK)M;QjVDL-N-nd5u*N?(jf2mhP<@*EW$7;X
z5fE&>lu@;#WvHmFc`<3p5~YpW7_@QfNZAyxt#HhL*wc7b@>}Oa*lv<)?~ls!5HsSd
z4YHe7JtaT&;_rm@pFbrN$#sF)z<1EQm!llKq2`6R=k=T=#Kh;{3OrBwpNTjq(HC@g
z+x#YhNMN5~ze=FkefDJX1@sNp7HX2Q>L9mvCpHLn5T_UwuK5x{nR#KjPct?`<YBh)
zazor$gh8x2?%j>TJVFE&^q6-nj?Bm{WnC_x<^>g7SV1qMC7<<Whs3@<TL55zGn;K$
z5E|9oQVKS^z!zW;R=D;VHTCPv;jV<U^zNgb6vmTY^>Ln}yOE2vLhon-&R<npPRlSe
zFMQL2jMd1;%)4|G*OKczS#4KX?|=JosB*9Fvu<R{v;ef=^Zqr{)OBNC`S>9|9qu~Y
zNiu80M^`4za3>LkW<ISYNGv4z%gP-d`2kc~ZLi3pr&${;K5=n|yS@`lg1ODHFvX6Z
z=A{0DOODj!gO{{i-@G2^@9V9oh^GywNc;xd94KoHS57ltI$F`l1lE)9(S@J;mppV!
z%FU$0ow8e+ezgfBU3Y!F)zBI`naAH+HJnIFxiKKkEFXbZzy4?ttU=(<{1meB^?)gj
zV;*k5cg0dRm)orty{woV@by4`2^vt(kn|y5Uk_}P@EoUo0SP|)3GBUp_>Y6a|H;w3
zuYOzx*aC1rTco{P;A*Wu;`MInyYJ^dj2n@Y0Ls7Edg|+ekCl=3PP4ni8NeNANqq|t
zj5$P!ff3Q31yDv5Wxw;*IdqTF!+W{v&cPRpIxf0Vb9UoLzZ|4dOE1M|m?>h7;$*p&
z-}wJ_3KRXu;u)6i$q!tR-wB*f%-NWK-d3M8(3dL(U5KPQ4=4!cmfEY`6lPDEZ6ehU
z<I6Ex*1;Te;$^OX4BHfVvJ}@tD6|y|tQ8EDA7pBpac<uhH@#JF_n@?CwcF6IpJ*QS
zA;C~?jH$qnYBsp$dk4>-OC>qFnVFhfpd$01v5n}iaL;o*8&gaeQozOkr>QgOyV@(#
zZ_M<aj)H#q?oP@#UwG=)`QLR`d>f63Z-OB7V)V6LstacH)PILTG(x;WV<p4cO1|AX
zNxFsmMVY!#*ERyw{>PP7Dy1jWahO+(U$aUCza|mxg3?^6kE$(+q29=X+@t}(xY;lJ
zVzS3E@S~m%K$;VRkt#l%fZ?mJ$NFB=A)`;Cy&&cG*WXmN*l;5AV2mU-Hoq7`eZ~&5
z71UkmhD2dey4cFZqMY@U1utS|eI+=kOySk)_tS3VaFvLF3c-@ScP{h&+nJYTM{0iT
zuWD<2^D<G__C3&m>>(!AlI^e*rQ54_1{@YIJxy4z+o}Vq2rX7J=H_8Js4FSCbBD7z
zn%{;+rXg>Su;N#vlf*HzG?o-$lptL_*UYUnQ>fHgYKF=!b}Gz8?b(^DzN!*V`_|^y
z16-=7HzQGBC5BHe@qG`-Nj|su_mh0igY@cH1|~dWZn2i@5Y<bvKvM!!6B%I6#e7yc
zZkTp>uvqEgVxM24x5cZ@sh)>}rB6N#UFX;&T{9DN5qXu--mhrG%}sH71pL#yyn6fN
z)&_MQM|?c|0rctEFbwLxa*yCQkpNFK2Og+7$@&mZ0k#FAR*n6PXTv;3+kAZBJ_!j?
z<wNg@KF~0q^=W+W`*)QYu~`v?CQSFo$x&EZ3#563-i*_vd4Q3p9XKH|9K1d!uhqY1
zlH}!2I_=@&jzgo$XADkc5z+06GuPC=9*8!H222q(yTUy<us{GVAt1-`vZ3Np0Nw3I
zWzD2#&H6h!zA*gv{|);7|MfH9)s$u3da!?DT^jHo74?=$V|CO-<?6KrZ;dyl!8tx5
zs6_M;$LnsT@P@^j<mC956yvMi%WO~H^t(D3Id`->19tTVkZ{C3nCAQTWz5a4snpkZ
zQ;qIM*p9$0=H6bxKJep(oi<Od--&<3hRZ*UNk$3HvcLQvMF08!?J(nt2bEH&Oz!ht
za4o|%WWv^rF(?ro9x~ZKAM#)0Tf0xAXYfOq344Q=SM8Q-q#I{mI0XqSo=znqV-1Lm
zEu;j2u$EbDck{)M2(Vy9vqtfZf2NRE=|e?U`*_pbl`F5bE=KuIJzoi*%J&+rC@D%%
z4riLMA<Y78b6rohlei24izRj8=L85Ax=~Oj**p8h*(YS$EP9E@eTUBz5xto9H%Nh?
z6JBDk<eVid%5{#}JUqdBlT7M%8BW{`+w?&3*HDY8-%ZC+!QXJ^H_qT|?u4E^Y6>Hs
z$bDK9e=Uxi%Hrr+eed^AeobMI-KxI&os`-6(IIXJbb+t#*{47CPvoZIOjSNBaoW&-
zr0*>;UhZ)^2tS?g<zcRugT-5-R!cc*g+2sg(pqzcx9Ji@&8)hj1Z>iOikYRY%p5Lt
zFyfSBDUB&#3ezmpc}`jd`%$QmaE6=FfF$XVJx12w-06BsJ=Ut!YW%d_#mZ^A&-Is8
ze#e4VUKAZUcx<GM*G8@T?Q9ykTu`aBN_ZHIcBkVI+wOCZzpz%<XW7=Ix3p3+F{*Fc
zr%fE;$?lk`z%GxP=8i*Mky^Kgslr&-_e;|~$PBn-!>XNZ6F;jKG{=55;{PhApFwQ(
zJgCl!+5nGv<W^^Cj)Ajo6jJ113IaQCg#|MBQs4#kYO~E1_>!okFjxblAs~zbfB1>@
z^dDQD%c^GPlgzLd_MGEJ^Zs&<qykD-E$m)vDE+C2I9lYgcUFAR;SD!nD{Xb!oHbDW
z4|H{?POy==-r9S{-=`h<AYX)+&{Yg%`O5OJnLI1e623BZj-+Pi&0E#JMeAjVUP@NZ
z;-*h9oLb)v5R;Z&ZN^-U!3m7BV<-9&4D8!sCdPdR*;j@m=Pq5J%62W8WnP0gRt}|Z
z*0KFZvNcH45|>%uFKS^wz(;`|&W2lj-TLdn|4fr)hPEnFF!diuh0oq^GB<RcAa%!%
zknCSKJ8t}o(HE$UV(&TTs`+viY_ef%B0f8&Az82<Y~3wFR)kFZo8#I_i^B_Bt(sB!
zpy)bOgurM-0{o_qOj4M#ReO_1Tq+AFHmeJBsx1x42I<|Za9#!BZ~HXvEgQrt<(|K&
zab>BfS)lSTGs(#4@(fG-!T_G3xz-$dvPW;e*E7KYX&ML8!162R@z(B{>b3EUUh}-}
zvv5X{_q!!)xF<BaFto-Xqe_mqkSx=F5Pv@;J^JmyN|8e_{#}AiP+^*p3Cr^&>QQX+
zp>G?}_WaVdx@%8v#2uqDtztW)f_-|xN^(V*BK-mXZC$J_qxBE2_)x;D0&1(Uyf`h1
zLyi{ss)f4+*|(Bpbrty8R-zMF!%ONpl5qA5L*IO`dZ)BWuuY0^MEY5Fb{IHV0quWg
z>%k;wPi0Sujk~Mo2^)8%*=(@Y0-jUzXC$7moQSw{MT4c(z`6f?!sTVW*L!|P`I2sI
zT(C~MyVXLnm(KaOVN)KvwT0aeJK2G3=f;}<*cgBK#XLAyQ!6$W#C9tqC-*Qum>Qx3
zrIt$fTST19>I$JD5GRA1bsC}y@vf&OAO^@4Y~%JZnq2luuH-j1Adop8FKr6<6=ASN
z;DTI)6i-Pl#3EK1Om`#TPzG(t9$a%R{XGc&lbO6>$JpQgPJz&|L*RD!ZxBQ9sd6`U
zq9QZ-S`p_O4(0dl=zLqrq1T&tT-I)a$1cX&^48U?8qzQePJFf`W)gLLo6s$}l=j7x
zhMdF8ru|Kppt#gCaoANk0Sn=zFNO}3LSYMX!;@o#b`^aZ7wiRRFqvxA9X9u7)jAfr
z93Qa16CG}waFg*R!>7hu#jXfhQ&f<~WJ@Y3m!!!MJ}y^)%bib|V>}1y5^x-SqCACJ
zb<MR5^ek$rJS-KgQ1B(p-}AS2Obf#6rWw3H_}itoQ4{Cg2iM*W%>QsxR6JHL1_3+w
zut?O;Dcm{eYhH1g)uHtn&hh)WfkW3rmPQYuk<bAktT>!G*_Kw<%qN?P@*wkwi1%js
zDMGTpCW|di*8)js=FY>z1JVh^a1e#ojL#v%DiO`XA`S*?SrlVAH2L*_hJsYIv|?iK
zl#qh(h|67`%8N3*WYl%U)#1}SCs62RPV38!FN2+*;B5h*0N(A)+Dq;rX&_@ov3`)h
zCQ12>{x?g><~rvYEQc5133J0VxHXVhmp%@-GhT>+=e|W-1JFV!2g?AzrU897(A~DP
zI(Bl#c-{6j6<7-WUVVvg^%zF~V(a*w51!WRozy!?76}Eo%BXd^Ej4c1KK<9NxK=7`
zYV%ceP>n!zD)q4fU`uKsz7HLu{*$xVk(wVmUK$GGm1eNA{~Pc=?Aca&Cf@UjEiEJ1
ztM_8}DH|iMossNnO57Y?T<}}c44=kYfdFNG1IN?3|8itYgd$0Lb9?4xrli|m3*Sr(
z^p3_CYU8kGNBeA(?PFr5VSx-=;9&GaKt+=L%#JbhcCiso1gv?Ons{k?=0qGwX=U#4
zo_QyGJt0xOf4ueNMBen3-nTEe$-69~Z{Z1@$NRV1Xcoiqh>b_l%iW{|9oteHr$v=}
zw(P#BO-mvR*(7<(Q2C(pul+%}X7X4rQnO%L9SIBOJb$4J;sM_1<JF<be1Dx_LO;fJ
z5W6Bj>hVtMDT?A#=?T5=E+5Af(u1aVaj60;Wt`RBk4qj+Ksf58XZV%{2O92;N0wyv
z)QDo}%w1!e%^)<@MOUeMMKm*!R8|_TQ~nXYy%yl{Z0_`ln2ifp;?;EOW`mvZRX|9c
ztG(|Aop)Q?!n1gSa<}I}Re|pEU@hoDn)zUjro`}g@+U}XEY0Z}kMj6><P;qfA=kb&
zP6!W?gn5VRT&y$naBsmd-N8yG8!lVn?x&K?Ft3sdgNXHRHsMpluYt)@tg8sDs(Ok`
zrkwzv9dp?qktU(!^td;@Pck=>lPR~yUA-C^T_r#o`|!&w>w7JBZkF)nf}isoYu{}Q
zj#(3S(Kk^$c7t*IO;D4J=#1mHfEm79@YS({a;<d5njdZK(8->i+9|Q}le+#qW$9(Z
zT8Ix(?@QL?gkZZ}L}`mfP*J2!ZQN!AYLf7#=rk!vJRRID&1GN|rO<HDuFCNkpd>}F
z?NlajhP-PNh$|PYb9?VT5=YIl<~v{=T_CSl$RiB{R{}D^2JWT$sM5TuDr-5>f%f*)
z(i1rvYHs+~L*<KUEw|Q)(zwZIbSJ-FmU*+?{h<G>Xx->lh0_Xvymp$uX3ZRnNER<#
zn}{0U(DwY7XYJ`aSouY7;C)`(enCq(oFk1h4U~FI2XXT)x$|r5%uTncK4xNY-2iws
zTIHMGhDqe6++jukH#ykEA!(aHP-CCoFghi5(%305oIBl5P3DfV!48boz<*v}$-5kJ
zvoYNJ#S0f7H{<aBcGbf3H*#+;ZNZm=zpdpBM|q@G;Nvy!jHgrZt+WD(Tw?q=n%?_X
zp_ZC%fleG~ZC&SOP?1(^RLAB58D(vTO|+7G=iM_32hM)IU~&&EExjHfLSr_``IlY;
zLXKp!O*T>AI-uGvPMnyx5u2fOGEO9&SPJni#5PJtBl*3-E8^w8jj5Mi!6^l65n+{K
zfFb$skTXmA5f#hHXQW+EzC3mFIPP`ESGH0S{%vNpO8ool4-<K5_^f)i3H+u!^$vl)
z!{^SC>QT~JnPRSqi1|lsHvuE$z=Rq;)q0Y!Pr`{go^(sEZ#ciBk><&8ay<*$1$!=<
zn?DIy2%MA!6_wSF{v0$qVHi3q3$pp-DO?SuaS-PW0t7h1Xeug{bj<bja$dM-tF0uu
z!RJM&FVtJbz3HCw5uF!jp>CHYx?j|fRhlecu3E#btwb#eLv^lLouHI<e4)5WtBNx-
z|E{@C_#s0rG5DDAo+&2YGb>4$hE@W}vs&m_x>@2ASLIDzdNbyz(uDWv<yk4wh0RvF
z9_Q}93fENF&g~?P`gBJJR;i-h+U_T4@l)E1Z2ioyG?PU-P@b^xGM%B(DJR&arOg<1
zxtn{uH0IVuE*a*ARBt$gY){DS2kRGYXZgn6Mb(;#`-I=_S(nklweI`4H9S9ixM?km
zNg1Or@NsjX`jM1wI~~tXAuw!=_#?AkeznD;UTj^mN0f(3db(_S>l0+z2Pd+KGJ=RK
zk`^;0DYW@RA1B*8z8>&!JC*rL9w%uY8K^3f4NHM5DLs$`JjB4$qJ1`KWhZfKYfrPY
z6#4+M-9zR^A3vYJc>C4<?jd|#!hdQ@@{Nz!T?IcKs>;`qao1Mh%X$lE!z2sv3$dRh
zfg($Ev-~H`CqZXqrqP*a{z!>(`-ss>5(J<PedJj<Pb^ymOtRHRb?I1)qSAyXy*i7<
zFbWQ&DbZ|g_lpjZmZ&{(rXtR#IT|9#QK2QPc45>e@l(JMb~(^>xZ0D31;$evdpoN+
z+dbl->7Mse`x|20p9b$$`4de3U45yZva@8QdjYAJZu5x43VNVTy+D1g7h0>|ifY(L
zcClF)Y2DWYaxGN$%#VnhE7e~Q5K>AUmvQ#7Yt{BQ7+JrI5~F-FvrFpf$OJYQch~?r
zEoHj9#(AIh=(jaq5&~_Gy74g?o=$H+EPtaP{xN8%SJLJ84#)B+7B=&-23!HA(Rxs6
zwpYJ=?sVU!c<Zqbtqv0!fo~(@^LY_IjdMiCy<52!{h_BX-<d4noEy#_-5^v|lo@LT
z;sw#j<1g61gF(JsMglosJhzO_%2U4<l@jMhb}MOJxtuYGlx=xpyocnW$@fH@u*^oO
zMQ0mykp5}}s-Ng>l)wV9cheolKJIU@k~{7<39WeZqizOm2}js;qh-z6Cd|<zdvH&4
zgM{Fg-r&`Qvy8sDU7OM8;>dY<S8Np&bY`4(9?jSbRU^_zLd{Eqt3$5D2V@(>#`C^Z
zdUZtX%sDaHWF?J18AjH{838M|k|`=|c<f5{Kf$&3f1!$l6rux``~1zr{Fe^rAXrmv
zR+`<8W!bX!gTHb(1vv!GmSfGhb64Q$L-_TU9PBI>Q)i_Ed1(_Uc211xrsdc3+c5*@
zS5;K(dK>0wyQrSc@Y;}`OZlv&3;Y2haQ^$G?zY>x3Uk3d1+Ujc(Knghp0BYlUFxOk
zh68THNAM+uw|Ay*^<VmXQg02vyuUe)v4QnY_hO%N&qeiQGa=>}-M`ee>05(Gy6ivi
zn<A3-VikH@#jTK2Lcd<u3B1tc6X0Ks8XPF{9MC6#V9ZS58@)Ep-h~}W_Q7=})kmxK
zEx6qk`jz*t1Q>^~v2DuXMU#~xIf09-5U$%KZJuN24yww9vCc{4@PeoUA7p4}shQEk
zfa|!Cdt}S+M7IfIeurmOal~?~z5npIW<W7=?G#*JRl9N7pkiIIHu#_H`*t=dKm4;@
z(0gIVxKpYDi~o4nt)R12FsW430AarnTNyzZFT*S44wQ^o&m@Gf0(%aX^2B>SJ3oLF
z1>Ngbx>0TROH0Vd8MMH{6R{rTXlo8lh&QB<z+J=|Mtw<Z<xq9*AlvmrjOZ4Tbo-a@
z%XSLyT(XC-*e7yz9{+XAh~_kGg!uBno5z|SX}r{2=kzRN=IeoAeaxr$=vJY}GO1Db
z;ja5@#TyyYIn=v?TV|s(?L>npIxc*mwh11sZv89U*ftZa95=tbYwa;4^$<o%lQO_$
zp+Jbck`moX)L^!oR402Qm8y|JwXW-Ef2mo@z{E?jej-<hkE-zwnFX1;n<fBxi$7-c
z9WGSfF`J{xBk5rK`wh>^Gml0WTyMa=zvpM3OjiTEhUtIZiix$6wXF4%tyrETxK3?p
zTvZq<PuA0xS%J8aSybaRdj?o&ScosB(7Pkz*PduZXnR!~#&ci)i!8UU2^f#b&{A%_
zvNwg{m!FZxnMVt$>|i4%<BVaS;hC*M&WzPecB5O_84H!T$Trn8rr>Ty+Ci0bJ=EgN
z!e(h1ZXquu1f@OoJ}b$A5*x<uZi-BLu_TK}J|Wn>4e07>A5<&xaJm9%X*q`^61)3D
zd2zB%T7Bi(l+E)}T#F3vu&7(#o6XI7Q=_rx`gt8P5EDDy+=*kDluyKm8e17f_4V~?
z-F%<#6Cxd@GB?0xO{;xzQ!mog{YH4!&cCsUir$u#$ESkdNtNZ^1V0hfV^?_zO9{=>
ze#NxiK?MTqkRs*1SHTIgjr#2;vD&x+lMS!<m0OP~yY8R-c(1Vy9mU7bra5cbf82^;
zRmNV-bc!1g1!{wC&)Vgt{qXTHe_j$V)_!e&G<j`-)Pn(*uv)*erKzu{mBoO*&llpW
z{PId>eu0(uRC@#q1OnprE(tI>d6+_fZkJ8W3`a-6mORmWRZ|y5!G~*_<Nw93&U%JO
zy*xyhWDd68iG2xPo@d9}IRbXf?qg=GH}~-@(O!pPscfl(<`+I&CVwkSC<*qkTL0zs
zg?M$p(t9?m6+G|ph{+p>V{W2S9^Hpw`{C<>U#mfRnI^H|OCj6+^8@cbpmj`Kk@G>j
z9Ev%^(9tn4Dj+QpLUF56^*iCv*;^uUv9{}x1dvpqQqU<+=BWXCY{4~cA8tUJ_b^}P
zgQ*o*J$A=pm|U47^pKtFI>^bG?$kmCIRLc8#>x2YGdXnQ*YscS=|?bmYQ<Kj$B{>w
z$99#Pazim!!hIcZgD<I&loOa?iIWy(sqLh?>#e`z7E0C|pi<Ywf*-cr-v-mm_+ykV
z%Sm;#H>tPRALywg!sX6<E`GWyb&O?lS>c-TgrD3qKGsj>=$eV;vEza0l|pRi65Kpc
z%+83C&9vPwva)Y_!!$kStbxez1_|R{uKSVLpmA+1KJTIsm$Q_L(k=s2l3OZ^N}@K8
zKJ^vEKFL4!;P`;Svi-^H&(C#VVLd&z<2DdFT81cJvp*jlh9{1;63=Ud>?9WbYdhN*
z8cA(WjGPQ==Of(aS&|X=fZ!IYBu2qmS9!3N%y0*fc|!^*vM3?0k`jtif6zEfF2($D
zcY1;KX)a`z%sA^QGV~?I_ml(B3yaAC-g(1?KlDUq<;0{{f<2*;d7;J#IYQm1V#qxA
z@rsJ-1y5AoQDpsJKUd1WqXm0^F#OMcHn2Z+7_3{JRvySm+Pfy*GNGoz<EX0+c9gDL
zF!xjYn3^dOYGT922^v9bG~M3mZ63l;c%u^;nYg&G2RyOMjRvkpvjV&Hw3Lk`k1EK1
z-G@7V87mr>+rwiG4F3+ehH%L>bb@%pUC70}j2bI-kzXw9cT9lj1mia^RAs6pf1|x4
zaW5kvK2~mMNZ?bRlj{XLc6fY#jnl~3;9$2<G>-syf*t@9{*Na>A7==~FB0PQ#iGk0
zcY{dGvkW$b24{>UEm<`ehN8sW_P-T0H?E%jbt5)#oOk8ps}OHTaMPrtQIvA8+s^|v
z=P0{6IBZp)Wl*lxS}M%(xhSqiM^hMd`>e3qM4~7;J+Hzal!SAw52r(IuN3`g?-wht
zArrgp*=tl#OosU_dP0KO7Tz=-iF{tSZ`<n!!=orx%|c9hIipp5FjiWP6ZnFJbh%@A
z!-M=Zv>Uck?&Ii<16k?Gu1lD7Fxa?H44C&sig9UwGF`85m#7do%f{P(+_g$W``1)?
zxOD_pFn`C#+30+c;lWjg)*;DZ=}{lfAYy`q1=A<~Ua<Wz=HBA|VCKiw0^_ul_PNXv
z*21~O*y(wl_guBEkf0~&%?#7k;l?OGaY3<3V7{uB1He>au9SJ^yGC*y9#z4O9jeK?
z#ii_H^68RRYX@e6xI4_&N1myvu$OQ55*3Oq@`IwV0Dej>45bYZFt#_=8NSJ;{~)}d
zx7C}ajMhO0j%Ek5Jp__UZZ6UC(6vJyiT+eY`=zZ@%M0(~PBbK+#Wyg|zr8b4^sBs6
zydZ(eEK-s`lJb!2Q0{CTAx54}-cl~I<TlRuoP}7q6_-}zN~%)ETU~cGDNMVLi9nJm
zoo4IR_k~%FUVv`0NR9yMjJPS&X=wT5&Rto{k=ALcrKR#}pcF1YhtHEiP?ePM5d5BV
zJx#Ed?JKzip)BJjgQ`;LrGot*?)*)ky6ZaBQJ-yREJnue>dnj%*4_;{51((FZlgHf
z&4`Nslxcymh|xIMZ>V8o^m#YYF=kahMKMY&@Hbg0@1s>Yvk0Kz|FI^1TQV|H93Knn
zOW3vL>UT?@&<;k4IHNVf9wg?Ez@fqaBTX;!nsP*FEDD=avC2H2wy89zo=(l3p2>?G
zNIQIMt?lgE=6sKkaOU;T<FhUy$It)Gdyyxyw`<ugm>6y?6y>j}YUrq^;`0z0KO<-Y
z51DPZmd#I}5BAlZG<(#&s$LQug-8mjutF`8koQF|iE>>xs^yusMqlJahL+iuiv!1k
zP=}WrD-krcwuie&I-lt#uD4nZ?!k}cZvvp{7^j;7Q`C{$s_4iwj79pMH1Pkx2<e$X
ztTv-bjx7qW?8mDmV8oFLmw%RQji$wPTt@OT4eMMz-x#SY+UOPPgWzx}Xanu=^+4~7
z7*~Wz4bv4>JN8S)kLh}x9y6IkO6pwUW`SeU!1C7vm$Li@N-EB%Wy>8WxyXBWGPDeE
zmM5b3o1%PCB&4ZqSo0+t+S4><_w$;<wdk1ch@f%#^@xxAcqtI^C~5&#j>MR=*4;_-
z6sz9mcVvc!M?cs!IcNS3`f;+Fs|boo_BDF~2@9(HWKnqaTD9*e?(OGh3;D*l(cZ1`
zr6UV|JL~siUX8pfus*NxW;yVU^3=tiGFQ7iS%mD+w&F+kKh=?a1^JF*Onj6Te)4T!
zb!c3`#64R#0^v^7R3=He+E^_b$DdNM^$zMl<yV`!c=uhO3qF<}IL&QO`6CkXN>K8k
zB-Cng41LEHIqkchOyCO0=A!v?l#9ZgM&KHH>9buAE4rtClhYy(8edUvKl>iNT%Kez
z{X9BY_guZpQBp)*s9vkEnR2`O!s?dwpa}(is|cG=@r7>xo8q^dprU=NknObI1s@zN
zh{MeX;Y|W%iM;j3sOMtHz=BRodU<}}41wh#?1hn5osJRb8wE_WOa{4yzfmB0pYXY@
zF204f8-$mG&MlkGQ?rP>Mt#zx^(pl|-R;-w+9_$3l<`(jJ!1u_`k<cdcDFxMF=uE1
ze<IeS_ztl+%<|51!K~ZSxOJVMwx{2r+Ss9FHX3VU+1XRQ{}A;9!g|}ValQArVEUPe
z9I$nbTZRc_^CtlDH3*{SI8S>g7{IvqSHfhg#%dR+BAtt9mM7gf*r>qfyx2W)-Iktq
zgB9Uc<&1FjiM4m=!}iaDBeah*d!mtSOLU`rG)k~;<8F%_?jO$4S=bCC@s8d$DLPiy
z|K(V|-^I`QKTJ(uuR=A;qB`A@(<QzyvNl!Q|7h-%97dK-yH$o7pGPf&0^Q78GXKmh
z>d`_trL<EzU^x;IJd$Idjd;BLw&18I#`(QGqJW&#Qwn3v;#Bl54ay7%Wz6Wh3M2~O
z)0t<1-4kb}MqBH`qGYICi?))IR=<DAl3}G{*hJv#v3EEia%=XqY$Q6}dt`siE6%2G
z0C(@vd_T2X0GYwgQM!x%*l~6H31f5x(bTDFsC%Ybt$(Gmqd8vp)*5$4kQ3Kn!SNNI
z`Sj)QW&aG#`^$#o2l>e#x6shnlhmTO2y+Lk=X29LYC5%6n`B2^DBtH*;D|7@2vT*{
z-8-2*g+`Wm2hx??T@v0iKBx7y4tGln0*(e${m_=(NH2}xmzNZQTB~qA5!hyG0v46G
zyqQwC=YE+xa5UF!xx@1jzj~=?-goqpxy}14G9;;;bJ6m8YH<{#dFdLIjOxfIYer2G
z01x@lx587Ln;kOS^Pw=hm_IyMQ;%N|Uvd`fRcE$<adolsYq#XraGni;PYRsvJqCwl
zu0R%^OlTWqN&vYyXq*NzQpZkAZx8hlr;8rd&VE=vZU9mBdme9BR%CgC9zv_+^8rHN
zh`@V~Y%N)Q@!Qj|_-755mXQ4YKwHRmp-+KXegl4C=wG8Li<ch_^^<@1h)lvvt4#B5
zJF62eIJ`GZs7j5Xt@TPTWlJ6-E~a?$#|q~3XB95+64_mZ^?bybNFGw)uT$L%i(+Er
z$`{}CNtH(1E6f`+?(&w#l-zMBLecA*A4$h$<j!Vx#F0=J^evp)JuybB=dREwjg^qJ
z94wEH#tdUNdx>K(#V1auGFU5>>%V@C+%(RfZ-M9s)eZMJW)<Fi53bxaGCUSe_Ksew
z)SSPjFyneLia#WpSV>r7dk&?4J%DD=t>0wb%rNhFA)FNOO2Jb*+-q7+)d&61w=hgm
zlDWpnY=ftdaxfl;bMBWKD>+?9OwQx+AvLP$*Lc=y`?I9l(PN&?{u8(%S-8n3cOU%}
zFOD=pC0633s$$i(v${B#UhsN1_t4T74#OFG3fb2#STs9h7gp%O=)LbT)IZ;s9dT1H
z{dvn7kzpFvx#&unT5@X<v%|5lHBjadvHy3lV~s+F+7>JuVg7hRB$<O|P;o*;jcC=U
z+Gl<I1@V%_bGxGWDi?Qe7&8!u&u#3%cH+txvco<u?Xb-Us$1`hO0v!G8OO!=%5-cN
z265C7YAL*rM`qf34JwVqY1G+pcKc`Z9oNl*TdgAg5-8h02$7I|%AWg0DB1Maf7(2$
zuXy?0#oflUJ!^GqDd3?14gAtta<5M;8?nx``eE4aPtC{s3bGs9aovET{S>s4=z2J9
z^!|M-WwK*gk*vo=L)T-ULi1Hkh~FAFHq7#4yCf#k8Qcm|`W>-xLQDb7`Cq#3H%4D1
zO(RrE*UfhXc7~zO*a{OkiwruXW@<hwx?-6M>%0L-LgT2z+{e+)GSZ{lVXEhklkS$E
zx^zq#>t_5r?R?M=jZ?=W<Nq7qrP~rv^UTji(j$Vk<eyWw9W7iI1+mxNYYvzlMi)H^
zfZa_=a3i|qVz;BvYb{M|xE|bPj78?6Q=?QGf%?YIHb-;1iH&ByF*F?6L+K$)i%-qi
z{RmSCO^a<=1qD?3y&q&>N!-o4-e$46Ule`NWeAN0;+9NhBh)VDNamPJ9Jk^FRBtcD
zyzG_6hwI}RdJPO-v`U(bkxd(XszXcnXR@_(zDwoL<cn<^@5`R)`eUCw*WOZ<I>#=;
zheej!kc{s#zPI^N^60a>cII}r<0>wQ)J<INz0R+ctA(q4Fj>)1(blShNcNhMGwpT&
z;T-ep_mVidByIpgul}Ob$v(7c6M>qEfS&DBmRn|j>n(CC11=()1j;+Rp;jqGr&E$m
zRNB_6PM6^pmB4M|JW_rkeEdT|@nVwY^Qaqgl<GHbwvL@%W|6r+G)5zEg%2MG7`tSr
z5y6nL1}16v@9|mJn(sCv{zTb8dNgA$Fr&DgY00s|q)`7cjK`?Vt@)Ap7s|`#!GXLc
zX|WuviM6`8Ix?2@`)5)jB&|1cLFC^{Up89`kIn1J50CNZ4^mGB%=6g6r{m<!!gbs2
zhfZ-97QDNwTh|ZXHUYWfOvbeOjPkmUfMb1v9}e~Q7H&HHwJ`Oq)cMuM44U%e-tI+3
z2lsm>F#~apSzf{+=K5?c;UD)udn^%lF<%dyB+=C0CmWCf8Ttn11}{k>w`U5}YkbD<
z$=&q*p)OqPFP~cwQ&J*)T7=t7tqfJ%t^n`?2RZEdLDJ07K75i@118kZxr%s;e}vxU
z3CSY9u*fe7UoI%TIBzX}A8nDcwiU3Y-`=Xmn!d*1Z;=b*R7{hW*ry6~3jgxw6Sa)`
z=QJ`v@tm!5e)#%q&8w+%+W6-axo5c|MB^S$Li$XN-Zm5{%o{#VGcY~F4G7uik{TUx
zWva4lH(dMXCaF~Ngc7sRFOVf+N`%W3bb17nE5{c$)Alu@WsjGqDh;U3vWn3$j1^3-
z$(w2)wQH!zht=_z!R}I@3%T;<B|KzP{EJ}^-ZG8PU0a2d96$T|d%EWS;dkA?^43V%
z!IHSN<CiAowokcj?Hg+(%B-al%FmK#v}5v0`=5-1x4T^=f5<iMitO}LeweGt;dT_P
zTNlA2%cKfEiPJRL(L?e2LQ0{Ay<>m1fn*9ju@JCSK@DDodyl}K&K%B6p>=hvt+P$Q
zW{HH_s2wlALFu+q(z=81CUguu7JS{?7Nj>HQocIK;+I6IBUuVi4(p+CCbV%m`kC-}
z)rY%V9lo-knYU?eZGZ6omWL(Z8(Xu;_sFatOd8^=MEKW!0@XYAf9Sd*e%r68xH%H;
zILoo_^`@KTOkB}6H2dYnTEB8;Ns01G|87?n<dZ)dH`b`G(QX>k4Ghe!29j~IaCGCs
z;$1|gEtn7xkAg@O6F{zZe82-*X1t4D;Pz&m=QZ|5K`#&d?X%#Y&{A3a@ORL3)Cy8P
z$7Hf680H*sl9g5WJ(zn(^Jg9ZGDEU<@-;quW;we4hI{+W_n8F*M<5>mBm0r{axgm6
zjof8pJ;kl@t<q;itDgj>`LK@=zgfq@YQ2@oL7AKDtS1B-h)j}2Rtg16@O*g6Y_}wX
z)|psk^?%5Tg}GlKx~GT#G}$-O9l~xQ2u?}ar##%f!A4A*u@l41v~Sq09uwdn=-{ip
z4AlfeLic&bllQ_{|4(mPDq>6KCfG^0akV>Bm&pHpS7Et)r+=+N<2t;99khO}1(dC+
zw@R~)*sh&-93QvM5>~hNCO_tl@0TVSE@tBC=Eu>EcQSPLJfMcbGBbSwI7uH+=D<O8
zcN<4m>>gM4{6QVW(pKx{*tiLgR|Il*?F}w{6fU31OGQ=MhP|gm2~ij4Eo;ort6`QN
zq&x^wt8SXC4o*#BmJg(sEimazg05WSGxweU_Vsxi9{AUgpJ>rk(>gL^zX4C2XwC4<
zFvv}tZR>T^MQeVXvc48Hp}&1!B8NrC%7d*Zpu;{ba^u_MamMl8uH<n&X;OEFP!87c
zP6VbMo`d?=nO(`VDTu~~-0)VdiFu)FIg26@VsXsLcVtK*wx<~QBtyfqFTqxh?-PUl
zCaj-?s&|w%_!QoHRf}nRx^S}o)$mgt_TbYmQ?OCEG**AmtC2n)eD%;x!KE_ey$w<M
z^F(>H@4tgWnVoWTQ;4q!hwb$&C_~VCD4m5^$^{G-uU!E^qkz)vB-~^<3Acb>vqcic
zXRkBknWo--Y1$P4dR>pJ&QGmf`>A8c&eHQD;^&nBYfj#dzER`TfYbq<$-yUIR1@e|
zWF%!S))|_pLic-0(fh1O4=2kEY*iBFD!yZFu)Yhuj~axk&vm+|Az0fav-2XBOdxmH
z@aj*RhZKfZrWVe`UhWrfWX`-FXb1b7JMi+&gbQw1-A3v&21rTO7L$w8om_GElHzW+
zX&6nU;|b0!+I|aXH1GV)0z}vqzbuS346ek~b{kkSjoAYbmvCMNii{`Lw`zn~fzR!2
ztYl~_-GhN^{}F1o^VMef0C<Eo6PPmqPQZoS`6R<tn*ww2MrkI#r;-vqkFQwB<<UUr
zmNd(n!~F!($xhV@Gs*(C#xGQJkz>`oWQoC+18=}9`)AqV!L2M#KAxhFVe{$aSAMJ(
z!GJpa2*IuX&C4>+lBH7<&yAuRH63)xx;L^dmqKo>MQM!v6iel;Nvw!zbMplLm?Uy)
zUuC<PnwS=_EA3!S2E2_EWZjxhB82R>(ako|&2t&&#)Or&u|@Us94Sl2mp+#@i7=?9
z3ESQU)XoHPRmtrg?gn7WAT-VeOA#4fAmNqv%ac?(Xgs2>C%-h22}DK#*vYDTT4!Dw
zdee+a;56h%p_ycZD*jA;xI|e6%n0>3w5-w>i^qj=Ks(cl)aZrbhr{jXwgQ8CtOBDc
zssecK=Y`h_=z3rsq?&$llCyVrG`lPP`V=+P_sJ!3L4{Q~x$_qj-a7iOt4hfKlsQlt
zvK`t<N@(l>TECOp;SftJyD6SphnCa0gr@E}HXwZ_<x029lL;K@vji7!7yvw)h5w^?
zOG{}HXOqq6TDc7$s(&5{5$Jd5H(3>2eG<g_@W)9AzLD+wPWZ{lFLZfjf8p7yGM$>W
zi?#<>g{2?uBf5Y3dcesXjd>h&7`S;Hnj-at26L?Ky;+jn33G?8t_!p`#M4tUH&ZKc
zMRzoc8Ib=fLF+SmswKd9pjCulYd!2Vin}S(L^p@df09-G4>f@x^8o%C2PkCGMFGlo
zG}oL>p+$0+#)(wFU%*xlW&!s@Pv|#a{^E0&=W*@X=2cgJ(&cFDLK0=lVgvPCfF*s$
zri0C}%f1@qfH2&KlhOmk>u=qUC&xsL@}~V~91r2tpk8+(T=Nlb=n8lG%SPsQJCOfO
zGzRQ4l~&CQ6-#E;nbCgK<jCmCnQAitHO<r0wAfr^2Sxa<9fDZx_ks_)zNf;SGIe{u
zSD9NJSWmYuoiQkgd|t~Iz?U?tZqd}p*#2LlOe((Z^N1R_Q}V_&15=}1{jRYk>Zf;q
zXS~+N=S<}6Aw`G@M1=q*v+vg0Q<0>KRr+WZj?w~J$K}}J%DY19)0w1iy|FNb?z%8H
z$MaU<HZw)FkW@STMXHm7fE1j9X}qVH9b^Pk80(UNk9t)b+)EJA8iCA;&*8K+BMEPW
zn(;sAWY3yJ1F93?;^%EB1@vK3DXYZR8lfXG3V-mZ)`(G>%klHE_&dj4<wq6n`mt<_
z#;#L(mYPK@iTcH3{h;LbiJ<^@Rcofyc_ZNi`;9Gnp{!nW>@y>MpxGK0_-5Ktm0lk{
zP*Wo^DInk=lbIMjoazavw&9Vn)*^2Z7*DZ}z+sJ(bE!ohoHXm(Y?F#M)k#vKOg}?s
zVy2x@TvJq(faL*`3E>)(vFp(v=$>hDJ^ARqalrU=;mU=?WG}tMg>!u?_g2Q&9Kux}
z4-MkkS01-rnX_V+p-38MRwx-46_QU<K4rN3&Waw(WwD4ODW^Z`{wX8mAdeKxxy`-h
z1;AX|typuUrjK%UFB(!6?AwM3bp_c5#zA`nG)^Z2>bV@7%i@S6sGM6=C){WpJ<Yq!
zm1d|Pi3yUHZ`M&0RCJ_qC&o6>n69LilAufKYdnJwX0H`}OmK&r2VQ1NuPODhLy<m1
zLl+D2^8}qT1FLAtsjf!Ks+YHsHdPUO4W`>1>^JWCTy|tzwXOT7^q6;(cK;n{vr=d+
zAw)|YbDJ2A;c|P|QF@$sDIPNLKDS-Cu?}3b=}gddsgt|XGaJ9@$_^TbCCyNaLtkkH
zDIL!&4To{0<}Hh$;1vfzZ6A%Lo%LpNELUw_sXWY9sh(Rmj8|<e6HY51d={GMfBtF2
z-yi?ztf~s1MpC*&uHx<XpWQd!Y_#H`$35=8rCS&GymZNdZQZi2!KOhn?7ENaDNEu$
zC&#X59gmz&6d?$?2s`Hc>f|7_SD)Ownz;3xkg31GxSwxwS8jOiHNP9+O+=L@X@bbA
zd2s*7lv)bl6dD~Tk{Qb?TGxgq07J=1AV}^UE3=}&-8%vC$yJkJux=vOA~J0D$1-~6
z#eMeiNJDz;mC;w(#|On}Ed;NlO6_GR&znP{dT*K1I|FOfS+_*+m^nohN}c}hr}c>+
zMH%i)37s(Wr)FV~2Cz3R>Ne}mJtZoQLMRVaXMJO$;Z|*^EdVqUiPhFJ5i?}5f7sq>
zq+1^BU{9ihKr3SNq9@i2|0y+{vZlzAd{tEh;lrBTP&595)c*Rr-d&Ua;RM%=q-Hrj
zB4^2BeI$DQ`{R1MuV(PI(K_b1?Io3G>_4`l)~PN7-#r=XU%ZZtNY(i^J4S1bIwLAv
zP&lt(%5)O)&Km;BAh}^h<45_A=2Xx@<Db|4K69$a+@MQ5|DKBG{T>#d8S7@IToIu+
zNMB_<$`wKp+)_Bm6m;6N-^S@|>+E40CmwRagVhEf`WNYI;mcmhP*RQ)O3PqOD4x=|
zI*LVvzKT2Rp6RFl#7RkmzEWBC>vH7g#rzNgM86Szy*DW1s_sQ-v~Qn_zHowx#@DW$
z9|W&3p5vdPx(%HST;VpnSN2Yry7fo<Pi19!<vr&aomgV~=Z;LrBH05F`WDt8I(h4S
z*VOyU7hmM*iGq}##&h~Ivb1CuoYg*+GTcj)7lgui0BO>_O>Oolk&HyWD>RCHgJt7#
zsIa=!-UQqjZZo}eeg1ReJd6TfeapMV>Cc<5@RzXMo)e2S|8rGW+2zzC4WhPsYl5=7
zwq~g}fEzurk<a_o_fj4}SYq1?6eND1N%R`+HvvIvo}JB@)Tvn&(;Bv{U#*1}khwID
zahDmoY5u%0s5S~CEVB@^!cM+4f$8+X8kqmjnn3l4Q*SC5dgEvnehQ<Ad&N|>TXYdw
zrqe59vE9qGh@8~bh`7Jz(Nv9ROO?6~=<ONRyf<S;={=m)sx7sWXE!SJp0iMA%`c(W
zum+#P8XLPR6r}N1%w{6IKw_oWndLNCXZ{fNl$y}j>e+`)HuVbt1#&0VKtcsbfgTUA
z6{;0Fg2p3pM-AC*kx{FRQWwcDY!cv5p$p=YzSzOGjYMm;vmg7ruc=(wbQ%|j?@57b
z<UA`KK$w%$xEKhq9PuYw*H0;VYrzQrn^^L5!kz6+PgoA*vP}rq=diP`%6&ERN_Yn|
z#<gI)d#xfu$FuPKem5%r$h8a)2_m8(>xLpaJ6p4Wx)riIrXXO4y78>9=>|;L2Sc6-
zN&CoX|KZo@2{1-QS>Io(d|_$UA&{4g!}Yo(XC7aVg{2A9I-n$rlC-0@(aK~0Q{@zS
z1A$fTY!`qvfxw=#<*VGsxVZTF<t~K^=+SQ&8QQh~iqo^s(J3)A>1WqZm6ULL6Rgcf
z`9WpM$;ouMspwJBuTh=V@7EojPX?_$QlAd+8~dR6;qB|98($Cnuod)0tXaVJX-9A8
zHpM=%{6OR)f((Ttu|%RYz$2;OSJgAfX`GnegX?)&j!b|Q=J18s32|1ReC%bia!q4o
z31+~+EkD$5d0eb}Zka%?ZVF{G0>PwyO+dT2BDN=vY)p4`>23)+3+7gzi@f2%-(!Ye
zMp_Y{&9(S`ls=8aCVEYTcE(ou;z1IqECE@ZYC5=>9D*3AE}Z?n5ZAcWCQIHv(Ppc$
z7sW`@%`R?X{IeLNuNXI7kl=yw9Fd1ML3Z<xYGZV&Vtn&8F)salqHrmS!187r1)s>+
zzDb2=2s#XfoQVZDaCfNT5nbhxNuu!b!X5DcVACgBMyiS>R&QOy=mfV5Ccd$t+a<4G
z!)E?1fIC))mbauU9)B0mLi=UP@-!!6nrNG#>0O3OoW6jsN*cAc?Q^7`9BvJ^3+wE8
zsQEzi&^E9w4L8gQ*e{s%nd_z`A1@B%Qy<S%Ka$v9S<<)Qg)lLnm?|mM;1;rZ7Br>2
z(IPbXbUJ7nX`d8GPfTTm9<Mrfxm@ORYL+T8>>nb_?*^kcv%)P3QM0F7Kab9!x}iWp
zrd_)*{JVqy9VV=AR0>$->8ciTCAxU;B@1;#+Xt>JcYe~B!?+Ea9)*uJKqcw|cof~x
zRq^m<)_psBDD1e|&yeB{N%f_e@DN8rwOU~op84vmzLOy<%rjK&5=LMNOo;3Nx^5L&
zfeoz~wmSm_J14E|X~2AtO332k8ptxe)fH)Lv6CFb6_8BXeG(5~R|5!g?7uaMQ(cJe
zUTs@9^p)7VMZbEJt?<TrW@pUv`NuD?_%o>7M8e_W7Mst+jd>hJwi~cB9uoiOpvMn<
zdat(-v21%<@~h;V&xZ$QonVSKZZoW{`n8O)eV_gIYTfvNe;LbCvmG8c>nMM_G>U8-
zQDA?wPudN_Nz3G6V^bnN=4eXs!lBO`F5B-zBAKCm@US$wN#=NLrhEVW_`L+El$S4o
zXMu|oFqKjBsLC|)$SCTKd01;eE=MOX+Aa-K0A8G$R-ip%HvZCA>8}cVJ^WHF90UfW
z+Zpf1&-n~S^(7}3B3CKvH*F)X+Wi4!mB)7M{TEOl^wlVAK8zUPayX6Exebn(QdROk
z=F+d@aP{6B!}&hvq`^07y(UcMruo7=iwaD|n4OlWf?E>7ubgQRMwtKJmbv8AL=v`0
z14f40+#Sf+o^n3X{eS$ycr{_J+s%r*`k|F6Lj67GVN!`c{jc%xSwtb}+`o6Z;5YYA
z=Zu=sO3P}9gI8{OeJOw?pMn;vo-hj}*R<>x0C*lp=k^<Zv-gYmXrf1t?mZ}tkc$&C
zFwnG=a?divqU@5)%tVmfNbAvO{6FlyX;f2bw>4U)8mrz?w3JRkv4kQ>Z-j(i-%<h!
z1jG<R7bs~0Nl1{s$5Uk?gwTkzNd(Fu0TKw(gph=$l)hC6A&?L#>7)yYG+H3Ocl&<l
zj&tuF_uqGa+#hET_A|l&b{<xqojh}{x#sHT2_hNsK;7Vhw5Md{58`m5-Jd0Tmv+_<
ziYGq_zGM|8qfa{Z0>w%I?#2Mpk<}=m?4xgW=9#?cXNb3f5M}U=U^^D73ht;c*M&3*
z2@S_m6!KRf1)pMvH!d3AcsMt2x6yCoXqO+o5tU#+pO$@KX8(4xCr25oA<L|K6%%tf
zeGq#ib8V}DqLz}!3Vc16tgmS(A-_LF0>E*SYm4%OL<wIi6iF;gojIJnPqnnwPes+(
z(`SHc&m0t992g)KM2bcMlm2L=jp?q-Nhu{_v!(s2`bVMRb3B_mU%R}1?4AoJ5AiJF
zB2*F&O0H20b%C}$sUZcMH(V^4x}%|-_+Q0e4nXnO)C%q9-t3DLqm|3zwpP&Fj`+2x
zu&IgI*fejQNI+d{D|4^$PiYi4PXN_lhbmnwW}u#{V-4(m8X{sr$`#t9eIozw)n7T~
zSjj~7uQ_X`f3x3Ib&ho81hZTE7<zRgX(h8N9T3X20Oen#a{lXTN^|ZCCOEueyF$7s
ziTpiq?-5Iur}8yT{UfGfA}ZZAX)SkpR3pz&tOXXgjW;|s6PsgKSb$dhhbAYR#tv@J
zw9qDYRo2lO3nf(^YQCxM*fG@i8ve6u7id7Pxwz@-KU>2abwQ&6n&#!nW(gFi06YKh
z3b1X{>Ho6=tQ)8R12jzlDP{v6+B(N8GyCyO2le!v^3T+uom<NE%oh~{m%lF#?HArO
zzKJu8ZfXgaKHQ{Z_QkJ+7aiHrgq>+^{Xa^JY&(9o!teMd7_Umh!d59!aj84Cj@QQ;
z-zEEQa9|xzF~0;X`0hr1s$$V|JR<-M%>+bZtECnt79(40Df3H99$fdV)%Bo2&vKSC
zGng)tELT$Yog!^adrjnJwNutI&v6hqKBjMsbuEAR=-$Qf+fQjf9Yn7_DyZ(T7^K<G
zCD>(CBJIa_$S<y1#ZAT#*J3W|t?t_lyHA|+L3XDwTKS<bmBS@$k{qRx{DGJns|VPZ
z)%W@Yh$5_FR>_g5B0*U|nvAx2H97IwObYFn7)gC;&@MH%OyYy_NHq-AsrKwe!%Oz;
z=(VJ;OHy;(x*gft-&23L&m07g+^oF;NlCZSOIB)A`KYrKw51=uaz$NEQ1pG(oNTTh
z8@@O@#PZ<P&9GGAV;jPhjZoL38qDG}14j*wjM<GP?lVIe>=H1LkM1%&Y~fC%-OptQ
zFZu_qrf|)^m#LJiU0NX*z0;*)IifHP5g(&P1^h<R8KvyFH_Y=68~9cH=JU$-q^lP2
z2^)Q0U3Fc@k<!24ep?X>*@e9SOE`Pwiyih-JEQ%hTH>2OH43K1<Ko%-E=eE08WT|W
zsRwVdn`K`lpd$Sceklqz+kc(-?@Rv23VJvhp8Ub|{mcLSpGyNV{-1#S&kFhf+b#2t
zuV1ZN`P@^aeL3?r|8I?C%Xxy+o$%@KN3}IU#@a_0FsbH3l9t@)k&@5L-2WW<^S^7@
z|DV>bUlYO(QW-My9Op5qLR=pQM>^Cb7hmsDWLSWV|2fWR46d)N&0e(7P%L8JmiQX_
z+&ZrLmv%nSpv);Rx;y~erhaH_?Xu(BEBi2WZiQxYi`!-#QpCZ^sJ^w;C|ui@S?)7R
zzl;>sE!ShH^QeKx>1}5eeT2buLKI5XKtcYx=Y9uX$eO5IEev+7)XNMLO`V?5tQvY8
zFIY=X0lA65<`estm~0DHDLa5+)v^BP-R;9yv1mZ#rS8=6!BGKtV;tusINhVA9RJrB
z8+H7DeX(!l(Xo$iHvZH>^aGNb|N0_T=dk#=#y$S=wT}1m*dhQb{qq0vM^|?Qp?-%c
z(;X}O_x|<8_qYGioUNj!+LXn|$4GV@5s-tkJ&?J;eu`aaYsaxj1o%y82w-0F^1D7w
z?6@BD+rZnMZtib7awez5ZH9mEM*cAgIpSu6-|`+i=rUHbcAmVVyD4YA?x|+Q#rk(=
z>HM?YR*aQU3r_P34!t!XhtRz@9h9n3{FRSrmPVGNsv?LC-P#7f1GD-XQ$9#PsV_?d
zeirS0%B{l3fH};yIxVA!dYxPKW(KY~k;}0wf6`?X=I}YN&m=EmT7-;Y(HC5iW-%{X
zu!Qn7*pX<LKSmqX@o;+%``LhrjXjsqF%<MOBZHw^U$v(cH)HrE$in?2S^wf&8cX+t
zr#Cz$YAOVlcqzf}Mu$V4a<UHRTpMl9P<G;kBL4NlJXTFb{NklB%F*JCn7$OlidDZQ
zf+Ue8479+bg&QU=TCM7S(aX*8A4Ud>b9dG|PV9Z^Y^32~l}7TRgXXpV!IuKeC#0KK
zC1b+I%wT@r)!By6<V8?5XWga+Xr$N;==1J7V>GgZN|I=Y{jtA>!s;v4bE^9EiQ)Zs
zS(n!xwVozB-hb#Hv!`5i5^rdgd+>d{Q3ncYJ>ZS^ngW;PKwhZ${ib2I`O;o?kUFQ6
z#)~VAe(Yl{Gt>Xpt?`~QY$B@lAl%uGR&ysH=#K9|*Y3ntBLNX+tRc@_w{~KoLU{g^
z7#a(=qTYK$KwpDyAhRMCS$iAvB3Nq;eh;@q{vgJ4V`#CK7y%OQzQ$moh~7*}#`x5P
z@RBTlxx)EZ>_{v#zUP}PiPy_+NB6M$^^7;FzY^bpZuV4%hOZ+|5n97pFlSEck-(dE
zT}5~s1_<;YU8yJM?`87e<XF8o>q>sfZGJF2oxlKTT&cG*a~^MMJ?E$MJ*gg#eAzS3
zDZwx~4qIC1kAH%ZCk$sB)3T6*%7Oc9?`f#eIl?*33DU2$YRsL0X`ZhsF198R-9wk7
zqXVXR0qQE6T|U$eS6GzT6dBRNs48b<1T;n`eX)0ve|BjWQH~6z{-ea#oX0)Ybb0n{
zZ#bFU&FA{G*LND%`(!^d%&e#>sgFEu2YHkK2>E+w!ia5uLrmoJ&9SR1XI_J5Ur1sj
z_BPX{i7pKA+nNveQ<=DgW|cCy(x_btnY|+EDJqLq88+PsK(L%yJKhN3mIq8J47}>i
zjDPdi4_`u$W4Sc?l5eSi6Y@TMJpkP{G&I8_C8}!<o9R)O7Uq-BPoOj!*hC~cdf(>R
z-=)K81hEZMbS$Uc_jRL}#j6eMpyy`Qn>)Lxw?8p_5_`gr`JL<4@)@|k;+UptT2hG2
zBN4Ct2ttG;U%f~6d&(Wt>5>a5O;H{1P@Oj4-s<hRUt*=|JPzJ%9X11vPYEi;wX2fS
z>ewSgQc9&b#C#tbD1>;kK!8ImZ%*2NWWDWM=tRZ6-3I~+hHoS}BqR*S6>g5AkY?qI
z=QX;`aBn&Qk8<}pf4vONmW;d_>phI^rOgw+YIMi`TI#vV$#M>(ClthRF@a}P(!J&<
zWihTXYs_TS-$wfJ<c0iGMrHB%gaJvR;WxG8X=f>6S3(zKUBg^{yZY2>O!2vmfYeyu
z7^QH)xui)}BNlbM-8Dq%-}sZB&IP`!!AhEos(@Y{Fw;?TT3O=l&gWb>^iusgU^0g=
zr$xt^G{!m;h8-4bhP7_BnV=8$ZKS97-i0~sCVb4qPfn@e1dvEGBTq*6pwI*<Sz{-Z
zazoC31GQ?Zy<RQag;aTMR)6^TA=X2**El^Y5&BbpI&)5R{-@q>pID7vJSwUypKEmX
zW5o~??zVKgp)#huyKqrwZL4vIobb=M;pWH~CscEiwZyB?=A#12d9uagL%gJ?AQwL_
z8{^mnxQ*{xGw&LCQP*dC$8<;K7+^V*X?D(r1M8w*POHk)LBN%XYH&%-Dpa4oVI%jF
z(+b1ufY4j0fNMN@xn}gLjH)_N1dii_Kv&aSG9GPLs}X+1ANfZ<HoMB-jWY2cRHTh#
zzCYzGXkE_;UwZp2G_cn*@_Nk`q2dLnJ8*binTF|yQ(AEHxxF10o><b`V0JuD04;2+
zBFOunU|h?d;w^n)aeud1xX%Ih2<XZYa^V^;K(o7#-Qx_AwoIvP%*%Mhn)0&B*SO+%
zwUlylu*p%QX-R)ihb;=Xt2mocT}FV|qF}CIG=GJ`*JJRpOpEn_kAr(%#D+0JZ?sv&
zX?KWbKts@L<dpA;@Z(lT<4TC_*_%K9X<&TedJsGBAIRyN!i4mPb4#^dy3Prx@=t{w
zH<A=om(>`;w*p@gVTvxwO4PTXK8@P&VE|<`?oWPYY&a3FYVHoS@J#>m9IIbhVq)u+
zhG6DqI-^pIZ~yKD<lH?)4dxPa%P{b@j84VGlVgnip7RnibhAgzn0@R-C-KYDxrMU|
zdpg0e=FmvBr{jhYmSpBmV%ePbZ-GsrE+diFk9u7~HyRAxx&2$#bg4mo{wK<OxWwmX
z6G1$ol#wy<TGWU<)%YE@<LS3!s;alP{M<y*3LHLQ!!c@l$gOpLJ=VApAU{0n&^<j_
ztcDqRF&3Srb=0n>F)Vtu2+#O*<%zqqk7Rg&`3W5#+pr<{bRtw*7_*z1Xdy{%;}s-Y
zsa$F>=#Hnvt6n0aJF3xrJ^g}$h-%^13W*^A!`cbn8H8>=VW5zX-B4wjgV2nGtZeaO
zF(@`i;y;Bk!%3!pjf6#>>~sI2Y_(7*SI@L8$2bKSQ%>%1^wifKf{Jph)Pp0|1ZH=a
z1$CjgcuVMODZda@SB@)=hFO?fK2TT<oA=|2;87@HM0&=$jJ`<QTN0SnPIG_`z>iGc
zhC{3kYOStxpYiZ)HVOy|m?lw;rUXu-XyvFO<k_x0CWXNa&Zq|ZErKpc36H;q#&XUL
z{v|CPv);b@sv*nj8K<EqT>tUQ?>L>(gw6B(Zuy0mCc==vog~#t+`Is#pupNe?V1TW
zd6-pgIJ@PtlNmQVubOIg{{>-%n`F}QbR#Lr8tv`%{1xY;O$752#6ZFO)F(4_#!_w$
z1%K70SFje~!0r+{A?zT^_BH-(9e=S!Qt$gG!m&V-^o+Tx(R6ROpeQX?1++_Ud4Cb`
z^Fz#o1ks*tXq&4yG7h#OLos2PcYFWAk^OyV6b$dgZkr)iNI|4GJ16LAw=Z?ZcL^@E
z#_9*9&BBs6st(Z#HHkhEkx$2~3r-A?PdO#Wzcm@V-u=>^K-{Z;UAoAtoPPtq%8r>S
zC)PUR@oh?Q73TO(Vv(Vg1?qA+y<d24$i_t@TL6tDObSv&(!rR}EM6Npq(^+$e>aA=
z4~-^~)gI`W47&oB2+S(3+k1BR0MUcZ)?<&0;1I-#f|;Pr!0=6K%AnW#n~~|Zg`{V1
z-@mIh%dne!d~LC}BwW$SJG`&)k>qqT@*#Ge|J#++M}DB$u$XuE9D{SXQLFN~^{e0A
zjTTJc>hJk*$~S1O2qid_81UDOtUw4~k7u$9=Z$r&R3}{I{-Q9Z%%eHS9{mKq%v`GD
z8O@EN(5D;oDuuJf)M~ceGRjHlV~aqs=%Nv;?!Hox4lRrpIf`asQ;o7iG`@1J^h&f{
zAh}lK#Fr7<p$zKrWYOaAQ;OL0n}tu+`GY+Evyxx7!E@af-<<l%{mgb_SDoVfS@Hq*
zYEoPl&m{KB;<zpcvyl?$(Maqu%&zMFQ}{OAFyJpNKgC+?Gxhw>p57k%;-I+AXQm@z
zF~0oztEen)rPo-KlfK7kD27?xNW=pzQT@)vhM{@L*t(gz^i@H^Vk}6Ln!v|0iihbh
zV%Al8g!b3!%vWvp(xvj@zO`prw<<7cnkDHYb+!Fko&ICrG^iphCm8SAP4DOVyl~?o
zO}wG*t2gPl$L1nVr277}_UXijRL(~S$)0Iz?LBO;xN4;kQj_$!)@bpdm|`9Y_)}|O
z%JO2yy#{z-_~Sw4=x#;GFxXTiBr2RZ5fL)AuaO+D@^)7p?0*=?84c3}I5Sm$0xCkZ
zs>@=^Y*slw;V%Iep976=8aAEMxbJ@uGiGJbBR$f!9)Wl0pIWSd_c$zM*8TXb-yrXD
zvbKtu*BacVA-N?4^CT6bYzUp34cCuP?#-IHywP#q8YXIWo+;*oG+s}HM?H3GO%{+S
zsTByky)e9~)vG4p&8j5>sVvGO6It87gg*AV7_gal62h)RMN-R@SEJ!g=Zuz+)7IV;
zRYNf?LIfn1K!@?FmTQf{8q-~<h$@Dr$V6#N(ZuG>kx=~y?z3^x0}~brs`r>h*=xT)
z)^qnQaqO#SEnJWMIua67Ty<2V6Rdsx`oPGB!=IIUeY1|rhu7$|6OU_BSCm)$=1@-;
zOLBs{46Q=2JE2Ftc1uphuqn(Hk#A)<;*Zf1%goh;rAF0NI&=?UPc^{))8f1}&CMF_
zDxXmFdLyk~PlO+>TSZlRuuj#~O{i{d!c0;vBJtz<nNGgxWenNc+V0~mdlwhwXu=FX
zWY4h0Ms561VR#Y!qH{O+yS`_~s2DS&v+5tlc{ZULfK*_~110^L>_3e)($jyprt?hN
zm1?EDll#Zpx-`Ko>a93QZv2DkQFGPZHcH(}Vva*2)3YPD68&;4qIN03d!#-qFD*Aw
zh1Fb5o>*TDm}ZiS>ASpQrcqX~Lyy?B#NK|0xqF&ldhVk^M*SLtyff9uMGEq8{AB^m
zOp{v4z;V%F>qxlk)PrWnhBnZobH@AhQL@kK8-*QoGBuX?L@GotTuBUcngCm{<xhIS
zapzwAj<ho$`dci9R#+2n#^pbPZO_l<s1J@Ar`)?iTRRc<*IHS1l>F#8;jzKGg2{41
zP&^8}wtuC(diC$~H<%F{-I94@k?5p7oCyNqR-3cOgizny{i2%!O041vs$qYdEor+q
zF-s2jiJq#@vA2cs{fl>#dQ)(v(}cfXu%1*(WmC{cVQY40Ue#ioM^>IKnOGVuDw=N0
zJ<9vhDl_3+>qP@5Vb3i8x#ezSH2Ais-iu`7g;X=`r%$?gw4^V$HWLyo!UUxkLY-{d
zO4vTHI0_xetl|4*RoIT7$Ln4P2BK6<<%k=7`*6r#o@A7c5kUw?X|j;RZo|;#O^F(2
zcPzk8w6P@Vsx%r2ME)8xO#9T_tr|kN+zeP49R(+_t9Fxyx~-9Cot(X-?WwP&n9*~P
zSJq!*e!i2s;tN((>;)|%ewkdjjla;coyiNcY-lL_n)0A>!ZF6sO(oy9)aj?bq4k%F
zN!P^Is97nii_jkc?-9p7s7>?~8DLc|gQEsLJ(Wh`19$7+z1waec%;)UL)waReVksJ
z>6FS!Q_KD3axIt9yZ8=Y5G}_UYU2qRjW~2)rWDgSQZ6oowP`&JUC!BPG_OZL`|>BQ
z_<3Tm)!wixb80<w(`olt*IzoD#<R2q%*w`rx$#aXHB999N!e+jUimaLLwb@fwXLo0
zM;`x4ZAG=t9rZlzd}NzOo0aw5ov2s1se_1(8CfF8Xu+>bB7OOliIt+gfRa#?DcNz7
znYfNc8>mGZBhk`HR70a_RY#(h-b|UDyEE*dr8hnf&WUCFtr<unqYhCFax6sQ@-7Q?
z8Ow2jRSj@bMquPdTyX|g02g>F!c4ts+r$jr%I}xWC>xVYc^HpRf2<<ebsJ>Op8lh(
zYVl(R^}Wk{|1Z)Fl*{*wUR@f8e43^SH*&@2-f9Dy>$D-;;SF%VQq7b{9XAvfpWwXl
zxYwa&|1fPsFdw(RJl>QkzZ!w&MCk2~+X+gT!N|0jb<wQV%}yaiJt;i+@+@+sy&U$|
zDtLtwW5PMgXU})!%!q37Akn%SF540lZs@^L#a8C*)sTzUMgRKZxXj01J&gX*C}xp+
zIJauFyTg8i*vf*5%;RAx=wCZqyNbtjFGZ{<L!n!m{-c-D#dqgR4z-Dn%8YmAw)nyX
z?9Jz{f1dVB-uH=nGUxJ}I?!0+^{-7FLVc4l*22>0bw}lznOSX3mdXsqNO?`tCQsS;
zX03zO_-=<K5<Rt*{mB!WoH@CZ`ts?A$Ul}7SyyAR+~~7OhYp$Kc#o)=o-8BgWhN&=
zZaR-hgRkoD?hzIcC-b_^ux}2|Kih8HIOCl>PT2Z|286CdWCy8p(H!k6uc+y(uiB3~
zh*DM=f0jLvZKrN@xXU}WbbH}(67z!j{%LP4_B(H4(m~{ypsS=F@C9vYQ*Syn3-Ous
zrfxh*y!CPrz?d|;ETNF7kqlUQuc|~#)uQ>>w)+@Em`rm=NMz-UAR;6L;U1R8S4nyM
zV&;7v&TaLB+Z1D%6}z?aC)2=7fFtoR;reZE!)7=tKO$twA7q|hOgW`C(D4JU^~zW3
zhO(Y&pS53|$K^*wNrg$|8pgp5W9*&Vi}3~(PkI~3LwN1p&H0}8<gVy#%v>Hgk^{6L
za{PVL+aV?4z4-ZwF`fC2-wy1*?fkn}l5!zbuCFn3B)2k*aP_LzRr|48n45IW^bqAZ
z-X!Hltwo7e*8|p6AaR|&^a%|Gk4bmJyfB@M`t86m3TGCb-DYn${!X8{2bUY|QVkXs
zkk+ESM2$o8jxeP<oyp%~_3Rc5ql~%-IY^&_tY>$1CTFDE%W8FzM{m79Uya0O%k8eW
zwEjl3sHHb%w{uvyB(DzjCr^AnjO1E*>T8j62N=l;gVlMb5)=*Eja52s_&q(5;T@?J
z{L8HLiYmc}C$J+@UO%9faNlvxdN=K$(n?qhtJLgrXB!D7WT24&)JnebX0j{_>*hsJ
zA3^N%+LAvoqt!E&SN+BfqlQ`gy9b_y>pqNla&vv{{A43UXd@TpI>X!<nu2JAPjvGM
z%wur#VLjYsqSpCUYo5E8Z>wd|_A&K~eOqLMFWV-03he6+tr3jM@9CWXthu{MtX@87
z-xKmy2q3#W*&}yS*1XHvLj&uZI}aq@C$Jw<E7m0|Z{Lx)bfSWeqo_Vs1FqUMZDV$&
z1Yu(0JAU_FfLqqpugq@kUnwK4nbaG|II9GhyWS?WERCu$G#*X3Zr3+H7hU~E|AdZ~
z#*(r!B1L{+l7pMkrZqWVWw+s8qSs2GM8Fyl8Saz9PwdnW<73;80UfJb4nHw(OtjgP
z@15E&6l?YbVucC|(H!N-?1pUUndAGGqiThkbfz=n5-mEcx{F`%q<MLIrr^_08+Bhm
zXUiW{Fyu!QPvB+3I_%VB0)~Z;oUU^m7#{4+Y`eB0xaZ7M271A&`cLHg<LI<-+fh>y
z5^*2cY*cO@lga%*Rk+uc!S}<wc=t4FzRLu9!HQSq2oh|$m?1K866N_Rn!E&qe~_6S
zt8L%dzAHJ&OTN`1av1)nwCgd(_k`brKo60-p~Vow%=w;|r(txv`n@2V<LIZ9V&r=W
zXfYP5hB#3c{q65p3O;Ga>A$W_Pf;cS+oK?PhoZsiMC=he0RzXQw%e$U{Av|x6jk%s
zFw(iZmznvGwR#J4i(EUG-Gc7Hd*igcPT7elJQe!ac-X4@c9dTIT4(vGPc}8UH|m#7
z2msmMQp83nCIHpZayDTvo7#cxC~9)!*%-3T9wCr{=JMUB0rG(Qz=y&nF~t3tk-@&e
zizmO4ddJUFpWRz@-1~&xqfRn*Z$gq^fX^<g3SP08Okfn0P#Cy`7<i`as;lr_*qt}O
zECu*5_MpBRzBVzN^MUg+I;}ZuiU*UTbU*VARU}_=%~U@p_PnNuwP);#WK9@*Gvush
z76Z#8$_2Sd)r)LOF*Bo@ieq3vAnh-xFPB1NRIATKH{&ZQD$OCs*6`m}vebo{<>!j-
zyT1><vguz))9*IgSPi(b(lPz^$)bV?rFm`cS}l%K^~_`Ep8u!(*&DUB^w-N)G^a`5
zp8{S7xaO!aR>mM!R`>PC?@fqJwwsazzWNF(XVX%+nxAsrF-Z57w%h_=WNd`d9UJdI
z&lQyw#2SD3sg90{SrG3e9iu6~mahHli%`b!;~KoU7J-69joLYRaYd%1M8j@VyYbMp
z-zNH-bqagOTfa;*9xw5(b9y?4!#qPzESSYHb!>ffm%DHmFV=_N&Yc36rdUkJcN%Ns
zexls+Zoe2&acS?ztplN;x8nvuZDV|znoZ&~`#MP;A67Qpz1%c>)#GvE_-t<p-t6``
zBst&~kk&Sw-d9?MMlm<N8gt!C*48CFSbyV3ZHc;>Udi?Ed)%KeT}7+W3b%I!@9XJ&
zOvb#5h-^ODjA3G!$zfG2O<g7vXy*V!BCHZy<)tgceN(77Owx&^3SXP4e%P?rBbG_V
z5rOGtna<lU+KxH5_Pz`GE^{6L;JO@<`Fz%)qs!$w)@j6_9qQcJ@;B+`DL<`e!oM~1
zdoJYI7!7Cjr_stPL^-U0Lv;|V_=8$uK^{^<c;AS$$QQzs*XRo=18-zBl=F<2jqC)H
zSA;(<KTut*`(DmM?QgB^=G&qYdSEA=UgGG4p`m!E_z^ql$b;7hA)|(-ui#fH_pTlt
zYxPpm&Ke%TJg@J2I`K~gGe{aQ;^e>jX~>n&uAk)G^+W~RSz5$MK7r!y*M6w&;G)u{
zk49)Y>`etfEwBHzt!o<ej6dn*Q-1WkzOj|2D1AAbY8<qHI`=p4%1Gz6(XE(Ixs}T;
z5WhM_cu%>iMk{rhm%@sQ1e@WxM3FLTCZ0+d1$t~KQ&KbK|C@%+|D8(v|Ll=U3i?p*
z#0UEhQRp|>|GQWHQqq67)#@_FG^s<er=jKTwD;)g{)@~_&y07%SKjs1CbI4^^QH)g
z!m-;GI;&1Yf}`BA=)=T+@?!Q<YghLTr8@}Xp-kuYkdWw0r>ru_oYymr$N^o}UNwVP
zOxip-xYXa-HN>DsY=t*e)8pxnX)1L>r<Yi&|1Jb(0+trfr5Hz|(P-B%&!L^Zi;gZB
zrKEky=j;FQ$Iqm5UrSHk>CAjk-jmshbqswr7=ms3>EZ7cKiOV6R#Q{6PWzj}vp7Sc
z#Ns>hiV6~z?10^GBShAga=6M8B*+Jra6p>@CZOWQF)^5+VVq3h=86+3d2xv8bQvp3
z2vc%(3WuAzIssI-ldUZRIAhY^d6-1`+1D@}8|ufBA<v$b>HgjLX81Wc9swH4U!5&5
z@~K>XoZ?h%+`=tOUaS?Dcuj}nyfQ}?duI<e6*YQ{jCN%$lPwd3fcb!c4LD}xiVfKm
zfg2{P5S9=`5zT#tA!7kU#zh7d%g{|vd+wX{)Q2r7GOdL!4i*3@3h{k1GFt3249NVE
zUQ{NN3CzEG34;OJ{05qjmS4-#IBULcn|AU+IgEUUF`1=iqENxm>b>*y3Bx1ksdh8q
z(?n<<%fF!Oa3f=)Enx#D<*9B=cjTrl#@6|^p;1ONf{1GNqLiyoW&7|V_zXdT*aV4m
z61fcx4L)NO)2Sg`I1mw-sqpg&2ZQ0J!n3yjKEh9S+4HhWMio?&#pmzcFl~EN8cuB8
zZaH#sp3R@jB3jOqt8Y)tO_blBD2M3U;an+{VnN83V%O0q1aaoWjHx$pb$W&wl>0gO
zFeM?}bgBFEbDRXefakFPVU0OzieundH-(_j|MFKV7)%_6LgAR0n9N|3rc9*-IL+aj
z(%3^nHAfpHp_%T$kK8vR{UIC%<r$FxNA?B7@0e9g{S0QtoPQM?<X^?qDP}RPx?OGN
z;@^}8zof+Y*q;qQ8{T_1W_~c+x@Dew4AYao+!{>DQ@_#l!a9AU`QX&)Kf~=x?<^HM
zn{FF%Zx=ML;T_zi>9^fepXTRjX@X;Ywocre;i6fBH*siAOh)ZAH!>2Bl=6st40V~E
z(fE0!03Uu;|MdYlA*3&l4<ViN7zm_^G6HFSy54li1q=oU7Z;aqfoF6EJ7VmMP8ob?
z_+=zZBlJxD4C(UOX74-|A5X(-%E_m_@%lj@HXAri2S!f_e93rf#ZA|za!1w80N=*E
ztb~m#b~e)3b-(%u0jkEAgElktZH$q*iHUL`K_G~|$ONWS%%qUgivrL}(+s9Q<MaIJ
zeEoDLEnfummPVTZgPT(qaC9S-?T;mYWfzsgj<T{G))nH`^j5;(h$nx39v+wXmvj2N
zpKK||eG-~y$wvC}=8y5m%+o0GP3DrZrV3b8;HLaNCFkA=lLX>)jT6f(yrDiKR-^vj
zc^!>IWv>qZLo%<xs|~ZuWO8Ff133DkG7=UwY;x(t#nC1*ax`Y5pwb(NBAy$Oc^0Qn
z=Y#ZsSu)0v@F>wq0J?xphitpHkG>mdXkPkyfR*vVofC7+;#4vr-%jVBCCUp_bk^mw
zPpvc@k3StYy1Y$<Hc=iu3Yv_W0QveJfoD#it;sYP@@|^x8SH9{RM*kiRUB3g^=m|`
z643NSVT3G7<n*3Idi~yqL)_)deL0_ZJXE4L4+fC};xJP&%+*u`_eCfwD$W+<BA;5q
zuLqo($#LA-ph~`KOHy~J(N|;F|Nb6!V~q|^^B5}L*l~Zlo5@ezaihQKx3j;yv{@pr
z_Md$|9{XVB<%@xG(ZcJc4Fzv~Rn1v23idJa?^r{0c`<2)2Xos#X&w2)hgz&9lAJ$N
z&%o{xrKKSP{82U8fKIjrop0Ff_EHf%<m7U+XXz15v!=lDp_{v}aKtf$hKT%9yCSfw
zX*)vAsmjDF#UP(^p|C8p1RIz}C5*rGIz?Z822r@~FxXgyjh~_t0qPc+Iem;a86N}U
z>vMdj`g?iA-D<WhTFMLon7xcudOV|=bYSe+m`h|?sMO9j`g~4B@rhElZZT7nX#{f<
z0ge88!PZW;TSJBpHo9Z${v$RMW{k<>S0jHYy;wLN7h{D8R(g7A!Sfr9r0<vCQFP|5
zcVbE0yoqOPwZ=~+<M%f>=T=&ylN|XewE{C}^{@9t!|R6jz9N%wV>m0{mjkS&2=){Y
z3gn@s$x$va39!e^s3zq+>M3IUd%IflYfeIODXsa^mfWWjNQ1Cxg>s==Gy;OKc8Y`}
z6eU3svjN3>NT0<kE--lRxaGWy6)G8ZWR55H@HFpa|NP9~7TGfRyCW7CGA`-Mzpt$d
z9tWK1s>_2AW#y}$68C66o<4)2QF8AWz*KdCSIOcum*C5w<!{5Iy2F(b)<{K-d+?7*
zNvudbsQWP)=>&JLlhEZnYA_WO!h^dFb-gDQ%WdY<{^;H5h1ct!u&GDeJl_J`4*O=}
z=RcGvK++<uL(IzG2Zeq&mXHQ-9`51K)M9>Dd&JI&kD)T&%J?(JuQ;`%Eu18C9&D42
z19Ze?<TXN9x1hrw=-Y&!a8*W*MwT;+nQR?W(UB{ZN`If-7JiBfWQI~X5isY`7)fV?
zuxnf*l^>i=$&LeT;COvoPJWkullgjQy$XWV3=kucQ_42c3o48z;SdDeQW(ByZ_Aj<
zy->KI-4LyCWgF{xLG*}!sP6@M(NVWeZM+eq4I7>;jn-P@tNPCUbsBv6G`l-?%V<)7
zq;7a|(C$oT2vZQVO2sgBi<kPkdj`2G4O=xq-Wl{3ZZ1Z)7Vcv!goncha15ZuLsdr0
zl4f;w9OaE@J|VKQC5pQaK6c_*cW{5^-G$+vtpo>)2u=0c)~dRJn)}Lc0@Q~ZR-Vm`
zkMe~?wS<tHMN=$d+i0B!m&_)@0qds57-LDVh?8aST95naiU#}}CNQwIC^R|N$W>LB
zp`W48EKR@_ml_bPTsqmT;$rhSu3jabxm91|!Dd=y@b#D?rD5QPv@x-cJZW;a%lc!H
zLgzMA3pNCH3ezVmEZ#^{*QT@D{I%BX-MRB7SQSL5HBd8%KxAp_-@fw4IClL!G_P)r
z|BaN}WQcbk4GVNtsgsRGpB=Vb1XvC+Tp;>i`zEF(ot%dRqvEm?Sp)x<jS~>I{(9V@
zDH7ZzazZnz7#J(3Ud3LbFBI~he-(Q_l65|-aa4>+I73V<p0IGSU6`1!?!H9Av(l7r
z-x?^1c6u`J<&DTz`v-_~=H2Q6EdswYS?)WnHp8`pp!9mmQKHCQz3^2592~(?S_<?f
zBYiS1W-LwHij@$`kxIQv^S}J^dUr%a1!ORtPR`59!gNh%U=sX!9GHn4Vrc87b>x-5
z-yHNm^U;EA7~6Re^$Y)&%C5vFfO?t6t(zMo<@^KkOOJ?yBEF!|IBNHtpWpOQbemB{
zdnQ&r=Pppx|1{-^m{Y5`lEPZK$Amb^%~t&EciYQX&R^>Y5GRzO6GaI#9jG<pPSpMY
z@A2ca1~WN~J9CTs?RBV~$BAZ#ets@~_O<l(()Nr5e%$IxWog2l!+8(qGpSs|8mp5%
zU}$h{kq$TO7Sh!W(zg<BVC1DohdC8ID?&T!#T7Ef=6}C`)6h0*_29mrc(vZQjg426
z#f?u_s<rtchR2z`47uCzKq!)uup?nCjM|D-mNa>E3X8%pe;jIHZO(pd?G)<414DWs
zfb5^g0O?ZaL`u6(wnMf+aIE78d-|i^e=+pbuBiBz4=E@;b-kQZ&Nnm1X8OqI@MbDq
z!L=70!XLQ3re$C&w7>U*sKM=rB37)*BSTa+re1MX1oWwn2%}lf)7AY4G~%uVyUbXW
z#>J~(0U6Xij6NGwRk`Qyua#=ym2xD=O;nUhDnOnz9d&nx&~{y&UYfv7Vlo)Tluhti
zdz%mOA7->juSd_<b%8Szm45}@9LhVYUQt_4()4VS>g9j*>KpZpx%q6+P<`WqGYt-%
zV*T;^$s|cLuNE(sFXE{^)in!K92LVroaBLTd5S2VEdXxf=|M5u6=Q^oXyL+LZBgjx
za9~R*&LNT0h^)Bc$q&t|EH)p!F0F?lhX8j^1dfWO^0Dcw#c$H*OEop>G=*r3AssVp
z;a{iST`}XB*N}i*|0Yx(^?@+QdfaYt>d(9jCus>zY2J5?6mNvTFU&tB^JK}#xWv_0
z^~}yuGwAEldi^#)T)}te04*~_3n-W%dY0Lv0jD8-*<@%Hb6p4#!30sm2qnbH?p(&o
zRIKKP?-3PbALAmKa=PU`8+OtJY>SGJAz@&kGo_2Qt;nR;#5gi%-h4*e?~Sv5j_W)G
zo%pDAy?6B&1nRwG*KY|f;VAno^}48szlG1RYCJD|T~TJsi{59{vGDPJ@Cgpw9yPQZ
z<+<-$P0K4QSgjpK5XEDcH)Mj<Wh#?;aczZ;WBO}RBIoA==Dc@3TK2+gqv5`F8X^I7
zpSfL1juN^cee8j0^xjMRqkT*U9tibm>EC(tb)S5Oz$oRkcYzNQRlV+fEmW|ue9>-O
zq<*I<k8r`6>izSnA8a3Ko-3uba#ONMc1LpfTu{M%B7d#?1Y8WUKI@>*a}!7K!CMiv
zo(xvOkO=K+3y=N`1m2gGexd+nakMGRl;g0e41eNmS8rQ{fAL%~b;Fxkot{0aXmV1+
zu`v5L^68o!$E>mO`SDSOc@n}NeVoy7`RSE~>Tfl^J`&I!IQwY3K?5qe$v<FQw&#&=
z=TwbH+t{N91y#V}Ay${A7<RAA?2SGV(aRhn8A{)9h7WStu+La+79BZC%(z&Y;jf#+
z=q8_@SZU0{OxRE@`2-Ixj1{$ugdl~)X@qSw@W9$O(rxHvd_i^OQ$8}+(}C9)_0pw|
zsH4jn=I}bfgI2m^EMvZ@8o`+8(7PtR!zd=WST<gVH%puz{+=5fr#_p2^(ee#IxORx
zt{;nIK29=FxaZ#4RS%CA#k4f$<P;HytC)+?)69!>pu5m2+aG|HDc%zIqTC49lp67y
z1I`dUHtY@{-tp-IRu2lzXEOaYH9?rT^z?VB^GH|Zb@VBEc0)Ur@fF%ujbG);Vm&b#
zyS-oh@JaJfh&(F4B099`_o)2x#i60z>$DPDb&%i4c;w~A?FiN`#e%|9jc8~v3Rl{l
z?dw1fit%%rmTVN#$JXB3_KLv0W|9k|snkw!h?XH7T!x{PZvBCtp{dINVS>}sUk7@I
zY~=d5Q1D?D)Np+8oeNWu>ow1kFrj}pl)Bcgn1CEuF~GjQW*C5T<nZn@)s7}t{9kv?
z(&oQX>rpwrEsOJrYp$x#B9U4+nMt9E(0KsIsu(P};)a>J=?rqrG9$y^9QaQ^u(5!r
zVS!9gu_7i;Gx5BLB^5}t7>NLuSM+B*8w|iX3T3<C;LESoS}v{WU+Au_ttD%GlbrP3
zP}lu0kJ-6u{A3=k^`u#|JEN0OuVJF_r0~o&b;3Wc-G0>E5D{T=*?0-#*H!>P{tah$
zcerG8A#ee(7~^ivcEm8*fbKw!^by3!k*Em%XY^l9N#1dVF)so^AxHpZ>&1nEVJgdX
zP3lEDu(ALcp8omF0A4H23pqi{?5vRjP?TI-Xpx7xk>J+<=$W;bZT0P^=9N!x=3Q8=
z;D1b$-Im3_aj!fV?%*|EadBvacIC&-X*{%k>wWuXZ*M(OMpH4^Kx0HSC;?~`$-#-v
zjE@p!eBFPCNta&CpaM^JE;#|36TxKFISBE<H2G_F->NWK%xVgg&!izBNVk_J5T~J`
zq3uDfXPsaQ#uVd~qhSSq)_Dy2fO4nsVWA|bkmhhp3V)(|o$y5W+7FK$ZltHDRz?8w
z3w=QhA9#lWz`hR;iJbn7@_`24029LI1GeByCU7zUc#Tdq?FK6ePr^;R5Ws1R_%{X*
z0q$<4z4Q<90Vc2H%D0C_A)kuWA2fQlB-|rWhdZz2)=Wg(XnC?$v+R<!t_ZLAZ{(b`
z{%1ylR&v;By6#GbR^(bZaH%eYQcvtjmvb%iq-^2T#HXjDyf?dcl$X}qJI8zhC?3D-
zvpooE9lWg;V<m@vjFoyDUw(IL-o<nde%y2ouMM3iUa(91p~18ogmpc0=3&!Cw%{|w
zUUc7(GO6-udNqAf^<puAZ2=waXT6gig9(t$n*RO)++eV>t<cT(GoI}Qg8`|mPE%Je
zmb861DS?uf$92I=*T>+ONtd<cH_3X?Z!+pZriEu6Yi?v8Aa6t$O-27Bywdx&5233f
z<hI+R%Yv}D`yCO6==yt!8wY#F687+c85h(TAzK5`p^=w>*ea<4=JgrMq7;EmzR=Cd
zsgqr4H!25YTv$tj@~9gU?)KN55Dh>PV1NO%0No$-`OpjyXu?y!A5r43oz4x_U)StQ
zz4G;>+#waJ#3<~ldNkLbb~;bR@wg%7<jgfmd_^tyho=|&<e^EKnNH+0aYdW{i~cj5
z+^oEu7{EdGZxuWPkc%06Gs(?$UOc#&A)1Q>pzCKQf~kp4jRIQC1A(IzZXl4#qi8TN
zxWOT=Xo#&40M`I90WcEQE?}GW@8xI5)JSjTdQo-o)%CT7#>?Mi#hl2zR(8ZEI<Lkd
z?J~ssvewNzT@^ikL0y)TF~@6<=vw+T<CM+E>-Z0zf|t=tcNV>P0s4yi;rKJBr&NrD
zrs&3$?)3oxR7a4g0KTVIu_-Hnc){<Rw(~a|BgoB7ZI&WwG%RLA`l1`)1%xxeZ_r$k
z5Zm_*CXPhfJk?InV{Ym%j-%cGY^g3CC?Bceq-nfp4~?d%6&HQ)9cK4%ATxbjI9z)q
z&}S?u-McuRM~y)O*=J*&Z_U&e&^<$)9kHtV8takWgJ)Fq>2g-?&=w$}WLlK!%<jn=
zafkf%SuSyxmA@m#$(5$LXp=X93{Mg6x=A9DJPXurR+#VRds7uI3nf-@M-dNev2R<6
z%>(LWLTGJx@U>cuTXb#FS*%0QxGvnh=?v99F!@u=`xz=jdtJdLb|cIFTTpSO)=?`z
zQNxz85FRznAb*%T6Kfy?=om&z#xfHyTxDt%lX9|0@+#)18M|3#F7*i{3SIu#Ix+Sl
zg~R}OCGO`1W}*nJYm@cT_XL@;hNCAaPm96q(`D@1Ou5GTotkLyi@bsJNm=>TrfBZ!
zxWgTi!EUWF^WmV27JWX%HHw;2QMGcWFj_(Xs+r@<-rdr2E0=+eR7${vgy+BR+26y%
z?bdl?<b!6G`q=`cn8*%cSCz4M$!DO<#$3z<jJls7_vON*$Q_ZgPz38^0=7d8YW0#%
z23;>3r0L$RQn2#CIB<Bxr1G?eRrt9R(C%853ps9{m!=<#56T-Y+uN-fuMYIRU9^u^
z7(pwb4Y%C_t+E{BHx+_UwJP{UZ=g{}0F#f%qqoJ;PV0^ZUTh;TW@}Ons~OLfMF}HF
zOcE6!uS!9hevm9<r*58Vebf5c#PP0f)zCZh5}p$hc-54pK&-kxyH&?gNb4;T$M9zd
z^+zi8dRH!cRalSTuj$kK!$-`#e$J?3KGnagxM$l&JKC9+4BdBFzQ`+HZ30}zM<dFu
z?8bK$bf6d6;}|T3XMV;N_vU28bfEt<C+1>ZddBkF3M+k;G<V+44tfRMfh_PJfwTSV
zkhwVW#D4WV7JU(d7)sR14&a`+nY~i_*k4hEC#`%^%=F})>%-UghCAqauBE?|gimHr
zyuUO}wU6%c4DV6s#Nx39x<?peL6fUALG)Vj`0KUAvb$@iqhyRjb1H*VjXOO)OV3!Q
z??NxX3-cn6yGAUUPG=+lir^IzfT2r|C7U^S@yzP)UaDsgo7e)X%X-YhsEMNKaEx4t
zavHK7w%#&1X>#s~yusR0dFW0fx+uW>6VgxPHm3L2+_8z7>;C>~@xMD^3?E|+AiBqH
zB<HvPu07s#Fr4>$nu^1RrjjtPR}n@*%U<mOLxOHJe;*N5a^2XukC%gtLNmJuCu2jW
z*~2J9ZypqY|5SjB5sGY`A{PApmt0leZPmZ46}tH_GRe@*Q8*xyL@CJ_O#dndldhNk
zCWB!y7qnL2W+nUxbs`l%P)PI7(|+b!G0?L`^|H9^pGS%+?@M7Tp?rTSaqs~>_M+*W
zDtMG&Yj>sAJ8C02MZ{8Dniq3eDn@WAG@9UdYn<RIFdbHm5w#AH$Gv%`NK<%}Ez)#I
z_)_swxw^}rolNG|rEoBwZHZ^8t956vDoZg`=4W^tIZL8sC$_d0ecu@grGtz8J$gKb
zzC9On)WT2!xpqqyfARW=Epg(fr}p6s(BJvZH;>HS`n_i8kxz?h$@97p<r&@|YYFY-
zND%d6U){~|#wf|AkXX%>sWLm;Fs!CgBoeM929hCe!(heBES(c4pt_l@PNB>Y3N313
zl9O_gk_`$WH;I&nx@-Z^4_^9M|M<kWFGg_B)=iYUME<S=-5w)WuGb<5F0_U_U`5w6
zA~L}J55xPnJuio~iFOi1fNteIXm0vsj@n*FEk2xN7OVnEMp@m?#CEJ6M$PvG9n3Jo
z5&(-#(a!!$R9IbUMqFm-S%WJbPhQW>C5)7L?g0Ben_zpCiyxC-$}VN{6L>4DtZa*d
z%b`b)467;(^&ahN=+z(F|IrxKt7lyF%_(@rO(%%g_5AR>A14E69YMbV1wzx>>LH7k
zc8k-ADaOJ-*7j$}&TdkcUdegnh%f*Y+8vGy>p6D-P9VSrbBRP4`ea0VsV0*b^vf?x
za79WQxd6IPpI(@s>M(`*0D{e!41NMXfK{0tx49^N-}9{1>KMi?>8fH+{f%w(b)iR3
z6>fQ|aQlX-={kcs*KM2TeJ%gN#HxGKkHflG>;lbqlkfS8*BWEdfcXz@!$#S=SHQw~
zQVRxE)|h&1GGxiO4awAFWiS~e29=pkAr(7AH8jp5ok2!WL&y*kZ9ZY>k+GzePNj=X
zyAdZPp!j0ahqrGV;FiMm0fb`1M@5AJMt9D1E+eqGTYI5uE1|n>dgw;eecg$G-I12f
z+mf+ak2&$^ilJT7zVoU~O@UUtud*O+@<|jYC$FhF*H^7f5NDPdvuqyzgyyiS5wMTd
zXaJhHXJ9K!b<;C0M;%_h%)}2p?;K!L6Ljlz)VlYpD=DQ^oB)PU+BUHPREYp(WgMLC
zAJ=^D8_AHp+8^C4ze3)ocs^e=e%JT$nt^=;zFuDocBh{RR<Szza31q+D!UWs`Ghf6
zmWQ0!ElJdz!SGnPgb}pBbl!VF6bVRO$Z<QU2)W9|4obqVBvcr)yOu9uRB_?4M3hoD
zQe-mt8S~ppx_au#MCKB|c7BTMPh$I1F+3nrLddWwu-JP|l!O4h4qth9?y@kHAMdr8
zkPOth|K1tyFnFPBnOK?c?#3}W&G7tZ-jBstXFg*}$FfPcn|j-hbLMd0cT+~yFnt{&
zC=bV;j*JipNfm1eg=JZ0Y#10lRZq&piEG&x>H54;2vSL?Xrk16)<TU~QWQC{2>9>I
zRx)fbc7|lJr9Z~MR8Y9e0zi_##T+nL4bHrpkK54o$;kBmtw_C^Aoe4&e`71yL})x@
z(>67gBTvq2iwkz0TM9`;eGVpAVkvWR7eQ*6jhrmUTcSoBogBS{i|3vB8|lqG2VG`q
z=`Biyu$DurE_EN9J>GS_wCBnd!<I&}{|Kat!Zh?~gwetwu}pse=sP*Rj4!n)!%er2
z)f)G&m+||fX20wsENbWXH@9|ft>JZbvj?<#9k<_zZ{)x8$QFiPcDAHr3tx>{Uc7c2
z%`RzupPfSra~U(kERMZ3*dM69gOzMl-zVXSVx+V&OkG`1L<KDltnnhSbP#`)$p<iM
z(Dv-u&JeA@w_o@y!Wh|wjQIqzkpfM0m4wQlnGhEsHFpvMBu8yyQiYE8j#(41ieBvM
zvhEYbitibP?b_%-+xCNovWw4tZ0Pb>*eNP~xMezK*(-|{ZHISP?GvuMjzdBRB@b97
zl<GGNaK+q`svy6ip$@2g6f-^3-^YvCo#J9-Dpk7$fB^e5-p4YuRvP_#Z1|y*jks>A
z#WDx@aqXzcO@xB;HL2L8YHDSAHjYBs9I8D^&>SNloQYNC->EEe-=5Z~^pf83kQHS1
z-uy>sNs(nxz*wL;{K<FPt*fF}*KXU7-aHxGVl|(zQrvvkRQ8%5_3<nYLp~j6b>F?K
zyB8qT8D#Ul%kn9XlL&Zc2C~D^NE7hT;DX@k3TgwHut%w!5%_GQr7dw{Bb3S6NF<OX
zVOFN6FJTgG?Rg0k$cfviu4)GRJ_70cLhGM3H6)F1k}eJxf#y5Yl$!(W?tC-(fI?7s
z+Tr!+Lf8@pgR?+wN4{%_Y>Xfx6LpL)A5PcQkU95o@{eQ@Dn{NvzsKUCkuJ@C1cCdy
zlrCaL0SWA40DMOwV0_&iy%0Je<z(38I@#LS`+I~;JoI*^V6hcmgrML6-M^bvpr@hW
z%TX&0C5W-0p(0Ri|EY?b@%ZS!PndIkqjBNg)ea_BIqE5!H3s<8af!khe}DfeZVs80
z)v`-IJvq4-LXHXbsK;L>yN7;8&)yv3=M1ht6`SFIUSatB4FNiqxRMY(Sfs}k6dC<8
z5Ac2>YXItta5yFaGy%Z>IdfeG{b$;7?Wh|Z(N@6EzaBC9;jiZpI$i14Ny<;_)eJmp
zw2#?moy)&acsl4iEaPFJ)Zx}6&1U1-j)IguXZRY=!orN`*Voq(ArNoQw)sp85_Y9|
z$U!bZFfccK*i3*}WPC=n=u{<FVg^CV0I+>A9muikaslD<XN=DPn2_T13~c)6BiqW(
zGMDaUVs!mo1y-ZDzd-v<;LSX(|LFG3)x184$hmSb;#S@38{Mtjiw<}2Cg0AlP|F-b
z$FyFXoiOxn3{DC=3Jo2FvD|BkXdKzS)|+vW3*dVbf8c4Uz`y|3=x0Kq^c=y<C<l#1
zx#uyEkk1S17of%!X)+{o8?FXO5Ma#=y{7~=O9_QiEaR0Be!<XGh;>jWH2w7Gf%Mvq
zN55at{i#A%)8Y2#?7N}b8gucpDgX#E{ND-zpU-RobCvR6a2{Ap6lQ#k>&br~XKn36
z(sM4R@7?V&T#m_TnmZS4zr*8td-)K0c7U#q*RxvgYhMWQ!u{i>m(ru#?s-}_bI+VP
zbF{q<K=W*BRb6tJ-ZC2~D#>_%g^T>09tJ?{NL06*ND+XEpMheqry3|rQBlo~0whEg
zPoJ&_gBQVId{ghewnq#0IYw4n+Ruin8YnT=#bavMM;Gklzb7o659+F|^o-`{a=pU!
zuBYFA^pxt<=jG0p)yK%zfWpBx%p#DD1yV*E5&^M_EgOuzSPYEOT>!csLc&}Dka}r-
zl|fx$QmZpo*dS^(QxnUiWK$SbK<cKrm;|(!vT^|%M4o`{M6~;OzBQ!*{Z;R4lO8fU
zpMyijUJquj`ex}F6yhH4#Diz9ox8JFnC1l_evbA{Eq#{Oz0Dz(YJCK^d85ClpA*qz
zXLHzJGgF_^y$X^^&1!vynKCJgAVLC_7H~Hd3;}mF9qOzt2f)4<KA7N6gGB+!YLwk^
zL&zxlbE+SLzyWhCCNZi$Pmc8F!Ozd?ekB|ULf4!wj4lM1rQSi`IYW6Ed_7-6$vYKz
z^F{;^tutPd<TR<bb92RoE?p4TDnM0V=UY$@Mg{iKy7m8H@6E&6&ij91GjmTn({)tQ
zs%1(`OHun$;_j5{BF!kF2&UHBYl$tEndxFH8C4~M%+Ll&Mb#1#OIu6S($o@5A`~Tv
z*g}E`&!=;r`#ke|{yOJ5&z$RA&vo|?SFhH0`+UEj_xgIjUM_>HI$@O|2->^$jw%GA
zCZSCPw~g%K!uz3JQ9++4$3Uyxvi17Q`d~Iwqbf30%LTL^5#+F3638))3E+pSI==LC
z(L@!KGU@ps#i^Ki_4@52OzicD8GK*k+SOmYo$h;{KX>wK;3ir5R7Xwvl}939W|C9r
zp_8s<w#VvLg=4}2osbFNCC#CNq-@X>zJ)<0IlgP_0FyCr;)bwFqb8DV%Oq&o!u1qG
zwNnQ(pMAy;n)=}Ud6Xi8mh;<oa%5_8CqbiV01mZX$j!~>hQRC+UDXS>NR-8*aIb-R
zVzkZ+B3?iHj{zg~1MRr<ldnR*x%(Uc`c=(uoMf+_^8WA>hvSEd3}5T+ZoF_ouB<}I
zXs@zB)ire{r6`~((}NmS4%c^tOTnf4^`LM&5a^4vMrsn+ks(ZSY){7|)$tB~nle=w
zO-h{v6O_q?s;VX#+vWR$RQno&-)jzDz~?1i3jc;}>@ZLlY~tuE346KVn{jT#;Kj9!
zkjpW%i|aR@jouz|b~hZo32PWIs^v2(wf(k8i=$gl_gLB<Kj!?fndsx;o)&ecF<9Hv
z%X_pc@(dNc6Vq9c6&`hjN!v*&oi5P0POFD5#2bCtHg7UZ^*t=d=!^xxNzh%U_?P7Q
z_!X*};C<ApyP^Ft=nvD<75!e-Pc{#lW$V@(7-hdKCwggJD9u!Wy(#QYCY2?zf&}(c
z*9;4S4t@X5?`jUT#6R|FdQOPNl>xo#29Zt})`=47cKMKNDb>n>)Ygz6@U+4fe9Eqr
z$-GI8B0?!K&p#JO*fD>EMLNB!3Q%U)Nx_MNS!!?@C?R18Y|x?g0i%T`DU%+m7g|Gu
zfAovQOxB(ux2<u%Zs%eB23@LuZw=i-$&AFC?`2vt=Qmbw_IGUMxLkjhBsKg>cR_(v
z(hwC{ljd-!l0sH2e*lj_5D(jdcEq+b$)%8pWT3ONBYWU19J8pbn1ETbX>AIS^TDQt
zX?x6V7!P@<=2AVS6ni>Ssc=bGDu9LPFg**^I^BdP+g9=^U%<Vic``5u%)w4*@FiP!
z<QKY_>&gCCFTFymnq2WaM0q%-y48A`b|hO;(=r~6JNs`b+RV&S@WsV5P{P6ux156P
zl21L@EGV3b-N*@KW8n-dDc7c^RD7zBhfnSlPDS>F);OV)24rLj8tXAKL1Sca57H5?
zctUKEs)zbiZc1)Sh8s@4KghJD<brSVM`H&~uWhS0u`Tj#rL!FKuZMDfTQKq-{Ab$9
z4w8kO-g|48hn-^?EvXmF@dLPKI2%iJqZJ<pPVSylD>*hF8%-fqmChRrR__!|6SR9h
z^;Fj>A^raATu=(gCFN<27OAHO!=>8LC4L*oo>8*{W-c{$hV%94G>^7$%N_%$lq=yZ
z!4KF>Jf=|94<9{qa$(iR;z)N<<zDT!l~KY`Cxsyy#j97HVr2A5<$=QMl021bei#M2
z41*!V`}K^v>IjOut<t+^T*}xl0B4u<VxFR9#dRwZi`9PY>aTFE8UiYG)1wy#1L^}v
zsR8RqxIVHUiU^m~vdOnWbov!*%POg-`g8G*u!c+EKWvb8P-Z|fF(r3CnUj*sp}uXe
z7Jb>Gn5}l`t<Hg)mNUP#4z0ZK&r8`1tgZ^J^}gV^PyLfVvwQF%`{suSn6gvX^yIf{
zIinOY)v$)$LJDxG<dSC08(kde1Leacibo!|>avg3JQ%pwIm&=bg557Er6+%;{q^gX
zJVNW+);2<2>gPPaCxxUOlS~M*wP8q_gZN(=)f|W2`?kum<+5ke<fbRGGaf$su5t_e
z>5YBSoo58OAmwXk2*25!hm}3PNU_ns5s^|I5xth)R%f_9PP{mTOh59Xlzao~e7OXR
ze{bWIH^VUE91>$1W$@GQ$g;6TDfR9!1$el_4H-o*+vB;pBeuc**w8Ew#zs-V6al39
z@P2~`nteab%-U>apQo{*)PpAj6T=CbDcPB{3rE^RxAKk@mKG$2$@dXcjy{+7$ojq*
zD~6O*pop{Uq=*v7QB>018)jVQVpaKjwem+!jCC8{wNM~!YhCdurn?z6(6$5*7~E>Y
zIn9@^hxfuH)9*b<8d@{6LBO2)xsY_UKkdcr{2X`P-j31F06*vob~O2})prrQ5?^(*
zQ+ICPpEPk<b_p$odDXEExG$M8S6q6m!Um5%rv)+o*oQojcckW|y6y32kBEIykWU}Z
zGzMH8HJwQdQ6*?}ke|(uYz4`3$x=}1v0{+Z8u^pK%dvvD+1bE#<SjdO@16&RlB)BS
zx`ZN3s+yWy&t{SOq|aGwHC(^7+p^mlW@fFeaiS$u-q0*L`k_XKN9O6Ko8E2hv)nUN
zoxeSO_pIZ$HpH#+hAro7E^4mV@~``yIzst}r7<!Hg&107Hj}ufwZLQ6S9u7^OK1Fz
z0-1UE`WpWkZ!UvHHs~JGx7D#>=sC*M)mL6FFV7Jk*&yPaH3;g2`~aC3ws1r!1ITRR
zRY}S!n3O`5jJU!*SMki4+#VmA*6WuIgtun)Z_ySiZKX@9mEV%x;w6t~`|h#hSzYQo
zWtKbZ?0mhVVdDFzlRfSZSrp-od+2akAilnnmKM1N>eGG50_z*J`b_vy7lOFUIK~iY
z#1JFPS=bsogC3b#y#E|NH^VIb0<61UR(JXCD#^~qrKsVNi!2wP;Zs_qPPGDUYk=gm
zlcE{etaNuVr|N2TlK<)Irk3$xay{c}m`nCB&no`Z*Imei`U}rn%F%MB&nT<|wNo=A
zlxp3+qs7<nC~qL4zPE;S{dxv#Tw%Am@eISWT@1ZfH3#wnZC#PQKG;n#x|^o&7&dLs
zN(J?5%6xJPp{x)QhX}kL9w4jOZcs)Dz(xzoN5FDgLP3(^hxgA-P9bLaC@cFWCp~NK
zKd9ez?$G$z{_Yc}YwR?w;RQRVFNXhAx*(Z#K%OPHsSAC2^_ps)k?Fz5=-(f!=nT(c
zX_2#amG_DPo5Hi7+B%o)l7OU2;T}z+;83pCWr!A>XdBj~LTUmMm>Fe-^0I4y_=8gp
zq@=>MR*r75HlAuD=1N5R>!W1QNq=QBX`c2lO%aqrobl~<=2n;wWw<Zkwhlj(mV3WD
zbe1#fe-I@(SxYW(W+SiMc`}uHa}XeW>BNB=D9ov=!s6CI_&S37q}QsOeO+7Sg4VQQ
zX{u(g+hvfC5b$hC=uta(_)*)>h*rtNMeXK~Tta>2+VWBQ-93gcT#5icTN&v4-vPOl
zCFr!%CC7YxP;ZB+ed(I-f=;nN#RmNBC94-(xIm`1gtA;r;(HWIGe!8gEzWfFfYt5f
z?sZZHA*0KtxQKP<YUEbJ`k3=&g#wGKY%12~_2*jtUf56#;@kzrT3*!RR(@b0)gdUB
zH55En-)m(KDl}oF08$w)Pmulm6PqBof&C>jwyBUxez~&9t^nZQLSi<HK>3_7<;7qo
zb5lL5Ewz}j&T+A|ng(A(gA~k;YSzK+mpuj@S}tXjPZym#2Gx-<mEn_2t6%kfd!vqg
zApN>6?rH(9fZ#nwAK-_r%P-`W5!H+?s0!R%x61I?2nI?`nVLglKq+TF7a;)P_;X#v
zddR0I9~I$5s|jw)Q}#%6xPxa?lzsej@oAx;H`xtVO1`^$v$MOp@8Ic%mpLKG=S*a1
zfcf;50$;!TQw}mxUgwgZY}dcyDrTLWmXtg@l<(Y~Tu;mdOizd7L*g&Uv>z&Hh;le2
zV;~uD;_%_S^BZ+_f~i8yVOnL5717>WuMKRm&uXON9SzFmYyg)I@ym<nlGO-Q^4&@j
zf>g#Q5V$pPIIO)4gY(}ckw~S#93tTVVM^)T;9dH~)m1wA$M0J+eho@~InjQ}cFNkj
z`gv`xceOtA$kiV}J}a~6;NXw|-R>UEf7IPgtr>sU)!7)XqNXfn;;A_m<BjmoL-F=K
zpkE6IoJQRYJ@auSHS*s{zke=3t=oU@T<1=L6#R46pNFynU0drr5AGRRt;KeFQ8g6Q
z`d#*ayNbg?;zp%pL4Ee!jq=T#)13)rv`?LnMQJg@F=uuRqnsx5ytI1}0R{z|R&*9T
zCYM}U>1hC50Wma$d>Ra#0>fxlhA)OIea5^Ec}^hb4`Ti+@dp?6$W4xuo@H8rGE(@p
z{@~9Y<{{T;;~J2bb~E5e2D}U;{$M7BPzYMgJ~2-(rOD`@>7Do?qxDMYF@OClnR!9U
zy`JZp&K_owa@XX%A3Fcw;)m~h$_hIB!;!1X$~6^k)XYepNEqzi{BO{D@EA1`&>OJG
z0eQswuxaFHoV-256|~9q^mG@_;sKgo6M)I>Kr8!mI~!EeKtr4M??gCMa?#T_;*u2(
zNgL}^(2f`8yHuDhevxX-CDj$XWSCCQLjyts^!@~``}`HOZW}zdIcCWFv!0ONQ*LVo
za6_Ob1U0Gre2qUx1}sD6t+gl-u5w^k_eFGq2lpr3)%pD37z`Fc*J}l!l){C)pjv#-
z^OO7Td<eq&Ji+>mloZ|3KOO<ygDca=L_p0SA|Cv~u!l>x!L33a?Ll`2E^XP`X6uNv
zhe^?K=G)UF0J7c+K<l9_w3hMeRK}?0xe(Z0d7aH8l?Rj#w|RB>ox7fyKBZqBc3SsA
z#E0%A!h~h$%;RR__29BA*P2R`(bZA0V5Dwqo$7#Ow~@FGfsRJ!oEbIomu$&C)3*Ka
zE_)v@H#|EW2C$Bna2=*^QdfG($J-3}tz6^W8b^6Y2SWp_G+c5_x$JXW#8NmJlhO4$
z0`}s6bqM8u<oGX;lkSYi&24F7f`t3;|075IpZ+`Nwy9h8O3Ua(!tj3i|I=^w@4vsV
zhf)$9uI)<5XbVceNcj2hzu@10g{sx1xgQhXn(q*QI(PT)xtwC{%sftkfJ6P}JY?Gk
zw~9nQZ)_N5FtR@kUoD#pFR=;Dt@Nz^b<UW-_(!#@{}aUPI|UU-%kSN!XSo=K)-gY|
zhC)pV3jsfB-OlTado=v2q_xEHnzim$jz?$Ab6I}}?~d28OS2Vsw|8bK=+{AZJC`I9
ztx!oEq18abSN^!wV_$7)j9OSww2GHc*zZdG_FAdeZDLd0i&XlVS+f05VkcF2n7)`+
zy1j>C@#*~DHOIPy#0XdGkA`Kt5+{GQC~9osi;CtgzNmq19{55my{Lp^%i3-nBqVnH
zq||3zzPy9<3p$ZnYEbU?!}Sl4GFgK&9gh+8y~<iXUWo3FG+c<U&O3ibcUtWNvxLOs
zK5-_)bfn;2*}zZ(kS7R(t@U&y=`qQ+k<&6kqtq}GHnJ2<g>7Kwic*1xnpF0Q#;Zmr
zV+t==(!Kk|M}i*jEY>9cDwnp6KCZN-xT3N>+`NN(29z!;cO}-!Cw3(^+XBR2!FMHW
zEsCJ8+}VQ_%a?2uGbhA{qce8`-FGEmeThT#IhU4Q34ZF()*j+cQKBW*tC?Ev{_!D5
z!&F-S%UN-K<fL5*j&#M^K}WIWPKx2X#3<+HHwUeDtdrKvC#^Rt6U7!+fyjcz^N_>4
z67MzK!TkDG%73}h!?UPXG~WYoStexdG$mdeQ`(itcWC}#;Jz-kb&7Rb%=X`v_=kb)
zrlQgZVT#gB8+a12nmBGJy8-ZB8kZ;j?%zDGwsy7lx`@Kq`Nm&+)y(3Lr6U$SX}c10
zTd=K7u|{;_PN?&)#6KB{1F+ewz#Jixxc%LL`={u{OJk=Jv9h}o{cE^&{+0vaDBhl|
z+4+UAD^ZH;RUOe@<T`GgEN$*s-brczx&wyVyAmaH7F^DM*784V`JeCd|BF2?`>J$O
zrD{L*8^&+QaRr;Q;M4dU?OD3`hu3X0WnI0)(_fPw75X4Xv(7&4ph7JzeT*wMu1{*2
z6PDPl<5^a9^F)T%+@`ECdZRX!uQf(WIBILZEnjQ)<FJ#gUrk8plob+obZazjJNAM=
zOeC*pI^!V*Ly{}OX8XU9d0FHH?@94?HF-Gw!^>YzV}8S5jyYB9WtFWEtaCEyN3Ys3
z@3z`%%vp6^h^HdECbF#)K<vUmuL4SaO6rAEpK|fhc^ZIc5sYb8e@1Kya;M{2R4{#~
z^)J2wfLA1gc}6gK*8_lxdSLv{c1#_R3K&ni-I{WE6Wd!~{ts5`l`I3CeP!H%lE>L!
zO6OyaEczN|v^5_3wa7eu6#re~&644Y(EG%B=v&914to7B_@u-CGp^9Q$5)D%iZGmF
zP{i=lwfXb;wzu=1j$8RWcC^kK_Vh`-=@r*^*^RCf86-}pkR8mo8^5|zU*<XGwq+oC
z;x*xv7KKj@j)0mt8LAUvHSi=qJOKqpZQc$KMLlJ*oMb|48gD+@zbWG*#v(m^+_+U-
zb+SR*t>K{}W6-yoBovWGS7=8B?titzE91Kof3?{{>;G8y|9Hy(L>>Q$;Qs#~!AaY-
z=6H49yy?t?_guDgJ$L=bK1BDXtbxXN*KWGm*{*7Ti}~)q1hlJW4$C{y=+)Gr1hZx2
zt!>{u##@|sHKs9TOce2$$jKpZ=BP=*S46|Om7>CdYU|$~&y-d<saO&=Q?5Q{9155G
z^nBm7!QRtX4V>g^(};JRkUw$d5*!M$ZIH3~M7Rs~2~1LYgxDw>)Xi6X#c%`=?_^?b
z_Vn}lEgy0^0FXyQ67=b?a$3D~n}KzK>{*^GVe&!`7+S4MNtv6x9e=2N`0q)TwHmH(
zy>i^BOsx6m{<Xg+cm3ac)~|6niP!G!NG`3reBt@`teM126dK>?y{ui04K{Gl<?8G|
z2h_1N(hqVlm#-RdxsOot5AHxqv(&#nkt;69`mW&gdFr>BC8iHLNWXpp)Bq&yKV)-o
zg606`f&@2h)@pKEyi`MH3I1t1xg{B6k{BQmH+?u%*Yp64%c6Lq!q<Bd?muX2TksNw
zYKFfUlKqf2ydxampm&15i}w%f?hI^OFXo&FN8ER#3!CVzZvQ;g0FF*XEyVdwfAQ&U
z9QBpjMwg=YPVBg_iPrA<vM+IggxQsdJKsEKh&u#QbWfzxW`?_1aVlWzlNbVi3!LQQ
z+_*HVafeRM7~3)4;6sE3i2enP0F|h?vlz1xsyH1x$>U7(1fs-70gOz3eX-501Fbcl
z)Qq;^g_?y+LHK#`RS-~lR*HhSkIa}%dYID?uK5j(g@+IO3Y#GAA!MF0DP~5ct3kZI
ziE0oxgeuNX_ej#DAo0bEQ}pra1rA0?KU7CG*_Dutxo~|RWwN>~x-D{>1sdaKQG7t_
zu^p;7C5Yz>>2FPV)1ZD@KNHVExJz!DDZ%Nx656j4CiY6Bir<^(^UWLkA#tJOEWIU@
zl5vO~Zz_3dDS4GsIM#s<9g_8J`zSzJxGZd##TPu3hePNVA#C|w3B?4x^EUm8<J!LC
zW{HO&l1GqjybZG?afSAF(@ib9S4)b}G748U#p<FL#`j7sEY6tlF+8P^mS$XzJZi?1
z#ViPoiF3OtFe~FNN1Hd%l}(!`TjL88;x_5QaU~Pqg~Jk=IPJuo97VUdm|clZ49)|;
z5-IYgznR#oSc=c$3BI`?;3>K`k;a=djQKMaDXiugplm7>kMij`N4F0(G&Je1M#u1X
zhC1WwH@`})TNS|AtZh#gT}_|Q;h~$GH(*s6Uq(~SAWEL1@g<bm*Q|x4#EAI*X#whj
zU{@lG)xlfBHJR;<hXn6RtWr12hH`ScKCMbB?YRFT`lZcnR)DJG^LQdiab7&1C*)6$
zj|jy6>unQ)?FSQci71``^$&$jPBLEz!&M1mHWP+$;bg_|=_Ryw6_=-&ST#M&?aC3G
z$#1Ss_pWZc&4xNc+^KXlBvGkiG90Fu*xZHFvN%lqh-_QfXs}{!jUOHmZ0xUU>uzb-
zw2bd(CMsT_<Ae$F3dQezmoip1!kK2xjVqtHX{==wM9XSdB4<RfnIvuf%DkyVC=_is
zRjB0o?i>}FS1t0EH~0%Ui@PDRAyqt-d2=~$85i>MvVyo*2xE6|hSPG$M4TUM+#)W1
z+ZrdoDKd=~?AZ(##E9gRH*>&VNn-I1sS8oYg1Gqc#M+o-VSfLTfX~}N?Eorhy+zZ>
z1pz(hUdy`0m<X(EZqyRfj4F2Bm54G6<@8v5MGoe3SX`cHgM-sj-zXcB&1p@5?F|t&
zz1zOnX1kEQBtRKws4R+n)R^J0=FOOHy51sRq%yH@y^R+g_drax*lx0e(0{tuOi^oy
zZ%)MB;S`GTzRXBLcWcxXy{nnpMbA)UNrD^McB5xyCXmT*J}z%uO08Q`95)dPXMEAx
zkQkmq@yBr3E8ls5)Qeo=Ow$xN4I)kw*e5Y&N_JgfpS*ONOX8uF&BSANt)lI^)vZY!
z<nV}~dlR)il`D+lbJn45h=Nd{k}Iq2D;k$PE)cAKy4dDgoV*d=%i>LCm{;l1Fg@lR
z%!ZghzQ8Gr&y=KR4n>Naf1;w|;+14A4yU~|CXyu;?G^;_O}i2nXg$b?NlCLFB$Ou*
z7cOZhNh)@H3szH{(Z7XTiZg!zRucvpPUdZj>cK7ruvb*_)QFxrxLL0i4o!a3h5=bO
zom}0FE2IgV-k=gXY~YJ%-;Kn>Ns^17W^i$$MB2u73@#BY<-*3c)yUQo`1aLCraoO9
zXGtz>@TY~Fp$kintvq1TwvG9shE0K><Rfx=4Lo%MnHRPiX5nQ_WG^iVtf%PFxHx(%
zI4K`202mNn&J|k52uvrJMQ)dtBY^p%nr9u_na$uY<W5a3XNXHg&0mq5e5+9&F&lBd
z<meq?+lS__#Xj#U=8+ngH91#b7Pm|ec}MqNi=k!wfxVjg%WaqM27MWqVy~)tcs@(1
zEi|%c-VCPp4eUx>DQHYeOSB$!(6sf{_^}?f*9f#x)7Ed1Wi{O@lEuU)y**9!;xhc%
z6ijkSI(AvfwewvSVb+K=6s5F>{h}Kn>dPvj^D_*Vy^Szco#4@gJ3}NUXZT2=gY>!n
z3oovos`}+iud!b?WJfDU6N?hwQkvWY15`Ye%z6?Bx2x_tQ?c_+zP?)R{b#9IEtN1g
zBv1sncIg4^>eu9>lOAtg{2WO0dl-AhAgkg`I-tqAcW}LJO?f2@xG0|BPamb_uB7dm
zBWMm`f$q+*LqI-cA<*c^JKda-(<gd9+#F_|P`X)@gScDN@bpSw_SsVt#m}y86~<A2
zSm;=@aMqErRXN>PaPfJ@)SVbzwMnDEO2&l`hJz(HzNlq7T)(J>e<pa5wKetWtE1S4
z`b_X=RaaMAR?92VZbHkoN$Feh%Mye;j4WQNT9N%tVoW=Lhw>IzJlWB5`~G;<^)YSJ
zg_!u?VA>(|BdaCD+7Ua0AGeJ8JJ5jb#D%cJ1WDs#y+TyN`}lN+6(M@{-X+fah}aDv
z16ma#F5D2rbnAv7TyeB6;F8-dT^sW^bB9UlOY+vV#lmr4v(SayjJRCcJQPfCSv1wm
zqM3<2e9?L&?d0vk!vxM}+huY0o}jT(F<}Y76~^K=<ADgB+JZ<mu5rh<8|+2?RuK+z
zsIFP5a(c$Sd|`RR3?eZ}x?oLyZ!WGY+(|(2$1Z&SHUsGl2qKCHJP|;}-Vs`B)ipNn
z{cTsGW*S_i(`p15r$2oaojuKr;gIj72oxH#^c3p(tGp%t>Qu6zTQtB;V{5CWRf)#p
z^P7)_I=-?v90xNm;h|qa_Uem#5;g^)t$Zb~l@;Pt8+4K<=)wWZ=RPJ<V=a6a#RHZv
z-nBVV9vD7%DRKgzGtONS)dRzKc0(cNPm}s+69)o|B<8<OD%i}4uVOGl`G46|QPWaj
zsN~XTD_~yXNnUR({E{cIX=$cF=0x>qh{>uTg1>M9#aTi(r>t_&79oy`{Uq&xmWAkg
z;Hx}YGr(_m2v92Hti{!>hWCEx4Lk4wQkz?*=(TZ;G=v)+-Rg=X?Fg&h#uq{?6{g;+
zEEe)N!pF=2LGQ{(9>mTfE)q@4=${skC#kt1n&&=p=s8;GoJrxO)cRHNSX>N$L!2m;
z7tI&%AT~oc;@!U@X&<inWU<#;G#*eja(;<!W{so7iJP10U4l4~LLDEc9TyYd><Esi
z(3(e1Pp`(9J^bvkW)nV!I#SG=7R^MW1bXRbdu6`887YWwb_V``BMw3nj*kg8VB}$7
zS41UUJVE5ru7pN2=CiMg5f_e$n?DIwO;=Yp=<g;tbdmenFwUn#>Qh}KN>Pa`xDYe>
zembr;VP||o@x96ayq9X5*7ofvq#wS=Tv#n3?wo^-Oi$CE2&cN{F(xHs9xSJsPRChj
z#qCN=Tf%DG>KZE!|JGdl`E%~{9a;D;6f#@5A^<E`o8fedPkduO4HB+Uvak`k%AZ2V
zS)wjbCV3m%iZowtJbG^$`Q@)3rGgk#!Uku9znqel=gZ<Oh)U2Cv~~{=(<-k<LrQXX
z+<$892d>yUUt~Oq63v*^Pmya?M#EuEam6gU4vQ_tp_z-bAbgV6n0sr2TmM$!7-Adp
zAz_FDncaXXG}Nz7k<I80VqBb<zmkF4T;b3Yj6vv^6GCRX7AlS}#p7D#9pytMr<n<h
zX5yWM_c6G*<(U)?1_xOIQP1M4V5JpRob#^;b{qt-Zz2WJHpWDj<n;K8Fnn4}QYlpa
zd`GW*>Ewx}DZ0pvF5gXWyisofs}V0%uHc>kf3hfuZ-^3x(^`_jXV1fc;8}5O#{@*z
z>#>s?=yxBtNQs;AS;B9Oae*k$5xPZGzF>+=K+)9nMXHGtC=X*ac-;)iMY{^^cZ(o|
z4_MeR|9!H;3k4o)X-U|+?b|VfOKc`nC(PpeZ3G!7!OQu7*a7f%GWdcxR7?z?PEn38
z7zJ;~yqLEU1<U|Ct{a8=aa=oXBe}U|WtFp_^s}<MFutJaeazCr)JV7WUVVCPLetLD
z{ZJOY`H7IHj%{P{axC^H%{MoKEMWM1F^OH=#kPoIf!K-=&sh*IK@v7ZJMZJf1v-nS
ztDK1yk)oncA``^89rEx+0je=SVOJtVacBFp84!sZc9wi0i=+2fBBMX=g61(n^nUWl
zw=-swkhxW1f`u>ab;gA#9E3eZX9<BYFFnqlo7T0_kdFqT!J`I<-SS@P#5OdB#}mZF
zeLsWS=;tel3&$-Uk*6MxReLd=DCNdwr3I%d#XmNd_b=dc%ERl<6_0f;nwFtyDwnKv
zpT56j_fw}nw)^+z?LVZNjX)dI8h&C8$Ij3OwLdu*g;m=GFiA2nn^=~?WbaS{+n^g6
z+kn-FQmZZ381^iZCxI<><H|D8wX2t$k?qsgxs(3F%eh}qXA@Hs`6zYk+z?z+M>1S-
z-PRSz($&~lb*cygNb20-I~tL%|CU@?*6p;|Nn`s~GxPVaxBiwZ{BJ(#mmu<eesNdg
z8{t;@&y|1AqDh3A%=|1r8%(_l(i2o{G3cPy%V8FB`X|3p<EDu`h1XYqCLj#T&RAa!
zaDEKPO8sv>DONu9{FIN+bEVsok{C>;EF``XZVT0e+W`>|C>##{4nRvia2q;c-r(yo
zl`nK#T7$`K9e`QZ0}v*_0wPFGMw?((@2Ad=AC6lxeDUD7c?ITO)A|J?RZTtHj9?q?
zVHG@cMt`p8^bg(X`&AWz*0Pj7WWe#JpQbk88qoIR1um4!`uHJ(N4Kxn40jAHx^HPE
zS+S52LpKb%;$!5}b#&lTk>kA!>y~i;X;FALhxh!WYR!u&C;AXy#j?FQ7+d>tu7+D9
zS3U%T>&2$dU~kpf*uakhPOeZUMUydtD>wG}iF?1gZ~lghV(TU{nE68~<{NY--+-1P
zw|{<%blDPfc0oleuKUr;oxLS)+!FP?8fYfD;&v35{7ZJ3r#CEm3NNtQCfc$JV=B!z
za9X$Hy|#>JvwTWhMor79yAsR(@2XNFK?ZMNMT`u<z51lW`O&bq^w<s3>)wt-7rS+@
zKb-0;GgXt{qHJDwGScnSYa6utqn72e3SFnP*gP9dIUK!bDsWdqAu?fZnXBFUf)Z{x
zmVzIOZ7-Iq(NB2qW3Jy5JuF*JzwsP!utJFt0yZjussnY<d<mgWG&L3n6HnY|9cI)V
zpaikoi2MAPJU(^XMX9nw87W8G3>?Ergi?I+3^tp5Nn)Zz9f~@#u3;Y5^rrtByR1#6
z>ZIh4K3Rr2-X0^*g-~;4+nR4{S&qur3Y6zGZVAyP25HC1l>$g#iP4d9xZ%3Oo7m~p
zSL&0cX-|HdK%~B4ftd>}uOWDZFqf>?e;oD0S_El2Rclc*n0Q>;26~I)O3|_lF6`HT
zf5S>odWdFk)9;h#hdWESKyU6bzH^f#$9`?FZ_pq`Q#H5V(zx$qd6n*I!z3;?<#hHD
z82*N)?v)<Zdl!$ST$!tFF1b;58EPne*z>wMNuME=`TM7FS{EKL2a<?RWZiS@6k@7R
zswz#=+}6<#U(tMPE&MHEfbmDbh~X085G}vfL#$OVDO_udUK^>mq~*iy5lix+W2(7L
z1M#Lk`o*d_DIEJOKTS_e-J2#RHKR|m{EB~Pk)KldtsyhNn|7+|qt`<W*BsxK1wvX}
zIIqa>hgbGqqK#g#gSRvEMZHH|)8Mqem%6bY#9Y&-=jf!j-8Ep?$|0oJ3eF;pSyecp
zW80YPRd<8fo&obl5sd-+1d8xxW>Ek>{sy$r=WIt?BaPTmiRjk>(=YaR2)KV!TD7B<
zp%TKKr9^Rm;!4W=^fnq4$QVlN*A;cQ=FKyQ{VW8IHUKes%zX_DO9F70N(y?LroDp>
ztr3SN1}4lU*v`?q<HW#kjq=Ge2?<M8OQw@M!{WZ}q69jb!xI?`ZMx%|ed!|t{!(&!
zbzl6W#c_nrj4oUE!!|{9{OM!F{*j8C9_vp7agdwp-qS-zHC?RqPODLfaCIY_baTHm
z>B?P$ah4-4iRv#jUGCoYobEM4(28_*7e|OYn)Byhe1hg}H+w8uA)=-_Jr?^SRhNcz
z=DPaz&8Qm;AvvKQjASn)wpY1%b0uR7`$dtISq^IfyVu#UzXBVbxern`7issa^1FCA
zGW}6B|4d2K6}!~5{mLcT-j~$<<*pTkNf};$%JK?uIeRf8$@}+3Ho1Pb-t8W7(qGTt
zgzC>g+GYk9MyD=%j28`B9R(~C2*QA^L1b$Eq)NqLZTm-7jSG;2T79X!uoBGc^s7*u
zMj#_x-c`XzW4XzI>I7TmsxQnHyo9-mSHz&e7rzi^Eg)%Y8;e)?ylwM-`X$b^1tlRK
zhi)=Pwuvi6&o?infD)5Z2$gGsjsqD5YNwdAv$2!k+$2cQ72VmY7wt+Ujckaf?u$XS
zhW7a+T_kNDnyuZbm|)9S>T1|V6}pA<^)?0QI&O25MTx~yBRYAfe*uT9ny}C@YZ|ZO
zaM4ZiD=p>P>J4TM5DT@Ww<p#9yurlyg>7#EXJK=5YDGx1iEpNIa273Ozv&@a0c|m7
z`n#+@?*)7`0*H#pj>X>C-cCFm0b&nuQ}h;a%<*O&7QW7JC*RYw9^=QJK9ujHcC7LR
z!KH@qO8`4vbGE?FmeXiZIO|ta*V~!;M)P!cJzoXhc#jgfQ$m=Y%}XfrRD-{n-*Ql>
zqjJx!+1LS1ykIOcf}tbT>uM)uhGH-Ju!i)-MzXhh`YWd9>KiSrx=2B>RCcHyOcL5j
zvSZlT&|Hu551brIj7z-XK8~_D9#2goQ&%=O5+EB>MN3QDkOYe)O3Z&-aLZGFS*8Io
z&cbX*ac&+>7RK?XHv~)arhTi#madkm)#y0oBb~9ZB4#})hveyesbu8F?}+oP5U8g@
zd8B=8yMm*`02X!n1oFuW;%<8N)oRQU(^1_ZExAp}b-BThy@mpPZ!p)dit#ARB|E@p
zi--)%aQ#~~V?0oaV&(<*y4nIaNz3tjHXGYS%`|3S#$}g|;gtz<?KNQT5<^TTtd9-8
z0Hh~2@JMPz_ll(1<^sg21&7*jy*}iKig-Wzivp=}>C*cjXY?Kq>1VzmgzhJH!5^Et
zs+oTGB<S}i<ZsVhdxG`c-)3>RYN;RX8R6F(RyJ2Mr<3(du<y=Nfu~}3_%zI*6^gJ7
z*J~YhN;t8UuVLWcOxDA*br_DlJ7_Gz5>#;ENwMO^F{Cw^V9^6ib#hK;>(6_2JTOW5
zZpJcx_SNJyzm8RpcNvp;31$x!PnmWtm_E?w-#DpdIBEaj4|NsgW<OHe6T7g_oX(W9
zpT{@>FS36IApl=G14cOu{c<rWB`GPnONMZJ$6!KfJ)m@!hJNOChDHq8tT6y*kR{*?
z{>(b1Rv5jCZuz_ca>zWC#SJfYk%=JQEz!cP%_2l`yrLm)dj;Yh=qFSAo&Urtf_gmj
zC#xTxUb*(nCcduw^mT)<AMCj)Dz<`CXbkRl#s%`3D*tJ2ejC*u0eJ7Ac1Kx&uG}Mc
z2~aMWSqn-!FD#ST(ysMZ1Y>NDMiE<wX+ofY8R@dZUMWCnIIS38{4{<vc*MXlz+{j1
zqxpu|TTkl~G?dPlSe>qP(T!O3>H3-ZjABDm^?H&Edu}6nF6HcLsfX7VY6YaFUh+dN
zAGdTjt1e@w?4QQ-)AT$_o2@8Xbf`HTY1<p7s?HlP9JOp6J&_gQ&vI0gm36?k$%gBk
z?XHs1YMYM-L)5_-5-HC^olDlyr<pf4PI`ht$bT8^SrKF$M_!m>>lt!bf;cfQ1hSEg
z93NNe&k>H%+uegeQf`5W=PzX_g^cjmVQj@Yj?c2VBtRydyf74g^n95@-@=}_&V^qQ
zSp3wevgJP<O*HD&)1MPAfdSK#Yw3a>SA!!*NWVUKj%$q2v<UkcWQN5Yo!`o)MunF5
zd-zxCk76sWT^CDJeJbGY!I)J+_*>3frs@00vG=2bGnJ-pxJas#G$Zy*qqIw!ox2rD
zjdXUsI`J(YwIZn}rRnvTP0C<RbCjAO@y%cp*P3mY2x0UpV0%FQsMaE)`AjHejz7f{
z#{23lp|uj|i35LqP9;Z+#AT6%Qrul!1)91c{(KJX&Nut#{A<ZMdJtz68t1nerU&{g
zv~4egmOhJ}H@x@Ci3_!kO0$LT?pYSmU++rzAa^B1BOn*MmSkD)z9aK4u|rmAc}~UF
z<??3b`oHWr=yQRk2l$lztS#9)u)_y$NPlNgrQjQiDzMKFu~KQz3`w6_J!f}$#>Ewp
z<A%v}8>y?!z<u{Hcd@R-el0wYb-0}4az8t-g#bPKb}i|<wt&1|g1lVG5rn<vFWO(H
z-#}IsJ5oOB!V0thF&UVl!82C~z5uu&<H^p>DFbIu9hXBGXH8L$MBL(gm|5l&b=?L^
zv8T#3u9|kea|u4;@Z?3xRo&A^>|LK-zIX^ln?Tj@`dbD}V$Vh2*jtzHdaluL*08U!
z?jRfHH&U}sNlIp1wL>0uw@$77_&#&rfof}SnAN2h)howB`DL*v%-kVlmS58!H@j%z
z2Yz^O1!dd365FQeFgl+0b!;7O|LxwT6)!^{)`D@;kwkdMukC-P7m3^20kiSH<gLA#
z<dbOg=&eSG#m-Q`$X2*ExlS>|T|ll7;6x@c_S%6vJ35OTp?@<PCrGs1m1tN8D9w;p
zA%F2(0#H#*NT?HW#^rwsAunuhl`Nto_=_vuD3$Ku-#w$#56;+SI-3R#{y3E#PFf6f
zJ9XA)K#tPx)$?@X<UXfA9?ILE+-IA4#HNeW+12*CGF;<=a=WT#?mW<4i5)Xfv(=*+
zX}c1z5tj5jfI^3poPo0BPfdBvxRKLSJOmEVkx9m*0Qh)gBw$|w!z~Qg@YZm9I~%&Z
zI8dCqFdk|A;KYcQH96hwMRDre=Qb~+zG*;~>il@Q<kT%?ec$n)Tdt_s(QW)5j$E@R
z+uzT`xI|Nze0;s^Qqt<zuGa%5Y%2m1aCuIL1A&1u62-3`LHGA7BH&5Al6suGUJH$^
zB##&Q1=JVLmB40<62IJTBX~A2!p;_zk!@mn;B7hzo4FaiFJ|T>F~>_EQ4gSBc4e;e
zw7fgET^mF(X|EupnAJ_Oc)RIcnnyyDF@ICcZrB;l`*&nQ?}T`6M&Z!mUGEco5+S`p
zg(x9yjUyITt|&3%aUpbuMFgrO=d*nY71BXO^L*MdDJO~cX?f$nIa%$cM%<&O(a%m6
z_~N-goh)Y~l4tgBJ9XsGy^0f#(OMJzQK)7m9C~80(6hOD`A_dX@m~Xxe+5P1f5stT
zONag|4xudzVMeN_Q@vvZ7r&Bvv(hz6Rjaw#>x7=l&e94#C)2OhfjrOs-Eg~JJ?LS6
z44@a0a}ElCT)m2k!_*JQ1{-v<FR8c7C={v+Zorv33oa>yvneFrJAA4M$cI$R__=74
z_*=O#F`F`gWUFF&+6|L-Yj}{@9Q&zF&wkCBEbZtBsCi+M(^3#yR%=IT#mb2;gs(cj
zhNi1-QZYE%>-r|kDF}Z1xX);^!*{BuF*>KuztT5t;hk(f^XSPF+`TMWjT09X-@7Sr
zoo?03%uzj66sL^>$7<}up<Rw=0Fz*9aRc0@+a|PCw6Uu5>iiPBy$k3UskR}V8$HC7
za==Q6nIverkj?_#2o%)7(gtQ;OjgAHCE_#5xBko4;e3t`r<C;Mtv|m!JE8S|kB%;J
zIV0PkhJ9GO@?SCHe+y!>)<cD$=}iV}t)t{_5%s4{4b7?MT&$rvvLY|%_-PgU_8z;-
z%=Aq2aEy`;WnaxaaZ~elYgm2Kb=(^H`AIJXy|@FNbCnPbZT_R%?nQvJ4>qiH2J3{W
z6KKPwru9cibyWK}zoG<kD4r({cS<2%>oZXJA=NjPG<kcx(I3MT8Mqdu0;wawX|Bd<
zU0yZVAL6|EtXJBzEYGNH*xcDw^#<qk+gJMtR%WEu>a?;Sv+So*A@N7aIsuNipEXp^
zVT*&f@V0U_*FNO_eg(H;DGND)HLNLguHLA~jM^h9&tJEkKN02LcL$qSm|pggnKHLN
za?8Xp5Q!=hM@|mV-)dl{(4@c@C%$~YZn|#Vls~ZvK|a<xm-6n9*1164{m8dzAI|Eg
zHdSJDGt9z~P4$xjzmG%B4P>t*jh6zIs!X&=9w|}_mm=L89HtkhIdv;7yLHW$8_a-8
zx+38H@_aohW^`^`v}t5QS%OZ)sM(7JAc;KLemV^!i%&HMa91XtK=GR?B-BkRdp3k0
zS=009am!0Sz{)Eh#5kI|x~5wuqpQ64bbhsHRq!R^^s{F=TBjbmde}srxc;Jnm0tIx
zO!ey4nPv|jHpY;|s!-$srU`ez^XF9WKds|ReS%1(_yc_=^`cU6VL(Y5$KjBAxQ@Y|
zqvWtO0m;#S8gQ0kYXOa^ABi1dV`cxRu7pi19BONCWBnvnc(is9KdN>jbf11X`G}vh
zz1z_;b={lxZo2w>?eC}Y9-g_ArMWYzYPnP6-7&2KoJC^d*;G|kwUima=i%Ym4C$DR
zUg8ME2<YgoJ-;g+E4N-dj8QC9eob&Jl+i=Z>d4@q8RoTlE3`)Vy|l|P{PliW+o^%c
zBha-cg^35|^L+cKu8VKqlgnqWSC+(UE6wqJQ*f@w3ajR%LXt7q!ir*nPfV&Oc}On!
z`P%V;rh8b^2t=A*X}rv?#D}7#J85>=z4NLZJ;`3KGHFFVoa6f!BNec;%14hYLB?Cy
zX4`HFcA^r_=j<%4>+EZTulA&%={+ywK#Sq2qw16l_Q{mdt!<m9e+0c;)zsOP`7o&$
zj=+1jA@ztks=C<6E%%o$&h)tNQ7pmF8+>B*uOE=-=0^FndpDR-kHaV3QwW%<EGRac
z;jsV)>*g2R3|f2bZNhC<1k--ZhhHCC2U2IZdvgMYCWms;66GEd^^SJC(Fp5D?GnbJ
z_OKT<2DYFqMwhlW{HxpC+4#ScN_1*M6J%*h05YKep7TA0_x2ncU?+BGm26#2uYc;@
zNq9wtUW$xhsGD{^{J^_k$ZdS};%C?QNhfTp!%{dVt9~^&U9Do*p|$jyU$(kjI=5tw
zy49eKW_m>9gJYPqhzL(+W4HjXo)K&62?)~#f;__%Lhym%{^~ftu5PSj$X0H>2Yuba
zL%rV7&5Gp!0xs1H$aR=rKM=CoEAP~TfqCDy_2{L&Bot`X#)Z2CpOCr@`Q&#F`-n1i
z;{>dF(lkgjsG_*<hbLJx@demw_Zok3*?dhbTQ-Q%N`q6MwAsV^?X7GX(v~YaQi||U
z^8zFbIb0vs(=k`?&x<74b;F>MGk%0Z4<B_vbTgyskys;jemQn&2|d>(TJoLJCU*g^
zL#I=5q`3MFV-d*8zu6S7tXt!x{+*7W)WnRS8iWD0t8+@hRdY)rU;yRF>mAX?wzc5J
zuEck{5<J<cNT64i!gHkK#1YhOU}Yot$UvAVoXOtYl^{|=K$Xy~Jn^hZyh01Q^gdGb
z_mrUj1CRTEv96yidj3)Wft`^hn-jcd4MwT>S!L7H+OigD!(V(QPDGQZR{1-u&aVOu
z8-IS=5p*tG!$(Th@WRRD8XYzan(b2l>bLsf(_YUs;S0$9x7C;_M^41mpM7$CKIfS*
zIqxC6U!RRD_CmCFj>>3z&tn{d;lbiYRR-PmvL;~`&_q8Ht&QQ=LoI61=&;sxyVxef
zmfm|U>3U)QrufAnogm-acZ^<4_~pic_#r(C-sm6V*3lR|ZQxRGd~Dr`&?9YS3*E*>
ziQqmNUg1V#6#ejH>4&zmM=Q=e?!!L26qfm2q2K+r8_>}Q*o?<N7GE<J919)Ur%klJ
z6@Ya1n;;k0!qQBe;0p(~A`VLJp(6e34Q>&(W4rphZ}kWEAGPJG<JC!ENDWH0HS2eT
z$xInWw>7LS9nhR$Sd|Tg8^By0WC~=J84hP^Xz<arDz0iMg^+rdkZ1125a~(73~Gq-
zlgQF@BbqU8`_tZ+x+8+BE)&0zIS`&a@A(JJrobTOYc|^stxfo;(+FDq{iyy6fwl9R
z+{n<~(vpzHa+H*93BxK1>3o;3qebgGgJswO#y}Y!W=s1;5gYPKxZV#%Ivx$b<gy0v
zCykMhQVL*J{UMWd_oKJpTVcUJ!tGr^*vGAQ6h}efjDG9&c>6YUggv_v*L%yWT9&yh
zk!R^0Ggdy*Y<^f>J(B~`IUve-MJNx+Z0MHG(m=NU{_e2xuP96&qsvjOvK|MOb?A#b
z$vQ53hMA!F6SI}TFq*6wbY8T)aXIpjQd|Fc*KUgHf=iMA4AmZPD6-Od|I>2Go}E7)
z6f_}G_BU!Mx&h75@L<o`ydaXMr^gf=f$aB)RizXaG8hVR-3%QW8BWxYQ?ZkL{@%Ss
zDNns$9@mC&U{4of{lCq4HZUN2{I){Br%CM=`5Em}<?$dgOZjQ=W0pqMYr56B@gvnP
zCqBPLxjkcD_3v}il04YF>kjQj(e>HTu9~g<bI$Wzvd_B1s+OZGVJ37k1z$fGcX_3#
zy>CG4POPDHIn2Jw@`g7{s7_9K5b5EAHLwjf2dp)Ilf)-h4rPGiY^OG09<Oux@rqLA
zyYj|EoZNV;2(Ic@>V?>to0(rYkmPI?H6NYzj~T?oT!W<XY;3XjpF*qG?@zcjscp$k
z1~tm;Ofcnko|jI|RXW@#dG2p`?gru`h|bry&0a(orRALKmml(P@J}RisZe-CY(2N(
z!V0Car!GR=Sa<&A8nebGExL~K)F3L}>bQ|wnnT4cssUUwvU#`$;WjnMurgR{a}5m$
z)=)&*SEgZ!)(V)+K_k_%THElOeYJTr3o;+jgDtV9-##6_C>0cO+}W<D2!sefvjzok
z#+=<wrxd+^@PggmXJF<`n5%+f$v)@2Eg9w&ljBsQ_#iuj(X^a7QW=)PFcG#V9<1)6
zPY8nYQPK(3EI}ADUQIPMc2&!(*GEmAF@iUt-aukOpS4gqPj%1?Oza0b1DysgeMCE;
zs@+;letBuYH&<I#io?z{vtXQyrgO_b2j;T<xeNL@kp@0-{DAfrw4{ZQILx-*@yr2?
zId#k(01QBn?1)Bm=EkFUB_2o#T19pIjn*hZJj|2|+dhtd<vC8zdFx9ibFbp|xss<`
zGTHwy9%+bvddmmXYf&0C5bkV~oWp)|dmrW1D;KTjgNnaSm*02veKn_2p3F-AA-Oww
z#pn^LLOUSnME2bJP{0L;juM-;4zeROx;Ko`MGgTBYSLC6RsN-99lCV44b9JMe4pW<
z)-+q+uE5IZe07(!zln@S3E>giF)S^z+!56J;K3oTDfLn`&{0+l$U((92t)D%3~xUS
zD!zH*gX`gbQQ1bYe!i#Db2sa!GUjJKnn!t>GVusPZi%`|FPBg@<S1SG&Gis`s7AS2
zVr9ee{?+P7L3VC6H{2rQDPi;Z#GGI&omkbnIkHvfS(1M$g;*aMKq8b7OyhFi^H%~c
zoUrfckGP)gd3yqp@9Zg;W{-sCQB?iwfhI<gvR~>QvOge7^h~8RVnbT&P<iPM!~<Pk
z)Q@5J9G%}+p6IgK|Dys;{d?!Ggd&s9mw3uOxYM3Y8U5v}j<<MRgXIzfRjEfEv2QpZ
zxQ4Xk3v!^8F1u_LIb0vLPbZ9L8<`LIbM2wwt=0xZMjmPXiiC9xfe7zK^qA|&YiXu8
z@ItswESJ#U5W^Sw0X0m57;76@iZ?B(Icil#>JANK3<F6oM@u9U4#M%pr-9e{?S;=v
zsh81M%l1znLG-$F_WA3~pI9rQPBw-u-qO!$p&^|<;d^TGus_5eDnHBVy~r{5Yz=UR
zwDbwfZy8M6mG+jbq*!$ZK@DbH05=AgpplGEsWG#bw&BZ*ZUtN$c0{^z%YycW-)Sd)
z*{~Q!B*Q!Xi)8?Hj&g3E389WEPY+fjmUN!Tg1@P~r&x`#D#ud#6%0G@h+BZ+d@zHA
zWHfq!&GW~rf_*HNV#MjOHa|>H8=6hXxOq)|FbspNo*J!C5_`)m-l4*N95`o{W-vWY
z4I|2wyAc@D?lr;G=uOG6bB;Ytlr7UO(@EvjGE7rt!{J%dbfbIt$B$#!p|Z9rf3^|H
zwFcTbTNU{U(NQ@+BbG;`o?XwfM%bDS*|b_`@N4}_S?B58%D^Ts`Lq-7Cd~cr&t#vP
z*Ld}fPC-yvV0lvZ;1AUmE{~4RUcdS7V6)sMtwy=)hblAzp_}-SM#DuPX(K>xsAcDC
z@#U_B4ZFP_0fz&}+lr`)G@W6GN_O|d0aZBE_SjHtV7ayTyv&N8da>FG)lv4Fo~mw_
zj-nJ`H}m{wg4kqpNmnVoR*=j00BwVpOxYF(H<xE{%4S_-KuaY3+2&UVueAs1j^Uv_
zpJd-%^G8X)3iKcGKljjuTXgze0YlXk`(yxV?PMC6F}IQ=b%n8X$8=9-cD(_@8J|zl
z;jZ5r7Wx$m@M|tqII<sZ2aVT<>cge1fPUhBT+jwQi{KZz-WXut&f4;^l626Fnp-C?
zxe%i9pPaEz%kZglC3&XAd4jS!-qbe_`*7l=;ESUEC|NQ)u$8P<Jm5X(Eo)2-Y^`~g
zGNj*Cd*h!ez9Dj!XIzIASf`Bi@voC77nD+y&yg<P=v_%vLJXGkdcv*Wv0|HS)+St<
z!-qUSA;5%t;1;voiI7G%Rn<hgtRsgcgEg?(R=R9O>2V;qND-h3Q`{{#Go<&B4;l_=
zv55HK9zw7io*ID!3SR--B23;qAXo86+g2CKZN6X9hiNbZRW1~j)&B7gCSTD;Qf}&U
zIP~gAw=7*K4Po=<c5)y4vG;%((EPxE$U?TB+Uc2_Y{I>Y+M}3W>bEJo;N@`{KH3Fz
zCzyB=1&|<Rs0PD<P;*IPDd5ret%ky~?riwtsB1Rm?t4w!u!Fiy=>bQ?(ya&rrU<B`
z2+dsZx%F`OVN5RGGh^l8OZ>;4B=7U*Rz@$xm<BxIN(Z^{72lpMdi}CI`1-*TYzY0q
z<9Ag56JgK;9XYcx*D|hMrF7bRWmN{O{o3fM%pNT$z<FM&k36&G9V2zann>^!lvOzx
z7^*g52p;HXJ43I^CSq(q22lsKSTuID<GM@aOmmS-rL@!flh%YD$71CWD+OB{XgJy|
zTx2_b|J3{;>Bn;qL!Zpecz%J5fha^lew{zkJpi@&c|O;fc^92k`Rw~lWtF+Dq$$&`
z*{$qn?LeP7L#v`>Vm{iatTCm)V-T8F2-r`p?XB$Tf&ierNQs*xsF#eieC>bCM9wYJ
zp#bAUbh;I_=s8|Y$5Rul4D2kH0SPLms5CdVESK2n;Sl9IG6i~~A=gDqUgAEb5YX1+
zH@!D+*gm@*H?9^E0YdLF0OpMI-!^-UsW;O=4xp4+O(A8}Cuyb5YPt0Xk3llXKTk%$
zexkW6x*$6p938Fp3HbF>)l=7?=~<wT{0(C;@7qiNSRgVFCa5hP2-H*#H8^*Nx|aS!
z*s4~47FVMvIMidBS<alta0GWURE;wVqkV$d0WQTY1;k7Yq1^$OX(+z@507E8g9QPI
z04h?_?Ai7Dt)U{gP{5yQsCKzeLY9l-`@}J{b<e}kNbmErO?MP+x>JXb1sk}Vt{&i)
zMYgZ@8rYcF9?Q_MF=FWtmfY<2MOYNOuT(%;tgKtCk=V;Y^J`2$vd5x<y=#S$)-+M}
zY$cSW!sypY;H+mmTJYikpWFG#UbFA%LlFlxT-z8r=Xy(RhGGYy2CS09a%nl`j96km
z{Y5TTKcVgaV(!hq+0Og_f7dlLo#{TRXsuIfFNQ?yTr(9~5=JA`-daL}sD0->(`ic-
zRaFw(G%6w?YK_D;(;{}&h%J`Zo)CLt*U#(v{so`&J?H&B-`|tGUgzYT=kj>mAGaI0
z<8&5ehS2Yc^ZwqAYi`n88`6f2IX@=%;llqMbDu*9lJTZ*9?hE^2|1&Z*X8GNH80>A
z%+d0Wrz2x2_w0HiKjlUeLbD>rC_ZyR+#oIJ`~GG)H4gBUc~gbC?5oqQ0Ea#mFw*M*
zSUk@jV#(wjddZhQVs?d6WT>VEoidtiLupfJ9`8w=kdpwYB;2{mvmrYfwblWL0K8Cf
z5782SPW4g|;(61&{o9Z8_S)q?qLOY_opW=}(aC$FsUm)0F>x`C6i)<bx&Nin3dHp~
zU=EhoZ~@cH>#x7hpF?Sq;?nZ-H-^mTd?yw`F4J32zLXm!k;?d&E|*89hfot&wOKVX
zV<l>r%;NBp^|mCHyZki!7_%4^m4kD?UIidvBX>bMHk*9BsJ;mF+sL+BflRcNb;u+-
zpTMW5qZTJuO7W?{qJjr(bg2ZAY1@G5xutUQ1Ky~TMr%=CRvLyv^ub~MyoelwF;;Ru
zg-odc^l+a4+SrQ_9fQF0cca~FyRV5<DqOCa1rM2XWN#%i-@mV|y}eztGLe|SqjaaL
zZ7{;FSGJP`t#`lwic{b7$8=ECX=I)3JxKE8h?&<|R0z9i!?XQKs1EhoKF6D3Z;=00
z3|o3`3?AlL2y=G{>4XR(UE&v`JZTJuVNi8ojC1?rFR1h~LDj!GQV}{n?069r5;Aa9
zr#QmF$504lMovyNr-#y<?Fp$GmLdZ>IVbj&aYacB)%egCx4S*3Zkjt+WobtQZzlcn
zp~etRd->(M>4owUN5rf%#0u3^QPEPB`JrQiVp{ofzFM81d`8gn9JA8ZS>v;3bJeq}
zH5(-a$sdrk7~!2gBG@z$MP5Dvb;3f+V|$wPwl%7iRfVgbiDRLv1L=*Fv>hCvn%Z!L
z#dWwb#3d{cavyqn!!O>M3vEu9qdG)>nQ}20y*kIzx|A1~_M#>x?1HRf$Hp?OnXgpl
zG|-$0YxMCQ_HA1@tJl-bdsP}pD>_rDd!fbr1HEdo-*;u!fke#-B?|eqQ6Bxz6P(uo
zM0zx!)l4O665<*8#hL!tpvcAjw6<{S@Rm;~fA&{G$l)N>xuT1$r?)Z9h#}>tJHe^B
z1D=3VbH=3EaI*;t;H$cszA<gudbBZDIE3pKng65Yr`jm>oBy%f@T+u}?Tx#GZnQ0E
zmNU3nfGev|H;b|`Uq(uz#wU#Tz-}07;X!XcgHObqFzDAfZrk|cq!g=|CGvdkfx_82
zNAR5I^gcuwPgMGOUO{!XGCxxtx2#l9H=|@Elt(CTChTRHyKzaR5bsCzCax68HS4DI
zY~N_4bMBB19AX1i6_OrIvwQVJ>@sf78|A*+mdQISn#$0A?HzLG`eU1EH|NdCgZQrO
zcuM718XhGqsJZIrvw+Azy4(W9=*3#)qtO5hU5f-XB1rhO=PndMl^tf?EJVwx;p$AO
z+7&@TBY(9f+ouwTrYCcf+I^~BW&ooV5lsNZI>fkad*e$oQKg~93e&BKl0T19_9;D#
z{k}fIj$ZnS0x9J;to>K9lW;E)tK?&is@W;{gK{jdW0mkdc6LEn?e|q|b;hOn3rHk-
zM(&LDUJ1IUApII6I@oQu5809PRqP?KxLJ<oD;F1{9JSxv&RfoOanQS5-@gDkpqz7G
zJ!}1`&_?B)fv&89soU!}kIU8ykB{7vQT8gX(&1VStD2pyo<I@ay|Wd@cA8u&;o|2r
z$!TlV4x5kR&E;QEeE0VA=a#TK7D$LXi?qglUf?uC!ciDGC<;k~(eYKxMI0zA_Wumo
zj16Dp`u|-aEctQ7EWj<&c`4C$WT~lLb&8U0_7b-o;tSG@)R}wt>e&rep0~dNX`w8C
zJLB0x?d1AE+FWB%OWj5wVCXT?p<_5PrI%I74CtYD|G;=n1Zmo<mLm!D`xlv-fP|LJ
z$Eo^E%p_l54uNU8HCjYlRF_USL#xZiNfaGdT2pGnYE{P+NuA*l>F6n~Mbexj9AYT>
zc;zn}WyVs>&8o2^jZXj3e_95jq%|rYLzZlbocCYFnjY%5>qPKWs+zV+->uRjuax1H
zU*5y$U#sgF2$IF_chVaGyY3W=z1yO-r(OediEdvtxQ5ea_gevMcmym0Xg3!3=X{yT
z$*zt?RHLsoxY1Kg9s6sr%^Wy{d|Lf)RCA6vNc<`r2$2lcUb|;_CfHrGoKeBn{>-B#
zP@NjG2J-&o)OekQS@hnsMMoniRGv`9b5%y?3ddFo{GWOIy)7^Yt?Yph|1%b#IpFs6
z<Isms*lAK~KjnB}pBQlDMQllkrsU(|Nw{zThZ`T?v@wGG93>q_WB<;4R4X)%=F`HI
z`k{D=@{vVjA#NQZVX!OY_$c84fcCGDurxnHM802>>O(i7wPTv6!<*Ks*#5CAy^VI+
z3pccEnU}$3e@$iGs6*USZsHDTz?*W+pR#vq8enpsHfK(x-|Ymuce9!$Se^y-!0|^R
zlGD-Q0{3o!It_rto$S5~c({OO;Ek@o)@sO2OI|WRijdp&Z14cUAqPmg9WcjXjRWg+
zbYlP95Wha8$V@ykv7Al#ZvEPkQot!V?%2ZolqO78*iURNoiK0OJ2o%zx6EQ9a{itk
zEUpYnivw>cS}HypR@1#+l5@~SKfDtmqrhGAJ?-<45Z4<zL0fFll10gz-)D`P0HddR
zfZ<oK1t>d@k+VWaR@!f^b=D%befKQ?n`@B0b5JfkwV2K78#)SAzxZr2`bB$_JS$N2
z0`e!tBB+4v;wCf8zrvB#7l&*KjY4`BU{282ytkz-WxE&8hF~nxo}taGbtBt<jsA{$
ze|~B3JkIXDJKIcws!$m?j+yJsv9|o>+!c4kE!NBY%Z3`lYz=Znj#}mVAGZBQ+3b+6
zptg*WnPD{K9pAnHv9MvFblCoVe|1f&DBShoc$M7?T*42}rZZSk!K&`4sh2^2WzvaU
zs~4?Wh84|}SP2Z3ZdzUGJpS`B`Zpxibt;vDv6V{umaSi1Y~|nbq(aC18a~dy`_-3S
zM7PRn{|V*^H2QNx-gm8xSFQ?W58eqhuIJrYZaJyx47us*E=x0-<0uVKxMy=d&bJyO
z<*AawKT);!{dB?K(@pqx1Nu{4mSPhLn(q5&qUSG`U0^(Py-M<h=wG$HTZn&gDfr4l
z4o|gdD$G;m$4f!+JEA1ED5XsVI}^<QerT{y@p8x?i2?lRgcQ%uYr2oHmrJuws)B%o
z-Tr|C_w?_s%&Dz4{IN{c{Km<<z$6BMrV{RnY>ptudq+Oz&+%hN$-E=Q*@SN)TJ4sT
zS=Dq_vBgA5j78Yy->prj{U@HMauFh{g#GQ3=L?IB-8ShXoln=}LATR;W~BJk^6Bo8
zgvnS@CVJ+K<9v70K2)##9{nfHD1@YBbxs3FVC$Lt{5tPxssn#Lst^F&#Vl2jX4;b>
zK<rgY{U4Ihxn^Tw*ojLyImHoGg9upInBo}122MpHIQrvjrcKG$)>=(}Yd{US=RD#T
zaF&pmg1j}~sZ<8M0OV~>CI!xwh$!pO7%dVo?U%p&Pi+3yO?KgXg-53gRZ+&TQ4RV#
z{~Dlbl+7`|wdQ;ki=|>f29BAzS7dlM=wuIaSoy{C%8oa=p1g`!ZF(*EaBbrK`?~mo
zC@8h*^->p9vLgcS7_UYVIGQ;~A2PP{Gvm^2?6)GkSc632T7TatX%Q4Y0}o5?blN#`
z8en8<10MK+OnhcsT8YM#g|b^52|tJr0jA0iO|$;<p=RLy$koJoaawAsIHuuWsYTN|
zh%nzW@|w9Z{Jst0DK~g3*Ns7o2zTqTG<<w`O11WCu4(8`pKwixU-pth9xg5RzB5@G
z0vMx9L6ln@Fr0xeR&Ty)gSoLATk(dM>$B3_XHyd-EA3fYsw*=f4rNBfCk1<w2ukXt
zP%9V74yy<xgJC&Ytk*_()Hzy&7&gB(M!Cg?2fc&ejprej{9+L5+`lmu+By{r9{z|*
zC6MRT5WF;u+&^ktpv}3P`O+_U^oq_!?@qxt(6NT1*1TsFmy)*#U!q>+WV=iTS5)au
z8dQO-!vQazzETiH2ykUJs03zgWW4ti=1et_XR-uwF0z-ZK(5fjbv^2gq$)jIeW(*i
zAb-3>pw@JPCxOU{0YcotY%6TJMK|IH8P8lya(Ym7_PP?}o3j%KUFIbE<-yj9G)erG
zTPmH;>#v|{E<RO=oaBKZw?mlZ;UZO!m17y3=abH;>a@2(G8R-ib%+74lq<4M_qJh#
zGe`xHbtDu4p)j1-0PPdP!>3Z3mW+?09L_e}^3>Hrd=h$sJj>5ayMV$rwHT`cM=K<9
zQqURM;fd4WQ&S08q}%}5L82o8ZJ)PHm?#?NwH5r<)4e_Q)FEPQjJ5EpOnr&I*pq}&
zC4Ruo8v6A{-3fW!JyN=$|GKUceR!rQn~GsL`F!z3aH0iCOQ={}AVA*@Vt9*R9hMtF
z$i>xk<&*^k#fiWJhB7UM{WvC|bur4E8bETKwyrwbfRM-`aq-!n8_zdqIvk@jjB96>
z#)E`qdj+Y31yuFo)s<SuE0>C)1mzh0;2gEms~;zfSe`>dY!-Sh{@CvXR~NE2&=X+;
z#8XC+oXyR+X(g)_HHN5m7jXaBEN;%|BolQb2fbM;n!nUKK^O2_9bpgP{b_Rx9VtlU
z3Z1=boLXLwbkJMYn<mERy<w)89=0l539nQbtp`Z)8}TD%U)ukLrYAl9$qky%g6pr|
zys^ui%3r8;yHcMWm3q(G{u5mRU$ydT{US(7|B)uEIa>l2Ih|GmM^4vyGMfmxj(U1v
z7r^*7oA%UXr2gV<n^Db=LRm$dLDC6TV#vTfzZJ~J3MAP>02PFsPbp3$na$l3k9%+G
z{3!AF1x#{nAgz$MYGB~3&Br3M`zlHsa{XA5t#&rvzEawq85t4hK!mv)r1N&*#_%}u
zQ{7UQ{P${C3JB5JRm*%y?^aA?e}WXvs}2stCEAmPh8Pwdv&%m5v}nZk+qh8k7qh9G
z5pJwr%5wq2=KVkE5Nuaz@(KUeGzGX3pgJj4?BvDhxe6SAB|Wmg(>!tWA9wP&S7_s!
znk<L#eB<1x_KjYYWblO}F~VB!exoL@BLB6z>vjF`i|18~{KxQyKL%I$eTjs!YL8)T
z!(^;|j!_)JqB`Aby80UnbsV)p{3^$jvVzAk<1)$lBr*<&w;Jc&Zf-FffzVVf%CE-~
zXt_BN9LIvk25_(ko<JrF4C3G(4ic~dAU0eQ_z~M&$67D37|hz6`6uD2zRUkpD49uW
zIR+{t`A2kS%E3l}!l1>~Q;U$yNvJj5-!zqdD*)^EyJH!%FCoO_l6296WS9<mMTL|(
z=)_ufad_L~(-U&o&3;&>b8qlw^NohGAHc$a2?gb`a<_?ip}<?FMPM;#PQ@EF(vVnS
zs4&lm;DaMGP-LL2WgucWU;;`aXJpS59u+iO*Oe_Go)tw4`=a%+T--{7%P_5f0Ehsu
z!6FezIXJU<C^TVc06gZhQ#IwsG^SkVNV!zy9gg_}61X>pXASZfXRu2!;digCrp{4t
zZ$D6SALOa1q(=VsYAQkr`;G0w1Xvzn|0&UeZSXueXtch?oSsFV_yF8ePOiQBGMozs
z^3#(<z){GyF_@;95aX&-@dibvrifXYr@1*fnDw%}CSBfM?Gm2U8He=eWFglJoPnse
zfe_gVsmL$S3``TqzIpgQ;I~IG!cx1t!R+nb$I`Uo5zXh-nmvlCu4zVGJC8~%!A`GA
zS~AZ=RZ{7n)DVr1TLt&h%p0xtzm)oYnV%Wng-0NJTQStUcWd|Qr|(zzsimsNF34g=
zIf1E-3@4KS@m}t{!X^Irfr|jc?8+kfx-zc1^|r<XcVvoEMYRG8Y8bB+OTsk60gGY>
z?1L*ksy4%F^4>cxHSzJa5qYDLJL`Ae3l-1iudT&%x~};kUU(qVLc$gcZ5&mNPdwr(
zZ|vP@+4i*vY=Ky&x;gKcn#^by7Fjy7Ny>)sK@u*MOJq`s1+NA`;4v!>2ab>f7FR%N
zIp9<U|56jBUF0m^U(j?getip+Mh2M5&N@gan1f^z$v84igGb3P_F?)|5HY-pGSpWw
z`e&uBqnpsfjT^_wjwvk5Eu!hpvO-e%UPsbwQZ{!2pf1G0qBj(o0UJm3<!7ps>hxcl
z?$;YLnu-kmCf|y>{977bYPk2!+pwp@ZlOON!W(=aYTiO?{US_FrH@WrFMrWBqp`l0
z@up#Y@c~D%a%Y3fj^yR131c=8NJPAbXn^X>r{z=Y$HZZ^-#<O_>f%qmuBr1$ae3&<
zAP027I>t~Vz%IoXLAhf<Rx)jwh{^ZhzMacq!iJ0(e~U`<-*a`QHt(`)Ze1z;hDd&r
z^kQVBjr>DPfzL@I9+~#CQ7-y|Tg<8~%)e~ne4Yv>c5CU=W|pGqm5hLJs7x`H3^k6s
zE|cq`9{{W+a3I(QIC_jTE1Ka5kUV(H%1XCrcrN1@M%k@u1(oyiLCz5K3=5oIDQRT4
z6X{7RW~wIgNK<U_Wk4CH|1L5iUHV5)nTC-5Sw}58s5~oM@5N`gT*rG0JPs|}P)~B=
zs_powz|JEA{06kZ{o5d0*pNwh6)+w&a}Ir7sx`R)Gjqnjo!gL28n{2EJ!U0DIQg=@
zE2wFiaYTa70n~tnPK?)0XI7=x1X&3Cr|e4VqPyO}4Fyvqnh#mXhD`$spHsQ9(BgP<
z@y1Yqc#JZ&xik(?%RY|?ovhBUA<Q^0T-R1X{bFaA2@C6^dOTV!pPTSdcC8O1zFT#=
zNjfSJsww|8ViO*VKcs&1!0oG;dKi|UAKQR-UZ0>xH@ndM9I?09(HPPRfq`RC$OUx*
zJU@RibU*H}&oE*71ey~B@9inSN8|7Xj^P8%k}ig*DFm`$B(gsY4w9Om==gx%f5;nz
zzS?L+A&W3iFFjv&h)BeF-5V%f_C!}O%vNPn<F2;oG|K((ag;7Fr45JS3?jsdZLbWj
z|5<j6IKAcKQdhno^tjyjv;P!}18zBQw5J_8=KJr?!o+w%7cy$Ppb$<b0YS9|jz<i>
z7?rSb5lo$!`DrP@FTiIw1wImioux-(h@z-f`UX2Zj7&@|J!!-+kf320!m>`mQSs?k
zBf>?(a7v+Cf#3K=)x$2B4c-V<)T}q#wi~vG|L!ZkioEa0dC^9~N$6$%0;veHz4-nv
zJ`<ichcx&sNcwm5;1?40b^VE{@cBgN;ddYpb^s6Lv~W0E$~J&U!QA=0rTDnD(zyK;
z$WDwD{nT*<32kOf0MVk8XxI{9M2T71av%$n+i3YjWts+m&N$VJ;}~2Ot2Sv}^+dnd
zfn-c`e9#&1_ZTfvvshu(A8EuZlRLC?QNFo9ly_s>qZQh&NTvxVwH;+KlqCP&6}lB{
zel|a|B0qCyNd5r7Z&47<XMBFr8{$S=^j^H+0KJOzM06g>&k=kAQZtPOT<*qgO4tQ9
zK4nWeN2F*mDI0n@6d3qUQ0wdDu03Bh-nTqFd1l^<hp0^hSD-(WPfx~ue%=PIsg>Jp
zT<Ox;lb*ey^(9l=2m5uxt=*DmkMGGwRa_L%s^}9@=R!tHtPRS0qUi3fDj`izvdaP;
zue>0F33s}?AIyHo#|f(?D5NF=W?CA916mUlOy9_os;b!>AYIR7OZ4f-Hp|PnJbT4~
zb+GB$Y<9;EtP4^i2XlzcpHEwEH*YyoUPVHx^U)T+2WEbz>3d-MWXDwQt0}&%G<C4l
zyws?5rFdta`9|-1{tNz(ne%ySFH;Y#t^5LVB9oRMS7G7*GzR-Fy7h8rs(HrJNe;ss
zhm-u$ldR!zYI~izhj7@Y&?-i7Z3j6SN3fs$xF@-~zo*wQ>$;mDbrR`a$6vvH6{D@v
z;q8-|N=mp@ph`IOA*k7_F9wlPFu(uQCe!!jqhN(GlXnR>P4!X!QFgZOn_mvNmm-SB
z=Nx<L;x7Jd?az+7iSntMZTYz|8d8#5l$DcXUrL@}3)pFpcye=e3Tko4J$x#j8Mj2k
zQ>vr%%Y#B-tz+3m8P8G_6)G>tGR8D}gf;NK=u#1x&R?O$0}5F_&Ub#gDN&Y?)Vg{5
zf1b!pTNhydyv-jW+EJnIfl4|7les1@@5NhrWD%2|LVMqD%`_GL4jn{kP+dP!=S2w!
zy<{4psl*QoVzrs>W$w}c7=eYxljrf}wj&kG?Qv;!ax6W;5?K_D&(iP+-+wT}_1`O}
zxMzLzlpWAtBZ$1r%Nbtx)2_SYYMXMD8Xz@qWXV-dqq8CChxVp=s<pqC60))r!LdWX
zKl}Yzko`x4X<ymP=da8-C!6E6eRhImrlpgdrYVH=0SOq(1?Ff5>+}k(Z9f}PyA!@1
z5JZ6FX6NB^B4p2;9Mb$?9_^leUX9vNr^!{I4^d9RWCExQAM+SxdWh}hC~4@^iTz{w
zbd*v}5F*Hf|Ca3+sXrA(Hhi`H{6*-mFPo~tXty7aW5(l61#d=2vU3|7d7t)@b3*Tb
z!g7#=RSJ-`MdNhm{T#&&^)?VgSI4Y8mh*uqE(S>B`jCC$J&i(bYQ@wip{Wx)o}kf;
z5->?rV<RDqjXfbWolrCjlFUK!PQcU@2L{*wJ1Hd{h^B>~a=Ufx)rKe~ZRW?Pr(ea)
zv`>BeNY%`wspR!pv&I;I>@Gc`@x4GWT4c6WLJO2T5S7%;Z5@19FsFDzTt7YDqBq<Z
z#U))@UUZmW&z-2T6buDL1C0kN)9(m?afj|uEW6}evy2PY|6hdb{x815+|uRuc+Yj7
za>R5${$G6l|Mwq}O7=X_yu5GI&I$f5^MB`Uzj>hxzFBK80KaHX@vV|R?j1Ru{nyF)
z?tg!G_-R{_b@10sn?Uw__uo2SN^7ld5pQQH6KSXs=H#MH<Vc{c1Bv7v{L=Owdt}E^
zMJu$V#Ow97n6-^*Sx8%eA}DyCTj!!Txuo-k3_WsmaN@{=XX;`oXcDe9<*E<~PmGsd
z!58?5{uzBlp3{Fsw}$tPbRhdD``GoonrJ`<h38WSjvQewo{@){lsQ6BX`yQ~v+s;P
zESlFfh4i3|{q$~2!#0EwY=mX9Y)X4BX*Wo;KBc`x^dG8P+2ou^M0r{)O--4YOAgzY
zXH>Wkv<3n7_=qbf)h?JlC)hyr3Ip!EOwk}<iZzy0QF<|Djk#p^se~9;8jjD_kK{H%
zutFG@p69bR-5fyj0ajFi)oet>@JYrL^xO@yJQ^4<Oqn?`PVfyO;CNh&XqeV0ECUfW
zGy}Kp*Sxyx4g0vOR+7q>9#98YzV5-7-Bng{jJ?xiySc%7e*Q2t{pd(qiBM536dvsk
zI1QLbZ--2F(~)`ujzCpFNr_DHL8<ln_?Z8=$yI$`mEMS*%qOHcbLnOdW=Ir8fRvb2
zRn;tS1&OmP#^+Gq@BNPRe%PGBQ6C#HvTf_VwargE6NEshhr4D?I2+4%X73JPzeKC~
zba(f9-WizRHvdjKr^F`YLm5{0j7Vi#lZgTsUH(24BUMV$VsRp2K}N14o}`jBPCQd%
zfDSrx(c7+^DG_z$$Cj5Z@!a>q9Sh(dOMF)ul6FKtJLddv&0f}K!%hcmETLS1riKF;
z*I1*Kxt~!tTpzyh+XZv9jRb$L8fCETJbLbcoAL@%O2E60@=vE;G(@Hoxwejq4|?3|
zT-dwv4K&PQS+DkkhiHb2E+?YZAc|W6soY|_82EWHu$@<T0Y7=Kk8fI}tlgav1Hr%>
zv<#BDzPyZyZQPnm<I&QH3to(bVoE%kRfk=IMnze%p%hBmTHevbg}+(aZ|z*5`6D}0
zhfFQUD~{stUG#r-Dqa43_zExD|Nf<keELN5W80c^7$-s5`2%r456lL6#zh~-68OaY
z{GAg@MV8K{aLl7U;QRk@YV|R^x{SBWF7Y%?iivs$gENW4P~zq@NBZ}+cW!@^^#-eb
z$L1GJw+s!PlzTZY`VOmhe_P+(l&-%$-WNFlSs^i*d=d^#0SdRzoDkv$4#yQ6qLv_r
z6B2NcA%@||F(laL_#6tjaDElTXU_VH{^Z5@Ib0UnDGD4<HrOTgrsis=y}XjW59d#F
z&|{bqk)h*Pv3!7lHM|Z1hPu^Qf+K~0vg~6^V(OL!hd^|^W*HcT6dj+GxG>{H<w5@=
zSi3!Xg8fir6}Y1hR}5Q`I_KtSp!*-7z!!hjK2JsaypM0r%*_b!!OY}<*-Mn$22TRj
zKH{FBOCWJFP-t3eX2?(_Fe3|`kSQi6=)KMe7WVq=-$C`s%rU)YXiR-R1$VI2KHk#+
zV>?(3adG8YA$t8H{Ov}%ugf*-8vav_7h`>ujMAZ6GUBhnP|9oelPK*Dja~61+gbn7
z66H+|-777=Mj>kj*O`~Ha??DUoA&CbGzC))GgE9SbsRD4i2(u$5@h1{>@KY~?bR4m
zxyY*G1haKy|KafT!p<-YCL@!)lZmFNntl~~pQZ8ThR5|V*~*fHri@lU4eaQbrZ)Gu
zln3vBn5aq>-dk0Q@hwc{V&1>^2)u|F-<?VNVNH4I{tzp;B{Au79*}zFwdNH(4v?Dm
zfV$`);&sOOCqnl+XcR}lBgHNGxL*DzF=%hE(SQpcgVa3Q0k{?pq7W`+j7roh@i&;F
z5zzY_x#%yg2`)Vk{mnw6_+72!j1H_$rOUfDl`9hj0_Was<6gJqw==IBU(I(}r#NE-
zpc|QcA?dHpzU6{F2ul^MZ>LV{j+{FC;-aA9>!ker)NKKKs3u8}AllfUXbPfV?bHiG
z(B&17aOPGZVfly?+24N>rZ93g)Kn_%1%SPb5FJS73WT+W-&KZvzVGTjm<+Z0!6f^o
z5`|fYc{KQ~=2Ka$d4%|3y$$-61Yld$!#6rleVPvt^c2RUDyzP~+aB$slv;o{NjB%^
zupO@M7`T9tK*ua%0AW@G*%%XQOuPS)-%U|gbm!c1&lDoB`Y!vq6OAdTrc_)4LPFd@
zu}=K`Bl@h`u;qnyg*-xD`0ByPYT`NVCj#oZ@Cl^EX8Kh>kwa}r&boxndBtd7tGkFk
z^%I>S@U;)8*Cva!I%dVeCl;A4a>3@-AzZo>ltP;in0Rg*&mf?S3+Qu%@_H-I+TyEY
zXtPYovjhoC6Ix;a^OhfIi%vU$EB2b<vMNnYR4%G=#iwp~bXrVgLbTF<Z3=N-U@AWg
z^}fNWOx9La#<YSuBTw75t>u@UY*wjdcjG1P0=``yYfSyER5q2zdb;d93q7xYjs1ap
z)SQ7Z{k7)p;#6x6Y-xt0s@aNMEs8M%_LqWOAT-{8$_eY=`EX-{YU-s1|1eq&tXP8k
zl%ECY`rQfAwER_An1r~56;uyql{{enhU6ySdV1OP;V1L>Puf2Y2z%e>dDT?xj>b0L
zPFq0mxJ|j4)4t^7h6;X$RMk<wsOHP$-MhOA80No=!LNKk%!o+s6B_nnJH>x4uu+#6
zw@(l8c;5=8Da|0$M}#{L1_&%gm7-Bztjb(cCNy(n38US^yV?=<oxPlm`y-EPwLR&0
zAbc<oJ67djBe1dpA|!BxSp?fAHZ`G@+~1J)Pjg(YhE7P>DtL0jP-f|+`#bF$QA0j!
zZRCh+U~_qva?@Ir(d0{76(?H@pH5AA=sIGCp4g8AL!I^YfU!Eb6VTAKVFxXtX{vw?
z*8$X@kkl9Cw>b09W_pzW{1|wqCMTJqnMOdVCZ}2L0mESph(*h@%S@vtdtc#x`H!AV
zk71oicf;*l#kKEUhhEV`ewul306woSLN$~`*B^x^l>YTq?5>IidHcN3`3149#q}e2
z)j}a?Onbi5U~ZzL6$`k^aTSA|xReGnt1X~%8oQz3`CDD0-e}$WMwQfb-$|t3bT2&|
z;GE&!)vD3uU@Sd}8NgY*C(wF~JNu^UaQ&CFcU4sK5G6N7(gqLj_l_hz>BAE;Y!*SF
z&CHGW_wk#nK|^t0#U6SAdYHwIq~v9XxfJnsLpdAq)4sZ|V!w6)G|rUtUrj!Vl7Svt
ztLSn`y}oWr%!k>X<>jrc^{t0VbvtXvW^EED>jCtHbfqtMYZ_wXx-E4x(shQux4azi
zcuwXEFmiJAZe#Yq;H%h1G=47uKsVWUe-+EMIn`MePGQ!(K6^Lz?z_Z*Yji-&$s=Of
zyKE0Z!`+4WJeA3`YHK&SXeKOmxG?)#|Hr-G{;bTn>09}!zCQ%Psu~K=nerfxnsax7
z>SAZ44{xNSgCjAlC)p2m)B|@6NnBzfM)V^$zKa%2xt(~%=(^Q8bqpoM=q@Am^kDxb
z^+>SII{Iz&xKGtTU-VSGI<51U(??A!ndWJY*NZLR-i-@WHAjT}tCx4Fwn667?eBH&
zs?2y=q(998=~#K@Wlic_(wSg`Zn+9$LimvrI0haMxPlK$@k>5&MB_ct;e7R`jdOp`
zlu4%cDm#37zyu46?6hpa3d}(r5prQ3Tgyko=9G)pPby4)zodM#;|`wWAOlj@x#Ry7
zTMl8Pb!>Xne?6<Q6wv(=I@I@0#yo1~Q)FN6DJKSeY{I7mv~ZB=Knlp>06^Crx7T(3
z$O$BlR*<<Q6EwXj7Yb&Hd^B1;tZd#ZbLBe1*4F#nje3J>v_}Oqry~WL`|<z$Ve@*G
zZ|nM;^ZJ)F<RvruMfDJ}CIOE%DHDf0y_nRhU^5)rpTIEHjhJZQtZEK}_~o-s?lqo0
zB=a34gilf)^hq{r&d5!ku>dTA*sWpvnkR6Ch0{O`nJ(#IPE@BfQIe)2hBTDs=}%`6
zSo+yPzuSbCP=jH<%wa;noy>1uH_>jtY>2YEIEeCNu8duM{ZiId<!0p*ou5Xqx^L23
z<#+UGf{PAkaDo$%ghzXf(nv~U=krRKL#Y?da&C&&1!t<^^M#TJHpMnuv(id4Job(j
zshzkOI2v-l*-{XZps1zAld9N5LZxd~slBSYh-Hi4a!2!W=0zu254w_p#NCeNya>k1
zLy0e`ad#BIh_)<}*v)@W8blbAmcpCt(EZDb!+EV20AFI*Oxe*TUanr|OV~%!+D{|z
zUEeZo0hcfc-1N4%IsRAAX6KTIcWR=aw;S+<xcLTiX93{Ea*?xQL09vPwwF@#w=+Sk
z-=hVHqNW{RQSDZJez`(gCWc2__KZS+h9o2>=>RmKuQ+izUuHnVgE3TF*d>B9qs~Qp
z2E7>SHQUZF{6p#vHHM_=;r?xt11tOAp*9S7gBqFdZkKfBt~lz##hOybw?1DSGoa3x
z_YGLhRNy?0%3sb}+XaSJS#ZaG2faSYrFC<+Ru?t>O5958^%$K2v!OMPG|ETSx82D>
zyx+meK@Juj8PJK`+YT}N(n-eC5sJ>?GIDp<iZyOr?2d*{OI?i0sqlMs<<*p2_FZ)S
z^+;Ez&X-q;Ydkj99-H5fOIAS}?HGdBLd&3pIaJ`X)Oxfzo}`wa<?4f~5OJ;|hJg0(
z+>pJq?z73VL!X&sE8oM_2yw}<?9lT{0)sH|A!=)<-a4T?HS_*hhK_Lr0c(e8!Wv#!
zUT}?0e>l$h_S+_|4TKNW1KKE*r}iyqbX|@K;zCYlmG)m_pTh3`ay@hB??kFhGG}^X
z0@>N(p&enUrL@q|?_QHp?(5a+4L4sVO-nieu(w5rWO>UBBYK@7WbAzx3G3&-G9H^R
z0!-_RM~6=VM`?shAfHC@C6JWVS`n;Up_NB^Uw-*;sA}on*0PKce(=`sH9EN3@{NPP
zo2kylgin5*D5R5<tM8h#Z*Z$TpIIKBNKh8F#n?sD{_UEj1UCJ3W_w1qUDs4O5N(3W
zdCSg=SMM7klTKoSXf$R8!@w~<tsUg;=`??z*C06zkI|&!R++RAPijNhqcv=igX4(M
zxU<2)hbKg$rkf*etl;8sXH2@@FoZ`{0_MSbh?EsN%xc(rZuvTCqK@{%VXJTbv99Sq
zk>RE4ok8TVX`6c<-G5Dhs`vjLGJBVFbW?TCHb?W<c@>E5&HC+6uL*YvHr{n%4J{5b
z^)gYb<6}bE(Wvuz)DB|E1O?Z)UJjWjg}9yeBOHz|8J8C)Rnwyv(AO=*gA4*2Yj1sa
zZ~q<2)TiRs*-k;sL=P%WXF7AoAgnWSFTeO&R<HwH=j>6P$*P@+t6guc$kE?XZ$--S
z(*jIVrx2Yz!tK}Sde!)fG-AW4x}&_{!=H_P9ntb=G8TWe)4BQqdUwpK(;z~j*|FG&
zqH4Ckla_yYR8@t~jXD3%z)-GF1xZlYKS`$I%iF1cmlvZc_&J||V%6e=BU575(_E`#
z=_IZ%lU;3gW7dvy?s+l~ZvU#b@xyj<TY+&#@DoA~#3T&g*r;)PfrtsdR`+9IyeU>$
zdB2DbF4?{u=7J<Q7`a8oL4qi#>eVeiamg&EMBoywPKP@ldDE`~X(%w~!G26NKf5(j
zMxz<o-7{D!BWy}7KTO|3N|)E$tPNu^Gzw<9I5Pzx`2`0^zgEAo(TwcP?kY@%-Lgg=
z$|&g&h5gZ26wvQ&?*K1~Wh}-z`SH~I4{(WBo3mp+Oe^VHZDB+qR0+$?N^5vj2MZ)m
zWUkv-IKHozt{QJ>Yq(pdH^9B07iyKD3&#cKA1>jvhgika+4#}Cx?R9;Q0uk9c<oyf
zlh+OusGp%vcdk1q92{BZ?eAr+mpl9(-s9Fl`_r2-YL;y2(KZ}weC+FfqC9{4MgUv}
z*J&;%5~~Lr>j4{L?Q4VWfE59VWbuAm+0X`VFbu<k_K2@ymtGo3)`#MUJiKo2AGfp=
zdkqDDi+_~Bpj6B$(#HgYH*9hZL6qkaes$v2=2$S$R-u-w7ACy(cpV)+cU8T|qS2!P
zbo~!Q&)`%OqxsT578p-i<rDMo7RNz<VlHXD@)Li<%Dy@N=61Bfr&}|sSH2i1e2X7G
zJLYNQ^hfr9y<s2!ahg|sYY6}qOD=4Ecd4GSe|b_MvTQw!5dDu`*0;fsuVUJ>qyGxB
z?|)4DC+zvGMh^<irD8Asx^)1)CHDmKs*F_b;pyyrWyZ~YH2_j<uW<+c3#2VjY7_t+
zJ))FP*j82<lqzNeCnuF%yboQm$86IUM!5Gwk@33$Pu~o@5WQ0No)`H|Msa=1v#le`
z|CVDOfKd5nUQy#BMd3dd-iU^^*n_<Ep6;&C@kU?8CTH5sQnt2B4$m0YeOcZE9LZ7M
ztij3Cw5?%anIp}foO+%P9_ypcto&$McG=mx4kaq#3yE2Ee<yagjgRBgHn{8ao1&ys
z-g57K8CB)X<E;(RN%B{*=^69Qf=umO#;<v(u7;-s5bK#&V4@jHYxpD9xwEbkCAW-U
z{g?}xO1sd@yKxR&u!3uQ=1>t)TJ3dq%-W)E?o3g%FK<*>(T+6ehfY?G#V8fWo{q33
z*)zWNa3>Lt^g41Yp0L4_RkRpE?rQF))<`-kauD(9(VFwcCO&|s9LUpC)Tx6?Z<U2*
z3zk_EA~W-66JOS3&vy*fnw_4eTvDc|JWnus(B@-ax40L!zhi!ssK)KZ8d|ZLaqqPP
z?w^5*cONMDS0HwJNos$M7fIm^ybw1Q!0&r>-sISO!)))C^6!!2{jA^Sk5v^RuNNDJ
zdbGGzP!$zn&FRt9L?o6f^nYO0&jCox&gK*QTP`IR1kFk<-akN|>Nvp-H~5h4kj_D$
zMF_fUFdyfWg3;i$V}O4pXVa~@K0|fraq;_Av;C%O%DWMdljeNM*H?b3iIJq9I7H<;
z`^%%($r*QSZ>r>(+*x)1^&sNz*^iO^AvRqDU&ZX<@{U=*mroNl=F})*krXvi6Tdv4
z635HOZ!b9o#Xq(F{!de-uvK2c0YkKkU-BjCH)xg;;&?!^bDpBA6q~<sA>!}#Z=c&F
zRF2H`=1E`gk1YL-59$qnvD$#(e|E8Hsw{yhx2k&No?HJ`{rMg6?x%sm#&~P)x=Lc3
zF#xL58LA_@()qPvQNh=0fcm1R3u@5-3OXSd1LF3y)M7%wphnofL1lm+mfB=UgHI(E
zfgsdjz@sijoOTjg*#NDDt1LY(o@5IDZK6%)>STaxN!YhNGmJ(O<-;TGU(?_(zI^oQ
z?+dHEzIyapZK3J$m4diG$DyQ$w8g_Gc^1Eids{JtUb@Q|!(GM8dL)7-)v>*N!H?TR
zYwBvxy%w!W8g&{T=qVi?GvIW<4*G3gnL3=?+`JX!zJ84Ka4eq~V6hlJI81k4#x8Z2
z6Qf8K`7|{^!R(>XM3K!y-<C`8_4SxA)2{6B3jYVqW5A@1JhJb#XV@Ln@9xX%dQRKE
zNQ!G_E3aM{v9`7-8I|ShwPnaib>fn|sjJVcLTSn*^paXBU0mYG3=lrE<V!3+LIOTn
zWFKzCMMVjBPh{_`EK{j{B`Nfw`JrgH8L2hVSFvXEN8ELXzeW!lp^D$X{b%z6RcT25
z+FxyLB{rcRoTSRPHZKlUx{KuXRr(UMuYaeqSk@?ecUNLMHh{>VlwwcEGK-8x5aH{|
z&IT=Z4byelg7Lgoe{j?=)QK$t7q@~%MY04oKl6^QM!~HC5g{!_CY1!zbmgt!bi6Z<
z9?e`^;0mLV!$?P@C&+0G+;8J_I)qjDDwe@<Ay&;g{n{OZ4JKFAr`TUAkd5zkbacLx
z>vH2D$Ja$AX%L^|{N>fNLwSQT^}G1&=wpTN1Djk;UDO~a)vnQ10Tv@$N9K;$kT4E-
z3z?i~g#^LPNlKc<-T}|!+fb0Dnlu!!j}@t%5t#LJI>zARRHEh*&WwR4DOD`R;nTbc
z^77;snW!U0UfYTJ8T!2e;xm<HMQp{<H>nJbv{6FF>sNCZ{U(i7w8Ah-%8r?4Yg#7W
znKIw{IP!UY!?e!vw9_%LN7_;n5X<Y204*+m$hoY)Uix{RZj6lYt%(XkU}Lw7n<FGx
zwX~ff#mMky#lUhey_J$MzvOY!{^Ph8K(1B<ZO+ef)0QXyDQex>H~=793WLWig|6dA
zt&eisL+H}FmpsFc&-D!TEL@xKBm=ux6f<SXS?#OXhb`s3>e&k@%f10RfJUS00ECt8
zfV~*+3E7g9m|LmMDJbD&;#=$Z>YrPltZj>0+mbwsQ{I@Gw;z{GeiiH9Jx!iXOq9=<
zu`FAi-v25#x3qWaTs<L5FlR5%7J7HqKB=T86Fx5?{3DVSYA0(is)Rf#qz9pY%2kQa
z)b;~Cn!rs?hheL!w(<5_WnUE4{xnf2iSb!EFuWz4YKQWvU&S8F9Zt^Ud8q`-oY7JR
zJ~dN0^%-aGbwg2T9KT7l0he0yXn-nimZvpfY04CoMmav8SDuQ`UltWSB!Bnmb4uf#
z2>I*>_uD?#G6~X4EemNs2ig6^iBbB+Ze6C(=;On2V|Vd77bqFl>3{PM`diOSQyp|t
z>?$)%uar0-_;+V78NjhKGTwUQOOhNtVEseM@uY<Q=GGL|rP9w`+}<84O;8jryvniQ
z>+%~l=D!S^S;bjJr_O22$uK`{$1?Vh<=a^Ad>`zW#0+hVj#?|P_a~gDl+?|x!JmEG
zOgM<IY0@@v@6JjnSvhm}yqvP5%0bSJahIG;q6|Z(WF3vQ2TKPa&$2Fbzqsw)`coh+
zLo3*y<6tyi@W-^Pd{ZikYS52#vO$-NR7*ldb7p0QW|wJc978E>P?xpylkoqd&Kr<U
z(<D_woQUOwob5&9i&q|(ATQfDoP%B>I&qENpDePz0hgn-(xX&qnNx(;C)Wy{q6{tk
zu5HUcuZttD7Ke1S{E0Vmr9IDjyONzt@@P1T#v8uVVM$Jsm^G-%_$4J(O(JdyN69qZ
z9RSIbuHmTSSyWC2c{p5Yv{b4tq<=sfLn=%qo#=sB@F+))xQia}E)p{;paPTAKP`j?
z#OB-e%h;o5Ed2x%)78;iJGUA=OsuR~t5Y?v*jdh~(eSiA*Wgd@5t4TiPhYydu;~x-
znwu`x7F0o69Y7wRu;WAl6#=c%9P&m^&f#KLyGTaMEqhBzh1=uu-vsYJYy$DvHT9nM
z9DbEYv*h&1`hb}mF^Jl5UDCnK20Inrm8ix4%l8;|UV(xR9(BAk>K4D`JU8S1o0e?-
z6Y(<UvH|n$&mYi;bjJ$J-3n`rUSe!#{JO)F*WuOu@HN7&=^jc+BaIP1<w5hw7Xa*S
zU}QNkEz|T-U5iFTsA5OKYi%2;HAJN=9oFC9DYZ&a&kShkN5E_h9OBZJo8{I}SB*Uo
zKOcmG?Q*_xjk-41gz^h1Xq4`b$d*>JczLIfUZUe!R9ets2lm)}I?Z26qm4c*1#)^U
z2hZpCDajx`i<x6sf>Bto9cOqhH=0Mq>8#3Bx;T`$0Iye(nUo_3h=14;a@)g#sfYQ)
z*db4>eSOQ0eU(4L_LnB!bzA5Ms$G$UMOu=pIG|11e==VaS*g429kSS&DjX&~^uiqg
zS2L_8u4y@~n7rJV$>!Eqc`jG@cP&g^n`u6*7zr0d0ujzMnlC9JGqF4~PIT;>_IB9w
zZ%i>a*4Sa>L_SD50sO4VhVQHVGdVp5J_Y#{*3BO1)}M^ZwSeLGOrJhR1PL5;rKF~h
zqDOgK#gPTW;p&XPCYw83((^eHKAgk(L#{VnoZFJ}Ui3jH#4~=0yZO?<6cI|vzzS&8
z*MZ&T>GVprC1+|btt0?gP%r+Vqn~&?PBa#>djxB*sk$71-r4+*$YSCJ!}5(h09YUV
zDz*nI*b!v4eoTQG&m4Q7bj&4tpFK!heq_|;?_x~)?w!ZbP1wx^<K)?bUR_6z_ex_V
z{$as4jrFh}s@@W^E|l9P65st<cJXI;;~z?QfBd|ueC6j_`l+Pcz_s|3YxO%$W9CMC
zpelrciwnqU%O;+`VG&{0m)BKT(cd+Gkh{gnB~JBFaXzX1_RuqHyTkGonL))GnZqaf
z+xL(9fM(Kd!)K5G%mGJDP~cWJ-LEKowk_;X)q^waqI><N6ZEr%cQh~`6xDvE^+c-K
zK@ON1Nm@!-7?L|O-AgRsp7a5uC;`Zsts%NhY-;go*rGr%vAS<snOjltL<CtZD)}n*
z?v7wnI`!1zb{@<7NU|{@Mc9|l6_nH4KhJFjoc63+p5)cGr;k30iaar|0OQNh>;X4@
z$Ll+`y>$F_MJ0i<j$bnCRYdHv+<t<Os;PZd>5(NPWr)t=Ztfek$j%=(w*&i22BG?N
zFJMg+ws#a~Ra`YEIW|eHkY;?l+T8yX74xb){IGX2xWKv?ek`<s+-#bA-D9P1t9<1`
z+S|rI%k!+Bd=(>x4Coh0!%4vrtKc<TQXH@^Y*lg0!R$`GnYjG<mYCu=e1YQqv4-m$
z;W;AUOSevj>&ZochLacdI8HlG2MHK~d1<DEX$KRrmtg(nN&e!eK~&jHV8%iM_pDi7
zK*GOO&CB#{6ZaTVMN*Zw6khaU!&lm)OiP7+E1jp8U5hNE_DkRT?y)x;p5L#ta*z^C
zl`A{fS<o_3#<Xds60<_&T3RykCA$ULdYv2eA>-axvEZpzEmdj*%m&#P<;#mfQ<zQ5
zxS(c_VF@lpkQlRg{sbLoshCre(=!SBDkiaha3BzhcGlj@vD%lH+c~p+%>g3&FA5vc
zpFCTc6Xj6uX~cg-=U2SO>soFvAOAT$-F`U30qV_%MYAQ17U!#dzKU^4qZ?nvSks2|
zuVPj1r<S4U{pL)C2>ST{N{|6CEO~QYws)qVP{h3d`d@j(|HnUi>$v&QNpUkPg))9#
z?tkam-z&5Qh73Lx>^^TkUp%#vMb86*b~gup6J1_O9{=IuDzGe-L|J$x1Cpr5Z@_`k
zm5j23fT&(xYv%NgeCaKB3ydAS`BS=I?|GtKu1C4zii$Pb34LW!Nrm~OR6MrtOK!;(
z(ZIFW7<v|sCyhm|Frrv%!YMlZdaN`_3G+OWuRTFnD%LDlmX%uQAI35@``Rw<zsdgc
zbfFm^&Y5bJaIUddsC$B?X;HLYkg;)%KoGj;J#CR6h&v!vCKfg)LSxivna}EhI>>Ma
zl&yB-Z{(S`J`<)FiaN@K;f{J8yIvU6)L(nyLD|S|Zs6;V*M+=WBeoVp$_CR89g3If
zNs?psRdo&1Qi9nx2nS@r=p2oToW=1;!Ap+G`EwWocASz)P%2h#<NIS|)KcT4`u3<F
z#=w=6N+k%%c&=1cM<>va@bPIj)E<&Rde=k;rBP6(R`KU9<_dD;>cS_bauIzmXn08H
zRA0j33NNxy<qfkh{p>rBHSe1`c30#B{>s&#58`RK=v48v=dpE<t<TsWjf4qotj1zs
zUF_Y$J`NBDF>om;_wn_2oGHeTI=PVbC?vdBi|CVI|8t_s=H53-l?A4uQi2H1)nR;$
z7Niq6+UVI^cpi-g)UzZ#h|G#PRa`|eDoUonl@AH3IGA~6mm`DvE=GA6Qefv)-xn%b
z63MrPf5&o~EAsg^#;MF;^oqx?uIS4~#SIR{F=J-tBPTaRm!`HHE?*J0&c5}v&uHFK
z>X`SKx;7cYC@`Uw7^>rR7Cq@59xHqjgT`pE+J%y9AWNS=9Ybc~5~S}9G-Ne4m;wFB
zL_Rfr1TDFQ?3^+qlPW0Flpp~V*%29L<LM}YWNj7t?9(cnnqi-qXONl*(qTxE<ExR9
zJL%MP|L?kP)=0{K+n5ToNDaIV2aQMszd6ft_l;c#IC-HizQGvAQM5Wqdzu|cQoDR)
zS!cLA#f{m*dJ!;$QdJ?Y7DdgS7PiJ{(W<IO&CJIWn(aGTqmkx@Zmc@_`v+{0eSkT_
zV4xLoZt*!Ykk)kCIM9L%MuHJgXJ*JS)S_CPZ6^J%Rj)3+{is^GkUPooqq`NoL3oWj
z#xJMimL@FHWC-Yx`lPFb7bp)Fn|W+K!=2~2O%c|;N`0n*(R&$mL;5rLpb47x^Op8R
zUHDpr9+X?}(Jb!7h9^2)f>(U$>K>l~z?mN(?Q-`5384W<<XT~GRTLEIXpGMb^`fIt
z`GhJr2DX1d9Dw(*`}ymOQG@TO<-9>{4aQ5-mHXYn^Cr1A>56Z2oA39f@owz--DFmv
ztOG;Js@jr#+RFXR<&R~C0QlwHnMD>=TJ{R2^)KFp6shoeVJ3l0uka$$2$hp8N)qZ8
z3>u@0fDbB7C66}7Lt}@^PiaG&dm(;^yj(mS5+$%xSVLx|X8}kxFxOxh1V3V2Wl{R(
zQ0dV@lv#_$lz?9kf!yx2!-DAil<eQb<vACD{^WxF$X79^nt5GX>CtStZeX;G!$~gN
za^tD=i7NXG$8rr@`c+J?@2gnpR7i^VS{~DDsAd-sdfGRqbRDZ&RHu9>Ns((=1-2IH
zt+d0FSq1Z{%^rdQ&GmHJ!0y~=>(6o3(EW=HHz@7p_C$3zwFo_)ca1}w&}lFjYpCx2
z{U&#*US<6qgAlJ46e>?oXYO8@R#d~Mk3>!8&jr!rGezr;{b3x~7JzqmO0vtsU@v6U
zRn1|>k0VeOq&3(UG&)ESj@vJe$87Z5BwLt~3T{nTe_wULa4%kmScf)xMp&J38<4HG
z<sEOo`gF0LuCulhsvZtD<dEmal`d%czau#s%-?+TTKK`PMDo}yooSzajoFn8?=4s8
ziegqnQ}G*+s?H<nHK^C5(qeSnQebLvMd49}&C`VS5Dek%0?#ei$>x*pLl-I1aYvYV
zbVW__0Ixi2qMDfA0)s|Hr)D|4*t#`N8{Ey({&Y!pe-uiVsVlHbCbC_&e+ZwzzJ2?1
zh~W#j#Vd0zrooh_-tP4y`^P^>W=!Zmdt1+@MBmOGZlEux$t)>&ZMQ?8UvPglW?hmY
zE~U-pNQ4t^L1UDrrhqdrLt4j^l14yN%SrjTJ=NZE84E8RcAe+i6i$=aa-6xh?$XgI
ziHt|n4V7l~_-pfAfOrvxbn4h~nzR2w-~aBcncA{ld?L@pmt0$J|FYHtU+2g8?U)f-
zogF!ob7Raq%FCdsbOAc4FP#}HsxdRJ10u5p_YX;V7=W{{K}<Jx>aa>wQc}+HOO(=m
z?6U43%4HZb3dVK?wu1k|+<QhfnYC@hPSYutK@djj2ue{2LzjN50YsXV5D@|+C3F~C
zC?PO2Hj02WApsJl1PBnMh9s0x0ucyA2mv7!r6dvvolpY!Uf%b!p0%EDy+6J`@B4lE
z!%E;{?`!XKUpdQh9!IndZ2&SBBVUC@HBQ6h1dPKPQOzFio0ga}HG7E!?marBvf%kE
zzpQ8#TPMkdNdOIFSjiB~`r3`%2>si(JrCBtsf{m;M$5($?r(s-(iOJP(kJZZ*Z=OB
z9^Ps<9L(`J7w>Tn=v*y~>Vcb{yZgK&P27D3KcKJfz2#sVqe&pddsSih1%$9#`><Uo
z(qp7qRzxW`hzfMg<9#rX9=vqW7ePE!j`yNX6Vmc2ON7EW36`9bjk%^sQ^9R#hEF0D
z&;9UPv;U;{zqT;t2QjznsbHP%^uh+nIXyJ$sm`y1QV1}yvZlH=dhy3c#~W+418w$_
zUF_F#ygfO;07M2=lWF@cjS%gec&|Dk1y0#l!vKBukVzteuvkH)P&dtW>gg$>!NIw{
zfBs?B4isx!;VbDRak&*6Crfu4BDEw{Ej0i`z^El}5D0sm;K-EIy(FiSUo0MnYtT21
zBORLtjc*4BxoaBocp8enEkHV8Blsh&VUy7FrahUS`-W0e=m6^dso^n8;~FPiv%_w9
z8Y{nY_3)-IJ1xe3kVvU!Mf#`K<7WVhV7YS;Fhs)Ei62;xnBuE)r=lVD!ny+#vngi!
z(Srxn%3c!gk-5YWF2$LZSXgXw4roKt?G|@;(nb`vR5!b?!^CVDQ4_9-m20<#5)|&J
zHN`#seC&^jt{kuP$MqlFosT{XTC7i-#6FLfIQiF;w#A-xe|}kxNwp~-_hU~V`E4)y
zPRrH18GNLznbUgEA4`TN=YfROmMv(s;z5jFy5W>wsE_yi{^t0F>b&jwg(!D?N(sfE
z9LVmwjL|T6ArSK2it)RiFBb1L$9mecw`C}ZkQyWO_Q1qd4{CbEl$1JaC_Q~FlmJLu
z5U2W|^YIvgt-cz*vR>yVa^gFDf~o#@&9V&}LQq5C!b5T_#Kt|aFT}XVJu!tw@ZCIB
zjEbk|Haulr2rvz>O5$CJS_x?;aAQeL%}bI32>#<elC3Du^GlPK&H|WruR4`KH|jta
zo9PrA4+Yem@Uy#~<SjS$)YAiMGaEjo7g!j1j-9TljWF}VqF}I!eo}jNA!tnYr%-C~
zux-G|USstu<Kx1R7Aw7m7IWI}J6K6mopo0A{g;lubrkq-792I*vvJk`<Vev|4B;;I
z4bg}C$JNMrdFg@bPO$J$459RjyZ{EY)#Ap19xC*zv88E{QP-iRjMdb>{7ZliGCeaV
zW)xyB9zpQp>WZAVSoXS*0mzUn*iw!g`$)f>Sa5ufbiI0Fyqs0x`n&XBNLWC+T~ke#
zttt3g!IZ91UyiG;TArA%9>odBabHqvDu2Vxdxkr;HCe3W6LM&zr$)+4ON&n%ae&s~
zG5mpE@&qLbUN=qvWby>Ye}o<0RB)b_<|e8@V#f{is+%EE`xLwyjZQKig2pF(=>QSN
ztoB1GpO5*p$+ed+3u8@-4(pM^FS_5jnT72*WW|R@s>pkMrhjE~|KM<a{r2g(9qJ3r
z>-wn{0q%QBJGxq0XWgK9)UIc_n(*?HE#rpWsgcKhr?Sw!K@D0@-;&WAp9)@!^nnL^
zQVOU4ZZ9eD>B)V#Cea~2EsxEHkTa^oO`L@(Je-EI*HqyXTjttQz~bF85VY`-bNv40
zW{3osW&0Lem}hD0ZX+^C+^_CX|9UVs^i#*+`Vy75-EX^55O*A6uO!`&W+ho)c(pIE
zI4#R?++`rLvX|ieyug(_96hqZ?J8>BQg0;8?)J7ei$?BM*!I0?w=Ha%2AFVk-?Kg?
z7kY9k+RsG-5S2}6ZK%4o${W--HydFS%_oDY5Ew}>2tOI+F-RbSn9KHZQR_57=sVVR
zQk(Rs1gSM<)NVh9OvAWa&`&h<cw3j!dGp!q5WV2@niKBQJ)`^5P7WU2{p3QcEHSin
zG9pno<k0dD!BclzP+Clw8(ZCjP?#UkC>E*?0}47Fihk$}FC$=6zTak?`YGIJAyC(y
z1m_kb*t%{c$RdWjKq*ni7!+hEB88bP_f(4KA$y>Z{P}1*^*+urW`j!(Z3?Ee>LDVy
z=5FcNz%503^1&m+M^(p5Z<EiTKUrEfUaerws$V%8=)VTEv4ZWL%@NBE<__X=gbeSd
z&1K#sUAlN8jBl_bpzC-rda0psL%2UVo>Ilbt{dBn%kK$eTvss{Uv0%m_;3W*`UWmv
ztbE3|{ZR)kr3JloJ|8q}Iz1$uB$;z!)S)KM?j?jP6)CH>&I7uZM<*GUYYESn0OE9a
zykQ{89b)TF55Rj>Vy8qlkx+$i=K@HQs8>OcA3#FJQjWfPW>4vwwMen7{iXJ5b<})6
z@jhX{roP<iMeVxH)ce5j^Iw-VY=2jn{uULisQVyyDB9BTy-}yviuFWC4+B3?lv<6Z
zwL3d`QO^M-q09(Sq@R)uL2Wx~45Gixra5`SB&Mq2I)|&xB<x>7R?~NzVN@@SqE;=K
zov7g;2KFTB&5gc)i-A{?fhPM<x-lK#qs2|9ImCYVDXPGgbskZ^)LeQwbg=Np&PU$4
zHw|hoIVuA1$Xa@M>lwX@XV0h8vJCX<yHzp8asXq~Ic6c*7y#x$|B?vU*OgKXf%W}U
z`e}!Y;o%9j9@MNH!1LOulT@EdUNsMgp8ZJUO#l38M{R)6c6hHkQc8W!530WmW=&&4
zT}giOQ&EvT3-BZfDh$FC{8C2y<kdX>jv31dbF*ve;}%*+zL&U`;ZSx4Uvg<<Ya;VZ
z_bcxE)B@Fr7P*J#V^MIZ+t=a)7Kc`;3+*G@534)xtj%aRtu<P4#^c0epu?;1KnY*A
z6aljl5xKb$1c(xKPn1yOA_3K%b5Q3<cT49tqi@!~k+NS`avG}-KlI3}9H+gmmR24y
zzqY%3?e0joi7Y1t-*%UX@h>P*NmYgyJb&lPr+-z`rIpp%U$f&F<iqF;yVy2#&Y7bR
zGZfw|2SuLoKJ|_ob?CPFoTNg3kT@A|db`-VD!I9#TC1mXPERsl1~E`<S_kp|gouuB
zE$Vj($b^DfiDBkAH*+zTdGs6`qRA6?2NU}z6dsOa((i3^#<%t-Gs``O=c+&DYuId$
ziNmiBEH(pNTQD6Tb6p!@K}q~Q^R~fX9e3(mjiFAgo4$|zxtMks1pl5*P$F&)K3-9r
z@EP|l)x3iOFK6AU=6F?|c?-U!^D6%MBwFj4mfvHevs@NOx|X`#Jz#&A6j9Ayhe#X^
zns5;6PVAWy3=w$fJJ}3C*xCb{)(WvH?dqYZZafs8TDv#_=~HXA8PJ7@pn@?#O~aV(
z&I-k<;RBOX1CnQ}s&aQWJEtyBDpLhJZY}gk-$?x%zfIo0K+Ty$+!cjye{+)IxA0Dh
zaCO6XuX!;#x-SA8eTj6IRA`-Y&!)e2t<j(NyTIgQNsx|1pl;Ej1;adT1jg@SutK~b
z&bG$gx;6vu=1xq1tT8Sr#jWCuepE<^_iX|dNBHBFhCQ(vM%0+)>uXhITm}THk~6k;
zu(4bjWHqwx!&FD}+`aAA8tR=br0e>m#c|!-*;iJdpo2ZNIa;sVf}e9uP^b4VC5V2y
z1Z9>%ki)J;WxbA5H*cjs)sgW~iAA}DX1@FofPSCf5Ja}H@Gz?>gs{azL!F#G>122i
z&}B-ELHTOxm^;(OO<bk!kv~rFSDb;2YU|1x7hc<PBlf6Ir-lMpY!;X{-Wj?Kq<#Um
z<{_wU<T-mlEGCVZV0A5a0r~UDAy$xDc<uEHsAj$F_bsXS=@Uxx_4Pn@BvK}UNLPBk
zT-LDNY<{YY%V|Q9Q4!}K+Wc1D(1)BVZi#815pROA$aSs^LWmzUFUywW!e#46!EQ>j
z(_GX_l0T7P6RIgBOB>bz%8AxrKIQhnZ(?l~%0VL7llQz<&Rr7*k|J8Ekajl!i?&4K
z6FROC)1t>TVK8@^Q@6I}PnK|_?e{&|%VT$1$|=hajOt<>8-Eh1t^IM}veN6D?slDC
z(o-zILAhE`e@RoTeCHJD9`eQW>`<*)*-r|-v!Q=c`0FQbPl_T&)njx7M&cvTJVY`+
zB6zBwKnZGeW&xY5Ut*A2wTZo)k^zf(63-n^Pp=XTByu4Ve!Tg8zNTP|S=k;WsfqzY
zVP_T~qv)5C+D2t*8f$qJg+6!y@2Q~HQpEE3@DI^d&MqedM)gg~U2V#7ev;-sc<H#J
zB)RGyF0=Mk9^hhrbdn3LS+M-`v0N_yqTBwoWtz->w@i|A538viv|?Vx*gr?9+;?^(
zrnSuSAP0r{l_bQJls^fs&eB)J5N#c7O-zJ2eqHKW@$E5oow!QA-ty#Adpy7(WwD(*
z7D_IyP>`Rh&f^fN(NMs#;&TZhfc{MPopJ((iV4D&!6#l<+P!L{J7}HXQ}5CoI3e9y
zuBHwd>JC@e3GJd+-6fVJi_&vedx!zN0ekiCE53-96z@=fX^%9o*wPcyg^?3+gas#k
z4XE=b4d_mZ9&-kG%@|G-ib@_k^7Tm`quX*G2a%Eb%AuCAvq0ET!j=%GES5~^u*5ab
zh08J%;lc8q<MtE}K;hWV<K3zV$z#qf%zWhW<X4UoL-RYSfPnNeu6hOcH)8j!0gF9=
z5bE&c&7_E7-jkx605<CLv1NJAD)SK14~WeyRnV7kRk*gV09krNfZ@=L57(!5imrTE
zQ~>J8Z_s9Y?SL-VsV@+zOteOnLT>x;!Jit%S|ir&S4P}r9P6u8#DDz<FrNnOP1HBx
zg+E?V9o2lEfw7fS41bs%4-QZQ`Mn2!)12-*y|IyA|Dmw1OLaZnePY_lO-5?+d8Uy?
zwQ>LX^0%A_%Kq!><ehI{6CUm4{Wyq;7E^Rr`K3@gA=)tSSA*4Asd9JiaE9XXir}%k
zzHSS9sBLY((mNrrHUjPP!W*3FbP$tceudw<?l^f_xKvXkm~Vz}mPNYsr@&PMrfE%1
zb|e*woZgU}edk-vwiwf#fP39|t>nwEY};;EA>PfY_kIdb9^&eIDjW3jFRGJ47K-=5
z<<d7P=~c`0f9b&<+jmE~okx%nD<yDO^t`rP3!|aJtB`&wW(>ys(eOUx)yDW=uF&Je
z62()CzTdfMUsbU_?Q-p7*ctiB(;Y*LE+>AUBA$Es^d6+ICGFb6D2Iz6HpOIL=_Tc#
z9RHgZ+Z_1pvF+I8(@yrt@STaTe~$lr%-3`g`2~w5`d!<j04Q=bEF-P=KVVVi<DZYk
z1d(2E&Fh~SqGxUdd2*9>_bkJbMBm6P?b7pc&4g*w&7(X136|J&&!iX;03cgGoI%(w
z^*1(kfZi+%O*On<1p_hlOlO-2h4V~2Y{v$lnmN)~5y&&c*c?()Shj^;cjwe4T2lmK
zfYkff3=bKQ7XaG)dL-fZ8FWE_Tw8Ybl5J1wF!WM*5H3#_qS3W9Xrwjf3Tp6mYmhen
zs!fRKEL{mC;rn|lq#VdLPWfH}&1-R^e2nr!QKo7#S;s-Q8)iJ9iAZY#n3vGhfRzC$
z2&cX&CR8`>nk3gxdYiD@*F{wHVsqr9Z8%E=Rg5B0kx)ApxIixO4Xi%>J;l>!@WuNe
zZtar0;Z9+F#c<_r_0asiTYmgGwnwLQhjzA7+u{WcF{<-$^1vA~Tl($(LHozq>GHH|
zzwFy(mU*<>Opq%lIw73r`&3&-P&5>#<DPuPV0aX4CUccCx?}=lJuBudLJpl=*jFb2
z>8F3lUg{2>_OT_0ef}kcO7zhf(m|79AMg>+>6i|6Y7cSozT-PL7;seRrQ@*HuE6p}
zU9r<mF^{QrUleHfT@EE;tw9XEpUWC0WxrRA8}-uiP1^NtNa|JBS6+ei)$!wtS3L`J
z+`}UT5F55>_!vuEH>%q`Qd^f{3A0w9!Dfz}10~hxIQX983b*ZRv0DsV$tpnj8KlNv
zOy*;}yc89a(Vrf6_#lurX>X9;KktNR@j(hHZrt++Hqa|9KDQS5j&$qORbS{t-D{1Q
z%ySe(_`CwY)@rhHvetLZzC#<Ae!Rdk_Lzy2fe+$xvVNI}^7ZD6N2F?aAO2*GRre?%
z1T;o~qS)+647{KqHO`0gi%gQk8p(HVMP__g=%;k2TKec)2N<I6P}?D;u!VppQ$zjv
z8JKGY1h4oK>Fge%@Oq>B2?tfgSKgI|OtPFy4?vu`r+wS^&d_3_3H&kp@^I~w`>!Tb
z#4akDG{vgxYf#mzU*tV}USer|8sT~mzPqmE%+e&8*U0JLU~LE%1`8|3AiKR(o2~8n
zX$(PNOD4`34@q!{oltWKyZ=Ma2qNamKB6=!6gI#dZLF{EPQR&_8f<4mDXp%MAG3)e
zwBmWPP<BKroKlkNrE(ilNL?nu?^Q`TT~W=e`~_8O+;$VPrZVL}Iq<4bC(F)1|I%r_
zg@=~O6~CdC@fXP~xRu)JD)XE+%kid_;%{W6q{;Q@@+nf*Ux{q(b90l;U<dG!#Co@;
z1kJ`l&Y2$X*Q7c9da603HsZ*|Tns7YK=LpR0qQtuTQFh};R>e28w-a6k7Gat^YRfZ
zc)4N?JmTg`c|1r80%iSEv{AF7lP54UU@a+IAG$ywr*iT(PC1mCR{YE~mkhXJsQz;2
zhl{230%}FKt<l+{pC&*vBdg;TY_}1jk8KLswsRcMHBT<!&a=eW+7MzDmsr4S-p*<{
z&>QTl)*G#_`~5Bzpj!SbOkvDi+*1Gv?LI;;mM9}FZQtX~{BYywKzv<b=(`$Sv*V2u
ziJFDp6AljyYzjFC3wiRE=(|uilDFAKXH)-sXDijq@m{*e$rtha>Rq>U16?d8vMAau
z!sf_G4;r}=ZUMPA`iCI?sXG$lHD)fUdWcn4q*UzpRa3Q@@%vtiSJelO>^?3{*vcf1
z!hZQBC*44*VI_XDa7XPep0WxNc{XYT$v%mpP(|-%v8y>WrKor-sx2xEqT?|-JagLO
zyT`MHyhtm(zx{B(Y)6!gH%>}^4}RcF8nFsFg1OO97td$?vYYVhP~?_eb<9r=C24Wz
z0{8YL0M0D^fbAIgj5*9XQcRmQZM>UVCETs5xgUhYJZ1d2ME24tk`dQehlJL-hJtJz
zSf1kab6~JY6c@2WK~YiXX+p7T(On9u)$eZfH)i<0bpqbsuhU);+gi9%Q8wL@X*6q$
ztxZ5jgJ;PAFvQ2y#5;`dDt}cVh+s?d?7CT@uI-4>A=_|wR>ZY5CzEJ!q!@?4fx72H
z{>!^dF4B@n&BEnWa;SlmZUldfB0)KxNMB}Y%L;L-o0dEv?Vr};A<y*gUJx1?-~0_B
zWB~Q!M$XJ?X4<(OAs7W~JH*cTg;5euy%<N3KgdK(9z9!vR}L1t{TXn${Rj^vtQvrX
z)!KqL338IC>vNId++4)@rBX*L$I;EP$kgh3ni4JDNX2Db=7HUtMRHz#THqAsN*x<;
z2)W+w0F(1!!6D`{_Go%~8v6J(lRmuj_z@1xw$Rgq^sTVXH6tohx;sOa)mhrwCJ-kv
zdShpOlv}8gjml~VCn~(ACcjLZbiudAL2%ANU6TKLfv2T~fD_VhsVx>ogD$Y4(#V$}
z^eZ20CAsKgB|YQI%_UI|`)~Ao+YLkZ`%s?4-Zpp|YssvhwPX~%Ne@LaDcq`*!%MJH
z!|MH$D@){c&+6*BPn%Ux1g70nNTC=nCiBp9Omw{5a?mI?)H>1-S1S!QMLj=(nV=P3
z*BfrI)!8Yf*E~7JHghvj%Q<hZX!P(FK=c95le)jD;Q9C-+$FQ2Da6cBm6YQ(H4!>w
z9v)GEZg&Tqs^kJJYhBop^z-)B5Vx4X@m(ia><-bbFd7QxR11T@q=mwr6WOLSPm)xh
zcIK_Dn%A{P1|+__LIlZ<E-OJ-8b$jtsw&O5dh~sbK>3p3ghwHH!Bxk5uHOwz9v+xi
zr=Rs5$V-d-=3gTrvAx8g?V(Ve$>K9I^6VKmn@NJ;7fdvzB*aqCZGs53)RuR*Z@DzX
zT`G_=RvFs597r~Nlny^K?@uIDl57bLjVn%?h$OGTPgL-_IFEr86O858e?B%WGNm`J
z<2Vd4`au{K*FPibFsXOm1d2$jbDKK=I_(LPZny1ZIt07kr5j^io4{x42scb`g89$O
zXhnF(Iqj<8cMq&OA<-UN7|!>6?VCeowO4NRW9%lUkVx^!PiTxC#00TAqzy)T&4ag~
zV1RTeig7gH(O05^eYYwcrRI=ypTeA^;h1H1cci=%m=W#d4BmtQN{qIqPB!b@+^799
zMe6$Hnos8<ocD1}wXffoRx!9An@{_+hFZaM^&U{4^MAc{e<lB4DQDen!DBzW<9BWk
z+n+pB26Q)4e?$*$jsG(i1+=)D(?|H#GD^}jaoaAT(L<W{ao7wE2XQm@fTs3IW)&d(
zf}E&iM=lNwMZ2vZ^>$V`*;$yH52|It;%96fbS)JojtJyg!+nwU<f7!tu4}NFR*Z9#
znbue_Bk&Lfsa4kvHjY=;PtDQ4Gx?V$Z{F@<zqL|GLdyJI3jPkf%sdVy?~cq)Fke4$
zyjaTd^-xo~;ldKMI!#Ig?#qT@W(kVQsTDvUJ=ZIQ4sEDR`Pl!$;tha83VeT_(o$XN
z)*Rn0o;X!w<B#7Kw~g$MqC2nCBBhiA1R`;Dzj681riw6Xm&HQ!OHr`*w^;5Adn^Wn
zMqf&gVoZ_i{`L2rAdG?I4`^|tV&1#IZ&)eMrd&*T3|`j{^W$UX#)q1A@j@7SSao(G
zYCNSva?*`#0yzSNGsWq~!Z`|UZ58&06*D#EW*FF<7O#7oOfrWOQr;4SN(}uo{05Qh
z+v|ci6X9e;lxGy+d-YV8`Q(mBmj+?A9dnDUNO!yQCKv0>Z-#C`<lR%z_$IBz{2P|a
z#S7aOuoW_i7-T9x_!rE<&Ng0B(oEANYADBBk^!c%WFfY?Lv&4^7y~g21Emu=!!ZMf
z3u3z7L$z?v!@6<I`szGvlj{{g^e3}v!jxH+367wb@Au=WU?k8O_4R12TLruJ`h%G0
zO#g!-HSUGxs?ldIgk2X&yEm?{OJK9;m4&mT!%Exuk@*+Hv*FWyqsHbevpa&C0gYQ^
z{H)LH@6#iRdCopZ#IzyvUOe6N+`JP!a1o733d*QX*tD4`?7aOx{^63_6LCkjKeMks
z2krN@ruQcR;6z|U9L{-S%;2=C)c9HDxD2mCtYbZTS~*3(3(@6wn@mv;&#O5V_>Io=
zljYOtcl$1mm{$GxHl?A~COh;b*5MQ^eCw=xN6nL@BYVCXOIPMW$z(v*3wHVvDnvVS
z3`pOvOU5nD5xhHPEzN?;4r2fs1aPxI461UcI8aM*hEUDkuWS#j!nx7GHcsLkY@FbG
zzDhqDk>B{iObutstk49y@{&wf&E}Y0%b7_hon%F=JA3w|k7k}6P2QBEh!4zlCFZLM
ziz{XM`g-}&Ka@iDV=9+9-a?6||MV$3&gL;KiCG3sq#%iMhr4BX@S;{(%f#C~Lu_p1
zo2kx9{|WqId!<9@lr;7OIWI*AhF>0o*F4>yBdB5hu^=a7`$PZ`Jr1RzBENBAA#7`;
zSKP@J{x0hcCDdrQ)X?sym*Hxd($u-aXaN4-mr$O)WgScXjcr2k=IBlzj@~bC+^W8s
zpR))e%$*CJxRLMbL;a5MZH1V8ptsSfLWAhY54U|QH)S%Nv~+TNpSBDIt=MWH1ZF?$
zo*HWo_Tt?et#7wONctuc0VV_jpMW8k#Fq##L4b%Sp%O@!FJdk)L;yMWpcPckgBE@D
zgQ6=F7M8@Wep)G!05=H{vvtsc$e9wChal&`<_?kos}Ui^7xv9e%#XE|?*m;l%D<~Q
zuRWeo=u<#TkL#<vI$zSaKP-_9*-e+!bvvu|`>5FkT3=2O=rZM7aPDAqM@Yqs$%&+7
z(XdWiF?2inX@V_OdZAtA#{m35Yg+Drt5hS1?L6I1_a}uC6_w%DQOvkiJt5-tglEnA
z7IRdpZXtlIw-Oi^4uOV7g1hz9z3$<05Wnga!a~%5JI&Ia?$1h8pjHh{Rj!1$2aHy`
zrM8>esTU1VLD|_Mf~>*b+30iW%9O9Aeoa|8n?ko$er!-~hd5b2%T1{aD!e?3t8g%9
zP@?6uwpbAYy~<S0Md~>w@J(+FwBcl$s$@)Rc0Sr4J)KTn@$(ytnAW{(_^6UFMIl0U
zrwLQei37&#+7fJS<8V3h>Nt9${X=7x=iR{e>5toer`)Tn@`ae2zK^cC%96EsM!Ni#
z+LzaFX57r)e4J6_Um4dr7<K!eoV|xjRA}e9-`1`!y6C`;SC7_*%ubAa>Z`*~=`FJY
z3-i1t34WtPVhfN+TZpBJtc@e3vln-&VUx&*Hiv{>oq9{oiLtMi1_W-1^HYS>h+38c
zP*8G`W!XT)jHA09balgGIeqzU#djV&cl6fyAgy1qJ95f&NZl-Fq{JMg<Wdf@ofNuS
z{np)Ke{qg{>9OstbQr|`1b(q=skoyJG@;+pkyn^xx-}n28FhvSp4T;ZfN-+LXfYXe
zQ1J+!4O;{OUJNjV=89EXDRj$p>}q@zcr^fHp+EFn_^ej|3A`0)4BpyixRIFR-I_KI
z;seHm_~?b@q^%A+5Bp{T@_`D_{*k$B--+XGA1+7F9I4F7ubAbtcl0uS?6OG@TD-JD
zQz5MEb)XZDwtdj{-9e&RuxrPK^`jdrM?iN$tpF;>EP?H!JcI3DBZBRc`b}yXN$Wc{
zpO2N_+dWM3p?6Om+br(*=l=-6?x(u5M>TmzS3L!HzCQdi!dIUno>oh}QB~J!V_ken
z(+s>;jf=@%kUF;|A+dIEVbrFH+?;ck(xH^lUy}v0zFL8fdhY(uO=0=o%juse_I}<D
zd10bkHF$^>;$m|^w`tju?03d}lq)MTA&RdzF@+X!`4{_Y3PIY%cupT^_$Suw*44Gd
z(ExUHf&I_w9h&z;MPflOdE=*^XPGHG=Ur-DPmykRcqkL05FQZlT%JD_zHlM;*D1$4
z_ph*@M~~^g3t3$LmE$}C&b;p$;Hs9w9{uf7Q~s>&1HCI5J^5#?Pha@7;Z2#-XliO5
z9(g2I5clGLw~l7V%X&}|ZeBOf7@XQj*K8>c>F&-Bf1m3xU;d+GLBecb+C@xM_DA}&
zE*iF$JM!Z2pa1+9ix%G={YQ_nzI!fT6@|?OUn2%R@X0Soy#Ex&e{Oc1C({>RXyr`)
z?o^vfnfR;aMZrt!I^4No6{Wt6r7Mb#C;z<VqL$~49#v0=??$r|t&+Jq65}rFS<Y7A
zzQTjZ3&hMY%V?;2sa`pBV$>$hP~th`-GSdkv~qyMv7Hacj_K$}CESTc2Dczb3X7*!
zk{2x?vIXu>Pxou2BdavWbB7huFI!-|ZoSp!UNG%Xr!aWow432S-F~*0oqYVcS7Lz1
z-===?qek|f?$WY5<@mD~eK7%b*W|>EJ7DLixqgO_j5!-HC<$sJf`2G&8oPAS;_Lf5
zRF|ekw>3h#)lj<CKifg}8t+J|o3*<AmD*2lOsQXau~STz*S_!P@Y{P&epq=*kP|2}
zWcbG0;p@--^`B)sW?`}_4HDlSnCj7g7XIDDfXR_b(0$7|2Q$d}`%S)p@7w-$UrW$W
z+16g}=j*4_FXiQ1kuYw7AfMkamf%NQ_6ZFem|&zK67vbs?DE^B_0ZKT!QT2>C*Rkp
zU|^Y`D+8uoDW+-Z*MQLBPvL)!`Tvg}|2HtRa*6)=_-Vlws_RBdxnMBshd<u-{^%B;
zU&Rmq0jj0s`14$#qcyth`2$nf#XtWv?VQ-L?(Y|cY5ab|-GiP+7bF4vJRR?U+Wel4
zy1|OOUHCUwf0x>?rJwMrA?sWPIReVBxiDQm=mWmJs0S}Qw8DY9nn|Fy6GlBW@0E0F
zc@Uy11)o{*aiSB}S}?ayZFX8xo%Jrj1SY2>7i0v)7+QQB-IB?Pv1DsU?Q^{X_jUIL
zEU3bIWy)eTPK|3shu?3Ve!LL+k$AEe`u6?(p2>h2qiJ|MT_!S1{R-l;TR1r2dIioJ
zeaY@(W;PRR!fKaGIM0eu>uapnL6ZE5g}Eru6oUZZZi#686rW#m4MX0T!vx?A???H@
zR&_QAYzqr>3{#cS_&O(uULXO}%DqgO(qS9Z!g&k2$c5U*5cv4*_>iY$ar~`&MTyR&
zF~ysMYM#y&^3mP3fw~F-RhBCe=W}GlWlndu1q1ygt)pJ5&EE2C$>0gRQZgX@vxrai
zQW_Wg=|_MIV+8^bNaqvBv>`DdC?uL~I!++q;fdpUg7m=Y7qjXT@h1tR5{_17C9*m3
zYEjQg203S{XlE((B3ISqr@6Dt9xukE>G@?CD{^cJ8D!Jn%TKMieAT=NJ)4HuEgX%%
zf{9mm!v_G}?WstFrwo!dwbLBZoh}~1k1ZIoVg1sW80HOz!1{AE;X5!f(*{A`te2t}
zfs1Nu>mLXjVWb+)PQWEsy2Fn@QEF5o-wXdyk!x)e8CGUt1;9`iI1GAkX{`0f=?95L
z#rlDZhz5Vc6;k3hu56M7F<#e!B!-4-GS)ebOQUV>>oR}bt(I||iZ5pb<=RSJb#31Y
zf~r+#cu|7QJ1VhCSikXgZ|!G++K!oJrP_{9*$f6_1Rm&6-Ne{Kqc_(PyT`%Fsa|S$
z#V;?V<m8MYF)^Tg(*%x7JOEO)Hf(3$3I>A-BsREkpgD*chu%CKSvstXYR_D4&p!Jk
z>xbEux7aiMf+Mq|_?eyQUS!2`Mwb<k;HUT<t^rLTWfKX86^H2~NnC@ZEoAzurL^})
zW(tk7^C+ysg)?@oxM2E%TzmTg*SifV5_upZ<=<lCM7*{H<5rwmOxXP~q=&r63@Bj_
z1$~vs{PP+v_{cbXfw#T6%QR&bWUAsCIol|ABx;CW|MB)Cm@c^L4=*c+SP9XoRw4{5
z?r<I*Yjg$&^7vX-Uvn9@8o{RQ3IU{;J)3E7+B-MCvbox8E@~0Eks|=kFo@XlD$NEt
zhYOoM^wyPBj2H9Plsf@%<~)BluJE8~-<Z6?TiyWDIPj^;*(nLPWTAtNlkARQ!mN1$
z)rMNb1HdzE5P`x?nzLbsCB<!Elky9L^<vs&PGtvOMFCD%9=|BLV2b-&ja79Q9fqoT
zK)zE|JC1MpP<ZWja{j62>*1i5aw)6kC}$Ui6eE|FHIDA0qz2oI$EuI0BBq&k1a6x)
zpexwh9Ii5Glj#gZs<fR)iyw~`nww)zNtZ9_Op~^Zi{dMZtCh*y?c(R6x7=67!}{@Y
z8Ay7-?;+5Ek|p^nK8QHbqxjBlWO7qu#PU<Io6G5UofscUgDXu-ublnqx{$O6XYqpY
z(xM06wR*0L+W}-9b>(Svi9ZF^7cxn%2)Kth1r2Jp)g5zDmMe;{2I4|I0iOZKR$<}_
zxUo@&qZJ-6YXbL13DnT(#0nhRmI9EK>Nw@)#~<o+do}`bLxMR>lT9}U+yC^hV!-$=
zW{HW3NhvC>MJOt2Eqqj5_^>QDX({h++4T`v;N=XHVv(gh-Ci3=;~YiO!-#R2W>4)~
z6w*GWt6fYHdyl#V>=O6g$4D)~yfyFbo<bOBVc7vj09i_@m9(zXpkh!VIps@T_aK}(
zL%WP$h_cs#z|5xcaI7&EQGi8m(tA+txTmqNE_^=bcr#6Q4;OT`<$-C!-iFSG_L_p9
znTp?^=921e0En3snlZB$fJXy3)O7*ZXmeZS1Y62CBf<K1u~QIR2WFBgS{QCD+yNuy
zZ=^hG&Cl(>9Nj@P#J{T$oA5z>hts{-*7o#~Qy{v45Kt1Bxy+#1ItT>XvNWc#5J&NQ
zXdVrQ*w{MQFk@2+QugPtKo^}i36~$U)%<b<BYQ*gQAJ7F)$B4gAgon84$qT%fYUNl
zXiz42c?A#@1BA47fc;ikg~Q=vm6#0$)hQ_%UZ|XDm2pi?+Fs25+()^aK_O+APF-l>
zTvsqZ^-}eQ_4ZVX{l%>F=-r>1_BOvSQB=fu6(kEJ=wf2iNN`--&aBA+u)%P6JU*5i
zi1ouxG8j7rB5zWD;r>#aDZmmL_Mq!<=M}1a8Ijt$OWP3~u69l_b49$_3fx0F0H>W2
zwI9UMDW+!zF20d<tGu6U$O9Fog!~YWYCn6ahOFj6zr}pzAi0=|uf!)4>I4umF=x6g
zU0aNC^amk`;H8GHPEDvn1Js<c0@-zOZCRx8-cBX>4a{xPH@B_fV(o0mxBN7I+$B4y
zOK613D`vjmgI9`EnoxHqOXp1h7i#ZJ?+BbjXG|jfnvgG+&=q)i2f#y*1|$SgLrgJk
z_Sz;cfw1bhAAhES^_FY-F1qO&HXqE$ENRr@n$ImT$ihf*2q7rrO6ui|xY^3o(x!EB
z8jX(2&2nc!bp`<arf`OREH5Jf%c2*dGT(i%YDGoR8^tw5<g(&o>`(dkW2pvGe?n*n
ziq|p;N+kIs<Il&=?^bb5wT+o$+B1L%5A3LN_sUr0MyySHm!kICt^xY{#r?LDq^Fi+
z<0hV-Mf5fFKDATDFn{<~93E?#0(uEj>E`rW_|C26hdmh=gNeBVH2iFNehE4Q2~g>v
zSaPkIN6|V!m%6Q-GXoibPMJluSIn*9%+k8x9Yn){o}iC9joi|EN)35phYp*U#3GfJ
zm$epV1smb4ba*PIsxc5l-b__X;S1I_GmwX!Qh*FvP{w5{B49RQsXD_*WFTDUFqe_1
zsif`?g55ntR9iyktU$KSeqmvze{%zC(oVHbD>yh1#qAwVhcg>HW8=4>?5dSdkOc9h
zuF)cRd;&8+Ok3kjK&18>F3j|G*H(}m%mdGac9vjcS7-YBn+}pPuljf-FQ$>aa4|3k
z`N8R=Uj*|)xqyg;w{7PAjrn1U3;j61hiUl_-WcFjJ}%G|zUQqb&w95pSHi(`Iuo0y
zO`p<c*!yQrA519!a!LGZ!;qRqx%jP{KFtwYx~f@2m#$>R+LK<jnf$F*f-(4n$Hc8_
z+L#+h3b2YsOwAcwlIaNBolOS__T@^^=nY;|r0Jjvi6c)>-ukguLww`cn{l(IblCm!
z8fH@_CXc~LPNOrs-#(Fx9R)$gU|--mhlEN1sU)Cl1E<sE^D+5r`Iq!pI29#nqyW5M
z-xY!)l1U$@F}uc><1QxmhM!h8`XT&>PkJA9#?`-H=r4IB73kD!;hvp9KA-<5l)sq5
zPxVi&%1|SyWe^fRAA_r8cylIHQbgp^QHZ9_0AxTHcv?EILvuGZ88nd*BT}w?xHP>x
zc0bG_gHKV;tsTYOqsWpmR<{?gWNivSiQ=Y-cL)RL>~$SNtL8GcSf(fnlHS#46u$NW
zzu-*k)`i-{2?<^*87kFHJHnVv9zXHq_1fgw;`Fa&PNUI<=~}~50MTftd-kFVIuhiA
zDoY@zssfCwL@J`z$(a!YxM^_Ik*B<}aff_hbthszY7F2srNseaaALe=E{a7Hb!c}n
zm)h&3a(WWdJ|ENHjF4?h@a7ueD8`4KClt(n*uyHV;xzoa?Wj!$@)`=6jg8xY&vAnD
zQJf~4<Ln^)<`%YvnNP$R$cphDk<v%rb__04oB2;zjYsWLU+)R_^J4Gh#ViF&e4JYK
z%V7a1-;z}Nao+tv{#58z=B$n_NtSL_8^Ozn!<weh&7dae&&U3e-|y)(wim<CftfuK
z<hiA}<pY@NLYZIcJPiylrLQC>Q*9VaJ{yV}sw42B4zDZWwzamk%1?eSzu@m0<@~7J
z;?V^wZKd1UB|7Nyi1Ua)j|^vp#vW!tfM9JqOZcx)JTqpDKs9b_+tNH+c#Tyvc$U8@
zP#q9c3_YJa){zsiO%<y;Rb+j!+o`r*7wfz2gSP4QNzXgpy}BCZCNU|W|GS-_Ff{!W
zqK1T>fCV~DKlZ;o<ZH7HOS94aaV7PQW5_I{A)B^%(&ovXErVd0ke=tg`kL1(sj8zj
z)?w7jA(-#rcxzVj@2+08br|}~o>$~+p=?(d%!jUgQ;PAUrY&w!c(aW!?6#cVKHcin
z)k~d;MbQnlkv+>>bwRs=qvyZphvOb)@eH(I-5YWHboQQgsLSY^3fz6ylpEn7!;}nq
zQ)}aMFw?;hkPempSLEb;Ii=Z<KQfo0W_N9NH0a)Fqpn)ez2||BBdAN#PX{dEa6iUB
zNhWbfc>p&ugFOC?UciGHkrs_=u=gSR^D!Z!9q04-!@fb}k!{fCa(YB&#~+#<iA*!)
z>exN{-t`j8PRqKZLR8KVt0x|i))9y!DyeJFg2u4#A~6-7<mXf0|1+eyzPibr<MCF7
zo%s@+w@#riPhT+KzID&B42{@zg!|3{55l$B@3d_o;^Go)aSk9@rqVxdf;XaTYBHxT
zmsH`syaMswotg|AB$XbEWzp#L&YUq_=$)?iNZn`BF<mS$+zO0pFY|#t8GrBlVOb!n
zGwVI%Y^xW{0v`Q=s(2M#0#jDV>m)H}7pj#H!&wjx+frd1!^bAfWf;r}TX9d=<IDT2
zLEXy;9EdrFytJNKnv$s|WKPM?yh+rgk9_3ed_<;uLKJH5;|!CTUNw@=gYi4YqP+b^
z<L^tB%D*$}Z%b{w1M%PLUa0fY);dpo6Y=uBt>wcz>|EoaYa`_29Um^87i%rCFc=a(
zx@#ngS_#U(LJo4Hb<<1O{W<_!i+Rmu<p8oSWJ%di8fmLZV~9$i+-gd|z_?LZm_sNA
z^l8r(N}!ta@!K99T$sZ-x8#iE<hPlZD+-b`CK<YVMw%D$Ew1{nzhyQR1hnkClhIog
zS&Dlmdc&;g<^_kqG*m6iytQDJWve3;M$!_J<TrUgNmH9AS0gAZHi(QWOj|okdwDi-
zwQ3=Hszc+9MvsT5ltz=mHGqK8;Be+CP_!KD&n(FFBW2F)T8aY^feG^275P7~hz{|L
zP3=oWqsvh31;DIeZTql_i9`jpj34W3AW(=)@PLB=qlV_Stuy5l=p6b8w?c7Svs*uu
zX5Vmypp*$pIG~Em9<FmQg}nUs;e{ssw{M^2mSqhsOu){Xx<Ad#4{J)iJ|F?M1n70e
zXyLMeAOj7IID$<&d1^BToU~^MaX@<k8oRlDv>?NGwR-s+<J-ZOmcHR;ugzoZt7?J#
z-iqQ-P%W;PT8V0gYH9%BTofdypy#kdvi3M}CVduYxv9Y52^E<JXn<}cHUpk0*SeeV
zjh-#74Ug8(#-_UK{5olMe&bVlmuZ%mP$;&cDNp?3TH*sUhM;1);P3%=dw_>hPK1n|
z%OO^#;Ng`iSUA=%Bfd4m8*2c3o9+B@Xt=^!i)~9Z9sFWt%L-u+p<kY;g3XKrCcS@2
zAd8gPlwTAgjVnlEYn%I2rs>$mEC-O#7}1+O_cnK8IIrA%|HvZmeAdRig^%w?EeW6}
z(HJ7ZurIO~7nfb1Rw<ankvP-m%-!uZwSiA;BE<pI7jBBp><OI;!84AwB=*_(2LEhq
z-3zeqyNi*Zw!Tm9FYV6<fzA7$MOoV0OKh#v?ZsFU!r5zc$tk>bZ0t$}Zi3Gx5{dC-
zESw03$MMhw1-zMwYl1hmRV~@iCqi1@x%aAusNHblyxqH`3^r#Rm@jXvKXj0wO8^4b
zNJYhh3?#5J<hjnq)n!5;AkzZ;7cM*wJ=erw;&7oy%qKA$7QKdHIqKs5_U86y#YU?X
zoZja0wz@BpRbDanB9_-qfTL|uI?;@9U7$EAGK_EKjW00eN;B_|nP}SUM$tAUHcd=>
zF|E8zgUvbYCbM}zdV1^u4&X&gNO-tzmZ=c7!Q0>(SZeZCG7_0QI*(Qy(+s#$pO1A2
z&3?wJ%rFyY<YRGRV@%EhPeoI{Hzm{Ti7dcwurjwiH@RM>dD`{-?KEjitxq_2mR+RM
zLa#BKaggY7GlQzP5;<m0*#olXHV{r{;yP2Vg2VhkENmV+OkFDN7wOu>;0Uq2&agUW
zQ)j`@i<|Qf_q#LuUs0al3+~5(1_S1OKY|7m9y)ljiHdA$R2+cg!6+rKieRN^UP@-t
z2S7HODePN;E*ui@vDoRo%~b{CwkNjv&+yX$70>+k>38yfhP&@6T(|H^AS-};ng3kD
zD7G#V)Cekrmy-*GfE9?(w8iX-jH*((XfB$dY%mKrslrXbv!;`#<n5P>Z{NxWXIr7H
zE|P%&QET3u2}FtrF{%3ziW)y-aRejFak=&xznL|xQZ;iB$>f*-Ixt~V>lq0byNK4k
z*~!1<DB*5lRhM%fCcNDTH-LN+7@1|*VKe`53lmJxQkqLi>*r&SB%IEH>4_|Wgj16d
zvzfd)Z7^TBql`@<NBHA*wE?Clxzawrl#F%Gn!q^DtTGxXvpX|kL!Y*L9?F#ct+>2o
z<Pxw?gvlpR?Qu^d;wkZz{eYzn{;XkpJ&P$zCy}OjvO{AgCMJ-my{%1Sq*+jASYs!y
zvoH40aP`cC*Bs7=wa>>;%L)NVxsu8hA>eB59&E~0%$S`6{Kb@w)wR7@<iF#m2H#TG
zrhavq)Wz12mWMKF+$85ldr<b%J)D*MfpS8BK?w#R9l~P*_*l3XAholJ#wxY#tTF*y
z!@Vk^Uox5yo1v(Jo;As(n>ujDMJCj}o9nqWgUX9GY4W$v`uOBe+<a(}&?aY!QjpS8
z0wtPyfm}(5g0v<F_n6ts_{)MB^ac*3(%<&^*at$10KqGao-9E%+B(^rx?2VS!Z*3X
z1Bms{&n8@LZj}w6%zfVbnDT?eS(~Rh{aLD1qc|}k;urJ6a2qE{LAR!Hw~5X=GZw%o
z8VKm7i8G-2DPkaciL9+jV-L79;?_;}gq!sI1|wl2F<RqSOdP&AIM?D<`1j!erCe6F
z1l*m~-`I#oL;(GQP+5osM6Lo~nUXOXZfq0IOk}NN{RV6xI!quwV?>JSFlakRBVgL*
z@~Xo3bMM#KFqb*n{?d!8d7L*PqqiDo>b`FQYimmo-D>}<``V~-C9zr$UYWttabnAc
zYtCfGbRZ8=MOj(#YM+2u1e;Jop(OCenI;u0`_%To48g$5ygy4ULUA6ZAf}+C$n`jL
zjJsHnp^{-TlN8nk#N!`k`6*a$EDC_VCefEN{ATN5<IKiS?gyPgoYN{Fna1IQ%{l(2
z2-N`_^q!;3ZVDr}_fS}-=|JWC+2<?kEpH$GI!ZjGL^WQ_s_Z>A|4y>`ycq6c`45gG
zME8_S4CXU{ghNoTAEPasl(S!EnP`BfQE8T@({M#4CR!jq6}^0m7ax+S)VVaEkDY$o
zN17qMQI6Y~HQM<&75jg!`1KW$+AyV;zFZL|T~3}t^}2O$bz0R|o&RpRDaUS{W9^jw
zrldQ(Rl`iSHlmHP0ZBA?$`)VVL9ui7WD|{kMBKc-@$GDJ&&}9`dw^eZ3fk*8ckhZ*
z_U$A~1nW_6!RU97yW@6B9IcxkzJ4jVKJsX+!=WY;DJ-*(aPR-6piS0WzF^Sg7d?82
zhi}DG(|OIONxB|SybOPM>iaXMp)$X<_B!#k2J+Dcmo++pvOM$K2qS=<&pwxE``BsX
z>?u&uS>K+byFoEO%YWT=J2vz_G4OeZzW{!-)ImN?l5=+}MmfDQR$d<z#a;TnF+W-}
zaHGv6h2?WS<@k(hH{ZWvu>Jw`G+P9vKDt2h@WW-`tN-{kR-+OY53uLK3x6GzZF%zZ
z+xoUtp-n*{iqZ?h{3YYkbg+y=m>f5QY4!F_pws?Y30&G&M90Z%ZNyie_|3H8_y0}R
z<yzpT{bZ~@uAw+4?bRWkQ96WgP?zqGmeCxWn3wCQX&Vd)${Uf?{hQchRurE&a;sn|
zH7}?8QFy+AqW9VQ{-WEqNwD88LKzvp=k-F~B+@k)8&@W&S9|rn>!b`M9dbDDJ?Ei2
z;_u&Fv}>kNnzKNIqkTb3ZW}zi?1L%|bvtAJXsN@KJO4E`Q`sV*)*AdIS!F+Tg?g0W
z_rkQl$fWak5WREgn-H@N?;pP+eK>aQ&rLh5{}nUNajTH7PH%E)!Se?4^>c#WrRT<?
z8eeIrpOlGdepdK<0m1$0gq%CI>x)CHS~Y{`FL#}+Y{CDP#&}ZDTmOpIm6D$I$H+{D
z#ME_$q9F5jM2K5GA$MUA&wuXv?8b{XpO#k?46I{*8>tz7b18~JSmaH;3d}HG3hw!s
zJJ_7*1iEOo@%{&`NlUct>>+7Dw|;~hMYm}l?#OsmyIOC1Y((um{-w)}o3|0J7F`<k
zHKsFH(BEcT#n*jn0rC5l4IqCF%8cjkOCRXP7O&tQ;X9r9(|<3FF3<IoYlEOob`x)$
z;^V4!+tkwY@Vd_(j3*0Tf%?u{r)6@VeUP+&!azk_p5HNQIk(xhT-`J<5P;thj+cl^
zhktM^Tbw5Sp1ap$%KppJy!l4O%V2kJ{aRY0jf<A7-M%hf^BFzuva|nAwD@~3o5zJW
z`%gXj&klSA@N_*{#!uD(j0rWkO(~_XcBD9+%SrE;(=j%ke>wlD{qz6EivB<Q=qEGK
zQQnCIr5S$9*Yy80v-s+4kxOgD+ZofZ(q;m-(|zTe7c;S{J<~0Y=<2?8j&{i6Zh2AD
z=8B38rXL~EIP$1b+B&D;SvEW#`@4Zd>qjU}5!(X^7vz{VUL#f!S1Qci`HOPmT1$$3
zw%I>$^ZGH)R-l0Sgg@hFK6$JkKU#n}?K}!w3=8{&`pX{p<iyc}t>aF5*y5ct|9J=d
zAMgI(|L%YO<$t}C2YW~zTz9yvHP#&Bc(GjQTrMj;4bE-N(LSQ=exJN2`c9Z8V1+vP
zIj~ly2&&CN`jQKMyuMV`&QJm$_eh7vIJmIRmw?86W5}k!srkQc_>*kn#$bpbJjQ!Y
z*T#}_D*fn*2jHrb8yyZ8zS(-V_09gBwatSimYGnHv@c@`P<N#79EAib>`Ekk(WARr
zm*An^5@^}>A`X?N5ybiJC^x}GxKgVzz7WA%CkatSo7^Nznbq-@W`&ltq&K`S94Yhu
z;pz7ESmAh;{Kj7k!r=B!|E_(9wVeb0@DV592sJ7awI2L_l%YSf*Iap+>vp7XI-Mx*
zG+(i0xYK<&fXVu}Eqms~8&d_pS+PUaSmaUkVv;C=cl1QKE{qhI{bz!np5}FRM`q(p
z&YRA6OEYJ7^uA0IX||u=ye|4Wsa)Yq(qOv=fPkG8m^uNcM<hDLI>>JU_(_@7gNYGu
ze#uPI(H@5s7FJaCWx7d`@UmDF%Wt~YG9SF({m*=thL8I5J|Fv(<fln*Bw_w*WU~;-
zy#r9VFlPUd{beze$RtOTg>T+}J~q<c!3-k3QTq4IE)DZKTNaoaF+oXZ%$^DbKm6AX
z)5Clv4-OAk5)wMZX4?KeHlXDKFBSEjSiE31{Nzc;)L8jpJ_zN(iJVS2oc<$aZ4M>U
z8!Z(Mh0QPz5&^=_PG<dorW{&^>^sr;`B)XwQz57MP<0o{`Ft$3_R8F$FEgo8Xu{gH
z>~K1i4XfJ>0(iv!0m!^wADErrx$)mC_GILavBD3riN7A@8Ux$fuCYvx<Seg-jqe^7
z&+T8EN#g!{+_@E^u_$5B?B!U8!b<Q+X57DoX4@KJv%;HSZgrf})Y&3nJ^@NCdtB6{
z!cdP$D5$kuE&(XpPV62S?J@r6U49u1$fp=?G5wC3>c)A%?i_ah1x&MZGsppe4E;kA
zCUprDFaO*6%qIVBM$4a%@jb@q%ir#neq3#g4oC{=#~qxY_kWZVF~$Gq9nz`mxG1$R
z3-sUX3XGUw%sZOc>NL}E^Q2<UCXQyRS1q@~J3b#9sI}~8mSmbGiToIceg9+3B)-Q?
z_wtji(vQ47b~zB1ZCGPHR+j-<Lcggm+c~z^T$S_zC^3CFGJA+zu>Tajnkg#ORhYhG
zW_4ogC2-XkK1nd~`Pi1umooyGo!OcypzCz*KW5YT#camEm`z>g`|sB+yJkXx6Vjh=
zmfo59QGU-Jm|o{Vt@Y<)Uw>K1HDJ&azzOO4&k2zLPKbv>vRV0=e;Yc$xsweAR<h-w
zLm~Cwh8oN_O<Gso0Jz++NeZvO%q>k+7_IbWZpfsY%SpFQ<v$-QzUP?_u!fO=EB~4{
z0@mn1_eBZ`1TI$JixC|Ddn#Wx-VE6Iq_}?_UKF1NY;*vyQPLG0fWiL$zwA2c1Hd#`
zq5zylsU1E3f1kxYBhTT?C+(fPgyAD&7<pS1a)|S>Z~Jml(vkAN9~#gGAfd+5T=_DJ
z#!AQ+^E$jFI6PDT#h{O(10uI4$p`9(4m(Wod0-t^lJ@_7ekYGaxUl>#T)_WxtN+`Q
z`9cnMlee<@cfz%q!~x)0)8w}Ezu0^0sJNDWTbOKih+qK%!3iEbxI+jo4Z)$Y;Ly+z
z+#!Zra2gE`4UKmrjU>1eq;U!E5Zu{cv-dvt-gD3R&Up8I@7}ZDH}Vf-G^?vuRjvBX
zIe%5PW}&x@(=;O|_YSuM(AZ`&GkptEe~zx@oj-fU+weDxE9IE&Xs*qhdB=G9MoILT
z{tZI_PI^&TvwwI~H9Nft$#nWR-0+jf{Q4t0c=GJSpP|3BiFBDho<8yNRsVBbXO`Zg
zW6j|AvRjl}nzp(P-K5FNlj|J(4fLb8+8n-a^56`>4zQjcl(-$5*GN0{%V#xL4!&NO
zLMOk65E`U@#|auP<o5mBkiCtf=Ic6>o8X-*36|fVL>1bg_xDHP!@PIKhOQzyPeo(f
zzXL@%^B+NbRClp!c=OnQS1f0q8GXM-r1kCn&=`SU=E^wS4g9;OPJ>3K%r#>O`ut}*
zw9D_%AI*A&j#jkglZ+2Ym3(@!zB1fk!><qOQ#1;nJvH`t0gE11f1=L{EGS6<GgyIi
zUz`(oUip$yvz_@wgOJTDoVdFokYgQja4#GktXN@mVO~(r(@wF$dde{|o_UX5O~<Q@
zzA6+qb>%6v2EC9O1%ldT&<Tg4CvW9DJU7aM<vM$<F6=;K014c&LTes+K(9Spq+J;-
zoB->U4O@l@qk<JrVojdndLAJIgu)VqNZ``AQs2kELZa*k7?HIX66vz;KVU{X&O2^{
zf;<i*lt(B-8QI4!mwe~w!-Ocr7n<6C!=UHu9ZGPXz1Ee0+Vvsr8j)IP=k_Ha)H9_h
z+sX=xK9zGVd55wuRAj`ldWq|4n(gY?2&}L!4_q7f@>@370~RJ*9<B$Z?IoX<ngh~^
z`A;cP_RX(OIwMH(Y!T)j5;vTT>a;;T5lTimyDF&~A%t_5M4MCbTLngRq773K`AhZh
z1wxhXuiXUhJqp;>MsazL-6UZA@k#6d<fl6cN0&hzSD>vmep3X8O4w4IVjFFj_RQ#e
z%j%<$u)a&-m?X!Dh7Z=h*t3_`4mKhK0(B(|8#Q~INAp#7O!XVkm9ngfYzNK9Bf8ti
zL)z5-P7>>DMh%uz$Z(Uf>a<7-#hy|2i4oQ+?^SU=6hIKv(`cwJt}Vags!i<WCF3)G
z`iBUpkJ)=hliS#wbz*%R9&`IQ+Gr8rzLt42ed(F$dwO6xzrKXd%2R<6r@vu@I->E<
z!Ru1^RL1@AABOE8V1*{sUZ**+>$$dDES^T|l`Cxjn}DU0Qy;$@zvzF*NAb*S<*Bo)
z^s?KqJS<j@hVBg349we{$S>`xoc%!=a=usVXqMViE4AMFev9VKe~5siq?#j%PBdZc
z1>CS!HwOzk`1$vuBbs;j-^d0{i>RHlox>9upTYl37<=cp)X#Q+{$)u?-^=UZzcCb=
z1EM7zZ>kRt|Kr{M<c{vCTTzCFg6ihz;~Gi)&098F{x^iRrm1k;Swr)xN0V=DV8Guv
z9KZ0hrv%}TD=D<Z@<K~2U-Y6s+Y%bkuY3y7%<p?LEfSsSIFXZiWu3WumiZqv7R^Oc
z&At!qEjn+{0!^mf(b9~1ufuPs7hTTeRxzkMp3P)j-;@qb2`Bu+<M<Or*!~SgtpA20
zx))SF3pcC%3a9b-*63{=Cry&C<!?I^&AwNXC3R@zEnE7rN3a)vE917%eCjV*;rQda
zIk<Dc@7t|>>_n4e^zF#`_#1{2HFZF`xAo0z8Jg7p*8S4AJ$X%Fg6{cXG*zSNP(1ot
zdHSsJ)2&4Ky$>aC6kX8633Ko;)0O#;7UiMo(Gl$>es)8!_y4s)l4xn_hb&e;+jn1E
zxHef^p+a`wf~nQis9&0kY%Q8+92cl}g)?29|0WZ}B(mRqKT~eq|5H@tp-IfOEc(i1
ze>)Q$1&8x!u1Cj$SyaG&MnLoTKXvrKhtTaU0|sDR=vKuVqRW2SqiUjd+5LcSobfEY
z5G`!lYF6fJPW~7-f0rMqg;7&q3K}P!%#;RP8PcOm_|N{qFf=zj+W$fR{yr8<2OVp^
z(4hIgc})kcD{36ne!~bqak*8Y)PJbZ%p0?rfV08iC5d%S^wVCvecC^~qwWPS(A=xo
zzm;0Q$3V^P=(riM&UAES_)|Lh_i!AZ%`9#29A9qzM^;0>AXelP@TBtgwrH7AJd9p$
zTp7H<obNk1{qeW5Qye#xh8H>8XMdy7T+jc&*?(j}5|U%OzhPvZT&bb^6b&kicIs#T
z?ajAI#`g8_@3g)6Cv8tuo-|hBZ%CsR+t%<^V-ly=`6c*LtZCr3=$~5s%<$y=y1@tk
z7FaqFzmI9ID$AWqxxe+#$xPkgf@5^4sE^P2t!75+3pbj|@BkXWLCNSOo1Hb*TZIxZ
zGg)|Z@a{t4R_Xa6%akuUZ)<r&J;=FsD-u%Cq-}Fc!6!b~O`yM%uYXGBdw+K7Zx}n%
z=><Ij$C&{ERXy}`w?pT50QJ){*!vA*UetYO=kWh^bld$~xZ3@vaGkvkSM%F&y%aB<
zzaRZOY?di!pmRO3##z{v*bVP=^=%;CCKFVR@dk&dr*-=WXY&3B|J9I1E7;#dHuGAc
zF*@RJ&~V$Z%nM2{my>;J!=8Yg$YXQ@YjQ?%v$uoJb`BRG{{hfd8e$ST=s1#Z-M)Gm
zfbLDnfks|5ZH%M4Gw1i}4dkG8Fkv*;b=}(j)$by`x$bw7ew*7x|Do;<(OK;8Bdqz7
zCb+}r1PwHQKvDA#^uazq0%%&q51>M>d!ZmDKaQ!e3`h)l0^u(jhL^%c#-1kxo+?)v
zQbdc<!4;I~9dO<jeLlsd0r<LC3g{&7kmgEdlR9AECc1OX{cLbo#RF&e3FIbrlS8L@
zmynM<+#Vc?_dOX~gFwXB00O8Zt;Bh^s#Q3oYedyaY+yV!>w`=Tbu6luv&H}3OyX-Y
z!VlxTnVKd>pLEL!rB@%#htxPAAFJpJ)eh<I7>2`TWT&ehk8Cd56rgu(+4jwmp6M6b
zl1IW?Yl~{EPS~}D$bGfPox@R@e17zF&;T<C+IW7thDht8I}E-08Itj?qYCl(zh#=|
z9}J;|_zuB7ur8MJ^bGsb&{MM^Ym#&KMYqS(zEj$vrXi|#^Wk>Va?1XVygVE);F;8F
z*OTu2wMY4%mP3B6T_Bq?uQDfhs}tYZ!dit(1mS7xS6&I;s~fps)3Z&|C(?%-PcsGw
zuEvsijuUE?aM65m39VLNmUnyQ>@}({#%wmZ^vUSX(bXN0wVnWN(S(jcfyjp_l;s(x
zN0i~Ba1e(NLEY#i3IC_@E@*vzb<hTdGp*iuPYj8)5*HMVa^@?J`Q+@CM{Q0YBtQeY
zZbC2gZXCaRSRQ+2=%5M?dYzoAJ)r0`@reN*{yMbJqM2N{hHP@j>c?fl=YlW>x0291
zW5!frsMoboY#)a>yIqsr1?9s=N>`^kE@s{IjxB&;$5V5LiV(1yMd7DO;jyvU6}^cb
zK`TJ)RZ_Z}R7W#)06)HE71gU=cgu36ZAp&tG{$jTc&d?em&%HeRqS%d2*~HxuTIt7
ztPMzYo5)s2j*4#4vd67p17j}M4-Czq69G=j$ec*@E-1qVWWQ<g9Nq@ZX;YC)@{>h!
z2BS8G%6ht++-2mX&XN!VSggR3ib!Cw*@RpaCkD|AsFaI)@MGc5YQ0;ecTtcH^J81R
zQqzJuINV=Qw;qb)5O=qS?__c4><3%H<^X?$T;bQ<a>7rDtFvzIz5w$2c5}4FRD^Ff
zxlSL(1y$vt202K^!@w_iTDq^HXLw1CZ&49kKN7-E+cEPpwwmK1UlEg%B7}?-Kau1{
zY*>(N_VK*xQMH}QY0!XR{)Hlvqr)~a)4ZT29cGx0NAS)L&pd1~zQEBh+dT<UpewEY
z%Y#WH{fTMSO;KovYh6?$?U!i>T4NB>CpEPH^6MLDep9=Xn12*I#VjamcomLm$nQQz
zd7t3GH<P$^yCSt7#{Jb0;cTuoOQ|DgO-KKUbFRA8Qssyn_^2_T+NLv7Ab4$~%XrQq
zd)2uZtdsWP&RCiU@j!(SQ8Xi?M9^3;ZD*YI%i$SQb&Cux^9ZA_obXx_{@o(boum5i
z744$ri>NA*$S=%PMPK0N1l`sf8TA|e;vYj%o^zkIUX)flOjC;9g-(>u41|r<&~_?y
zyU@U#E)9*g!n@B#FZHEdoa`5CvV0i@hIbL0Y4TMnr=016#NT#E3#%fHK5Yhpsqx*#
zf|;6T5?+}Kjc~T5r>(NSdyT>oPWYDn0^U`|y{Y`_ynZVwMR~~xxuX<A12~>$8Me{o
zwMb;lq0s-zwkft7wHY$>kht30K3codOTiNNxRR3x$-^AF_E`pId))c7QDAyC<E+~r
zO&U34TMl{sGfcbyi(aq9@Lk<|#jvi&U$gEfwi!;9QFro7k|2eGluNx^238Xp2J-E%
zfD3sF;m;%z!~}%GY~+cqVT%SC-@J%kzbH>)BOS$+Z1ARsSS4(Vi+MWEe?RAP;!vM=
z7A5*_Ff`q%tX@fJHI}kd_#;7h>g?SBS4#F`dvg<ocS2Q(j}`g$T(st9DQ@w&@UrZN
zd*mifW}+Va2KQl@lqA&KCsna{mvmf!<Ct$xw}QD|#Cht|5<AQ1cVT<{d2Q4M!8>zl
zdYf%3YktDJ-(51ikL_}&G^1sPVDK*^{fw%NM!}1ELD-QSyRsRUA|W@0u3S8Z&L?tQ
ziVqcgBLx7ZjLg(H(X0NoRr{Z23yA#gPp_Qpx`1_u4f8u0`uNA<wD!?9V9>xkJurJ;
z9Fu?3(1iex*vpP#LtukEd1kcXi_m*dyP@iwI6Wi&6z?SFY^3|Cr+$p|n}VO%OsrC+
zIKVE*XOc8|u5Rkx#t5?YP5=h8wQvhV|E965zBr9NJBGF%bNT9|bvugBG4cY#Rm{dO
zwNa)WW}Xx0pE+6*N)1AGlicdSmNKW5@bv5{u&!F<h=Rn&sO5~gcw3aE7*i@pnpK5m
zgQ+{1j31H9!SuP;$LLTq&BT2=eOz(0-zvCnhVU*@Ho?}RZegOzg;~~gV6{5XD0A`b
zMzT-{HK>u0O>qEr0I`*%fV8g6P?Zp@*M7t+`wvTI0<hAJ7lq-|6gWIBoffHKV=zJY
zRb1korgTtZ(neDjWmqxj@|k;W|4vXQ$el_e-+*_k;F}RrDX-atsnWDV&o@vouJL+s
z{|@Q|zUDran687ea$+?kF3BRby7QjU$=+SNWXii)=xCaxbst5kD5&Fb3WBP3$h#gf
z^m-x8vp!Gtr#m4R-O9zORNw6gI!D?X$K|jz^o@j-$+=>Ss9QkDp9;E|ukGe+I2gr1
zRAlYxo>@D*F+XN<GAnlwakw{r>3P{Sgd4Kcs!=0fYaipen_`<Ttd<r6sNkaQF3qnG
z5!!tcj8aH=@Py1=*Wn5nCbH5S`gplZIMJmTM5m`OjStZ|Za|eihK+zY(r^*00Pqt<
zO9NKCCj}{h=x-J&G1j=wiL*)vvnpMM90$r`8H!>-nbGe%Th@)#E65%@XIiW&;@}rF
z*{DS!hFrqD$aWXql%F!vF{Y*S%}S)`Ec25AMABpur(@QODe?zRbq17T<Eq+tO5DFX
z%`nKr!Bq<QS&qME@L$MxV9N1e-=8@0W0}d1<<}TM*vw~GTiz)=*^KR%_`HFN&nWeq
z!}PrpCddHUJ*aCfBP{XX*EH|ybXG}Y2=)?14uL~=s<dhVWv8V>8>I=~pkqF-)T!m9
z;tZAjpxbuXZN<yoLH+sqNpn5#q@^EhIMGo~7RFhN)I<3$@4q{|J2#$Eya=hAeSwV0
zCeKLEw3*aXkfSdkm)4E$=_SxgGLm&o>(wq>Ut2lVC(jX5EZ5;Vp$Zo{m2#@EqOv4>
z{3L%!Y<Q0p7Q1=|SlzTgZ0ufVQa#e={Sn#neI?RD7)xyNTWAfrY02I2QH#FWY+lyG
z#w{;M{!<6qk%TCr%iff4F8dAQz>ecX4~o2TLjnG3F-f2Y1#Ai)*4h+JO|@+3s1E04
z?z34kYzyY}RY3G<&$NR&#PmnZE4aB3ZFL-sT<j94e^^ds-;o)3cGRhM2vh8yb&9Z4
zd$XM?)MacdE~uZ=w!v&NEi&a5!_x6xw+l$rS63KPm-L+G{<Zs<;k1(os+jz^NI=bC
zt-iYu5|p<xGVi8FQ8Sr-w6%pb$xo<Q=wy;s2;T=vt1o;t<5TG4B}+_s;a7h4#{Tk`
ziJ)%x9Rxd86tT33PyWP|i*Usv6MJ8c7RMId)TR#DwG8TYX`lH#!Om<!Vk^%C_f3&0
z1;5)Tx^jvJ%=rG%^Vr!$hc%!2L8QtW5bmh<yx*#;*hLxea7YP<2shXOA__2|5U6z&
z>H-~_eH!qKlPT{jrR!+5W0bG$;(W-T5<0Ql`QUOAtJ{?=xM{8;vI5+JXJC-$nUHtg
zufjgagY1VHRQ1@W)ir+#$4;7&_4T3fSOE%aDwCAP6q8l0cEHS3&ZEQ#3Obge5Me*w
zf3a~|HV+x&7z3r_rZ>f}FX<?j<`?9J)4GL@`sq|h!Syv>ufrX*L_nJpTa5!S5X`n{
zz9`thFLr6+aA91HueBrKjhDscebFNL&;`<CxL-AW=k#Iw4-l)fTkGcc5q&81QiB<0
z_vTZuV^#PTv5T~A!jRG<DX$%$enqEu(-{W(rb%|o;(f38S1XbIg5wR+6!lhPtX}*q
zg^)2}-N;P<!_C^pDaG~*4|HijSz~Qz3X!y^h@Z3fAc{lOYjL(_+|5P7AR!1uO-zRf
zHaL=gS{^FOI(!9fAzzjOGRZseI2%&XKO4h0)vxv>-Zri<v3%$T=9E?Qk=CIgZjL;#
z3C4>o-4xh7KIzdUwR%7z1}t&^yeOOi!n1n=mu+`N+v^4L$XHn|>pxyDdM0QLA_i#L
z*+(s$#k(S0%<uIHdP;zF!v)q3w`?y=OTwo7lbgkt7rk|}-1xwH<5C)9eS}S;(@@2W
z4{`IQBB_9gLU+#-an=AOBUO5Z1b(v5gGjDJRYRb=bm19Lu-t2Px%wWH`{UFId4qC=
zGl%?Q()exX)OcID{9>w(o$X!vc7|pw5>tBiHv3rBm&#hs?Bb84@T~YpcnU|2z+)Uz
zog>_jh3B9k5=(_tEOl~k5RWfnZWqp*gE$JBNPboDTti=aFRLIFoV7+63;eJfSQCKG
z-Lf(pb|C@36k)s>@5t%51_@vm3pb6QnRK<iZX22Q!O;|*O*p0!@GLH=%-8%@%2q8+
zj&vA#5EFrB&o1o2gZhXJPF+UA+4jPPp$=$|^P@FFe2O`-1o+0-iC|#8$Mb`7+q)38
zb~g5-!Zifi(B;v9cm_5XU=-weQNR{OP<`Mf=Poj}Jb_mrMCJu8DRWL~e!qPfGRiJe
z_R5WE7>_YmG@%W;uy6#CbF%skLtQZtG{8vQQL!oNTk6xVX?AW!L6Q=>RY%Mx;F(L5
zwkr2&SL?qe@%rES(Yu*R0ZJdxGh@dFcg)}Z@BG65zsl6_b8U8;EVkdEFu1;<n?}PM
zAkeMrKKKX*n^o`vb!-dRs_Ubj9J&#uoBT+<JPR&H-*XnnFUSr9pqGo1Il{_%m{Y1*
z!nBt*+vC?o-Y|Zu{7M@#hbbPi%D!_+k+7*u;1*~pR^<i;Eq-VatYyPBI_~`bO)tzb
z*?psIl6ZhVi+nuMJT9%c%PEq6G$JghJAcBfXDh`}WVNT%1*~X$Hc&_CYQ;cXhttp+
zW*@tyDgcrS?n=EGR-=gCJsD^|owkwx4dY=s{-vUrgCueq?FTmx&Z4<u{o+S+QqY*L
z8~8==@%&ur3jQWZrP<|!Q9#S`Ie&s1pD>w~-N)}<LF%~O{0Wa(3#XiW{MbqT`U#sT
z87V>)t3MXWVJ;ixF-F+ot?*5X29L5-3t1J}B~qq4qwH+ex+?QJczsW5;k>U(^5d0-
z)%r`JRmg6po@XTZNZKT-T}6jykT()l-kKe2ZzY0l-5Ik}6dO1iUX!bTFt6BG(Z%ej
zU_U@oLbIc~E8oNz4`8WuX|);*b1*6~uS^5tIQ>!&R(s=Uznz!uH-%PodWLOLo3?iH
zv{dd?Z~BVaX|ODAthjFa3gYxTE3GiUs=)Ej@Z)$eqRpM~vb-6V8)!TWVDy(lAGneX
z3nXE9eE~X82XxJ&tz;4XRrdA$z4pELTg#3_>5~r<my?ubTqt&g+U8Oj+gT<PiCj;D
z@hd7*4ic~G4jCFQX1p9Yqz+0JA(z&aaseaF9+_8<L_T6}f>&Knd~z=J7SyY+qA^=o
zS&p%tyrlxS=1SL;-aS6pUOm;cBT62wadAx~;E`qGiZasvwphJfJwB!*vzs{KJT|b{
z(<Tf)1Yg{FYnrlA$<bym6saxX1e?%7e7{JF9g`EPbRHS1t5;@z5ci2OPQ@01LV3qH
zYLic{Mhye_$S;a`GRv}xqVMz8B-?i_*BP~y5xC^OSpZB~$rHs~D=|!R*Vgw^#|9wW
zpear7VrY15ElLgJ`snWUyYDJ9j8SuvIYjhSBsPeGfZBOeE|FSpfotZ1-XDV0w^U7h
z$UZ^}2H9k-kF<1R<7Kw`%=n#j`)hHMAGjUzhz@)oX<f3!w-!09<9Ip(Ufr^?Oz9b2
zJAhJVw2IyG6;0QC=NtB4i2LOz%)(}XEluxWQr@M+sdHXHi=AG$0ko_rREs1y<-_dS
zJ^!boYHe_N@MY96mHeDl<l01^3iwJcoEg^dB2M2;4be_EO(IE5b~EtMs=Yc-(9h$u
zW0Oe7MAZZ<j*>`wP<Qd?^y%)(U1HC3F1mkZ?Qi>ju~D<ekH-%jtuYW+9P&j(hFLEn
z@+h3}RPhVR&L>5NHR=lVRj04bOglEEMb2vjwtHLWPb<+=j0-vt43rwr-PYEZr(Sii
zPP>~0zAaLeuL-C&D7EG58~cMG62(BtqWWg*<kA|hsw{*Rfd6QQi+aT8Nb3tCP=eU6
z6krr1&??Ax6lZ>KDCh1J9_k3Gv${2qyfZIXy`;DOBedSpu5bb`eeuje-RXm#k;XDl
z1u_-K83ETh-mL2^vaZ%qj6bq|$`>-o(=!YH5>)zB!kX)2>GmN^dzR$8==yrc=e0z9
zOYa;Q>FxsiI&Gc#OhxesSQus@{&Hl-0p}Sv?TZgaMzZSBI#Qt|IOOX?`nK_b8KpvL
zSZ7)`R+&4F3N^$LYsqsVi~P0j+;4Z1D3VGv!6WLVsTK9rqm*(@Sf3_Wg*nRHj8>8a
zAlna4&pFZ)8U|7iEbHH%uk6>)yU*5s+8gAYHJ&;O<{H|T=WcmEWuJ`{3k)5zge1B|
zFb$MC__!Gb0p~TPQyn97NQ(WMB^|7UcF32z%RT3nUeUZz<B?<$X1+Tk5Vp$h^ALLN
zG=Ak$%gi@ZtI2B5b3gHrwmX~_+w?5fq$y)GWFV^pz44xLw1|`AsWA1SiRXY|=)QFc
zw7ud&#j5-yFDEt%NmKL=fUzPT2XZ!weY2GuNjdMv!IwlWaM7MuO&D+SG3K87oQ0;!
z(GuGN1^M{1kt=3~YVp1;74ypG_GhKcY3{TOSiSdnI(~&vDa*l+!EWh55vBFqdig$j
z=6<AuGJ|Ysw?MG2IVTgtit%R$HUX&g$5(rfKJZAv5k<$R(RpJ_&mky20VR5HOAoGW
zx6`l%!QQH_TWetA&|&(QT_3-x-h6Vm()a;0J%O2&Q6&Rpw@^FrN6~O9h5mh=EpDF^
zP$iqZIG#Dd0aLD2?K_o-sBEG@Oge=0!Z$$*2YO#@&(I$SGbHh49zgmuJ9owcFNbV!
ziVNw9RYXT7(1|tJ8OJ!9sHx(s=9;K?v_QT@=CVDjp&p5A&0vl2m@YNLx>Tnwq;e*5
zce=I%Os?&g`l;d?+`a8SpZ0~39nzFD_VAR2dvj9!E~8us5JGdOYbl9W&0|Pr>jB_s
z*mT4?)q6ZGC`n0x4S}|!$XW6ZP6ORkTD^k!K+RltI15_#Dai;^rH)qk#{F={<TW<p
zT@}q3qCS*W_Qt#yZs;Uw;Y)HB&79%fgohIq4$y<C!&RPcs0zo8wAOYmZ6ct0t3(T%
zgmeSaLm6Fst8sw59Gb1A`jlS79uj$+rCIe+l?WOcqp=-pS&y90-jOx90WBMceuASB
z{w*oKqDt$UdEX|R@uZ2#qPUpM@uAT??Mw~f5$3dkD78X(L|gJ8L${)yFaL2kVB6J2
zmtRexlp!D7l&~q`R6tVos3kk!C<;FgXij92&pPAf6nm9YW5o3M5NeSKC~8ju$k8}*
zQqPw^F`r+5hsZqV4#b%LLGIY5AFgO9;W0@bV}xBqAvm)|jJ*t6pG0PRA>R-WDaJ;}
z=kfW5bPcxoe(@DAPbE)@7$iV6h0l&^dx_tT93@)yo78+b#*q4~Ay&?M_R*AAhK!u`
z$g+$T=3dGOtZBrNXOYTs_1l&zBHO?Lfu2OI4b1IleK~BAo1)UK$ipKchk1l{J=%Y=
z*p8&k@ez|bwXQXc{kjpu1Ag7Je?>yKQpd`VewWc5Z$ZRP@&2eqFSxDM21J!o7%3ku
zvIs~oDi0;WPx-JZ_2olg)Q2AjlX$_v(eFZS{k@<Q$Rc@6X??puLhzxFN}OP!-7{7E
zUpAeo$_rXuZ>Xj{BV{?%QR8~hQdNK@(Ch9>Nz#WG4Ja;bBdqbApv}_beHCodHV;X}
zGp)%{m{jM6eo5UZj8k?Q6fu2oz(OZ0*MwKw#CsPUWjNhYj>xU(q%11SG*XK-AzJ8g
zgv8Fe>mu;DeC&#ftCI%^>O|S7YCzi9ssyr4Wuwm=F0kCj(sf^jYLOQheKf4%=16OM
z5ee2O>zJxZOnT|%JH6Ym%<NMEDF2$tD)4Mp6>2x2t`pq6tE-QFXv=qrnMo%pxmjXW
zFOkz3vY%w9Z$C<KLW(23%HxioTi;rg8Zoz9yr^}aw>r=M#z^>EtuV2SsCSxyINIQ1
zMOY{M3&@$;GHsB&ptu8CVIYvQwV~g=XDx4_^RP7Gv1rXaXYY|oK&l0bqhc$H7vL3X
zeb(Pbj@|pvQwMS3m?o+zLd{lGYgaikroA!&j;Pj0^@kj}b%6<?%q$n-qNL8!3-H)i
zgYA`U+h(XH=s0D4bylTK-GPIj%-YtffDxTiQ?`r;q2aKcFH#}8{N9917}Ij4QQi;I
zu~~;!b}LOQeuEdoS45AIrEgLyv$~G+A|-}^zCjhc7WR|oBk%{QwLFAMTDJT<T?~a&
z>&d=QA+WiRNly(50fcQ>!bw%f^hXPIFHl|qYtd<ri=^ppiCPVy@6gx!lj&AqOMpq>
z0Kiu|xo~`!Ni0M)#i|n@%9iHliQQPRIc)6_jOgU6@T~dlW}E$k5~lY{d6{U#+R;-_
z=j!$Sk@lX?VI_rPyNAUCi_c_<>gsqfSIpa}gB|qn&QGZ@Exr2Ko5636fxUF!?90gG
z>}Q^N`W?_(x{e-=aL}S=vq;5K<(-6Lt;)j*btICE;y7+xOz<n^anoC;b~|5-3ePx5
zvk*D=8%!e-E%o~lsZ9C-uhQGuvOo_fY>*Vu>tj?K@`|2h3{U9pAYc4d(whGPW2L`J
zQ~y`$ad$j4Ku75}3=*!B!aM$N(M6V?^h37aujwH|vqk&AI!)FHW1*@HA~ZF1b%`CD
zpU56N5yPkfHULHk&KlfN6%Ug4ACyEM(XWk@Xz0uc9T?dQ8yyNzYLu6fiEhW=)Brh?
zz|(Q<bQ==I%YK?=og-Vx9=L|KV+9j(VXY&KTIK?!2Cpm>qWiv^lO?QN&<`XilHk0(
zm>!)ry@(8bS(j7Qao9k_mj$u}!?b$(G6{RRQzwj=a$8afx~osCb%}=`02+ELzpDk)
zr)X{AUB9w*hk`wVsDod7krdf+P7845r8tGY)Q$qAtzEEvKFjA&#AQ-sws+Xw_k2TN
zwDvfWo~Pzw!`TqTXEpn23tT86fV}DDOBJ8jIpmG})%o2lUZ0>^*`!1iGEAN8N&C$3
z<*Zgu$qzy9K5);J+c)(8;&h-XmZM{=m3KOW+<x?Dtl3LsQoI<x#I-x*Ank}9X-hKo
zDKzAS+mUoW{bIX8WN9&$Vwu4mSwYU{ta&i!e5}18BNv<3!~3i)!L1=eLTB2kov&4N
zhAHs%#dN&>>kEF!7G8>(cSWjZUvU|{)peesuc)ZXzOvqJnL@;-%BjJ@GGWvM_FB1A
zx?IDB%+SC5-jG`vw`6Pg>a<t-=pOA@`b#R?v0Xm#jv25mBW=f6IyD~*TQN2^7m|l2
zH(vTXIr?RgkwbBEOPH7O%&};AlUqLd<cDU!6PA%#<+0gH+v9evzV}g8jQQ^is;Ic~
zN|Ge}82b%RK7Yw%1*-EWrd!rw<!ey+yvd!!di&#g@`U@df#n$XwcF5&<#~oz6i7fS
z##GpRtdeJ%X&g-6bg2d^+7CVcWN$Zb87LiA0}QCwkD=`};CbO2<eBe3*-4Ojv^R_1
zf$%F{7RKe1A4B2P_;?aDKeG9P=-MU)P;?!pmCHeT(OE(iD%J1z84=&&a~l-t?sNsG
zM$vJIE}$(KAXYwUbdqEQL+c(qYs(U{gS}l1Z@J5oZEu>L&|4iC!1u0CT`*|b`t@x6
zg3_JJW)fzSofpSIClp!aS<$7+YP3c0t)XpcH5ikW*Dzheqk~z=Zc;&gjY%s*uU8tj
zTC#|@u^A8i(Q{O<uQU<FA01q{XA+m3()Cqs3+n@lljYKkZCRE){Q==Cnn_VF&8SXN
zO-x=muL11K?bBDm?8~spCytT?itm@LQgN4NY9LzQ8>@UJsYj`XogU5j2bl`%nBKe7
zdEoc5*Ht1VqnE3)xE2|#FbZEEiLAa%;Cmu~Cr)1!OwD(1CXO+7-L*x^lmCa#0ss#9
zrn%c|)*w|XXareGX0nW~ix}O~nRSz-WcT3?Bu^P=nInl`dcKz>Y<T#sKCe-Njk|=T
zl}KM>EK!BFroR<!$YG4$58as9jGp(X?!p>A53XgH3);u1S~g<wuE;6}81u4jOG|u`
zv0kDRl_m3G@_$7&jBz%$Kd|UlDkgi{3d3I!hF!R$EwsEdQ#NkqT@%8ERTey(XoKhU
zmL476UE1z8vPioWjm9{jEMxctNwivt^V}BwrK5=DsWe{6na_D7tg1Qh!W??JJk9|a
z-5%s_t%w*HS8xpisHSg>@w3qth1#!j*g=p)41I?mbp+Y7kv`GNBe{z4!B5C}7HMlN
z?(tLG%7Z9WC%FicMI9$s({yYd(q5;!4SQ4?T48A>SP@YT9w_4sGL+Xg*Z1?_q|!NW
z_CBd@(^=9ixnMM&bs5U3B3rGOnofe;7l{54#e#yd={ryuJvhoePjxt7le2w^c9nfS
z`6m2Q{*nK4$}F#n6C=Cubh;U-JVN)#9Iod%KZJ&zul1tr)XLS$4+AcTA7d5`a?dgr
zoJq0d52h?QWCE8YOs4AVBgK<tPa*cj=sn|i?%ja*l-I_0*~~@0y;^~maSef$hmQxD
zGPJPPHU?ab65zOw1QfW+{t`tEhhP>(Z)xsX-~*e8Dtr3QzQZ3nlVwk?G@bVOTaDa`
z0sM9{Jd~^eW5&EBau?jxMv=ZPFu5I_H<J77LAau-&T_Tp7JL4<lEbfjdK%m;K@6%<
z%jKCIwues}o(XK_O-8}Df=*8j@>8@6j{6O)^as=evSXg@Uz`&yYIpkE^{@91941(s
zV*RS4XF0yymqg8rD|FZ;73k^r1MfSS!r8}b{5s!BvvC}zG-PBt0yb~uyNEV%N5%~_
zbh(bR7xhEeoyB-5TTIv3btjvZX6*D-s<iFhq7}4s7m=v07q8U>3G#J0nS&E6+e2!c
z)(dKlc#vr_O{KVf2g4=BBq59SO!KN^xDKxm8Ap^Z`1UeoWvAYD&Y4k|I6b6Nnb(=(
z+@){ngHsR>r5k-=wGiuNtt}m4nF@>mv6|+MKdRHrDtPrn(KeI}ufc753INok#x8#0
zhvh=qOg6GD&oby@P=95I+rr)3hvewm$y5QB1dw%XF&fSrO)Bb+c9ubl16f@=QO6we
zrk(f|OqdIj^&A?(r3R209r$?O=!-F2H&nmI`{1Q`-8lSLE>|DeBa*U{cp^6SYNUj|
zlEtyn+cz8Z-J2!~uGzjjaeDC-TyUde`a<C9UQYLZk6|Z8mlI_yra<zpraU;xL`RX+
zY1`KEsX|Hh+jCC<<c*~y)$o2+`y>3on<Ih7wCystlv6lZKH^eXCj_J_9VpJSOkvZM
z6V8m}#;GUNnpP4i8F>VeO&k%*hZw*GU(Fh@7g-HsZiRo@3tf%gXP(6RWHMwmsv(G-
zH{}-m{W5lhcp$99B-}cA#XxAe?`scURekUMQCogBOTEAut)AjtJ7%ZxwM1EIFO&O>
z^K3o)5kzE7{k-;JaRhDLK*;I2fO;(OWp3yGg-GpyN@Y!5O@+7E!>I`Zpu0?691Td!
z-9R&{Ha@rSUXMqJV2a#4mV8^W;pv_7ol#IeWM`2IMtv`r^w*Y9ulzc^CvV5Nl^Vvz
z6{>o!KH#!($TrH}7_EK00#O?fCSE`@Ttl3TCPB~qaL>4R>D@2k3zMIH;^!ib@2C}v
z^y=ECXh@H28;=u&(kol(At_XyDK3gBKhQ>$eBjFtikPo2)YX*P!IFP50~8|Ehm7Cr
zM0%(2`_)9wJNQUD0}D35N5#f)-TU5{8$2r{;rw;Mx}4~_V12r~2^zHeIx04!ZJQfT
zR`S3JKX{?Uzzm<D)cZYpQ>5MuXXb1>64rm@f^Bw{bFAPU{)!vwS=*6gmZ+Yh5l1*k
z+7IPovE-2h3(wdj;W|pikw|CJG})TCcydnHO77;9sC$uPY1ZO^1m^g9LN039dD1e~
zt@!$v6}ZQ$Z*r6iez@SI^Z+AL!|}2dUM7jC+#kkVySwV=s!KPn%AJbWS+UNgBu4~u
z6F~*AaMTyVg?2+)(xJR-YJSxzbXBz5#*AWuGy+pGPImgnN`m>7Q>~xp1+MdTpYyMo
ziS#(>l=!d=i|dXs^uN?WOw@gGVZ|wOvV0a^Z0h`ZQ`u2UIZSCKpDMIkNrgG`PKIxr
zMAPDf_L^4s`6>6p8gr~|Y2D(HGmR-BS4Gk$_>Iay3+%(2dEbguQc`7IEC3;yD&UnZ
z@WPQuYUZYRGYB_Z#I(@6yQpT!2`n<)z{WhvzDA$on?1ak9LU$Zwe>`y2{y0-dq+F^
ze28TOj}eC_uzvB;=QHt5xz{&Yn@m2LJ$8k0X{!)bQ`)F7qS59!W7T>u^cyEd1;x&w
zm?6ha>?DS3wWl;de}NPe(_(*XO%v{(!r0Y`S765%{KB{`wrlesyiXSOiqI#+@god5
zLKRMZO0(w!LV^6`gbYe=rkO;>T32aXxg5!<kbpCAF_<E9uoC=ySlnS?Zlt$i8o#0w
zR4I}c{IrvcZe!GYa?(il7*nZXG|)`%>Z2Mo)|qyyE1E!kwUJztc5ePUw269tR=J@$
zuo}J4@NpgYZt$v)@4&}Km3}_k8*Yd!HPm+1)bFk38Hq`mk3kEMQo|r=b&TRCbC|9b
zxvks|ECEn0tbItHfB2Q7JL1xhS0HH!Z&rX75h*Y^_w9%V3=~hB1sNx8sFI}E(};Dk
zS6qhM=TKK7s=B^UDZtxbYcC;ig;L`j9@mGkOZ#IwCBA>tuUaB7^S_co{7-!6=c?M@
z_ws*URU<(-eOwet3bL$s%UpsB`(vjsd3gF;hRsVm_%SJlTO?H4?iIqa#FRirZZP!8
z<W`+;>Nz|8+Gse@(`B8Wbzbq2V_+A-rqLUkvqbay*{W$_Q-=<01$9pJp>qnKR4L|y
zO1P~Iqrt^mT}s1a;xVt?67{8b1dhZ~iG#M8oN+!1Zw@!EIX##-PHVtaJf*LPRVj58
z^qGyTu(&<Zs6Hqy?RTbf_Vt<NjWPcvfe6oR&Rv=0>>!;5<TU<BEhQdv2o$w^I@oO3
z6ikImx0>nLK77KMltn4Z(0fdzYhA%NZD2$xjje^u|1Oa~=f>(tQy2w5-Xo9A@oLor
z{Dy%Sz*E#_;+f=mqg|__YSqI>=8`S$<5j}%OuHqEdTFRsAk$K<K9|Gr&K?nw0<y}_
zc%ao|Ec|WYTwPc1%<lyPUV|$(MKW9MVehS4yvDaB^9E>$NfZ-5$?q7eQk-uw&2pNk
zzNp)gVE6iBK9&E?*4FoY=`9DG`|6MStUjzFSKs*#Vl5$>YmIbfb%yy;=}_07H=EQw
z3af1SnrY&E;=uDyd)SwO%R+vuL<*232is~p7LELc^GJU4QN3%Tk|klOksc_lqN`OJ
zvA-Hu+l-!mEj~F8bfCtaV-6msX8t&U0t>eBV~dUS)G7#-gEkJ*GD>CP;e5DR&x`C1
zqh;T0Jnf8Bd&uBj`24#(j$`=%ahkN}r~{9jd%HlEQX9XZM0GcO+*_mw;Cge{h^gi4
zB~Hf%+2ez^9XZx)ZmJaq?-yM%Qjm(;6|nU7ipm5Z9U7yZ(w#wodw5MG<+Q*j9WPXY
z2+O-y-}QPq8M9M`Q#?aeE-!=0q`^I1&f(U~%MgZxm)hojU~}=7Er!B-9lBvmPm+5U
zE!$L8?3I~f$mK3--duU%yqXeg-4-vqc)O&=v*n~U4HKD)Ma;a656qIug~Y_Xy~KXA
zXt_Qkc0xfdt@`|uf@kWGtNajTyJ~DF^V~vc1jZhVNCVqyP~PUU7PjgHF`w|{e5X-V
zvvV15HURwMd6hY<KFL}i$8>u$?1`$q833v6#7ThMS{wbR<3o%o#J!m^=UDuQ`0;|(
zNe#+)tBF&eb4S5X>Y@hfEq;j^?VBkx81>NPaGTWy&g-L&>Z!~`4C`%IiC9+SJfE!?
zYPw3^MH$t1=84bj^l-iyV2qwpEXy<3KU%%+s@mmize?;-Vcqn751k^j5b=>X(bT&X
z&^%C5CgaDkbeJCv9;5+J_bdi~rO-bFQi-8$fK2*#<HL58hWZUhR>~gtDJ4qxz07ng
zn7kH+>kt-?6>+Gq9^uN>ST$K7Se*Grr+OV;`=k#$=7iBuHd>67u!*rRIeA41gp;^d
zN%IM+%-GR%sLsAgZEh@UDThwtBbgMr@dx`x*b*#Ro8Un?-N+Bh#0WC{&uftC?@6D2
zR1c~2>IveYc##b=U{$dWU3p+_BG@J|O}iEe+F>!L4ZA8#JSWa;-8SA-xv)-nvD2t3
zEU$IlV5puJwf%Ihzux?G9N6A^>^how(iNTWeU^mMHxTgF*Mz91kP{wcS8fR8#cJge
zr4y+$${C)M;#M6Jx+d`<J?70l_(|M2C1K=B2#TtAoWIPyeC`{=YM>ZDqU7{UJ%&Yr
zo}sWSqO0$%`wix@;_ACvv0eJtsb9xAEMX%vNi37QaFDxNQN3MB^7Z3T&mij6T$90W
ziw!w9*4jMo;AOqY#`qckce>#pDEP7hjXV^VVrOd$bFr7~jBSw-`KdNRm{l-}sxQZ&
z&5W=rKy?yZ?T{UZ>I|`;;eAH6u%(F$J5Ic4fr{MsGYYPImi2xRqdpnZLU^2m#l8eN
zxNRh8kO<r)H*7Sscay`=#jos>I2D`(^jZV+3XWc2r)iVMK|_m+pdYUVXuO=%<Hj4q
z7rwDuG6mBVxWAjwyMmH=uDozBT7B;3&r#aqakoYpbm{H8CU;89B^HA<^|+8F?)?@u
zQDjBrz<Wr<^BS#)OjVcKs%eXK88d)SX>FhyxbptSp*6*?<Lu!;SJ8kepLZ+5P0?fP
z_D1+)q7hbk)1E21jzQLS4%(p>Uf!mRlPJ#{yC<4V!|hEmr0c~y+;Uqb+ppD{vLmf1
zvv)+pm<eE6w)yfkk=YK`TGhpB$IR?vVXatS7KQ-jsgTL9r8&;M#1)We14~cTY_&*F
zG7l<nazp_ka&Zt3;gXBWy@#R*lf>K1;JM;s;VT`s;E)_s7PD4!o0jFdO360>zrLuA
zP$#@v>s~<gII!(sWoO@9-s>ed<2_^{M>w269z%F8lXteRET3?+)4cQFq_@I*1-cpH
z?<972ki5sPElmRSDXhIJFz-zVkgTlN)N7X{<F;%q^vCF_S&V`*4G1-d7Uza}>VNba
zd|R??=L#F33E7F$tvYpKm*}O<i+dthEw5s?KMt4VDFdp}xZe~!Docv9s>&XqD~EPa
z_<0`c+Jrk2Q2}6XhJ{k7Q^?!zcS<PCMulc%JQ_cCjWjBMh%#Babp7TFtxf6w;b!^h
zYONX7a6weO<s_D<%1^gO)c$+}=5_i@6>U>ZSGJa30)xdpz>NNx^XnN6gnAK$e$V#O
z5}8<MmNif?Qz4=Vym3;#Sg}nMYQd8u5KJ8-FCw0fU`~{e_>gu(8*3t{hv2|_oR7I|
z4!cCV9^CgE)<o~dNLQVfL)tpLS{Z<BMu{OwACH&n)L&Wg%yNv90+_;iJ2<k6&JKKn
zWNHuWRt7trp@n;_a|z|K55*7AuNN<+KP)gmyD8c%Z8oim*7`EQ@?%GGB%S@M*iJN~
z8r-9-KCWOwf&#ZegNzZvW4d~}nTJR<GOUgFAk)ZtTprfAx!o*2Ik)MCv^!8;dbPS9
zWp(nvGF8JY>G_7UGO+8BHhS%X4BX>YAIzQ+AJ*!(qG(7m;jCovsfPT8Uz(6V&$eIS
zc4R27Q0q3B2prMS+oh>mx#i6tnAMREwn#uM*(qk(02LK90IYO8Ja=kdV+e;IO8HwK
zgogzQ8r9-9t96#+Zfzj=>V>!r<J@Y4noPgx)M?jpy4d%!IY+Z(J9USCaS1kvAx>Ov
z4z~nD1E+lY*d^?|<Fb9FJWaOP3J08wy6rplHfSJ<Mv$QiE1;$6>!G58yhH-1r%?oJ
zT78s$hvh+KD^F^e;tE*}F!B7k(hJY2**1z0yj`b@Ywq#%>0LXceg0;HnG_cY2;r4z
zPNJh7JY15_jN^>j^g*r{(>sf|U3AwlY`}(-9XLR*RBP^;@ZLp&S3sPZpxpz$J=STf
zfd$pF?$SkToaS&ltSZg67-}<a9A2zX%l0*o@_Oix^dN=8D&$P`Lw!aW<Q?+-Drs*n
zl9osZBm<Z9PN+ghbj@o^ndLPjmV~ZltQNF}%#WVaqee7!q%C6mlrDPKE3#fj3~Nji
zB-`dP#J+z@X?C8w#|ta;85m#}+^O(p?S=VxP=^z}A*WkDftJ8yB6Og*a)Rl$ZK0c>
zEMT;5<akHtUWL~objgbE=%~`hn7xAmdetTr{Y`wwNqzx4ZB-;7w?4Y`QnI(CK(=DQ
zHC*B=B=i@fTp3PEgYccFWA<`u)S{|uwy#o=0VFD?3zZ>x{9Pq{Q<z`-;k==HWV0R}
zSt}&K%SM=$X|&M3w-WFk2k6J<w4yv6uz0^uH9u4BXzac8jt6*WF#T+nRsW-A=~_0w
zw{aAWJQpwgON<!H(LBRgy03pY{ahkqseJt)Q^-=#HhR4@jT7KExgjCxJT7#4iqQSU
z27^;Dz0Vu&ugc5L>ColOp)poW2)fzNTnGqvdOfMCT~kB?uT3mA7bI{KeBrVsH?Qr$
z1$E^gbNi9CtQ<(K{cZu`#}6-Mi_fO?ik6DBe8LcFU}+bHYSZ&g=7n+2uiZ}{)mG?U
zhdbK(d>WpHRF-ddP&k9)UJnDk5k?&Z3FV=B`B|pA2?x2leKDGxmroi?Ae`B#Ivk|H
z{RiS5r;43x{l7{^ZLe|`faGd6suD*T)6)eWytBm%e+36#s?oELH_QXd?d3}SEamn3
zm$Zc0r<3f#PsbaB7X=_@pqz$BhxOM`D<8{QtVP6fpM9K&wUk@4R<%>Z7!$pn!4po^
zl+|%hCyODwp|!Br9gT;Xg95bqElzSEC9>UaalAUy6Vy<Hb1KU;NrIWMwisK35B)b^
z3d-(hoOFV$^>ZonYS1lCmYt_vN>gzUeaQ$-0U;NUr~5><ICF>d9^uE2lSW}6>lYv}
z;UKp>U_d5lbzs%Pg-4uaY*rEq?<&}s<&CAPE^9k>5GNZ%wm6G%vpCe1O!XHgml}qA
zD}{1o&nQ=Q*hr(6nn0KltNhu_*#xK3g#AHIC;Q&lTFoO7)LX!<qUn&r2O9hj!P2oM
zdr@?utbLXDb%nF@Xi1tv8qT=Ha3nJx&0y8MYnRmDH#}5#qLLnN%mvh|nnOjS2a9&L
zW=EA0^lC(=q7{~xA4K!2^NMwkTgr$b9!!{%_Kcf*eF7>~2H&G>!Y_3v3pTs(2wqIa
zB9J$UF!dA=ln-7cMJeAUFQ#tahjvw0KX!|W>*wA=hl|WR+*n?woaa-*5awj;_`lEE
z{&k;z=FflM_y0M6#=s!S*MI-|d>gZP;LVIKZwZfvDw4eu%JYmVA&D_}(NG#flbwC-
z<mQ{ObmwX@84vrt>@FqE!?yA+qhB0r_f?!Tsl!pZUBmM(qtZd+FX{+ly{QO!=;^la
zxWC1~=p>mE&ec7ISw5-PEqJ2OhR>=jOpL&f(3`8GHM~%RgLz<ESaZ6HM$GE8oHV?v
zIjgUE=)s5)8{1sk{FIjlV8n2|z|<p|_^kVigs(m_gt90pgfiS3Hez1>-mp9ND?-K;
zX;J`VyHIoCKo9@~a95Ys{v@SX8><|qqHI&S6a!;X(y^H5CtyhE3wax$*F+IwN?e%V
zvTx}4sL(M7g)%#Gb0vlyVaJL*0%f;_WrVsj$H{gfvCHpvq3?lt=7g97gGHvOY)4sY
zD4bXf90aOUbkHfAEs!LRuZWeNkpMzQMv6^OKD@)*%E3IlaN+TOF-a$XVai}hFG^C@
z*20nHq3haUaalO0X?=dGW-#rz2ndW-YW48|dmRJgZXh5JXcA?j+*7e5_@%JYYfqTY
zT^<m;gz^}%5SgspI@qPv<u+wp1M=E-8h?Vrs`yys;&{-D*2hVY<0o-_js=g2lM*yS
z8CfeAe>D&}XGkE7RclE>+h{TrhAJ|Xs}qJq)RAG-a8vP(L`t$e_=5JodeEX|Hnv>Z
z=A^`KpC`0=(fQ<Q(dU=UV~aO1JC9tD397K~@`dpi3~?;wu*LIq&1(H1LFt5H1r42R
z31u}SAFYFLN=hWhez{)ilo}fFSS^H>E#tG0u6W;;<T2)OQ#X|%(+Lz7vpkiO!Q+X3
zem7!GHsGs37U*W}kQrYdBVss@HXf2!jdsC|d1UzevlQ3vr{a`1zLt|#)AiA2&epKw
zv|9w~b}>L$7_M0Qn#txuPjb_24Wna{=<nE@Ug+nGcor#t;akJ}nj~!J3@X2prs>e;
z?j#;@a!yb-y1&d}cL6UN0mhd>)gdQuo6$D`))`}ewEpC~HSl~JJ=ia@50PxQgPlhE
z;x~-TyawTzo8lytf5CKb^}V=AGrdwWV2ZsYS?9jU$6EAz3gr2)Y?yNXM*#;WTrHal
z#a6UaG1~4#?jtReZ*1(-+7}oGUOE)PYucck{(TiY?mUB!n;QpE)`lRMLoCUp5cJZs
zijJiP-1$gZ03Z_xWAGz>79o}vK+#ztQhVeH$K~W#d=!zbxpe%-7P=UV_Ltd0Z!{!`
zRgWP%wl9F3DcnP%w_-(V=7>l{yc0Ha%>ig8&?;&0SbVRocmB1=PB~n|L8+ufBTq<X
zF(Nmn$U;d1#QB=%Z$^&4_wv8?&;Nio;J>zk|FygS!yE4Zwc>xlZ~xq7^mh;CA9or3
z-R1uqKU5v;ZM+I-5x*&Z&Hm>9#s>VKt@9IO#%q1HE4}8a&v!5W=}-UCiV~|dzMC%+
zo&F-6+Wv|Qv-hCuRX#-?o<)NTO+{@aiF$dpnEJB#16&C(2VR`Au1<AX%qQP<({qI1
zc6Snq2qFL;pD&Y0rV(dbY{ZZ^9y<wf>FukFVS~7=8eV^)QT+T*EASV&NaAmF&5C~}
z-;A=q8}%2d)4#A9zihQN7c&NgKe^PxXnyfeD`FbidD9u`b-l_je}?ak`TbJ++7*U=
z8z%VX#{_<}!apth7rBVzRY+2zU7=Sn?KjYPeqV|w{@HmRF4-YtZS+n^coHu!5y2p`
zH)GSPupr87x5Gay>msUi57)OgBXdvlG^Z48X6kR9UFe@vaoK9ZNhLW~uqU!KD;d1%
z%^JKZO4Go9urPAH>3{h1i$A;P=b-o*3O|wJCqDjUg`cGPQ!4xvNIzxaPu2W0SNNGB
z{Y)T#rp-Sm3w}<K{G37hIdAsgI$LPBb*Dbk(o#4`KX!QPqbggOv@NHRi{i4;JrduW
z!szOdI+~9yw#iXmq0#Tn3L%Tamg-8M|Jg(>7U@jDw&!@}oVixO+tbey_;8bPrTP7{
zfIZx|mq%~^S<%1BwRVxzYr)6AVX(1X+q^gLX+9_RU*Ww*1{~y^wvI(Zj{hq4{MXe(
zQsoKy=p751Wwd8u)T0Yq^qJ!YTbc8ktHsxdGN_~*dl&7~e^xsj?%LllCicrD8#>U3
z0NvSp;Y$N*E=_1&LLUJ%M{(zL<DV7&t6aMi<|cUIz;dIbdolEDXF|XM>vfOyCHlSU
z75Tkj7!B{{U!|h|x_bC0bKUTCNp_=;zM!N%8o5EKjY~Ea^y^-NO|<+s(T8Qe_-Az^
zHFx?AqhiT8z$;Sy26OK1nFK1ph%KOV;Wvy10nEMOe^&Odat-tGrsL6!=2c1AarT2~
z^ht5c+L!TzCp*7k^hD3&1OFfL-aVe_{{J6eb(OBJE0UxGR)iE5QjVMJid>aLLJFf0
zA;gf=#&vaYN=vFawG}l?$Z^fFLUWjwmh&lNM4N1wu}#0{x~}*8^Z9+=-|zL`@AiHF
z-rwDBWqTc-ugBs3xIYh%$LsZi`+od)k@WwZ9O6SBqDyz<(!Ev<k4@)`NraL(v|Uyy
zf1sg2{IO!sjrhM4?a<rtZxFAyb*uBc@<dYgYv$0T)ta;5#pK{g36WQ%MxXxO$^S0<
zHfVkvoqn0WeBR@Wg`Bd}H%M#Ik|u#O2;RH_ZBVX^r5F6WNcw+Fjx{S68U-DDzCqNf
zqN@oEJE0T8ZFPlU;K8Xv+Y1YQe);ilMf*?4OFI-cGD7Tl<&8aw>Zag%6lu=04v0bW
zH^>1*tl8yd!oN`w-QW7+|GS;*7GB43AO5mxgB6U)5D9z|h9X{s`!^2yujDPAq5ZM-
zZ!g`n8!N`{nitaiv2)4`Qm^?%E7Pv42pRF%3n=F9-=6w!vU5kiLmZJwU|otU-eMi)
z{0*W~Iz)^k2d_F}1kqjd@W=lqa{iwaLu&dDaD$cBbx}5y`vz)*;5dp=!kO6axa#gj
ztJ!CnKkWNE+1A`T3Kc3FOf6+}LqC~)mbuPxUi?57VW7gM3vltCzdQBcWuNX@qQcP-
z56^Dj*#jEfe;tyX`+p5dN>zE#mlyT};n&t-TQOttk2moUdxf{w#zPLaiV|jbY=V@k
z!hYO|e!LBo{kmICAg;dplZMtuLE&z-7|D_U<MUr9ve|P2$Y}^T^scyRjrl*0lb^uY
z<p)n3+9qjnK_I3>)8GEL3qrsp`QQb{Z`b@K0ob5}u?tJ#W^DSi2_SVOMO?c7ub+Rl
z3$~&WgccST_2;GJC(!kH85uZTrdlGTM659uUxe@b=V^tqg!uS>vg&_{DX*tDcUMYv
zPK=ml3yk6)g9~ip!QDh2cCBmU|5Q)vABOv%cKh$U6@f{WJ%ZafpovXsI|V;YFC5w?
zvEoYTl%Ur9pN70Eeg(`c`ABy1MDAOpfu&I$pO(yx{xY4!W(tTZ)?ZcxKaZ`1ik|+f
z^(Czw@V+#*g7K*O^p_!kIyeTI2YD>f1EB%F0>6Tf#E#48#yyfB66=}7?ftH-WZ(Bg
zzas%?gok?s1J%E8{{A1>MhI@16~GTkB)$6Tzr;j8!b2f4j14;HQ<s+m&I#_w{x68y
z^FN-b$!8jg%u$xtH;DWG3|{G~r3b|tyy9(CR3M(4Gg4r1B1Lc&E?8_h*fBl3>XCn0
zN`=igg=n_c;o)(??Q2@=Zq9CzQsF637I+FT*Am@Cdn7w=&PJIkf0Arx;3oD+TTZM*
zV18YKIIcHRxLf}QGU5K~(#CaZp{z01r%BMp_qNtH+U4Uo^J{D^Kk=i)W_OVI-V2}w
zv{9%C8Gm4HPD-o$7$g+7Hzr>3`&U`nNHc8Wy0oaDhW5yR_x74!RbXo%1&~dWqi%q9
z3AUr{7dBX6w!)-fRSQcHBZc&7(JBIVdtDj^fvtSCF|iJEuwz-I)S&Zz+l0Fvj7ler
z!(Yj-u}zANT^*g6i@CA(F;M3QW;A$w&FO7=s&jWY#Y6H{Caor>xB~Nyi46oh1#Ht8
zF9U1_pM0S4OY?)RDTInK$OIBkz0Uc^#y^D))rnn38_kQ*j}Jg3#{0zL*Wup-5Qpl#
zqf5oFON)U|ubu|#029rH3f=84um5qMJFH_?VE&i!fEX$&bKeaCRHTHm`e!k~!V5qx
zm;yp7aeWj}lD!;b`uM+liD2v2sBmFT8uUgyXb0nwUww(T94USUt^^FuofmoBt#3d+
z{!_?OmAS|Uf<oCRZsd2no^Hif3?TBE{JQUc_-x6qa|ebFh2@iJ;!)`Rb!q4lH%erk
zsYn@!KH&jHkUMb9HOEvCKLTt_(Q(1|fmC<vOjk#7gWz)XY37Q;OLz0NDlmw|Zhu<s
zhkGz7#`FSPWkT9=u%9V040Qm);@9Vady%1opj|s?Ldz$9tg`ZZfGF72c;LPt%>RLm
z0$@TnN<<U@+0Tihec!z=(fOYc`pW?lXh6KJ#-?TotTpbIgMG*bsINq-4xA2s4>%e)
z@BjqBt!;rVy4fdjk_6#gz7X9Aa1#Y5EOPk5*gX<htwgLkKvZBF3<8{9je#-ONT_nH
z4)O*#o!mdSD)EaD{fx~g9QspYi4W5SBEsExM{qY2_)evBmd%Z11;1fAz@8e6=|K)(
z5M%C~hK7%_4nZIQmA`Uh-64_?LL+uA+hs_aU@VL8&;u%(-F^qLW0&$uBILM0v_}yL
z&FbJr{D}^Umau$ZN&L|XI-~m~u9zMd5!cp%_>uMqXyZVh$SPcf6iAe25Lk$?S0Ft9
z>7@m3WUVRT^18GLeq=0Yy4Ayu$qk(gW&aoW=2?R*p&QTz$TU9>$9zo~2f)>v`ZUKE
zzzI9!zlUOeRR<>`%ILd#tMv_|!8g`N#l{8nfi<=re@5!s$B_L%Rv_mXcNG>e&5qpo
zmwkd6t9Uud^)wh2JPFUB5!3ygh{*3ul0FSk#)g1skcAX%+9@k&B!KtQz5!17GnU6d
zs7OGQLhj(dGLl7um!=bH`JzX4q6-o%Na!hYS=<Qh0-DWknc#k>4!~4)0TM<5kl;jw
z!e9e``OX=4y>AY)2Q=<sM_#u6i+I3M6Dz1GfzbTB75-#5TlZdH<^*6v;m_Z_JiP!1
zTR?Q9rWRJ>j^3q6PM%(l#dxfV{|8q|{P>-0{vy<u<`QpC(YZprgx)4RQkZjpMT%ky
zP6vrM|1)+cR^x&@_{+f75Qr2kcvx?aA78T7tz!2)uUEHY_$Hl){g)X?2lGF^V{1Bt
zAbLSr4&NXyhyG{Bw%9a@HV5YVzxyNe^~Z0JmZi~&NyPtZQ8esV^S`stf5f8bPJdk#
zYA0B@IEW*Exc#q-8P|Ssx4YfAf0<z4<*XBn1yg{HQkL#^`R`YAZm<7yEa{$vOPQWn
ziHvm<LX0&g5G@Pbh)AKiu?ACBkH-^ygJ|HI*FKWtK?he94ncd(R|!<nkSJcW%Q?Mm
zjJ<+3;&CRV|EMF_<qvS@Sp#dj;Bf(&0uxiX`)F|61n+uoEQPlkm+uRM!Cy%y&9gc<
zg8KL13rr!{R}UfO>xWArq!5^sYq}8mWHxFA1=}KJAv7_X=S7GSTfR%n6+f0v5;jn;
z5AtGtnUK#1+qgIu<a2?@-EJ@1jBtBhE=Fb?=EcKtAz^l)*W;VS7zs=h*eEHB)f_R_
zt6zX^U7G{u2#A1+em?pQnBOeeO>=l$-pbA>bBDBLnW01Sp9zd<?kc$-^0^4~LG<PN
z1|aULhXPD0fl%!c{9@hXQjNDB_t)jd*(n6;j0U%K5D;57f$OowX7nrTqcs-z%jkwt
zjYI2BrnJxpIICBslfFR~>cO~8P>+8loeWIW!bCi|3xp!r(#fsB{^`^XPJHQ~K3em%
zqv;dB7EKQzlkaW{MQEfD26*vdcI4>=K9)>=wYFY93fRL!+z{6}wiFAO*$cFgr?+tC
z*`mhpf{d-iV1AtfW6vEDy+v?4De5P_xv}~sI*u_ddZh%;$8u4S&xE&^SRyw%FeD+p
zgBxQGc9=r#$ozO?DJDnI2;H;x1fjN7jP3XavD1WvIf0J5)QilV;{)q*1=z}tX17)f
z;RSG*R}(ifN@%Vm$L62P3{^!idCNuuQ7m}wqC7MayjXsGjr;F{S`6MJzm}G$B*&t2
zB%b^RIZ9cM144=jw%#7MJ0!ZTk5QI^wQxl4t(&&mwZ^4z<4es~@$RwpbSOe2_{G{s
z@?hNZ>k4Ymx)aMkui{~_aX!R&6EF7C@<09e7b*yMmxb=I;0|`&jy3xLdVMbnVjFq~
z=L9AI(+evx9+z81N{^M~IBfX4Q5Y7d;|=lC<D&TWjo6CmKA;rzVnl?<9hMJVjiJxS
z#$Bm{+g$)Qf)j$%L>AxxDi<wNfrHZ_QWl{7rIFrw9&3C8SV7c4B`<Jye!miMhVm#v
z0k%=fqN54CD*YG&br*p7*f&TigWNVQL75=lAs1siG75$+1ALRHj6bKk2fN^O-?oK4
zl2{p^6U6P2vb1EbBAO-U9)`)JgOQ#U7bWt*P17hhqc!+-^^GYwL<;aG6tu&QT!lfw
zxnfiI*YjK<1n4;zqfEzEw0}RJAB6#}x6puUh+r<dG_2z_M-nA)cNkNDC6q<a06*|B
zZ;=^ZdDJTdz`LnqS`rP|yUw|B*dZ9S611~d7+dGwi{l8v09}+Y3jwYobb>ZxDPnj6
z6M$BJLeTLYyL~Hhh{`qZozur*5&@X7;K8}5*w|%3ys<_{KQB@gmtN5hIibH46$Kv+
zhULR!<BbpS&}OAF-}QY>f8#C#OLzrHWmL*HVR{z5aTx?4wZBzpHi~NhXT$K=`cmv5
zixY#!Zvo(CP}$3-4ZYVZUm0&e;_WnoQ--G|raB0~qosSrh%W-stu<YGD=~_zdUKF}
z2rDzaup0CM9$O1w@`*Lb2SfJj6`U9!Aht#Vhd(c<t6M!JolHT2alI551ZLX;U$lM}
zp!|@81kiB&I#m!+!FMTqp;as7LoUD+XzKtNqCsC^=#p^B_f78rpFwa~*~$oRK=k;B
zo6r;-c>$0ga0?DbrsKZ*ak`(s98({!B)0-h=;K9%a`*r$aSEq6kWve<M&ORfSK$eB
zUI9Gw0HL<g3C(NXBfQ3^*usQ~DH<h|2MLqm388kV%Gwz7&Fbzj6kw5wHUHHISS+Jp
z*ox2Y^yam<j(b3ZRXKXFH3|t$eQSy7Q!IX?#`~1lwDras^w)!*5KPWoq^O|}a3)|-
zh4gU_7o{R0Vdm5VA9GWJrDGgEJf{`|yR9TApny(BDxnYP2<Ae<OEe`pKDr5(pFRwK
z1Gc45+0ZgTB7}?RJkV`45NzGYK&Xq51O@CMLW&!Knv2i35Mvb*G9&G9w<RzkFmO2F
zQaV5Z;t8~j9Vt?zSG)kqOL)n5hzn@Rse=Hk8ef7)tW5#*LS{>73-JWC5@Uwn0{RX#
zcnmitN968t?XI!LAPc5|8xYiwjF@Xmz#I=k?p>3}V?lUfL%Pmre)=>=82kHs=dxK`
zIsqn=&<AGIpo>_E!H?brJP$7W6ZI39CkeR92A$DFNtci>HjR~pvt2%8oIMv`3JlBz
zEJD9N&W{CbBOfL6nm)}IDsPehe31NlR)E3N_dw|uBDnjAZwFa2B|!1OS*v)&ih#nK
zVnX(13e2!Fz6pZH#C4BB`_x%UKMOd4je?;F*N0U3QOgix4K4}{uVrKaLCWz(b@g}L
zBy*AlJ8p!aHdc9bK;i|$#MCE_pw28_X?+x2PXwrMB{B-dqT>>91Aw1}?xWY4lW__)
z5{{Jt;m(eML!^Pi)Cm!0O42I;jpLvLlOWm{gDlbNdkO6+P?t?F0PA7%_^+HFmp&1j
zHH=dGz#;Q-i0gpe2)<l?D3~D0qJZ)E{73|W7i)eHj2XC)d<pN<$Ms;K)BKacB}DEG
zLck+{P&PZyTHs6|B_@LtK<EPN#!6fY3nuLOKFkPy&7B9_uhG#oK8NC)HOi#mZb%@=
z6@U&_Q>ff3ArLqL_ZkdE5PQEtdS4HY^J7fRuv?6!L?ei{j{YeoPtujG*5h^J?6BWi
zc|`${U(r7%F;{zRtx;qYUlQQ`gG|JV&;tqv7^Qm!QD`%)F94x4=rl1N06~z$nF{(V
zkqB5b##}lHIE_W07Kn1<f7GD!_=4p~vpeX9-pk)dFxVl1$nDiXI;=lgXkp@lgzKR!
z4ja(nCLTsYk$)QK8zh@L#uBXJr7yaJq18^zX^HBRKS|=!Yh09of~)w>x47Uk5Um6p
z4D_aAg6|7(;xOz&M~390Ud!JggjW)doaRJgbTF?Z8=r{by|*QR3KrBg*p7gK*s^J(
z0*f1kE|GkfAvAj^1P=%$vI@RO5;tbg!S@I(Zh$Ai0K|S@jXemEFg-4)`%cy|h|UQi
zc%<*?FuUn77K<fWR86kCBm$V#LFc^gLP){#8BI)>ro`sWR(v5GzZDR`KgLs#d%pZt
zJoL#tH<a@ot=Bs~3F_dlz!(rt=;TE3=XK110Qrp180zIsAgV={Yz#sQ_Xc>E3jDA|
zhVhfm`M@c}kIog`9K_^5QV~<Uf94_{-uREtMG&9hj)o9CJFR99{Oc-c0&OM!?`A^(
zSL~B-I`z;s663<hsNRAn#|=yiX!K!y@B#<`S<W))8ziu320MJ};2-9HXO91fN#kyR
zS=6=N>~9cmrpP~46no=;)?CH^nYwzCDZ4W>O;$EDGjm^JN+pdZ`oVJS(2k}xJ&&}z
zcUx!SAf0}pDb~ggS9!lVVGDbUl*6f6mxtB%{!smp?YCL2C-v>h+q3ERI~7h<eGD^m
zK2entXa5>J0Psj=>8{#Ju7amZR{ov&!!%1rr{?AelDd-ZqmpUkDt7Izi16FBeNHVq
z!(588ik<YGbh`(Pl8IS)Sy6Yqk_4C5bemlGJ2(A@?~|HN0A+*n&c3X8`gr1ZGy|@`
z>H5#A!Giz%8V(fU8^nf+Xa}!#`q%ZbxoBtrbxhCz<|zF62@m}lV`Vxbg(o&Ec^_tm
zk5gcQ0faI=Wt?9}xUvD>Y2a{jVOnr{@hAHi*6NmdGKjV69uH5Cm!6|u{4MBI+Ai{=
z&7tT0Z1<9L{0<&F>2jt*)qV^6YGtbH8Dec#L}r1ngFnfS;=Wz^xXqCnlDc2=OhwSk
z+``1n#LT3Oqzqm-W#w%e(aEy@lmlb#K)sEVYo5M7bXU=Rhh}G|Z6g$mZW}ZEPJ207
z8j7l=Mh?^)IHGxXX1jg$5O!x<aQU^d6P<2URI5?C)6XZ<;odtel{Y{7<A5hEd-!V^
zL;c+es>_z~#M_sHn8UkDzI3*H;2zzs?xU9D5|A&gS1HVJh$ihP2eCq&I?hfmI@%v$
zmKc4WN#gX53=Cx1erRg{$m37CD<egu;_ghtnwvYje6qgOBDXCJ)`n_zg*Kjd(lrea
z6Gp43`oMIib(52pmepF2>*Zag)Y~U#VqMxvpQq294%Ikk6O~wY{$$Vn?9!SWTkl)=
z9QU^JJ4bDN5Iq9d&Zg{C@c&%l*-|q*LUAV8)J9B`YG#<~o|;K`73Vw0kHginGmV@%
zUS56qB^=?>{wkkgU0o$-9p}ScmZmiDKu{}vS&pX&dv!oA@%7v1JZ;xk0Yw%2<oceo
zCLWYF*Cz+4*ETd*7;W7ZU8bYGKa}>+iSyDsVc#>4Gw2eX!QT0U?NzCHGYUZ$?QD@R
zTWW107rFW~vku6^-8TJ+nJI43ZFjiVO@(bkgE7)0+Cz_gZL5Nfjmvawok#_(d3du+
zqyiGjzl!cxOD{@dsuyI&C{kT3W9;(n&-&CnxHW#(2zHvZ|19psxilZzv%R`3>Xn=`
z?LW1LYz&Hdnt1ZedrQ@<dsZzsAAd1UtzEhrMjI-MtP`u$o;{cqXk*&x=;Ul|^*k|&
z;H=o&Y`wt8=>#?APWS2-`}${`BZsc|Di5<$!g?dJCRMVg@A#nIXw8;w%u#*W?unP7
z+7^tVUjsJ>oqseY=)7}gtm5`BjcuCKtO=cGt+ed0v?<!fLCf@ti`AEJB+#EfP=c03
zl{?c@iT>yBd%HEi>TD+`DV=xL<p#Uw?J^{l1a}7|7got`b4K*on6}8HSf*$r7ku_j
zlyZ;Hlbb0hz333_3Ge<2cN@1Z#dAbfyAOLlG?AHDR3Aotbh@&lGp<YNPzClBVFoXs
zJal4TcKBIx+O8MBD(deuQ{SZ+VPV*g_#xA#OmR8RW)54iF6c|mGg9gFifsfkD{7LH
z6FoUZi(*)!wsj#rhI0LjJ_!~#w>_SAE~I`mxT0vXcrE2<`F<kXV%f<_R^JH}nDg72
zSs1fSIDQ~+Z_)MeW8keKc0Nt0gT?+t-HRT^s%>f&D*IYXOq7r5JZ`%T-q87AUu)Yj
zQoRq~Y=AxCLh)QA5mO$PbU770{E}bEK*~2e+1TrjsH;RRnYbQhPYjULuo1(;qkhR8
z_X^}u?V;R)$pg7H>VBlGymn6?HSejF`e)IO3>km;W2T|!B{k~ainy4y4Ob=%#D_jP
z8!ugZr%r3d4TTU6&#UOG!bfXuyQ%_RNQI2UY}#y6V#)I>F*xnPOx_|+h2R`a0I#FP
zqp{cRdn!{^viv_TYH2AKWaj4Nq(scRCC+C@%oO#wjSqI&xWza!#O@#Jo&1m0gcO{+
zbNT9EV1HNHm%i<CUj`#PwRGNA<==TzFI^}t=c{*6>EaXbJ3i@~tt`7MlZXapCi+q(
zKHA-B*wEqHbXhq^*V@A?Md{Yo;g<5<x-`^l;+N>-rSk7G3orVHS=h9Plxq)Bf(~%3
zhpTMd%X{_Zwym`LO@dn*BJnyR@D`A=U4H%Bonq`$PoI4v-H#kS*Hlm#KZrf5R7ESp
zf00#vLVe`-uxGyb-f&1Yqw~}WC^a=TJu~|D73-M(^9nwm!?$CcM(RU7vq%}a-8KFt
zg~^`&8N^Jcif_9SnqqXWFG9zgHF3U$S|#f=xU6uw!e_|3Mb}`6;&j+S&dJ`|#zB$R
zY!x-;GyFyUo>#{0pZQ<U%q5Es^sAF@KVfG*oi5q<uyUt^t-r(VK_f=s+cwhT2^8_>
z)4M+DpT*UOZM-z}CJA1r=#)o3k@YD%$hl&Cm+2_kN!BTY=@~`YMwOGJN5&pq?A)y&
zJ@O>Z=~dNrj8PfxtW#zg#waVUkRsoCG@qnyNH%cNms3P-6&;>(S-?8mhKA8kb7MPa
zst2RnwYAfOJL5ZMLaLQ<>%E6s*>%-9$82krHrG7Wx|^q`cc9Abr}JH6x0$Snv+_rM
z{Hl;bl#RjxqKZE;c|7YtmZztGBFV?-j#teim`7Jt>G@q}pCvtV%1ElUOsVV+9dgpQ
z(MM94HJ(1|Duww;8F&qH;4Y>U-X{Z&$hEs@6yBr9A?E9PT&_7K_2fX9+OIC}w4NO*
zKe}^T;nd@XOCMHBH#XKCV!S}{4BAmI*q0mvtiMz!a2Rf`lwK-Q$==jSmV@;MRp`ks
z_y_{_@w*L|L@uTNWkKw|TBn9uOo`9sQlb}$aHX04^@M?IVQ6`Mz+If8+l~~Mk4zg|
zZTmq3O=M%qnFll)Z8<aJtwP~+ZG=iDQIq0k5i+21AR}=l7lG0|Pg3tJ%*yo%?9$hV
zqYgLI#El4<OWi`#RT!4cmDbRF&w)X)uep)2@TnU@bBD*Vb@}6*IdmzgX)wK@t0vuW
z+14S)XV<0F0QiH6U8(uSr{c_B6&kFQucW;o?lES_d35L&L}1;#I*&z7X5DAi?Z14H
zx+{?J>qCo+*LB~im~8K;Qrl&DJE-Xx%wV%{y<;=AsAA^4)Ws{f?QI`>T-ATE?;i|r
zb=7=cT9IxLyl>#thVsKnOf+LjHDif7x;gzL+KyqS)Z#tmQdJOXn=y8JqUol-&c~l~
z(`DXz(#p`fMa;&iySAmzeJrLoos!)ZddBVE%)P;lF0{n4g9*D@%V>1yvyPCTR<7jB
zw;kMz9Pd{*@!d3Yi-&K!#d02S80=|0u>L{V=Xcej6E7wJv7I_rY*@Mf_M4W{UTwMR
zk6&!I6?KdC%6p8BmjX*FNgKj%cYO3uCb}kmsXlwoVS7O9r&oTIIk!-^tjUaHD3#(<
z`&uu5w9=(3E-h&tL^1>6IQfr14`3z^TbBjyr)UMAnZ4+LRxLnbLz}#c<8f}?1A~YP
zD5>?)gH7H(moFVjxi(>Ouj9~ANWz)#W>BOym_#k{Fw$QS`u}`6Cy-kI5~<;JO#Ou8
zj;w4|-6Agq+~oxPMaPg!#w$v3do{*SU_VDabdKtMF>m&|r?OJ5S(rbYF0tiLS&gXO
z*D5}2%#KejUU5m?lJ&ve*J*0<qx`w8TMIQdW<R($V4HA$!>>MRO`%%-vMd9wJNnO`
zFh47Xm~xxvXZbl)#reUJMal@f`ZW<eC;B;MwsBh7;8~{kS#_q-glJyLG~iH*=F`^X
z*}Y!7!fuu4rYT)gyFX6#CaW$Gezm!GbW>y|-f$Krt2p2PwbJ40fIQ{rKz8FVi(c!y
zuS9G8kzm~PQ|q3gsD53W^umN-*@T#eT+i&oFYUFM&3B)o?GK;e>}#gx$m*JniN|Vh
zy;nL;*F$SDIKdy>#vKkl?JLq{44tyq)k-<qB4_W}Vp%v5S!G#QTjSW{dNTF=g`$o2
z?P@n}sUW%=X13(Z%@V-#okyWDdSPY+Y|A(|vQC+vQ1NQ7)(K+G{FAKCiw_*9H<V8Y
z6}nuw=k;r6QP{j$ONCM2jkeNry3#tQ<diDQFn28SXRh?$?!2e^Qdq^f`@ylGR83A|
z@wA6EQiJ2X$VMzD2J@P4UV5xw7}wb$5ECv)&CO(8BrBtP`Y$~#q`HUu9SeM-_TkJ>
z$h+jzH*f};C-qcLMBvh74Xhokt-JMQWo7mC^;xpABR=Rfr7rz#&7BUivU2+R11hR@
zG#XtN>LzB(=!ew?{C<Ju)*1Q1$5)RzfUfe*V5TWe_5{A9C2f4+=;Q9!oTcd3n;lY>
zcJIUS<Ca(LE0@eV?4wiHIg^IChiQJxx=!}GG$$K**W?+Zr>7^$FC&AO$qII^t*)xA
z@TqUfvuyYABN2KGE4FpZH8Y&-L;Z*acuzl4K?a_oSdLfA9HVgTM=mV;9L-=X&K{p1
z++JJUP(1rEO0Mf&pj)PW(&X*dhoi0xE4}G`b!b~3=^(<j&&z_SbIvuMr4BgkHRZ3?
z%T6<2?Jd#zY`h^ZIDv4z@chbgLUwAl7s+q256AI#nZX*PFOiD!NM@DuG%d7wZ|AlT
z6u**>W~bk9PZ^Zk5s5vsD%zb7LO<7pYZZUl>W5VMY#4GvC*t1E%TQRO?cl&<u~ji6
z>^ZyE-r26C%Cpg-dv>wg@(2ZcC463a1iB&b<${t*qf?=OpW(HaGmA%qyA*1C+tD`q
zP9qxqW~SE5h+;UnYWMk+B1WhE8(xP`pW8EsZg&f0+wm*!ZBNWUVpaHrWE(twbT@;@
z@OL>=C7Zd$-z3KE!{wBkhYubl`S>TJ!-t}oR7ct>%{?(2>Cnuz^U!bR%DpDenn^iR
zj#Li2O6e_FbRW^0us_O79M2M4I@pXPP8t>E5!F4_Jul&El<5p|AIxi9Nx#Opfi4rw
znC6RO<1u<jth8xT=!C&`t+U8<b?byJ4B7D)136Jkk6i8zT2~G@j^EL?xW8BZ+5UT0
zt%p)KOhxoR7|X=ujqm=w>k8H`Y{hj>zitxic&PTusm%oWhu$Zy-=8=gG+OvLqhsri
z@YWubO5LrIODBJkOEj>Po5imp>$+$$<6mzk<kn90kBsXjy|hgXLY!FCI8+)qg@3C%
zMEZ%0E!*)SRIR7ygrncJEg5REwMU*6mhDzO`Ecco_eBuV*6177cK6HQJ^4aRoK||b
z_l`gD#q_OgZ{yz$!e6Q>M?GQ+9yIt}7|Pg!laeoPsQg`T)kWRqY_nT5@se@oku4dA
z<?fw7IcC}%MgJLDyn7hDn<;do;Y0GQ$%(LD&9WIMsA5~_N^5WGi<KH}Yr<JZsSPQ9
zEG_GbR$syoD+l49*>Hzu`h9CLm}$;GY+`2GYCjsUT*k=Vv%l0w{N=}#16sIURoAas
zh9EV|BTQ#HPEJkynk_%EQ!{@Y{w3f<+PxcUSHomF7N(X>Fni=@#DtXPSH_LLOpv=y
z7+8S@WBWiBKYrblF+nYw0QF5D7ySNyPJ{%RQjlpki3NF82B<lM!AIQUjiJyMP+|rj
zy&!drg92$`bR(3)j@hFmJ?b_64dNkH#*Pp7Wx=GV%VvZ`NE&M;La0m~=bQZUo^LnA
zVJ&02V$79fU<~pgPXQrG;J{(nRRYKefK_?4X~P+iPD8kkPcbK@OZ!)%V%Gza?x5p8
z$G9V|h;l@3BKKeU=9`s|En#}+MWCT7AZB_$NG6#JF)-h0ZVd9N{62IkYM3oTf)cJn
z(g_UivROIb)HEJwJl(+))ZxKKaTLgHz{WXz$Y!NScxABC4BIP&_xef_S`vX1sH~O9
zI^x<#>sBI7J%T%!TuA9LfMGavXJlPINZz4M6~!a54REmB?#x~`!;g$$Z0S-kjo^Vb
zs_k)$BK#Og%E|=4W`oM)WprsZ<T4zc&`&21ND{-X(g_s{yhu1AalLT`4@X@9b)+Tk
zpf&+*6o*J@S~fRUtS6*=jtfYG-fpe^<^2kZk1lyf4}P=eJ(qxOM&cdVs!MQvHhWLF
zFa3>M0|Ev72^y+c-~yX?RCJ6BgrJ7NTxfEte+6yUD7)^7Bwu%jE|o%Vq0w7F0%s*M
zP9p)Nk$(eT2b#Kx%<AfHj<NS_`wT5(&WSKGC{BD~UO(b(toY56wfGf~cu+Lbi(`UH
z1L)#PRE&u!vc3T#o$wyp2ZL46E2#Ygz#YER3mopUsT)o6JtBA-q<vr0me)N2X(?ro
zoq}2;ehgfW4H9bNxC~k&wu8CGy|iD@m?JsD>`tu2C1n09uw;W}iW==M15C;IQU#Pq
zb7N&0>@bXpkSPQ8AQ3SLyWl>7Fjn#RZlPJ-Doj$q;Jxk%U!<&j3EU{CFfa*n6A1~_
zKFm698YoTGjEaQ-t1PU5Or4I8bSb&1pDPCFTvx=N<AWmf7B+y-H^{U|6z^QbkHdCw
zVxT*uj!EDLRd~DR{WnN{I%rn_M@ZdX_XI4cBCZc{mt%pts1CY-5(ljSFqY|Hqf6pa
zsGRt%Ak)I=<3SUKq>h0yMl{|IiW`tR#*3*FVtRdjv1HEmB7`RTCmulPt4(1hnBM!8
zOrD?t0ok;Op{Ii-zU(1jWo5cJ3JU7<u}vNQhr(pq-DR*tb0F;xH2ns#9nBv^pvY<K
zH%rkgaE#&AL-PBgFgl}c9piK8yiH*;)7(fSabw%S+Io))Zk(aiv8F*z9D)qeJeu8v
z`}Kr;EHHk;pfJu(Vp)#layTZdV_49*<1c0?1!u%!KvCi#8*(_xC^nADg2yS8Lu{7q
zhF0TRKx!PJkkCIh!RDfWezjTIE}zMr7rN7LD9LmTaus~3AOmO)3hKf20q>Z)dq^zG
z0_$`LuRc(s8V`Wh!l%qH#X><M6F%x&iIK)Q=)?+5b82Jv03dPDr7#$*?9&?CgbENg
zyf_3xCLNb>*ZGMEL-o~p_s$)vDao}pY6<`X1od~N6R=D^Ca2a2RO28YJ)WTxZct(|
zQ}gj*rU(VfG8(H;zNVXSyMw!GzRoIroXrx5-;G{bSHzj)hYA7qG{k104XywVlfktC
z?ck&7uem}~VHD1Hss1Jnh(u@+9hWcy+HCZ~vfxzO&1h)w>v7h^6cgb)z>kFh%;yh)
zd~FmzvKA7DT+15cfMe;t84gd_;?gOQGNYESc`wAsQ27*o1mR*N1+ZUl*#sZnOJxI)
zgGBzaB+@{#Z#4)_fTmP{fJacr5U7>~u#ySA4JuTiSYD)wDKL;!FqtDZF>6fx0jnt_
zfMPBoprT{b{VV|;SLxm=<=j8U5j5cGpk!br4hl-YS^Z-y+PWuw9?(HvgrFfuB$0PT
zU{-?B*SSPb3+0<Te~$-u#H4@@Xb@e(SdmrmO(25`DAma|-9m%{u%#IxpTUhWi=Ux)
zEUbjCC|m}>fnh-*0z60D=rth1n}d~C59q`i+u!+7rUWbpHT4OO0YwF71Q655zV`!h
z+1&Xh^!sgxq2gE%JE`j(eaQHOV5+w>a+T1+6-2E!mRbM>Y1tj)6Km^b$aLmhOansW
zdb^|>K!AiSaOZ4%I7WFiU>x*P4U(>?e~JsTrV00jrg3)TAk7NJwlHDj1wbZHT3kwI
zs1EYxAUi{XKoI$jP{bMt(1f|F%Hj&l@UQaCS92&BeMAap%ccdBI9cfRfk`X+>w&e8
z*2h2&zF&y~>1RY$Yb^lC@AEtbDkvI}<=+LZaRd~hSwjDUq)H$L@vF#9oHrq{JzETa
zRM*gp0GNNx6`P<-H-dw~{UwcCq>g>Y^uEA&JQ2Kp&&Lp`6Cfq~A`DIlpboNls3jxN
z>MCC5-6{g~5ex!lu!9q6M!2x9h>w9yL1si|^?2A6=u#m7j?{h;M%isu=iTT9z)Bzr
zP)(q_hUbW+W+O!An(slePr!5^uwct41Z-fcAC#YTgFa4TEkS*wuRvTr2iavxfR2N*
zbp=qpGbgD;NFU@0;{av=F~AVyPjI(7%%Lz7u$Ugu#HYVzgRqbQ)`%}`l$~cw9eYt%
zC#Hh}e1Z7hq5n>-K-@4O6h978H6akkOaCsyxR1EoN$JD|-<6mLRQ4;R18U{R5?bh>
zgM1A77Z#r)0QEi8PoRg0!N_Pv#x@G+R6wl&bHVem5_rD`747c&bAnjBjDp@Az<|RN
z95IOq&WwZ@D@%_CFN411aa;hopTlNPwKigLpautY=c5$nQcO%NAS0Q4Y}0H64x!-d
zTYf<kV0w{@M!+rr5HrZj4G3UJz)b{Soc@#`rB4S*1QNz&^MGIDrJpZHo+7k>LZCxy
z@>(%p=OL@ywXFNYF<W-Pmh7Md0%{kyA;sA`h8QKB85cC1zoXN=h+5)m<^Flc#s4l>
z^&gRHRQ4q>bieYF8+-2jOL3~{-tYHOZczHjp?p5_@)W78o4r2g4f^M(L(>x`fl+Q(
znY0kxv%~it2YR2p3GdnIabm*o=|Z&TncqHIjef2rxW3=t7;v()X#4Htf)b9df6Kd?
zVoQ7L`H1fJyhKtg(NtK>2@bfN{Rb<%rh4<v@bNBWt%K7be`S9OnEfSIgeKzsiKa{+
zLyp;`$`OJ0nX}_HWwpuvk<T`f%IBqjM{apIep7jyEBaS|{e9y_A>HXVYMN?>Ra(oa
zGxvRbq7Js4>eV}(O28~4uLbpYQbHS4jEhL`9vRTV<d-(jlAY?mEz&3p-hFoDtX5Z-
zf4AtTi0MiHWWU_mXR>xq4*I56$|(J>c-PnUl?&C+N`9R1rT={G1!bO7R?7Osfdw}s
zbDR1I{L#nXAarnplYDsG*HNKe?>ES#LELy??pF1lA`P~N&X7(nt+M>%2LF&SKi8kK
z#=`|f8M}uLT?V^f4o);BT$8?(Cwh3?=2>g`J^W1Q$(&B-qmKQfFI`c#x!unX;zDP-
z4=jcluzZG_>(dMt=^TrG(r`^o6c>r;Pxd^Ny8BG_;5z+KfBB)>M?pPmyWDJIU7KH5
z;e3^~89}PpFMgDWLVn2MZYbG+#?T*hg#9*elJpTQ^{y0$)xnRbkIyU5y2aSdEPH)?
zaQP;^P(O3Su1gB1%Crmfvp!SyGLCF#jmO#!GK%{vh%Wg$vu+NSX3q1o^bas(_M@4}
zlm`pq2M+ozhx1IXSh{XMdgK{$aNya%i(;mY6Wtb0@K;rCAzkKWlnlQtA_pyWD@|U)
z1h))j*_?grJ3j24Q^`#VV(h<%bB~zE<4G@DbM)1;CbN;sxZggOv}8V|6P(AyJ1X5r
ztIJh1n^rCz9>MNUkzJwoT*K~PM463ajRYnhTZ-CC-oBz>yU%)iIwLbzdyg(N>OJ_i
zoepkR1TbT`-pm`i4ZU^e9__WZV((6J)Z6OG7b7nkI^Ak=EK|MZ`W+H&P=Us`d~>97
zj)msD2y#7R{OqVw0qt4C*@xGXYHp_f+F86c#n44ZzO~9%I{U~<@OYP_)ndA?z7<ke
z-`X|H2VWA9HN(30DH56Q!6-%c9&CJ6Jms2Ap7(DIRex(4Vi4yf>lnUZ-RyMKlN7-8
zOw7fzLCph#@`h{HenFuEr<d>>i`u4ndZxyv$!KGH-e@Hwr&Fu>KHm(fOWfk!{~IRu
z4^_7WM>ILvCGhr8XGMY6An|spXBIb@W+m^;SU6cWSumefn6J)}b6lDxCi}$s7|xh#
zef;`qEGQMBUQ<Nvn#t2?cA~W?+2f8|>cSlvPUm`xlD)`I&Ihp^YkEPyPr|q>rOiM!
z)b)1KLBnIqs#WJ2G>zXzIX1J6PX0nPy`^Mmpmo4>|7q8#jJLTVQ(56b&#~;0EY2`4
zi5OU4?_}s;Yl>`bs|xBZE%D+<O_F>F=c5ZqwGU`E7{})3W8?kCn;(72ESyEQSWmW`
zH*k5|LW%fVn5)4TFtc(?oqdcilP#<rrkv~@KdvrSJUQ3!0an?B+8ye77II}$w(NGk
z!c91;%9f&{SbN#7`eTxeDXHUJ_mQhQMI-knj&7fje)cf&<L|pQ-0SPgs*qJLeR5W>
z&a}R5rC7eKw{~zkLQ2XYYA#mK)CRFq?lTlrlb1=pH@>3RzH4v)K!y(`zc8fb@7CBg
z+>!3-bD6YYZRup=PU0LX8Eg#l&pI$U#IP~7J}MWbKw}Aa(|Nac@6%8FBnbZ!xV6Un
zL!eGs;jrDQw6N5YmxKB&TBpXMI{J^6enUn=<6)5t(I*5of5n;>(mnHhVsHxU>#XGA
z$X_b0=RVk3yY6%c(J=qux}Q`ur|=oIF}uFnA%-)V7-G(q^h=8?{qdg0RTQJM{YT`_
zpIbG{8@PJ*%*C^LN(n}*95QZ5nwqvtYY@Yk$-FJxnr}>xu)bUHIq~&q_GmQbF`f2-
z&)Hs9QHs2s>Uvs|qkNwHIX|}>UY=D*DvGEvBGvfkCi~_lM-wUMOm=(6-8bA3T{2!W
zYr8(hAvJG$K9r<wpK94&Guu5gnN>(+^-&l$@(xIpQg}CG`&ika?EumB$x7GDfb)j6
z=N@MH6gL$!7MsduTXM?^dvvzi_cVy!cJDGJ_S|fb3evP17~`?4mhjU}JqtEu%bf<7
z{0$#CUxObcAGR3I(zY@P+16zi?x3q{Z3*v7WQAX;FUoupQ2Vf^tgwkuu|KzmaM<k&
z!P%yxzHpLMbSEod(s<H${A(AQ$C2KU=rkRE+w3BJe!K%!*<}LN%vgBncaQCL*WpWB
zKQEfHU+**O#9<%2TW>2x2|RO$92<ZOH;E&KiDaI9ie7zwD#T(Wqll<I#2i*+z)>TS
zOS*hf|NN7|4<*M-3w^~f#Aor&@XoOQNqhk%IBSMkFrGMm2hAIvnccMw`}+4iOIF2N
z*=4Gxfhq&%)ha@6XuZdZ%lhtX9>0N01dkZI_>;=lq%@n3%5RW#mwgMpNYl<_AKzr6
z5n;xJ#-#~+I~#wlDIO#pJyNd-(ubD9oNI}A)s?S0>Yl2EFri7C$`+d^o71wAemwc3
z)xfi@zopaYZ$W`-yO$hoULWp8qmOiiC|Xz%WM!8SPkw}4Ej9^o8LSJ^-#t0_4mb6%
z=2E=>_QyV!FLSEP&}X{x9vod@X!~_)y6S03uXa0y-*%{+*A3ifY~xz!^kdGzvfnt-
zXq=4jjeFhXY~~<4h#!rEHi}%Iyp_9gtAmd#+n7~Fwl>+@Y;fx2yyB*H%e1<3{<|+n
zQo}EITT}Yi?mOk=cxne$yY{zxx&t+~B!?6GssnN6)=qM!RtN<A14vF{yrw?UhfeF&
zBCS5D=2IR_<ogaygt_`x5_G$FV^2I}b?PbIof$UReaG;G!AtVy<Il1l<{3bBZFE~4
zls}6eIEgcah%n-*giz!91hyDDz@eYmC^U*#`Ke&4pUDCL<~ALe2T$6Gu8FHqc6X-}
z5OFO(y8N_LZSvkxy(KH0+4@$QraphC?GLs-VDEg|T4zdcwXpCf#g*ChE*D}s-HR9-
znX&1nmD*Tg%P3m8Y|tb6+`|0TfoJL7F9Wr+yKPumd6#g$@`h2gba~5t>8H+#AEe!L
zw3u?dwOs9g#@RzxRz8G!d3=-)D(iD%t?7!_`17K(r5hWR?$^>%UK)2sSGVT}np{~X
zKFzbO<kd_f!%T_;-)ZJDI?0$*n~rXKr+SNNC|Xi|Jg8quj(@7Hb*jemwZ5#-L1~Ii
zCQz~0UM6O8l-=W`G9Q=Pq)$d%(>|}ps@+xVI=o+7yQLscZTV#pc5kkI=HYagXz%;?
zq7R&T&^3AKorW-Ee)JAz)o^O|fxS4np|OpebHLEbA!zjY;YO6(FM;KOR&P8%Rqe1H
z3^O=d9rjWT$BlDizmzmq&Ewt9ho~YI<4g}0&8YR7T$rg_eVMK{ec}#E-^Nrp)u`OR
z`deaFLjj-0znSpF#m2SC3u~!79CmzU@<_@T{E_yK?U|a(oLA4<%7nB(%5wuc;N@N~
zStUmg75Sb}Yx7qm4n|RPjw**9VERopHhtlJ1(UaZ44PB3$V<TY#_xY=IlB3s+8~;F
z@!2B%IAgPSjh2JkfufqgU);|Q?d$a^(@;~)Q2D5+W;5}izN;szd<erE&l{G_ICSR#
z(F+?l`gMvv^+y!31T9{rU!(U<W@1!p>C}kj7p7o2^y>3JOjDH1-;CLR=G?bR%y|@3
zP(>V8t6X<tA!&w0(JQs5QI4vt3p9Ig5E0UhGDVAGx$_)ngAhfbC?n?gk}mPV?Cnj#
z<;LTwq}r5>;vo9y-QHi{7zB!ZG8seR^Jh+v2ljYsx70+dJ`G5Ux$mL+)H_$+*_5vQ
zb#z{2ZxJyyIu8_<snX58d12%I(L2F8{`ivVhwsAp{bAGZMoZJOPx}1Y@9=_H+@|U2
zSN+tGg$y@6LDuuUm!yCHgNut(_v2fhbMv;Z8HXR)Ma`V`Tzap<+HtsB*3kheyR923
ztJ{n$Sq$qZHX5t=6wE)|$yKnYcN?U_F8Ps(bjIXxRJVg8L*Ei*U}?Z@?ktM9Oj>kA
zwV-6Jsh54Ua0xw(L)tsrcLc9ZPiJjAhpw!iKNZkX{rsuhv66dSyZu+So^&7XYSl@%
zPyVU+$-Tg;!toO?8_2txk@SLq$op@@yqFU{F1aCE{S`BXRmcIqMYOZFQ$%61Z&GPV
zX69E#)HbSb`dD=@V?nDhXJ_LB?9J#2ANP8UPUnDjx3h<h!u~32RC9}xoU9@ksE6A<
z9rNiwS{<sQ<2G-7TZ9lTH-gpChF93_ahzPl<c`tR53#D>AXR<@Uq*S{k2hjJKTnl?
zwY0uyZi#i%)zznmqL^^7IP?b>#4lS>54UUh9dVT%x!88scdxzl!(cJ9#vo<`!)9yh
zog%d_jsfShV}2|3yyQy0>ANdA=-k_EM~{*rSMp3ju76T;K(6P@`e1&Lb7(<krnn^J
zPLg|4!po}YYqM&DnZ4x!B|XQj-HVh{(jz8@OA3euSp_BPSO*V|D{<J=0!5XR%^T3H
z+_C%v`r`XtiWsG~v+YH*qBE;k2D;x=?4OC+`UOhReTm)WMecr_c`H}NKHFb^db>hz
zmAD_ZwOGM}l72)St`M!q{Y?Cd#7Ac7v~05pTOC@ODKV{WGWfhZ?($$qp35M=%o7yW
z_yk=XEh$1P)cAPTIv_)AJSr5ewuR7^s}s)`eO%+`+kEo1{u%fO(^IOd{5$xIJ0WA)
zH>PBMrY3H1D9X4+oa#S*usl_(Cp(sI(yZIoc3t(1fqToX$Gr}EQ;y?4tV^3?jkaEy
zEYp8?&cV3mmcs$2^$ipXWF7}1v2%&bf|n}lF;=})de0zSbGPc?SDDmKw#@k>F)4k1
za!W3+kKLKr_Fiwd@N)S`658fNwU5o>vm61>x4Zqt@17l@4B5%2s33jy^vNlb1}AZK
zE8f@#zB+n4c3E5`>{O*>m=u{B2ZnKRN_1E9N28Zs*5wJqXC2Q=r~O9F_DjG#V=0xQ
zww^y98jjp{Fl}cBag@wn)hzMNSZYS`MNO#60Gf!?^YwVSi8gb9HPBRxc7L9Z&YF4p
zym}&KZRf0K;etw1VtHoDRn}p0mggHA#t~hgZBACDsT)+U?7$Z!C1PQ5wewciG#iIS
zG3Fbju+-hB*N9diTfe!;W2mjFrSg<%^Xv<va>$uog=zQtd@IjimsL8`li1DQs(aw4
z-`<~B@3Xm4RG4Ii)LtYVOyro=t|Ab%gRIW8;znO?_zuIN`2XGV#Q#sX%Y4}i?ip4<
zFtt_<um8(4^2)E?_*;elGoFzLf7g)(miBnyM$BSppZ*fO?|<H3J}33(EdyI#mv27B
zC_da{(`!<&vqAf2?iY<GHfPEMYF0eU+<wYoIXG-~81;Ud<48*mewJrnG5qsF+dii8
z`6kSC^uxluKtslh&<f(aBNihoYALvSH@AxGUA;5Rk_^)2()^-3(Y}csE%~w4LZV-E
z&(}w%5OV8!FL||OcC?bcdRA&z5}q%jB1qAbcWR7tGc)r^c(1ynwbza;8Yy>$ht4Fo
z%eR<S|1|ud`a!u%_S!J3d~*jq&yyDeF7XxZ)b_sZdAh&$x3`G?o?x$FZRNL?MGs;9
z^w0JFi~5!w;f$cKP6l0$mX6)ZeHoKGWuitZ9+WQ@36?H5dh!Xq{v6VD$#A<mi$WwF
zB>H3(61<zax*Y#m^Zufn{=4kwb@=D*Z4@abbrcw@wbm)sVbdaF6k7(dx?=5lF!G;y
z=-c=W;ueben(1OB`wj9T0lPXm29^rw+y=06FVo@iK47f$#$#c$;lu1S1$`ZhhglOR
zc(Rge4xKKOh{DzM`Kdu8<NN4ukV}Z*;Di?2aqQ}YtXPY<MV-t6OeuAYue*Ug`D<Zj
z6K|m1;ZOkS@!LnnS8zWTa8-_H=F7CE?`_$sR-NIuDWs+7#XY?5gi;SZm(|DJ+D-B=
zVRdwO1toj>kcvuDk7xN2Ro{_F8Hw0{XCnq(0V{^}5e4L+ZqmV8B(gAfSl*A}<YcE*
zn0TiU3<F0xKljwnT@J!u=-~_AKJi?hbv^L&tum6<@eyeYVNdxk7f*zD&9scJ$0q&u
zW4zij!B@i@+1Azh3X6uqV|~%9&j>y}?vEbik1;~C%FFe&dQ)Y6vT_4*KO%_-D4{h`
z#5>9UMD=JOkGJ+U3`eYQSAR|BO!G(m17s(|48gVHiSDC5>ie@R{7HpXmijGD;oa6W
zD^ObUv*rtA?XbgVpJ(37bs7Ps*S6%_6v&<Va%N}qlULOVHb<<l-NmUM7-`y5mvsZG
zAN`y(p?N&tc=M3^-dU}fEPh(jP2^Fnz-c0>DAz9|hg5RDgtM*Nn(IL#r}`&lepNsR
z>au8qc%?vEN=)3jj?UboYfIalwsnQ7tLyG3M`s)KTG}hgMNC@GFwLro-R>i*+u28*
z3XUv$748o0$#xl%+ViYKt(}y8_k<1d@YHF`tyjWa-p%YiLP5I96*=wgV;wozHe*cv
z?L88g=mI}GMkWT?Z0{bnQl8a5>`qR#dfph6iLVW}A*mOoZ<}{gF!m$uC#eT8P3Y&x
zda6>$LH;qpZuP;otyMT=M3A@B_A72ARbq`<;JYjpQttkeQ3ETayoHrPM4an#cTTX!
z#Pq|#FrS|7k0+dW(ssubM%Gma9vCwoI&pu|$<j~O&iTaMFAo_S=fk%;X5Du_lo=4%
zwqZcEseU)iJk=pI;A{Y+toTB&54cam$)D5iQv-9apDCKASn6)SOb#GrWHTb7RXkOE
zVy$|(M}L@h^|h=&o|k?NJ}g|)4M?djAnh+{$(D`42lgTzqHA%{{uve9r;`O)FH@fn
zPwM4Zc`d))bMwfw*EyZt*zDr@83NMst-%^<*N2l%UFg=CaMKSD(C5)7?OqgXKYFlG
zFT9D<MX;ZX*onz?%<SsWAPYO?&lyR0jOkqM5?<pye?yR>9m0|^cLg=dCw8-4)Jt`|
zNS^*Y{xYmM{a%WH&Hn2K`?H&ePDahM?FdJb=b(OJsmJs9*3{RHCbOJS^s-}I*K-r4
zn~H&(WmW?`#?qn@*ZsH87oFT4(o(oz?CI#^ji-Yw(F>P@^$H?;e+`)R;iiM0s=NTM
z_UTgGmFg|0_NRIG+MKd8p>^aYsOixh_tW!w;U8qup5HDf-&iVsn_uMFjXXTm{$Ykl
zb29r<l3j_ij`hqUQtm`$`DJ2t1lcV`M-gZJc7`#WBDe-F2kLdH>@KI;M+Q!_NhFmF
z0@_|~xa!>MOaz}#aG2B!M{W4sjT9NwmZGz|Nhi0};;b;YzWs-qqN=_ZW7-ve(4@Ry
z`l;!4B%X1sJ*MF|YxuErtSfJ7i1Ou2vVWfPL|;i#iBEP~TJyP-aEs=Zw%To0)t;Ug
zBRyxuWaGfw?Tf|m;l{i{Qq95ETv(Tzz0%qIUFyLr1HKuHAwdBI!=$97N7sugu7^7#
zQRluv!g>bH6MQ{1io$c#HpE7-EaFJ$&G25AreaEW-C$7Ov#yYh=XaePs`Y23kgp)~
zb2-@$1}JvA`Zf_WSqJm8eEjeSlJ59r<wj1bUkn|n4H6eQv%7o!7i)9|N{OWC;vwR}
z+>!%{Bs|GSBRN;yl&BJ8%6&TlL1{OY-8@QFNQ!QyB~6C%oK?gj(ME6OTj^;L`&ub!
zS+!B9i4(s~;G<R9!dk9KF@)+Su0vx+nL`cm)bODr;YJN^%ad6?mpTY=zjm*kp(^~}
z>YlpGu~nyRpaWR)(n)X=Sp{DVCRNOj!&A&7QHadR*AC9>mpk58l7lk$^{7;i&+4C-
z(Qp2A?$nIyu@uAis{JIlmaNUPd&*Q?-#+v*tn8K(qPUaruX7I0d`+<KP@()$P<AEX
zG7|L}+2CCCQgr)^?csFA*S?oK-FlwQJY|!&7Fo)LTYElVOwaqJDq`IkRd3WT|C<?S
zPS4LLO%C05?9?}~^P49W@towXtWj`#s`6+pGZnsVrMNI_>Fu?%2Ki@o&!4k(F38$l
zV5GXEu207tlfC0boSjnh#l@iN!J<IJ{+>b#EUjL%<i<_H(QUp?_AR>59b(23`lKe_
zKXYlGEf6j7#l<V1eCST3*fSd`ylWBUQxg`kw>wln-oWWFbl$MpD)4t7IFl-rDeP^F
zRd8eKzAi-@59!fHUgXf4fg{yT^@kPU$z%`G9R<YBQo9$GyXw|61J(F->K8q7&>n9c
zte{oY#yNi6w)2;NaK~l}hUU%WI6WWiWpS!1vvo)}?3=hV$c8gI@7#oz9&_XLuN+fq
zEn5UXpo-62LP6<*SYGy{MDCCMAgmV$?`vBQP2T-~xO>m2rn9f_n;FM4I)Z?JB8Y&1
zfP(ZQ1V=#x5|9!Qq^NX|F!T<Nl_I@_1W1$;AT*H@LLC(hJwhl#C_?BEI-v#TIdSH?
z*7dA=UC&zgi|5UsH-njQe&_7HPyL<!{q#=p2W8*SvW2}6RmlI)cZJt?Gd2b9zp6eY
z|AFYQO>=uxQ|@N1a#`})<=ujpO4?nf%|@e;Y2QvAIf3R-wfK<RJHdTU8h2`N3!a>M
z-?AE&-UY*cW%@vI)-7fh#l^<9bY59oYJtDDtGU<06m`_LLXadnn3{QCFNrX0TxUjt
zX$0*bju-75G)~KD40eW+kdlkz)&27-sH)ReUmx7LFYUgkH2Toov)Wz#n$DLt|Mb`l
zH$A@s4<G8rhmTiF^7IPdXPq_>pSo(Sy`}~3Bh60YYi1y2qjayLs#(9q@LFWWA^&iZ
z&9xE@s_k3HLG|cnlAwREm$lVbY*TKNt3+QO-sjbf;CUB+lisZf$ZL~VyX=M{80Xog
zy@PIuQ(%0p;QE(X{<AKt?>>yU{6Q<{y6m5H*Gc1kiE!kS;)`{Ce#f-0!)LU$Q{F~K
zM!r6~jlza@phY7ix3wPq%6403icbAHc4^qYM9nt{_J|h_FOm9|?!}<pfs5h9TqPR|
zWonw#G4G*-Zyr_pW@f6!!(N7$keSL2QG@kr{fbcag3Mu=F)X~(RLO#m?notQnw}+<
z;zoL4rgFO0&HZxCpN}qQY|cS0&j*irD1ABVztmz8>{Nh5l?U+By_%uqLj05kTn;+e
z!%YEaV>5KhiUe2rXXkmuQ%h36;O%ayyiY1?^3v!`?<8#t!%v@}0|_-o-_G9h&3HyF
zmEzL#eo}`z;aVuj-ZP|Dy-~7I_pCqIJnPTk=*i5+tF~=TI>tSS3_qv^tg%szC}&hm
zhm{th@EOkoJ>N*D^C@Q4`DZ;a?z=juYb&(H@G7>@-s?!U#)(IlZu+#;pfbF=0)hi0
zi*_gSJx^lYw@s(~ikx1HwHHfNx0}!op#CR6I#Aq9el;vr52}25?Uy`$gWM#D<Hak(
z#0X#7gZ#)MQ9j;&ViCh0UzChc!yMG&)Y7BkAryN2$rWRJH;0ijRQ?578G0*Dtq6w}
znc8e#>+F&nR*O5RQ{dkFTs}HL`%SmoqjF(xV9+cbm79J)+|z8H(B|@xZd+Qea{u?F
zlOEb_a%zcN3IbD3>eOH-=5Ub0<`WP7Q8|sh5q`JM&vgQ1ZaaKT8cr(xuw}u8s99SO
zwXwA9mT}6TX1ZBiqD(9z;JZuTT<uVT21JKFq1JI5X;CZS9iil<!RztA>GFTtv=8HV
zkMb}we&Pd_70uvOOSLw*Rf{hvi>QMq_UTo;dU=Zw4~KrDqf#mon36EXk*Iv(mo=8$
z_gXs#7#H(VwZ8j-mNg6n{H6aiiqv4!E)t(Jg6k(QUhNNF+C$;Q(Fjuo#U6^Bfx!IE
z4*KFj?|TL=&lbT8qx#MaSB1>;dJLDk@4@h;4`yQehAS>sgC%HoO(edc6!X~sOQZ6o
zpWJ4{Y9J1_))qij;02WVR&9Gj4R%3YvhMO_^4pRgo2!wXyFMBQ7epbtR5BJ43MLA0
zf9B-kdj0}4)1S4Y4(3q&?W1PuAAhsWf&=<L`rTy4KFfc|lRfXhKHIzMBXTLHJG_i`
zl<Z7p@=9DhW%b6#Yvii;lV@K9HQ8@qY^|-S_chX2-K7>QPM<@mhmuB29ylqgzR?oi
zw)pnyUS8md&YAcf>o9CQY3HDztQ%5Blav@GoP(j~%;TnpAehC{On2qHZ<Mas{2g25
z^x55!p`gpH-2_XS-S%HAM2!0e+aGu@Aw8wezUKEZ&XJ#beIi*GX1atO-S6)*<{Y-4
z5c)pQ$acZFyVCvUmXViLo{{`>rO`M$(`0kgZ&l)oGm0Br(pxn1-nSjM1fMP&_#hq4
zjWgtpJKlN<rRV=PDvl~`atJ-x*xBCe>$&K|2#@%vB^ulM|Ic9pP+I+0bNBzM{QnY~
z{lEI7|Mklp6mFxd=<m$;mQ1g3Y%=M;;SC*H^s0lVkIBDC{jYz^`XB!Rdh4f-tI8nm
zh43DWI8gM0wAVM4TP$N`<5#D_y_kQ5TGxK4sshr7X||m^QZ^n-sldi#!M#@`xM~*y
z8Glar3~m_Rt+(aKr(mBRNfe3!H~KK&PWx00mG+hvOISZ}$%9f2)Qn*7(~Ig0tJ+Y)
zEgJg^25Lt<-j4X`Tjo^?|Ki+sMMOlTuEaBo@_4jPxyrroL6wd&y@$8l85b=giNg+)
zP#6)B8ygUtUCV8M@qs4)>BU-|=;hJUBBXT2ZlaQf*%sWaMP*VHwIavKx$21Gcg{Ci
zjoa$=3hbUQ@2k2Mn4Q~wg{1#c6#wU_UYfb!FBe>l)kYh<Hcw^bQ_0!+aB_~xfv2I4
zalzi+hwx;IXW1Z}3=dP##m@v`Mg>CNja%e?x5zupFXE4)9O$Ty>!3@0gVBMLaiulH
zW-r0Ly@t$=mH<89+B-+bP)vw4+HN*9xa`((`PQ7E(bF@I%`u}<nHe%M3Zn0KW$S7w
z%573sNp5=UrX_A>3)s@a_EIHTOZa%6?=bqn(HD-u*E)^kXTW8cNsl!FG9sjAp+~X3
zvaB!lagTaHQFh<st<h*r8y>Jx51yPYIY7_j9tZEv+-OhR)6GM%^IU+Iy`LLU_dBL|
z<&Sd5t_nAe!2TCWr7n4Lnng#?+D7@B5u>ekeY^}Hq2Klf@?#4vite|Uj=l3PN*!s9
z_JG0R0S6y~-3ZFGq6uxUrSGjhik&6$u@?`;_!Rejt+L;^mM$5?lq1RoeFym>oC(}2
z;?7jEujAU$W;xQ9ic5O1Tf!Td`L^ocY?;kVafOPAv4{5ze}#AkSZHTU7cTyKSRaFq
z8pHPg_HJUp3F=x$Qb$0iVRdR3>u{JdJ6?t5;BFBNCO8|UA=AdW1=A~<L$3z*`8)k$
z6`vj4@ofwzUMkXJv3jjb(8k7iUw|ZM)=we~k7YVLUoie$l4INJGAdo2FtRv7^pqKx
zx#}EKCAEC&#;^Bm=TKsLW|3uu$FyDK<wd1Mr9+oP>y+--;m2eANH8zA40AU;R`T#e
ze=Wk4yzQ2sjyarOiz!!b6kG5d?)93P=I-M5FB&Ywg{>jzeh*Ob&Ts|Nheze5#iMq2
z+ayIU=be6c`4#oniKO8Rx9%pr5EkpvH_^6!nFn%i6f`<_$qj*~oKPJ_5+`S~c8GA~
zmX_zRU`d=d5^ojuAj0#cmiL{eM{V%I(_dr<DyM3<X)-1c=;D-X@+L#)w_J6Vd{#C~
z$R%?rT@705^cXn2AC8Kau|hG(I_+-JPGJp#3Xq}M+Od8{byTSZtn2pA{ri=sij-if
z(*AGpI}QE|IQMv8i+;|fNuh2eedV^I&b+BF!N}bOfxEM;X8GG|F-z-GNz~@>W5*X!
z(l3xf>tSb1IXO1(8Hx3J2%hgJ+#!_oz)B3%rxz9b`(c<X1ShHWDD<@+sD<3$Y{w9O
zo1Nr~KJq=_`R@kdqc?u;kLfq<fiL5i$IxLt-qUaxr`BrrI22CBJQfah3`B@wX#@9P
zdFglGjviH}DsQedNZl|+J2D(Uyli6Ew!Et^rW?0*!Q_I1^6q&wOwYpfWmYO~+Oq3Y
z9sErOx7lzV)Yz;?$rR2l$3uZ@&f_107kg^6uRXBOV+tp2KXz5Xm=887Fe)78OR+Y!
zQKE1w!eL%5W)!nEv!IYVEo*14miNIYJB?1^+@AC0yt!A1KOYQz^;T!TgyP5n4Tq*C
z*3(LTv%*?7ERhC@oM)Y8rWVdyRQSmBhe(K&)(yHvKX%7UMhB&5qg~_rn*DL2ga%dB
z`@{ZXZ?4`jw!d%@mEDDR!n1B!6h`7}@ud-Ng%~%yb5jMf0Z8!f32FIYWB!6z3~UW-
z`kucAi_=W$wQx|~s(1WcnO_+<l5Q$`watd7a^gH!LT;W%4JMlEb-B7PHStPgCba&I
zG-KV%(}vXKR)CMGE<#J=>_@@9K9E5*CQ;e=q#hc&6p~If_ZTnt$7H0HQh0i-LvZ4n
zOeoY#u~WNw<Fu>$;|IS#9#y&9>z`JsicB=2_tR2;^^%%T%yj&Q9{If}DWxO6Bo)~w
zStYFh`)I`^`vRVf&oXRuQq4$oYPN~QuC&zPB$@<`q0pe_4NmD>SawF9O7`|byF}D%
ze}C|SIdOXv7D~pZ^f#}KsiJ0G?3-VoekJ@dx>Hr|JVH^`>+9WkUeOZ<#x(gMKAqp*
zx1XW7iiqf+NkKmH=5_Sz_pe>HW}S?Jg9^Bj5*+TM1UDlioWM^=K<XkQ`kJv$k#CFQ
zgKYE_T>2#1C0hjT8p+TGKOrJFnh-iuvvkHG<nc>q?fREM%i8(Vkt2_8+^1sD3ync#
zs_aN}dvq7xre~Z(Bi=QkEh9Ol4QiNT^3afoG`diV;+0<5Xsi|zb(A#HIp(U43yO@>
zXsBoY%_b&i!~@-!-Pj5b@23>23+3K;`LS5>^Mln$tLOREA0Bpn_~P7IRTeAdXg@7Q
z+|ti`U;N&~*db5$5;}d@Ec?J%TIM6ChLMIhNT(GSUUsQ$f|ZwOyw2G=XCvvV{hLka
zMaa>|pO#&$KifQ;eo(XYZJXW5Cf29Fo5;y17EJN1%rR*9L2dZ#TQh?&T0#wiS``$?
z3Kr{Gm8<G}0wZwlSa^3Hjp~I9J{0(yO;c)NnZANp3yzxF_Ep~0KXv2$bsGZ?F6&28
za~XoE!pEq3a}r1i>yZ3aW}mN7qn1!xCC`lq)YM}H+hwa5eY{lFqC~E_Qb<AK6WM6%
zg`*V@&ck1t)cMiYzAd{2f4Nss^_#}9$fbk@Jwx9)^lltP%K{Qsa~MumGS=lLgB#pM
zyY&s?RHtVyf^l!DbB!5;5B-BHk~Gs7pf$=W6@#gc@<Lf+i$QmZRUZ(dMd95aKRZ>O
z$xduD=Kdu`IYpbS5h^0O3^i@;hwRSIP97=-c&HyGME{o0^A+6EqAUfdwHA9bKb0O$
zrN;!Ibn35E4fwA#1bDcLAss(H=aNT7ZU%Pcr4-^WSfy3LOgqV+%b`$R?iC%~vx0=V
zJHx6r?xk3deZ#k|nDa7|&lvF29<kFIvLn#iGB>mC#=U{D$GjVIL%&x?={?f4<hRm8
z$HKKc`^gpYpP6P*U&?s5^XzP8N!Oo$6im<y_qC~zVmHX>qI~g&bZ_Ww;r1tDQ@a_Q
zh%$+RDsjgb1!ugkBy!QpBn^~BtD8-fB!7*&tDx68*u2)v+~w3J2c|P4E>BMEm~-Z<
zHrzR#f9}<dKPSwtT~OR<l!dcnvbo@U+uU5-Fy%rHp@pQ&#c9mfFvG^h(n14>8pg?a
ztW+{`hupwdyZ^=1kzX^)$>^(_srsLVs_P{S(v^z>#Z~aWH>Jc`>2H53wOyxuj^a_)
z(|0gf9K8j^0Ju`u_2F_b>wWu4n6Z$M>c0Kh$NF=Oa?IDi`VUqYr@0oCkiJ8wUv}wO
z&*`vx96K%Fws_mhUNX(*MFufWX{4*Ldd$C60Nykp=}=Khq!F6Pv$M;_+CCclqTBfv
z*>;BHH$))e_<1c(LaILI^LL=tPZI%(VGlueg~9yy+3&L*AIWsK@+uDfO|r!tcg#Y#
z=n=^R$HRpeo?FF+#YwK0OZIlx(mHE!6P$jnu2bMrm5yW>;aADajA6s7LKo!BVOFXS
z|Fkj1&WOj@TQkw7Hru6WLP3C}RiU_{v8CP^g<JF&6*6fdBqLA@`9*j%y?=$^9Th!~
z4b@(=KonqvJzR04l@fPyEOjYakdl%_ztek*?kIft7jQ`jw}-l?A4;+bS9Cpn{@u)@
zOs7$;A}=Z9%N<Mo0p`SFxSLN!=^!j4Wu(e2F2i4v2;G2%<D(x(FX95{@r}g%p;-;4
zx;f=*r5oZ7`_T+P+w39>M8A?vgH3G?4lh3K2Ltny509y?wh^I8Hsa5D_LO;;N2)k_
z;*av+8-;FIFO!jsQzz262HHhc2^m!x5A|HXmArd>?#b6rUorZCl%y4JI<N(^J;=<u
zEi(gS20xBbaf|V_(2=tq<m7gX37c#hj=~N{tz%Sa^xn;DvbyCF7Yv&TvGyXw(k;9j
z(K~tp{Wu!;cq>3PDxV>qK8X>HyHU;y`yiM*O|dP<=6=PCH|EytmCrNaas#DK5yr<9
zKRo?p^p{IKPs+ON0Jx7oHn~El#6Tb9n-cw_?IZVec+22I%uzGHe)<aBXJxNepp1i2
z4fnGyGN^LHH&tzhAfK3SwVbu+F{+D}jHC_Cv<5p8Mo;(XbSCEwhgtsauK7|M8NQc$
ze^iaLyR0O-bN_g56R*O7X;=s+`wdxsuKWj9NvRS={o9bLQd3xit}=bmj0{&6#f_#H
zHSxgeykJT|mxL1@W-4&l`MI)((fH3I<rr9`ka%n{b;ubM4gHwjhQ~`{MDb{K7mONS
z+|oJQ*3h=iU#MCA=qI2TKeerKQDmDfaQ%BTmKS=yGbP>BzJbgBq!O21j98wWa4P@V
zmx=1V1J4x;rS@_Z3*r->epVN*_GqGGibS%hhp0i<63C+j%`}Hq>Eg2D9?!(q*ozki
z;z|#PM`x_9?-wgR`QjDX4BJflc+J2Bj}M{|kfc;alE6V&WBzP$#zSiNWbUJ+Eecdx
ziQPQQRw?-**{Gta)Lb*Peb}q+#I&xb8S0e58~IYatA1P<yS_tfpT71Q_ZpqFHoL=P
z+E0p*(5eq!UHGlTD^uzE{@&Dmn#wN2IBuk0auBWE{=mB9u5|t<)6Q$eYt>8%%p-*S
zhGA~zyl|gvQhQPOc*V}YCm3FoW$sXw$|mvW>63NV&AH?Kt<d&-Z}+RolBLhZj_G^2
z<~7xo<B~iIqC7J9cHKN243%B&`i~dygln~dBaod4#phRg!?^S>xK@9Cg>p<0!`*kx
zP2zu)mZ1F#?eVMrnc)*jDM%4v?M`xLQDkJCXdCl~roFI)Hb{Hri2ul9dPIj}jVAZ4
zpZ*wFS~})I29a0?QZ^rCrvR;KAa(zPF0j7pLtlM*o)gIL-rfrgxe4~OAk%3*F!E@y
z7yvai8C3y2O))YyetvVOUht_P7l1X`lFOG^y27#&1Xc=R#qdMc$In}BSDt@f3t<z8
z8D`W$Fh;UK)i(%|`KF1F2BdWjKCk(EGaI>9g`A&*oGfJW{3g?R`Z{}*zdsP9q>Qq@
zLS6O8zODetsel&nLvJ4xpdmq~YQ5Xh5Nu>StC+Zde45#4SIv4h$jY3QJ^uACi^Ndt
z#;+%R4c0?04pP>z*UJEOVF)O_g4CX{x4&rH{`@ceBpW~k7)&QXZotUcxzBfe?);yK
zjHtjMzW5fFu&r!<Jarw2R{uaxV2l7ZA%u<9KuX6n$e&CAbcT?NKyMTfg(@mWmSi~r
zBw;T^t2vYyVguk6pk|a^W;jR^)NaJyUftp70OF!Z8SBQZCS*nHPdzf$^|rH<Oj}Sw
z5G1Q91ppxar7FxJArZ+PWNZup04dniCD-}B62mkx<0nAduv#3?-?swCA}OTKqfj7K
zok80U1&Lw+g%EZ_vy^=NA?tab>)17bZs2pS8k%B0J(k35<evQwq&*u4J%2bO3t%u<
z&jZXApk0Zi9A?&QHjV?uPslCMGzxtO)P@4YWb>z4Hrsb3uN|l{c5bSRO|AbRVr=f)
z{V3oEL&Nkf#+nF{rYVyQHj8Yma0EbUOs@gD0$^(FklTR|E4~4>VuQ9#W+VCa5Zjc_
zcYu8;qwRxmS!3*kYBmaSyk~JV{=wO>=nEt&Z3<-C(wGfzC3}oOh-`<3{sjWqD1LNp
z-?;;+FvD(Zs=(BpURsU#+JJcoxDfeB0+kVy09v(s&?HL*Y@4Q-4xpb3<k)vW#81hL
z%Oa_3ERY3X4Vx>A-1GxrGe5~_cnVygwjWMDPzw>8SOpTK5Z10EzkKLBn>?U+B)8H3
z`LmlW@CeI(r$%JkXuwdH*tKnC?dd{~4(&V%K-beKp)vDsHd#s{8zAJIn*Y8R_7CEs
zjF0$z=jmy&@A9i*T8kjz_ea$006d7TmOb*lP5mx1VV~{CQ4-r+;tugZ3nZYiO<70Q
zXA(go8DNN%cMNTZY8-xJIT!hP;s*vrAZDJvaVSfmu{b_hBU!@IPFMna6kp7oo#8kD
z?EqH*Ia&9X)+4}h4|kD7AkwY@<nLvy8=8(gzt{yCwxA+yfHb4VH^tB_lNbihHUb2h
zX`p-?7@i5r1Sx6WQ#^`^yJ6$@NJ<BI(vPsNw>1KK?v2=ZDg$J;a_xmQG}?xffnN51
zIWZ_)i$nIG^70`O@s(_mF&~$h5w}>rUs~U<2NKmSLm&}0%6}iSQbC=duQ7mzaNP|p
zEOq@yuWw^PXaGp-zc4)~eecG9r0z8RONsdb5dqx+3|&CHpRG(}x&!H)z-n>GQ_w{N
zHEc!pZGYa@xCCOa;^TJ~jpQHhQP3%nDm}61AA+$3B%7T>h1tz%01#0Z+trK*EE;CG
z5$JHToZ(tue$L|Bw=?pgdozNb`kI2?rP0)E{jB~6y+pf#+yaJBcMw>hcY0Fjz)4Zw
z)03+^M)N~V+v{iN^4oU)GV9fy*bg^lkUxdbom}4wYYdltx3l{+UqqHOj04LW4U*e-
zfy6u*ApnLctf2`EG+~hOVUTSuz;wEE6F4J3d6=Pb@h=dxK@eFsR_qd|SNGqtnm=3K
z%9f-9LJWf8o&ne-faS?w3nT`r0YuBck(A$Rw*{1@LGZ5mgXC;ZAYfdt(?Mf`u9O(0
zf#rL8BpW<}@^MNAD8hrFUjx#&Rg<hB$lnMAgCppOfJv~r8+sHFm28civD8`mPDtoe
z5P0AQt|aK;96<OSjE}5H<G@nB0X~iI-9`+DJCe4#4+7&5(QOxltpZ4)B~SnhJe6%b
zE;_qYe=Mx=qVwcGFqnVHQZHDMwXq*=SIuJQ$bg}N1uxQyA6^H#&_6ouO+hX#Pny=l
zMhyF>Rq*V5$~?eBfp1Sv)5A0xfXufq_tXxVB}NS*H=m{@xdCDnaG(gEfub`=G6!uA
zJsbzfBtX>(0*^Wapf~G>hIWy_j+>xCva*1<6Qt231d_-@nHcuJe#}V_Y&>CfJ^i?L
z(;}78s6~S~Py7Q`a&BhR6FHQ%A7=bB$^+yTfF4m;!v`?xPW_9YItbVb^5hiw+A0`L
zcQshkCgY&dRw&;Sh?0V5L4cgJd&NgESr{cWEpG1HUIJVuKw_ePU!wnuzEKxj+i!CJ
zxU0qh<N3d6Eu)#WpdS-ApmOowWojAazb0UT7p!-8BV-}rbH}b)e{hfIJ7~j_&JsDY
zv5cC%zuDN`m_f6v|1B+?^!p{&A^rdM^Php|e&qBw8|6v_ssFI@kJ-o~8q9c`)F~1=
z_ct5TXeIFZY}loL+s|S8D0b4q`x30R+|6Kji(PWTvO2-uAo<>23MJbARN04hT8SYk
zZXu<Ys&wsDXJ?nB{NCC|h1|8Vs@UY<mh+sP!5J1(*%^tPW?nXO8gOnNxPZ}z!tF-r
z8v$rt8p45gzriLNSEK|K=-Nx5UEB%?e6~{*3J2~6M$s?Mg7cHb|I-!08u4eFH>{%W
z&v?EunK?JI{=3oJR%x1LKWrh*P-*ncbma#77D7x~7gptKrVaEE!9_pF6x4_N;&!ru
z&h0n-7POs)YY4S?GaaX+nmpR(5BvLX#yvx5T#*y41~N6*pIug_z}}tE;}pBB&8^W%
zu8*7)(Ih(wU-K*+^6#9hP^se0aksk0aU=Yr`Lg4QkL4Sc8$+b-2i1BBc$$^C=9TW-
zE0?~~y4S4+@;J|13@*oMju;`o)3zdnKJQk;TF{-1d8hLv^)Gnj=ABj?K9&6n4bFUD
zeNDQYCkb9@$XLGwhrXTvu3S&t{T~+incU_(4&8{_G#?tkOj>4DON*I$J<y0wYwSg+
z$sxWZ5Ea+O?}UBhDAm4VRM^_#>NItS*U8QG7z!v{s&2$-Ma+OZH8m@wQ-9rh`#;M4
z&-nHDRt%8QJ?sBB+oVUR^M&+1$-mj$^CuL9LOT!t8=L*%`hU83_*Lt#M54~p{qy_L
z4^?NRc1+C<dN!n-_kmrebyr@$Y~A&l9-0=qv0%Z{t>-{ezR>~4>tBlHc_FO;I+&GD
zfLV7Jl1V(20W$K8eSt$+4nR^1zrFq2k;H=&hZws*{yOo7`3pdOtcGc5B?IWERz7ig
zryj=!5So{G=87sRK5x?3S>=LvwgT~38JotN?7Gx%fan3xJg<@R3@z*^i((+zy34t1
zl*k4MIY6ufs0py}JwG1BrY1DW65;P4Bc~Vu=m{2OJK!_0?xi#80EK~tANUN`+)s{V
zms;-z5}$&Hg&Q3QdjS?JKz=Pkg$M*=s%@T5IZv+ntosLrfI<=u$g<Bzf3l6Pe@z7y
z410feDx4k2-5c$Jc{hGempVCROV(-zzqbJ$nSGl^_OF1{x#uI#bol9q<p9ZrMeaP3
z2yiZp__e@ru=E8g5EwGVnuvtl6o^OhA(67xfWfK}&k`RYhH0>|3R~X?^oXIKY=Jc|
zj!&)aH0136ZU{;eYMB5kJD^<vX3DO+b0|6$K5=*tX!k=f9Pt3k3JeqQacLhEjD*-c
z{U6l=Ms5vfk1%rn22_zmcXlHHabw?>dy35=T!=%9G^GXRb`}YV-MSuZlfYII{DbxQ
z5D2@Sp`DS0%gOWS1MmK&CN#W;O|HoM&UgxJ3zj82#HZenu&ny}EMQ@ImI%sEK6H6N
zshOAt<QPyNNH>`SytHMN{r|25a5W1(yQ|qGK?Q)j9|TrYy<HAu%n*3J3n7E8no-*_
zBglvZ*n(Qlgu_M|%K8pdm&P*bo)1`Z0^$dKX*DPW2x@3EUvD0Ue_jvVXBmgKQ^zqs
zMPK!Q>b)AGg-l@Kb=QRmliQ&{L5c17zfsm2Vbn3Xj(r5`2%xi@%(nvZAP&KvA>1fn
zXbC8S01Hp8`Rv=?`^Sgbj;#02WFSfP5Av4s4XiQ&#X&2dx&*YZ2IE=arkOQt6)QGa
zlwSZ$Vp{JW%C>;r3O0Ejh|RF}XrQJ{jFvFt+<(>MN3{Pes`b+k4Dh-TV3Gb~n&lr~
z0SFJ^4nWb+0E-l7Pq5&O=%VcfYWA)_^=9!)1mZz8fV2@S&VU)6odGe2^^ij?Az2QY
z@&9fvi)*z8EW!d0|J1q<7+o9=N01ai5RqTqVG4rQ&rhwdudXwKfVqW%aL9?1n{YBx
z^iVe7ub(sDFm*Tg!nH1e&!)7F16~G@rZLUVUK64*Z-TCsz|;WC#0u7SWLHD85h!s&
zWLdQb6r2b?7QI#h{KDdS-3Ay_q387d`X-@=$3=nr0>iN}9bmx>WQyFuZuvFPhyg9h
z{|ZpXfYQ4UK6VC#DDK&giX9^0OQ{{Gv3f2rP)7y!YeSO+P_+8-&mM^NL0p9U{N@tm
zCh(wa3(HLy@-SFf^0~5j9#cC^As43ZcdaWwT_%qJ{h4vFRArg<66hwhX|Q<xK{aNZ
ziV*tIApzRRTMMiT=WOSLZi4UxRIi=Zf5scomm;?4dqH3|`j4mor%|210P|xFLz{#m
zk03wTu2k)Yi?07LX;7#&^P*7xbMOVQc}oU0WFdoJfHDE><cqDz46xG=EO8XWKQIVv
znszi<!^sb~Cby93tZF<TTJ-^>H5Pfu^C^o*v$hW4RrW(aF`u&7!cBa2-*^7{?M=N{
z#~gq}wSr$8x+9QtwqRt|r4D@uOXBGIA1#xZTJ>24M?r6YNd|)^9gH_Gn0DQOt8rYG
z4{(S-(>L~6#|34=jDs7Y*e8HjHo5A*6MB*Tp+NjFU4+gWV=07TRw4DQ*a72TkU{#7
zfr$lQrpe})f1h3YVY!39j)=-~P(JUnrUS^}H0Z#aTKQV`?KXLxPt|~$71v%+P)H+(
z18C0@5U47Xty%-H98goN5ou!I&V>a_n+GUW?C}pjwH{`H<1krCU{Rn7q#^i{u?x^5
z^xYc}?Cn`r=Y^0x|G|^|7*fCoxA%l(#w9$RT>u&f47YQD>Ge>u<FIk?^+)ny7Emku
z*TjIH4f<)qAvk^z27?-lJw{zx0}-M%$Ldh5aA0vf!R!TOmFzs{86E9x+Rop>CTdM6
z;V(cI8u||QRR45IR`>rmD434ZJB?Y5hU2V4t#|$ZH3^G^CN#Xh0Wi71qW|@M5KnE)
zvGHUu8BRG)PBXz&1EAf211->)1q8ajg4vJ;M8O8ZSPwi7$l8Fywh0u;esQ`vhb7RK
z=zwY$(Fg3<ft{XR+Nrm_F4X!1MvavPfF6LDwu|}13Qr^1s-+zuT=b*q0PbbfHnH>c
zUhAQVR>1B)8;Bo3pk)F=j(a=kdtiTzeL{T6LIDM7Hu1^IvU)B<b9^s^HH7>-1B3k<
z%M2`c0*f5JZWV~1oMqHEeg)k#?1mN}WL)CMDgiVD_<SRfX_xZ!9)N*6S9}NRbUZ<<
zX*?o^CaTqZ^B;5k%Sivn^uEwTZ1P9oYTyH)CuUdGc_eWs)YguVLYp0$06XJQ;(-Pz
z5d^HnS8#UM{Db&+@dp*0<u<_L{%4qI-Cdsw?F7gmd2qV27Zkw>+{8L#i*F5mzWL=;
znvTa%Yd5#X-)sjdVQ1GC=C{cTS`PhZcAXEZn+QAd=l=8WYl+IsxaVqJrZJy{-%Mw%
z2o<((p$<ozc8&KZXdGEvVEM15lI8I#D{PvJ-?6H8f&AAis$Pkk)+0$#1;z^sw^<A4
zm>b9^3hi7IicgZlOXBeP?dfP-IHv%4;zSiGA8|a0c!(%AGIes9Nq&@OWTqffR0|g|
ziz|nHdDQy;suY%($@Inh62z?TJU=b<h%>R*HzBhwKnU@qOXZl4FV*Tx2?~+X-*OlU
zgh)i((C$P8UHJ;G>>Ey!0Qc*JE0LGBKM#CRxwzHmlfz+Fm*c9cQdEI>;<q&@XVxMI
zS6;TnZ}}N_`SshhX2$^v!E}wf<h$VbT;=Wf>ELOvqQHaH5KcD{>ng|d3sEI^C2SH#
zTmz6jNru<&v}Zc$^HxlE<s#*IohAk<8?3%)eXWFsi67Ke!Gr6As;q6ZT_z>bbLEc<
z(~D4%XmQbTg-;A>t$0zcUE7{NF|o*gqzb0&(u$+=Se8mhEkwIFg=02J&`xl_guJxI
z^{vNE=CA*Kll6<YrzzqAc_0@0^-0bX^a_W8*KZdY8nL(Ep%TV=Djuel(7J6pzPc9W
z5z?;vUDGlv4sO@JpLDvr5TSI@q427=F{hly_TE7vQ)8NB-3j$l5UShFwaG2HR*gJs
z;CJPgwI!c++6O(Z6rD*iX;fk$aa{nRJNJK7=m<u0iM|*qK2lj{PB_mU>E!7C#-34B
zQ%2dV8RBt>><<{aV0;O#^Iqc_)F~RL`TWYvy;#D9=v=XDV%oUfY)OWfmT!$3HuPXU
zwc+Uu(J6b0>Pc+$OSnI6>96}t;ZaBewbiUh+A`nOp{*SytlJwzbHydND=5hS?vH)p
z1s8yl$Ye6-)J%T8E@5^Ca_!@p6dx*`64blBrBlW;Og|~pf9b2#H5@YQU_$en?roU{
z-`UqZL0@I`KOE1t5n(Qcyt~?BA#boO0JmuNh3-cLglh$mlj5Aq;zB|iyG^cNG5^u(
zKFu`*g5;7B)-+IC3=#K{Pg@ki-_cSZ-BVq`ExR^9Sa=`*bK5|-T$t6pbo?U60_?6#
zkDLHmU^O~F66K_Myswa-l*4b2|A6Ozn_Qp@XE#5P<{Et1^?-1lKfMM9d%$U(5;-$b
zjYOV*sjAbfoZ#G=3|dzcH6s$X^g;L%8^^J3#i7)SN%Icqsq@#TNYB+=7CG4;bV=5%
zn?Dzw7e=PMK8vx*O>A?*U#W8|stoS%9my_A@gdo8n!(IAmS)sm^DH1Noao9$TLDr8
z8|6h>Q8qCu+KonSUeizlH!=AgOM;z$Pxxyfa{=P+=;&MM+wL^*W$ybY{j;z83J1Kh
zAuqZD+ZdDYs)V1(P&}Mmq%-{7)WzQ=$Jj<aS2Zu8ZQ4dvAv&$g($Nu*`{AZtXBlYX
zK_MoD<X2I-N%t}Q(vf3LezS;_-TIbZ=e=E&6SZ(qR|#@3ea5pz9;egV9m_xU+dTa5
zFgtBG;PL6ct8KwNFXl^&y^U@kwKh76x6HEP^ze(Jsq3!1`*IInig9(N7%V4wkZ4JL
z&z|%pLDK6KH3^;;y?$Bx^|i(Dyl+kZ8Hz)=CEJYq{o~PYc!a6)9xe_8a;ijvCR~B?
z%S?RHT&?4|6uW$QlT()kz5!la3<RX)X7H{temY@t*8MGT(@v5=^;0YNN`RXe8qwc2
zPM}La8MHh+E#9d7jyl?BfB(_Y?N~>_Ldr!IT$gvF<+jF`iI0!Q$_c;f8CfIMC3=?d
z=`-OM$G{st$h49XCzZUgyUPTZ*-m;)&4jdkeS=vsbL#KxW1dn+#Z4LDOQMV}+9+kJ
zGIqMq(vP_8-&0C%*x%FFzhJu3)8^+gFO91*Zf5MYJ2~Ch?|mj|#>N(DC|s|z-gTr)
z4pu?Oaz_=^p~>}&;yLYa>fGv;l{UI{^B(zmIml1rEIoVF{I-&;-Hv*=R(QFg<Z`v2
zE7!Wt*6)UI@xVzg8wZ{0S%fvp>$_jxl19Z73_ExPdF6faMwxAYyh$G95iws*&p&Dt
zg-$&`CEm;99$E_>-?&^4_tkBT;AqpevHhFvla@Hs-8^!+=@0&&a{5%`>YJ)HJxgPw
zrufT(bWZPj7*=aQpjV5|K03>1oUX0BUO1oGPV3J2rRV;zrD>t55s%_pzs4FBt0fb(
zvx~<&<4`j}^!>(7JCEc`quudq8c|gZ&ZE4JrK-~i?9dcfzq(^)@CCknQ(f#Y-MCMo
zTR5HIl-%U#Ag3G!t6(iN@dDs+(Zxt+Q~b)r+0HFpkarVWKg!LnBF*LAEnKk6c(Y*3
zmwP@$l|A2~CalGf6WPctcw%g)_a^sqO5v(~N0uc;-ppb6V~mua2^RK3b*1eM#Q+^j
zTZ|uzuFqe<j(pnZ>pk5(7d}4!-iEwzn?I|tc+6&qr%pxLlD}}Rts{2!j`E?7C;#O|
z<agEPA*cN}TgDgOyOm0pwLI<UKxd1DghI$zMPmiI@o})~i}|BP>M5%)a8isK9~9XN
z6M0+VxUD6V-bV}VoCPlv+HGFE=8iOTIA|>nT`=XTyio1-dQiMy|8gZVsxhhdLs32>
zm^j<94l>fPAB`;NzH5ww%-s7X1*}4U;t4e%j3eo86~FY-Io`&mAWq#EOA)wpInp*%
zNr!X(Zb|&(QGTA@>gx9uK?P^OVs9t(?Otnxw(9==S}Ge8Ua$-F=>Cb7t_ywWZsmPD
ziUYz<en8{6$y}lRmsX`C`VT*r)t8eb1x*CfV%%_l=&K*n${w|dEE@_Y5Ty#QSVxw%
zMRz1or{?lUcgQn2rDYKlLP@5Iul-g7)IL1PxBKH$0B>s#PW@@o6;6k(BspuEz?UZ@
z3F+QyTev>=W9W?g8(Dn8sY(<j)t6UyaH<%Ij-61ZKxQ*VVHFeJ$iMONwWZ$!YyYZD
zNAyRGgZk^F!Nsu$@i*(*oa-DP@xKy3g)=vnq9$xFyyS@*7%+L@(z(YDjriEBm;!FE
zPzjziF>VFH{q4guNMZfA$C6g4Iw}gK!{T3rt2~T74XTbQTbg<}Ds{dZ<h;>|xWSD=
znhoe0{QN0Y!YV`L=%2a^!!dH4*0-0pV&79hXC>2wwBo;IH(dd*5nfbV-1X{Q;plBp
zYf-v=B^JG)toG90jQHgNM=cMZNV~L~h;}W(aR{cHDyEQMxj~XQb4hZR)V;RjB^cy|
zimsns#E#*ySN3wY<*&SW3LGj*4!bc*96Bg{IS6+hMfk?rb<-w#$KJO(_6T0^zWgbE
zw48Dg*QH=1l9tEKj+g)KYm&-|6gO}KsGl4TX;FexNQj`jS1k;+UDi))9tnGuHJ7-d
zm=T*7b}{T#7`KItgAU#2EI$0ZXf*o>A)9o@wRT_H=6hY`VEL@5LSA0NSjiyQjJR3P
z+WIH>{mk&oHI@Na6D=>8#{@C>ResyAev*C~OtC)uJ8>YTofOxz6|%F7X;i~N{$_Lb
zjMxd<H&MUh4I0A6c8S9)q-MMy0~E2V<EuvE4KL-HC6svW1f{9n#r=BCn5PQWQJ-<l
zAuch=9GZ961a>El(j$u(NTF~GkiokBwb1LqU9T=vZVhPA;R-eaXrmg~nGph7XpT<?
zX6sNuJE!losPvV4p0efs(kYbF?MYNghmzI(SK}n-yGmPditHL`Q7FixysvRD_NXT3
z3ik85uJwM-8k3UvPQ5D%!J+p)>)|J}zuDd~htZn3F1=A=^29Apg7OD9)7EISU+?-W
zyzVhFdTW@(g**`hkag+XY9c{#mVey3$!g#~Kj&sGHjtH{^_8SeL<f~TMSuS2UEhMv
z&&6L*Hl(Hacw{_Vs#Fm6%19}A(LQd#A(>k|kT&3YIkVYE?&5lgHo0#15a##IcJb<V
z>f1A0IO?gFe;TBwP1THNDs3B$-w}7Qj(ml~g`yV-Fq4qny(f4{?mKZUW!T58>WH9b
z)b^=wg|GcKzWf4FVY@lbPO8o-O03M0#C5;5apbaVUG@?ip%uJTmbY@7iLPxoZS)_~
z<odV}Fzh52<H4T>M2vVff_QqG$$@)BW@F%lj*lEC2@*~wW3lDLZ#K5Kbt5154S93~
z@tT^(n_4&obXMcg#QK+Z7d1;VT&(U@X8R6|Ct%QLDkM$}44-vM0fK0xM!4y;kXiV2
zyYQud&Gi58e~u~IG4D}@(nRezNqZ4Op2B2kc#Fcy>?*R$V&B8gEPTO|P5<WqR~`Nn
zbXtFI?>y1vn`>LD^`$jL3Z$lKX=!;P%*YI&Zr&b0&T%Gm1<V>hl?ERKm8PDtC*-EX
zoZzM+X`Aa&6*DU5hZ~=tPVss3vT^XfdmItL5OeNzl!RD!WKpVDH?N`>i!8e9VgNr&
zNGH3W+uCNv$!*~PMT^cxhjd1`x-DBb1n@^+>Aha!Xi1Du+{7<iob7jyjLgpmSC*Zk
zO2ZoFUqpDw{LSWlcITD#K+^m)#L`GY8702p@nRsmj&}{7TIkPphIkeIb>wl}3N$Hq
z6Z1z>8K*-S^)*5M!|JR>pX{Vf&dKN}J%}hgy10d!%tKxx(_uEog3N4*3010#69Q_j
z(b}uHc-|l3Yyn?Hy3OM=#_)s6yJ%W^w7T0YRmtd+h;iiyl8J}ho#};YH4fhPeuxzx
z{HM*fM@n-GbxFGSU-R0iK*z)+^1j`b=^ney<*uwa5;{c4bFfH_;++F4UKFm(Qe%q(
zP?vEdMuL-Jopm{OeIKr8&lJVpk+CZ&MO7h6@sCGzG<p1dG-pw>O%)&1y<wz9TeTWZ
z)4#ls5>6&163`aKm2*Mq-8XSH;R97qzGXZiAq?n$6*;v=$+j)rGmw~8jijqCc$<q4
zd|hWOCt{Of_39o}0<zH26z=7JfPOMMGYE&n6!$u+tY%#E@lL9&EEg|~?fjsZ%Ej`0
z8r(;@2r2W}wWi{1R~;X|!OgX<*%W5VyF~HVaw|=r#S?*Whz86AKe8M%;`-_~dQrRR
z#&6@-%1Ymp=KJ)b%wM+0Ev1+GqM*9N3u#X{KeU_aDHuyy*V@nZE)3Up#Ts2WrBj6%
zV;Ad4v8x-aOQygT2Q}bI6b+I^a1ClZ2oh$ZVjde|(AGV$JW_T;X0&#V9#J(MRT^EC
z(;%l~+}tJC>?07=sZ8)i|IOx^>#N#*<CTsZ*|>tBe@#JFF`oi7q`d|mN9LwSxok7{
zIp5yVD)8%`v#L9HxwK<A#Jjs9IW;-MY?yLg%Wab)2OTMDFm}gBROl$r<BL**ntin+
z+yWZ{5K|2~_K)qq_we~?-U@!AW7cKt7AJzLqX<z;5pdX9(_sm8K6Ys3jUAP-E@EHQ
zX<QbQno)I}%y+#{xi{VN-BDpbSi{KZCt?_v^4)ex<;h)PX&HwPE20mEEadHvZyn`)
zsiJAuJ5v}vDIOf=Vcfqcl8)IBv?gH%{C$;NLI&PM<VU3N45H5Jv^7WD<D5jKGK?*B
zi_kj@@k`t5vxjFxLXUd~?8OsG9#<w+&ncP0<TU;?KlNhx-rbabotTvE@^4|%9t1v_
zMiZgvY^(E$XFEIaVcA0LIaO<$U2WMzI0;d`e)%Pxn$pnujDg9mXgq+1Mavi~_76Kn
z?`+`g{4;7l_`#}5#_;wREPCIsi%*-=;)i2YoVs-sJE7WJ5Af*16;8O|KK!(;vx$mk
z;jo-Nke%Qzlsk}mK|vC-l<lcbeDV@+)1VjRG8|#%q(p(((q~p8Vp|Pj=%6tnymB?N
z#PoT<$<h72DxY%<Ew?8V-*DrI$M~6oj8s#zP7*z&q27wKXgSMNrwlUSN99#Si+Ib6
zxZ)U&d!+$1TpS8F%0uCXYU?(kXjtbYH_Y_bjk{CsO1yXKRZdbO)B80ju{|+2l-kt>
zNUa`U@APa=-p;z#`22FE&O+B0b;9L$y`Hc6?oj!iO7<{^icU@oe$%!Do(=f9``7_^
zymRFSPd@@n_ubNNhI60o(%5iZe2`<->ugDCFFJJ!mc!rG9Vc0)BE47Z?MXt2kE%%1
zp<p?s=%v67)fY!PD0p|NfBIm+Eq33Umg?htP4aSH8=^(mS6xr0Ik_|s!DaBi);i|>
zE3y8yg3o;$YaadtdvQANq%-D9bd!@Boy;jG<UHI_dj}STwmk-KClVug`vW980J&=s
zBU<TWX5{&pDd<zq_Pf(mKr%7*Bf`JKsg5gL@Y*Y35#Q5y`z=s!wzDdvY!J%YHAzn8
z;$nV=Di0^!E_W>J84Xz_mrq{KFbfcz!eO6~hU9jb2MBdM#z}MmY(iv52CmIaGWC?z
zaGO?HVU3cgT<c_U=kvk%q-;_7cSgu#4(ncq!~F)2yn2m!<NGc+SLsbz6VDVyht`Ez
zZ_^oIo1hh1I8(4q8l*E1$1gW6W7$VA@93}MFH__24(ANt)6z?n?;A@E75bgucY7Ci
z&~P>Pn|Y@PYW=jKL6O%Bc?@0edS!;udGYAS>Nr*UIC-5$US<^Gog>rtom3CQ$9);D
zhNCrvNlwPc3|n{Rv)D4U+L`&{_uIAj#n`*GN}4B83pKA&#G<$)t{3c{uO5frd!m<K
z-8=lv)itLcDB?AZ&TetqA)M<;WGW*l1U2}oXrItu9C2x+drFCmey$qVic-<rd^2Hw
zE5DfJ4D~j(<hnQNWA0L7Djuo&*~CeesMM8}(GkXFe@#DIk`}L3wgyx{q9su%Oc;d`
z3T_RpNt&EhV>{>b$8hd)l#5gG0*wa~+D_>_CD#59e!lQD84I`CUlqM;1u0BNbcu7z
zO))C?N`33@iEFSnn1GS$Jn}57*mZ1g=h+0G)x@JlCD%79>(VFNnHM-6Ul5^7XDeNn
z+ZQ%G{1R+cpp)ftZL)11QlVKU@HT;Z(~GV95g9S~LI!z=ne@40^z`VR*An+634CQ<
zRg-BSm%@U5o}{<%``tD+Mpz!<HmT`qEL~8U?;FM>UheW5Q!M-y64hg^i_om^EXsBY
zCSqkHs2j|81=Ni(ae`7w4Wz?AA-!iSMip^ki1+aZ?Cw=VLFE1Fb8n1<=X!FLenJ-K
zLzCL~)3ki@st1U_%9|R=b?(#RY1^rAxnaoo#0GPs6upJZyi<Mph551_<^BxO{@m#3
zdACw-j}wFD&F*cTrq{PkujSR#Z>J?s{AKd}#!0A}i73oC(|qiSw->aT#K{Pe7~k4I
z>thzMGCod+&iq>Z$JT2PKlEUfq~Cmp;wRFTr#fv2_=5@~MVyX)rjZ03?#Bf+tI6fJ
zQEi^c$u5nYpfw30XGyeOHMJRxJ;9YJu;vjF#Jm+CW*Jc3b-4+f<W@Bh5)lzmf--Zf
zo^(w^l1j2mo*$EIDSrHmg+Sfov0DxuQS$D~2JyV7*!ld3e8_BU08J|+9v?^>+F4uP
zYYatg+;T)vZ>lCgtaI-l?HF4;tvPVEzIL~AlbW68pPD9nC)(Iqe(II=(}M^|W&rB;
ztJ%J)wv%!&V<VvKytG0lSCH41(@_g?K`T2O(+nSG;G|^$*NA_E$42zU-w80kzKl0b
zcdGnFa_3K$+4p^ky^~q8QSoW;81%}S@gUBn@7AtMA5s<@90uA=Vl{R9LEc3=jSwD4
z*8t)pi12l(^r9%N8jZda6w(yBnr0K$zo2=nr|VXK{hwcwkN^5*cDO>YB{@9MiaQLZ
zX~prGQeNqQ+0a$&f{Pyl<$J+6l~bTox1mj;8M>cH0{7P&2S@YVsterMs~&vnR{C^D
z=g6(A*B0H{k7dI9RUW~QOGmNuxU*|?j-a7a)qNFdh6A+zUP~nr%R)agr{aN+0)r7h
zyF#l4nU6)?Az|TduC}$7UhG4M1bW7~KAWk|zFJM|zD_I56MlUu^xQOjt~UB;?uDaz
z`HKRGP_@zzPM>4dQ_!_1#R}fwCsDI;kOIs&d!T8Tu^d^m;c&%s=g=_Ol$?!m?%WgW
zWahWlg=36A0m{djgT;JjD;rDMb44*pZrx#;zBTNc*Es5UWR&kp5_~M*+}TWfc%g?H
zd-UY3C9(QfdfLN+@_2M`J+q-H35}nGC`|SJcEh0_=I4ICtk34&*z~J3CqlWY*Acej
z7lyaLJb15wy!hd3(SsPilDN;Y@->ABt9Pm|b*)kQPe$!7w?J$f5@*Mkz_m!NCpEi8
zWYposl|tct^Unm~Tz<#ofn|!xyb}9z+~k)}Icl$}&pVEY3mn6!nZFoO)j9ud!OgTg
zukV`6!zhD4cKPF?3nw3oQkM?KwXA9-(lO4cC?+IG=a_I&$k)l|Me31X#;Y{Dh#myP
zV#D$`sL0G^PqU3AQ!}EcnBc%cRk1mHTPvrK(=`o0MaptLEmEn!2dW#<)CL_?(5wr|
znN6kVTJ_gvAy;l39rmBAx?YuIxRrNZbiVs-+nbbcx5}%77SO*aFWg7`ULkWRXZ-Cm
zdJ1WkdabaesSkN`5^o=8KQvx>2;`HRI>piX7>R|w{Miko6Ett|o(bc2L#A;*()HDa
z2W0ihsB4uJpwqn27n+pv?pGyA-EZ1=a<nepJO--`oAIRlfr^WH@@ns5s_PwehuM=N
z^StIb+DaagYf{`d%2d=qnOc*`drNZuVT+_;Ip>f}rvtHo_&7asWjEZ8L(4td9TVao
zU;dhKV*E^4(t`h(YlYzS(1NWNe(z;Gr6US%`s*`~7vpBSl~GFm@c4SKX;S*4w%5q5
zwnit6zD%OHWI9Gff)PkxJJgECxNVciB}FGKbnDQk6`z+<E&8F^X?M`t(1zO-{sJX#
z+EjX-tU#OE)nnrD_Vj~THAzR0pEX|KJtQSt3!v!^ujjKxf@UCLnc*4-(Tg(@1ebOC
z#&$s3&sDVJ&kTl|ZS-=kVA82Cf`5(06(>!#zRV;Fpv<*LE+;lJl--Q)|E^%9;!vpM
zrlkH{sa67`2@X_V&u&a%YUXDNkd^ppI#UhN{qcQQd(T{7Xov4GGQkNQTR=J`Y|jg{
zR?f$2yi*?c{_qaB=sqD$wxk;@7d=Z@diDtR`BhKuWbNX@Bsj6${lgdtg99iJ2X^$@
z^jT>(_nEovQ!(@q9r=TY?e|R3*>BP-Qzb2{=?nvnw0@-<?lNJPdI`zix9f>joKfug
zxs_s4>2c04pljTaDQ%Z8X<lOE7@4e)Hp2B8YUQ|G(Bd_fPhR-Ll3m3vb&s2sx6Zo~
zV+ZubyP&ODwQUd7oRMqVxuh3ZS(w~7bqR&J!Hf{Xbjvt7cZW6Fk}F21#v8I-Y&R9b
zjI*$Px{RlG8t3hLiI>CkD?+nRy0R;Dwn|Fwqm$nCKVsuz$_x_p|47jjp)24DBIE3-
zTFxo>Y*uFpJ-}-2?|Ka_sFtY=EU8%cw374e;uDi1Z2o48T2KQQHP!*tX%#7aHi+EM
zd<2ge!d7g5+M0MDSN6EhMOVf~@P-4?M^t3m%gvTcLB`f_{J1dB-)wH-<k6!-LqI&N
zLk8gv^J{`D_AJ@mRPu*JJCnRDmFl};!B;b+T1(OSg8>R}?#lRFkk^wfn!&`sAa0xt
z1S;6J<3Sc0{;Uk0PeO`9kg-?EUXsm>veJ;$@&SR+^7AMjaKOK?Sg^$2aVEpd`kAW|
z^qcVT?AM+Wo4%BqNwq{t@!XU6NLu>#z64=(dezuS!6KZp6*jUPBHA%Fcgc%*4_U7{
zj1V&2Rxv6MLy=?RX52mV3(!fEmR_A+xnH1EZBh^;xBO>8f#X&ipMED<xAAgR*`3Ys
z`jRE1?-~{7JWRtEMonsBIk4Y91<%8pkGz@+y<sC0aFsFXTQZVVongFIRA8|1j_5RL
zJfIkL#oH9x0jbsWjZ6Y$19o-0M_>ATwKpjmUL?-|p=b;~hpOed;n~gHKAL8{deB<6
zt^I+x^O9cTpFU=f5;&G!r~KR144E%EP{-5V@#69+OOorCY7y6)3a$E{>(@%lsKyUZ
zhB5sQBZNNMnC@#d@9u|#6LN=~oZDM`k%OGSiSm#(i|$!uql5mWRaIIZR$0Fvda!WJ
zzf)Vyt!D90p1j&2^+`QrDFAsvD`zEN%JjP`og^7K;~ZbqExN)eSilEH&Ms|Wv1`*A
z)jyTgy`^-GFJMRn*$hlxLo7|%@}g{--duwcm(&O@rPsboQA9gBrd`_kD!*HVrPPHr
zt|5&NLE4dc5z%x;XPE6cb_5@V2}Eu0?ru=WmzO_fnk$s74$m7FJw_cy+bcfsiZAf%
zy0;*3<L<hQv-xde;98XI?~?6aCo6te^j0_SeNw1D0~X~tbZCADf!svi+2@3tYLLE!
zk1ehs_=zp%MZfz-6Ab;Eo;jR5D{$mOY~&cMDe9tQ)e)u6>}uaMxZ6<mW_t5$f#0hK
zb#yx_n$c<~Oz5G|cO5@n0Wa!~_}^^$SN(n-H@v>634i-K#_j5`>Ia42f{)B4vuDpc
z<Rr6eyii>b^!PA$B?ZZAlfQN6c3cRQRO{Hr-Ko7l^vkWA{G{UG2W7jk#odb;?!MN?
z`D5<BD7)xiUWh35z1-$+mhI;?BEOuQp{M5?pYwG|mP3E_Yn31|HpUQAm>vGUy_C6@
zs1hu4ZBUV-!);_8;mpYIXo9$_*ir~Gci=(%<^<JqDfe;CLj5|SBiYg!sQ^oA?MH%!
zs;)DxdL+8Z_!X+M!`v|rJ|7HDptwY|lhrsA7C8<1t(8qj{vYPfJe<w+5BHtlOxx*R
zrHg%PElP*jmzY*iOJs;t?PG6fsC{RqRZT2Otr1&VOAtiuJ5xn0A+>}O%hVQID6t1U
z@BGfWu5<o6|DEeP{&iiic)i}|dG7Ch-=DHyUl}#HQ$`6%Cq|-TqEC)TUMLUf-5Mh%
z`Up&ijMxX*W-iH6<mP1On#dU^xmu7d_B<h?`8=FHy0hNg6blViyVn<YkJ6AQ8>eFj
zgRR3&Lpfjv#$j_ARp##<E%DwLObnmZ=d#M~{4ZU4mqP0br#liQYf@u#y;5lEb&TQX
z^t}A#Ohm8#_&V)ARL_ChHCSz6pcCZhNtuqc2O)0BkufI(3|1DMu}K9-Gz+if;ESH|
z^FLKvqxR(ODZ_)XQpKd^9ROB?P%#9$(Q;k-`*G@@*T>=e>U%$6y!l&lRxc~*<lNG-
z(i?u{(4Hj1FBQqimlVoO5}1k5D7HQ^3b?k*SBK>1y<{*s%5q1obmfaHH3x4iECgpz
zNli;A>spW%Ld80-N&SEtnyxlD77Tr=9v++BRQaK#(M4cP+jmeu!2HCGJH#`BP@@28
z>1;zGh{+V|D5p``;G^2rrAI9uFu&tf$y?&Dl#s?#d2l~h$cH*CO4NZd>e)+HyS0*;
z1y>T^MS&-7pCjg@GLC{~j+EUz?_}IuaAP6p+-N|(na*rvt9my3)WAEVMiI0T?GpiP
zWg}}@FwW28wpl+tuUq7-c5r1^Alz@qx67d|9znOV!rz3C+|-lauzF!Vlj?=NNY2vY
zzZ2gu{Cr*U=ko4VZ}&cDmtP5?T?RoGMsPc{ci^?VgD?3W8#8*vizm50FvgD1L8~VN
zV8N@R+!2g%=QyCLHqh{<zUY^@In8{m$25)V-xRFyK7Z|i*L`>LLmTEc#mSs+XSUH^
zO(svuolw#0Z>R&Jo5FZNzwT;)FSVA^RH@b^kCHADBzEy6iZIS^ND#=W<Gy3s{?uO<
z4EG%{@#Cv)TOH6&oj1Gxd3HQLrE+VV{j3JLg@+>a<1^D{C*w9~JG-5`kS>tbv6e3s
zyiD|=;uG8a6cxMzS3&Xie1rO}O5;!!b3j&#bI--(R}>L4<uF8G*s2Qh34du2nWrl7
zDzltN>MP1xC9T)3%EU*C{2(aU>;5k^%!F^4>?%)x7VER&8rmithZ@YGOpd&8_R7~R
zg=3)b6GF2ipi~vLjOa(4j($J$adX|C>E&lo_-=Z%nsl%6uiMCeig=vou4Q1CFAs=|
z7^3<djC6+9vi^f_gif3ED6I+|)I>w&%Ir%@a98qeM_fOo&#We|s1lT1mHXKms#bPW
zuJRoTOk=p0^0Y<+U=p;YcJ}_A=m>B0R+yo%?G^P^OgrhObSNLKv#Uwdog%ZE+1xSO
z(KD^py=35^@Lm&~<;^13>K79-$t1B57SNlLCyQzHC*Q;5W?+Z+Ed{d5m_cI#SLDWp
z60>Ko4+vuiwFp@f%t|v`>Q%v3PeGr8Av4vy!G!N;pko&XS54x4K157pf-wy>&l9a4
zQje~>Ur}Ksy}9WR_0<Pg0T_*p)s<E9wB;VzB_A8nmM!mL%gh`Dr&PEZM2Q4ij^MJC
zyHsp5SC{ijGSKYt5oiEo2A$hLz|xp45wi+^D|<hA7&Bu;-AIZwxv9#-BTS9cCg|7>
z+;FYp+TXqrk#Hm|_ATjc{hrx5k^FKvF2{KWip!s&pr$hOa>>Xg0`Q(#lE#9;Ch_5`
z{F&Zwo|s+L4;h~q>jGDyNr39zCX1Vl%Wnx{i#E5l{=*5@Z}QCW-->9|MEp4l!P&K7
zZSB_mN}tjs8;8;Jb#H`>b61KPcjmjD+|v&=DR`>@cNdfn0R8c+6t|?e+FOZ>$etG6
zU<8^DLG=AQVK5#=D+Iz2MDweCu^sHNR_onjN@s-JdPW*uK)}@eHqCjU{lsb>(10Ch
zNO{Eclpea>q#BvCT4(Zdl^dky$0^IYRfBVkuk>EMaMc9jrM$0ZKZJB0ZRE+aNMt=h
ziajV*7#6+C83{;08S3w<-ut_YR*wjUyAv>R?xjE|J!C1<D+4Pg-j2W4?&hi4W|mVv
zz}#iIIoszY82iwC$NB6}Fs10+%u<rZDq&8XK2R36AK73l{KwD*!rl+&`UERiKA%5Q
zw<K$J!kZ=Bt_#2HEQ)(H+$UNC2`N=k$G#L=@mCTzx6eE$HB*_tmR@tr9sjoZwB0Ol
zbvc9F)!5Yt4TPJH2o0!`$R%^q1PlhfXJh*c^kR0bMyn~7=-!RWj=%W_@G5nW-xpVi
zE6!_-E5eqxDs!0VT-%C=8Jh<B)9~9zYKear1a-&-Rom1#dFxledt@xGCug`9s`fV2
z6ePQQ8r>FOx*w4v=Qe~3D~oiIU0g*mRG($Rc@TqyOc~<DHVaMsmc~gW``B#XA4$re
zV^$*iWSZ^li!$X0%x{j1*0Z46v~V3Z?G{j(f98z;N+-j|8y0R0oxJJXTt#kLZT@xO
zZj#K{jjq&M9q}e3A*NB7U`e$TGN4h?&w9mAGD%cYIIrZ8Ed)4~Z~D_Z4QYU)cuV_B
zpzm{P9Jqyz^z=K&Q<8K)&)HNv-vpMUCrb~Qv`McPgU5B=SJLdngVdzHsCMvq!hm7N
zlE%LOA&fo-*M{-3*?Y&^={d)LG-vcGdqyE@Bh7lg<rsVvObOWo8HrGgK}}C=)TGC1
z;suyz(^{UqExB3j^3E^aIL@`O;V&pq6;5ydwhau>+nlxhIg4AHyVSX1ottyJFB@x~
zmufKT35gBD3JHTCua|xhxyxLMWDL(U?-J*GZ=b!^Il#$@n;?~cZx?{s8F@m(9H6{(
zo<;C%@a)EM)V3TTQ0^adqns-3)SNe{8XXk19~xT?+p+}VGk#Pv<vNILk4nglqLfK;
zzU{8*6Y7VDCdf`j3mQQ|Zg7m5ybq=JbTptd=EM)UZBfG*!ji{V4mXMQDKD+udk)u(
z3hloUZ*(A3HgSxwuwA>@*3Fc?|3EZIqwJ6e{uQyOGGsbjsOp(8cSX)kDibKKcbNb?
zKIiP79-Lkj(<E<h5e|>0#da9OD?4c;EP5gB$nLo1SewdA&;yxgCeUbfLMDztAOeZ9
zZ30L4Jig~YXZaJqpBW#*19%hKwh8s09nkHMZ2twfe~$9N5D(fKaUBv)(!Ikv_y&Zz
z$B4RH@X>pIcRAq#8tvJr^5;{x@S9@bn}rRr4+66p$b^V6@jXR3-##m8TB&nHT<<-N
z8)-KemDl2JEb1>6ezDI^?E>-8h3HhA&)8gPDS<Q3p2t1t%4x7)tt(hF`&@aM`$Vbt
zkmTW#vNR;+I3XtJw*aBECEdED!1wc(_;y0ag#rO<$$-ZgPjfubFXP`CE~Z1IGC6zW
z5E?Z)nsX{&H6H2Qt@`~;fyQ2?`?>1(AMd%^w)3YLOH7!;6ZJ}p(OzI&Ad_kQ+E$33
zX|YpKu0LcP0|CbFJHSH>T8O!FrRyNU_fCG}_=30G@C50}t-22ZBmD2s)$txIyLuOO
z?oS`3o@XAw1rX2UHMX-gI8L<HuOFZ?=X?1^csjS3?alRz2kIo?kQy}@MVSSUeOkZd
zKY4QLg+OVQ{FrTJnx|RLN4!6VCI6uxwh)M@auNGj;+4R>{1~<TBrrW-I6SB1DnzEd
z7Y2=y0iu$^^X$VzoTMcKqz#LTii*y%tH-sqcA@6cXGu3-PdF?!gdbedz7K!w(*0*s
z^W|A7a*_bqcG~i+ZCm>K=x7nYr{9IaGUicgX_Z(;W@I)OQ0;Nw{>dcC9vD`y+uE=S
zrmS}Q4Sm#JXKYyC4Eplv#dg-T$ou7ADqFk}2?qG@X9CqJ!=~OPgy)S;HuEqe$dd(v
ziSHOKr-DQbrbV2zF?0~&rvQtKVC2BlDBSHO)m1~c4*<cQfpSp2^W%oKolCt?looc(
z;pzM#BI{z1-UT;V%}>>dycHmnQi`xtTSdnF$_XVGC<aOMw_5MAGXya?Cp}tugZ15X
z-YKF|#gMkM*|LPwC5#s-uG&dCso3Oa5^lQqgWZD4*kO04Oo@@Pc~(V<RIYnLQMyaU
zJga{GfU{-J(B2JVecS%3mK88MBW|r)j2PXx?W6Af14-dyrQe8TaSKj0z&cTJxB_`u
z?T@2W!nKn}jjH)#;!Cau1`+;-JapYR2jsXVcX9M4XVd8V`<{Q@0<65fA2my;vVWCU
z*{6A@J3s&O<k8!<s{1Av?NE!qZu2`T7eb*8o*oj(VHQuF7UeQ|>FHYLAW@<yIQr^I
zVVB}SM5JSL>d9iDuis@C$^1$u-alr{Z4t1zL*7K=c<`mI8Mo0=f{dI=tRW3opR~O$
z<f{$S6#?KTy^1tnDr1)`wDh7naw=T^0KKB*7(IAwrtq%S@d}l%F0U%vOQv9{_ugqy
zne2jKtU<C(e}Nl46Wer>!Z}o5*{)bUy}@V`KFVn}#T$DEy{a*;^QGng!Vt>0s6KJ3
zLSCYDD4N)Lob6tlbYe|LDO+#$cb84A$T!|7^iQDOS-`GE#*s1r$KFvRo3kqh=sV)G
z{hsrlgZm)?8`Z`;T;Haez511N6;Cby4*ELdC$1VFZm_cW6wxR{Vm-0`{M5+Pvd1fO
zxvKPSr0w(FVsm}VPO2h{>U|m!b=sw33X6V#qYg$zfp?a?O<*cnmH|ysa_^cb-3hCF
zS(>45t3!X~ah>v9_RoqMywK%~xd8eQix<CfRIMy<1aVbgna@3Sbi8Uqa@N3x?!9=W
zvp8o5RrM&iymn5%jj>caxLDJG&)!i7cy)h0Zo;7Cd9}!HT1kCc`j&O+vsPmQxhNyY
z(R*pVT|#^F#F$fyxus(4@z#tMwN<C-Y@~0>GB|GGQ214P@Zr^DWY_L)*21LlhlaXO
z$TIo+q9%`#sTvB;-z=y@>zGJ*oW$J=Vn)6Umag6DzC74`*Kk8m)vF<Ke3R*aNSfZm
z<SOQ88Woy^wyYsCv2T`)Es>OQd6uZ^Y)FdBADzWJwQsd@#(5m2>@Y|Z6J%hQVyEfW
z=H}wJ(}>Q7RWeZE{RtS=%;d6u4^H~z;usJwGIsJT`~@?L!E{$#w6})9axFnt3ipvx
zBUnk0i#6~&l2lEiEH&QpND7~`bID3k2JXG^sqo?PX}j(en(Q3T*}|Aav4geZQ1FLU
z_8x7|%Wn%eFhWqkaXBs3t^R&nlR&Dv{y=fRiPHV&j$$j+lD~9zB+XP~<gnrd^nRmz
z2!`k<s~ikDNZs_|O|0P0T@hgp3TEz5YW9Dk%3Vp5N@LtBeXiNC;tdZRCK%f%G!8E`
zJAbU#kXIUwsf%b;rZy834(T**>vT&t(AEc{9f#~iaF0*rcPFt4n=TMn&%!6|UMgm-
zE}rj8{2!w%p!(IE-B+!MvP-I0g+eb=L!}L3XOd*Cpq|lr&m=@OJ<$N4xN(y@%Vhnl
zwruGf0Io^!;X@A|ZB8V=TldWEhJhR10o(Rw@4E>_&qr(wa<kt^K#DJUw%KIm{PG29
zM$(WmUj6)7g-}88mn|tO%01pDk%5lLy-<mxe4CIqxxwz7Nrzn0f0`4og01g^36K*b
zSwK3kUu$}$lElNG9G{fL!;y`2q^_SvQ*o6#{u?+d=kYn#PP1d63+$hHZunS8lVB#b
z2?O?q%huj1?<J$^d8Nwl%VT~@31r-jbowO$a}JWXbgu{-e)uohoz-g(2P|3H^k!n1
zL|to%VAsfo)C0Dfef@2>GG+A{ex#45=@`T6^F&|uNZfjN;gYpCMdg**6Y_NI%|XM5
ze&Xs8sgsQtR{?o&o?g$xI@VybHn4nf5Tla@)F4cQ*rylP+sg6&hW_UNnMC}wW>b1`
z+3JH)A(Vty6(7o<>@;CF^%ab3uS$L}tC=?rOilh=!!g=w{GI;EF+g(~_$1eA?1$a`
zx2K^WA9zfD9S^?$F<!|n@^b2kYXFV!o(io8{ahGqVqEZ{Ug8nn{{h7=05n8t3Yl}(
z3XQDfaQfn82}{yzr=q%EsDroO-tOZ;y=jY2r!iKmQoq+Y3q-75D3a)Gsx->ZN`D%7
zRkDNhC%&N27&*pFUjSFuzBfpTb0Hjr!x46d-4W?~xvOLEP+or^GMP<wb!*zA=!Q`D
z!;9YL|B=MlFK%DokbJ2;b?0q<)w=4^$Y>A9%|!nttEuSLOPB<e?JG(>j(+n6O<2h!
zZ_W~_oM);SWcv;I<GJMhOLG2(^$D7%maLjj*9v2*-YDI-n)cN0%9G(;Em=#MT9k6x
z7Re-K2a{3S$r|b;?E`LzISAqK{R~;-WB~n@(bVeS#29d$s9Ra84lbQkU!ane_A;D`
zsx-;H>`ztRAHS4W-*QgvO4RSV40P-JxoH=Lm&9oCpKFpMiMND%v^R71qF*-264>A#
z&haGcPBK;4`bK(9Rm?xb8ZEVjHA%TQD8RWOln4K2wK+_?QIsLzarzUTao{WZ>0oag
zXh`~ZXJ_l!mF0hA`<flBJH)DzEyMr5>s<Z0_S{`tRh8P*8_C>rC+wJz!4|g5ufHSh
zZi3UN3HO<vvbTGw0&yBFCN=vcT9k<UP=N(&1I;I_jbruwik!riHm`Tun$O7I;S`(_
zj&wK{@JZ>qzGkdjS770yiTAG@_w?B}-_Ha^Wcb!fO1a{%M_axn>w0|!3AZ!?MU>sf
z(Un$A?t304bmshZHe;@o*`2z`TOKIpj_^rx&)iDWFVnlDN^Er0zX-%H-j9LKfo4J~
zeI?Qtj6em=k_}r>nKcLZwz$aIgwtnhx$1ks_6MKhyw~d1G{86KDC#x&v4P2U$@hVI
zEUHO-Em!t;n{ZCB|II{kscxZ8s4Vmpr|N*LzXg<_y%{BE=BY2Kv%$W{aiOc|+$hTF
zroXa}Nf7tdK=1?Hx{C+z-0!P=-FJKFyI00+Lyv8n-sesC_@>EVc*H>VkhNRs)lkgU
z9oOnBG4E$8jm%*>taWH$xH$W`U3be_2Q_?EvRS0VZVIZfil{gk6PA^1EBa%P5xd|z
z+QKm}g59giwA~o#_Ireo8=Uynq2iD1Do0~&55b!OaGmyGVtxM>w@tWYm8??@#9dXV
zJ%hI717Wf09*t~~dVGvKrelD5^T*;8H6Gep+1-yV<sQv0y-o{NN23eUj!uz*0mT*S
zsp@mdSTJ#N4ydC1cCcrq5=B>Jvv(u7;LdRNnq5e|&;Hr20J;IkP<dA5V+jFMW{d7|
z?uL@|Q;k&mpn-mK9ePrv{F2zIzu^2BVOq^Vih_)Smd|zoKx>#5+4R{w1}_T3nX7Z*
zho_G1^K93FGVdRQ1f6DD?9(=i)WR)g{UJ)(YO<-ww7XZ$FbeiJYf<ruit>R^B0s50
z98?eN=bJhip=OK}a#Y|vLR4>XdqhOIEUAPr7tY~qoI1vPHKc9^YcMB}OK+Y@x`vm3
z^fGODTgeGcHQmv&MLx^C*y#I*ypgxtG<y<nZ#MdB!(A)WvM6~j`IA@Ne{&Mw&n#q;
z=W9Z_C$m|Wfrh9S*dDv^mLS&Po2v&QO4Oknk%NqzS5qh*9J|#&_F`^#$oADA=%3+5
z8j3DfWJo|u16JwTBmGEaZa#r^xW!A~HfLB7LC3lhAD&ZvpL-L%Gp<$&8I*1)IQ5nH
zc|DaRX<sH`O=|0*`jVMxQ{Mf5EPhnEH~M1XS1&M8NeC_kyK3W+!P@+7?0nnFwrkT?
zgK!fvFfF}sRSHpR9V<_=*c3+yOQp|^@&`^C)#?2FS=`V2ikD)BX|tL{wQyaQuPe6?
zRUfJ;UXK@<3Ee!^eHBj$KaM;(IKpFLmm;`kGiINEEw`FRjTO{W2xAH@dAY@7)Az$V
zXsezI_uLHeeEBUH)2He~?RK5I#p|%&25N60vC?Slkgh0hWAocloO>vi!xEX;ub)Mm
zg|`XBq3pVA@cO<9rN2onv*(l|_99Dp%1gxusT6+NEt<+8vWramVRs9iUzTH~ByOc(
z$Y)s3<kf9N1cqz*vX2f8s8`sAbNm<-Nm!K`^!*HaRe?~O<p%5SRY1-Cow3;Y`ftB6
zY5zu-n+BtZEc^3Us#FAM=uwC<FG2@S*AANRKP-tOiKD-Q1%q3r>U;{uHoZkJ77n8<
z(v&uu$Y<}@t-puFG!!s&I>x3jk1d8IKZT+M=meFX5oV9MawmdkbDlVTe6tI#2wJ3q
z&4CF0;a$<Nn=M9rh*+nlq!gdm$8)6$qnT4T<2)~ik$x&JCBh>kEF8m7N|Tn(@sJ&r
zySi3mJxE9-U$GwqK8vfz5UEXAPu%m9_JJW);vfO}*H_gb7#Z;zE1oz=lyObABP6f9
zf1G;M3-Fm?;JGRDyD~-sxV&Tqh`L?Pp&>)Gv7kjim>~X4zyuEBrQkxzlZnrlKt;7m
z@#P7lq{&CD>3*!qqXlZY$(9W=rWkX`ar(7Z@@jD(`QLTW{J-m9<}ZnbTbA|1y_$Qc
zn`K@q2wLQE-#IJ<C)&v1ZH(CQ5`IX&JF4crV_BFku5|@fxXY&|Vm#`Dm*zVRI%?cM
zY1e7}Jik-{_BaS{BOSb8PA^h6YksbZ=+?g1l=9kM3&S!hcQaorlOSpN%T~OnniKui
zP-j1_%}W~Vf##qlhNHW72dh}=uj_umpo%()EvvRsb81CxdOA!h6CFzTrvF-bwMSE}
z8xhW0Pb0sXf0$0P^%V)RvF-_FNfe@eejRx3x-(_60<sMR&f5q!@%bJw{B}HOSON%k
zAZj&^=;a&W@7BCDRBqFu<aK4GkC7<*%9}|T6HL<c1+AAD)uxnFlVnOFW>zw9Vc#E=
zO7h@MhkAcdgz@S$8&-eYHieoi@`7euawFSyhAHaV@BTgX{lEOWwkD`xwfANbJv*PE
zv6B0Yq%(C2dXO1DMTsgt7T<R4u=v0J^u(D5+yts>$tUq}oBkx2^TjpQ!{QL^J$#`W
zPmt!Dl2pHdd9&5|a*q+)YJgvqbX81K_$~Ui)FVOjU-N%Cm|AB&_E!6$8Cta*dDfF%
z9_ZmvgNNlmd;~QL<bM)q2$PwRT|l~De(G0X@*#Co>h4EpNXaF=%MCSerr-TvKiK~#
z-#nX@j#n8GO-6FFCzgb^ZU(XTJh72eRN^=8|F*&3eh+kF+zldjJBZ#noHhS^a4bUg
zNNB}9c+3qwBM6k;05x<)9?vvI<G-IF9sj9g04cKV-j;aZB`H@>vvR|<zp(G8mw<4l
z#Mb563ko$Hc?04PnSy;kV+oMax-AZ?#iVabc?>FUM$fGx)vi^CCTfi;p~kNpkFMP;
zPR@rhgm3bPB=m4^0Hq`$kGpo~<n2~%f_Z@up#@;nAAYc<t2^%$;tSO+Rk;UfaacUK
z{b+_-zr8-Ic7=)<wi&HnP6|kO&tGu(&r)RJ{Q3bm^p>`(Yrr$tdqu76pzL)65&XF<
zF~2&KfkN+xNn+p62*vGWkZK4u+h5zI9@xB++<nkAYmzq~(~?tR*J#lB`B`3532BjL
z?5z)?A{22ccKUXqRD?2v+9np6)cxRdxot(%+U~f%3c`B~@nkgvlj)V6?J0$1^#R8I
zT$j_I7^2^j1;y$}pGvhn!`5CKz96DY@01y<C-e5FYz^$FqcrW%z0a158+ih_8VNpL
zV!P;y(dkAnid5#292rd}2HJy!PFmY8r-Tc19p?Gi)cZ3S48yW|XrxL|E^-w!SCJ6R
zO{#24CA|vNis+}Lrn>NCcZI<;Jt0p^Ndo}2!h+Ucs)X%azF{X0PD#uZe1WMcpi8kY
zWN~4>&R;Ou<e~@YqE-ZSN<gMoDB4qsTs{sl4TXV#tdYx=uY6yonf`L(b@GaDgy93q
z=BoY_tKb&EXYXxGj@;^mDPEAX9jKWkDLk@8)d;=rwHZzyj=?`IO8rVGc>7q~yo`=F
ztU0{iHFOhdZVDOjx8f1<w`9=qA`zVLXU>0fGa_f?c@ofo8vzK@W)xxFL-&2zkaHQz
zR?y8?_lm{uyF2jWmFW+wO|EHqo<3Bn8P0#`uloKIVF<<$wUS7UXe-Kn#wcPSDl03Z
z$1OeY_*;Q!Be#~Tp|j5c14`2}(baTh_d{;dNuU--ihkndK+NLl^t}n~+Z^`{-S6U?
zn$uzUgI^3jEiiCdKQ~m&R{f}xXw#d-D7>e(s1-437da18$ci&T6^v4nc79I=DH+{a
zu2vYgLWi6npymcpG7{*ycm*3;Wee@CU_I)+Bd7Y?Kh=BVuKH;r(QDn!2!Eava$5>O
zr-RdZgb((@o7-)d74u#kQQq8)omUERlPD5+)+FQS7W+^j;!uu#RPWmVmQ!j@R1rY8
z2*19iFz(-Jvq~;0O(@-g(~kw@=dGYTTWC_Iv({k%H-4|2`E4fd8R=eJ<#Q|GJghRF
zrXNT<_9T;wCTMi`iDe7j!>cPFg8&utMxYz5G04v)DN5`})Q!*(y*^X){ft&K_iGrc
zoqM+M`<VfGV1)5>IRp%jj!%kuv692&1S&`tT}e%zks3p(<}U+wZqv`N<?8c{tt2uV
zMGcAI8T<Ei+>iQ@VA<)ewDp(%zps~n!%I?p^ytCA=(=P-?wMXG>oe}<R+nk%NvKY`
zO7Un6TYP@JNC`os59e9me9eUq_ajbyT&NG8z6gDKmALFGvr;`6FXy9H``4ART&OTa
zcIB;eTFHC&I!S9C4HJw|vUjkym5A@XN0SzFXb!QO^$i*kv`^hQJXvY#ey~!hBh~E}
z&Qz~e4hT7Wr97>mD0%eh3ul*avr&3jQHkX1la_9P!D(0KAwPj<hz+Qu`+yumWm}0(
zU?YdP*{?Sq<?j30w3mm|hsuABX+8fpsLXdNr95GE>p1P8N}QzMt^bsk;YY-zC}>)x
zDw@sl*x%?)i83`Tct5v<4Z%j7P_C6Quw%iui@5KHrW`E-PI$<9vO=EmWg4|A<NwqL
zZe@D9E3DnG7nZ@DUyWFMi{2u9D77=9RSi3=W_>Nq^&skg{dN#>*jO@;o%I>-^!xJq
z#_wtJUdGp7s>paM8C#_8l^cV=Mq&yPG#a|mAN=tmF~;b3R{i)RyofC%|6#N6sCG<=
zbWeZ5WGW<Y0lCQBYk9b~E&JT>G?pdW|IfJ1+{D@EeV1R7r!o<!a{Va}S=N}TUY}I9
zdl$wlmXy|CwuY%mD$yTw4#^<mzOo@yMa9iW;KCPSg8?HFIVHB|rfwOM(wFMp)rux0
zr?dHZ1Thh%<Z9!ttmI#A8#d05xhZo#Z{2HdziXjs4EA=ez<oPz6LCC$jThFY&ZbAt
z^c;hOvnmejDp^c}X_T@UV?evOx;j2hFaF|BB`Z|cxRvebY&Y3@cpw8=ljs5ue%z_<
ze-{Mwz$-kmPLsnng=_+<m=4jSx;i<N6mIgU)=~d~f7=6%)4Oiv)={(1{d+3qU*za_
zLA$*rlv-rdg#^O(1sV$AJZ#A)*1uU~5svaJMZ}+vetk<Ju2Z!QeSGk=lWfkqFXN=-
z=rk&_4aMyXG4{OW`_Ln9DFU+pL+bo(dF6pN-DnY1txnec$L=ojgB3(#PrT$66KrSt
z@S3r)h+B~bWFBcPjwWhG4T^6q%temun*`UCJ}Rx7t@TiL&VP`SjrjT$>~rI)p#q;L
ziv&_>?$>{BE@qP1Zx6v#fO>D(ft2|Pg$76kjZLfS7AoU78ffeCGu#9BwZ~~U2#kz%
zd~OYHaBdS+3ku_SAe_9JKSL9m$PtFO(yVNb^Nq9@)Zq&9i&Q5Q)z7;LZ>}4``&st>
zRG(;KJ8OUs&gQn`=-Z}qTM-CmfUA%5&I7kAO6D28(`7NuW8ysb;<TO^|FLv8S+Xp7
z(oy2Zld44~)6i;t2gp~qkbxNmC5m1h1KCRYjGS;!{A!!{i9aysAv4#YD*^M(sME~<
zKTX6d@F!kaWaOuAzsEKE=2I&He^~YPI2dQHWj`fj*owf>{2H+w>N_I^4%s{AW=1!)
z(!OktT$<RMRPGKv5+4gyKiMcJ-G2C~Fy{wjO7`g3?^Ub(f2J$=r6f=?h5t0SS@QV>
z1Ehg?L9;b<kqz(+0GI^!@sLPh7fjy_FnaaXQXug4T*x`OD^15<7pby}p7)@)B@D{)
zUk*e-f_MT_NfXk-)+camiTAzH!eXir9R%9+>-6RM88O!@S~ex}$u!yHGo$ZpW{c7C
z(m_*;c53vY$=~|4V<gx>uAa@+jmNgDAGJkAX9iq<r7S&R965*s=@xOYYH~z{2WCc}
zf=#Lh;Jx#$(x#Pmu`f{_d>sW4Uz(4(?6y#x=7jO>arNE*@JrU%OKJ|KxK19~y^$Az
z?|wh?v+p=d7|>^;b&3A(WSfn-N6WVZ;rZ5MwtCJTXJZVLq$GNhCrySH3?mC=>a7E7
z3sL2ll#~g6i#;^oNN-pPx?UbP<7OK;Ugp8CH>@OgA}~5>qh1!&ljh|SYY+MB3U*kz
zRmo(a8>v6}M%cK26&*9e619ukK`iQ8p0c!re%(t+_Wtp4wMK^fPN2s_dEfPwgZd-i
zKPfKPg%hRnD*Lj@#$t=31)M!E{Y0mh1xfh}Bol)b&f}$Aot6lVV>?B9HL;rbPtv&%
zLufEO;L{V8OW^{gCBaQr__;%{a`p_N)av&QjpT<tdGJpkYG}_L|D+v$8=h>6(WX@H
zYQ_o$uOft~yh7AZSaHC4d*@_01H;OdrW0k3iw;Pa-v4vJh4mj^t55P@7gF#*W8T`O
zF5Ft`vg=Ov@{@H$SzJ$H^^6V4l^OXx5D3=_KjWO!q2Y1+QDCgTE{v{P!vxtmlpDOc
zQS~m{*ydKQu`$z!RAug`q|8w;c!cEO%5VDffC0zkRJ`}OgwIr_LEiq4v<KpZ(mD&J
zS1a$~=oH-V^_Fk&17_LqgtQ*l^hn1*9TlUcuTA+BAUWc(`R~K+94DrI(%yU=6<HoW
zS6!wq<<)pnj&x9a#;DUP(o9U8bffD9;@`u2>wA~s|0x`QTNuq&aw;@V?J&jPO#1+n
zz#9Jt8$#f=wX>KTK!FG@n7)v!gJ5XOL2CN%F|O!UIoI=aA%%0E)hFW-8gkDF&UyFY
zY+yCkOB{dr=Y&jeva4#}4fnP_@mv?2u1hjSJvS<f5=qffJUTA=fsyTxzF7D5hQic(
zU`Kb~o13;~M(<*Wpkc;$4V|b1P=!lAf((BLXC1-ut-kwwiu?rziG%+HsgP{&7%!;{
ziJoRxo_7s`(w`XAlxOszM`N^Kt~vES$7P4U$dT+y<L@v@c#P~$Pt7!abJoFsm%iyw
z6*3gD0|uqv&%_-A+lz6a9WA(ia$87Z^@Lbddwi<vQge4ck)#`qsv$R~q2KO8RSV{~
zD<++)eSda37@2>RCY{9pK31ZkK<1OtUD?Q=D8@FA*Z#|s1kjQ6de(mZxW{3)WO1|W
z(B(H5qXdexMzU?vt~<#H+uxtOSiG}m>8<?g`<dpRlAMbW=}cujqjtooq_)rlzz^v<
z9z;byCl_He=GjN5(N_f}hrsqC`0D<yg(J}iq8<qrB_dPs3~5zRc`?}C_%9qD5)aT)
z4hl<vM)CblhaZ7{KJXCYe`1n>%x1N&$L4F+Cjuq5C+HvI54VVlUn1L2RcgS)TSU-a
z3s}K!u0rF8wVxX84xXKr9|P_|{9&i;!*hC~Kq){Pu9bF}j{nWLSpyH*SwiyoFGPyW
zxZCwKIU*(_`I4-@pZRficH{6=LH7g+?4oaKom?GegB^6w$M>=iv?E?$HM|c*4E^Md
z=HeU5by6xO2B$zK)*7fbE&SKzD4jN|ivXVDtjl_z`_b!}g`B5QrkZfcHIhr9tbgg^
z?Tn=&uHxe2`XOik*p8PDV&{e&4fAHch#gS5@zOxPSo@I*6VxrKD`43qly|wb$!g~1
zmJ#Oe5m)=jR81uA=sM@B`fwBex`~m$f-8okQ^WZ-ryJ)kOH?OEwK<HSMbf&425+ey
zK`)7iE4=<|bj3kMP%4E#!%yKUW%ROJ-BRn#H;W(oreV%m-g@k$+u|1SL``vSG|I6H
zbP^ZUK2YDHyINWjWLqzLl%Xi+tVP<r<;NfM$$Ut5_{;Q%st?KO$;m;kE6E~JdTT2(
z)LWh^`~*%Mi*VE|n<tKqtUk;q-u|N>>+oz-0a7kyb49;@6{G3@$lZ7gkL20oEV&Z@
z=quSYjP_trHy}Zc+Tqa((cvB;A3;j)RR45Ac!la1^tE=4KggKhlewDiN{QsFvlp4M
zOQ93d`n=j(>wCecB90!>VP5EYqR-~x>6P~rD!4$Dxc6X3hPR=P)TiHHE3=d(n8n!F
z?;7P+fgABK|HEgbGJyww7kC<VSc8#O`aGTf!Y8fY&qxgvr`8Ij=Ud=909KdB{E`Qa
zt1aSKPqQdLPS$SMJI?)j`s>w-iN&qsNFw??rSaiS{SwZ%Cj(rMsX;CgR^RV((ndcS
zn?2n(iTi$rtG~%jqo2%?&@Xnp@<{|_^x>0N>_s}^qaiq=G3hWgBBDj(d?$tl?;u7=
zIv#Xue;X51ZkYkxG2|6s`d(#lARvii%01i$2Q^gg9cU>DaJY{AJbQcC3MZIg%B^Uw
zf#a!IJ+gg1ry4kN1a$&F6x&m6MX8m1<(a{n)%)%?IN_R=@kd%nsmX`rmv9O}@@!8c
zb=sFv+~{vJnr}|n+Yf)Ti14uos4e7@d2`@Qm@Cthm$9i$eTGKkg7MGrhE41dgWlm~
zQU)nEU}a9rFd}R<cZf~M^}zY)oScq_eaPz`$`ShOqNdo#zlzmepPBO;NfuW+JMSkL
zx}{2gNbZL~3*w0$;1PvSa}#P_jg+-uZ-k1KKa55OV3nDqvQl0UXmK{T<bd1g?-T!o
zy=)`gBop_nzQhUGb_nS!nn@nSIqMAZP^lQ(qZ0dzpF)n#4J4>aL!b~DX%SmMo7TM{
z9McmI_n;+wN$tNO2)2($%C0*qIGQDfHEV%hSb3J@wHG*`$%KpyQgixpDF)|e$v}W+
zb7k@UWfS$`VeD0fA!vY^4q=5XR=gCaFol^5_x~@_h>iB1saU#s;E?OCAwSyv#Vt<j
z7IW=#qJ>0X+(O6}zYjr>(z1p@)xEW&f53oEE*W&T6Uk046LDm68Gt+dmoXS87q?3B
zIK2$#0nw}3j+0~diMqxE?P0~FIK>UPDT1oPu;<lR9ue_Jz!inUql62E1!Vr3RjitJ
z&*=A2+OP;x{cHg(9@|A-bdvg;`K%{lNLHCYChc9K+@mHVkP=9nK=E*4v?*)KW@K~$
zYWOL;|FoLH0Uvo$ur7mA>YL1LkN004i=5XOBRAn1aC@e?5J$DFtGJPglP$?>SP~1x
z<f?zIW(rqRa_VVeRf6Tgh?|4ZtJA3!yH8jqP55X1*ms9AQtao;KMfK5l<!KUETvtq
zsb`ngc+2g}#kJaV?=U6^*b#lzl3FX?MVei`qg6ctlRF|bohv$B?J!m8Jw+|6jFy@{
zz^PM88PoZtYqj=5usU&^%Y{vbeq{1;y0Gas?7ls1T&ys8t$+$`=?g)43mS==i#N}=
z%p;Y+$>?L@SS?`552bdFaUQ^RA@P=wfym-ghIvr)5byUhV#S!;p<SA<@OgBO8b?{%
zM8$A;iQQXNv{za*HdXCCV6VGx?3sZDoRVC>LGGl=Gf3yQihu7t?X%y{Fkkmm=UM&r
zc-$9ZwKpT(?Z5?>>c5}a#|O^|MQ`a^!cJRVfF<Ddx9360kofZ^0jEFRS?_o)c&k_{
zuM2tk!)uKzMMBnhe@b3yT1~Aj?inq-N3#u^X5s@;kDAjM_KwD_Fhn2}-f5*M%=#)6
zA)bNl$LrPVmy5&l7<@yf&J!n@E@)PsECDdNw9sg<Yz}R^E&s{rlgz=)+&VnP(O!D6
zCjqj7KP2qZgSDW6*xM?8Hb=%sg+Y6`*gfbmme3lu*qf)G=S3kOxzu}xkwUOD^d)ZD
z{8Q>$ylOc!Zgr``N?%zYF0@}t5+{cr(y24JL}o+X{`tB~7tAuBCgNQNW%c40pd?ik
z^zO*Yz0-oMhrF?FN5f!ux1+B~LBT4hfgy+~1M##`k&mOzd2)6Yxo5hc8y*f+w{&y+
z#BzFTy>i7e{kBN%A)Ev*VEzLy18rcC!w&t#Ln#BpyEBGg+(yjtkYmXn`o}o{fD)N8
zLIC1z_j5*IO<)%s5f`n7`+O9w=2QQ5PvdsXduhF8kBvQilVLye7tyRiHDSDCAS9N6
z@@~q1FjhBt#mk(qEw%uOXt(``y>({M#5i!o+9!CL!mS-{t@6;yOuV|pz5LFL*Kb9o
zVZbQ~ZbA18G?PzNWsX@gbS4C3=c`a;myNn@^q5{27CGNP!I_DwP`jpEOQ%jSIh?(6
zodqwE_AlPBNA<zEC_`-~G?&K>V<Bc{#$IED=H!AxwZvj4%htcGTi+wH1QroSv9YXu
zdx)~caQc@(xqltPQ<z+LkD1l9Y2#0&0qj<r53LyMS4FqtGZpr1G2U{@N~Kw~p2qad
zCWD3wwa-XlHDO;!|1;mNLg`OJ9HCII&@@SBOF^qPIYn%&+?p|Puo5+Ux3)e3_SAQu
z?YYoxv&A&}*$-4DF+E-@gP-*hkK%noNJfq#M&aAfDs$fsODb_%zMrxEYm71#xc1{z
zZi@3!R?)_XF1v&KdO5!hXRcM6Jbhj_Iu)b`m8_{{`SoNb<hiBl@)p01%`z#NRF?%h
zL=H}cxVqQ;b8|IaaKZEJ<KH7|`fu&M?{az31kIA++nJ=gB<pt-^`gw3jqG=s8mB+r
zeW2<3N<Y!G#BxPBFfpg5Dgn<sN(39{XNDeb?w=&lH+)AXDc}Zf`^TXY@&5Ib3TXw0
zaa{lURcRT<-a$B6x2SsS%uA9BamPnet3N5YV40(PA<4zp<ioL0;{m#n!v={h7}|Y5
z;|%6q$3;hdKNC%4TvdqnA>u^eTo_xveG(qg(%-dl5J~x{msXZF<WpFa*e8L*n?QE&
zN_fnHx}KwPe@B5m7QTi5dmf?ahW)QCWi0tyhzt44;5i)IZ{Ro9B1k*knvfKFnQP%5
zb=4r^ki+Y^SG0ceIZnH1?ezMdwtZzY?UhRyjwLG^*kntS-P>wdwipfYRl1i{Q2q?2
zBFIw+eJ#J3Y-w6647-o18!H$P@!yTUBe#vKI508Vv#q>8VjV;;oEZGE^MU;fS*?qh
z`>lV@VZYSEV(bm7v!`ct1+*i!c~O#Cprm{FDaETUepw$(9z&{B{9v$pTs>}X9vhi`
zk_{|33(My(2i1uCs)Xa(r$qwGG^%f^DRac#Eh}5Ldc>^Zl$;BMP-*FgD!@A@gE3WO
zXhMq1)rL`R29^v(MV4s$zHFW}SqoM_0bcRC!jMPd1*bySG`|E%+LL_C{^9$=?(vq)
zzVH6=1?yM*^*l&1R<6uxbQ|tzE&!DkX9;Z?Dl#YNn}q#vmksr$HtE>|-@_my;GB28
zsLrJko0K`al<XF%fJYV&@)M7y65{KVlK4Z<D%079;ECh#Q_r<i+6l4k&a0R?I~NxL
zozFI(3o3pUQ!-CVM&Rv=(TyROR?&xk%S7Gdu1SAK9mil%UB`Pxm?8|$hr`vL48-Y7
z2?hGG+}g|+tXJ@Uzro{|72vFb0A36!H6pcENWIE#S<)C};w66K+@u2A)7Yf*u!#}D
zo1BeY?UgyrgZcXAb(q+6Wao}uXqUucq*3vJ&Vy@z0*gaDwLfEiDU!3_59IW=-~97<
z!H(9`$OG7%#nvC7TOy9^PF`Nem%T#4pDo0eK7ViKCmuVMs-N7y>f{=cc(t&euRunc
zmXHdG=;`nIabh=$-nFy$o2IDGvfb(I8$6ZXd9`9WGP@OsoIVhJ^M!piY8TnPBb`9W
z@!%J{RMw^9AzV&+6VO)PfP*%YQ;aEDB+9V&wo1Tli>ilrCm#D36iC`NYA+w<&u$#G
z2T+&hwkz0x;_5?h{*pal+%NM@9!%qytRy}x>aNVn!VFLQ-b2+nS+#8ujSP&I3MG<X
zf|d$pB=qO`3`5cTG!5y#sFL64myXT)+?m1sr6Y2VnDgdOM<JuV$wn$PWP;ggho7y@
z<z6I4q03Yg3C7u(EEo4DzYg3Q(8V!0{aB)GTgks7)co<p+FybC0ar>_QQ2Y3NtEZc
zTro;zEh<JVVzMdwn%M%wN>?Jj+)p^5G(LE)s!$N6<0OMh5*(EATne3^pl+QywkktH
zxuK_`12GL``fExvW%9|?uT*1}e0ks6f}`45NUcR}%8X^oYgIjawL9v3ySS-dWk2;K
z3qL;BDhUDX5NI980&#~Ng{bY!V*u**Zl@NYT}Rz80=Qa1x5f8OTMYZ>Slx>dE5Um!
z;qHpVZo>-P1EXz_k@V3>`fVf;sAGx(N1ziDn$HIOH=nB?We!M57ktuV9;T^zhN+#q
zn8+*1-Zab8Qzm-4=@#*=Q<VvX+saKf1b#vY0K4kT{68A^H0WPfY9ol&oRe<nTCK8C
zb{#9D&1$%r`XZnYBQE#nle+c86|IuqVxeR*^QT@XQ$dUHBwpJ<Z?)-ONq?Bg40v|w
z>;L+{|0&U3NaO)%(Ojz}Yj5oNwv^)Yu4&WfeZ?DJ|IY3E967{)>yXx+*EGk=kXcGj
z3*gIROUlS%k=XDN9@e@#IbzOZJ*~c*dA-eGbSLfi+fpMUpLQR`DLmS?7B`7^kPNnl
zp+Kc;y!^FI7~iH55c0u9O|QDQ_V8_}i2vdS=jgcg)Xex8NTvbd@U8uDi?llJdA`xe
zw&f-G>&NINEormLJLb+E%De#^qr;C%2?ljZU;0OZd7#~zVSn;`X3FrxLSs@p8#cjx
zue{aiyLS>b7PR2@{Y=IoUR><kwk{9eC$%LeG8=7Ca$eg#tfOx7Y$`Q1<M+U1IZ98U
z<L@8Uq$+o1Oh94}@79gYr5ivXSdV`c>{4Nh+B7bgMy+PfPV5Hx0&_86k?&{z$vP^*
zX<6CL`{mf0{f(}XQQ$9pY!@_|BEA&SGaMOfI4`d7(GavrY_rQz=FoYL!JE=+pSSh^
zS>bddMC;m?DA@VNUzv9n?1HO3E_IZv$JiZ9L2RzXC2`gj>>8!JBYKWo&_Jfaup>Hw
zy5;-*%o>b%P_n#ydvY7NMF}n7kD?wAL>2Xt&a^<kF;P|C%|78rf4?tj7@?HDm^Sy8
z4I{ZtS#(-H8~oA~lR3bjPL!pJL`KF;Dq5Bj9;d#tlGm+;iOfFe?`KuVMY~g?X4^cf
z$2|2HJ$Rz-hCTe##V?Ngo3cV?{qI%QSME)f*Qd{4uNb4nBx&$72Ge(S1Sa6IFo>Kg
zUvl8<>u@0m4fwkWZ6mk{Ur{UiC=;yIz!+HSQ@UM32z;@%K7OIG*;&h8nMB9}|0CHD
zla%SwNl?ACW-gsK{8X9`X$;i|o&#V>L)iJ68a6rWUhwiB-Mf(O**2@(toBho-fKo0
z%XCph2wv;PJBbi)Bu`F5)}!tlfO;?q=o`JGDIT|!QquxkEiQ~Ed*y0uDxXx+PWOcZ
z!!@`Xr&sEjxGSovn-!ZIx0wF?ln+aFsbguk))#ktZL8WnQ2}8!M#PXkBb0|sV3|P-
ziWeP#;kn~FHhCl0+>#2*Hr?Ak1Zs10U23SU{6|~$*@k|{_6j={8ZG^m+1{JC;+5>%
zuBkRvi@ODel=QhZ__Jkw@|53X8;Kbsp6(k7_0|GSbdsPh9afDmH@|*@E_&@ROH+oT
zNOZ4Um|{nTyN!7A-viF`)<I;qT_ql+T-K9qsqH6gk~NMcz3!P!r6N?3x0Bz`JaId$
zsTRQ=@(!i<4W~(}P^(*x50fC=@F-o~aKG9}siic}P@=dBa>E^xS^%&_-#3d%6tw)5
ztk9ugAEUp{0WzM&tXa`kJ1QT|uqirD^5E9U_-~VY`8I|XZv^JXZRG!%uu<yc)pWhs
z0=p4acWFE9Cr;8I?H;b2%O%EoX0*2j(WX|Kr9_X}W4<9F4rn*x^g`0kgSgO*(7SUx
z(rJQ{f<s1(4i2|5tfqEqCUjSrIhrCP71q{}#NbgIT7Fwpx)&D0K0ZW@aP93*h#o)l
zG1toeh;R?SZJ()aYUloza_vTzhOA44^JLu1r8;G4JB3%VWF^h9L^<)$ZXu+Em(ec>
z-7i?}<_KV6a%_b}`?R-304v1T4qLUfB$oryNbp@`st$is@Q#(my|K<Bi{bl4t{H_S
zUGv{3Q(f&Dz9_egb^JjDqtxN}q@n^#INj2B^H`_-WmCgkC>PxBd_|R;IBogOZc01;
zQDCvvgsQdM;}Ntu?_1mZj~Z9X?vJ|9TNFJ^?hBQdz&oqE^%^Szdq{7cHv87uuS@yF
zFT_N<sFl+T9$#NUtqd#KH*j#ri?z4-s=e<Iir@QVUEbqqUyywQqfiFEEewVf38~y4
zPQ@#U-MvYV(&;P4PH~5N$<<;Tq6ygKj=!fpSF80sJ+1r8od~<;B2+?iZ4-)DKr+wb
zE@h@oEKEVN#zZ6O3Cm`ZDcf*@!LYYBML%V5+qCx=P3dfZVeZHxwZ484cy(2d^RU&4
zk)0zCzqF6qmj9ABi2PMueUJ?Qb*{)M#f52$=&q!^B``1vJK8z-%U5J8F)Z64k&p(~
zlhn*n-w|COU0n;g)6o-RlOP@0QT^mWskEitmgJOMtA35zy9ukOq4P>-6ZlbC7J0OU
zy?5mp#e%9r)p>>Y&&w~F%Um>0l-8|0nf)k)*dR99zNv048CmBTNvsHqyc--11CZ8g
z8nOx=qn}7YykoSFkj$D-ufZQFuVA12%o~b&85)X6I8uVA6x*i3rsAvF`x^JKV&T5C
ztb_aqug}Umt3lhi3@bgvhz&mNH;gAq(5~mL>*_g!_2Qv|OwDd{Nuy^Al5IHHyJYJ!
zG@49t7ZaZqg0)X<agV>B(L@5;)h2cN^ul0Hf^!|miw&UTFU4mDYp$Ba%QpDmN9Ib_
zt=gC4@4gX@opVH8OU`c?#*9k3Mv95cin)uq%utX^&Y|<TSG*f2C3~iOTQ}01Un4(^
zaFlG?lj`t_a>dovkhv>Ci4t*j0P5K2imcrTx~O&<#23`*-+6q3I{#0J%K55{A+dgt
z-#M~J<zf{x)}Zr^ai;T)Rjn%*?W`4Z6_7?HS0y2KJ&g=I8Vs4iJ}k)-b7={WmZ)j)
z){K-EvKWQflflu?8c!+!A2<MF+m4a-?Zl0v#tq>89~-h!4gf6lM~_}`1g%9!942<+
z8+wV52GqxP{9Cp85&A!OBPLiveLr;PAigWcIVR#H?v5zzP9fsIiICqa1S2$2q++)0
zQr`?!_|t4!%6yeyhGKZZR=C<mrG{74^Egu&chgsbq>r+AszQ3tct4sVXYOaacS7u<
zI%Y>YcH6rB7#me#)A0uB*vavO|MA5Dr^(kd%uesjs1lHV8u~_H)FF-#So{8&a*%*&
zvngyH&`r+gMhS7~<D3KWy#_!<p^A#AA7`Df62{DPMaxHROv4f7D)QM!pM<MYr_(N9
z6wix_)9DNdM0M0wt6_VaE!Ox}Du?b}NxtA(7=WaN&SNk*D*JGs{NDlL^^H%j-IKgM
z<OpS99z>hv+}i#~UiSq{q}+4(-FkFEfVB0PKEc+>-h1u~s;C@E;a`^58Ij8CFuqm>
z*sNXJMXf$<qqz-BmIEN(Nhy~J_Q`-q$@swWw>8Fq`pA_Lhu_8$Md|XHg@wI{XJxVe
zNf+(ZADF%CcdvF#G0d_l8G5p->T+|>9-Wp~o8IhJs0=v_I^2)`>n1HCs-4bR90FbP
z*1Oz;G$0X?^;uR5cAr-*YDC4^Y-99cNQ5|sb`mB+$Ib2O5PgVsKYZScCeEz`Fw=Gq
zz>TH0#o_)Y>{z{U8hzUb;QXv|wE=L-$i%^)!1q^xXaDvqfOX5@f_3A!4<g(D`0+jP
zT3y-$zhUnFvK+1pd+=%wxReNe1fWN`{`h|9+i_&>?5nQ+jPrW+efGKFSr4Me*4v9a
zm$pvYrG4V3_J4!FbD*cN4<jr37Zj3Shn-ynu6(9QZ{jHEpp`C)|MG*wU!hS%*m;0>
zd_@H2fB5g!t2Y3)!(N0g&%w*q{~R0whPL1Mmcu0%xY1^G(b|O{@;e<iM~HLBj+TIe
zeiW%IS~+?3P2(2V0pP0i{qXtQG4Kk`0J_S1*WaGko7e(~E<gSc>fSS~$+T-1MjdrT
zg#koFDN2<-Q~@EeP^3nhC{+Qe0@7PZ94kFi1BgUHx<sXisshpp5PB#=fY1VjKqv{!
zzHr{>dH1pZe&4f?{pMFf!hNr6-K$(>t@C_OT=t{xK!#{6{+jg&yKXx{@mCnYHS@$@
zLt_VBK8lPcF3%hSwR4bf3<1;aJ<RXNEtd~?hkDip311OV?@|iOr)xNdV5dDGs|vn#
z6o?6p23r@pOeY`&9dLd%xJ>UqX8Y7gq#y+q_8)W4^AJAQgKE>SX78uSbQ}^|y3BY2
zbU(ZAP2SRceMI;5-sIdx?!}HDF53;4S^d=za&pwbu<+Wt*KDs|P46Q;Mj41}{u&sV
zcj+k8uI<SCpA@kAc|+Vj_`$Pp%iBl{m8a}b&n$HtxiCe6dNku-|8G)P9$c)R<v@<o
zwoyAkXj9s_%<&}8av2aO*^yE0hXyc7fWCTH@IDyK=st3eWggkXw9PaH^JDPjx9cIg
zOnCheP#)7BPhkK)>{q25``)NwmicfsU@F3EfLB@WkK;F~OuNZaieU7}A>+UH8h8NC
zKSX<l@lL8|A6h;&OWCePOnyF4Qmt*nx#Ks}c%Gw5F;Dgy1bYq!QVMu1y9u*&+%LnV
zB2^EDU<0_7|Iz;d1Ij~t5`e(d_F-T<hXuTAUU|ZKH5_$3v4gll1vDUO&)$zL1KUPX
zc7Rym4kLh9_6iNqYx8FcOh!|~H6ULPWw%lHAek+)J^pIj`?B^ta(QwtpO2vcang6B
zcVkEMPr#K8d;;BW#|fz0P91qa3P^Dq>5Nki(ap=LmhbHBc^R2<<YoGk+r)K2@6wF}
zpAu|l?!){>0lE75t6>@ffcJPBids)q7k~xr)Mt+PQdk4RGX?$DX>!wt?0~8MvU#A-
z*||@SfJWC(P7aWiH-V=;1{jShODTZO+VY*fU^5jc76mwOuZG7V*g>x}(xU}XwE@Ps
z#vKhfQ?s`A8b4bj6%*YIxWgcjYsXN>iDagIB8E6S#WD|=)@!C^0nycokZbDv^(}w|
zyI~c#PMKYPRzwG`#v|v5i8<;sJ&Sy}FE-ia6A^v}zv&~eu(xO5wchCy_^Alf5NUt^
z7a;$yI@m)fWlKw?(i>|Al-q-`I^X^CjD3Pz6fboLu#4w4$y)_p)nmnkgeja3I~f8`
zkNS>TnuUDOjUj%LIi0WCH?2L_S|)4&-zV7dsrxNF?;2i7-v>!H@9@@{UiC=5h4S%K
zEH@#i-{{t&jed{uXie@$)0CphZ=LD5?q}~+1Y8n>&NJ#Eyy-}sOea~YU&gvy+0y17
zu0hgWI`D!;S$eFW*-Zlr-Pj=uvqAGd^X6FF!y2w`{v-D@Q<*qyYs#AO|546+Dx|Qw
zGN+pNFDLb!kGqjC7VO?kKOTUJbO#&}DQKd)-z|XX*)WO<9{>DI>h`}E^{sB4bEdee
z*DCa8Y{k2sM?2YAjn~`n$mUo5i3VHBHPXGc$zxG6T^~=JsZ`%A_)Hom==NV_O$$HC
z`bgku?)UF!GaUyX-N3!RL&vFD9FxgtcjvpU($EQC>!=^J?HQ+@?T<mgjf99&ced+e
z^e`Mhic0H3Ixgq-j^eS_t!8Hf(k-5k-EA*1y`2!7Ivebis*o%h*Y)zqg8%<9vi~PP
z9>%wWECSnC{q&<Xb^X)6wmr@e%#kOV`Fm^Bmd@QLEGz=tNs3>vm_xE$`WRMorRpqY
z(bA;G7_LGg1x1nbyE<A<&e|QAC%z+}1}4h=;=*l-0n}N%wuvPlGnio$l>a8hix@fW
zWrfp%)nFs@Fhw;dJpX3V5}UWX`nt!s^`eupy;Glph~B4&4VCOGLZ{BTMO91$kjx5H
zou`T=Mbd5beCKprCRaSY25XXI<K3GBg?@;t5{v=}HUmmyID2z&dZ<gGHf_zFT^Z_G
z>iObJU9ry=@>P^<i<+;r;LujKuGBJVHq+=d;?kf|=z%WpJ+aQU_p6UzWgRTOHpo+=
zzw~=Zr^J_Y*$x-FjIicQk{z5jky0T>)hPt4T=QFQ$J|3o=d==qs{{BEgI<#a|IkO4
z+MVYq0~U)|GYjSB8cR>A5eL6H?TV0N+Ir-6vF(*>xEECcnhw5~u`x?QuDMMB^_bEX
z^k{6ZaIPR1JGd9mzqmZ?;~|?XJ+Q~$tX#>ZVJ~X-*-0y*hz~&*GWC=A*p>V_C+~yq
zkXFLTdxfn0HuekB()e70J1pqEM~J$B&CJ~eY;;m{4PA<yMXArv6d$+HoJ+4I)rMHe
zM%(p#2^hQ8(z=*i6JTK*-LFd-!O6=Hz##s3Xr*qljhSX{s>jtgwqx*ghwl%)4oiN{
zeE0rxyL9HVZ%zEgpKqR*btrM!o7ze^>*)4as&u}VZ5?~#@P>*p{_BUgr48(3c9xee
z@?H~7SJP~g9s(wMd)TELf!*Z7mc`x5-@4fiP2`m<h8D9^-@uzTsV(&dD5MhI#VWR*
zq?!^lQ+?%?;mX5@N|m>zORU9q$+@<+ciy^F#41MK>!W!2@5ytQ=g24|-Z+(%BXNhW
zLuz<~om$-S-!yv_?=V|)mK>2a@aBu(K(8V`v}V&nBft=#P&{mlt@yH5#y2i_78_M6
zze>c#`>GdWH<!$`#<9YY(`s>-QOiv}?1xLQy9~K<?ef(o)GhQ$It_+iGSns5CAWTF
znf9N)Ft_vmglvJ{(oTWx2PfNC7JH!pbX5m;Zj<2krfh7J%U08m;O;stMarw)iX?Rm
z=XU?)J2Cv}W=51G&LW)>E_Au@@?OywdriJB8y`Mc(3Z=4u8;nQRbJ!?e{%N^Wo<WW
zP3G9NEN)+rpDkXGfN?bu>-66LekE2<QCL|}_@w$~QCNDN?;}f<qSl(Qer^9|A0C-5
zd6+%FqWl24DJ**QXvZ6eSXPD0(MhDyay2e(TP~FqGxLGo0nlUX!!dy8!b^X8I=wQb
zicpKxR4zL1vzwY6XY_#8GfgsP<l?wzd4c;4?rHs`Y}r0Ve%YL4<3NO+t&-8189VH<
zi6Xa=yd^>`pP~r*uC{_gon`K-Bg8~4n%loD6Wl!_(j5wNq2(MJ0EbO&HURZv>&d0;
zJE}H|0nXJk>7=sL-Y&XMF22j+PX6r5qR$5mY7%_0=^q2-?wz%eTJOj8F6B1*m<U=~
z#Vol#cm})To;|hQw2y=spO{NaB`<B)dENU~{Yj!v%F5uh{hgat(%G^F?F!-dJEmb3
zF^}(Sl{?sJ-aD0h=Eu2|y=YF0OTOmQ11D0tB-~Etu(`!e*lU1V6qq<y3)CFJUyqlo
zzhF^Sm$h`KX8vs2JhyQvN!8PcO)aL+y}8%aN>lhMl0Ux?Yb9wS5T<KyqHd!%DA$-&
z*}qVDCg@IX!j5cI@FGUeTM${gY-zsVCBGWMRP{WY;A)pK|Kl}b{-yHG!*|#Z$L<>B
zD^8a&LQF<<&3I^X=dH{m<80obGLotdV0WgJY<e=UD{oO|1yMt5OvERrX8t^k8WR#;
zEJj4LXLPuS?x4xlx<vt}U3DHRZ%He~O=H`QK86nI!M?>c7rLA=UVWdk)H!HmW1&iM
zOXLhS@EMRS%$1A_iNruv$=iFMb2A;l;Tq=4r4j5E;Xj+b<@5cR%Zc76e9}^#vu^N-
zzVCA*=p|G~YTEnOYr(tFDYmsc{AZBvkjZh$|2l$M(oxF#o5k?KnAq+Q;Z>_TzL95#
z$bxZ=KCh>ZZ)8|m%a*Tvj_);i*!KHwHnc#fu>NWo*hIvmM{_4Dm-i*T6typm>BKs#
zL*^_WKFho}A2O#<_<ZA|q!>xfNX+`^{Oq98heVpIV^fG{HL>{1rJ?ip)RR(kQ50Pw
zBj>-MFV!UIj!mV%Ma2Z2oM<+xSNhH-a^lKnpXfcy?=S_orq;~<1ysAfn*SiP#oWu!
zL^?t5y01S0(#n7)M*lik-YczpewL^R4%jxOOlBANYu!(YFRqkS4qN|GROE73r>xAK
z7g8w#mF<4SN{bhr6-jp%H{m^9Mk}yT295*HUm*kLS~NV@?Ai{a!FOuHxXFiqR58wJ
z6DM7)JZN(Pl#(tUb?5F=;av6qV(*^1<Y4eo2i>;qU9v71QH~qP@0li=0#v;;21L{c
zY(<N&?;APjv{%JDl~WveQ|LeYrzjOsJ7)7mZ>5$bON0zw`ki-{&dt5xr02P^pr4R6
zS6=iXo7=}QEE|QXoZ1#n&#P2~Y9LY5@o(jORW>SbES5pUY&}y#24W^0n^R|AdXJ8@
z*9Q)oM_&*YKa*!ZR%B^_zUyIj>Rl2RqsV90qPRjH*<?V18Qe`UUA~~{u<&K^FP`sh
z9!(ni*>El?<#o+G97VZgDn2l<mMzh^Oq-vV<glYDis<=_oW>cF7DJ@;PqSA%g(>l7
z49P1J8vw!4KQ*;pJvmkr9X_FtCeOC6jYSEY&HEM_XdtF9%!{8-wDmkaY&vRnMn_-R
zZ8ne>Q5DhA<N%7^lPvg}zW+&weV=3N4e$^i@EF@gLlb7r7j4c;pLerZm){;#=$tjk
zm`PJOAL_egB5+BKx56|z1I?*roW>(+;lrP+MxK(D1W6XGsOP8sgj?WY{GF9PemT5s
z+=WWjE<QOaoW&Qr%JR^z>&q6Be2Y7;yQNb&J3c?W>DIql;3H|m)4wh(jEy8^2BYM>
zx8Zx?RV!hT$)62@HI<Yu*V#GPLo6drr@Eimul<+y?Q{!^ofe1Arux13V%^;WO<u{U
zTe9C5{A)Ej2ed&$Xl8gKQ*$!tGoT#_>ueFUmiD^pIqFvJsq~p}+AxB8BVc2=XY#8U
z!6$m6Z1j!8U^$r+bPn7rSCH&XO%r@08y|k7_sl@k);cEB%vNIyrMlNz|6@B~b?MfF
zP}W}0IeF_m`N0aesjk?I86Q6>Ux<9bA^PQhiji23sL}O!0d;kKm2}<y_1V3&E!a3^
zuN&g5St|HchyMP(Rs6fVGR0>L-Iw>AI^}%D*Co>UnvGQ+1ay|Pdt9u97^2Td^_CBE
zURX?DDAqU7Be(htw3r+E%nI1Ehb6>;4h3%{R-Ad>zF<sAvDVx2Q$XWG_3pQ9wH;1+
zZxbJwW)%mMO2z7Da0OV;57?^~B@e6e2-ZTvA8#o0-caG{6|vwlGGW;H(8KDXiktK-
z2e&lZmfr-VrRrOSRL!YZCVv`TAFEiUl(=_kcFNaId^*zTfU~!7c^G5USQ~Ba>$%0}
z7Lf4aoBP#0Cl`JLDZh7aCiGm12+U<rx4gK$=VaK~lCc%ZME+J3C$4Bk?K;{vU#=a}
z|7IIYxS|sNl>cUTVS9*?=Unemx?%P2dNS+cck<3A-W1cjVeU4+j^`I-)7)Q0RfL6O
z&yfo?nOpnN<}ZGino1&N>%U+1y9kS6O7sfo$)w$$RjLsA5paAs7xSRGY#M&-1uvVp
z(pDr2Nn`e)uD(#IbIZ+|<GE@&)(|EG#m}e)_ODx|Hovfdzsi9B{Fqf>;*xcuo}s+{
zdh8xW*VJVEH?Wxmb+4gv(N(XuXl}2{kA`<di~XEJot4P_=(eB#`A5O@2<hHjctJ?*
zi7h|Bd!GIR&ZqoKmD#mW0v=B5M&wD~>jf}9pO*e)A)&&M^zaEd9aXf06prhKv_0}#
zQPIvPR@V~F+q}B#E+6~HcTsa#+gh;l9m$#w&+N1-9?k``GJp0_Z!BHvy&`nA*fJwh
z;m#`o|D-6_X|*|oBgO^uv#_4V!K~evX{_(Hm1|Rasq+lh>@nKc;*&zXT(ul{`(=@b
z<Cp%%<P$oHDj&1d_0q>ChLROOFruw89uM*R;Tw1M%+|IT!BghF3=M$!yv2kByV3_M
z5<iBWY1w>SR4!I?*S}ud`Fiq$QAbXIz6@g*ecnDZsQWRsz}F!W{l#kK<lQgRIRz@3
z9r7<<FVuT=*}%BCG<E_r#p96E40D2pA3rlUm^4lW4J)ZL+s{@|Y3L8^mbGHxUy9*&
zT+qE_VrwVrbCL_O756%{0X4-WEI|ZO>qz_FW8>@t$j!s<fJO{rzh!MYhY*?xc-VnK
zV}z*NO&bT4gsqzx_LdzHfjI8K=`u;_)AtED^V`Z&JfSMqd0PcmkKCfnu;pVb1%o|H
zXF1Lg6dme~a()OH7>m?@!1W52TCQ#b8nLI&>R{T7S&h_=f$-E|dTJQup6mgiriPVg
z`rgiaQ7RLKgt`o_U!c74TXcf-KUDc<==__T!tIWS5?&WP1K<s%nhcnr<uqwkV4E?8
zt1%`XoAURN=uj+1j2VtQSGP+{nd{`~8N27AaH3Zw+8*fy%v?7_@AkQQ={)@&dqT)G
zrdq_@;tZ7-hv3>|GBhXVEp-(&=YmN@Jbr4`*>hmtaIbyd*e_cdcj}at*8_pN6glas
z2K6(d*GeNDoSAo5|D+WJ$Qxyes6P|F2WVY=)m^8>Yr%02w1Fx14=r68R3{keeN)%T
zZ9o&Tby3dn^VLG0v9I?X`p!jO;rUqQ@Poh?x3r=`axIH8yObDLjs2@*;c{of&4kdf
zt-LfkKu{5K0dU@Nj+F}NzrN3D3y`V;mt>1qLXGD(q92!~Uq=LQDxdZ?e(L$6_6RoD
zR%u9IUnI%miYEGw?U6Wl17|(OIhz(4MSgiM`AwL>7Pn2a8Etv0`ohD$Sl7N&=ZE-*
z_+4du+FhJVDDNH~wF!7(e`_<`$yq}0Mc}9Ww6l-uOUqC-YQDmHQIe$0v5+v$#r(9?
zOt5!~Cy#y&YaBj$c240`n8n(uKI7L*6BkX)Ox{>lOzD<6H{=#>Sn+9pk-@!g(`gD>
zR`*yjTd^eiLiLVYZ~<^F1RU(8v>BOGHCoV(BhrUgTB{<+RO=tgurSYo2X7C*atKa3
zE;M2F?K-DXH|1qlSXZY{18d*zAMT2x9|)bvmS<Y3Aq_ReZN^?<NE)RR-&k8QX;bk+
z*rFn0Zs4+C$;%~+y8-4lB&mqK#Dv7iNei1A*={Av(U1j|6IdUtTX5^!5J`I@kt~Sn
zwB)uH<0_N6hIOP1XO1DCBcZ7-n!oiesEcQ-dwd|sWKHT5nEH9V@jfSB2{nBQ#69&i
z&29;Av3;1wUVho9xd6Hoqy4M#z(SYqNZDH-Os_-RT6$M3g}8SY@a2>@@kCcX%Pn#6
zN)HJh_dFgT+c}e=tYm!M7j`GL&GOFlZE&!9d2vo%l}^Qv^n)fzD>x{A_7o^}V-x>$
z{vp~;#~mgv*eRJ=oG2-$l`{l!HojRB;UNt3zw!Fz#P!QK=}04tLrk+0Sl~<)=FdSO
zO=i=4nn9#&|3X6exe3jO&gg;#yC7xtoB`vOX}u5w(@V+{xI)jy8~EI@d#9RjB-nNn
zY&!AiD3=CrEEL5i>wMKuer@%HLC;&IhQ8C1O&ePh-H!sqh2DYMRYhwH2(B{mqo>$t
z{cg8GpMZg3p||f3Pj9KBm_{SCwuA_p+kP~YxwV5TZR_|)B^^(>Sv1_@yKVGLIJ2&|
z$1r`ZqSG!;xEUqHM5m>R40N`l$|CljaubWPW~;Z!S=H{ONrE{{(3|fv&Tz<UR-SCZ
zZN{~dUXc3o#T4Vp;5hV!P|c|^erM=mr>DtcO6TwKTlbydzo(AHkuCZbcX%8k>f+!r
zP8s7<?-rE_L07M5yh$?a)ecgSm`jK6I>ue~8;|*1fvCytwfM;YO-HkOaS0nGPmU@i
z#Ub6@!4$};35Mt$DOK<IP@!Iz%jLgNvz>g0i${M6ZRV&ZrtmiZEvMs(#LKr2oqglf
z+xSv!kE-gx`)uNRdji7Lbw252t4QM9(|skPl?KjOaD+xJ<f<3WK^;@+K@r>ZTID?<
z`j_mXJRNNntmPez0YtSVHto$739PeQu0zPDyvfn<l8Q~sYh-sXl;fxgk0>@j`R2Sw
zo}cUmd}BiJrpIf(#cUOtmw2^#{g=`!?C+i`SSmss7kkSYR?fJ@!XcC`F^D87G@wi{
zGT#%TToH0Eh!f8hj-*UHT=%d>tmM6$c2k-$aj!*Y(LxI)3)42w7G?$KiixS`0y;!j
z3SL8_&Z}pnx+T)#?njNNqg+*31B=&JRs40G%H`ABiLA5DZWWoHMpf&1qW#7Crlu+V
zI*E$Hl7ag*-nw1g?Kp>;jadh~mfR?*+_YRTyK-O2v_kPrp~yJl$TUo`-Bb_OL-m|$
znUjgbN1GZoZcAG(8_PwCZKaZywh4#g6*V%(rN6(pnRJhvPWCSnY+C<~q{f$i_QAl6
z*l0@n&MEo(<+tW}+!HFS0<b8^cmGa?+Ip=@x3Tiu;=&nc$35}$3Le-MsJuW)dQm06
z2429-95!4Uq@!X%8Tde&nhEET^l{DV*3#-Uu~iyND~eOgbjx)_xj+UD!d$9jGwuy(
z7Vl09OSLT78!Y5t<0@>zZ<(q~K%B?40`hgQdv7UBdP+iy;r;c<-@GVN#^r%ZM&~C)
z8A)A`>gp^9NPe70rg#I)l1nphYS<V*5Fk~NXEV6se7Dlx{BETdVW6ocW50`|(~ry2
zaDb1VIu|nTUTEchK2>#3^O0+k^6Oq}wqltNB(cMPcow|6U~=BZpw+Z(KVHzvOS)gT
zPyYU7M?+$woh>oh0#QA;JeIbHtq+OHZg7jljF37awtQPv<fR>_dn4V)mF{MZS?7-F
zE;sY}G&Swe1mIgmX=yaKrtxg&RA)WU*Pct}$<GIE^T*R-tdwR_cms^j@8;TA-UUdu
z=_p)3?YL-}RK9puSv>!y4uw!IZLP|9W*9@PfI3fT<3MhR6LY?5!EGTBu584m2}<ej
zT2T0PjoRyw-tGbSM;j(;J>r!KTo4-zjV6L2jnEkYz>6&ff}GKwfjUxew)zJi^@hgg
zbM$+esSU^!F6NoHr%8wr3+?YRom8i^^f?Lg1HB;?ZGECB)SdA=D?g2sKRh7h{T$%G
z;uaZ&wT%l&cXo@IPS1<PIy)EE5EKb=mA6KNq!*#ZoACB>=|z_}0Y-{vDR?fvW<|Ur
zOtF{HWTH60F)$t<K)U7Z;r!b^*Rgd0=q2KL#$gOZu_RUGTku;m9fEOLed3Ev2@|OV
zSXnDW7gn(u9V(X1TPZz&NG<oB)aE8=(QJJ17AEq<@fKN<pX=sQ`lBi9I%7?<&~f}d
zTP|A-6AdcC5UgUzn63mLcV&|5xPZn?%AA^kC2P5F`t|n(tdabuz+@@g<h_)Ve?p>Z
zQUXNBWs3if=XQHXTPyILH$zxkfcKN`P@NkL`U?OkJ;HdBwP|{a<T#*d=FnARW)8U2
zl+4c1z}}Nzi$*ve4zcY2rcQllzDvD%@90thDNRgMW}%9*cwx!<p{Z%#L+zANb7Itk
zX5AQuhi6lDM<++w!$G+mUrJrG-?fIoaC0X|C&C%~FBUcd{xpJxqP#q}mXV1@Q?q#|
zZT6&#FNJ8eubIDDib{;NuyCA=Sd0p-TFfnUaBxmbcgu8eD0I#o&%K)fs=3gwH$Cl)
zK6Oc*{g}wnVsYq|$wL>;3&pCb9pNH`g(c2j6?dB#=?Tu>7ors9Cv1l*%Fm3~#lCBL
zXq#1qx@Til6fk6DGH?bjui@9JWyS%H?;;}n@07_9Y(`2GKhdJy3)Q0iJ8f#TxePZf
z%(az+NO6Es%Uf|A*Ogu9d3qx@+=??UZQHU(Py7R-9p?vq<hYxcmImq|6**Ss?y;s$
zr#pmoYUwJHM^%vyoL`i0o@*`jRPFxE)ThhJB|cQmS%K;%7(G~!&;MAw^x=A}*G3@U
zLvmDIOS`GHXTOApn}|N3cl=d2ExC><MEcpOsFh9X#l`Crl4r_C>~8ub7`Lvj>!-3=
zYNvZEG`Y{aqAs;eid=h;XBg|1b|JxLp3_w8Ms;Ck7!19-FDa2LrUozu$Ze~Y*%g6X
zFWv`^UfWyd3!GW~!puI~COxJ0fHQ6-!K#P0qVwQU{j_sWcTK+Koy9W=pBF$PZS5b>
z*usnpLLXSjb99C`Kp~!Y`rkhF=>FhYb1Se->hay;JHk(YIGhdhydYR*`_l7pebac}
z3%<CXL;)FFhsfu;=pm(zeCptIXhoYSfWPeQ3<8IfyK`_xAcZ&uKAQ0}^N8fhZKR-_
z*xvn6@!WL358PH9H)Frrr4KoxrfCwQe-b(;v@+XAacN#UaWQQ}?CMNh0-vaix4N^A
zc577HCpr!#tIE)9K(1Jv38d3nM%L(ln`Li2tdDkA#D5IQHOQ-OfI)94swGS!4IV8%
z;`Td~<3Cmys+?x5stqY_;r?SLNKP$a;fBE6^a))HGcNu->b}6_)V9#UhTrxo6N>VV
z7Ftz&FwC}^W#V+ymyYpZJKy0$HM)*<?VtXsn>4ds&SO-OI}v|Yglv~+FN(HCAMzCm
zIH?e#cR3e31`05%cF2|e=?kl!VSt_ksL@<!pz33TarYtNfV+l<7AahIA<v}MYu^ta
z-NopUHC7FC)cVo7A0_k>TX#0+$OGUcf!u^-59-e={3PqLy|?e?NLcp5pZ>a5tFuDs
zOega+%QHiX!olNaLxbj<e=g09Rp$*Z%~3K^SBxG{1hg2z1dZG+l?BgGJHeq5eq_a$
zaKPp}<Fa!6m2-u=7g>5sE}Izjq3;5&-y+Gx^dlnRMU#rqbwi2fafuX@Gh;I!Vxn=o
zw_vi)?n1M+id_8Lv^WA}9)&5iu|T>b;G`Pwin{D{N_HyU=pJ@=IX1AWsxn$7ZbaF|
zlFS{(ai%kU|7lGBs(-7g6K4!Q>(=b3v$Ov|i)IxGz-TuhCFj#JLoh+qhngM)oI37<
zdRk7;GRe%KpHEP%N!Z%(L7TjXU_(K&RsV?msI^ypXF#jCja%g(<?qFk3dEcoBz}2K
zA>e!88}Ieb%D55~`95Ag!<&7g&cYbeW3TKZ%vRX!T+ZgmHY^m-q$O*Z#&+or{Oyf?
z4+Z_VrV-Jpvzc+UQN#k~)#y}Y>Uz@Gi5`<#^B~n9FiXx=zScsrsa?PEqqmU{EnIk8
z6SE*A+qY?K#Tu<`w@h+eP46V-S7m|C9KUqO)YPr_qsLFnw7aqyysj%X>I!8ejHUaM
zufCl9s3sGDmD^O>Ko97>lG*O`Kc_XL6RTqIsB+v(-R)6g*C<vTKaL+~@@bM%*Xah1
zf#lND-h^?B=wjm2v6Y)eI|1R!#WHqsym{}Z<L_2^<_aXtD62<|kMN(cG~Wm|k_d+w
zm?W6qh^lr_jZ-62P}7B(1%+ZEyW89PKjTZ5^0qqIgLB^Y<)*6$1b_dGx+)g3URk>o
zZ)F<QniIsU+=s*}@p+YIWugo)u6`W4USeT1E~sEo2E5&6uarm@u(w><VqOrgIxU@C
zyem;Ks(oxi=C`h!jVtoOLuT1Eviy#NKK*^7Z806O>I@mF_~MGO8%esYNjK7w<${oA
zi@#Z%5%|fISKQQ(JgAU4-stZ@5dPF5?ofy`8%lmWbM1v3-XpRiLd>BZag(xnIVB~T
z5MnF(uHPoX>AF21jh_*wIiKB*XqY<vXnc#^t>&-(S_Atsq<Lm<mrs*p(@V=dUh+>X
zWh-xfr&1eT-eiI9t89hE0|l_Av^&@tH2)9N+%ar=cAOZzeqvIy%FNsl20AH%8n?hk
zdrYk-?`Y=cefPLam=&&oh<55nAx8k^n>usll6I?|Dyr#5@mO5yquLhn>dg{{9NKF1
zOoD!5pK9S{uw9nA%|w)>rKQffRC-f)$y4JZepB2@yvd6ptqX5?=c8Wtt_LK%9!<Yu
z4Lu=Kk2VsYR*>!0hWKo@tor@}Fqli6Uy*QowPX~h(e><((ZG%Lt`Aj`_r8Z@d!o-T
z#HS_L>W}H`mFIV{B}NJ>i?oQs{Q<mE!`f`KE<KyF*Eb>GG{?f?X1rMei(D0F&BumP
z?>c8qmDs&~=ImLL2QucaxL8SZMG9f*2_yd|1;;-)qy0LwJeg*S8`7d`01Y5%K0~ky
z$Rr*Vn4Ntw!w0Q;wMf5u?}IuXH|E(T_;gtZoaI-}5^e8%v*PBjIcxpwG~;O5%brw9
zS9*n!oBo`b9*x}Aq>JcfG>CF3YWPi#C3Rl$8b+j2sC%_;qwUz3wU}pIC{MK#A3ztv
zZC=GyBBi9*vxJFPxYzOeWf!#Iz0AedghW0+c?%QDDhC~5k3!ZY2uHb8BE&H*&+jeZ
zPwp+aCZB#Om`%jzH2GL}xCN<f2B^-oQimK>q%U*2#OwQ7&x$_wc-LY@s{NKORH8R8
zoitsvm$s@ArrBWcje!Qa2r3Gg$<r&AT;W&c`AcT+j8?vJE-gpdtlmrM*eT97Ftl<S
z_A^Q>muU;tDmVVDyC^rTAL7fwH^7@&oLnpg)^%mbSLl%mu!>adq!DZJT+2xjY~qsq
zs)I9Sy@8EUQ>)!kmbkNpw}-!fG;3XHabJ*7fN8i?3Z_*((oR&zCD#nR377aS(Gm*!
z9`-N^QJ~I$pEHS{=Y4Y2JU!8YmL90I>}~W|pEpJ$^I7k#+4QEx`uPBXGwOuij2nc%
zS(4opsI2ZZ3>Fz$2+v+bjrGX<So>MgRgNSGR-C?rys?>J+P`umCC#tfpiQ>Is4ace
zVSUjyQ_n4s<DtAty+QGr7rKVeD@Xy&Ch|?xS<1BVl;EG!sp&428eyU?rJwhf7GESk
z@DdIw>5Z_&zknDk{?!zCk}z~=LxX5s-}08C-jw-D=ceh2xqf}}Kstd4c2STn4b+`&
z>Y7?1b^<Z!V60dt^KDbt*jp+4w!2CER0#k1%Q9Jd%i)hoUe?65D>7q!JUwQ0QI3Zl
zk%d3f8@(^?<i%3~jRZ6-A%9L4<&to*aNEJF<p!mAnsPG`J3Y)@{*<hI)4#DG{$yU?
z^?+rGf^t-SO6+N@p=+<eYoqtD^GQkRAvG40WlZLtu`G2J^{u8xmk6nukF5_&S?d1j
z;&Deiqw)QV^Q>{cl3ywo?90Pjgmhm62J=fh6Csx>$0FI}WjL%vHzFlOKwFEJ?#i|<
zKxU0=pPzFMNlPUWHS1j>8DVlplb&*aOdLaRHMADn{aBIwXciXXw5*WEnv1)s|3HUv
z%=OCq_6N@WNtayji9~u9Na>x65|7+vz@aT;>C?^tf(=imZI(IYIgDxDEx-~&YDQkM
z8+yc_ty59%5lzwA<a6*mZr9(Oqau@j&QrQ2<c`=ncQ5E;)D*G^SK6TpHa1Lcs_qm`
zLg6{|sA7MA|Nb~*L|1!g;bNuqhXVcN`_<bqP38-%rb#(a8N-4t^Bf5sQnaE8#&yb1
zH_311xrIjqn~_()ZU1aob}!|JiQ8%Vb!%;lLjtc{VAaL3mM;m@Dl%84bI#~mml!#I
z^1VFs;ZKGqZQ)%%*2-DkLCkS6Q*%YKp&SbNu}Nt#3<%OI+X?<r?!Mq*e7~sU!kO8H
zCjm`SIKBJt2e#C4X<I^X0&F2R1B2_b-_Kb5JL}8pgT)?Wa)Rv7iuMb=lri7cpDP8E
zc+$QgJM(A3<#%mGB3B?U=SMs})yr;Xw6hv`s*g^-TuvPqGH4ez(SQlG?L4iTgJAz=
zakLBM{=EI>a*Lme^m;m+tkm_<LDoy#<j$UxYgs_6rB8oZf8|UAQFJJ$KHU4dfoZ6O
zEJ3{}vzLRiN)Lk|STD!9V4M(tv&=HKz1gfI9Ya>+!(Ys9KC*S>)ifWqu5ObMRhF#}
zV|}pUs{ncSJlTKraeYCh=_^jB4;)cK!Kr9blTO1+`)^*(E%_gtUO(F#WH_OiTS>D^
zosdax@-^FJjo;K%Sg4e}P~uqGtMBx~SvVn-UqXcR(u1#|)UpLt+&H00-B$7TysA8?
zDc9dpUSzi=^Lk0_aka!YQ~lT*gsk{s(WhymZCc_2ms>O2c|`)gJOHUk$16QnV%Er3
z9>%ds0Z{9c%@!J6CYpP-V463wTUv7RM`h;<6$Z&xx9C+f&A3>YyR%_jqi?|<l<6Wp
zwds_7W%uZr<lfy(cx#g28nkheV{MzU4?seG_17HY^U|(E-~J@gtN?#dAkzrW0<rnl
zDT`-F@|ND#Vs)H0o?^)|72Fr!T5*0z2OYHvigRYLjBp=DuVvDn86!RIi_r(0M?-qv
zN<Wqq3%VcLxCEvp29|BEFFCZtJ}|mz9>MDHC{ngy&eCuxm#TV8Wzth5K}jlOs70fR
zMx(8o=_c(VzhB6gxYJ`a{pf<gICs5#LGVitMZJKb6Iy}t%$39Zs_Q=+6K3*#iqj)z
z+@u0e;x!)Ulu$KB`K(Wh^D12THIr{K++PLAWO$~HMH8(P4ryC=pPfA2HM;bs|L)+z
zmj6AS<i2|?Xhpu3BD6pN%_QvhVwX&FglWH1(xo;^*X4|Ura#xr3osAoXXtK&D#Y|m
z8fay*mZ+CL!bR{BRL{v%RW|dliWAOzJn!m}CZw#$#y9VT7P!Z0tkni9ZcDT`;hgCR
z-$HMCodZoLRjIq#i=dK~7vSFM3XEV{h+94_PX3<OE;VDIZEb;Bdpy{U&?VKL&u_!0
z$-Ulf#K&Xzx+9)LK|)cf8lK`*RVrK{HdP=gEV7Pu$PCF;+h%T=bW&S1>)LLQ%>CFP
zxcFPLeke(M;|{rdOQYtt%&5MYf%iuVRxH*dy<3S}U-?#uyVXpGe_I<HFZz;^-!cP$
zMh|17!b|B0Bzjwuz8dJ4cj3p^2M)OBXzEze`DCGvqum~(o__N#`fjH?UrC&t9Z#wd
z>0m{l8oLpHDPn!tO2B2Y-(ne4x-OMR6!a$3cFJDnIgRN{6$<~9UUkQw<vNsP*;J|R
zx4vpz@cZTYIA4*H*}u|sVg1$1TOx@s^;hED<|NbIXsO!_L_;}DIm8cw^oBMp-wGU6
zI=zQ;i#{KA<#q`_tSoSP=u%}ziHmLM56_&Ah(I$vpRJkt{Erf4xlpV1vBl16-A-`M
zYUwRIMs1m*&0*sl#48c-yvmYuH{HTytBGPc9xLxgUHox2;T|0y*Il`j1ALbz-#h=<
z@1w{&*NGX-r!096UwTp1XMMeDbTCO&QY6^|R4;Jh$G4e`{$4WFlKOIo?TXa*lHyM|
zEC%Muiz~m=lXGX}W`Ia;&QSM*)>22+#AP$@mVh5Ay2Ebw;v*xc9l~2v=fpJUUGk^i
z;<^sCm5hWz$4@=Fd9>4$`++cL+kJIf&#qHh8EtXC{ZpdKNj4#2v()Ib=DEVUD=qvg
zJxRX%UG@SHc*7O;;5m)DgxCB<^Ug)CadHVQgSO6?VzmnU(D7wP!R7E&<=@7Xv?QzZ
zyQ@CPUswzs;%qcjsx<63;*oE$<j+#wrcP4nOeSBF-y}GKfiuIGxDXev*G)L5POFEb
z$Y-|Cj9K5l?;=@IU5s{Zdik!$Nh#-Oio>lqlJ6M`$&_<F<%1|kO5%-JtuuyNTNL8<
zH~XY;<o7$>BY|J4Dd)N`jxGLJwi@Vy6#sa5TkGq{d*8Z>9RK^u7HXV)rb)N3D2_SK
zkpg{%;Fx}ROtXbQm57>zx>WA?=BA}(s-*#2+^M+YL-ZvS_qn~ONO9M2(LXh0^*N{Z
zWm3pOti85D<SS1*QltrIAKKQzK(~$4p>~NaTpyPV5Ki4PC!x@97XGve4bm2zxg+d-
zZWzsRTB*=D)2)K$?%-A{taACo@IZ~&U@lHl$X!?>sk8FU4XfQmH_qym9e-XiLhCCz
zLl{C0fD&#Z1dJj79AKQD9}OH<kg4dIZBF+PFev85BHe8L5+)~$NwKUJ{vn1}IFoVg
zt>LM%d_x(FDP=kJ(Vi90MSr^)SAW%qMrYLZE~8|})Gm`7G`0XO&@$l!TyyJ1?ud?V
zP{kgpv!JEKphM%YPF<-8W`3{8C@lW|jLsj_;TtH9@-~%(R3Fm$Hce%k34Ms6$iKbH
zv9`_t-9B5YjFx)HHlm+TCE$~}GVE0U-n<ug-R&27G~s)sRL$SGn?D&=h{(&~Qr6pY
zd7m%mcdAJk{!oMEXqsk<#q@c(<wE6<=rv3bGIF9|t22ixp?@pkdBPt_38qQ$oTdq~
zdN*$2e1&ASw6t3jY+nq6A~aB}`k#F{0$MAg9Mv45N}A7OoWT!qLA|{j$)$5edljDw
zp#f4qIsYxdEIRIQmh|5T>v_2F`br*7>G4768F_0L{;|QT+xU;8i{y^yzJv?EB}NX{
zT`@0_91OJ!bFAQ%&1lOGm${WFFnvB2eRFQAcI8r7`}B#oTVKa3>YPW*A<_k$^2=V|
zJp$B2UI@YTq+*}B4R9$}*<F6|ryEMlNLxBgN=fOJm-yZGh1R=nPfn;3#d+<-!c_Xy
z!+$5LEGzPh_t_xY>C>2#xy>7ceZO6hPHAIAw;6S|Jua$wi2dGfu%2Ns$fhznru6yc
z8<BU<?n5d3)Gzz1Z*`c*dy)HRp9s?9wuiyuv^yJdWWM2VmXp`m8M^RC$QrUq!=Vj(
zPz?I+_+ECs3Ur*7Klx-=4s<Q`9NF)kydb!;)9?K?oab+rDc3`Pv$(FtqPF0RSAIeO
z`te#B%ik<zK|fj;^#K>w1hoMcm=z0h^eV}5Z&E#Wn;q1rF=bf(FNtz!x{Id6=ONx4
zdV#?lI5rx7>;?9$#E0aPghXnodJRmdE3QE}enx33IO&@38=tG~<8do8r16y?pV3|{
z?S8{J_oGEbmebBdqYJrxsfsPGp3?T&Eq~-@1W39^RoCTQz!%?Au3t~#%J~K;((o%w
z5ymD_?uaR$k9ZGj^?Oa#pS#x!pZB@HEF`4&tV=5Su{PXawsu&Ne@9w$HZz>yX8R4J
znSZZaRLA1vw@o=B{VzAM!3{^_3dVcYhjy3+m1M`;Uu@*K3vTR2DEF#+HAp(;CCzkr
z)I_ct;99*uW+r+f!yY&Wc*@!)EtuM#YWla;{y+TYFrgi4Z?><BjD1H+Et)dKkL4mO
zz4y%#+rf_H@N55)oRtQ#0ic265l8>|i)9-6NoV~q>wkU!NpT%heg87c;aYq<V&TcV
z|Nb|q1S9wG6YyE8lbN@Y`p-Y?C*)kxGleB2I5~yCeQEn4ybi-NdT4fsq_{-hI=z3t
z2>S^+C76);iwXo3(n0Exk0<Aj+u8x6hYDw5f=JLe%~t)>{pA=veMkc8GY*4}=iq_5
zJ~!;dD-;UH_3VeF1)u7mM4b6Yl0$aR2rfkikdWmc8R!2cIZI>O*;l4Bqd8HFGZ|5j
z3=FlC<<G}f#g9cR#TFQDWWFv>q+oQcPzirz?Z(+b9s*|7A$k8tT|&~<g)zAF!aC&v
z?%po*PAwC`Qq6ffLwvZ>cjH2c)u6x^3r5mdV_xnXrC!5c`{e<?E;D)gJ=}WQI%%}i
zVbbQ>KccbH{;i^4fB4_^#PUf|?oe9z=U=1v*BCS|Tt4n$tNu9`Y7@vI(kMYiB}M3p
zCYO#I@=9oAI_3?kIDNqAiH_cI>{i?Q{_AAHckKHH<2+%Q^O6Z)TB&#yy4dUEomVd1
zKG5Rp62E(%*43}p7kiB<G_5w48Hm4JSeTz05lJStN4ceDx{S=8{Cax^$Yz<g0~`PI
z%pp_qG@0IzdX&Ze&O#tAX+<+bnR(UTcl>DW_YYPL0z~dt>eeYak2;-W3m`!c&z0qd
zrJ0ykq(pl204{+-S$MO*&t#@%CLA8VC`GR1W599SVd-tTwNjU*lqI7ruJ8b?-C3+p
zh)G_rsZ-CaR-ySkOw3RYDl+CY;`z4enCN?jsB#?l>ksN${d2iZW?$^F?zx(izoW(h
zNIZZzhfHD-VL945?uz&5fJzbH?hMD1*5Y^+o=A1#zdIuz9R)-Am-+s(++PNIV4@vg
zo$bNb|GX^IFwZ;phff(`L!p+RLe0|LM7zd1eJhnds^cvGbX1hJIkr28S-OY1<XN`3
zQ>Hgx>^h@Q#;2Xg>xNYAEqZ|Bo6_oZp6r42<2kB`1T#Cs)}=q&`--Fe)D{1jHP*Ev
zjeGQF%ltTRshl1vli#PM-SIm8V2%Afb^21u3&|(klv&a&mE|1}hn1$V#)~E%x(42J
z&mqjVuH8QRkBR@cvw^SvLofD~=M8}D;cpgt5%O;q{Lb-S7{eASfMhLXPOak^o=ERC
zyLbO?pE~5cc<k-M{>Q1I-Jrs|I>H6#*es5=cfL8%O>K{<<Z6|At8nGwxZ_X{xqqbD
zG$Hu{mypJYjKn>Wq@=Xt0=2i8EndR`EqKbUsm+W~i*iXzfBfq8w5tJV&z1eKw^zc_
za{+$WwLa`(!s2+y;wCw@U+i{sfEc?Gx2D69XSUSGofE5U@RpvG<qaNLLY1Gm=5gMB
z)n#FkKKQs%Vw}pyo2IVDfl?k4!?8(ZSDVj~NdrFuK0z-RemI|cD-bys8qD>$YSR9c
zrqr0-Je{a$8t_aMU;Z><Ny^14V3pPDp|eCj&l!LD_GgFtmmM#}JuVXZa&98}cE`HW
zv5iXmaD#yBR|M(zRL}j*;uDJk;QiPo<aQ7Xa>VJU<K7PyDATv57$oa4kgOjj{)O5)
z2hu*ZeFea!y?T6nFU9y6bj%UViCUzpC-?ck{!hKq5qj*LRn@tc@Vq@y-<bddgXf)Z
z+q`f;_>?dT2E{2=!XI@ZI=TJfN8>rq>R&GwyFjR9`@=?_p8WWS!8>cQo=1w(FG<Ej
z&M9{j{FbNR51xu;Gj*;;iOaSZoORbXUo7;Ty0CQP?B6VRKV0nNy!Icm{QsXXdw8HN
zI=v9pCcO{&D9%tuwqZ&;n2-U*y&B|t_EzKCJo<+Y0x!IcWAMG|-hHMgNSKxrADub;
zmXl#ympO?79mE%pUx&i-2<XaAeooNuY7eW#H{wAHbFVC@{1TV*Lyztwde|aEPPR-#
zJ2uwP7ug~a(783*8qWh3e*+9u(7R=dMJawe49NqP;E6brW$#g18V}&Y*k5fk-|=%&
z=<vFFS=)@S>Y_8d!BC!-PG-hmy>jBzeFzfvJFEZ61*Xjo6TbJ3dp-{Xu}@ssSZDD!
zo7oLR2*w-%t(?#~dTsq}a8sBdi5><=Or^2-->R)=vs;bX2}Vr1d}%%ME)xOuz_`qU
zep1mH$~xn1{SBzc`jdmJ1R_9*Fpy%Mhk%}8036=?O7RT)jV%p698YxXp&r>ucJEh}
z3nU7CFYfD+>w;e>fh+~_f{A0G#k0vKx^4Q<O}*ggc(9_FrmTg3VmW@TVPP3f-1p`Y
zCU;7K-X9GOfDXXI>Tl3$FZc`tm3uX^FZ-1(GLRmO6s*Bhc;whoO>B|DOz|nGT;&d|
z@*A;}rB8P`81iZsLZC3iYrpl%MWf|-2-^&}=7G9qY;$dD?}KMSXWB={BcY1aC;kQ?
z6M={HB<8PPSR(j*a)HWd;E4up;$|~)?9jX~%fT#WS@tdX%Ek?-oB}nF_9t5u920XS
z%Vvk60TjKlz4J>Qeq`pv(EURq^%)(_9pC7k48)(@^JrOZ>JGdf-&4!N7u~i&Vdet+
z0GubXb}&H&IS!WdTsv^c6m4J5|45b{`!roG7+%lr_w`V(5S0fqi5gxHW07jD#Uzfc
z@r+(PGtW~xvO(Dm-oswEi-ZFkdNVpZ(($eJs5-!Pw6C0*o!!64_E)$@CXBqsxX8~r
z+fg}hQ$K_zvdrhTZS00$J^5N)bU7?jFa{SK77`5Se|`uInzW0scfq(GzYf7a15{^7
zp3jGR<??3NnD>t*04(CY@IL1K-xB0@eVKWb!0iyhS7ir=*f!0ugO#@)$-;;e7B-e;
zubb(G+0<q<@B$0NV}$nQ{Jw&rGVAe`%n3okKfz;TK#lPgEN*(+xG-BxC%8v6n#Mr1
z9>0d$SD*^^?1b@WvK#?!1R!DUAMHHLvRj2boSs7{tkL)F9m}M8VGwvUa4Hto^S|$*
zT#sbgzJ0Mw4@N*+I)OZ}=wYbf^B4vk40{U#6^urgA=vP)Pt+x-JaQhTZ9BgM1smhV
zD>rt1Wv>BsFp($>4LrK@25nPov^jXkI2{T+%(fo?yw;BVY<%v(OLut6r<q8YcNy>&
z@9BML69TWuGLKHA26vz<e=-q2)kU}WC!y?$Kw12pbfoxa?Vqz#avsx%KA;vzRKe1Y
zrUMNx&}%fma?i&jhuS*Y{${B~@0~iTequcwg7I#_1I=sIX6$`u+SDHyjLVlD_R>+x
zEHI%bW?7kd$C0%_8bITX41v7GBzBN~y_<Yh1}LbZkHDB<<@nO(9mpgwNZSm?t@<It
zekT4;e}lG;#-H?%eS08)oXY|j;u@xTWOnW!+kKa5Qx83NTpJ=_4|8Sw8u!~wc8^LT
z`B#MKK+9E<w!$&HzMyA55FU;(294nGTF^5u7rnz!qs@}GuAM$X-k3qGdQbmjTft07
zE4Uu)$HDE!p~n+}Wl}&8qwK&DP(k84lINnoL3?;kN8<*4e@O25{a>MRlnxi~WLyKG
zMUIEKP6Lo1cpeuJ`));-f<p_Y0K2%ii$Udg`=-Nnoa^CG@h?B=EN1WQVIJr&6igBb
zNgKPtaLsxF0%gtjxK=d!%Q6q@D|{zR69eqS-q_9zDvqXd1hc41tRw#Aw%W66G^${8
zM?B++M30=yejNse`g}B0l+LWffM^j-0uNZ$rXB;En>n1D06oEycO3$2+1LV7j!aQ$
z%X2RH(uIvb*`C5Ny!*F7WNoz*K*Ah{*I4qv`14$5n26hKPZ1e=7<TXSp8#M0*dQ5n
zic!O!@E<0x(8xsmx*z~XkSnL|U4@Owfh3_>FBtS~gB}8+QRy&+3Sxjgh|ifexbOy$
zRelZG7>OR;X6Cd)v-JdD?blBd(t@Ad?$e#7`T_`wlV~CZTz(!+KtF5y3A*lE0mwx&
zX+3x^4&P+&U3L^GcpYwk?zmR{q_?-rm+9G#=BX_Wagdf~JI0`g(72BAwVhK>fIjIs
z!I~-ZFnEf*;hJ%i$06aF_CS|*LKInNJ3$`w!Yn}X6P==WOwUq*N+Uzm0Pe+PB_6;o
z9E2DiLfR9~TMcq3^s8|)X+03A{^-U&0<qu=v^u&TR=@hH1%Ch+fKIHOra;BLpN;Kb
z{G_e1vo1HirNMF_4{;V>iT?8p7)&LV27c)BWxHd)E{76adl5VsyKiWyAn=$5IRcu{
z{1dn`+S>&;vW;H{VXFsLy0WJM;6eV0-g$da!KANd?=W-w2xK#UWcFZKd4&CX2yj+v
zu!csh?oa01OyfzO`>Y!Vm@hNbz-@*GQjmDa_@4+5EJzT5imZhJ!JsF2MrG~MmEdux
zAMHS+I!NRlO%#d3A`<UQ-`U<DLNx;z43Iyj!QlZ%eAw|3d<#DHxE8?f1jz<n4wBzN
zBJ*xu@g>voZBqczfg`wXGB3JqSeFD6D%T7c12BMv<v<=B^&qF5{9$hlen^M41F;?B
z1H)?fKr+sAx#)iqEf*BOjf`M~!3BW*fbdcGiL?%2+JQ&R0E!pEO}TP9cu$<7&=DE?
zlm9YSw#e`Xo@SuojxjR+$3eu>#A^Qp6NF5Jz~uddP)<5z7a0PLuB7m|JZoO2v*c;v
zI6~ML@C(!f@!N>ARN6W^@xZZ3>pU$Z+qgiG({se9<(jvj95^OB;akTj{i`KNj&!CS
zdtNl=S?RZ7kO)BXK}<sIh7>?(q6E^zH8Swy^xr{Pg^|e{%j>iQ?**B<xq~!!;OjiN
z*?kE9(ShOaf-#rCzxBeNwRI5XyovON=+~!3;CPTGU>;y@6n5>j1O(pEIJFzDb}&on
zK)Y0u*8XPckH$dkMh=v2vxl|cIPl89f;Aw5-s7?$=z;W&<7pwzFkeh7(AB^oM50IN
zJANQ017r(>;VobWqjJZSSYF#rChjmB(8J4zj3W`oct^Aw6Cpc!4j2rTi4eda2chZA
z&dJgR;w;#*&=Xn<*APr3L9PJ<^OJMk0k_A{!r(P<FyEPxJMIle5BL<|I{*TtoO+O=
zcf&BzrBvV@0Oy7pX3rZE2HcCdymJ*Y<PtqROC*6jB<Hfj)UjLGKKU9r@N+x->pD!t
z6A+D-flh~yYioiC#_l4wyIsro<iu|_Ol0B$Y37<#uPoy579<l7ZVW^Z0XB17b^^}S
zU|y8!m225x*c}E?HwO`mu)Ym$DuVR*Hnq*Hhk2BOlp(y2lyd>_C`1x~4*@cd&e9v?
zqA`i3jiW%G@)?Fr^a6A2uO1Mw0d)(Za-4oJwE#&G{KGEdz$WWu?c{>KO&vsw<ur5*
z11adzd@#NG_H&&cqPDF@Qy_!64Djy=92|*CWlg`u6gjYyFfgs*c@T0E%z@;@38QP_
zfPod{Tpj=l#LIYh5FpS*>XW={wGGDj<22gZll^K@;s|*7npsdV<u#}rA%j_p8X5$-
z)?XHVFo6zF+(KY(EVE?GTAG2Li`#ZnpVj8~o8aPGv>oIt1j4<6=)Rl1Wm^0E!+)<L
z{=ck^k4cQ~_m{CdTiuz6pya`(+yaxd+Z8&$m;PQ<QeBdtTh+Dl!Ab4B+d#OZn@T;_
zs0n8}MJSanlDrUBB`Ngnql2`{5r_X4<$u&WtDQ@J7Wa_!5!#mZiA<qSv5JCl2@Hh@
zYGVZJ1ypw?Z~u=P&VNGK!)1}PJ{rs~@zgiRkg!ML92&ibVYt0Lj5K~TdH-*gqPR)O
z>;&tR|HQrjFJJR)xAyKXo|?_{{Egad1Dr1tQh!xA4&WY`!$^V<q`5&eU%0>yJ<M?m
zG-|J6`Rb7D-cwsabVdz4?5U=GOD$fYl>yke8wBHj^)DXDqmqdHI8ws^d2>g%b_q1&
zy+TadqgUtP$uodZ@|3P2gS5!NZ3ogj!>V95a76ed{#pDE`<T4u|Kjd_qnkL>d_fbK
zHT2wT$L-{hzQ>XXp2kUf4{l#~?_gW1A_=+f>7k?MCVkl1b#OY+Q8LMmVq`1}NmUT&
zD?4OWFr?Qe11bVNaT4a>B&J&i+bV)h3Szel#?2Tb?827iGFVcnRB`=rl`8v`NzP>U
zoc(z3hy4~4yXvj?eV*U*d;Y%Gsh0xUkVIVaN&axJl4nPj|Dtwh99?}ZayyU@G&#CH
zE~3tGT(rZkZTj8H)14ymSb9<Wy1+(OH2Ax+`o(!E-{6&1ail(@T_3;YFs4{W#|oN-
zv?$XZ9b_NF9sSn;`rwB4uxnBj1;y)ZzzJp~;Q!GToqRSx*<Ng>$8WW^@j}?uCIuW*
zO8cICL%^r`x8xfg(^vEe(&)ZY{sG3|gK>0wADiM!A6y_<J@$10o6YP-AJzVQ#Rm=;
zcKk|PE?vdgRjuFL8zwCU$C!(=m=nT}kCr$QewV6ksF9t_fTIb#+V|eti%@7{(cdQ~
zQ*E_h7c?80fPpl?jm#H+QEJ1je1cEyp%qbe_E2}a;R2f8Uti}xF6YEFd*EN|FJLx=
z8JZ28&!ut)x25K4B(EaaC3WW6UL9_i@)0O8b+S3Q4Q-^O2sRq*(MEgKrPIU1abD3M
zo(&E}iW&@dU+vCLo~cHgon<28N$2>o)7_lv(VwlKMc*9!08P--YNTgc`MThTtC~Hf
zL$bJ}wyAt}?ome*Gbj*DO8*snVV5iUsKe1GBw4<b&k}gYtLcgckC@dQhGE=XKJFji
zO>?#%aOarp3_?MrHJu?CYC5BQSW^EgpGwnZJ}7t{+3Mf4G<qQzdEqn^JMZl$u3*y#
zpG(gnKD}R&8LaMuUScC0egwg$%808FP47Qd8_={?F+qk}MIE7bo<+KQ$YF4pyeA_>
z;BpmD_DrV41blUEg5!9-MttBE1(bS_WpbH4wDmb12^iE{AN-dc-!CGN4HZn-;Ct@L
zZ|<SpP|!Zf-V=Yhfe^Z=z6qXZj_2ze@XAV>$`C*G*$(Fii%3M5+Xy;!tS!geTt~)j
zlp2Ik&ZnE_Mcz8kr)B4OMkL;@1qV^fH=^6^{)dwjK(_3W`DSl|beG1ND55ABl6h{N
zvtwu|n&8;z<<0ZyG{+^hSjJc1h=&-aG|f>R8yjtI0e`SrmBV8kRNm7lMU{NHH<5tG
z00@+BKcd!XFHna%!{kOih>yM$9J34z<g*Jzl9#mDRX#OmA2knl=uaT7f3UqS!>lZR
zXrXK7WV174U_5byF6F)-=~ayeow@Yu9q_w_=pT~O!DdNoC0uCn#xZ0xkiW9P+qRF>
z(cw6a?My3=ZFmKWdbM8iQt-IG`m6(9d0zJ!-^+n!yj6?AH~ah=OmVVa=S7fbk01{P
zm|MHiild3_n)g~IZLUrOG9B3B?>m7r9EtZ(Zy1+Xo}P@eBXpMdV%x)@@2mABEC+b2
zBgPcJ*er38I{MO1Ha^WQf<j|(p2HJq{h^{rYeLEA6W}|u2hO52kw6=9*Kd45gx$zi
z#EW3i0(KTD3pO?mEfDVnFkhao>?+Ut0@mWLM=j_hq8=Ods~(zWLI715hHkr}hcav4
zdubDIBXKu2m6_rLLe;pKS?YtAzOMBhe@9>Ck`8>Z-p^PDdV~msOUeR6pZ`<>$L~X#
zrpO3ao@=eqci6iAkX$#fNa;nH=wl|UZ&gsQI$VPs2^uMtnUWr=?`V~sSe-U7V6-|L
zym}oTnp~7~HTNaWRmCvwvp%?|-W6)_395!+Hy(IQp(lX9KJM7%NrL!z#g(VqI3_wB
z2VqabMLAcW323M1f-!!hD=UM5ASbB9gIx(ukXoetr<ue=>CFXVkkcJ8rN(xY9!{&P
zni){v5wAkyWKX}PDMJp5HMWg`KKrS?!J}17B9*_+l!ztcSu%omF#ib8Jg#EM#k<Oo
zUtK!o%0RW26R>#>;+pMO4Dz+y;bxjjrI~1a%5jvr<j5ZxX9g3TZTVJp{wJVy0{nMJ
zN3@Ng+MJC_07ZfZ^GWG-AtC57?5r@J<+8eg#YT4&yFOPAHv8m6*lv7yjGKFIp?q>t
zmqhJh0HMmu(?#T@m9Nnh!rj^~HC-LIRXI*GyyP44n}FsSz9T>NN_f4GPxJA_P(5k;
z`&z@s2x;X*ODP`?LS@aTj;74-d^VR0CO)3CAWb7J3FcfqiuO^*ElnaT3t8RYbbEbu
z_QCKmBE5BsKTi&)^S4K86A4kriyZwDY!&Xb%~m<kgWjB&W?oH9*j2`AT-!q@7VB;A
zFq5{I_Fs2sV{wLJsEHI{&;`8sPo{9U7*7`wM3Q;6O%a*6kORPC;)`2?B8bP;uaYoK
zGUTjwB#00dvG=LYWSlA)kH;B6BRJ2K%>iFJB}@8@-`klx&^?{jk*^DWTx$pAsXU7h
zM)9{^qUZ@u4q*JQChmzGT5*WHl?yrhC@A1L=e!W$$ms1oP@FtC>Pn{z%X2=ga`iLl
zBVffcS|YrY5%$o19Je3^!?4!foYMF7sTu(IDd^v~3{bX8WJ7I2fR1(AiZ?K=eB4-j
z`5WVDspb_~+u+N?b<~%h8@Jg8$({*&K)ik5-Ur9ZXt)FiEvh@w1gq}GsmK&_7IfLz
z!9FJAU-Bg=y7DZ-;Pkz;J@Bj9fVw-KD|FIPOjz{y8|)JwC5XU<1k6<MN%_Xco(pdU
z8)?w130^w=uK}c0&&T+e`iY)nySZel<RKBke~_Ri7%F9o>4MQX3c?|Hv>m^^tB7p-
zk`cqnjyJf5`oYOGmr@BNAFhYURzK!^q*;Z=tfJc^LC>woWIpoeJ;q%ve}u*<a`ryi
zaXPdXE^Z51PDjb+MLMGa&^Cho`K7ZSrzTSolP2^`+P6v|4N4%$x)8(hHXOyZVREL3
za6j*3$ie6!RO&Ql>i=26&@1;1e8|}o7vYLPnhce7okj3LW;!W=zTTaY13MZYY{vH@
zzByzRp>XRf3resNwGGYyz>ZyoQJgTc6LaXQf)mF@N1r5J$oB_HwD?R~`eiMGx{jOJ
z3YTiVad&|rD-&plCSI&Kosb0G@HeIF8wXR|J=45=zN0>#v2Cn}>6s6PiH&&<u0}W(
zMy5)~tvsdfKG`fU&0(=}c_hE%iezb`X=0=9K|99fkGHt)tOJ;*>1iumy67&sy+AY@
zcrj^;EdE8H&%=~ST(&kbsc0w~nqVBu#Bo)#9@LxCGZhO;Kp5G3MGlCz!N`<R_hH59
z{`&3AyJnXH`HI3c4o=1wR2j?GR>%Pc#Fs>9(;#%F(~6LsG>zDsL(MX@Vetvpus+9<
zP2)B=hf4R2I7R+u)mKfLQ5FpN@__i7(d~^h{L0gbjP^k*$E26~G-VIX=f@DfXF;pE
z&%z!OY3mM^Oj?7nv$_l%ZCnb99P$E)Mn&|A!g_j`$?&S?IH6x(=E&QR7--vh<55es
z2G%72AR(r!IJ$b=dRIq(+{Rpda*U7Y5*Lj-xQUVAs|(<7Or%penfc_tf?!o6^|;qA
zppPgmnJ{&RNyDAN40ZElq+OPY5J!`k<3vC3JEI{VV-EtcL;#CO-0s!tKm$<Q?DzI=
zsNt=_{PlPoMc$0`%0KT1Na2PBY0bPi=dV|^ni-@Ou*e<AUzemgY+gC|U$6(Hr@PvY
z9b1aU*y3&`?BE)0_^V5?kAQ1&_e}9iN&{r!XXU2}w1-V=HDt5X%s+QP2UMKuId7WI
zM<OI&|8B4mM|Ly@S`Kt)Nt{Y01j%<C-k<Yhbm(SLr<@06!nY$v`AnE1qw)A;=Gao#
zITQ6@_;APNk@~GZ4+Qdx-97yyVu{9e$E7A6e+=#z{&e{zoiM{p^Kp}^rd@^M_LNCj
zEB~<hz1FZVv5*!L3G|CiODBTRyA-@F=a#xyjxTh+5}D-K_@pHdqex=RugA-Tlnj6N
z<67iwKIWIxOtK8rlor!)lo3@L2d8B|FfawE81Bs7BSJiv;WNtp{XKc+ba)JX-WLqU
zu*Vl*$UN|S*x^J{(%*`Zgq6J|V<3~>sznQMf0Uf$c*V5&+}IWDPK0=zi;Gq0>-FPE
zW`2QiPiRUWromj^{N7HMohVZ!K2Fj1I=SJi<C2G&(9X*`TupKt)KrFnMt*52X7^cp
zrh)fY(vyN|<;iDv-1Yse-buEqBI%;1nO7tK@(XMfiQ!&RcZA9nHAr<G1<WBUdm`{I
z>-sv0M=t;el}1`-ke88?%r9%}R64;-X1KWQ6E^nB&grDA!X8QRg5ujnom?1ihBnm*
zTV=hupy0(!E^OyIYkrV-L>k}OG6vJ*bKLM;pm)x7aB<njJ?XHH?j&FLpO%fn2Ct~K
zYJK>_%QU)VOwfWPj*M2oe6XTkF%h-<WNm<b6<oAGrB#G|)M58`9ZU2JT(Zi_x+o@G
za{zGA`z{DYB?6{G_oSB2Yv+kwdT>DDEx{OyZpY7@(`(Yp0C4H?w+CW3Wxt~z`?D(8
zTfpv9YCSUQ*PaKSAsM`zokFav2h4>}ABcTj@Z$Pcax*>yeLmRu(9#vzt0UOuZy0|-
z$QX90x>ubyrdl_|5Jcf6Z=zjIL@Ua3!5WYL!l}J}mF~d3WSM`H5_qXKap0_DOu<}0
z1YzITb#?VOIbd|ymz)0KM}*r=bKLE&N+PNN``Ln5BKfY-@=1h8H*OmjOH_u+Dw1g;
z##DVr<JzhcryB}Ug{AXy3tVs}I6wgMlou|toO^5U2FIsh03F8I^6OrV5Jaq1CB0>`
zw(T1rQj2M!bqnyZL?&$P2r%yk{1UnGH{+s&IgD;stQD+=NekNUCZfnZ&x5CjOpVP&
zE?_81Dj!yy3JQu*4`$Q!yve|QcDO7LKfW7oZRaL3JQGhXX>@6-eUHA0q@tyYNcTmj
ze(;0VpREu3MVqTy#ecGw2_^J*i`r$HR{bwyK+~E;5oWVv7wergqanD<ULO_Zt-!ge
zB$>;|`A@n{p02{+<gTh}C*H$EZ9ixylqxQnm4CG{c3Te&^fyVMV(<80GI7IXO9!75
z64P*H23uWblzzE^@-k#DQWyODHSeu=^khN%6SnOW>Gtr^?)a@b)eFW2e>Z=!y^C<e
z{l+1*Z((L{0zw8lR(9a@@Ud`kDucCh>@7R<ch<Wq!DRU?e0g0Uf_g5McMVhHc1dnx
zU<rfmK;5Wm8vmo89*mNt{jrKu0Tn1K6vYbyP(>X@kWuV=M}DCE<yApJ@2!6|5tnJw
zia)PyBHg#2X-XT87_~8&kDYa7O#3z1)js1QkeB>k!bMM@L+>TRQ-QpoBU6U*OXrzL
zos+Z`*W_f~9-Df$7xdYdq~nTr<>_O(t`NiL{6V9k9Aq0OsELenW*IYoQRd&tookW2
zlDrg-O*MvCHl9k~yl-*fM&U2)=XHO*-LSNGQ|?OoOPsJBGygDX4`mAL<<oJU3bg3{
zE(k>gNk(<v>Dm~K?6^3e7!5{nq$wqf^sYTJNJIkKWULK-A8Z!j665X}yjil<C&uqd
zE#UXXMvcE&NGI6^6$WPepDzkeHD&psCZ5wbiMcxqM2|ou)452XZ^d;Mb}uo3d|g1i
zQTO@yV(<k6nL<W4NveA0it3pSQ0q)06c{k@KI-D)kqmiEr>yM5tG=r5K;Q8!Hqn>-
zqSR8~pq=ItK)vZ(tLM|D3}URui!k<*T|Ky{8Wug#@3ed5der-7t=+(_>nrZ1h0fYU
zO6Be?Y5Mxj^NOmrD4rAtW2@q2YI*mxj#qYPmHmJtf@t?D67e`JT=7$+n+o5{RcF_t
zU1Kxg1r7Na_l00T1bB=cL9L>D8l1>{qnp`zL~3abthjCk{6<~)H1nBX5Rw)gF0a~2
z0@a&(Yi~sqF6rXrinDd9VZ#}dy?4xH-`?v2TNxh>9$^U-hRK^AF9eig%s=1f$g9RU
zxP3uf3^vdmh{slXwmi<f>+I3}(A!SF(fZ4++Zc-`(j!gK0lk0R7D_CNdci!2I~39t
z!&fKRbZz2ZIv!6c0Uuzp?f>W}qMvB?M;}r>c25cn@;eZ(Uo~zWg_c-`u}=LO<YO94
zUgRb)yLU$XiExdg2G;-VEaJxk0lQ&;(iEUoV-}63qQX*wVuyado(g>d9s84Idd<-z
zlGF-M_j$Ppm_K)joEXyWN5^e&<Yy+q%lmT?(G0iE9t>zoZuZcTG%?KNfT6v=4Do|-
zS)nx`=OTtt*P@K5fo<b)799j15^aJj)DkKE*FVvy!DwrwX3jMnb%$C(y`x}>Q)y9>
z_|0#l_=J$SM{Y{Xe+J5@B~l6MvzK;(N9H&`gBF2KeYilltPGCYTsGo+IfQY2p2-zO
z>^e2=YazVRerf5vbhH7|-HkqSIF;P|c8x^akj%~$w~#n`*1@6;f)>lxI*hLi{#ee^
z(bP9gH9E<P`)W5u#5hYcwnpyr@iL!R#+M^Z-L~`7$&><Ycqe#X0yAm%odB$qy+=)L
zVrFv6!LRh4QM-<?ZhDyTdX}O8?+fdFe4Nbyb`paN9P!r;0h+u4(^(rI=`I&qAV}0!
z@1)N8efjHfnYj(Q{Xs7ez9ER1&H9+h{(a6|1X&HP1T_rG1#wdL%1gC@fdNh3B4}Ns
zoA@;OlSLY>E|augexyl}z`>!}R0O*s>FRw-R9Jt(j~DraSmWx*?TT&|oQQ6i_V3^S
zldbhVost-rG5gi~6u8#a&M(5njIH^pDV^zsZN>v$a9H}HLq%cTqXfXQafglzg^h!P
z4Dq)M56}K)x*QyI>)mz@@Av}ka6s6S;L>`HBuv6`omW$e<A}v(Ws75uu@NoSy%(Zn
zaNTvDE>AKYu6MVNvEkqtd}bW(0A2tuht&rt{LXSP*8-JZ)<!(aF~b-38YQ4AIq35)
zMrbnxYe9g*cryU8ZlJp0xsQOV`OAz5gD~z-YJrxemW(P06;2E>rl;d(@XEe+;L7A5
z;>!7uGuFyct!=WCnj!u!tK&yNA?`pFGj{D4Hk&{`CIK1UnU3^w9Lvf6O~AoC%rw(C
zi8dmcy4~P_!*z7t#HHnEB}B5y+r!92BKnFV*Eb4j7Cc^F@&6g3T*t`ncvVWT6EoAu
zXO1JHqBEZqwQ&=?$Y*2-)7b|?4MDiuf~8p1lZIFu0!2sfHl-w=W>^y4tj&wlQHF}#
zm_GvqwV6hm=x9HrrDFi`aw9n%m*xEDM>-CfDau#!7e9yqtcfG4k-6>k9h34Oy=Sy0
z2n2$Zsp-x0WQ6sS3E2#w@Hrh>^!a1X87WWvVLAoc1?g2To}`=ijv6meHZnAGvaO4a
z%i4LpUlZc($?67w1xZ#i!*RCrPU-b3I>CZ`N(pNG^KlD!_I(So?qL7heR9^xPnD-R
z1%%z&_&VJfSMW8t1dHIimLpimK?`#A7zdPpQqNBrrQ;c1Ox~(v(8h84cV^tnlBbH@
zj`O_3c)18DkKJf|C(!!64+x@)O&OMf*p3&DjAsqJW2U>D8}4u)u6KTcSTjh>K7-h-
z{+|O~Kz87^GjLnTW{CZYaDpI2r0ZY7zA`kB1Ddi1yqyqWGDF90vgY?;n1^}8*nzoB
zZyz}8-c`>fGm*(IMW7*k@#S7!kpMVHeHvc8@ih}Y3!N7}RJ@WdCsNAk?rCvRlzaiq
z4A&rNL9<MT93S|l2x5EPU(?Rrj8H7{T#^S{OvQ(2>rJ3{=xb5IQv|`xv#sQ0JYkuu
z+d`jPAUcgEre`uuzO{c$@(<kB`c;ULnO?c7A)|K9r+Qa?wZ#Wj6vx19bCbxG6bP^Z
z(^=0XIDXkVx?N?lJt9jvMe~7(dHE5DZblQTS5?mcK#5iEyqgYftv@dDLM~EHS+Oag
zTuQBm-}$-#nPOk}+YK2M?EGw2dux3p_-WFLso|Fpzg}w2siuY}M!XP)gncU=vk9k=
z*lma!Xq&i&CB;-E48#9Z;TCQBe)#U_Eh*Std8=fs8t$-~rpGyY2Ap)-=FS+lmG^T`
zmVSr?&jZT08;>mz-Z$!C@-A^EvDoshprE+pRFJjd!c1!tavc-kQHHWym@C&K8?kYu
z!MGNVlrvP{<RWl}*eLhU;XoHbXAEGMc8=^mCoFo9jDa`Wy=EW>40hMpm-W)2&Sq6a
zCT7a#{Cb^l$a8f(sX&14V368>#fYG<4NC`3C*p$UZ&?D!c`3Nd-^D=e69TDl^p8AK
zrf!AHp9c9fu*gZ#Kts5RO3O|~GyR4jtDvZoa~dsrp^43E{V_$~nYegm_);U_x5ch*
z4Q@**o^&E2YTT#e7G3i>shn!YmkkI5JsNU$U}HmCpWj8ju`?|lZM>0|#AH%(-rS7@
z&bL5hSmHXNbJ{Y4phuRGyZVS~N8z5AH=R!Mr(TKvV;yGqWAoh21<1vK)lMqZ$Bhg9
zys!(G_g~#wzr`6nmIkZreq-uN=1yniLKn-oD}Dt=@h++xuRrnsixlhBlS@mi@LTME
zJLoh!yN|#n2Nymh&#SvDD~`@&J-M7zUtG=027+%#3;%ix*5ETTIylym>vQlcP#!7l
zE^@`IA#pKyWu71ey;rLN&ffpqi96Ct7%h$?hHFa;K)PXhcXr29PzidljEP}BJ!!F-
z`2=)u=0gvnYDq_l$Z3kcqHqgl&qsJpQ&THXp%8zGAS@(Vpd2coGl~{|W_HOInk*gB
zI9~PzugpsrM7!dUGtwWLeP%|3tiq)-D4=Crle3+t>O3mOkK(&e6}8PcjES5A^?j_n
z<2N;HH#YRy-||I#9MwNhoFR!Xu&LIZ?)SdeR$ER@vnuvt`3+E1zaV+iD;y;Z=_aD7
z@Fp$!>@%~KZypq3$+PoBz_1;;5{vLOlu!*GQy~@Ym>N8Vn9-BL-##7DjSSm6Ph^dN
zs`<6@Pu9)z67z=RxJPM9r?WCV1uQe_?w=1MK(nj}Q;DFqV9i>$D|D&MMB4f-9tqqa
zY7#isZ?{-%+ksdUC<=SM@)WpJ8MN83Yx>37a!zdmGQD~AyaI6*E5uILRlqVBzy-ZG
zA+Nf>T!_@#Dnp6e4>s5H28siKCc?K1;M5o<dC}Zond~D7%f;q;>s`g8zwFZsYRiQ*
zc*L*j#>8Q;0nD^o@Cz&7CkM2kvHSg!wfhc^jHp*lXCMmInz6~>!5OEr{uOXM6p8<&
zIhZ%P-PUckV)~QPYGilkDaU{jWFd0}HzqtKudX2j`CM~-Ns4qwbNMgZYU^5zGBp^v
zna^wcn(HChKuup<b-gqh7mO!B-p1}uq+43+;CmoA8-c(xj?9}`suZuseBm!4BoU-S
zWFI(SKKT8OqfFs}yQ$Xx$fI@!gF-&wO4O}ilq?1i$fV<+Z!^DH7i`v7&5CiP#BOf+
zmg`m~f)*JQkdXHr@%`oPb>>yZSFhWMX$h|!ecu#3m=uNLG&9T)J@Q9apoH!$Q*rbb
z{Gy(R5B|E~?6U-MJfOv>zM1azxwQw(^2Mj%TUBTnHqJHIk9=LAQH>+g1AFZdhQ+$w
zSunT#rnoaAPe&u03g+RgLL_}p@u~jbuWI%g7D$sb9bdc(nw3!W^*%D7p~ZjoRlS=|
zh}q|#ylz8jwMl}-M4X2%RveSfssq@5@&`IKD9k+ZsH0>AOfwo9w-t>+U6@tX(RD>_
z5H=4&NIh--%hv_n=P)bZxk`)^Bx!ymI!JoyVaPFDgXdT4DEY_WwXXG0=XQMHQtrWs
zixowNMbz4INa9Vyqx|b;Uw~TR!m+zExd%NCO;vSW3&QyVnZ*zVR#_&dKy(V6TLSbK
zt{ibDyZkt^t^c=Q`bhA~crqYbK)q>8e{3|+xBO+`^d+T$n9OK*3hVpMX?+at<v8um
z%2N*5ql6v1_Y#O{*bxH+pMAj5NaANyqe*@ob}C~o5)E{nKIiubf^Zj??l{V&cUL~@
zUj`vq`EUfX8c@yPmAtuj@*|K4Y#Yr-d?b~tS$i~k7g`by!fxuip~}q0HXVg+8lD`{
z3YLZ9q53F!W>Qa3_~*|)1CA#aDRx5@Vj1HObui?OjPxeRut*4b_U{3K@@xpg#meJX
zppf#`Q}Df%AlBCTbeY%sTC~6@0E_2t5^Xk>0%*p(cH}zHeWUSD+t$2Y9_P7eA5lqV
zm$YSVPFk#ZNYr}MR@VL5@k3X3Oc$Eh@ID8EfZ{mb@#*sCFC(kPW+Mf7V$tG*$4*rY
zN147{0#-p>+Je90TppTqqJ+=0&+FP>{hLyT336K2UasgQ9((?&z2yl}RT+2RE!cEX
z!G3P4Vn^lWn-#}Ij*H;F>o$^4-$JmPuxYnbOTJ)?UG0H=r6{KFUV65<o*7qL^czLy
zX`Coo3^qc>(1iLR9vZKTXW9#(f`BBV!{-7c*yh@DcG0lz;D?s{r1Otudj1taHQj*2
ze$!%Wv=~%~Aurv>59hbv+(?eo?u|93BG{%?K@tT}p-udv1W}=BUfb7xdQ!|B<N37R
zZ|^-^3i1!Q2j~rDk^b3bV{#PcCm1DQsJ??tzbHZ1zzlwtjQRhxrF2VE9Ev`Sj1j!J
zp-oknU`5SucCCj3ZTGv*YPw94EhUiM-UwF=r^UCyj@;vwrexVjUYuMu8s=gLEY_gW
z4Yd<xK@k15=fBFW?KSNoz*!)Fcb>o*_m}X)zAew(SqB4H-)#mI!6^6O0DfquWJFd#
zqhvk>`lHkC&ngn%Tb5-IP=>9jC#xhOy)-})9dFogK-!D!X4z51(v_Yoh_tB&I!w%T
ztxu&kG6fMA<>Ti-Hiy$-fv0CbJa7u_5If1qjFMZmmx3@BTx5xiW<Vl~(P&Z{GV#`B
zbDakxEd{Vr**v=dAWV{meTM3y0iB6P*xZwL_!Q78XsfY8;0xWiZr(uclH#;^J0a5)
zXF8@i1{MJy4e}4r|4fO^)@|l7Dqk1PRUp2cc~NQcK2zreI+Fs)fo~4P?tiGcQ}B3M
z12SwQ<0*69%L(P_kzau1K{KJc<=dr#)DOnlEXe)tQ>8ztcl7s}M>DbtFcGB}6n)*A
zc|tIrE3lOCVA&L%<~b;vxmfrNY8&PfR{#ijj>g?|1mwY4Nl$;vIEJ=ar=YD54!XGg
zO8<#iBNV-ZT*qNVM}N3&E;uz>2M0*r{_tS4S0(Wn(mezjI+`5GN3h1u((6e@jFV@W
zI7}vg9-iC<5*I^WGSM#wuFUb0EiK#v**6@e<Hi$h^=A|vY9}01*Vdd?ARp(L;@xg#
z$5Kg2%Y(;`kw5vWp5~38gW2@gq<Jk!UA(WKo7X=kF_X5wV;a>4pBZ@>4LG4bYsGc2
z!(teyi+x{hIRR&i_6hlmroMS(a8Kf5VP%EuTm^~a-GKhyC@%%fllH^6x;Q~=<$7UT
zFlDK)g<oWDq~Rc2&j-8L0m!Dgv<xUh4gzc<fxf`|%TsHHu3I2ir;>@Z(zN55*)&wz
zvX6(Vc~#9{&xF@MCg*QM-qhJMtv*l+5W-Az&<AVs5kcQ1d2Y|{D0(5{d>?<%V;4t6
zUIOHhMI~V11w~N%hM{&RJN3hV$H#h1oU?g*7M_3dC&A;g?t&sQ2otFH?c)57={0jf
z19~p!mc<CK(E`2l6YF5$Wpr2-Gt0m_JD(QO4xh~QMXeX#`fv1S7+$apGta`|vV9yS
z)|9mV<zY`XSB<`U-DcPGSYwgn{nhi5M_lwfj;kX52?TL5F{=bY5H7B2Zc+4n<KG`Y
z6rIQ#Kt$Z3e|VV)G1Jz&km>u2E#;>Zi}|V1DFYiV*!q7GuQL-On=P1EWJQJ>fW@VL
zI4Twm&91m^fnyN(I7hu$yuHlFprJ|q&q;j>T9uqA!%*Xs77G-J3}^J^EjeI<!sDiL
z#`x&(o(6rA^phY9I2r*{@IX<U;n2ILZ}|v%LRev%h(3-Cy$P%lfmOR8&<J&jmJ&f#
zWq-`R55RKYva(LtNcxleG}G3pTn4SXt4~!{96D^$0jOO7LR(R9UOKBYZ)Bc<>gsF=
zuD<E`hrQ*58vzTQ(Xl7mYEdtVy6jDG)5B)KBIu?&Rfi!M?}{r;!Yr_4xPFy#^-buP
zmQkV^0{d1RC1Z84Yy!9I4&&w$<?VR0sErq)yDocY<>{2}<E17EZ-XF>o9w2@zESL>
z6=0yLX;FU5R4=<>a9aW@h7(h(Ys<+Ho&=6mWV{I9ex{A!swMcXWj0F7LfniOzx8is
z_1QLpATL$GL7KQOh<s9JOiAT5ij2qmn_$PGg+S{{R@oVf<A@86YriN#UVtPHec=?W
zC`wiiGUVm&%BtdlH0f|WPJ|K;CA1!D(_M#a?)xK(VZ-CHkdjQL0Wmr}hgXJhS06Z7
zG!zJmTy~YEL=kV5M+|I!slWFa2=3Pf5%yh62_3`yO(SrEsN1rC)wVLPsOB$PN)~m*
zMdHy(#_oM!Z+S`&<PGpyE*(E~HNi>xDcGJKqUUzFEoU$ZghkIlkWT;Mw!$+6!HQ~N
zQ|ddA4TyY$rdT%nyvb8?CYxq*TC8Hdj@v>)s9MY>azyWYrp)GQa2tK=*CPghX%k?(
zjo<+M(g~m@CnAhH0jNpeh|6y_2VO#LteVFft~_}!!J=rD=m7GM7L%78<lD=K0BS4B
zU+yi3<`#dymd6J_>j%nX#oaLb0sdE#p2#d6xeh5NDnU<_1vb}~M<K_rQ*eY=p|$d8
zx>Z{?4BVn-u&!Or^>JR6wVn6mt$*4AA`g@a4UrdD-k@VR3Kebz^KD?1&|R9nADX+b
z#9&|*iGEQ*PSVq#1Hl*+HO=2^l5Jp^-%u<_RRR9Kp8vS57C}Rd2;lz0K*Y8>;)>#x
zCQo-$6a12<I?w-dE(neby&&hi&Jf_Eku-(73f_9GP1CSo465K7AFf$jURjCYuRE(p
zoPA+rs2LBT_8{aOv*(|G>ZLY?VL<yqI~4=Fl9<kDeL!|eXIBD!yxn{_ih}%sj|aRm
zfG{6zTMY^YxY)-{!QsRHqQ&l;D^JzwcyXvz2Q%JQ)CN_2TE2;Kf3~^4o<Y4BxT3^G
zfH8Ov&N4kwavg}hUY^nWv{qni1Q}`7&TG9Zu9pb1sBcoqjlqnm1OFj^-Q)mZ0GosK
z7OZhQxY8a5dv#lW&k17%`hOky=C+4B{KGfOTJm~RS)Q_QJn~;p{!cC~{~z(Xzo7g}
z5C{_v=>Ki?nJ=-A@?W|luM&YFtZI-hvpX;BOpylzy6axhPQ!4gmF+ObN?Ou<QVp;^
z20ue}yt#?=LfzQ8<$rOH7R&@$l!3tNZjTi<G?|wDtt&qWvJPrf?P*qIbB8yP&3U?t
zC;E#jCgJC!ohHcNfgjN3<KMi)v*^v=K2B3aca*<VI_8L>c4o?B(NwHDL$sAciD_K`
z7}vq$Gz@m-IQ_*Uw1dPy<RERh^!b`q0&-XTr?PLJ<-F~&D%$5|&b~OETT&#ScyrDn
zK#4}b+QeQd9a41--Bd%+1gj~jRa+omuLklkfV_MZ1Ebwp=DdL2tsw%zY+)%oUB)>2
z{F-oct}XzT>@pMx^jdxjg_y0X;?0@gC&m}j+|8njaBG^c2}35L*Zi>U(Ttc#%zCWG
z6qqS?ws5ORLvT4iHTLl)E-UE>R6~?G8RinaD%`57ybXVG6Nw<LoLokCpmss~+)>Sx
zK};H<cGsQ`Lm8ry-;u2^%}5@}e*->HOr+)Di%kgM$>rWPcm7xc5?*FGK6{T(-o9s^
zmb*JjzPKrp@9!Vo8O70c@Sk1LV<RoRMc)UfM$ZKWuCR(%^?|Xpws|HvCjWdhG!`Vp
zdbDv8JzVCrH?eci?u57VrW^nPcK5|E`QXt2aQS^(wyIlB!M42!Uz~>JRCGcx0(DXM
z;Fq0~>y{57o~f_m%{5<~2~J^*dveZPYo)t~&pq(c&e7!<l<TjyI^Tt<sZ&!0GyXsT
z!fsL-jfY4`e_n$xCP{%9Fon?vGkXp@ZB#f&?ybq+7Sh_j_g;p4GD?&kdEVK<B(vN;
zkj*F&1qpy{M_s1G=|(qx)BHKPsKu^r>P$m%Z`qIxiLHw>7#2eyKK1NQ?jEEp`8!3&
zl%SyM***1ih8*S&*Sdy@NkQ+DPi*5DUcs6)*tan_B6BgnF9LhdxHB+UARLGxtP2Xl
z?H6bKQci?5_0BmAtPw@q%ACnte(-?P@KM}UonHm2w|__*07voIh(FH~hf7Os?l{!1
zmks43#%(~xS4<J3m<j2`^D=LVnb&uJS6Q5s6`)PA0@%?hWbFQ?<J6aMfTq%!dWtXq
zz}%Db$3dyoAjU^wXPWc>Vrs^BHSYu8UFpO&dON%T5~?YlzcYtmXR)cN%w{+WyoR{r
z0k-xf4qtHrC`k{2_<it1POr<qpzpg&cHdKZG2LE!`reY^4`>?-?w*-6Q#|(1Z#cmo
z(X#gYB8W~i+%S_h3ZM7#NfE3olsm#|blsk31OCySphQ~Nq<vXx?tfV)^rVoL<lGi0
z2x3Sko>}<;W)1jg=sX#@M-Uxm{w0_D_75aMx_>TC@JN^l`*)Bf=ZQ+FIJFojH$^7<
z)kI3H3Fq}*E)9RA1#demswBS}Hg}Mxho=qH(`kmHicvxd-kAvw!~!wupiohj;ze#T
zu(R`lS3AYLpqf&aA+Ho|z$h=~iZeIF#q&g8Jl-ZXEy9t;SQndAV0%(hehS}Lw$&$r
z-x`AAotpL>PxmE6$-lWR!ov0CfCRgs931K`2YFZ$fv}A?7>53I#MyzLSx9ixw~O8z
z7>jdht?1*Rz;Y?=$uj3>8?e)z@=!2HwN|tP%4$LS8{N7Ku&ebO=r&Ua7M3)>Dy^W&
zLD=<S>d&5}po&m7z4+Gm*Ub1^#9R$T?TUYFqIlne;?qk;{00c!c6xXQ5HH$c0t7c(
z2#R;%*`2H{uwqE-%E-R_RX|Y#X>45vNYH9k4&ue=zWZt}$fjIOFcW%UAIKp9SUv>5
zk{~D3-yRc4Y-eYrVx+x2yz@#?#RSWZ2Cpw>J=Q4}*~XJ+K--{AhMng4nYt6(L|sV>
zIoaa?DPyH3j_JKeK)phe2A%#6)*}9VjB%ZVZl;N}q?rc_G3^o1J)VYlOp0;w0GJyY
zl(i!2M$0>Sc6j8w$^?c5BrsCLXJDEB8qLAMmV!L(zi(8In4+|@osSQX%++~CQ{?cC
zL4%LNQz7`}fVcH5+qZ8u>)&3oVUq$lBWA^w?OeDHyy9gbWq=vWn3ZiaRXkJs5f|R6
z1dYv2)lj$q<1qTtHEog!qJyloVn|AJ_F(L;9PBTnsV{3Zl(@NkEWrWFRt^=OY>~l`
zupdxMRNa|o8Tev=wcpR#V?pxxiXjGWsvms`mH|B(`sQV{V4DDPj^`uU^~cD^Hhf6b
zfMQ`-cdTOg{0$ols9y!P47EL;YuA(--JTfTnVBtuizjdIsgEJq4IAtUZf<oa`M9dW
zS9{$!z&PLy?geuPTN4nNOvZ%TNNYY1VlXEEjk$mOHt}QWJfLBSwTfq=4n<c~^5fve
zCgiRP<wkae2Scf-hIKS>K0R0Xiye2h@W-uOr-6FG8pqk_4H0b`37+W7<5e=~Sr<b@
zp+p3X8p}?C++Y0vKK0VEDiNi^==Yy!kFbNXhq<$Q1_S#;{Mlpkl7kE8p@TA1gZfjp
zb(=^MMeO~$QCI~zI)?1%7=j(;p!Uu;G*koO<|vrfs{dSs_>IQxY$58P(C&=65%(3T
z2#}0liXn{;KJ{dEX-3xa5q4v%k`3=<GqOJz33#%z-<3U6EW*pyb=6;$no0&rMMTg|
zE8Ep2h0pLjyz)$B%9y&d31(8jGVxE>H8UeLGYFL9Qtj!v3NEqe&6x+Z{1G4DmwEsW
zxMqW%p4;;rPk^6nTmCRQVfVLk2Rc<9DPejdH6YYkh?c?vqVzB{aXql!_2rsQqGs7D
z2V(fK($XY~;HZ0<&wNYN_^i=LMRu~OnlNkQv|iZxYu_;}bso{y*hl`F!4TIySy)zm
zlZ`?4Yco(4D$45)Js1FynPef>p25yO39t#I$7@<GJ7$<Nq5TsWd%L0q+5i1ps3Ay#
zc^6Dc#bk<0$pAEQ8f-$lUbA#(4z(g5!7)0aSmZM-0oIVQvVTOf-&qyF!1A$_0un&B
z16-&z;Y})^0T%qY>|7o3scC+R+9cEK;Bx>s06k0)F^>xEh)z0wKj#2%#iY2)P<RyH
z!4@>3A=Cw;6q@ze2%8%sRQmoQNfDLZU@=Q+R-~suSC?Ywb}9uQH=bR}v#*pn6Lf}O
z^oE6zGAJPC(~DAMYzRa;=EvS#R-CU3-rW!7ST>Sc$arjn9F(&?0I+RRvu`+QMpfa@
zM03MTI)2zvz3B5vP))P)V}wFC8jXPBqL_l2uU%DULW99SS<_Cy_o^jT$aA6yBJ{=-
zFcU=lxK7T22*?k7zX09<T!P&VPwgRlrdH85v|90Z#aB7O&sE^2t*aoETy6<Enm|JD
zOZ|7*DySf9_h@}^G+YEVxomSDW2j+*FcI_~*u#Jh+z2Q{Um7R|9->mqs?bzj!u{AJ
z#}<W40nZH{7eoz=Hc66j+kHiqD2gNgJ4GneSc+=aqdTTpau_u3J5(m42SCAi^YP2n
z%)I&G|7E<sIWZ6|I7&gjxJ+hHSIqw;ERM@bT0bl%N{a~)Ej*NDOk418B+E2r8UQfV
zW_#a!YIDH~VCFWvheo$&GvN;t;A3(`w<4K7&wW2=Pd&w{e%R20yCmeM+);wuf5S#N
zZ<)*Lqk7=%0$^|(mu)QYL}3}!GT+Ko;|(#M-tdxL3*rvr;)Y|yru83^kFO*m6;R9>
z1Q(l*rzQWBV4zT-<FX$c#U7i3WA5RlV1EE8bW6mjuVVXKOi1=&0Zi<aRinYUV*(X*
z56r={092VJ2v!xC&cS)kWJi;_k}%m3R1JK`75ffc0Rw9!`i$)tfq;?IP~vt}_;=f6
z(qc@A|2ymYon8Fa7Jv=8(is8Yincp#&>3@Ot|rmRaf0GAk^S3ExK+V2cRW)&Cuxx8
z3>m>6hdm1jDjZ$k^4^YtE`w6KD6oj&C&9bR#JH((Do3yr>$Yz7(U7Mj0_!(ae2HOr
zv-VH<U#&I%{C@{14erq6?V+DmK_BZJzNQmu%*T<&zxJK|2xjq5{m`@K`Tr4K_y6o?
zezo>De|pmq`ZG3tu1tzJKCO~39ZUbJ6ctWK<E2UyGd?KH+un^uc=uRm0*ZC$w08O)
zC!bl4U=948nC++}ZX0jH|Dm|F^!~FU*U@+idP|tkEZZ!-VZ@T4s*!myAr$S;BavsZ
z(v56*Z$ig{x^Ii5uPFZ#tBj^Rm21zHHu$N>dER*P8d`B{ZU<x@rSsUj1%k_Hc10Et
z=HsVN1|xS3=h}GM*!b#M^pON$7ldtOFSqX%bZs839Q)+09}#ZFXmrfh_DKKX>w-hK
zmv#%Cd#@-%xQ*h-U4@(T)N4YE8mP}44)^k$73RJ&uQ##hc3W0veSbN~T*8MMUz;rY
z5N7&Ydijz?zbi^jCTA@xms-(ES0!;1I<9bN!UWjQJn@RwXq&A@*C-C+L3kP@FrmV?
z-oC}l-LkBdVk|ndZlR5fLeJ%pGl8~tZO#*3l~D8hms`+}<CSgA#W%BtlhY$VULb~9
z2B&zyn3X4gGm9R&s%ZWi$@kQc2(Pt~$>_n)uc5R)#a|W*UqJYTK%&h8-4R1STXjJA
zx?n7)VjG#&zi5IEG28H@BIk<si^A?43|Hxxk)A&5_>@hNt~Z2iTKV2rc1w&UZkLQ(
zG>nHpVD**dl?^6y`R>=w9XMOTS}~?CZhO7(Y4Xr8b2+v4V80c8?O1D|`lt2L@a?T|
zKNySd@`{nbZ}(amVn*rPvS&!8-zbm9f&&~mDX%E+i4LZ0iPW0ziJ_+W#G)T1Y4Y`O
zs7GfBvM+ORyi6uiVv@~?OI;~vHJjk%T(iBwU=O9&9yHi>3nqu5%|@b!9{U;0`4Mbm
z9iKe~6@E0?QOYY&9jRVyE`|<t9R|47H#8+n*>m3T)K%C!9A+#%wfF8)Z096H#%G`S
z8I;%1v*>Ga$^JlXf(b)=vfGOZj5Xj#K7Dq-ij~6&>~c2vn&J!XX(8fgz*UFy%XQG|
z%CnFUCf>nU><GC+(aW<MW}8ulpb+$2u!nZUru<s4C-u-gR0%`L+Zz2DX|S6~Fyr7n
zF6rDc>0<ZOpU>nZ#g4F_#+h_sBN?5HGsE{3DUpW18Z)TN$yC;}hqenenboP)JY9i6
zWe!&XlF}sv0{sozs>LzjcZxsehtjd$=HmIr*E*sJq(NQ#F7&<5VN+Q5E)QLuX)D?<
z1%ff=P`V{wQ`iu^J=7xk7#1D*kA*YorL{dN@_g#xcJQF8vr92Z$Tk<}S_l4R7JsJ}
zoXz~$2zF=QmPe0YfzmdCPa>nxJ~&N3dJPKahK1c|@vm!V91+$v&dUN#QG@MAuEBt)
z0mkMo--yxfIP@cj1v}IP)wLWWy2F8iw+2>9@CS1#WoVHe!4W8%`RC_O?%H8>%^IGP
zhiRl?3r;})i+^lH=DE4<r_+8g8fp9o%CofzsQ)jTYS^+>dl^NpfA2pPM`w~cH!;m6
z9({hhZ9CMK(332?0@fDF1uzg6BAV5>b{#qU^vQtU(1-tH+W&#s1Q`_*_mt3P)erlx
z{%y}W-4QET2nPjl)Z<wqbi73@3iW_!4GX?4duWU;X1Q!GEgXiUx%*XGs$m+SoJLZz
z+uxV+Z=q}dh@pQN9!wQBPV$|Sc4~cVBFzoYSc=^puA`qP1jQdb%j~Leypcpqd($&4
z@Y#p{%2v(Y5rQ=>8d5x5-OyA9lPz$BEv|a5{&1e{Ix6^&Jgg-$stiQ3zZoUT&_r_h
zSWEsBvmZ1>T3;Br&aFRDT-Xtp12061cf}rcU>JV}YRg5%IRmg1L=O$8K!g}ZkOQso
z5BOR!mzJM~{wyNLMQ>YH>O1J8<U9FDBRzcYL{nBIH^RHlv~!S6eCzjNjHcsXr7Ev9
zjBe=VShiEC?@2@5z{n6R-R0iMR2;Zj`coRaEz7F@^hxZ$!Vm!T(9i<Jv0Rf%poU{^
z{YTrOC@mfO<UAR=slU7shCDaS&+s4t$W_tz0QF<|YcL8#XbtXrJ8_SIK5V{de@E!@
zj1-InBkZ~LFmZT6gp6ez%7_3F?|zl_qnB@0oJE^u;4D)PNYKH*4$q|Z0JNQ1j2PSa
zib?3VzyDPwCy=++=Q@F1P1+?<ktM$-jxgu#Oc~S@J@(7%v-%69bJuyze%@0$&ga3m
zETsn{tzcuiuEp334f?8kUDUMQOri!LngIizPoKo-LuwviZOvK75gNOmmAtC44h+oL
zT}ijvR;>D!H3Kz93goaX1Go2|)L)>SyYy8%u69D}P+H7F4x|Kx^YH+F7<#oRZ_h9{
zk(#Y7Ou<nlxS8ZFOOth*gxOp-k^%=`W@0o9BsI%tRUE0&g4E>7CzV(P-AP87bj$9N
zaVy&c&0q<bAD%Qd@<2}e{Zo5qP{IaO*{J4cx4%CO{U^b{G`iw9pQ?qcaQRYBN6J+C
zkBI$yuyqr6(C9SG3=yGxf_ePm`PRB1H7CW;7tQGwJ&^gYtZaU}xi}h~fmvZ|>4c;@
z8(PweirWS!7I)VcDjHhSYOt?{8d@Y!Cg`N>xnn6}q1XOTU&+4K8oJR<j2O1z$V;tH
z{EVAUEwu))JIA_>|J5TVZ4b`-5$g+3SMY~Bm0w`R5AD&>=9aKv+m4I|*;`UL<PP=p
z#FsqTV9kZZBs;e=8^MR3JBc4k`vUlD6S=hq0g{r7-ozi!<NjXa!!?C4MYRvLja^Q)
zzn7SZFYsC{NEMhI1-){=B@Se+4zIYSsD%yw7>3|hEOyOQ6$g8W+z0zJk<#pa7=Pz~
znNa#LbF$XDg)k(ew&RG4OE?Bk;)}-4lw?PcT`%k(FIlYbV~CA)w3~R>=L>B%A_)8X
z`4<2B+SBKt{m<T4AB}un@J*^6X2;fO*6XUz7x>}u@kJU(4|*8#kFAORAe3BtLF_8W
z?&>?3$IJ#P7~qh|WNHhOl|*ih*_LhVY%cy1C`D0kT9M2|R@ILEn$6~R6IUhbczr|t
ztCCjNBqS*8W!%=^j6}>yk&^#~u3Hws6gRa0ElT0fZ;<cy+ibIv$(_6ks6olI|E&KW
zUz;rIm2ul}91Q2@l5TrP4Lrk`=zPIJhB``^gxwe1e+@09XLQQB!ZT-S;uJq#N!BFx
z@93UT%=hD$_2ASFW>=S)ihlwG>yF-UT#NK__arSCYw%>mxd%TwJt+n1<7sV&m2r`e
zS^%wB%K$j5`Qy7E2d(fd`E6fm!J{xc0upz`VDJ2oZY~AwkxJWtHjDmIkS=#myU7(R
zZGCCUQ`D;htu%vm?iJTi+vWH~Jef@SV^x-B(MB&2XK`}HH3MJMdgyZ#KKh)uBOl83
z2%|JOJNlv%DUQ7jlKr_!ktSgS0(C-`V^Q+(N2jmgB_>Y0FQ?ny&I<~i<Kx28Z)*z$
zB}jBdrGC$iQgHU6iPBw3T>&)>0vp50PilcbL>wn#y)g2}`|!N4M$r+v@=@G|Oko$p
zl<SS#etUmj%*ZWL;F;ROud-`rxb)f!Kwh}8OBNQH$wzOH?+gpZ<^^KcR)!f*Jq!dQ
zJ}u;g7N4(G|BDQu#z1T;-#=hv03W;m;l#=%{}1*0l$6W10(DmNpK3E`bKu400|*$E
zUs|3w8y~;0>*v!9^o>kB2$T>A32`gu``dFHdSy>MGiRrWs$`Yr;$aM3^5f2E5~!l8
z!hD6o&1$f_5ShXcb;PCMeiNMZJk-rj(&qrPKx{baynp@PyE+^qTu$!`_+Y*9xd6aM
z)iH6&_Sr%hvw&-ufCi>J&z%g$=mQgqw({f1)Lqbvq6)IIn>UScD`W|rUCBLkbvz2U
z{pKtyD<92CDmfNkeb(N89-4;$`BtZXbWOq{&O%yPeUI?=O{Sna7*g4GvAX)vYfBgm
z_cO=pWeXdy9YJW)gck^@bpCzdaetf8-+Br9qz4IdP*_=ipre|b7W~+=JLpjE)Q91~
zX&7Wwwgp3(Pv-p<(}JYdkcWh|?oV%(G+KH5LJg0+paHGlB^|kju01sLo8jCM?fr4f
zdFT;#$9bZ6`}-*vON+3Hl5uGDS=2F`I|uY3k+Y2rO6P$}^hz`thYk#Yr>0p`)WHCl
zGcL0F;7P2qSwkw<0udDelc3rQ#xM`S_Po-)SylQYFn~qyWVIFJre$%FU@T>z`;t#@
zbP~hyJ0+kjG!!Ix(&XR_Jh}36LXR*oy7HW}p}t4Rg;^VVuqBZwT|ZFloe}a>#i2<7
zq^Qpefvb&qRh{kHtOdgj1JD{XoEA&REq#8YtSKoQfh!-agK9K%&Zx39*;~xchkpI?
zn(u-0GCQ~7wXOvXMgHP%tUJJh3(y=SZLM$YYm>a(WR`#uvIh<x$SKB&&~&9I-SOXF
z?v?$BpMKZeM71b{uDyjY6_u46G3&o%gJ<7}IjGmYvY6qBjbogY1*@Wy9be$@s;w)}
zqFxH>@{?ui+lx@+eJGc&3p9jljWP^9ya!`0>(k(>_;+SC$I&T!>}pZyGYlkTI@=mU
z3--u40w<wL9%0c?=l4-^24UXm4^pLPkzUdp=N2XM3~>$|9P&^itv3uUKBos`K{b(;
zJtMl$vXX;wyKWRlb?jhX+(VB9DO)9*E!v;U>Wxn;=S87)@N7*^$94({0y@kJzXE=B
zbYd2UMA-cX<(fpn@Z^Kt715i)uGHG!joZ*h>K<=7oj|4mhhyZ^OSYZnxG-$zTwVB|
z_4<ra`3Q_PfV+)l%?_5vR{4jd;chmS^cFryeHJi~4A3lEko*zbq2Xk>R|_aIolOi!
z>IAvyzL=YhDcl^gvv=O+BH*eoTgTw=7o{+WO~<S^Ax5&01`jOlha!)6RL%Op?E2S;
zW+-OCUB_wnE6v5^A(%xOfT6x%XKwkp1sKDdQupr&jgbdt(hB&06FM~Ut7&%PHm4Yl
ze|i>-&{LJ}D2cvkj*w%{2K7Kix(NRDS(wN0+TD@b)58lK3`ST|+a*u-;d`*xcbOUB
zEj-SQ+v}kw<SfvT9+goJLHD1w1?*zd_Jxnngg?KA?!5bz-rxVmhf=+}XQG7Z<J!U|
z&ji<R9r*2#$WxqAV4nEXOTpFoeG6UOw9QtNR_I<ZoEvv=!XtZzv^p7@lO+$#?!1Oo
zF6C=-T7&-UhiSJfo)qJg9GoXaY-~Gop$4a8ATm{_BiN(&6%@Dy8k+mjwQN4tc=@a3
zo!lc~s<MYnhzuE9c@k^vBJQSxzX!d9rd>gYSkoIjsq*(zqSnw4SP=cXU^9F<eZkkh
zq$`?3Tr7P0ay&(DgrIt*?Ni}Yud#z4oVEvol<@QBV!;<U5vzRD7(cikbVcu#^BCYZ
zP+rsMmySD0i=u{c8+Q|cSFe|Jm@N#1m+DR+JDy`K=Vc%tBnW<L)ws=V>5979Od@<8
zG3;M{g_6;ZL$^>{-u@Ca4d<kQ-!p~qmBc5{nct8{5T@`Im^%WLf|zC*Ae}Hk8Gr5a
zSAa%dj><%KQdFFoX5%YQ!c^RzxTHU|0rJp<dqjYXCkxHjcDZ_B*4bnlsu@%zykE)*
zzQ652yWH0_%eHuD8g<(oU5OM#9pL3BNTL&*+7`Hplp}ixv@&GP-sflQUjqqQgsuW$
zOhY5SzX3Koc8j}(Aq|6P9k+!f3>#3j{0YrNHv?g7MIYB2-7vGr%ErneW%!C1so_#N
z0O(3$259q-;U!J+a4>@5xB#>^*!>6QrjV<20Y)bm^ZkL~ab_^PvOd&-?mX5Z;C~$=
z4;{In_s1MvoU8`?;DfslpU>sU%EFGpuRuClWQk*-_<tdhT~e1n<#=DP+kG=Rohmga
zlt5sYu#Er;80Z|Q|N4Iatodc6(?v^(vvp2Og5h8&R-OOF?x*iEw+#HjJvwlYAE|zm
z3#29eH&TDZ3Kh$}T-2f{!TbPO{Bey4e8-M3pE6f=6pfAv{|L@oF=16j{+8o#ZK3KQ
zuS$a-+cmGKK(!cSTVp&hP~7F-2q-9sLat;ym59D8(ye5cFKZIut4u0fah7xeCgL}s
z2R2mJbZh1N5E>f4%G{Ltg8LQ{kY5oN^TB>?|2q9Q3S~nNbsRm{_ta~|F5~NhNf<ps
zos(b(xP;Ilh8jHdBs#1=eFm-UPzjhe{4?M#z9vN-f@+|$-`yJ?PAF96ftzyO%dbH%
zHkiv0=3pDg?GOFLmN3vq{gK9<pZq_(y=hoecl$1i)>f_W3q>5NLZYG|TA_%xfEl!+
z2%@Q?#Q{V`i8DoJGuf(TvWX0qLCmtEAfTXvM#iXs0fR0P5g~>M0m2kANd~f#o+oyn
z>)QWwKJV=(FNI{S-|&q0eLuf1FXTr^Jy0Bka#<wEw=Cw2Mj}T&RV6{B%Bogr1*~Wt
zL>G^lZu8<%o?r}I!&-=?QDzRut)9=MH(IhN{sg|fPc?7)Z~<wIp&17)1BCMk?&=N*
z?>0xz7D8EqIwb6}u43@1Aojp|24zDy)KJWEW-~NIQwGM2A6wCGBty+k2lz>a(Be}I
zd80-gnVJ*T-y|yq0Twn1rgAe7Yb@$L#93G<DulvM+t>h<zDl@`5|KpwFE$*iE0#*%
zym&9x>i{?qIsrLR|2IU;mb6Y|<Te38QpY0q>#)xoK<?AmmjWp$IDu9Pu*?sZZ-(o}
zRFLOJxIoSil7!-atmZSzI0u&CIl_JwxB+KDlMh&+F!l!lPXRX!)TB`p^MZ3JEF$>=
zKsk&`Fkle1PR+{7tmRL{3PPWOiq2!TmIBlFEXU!(IA{gxHKP+Xb>yEdufy8lMt%CS
z1CK}G8<m=m@NXii5h!HP0))LFK_Q7tf2@ZB){uctsMH!mxUbp##4NrTtV&{spgP%7
zztY#>8#|>I`4CI8!eJerbY1vVq+)%W1WO&piO55l(2s=hrwoj|zP9vqG5xXJLunG5
zx<kPaS%w=(l}3jf5E^6T#H`P0eoIr0ea(Ie^>U?V)ZI4Nm;AzsZYBq1J)&1D9c-p{
z$wq0Rl;_T6pu!{_$IKsx0n8_wV7>MYjsxnkiIrFr^!G<26LsTt!pFNr^^K>Ea^jet
zvR70Tfyyn{au3Z=VV3nlS(H#fQc$be&Yg;ZwF3UVqWlW%T3#JjkSIT+)H^UkBh(zw
z`ci9Pn;<dL5tfc2ecZ<Y5LDnJGR^qJGqqok=h0xAz*sz}%!tjeywn<*N(bJ~@^Uty
z(GR6p2{|rHs{C#LZ{V2pLLTrL&VzHb6;zMvfcL6}j;t-juQaAxT)k@Q|I1`OaFpx@
zD=SYns1d0laUI6DHB%yb!AQ~g-<l+G(lLDdo-+@!`iE!wYXAIuXM%5aPPQ?es!)i1
zSf|TM)WL^#o%_GtOZY$VeME=mdBxLNAiCi72<LxX!#10{4O%8i5~jhg&ALHj0<0E<
zU?;VbZTJSO>>G@_f2^uJCm;w-#DzTZhIm~IF|C;(u-|Ok3Kxb21$LH|MRv%5p?uz>
z1ZmThXy{$C6!KB|D0dyjS0YI^*2wr`TkW~6xFO>|R+py@A~J(762Za$Jk<xN-!hEl
z2B7g9VNwP3Aj_)7xF>e4XuU|G(K3*Th5u4wG`e%qmgb$%V79`QNj2lpA+LW7*lCni
z20i}kcD{IkdXvoYY6vf_93X;+xrd?jsJ_YG&KQl-gQEAQyu>M@;VQ-<5heUzts&%A
z%0YZ%;)$=15tKE~f$9TT4lap!j-70hK)+DZH~B1yOynay0OkfzM7cqWkX@g^*P}f~
zeq2PUm9Gyp!ysl<PG*D(l;#7D`6HisSZ-t%?**~8`dl=H6(?zzm&j$%#XiSnn;{|=
zYt8s?;K|8}Fb`Y{&j+e}Miz;2-8r9M50ZHiEdJOeyVHyed<41L0Ke1??`@L^nr$!$
z3KoG60}%99(X_f!Q?(JR;j?OB$?J7Yq6wMS%lPOF`0EiPpIkpYHf1Y5lm|JKLL$;w
zIp+h=Or{3}+&NnxV=wUoU0~&S13o$k0NU%kokwLa=tfg8#)_XlXvSCb@e%bxHoVde
zCZ1dmS~chaYmmq_R?5fYa+2f&lh3ZDGg;G(Ccc$T5Wyh8bHvz_#cVTkh%JWH*3FHh
zoCHdg9D#Zg5&x|Lt42Fv6?GH|t{V<h79Y9&g@r*|CL7gjC$;v=0G%mdi2;tpTLrIE
zXGJN|DCF{SI+IN5enJejumcj&e$}YucpYX08yR^fBqCy=U8|W~56B^7^|?XVc|4wq
zT9!jCSL)%c#gPUl5$!-9hM)&=kG?M;vetn33(D-K$61X-LCyesZJ4MunfQ1Zk)#LV
zFNm;`ruvxnLv5!=Z%D9$ud`vPtp;J}pczkUmgca#L!LxakRG{tKg!|_TMcqycicEq
zd-zdQ527bHE3PGtU}Z@}Ig?stT9s0*kplFK&iV$9jzq>mgm7?qIwtFaDG%OV1#zHU
zr8ku~O@nWlu44dga=dB!pdN!)tBC}DR&=Zk8rt<TsqGWEJQkQCtVV>DgPpN}b&1CM
zb$Twb4Pa3VZCf!ua!g0ziTfspbD}k5DhiCLx`hsSQ2aH5Q?Y59>Z>x=a>uAW_)N<U
zhR1~WmdRu=9GJ80G8_yhS@4QPo{oK->H{mTlIdA4U=8A!-I7F>TiLMw91;Y&9mvN?
z!)9_KOwrH-<qROcPi$KyLY{=gO3ZTC`T`yM^Of>;9swqYt&%GVs-w#uPiYnKkn#aC
zzCPz1$ZvgsScEi2?Hm_~lvcS!_~jSGI%uX$_3g*tURjgy-IE;P1dD-vIbkll1na8Q
z$mDVX`R8f5%OUv%4V3H7?HJq$RzIN7*7^FF6`YPSwFZtHUS@z=rS-@Eb-BGETB%d;
zbz1GY?0kuA+-zsh90F>zGeYJ-wdQ+k*dUuSPPqsk2*A6xz^z1uLXGLeIZz%$@rg`^
z89fS%PwfPaRkTMT6F#<Eod7(<P>=`@VwE14TB29!jU+UpNAeN&6d1=ay%MN1b`U3+
zOttsrqm$3Vf+5tba#RhMz7vp?n3oBIQ&UIAH%<A(gf;>XAHnMx)t}plGZ^bo%PogH
z2HtLxt&c_`l;BY~hzKIZJY~`+(+7cpCIwL-9Rn<wP7n26+!Lo|It)rlE}W_Za0^gA
zXBj3tXV(gB9R;wk6bjBlYfgEO43G@QI(xYFB!U_L4Au>n+s5rhg6_%m6p}lz(c?HJ
zDmf1RpDYAX2(S~(cnpGk<6<i?r(iSU&|ZseBnyYxtY0nKXTTZ|j7njI1EhMLf{3_0
z<LBRk%uOU5QH)!G>0{9Wnt{dYncq*cxDIKH8OI9kf^$S*f;#mFSd$(y0H1)8rpJsm
zeEgFy@<%u^r2Qx_38C_NV6jPJ1GZMH(NaVWMxzy*EtZY3naq#VZ?J3xBAjLYo=jeb
zM2K;$tZHCs=*@N-==45u&PSETLZu8CoyzI?VATGxT4uo30lC&xherJCFK`x=1<Q>L
zcZe)Nl~C!icr_y#`F!>9<nm4hMk1_()dOx`j|!N={i4ebb}|F>+fb$x&A_Bv_If-D
zXd0w-PN$583Ru(2Q+*)?18yHG;5AGE+nr~i02KeD<tPAsjv+~cBSnY03~Z)SF6ogQ
zSvDn2U*zjq<Wk4ZeiF+H9_Ny&yV;<pgNc?2p-;s;VA*6d*+Z#;CF8^3Aaf`u+wuVh
zgw6=oPSjq<l`>Q+gVt<i=UY9h6&%$XX=s%*$MR(tVlgJ&TnB4Xff<P(C_NtdIB=Jt
z4s{Y1;-S-N>6X&~q(Z12JHiWE<VcjT5};cA#Ct79{??4Iu@DxeFn|P3EQ!G#^8}6}
zA-`+#G93yo062-pI-CSB%@Ve!C6K5#m?U6PRX}q_I{Q_Vh`7LUmX`t_8UZ{u1!>F~
zxCcG5QKWa$ESi~KO_6pJ;qF`R?OMzn0(+IXFrXFSu_$n2KuQV>wp`W?MI;bXpF}E+
z*RWZZP=%cgTV}(xI4={ppRyVzz6+y+yJYIILYWkIz_$cBDNtbN7AFJf-<ruUUPmd>
zoAE7QfV!;$fjj+RttCJ)z{N#b6>ytP2&z|U;Vt!G95JIw{DhrsG=olq0!ds5N2-SI
zDdPzEc&YM#c@kj8_P`<$%!7c^t~HlM!=p|>R~y1}oFwS<M{iOkM$9uZA<a^bebJ7d
zR;kPctFx8b7Fb9y01S{|1#esp5v}08#W}`fCa&V>R0tj<qA^(6&%s6ua5;e?uaeoD
z@uYQ*pF=iOHUjojEW?-%)lqVdT;BJ^4F<vypdwNz7$b57nQG5uB|d;~OfWWKssRxu
z^FnVZB=OkJ<{-dLT7ybe8BfrOg3cuzV_=cP;pmn`+yl?j-A$JXMfgk|692!g3JZ)T
zj`^Mqm|(5dMjX_!m~C3LTqWb*gIB8~Q9&&Ojw~_H#Nyb;8l~VcT2%&sMUl@<*#`$j
zGY&g@4h|#&BXu}@xDj62Krud<U`Q|~3?yJ{&qdP?!MX_f)k3MgO7N@W4HiqyDP`1c
z$A4qjikZ44ZlJsZbkj)=_B0|Fy+D&mJmef!*OW#aoS9CAd7UNvpyCAvD^3uplOr)$
zf$PG^KsDx%a5!T+N;UdwofAh)G}W<oVuTt!_lW}@&mV=7vj%v74rV+DMDZ}<Iaqkp
zH4s=Rmylk=9oMqR)y@8IR^88JIv(aP*pQnukUjC(SnX+y*`+Y(eNp_5p4m(O4+QW3
zw?A^y(5Zi{LS|s@Sh%FmeRcw+om?3p@CcwM7z>DX+Xpb8=r(~p1QghZYrCRQ4~(Q(
z2$CeIK-j4P6^8N)<ETBP7W`oh1UA7(00-Jx1NxmE<WPwL)a1qewv&J|j_(n43PcX<
zN&}*jfl4*^V|3scM%8!^>&&)>l_Pm&4uZEAg7-c9U3&55e+fe|#cpoDr0;0rG=y(N
z8(SDZd|%er@G^Dwha|33aaaX~6cUnOVY;uYJ7IXQ*Wl_*a|Y-|#Wb{22EZ86$Gfh(
zIEVccs`l-EWbOM?Fyqmz@JE+#Men$Ev*}V&`#7<VRBy|g;W*#1qpMWBhGM%WAwb}e
z4}z5>^w7z0Y5(}y)QaHTFzpS*`%prX3~_aeA0U_>Jf4cS*^TRfeBRB%pA-%bXHNh?
zzEjXMnl$B8LxN6^vJ!LJJ^Dq%ayXxEtPj(MK+ym~SSdcIolNy0P~sIDY~2t?MhGxE
ziUIVZGn|Q`^xL9>B({aJ(M_KOVuJ~o*1J&BulZothTD%u>o|=LfzZnk9K$9#0C*bY
z5p`0Kvjg@<$GA$wC7c`LTmT((OUaw3bSl!JM1C{_1AxS;>AetffvN}St$P>iPy{@9
z;$_DIvNdRzN1=E!E+i08wHgWA%b@?&I@=t|k~!~$5=lh4n$s}-hI7w&?BXh5C;!$)
zdEh96#!_5IgE7IoSlfaDh+8pH$)Q?Bx5<oVawxA;qPw;$fvHGrAJuA2Pof2SQysHe
zBY!`PD2;^L`~8x~4({+8mfm^2%DYh5Ic8?EI3K4EapLL-wdFE=XfIqYZG-84-?Sli
zg4U=9qOTzV+Xiy!)ao3-upn7aKzy05(->bB0nbfsJeP9F%V3ap>NFmL>BUswB*MTa
zTYI&2wn>s0@qDdcJDIOi=qW&&^Xw%_5KGDy?_J2HjsinBuKbnPWS1nPMQ8$Q^&=Al
zfjiM^IO&+Q(`-Kl22t{cWjvX_u(T8;5*^^VpPsJF3R`*jzlYc%9r=z4zh7=&h4@h-
zFD~8UyhAYZCvjF$FqTYS1bSgA4)q@^58M83XT344oqVCe#CO-ZE=)N@V%9;Kg9Tuh
z-;d`QCaM`BHPE?S@8CX88I$$&3y{<QSnUljnvm*LDveCYk0=_sdv|2niN3B6+c+Wn
z{yU{I@W`^O2mg5K^Lt<_sUfY%CtVeNsrFp%G0?rlgWf@lpaW-KIKnlfagU%Tn#lZx
zu=3y=<GjcN>t98720Ar`f4BK1<)>?DP8)6?`C~7|t$1JQ<3C<+?Aei&MA_;t;z%Gc
zsI74<X8dDC<hx0POA<Gd6wg^6|4wQbXJv(b?{+}y;W{D>?Y!AB`R|T5IS(#x-up7(
z0i$H`h7;R!;-@<?^3E3^WgNbw+fm0z4B2m)=DxJSVV2pVL5&4=K)l1^5SS#00HTz-
zI1scnZ>y9_6>?zXV9yR^b6$YFH(K9mPqSSUU<R<3)=cV@bxjS3Li+_`fvuL2xKkxi
zal_=Y9Ako!ac-y(<i#BLc+!&L5{br4ggC+V;L+FAZ02@q#8yt<?01i%w>ZfUHqh?G
zF8=d)#IP`JX8N{8r;0YOS-QgSW(z0FbEp8a@x*q6P9x(JBCU&S;U1t$DbF7L@mFrP
z-UJKUu4!&3+f<7b6J|J3NHmRwN6O#Bp+%Gig+hCG<U{%V=d-```m&wz(;dmOdtXNl
z9qw=P@Usihn9drCmI^x;4spaD0v?-@rcm^29RkM_R7$ouf4JvY+w`PN-<C^BO~U8C
zUB22T?=Ttf5e=qq#7?}bG$@R4$IkrShKEZ&HUR>J&Fuv6T~MnQ;9gmQl>r69NI?Hx
zi($EOp<Gcfyyq|rn$lTPlBTpwrZclH(0+5Ug)t3sP#{v-bpV-|iz05+RUl6n)GiVm
ze)28^aucYf8K%5+*TCyaN*U$OJ2A6E9S(%{tk~IxBoIfDkH%11P!)nuWB${W)Y6ia
zbg8AKd5KR*K{)l}Vz=pu0h9#Acr5UjLUFR4f}=1!#Go^=tXn7%rMX*ti7!Oo4gZ;Z
z=Z=Vx**tRgVp`7pkbQppj;&g?s{aki_q}71Itd|pX_94XeM3W}BBPip8k6^HR0<q(
z@bG;5-?LnbQxiCQHu4IucEqi`ux30>G8XsYlJ1wfjV)c5g?GQJoV)bzjlFVib}zwf
zu%rQZMoLsN1B3FbgF6KE5R1TojI?&5nLG_dFsfS&VB4ypqlaz#yyHPxkQCDsbik#>
ztSj}AV^XjHzk$dDGn6V2wdlNMG6a?DX`bvS77DpSBKgPaV!!qpurRA*b*zsD)46B?
zItKMKgS2HBK_ue_Mk{*R!B&ED#2U!JxsHiDO?5!#1)XaYjsx&pghDVEPuofGk1Pb~
zAxV;)vJG-$Z8!QjrBKFKfWJ|CyXcr6hHEiU`kV&jKn_R;Nr1bjTMutSqt$+3z^>E(
z|F0522Yt>8Fqa@qMv2$KI9ap<G-%mXFsFo!yqg-p{KpCqsE^e;hd^W9AoQux?rDRC
zQs|3@4eNE3kWicI9t}CP{+{y<08^23|8>}t+HhkujYKB=7(?N+9i@T@B!YVDc2kkt
z757=%?bPBgZnivN%K|$0R)x&mbH=rC{@m8>QT|oVO~$Fj(Jk*5C)XaY-v5AmG&i4J
zd?-u^lErEnSNX9#D;(>XbGq%!srE*O2RfTVzMyg!Rp_M2W4aBpToNeBfdqOOP~r6T
zgDGa*nB$AJG-n{J=r=nTtfi9$I<rF|vO0#s;^>(QK}}@e)U6()X8hyfVj^FlgMB0;
zpd4fkug{Fv!}3E}g#97TXh2vM!`UGGOw)+QdWqdnyu(Hy*T{k#N1l^W%q%N0XhzMr
zrlc<lyo_c>o}hk|?RdW(%uKzemq}uF4$yvUv)5<=`>0U%v=cz_7C`*C^=<e>{LVPF
zUZ-&0Y0{t|o)`<oAbX9OM28$Yg0JfDAVivR!I7Z8UakD-Fw4lFhrOx-({q8jE?OWE
zvQ6+6?6r}kPBmP-dK8p44H}dg)`??08Dh5y;(;_R6Zem*y$frix)>(C0TOunlR-{7
zKZ0wjMWjL?oQln~I1sZ)v;wbc(7;{r1>=ck;n9)p{{Ay>x%gbT{-^(rOUW*Nb2q=u
z4_XX=Jl4#04Jm9&^7Wycf|+Lc<@B>KOGT!O#pvCuk6j+PwI$JYb;Y8Psk7&45&zjj
zWTXGF<4a1?($pS~{aro|wjh^j3XUAbwpGfN8a<yHa^#P9nz2X1tyzAjt15ozn(_4I
zi9P8%Hid6&X{nr^drZca;NzrN8Y~_vk|2DUj-0Iy6CQJGWF&ug`>o&EwYL*)xi~lb
zWG&cnIXuJBvF}Ke5CoW~2GD8>;mO!4vYo|ed7#jn*O3oc6kWsuK7^!#$__FaMdIAY
zki-M<LSaHa3GD$ayqppYPMvRv<uad80^$sE0W68ecc<2%K%wfPK3knyoX4!%`0#_i
z{mT8#Y@g24cH29bes?(|vVPRhWs{$q;!=`#*r<-NfB$|~gok9r8`4l<BjWFd++wuT
z_wCKOL@CW{CC@y3B;(TNBf-tCs~?Q+pronEHWd!;Wa=}LJ*_5R_=>1nvdi7<VK7o_
zjEpuL0%%v%tKj@`dY*QGp-ODW$hh?mwq`t<IgUb`-*wKHb;ZZpzsxMA<vmY%^lGN)
zc<{WaZ$2cA5!xau5*MuU@c&Er(lcF|YwxFTP(d1ACTYN^9F|AFR<LWkNUTqY!UwsJ
z_L6WT>F1aH{o;T@#_p<rTs9IMR`s!p;@+|P@U}H=i~HNw=R_xAlf92Y<W${BdPOGl
zc!$R+S@jY*2<o=FD;46ghDLXYGu&LeQBIZvbPHA`mxxtx{v}8v0YsA{t+?4bT7r?u
zY>1%{Ak-u!7_jx9PdepuU;y4L4n@m_J<h?H)?hLZ!HhStKs*i^em%$S$DKw4Gz@z5
z8d=}80!JQ*d?}>5jGfS;EkVS<Ci-laayo+eN=}n?HpfkjV>Uxc2f12IMRjP_Ba)=Q
zzYhXwMIzH28v4#+uQ07#H`zN(Zeh9DxOsjYwkWD#bgb~QcbH1&n{`?&g{)mYjMPg=
z45JpQ*GE3_K4yUe*pzb~TLKIWK;%l;d4qwQAWM=p4I@0+fTxsA<$V7?4fP<$)fsf#
z)berXQf9nV7p4FGY?w?W;(RD9W-_iBOdr{>I`FF8MZ`b_;)zrv)XNa$3lP3gsn$CL
zYS64|RHf1XHmiZDhyycQlUu-I@%d=^hiM=_0|F;g#z8iU$x5<#cbSe!xE7@m08Yt%
z^4ViL$WzJEWtJy4*76CHTJZpBal7OY1tS_X`ppjR+62|80&u^z7M~Psq){wj!lMw3
z`qrZgGp>^zzO4+DpkA!CQ0no~%BfpPg9e$#&<o*_Mq|eFB@3ekRV11P29_8BF{8#3
z-p5HlSf?-~m(0|EBfo`CDO6A*hxvezOdysje~)F~ApEU0n@z8(>cV6rI&hy5Qo%_;
zk3y<~LB$z%V~kBisCf{@b4c$s69sE0fv<)v0~QRt`IuB?QaQu<z{qj<8b}OS^DtH%
z%PoT7081&x2SrGrA7GtrkuL)2lTJxxyXEy*V$d;;of7R^$FgZAMvZX<8h98E)QjO_
zSyxpX7~P?8-GTz*<RS%^Dup><a#PbVmmou`*3v;q93wo19nxr{1SEz5oCA~D?FE_F
zH?TMgM5<L=Dg<Uc)ry@GuQkzimK|zG%ZDw21Z=)gCgi+eKK9*dzziCLpnc(1cQP3S
zoyXB-#bGkpF!=Y64(>)%UMx}~8kO<BkD-idHQgjWmB+?5JKiAC`Q%1o6qs?QRZf9(
zCrF5x%3C%D;v}FtG|HX~LZF-%q%t=uL9>cm34J%O3CvUz*{}{AWUSht5afCn@>ygg
z0vZ9nz$(}JiU+nvk*y^WA~Fl&6x2haH!&d;q1xK5fqOY@z!Gw%1~5K+E}oWE!EXx?
zsGdS#`uXlT|9gpZ4~0AP-C@GOC{iXw(^q6CbLb!jL8Cv0)g`gAz!BoMRT|86GXe92
z#n-b)kOoMx#}<VdRAKb=4aP9we#54BicpA$sz)c-MH3j`#w`baoM6VG5HZ<-yVQdz
zVPF?N2?&@*hStm2&dHzs87>&8*aKc*J>bw-l+DWEUUCz(W_wy)8S*{E)8a9<=R|F7
zbpl)~g$^Z{n6FG2xKUH3jGF)|Yol^mb34i61_WH_y~HDBEz=8}6$Z`)`oa`wd*|rU
z=j&}jv5)|%K?yHCymF#09*hd>d&*XVrh#Qs-R!U{Qs9BX@_p~^C>8;r0cSE@jzRf+
znHdm-ho=|hXc^+9;@eu2=}GP}<S<k5A!Zk@(?~NbEfxkdGwA`JkH2|;aKrfm(jmk=
zWf;jR0VyiW?NC^<_?AcrhX+1~Ze@Idrf(P#k3+>hcoIY#b_(Q?r;#DsY2H6*B-YTp
z)Rx7=SO`u*&8kN3!}@8)P0UWcYHZA|jZ#H;Y|vx$S8$tUC|cjsH>J%^sZNseBgo~l
z;l)LuVG0^81_qzJ8}6GyhV+leLTLh!-{{B~*dGKbhbS?igB*mSBIW4xAvU3oN!M#K
z;gw;SmB!w%&SABYpU@t}*L}7}@U{Yb6msmRpUu^n-0<04DHH}UPUMIdR4fQ9g&Q?!
zG>m~~uavt>KyDCAu$XNXq@xjN{kiBs=zt-;G3{W!vmjg#V5n^Mk)jV<19btwC1S<H
zSLO?W)+~SkoNmNm+>E7e0L9K%5FdmpOk+hx_=~lvRzeQ4c%Lzy)&NLUFeP9es`=;x
z`r|JONDO``pU>)xp}-x`#mX&DJT7RrH<?-1AD965HwBKP__!<qpIkMGxY}3;p=B;?
zj{&l-0}k%ckjz3lS_}XQXN0hevZX>)7l{Mx3_0{yUmqrvGXb{Bf@GUp2~*&Vb=}>1
zW8Ea8yGkupfgCqd{*hgzPtfQi1?>Wn3d$4)jLCoICAZf(6(LCdu+B&Y0R90`Dh4__
z+nMty{H%-zO<n_=3ElZXmC?)$8yWB(HgLfB1I(J60;?*i7vArHt?Gm@@v{L2X0tT}
z!(lQZmN0o4hX>PA*$o{;lO(IXK?iF95P3cd__3s<7|6!;-~sVomh0G|gIEOV<yI>C
zagh9)XWGqbU|9pKh#aBuY4^ZrzDz<bh<SiUrjb;}v9qp(zcBfU?h8SF4)jt1@d1fc
z2`J$k=&%C<gevauS9%vJVXlk;+BT5ilyOP6<G*eU1h{|))Sn#c@SF~X#rP;EYM+49
zfjekZEe7rOF^j!5U^Uei(<jquME9Y*pc!YvP`jr@IJDC9mavcj2KzUN0$}ZUX`fRq
z5Z%g#Fr<#f($lF~7AuB`At|&NU;eaS>mn133jsp+|5m1itOvBgMxh!tb*BN=g3RMG
zY9QyT0&l2>W5`N8&BRxOOn!oQ;V=v>0opoR-TCDI6V?BZ541Nxg7(;iRl%jwC}g@{
zZ6Vb(L+ZlMIy{b|_$<CsE`vIteS&Wm)E#u@Wy4&$43&)PXnY&6X`?z&(2CN7{EMm1
z3{H+qwrQ}=*3wz{xCg}U{k}Vmunv3?2_6=HniE0Lpgg2Z1`CN$?IQ){oD$g>94<D&
ztZ<8KACHvqpX3%$5jQ@xr%iONT=hi(Ij#rN7Vkof066E@{}JB&Ms)Y+@21f)kR;dz
z3uvi#M$;H(2~YOQ6YK;DJxPOZsU8*w&heNwQfV02Rs_cz#3H4dhDnez!Ds-AyW9w-
z<rqZ%daV{FF&Q6AIR_vJF|`^&!MYL_(<sxmV5XlxTO63$3+_+zIl>8(_6y2|a$SOe
zH=_34X(pKJBz!0n0SV1iF0{C2W)p0vff0zuU((2-l!%!hBZ<#E5jTrPsAdvF;64){
z$524cZb&``{;|my<Uh%*o^d95)_A*8t%Doeqnz?dP>rzifXqWz$$KEa0D=wC#AOkT
zpz=XtDxMK)1v@dzkueyb5`3#jgT5Z1#S3*kj8K~wsl(t1TED20!Gy+FaB@I`8-=Ab
zOc>0eJK&N|JT|i|7M24#Hh6x9C5{5QKHaZysP?ls9U9QehcYTS)}C>?Mp6M+6N#n^
z@S(tAt85RdV^Wclp8o##mM}N2r;lCqS^N2W-vIb3oG17<xq=gC*A7`wj$Fbv4*}KH
zVh&M6X~`;KoGS*o2J$O-Am%?-DCp6A4zc3;bw&(nj~k~WN|%4E6h9qP8>zY%Z1VBr
zz<I}K)+#g4{oig^{2%x};>NN+ob`d&(JwvF)h#q<eZ6LIF7>_&dc2Z|u2uFSfO?0H
zZ6`ji<3;}U{h`xM{Hf<v$a$W8YOt{;e^jF}waP^D*@JV*Q_q(-_bZJd*Xr;#-;W?H
z82&=#&^3evT6IeFG>%Y?kW-&(h2`%(<kodzA}u7-(qSpWxyRfwE8u%0mNfBKUR)Nb
zesIy2{*R+Dv4a$HEm4RZC{E5Txsffa-A)nIG{iw6;komJ<=oukC1N{r^qCH#N8|Bw
zKZDb*zrGK0Eog1pp6|ED{?J4wsqmK!(w4zT(ygnJcMFq4V3O#ep|={>1ACu0Hs+Mu
z9bZv=lp{j)q)5CqwITrJU%X#%Xy2>qjI6AH{m8!6fm`V>N!x=66vTFQgD=6?Zs1jW
zcIg_wMTY21)o~=VYtQ9KyVs|lzbVb#{Mz@m?9D{!;dByelitn0mX@}4_5R9`_cuz9
zr5IZ%_OXDf5*sve+DLcRNOAQKMZ8^+je!A?cJ>-uLDn%NQ<vF(Eg<>KLU-z<&erKC
zuKc$sO?+kEv>iV`Bs;h+xV!Z0y)*M8U6vFbTZ&)fjM$6#TW5VJe)*mkYJ0l!#_fX2
zY?=1BG^G-VwB0}Vr@kvLEr}NMcD?Y*9ZVU@9ZKqTI$q#hytXB(#W#<>A89REc&3`u
zI@_=5$LTk(7p%T{HLrO}r}vcwo9c2topqpYxw(lO@`soC;<eU}Py57#t${mR399Y{
zX%dX8R=JZ-;|mXte_}YHL4?Q-3Ia6?LKF;PtFMxg?k@Xr&dtB4``z8LCi-W;#RJFJ
z%u_E(i_OVNJ0mAA)DILFoIT`3<u3LdNKngdanQ@vlS(a0H8Ql^qHJn>C&&->?2B_b
zTu~dDbH){>T<7Gdmt=?!et%*|_4FmOEy%%djHL5OTl=RiCgML<<JW>!GSto?BwyrC
z#)Ue!DXXfOSHcj`dSUfMwdqTL<X&osTS_v1_x#vH(r8x5hLfvb%sjELB*EY5^_!nJ
zJ@kamSo+X1!#4((@hd^I0^w!Rfb9WMJKNpeZ3Xr%E!tkk9CebmuBbToSnhDKOF?#S
zK~=%x>;lJ}oE*8WZFe}8P3X?uw*Aru&Ya-X7ms>d*39yKal&<t<BjkuSGWD}F#Yw#
z*p+@0f1vmM7onS6kHiFR^wbqpE%qK7kFuk<w?qXrfJ~FNg=Rndh;;kbHSNr|+e5X>
zM+%OWu-h6MS^|8uIciSC!-?H}o+iJIYgeS`k9a2k+-z8RYsXAFf9DaR<C@hWOEht5
zkGMzP?1^%XJb_=X3*LJ_wvy-G(MT-P2ehzGpxT6FZ9`<Y)4(<&7uk1K?7(_IT<lo9
zHjYCF-53%7s(8ueid#pwul(A7=f$Ny$9C*$YCg6jAwAiz@7Su*RdPqmX!qjVImv7%
zr$b>J7xsrpQlOnoK0vJbRhBGEPL^?<oSeA{LDxemy-G&iVWuDx0v)}<lovo}JiGb3
z%1Go&%6mqYbjBZc!A79w-qz8nH74=Lv!X5jDW2Qji#wkbokKqjJ&JOClGH(Z=Dfr&
z=23b$y(#J4ChB`&<LvSv_>zNU;L?*G<hH+bN$~$^&D{SiwN;bK9g7ZJGN=cx5SLcm
zsmz(XdS2nvF8dB3sDg+JB1I2I1t0M&B{l2`>U5aB<XGdYS*a)DXSZx|{mttitJv6m
z@<snxy=}>QdCP71KbiDqpYY@o#FeH~%fuwfVN@zgQWM`&^S2x<+8!DnaBStG_iD$g
z#Ows#O#cmcYC7*)KdM0dR`?Y;Wej9Mf3<_Oe+VYf$LV}=3f?g8uQ`D!oqJ#8mo_h>
ziq{7=U)^rqHRbB&<LvTNpPGMfx%lEh=dXX-*xhx${$S88J*g?6eWy<>^m0|bq*rEh
z<I~q}Qe^@Iv3V4Xk=8pzX~^|LBw8ivb(s7wq5Dh>ln#K<oj@c$9RG59!wOMVZq$8N
z<%t(>o-Jj?Hc$NSv^cXc*K4NNU3}5%6MJ&o#>t?hT7)dqqjfYRQ8lWT_mAO<yk5;2
z*c6|#>TJy6q5=O~E_d>J_c-7AVe11@$m^}9JpzxB=C8I4PdEjOcd1ZYflQmy`08F|
z=ZwdB-*MA_+4Pk7uI)ZCv2f}3%F-<%d*|HPnKbDC4m;#=S$OL1=^y@jo$tIRQi8DR
zli7}9CAS-@_BN3bky|;>&OFz4Rpox+zc^UrmG<$u{}uGcd1m0ByteJyMN>-CE6@1N
z%ib6W6kxNWUk<a2Z6|mB{kC>(^T3suuc+*<-M8Ov4Z2UYFQ|RJbR`w%$o>JlxpexD
zKTrP9JAL|^`IS)~0Ue}f4%wCLDiL%(sjGURHhD2an@5wzGa$25W9G>n{1)YhU0ZqS
zM^?|GkS==G&-v@#-+qjWoTXdBJrs*{morxRXNrz5K@MH2Oq)Jjg0jVRVI#&cp?r0F
z$yq{HTy)&-@w+kYXH?I#1137ZYq`XmpZMLi6QxH^JKi}KVx7Kr7B~D#bU@3MohV3I
zj5C;K^0!c+bv)b0pYxx%g01Ndo9~MPA1rB3c;xut&5Yc<n2yZ9ME;vAi?(dHolBnd
zZGJJeQx8Ms!`R(W=}A8H%kf7!!S4$S%DkT6djEKMPxy*o?kArK30<{e`>cnzDz~|%
z<=Q%IE24AqTT-*RCpc6RB#05!Vr}XJ4X&WMxW85M^!ST{v(t+%bJr}0c;xTjyu9x)
zJ-y<lFfHrcg&3m|bQ3x?qlnw0zxTEm==a+cPb|CS{l`kjz*T3R!~edxEP1><Bk0qd
z!?#xY@7<23rQJ+QGUVvLj3jO+*5Oz}{qz}wbF22t5})>~5(rG!r@q0aHXYz!YU?2X
z<2`=E*!U+Jdv>?Z^>FUEW|UOF+_=NhV^H1yGB~w@bCYo`uW6(pflZYg0b?OVZl9Dt
zDi{*a{(Hy1fp~e${X4nA^L~j;{}eZS;(7m>JvSDlIQrdMmHFmUX2xPtT2#^q*XdBa
zJ~juTx`VxlJ4j!B%DXG%Sc7+4k-sSN;nIec57S&`oI2{^pMGQ2_GOpe+*q}0R6c7Q
zt=BW@zy_LnGcohfqnW1`J!u*&Xt}xXc2(+}cV8&pmaZ*6($~4FZTQUBxiK@2ylvhR
zvM*`J9G{yxQ(lICo6P<|r&mKVjgOoBKDFd*sCQ@a=7jDwi(3sVH*jW`{4nRt5kl|o
z@b9QwLwu60Yi?Gw1w4Luw{dC3tt~q?4|tnCnC?3m^pRRYuj|?mWp<72^SoTQY~U1A
zci+Ak6EXQ-|5EO6-X~{u{wL;U=iMniyf=gM+t94dNsEtjRk&BgYXvmW@|Whg7v%JG
ztqe@>Ggx001&a0ur>%|6y_2%?*@64Zf0C?rTD^yo`NOR(8`#lr(?@1?Ue7!%@hK?E
zPU`hkD7C()#l5Vow;mDe>hjDaNn=xB2gv}5`YW$wsl4xO>eUzaXPn~>&s`J!Yt`Y-
zEsyS{?thjSG_r5Y8}YHPV|<$~cLlu5A6;_0<$Q5jT#tySrx49!U;Q~2y0Tv5(zSQ@
zbKfoW$9w2Unl~QSYvwx4i22<=Z-f6{(NC9>uI{Z`HpCk6m@Obq<acWNNeok(g#S3=
zPV|!e`#t7#$pwu3Gh&k=ryoixruk+8$(}|^m=nS+IM%qjy<@Lz_nql-ulCxR8~cwh
z`CdjIIy2A9VcCd$@#>n*WW~ndcJK%=hE#(pq_VWJ!3XqE=YT8HgxO0c4qjcla_6Ia
zcU~+<mR?=Du!o>pRhV43^+w!{6Hf1rIS%5N<rcHO*kleG!Fle5zop&CIi$ptO{FC+
zi*hHXrBU6plOpToPPsq+m61^I_Qk1N0-I-kWUGyVWPwDGMbv9`a(B$P<av3E^Y@YO
zA}`(eYe;MRjz9B11Hrj%A&VnQ&&<!v*;?MV`sTI3C9QXT(u3M<=N~U707E!dK@O7i
zqj?QY)e~O=HqwSCd<qk$d}5&doxli&zP*0xx@y+YCy*RzTHiqiiP+Kv*{coLL%0A`
zR|BmK^{^n8IFRanYj>vdzif}2IseH`?3f>|Tln_JMGtd>kCjZ{>fy3^M}o_YvB3Cr
zaA!^ow@F_YaV;qui`fmPm+XFGaICwc%5~Cuk423mYFc&GyYKFtbX_oZF@50Cwv6qU
z^ZSHHPrljY=bC=y%^H{BhL^!KCsrJ3jMZ?G^keGJ2S2GIN5gBYs&;S5e0+KDngs#&
zoe{6nH?LXNl6>Z7O^+?I<db3ULEG7Oq18M-h}^|%Z35$N@3vmEtS59}cYDCtGf$7w
zh3+e-+;H?OU5vO)bEG&fy<3r%mIk0e?s2YXaxaNE5XQ@cadWs`b530BQ`@JqkHq(5
z1@B#48*f*Z-b;VW%>6l{s=>>eAo6*5%VEvm%a#_ocnlm~?_qJe5}5Rrhe|Zg{2Zs*
z(Lu7^TmJmY{+G9(ax(DYt9OrHue>XeFg73g>vC|*yWVCO*Y;(;JJHMl$bRwBkhlKL
zPP^0PU+HT*KQU$eW~om;FC}~6Fuk(i&|g2l<ua}<sk^Bsl?VwPf~<0A7vQToCux^p
zw(x`Z$9MXD_aeo3Sh^~#N;FF962g`YWEOn*DRsQ~r$3#0992ofC4lX1f^xeAl6uF$
z)gz0WoyL3v2W26Y4Bt@O6J;f5{j;o>CzVh#zRM|EIdA?4d0&QW4#O|~Qse&Zb^;GV
zLev;n<xrD9Y$X2i;nNNf<N%2Uvh(G@_MI$pwvJ+}&~f&FM_xZ&xInmf_^dWF{mhZP
zCgq`9{<7Bf{H5<&+k$?ZHWZynpKu}3e_x)XW@~VCQ;4Y>726UomTxFJn~>46eAr&-
zP;qa^Uw?hvRDLN6_d~OPRAuku3i48ypJv^1`Qrm+y;B_n7w#0!2Q@#5D<NM9*lp2!
zw~I5sxw6uK$s(6EF8-?)gYUcjHlO6@x8za_`m{&Z1pb4SsMMJv1$j|Tyo3GvKQAm;
zuzB7l;;Kit!o{~PY)KjV=0vc+--*-Ff8AV}ow8(E=26C}oTZyd!Ivu0HL=N15A0e?
z$pR&8iN^qOw(ISox{cxDHMnEzkAx2uefR87;w|#RwwIf>?_kVKKb(0Rz5OF?8k)6u
zlzTiWC!6D`#XOO6?pj~Tfa`oLf6l}59KStjt?xf{+-js0oj>)}uFNI(hQsQLe%*b!
zFS|iF_Cs5NX7jLk?}9z)nVESpyI^3kuv2SbRA>Zt61G(GX@;3=LoII<ipVD#99Ym<
zP-~3&BU;WzO1jkRi?;mM!?SohR?PWhba(N?^9MeCSylLRm(v<s*{<2!Rwlmb+VAhZ
z<zPeM!-wg??e<a7ekfh5QT8S%1jGX0+QGSt^;n%G<2@)V(O-DWaxkJKh!4U;kqzwM
zXyg+HLt-$LKGYx2UY$`q|JJRm`_8`5Ijq0?s1Qw^eIg>~=G#otNchdz`CI)Q9Z5$z
zyS(IkUer4Lax$bcfMWN9yaN<G20(XLj0Fd|y?^MwRrKy|fWUpO>vo?OTJQ4gfjQj^
zI3FLbC{IiDPjx@@;aK3SHCKOexs>r^kOO-wp=Z^_c~v)ORyjTJQ+q6#=REp$Re5P!
zc;1T@yjHu`*<DuygN~*Y4;=C?SnFeG042wGdc3Fphj7#GnAYHqoo8RA-aq4_0qED7
zXaR!LbADH4EuT%Bkww}WzkToEc=}S(hgjP;+ZQe2ad-U8&CSeo&P~0U58eG$i#l58
zo^GGD_eRMT^vdeMHM<DD9*wI3MQ1n4?S>Jbl91?q1t%&)R&U3}bnFZah$8r|+2wBA
zqGg*Sb~!se?{WQlX~?@(B~D$;Nv``g?Z~EO?J#b-GJmJ=gRq!}?35i`w5jyz%WXe|
zG*zyCU+sSUc!6wIz}DG5I*4Wx3Pt!Xc39TzJ-gp-wcTah`LZeHt|o-AE5NYJ4t4PO
z$Lg!H(e}}-5s!t}%W~$|#nenI51$`6f5n@98(u9)Y%blrqw;2+M_*^zjjUPpv^OLl
z7hc$Mc~z;X{7iLp)S94K0fN)xaoGyNqs3lHa$@C*7Xn_$oU#49jWMdWjk|ny5e%_1
z9_dxQtW(&0<=C#Te2!+EJm8b{eansK87_a|(hnc@U)EApaaX!@rAuGe&cos9q_)0j
z(L<(u&92k!F9!3CyV^De&N5aF7gj{sdPvY@X-`PpseQ${E2>YYjF=nObZlJ{)Q>L@
zc<B+8z<%3csPA~+|82X^g@*F#(nObF{+Po5lxty5?1H4F^K)VMm#%HUiPa>oyb~<^
zG~Oww4zX8t%%b?T#L3ZPT<03!AekV}ZT?Hf+=#9lW%`FfwnVHT?3JUG;LGn(%-gqW
z{wb~fhL);KPw4w+&5ya2=3@Or?|(P%n3dt^pH9D2IS=pJW+=`bEIwB3l$)QO8=jL~
za9fj{tj<PQ;?byv#_h8M-B-8xG+6!^0s>;Qk;4LN@uRvM`GW>UL}>5{zoJL|e#Z*0
z`QaS{1By0{6gfIN{_whVzcu4LxA9V55)ybw-gWlI^4w!N1)j2scK2BgDwrHfAFvNP
zb8@$|V?XZ%vaGb~Sk6RI&H%G1YbYpcYxKp+K<ZZwR~L`?FZruGa8oG*_poBc9qO$n
zgI%&860vx6R(E+C$9?YI;D;^R+`;Stk{!7`2&d113F~dgO;=SO;qvM{UuMdUA~wD1
zhI1~{zTu3<Y0JfsdCh;#l>U^|u}++^av)z2ac9Rq->tHpf3&tm9{tHHCpxp@X5H;K
z%Wk-alrDR?F>XZ(H+NaV@#6SoZmWHZ%C<#=X6mF@VmdZfIaOb2+c-=0a##DTHv1Nz
zs9k}!_AL#*9^K)NUCO1BpV$A<eyKWTU}2MMWdHJg#DL%Ba`uD0&Hq?ULWVchJtI52
zH*dUDa^+?5cfrN|9>(ZTyIzvj)xH#a0fJ9u*QCxFOuf?maBCI6Jh+3hHB?RIF8{}B
zAYpmLQAPUj&}h&5j3vL*Z`52VN{_nZ^z!*XR{JFZuj`WUq!zZ{&Op;E-F7!W9~#$|
z?j6n<8YwumT*{8;wg%a&8WN&p4yj&q+%!L&ZjCBrm!{kZ=O#$AHSqslY>e`su#!2y
z#Jq`|c^7e7<oEbg(RB9kRp;EHicN{$PCgI%(R+q7j%1$b=)1ci`5y1O3MvcUz9eo(
z16p<wx7+m-jLfj&cdlaIFZO#&i@nQyg4zf%fi1B}9T;)<>*>eV=T_$2BCV2>Hlm$!
zhb8Sh1GL=5)MRN*Bx(<;cm3kPe*8|ROzS6C$DSEA`u4d>Y_dAlU!Se()(|n%eR~yJ
zFV@BRy><glgUjABF`hH9a`Yh{^FRFI(i>)CY3025&6qr+32;CfN4eVI6?<l5pa;yq
z8KcSoIc37qH0392hg)9gtdT|So0|W&Z<@~N|1RT%e_`IfQ@5z&%QJ#M-Kx2nGoF(4
z6#v_j6(!khnB1UHU|#b&;yRBHRt()avtT$pb`R0^-KobnJ#z=Y%Gt3cmsI3%qS3)+
zkH$aQdSR0xJ0<)$sLy-CU`M`t<f-*DR?ZSkbl688IdVCzsBOnzhl4!4qW|=L<GN9p
zHZR9^?qUx|$FGZr5I2$;Ci;9_b<z~nHotQC4sxwmC0pH_bJqKA?=oqPWYxYM@7hw2
z&wou%_Dg@136LjOdVOvo3^z6s3OO(y$hdaKu>Ip-+;?9Kab2z7adk^;_y$Ii`{C6O
z1`>B)ewKIa-P@3&eJ6J8GTl!(Tj0z`m7s~HoW5hx6ZWT0b$)%~ysys&R7R|szdAOp
zs}Psazv!>4eo<>y9{1l|b9jBbN7rI{9kg8eh$|U?p#9XdVOQ?q`Ma*X*)j0A@oxUh
zjZO!HJ_NmFXio3j=5~97n`hR<GXKo{qwoCuqDULNmM52ngY>N#Uq_=mvYk=1CWo>v
zckt){vJi6-+$g+Py>^z26jFPw_F~Ust<TnAN<@SQX+H~JEvzsaQ}Y+S{gd42G@H30
z>*u%Vs-LpH+I4=Z`?RR6U!<*<Z`IVFaD7{)UT{*Fn;e`TzTIx8;1QEpi%{D|F!X8e
z%B>HH*ql3qqaQ^bO^(CVUxbHF2L0FR@Qn?v@1)^3Z~wX6Zy7A8f6^sRM)7TFwn;V0
z;V1?UZE|S;?}9ccJP5ukZ}|A48CP&^r7+0*RppMq-(7XzPfBU33V2k%jIt!_(6rh#
z_n^#37`f2RlvM25M_`{kw&3XYc`Hz3*eVd4yqYvX72Y%G2~(KQDEC(6B=<VLS_>_-
zAiZ|Mj-zh2dLwx*^%7Rskx|jGF-i{P9Eqe8rzn$jeR-Z$JU=Q`7!sqUa^AiF;mEsZ
zu{77MCvHu=IlJQP80{sDkaucot1IHlISPeXxY259wiG6L;Mx|vy!U~B_mg|Zj_y)#
zzccgh-uu&O!!(b<bV<s^ygz=v#qe`^o}RJ&(`JKq%FFG5|AHdKBdQ3`X$T3KtYrO|
zWIC@z+V1%V8%#iOpohI6(SEyL6N-N*G=RP$BG79vpgi7-o-x#7&@)OyDUU`Ft|P~-
z3_<P{ElF)0v&f(n0Xt6wni_NaR5TwZz_Ii`<ERWuR)oiT-^<a&S4aM_qS)|6J^ko2
z#?29<9#N?$suQ9FnXzfFUAsdN$^_#Er~oi(%`AZ&ITt<S1#51fb*0W1SpDlsO)>3C
zXtuO<X}#h>?=@ix6#hC?vQ817q}ujdEDeW;(Pak0;<GY7jGxd&8Vr-J*MuPWtWIt1
zlN1>Y1k}+<{F<N}Fp3z&fEuN046x4<5<t(jMxrEX{tSeAUW^^-QO08HI|y+aBR|@Z
zpur~<uB*Y3C!#>BkMTD#N(R~>xV4ypaqzX4^M1R&2Ikkk2i?zUF{VgW!CH;NxR%bS
z5nP@*0|xd!VY)R--M-x#h4u*YXe<~j_2t>=Kn*N>jlS=9h@rzgE(0uBBYdWw)O<!c
zZ#pb?5A@vV&WPvb!MT}cr16ui-%{d4dX&1!>*I|-oj|A&0@;}LZwz9&Qr+L%|FK`x
z+us8}(~b+)KsSL4tghb{))25d>sQ?%+UM45+>{IGqd0{5fyRKAAO6wQb^M`^dgF#t
zmO!jZ;(F<yM9%<f>D8xtjB)Y3LzD$*F{X{<_PnM%fXB3Y9h%x}CNTw$TEe2)r~+dF
zswYg^=U_9Toy(KzrlpFk`AW#s%N04u?V#)Xf6}g{s4>=D>d6=w2}R6T1CY^_5<R1Q
z2py8@D`S&KxX_vh!95&p>bjaj6?5GP63rxM&zvCunBx!@!bOj6jT$vcq!`Q}pSrFN
z*%mrC>wVt{0#bnZI6>NLw*j*&Q;~_`+-8)r5S4UH=Zo-~q6ysns7wa1eriVfGY6)b
zNfxrX<ZlM&MwsaQ7UobaR0ARsrTv(>G)y%4CO7nDr)sW9o0{d?-D(#^M)^`QwFBUb
z_X^k5mXE3c$WutBiWd>1#(h~-vQ$t%2D{=K3Myp%N>l0nbtBvah2)hJlZgY!V9#qe
zw>C3^k2rr1yq0C`c?1dmEfpq2gTf#pPy$76aqVO*v{7zM1ii|n5k!K)pa@SRZR$o)
zTf(PysTqXz?Lnu3_1U2?Xj7sIwfwofkNUdS_!*DRn#47r?*ULVG*7$ysH48l^%ZI(
z0A<3BAeS#C=+u~}dGfW%15Ao7GxS?-rIb^WqX9rauiIfVyog3)T?w1t4|@%x-)cY(
z*juKA5km&%<l0#VlE(xsua-Z30njcnM;L-&BourN7v4bNcDRK-3kJLn05a>Cq`;=Z
zxm%19=SzII28SqtMU~&4@%L-D!ZwgIv_Ee_@hB3JMXz1=2P$Lr1p69?LX|Q2d}#vg
zPz0GoH~RLBs~skj!}PU$x3<oKk!;v<JYYkVAK~k?wa5XJ<N!2J<a&D8m_8+UMi)#j
zh3WI47a^aP0*GIUAy(Qc>z7CtUz@B(8Vvx1tRmxEcD_MQ+%tDn1G~ZPVAlCg+u6Iu
zbsQcDKQY9;bn_%IG4&WIoO;W6@qo9Q^xAD|d5;ot!{1}2>cvr@XRBs2kZ<)y*108V
z3<3$N6!oIfSY$B}jATOTn4(`K?ea?7mRR1WrSltw??>f}h%6gcy%<Y?fsKn^yHSV6
zv{38~gcfk3NaFy^k?3{NGgJ&v)=(yVQOaVWoo149aPC7D=#z-FUz9TW^_q#gvLrb6
zqE8dF{Y`M@<r>X6_T||x4P*qgt{yl<!9-(?pkCy7J(L4eBq`spD@_s`H?+$ImOOPx
z)FbOv$D+ZQkx5I4l<TyXz6gxk)O*26M^W(@hjiF`x|WGhQbhwJB|RFTcmz*gKR1!r
zYyLiWfL_fcKk@!jK360Iu&IH@xK>&@Asdw{rApbb;yr9g7t`jUkjE2%O(YZZQMrs4
zK?Q+w##%sSQ3a!Lh7p8Z|M9EewRK!K*?`jufSHJ}$Fm2s&g*NrOuWtc@@J04P)Oqd
zsux0^pZ@qAuG?!vc6KSV62P<Pz@+QJ+(4iwvp}4QJ+CL?S$Q>F9N_C0ju_PDK!sY`
zDX0j#9$Xm<3cdG7<144Zy_4%PCK(?A>psCEd$?5`{0u8hHXaDP9ty(NtgKGu8Sdu;
zFX&fl`pVXm(CSgsj4mzB2RsSjwmd+5>g$=)YBOMUI7o%li5%c}qNhps2Iuywqm)Ra
z9E2wki9-5<j4uJ*saQQd6a`mN`tR{HhsHH{fB8wt4263nL)F+9(EI~wr2Vo7@F*~C
z1~fgsahNRYcZ+!XNv$EDcn-IvidjZR;%vu3eLIvg^&w1LbTj?8gG^$DK3GfRKL@)!
z@=;%%V0vJ=TV@=KbDGqQbgyW2E6gY_7xt3z=M2+;UBY6T6(?x~Y1%gM0;T&y-69$Q
zXUwc)bdO1rW=;c<E#c^e4i61XaQq@)9&ganxS><aVKSPH)z;9^a?nFyRf9{)OfaOR
z&lm(DLHzmV9_6%Lkjk^C5lrU&hpxdfA|;N-f<r=(Y`qS=R0R`c+AXFKplPZ!6MPL<
zqG3E3SbQRfy_Sx(Fan=MZ|T%QuTG_n)gl#+rA&L^oSTW&z7^g}d)4}YN=?53q`K)%
zl%2rdS{n(ZuG4g^5mxaZt0EYpX%9xmY^M=e>^l{tm*%?7CA}Vj>0=0)GiD}RKRN2w
z*r$w*QpGbt1|CV|Nr3j*uYs%FsKxl(VA=UFWL)VmSw&0KnmZX@8i1RLV0sCBCax7M
zlB}@@#y!-bQ5mHS1dlv`<H=fj_RCNDnmQYN=D!U_y|Ko`9aQVU_`)c}x#tBcQ(4$O
zO0bq|?xq<4k4nO@b(i;(BsQ7^(tFE|?Y97c4wS8{q>)0YUCJ|B2u;W~arGb<C{@fF
zj5XpYsr94r{M2X8Wjr3jQtJ&A6)0j-6Xj~Pi9k28*Mm^zD7cJKskG^TggxN-Ku?a8
z<>+C#%_!45A?df*ZgT~qUac35-;IF@sB*a;X3N)qKJ@?I0XV8OJpQ$b-`M<_M$p0t
z<zGJPMs4mC>Qi|lBmLP51oAtK%2Oin*wE)l9Muyp_S*tGj4r5$!R{Xi(~wml4RsKo
zn)S#b(6+&ZSZi?IOi&NCjn!&tdDS$M<S4fi#*hIW38;(IgZq;iUV4l~#+SF4kg3at
z5iSvYF||HPpP+^fo`?sP(9b660m~%^PE(tT2IdV;Evp_?g3`X7Pyp<94Ecs#RjFkJ
z$3mO}T2CO&Dgaj{t0qQiVd;4iovHqMFm_Sc*dx<I5l^Myf}AcR526U1B%&8$G~(qE
z1VFk9m@)=ph71z&lPXGX!66!u2#536LsJD~YOPt~R^bq3G$C>1piZ=F`BAsOH6R9h
z*o5#}e1gdCg7Q(lZ~Z8Uw1!e-pf}ctBCefq6Z&!85{^CSoXr?Q)utH;CoKcu{a78q
z+}$0ohmPw_Z*+tuaC2dhJ(-o6xUP=tiarDtR+@d~RIt%BQ{Ag>7B0a$@nF;_84qHh
z-^Qps1RnONJ%$;jU<=)(+8S~O@2MP|H9$z2q2M!1;NQ!&U?bbaKL%UMmLSjRIeT?h
z8?U^+sb=FRD31eQjK^}NipU<gr`vv`F@I9dy+3NY50{lDd&MN%-1Ght+uuP=Y?Nxv
zuVYXZj3cEnkf1Jy$#Q*7FdyG*T`31lrAbob{U!Pocvfnr8<b1WLolEa)Ju%!cx|HP
zeD4QASCtXMV~i-~y8`(C|Do=^gPKnNcwc|Ju4{Jz8(l(20U>~(5LO6;KoDu75MZTB
zReA{_yEX_y0-**ltd!6}=^d34N-)xeP?j3$Lg<0)efXWZckY}ybN)Mb<}wa5B$9;h
z_xV2Gr@TL}*ARGx?J$v#N5iTN&>sRK!;0UJV9i&n5oqpO5U7;f_MwGqf&)*SQLwu}
zT;aip4Qto`INDmdQ04m1@#jD|4$IJ8{vWqKwlMAYhNgvZT`<Pd2BU!S9TOHG=&4$-
zT{+xGyzjlzJ~=BCr`c-Kbdk1Qz)|OjH?JMoEgSy*e;^?C|Mg4$TKM5G%;ZPZfmh^z
z9ikl+jL#Y&ywrm&gyo~rSWK`ivI2xnc+i%K#6M1EVgN)7+azKSs45_-#`tkr-#JoT
z(wWyPz|o53|0s<7yz(7yotsfQ<@@bPn61_CwQ+}r19s<~u!M|6^>RxQ>iDFME#Dv-
z-lOR!TyG-eM_<=8Y5d{pfR#!qP_(UFTqv!qO^KRzkdDff(%6`W)v_Nq-ucKA_uwJY
zPS4}MPth2D;FC3pIez)7U#Y#`te=lQ>vu--U#Ap*TT>O*{l4SXzlpgOTC9g2izE}a
z%v%E~;$Z!CeT&_<P$Ozk?c1DfSJ@xvHa%|6qg{?-Yoh$Lc0s2?0Dx-Z{am3EV_#3_
zQ69~i@j+GnTw>ApxLfq0^6%pBwj>$viLq51tA}!H#c%wr4^!P1vLbV<E$Hj5K@>+;
z*Hrt5O^jXXkX6t&wWK~CYK3!!LPrHQR(>hb&DT)X^RK=}dmnD-Em7rSdzw0n@Y&1|
zm%Lsaa6>Jr)9GG5oS8my>z8glj{2yJB(p*<z&ev$$)_~2*>9<&X`(br$tTZjJ7^H(
zX3~6%Ftq{{%d~)_U|L?I1#UymgPM=u<_?;|5qv^>CT@u=(gOVI95G4UEa%SoPl+Sl
zq2^H`^8)4`OK54KMT32Kma=nFH+ffspS`<Y>AsXs%(0E6*843kED%y%N*p7j9OS@M
zB*ubsFzIqHUQZOUFJDW}oP+x2mxLBXrp&ui95icJP};?2<j1fUk<WiL<n=IX`TZWe
zl8I^Uh#xa02W7g=f09?cy?9{}t)kkuY$BdzQ0YF-&%Xn?o`$vM(F|u8!^-Ni#w}<(
zC=SHM`F3S@;&99>J#$1!D%MpfXt6+sN}k4*Gc4#nnsh|@mjyrM59h(R?$?v>62o^d
zk7LFM%uY$c_v44QwD_!+$WnDB5y}#zro6EABA2$cN4K<HQ@Ev6DFT^Iy<?2tW|LVI
z*l+u}2A|W+z2)hat*y3;g(Y%;t%%wLFe1uT=G71B(I)qOu7lKd@8BaT$e39|_IuAR
z!ltiT{q3Z5U3X_mM&`ZzUwU=!$9;1it=6au^r{QRVwY@vc!Z46y%v^AqoKLvd29p5
z#htwLL1AoPBnuyyU*Kt7-w%b-%~gg`{dC&48z@@w2lWX)7m!vjE1@RE%GZPJeNhkF
z2hZ3(jyH8Imy}{tpF<V`x_v&+gKB5!iXp10u(to{{qh+16zcWDLP#(<bul$Kl0_L;
z;u(c$`T<Ger>#`%SlS?JP*3O~!8OW_TH+GCy@jG<!=Enyn4evs7ZuEPUKmXXMw$IR
zlwEzg;Oy&nWrWfXQog_4%a_RdPC9o!j_-|i%;rRl{)8m>{eeN>`U&)K=p5A6mIL?a
z(lL8@Hi2(bz4TtG+-8rJLz@4iHKmKz)k=Dr2;2%ndwipisr*e7J`mi-Uo?<&$<qM}
zg)SR8krXESXW?+WPr)ax_Xhlu<7hl0Ri%@Y)goM+w);%WzsHI_abtOh-H~C9qV~gd
zPy_@&fK~SaxC($2f$5$JbXG<G(v7#Bi^WSjRlMZ>W!1sT6K8KF5Gu2<KA8bwj7+kX
zFhk0G1E1wX{SBp7ZuA5?8#xWpYS|ro9qTs4_u6vA;o-Uk!?ycoHq3l^E?_;8e=GmG
zx6|;bPK;O9BGao9V84*5@R;1rX98M!{kz|K`ejub&HvN2_m<bZvg(|UXO@w(9j^*G
zRY)N$-OWR%Hb)<ULI^R+TwGbrq7mY#!Yx5l|D-IjzEeDu`0#<3Rt`h95bJ+#{V`Qx
z6qU%6w&0_nwqj_I@S5U}N2)QU299c(Kvs3{@F1qJS6-4xeYj@pkkCahHN%IT-MU`v
zlc?^p;UN&bm-@A9C|9Dk&$J`qDF1W6Gsh}8h3E>(qq=8UBK8-8d!7UI+}<7m+W`Ob
zJ4JeOd5`sDzkV;~_wjn)LHk^SRq~w^Wg3WI<NT}w((ZFI{>~W6*Wa*#7~+x86=mgl
zgssXtDDRk9i%5?)DiAPe)WjG&_o7m^u>tZM4b|vm5e<#wgDTaM&_3;fPCTz2QZr%m
zt-@1{9`{+xkOMu27uE8%W1)ahRK#4#mo7iK4k^w3F^%o>Mp9JROa2Wv#ar@2mtV?!
z^^(+*$k)$FNBQxItS@`Y4wr<kc<99%l7b)nW^!#wyu+tzRza>aBDzxuuV|Fw?#3cz
z2JnnZjhb1wIEYsS5}mMN2YcH)z!7xCyZp)dGe4jH(D}nKzHQ@%@$*{3<|*@Y%2`(V
zib$iE2N45!rI~jlJWgfQ@ujcusoFlnv^@01+qJ3!CyN$+<D@pK1h=Hht&Hl5v+^=O
zh)smm)>P!zD#mR{7AyFECv_(a^-qLz4|iEin)k$N*e<2-hP=qvs2kJ8*2DrVlQinY
zxT7q@&*xOu=MeDvC5p6c^)R+t!vAd6`-rs1k$5kyOR15mvm$lXXWD=CFa`#v2VYrQ
zL^*ZI;6qU_a~4b`__~bB-eM;V)zTo6>X|0uh9Nw0CsXF7eqFC=Nb9})Zl=!laiIUz
zu%~yZPUW`;)tIIC%(D5rbVA}+?v@%Uw7S_wDC?xF<c)fH>8Y%59USb@{Z>S(O+WNh
z*1s@zJpG~MMS>*)m74#7`mkQ@v+nh~H3Lnx?p}jUFRVMAUMKXo#yc_Z3K+gC+rL1%
z*X3U?@S!}w<Qu}<)*^Q(%Z4b7roE$u4CeJ1VZVE|2W;i-70(ri|MDB?++D)gSe$x9
z9cJfh(6b`Gx_6mNv$kp*r{l$8SiEM`f3AN-OwH?UcYhkq?osOJTiI_`#BB(^l=%+I
zc>I3W-P%5(2enzh{NWaw7~7ri-Be{{Xox1v_9gP{oicl~r7#~OHVjIWhd+E>5)`)*
zvsD7Ns%jCt)es_$0uz1Gyc}5k=ckV=k3vTJ*%DUB_dzm&Xo6w6TMjZn_N9k<_v@Mb
zN7qZe7QA^AvPg*n$V(SIDvj#TXlcgVYOq9&e<@p<DsC}Lh$P+|X#XZDL*G9a(4HW{
zF{3c9WxC-rf<|@@C$I;H+!<S+7460WE&G9GcggJ3B6k=2xUO#dC^th&36aFGF+IDt
znUH^hEOriOBo6WEGc$9$T@Md~XrJUylVes0gfUP>6-+_3>oRMLp309(zBMg<$^51D
zlyKjobEg^dNmO^YO55Y?8}VNkJNlkWmow+U#AC8>S&%2ijq+>WndkG)I%&z*dH1gm
z`FTR*Zaby9$C6h0eL`&?myM81sm@>b;~`MiMNp-|oCBowt#fY708Kf0#Ng=S_qof*
zZ~fDb7fP7!%v*VAol_A{$TzZ=Ac69GGh(u;dZpiN$|AAOk-^#!j8QPUpV^tRxA_P!
zyr6iU^B~3RUJcIeB$T6b#I<aOO~&>{i=KHfUG5p7#*n1@UCeqrc1dM@R-QTzonKP@
zu99J>8R(d%*cCw1?Hh|l3`q=lm`EoCZX;Iz+V%L7^hb&w;caJ29uI1HIq5ny6e?{b
zd(vI^%-v1bZj!;H>#2BS{|n^pWCbT74ku)flf~RRTUv84z4$^Pv(+crju=FiH?>qy
z?%~a{x@npkP<1(@(0OF7(_KlEVCa?I@m=<Oel-uuyv-cO1!a}P+t>*lUr$R_HTB}=
z{h2w-Go4S<ozH*k2@KmU`XJjXdGEuy6yE8qeCIHx;e@os8iJaMlu6TbXdm^I@{n*(
zJ14<>oM!E>HVIYsO0o<yK>8DNwq$1Y5i<(md`iZqbS#=NB`&Dy72)n(JZbCRw&>tc
zq0oWXGdv)lZBI&)D7b^l$}<%Zc*N6{lr3<1ap0w3zEsCc8jo;8lgMN}qa)n<@8aw2
zPk$^k^yDAkt@+EF8JHMQit<$^8=6%~Zrb_xRHf_Ky;6Fq?rG?56gDcpWW2yDdcx4<
zbA*x>KKZg{LNY9j#4+%1Y=W1q%#EtoNgs$$Zmnx4Pkx#^^?hZ1g$9r$n5G{PnT&Rr
zaAVVv29NLClc8Ca#7h-BAtTPUzeE+4rUb9oDhu6CX`I+mcp3ObCpLj*%2T;dlt@;F
zAo<y|gkGK>ho9%$)J6v>stAYT`#oYi0T%qWG!df(2iZp9GmDQT?wLq27yR$T!B@OQ
zTx?uI082gX&jk*{qAuIkx?xb&>ACjcLz|+ST3+UUVt5JN#nwhnS$Sz+sKcjC(-N#*
z;~3+x&UjI`*Q4)W&BU&$o;b4y(+&SDFG)r1(UmSz|9-0zdVlWbCjPWeC(X`hkT<DP
zz~*jh4Ep(S?+xGbt7GSL&HLc1w<9mHvpy30Lq0nvd>bs5=PNCyT}CdDd6RN(?2!c$
zyURc<VaEhuURktVY}V$$fVIbovY~~$4-HCN-E3rs@lgwLT~Y!lym!qctoUh8o6>3i
zCT4NF&zFWYs9C3>BUKs{Pic#s@bdGnlUZ@Ue&LMsNiX4x#_%PHZ=3sCd?9A_?+0cg
z=mTy(*B!H6;Jwm|db(el`vkhf)Lxv(?^Ik*9rJ2Ge|B{5MBoJkp=BAOlAEfYJ&np(
z$lH@njZNeYg&WE_NR!C`rg7n@6t%<c%hB+82-SaHlL0AZC*2_=9w2OubN;$s!+0ml
zU6j5Ue=F?7Vti+Xf5<l3#9J#pT)6psi?<3#-#mBtsclj~yc2cHZ)^P1Gfs(~XGXXL
zVbKtsuotkhBW>|QB3S~qCaw_2j><~OaILPc!bN^e3OF;Y6<k5vIHj!h%hd5j%SXan
zRb#JO&G-YHdzZc8`2_i73Nk*AF#20xY)KqGa{7+n_Q5vi)p$|rf1dyTxo9Koa`w7s
z1m7E0YtY}QbmOi!)^2f<3VeN{W)kfLCdmQmY+OHF^*q1`NlJ@L@Q+YPOI`3hW23rN
z<q1!-`UCJF-Q%UsXxaPUIMDJ~&d(KDQNCqSzD|G<4q5@XGjJM50|vv7{n#sgg~CSf
z8}mU#?2L(AX|hKEuiSpAR*BHzy&EcQ6IIQDKxCWdA{wQiZd&oOQA`T`LNiOpG}=R7
zEbj5u(d+LQYhDK`QZZ}!P-C)zO@VBgJ=XA1-E-J;ymTK_;>uGQHnBuIxJvbP!Mz0c
zRmH-AVgWftd-V`O;|#q@kkK!c5`yfL*rUJdJ4!X-cD|D&d>7Q7tAou_f_6kbQJn+9
zARfQB&UxG3UNO`!SwebM?~HyeL(yrbc2#<0@yzdQHH><>?1fw$kpwUl(~BW~bWHfF
zR<pfk2>YjROv34O=)Fu-LT~1heDH?gEnx){z6<YU3Zk0x4M|ClC9+d5+e*zElM+Qq
zb@?EXFxd`Bu0O+~LSfpjZ0_&kS<UO5I~>Pqm~AfQ^LS0mdKi8_dt&h7&Nt5yXKcSb
z;e~!lTvnEduB>*!%6v9?w{~M$G*HdufjpmHTD+dBC7osrrSHLa@F^=W+Zy(66Dk^H
zOfn0H>K3W$_{9>+7@M{{h%&*=w}r~{*OWokQcdnw7?*lgBtyTys~jcq^%z27NM#td
zfzBY4*}{~1sP9deu=(+kjA;5JkZc^^+iAtXwCo~Gj30gAY1IqN6d!M$0Fo)zwx6b=
zt(l3@ueFMbJ{ax848detcNa_;B9+d8&Vf42gEo6n|8<(?-p)+@>Y!u#ne{b*K5jQ}
zIs0wOxH9*Kvl^jkYiZ8xen<@#ZRnq$a{r-JM~{?8q(N9BBOCq%GF$#^c#`>rJdn#f
zsM40Q)$f;z84fKeOqNS+;8;P0Fu?$Mh44EQqn2J6Z!wW!e$h`!$v9=98V<)$Q=%HV
z_9BneL?R#Rc^m(+oV;=8`bn3ir~r79Sp8+gl*jVqO|uHVvxV7+DOHdodFkj9z_M+t
zyx6W)7_cq`-x|JTK4_&?%PPReF^T>gT`k`~S0@I#7eZFLZWpk+qVfvdBJmW6gDh2g
zWmiGi624RNzU||J(3XDi44H>NA+bdIt{#KdZv};F=dW0~#r?LYg0nJ?9=-nWG55vD
zLaxaQCwKe@*iC8{eNeh2#XZaX_n0ThXZ?FD5sDCvTE=J9a1u?b{g=rjXSauzwrOIM
zoFbYjJtmtMukcR6AK8yp$Zw4!Sv;S0W)BxlP*-4?vRNWyZA$L41i4IUCIDG;Z)|U~
zXZN4<77e?U+NxT``QD~34{DdAX<Av7bw-AjCPUL_<mrNOYr!%`Dn`=Q(ZVI))W$3Z
zdAGRs#1cGn9{x(R%K~Zgmxm7=J>VGske0dqK;ECm7Y-Q@Kvd&%sbRH*ei#ntdOj|+
zK6u2!QnR1P@ki9B6sBxr+z8^`snoD}QA&oxedwhT=Mi6_%0T$=!-z{PTdCLt0gKAQ
zzbkHheF2s|Lg9V9FXms&lT4)Ek?qb}x3o+Tn8S?&`i+&W!h2bCKNvs<SfXhnW?3dJ
zS!&*MbFJg4g{e<63knt@iBB^CY!Btv>Z7EjxqVO`9V+#+hjGI6xR>8~qt%f2zp=ZW
z(UD_j1__n2sD|DfNuxs*y!$6e$SYzGF2ffuKb>}lqYGpfGiw(>Upp;bfe5Ckr%VUq
zSrgG??*2<d${#|iQN#VVHRO~6@onlIMyQL%3RH=&wH3kBvy)%2+NfC>AoQlh``v7^
zvhset;noK0A6Uz*-V98)sPD;EXktpnuBu&bH}Sfa<1VHtKk#-@k9EKTao5x(ytCdL
zN9z1oF|W1dd8g=;2_q`$w7k3(dIpurXQaGzWgzXI<oO#3ZiWwD-RKyaDtsT|nCEUz
z^sQI-xo8DdwX5g3WNbMoI<W%f;h}RhGa-`03;Ui<)|uPrL0csg<1{SV{33jtGG-FO
zst94zIhm}F&pLOP4;qyr>8Cje@mmkTOzfg5)ABFByJ>Gs1H76#<BG%RbVdfu1!ilc
zJYifUOKqx8s#~nRgm`Oy`CvDaz~&RqUs+viS?qsb;)w2v%1@OaPjwhzPg14Ul0#Ca
z<p@%)Nve!z5?`B%@2+eIQi2~|bVLspyGI%MWT!=^jz8<mTmZQ(lO=;%_Us+hS%7A^
z2j745;LuhWZ9}+SgQv(%LxPXE+6`NEuaAE#6v)W7tzS~m(;O>s&D4mT*C=q1ZiapN
z+9~fQj#L}d92SGw{*rooa03QM^m6l-w{bka!d})5+yU$bL8j;kkqg`}f&L57(eSjy
zWy<YEPR$Yp*8WiVY%gnzkqn5o7(eFZYKHW)Q~$-r6Lk}kjjnd|A6A~L#Z+#bysf^r
zx<(t^25c8t@70Qmr&ATyUWEaXW@vnh%^*`&!a(}GL>c*)d%nojD755=KBxjo8VHOH
z>w!DQIu{!(s$%YR6_HI1+n)*j<X;m7(Jj(=tiU@gY}?S6!?$s9<D68yZk(Bsvhu8d
ztDe%<`j104lQs48uu`J+G{?1FPS)wpWtvP!$)crfL>*b_UC5kjr-xfRJt48EGq2nF
z+;trh>snHG3I!yfXV&+)cKv@m>#1~e(!6&O9U*$Xu>RyZZCB5O+L%sn6cyhlC2cB}
zG=g%p*DGHf@)po&@XlR?M+ar)IOf!K#(!V`67GFVyd=#Kf5tjWTiu}cH3#lGShe&+
zV&}$Y#Y-u86=`Zi4KJ-{3RVF}M4t}sV8YkL+XgOs1++XgLfMpsH=l3E`zjdTd0V0&
zD_%Dzo25`Za$@=Q<IhW>Paa?Qnf-g<9`Ah0im27vns&?TF43<j^-@j>HO2Ok<F~#n
zo1t>S+M?md&Qd!2CKY`HOA1EnC4!OA1zyK%^{7754<Yv4D26L)AO3mzMN-!%wQfX^
zeXns^@Yq$=*V$|%sB4o?T3CN49NYD(R_v6ifCbg;RiI3pkcpM{1O?W?niM{04m&{I
z%C4w88_Y>g&GI|Fqu3Pg*#D@K?BqQx)_GoyWJdNm6ByQlqz4)i<i9)>ll@-9d}DQO
zXJ7Dk*43ftWS$nUFN_n7ucVya&kdu!^6!7e-`JA#ixVtrw@B%;?3IKWS5%5F>h)9e
z@Wk#JrMdMCFgzOb+V-}+aYqIT%y~9E$r10Tw1AVPZHnij58zbH9fXH0KR%aNVeIDU
zCPBWEk(6y`zmUJY&YlHq@eE~%;uWYBA5@s@%J}zK2&(~L37>hM|DC*Qiq6dIQS$&v
zRyK0O)Y!txT=3mLvn-ph-$h^K_Rd3Or!j0L9zF&S*dkk_95#L9naT3{esC%%F}2fh
zaie-gRHeHa{^cLHL|)@|TituP1qo*728;I$9u=SQoXJ1A5L<!?L4eE{V@A8dyXjo+
z=wkvsk+iURwfEkT+~W>g>>KtUHC<@w_+GCCUSu+FnPR?SEK)Ilu|v;xhM1Dkq`tDY
zGWP?o{3I(A#6uisMt}Fmz#MS08M50UJ{`w<wQ-suTr}<`UoW82G&YN1#||3|j)>IF
z-C%PQ0KoUy2lYYQJAjZM!T50yyzTc5fXbPZtv*1*`f6qS7zMu*wBt8{o%jM%+WHd@
zxZk)8MG6-U2>iPTT<qcc*zrvaCM*On3$>`>&6HyRL0*3V)WH8WS7fxTuz+_1%u>Ow
ziYs&1WR(I~_$*=sZE2eWyC%%H1;SAZ<24-C>akR5qISzJ);_52cL<(~!<AQuu=oD?
z8R$J`IQ0P457_A@PbpysfHW@5{dk1WKf<X4dj#0|<^rt)8>AI4lmg|+_YBJTE-`w0
zjp4?||H%iW1WN##$==Ue2SxynE0M-Jf?7{9s%O~i2gkJR!OvU~+k=l<Xh0$avwr~C
z2I#ydptTjkmhb(g#kWO+`K+#Ohl5?Y<RhfEXbXD-pk$P2&!jVhcNj`@LSRR?beX*a
zDBwr#j-!J)sU3T~7l@|p?fJAL7|mV3qk5Uu8Obe66r4W>C`kxWO?>Qi@aL`V&1BFf
zg?{`52C+bFMEk$N=-||F(d%0~;1m!kVa?~ko7YbRcKxI%@yJsVMB76!hL3C^T$r6L
zh>+Q4s<USS58ZFe2Q)>l&@lp$I7i2T1P~x}^Krlg3s<JG{}H$wB7hKpG5isj0dPbn
z6GRrA{`5DQ+zmY7VgO4KHUZk>+{ytA;M!+)aW?@=f9inSgr%VUCV_7T+t{?g1y3~t
zAI#pt&aUrIPdGTDHrPGA(D2g}U@AHI^9Y4cTRT#d(Eb$Wn_OI(V==#I>>vIO&QUAZ
z&TnpOXNK?(uv`G?_XH>&1%{K1@l_5;4;+E+MYU-ofXq2}4FLyEDpBIIQ-I2@o!qjX
z5e!b{HhX1bW|4is***qtDeK!H2B97F2k^p7u_qZ6_A)pg+F?N;*ihZFG&A1u>k<83
z6ln8~P6_MBuFu*Y_(;sv48~sA$MYGbBg+)61`Y(EEB~b6fvu&vc{lvX@6s%an`17m
z{rHpbBF1%<d*q7vA8HOMZ0`;Oh7&CWz6s#C7-56@MeN_gqO_gugF9T|KM{!WogLQl
zK9^_f|J^p%wrS;)2UsqHm~y;#^#GTEQ+x^wJNi(7{@(=q9UM<Or>QjT00S;ba43Pt
zRQvqz{wCvFK^vULMsShLW5EXp$u}*o&|g}jxQvz^9~KiB%Z?f=HyqnsJ_phm|NQiX
z@?TChZ1PCuKyhS>22C<EGX6OA9bD&)+{xW$kz?>)P_<lfe7`vicoU|ygO6kiAsmg2
zse>lqn9z=xS>^r#PCZbAG;2@(_zxdY3V`<U-}=?`BcDQZatqj^@89MCpgsTz&#`FU
z;n<AGlc~}NT9fwPNA!OaoT6(pi!xz*!2SC?9$1_YvGtf{U?@U_i&FqO9$+;y3MY<m
z`z@^I^6@#?4?Y!4?h$+*0GGSh|4Z#ZM02^5#Mocqz|GJ=8v&s`@Vcv9&GpGe_HIZJ
zICdH3AXIzd$*Ih6oMJa+Ql#&An<#D^B;h$g<tlO9KM-gqCfRFibO4TP;@Re+|15X{
z;4@QOoIRrAP0(f-+DDcUJ~)8Br2n@oVFKo?-!{;BfH{_l@F75v(=plqRf24AG=Ss*
z$Trvq+9F!%+rX2-hD`wRKuh&H_(RQ(Zw9Wg);EYmH|{miekj0uSU(1AKRcAMaB$85
zyYo-~5a0zr5Z|L+6Bgy?G-u#|SYF5k_JgN)jRg!7>`3;0ZHD6bHR;R{PFT~VV*SQH
zKWA#OK}L(g>aISriD1Fr?bEt&kssjnumQ$N_+w-yPLX`DyT;H0iiT~FW10K`J{o0`
z*unw^lSQ!D#6Fe^1K(r8WqkB`lYgvD1cR<dl*pW{9$`*`>ZIQZ|4&yc`q?#dW{Gmt
z)VQKgo5#QX4DKonVte=+a8c|Wh%?rge|y8-2hEY439_{ICT%mUqxI@lGO^BCD2U9_
zj;4)_hELXyohJQ%y~+Q7;{CR&exc7`&hvBL9)oXBhclWU?>_(i*f9~!KG;n1H{1YY
zm$+nf$yB=<A>=n`B4Rh1MQpMdK8$Lf9pvOh`zJP)Q2qg%#SOTYWxXi<@b4?CA>?M)
z4Cg4^P92wKnFViq)2qF|ZZ+QPl^9{R@(p*%vp=8g_liNvG(!G+Y}5Ro8{Nr~(H)I1
zls5=01>B;Y1TGx(4{JeUebb0TOM1qt)pq!JGi!2r8_H{u&HheK*1n)9IwyRi)g=86
z_3A(h?={<*5R9>4kyWFJCWXjFK)N)F{D~EPChurz;t-h2B%B^&AQ=?u>_7G)5fAOR
z>?kM*SF&ujvJ&A+$ri5cng~nf>OI&{Dx0RbRjZg=)k(d1>)*<8q`0pMVQL8uQd~c!
z_Fi<Lv)4J}09%Jg(#+LL#big}&c3D2)#QQHbH|<c=lI1JBkZp}g8B`HrcQ^}q}I7+
zYBvraQpjvpMq#t7qee|r>Cpb7lsbzlM`&`<SoqEzumoVRmU=$N#6IWy?kDbX^hf3Y
zdGz55Sth1sake9=VzZX*r63DZ1ujO`LXcUn4A_j+ii&;z2e*`h!dqUJO8*|a3~`Hc
zTv>yPbfoOGP}?dk1673zW{9@lD=Ir{S_K6q093L}gPALVMtrj-Q_VHIq)M;*tzWC1
z{bJeN)H>&F(<fpo$T}8mGPL*m$n&^aKSn(<s3oG`TfKLYsJIC5k`o-vOgHfb=lxR(
zN##{=!L?+l!e`~!&fPm<;fkV69&g-YE2Da77&_T3Qj)jpw@i!z!LDziAv1>slY@sf
z2ingM+emISxA?=N6KBRuJ}jSaR^-+;|9i~8aHsyP`po54+;$z#Z}P(1%{;dSo?LRW
zL3C~M%i3fG=jG&=jd6T5K1OndoBs0&OSdn&ZyolWsBRLSDKvc)TZSH_2dz2_YateH
zNu;$&nzXCde)W+t74xu(yd@Fpg<&wXoKp{c@09siBRf&`iSH{A<fmf!=V)Guw_lQ#
zUyqQC$;r1a^*ze{n@WG->0K^Izjgdg+`UJ#C-X{IWm+?rq>aM)5SO8v^g5U6=omxs
z+1H&$W_U?3hlxyK;Wq%FIV4D+=nKJl|D7%-Kd@W^n((dZ!PPeF1mhjJf$O(Ew6$4H
z`=X$Zd+2$!5%~(ig72D@YAbWh?(V5A%<F(&V?&H$zUBGfb1IX1t58PLU#XNbLT@}W
z<%ub8l#L<8{ZrO{MT4yyw-J;v4XfGQgfuGL%_9>Csr#}3!b9dYBh8}+D)}WP02;ZQ
zObQ|YTAEjn_l-ytR;Z$0Bv}h1c{Z#7vBAv@4-bnbz6bxhKe`I-%&hKmxKn>tU;Nat
znkqW9u_^kPf7kNNrw&G(p8_^_3Ze{iLeGZW{?>UXX~D2Zs-#q6!4Jaf-KSzq-tS#0
zAEEDEfra7k&baprdzU=vevR%aw8#K8InK^}pGLahN7MH*V4r8vwPe6AM2HeME~o!G
z(uei$aI)KcZ0Z7Eetm<UuKr!NgldQm_>9as1>eBd!lcZ+>dGZ#dK7wjT1^JoKXy3S
zEXa~n_VoG>kJQlmRA!nd;pZ(S2g@%GKAD>HJ!Ic(1#zO(ohOkq*^GzTCAy0`0*3KJ
z`PGA>qKcrlAfm#NDX&#t$c%Vv6uIacu<Xn`^>@Nkwf>SE+h1)=Bsb$Pj+poGA`SMl
z^S~h@fuUh@1Z_8H_e~3G%j4JlemVW3RdJoS#35BHNRGa%3+H|BpmL4RDJ{<m)LpfT
zhAxYKk7r*wF~#ToO;`5!h;`z-XU$(tj5^ot*u+2j=v$Nf8EYArj)3=@r7$4;`A%pG
zbao-%#kHs3>i4=bNx_k|HPL3)l*sCx3)mtRA0@19`_@mZZ|l>o$Mbq`FhiA{%E&z8
zcYZQgP;+vV=Cgao(pl0F{PKS9)}f?_EPo2)>D+%VM>(3r@fZi>S&tPOjf)UX{x&t%
zQ!IhNT(6RsIvAN=r26B4tOJJajnDFTbKaiedsY+O@-7hGO7n_z|Fdu0K{whUL@DX@
z*j@D*eJ5|VS9!Gpeev_~{as@9-tWWX7%6MUCe<m~R7CT7B(D=}5nq+^_)Q-1_EJev
z-MOm1-`uSSvPlNfdz+zmSGM;MgDXb}yFb3|r<Y!B!uq%=pZ0JwR(V=0Qr9Ixi|v#>
zdCOuT2yLon=yOVy?0?$`={N7(0|KY^N?<Ji=g<l~@kF`(k2^N}sJjuDd1Ll`-|F&=
z7-pHco$(pIn4qRV6NxZJbi|-*H%#*tJrCnKGrPYxsv22+h+6jc^49h^bFHC<Fnbl%
z-{IbtbXw<BWk);&GVXl*rAB8J(jCh(meCEvd+1WanuHjl_6Y~HA+4&HQ_Am)c*WJ^
z<=<Qn8mmfac}uvN+$kH?pZ`OwRo}>|Y+W>>ntU-jB-6B~*kZ`)WHg@1p`aUnWk|j^
z-B9RThAO!Bpa)#(BqyZI#thhknW3!{T$cqO7@G3L@yCfqB$v;fip)?K?TE|WHwn5M
z<E~%O(O6rQ3um5)6Atuq{A9G`9`IMqYfG_Kz5;~xE5%IL?i@xYB^mj+LJc7bY4?nx
z8u6u6<x=SUHz&NdCDJ2Hl}F0h?&jWN-lDqMi2@;P1?+3_FIrqXoD&;nZ!cUcP+FZ%
z6K-)=M%zezh|pD*Z`nR+`vM;|_A=-WkEpYPI?`~iY}tO{m;J0rA7L}}cg4)zosC`Y
z8g8Vce9KD1@BT)*b5~^FnTvT%5ds%?Qf%BlQ1Wl&v-*eB3=MyKnfhDjhL6mAb?j*R
zU9rh#FjHx6uAcnZv~%e^{LG<hH{lNHnq`x5{WtyLfYxN*Pt=QHwPGWvtxhjF@umCy
z*(BKWT7%9JmuiQvPtq_~(<koYJo{|&ik>HGi{&dD{oR*akw9*@QIG*!Q@ycEFOA=w
z(D%63&OsA)5cj{>xWn<#g)deHwXzENSB`I3py{<ahA=FjGBN`m-3ND&3g+alP{!ba
zZ1z6)llxR<Fd<{n&MMknRo?Nv7gfPnva#a9Fi(P(*wfYpgJ%iMvmI^|f7M}6n2P1Z
zms}elp`+LQu0I_(E4df@KH_4IctMQyfS^W6aod##)al`bh|bgY;?hG2H(%^m);$+D
zMd19g?7g#XzkRcK*cJKQTQ0?EMlE6URMokPG#LrpoJ4gmQX1ev6zr~T2h9ML)y(z<
zq_}ixjEm(PqxmWIq@;C8scvsYuuzXxwwFkaOmZ_B$bi9}^d0nCe?I1ISyGS6QFF4l
z^s;a=oa_Do#_29_=c#2@r`^x<hz>+Ldy6(B8r!w26M8lLCIwxmV)?4Or8`0YO&2b@
z^$@3B^T*0p3&XdOy(91Y#$D2`xx<^HLXpnNGY#EOZ|ntLvGan0bt204cCaIJZkfGu
zgd6{}*0kiRstcdJ`<oNgM^bf~E+2O1U7h?!9-k9(O_h+DzM0Ik<vRg+LRN^kb>Ej)
z(1(1DSebL5cXW+I_zL&5p~z*Kbt7e-CFs!wugU?%#Zg1?Ev&i6?6lx)@|hw@l0(=6
zNQt~8N~>zzwklrt>%2LM6X{4bH*-RGjy<a4fC9}5teKfF1om1&@|8@&k{ePYa!m4+
zpupSv{gpbhbL5%U&#^||3%)2zZYZ;QZ*Semjem)K-P<m0mZ_3|JX85vz1+&R@nPMn
zB3zvZ>ph-Fg3r6+URm={X(6`iEkav7=AE5I@=ewVNRwx~zcr2t=dVd`Tz;-=`N8*u
z09>}O=SG`p^LfD)dD?KVSGKHYy<qX9ob2MJHw2#T$Q_J2n;5|-!qA@Cib;$iJCu7p
z%>D~!_1H}^cfwsbjLN!^Mj(6`)ST4~Ng0dVLZTuB%wHYeNRse4iy-%5gFB?aUGe_B
zL8$I!7mvl?$fc0GHR~JSN!IDfg0W7e13ZH@Q@jrng%AmTtk#e{Hf@c~FZysx@p8;_
zTg$<*p|Me?0LDNo-bxj}C^IB0(=Q3jdxO29GKLsetNWtQp}uB{Bh{;BF~ve{lGC>z
zpJ>9kw2sh?t7AXE$()h?V?<CUL<;WapQYAW4EP47sh75pH!n+lOHq^cGgDSxa5j7t
z$DXy{(eKmO59d`OVL||zeWLe>yYvGTi-9&WdY%+nr%8|Ggai{u&i79?&v2S!%T)Q!
z$50**tB<|ReoOjfP!gcW6zP**{Az~;J)+>N!Q04LkH^C{P>5*+Z8;<3X?X8eN8i~;
z>2L14P6=2&W5x_SY?q6cfq+0Co~IA}T14@n_~sL|{=QV!-6o1VL_r6R7pqWtsnUm(
zuXtHC%4>V^T~!}a#~`wnM^YB3DV8A@GMd?<UymI-%TFD$OjO<|)$|m*yNdq+c3Q<!
z4j-${C7Zk(l0V)Qfz%R*%<E?^h(aL>9P`S+Ep<xCcn~D@IxCZ0NN^aVX~Nf!WL<L#
zFN|v4dqUy@=Pu6|d{0>_NDZ0xX49aPgvl@rMlnX$-0WGQUd1%pZD?Blyjphg8Duvh
zJO8x!_k6jzDe<}PcE^AEL$YoUpiq{=U=K>Eb<>iM5BMe&o6V*h1k3sOP@5=Yt!c-S
z3|WE$uBB|OOK`!>*an2(Z63w&kjcpq9kICWT@0@sENssg=H$50N>TMF|MwWvxB0+v
zzAKlTZIk=&u}Vjue~*1ZX}{I;DEgS+J2eN%?k4k)H~&B?FkF+he+7eVMV@<Un(kLX
z3DuiWZ4S^&HaCA9n0X6h927$57@~uN%LlOGnDqQAU!lvoU&NA9?iSP1ABk1<ML^hQ
z10_g6oFS8U@1SZ$pA`OF;OPpG2lVYod}{8t`!qznfll{n4Ts&+<jQt@EO3FOx~)M+
z#T)rpqw0}t_N%E_mVaU*qxM9*aP(GHMT>*rilIUy683!3M$TH$>-20+j=N-0Q_d-<
zDLE{7_sZ&6WUY7ogXxfxf`tO&2vI#ZW!?qT>chvQ#p>K~a=$EZQ{Xk0xzb6Hp+;qv
zj;5@r8iV)yf@$T5*MI%v;hXjKJCLLo?hg!@QAXZu<hz$;Gd`CqMBR<&YO1kGF*&$-
zh+>wuLTdSqO@^g<KMzzG?7q6=3yFK0K3lCj1hI{U-9+<<P&qul6Q6=-FD#_wrer9%
zxM0|$j>&)JUQv?y_BMK~Y0$sJ&)x<h@6>Y7pBP^pUy^DgeX2FJ?}fF+^kv)Q<ac4U
zs&}p^4(nT{%GK)moAdR8igclMEu|i`4@*f1-pW?uJFU{ay=-o$x}RTZdx_pUtb{TN
zUmw&oALOIA`b|hzD+w{C@}M(w=CbQiuGW+ds^rM|_D8o{Tq@F>D3!ueiC@8~LM34B
z;gO8oEr|b8z;r)GZVT!SA*8yMNRbNxeNs9qiyRrb!=B?|2_$y;7L><06nLAU0fwYU
ziSZ;dWd(3vG()h3&G#Mo)Fq?g!TjF&B!HXSf17Ze9487<-j9c&yUBrp7l_7%Z*>OE
zX`!W-JIq>b9Ih~VB(J|cBajG7^@3xQ<y*DB2*lr$*N<QS{ppNZ;(=qtEpC0s?pVfW
zh_=7Rd&g)a(H-@SCKs!_`Gyd2sItOmv5i3D%N-D&X=igLZFt<vuS@wX{j8Hjziwte
zPZF2d7IOnN^IP-rhM4d?&V_HQ{;Bn|Oi))=+3E+ha1+BOh5ee2=xG<3dOqNFn*o>r
zlSb^VM90R~v6k9WBRMZQ$ZFF51{CS%Im44qsFVmx7`7ae*kdCvJoix!osTT{b;nZ<
zq5VuIIPNrBA@R*xbi-4tD_Q}0xsjCHnNl!e(NR2cx500DYK>`at0Q!GGj`gnLR4JA
z?GZ2Fs6=*?`T4=vFSsK*j4gy}-JW)lo@(9UWWZ4<I4z+Sg*Bh$zLZ?>)sb|po}0dH
zqr*Y`vhknd+paHE-+tFz1_=ZHh2%iur-Y4jUhNr|Z4~uG;Pw_z<|!}UiWZnF+^CgF
zWQu${2%l6OZ^hz5Ru}!r*3LPhFHWF3hSUPL%*R44EYx533#WLvYx<`(BzGjZWQi*y
z1?tbfNx|SEfDxR<YPjlraL<`Ku?{GybCP2=kgBrS=Y}!JzaXD&p4N?q1ip*lyrpSm
zNW6~_QpF-Xm~0>@q75=Lio9Tw1vR!+N+-G=AWUr_nwQO;PlxTUJJVuYWQ|%)Gpt2J
z&fkdsW`5G2GM}L4;g_nT-7*d+;xl{CZZ@S?ZNFPeLguYI)})?Jx_9H&xDM%F^4reb
zKar9-a=IG=&de(U`&p0W)P}N4Nf!^erB4~d+eG`AZ=d>mFm$5iv%+_iLN96W6FYB+
zDPIynBhoQ(@urhHmtQrM8{W^Z^4`1;{~N!;$k^TC0s+P9vSzX0z_$=B6}s*PZLKLu
z`Oc0K$7^GMwmwLBRn_Vy)E9yG&d>BUcee=ddt6yFojBOo48(g&GeDp6&DN+@a<1N|
zAg9#p-Zv$^6YUrMdA5l4c6&{;fsqM`f?wKJUw;FXh}?vixOrt>t$Q)|`UKI2zqsfb
z<3D%G30i1yF_kB-$M+@HP@Stbvvi!)t4_wt9-pS@7R%EFjf}hu@noXw8ka}>#{EZ6
z7L_9hcB4@~^yr!H*s&-Xdehhg8_zBi!!G6Xat~&_<;&fwht?#o7wPaixe76=?%+fm
z#F-*?xF59}1joNZ=1-|QkRgERf^b0-t25N^?bd7Th6x?~6hg!?G{d<qc2!%6(<$4>
zQaNEHLdj!m2Cmk1qy>MC9MCH0ou*jV)d`aHR=<-nO_Wu`Yi2!Tif2!|yIF0ZZ4$pu
zCkzWmA+bipPC4mE5_zNr0ShA&J#+$PfV}N@*4Ab9wR`qY@orFogry!?J=xA8b6Xmc
zhc{tb)K_A=?$_?@T+_kAOmvK!GE3;N$(Gyv8B4L7bUG`R^L}P}v|CZ=Q)%|3!+At(
zzTuJ~+#=qSG~l%4hO#1vw<q<>K390pC^RR?%6;z`fA&AtMK8^UfvnR8*yEp}aHsBG
z`yY?~Z@&N3K<o%-^>>8-LvWk~&KuF-zhnRB|Nb94#H{wSkN5xQ$Nv{U{}Lh^d)V@y
z-~Sg+{cj$+G#jqFa~$5fXTo^;<(K1s9=ou}gPWMp<_4mh_f2xY{sq=(rZl-i;X75}
z$^ZX<X@yPCgYn2`ehf9EgKty{JIUC9uhX&XoaRV%h!lnvu0;vA2epE(2{{Lwe6zcZ
z4g7rGOmoW!O`CsXUY+<Jhknyxci!HkbLH1dLHh@sofJ@EKQG(Sln!Q8^>eWP3#05c
z(Ux4MP#{5l7lq)N`{|9&gG4dsKApZN)*)m-sB?0rM+0jjz5Wg+>@J^54|Fx9;11S~
z+ds#i$43@ONq0HoId;DU=-KLems{~D@nie31H9vY)D66>8GXX1?TX4f$&Tpw5k@i;
z3d8KI*{8bKT3Q-|Y`hXrYfwt$LgaRjJ%%ut_HbQgG%HiCu)uLUH<c>hyaMW@U_!dz
zh#C0Zy@QZ>ZDK~qj#1*$)$F`CxAn50Al=)oFOKEIXWDbka>U#fo{yO$=c|}jDq_+-
zzl_3CHx$v8dQk=B%(2#fR2z6RRLD9I@P-EyN7v@3sBNhVR-vg71uMQwmOZx9E>alO
zy707AaO4IC0s&x#jl(-IG4>8+f}x&EED0g5@4usSg`6Ws12tdUgx~qDHsL%#V)~pw
zM<*om#O)dE@yBf^KPgE|mMOT7Pf4LTI99aov4woJnpW>X3cvs&HB*j?IU;}UejdE>
z^Ev6dHH{9C0Lq(1s&&&CagBYmF1Goq0w&3BzM{A#&Duz!jpjP5A7!ZtXoTeq!sL3!
zmBlew$d=ti@M0lxk|F9oA1HJyQ?9Y2=i)Cn^9rIy0*r%{Nl(m?Qaw^HQ_uGIoQMaj
z8@khPsv*n&E&5uFqB^)Q0*rRE8zq=U!JVH(Gc%v0VU-r=$y7H&Z@&^vxMn+WdEFd&
zKWZiyE<w%v>Xi=ru!^s8?K2B)2`ZlT0Cf(oP*z9%onOr3LZ9lx2Q9`3hiG%Ytu~l#
z7U2MMvF(kL$3Pp;B30_P21T6=Kq+CIJ?>8n1y*Q_rI(3=@orA<(un@ZY<}63OUN(v
zBnx<0&liw|kKRksH~!XmYr2j)%g{3AVQJE|H1#z1_TZeXu2Ub}z4Y*zQUh?f=~3|^
z;ebuYhBdT{)1taf9o38xk0oAPO9I!PL@jTqk*W;m!oSCqqZwU}QFR3+b%kU~L2d&>
zm<9AenIZc%yVt3yPpGm~b!y7A__)X_7&F<ID(1TOcdw@8M%Ap1suOXz$<@ehpH?}I
zE3u8AL@QTVM}_I^$)&Z)$(1;V*;L2pbNz;0+cDL-abgN6!qNvjQc?%$WtZCJmrdF4
zWIf;dNG+%(R3)D}C6Ro?a*mIWegG4qgW_a_I2~j1gM3lmVRh#H!9ULPvzd*RuunqN
z*ldUqkrtAF9%f+O=Pf^%*Qu}XI?^Ro8sW`d<$hxyU_WRgKI>8b^LiYfS6Y<~-@SOF
zQY2Hn{Kl!2=rSg%BHmE)?`#2n=c+p2(5?LA75RqJLXw%mWHLD8!dczjD+l2sy@6Fs
z&v#X19c#CQcU3lbdGGgoWs_ctNq$RsCxJu)G%7|bfnhPYLM;3u4m21YmFCYMfA6@G
zq^jS|&qq|yO+p!@d`=H}qBid1GfIjqycNeNS-9!6k&shoWb?+1>OQZ&y%#jYSfPv`
zmd`4<@uVd6$IVvVE88$We~po>471I4?&!-t>Rg;V;fQG-6D1`^a`M^=O^d_8X%IgK
zCvI+SowUC8&UuI|+g4LT8tjNBQJj!SNlp45uK}G=!HDxs%`L15;)aFbv=zAPQ?}FJ
zRt=t=-~Vv=S1TFn@a6=G;b&wRn;D|@g`))(cPsy1+lLt7R=0cs(j|2=MVwfmRlfH_
zyAb8_e8V`bJ9NM}@jlpL8EVMQR`Ici*%yf2Xc!fOwQk>z;G@@W!dut*gy*%^C)B64
z)<w)UeUz9Kt*pHdZ6BBzx%#x*?Q_;<8bbr!=6%-aTesbwB_xf6RHZBXoU)f@K8x(L
zrHJuZtSf0Uv}Tw;4*xyYfO~WTN_4&-Zy#sTUK#aDVAQBJyXNzr@4`)vq1QZs%v;@U
zCB69ocbnHKaE)98!I*<Pps)A>pm*+l3zySp*rR17G=7WktMR>*)C2Mp!orn-;wsS)
z(|xa98y_$E_8~ZRC{a?%Ton#y0B{d$H!y|&lACV6wXlw`F5X($t2O6|Pi3;+6Q9%K
zhJ6*}9Hc2gb^+<5&z6biP6n<{bC+u`+TaNPK19Cd%qsASod5UOLvN}wCH9#a%cN)_
zNZMvFH8d;%MV2MFL8Lm~#Lg}fG}h)AEZX48ZqOC84`&}+MbA$7y-)kzn-(E}@iusx
zAK7Po?){jPW(wZxq^g+YfFA_bvVpJy6MD8LEtw${$l4I;>nbRlKeLte!|~y8=d4K^
zX+(XrPP*s%T~Tq}NK@X}X%NH%BS=ZOzAV+9pfT?dUV_;RU)TxLSM|_UsZH+Lpe0QS
z;IUtp_h`aWh#twWvZu^QcITs#vdMn)wwX<Nhu}(Tz}(qcS?5<i7Z70a;BU)MZ){I!
z2t34}OPYB<A_Hm|yee`ea`JjStQ1H?E;;=9`mQu66cogl11P{jCQe2q0`lpDP_0jT
zr<-*pS|y*yViq4AVDN}-9RiORmJ^3f7E1cA`(85F#7F}^`1*VZQI@@|Erc_9Uv$s0
z>gxCzRX_X3rgpJoWQMFyLe-eLK3O>fA?`yRFq9y@$?8aeyXY4H61#LW$R1%Xu*CZ5
z(l-*umS5uw`HNLh)z{|kJw`sMv%R<}Zs?rjyKp_OxKh+@8&546Gl6OzNFUXU!E`_6
z^u4`9n}>WZzw(mYBQr!NdmE%vAGjd;z4$~zven~}0zyVk(<YVI#i3$QD-l!ZHVXHj
z*bCJeu!(RUnid-1v2jXUfgEW{aNJW%2lYYWn7563M$d{$C_bD=PeP5NQrHXyqe5#D
zY<pZkwyn0;?Ob+3$!Cc;9!pRq&cnyYs2-VHA(k?B4Qe#5yhx6S!-O?7)MD@a0v+UA
zS(~G2)dGGBi*OJD*}#@w>HoBPC05&mfd?W$91K=_b=4)#EhoB@kL3_INtPcLC9Qv{
zMAmR}`vSha@%uVz?7jkf&Wq+4)a&_J_^eE;6M}bvLGEK?!D3<lHNArT=#+r3Hd2}G
zI^)mCo(VamQ%~h-R-J~XY?~;Ui*3_+3p#y}xk#K`*~oA@{7i4`!)kRbTRrt4Ruukf
zk=PjsWwfE%lC=wCUkmA6ENAt2OT;eH=|h!QOA9CYM7R@nYt26>Eu-BZpXK+;70%9l
z{c=t)`w;cO5&|`q;33tmM!SwIrJIow<xRGzl~<5hNk1r|^d+N8v$1z4<RFaWH!B~b
z)3ts&7viv&V-)YEFfHq{SN2r@al5Trto{>~rtS{$g=(=^lI;fn9z(}lhvaXA;@hyp
ziCwOCcP&wT^n2h*`kMx835;N$Tg#vE8s<*#_xqeCb?ah`+a>b)?rD;ikWJ40%23wB
zow`%e_j2FW)P5k;l+IQ;D<z=~zp{K0MtJ8rF{i|=;((n2lCrp&LGnNLVtVFUnRnPL
zdn=lp<_ltpzUW(5=;GCa5_v=i)FwWHkv&j#3bBQB2MYrOe;<H?^Qw|BB$O<f&@T*5
z!(l`2#_qc2q0UG_Z=*^sR@rd|WfXE+_1y!7My2dLdLom5F5P)E<ch+FS7cw99UnBM
zK)OqMb&nV{`LhJ+H<H9VSsn8ZQbHWB8K|yG$bLgIh>VOR!Qpx%%j?_$Hfxo=!*0cB
zCl6?u*Bt7yqdJE!`{I$0vXeWLh3@i5(s(-dG9L3MD<e}wQc*qn1=AQrh9*Wl!_^yi
z!P6u;;euC3!>iTbvZ8XUaZ_3_hX39(Ev@-Gi1^<vZ4X0l6b{XJ7hESa^;ygqE~u5H
z2R(XLC$}O&+L!)ES(y(&3QB-SY-X8=^I3-IzBvK?EmH;67S->Gc#@D4fT|$rK9Q6+
z1PWdfL{;7pW?OKRT;LMSA_f7vGfl)Ec30u-^{b_yK19*Kn=#Fm9Xh=f$~cDN;yPb5
zRmgrT3UGaWxRVn$(=mmx-ii&wXdSS%t#B=WJ+^SWoZkB_bEsL!v7Y}~dd#?qwtQrX
z#NIp4o18+=q!$nsc{l5?PWp+wwO0!ZU@(Sja~DojQTxwYfmWp?Wqu7uRpZJyx`~cu
zs%Vl?h_J3tog8soxbsjfsh6H24mB~!Fg7%8rT4(hC##{f@`fFbRxfj1D=|mQc=n2i
zLgDntaD$V_8`&}+VTc0K=hyxDw@Cg?w3mU8CL=Qw3gzLms?wMb-UV*e8L$GpbJjiP
zRNugD6A{?bOMQLQu@C9gGjZf(kTm3rm|CxKo!*GrZ)%cewytYiHDM0NIy%34{4=B*
zV{3aU@7d#u*O2xn(P@v9x=DfvC{!gb5cn3#>EJdM0wV03+|`V_aM6dG{0iqH=Vjb9
zByV7TeW;V>U@F_OG2-Ro^2BIaKKJ5tD^<zrb6$Xw5YMQXmzxI7dJSg2Lclh=);HG#
zs}H3<zWOrf-rrQC%cs3iI~YbYL`T5hF{aA9hRT$@GLZLm)FS&3NzRUpaw)+C;pbsZ
z6P-JLG;isZzt*09NXN;(wA>R$c|6R2q~2#Q$rr7`-13Bt#&Yz-e-ZSSNI-Wwxk1G)
z4#HN)_FBVti4pWS57G=O=AZki1?<iSe2Fd(gtCR>Ze-Qc=?P55+thOg+67*171h6$
z8TOS~BlWDTK8LKeVs=>@``n))^0Lp85|a&*3=Jdqwt>urD?Z0)=`K&nBD2UFEIcr5
z`|JeGP^8yHq9QKRd;;g7#0QYd{J!%R6_R@<@qOoH5f5EucY1`t)%3dOq~Zkx7$O$r
zQ^Xq(8QKx$>j1SX-2{PfwXbOVCAW2tU*rzZw04QwPZ=52)xxz`t2kCa>Yl_~43eZD
zwo9jaS1Dh9T43aoEm@gLO8QEARx9h$1c$qJw}SK<n7LuOTutsL&nI7`HF9iSlY8sZ
z(y6*7Wf1~K9uEH(b#EHg)Y-NTW3@#)fDtXV1)_lB(1L)rjC-pE5kyl&%WWG$g@Qws
zG6=*>4k#cr5kv!4vn?oywxEKBF;PK-f^HEhLx2bY!W2Rh$Uug^CqB>jtnbIS-u3-`
zSod0YSA?BwU)On^^LZRrtUq43k<QLjHs_p1qK~fMQW$~-?r!r5z?d$0lUUAX-i_gA
zP#Mp&6VPAdRJ+uyZaoad%SAnMnatum?d*j428s%rRGG&B;fE93Zv!E3pZOBfeY!Zg
zd{S>(-dKCJt`vXT=&kbANqBr;gAi5JO8a`3{WP(!;|OF%Le}5>`|wY>@hh!2+-imt
zeU9IjXn>!XR9i3aVBBv?PtR;iV%gjO!63&GC+WRp4Ld<{4P6u+{cA&WbF*qUBSHGF
z*(U~HcKoz|R337CYx<5Krxv!yf5pl<@i6%J8&}@+c~5_oiXr9s%9%n}Wea1_nx44Z
zHJ-w)Oqb@foOo)g!nbm!R_y=d(e~K4ZGPMN_FfchRexkHDeRboNF--kzS$L2cC5(H
zvF1N7mL94+Cuzv#eUTH=`p=3di8BS8Jy&&aAo^8r`E$3Y^})K`Gly@ly&WB0c|JP&
zw!Tfo@>aQbdB{7L{O7^mTjjfLy3&j6xrbTyEU(DCygNLCY^E)LTWBPqS~r=!@{=`3
zgN322KUalrFDc5(6{YW)`f6tm`JZ1z&ANO-?$3>N(NjGIB}LT*2cqrs^UoLOvsQYT
z-X?VUUaLuV{?CJ??w?E3Z=a7|xq2uvqsW^QOv=T18R~nbl#hIB)0SkZ_Ilj;ZfRWU
z5&Lg{ZyX*@d;{TpVP`WNLl+R)4_ixjZ~1m_i2D-1X@7F4yVcVK+%6QF*h<<)7yIw|
zB6qh<OGASYX}l82q8TdH#<8xI))|i}iyYqEjs3Rt;$eF1*3FhLw%1C2U&=YVvI1iA
z>}>YAp+7I}br0?-%aFNlcO@}P^NOgdajv{N>(gz={daY;BAaeU-<r|hj`q|fk%yuS
zhv>zIh?i57d0MACc&(pCFK`y8>am|&evbTNdi0I#?<28W9dkYoS}bw-a@zyn73(XG
zpTi7AeP6%vpZ3YTOTV1p_O)NDf5KuRgWDD=9nZ+<r&2xSxDRrqBTpl}J$UPOne(mu
zq7rs8#6D0qW?b@^wzb~AdxLDt@h8Jo%U&Mi-{7QQiwoL0Uw<Mnz>ocG@995u9Hm3%
zrM6v3T(13URwNHn8h8ztA@JB8^Z8R&bX#Rn%&St>E@L@v7ar0;vx^$lej9h~Y^2SN
zdg`D5+`U<OI$~Fam$+?x(LY-_Uv1{!^v4VQXpPP6b1y}gkNLk?d7zrof51N4Gvob_
zY?n@Ga4C77)MI_{cF&6wx9+eG4MlQSu3-T|Tbz}&e(Lu{Uw<L<JNMjmNvhSa_k(t3
zR9K&KyjgM>33u3;Wz4pACRgoE20^@C3^Zo$FqO@zX&UQK0U+?6^(}sJ+g+c!_8zeJ
z*5PQDfRRB*xZ;>I75Wc}4<@i=4o;DY@uuqHzLk!Fe|diXYxGV3lxpX@qYJi%08M0|
zKYlN8pg=*qnA6sD`Of*_D+RaYOG14mSbj!^)<7w!UR4#`5}#KJC^)x<R{Y94x&I}y
z(<=0(UqQ(F63dd@d?M6h60T(CCk)?ui!E3h@9v<F+4<s1${!E>C<(2>!77DBGOqXR
zQ$D+vvE=xPQu;ls-JwK5zFeNaH#K#IW3D5(CbJb&$LpEdj#cpketXYO=;xM!l|L+(
zA`BA^I)Rx;dnW;Yx*RG3s+WgHkbc!+HUFCJ*o+?bjsu9i-iRgvgTmTrz*r5JBiPID
z$7OuxQ7_BSx5}zM`|aqbZ=woT{23H@C78|2UrIcf;0xJmgos3(oX$;=*}2czkho;H
zv~1g}#^5L~{c|S_x!5%F{J50yS0lUN$S_ni*KD2j!d=@4z5@2a0ZQxj%@t?YR6jm_
z@uMT3WUleL+`ph>;>K%9^ux<(gg1f8p{_Rtid9+ZPhD{zg}XRkqN5tQIN~#ql3TI3
zQg$S9>XK~p?O(;~H;>dU=g0rZi%NVI%J#T_{OR98EB^KuHKwP(*<2cZcnB$x+x=@+
znj+=3`!vUYy{CF@k=3AcWJlA{%F!8N{_oDerr=-W`KtXT3%-vEc(-ldo1NsG9Q*Ml
z{vkebc+cRp!5F!0_uJhej}DYx;rkv<Jm7Zt>!?IRvo7&+f7O+q<Mc%vYDX_Co2&jl
zj2wVHr&7HQR6V^;J7X}a7W#e^a<bpE_=mq(f0nSH1^$`w``?Yq3td(D6~%jYZvUrB
zHT+QFak=^BR>yqR$A@L&&S^TvLBe$tsK<%pXCB?S$l4+-C@4t&zNT}`G1;&6=-$W}
z`$I^mbz9kJ4!8EgnGK5;<%DFnwb_kG#0@YD_QBv^f1TR)(Ko;BimE^LVZi~LsK14a
z>I0=Mn}1YxANM2U500%eX8Wn|T?s;0sVq;zBS;_JkDcO>;t!SX>3sb#L;iJlw*4n%
z-+MYk_$hgLzoMczta(Pb{r>TK_YHsH6=7{Kj)jv%BC${?>7rU5Wxex=c8s3zWj{H!
z;?T~p^>0e7wyznUdcE_;>h>Sb=7j$9ZP$gfQ=xFG%a`mPXA4*@cq~d2a(IGc{X-W&
z_i2>-<+gZ#8M<Sge#_xNkEW9K9%t6CUSq%W(>KwV2&&1YOBWWbqzqB=)OuQN4Rp>=
zQ<Ab-)TTpW2;+=oDa;mne{#kRp_v-hj2(Dv&f#=bh5-R0*iTR2?$Ru~6nVFU!K-Rr
z*oT~Ec=g|$*Y%$`i>c=guyHf|Y-l!WZmv3iAcmWArLa#7pg#53YL(no{zY$e`k_?e
zTb)F><VeC_I}dEX+`6Ug_E6x#-k-h-oMMmu?C8YVA8&8J<TxdXYkK9awq|^Hw!iX>
zb2)40)6(x6&-<n3a4##jHa@oqr0noHnw`CLL+i?ny*ExHElYMIAvn;h7)bR3LC>I1
zaAsbdMk>-m4;|VNnoaqndpY``IhmUx6ANlO<w^}{`79*pa~Z?M+S&gF^GFXs>m!kv
zFt4BCke#H5#(8DBu`H&%N7|s*N_z&;FwTYH;>HNl3<F)AwjtSA#>VPjdFr()=?m|q
zS=Zzp<1H1wj*W{H7XAY+znxL@<$&ur3H^Vy5CNy?jDl5+_SkA4eyN8(+S@>HeKZh3
zisTF|GnOT4G(rJ4vd1n#UtfSNPckCyT^2{!2tyu4)k(R&Duo6fVO_aNPu5eo4ic4}
zIGNYAwvq2lo*nm^pRh|m<RGwnCAM_#HfB#QhF;SYK6exGffdf6tEOoTGe7@+B`ti8
zdz(xuBE|yGfsjF&F<p0anB?@MN1lW`$Sk{n+Cabsppp98B#b9&hHqqk_h3N3>1jCK
zP^D6n+#x@znyLHqeQ-qzyV|VA*aU+IOf6lAcX4XvDf38S9CvXK46N`vtrW7xc$ekW
zRGT(CXzYYRFXAr!88``{LhF~TH6eJ@<l<q@pqj=!!sC@+;2xf-6DKl;HO*?(62K2e
zot(un_2XyvhT<6ikUD@Pki+0k_pSg&iV~}lOtS9)x@2|=g#zG=t+EBiQs3GnrNKu)
z1U}w@aAIVS+K3Y~JJt;yaNr$lC1%mDYOrS>7XBi91W-I{5#oGC`!R$sj_9#9VwTR2
z-@g>=um%_Kdc>2u>>s%FR#7KhDv`l|fn&tQClN%ECYZE=#Ej)TWMDM{Vp?npymrjR
zrxPjjlG$mz9z<R?ZsFg=x!2U8@v$+zPFBhBdc3Jm4Eb3}mknTK@8xbBDy7%rohUM_
zzwT*$al~LkU9x~xd2mvvN!GqU<jti=G{HJ(HcK^7P5VJiMH(N(d9x?RZUZc4Z+b6H
zHLbA-C&nmrOzTRUU?Uf@hY_UC=?1;$1$+A$ijSQ%Q<<uQ#D_#CWzn>TnvJl~Gn}XI
zEW5NCEscrB2qoR{5sby8rNER&rY1O!xC_1<HgkU5t^^fsBaFS1p|3N1oZe6G)=S6w
zf#|UnO)s{eNn*GhQBXdCL^8b`P<ydl14U02+U;-5eB3z&RDJUt1XoqA_%RwFoFd{{
z$g%Ue!qz8@P^*_Ab_v3Eb&@G!fyUeqK8;0VJJA<5Z7OgeVqNa=CM^6FDm{}_yg#1N
z{?MssT4ywM%7nsK2Lk+R?Lwu4Fst}o3`dr>fnvOBj3CA`@5ca|WhTeP`nAbHp(Q;d
zv_bC>>cZ^QQVlig5A0ovD<clpd~kq%&sQ(Rgj{<D!hC1LznWP;IGMt@diq4&Jzt%^
z*2MSWQ~vwXR;h7{M$bG<Z&eXsK?5@Y)Le?2@5_+B0CZ1Zy^(Ivr}-nBHS<(qJ{sw$
z-lvv7r^T6Oa<K-acynhuzat@xBVhA^$WLo<>gy=8mqH)8QrnO$>q?OW*sTxVDOV?a
za5b|r0zldKC#B3&!E7QXti=Vjc!Q_4L{HOGh$4N&kvv8_Ud|Fnsq-|(TJ!e`z$#4e
zDrLq{DMKy1P5T=K*0YBS({gGOZ&ZVqH(?cFaL4L#nT~!Jbkuzh2^(o4z+_$p#d!K^
zChu+y4J8%F9!v-m=E+nVBOSu18t+fa9}yq;O6~~Sr8>2?Ip7$rKAYl0Gl_H2Ba%V6
znqq=DEY^nVvBY8xOH3^9v6|>-5)T&J8yNNqBkHm)pWD))_fSEPNL{&_q8PWxpnId%
z%ZAE&YyoytD8d=&$uGAWB1j92zJe%{tbtm9l-&_YO*YiIuwDRz&0}{-xP|~;r<5pU
zJ#GZp-XrCYJILu(^M@!Ia;=dW6B~<B<lVc_!$9(FWC}4}_6|563Pk$#;~7RssCLtQ
zdn(yHR*T-()Cc!j1az@!GODMRN@ZHc^davC!T}4oe}G)eUB)eo1(&j3&?fYWll}M6
z*Mj<<Zz5=uw6NE7Z|3SWX1RcgI5dG~Z2?Q>3bOVbkd}*M9(2O4OGK~+u?l5ZdaW3k
zq%g27k}P>fub2iz<MEkCIsRS#Qzn-R`xCWQ`BF(U&J#S01ndaVl|w+7nu07d-qQPa
z!!P5!SKbunoF*Q2uMs_xt_fhGzZwB6dfZrxUzlBUYkAlBWYWCp089++X=oW<yY;<l
zbyD-uFXZDI$l)SBX@0dxL=%qHXsf7k?qLL}P#|cNs-)^oAvU`Q8x+#PX{HO0eX!WC
z79+(psdZ|#_WOLGX}6))b<;d%bu)5RXF^)s+i+-Rq(@ku1Qn6V=~9}@NXMMG6ca%m
zAZP@o)uhmVuf9D~4E)`=aMHpb$Lyt3S|bJ;)h3zCu<L-d(zt0;V=dP?;$Xd6I*G+}
zYE1IGo-id*Lo4uBKz+LhiX~WVebe2<ZT*9inIuy<gW%IfP^-q}A|BBvAa>bsF^2dX
z!lftCBP{q6&J%>0{Ajjh_k^Gi1&a9Zvl%X#kEaa9ufTev$$T`6)FZ5(o1FktbT5bx
z2E{NoK56jQB`+BW-wcO#rW-6Al_bPP(^Nf4c()kt<w(CF)mE!7?i}nzI@HNCl@oBn
z29ua0OP%)9Nb@B)2#6Szx>^?xh2s=hyh>P|B6(#`?@l)PoLAaUS_5u>zso}IN(M?S
ztDq<*WD4#1#BCA>%_rIB8)!5>B-^n13HERCE|xCE2r~;k?9eCV*yxh9P($|7(LnJ`
z!MGxk^UNmLFBr$#C%A|>x#n)XGfiiW77NE=vn#a$igXq-Vl<fhJN3If)t+8i=8-Qm
z#AsB&aRAw4cD;`JNr%gF%G@h;t&>CW0E*VgXfcAA&@C|-@~l5!*J?w@m~FOoEp~n1
z+oKE2-R+?8Y#ghIx5>oN;#M9|lT_XjSxp$l%yO}mXg18bID%u6nfUov>bT(%L}#Bu
z!7*Cln99Omrf0ZBb~(p@xJpO_v4B}h!BQcOb6%-7K!svza2!mcQZsA)<e-@_WO9|C
z|GRNe(Mmm+v=Hl38?W(ujWr1kMq2W?P3Ew_&R3Awkbv4T#A#O2Y5ME;$({wqTU<{p
zkzPwn(a`tKIuG5;`b$&faKkhSgM-AhM{x68enXFleYGqGU!k&#IGDIiK?_HKx1+sc
zlRlh@18<^1Z;X?v)9hhBRZtA@=L`DF>e+%$Sm-GGuUR8JW$+(ZPJE!qH&By+oCMI}
zJueJ%x*&RYammESs*=)nDFB6?YC>oU<H~pNC!7bJ+N$|5{0zO?h!YQa&ic?#v5V@|
zIWB#X!e(k|A9}3@e+{0=c1h#$Bs%?or*%xsna5M6b;dH3W>igR)oKdWL9h>`bB0PJ
zEzkuAY@pGziN7X6Y!)dtiwfvCz&z{(IX2wjf=t?E_L$tIqRu9vWHL!_@0l+dmjJ|e
zQ(vzWSJ4sgOs);@1Qu$NyA@0~=f)u;8WQH}3u+wp3=BbR*()ZSnb!M$U+2paonG2w
z5SZE9tLgmS-d=HR1!)6~;RD1F1Rwg|LOA38arM^B?#cOb8Q{N?kZ_nyo>hNJYPj%B
zhgB~{lLW-#SXtAHC{OEtZS4o%DxCMP*=d8Z#V{9qSn6r!ky@3zkYY5@<3Q;dz$LAF
zG*k+Nd^#c4=5EJeyne6Scp}-PI-Sd;3Cr`;iY{-Mkpg!G97iX0VR{A@YZGiRFiHmI
zr`OK2YV*<IFjAc&Z5p4c>zy2kh;7o*Vhp|Eo5W48e0%XIm0hDslQd-ZCID~=v<;@W
z!Bl$o#7rH=u6bvZIf8c>Qf+n%5&RC;H)vxF!S@Y92MVnrx!<?WR0$L>Y}*2-rcUvN
zfrul#@~5J4P<{7fI&r%ZH3{?8>*;+h;OLHMGaEO(B-0SQM@{O)@52Vb(SijP;{srx
z5oxHt{7sFSkNhGI2FR!s`LsrxAOo8+*u;ssKw6TZkb!*4?~DR5%YQlb5isb@pz%$u
z(9izJbl_jJ-M+J><S)&?K|-c2qaFYEcRZPyrG)x?&6LMz-vAuw0lFAP-HDI>|GpTZ
zfo2q3J^8*dZH^@PRK$T7h5vb+iFnn16Z8u&^k1{qo8-FJcITZ5XCGh4|3Pv1N5|!R
z9lg?n6EBplp<Hj~x7$nm$LAzaJZDoj1V^Yf`9ngZ24u?r-T&9-&N=$vY~No^yVQ#F
zi7$UXr@WE;)ylrw!y_fV(J<ZOGMZcJ7tL^&&50+bm1n<yCV>F`6@;YO|5@+<pGO+5
z{!SQtIHuS`847;MA#dLGFza8l3U`O~YPEXl#|70T8Rrk@ofkYH_#}+W<udK+E30lF
zxLuNwr_5(XBgLbucm2<!{<r+cU$LWUOwH+k%?y5iahhhM>II}n`QFg!M0_gE*!sWa
zHU2N_oZFB~ep$9ZU_uqh>7Pd`7vpG5uCZId%1{AbsZvAv_(UO#Y0+)$zA;Qx#cO=o
zvDw(NF}4&JD`u+F2fWOuW|-ey#A3c9GhHUt2452`z{es(HA8fX<sjtx+XyvbIzi}T
z?q}q-4(!F2s@2x7tI~E+B^3#v7v^hVGMRUB7`rEWXJ8(|T$&oa0Lle8sbvzZDLed>
z$)y-R3m`N8HOt4EE(|Z#Ge}1|CsRD*)zo>@0I{Gqav2OdfB9>KG1r71d9F4PV}ZJi
zDv^rhdLPrPzSn4GCkjRm&AeN8eT*&avlBCY7{1c6#luABUqtxhad;$8=ZTXO8q^XT
z#&=CPe(H5ttAdH1>9nY3FqujfP*PY*2GzBEVmhn^+FRu=4WZW=?Lw`9c(RWF*f(z4
zL=<Q!cw2Yab>_6`*a0tzRHqR4*_=;YGN?weVRB~Apl_m{k2Vcau$r8@xoYIBwn|N*
z(Nd2K`)~;$#MU=SgmF7Eu$Wenx)!s}3T`8arK2<8T(a+atx?iFl_utq3a=84L4hl0
z^suB6wXt4Xhnahfu`%wyW`aKHqhVsJwl-iwPdifzP$>@)`JL?V1372PJ5lq8bU|i^
zBIA4kbh@e8=xuM<)K~=m0@zsaE=5YP*N7p-#QY&~$$+*=nYI^!u1@B%VeBGM%~<oA
z>ZfGm7PmC|x_h#jDy2>aZ_3ySF8nf9rH0g}OWO!e@KYW5u@%?i40QsIL<_)1Xt4s~
zTvTFFsWTCA_B-3TN{&!Bx74>)%h(V9`anlTmmL@4?Yd;F-*zszwhz>GWkVK~0D)2_
z1BWE9*iV=6uUT2^a|ybPdrQiO+=AYr7tmqIvjj2}^Ih~>g%IT|ZwW)lY>|R0ASK`R
zDv@v_N7Y6#n*7@^5qu&O!@tWGfK&jU2HzjLNht;Xi;EbC_k27x%*uyq;72>m2C|d3
zbJ2Z#EEel>?qyg!wv7`CdsNK~_YSX*)joC+nIdi7g|T9Zbi%+W6TbWCjEi~fQ^Wox
zLU4-oXRb3zmIK=QQiLV|7A5m3Jp-)>m#2du;GN_8m<i1F`#?Q|=#p7JqF8Xga8#+G
zLSBU3zZSOPVyBvZTB@xQl4%gPxnwEy+SM&#7bPw{iB+2&Y;24Ok;hm~SU_y}O8hX~
z(gi)jdn$s<0(8;i7@ACsh>bnwf5Kceg(fHr<c8zISWGCDJo@uSqajAN5SkYH#4H3L
zrPjU<yEmalxWaW~Y<4{lYn!Hi4=b0LVw0O};{YiHU~N{npiUQ?Adpc?bxyZ^P5+uD
zmrHp*j9&sjVmzjhAhY5Uwel9btcQr<Z*Aww2Ky&<KH-*2wc8pTO1%?gN{yao%pOEu
z53l7tp3pM5_;TM6g~qtIO0K3>Egr_48I~@P4?Od@X#~IPRmepqJk-WIDX<7$7$!=T
zMCVuZ;!YWjvEx`JG?2`81q#}=RPF3e#a1=dYnvKl`dF1Z8H~8-i^YTixQckDw)U!l
zA3;zp_l=u`@e(4_PM-YG19XVm5RMBc-e0s>VV{-`s6@+Oqkn24SyGi!%YAoT*i7}%
zSIsJnr(Wg?L%4(CW*<C7CD$;*pKH?m^i<1!gCH~Scz{4Kp;3u63Z2F2T5Kufhp}R@
zmNWqCT1~Yhvdhbr>iJ{rL<2w{8BI~4*O^l?pf!cCv^@<tgP80?5<s@8ktrwEzD8(#
zbPkn>YYqF`3F1kWP9i}Us2?RR5xvtuRwL#DtD1o(bJ`767N_gjMD_>}0hTQT;nUD`
zRHlVL2$+Hq@c`$rFK{b$YEwk4jox#~Foux%V$r0Y>+%q!7wc3=hCCOq$1WJqDVd99
zd82^1Y?Mt)2AvYisn3UXphrZLS|6@UznT`;PK;<9l*(r}FAX^F;QyY=wCu+d7%=mo
z$&zVm<a=@+%V9_x|7+%&Qbo1*l1QfGhzQQ)iN--}mK%K#E9VHIEGc7@jWNq#IB}Vl
zv5Y5mb7z?Cz<T7HZ40|&6tmm3z>vix&1iiw%ex?Vz;0#flAdS@#1`Mh#y-{Ryd+8!
zy@r9-%<2#$%??lN09}lP%j?Gsi2b>XE<@6BsvX{N;J__R3blqw!#>bQBzBkDS5SP;
zOCr|N0r5!M2#;i<u`x!P@710-m1ttf+aNq<0NsxgU0fUt{*p;G@dAk)r_dw-AX0;o
zyC3eI!hAO<zh2u9DTtdzH3&@V`6){vU>hK`+l?4d*B$RqKLqsozh(;zXHS?b2ztIt
zx)EOpstgH#u2idMMM=ltT8#tUkKH0re>y;0qSc&QhMR7xYpBeH+vA9RXzVd@Ef@@9
z8kX7Aq%*x{RT_LM)%q$E#D12;;kXNLJYI$pQYc=MA>hx_&?qvyPUS%2`^8xNx85~j
zqy)_6f#`WV5k;9c#=5H2uxF}-8G2~~e1x+x-s!scdUxJ&VKvjT*~`2MV?$~HEe69P
z;R3H~@kRp|7j~)*F$2mB5b8;Np@!-J$%51Tb9)7(2jgjbQCD9+LwvGi8D8J1FgfE{
ze(i)o@uU$%Y0i!eoi7Wr`9^Pg*%#bf8UsbbBl_|M`xaR?gd^nIj~MOgD=Rfdew<n}
z5O$r85%0m?bg(Mrkid1pZiN;w+IdssddRavVyJzZrqO%O+7~9t<)R`<IoKx~C$!rb
zX7k9*V-%&V+smo3u1sQ}-vkCP>fGC&SeceF(K<LO>kchoPQ#uPv<AL7HFY{bn=F&M
zj%(=QJ3-dXg(8@m=A5Zl#LFxyRZ>j?j`^qzI!JPmMig3sH5!RISlrSXd<6nF+vZWC
zr6R=_i{+dV>MZk)!*H&aP6hxm*4*oK{!S!<qZP0@%f?u}BP^zi#NavL1t_qzo1)G^
zml>rpZ5KTvQ7xaqCZ*l8T6fAIm>zH_#X!nI$5;~C{T7vQ6R`45_<Atq1|r_n=rn7!
zmC{~VtdmeQs8p&wjr`4E^KmpR7Q6(>xWy^sr+QG000jrsOGH=OIxZ;am~z?T;iV>Y
zDf7E7sVrkG-Prrw#Ed2WkVtwKpsry!?{Bkv&IUX6D3Qc*b7qnjB&m&_2^P1c8C--`
z%!}owEEy(d^6IrlE}4)11|AfZ0@RH@CSpplpWeroZ}CQrT)S%g(>DZA=&(atgGM6i
z1}!9;p)fJ<xM&O}8|TD@pE9DEJRmd!gYPRyUOsDXj6QZKXwO#Y*g;^}Rd{Mgh7q~O
zB@)3b0G^sD{1sTPewm&u_<~glR|j&|U`=D64vZ`H4Cjbny(B=&st2sGSae^aC9fYe
z*}x?5rkRtI^qwlEgg2@80kcB9*fnE~M8uYNTD61CEE`IZj>{*(WC{RY$hlW+?!w0I
z4IB^X9K_4HJ+(T6hgL7#Xv7yVt$E{&ofegPLow73Y0T9Y8CtzBUd$VCj<yXMmun2h
za9#i00-E6?8oO&K;+1WP@M-{HbE1u2YorhebLywm2(TU7?o!~s##WJ3qD}AxwJ=Wq
zWH_SPp%g*gRVZ9LoRisP=v_Eg%#LjnGMf~O*ArVI#eX+k9P483>Pn_5gf$X0>=?Zk
zoV#ZcL?eT13GS7e)HiL6Kp2z5L?iPL5Jbs1V2XjQ7`JZi7_yC10Ib;L`2X9rI&f{S
zfoa_@?Uy!^$j|_&QOG3Op_{Zd46c~hH~7w#0Ow0HMuc&Y-)!ds#iG+QP{M;cAjU+A
z_e6c#T9o8;$V<|xkogLTE*SIB>K25~7soiD3Hb?XxL4&Op9CMppu9iDfUeu>4qErQ
zjxw*EpqLS$vyQvbw-mH>JFf%Yp27FMYGj5nPZn=9wPzV@GMgHK&&|N_9D~}L9SyGm
z>SZl(znZQZYHP9BF1VXINe?a<pV6E9nXVdfbKu20a~ny~12UelvoSH)V27WC`j`t)
z-&x#ZQM6QFg-i?hE~jS25917=tH3=lkxg;HjB8+$%n`ULD9(a_W8gD$wN6LMH~Mfp
z^YY3@O|B|LjIKJh<8+<iA#~Hgog%fVNkDJ<eL-0hIC>Ee-c2SP57p7dX*EE{`L1lE
z2|J?1DjiBuq0CSyhx>@KbyaG$IhkfKnSikpWka6X=<PS%YDkf5XgD<dg`}_*re&i=
z^m}Txu^+4<9kCx`Jc32)3^D1m>gTzkchO7=)h7xZF2L4&!GMyB<E#cu#!x309$wB_
zH_DEUMay~y<wHoL9Q-EpK^L}i*55D@^Aas}nH5m{0^v|`TMQ}SWR<^@Hg!wY20QUX
zG8|Yv&SF5J61_9Zg*RJnwB)pQ$R_{G(1EsjDw;?Dk+~1A#f%lOd$fiRz03z7wNEnf
z@q^q-jY^{n(9*mCozSPo2U#>p6$sV!w|%PQgV1}OJgJ{szyvSW_r5|OAnLFU(E<D4
zE<I8?tE+*m9(Ni<f2+TZLPcB2uJv*fjT7zFR0<u<^D5*$^Tg7QF_?L<wOldhxk7K}
z3eqT@#i+0Z3-$gz&B@f_NkctsNWaqLOuq_#Hj`bWQHW3NIA!!Sksc_tpJtKZ20>ml
z?P{?3rf^c$nNp0|x9pG`>^9dTiu`2FEDevUKjyuZSfeg!Pn*iSj|(|Mc3S&{hU)LN
zG8~U}v8*81Y!eR3EN;Q7X#6T^h2FOZF7p6tc@KOuMEK01)K@VG)xBCidJ;6Zs9!4>
zlMKFWC;HY2*jlRRyW)*{Uo2CkTrdVAXDJ-jfcpSBRuEG*)OV6n{EiFjp>ao{rO>E4
z^No5s+q@hSo}@qF4|>7fu!I7IZT<ffUR5*fugwKK$k|CxBI$wX#R;(4$sEj%DbT71
zVAHM&TwGjE;!ocZ+KE0Uvc9(8mw({e5(>JITTQDn@X;EXOlV&>SF1}>CsF8S#RtL9
z@SN=)7}Nca+CUJp9!1f#Jk$HJ$0Z4NTK&6SZ<s_E-qcLOkUwD;#Jv~*T{b)q&rnSQ
zspKoN$pN)a^(q|2*m9W<o!lJ2Yicg`gK))k)g-?7%r?Xz5VFg~<6v4^hncZY{Kv6<
zvog=_bDw1O&Dz8nT6nwXSg}+p#mY`!vr_wzfhLx;>UaQ@+3MNMN-w8z@MV)llS=n-
z=kS1CH1Nw}d!{J~1W4Lm8eZF>(3-f!86}~eB%IIzm!nM@+|Q@|nqYRvAcTA{Lm@~E
zrcvNEfe9C%N#It>6gnsq+nH}P(D~09B+JSeHH1s$yeCZ5h3^3N)r3ZN!DC&%I4+#f
z8X0(uWE_l3D>k65A$2GgPYZD(Ch0<uPPjIZotQ!+E5Q7zX1x;&6aCC}hOkx3N5Y-y
zr`n0|n~*P+RDul!g$SF0ozcQ&WR|hn?Ltfe8<WdN8$qD{D;{=Yzwl!|ix4&&4xEr+
z=#ywP)j%ucReG6g`DkXjNUO_!p9`s1$fp@L!K_niDrz0W9+@yFCu<qLa6T!CmRg8l
zpD=%|N+UCgSosU&`MV!-V%{m0lf;3U6d&Ky(YM%WG7_gV=%AkTUnYeG@wTND{GE}0
zwN8m<$`(L$hzdp}STBUc$>mQwMT5@u>ZWW;Z~z&=P!+ut*@Q|qF~nF1hJkE|0B8Ea
z)Z~IuRtg+}o-e@GfdayT^!)Kel727WY?Pg5W3r(bjeLIZfjWru%VJ)eT%us(waGe2
z%7EP|hJ=gD`$PfSguPpJ5RDxfn4x}fpmK%}R|06evDkZkaEQH+Mxku$f?+k4iA$#7
zZg(aeZ&j(Fr{lvDAFF-EF~f}Zaxgm>yJu3zF6#rAu*|97s*NOoP|jGWux8gQtncq{
zBa*2xor7BZ$%o*-6Yt3=OsBLCP&`_pa7@Dh&TP=^zGdcBouZ~j+c+S#$9WJw0J$|f
zqh6Y%4d%CK8$cFPxV<5+(;#)+2q(#)EF~4pvlc`lS**zgsgW7IDHO<5;rN6UvKBK5
zYS01wkn;$@i9}clJ0+3S$lNTQ()cWIv(wI^r5ft^n_9wR_Zv}Sm&I*rG8^krpU<((
z)|&(6f((7GNJ-ru*G!8Xl?xAoK_o(`@La*+JeQ?Z83ho52qv*)!I)a1290Lk`A(*0
zbZ{%>Qc<7!<13a;gK7#Eb8m25hOSOOFX@@k#@U*X31eZ;2XOsrokRmJL=Nm0I8eU^
zh|n}hPq;x=#!btH<j(h{PC7Lg=@qj%JzyUC1_!mZP~d~@8N_Q88of8;{n-|)!2Q$e
z5qYfO1w?rqEK`Eu><-XQ96Rw7ZxeChF!649r)e|%7T!j@nzm7o0KeymTneKmh7`Dg
z2=ic-X{Y<7Ihiqmbmkd~+3SXh08Q-zDQ8F*wM`4rRLCD8;>=FN=@Sy3WKxE60GNwY
z0UqfjBO(#fbTpkQBC)a94`{xhHdWM6bX4zo(*XjhM%6_HXk_qRn+@pG2|cm{C%~IQ
z)tXq-2Z?L;Z)Kapg*8(%1ig)9R*Im@PY^L1t@fK3vwnn77r*-TY;xv>)2x9(FC^|B
zm|?gS-=$sKrH9@i{F$e%74%3!{_yd58mLWEh*UbcWM7p}nNfnxU=3>m7%@E}cE4)V
zvj0nJ_J7MiS=uUxj#59C%9QJk$m<N&fymwrbrP`d?IK-0%}tj`py45$Oj~5`N0m+*
zPI|4@;&}Z@t5#xcy9|EZ2eplV0#^pM-}(@CA>GSCDg&XOuld9mInn!qc`NU?Wp_(Y
zeYy4J7Mgzka|?%}LXv~~&gumY<Ic2nddugoBgMDVtKv&R`Swm6NMX^hCWu50eB_w8
zv~<mIfp?OW+d7#yl%jVi?+t6;cAEE@-g@Q3ibX$Mnrga!Kg>O3#SRa0bIqrXdQyXb
zhlg|fF)8Wul7d^@=mYi?b|OyjNtVf=n?vRP@^DXL@HF{~gk|r<72BymrWMo>)l{gE
zHzpMfYCTW+x=!dlxd;RO%{-XOFH4iEm2+B&Ur2F@Ovz<>Ow4JuiIWnET=72h-~=4!
z$fW{gpDjT{p&Ot*ifFc=$O#t+Uob7qydtFn22HP;Ce_X@@+m?sApc_UHe;oOrDT_L
zsmc#dKvx)@mVn0CJ+c5U8;1$D1eY|<K9?Y!kO^zp0H|%|R0OlJ27?OleSNP9h++Vj
zW`GObG{J8EV0F3)E;$pg=UtdT%$_vRbOuQ;JXSI&>=gD$#bdBsJ2?|L(B(0?NoK?B
z@*aJ$fuaFqV({$YDUgVPFMwhrP>T-^d~GfLkCW$tkAJECVfPZ}H7<dEEto&O&Tb1g
z`WHN(=sw`*l%6NSbxHcC28SYUbc!5dq78G4kjOEOO0HIQ*#m?5yUs0_oUIgnO&)H0
z1Fd$ZXT2<3LC9A+I3Q75xb@EU1GcU>8X+;cn|z<G%X>u%JR00^_1yL|2`l}b7Iu21
zTG8K%wry_wr@-x}#dK%CoXLFaR1#r=UH#EsLJO%zsxgWW-u-)J2QTk!L-fzXsgmf}
zXF<C4g`-@z4aLId*Wu6C7DR0!Hz@3&r7fbsE3!w9r9o>)=itzsR`~m-F#`jkKk|wk
zXUC|(x9FXC2fw4zBtusb?c_*tj~(nbVz;@qutab+CT8WU#-hzRzjXh8y!4x`GY?j;
zS@|~Sdz(MpHs_pO`gJ2aJ25aQKRvJCJD`R)DA#^+Hz1f=rIYp%CS(A;3@(s}+hbzR
z$2>l_JLL83tK6F({d_M&S8}xf;M={2w{M*}yJ)LU<^0RQ<ef*G76Y5^N(SV|v4i|<
z{5mV9d1mzxBXod_%dmrqLbSe91|@n%FLKXl9t<?wJWg!XZy=nh!of)sU28T{JZ9Qk
zDHGy(=E2Fj8IwEC%b%%)k9$-qnM!*yu@(GeiPkhClbemOR{EMij<82_i+shxG4>r+
z%Y@p`EKz8~WG}OIRp_NnJ<}uxqY8eN!c~F8rJnsR%NjWs;Ma|Y$q&37#^k`$6=SRW
z;KU(KK(k6UDI1srM<kXTOf)?_3g<Q9;=yT}nOCGz38T7tirH{+kN#h?br$J56U`K1
zd2s$P%O3qg2kIhy!IcAy7>T}0chVwHYtVs2$t0)7jyc`F|8Up9deUw0sQZV4h8LMi
zR=aIjMB?nv*ZS<OyY^?RU4qJvoFtKJlCpy+qZNN=`X06VIOoisil7ZSp5Fux1nvAG
zruia1nsd>?-ND`s<gd#^riGkWy=3y#ctc`<Hfe@07m6Am#CzQtb6VTB{lE`D)c34*
zJXYBr==3haak-zuvF5XzUv6z~;$`33O(2Z3CjlP)^hh9JyT&v(re_(7{JWUH0*9**
zkBvoMAqRY){0VwO3$=pZ%(o+QOrll+7Cbu{?-WX!4I8~8p@B{>eYMd(Z$_^YmSd?5
z+R4NQ5+Y`Dn9@NR#>+PwVc!|W$~8<PP5J(aaF)>sQ&VtsY_?d8V`BCD=fG%Y;}3ZD
zQr&#Kd$~Jw-@7E+fSIHUPCiuBNMrZpj@`3bD5<^td_iK*$fAsZms<`7_?e0^XR?lX
zh@hT-<p7ft>*7+w<4okAicbrs86nxgTxLrEF6!QJoTdj1dcb*4u90Ye``|>RLMj!j
zRZ>?oTUQlfa6)Y~7Wacn#H(o%rG{qsur9l&D&WiY=Ol{@89O7de0Dg%FISW~l6~_N
zcjTaBHoydOQoUB2B5+(`H+^&Z_iy_+e%(hu83<_3ypa=T?`|(`!eDUM5oiX*2DE-`
zfN=^ovVkr}*{Iz{uS4)1MpN%buLHg&F+W8qVfpa)fdXRSO`4O5LSC&N#0vg75-bo^
zXY4lP=4t3Xg5Tzkuzv+8ZE6x$-Z;0&Q`#r2ht~fj4G1Z(snaHopi+&$GY=*yxng&l
zghx~Ag}>f-kv`=aRsY%2<1erL!;Rhs$gsIL{rbK{3JapccRUFxLH%$4)#Cr+Na5|<
zwZ*(+j0q1zT{r2&F6Z5Qt^8@Bc5%%@Q2#&|$8sv~w#16C<Yt<Y-|k^MYhlzzxWMl~
zI}aOZeCXMs^CuoT6hIMW_EILdY3G0)(#drIO;1qM^zfOW&kW8)rP6>j0C}cLC_#S%
zosT6;6&kQAcZz&l)EbSBT5~eaCQcZe3;@dp;pFJchuyzr=KuL|gfp#bYAar^KD{a<
zk4<X`Kwib@j10Q5#l0a|Ul22qb}Gy_R8lfnvHeT(POsG=x7eJltTz`G%E3ncwkMG}
z(^rnRwOYC36t2CK07_LT3<h;b9>gCDZF7bGIeC;vb>z4v$!JK>nk;?*7eG3yRjcGa
zC(j>1V&u>Z>njd3KlMwb-EAw?jr{508-Sj=h4(o<{m^}P`wf3XV`V-esbouFV7}J@
zMBoy|nDD8T@)o&{2dJ{Ha@;=6t9ZTO$hOr#S^jbE(yFYCy`iPmhungG?7Mh$>BV!o
zOO=~rpi!~sgE3*}^TN`|rE^;bzdaq*7sNW6?b+Nj_CrC?mkC$qUqOzoKjRkWS55wX
z3#dg~$c9&tpK!q#ieqt(Ryk*}AB`cv#ElC^*oko}NQ4nnzzwK_8sFgftgy662Vn-o
zZtqF2NZu2Vf6W$JoTrPkO#9v|OlY?ei^t&Pri~Eo0Koy9i@IR3KKL@PNEVDQOr*$g
zEj*+ek*DDpT65BNu>r^M`cp^07sT(Knu$E0Yrp6-|7BN0uIP@_s;<BG+v}XK&S7Um
z0gx;yLQIhwoYx~fq&99IT#_iPk%EdIC#L3ljVFutKRKjX8uyt8W1hziWf1$W4#te6
zjm@==o4q1md8nWVu~TM<M&`CeZs2XW&>KA(x1AopSh8iuHxng=LEEk$tIWMMvu#CX
z;6T*rJx*AReSSKI4r(Zj4+aRcR#Pgi_-$4VF$Zm^Oc&=zwyrR$T`rPI*v;m_U_-?<
z(w;}4>V_dMCvulD*Muxz^AkJghl?TDg`6$X%!BDG^EA++fPu5pvvS=5{F@gyx2i5_
zr+5GL^&8H$&D(bnrmAh)+G<CtD>%W==cRs2&Y`t6*B2pzG67isla(d=1}BRb9sByg
zf5QH7P6+kOiL*=~?Z$7~a;V#G;yIVuF0l?)4OBi6`o-#=NbctTcKyZWT&we}de+70
z=$t**vu0LC1-VJz4F#gs`?av2Wp^fk=TL8~Wq)+KEqV9v|M})LtN2IdX9auy$~o#$
z1T*kady5VsbnYVkrdHb;3QYq>!_D_K3qVDU<!Ty?t2`nF(-eafN(0Tb%+^(x%oVc*
zg049&L`<mEBiAkRRFZt1I*I)VTs27r42s%<U2L7un0)B`#`l&PMPu&HEkc<RLoEnZ
z4>7r11qJz~`4?>0gF6gCa}<*hH9+~gGUw-8H}|`J6XBR$^2?i}+;VJVHM~zk_Lt3<
zrsr>3lpD~T-a`=aj+vk~%9FC$K99$0N8kQ<W%cUH(^IUBzklEtW@8QmT_@tcV!s(W
z<S-;T=o~OpP?eMV-cySDVn`v}H$Ygq>Hc$m&R`>V*y(Tk;r!xkmm7P}{UUKd=ll85
zjNN*znp$~E|M7{{P<w@h!yN_q6#*D;RHg|a3dRPY)lP_cLJK>i^pR7Mw}EMi!Qk`#
z6tG%p!XXMY39^~9zG_;J9mAAjKyc?rP6b%91kHmFX(r-@1tS@1YF+s_%ubm2Z@}a5
zp&>-eXY_Zr1o%E;Qzs4^tzf4iVj3-=q~_k<?MGA>UWET_Yqx!Mbl2`TVF3@KE-j6h
z;?u<gkj@V*^3ArxnKfg2AHAwsXZcT2{9j}HU;Vr{=%yQIdqK#p>^F)<Svk@Eg9GHI
zr~MXr>Lf$mJ)~1A?WZmT6Kq(Cm>hpfZu6X{|FUYyw};X`chA`MFn8?toNDKsx8>g-
z3_&hcY-xE|Ztp%mlv=WF_4xyyk}l`RU>@ZfXajIbtA+ak`kUQ8IIWV3dckdyXlV?H
zAePOW3czbrS{?M<&RHz1!7wbXdfD6pl!>FLAgBFB?6n)|?eXh5MNJ8}S(ns*I(5F;
zb0h2b!_HA#1Kd`ueL_oboYk`vYN4R%wD@rG4+C!bYiiSM+ZSyM{^h3Q(VUCTZ=|~k
z&$hpc-kW-O6Y=a8a^rhX`a{L8KNfFf{qf84qy>SwXL+xOSNjz@miYd(lYKkD0SR<&
zMV?G9P82d)#7)vpakua8o2mzzCC3k@b3*pN9^oHby|QJsGD_!op&*2Va1Lia@7Vk2
z`i9y3lUyzyQ%O6>f39x@Kd2P+gx1(HsnQzj^hsb~6DBpu2Catbtq@<Z$O8cv*Yt@-
zwHgqQJT|cBFpMB#-Wz9qCdi*ggzbX^fxHW}z*Ky4<STyCbV+8C{Q-^pt)q^d?Y(y|
zUs{?!bwTXBh3ntv;4rkQ1^7iY#%!%Z=$o1A6{%9nH32vQy!`-wlqir6#q>S0NY}oX
zGO(O>5n;Ts2{;hxsH9UkiKT<q$FkxOGR=4u79`dh$)x(`S#!(5Me38=T}6FD?8!=p
z9=V<dcU3xJ>=A&c<f^X|S4XmXO)9Cl_k%+jfG^$#A(Dr|;iN4=t1|Gf&oA=OOR!jC
zY&{Q=FSE$wqG<;)>><6?RySvH1C4Q)&|=qXES!|E?=+3chi2XBWSt&O>i|Vkg9~af
z0lRTtVyHrIBDzn5vFBc)i)%rn=w=LX+w=OD&4#Kd#4S*vKI!kOH=4xkC*BO-li)8g
zCN1&|dO!NL${8bnqwRWq;SfcqO;VbTq|5CXc&x1fs&Z7g!#MK^<-JDIn;3*g5NZLi
z3|EZxe{f<o@FGDlBa}eP((HRfN)d86S;^81aJmo+fOPlZ@NIBIYrylQ#)%CnX5h@5
z7z}3OZ=exouQx1_Y9bKlvtusIYqe6Ta1jK$@Rvd*5bU&0Ku?fdu0F|1hY<ui-Nbjo
z#o~7#loYMB_fRTzTK&e*^?te}-OMABKO%p|TxXG|)5$>F&BoIP0Ve3vLOAUp#lf!D
zH*tNhgOfGe1Ch)j)&s>=BzPloX_xm&R-RI>!j<~M?%6;^sxcU1;3dQoHSD8c3_NrU
zpV0H&W&=fuflsTFO@ec)l3_4^2D3#q=@1M--kC{kQY}7M<yEe{QTGP8P}k+L9N1aL
zZhXEPk9GD2Jyk!ZuFYQkza{<ie_JQRIXnNl;T+5pHJT$h79+JBfL}L1vms$o&Z_2{
zZ@;t|+jZD}SZP1lSjBO78%e2kB<n1;#rff~-3{<VjnU}IVJj<T+W}442U3h_H~vyR
zz8t>l-s}GD)2pK&#3SVeIXfu?b-exf5LuU>L`d<R@UJf7Ru_2`r%P%TVi;VOHkF_2
zZYTM!ZxFh-O&j!#1O2Snk^;|FD|>*Im2`-Z6t94WC2Sw9EwRDW+$<F0n(s%74(Azf
z^)r~n`wrbxNB-DQb^kxVzLjr%n=V^QxVd(N<D1eaL-fS}Ytfuum3?alDA%7>uRI@9
z5!cY>tIO+$PHBaK=+j-bq@Caz=I$Cyb?6b=F_$vkzPx|Fa>DcLx5d|j7N=s4#k`7I
z+qpwq=q2{3wY&TLKaNlPg*1d(x%vb!kzSfkqw8DZ-xZzz$3=Z8L0BWUBM11_T@~83
zSKPZiko4x<wwuda2UxG@i-!D;48C(8SxT@*7wG4*+6cRv+l+ajP*4jx>&SzbuC{0w
zwn{Z)#FAtC6+1qxawwgUjAi)D6gcK5xDbeb7Y@6P8xL8dUz9rU$<G^Rd8#z@R-72&
z>{&47*EZ&K+q;2^z!<4Wo;KRt9&y$YpY{y?R*IZw)`cvO3Xr%58Us1|q3bj2?1rDu
zA1M@V-yJ}tB@koQ-M%t(=XhM#?MPfhD(~?T*7r}>9le@<?U2vG;E0hs<12j1!<-Kd
z?6tB7uzS5HBm$A|IYn>7guw}6xs--DvB<YdBc#ZtE30bOPy2^HMG|lu6(Uch3WyuX
zjAwocIbCmCU7uH63GLwZIJ3700&@NugzZdYSFo`|^Foj1kK_S#NIpB2luHw$#KR@G
z2U?qh8$ugM{1UasL}RS0jQo4Yx~B8!?a+feYi)}TIWsJ9%L}x&fF--_M!4|a?`Brz
zUpbVo%40Ej<Iv1p@z<|k5581A{3GYH2di!bA{ke<J3aTih(x76jA^=Y2+2u*?xXii
zyHn22Zd=EBHAu2zC3=0cV%{&)AN?|;&JF0|_GZ9=5BP1NogTg){di`|ng4TE(9w#j
z=kq$JM}`T1JGE|Zx(SW1EB*ZO@<-8g|9R9HDowcaM`O`blK)Zl4ROSwWv>HpPN09w
zo}}sMfHwp5lRoDiyZAfsdpYHYrk6fT_d9)bfxG(#>$fYSbI_7s-!gaQtVptU-XC|;
z>|uQ1#pOjG_xn7|Al)f-T(tY2<&KvY?S2z919^!o9l*($N}O8O+5JAkxp2epe(vb+
z`PWtyd~(R$@mt#OpXC!6wXdT}W197rS8}%O^Eqc*$399P==`YYSIf~t$HqU7ezGBc
zQ(-~S3Ww3pw``$R#qYt8N2<T1|9Qc^a+|a0MJATG)h)2Z_S|8|oT}`^`OkyKhiHxI
zO;6@Ny=1-oW>9oNgH#yQIN;WoU>}qP2neL$riLw-5*)%d5N_lIvU*>6Gl&<ezqq@<
zV*!R2wEuW)EpNJr`_$p7=*>(H@``@7_JIFiZ|;93{d_9WKjOnRKP`kia&u8uRnm=r
zZZ69leS6U{hhNA^ZS-07+sdaezL@CrckMr@YaCXF@Y-AoSY&{4Ds-2&{A<?i->g$E
ztBn16K=|{zWlmS0lhd!dIjr3I;L9Iiu%&o$RH+~11Jc%}>}<1Xx3x;IXFijvw)o!*
zS2VH;NE`ncAA!qNl}$uo69Th=sBmC^(xXc~_rAE;RB){DN~`fshI>%!zh<xh%F4uI
zoyGe;{5xme4pGQjec;#m2_|*unn~p($2lt8N@nlo5ON{>`qyjQa>?fwPn`0&rD%T6
zr)9k@`01+wdTsjKm^UHTZqD`6=Jakr?fTd3H`9&pUxe@fF~CD|g;JZo)jsg%{gI`P
zjXBY@i_@^1A)Hjb=R$Js#rFLRoU6&)&A;;~*7iFuz6%3P&Vqqxa%PvyoDWNX<gJMO
zWx?4JsZH*f>!N<bYgP{B$%w;sR!-v|&JjP-a>%)*v|NWy{QYEO4L+uwo7LOCXSZf|
zAuG4)>x(Z7v%lT?-GxgP1tE(Vw>T{$z-zQOHHubM$?^cdgL<~*(8vGmDVH8Sy)S1$
zlze^S%jnX*A6oBM4#y)n(bb_%BJXtGd&YgxIpBxOZ%(iObM=Zv-9c};=NB<|H`aV=
zuz~G$UswHT&zmAq*Q@v?-e;GdBVD{-?Ra?RqT9~uAP4)v{4aOrasq>35N!qd%$x3A
zo$JQu-n%z3aC`HK`@=b24+uB(R4e>SdRBT?4b7*q>Wgso<rP0KylTIqw(nwy2>qbj
zP5P2tWdvyaJ5TYP;3mw?a^q7*i0A6u!7KZxOn-lUG0N`{@|NS5vE?+nAm{ghMp#*v
z`K8|juQ}ATdIb%v<}BTIcxRxaA3bhuVaw+Ex~Xj#64N<NX5TgM`*!cmo2Ty_v0e4o
zeWoVkBl6oQa<w11QD^NwK+cJZks!2m)X6$w)0)-KFDJZf-1BYu9_#f<*K<p=v%e$*
zh?})TP~)Y>s+=kxll_##3c}KqPpdehIX{?ve)I2KP0`BR*WZwo#_)jkSN=fA%{iAA
zrB^SQxw+aeDtb$f!$9G(>7<bOpvs%yB!4t=<#m^6^S1K*Rl4jFes45(+xeF+TGRB6
zjm|fBcwG9rnR#RET9a)+=gS?{l7kAz%m>?MwV1!C!*Am^9e3l`M!!n99I#;RhCe?2
zuK8li2z_(aBF2rHPpk6Fu+$oFGiuOMRrs#F&sY3wb|Ssdx%}?T+na}vZYd00M9FCy
z4Nk8PbVFA!JP{^4n)mQ-))S6n_s%P)ALo4b@W;c>K}ZH6f4O7fz|91=p_mrOD_N`E
zz771(xpDISG5+7azcO#{oS#ciUw?DnMF*fBns-s!h<AKa7kZ(MnOwakaOI-W96#Dj
zkiGpT>qDvy`Hq<{_I;Bw9%Y?DeskALI~ul1wk)e^Dp*hbBwwuzED4$^4VYel(CRyO
z26OY>){M9R!xsi__le5+ZsDbPx3GY0;@c~`&JwKG=c^p^6Stx^=V=`~$J&;*FTz*+
zt-cWGmY7qaba3C?oYgqvxHpw<k6Jn`n_RIWJLsITVCd?~oo9}xr_WqU^vrgAORM?x
zxgYH^{f2*6j{7=j4Epft?r)ylWIG=@|I?eM%;SfLr~cU*7e$_N%-OKiF_U9IKu>!&
zG`GcVViRqQ)j+%RZbeYguBa_H^H){Hy9MR)e`%gM<Q7QCw%%a9???O6<-GxA`3I`D
ze|30|a`S>u;9@wq1GG80>dICF4d}rxShg?UUEWgkY-0VIx~RahD&-#D^haC#f`%f(
zI$t^-9@v}VFtE?(r?gl}dqDJx1m`VBKiTH|W%ZgpiS2!WUbl2D28g>&6;+Ii)3PHq
z5oezF2M50Twx4lh=Fl(o8>|jb5)uEF5YC|ljoD9De>C0+Pv7zL)l|OfpyQtv#iN0T
zf?kqEik!@fD(ge}+3|B2Cwb1}H`hycPd9MKvKxN~QULos%E00qK@MSOx6qg5Y|GA%
zADY{ahUL^p4cT`LREt;C*tEV}8yGKb86u}YUs)LDm(I>!Wc6;HX>QN=_sOHYffc^H
zF6YzNl1}C*zY59bAiuzia}eBR-zVYo#!pjM%t;F#_#-pkuYHkU@ckR#ZawOgbK2#$
zljWl$Qou~Acw<LSsJ!*y_s@T=&ueG{B^DjE<-wO;euoEcl&oFPv4^}_$7ZLB1Y$(z
z1iM_?<r25d>;qxwL*9=&4?N@M5hNvkepvh8tHZv3v27>&!Ir4Ki&f68Zf<jA``-3^
zpVQU*&@+46&3_)=x8GT7lcl?%-<Biz>4vH15iJ#lIW5_QaXo68Xt}0kEa`G;yWg>G
zLI0<{cMVH2d&9-;G-gd1YU9|H6yT`?7T9zceJR5OXqkudSAkZel9nCROj6KkEHO<4
zEy7AlF|#H+=wu$!(n<`9L@f^$sUT{Q3W_4=?B)FSzV?5A+h6zfr>^->PWp1cYpv&5
z&vQTbeQdIA9iF^27FQydpp%4!LT&<8W3g6rM$Cv7dc>h>&qFyUKl`I3VV7Yss5^{G
z9uy9aE-V(i-n#ARlU=u39$#d<v17S-2m9R4_mcQ)dFU_X8_)9R%!>Zc{(Ud2PT$^J
zkrzn*;c%S)dxtBjqgF{0AAgI=dxLw#4_lg{urS*d88*x9ZFu%gwBN<IO65lja{q3L
zZrpN57kAdlC&H^s(nsSLLfg=~9Q}o77%TGcw<Ts~(7S7Qx5GEa3|pTk%0gaPM=n`N
zW$fEnV7Xy$_`C|}{%y0z^7l|H55#_OieB=Jpk3gzI4f;;xlU!Uiyv(bkmbgrR=zcJ
zClF2@!TKTW9CHasFjoL?vtgO?1vL}7FXI(2qc8k&#mD`9VZfF0C*7z@l-`UGAkeWh
zo+o9zv{jW*H@Gw-9_4KR^_<^E4uogRl1t^C`5r2*dgSNCfimKcc5hDOf>Nzpsbv2c
zlnn>+Xi;Jjkca5`0a$m~-aA|Fm2Ex}J9zbwV?WC}<M}x#1hHumx%jIKYl-=x9fvlw
zH9dILS9$73;r;%&voCMqJ14y++tw715dyr^y@OLO+0W5cm0$I%I1t1RYWy+%I5#f6
zz~4rV1`wmB{JAW40G?EN)qLO9#e^Raedur8o9v^}<uER=hw_XbrGm-aW^@AvLtv$G
z*9pBg^q2kqhH9}t65ZrtR~RS=cDYYPo?xB!YNEVHS$h?AhT-Q(_xOJAUkjBs?nH6G
zPTQLGiid7M_)qPFaWF&>`vd~h+n4Q=?zO)^1U)(Y8vQu(<~4}rkaN+!iU8bv;`XDx
zM3+31GBY#hNlfNwPA}fv`CiggVU>egE9BaRATcB9w-4Lf{|+iI&z9mvzW5QSY20mL
z5yc|N^K)LYM03Z^VmBN8$2wF2?#=)$*jw+dgIS-5)GQzQX|SO3cGl+fUM@F3VO%8U
zz>kUa)rPjSZ;sx!S43VNKLzW`r!s{M?Ko+Ng65MT&6idB?>nh`I>?sfvisLdew>U#
z_VY3d*w}#>aL{5flUt)l{Yi~SZrdjSXK-wzR5eZ?7q-P5Ed`^BeWjnk=<x(CaAbyP
z`0I*CCqJD_qdwiW=Gq_>aUn-^8EmB>aLMx$HgvSF&WnEO(ZuZN+04jl0Tw#8WY*5$
z#ID<=z)D0TeUns6%WuqCGyt7>`)$QW*B4v8PwJMX<E5C97%WjB=CXjQ#qg4Xs>re2
zQ!p5I`Acf@(TSJ!Rjcnte|#I*=I6RcmG${qj(linhz^rJ&g=Y7IBhA5`^=7j4)aWE
zvJ+KWQxNhoA{0gu1@HOw4)<=-Q<Wc%xqNEcUppSSGx%s$Uc~RxU`<Z~uoC7KqIaF&
z01Mw-+<AuM@y??|%(ciU;~|1Em|#xls-6BGai;@ezJSBb>w?26@R-QF{!*vXvjqu5
z(z-^V;n3E2gc@&=?|$u)Rb}lE`U{9T6am<W3_!Q`U*^paSG@A5LL82|r{6GPn2P%T
zG3eAzcqflXNrt9FLOy<#t=YTeM8T=5@~OZ}PN+y1?J`CsCxyWrduFM&N%(eA*@pGP
zdtJxt>9%Cukiu18c;3=uF!|_*SSgp;{e1|ERp1s!gw$K*4<5V{hJTFpNq3&*SQD<a
zv9XEgJ^B5je8tjJ%KTw@MJU--|9kn=31Tee_OaW$BZhX$H?07US4XPoma^PCI=j{A
z7Bv?bqr{H-yZB>4vu>kbUAcB5wyW;@(1wPOq~IVtkk;Mc=hpL^{%K#rWN(*Z&fEXI
zJd3z-_3vxqRcqY+^5$0&2z(tsgT<Qg>43;L?p)jR>Xen&k^JevFRuDCD#Ia5Ysb4>
zpiEh!R=k054a_8cz9dEgD}nVEz0PfBe~6Mr`j4i22)XdXRQMrynriVfMEcO4KugQz
zUUfW2HSdJ&t^=1zS&_H%BYU5PNRRo!>)zlpTK6v37Ums(n08>3Rbc?Lpp>OXfQve6
zlzOk?HKk{jIDTDQ@WwqZVweiFpe<GMS5ZfAI~3%XPRB}_59*S6o9Z%4Rv&v87#aDF
z%O*I8F1co~CV+$i2Ga+f?YYY9$p4*Lp;DO_q~ARZ<~*lrAIOKOi1P;*B3<02$7?sl
zx;V+J%2}+~fU}F96lyJwJC|{g0c_83QRVqBExz@cLNk|NUq)n0b`nwi^tpt-2#PQ-
zv&u6fNhGe1m@~HHex>Nz>58q8?K#O~Kb$F)&I4C+4u_ReuJA1@d$=e1z_T}9_F)D#
zYP7cak*6mgocEJXi0!fdV8X!LG_$Alk#56f`$O%|X|#Ex?5IGREg1p^O=D@20oc8~
zt@HiikxR^@*~H7XbB@>3j@|b|RUJWK1FSs5LW3@=2|U0OqVGEBwjbP@_@9`L_X<{4
zt-I5NuUtTg*%(W(U;=49C;QxZF3lfilXfjjlwTYX+`jg641&=WC#doyqtp4Vq^gkF
z`+(7-9W0&|J4@Fge#CLT8?|@mI=ji59Qj*&kGDxx-u@Q9pz(zSo05nlXiw7f1okh_
zZk^(ycYU9T3ii;p=3ocl2n69yCBLmRQkIm#&R}U_jlA5cdL?_{x^?~`+Pu}6Pm$rk
zS?I}8Vg}bH7TusiIy?Q$3j4ULt}Kh;{~42%`#0hG(6-6)HJt5p@#oMItiJ=tTgYj%
z#?0z#TgvzOc+>ZKW%R1cYsv-er~^Ae!i&3y@bfG1BwO@(dU|!Q*OackeLZ#W9_y)r
zcjVunS>3yfL}Czj@h~+y%AZ|j2h?=uUOij+9^(xoUA<T4&)IhAG10u>XfR&4>>4UQ
zj-i`K{4s;K-geBqliF<7e?PbB_^Ka<_D6!-RZv7x+!j=>S2GeJ^XdX*hoASKaveXe
zjk|LA*I2UGppwg7$)<%y`SGF{=_`PM?;&mshOlH|ZLLQO+bU!J&PHb29Z8S%TTcZ6
zl?wdKU~yX`DB=b|Mu+)p$)c_-H~Xa}9$8li_{Ja~Okkw{eLQUr?yih{1BsM>MN>pS
ze@c%zcs~|sKLx1SJ^=wa{i91tCEG2MR)dO^kc0LN7l_<mHaVu*3@lXHXGfv5nt&Pt
z8cz|4%U;N5MS5gpe`5Eoezp7OfRC-1)MC~sy=e6$dt)4x6PYLGlWS_``nx73^!B*k
z?f9@>v2;&a>Le&-?`8w9yFOBjs6}wep3DnkyKrN$!I%xmk?O}ce)zJXT$5&XmB@iP
zd3h+7ek?*ox#S7+(w9#<wByOwMU`nx#811Gl(oUG?fV0YS&Qc8F54X%L~v0DcsJRN
z;foygzLTzk6w;o6<z;u>`RubFmjF4>L^Cfziv)qLA3%X(MD-Rhcerk~hI_H;EK*t0
z?!4v0>2&A2%07ILc>%sG7~`;nd<Tmr2__?~m66AMPW+YCvzKRM{f$klX(9}>Ca4ok
zmKr<jMCflPqkRuCqATZ}%DoI7m6fyEVclx1e9RS009_aJFtRzV&86y%iO1^Cbu~AA
z%)EYzW{C~F5GC`F1!EBX{L7QyBl~8nQ)G=uw&L+s@o1}Or(Ygnp0J=aN!-v$k;!I%
zJfIzKfgi3S-gM6@y_gSLF|@mgU~hpqPx!YPfkuGkiFpUjRUB2Mvap`FDzM;9e^G@m
zqxHNKkw}Ff*1^<v7FA>mSwt8qvq3AMh|Z&*_SW9+z??o&o}acUeZesSlkW-634}lA
z8x>p;uo&}2ryd`*Z#$UoWmbFf`e6&I-BEX+JTvk|Oh6`+<1G;7k_6xm{<QHJ@6m<Q
zL0Js)PWIod7zc|2aC57KS%}!BEbE66cuMoir~j6_|C69RUaS7_Rzi3hb@CbxajETB
zBGl3sCLuTmYx$&%hbbZ7vKuH(jh)8~8K=x3=^-LgHSs**gJsTwU)YtWZ{zdtI}LCC
zTZ^h9^YL_B+Ozz{;3n}?%~?ZMbLsaOQTJ)y-k0%0mPH~p7M;$rOAc%vh=r1a!tgW`
zTgiobQdWN&jwQTMfIOMOdYaGTgUW<8VKsQgj9f1<E5YQf$No>7!ozJYuO>XzJo@^W
zzw-TyoX;6(^3pXorPF)hXu|$VHFBe=0t0ykwzlJb|Cqx><%i{x_nQ%>uY~b$xiZ<c
zC9YsQ%iBG-Id3Xnb#=w30<2#^nwjlJCva`ypp#1j7#UrvSx3Fnv&nXWyIy|96heLB
zk7YTsB!;+f%#)<!Yy{rA?2JKgw@WD9J=W^dmy<EbWtYrjWJifX_=0^b3*YJc3%9G$
z(cEPj;H}$EzX@A7-)%?M8Qk9EP`$;UKA*0Xs{?Y$sD$c(uq1}z!kp(<-r<$O(eUec
zqKgZ9cXv#lw<LzTXk=b*$%{Q{*x|SgcB1b&`5$XLZ%2ex>})u`_4?^oF#~HQMX;D*
z2Yx9&cVY{-`1x$<p?!<gR}7I(_b(7#Zh|T}s8LEDlsH&LLP@QcS+RYdAxQJ@vrCbo
z_O)SS#PD_Nq7X#01$Gq@!t}%Ilb5ak1WtKS1gKcrAXJ|(VYU{my!V!Xu3?VtIC9CJ
zBD_t$t8#{MoM&m2<~^mJ?xOjZUOxzLZ;C{<+(4&aOEO$2Q}jApL*bKBlZE^o$b(ws
zc{TQPG@gi#@<GRCvguHApa7_ib#%^tpS^Eh*yQW517~)`gt_>jo}BCZlbu>(I?1&e
z%`WrvVI}CPqrT^+XC3?8{?_O1-aRa<#!1)D7{VAHw4p3OZ47@YYF`mP7fe11Ty2(e
zm;aFXs5&tBiRe1@PrueX41ECGWlf$B58KZUN~*L*N8Gd>+;~yZ>gM$LSJ|pGNdW@m
z0(C`%9U0&Y1k*)?J|dzIpQQVF;o-TTF1*=#urXm$ed?9q`k<NH0nadR_p?qYjm(pF
zy-g&nC+SQtFFmvIr;(d)uaq?RS<L@X&=*#*C=%3#2_8)^)xi(}#f70NHS0Bx&z?E9
zkBQje6tHUThg?4JDpEt?Z|%LBT{IY|fxu1JmVA2OY&hz5KKboB`0gT`W&8IP7mD=M
zU!tHe+B|krQnf|I?629y>1lgx%uav*<C-IO^71`SD~qSS9fDi=<LO96nf5Z1c%jqo
z;gdF-a=6Rdre2rP%vKVk4P^UOuDXdI=Sf~(PxmY5+~e-}p@;C%>|JYxn4Rk(0^SW&
zIJzD?%w1xRTt&y9t@z9s?4YeV$nLSY?3nJiYPF1r^o#Np7|x4yoI}|MMy?B-ZK5~u
zt$zI_+AJo3QFY?Ct`^L%!odt4V5V-jcD}2SD9wMfj(-vLC@Su<;QN^a(S`fsV+J6Q
z2B9O4q8y4uB8H&Kq4U;?$_Wu2HjPwspbwImn}PS&Vw&%p`v6!161S3{rAqzsAbxJq
zGC!7jATBH+uu!&#CMrfRlGUhp*PfeSX^l{buH-m7aXBas1r<9h_CL7eAN>_^p0c8y
z(|3!P3M94fNHe!TVMAw#ZBq3XlgA@cx1LRFZ|;cNzOMC`*MHl&%^MAhbob4y@@OLI
z39dQKh+*~0+M^+5r_nbu!#w?+-27LO5dn#3J!!QTN@rl~A^20${`N7?b?`3d)6{It
z*<b{eu7+3|l?&VaVJ-h{loE;J{%mg3Q+?~~Z(`p)nVU$c&6iP?h5XVegumLw#^LY3
zZAx$f1;J>}K{D21<6=b7y}_Jj!V{9tqP3R1FBYmFh+X;9J=o&|F+l~7eCL0Ij;)%H
z_RR7moMbdbx$sgi0n&m_EyK?d6ve#%tFQjrhNg9|!>-xctgL(*mNUR_DvxIYjm^|d
zkajIeKAq#>`6Opw_^KE)isWK%Y9}=z=p6c7fG5Xa)FS3b2AH`D+61OmPp+K238~F<
zZs*_6asFtTp(L;p-r`E-y>cOohgd)D6uACg2Wk9@A&puUbu#<%s)ee^<=MHM!&@Bn
zuB$Q1;U$UxH}-S?M`LyAvC7R9Il<tW4gYTl`aiz<|K1<WyF;uJlV5LJF<S*jNIs_V
zDn=G0AC-=2vnXO^y1}hRP`pD$t4)3x>UZ=BxSl(Ai?RAY?`i4IiWy#0LpfEe9T`8f
z!rYX?;hi|~2xtK~GapI2mD9<=Qry8?HzyaM=~i4DYi3!GYQ`new{MH0jswyAC!Ye+
z!e;fwn&Ls;r{Q~@gPo^QlumUnMF3Ke0B~0>dzb<Eyv(xaThQW-UiJB4en&r5w|H7w
zN{TP(ap(mza_|hX%Y=Ba+Pq8aRyI-Nb|GSOF^^IUeNUWT99()RKQRBK5U5&D+J{fJ
zY(#STm1-41t!Bl7lzn*_cCdtb56lyEbwOY3f18GNE)RNn_RU?5rneg?JWtcdAN8XE
zeT*H9M54pREVT*)JVcyd<2Uz2F`dTs31A3lsJEJZ*x{Y+W1qRP$gPEwxfbBqnb-6#
zyh~8+tBdVRVzR6e|CqcLdIS!BE6&o0;m1ymXnH3M$cs;>ALkXn*{s*`8in5y(acCV
z&h!MX7x;~Pb&{suy=c(A1k|I&Qs~Qf%H8PPe@vu71oXwOV4RH&4`MzAF+D{9NS!al
zL1HVsIJZ$_Ah|SZ6;ey_1oaH_O!XlB4y7<8ngxs5J~5wBMG{tU4kgUb;4yj73rp-K
z>JV_zI)QfRhDaJI554kY!rwqHwYO2cd?5u)vFDKLbo*kJ=!impNAXM2+Yp>77KvQ^
z;9FHqVLooLYcDXmL(>$npPqtAd@-)BOc-A43(f;j4{!R44iR?_!#RS@gVU`#-3Smv
z(=>jO73(+1<w~7q+z?u^t)_T{zTFRKsDbJ4#>UIUI9CWel917fGn;Z<CFM!KYju9d
zr>DPk_d6Dz-nDZ-qw2U5ZZD~l*H-y~lPapc!={KbxICY{N{yKTh{SOeTIwm0-^0Bl
z<uxtGtlKv4vb+wIxIhcZS0G?k@X(?672>k1tnXcj9Zvo{gij>SG!o{ZPPE<ij&_c#
zxC-1kc)ShYdpk5;H9sA@L+fxFc<Y5cv}UqE7Utt(1GCXg7CJ2Yfk@oKZz5pPQP6q!
z@0mR)7oY8MXK}luwX9$H9CRAmqn&)cdSMw0lnzt<pm%Oyi-ZDxOb?EpmMSvT?Z7$o
zYF%eZ$72nETw13)N3e;vfv1&&F>Hq5x2e!G4dDG%d~#dS8usiHzQE2RELPsE5Y_TC
zYaGDQs8vH`3yAdcVwaag*fv0ecK~!q@B`-y%Xlf(3tx54ezJB{d32AM<B?==tyzdv
z$?_$#ki}u->*9{A5x_eIW$UYkgI<Htx7UTy9L|RMfEsEQ8Q6ILWAf&_MPi~KOaDWR
z+x>HvpD^q~<pK1jZcN{Clq8>2r?JQ-RVR^jENzaWqpMpZ{wQ->aCn^+ZTmka>PEkZ
zl_>UkIK)iRZ{f@|M|o)GMi4;(c_{h!FCQLXIQTkpdDZ-54B?BhkfpH*h@{gf#=HbN
zPHH8P>Ej*$q8sW&)@T{}-&up{<<H{M-jgJ`ztE8AhG0B!K&CEa!Mzy<KnJkEIlT{3
z-6d@bdQq~-0uC<v%d-m*NIO0wJ-k=h7ZDa#>bYHhmnARqCSNbQMY{+t@ymQAQ3TCN
z9lO#dUV~7y1D!Tg-Vd`^KJ@R}cYkgLC{lo{qkjoGHG+Imq2lr&uTDgw{<lOrMLO09
zPZ`smy3~|yE#&PTw4#V~Fv}7w!Wj-0mymZT7K&gYzr;bjSK_h;ps%N;ZyqxA{6_n%
z;MKa7egDZID2eH_Xzkpa%YmX0GU9fpBdTF;5h<4dbh7+oFwhA{o!0BKZRmX4L4Wqo
zr-?X27R-T@mgftk<md>6GIqWSl5XxxPSQ_{?;;uB-N;s0rp>zu%7>~bm2w!B(@wUU
z^27Ay*DK}SmQn>^{7Y~b1us9J|MRm2_SA`GOF^P@?afA*4TploIhT_yk?Y4Vw1C_1
z`}0l?Z|%c8BQdhD$UgD~Pdr_UR4=(r?MX=2j(mZDv=Vc}RGb^nx=fd33x$M8O%I;o
z2|HenYa81vhZR}9@AovKc0TlnN44hD&QDceAQyE;ki@{Z1l`il<!w!?qMEi=5vI=U
ztqp;jvDEr-?NlkPc!iExo7iak-L(=3(|qfFIA(P!a7FrD(&W)ZK1h?*{c1vj9~-!|
zge-39pCV)@)&Nl&rUrhgYKf+KzcS$3={5c=*r*SmSt6r!7OE4xtGzRIOISV(zGnDr
z1TDZSC-~0-(x($eV(yyyT>XmkV!kK-AlfwO!u#WnlTO5>xk)<pu!T>f5@wB6D*F&y
zhQ2GhR!nbD>RrE{UG4o+a@#S3^Plpy883v~ssf}^sc6CW^CCfcMu}y0@3<SlV>Mqf
z1TXTE*b8z#SpVlJ6fQwx?LbKsV#Rae-l7I^eCij1-8ZwvMfN}}pNv3aWKpWnw0iRO
zjDm=KQc7aEi?_R_1dA$iLXPk}P^4)o5+c{O17g~IL$WjPUggv>A0W@P3&@SD;$swY
z!y%~x$cg2H0|nNm2Wra{phfU<g-&~F#(1xx?rOzdB->%vX;1Re8=zH_Bf)$NI&Py@
z4a8N#csNtU$h^?`U^KJgv(k`g{$;>1W{X!JdE;LkfdR#~v&PNhV6g~vM2Iczg2y|r
zUW(Nx;13E@Z}-{=gT?e0Jw~zE!iVXkOo8Zf2k6vsO{TAwoc8Y+R2KaT{MJz!kKC?L
zx>kh?S=xjx)IcF?33dITr3F!b#{J9b)@6H(m#_mW^L$&QkbOXZHjVd(TamEb^@u-V
zlVu3#e%i(P3c;QbRC$Qo`v@)*;2tKUFVUcFz8!qZG~ZYuU~;GY9jhEVx*-;f^XIG1
zkXbPCak{IxObiURJ6#F>xm6Kee$zo^5@!|rG7Rp+vH`seHB#P<;&PCQKw`U=)~iX|
z`}b?gy@%Nn!t!)}yxjtYNLb+Gyo}02r?1J&4Ihg@QBRe$*!uef6^8($8SYcIos)eI
zg2024a>QI~XIQLCHLJJtwNJ~zWr9g3>@d8Q*R~QAEri@y1lp^|UyZ?{4-MsZ26<kF
zIW4b@v%*Kpg)e4)w}<z@q@OKE6j_k%7kXX2>0cGhmXzLgD0YrZ@lCdjl<rwt?@Zw^
zrrsTbL!r!%Zu*nB{T~QVfBZEEP4J5399mXmPw7mWzSqztSB;qEyVbs%n&QE}ZFM|5
zPyr~y?7$=gJSdgxD6Xf7`GjGMvjm7DK_JW{k7VWAn4!N!^S)*3K(NWEc-4c1zP$T_
zKPBa=dpm|E)J}WPBNm!rWsyXNP(a%V<ip-jK%?WtLrb`;h~BNp+cF`3Nv?y-fnWy^
zm+c*+W5p*ntB1N6;5b_P^7!8;nVjEG#4(#X`7w%a!2dH80S0FY)>2y6)(J$2hsO=3
zHK<sG)~87>2X-c*C*EwAc7Z{HzU1stG`6`qTCSctWVagQydXOaoD}S8%jU_VI3@>8
zv~cT9C+_~+MvZW<oJa~60x!1&pt^(j&EUqXs}r{^r>f5`c>0==BjmI`_^ZE|8_z9I
z())1LfTke=QMCq5e(aj(U1W<)M<Dj|Ag&tWU*s6=L;&?~M#e+tb3lmFAUEgJu2cPZ
zbsgHoMtS7&>(@vMJ76rZ8twH}rh@~p$j@$jNdS+Y6P4?&=xXdRG(H=g;#C-=jbfE)
z@)_~Rk;!V)U?2?+L=eEPoS2xBnkREA2Ip+IJ^_g-{oAE=-O_O_*u6<~o!i<fNA(KQ
z)$gla(zZAe{Wh*5A~1IG%YrV)b8T)VF}WIo28zMiTSJzD!~2{1uF$NUyd?XyY9x`3
zFN#qPoK7X%a;>3vQGOpA8x3A{E?ksu4d;;Ft39dK**vWWCemzZJfa5qO#hvIzaC)c
zCuM#j)<lr`g-R@75{V4_94vuep6Sq`GZ?gLuI;FVBIHg0ZS{;1Kpp5&n<nELbbw4N
zcg<%M!oK|ShTxc9uI?j1M3majnu1XdSU&>OYMc$oFHX76k6zlnoS?2v#}ie~vw&eE
zH+ER95Z8l+GLbk<f4~^sbgHTi?}r?GU(3?kuoIKXpbbW~rfdY8<{e3t={<h8|A~}9
zwz_5J?2I5-r)2PQo|6oAN}l78lJ{Ac2z3lW?OA>vyX%QT9|<UO<S0YP!4j_=Md*Ps
zBUG<MsD{(3vw+r>>oiVW*NHpuVQ2q{BqGAY&OWKe>;1`^c>+<UxKQ{)p!dxvXNfD6
zLy-8vxZbNW(thU?K&zf>T|Ofv4I~`iGxDX4U1O)dgZT~fBctQ2x9z}=g|#&&6wybv
zv~#Fv2Na(aeUfViH<#ITu{d}|+3|P8fFo#pRu}=2VEi^}5J;$A%{hU=BF*hrxF05v
z@&``KpI)|Tsz+is#zA-5B-npU1SCUom#F4YcYPM&=@oV5=`6djT<*mD-25nXi;R{^
z(!q+k){*cd)DKPDU!E$QELxkvrWGQw;NJ>`FCW0Ta0>eHZ~6qCVn!PHACp6qA2IAd
zR+JH($|yIFV9=iDVH|u5puz#}!t*d%s)rzz8MyTCJx$q-$jxu;o_2usgg&}`LchrV
z$i`T7lpj*jj9jlS>5g0ATVdHWu!}pI>lrmD5IzKLQ?UOqz_lzFWIUKz-zVpdYAdNp
z&+3+BO;6KTJp0FlbD~Uj?gG;sIiBpSq3<5*`ZhF1BO9il_KLr*0Q5~DBv;Ej`ee&J
zo?|cg(o`t3f_ZRE(qV_Q0j-^Ll2S^ua<Y74i)MraHH^R!ePO@nr(Z3o%k=f**@FN1
zv$(ydH2-eN(91@69Z1pB9`{?9ss1ss?gc5(M5B3fDjoXA<X*R-=F@{vn{)UQqKx!`
zGDfx>h}p=^qfTsrW8~d9bEqCoYHNFI<<&NJ3I<%yhsI9HM|ujbUi_B-C3Z}Ml`0L=
zR?h%HZe}#?TC2OqJCVlA;Y7xPs~onE^lFoRL>^MiQ1gJZ987#<+$s>ZC(ss1*^Jly
zv`;3gA94KHp5Gwjb8X_)&SqTA2urJ9`Vrf9qJmSwi)*wcgQ39hfZ4wZDb>+|e98y)
z%JM6{>;O{bea})7@aBt;k3rYZ2-Vp;R%18Dsu*ewgO}FO)m2yh>F#GYm#J9AknXzy
z^_WxSg(f497SQr%_Z46vPTE6;xCzK(tS%_IeX$k`odVi$U0^agkmRwau>nXWYFFR(
z{wrXB^T*s=E(h{z!iQj#DlimfOJbJm5R^7wuMMu_hArbG<0A_3s|{1fd<~m{!YFu1
zAWfN&Xh}e20(VR}Zk3kiTI)BzJByOlJU*@%Pjc<;{T|aM#|2F*{+N}2E@Z@R*RmWD
z^I>g=MoZOEfj*6foPQR0!-6UM#<Hw7*VBHrbh;bphuWmxKE~M+&+{ors#)HOFGQ-4
zSd7X*I#<d4+rw3l<TB9-dOaXI-?@Et?6Myi=#viVG>LPf?UUZ3O234QD_9z5XWWS{
z=dhD6#g+<{3TR*OD4#`Jg7%x=e2j!7ka(7s<7}Q$)-G3?F*i^4N^xmYjU2eTx8W+g
zKNj8;d$<bOOs<VauHtgAz{I8*XwrC`iHIl#{sqZrAbQA`Pa0uX>nT6m0o{;F$2<DY
zMFX{HYXC2YmjP@Jyq3oT>03oz1#gQA2T#1bE3Hgi(W#u!yHgbv71GAzZgsX`v~a=b
z!v5s9qq2}dP%|;7oB*mI6SUZ2+jDA*{Nul+%O}}ae6SeM3J6G=^vscw-iOiA#Nl|D
zX&uo0(scDsFpnI21Z}XB4E#E1*|ObH5*du>BT15gxdW5M$<0LaPBe~db-mxU`2|_<
z*wRB5-fY_bKu~f+{qoF80|BxNOw*Sf6hRN%VvAZ`VT}lO^7GtxpLU<kE}nlSjD{Up
z`2im;nz0SzU}T<$6qWd5274?uZFdfmNJ6tv!sn%Jpf&I^3aLamJ6%l<-Y|Kx^GNL-
z3Evv<M&KUHt3qrWaAq+`h(m$AL<NZ7ySh3I(+`|+&Ih&1Cxk<M7|d?nEfgD^?|?6N
zz1ocpY#XRc&WzKc{t4|6^xf^c-s|?C@IuO#cM=D47QLF4M{p3c*xq7xB^*tk4Mm_5
z)&;NYvcZjW3Z9leZQ2=s`ue8<C&-GcT}Hy0V}@-|>^X1a&-<<Hihtu%f*pt>Ap*R0
zF$DDhi{_WcNtwP^#^t7Hk7n+C5oiXn5V#=EWF6yLKWxIUp&ODOtN}g@!|G{QP<4Tt
zy93r=>E?O3Sq5IjLZLyC!!Y=Xl$z^$*X5Pzwx57C@X>yUzo3SbY%&&IuL@3kPO5Qe
z2h!|x2#15-C08rm7A4Uy@jvEz`h;R;ZU*1HXEL+G_roy=KQAt7lI&!=^G(2=j3fG;
z5&3xXGsgZsB4OE!Zj4ItRf)!*vJI^he;(24skY&XwE6xLl7t*Uwa%#z7w8jR^BXf>
zG`WhVlU?=2y^YsTeHweT&*hUa4iH6`GeniK19qm}2w4(nSXNWx_H@g(Z|Gw1s=JjM
zMYmh;)A85&bli!CPTGxH`0o#Ew!fTqow}i?4}F$qDS9+X9dVTt2To>y`wkzZu8lI4
z{*X$iW*8F1ub~p#(kq(+mwFG=|0=|pvg~?>X#{I$v-pIJhkQEcN2m5R#D%=JClRF9
z(16RZJSHb{G?^3$9;I?Va8Jbn;Zy2T75|pk_IQhB#3WZyF8N|#csF(gq}Yat?XG(0
z3;n4+sj|02H7O20v^>8owVc<ssmY<8b{$xlw!x?t$$>(yHeRiATFQZ;$vaLHkSMz!
zNJ2ioV3Sp7sWXga6N8Zpd5BrxQBR_>bT-_)yi*Q>nmhhB6Vo3llx))n{IIU`%@~)M
z(_ioupv#vj?nEdIqy~lJlUw0mF8ZiQ?fY*OQPmnW-9p|y4EnL?aDnhux4FVlOGk7#
z&f6CyjrpsoINkoXNH^vFv3b95^%kCY)e)!x3)m9;$?Ye19&20dPURh4g9c^Hyu{zA
z+87A576XT4VNxCi9@;B6&_{Y&%38NaKOS%>s94CZ!iN#}{Z+}X`2mFL@MhH{uSEV9
zzyYSgv{dRD+eFnbnN_(UB)6m^RoDUGyCxU&GW82b(vlwxGp00pV3-=b*Qs^uPSM%b
z*#*RFN2*W-{TiPO`K;u=jj&||J3%d9z<lC${jS}**QEHwch^=Eb3UxR{1tS3Ge1O-
z?Gi6|LtmIk^S(|DnG-_vBg`$cvqmTBcZU(3O}!q9TDq~Qd(ZO^x_!H_lmDR*b!(ou
zNdW8<Ld<6+Z(>07q_l#9^foIFH7;PVxjpng7P-IGLkW9|4+Wj97$p=oN1EpiN_Jg9
z6{ruCG)#)?;1o`UP}-HQ_&WC-rq4^nd`@qy1nx5j{gTtfVj$}XjfDI`_5E!sB!n)7
zK0o`C+zYDF9A_Inv^bdwd>1y+4O7?YAW^LzPYdqoR!!8-WZ%=Uaj|UOoBpZBLr#Ck
zTfW(!!Hr8oFJnjcg%Q!|+*qP7U>=IZFIFXz8q$vwU^#D2Iqt8Dz~65Ao41V}0qk6J
zSvIf-O=51!wCeCLmz#6M>;pbDbQH!#eT-3ir*f3Zjq`grQ>OMc?SJcYF%qPJ0sd{k
z-cjC+aACpX*YH|7f3i4{8qA1^dR43aFW&`Di?{Zc$ZGR^@@g(nBv|aQme;$|)-DN*
zMvPZ4@+?kY?E>0kHJ+tj-B?+;Tq*M`Z6KIljuRi4`#xNJ?Y@l$s{*?yb|k@%rJeXK
z@kddC>$LdHao`kGr({j+%J|UgB6ThyBJ&vHvOxeemxn?4v$%Pg`ro{Tyv;!VgMMdX
z8WiNj<@8fgD*_iSTU+!bi-|dwIek+E81REamUX(lmwe=6TF$Z?u%hd={Rc`*>@UaV
zbd~{{Qqi(jO0ylzp^OW&j?~7O+rD;k3!mG{4hVd|5l|MPj2j4WZUL@$SGQbr`Rn?p
zR-He2`VgG#Ntaen@_8u+(+&ZXM{y1d{IUf|9Gp^gNu^m|l46H!3%c3g=dChB75;T1
zFThKX4cg{*c5(z%4|#PpL+(k2-VRIcxZ=Al>4*29es1zZ5cd(d4i5n*$d0RpX$EY#
zJT#@NLtLLv<8^ChuAB+BC5vu#mG&c$2x6XH9Q{tyqESDVM$ruMpjP!%<b>d@l*xoM
z+ubhR^OTu(4YPntwDqAF<p}3b;10#C=s@8$n#Nkn1U|N%e);~jlZ6#u`+v=i8L>v-
zOuKMGk?;`|2nyLC?OyD4^@yUA4nNbWPN;@%luo|`@(^F7)x4@i{3+Slh5wPs-pDmj
zbk+Iz1T(iNKuR4$c?k=<1g%7wSGNL)UV^FOLqk=|r<l$;Dr+ERbAAQLQ_j>Whj3{_
z>(pVaq~j0j%dNMQ7*+PSgU*4t57dj7$|l*E;4QXeGJam1`1z4gh$4O{2?q+OGv_FY
z`5IYwq7exG0kzhogs-DZAML!WYm(v9%-;QhV!m!di7M6DK3XcxmXyJ=K4fMazV`58
z8-BW-<fN$8qw_JUuz&*k1*tk44S~yHh#Fo$zSQ!#6lhR9DUZQ|x^{URm~ohN==iD}
zh@Co{WC|10lYy`%Joq7LZkCsr*V3!4q}9?_^2&H&iO;T&kDnQ8Z205b+Ni^C-LKr1
zq)Br6YuE@Vyc1tan#4ztuM0%gNw>X_n%fS<4NtncImG^(z^Zzu>!h8y;-MIb^_ADS
zJMcaaYa55bU?z29ALeM;gOi3uGghZ%1&WQavV3xlu)aV&QDvW%Fr?6EbYd!cu~!f7
zOs9-oD~(jkb-$-;$C(OCxj`Z9eLlA7$6p?jx5MsnD@4_<bIZAEoHestzpzm${-_yJ
zbf|#VjkuM%o@?LZ6^-&H{sOi(qObHzL@AJ0o=Ei7aWz;AT(FJ%FA1IQON#6~BiN`s
zW&!Q4%tK^I5`%MY)<A#Z*v_Ut{4S@VyAabW?<|Tg{gx?afh)Gi;F>}|6Y0}v!ymo0
z)2#&zFpT=P2r`O#ut+F5AcCzL;jDkUJD=bODnIcFHS<(4N~M~mCzp#Maj#-bcS@Q+
z{e3f6?Y8L0*h3aMg7;CU%KA0vr1aGY1z?%N4`=G`ahGl3+Mxn*z&moBIu$BN)Og{{
zGzL1FR(GU^TsvntExAFds4E+v%0Jw4=|4qpId6Pe*Y|%4{uLZ@x!w0zc{@$JC(F<g
zNru%>gctH%JmU*t6{u<6auBh*m$7sD^UmTu6^LM@3K#B)&!_!i0DRyDYvb~+T$r2>
zCg}^@x1t`*;z0Gd6E8%1lf>7O)b$<Kt+y6AaI)FnNHsUP1#G;Fi5hT)YXpvTeqFBm
zL|gVV{^~ae1{6*8kkrY7!4T=R8bI|H4pQMh6|5UOMk~)9KW(|pxvjSZ6@_B)wq2HN
z=im%)?*OVxsIEH6(|@@huqj!+Z@luEJF>#^OHdphVk5mKGdH{&aN;~YBz}Nx1ce2l
zF~=ATMkUM|0bF!g@z&1eAs?|?0bT>jICO!cdn{J<bt^pM707`U;uk=_PdD*r9T*@U
zoX}`bu6gh5WC+a$DWeVyPYbkEpdI~Lov={Yt-`HJaLXP7ia<=MRH~(d?7u>#*1G;7
z{SHdIBNN`mJ34R~O})VG0+xV}h5FS%@m{Ug0AI!S6`^ILa$Qo0-_Q9Vs_}RYZ(McE
zlb{9<e@d<$n3%mqC9baf*!8&!TrX0NsI(C7=nG-5o_wPectp4=W@en9KG;?VZ&4=}
z%fMx$dWK-z2$bb7E^|%$KJ`EWK7ojqY%`IhTiS2^s;fseBlMm!)2pI#CwrO9YUPk?
z;;XIjR~?4rMgg-E7*1DYQF+d1#2O${j|R7x5w%jObbf-Y6Aw?*JHQmCeCvvQan3}Y
zb7co;Kq%D!j-fKcytDH<RN#|{HP(-mY4oY-(l4RRWHHbIraM!F;wEK`QXc|?|2?do
z%I}{rP`>qM)!#6usHWWNPFBLP&|$DF{xSJkOUsc;745<IoSmjz;ePOkD<;%KKmqGB
z$OOt7x(xrA<h+stm21l9Msb;LWQ;l-e3N{AT3QJVFv1-2-CE?f(tCA{QrwZ^9csPp
zM~=#{UCKOxgTCX=bU*>YH$!s<MiPs~FPfwZ@EWx7sy?UpGu`cxsV$~-2$IUXh7Hv3
z;JtuJ^lQNE>*f%6;mNp!2Rk~{Ng>Y8&J{9sJkSWLb5e-w>Jzlo*hZ{q`>wRKiq2`e
z_@zT#L$6}Ik@tBbEf_!pR1;$%X$6oe2iAoPUx|CBPrP0+EpD8c3LFBLAmZr4h7lUh
z8uui`F{?zQQu5?LoL8eY-O&$t?HQeYA#n4^goZloG|o2JcC!krSnf4D6^N!#a22xQ
zX}@EsVmFXKS5xV@w2$q<H$xg5LF=Xw{D=qDYT(Q}b>rDBux=T^LbUENRZBbNI%S}q
zQT*{9EiDF)0b+)Me<WQiPEH$PmUT2<A8EWAw{YO=aAK{k)XZ$^K`I^B0E`SZ?ZrDl
zDj`vH4})s|Z&Yp?oxY;BjtQF7aLRJO@;F!cs8X)g`%ZsEQ)Ec+kKWOxVey6?w?-#2
zfMWA@0CjhlP+Sua)?nG!2}4~@CrD(|a3?;GOzLUd-#bsnPGi^SC+g{zFO$u`%)AiZ
z*hCE->la+cyvOVnJlIx~->sdfd6w@FTm=*LNi!f#+aA7Q{+h0+HiW=o*>}LJ1^zFg
znO=0d?K@Y)vl^|ANm=zCIO{cat9dCi?N*5X<u@G_C}Fh+-UI$;$#1o_H=bopfjfF~
z@`-Ao>;3}yoT3%#1oc$#mp}LRt|+dnns;`24IGXHhB$wAI1Xt<iAnmziE*HA7~%RR
z4##W+4ldTrdWSl3nRICAyEbupm0go(h`Zqr+tBFHTCe{9TYswm2e>4=85ukf6_dm)
zSS-#-J$!XR$}KR96~;#5%vdlRJ*9j7(pfu*d)f<-&Z}7PUK3y;9D@$oMb7`az#;%2
zMnE8kwbmyp<od%@UeZ*5k6ZKj_GW@Z0m08%MX+lg2RnZLp9~<X$UO6~znJ4ykRP%u
zJ!ah11ju8(1X8fLo8Dmf#b+l<B<YF-qf}rX6$>7R%ZkNfRaoHTk#g%{edb%=@<N<h
zNx2QLLX<+kkjstR=>&9nqQX2wy0s>^R+*tM?#?Z)7$l%PKnWiq@9I@gJojJsB#Sn8
z)e6wc<mZUHWFBHw%Cx9b`||=A4DcvA5za4XUnK4k@^kNj3LzM8)A?}YO_nVwL<R&z
zgv@baaT5VOzxMrtynqOz3%@mF4hD(QGV4Q+%v&@cJFv_mN;wH=>e^U&myk!n%4K~?
zH7ggm|3tcy`d6h!#%E{nI4qXdaM1C{_Tro1KYbt}F43GDAA?asQqx1)iETmI=g{e^
zP&^fQ@b9pgS%#u3*<nW_%ee`Q9ErZq3K8gSt8aJhW8I(cU(Ca|=H8i6W#fWk3KX&g
z-VQzo2z&@N%o;5{`eycpR4xlV>c3AOuTf%$<F$I)P25d*)2$!E3W+;CzWTFb0hc2d
z_z10^oqe8TcWaF^f=zTjDR>^lV1prTMh%{np@Nwd>`q;F$&t9zGop~_eaVv|dh6*S
z5{b)2Q)RQJOiZQ&Qo-v=<%{bh2{7K1Fml~7)6qE-K1ckp#3eZp83_h^xqv7KN*C6L
zVikgvWH%qzNpc`bTv60k$g37!H6Pgdq?YWm3iOHx>{iP~pgsg9m^p{kBk?9qPg==K
zSN*^<^&k*9e+z`!3yR8DW@e|}WP-UW0%UHQlJTNPoshCS`<(xsY_>lS8;l43McqKL
zZauttrg$*3RBaMdR?og+1+ENedWeVk^Q_(Q7cR|fZtr~JGu@ue4o(C9jj_Ypi6pn$
zC04)YZH&8g`z@fHA0gZ;dolAm!5Zq(nk3V#)cvDAg+h)OD@HDpf0Zj|z_CWzIXWW(
z2+k)NjCoXRh=Ul2kpnkhxlGY5o1NNe=K0v&ENUty9Yw@NgXV+k>TLF9jz=VsO}Dy}
z?x7XvvQ4+Iuub!DTTHf1!&kZ1+^{88Cfb_Jy)|Fy9&irZ7naNRo~p%fTO$A~iDam$
zroFdv;fsJ09H^3kzD;=qNvJZJPT4KUB>}Rbfrt**3Z5sUIxX@q?oKV%2M{c1_2|$g
z3+%20&`{-_d!`=mX5->Au9_<)L?n@so1+9*LeG#L$Ud)8<t+TI`DkzA?hrhm9nuY3
zV%~46=C2}E5r~xuTM0<HDi-}5{@%QM=aVwz{v~D3vr3%&!6SIhJi;Xu;Mag}gusmd
zm|45hint_qG$6IeL7eyRX6(ZQ&Ix_@+#)A5ffx{iXTcKB3ubw{XI}CqX4d?`|96vh
zXj786nV0}G&~`jN#2swRIl6$EeP~e#eioQ-=hH5PB~^PfyAZe~=1SAyD_w*<b}9-i
zK%%#}u``SWDs$Z<DbvrB-R8zs(X2*wHgQ#U5>QVbtc+F3Ncu#ob@t2}JrMnrbUk_@
zV%S1*8AdCDu}n&C{eaJL_XR`1E-Y5eFmTMw;A9Bq=zO9Jxwqzf;XcL%5EMNsaC$t<
zXpm{QB@<_Mc<x^{uDNXg#0#FY`ME3$?|)qb8H{HIa>29ba|n)=HZV@MAoE!R(wQ;O
zBK2nFDklQff=vJ)J|!4<(F!Vc)Oh_(T>kx98v5MU(5|>Z@n^Znr0bZNd+P^!AklZK
z1$_`#fFQ(x9X?`x(7Zd^MG@5%9TBoP&!Q9%!V|m=%SEPI{J+0KW+P-CDr?3(fZ}ej
zMz9F1^DHp<kzoGNK<0aCQ-bE1IU%tK1mN7lz?P-2pH1}Pf&+fS7X*MOV56{XECPYN
zT?P<U7zzmu^TQ71e+iv?YK0!S4l_TL01g2H;k?QkF+LZC;@6lVw2X`qlVt%Bo{`A^
zxgrn#fDac4GHWi7gr9(cC>oFVkRfCt-6j?Ro)LIFr5l4l;*J>(t~QySnp^xlms5@p
zW%!E3*F{rDdXR4rOP6GufMr)iaQQ2pfHQMs2teaj0yU`6^qGC}|6QL8?8;Y)-lwA4
zx?T8e(D>A<(O+7yD)rR`qcF)pT%csStAOl$hO>St28s5>hs(f`aB`ki>e{L`?t6Lv
zJ@N#mqaMPCDaoi*i%a-aHnY9}%rJFf&=l?Sz<bBqB)>LG6yl<0KfvK2=|DFWT+D%t
ziRnS}enIBHZ&%1oyf8}u=EIR_5EGT)bdYIH{2hr4*%rCL(U^^A_kqDV4<WBp1y*KC
z@`?jAG;ePfPdhbtv?f^x3PC*PYx-_*`1orAcXGAXSa)B&>pT&lBuzjo;T?d;4J;%R
z7?swYh*^fQ=+Vh{_gCOllTHH@Tj<psoGB&k`5e6oh#B*@FWa4Z%RVB40Gd2tS_}`4
zoY!>l;q9ye!wd)0ru_HJa{Bko0!NnFlP1r{(R0g&@37ha`Y7O^(Z3*XNt9ilMU>wk
zp6rd*<=`;xoE3=hx3Ph-IDkDFM(+Uck=<Y9gwkq3*g28*u==;8gRAFE{MxlK?(O^m
zEiDvZMb0<BD;Jl&>YOu)K_bD=2D~KtdT=CwIBd6E%pDCN^Z_AV57AT(h$GWxPb8)6
z-RJx@;GRkF5>+R9GxStoCgJ@1LtG(YO+B985jz~uc?}|)W0PlUUdHnXHrDxw2t1of
zz>>%P_s#ZEt^VDF+g9fm*)=(#I4l%Cq&sHi9AIP&(}#;kQvMt2K^QeLnKKa>eW!`&
zZdD*}A`*Q8(<KHJhPxs1vkM1T^NT#12*eCQft~f!gFmF^o&6!V#I-Wcm(d~sy=Nf@
zI65Dky*t<1!;EMVGMixG^9g}ixl{(028b`@+J1A{HTR$>;*!g2cmzD#ZVgwX^0x+6
z@e|B!kUu<@f$!ff1K+$+loDK!&jfi86HT`DXr2iwvf1hGOGw$dH*3-fsAu(vSs-8m
z?`*le>wC(rTMPW1yotHI4|gP<ky=2nPVjr71nX($<j?(|OZz#9h5tK`hymX(lFxH_
z(&_gm7sdRHSP5_D=aQGg9GUf=c<?6S2N&2mZI2-au+=C=#wYOgj|k^kFq7#ZfHToO
zh(BH4%EltKKMFwhGZPDrxHI#zXCe<jQ>fy`&e{WJK`*i|I|x8lcv5e&J3F<wAlJ@G
zHOVH}xz*a%qsg_<^V!7RBz7*p#ue;@iOMCjZY{uU1PqaXze!^|h+h>I4^2mBU%-cj
zg@K8>f7e*TbALzoe>V^dyv(QgZI1sd`x(2<FY?R>#{nSdNRY}A7Cbd28EkEVppt6h
zPaq1}lFQbhfQ>|p59&A=Mfa=#1OZsOG609P5|)T3r)~p_)n{iRZ+p1|FRdW8m{%bH
z;ZSiWOw`uq5{2!fv=wrBxnd3$7xtWf6UTP9h=_C#0LK={ankC;$hDT^esAVX{`ZxK
z{%qcR??jQLA6Yb2;o%0(8eFF7DA2pk1O%LL(X)nG9;+50>{hJ?k%sN*p@IQHo46&B
z{dpFV(;o}S@L0}FT<D7QFIjhMHEsbrmj&`DJUDy6XDYye%>%+Sawd<N*p}=R4W=u8
zP30lb;CYfVo`cCGoi(W2gUxUJWkkS;fDr*B0!9Rk2pADCB49+oh=36RBLYSQj0hMJ
zFd|??z=(hm0V4uN1dIq65ilZPM8JrE5dk9tMg)uq7!fccU_`)(fDr*B0!9Rk2pADC
zB49+oh=36RBLYSQj0hMJFd|??z=(hm0V4uN1dIq65ilZPM8JrE5dk9tMg)uq7!fcc
zU_`)(fDr*B0!9Rk2pADCB49+oh=36RBLYSQj0hMJFd|??z=(hm0V4uN1dIq65ilZP
WM8JrE5dk9tMg;!<h`_9WI{pXrZv9{Y

literal 0
HcmV?d00001

diff --git a/docs/tutorials/batch-tutorial.ipynb b/docs/tutorials/batch-tutorial.ipynb
new file mode 100644
index 00000000..20e320f5
--- /dev/null
+++ b/docs/tutorials/batch-tutorial.ipynb
@@ -0,0 +1,1546 @@
+{
+ "cells": [
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "### Noisy Batch Optimization\n",
+    "\n",
+    "Many high performing materials suffer from poor corrosion resistance. Anti-corrosion coatings are a common solution that can be adapted to protect a given material in a number of harsh environments. You have the task of designing an anti-corrosion coating for a new material that needs to withstand a relatively high temperature corrosive environment. You decided to simulate the corrosive environment in the lab and test a number of different coatings to see which one performs the best. Based on the lab space, you see that you can specify up to nine coatings in a single test, allowing three replicates of three coating designs to be tested at once.\n",
+    "\n",
+    "You believe Bayesian optimization can help you in this task and decide to put together an optimization script using Honegumi to help solve this problem.\n",
+    "\n",
+    "Looking through the literature you identify the following tunable parameters for this problem:\n",
+    "\n",
+    "| | **Parameter Name** | **Bounds**  |\n",
+    "|------|--------------------|-------------|\n",
+    "| x1   | Resin Fraction              | [0, 1]          |\n",
+    "| x2   | Inhibitor Fraction          | [0, 1]          |\n",
+    "| x3   | Insulator Fraction          | [0, 1]          |\n",
+    "| x4   | Stabilizer Fraction         | [0, 0.1]        |\n",
+    "| x5   | Coating Thickness           | [0.1, 10]       |\n",
+    "\n",
+    "Additionally, you identify several constraints on the relative fractions of the different components based on prior work in the literature. Notably, you observe that the best materials keep `Resin Fraction` > `Inhibitor Fraction` > `Insulator Fraction` > `Stabilizer Fraction`.\n",
+    "\n",
+    "A dummy objective function that emulates the results of the corrosion experiment has been constructed in the code cell below. To simulate sample variability, random noise is added to the output of the function on call. Although we can easily find optimal values using the equations, we will pretend that the objective function is unknown and use a Bayesian optimization approach to find the optimal set of input parameters instead."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 6,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import numpy as np\n",
+    "\n",
+    "def simulate_corrosion(x1, x2, x3, x4, x5):\n",
+    "    \"\"\"\n",
+    "    Calculate the corrosion damage based on the input parameters.\n",
+    "\n",
+    "    Args:\n",
+    "        x1 (float): the fraction of resin used in the coating formulation\n",
+    "        x2 (float): the fraction of inhibitor used in the coating formulation\n",
+    "        x3 (float): the fraction of insulator used in the coating formulation\n",
+    "        x4 (float): the fraction of stabilizer used in the coating formulation\n",
+    "        x5 (float): the coating thickness\n",
+    "\n",
+    "    Returns:\n",
+    "        dict: the measured corrosion damage and  uncertainty\n",
+    "    \"\"\"\n",
+    "    score = float(\n",
+    "        1/(np.exp(-40*(x1-0.42)**2) + np.exp(-24*(x1-0.75)**2) + 0.1) +\n",
+    "        1/(np.exp(-30*(x2-0.22)**2) + np.exp(-1000*(x2-0.22)**2) + np.exp(-800*(x2-0.42)**2) + 0.15) +\n",
+    "        1/(np.exp(-25*(x3-0.27)**2) + np.exp(-1000*(x3-0.27)**2) + np.exp(-150*(x3-0.37)**2) +1) +\n",
+    "        1/(np.exp(-400*(x4-0.08)**2) +1) + 0.5*x4 +\n",
+    "        1/(np.exp(-(x5-8)**2) + 0.05*x5 + 0.3)\n",
+    "    ) - 3.18\n",
+    "\n",
+    "    # simulate 3 replicate measurements\n",
+    "    scores = [abs(score + np.random.normal(0, 0.1)) for _ in range(3)]\n",
+    "\n",
+    "    # report the uncertainty to allow the model to learn from it\n",
+    "    return (np.mean(scores), np.std(scores))"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Applying Honegumi\n",
+    "\n",
+    "We will now use the [Honegumi website](https://honegumi.readthedocs.io/en/latest/) to generate a script that will help us optimize the coating parameters. From the description, we observe that our problem is a **single objective** optimization problem with a **constraint on the fractional sum of coating components** and an **ordering constraint** on the relative fractions of each component. As there is room for several samples to be tested in parallel, **batch optimization** could make the approach more efficient. Lastly, it is expected that the reults will be noisy, so you decide to use a **Fully Bayesian** model to make the optimization process more robust.\n",
+    "\n",
+    "![Honegumi Selections for The Problem Statement](batch-honegumi-selection.jpg)\n",
+    "\n",
+    "The Honegumi generated optimization script will provide a framework for our optimization campaign that we can modify to suit our specific problem needs. In the code sections below, we will make several modifications to this generated script to make it compatible with our problem."
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Modifying the Code for Our Problem\n",
+    "\n",
+    "We can modify this code to suit our problem with a few simple modifications. Wherever a modification has been made to the code, a comment starting with `# CHANGE:` has been added along with a brief description of the change."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 8,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "[INFO 05-01 16:33:32] ax.service.ax_client: Starting optimization with verbose logging. To disable logging, set the `verbose_logging` argument to `False`. Note that float values in the logs are rounded to 6 decimal points.\n",
+      "[WARNING 05-01 16:33:32] ax.service.ax_client: Random seed set to 42. Note that this setting only affects the Sobol quasi-random generator and BoTorch-powered Bayesian optimization models. For the latter models, setting random seed to the same number for two optimizations will make the generated trials similar, but not exactly the same, and over time the trials will diverge more.\n",
+      "[INFO 05-01 16:33:32] ax.service.utils.instantiation: Inferred value type of ParameterType.FLOAT for parameter x1. If that is not the expected value type, you can explicitly specify 'value_type' ('int', 'float', 'bool' or 'str') in parameter dict.\n",
+      "[INFO 05-01 16:33:32] ax.service.utils.instantiation: Inferred value type of ParameterType.FLOAT for parameter x2. If that is not the expected value type, you can explicitly specify 'value_type' ('int', 'float', 'bool' or 'str') in parameter dict.\n",
+      "[INFO 05-01 16:33:32] ax.service.utils.instantiation: Inferred value type of ParameterType.FLOAT for parameter x3. If that is not the expected value type, you can explicitly specify 'value_type' ('int', 'float', 'bool' or 'str') in parameter dict.\n",
+      "[INFO 05-01 16:33:32] ax.service.utils.instantiation: Inferred value type of ParameterType.FLOAT for parameter x5. If that is not the expected value type, you can explicitly specify 'value_type' ('int', 'float', 'bool' or 'str') in parameter dict.\n",
+      "[INFO 05-01 16:33:32] ax.service.utils.instantiation: Created search space: SearchSpace(parameters=[RangeParameter(name='x1', parameter_type=FLOAT, range=[0.0, 1.0]), RangeParameter(name='x2', parameter_type=FLOAT, range=[0.0, 1.0]), RangeParameter(name='x3', parameter_type=FLOAT, range=[0.0, 1.0]), RangeParameter(name='x5', parameter_type=FLOAT, range=[0.1, 10.0])], parameter_constraints=[ParameterConstraint(1.0*x1 + 1.0*x2 + 1.0*x3 <= 1.0), OrderConstraint(x2 <= x1), OrderConstraint(x3 <= x2), ParameterConstraint(-1.0*x1 + -1.0*x2 + -2.0*x3 <= -1.0)]).\n",
+      "[WARNING 05-01 16:33:32] ax.service.ax_client: Random seed set to 42. Note that this setting only affects the Sobol quasi-random generator and BoTorch-powered Bayesian optimization models. For the latter models, setting random seed to the same number for two optimizations will make the generated trials similar, but not exactly the same, and over time the trials will diverge more.\n",
+      "[INFO 05-01 16:33:32] ax.service.utils.instantiation: Inferred value type of ParameterType.FLOAT for parameter x1. If that is not the expected value type, you can explicitly specify 'value_type' ('int', 'float', 'bool' or 'str') in parameter dict.\n",
+      "[INFO 05-01 16:33:32] ax.service.utils.instantiation: Inferred value type of ParameterType.FLOAT for parameter x2. If that is not the expected value type, you can explicitly specify 'value_type' ('int', 'float', 'bool' or 'str') in parameter dict.\n",
+      "[INFO 05-01 16:33:32] ax.service.utils.instantiation: Inferred value type of ParameterType.FLOAT for parameter x3. If that is not the expected value type, you can explicitly specify 'value_type' ('int', 'float', 'bool' or 'str') in parameter dict.\n",
+      "[INFO 05-01 16:33:32] ax.service.utils.instantiation: Inferred value type of ParameterType.FLOAT for parameter x5. If that is not the expected value type, you can explicitly specify 'value_type' ('int', 'float', 'bool' or 'str') in parameter dict.\n",
+      "[INFO 05-01 16:33:32] ax.service.utils.instantiation: Created search space: SearchSpace(parameters=[RangeParameter(name='x1', parameter_type=FLOAT, range=[0.0, 1.0]), RangeParameter(name='x2', parameter_type=FLOAT, range=[0.0, 1.0]), RangeParameter(name='x3', parameter_type=FLOAT, range=[0.0, 1.0]), RangeParameter(name='x5', parameter_type=FLOAT, range=[0.1, 10.0])], parameter_constraints=[ParameterConstraint(1.0*x1 + 1.0*x2 + 1.0*x3 <= 1.0), OrderConstraint(x2 <= x1), OrderConstraint(x3 <= x2), ParameterConstraint(-1.0*x1 + -1.0*x2 + -2.0*x3 <= -1.0)]).\n",
+      "[INFO 05-01 16:33:32] ax.service.ax_client: Generated new trial 0 with parameters {'x1': 0.540494, 'x2': 0.28639, 'x3': 0.106513, 'x5': 3.150732} using model Sobol.\n",
+      "[INFO 05-01 16:33:32] ax.service.ax_client: Generated new trial 1 with parameters {'x1': 0.369966, 'x2': 0.357531, 'x3': 0.268685, 'x5': 7.204235} using model Sobol.\n",
+      "[INFO 05-01 16:33:32] ax.service.ax_client: Completed trial 0 with data: {'corrosion_score': (2.171497, 0.055504)}.\n",
+      "[INFO 05-01 16:33:32] ax.service.ax_client: Completed trial 1 with data: {'corrosion_score': (1.180876, 0.036845)}.\n",
+      "[INFO 05-01 16:33:32] ax.service.ax_client: Generated new trial 2 with parameters {'x1': 0.881586, 'x2': 0.096711, 'x3': 0.01376, 'x5': 6.697699} using model Sobol.\n",
+      "c:\\Users\\sterg\\miniforge3\\envs\\honegumi\\Lib\\site-packages\\ax\\core\\data.py:284: FutureWarning: The behavior of DataFrame concatenation with empty or all-NA entries is deprecated. In a future version, this will no longer exclude empty or all-NA columns when determining the result dtypes. To retain the old behavior, exclude the relevant entries before the concat operation.\n",
+      "  return cls(df=pd.concat(dfs, axis=0, sort=True))\n",
+      "[INFO 05-01 16:33:32] ax.service.ax_client: Generated new trial 3 with parameters {'x1': 0.509931, 'x2': 0.485848, 'x3': 0.002866, 'x5': 9.512716} using model Sobol.\n",
+      "[INFO 05-01 16:33:32] ax.service.ax_client: Completed trial 2 with data: {'corrosion_score': (2.240028, 0.043211)}.\n",
+      "[INFO 05-01 16:33:32] ax.service.ax_client: Completed trial 3 with data: {'corrosion_score': (3.998066, 0.055256)}.\n",
+      "[INFO 05-01 16:33:32] ax.service.ax_client: Generated new trial 4 with parameters {'x1': 0.354248, 'x2': 0.292253, 'x3': 0.232996, 'x5': 4.822508} using model Sobol.\n",
+      "c:\\Users\\sterg\\miniforge3\\envs\\honegumi\\Lib\\site-packages\\ax\\core\\data.py:284: FutureWarning: The behavior of DataFrame concatenation with empty or all-NA entries is deprecated. In a future version, this will no longer exclude empty or all-NA columns when determining the result dtypes. To retain the old behavior, exclude the relevant entries before the concat operation.\n",
+      "  return cls(df=pd.concat(dfs, axis=0, sort=True))\n",
+      "[INFO 05-01 16:33:32] ax.service.ax_client: Generated new trial 5 with parameters {'x1': 0.421201, 'x2': 0.223665, 'x3': 0.200311, 'x5': 7.933213} using model Sobol.\n",
+      "[INFO 05-01 16:33:32] ax.service.ax_client: Completed trial 4 with data: {'corrosion_score': (1.889654, 0.063915)}.\n",
+      "[INFO 05-01 16:33:32] ax.service.ax_client: Completed trial 5 with data: {'corrosion_score': (0.312144, 0.108107)}.\n",
+      "Sample: 100%|██████████| 2048/2048 [01:02, 32.59it/s, step size=4.12e-01, acc. prob=0.905]\n",
+      "[INFO 05-01 16:34:39] ax.service.ax_client: Generated new trial 6 with parameters {'x1': 0.58821, 'x2': 0.205895, 'x3': 0.205895, 'x5': 10.0} using model FullyBayesian.\n",
+      "c:\\Users\\sterg\\miniforge3\\envs\\honegumi\\Lib\\site-packages\\ax\\core\\data.py:284: FutureWarning: The behavior of DataFrame concatenation with empty or all-NA entries is deprecated. In a future version, this will no longer exclude empty or all-NA columns when determining the result dtypes. To retain the old behavior, exclude the relevant entries before the concat operation.\n",
+      "  return cls(df=pd.concat(dfs, axis=0, sort=True))\n",
+      "[INFO 05-01 16:34:39] ax.modelbridge.torch: The observations are identical to the last set of observations used to fit the model. Skipping model fitting.\n",
+      "[INFO 05-01 16:34:45] ax.service.ax_client: Generated new trial 7 with parameters {'x1': 1.0, 'x2': 0.0, 'x3': 0.0, 'x5': 0.1} using model FullyBayesian.\n",
+      "[INFO 05-01 16:34:45] ax.service.ax_client: Completed trial 6 with data: {'corrosion_score': (1.052019, 0.141698)}.\n",
+      "[INFO 05-01 16:34:45] ax.service.ax_client: Completed trial 7 with data: {'corrosion_score': (7.553051, 0.057187)}.\n",
+      "Sample: 100%|██████████| 2048/2048 [00:53, 38.06it/s, step size=5.12e-01, acc. prob=0.855]\n",
+      "[INFO 05-01 16:35:45] ax.service.ax_client: Generated new trial 8 with parameters {'x1': 0.25, 'x2': 0.25, 'x3': 0.25, 'x5': 10.0} using model FullyBayesian.\n",
+      "c:\\Users\\sterg\\miniforge3\\envs\\honegumi\\Lib\\site-packages\\ax\\core\\data.py:284: FutureWarning: The behavior of DataFrame concatenation with empty or all-NA entries is deprecated. In a future version, this will no longer exclude empty or all-NA columns when determining the result dtypes. To retain the old behavior, exclude the relevant entries before the concat operation.\n",
+      "  return cls(df=pd.concat(dfs, axis=0, sort=True))\n",
+      "[INFO 05-01 16:35:45] ax.modelbridge.torch: The observations are identical to the last set of observations used to fit the model. Skipping model fitting.\n",
+      "[INFO 05-01 16:35:51] ax.service.ax_client: Generated new trial 9 with parameters {'x1': 0.25, 'x2': 0.25, 'x3': 0.25, 'x5': 10.0} using model FullyBayesian.\n",
+      "[INFO 05-01 16:35:51] ax.service.ax_client: Completed trial 8 with data: {'corrosion_score': (2.644857, 0.151135)}.\n",
+      "[INFO 05-01 16:35:51] ax.service.ax_client: Completed trial 9 with data: {'corrosion_score': (2.528284, 0.04713)}.\n",
+      "[INFO 05-01 16:35:51] ax.modelbridge.base: Leaving out out-of-design observations for arms: 9_0\n",
+      "Sample: 100%|██████████| 2048/2048 [01:02, 32.76it/s, step size=4.35e-01, acc. prob=0.880]\n",
+      "[INFO 05-01 16:37:01] ax.service.ax_client: Generated new trial 10 with parameters {'x1': 0.615542, 'x2': 0.192229, 'x3': 0.192229, 'x5': 7.587642} using model FullyBayesian.\n",
+      "c:\\Users\\sterg\\miniforge3\\envs\\honegumi\\Lib\\site-packages\\ax\\core\\data.py:284: FutureWarning: The behavior of DataFrame concatenation with empty or all-NA entries is deprecated. In a future version, this will no longer exclude empty or all-NA columns when determining the result dtypes. To retain the old behavior, exclude the relevant entries before the concat operation.\n",
+      "  return cls(df=pd.concat(dfs, axis=0, sort=True))\n",
+      "[INFO 05-01 16:37:01] ax.modelbridge.base: Leaving out out-of-design observations for arms: 9_0\n",
+      "[INFO 05-01 16:37:01] ax.modelbridge.torch: The observations are identical to the last set of observations used to fit the model. Skipping model fitting.\n",
+      "[INFO 05-01 16:37:08] ax.service.ax_client: Generated new trial 11 with parameters {'x1': 0.615612, 'x2': 0.192194, 'x3': 0.192194, 'x5': 7.586841} using model FullyBayesian.\n",
+      "[INFO 05-01 16:37:08] ax.service.ax_client: Completed trial 10 with data: {'corrosion_score': (0.55539, 0.077393)}.\n",
+      "[INFO 05-01 16:37:08] ax.service.ax_client: Completed trial 11 with data: {'corrosion_score': (0.761784, 0.067236)}.\n",
+      "[INFO 05-01 16:37:08] ax.modelbridge.base: Leaving out out-of-design observations for arms: 9_0\n",
+      "Sample: 100%|██████████| 2048/2048 [00:54, 37.74it/s, step size=5.73e-01, acc. prob=0.778]\n",
+      "[INFO 05-01 16:38:08] ax.service.ax_client: Generated new trial 12 with parameters {'x1': 1.0, 'x2': 0.0, 'x3': 0.0, 'x5': 10.0} using model FullyBayesian.\n",
+      "c:\\Users\\sterg\\miniforge3\\envs\\honegumi\\Lib\\site-packages\\ax\\core\\data.py:284: FutureWarning: The behavior of DataFrame concatenation with empty or all-NA entries is deprecated. In a future version, this will no longer exclude empty or all-NA columns when determining the result dtypes. To retain the old behavior, exclude the relevant entries before the concat operation.\n",
+      "  return cls(df=pd.concat(dfs, axis=0, sort=True))\n",
+      "[INFO 05-01 16:38:08] ax.modelbridge.base: Leaving out out-of-design observations for arms: 9_0\n",
+      "[INFO 05-01 16:38:08] ax.modelbridge.torch: The observations are identical to the last set of observations used to fit the model. Skipping model fitting.\n",
+      "[INFO 05-01 16:38:15] ax.service.ax_client: Generated new trial 13 with parameters {'x1': 1.0, 'x2': 0.0, 'x3': 0.0, 'x5': 10.0} using model FullyBayesian.\n",
+      "[INFO 05-01 16:38:15] ax.service.ax_client: Completed trial 12 with data: {'corrosion_score': (5.661515, 0.075103)}.\n",
+      "[INFO 05-01 16:38:15] ax.service.ax_client: Completed trial 13 with data: {'corrosion_score': (5.415819, 0.045722)}.\n",
+      "[INFO 05-01 16:38:15] ax.modelbridge.base: Leaving out out-of-design observations for arms: 9_0\n",
+      "Sample: 100%|██████████| 2048/2048 [01:09, 29.27it/s, step size=3.20e-01, acc. prob=0.940]\n",
+      "[INFO 05-01 16:39:31] ax.service.ax_client: Generated new trial 14 with parameters {'x1': 0.481826, 'x2': 0.172725, 'x3': 0.172725, 'x5': 7.767653} using model FullyBayesian.\n",
+      "c:\\Users\\sterg\\miniforge3\\envs\\honegumi\\Lib\\site-packages\\ax\\core\\data.py:284: FutureWarning: The behavior of DataFrame concatenation with empty or all-NA entries is deprecated. In a future version, this will no longer exclude empty or all-NA columns when determining the result dtypes. To retain the old behavior, exclude the relevant entries before the concat operation.\n",
+      "  return cls(df=pd.concat(dfs, axis=0, sort=True))\n",
+      "[INFO 05-01 16:39:31] ax.modelbridge.base: Leaving out out-of-design observations for arms: 9_0\n",
+      "[INFO 05-01 16:39:31] ax.modelbridge.torch: The observations are identical to the last set of observations used to fit the model. Skipping model fitting.\n",
+      "[INFO 05-01 16:39:39] ax.service.ax_client: Generated new trial 15 with parameters {'x1': 0.481717, 'x2': 0.172761, 'x3': 0.172761, 'x5': 7.768358} using model FullyBayesian.\n",
+      "[INFO 05-01 16:39:39] ax.service.ax_client: Completed trial 14 with data: {'corrosion_score': (0.681955, 0.01967)}.\n",
+      "[INFO 05-01 16:39:39] ax.service.ax_client: Completed trial 15 with data: {'corrosion_score': (0.802229, 0.073447)}.\n",
+      "[INFO 05-01 16:39:39] ax.modelbridge.base: Leaving out out-of-design observations for arms: 15_0, 9_0\n",
+      "Sample: 100%|██████████| 2048/2048 [01:01, 33.48it/s, step size=3.64e-01, acc. prob=0.899]\n",
+      "[INFO 05-01 16:40:49] ax.service.ax_client: Generated new trial 16 with parameters {'x1': 0.71564, 'x2': 0.242649, 'x3': 0.041711, 'x5': 7.952869} using model FullyBayesian.\n",
+      "c:\\Users\\sterg\\miniforge3\\envs\\honegumi\\Lib\\site-packages\\ax\\core\\data.py:284: FutureWarning: The behavior of DataFrame concatenation with empty or all-NA entries is deprecated. In a future version, this will no longer exclude empty or all-NA columns when determining the result dtypes. To retain the old behavior, exclude the relevant entries before the concat operation.\n",
+      "  return cls(df=pd.concat(dfs, axis=0, sort=True))\n",
+      "[INFO 05-01 16:40:49] ax.modelbridge.base: Leaving out out-of-design observations for arms: 15_0, 9_0\n",
+      "[INFO 05-01 16:40:49] ax.modelbridge.torch: The observations are identical to the last set of observations used to fit the model. Skipping model fitting.\n",
+      "[INFO 05-01 16:40:57] ax.service.ax_client: Generated new trial 17 with parameters {'x1': 0.470028, 'x2': 0.264986, 'x3': 0.264986, 'x5': 8.34103} using model FullyBayesian.\n",
+      "[INFO 05-01 16:40:57] ax.service.ax_client: Completed trial 16 with data: {'corrosion_score': (0.607151, 0.094466)}.\n",
+      "[INFO 05-01 16:40:57] ax.service.ax_client: Completed trial 17 with data: {'corrosion_score': (0.344838, 0.043934)}.\n",
+      "[INFO 05-01 16:40:57] ax.modelbridge.base: Leaving out out-of-design observations for arms: 15_0, 9_0\n",
+      "Sample: 100%|██████████| 2048/2048 [01:01, 33.31it/s, step size=3.33e-01, acc. prob=0.871]\n",
+      "[INFO 05-01 16:42:05] ax.service.ax_client: Generated new trial 18 with parameters {'x1': 0.5, 'x2': 0.5, 'x3': 0.0, 'x5': 0.1} using model FullyBayesian.\n",
+      "c:\\Users\\sterg\\miniforge3\\envs\\honegumi\\Lib\\site-packages\\ax\\core\\data.py:284: FutureWarning: The behavior of DataFrame concatenation with empty or all-NA entries is deprecated. In a future version, this will no longer exclude empty or all-NA columns when determining the result dtypes. To retain the old behavior, exclude the relevant entries before the concat operation.\n",
+      "  return cls(df=pd.concat(dfs, axis=0, sort=True))\n",
+      "[INFO 05-01 16:42:05] ax.modelbridge.base: Leaving out out-of-design observations for arms: 15_0, 9_0\n",
+      "[INFO 05-01 16:42:05] ax.modelbridge.torch: The observations are identical to the last set of observations used to fit the model. Skipping model fitting.\n",
+      "[INFO 05-01 16:42:11] ax.service.ax_client: Generated new trial 19 with parameters {'x1': 0.5, 'x2': 0.5, 'x3': 0.0, 'x5': 0.1} using model FullyBayesian.\n",
+      "[INFO 05-01 16:42:11] ax.service.ax_client: Completed trial 18 with data: {'corrosion_score': (6.691786, 0.072006)}.\n",
+      "[INFO 05-01 16:42:11] ax.service.ax_client: Completed trial 19 with data: {'corrosion_score': (6.70339, 0.036864)}.\n",
+      "[INFO 05-01 16:42:11] ax.modelbridge.base: Leaving out out-of-design observations for arms: 19_0, 15_0, 9_0\n",
+      "Sample: 100%|██████████| 2048/2048 [00:44, 45.67it/s, step size=4.54e-01, acc. prob=0.914]\n"
+     ]
+    }
+   ],
+   "source": [
+    "import numpy as np\n",
+    "from ax.service.ax_client import AxClient, ObjectiveProperties\n",
+    "\n",
+    "from ax.modelbridge.factory import Models\n",
+    "from ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n",
+    "\n",
+    "\n",
+    "obj1_name = \"corrosion_score\" # CHANGE: update objective name\n",
+    "\n",
+    "# CHANGE: remove the branin dummy objective function, we will use the above function\n",
+    "\n",
+    "total = 1.0 # CHANGE: update total component fraction\n",
+    "\n",
+    "\n",
+    "gs = GenerationStrategy(\n",
+    "    steps=[\n",
+    "        GenerationStep(\n",
+    "            model=Models.SOBOL,\n",
+    "            num_trials=6,\n",
+    "            min_trials_observed=3,\n",
+    "            max_parallelism=5,\n",
+    "            model_kwargs={\"seed\": 999},\n",
+    "            model_gen_kwargs={},\n",
+    "        ),\n",
+    "        GenerationStep(\n",
+    "            model=Models.FULLYBAYESIAN,\n",
+    "            num_trials=-1,\n",
+    "            max_parallelism=3,\n",
+    "            should_deduplicate=True, # CHANGE: reduce duplicate suggestions\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024}, # CHANGE: increase\n",
+    "        ),\n",
+    "    ]\n",
+    ")\n",
+    "\n",
+    "ax_client = AxClient(generation_strategy=gs, random_seed=42) # CHANGE: add random seed for reproducibility\n",
+    "\n",
+    "ax_client.create_experiment(\n",
+    "    parameters=[\n",
+    "        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]}, # CHANGE: update parameter\n",
+    "        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]}, # CHANGE: update parameter\n",
+    "        {\"name\": \"x3\", \"type\": \"range\", \"bounds\": [0.0, total]}, # CHANGE: add new parameter\n",
+    "        {\"name\": \"x5\", \"type\": \"range\", \"bounds\": [0.1, 10.0]}, # CHANGE: add new parameter\n",
+    "    ],\n",
+    "    objectives={\n",
+    "        obj1_name: ObjectiveProperties(minimize=True),\n",
+    "    },\n",
+    "    parameter_constraints=[\n",
+    "        f\"x1 + x2 + x3 <= {total}\", # CHANGE: update composition constraint\n",
+    "        \"x1 >= x2\", # CHANGE: update order constraint\n",
+    "        \"x2 >= x3\", # CHANGE: add order constraint\n",
+    "        \"2.0*x3 + x1 + x2 >= 1.0\", # CHANGE: add order constraint based on `x3 >= x4` and `x4 = 1 - (x1 + x2 + x3)`\n",
+    "    ],\n",
+    ")\n",
+    "\n",
+    "batch_size = 2\n",
+    "\n",
+    "for _ in range(10): # CHANGE: decrease number of iterations\n",
+    "\n",
+    "    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n",
+    "    for trial_index, parameterization in list(parameterizations.items()):\n",
+    "        \n",
+    "        # CHANGE: pull all added parameters from the parameterization\n",
+    "        x1 = parameterization[\"x1\"]\n",
+    "        x2 = parameterization[\"x2\"]\n",
+    "        x3 = parameterization[\"x3\"]\n",
+    "        x4 = total - (x1 + x2 + x3) # CHANGE: update composition constraint\n",
+    "        x5 = parameterization[\"x5\"]\n",
+    "\n",
+    "        results = simulate_corrosion(x1, x2, x3, x4, x5)\n",
+    "        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n",
+    "\n",
+    "best_parameters, metrics = ax_client.get_best_parameters()"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "### Show the Best Parameters\n",
+    "\n",
+    "After our optimization loop has completed, we can use the model to find the best parameters and their corresponding strength value. These will be our optimial set of parameters that we use in the 3D printer going forward."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 338,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "Sample: 100%|██████████| 2048/2048 [00:42, 47.73it/s, step size=2.50e-01, acc. prob=0.877]\n"
+     ]
+    },
+    {
+     "data": {
+      "text/plain": [
+       "(19,\n",
+       " {'x1': 0.5531141417704178,\n",
+       "  'x2': 0.1846591406143514,\n",
+       "  'x3': 0.26222671761513805,\n",
+       "  'x5': 8.2901097602077},\n",
+       " ({'corrosion_score': 0.41040928033269086},\n",
+       "  {'corrosion_score': {'corrosion_score': 0.0008997393689987544}}))"
+      ]
+     },
+     "execution_count": 338,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "ax_client.get_best_trial()"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "### Plotting Optimization Performance\n",
+    "\n",
+    "We can plot the performance of our optmization loop to see how the optimization task progressed as a function of iteration count."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 339,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "[WARNING 05-01 08:58:34] ax.service.utils.report_utils: Column reason missing for all trials. Not appending column.\n"
+     ]
+    },
+    {
+     "data": {
+      "text/plain": [
+       "(0.0, 2.0)"
+      ]
+     },
+     "execution_count": 339,
+     "metadata": {},
+     "output_type": "execute_result"
+    },
+    {
+     "data": {
+      "image/png": "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",
+      "text/plain": [
+       "<Figure size 720x480 with 1 Axes>"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    }
+   ],
+   "source": [
+    "import matplotlib.pyplot as plt\n",
+    "h_blue = '#0033FF'\n",
+    "\n",
+    "df = ax_client.get_trials_data_frame()\n",
+    "fig, ax = plt.subplots(figsize=(6,4), dpi=120)\n",
+    "ax.plot(df.corrosion_score.values, ls='None', marker='o', mfc='none', mec='black')\n",
+    "ax.plot(np.minimum.accumulate(df.corrosion_score.values), color=h_blue)\n",
+    "plt.xlabel('Trial')\n",
+    "plt.ylabel('Corrosion Score [lower is better]')\n",
+    "plt.ylim(0, 2)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 340,
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "application/vnd.plotly.v1+json": {
+       "config": {
+        "linkText": "Export to plot.ly",
+        "plotlyServerURL": "https://plot.ly",
+        "showLink": false
+       },
+       "data": [
+        {
+         "hoverinfo": "none",
+         "line": {
+          "color": "black",
+          "dash": "dot",
+          "width": 2
+         },
+         "mode": "lines",
+         "showlegend": false,
+         "type": "scatter",
+         "visible": true,
+         "x": [
+          -3.952154604781554,
+          10.903313786058435
+         ],
+         "y": [
+          -3.952154604781554,
+          10.903313786058435
+         ]
+        },
+        {
+         "error_x": {
+          "array": [
+           0.09697307193758328,
+           0.09432580615956097,
+           0.03916477554641934,
+           0.10977525367163038,
+           0.18041947262022728,
+           0.023662859070073566,
+           0.0316572302731216,
+           0.16109585493320808,
+           0.03103124681089616,
+           0.15511805238701318,
+           0.03874663888694734,
+           0.04354174301098419,
+           0.13657119941037701,
+           0.1018157587494321,
+           0.1110139250248606,
+           0.1641737784542783,
+           0.08661683840337824,
+           0.07612889196206886
+          ],
+          "color": "rgba(128,177,211,0.4)",
+          "type": "data"
+         },
+         "error_y": {
+          "array": [
+           2.7256679784174875,
+           4.199633838796122,
+           3.599793582557131,
+           5.2253166045607005,
+           4.516845300842414,
+           3.0953335998152287,
+           0.16104538196927914,
+           0.03166602726203759,
+           4.657165685333809,
+           3.574701887771811,
+           0.045083000885533506,
+           0.040628188869374526,
+           6.458009708196848,
+           5.374882199652275,
+           3.9081294574205843,
+           2.977922279320509,
+           0.07742037877804649,
+           0.08754464854448725
+          ],
+          "color": "rgba(128,177,211,0.4)",
+          "type": "data"
+         },
+         "hoverinfo": "text",
+         "marker": {
+          "color": "rgba(128,177,211,1)"
+         },
+         "mode": "markers",
+         "name": "In-sample",
+         "showlegend": true,
+         "text": [
+          "<b>Arm 0_0</b><br><br>Actual Outcome: 1.164 [1.067, 1.261]<br>Predicted Outcome: 2.102 [-0.624, 4.828]<br><br><em>Parameterization:</em><br>x1: 0.46203490160405636<br>x2: 0.1405540555715561<br>x3: 0.12007084302604198<br>x5: 7.49138588104397",
+          "<b>Arm 1_0</b><br><br>Actual Outcome: 2.737 [2.643, 2.831]<br>Predicted Outcome: 3.082 [-1.118, 7.281]<br><br><em>Parameterization:</em><br>x1: 0.6811950104311109<br>x2: 0.08317248616367579<br>x3: 0.14701138995587826<br>x5: 2.201754858437926",
+          "<b>Arm 2_0</b><br><br>Actual Outcome: 2.173 [2.134, 2.212]<br>Predicted Outcome: 3.007 [-0.593, 6.607]<br><br><em>Parameterization:</em><br>x1: 0.5404942771419883<br>x2: 0.28639009688049555<br>x3: 0.10651267971843481<br>x5: 3.1507322339341046",
+          "<b>Arm 3_0</b><br><br>Actual Outcome: 7.654 [7.544, 7.764]<br>Predicted Outcome: 3.416 [-1.809, 8.641]<br><br><em>Parameterization:</em><br>x1: 0.17495027277618647<br>x2: 0.007729803211987019<br>x3: 0.0571193927899003<br>x5: 9.218316878192127",
+          "<b>Arm 4_0</b><br><br>Actual Outcome: 3.357 [3.177, 3.538]<br>Predicted Outcome: 2.838 [-1.679, 7.354]<br><br><em>Parameterization:</em><br>x1: 0.2820521341636777<br>x2: 0.12209852878004313<br>x3: 0.23468931019306183<br>x5: 3.671897889301181",
+          "<b>Arm 5_0</b><br><br>Actual Outcome: 1.163 [1.139, 1.186]<br>Predicted Outcome: 1.707 [-1.388, 4.803]<br><br><em>Parameterization:</em><br>x1: 0.36996550764888525<br>x2: 0.35753058083355427<br>x3: 0.2686848407611251<br>x5: 7.204235409293323",
+          "<b>Arm 6_0</b><br><br>Actual Outcome: 5.56 [5.529, 5.592]<br>Predicted Outcome: 5.557 [5.396, 5.718]<br><br><em>Parameterization:</em><br>x1: 0.9999999999999886<br>x2: 0.0<br>x3: 1.2886717519318773e-14<br>x5: 6.657871762887141",
+          "<b>Arm 7_0</b><br><br>Actual Outcome: 5.56 [5.399, 5.721]<br>Predicted Outcome: 5.56 [5.529, 5.592]<br><br><em>Parameterization:</em><br>x1: 1.0<br>x2: 5.879023101716506e-16<br>x3: 8.221970846891384e-16<br>x5: 6.658104954751326",
+          "<b>Arm 8_0</b><br><br>Actual Outcome: 3.228 [3.197, 3.259]<br>Predicted Outcome: 4.24 [-0.417, 8.897]<br><br><em>Parameterization:</em><br>x1: 0.5000000000004156<br>x2: 0.0<br>x3: 0.5000000000021749<br>x5: 9.999999999999234",
+          "<b>Arm 9_0</b><br><br>Actual Outcome: 2.851 [2.696, 3.006]<br>Predicted Outcome: 2.316 [-1.259, 5.891]<br><br><em>Parameterization:</em><br>x1: 0.49999999999999983<br>x2: 2.9442422767142952e-15<br>x3: 0.5000000000000115<br>x5: 7.195932963944851",
+          "<b>Arm 10_0</b><br><br>Actual Outcome: 0.954 [0.915, 0.993]<br>Predicted Outcome: 0.952 [0.907, 0.997]<br><br><em>Parameterization:</em><br>x1: 0.4244838630578679<br>x2: 0.42448386306381597<br>x3: 2.1289729987226923e-11<br>x5: 7.339947666832862",
+          "<b>Arm 11_0</b><br><br>Actual Outcome: 0.951 [0.907, 0.994]<br>Predicted Outcome: 0.953 [0.912, 0.994]<br><br><em>Parameterization:</em><br>x1: 0.4245136618061755<br>x2: 0.42451366180615974<br>x3: 0.0<br>x5: 7.345265540626473",
+          "<b>Arm 12_0</b><br><br>Actual Outcome: 5.698 [5.561, 5.835]<br>Predicted Outcome: 3.77 [-2.688, 10.228]<br><br><em>Parameterization:</em><br>x1: 0.4262126487368945<br>x2: 8.147112918011196e-16<br>x3: 4.376823862696568e-15<br>x5: 0.1000000000000202",
+          "<b>Arm 13_0</b><br><br>Actual Outcome: 4.724 [4.622, 4.826]<br>Predicted Outcome: 2.098 [-3.277, 7.473]<br><br><em>Parameterization:</em><br>x1: 0.5000000000012181<br>x2: 0.5000000000106196<br>x3: 0.0<br>x5: 10.0",
+          "<b>Arm 14_0</b><br><br>Actual Outcome: 2.555 [2.444, 2.666]<br>Predicted Outcome: 1.663 [-2.245, 5.571]<br><br><em>Parameterization:</em><br>x1: 0.42085294074668167<br>x2: 0.3440857934253744<br>x3: 0.0<br>x5: 5.928959908522031",
+          "<b>Arm 15_0</b><br><br>Actual Outcome: 0.453 [0.289, 0.617]<br>Predicted Outcome: 1.004 [-1.974, 3.981]<br><br><em>Parameterization:</em><br>x1: 0.6191508786933483<br>x2: 0.23051694149204408<br>x3: 0.1503321798146068<br>x5: 7.898638178540372",
+          "<b>Arm 16_0</b><br><br>Actual Outcome: 1.248 [1.161, 1.334]<br>Predicted Outcome: 1.246 [1.169, 1.324]<br><br><em>Parameterization:</em><br>x1: 0.8194301997175271<br>x2: 0.18056980028249606<br>x3: 2.299616162607646e-15<br>x5: 8.956439436443388",
+          "<b>Arm 17_0</b><br><br>Actual Outcome: 1.244 [1.167, 1.320]<br>Predicted Outcome: 1.245 [1.158, 1.333]<br><br><em>Parameterization:</em><br>x1: 0.8193333945709272<br>x2: 0.18066660542907836<br>x3: 0.0<br>x5: 8.951647627618732"
+         ],
+         "type": "scatter",
+         "visible": true,
+         "x": [
+          1.1637748514190176,
+          2.7371261343507274,
+          2.1728861476173438,
+          7.653779219276586,
+          3.3574303343959673,
+          1.1626663489625728,
+          5.560470369357521,
+          5.5602897702587555,
+          3.2281845802895037,
+          2.8509992024517383,
+          0.953817334433674,
+          0.9508013533028197,
+          5.697936838066944,
+          4.7240923300902775,
+          2.55463885701626,
+          0.45323031180355544,
+          1.247548031754366,
+          1.2435919914412081
+         ],
+         "y": [
+          2.102091278956899,
+          3.0818544551702387,
+          3.0070574018667036,
+          3.4158654322816258,
+          2.8375954991177417,
+          1.7072225760669248,
+          5.557351121990914,
+          5.560187764542006,
+          4.240162969938776,
+          2.3160229549868583,
+          0.9522162444816504,
+          0.9529349604944826,
+          3.770055514641586,
+          2.09797615809072,
+          1.662826192337171,
+          1.0035465684178826,
+          1.2464475640682549,
+          1.2451972626406453
+         ]
+        }
+       ],
+       "layout": {
+        "annotations": [
+         {
+          "showarrow": false,
+          "text": "Show CI",
+          "x": 1.125,
+          "xanchor": "left",
+          "xref": "paper",
+          "y": 0.9,
+          "yanchor": "middle",
+          "yref": "paper"
+         }
+        ],
+        "height": 500,
+        "hovermode": "closest",
+        "margin": {
+         "b": 90
+        },
+        "showlegend": false,
+        "template": {
+         "data": {
+          "bar": [
+           {
+            "error_x": {
+             "color": "#2a3f5f"
+            },
+            "error_y": {
+             "color": "#2a3f5f"
+            },
+            "marker": {
+             "line": {
+              "color": "#E5ECF6",
+              "width": 0.5
+             },
+             "pattern": {
+              "fillmode": "overlay",
+              "size": 10,
+              "solidity": 0.2
+             }
+            },
+            "type": "bar"
+           }
+          ],
+          "barpolar": [
+           {
+            "marker": {
+             "line": {
+              "color": "#E5ECF6",
+              "width": 0.5
+             },
+             "pattern": {
+              "fillmode": "overlay",
+              "size": 10,
+              "solidity": 0.2
+             }
+            },
+            "type": "barpolar"
+           }
+          ],
+          "carpet": [
+           {
+            "aaxis": {
+             "endlinecolor": "#2a3f5f",
+             "gridcolor": "white",
+             "linecolor": "white",
+             "minorgridcolor": "white",
+             "startlinecolor": "#2a3f5f"
+            },
+            "baxis": {
+             "endlinecolor": "#2a3f5f",
+             "gridcolor": "white",
+             "linecolor": "white",
+             "minorgridcolor": "white",
+             "startlinecolor": "#2a3f5f"
+            },
+            "type": "carpet"
+           }
+          ],
+          "choropleth": [
+           {
+            "colorbar": {
+             "outlinewidth": 0,
+             "ticks": ""
+            },
+            "type": "choropleth"
+           }
+          ],
+          "contour": [
+           {
+            "colorbar": {
+             "outlinewidth": 0,
+             "ticks": ""
+            },
+            "colorscale": [
+             [
+              0,
+              "#0d0887"
+             ],
+             [
+              0.1111111111111111,
+              "#46039f"
+             ],
+             [
+              0.2222222222222222,
+              "#7201a8"
+             ],
+             [
+              0.3333333333333333,
+              "#9c179e"
+             ],
+             [
+              0.4444444444444444,
+              "#bd3786"
+             ],
+             [
+              0.5555555555555556,
+              "#d8576b"
+             ],
+             [
+              0.6666666666666666,
+              "#ed7953"
+             ],
+             [
+              0.7777777777777778,
+              "#fb9f3a"
+             ],
+             [
+              0.8888888888888888,
+              "#fdca26"
+             ],
+             [
+              1,
+              "#f0f921"
+             ]
+            ],
+            "type": "contour"
+           }
+          ],
+          "contourcarpet": [
+           {
+            "colorbar": {
+             "outlinewidth": 0,
+             "ticks": ""
+            },
+            "type": "contourcarpet"
+           }
+          ],
+          "heatmap": [
+           {
+            "colorbar": {
+             "outlinewidth": 0,
+             "ticks": ""
+            },
+            "colorscale": [
+             [
+              0,
+              "#0d0887"
+             ],
+             [
+              0.1111111111111111,
+              "#46039f"
+             ],
+             [
+              0.2222222222222222,
+              "#7201a8"
+             ],
+             [
+              0.3333333333333333,
+              "#9c179e"
+             ],
+             [
+              0.4444444444444444,
+              "#bd3786"
+             ],
+             [
+              0.5555555555555556,
+              "#d8576b"
+             ],
+             [
+              0.6666666666666666,
+              "#ed7953"
+             ],
+             [
+              0.7777777777777778,
+              "#fb9f3a"
+             ],
+             [
+              0.8888888888888888,
+              "#fdca26"
+             ],
+             [
+              1,
+              "#f0f921"
+             ]
+            ],
+            "type": "heatmap"
+           }
+          ],
+          "heatmapgl": [
+           {
+            "colorbar": {
+             "outlinewidth": 0,
+             "ticks": ""
+            },
+            "colorscale": [
+             [
+              0,
+              "#0d0887"
+             ],
+             [
+              0.1111111111111111,
+              "#46039f"
+             ],
+             [
+              0.2222222222222222,
+              "#7201a8"
+             ],
+             [
+              0.3333333333333333,
+              "#9c179e"
+             ],
+             [
+              0.4444444444444444,
+              "#bd3786"
+             ],
+             [
+              0.5555555555555556,
+              "#d8576b"
+             ],
+             [
+              0.6666666666666666,
+              "#ed7953"
+             ],
+             [
+              0.7777777777777778,
+              "#fb9f3a"
+             ],
+             [
+              0.8888888888888888,
+              "#fdca26"
+             ],
+             [
+              1,
+              "#f0f921"
+             ]
+            ],
+            "type": "heatmapgl"
+           }
+          ],
+          "histogram": [
+           {
+            "marker": {
+             "pattern": {
+              "fillmode": "overlay",
+              "size": 10,
+              "solidity": 0.2
+             }
+            },
+            "type": "histogram"
+           }
+          ],
+          "histogram2d": [
+           {
+            "colorbar": {
+             "outlinewidth": 0,
+             "ticks": ""
+            },
+            "colorscale": [
+             [
+              0,
+              "#0d0887"
+             ],
+             [
+              0.1111111111111111,
+              "#46039f"
+             ],
+             [
+              0.2222222222222222,
+              "#7201a8"
+             ],
+             [
+              0.3333333333333333,
+              "#9c179e"
+             ],
+             [
+              0.4444444444444444,
+              "#bd3786"
+             ],
+             [
+              0.5555555555555556,
+              "#d8576b"
+             ],
+             [
+              0.6666666666666666,
+              "#ed7953"
+             ],
+             [
+              0.7777777777777778,
+              "#fb9f3a"
+             ],
+             [
+              0.8888888888888888,
+              "#fdca26"
+             ],
+             [
+              1,
+              "#f0f921"
+             ]
+            ],
+            "type": "histogram2d"
+           }
+          ],
+          "histogram2dcontour": [
+           {
+            "colorbar": {
+             "outlinewidth": 0,
+             "ticks": ""
+            },
+            "colorscale": [
+             [
+              0,
+              "#0d0887"
+             ],
+             [
+              0.1111111111111111,
+              "#46039f"
+             ],
+             [
+              0.2222222222222222,
+              "#7201a8"
+             ],
+             [
+              0.3333333333333333,
+              "#9c179e"
+             ],
+             [
+              0.4444444444444444,
+              "#bd3786"
+             ],
+             [
+              0.5555555555555556,
+              "#d8576b"
+             ],
+             [
+              0.6666666666666666,
+              "#ed7953"
+             ],
+             [
+              0.7777777777777778,
+              "#fb9f3a"
+             ],
+             [
+              0.8888888888888888,
+              "#fdca26"
+             ],
+             [
+              1,
+              "#f0f921"
+             ]
+            ],
+            "type": "histogram2dcontour"
+           }
+          ],
+          "mesh3d": [
+           {
+            "colorbar": {
+             "outlinewidth": 0,
+             "ticks": ""
+            },
+            "type": "mesh3d"
+           }
+          ],
+          "parcoords": [
+           {
+            "line": {
+             "colorbar": {
+              "outlinewidth": 0,
+              "ticks": ""
+             }
+            },
+            "type": "parcoords"
+           }
+          ],
+          "pie": [
+           {
+            "automargin": true,
+            "type": "pie"
+           }
+          ],
+          "scatter": [
+           {
+            "fillpattern": {
+             "fillmode": "overlay",
+             "size": 10,
+             "solidity": 0.2
+            },
+            "type": "scatter"
+           }
+          ],
+          "scatter3d": [
+           {
+            "line": {
+             "colorbar": {
+              "outlinewidth": 0,
+              "ticks": ""
+             }
+            },
+            "marker": {
+             "colorbar": {
+              "outlinewidth": 0,
+              "ticks": ""
+             }
+            },
+            "type": "scatter3d"
+           }
+          ],
+          "scattercarpet": [
+           {
+            "marker": {
+             "colorbar": {
+              "outlinewidth": 0,
+              "ticks": ""
+             }
+            },
+            "type": "scattercarpet"
+           }
+          ],
+          "scattergeo": [
+           {
+            "marker": {
+             "colorbar": {
+              "outlinewidth": 0,
+              "ticks": ""
+             }
+            },
+            "type": "scattergeo"
+           }
+          ],
+          "scattergl": [
+           {
+            "marker": {
+             "colorbar": {
+              "outlinewidth": 0,
+              "ticks": ""
+             }
+            },
+            "type": "scattergl"
+           }
+          ],
+          "scattermapbox": [
+           {
+            "marker": {
+             "colorbar": {
+              "outlinewidth": 0,
+              "ticks": ""
+             }
+            },
+            "type": "scattermapbox"
+           }
+          ],
+          "scatterpolar": [
+           {
+            "marker": {
+             "colorbar": {
+              "outlinewidth": 0,
+              "ticks": ""
+             }
+            },
+            "type": "scatterpolar"
+           }
+          ],
+          "scatterpolargl": [
+           {
+            "marker": {
+             "colorbar": {
+              "outlinewidth": 0,
+              "ticks": ""
+             }
+            },
+            "type": "scatterpolargl"
+           }
+          ],
+          "scatterternary": [
+           {
+            "marker": {
+             "colorbar": {
+              "outlinewidth": 0,
+              "ticks": ""
+             }
+            },
+            "type": "scatterternary"
+           }
+          ],
+          "surface": [
+           {
+            "colorbar": {
+             "outlinewidth": 0,
+             "ticks": ""
+            },
+            "colorscale": [
+             [
+              0,
+              "#0d0887"
+             ],
+             [
+              0.1111111111111111,
+              "#46039f"
+             ],
+             [
+              0.2222222222222222,
+              "#7201a8"
+             ],
+             [
+              0.3333333333333333,
+              "#9c179e"
+             ],
+             [
+              0.4444444444444444,
+              "#bd3786"
+             ],
+             [
+              0.5555555555555556,
+              "#d8576b"
+             ],
+             [
+              0.6666666666666666,
+              "#ed7953"
+             ],
+             [
+              0.7777777777777778,
+              "#fb9f3a"
+             ],
+             [
+              0.8888888888888888,
+              "#fdca26"
+             ],
+             [
+              1,
+              "#f0f921"
+             ]
+            ],
+            "type": "surface"
+           }
+          ],
+          "table": [
+           {
+            "cells": {
+             "fill": {
+              "color": "#EBF0F8"
+             },
+             "line": {
+              "color": "white"
+             }
+            },
+            "header": {
+             "fill": {
+              "color": "#C8D4E3"
+             },
+             "line": {
+              "color": "white"
+             }
+            },
+            "type": "table"
+           }
+          ]
+         },
+         "layout": {
+          "annotationdefaults": {
+           "arrowcolor": "#2a3f5f",
+           "arrowhead": 0,
+           "arrowwidth": 1
+          },
+          "autotypenumbers": "strict",
+          "coloraxis": {
+           "colorbar": {
+            "outlinewidth": 0,
+            "ticks": ""
+           }
+          },
+          "colorscale": {
+           "diverging": [
+            [
+             0,
+             "#8e0152"
+            ],
+            [
+             0.1,
+             "#c51b7d"
+            ],
+            [
+             0.2,
+             "#de77ae"
+            ],
+            [
+             0.3,
+             "#f1b6da"
+            ],
+            [
+             0.4,
+             "#fde0ef"
+            ],
+            [
+             0.5,
+             "#f7f7f7"
+            ],
+            [
+             0.6,
+             "#e6f5d0"
+            ],
+            [
+             0.7,
+             "#b8e186"
+            ],
+            [
+             0.8,
+             "#7fbc41"
+            ],
+            [
+             0.9,
+             "#4d9221"
+            ],
+            [
+             1,
+             "#276419"
+            ]
+           ],
+           "sequential": [
+            [
+             0,
+             "#0d0887"
+            ],
+            [
+             0.1111111111111111,
+             "#46039f"
+            ],
+            [
+             0.2222222222222222,
+             "#7201a8"
+            ],
+            [
+             0.3333333333333333,
+             "#9c179e"
+            ],
+            [
+             0.4444444444444444,
+             "#bd3786"
+            ],
+            [
+             0.5555555555555556,
+             "#d8576b"
+            ],
+            [
+             0.6666666666666666,
+             "#ed7953"
+            ],
+            [
+             0.7777777777777778,
+             "#fb9f3a"
+            ],
+            [
+             0.8888888888888888,
+             "#fdca26"
+            ],
+            [
+             1,
+             "#f0f921"
+            ]
+           ],
+           "sequentialminus": [
+            [
+             0,
+             "#0d0887"
+            ],
+            [
+             0.1111111111111111,
+             "#46039f"
+            ],
+            [
+             0.2222222222222222,
+             "#7201a8"
+            ],
+            [
+             0.3333333333333333,
+             "#9c179e"
+            ],
+            [
+             0.4444444444444444,
+             "#bd3786"
+            ],
+            [
+             0.5555555555555556,
+             "#d8576b"
+            ],
+            [
+             0.6666666666666666,
+             "#ed7953"
+            ],
+            [
+             0.7777777777777778,
+             "#fb9f3a"
+            ],
+            [
+             0.8888888888888888,
+             "#fdca26"
+            ],
+            [
+             1,
+             "#f0f921"
+            ]
+           ]
+          },
+          "colorway": [
+           "#636efa",
+           "#EF553B",
+           "#00cc96",
+           "#ab63fa",
+           "#FFA15A",
+           "#19d3f3",
+           "#FF6692",
+           "#B6E880",
+           "#FF97FF",
+           "#FECB52"
+          ],
+          "font": {
+           "color": "#2a3f5f"
+          },
+          "geo": {
+           "bgcolor": "white",
+           "lakecolor": "white",
+           "landcolor": "#E5ECF6",
+           "showlakes": true,
+           "showland": true,
+           "subunitcolor": "white"
+          },
+          "hoverlabel": {
+           "align": "left"
+          },
+          "hovermode": "closest",
+          "mapbox": {
+           "style": "light"
+          },
+          "paper_bgcolor": "white",
+          "plot_bgcolor": "#E5ECF6",
+          "polar": {
+           "angularaxis": {
+            "gridcolor": "white",
+            "linecolor": "white",
+            "ticks": ""
+           },
+           "bgcolor": "#E5ECF6",
+           "radialaxis": {
+            "gridcolor": "white",
+            "linecolor": "white",
+            "ticks": ""
+           }
+          },
+          "scene": {
+           "xaxis": {
+            "backgroundcolor": "#E5ECF6",
+            "gridcolor": "white",
+            "gridwidth": 2,
+            "linecolor": "white",
+            "showbackground": true,
+            "ticks": "",
+            "zerolinecolor": "white"
+           },
+           "yaxis": {
+            "backgroundcolor": "#E5ECF6",
+            "gridcolor": "white",
+            "gridwidth": 2,
+            "linecolor": "white",
+            "showbackground": true,
+            "ticks": "",
+            "zerolinecolor": "white"
+           },
+           "zaxis": {
+            "backgroundcolor": "#E5ECF6",
+            "gridcolor": "white",
+            "gridwidth": 2,
+            "linecolor": "white",
+            "showbackground": true,
+            "ticks": "",
+            "zerolinecolor": "white"
+           }
+          },
+          "shapedefaults": {
+           "line": {
+            "color": "#2a3f5f"
+           }
+          },
+          "ternary": {
+           "aaxis": {
+            "gridcolor": "white",
+            "linecolor": "white",
+            "ticks": ""
+           },
+           "baxis": {
+            "gridcolor": "white",
+            "linecolor": "white",
+            "ticks": ""
+           },
+           "bgcolor": "#E5ECF6",
+           "caxis": {
+            "gridcolor": "white",
+            "linecolor": "white",
+            "ticks": ""
+           }
+          },
+          "title": {
+           "x": 0.05
+          },
+          "xaxis": {
+           "automargin": true,
+           "gridcolor": "white",
+           "linecolor": "white",
+           "ticks": "",
+           "title": {
+            "standoff": 15
+           },
+           "zerolinecolor": "white",
+           "zerolinewidth": 2
+          },
+          "yaxis": {
+           "automargin": true,
+           "gridcolor": "white",
+           "linecolor": "white",
+           "ticks": "",
+           "title": {
+            "standoff": 15
+           },
+           "zerolinecolor": "white",
+           "zerolinewidth": 2
+          }
+         }
+        },
+        "title": {
+         "text": "Cross-validation"
+        },
+        "updatemenus": [
+         {
+          "buttons": [
+           {
+            "args": [
+             {
+              "visible": [
+               true,
+               true
+              ]
+             },
+             {
+              "xaxis.range": [
+               -3.952154604781554,
+               10.903313786058435
+              ],
+              "yaxis.range": [
+               -3.952154604781554,
+               10.903313786058435
+              ]
+             }
+            ],
+            "label": "corrosion_score",
+            "method": "update"
+           }
+          ],
+          "x": 0,
+          "xanchor": "left",
+          "y": 1.125,
+          "yanchor": "top"
+         },
+         {
+          "buttons": [
+           {
+            "args": [
+             {
+              "error_x.thickness": 2,
+              "error_x.width": 4,
+              "error_y.thickness": 2,
+              "error_y.width": 4
+             }
+            ],
+            "label": "Yes",
+            "method": "restyle"
+           },
+           {
+            "args": [
+             {
+              "error_x.thickness": 0,
+              "error_x.width": 0,
+              "error_y.thickness": 0,
+              "error_y.width": 0
+             }
+            ],
+            "label": "No",
+            "method": "restyle"
+           }
+          ],
+          "x": 1.125,
+          "xanchor": "left",
+          "y": 0.8,
+          "yanchor": "middle"
+         }
+        ],
+        "width": 530,
+        "xaxis": {
+         "linecolor": "black",
+         "linewidth": 0.5,
+         "mirror": true,
+         "range": [
+          -3.952154604781554,
+          10.903313786058435
+         ],
+         "title": {
+          "text": "Actual Outcome"
+         },
+         "zeroline": false
+        },
+        "yaxis": {
+         "linecolor": "black",
+         "linewidth": 0.5,
+         "mirror": true,
+         "range": [
+          -3.952154604781554,
+          10.903313786058435
+         ],
+         "title": {
+          "text": "Predicted Outcome"
+         },
+         "zeroline": false
+        }
+       }
+      }
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    }
+   ],
+   "source": [
+    "from ax.modelbridge.cross_validation import cross_validate\n",
+    "from ax.plot.diagnostic import interact_cross_validation\n",
+    "\n",
+    "model = ax_client.generation_strategy.model\n",
+    "cv_results = cross_validate(model)\n",
+    "render(interact_cross_validation(cv_results))"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 325,
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "[CVResult(observed=<ax.core.observation.Observation object at 0x2a5c793d0>, predicted=ObservationData(metric_names=['corrosion_score'], means=[3.33038581], covariance=[[0.0086991]])),\n",
+       " CVResult(observed=<ax.core.observation.Observation object at 0x2a5c79df0>, predicted=ObservationData(metric_names=['corrosion_score'], means=[3.38045204], covariance=[[1.14749778]])),\n",
+       " CVResult(observed=<ax.core.observation.Observation object at 0x2a9d89430>, predicted=ObservationData(metric_names=['corrosion_score'], means=[1.57425067], covariance=[[0.00012101]])),\n",
+       " CVResult(observed=<ax.core.observation.Observation object at 0x2a5c79cd0>, predicted=ObservationData(metric_names=['corrosion_score'], means=[3.74291177], covariance=[[0.00775312]])),\n",
+       " CVResult(observed=<ax.core.observation.Observation object at 0x2a5c797f0>, predicted=ObservationData(metric_names=['corrosion_score'], means=[3.35092202], covariance=[[0.00104975]])),\n",
+       " CVResult(observed=<ax.core.observation.Observation object at 0x2a2ab1a90>, predicted=ObservationData(metric_names=['corrosion_score'], means=[0.75377368], covariance=[[0.35887005]])),\n",
+       " CVResult(observed=<ax.core.observation.Observation object at 0x2a7a8c8b0>, predicted=ObservationData(metric_names=['corrosion_score'], means=[0.03287673], covariance=[[0.01573419]])),\n",
+       " CVResult(observed=<ax.core.observation.Observation object at 0x2a2e09e20>, predicted=ObservationData(metric_names=['corrosion_score'], means=[0.28427335], covariance=[[0.02445275]])),\n",
+       " CVResult(observed=<ax.core.observation.Observation object at 0x2a7a8c730>, predicted=ObservationData(metric_names=['corrosion_score'], means=[5.1717073], covariance=[[3.27016158]])),\n",
+       " CVResult(observed=<ax.core.observation.Observation object at 0x2a4643a00>, predicted=ObservationData(metric_names=['corrosion_score'], means=[0.23590281], covariance=[[0.05287956]])),\n",
+       " CVResult(observed=<ax.core.observation.Observation object at 0x2a5b707f0>, predicted=ObservationData(metric_names=['corrosion_score'], means=[3.954885], covariance=[[4.13191409]])),\n",
+       " CVResult(observed=<ax.core.observation.Observation object at 0x2a5c79fd0>, predicted=ObservationData(metric_names=['corrosion_score'], means=[3.79301819], covariance=[[1.57803025]])),\n",
+       " CVResult(observed=<ax.core.observation.Observation object at 0x2a9d89550>, predicted=ObservationData(metric_names=['corrosion_score'], means=[1.57698842], covariance=[[0.00856682]])),\n",
+       " CVResult(observed=<ax.core.observation.Observation object at 0x2a5b32460>, predicted=ObservationData(metric_names=['corrosion_score'], means=[4.39157658], covariance=[[4.529502]])),\n",
+       " CVResult(observed=<ax.core.observation.Observation object at 0x2a5c79970>, predicted=ObservationData(metric_names=['corrosion_score'], means=[5.52922756], covariance=[[0.00146855]])),\n",
+       " CVResult(observed=<ax.core.observation.Observation object at 0x2a5b70460>, predicted=ObservationData(metric_names=['corrosion_score'], means=[0.97394723], covariance=[[0.36594216]])),\n",
+       " CVResult(observed=<ax.core.observation.Observation object at 0x2a5c79ac0>, predicted=ObservationData(metric_names=['corrosion_score'], means=[0.54739115], covariance=[[0.2633537]])),\n",
+       " CVResult(observed=<ax.core.observation.Observation object at 0x2a5b70280>, predicted=ObservationData(metric_names=['corrosion_score'], means=[1.38298294], covariance=[[0.0981112]])),\n",
+       " CVResult(observed=<ax.core.observation.Observation object at 0x2ab3ab3d0>, predicted=ObservationData(metric_names=['corrosion_score'], means=[5.12117681], covariance=[[5.41822595]])),\n",
+       " CVResult(observed=<ax.core.observation.Observation object at 0x2a48c7820>, predicted=ObservationData(metric_names=['corrosion_score'], means=[4.41398979], covariance=[[2.74877288]])),\n",
+       " CVResult(observed=<ax.core.observation.Observation object at 0x2a5b62730>, predicted=ObservationData(metric_names=['corrosion_score'], means=[0.0970607], covariance=[[0.00400847]])),\n",
+       " CVResult(observed=<ax.core.observation.Observation object at 0x2a5c79f40>, predicted=ObservationData(metric_names=['corrosion_score'], means=[3.74686509], covariance=[[0.00016273]])),\n",
+       " CVResult(observed=<ax.core.observation.Observation object at 0x2a9d85d60>, predicted=ObservationData(metric_names=['corrosion_score'], means=[0.12282408], covariance=[[0.01368547]])),\n",
+       " CVResult(observed=<ax.core.observation.Observation object at 0x2a5b62580>, predicted=ObservationData(metric_names=['corrosion_score'], means=[0.01725345], covariance=[[0.00132096]])),\n",
+       " CVResult(observed=<ax.core.observation.Observation object at 0x2a5c79e20>, predicted=ObservationData(metric_names=['corrosion_score'], means=[0.64291407], covariance=[[0.42318268]])),\n",
+       " CVResult(observed=<ax.core.observation.Observation object at 0x2a7a76e50>, predicted=ObservationData(metric_names=['corrosion_score'], means=[3.75959764], covariance=[[4.92338383]])),\n",
+       " CVResult(observed=<ax.core.observation.Observation object at 0x2a5c79bb0>, predicted=ObservationData(metric_names=['corrosion_score'], means=[5.90193752], covariance=[[14.30188832]])),\n",
+       " CVResult(observed=<ax.core.observation.Observation object at 0x2a9d89ca0>, predicted=ObservationData(metric_names=['corrosion_score'], means=[1.25197049], covariance=[[0.43754259]])),\n",
+       " CVResult(observed=<ax.core.observation.Observation object at 0x2a5b62880>, predicted=ObservationData(metric_names=['corrosion_score'], means=[2.22755944], covariance=[[1.44631305]])),\n",
+       " CVResult(observed=<ax.core.observation.Observation object at 0x2a5b32280>, predicted=ObservationData(metric_names=['corrosion_score'], means=[1.10185481], covariance=[[0.3645476]])),\n",
+       " CVResult(observed=<ax.core.observation.Observation object at 0x2a5b32a60>, predicted=ObservationData(metric_names=['corrosion_score'], means=[0.52667369], covariance=[[0.06076658]])),\n",
+       " CVResult(observed=<ax.core.observation.Observation object at 0x2a5c79160>, predicted=ObservationData(metric_names=['corrosion_score'], means=[0.81682378], covariance=[[0.00202599]])),\n",
+       " CVResult(observed=<ax.core.observation.Observation object at 0x2a06d0ac0>, predicted=ObservationData(metric_names=['corrosion_score'], means=[0.30203122], covariance=[[0.06494224]])),\n",
+       " CVResult(observed=<ax.core.observation.Observation object at 0x2a48c7490>, predicted=ObservationData(metric_names=['corrosion_score'], means=[2.92546491], covariance=[[2.91704295]])),\n",
+       " CVResult(observed=<ax.core.observation.Observation object at 0x2a5c790a0>, predicted=ObservationData(metric_names=['corrosion_score'], means=[5.5097873], covariance=[[7.68646695]])),\n",
+       " CVResult(observed=<ax.core.observation.Observation object at 0x2a5c79c10>, predicted=ObservationData(metric_names=['corrosion_score'], means=[0.81779914], covariance=[[0.00204636]])),\n",
+       " CVResult(observed=<ax.core.observation.Observation object at 0x2a5c79be0>, predicted=ObservationData(metric_names=['corrosion_score'], means=[5.52898111], covariance=[[0.00526972]])),\n",
+       " CVResult(observed=<ax.core.observation.Observation object at 0x28ecdf670>, predicted=ObservationData(metric_names=['corrosion_score'], means=[1.73056953], covariance=[[0.32395525]]))]"
+      ]
+     },
+     "execution_count": 325,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "cv_results"
+   ]
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "ax_env",
+   "language": "python",
+   "name": "python3"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.11.8"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 2
+}
diff --git a/docs/tutorials/freq-vs-bayes-hg-selection.jpeg b/docs/tutorials/freq-vs-bayes-hg-selection.jpeg
new file mode 100644
index 0000000000000000000000000000000000000000..42a8243851bdfc13e7d5129d6210fac7770a6244
GIT binary patch
literal 1260172
zcmeFZ2UJr{*D!iOP>Kx;JfIYjB4C1opnxDKMXAz-kYJ(r7HUFMq^Srhk&d)b1eB5h
zp$bR|5F|i=(0lJK)Iaz<@Atm%x9)fEUF*O9zt;W!XR?^H&&-~gz4x3mv(MgpIP5tb
z2F|J~KT-y0sOdV*bpSXVr8Rr>;DPC59WCWY>PplK0HD>gdtvWNcLM<I9o$@WRPOUU
zF*4>kF%8fGEY#mEK<3#CS0{zXkJW+yGHrGP)Sdu<2s-NPzeM}Lu3$DZcYOf>Gz`>g
zxff0@ZdCXo6;}6fb2@@QQsE0Pte%-s;Y=!g-GxdZ74AGLfBq*teFR(n3GW@jE_yl-
zsJxw}!aSD$0zdy3_=S~=Jr#$_IlyOc2dC2TI~{Ry(%eB`mwM%;{#^!K0UbaExDW6E
zj{qg$0U!z71f;0BBme@$05J6q_%GTY^?e!81RSYtUr=)=zzc8ztN}~F3UC9iQ%m90
z-p#333u?JJ^=Aio9m${Cj+*`{S62&hvEM~BC+`CQ{ovu@8b1J>_y7R=?+*`mzZ@R!
zrvSjQegLR;_-A{^WB|CcK`np(&oaIw066ms05G-xEPIXu0BkS-{L$<5%;nkdagI@Q
zS_=yRSVaTCDI)+l-wpswkUwQctvjj*Qb_=y=R=*URse{J1pomm{uz@)GF3)L&)ok^
zI_l(qrv2}>ht&WJ!-@IR>c?nU09qEBV=Od>bpSW@VWpv?Jw|iH!k>wT_821*9RvOG
z6U<bE_teLRp7t0W-H8*-jLa<OY1jcO0v)xMfrathm7C`kw4X6ua1Mz4vV4)1_j}E+
z5w<_r#dHh|KSdKeuZrK-kNVirwZbQ*_vE=t;0IJeH^0EGI}cQJUEv;{L9bIV*g6iu
zYZ8*T6_u3T++PJ37M1*{t$+B!;^pV!pCln^GjmI;)U*tA`lwM)XzAE0l_IVqb^t2L
zM{n$Zkj!xW1S35i6SW?}LZy(7_5}TLI$An<W@cK!KPhH8e&xiun^ba_&mWP?aN&qp
z#*4fwVmho>?>}ez!zHSNPh9f$9fQD6Da9E6Te_}paCeV_BJ7V*0ST!G%KC=pA3hec
z3+lZHqVm@D<Yj8TqLSGwRNBw(;E?o^I@0Q4A8?W?N?Mj<EWkZrX!sxW{eAsg0)I>3
zZwdS@fxjj2w*>x{z~2)1TLOPe;BN{1ErGu!@V5m1mcZW<`0pYC`r`#x#U{-&D*bNI
zN^35ewgYYngH_4Ej7)`}JWYp(_ksVhw8f2m;KZYWxgFs22OMb~;1u-8p9cO$|64MD
zKLmfj9)GU{{~xcB2?rcAwNR;j8pJ}C)z$wWufB9kYqC<VW9#&rf}H;E{@-pLjQq3S
z*3`Y`qW^!FP4++Yaay{{#B}rCpg+J!(z*u}@39X5FXxqpA8j-<*{M`HUiCJB_J8?s
z|C#0gbk8;Tc&_9>mHqF#rGL2ld~Ht|DF5F#<p1W8?#%8x`}G_G-qdj!iCXv0R_w7>
z`~L?KXDt+!ExcJ=#R22ny<i8y)~r26I(>}!2~gl<m39CN2?8O^YzhASET<-_5a2lH
zV?Qt$qBGvpQZOGpXdnHDy+iq0j+$Ko(OLr3U}BVh>I>slosWflX(25wA#bYFozoiP
zI76MeE%b~)l6zh({SR8?@LBvtb6|{({m=gUQAHpo9fJ@2d@XO@&>(1zo0(LzcCMZ$
zn99>OwLXn_Ai-u?Bfh!=u(qZCQQTEKatO4L*533EwN3}&gJbtu7`9Ze_{GkR+xrgv
zA%NL+><M2P7`=ZwhoP#K4ZWjt2)wMeIs;GMD63jfyZ7Gz4E*UKkjlM@_VecjZHfAt
z^3%g+e#mLD9|G&Q?*%eoM-z7Cw=}tTgbo2)*KOXw(^(9q<>nLigkgW4+0NA|e*XQJ
zRlK(Vm9q7v>9m89j_r0J37$3d%bs!zR{3^7-ZZ~dZkBimq{JNp$i1_4KDn9);kNr*
z%kmd}vNR6fzUasw_UD=<`lUYE@2Cm{BH&p`zkVG8Zhkd@Y*&V#ipqZPp0N$!ldpDf
zSMK&^tDG31z1Vei^QYh@AsN`Rd2ceSeGjt_FlpkHMRzUFG!a>onPiuLF(>G-`>mV#
z(>t{j4*GtGip&7A%k77N{WI4?K;<EAzDH&f=9?cj4cPJy-L*Zj>n(rABNN@6Lxt{d
z_A!N(D$D6VG$ls?U$fWKk-7JUt8(2Uw7?0;j9bRhCbKMRRHoQkfaE#7z$u*&n4zyf
z{Rvnq%QXC268Ryu)9_te2He~ASs=n=+Pp3I7<GUPoDB*fc#Z;f+FCmR^biO*1QyD<
z7iIv%rFQ|l&d0WX$kUc&3O0YILcTZXc1OPW_Uj%32r8pB_xG{rgKc?Q<nrSx`(fX0
zdFY)$Do0iJGlxLsS-^*iORhN>G`&Hy^g4(z;3@5gMzsJD2t97HCpWo+3b~yjhxskN
z@)wec9R$O(t-Z{WX->F7f(`+PnM!Mo9WQ^OtuFh60p0SB$9^n9)mvQ-z5}{dh}0>X
z?jNc%SglVMiGJHZ<eccOORq5v>wDg#cAGzR>cr+?((VYrD7?==(VrQrjGJ5Y2TMRw
z@9}5a=|7bLjXJB>JLsyupEXjrvl?gXgES-BsP+$B?N^BU^o1+M6#nUMrE2o(^MAgE
zv_7{+@EwAerm&6Yv$ddV5f13Oy&&|>>*K{Jk@*6%ggK;tn$<akFKk)8(j?%o*Oyjp
zY-EOh+j5)<A5kL)reFDE%%n<3c(UpOJrnO$u`8JDY~D<s$G|xi^f`L9@nja;>Hve@
zOun5Ed~{_cufT>9*-<^<Ve+5d@eyE%T}4GqYdj5gSBR3m^Iq7LpN?8b3l{@Pzft4O
z-AX=DTh%CRSuaxqFv^V*OU}m*eddh6i|)JZx~K{G12h8Oi7C8B*BLIH&#QExvGvTQ
zB8rnj1NjT&<^LQ7NNnTMT{tbZB?wSU64VK>@7F`3`;VjZDxR4YV`=HBXqshS+w}Rv
zSUNy32gT{N7f|J3BKI`w@|_LtQ$D-Lxf3Bo8JAI}h|n0;SnIl?Im+(#`)0F{9G@Zn
z$|dymj~B9}S<7*iFF0)TFl554f3n@t*qXt43D4Y$w*Z8k>9x=#e975sRNjhNmKS-!
zzWvI^S&98!hEK0?P*L^Ey{#9}PusF#`=%jRVcu(URc|16n7V;wIwop$f)~p12Bv=G
z8#Th@pNa1c{(V>oPF`Gmk|89HhBgpcfvY80@c@gnr+sSpEKJJk^pwa)qcT!aczFe?
z<&yzHD59UXdj%5syB{sfGeVoemv=w-U%n;RLDXCkFLQzny%*-?cafEtZ6}cCSo9&q
zqV`*g>+AIV(-#pnNBg&+rV6=HwD^ua5TVJ(glbYa>g%uh1GcxkXv9V<hS_7*SXj<K
z<$0{9oTuo=)-0Aeke$T+UWkKkt`R3ikICmurgjBr#c4S_8$;~;IbDpT0F*=G7f`Gy
z^G@K=%mC+<&Ffz2DK*KDrjGj7DSXi78(MiRpX~Lr4xo#SabeMBvpSwCXTi{^5yEic
zsDWnrtw~c+$9De0%6f{^ZXiAVBhVP}{X3{$Wq%FlfcF}S+;MpeVqUg-tTi8d6Jek9
zkb&BaZmvNGB#W!V80c4-1Q=wAjaBdZ!U#87<HMt^YbP4&&U&-6d7KeqGj+;x)1OLm
z$y<DyX7zi(Sg-A%rC|*9;nAHK%8yqQ7&KU>Lp<N3ku66A^<5TQf>EE;--$+g18$oc
zJyh9E7vnH(hTMCl)Lka*h-ufk^(Vq#d4iUP^s55cZVhdqr}I?4VF&v}tzXa4<TFKX
z)=zD}q4E4T{Qpij&RsjBhrsuPb*kEEy6Ukt)5AODv4sD$9q2Y}nUUV0WQ+E=A{xs-
zUkb%k$tM3f2Y(8HHEG`(*0(o&hG8;`l+%jlx_oPUR|bHS0vfbq=Dw_qpidtHsKQ8B
zxpuL)iWsQ<3_^-E)UprZHB-I59nmIlwejYDxSgKU<X$kMS@DvXd4h`~PN1Z?ASAjo
z_U=y+7maa#HQV}`wuc_+_DseabXKYn_isqV_TDbWomdcYWS8`<DMLTEABYy_Uve^=
zoh?r_`NJZ<Yhov)Zth)wMte5f1TSCoqOF|ZTZje$8N&ir)sU98c;A<D%UTMNkBtKD
zzQ$kqypRy5FNQ9PJ3$QneADCR{SvE!^U8I0f)&CSVhwnn4K%;X5+#b<T5p4hCrgiM
zaS9BlO9`O#pA`$8zA047!k*l8%q8Z1lrC6uR8J{|Y;5p}4{z#Uw5-kU+;!X>&FKa>
zL#<62I=tt4@?_(Sw{b5m<X74xXW-nH=SZ+xlBvUh;#?j^@bgI4>5V1t#y#}HYPiq*
z2~KO*h+{-p-d!nY;^>mBw?C`rDtPfS;vi39d6v4^^cg%9-+_<JUJ-3Flo33=n?9dd
z9-57+wEZ#nErB3Jxh48W9M7k?vfgxxjZxa=5yg#85wb*hKILE%AkF=3=j7vC#KiPT
zK(|b#YnrNqZK-{SrHUH71`;ExI)Sfo;EuH<uC{!~8?>`VSLJ+R0<a4BR6kIzo*2I|
zZbE2Rgeib(><1iW{qySyrkHz=W>)5AE?o5dzL~VU6i;)!Ieb%fjib_Ooj%oSSF0tJ
zB7Z?Bo!huPHfB#{6}UF7KQ!|gDrqVQIEXApY^k1+n=jj-yJWMQPKh^d_GX-#A)z`?
zhZ^sf9Re6pt%%2jz#XLB3FZ7ULyuZhLKbDOCJ};^^{vhoGIAO+O;=gn3mo?^d1!Rd
zWIkc|JrusQsOFvB!^S8KmYF3rTMm$W`16a!mbO#OXOk1^lQ<{1KOZ=|3^*(<nx2I?
zGl3*@%663|)(CT>%re1SHtRXt_bqW<hSSMW3Ml@Zj{K41i1UPb#$SHZVe3h4*90Hz
z79UL45xmi;{_|yd5d7HwV;BjN9_eYf(Z|#z0xn$3(FTW=+KTd(3F7aa)i290yJw2U
zUeW#9<-OS3a_*t{epE!${L>u!=}^K%Qs>1FDt*3iM9tLD?$k6>XsX=M&CG+xCLd-o
z6xx&*0ZtR+YiqK|(&PJ-RfuVoWwOXLB7I8q^Ig%2jbmSAy^>I-sgIYJ@v}5V$P8b<
z_?I=hmH1^yptJ;wt0CG{n<439WrF~6i_Xnv!Ekpe0;*saw>G&nAWNe;vLR$zrj5t|
z`y9w1&t6y*jrd`>7}2#Jly8eppf261rUUs~_qRpSoh&6)Wu}Y%ryAsD{K|F|D+%j{
zX~&#fZw;C*ZhK3d5fx3SF;tuy5ZzK)qbHVnLct7@7m$^o5a|ok7f_0QnU8MvsZ5Kn
zx&!3zE)#RO?ybAJD;&^P>oWVuQcgDE_KJ?K;|p?^gyoF)cn#vu<V0{j$Q?{_N!&Cu
z|FZE%NG4We`oy_0a{g1ZQYY+L`Q*er1bkxk^?r;qg%&vT#GBT;bcKX!zO-rHFn+Yr
z{9DpYM}uX{_5zI@v0zq;KXeN1nbRcN$kHh354a916iU!T%AD`1I688DNw6JHMBFhi
zKQ&0_o46>=oEwWq-OXbl#%#K>?w)^FE&m@KGx82w_M*Qknf$sl)PAIM{3G#&!|0eW
zt@7>L9&_l=Y8i-JKPtihsHI;2LioL(16+?6|E!{sUMA*1@AnS*{y7NasIm7$_2>gt
zM&8c1#v7Zi_tILFAi|R5l6kEH9-&praU8B1(2S*&1TBnGx5fXPMC(RstrK;;_>Je^
zkj|sU=O@{E;@xTgEWT)CY@JcDvo~}n8Zx*!;h5ikOja$x>e12y%Q<}P02J>pbfiT+
zvL+4GVzCjA!;$@KBekGmE_4GH73a_Y0elJJO2<T{cb&zc$w}pkyZ@0EW7*px`u)!Q
zzf(x7{-*SONG9&WKP1P<OA4UPEqpBhCoM+}IW}yQO=TSWj>qGE?0Nyn;I7(fmOB}H
z2g3nWtcx6OymF~TQYWwUY*xiZ<n6%7-)(8M7ng~_r|)hjS$2rU_72@$?TvEYrM)}o
zoX-tVF`__P3K#8Az<TGlhjsqdg(bR*4{{!u&XLC7ildtxa>uK#%x=OD__s;>lPpHC
zAVG75jjz{No}0SM@nj`kSUO^iR!hUvg+<hEGmxH+QheqTEnE(nh?H1eV=u598}ae5
zqrWPj#cc=k3#n%F|Ff(Ae)$Pb(>&2#?j`wiiYjO>%Z!A96ya=BP^iz^Fw>iG-S*=4
zi2TUC)N46rWqq@87<A)>Q~p4XhMEqTo)BRver^*YJed=1;?5)VTg9Y?d7oy@)OPLk
z=i(gSL>&_+7B-<<(ape!kg!3T$r)Jc$hi3Eu|(ghWAFQ~iVZq#=@Le`;$^lGWeE9`
zatjJ%?N7O}m|aY<fu5at3-WpFdXBEBL6C1GN4ec1w+f_Ax-<QrRxMsbSWH4Iw_rji
zLowspE$^=%_~`RdeyDjZO>T_x*{=zSrK8IM-!{>2cONc5EQL0qI#|c<f*k4Y)XqQZ
z-+*f~vynTk;9mt(6894Xj$gr!KFWeUsenfBjNRzpeQ?J5O=Vq246iBf7N)g#2XWkM
z$9KpNt6FIEjfqu4z9C!rSns-s!*K6dh^I$TtM0PRz)P!9n8hjuTpQp%a9gPnYa?!T
zv8Ap$Th=7$LC(YQhe6INd}LxD>|?LfDA;|#ZRqD*2N)q5#pcl3?^2br;}uiX8&l{#
zjE^?VM_Bjf7m5GjJm@a^tlr=_hf!%>ri@MvnPWMavvK*4yeFOaII@kl_9LBcEHF*H
z<i|b<8#`|e7p6&ziHgk*#X9;$)Rw?WCWJuqm!b!(8Ut^OO4w0t?u^52Obt07$tkmF
zBX*a%olp1Q)lTbnlELl8^LOAU%rvb9TFm=D`FLg8OFLHaJMyG{4Mz^0iOSTQ!$(98
z>}TLyM0W3ePi(|v&W}fMyXN~<C}ebgKbQZ!V=>(LrTuQ^yVozz(Zgb1V?Rkwl-e$;
zpd>YfKYxCFZKdg&yQ%sb%S+KKb0qN+FF(Sd^uT8pPYX85(qlib$3DS)!sPsTz=k{o
zUK?&S8`LrA2utits5YnTEzFG$hARcul+SZ?mDShZAfyC8fD`W<U<Q$c(CXQs!VxJu
z)v^iV%ycW|v4vjzv)9ejHFof1#dqVM(ogMv(L)8RsG$o!7zt6tGO%1Skhb#K$Di1h
z)K&_1ocnNF=I@o}@_X7Y<~a4CHgX(>4EXCG;_S$kZ#7F6d=dvURdOEZbPAUnOcI5Q
zV&iXL4wp`nyP8|Yp>|pbdr6y}ohwae7N^r3?r_Q7xy)`=iSW#O?knMDyLDG>m1)Sf
zk1NN>#t6Y%V85bHSfH3R4no9+ltQ)<1_PUo7eC0EN}*PR!TWFZR`nY%u5+w{5VNeC
zec1^+RXi&9p@SdTY7Fs{namc03F##A3IO7e+T)gsMe;CAH?D(j&8_)(RbNh15hp5g
zqqN=^jpSRgE47gbLAj1^J11{nLe<<;0auQanr7yXseo7-%StFM2XZhc`XM&n@u;O;
zDDqx2r#;;|9aN>EyrR6CNv}cf^4%n%6Htc}?Aqa(m8SPni(YMS3{5y0>A8Ya<nk)i
zx-v~)BfDEz%Z|xCx_A3~*op9*?@9~(@-zkJwGnBRI_Da3AL_x(7EPNrzFMdd<NgnF
zusF7cr`M);-b!<oS>xt9CAXZF@a-4RTQ(N0A^PY)cnjVz5x_+dYG+QD>Enl)U-gx;
zu*LaTz{}YleV8M5_}yeo%BaWRV-Bo)z4b^CEiOk~pS|Rs;;bgTfA+HMkEN5_6m&;*
zl!8YR*2%J=Ne)Y)^_IQ&-1p^yJh{SqkscDf#XozpbYVCrDO1pCV;rWtd87CH-LH)O
za)qLf@)ZN#*F^Z(z?M@KI>K9A?wY)b>Y2j(%#hfXx$)?6<}Al=#Y^|><sM@%!p89=
z#~LZxS!20F;EIc=Tp;8%q)SX6c6X_5mN|{^J*jms6^4ZOGnL6ov0G1XoXi$p)?(ht
ztpsvXs7>U3lNJ{yx@wLOS#k~dy$50WChSmzP5M-8mBtWU3F2sjawD(kZ=+i-#HoP=
z(P^RyUn_Pic~VnIiA&hepv0C^;XN_ULVDRV&N~mKEGG(ocCW3_eSp?a$@)5B{c1<d
zDV~8Twn?;Wtd)t?y$<_}rWLyZd~Ib;hk(!OXWaQM)3U&ko0Ds(#`|_Xi3U5#f?pOz
zABE#Rg;#tR<&d&GG!5;=6g;tEie`BZ)$)DRxb5Y2{N9`Xb+%}o7lN?6USi%9+^eDJ
zIaBaKsaNvdthJFwz23xj2_m5y<VRm3=!wtA<z#B;LCH&1o>ZgF9J#9nR4_o^L@FFv
zNrXr6u*)0ml+BY<nBMtfqUBiyHm(k^;P^#DFjJ|ZCkK4PHls>KP60_?@Ih|c(<?1|
zqRvh9HhgYA_=@5uH-P-#Rj8M)5Me_+xUb=bBIIDtO9Sx?5jvs{7U29#*?FoJcE@xL
zedE0Y8wU*`j>pRJIuit~5Y2}eC{A>RE_$(ir+97{nXuzs2eq{uvhPK68LrcDWy^Wb
zkg7|VqHXMoNfx%Bd2lkGw*AQa`CRPeSVwweE1wJ#^=+<-a)$MDTN-92A8a76ZBTUV
zEbv+HRic8nzK7AB!Qr&Wzx4zI>-J(2^bDCIbDV0sH`3X6D_77GA2!*kDB|de$!py=
zIYxrrtpV9#X!6n<BA#s`cD91PmYUb!*#|XVbaZJiB1X>T#jc$<wC~>Pi!NUH#**y@
z=1Ai7jG<3}w>5n#P@;<};+6H>KpA>CTem3Lcj@#xK=&xQn>6K$(5bv~x4B`C?3^Z$
zT|Vvx^L^9Pu*6&ihp%m0Rqnfh4JqO(80tOCKsGOCbC`R=QqiY`cD9~lGYrh7<n9Ak
z+qV(t?%ffFitno?0tp@;O>R+p^v2y`v^U7J!GY|sD=lX&zpS?J>`-=s6wB<(90m$I
zX~iE+BcUeJyEWREQ5w5z1C8`t<!(f5NUoZ>&vN6blSS<%D`Qz)Rf3`qQA!o0>jeer
zl|G07>6tI9yd<{@l?GCs8f>Q6pbtMA<88)Z5R}ClJUI$;hQJGQi4FYAbWtT%NGP=2
z$Gd1F5G#|hzb7_QdEg%fx|p?hZWaWe9b0>oJ%CXw^~lYw5>4j$kXOciddT5{>YVX3
zcBH#qfh(zi2t!B^HRZGX<_F2*#Y;tJjp55nqtA1<Eb)B6`m8MCr#xyfdPRwj64F?9
zK!W<f&(~`kdaM}tUuPJp4Ms)W`)smCz=4t;YgQGIG=EGDOs+FJ&iCyZTHfQnp`4+F
z)($t*uUw~NVj_5G(%0>_tZ&Y<YuHYf3i7FK)SX3ISY}t0;e&ZJ8bdM%v0xvChwWev
zd~_^|x`aEYvf0m9kSKv?G`(kiH4))C8n-sWP`$Q-zO6ZxTUE(dGuJ_GbE{kME!#%u
z3|y~ranYxgY(-y3vCTKAx|S1VUyLP>({RPcu&^bWs(cU0^5M<S4=Uf0S6-Px8-5R`
zS{)AFpFS~ATfY6;njrV+J<nq;xjdzY2wAD5RmMW!E<6a;lT^Z2G>Y{54nD}ccPAY2
z!DAB@y;te)khw8JYCzrSwt>j@+V>n^;$%*YhcC?WZmzw?7(!9a{kMfTh8e`#*KcY&
zpycI~KM?P?Z*RO0^~oiIni(#fGf@vOs&P?*0oGllyx5&mM&xd;kM{!g3uJ9(q4BP0
zk}4xei)x?xAsR8p4iS92I-0Y*O_RkfCkyE>620Pmzd259C<ik)SSio$3{DwoNUWf)
zshOqVjk1Y?{>ovlC0bD}?vOeSyDmh*b~2rr5?+^ta!;r*K@4YR<Mhu*$Sn6&0Sp_m
zR2^bue13^)wO8ZM^~po-d0CDhkG0O{RKse>R+Q2&`IwuJ_-u5y-|EC`-ep)HnHC>u
z=w>Sbt+wh=C@GffbTP4>Y^OVh5y<DfR|`$H!Pcnct1vm(i6QbhsS{MN*ed0;gv}K6
zi)69>vV)V0l#tBMU0NC2q!%#HD}&FGCq(ab@b-}g960gAjN;%F-nkv=3i@hArNL}m
z<?bHohR!WU((nu^uks6MC%>$F<CxC;EHTN{v7HqrvsCDDrbPHExH4^qkhJ<H8@#Yo
z*_!n2s@{<;@3Q)KxTSf*8i7yU|0?g&Ro>20=N%et!>E*8l!K|0az&a-t)@wM3B*ZX
zeF<x`Mwb|bK5d>C@3o_-mb=z7X9`M`j}n{0n+BRs?4umxx_MLP$&Q`j1N(_wa{e~P
zk!!^J#G>%Xxvp_$&?iaQoG%kjhrC<eSu|>*zOlybg-|fBREy#-*NxhfZ!elcbG21S
zxsJ0z=JoOK<`#O3&iYiCczU3?hLZKVhoTWEFt?L3ItxB+>{T>2+X>-E!MhJ`<GZU9
zPT~XLn;Y(oZc42lPD6F4yx$b_P7F-(VAdhR3sb|%AGEpRrC63(II<9%*%nM*Ru;yQ
zbKMe{>2Py32YQo}5u1F*wxqqpXYs#ucU7^NrJ=eDGqt8#9+oKWoGC{z!sjcEZ9}6s
zg~lG`4Jki+jqgnv-c&A77(9Uus*y&CgVguZ(NCkHAsEF)$O>KL@-EcFXV>SI9$`B-
z7ax2>$J@5{eFv;}fo^$Vs;-nZVb3!(5z~UQ8HzTdJyTM&$f0TL`!s=}bXFN^vN2Db
zU@yzV3xoHg$Nb)3eVvfuYcJ>QT1BUIiMr$x+l*<`sc{GFR+e4;3|PgX7AieBg61;p
zk*~ZzUF2UrqutwX^J6QUdoP6=j11qIp60%xXIX|&(_NStXndvEtsdP}>nNMe1Rqw)
z|5!B9P-N4aO)jOHQGlqiti?He<P3T>q|`_v7i$AM1io>gBWhMz(RJ77s|>w+#pt*e
z<C!BY4U@)+ZH&1H5?E$W&SRPW5%0Kp9ggy*)X)a3hJPw_^{O17q6wU6`hi?zt*tGl
zuis3<67ql#unZM6$on9GC0a>s0kc%XF92!y;z}vmQFH-kL*0=stMmdEX2x5a2<U@{
zAz_$_hOu&%#%)xbwtI%!*i(aun=Gy)1Q_uTzp(6if_Vv#ka*{Y*f7%cVF!OyV`&DJ
zV!B-B2Oza}CUWjCyUM4mi@{koNE3B7;~Db;$0nNN#$16Z17Jqpv4%+3wbgOIvtgy?
z#qOT=eMx7P(`$nr%Dwjbn9T`tta!`rVRxt1B&+fxmznK9d@q><9}Q+y{Vcm;R|Uyt
zXUAU|GGyC;y7!z#8}1hl!bv|qXPyx@(NenSKMondxq}Ixc`AE~Ca7*U-;|4fL5`U2
zmEPIp4_yMO#DH0oT}3r14pi%P+#Om5e1u^Q5gTEZ9UgrMbosJ7(Xn`~PeSWS`QO8G
zMv&pH9*{J;JuYvrl9i=St*(OI%?XqU%@?ne+mWuF5s|iQL8OZF{LpB<KFkiS1TiPV
zdhaJFzY5H_Auj=va;C^R6%09;-b)tlY$>%-gjSf7$dl{^ybA@{%l)|Wt$;69Dg=FJ
zu*pu~99$#XK@DrMtQ=goKCFyH6mgW9p<3;6_SGX__)20DD|2t1m&rl#DPnUM8LuvY
z)N$)X&Aes3VT{s*&yXLF)EY4+CkB4dkVGSCS6Sy!Kd_%06PVz&JkoHV6uG&%H)hbA
z=moVezT>n57)LHo;WH2XA;Ol04(q#txoR-}FLQMpa|=X#gTi%8lS5n-xx~~_B=F=m
zH>WRa88y_1Ua?gSV^Fa#H~fzCH9{8IcO0QDzCPu2Yhl=UW?sRUJvYB2pmEo8N2xDf
zPCiH1P@Gp3HAXpp{cTTJ&gwRzM!x4r4f`5WWrj4K86dBp6wcq&MdlNN`_FmXC}nJx
zSvP2MDy@I!pU*);goi|?DtsH6)Ei4#U#sNDmz4*m2rocV(#zTWJdb-X^G<qG2;@eF
z=vKq$kQ(VFx|nH2C2F9=UOIvNo)@Zb_<Z1esXM&reEvBJr(t7{t>x`E7QMo^?{@E%
zTCT|7N9EH_G!SNoMU3o74lu6Lz^GAg>&%XER7HSH2GoS0ncS^3R46-a`(<@h2`yJS
zl&l7o9M3_<l_0C9s>4HFsk(h{Y@XV8(E2f-U_a4$A6Yr4E<*Iswh5vfxH@V{(2?`)
zNBCz&30RwJB$y8}jwkY06LiM<Ol03^<Tbe}7EIK*&d|vBh>UDWFOwF^aLVEbww8_~
zL)90SNvPAUzS>_GuyT?V+FY#%!;rF>@wrvXPdk)4$(Xb>Ffh!(Z;gh<PQ$d*?U%|n
z%eH8;kc6)kA2_A1;1IZ-Vn65MPv6o~$p}x9m6fgh1`P?{LAlsfp@5Tb4p?K^z9XK7
zh3+`QXMBM`;)8_|bhGf_Os&jb0Dnes{;>t3>20TQpY%daHf+QV6-y5+OP-AD9FOKK
zPS#BP>!4iNZ)+|`W!>13qgPmk%$78IgMRq8p8el2qiHal3qNSsZKp$>=UupA5^3m9
zUDjgfW?5Kgd?1~_{njMSQQmciP}?<q(;NS=dI}l#C1IZ&dxN=M>LU7LYI20!ee-J~
z(ju#klCu2RJ7~@4Nsq3jiDz)gmIf2Pz@F)D^9I4Kzw6FN?~6<kG@oR}zUXAO!82D<
z#rm}L?-WSn!!x4Ba_(Q~?(Zxxb$p2PoIG_#^3ArcN5-=i8&*wsn|{nDU9Vy5h(>K#
zo`<u1QZWB11L13FswH+z=Zj_@720Fo5+S!gIdI+|dH*xY+AW(s?&7fy^U7U^^4lv_
zb?hy@F~jo{)PLU4*!#8t61lf-(LEg7Iw8pU-Qpe&GXXA@x)vY1A8Ol`?Q#Cx?7NZ?
z)wf9KMUmITo@I<~XscI9Mir^^j_)N!#dsf~tX;)NEi>ZPDW6C6_4`gf8TT&L%CtOY
zb6p^<0G;_{Bj~1Jsuv&TvJWG>SLMfu1)ZYA++6P~-q)fmtRIiVyTq0TqD>r0oG<gr
zbWi&zirYKd-6W)=V&;x7VDn7SQ<7u)+>*UdFpF_s<UD*`W2CgQxZrDpORn*m50Z4#
ze!B3_heYydFB&<0X-G`#5){{YS4@=I76z>rQ$Pm@g?{6Mq40Zbb_nIVx3N;%_4cV4
zi<9lo4OZN7PhpY<r%Eklv;!BE`Gp0zp1VZ8^C<e6s`4JrZeVc1K#0ZcmeUJ8blc#&
zF!EAm8$+DT*NR^La1s>B%<kRRO?sQo4y|$)1y;jf`$*+PeL68Kim<>em#=6VT=}Nv
z1-H!Vfoj#p9Y<Da_CI5}GHCb!&henY{6o>~Dd*1x2^ItAQOJF{5KG)Uq-By<$L?)2
z$=-@&_tS!XyC#^A(4H5;*aw96EDEwYLHy(;?4q$0$WhpQ%;XEiU|#}nKIMJ9r)Y|I
zEXgQD^OLu^jPrHbi;Z#Hl;x2Q*5*nt2l7~=$KuA#DdlB^SaB3GL&^%DKKHm}Eg-1$
zb@QUbqHq`1_sQlbUYt_SvxF?QNR3aH7qJ2&AHe$H5*I7i%W<buRkI$vDRFHQxphll
zNuv_tzB7ett6@3MuO%~H?(@r4l5yGZyO!{Gu^3p>CTxxV$xrWhy3XQf=V$9K-^b+L
zy<N}5Utv<l*MCbg$I}$CT|D4WjZMsX^xaa`<=BPb#T@UVg}X5kU=iiYmhw#s#F1~J
zR0XsTpSJYs*<qiC4eGA5Rz^Dd2qTbkwpF0v0uASJU)PaY<}C`+-J5C;0_STZECeMA
zT=c8b=fek($+&iktN?rdnjod>Zq0GRD$Cxuu#=uCrzxtPwX?wk`!V3Qm!lRR>;u)7
zcAlhha-Zv`C>nYQJVuC!NhRUCHJ1nORKy^(u-{Op^_{itZw(lV2|LI`8=SU>N;+T1
zrpafuxtOfJyr~|=n^mB0X)sDwg-aE?zzn8pt4cVTc4cuqM7+#rv%Fe|pipqyHtL+N
zyI6Rm8N$2xQWxx#Txz5ahGO?xfjvj2#O7n9jSckJxsE9Cn#_eL*aKO`E*Q*u<Ak$V
zYLdy9sW8xjxbTw+>#@6tJD(l|`DFJzuV6PAkBhoX(UU|M!&0>eMBT<Zrb;i`eVOdk
zhN^xH;tSPwfn1Z4$|q>*u1YBu=H2anAyM_x%|t#|&wfr0>6_rv3CT0CzQK{L_w=U{
z_CaoW$7r=z)@doBl6i+*p$FMcTXdC+WvzFA`V`L9&1l>2_W~(hQvJ1EZ19{g^sER#
z&^!d**x;M-tu)5&dv>o3x$_;*oKtEUC|=R{crZhzp-20cxrb+02qP7S-V|7nz()9Q
z0?*$WIU5Yr&oL!Ab%Gv|_%Q=UpCcWFb#HQK>X=CwWprc|gT=*Dq>Hj8N8^yiS8Vv5
z>c8UH-gRN4tKG$&<+L1D5Z<Fq8bwK{st16IB6Kh!P0f*#Tbbye+`D;odxI@s6tYia
z*9&fC-Wfvq$KP3OBQ5XGGfAIYT>qTxCR;yqW6FoN@>O^Y)r^stVph%;W>5da7Ij8y
zZ%7_0uHM}*S1Oo))<A=T3jq!I$|4mSK6a<Qm6$}D_b4P?ST)VOQ$dpi(jqr94wx?!
znmlN|kL@Nc1<oKV*|hpkzr^hhSN3-d90FW`8|Pp#v)d@m&3waDUCOJ<`TRkq(6`bN
zTla)!uL(~_z*LE({A9%`dt6x`Q)u3jR_i(w^Q*wQbv7RQ_0(W4&N`q(QK>PoX=sp<
z8S<3sHoL9%?PYEi$p3BKrgzI-gf!JoiA;RCNjX-PA5nK<9on?Hc5=xiVG8xyeU<Km
z@BLM_Rc5UF{vmKVj|t0MR4cnV*Km)IE{XxSmlBv<V2?q_0}#0KD~@oDhwUa_2>x`z
z*&8P&6T%bEOLpOHS--msBtI8mBxbU9i;Thx<ug>PRWI~j*x|?en&@%dy^WT25>RYN
zC0c)fKOb}L_K;J&^MZlRV>s#IDb(I&do+7apVqTU%=cL1;@~IQ!;4p*B~)1Azkjvb
zl<}Pyz`z}_dKx;{ZPGQe^0((14#10R!sZ3W8UmUuSH?OV2G))@S2K!F<;!~Q6cOKt
z#DNOPvX01P+x$3x3^uI)GzW@P?RB{<|Db@o$=6cyJ^t|%r?lP$HE41LE}oo^fx1-s
z+ETycsDjJ&WR1Ti2i-5*dnwuKm_T%-I<^Bn2+aa19<%}9{!(M;mgnadX6JVl3GWs!
zZSM!r2BsRMDP_bWYOj%nbC>)wd5TC8R}41s_BGOpWiivp3{xH>keM2q&h#^<N(0ml
z<hd+1Mk4sf5k-vW><X3e?kzPB+l5I=wi&T-uZlMoB>Q}yrIi_)hwXob$teid70>OX
z>R@%eW|)Efq_FjWiyb;5>``?5NfyUg8gfG;^h+(6YU25e+`ld~g4Ae4#@~)rJ{%r4
z3{!J>>i=&jnq@8{wVDG+UgKTYwms1J2^hK|g2lr~Y?!I|s@xBv=71Uu*8lo?X1u}H
zh&g5u6L>N_V0Ce-7Hj%mPrp@itv87oQ9Zy5Ql~_xK@a*RPMPZ$2-FwkgA2b=Bj+M#
zx*DKx$}JY1-PPjUwGxgM@r6lfLrY*UuTxJs@p)h#opG$ED_FJy_U>1Fm)$$Ub86sN
z5}=cZC>n+h9B60ElM$pmX!Vhs`u{1>@Og8k2t%o{mH&!KWBjlrX2Y7ida~4zPh)Fj
zjgD*S6f;(*lj@uoz<9t|A%YR@XORi<EfG+^_j+TAJ}f%j`!&*Bixnz_7T^9?WwsD+
zyA@4LU}Q3I(V2IoNog-lJSxw}xT=6!WMGyQh*ZfdKLkELyEv3}J3Y5>lb-8)mArmA
zLMtDu6<+AymTeY84m!B&N=FSbq|hrW_O4NbBx!x_cxe^gnd*=KkAHPVjb|X3UU9j?
z0vg2WoxLUqGbguzzI)pJo4=nCBz<HF0d0ZG!Hk1r`1?c+nO9wSsxi8XYMJ_zbJ~Y3
zr3{(jMSn#E@P(9kJuF&W4aqLu5*|u&5&_+g(AkBlGz75yyn*df6J4=aFVx?r{luOa
zOHpg|t@>!48S3L9>g|mPI5ESa#UoD(|I3HK;m&P<?m}CkM&st=ma0z+ASc;*UE~eF
z#8_D<p}B7x#wD?Yr+|K4bVR*71UO@8-+I|LXmYq*!jR436`TIvaMnbmf@p=~$kTsl
z$NxH6R9H}#sQKkOs&biOT~<M7=IP0eshNe<Kz@uVF4&3Ac=?LHeQe+(>D945XhNV!
zE*8w(5W;Wa9RqjTTz~VO>g8ISqk6*wQy)dzARk;C*Dr9S`5i6rGvNP(5JWy@#4CNp
zJ|-+;pnrSAWn86CJa;T{AjXkgtm}X65>>uZudAu~Q96DL^Jd{`(_{-m7E1bTBL9rs
zmwRgK`pFALY#i9|KZ2!%ZYe(xx0hUz(n!OJq(%tL<_jmuzp@cq$r5|S!a1rug#6%f
z$@!avTaNjV_2M8dG3X6hS#&#2O49hdZiI)?_A$wY^j_3Y9>_`2pIN$d7iQPin{+2T
zN29p(*Fqc9JF?#3#x|8c?mRlrMhz9vn?ja4F5!ROK1tcydn3CLjFA+0=I-Jm8sx=}
z9)b71D)G>F2#w<(QBvNoD}%QJc{XZv_3mrwox&YUyNOCSFW!%Km(~0tg*MSsZVf(t
zORuXeh5Q;4z314q8??3ys(kI>lq45}!yo_E@<+3BiwOg61;y`43V1gPK5JGb{v_4t
z$6QDKmk#bc-h3p4bReNCYNuBN0Vg|EbYs39tctw8oBhBgDZuh^KkAanygHE&8W!(0
z8H)AJ_cj6hT#wc?B9od2rQ$1cDF<Z7Ph`2i3IPk=`sR0{qvmy)J+2?GUvHkky^(}q
z>fkX1cbL~rp1iyFN|sa=FUaU5**6wA&Dtx_MNeO_rbDqRY9yg8JSsca$b;F5^m;wm
z<(&yXlz7AVrDyQ0Vy$C0YON&Q#Dm^pVb?8q5#eL(5yx&i1gl}3l_g|{oTuF<h-zsg
z={Xv;-nsd#v@_?eB-d4m^O9?MH;ux^8r7ur^(u$dOPw`0cB!%G0)F<6Ck}xZ#c1=}
z&+8iv7`tbhK5*EDO-yl2RydXtuR70UjB?_g^+Z)Jy)V>_N|t~;|8zyAY@S~7<CVKM
zi?AH5Z@uEqO9t}80!M^7w~awdvt<1qPxaTbF=WH*?Omf%dmG)G%T=duhPBqdg6KEr
zMuvu`RY!L@#};$yOCy9n8F0q8%-A(=8K3_uAt|>Qf>O`R%qgq~=U~%@Y%yd*94KK(
zCQfUDEOvUdILBJC-bA`gY~daJiB3%E4V816GVl^CJxifTD61@<y&<<;xA?PULMAl^
zLKv-Yp>)blJ?m$Bqk867m8+=4UB}glLSD}F%jT5^{ZC&=t+pj#O<#O`6{{1Labc=5
zjKa>E$KQV|S=GKmo#OWWy*HZ$I%`_WrAs+)y?3yRc_Jz1B)h#*dOt)~{7S+?`XAzU
z29cqgi&jq3srE12IhzAm{&32PV6@v&8uF)v)`orTGZ$f57P<V{hR+?WdaK?<61u|6
zZ>)cV?0inZT;fw>maV-D|1Sdy`g(;O)?~h{8SgWQw+}<98A+nUGU_0`8V#!fF*zrs
zyv%FtTA#&#l(l-n8LWaEQ5HI-<{qm3{#SiTt{34(dQsV@9B1WhNEc^4yJweFJzwXi
za}n;vUE-Dfza$+strOS9+(lA(<)kUcBzm&UhgPv?P59=jOO9)(<#u^4Eoy~;qaKE1
z$(`bGlf_)EyF1JMj04~w99^c95J!TX0lwuzEOSoqSJN&eR*Qdqdk1(3@?0UJu-Hca
z!d?gAIEwSi?zNk?81bvv&35%jXb*i5+*T(tB&6co+FASCjz+~y`W0Y!Bvh(UC@Q<N
z3~sCs+Ih<hRlX$@td3$A`$!b)Xt9is`dKja&4W80W45Xr!7f-Y@a=_I^jqons&UAz
zoS%HEvZM#)Ss!w<>-y5=V)R3wBtP<%9M5vI3HKEeVi9I(dX=rH9NN7BA4i$0Q%?`^
zmyo?v>?&2mi(O?eF}VzddBxPr+@p08=#Uw@c#~FdTe7%#v5R#`aZBi9=tkd%{BE%H
zTku=0aM}p<gvU{Py<gL>|I{%SG=<pZ`gkm8N_k9_XRT4<7Ci6yfSh+ND%oC;d&8-g
z7UYYHF*DM0@@GHO%mJI?g<Z+MIMoQvl05cXVX~4?quF$0^tIH-u?ZTXoDXkxnV0Q`
zybvC5V&gAExWG6!dk4=rl1B<5T#^w_PoIStc_swE9lE0u+GX3jBgfOPFQX~p9col$
z4k>u1ZQE`IpN^c(RlB3ZBx`*4;km2l8x~)l5hFWd@;deKq=fAZJg<<;kl&dMh)?_8
z!3t623GO4p3MaM^(Y-DMx$w2<S7Kg=wFjr<bcO(^MmfkdRrMt_wPc+At7cl~GraPw
z(fbBfJGBTdHO->8g4}2FML`MUdIb}SFyD$@EjxUo)(TGQbJ`$-d0SVI=cvr)0&H4Z
zUuQPO_y&7&`#@m5W~xA;6F&UZsh_@Rn#!Q0KcZ*T`1;<YrPp7{HCN9oze_>Ordscu
zv%A?SdBbkC>E*V{ySYN`P^^P6tOR9>?!8FyP!f)6_3Q6La(v!**AV;IIr~vniI>+>
zN-NB5y)ykaJ4Hw}(a2ynydj}b{cF50bl1dZr%GI0{?A?R6$1Yl`v_GHU?F?e)FOIj
zkOuz*1vvF%PP|jG%Ef8Mj<H|M^6P<9!HTI?@}<zf!dBk?SlV1;$i|&(E(h6_*T{s<
zZv;XeW9{3TVqMhsfukn%)E0a@0=~#<2f@N_zdNA$(%ikf7@D%gYPDX;q1tIizDW4V
zJk^jqlLfZ|yQStv#-lCx-x7^#lp2UQk;T2J{t;HdbG-C6Hjq@%w_EN&?#p4mxjais
zC%Qm4qJ{q0FbLM%dOnj&$nVlWErQuY^ZwMopDv`XBE&3UY-~IP8@WakS&={niD3&*
z2ZR|c8nM^Cx?jJuQS3Abel9`br9{u~YJQT7P}!1xu-@kwaw7yTB!o=+LPR3ZeVlxI
zg<muQ+eeLbF>|AaN>Ke;liQWr7wtzYt9Z_6<`EtRn)-3`{XX6kH3`#ZdrmzkL}+IG
zF7xk4WV5I$h?+J^IK>P@$1OgV@0s;o49oK<l7-DW(usnsyuzZd@5!6z2d?Bvv%1J`
zosqETgBUP&VPV~>x?bYBRN1a`O)$w2Rn@x2Sm?bvF}CsNki2uUcl$!JaZv}70RDj(
z9VF!%)kAK{E#!XUG7c5OS$;RT=d$>cCJ*(Hm9};Mxpv&0O>eIdEJh7JhT;bKUkix{
z*yr;7&XJuJi`=_S8=0&nK33a0&?8WXzsH4;C&jq0v69(d{}nP!igb5jwuj058CuL2
zLU<pR8TUI(=4kdlknYcqp__S)=Nd%ze)G?V*F@km6(RFIxSFG2nxmB9Y2B6g#crwU
zsW9ka$$r*e1+u*CHEVOJ-#G-<iClixvJO`y3Ql58q1qZ2D@fZt^sH43rSO%x<GEF;
z3j@)^RE3^ay=eGpIHH|;M9srbjx*LvSznY8-9oiB(Yr2mh8$?Uzf>6XNq1_E8W;O}
z)ZbZx#F{fRx!&$SeGpPl8~w-OpMN6%h6J$@?{N=aL+7fgrhMvW@xyHtr_B0hr;<8K
zWVOnG8VW!RNU}(D7L66Wv`tG-7xjkvaAa2ToZOIAIM?)4FQq*SrX8Kzh3{u*Z!De_
zBQ`Z?B|}nju_;4UBKy>D-G3R6abav6+&+TxUjDaFn}5OhcaV`}dK)M%HyA$pyf+>c
zanhWb$;Tw5N;mnyeE}<D+k8$5;)CrUb3PD}zAdez;A$E^-%2irg0;<%teK>D6U)OV
zOu?1LTdSmrQ)CZ&EQop_K)q<TeS?%(E+ZGo;#fIjYllJjR{!hqoPQ;PM!?wVhEk=~
zUjluUSgy_jiUQ2h$LJuF{z$I;=Vz+&RT=*ea)F4EqRSQBL6}8)m1@HMVRN*s*T()a
z##nyREyz5kb;#+lBtYxMeL8jcS+V#~G0;x<m2Y?j96Dvc``3GP?5#xBnYxcHkK_b}
zdX{c;7!`!lvu_lo74nRgq<vX6uT{%v?))f&D@o<cY@wmlTPD3ww)j3~$5|Lq+@cfy
zsNs#$4|Q0}sh6>38v{;*sy}mr&&&oH74nj<b40ISrVF-?`<Bb_QRb21DXy#9y(P~*
z#}><K+0`<-G{>9H-~E6GWiov|ne}|KP4#QP8QH$zzD4_qo9F+qG9bp}eTWUYLp`7@
zD?9;ln>|&bfFDqu$r0FxH;_eH)MGiT9z*EDeA=!sTED^$k)9j5JN6ESozAm2z)2cI
zDSN#{!``jen}f&dgM1!m^}ei?T;^EB2N`shaMWG$>f;p^kzT^&hPenwzVKYPwA3bP
zhc2`B&4Qz})5$)RbYc?Pb6FEBdewsOsd~<tAL!bCi2HobK&PE!Qw7C^+-i7PZ`!@U
zKF(L47RvD}ldnxUlYi`b>+|dOhrk72yFz%^cE_#oDB%G^nFV!X)Z&EQ(zgopq3+$q
zig$zC8p}MYJu@-Z=SF{e^_j%q%dUZMzAvszI=zIJ;?gtE#KuH*lwejbsA{4!(yPtm
z3~`q3uva&JeUE$Qew*l$9w;@?d_Y9hnd`Eyv=Kgd>CMkdytOp_EJ+Q-%vK0GK2y81
z0OOR~K$o{1A98N)&@?`^AuJx*{kqdzUlIRY@YIkfvulq@*^c?jPQuTcC5Fm5e8!9K
z`GPpr<rIBwcE|EBJ#3TS1x?QeH{wsLXTO!?*9kJLzb>O8S*TZQIQw4UN^^LTNFeT1
z*h^jG0ELx525)8N+N?q=1W)3y;oLqAm+?=z!1l4dN^eW7!=vdbuWqj9Ctmuta{5=v
zGZTZUQhTE!Hon*l9Yc`1df=~I-K;2f8#Loj5{J5STDtnD*7LI6y~yUo=)THVihbB+
zXuMc|q6pT)veiA=Ub&YuP94|(SwXq&?Ju)2WU78ZD7#<1x_GsEyP<H8<6F#wupFb}
zW{`XT57Tp|Z?s<Y$9^;r{>q!K7AMI&q^1da)C|Q|&1#eigMIRAb0;9+)`6y(h6am4
zLDN<I2*(sGGOC~JXMC+)PQ#wc%O_CDW>;}Z_>9F9Po>b}!ml`+T3m*>E2IT|hMZ$?
zb^EP_$(F8l{ny*l_l5O-vg%EnD}y}>?C_gq3j#>z)wzIG#E9hyEB;@_s26P_FER+%
zj8uZvf67cTEVo5?Su~uyDXc8;uzuT3_*9K>s{W7Ii?vItD5bQlGgroc+5Vb4t)}SN
zA<^FzWx8F^>qee(Yo3Yay?BmdK1E8bB}(F-9(NWoFz~$z-|K$T4L;Q$)7{J&>OGPD
z3qNrD5J)4<GQ2oG)2%E(au(ak@tM@u3Fn<et7Z4y0QFIwV?MpXnz2#gAR95LuKUC2
z;x)4UOf8`#yQ}y6om_11MWoBfq}iAy*YnMn6AgDu(2*x6M-A95T#xrTG@@SjP#k)q
z4Gh6{cd}qNg+J~p6aB!~b?mA*zb}^6pPZ&Qs?FyA<@n1nto7ojfe$8ms%YKyXiUq3
zefWTFHoTIOby0F2D`GhuR<Zx2w?e60F=b<UQ*7tBF{a5e(K>3PA+0gUnwNKkw9sO1
zXyl5oov1s<3s#M{p9Us}3LUztPWc}lc-}h%_5@E<?V7Qw@qJ`TH2Buj(paQZT<NS9
zM)*ps4xvO^nfQmru?3MvR<=!}5+3*5w^$@0(h96~uxB{T*jcz<**|sQL3^%GpdYnt
z+jz6xV8E`ya(CAtN$JimqE`?P0`Wc&v6th%54MA^>X6EeJXeNxcV4d9EmKeH>-LRA
zgz@qY6@8=HJQS5JHw23r-t4M(M2#Y;UIG^f%?%1NaB<9bCmRuw?<xN7{H~VpH6uqi
zBg^{RkCA4nWy`G<Zdb>xjYIY_UYzd7t*sTTjZ+;cp@kwYnu61IcOvb|nO?YBVLIar
z?7F`TZOohJ5K6wEUI>@!m#ne9K7y}f9(k<aUL^}s_liXp+SEm-DQ7~l%)hd7Wjg9J
zum)L0Zevp#B-2aYJ@D$z+Z=a=%6Eo}jjU@-ZdRKxV3>5nGb9X>Nh^V?E4s#I7xZn5
z%jyg5au*|9)0Cv!-o27mlN`VEl2-#COA#bh3tnu!8}j1}lbRKOsNB6SpHpJ{?+msy
z`=*MS`b(ChX7rvLUCKn0>~|Jw6`pln^#Wy^C*9-}nEBeoXXZ*I&8&hA2NQ2o9Pqxr
z8Qo-w%^&6-*803(9)07^ewHCg*4v!D{hEHLLujUKL|!L~_-RfDO;1c8GDzXe{0y~t
zqoL+M8e@ZdnA+%-fzrPyhx-_JdnjgN6KC1I)7$mw&h|#I0Vg`d<L<mhvE8*!weD17
z^y4QHFf=LqoQ;;Se@3>B>AP4l<LBZ!S23wbP3f5rW<k%eZF>g22+3;_p7)q=GXI0T
z_x@+=f877+E=5rkEv*_gg4#9GQls{!MrhS0vG-_cwMJ1hc1jVWq(%~(mRdpVG$Pcl
zz4v&2PM@#&{=UC|!RK>+I3nVllap~fkLz(=_si2HR1i==R25UBc(5^EH<Ac@nsU=p
zn@+{<P#)B6Z#FEg%l`Vb$Rt7GJ#@=(N*dxfFID@vZ>~1XUoc$71|s>^%7llSaeiE|
z!Ey2JDRcO60G6q57|`WBgXp=q(uU%+fl!8v?P7E-&j6EGFGF`x*3d%zFH!}#3%^_?
z)c480`TM)x?86`<C!%8I*Zm6{qX_U53G<1&V`bC!JeC9OW|&?L6a!R;=Ccd>!36`h
zyl+SoyH<^(hsMZpp1Yzsk3{4arV-}}p2t%v4ly)qH~D)C*<bJ~{M6BKT^2EyIzCW~
zOt@RDtW&R)ty^Ne(zV5^m>JtrO6ng=omMXx`Be}i0QlOP9Z18qA%6Qlo0y;~J3gtK
z53?<)tGv#Kp%(;Aw5RAbHUr9dc|6HqewaQn#lSQ?K&?nq$Dym8NqHHgu{(?(>XrqS
z@B%RvED~$hl6bJwEN%BS7khX&VKRs;b86Dk5;siKD0BHE+o|RyS9Y)EH$7Q0oe=ZR
z0z!gf<;I+6#$?p4T75>8VMs<^zhj70O7l4%w_GvVH^bh=w47X^Z(qoBQ^Ko?^@4H6
zWznhD?xV>C^cAjpqEJ(ADbl?IP-p3cAH=-kK`)tMxmS)-Sa3>`;~~eC>~af?=(CRN
zcoo|>V+a3U(vE}GKK;DFLgAV@*5@<&(%5aBTA&P=_)jxPggeLmgiG4|!U`KYqCI{~
zbgx$2b>iXWy6xqYB8NHeF<x&kep%e?Ee0d~TxC_W*?nm}YUW0FucYRaihOLIgZ{F_
z{^0nSq;KkBmUA|5kPz<Ok+y(RrXfs3AcH)Coxu;qQ-p+<8W;92rcjO8cxym-lc=;+
zkp?-(`p;X!uBLRt#cYwu?_RclXobliua$wS$8MnSM}APnkbEaW=_C~GhSLCoj!O8m
z>!cqC>#PWkYw7n`H*L5YJpNf_^oY(ID+E)=J&jj%U(ep?e|nnzH16nHt|k0-%Wz=M
zf;QXh6-lY%`b!F;N#EVZ6zwew^xL{k&b`$Bx*hq@x9y88{9RQ>oDRb{EXHe0a)Cgl
zF;K!5TBeE;8~czv>u_<y8VmREJvH_`2uxS#ee`ydmiEeX?e!#_z_;S*UV=$ZpUf0k
zYS1n4?YP+cOIb)Z7~B|LlA6L>%Z4=IM(fn%8X+;WhFnRi5rX+oY%#m!`d?a!`x_Xi
z-~@RE-HTNV-vxVtYpd9>T8}n)D%@_p({$P3?zv)jbG|5{@f)K_l!w~Q+A7Yg#X}~G
zbpv;uQdoy!Y#*w{*t}RveT<uCts&{FUHS^X3_u)vXIJhn8sMzt0223nYWnM+9$_~X
zBbn!(Tw7h9lBU_VY<gj0S!6tKbWh{;xJAH0^^S719!V-|!e_4dhxgmq;e^dRKCn6t
z#W2|nn?2YVx0&K#=5S9AH=ENG=2o>UPWcgCsH}#6Uzep-@54nXasmZ!DM7^QNZ=sT
zT9#F`b}7cp0lbed*IDW5w4up#c?OrnS=Sdcl##&3Hg;5UCae0QFCVck&ESW%J<Uy@
z9zg=0Y8*)C1x!sAbVgM!2A2ixPBs)anhk8EdOb{S7?~-jBodul>Z3{DV~*@{)yv1>
z!uRrOO=B2n!0Nq453~ZgFL8JfVcZ#!W9d&E71wJ%>&J?i`E1q5c8Glc(Z^0dx+R8$
zU&zpN9>K*hCC5JfaFFO+RuuEz31NMXN#3-6Io)&N6&<RF6pick8Sz@#%UxhLE>?W+
zHzGEj<(TAv7S1d6$XIBUF2#M8e4y!B;+>RAcnnWQYv1o49vMvi%zoPiRoHV9gx$Pb
zG}s_GK3`DOEnhkJA@x|OF8^Q+RFnvlipY4{y?9TbzSKL});z(@sY4OHUIUJJd&Wd-
z=W<Z+Uc%L=WgKnx)55d~R06w8_FnLK;Ufs^&NhkRm?~3x$ik!~Yd6FXQ_mQRH$+Nk
zJjecON|MMP0i3s8!7TXDIjvJwGv0u$$<ggqv*`_FI%b_}B<y1YKYD!$#P1yWZ2LoG
zZ5mcEvD$-9!S+Q2KMB?>oBI7oJO)y?M#Zmd{dCqP;J0#5?ZZnCggCFhan`Q#ZZ9W$
ziBdC`Rd@3ARF>4pXX>yI+?Y06zFy9*IvWK+KK-o@tpM!{^j24C;uX`LO5LjUTAw@i
znwK^wEbL*nowX=Qy@?+TyCO)LuBx^AZ^Spg?K$4ed?V?cCHLm~-HD-ReD`lqQ)OSt
z8-8#Ov{dzb<tR(m_FM>5*SX;by-{@L2cgCjDc6lI@RNmItff@BYURVIjnCT)k^2n`
zl_bY0u=Upo?ymcv49YbhelTxiOOYQJixpTynFu6+B8@MnE8X@@a@AvG$XYIM6PR<%
zR36GcbzNVYxVMt7y-~8wZbvqf<?#I)_^qu$3N9<5F@N`0BlG7M)7n+EPge7^-B-HM
zC3CrfRXVn#$I18U%y#)fxlV7SjL?$wI4DVMvZK2zlv3LM-NrM+=(K>in{IYCYmDUN
zpk<m(1~L+20wpOX%t+Lc8H&Csuo+T`J^3O}%$U)611jVj61b*;G`gpn&HRMs$7HP+
zY5VYvc{aM<vr4hbP>PxN!mo)Nzf@Ko$v0WtY`n*sc$At8EVT)xECt)JN9P1qgPPvx
zz>?bpP1t+y{?@@6r5hv{^Nv|KMpYiJHN%nN)Vp)L541WWqH=zy@-e0M6wA&prWXsn
zq;>+cog$k@eutPd^wex9JZ7g}ksK?CmD6{<)e5;e@b2Kndms5*ZPZdjc~$TDd~dKd
zg6GjZruofADg^Vgr&_-gXS``{sg1qxNY${gpO0q6_~Dm>=DH`W?sd3<vZ3PW1+e^?
z_5CJxso?tLKmJ7#;S3Lv`!VF8*~XSX)4_04B<ZVa9?wesN!>J+>TBRq_o8x=*l@jf
zzltXp(ckLVn|s2~qhr3(fpf+qHkv~2-=Q>Fr#%bw^F6)QHi}PK4Gjh#Q$GEaxX_x@
zZ~>COf_V<i<?DicgxBvzVMeV&eqZh18Lmmq3N@J7jPh=T^E6rW;C@m041F9yRsXHP
zFrql*Lm=i?T`O^fkiW$>?wuK=8*S)B0>hzohlAB)r_SR5>X71A$2*krGCq57tYi8@
zefVz6pJ(*XFg)FVQsuvc<#va9c!t0x&5;26=^y*}uK@prbjQGRput~&kjM|xic|w7
z0F&_)RJ93i91$a}L!Ayb<s}fmg}LhM>f+PN0=0}hdTPR#gva|CEDx9)+qqVHjURpj
zHq3}T5AId4v6j%020}-`ZnEjBn)`W%;*JAe+ZE<0J>qSA922x(;z#noZCSI7pJfRh
zuL%ZtP*bg?SQLKeYEIArY~whr^T3DmJd3kKvonQ5d5-o4W%ThH;Bf%b@q6<eQuzGy
zR%`&I$3kxX=VgR0OmQmsNKbMI4py763Im1Ii+ImAJ@*)@w|-_9>brAZr%9zeBVr$F
zVb$UYX$<sST0J_q-0HKpJ_TIgEl-=L`*J3WlM`nf3~T6-)+Tiq_ubF{)kmwEp7~dp
zBDeYSxN2w^6BOngHGc_^5@Ui6K(9fvcwLwhXrJe33^=Zf7Rd4pkA-30O7s2hy@@Mg
zV)Ivu>L~!S9=+SEi~k>R!avSF1%S@E9sC`%h5h#n0u<jF7&dOD#rdMTD7n?exCl>X
zP;~GbJLAq0Or4$UndP@jsw%CkfX8BviHXVUjGNp#dv&S~(rA?<yLTxNhasI*Vz2A{
z5l|!1GdOLzzbHN`8(Y|6_U8bnJKZ}VyVumr$MI5HYxh+Yba67Z<=QUDTbQpSuAxaJ
zja~>-H=Aw;ZRRZcd)EB#<5c&{$^O-lcoO#Kx1y0^50GyB?NG&JD4eI#E~V-D|I`jC
z<-OMkgw2!iKj$4VzW+-V{(Do2px`!#-}ELE0#Wn-_N!1@MfIZ<t{rTFA#YfGM!USV
zFohF$KDjcnq1TqSeXo6n;^huD*Z&B;eD7(S%nt)QwYJF_q<YWS%7HE;od-u@#v;&T
zLlHeStYf45-#hwWtKG{*e?Am&@g;f_k}vOu#1O=v)u}}!$U?&6pNC>%p1$!*WmW-z
zF5I->a>zghT9HYVf6PWu=EF~ak{nz>lH+fw^MT_31Fj0dXtfP4mk*QTJ$wEdIH9^U
z4+D1B4iBQT_YN!&Lw6p3Lcp4nP`DQPy{*HOs()zOEuk+sG&!zg@PqC#hbW%fJiE^d
zK9kBv7VlI#auzwy3G*z_`y%6(=SGs+I403Y5LskQCcpbco~_4v?<eLPwV;NX2Q&T=
zDT>9GhAel9f=-FQhi+3r1x$`omDM%kU$^Q*TNTa_l?V1TUJXJnyx-g8&sNaYg6_qP
zi_)zr7M@E!??1MV2!<9!F*opEw;NtWspX{V$iL6DYoEicgeGXj%?N@NMUKv-A>-R*
ze1fM1Z_oA){{3opPYvI<5%kt>T(wrm$`U?R%I@4>6jjipxpP0zD9622`;C&p&*S&G
z%1FbRK{t>oFc8W@(Mb>4Y*o|{!7ug#_hR(p6WOjtY4a|4XrH{0g>cWI0J3A}L`vA6
zMrvR+wueiU@j)iHSLXYll2K{<&%&G@I?O-ehu1URdQ>-NQ%6?HAuqpUVi7Rb<XVEW
zQNMqd*HT_K=3S5r((69|q0cLUJ$uit3h~a`kk@06Yc-bG)~B(BOrdTSXSQdU>AO7@
zD?3=@8vH!gy!A@CiH(Usv7xtI-v-i9$b8GTtS#B5a2zGOU5Lqa9EUy?dMNTMhMcmc
zRTcyzzaTvIp<~#77?~;TbsJ4!MtzB%4-&1>&XQ~IZIfsDk;Pi^`)1f{<5K&mw|#of
zw9`Dj=ZXz%!XCqrAeeS2gv-VAg>o`lBDrR8d}M&=ijUpj_*pNrx;tdjS!;Ft(2NEA
z(0a|OH$|&+rk%!Klm8hhA^1Yuyf}nyu8ow6Qo{<d6~M+jWT6XY4}AnxyQ=tmjcB38
zmq^A|V@1L-3=hSr?2qKlI<E<pSj!KX3c|K^P~-#_Uz5|bROz#CG&MI*vv9hMWhWTd
z<Q~a^NSQ}EvCx;Re9W6Jm&E>}aD4$iHqPx$=N*TOZTc367_$R`m;6>Extk^BN}8IP
zI^kG@nMC5k7mU&5t7yKw{ht;yFR$&{bxqy2l9OE}SqMG540o-WOPr{D(9|fk5BGo3
zbZ1)iuDS8?-d_~8%R2`iunv^6fib1mvP4~gT6IyCGyjX|s9{z<5|qW5CBxr!owT88
zkybZGSYB9qdH%BJ`JfjDd$bFe)_W%SrfH@!uMPclcN`^~C}$L4-;9zktvVKMlm$qR
zrC3>vm_{cw@?*xrRGlIY@&gSt1gw0#Eedhg!uO-dRbqK&7QY%h9r&E*J>A<Q9y!tC
zEfpc=_#Vt1`yu{kr>~7_jfg`|HG~%mdM#8Q4;+oA+7p<WI+Tl_Nedb{q5bY~z0?ME
z8&-0}2t2ft60i_710f`gNl@%I&W8DrB$Rq=Xm1A99P2_1cU!+jyiE%+=@bn3nC?7i
z)vi@F1G9F!CD3LhO7erG&Y7}(k);`8NH>3;BwCyjt=vH@9?S~y-Y-ylqbD<`SNJXQ
zTWQ_zS`wJ5mKbo_GjSp>HseBC+?8T`VY>`^*s5j$owyZW02OA1ClA|0GB#L!2D<CF
z4z4$FxdNC(*D`E}&ubQ;1Vhv4`KH&j(ArObzUxoUg8ANhTc~fWsuZ&mO<D^$Uu3wQ
z7&^wB{;qE@TzCb~h4UP*V0IonN;nxsa@96|mnW38_}yS@y<pWD*ctDMb!aq$ZB{Ig
zI%WEuuXa9m@4H;Dq9^fCBRL%m0zh{~nfY5!5v~~2xh?$6{&Ga`2X23>VJQi~uI~v_
zHou>emLc%8Cnj#{RO4Dt4M%2|`qpqHUXh+Lu<Qgn=UGtJJT&vF_`!*YtWdUr8D)*0
zy;^RL&F_U=fa5KSFa!l`3?FR#vIQl@t?bQ>s(Vk%@9!>RI|=t3c|F+hqi!CY5((~S
zRyTh%75geYj1lnp!LgL|R8XlW&m3o%@^iREkWoHh$)bv+>HYKcyeGr=*bylti@Y1+
zy`VMt%#UZp&%DbPa%^n+{VxjkRWa>fT|zi#1vA+YjOnIU-OgVWrPoBkYmXks6DtgU
zm;T0V5V39_RT_^+==!A@D`>~QYsoZM=xikyNyV`f;R^d+J&HU!s@PQlp|Mr7(T@cf
z9eKS^4O&u~8d(8Pm2{11c?$^T>+)OkhFdnhIgSq)<yqr?T<e)mUZfj!TRK4&NjPwN
zEQP>rzK6^Uk$9s?(FQLKL1TB-K0I~%P`i|En25C3N!&Vo-`X)n@JOLaW~X!O%^ywi
zyIH=Xof-C%P6dTS|3=qIR?yBidQjD?OJFUvYet+7b_5hI#q-$dBw1y0RIZGDW_+QH
z{F?Uufxa;V5eZiCsm}7+!Idw2Evz3K_+Fjnz<&`v`X$<Yqa<xNTq3~eDD+Os50|0U
z<72v8B^p|6P)~=(eHO(U&_l!Q_)A&`8^w&jt9Sq!Dqs6Y&!0s_rK!R4`()5m3T_9#
zy3?!0{=!PwhS_q{lurEo8{=ZBJ?<B=Kd2Tk3vC~luO!zjaUqn7GO3jE``2Ux^YsPY
z;SD`SjJGD#Ye)iWye^5W#(hgUPECt*=!X}qML6GCGww{-^x|}gmTz82IxvIMM;avC
zsbToGdH(8{RODmS7)zFOP_L|eX)?=;W;Wy#xLB!nKuW;@tN@1Q6>L+fZz=b>jk9r`
zvg>OZ&~%wlbB}jegMIq-?%L>j_Z-5oHOO;c`T=dRM%n0#VHYvv1Fjo^DFRVSPj=>`
z#pEGkc?zIQ&U^)(;*;zubzGa|Z=hN1HoJH;X>NPkyNbe{YP*+Lb_&4H_TI?1szG|>
zP2_8SX<9K?kgQft(l0hHfvG`>U_Kq>lO&$@&lu!v8BK>{B39kWhcMcvqOyT2Ysb6!
z^Hz~6vy$j!i{&-%M+c<2J<sy}dg^n|v9~U$oNGUD(OCd7fEDNp*}qQAT+HWJ>F3ZN
zH#uCxId9?}B($^EjAu)$QYQjj$=LLg9MKor#PGD-D4DyF16*xQ?%$Wc5aEK?CpL?$
zmh|(>#qo-XPKwP8iohHdI+xa4Q^QipL{+6RJA_3<%*=BR>o;WIh6|`_4)=;dXTa{-
zPl^}{pl(<0^7j1D4E=~5b*-@URjv<g<P!4XQFqN(eNWS+O8gu$Giji$>MY-EH%Jn~
zuAztIF)LfY7#~5h4@Ka99+iC~@-wAeM6Pyu|3Y?+MjL~SYWG38wv@XoO7+^^(59u2
zusfqh&GQ-U^UK8^de=#@{OF<=#ojaBxaM{oFoJ3bm8ZiwhiEkWnYh5SXG}7!Yu5tH
zl%{LTsq!Pb5yoZ4oR(uLs<v<%O^=RzndzQkhZ^GD{NHcy#3Vf@5mum-A0J&hD#*}k
zGJkd$=fCdzYhi>T&6vd7TJh!Bxzr>U)<qbv`!Ql#2T2VNEezjGEvi1Ul~ntX&pt#p
z@5J3-o`jc=uUFNB?|k3uzm*6%yw;OVhR1YQ_f9e(s4j-~HJTxVlF1||6Lx9-U6$ek
zu9xltT5QJ5#q?30H|-7R^4vdG=dxo=wKK0x5TWb1IXi3NOUA*hJ>%ogM6zVpjFv|D
z$gQ{929}u2wI9qq^;BSC1U<_UXZTF{Zn%)>S+B9^G&G#2sVVjG_#G_azMWeg*(V1w
zuFaacktH5)@$FjS7^`nZDT9-=mHqOT<34v9Z8KYNuqbv+3m5xt5~|KEZk6adXhl}o
zb_?Mb(@5sBhPBOA0rBVCQ7vpY_3j%u^yKUP-q|ma%;@<o5hy9ZuDX&wBKu}l38@&_
zO>Hc*^WNMpKhe*yyioaVPR6D2H0fU@1X$0#PnHMu8Do8%({rIw9Hjy{jA-{8NcGBL
zl}+K38?tRjjrKgA&BF9G@0=^0*7e!=2DkOu#at0yvP2T105kF1vIaKO9EaC6gnaou
zi)XH6Orz?v_Q>w0(J5m<aQwoQyErRN<U;le6bTlEgj}^m6{+Oob|M;_@F-Lptk15(
zWym`ou_&%dnnz#*q@YFz!McsEUfg`n?{uzbUwvfsS)aT5mx=})<-pN8GdnNU+!BRf
zZhdza+{X(VA+C6OXnkqQlJ;AHwCxSMVsS9%uT}Ko#zB2;f-W~jbuYo3nCEN;YH}-7
zzbIp89+@Er%VT`2)Jwl1u8Q^Bx(&gf7VqfDj5%M`sJ;a<uERo6IS_|wPegXC{E$El
zhhn0-j*8tjS@;6d?8!R>m$s@30co~<E6NuC&Z*bm@0PA**$GkDIXZppaAk`$9~zTG
zd)K9sM5T4`7e)LBJz<3ikI&NzsvM=T-JV%pR7XcSJ8d|0jVPUhbGWcgU(Gi1>iTLA
z6k<<65sF<Q{6$l{tx^RBpGisC2~eZOLE9lt-*Xn`w>}$xs*HNNI3Bmi-##umMucgX
zZgnZR)=RmUlD*V5BS0xkyQx`h`pd{&g$}~}w5967zT3L40k30BPTwPJYt5-<C8e!}
ze|#0Y;2zE7MISly?QU{BjIg+vBEXo$&s48{eW63EM)!fRSa$r~wt|Coscm~(%xFpi
ztH(pvSu1OU48-a%@p6j%>*w!SUrJfoO-IOv5M%V*&kCH$gbX)o8y;qrD(ljMqnBt|
z1DMQqMuJ9kx2Zt=!wA*I>xHt$j6X2s&eBCVySw)6QHU(=w~N8deM?_+rwd41eoO|}
zJoc78&(-0&UT1X{8qEYMu`cN~?Zcz}Oq=H-;Yg##jXq+-=6h)!^L>>O=)fBDYOw70
zG|Br?+vqw&CMo`1fm4!4a>m${|BP0jjD+*uvUw{lKY&l(HfNo2%6!J3I(ci`SOP;t
zC`2JZ_44p{tyc02b9D4#oasRX(6d~D+x8(1y<n^hdFMda(8-h>4|%enShqO5$Y8Ia
zy$QGJD;zG?(aI=+T4l-#*lRvdOWaG6Xj(^~!c@l7%a6xMTL4QBEd8c7xPJRV+_RTq
zG8VL{XIwUDqIOjowq(8JwlMoH-wn?aNwYGslgOQDOXdQj63sXqKI<upSKbdHm0|od
zN3LoK@UgUM4P`**;I}jP`Ky!h7up*yQ!h!f*?ZcIJV?k;nm2Il{x<KFW+(M^+{@>N
zdeVsRu-m1VT8S$3rBR<p*<G{<WwrJZt)^_?Wjl;Ztx&c<8#V&7a(L)wb*Vr)JJYIK
z*aAl<;`}68uCKMWbPq~Y{#Zy3jrSVcRrl7Id(dGf?lh-V{5|Eih?M=%ZzXG&U;OtH
zM>jQ3ugx|4g)W264(FQ1ed8|-v6q_am{i?^k2UNU{%C1%$hX}4Y+y9=$Jd@3A?8|Q
zbkCp{^GVu|h`nt_Z=rOxIHefBE;I*|d`KJYwIfzl&G?c-$2V?rZ&c`Io`=qx$i0Ge
zcDhBlJ*#yI1_41aU=2}<Vc;0U{Bfz_eJ&#x9|zV<u8&>}!wY7HIOw$`Lkqu0;UR}N
zPKJeJVA-x@!fvrnpjL?wGT53KEo(VjnLha3g2mV-j8^M=T<TB$`1V4Hol3iEQBUTn
z$}u*H0tpWy!K!fD>Pn@-2Y$Q(<4CNGR^ThhyIpYt!n%|9?4*%NvF48t?~i!)x0VDH
zPyD;w<igf)-9f_1cWC8Wm*LWt8%&-{(+-WOfjLIlNzx8!z4}J<)&BZG^c;b>6Xi42
zMDp`Eru=EXF(3Sp6Zgr(!l6QZ3?Ue<bu1gB%O9ffI;taZvNm-5@sPIUa{tM=f553?
z&;%*PutVrX;W5Ezaa3)oU(<{*UcBtChQ*)c?e3hu9zH<8>JGnmCAy;<DF5hI+{f`&
z8#+#~Ke&^BGRy-vALxS;&jCi!jVS`&U~ZXFe&4lnb?2ga>f+(k&YF#-1w!#vyHVX6
z#yEvzX){H(t$QGE0xBnu<eWFQN3LvnGo-XSv+|1XjwadPZZLT7wOZ~H6<m|D9NBhM
zdtw<0vh9*yIat+`Bq)JZmI!I)1f#Jg%+E_1TT9d=*oKByWamlUk<67@V-=m{fcdM#
zkxg{x!R34e67FNFvGqs>?@5v$O}*mZ%=p+%e67qm0~JDB9li>Iv^p<rOio<Wk&`b=
z1DXJ)ey?0ADREe6yAgZNM?qqVp(|+pPe#ixUR&}PZ;91j4^04z!(Axae-%($1A~s`
zGp@R>&*9=jS_ok*_Dx?g&JEDtlsS!UmXY+JUQ_4rJPAN%mT?XaeUYLD5ueFU5-Gap
zze})=gczaAHZRcoEv^!Kei+Y%)~r01>_70JhKP64=*GW#;F|9FlPdAE@Wqar7riRI
zp7U3%Nt~I}dplJ4DP7KZD?nvrfXt!XMR>5Ha{d8yUu7I7vJ~}8!xC}|T1&Nq0-GK^
z{w^(Hq$wP<4r4##cpfJ&^V`x};mQx3e0eQIxai7o^<44Z5-9c9HB0C+?QIAWKZ3-5
z5y!u}KwJ)Oq~Ic;P&&=h`QJ=PSZlxdL`o8gxQcr`)f#UBqfE^Xje8suG7<DxK{wv?
zjwY|gUbt@9*3g$2h@!YQcD&T(+_ha{`QxB^N(NkN->rl2i-5RGn5Fmg`*VwLcF2ln
z7Hp}kb4=n{*RDMW+$XQ81h(8^<f$tp!vUA+6fiakcL+)M`1^Kf6Kal8ZVQWAxym(b
zA1t|mk5`bnQi`S*$MmKHm>Ks8g4G!$aO>!af`vsYd6cwAfwz(c_{Dz7Mw)i6HQQ_F
z6P!r7dxzCSX;6!Wl%e;rYgY*I;9xwi2b9{I$yooK`$Oc1YC_KY`Tm+KMI}X83=SFl
zH{?lN!H0nFR8~fpO#w+dKApjNOPRf`2O{5mg#+BjUKb7JIy=noEz{rjspQ%Z5BoFx
z1_=89u^|Jpq3@nkQA}&>3gtd?_IIRYv*=^sLH}NoKoaYFA^>4(;-t=M88{bzFO;%d
z!hMR+(hi^)0bui{^XbeC;}GXTk~{lbH&uKN3le<oV}b%a$F6WqJAp$xmhT#Ss`2(E
z4LsI&8d)G9dT1=xH!%1jAbyGPa2mhb>(=wPqxT<M|6gl9pj1F~6BBX=*qbh`WsJ2j
z;^8ZiR0pgx3803noYuenGj#!cLkVvWHLxUFe*_)}ER?c28NLA*Dy>d}fa!tYqaaT@
z>V?5@L&B^$cxz>iP;E!5A7D(}T!2x;2G*eoiqPheGIt|Y<)kTcd1FDx;U!}x%k*t@
zjj(kMps-W&mA`esC0t3Z%Y!x!(*!U#NNyd4lT3)D`r-4B9))I{<9jbS{?9z)-v{uI
z)n(DYaWa2?`5(0z!0`frB6`4>{0H{L^swwOT>t-_A_CdPg<s6g;o~&kbHv>JD3{e4
z@7YU@p}MAbq4muKN49N3Tr;O*D$%uU;}@kD=;3mU@lPzVfmU{Zd6`kr6Z`3MThFnU
zQ5U@o@csU0VIwX~o)ay75afsPeX%AwR)J#I{!lcpGovD)yo44V`!Y~sgk3ZwW-AvN
z3eY^FRosd7_Kf0v1&d&zWLa*8)0}qG)k3|0#;c8GzI@KiUXd=?EaGc_PQ6qoZ@>{#
z;x_8+wP0wd@kVlyDYGO=RzBMDoTYuXQEgn!P&8<7WTQ<xs+pTOo^^aQCUKoaT~Th;
z9o%7e@0w*h4w&Oz@$m4P8g#DqnR+|lm+3!nA8+nnqT**+_SpE_TC_Q=dc5RSEs9_4
z264?T!3!304AvZc^V0R)Pj<JL!k+F@_I$3H4VCf9f(E!IM0*jouRpL)LF<Q0c?f;8
zzm{T*qcyIkDB(WZV9{2R>Gy+@`#w%Z4CYXw5;IWF>An_<yx=+%&rotIf<wF)-sY)^
zd1rA+%4)bGoSgSOBI<R`I^)%H`=Wijp|EG$-t0fjF(pt#Qgub;V&&Sh{CrP<daU3$
z^$(^?v5*VfPJDHKKv_h(Nazk8?1$w6B!-`JE0x5<MAw5A|IUnU75FXjcD~}DZr&-(
zwH|tb?3M31V0R}3d=)s>@)fk}_(33QrPLmcDo|#zW+CBeaypY=9wd|+W?@3KIt+;D
z^}>e?frtU_6|)}fHj=F=*Y30%5>eS%9yq30)2||SmPF4V)=WL^mP<PYdP@ufdcQI4
zTuqskZC~&&1fw(aB78ctnRp{^0d9~COZ1*?{bAnrMpF+wyKB_`3oBb6`A2fF_0NDH
z`&J6KYZ5J4jNjz@TE14;3?(<12i;uBQbie0X*D?cs^n@XSYr@XF3!fXD3#>m6h5wp
z97V(aU%vEL85<kB1@FOh<|dn)hX*IRQq4d^T`XS?StC2=i3(W<E?KALgx%dZNzKc&
zUMeBKdI#S4$i-i(s1u4ochfQsLp950@`pr^N`>DRrKoXzt*l`yR1h~vU}a#vXXcE1
zCCsUtu?qC{SkyaoxZ!I0&bllx5pkJ(Ur`e-6U&m|{`O1`ak0e_HbvQOmc6@*pL2hs
zJG~D<(t7!QpLKZ}Z(q?g?AL*n(dIUk5Z$^Aj=k#a>K$H{bvnPwkg$KmoeulS4!88f
zF_D!4g0F*TvE?TV!8S`?_|qS35P_X?<NHSS&Wo5e%?yRcwMr2xkV1!Wv?aHp(YG$Y
zcD>I|nb}_=;j`%w`7b5)QU|fMKAr2K6QdqDRxyjHn}1QHYbTEGF;3nTWR88-IzDJ_
z*4A2zxan@L1RHba(g`0adsodyhP~|}UJZp-tnHKp)K#Y8pju91W2u#776puzOSAkC
zKHA>NWRYw;iJ&zlwH(a5ipcMc>@k{nFfw>lG)&ovaF#Wo$M?W@rdPqdx8^4=xG+?d
zcObKyL&RaqHD6Xds>Z(zapz`Yv_mJF%wd+!ZwrwshF{Z)L-v}Sg61>j))7p*L}|YO
z(Bb9=xjA&FI%M6`DRUCK=1sH4{ffHnCn3rmtYFu2^v){t<Y?K6j*CWR4RS1|$vZq5
zYGU=AhKpAtlVMLMXnTU9C{@$yqy<R!Qy!ii=#J$K2FpC!kuRQ_JYG@$0RZKeS6KX_
zgS);1bPc-r+wE+X9dPP;nZxMs4V6EoFV?$2epVQk8Q$MrE~|?frgAPaLHn+x?JCx3
ziVWY<_@!qj`u!fpygMK+r9GXGIgK88^SRH2vNzegIvnt&k0b{5*js}fha-Eka&2BI
z$ZRSDxLuA2q=Ibt@A{!{dPokwaKG$!AL%Mk>PznBx~F;5Q5iMvBWhQ<vqK4x5l6N1
z_nkj*(T5>N(=T*oYBid~c+%Ntg#3X4(ta%Ir4rgfw?=5*xuS?+tR56?jHjW$`6ZX>
zLb-%<mA-A}kGx4InB-C7M`h~G(+`CQ;?zUVwx^N<W-or&{6(=_N<$a_u~P<00u1ob
z3PuFe!Q5|XFV<u}+u1dz&rUUKnwTU<SH%21TWK_+kZCf~Bm!;7opCMRlz)UD<m)&c
zxqI(H2MxWcod4lT4HG?+mxXOf_V4t&hi@68d1wfkpc_Nf?loRgq=_M&M_xE>+3H*M
zhm&v%trFK?G0pOcC!u^xeig@b{5S2Y9`Q=8a_}p|>@PecXscYvGi!u2@I%MZ#29^g
zw}--|hRH|6fkng3ZwojIgLTP$XO;><IP%vOn78OzSag7TPdVrKz`fqj;|I5+hWE#k
z$8OtU2OLN2hvs22s&70I`cVvCm+}Va+M$67`EPBAT?ElkbysYG!hEl4UK~<S18lEV
z{Pf<|3C+{a#@p6|vqrtl#GQt1D;B&cAW2<Y{k7XsiKi{9WZo_}yBQlJ)#~CCG@JIp
zYDE2Gy2Bc)4KSjIHhL@GQ|hn2d$K9*AJ?!gIlGa5`%$IZW1VJ~*Cu=gh9wKR0sfp7
zajN#XjhT@2JlEy=P>72|kmvf7Qub@3FsM!U5MN1@m<VAAX^>I1z?~d?$oS5DW4b&<
z&6L|%MXUY{Ax_gcy01H&Zcfq<4aRyt(z4@-(%+*;d_r^fcp}nth-oX|#t@kYD|bP;
z%SZAf1y{KfbF!aC++Cp4zA`*L>TFP3gx5diuP90P{MEQePCfoKH@e+)MvW_%tfpp8
zyEjnmfARh}7nn*)Sec(KwNbUIzunjsPICL1$&!Mx;NCoBRo(bnPTW~3pR4SN+g@#{
z_oys2D3*Ai^@KrT+2Tth4P9jfqZ_PV&!hcF$yBxcet%C-e`d-)(bNF~Gbjm7v2skE
zDf8ZuNP0`;&3dDb4D#VM&yRlsx?z2xb1+YwYeGFCAvM43>5oe`P{b8$XPPjVexWhm
z`%__htAPqqVQN|pG?g4F94YFN!#9OM?X1x_oV_?ruI?>Q!@?vFTNg>zvPsxW1<Z14
z>nAf5UKQ*RbA3?;k+z)r-V2h<9>V-Xz_`&--}S(wsPm&|wAiPCdlFI9(!((Jt<>X!
z@}(Qe!Srs4NT#|Z*qZOJ=4J^nsgkN>TB^XdqEo<wwmxRyWlcWor;*!B+A3n~rVJz0
ziuC;Q$trE}ue(>Q*480y+L$d;B9{lc(Q~>`=Y(Ig$;VGYM$TT{J?>&c*lO3?u6x>q
za?Q7MPt8hC`thGMG}ow&;0hPDbIbx3ju#LCXd+x5HQQ*}9@-gmDJ3dX$Me05={<3=
zcshE0Z_W32^X1!6fg7A9VneYtDSS^tXo^0>7#KY5?EL2IMIJounS>sYd@ZF<4|w8-
zBOW=cTE@pC<GUYaYlMbc-QY?b?Qf7Y@CW%*YD|L3q=-SC##yyi+$?8*&(BQ5T|Tg!
zP-bYSN-q<b9B}*G&%r5_Ge$cSZ+$q`5FA7mFM7B+8`{)(3Ep(F-B>q+h=`Nof^=2w
zO2-d~TQ;nBp3+LitcTc`eH|C)IaPW<?TrgdI1b~B;!Ls@{RMMt4(JZmujWZ0c}N1E
z-N3&nrh8e00TPN=sh82#!l1*J45A6w$W}b_N;`*LcOdiIH>)$87LOR1<FrC;&{Qh>
zO*F5W2>ZTm`YmMrffBWWl)=25_l~5%4>bizo{41J^nlKtrL8DB=gm&xA5Far^JsUW
zzR84+lOJ}(jwgG&anSlQTYT&K?-xZ`QY+1LR$w_`Fl+BDEfq3L<TiiRzYw8@wBJQt
zRhyd*?l`}95Ijtbn51g?P^lhH%d+AHQqo;pacQeLYOINSW+4T$939m<Iw~BLv)OG%
z<z)yfnjVz8ESE>4`UmaC)C{6kvt<^G-NCnysB2C3v=_%Sjpg<?4S33no1rV2hX;q8
zUwyugv10GjNF^$<=NVj2hI1F(_*u*ie?d=A&m4jKc;+%b@W~;9la=20!<21X0#Ell
z(=u+EJYfMXxuX+qH;x8%uxJ^z4qMKfBlG5w15#7hp1=xB9gV}kG(_Nr7vHreWAC}B
z>SYAT{nl;`ZhE`T6$oaTK#4Wd@cp2@D{${uU}_yQEv7<lUqsG6rbC3L-QdIxP-q%X
zrRnz~&(Pq!yC;h{C*7`NM{kp9C=*_b*RZD5EWe$nfZLq@R%waR0^PB+oy=zi|Mcab
zDyPo1QbfCrE|%3A(1_P;RF$U%sCM@yP!f6}F@Go65Fc?B%^BTOEz10LNm7AUi~SAb
zC#lA3wMm9#P|(iZ_U3CR4T1B@=%iZtSklBP)XL~pXhPfgcn;4a&f`N36aJU{rRmf2
zLh1Xg)1aZ9I2Uu`q%f){s@d7TKGZU{S)M+8h<ba>9C6FxyCmnwjj=C@Yx8ks0bZhz
zriT`BuedVbzWy^p`PXtT@Nk2G=w`<_w9Zl{z&Xl4uhKKIfR2Y=-_cWNZn#pjO|Z&k
zf@YLkfS`b(O<~M~fav6tn}Z?%Tw|g>oDLeKY?8MyQbUQlt%csbSXl{bd<;;ahww#Q
z!n@580Qt#h7LpZCGL7qh3=BFaLdlI`ttjC;uK}H7!hHYSmW3Bk7zi{dEN@0Zi%MTn
z2UFgXSm>zIO%`wJshOOm^h<60jsnD+$NwP!0h-I-jXe*Y)p~Z858smPw4*{M^#DQ+
zckLFT9<CQt1+873paepvMw+&-|JhP`@t?apuX72&+`oqRGLYQT?&#@c#bZBOJG+ko
zQB_9dz13Nku`H=@3dVm!qA34qV|xvrAv7>%Sp0KD51c_JR_0AO*Y;kZs=lc)i}QiU
z)&*9lxn<#R`flV}f1r@(UxETd@U_lb_x0UFWvu}w-v!)I&5dZ_|Fbo9%n`6WGdE1v
z_ZAlT2#TY{YTd%F?Yx2({v($ey;Rr;F0SEB(Qg&^zK6Cf=+NigvBPBj{`vnZu>{Ve
zz6Y1Cu>4tAN|!Oolem9@^Ss!~gZy54|8NQhWrt1`BkfPDPRJxZU{-$RQiBiIvM{CV
zm(mVVx~TNMwBkfA_w!E{CXrd1g?*+Q=K&O*h{s+)_(^>&lJmO6k}#NjbaB5N;#tGp
znbd+a6eB<af{(65m-jMv#<*Cx4gT`t>aW`&SUJqnL}MWYG=|Ur4o>Ix7{brJ2#e27
zLWoL;fdW4WM`I>lyO25(==Yp9BWmn2N|5X)e~YF6dMg4fqyO=|{(b+2oQ#~Tx0*ze
ze*v2Q@fWI^bWTq6AB499w-N3D<<m^@C;wfg&&pg-e5h4@rvKJCuPn7|Jm%HU8#&pQ
zjkWu+Jo~GEvf2OMG|*t2u(w6Ms>A+Rgtu-3u<}m2*f+MYvU*)Gg1r70MN1bzxn7^a
zf-5Ov5cEvGw*PP-0=71f)nR}<F3kK#`wT=60qqQ5wD%g}u-^Y40GE+kTU|`T?lNgb
zNx&AUA=4h68Vr{`@ZC5(?;9+B7}kM!g-Wxlf)}=^Ve)pLQ!t`vWYNQ*?EMu03kG0@
zYn;e)ODn8Kje|sCzXdYj#l;Vl*HJ|+xLDHo(>vi%B$6J+8B3}!|MG8#<UgOMZl*QR
z!RZQgaGraLrZ;I}b-aQF<ieGyIdK0B!qfeHReW;&`8Wu!?Vrk5(zJa@;Z@+Bae;yx
z=Y4|c<`S;-J8PkLnOhR>H^%zz{-R3Sn<LB*m40P(pkqG@|NZGdr}Y1>TmMyeEhyIC
zd34}@5HovIrA`2be-wT4O?~^({XtiqUqCM$Wtu>L54M&CfDKy8H)g2{N7mJrkj3Xq
z5q`JA{(*6(ptyFy{-G-3iiLxH#%5Qa)0Lft+@6BSdY0Ny3VaQ3+uCDC<0<>kL5`|j
zWlv)TN}q~E?A4_Tngs(zmq6jbr6c{1YPVx0M25|Q0_1zC4ULf=pUbccn7Qdpx2+nm
zVw!U>zYP0_=oAL|GSTwCC_vm|q#xghOVS<T40)hyHJoB>&SX>H(^9wCM#?an1TzU-
zp`S1JEbK>*Pdx7)W;FvsHqWb2Z$M=-f7UL>zM6PocsH6H<zZp`7V-F~AEV>>nV;99
zCnNGh`HxBmA1ifuj#GsvK2rSghs&+E&~}%0zkRs%NfRTw)%X32a*W#`@3=J6kg%ZP
zedA*M&hBw}u{mnk(h^R#*48Ul_uzJd3TQeiF&Yp^p_Vd7r(wl?Jpu;sbnDB{N^D81
zrFU7!4LwF%U7lxa2NqS!IO?+*P0!RM8N~8!ME2kfr^GZ%xNH$Y?#q2-D0EUYnC55P
z*e9WVg7W8LdzF$D&9~0r+l{oy1#s0(LNlOYT(5m_WY?&fDLHL}(8zC1(Mv5%*K}=<
zqL1iG-dFB^D1~z|OJr~s(>*7?;2Nh4Im~lifC99{&@VJh!CWV#XKBkGE2cd;kek`j
zqtGBxmTdC-)q)hWq;)gqhy_<tSK+t)(H<6lZR2I5+KT7HOJ!=WF@rt3MbkB7eb}94
zUXsH+!iL0cCA}jjjir`4)~tABU+5y#qeWEEZxP}wG1u2vZl``DBl9qHxJ*OBB`*c!
zb7N{Q{fVcYUB?Za$I+FPVfNb(J10YEU~>B6kcO?F*d}u3Axi9KbK5Q4!gseft#8s;
z0$g_k2f7a}exBC~%X-NAv2_f`B+FGrty)AH@pZS#)Rg-!I1*i#XVDoUkf$@0SXHVa
z9{c&iWJtZi@0eXIN03m&c)hfH^ZTCD8&1kjg@%cCxye_!ZPm)(>9gJg1@6qDo{f@3
zqOUJRMn-+5^!@uYfBv!cRdS-2<7#+$Wq0ANgx$5G3OXjd&gGwZV4m{qF&96P3r6J0
z<3Jc!xzl(9<6LMJ&Ot+pjWw0flPxhtk6kMV-T!cqAe%+g_D1p0=hzCs&Kp_3JDWQI
z5jvaX&Vs<!UOWB5DzhSy?!qJM6o;H2&+e!RSiryUe|;pSd*xXjq%7bK@4nUm^!axd
z<awSDb)h+qy)3*p&YH-|>b|nDj1sOCSX(<pLHGgd+9juyd(J{w8rgNb&rAu8<r%Ny
zf~8D_8?~y3N(b3b1AInj<m63JVYM7TUHM99ta=(|e@a~G>nW3aGa`q{qz^ZkOcA6N
z@NRgbWn%sWR<_SA>Dn-O6H4VV#}C3Nqv_$dB&+Jkd$s=CDk^J+TOEb`b+I>y3f_k1
z_tBu$A4nB?pM-SQ#iBXO2Dj|BM$q0y?`4R80HR>Dr2$7`eNC(n8@`zm9kgF~zCZ1o
z+}2xMEUyo<U1B(rw8g8&lNo8>2<1(A7ID{+TODSAH9@j_Cw0(ESv;bZI-J|M{Cp2>
zANsSqTuGcJWMpaMh3T5brRI0fLUaRGi7B|r>rRPf-L4uh)fM*h4Rd=CmLHw&X*qg#
zSo!^0OU##m*w%rtlXm5qFe&>`a~E^2L5Rf~4{v@k?EVhdJ(EK8l3L;0&jufkVj|BX
zPR$Bxz3NEK!Yj#{ArIZHKA!T(bKJULG-co2(^yh<(Y?e!W*y691UKV_cxHxLqS?qa
zFz(c{x!WAd3L`m-#cZsfK2(b=MAh%FSDrTb51GYZb&n$FCl3k>ygVm*D7dhpt!p%@
z9O$WVEuN&`*^>z+^y-WBf~sc74XufLy&|<@hbNC-mey0Tnqw;G5(4QNU8niCSF0zu
z7x7bG{V_VK61!ca+QnPVw{wDOpAA>mdKsAl{VySbwMQ>dyJ(mF%cwP7{gbZ5eg3NL
zTO<q4&tnCdl5m>i%D8d`yY;}!9A}W65Vaj^5jgES$1AzI<+6p429}~|`Ge{4Rwb-i
zOBBsIa|V6AM*=-7r&(fP!wjtv8vbP{B-xAHfzqu_wWdN{b0$b>y1!?1+BD_jxG`gE
z^QElPJcTVgaF9idpEgl1Klg3d>oMk~;j!y3ifpkO>K7!t$%~;Zi(b}XdS8nSSd&vl
zuvCWdAW~FICE$05n4GZaY?`_cFwyUu1@+`Am}RrH^EA`^z%!g-GMK8Vo|vd=zO&F$
zJmp%qCOy%$*FlapYrr+JzKn<nvpROd`c!EKR<4hf-W*TYtjl<ZGi72x(PSz9MZxI&
zQJ^^S{jWL_?Ooo8g8b_It5xyrC%z%nJ`?q2u3xf=fva*7lW#gIIUWLphDtgmar2az
zw7&T%7Dj`8)HZhlb6v4HJ#$@7m=6itvC_<7CWdnWWp`lgKHJ)NX$iNpfyL)queh3S
zT3m!N8flOlKbU48>_&OZ(YFthP2x_EtqNMc_~(rq>Mw1`+tFlXJ#l*UR&>tO>X^mG
zwrmSSu08qaQ`IuM{k^a?eQys|xf`uO9g{zKQ^gR=BARd%E#TW5Anc!)#gl`&B^{?a
zl`AWy9GEueu;BoW#LTq5@<Y0dE>~Hw0iZg3-?Ilfi*OE_0yi7(D=fyU>Y=&k-gx91
zIaQ@rc~n<NHi-?D1~_wz5&3^4Bx<pMEl%B1As&yjPQZXoulOjxrAQG7R_Y@)W2F#a
z?trhnDgm|hh2K)UIirLHSNWlx@w{@e4XwTW(7Lg~>ieCYuVnM`)=6;D(7+}}g(2nI
zy^n*AJJBCxKV;s!P^O^*Es96qH5$v-$Q3SF3!aG8;N`vhR%#|;d_169hFEq%hZ}$N
zEb;?Bw&`J6itpqibigl=2zBGIfHr^ATV!r$azXp?T>+P#InZw9_0IjC8QDBHF*@H!
zCi(BCx06evYCAJo5UJMe!WNHIukGv2Th@yAkf{*2y9@=M@kHu>?r`Eo-gaN@D&gkg
zzb7tER*<2+^ok{{J5<Z0`^08Edd!E1_l9lJhbDS{`Z4tcxDzWYC*`T1Ced@OcC#{|
zJlK<1B_7p<CVA-!`>tVzb{=my2p{nT+1oX{eJ@fCtgx9L<?b)-XE+vsUtHYg-z^eI
zxJ}DH$_8gPD-oM~67ooC6WyTw1$Q{TI+g`(2Cp>uulLM>bEmEoif?46J|hFY@IvaQ
zJbtuA$-bNcn2Dmk(Cz%%>HZ9Z#9E!))U1$UG5C!kP#NJWC|84@M%;qTSP!i_pao|g
za%TPri71I&x`LLJDx-e?A~h?xF*Gi>Yjgr@XTy^u=P`rB+l1(jB-wjnF`=^yrGCOc
zEs1d2OCxCd0v6g<MR-SU+)c!fscnNan~jp;q%MxDmG;U=alywD^aw@?GwE==(((Ym
zIQUy9ero$*cUoGcH7?+gH}qZfR|`oKknL}KHFrI{(!%%R8z*l=LwQ=P>XzqqIp2F6
zU$uDO=++;fTJ5a9NGt<Ch+nX*<o|?^;B^XeG@-vMZ_Fv=JE>lr8ZkJD&Hen|q>&BZ
ztgZR+?A>1!=lJAM(bllJTk@e0&rMnFuivT^%7`iolH4W_u39}gO18eBG1VCf3^dmt
z25;*|2VaKx%)ej!5?`&Bz_ZA8_gnVi<S=Kl?|=bxHSh5E$AKkMdhU<!8nsJMOIU|B
z2OC@i_ZIS2`N@$qpE}y>02wAiSf83Md=<=i-&yCu!aMt^-@ccwK4`LbX<ux+>=z{c
zA##LNSm*QMp{Lc{GNHAta82Y46Z%Kgc*pD5FrsjS^P|o^^uTnv_~(SrESny(oW9%;
zCs0r)?#V5(&v>DsGZ2zI3^p7+T#?$}Q0s?I6$XApVdc%g5rSf=TaTlznkEmAb>dM<
zj3GypRXqbW{JS26uz>m+-z%A07?&A5_|BgIrKzy&_Tl+Ifx(t-r_~<7-~RN^c>nLL
zAWd=3Y0;%ZHI!MDtr2GSy#H}v_4cI__sLh+CFhVdU~oT%wD9UZg(O(vXnC^8Ol;1@
z^D755q&DWym`cXi*JEb2P3vG2CtUfzx~=%EZf6Aq)GlvbteU}h*fl+OHg~jZ?G8Q;
zIS24WERT~M=>8OBO|^z4_$+R2GRkk)O$wffQB=>MwvK(mCj|i>jXPToz`PCzJT{yE
z7`pz4biD2kzNM?J+chp;n&FUW)iv2e-&PHtuzKNXfuw2r^UNt2th7TO-HDlW6G^)z
zv+K)(eeRHYCP=t03CJ=Ln30+4a|$va1+00_L+A{qRws*6>&(2CQ(g<JbyU5}I)hUB
z7Fv2W;mca?4K>84pDQ>NaMhCpEI-sd>ka3`g&zah<n!PQkc^{mjFMJT{|IhUP*6pw
zLXNfxUsRk$1YI$|4DSEAK+N`EvP1caL~7dET6mE*&%dWF4@5lDbI3VO)z10k3BHW-
zhiVOuza~I@rX7&i(Pgd)B0#O%-`AED0M9LeQ;go=Bt_~z&@U>(g2H!@dYH$s7AWL9
z2q3lq4^wA_pPQSf*JqI=O;fV^q&9_6>iYrQ=gU?Ft?`t8W?s|4YyM}^H3iFV9X<)|
zsR6=?SH%g_JLh;B*5|KK=rnWWSxh}w1M>>^e5AOL4%c^D7zV1$up3|hBvj{zNNN6)
z;E(?rCMyIsWd9)UL`FeRPYCfp>A+1us23A*pCd$}_HxAb;$trHS*{Y^{y+RR)8r=i
zQBBe-&r=}Y`}a`*t8=<I6jnr(&!=iJSn*$X{`XJrKb&;tJ)Gm;z&?>{$3v0aBgHJJ
z=EDoBMAISo8kGv*qRE~Ze-bx-)cHS*E8Ouq+*I(c|Jv-oNa=t@&F}r4|Jn^86SV*w
z<X*+gt*?u7o&$e5_jG6aadB|+zqfc*Y;En){m6zxr2C(j<ic{RPMHt2L&z~D1r>U@
zde-vT=TVaN<J8JY=ZSG4gj~SirPy?CZjXHkY5zmmxv+>tL|UCuKYvU?fdJ0foA1P(
zPp^S}0t+QiVF3zwz826wl1e$RrEZv?=gK;37Ylc9!0Jrr^^gvf|DczqClDIE`kxQ-
zf4GKxVd+bO2%2lVBBp)f8C+&ejB<HqA)&AH58T<bwl6+Cw%I;WrD*ua-u^FK|3|K<
z2_KHH-UV{AG7}i1U#S-UMRz`&(X(!Fi3vk2FB<6%7NwWum6H#{!@dGa)+W&D#wfRf
zi#>ZBrl`+Jd%7C(L?C@@1Y2~$9O3qT1;{;9nsl}9Y)-bWT`@iX=zdu*W6htuIAB}E
zEi#7UIltb6pJ#zNUb{Fit$UaY&3+paAl>@_o?JLs!xEa)cCS)pxx79MVT-JXNK64e
z1RlE|mP3>K2-?8>VQt`fB?PsZMdo8Eo%=OTrYX2npOxl2|A)M{42z>{+C_mtLLfjO
zNbum6!GdcbK+qxB;2PZBB_TKjf(-$J`yhk6gy8NvxVyXM49Sz{{hs%G&%XBF=lnR=
zKK*05r@PkbRkf;@R^3&n{uANR%F}ru`SX&2iU8qbdXg8H)<WQF;xHqa|HP%cm2?S4
znN46HV$lAh$E$)R>kOLnsFPXqVRY4zK~=nwZdCzEb^Nq0m3lqxycs?yc6P|RRsnHS
zrdm&Wcio}TarU117H^M1B45kCF!8SgkJP0XvX3y5;BHl04lDQxm-t>@?OeHFP21Vc
zo^>0fJ>cmDo$edQr4=2~i@?WN;%yk{7$peTFB8kk&H_@o+BDB^B1JBv*MT1Xk@F9)
zl#FU~&#eJCD&K?Rh{@llR^$731fLT;Aesz^RkhhzrBt@Pv2M&~Y5BZcJ7>I`{4_>1
zwq#q7o_c#oZ&g|zlgPCKm_nt?En#f5Ml`#WLgr=5s{fvh+t`(Q%Ew8H#u4NUdQv{&
zo4ytTvvCc(*wU;`A=}e>`0OH**K(=>?%g~Wl31<XHP?j9&6ks0OhInR6ZjKhqYqP*
z<ohq}f5^hsvJSfL<Px_dCO@dfK=P)Rs3~6SEJNGKX;4jjlQ+-pFBHL?!xz<>3sm;%
zPsal(tK|T|Km)28s%-tL=Z;wYRIl6?IxKe$mh4j4s+5^Z&&^@kQzw@(b=HxViksjH
z$~2W+CHm1TIW!WUAOn)xF{$M*SBA>eMWS(BTK-YW98E3OjQJD0_3=<Z=s?}T;7L7U
zuQAPPG+}}4_k~zz!J6Pn8zCp)M1frT$o4@7uTsXS)m1dyI>Gcv4;aA}sgnD~DJOmm
z)pJJQN+=*nY`aMRSrPwDq{m&|)y}i&kKrCtwsoYL?{|q>LR}VO4^3fq7L}YtUiGtG
zl63%8rrbES^sc<ow6@%~Bz!~VI@v?XD5_|m-XN!nJXZ&n<i0tnvc+Mws%>9mH?Kxw
ziNk#p9bH@B?VcVk3j#2MK`V^e%Kfk^*0Cd9!HmmGbji)Jn|-atZ&lQ?GJ{o-E?ep4
z+c$M3eI1WYIV;!dJZhJ)1_sE=<U(pH+It3Ad3H5}?UsyAAl!AQ8dhj1P4Z3<44f3}
zVYT45S#Kkdo;}^f%~c2nP<geO&U<y}+I8U7ty;}j{zQ1`Uv;|JVr`Px6JS#^K~r7Z
z&u{c&H`v4g*TcHZocCQWj6(*J!Z=q{Cm*Sp-JLn})TCKMCt5nEF>r9iUd^6eE+<N&
zRC&8d9ZMHVb|U=db70L}TVGl{k;Kb2Jg!2GyTD-AxPa43=hFtFPZCfidOT6%<dZtD
z*akHIBl_d`U<EjyBauQ@hZMML^+ok;Y@GM$-pL6g%r1v^$}8r1blC_=(apQ!-H=!p
zI@f5s-l8dPFJ`J^3VqA1l~~#6fT|zyGJ$W5<JsSo>@wc0a!VHudx98xjG?3#kQMQ_
zbSM`tcd%E}Z-{DK`zrNIlbrjTO2@)f1F*H6zS=E<QYu$VAx3Qv*~fPzCr}<{5ji0>
zkTHc<=X!4{tunJnZ?@T`JN50-R5EG{#5KEEP1((EPhe4X8*B2_B=_B#$JwpDC)y`b
zOA@g<8Po@@v<KS^K6??pER&{*Yg)eT9J5o{>N=Y1*q$XxO`YOpDlG|~3pVWX+_zxC
ztkNP)nbI#~A~;J}*2YP}2?V9BdJjRxd<aYQF9zDvrLsl8z{+;e`BF1!5?1GRjTm#J
z7#$+mtePV!oy6CSae1IJt5qox<SgV$jCGOBb^P_-y6?TcvRzkgcTT2CLaIN<1&#9S
zEd3xB>aI9w4vwLZO`t(dg1)n=PNKnH$$nC~I7~ntZ*(?Bl0rM<l$BW}k8vxV-`{n3
zFd%SAB+M<_X(CVJkQZw6-rB%gl0fd}_F=R&sYJAu!mb*I!bTwj?nRhlF6&~m2#1WF
zjGJ8HIXFWJN`GnN0$ME_afUI&WG~%!u8vOxKqzbPaY`#bb4ds-WZ9E81&xygLx&jg
zh&YH&IbeBj>`_EtZi&m~==Vl2QbDXF)p20z6zyPGE)AY|mwlJVvGR0-{q#qcCVpm_
zdGJ1YaP{d_r306|<(tx1eUbu#sm2dC+Am$@m<uucjngd3fZ3`=d=6u#;6jii+*z_G
z^le0=Jm4aXZyRFGC^;zE4$-9sdW|17?I_93nDiP`8NYwdB5T)fK4TGx4V~&!wwA7E
ze2MmiSCKG0s368&j*>&ekyb;$oC}=6?zfS&YD3@mR^;Sc;kpT{8)d6`ID--7B+yvf
zsg}>GyC;?(G+$>nroj>x1I3O8R&S0>SqXCGGdzvOF>v`z|0Qar%LTddE-rk}WCJhS
zqU5%im_B#UI3&_q@LdPL#6~%bs+GH9Uirmof}HpT?sY<hth;P_j&f>W1U{taW#@(h
zMJ&6o3^M~2mzce#H>ii%)4k96!Rwjr6<u;J=l1~J$9A=|@%!#_qWX)Cv@K!RU}q{=
zm)qNiUX%q^v`kK!;v3y1Hj;Lb%*SF|;rx*?fe~V*o)OQ$o;$`qbShW=EnqG4+Jhn$
z!^Cs@Xy74QMWK$8_*H0Zoy4NQawUXa18(hNGXi{Xc7f^{c_7}TTjF`OcS&y0;gxqS
zT1F+AWV1A9cF0v?r5wLu4r(i7P716XJ#%qXdX;;IA~X}s8%uqnsiZ`a1<tmLkB|Rh
z-Lu~0?aAXd+55=1$J?NR{FPdlk-|%UNR990r}q%Ai6CqSPLwHd;i0(zP*Ee%HUays
zbD4Vham-HyY-_EMNF2F|_cY{B-ZH8?J5)sH-;Ki;n}h{nwQ}uk{5H;>W3SGmuVT@$
z(!?V_Ps%CkP%LWbQgcD-C+)zq*24_6V%fmu?5L%7#kirE(PC+-{Vm03yKyW{9ZfBi
zDG3D^WBKNJnq~V$G>Wq|g^(}^X`79zzA!L<aBJ|J(lZ7p4bK>gi_%Fu+StmY$4tQR
zatk9<J1wwwUJ0Ou24;hPT0$=5rM8dvse;86guiI<oGFmEq|L(xLto^aIrdGUvfs@W
z9*a129{99}9XT<<UtUx>vkX60HDtyyKATVdg(kl}<zYkJ7#CeHDSf0iWY!5%*kd=o
zMS$mB$hPSeGOLqATcVPuOr{PIgM0?QA2#a!y42gZuItwESqry_=y6iz`Azg@RU~_3
zf9ZB`WMQaZj&u3SYWapoRiFMTf#B1&IzeM^E6&q{M=U+ws#CG07m?4nU2-O%@mD?*
zxnGz)+)J#-h~VyKS7KL!BFW^CX2=_7lI%MikVanMxHv?mLr(G%lZlb<Y&w6@TgaP>
z!&~uMq}0dD^0u#$Mo(o{JXgHjNHw2;7IUUZ!ZnW@Orl~U$Q8BAXg_Tg?*Ld^!NaV;
z$Y6My6fiPS3S*yCtZ547rrfVBJeogL5o^@yU5nS)ZW9uC)?t1m$OPaW>1_vEOIJ*R
zWsPczAsKcO)_@M2s01hJPqyXl=b@z5(@+t}J9Bo5gz;P@>X8X{q8KQ?SE*vqkp*<}
z`1$9M`0)WFjr~?x(dxr|wvZp!&Q{B=D%qWR6NSB;6>4+5DtS+dc+UFbrR0BzqCMdO
zN)O4kN4)5gqJ%k}t%AM%sxTWZyuH>{j6hUf{zINrG*pW0(h-Ur2P2{r^$`=PP2@0h
zd&=}#z;cvPhfQR;M5|m${EzCA;|KQg?kWWyDai1#s*j>+;y8^^qML%A{vFvw5xN<8
z`y?)v(u2)#c)5&0X`w5Ykxm-e$?c;;s4TY;e{sAW&;pMS#^vdq--D^XI6a>@b2E~v
z->+w#FP)<|^*mOyu8@xus;98~mZDeRsg{PKYYdh8RMbr@b-<ae@}{tZTqBb&wl;={
zk*4wm+a=>7&t@nydUkIo`|~*LCDU)TnG$wb#cRt!n&vZ3s0vwTiYnP5RPqQ97hxP+
zizXK2w%n!wRlF@8E6t)+O}a5z{T9YseypXz73R&f^fdYR^Bm7#C1#B+@hGHA+P`|Y
zNMT*M>52bshBBRb`#hb%SKa1tz*(zL{a8FX%xeG$5N$8$oW-q6zW8J|ywi!33vQqs
z&2Fw@a4=V)x1}21h|FkLjOKz0p3`)0X4P9q;U6WMWzy8nvS@u7AU+BF@V$O;$ytT1
zsHK<o@wnc*5+=;-p2XKlD(S7z5!{3%Mko9noDAt)3We2{6zSaFfFmh6p%?*ci7%~w
zTwCJH$FtUC+6)!WY*y3#tHkf`w2wV(23nYO$?TXoC(h6(whmd3XryPNji*LK8$I2=
z0BT3C!szi_W^)uZV(C;9Z$+im^)p;YDLHwZ#`_WN8+FiUzC{<g5e3eE?Q&cie=+~H
zFo+X(oaLKL-5&bd9;5<^le0Fq%Tak{Z0lGf=eto5E0MQlj`Apd7$^QKW3ZYPMjExM
zK=q6L!v$wy24YzU`Lg!KroDn<?aJN9GvZkasdF(NrQPYXC+^#z`jD`0%ws27G{;Aa
z{3?0WN!bGdISF5n$mVA)VV*w>!CbGWovKU(pFc9$IrDi~M2yr?X#9p^l0UvcK4Fb`
zu*kvvv=x`2_UXnGEfbXp$<%gl<Tf_+G_PT6C8@(UN9obvW-dB#KX`dyb-jBoJ0rhw
zb%v8~*Sq*!UaflDDL(*8!PXTO4T6r*f*?^+*<Zj3vI%~hEHv4PZ(c@qWe>;DFua&e
ziO|GcG>o>OHLEn#;&!q@+yYmKwl^`3uR7oqZ(UYor~<AFb`7ON3D3-}7%@7ml{GVg
zDNk~SEU+~-aJ$#UzuLu9BVN_Q#=cgOJ`A-LQB*~%k;)#Y1fbERDVW=SO(K<2w^fvj
zM;#JBYwIXLT>H0xUw_9G@O7_&J1lh*pBZ(m{I#C*?-TiFu=(BZJXSpg&G%2D0X)ws
z=Q-d@2LGcY>d0dA5?SZk?HBY^^q!zcGjCSSFNcn^qd@KNztTqS3-UL+(Olm9@ompS
z2X@CgWgMzT3jaW~!-?qrh8O!!g+rv9(QrX>&^&&SpF-;9nbu^i1rLf2cMr0iXp(;y
z*FOV=`!L?Wiud6l0>Xn)vQ+prlH-n`SGB_#1xp)X+=K5Odi}{h-tSV}lm2bB^uenw
z$CeK6Hg)gBgS=L?_Q(W_;Vh8kI{L#4I33QP$2?>W+WrH*v}RV9Yjpx5|3cb&Of?dW
zly&s(HJtkQ@%Y%&)63_nD>Pdt<>_mLe_J<#)2tAS2$Y`$z%CI7BFv64;%#T{*+xFK
zUxo^Q;l4LqdWsMQ@P4&NfoF99G=HSy-vM8W)3#qi1pC8d-3j&aI2><L0w&;7O6~y@
zI1U`Jo?{nhY94NX_PYen(wtA7ozA={wOwtN>mnxqa%!S03yisb6YkaZ9kWQwY5nNo
zZH?`v@9!kPesSD4#5`-h1Y1C|XG2hNpZ!6s`xArU?@0775EX0{ERIiYSJ-PRtWM^W
z2c;v#2`~9hwy^V@(_Qu&4;0_#;1DY}{{Jdm!2_xtR;?kWphvK2V@lgaJ*i0BtqZ=(
zn;(=#9zPLO>IFoI;=xH`fJpr7QOpJzn7kx>|IO!xvY4l`XIKN@sfz_{lpPY!$(4So
z>BDb3hYFCquzj_nb>V#NCJv9Fo<KV`m4f*bVe-L5gu2B*6D&U$XUeX!U)yF5&8wVf
zYXO0}vnK@+Tfx~F6_@CGEg-c`_{a090!4Bozy59svlnH<h2_M<B+;^5lQv#b^G@z$
z%cgU78lJdcdn&zlug@RojpkNKUX?S?dak&e&2Ex_CgbP5jvo3Y<42yrM9$<o0Guj{
ziq-K1586VRQuJ0!r<Ba6o*8R$!%cF*ujfIu!C}qjT7-r*XB@tlk1%^<$k{r)@O;%8
zqu7@N;2W~cb)W(r8+sVmg%@HF`NLx|%Mab%E?>Ne)j~ew6v|fF;V+NTwe~2kr#TTh
zt}&HWj+R2Sl>p-hI#iKgEiDz5K7EDOyJPED9GT3BTa0mz4t#X$^)sG_3nkE&^T|{N
zg`Y8Z^OPulR{fr+L#8&Gax9<{_&)zl5p~$8q~L68zNl0R(kl|MxVYTB*5!KBf=(kj
zxp4A&uJD%1j7f`}d-W!qYptD78*=nm(8f1sjT!Fw9;(4Z9T~Rbfv4ZR+NVS<JxV03
zdHnE$Q<a}w<cVojjDphB?`H7N*NeL@!%Z}laShR%+eKqGrL-5rMqF?Vt#LV*j~_M4
zZ8y|Q+Z9E4$oslun;O3NNJ=V9I9&E2uLN=4)ab4oP1?Mv0#l5hy)ju0c$$Yv#X?Zt
z|2d1oLfSx-L1HN=wrC4ysL*7r)}1i<6JWr4H<8<B*(<mYwX$db+TrRsZ#(Flyzl$C
z4ItN+R;1CY-IJ;zs9T4zaRQ8IlWLZIY%jLV1}aWH^1Vmx^R4^T{@!%cd)+Cm;q${v
zZCB!Nt-ux24R&^pj__!z1*6&)qqI0Pv(!~q``;}D`NLjkbygOxNfVYwW!QoEXAFz2
zTYWeSkrrA7e7C$#iT3J!>7z=60xf;rIbDiL;oH`k!&&qZ%6U?(Y>Q_|m68V|G?NBX
zaY!3lHAIxmOV-s<ENsr`PCC`#?QI{NkwipiDZXgYQk%e)Sh8%XJnhcAvqk1d<62=N
zPCbI54z+|boz_Z6oU#pOgzC+8HmC~*q`+v^`h-+oB`>Eff6QkmSIWWAF}eZxtdBbH
zNwFV~9|a4KM|KwBM=Xkjo5<Bq{AUeNW8=Wb8)_w1Q*r1MDrK@o!b;(@j180trxaIy
zE(}Vk!o&PPugUy3;e}ZgTAFi|><`j97A|%ICzMx~x@>M4n>@>P0`~kO-D*rV?2%?w
zg&JX~!jD=E6+^`huZ9#IDLEkPO{38x6?9Gs^8Ff!PXgdG=XjScYZ5`fc#{@sA$<Lt
zP~SW!Ve28C+|IC5@3KKW>CEjh-smjAB@Llp!RTdX)UaYqW~fN4xK>S(Cy!yB^@r=M
zJ_}8kEVLX;?NLVYK5I}~lYIVQe)&a(h8*xo46!|>d``dFMbxsubW9+jS5{?td}L;4
zdn7j31GMUvh3%cfw%V}i5uV4~=v$|cF}uE;J=JvyARpm-o1|Gp%r!<^0BW^beR)cM
z;v5v6{;W18Gl*}i<KvJ0hwDZ<B%_27m12#RHYvLltBakbMjD+eCsIZ$Gj&?=+EozX
zp@(tip|zKgG?o$-a4SaQq2&$<e{nHnEK_aL%2X3Xlk_@dh7nL=uCRGRM9N4dMO{UP
z!Jg1+8VQq@tmYbaTFZE~#Ge0Sd)C8k=TU|+7Rz(~gl`Nrwr|N^);hmf{5<mxZJ`2U
zBNu_6>e$sgs#hpKveNQ(cv&`lq^~d-kbLC$iX%R<RYkVSs&=(%-RJ3hf;*E8Lk5E8
zuG8Ep+gtPPR@$&~04J6Bqz2(>irVBrgl>tIHx_fJ3A(runSWi2sN89Au>79%0p&z}
z%@+)~8Q`reX;Nfq4uySb!|SfZDW9qVX8gYC(^K&lKb1Y_#W%Gp-ASBRUCKk6jIh$=
zXoaua4V>bwFXz=69%^R3c4`p9O73ipPq6=JmL<JrR%?gQfUK%+>Q#$SqwXqnO^w=l
zSFWjdF0UxPHzTScu%R+h>-`46?`n%QD+^wDsB_@JIH_*EQ2AhTw1BULU$aB7rNT6a
z;hi4D(_iLmIkRD~qPDYaY~h5K+i9W2t5@$@qPFtNe7d(wJHfLZkytcguB4yh{YxQg
zPM^Oz7%zFnf<&^ds@w<jPPg8N40d5r3Y_WC(r}o-L9y0G!7AyAPK=c7V@Lfr+|clz
zi^GQ`HYpE}jhto$orpkBb<{kU`|~CY)YPhilr=yQK^LlFvZDSR5yMbl^*};D>IzCJ
zTY`=Or|5xRvf02!ES>Hil#S9AqdS#bF3Q0rc<{VH@40u{nie#XO=5X7XlYHY<s5ZD
z4hC8?MLAkShdys@r)omqmL0y99fpccXAqJ4cweBI!h3F<SgUDe4X=lv<-?|M-i>NF
zu>~D!NMU=oHLN-)B!uZtch}i#$FJ1ayRBfkj+Rg4uP?C2^J2Tg0<}c^WF5P~?W9>e
z`R<A~+_rpYy?BW#(Z<@xmRh0$LhY!W2{z64C@S3;p!q_rxAGdn=}J4GwDJaZx+h(d
zBzZpC^_Ay<k$|2!<L&c#VJ33AiW`!?m<thFt4nu{iH0HTAq^EpWlxoMnIsrvA`xvV
ziFf8iEr+Y39>rtf=jM5ovZ_Q>bcwcMr*X9hGAnsaUR5XHYL)&=I_jwjO*wI{%H=t<
z_eHiApI0mAE3IY_o3}M0{nV_;1~Y!7YsR8F!y+J1^^BHJWUox*5)~D=(L}PmIYyPD
zrbkCMHO>viW+3@aJpyNzJ%Qf0ezmJD!a`7o9UV+CflTXczB5XD5OZ((!WsGed7qAw
zsh#Y|iex5ppqSkRj(-}Odmb-1QF?T$VAKQhYJozC>8SIT4z-+;`3L36)9xB<tM#5D
z6u!lH&-B$69xl=1W44R8CI%HnarMTy7T~Y2k=$b!Ih1#)<7!;yCyXID#*#@=i8#{k
z-|+Cb0XNo91els5;GqZNW-AMB^o^sJkvFF$+pjN`pIKIavsyO?f~2BX)6Fv%z1v#9
zq@T@t-;uhoyrOrND(j$oPW+LeNRzeh0f1N&m5nm)U4{W(9pNiO*@i_G3TEWt+9F>2
z`RZ+Zl91)Fbmgmz;pft0OWUTsttwSmc-yOUA2uNZ<bcQ^X+?F>im)%fRt(xxK{8S_
z)MC0Qs``<+LFMzpeqO1eUf0#<s}`&ERXR-9yT?LhA3dI*S*?~mGkf0liO{Tar$9V)
z#G*Vwt^>2dRW=e6rAI1Gw1_q6ls~}o)wyUipE4b|dP`UAC>Tv@!*KpwH((%0rB>b9
zh2n5i2TP*U%$_Xf>|J4F!kYm}H@P{rur7v<8O@>%aM36Ys^WMC?XZ|1I9OVWT1;6R
z$X8E>@g;<Y`Y1x2IzNwKjqXKgr&N8Y=pKL*iQJ+2b2vX=>9ysuQHo!UTlTc`odN5b
z@w259d}i2|OJfh3TfKcQurqHa+E6LcPQKQgQ`#Q4l@PqVS}QW{c6vm--*k3GaqzI7
zlr*9@4Hc-E%4A2Kw7LXh%6pr$6Jz4Ol@))io}4q0iw?IE@afDOc%=0HI02_L8aP<N
za<yUrH_~oEI~^R=mhe>cltgzD+Et$%DhpreNs5TMYS(6XRZNV@_OK!Z<P^R;;Ti21
zy0#Q5glABm@Pk=wf`-x2{M&d^Fmw71U_z;elon<jr)^8yJMwuTFsFHH@+us)8>KOs
z@1T*5CAs!M*qbQQN2xP3@C9HZyf4YAMZY6@VZ+iD97}VPG6;%|%#u964BoVoQZO~s
zG^-D01b*XSA4K)KCamPzS?qmcNf8_xZ(g~NdV%fdJcqpaUdf`$({XxI#WWF81)vp>
zSO{^TRQ!BiuOa?=RZd=dkx@F>d{jVH5LPP0obo8g5Y2mnwo=4y;x0v()~o)>-e<|W
zpcCfr<bV{Mo;xhEj;Y7&KD1m^t)?7=g4&JvNi6_e4aN=jC5<i5fr4hY!$waTCna|0
z+lR3h?{de-(4IMN0h*+;hZvp`&TpiRFtC2j8Cj|tX0Vuvz<l*G!%Z5rZB!4BFdgJ(
z`}cJ0j~~jtLCUl2WN(Ul9BlTHkG_E8*Bldltt>Wp|BiHxya`7sM>)&E_-z}2$H3o2
z5j0`pIrtPFL=F!Qto#BVo$nhb`ue>c!T-b&EfvR(O%hu2U#7c%QYG=Gn6NgsE*C%j
zbLY=FiR1VXeZKiVF5t7YtCQnk$CIY}nJK1z@o$m`oumtYh~XA4yyP}78b{&yC;z(l
z*A*S`Ev%EJg2xJ>2lx=Rst@sRsP(u1!sPgGsUCHbv;Mc6e{KM4-7Zj|3suMnbih)o
z5pDD;m<UUaD!;a?lkI)9@W)%L9XdWQY+!L%Utj!5^!p5VDQW$vcu7xWk^N2fbnztE
zL+6rGQ~2J<1pzru|Lk~Y_b8}dsV4$+;Ys-`=02~68YI9F!ROj-(--fQBPv;hTlLVt
z|1;13bN)RDhfWaOMx3WUo(M?NvG)E5_rkf@Ieq|a*_tTp>-ef0VpBy<QhATYqe{8B
zE~Ety1U<h>I9+o=JDW1SI)h3Hcyc-dEE-Y)ZGaoj50Il|TU9CDe9?9IEOONHEumHS
zHZ*wq=1IKCjdR{<Nt%4u&!BP(nWqyG!nD3wh2Z|{PwJ!%#Mo><5zHm%jm`;dACXz^
zkXIK=_t#pL2HH2*s>LeCYZZv+izgwiRMJ(@f@NK+UN3oypKg^p#-+PpBzsgd&s9Ij
zxM2GvgwG}WQUi#t&Mg6Qpm;yE<&|jP6hvTe-4pCMfiGj6yxQWd0Ju)ufZI$^zRi`7
z$qgFI&U7@4!f0^2GJN8^er_V~7hCQ^#7ad_(!=8V>`mPPVpVitg$K?_@h1MwW6X&c
zS}0s)XkHqnw^HL<9FyT((@Xs1X`+dmXf(O(QwN=-BcBuxI`!pqxgB*Txh*suO>Jc3
zoux`y**9|diKwKV=AULvJ$JCvt{|kHev<H+Y>rMx{h6s$ZWpcc{$4}n?$r8HT}N8P
zlg?d#n7_ZX*NC-le0@tDSgg~@!@9~!r1KUYoyr+^gIP48=TxSS85~1z!Vr&HMrb9J
z<E}ks-tA}$p;a&Hn{f6{thspw0@VUWdZ2<iep9jKEy9FmUH#A0Ts0HI>NP9dXwy7;
zvi0tmR^JmamgPd<Lh+<CM4<-$87AFXth-K|{apbSa&&s)N|`HP*LqArR&}a??wa#+
z7rHc-Woz;`RqENWCNtyqxpG!{zC<qb=jF3RkW=+IIGS{lN5>OWPIm}Z_JgCmlS+C=
zjnS3*(TZ;xc=k}vVf|II#XZ2tcn*rpUTL<bb#A=o*_xc;WPXZ$PHTHM{!%iqgi5%^
zI?ag6(3z_HqxGUQlafWeL+!C=A6LE)u&CW|gPSzcOGm|3IRSe*G*E>*ceG2&x&Amp
z;td%mNf)Q$6pM~-O+MpFTyMWNH%9Sr8Kz<4ar1g_8%j0KyEa7?DgiL{hMko0luZdE
zkREo&9<9_RsyvhxD_B1O1g4q2oW)a|=s)XH+C5~f(5w(zA5Ar)d(XLz_qDpKY@LsX
zK|v38u0aaAZnanoMr6s8dpQF7LE_q=@wJ%+WXL`Xcoq|f-GB^pLsM2((XNy&NSe2&
z%yy&PSsq?V4k+{RJa}WAjkt_2y?HHMm{8v%S~R=Kk9%;JX19#TVBPi6QN3j2#5iv<
zb%(Jbjxv2FWQN;NA!DR-eryVpc!*fTz(P}q*clo}tZ6gF2r;sn^O38XTCVO&reDoB
z?vsC1g7m!?Q|VZ{Oc$5Uy`_4$w|@5X+R{$H3e-gxLoxDQjz;Va{}1A$E9k{e#{?do
z)vPm9`Ia96spVGrWtU}l`fAjC4WP_^PhPIsd8dl1rjEUmcgO)RroIa17SSc4&&s!6
zVK&~NvDNA+-)Bravb(iYr+5};I|!*vR#U5G;+PDL2S&)5IVMiWnXpNIyMcUCcDdq$
zd1ucylHL(l8a8S^o}HK|=z~d?pRH&Gg~lNPP^$@$aB=?Gz9Mll51)LA+dtAxrE;U0
z*gZtKWYOOCd97%T?M0W$cxt38hLR;Xe?$OxZXU1^*dI_4aHW4qJlCo#+{r@Pga!90
zYqpX<9?_Ppb3FBM7aIV~{6q+aI|udR;`$SALQ|A3+dn!npT*x8de5~ZtUQ}t-=w<u
zj_24k=3aggB#&}(^+9GwHD`^rr!d(=0I?ms8KqE^VD44#N}#9dJx9#G_b6QXGrvry
zJ9)nDSDo4!bRG6Gcu?ZHdGbz#s4qojg8ZG0>`S!stoE9bJxXVLBV##a!`SZE>97Fz
zIW#Onm(hIOYxeCt&0>;#t*3z6l`R^EPo^n(V@ryw!g3$F8ndV9vgND-h`B)lT{2Ya
zS$rFnrgbAP8(79~_AB15ScGG<i4L@Xdz1U#D9ue@hgyqZcgkLcfuMpi&kU3?%EWm&
zqJ1bnsu2+XnR-sfQZ8FE8vATGj;gV8^poJZ?m4kHtmd>qo>UkQ8d6Aso6tx(SD`_P
zBgLqqcnBS3p!U0kNI$8KF0NA43e8|2`2sgdAlLM*r}^P&&|A0*Y<XF~*f17KP^cqH
zC6P5%J^@i!bPS>GSFtah@(8w3`ub`v`dz^Qe8Nfd7BVjO$*&@6LS=to#XG1DhwI(j
zsWRV++C3_{Y%TTJb}=Q$^CZ=*6nU8nGuxi*%*E9vwpN<WFL^8gMaD4zk#)W@rCqiH
z*(=&=NS+1l%>^aq$^7v7t?V4b@YborWurnO&6w&0?YT+9`E=XMyuDk`tDguImu)~P
zt0|horOH}Yq<LAdJ71#Ns*ZvWoAr&pCk|L-)xJ`5cV3P2^hihA`|OLWEZnRp!wyw)
z{7o_P#7CNzN8(%G3u4|*<ixPdujC(~h|cnS_@kzn0y%s<!vxT2)}za#88dSs9eEZ#
zq$_zhdevEqN1Ia5gO62?HGReZ(8C8F2RqBFSI7h5+Vj80FZWNV2XxX|UIc;X^5_2c
z1m}To2b?;4QJQqPVBC2B2UpvFW%=({INIwut}N6|FJAxk#8dy^_Z6Z5KVO~uwVQi`
zfk}rC#thfh6+CKla?)F9q6p6Q@ErGl^&0;7Z2uSkC!q(D&AQrOp5kmLYXOqyvp)P6
zkNFQ&BR!*DAM(6+H(UQkQG}!P(V~L85ME6L`_h!Yehq&ib&ouqP4w5Gs{O_J;{Sm3
z=JU>e?ti9SNN$^HE^e}Naka;LukVMA|2(Mc27m`TQr@r0$@sRsj8cQapi*niU*^{G
z7;Xo<&-g6|K0N9@S>3#7=At>f^hBipWn)Xv^3Qz!5Bi76*QBL;COr3n4mNr9sVLg@
z(EvEv%gcbauj-%*zJmqV<TJag2jRN5@KIoKD10rGk_Ora`iWqJfI2YJvVkIYq<P!?
zd)Ez#gZak^0yQFW)})Rvf~y&570Es(uUtc`MLizco_9!%6Lz<eY-E_RFDQ%HX53$%
zQ8CD7sku{DpLHS&j~2>V-&Cb4F=Lpaa@Tq8q^kb)!N^`cMswX#GDKz8|ErhX2zX0J
zi?1(DB=@aS=e4R_4Xc*CWOCMRiSq#fzw-NgC0wp2mZ@zCmow-1@~m4C+FWxY$~y!W
zU!54xz8pu9!2wxRq#3xb=A1fkj}(Ck;~8@o5(uk{;MTCvg*;^iznB8+;`qE4EyAaH
zyb<kF98dy{m<(z9Q<-t6mod_Is}v?rN2lkVS_yf|-aVLEv1}-&;HOP9Unb*H+?u0f
zq+yrSI8CxGli5_L!<{xTk6eEF!|A1@ta_-oKnGRjtayzj!<tq|+`M(|lz~lCpm0_1
z_dS~%0s*^z5U>6n(~|X|#+#EH;ZI0$GTBuhFxvBZALl1)k15A7J2IUK@`+oQDS(%w
z)=4>^(Afn{2LwE~iej8%Zg9-v68gUTE&g@S9n(C|vED)-l$fZ5?WDiKjy7}TG~d@r
zWaC}>=!Ar}_GN)aF{HoQ-0{k&dW;MG7{jKz`<p^;R~P$i9rNB#1jvJWznC0f$xvWF
zqG7}F#f{CT-Y*mX_@gQx*c$4Ewpa>Z`RSf7BzyJbLJK#(Zq(1mie=%et}%QW&F1wx
z^T<({w~NZ*iOC@f^i1U%X$oS)QTDv2QMB_`Hl>x4g{CRNmAU^mnGs#uK3ik2J)o#>
zQ=WjF0}kVN&^XWXPEBd7Y4S&o#XOfpQpXR55CPSZ4GpDptR1tHjH6NNxSOy$idPkJ
zHD2DrLa&RhXR^BE-wFeO>f{hCwKyI00h>PN`od*BP41#XSH;`}A4r+-Yc`vC?{zOk
zsRpN`bH?C{ahrDKgQc3wy2X=fs?6GmB*x&`J0>~d@MM|`T8|^$@LUE?ZMvTb7W8c^
z`7(FRD4IL+9Gad&`~uA)v14Ph>6cr>@BxOu%zl;s)hWuu*-@Xf($I@(^&7__5mM5T
zYjyD(3)V(HQ!8Xg7x`T)0v+g~qH(grIse+T)wu?DU#Ys>h&%9*Fp=yf!|I{`*-E3#
z3Ca(qj5FnF?u3k2aI@xAyR$WV<}7YHEi@^+t=yhV{imUo<4r!hhQ|jeE3{45JdQt@
zH;-g*ZSUTR$#vPh)^ZnASjmiO<=E_4kap5ICp&SBOMRy4Lr#5!65+u3C_67)<F;pO
zarIWxuXa6Zb-TT5teI@3Ux=?*_x5S;!0;4#Dsv4b<*g9IY}FQuRKnY$f=h5Zd{Yx^
z%)+o?+vn3eJUn(c`wgEe^1BhEttaO<cuGGJn8qy8`&x4=Jr#!{og8Pdbg8_fFU4gG
zw|fR$kmWv{kcfTXp=(FejL}m2agF|TQ{KA<>9BI6irOhEx;8V>NfJjgXxQDBBSqi;
zO6%ju8kd;mV1)X)Q$f_~*t1$-oMZ8vuWi>4&>?*)i7w7)M8r0-Mu;X9k3TQ4n%}=#
zN=h6Hy?C#?<?4<5o>C#Yd8Ex7V5XGtsF?y4^{EuL9o34)M=TGg;E-!^wjBLxL+_*(
zme<{dXX%#!cvw5}VxAnLw7Wo3-HZN|>*$MS9=ORKkk>e<kv;_dEm2-)Mu+f?<1Kjs
zTAnv(?=szYI0Gd79-lnfJ+xawhu-ITe7eq7J;3O2GKB-B%%D~)bwp(~uU;qJSKn#9
z3lD#Gs}iNmd<h`Oaiyx0Fh+gZv}+fdf^pp_Y}pKq6S2=$Xd>!ixg*i9<2ZmFT_VZ3
z*r+i2uS5DS8CD4Ps~n&Q$c|MX?X%wTht9gew?2eL*cWQ{#gtDgngs<{#M2Z3$NT3R
zKJZQ6H6)zNS?l-}u=8yXcn%-=*?Drk>DgY&R{2ka!G}fPZsIo3hHn>JG~^KK9P{NT
zQf?O~Y4$5vT4+vtph2~Y*2C-4X?L6nJ|qZbt}MR1La1onaw8oY+620%K;yvKyVrOs
zic?mwDi--vvR96GV+uY*GPejU^HrDHW!nRTc&=?j_iAtQ>I9LcOGyZ=YP}Pe_*?|s
zxTfH=a}G4Y5)Rs&@6<g1iI9+sDd1a6X0#T9LMX@t<ZRdrYa`+lV?Zut!J&KwJUa`}
z@{sgjO^F-O6U-`4(6mBh=cK3{&Hy-a(jZ5(PlN{;x|<{9!aLz-P=I?Q1)vgk_rk12
zr_k|I^b+O;J$mqFalR)=`xOj4^N}t)X~o@ndyZ1)CYI&L%ty(=U!6CdoWPCf`GF?F
zUzjp~&`|{IKTiCPYDT9cKGg|;gFeGor2YXZ@^~cv%mIA|$d5oddUCcxSr~O|_27O%
zOcRd0A_50t2@h-;r#Y%M+4vi}1W~YIZstM2%+)V6UU<;M2_TX-i^X3d*QCIKsX<jT
z@L|Zk<;>7LYy;;KHV}se2PJY?5Gmt)P|u6FjrIDz+$+StP6!CfhGOD*hwKybak^iO
z@KiuuQ2^*sRWI)6d#c|DlH;pmH{@En2-jb|_=cgz>GL>oO&j>QC_ms^9nY55$=SlI
zq_e~Bp3+<6`}mIXiSlr070;{JaGw@<^uOcE+f<8xc}D%o_kVOok}Y4*3xzLh{!N6q
zin#zELu|Njj?)ZJ%)o6|U@e_qA6tY@{`jx`P5S@B#ovZ8NEwvNbscfe$;s|l0obYl
z%gW;>aF4rhgthndWdJ--cfE+B&}2Q_VE4oSxnnzY627;bJHTXZbP@bZ&2TSRvlCst
zF1Zh4>&lpJ;LB#ylZlVpYlHS#y9t`@3x7TVcs0O1*8*WJ@-c{iS4k39@h!r`z)5|_
zCHSxRhrjI4=-%uIzeY?uBTxN~0jGF~VCogP!TJDxSL6c`x&GA?gu!d9LB0DVpAa~N
zT27w-uM+=G@c%fQ|Gff!K;G~QKrl7w43kN>Qod*FwKmu6%U|r^FZ|*ac_V#GNPzd(
z()GVHk4Peur(5&tA_!r55;~K=hQM>ibk2DB0Y}29Gxj?A(ob1@8ZjXvp6Q(K7>Sbh
zJC@SOLVh4tg=#M<xQVlxD#D=29{Tm&U#Tz!LvFhp^&b4bi?Dpoc>fo9%d;Gmu${ft
zs1Im@qipkRW`~Sudyna|R`y??Awuf#dgLpFbRsfk=n8M#50K%AFMBG-XphL+yjOKu
z5j#qBYgBs?e&0nP@xZ%|_5A+VLnPTz&o13+;jdz1av#WYMxit0y^J&;|4H<pb9lVm
zTCPgwgQCl-X>c(4NOw9P*f-|+!^EoQ@Il!E6I2Ind22CM-C@ALbn5hPM&(VKRWW@~
zw7f*&Pptc26UJ|*D~oAsLWI9y13U`HTz39fivOQW_MaK2((%M=Noj^l;`5SFmiMAA
zD-x9bUihex;Dqqsf+vtIgBAZ46yn?eC_cKfh5Y(+m5xyohdm&|U&;QiuYah(pUMn&
zy4Q$Lri`y95>%kpTa5sY)$d|5mRi-C>s8izpGJQo7!*M#^K;90kY(Jj3Y&jR|HsF_
zD$d8?ABv^Dsc;#4j9N8QdGJ7K`9P)P>eI!!D2&(Th8cCaNBq9Z>%ad~8UK{zk6Qhw
zVj{=6D|^8pmml2#H6OgX6nqRaE|F#SH80dKq2-{4TBfDJmsPMhbkAlI_$%NFMb9`<
zf;M;w-ilaRdu23NlfF_ZhD>g+tUH-|C!%;geuw`ce^;dlF073e=6hqukC5^aakE?f
zn;+&``<C}yc3j1J{<!&Y*Y;aMuVe2Ek0jfB0smYWFa5+SB~gEW0{XB2VhDKO%CLle
zz`*NyGty~b7+<7W^vL;}(-{D1(`QM>wx<_&OFNUED9+6N^|PXS-_6!%_br!94(>7X
zhL63`=86BvWu(LE3hI5M1#jt+O@Dpq=x~BxO`88mh}*~eByF|KZPmZpH+bS-k3v0m
zew~Bisq}ue1GRSYtiM{kD)Gsx|Ne~QpE>^6Y!ev#)r1lN8P@*!+EB;kD!o5%gKsh3
zH>27-ISId-(zaT=e`d~RooEyH^cPVaJgX1)g<-S)=4|y(GFCVXuX$1wl9QUFY$`5)
zvGvbF{7QMxJ94CR=D@J-i;#Mf;FTs;iNJF@c#bK(e~Ab8ua@7(-uu%6smrgN{`3Y#
z{~Im--%WM&oRFa>f8$4NY%Uw*-!iqV)%w2?(%*ISADCeM?{(pSZu3hXzk4a>Z?FH=
zPod}hOM3sTi2HY>{;G&~f8_g{QFxvGzM%iNYD)c8?7vmzw;F~1?n6J}{9WgOyVTD@
zdQI9e5x)CQ^IjwWt(t#v^(^hSD<x-`Kz>bZAnps*eJ}f~NW?jC#m6wb*Vj$BcK+D3
zm=zeYLEwo~0N3+6pFWgnz@OS(tjBEB(QDN7{Ke`YE&KRi82_K{paZlzJAV%i_`DJ@
zQVE&6w510#jmy}d8-2uMy28`RG8Stk8s5a5`C`h_{8>F(hK0b5ojAKmDW`%(9J@R(
z@)STKl{iRhmbu$K^-Mc&#MCh<)y*~qRG>_unD<V-z6linF1O2?nkQ^uRN7q~N5?G2
zhJj&D-8>y2YdoKa$2;Hi!br=M9IyA`ds`O9Y7517fp!D`tWOu(Z*vtHRQTu4a$}kO
zF|%|MX1O9BPiclTwKG0TvKBBN_GhqACB=!Q5ACw*C4B&zy^Y{+H7e%@c<XwH2<>l>
z5k27)mQJ6q9baFzHS|F(@p)2mj8XVfzd3(AIH+3RTS-N&%*Mi6H0xavpQ4eZuft{(
z2qx{#&=X7(Q&%meUY_-0^pkXiWR=YH%*T@$uM)RADqMmO@IloO$>=U`Cs5Lc-v{fJ
zZhWsmGy3>eC2<2_+xdWRQywziKW6U8wXM+5ouM3mC_a`u#zQmA7wB$keV0>Hn}v8L
zTf8>G%sbE$%u^TJt5M08f|0Iu%FiV?Z<ZvsX*r8)>R9Wq7zTA*P|CcRp;v)u1v=_Y
z&m{D3hk!1rG20LXoP?lK`6MU9keIFbVF)_|o@h?rXO7JZ(;%}h<(gH(2~1K6c5;1t
z3dxO?(!d0z&(z;UzLGlj0UhicqL|1&1@jiQTgW3@Y9?<-ydPxNtlu({&zp_P!*f4z
zC-Je*8>73in-)YW5fr4Nw@CePSA=tplPcsBi;+o4G61mNuRMy2M}I_;55{0j)U0Qy
zv>WdiM~YSj-=JA;Hf{|w<Zc+thFxPgGC6tdGPyATo_yb1fDb|WCgNW%ugcSbM?I=I
z-28tCJW7|h){HS)QaKc@gt$<qM&Ypw=GgRs6Mc6R37NTa^3KNC;Y(*4%62D$ub&4(
zToqv$o3Wrx|8F&8xJ08fAz3WUH7T*uL7B-3)t3t}`EHTBwrMT%PtwdX2!>1ybt=xz
zCSXT*T9AX&r;8kC=?F?KEg|bF#{M=!*0j`_c;2v=t*H3flnoDwjhdJ(oPVUZNu|lx
zP63F&8=@i})_*s~ddObl=VB~i{OqZztEwM9OSRB7v?sa)EmZV;z(Jv4`cZh25H-AC
z${Mg8G99|2FZzkVIr@B)n`qt+!ATeZ(>Sv;zQ%pAP^nR75x~<F;>0IsP>0l05x&G%
zo*<6ICR@;NGs}pJ+JGU%5IwLjDSa+9#3Y391dhdu`EshnY%pt-P9SJ7RXmc*FpzV`
z;3Wur6Bt)3QP`d!I$nr6AeQWB@*OcaU)~0@Q*&*zvLI1L=sBBqEfnpRXuQ(&%b4@U
zs=-1vBU>Jaxv~y;(hN$NtDMyy)FHWSbpoaHG6h0ju>aRZ`NIy?%vg1f2N0!@+FFNL
zM+KmzK|XWTD0q=yS&l#JuuUoP>oFTig6ua#DW^)@O}_Hf@2B5nT0@B+vF=^AeFN6K
zP@i-N04Ftj>AD!H=A5%nms^=@LB?9Nb6v~Wsa<0kV=<Md1Rr4W50*B5&dL}2`uw!H
zu)GfN>Tw9ZG5SKIV{zQm5}Bu4t0JWZwM0cePDMnu8Is_yPd)8tg9);Eh71L-$^#;6
zha2&0u|NI`4&?7y&i|(`@UcQex2|=A2&k9P%4a8_`8pbem)CslEQMCE-O7z*5*MV8
z#c&P~Y27z-6tGc)E#}=J_j?=rTbP>P5u~umYLmYqt0+&#^}+*?Pnv8lr|Q}O@)m9N
z9n6evj~{?<)LZH-9v6ie;e6CBW*n<6;2mEjD_%ZTtGF0c_rif->$-u5@Y>R3E6Xdh
z2m#_n0ld$HRtJor6Iy7&PGSwU7b#{_@$s^{y%nLsW4&61&H2h!`R$2P84bWpr99-`
z#~e9^^{~$)%N=jBZ73Mugj3H*nJ4WcQBn~Az_m6FXP%AM(krEtT{#jNYsYg&3EmVP
z$*U(#eNVtK!#y`s%gjdkp~CXyx_YU_m0+K>@Pi#YPcJko%vzneFE10Bwz)kxH?@{m
z1b-sDMjWEA-daCfcO^H0#XTi;nv>#c%pqBaZWrbErRCltBh<@5_rTsN<lSj6Lr>?|
z%klP@_9D)uFWXRlP*IMkcyH?RAT23o2+=kO7b$qAqMbXkU)#*)2Yww0LNKG+R*jct
zi9r-8k6wnH-Zkj*A^`F}d&oDr-YK!7d<u#dlF(jl7#KNxZe;y<MfSptWxUn3ga9Cu
zq8?Y}qet+gO{_o3<T{HyUmF4M#3(zl`*vt^V-!)4$m2;<scJt$=JbxjW>J%6`b8L`
zRJo@&&Oqm~|LnQ=HiAeAl#}J!p{WJye%D&oZXt*FRN(Nc>f!@4u7mItpV7EBl|M8~
z53OEB5OO!YfAF2)u^#~j$mbS&VDTpP@rJ^!?%9M-0D>d^9rW6wsRe-#*BN?kaA!+X
zcQfk9u&cW_cKo5rCn4OQH2-$G(UyS}jbH0-vh1Q4Q4q~T^}51y;gM+a*xjx!58_}>
zQ@^pv<|v|-LR3-N?QrNS>w(JAB?M_c1p9b(&RqC54S%zQWm!+rwntkg!ucquF-1~m
z6CRFZzEaOM#U;@ru%Bip5i|ojKLs`-Mf>m8`(P73EVYBR|3tu>L4S+!_FQ|KLBs80
zdw2bQEZPJfU{zX@*;U<(yP?2?Mf2~A%bfON;pm>~*6JSki_g?B_fFWP)5Q(K@9CWH
zO&ot6WX`5IBWuVuxncPtU!;EHeu^?2VwAS!etJVSGU|Tmzg<LvXq$1kWb89VNYDXo
zrJRG^0FF_-5M9naX*`!s$6q@>-apB|c>Q>T_%0u=BHuATHf=RILchTU!v9Mks2kYR
z>t}605m5hwj54mu;4mejEWu)b%N%ieaomWGv~1g@_c-hDB_bK){H);h;{M_XdLG&Q
zV9479S>K-sa|i%ZcHX;wbvN2>mfH=a`5J8bGY4TGnVSQo^ehP)oZzm6$n85|?AT9)
zq1WU|7?A;P3%%Z>+W^?!NUxxChHjr6<9d;NwOS(}h@+P^btz%1!fo_w%&S|4Yrev?
zvXC;ft$v1K=sk6sSUT);lcGH`LDNX`%htQTTK+Sk))PiriXkW7sA;S`g_X?{w;kbk
z1-%An<4>I^6?AE&)FI`oI^H#ch$ibPh=+-6_#C$q9Q3T&v=k>2NmQTnHKe{ICPqWw
zsxH;IeJ#n>2a13VzZFW+8c>RVm9#%tUFdH1dH<N~Gl;WTkFDR~`R>TELsT_)xJnth
z=^`H+V?zi5*Q2G_N0_-$Y`)t@kcxd!X+Eo!y9IQ{MV9XYps!FEjc=x~H&oh+nP>}#
zFa%hga@>__`R4Tx4eKF%#x{p{E4J+<!=T4S1`LzK(XR%x=ws^N0+S%JVo}j?iHV7Z
zg!I_yd>R@c0|7@H6EtMCa0`N(ua(l)8UA*pLP;0~1O0<rrKRr@vA-nYm>Wn7&_A(M
zg9c&}HBwEFdP;`9$5ZI{P+(5TF?4yKJ+l3XJKm*%4f??P+U0R!fnufk#YZg*GTUU$
zm^y=(4OcO(5*3FAX||2o3Clx63*Fy}WZ5fh$!%{l8%l2@>>59oL;Y+)f*Hze#uUve
zA=*e14<il>`8dXa#t$>gpF12+@<{_#2ArjcjIDK^an&*@y@FIzNoYuA8SDruKon{9
zMphtEL+g}{B{JcYw$HB{dLCj(zf=nfz6HPWh;vY$IyaS&+NzF@y74JteHiOCbJ1e2
zKD00yLr|iT{}_+2mYL{f+;iog#;t74FdRKLw%M!li=k7?s5p#eGscU7oxCvHfQW`P
zN6oO3of{%!tW?#depQn-hT{rJ_Tlg>Wuxhfm@CI`H#;LB0aL)v>dJBhUgxO+XOV*k
zXC{%6blE41lWsABw|PLRO599*34i8|NE`$7u{<IHHP*DdfEw$I0@*?j2<3KU*(*cf
z&<))HKC%Jk`vcH!1*^tvuFl@vITBmX6ULC7l6Niq!_g87B4unqsDtn_0ri!@8%AwD
zX&7DK6!i?Xlv`6_<7D!LcUOY+na!~A<Dq_@OqG%>XI0FWbfwi|TDHu$@xe4<wx`l|
z6r<*@(uq_ri#JP^BDG4^^{6FU9Pyc!S`E2RCnRclhH0>f-$7y_vaJowiqu<u5?(s-
z?Co3tE_-DD?;jN+@XkNu6DtrtzJlbgidm>A2`8$s^SUo%zO}Q!Ds8dWQT65kLDZ~w
z^KGEGNtxM(?jRZsCkuZQzKc5g$`K(Z>mLwl$`l2CYhpoMc4zbD(x$lN)+>?ke9UKG
zzE{Jf#-!YZ&@;OYHk>s4#uzg+K0c&*`)Gop{KShdRz7dwty%?PjhdVqyMp4g<jC{_
z9$8Nbk8NpX^9CQ}u{>o4mp<h+N~I(gl<224T4&99(jA{oN3c^xtP4=>NOCs{$HhQZ
z>fN8UsCK?g374azIHZ})$LIY(Z|rTp02#rN81u9xxEE)&WkHO}kKVugN)IBVQ)Q-c
z?Om4fqQX7xu+iQGa*XG*WC+a)jd}MZsGWHf>CNITORHEA^evIF@G?;__}W=va!B!_
zWzf<k$D%djXGy)0e21;GVe{?%Qo8vY!?(<kD^;;+!H#MdLg~@%lj@`LNqdKbbctZ)
z>^@FPi4O&SbL|@528{VBQH-L>))S1h#@dx_hOh7rd+6y)*<;P;RNI-rj8iJH3M6gL
zrmdh>RBmFY;AGw)?#~ZD=3`0c<fI?|L?Bb=Ga-z_iIXG6K0_G}4|9Bz6KF6C0MVGb
zD%b)kvocRs@$+>)U4Gl?qdBtmYX2;o-*3@sSpTDvyVm*LDIPgq$XP~1Mb;IH3ywZ@
zj{cNsv=dgJVKzU#gW^&64h6eDWnE=_1OTlQY6P3M;Jf6stZa8~O9IEQd>0tk;ZC+?
z&E79zZFff22Cq^Uhnbh;2i)#*UAAa^W=>lA>Jm(VVuyT+A8jXj82mIbo2t#7dnH|_
z80Jy!6Qo-{U#1h!LG${@G<n-e_ZeOEur_85Hyz!l5+Hd<H+42~+mVtIW>TR(L?Bbr
zGAU0ft)PrHg>)}uGVH^dNXUhE<={AS$8!_heq}ZqE`dxx4U(Bmo)UA%%l3j%sVz%2
z%bhHfYy-*ydhDd4;-R3=p!^}NQ5Vn3Y0K*l&+kU_=!Kr&)Ls0qeH0GU@w`{FS@849
z*@I2F-_OOY=iq*O+fcYAlbtXo8cCk&&#)DIa!dLCoAC<Iym}L6AL~Y_=oP5l^^>gA
z6dwvwPR3LSC%d`xM~zn2We|BV>&Z3^j7cmLWUSX<n7K~r(W>tHe{uEJQEg{$_+Y2g
z(kTP9NN_1$+}&D=yClJ(1P@lU6rHxXyGxM}lHd?r3JLD+P~4#u*KTIMyL*1S|K*%}
zZmz!1dq2->I>eRJ)ZHoxls(tOnf%kt*sH)a_D~z9VXZf%)lg}1{lsduIqSsi;0nVe
zeZfI@?<&iB-3pJRrujN3((4_Bd@y?5=9EI9x<0!_sQG{f`~#E8!#SK8ZBy$Am;Djv
zA;N2|Z%24Sw3geWti{?qKA9dzj4K(qmwWXmu9#T(NUcT9bY>-Q?5-qr7x3UMs9<eA
zMyckp`tvnY$!lMKsMO}SE`Q%G`og#Aj>RC#_-~3a#HgDkLYkX6^3ihqZbS1j<8<ug
z)wug9aEIR}hrEzkGA?>9*)KCXIVaNtK$TlFKV6P>FIS$wSa;ePEVpzfXnETg^3})U
z+!3;Y<@mV&ch?@^2_pF>hj+^M%Pt>DMYFebQW6`TCQWS`tMtmd(puq=O4!Zu&mR-9
z!DsB|<E6_PCuFe<$yl;zE}l33M>G`>mdtecFt}i(n&!2q+QcT9j00igv1P<N+jOYn
z5?<;nbm6%WBA~^QjQLrKO*ZOH^bmPA%)+s&$~%;huU3cOXuAc|W2?!>DkyPH)q#Kt
z!1hunzPZNC=;p>u(&1csb4vfL4oaSLQDG~>M)OM}@J*G~N@%;j`pvp{?kiBUg|>H#
zHv9y0Db+1Wwe;-Zfu+fB6r7cjpFCQ}8js})H7gRq=B|ed;L%<xlomg`%R}E?d{w+0
zUipPAhBw`gG+*!?-qO#%aJ-v;*-2T0U(WWBtkRu$c~<Bb=P1W6HSbry)pm+HO7bvE
zg}11$)E=Cl<ElC9&bPgAQ5w>ix24`0C>l&#?OF*ZTQL3{Ay9Fd7VX{oY_V?&8k2e0
zGRIC*^>{H*J4WDfc_E0AWZv=HI3&X)p>JgTPVD}h=kt6{(NtE?v54I(U%Qu6{L%ih
zeTzKdRiwq=^kG~-_jp5Z0FP_?6hcVcHEHM4)SCpcilMU+gnPq;qCXBSB5f<rGz1u_
z-4b;;xw8#H&pCL9PS#XQ$$(keDTbMS_^c`6XU^?<^+tT>hv78RLZoSf+&$Ey?gu8F
z9gXF*Z|b4m*20iO`X&%QnTcC(%|_O&tN)^op(qkBwMjB*48~=)m8maB;6t${Qrt?g
zi=y6HL*!?_Wew*@*5CofmnkJ;(1Vh5pROZoY*|-j3EiWVq6D|q3tE}Jhfvdyo=GyB
z-DuB@7L#MI%+pLhM-;8~^<?Iy#c<BnR3{;?xYNKgy}PQ2;ws{6XzP&sp#pGl?`Xj+
zooZqb@BN43&Ax7mrOU$zrH<*>1~RHXBetV=T8EnZ-av{cg%DY0lPLqw!Xa4uL`USg
z^o>h&5pA;Uyv)dz*R86bq`igW?EL3q4I|BO2GH~yR!AC;RIr;UV+3{Lfz-Q1m4nOz
z%l^R@a_)_yN!d<rAKe7QDsrO2*~_c$%nwS!U4R?l78P7MJ2|aL)EL6T86BJ`YRPW9
zw#UVWi;+D)SQD5(w92xnLyzf7ja2EsN}u^h0^C-*_+#!eCM*KY9ne&6%tJ?6h1a0R
z?T{W2GTCpY-2HR!tie7eYidtsfTn9>$fYEXY?mQAJhs#8s6LQ!-egQEqjn~=N+}HG
zo^mI^T!_z#=k2Z)Vu<;fKf~&#C6t$$>Y3t_q?{E&oCuI~0%e48UVp3;vJ{jmcpF7T
zk+ow-v_>lsvRx#=U8*W^h2BLp=oAbMXi1DGaGW*pFDKa8vRR_Zh5tG{m}ZgzAU7c*
zLk3nRPO<%q*H(3|OcqqCohD6Jokf<cX$s!8EsmA;#ReMqWs~O+iikBLRvpJ~4Mp6?
zxJEkr9|58XD{l7w2@50d`JjBZW-slTWJ$)C1z@99CRNjnC_yz3JP7g(1k+O0wN&lC
z+!Lil^%XKHrerGoQ{zBQU1fTup<HFQAH)Bi+bcrAkyG3|i{2$SIYYuLiU_Wq<k})1
zCA<@TPz>KJFry)%M%~JQS`+$9N-~`1NE5O(J807vyvd=?S>63zvuRVex+OXZk@lse
zK;slMzR?Yz^RSOI&YM?*W(H$5I;N>-OUVmb{={Q>E<>z4>*}(jASbH=WDDB$FAd95
z0cJiq4rQdyk0x(<#Zh=ZOtwN3rd2h%R+d9N?i>aP0*YZZMjUrC!%Q#@lI9uVta)5|
z)@OGH9T7r&q8lO9oRXx?4)&)nQ)m~*5X7Kn0I@rN4%*dvc#ObqgoqPDZdmDl)bhc$
zMg<efll9qfHFWZ825(B3_Ym(7<!}qie1%5p6XQFzXtUkgEt8K0yG$Ijkv3EJ#1||V
z)fb>K*P2-T4rA@^+DuU)*vQ%}TGD^|S=V{VPfMVi86;oZ11X%%0ZfQaux7I<Kv%*c
zrEIN?3My`tJI-g_$!DL6@E=5xpx45U5t-ILz^(5wL}i*ORylZiJa{7CG=Q%ZZA)!s
zkmbRX?gfoaLqPZK8CxRtV}e(`bx&5OtzUmkfyIb!IK5HacV#DCL8eIb=W<zqGD*||
zGCZf95%fY%K>HdwY=3exdYz7-5bEJ>=(t5urzG|LM*Sjj++1v|RqZJ*7?XMRqb{f5
zWZ#YS^U&V+rDE#2V0~nw@Z7b_Fq~QHX&=4X<1SAzrO*UMg)qRw<ya0IO1^_k66Esd
zCNeQ5Pql<E4$o%LaLQ;m+#rsRT$*D4P1N%~1mLD*!?AP=rEeNgu)tWCwwCKv^G7)4
z(=P_jBVR8?=CrdQGbyp@Mw*f-JSgMOMDKd=4!ev&T@2oEY7eOtV}FB@!WDVK5MgwH
z-G$CY=Y`VPaZPtrrL9~K(AS**+vpa|v6__fZ>lLPbxnmb;8uP$t9(u{ltf>xdhCY`
z1D^z4c&rlF*rA8XObny_g3463ojGD`S4fC#uCc+5m07StJ<V~o4*Wy}_>~%l$1OZ#
zH9ivDWnw-J1kKGbHA>}#%}EZR!(t{E<`=u7ObB*_vu?tt*+H<|47V<Im?)zbx*Qtw
zSXI<EZ^}VvQ{KF(_SaSV)zTqP<j*>ZiRF|k@8>xox}82@l#_eF=Q||&p0h&0*(1pJ
zC?felu$nulT*2gRMsix@%HcbQIe$X!yrNYC8BL_TCw_E}=#%HYT2Vw&`903(^Htxl
zQTR---;W{^=%*GpMHGhVyYCV8-HxZspDfv)EKLpaHo3KvI$nJ;GMBL=Q=D?@N3Vdh
zoF@U~t6e1`J9}xWzFb>Fwq$Bqwyk|tbMab?axQ024`Bzc!NLp$(p*t)_vFfCmP52Z
z_ShH`V+t2t<;cM`n%{c+fi`PeeLv<JRqH!4LjbYqshR1Ysf;T^sF`Ww%u6Bn6uW}3
z$SOMd9OcXyMky1-kgB4yZOYRs*21B#sHcY^<ZvPuVNWp;-2}#}Lyhc7AW-bj0ea9Y
zV+rShB}Dj1g(R49L7XA@;{ICgrf_C6MewxH5Qy(Qrp+Pj_{E0)t$evs4rN}^K#Jza
z!rfQBft;CM7I?K?CWR0)UI%UB)xJ>8)xC#=lP;I9BKYRg*%Fx|2BaW%gvZNGwr>L(
z{nf(JKEzaf0(^Y*f11V@z6~KRxAJohh1nK;*kVY`fnaiy3iM#MtBnWYJGiI)%U6Q}
zkar$)#3q0^XltAwJ)(3*0s!!A?jDM2HsK|(?P@yD`G>;kR;fN>=Fog9+voPnAnk%X
z%nM{Yx|s!_wTRQP-k6||JVDD!k(#V5CJ$&Qj^>>)7FM`0Kq7hLDtYK@iqvRZ&6B=X
zb_t^XxDoI1c=1^;zCfja<gGuJfRy^l$9Mm3bqt=cp)Lq{<pvSx-P~=pQ<T)pECt!x
zl+<XarX#eK(|p0Z1j=usM95j#nHfNi>9L_(NRc7OEI%$?U7cmNjsXx%vx*z(^`YYR
zR8s)VMZQ$|=9}q}CqSjv#|cQa%QS5k-4<!$${Z*a%U`y%1+E+7lsk=x*xIx^;bqKC
z@Ks32DF+pAtck=0&nDK8o)W{BXLYImJx_|N-kz#B2rnvqhN<8qM4skHZ(nC*<ZLC3
zbrlOmN*G_fwK9*~FknuhjZ9Mnnw20kGFc|PUpxN;-Ty4AW$uEVplN{g2OO9G`R7#g
z9{{!`uO+M%NuCUIVp>x$4ZM#-War&I*U+uE9}zZxupcun^W3&TuTUKWv7cZN0b7p8
zhEDI7<0&zfDk{tvFzb34Fxh&#(?@ug&ZWDNM~Fr{dAAyz&dX}3gm0R#8sCZ;-<G=S
zM_k*W_pThmJ+qv}vYR6jR%W$(1QtlT;hE^^;!#dwgxie~y3VqGLdz@kM%7h>ie_P?
zAv6NYcWU|7t_YFdjz5XirNPV5#wJVbUX_QD2wAGC0XMCN6g*#RM38eCJ*HLW7XPrO
z;LC=h+~ySV!{~>Q%ZTPbe#}*<l$aIrv}6O?#H~V1MMyjhMk6bL3Le>$ibEh$A(gww
zvlw*le2*!h;W+w_W`{b^pX8Nm^*%s#27buf?`VWU47FE>a5`-5EHi7HJP}*pm+9$1
zy12L)j;SVFfc}1@=&OCPa=}?)KcnI)NguzHB+@jD*Hw#@Zjqr~N7xO5pu+gUIkJDV
z+42MbA~}u{qyC$QUfPG>k}Hu#rE+#gM$+W%7&w`gE7_>ANqB<Q20FPLkyhrCgp(dT
zcLJ4RLnoxnu@c5H#jsmqSfR#v6R5aYFVd(R^J&vV5FAK}X+cq@c>SOkFfsH~kyb8O
z;Dnu+VFn`8942DT+<$~9M(d--5ht;V5NkCP89r5=M65!~p4`YqFlthor`VZz&0kxv
zPg*QYIOiYKY^XbTE%tyzjgA?h2L(8!C%VV86&KGvmN$oRZNgy1Y|LmBKEYpA^s_EU
zL>$RkC6>|3jIX+)WTf7r`c>aKDam`PHfH)`sk}wGl`&z27E}jNg$UwkNO?`}bO92|
zmDn2FtjA5Tu-SRhP4IWKy|@u{oi5A_XV7+A|7)8LSFOfiRkd8}wf7eHB-gmVIX=Bd
ze$m@B9gmORTtt!%s`oP&p%=p!65YOkk57|!T;0UTMOKkN!cIoZu|89tozEkOR>-KQ
ze`bD6dv-*Wl$`9Za%^QnkE8SmM9l0H(kBhr^($?og;O-T(!C4y3kN&keP9+_JEo=@
zwdh&*pM_N?<~uq_H;94~8UTt8z2wD0%atY@@A{VXNva*wO-TRb5KHNaN|R;KEH(s+
zKy1`<L6WCQ?;8fKvOG;E2Yf|OdD_Yo%VfFbm&RNYV_1U7?$ZUl$-esrjhXZJ2#q2d
zlwYi6e7ZS1`$W~BO=X0qK3H`EQBhIT*3uNAzhLT2`!`gl>75BD+}>?dve2O8>Pwbt
z@Xk3^-?NV7;NH=1k}M#95v3ygu~(g;b&B`e;oYYqxpR%x%nn*)+a8Zjp7E!1d^DaF
zb6%t%$XaWQ_UznCM2y-P&L~R$8G<*`rH1ZfhLvd)_bkCCikcGriax!L4qH^7WpIph
z{Qbx}hv`<*LF(zwCG9w8Lw~C}Onr3s*VPLuJu8i#G~5U5TqvewvNA<d$;gD2ONiJ^
zK<bYQM{-ir?bgbb=|ejE6_r{Ihqp+ugOK1^pK3(dQ1`!t-n76keVud^-cu!E!?Y;k
z2wQpI0;Gg4F+ksP2=R%%<s^(LeI?AuA>aICpvWQ@MyQ5oYAxxk2#B8)HBIQgJ{@rA
zHCC1?tv6Xb!%oozdcDNa&1UI2u^5SI=M0Fdx_B}cR$Cq`(4DyV;q?_;czOpHhze;V
zU-Q`Fl>1pE2ic*oGp9UUPGEitzVTkHGBIYhjoGT};74Z!Im3I;JQb8drHo68{6E*3
zMdK3z^8^y%g5^P$2Jsnj_NO>H&$hIMvck-#Y+4CLQ@SmJ`eK%p#dZ<$zT2B07auo7
z|HACTmPntc<z0k%MSU!#7uXaUqBn^8{!Lh?R71OLG1h0u>?)diO@GOz_5RusDj~iW
zm~#8!*5H=g!C-Cy@V=G8N)2R3ImvkihREk}eG%{?5jOG-d=;Oa=LZEiaj7Q+#~1zR
z3g1%54&5o43L~td->M3rD=4F1zmtpMI9I#pg@7~gjKB3(6@VOSG#%GfyF#9l6QJcG
zT<8LclF0ZJYSc#@v1fhJwjo7&(##H`Vc@uE-SvYkG{jhem~|Grg2zdk19Q(6L5=1T
zLCeBYKCzN?vDWk3#m1MlkrScnOxO1ZG)0HhJ4xE(KJ9Af5m0)Pu;3FONtUxEYLwZR
zl;=BoWr@MnWWF}WZf3-zYtSxL#gG@^6DGmkpbxw5Urn(-vaDh+)MZ@?_M)S$zY^J+
z>ki<NU=Pg}aF;8@TA39`>q&V4yIi`<QZn|w56;c-NiVO;tU^|;qp4b6>!38PoTCi&
z)BN`nds9?D|I6ut_I|0}c*)l8wj8ibB&^+5vxuH1#BY23e)Lk#;rAok0RA1(oDgBf
zXpWRc{bwq|_*e%HX`09qViT35B8Y@T!@i6$MmR!QAc?eE!^nF`Qdlv3ruy#%!XrDB
zr_?`H-Bev}d*ZPf)LL)P9QweD5qj`p5nG5xpDc|OhtY0&II>sQz|69c6r^h!xuMu=
zJKvgng-caugF$JtlRIw*t|{`-m532`sD7ki)J-DqjSIU<(9F^05Kr(kIWr(z@<q6G
z?=TK}mlPzsh*0U$r0>h<-$~WB7B~|)M3F<x1PoYTQ@R%xs^t*niAlXJU8B@>bUV=^
z&il+eC&|iIBZALd#4$yQ5F#<A#j7LDmz07ih7qw>dX9vi(6eKJyR;%)B<|jX?oSr&
zab!YK_8eFR_`Y?Y+{M<ICV{Ev%kof#0kV-c28`~{5uQz}zLXGu7IW45!H$oB?{B<;
zvQ?n33|z;Nn)_yKsEPEWE?XnPEXGN(0qT^XS2!31D^Z5L*z*fM>D9M4cppR@WdHk-
zU4Zb;`X9Q&Vv)O3H$%)cipVM0MrA51X5_Pm7k=UE*^%^e%X!@Q_hreb(QHQyH78fX
zeWF0Qua9>~rIv8#c~i?YE4(`?B?BgEZgv6{6e_ve6?@Ev6x8#!88jcNFtbCK_0P|k
z(1kDf)x-&i--o3WYV-w!`K(+%ZT_@$T9H1yc26B~<VEjHlt!O;Efw)`shRjR&{5Dt
zSZ^tbl_qPP#<@^FWiJjBr8!SNTFf=>{7z#wRQ*M-!H07^tuXSBL<Vl8wstIg`c+%2
z)c~R>;|T7pw^7drNFcO$3W`%!%eW7%-rAWEHzf}E=(p7j@tOn}JKNe&9oa!qVl}PG
z%Pd0Bgopc3c;pt6*>D@lX{Jl8b3Kg7$+$)g+*=8{ZW|D}(#l16>3@KgN}3q<IZxkD
zB`Bwat2ezYz;M55$c|Zao-Dy<E-hxFJ4rer4bU85YpFkBnt1qJyTDjLDM^kUrwrw#
z&F(%CAhoV+W5h&p5TTyo*QHj`nSC2G5nGVVlZ{rj$uOi#O68PSshnotwpV|)cECeJ
z+)UcrmFRRJsc)UieV2S`zxxqLZe_5<&X!7!R0{D5k5=AiE3ZS9fft=Ku1f}=>how@
zqj1xG69xU=m3l(uvd-H0?vmqXFj@kHOipE6d)m|}E55|LDkLdr1FHb=F^CfOS|0@w
z4$0Xw$O{@<gcN6MGKLl4csW8+qPmZNuhA;ieAwl$a)hD??Sa%2Hc}Kohn=$wBb&H+
zv?U8LgtVr~rQ@m|ponql{3QUVVJ0d~Lg5Fhsn-~V>8Dy4v2C-0*{a{hxQtlngS^mQ
zS!KryIA_K!_CL4@{ED~LI$NCZ54(Jfl1dGSte!H}fjAYYc`PF_dmaSG3tyP=us!*8
zD#%7^`OC{f#_otN_O5o~B{6#wO&_=t=?x`T(C~}=*Ed)auiivM_vSDY9x(~s_!zd{
zcvP|)w`!WgS3KMsu9>--TEzIWs(<q1<&HM~S~L%%`PYbzg1+vo?8J!I|Ks;QeCu{n
zR94f}6v-_2&^5jk0AJ!5zWuEQo;0Jbo@~dC8=Uu9+i`p8JRA8yky*`Bca&1(2N$oU
z6o;61Y;JtSB>?saAiO|ke4?=ZN&O|+iEK)~?8Kd<OkqII8BbwLfYJmT7hH!+B--&Y
zLNzU?(=hF2n~Xp*$8+Uw>#i3>)9=-**zy(atf+-wBezNwLP|D#AMi3OwS>9=NQG%C
zcu#g}Dy;1ZW))~{oscl?!mR+UcJT&{m=6t^*`kN47mTmzFtZgk?bIK*cno8p+iZg^
z2i!`C^%R$OSG|m`S+C^&YHLIBPg)^5C)uK{`)A<EP4A2kN;_=o(2~SREh#zy+R**B
z&vy!*tp#R6yI-F}G>K+d_OIM?3XPqK^2J!+magu6Kz$HU;?9fV-IZKoiV*=ECX`I9
zaSuw^$J5T11<xKu;IoJaTIBnSg;#ddl=Q9{sU&(-?I&#gz(j~~iEpd5$>%+-P)SS@
z;-8NqC$Qa<IP~QL^-n5`7%-SvU+8bA(F$3<X%P$xo(VY*ajP^R&DSn%j;Rx&(_H2+
zkJ>XceipeV$~x-U&&H90Ptz)-lcs6ZE8G)bKgL0L0snb>k$Yi3dh`{UuCG4ZmLcCI
zME$M;{9aY-{i41_<eu~r1wntm2Fp1G!89v%&tUI}0PB+Q?9XYM+$(Nt6_=ojvmTr}
zVe4`~r<Kr%6MNM!oQ+z;ASO2EllZsb1qyaHA~HP0Aja<Crd>kArrz*Pa3d*)38Uci
z2t%%<S#q$`q2qLKSN9416;ke<9eHGF@B;ej_;3_g_s@r4<tE+kB$GKN3hU`HL#)Dc
znIgM(^~FiG4zMBF!L@tRX;w!c)hlJfi3}MICFJP0G9|?tgw)@QTx9baYmy9(e;W(-
zv%F5z8fHERFOQDV;WdWwf8}ThHD)PO<kTOVCR0Uxs`nNEQ&=C(8%txkq(IZI4y6mW
zlTDPK^O$NAdtQNU@O`^r&DZ*X?XO-mb}tHaP)F;bbb2#vFwI(y52%A|sZ-LOdM11k
zqYQ2mnErxgr_!BbGwo>2@p0ug`eXHQWNXgPZi4YtVVJI4bch{5V?;&sN3@wq43TlZ
zd3IL3ACfGa3zQI<V8ev*Ga_<33XVb$S1w0l&mK}d{o+pWg!7O(93G^=-tozcw=qO0
ztrTb*LH296gja);tZVgPX46R2bK-$4kgkiJm`$7F(I9j0@ZXjXwq>{Tquu>~&nBb#
zmmL8x;MgAkQ=>LV(bZz-IBzgdSc}V4i7s1SBbo|tu2;19OlTn)6C^5ATXpg0cu;L|
zf<Sllp8I(;z$onNc${tuZ1VLe8+q2}N}NR*p^5C(XWzW#ER8;(@cyOaXNU3hF)N~s
zUu$2c<|~}q5vPw)EC23~)&otkW_%@q5KPQ1OxYE`GM8?#|7lUiM^2ig$A#AI?SVET
zCue#b=zBGb(hGA+Kef9beo6D1_(}nApHHwdSnSc@gYy>R^MpYskXoF8&B$a_%2#M3
z5rW1Wr5(K&k(-cQ&YA8N1Z+(RsAx^B^*Gorqy-d7d#J#yBmtroj~lf!zWel-Q5#DS
zvz3jgqtEeONJ@GKQ5&`}2s3Qk3LWDs?~z3uXXX3*QIRvvE_r>7<CK~Rn=K?mjNWKG
zA;X|p9eohS%=L6H-fHq&SD1a}8f)J3G+aGaFp~2C7SElkiHXkI{MFL~RXlWFyGL6p
z!!2rsAu(WghZO82l(&&eh)2Bu?$XN?d*J1nobjuG{vyJ28u#>kdiXA?GFRr}pYdY;
z?=_=PL*<tcpNTW7D-zP=ULir=<do3a_zg~yMAGUq!}v9J8v%AcVji#1@DkJ<!BfqW
zaei-x;;DSer)pYnD{WV0g<l_q0hkor(N=FGsB!e2N02^~jF<&8xODHpeLf=4c10kq
znD7T^!MqcdM{(@COh26#N2yBarLGm=qJL(DjBtkx$+pvO$eU=<fWmd5f6b_Ji>`98
zUKGx)EJ)4n<Jog5G-y+<?+J%;Vj5l$=?OeZc4n}G1{)dIrbrSzpCx85s`T8*URdpR
z%-s1k;qel}>G&@C0x99!9(`t2)Ktuk1yMUQ*^0<|S4au515z@9<k++*y5Zn69L0yv
zE$jiJ(ET~PF<bV|xuwL0h%hUq6g^K_NODyOaYp?kF5$e!Lnof|K~0hKHh*=gQKK1w
zNRn^Z8r>)+>>$r4%PZ7KQ}fB(<rudJFE+F291@!;F+^!fT69sG+tMCUVKa4l&JUy{
z@QopAWJ~x}k$Y443WE|74J0*s@#_;?tZUWAw#Md=Kh5I{-+U#G4PBnKG?Hivq`<h7
z9vJG63*$>?AFwM}8@$U(g{yaDSQTLWRonmx=@+FcZuC7!N0B3FVVaiocPGpF0r&M`
zQByG6A7*GS`u3>w1mhYgXU`&4EHVTEYC#E=%@uc<P!vp2NdUuE+ngo~l}kijTr?~Z
z&)sV8@5fy0F00J`R5F&#H>eF$NG;d>z(!m&ZBt;sO{b}lhMcIpm(?o#<5cj9Q@RDS
zs`WO8aO{V244xnwA|FK@>o`*axCiXYyEhwWB^PSS>Om{?GE#6<AOF~|lWk_jFB_pZ
zh4k2e8Pzx|(@8S}5oBaUvnXKGcdF1bm3G_(xIB6!X%=CxU;bAf&=EEsW%AW^VjX1V
z8IOvS_xP4O8Ct(4eE>Q-=3ko^e*OE=aJNWq0@8}xM88<U3F@p&Y5+%^WD&8DcO_&Y
zhcwg_oG42}$Vn|8vxu>X$FE{TqmUN@y=7WEdaH_`HQ;a3p>@(!R4Tfe<&G&;mID}T
zCDgw^>0>o*8;zTFK|;djr3X3-e=zN%G{e%e(ETPe1iQNC(%;|vYJQ+aNbVCINkpsq
z$LLO`95n>D81EF3oHIToIV>;bVk+$u1*>Y~<eertfwe7Q-WY+??1Z`3y{=HK3!Vop
zf*1hUIfs_pZ)yz3D~<7{n2C@xzfN)_oFJy3y1R(Y%ZG4f*4@nk<TR_RIi0FnTTfH{
zB=fD8U%rNY>Mh0b_kVe&|9%AILeR-*X|kE{Q>`Q!y9i1esXO43bUOzhL-)q#I20co
zDi<efQ<=UE54SS*^@9-}`fh;O@c2#G_BX}Y^U&_rW2H{3#ZoK!Z4*$LJCf-#)|Y0j
zPfI>Hawmb+mJtKxWJ0YxKZGzf>Xslk6`GJt&L9P0_l|(5EE9XmuHf^9cD_rezqO(w
zyjhKRvJG#)jAToB=XmNBgEy1N4lELMeAXw75-%0;q)urU7b6EMGGA7l1r_nR@Iics
z|L7r$ludZ;wl`(U7~xRcKrq%OmF5qPJ}37EtC89=|48S5FUc}!sNrNkmM<TSjKh&&
z&_+p3r7^8!m?Ie?Ba?fAb#qToV2d8z^lq&3{g8YFpuc;{bi2@Sh8JbLEjg`(EQ=ZU
z;7gIj<!q+?ZrtES{tLNAIPk@=%yqW51`EB@LzsiINLsb(UeRWD1=2A|Xq5GRrCmV|
zL}}U;?Xs&t1kJqUwks3>NKu;c>8L7lrgZCKhHtwRPo_~dx~%I%nAAeO6COcdxB!Ws
zF~Q*A#H+qHam3@lex0m$JA!Fn3qT}FKf1fIln!3dRrIL}K(w@Uq1Dy#^B3<twrp9f
zI{(2&5D!|5cp6ldT((9-%WH7NDsW<(y_r@NK08-2V6z)VN$Neu-#OVM&C^gtRBd<~
zOsbV)Gj^;de4F}B_cbM2w(<4q*lg;xa$#qho?qe4s#2hQma~Pg=OxZi;v7}W2wUU|
zKDj6NBZJ7$tgErnCw(m(bY&!WK?5<TzdYxBJZjQK_4IU4;LLABaSqZ5pO>IM5^f(2
z$J`vHPD#FT8!3Qge~40?2)?{py+&qE<lYdckY)eC@Y^W5qYxrE35$Z0Xd@_owe<)`
zuOdrAXEXNt2^JOmE+D1;fMRn4B5pHrGdJA>YQ{A)y+43Qsv4}qS7Cx%jU07%5HD+M
zmnk?lIPcUh{@vQM8gtIhD+X=j*Il#I-T<lRs|R+fh?(ueKF!H*{S&NfsA6V?`1q8Z
z>`X`F#*WnF^vL^|;B~b4TAm!JQ#>?BJ$Pq|s^1#>TwEiH=+6RWUnT(rwXlj#KJxb?
zp)~<SYcGy-uBo?Xn;_sFF1Dz+^Y7rQo9no>J<gE4sbHt$FE8?(*E>WWC|SG~&E=eO
z;lT5vdYQQh)8CXeW4p;>)7zW4Fqp4jcriz#B0D0f+tKihZk*Gnv;FO74*B@B?}o!E
z)V9uOUnk5erUdvQL{<%?XZ++Nw`xQ$0T^u<XV6&o(Wo|{(2QBPD&;ZUuxRjmald98
z%c9-Xt_PvDNNOA*VD$ZDp<;OEG1MqNw{dw|y||}YtJ=hS?MTUOh3MD%yO#1n`-at7
z=Q+`vDp|$XNIrcxwVqP&xW}<nah$XAi60c_$X%udqvG{xe@C{IxnXO1&2``rTPYhO
zQ{?Foa~7xjS+&DcB3kWcN!~U~&hRsL-vL$lZX)N67q1dbMra<mZT$`XN>$lVD|~j0
zI6l9PONz9pNfX@og)14!@@#O5ZH_6i(ukg*CbtrqsJk{VjWv8}gnCS7z}o_LWCkHV
zN>DdEoztwgvQH)0j=~psA(Srh(dbG@5w`YDOy+y^2Mn|klLW%}6{olsV!8cw%1q;Y
zraA;OC=7+SM`tw02lRDc-j>`19ik6y4p1;YI|A1)o>nmkk!(9qPN%Me=e6mJsGX{H
zc{pUAN9kK>$y6nbSLv}lynnFE12WucSO)0~O<<CVU)pgd-#JzxBXCwj8X!DKCi{Fq
z<UKJYD9ubqGwK`>Ji;FXlWa`Q#==S>fXeI`(6l=mXZWd`-%iL@?_Nw>A}0csWp$r*
z?WVPU)hQlLiF7E0?a7GNWdQn~N!~T@c>j_%rt-|8nR=r7;ei0W<nbde$GkIBZLB-u
z{@c!Mk6MZYl<j2-rXg3&Hqa$Ozx=KB#=XRQ6^wBJQx+)4uR!UZC%jKgkmR8xE_7wm
z53VjOxy;ltAv`F@dyPz1o%^d4NAo=?OeLe7A2Yyt@9{k+E*mng(lx6hHee=hru7Ef
z4Lq*#y+6(}Wo<t2=P)N-t_6Y<Vp0k1hj;RH>*xzHC(^RmrgQ&@tDk#nyRpyTDk~_m
zUQE}i2JBzlUKCtdk3W1~UF9KJr#ABd)YDq-A$$9N3>v(RPHyB-!%Fln-y31bB4iBR
z*lZF;Wsd72$|HNVl_X5|oL|1D^$I<4M2bgd(Cg*aBIZ1!NZ{Jpt}TiY+?5M+Y9oVP
zP3Q#fhY=8ujk_6N;vrh%X?JNrKP<`C`ewAnwM7%YK80`x+wXUU1LfnaIxF2Ivw$U6
z%S_9u_rw8&?iJ46Cq5IVh>GqTC1zeOeKR{au4qGUYOQ{LnJ9!?zZicJ_KZMlr@v7?
zDLXqKbn@hNN@J<!SNNt)7OP^>G}|iHKSi*Rn9YK-m%xca-rpM=^E^Q&sn}F1+)b1`
z5x$BrO!D}Ns;+Ufl+Q{{a#wp7kLV0Pj}j5!A?!`C`tE9*0N5*6W(H7ix+;pYI(;85
zx2X9$GELSE$2e4$N)8jMeOnL_wvx#(lSp$v_LKd#NE#PL8TDbxXgUV3jJY@m5E-;b
z6B(~(K!?6CI3c@8yzrP%iu?&?7~SG7p4p#lef6dkxM^J(%Z3?yHBJ-8^-C*u7r9JZ
zsNRiJ7;5h>ee{W~xN;Ln@T+N-NAO15{_1Y9cnM+`5RsJ5F6yEiA7xTRsAv@ygT3`Q
z1sMCahGbB4@4}motW2!I`)0ZnXerl=Pqxhb>Dph&OfzR^LViE;G`R{hN1E%*L?!zY
z>+g^_*Whwd%S2ENH9tLdFR-3b)C^Rpy{{=_%?p*`xN_e`y)y>%P%9yK(_X4t-k3S!
z>vNphpEme5hGz)KJuXiYSQUIq6k`;_yEX8Dj-n@{bkQE2jXWOzq?P%zjjf9e$=ybk
z6eS_Rr4vW(PzkqTNr|`MDe2+b+JYIM8EhhZB=n(ZQlWJ}!aC^I3h$*Y#CBVX$mLEB
zwj_E)+f`|<Bw##s?72OTH1I%Q#85hiQ_!5WfTj2J`asqS4$FBekzV%xSr^aN-X1>v
zCL^kIrq7k{udR(0GW117v2h(FcFsl-m6o<D>4gyG?!=Yrye`dDsbXcrnM6D>`CI^q
zD76Bl{p5`GYgHh5CRRW|I@m$nqxx)1wbVE`c~8OruOye~@WiAfbwcCW_@fFX<_cR>
zY25`+lb^VdZhCu}>pl-=p#5dQc7RD6!eq9?MMtLh5_8UUnFF8VeW;gq?YYM^p)V<V
zd|#0Keni${QT%!0!C<upUiQ@f-JQ_yN1>(ajA89XzaRbBzY#f}wR#DW^S?{G@wy(d
zxd!0Z9|ay?qq!`WTNjq4yuIn_EorJBV#HQyjUw9%8OA<QPPDCF&$N5%|8h6WYYo&t
znqR&Y<C@>UZRu*3ozU0yI`>*A+xDh&+8C^};GylBIL)^-Hw@5kDkqRkdS5-vKgMOJ
z|5?3wX>gI~EZrf{R6|j`nr%E$6(s@KmUM=&PBNY<Zw9hZyN{-3ab<F3j<zu+U<Dg|
zQ=(Y+j}tLo2a<_30IDW3H9v@8vqC|e0|b+t32rm0Aq<WNhVMWxn6J-Rac-RCDOmm!
z>0~4Eq$INKz9OSt)#b5x%b_~Kf__a+78wJDFi+nqGe0!>fNQjK<6K0xbYPc1N$JJ_
zA)oXc^x>U!`2?Rt#kwX!2_)Ojv>~AbN%BbxPd&Tr>;Jgd$|{mUOnOn?l-Bm>^!Fo7
z+C10Xv3vuw%4C_~8q%8q39d2|%vbs<6@0?R<c|%;mseMg>^+l~F%IFK9%L)%j3Xyd
z=G%7?6FX-}C~l#Mu2KX?34}VFz*#pPPJRqF0hE+fsi;mR5B^8*foy#T9W%24y6#dY
zRa|z<(wnz@J_si}>VXy~u3uHr)u`zq0m^#Ka<TzMZ^b-qs9P^4YI(J&*3RfFIBrxc
zdFb6a;kokbf!OujlsiS?t^yum2mPl8XA)*f6{8Nma$l7ZgMB(&hQ74a&2DtQUO(x$
zObW+}nPpgtH*rqLuZiEI%;U#Jo9VkhDTVx;X6`)0FUn<0j$#)RF!UV!%gv6}4m3W%
z%JG~Y_olZsoIJF|YAVZAG&w@&bPWo!px!2SNendxHs+t?I#F$R80x^s^WIC4$`*&J
z65Fn0{Cn(PiXP7MJ0gPo-mTL%%%Gmd2WzjTH?tl<Lt?;zC=;5qz>Lg0&egm#CdG=I
zXizp~yv3eoYL**VZ*22>b6|-fI$M>)L<uPwPKUTaFIGm3|G6lPkTp-ZFI1AWpzTk~
z?%=Cj!eZ0su>ztTi6xdur%#Q40Xd9Rsw{!KznoPYJkG_dC@YKMumTA^G9RKt66z}5
z-br7SX-uDY4{g}m^i!MXw+y0nuvKeFbrs~ij43i36lctg-Xq(``I?W=VQ#xy-%ec<
zbE@uW1;vStIkJY1)@cU3xwsuqjrMGZhx^nkpfik*$XuRbCSf_r1yd#bdI(lA!w89I
zi;lug3R|_J8Uv_87Pu9u8F7Mnsjt|a<g0H6mCGbvQpD;C$t65sZ%amn%(SeJ&0S&K
zvs#BRtHI7{6n*&egi+lP!n2LpyYZl4{%K6(XcONe%JS)wuO-0%TL*l?5!t;cRY)_H
z>HM;AhB=$&l+@k;-{07_>?}@yjB2DVa=q-1ak;`$8(LvAM0zW$Tj`x5fXfIMwOK{a
zo|6%u)gfNjNR{Yu4v1kUhLT0|fD-*6Z|^lj`+8(a0@I^I590J4?38D3iLE%0xeUe1
znZ?+|_z)st$udb@6bS*LA!7DlX0HH!&6P^RTvt#&A$rSEts3}#SouTY3n2+tb5U<0
zU1&q>`G_l=%mAz*K4q%h1g*R255VzyW%Vu4l?;t!8@z(s7@8%rYN#>24TEHBU2Vue
z!mvYzn3K9_XySP`6)3S&2jTVH7^D>SSlHN`q+kO1SHG8Wo_g*-l}M5r{_LG<okXtm
zD$7z8S8HWIxMDK2>(3|SyI#8=2Y{PvlV$>VEZFC8kt(E+z&B)1S`0nu`8RCn##H1X
z(98hZ2tfQ}2A5*4i(fnXRJB4Zxc=qRYwpV>?E>)t+Yem&!Wa$^xkH1QweCJtoFbP)
z_sH;p&wXNF;4h@-Z2yPCvoY0KEu~K*@P{7R8S}m-;R~0s;=jga#f$XZU_=%q)%7@H
zFTJ~p-sp68gm`j8_ng|C1feo`k=!@;ddQHpYMc}zN}y<M=o&oo&Vf&v`dzt|?<P1x
zEj<G^)TU@-C0L=6>NpPIReNIcCcWf)^pN4t0<_nvkdbk@I=Xjr_Snuv6P0d}*wQ{L
zeyiM4mTg(*OjYTM#CH=7P(h$Ui&Jw>)%s;pxhR+3Ww=W4(LHBAr-Ww~(nvZ@Coge%
zs$t;S?z4FsHSvK*BOS>1H|O5wE+KM!+Bq<sx1J4s7WX9vK54ysT3wR#lDVcPJ?Wow
zT>M1jjL7hTR-8i$rzpWUeXf^gk74YuuRK0jP?KRYh;6M}iLKs7QIlzpUhvETc5To{
zkrAbMzI8Ira#+(JTb$P&ri9HdwCqg+O6#SD0JK#b=^M!$G{+e-k9*a0wo0>snWltU
zHnQw7H!fF9X7$(6faD9ui~bAQ?>A4!l!&`$FJiR1(*5wp5h>S{H*($R*W+jo9~UPd
z0rm}k|AXv_bYyrHtayu2>Q;c(D~r~CKvl%lEF21KQW%7KZfZfKv%%9GA81JaT+EDB
z5kTq`U=$w=Fjh%^It!%i8S0ubDFk&!K2><9`D+1)&kbg-5IQRHB;#~o^LU)C?hn)2
zUm0&2;G>&lYRx3G>^rA66ZDI_uljr?v)n9R3SWj-KJZQirn!MhGs9#;PUsT_a$H|%
zbtGrRhlnX6*6@Qj1tGM_8!bhPh0LnmA;gIjg4HKK3KMA!U*^sTNa+V#!yRS~i1hdP
zmvEN12#u7S;GOt&N!r1dY)akPoyt5CT16+NM!-S%g@oRrUxG`MzpaUt=+u6_z}!Dg
zmB)YZ&X+SR&MrjOKaB!P`*k$H^=eRJOTL9p+Tv5$W)TLl9<LMM_lj~>F->OdN=l5)
zC8wEaFPFXabM~WHZ6&pDCR*cR)MXWT`b^n}Ey^q<s<6~UjKsJ`9Ruv6&Tdg{bfOUA
z5xx?tWXAYoTd+RF*;5ht!iFx6OOsBj1w-s35YDrM6ly3-_9he|!a<BiYKXQOLwuzT
zU#Peu1%8f==FTwFcT1Esl{<m}q@JY|{pam(I~LjU%?X#|dFOel4U@A9o<n&0Pu?{k
zrqP^qE*vsi<-PD0-94v1yP_|*KeHECqW7bz$%D30Dj-fTI|i#rTxD^qacDz3$t1#U
z4v)K;Ss#wluhG*r;iIRe9Hk*K=4P)Wu;(x6R8&$3?Liyo^cB8|<5GsHYoLvU_%stf
z^LKfem&M_O$;8VHs)ik>s;D;%M1~#9JlzWlGR0S<$m-fAT^+mg6$h5)nBpXN8`$)$
z11x^NZOhHzEaSO);iA^fwowR+tV?TTtXt8#h|T9ENwnDc{fO|^%kM?+>im=Bql6Nj
zZI+!{gi(XbSh;NYmc~Hu_Yk@$68dL;Ok9wvLxZYgs}zyUZKi?>Rd}r?%y6P6!BUqk
zhlqO;s{&2wf2V7BRp9}k-a|^1lmSgRnTop<Ok)oBB^+z4jlJwY44=6p&zmKM!P+m&
zYmX;si<>40T^o(npSH59DxF5Dvz8-une`dAGV=uZ9yc<J0ea^QMT{K6<G>QVi9ffg
zh>yU`pd*FQNF)~8sLui|SN2)2I!cKljPQv~3-7MmN%O^k3(P>IVn%$PTDDO>CRJ9B
zWi+H8_A1<MZVp3S^eUQ_U>2!{zaM>Z$u{Pu`I5(pV#)j(&8PWc_RGXk(uX;3y~~eQ
zwd=*fc{f_!xr@tKFKN0cj@lun(RCn{657nH{bh;9s+TuPN=!X7Eje^{?1RE<o$RX<
zMl_Q`1Q89uEB8VK&^L~69&rYH4A+ViP{?%pi}Y7Qp@gGnSfS;+^w6tVa2;}6f;a{Q
zMwf%o2>^F7$3Y?kL_P20buYB#?G3_whUn!Qh-kR@Uhhbyufz=GD2RT%^2K$E>%_q+
zxv4hXZD>9@{U;5(*p}?1XTw_1a9b2~5{Z}O>19^lV~r+>iv1a5Tq8vC#G%Cf4C(R;
zJ`7}TGd3st6cH+){FR#d+ndkS!nSD3&Jf}wM~WuE^{0q_LKi*VC=wTFf?isRmH906
zI{cSg{#%I5kricm_*Q|;v7<5SBhG0EQM{#hA#rF@RimHkI77<Mds1MD&P(T}YI4J?
zj`r`kh87u-KQNnkIaZgXCv%%-nIxBxQ6_D&$<UMaqqfFGw4<1eGaQ0~JXa3O4(uXp
zd0+<uA!nZzbFi~XidPR`!>m{Bk7-pEv&yIWVa9nXsbNxLUs-KzaOEX5)H|&w2M23g
znzIwEl)n<jhA-!>nvw5{=0)-^-pl{T=(Q?m;TXLsHZIG4bs+1y2G&q#&y7hz>8;gb
z<toN+IRZH?(H~>haypR~IkcWcFfw<gXK>gNUQhlULdZK%0DtLdeXh5f<f^GDN~#6s
zszPA*kB{o&XcH$mbz%5MPz6~kKe3sg4}giSuoHT&tj~<0B~9--&+!*GCjXky8=su>
zEov^dQcnAf2jMp4UtJobB6<P>7nB!lc{kJa4FmZ@SrXwGe}`S9@Ew(*&N42v=;L=^
zfl|N4ZghV$z<jdpR5d4Krr4|Ug}@6Fea^=4&M4?{4N7Z~EGkmw)@cmM+o3L4f@VjY
zUoh(r`{`LpQ9o7xc2fR%cQieyAc+wjkb9@TZZAgWC(CL<H09}R(49<c7M7!&f$?$@
z5wi$kmLjp$LUFG2Z!u{r7I%FGR29QgfAtAG2HPB}#4|QZ4t%SFAPB6hPsh*upui>3
zkIP0fd@@9uW8dm%RbH4cl?Bg*Hb(ZkHYrSbtQMe5H$)lj1gR2Pes)KlnEr+1@=;}W
zn^OvQaJYF}{zIUbt~aN3&XR8qwogB##ijW+3Z7xwpZ2)haUqQm`lNhW$vIB|z=Qs1
z;k(kz+xD`ybimf4g6i1)7f{d4Y{+qER6$|6?o{6b?%cwc*C9hBv86A#{SY8F_KCCD
zg5(K>>#v*pNZCe$0fy^yi<MY2zVI(I^SmpY=dY!^w{))8HSezuSH0Zl7d|vrOxX0b
z1VzaaL~09sN|g1P2zro7IE|hog)nCQe&i(+*!X1-Y%O_``}>jH#fA1TAC@`EPj6{O
zucDkwU?tT5&7B|fmHZr{dw=QnymGoGC~Zu(=-y2t=1$n+_oFx)|Fu1~O{mX(d{izL
z>Qfu=f3*B4)8eTYVS*E8PqpiwW4FY9j^PGRISs84y-pk$7?Ldta5IX%6Iy$=J}1{n
z3Ld!6Jp=MIT-h$QS>r~C{oe@IcR!-Fc(lo`N&bIb7&=L&S65x*>kZ@s8?t*GDyDGh
zgzx;=EtCq45nYX2C8+j*JH@7^7v=1@g+%>+baIfFO&qg=asl5^_fMOAEG|3TlEw85
zq;~!NsHZyc1J3%Ywv?^+5K&_(xKxu_cX{K)ydOFEaW(Iy9DU!z%i)5ck8L}|q(KGs
zu6d6G8%qBjwD6^L?or>&?JDln_f|0LH8X3$bwQq22T#*Wdl;_IAb)&vtPTqPYr0_D
zdcdpAfNyst;G(X4*uVE?t%mN(9{U)V?cDk<@`<av4?Mo^qq@m$f8O6o=Uui^|Ls@O
zX=p)^$0)D&1DTfE<aNS-4I|+Uak%tRnhHNwMDOL<J{|mi#35Vw|IW4d;bl{Eb-l02
z8eQh5wbsSG>dl%5Zmf1(gmT|CMpwLM=rI%g=`rx;-@g<8{b=OB6<z~Mphx{nyRH$S
zPhvx|YWn9`+$1K#E#h#j#$TF1#o3%4?eT2fZavlCdLcj+yVo84_SbDYOZ!FiKTSgq
zCYJ^59r)V(*YR?vJJ=UJzuE+K@r+JPJd36^em_!ksZ&j(r{K6*^<cm)8`q5EOdel8
z^!vg))I&082N_7an{R7EEJwnlrg`d@NXrcfr<QZYuGtq+`>}(F({9w=??(&psHsWn
zk_b;=+k>><9bsUy`o`vV8r)!Bdxj=M4x{<|5&Wj^vIpMFgcfI~Tf%uHK7bceDsHb{
zlnV(yyO<pb%DeT!Wi94pa+=$8P5l|Uu$8kQ?!(BMp54R1fKvSZXyt*ok2h$}VTaf-
z;;xNcC!+1(#5}1tuAD}o=OKXda}SP~7(aWseWL`sVDeztvRiHkGVgyKbp7|9Fbq6U
zmZY_LuX<Gx>-(}^nqT~4-v9DnThMDZ3T#!7_GPPVS2M*e*XnJ)_RO<%*oT{1$cxP<
zD|}1g5nc<fQeMlv!++02XwN)N&t*8>!t%0TVuU7cj(Jab{EyofK+^6@gJ#(j6(leC
zB_d`XTCPI28sIob&zz^4Grufy$Gkt7x2Od+aCZe#x&8es_T%=GYL8PO_0R+G^p-4F
zZj0r5U6k=bPwVc7k?8q)4|eFZk7uKu>}lLw_<py)0^g;w)ah3v*Y=bLdmwPRB=4tw
zg?vHJ0mfb07OUkyype5Y+HOYDi>mJe>HjD+5@*`(AoU)KYDw(5!w^Fahr)evCyP{_
zzaA*!Wq5<kFu%Rz{p7wH4^n&>_IEu~!2Q`B*2_wiaTm_#pM=I4gC7e8R)4f$`B~2p
z-sDtY9_W1y2(YLjy|Ts1^{cLAQyAP1owAuy49R}o|7TFbquWWIc_Qp;BWL&Cv(4gF
z6C!7duUYoK)s1HXle%K|b^r6A+-la#AP)9GPfd$su|X6vwU45){{{c2a2~nsfpGSX
zhONAZpNi_Pibp1ihYuDYmec3yWg^`6SO)esj?kc$_z0IEdTD>cz??d$Yc*dVPu>!c
z+|q+GXnQ35>1GA)kV#oh6G+>3f0_USK3vBz{eD!T^(s&<HR#qjq-MM4i045$^Q?^p
zhll?n+>7;HIShqC^H96)j}hUU^EDiV{y)&i&`E00w6uFpu`!}u@*gJH;e?}FF8Hqq
zG_l~3ykn48n}yr|@Zevu$BCi{vB$lv=t#NER`}wWTpFK6T=@SDCb*74!Mr$}YTBW2
zhgGzTF09Goa4L}M<j<3XxYg&#Y*JqP&pb{SORjeexWdT)P2)c#xBhv4_s+L$KzQV;
zMzg*BQ^f@4_ao-L|9VsN{RfbN?VD8f!l$d#250<@3v6(&R*$FuO3n3|#pY%?4j`c`
ze*PfYjrxz7{$Vqu>BktHG+p783NO;eOL4^8yFPUJ>94t_1=-f?xc}5k^cy~$(&;TE
zCb=Dx!sYU+$%OU%*AXYi|LbU7?|ikgn&8cL^r_(5^_am_%L+ZQHx_gqu#^P9?1Zx7
z%z-PO_XCNC)gPw7@SMs2j|PfXm@PZ~J9pqy-Tub^p#Fa=`7o({<QqUR_vsTH*8z7D
z|J#DQwL<SD{ZxmUe5cgy2Yu(0e}`))6#M-Mr=iH3LJQnn^JY_rXSAjNKg*+TSx6Jo
zR8So5&pEs<;l^(Li>Nqd#K+^U{j(N8y<ppT3138^>V=QMcf9|H%F$jE0K~MZtpL>T
z2#yErE>mPN`w=*-_)n2>Z4>@`cgIb9#V`h2-%d%>k1W7Aojqih^otUD>YBk9Fr@HP
z@n=)94TGr?Zkm)KS2x-}L%npx*hAXhofUt&MhE<{#V)X9I2JA!w}HDioqgYtTSu-1
z0sVmu(Y=AxZhxNKRYWJVCVlxtHT|&f-)67(DP$0r3O-Ni)q`Bxoq*!;ee${Xmy}d%
zA#USWRpk#Xa0U3Ck>`l~4+Xn?gz~Z-;#5fJiS`S*21WI)4ELoO1Kh(wOb}jvA+oZ)
zQ6_Edi+4^_cc{3TSFV(#x1Z^<<Y<z4vaCEqfx~L^SAo(#mb$pQ0NZta86ugH?J&uo
z45sF|D^BS1`Re32DxO*QBh$x%Ct#fU7oXjC^JnGF+9}7CBThi#{J)n$dBN>u3RjC#
z;A{4j1X-I6TOmPZ+&k$c&*J~0?yaNR`rba#P)cdh;>8O^iaW)jSaA<7#ft=Yr!5Y}
z2~KecuEE^{#WlFQyLIw?-#atEx$C{}thsCMT66aw$;sI(IcLk+dG_;sp3mpD_B!Zj
zhj5<Nj|>5h_iw$qC;Pe0#IddKnEE7g;<kKsdT{XulIGO5>?3NYvq?BJh-Yo?b9d`3
zPo$n)+DC-fCwLM9vbvahVSREwD%@yz=hwSsq0TY#VReJxSf3wKzK)P?KgA3lCpj>$
z5jiG10rw1R%G<%JA3MBgyS?>+f^p4sQ0M9^&EI$B>aiP|`5Xj^e3J2Mrkp>k=TnG}
z^XfSIc62~c^Ko?+gbB$vYYk^v)*&z9>%Fl<_`aoG8axp!CXjkP))DBP(X16ZdZd{?
zng4@xW-hIdgn{cUyT?wdN8OUwXV+8T<K?ZpHdY}31xK%(GJ~2wwDFNGw*{$3KP=Th
zo(AkG)E;3Jieyum<g+a+%2e+77vKYkbCYhbXsdV*Z8z!y=*ve_eoE&&+#~eM`Uz-~
zUT?UJIDB&Sr+V}e!@JLRq}QLlzd}r*4*X*KZ~l<R?%=%j+KaP00EO=NjgnX@EWY*7
zR9YMcIaQpvS}qoa%l*s9k{ZR7vc}`3?SO0J-v9t|3<z~6Xsz3S*2WTizhpuAjJqa=
zEM?x6yvWPQ6vlvmr<UIpA#5v<>|`Nkjd!um@o`K+m=TYF(RpGHWbMYKzQ%X{HcWvu
zz;6Z4>CBwIX8SPo$}A(PXQ=Pvs&J4-i$T<fz7!iA2K=P5bh)1)&-0eb?L}Z36qldi
zxP~E#LHS}`8BAacZ{#p}Yk-d65D%km($=~k-SWEw_bS}a-192z1pyHPGeMeGthDnS
zOgxJUr5ax=>z5)nX6z#y8|i4caAB@`o6UiD(=tK4Oe$H4xEdCWj#|9wm!Zei)8cr}
z=96GnykArw6A1{=d~VTWIvXY|y1@7|%6*=%e?KL)P)6@?*&^&1?Mwe|^gN{G&3~it
z1Q;Ty^-#B+*Q);lgg!8d9`fY-_uH`pc#%(e5P$5nZvHhy?+}qCEQ|mFlr=q2>L`|)
z<Ms4Fk~Z5myGUP~(~XTu-9<RB1gEj7qVYrDMNf(1fo6y|tJ<uIN_J1x+K8gkNXVC#
zl{8*?bC>>tlgVXdCXqMXcUK0#>V?&cc@5+W)%qVhm5Zh<2ffVhF18J2Vp(nHod`F>
z)V5?C`!@%D`zd^P2oFlsrt;rWN}tAETUpp!(<!qXp>S4V^{08y%_nBBC%QIYA4z4&
zE0!RmS4mIAG$f+F*FaM^FIlSnnM2I|L&H_ZKacf?tC&kfXVzR*Mp@!*d#d~-PwgT5
z56RwA-b#?SlqSma^cKTVS+nyD<222bDO9eIj1G-v1``}*=Tc4*+mx}SfY}}Zctu#h
z1G1@ilQn-OO*{F1qP(SheMCI}t9f5n37c0M?~z5?Mq<hRUg&V>Mc!O82u*<`(MyPk
z)UQexvQVG(YXZ-i+%6e)^N!Kby&kgNvFN^ElBPdxY5@8Z6dZTt)W|}25nY^IO5Y1Q
zz{5WBIBBQ`RYRH4u>vM${MN?SiWLH@Tr*bnj(96{MrT=lYzLRZT$>5Ndu~$Fx4m*3
z1HM+Y!e|=}pyBsL<HgF=Ucb|`S&u^ZBh+FNC~lEah_C0vGRxxUib3NSq$V1NA^9Uw
znq7zX-QmOcZKMW30n0B!Sydb)85Bv6=sDjXq^K3qiYb;__Q>(Qs&@}3%_kYT0Ly}%
z0q`}y9ZvCc<n)6oSOo`fSs80{z+5MAa_P{r8YJhh?4$$UY(27hmu)e=(df|Kq*#;Y
zqx-hRxf}@DuD6`EJLL>p`{K`+E0n2U<9`qYUVZhe?e*xzb|w7DwCdm)^*7RePXGYh
zu(`O}hBssM%TF2pL>sR<#ZaY+T-(h*1%u(8eA^)nKU}BmD<qbb3bS_sj|0`Ct63nM
zu9p6-2fX7ZLc?(z<38Rs#Rn(WtSF9cWvLgVZM&Z^ja9M}8|Gn2sjdRDeUVDM_AN3t
zzJl-`)V)W<1M%+#*~sV(kyWtqwc)1HcyDPHk3~-}hEBgV1*&v28zWJhnYPBzHrcxQ
z9_u0WBd6IXiJ~PO8tuxpiAm?~GI(i$3wR%&q)>YD<*U|A)Oul4CG(5<TD*5PNZ~~F
z>?XdGf!YpNPFL+E7sn)<c^_~rA>4#!O0GTXMVFg=MVH~^JPj_^+?<jm8m}D|tbQUj
z9go=(wAOrx724Rkz|_lc4fR+`Qn2qzBZn-B6?40@!fvqaNKFF5%M=wwH1541i5~~S
z-19EE>^2^7!|lDm+*gBt0k5>Iu9e*)%s+AQ4N7+m+#3IA$ef<{6b7wt5G)+HB&gVE
zTmX4vQeOg<V>j^TsLk?tPBE>QZw1~_=G^|*p85aBI~M<$qyOArfEOteZG{%+#_sX;
z`Rn?XDyNH`S%e<g2B)yA$e3c<9d&+bI?{{p9x348KL>GqrjWP&)L+hJ47|q2THXGq
zsf;W#b*z?~oBB_tOy+v5Jtqsz!bl8bqwRqQh1j3Heez+T*v5h@Lj${8OeeSLIMWS^
zH^Kv(#v@~YTCR4>Uq?Hw!P<@9PZF6NDyCH{hSfj6KP2qM4hm|RYZG9?`F3vuTOBjp
z+9UNp^Po54ge;DPb^=~-f7q4;L`_}$WW9ZXiLxw)=8j!kdwsnPC~;_304yTJaqQV&
z<WRFp0%HDUJdj9;jJ*Ju=#{N62~RFmWOVrcrSg9~)bl(2A5Z-2CvvF>2bC>LsoilK
zjbb+bq+1NCBu4gSM<<eA>+?lp%=+`B!m;HNJxK#<73o~1J!0^UWWL5~mrm%^-mVJC
zOrxn}Lb4fM-B9Q+?O(x>(v85PX)PU~DxRH>Vl>hpX<lx6e$9cI9+O7kCu-{D9lMaz
z^xyw_1OJ~NzhsaV*U0}Emk?Vlry5JkKs_~;EBrb%w2n*UrM`lO6i`wUh~R{hl!rhz
zLqk&_kVCPrY2OqLBY?vQZ699*mABT^a%2>has=};ieXb7VWhy$LR{$doV2u;wP>|4
zXCY&61dmc?()aM0>#D5}ogt-1p%r16Wm3RF9L@b^EV?g0a9njr@Rp#!*mqknI~ho!
zYTVB_nc98I9;bw&ID@A90820?LMmZ-k5U!Dz@684dVP`nV)kFh`#<g`{Nuu!Erb#%
zqV|QndvMduIY8@Fxb9gD#R&O6ORIkhP<SN@62qs+&=>yt8t^%VhzrRnLv?BQ9zcLo
zSfWhj)A0uJ-IqxjKE7gtO;_Ts%8OZfZ?aa7ZW#Vu6z@O2?!TUu-g6!Uj77x0eIbTm
z@Aah7#fkO2?;sj4$awNw%KG@C5+zJJ&USPh4htn9Tw#UY5DyHnX6YG}n?EovEmSCW
z3vRy~)Niz_UWq<xVVyn^%;Xg_`!bBw>-k+xS<&<UnJr=qt1T4!4OI|MmX728(b3`c
zMhDV*x^zed8=DaYnLK;|P-!)Fi{o}aJDD9o6C=0k!@V(M=quuYO2r%-`%Pa6A(Z0_
zK;Im7+*|3M(ibCe5$^GGz7r!P9Csdjc<^mw5kxL;+FBM0!l(s*c)r=#*5XK%pZ8$&
zQ5ASVo(~Q{wD&)-EOqbe^;%DD`MPcA)v1Q`OF0(Ol9@4@vLeVb%r%5`hYEo9OxET)
zQ?*djzTBD0`ZDE!`wVLa#pYS!2`7aHNilnzRr77fcUd(sBF<U$Vky7&#6_2)p*(d`
zZyg8rNwW9+o+{+A<nn40VIE;Jz4_etxWVK!E|Qrk;G|WO@>|WjB%uyaU~ihz%YJh`
z9kWV@1GXs;&1L1Rdcnk|3GLwpPFx)iiH6NA4pY+QmR&+fqCKZF#RfC)82M+oB^bMm
zl+yzV{3(}<uqx;dbBLV!*iMjwaT5F_kA>&3&RolGVQt@k0_pAMmt0TSgM2RDDx1`e
z#GbV2ynS}&9c{mg{*w0SVeom@gZnewpzg-%59*P`pFeNbu-wj0u-MCc`{;M;UDLiT
z9jg}CFVW&NqI~AQmmfFH@sY`Al{j`$9Hu!IsK>rSAr-@(z_TPja$DCN6$&vy*rX2A
zb_WPSzriG}3Y(e3a+s5^(@wKp>D#IjD0#O|`dRT_o$j=W&rqVYUV)u#3hhpkMSe;E
zkLqdL#dx|xx*<PiP@>Pk`Y}>Z`%!w&P^;-$(3*N(YU8G!s{)Dj_bYS!8$E_{HNCRw
zSQMpwTe{qjv4wKKQLTM&4wo(b`H0m=q~NK{Zapm?jb%ce<1BFN#xStZLr%VAwrB{x
zK-v_<CH2QIeiHOV)aB%QQqIq9`-_~*ssZdo|92aki+r_ZigiKx$5oncS-_Pw&y7Ma
z#>y0PG2sF&zaxiRR|lCuyq&O^9WGgF(<Cr>tvn$#s$8v*_GiC%1~*N|<KqQ5EM=zG
zR;xfi<-B}jt26TKFja~=UXUZkZr}GdQZz!VfmZc{xXESEi&D2q*G4E;!VSsSRmkrn
z9p;9@cq#_uHbK4`ocdQCHm__W3zkz>a<`&}EwUl5Os_sR<a52j+|S>VW!l}8um5&_
zw66b?;lxMpJQOZly6s|KgzRmEuFP{6KSU_W`4`}C{)5z!^o6}%2=Db#mVV>u#Ab%A
zrgeQwI$56_B3Y>BD<8aKlA=V_iU*atbfiwMu-66VnU{j}>*S@p$o%LZ3V|KEz*PMa
zQ$k@?+ZMWRf!0EWo`XNdE)q~kqI_2?TCPaK$mdx+T!n$y`Dm({frJf3Dzw!&fIVUl
zBg)K%@r6CHMRlPJ8zsXR1SdA_@ZqoBtDS1GV=qa=qvnrrXYtMzouSj%y#c03#eBbI
zim;hnn)I~hLjH=4aD{)o7LJZ9wfrNwKxTPnO3v}UIq0%ENkJOSnfUXafijfLI^;{z
z8q00$!Ca6=BmH7%c)fNBznyLn2Q&k^)-Xq!8NiWWkH;`0;(QIV*g&a}BeQOqTAe-%
z<#Xg1h?30$nJkm)x$3oJm6HgTux8Q}l%tt>8Az_R65mHPYL=^H_+$gO>Q^a$T}wM-
z>kU~radxK&VC==ve(mEk?!6N2G19QK<{hEiE`*#xvmBXhr<Iym=$YJ>bbjdKFoJ<8
z#Jza-A3G=^{z2rA>yhW4P9uK-<%C87lqsw*mh9;#@O-X2&$S=Lgd^uw&YB3_+_gcY
zqsG?6Q4(v!dH50`*r+j4=}VI<33*!`zebuwfo%St@&p$(owBhO5YR)+{f2{R^k-w#
zb-td!T<!Iw04x4T81=KoR`qxkpH|SdZ_cLctZozNQLxewYNhG>12>wA>w}>p-0|b_
zc+AwD`^);6;$Tz|uYiYYO2g1H+dQ96<|44UhArU%QlSX7lkdjamT#_#M_DM6i9+EE
z{L)VIiSx8=7hf{f!hq0qnsq&kHKF@t0e3GkcowZ5b#KdgRy#3-2oq)pe3wX~7_V7A
zX5&(<I1oh@Ql!R+-;~d5>snSg=UK<WN>)HvzX(!nnwUzY8O9#?@-fY#zRZxn5{8-U
zas(J5_nK>R#{OQv)h}MWUe?W+ru%!AgKnV!xX1IlT!weEvuJWlCWIv@HhcGkZy|Id
z7vR@A2M)a!!CG6oir`CgQZGFe=$4mWZ)YnhFJE6M&`U~0<OAzBFsUajkh%93Pc^9J
z^9Ad=-n=P~3eND90g3tPrYSwrr$6sR5S3`V#|!S@R9Aqt%2D9&Y^=XNrX&b?_w;ST
z6j`J7S`dWhUKc+c^(zGA3+IyEHS*M~WQzK-<=svpol@QGSlNa*^goSbU~bVV7)mYI
zzrPxQ>9SF#SNLMRHsY_<tX;n-2@7DopI}|+HGzk*7E{p-%dt{V5Ig<8<A$qbx-5J^
zc8W$7?P6@?4PRt4o8>pRovbmHuiq!4AT|NEzaX#qGK~2G9<%|LV$8P@m*-BUV}#e%
zn=f%D*KKq`?QC?(opP0-M^F}>(2~HKV4I82KXq7P<;>e5<OuI$B@`8;LnJ)NDkvMT
zRwPi2Q=C9*{R7jkGVF$B$x(xWx3x}L?`$E%pk#7Gn2w9Jk?OZ7bNczM3@Okp$mHTW
z>4ga|y?F&@xpx@vHvTG>CIq;zL=7A&;cws;Fn51j+Avln@%BV_`BKKo>e4D9&CPZe
zND^ex!JtenL+;bgA=ve9z<}aMq;^GUntJt#m8SKWL5RzG#!B*#aF}X&Q(Gn`<6e-r
zhMBG!<Anf9G4J#SlDcIjeAj^EcJo|}G}VL84~ONVmiI6t?s2d2v!T^{__SUPnxp1&
z&cu<m5-;3s>Q(-d=2gG#gpX#lY|n!qva?;1VV#xBf!$W^x@HC=ZL7$%BRm~bQ|QB^
z{sM~gjyKOc$EX_$^TN%UsSzFvX=GK77a89D9S~m=+Gb#oxFOj{srp<UjgsABmTR37
zBlA2KB<adrAud@m|2>A>W;~JG4`cFi9}PF(MbBg?8(Qm&4D`2|Fe%>KV$Dsst%l|k
zLrF$6MV3m!5(9yR?pnDM=`W`d$%&{D8Kis9H9GF?wy@Pd%ME7{T5aqqT%NXvPs2|)
z8a?P|Lr>VmMSNkls*bv@RgP0xWi?o91&c1hdsDJ5gPijg>X?1sFQgh2KXyz~nz+sQ
z=HG7A>Cq%BM%=9p`+Zo;zldZWuRy0b>VX^?E+P^E>W|-iQ7)M@Yb&6W;>St$Sa7Aw
zjLl<GR1?Fv^U!-&Rx3zXw^(aY$k13lOo8<^lS%@z>PM$D=k7oIjWCFa;G;Ff&`ojh
z*hlNp08;*bfo(}+-eOez&o;4?#{$24mg|^E#cDiA4Uq`E)@RANm6~qSTop~m9=Dxs
zG9B-+PV2gVwWQyrw@l`3QYRs~$EcCQJuKX^%bw3NFHjcWvx`IGIB>{2^_ACnG}F4Q
zCB_WSNnFw7qdLd;AxMj7MXpfMIVaj-Sqi^c+t-TpXZ;A)GSQOUVqI=Yo^%}J5yiR7
zRKR1$^|zf1+sv_+r>$-#6}RB+`6fG-^`G>PSmvbo$FE;QE6Z2T(qIH7Dtf^i?0XUU
zb1r&88oaow)QqJrP5Q2!jT~cuC!*`tBc-mWxD>@=(~%>2rj579a<STTct1mn%U>>j
z=^<JZWR6!Yc2TEMvwbj3ht<LVtnr<cbJML)7{}jW`h{S9rA*mcs(r9)n;CX0g7axb
z5|^|bd!;!aI5<$!JJ}zKR<V$p7wY_in&~zeh~Q*2%V;EzCAzKy4V#6rVCDpZsgjL(
zg#j1!+mjisa4JVgnEeJX<2Gcnv>r|{*~z>u>}z63VZ=K@;Y13_SLa~NRdLij?4NVP
zo24UIOkfxp`n93T`fg&*Ar5uVyqexxgaBkV1jJ*nr)A+)yb_1-o0|IJiK<JfxsjR(
zy^h@$CFI_)V4C)p+bEQaaL}o{l8oNCY*bP&s3U2YDxX|MO8Cdl{y)Bux%xG<ANscf
zG;$`zk6C<jc<f!b>yJ^Wzt<M39SVEiwZ(=Az<OC~`VY|*YK5_()9~3T?g*DAmzGcV
zfV<4}x>ZJ7`J?66|K3t%Ol)Xy##4@^%UZ^r1Bx!iPcN%NBVn0m<Fx!~-W7Kg<qB4@
znG?|F5BqoLxAU37SAkwd?e_$_R4Mm&UZ_mCFteq-ESXI2z`54Y<M;;IC<~))j|^qA
z(csja?-oAIGuY3xr>{W4V*guvmNMONsC{RCb?5%cH2M*{_w#d(aOGdXBK9xQGxyw;
zyBkz;wDH0Rd$snH5lcjBU2#_K@4o=n9pC@GxA8w~6Sib8+x@%=+w%*9l~cdpLLmq+
zO?3;$U%*dpqq`yPXGZWd43RH%%uw-6_4W~VZRDRcwDK3wi4OY<C@!MhBRxYz7@b_*
z{{k>hwH~+L?$HH`!m+fTuck$BBNQ_uzRRocQ$1T?T@or>35($~NE#a1eMJ{I82+E#
znDn{l!O`Y6z@Fy*SveBvGn@EE1vH_<^5MBB4ptHHqRmI<|N05Urk$PkhR){n<Rj0u
z!^u5BN6UJvQ>59CQtzt@0EsoCN+nJetn>JlIji32MeD@@Tl~%oYXmFzqpy>i3_|81
zG@`u%KKx=42y};-+hb<zWx1K9)(UfrvW+YtH4TCFe26NL+A>#EA3O9a&MSl_4=&$N
zSG+tEwG)Nyys&6?QquFTxI9(+vCL}Wmz9i(Q+%D3*MX^A`$PvQN-`gf?gNx2d|BE8
zDCI90#=Hz<O;;aA`hOng7oQ?(8vq|Xh1=h~_;j1u+QR*;^74cp3Wrz3y!ef1v__)U
zrl|cEz;LT4far~HTh?!+-lUq{1RudxzZG`Z+1YiB?M>9jGQ#ZZnexfMfU97SZrfs8
z#6Hv(xBwN*;YD1V?@v|xKt!~?7F6qZcX=r`$risn*Q)M8A@v6Qv2{rdi@G?1RSF?u
z+IT-&BZeYES9_t9KieIhRoN1>0B8waGRuly-#CBc#~3y%+@_a6)I=^&J6Ykrg^$lr
zJ!yp4oulIPSRD=jd}=Z(I5o;mibD{t%W#TvB(sj<-~JYkR6#TZ>2A%#VdZLJe(M>k
z6M~p?4Jv>;x1mGmlZ)Y-?P+f`8V2=pQc<3FCBJxcU2dOw)C7)^S2fkn3{)A$2Mg2v
zfKFl7GoZG>;Aut)g2K6oybxVgq$#$S@dEf5r8|ECM9e;76@59z<$UH5c#H%R3DxE`
zO{Xuw^dt;*6dC<*yRb4sG%)l)EUno>jk`1gLHQqDFs#hzCMu5j;d+5XEhagVi^G<h
zZhY$o0-8_2LKe3Q-DP2uM0(FI`#zGJ42edoVS#;I*f+XG$6tYUoPd+jpg+1GaE~|w
z?L-u?a;kcUV;A~+W%S?u0iy_iz$1H;6zaCr8sINLMDH8h{rwOE9a=wteC0K-Qp;a`
zVX+TaMaT;%DMvBN#>U2O9bnzY#vm&gVF~EFuJitpVQ{5q6i!Sjw&ygmUG%J5#IfKi
ze3CC0OG>z@u0(Zdd3|!8VoCoLUys0I0rh@kD0tYmkG>uneZaIbI)g_7FasVj{dZ&y
zvJQLuL~oIQyvMZi;cDNhGzLGsu8e;ERPz$BP|->F>?}g=Ur~VY40ZL1k*l1}iDq;s
z`Co$;R?cvu0T}wXZz;DTM?>z1jOiZ*8agO*5&qT3U>}AiOr<C&C0{%vDwXQ7uAu?b
zl)I65oYEvZbF!$oD!YYb$cMh-aYmJx`YWYNV77%l)!3_3iZ>`nvQ5|Q#)@d^!)^t}
z1>B}b@rHUdlAGbR59iV!sxMoy)nigA-mB*3O=sVh;6y0Cj_jCEMmihrHB4Ci1UnX3
zr!`ML|9DM^W#YOAjAo%Jik<}7&_VUnhx3XC8GjvWvu1Cm#FAR85+i~ml;3K@Zct3(
zKj(jOGEjr62g(%mn4>%Hsqd?$hF}RQ)+h39L=Sk@kNu>TS4)tUQ<RnjJh?s6-$H^a
zZ3Iq}9^$~btXd`z`Tmj#0?JP*-d4&@<{gV3V|L82YdLn*ou$>;t<3Hb);XaHUA_#7
zFlX%sujUQX{h6cB%*`(<fL&=KC63y|8q{qe+C}O)>(neHiE>E~%2E*Z)uF;+D~HWC
zG3+_>ZxW&Ec2rCtD>cHPyo8`^Nh>w{M4WtCHHAT0PEJm)4=e!6i05V7dL144g^i3z
zf^IJ+&8;8ExF5k+#bV)b+sThE!<OA}L1MD(@ptC+DPf<pr`QH-t!ZAVj!#;}P)~-*
z@df81U+&F~hQ~R^8y_p2$i`D<v9hYFhg4h0x%O|iJ!6YCID}2NT@(}=di(WrHynhk
zEvk}C<&#7XV-KRzP;N~O5j}i#Kz-&pVn^MZTyro5cK-Axw=Z#8Nq^$^c=+{B_<Nlv
zhZ#SiTG@@+?3rd63RH)?p@4DNSXGVD@g{C@NOJ!t|I1A)Y|d2@<U_hYb}GF7JL^AE
zOE>5FLVR_#LBY_*H8V2qIuN|8S-Q{q;SvoKYtt{@WtrE2wn3wWF5CaMPGKL<LY--J
z+`OX@zkZ>360pQX<ijc)sl6AJ9QNZ~?)UlOrqdk)X4tQe)}x_cSZg0D4+ux5<r1Y!
zx*Vb@p=o+CTTQky-^QtFzFTWplElF5Xs}0bEG>9}PV=?c^tW&;jeQIXvTf5h`+o+^
z8Gdk<SvLDj6l;M9)n$Dz*?XtWN6S<gtKw;UZWDF&T?SQ&1pq5X-9~-h-zNv@fR!Tg
z{JY%gro)OR2TeE>5|NM!g)UMOt<doT-XApb`5E4Ci+TFBJraISf5W$uWsMWSkKr&_
zRoR;8>%o{sP{GXdppjeb!Tt~!PZx&N&{GwT+fJYTsq=#BHZYVn1u=xLmko0i>y7W*
z7Gp{0E)wkOt7@p`VUPXbCU)#*UN+x(cRm=IFbnapiK@r!4!Mz|D#$u+<>hh0i<D?&
zf$rDd*NJ4gEhKXOa9-f?<R0A1)t?Fn0yzxzx$A#(V8*q-YBsNzs#3>3n?*0xH)=jB
zLATzmm`etS-oiawwX$dAXu`c~fS<c(Pmg;cP|lZwmGqNLni<kj0{DX*=IR;7)4YJb
z(ctFCBL%2`O%lR_2o>rEa0ss=7R#gvA^!u^5X!R<cBH23Im#W?BbvW}ugl;50@h8=
z0DeF}?AeQLUSI!;JpWsI=KxXlqe`9k5w%>Fn&#&)ikF_?X}|th_tlLPyl+leT*e1s
z&8K_JFz075mr=PIyyie%6*3N5Oe}EMmpHKc>ZdD+;tXbbuczq*54{nZC)^72Fc>x8
zAWbfEN!T{lPOf9+DdYupg*4IUey+3^uj6=?*Bwfq<^-cl7@Cou2!JbhtNiJ2X?YFH
z6fe_lm(uSm$&?mcALgq#Zd)cbZax`jzm8S^I(=WpdeS4CXnhH!eU+zinv(nJZHCGz
zCDAFP6UVsJaQQfTU-ao_SxfcTgksxoX5I8st!JrgKSL&SVh|7%^%4`XcCe(ChWvV@
zedgs1E~-=u_n~zI>r!@#f5)_vr=}g|mRr2b<enu}Z@<Y^Z;M#Vw5vuerJ6a|QqPT*
zQqs{Z-m^z-%#N>MFpn4I1T`8Vm!TMU2kdDkB6EPT*QNKEEVC&!+v~y6+1EQ8ik(YI
zXkSE2BNz#Run2-9w6sXQYOv=_Eb5}gMqiP^^d}rT)%FrG?_VoVEzKNm6=v0mKa*vH
z`J)TAiLe~S=i41p&w7Trx))GLqEIf}f<hByi3nOFpo$i>BI|YAx3Joo-I}jc2T4!$
zhkbKy8CX8=h3o?*qNd?C!o=CIn`k-3+z5ViW|>$edY<5|tsf21-^O=26S|^Th0>e}
zPI&L#F4f{2CX01k<jk-q77<p&|M3Nt>@PsyF#Q?32YvJ_lj=c13r6I?lLU*r;e5`#
z`V>C6w||Vgovfu($EMhE?RsbL`uLrmj??HZ9<2Q9$AN*{*lQvd5p|rRqT;wh0=R$N
zi@DAJ)@YJR5X{?|jB{aSX>EpzGJ{9P^78tW#PZB_AFI!n-DSCR(OD<6$@I;qZK<~Y
zAq(thY6v5UE0t3W@e>%DM@|clU*^@57Nm_!v{f~Fcc*Q|vbDd}?`h%Nn9m*0z+={}
zm8NJW_hoo<6P%UBlHMpWSW)YG1_%}>`nX1=3^#r)|C$Y-V3BF)4&m*4&jErZhyM&`
zawivV%vW-Aw(Xa!lV0KW=+d@gO$w(WY{FOpIyXI*mrU5h2PplBbXdGM1NweV85G`f
zRM#LMMS}WlOfFj(W)ttQu|W%%aXqwCv|6+X*IkW%G1sEPH<gsa2@Af$gb<BF+4{KX
zkQOB}qL994O;qD^)zkwfA!5TWvMJ*w>blQ0Ro?@7o4@2P()2fTPfeCg>v@^!^poCY
z*nJX+r5W3xk#S7qEyJ}iBdVLW2eURc2r7O2r(01vFCw*jAK5!f2<yTcx_#mA?7zZ`
zel?BEy8}(fi(;_MlCmX0dvO}FpSI;ai^ffBl|-p--cGWVSySWQR2Pgc3ABmbp!>B;
zHx+TqXD$`>mFnG)Q>PeaPiM2`hiK%^7;~uQPnq!Nc*#uu9yC!$AA_d4qK(y!;n@2&
zeq=YV%eA*awoU8B;Ob(Q>=Kgf)oaJGWzbFlE$}cZVx@`K&R0ya%$=M4fk-Di7Yrgi
zys4ePV7|!A+)#40y!yB#pu!9+^N+uB(qd;?8t!qS(KRsJR}~&JO?OdO2Cq(D%!{Le
zDmQ)`vvG)}dc!4WB&l;myO}o}xArmg7L(q<u~NA7pPu-CQIIB^Tr<8Qa^qQT(RG)8
zu%tJQZ}Tvc<ED^EsMDBa%)rj!zR5+dxEYq32EL5+`c~2lNmg7`_f=;oUKYrC+cvJA
zVPCp!C>bd71qpSAO;by4eoIE~D%lW}Y1Zjjl;{>$u{;jPCzwurpfyjyA<&N{et~Bg
z-pH6cUbG36(knZL4j;dN%)J-;U%hz&@>+k<WYZcgCNS@}Gn3^eR+!xnC!Om^d8aWu
z1_To7VW#33$S7c@C`u4S@G6XxR^AgU{+sznd27Am@mYPOj`_k`F7)fPt>E~SMf&<<
z8}UHivleFS#5%Y1FHHCm-$ybigtqnLU&oT@d*T#|r4e7J*{{m*CdKK=aSfS;LrNxm
zW?egDUTiEFs`5-cHk{pRzqKq>&5EhRG4P31#S7osMEf*aX$_Bs`CS^tFnyRUg29W7
zXBF>ud>-lxj>TeehQiELeBvn`b$MT|8giMM4M|+xCVvj{2n;*_rkY1h77in%EzxMq
z$md|_Te9PteyOg+<U|M$Ggpca4vKSER&;{EgMC-~rTamUDaA4m)KI=1i7t|Q%-$*v
zUB7qyVkL{QcJ2`D&-G;oq5EZjB9sfV3l^n3Ia;*m)Z&V^TQ6v131jQ)TAcMf^@*KG
z{#OS9fM5375e?Ln4=Q66AFGXvsr?wmC>NK*UCbz(!ayf2Ep2B0yUaOt(;29{4dE|G
zXbzN1H8C<;gh!+@T12Fx$Zt00Y!)>SV}JnlEa9(AcIhRcAGyta6~Bcq=c_SXZdq>`
zwt=y=W9NnS<<t^0+cDxK%Va|R@|L_BRlZkGUZs{EHZ_)O<GB(KCNplcENlglaLtul
zZ%=D3m84Q4P=9LZb*n?LF5g6+wa(yv9k+rS-3aq#@bG>G-j}rXd(<t;F<pRT2jC@f
ztn&*aP>sfc)#r+6ETeZX+op{$2QQ2`Gp;;^0je$KEAi_9=`)0{07-EF=!$6gB|??8
zq9<z0Eeq=Yo<VTv=!IShB=#R?0-Dl!v#;bo^-t30%?fCrrs4Z+$w;Jqz7~LtEw@K8
zdH1_wk}RuxvAmNK=#&vV?$Wc()%z*S85F3Vbm4+uJm6Dw?bG{L2wd*q8Q@i}NDTl}
z+6SY<6}nZKapiU&v5mk!WN`uh>r+$WY4BU#br#lBl6QO+FdCieOCW+GtJxy()QCg?
z<j=o8IfITBVZQa+>kbPf8vIoL;B=6={wD+>R1XMGD_i{pobxB@Lpd;ieo4$EF&1s~
zo>-%`w9!bURY=~N7R+_&FbV4jpr$@HB73%(4MX%oG7*sP{nUf<#8>92#js6%+o^n9
z&wVp7mJvS@t@+bU7w2zw?$n!j7>7b%ij6l*rA%&`8I!6RVr@WUiRg<>A6)Z?_*o^B
z!I)DG3bf^sPz7k@r~!a(dhKFm2kLjxIOpxa6PoO5vX0M(J!xu07}V~@p%_!!_9P5>
zV&zeMyX^T<oLt{*fG%NQ<!6Hx53ao0`zM6CYI8AsRvU5M)4(PekG1O_)T-51dRt|&
zq?ssCgrEd!CA3ZmjRvmPwKT~3ep6EcE2J;85H2W-y@syrU9|a;V$=HSvpekfw4A+a
zm;!&Ckt#E2wl3|J12PP`wkyV}B}|zZhH8Z9{sL9Yaek4hwWl~Pp!0pQ>^D{l-Zc8w
zM!qxhs1RkM4F-a*c%8zFvGkq4$^+GS9P$z@QzZ#W6wK=#!4tf7&BY9Kax54)EI!lr
zyEp{)u)T*zg&+C4ry{;N-z)#ygGhiI+0=nB8Xv`2R<?6J4cgx#A!cGI3PKz>YU2pW
z4k1(iZxab~Lxd!N{-636|2B2rENfZ)BO4ssQa*_^Gu~o3gIt!Ye#DOK@t^=$7MZ*Z
zp!YRs%ef2CYafxBx|%h7k5n`on2hdzS2XpYfb{RzfAh}(1L9O;&(`%@n>&bPu2^J{
z9ML8N&u;^W#K?5~%e=&hN9X?no;AI-p7fgA*D%idM&*5oj<rCi%HPQ+$$lbG%y{0C
z*CU%X7I2)RcTf+dR>#!(3E}-!k$2Kk<6%6-oC;}$1?B>rGfS_8W~YKvoN`P12d$iF
z1=JMkz(ae<93Yv6AJAx?xA9KqqUKT=KRukaW!vqSh<~?2@P{^&?Mm)a#rHV2)oH%)
za9g>8hM@0uPG7@5`w8(oq6)Zq_|TBO_=&(62!7(ciPkJ7HwP|s_=pwf9Gj5Hdv*sA
zg5=P#S!!dG_8Bh7)M{zCTU`L|*wli4PVs5*wbEC;Xgq82j}z5rVkv>iPdJ(gi8@YI
zUPqRQU%4bHSfv*wh@_8RFi94jY^pj&0r9@08UmX>Ks>!WmP7j?j1qUFTVh?;+p}jh
z^+Hl{(QT!ZEvZ9C&1GHpVr%{gA@v^C%W1^6!uH-VNtYO``k3=0;Dn8{jY>#BF`!w7
z`V88fayz<144fbFaB%V)+{gY(`p?)<W)g)U5LO0dnMMb$$^N(UK>&=zTVifG16Z{)
z=k=BoDG-gNGt>5K4scaOZ&)dR+G#%NHNk6p`fEkm(1+3ngKs)9$QO;nGk6B`!XtL3
zm2V)4UQzm;`oO+)ta!CO!vY}V%#uj4i}8r8eIr@!<y6V?a4c=Hl@;xa8nN=0dD;b^
zqv^m}qCj^r8-&8YYJ=lq&gx!TuM$Or(F9DH=SKeUEAevtW4umGzFq%cz`Gek`VPl8
zpPTNjNoaex;R7zkJ_+Q&)5i8Jw$QADe{;T_zItxMZof#W_e0?A`7Xa1v~P6nV-dSo
z0Q@W|-foc?po0kg*A&I!o{xrkGb)MMiYYw}mFoiy@0sJuFvd1X(%ngo78(}f)?MX7
zJ17h3I@1vZDTQXd;x@d%arwq^)2$V3cDFr7mpIo4x2rBDKg<wE-K)?c8ri~W2IX;G
ztq3`*ns~){r4n&|d`O~*Sq!)5>EH!dP{VlEAz_ygPixU^>0`s!p9hwBttTTU`O39b
z6sN6_y40W)URy1L=##iud-bwL;4Nmce4eOUv_g6TB-Z+EHd4yU3&F<%9R!lq><E!n
zNT9ikj7LGrcG1OI^W3eg_2{+%K>jL<Y;-rFW%<D282t=b9YEHj#Kc7(cPll<2tmIR
zNRnpH)OVf?te@ev5Xh|N+G<f2cNlJr)5wqyBMz)7{LbX9J1{pZE!i9#f=Os-;#eK;
z*vwe=LeNFll_aXw;QNSyy^Oe3h5Qf0!h!I`ZgZ!1iNxDsptgwwRKrJJ>>Y=#tfw?L
z=%b`*2Qrb!OVm44uC?dtWZAy}ta5hL-9HGpiYOZ<spZiu<61{qo+de#tJ-v!^=DYQ
zsG#CcsCFHVSOXKmBUNh&`Y(8dHRiz6w3BusW6wr38i6!8<Wf>#bLdf80Yl}MSI@ch
zDGB`iepSSRw4ya3-MiuAEO&Fxav)u!YNvvF7$Hl`(fv^~<`;oF%f<RdFA(Qblyrn@
zW9}y#Jpv}bG&OzV%kbZn`6G#0`q3uB3C|zqQo3Q7v&KE4!~0S5QvJ3ZDH6H;zBGB>
z&HT+-IoTpVM1D!ib#{UYF3W`qh<%}?G9;n5G88y|`rGC!_|?d<wqJ5h&7UGKMLpk5
z|I&8X3Nq-j`G8>qB3`;SNb^=JkjXc?@jd#&S$CiY>eV=6#H`Y*kaTM}nu&I{dT&Fp
zx^?)}RI^dLf}Cl9<$cSNq*pyRtn^9WkEXCFX7AQl$9lEWHsze$z9&5;47^00XqQN(
z3{^?AuJhFV%6p-CIQb2GOt0(Lv!k@9P@~0Pz$`jj2lmLMagys{RMvxX^+@utdJ@DA
zck<v9PhhyK%a;tc<)`>?!XUfeG&g?S)Zyau_4zuIY1ZQ+i6AtW`GkA2h8W{kp509`
zzd@;pcZMEj+S6CkRnNKomQN~vkPufD7&cn>4ck(?pQR_+IyBxUk@8VPU~dX7S>rEY
zyCzP-haGhTrQU{jCfKt+a|e~n!{cF+{rsUPl$+^A!S$JUvzf7ZnyoQyv2!&ns`gMD
zY=G7K!>Q-zgW^yKN$Hf8)mcF%7xI<yyw&%_$QE0tAzs`?7c1VQ(a(d^tvXG+cqwO|
zmx9#~HTml(lQKnB&!%dH`b1LF<ezi7xE0MWQxnWAzWRx!yhlXZl2qE~B?3gxDy#xT
zRT0TeHEgpYLWfR$DKkN6QoY(8HYb=DNYpR6@oGNk@q(m%n|-x(ZSVT)ZEgN+Hb7t>
zWUD0dr#5|9Lvpn#vsQkbSRFq1EFmH;7$CB$GQr7LRaw|}>7te7t;Z3m!}Xw2r&z7J
zgL-ngVn;n9#X@SL=BWkpabtDTECDHAV5f&_t{?$hBZZlY=x3oWTjo>0>h|yN@{{#V
z2Q4Uc=gfZ+0-f%f4>-@1$#l5Yzr`%Zg!}U2R=Ek#&z8gu@!&4z)}H**Jh&nOQM@2e
z6A+X~es#yt`xo$1nMZr%r-xq1JYjDsbKm0Ub{0KLOi;=7;xLPaKJ2_uE3~(*Tm3Yi
z^von4m&;Yc{N{kC@T;nogRWPxVZ362RO$o3U8`Ffwgn6I7Cj}UWZ8FtBrrE^W8$zm
zNA}BvlY=8Q7ZEsmnw|DxDjq+TQc_C<QH^rbDq$+ka6kw4_cxh~?}swo9YbDnl%Hg;
zUA3m!3e~Z|7*84xa6TrTOxqxsfG=zh#PdadH(cH-FKa~f9S4eoSxsrFHzU44P3ov!
zdkH=wUrH_zWR$YZ-z4aN8+k>qkAO=Ol1Dqpci*G9dA;+;<q&pB@jGQd4n!%&m|`Lj
zEZ@8lL5@_%_RErU$~MS+L=$GsRSSJp0B34pI-K8$|D@wLcy$tcNZDW)32M!OtDo4?
z_e3mr1exopyjN_nFS&wCALg2LPba&Xr%6{V9*_Mx%wsJGw`|-i>(u`_Bu1WaI`g&V
zc-XdO^t1Ia^R@8>_BUyZ__P52Sl%RG$?qBxpHrR(dx*4=vD#h+^~wgND{R3zt>0>w
z%sMUE0QVAZ@++tI##x5t3ps#ZWKs>_Ev}cxXr9BaNi(5sT$BaeE+Z#M%@#BF`J8nK
za{DJa1na0~p=j*lRCd_qpyN)Lr)uT*>j)|qx$uorzM}JFFRq>tj0In?jt*_#OdRrg
z)2Uy37Dvs&lkW#a$W|~N>ERPVl*;~1A$4J#B1F4gWFa?$?M;o$_u&~*G7HQmAAabT
z((h#kz1W!vbzH$(ef2<B<q(-jMFQ%X3f-tn%<4yTzE?tAqv?l^3t=hn5BO}lzH+T$
zKxSrGpHW$w6W*Wcg6WcLsSuD3@Dsyp?)+()o)#^?)Ap1CzoG_LrmcM|QJT545V5DA
zdYjGTpcws#twM%cGjWY!+HN@o(wTff{QB&xk6H*Bl5lrYrX_kJlIBpJ=3lvhO!Z&V
z1?^~*a<$T*1GoiGXRDu_;VF|t&lV}!GIpK4LXCxQC>n@u(|>MLt?|A$1QST(-@KLM
zi_cqlmtsRMj2l&5ER{FP_wsB4BHcl$cxqTJQMK~TO_zfUlj{o@hK}nT{Zl(yyF^2$
zV&yi)b`dI#ra$)WWI>CCFv8Y^y!AzuGSrUnSkGcLRzA0Q0WBO3blv{5nOm++EVB?1
zck)WlEHNVJrrf+5w>i#`Ww~NOGm~<S<t}XcQ94$sV8VRrnCW2oX*SuLjx@=j^h}k3
zXCVUUDrZsbqGBTPgJ)2Q_O~EyWnb><^M(@J`YQ(8+zuu%M%64<<sSF2M2e$cN?@Tp
z)K>4lf}no)3T<S=%}`t8vgt{4UZS$t!REZ2K~cq8y*QFJOTI+*D`r;5KX%>!>BVvM
z3b9I=o&hMtP5?-%ICd%L8|O~MKm#+yMxX2NvA_R&18jJpDmRWROo+JVB(@_9>YR6t
zamC=wC+yo6&dV!QQ#V{DZ<MpyX5tzBt1bIJoWFpcwpZy%P8XvUc!J16k(nYTnguGw
zH6=WdNJd4v6oR{Hk_<WZx4*?gh-=&=i{m!S>OXujt}X62^z7kXcs*5zLzwNoo$>=M
zxn>RjFr@RK4H)ZSu{=|z#V|=0s3FK1Vbp}6(N(C(b%olnQ_*G6e$;F)QC~?OjI|EI
zr@eMwJ8Rc3#|SYu64Fum(^^4yJoL;rH`gi~ko2o)q)}DPM#I!4!l8ul>2r}Q#7XhX
zc>dknc;epJ{K6R@6hyMrJARzQzIwldgCE^T6GvQ<=b|?SUhV;crb#T26tsczKSD8T
zH9X01ijJlX7$L!6v>ELl5I&9&D(l-#+G>?D(QkO;iPF-wZ;66kxW=6i#$g7xHX^q-
zvMyA`JR%aAi^n-Tqi1?=+*ET}UF8Pj5aMws7MUyXQIimD3&E>*prrFGzvRXs;+HmY
z3b|_-{Dd=#VcrE#EjP$xylv{&?gK+P)^G&|jiz5lyF^bV)HLCdj6MCu?qzb~yO;bb
z!d=``+P)K6)E~-gb3XU_ZZA%wa|*zY3?2lQY|I<TA%irw4U$8@UR5Q?)HZn93R2UQ
zQkg_i=<7P_))b5+fs;+VRC&<jRo#5T=oy5$;`BXxQ&Q`2_M!Ef4^JwT<bNop$%4!T
z<nbf_Sa1$>A3)4BMuN=Zt4n4RLQAF=d^Mji!>U&+?{E{z_p>|gKGRBfrBe$t8gOnW
zt<@jY;kjOnt~p*{<Q-Mamt;gth>(62u>di)QIEPRs+C4Wg<S`E)mKw<O5s2@m?a>x
z4x_Hg9CVGsS3ZO$rLfQWx3=x%Sv(5Qt3&-B>rzf^!WmDVKZtZ1;5;?nH13jV*pxM&
z%*Qxh;p8r%<9<(>1&u4LaNZJlXwcF0=ZVCrX%x9K#1`Tb2Bm|hak3bVSizQYEJxrX
z;amn%-D_{l?`62NuwV;j!R71JsgUH#O<~rhmSqL$MC#9y*|Q(I-E!(pvz_$qk`HI!
zcH0Tm;*#nVvRaqJysmaJy@Z%Bf9oi>aPoVbqjmS|i68dQa9m(9%>H_(7gwNErZ>*q
zSjTwOm3lGKqsYlw?6jetI}qxe9uxj!_=~#aI{7YwxC;*UeN;yP6dIH(I$+YeZnKAZ
zb%+evKtfj=!CN-Da=12IPy$*VV#^%_X1Efd4YF5fWGCnRohIW(HZc0d@nsE7SskuM
z4mH=G!?r~MF&|5A;<QsI=RL>y>GUDfR~j6U{aVvT8(pAY3eS$sTdS+oT--$QD4L?u
zmMsx;prS|-goj;?Og#(L)8b8)z&<=M{&61kYVdfJh@Eh981zK>rkHPBGsVQ%B}fIr
z2`VtCBZj_jC&^b07+N2*nwqQI=b)OWZBiMTjhU~)5{GxI0txpU4@doAXIH2`q>fM(
zFo~`{$cmP#=OZG*E(OsH!bxF3myL_d;i9D(fCWaBfr`76yeIY>Z|OQ*`UM5W&Wmei
zy6eyIc=$4~v39i*yN=4fRUP}>_gZX~O9wQi54eOY@)Vj4D7<5kpEN7UiF&swY1J>g
z=LmPoffP)w^$R+|{0GS6<HN3{xG-Ua<&2N$;vIG~5uQ_aOs&^t4cvlWK>pNhr3zsU
zJ<JK&&<f~=B21UCd)skJ`Z!GJfS)ym1$qdt31A;}_mCANFFokgbZJ?HU+FL_eB@%{
zMmO1mz^f?f6~SI|ISgekk(IS2d4AlUm<$3e`2z7Dgzv7FdWao5Y`a&JEKm1iBeapi
zk}{c$mz+I{Z=|x-t#e;4i>pA5hiz6B3N*Ckqaaee`6M*kxEt1vKl8t-<&;AVXGgCu
zQAy4aT(}N5xGc65ZH~DZM`UWPu>>%=j)n?Ga_YkdbrFfmz2_}ei(w11d<{CzLpd+p
z@A`Wm)j4o4f|}fY#}SshlWWbNjt7r_S?B-zivOP+VUMO+Xo{NF&DOiW*(H+ZF<BZE
z&xqt3wEL4Ydq#0qs%D{g3A4v%<le`JxJuX1k%jY!{GL=_Q>QR@$Qib3RA7hi%E6xK
zytQ|0q`B>lA1y;a-t%?1KrS(m1Ge95?lnkY4@|EPE6h};^M{^7)M|ItD>uHFeH)G<
zh>bndA!IV#q4`368S=&T{_;<vm6J11Q%g_o{*{}Erm&n7b4~J~TzJ*5KwPeD{!&du
zCXRJ3uuMZzL*kN{a(npr^gn6Ze^BL0`icK!$U^e)-uxuB6jXB>9%YblN$Qz;7_zBZ
z&JYWyr?dzYE8yTPQ;~F1w@PGbC=SH5@0V;;V3eR~^h4}+nzpebSNnewRZEbSzn&>8
zNoT^1Hp*gQ-y5s&0K)iFM$4?k-!$cGBB^Q1CFG0`*5p77d2k{QvPL&Eza9}HavbE3
z72-@^9^Ow{)wL9>80$hc-;2^VPQZiK-OVv0)T$JnB6EuHw~qy)7}#kh7iqzfsnC0r
z(m$5<%jh#kluR%BhOISatoXUQ_R@skj~Q?2Rz7WLRhy3)Pn&S4L#X#k)a?Z#FUo3{
zbb%V>x<EN{agC|6eyx{rlfT&Ej5vcXP_Q6^Dq__`pJQP>hD;Nd^xmQ^FgZ;G;YZW)
znqt5H3uq~iPqg=5Z2n!qH8M^!LCXy(O_`Q){-BimrLZ^YO6fy{8<*w!`9vQuQ(xfp
zWD~1)_2#E=wK}OIT}Dt+lJ(IX5sO{r^s7MXqZ}<3Z@)T?m@^KXP$-HVhq@`9dN6{y
zJGXWlV=B&Z<|dRN%<kYE7P7L5F<W1sAx;7M<_LWYe!GsJ$Xoj{OjcJT{-3DDKkJBW
zY+UGYv}zi}yh`2g!J!?i9+|kYAC6tqPIW`yV%u*aFj?pLGPjClQ8cekV7EC(Dkgys
zRwrOU0EXu=%p}1_q#<%)ECa_pR5(+xP@O63^2Jtnh>y>t@DR3a2TKzvlQTKaPAT)+
z8i><YqhWJM_%A@k8++ezLHaH}I4!X##)vi!V<Nd`#aVp)I~TQJaX0P$4cxpz_=b}|
zsY7Vw@L@E{2H@z_(+d%<ON#8&yV@^{GKeq=lr615J?1Yi9El~;Op7UUdKV+}L6xCU
zC5P(+i8Z5HNewI-^LOV?U=7*E;lnVR%cM82J4x-@El+o~h{4vmE2a?}L~qns%Uu*w
zI55w}-MBDRE>|W4&HU!r)z`UCE2Z>F?q*a?6==iC)j2g@#7>f;Q=Ffls&X-I%qobk
zruHzB5{t3qJt=r+L<7$}c)|}6cVO#6nHNmf7Y_<1q_73O1DhL~y4>S?cV%hS>`gb8
zHy~7OQLL+oA-V>h775iU6UHtOU6mg^DKtk9k4+cz!^A08Gb<oyiX&F-KJJE6eO?)q
zR999;moT2Y?4OlCFC`xVuj2af0iWnUco_eaQxu6|YLsCipB{@Ugh>=^824i-B;5&N
zXDb##!hMG3+f!f?&q7=<sUw=)SZ%(n>c`Oj6Ju?M4HF-JG-h6@Ai~x!A%ytSb*JkR
zzTyTIAvnLJZ04>Zbt6tp`qLU38tS+ESk7kj>C>8x)y}LosgPebap=<o1YA;5BZ!v~
z?2%E+KRg?(bldgIHn<%2c#bF)p0j$s<R^YhT6(@@+H{wiH%!wydyA15<pNzUOXmHA
z8)k_t(*C;=XY}L2Q4ST=S9TS_+@67zTbdNCou|a=zt<R1D1Hxc?6cW^W6mg?+}hSd
z<MLO~wjv<LX%r^FO5YV})h{xgx3an)(RDMhs-tIq(-gTmOyEq#qx&z>Ga~M)Do5YK
z<Y(1q7d#VpswL|28)tmhUcOZfz=nt?2ZEHmx{r~ZuzGs_YfrQ3o{2#Q^MaYQey*B<
zCqCK_8!5GA-Gq%l^-~OtdAF0WM8<fIZdIm}<6z(5TB0ND!?f&$CF>~CH>P_nwB-JU
z$;#Dn5XJrwof7fy_fH}pDCK*YZHzmEaJdg_td|c^;wI{>_(0as!nrKDxr$d-c(_9_
zh5FntA*%LIGDQbc;NV#~SjmJEC!>NaV@c+wEqy8U+l^!x3&9X{KpF{z-eGHZ#waXT
za>dbRb_w#HZ`tZM$r8}LQ6E=fEqj+Yb+f3Z6GdXBdVsqmJa?HGYfm}KBI-xs<^5l%
zd#k86yYF4J-a>(XI0R{N2oM~Kw>ShT!M(VI;K89S6t`jtPLW{2Ed+O`xVt;Wwe;Kk
z_SqMEjI;M1|9yAPd-aa2_abwwxz?I%&SyT)NJ#S(+eF!M<~LM>Ld%#vB^wW|znnVK
zc&a7Hk+wv6B6wOhIXO<hixxrzl}GdFXo(L6o)3KI(XnI)n4bdB2g^={o#yYlLR)6z
z3;e@uu3U0-dz%mW2nVddrQ=PV-X|bNUxk5UH*v@;NGTZzTA%;?e{kCWX^8qOcUQeZ
zh>;UlWqXD?k-~6vu<Y>Al5gnOu@a0E#=Fc2wl5176&A)ouY>FslEXVhT{he6gXCiU
zM)qq$c58r4kRN+F6NQ@q8GfvKB0!3i1~%4j7PV1`!})6B$t<k?P>tp<mg(?BTmxS)
zW`mvlcY_Ze_rH&LE@3X$^?Nos^9hb5r5fY9MIv*EP6}5${@(IGR_Z7f8R=h@j?FS1
zR{7fB3&j3h`*{yua&%Z%u~2H&#ao*r1m`c)*PN4*4pFpBQW&R>2qw|drOX$pG@avN
zTSIU7C;ISuorAJAD)SJDRH2m)=gX%CXX0osZ2o|%{qCf9wDE~?U?{;S5x!Cg1HFpi
zYA*lfssIp1mOW04ON@5LXNEre1M%!j6br$NbCsMAp}uyUGd@CYZVp}?3~aSLatR|C
z$&5jkHP(4-O64opE?oiPYCTL4s(?djKkhgb_n@Yx8DLS5Ht%<v;!zUkA;ONgspSj9
z?}~|und5CKIIpQDy?iSEIkfJNtqR>1ts*6~eqEK$b|Ul|k^3RSZFaK${wR!my923r
z{q|*Dl{GT;^;?dEswJH>enYF~SOxdl8sj`oYG63DSduv|({0Z7(~P$(Pr8PJQ*R9;
z6PQ8gb^2{n->|nyRMpX<Tk{=DP=(Pdv%nfwcVJC|;PfQ;^Z~|`DR17!@b>|mHy$_O
zP+I`rPK;;wy*=*lHg-Z&ltUk%LXnNKJ?wwKg=j;#b8tUPIxP@1hL)-%LO;L@@B}Es
zh)~LYp}$LNDJ-E714x3cA|R*<aWmJ4=&Gq@K}_`DT?K(rCGf>fOz`APUSlfv*-+vD
zm#Ej@?gjU5T=96RuKo@er%Q}+pq}s1D;BSXct~hTYoNf^Ka)Iru1(Ylyb889iVs0L
zQ{sR?AK6~|e6w6EM`&m&Ci`R_Gd8to{ywoOPn7qlU%+^|Xe$QKE&*`Qn<ZPU0sIG)
zNm?*g6iXBI=^H0%{+2CS-NYG5<crTPA%es~9E?3+vL9$T-+59e=-)}~Of)LQ|JW7e
zxV0~;;h_iu4tism*a|%Bmr-&Fhc2Q7bk*{3yXa1H0~M5fA>&P_F`ENro~7ZX7rp)U
zMct9hUmT=EbKJq%1+FDzcM9AYVBE+rA+faZC^^rr(2<LAeLYzhFrV^2Ex`ZdE9uX3
zaZ!Kn=OfS0Gjc!3ETxfy4wu!QMsZ!VDB$leWbz58N~#%F&p)3vY5@QHXbPtTCyx0-
z9;hegy<qFAWq7vcp=<y3@pn#)sHy{A;7Unm@byditFv_Kk!dgi@2*)_BpgS*$aPT$
z2U9)sY6-HYP8~OAyX*CwaxcKp>mtFE=vC|W-+r191c9EOqkwRai-aZLS=g_;)0PU*
zgUe8#*aMp^!tEnm`pt@vvyt`Qi#tr(IN3k$hK}TR%mv3Sk(>>;$MszKUqy#%&IGR8
zV*h+K4>so}8vf>CCh$!qrzi?HM=4ehHO4Rjv~No`PFD&w7|VuG=)KR=G<pme$&Vk+
z?X*Kz8DcD}rdGO_1N~}P1PxS=J!nhQ#y^#urH7DT(1zF<lXc~#$|dXNPwWh-B2^7_
z4+UOWq}V4?0+r1mVXwiCZx!UEKlf&g9>!s{=OjZaci$uJx~fkcesqi-FVh_CL_&A~
zopq_8xJ>mJ-C%^J=1&Y8XQ`2(c`5&$g-`H&ZcrY)g9}u*K+!sLk!jfRx!J;TDB4FB
z9f95=e)Y4+-MAt;8>3J$|Jp3~R|=9vaq&v2h9C~XD3{<M43vJhq4*3-m{9d|_Galw
z&T@OA=2m*I$@g2EJ~q>u%aJt$dEZRK`ko+DlBM`QRQ4v@%fGCEYxJSppX>6RWrxX$
z5O%ARKAt|ca<=~Y!DpVTy}@O%F!ruZH6v3)&x&{8zIX>Y4j=5sl|7@!gb6+mcALHO
zb7)JJw&D{??~wJ+g)oc&!SET*<_U9HwODn@1Y_%O`9Qv^ZT<$qN|p%Ex2aJ2cvCa+
z=~RPl(t?C=fo<Dm9*bFlMw-?LsdR`-iS;C!zxR~3bY_}SOCvLGVJkIw0!g9l<(&4P
zIxOHGwt4L0{>^IK){1w_k-1JDx=f)H<;Edak9OIsD+?ABtuB^B&5_K5j`FJH8bBhP
zTHSf+XtjEp4E2l%W_TA5UY~m9ItQ`UhhdLE@yccDT2srH6r7DC`V(L<t#c~%=3-5_
zEYVu+(Ny8JkG;O&MU$Dv#)4oCq(Uj}-$(W6W_F8l4Wa^=PHN<V`?B;eu@{7mbVh+b
z%7R}HnVa~?6aSFc{8nZuT6G|v8)2XHHEACZF%0i2aRiat#$_!E8G$Ot)ltGDSy!w4
z3xP^V>IEpgOIA~*p{=&N*SFEER-@M5k{sl@{viUo_KRpWq{(OP8!kq!ZcW2(2Mayx
zOS)+Ahsnh}lv1|3{J7du?z}}ui$_kWVI+XR*YdOkd2EL-J42JF`!p^m7CPw1a^@&1
z0)+&$GwrG(MqoC7>1-F>qc-)E2^Mor_`T`P4M+hh)1btUr6X(!I_nE0`cZ^4$aetL
z3Rzp(D^Q?^X0aj5z?S$5gNMj`&XN$Z?k03qAci@qiR27J;~p0FY29!lX(%Vxos%s!
zT4O%Uibg^+QLz1MpOu~OKgIK_4w~fz%!Y-UX%kkrTP+U#mlkNvVwo!L`8V8uN3UvR
z?1n3gC5wipt;9We7uyeeKvkYC1e^}FH|<OYQ-=~%D#K6B9iJ~-6F9lUHgkBBJkkug
zLV_^vbn7VE^XkGMWeZ|0qZ$}`T@pL0vuR0!qtv|2=L&_Lv2VM8L^;ap?=qxuO%<F!
zm@l_~nVGDXw|(JMPEN+_^KIfly@DgM=0nE-qJod~IJnK_`CK=vnRxFZ&qHFo8D^K;
z$sFSZvGhD=bz$2oWTD|zm3~t_Lg{l1<z-Jk%<vIA!h124_g1z;Sv@nETCtO&_CU11
zuQ|nKybeaqkAm3A>z6xFr344{r8`W{zGpAU^Lb7QQivQ^>oNH07SZlHV-S<7yx5jD
zWbv1&%5{ATe^$#L)KFfzT4(DXB~9#fs?xhTZ0j>r@1S?yAed@WiJsh_((JdK)XHit
z9V^wLi7<{(&yvZV{yHv?Cq|=Emq>M*{_*EV>7Q!mu~Ivgk8_91p%fa>*-t<>N?M^8
z)%Ok#@92~DVw53^?qcCMQx<iq5?wrnytIPaFRq%2gNM@=4;lK%o~$xM%IgUj){L@}
zQyqyx11>7}pY`w{FwZ3U@TnmwSFPfYSeU--hjpE!6tQQWK|FhbJ8=(?cW2$W{daEW
zZLnG5%%PN-;Z&*yVG2&Q_Wl>FroprZ2;~{wtRRZ7MKxxD@TnrW0y-i@r(_b8roO5i
zy{k}IK_uE;1q(5Bel7K3;r!7fZvvpH6^*$+wQ-$v1Oop0AE#l$lDYJJpIHwMMzw51
z?R#U99JPGa=p3a`#2jWOt|M|Vh^>~nEf!$m{?`7@u!+s~H*_@F(7Mywiy_00wLC)j
zMdRC=pH~P#3oyNzcnbkKu)lZ{)J(u|+Isoib6`2@1$-M)0sTRD9jTu%kY~oCv-I>a
z{L9M-7!6r*3dd09b?bSTSS`+;tPTP6JkG$%7LsU^Wqbi7qU7WHMh7P;U$x0w&{vLZ
z?afma`==yd_sm5-cS}_~DBhRX<)f$}dU0AD6LyWLH`A1{wqKp1G9<M#zaWo+$QX2A
zS1O6hU<s(uU>>H&x2y;Ivuy~Kq-6g*%YPqn3N{>n7P%vB{LoVRQm?@51^ZGEP7<W_
zZmDvdaLyZ&0{w|nw~hU2P`g3=o?pa}8A%6uw7<1PjW$QPyF<2eIJ{uh6fMsA#BZK9
zZRxu>fXznC(7*Y3u0Mv7lihcDX|s|oN_N7)kOqiay_QXeV4ZE?oBT{>4(qRDYS!BU
zvgy7GO+0f$!%LK9ATpQ<NJsdyHqZQZ;dhFR26<jW9tPaJ*(2>@L$S`3F4fgqJ8ceM
zFv{?sJ<jpnzuPKay#gsJzareBAKWBqe5u8;6yzF3D@|079mSySJ?MN8Qm6F~B+D>K
z#jQ`eQ~<DIRH%t9nVWn8`4OO>(Zo^zY^SsGTFGW?uXX4}SIe_;%H6sy^DP}ae1h33
zdwa2B4Tk|W;K=^v{e|=L6(_TdT_T}p8>Dtq`=ChsZD(9ClUn8OMg4`<p)-=zgHP~~
zj@Wei(eg6MR!^R3FoUKgYigKUdFf51PHR}*3FltKZ}x&}Fs$Gv*6EQyjTeRKtlnJT
zf)x?eTh8|Mx`P!2g&*iw8^D;BPz4dc9^DU>LhZhJ&3io0gpaNm3f4@^6wT{!J}IHZ
zCe&{}X?P+Fs*o>~aZ8veAO*S=a_7mxfw1e}@5g$_a^Ah%nylV#eJGr01J+b#Lk)!d
zI(ruIODBK=Ma-`kbP|bvbu05k-t{mn<th!`Xywi&#XFA8>(6IR8`ZPvBqT-gDHo0a
zFBJC5rdqo@OTse<^F&4quGzdlgP;y*+?iBDO#@MKJ{%Zd5MuoxOPNO`?9D&U-%>$5
zQqlWi;Byj@Gv&hWhYVeF*awHl<t%N*5mi=~=7KjEYzdyqP@2A_RC2KPWEe9_y`%AU
z{(_qqN$NwX$qucO18z8BE|D+Pd}#E6jnmK9PBI&rF_C&ay^uqfx6YAMIT_$pQGWLf
z`=&c&tKnG8(PaQBvs@-a>EUWB)!<f9s&CM};m_M>Es)j2NW|)VJ@-xrl#r3?BY{kE
zBzl~WZ?X?@kEPDGzv*0b8TTG|TRAw<TkOV{glixzvn<w8SqcW->-^C6Ko7MVPJ?C@
zail?TmwyMkg?Clcdhd;%V2{G2*y9*9U)!(w<Ve1P%Y(?YH34R%EWDmyDVW_Xnrx2j
zqD<9b7yPTu>Y&r+Oh+NA(Z$GW0;NUk_e#Ajkn?bA;>AV7vp=*x(iV3*FGS4!CT-2-
zoXhmNnz~C3j5F2Q-#oY9jpDA~EDYgA4@#kR&^jF55zB(_G+7Y`v;0<c^Zfdxv}x#Q
zqnVffmoo>m@ipLgS4X-f4bef?Ymrph<6X*>*rVJDQYhqvw%^;$%51?}$xKU5oPMhs
zIt|aH&nF<2y#*YRp?gtUzmks(TP67(iYI{~wQIeNI#5jP!pC%oBy|*^#oCO3aOyDZ
zM<&a=`aOp|33{wt5Fm^60;SGQY+I#DbtEA2`UQmY1J(5=7oA>LN7;+1^V!3&Yd__n
zRNGm_U(7_GxNm8$EGOv|>v*inB5TCBF}X{l(1;2%1rhk6!>HAeR(BDpz!tS(c!j)e
z&)AEyyLW$R7E38L=k((LbgXCS6_WWAnaTmHTupTNDE7aSvrl*fVYdgQV22>+8qBOk
zLtQDAvV|Nq&aBAmG^OG|XQ%KMHS^Sf3n#{+mKFA~-+_E+Kr0>G$ft;000V0xWHO!R
z$-<8!g*qutSama@e9lXx`p>bCw=6X{8dg!0`|1D}apb4GR+;d0eRhIxWdsjq#YoJ&
zt@e24`ZN;{_vh*_+|N%F4GnU*_DTY00yf7Z+{c|~+xyuyBgLCR^^Gj79$IoB`n*EH
zXWkHkE}-kUItYam_Q*f#GoCL@XtpDO3gS`-e;{BY7YPj`eKj2nFykZ)Mj7gAg_GEj
zz9&2Ut@!<JbX{xKjLEoAu8wopg+acKv#84$5jK%L+*YFT;m_j9a9ilSPJE+=kq_4t
zbulz=1gMc#S<&!76eU|zYL#6`AOv{Ox7{4IT!WS9C#rhEG|k4cwAq|9HH&<q(M9b^
ziQEJN-&QtYTf?+$5a?`h$6f!ZNDj}+l=Jla_q!@})}w`l)EZGTi7M=ve=6c~VR<sB
zaodp+mY>W@h%i@M8$Za0x7k*`_lXotl{y^i<wAJYijBVqYJEjps89USBw1P&g8x2p
zNLzSA5uw7eNxx9Vxzl+l*k3)H$VL(h8chF`s#UkIbvvnym-m0XQT~5ru>4CY+E_nl
zjnJL5c7`a2VRO#k&-?E;KzOPrOAyAdO<@GcWZNOdJ8jGWoZp7a)>J-wOU!r3liNr5
z{>g(8kBtlYcYFV&Tb@Pg_3j^nPZnA}d=mVRbpk2#MFuk$iHzhz_J*2j$-v;2_+Ftm
z0>>sThi5>T>bWRod5CnIz2~?|VAQF>n)iup3$ck@woIQE%rApE6Vv*ucinE0SZomB
z=c((3=`t{XbK3F4ki2Z7q#^!2rmd~fYBX0dN<G~|W6y05Xd^E6APl=Yu@Bb`FQ^Ut
zYa1|L%%#cNYhe7>#6+@=*DEgS8<mJl#aG0`()4ofh%{^gl1&O1t96Pg{OqhpPpfp$
zzXq2O6R~8;D(BP)JMyW1!{NwEIuu0N?ICiq_*4_7p4x&85?8Jd!Tku>sotl8nH%4p
zyBZScG%o$GA87X%n-FS){^1jA&Ag1-eNB<Ts$F#>rD^$C{9#gv!u;Y>&cEbQ`kX$N
zd9+i7F~k#l-J+)H5aM0QfCFHVh{cY><xVZFt-rs>W4?O0FR~-xCIf1^)zKK)-Pios
z%7?nVJu_+0x!YbKd$EQpf`X`_GTBxG_?)gt0Bb$6_uBnA2wyO%1E08aF4JwQa*`8O
zooA5VJm8o1JHwLaDJGxE-MuGmcVpbfN_KZUV4`7wBzyGX@Ec=cLsWw+gV(h%I?BQA
zWAh}b-CrlQNnJv-4apZwACfJpWsb)6`#X4p*P0j^@Vr?ECXEa+C|dT9%b}aJ?ecU|
zgh`n?@MMnfu68yhcrxGg<dOTokLa%deFW}r2|Nq%3NNTm<}Qaa35b1skJjFfR;*oF
zNPKlX-dST<SR&Z3QJM!GPmU}y>TTpihF^Ky%m6Cv#Y+x4dqlS>1%qHVg9zfKo3c(Z
z^^!zXFvYBh@19DI;)y#m&4^|gH`7?I+ph`nt{SuYwVY4339`g^bUjx=n$34l>l&Qc
zM2VbDILT=xEVj8RJ=%-6SAg`(x+>aj!bKP;a<HGIcC)lMoh(Krbg*ZOKg9UwF9H5T
zk>dH+yL}->QrRYcRjbqfcbiLiSi@AEiYH+!>l&A_!h&MOHieu0q&VU2sl3v4lb}7<
zXUW)%$jkr~=om)M5xP5C8mmgj#ysGb%{2AzBNdFn(yY@U@F53tv3$xiy0Ua&H*m?b
znTM=)6~0tf?pzNxD@`5Stjg?S@ZQQBfqBFIQ4TH6N~npU?jj+}Nhb}o0J&XN*ja?K
z!)tEEYO$jSV$(Fo#Bb$dP2p5RbAtrLR_=oXO)Ww1y8=D@K$kTRhRn$#tqw4|SSH@=
zysf-X=`k->@=Fw$En3sYa{40k85)Cl*T%f2!PY*f0|Vj>HwP~&GRT)yD~AK?m;xyS
z+7HMoitWw1smBai!rzBDLPnpgH5nDIYOU%gA06Zk-(b4;CPLxli0q}4Zk)AiRgU3U
za&O?t{FK*Kpg5()Oh{$M&48+L>3$Bw47mmtl=i&scGJJ$uskqPy+oUPYra6s4no87
zY5^(YTS&CgD;a$Dn{(2ai6R<a;(J^F=OdaW2&Zb9Rrp<JQ%cmP!PA336mo3Ko@GS0
zDmkRixN9mkE-Uc~Z)lVj^f)zh>d-jh+4TJMcB5UQxtdgL*Pry*zX&$4R{%F^3$4{V
z4GXn?S_w`nyK8PrS=K-`m`9zpqdTJFGQcj}M-{qdRQR!JVV6Uqn!g1DIWSBH%h4OZ
zXIB^f_TH3uip(DPh#T;5+pX5Kv*>mWzLrk*?#ka#VA&(Z-J<>Qb+OP@Ss=OBYro-5
zp8lj4AI)@k(YDU0od$TMM0S1`(-7j(3AnQcHaAse5N*A%`1^VQTZ#Rlce=HPr|=Ng
zBL3VW08LNMyM^IoBM#*-!=&iR*uXb+D2FD3xJt*1o>!V+rrL`niS$(?tGyhf9%{rv
zgEiS`0EKWw94u0KaPd5hh$3IZJ|g(urLX2!bJBN=$xr=HoOkWViSj#!RyjfoS`y<0
z;d37eMUOCNU|*r9@33ExD}yV+Pa)i(=}zh-BliTLBv$Ko=yK5ee$uPYUnCY%Y*<s&
zM6%~<Cm<2qL#s~><A@s@c=nzDdqVAgsmHKHE%O3af2FH=fWva2S@H7DFj>6UL2S&~
zk^W}1V{d3&zMAv4_;v^zbY1b`=X9CE1i&{LCV#IVNU(@yvsbjN@tkZj%8^gn|2_Us
zwHEkL<)P(b|0^FjBmJA;Ur)i5&$a%gVm@VhwSUP!wzF)l4&g?FOSI6}vM#-KwyGLO
z)^8Vb#`c24eMq12Q1w7^2c7fd9COE|mOYHvD!?kJypb{GA<o<%ayfkJabOkY;jrY3
z<TLK)t^YZq{_D@<D0V95*-ssVCq<_fy!(N*LDO<=NggG@HB2#;d!lZ>(d5C$Ce}nW
zg7Ml3@T1G2P~Pv>VlTphSvg4+OEXO}mJomS++&Vd)iVcw@bjQNs_EU-gL+_Yal#aJ
zky%RPCcsQpiKEe0ShxE6ObrqCM4O|yQGt6AUl0G-^g)ZQd%hdR?LhJ)!%`JYFp9GK
zCtD_h!2P*|KFK(}EOefT!JHQ$Bj0H+UvngH)O$t0v}aN6?R)wku9yGv^JFk2o+c#I
z_;Q`|{oWPCPJ7=*01anj5TM;tlZmA)IHzaed-rmaFR7%YPAI9Q>dM&OwFhai{?=tU
z=((He<rB-wns+6qm|em3e6=d*)$=ANX+vA~!iiM$vJexkGQ_#ROFylnG?m|-=+SQw
zJ1lQTC_14MQ^9ARyM-A{j#95}+Wh43UUB!wNxDdmOioAIeChjuC9r;>A|$qVSjUU2
zLp+l!ZsAsLJQF7Zl-xZ#n}nBYTB>pKtJ{Ofo=WgnmXVIkoXg>WDpP~E-lVrr4X?uw
zq+!e&Vvg-3ab1OTSFKWJ6<f%vG3u^$t(s(8bw%g-Ji+c#jYo&5h(ZhSCra=F<1ptZ
z67MYRVJ5!p!#U^1Cb%(8<Q(gKmvKtUc4!>AccIAc%=DLO0}Z=U{o?=FdwA>f8#-)a
z{}GMZ62laqC6d~5aOv5k<C9J(_)&4{lqFerL@2aaQ%kjUO`I~tcSgq=d)XP7t&!bl
zi?|xQok9W@r5^P%PvU1$3G&5$#f|#r!^>q~ZbdL#pg#s1G(vsM8)4Plx5AhISv+!-
zbPn}SnS^BWK1=LI)t}C~(CDs7geZ^J<;NKLK1PVAN|0_caNWFbQ^p^(GUPb}RJ_6E
z^sw(av6?FALB9B%H3B4B*V|;xQ{U{>Qc8EAM@ko|xSQ`K0-we$T&or`07KFK<n^$*
zu*(-@r&Kjm%<qMglqa_gB(MSvd>&DC%b}C6D|<^9YjCwb6{q^w2%>lk!QN4>ki7Du
z&JdJbq?R_N6ji>oSxScr{}_Z{WcF;v+xR~^SoQ}G5=!{!&8l~ah80U>7gOiOpGwp-
zvPKs>eI?`ST{#R!3eSPc6nki<ts=#I?FGxT==Gt>IWk*k7UC98#D4IvSR$?-Ykwc*
z<+kG>-ohYC@~(x<%@)f%;gz5W&(E|BrkfI2H7jArs#a+OZkAp4Nh4#8b@{!K3YJp<
zcSzfQDS2y7KWc+0zCp?Dw^_*d$3X^cr$eFdg*+@fNKX9WXVC)6Qlh~l#UFA;3pM)?
zLHW{G;;JgSYA4}n2lcmc`)u%P6mm{ImT(WQf%W@1?o$N~ezcWi+fBC}A#nq5I-?30
zq`UTfjT9#DlEAMCnb_f-v6_!^>dJf}%UeApfoRWF3!~-Nwi9}W+2i$A>Hd*+<#MY0
zcj?^`t($x2fydCVCq|wYO9R+zHJ`j7zCF`$3O^@ZlGoww4aw1EPJ1fKxu4b);=y)i
z4C$&FdlSjAM743YuFjSqv;{_>-hL@;`S^vG7#U637090Zmz>N=p^(aOMAB31CWGn|
zIrC^OI>Q+~qU<i-ctxx=+OHSM)Ci%->++UH1G|L3*JVzwOQ3A1g8$3=#()3t{Wx7$
zK}VumI!&^ZwVXe7ZyXi4VpBG5EIF3u5+N5t>6KW5F;E9+2eAPe7<dMrh}qg4TUo9(
z<8?ObI<Lrm-?UIr&l0DXioidttT|qq!bZ!=LSsq!q`Az^ct(~cd}yb64qc2&p|R>i
znuZS=BkCI7fQ@Z(pqE{%a<@LjN#ltV8aLAw&s0HUiY8vF)=*=7|B70;G-dZw4kZ@I
zVa3o~0*50EU{<yt34Kf`Y_C-6rLtu^bhnje?#M>y2f;(KIcC&qT1De#%Kf6%i0B(d
zY*M)q+0$ub;>r3!_++g{&jUK`wR&EhdwfVC$!>SW@jM?39VZtJT_;#PjEGtg3pnrV
z^8vyHkaIdw6kTRk>izh!5fn(Pk^c4?W>Uap^HHjI0%W5Uy-pOKu29qw3}GaOoe#T-
zv!{kK6&e(dABL=!WktO)FSF{^T5!yc0;~h()(5qiYtpp7IW^N273Vgvfw_P+xJ@7z
znB@wGLV6^@d}UDSy&OhwdT%L>M}jGLJwypof$x&cej_Wv)u6FAehRsN;68XuT!Bam
zd1JIQwJ)ioH&+NZpo>{9h4M;{zl{3#(W3F6zmhZ$wL~(<hnA(kJ%<l`bhm7+U+zuC
zCGSPL|9$kkhxE(2w9dbe_Sz+|X}w>R3534Dp=a0El3l5@4~=jYcT$pn!=2tf2)xLj
z?^FHzsD)bV-$$ePCjUM{zlooQUd(5giL|(C7fqA8=j}>DN~Z1jfEV6@8*>1be9fXa
z3dx*fwzUs{I<KyH8>v@UkT@>;PX#0#oC>i&O;b2o2u5swAf5|`N#*s8v5{O8gkF~h
zGN|OYA>$LiSD49!m#=&b?g}IPtvThI=<Ds^;^1$^vGHeO@L<6M+LMwRLuO^gH%eGP
zYToU-*6EH`uOD_Pi6y#}5mUG~zlpxEbQHSD>PT4xbU%oHxjVZ=-bM8U1q;}~Mbl*|
zUnCSZxGPJ<VI46vIglSoj_CxzghCBl7ovC?qRYHXiA$p~0ivpCw{JIsp=fTp&ZP7)
zKW3NYU_cd@yl@2%N%gj+bSEJ*eUAQ>N-NC877$mdr9j|9JJ$+O?5S2P)K_`C<V;;P
zNr@ZCgDZBs#}sS8c<nvzV$;2-PS<4lFz*VLzhyKmp)VR-7N(+zFdnmU$G7VRGT=$D
z9Lx-9>>u-P@Bw-zQbW_#j!0yIcNWLeY{b<+kezz5+qVocOZ;3d=8HhPiP1$d@dt&l
zojC-8yT&xVQ7@jG85LwsqnIo3<13@(p*({+gCAWm)yU)+oT1RBjhvxjH4T(ik!D?j
z?aavNILS;E@0laQ`=NJWi7Bl-n;w-&Ht1qJ9Na^nb@a4jc<jHjZ@li^98g0x;<22R
z-mxG-evZ;JtQPWgmFi9!p$y^zTh+*}lh^fsUo=!1)i%|_PX!43w;;zlLOwQp_crGD
zmG`^WdgP@dOCtq}g@bu=sZR3U`f%JQMGjI1j^Ka^K;)Vs6Ke{5-PVpluvdGlAl>rL
zw?CB_ORx`%>jh~aOEbou^l-iZXt%P&WC~(4Es?^XHRrvjFW#iB67-|I9<!)Cx+^j(
zDq>P%3}VfQVdg2RA*^q5yUb=DLZ4y<%Q#c7|J2=POjmNfn{^vx)+oR>4Bn@7o1S#g
zI*9iBIc#)U=Nk6rX>g*8(01dd<2hib5#&adf+qG1*|=W>uiN35kDkNjAr^4&Y?<~z
zF1V=bxy<gkZ7#6a*80It+u7efUB}w2YV-BfthZRn_-$&Kj+~f{q)gntmRXBjK#iB-
z`oKjDz{@8j+ynT@Zqn>EF17x$5>2b(?I#AWSs-!xL`Me}Fge5bdFdhjitK~de}3+p
zaHc!shDP{d{Vv@aqQdNTH?+y9>k|;svt%jfAlVkFuNKB;5`aGzbg;zRS^uid>W90l
zXmRf4?Z#pHYgyUWqv!eaGz7IxYN`(?L*G}L%~d+^2Qd@;jia@Yr9aFatT#3T=1d)V
zUFe%3$+3ey5rG3q%<mD~I~8G$hZ{pf<yDX;T_D9xJO@_ydNoA*h<P^3l$mUr5VkNX
z=I{h~Y+qL*lQAb0m$%6s#+5}tCu4ymk#FsGHBqZbkgY6fIGlc}4GYIw4CF~{c?cnb
z1x-l{L#JvIm2B_i;+^>KDoT1(FUwka^Xc-{vp{$@CsPiNlyt#k`o5exS~XV|>7Jyd
z#CC==p!?a7stQ4)Lm+`BcR!O-riF35ZJfDWDrN5^g?pE_CJvimE^;q+lY3xC9r^rc
zWN+sHN6GY)qU^ao2jOd9Jaf<Y#|$+gvlvrWMrIJR0W0iz7b}ckt9I^nb)3pkY$h@N
zn)f$84Cg~vxha+@r-*AWCgkTU=T`DDE6bi6pVk4$^y-&I2^zy6{Qs3D^N?$U_|G){
zl&z`yeF`}yX^b0bZ2ZtYz3gH+Z9*LLA1Df7sp+A^Ilk)PV!)`*d;jnI2Tp<OkNnQ4
zQ9{o2N#e2nAq~sF!U;#c8_(W85f{E7x4j@wYX#(5c7L7nX=I38GM7DyE7fgyfBbi$
zEsZW$`C}gn;<$I3+f|1qB-*92DUyo^pt%tVDX~ZOL%@;vpHznA1W!TW{FvVh1js|8
zo^6+!IOV=A0`Woz6JBmK-X2oSNAA#8ZC>jJUnvMb9B%vFSLn0YL3Qd-{wl{0MFs|O
zIXJo4UE?1?Pix1=Q86#8Y!0n$1+|^oiT3Llg5Hyhv~^h*XZm<btNJ4yH1*sgh>9mw
zc47*_GV3RSmXmh2kWcZ7gcMU<0|G{v0mf!87Sb1fU8?uVoOM@$S}izGT$6ThRW-r%
z)lNw6T826wj#)d_0q7s8R$%Hkt?eMLl&KD{K9A8@rBu6zD}fc0>rGz5=>x6y$@aqh
zbp_!kYKtdS?NdLE{;3xLjH#n1k`lEQfh*Q^VKQ;zuOXn)k4tX^o7|4}Z=Nk>q+Q>z
zbW*c?vD~_Ow&;($dHs`Xw59LOm)tLhcw8sf`MtF5!XN&h+19kOlMD@3CoIG?eN-4>
zjJ$2CUZc-;+`D<AJS4kudAUbO2xvIlFIizYj2J$l7~%TgI?11%Mn>@_z5TYx-&<ea
z=~;E-YSXM(z9g_06K>iV<w%u+AA%}Ebxl<2ou7wu-fN=NO6T|E!Od$1jsw{WL3%%j
za+)u2QHq63yUZ_|8lIm9g$cLLr6*AN_tor8rfR%8DI=&jBR_g|!c-J!W7_>@(QV)|
z@4idhG(<NyaTN58sQ;(izGZ)6kCO)5i8hi(WANLZ(AWR%|NZ~NK)H8n1nB(M!;SEv
z)BAi5H^KXO{1AQtzj7fgwEz7_`(km}hV>%K`+wL{sP8ncYNQV>`=UGGnEo5%cj#d7
zjr#rZ#9uC+*BN@V5)#0d);bs-&)@r>8K!8zW{AAC#`pioTkrg9nH$Q3;qUW`>VNw3
zFVtck$4jx`gnW;>yd{F&1b%6aDYcZXL`8QU(5wf%|ML)Mu>!^==WI0B#$#xy=5UCw
z>nuB3JbK43uPNuLZ4h7zey8Iq&b$1VXKYpSqQ&Z{Ne*hpW+sCxzKxDysSt?SR+iC^
z<hXT<T)C55KcK5U=g^q<N5L`W30uVx2g`4slObLq?V?WsPR%VUSbJZYS~ejSqC#-O
z;3ZCYF*6eu=PD;<Y~j{*7({Z7=5M0aAgEm<vUtCMj$_F_Jf3f;1AMrKB(+En8N0hY
zuKNAl#NoC}pO#U&Y)P}8xFGGYMZ3{1VUFoZ1bbJ#nO*iK_a0nPeSsfKAj%1nqGE|N
zx6SbN55pvhuC782hV*@yk4q<E>s}3xl-zM&=lfS2@*-s2Q;=_v<izBjcY~j-Fy;cY
zjz)K&N<}S9(GOsGQY%;_hH1_Xfx-atW$a`{P%jjFAh$TN7CrhV(MRhA>WEV}b;{Mb
zX<3=sv)UEt>s6a|I3JDkI-RJvQ$5(jomHXB#^ZkALrzKIc!BGY5JlxdzhSp7WYI25
z)2>XXfSEWohM9xEe%)5OgiZgB%iqY+Xtscta#h)Y$^1a^3OP;^U&CdOncMxMGUlg!
zVfRfzC`__qS4<W>9{t@{RF?i|-{m6tilcfZ(VQm(?;<mRpNtf=-kJ8@Sa7JhT_GVK
z>RNbZjm6K69B+go5dS0~Qj!rNr2!_Q@Y+Ssgie!yfj}QgKE;GInKtUS&G~1J_9H9C
zoC}WA^d_~5f?5~My=C(GiGt>lzZi!Ri%(Ybrgy5Y4D>XpPvk6!{dfhqI*>a1o`PTM
z+zk#YYj@+1sPb8oKVx=ADTVCdDHq8OV{EbJC(AB|w*hM%58*)`)Omssp2_(z&+zsB
zyRsf@YQ*BEYV)6}bCI{T^B_;3P@k84<iWxc7#P{{uZi*<ip-D%JXR}`0Xf;Kkh)l<
z&eZs&1~<zlB6Hpu)U*{*7Uxlleg?#2%1Yj#&9#LlqtVsm2^XjZwmJNQi@(k_rh%24
zsJ(Wdf6^q`&1|*ZUapq%$8x=9A&22bYVcgK=(EXnw!8(wL-%K)RZ*4Ef3|=;ajOlK
zw!m{ORl&0bygd_&mY{1}NM*syw=eqca1+APP(h;*<+}YJY+A4D_X^ov@Nt(yIJY*m
zfDrQRPo3He;+u3ph~^rOu3XXS?_NNvr*%SAY}Es_<RXD_un=3D-vjAw9u=S_?dnXX
z%t&zrvbo+Pf`(PR0O4U7OLrF?tyDr5!LC{yNt=M5qbaeH9PghBuw*w+$q~<i$Z14x
zv7zV2US7=e7v%ttYpZbjPQSRjJR>?YW)aqim?yx$#%>Zc2Hi&MiYa=Y)*JZS?A7{O
z12X9Onp<D}CW0oVRM`?pK5y7RMG4`>Ebk_E$JXoQ=1;6Y#?!MO6eolL9+na%h5mrQ
zKSE}yG1fEiEV_@`#^l)Z!G@Xv)|oJRSxJ1uc3;)b$flkm;?j<~GBjDNMiXzhJ>yl1
z`tl{y$|CBMU$(g2IWv8;cZa@)zmEiihYs)TO3=n_mE-0%jeGD~nn%(5oYX<-VhE>C
zcyUmJc4CO2miNz#K0{?pMBcxT-g=@zYstd;M?+>axe5ZDqQTsomG<*jT6a$iXo*o#
zQmwF(J6_k0@F^9Mqk$k*(x?#KG9s*kirI4d8@v<MDxD@<Eze)5-lHZK0v?jRgx)d+
zud{4Lc!0CIP6HZM)-Xb)J}1h1scV8j0gkFFzaR|a#u^IC_s{h=$-1*N3$=<Y)xRr-
zL4rhpMpn{9>vbXjtb-Rv&PO66!~X~cE*`hAioUPBi$oNaqn}rf2VL)UdN-zuaAH-f
zcdmaV<turGPo&WB5aHI_XUFnf?!D=bs0)SA>}sminG-Z2238Rhnkr48&JBWrCB5iY
zWwDkLLyqQ?82l1E`KC!4!NeHnC4LdvQ77u2Nissn&N#DxpjwQQ)LPr3O{2RKNvA{N
zM9MQ67f*ZDIv3*uB+q4R@Bi(wIBIR}KT^E4HZ_AM&@m|6(@?Zq1)`Sj5_?}fVy$r)
zoIP9#vr}a;MBF#-|0v+GWM<;UQiex>1n96}uWHl{K@Hma?^nOP`vma}Zk>uH&M{iD
zQ9q)XC`NA8*M*AHn_K&8Dc{5)=+!W<T|`GxkzC@#%`iPB_}r}_C5&>e<e{-n#Os?|
zNaS9MQoefr{+R}U@1%a(t2oCH;E!e8XN_m8JcfPlomDir08~bJeIZ=4=_>E-_Vp3P
z|CO!c{Ni5l{Q=dc4-|KU+8=BEw*w-a^(Z<5bh*OHM)tx(43gynKbW1?fSxnG;pBx1
zE=wC^YY)XQ8Mq?CUW})dyhyZrntOXYGdkd4+(Z3-rsegs%Ye<7Wa!xW94p=d%)uY3
z&K>^!$s&P{BbsS1>15Tyi8<%_fW_#cjC1`-z^a+q+YpF`O!d1sB@+)xqh|LP-3?|1
z?nb*Szt@j=CyxEwI`&WhebkRII(gBcmNQ!okeN5zZE!fCs@v1FIc4}@WvG!iY3K5}
zHoy(>-9Im9VnXg^(y`q!8qWb1hLtWgobfO-gomUI5~ZQF4`2HiYB(kL9O#&d7-foN
z{K@w<3XY{o<Qx~Afz^)Ie`jJMx*T4$oI}<BX||#e!%GDX)$gUM+@I+`=;DUM_P@f$
z35|b_?LSwOj^n9Ie=o7q8NZf$!`p8JrutrKS+aJGGxlR9(@~pjwM58*z;}D>fSoy3
z0`!_$Mn6ho;%6I@X6CBq=e|{EK{>z=Ns@cSq*52pptoaL80I_q-e7{p;@HILsI%^&
zfEiurJSzzE&%Gat;qXD})Y!m*0W(k<$DN84e~Vr&ML(XAIBA*n@CBP6b8bqn!k{M2
z`)4n(hy@Pzx#u+<Vl(>}$EH@ld@wl9loV6}Nq!!>Q)xCHf>g5&m(|Gw-wc9iAAs<r
zP(4yWr`Ibxk9xR3y#refS&+;SxnJY08it%e&bczMqKuiC&j+?G%|m(Ck#=HsnA9aG
z9s`vrTE!=S!#as;D8D+E*I;cx6jq-#mdJTSGDD>6!VZ5)EZn^qH;0#FAC`E%?MOi7
zLF&C^Eh<q53)zh2*i54m91qRz&yy(S57j;-4Jyw&uH1$r&FcmkQ}Am96KiWH4Dz!_
zEG*=JpkJHkl^)i0!F9ARFW;ITG1wpVj5XOuNFBLIKpn2}8e9zc^%p{&QLB@|4*VS#
z?|F@FbA;8z6r{!iG{Siqe^Uk=X~~?dpe;9Rml~BCu=yF#>Td1!G_1wt;GAnFXU5HC
z>maDHf-Yh$pZ#eB$>EMee0=eK@JftHwm}I0NYGsiY{Eu}hL*$6px|Q+P{vGOpA&pM
zlYFy@=?&hFcbB)X2^han3RZNPCE!;$5(U~15XF-*+6=v$S|6kJQ|YnI8<9iR?~Nre
zMk#3rb&$K1cJ--ifPA<b4m0jzFq`Z+mEg|?lh|3lDkoqFEK%F2!Ki4ZSk!ig9j(43
zxHZN+#VU8xc>Z#*Zw8%EUA>xufs4zP?jQ8u_ov@!4A_YJ{}*q%|6_Z2^QjZ?y<a8B
zXVP}Klv66%#<1oU?+-<R?qu93L1mp0u+k5x<>knlqAf&uOeQzkPWrNZp&DmU_CtJf
za4aehtyyRrs#-6|@#<`ZPBwVcE~og~!Bx9I!m>}@RTAO#JBUG;h$^uxgq{0RSg__G
z;7Xv+w=_Hyo0$=<=Gv3;XKX8PJ|yOP+{MhosyHGV2q=-ln4x%f?VCP*9i^d-QqexJ
zjc4rrxZenc%Tz)575#?(d}u=WWuB`YZ0d9a#75J}U$5y><m2|Xp%n(F%m#TO-d@LB
zD+2K~!UceetZae)^2FY-TP<|*B+VQ1Kc=D1#WjLeMk$%GzClv#sC8)sl#G#rb;q%h
z^`B+;Qrp|&hEe6JWlGQ11Lcyttxi0SD$xGd)$Xrw??Zst?lvxqf^&B2Q7k&^^P+2x
zJmYV)s`>iX%gdl~|0q`5;PMv<&IrPvj0>H&5LS3t7$r%1s)8*a-;9)|BaD(xS|Mic
zP;S|D7FdekUNc+fU;N27k*zDZSK>^7qqo9`)w{nN+90mIFRCSEq^7K`Qdu!rPL3~@
zahY}7{Oat@N!KOn-$zR)xn+ODSR}_Jh5i6OzrD(l{`b);#TTi*jb(P0fyc)aF%KfZ
ztGg#Juh;G#DE5APjy)umW#Th;iT}svM^~)=$=>!yzizj-JaCu9X4j%AO_`$D&bzTm
zK@PK1a9Wuf?*4*_`~?S6!KbPvbvy8FTCbro%Q*c-FPo2SvVS9*lFsvSn)_0?#vG75
zoz*V~O_R>*A}H1YosDSoZkQSRc`S>a#)&B}W5?>pkQaF3mQj?ty`Pe<_G-KhqIU?4
z=%x#FT@nd*`nIIv@-!?n3H68FkuJ3)uQ*+5Lbus7YUOJ8wt-9&NB)lvel{GvA*$oK
zUAaPnP2?V$p?yrR;p|09Z)DvB^ta-6+{8T{g&vdM;KGz((0m2e!d?H&7ScZqA@^5r
zcAma@K!*T}58r)J<q^Z_cOXTCA{}8$SxXSC;iSP{SBBpcy)vj7FO8hhlt)#0NY$j0
zYh_wDaV7#>`oLHS{Nc+w`4Zz$+&P)8q-5e_$~Zqss&4dn7hzeBetKN>j1~2<g9;b}
zmDQ(`nT^%c3?I@{xkuNe9CRt5wG0_16;2+c45QX%+-DtS$h*bacUzR|9ZE`k6tAZ$
z!mX07W8&gdB4@ohnJP%(S?UE47&3P(t4OcxitRlQgWf}o#1C%r5e_M@rEMQ;-u{Rh
z(3mL*BVz!Rr6g$|xJ@i*`3|;cZ>1OO2w<xDq4pt@FN@UZ9z<c?_%`G|gTiX;`6W75
zhwGk?M=z%JK2i-f>}^HL&J=tgNq<v*?RxasZ5?CH`w0J$;r}{jz+p_jpdd61Yb=hP
z-?_YLMQY6=1=Ipy=M_%4Vdlf?5_oxP6%%jo&cDv3McLL`)+NAl<{VikWbBYpK8^nK
zxr6xpkz@WcyLoBv;s1<(2@~^w%qxR2{A(rUJ}ctlZn;HS#Yb<KA4CMM(sri*9>8!K
zzoaVdIP1{py(G?e2Iexht~vPhC1>^vLfa$4guGKWBdvF_>vdE6bO-_N#jVq1?IGs&
z(6Ght=A8-}BkEz#1Fc5rn;y>f+ApqsOy}bIzSeMmvG}R(Dn$rlv>-sqWd!fNll;DI
z+goIb;h4^OrJrq}Tje*h)tHbH*D=kBl1_e<rtT!KiuQ*9rd2NZD>>bG3>)s^z0aHa
zE?&4hk(DSnSKbGB43SybrPE@9$eydzN4}g-d=++QoV4Cq$jJs1%%fqdy)}erj)y|1
z8QFmJ5wN?eSG7YTmG_^f41|zPWp17yybO(MoJNl*Y7cdu6_8SDDn;t2DJ0q;7G*Wq
zO8bM?EB?4JRN^G7Fio-{73Pii#&Q!KCVs1hX*foYSNjt?qyAtlQI=UpX5H`dku;<>
z5WHa{FG9w<?u_K+u>3>W&FRjeZV2P`(gO@;-1>pdkiupRY|M<BtL~07e81nA&G`mB
z9l)%jdA!<y?68K9-45Ho%>)0mMv5Z_w}!Bzwi*UraSQ?|%>(%|{#}Q&Yv7MKzVQ;V
z8bXOL)Pt$KYjkN~s()yv4>=6BJtLxnwbL!CC5Bm^f_{nM(;Jbx3}n(a8z}EH9fTG&
zdFCUt%6<o?+!9)E-2k_?=;xvtYxYG9(>>iY^-lkpny4tB2w)Pa+T)w<n;d+FJ?PUY
z<Elj9JTp;v4Z1i5a}5@}u44ROJL&)XSJ1oqKdmb0#mw~avL~ltjidd`-;dB8b{m-{
z_^1^s;yvfh+*^jD>uSCm@pqsMa!k~%@%sA)i7)-rj^vg-o?qH=ayVy~;TLR+6+b_o
zp2gmT_6|Nq=XGhQU~4oSg2Hs2Rg7>4Yd~iXi~;<;-gCuR4Xe~o_QQwkV(TC`EBVd`
znI+yQK;wl(^4>XaTr(4gdW^fAb?dcL8=K+zcvQ~^Q27f&-`8<m#nL=mhQmC4(?eF{
zZbbh+Dq4X2mBh3<ZK03v8;5!WMGpsDbOATls#njMV=CEc*x>kG!%YtM!d09}e>Bfi
zqsyFs?jFj)Byk-7W^e7<=C_aB{Aj%Yn52vvE#k*3v0RJL7Pv#Hw2ASC@u#WohyIvM
zS7f2mSN@AT&$j^adXVCYTs8w5hz~S5I^2~W&YFCc*ju+QJU&G_dig-hJX?(JkRjmo
z!wSfHcVouO!+|hw=`T^^E6oPJ_OGj5#xyqneKaU}83H&2=kL{9cS-siJU)1u@SjO>
z+ZOfOZB-611)R9jpj}g9B+RRQ9&Q8-T?I?i_wAwmac)qvM3BX+0oDT|lcLEYzR&6f
zCkvKvoWTYlF;Ua%Pf6=01|NBIxs@hb1#mNCBbk~Gv__!<WkcbRv{Q(~QZ)YKvvQYR
zSD2@Zi%aDh@%X6z<p8+phZX(DlSqSzojOp$ZHb*|@*9In1R@|ROh}Zcs8oTdJnfPp
z(_+eJ=~C+Z@-!I(Rq|Jg27lMz5xt@qPEocK;AO|lj34zzJ%EtNnWROfbo^<6_s3~_
z4w3x%(M$os(&VJ0Ny@J`Ahy1wT(_cIY4YRMOV>7#l3=1=`EkihvWg$6!r%zWZ%#QV
ztDK6m5FEDa2;#wswGJ5<cW3%a^|OC2%yL;@9m<?HxQ$2Jne3$yM6wf4Oc&KZ8<Q>^
zVzg2GGCij_P6L|eL^8`pPMC~EdmrBY)%B$pldM0n%ltY!GjO;bZ~w7y;xmMP+Cex`
zSo=FLQLlFgm3^JJNVb>-U)>Nciks6%s|Ffnbvr5r8KV{QQY+1Gm%)<3s*E|G`_k!c
z$Q3nJgiQ+-^}=Q*ZY?Mkw6I}%f}L*j1;m=+D7&bzoPk^x_KnRgw#iqejnxTh2UEGM
znV1l_Wd<S5AepXewgU!8^d;}>d@=@<)~MlBn_`S7GCn!}(6A`dSK$rvUdY$oxUUjG
z7V=LB)dL$*C(1^jL`mv_@t4WwA*jqPWR^=LOs1ryqz&9`NZG>bZ)jtSN?Z-YqoJ|(
z1qImb2hQ^c-t+t>g0vNc&*@20aqGIA5x2`my0;P5{~O`Lr_zo*ali;l$5O}I;AqaR
zzk1ws?$9n#1Sw3KzgxbPEd{TZwkqLuvFZ-|jpui8k~>7PGyP;&k{YtP!*9em*lk6W
z-;-*gzpyDmDq@`5Ke&+aEVSS4`$pqJLoVHryv3gm>WXESCq|!TlTK1pL0Qx6q~1&*
zTdS=dMf@6}N~$>HXPy=G?TS`%F7$Fy_{V0ol9t82YK~TAx}leRSRn6%`q&$}8d72b
z>_%tKe=j!tZ2X^gti2<Ls;IA6+IGeFLrJ1VvRJ#jyvZi%Krm{c)X)Py34AVvdNNgO
z%R?j!EEAytX3-I-#;fK^#SqE#*FZ9#hveqWdy|d$$GZLdD020)5kP%)lAU(vHc`sV
z!2w?aDVx)lgh)(APv)C*UGPFqwwbc9mB)Gc=3P+)&(d-SJLRN2qRk?;WZTYrw9xS&
zB;t_h8JTqO&N#6K$b5}g-<bGIIQuOAk*rTvySaX81Q8XXVWw@z>(lgbPuq~BbcGPX
zbj406V627B?H2j*C=vi){*b@$PuB2>j!D#dg}J7_nNCFW8I1__PP%i`WM+Wbf+*j2
z9&P4!*NPPtRa4)OdL31+yiNHH%F3faKK_}L%?54hh#A=kj_V&jy7pnS*!lx#>BPOa
z+kr@;gj^0zObD*Os<t~7y+O5wZ~<O)x1^TgqI)epLyZ<>r&ddOP}{zDE?;|&G+ZyO
zBPAMFqr9ZJNj0k@N;}BEuaUYwyu#U7lfn6c=s<hTW}8MZn=#Jdfy$Y{B}L#}PJE2V
z)47hBzu?*YS2md>O6#-5pYNH$6AWE(Ngd?{Q8{;sydmmx9Wew|ntDg0P=fD8JCOQ9
zP)t>WEL0qpGsGy{Bfud`M-)qtlEBhtZdM5&2lLJKqy~EMjE#tLnz+kb>>#IW^)ZIs
z1Th(t|M7+Cc)kY5--3K}uA2<j8QSRCj;;uty+*3j%`oP?2IdIOEKyn-J8@>x3V`NZ
z7yWWk?-fCd+Z5D2h)lpQ8>K;m#sHZ<F6RMQE}vw{jKW!z|BeHU@@=GFtO+VAo6c!^
zLJCbX6AE}4y<Ww~?u~wKjY}_7?D*53t`Pg9WVkC|a+!ISaYr4MyfY}9hf`fPXc~zh
z!eCaE7T0BQ&|ySqvC<h(Vqw)`(O`I8{!`Y&oYsoNR+lq6Jx$<YtxnILa@IEIL?!=l
zWv1*I=z_9g6A`zoa}&2|DRHm{MEN3j(U5`bXTuTO#Cc<^em>FOe`D^f!rJV*H(!Gm
z3beQtD8Y)mOIwO2NCOEj#UZ#meOufK5L}9r;1XPlyAxcByA>@I-pTj<PiD?$&gSfq
z<VvnQ*=w!6_q~2Mh0pY5$ERj9T9pr_{GXHMBlZ7QSgU_&w8OzBvDuFcDsfXh5x=_Z
zd|HO{+2v7_`I#csj;If*`NOJH`xk%k5O!#LBvJb3JbA_L6nHo{u8&aY#g}Csxsusg
zy!9qBJHfCwTjn(Y2>@tVUdcZf)S5sLN@FBsX~c|Lk)ke~FO8UOA~mxb_uG`7c|?U+
zV2C@)!1%Eai#YN(zn!=4djsr};%q;2sL*b3`n~XD`6g86XRXx|@!qWAzmF7hGrKe(
z+q<P2g9Uom?)_kbxUh=K)kxm43-EW7Jx=9YsfkU!$i3HRqrzQcKo+exD7Ch<;ZM2=
z1rpxz4zV2bMZo;YKD*{Zi%Ef+8lCZpg}s}b5LB%Mp=I?PyTHCGk56v=aad)%#K4eT
ziV}#Jn%E>?Z47D@6;O79I?#2$9lToRP{Y-kvd_r0v+EibuB6lyntA(+PRoKMRHC7a
z-EXO)zq%5yZXey7r)J*;*9m|s6BpE4t_hs`t{R})gRWxV2@)ih`+^6-g}EJe-Quuv
z;+&~M0I2bdQAG0XsM#Dp1I`3?Vn0^~t1>2HB3>AFVKdme&#~2v(uq&OWu6vk)Y%!D
zw%n+ykW=I&A4d|w#A&usXss(3?$Tlp?sohqe`-<Ax;;mkvlcl@-d716wqEr*h12Vb
zx=H6%Hli{wH8vot(xBX!Ez87Eg8mqwckG(y8fWN%2Hg0&f8j0$IrGV<BiVx>J}oht
zoES<?;gS%u0~q`?blGhewUHKvic2;b%$g94ucBEf$4Nzml*_qJ7E!!WKzo#!s7#vF
zF*OLA$;v;Ni;(wU`Tf4HeU4ApBlV(en*3O3?!tGyw{g9%&{zb=N5zpgR<&WzxnL}!
zcxnz*0X|4f#n6zO5Er6xe>B!El!`G3`Tvm4tW7vvj?xVh3TNcXy5}YFeYK9SVqN3)
zl|(Z~{?JraV|%?LISOu(%SkP*o{@|wR%+sE-eyVX6=kD0Tk(2N^ZBfm9|qBxc#rZm
zQB%e~wH;<N@urbbe!>c0XodX@nYDU2E{E`LU&&gJnM5#J8<weWx!jk8A{4}7pKHkY
zn*Oy5&f))ubx2ZwMdhz}&|(ytezGy5;yBzmMy$4M&@&>FACKkz1`P7m{f^;|VZc{o
zvbIDcy{|@x+s%<Tl&Le2H`GL1qvFp+9b3$Hw;ynX#k>{cbfAartM+KS+Jc16<;tJg
zGw-qTJWA<0HJN0@kktK|X&lRRD{8(=-=vBNo41LXt-Z!Tn`*Rc)YoV3#iQsLq*4FP
zX{JTCxRb!F5iZs)E|r2cG6G}I*Hy%K<bzUolk)M)>~;32g0IWWYO!{P_$g?P46q**
zY=%_}wQk^A?_!o1%)VM6)nP3b5vTbpLMa_wT{@=pjCL1jd~)xp7=JCb^yBj6NUi*Q
z++0;Ihj}J>s20rQYl*J3eg3dNsp9mYyDT|3LOv_oExyxlv4Dhzxkc$dSGxL~X6W9i
zyu8F?RiBebsdDAW@IKUo{9==Aa;ep>S?|bDn1o1=eW+pA$rk^6IM7lezYtLP#;srM
zkhXhX%&r!;q^@<QO5NZ}YRV-wADnKF->jV{VlzJZvQQsXtfG|P*`e5@1fqWH5lIM;
zQ-U2qe2+@o{22vmss%J7Mn^2SSPHdhBz38(epPc+(ODkA4)zaV1xLQDUCYnc7-s&r
zdH^d#iS|d%-xnXQ4P`dRk`40)M{003)Z4E`l%>wvUYYTyRV>W5_M!%16GsQB4vpkm
z-vO+6Q;VR1tfClAt2gPswGaJ9qa-WSdg{pCV@Ha<4lGt(pd^x{j8`Gl&`)~`c7boI
zRGI<IWG}TdFOv`$_8me*@<bgPz@fvZgv@5vDfod;&_o%7>3NW?y_MrX$v`Fzsg})v
zS+pHqFx0Ur&s~`0cE4$$#NwF8unK(WK6Hh%;TFN<F?G`ym|_+YyBIYFg{D273pStN
z*?AWYr6hT>uR58m4IxuEpaP~O;mhVKh!Q6Hd{Um1PLrNgi>oAHzs8vS@^g;l^OMNi
znpC}HHA%a;Cr>y@w6jSXY;&$w=G*faR#bXPY{G0^yzJ=>#mz^j(a}`yvh$!}P>KBN
z-3QEDu8R!ng70whQk`%-(|smP)GTHOV}F=!Mo$z_{K!Xv)$pL}P_&!&vH=g(Xs%0X
zxU$UeLPW#CSywE0!2VC8jU_O*&@_k4T}j1r<h#={tD(>_=k;j;c|$UpNsGi@aDsh6
zyJBvWePz60fj+c&NU5~zOUW8D^dI=0Tu>EbzPv+jYx0cxX0a+)RM5q~V6CpWa%z4^
zIYudx7v~HdaUiL!uc@sMooPL6CU82;UQpIt&<~nR4@avoeG4L5y8`f=iYk8dQEE*`
z)m@^24boEvdlWt8#kN3{B4rOL1yGeMVTDhjEm}JFs{cwa@e0zbqlphRQ;xD52894j
zDb-cRb3PlgZbPlu+xF+Zps;F1SUyVOC`pJ#KA_0z);;;Bn^kPC4Q?v9SiYb-sao(T
zUP&e0Z9z^xIh3JRyUEBH`n42;!FUWESeXwN@0}gSFP7{sIFAVLF6v!umG7(S*<#8A
zRUB#-78i9#>~9RgJo5U0if7Q@pI-SgdlW<KGA5m8iXa_&9z^tn(@wvi5=@;8h;|s@
z`1Q8riyf3J)gC#zs->TRU&y`=Zxh}?X(aZ2kUQ~|uDOphZ%rt2^JQm?Wv%Ko3pY>|
zBw-$<3&wirazDiO<aF9n%|X-JTYYDDh_uB=zs=n>0hrY$pjo98A@P7cB63S+Jvj;O
zEikjrbg0DM=;|u1$QD!^1oH6DZUz7V@AVD8hH6WCfzfJZ;rbfR>Q<Z$>}vU%PV3}j
z{H3}M_X#5II!B9^-<%5Sxq9@#W}RnfEettOb%-tJ-R2ofm)nA)P9L^5#J>Zqy|TG7
zn@7%<YVs@p^OxpO61ydryzQ=s5H?s?*D~#noy-5OAg2f6c8@59IuzCvh${2LXcUoO
zg!4ZI6ManJ0j8-no<z=pI$|vnuPxDOLzLPN4en^i%)^9QP8dR;$vb7R*Z9kq0!b34
zZ5em^H|d3jcdRo$lN@FLJ_1&&_OcT!KsU9bgCH@v$}V{-dlUom!f%rt3l1j^zp)%W
z&CTT9<FKMvHL-5uNzRJJhd(9iEa@uq&7#eyU@}$fE}hC`T#$gWK=nCAQL#sMTaK2c
z?)fPVlCMn$nb=?N6IRHO8VdZG1##8eND^rr1;tOqbH}sUq^~V&YI80PB#Cb%QCQUA
zI}+IUmejGthIqW8syVhzGS>^*{3Lgq$RR8sNOlWR#z;iVmgpii_k;%hW!|JFpagS!
zVg{GonI-b_ju=eu**9odAOCc-o(m*jXAhgaKtWzqPaenY+31{~JDaMEDkLW+B_+TS
zsTt|1sj=8!xwb{8ihia_-9*b)Xz9IbS_x)5olQO7%$T3=s&%*9PvK+8Ojposci0Ww
z^iwcl+}uv<iLG7)%BM}smBDO#<8i4}w@K~G@9C!N5)R5D)3$}oV&q$zM+a@cYJnEA
z1k{*!;~22ds>hFR^AZ0Uz@SMZYdE`men4Att)cOF8=Rjas!F9t#P;=?r^YXg=_Hdv
zN)hu2FjBRQ3(qDz+*_w=5{h62$wgG3&XWjH#RSC!mHM`)6V7g)$x{HE)y=(JTpv!c
z@eExWjeG)=DsP_a>$dfk|EG#(*&Sk=7E@CmK1Y0DcKv_kh5sK;h{PWn+7AQXK3zY0
z75fO1FpikBD9Wz17;twa)4y@XVfygjZ<)=1eo4tM_0@gu1=*u#+1i9JOaAybhNYi3
zo!mFg;TH9^r|1f7RniRF4+|<-x|o-_vFLg&#L7!>L}sVBbk;Tbuo+0qyA)2JV9D{2
z>(iGEOaDIlo!G2*E-atBcYiAXAbg7DrHA{;;Cei#PAE?;87#-!k$mRJ)oym5iGR^m
zGK1Hf{p2I5=39^>w4)SUlplR-Z<NmDp~7A@!Xvc*GEJrbF=|1I%5LF{*=0J2SyG2T
z9f!{n{n>xVj$=bO{P~k-2qY|=aFF0;e7a7<pIT{`#`zv^_BunDx~Ki+*@2gUeDzMp
z^-tAoopi`>7QmEnDvb*Ron@b1p8<%3lG?RPMGND3Oj9N^ZbB{J-U`oHNgZD43JcQN
z=jh=R4cRBp$7(MilOm0Ig-Smctx39NXrfP=3c6&%0WC<fCPpO(3kfnUhA(jOt4-=&
z?@RU|vG~2u`Ke*3aPy5ys>sw|U8V{zHINXgFnfXFBm?VL;J0$k_BIG{bh=B4k*T2l
zbjfu8X?|sscGGVnx#TYbiB)u<Pj-LFQZt|1iok9p8ieDAy?sTg&Z0z<9L}YdkfiTK
zXBg`gLFj$;S}X@h?HkL0M&axLqO1Wy6F3AFmMnVA{SN)E#JuU&`M%>f@5|0Ae4;_+
z#J>E>;fNW@HD(<ZiG6(!&A0<taa+m}Qdx7*Qc98DDGsz`&*7{`;u!=(Eo&oMUGD-p
zB`Iu1r|KY^T4YIy!SqIDc=$HcyOzrz2cx!!Cv(@6b>oAJGz2ur6uV<JBT{p7jh5Ns
z_#?9_aO;VjyZFeL<~G~;l2;VZwhxj8iaeCMe*249bQxq>UMwso?z^6DPi{dh_E<~k
zUl!o|1vj(zDqoUBemRhS)e$p%%nN-tQ8R_MrLXGjEE;4(QOF}X8WkdS!X`z*>Ipn$
zVBZ}33%JN#vAqLJo}jIw5Fd812HJAb|5V}m+48!|<s`IjTF--QZ7Wv2YcXH}@$ioS
z1on*lqZk?o$dYl*#yCJUZ5Rg>j~$HL2)@mUOIm22wc@D@)G;tVH7ODEj=5HlvFjS0
zVcztq@>WTZDKnpf%M@8nEhjV1YnXzyRTJo4*cI-M%L#eBm1045Q}lRs1JrcS7X|!(
z{hlrL8(HghbUnzF(it7){dL);(&99e>a3KhR%O4~o4CF;<v<!J0Qy~^ls~589cM<V
z_-0FBu^MkQ8t;)siA~+;A^U<S0+h!khVAv#=eL@%uJ_OO8P|`ww;XCc>N7UvXvLSR
zMG4h5(pV@Uz#)I!G}+z7qc}h5sr=JfsPUJswD${z&@YQie|yDYmb(xg22Wk#TQ!m?
z<81jH%&ym0jyEX6WU-920|NbWQMFJ|Dsf1?wlW)_P7cUT=XFNs-fpQasb9Bhy0wN*
zx)@j<u`>&P`ohd2zCBFXr$ukMMrO@p%i#jK)~Y69QxeP}!1!KWl5mD{gL%V|HFq9P
zEYPa#f9Flqbn=Jx$vu58;Bv1b<8hh7lKS(7Sad8IpCV1Eibmd-Eav&duC5lfG6?te
zfl@19kMMKYv}V+jqK>AgYLSS$1xh7V;n3<A6hk4a`g+5mMdB9$mPGxOi=gI902y{&
z`=yQQgwwx|o*#l>Kha+p*;xt^_ESt#iwRh-*u8!7=QJL98i~20SS#0o_&sDF#44Hf
zVzGHF>%$)r&+7Nd%Rf@^oVRxPX97K?Q+2)53OL-Xq}0}HdAwBGFampMR6kEvvQAJq
zZ_T8LKUr~Xe<(W<-x>VBy~vn`r#el>fh?JJ0OJ<FT6_QAVbP!4^l>9HBgKkkbZY4!
z%|T7R9AAV6EPhhY%(=E8iN-uBK}aN#n)$QED@&110=9l0_Tk}j#tkiY%WT1OYfKt}
zooWw$3-a}mp8e0dDLvRsZZkmE42bMCNvVbjP#|Tvit<F8rb6PL4|;@ny@#}1ACYt#
zoc9kKc7!^YKIzzka;4c5He&F2*XVkVH-#55^>lz-*76V?EXowkU4#LEz7lrwgQ3#W
zkE;$Xfs%xLApM{zYrCWx3~X7M4a%UKm6jcl`{e(T6aJ`+i0H6>fc3@E6NPssbtLLP
zR%Xg+Y9h%vnj50S-96S;J<)4nGnpegP>C9kvBCidf1JxMRRM@1(rLPaI+#c*7u2&*
zYI9U|#FS6RFt%GmVb-XZ^c*?GKApPoi7$o-9!?>fAjl%fh{K^lMh(jl4y4I`cNvLq
z4>!HSa&&!7<rIBxlXI7d6-z%>OzrW0gMOnI>Shl{Le}%lR0KEy?Um6YH$;jeLUI^4
zVOw{&URZI`@aaDC@b&*o-hM(5E@va7prE)ZF#eNf%}W1^p)J)pzt8qHTk1leRGJw*
zD<*HY3OCalV>w;=^81=;HVMh`GcYMTdTM9nT86e}-h_?(y2bJ%3y(}L4k)l9u>yCM
zLhyNbgRA$E^a5v8>O{>cs-C9RZp|*_P0j)Wk899=)=bMFX`@<`#`c9+4dw-$Lq(K6
z$e~d1I0y7pOi~RCa2?B%<v(YgiQWKq;oY=5&QEh#U?k!%)e=6KV%m>r`*K_mc)fHk
zQYjvtNN$nN#WR=q#UTA?hkvAS2Zy6)uRP7!o>#j%tzHZ(mzcnuxj9?eg;IsQ$jOXT
z!aY~{Cov-WWBs;EGsiCfY~K}4@vt{!PM3z&_;T0n)VIs%u1=*axbd2(^_1+Gt1Qh7
zwK(@K^irHtYIj8-q)U9}k@`IjcU|x>9w}ES7{S)NgaTZ|D6@)^bp8-8Z{_%?wR{H`
zxjwnE-z3O9q8UWnN;WbNYe9|{>#|K}s;av=6|y}r$P*0vv|@r$?u;6wpV$B<m!!4K
z?i|h45K7{h5cKttMa)vk5>b|UAdhf#L^{i{{*yn=zsxjKp_PKfQIi~X7#a)W$Pj$C
z6`jguA}Kue_bf#kjfx--V1~e*%`>?~IN?csbc@$c`-I0+1L^p5fWScAB7T*Xje|c5
zlfFoKh6RoCuAXHFCzE>Mw0@bi*NxUc{iTw>7+tr{;rbWL>|?h9B2+lsazzc3?#ju*
zVz<W!U%&GCK1?$_e`%Eja<NZ&GqJ`axRHdrqysfoZ?M~~aTYM&mmcuWvFocl;)yVu
z*y&Gjp3Z`ia!<`5?bBx2nN|7r6Ntec#`br2Z7==mpRzUM!ebh{<~Tu>RprNJJna!)
z{wXa~)y}xP;!xDuoO`3Tf3%G@OhzgX4zY~HdU*Zg(2$o7Dw3PPBm}Kh4E{kGhXk}s
zIzP7;0UYKx)G1rsk?|?~>cuzW=X~w#<G|KiGI=y0av+osLLVJCLwfoUfKRcd_Z=z)
zQb1;So3*kvJXw;t`sl`Wbuw_=qu`Ky=y;;!YIT^d*>8E?c0|6oTLd{b72;1zS`uFH
zHa(wpCn&)}V@s3fUY6UGvkWpp{34`CtHU9>uzU31M?%Z%1O|{68o$5mX$y6Ncy)@j
z6)Jqd`H&_;;`HE%ylG3cdZbk2rSWV1!ilu^5ZMLrK^ubk;-$+$!DqXQIF+z3g`=fS
zi{ef{dpVBvg_IE-%SNrvBT(7(rsIgsz-0Y~5`nQ_D99gH7>JOVpE|LwOTJ-2{<-`A
z)WZL_!=GNJ3>2(i0NF*eM&G&{WeX%re|PRx-PPO>C()EU;I!Y4?xk~nwgLJSCb%Ly
znajkxmyjOID9S0#@h}PlT0UE<mA9pj+Zd9!v*y*3E;qY0<uAwh8uaOn!g5hmk%aIZ
zxj<s!oF22aRvx5>G-{S*0<3Wa1?A=zZV5Z}WoA-;8fddP@=o>YVy}}!m;`FhG>UG&
zH8e*GH;r_Fip3i}4KUrr()iib@gJGvCh8}J5+%@OhG~=ab&j=a>3l)G&n6Yb5EUlg
z$UT}Um>(Bgkj*Q%Wf6|vgouP%*qfjO4KjIcE|FnovE*8XMy}BBCP&m?TDUbV^umFD
z`G3X0Uql2^<m`z`&I{3DnTf+o*2YMrZ3+3nJ>t_~!TaaQmhDRWYdg(<%E#)*uHKFt
z8pyn5w@^CbepSDiLfTwjQQzqLX}b7IKF<MKj(<<GkgIV}Z_t_TGP9<cTccG0`ea|d
zh}MhHKO1O$swy(=u_;xTSNHwX3wx>O2{X=xJl#TRI8PeGIiY$K<O#Y2!t$BL!@c?A
z1q>w2)^SM4j@-34OEd+SQVsFjkLJ2zvD$rpik4rWp?T8T1jR);nYg*dKSwUAtIB@w
zRT=44$fMakn~+p$DuW3}yDs$t0K#P|u_hWq>jrI%8_%y-hAd8v<IisrA64^AfBZoR
zF<uV%fmfXI_sY}1f4uqS_*3n>(Q{_pT-TF!rAIK)lbH{=d0syM&_u88V^2ekUq1!6
zNNfN17<K=FP~r5F=CLqo#Gm_+O#d&2a_e<#e*chy#V>z7YQnxn=-}W9%pO$MbZX4q
z?Et@gO{ulGymQ6mF8z-_?>TQ#XBYN}tI=?1&HREEd3g2v{84c;HkE7kJFwNvd=(=Z
zeQ=b+-gs<Lf5sah@bF_in=K9}v-f`QVtx|dU|9GJ>f)egTfb-0VE79Hg|tp}2wp;O
zh$pUZ|4I9sbQiwFGxcQoppmqF4~@61{^cdlHSc$T55ii$zxFXI{kV1i5vHs%UO5}^
zgcB}#<H$quvSRmf+HZBE2fTgn7mgX8uh%G=OIWDC>X`?uQc-^O#9FoRe@%+cCUrT_
zlp5Zi7zReW$Pbgw=9SGC=ZNFoxs@+{i1cIOKq9wtjl;a!!)W3cB>XqrhHGhId9H2K
zx4iRWzn_0OA4&$6I|$T-r~2j@RF>gK-)<Ej>WAdv?VbKWEjHPB;>Z70I<VgXvy)va
za`(V<Y=L7M%J88Bo&(g;K?QC6Y(2|Cg<v||4_noRBZm2gw)D2cJ*T;FOLha+iaLF=
ziEL($VQ7D80>zn)GFAx=<rv=sFw!0i6qZxe#d=p&?ANo;5KP%^rlt~{{!(z#>0z~p
ze3_zhm~F)wf0NZT)<PSTW1w{;mnF-@TDY{OD4+=wNmN*aCk5%M{BfV{cfXQcO!I3s
z8Fg}!IQeM@hAHdUOPCE7VNBct_7|>MgqB>QUBM&kB7^SYjxr1BPEG7!=7ESaT1RU9
zx;lNwz3i#)_N<HE_bmZ;QIzv{`25F%Y<5lb<G#OFJLYXQHSN<~$OE$Z2VL6m*ZU%m
zD;oJ{4&;=y<p!yNi3KD=)8B_?yV{eMH@<KAKP2BVxS~DYm?)513i9e41g4}K=hg((
z;{Y~Vk=o%(jr*RLmWh97W6DRM4n<3i?qtgL;ouP-`H%dQwQ}gxh@>XZLeB#${m(^C
z9vPb+TH}5KT?SI2VQ}m$3}?-_aTW|$bg54JZtG@Un-`I!@118pcjl1nH|^WKRwOF~
zkLKMsYy`NqdLf|NHp>L`Snb@fU{{pnwGzr<PQ;=HMyFe-#$hGys0o-FbYaT6{-BN$
zmr&ie43vIlNMK@@tj=<LdrNZHbi0E!siJdbtb(P6>-_u3f1C1&AfwOLN_)8AQi-7u
z$da}QZ!o`y<+k~>@F#*og9%bZvV$Usfrprjm4`ED;!Y&X=?Au?st#Nn*8GB=1R+XI
znw<mM{iQudkgJptt3E2;p2ClBX;nXk-JP%(LoUQZZ%JIWGl0Vv!eMowT>G|*F)M>V
zj)0@pUs>9Vy?YHey(5R;q1Dv#J+~Id#6L`5&Skr5az-_(<tgQhf=tuN`}<?gl)SfO
z4w~-<HB{emzf|)HYvxY!Flc1!3s;FxC5hGKk#UBvBe%}w+5a;l1ImF}?3d5JpYTzi
z)w}r3?n0ZgI<HQ)pJ5)(TOPEk-U{n?tv`a*hobI*o|<z{hW-w1M&3nFUp>Njp8NRs
z?)%E{Q5bJud~C=X)BXF4>ty@lgOal6pAQ`3U(pVm|9Pd~m;YRq_0%vl=Iwn7^=$ML
zz{U~Y*o`NPX@3^i;{FrsKByX>NDezBn>_mO=XS{+@_r!v&n0Wa;N7=0?JJ%s{}U3^
zll%5e-KOFQA%^Ybx9}WFz*`z*+rbJ-iBRkGk+5dL=cmyBL`5N<pESr3&EVg%Lfr0C
zVU$pA`|y?gZ^D{ZBMRAkgpi$O|H!_UB4=#OI~qNgy9766J38hy;AGmjDbLQP(gs)X
z36^dSE*?fSU;A7~cOLNxd!EXj+!1!3JPByi_c5y3R9pWL(55`b+x~+YPT$szQ<KbM
z(8y}L+V!*eWxR{*ElBp}C#(0_jeC9eH-Gl2!Dn@7golHAT86hpiQ*0~IJ^G~ImurU
z4yPvWR)-KLj{u5P`}(1OA7z319$t@7n=VTx4J^a+nwdT@&_D10NdvbM*!*fiv7Es*
zdOtuu=m8P=HV?7G?UjTlmZ7B6Bc?UOaaYP_mYH@_KaG-k55eD#W492~w$Bnp-P~sl
zyaVlN{BGQpz#Kw*Dyo4Jf+TB0QR_z7Uc02pa}p|dTx;j#G(7{G?$iQn@Vlt~E=w4J
z^3k5Ny<Om#Px_r>ac9CdbjksKCUf+W;?EDhl3xZnHvJ)HGd{-^{K)0$2Ku8+Ri^Q~
z*lh2x`X=s*-(&C9LsOQ5hQp$UF95_)PnIc*sg320dY`A@XcT4VRe`<ypw{4DhQzch
z>aF#w!fjD8N^WjN9=JTT@~?NIzC<2lg4m8YLZHh*-r0N&l+5^@`)b(e-$(THhV~0T
z7G1J&Sp=R**ijJEkiDo{>yDXssO#1#%?Ufqkyf@(8vo9dtNdD=w{3ojQY%(b+q1r&
zSlsafvbK6UI$#^v+qU{1+rnK`@{jxP^M_H)=pa734fBxiJ@e)L1VQ<9+m6W4(6u^U
z^W&Z8Z~s%pP3d>iX5uE)hFYPm7;=<~Em*VVYgi2`6T2JZEIYr&D6;YKgC8g5zZoD}
zo1Okc{vs|9+j<fiOoeuK9$|3#O0ef0K4#HwV?#&V7s_2|le&M14{-~}f7botoJK}T
zPb8w|m7h!}E&uxq#N@<Ax01kFIz8`c9e`J>_ScncU&<41=jZ@R)1Q7Gjc>LaskOhO
zYr}f%6uLu?-zS%>p(AdhW<%<l$wjt0vC?IHXO8ZA2S4&=F5e*{b4^oWduZixg+i&`
zK%u}OO*m1;5#G)hIT-Cb71N?0sSx_u9<448>qYgRVdIdq81il?P)#at9iRSPuqZV~
zmrIpJnnhc;zT0&}G0j3gDmCin#HB|!**3qo9ySLj-n~2=4%!U*n~#@fK3uw#sAQ+>
zHR)_3pX|mBnLb>_F&Kf^!-*-GLAu;#H0j17mc0wo-$j<&EgDVqQjWeXCk|#thh3*}
zR55j;i^QBAZMAa1^}Cm-u8XEpVMSDTQKt0G!^s3Pa_@`CVI$i@w&_>c+{#eHEmEUG
zlz2wSJ8XB#B_;Ib=S+4QE}xcWUp*XCJ*@(1MkXQ`H3_+dSb>Ds>ks<UBFoC5)db9^
zbga7%6w-Mg5#Ouh(ca=I&gL*d?(lmR%Eq#ce|A3F=1P!xDa9O?mg<cEXmxjYPuPz!
zzKN}=u=4+=)idOAID7^8A{`lW<LjbX>A($abTy92ZB(e-yYvwh4pV_zl@kIC@X;P}
zH^{`oACRr44HT3A*lUM3oexu4Q#$eLIMcfeEL{?E-C{)~liO9pHhS)CL{-y&4&?<|
z-jgUuDT`YUt97I%6m=#=RD#b*;&f&J`=Xbb;*#G_4f6G(@D3Fv72}HrR0I-m&PmbF
z^OZ;pTd4TS>93iL=EL^4W)uWCF@|=Rfcd~$vjd}%6(9bu2gZGW4Q`ta-6QH-GDPbA
zr<+dC`}ieJ8~*|?L&qz+E%=Ye8eg%w)1u!AEKL2fHnzGilTP{CITI|Mb*g@&KhgFQ
zyB3z{P0c%Y2Gv~JolCFmbOS0nevB>8I>TpcxI3UYZYk}iT)m{ToS2dL8%nl;O1`iq
z#=Ms(mN8q}i<dbQRclqqW2nv!F3pURyN(&qIJmca7rkVkp6Opm<sUD+WluRnYosl3
zuIwynKOPY&6RyT7qR(c|GF$HGK_6p9m(ZF=`oIcuuW;Qu5j{<3pmBS7SQPGdv&_>>
zw4O%C-UKKVDf#@j(<s2kwh|B!P-?-sWYQxSN&5RmYnq}~)x$aN*g&5}m6mv1r$g6B
zRziAALxguc$-(0<xghzl>fPAXqufumB3FB*r86$Did{__s;M>pttlec$C{mQgnwd3
z79H0u&h$|ZxaXzXQO@K-#q@?+tZMuNt=?~TVH_j3Sym1=nB1RwoXQdRxe4P!+Ucf5
z7~G@&aR6D9dJm0QV((mul8&n)x+$6t-Dtq1n>)?As+t(dO#vg5($o%?<Vs{O{Hg3!
zQ+>*gwocrZk0!FG7G<Y^`n&60{nItQFbz|Q2IS}la|>0&tqROOgc1o^%9)F0t>NtM
zs8%s72+%Iw5XM{yn{fW<u`?6=(>$hT^jk6rT;<Al&77{9)=VS{)8W?%*Xs5iZ5b9}
z>96BX4<Ne+j1~VB=T9k_o`sms<nqXQ?q3s#xtJ;Y=H8ND`Ar(v;tEH8ueAaO>xQ!s
zmzS=KRWT{A`H+jBt$z&FDuEQT%u^U?nN>%U0KqH_+Sy8A#sN*UzF-Pc>4FG;#H;5T
zi;(hR30(kMLrb~LPWey6ZntYe!7u*(*>1}keF{LgT`jq*7MBPAh|3!3gAKrJJofe5
z!?oeW(x>+IbiS1L^KLmplY}r))k3FZ`P7MmkWBZU65{bhT3c>aJk8+M_6Ono7ta4a
z%5$Zk*5eAyw-PV4#Ji^8@@$=vvur*S;yAAEcBPsrq;zT4rRd*M;4OaNb+9xRVEMSB
zDIyGoR>QaxlM3rI2NrNe%R`~rApH=Vbi2fzcAnImF7kD)N6QWna%ISi<eo?)6O`dy
zC^RCHx0M9pF+Ks%v@0|<RlVjwgHko?Z7py;8_8{AK4dK!1@7(4uPQ%`>QpsoE0b9c
zF|VuicHM+z5}&8Ik`4gdXqnW&FrrYLF`bt*&)byLJ|rUhxw_li20qwzH*fg*zSsZ~
z2^Vq8h6_~a7^Yh>=Ls{8_RELGXmzH2k8xpbE4dRJ&n9_huXqJED)kYqNNrz026+pN
zXg%FZAhpMD(&-&!;n_@xI5BYJU_pc?vd_vpEM`KRG&PwA#Dpw&HBeL8Cev?oQdwBl
z81?QrWdr(drqt3HCRvmCG98pnzi<RC^H303@Z`KV>c&u!rQ+weWHMUt;<2)2ehy^g
zC1QHB6_ac{SsXymKB>x!w+`f?FCOf{lyrph?Wgqx;%X?o+%^NXaSh7d>&3;t%x%@H
zn;wEu&33ornf3%5G_Y=6v)nAkib=vCbioF_y29RG{vUUlBqv&-!KzlTD)OS${qY^-
zr70~mCQ)KTBjgIHK!2bGT~TLyuq9(GqiXV`T4p6t%vr0cZUPAv4Ozw;5cmjCqZ{wM
zogL11Z7WRJzw64A2h=j3Cyu?gY3#R#Q~u<jA16_Xrn``0(=O8}$SRH2a?@7oz`*7w
zbP1WQ+)Yrr!^$b~RphA5d@&F_kuK90QoU0f5b7)%LbxQN#yZcRR&3;@?t>;Px!C7Q
zRSqWF&V$;N_vafVE&g6|AguvN(&vFPHuw*OB3#%ejgan~3$AnjK2m-dLg4FH+3q3!
zvKdGGfl&zjXcOJ4(LDW8>UDwK6EFOfS%6{eu-&MBW+{hAPqtI47OYWA&Isi^G{DP`
zAd(K`OWcfQ`<wuiaYz-3K3eN=g-4|Xlh8DgbvI`f_qdV+KKbOC$%e!RRY@|Yx2*q?
zQ9KRFi>CVbkvqpXd&R_Mj!=a{zGbzO)^c$|i3^bF$Pbi?JoG?-+T;(9TeZoQO{LiY
zat0=}V6H1jbLkQ6#+a}AenZy)FV9B;0;~Gc<dP#TO8h%fkC2CSo=8})8LR}32P%*$
z#jS~Ika;7KVQ~e_7M6wUo&R8))q1E<k%4>89GFJLH=bGMcYHZ$fU25`IJrJAJC4E1
zj3B(+WqCN1FxpM2im}T!D}}be3s{U26}hvraxl%`x_bH>%tVERuA@Y$!%kWPJ?zNl
z_zzPY3qt5!RNU8n;M7+}X1xZVfN{ohR`2m=filLg-Y+;Rysk~)CnD+md7ikH=%)i`
ztOCcMzYJTS&FgYgio&LxPNm(-mHx~GgNq;^Q-<0na*CscI295hed2UbtPDm0jY4`>
zF{iWk!O+kVy@)(0dDCE6xnxnuqU-OD=tk7gjP9w?a$^;)j0tx!`CtH(0#XC%gf(Oc
z6Ac>7O!ZxGB|#mD?ET2h-GEZrI6JR6AXSl?%%gcjp`Vj?qcm^)_G}+p`zrfr^|?!E
z(M09}OGjj^UiKg|(qqGz*RZZJz*%eZ6#T?LD0z8KRlibqTT}kY1dW={1v_8*k^;AC
z@A=mjdRcPT_Eqr0w`W=AsS}RE@oAaY8xz#S*~PgIdP4=DE<*tVw<tJccgRksE8zXB
z0Fq;;`@38bm4`gBNZH#MGX;j#RV`G8|B`5{^l}G&4`}7Sk*#}T68gjcbBmsal=BgN
zJ#1^D$W`I4@dje8`1?5)4tjAw0C`_TMM+{Zw)-Yue?5(IYwlJ-`<x@6XL+I8_5DQo
z%*$q6Lxec*bknRS>i&nbb{lL>ebDOHgQgZd^Y04N*J)CJG(DGEr{O}lSLyYUHfz()
z6CPcQl##eQ*@vbQnyH`JWmA-u>H7@HEqt*jw|~RO^=4Na+#&-;_qaX%$08?lzkWI%
z*4$g2$*ARAvu%@*J}Nn!czAw>W2W?V?LCee@9DSvQhm}}@lZ1lgrD@Epd(;oiMCZs
z@3dQ6dqdAdIm0coTm>tGt&2PxyNFlQi~d8Xk@`oPC}Y(>)3UVo#0>vg*;X>RUU3oa
zxNp3VksR?7QQTSOU3|~aU?cL*tPJ*5Rlnf~Xc%mw57sMQnF}=x_FrELhb($L3xB;P
zNRu%fHpvMcY}t3Gi0ck3PT<=aEBKjxF}3)9eo9$pQb&K@4A1&XR$oIlSv$SkE24z@
zckLTNwb)Q6b1_YK@i3>2K0n=*xb?2Mq5;tG*l2es`S=g!mfEH6`rHe%XlpM(PVByL
z%2<m{xV4Lh#*tEj>qM$J#5L!qS;h>LP7JIkn*1;j@6EL9<tkz9K<;1G1@ao_&-v?x
zAFHt5M@d`m_3ymCu^vX<dsL}_-z^n`k0dU`ZS*m&_AKq=Q)krMN#Au7%QWQ5i$za=
z6l0Kx7dr13L={ugsV7smr}>s>q;%Z=u;*8c7<tsFbql9uBvqG}w~SJTY`UN+$dmRC
z8P0+OdFfxGbU135R8k;1f}?4+dGeuXV|+cK#L^daDu)A?AmdACY+QU=)l5)B7j&f{
z^6O}f;<PJDHZK=#uWefQ*VQ}Tf7jJS4eP68smXB>k*J%PHp1b`HBx{vQTfK5w5jNf
z^ZP)F>N)vtJJswdg;;)Xu4$C8)E<7BgkVnh;X0R?^Do8ZSe&tMqAZe}G|vvBHd~b9
z&t05<seJU!>4sIbzq|DD_eo{bE>IDH$@-8eBu7naE%TK5()R&m-spg7cEKjn#HZ_3
z!#+%D#r_6f&bJ>P=gSNlIMsWAkLX6~v;`9{d>>O)Xns@v7EPHjv97T;T3bR1?K9U#
zVfr+dtUa`N^gSaH?n?^iPU~75`p)}j>mu%WLSNe+=F;9H?@aG_cFjO(gyt$r4I&zr
z9#&7r*LtGG>5Xcb6bokeCLvdx>wOhpnhFT3RSRJfQZ_HpO7UPOah4pl%clzHFdCu0
zhHN6!9=q6WXKFlzO}%jC+#l;2K@#WOUVY(sgt9j}qUsV}7B%nDW2<R6-dbMfhej_t
zWIbMAfxbiRI4N@Qn!GufrDDhh<!J7ET)kQ+9||X$WfS}D=fJ`Lr0TO_u;~;mO1=V$
zEJG+*uo8-pn=Q5XCUl)#asHLAjsAYo^|?rpVJ|vm66g7|uIxZ?Y9eyOP8isESTux=
zoa-?VW#U*;46`f%C=@A=v|9X78&*|O_T9OZ&rp456CV!g`fOIlK*-H~?ymUJ4Oi|P
zFe@cnXAF98f4y6w%g!oiYt|hScHJ3D2-l+D%Ui=+{QPK7ZPIwhXD9MLyz$$=k0OSC
z`0mVG-A}HJVYzHu6ZQAJ`SeWKX6QK)tPf{L+Unm$K9iFUTP)%!M)+&>!U<MCwDg*8
z>91k?@+)krI^&naE(TEjy|k)Hz9Ybv`D;?Me=&+8{B7^F3`eXCE(jC~^xcQ>OFkh#
zwb+P80qyQI{2Mm4rDsgn!gW3+79pP#WI{HlUptf1aq2HHvRa%!e>Kfni`rt%3*O<V
z{j13K84>*{Bd0elxfFz<rEc3INu`k=$uK{0C{=b(a#7HUnCWUp5c{8r<7${{m_GZ<
z?llSi(rvfIUiZnbW(!lYknbTcA5w|tttd}xs)zPZC3h_6)4jJ*^V9DTC*8JiNPwA=
z1K`px^>%-6P)JNpeS4)Wvf>?#$=7k>P2N*O&OiGsz0A6-4Qi`sFb*$0X}ROPla!?I
zO8!<bKiEr?nVF=8X8VGrkv^}n%~x_luT?Xd=a{lby(h9Sp7Tuo+iKK3jZs7Ga!?e$
z?+sUST@7Ufsmli4RhVx-xW4)gbw_|ro0JnG38C5-9y12VR=zfclYWf-gdNFgJnw$F
z>r>^M`3-GzH*%z0R!-VO7f9`1BrL^glj(ATA;~_P8IIkAHGqP~DVSyN$=E9k28$@$
zBSw`;i%&21o~S>Q{rQ^jFD#$B5SoP!nJGAX<L%M;w_1obcd#y4VAi*{nq#D1qaV(D
zY|m)(B>6kj_^)Z}Ggx`4gmd+ANcG<F5XT?ZR>MW~(7owM7i7Brc45^y5ma|n$Hi&1
z)>@%=&x)z#mJn80Yb*yI&al_Q$cYlj=g}MmYJP}%dggavYWM3^cN-w>$JA^9ol@=u
zY%Zok<;Q26DMZAwrH@`Jut0NbT~!^zL0XmiL`zM>h}4io$m=Yt-)XH1Rqh&hB{V%<
zn^@~T(0^J`U|}K`%g;}J6pBwkrK;T>PJ9fp)#=SGL1^`d*%pOm5noY592bsa6PEHe
z${iQ$XA=7x%etozZw=MN*}Hs@iutltRZq@TnQ|?*<PpyLBAcn4CilBsd<*7{*cAOh
zsZ*^@q<My$?lzIPjMtWr(Q!*d4Qg#~0XPITG<-Z{7`xnrGiQ8BglH_ZkoY8Wp*CUg
zH<jtT;YZR_xoW)$oef4c5gJblwps3!b)0FlYLz~xvPp~1>rz4|wtk5Z#@rLan3Ah?
z-z`=l{?42FQW5$sB$sQS_2z-Y<?HQAbqzih7<v-)Byj81f-*yF_M_b;8CCglgWbT~
zjnJaau25Nt0yV|P-Al`?_E)Oh*0y+ZldKG-pWV=Te{`*<FcMdUN-KX93qQbGFC&6k
znYJ~!>Eb5NSUH4JZ%e6PIOQo8=jZ3Y_x3zlqZ3$LQU%*nxaheO&6vvBQIWVy8x2}Y
z&J1wlniP-dV&$}NhdYwpw8f*oR{<o!f*_*`Ft;9^<xV5&VliN;o*jyR6ws=hPM@xr
z3$`Jq&xVvhVS;Zgnep8v2t2pd(%EP7dZLCZTGnZ06Aej<G>Z<%@n}>zfUp7xN)4pB
zSGz|=Y*}QrvtuTpr=g~%_tKcg6b3V879%-j<Ua%S_%c?NyTU0K!5bY(ffS~uT>rVz
zCge4dHRtrBBM9TCRAnj+F2$C@+S1s60#>hD9A+CFCPPB>W}2U;#>9g59o0s<k6}Gy
z?*V0SJ*#^ejXBjMy&MXQRG%!h`_#z#8A=_>{2|@-Lrp$t$t^1wFOdv!FV|%pJo+xF
zGhQrKmBx1BA$WT~E@h8~MT@M}h>JR`a;=G60;!qNO=X$<Bo}TfLe`BB>2Y{NXIaA3
z7q%6Y6_ho#8MseuUfD)bYU<^342cxx*%=ZZwnfm<7(Ejg-wW-)$h$J|4M06SZAmYA
z*geMWC4SoHP|J2S`G>1E^DR<TuJWCehIw&^jn82X%1)Om9jxtUP0BfAx=P0>r7)@i
zAR|uA{OR~G^n&!e9Sv>pF<WJJwFBHgt#&?!)tLZ8OZID6Gf#{&#Ce4>mc%kxBW6ii
zIp)^Sqt}e7e>$BjT$qUT&PI6v>IU;WTnCMc=+89iPSR1nF@ZtE#axAM^Ml=2?v=7_
zbwwXXvha0BjX<?}{;7B<8&%bb%hj#Q`7|3;C8{)39WL(s6a8+0U5(HrBC&v)m^q?9
zmpV1|6aZMa^m`H7Db1Q&9$))CECSWALjeF-qDqXjs{xKsNn6=}AI;M-JX~u2`{-PW
zl~Ry+4R7nJvj>=kzvYPQEGntJ%*b2AVMHPYE)7SBjlJ=9y9;Q=4fWD4_xRE}kqh+*
zv{tWM7M+t&HAMv@l}<dO+6lQ?Or;puV}%-HDzm=v9scg}qY|5lq-cMl!6~UED9qv3
zBO9)k-)53F8mL}8v$Y?k<V@lLR2@-ho<=N-f8Ioa*{NI-KKGng!oj13Jf-Qm!qAG)
zCxUNB6jKzq;S)cAZHDd%)se)!jx0&)FJAGcltq#azZ_0_y{-o07pQBf+*4iOC;*JO
z)C+u+kCa0#X}+t=G@XY_-*4|>IFlv|^iRZwCbqTS0>!GZYB|=V_k<%XUk+(S^SLQD
z3^<GY*x$nQMkR3sYsy7rpYbdR%c`Y538-O#hVl}PrY06>S#hY?&XW)%c9TVuyke!F
zawjU%3{|L@+@}KOQS0?9Lkw-?qpo3f1Ue^yoD_y*207g+h~URi^707m#BBajBb*rq
zj@O76W(NuYl!;-M*rycMkedC3@f=PJ1rw7)wfe);Q03inD=6Q}Sx<g?VtTxJsz%KY
zsp4O_EIV3fIVxBj73B07dCZat&c>P90K$fl8r*)L8cwL3S&;qJoS)0y)l&12z8Xg%
z8c0&pzN)%e-B_3{JiJWoYciWuAR8AZ%6al^xU`%t@%nbDw55xI;$_KX4cbo-l&ctx
zLrX&z<TDM6Vl7p2$5X}&<Ne2nqq#^=DpQSxi4e>|)dYK)(GO=f^$>v5_b4>70A!Q0
zVufZTj@&GPbx)IqaG5Q|>Gg7M6sheLDA$i1Nu7+zojl*1re!-ohImDn!pKTO9xUr!
z$L$w_^l3JbvIX^k3VX$}%pPdY5*ckk6<{S$phzhg97%*a;)TBIyaU=sIzd9n`n=x1
zvA&}ofwH{sw4@6a6j$kjGqVfRWRx0w1R+w={lqy+m7f(VPqU+G5)(NRyQ!7ueplrU
zR6Bf}!aQcjC}HsBfe76~0VB3e%@}jB^Rj5cCq_)>8{OtYZn_@5x?z=O9pA{usdnSi
ziAy6Ye`CH;x7PD7vO>7FHSBJeap%j|qSwMp=iRAK&1ZE2tunhcY#bm=%*g#@!Iksn
zrvvJ;t5L#}ls|R$lHxAdArsq~91<@x%K9O&>A$*iosQ*0pkdjf?v%0z-FekIJ&w}B
zkKg`e$V@~u9Uo?wpkG;mtK{%UNdqI+fW3jdG?~I+Wu;iNJmL7(wpwA|uL?DHIMunj
z!tr-xU72Q`+C$QkT;pfGBCl5B&xqP1D7D@!%#mEKI#f9UzgaI_A$mS{)5dRv;}+1&
z)|i8u*Yp$n5DF*^3>}7Gp4_W3oh*4ZCE3Arqh+IK3!J$4N=f_q%NojqByT;UEpA>8
z*<IL?S_VB38n6=v74)VcqP5Cp1Lb2O>89$+0|WiN2;IhFVk_pH5g!z9EiQd%Xu6I)
zkNwmHN>G4g8k+Xc+_8CSZK>y&ilQw*L=QZgX8yu<UM`%1)6~OhWzJcP{zIMTk$jcl
zMskM6u&Nm$XR1#vAFM*7U=@+Ru$!yRWjzAsPI7$tJrC4M9sLmwpsR8m2>#q?Sh7e)
zfgGNG%OhDcxfvKab7E6r?k9hEl0VFpN6iIQpEPuQGXPXwo2S8H&0!1!wB=~|lcaNs
z%Z8KW5xwz21IfH$n#?E+<%o=)2|z$sACJkq4QEm)NiYXeQe|^itWld%#W>oNo)So-
zio`W!&Fc;oiT=JLV)QiaPcXJIiZO4lOnZ5u^;xqYlbvv*LweH8Ag-DqQ@XsqqZ_b%
zF|pBOgeHF>y1Pmvk|fq{on`q6oL^OZ-EoPcjN->{eW#^@YU^d0-ih{?f1$v94b-jO
zR7|9h07A>QE)$XwsFjUyVbWqQwVp!T3CgEt{uAryNR+CRon+crJl~MqxF)FQT{NsR
z3^K7}H|{>*sXO}d$*tvGSiM5GtqEvsM1b_L39ED?kgr~$p?BrYB6fIUJlFcEgKOZI
z2*(!1X~pO_c~k#0rd7In#R^PTR1YBAtzoOxdRgR9BE34zK4No79+G0d;c`I2?$(x-
z$8aNrxG}Y%CHj&UA~RtLcM3?ePMr5ny@|gGAer2+shww)50?~FG=o7=^4MKCM(GLm
zQ%+@ZTIjgB$;il+%Y=Bbk5{f5=Itcsg<cY6TFli)#Zk!>mNmtQGCNik)K5g}Km8G#
za#4FmBgu;LI!>3lq9%I@j8i#7#SoU>Z%>|CZL4MSU*&d%5^(|pR2p53HK=20B;V<e
zwf!ES;$~D6%g@11U<97MI0ye40)33+7v!gSgXit_dX+{BThPX>eGJ5>kzm%TG8O+%
z$>O?l&7>ExB`3XmT}?|z!j#s_U+#;&rqe4-KI)pWC1<M_hT7rC1prWG(JobyokE>D
zSoL{lRL)i)rE9vxHy#KubHIEQ?l2Au#sw>;f~)g1xR&G_9rT!t2~x{c>yLIxmw<1I
z8hO$I!m6vqR7|H>Y&n&TI@MkYNQd*I%jG-1JfOzG?^BdJwYpc}LPt`Cd}g^x1eE8;
z$<%_<&@{@G<Vah@qS7*_C($Yg)#P0$P(+d%e0rowA@ek2E`h|+?FGLu9qM|Zjx7Ig
zcQ@JSMD~-X^u>{chJ+&#4Lu8g&{j`e|HSzOFo3z!7jd0;Ub)JLy6d+;bC#w3d>mrU
zz1GXs1$Q1V;BUZ~#^M%0zNw%9Dtht`$vg%ukatyU5StoSkG{FQxpSp}QJIMZs-zs~
z53@45bwd(UZwVfmPVg#y+-+o)0#Ph}^5;O36L?k%A|WL`(e2$-?VkodvpT&1uI|$5
z7xXQYG4aK7i#P%b1TD*~Xr{1DfGT*OTR|oNSsW5=`L>Y@j%9Btd<u)PdpDl3?GdNf
zAGdN}E)vkNsQYYysl1G^c18X7!bXf(?kD@YD9Rc#bQw$l_&NxV1cKqrVZ*~=p?!`v
ztc?#73w~xaJFU0&lj(Xhxs9W)Rm<))GiaW2R|J^&zxmZe^C^lIuM-~?79Q+$i>0fC
zC5D4g<(r?#shZb_w;qziJ#k4b^dIqeiF-;nHQGsQb>8}z%wKx`7!ZqJsoVKfWSFc9
zi|WqojaaiBc8HcNfT;ou)n1+zv@UUx-&oiH9Pu@slaeps{|HelBkKqhl-KtVvHJ0$
z{CG?sEi{aTF!XRnX%dQ>nUvduqJm2Xd=G-sZm~*}+XZBi9<xM{k0ouc7QaC^j;j{8
z1bu!f)C_G<ePSn`nlhet*zA_5q~0V!n#(`p^LM}NG{LC}5hld-eq<C}HN;-caa<RU
ziP)Ipxw+cZUUCm95stu780=b{*ymsts4Gy)SriT*p`knTvFoX=rA#p8?Hr<1GeLzg
zkGG|{`4{TAsJQ%u7gZCIaH|Q4R5NXwqXaOAAX)Rf9O;Tc{2VQKazKrwZZWs7L3<ZV
z(Y~Abu7FtH%NeJ^9>{DYwXx6~Ayg}0r;>q<Nu(0oU%c}%IpU&vzsn~#CyT(Vbkaep
ze1p5|V{v!M_`%%sFhrWLUe+!1Pm~5M#aky+t++FO9@$9D+U1|hpKY%Ys_Rblc7%h8
zK}yJ*NgOn}f?Ei&$Ww-}FDP^g2u`~4E#sSNxM-3pgGNA#K^KajqMt~Kc19BR$C%A+
zqX4;CJEgtB{GqU_UvhRj{JS+1KYC2_Qc1{^sVYK=V^lI`tNOQ5=EcQ@slHSmS-)h)
z{C;e(h+G#ednHHS#HLsqYnGCAHAD|LA5+;>9(ld`U(CH_R2$J7H%N`P6e;ddyaabC
z&=v_0tU-bXO7P%L(Nc=L2B$a#mmr})i@OEaLV@BA#k=|c-#z<bKfGt(-93ADKIEL4
zxs%M?`^?PE^ZZ1Yy`4qL&47f1vUxnmSqP8}PeZa1Mr$2Qmen7k8J7I})8#IbF~+zV
z-_*fJMbA6tg<|hGdQ-E7zJWR8m%@;km)l279p69g1Yw2orog7Ck052xPEUyj$m7H=
zea9obK3w6)tU(2#3*E46>078(lP0S{OnS_y3j)^TlXd;-r{X^x$Q0Bo9tN#8m7ff&
zK3ND^CT^hY?TYjv5^_&kbnZXZRh_U*^DxHCa2~groWo-dG-=vKaS8rJ=Yf|vSjhU5
zUikl!f0Pia?I<f(M~tD<x-?naN;N1BujeUJW87{&3&XoYix9AFyPgVti1%B|Qb|v)
zTmwkiY;rMQyoBojREm5!1W4%I&7b<n<P@j7!e6pF1F6eRwbT4^ht+I{WJ3v7elkdj
zCnRX`d~I}S`SvaqXv@6WrAtdbzI&#Uuu`}rx&g^>T1eiJ0^H#?12bp}l`6&oap@%!
zJN3vYMTStBEX^-BB)^6uc54cOpNi+4?RfxiQ|l(@uAv~8c<sdWqA5-chHO4s#{03p
z{@n8`rNiWm#LVQVX<$lva%w0Ee%G^db3FmmHmtV7zP5%>M=Y+?Ggp2ra+{Bal;vmi
zb8PWJm?}ysZcR393L--X1W3ff%Z#QJnUS24(3A{ir%3B)J{q_<IPlz#!WIueL5HQ;
zDOQB4uezMXdbDe}=lD3Av9YPDp#%uMR(XJ~@na70Z>k(xEALXAtLj*nA0_vQ-)qpf
zQw_p6BYf89ATTXjxk!aj!M>byU<9(xR_AlYo;z52$w#HKu+=#~DnO;zX=j`>$*Vr8
zNJcq6fqAgXq0T-X0`O@uyl*@>RG4l*E^JJ-9l+)b-S*S|;ha-q^|#jMyXZG2{@zd6
z%IEy`C;Nv}C?0kWq{B9mVaBRFYroyR5)za+n%DR$AwHyT;3WDF=f_S1yHV<mXd=l>
z%PVO+F@{kn=&Hy7znM*>q-qo_AoFcX>2B1yZqdA7QW~8*K57+!&1)6YQ`_h&Q8_A+
zcxfW?!${+ilb=F#vM3tnjNXbF`g|<^YVERB!?uwF$Ud22ubaRI_!5z=nc6GOqe4%j
z@joTUa8mMOS-bCp6i`|~4F}UeUP}YK{j7DH^bBC0sUGMByz2a$h%h{4HvwlVRQl=J
zsd9D^l|cn%E@<bn*UW}Ra}T}O64z|Ec23)h+xW;CBu6S2wZCjlwdQBV-SXA5T-_&D
zA7s5P)`R8TJ?qlzB>0E(u4rw6Y$kwzd~&gI)2$EM;$DEg^$J?cbv=mcI9=OIMI2V1
zV@3dfLHGxxw?+P9o3+~ghr_E4CnQ<XaN3*ap>UqO_C<Gt^zBS9edlw6U>^l5?W4DR
zY0QNip;WBe``mB42cq9lttayHqyJ=qDonfrNfL2um^0Rn>5+XYOXZWro8m*>wNS%S
zO<FCpi}shXmYknd0}3i+hPQ1WcP3T9jHtP=M#Z~BYdfmn*JBtuor^75bJaZJ(>MrU
zAP<8GADK|7=}iM9`cALh#9~%RV&yts9_{OlvzcRSWm8Z~ts+K3;k{E+L9W~#6ilL@
zJ4P2N3IJ3pn{nKTU6ifHj7PKy&ZjKX8jHX8{aDv<HB)}GlMhPMDS1~3s=%15e~o|}
zXequg8>dvM1SU*CgV=Lb*s^Q;Id^nXj5P<`7atRTOhpuN=Ki*O8%wPTa(5&Gjl4@t
zNh|J%OAqCvw@lu`imI6-2{MYizn+_SaoV69uLqDflsMaftXSr$!ouxSx8?!RkNrX`
zljq|wL7{v(*V(we(piNs)xSmhB6{UR1J-KIdk)Yg6g;Ka*_n<)gmjhV!c-Y@y{unI
z6%Kv4T-0AS4G<(VNE5)#!tz5^BQ@;{sdlhlt>7+^2(^c?TqNFAMx}k=ZI=%T$5sFl
zi4sjs!Rka`<?z&_TmiA>EgxLC<>RFKuJs}LU>?qg!b4rU0r=uCnb6#NG3$MHu80Jx
z^SPS(FG2MwLtT6FwMX!ajvGEA-T^3rq|R&lHA)bD@Lao*%RNkc)4ET%%Vnu>ppkV_
z`R@{2R{wK+8Z;%uFXf{6!l>kid&Vi7y2>n_kf(ADF1+L|98CiB*E;dG<80_tR!sLj
zt)fOrG>Ir_E@g@uo-`f!p=QH>%|2p$-+Piwy$w=blGFrR(&fX36@o~Y&CoPG5q&n#
z--3RGY!QZ+mj!si*t@mZ%#gj+|5w=>|DUvb*n5TiUUC_xFxxut;Pfr$?MJN1k_?XL
zb!*$Jjoq0p!aJ|sJ=1|X!-B_n|2ZK5>yE*0F}>7pz%EaPZ`H2{*&!CAvVy!7BAd?;
z@=%1f&WexSV;9VXNW=ZR@~I;g%^9d4)GV6zI%Uw-aK?!z=AA8VqdjoMOwi~69T08j
z@ILL_Mk%V&PtFN(&VFX-a6>Ly>`fm(OH%8~Cw`kJ?LM@${SRl-fU-m{(zD9wo}tB+
zt36!xC`Rbje|zYGr5xx}XqsFlf3cyV&8G@UV6=^JVwlimxV6x-#<~2vQonCgeB{1V
zU7NA|SfS4m!hky3m`hXUoAPC0Zg!+Hj11rgN(QOUCOu452S_Q^_toO1455928<nK2
zF>%qo_{%b{=J|;BRCHgI+rA|zhs%2tq?sBDzmb{B>r&r?>$E#y%rtHIsd+6^$MR~I
zPufT(fhg)ypcOZw?a0UL+kemV|IvqVe_!Q{(Tdr@v<m67!>d%MjeO6=#(RQkV#Qj#
z3{9Ohm{2$(`biT8g#3r|UQ5lA3xs%^2kwjh-YXxye{$ov|6-5)8xPG5*P1<_X_~L`
zkSVA`98~xEYEP-oZ8ED|hySd6tt<X7mG@B!xw;Kg=e`zrkiY?*VqE@fm=|`v`8K{Y
zakQvH6qXN835zT65x+|+|GuZBG}~HMzHj~V9HHZYP_1;ZlVPR*zhRut2~Kn2tx)Zu
z7--O04hzd#cyFsEQzuZy?C=h%oR(4~zPPf1KHm7jz}x;lA1W-_jdIpvnl7Ov;LVk<
z1n+Q*0djN-bYRjC3)Qm7C$uz~MnZ1J3#!-D2y1u7@&v&V(@3bx!Og+>0d56V;;-Iu
zJD=;#wg#U+0sLv~#Y<C-wmMRR_3+Utr~Aj@FAPVKk>r}h>|B0;xJBy`ww~Ye_D%f;
zar4jLpbs_hXGUgD&-&;F#eU|gv+G)uOMjcY9-4O9&u}SO(!^R}?H(*l({1Cw&Wu6V
zIcFWuDXXL>My0D}+Njv2e9??j$+IkLcd=h;^)!rbw){L-N9eTE*P%i&)u}4hU_)Ua
zZa)6XumM%`K4g+cjUrmrlsNZ&D0NcdGy9adgvoQmh-K$)PPk?)&0zYR#O+Il9pQ|C
zP!ZH6hbJcG?i)d|;phL=P<vXod~W0Rg7>fzGhCaafZ9E`2A)VlKv2m<UUM_|R;>w9
z4Z-L$M~Z8A^}@>h^AZv#xX3(@jFrB{pPk7$@vUC?+nV&upohHq<><;u9~9K0cFwNh
zZMqBBFXdrisJ)lq_*(6@+f3;5QG1Lxlw|w+SUaDg`l(#D>Xczz%>*=*>8l<&(wV%}
zI{AAQ#W%Y7E~2kG)a^*JfuzJw2L2^;xilP}UiaNM4n8loINqV3B5jCp2G{z$BD9E2
z(dLv80g}mr*6C95EmEaz6PLgWV!g1qm3j8OKNgZxdBa?5izbmD!N|kn@+4)>*#l)8
z)<DYoKmvNg@i)5q#E_koVL`!zgt0JR<2O<t>dz55f3AMm@leAJ9#F=dovYK^al8Kd
zNY1y?>_z9wYq9wNMf%ex62L?B?pfmkiX04j_w9e)=TG8A+5R&uya_L=az0YusnoF^
zpsPX-!}Z9CNxQIwq=5&JrQI5FL|8o!4l}lD`CY;nq@5_k>hI|2a72BH<=+&<`T0tJ
z6k6SE_MI6kiO)TqY0)UagW!CRR%t13d51u6q&r^eHGi){>lXYdlq?wxzmmw#PW-)t
z^TE7sm-&P`j^q}B*?G?<@4aJO&H1#w$|M6SEoU&I_7CSFRD9ukv6hUbrJ&{jW@f1Q
zcn8N&IR4xlCuP?Yp)%Ci<Oeifj$@!aXdL-5X0iP67Ob`U=XCbNlmGc;{~voRo|InU
zL}nWz96eV|``*L7v4jmusiu9b5)fYjO6H*YMS?}6i+ksuUQz=4By@!-?uGN&Zfxg|
z;rr3Ge>k}f%ZE*ejF$@H7NzFrr#EB=)nhn>MSrZ=z4&sZT2?+e^@8vK)6p9b6bK7C
zXaP&bPCcMxAi81>&SE6l%B-AN!GoLB*55+DZ{RMEGwq2hw3L|@)X2ugX{fxnutM*m
zqmtt0`*dU~Q8~w{AGLQTt~z0Vs~j<h6`LsN71h^`K@auWDCee;=*hkWyDGf1EF%`8
zt#gGJX*<7pur-J380ca6ot$$BtGy$7X)U_3^dOvsX);~Bej!5sx!oiSDC}oxaIb+5
zx*=*3Xek(0>NcN&K0TR54EgF>`QzCAolQfO?c&==4AcTr3I}~aU4SJYk{XhZSV$z>
zEcXw2z503b_q{sQVKrm&_C3xs5MchAH$hOIe(^$9B=s=bRzh+%Ib?=He9g~k`5RQf
z6rq%?_H&3cko6Yg=YW7@cCd_O5hPb^VG7cf3lof@pZAg^e-=!uRkZNSt@C_t8@nj>
zqNyV7*1Y8AXox{f=kNN|!C(<Qne!Sdqn&8fP;qUJ$Wn!*v|79=FIGw_DEK^Bf<ybh
z+pksW4(jjdP-!ORRiy7}f2_`xkmvZ61kck#x>I9P<7PG4aDG7~;qKO_(es;B$vQSD
zG<3VqZnK@mZv9u@;<&qt=?nkBKU`IfdNo9!sVm*`+kbcA+}jN`!Mnh_>hPu}wP_#k
z=9%*~6kKGC<&S(SO-e7Y;C<qN&?;_PSGaMNop+l6es+$Mvd=eIo4`(Ao3#syzBPD6
z8W+iMF?SVCZ`aWExhBMh6I7YaUXms_gX;uIUFvx3EHCfbI?hxL)!YFs>~qZ)-U?sW
zT{=@!VkRLCFoQSTmjVVe637zdOwC9rc~!_C0r#>l4ctTiS!xr6|K8ciMz8g>Qku2_
z&%Q3&sbQBL_Pvp$<5=;z^41hW7@X-<yJvvU>@|(pU@A`f9iUx3x|WzP{R;DKb=1c1
zf#$A@SXg}VmY{YNR8hAlWE#!l6eFbPY;Nc#A4x&TRMHE9my4_W3fs0mm-q%>q7YUd
zdKy*C%8ZoLMr1_`uu-t&wi;y>%BDoyLGmQrnc6a=r>7O)=<JrAz#GPS8e6yy-aw<y
zOi%Uxd`U$1syh5`$~TyzIuDB$RuYF?h^^-1NlN##XfOm4k!`z2tZ(XP6q3!#uFIWD
z@L6hnd~3S{HRh;WX7i7&S^u8XDv6!H>n4Y@P><>;ol8B4$m3|6SpA1HeM)@vbun57
zdKh?Lr4*wC{)ZzeDmifx!_J1G(OR8(T@_ohy&lTD@8w<YP=J>uT|M_uv@ZT=PCHp|
ze2Q75wDitgx}kPFHB)pr1lM%>V|7B_Ik3Acoe@3Rks6YQ)GX>oup^+2w$5qZa$wt(
z4Hp7nsBR4AusZyRQG9|GsKwWO915Bj=$53NAmkXY6iPO)Wa^kAa?Q{q^x5Ee1q{M-
z$kB?6B#z6W$<Zzz0*&^|`-(f$^9c~4Lh%4Hv_3qt6_{gsW*|x>azSbGRe?3HA<KNp
zAGt_eF0ihq@Ux+CTp?a`ylHErq0w~C$Hvn>kGg<L`a!Hy5>=86(w{l!z~DmoKM){|
z`<SEBEhs@i-7r_cN?7~pu=!Co`5g6JGYi&y{Y^f@v#>}lp}$<uFR@-+PIp8<hrwLw
ze(3qhEPCXLhHsC4>Cb)ki2Co`*a9$$CW4tAj;uYokXL8#owFH$;C-Zk9#uWGrZTk@
za@6<OC_;awzSv+cK%#b)Vrg<Ek1F1X(hihD1$j+qDv&4+e=~AQZ@m~r71i8!gn<@D
zRBA#du_h#LN4x=jo#kDsuWS<z&)myX9<zceq#*K?1Rr~_Rxp!u)x;@NK>|s1(&ID8
zO^8STc2g2+TAzXZC=7A{a}wf$+p}%>;$fFZ?lA5y<MB$p#25UL^ky>0Z**o9S;~v@
z3{_6azfeUt%0|8$aqkzo&Dk)$fvxy2h02#1#TB*?5tLp=Yv+R+O5!cswYXX+;c@mY
zKR5V6SymIX#e=^2T6v-?^P?8hRKIWeE`2p7>&ZcZL~3Z0P)Mm9Qn5r?^Y>44J1vpm
z^*`UH0K^un13Z*Z22^!k$4xLd7&l6gSmKLBicpX2w8h!hf4UQt4UDv8*y9SBHh%O?
z6kl)8r6GsVpOr)9r|lByGBTc;PZ-_kTQ1t^8ea})ccnTC^3domsCyqBj!^Zc^Y-TC
z2B-cQdT*J~%~{e-fHV`{+^_F+$X=8$0x&Tp^~l56N#=y*<4<=uRx9>nDV)!#nlxi6
za!D*OH5_?1PYGmPSbwjTdk(kROew@%A%HqjTL<3VZaR7q=iZBdnQc0O_Fv-+!D)6B
zHB6!;;Nl4qS>O(<a7e;+Ntfd`kQtuuf~w{Bth*8PdTO!1;fCcaC#nvp!*z`sist?Q
z0wB=<HegMNt0ODe<^p9*E?z27;x&WCn$ZLk;(J(Mw3=Q&VN)*dOTY%~JLv*9-2(XG
z>TDMl{X({!b|#Z_UEXSSs%$X1Zlh19%O~e{HVjjT*;p}8@?(=yclvs##EjCq@u8ZT
z@S%=i62@XD1^;l)qTu_mUN<R$a0=m;jn-w>a`vuBb>9_Ozjb-&oG5`Y(RBtmT1oR5
z`j9-gx$m&ZOFq?Wns;BQFNI8IJw&hThLnK+_z#SSV8EtqhVj7GzL!OO*P&tYD{}$d
z!a;cL<tm$sy$y@<W$|iac+QvBB){C_B+*V^Rkko^2zyB1u<NSy89nU;vce7`Fyxxs
zJW&#BYy6+bbqpK{@Q=wd*DNidqgHikrUfS!_aRYWx^OYWRD=$8FH1bLIWi)w1+3QX
z^qXMMh`GkzSE$p_IM#AD1xiEzaK6F%!+elx;l!`n6n%>#17)|E=G<0Z!=ClhiA8_&
z-khYHQNA_8*WKcwG;1_%T-`dKb~c$SM-@wT;HYPZsTmm=soq^aE5tkKa{KYPkDb7u
zh%d8FuLS9+?96<?OiXMl2`R;xja+OlS=Jm|^Wp}|g%X6D!$m}q`>rVmKFP-Cu9--5
zsD*leuElPLT@N=`r?;Dm`e4&5m|(JlVZ}|FT&!B_5qnT{!s|VZS%!03A+R}_#>l{o
zhy5|<%hjaUdvzp&1~*zoH0N^40l~OWd#H-wB9<*`X&qb1pp3hNToQO-0U+gJsN_`%
zdK{*%P{^t$P_!m($<h%^Z-!O28`rrr?$mU$HXcJ6*agw@aMCC4)E<%6eVKJlvL+9e
zRkE+sB()^3v=deXa#5E|0vnbU;X`6VEsZZ;H`BLlOw#|nVIYs0`f1;pzH6OXmd&F=
zek|&0nW+3e$4QS&#baJVEny{`CTX%k+tz7;ad__&J^JAV4o0{B3B}_&0hGVYD|f*%
zhDUVUZUx=RKW8<eGYWshy&YIfbO5DK24<bthT9&?15pevHk^@>XFr3Fd@*Z0ylM$H
z4A3fV#hXr;_70K8y06V^ezx<3o#_rmc3U$?xJa?BVM@2;K;u4|oMfZ`JlkjkDP&K4
zm}8HzCeBNuI#Uu>iV{(aRJJ4mv1nhIF*CCe^QVNjf05zTfWXf|^;6!3+9jhi99DVS
zf4E2!>qteEkx_bs0u=r*e__R8U}iWyL}(MH;AW*@JZ@5vHvd|1V&ddz<CV(KAWl4z
zccDCZq)daIpQm?3U%Wz<5>&_M6jDV4u{U`%AZw65Hu>#yE!;gL!F<Jr>Wjngdfm2=
zxghE=w=(C*L%c&}_W>RfR&5aq@`nTqB&!a=Y7w2pHe@-Z1wiZ6FZ0QNqto@fXiRUp
zYDSp4UB}@~DdBRFL1pCnVKpoxI?SefCbB^qIVHfTKRvOUsrkSnbgl>mt5RKE*7rSK
zk3>TG9gQ<@4FaEcng!cO8GI}v!pWRA9wOzz4>)WR9inkx(1cCDGaoc7pt83$#G@!+
zFDfIsB04I1J0KSru!n4HUyG#-)o*_maB(+5dC#$9a)x=r;JQzL-}QRsQu)!n9g=#!
z|L{sv5Ov=xMf?vZ^36xz6O8AL>$c}V9O}!P`{b!uS5uP<fzY>y1dR8N2$5$;ekZ4}
zwYG7onYGqaTTR~lA1+N+b86FT0x~@90me-Tt#A+;CLfnv>mC6mTy8d-*l3!^g1)}F
zuAO0zE>yfRyiBwadim?nQkvk`-do+Q{35KgV@j2ztdz1fid*7+M?!otC-HAtNY`KL
zfnfDy3YQl~lj)({3FD}Zyxw5aGW5$_<;j|5_R<j**|b|kqi;G8E?v!GxkF5gXHo8s
zL51Qz?lhGE`-sWU*yoAFIE=s1tVWM^w`S{iG#&zYWa?;w5PMo(k`Id&<pv9k_3SMn
zjy}NV2N>QI)_HfAPy+K?_uTGoqY{U+cnc-dl?0i-RCME)yCPv0ty~S&zV4E3j%TLn
z2`b%gq9bE9Y97EqgW@5fOO|)(#gi<JbrCZnH5s^XL|EPj6K6$CNJ~PB1D&;>H>pO<
z3QH#45_qiCW)u%7VCya2^D~{%Pa?~w6d+}(jkcLnJ;gz^PLsBVl9lOpb41cK6GI1V
z3zHs8&MNXCw+t4%%0{zZ<s*w^ol1B?p=iz*7uL4J1<Bm`obHGW9;W>_O6h?FQIQ7Y
zGUTP|&Cx)#@DSjhFTa*mN8`7*@`S~!e1vc)9<8Vdk;#ozZ#eBk)lQ#eL#{C7OD5)u
z@#-ZGwKPcM7mqq(dh>fsvS$WKY}(V-m<W5~QGa^7v%xG!I>#`=#i?c;TplUm&P!MQ
zK$+{d9k9hf@x+K(sn6m(n+z*wa%4g+xDc74_yr5omdnveR4W?Z<E!JV>*2USOneQA
zJ-Qv_6ntrs2XTh5#y&VDne`zY00!b(I#$E#mWwydRNuuHDwXSTio6jZYKZ0Iz=Ok0
zI-RG5;(kn0O{MkO3Q&*Ja%seyjmKtYh0Qd_oloCoQi#Mx3Q+3o5|Yv*d%N7k9Vah#
z=Qirw?Hs6<4P|TcYRnBOA2s<RT>66qDj+4~baJ&=?L?6#H(KBRO>;wvR&^0jix82+
zi=8^R89th!z7ZAgWE-7dQ1}fK@FMj>`+}i6>8)kdbCxi`<TOmjGDTq0pmRsF$+A_1
z6EBA$7p!2e<PQH{pl{%ducFpDAspdH2z(6rX}$qhw~WcMO8YVYexgFsk!~ulf3R1d
zc6CR^BH^*}6?XDV9rwyICM6|oHa>@i!G(Lsrl|_vY6|K$?~h}AI%v)QV$4bXqJIRi
zAnN~Q?cmXm)~ZqztKrz*wpXdu{O~B!7-idB&>b$A8<A>*hhNgD;?9b!kr^xMC>6#G
zG#*$nxi=njUfA`z)A_u^gb<bK3I~6w4;OmhdBg@OWN3c+46<{8Ncz?fHm)&ym(oFn
zl<K0I3>`o!4Ne#3547%qlG$o*SKeQ)B|eOo2!?B()WZwO6BDb#X9dCMociXz%1w!C
zQ)Vlg#WgW@1M=!qC)IPr#E{1wjeIu}4LPg06X~L?Q~dda&X(N}GUWzSs5*>ZbjC6g
zU@;r<d$Orv3bumgQrF3`6nUr}QKq?FkFeQ_$C(Z#6_s4v7OVe<QwtHj(9Q0sP3<N7
zxG0<2Lv8?zSrih3giGT-p`N^Sbw<{T091p_bGstJL4;~Ymc{ylTduEXL@d@&ZI~b4
z@`c+E31ry>WsBMCdsS*Jta1knC}nI|xJ3uM1BIwzNYO`G#PozDor;26B^Su`dUx%M
z2NPKZ?IY)-lsBZVkN)A1Bi{WTtsT87Y|X*42L|&UuT7Ms6dC{FRNI;U!wC{;ziRk6
zv1@)I?-6Yoi^G?3$Z@q)wvkU*($uM6Uv{DPmeRS7Xs+D2dasP6Msf$rLSWjL2XcYp
zH9Be%77AEsy9=8uxU-S4vgj^V*~MB1SP~F#vk<NBi*DhmfM&A;9xF}Tpsq5>Ki9CV
zTMO*YMgWat5}S}xoR>U7AL2^Y3(9fJ^C>>TYJ9B6$0gJ9rKn>C=t*2WU%)P#{ACFs
z+~uD{bpR)v6!x_VU=zXW(mgjyu5UeoFY3K7BIn>pI4acjxYP7vKI3E>smYspsa10w
zD2O)>t~sO|+GKYn&odv86J8&$$}m<a<Rtt)XgPyMi4(_j<vU~SIkmRVAhCDttwGiM
z#+n}Sdxs+V-cip((1tHRoPqPHy%GR7oxTI3rq}h7LzGy`a?D=|<*>3B&?rSff!HSt
zGr4QStdwv(D;l|qA3o&F%EK_4-3gI)O}RZQwk*K9j(npnarAU;MSD0_BQ`)nH2XR1
z$<NwRs*_#&jdp&~JL6Tsrq>SYT9<YH^GP#|&ABj^DfE)R*34kTzKHuEyy{(rO7-p}
zRf5@kQj2Bg6fU{#R8Ck(kgfO)0miS2UnUWm4jxB=K&g>xC8AKKUyE<_(#<8ceT4{f
znAE-u$qZy-%mKAJ)KpX<S>RYa9+*R!@O9KD#+9)iu9W;y#qn-~Aij4MKOz~gF9CPh
ze{%IAZk-&{!qx}+-GKyVD&2w9C*WVG;p9+0f>(lKxtHMf^RBt>`wBB`L0VVvL@M$K
z%U*#kPn!J2jeWbh&z94@PyG+a<RjmWJ(k51`C(+jOG9r)zxqT!L=RhR>@TdMb55t#
zmA4)_^FNrK^?YAqq*ZdFY4RK_eKlMSuXIR*?{kLCd|xP2zK%Rf{h>(}sX<joqivq6
zRup^0?K#2RZP~bS8APgWQR;G9jer4fBXC&~?Z*S;*A+!a%ygs@=b5W0RGw><I>FF1
z(|{r66qNmfa^HJj&koPOe)}f^x|yO_lPvBVv6KAJIq3mheA8fr17_R3IfbXjZO#4@
z+E)oRI7nPcBV~XZHE3e5odabzO;Oi!3InleE2KyZE<!%)xcl?ib}@<}^_VvUivX{&
z_bwI^V|w&*s+3AIb31js^o(0TH9L|`Ij-o3PYQk=Y+$e>GspM4IOkif)Qp^nk$KY{
zA2FQOQaY~SQsHeS{7~u>ms1g{ZGwtuv9}DQupxz*Gc~I+<MZGr41c&^<A**eLw50Z
zuL@X140}FsT9BlO=D}K%@K=%98-l8Q49DjSr*}Qvdl{U^jVg40C3#1sC#QLUO9&0f
zpPKXRaOrqYdKG<1ok%^@SXC>G_$?eA;YMnZ;cocA44*}>ynhVt%qYghWK`IN_`BKH
zuga~gr-_;uK37g|wv@HySYczL5N-T;7QW+XE%oMgy@5YZa!A6tw>{7R!o7%FAU<Bm
zPIKQ4>$WUclut?zeqaIPN;YNHtFT95bf-Patss{z4KFtmZXD1%hf+&&KXxfecVAMZ
z+hQ_f&1S0bwG3Z>O^L(iEH^@XfhCwU_WTy53e{q%<Syq;x;sa+;YVYy+r#`*_)klR
zKor6G+mBwFN^qu>PfB=ncaiO|zqDPVH{Ob0f9-`>DJ+6<<<)w~KE{x<h{V(0UMqKF
z5)G(2F|FwG<=6HM>iHBj7w~ht%umRKJi)%GXc3k<`H&QNN7SXQ2NLCST>FB%fnDF_
z2|^ND@h(UME%4^WUZ!c+&hfj5Zjaxs0?aGHUdj&55-Ah6X9hL4b_T-2ZKxu8ifH1Z
z33Z<HmWe8RKdv=;t{cT|8J8+>YFdG*6x8^qxO!$CF-&XQjgCh%R{rabZ_DzQ%SgFG
zaU1G`o^>RHx3|zDx*}xSM^T~4TZQOjIzFd0w|O@(i&LkhXA%yK75nhHznIl3kGYjC
zgPQH*k}iadJ<qaI=ch(6OY#C<EbxcjqM;iV*_-9H?^eH-S+#AFE^E|J<QgV-v-^-U
zCNUuh!N&KYwu{;!aFPi5ez#9AIS=|&K?KrY{a)NMpEZX4v8s>E^)3X$`~y{U2hF1R
zxI~0##MT`wAjXL%NyTzue;OVK14RtdB(JQclb#8U6|w6Rm}y(W-e8Q!n_NI-0N&S#
zFMUGsn>orkzT)<dt$BlOFm-M#5n?FKBa7c1iY@wwUpEDeqCR)2X6{S%4(9V2lIu9@
zoi|o2^j!V?j@x0t1`~{KuKo=p;GHg9TK<DhW}BUQT22Oy|1}cRu%U&m$AGxP;)zj|
zT7C&<u1XUojX^eEU*0E6GBEbi$oIHErz)*@C}kufrkj~}Ge}MzZ;=+Z1PmA8Bd-h+
zc@mo(H6s_bZeqGkf)bJ&XYNgms%HPgscm3RRbeO{ogK<UXbX0Mbm_CE*)mKi%DEH(
zkrE}aGP8#9yW~&L2BO;?jc0;g5>s~%EMTiWHz9<?%ja{WDmo56q7`L?>w)j$5cUfm
z)37KtEhaS&#9@ezcl89G8u*m{=f(1c(Z!F8C=5NQ_+6gXA7PWwZ%P<`0zosHG0u~{
zFLTd5r*?_AwaBU916i6?B%e$~7#^|z!{H8APw+`0D-f2p&-?C^Z$}>T?OkR@G+w*p
zWz8<}oJwZYxu@M(N{U_QD+!9n&&`9a?Yukpmly!dl3cM*8v?9*1~acWVD@hac+*>5
zg{XXt8p_1q8<m){Z8%KQ756p0!W>3|EatSn@u%?gUsT9e8U1B9>iMPI{MwE4angVd
zDrd_F&xhZ~=tk1U-`JL1leP4kv#b8PJVvy5&`gqEUB8aMw2oBVzwTi)CpfGIkT4Z)
z0|;n)|9LxA8yzf5g8b=|?zC;iVEu<?wXw5!SRp#h+)SmX5a`G2@MtO~rbivrX32`1
z|4~N&*vv!STLAO&BYM6s!sB{>6v8@k<nZcu-Hi5CYP;*|Z+-VHcRExJ^mr7xI7Cu5
zU8qf?8%^KKM^Ku`#jumxHOYt^f|vi^O8OmBJmF?yRnGJ17?e7`E-CdLpH#W)aKf#X
z{#5}e!1NA3Ft(t;R7Ls?S>exaAtzS9;bDDi7bg?;S%<IFHNHi*)Y^>S*dY*$L!_!w
z;qz0;fp`>FtkaK~+e%(zYN0KM;MhkSD^+tAhK$kC-VX>`N?v&M#KV7eg!f(JFAC00
zPEUe!UuczV1ae_gd`7`I(86FGZax#!Xosn|X2o03E|hUjbeiExCb2_shcjf|g3g-5
zj{nrBy#DD*&Y9dnBLK2w=sdGjj7q>SQ7BN>ww>~tNWVYWDIW&;!``2U6EIhZ=4hl$
zLw-UX*~UER5(++s<K@2+It85cW@e^C#2WN?(jsxy^=o)46ZQh1#3#_7`*@A#*tYP+
zUV^9%I``!IrOJnWRsu=M`5*4q%sU*0;&QorR|NV&-EJsyt*46%^O8lQX;#tJsB0As
z^q*!CDvZ>-{6>Q^vMSMAZsFE&r;+SmTluG)sk3jD$$a1a1+xqReUq|TJ10M7%JCS)
zA#}^BKW){|%il>lAI+}#CPzM9exs<=U(Psyxu=WI?%?f`4x-h}Q|-@F6Npx-VYOn{
zXTQ=LZKzunl^UWNi`Vga9qmzs1V&cO+GjJ4aCMGeaF{v%olTdv+84s&-JW-?TyFd@
zxgTI`4Gfrfz<Lei!m(+`BY`Ksnl^F>D^Z(o1sk)nfXxmql8xr8l{7lLPG;JPHqGo$
zfLpB`<7PuPn@ls7g&uYuWp=sb;xF|x@W@hFbavGG4|>-&3hgu$)hMkigN?Ea!DUI;
z7l3X}O;TbDNbRf^$z%js{jDC6xzgDs-u${^pS)`!yCw6xSOzksTFuJfolz?(X(mC`
z-g_}lJeZ}XaF$FCg%B+<Tk=r2km#kF3)XuNi%kiOkADLcVCa#kh$N=3o3?TZ3{8t4
zjN!DXA&(fr)4yywB{`IwCj3#qkKR!3nzGS>mkw($co(WCWh5?IyDenPRbw>Ez~MW#
z$rQi8uQK-gk=oReS~r5I<UqD;EU{c(JXoEz0)g+tmZciW@ct3y{)^{)T1jbN>f8Ne
zMU=nx?Q(HQmkQ76%A48GW-0_1*q+fOKVa@I08Nq8i_s+Y+jxyr)`wE?wtasVw{%FK
zXgVaJKd%t%WlKz~-OG!+*>0ZF|D7;@!#^v1PdbOaGs7U!Q?(&?xXX5DuN7XmoTgK{
z%(Ke&aN}d8=tM~pS&eiLVI-CYu~h|NH88ZVo2P$e?Yi)GfkTNDZIC{;$&LsP`=%tD
z_XF7z7>v~n4dUSLBXLO*GXo_9I5r%#qrrL!`(cTviFK}fO5;U~m)TDgyU+sT$qW?q
zq2cDn+Z;V^n2p)1tGTqM8oI3-_6fv8(>HEvdZPYS7I<XhT*z7L@XCuN`od5A^N-|y
z_UZdIIOAF}t=f+o17g^hAPUS=1ykIhb3R3$K2l1PM2`_o+cC~_rMNO-Jh#`c|0--%
z7OD2DNt|JhE-!R2Rdw!ibgDLhijZ%nnI_7`K9ywoGk<JJYg0oqBs_H3KZ<7HMe;_f
zr==xDGXM)k1n8G{356)go@rR+)%O<Qb434|NRvZkyPb<#irIWfOAO}0V@>k<G+1-j
zQk31=!qjEU{3E1?!@2@$)l4b<2I~6UV>P=QPsB17l7Y9~YbWyqpC`-E!M4973IAP#
zP4ezWT^wAj0J&Db<;m9mVH;xWP3~Y80WNR4-MQc%FQlaK@zzipRsIpHG``6;0pvr(
zxjnPxt`m1C6I1UzP_ho|Q&ugX#ER<M1Lxw`UkMP?H2V~T$orB0VXF$k#CvHKO5>bW
zQL#|}5pQEUo382kgJlR7f{Pf|r{1l<>FFr1olOgv`~5{sfiUn%<X?vUlBU)$gu%BS
zhh`J#$XV%hu&Df5<L}5kLywHJCO2=F6(dpBb_lY1P;l4%$(!BLE)LLTs4@da{5t+a
z$&h>}S`yj)Qx6uCNDS2|YDgr+Z-qOM)wWdI-D%blI!Nsenl=b4sU{|dldw`hH@Dii
zb|g~%aKyu4iBG}04~L?Zud;vz*1Lk0zU5*v^Gi|Rk)7tR;TtdsShu`wzRcy^QEqL^
z$I%}{Y{RV#LAV*=7b6^6{Mm;UTHD8FgIIsZhJ~h`q~yiKbbCQK-#qz(CLOK2ruLtR
zPIMVQ>DM`+4Smrqt6Ou}MtxA!8J{#bHN8W}jzcV}q?S}jAnzwusZK-_!6}*4?AdIp
zB3FSaBi;nqJf~ey%<t=(()Fk^!MTX&Owh>32uQu=4h17h<cidQ=Ax)mL08S=HCg@8
zJX?8vy@_FS0t+ZawcCCl8J<2=wOXb(SlhxQuEM^9=JrZWj$0(C`2$T+jj&T`2v%=C
zqL|E|<S@hl&IY=kxodZ%Tr&@OWHI%{a2Ei#EJQo^-NaP9N-t^Wv4%ZOBjgP6Cb7YO
zd2WGWY+F-XTHr0z^K|uiThfV%b8!+a7a@W5X&Am3@N<8OTGaFf*!@=<tJsYrq1Hr2
zzMZtH+t5%3oQgM^{i;+<&Tf)I6>O0!eZ)vX2N7h?&y4o&bV2FrH@L7Q=3m*NMZtR-
z)m5r^C7V5c9+yZcrm|eFR0sV6Ls`(KDux$qzc?xN`r8cGOBO@zO^OHwLLEWov_!2k
zEYGZSs-MeeWnQbm2X^BPlcb0%bQ3diU`)yz96uT=PnXaASn^?-r3gR{(vhQ@v^mid
zfY%AoY|nqCZrZxktaWin4IA*K=sd=IJQ6Lez$OL^4=<Q@Va04(<ta>MASI@O?-8j%
zs)00Iw7fk?*)Lmmb$JRAQScu3qtk{_TN&B9egsvjSP+@7LTioWOc$;H5mv^$G&F9%
z%sBWVr)tsSF%8m4B$cw0KzCcKKfhSaCW*R^8Ix1O6Z%okz^7%8c+X~^a__G2+3X*B
z_0<Sm(y~=_GD%A?n}9I9!XRIW{G*yCi;Q|2i%{sAOqK+1)n5bz0yBQBU*a+x)}Ol7
zRyJ{v=mZh?){v|m{8j>-k{u|*2@YqT!`6Zv>Lo%7l#cbq!x?Dz=L%{C4(-I`wIKqk
zSLO*7f~XoHyv@;1b3~{FuIv;9gS#Vl-iLCPl<)_6h9PkF*+BKEPM+PSL7v={z6@EV
zHW!-{^~a%&4g7&T1aYhF8ue)?TdHZXh7uT!_U+)gL0I}YJ3Fk)r#<}8HG4x>Li?{V
zQN-mubCL88a=nA}<pC0jt~TF%Rm0m7A2%PCK}<|6C`Xw6B1Wg2rnNQZ(~oBBms|+c
z!%MR#NAvkwK@=w6yE;*&Vmik<-<ob5D#=}12oD=MiL~?GNSVifj}~L4G9q*Vb}h?{
z-GeHZF`(l7{pS-%Ewbw;ibP$#UPTa1m@wlgl%eX4;y2|<LxLeYZ$C*9l2XZjz1D(f
zm~o3unB0EpNtG3{HS=gON{>8KOW;9#@hf#vtSk|6v?rRyB3E@p{!698R5W#>=1~fZ
z-G!{}><qjoEHZPr`jdI}sZbn#_Ybr+fANx@Zt&j*LDT4fE?B#9BOje;<EAk|ja=p_
z+?!O_PLlQ_+PMLeq*|b}&StZn7O~4H(;6;Z6IKu2XUcz`t^kxmu~4qf<38<9I`oPp
z%+k(XOHaaMS^0o4TOfS1HNNw5SNm_5f*xt@hsCMA73z5D?m{VPwlF7!Pl^0`SLybW
zE*0vrB{M)oM)42d306uZS8Af_mu<biDK#&!QDnM^R!JM7hxnpid!4^@otw^bnfi|?
zaBFo))H=#7(QIz$Lxt_yT*&VMA?I1QhARCeN%gC3iZrE2HS<J<N1S}_{v7jfi$3PS
z#gh%>6Ilz-!oONmO@FG^f;<MTgMv+nrp!>`mbFp~Tl@gh^wK-G#XXy1WOu4}Rpb_P
z4MXW?O=C+yw;x5mgv!@S!`9m9ARY<cDJZhfM*Nb$>Xv_1l*T`WY0Tuwi~B;MpZ{>a
zN(B7F*>T_}?70_czmXUty4#yjym$Yifp=l^52vCg^sX)ZA5Pa!VO04)oWJUG*t{m#
zneU#b{C$7n;a$OW_}hb9PVA%1nK)D8e#=+k9}c`EVEDv*|Nj0H!w}KA!`MHZ*bF0$
zPM^C$<LQ;ZZP*!oKDr58Kaw6_xoa@I|AY|iZ~S|O(J%Oi)8>Soo#~ac(~I7U{li&G
z&f~j*#08w0MU7f+H{O-kFkWdpUR|g9gmeb{(Xk-KPCNx3-HRWS`7^FPy5FN030U!!
z_=h9Q|MSj({vVF59p673*U^9x%qGb}$%@(24u)@w`vGBkl-V4@>9xFHQYM_2k?Ho%
zX(TfER3(jdM6$(8niJNtHWZncHv0TsNy;Yvua5(IFh@fr6UObSHc#j78b43n*V^yq
z_lu+#x6yHLvPc%^OMV?@co9FMKR@){-O+}GG?$~K$nc)<Y#Z0&ZE3z4(QoZL{&6k=
zAMw6F<~;>?d(T{X)#3`X2Y2%0d(@<u>Bh4knB%sGl7Rxd+JvN?<vI*(?cOv?lrbfd
zcZMfF7AqyKW*y6wuLW?9!&z5UX~3mt(dTkUgX4^siWhN3YvET83q)J#X^hmvZ7h7e
z*Edr(O`7^6ja#X~;8>b=KqDNmx`i-8LHBqh*WUK?tJ+{$jT{1G9vG&4$f&Zg`tr9k
zwLLcqQFRW^u0d~4Lkkax3P&^P4^_qzt%w}&7F)3Zm<z?Yn9xpi7=SDLcc4^X;zVBq
zp4%G%L^P)eX1_$<+%ge-{hybb%F1fXJgw5~yn)rSHFx=2LQ^r3>#?_amo^lMDyb<d
zl?9TGExnbkSkEdUotV;w{*8ID&VX?vYK21i(7ECATJ=e==IGz6!*lsg#_oyh{Un9u
zeh=yw^wv`UL8ATtXA%F?-HkYJf61%=MYaAc&HJ%Jdjel7YdtSJ6m2`Gxr5mtc4%fk
zYHgVFTv>niR{i^J+p?+Fz``5!@&AD4bF{aoW;Wa5+{HG=@~bC?7~%-csIM`fJ!GpI
z*75y;iMQ!h^KA{FY{>O1^R05&UtuVJK>0{!So&jGItT6bBR!tB$Sr|E_7nUw(TY7d
zi%fecFec%YI_Is~BHff|;T_49r$xJvT0^a8v5AIOO*YYBge!0A&G)Q{>y3jAj#jSR
zu~8F=_-F+cX+H@jm?6~8>W6dc|EoRx$i%31&v1~}s3Ckz06ow}=UHYn9o=l)lkO+8
z)l_{CeR5NDS+$&^1IZE<;UIWD_4GCIwz~}p=W%LjZwX4e0h*|dpy!JC<5!=28D^6K
zGGIb(8Mg<!nX!d%kr{~KS>`Ggz)HxsT#_WvcI)Lzq(V&t{@HV3@is2|scs;1B?q(L
ztl}RsHzgTMO1!B$h4Gd7!q<1mXX{QdT0(X~LIN7Td0iQy3}nF8xYBAfF5)^O2(uzQ
zFW)_L9nJewE|;V6-T!)*L?DI*jo%u-ihkBij6XM!qN9%ZK>}chK$}-8f8l<{d2ECU
z4gNtrWq7PGpT<6vu-&%wgMis!VZY`b+5hsdd8Oach;++V7@ivWmAKEVfueb4k@h48
z(OMyA9E&lHhMx-BLUIh!wW84TMy%-z21*kF3u@VsswGO=-$ZGDWByoJeX9jXn96V$
zc&N(RX`7{<hy3Z3_ZHc6LFv$JRwRq(ci4438)K*h0)t6NlN%ISU;=s@_TyUm#DlhZ
zg%^405lcRtg6DDE;~zFx$q%CtZz@(Jb<`G>D(A?7*FM?BPGHM#1GCjUY;J16d_@;#
zWy3AM-pP{{yRikoCCtj&lOjgzX!_x;8R}I(nu!wEWDpd>y-8IF3zsaj<P?~O8Pl@p
zIQv<GUv+sSfwl#if5q74Lz=V8J80-T^?2(Dn^kivRn<BJ1$bv&Pf3W`&e>_Ayvr=M
zYsW>#Nzaf9@(M=dS}cd=J8rzv&=m>7i0s+dhTX-zy&F3hr-8j%>rij+8y_$0NAac_
z1<l?*H|QoDZq9WCy-KWpPM?W5(HDy7$sMc8Hh)I(he2*Ft&+Fzc>KP)buVr4^B}%l
zz21&&8T-inow8q16}lYsz~Wb%ToLc#`_HwNT{`pvp5FM+Bw+_<R1Ywep5J~n7{6Hm
z@9GP{o&xy4oAkBd0K(LkR@}$`ZH=?-GdOKR^<d&JqU!fLPBx-dX4&ac;h{!jCeC(b
z)l}8c$ET_}_WmCq6=^iI^FQ>~JUIJI&*|yE^_eiKrQPk_QxJ3o{pwL6@8QzR2NMX-
zf(Oz7zo^fo6CCYRr=MvzR`&nkDD$xOnt(s^CYBzY-G<<lq$FNkp>T{;zCtaErKOt!
z$w{_a{yH6R<Ds?xdMn{+fO5?EGI4#{M6s6(UO4vsrFzMce)kj<vl98|VKQm>j?Br$
z0Zz+{uYW!<_8C{{QOm2_FQJ518}6S#IUEr=jV6ieM?9B&c~zaOOVwE?hkte+<nXGr
zq!`^y()6{`oO|CJ|6m3j%R@6A+H4vFH9*kiI9}g?^{S3_x7O{8)6d#ygTYnD^ps(9
zU-!w|oW4hvkii(p<zm4ikBq+w6`{VTU!-C8{5wtTmt&iAim9}W1iX3}q|dmBh<l)Z
z?c=9-ibK1-f6h3B<5)bnzl;-oa(M3h3-`+Ny3Og4Z_AyD_=Eq^Mf6Pn+4Mvj;9Y-)
zn80x>T)N4_k;Ha{`(@YzdI8Fl-Q$-JFdPdesSl(%ZKk%baEK@V7B7t8#ANhu@jv?S
z;Q#MK`Jd1*<a=`R-_<ABX!$=6bihlBffiy#wPj;$<Uqp;rNutwq_%xvtEhKt{ZoY<
z3TvFQq7!d?ob8suv5(KTTh31dUcDg2P}(r2{x_JEC85h%(u!TrpmO0e#w(G7eGdj?
zF}>{?uOD6;G4HO<^(pLf6&t^*;hETqI_F)RG2y(!iJVH7dE<Rd{d4X%4`NVJ$1-#4
zZ0w)V^k;XMxD>J1A~>IA57<Pu`s%KE>RErr!SQbWZ9Y#G>^Nz!vNM+py%^RH?kOGP
z1!>bvF}rdpe;a3i@;;n6>sS5~xUns0Ets`0>cQ7Vha*Q1)!+6V=aUD<%N_H#p(9%7
z_*5-idY>=@;IyWj`0bdAFNP69nrPN~!^N=3&%WM&PM;Rc%HRY0BY7LTy&J!Jj&Olu
zZupqlxnX{zURdHQ82mcJ=($5Kp=dr2MKGTN#n+GcbH6qpd~7h@*J}ViTlGr}o-Pnw
zuu11ZIAXS(wNsQ5#{0#ia<5{TSmo6R-Tw;T#XZ{nhJDx;S|Bwr`<v9z#pxA=Ba>9|
z_{0{r<$#Ku9$$H^XN_SI2+`<mu{!J-lrH-8)jaly$yn=bn>y*bd^-s7@0eF>4dM_%
ziEz6yKv-v3qs5eV$BrNMf&`UggfPfGY9B1xRaQAsK_Xy4Cz>ntL1en5jf*wurXZ(o
zdB~e-oUi_G4X*I`J<ZSgxA7N&9aSa$m9L|L1z1Ugv`t<52&bsq=~w1?Q{TpS1p#}E
zG;e;l>66I%G<+>3h-kI%)k}Q&5&C+3SLhtjrAi78C17OltASy@?Sjga&e#W7TEtXU
z6zl+H0UCCt<oXXr)VwUm#NVXilMYGeYP226k(Eb%KmouD3Ec*xO`PX0ACnYYX-c1m
z`i<pHnw4@{+Vj>>5P<MQmSP22h#G#|TyaF7*~Zpe8K!n(H9DWI;j`!`M{I#v%iQ|Y
z*uw4OEm*p&S~Lc_6zr5pUBE?bmQ2m@AA#Wn!!8$iX^<j_h3UwI)doN?UQ0iIgQYX7
zd$D9+5m2+Q4QY)9w#F#f&c`Uoci}FTuWbYn335fx(t8?1>^CB*2JL%LNB{9(M2Vlb
z1d<myNOE5|N|u8RHUJC7vxl-=+4LYZ(ia{KizSJD;8Lvo`ge;-Xm-#UD!et<i7@VU
zz2pWF?k{+`K^trLsU}v~sEqmA=(c4`)47=T37}3dF#aD7XDpY7a%$~_(0Fl?A+9kc
z+`Cs*FP{j*ZvjKR5#kZ}P!5t&&rvtapfDX@yx_9TB0!emk|oBo`&gr+_L&4p$;Y#M
z-fYo6Jc_yQAV`~)z{A2mB;&Q(b`fnH<0UAvGtKRW<ljclH&EM;W66pREy`itQ`E%-
z4oL>5T%LK(@R-r)dloMY5f=D3<jhDgq%tyKy5slgxrW9-Wj!@=iZFZD-LiohP@t&w
zhef_X!3RuLV>1f0t*Hq6hqFI6#p{unzr+3f(lLm^)QV#Pml+^K_PoEOrf_^%TLIqn
zr97s>WO8|Wq;{z^aw$XyuG5HOf!}2rmSVOW>4s#x6l4ok(&)oo!M2+ZZz7@E?2OKh
zh3x73>O|3{DN|$2a<)FsTs%IGmtXqTb)@6{Vz^vL5{0yzsBH0BKlSl5cLnmI>}gH6
z>epjQXrY88!?l`bHA*p#)?4a8TQ11*_pviQr~2SR3F~;O*~2n5iMYj0)S(^Tluk|N
zG%`P#sa((dLIYE|Fu!GRR2G$NyKbs1sHx-mJ0hVVjYd0Jt2(K#Z!j<(yC4W~+n*L_
zm)*S2%gZaIRNN|QR8UYZw~UDNJ-~VZrC&t;!zr&0xVy;2a`-ii7Oih!r?)^EKKcaV
z+);8=$ej4Xxr3R94gbJ6`Xnb6lMk^3+}3T0wI(@L((xCsx=%Sxav}(13PBQ0J`v6=
zCL)_);+<}pd4IaTa-M<jkK+GU&E1zp)c?M8n&Kx6tFeH`7D-dyoJBqEoI*!G{2ssE
z)L=!a$(ew0n@iIl=_FZwcI<Q~u}ec2H4LHiBF>)fSM`7@^pKoTX{jZgM=5Hgq;={X
z3~;5mKpw9b6?nUH)(Nc1#q`CUuLF4O*4=>RmMYD$a6q?Z14cVDb?>suaEy1JL0iyX
z9|fo{;;J7^PY%>$`_Q-}tSk^==iNm!q@us$hBiE!Ov93z`dwIRk~DHRgw1nLXk68^
zC+c@HyD4&uq$QN5S^qF=ZEg#t%l1x8aq3@!A&>Q`!>Oh`)ZS-?(o?+H$dk3R#43$$
zvP|@vt(r;^(L7@Rwpz|fVZbUJO$Z~Wi6?oi#Ou%ktL3stq0Mb>FJ+kF(rM0H5F)wK
z9dVF%PPCEM1zi1a%zgD=Q|<r1w+#}K(n`bVZct!!vyB|e$c>N&DQ_i5cZb9V3wpGG
zbm!<$NonC00YUMz_y6$y;d_2QKb*%o*LAMf>-l<y`O>%Sy=n0QV7GNd{%<9Mcpj2A
zu0L1Z6?fspNcn_(6AUwp`<XKy<-SU<Kd&w-$N=zg25g!k?v@8i>0doMkNNa@l-Bws
zk($t%=Clc0^jms(Mb|fncT!^s=MDJq6HVH<{HKvt@;~BVyrZL;gEyQp8KE^Y^HLKQ
z8`4ISerB=uw@6kcZ>)cmdsmt4q%gfHML#q`eAYy+_rmBif?ggp-|r}05aNj1?9TVm
z_S5mstBRr%Ej+5e31b}j{?koV?P`9K<NG<q-&<bABsAksH;VcX)o76rYVv%EWHH+o
zL9%qW^Egx$WG|H=tN(lJ8fQbg3z+@)R^=ap>iojn^Q6nyH{VIORa{}_mW(g-NRfog
z(7(4dP!o(x0{&yEN1nsg)5uu`urOy;tMl~nuh|TF%-FoJ971;OLeDQ{dD&3CBub7!
z(W7?sm=oN*=bz(&qL?Y~EP87{^dC#uxi~Sq;ckaIMYT+dhfe@fOn#^Qqb{4~ur-<^
z-?<_0XI9pLbZaGpORrm+I@?x4AfVHSUS=ROpY<VJiQHkZ+fUEdDI-eyQm1W?Q5U8^
z(PP7##=Td!-f6{#Sy5zzlyYhK<QrQr#eNDQl5cJ%)+I3cf_up5eQLr?^V=)&)+T%5
z!Wg42Fm5cHLxy?Ly=11c7l)<acNo)2G)ZJJ2wZ->#rsmrTJyu^!!|YCs{&L0#1dl>
zr7O{W;oo|*ZzkrR)w303Fwwv!C6!hu{gZjpX@9sQ$%P4)wWYlw?VKXe(s81VWj0m^
zh-!QC`uiT~RaeE}@YV4Q$8N81rcVftQ0-ojuibnr7KsK{<z34usM6d}!6sPs!N!#E
z>}9W=?y9E-S;v5T@sSt)g4bExW-RvSg4ql~2jDy<W69ZMAHHr|bY4btp7t7`DMqM`
z^0#Q+1ZBU8nK^^u`ZrjD_OraWeJAb5n=HaxiXfNXP#tf+y^O7t)S^)N<f!kpe607r
zzh8VEXue%1j20qp11gMz`zm@1F31FyO10nd>gPr7MVF;Aa1JJ~Xhp`w9uFT{a4CT4
z1d8Yr6>SrcU&n!2<KCFoDnhD+?(BG4VPwnD3Y7ri>vY?+B#JpJscpnh<%V77H`mUG
zp7ZEIvzh?BRTb4d^q(gV^Dmjdx@f|@gC(wZ^A`te&W&LqTD`ut9(-K9E7Y{3+`rt2
zfXcb!HtXMDx;J^hlUC1v&ZQ?{z#hxvk`4DEwq8&}HKRgCn67v2{xVV`hDxeB-RIeu
zMoNWv8NdBFe6gN(<*4f3RF^Bcp;2-LJLw6xoVT2x?P;@@t&i&u{EJQXR)I)AOgyse
z_Y;hza_sx4X{d9;F*G$dwYnhI?B&--Rv;K%`aGUWi2ZrFm3>Fecs=crk0`dPV6e3A
z$aeV*(l!zo-DF4k@|R0M=35-Z`1h;LTUYs)m82oGSbg)KZc^-1nI+E|)xT;>j45hI
zGsVAtf|eerz1^mpsJW(XAYp)X%_+aU54Uf2p(J+ZW<Cj?Jo_9S@}2c<?HFmpMKJum
z^$Y9XoBjU#@9%iCzqg#`Ise}3XS_>wawGEh)^SFU+GSR++8I%Ygk2Y^Evo2~GL-nc
zcG?8%AOG?SIhhOTDe8T+u{pY9En!4*zLOp771r%=&0at@5;agGebzEfb7j^z+J|fL
z46RuxqbLraVF?<O7W+^BJ(;Heu5`V8`_RBDW-nN8ruuPlqR^ekurZ+*#l36o*YZES
zy!w8c{=|@}xXDVpJjd&SxSk{MPFZN}90v#mp*MH_?=3jF-&QckQ_Kr@?O`I8&dWuH
z_Aq+y(rizNMWn&a^uOt6jR#k>JOOnU*4#9{g$>^%7hmLsqF)Vgy*h%#Z37sh;MUE&
zW%{X42ECS#nva$}6dPa=OhVh{EK*u~ad&TAN8hz_`iUW)ILDQ162tqY*ls2dFtGfL
zD5}`Y^w}sZJ)J43(E>G(Ue~$^HP^y`cF>OA6p^c1GU>Gy>wWXNp}f8*ml=eX6X*D-
zcI2e@*U%UGeL1jnH9aJmy`&z*S{W}VXQz(1`_OA+SSvaXM>Yi(?(^Z&d|$iKJ5REZ
zg32m$URnX)6$HsqSswpLO<oY;Xl9klI)BQGWVA(WwHVH<Po~?V4D~Gg^owFdaRz-h
zh++<Rt0_~)60btV4l{gd#|8JvXb#a}AObCzXIaW4!a7so$hBKkfrRh?4y{9(j`Nen
z?iVZk1L0#TpX^T0>t!Ih>~oEdDGE1A@WE&5MJxREjqp8(BscXKbR*99acT48PpNZ$
zWey4$%49#+*MFpbJciP;owc)^cf>9X9&Yx<r8f9~C2rVRZf{MXB3s~=n#!N1Yi!bf
z4z|Nu&SXUaAF&^s1Oxh4JO=bfgUep^6TYrw?Q^{9>dQ~#wZwXtbXZ4m;ig{p8<WFf
zRE50WMM7zoSF%MPW_A@7aOO)Kypnnp8BFQ^2oW<tUG26voAp?~8G47LBCCOWm(QCC
z2l>#f>t4TIx+#e%jNC96JT=W6GWVtP#{`WG494wNyDPlG1eD@Rb!Dy6PH2X28pRhe
zG<eG~KWNH>(ey7adI4#FZz(M(zdSS?qy7zn-ua?Bs&<@pP8t{tiY9jLg*`iEdwbJe
z7jr{FT8leN5oD?qp4|3F#i@r*70~9jd^Zq}bJI~?-=n16Qd2%P${^tTF)kYkgl}o*
zmQxpphG5u3b3Ol<eBQieEB>yh@f=Kjja{x3lrmbqb0Z%KDpb=SG;`2qF5kXtB(1t7
zb16RB)En)OUaX2?WSwU%NY>JH$ds|0dubY?JUKM-^@^a7J8Tu$PCKr`%mJf#f@93q
zK`5}&9Dzx$Q?Xn6PzEtSb9*<Ov=zK=m*~klC?3S`Zn9Um0Prwtje`JX*#c-j%~qfw
zE4mmxNsp%KkDJvPO6jN;!iO8z=`(GV?F$t#gOc3^jyWuyH>o=?D~@#gVY<;5uv1SF
zkx8<?oAkvGNU&YB;@t|0{>$yN3^$_N9A6Nmow9GVy-dYQdM(bQuiGR8Nh;A+E<jbH
zlDOnSi3j;fLf+uV4&r|DtDm|4J)MB&e$`aKBXg00XFU&P)kT+Q>~yj0N;Iz#V99qn
z3JISIMH9Vd;`{{Wo<7xo``Rx;yYgB(oJnkYJ`L@}oOnP8x(qj_9-HXS3GJ|8*|dtg
zuB-<PS}RX5eV8zutq6HArr%wJo)G>RRNr}Az2C^<Or=2VfmFK35iq)0M!*6`ehGFl
zeN??6Uuo&zTXz+H4VipZpVv~ED(hKrfVq_h822Rpy;aROA@<Xa<Ly;zcIy6QkJ_L3
z5YxA$0EoZ0@(Eete{T_7{)GM8b<|jMBch@7dR^M-YOh~7b1|qHZpu-mkUeG&3_dvp
zouy%z0z*t4SNutfKv(eFMt2)?y}22mS-7BaxFJNh`F_6OPKVf~DV#4gQu;Ku*2y_6
zjdBL^*k2L3T#aX5HGEMV1xiivoik0@G%$N=X9W(f@Q|XSFA#wvE5*NTb4+C@<Ym#O
zVMe+wM>Q}!@G4MyiH>2n-@`h5F3w7FNp)uU%JQ40e=g-nuKH4O{+Pmqk%?ExWBh1Z
zzG+)!fD8@}>#rtdOjz8@jP)Z|q7ObK|9yYn++G4wXqai;%x93uEYlwD2q<v$Y}@qk
z=U;A6Zz+4x<~DHY+)na!D+fv3gEdBAk24eR=iBv?s-g7RUwMALd9~bq>DhS^oS=8N
zg~2J9=Q$a1%?`FCdiU-%{SfS<h+a_!$z{!%qJU6kWW^;jz0LnY-CbZTp${4$E41Dw
z-M%_s?n>>bbp$dH3o^*X)g|9O+2lEQvWl-^k+?&5gqztTcnllPVbA+<o}Y^HgNvc%
zpH7j}2NwOqFH5@+GX-$TZ#|{6&b&#9BR;6WhLiHCn`bx0utXIL#|_gHK9(M^H6>po
zC~{8$&<LxL5Z%uhR}FBe&yMGe+4Ebv^hXYKyc*Rm&iuDdCSN)wO#kQs(vg&<P_fRe
zQ<aM|2088QIK5Kf>d(shcaRjrRU&0|l&B!VK|87enh$x_=Z~+QoAY=+H2(J%a8V`O
zT_sU3Rq}fH=hFw?N{=!0mvN~gIj4v1@0a_XnLQe;Cm3+1ePn^`5I7B;pBuy<xJ8&@
z=P`DyaTS(RAq@xQ=zs+=cc}K7LRi<8Wt60tIiH7GHtu^^JkX52Ks53R(YAO}iW=Qp
zunw*h6QZ=ckbwFjBPu@Z!lJkqjY5wZqx32#)uY#3xNQVE?8;{~!BAb7NkwIg<0#<^
zEEm^`z%{EBJSG=4<atNR(^mSI{JRJyH>X$ULe~0~{lVMK#}sNMb*|5=<}gS2@&?0B
zOvZiBJhq$ozHJ#ruuAisN%<T%a9^&OFWOr+mc;ppyFy7iGqMg^VmlH~{GICS>}Rl`
zRb7*K>ztIdT;C9Zbjx7R(kr04&CBM<UqMo$(P_l`REH2&AXAYx&DS=7{-8qzM*5c`
z*vJYd<NQoJ^?e#Xks-Eim@0tnZtnYQL@UP**Lg~*OpC!unXhR~CNAWTWQ@D5iO|Aa
zdzjvUvM$;TOXJLt5u4R$v{1h%cgK6x-vl?enC;&Uf#vd-<WnWGu&Zm_qir{Mb9{Y6
z)Ap=yhC$`o9=6ILx59qEh$G_N31M}M;QF*!QN_-z#;yR}#%;NHwjYWRku~C36$zrL
zC>_j-qH!?aS)!h`o6bqok|vjZlASa}OyOq-S87C`cdgOXPxGQUwY9J~?g97NAPnVv
z2&YW44`Lu1m#$K=c*k$&DXh(7g)#Q$;2X;pm46qTHOhd-y0VV$YUMTztSy*B6d=G5
z7*BRkt;r-Y)#@)O+-w-TlIBE+^4C}r(xne$4hg)0$08_DQ*~6bGkg@4#z3f55REs_
zku@a#yD?ZxJrfzbq#Y$8!7v^nM0T8)u+#JYg1?p!ubErDn_Zj!eh=BI2xY9e|H$ln
zh-1lUU*4?2+@yQjSZe}8XhVUP)zo%Nqneex%>CDRmZ5t-%faXbq`F}6D*;i6i<3U+
zy<eNw@OEKO;y|+8Iwx|o(Z&ZH{<RFW%fJHU^6wS+Vjid>i@EKvvQ4uWKK)xAcCF{?
zX5|0-h4zeHHk?ItFx5$jNeh8Y@-9zetuO+nMo~3HTMsLi13YBI%}_#0q?74=%Jgh5
zMj?k{&T{PCB@2rIHL|v#AY+`G%3xtnBwoZcMn(?}EJ`G+E7*f?Tc)scC^MySsYhNO
zx0UzFCj@^1Ll=)w7oW$V`9a!x9mfetPETm+_>QX0_hhLk`g(+)G`d}AO6E+jF0jtg
z^}1yNIo6usyqp)<37a2W2GvSh&A~b~;1PFYi)(=W7q_PfTe!mW@{s3y6_3^tGOEsj
zG@Ch>ME1EoLhK(>_bZ61w~{af&>tC9DJ=EI@GWONoDw`N!t4(?rhQ7hzIstOef$rq
zw;;Fs5*8kgkd`+CWf27>r+2JK*^=zA@R#DdCcNuz3NZWwnbR!#M~$$9ri8&)gA0Ro
zjC-iBN!CZdTOkrcl&&qO(U$kz&0QboR;YP?K3g?L_n_4qMOI)*xEQ)47O4sIsWsY1
z@s)NGOc)VKa{aye__~>EwU96~zSHxJeQT4no)+v|?VuW+IXE_<Pq)%v@$NirUa|~F
zojG;_B}X=2QBv<TS)AkGRGC_;*Ep3$FOzv?d^V^_C%#Ajke;cLJd|Zl6)8sb$c*Jv
zL5ykikbBlUF*1i)Cwo90?IVVl((`?PtVcq>bZ<`V&~79d2-}hdJml=uY_+sdqNjq>
zJYu@U!t;O&=Qmp&f;en?HDWG=HUA*v6EcbEuK+lXfYl>s^GI=~3j!mV+iJ+9f9r~9
zh}E1ktA(^BCe!DsA7%iFH9EunITfaE|M?-U4Izn;ocF1$f>Pr97T<hQ(Pnqah7l3#
zdVR?(#mXsf)n4RuW}EhySqAXTGRedVp<fA?TUI&3G4kmWd`>&4#RFj-cim+F*hR#J
zI`SSn4b$ROW2fml_qV`@M~67w;qAI0lj8)teq857C3JqAyoNCp$I@GOFkM53feO6G
z2%;zPP<o;8VqhtVf!1-cAA~L8>6akO+)<0F{<jWgZ19+{1}T{pWD>e>DwH;B-KeD}
zO<;Qwi#yvT82gvC*>eide$3=6G0QD3&eEBB2DH+Z6NAfcNBznR>pyE@@>qn}At?II
zDrAIc5q1=iju{^_?|hkMgEE<Rx&GXUe^Orc&4sKi0}V>h&EDrhZ{=&K)fNjAyS~6&
z&jJaPH?#b7Z#T35d3!DJ#`zwp*J`Ey&5R>)U_R}sz3;dI{T?JK4ms3*+gNjdc^m)z
z+IzS8?=42}w|j<;r_|1WZ=FlM{Clgbic}448xb#oE$hfjp@sp2PzlVuz@z>{n?~{I
z%bIDPpz<LIWx#xAHOrCoj?C{RMwy8}RmM&GfCRKTAQM!qPpZLRn1JARAAU<nzm2T>
zMi>)lsW;OqU1fNGz_dV3bXh;2;@(gUdr56B_(S&}61!O|UN^=f7dy%Wimi3w<wdV+
z9Py?JKjk5Iz{+$!ReO*ux*vhC`exs_y901b+$`1s%C5nCSIbK=`jU2B%@)>d7QMbz
zbl@+dOA_@qE{1OWZ-NHflgWh(`eCYkmh?)?$L-#?ZSy~WMz8+VqWu}(hjbXtYNrx{
zP{O6E1jojkwjofASw%>ds$QF{<$vGZyWA(?OYCxca@wGe4lk#Zp7i9Tl{_z<_F82E
zk!Q^D8FBSizyvPeQ!s7_O!lQc60^<@B`+6b)SJ-1&LHs2hZ>JhXC>6KxeKXR8JP15
z5q~l>Nu`^cPxx&;^rf*0`p+=kH%k}8`@{~syGs>U$EnsBQm?h?+hfyT(7ECl+V1%6
z4$OZ1KIYYGpb_r|I5|2)KNzWB2Tj$eu$h;Uth>J)&seBm4O$0zzWJu%jmC=a8P4oW
zZc0(r=ml87)yVP8<yicAI$^z>X|c3SFm}%`K`6@mzECk{astMvQE4!HdUrLWA0iCL
zw1t-URq<H?C_}a#CKNrbDJ%xl3LM4$Ie&+l+Fm!Qtzxc!4Aa%ltpxGxoxFMx!-u=i
zMke5GB|2Zv=9++%yy-XP7}BNU&PycOjAw|TEZKdWE$qSXk1~x?ty{f*+^Jtmmflwu
z*iQ$P=@amZ<kjC9JgUA??)%3>TXBX>qn_>jtCO_w2HOmbitagN=AIbvWvmgmQ?N6A
zq7WVYj$W`*UQte5%Bzv_`8o!mKu1HdGs0kl#h>rRL`cq%-Nj1Ba&6|5#kS{VV=zj}
zVc~K)t(?@dDIt?}iM9zO6-1AG98n5=O=TqA=coy^2ndeN_C61bup;!X_ByhXiEre3
z5pS?>0-7y*R-e~@t{i}d7wUb`4HrrGZ8VZ8_01X&UQjd_-!Bq2Nwto!E-BTR5frTo
ze{_l5b)nszT$W>jA=?}$&op!!<#W0y7p#OB2z$9B83Nw8v+Qv$MQEe7|I062+8IY7
zfjR~{Aa*J?g#nqas*+G+ZfG}X_@VdGuT1jW<}OZ*+z=_w=RQqF_MB@R^{1$Iz9zov
zQY}YU0c>ri#nU4C`v~##;&<{dGpu?0jj^!8NhBfE^kb#;a-@~nkq{%X+4G@a`RYlo
zzZpEOdjIF7VX;S};6u<5;~WgKF0+x;z9iOTuE~_;q7&3?D0McRYrv`7<J+PYW15$h
zc%Rbk@jSwB@khLF(s{XHi?p>rpQl~rC1C5F1UNOKs%Ax^i28^Q$z!EeVIL&1M2nIz
zjWW%TpyIaOvr@36S)$~#jIW|3nxZ*7p=?O}a8Pw)BSXl$90T{hj~$mPmfpxqpb2(b
ztSwyUzc5d;V}uYU8PsV1CF?V_Wn+8X-8Al0sxU>QH;Agx^gU29+ti2GNNE~OJ;;4!
ztd;YKd{69Dw4X_zWmI=^eW_3t-h`3UD~9!F;I(@j4PM9c2&Riyr<9G38xwic5Z3k(
zJ5O*OimP7A;haVp-D**3a>i_2IrD{<;y^r{Av;Fmo<B@qHoz{Emqm&Fw?W9W!M83n
z#no&;M=cz<?PyQ5=P3KlqruQ(YpdJ=R>p`NVf(i4tUJ7uw@<3)0P{DjdQH{XLXLd+
zP<;u=ewud{F#p?{_dpf6X;@{P);~$<#STmRA(;BTtMBRZmt`UCJP-VC#a$RnQlb}-
z0~xbv#wU>ROF=P9FalgQVl)o1w<y!su6&V|j?pUqFxpdUsH5M5)Js~*(kyV8e-gXv
zt8FRT7=`ktgcwo!;G|^am$D4C;u-Hopmt#DF~WPYn(j|D<Nv8ef8ML~w%Zw#u*6WB
z+Z3+TzqEQeI$<nA@!@M#V`PqhEK@47DQXVVD=gmVK^nNoGB2eF&E2y&n2m?rr8ixy
zgz7hy*le4@lrm=<ZG&@sY)WUxi@*JrsJ+P|<f5<U^Ch+|SYV-YQMD+~RQ`Ag4_KXr
z<GE1vm^msf->48bW}q0r)5s2}+qXNQ<%aL?hw!D*RiJ3DiG&^Y5_>CK5uI$bj<NA9
zuQSXXGcz+YN?t55G<meCp5;JG2Q#>Ce-;osm*_f5ZS}0LN$Wgx!cg13lSM~2HT{iN
ze~z;1em3=EOq`e?fHU<XW&eVJZyULJCQmw)0`k1v2P0QQWn_!s%=AB2DQ<`wv){aW
z8D!s{$WCgIj+>l<6Wa#!7*L2Co8&v6Rp~3LXSQ-b<VT|k*o@MI_6_BBt!$5%^vl$G
zx-g*g64Ggkq%TJX?Mq=8$TFEsUq(I0cRgx-8J`s^q50EB`^p_<&WDgk<i<RGR|nd{
zGsLZ`F;aSZ;-2SK?gRRwxWS3g%3<|<Qg+A3LjsxQX~`I<+<x$9aYGSLN|cyo50PfQ
zyC_*P6A>zqG-ox?24VZSB`^0!G+uN`hXzXdgb*r_96Nw}2K1FOZ*r3j3d~%M+MHEN
zTdx@Gqj2xFS-5{vJq^^7k`w8+>0{wOk1}i3`?rxT8=fV?*ZF7CrqT~cC28MVA*v3L
z3ATL$r8<8yHP1KmjQ~k1o7roS5dVC;AH?Pze_)guPn&6&(rbQl;FIxdIIC3M@iS*Z
zg+}?1zT5dJMEOltuOA^}RmW~+W`HlXH%F8(bh@~--=XT6ZkVZ)=-wAZoPr9`O7p#?
z?M;tjCeoDJfZGyc@TpwA+M!Rok4Jyby<?RVxyszuhSJt0U97DnrMw#E<tzDawDn<m
zzCT~qh7?`Nc#w5ecz+8`$ZW|p!+Digu7v9!${ZT^)##Fxbs=W6C|C3GN}b&w)TV?m
z2?;V&r!EE5EaE3(>mil4nCl*Sp${xu`llP1pNG50(ABG?@9oB}VvH2<(_@%nkEMj0
zc{(^t;>+N1!((35Hpf|$F3VjUA7?3`!ihP_MqvWskld~Q_ZG(B<n1~YsU*6}Pq;%+
zOL%=V;YWhxL!?PV_rh1{pE~m19{kTE;dSZTQ$>t}klG(&lM88e*^(glc(=zs@DNf7
zCR1vB-9`R(ra;q$>TTOonzzf?r$0h-mELw0X*b;YV?$8e+*zPI`TX~m&q8DKpYK&~
zcaN*3*KQ{LNLDo-S$}U8IgPb?99QUL4vWSeOjhJ<6Xq#qZD05bWUPK#2Q4K%xzdE#
zh1quAK^#maqrSvw&N3UjCsL~_H8L|??trV;Rt#NNn4bUp<z|V6o3rAgD25<CM>#F+
zf;7eWBg|lcw*l-vc^5SirWUP_+8=s`gg4iZ0#5e)Nk%OM>i}Mp{S$|YxU*#^Yd*0J
z`j}TBF4_I!7!i}gWK%ejvgv@Z<jr65fKgwx=4q@yABywn4f-K5P0EY+8~UVCT@NPw
z(OG|dSi9R2!OG`UrhueUrcWA5t3i@YIPiTK$o5T>=C5nkkb2CuyGdEkX8Af)!2h9y
z<wE02Dj`I}F+`=ZZ7iU)iHqb0Uc0(bifCmBMy>NXg=+~r&aaFaRxx~k`v%<MU@K<)
zt}`!9eED~#ezRwNT+4O6?lN$cqrWxONt6ZdH=^Qc#anMvBX>-Sw$S%F@hrgRN#F)x
z>ucgBFe0;v1h&4qUp|;(%_ynPhVdjboluz*ow+3RT24lV^B#B^AN`$$c+Tc;Os>D+
zl-GG3y|cn@H+o*2J<-NSRtSE*|4}e{c+AaiYP~1nuo|1w+FTd*r}5fQHfY6XczJ@=
zX+^9hEH3EiA9u{MQcwgCsoN=$h$6V^dzyLX|6=lpN~a{qI*$a7yL$5xRa!8wa>BdA
zM^uS2mVhNTaQPa~hAn5=a8?Ji@}Z0xi!b%pn8pi>CZE(8Xq>knrHkSuNYgE^^5Sae
zo~d7Jx$vWZ7UGRdOtui<EK*vV@e6l_5iG!Zl-xP+zbm&Rw^9$_u$=ePJ4T=XB~gR^
z|GEAdR9krRC%NxdA$7w_d-E7CG1>Mq=>0b?UH_+dL_4`z!J{2gz_si##>scN0Zdy4
z(@$cbkshC}ce19^;J3*BLwbf{S}yMWg~_GuL(W1{_+Y!o`r!Tt8pfDUW7jH#gB^05
z-^B1B?s)OxnhM2a3~pAm9!ewi^XFY3RmEQlZ&aTz=h}3=emlmLylmR={^#VqwFrx_
z+<1C6rSvGm%1Acfy>9p6nm$~ke=X5Yd}O0-rOr`9Hr~)*11ZHlHL$qtP@Q|UGUh%`
zjX9hfIvPxM7wi&Ns-S_lBwKXW$yUYs0Q}-u3FKjFYk*KDGPHK=by6FFV1n%R??<5I
zOYDAzWRkeI_qFf;zx^yMKK9v>oUse{!(16N=dvF`mu30`fur7;x!&_nTURv`ba2m`
zv1dC%0O!1|Uvl>#d8?Q8Z2AHfe5U<hGCZE$j5*ZwY(ARu<BpZik2x!OnO{9k2z$b=
zj)DmK?b62{rIg8bnWZ76c!r3HzU7=o>Z!i=OFzO7)2U+aO~a({ghEvizXYg|+~*xX
zl^LRbSs%@d=2o<iXMsC@1!@5yy&#>5=-o<OhDw;GZ)z1ZsqVFWHY3Hl9S`j1O4e^1
z!<uMD72;c44$<v4GX_n9YlB;(QRJ49Q>4m<m)-7Fpp9#Gh0X>YHh!x!*~>{oXYt2<
z*PC=1eqPYPB04HckeJDkuta9H{Kv>TTS!82&sB<!Oz6HekxSe1F}IeEVW0~+_u<?j
z(7ED%ye?**kC|4O92B{xr7WNr&%z@pPs7JCq;nw=PjeJq1yeafw3;#uWrO<d?qBLK
z2M?~=^?rQP;PWjxw(V4O4{Z&)rx!xK8V6Rytx@=fNr<|X=R4Lk@iF<2tIpk5`i`RR
zePW;pm_8PtQE_A6oiEYt&JbMwqpp`N6~|7h#8X}1v=sTyC$sp5vJOd@c5Va}Y0OJQ
zCp8OVS6auvutHDs!exe}7!+E}ykKLJFP`-x8T_qx<EetVf6@f(*G+t#lZ?qxK2gY8
zebPta+eKn^#2*m6c9PaSTS$RpTlyevgg6GnXQ7B<;S{SV7~cOKo!b7<W@VY??5(D?
z6ds<e%cQRbBHQ<ntV;lhp@)ni?9wCd1ZO>i389@eUJ(W9lzBGXa{PU3sm9eWM4Z1d
z0;YJ_#9Rt;Ye*>@%tCkToR;YG$RwTPb-b!5xjNZo{5QC6DjJ333;&ZC&3gIJ<(8Cr
zk}Tgp>pOPdzw$>PJ3=;1Z7V}Q1+ZAWXmYavP2|!al-l;w5G&2e*eP5n3Zz*sY<R0w
zhKMiMJ5W{kcYEpg%xN7w*ukzXMwF7$KUI-ZNEn!p6;wrZ^)fR}zYNeag|LkqQCM~m
zW_ffTZ{QfFX%YhQC!P7V$3ivcC%-cRjbbvUegW!<(!=!GuuqxE(#O;GM*Xp|_<whr
zA*@|yP3QX9z#lR;FP=Hk-8o(j>BHX|ki)Mi8aA_3#<|I^n2nDXZI>_4-X!`1@bXS5
z^*Z`he#Z~v9UHI1di5A)kM8awx+;$Nr=%^S0e-<MZjZb10~wVT#Y$A!_Xv8ze|+ru
zo*JhY=L6$KjN{3vU4&>DmDbDVQhpYFvENuV<d4{u%6T7tcWexwj>@H_C?%;pg=c4(
zAqXDVTKlD%gasvQuJ8<!6e4~`x%}v7H^r8N3jdwjb}^>&xi!l&i-Xi`vy0d>Go<Gy
zu$H$VeS4N)gH>(%s`Q}Q!^<jNW16KZwHuBrv-7~|k2mSw5Q$F=c`B+XT{JMY_XkrP
zukI=KheyUcT2=&zJhaA%@_0V3I@fKw_|rnQ_WQ4YBuyj=Jt@hH4e)wn1Mei*>s*+5
z79TFTy3y?7Zu>|Zx^P)F!BFQ|&K3Onkzk2q*+cV0rmF1UF2VEQ1x*4OeUnGDQ%D6A
zhT>vr-fs+|3FZPyYW<@7BbZy^O)j!2+Ig+y`hf0Q7f$Bq8;PT5;k6O~14zoN4f)K6
ze&793{pnFGRE?7AWm4a1*p(>l`P<JCiQyl$s<BFA-e~cZZhFV6Sts-k>d*1rVke#y
zC@UdpvD&SQLz)r6@7VUG`WGtISCN=D-nyt#6biww->2FB!Cq!Nt4~`ME4g4H#JM{N
zEL-rv-Yw^pj*lzhDOPAURGxCP7e1zotT1cv5W09t$gT}f`;#lquB5}%%lH*>B#Mf2
z6fa;_mf`4P3}^ePo^6fo$%$J=QRR!KB}8cE#L(D?@2=r3A1)ffIu1`mX;5t(O1ObC
zB76?6bHc+7f-=5I^LStTVK6h>m!<xPH>r8t$*GL%^TfRj<XU?2U151S9loVTb>yx;
zK=IE&E<J108EgPunzU9q7CHB&WFEI`USU~fo+BG-6eXOH@BjLB9!0+NYm6%H9DWY(
zznNv~Y>n{p@THtW(6h-sp)(lUaBY+~^DbWp?GY-T|GN*CtzdSB$QV&*>ZI8B9G3Kh
z<C3)+Kfg%w9+&&a+zH^Xpu5*1<vX70z+>~zv<<@)_Zau*begYn=8!w7V7BUc$W1=f
zeB0{jifPoF^Y$7ysRyK=P?9Mki?@~h8xC6^FR}x^?QckrHXAlQW{Jmml?x|~@2t2J
zKV>1N`M(tr?fMlp&TV=x-Dn(?O(TT<%PFms0;Weqb6aqYJ<rKZ^ZT(ehJwT!M5jP0
z+Y&8Jxw>Au(R5Y*PI!c3kyR}yN<-^_wl;M1B@sg3Rs`l%Qn|``+|>f}lC}ft*ZlKl
zo36J)n%)F7Sz6sKTVHcy7JtOb3+%&Y_UeDgq-h7DN`Z}xnWAro6vqSdqIgzt1G`uE
z6iX!`S7uu2YOLq>=OzQ6=NKrUM%_U%kL)j7Wc5;&)adAexSkj(sydGFyX`!GZvjxV
zpKGG`LoKw9^6ynH^v?Ru*(Tl`29^ULw;~LaLFtu7vK5TZ^z<q24TY35W04nl<Di^>
zW0Q2owf1pYyUm&-H)32XuWAomZ`T3krT&#)qd&=&HLF$&SZP?>BfnCr#c+rm%Y>ua
z+rKMT5Q8Q)s((PKb;~-QUWAy0Jlm{s()t?XV5=Kh_u%Zvo6G!d>8Tq{81YIxgQ4K!
z`#46r!|uz~3!&rB-&elW<s5|wX&lMdGNuVCzOcrcZ6PGzW?Uz~{(hA88vgy}VkyAx
zyM#oTu`hDjZQsT!uWccq=8^J>qu1m!e*=Z`O}<+Wkr?^>+uM~>MzQfS<t6cUKo*(j
z<qXbW(HK4R%%Q@k@u;m-+AED0$FyA)OEHUIo`!&QHpzZJLDoh;jf&vgoyoipi=E3c
z&O_Qnm?nzuNu$gF9uRd8S%B8j+C5X4#^Ll{b=Brzo<Z5}jfgCznHEZZzd)ff1@fZ4
zS_cl;r3@x{xaRr;eFi<NOBIBZI_kTcw?Ys{sl_3XU+uMlx9YGWvfV9$^V?1`=qi3F
zXunVaEK3<tW&tpyb7JSy<Ih@&u4#Nx5qtU_-sOaMIc$XX(o?v+>}B|cH`mR_>uUAV
z0P3FUe@ZXV6-e31!Wp17btYol#&k?J8q6<?l7?=^9wlRPuFS%Ha{1Nxsfs12tc6@w
zL>D?v?E@-z!@qMV`J1+i);GScV~|p{DI9OQ)K6J`(I~#jB>5*P4VN93xAxrt%D|{^
zO4Rddw)l_+HH2+esJ;%`W-;9->brSOLHT|O3jQc|PPRYdx;9x?JQwS3skvw3D1O5g
zz?zrUH;t>^*)(j*=LA`v&}ITL-B#fPCFCr^I<*_;_}IdtYFyzj%WrxF+k2)N=-IuW
zx-G;sT@}jzC^GzHdo52tAXJ?BJ}AAkew2OIXoJOE)%Y2=IhMiuYl_AI)f?wDulS^Z
zO;y)7xEOOAe50F^z=2;#@O&#z6?r)XD}juN2q2oJ+3bZD=%dkTON7R&u4VJ3Nrr0)
zhc>%^1{m-vc9=P@@@Ua_3elyYQXP%dN-!=X$ZQ%4svtVkdZg}3NpNvLQ>Aubx!ojG
zRL4wxQL8FbvDs$O56c%)61;6!!cws7JO1zD1gQWqbDBix4<t>DTgg#k0?Qint{Z7b
zI^YvVYhzD_8O0jkU)E%=2B!OUxK6+3QP3e-2^VcIIDuiBx}{Y<=G%UKM_fj~((g}~
zPq%C&2c;jsc6s><${(V;d&QUPl_d5lLl30*N1qCx#ho0>31aljy{!e4cJz>pGRwLM
zN+^y@$#kaWYE6|XX{5n^XCBD2D|r>%>)OcTVTZ7v@!}7$;?r%Tc40rwng6W3`rx3n
z+o$~uv1B&Vy3+JZW0_xjBH$6xd-RbP9aGsdSn(+QOVS(|2VaVkUCT1r$;;HkP9j?>
z3=-=NBCMXQp!h!pKUjY7A}4)mLFY%FRKLG9FQSfD9|eXP+kczvL$^`dL7kszIAG1B
ze^t-3(m_NC4VFna4^*U-yB@%X+2EQn(2CJ<y-{FJPPV&X-+K~m&ShU^B!z1r2D?4p
z6kOF*%Fa6;7<uoV!t+5^+^th7G0XEmo34G$e6xk9>mE<qyw;;-r|~<bK%;LF;)&vb
zx|whK%y&I$6B}%tGmD)F)RzHpi{(37(jtQEp)O_nDvVtcmf2Y;HDfS&e;u*Xd45+S
zRnl_6)8Y&gn3sxVq;`@1X(<NlVEW@M*InnCCw@a+JwC<1mzkT3k=0k_fy?T&-60{m
z^hBC?79R@z##HMCHx1+2csEpL$d4?FOYQOSdy6@LZ^g(zh(BD)dB67Qh@mZA_8{FX
z-{+uunXb*Bi1oE+E;Ra*hrVJK-JSkWu<Dl;`BUGpji=gGOpmDoXqi@~hlA7>=ou*N
zM^XMrMvaNsMd{akva*;Q(HX~TWMaD`o-}1fVmR19FjW>w?vAJ^Z~AKC59X(LGsuu{
ztlt|53rOiwayP7@cX+-m^Dw2_-&Cd@`})I_&J>y(s*y>`{g;}*LPk|(Htyeh*%~zX
zt@^hDPP=S@?O(i_Cksj!Mcf~frBU5QiU<nKF+Lo=JQ51Za7;JaPj1P3_FG1}9qqxg
zuEKBs+q*2ziS@}4uu?|{!so37@R|BBX*i^i?vd7;o`#`;jcA61Fz1XPuJ&7SH|7;J
z_T_o^=D=D_KTMe6ym#OvA75#>vySkC)~S*KV04{}y=J{E7uSIX;(%-04Y2j&wr|e)
ztJ*-Sy6WLi2DYB_eeKTo=3aRE1(Tyn3r%=gc{#oIC1HX+JRGNkDF-gs3^{uo8D}Bj
zY1Nd|=)?oxGzOL?iYmjYeL6ygfoh?Qlt7hsf4bl5M?CR^0`!Ven!uDBj+Q>pyxkxc
z>q=N&#S2V1?fca3eC^56;&F9HhRAYhglf6*Hq3>j=f(-cPSSmr-Urno<y8WlzKH9$
zo^)Qe9tyqlW{82uwT)vevw@XZMc%d4EmV3oCDFZHlGs;BHZWDC)BdfjtLM8xo9-t|
zIUVVQkTO9QJK0`o19PKvj3Ep?%j<~C_!XQ$_;}z{J2uwtT%w-i#$8SpJC()_UFr{U
z-~P$#W*p=kV}Y4Q!$~4s`qp2VhtscmEAUJt<&f&5FJv!O2k~+WargxXC>DdCwN2=;
zPjcPQsfs2^Fl3MKZeT@z>((79T}2{=O~IV>ZlvVm4K(LZt-LdT)f_M*8PWcml^mHC
z5%=2P<;M!EJ$#ii;WU5@W1u%a>lMylV-Np@VTjuk%w1Ij=a-Kg=qbYJA07y?D3|M`
zFiJB?K<G%p)EZ%nMO&y@PB128AAesn=01v(-z0)c6bp0|91ijoU!r09R>viiN>elE
zn~#sjZtDZq5$7ibKCSkj1O1aNN1N<3Aq}Fze=s0(K(CJzm^A<=nkr*Rw-Dl)uFc1a
zUQ%4~q?P`n4#w*5-DmkNKv;?s-+9_MWBp><kxUt=iL!8dW@reT&5aPTNO@=Sj@+Kp
z-nBMOS>Bw&(J_XRM>;l1OOy0E&oV9ue2giD1*u4_nogdK*}ih^-Wa3VknN4mN2}wA
z2x#Ru32@uE8y@2a9GU=jB{wq+z9_YQTq{MNPSvi9o`>;zWBBUh&7;Q_w;DXHQ{4E9
zwN6X)jM|PXM>U>G0%ms!aVVp7_LB{5)GWW?bgBA0sWu!o(9DXx)H`qdlV3~;&O3B#
z1k2>br+$YAjpt^+6Yul2lDJRr+&Dk4)$d?8>z_#y3J=9Qfjaxm?@D|Wz*ml72g{nX
z-TQ-P^CCdA%s5Ip#lUhjz~JnoVMqs8Jd_qeOf$Hot~~7`6Or#ui8l#I^)st2(e3me
zbDtMGVq}vSVz<ujsa*3%Scz_VHeEXhQ+7@VnNHi>6KILCnx<s|^rDydk3itRx3Etd
zf7uNLcqVi&94R?(v?1h~<Mb+4K!A?$VCzI3eg>*U<5bfO!~FEeOTvo@^)Ffo&oa?h
zX@v%z(5XFdTq71pYt17s#Sm8>1-NT@wRf8y5xFX;W6CYWbWzTyDg~u8O%5*=#o0Ci
zFIH!Jd3l|kOSub%U1`u9^EOsv5tP$xiSv+VD^n~okjriXWH83Ls7eR1sH45Auf2)M
z4(;V@41+dp7wUkb?7M68P?MH1n*UZH<2$~MmyZ@qw-^vM(eEWa8f!F*iEI{^=5XA@
zb$9)^Z-j`_y<B05=Xe<N@EhF%%LdbJSHfOgMjx|dV$}KzD-v6>Xh@JzSip^DT)WS~
zbL-4WY55Up_)jan3ktsGnw-$rRS(tiJ04HvE0es#P6r8_m~|`VY6n*bU`0|bxFs)Z
z^4hYf*8_Mq#>d|=Rc|*(ol|bcHs9vbE=B9@W*N4dwjwH%Fj6@Q*>ksZpXDl0OLoh{
z<)9uN+u0NDWRWg)H}c&rRM}F3{AL4-HB{>?eydjl?g{vF9^5d=!_);TMy?z%@;s+7
zhSBNO^F(G6W1Wg#d%rGh&Z4rXYzFHebckfU2PjrDO{Aj2_9Mx)tcAz4)U{g^ER!91
z-DEz`9{}7GF#ZObuWBL_vkv|GKgxIGG{l(Sd+zofj6^t2JrO@S^)`jHASj3>0CoGd
zG+x2durn4SdrXn^zK)09RLm4R&&`u7hv}qS1Bx2>&2i?*nOX=tc0(^RQ__XLvqVNv
zHv=ihaW~(&c8t(~hDp*k(Q4;JJhC#k<7V|YHa0o?5$!e?=fr583XLG4a@na7jLHgk
zBdB=#w0a#j@Cv)hS~9kA+YW47hKc3?`9Fm~C*u9nN31oW_xcyAlx3Sgw;WV;CqLJ0
zd|`jijnA>&&#*-LhO0}<eSvq#%sB?TO+)EjcdAfp-1h7&CKd3r<xUBlMj7!Nb|Tk9
z)>>J|6jp9Es5?47)<iB#mrBwHSge_i*&ZDwXxL=;W>_csc+$K6H#;0?Fc2l0GOx6_
zAcI|mGn}mBs<)aNmJ7yQwtIWHMvq<U>U%yVsErP&H3z}i)k};@6-<=Xd{|zhr}i^{
z-TzPU!eBT~jJAB3C!MAc@{7cgd~X)tX6^J_R=TxBg?XbxU_<zISsse@;fS7*v5u*;
zx72>RZ~<0lrliTVVbMp=)Nb*|NvT>>-uv0ep}0BUX?!5=WTHU)Sg@M7OY4W7{Wzz^
zqvKU(P!Hr?to;DY0ms%>z%(@>-M;1!)@&=A&}vM|$LVkXahQt(F2?Tje5I0Tced1a
z&va_@3-QLpT@VAOgxdnuxAk!3G;nas2h2pCelS6w?_y0_wz||+PLttC1+GouT1GL8
zp^y74s5dhV5~?YP(aZjwObxDMo9s|67F3Zf#G|;BAF^lQZN@YsQ)M`D8&ZpxHMDy6
zEEMOQMen4fggB2o@XVT#4b&Hy;Wq+dd0I$vmi@k5w=Y#RO%71EwxJG&k?=38k0uyp
zRxatH72XWa*E~wg_nLcFU7An18^y6wreAIR#WUO2C<h^*vc|*0jp3=H?B`{vY}c*m
z>K)W8n#_~N?P<_&syv++n0+&%nem4F?f^H^Jf60}EE}!g_6Q*#UJSrTu{hOn3+-5L
zMK2&^jvFjKc}3^e-bhyKPr*qhWQ4IX!5n;|{4n1qH>VG(O4FnFeWdkU@gR>9tQ!Do
zDy5X3zgA9foBj2$z`fs~Yg50ez$Pimy(P}k%?W9z<|V4(m5>d_bt7B!5_W#EZmplJ
zs<$MyUaRarrM=eLW28!~EwqjS;s(p+gba)N#7H)3<sE{xP!76c!_>;La#tiXhK$zA
z5*uHAzV)VBvcFy)5VyarW#@^oN|*c(N3Q#<O|jG7HqnPCsMdbJ8SLK$J-4FYA25_X
zd^I0<iU^>x7qzR~&M&=}YuU*G1<NX(b_|t*D*amFJZ7SLk<}Sn7A40R13I2GE0Mm+
zsmzdtWu-+S-Oh2Q+cr8uU5sC?j{t78=Cz|dwX4Cuts`6jW?+RyiPdEkdA@MkSd3o6
zVqE&8Cw4qM?s8nMHXA|w`UGKLOQaT>h@DA)yj5l_Y_+4?MBej`@h-mv<K5N3^SHfe
zAydo5dc|X4y_kJNRU_m45Sn?V1rxs%HZXX!l6U<sW{OS^HlOqhf|dO<>|KkhuCb~Z
zkPX#ZtfWCTXZ*siNE1BL<FqTo@;dVCUK4I)@UPt2ZF&hTnNJ=W!iJG}L)r>*E#U-|
zb)q&>?A|Qa-$-%1O~uoWo92Z1L0uFx1farMccYsm(V_Iu%r<sT!D^|64*bw>M<dn!
z0nptt?T3`W1y7$V#o=^X4FanI#qTU(vq6_ly64O5o*^|Moju*FA2-YV^F`-8zheu*
zj5bH?5+ff05_Ir#TbT!G-x7^o;;Txk*7V<oXcpJ~L94``#UHjui=PV)1XbKyD6eCG
zb4I2cD#6jph`37>{2{}TD4uC8B1!w)(yRt1EKtW4kc0>bZuSZ1yEjw#gky5Td@tjw
zyxdAqvmo}@`G_oQ3EtJq-|F~>=~_({&6A%p7)RUhsb_4*NA3;vPhMLF@=0uOva-L6
z1?TlsA&;#^1P%oKEG1k<_L6F>|Bd1=$(rAjtf1Gf{!fP>v3@faKGL$VoL3W9xhVD6
zKKSkcE$IvTPJ1g`69l9lBg{>=?KLv2!J_C3VDu%GpJu|4@GNA%b=1#zW-$0U>MI!{
zbll7++&h!$l*F}<a9Z6_yeq5e)2ADILc>yd_CE;eqK(VBvmuSLK7cW72c{{^G+HEw
zS`|w~jI@N!feTLk0z;Y~fjT2r^R|BIez)<xGAhX9zGUaPiC_JbOtRE+7z;VXiAC^v
zSM&BTz03Km5`0gg#ZJaq2nCH<(g~fJxbe_ImDw;0N$!a)3j2d&SP43B%U=Gf<w(P@
zJP$L>Xl2RBo5>fr993O{lxr4PWyCCnd>rea>pc7u7WQ((v+75KGOOdEL8M-<YtDGK
zvZr}V8OLF|=XF-Q!h62M3z2zjp<G*I3tyRx?d(nR2Y#K-r#h}uT4(34&Q%7B=1e5Y
zLm7}>iT|{4$lRvXZ7G#;rb8FTT3Ll+hr`Wvuw3QzrDHsUp0!{`VUBSwD9PWx9IKW5
zMc?Uv*ij>7Lh<&Htkp0jPnH!|Kqsl&bqef|GB>xmgjV4)f1m5hwk*H-@57gDMoOnq
z7UBbHA$o*}?;3DfOlaUB1#N_!W*2QxAT!n=TibYJSUG7pBRM<zUs#1orbJEsxDC1K
zDf(H@+^+<v5e8Havkz)?t#$f9<a0tfNO~X$Z8kGJ+GCAVX^*%XNj<Fs&hh2$r@R-a
z&0oQh<Dy34h8epgs;K+5)AF}ArKdDF>a`v2(ngxNq<KiyoN!$5egE^7^(M_48`b^+
z<yMFHxF=QBKUZ`WZJKNhLvqHgq3n+9V4>PVIb{UvM;rB(H6r*s0=WHo?4~bI^SrEA
zKA!&7nECr%$so^ms&{f?$jt>#gTY*+adu@n+!1Ap@XETap5wp(>yM!$xK&ZwxQKmo
zs1ls^qBzTlGAafYAXNk!QocM}(w4xxyL3akKJ$H0c#7`_&AFwG<bL5kR1jItR-e>j
zSq(x(Ecp#Mt~B-kko{E2U>HzVbKQ9R-D_K876t>H;j7$J)~tVML}uIczeu^KxSDnN
z9|;|<#CPOKRI%RK1bk+%0l5X)*)WWbujgdS*Nc~<v2(#1yqk(RBeb?|(_ona(AVmp
zlXj+BFc)s_(B<fr`)i}YjMls=Nl1X%UHosoko8H~JFqrkg*Wib5HmaJfB?iWT86<c
zY2K}R#mb_lb!deoTj1&sVp5CG^%<0bGO$XfhSvHm-~-C8ldZ89a0$6|vzp%2nalAm
zxuD@PcVcT>gQ;-((WDFYs}8;pgICNQoc0tp?_5Ex%KE0gG(2^}!M_0!6@vBBt(Vh_
z%TRKZaQ$Qy<liAg0~rHMgG83A(o(olNF=nXf0_&OLu4ve$w@G3QC!1@>J6vN=z|&_
zTlZW3@Orw;jrv@z32PczQGo0gei#v*$+4Q~K9|6eOKvB0o^9FskJ{fe0pkyg{RGGp
za@(+lpj=j8D-X|9m6<2(XBE+%3FlE$vV7;>GdxgM*hmL?vqh8^YJN>61_t6X&$7d&
z&fn$n*V8JLTG2RUUoIWxrahVsX=u)RwF>1<tgNacqR`t;Ycd=b^@Tbj$u(ja{K4}?
z77pwIy5^vBrIcNH{B`7=2nDr)xbyq4B%9NfFHs$a&HL#e)q!!VuNhnRXxvQXXjHox
zb%pwQ8@oD3%n%D+vxOAb3?I~bNA-%P@?=nOwMPyjPp2#A;QJpF$S7V&#IDb-_m|Cb
z6VXo@2#0DWnOea3e<t7Kb3?LNnxT0A<sydov+<Y{4<A^B2^zxZ>{Ox;428?;D&EPL
z`X_ZU(UO<~gH{D|*HF?~F^*tzKeN?%Jg{@GRUI1~lNpVc$jl6_-U2=yOj-&sYU)_9
zZP-@<d<CI^YNwEQa=xW+g5_HD*SttbQHW9S(;8PgJ^5^)d>%>%H-FOZT3t|SQmjCm
z$u5*ZZpWE*;WEq1!#8KIH}~O1^72Rzp-&z|ZAU_ieqWQrtPG&fQ~M;vNgLl~9?tn9
zUq@Gb(!U^?=Vtm@Up%EuseZ@!T<rI$s*b{dis!!U>oi|gExoI+KP704W~p5N*Sj>k
z$k6-#=lj?W8M7$87KbyPLLXoTa6=}j?QBfexnon8g%V)l=_hsRzW3TNPpT)+w72Dp
zwsf1UNM>PvYD$&ChBIh;i5py8-set&?6(d}RZjO#VN!K@afRvYxXc<0iO$x(<=4GA
z*6k)AjAn4YbfDdolKw=%FvGt;<HgABag?x$p4iHSg!yWMY<0azbryKo`Id{P7VP9T
z^=aGC=&_g_9Hw#)7Z;hEuyQ;iAUd2%kKFFT=)|GYQwwI7!usIoxCJwI(V2lHz%u8I
ztjIJfAfV6;_w}k_kM758j6r9<6nl!jwq=@UL3T!-AFsP}X~4|XOMFs0jg(4pjGND8
z8q%m*Pd)jzHA*yH+}@yEI%z<tVB8W(cFI{+J$Lh@Rm6SF*1;k&K3kZq|D#qnu2*#F
z9h4Yt5#=-be^kA7SX)aUHCm;G0xiM4cnL1SrKPw_g1fszDSBFp1b3$tf`*_$gS%UB
zFIwEK*tt3H_uTt@H-F~I?3umy%-Y#&)^Du}IQ_|G!ytT|Mf90nCG?9HAYq~zU5|Z5
z;Td^2p7>jI0%p?nO4(O!;t?OsL`h>iDwQ0~eaS~Kxe52sEvUuTQ$!${f3;^hmXoKa
zICcw2%l60<667+(d(jK?iK`h-_EP+s0eb4eWxUyYIK9P^Y%A#|XK>{cPDv<pu+7zY
z_WFQE7^mQ}+TuoXr%wL`2v6`E*IAOotfxKs20jI*dUKI3^h!h^c@_NYdb?Q=6E87^
z-$c@_DR?6h9&?Uh?bN7QU*s>oeH!cf<1w~sl)oLKkghQJ^L0;J<ZH0?CJA3YbJ5AP
zxqGxAALDK~?>T-_EDsZay}Iuk7H@|K2|Jm~j730)e90#S+Rp2%w{*y}Qq_c#k748#
z@wpv8Q$j#^@ftCH00wg6GGCq}=W5Q<tQ1IB%zGadJ@c;2v#YL!BO|`o@MY*1!xSqs
zW#KSZD~N$HQUIzhp!qN}**!7z{j^Sve=$EAyPM88zcpfws}rW;(}lB^Wk8pYM=~UD
zo+T;Ut*JTbp<9J?KTAe=)~XugqCa(>{>tzEz{hTB8A8I&A3n=KFJ9+<s}DPT8M0f8
zed}p29&l=Ro~Wr|f=?}M*2;+Z((aeLuU5Md;|bE<2bE8Qn>5_{%p9yG)|;F%-S-NA
zldvU=Z0BlnzJ9Uk%hRA<%71hDA%Cr&$0=g~%5hBTppU12RH86Gww>&qO!LhUp<b9L
z70o9LpJL@BD7c(?A%8L7fOa$gq#G4RSPccd$;LCmi=6_`IB4(uJWEQq+(gtY2{a-y
zQiFMWTNe&3eVig@njjERJXQEUu;ADVJlqkrio*pQx34b2-<srp1mXXB08O&EE`BMB
zWpdNWrCv(5L%+Y6g6ezFOFZABs$XL&VlBb;4d#dDg?6STg$T`vxR72&hf%z-s4bW7
zj9B(`h$ki(Ty}~94_~l7<W=SiReY~8{EX<235Cb)*3CF1!%KWa4n5cWocXE@hN;Is
z<!Q4aT}g?>R3nv&Aq5ExXzV}^mGFzc>o40(l)aO{<rRIcip39+T3jhpjjwC<vqe&0
zo~-qjs42=zL5U490Faf2N%uJ=j`m2PG$Nq7bTB*w8-TR?{jS8UhuWj5`{8r%@?_lw
zX0^M_TQoR}Q^8INd(Z&5>DACl)>)JPia>Uv!ce(cqINT<J~5^qqO>d63iRfC2}$DX
zWhFt63Q8+x1KDc3dMf_?h7dLkWrR?;QQxh{U0Zl$%tDhyG7OsCa%eV!p6NS~AcpPi
zm7~7ioD`rrNy4x#y2CiHq~zo2x^x<gudC|aed*;E2;Zow4$61I%(0}iGE$R6ue*vE
zDE@{czR!vqs6(b5en#xmm&H3qG?Eu}H&0BvBJob26CP2$>hT)q-tI>o53V!x<rPX5
zUxi==bZIH!#0#fwp5<!gFh^f!Sf<zsQb~a{{Fgg?j6Enfs93)5PHYv}leH15S2q8c
z+Us%5s@V!GVN*zs71((P>r=4h*`5c8bOciXr5t9cKExtPm!}nOz;G+s66$#RPlf#>
zW4&Q-Am*(^S<MRJQK7MzWkJ<%KjR==HUpk<XPW9-dF!JvZ{ORj+hs_DNuNm*IC9iz
zEpwEtUrIrUg2CF(ds9GSkh>P(Af8p3*>F1=9n&>N@p_<crUeirbrE)BWS1&O_LhA2
zSy`;7m8epOka=SSoY!*~0p9&v0>l$AFXH23oX_In;M^@9x5~kPh9Qfv&4o(<Y?NPE
z3X2!~x<IHF-dGO#^3F84+Mw3&?r8j<ab`W4G5xsy5X-xjxP?EkoZ>!F2q~o8fPO2L
z8X0wEQt3zFR({?drRP1}C29=w*2(<dgAJHfDHKqZNZZR)k0#(uCwI<*;t5_W(T%r|
z+TNBFIMMxnpweJrWW98>F--h5`Nuvc%}PYy+H0@ht%#bH3&cBz2YzsPX3Vb?geyq1
zR54?PG5;?x$Oxg}Z{gKuT(Z^N>4_$YOW2!JAUK(;Fj^S1cLVeIrJB5kaS5mPA|%^O
z3AuD1N0!4c!-{JJ?P<0C(B)|n4P44({rHF~Q0Dh(5nrb_7cE`k-%C(Ei)=y}-}7HI
z<(5&JFSD5AVm4l~X<uutP1)psI2TI4WKGksv9i=ih|caxa4#jGSE<^IFEkU>s6Zw!
zl$k)*T7=DpRSb*N^LHT)grE<j->~?H7y-|z<}~P6=sU9`0fTYVgpe2mxRv8@oVA~L
zygTgMmu<;)&gbjrSI#uZPk}YlpvbR{r7~Tf<A)!Np6!9;UBd?pOhOAT&h!~<Q{+KH
zwo)jMTn@l2|84;;M;<~@qd?O^x0*?tC|Rm*w@!0wCh>F0EDSN3&tzjgL9e8_@hN0(
zum=(g0EkoK8DHIT5v{WtvV1Bw$KD|%1k`K~q1ef#LwCdOx|gtJz3nI%KLiz=<9gKQ
z30^Ric}dIc;Dejri@h4T64v*58T!?)um3uFjbt-b*}GVqFfL)b%4ym;B1|#DYC^;;
z<Ci95L2CqWe-?jcXy||v=j`t4$v(RLAASpC^iP)F3<-pW>%A|3^Rz@TPxvZ0YH!+H
zHj=aYu-NP`wu~by^hU$NJ5e2Du7mN4e$W=3ZBSAn3nJ68`T!t84cAO<Fq`1>FES=P
z*VNW$pPbnLA5#=(IJ4JfGF9ToG%ebc*^<Zam~PslMRh9o)Gb1$K%s2>1FwEQ(n*ke
z+fXaw1)(B5*@r+7dt<s2`~xw01=TpAw!I1(cV@VPqNQV^PRjJ5G6=Ov;IzEQ+_&?S
zT|}^aieB-_%ARMkltXQJ_cVsa^&npD-DbAQ+Q8Le`o0tIP>cPwq{LQZYis#`WD&}m
z#kxm2mIPNEY`FYoWsyKN&vMoI=#7mB=~9P3%g~W%FUp{%^h-n*rhH&i&EtcnUA$F(
zqC%rR0Hx(#O6z)yA0D3ETJv+4)>J&N<r{zI!6IyXmy%@t9@{V<)uPh1%M@m@{yLc`
zqogA{8cU!|HAEG&Iqei55wEi%D(#V8g4bWJeG~Ww{qTlzMmtiqaC*xcpegppL1aLn
z<KVXlq2XeCX3T?^ZrSt$MV(AjBHNXP$f8PFR)m=fA!*f@C#NYYwZ?5FLSOJueBMo(
z5%s#i46li-${@ZyB7z^)ofM($akn{wL=y!C@3Zv9ISv*Nb|W5&c$Sr(nBLCN<U&}}
zw1Y@`v%0H*>iP8YlY2Fa%6dM|m@Q8&nyZ`30o%9<^x^l)T)CNu*?fVvL7rYv(Qid%
zoE}AOZ|4_1Qh36Zn;4-{0=wPUl9ZMK9*XCp{FbnVI^QO}VI+jJ<RmJ=sj<`WW}SFC
zbWSpL$s^jR#<9;OBB7MEI@=7KQkvw8lYq9p#*;bo>tl0!tE2x}^hdRz;0U%(O+djz
zm}m4s%MV-KK5}yep>bTrx{!~(wgGNmc~4fbFd$t6&gioX$<unDU;U}DC$towU%sVu
zt;nVXvHggJg)SqdnX=xk84S?W<j+bNlZHyn%XWG4$6}di{5%FxFWgQnw3QSugH5LW
z=pVGiD0SEoKNh@Y)}nNuM{C-p9%3xoV1-f3g+nDdYK3>jf1ABBmO%vqZbMSc+8aMA
z>OQfApO{#iST)-mDwuT7`C;jt^gMHTgJyEzi?0&XN3ec{er=qJw-kmHXzs<6@R@TW
z(d<IXQ;(#jZ3?$@`?OlzxcO0As;YtRRQJOQwkmOsO+Wj(LKaH%{JMScrnNUS;9IRz
zbv6Z?9**UdHhX2^HU5-CM?!Lmp-KDh1;!$1$63ufPCJlKC62I0=`<6Q^cIDVyi1}2
zFTbZXa)t3PZOw62o#d5HTqiNE9kenakc2l2p)ffgvZ0;VodT5!R*U7{9UCdWsH=@P
z(D)JO_e$0sqnVfpkJbqyd5*SXkSDNLw{r)v3Tn{}Nk<_MlkKG)Ct)Dyo8@f_c&YaD
z<lFu!rMkh!pK{`qrj^(K)mqf=jieqpCtm7bA0N9cf#(m7{JO4TM}1LuNTb4I7r*m@
z3ki-%JJhkt#NMpJec;ezzwYru#jV=|PPdQJTTE6t#*axd9peUXp-tzvdI7urbQHMp
zSB1Hx?C9So6Z37p*D;;THEmW4D=anxv7WIC6W3cN846$<Nx|!*t|dzLqp>CNvD(9c
zb#vo|$Gf-4so6);ef!63=Ri5{hPm{=`Rg1OSZ5E>@f&6ZQcH!iiq(-aTf+3)(K^Y`
zdq!goiwjtlJnf8FR11@=ZFdHw75V>tg1M5_fr3{$Ih{oJcHypNft-9v2vx1skoD>H
zuH-P8Xti(L^F5`SCC+j^Ub29W14?pYazZ0rRPmAWt9%h{$zaWBrG#RIx^#d(k!6k}
zQk#B6fN!p{WyM?s2zuXSpr>E2Jx?m?>;0u}axGA*yoqM4SFF_{Dx%=dd92ZP55_!X
zm)~9lZW;3V3GC@m^-`sZVHHr{@N09-)9Ij*I~Y2E<a608;xl)X@e!F0z9=Ut9h@jK
zUH6&p*CHe&bT_{3uP#x9f!7cuT@PHknDSN=3*xroM|LlaGVf_-d}2vI_z%6>kaYTw
zU97>zC`+N1ggTaHm5<)lP5jexOm$^v5)}#JB?2k_u2*~AS+wo!Hjz>_E`pA}32Z70
z<^2#nI1^2C0l%m3h|G7J?tP8d&{_4Xzw9H`vg^seThWgrWNq_)r0SL4yya`Ci-#b5
zKERMNhBuP0cC}mw?-A8}qn_4x;^kGonwpy+rt5&dsrCmQf~H4Y_Dvl<RQhUmQOUfT
zzM8mMd;Q&(=cWP-d5o|5C=L{SP8Ul`sY^2B7y6GU9f0=?B2Dfme)J7x(w$UBLq{S5
zjSe%!EbrDDW!dP83#1V`M}FQwl8lxf)dZKMMa?`1e(plNT7CsDRl;z&nwCtAA#?Jf
z@He3DzfWG7593J{pijY%fM^z|#q63|_s=49ak4FvJ5wolYHLR0gspmft#qS~w<nmI
z8XF<+0=*$yRRW7)Yk!j~(Sjr0HYR^h831GwoWIAxlvR{KF?!I3w}%90@l?f87}_|+
zy=+1hl<1W2{rH9Z+pt|{*2`X@I^-88=TQOp!6N6-wQ2}OI<t9w3CQ<{arguzMxG(4
zfPu*+m(dR7Yp}rbkt}DvUY=uOloO%u;ANw3*YJ(SbyM+&qPybMhH6LfEBzum)%5Fd
z=lJj{t>rs@Iiw~-svhMQ85fzng)i?+33}4KdRLf6?LQ7Fdf}rZi}1`h3O*1qeJ$nf
zWqG8ha>bP4^5IGQLAr&C)^i>?u7*qI!~BU4#JikmEptKKwIy{$e}I&IJwJBl=D^Mx
zTtwaePKHBF&NKiM^zb(N!=3cbikw244!J1>2Yq0Xc0pHsH^C{9NFswKCa6tJt}gpi
zvEr2mO~!aKM($95v(VFja&K0si1ntdCl}0`3@u6HX4<BvP)Shz2az7Oz(Zg$`U#4d
z8kjX_x&jv3eBp)V?0LJK2u$@Y@wfY%#@ZMIj^vYi;ZX)Ls!DY&1K>I!HE=gsZ#*hl
zJzqiKmX(dyaa<KZwIby^Ruqjpd+{`X?<$f<Ts2m=%<P894hj!>d0tV_>tr9xpr5Mn
z1<bg~JUw3UXH6=4L9q)%_(SoGM|7G-T~k;Rdx=;T?|)#Y+v)WUS5uMK{D@zuM)Fu0
zSIF~Nd`9NH9Cb#>JOERxmkEE7|2zqC|My8UyJs8sa{~A4*s&E|--nkGlQrl{%I~#H
z0=4*uA>;Hg&&~{K@WKUu<D{(Z2wD#o0F#<tc)$*%yw#g5j6A7Jc3lyeSZ0olq?<M$
z(AxJPOHNks%%25%lI`-(<BE%Zhy+-tQt)&3+5_2N$0hI0MO+EaHWaBee8HZDlLh*w
z>ZfHCi5OK`C|1i@tsVp7nCIsac5@n%%EF<w3*(Eo6Ei419Y%7m(LoQc?`c6Xbu`b{
zHw0oB(6nXh3&nE3gn0;bOU4S=b1;SX4G|_$S}VLej3UvK3N9}8GLzX{JA+lZWtiQ5
zve`eowIaRmilvIw#wM2eZykqqt2L1${;(pe{OuHq&zTm)7D`gyO(wTqj$2p8l;uO_
z_S7OREI4A=_DQ6&!B|#w`QD%^b186mTn-6^s6Iu-Zun+(;TIG$m)l~e=unJFRTzP*
z=K0X5sS&Djbf<X1>knOknn5@!n#Xc+F9up~XlG|T{D>_dGgNTL`0i*svJKaXd)GiN
zO;Y|s`3p=xfQ5W~bH5q@(aTaFj6&3GK*oQM3TWY}3^97=1=a-fP+%MYL-u^j+lm}P
zVWl7-<1}+4g?Ty%pvkZ2<NgMVVT~hsJcaQ>u$0srNztr<TWR{}^v6P*aIKD8zX2~&
zzRpdyr?%oQ002{;WhHW`i^OkKkq+tkc@L3DOVnpk6+Py5m|FC*osb)Eq5F*rjXvGx
zn|RQ-SM@FV?k3}bGiAd%8ncUCvVf<E>6V9aH;s1^KRJ>@@AN&p&pQitCj0IXylZnm
z_h@jiLB19n@5y>6#?n#|{Jt}`I6k*C^LzivJn{duMitWT82YI8(y$~wtXb(K&3z!G
zAv$O+x+~SOv%D=#uy)#9hnQN|J#7!6PdtD1prXtyvoQ7Da(i-B0D7hG6OU4@3pr+N
z4K;1+tXC0QePt$Qr`mVRm0Yxb50dHu0FX%CGGsLxa%)(oul-`t)Xe9E5`^_)k#3lg
zY$BIYyM?YNd`LV1q;>UA=LI}lHkBsU7)p^TE=Whh_`&ls48We{lW<D=);J($%3}?V
z)GKqZ)zT|7$W345<AW@&kJWPN%RX73*K~PfAL2QQJ(px90>1m6N>cVZOG_)ucIg9p
zfhkkTz`9K3ji#r@{Ze<5a1>v7>5a2HJK4F9Z!I`?L@q@)l8eGTQJh9fOwTK#-%T4=
zzWT_Y35ri86IzCmjCmI|<y+57GO>tbS6B+q>wRr-45(cg2v?5kZG#vtd)My7OuH9M
zIKtD@B+Gr&=XHK*?-Q}6dkzZ^<xK;L(yhZ>=MvoQ(95BJivo`nq2g?8%S=qbNtJbm
zsjx}7$r@V`MHz24_*=dTQCF~Ld{Ry@KKlxnJR(Gx3*VfQo`D-S7f;b2J_|G6-lyMd
z9I!Hxmb-#>R-bNs3?O~S_FD$}a&XRb+8#x@)T0fbB5;H+7EF-zdTejg>;)^N>VDyx
zt8p`%%qc<1;KwQMF!UN0#FLLKt?Z2Zk{2gd@X@bJ7z_<`r{{cAvq+mm62|2(TswF>
zK}q;b8M2C}z4L_oORY<5y+Jw*t5x**LR?YBC`9NE99lPmsj(8C$E7r&Aw1GK8cJEN
zwfpKmOpDS1dB?j<-s)8Tft@alNQsMZ{uT6USc4TMten!}=%wT*rSHgH;xyf;?4rJ5
zp1{_gVR;ScDSe1k84u2j8IHsC_98{utY~~3>CWNMDii!SnQv>td)cNzKWCN)cJ$P2
z9od9I9IW`!;gkS-bKEAX8_!J(6B-6HkYNdiKavwBo{=#y?k|ncF#I^ev=Qb_eHaeO
zYFLjb_NLBtH%`u;*FJyim*_B2I16MPQ~Qv)+pPL#nO44rA?V9jX^cwCc8y(BFp5W;
zyi=24YpbYA^*zjNUKm;%o{(_vg-WXBH5Nm6gi-J+bw=~|$EOf|r~*(LB0{hAN{@k2
z2O>@qfiUt4f1rQX)PeTrY2DZ`wx_2nbH1YvUQ2WK(K8+fh>|b#)@DQ4Vod%|dX*tv
z?C)#c=VM+sp6a|q3=7mP<>*#Y{)o6dch_=?5ZM+Qv6|H&2I1s;#MNcEb=N5I)7jH2
zJdpIYg%qY=1blNV6o*?Jl8tI{%Mx|#w!&EGKoQyEAiB`v^W=1lq#>;faQ2@eB@=)M
zY0($_(n1_c!Y^QScu~us)?kj*E<iKw67Oj2)sh1p$>MV-?>@NM6q+e^`5P%*w;?Uo
zYsaQ6mf?}`S&n%3<5<1&Hg3WfRR)xt%B_Y+>Xp@9jV>99ubVhvUORP{oHCz!p^SEt
zVdDvqJg7lo*=k5G?WY=@Z_F34{HV9E64rjA`DO>`&W$_nJzCncIdczLSpYBxehMRD
z$i$|pK+?59GypS(EiERcUB5JdwZ)Y=6x5tj)?O!Od;8rAKB~~-WgAevdqH`*>bbV^
zs|B{$JDv0FkJWZJl3^)EzQ)Ah^?H(Y&lyElsj3{r(wmf$+8-2vi^Ouc$h^SCQp|i^
z*%l&f;@BamYZA*(XvZFwzrY!HD#4hOSUzsYUI<}LDp8Q@Djg9qX4jETa_^PvaPUf-
zQqciEBI<m#<n*mdHVjp;LO?R{@gXFd&Sut3+syRRGO3}lvU}aqPT79ILzQ0hw0=%I
zJVsTiDl+jm-;Y}3`Z1;Il#Fsk6~+GfqAxt&*lgR7jvPoh7=dP5m;wydz_puCbgsmZ
zz`PADO!O*nGI{4gxuxaPtZ{)1vYuyKABYTFvCle`JomqAC%6*vAhYVNH+|&I2%-#n
zPl3{(84H*W5IJW!@UUhnZ#4T0=aUpXd-}71^o@;xneTob%LX|L<EnY_+;@2|dp}4S
zbqc(rnh$<xsR|++d3=Ir`Md28*K9eH3^J3O2~#)k7}N`R1e1{(wXK}D^~lX_55GD*
zjJ@eKIeEgC(WJ2w#1>mMlJ<6Sk@hrd?Dvx<LUUoUz|Dsk$M(mSxz#A9M@KBA4_5^J
zEj<2882R+h&*G6g#RTxgf$#f`)rrRc`yEgE{S%6j$Lov-1<Y{1VTRL2rgT4tXMF-b
z8u|^xoXxANtKNGvuJfRGghr$e0WNnwZ!sNH2kjn4-*g)PjEUWRwlDbC<6^%(Ajbdy
z>l_=AGx#&-)0Ob~b4=HP+^K&nDF*IKb}wixa=0by>jXEjgEF5YrWaN@<Nj-Rsl)El
z3mE$MiKi*?$kI>fg=S!%tm&J_W!>I@)}0})hdE-y)vX(r5?YHQOnVZ+4dI~7XGWO|
ztzT~6`H6VEoYtv|eLyCe+*L9y-nYW#=JtkZ4{M`u&}!T5z@j%EEe}r}4~NNv7TKtc
z9`__n?x+G9FL5HK7E({t$}JAQSQB1#_TD>UZc%&2>at7xD`XdU&!AUK&|z24zfYR}
zenjmz5)K_{U5WAGJ&;>G3Vk^@B^X(mI+3Qe=;B_TNRiwK3lexS^SJL1y`v!yd5o%-
zq-y&63#Wc5JbU1@>Ls#|d)Xb4)2A9}`%mVTRX`4XTh*U5lbuYq)5V<3XUlDvlV_Ei
zyBKD5Pqh5I?;PmZjv7%rxLsZUK{IC9g?Sa9XL($%IR6Mr{Hry&qv<DpPB^ud5OlcL
z#{XQe4Rf`n66ep;(3?#k?pjuPF?nlGlgF;Ai5_AReGcshKmUDl+G%ylOHKFsj<zv@
z{!WRuLdBWk@-goT$=Llse9oRW&)L6EP<&ime^YIry?sxfvH{Z;)Xv$6`|I`ZleDm~
z`H#2lNOt|lT#FpQq*Ahx|1c)D<4U6PB}<&xup62q!cw0gD8_?fvf{q`>8rS`k@%FP
zJF**$Mg}6QPpcg^?tZkl{-XLeZ)G5d^Fqm5^Xr@8Q-=b9UX2|?`~&*S3lX>V!18gD
zwN$TG+x!5tuQ_~cN>!d)6|Lw$i?_Wk?dp2C!^h7c0zDj<RJfQCpMRg&-ZYQ_mi=!`
z(mnREhi9K(OZ0GjO>KpV`++2a7l)r}UJa18xxI|@AI=dH5NP(f+qt>4*jj(dvj1{;
z?1DJTu<a;U#8(6_!1o)6->Qw_U~8q=5c$t4NB4t&pHTh#1Y_yG{VefH@|@AvekA7S
z>ghlQ?csZzN#r?C-_=01(tsQ7_Q&FqHc`j<J;;>=lF0vI>#O+52aK)QM}zvm0pNuB
z)3g|5wbDbI&UWiZQkSzoD=ogn0kQtwq#!++$5pm->6dbXe*=R65`Dd9k4FRd{uLun
zX4JOS)N)@U7kQ|i=b0W?{&zfS!}Iy29}lsP|4juw$(U)3VcE#Nse@k5&`$7ypEd0R
zW`TKEjQlfWb9Vdwj;1l}-zRd_K{Q<FA62Z6Tsp)(ULq$k6O}~at3ynnZF{QA*=^#L
zINzID=D~ZPU5q_;KTgEz7WfEqoBIkm#Em(;)jZSk(PR;3`g4QIQ|4MFQSq4Rj~jI5
z6l1L4)pyE|1(|&u^k}twh&kxY0ft1G9F}65H9aYJ-Y$xG&yoH0hlH0BRokK-ErTiN
z3QlRDq;g{8yPT@&G)~liYg1?Yf5TWRJ@DLq1xH#0@+SYS1R=(@F4Um^0fkAuH&0L<
zZ>ww+R_mc*(_{C;e^4>|A5@HCpyDGav!yk>?f-`|%+*0a-+)(he}>;4{)bQ|+wBf4
zXGKNh0sj{?zNr|R{VmbQX)5W0aOvgyZ{4cmlssxCUI}|VF}b?M03P?Fz(1IVnB^CY
z0m0{2w>=O4KH>X6tFPjjM4y0Zkc%T`9REHMEGDMO<gnaA|A&hh50v^k%$Nx!WZNAC
z-MPNcixUOJ7`6wl>aIpvKRV*>(;vFr7tmUya#Q^KWO5(FB1Wx_>^(btG=Z?*QT_eC
zb+jB^<bvXDku2Lk`$vx6Sw>}%_X3stf4~)U5T-|$?dSR<kB0Xx-v2!UFe9k1lia|z
zIsJtJ-J$LJn_OCpOm5o$z9bJK`TfG(6?0E-EHS6C{pRYZ=iUM{BZ&KpOG=%O=ZfpR
zt??wUR{C_lpxh%jT&A!fS|P2?fjwhYhiUu{`HQSJ%er2Ylz|nxHXE}h9NnzOW6eE`
zt<MlWXTKp2Zy(O}Q%l=E2^PgX8kYYIBz+YfA4fGnL6r2VCTnsnXqrud=_|-Vlke;`
zytt&xW^})a^HJ<+5+ZLcdhv%4lT@{fZDsBfH++_AUU_mmFBe15Yg>&()hIZzDo)i=
zOQj^xFKJ}Fh2mtM)$83pXndN_{rG=*KfC~=fX^lsc)i9hTEv>aLOsDU>t$+2IW&kj
zu>XGd{P%p!5Dk{~BJQ4$Od7j*-a1pfx@RFDPGRZ>*w)K?V<3{^cb!8eb);@R+sRxi
zYudE0*jyr@JDrEht=~Pr&u=b+3q2a=$rZ6xRUYle{z%>p7(T#QmFc=!!2zm~IPCgP
z+Bv-aXpo|#Pn>Bhf6~G0ac#yA8W}Tv!8mM^^o^hL%@zIijqGJ{?~@#7h`?Y^!8|qt
ziWk~!=0knv!d?uY{OEc2ZS)|d#!dSjRzo0Jq(E+{a7yp~v*$-54K?rwnz7cHpHON#
zQ2<Dm%8%Y+96cFEosj0vL`aKeI8ds&WP)B2ziHkz)|NFkAh9idvaXQkXknqMRpj8~
zw_@}LqDRAy@TtAQjj6Dg2{MiQeU;u3%sC}nkLEKTZLpeHrN}WXpPU%nvIKeNcS_0S
z!P5G(e~`gPAll4LE0-9d=`pHgPPyDiD4nLA4V4~fF0Oeyob4OQnqC7ZVYq`Tb=jUX
z+zT*`RDa-BKWyJ{{`t4P(HW=b>H258L}?#%0oM%5@Lv7l7k{)&W><r;)?3@^Y)LFe
z?7>jd`%V|4?qJf3?qDvH;$h*MnZ}#~W>)Y#MI`a4uGRqx)Yce!K^Xjj&-G7Xigy#w
z%R_$>@an6jT=9D3!>BPXbFbITEmDJ_Q^o;OIWd4}dM>c&ShvVt7n)&*?NpelsJ`fw
z%a|mv0FX|i!=_i%OkgZvl$MtMygixqr*X_L>$+v|Ps?DrN7hzT)*p`PQFr+NKFc2B
zXsBZ%qjEf<+c&CJAF+9vkx)Tn-HVaCNTE+(*2SHM67FI1?i`BAxq1zMgo^fOsYHwd
zS5md|XO4s&++lid@TCZ?X%>^L9_q0Cfn7|tquMF=q-tp8%eQhG+}pUlQ_37_PHgz2
zvu4^%4(|K|-uWV4oMOuYMOd2NQmHS?QsON(=~Q_kaa-$Z(v}~Laths4gHOsoAXKzh
z1tEnhW8(CqpW~sx>IoPoDQMtF7!Moy$wu#hc`5yss;_r;g7_)b<A{xk<>)(`B0S#3
z(7A#LpR&)<DR6d_Oq)r~QjI=hIY#xD-W0hPThVf+eHrdn>vAT*J|dao=WMiUW}XP=
zx4IUMUnY(Sr-SQ=nt}+|(03ivuTkKfU{TE?Z9RSeX!7byiY-~(GKMAlY7I25ppL~y
znXE|O$6jMB-v?OE=3G6-BhrdNI=*BK{HreX(1*Z^7C47|dtln3qwI~In5&onuqE#l
zcAIr@XsL{V9Y&6wlI`vuJTE109?&*I4um?Jd7B|@uf=(R1w*yv^7zTX>&jBz`Olm>
z?4`+tS;_Jl)z-km&tDsBfM|M|NLNkZiT(%Jk_ho+Z(=}Bm4Tag)G4mM@)9^0t3oNw
zd{mu>D2gn`&+c+hYh{kFdUruao?vc=82goMBt@nu6Zd75&kSoPKf8sRzL&QOGhMCW
zu3adXl`s==e|S{^GXe4V&pPKO2rU*6R2GZFMhNbRip6s7is;Uiog34(vi12SnSZT>
zw<!%%G5^Y;D9+fX23Ic&tNk2PF#5C4FJCLd7|NhYd_h6GUK<XR4$&|56W@X7Mi*R)
z&k}V-i$7yT)e;J+Z(9M0ZM{8o5eKKwK^n!W5NUT`x0SI36*<MB4fao1Nq1LC*8+qS
z*N--zp;&tEke1Dv>*Y2+Ve0v<lz<#<4%N|iN@G7SrEFK))+)eZF@00eDsM@&(9}!q
zy~CbhENT&KaXQbtz?DJ`ryI4wbWny3itx{H@<godTf1mBKSgtL#qIdLv*V=v{9!Ti
z&=2U*qP1p8ET=>NPI(0p{%)UhEsD5IKFt!9F)*wv_}TOAhXrHf7qkw)58i}Tie)M2
zGd7{wJF9GLwtvPK6;b06lP}81Rz-ylHR?TkA*4?BM|}5fjo->~lJ$_)4AJ%mDBW`;
zdn3Fd#L<}CIn{iqjw92KWT6axcEB#@2PoOCpw^5RC|1Y>e+B<-{WB6C_+HwPQWNC!
zH72yBoF^vaLNxyYPyM8qqEv&w;MKd1Qiy#yoJl)tNhV}I#9BKmA+eH)*i?~++0<|N
z^}sQ(rLdW6JkXhWVnmS0Tpyn~Bj;yo#xD_-v$L~Z&g#<{#Q=NU!zOt(s7<$#+@5$F
z%qlxQF@9=e&YGvhfU~kNvBtNyrXWkEo5f9{o_H>H4xfJU`o|!3rx2Nhy_G^faaSu?
z-_J{Gwe&;Fl>U$!(Xwfih47(UlubaYrFimcDEB0H)(`96Feuqx(FOF#nm@g_Kh}P<
zOkhZycUN+fIr1plezmSy=5kY<YSPSNXW^s5->u?a`zlKa)58FPonwI!Y4zdpT$g-$
z_rFgJA&+CbViJLTGI8E6`;_m==j-i3KJ#h$i<}G2+GT=8y7*vcht47k^=z3#jEcQ1
z+G~9OWP(+YPg!%)Nw!4GzPO^#X~2f^t|jY2;gFmV-~6o|-ZSu4e%?-=H+*o&l8rsJ
zJ36^QgNf(mr<h;ELx{rnuqoE6<(GC`P=Q+A>O@WFB54m(?<L(?g5UYsi5+CEFXYR*
zxeI3WER=-L*Z}7Cj9J7A{Vwte9^4kS)W@&m&_-2sJ%<YshsH5Q9pUpgnT?>;AmKDf
zwUW|)<#dpq#HPqC>4`4HjLbGNsbm?Dg3(5qmmDLOWX!OqVF%A1*;~;LGWtOmK{Bk6
zX<esog)p|q+kR(fH-LhhCOXBzh+@-sfW5F(qdeaIvF2p+%BO?oa?x*_PS0<jJ+*b<
zO0l@Dvr6-4B6W6o%Auc985-!wQ~#~OBzUr!(NDE=aI4k<?JCJZq)?Oke%>j6ho7OV
zxlyH#)2fjb826xw=;fSElxkoDk@N0(8D#IC!#%cDzwUB*r^=JnQw{}^)Mm*Q43-&x
zq21LEe6eB8R`F#U^rHJ#?HeS&wh}`3IWZCXeRoK`uzyG@n3Yw*vbh=2jP1=!D25fS
zS|F)bjn;KH)+eTY+u;wg%%gArUQI+gGz>%?dCHeSV&W`%{Bi({gDs~t*YHN~x=|YS
z$}eKN<Zh-|f}Mya+gWC<(s#hC=}#GuNqfT@FA=!Rm=l=kbIC{Z(x<#keaw4cQ~|Fg
zIns2B^(=4D3wCpuOAXoxvUzwv$shj@Lw|Lx+BXmb4-+KCe~Z3K@uvf&r~-WGQ!E<~
zK-J9bQ)C&NAgkar^B34qyEL9$p51vvNXV{kXEY5|*AkK(b>pT-VoWcUo{|<4IB0Hk
zWGO4a2FhBL4g*lmT~{*-f(FArU3nZ>b}na;jRdSeQd>APuMV~#SZ{o`%6zmqe3JTW
zpsCcg?sCNgsWM>{GE$9diB3tn=tbw8x7~ZmFF1cXa}-e?*wkSD1*>vi<+V5Xy)^fK
z8N$7;&O=ql>|_iT)QJXBxZ`n_Dni+CaPNI8w@S{r;gJ;Vq$1XmPOR)@>9VkeW9)Wc
zmLgMN&6cKAg?je&082ljW=h!lvm=XKYMP^*2qDFvUW<ZD{Y_auO@;}xmBAd-e$^oa
z>-@h@!iBEGeZF8|%MXlv%9+Ty9_&lMs7yh>OWc&#FlgJlhGwnC<LpOSRUgNSGTxF+
zT&u9{g4?jKBCi^HCD3FJazYk_W>KJz$rCl?{rm$F7?&;{iTQaP74dz^zdsJ9__aUh
zfASMwj!QIVH&{l6uTFR93m5)|?k(L%-Ct}fZmda&GNX;iUz*9LM?o=_M0oQktA5np
zLB^tu-*poOcW-iZnrI{$$fy|1N~CPirF4wzK&-DID^kXr9Jp|4pq;B;NYra7E%M_-
zvQB;%tFk27X3eVVP_zr3kxzR=wZ~Hnf65rUGxQ|#hL70nK#ze%cp*P)X`8xqV1}un
zFy;AL@3rk|{ydJP1%S$U;OJmwc79;n%idF@1z4mIO>pvSn)SS9r^?L!DsF9bNIJz|
z-gjqvzp4r{JY!$42iPg`6lp|`y}$887QHxPf7@k%K5yL-q+y;pl~OVPt3m+s2G%ZG
zkJfH#r2$_Gq`qCcGS`c<<aL%a%ppr<3)m7I&uw$QQ}&ZJH?dptQ$}L}e~q79%BF^}
z-Y62K(CS?)NGBFemnPeKXpq1_g0w3p6iqzXw^aaAD=hp{f`y|mF{X~h4FBvB>|Fe9
zY@FjGCYvUe$&zO!uCh=XR1%@|d+j=39Gzs5aH!)TJZqpw*wR(;D*b1Ac$$~F`EtDb
zS{?HOm~yYPnKfqO$U$|xNDe;1#{VZN3=bNYpX*ej+&amdCoj91#-|%ALOHAy(>3yV
zMBaT9hNZKsTZE%+FXTs{!8|XnfIwU44UF}V@KCLwU-+mx#M9y(2UAei<|==Lk{FB9
ze~biBwk7yVsTQ&Ds1+-AWF=&TR$|=Z=<3+XDrbh^SHi@+U^&o$4bP#&T$zW4mZ8|^
z6Z5mbz}?c>N>V<bjc=YAn~nDA_T?bXzu$0TdA3kn^$Ku7b)@r4O^X5<8oH%vr98LF
z1$-E90YjkFW1d+~PppD^@frA#ekVunRw}!w<9i$%Hi^sI;bFCp<~+3u^MXn`tR<oZ
z8s@$CaLstG1>u$bXWBFffm;&4`R&l%w5pwk&??yw^3j=;=`f*QDPK6ESqKWy6zQ}N
zOu55g_FkTMZ1{eo-*sGBq@H-ckrC-3Ql)%-o6jdJT*W+v6`rb|>cS%d%*8x?>!t->
z0PzshvV5_!@#+Sk$X{!UnaZNc;Ku4)(OFvCHs?b@KN$s1D)@-jiSf(BHMu~7PHaIA
ztRjFhP%bGl+#adLvj_SzB;!EemEy2jf=xq&T~olg`C1G2+>kWSknQ_={z0?Qv|?Rp
zMlvwfhTw$YP+QtDsab#*w^|%t<}pn_yeE9}uu<)o*1SYDx14l<HO6Eh>SKmqYd>YF
zzE@K3`aR}U%XNFS<;Z$_@@S|tnC4BCyo^YxblL^bE0T_L;p$>{T0fdsMDMyuwJs=!
z%Fk^#QU2!`+V^=+)g~LD5)N?C>YYUBAv6qsX=XpOj+HL%Nwbt!O;61Er~h0il%l#q
zGc`3aIWwZ-e$>7cr5)iqLQ2-tbFnR(628MDiUVdk*kod1COAmX_&&pC7`NWxUQLx4
zZ}m(3)cmtk%L-MUUZ2Wa>l~K8EC$%d)$8>mtR;|0HXt}k$jNst+rCQXLy=Aa`q<Aq
z*IBJtbI=1|MlD69A6e1O>b4p?S+}Y@6n*^uen`#-?I%Ze_tFk4os~^eqk{|{9I<I>
zlftW@sxl6ApH}#4SDOmO5mAzh1opdO%8%BsoV{TS`@lde`&68IIk?6~Hxvs`J4aA0
zl7ytf0iSB~b;7Z~xTgnbZZ2YbA7tREl$To`c0YKfC0QuSBe~!@dyh}67IbrrvikSQ
ziV=+K5j&#o_R5#`-r!{H@p!*&dlJL-CtXqcbvO+0vx<X;V#t41a3=%e0ojI~d(Ryv
zGZ<0L+=$k7OkCW`k+dr+>+pUsXZ^6K;iQ&TUvdH@LZ1hNsmPn}B_;;v@-$RSQ5xpl
zPggBIO#Xa@j3Kb7;L+{YSZu7^(E3Gx=3XYk3lB;tGq!f*jgCiDZMu)-SZ%f8y`Oi~
ziOD)#bQ{c+X!&ZK$m%p8P5wn@OeY=)6Uu`-UYkxblvCp*ZArOl?1HZfL>5rQg*mp=
z?0z3-tx8cEJzB&&NylHUNK$cL)QW^ieG}MGjeVA=kPwEYZDhfS?X9fR?dbbs!>Zun
z*4^qPZS(4gv}4Zim7;{^H#N&V*@GE-voU|W#`-Qf25L+Y%BbI^*9szZS)P0P_#l^u
zs@Ut?)*15zTi`KlhaWMSbJFQ1aFZBqajsbR%G4D)D77%_5wb8Nj~?iYtCeNurY@vU
ziDkv-6GbdsZwX!IjfWp+$@=C)v-3r~iMz-g-gTk*DX1utlt1plOx`aL;oM_;#owd6
zHQ-g5;mk&qD*&MJ3p{Ci?~K*3m>P_TD86_CPI!)1lJQ*p#h9T4m%XkZXxSyL4<QIr
z9H^o!DZO`YIjfU*X7m2`MV-yNY+9zkFr{Zmp;+CS&9_M{Qaew_HrY0&9j*zyXdIEm
zfvFUC2HYh68nT*0!^Ae;3e?Ty=jgB|i$s$Y(CNa)Mm6`QN(Stsd5hH2qgf=?auT8j
zpu+R4jbFNAIY4YY_*8aZ!?`SVd35LEmS#c3Vz*+i3Dp^tlsm$BxOi|FU!O4WRGIb$
z6T|!C^16K4=s$<|vC;I=*N#ABWMn`aJtkP!oL+2aP2s}U!&t&wvR;8ov5n4p$c2VO
zMK<1`fuw!UREfoAGxR3KPq%kKoosV48HqZ2rVL+(4x8PWxAz(sIc#Bea}J6N0uxth
zTf3E+c(sZL$lYY<pU!`u=%$yHTEUZq&f^YxQP$caOeDOY9p(1{Rujy$JE)X@c6eAU
z)Xes_Z15yRleLXCc4_o+L~qvGo3^&+@#f*Yj7u2L3jHa0+Zn;id4@TyB>$i0xVUy@
z;A8gg4JT%|X<MB-c3|2e%}iOrW^+%E<H9=&AfFDu&AZg~B~oI|?UG4mfXxJL{;(8{
z8sg&bO^6*#WzOE_Y5$=5g~dNMRu7}Fj~&sOmWRflbho7oA*EUbnryoo`deWXnaW88
ziYo1pl;Y1>|NP+z`tsV=0oP1#8?xS=Sw6icSsN^@rR#vB54ubcYc8dPC*m0{(ugU9
z$F1J}MzCyK$N0Yn>GIim$!h??gGmb9K>pp+^mHtL@~CpL1}$@pL)MmbGIbwWBKg&l
z)s>-vYOcIyLI5BoJuNZqn;<I<KVk`CtHq1doG^K2u5bQ)kp$O=AXm8ReQz*98778Q
z5OlIK(-I>`&w5x?6|Z~aMm86>4jP8y9m-W5__7)H)%KTp7*1<9mSb);7!1}jYPL9W
zJ==9%4^Seko{zcdORC?m&L1dUsKc2<EUFt6%WF-4AaRc<a#|7?;WTg-S@eVw?`_dz
zojXqck}0el0ZFQyNUlTkZ><5zoxWNV+pi}1MLfyW^GnI1s~GK(_$fc4lGDmrP!QH+
zy#`JPs@*xyJ^zD|tiS1LNDW&4Rw#T25KIFDz&;|TU58-8qmdkpB@{;C*ACK_6_*L8
zTWgKP8gwH3w#ox^Rh0{IW1`>_S6&nK_JmGG6S9WPK^SyuSf$BrFPDA-v7X(32wZO5
zAvoyWQmspt&(p*4-Ay%0Rw-(~)+{lAs-_zurbn|?KLlm7+~n=f7|1(B8-<<dj~BRR
ztq|tz?CDe6%e~f|7uL$3GS@M}F@`55US3$y<4F)r4fmlG<z>^qsl_KKr2kyaVE)FP
zxwD#_zooUZ4mO*zF4drBBY|;O{#O2|Anmbr_kq_Q$r-idh%tHPGehy~qN1k9WMb_)
zs~jVxdV=Yt_X4u*7q7B}wG=8S(s5SurD1*X9PGJuV&<O--fe!js{x@oEPpUjFt&Ql
z?(>|!46F}lrt3;gFPi~E(oRg-?#KxCw5p2CKYy?A$n$nxm-R3!UHY?E3(6`-{oGAV
zvCPUsWK#X9*4BT4G`w>VkE0VHG~B0WM#dMHSJNA54_IIcC!Bocjg6d3mLxrqz3xW9
zRzWPJywYMc-YJ>b5nZKjef#{88r!B|r*wddW$JES*`!o`T*`Pczpk1v(=P2NPraM8
zUD!)AJx9XWVlo{SvOq}LHyjB;%Y}8J1a=1ld|B@*ymLUUu4_EJa=Vl$B18ppqyYUJ
z?)0NSFsovR@69yuEnMRdkj&hjzBFr4v+J5M0uHY-Qv&ibqA(lHaH^VNG7{;jA3Zv3
z&8jB%)L9j4N6eFMemp{D5%aChAI?gq$<%qouUKa+2Mc&E2BEy0FhPMvW0XZ-K%okp
z+W9ciS<#}GZf55)%^ln+S+5`oZB41-B+Td#;q90cmGTtc;nS?z=1vkygv<dFU1xA@
zi5ma37foUewz~XPe1C*c5Ni_N384wAdDa+Z)^m4Q@cXCCrX6a5KUp(>?6{E)MkY|f
z)<3%o$EN*=ji#}pC|>*&&A<DF8EGLE(Pv+1e|k*ecu;Fv_ueAp603!|w{ys8F=H#z
zPIC@EllJSQx2TroWQp@|5g#4O(gq@;B*T&D{6g=I6=%Z&xkoNaxZP_6XweM%hx?*O
zgyenCrL3l%h6A4)<0d`Wnw;S5q|kc)nLVSRij28ZXOQ`YXSc12Z*94!?45h9PIA{p
z`d;@214=cM6G@1d9Ex%B#A4Gp3d0n+QN=Bwj#Mm>6kAGgdd}Acj_vabb@;34SyA{U
z=Gv4Btq#-Sf*>6n+I-!Sf!5;c5#j;CP^#s+vH_JsLyu-_Ce<SPh-m_y=N@KD&i$#J
zrOXAQJD#)-YK|u6ib;5{?4If1tf%P)<}mOmCiZgwaeJO?wtGW?epj597p+?`qJrMS
zT-WN#KdTZ+v+c1;mO|WCbc*d4T{T`XBi8mnHduGD*0L}K8#BF@Hc^I%H(?q1q&uPJ
z<h!+DNxdq15u8tl69+Nnet9Acr<m|vq>Zdho%O+esk&=9NV6%%XDfmPuU4j|O0`xV
z#UtTVYp>vah=LiYpo>#3feUr?*_>Hls21>^{q$g-w&fq(QZ$qN9KLE95EW3=k}F;s
zkiWSn)XH8UDYqcmsT)%&`LPo~SQEv<1wlwTQN)tKcL#2hwzSL0nYT4lpTRWK<pF12
z<S2i{%P>ju6@vVOSHYfs-|0IF>lp1v%wbGiSk46$L?be1XQO)QDMR@mxapLJ>XBi`
zZ-bv^97D0FY-5{A3OnyaD>ap!lINW|9BO%YI9G&?iY7}#h&c+2<YObQmTrfWnHo7*
z3*qB=%b`WO34`dKPqo#zqz%<NBX8{xemGpUS4mY{#HLevpV6Dc)8gi(Gn(Ga`0%0r
zP8m;@+T+BuXjvq^l%F;uG7~_ViI`UNHC6na3G;nP@1Tz}Vm8AoP}U@`QZm%aSM9Dx
zUxgH|?MGD+Xt-F39gSvZ3ppVD@G6@5y4a5Xw%L8TJc-=PRd1dAs3JxDiHO4LNp2Tg
zV5(w_>!>(K(O8jHAIgU6LCeF7^&XRyaBOz~IaG*nJ!0$e`SNycpjr5%H%&@P*Jv?H
zUknN7gD}O?p0hSKw6r-0+yqUxRNe{HU+f#{WIp{Z68N-ebaM9`nOOP92H<v_CI8_{
z{C(HA;?YhjfxW#hnGs^=((IVY?dMOpVf<urKEJ&M^s6eUgj-E(@oVM3{#lo|+et{z
zQ7+ZA*Pya@Vy=}fk&V<W0IxMKg(ou9RfFoB`TrF2JLN}N<SE;IMq@og$mFj3OLE_w
z5mwIQi?LMme@K{g_eiM9xFJj+L}SFRql$7CxxL?rN03JFtH0FlP~~pVM!2}K9@3pK
zN>=6ve7ljjT&^n;hXF#1!~Y@LQC@?rc)PKrmNW97OuvqfdhV+VaPSYmk9=k!HzxX-
zK!|2{ZIMj=6ieM%rV__!1IKMo+vgm*;%Gkdc@M|0ZxWPiU{3kJGHB(1K@l8UfB}VJ
z3pMKpy!9mmx@WlL$}&2s(ZU>@JGNVX1-rGZz@+<T!CrrPL@-x4odBurg`BWq1;#h}
zW5OAaId6LODThcl@sXzjt4$r{N<?vZZ0s}k*Ie#A1n@rtjYgIR^h`1Rv;MUHVQNFl
z2dx^cvmt<u(&eN#=$5qf>S94o5$%1j3>Z+CzXXY(*d^MdR%Ful)SEeMk<Obi(Vs6F
z2w|=sLn&Q4iJmfFg{PovS`+pN!3wvA$rJ^jifnsd0HfLDssY3=QnXRTyt2RM=QDmF
z?1X*Lmm*2`_xA#|XK}F3c<52t!_C6m4wtUu+>S61mO1+{cEj`^6mE|-zpcnl@y@9!
z;)AQX0K>w~yZ0SG;3Whjp5Q}L_OJdO4Gqt~2^oB2Uxw`8CWdH(2;bW2a4ap0E-%zJ
zId#g=hmM#L%~i1;i(`e+;IXQLh^rHZ`9*HQ=P@Nlf@p2w$ov4h^+USWSk<n?w61q*
z>w1F_;&YQ!1@K<{jQ6j<EGjb5yuHd`dSz7x6XNB>jmEfXvKe|mfiOiUK(s21aoBDS
zJTJt{z0I6J>qPRR2|z>?5w|b{tbU`12cgL5Uu*(t;RydY@#dWjPb-xqtKN-l$W6S<
zs}+IBmdbUJtg9jEhca#%RlLXZK@tmPdLzo7s|HCKR>ND3xYN`bApB$;pt<m>I7J`}
z1*A(z!r+CS<V!PZ0W>4iZ%lrYd7o@*d2cWmp4m`!y~tl*M2Oo7S=_?>07?rOUq)~~
zjop8$%@-hNIIc>YgRzbN1P8m;4C|m6b-@>o3bKv+h|A<3v8mA?FvW|VVceCJw+1-4
z%=HlGs^1%yo0Y0ZUHU2S%`ZE`uMZcUjYwrp%gF`GEC5;*E7CAF#RcX4fIUc?BV>C3
znC%)J5#6DiP6Mz{eK7}K0Lzm>Z+`7|-3CbaxHq`G)N*a5f<ah`qA4cnv>9FDJtYXn
z`)zpjC9^6Rj#j7hD--wHva437@6%Ise0{DJUkJ1%#d*{KsX@iTv67;ETs%C6F}t?N
zow?V?m;5M+XI<a8h>s2#tN>y<ASgg|9-mS#olT45#zlGvFE^<vBTF|jL}2vw!rU9J
zU~eS#tn)taTy@Eez@^(e%Fr&4J<maR#(}_$a5)mcFy;(5DytCC1Zp&?O8GIR)YrN|
zxty#QWEh}>W=Hf8pOOu7c$XPxbdwaJ%qG@F^fjIB#&bVsczQ;2t_(WVIjl20lyQ74
z>gcL&FgL@9n3PjtLmKA1b`+%`RxTEN)zeyn@Ob*qp)nr0Fpx27{1QUxNZj~)W$iH0
zZ)X3m%jl?$`?=kc7T#bP%mGYD|8)|Ni{eAna*h2~E-(0lVmuw(qn=pgcgnCJUV7|U
zLeX~H1yfsu1arkWBkO|ZI5V^+U9a%tbYht{lQtDfWjSN=7^bGR@A|$kCN8(tDrq3Y
zk+jdb`L+9a0@drl3?t0zkg9C#$DW(e5H{*-C>#5Q$DP>WOXghJWWTf?_k2(+zUV}x
zP2*A1^1P^vsf1jy4|Qb2|HIr_1-03RVY>2{Dp1^^xI2O3#T|+SEl}K{1(M?YE$;4C
z971q{ON+ZFKp;?@pruHF0NwmMd$4=7Cws6tNoMlRH_1DbH_v-L_jBLohDz6wlnHU9
zmm43hQ~IOxm`(;w6*DhJ?S{zB^Mu@}$S>-eA*mI+T#7IFI{?EPeg&TqE=2gv$O}9D
zQ@Xi31yxFJ3bhV`btwN=HgoEfqE22;NyQ&?bwzVK=+EDc-e^oolp4Bhvon=YdXcPd
zGSX2AYbGVd#(Q;svy}Id!Kbg(j!tQuyRc`r7knbm!U=)0i;9Scw*#j_6ItO*^6rgU
zPhqieLrtaV0Y&XIy9}1*Nt1>EXttgp{dL!&>iH{WSOo^Kh<a87k$y_dP@`3$zgd%^
z6{Z!~t|h-wrB*^4sUd3Wo9V(j5Ig2{aGWlJ0nT>k=xckmS`tnRzA}xcteTv4oUGub
z_bU!tpWEIpvlexCmYNsS*3f4u)GpyTQcVcII)cZM9a{?ZEy`OGT6yf1a9)q#W;6>j
z!UTm919wQNFv|Xm@Cf`wIHQb|kvzf~r?+o$(S2pDe@nBH*8c<e>i|>6Bkl99A6dg@
z+Isr@Kt3g!{4CAI6Y*WvPU@E1vEpKxs-;>rZ7+2+O={m8vND=GN@kx3NmDEXPdc2c
z%=No5wb1kxenh<A`IMVa3Uxe|<koLh$C-fUSrij=;V13q-|6%dL>R|u$BA(ZAEzY0
zW_@E|>!4?7R(G*1@-#F@>9%qX1jx2i=9M>X8b}M%{kQ(3#yb-gZaXZ;tRvcf$W^Sd
zDT3;h4o`${maT>2pHAAtCgj^3<pHU|O=SW5ol5sHh4T(e;?;;pBl+V1s+o0E(^<-=
z4`Mvj76;rIRwE^uwaV(^1jC|<k{+hi6F!++9q%gBUav(3r7Mzu98LAgpIA-6>x!Aj
zjY`+3h}dm(%k;=ZbJR0)I+jl=)v_wp0cc|NpntaBdc_Seyju)1EPhdHQ3VuKR3%VR
z$f)498kMH(fr~n2&uoqh(Q7R6^u1ARh8pvS)`WIESNTdnipTkhY3Qp?6n@3)LV2qS
zoC&K{>QDkXi~aj>=Aqc$*K@B#kkn31jLp-jERo$#16c|5_AD#?j*Y^C#ohK&KBo-i
zB0i@~JjV6KO8$AoxM#AdD57Jdu45zd(QTmg)}pJeNa$cQt!O=CeXKJ8YApt=cF`rO
zb}@~Nb7uPQVLVtfag@=j->9Awle)s+uPNVy9DI&2*q_VT=WS}!@Eo?@O?XJ!<3As*
zu0=xPJ!a5D!w&DFmX&wQP}expfw9Ay1uh}kKfQcfURgm!Z%OzqGLqcsK)Mq-Zb7VU
zEPZ7-O;}#8f%lUVIaTTCJ`JUqn)o~F?w%hK5@mZgbb53WL4LkmwW$HXJT-4)7q+GJ
zddADwg<F3g;GFegxM`N=<}bwrm*!T=v~K_O^A^sJEA?|;B}FE2pK=+BoSWI%$|?Pn
zD}K)cc~LxrtfkIOP_nv|vaUdTPj<`_C$*-o=-~RvpH&*|sV?nv`Y4^vUam*ZFV41p
zGl%O;?a<*2w+yUUdKL2m#fe!=t`axEp0O4avG7J%njW=y*|1I2A&4Nz_v~yUelgiY
zE!fM!;}PjoZnC;$QLWJE>=_Dx*|sga1g_g%F6ncc(HYoXPx88&CW3Dkox<DBTN2}D
z;JlCik|uNR1n7h6agjvvX_o_tre3Fz-U5g4*>v1nwf^z{U_<wC??lb-ZQr5|zF}py
z_E)h-Rk9`^UG~a}oWhCs`Az3ZdFdHI*Uq_3u~jQ2?WO9jC2$pm<4#V@fr!NX*6lge
zkEF2O;uS;#$}vXZmg2MwuFJn&!>4YK_do;JvQ`6Jes;=?vyrT|az;1l1ibZ+5<OhD
zMwgUn{LT+55J)bOec=skVSA!pz*})aZF_Z~F%2PNR(Hi9ndkz3_-qpI%~7t8`-klQ
zeZbS-YFf!}Sfb6t@p*~B{t<-d?Vr=^J$>Rv0ZVdIoLwI<X}ehI+sW*O)M%}GQjI>l
z6B&neXPo@w?Hwx@@#q~{e|orz5HoFDfEdhu2v(yx-6yIIe)7v+?M&a!-s{!p11gsU
zW7T%5ViThPNiI#_)0#2)UyTr5cj|5O2qmCg_^*YP93p{=^HP{@7|{i9e{nJ0-dx0n
zSP7l<UQkz&t4T2wUne3Gr5bSI7e(Lof~Uz8;GFXzEl#t1%78uDlzE4<`YV+^dpY$r
zE>N%7v<4xq6DGo`()%`9=k;6_&~c0<h4Te212f)Ld3<yzHFELv`_GV?$l?9VefjW8
zqBszocu-py6<{Spx3wTFru-I0ZBHTiZ81!<y<(PC5dR=wI%v7#eIe4IWh*&H34Bm7
zZ*6u?g3_Bas^~L)7AH9{zvhrD8(S}v+oZ2LgDv_wFeA81ol`s|A4V+6Dd+?<aSrvH
zUwNw@ZyHE%Y6ws-%m&z&7Q@_Q>@MkrbIVlMH5CMnmD_W1(<)v9xjQg>8hZ2JCfxYu
z8E(SF%pc{}kX_L^uSk3ZmbAVQHejNzVQrON_+RL2{xLBQI1l&sg$ng$%`pS_-{%Rb
zPY=umcj$i=OkriU`#8HTTD{=Spdd>gDAY)d=6~*Jp9VncpL=PUUb05u|9wz+)kv!G
zBZifmE)o(SrP}A3n)#K8jI3u?yj9z6$!kt#^2*NE5k#vaNH9~SJFBhqA*)nI?!?tN
z*1TdV3m!Ry+Z8;+w??#zhKA4T|NY<eivJ(H{5%yBBYP0^=EI+Vva)VKeCtF+Qqazy
zlE@@*(!R>kVK6Pcq4tRe7ic!%3HwmwtX<-NA3FIA<Fq{g61x`uo3N^7Zm6@vUsZ9f
zdEt00`S-8yl`J1_hOfT_{CzOYoCbC`Se`;-ze!AKf#&ZR-nwfb5jq0uEs>E4Q`?7U
z{+>kb5J=@ZYG|nXdx#94sqd@n>v03k8IRSd#-9&Wr~hBCk-)X-qD!J*zO9vge3r`w
zesScddi8|Z$WqgVxq6<kwbYm*zy^rWA4YAi`}yO%`?CR|K>C9BVVdf6V1NWbM!c{Y
z+s1hH4XZrr*enxSwtwXe3uhAuJ;w9iwj>k+1h#sDlgJ~E*Rn<@Mw&(Xt3|nfX5?k3
zXXY3QCfOH__p6AnLSqV7$cVegRR~P|3<J|RfVL65Z>v)<i~77dN0QGHH@HoRB|&S?
z6Kq9rSzoTAOC;<*Kf}N^V(ZGRAK{1D5uJy?diPGr_CVZkl}xXtmWyRMsQ%b?yPoy+
z?p5eDJvXQB8XLokH`42i_f5t(L6Dis)!2(o$kX<F-;jDT29lY(n-}9FIIUQP#(YWO
zK2CKbA=T44rf^iqV(dap->7@kHMf{%>I$>YdS5c>g00q8x0iQl@D$>=5_|HS>C(dh
zTztVKsf(<EgVM+9`Vn}p_`~)VpF;@hdkABwrLk6laLrD)gfE|o$+Q!W2N;jD+1o|V
zO=l?h`2IdPfU;@eHT=?M&pftI>>=uLtY-Tat6O}largK(ghF>`59BfSyWdys<a3^V
zi3cbPWFh=<5PttG<^>bL3$BkaaC`Bwoy<_ap82|`TffX|LgRAoFzO{Nw8g`*Ki#>l
z1(`h(#9>D$aP{2E{>zj4rw>010>0G?^2MTFY-2j!yQi3*l0yU_Q5cz`Et869Oq~Aa
z{zg%!^6HHk$nX?Iqim|wFTpyL>7u%*vm#q!tmg}-Hrny5$xXyWjE<}GcaAoTJXLnE
z5hk8con9~ldZcA6!jU}bOeH$H&eAPy28=9J@x4s)aTQ5aL;xmki#T(8F=^oFo{c|G
z4!)3VYNuD-*~`%FSOx=H@gy^xd*@EiI>IrK_ou62HRrvoraSWAHZ=@ekL-7rF^8hG
zX({_6*JUH8rpBvpIBCA-jZD9(@Sx>%!XAPnOeakLwj3YVKXcZ1o+YdF_>vuHXjPZ7
z#@T=>f>IpFA2F6eI!vS<6FPc5waV^K^UO@M3`+ENc!!DF83^(Db6RRyi_f}Pa}SX`
zx=EgF*mdE1Cv1|o!apg4hg$L<Bwb<9$0D>^xL_xG5I8~GEV2c_c365<s<jCGN1y=Y
zSa)k1&zvXXo|P;@6T0EIT!`Y`n7~8h#gvQ_T|1nohrFo5Q#gxaH{iTUi`G1%_G(VE
z?QbmNZ<`uAOk)<)<12>K#=$Mc`ddBoqW9(uxN1F)h21l}J;2e4HWsGVq~p@Y(L=?-
z)u9s}vn_6O+$(52-DcjcG1j+pYC}*89%ISl7^PMYfjf+)Aosi_QJd32*?haRm*J7P
zyY$AGEXAi%mTWJjY(vX`a5<w|7N{HVxz)#cGm}#H;kc;?2iz<)<=+R%>wELK#q}xG
zY<4!%bGlF~Yi4ksKcZQLbhR<HIIUk<Dz@r8gRdJ_GGt*7IRt+Rua?tLlb7^KzRd_v
zWQ5c}?$Zz>UJZF&lcHLmyrdT2<-Q*J2gPbzc2H!YbRy9-(PNCU{&zrK3g4MfAL4kb
z9bd&=ZZiP;Y)+LDwIXCxa!yyy<8g?ITVUAx``}OT6e2-Z!E%ZdSAhG+sk1;$3NA#%
zvsA#y*-!p&^QyD6M<DG>zS`WqsZcxh+x+?yncWisTsHlDeeypRh)0SpvZf78&Pe8(
z?6&P};{!JJOSx?d#<#U5gCCyqyvJ;s*K;j`B-cvO<Qt{_sA?nLs&Eg5C)rPU-KH?|
z^QGQ;omPs=^g=UD#XhMCuISxm8!efcOi`6bXp$S|YdTpXHEZACVMoErl!B~b{r#q4
zx{`kN)Ma!2V|dhrk90D35m;wIdvfyhkhy4u{~P49^StGF;IT-m$y4mgvQ6(HvvNi}
zR`892L9i3=i%G-?AbANTbCK5>@yjr{?ikVcDu8=zb)vveFi1~&Yvt8)8p9Wx=>x;F
zAOP>aIgq~>Bhw}8CoZ&<bN8}~lf*vEZ}*+W47L2Fd;dj%S#iQ}k7)I@wbjVl6gr73
zc`ZY+Le&7Z_%lL`t!@*Gn<tzxWKLqdjGezT9l;v+;OW+PhQcITPJ&LjMN{g!W+gQG
zk8sf+{~$d3qQK6|!RQ7DyKY0vl&xSpLqpLZ^99(^zpRn>N0ygc+@?Jzt0!@uiWt5p
zT4eT8YT=IF#c}L|(|MTG&5|;h@;XB=z1`8nqp(sOy;bLZ^tB{mucg(hy=1^5@A?Tb
zFP~V~_$W`Rj!zVE;oBrDFrT?0vvI}W2XJTy;rP!R(LNt~XHU<lle?z#4%TAq)|@@f
zQUjBcd$HatX(Jal7tk0t{ZYPF9@Q5liArBP{YUkiQy#&}4*)ZVRq648CQRIiQzu=n
zpuN^RCWzkhp5yOZ^XD3Dv+Y3YjGm(4r%M>5Kh98_WHG<6*m4ikX}x4>n`d?-rxpct
zuu+Q>r`Jd25P-El_E`AII&T|Cp3iHD86MpE3_YB#ALo1qm<pUP$2m4xuI$k8wa?*)
zF6I?MIod}Ywcv`d#^r!c3#Deq?OKQEdL@$pZk#sNLMt_=V{KxNerMf^Zqb{;^X3JB
zy}xx8R6w_Snm~1Li`r{{jFzoMQXm`E^XJq;%}RRqP<46>=`nZo#RH`dk+ZRB4E{&#
z<^I8UNvfCRhgklD+FB%AEj|9sCBu)N%xc25p8fK6|AlVNW}Ev4OnzTu6NhQR#fo|G
z2I5W<;Pq=a!=AuIGz=)PQ=3WLFH6(AN5Lz)fJp!iHD{l?K$NI~MkrEQr7GGM2~$!A
za#MpL`TWc{ZuGM`+zduOqTTq?A$5S*6mTDV`L-QbGaB6|>K4H*ku&Q$@+lZ2g@WWm
zvr65}BhyL}a}XI>dsTo99o_n&7qL9A*BY#HJTrp#*o!wq%OpeI+hT_NmAPFfxpVdG
z(ox$n6~oeL#!&&d`pn_uLyP;(4YHrwZSI6Y7-kBVTJK8lZM}`Fn8R5oC2HB|iW}wB
z<+2tL6JA+oSrX)9q>-A;EG$;`!>q~{{atuvqXEQYi?lVhRpvlUBqb?ugH~;1Acmq@
zXZGBSh~@@l<aB2&`*`}dRMS>9$(khR-tQe|>{hR)WK#rTqkB;ahiNtZ!|WJna(B$D
zt-*zh>hFVo8hzr^-}U@!tT_%$gT>g9P8&1tzYjhwSHr8&r>(W`A&?-41Vq`vn|#-}
z_^|F6!wKg<M-Tn&&<RT|_=ca#;$@2p(u`c<`-RsErZ?bh!2;f)X#ju9>3tz5f~Os-
z@9wLOw~qj*+8by=Ud*{&czx$NOupqhqElU+C70hT$F)O}G>%+$-}z@0^v<XD&PWrQ
zK6-%BOx+6prvX0<xI+8Gx}$4%#puq6*qQ#$Wqbu>834-hLp0lalVp|6&dw8hx!$NK
zk#1J2CT(ixjIxj5-(Yl1gNIgU76<d<MPjbmcfH5O*uzSD7|Ufc*snjtY1bNlMQZeC
z=X&h~9=Ax=(!iW0D6A(Gjh%OcCW(HY$C7BpsTW}v+K2LE-bbB$o|V$^iFssde(9$W
z9yvK_NR|U>1~;rpvYUFP)K9Uup$B&ATWB>%?`|3$XIxACC%cN4&l5g73PBp*xI8=v
zDr?M`Gi39gx|)n<FH#fD>jU^3$>py|xL^AH<o_Wdb02Z9Kn#<$__CY&49>Y0T=6sI
ztDsXRNL;yuPK}Z-S7!9c9E31;VNclTBD4BnXP`ddaK`ViW-JtSoD%o-W&OW<wGMm6
zL=>i{mS0@#(?@u~YaG5Ry&sZ)4TwGbZBXOk)6{+xv=MXC-T?P&JPyO@rsw_CV+1%W
z%;+RE29T$EqDc@fmmO*TvM!NLX&IZgY9^bt(clu{R9q0M*j597xhfXM_u@wyJUw81
zE{CMi@_=EPOWI4h#cJ!y@|C_o7#kgV(h2=Gfr!r$rlOLb-$~%eCj)2e7@cuX3Lg=x
z#2IP1YK$0q)(`HF(+iUG-o59i1g~%nb!q%zC5V_dMH~q&lwMT>mOZ~WQykl|vBy?y
zx6P@dEtur^Kg--s@3su)ds7wP$ixZAp7Y;`DI^$VCf#ll3{@CNst;(p*NQD`303ke
z#$nnPQOPXi`||VFRcY;G8<aVUZ&k<**<8{rC!}wj8V+NQ=c+^SW~Uj)6yV`?Ox#9|
zp(u~}7SICs;>oJfA5;XD@{bM@eoA%VN#!ve;1j^Zf48#5wC4Cji#hmDrO%yK#U~<;
zyTw)rJ)cwLloHAJkC=(#&w-z2S5xbiv%SM2n{s0RKH$4g-uN<SsefChJ1cVM45Tkm
zNW=K53}tz<V5n_doMWhNPCHbiAPLGX6Lm+A6kg7hR2X!xt=JMvyu(CBQ3J}@Y_$3p
zBIB$jZx&Ah!RjXMm3cL-o8uO}nlU>O)TfUb1f}Xz)^=Ed#fTam|JD$K6s6PQCP*M%
ztrvRIjKqknmZWTO)lDy7_)Ce!DRC6%5qYfa>k~&UqF+3K;#k%@y!&`76OIPb+VS7>
zzDxJLJX_;ucRNlnC-7jf2Y`H1Rp4;KaGAUIt5K71$Y(kWzZTuw<uzv_GisZ<d{++B
zqwrz1N_@N3YAsc_F%z%$ESVnxWDS4NA+4aVtze*~FQujHyQb=dFXc?-#u&C-b&mVJ
zd73yP;;p?#G`*(1T7axR<9r^I&FFE;=^*<o0UL%NjmKN1iGZwSXte98tl6iL^S6&l
z@O+UA1D8PoF+}>;sh7OuNt|+}s=6hTmR{u=4ThG@eu^<m3f;Ai=PtUFQvkTeZ$a7E
zAi^Cv!71|3>{KV;2B(6Is{%e;g?M8VB7aw&7X(naS(|%SAG?wv8;l4D6SVvXY`7Hm
z?eYdE7Vss9R&w#LlCdg#7(A;F&m;am_%56D@laN+&mcIc_l!iiAbrO6<J{KWz7;G|
zD?#x1H*Wq#_5I8V!c(xKS;qw4#Qe<#AQ5Kjbr$SjsjaD_d8vPBi<94R7b_huHcT1{
zn;$?>b`2(cWgqK?2zd`#`aVea;^Gvs@IG$(xxNDzl=$-;Jxq`Nd(O*k8r1VByU?+`
zW00s=+=#}j(rY^Ak~!P1UNk*#Ka9sIx?WTeitX8@Y<z9_nb!!#$n)HCyb4B@U4rR&
zXVh;jbbkS$s1L1Y;(6-_HEPp02`tx6`l?YrJZA^GL*}t_GxH-QE!eF%A142Xt=S@i
z9q`*H8(bs~Y`bE4S62y|r)ciaI%?amZE8DOs;4YFTr60l5^WSPrdyCPs6l|9L^Zos
z<f^#f@z)ekpHp=|O^nD1LjVUu)LPRdO`-0mBe7fhd-HcCbB+!4r(tYNLl`cDTq-IH
zB$r`k5;OQU<7E<2tV)l`Fk*FNU#;t(k0wiPm5YBL<X+Bs5tS7?8&D#^jYM?hY*m49
zkXg+IT%$U91SA~yu1KzV&%g|vt&Pq5@9uR{R<;NCtm@(8lwHClsU!*Buf%5Gs><>D
zkFu1Tzuetc50W-4mNm$p$3RjK=WnO%QL2HRTRtg79irw)=n<17uzd`&%!*i$9(eSU
z_1Dj@1UHF{3g0H|I$(_TlqkE=1N&0M2uMJM=4Z#!2yr!i!;H;BymAFh-@~8u4QDNt
zy}BK<Fa7zmiS9I3h**rl@6`hag^hboi+RD4B30R2+DjrkOO698&eUMrD~%_JFT1{y
zx`va@D@U#8ecn(0%^|ASKdZLyVq)FrX>O1-5yF^fpu8knl-jgAs;i>st4LP2Kb&*H
zy7W(to=3^-BEGAZO`Lv-BOu|wYRZtH(0a+Xf-)&Ibqtn5lZj{Z8fKX!AKv{*d7oT4
z3LZC3?AnpHr=2H&NA6UuU)E*S^m_p(+Fe+W{j;qfvz2a+DIpGb-VA&t0fNjcsp7-$
zfDKE?8Z+MXvAkGjV;O>-DNV8NL#SfQOWM2Ni$-p3e(wW<iGPiT98XjmM|>!~I<!T)
zLKJoAzYWFKWn_zGt-0pPRrz#WN6C@|JH-McTo$b>?+Jr>`n^p{@rsWZ^{ihFwHmV6
z%<W*GRhdR67W;ShYmYM3?+xv@86TJg$}xY(;n2L4f{feUmgMY=oUgU-ZL*d%lx!XS
zjxe?e_YNB9a_KDETUapq%Eirwjlx>{wFRr>YCPAXSUe3%W?M~@y1Z*_2UW35Hsxhs
zne{wEc4f4rmL#>}q)aM4Zq9C2d83^-q#kFwW_`#VR}pk{jUepzmeN<j408)|UY9Xc
zj>0SNFCH>Hw;A(aDWP9|p%`Az=pslz0B-QNF4xJZ(sw3p>k0-{s%g#b__I4JVtj{*
zSU15QjWC9#esD3&2#7G_g<7*e0Y9>3cqr(Fe_GSW<AOnlWu)QV!0wR!jzDUaj`VH0
z>@-Of$1_+_jjb$|Rd>FGZJm2d=Qj}bzByv3CyKDiJ?Xb0=MUqUn<v)<{Dt$ME-Y6B
zI{Tp~+ZSq%$&D5V0{i{;KN6M{zc`RtDJn~hRU8;!fSBctxuOgbt44x6PP~^|g_BWD
zSSI<S@6t>PN0iAEM#3f}Cha5MtKkNx*vGB`Bfk)gzYLkvjG?aDEvwD$I4{YZQp3QB
zz)KeK<3PFLMj~PrHM*_^AlU}X3HdIaMtQr~?2}+6pMzIu!k7LT=Mm56*Ws12nPU>}
zDyxNkQTW|wJd1U7)Z|bj^C)U+$nu4)KU<q}N<V4VK0pD6kqCW~k$Fg;xCwUm`p}iD
z<T4ZAp)U;_`y6}Ne8@eF0M@tv9LFP17xcgH^9z$NclY1we+G(XkDcEGY53{1lOm2X
zxl`BLE_?&2NjOs!?npWYRZ&JR=jB&n{&G$c53IFltp7gPlRft!9n(+;%5wF6-YUm(
zs)y!jQ*>xaR0}NfVX`==N9!!B3C+{BT*_-5si}W1;-W=Al-;|W%wHu$@h?w_^3XXZ
zTE7E#iey-~NxrRZaA5<RBZL^TT^p_PKMAm1jWzR?bsGCevb#clVX6(Bqs$R=9eWgA
z?ETK}y%kJiS2)@c;9?wqK7DH~?ro2L8(+T{qP|!qFQM3Xqy5t|{qF;{x>;ay&UI%_
zIgVXi*hlI0o#mf$4F8#ox&IEC02{_mPRcqEfs(Ro&N|g1Dz6}M;&Kr1{igKsT~CLi
zs$kE2<9l{SXGKG4nmg}KX&hj2Z}jY%D>BhsEKNSl*mUbkxDqjjo!l&ytr$QYy2p-L
z!BHD6(&Tn{<(s64>QPVBmwtBHjRQ{Az5JC=vkltom&}^0tJz=Xx*t~Vt<eQ5$QGwL
zL2S3ywidzCGQrIyQ$H$3!TLH7EkGP5$|tbO@Fi^h06^B#?qg3mLr>fbj3jH<*$Kk(
z6G(9cok=WCgY4gQztVDadPEJ`mj)S`@HBt8q#K_kcrOqkc_qgF=(MRO*yqo&1ljCK
z*NZuW7gO!&`&;5Vs0wMFN=Ct!Qpvf-jY0QxH1SM1gVLUgR_4Mx${q5=I>K#+*=Hi{
zaf}6d!!o7J$7!Oyt`pl?!d%-ijy#8J%WP#|Txxvnxq9b2t<*r&IKZ-ljuhrT{*SNB
zK5DbNhuJE^?=+y<ZKn#5Ho?V<RN6@El|!c=4Fmi^pqG60R>PQGzkyaKnRO#B0}1HS
z@T=wXfijOcv-$gQmDnuN1?><gb52hB^|{*S%N*$<fe+n}XDQAV{%|^<$1_D-li>k`
zzMN49RLg5ynp(4}yH!1|uhcx~Jxuy=KY>J7RGd)v`f>Lz3RT>6(co(PfD#jS|CLP~
zTCq_KA-lAG7hZAa`1<h3H5qN!witW3W%Ic4I$U&z*k;Cp{`|=0PeEi;k1z1jZM2}9
z%dN-LOt915Wq!2C;_I;+9EfI0d8W7J)7T|Q!WMQj^e$-H@7<-x08G=VxM2A21M#Z$
zrCffo>kGhzn$8#-!~g=d82v}Q4PZOerMq=u4OPKy8738wbFbhp6u<UmM9Vd14}va^
z8|I(|5ywfh|7fq$K92$)NL%%HNQ!4>xI@g5RRPc*$`SHWT8Qg?x5q2_A58tYs%xME
zm1BlE@lkGjgN=t<;JeANQ)QEJ>TX{h$;i~^)DSgxuc<&<VkXb;juOfkoTOwG`b<^s
zqwmck4)|*H|2}wAq*OZYf3{UT%<n<YD*u^D?vTyelv*&R7INeY?MVqacM7jl+?u~2
zIU`v0G+|}-#}O8)B^E>FJAB-LM9o6@%5j@HZ<kD5K5xFbKN8=4(XjjDiz|#U0XH!E
zq78S1u|_6sUJMI5oplB-paFWRQ!#xGg>D?S%N93r>Z{Y8mP=TqgIdr8KvTtNrsX_%
z^;9!7PNMSeO{pTJ7vmSu@01k`wiO=%gpk1E@4j&7$GfEfdA+cw@lOgjGl})!Qvrut
z%t{IT$6n9*j(o;l#SS37G0<$N25a(Iz?9kW;CO7~;aOIce#4D%$lOMgi#99#WM@CY
zv$&{{t~I~dKQo<UzJSE=Zc6_Y#<7MT8yJr2j>P<ZppPU$_NlMV2?&&QC*#ay%=7VM
zOyX4T6J|DM^J>lQ1XEVe>v5pP8^)wxuc#WBeHq=3L|6R`@%D|V<K8DM#9r*XrMphy
zK#LVGM?HeV!SOwe{EP_Dd=QkEJ_$WL934}h9x-+h7Ep?$6D1kyurbX)t*N*imEeYt
zxgn?8or2;FPX|0q=X*#+a_AoVj8KWsk#Al4U)IjnHj2<1JytH5sn)b3kD+e7dsEyG
z>9F{QEM}30TTQ8fpF4N1xyGyJ6<r#_=gsYfW{w$#w9DJuP8P~Uof=kmg_Z)!5au@E
zj4%pIaYw(NzN(w%HGbJ7RNj*B(wWEFCE@|wm4rJBf&(^|ur@G4zn!!>PpSX$9Yz}A
zBvKSTB`1BK1!LU&>>tg_{zc(-&5^;WniZ>IP=_*7lb!kdz^}JK`25E5JSdGg?Imho
z_%)lo^mY=v1G$LytObMpmKRYqLWB1>_e%27^ly>pXg`K-Fz8vGji&#9kj3ILJ<beN
zJ`9pwYgU?m;1)6E)zJ{cmGjQSMdI&+q1pa#wTp-<L?Gd+(i-`4$g#qOFR1it%b@}Y
zR88A#w3zt=Dq{OWvv;9=dM)~|bbV9b;nKSrt1g<q50rv=`vKubTnq&>OmH^o8`@MV
zyh|dhd6rpP`d$mV^(y*y47eFA8$szfhi?-#qPD-bL)V$xF!~cCUN+X)U|bX1eq<~N
z$g8=SM5iS$Pc_))FfCFgY6`ar6g<@Tp{oR0>emxjo*=sihV(Ig?^}+RomFDA?6Wj9
z1Og{u=$s_2uu35*yVcZ@4S6XV9b9P9yUWA@Z>U{~$A9}r@j5y{ko_88l$B5a&xmuH
zUKJtiE3x}$OTiPMZ<k{wQAHh=&6%yNvB&7yN2JX2@^O8Re9ZBjaBabVc*UCC{pvL<
z3UVGnFn+jQxYTS3hqg?PEI@Z%{YMG>Pn`?mDEEwujPgWBix&mEb}Z`MkUFg*5~Y5%
zFH8Qt=+W0JyFv}<qD!uN^)}5PJuk^oF2xGV$t<_KT!Me5yd_jAm?0c4JD6LH=ihCj
zgp03@I}>}OIpa(sL7d7Z1icyDl`k8BE?5<pny#EH;})e%Girk+Bj~-|$){~bH8B+1
z@*@f>JQhc22l+orMhD}H9k50%gJzS<0`?{8-d72~HYusTskKcG#d#XJm)%_%nzVAu
z?0?{jhJ;_pK>`6!X^dUZUtQJHSP~jhboI^Mc+9y2F8>`86bVb{;kZv!D%IaRzD<@p
zUt;hoW3f$dYTU73NphpRE5@41as!&DMg(feOMmU_=~mpI9lK@53FTi<F%KLu?_U;F
zQ{e-G<xK#dQveCK9IfaqzR7q|qf;!E2D$oO8)y0jTk(ndO_Y3*PiehMj6#`D7Y?!%
zHRqjjxL);6>fIQ%<g$EEYqvO?9%{+YYa;yD2`ToYXL@RHo6o&mlKXm+KJI<N5oE3&
zLK0Xq$CX^|FTEj*Wh)`y7n%TqYPvb2&Q(I|-7KWttFH^X#}&T5U~ivnenn^EqZL9h
zjrd1NY2Mm}uL0#XX0_T0wM%)`;Q!T7+x&MJ;H+=?$rqtLnHxx);X%iO-LiQXSGT0T
zGnq882C6E$a0996QW!NY5Xji%&R;8GIor)O7nFE#qdmCxU%gb%?8IAfugivh5m+&A
z3c}MtZ8Ol@P4~=`wz(rt3u<YaBC{*bZ33q7*Y|_Siq)no1AAIWczS<eGApUYZpxfR
zR?K}O2bp_n5=e~BOLHJAN1&l}J5lm24?yNxu>DigG0LvL4>lC{?UXk$u*lbY=dzR~
z)10DPI<L#xTqi#m_~$7{%$DlquO-cPK-7BW{Q*Ti{p`r63ayU`92cJceISS1g?&sd
z+zm6XYiJTA-yfE938$&9o{5Y$b=XL*mKX1H$=O4tl}&I*&tI(}3O<s=2zHk={)xbw
zz~391{+(x5a)_`;q#YneXa~G<sn>1}$5~G5lW%)I+Qp?qsrpu)_gPnA2m81lgHf=P
zZ~h?DskRBV6i)iRd9xUN;w|rnJox;8ZJcCjTc7@Ze6sgf%h?)<ex%}n#pWI3KpH#T
zxhH*0LQiG2-~}nWKeVxE$eYqH(Zrs|F~+~9*`dzS8}~1)mQ*x{#nBcyHQyi<nwZ}{
z-Ydm|2@tt!K9&!$vmv`$s-t0~yo+f~h|56R81-oCcXDS8%p9{Z^_I;_FSj~KzOV4l
zc?fPQ?TgqX<o_OD$)BUOmjbdkE%@|_a6|5hsyKiwtd4;8{%epzC`an8tSaV~Ul1U9
z9Vld;JiB@!fb1t9OjxS#Dp3r>r$2xQb`Y8Y&so1!fz)m~Pj>DxdO308!i!T4w0uZO
z1T$9DzY%5zt+GGhDslrN?S`J$jV}I-<?hp@Z`7Au4JWm_%L#9;ZJ*VZs5)^UXyI(`
zB1DM~TT9p*NRO8eoGz)w_u8tV>Df$Kal-aayJ$%338i(XhdIV$@-&^LW&<+R_S3{y
zcH%R>o^$Pcj0q?J_?7p>E2-ldH!)Lh7-Ya7-adBp!cN8_MXP+a6mW;gV&>FIKQgZ+
zL2Z?;kI5a;N6!0W-?}Y;t3Y|628%%&!KXi0mky%79GygBIBg2LrQUTMn<0CSgIIq?
zJw{d4G>UcrPbncys|Jqw($aq)kjHySxHh&fwKi&>VhZqs*j8t+^l_CmjQf@$$SAxb
z74k%E^;1c|Q^Kd>;UDE{*nZn;K^%SYbqs_m+7O)ra^u_%^DFYF4e=@_Ck@zS#W7Ln
z%BbO+G(13F)40Q6OvM`5-0+|aS>3+GlVM)w=4NJ}<*-b?U8<_Fx`mvlmTuEHEiJva
z*wy7cMJIOJZ7A5DCy+-kXre{GeV*7(CaFPtlh+$$*8m}DmB&+S&KFe3k~QneNs2B5
zHtUIT@O${Ap~8`Mk8buP+%3h#&1-pf_V+ex>9&;%SZ|t9qYq)$(HMeXw*+n0=m1uN
z8qB@wzqNJauTH96n1C|I=sP{_<Zg17GYcQ``iw_$r|xV9ofCq((;V94HfJQr%f|y>
zX58v^RoIn$S5?4S%}a-NLd8}67S-)6m*hhid2tBF>5DUf?|^kd(p_5Q?%_xFes7~7
z5EPdlipv6Y!3h!@{A<Q)`Xu3_JT&1^#@`Ky*WcCXK}da9`ATr!6jAdV(l2!6(WSP_
zT`v#z5RSTxejKyaBC>hs*L&w9*VL)a<(+f8tmUOx<pYJIZDt`UUT_iY!{jUtC_7#Q
z2QQj1ijG<dcsx(n^hDk!7T`JyP~|E*yU8&-Gu{S$yi3cF%EsQlPB+P&P~(H+?JTeN
zHKStj8@5yeckmYHBp1>Xns~z$<y#M$@G0$)w@dAao9B)I4~l71%!cm~AaW0sE3J$B
z>IE4KPC#lAj8zyLqY6f%B-e?mW{?SypSW>Jw6@e`yk&E^?NV)bzE$>3fG2eYy2UW*
z@1zzfECAc*@$xq&q9Lbm)H5M7tLbky=gZJ2%y!l6I|wEhSbHtdaKPDpaf)-o_<g^H
z5t6r>@A}bMA$)`!YC9vHL&r$2?57rxCYRA`5T3i|XcK=SENs;$HmGyw>@)68z$H%P
ztef635bJdv6Az``jz;LFLsYTjqyd09{u8~r4o5Yj7@UxW=m?0=ahqrPXn;POU&;#x
zSE!8-kAvOkqZYM(20Uv|XoOkffJ+s$y{2Kw&4$oe`q$kS;p2K6{Lk6LJg%6{Ffn*_
zB)GPNGT}v`8m7+yQMgb^kCs!aqVjrkbxSVgEMzO-(b&Btv4mq*UGcKS9*%c!eysq%
zy8L}G%`ki#PhYj3J=D7H8|5@L#_dUjp6mXPxfuS)#P~+$#R9{K82jO#^@xU>hP1+0
zB+8X@`?JA0Sd^%zsh=6Fz-Ar%r{08`;y+zWSN5Tn{Q)bETp{gJ=;|!S$CU}K`JP%6
z0Ry+J!)kS#_<}!`Gi4cLm@2-x8@0io45mjWx2U#@rBVHnJ9ASg%a$4fPEc+J2$sAO
zJ`OHer|ztFPq_~5DBB9W)EeEBxxhgeb87sB3kB6W_}X|xBezJF`UO`|f;1AFr)<gs
z9kRZk<hLW@IiA>GU^f2=GF*9OvAt|XuGH`|&Nw*1$J%)0<P4vi(D0}yz&M1UTIqBF
zol?n-k<UfdKljW!HI0o!%yFi$`vFeqkI|Q#YHN|fO>F7E$ysM6dG14kRX%@6l?XKY
zGY+6v4@Rv|j7^H6%JK~37x|A$t<1;pSeKy74&E2=rwWu}1n?~v?~{qH%L1HJ{pHz2
z%u92tZsLPJH86LIA#(#l^UyX#c<EDL?(#45>x^6zu9JBzk;(V^&;E;*Ey^|L;O#ku
zPR><EeK760l2QBRooqQk(Z4iiu1+z;J6FZ<T%(j$r!bJw3dQzuYOJTE)q+)-Rrt(F
zMMwVvk{TUG?6S&AMQ<{rzXH*h&StT9RzBR|YY;R&k?A*h{z^wVuh0PN@wot({&{>#
zOzQG3D+EBO#{}D3JNPJUc3!>gJ&Mu(cU!|+Tiw$wUNqlIz`NdS|Es-H8{<sZ#aa?~
zs!aHP^L}%+{vf*A%=@Cd$Mb+h%0!8`J6uQcG<gIs46LFVv?k#@m!D82Au_};UAqSd
z1NfvcTUR*DQdp}3QA4m`smkH@X86*_%1@nRp4|L~EoMWGY=6fBifEa`xQXGTMmza;
z982Nym@LF#h*OqN@(iG`q=U?;G5W}IJ{^=HX?(on-jEWywD{An?6Hm7ht*TAi}oi!
z`B&rYAowmrM{NV|%g%um9}N=2r9Sp#)4YX<>Hb!8+>w4<XfC|T`^aWQn(3D;t|_Re
zN5<vKFj-=$n}~6SwvTDKj0sr-C!>~XFL1-9x=RpAtL>}eSNej1Y^>z9xDz$G$g^XN
zaU~vKXUD}llBt1-$t9TBc_(LjpHJ)Z?G}dVs=L$u$Hst8(4;A-L(t;Z!>8p5!;155
zwr%CjJNwl#{=pklP3^CLA5hISotL=P1{+|jv@KTr+%H3a^8TQHhijm|Lym`mWSPG!
zmrY_XL`heuT*mt=CtB*~^Wu;Shc(;t=HBpzt8UU6ZL*&*B`9u36v7Oo$8mIDyj}d=
z!EqeEdms80A5{Ci8Qg5L)gmdF_aX6Ddf@M}sF2LxMqfIyL3bq^QiQ(19XDicw*=tP
zZu9#2S$-Luc#F`hX1_&)w75`P_0&jhD}^mWTKBt~tj(GOkBw}+H|UywH;qSJGHfSU
zrd*~R*MPoLPBBPW8tcnW@v;xxk72E(ff=~Uz_X+4i6;Kgnunk%5jl8irkGmquaYkU
z^U-u~6pz;5Wo6i{)@_*Q<<6V60Qru9I0Z}SR*LPQfzabfBSW@dRMwW=W`pD4eyIOt
zr_G1|+W7f3y2nqhpvsO0rGjFp7qoONG?Ofu=$yUS71r{_?7++MIrgS_s^WENX51-L
zma}ldB(kLEP2oW4`2EMxfD1rZlVciHKd$_MkWhpsBD>0pFL1J}jMKXC*t*GQ`B@b+
zvI3Z>N~PqO6EC;Sh()NQ?gFq9kGlBcSL21bXS<nXIb8T>;{iWbj#rXtd}yPqYp@R+
z9XRz4lk~aX4HvvptcBrHNszhVg6P>?)gHL9HpdYJNT2b4laj?*Fd_BT=qEtAa^^EV
zI0;0QQ}TIs#_oA*nyL(%8*q|{b34ItmN1Iud-kdVa8RBBGALc($GPTbUUp71%;c!D
z<jx!a++K~2rYOP&a7Q|+x&8j325W|7eoj(##<k7Tkzv0n!ZPOdtrsaLVw-Ola|%{&
zGdXp!(6TV+U)d5=Dy3DCFkM-rGEF^|$`6g=*t;+~#>!#MoNSO58n0BHxNn3PtvDei
zDE1p!<rwu~7o#v%Jdf2p;vlbP_mWV8k&46%3Ui=&we7toT`^ShYthtTA@;IMnYL3r
zAt1}O*Lh5WyYSL6vCz4CQ`W?*+l<GT_WW>e%zpasgG!HuXGVV?-0n$Zu``$$@1f@4
zgRdtlsJ{<D5`F-)q2PVCm<`Ep{*AQ>e-wW#h01OBlm*;R|I+q8<L`qIoxQdD!L`I7
z&>drm66EiL`ku)iUX1pnpGMu^2TQo=(@x8#qu|Qsa{Kn%%rTYoSD<T;-EN#4uHM}l
zDsa-S0f(V6xb3ubU1{O(1L&{)V1nBH<?4*YVZ@|shwCLS<cWrVAkZhD!@zB{bn0Yd
zC&9C8)3C|8Q{A#{Qc$%dRr!Vsw|4T$r=2yCBdUAjX}8_eL320Sml=6|Bsc2!@t0TX
zhLxOZ_X)GsdRa@rjS=M>Oj&<k?onPX<KqTT9%;n|B^4KQv5-2tmyhKOLKHA4^;k8Z
z6#8p{l$X^!i`5ZJ5FBzzkG-><ymqZexBXf&Nz415VVC5)d2KD*@iy3fs+oiS58obQ
zMaPQ^zzi8V-Jvgp1KCvK81~&$;)toQXu^Y^1)Vhheb7vh(UrZo_${nB7$@o4FwKRT
zmR{eH^XiFpjgDe-`M*RF6tX22f@S*$T;P^BuoSD?%s%Bs4_Ym?2?TE0U`hX8JJRop
zEP*dSGl6RsltM^Mkjvy~+tDGokXqv5^I-ydl(F5vcE0{gogQKy?L$2-VKMQA#|QK3
z(%n5|=@WdZ>5SL{g^&_2yFT_C2MKdp7WTe#>=8+ctP`!4%0RZtZOoZ7-SbSa#-A6q
z>K|IR4=Z#Ga+c+y#NO63V<+UTPxYAf8|UrI(Fn`gUP`{WZ!Gps^E&<*-!@oY%t_L$
z)m_8Su!o=ziR(+*i6%F$Z77etgwS0hTQ=RsljwO(+o5kb$H5@dmoHKxUQ2JUCb=Pv
zx<&zhyRGZzzb(E4$#D}d-tNd24UQ64KEaJ@5>8PqcYl(1tH>LtOV6KkS2QY?GyKE*
z#_?ZBHSYuE;`DQGXM^(8N{0@280SlnOTnO`5ofu>G&2%dP<x4@OE3+EW5>%)(_HFm
zmoo;xudFZR^3kG;ue(-BeWSw7V;(V8Vbv;Hz|BsO^f8%n+cmaU@uYw^UjKZ0TT|=H
zXAY~HvAA_{T_FnwyPQ^jc0fyKg9Pb0XY`o7U-{Hm!o17eInosi44lTajW>}bG-3?s
z$2dXevT|Y}`<6Eb*GM1CA$R9#xyd>cc0Ypc+JKWJ&BOsDK!DdDTFx*|W6|^`KiN)f
z_Jnc{&hUIt_0-?^D170*DFPqh_Pp7>Mk4^RtLp`-|H9_GI2c_fwW(%3)YINY{MBO&
zQykF$eI`JD`J2IwhmtlJRZ&fN)fOxw;hCykyH5Q6!_)RA?juyW)}xPVT67;}_dL$Q
z>V<lh?`{tWl`E#O6NWac_OKlj{WO=t%Yp;9LtW#Qv`M5-H^&Sg$-spV4AO&Zu8_Za
zMu;&P{1H~kH#tdH2z?eE55H$cnAJY+$kH3#q1ak}TrJi85&c0{A1ewR5sEvhfucOy
z=FE*)4bc!Elv;(sxl96x8&{LC5_N}*DUM`M;B9n0DU;^2Mv0^G8cRAq|0n+Kz{_~#
z|Hqeo4t<mA_aTp7r#>y9f2^=$W$PmtQlD2#VZIlq;#{7w4i(q#P4#uA<8HxFsc6bj
zbPlM_)Cvm{+p{$(+KS*r^_EpPdJ_!}E_n~fkNaE`?U!j!#`~;CY@WiO5?Ov9ZtXg>
z{Jl3XlZ=8@fWsvFIpdf07szC`%+Bv~(|C{5jkK?+pZB`aCC>0~JTJd{xrZUxo5Bm0
zmn}B7pi`z3K64eL_baS2sG8^YvN#pRp%&}f?)fr*fGW>?$Wq^g3ACFivPm-JRN5Zg
zp?|xuR*raCjQtO_bUt?!s=N3Pwe-N>*|+SQwI8G%W-*M_JX|!f{P{ulho&IcRHu|k
ziL7Gt-Khahm0_p4l6IiUls6aeKxmYX4@z12AS!x<#GG18bFMScY&IyVo;>~Zm~~c^
z2SbP^=gy0e1>QCQXZ7eh#npV-uuWCPD&FK^p}HD)9PtSPbk^W|ffvOZ!AUHl{>AOs
z;7U7$f4N_^9!5uWYV;p**+xVjUg&eP3sWo})X6foMGn#`4bjT;IATtsDYTSN+U|P$
zRSiaE+b>j}k67&LFSuxYQ0?e9r{PMz{@&JFW7B*ib*tP}<Zbiy_h7>BsdK_Vub5TS
ziu8g7`bbE;R<}9euKa#kZWb?XlpF3VqFzu-UcbM9JIH79bpZ>YIzetHX;hgh`gE~a
z#@-`qUer>(U2e6<KM1qWLL(uIM$W&$B<nS=mY(Qha&YtI)e-3}9mF&5tNd4+IO_I9
z61>Y9BZspj``^@W>gm&kc3)yc=WMHI(KT-eE&|_d@&Mzj?hmea1p5I2TE-YkZeI3l
zT&ke7@)wjHikSPm2DZOndyse_vTl7JW^rI1n5AAeri)I!(JXK}PF)grqH>3$N-TN}
zN#67xG1wQcRSX_T0{ITA6Z<AvucWXt^*HkJn6##0h#`<N&ENdX-N>;$t6YROCg@Mi
zRBILeZXdAhrl=RAT+BSc%!!E$mk^D^NrHA!hn!cLSoL2kyy10W47UvTx?zr_0b=Ty
zPHsVZmp@243DXSb6T#$0d}d95D3lM$rC_ZX1wFUX1?j`n@bNzS<=$SXJDer>R%mNZ
zj)OsXLJS^aRHqyJ0NnaHI1wSM5oV$zAlUev2{?|vud!N^IyB?1uaP$!zg+orL7&N?
z&G3kimp*UPlXzM2wty6O+&>jrk~qE@X(DdXD1`5i?=aUGe#Td6xb5vQtUx$);PG(K
zUmN7cNdp%9jHIo4k2ZT6t-D4CO(46b2S;sc#(H|m4etRRRBAp#b)W^4JDd|<Zv^y!
ztr13<J!adEat$Zo?k2#S^bc<_uAP5L9}v_AtT1HvrYv|^CgoYqi*E2EkDoSaUeLU`
z5ur|NjU(ywa}Ji{T#J@%x)Um3*1}=hrxr&#Dd|ZWM*^v18(`u@6aJS?o`%Ty|EV)l
zE=A$k*HAgLBjLlLIC3eSN!X@hfYJWEaa1kQD8@v3fu+4V#gl3!M>Rqwg)jOBxZ%Y)
zHM_bZQhGv`MD>|=#mjqy&s#$pew;<zWQhL5Ao)vcFe5KSme8o>D}Vr(cn$_n7nRtf
zDk>l-aOn&gUzWLO{%(M1H%XzyO>Ov|i(SKh8wB2PY{s|jCSc;?BknKBgUw&*aC#Dd
ze_(pCuQZ@>3f2Di*3#QC?b*d!3!xEH-kfotar<`rsD2upRbF}?h`$gNJn$3=#`!Jk
zIt2e*YlEZbn%cUDo{#s%95?=%v17vp#+iBwvD=Z|Bp9|)AMintLuK1d_+MGsSa%A8
zK*%wbFhN4oq9L)S5Tbf2X#ReNmi>Rqmf?x!H;d;|8g;b+Z`)p2c8$Ml;xl9FIrc|K
z$*ySB#EFIgPpQ<Mwg$5{gx5};Uf9jr9Sz~blE3JDZ*}gG4HKk$OsdyGI3a*?PqQq;
zjsxO&PtJY!$f#s}zQtP)d_5ta#u__ba^u?ew96>zE@^hWVpegT%ms`u+9c}jO*}l<
zM)=8Z>wSqa$el{Ze_9sYlu(}q$(4+m>~9Cs-*c2avx6srSGtkp?yRQW2iA&I^F0CS
zyaQV896@M#tx7t5&U^Ii#?_X=u$3P3d2GrAHD+!#!0svyt7g;55dB-}%gw`UHlaSy
z3b!#vidT?d{MwMQbV84d{y+ehS^3BT#m)>SN~cu}qX6y?6@!f6q!Uj+Mr)}T08?)_
zS1VIy)1JE{z7O?bqBtYH1pJj*8i9vLj4kHv69~6iw11Z8;L4jOLF}`7hJQ|GE3sn(
zk}_>JDEbXeCDGbr)}POjfeI^22ML@+Y(5=hdn{bSTgY?WrUqWZQlF>Ph$fG?Pt7uY
z`RMD**ERa5Png$!@yuuJkdB0Fj``1bSj%}k(iPJn{{t;Y?aa$fmtoc!+-9Xyv+Vvl
z-j(WuDOrb>O9O{C^|<tH9BZ+JR5>_TXL#l@j983FWpD9a)ofTiHN|*C-_=Z8!r~8m
zY8H;8<JeZSxM-1ad=XR?-VmTPfKz0TJ4d8)5>D$bsHnKJ-o++~`gd-sIVkt8Xpi)e
zvxK^foPfIAzzRMa7r{}7CT>i7``YG5T26&hfbYmihI(&Srt7npA|uM}T!Y>n#SNs!
zL3E+sHle>iY&nO~n>x+Sd~Y{fEw|4TCMJE_mJmhoQnzkhxXD|Tdu<7-q`Y4|6TMTT
z8DXGf_5XT~@Bel`A|lcM^O6^@ZMC5tX0o}1U%qfLc{W!69wv*!UBc4DyxAX$%M?%k
z=0n}_a%P^*ag}?wlk~s$xEB=DbEhn%B$au;@LsYuO#RD2^F$SEeRiw-p!0v-UT~zz
z(%#}9H^#o3CoNI4RiftC*25JVeNe-u%yQ|^X1MJ5@O=@}T&|dxai2{ADvKearE;E1
z61rpMN=X{fSq5>w;iIRGgfMX8UHU1jdS&RZ;f1<m#t7cWB)|6O)ynj_qBx1nd8N7O
z4VkUyxb+e@G~cCFG?UJgeg#mIZp9vogbrIgNf7Z@9-0+k-6x3Qlw{d$moMj9bfLFw
zQdCVW5@O=E8SVEnzDfj6qo9NBw8q*9BkQ-#oNm(A(Xj$?&FrKpp;6xqg2~bKpzIWH
z?Suh>L=+Ka$%ga@2ixb-c%yHeD7?fQ;)MQ=qVCn8k<aX?7$)rNb+SPhfksqK*^7Xx
z@(ZkYo;SAce{lEKVNrekqWDmvq_m)fq=10Z(vrf^-6+xx(yfSe$Iv4=bi>e~fW)BG
z0MacmNcRxlt>5=K&-Xp|+<VTw&-p#S`@HLq*=w!6_FA8wy;pobQ6_s)Pi>+^>J?O&
z)wWM2S;5=H`w9i}pOxOYmuOy9P*2-!>yR=&Z-iOIYZjDD+XW_v?wMqF$|V%(oVqzw
zBW}LoJ1vVzLJWHz0MXOFgay4OTo|(AZ18ng#N`!lcOGNp+SJXij40B%L*74TjkKG%
zwK7mOiMm-IH7Yzln+cgYQ=~)c7@C{EAiZaQYk9Ng^{XKgS%3ZHE%xG(L~O%sB~M;n
zajVMB+qEM<@;;}gB&NFSYfs<d@q@ZA)>c!-Y^L2Zm27I^KT&n9u-MVHrTISks$pj5
zEhd?c0T*}NwbO?w31+=~6S~Ky-|%DxicxQWGW<gr{^JW)mx*pev;Qm=1;l%l9nMwg
zIm(eb@zrV4yg?+%<fHDgq)OQshYM6GJq0r#6MVp;;K~}N?_r~1^ZMB4#J7mPOpCK#
z@uF16Dn^oBrvSP3TEf%~su7<SCcq-fNk@~vGz3<b4x=HEFMXD9Z;G}1ot2@fo2gg<
z^4=TLI(2jKkW?hn|IrB3(^tKrBvDD}B(Z?^!T}8(LR-`V@r}SsogMeNSy9hW^Dl-w
z_LT}Ti0;U;DW&{}V;+6?A_P%D>{mV^TEVO^Jgx_oq8#}KZz*b(Th?{KRZ3<Piab40
z7+XYdUld%bT=FmNaOiCAU)~eWQ%Qaw`Gb?@d%{OePT<sba&rkx;I$eTcRLr;Q|hx2
zwdD8xE=E)5EhJ^5FjlK4tj|i8FKi+964H!AoElE0G44YJJ>+pjG~DHM3mOOd$ZL(f
zXF7=8G_Iu^?)GT+2c9$cGnHk&HGFk=5yl#4Wo&w`W!Hc9tY;uZwRsz2IsRsjV*6D@
z&{ter&9nw~(&Ry=3Gwm5K`1!kT=?jK$x##w>F)D$I5&&$qn(N&xwlrsjP-h%e%PLh
zJ9|W_awtuLV$pc%C_|8LN_u}v#P>&MvdL$PO}s}gzd)5qxpWI<x9VbD#csIMAa6|Y
zG+)CNtv&5xEuMvs^YJ}h1?P9_wR9`msMZRy*%A(a*onXI*qAA!@{q88hIH1kF;T{`
z@f~ySZRr<@?;kQNDrw3p{&rFE@Nnzfo4JklZ+;8wKnxPSXW!iBzr9ICZ9LQYlbm23
zqa=UQ<M4DwMh97y{?OMf{|l)zc;L4m^i+Y5I;{u1O@1`)C!TuQM^#y_TeyVuwVX17
zI?A<Hw#4*`3^ZmoIX4SBR?drXsmgIAxj)a<CZkSt7`M$WAj8_e6&pk<#*~*$zMLig
zUOZ>>?s1{aC+%o@p;&X3(KQxSO>w=Cp#|QF4su8wuUXH3$o7R+QgV7^bZu`$g)btf
zJ$I^f^1YpY*xli8$eQGp5mr=^fUtXV7$5cFPDVxq?V;Niorfh$LQMFkOOXK%y2SZJ
zAo0tWCd&_vf<&)QGDE#>ox1AF?_(|$%}i!Dt0)b!a6~D_hL188_ND%e;3;_|RHT}A
zbW}=48|c18%K9Q;L&9vOO(*>67naJ`1Ug!2<tt)}TA5{&c1cPTG0!BQQH;Kp)vYNH
z#iQYDU;uxzc;;h65XQrUM}Av<8NSGm_Rr>!*UVH49xz~kba%b5aCuRe{E>=$16)V%
z*|f=?Z}S0NrQh5}GqCdFeQ_Q9A2Z<puUgj$^a?O<cXJE^sp+=Mbt8_?uFRbqy^Jqg
zmTr@SKp0~D5(~L}EiMZf9to744BYEF@S(ep*g88`$sjzzG*F4^T8BBjh9z0woR;kw
zL$~qjOL@nPQUfk~7o(j{I;nzV)YLbA$f7gI-=6%4@a_3S#waSPr>gzkB|@Z@93OgT
z%|hfoi^F9VxLgvrdhTz<wVCk7%d4P_zDS3BT~2+n9LT3pXe_J~Xn1x_N^r*hh9kAY
zl*^plF6sq1_Os7<nW0hq2V_b~;byH;`0+?T#_=Dw1OI&l^(VVu#~y0n=Q(0c#?ioG
z59vPXKSkRfM|hasbDNLAbJT(zXhWk|QaMOe@VulwRGsp*6hlwn;P!u?xsz&(Q_5XB
z=$F-4Y9skP{{rN3^h^#~UJt#VoWMsoYXgz{JTWgl*C&k?(Ko!~XY-m@IVCqX7`Q+i
zs_QB3PBxTTCWf;~X#Vhx0pe!c4%Sz?P~u<Az?FL1=zsf?f!g|Vy9}B*S3-Z&MX0mX
zA;(sa|9Jfysyt8y*%^#En5k8H9o+cVGV8)bCvIR6vM<HR@ssUmux0*UQO(6Ojhow4
z8gt@WD2e`Ch7xFd`lJ;4rO-1H!xA_Uz3g|Z(weoiX3ZO-x67-O7W5WaT=W3~Ku2-=
zQjOZ+U5!&I-qEuO{X@UGk|12E{ex{@;#!Q<xQ^0Kzd+4{kiqzUxN&57S#>}>L4CmS
zq$C@M<R=Wv;nexRdjAIucmIAHGhfG(0l>`Wa;rR0RXgY%&V(27TY3xJ-$O$OS|IzN
zuVK>|2!%xz>qDo)D-aG_Juw#!Fl~5JcW9);vlNfrvTM3?17W{pXtv&N-nQkQFcM(o
zQQhmI!{0jLN)gO8aKu79-RCT=4ch(P0i`Man%dub#Ne5LRCMcT@Hj;L9gauxr9aPM
zNa$Y(cCM|1&oRvs(E-whwHm;x#d`i;W!1$q%^Q%3HxCT?{tr$ky2RJtv^%t;mYtd@
zXv%8vIy$fFua75*2OI(Zu)Tvp2IL^`E=Ybq4})}MeCs^_&6L6kFN{A`dwF#iLYg<X
z3H4mg@A^scBF%uKhmtICpbb#P^fP!JwwD9H5CY2HW|bDT`|v<B5`ZaF6K+$$7}|j^
z0S3Hdtn2}Z@yrp`WZyTO(x8Dl@_+fb-3_QSaK7$nQX6nUO1z{gOEpqbvtcy`VrLq|
z)*4VC0X+b+)8}r>mv7z#g0&@FDRcx`5T&~~k{0yb^l&PLwFa3D{3`4tv)oK0n@ffZ
zS5(>Dji2d9(_BCFv6Q0YEd<2G^atR^!H?{d56cCnM8AY_r{k{kK*SOtyBBWP@aA72
z78D3~`ZgkNr+xT>M;#>5FK-M4XpB4)BJ}tLx=)4<Cclm^Z@QT^{sl@RyK%w3j?Zce
zoSj=tJOKfDf8fp>a3cQ%gd<mfpIh7`Dy?&_?da#}5s2;QyY=qVA3u&}{@f(y2KGxi
z?MnXwy`dn)HrfTB&#v78DByD4*7OU+$|3nAeBF4oqu7WC=TI}SpVA1uJ$1*&F(Y`2
zsQhsWWF#Q|apTDx#o^(s0DT%xiV(ZrG$ulCk;}a`cBa_=05@mUa(A!RUlAa*?d6Y(
z8x=rwS&kYYoi1=axpU%LtW;=+-QwB=_?$87_YhT8zd&4)0Th5p1KQLQyNrVV5hK{Y
z@m#^>M6KkXbk0sRtTaGj8uVNEUmv>VU)`mtT>8821|z+Qwb5~ZL^PvsZC&A<L4dY<
zLL&bk%KS&=33Udz;8vgh=<c}wRTgl19Q#&&lP?VUcRj=?-lZ!!`A>#tY|s*rGLmu|
z2|Cj<xXLUE!p4DE-T3$`q5-!AW5&A~!+&5f(NXCnM<>&bhyi8|sT&7KX6HL8OYT-k
z<w{&FeyZEXo2lEas?#MlrdrHl@6OgwI&&RLnw6q#SPj%tru2XJ48jD&m;<2hZ^Ll&
z+q^NBvMzLlPIwWuHYS})xs9IvMy7=7OTvrrvbr@jU`+|IL@j?Q>K~m9+5YvLbS7qp
zZh+di!vQrdh792Tp{EW&PoT0xuNI=??Jtz@56ku!bx52bG?mNye<y|<%^!~aheUkv
z8_UTcR+Pd4^R2QZx2r(xfcOW$4Iofw)`e-{&65~1g9D)L{<|iq3lr7)F8MlMx%T>G
z5+C7+M661YB0d4=(;0~DqH=v4XVQaqFM~Na9Hnb#FjD!J_Xcw};J|+aCPiux?~s2L
zO~eHk8d7%h1jFDGp+Q;bp_6QeG-?1=UB5HXdUUPw*svPL%Uion4XYwd3w$pm?%3f>
zn9{np8>kiN193|abd$6zUwy?yx3at2jW=Fhs5_<;7Pe%n=OQ0ivIq3A_ptJB4hm@+
zWg_?kYL(y1b-dEhV(CAn7Vgur7|Qr`rJID1ahjKYafM2hj^E!~P5Cleuz+p8I-jAj
z!Zh$nvppJPH6_mLZ9tJ}O@P1e77`g0@EF75;P4V~gocRasv7L2eIOc0-l-1KKM1$}
z1sZ@;)bHusZdc$jWx%_s{2q4AbzW?CO&W3teu5(xoA%!Sa6MA)7ijcRP$7srRoH9;
zcxI6tf3SsZOA0?U1`hZqKJ@KbIjoF8t78UMK+cZaSDbe@p#{Ml=!Qk;*kwd#OWsW)
zD8O)Ucq^c1ab6=x2_#YBwQ<}5v@7l`1kvlZ9-+-#h0e6i#)65<OfIf@nldj-dA6i*
zCn~RQAag3$6I90^8O=lso2={q-VbKNi9;t~zBqrd+0M4hp10XxD6Tr58w2j9Ha7BZ
zl0o^X<-lJc@Rvtu8(<)IF}Kdn!ddcPplbvGtVE^PFVJzzO8Uhn^l=_Y(|aoj@(aW!
zrY-q!R8k3~FfsWHbOU*;0|bj@+j@!nS5Zudrw7XX5j(&Huq#_`z0PuE#~*QU<L-q_
zq+aPtiW5XlJp!RySJ-roYP7zs-|k$Oy7Snh=k+hp+I*y`*Cx&Zf~aMC?GC!dT<92X
z9=8`2bX<qYG3$Zew#&6SSS118Rg!c$e4rocuZr=S*+>sw7coM`;42Y9$G*5*FYln6
zSJ;l5%;QecxLa0qFB`pg!T}c+190-;{u8X@YU}{hr&MU8r&^be*f;Di@zm&C@0W4u
zmj+!^zB4=gh{G1AA8U4%@l^jRPF)g=#CX2C%S`pS&ivB_?MHu8ewK0H*>-huI?0vi
zL5Ht<tN4<M$KGxAij5FgVv$M~CQ)`ass}0I!^rnq0-n<w-(m)+OUd;M<nS>tl}eTU
z$8Ge|Q_i?+0{w;!-$kBSbrWG&8oe<QOAwfd%Vh)I-j03=R~7ZdlxTXaghXt=?Cqcp
z5NHb`AC3mY$cJAtZHM$gw_f#D?EtjjGJhx0q^x+8`oQRBR_gi@hNLmw@9~JZf}G&M
z6{I$W6Ja*i9ThZN@e4H8)8eZ3<afWo1BT=9BKYy~`FFSqWH^{pRdPkM^Wy1BpU?5>
z+14v}m9m#QCjmW408<c=H(ooy%XvyLbIsoGptVWd92aY|y}v*lAYJ{F2ooRcUg_O~
zOuu7*=cWg2XoCa2fmKQXIz!(sk`q4wT&rH^ol~IVC0<9Be!?}^y^8COw`hILz$)P+
zpj2#j{<ZT^=?NC<AVP9%JTj=mhQ1KQ@z|>;WFxg)ljq86@%oX-cU9_uaCDow<Z**J
z?DR6|xcbpQm2f2W$`x)>vvhfGVx%ruf^q-iqpSW^><A#Nn%)=T>wp+QDT({n|A>K^
zD?Qqv#`@9{DB;M(dB+9;TsLm39FYVAh#iFg*5fh?{6RoyKH`k^Da%RTsGR4S>TLqR
zaZ^4se+cNcThHNP=W~Ga2CoAj(u=jMxCR~Dkz8p8Q{mr#3It!i5L<cWW0Dt`2r5HO
z1qFc@Sc4pL0kwmc>0AR!47sk2K)(Qtg5I8~nJeF!yV_Vd2;E@TZV=EjJP+s?K$D5U
zAn++`;4WZ2FiwD(?!rqlNiW&o%+}2Wtil1$nY-DT%;$MGEYvvA^oHeTHwyg%<UJnn
zTUW2Zl8nWR1!2LDbOje}0d`(SY+O4xp-#N}G09@|-&e9gvJ{!2<QawPn)%%Mnnb#X
zfYZpH=6jeM6M(PUZNw?lMS*(UO3EZ)pPcu$0(+o4^=e}>pZw6nBa{;{kPA1&0xNyO
z4BHk1zIDv^cgD4Q5ir0S7F!++fDFaldU%0B*CTP)d_w_?GFG#C7dYHB&`B|jUUZoN
z3=>FSEU1S_#IlNV=u*hW-+zNg3a9V~E=gjja8|AP%z&6X%g7yhxuO18GkVsK#Dy<+
zx%zrrRJ!>VdIN_`4dFVbBb6ggtlyQtrzlB}Kw^l#7=8~FEp{d1-?roB)hIaDpuQd1
zdbYdl*vaV?vc2+8AHH1qC+jDhqtJru@K;_y1WGkIBXlX|()ZtAZHZiJ&nX-r*m%G5
zC;KCE$KRITKG&4^-u4Sb63p|v$-iX6zhy?~56%Ez?@2O1)>?T^_psA#kpb5^PRk_D
z?ay2feqbChqPP{NtDGTf!X=57Co@@&-NRkvk-y2o`P|O;ltA1eb>+4=p~;t^K@OAq
zU&&5!tYnkZU3pJ0<VkeC$$fA4e{19Ru9g`BoW5B^e4Smpr{neuM40l-@nCrYgRaP*
zpJ{n*`S4v~0$)o}z{S<uvRjL#Dk6nx(J&#6!g0HbD#Dt84>P6GVdC!}FT#o^KNRsE
zYHl%^eVAskL{{9&Y}UQnTkH-aJmqi%|7@Oy;(kddN(r8hi@<weqs+v6hNnIxPZK-9
zBG`L}Il1w(`l!>vrFLqOZ@S%UryCZ0`_jUl?#jmd7}5W|AMpeVj_R?hWjsW#i%!iE
zjs@uO9Qe-H9knvk=%k6^evClsw2=^1HTjU$K2<AAq{7r}73peMFy!3i*_p3Tj_k*e
z&2|b?haN`~5{@9A3<tzY`?Pff9t9z-g;hYtRu#0Wt{Q*J<c;v)`zcy*T<^{-tQZHq
z9jGgedN&E{j)TrlbXVhh%zAJ=phU#=PNR;GFt<hHOq?Gbu*T(dbY|W0FexqYPLh1H
zwy|^Y_RfJ|;95I2PN!F5z?}p6>u~(aU5SWex<`vOpeZkUSu`=Z^Y>u;*0{U?+(@CH
zQUddRqY@##_fHvUc2_r{&sSpn?h;zN9;<ykkZ%vaRo8Td(QR|x5s$ZSvgVyBH^8#K
zecqwY_)N?Clf--PB*X0%)46ju4FA1DwGW?Glu!M9rA&m`s*v7E;yDZPRLc;Ln*t7e
zR(V7JAYuF}iX-+%To4W0=^mV^$dvOxZBzsNan3gu%1-E*hppiSMu%!<#VsysW+f7=
zQ$bZ5#Mo+*88Mc-2WA;bD>0XML;q`7T;3)0`U{Ke-&n+v+t-;14ZQ@H?&xrn%g049
z4Hw_7>LvMZ3jIiEFB+FeOs??C&nMXanO9Z!l9&c<=*|yJ6Yr|O#=Hxqu&<6A3VIma
zewy$b2uv}JjyxXcr@>6U8fOx<Aa&QIeuSD(Zx4oIuY%7-Kr{v&$07_0*T?wX`p*yM
z%4;UIhACBA=snCT3bnZeAMY!vI|s6!RPZ#=Ul$Bq>RJ?!+6V~t(Ro*j@wnuKj(ckE
zZH47@5GPv~tk>W&M?&if@AF;<bp7FP7&^@{w|OXj^A^M7f8ygmQJ}R6%{JnlAJ~Is
zG8aI$DW*|nj#0iV`3JJU5$Ye!Ona~l<^r+)O)8|tG-}a-+3pzSYYUz9;Ro_JYwc}a
z0+0egN(AH5{an`<H!;e~;1Vrg`F<!w@J;V*VVb=&YSu$N|70z*j0~LX0%)18DCN)+
zr6!E+mS<EYUKsd?-N3?d=K#)>Z|R|qS@8j@pLjb1cNW<cN-d&caZ54_5R;1%7LqC_
z1`})6U!dUM5`F6*P)Z48_ju<6zkN#z5p~~9-(i~g_yAZYiHB}7Q0*S7<w`u7vM$Ci
z|4RWz0I_EOYXmSLh?+`R3ImdK(0jk7lDjBpc<x|#YJ48X@DQO)870lQT9hRm%$>y#
zW#l~qe11uBO+{vTtuVP7CP1+^N<7gWrXa`4)qsor#i`shlMMNpYTrv~(!|Xg-Rclt
z#&d;T!j?R~9bUymk0oD{yWy#_GCZDjDgI(Dh=ha~{OWdW_Cz5~|1s!u(Eaes<eRmQ
zm)N@n&CF3lQnx|@TtNKXpO5ci<a$zl3?}^D9AskxpoJy<ljk{zhU%5?ZO}+7{>cA>
zjZ@`j!vCOv&xU;{^#=_DVYc``Fh9;B>nB8>Qhj_k@JTFqUTqFS=YQLXjW84Rk35(S
z8}X9KmEH|H2N`;FzvMC*eH%2VFb6wh0VZWU0M0)AtB=EZE>Zr64w;V7yF!}AiE=*}
zYPpe3|4$R<|C*UGJpqG3Ze~v~$&+_E-v*SnL{)du+9F!pxvc5Ini5fTJf7%I^QO7u
zW7ielh>WwU!Rd2{_C^O>EBWBRz3*VS4N=~g{1Ju^MOo)=FYdtf=$1ok?bh4HxZwxi
z@mF7hK)A`GZ`oHY`&9k?KZ`OOSJm;{mHG$*-Qm7vxW6LJZPPf+0z#;~7=|FG58rY9
zERKRA;`&MB8hIpp%-`+*ZLEK->?(9S>ZWf62Rj|*KU}3I9G<EY(Q*TN`&@&v-SRNJ
z%n@;jO>JW(73D4mynA96KqbbR#t}!S1kHjja^N$MdJ=@Dq@GQyL&<g*;p&dWw}+)!
zLW>LvUwFDGU=au6ez(uJuR@3hc)_mPgNsevl-Y)^KdD;`JgVFq5&_3^AiD#2E_Pdw
zB?b*iW}_}&LH`9cBfr4bPm5<65N)}G?UBevxIoRcO7IqAmw$Kq8!?0l&TQpQtIYUl
zNy3}ukK9U~+;9i>oKOF~e7ToC6K|pN660ucwTF?*o#!$uk2C!j$Pn}e$wDQy5LSix
z!<abi(9`uWo_aW7XCd9I(!h&*+bHhI@Wj;4&W|OWjjhQ*Q@zJrvVqc=M>StGo1yAt
zEA75)U90IRtCyN%5Tq$I4a4Do2K%mcuKmss26nB)>ghH}`=1)zS<th+3c82sl`^}t
zwgz%8tH;pbsc)xSd_;goUq`+7Jz!wuy0HNjP;d3zy0l(wJ>$w4E(<HiP)e!dW87WC
z26Ojz#{2@AhZs<CGnlq)x2%h^d6i=@1A~Edv0GTG+KQ7c5#^Xgm$Qh;b&K{O^FSxb
zCj=Kfuz^5kn#*=cp)o88_GM_*cUXX_n$tc{40`Q(I+PD2Utl~eg{29E-}onh|E&<R
zBp=Xqn<e4Q&(9P1R{y7j;mWb5@Z$;2F<ltU1}i>fJj}T<XYmo{CRPmF4%PLmVBXm}
z7QgR`;p<g1D_>vqH>|fi%*qEbc+~yinevoZ{HM<Tzb}wi?ND6XbufWEoVmg(?4Mju
zzw5-mzH*M2HEG^EL~Ofwrm6o83i9G(MYK30_u9d}@=e4F(PIyLmT&De5?}d|H?LA|
z27%ZGye7r%PP^65LSBOC@A+e}baxKoL&<lsGa68ghj+tgoptuq6$pu;Ue;NBEe;!f
zU3VXW?(A$-HMWEi(HeFI|NAX9e;fT@BrCk}XG!)vm^5Hv!SsL(YBKkh)zEgqFg(id
z&<GRCmTceI^cz<tcVq*}RPA>xUOqksS#4Q4&^bNUR2v!pSFZlMkC>w`n~z31o>GuO
zJ*JlM`upRg^(0vR37S3ZXJ^-_8;-QsA~Lw=%W`&UmnM#!xnb!K(qX$l=bU*&Y-@CO
zir1}>+b*J<)dcX`>oh%rrW|W$?1Wz+1F4TxgPg#zRNp2(<deUV9BcgG)^)v0?U&XX
z;2p10y3rwuXs&D??%1pOCJ#2sg(9FT6|gIq;ur59d8ObhgBUqe4|#&ySKFLf-fAsF
zRJiiRz*T9cw@vb`n`%9)absJG;#<1aqEAK0qOmN)kx*9f-O}R=d_~wvGZ)kb(m2<A
zf~bJdCXj=0h=f1H!yjm{5Eg<B;}SI!1s)AIg+~uJHGn+sa~SLoe()j*FGbwqooT8s
zXDc7y6n-}Wf7oZ*Xy>6bF{*5yH(M;7$&Sj^Y16SRH94R<mJI3bhVBwA{0+VT!z>o5
zs8&k8k1tP4OU)e#vx5pnz5WMpW18dx@O+PfulR4w*hj<TYmslXE4--~sP~SyPg{zt
zY3-&=Kc8%3K_TGwz&YnHnhl4b?e2=s8cRDAkK_=JGJ1Aut`TG>H*^fr=rFmd%EEM5
z5<*mWk&Joy-UMS6?4DeMZnHq?YR`koI_J01AX0noyeT9oHmP;$7(~nFg7K#zb|RuG
ztQ;f&(XdB>v^suUD{8Nk;9L2;<H4Bm*r<RaP@+L<w&lS^7lreg3rIwD&uj$bo;@et
z-W<&4-+FB(j47hJOSdrBg=O~2H_*8N<NgLxZf6sudU5I_-WiV}x<09JaMg8tOB8<=
z35o%nXSLRYw86;peIBQm%Hz<m?jJeEcwLN(kwy`Q1ScbHOB6ysqIxC#wS3lI%XmtM
zCnk-Xe)E6P5f?6w=NoVkA*>QQl=(d=LRTPK_MPd@_?t0X?sRG>X&i?V{dHsn%jgvo
z_XC+qd=#-p?bAyh;lf8H=dLknue_1kR>@Mi3)%)++5B{ota;~SbS@{njB`rS=UZLD
znd7~_N9XkFfgG9cBNo08JyLOz<`ZgHCT`0fGKr42Zi@)3n`aK^4hP5ftz<4`#H_L~
z^IJ^Qf|k<m+QSieKblICZQU3url^%0^S{QAmqo+Uee%4Mnw<FSw@bBUk65nC%N!<j
z=guAA<YJce_xJJIS)Od3PlbB0+g<7(@MQY*=~T8j<@dc98m{hOFon}}PR<Tj=S4gB
zSvd&lLkJI9__&$AD#oJJWSE>}@Lf|<!>YOrPCSFR3&fR<TB|P!R<*2F<o4{ok8%=s
zQl>C&SA5FjQ%|C3oKNf#ViOrN^AvUFU+s4onw1F`ux&F8Tda~9lW2;rgFNn9T86wb
zJKa4CEL_bx_y2)C?XkF+=n+IXO&a}a{hNGAMBF{OhrT4c0)exGW}h{f>v?Fott@1X
zPf*0^z5oEImCD)FURp20ZnJ##O_c}kl{LD>y|8i366*r(tZ8n+FSW{uk6hO#;$8xn
z-X~LDyzCx^mTu{!4Kl@_%E#E?<IhXR)@wwt8{sjx)@(IWw|N+n<&(7RHAut9tV^hG
zM+IfLn6?d;bbs>qeo)A1iqqBJ90|+};>Z?KnZd#_y@p$-jsTHFBi1gui2fnvUR9gF
zMOObWVgE}ja20>6!{1w|UI0O+gPve?n8*g^frJok827K|1FNz?0;g0&``FG<7mz(4
zvYwk6^brKo1@rEK)c*8A)?r#@W<3UShP(>Ku08Z$x^>|Lo~;M5B;;g8)pv0l>oqw;
zyfMATXFz2pKA;xXiE0ptDzO6G*4_YeBdq{;R&9VxSkm9D9t1yrp<7YbR(~NyZ95%a
z;Q}R%GVyXX^}5_Pot9i15!v1)ND2PmTAt&}#k&p;0(see;X(n;mg-8{-U|9HoKGIP
z58qE+7N8M|MVh^e%|0V`Ejl}!th~bn@`r4<3~n5MUK%#LSThW&0cP7BMS>|BGX1OC
z>{#pWgmBYnwdJWY8(LZ97;frFZT}!Hm`^MDFv(IwbkerhsybP3hX*SnM3_}+*=j3V
z>q^rWk=rE?5X$=X&AJlFZqDnehLX-MsNJ2f>!L$A&uQPUjZtl?i0tKE`Z%##WWrsi
zkV%)XTr8&vZ0idX<c#f09H_uo6dsRMEWzXQV0uX6z1B&$<dg9H@bsYWwrLOBetNlb
z4J4u_`Mu;T){O$%xv@EzU-^5Ex2vXGq!Y?F&dlc>9c)*lFuDC;I<vb^`G~7a-$y}7
zoxxVUaAPP@YUKQcL5NHdN4~md0v@RqlI~SV61$?tlL4N9uM<uQyJ=X%vv69>-8?OB
zBhxzNCP(y;wras-@%U?`t;ZuLn#2fon4IhPLw&>eJaC0@0d=9QLJ9R<Ivb3GpL#=-
zx0lSC%yb{?RPT>3iY8IRBafSbxrH)Stt)L9V@5{w2a3ll4>m2Ca#_d}SP~5qnzGHS
zVtqF1BG0k5w={K=EAUtz+On|7`X(mkO>i)6GXd!*{@yY?UV>Vbui3ZA{Ng1Jn7SI3
z9V6e%q2%%<U=$qcBi!pr0~^)-N~yJ@l(1XC&wkGSiilepUw*8Rnxsvny<_RJ_+ueh
ziL|gnQJPe~kX;_wIw)6CBJT@q3ZzblVlwhcfQZ9iO<ErQ0!2Zxr^C`-MGN}{nND(A
z>?}GO>WTBVYQqh+H*6!^%II<~H%Rbux<^T21QBYD^(5&9J)ID`Hmx(<LSyR1L5eby
zidiaFi)!--mst;kWRVjZSh(PtNDO0@nvd&-mK_OwR#?7rIGS?k1Fe#=Es`CdQG*(f
z_^B-rgn(0M?p=h9)%{z$cQ_0lKHX}HvX?lV^e25Uw412i@MtJJF_Di?{B|(yhbZpH
zn>$Kj57RZE(#BS@UPb|7MjCby8jnJD_}#pfl!x+3-149`kc3gR@a%h^z*_4Sc|mio
z{LoVtu1S5UhiOW&y9mG6+8!CfgXY+W_yIe;?hZm7r~wXM)d;>vW%kZ3vz2hH*$Vv6
zlbt(}v|AQaTS6+ZH(KVpj+Y<81R`S?Yrt89W?>RgLEg!pv@vQ!W7|ojt)_Cz$;0Q%
zH9X`lN4I{WDj%9>uR4F%l_OZWtZ|>Gpk(J)vW`=-Bj?MMP>SzyOL+R@HNO)(Gb~$L
zU_utu9Oj)F)&zHuZh!LSxUp(hHP4BaqT6=z<rF(9`B)55ZgVxAPm*h4LgAOMvh|$r
z5~YZJK>pVE_{iRRW8h_~e`zzQcIlwCQivSOZm+a6!J4N!Vcc96l8{jy;MLeouMcJ6
z7-hBa$J(9>fA&2z_WQH<r}LEiJ>|-^KaEUReEEGTd{Is8NF@`Ui-d7*K6NWKpHvIs
zD;<{8v6*P(SVWMW#Y^HOn0UH?-pmg_EVWqZ$zVfJbV$_`A0<zBw-jBMX_%D^4oi!n
z|1DR>3C*T-=J2R_apj?*@bKpJ1S%4qZi#LGvq!)NWMv=M^mK~T0=Yzl8&kiftt%wW
z=Z-7P(3REJ$VspOFa+uB94sX1)VE_>W+u0rXl;9pw=DVYaBeo#a6a`GwKTTZ-O}L3
zN6)|2l0Jj05X-uGDSsZ2%gtTOm7A31DBT4ge4Cs;!JR2Dm#aQtO#bTD)m9NT7JF&q
zQJPX^Tr~BHortoSFF_^g;T-`gT1vp-;`7&cYeb2k<b>?ylkOQ=^}zid!|8VBJ<|9^
ztYZg|`MvBZxM?pJ{oK)l@UyAh-GohAk4vl}Lobh-fZ6)8#;;XlPEl6P_lh*Ok?aR3
z_l#sM&3Mg+s^M^^0gK5LwgOu_)dOn#F3Sh-T9|L%`su;_H!|NpRAxQjKda*ketM82
zXaeoM#3*-5e_`S)NXflkl!7{WSUd4XLuRYZVV3zyyjxMk=G<Z$77l5z$UY0%#SEFK
z@>>||wwGV&bVj^`*cmjYaG1EQH**3OE=jK7xpN^zy%_8=K(f;30H60TCe2|5)x2E6
zNs82sxpmWBRutwt?IkWAS<(5%DCORSp@1Qjg}KpEI-A0Y_Xw<bqy$e4!m(8f@5k5$
z5oem6_lA1wUUyQFKE)fZS)2F39;;|Nll!ES*J)e~n3q>$GuaPQzQ)@bl8siU6eky?
zW~7585*fM1mVPhe{V!8tkWJM#;ByG8%)pV0vpMn?$bxV`5K1nYOtvv)%CI90JSom}
z*OjB|bCh|Zuer{;<M+mFI?>uvJ;HJQ5ZOW?GT{B#is_^U%L>i+8E{fNm*b@*(91K5
zE0*?TZ)g2tqsbfI`FrDr!Q15(StQ+MbtEr^w>bro?{zlo&py%}B4i(B&+BF+1dA(u
z*@re7&uOK>OwUZ8ylfrI0jsF=`crps*)iUcmrd!fP%0`Su;4UNRBbZ8eZu{cVD_cF
z<r6BtREx(_T|kh*8uO8y8DsMXOXIto8e>Y=(W$FqB_t1+kTK0%{CT{*<C>(GLZ%BJ
zFM~gVK=?#u(i@0X^zGM?2fH`q%L~nKS{@0z#2Kjm0u?}E>K=!kM{JO(&5wt#UjjtH
zbYC*D@<qSzux}<^_?S3SFX4Z}6E4>>FX&Kv^;|94z6peVYG*7VDVeM)XnaFQ+vQVF
z^<^nJpvLEh31kZN@UNL3cmy_e{tRjdsWoQ5nm7wwN>0ehplmytiU1~GqMJ2%JQ@*3
zEZdbJ>P@4LB&LmGMiY;=1TD+AxLc>&do*g?aLm65L$-18ejoj=PSLi6-ppYCC$$3C
z&yuzD18jOyUU6aDqU=Dm{BklDIq-(xqF2zOi$KV1Tg=tq8Vud4^*{@DsX=Xg8OG25
zl?*cO+(VoA9b|U=%nHZEI>Y{ger9v>P&Vj*C#;KlgBq7ods98;T{^dzR^p+4tR|gf
zQospiiEU1T;ubA06R+ca1n-(WNHLR?nzT^STtquhN#!ilB1Ybq^u?Se@Rji4y)O#1
z=3Fq|ro^Nsz9RktQ6Jq5F>TNC0f}i*pqbO7V0~rdW;qLIig(gSO;Q&&W}W*SC&aQB
zr|sU2mt(AdUaDgRlNcI}nc99ZQmp7Rb>(V&4-w$9=PzXEuG2^W32UP#7)!w-mYPs?
zu$An}$Dx?VnS5(t;oasY>xp(46UgTY4rJ{qMPWP3k2K5rj5&`D(pej6pAvh6LUzV&
z0qVP0xuVJ_^3reL9vpkeNI&>Nks{{d=+u+gQ`A39R5nMbT+y@Lop2RQKna}Z2fT1p
zb!x#rKb~3PJ~-VD6V`yytvn2ynO^kzwkR6n0HbwGX74kWv5{6$GDN<R)=Weq3m*-z
zGcx|LlI5MZ(4I}4nZIh0@*eaEjL_5SFPtc8ps2P`D<aseMNYz6OxP7Xx}CL_jMlQ%
z!k%_cSbXK_3@bXu<<9?ty>Olgj$>lst<ULr{Zxzp+X9{Kp}Nh*l09!YTz4{Dr64i<
z`#^|_pd4~mC%UUP`NW0Mp<eu%ryJGsWx!{TQ{V-=VceAEFqrx83Z{6ldkA9qd(D56
zzZWr6R3QQ9zn-5C{{`A}K-qOj-4bR9tcE_s5W2W(Id~Q+e!Ht|%JUaUCbDiBW_#%$
zta9z`kH8bhM`FCWmf)N}19q3;uxr+@58avHI7ZF3R#2xOJz@V(2L9_dW<i-61BvD(
zC8BAg#~J2xn_KjY+!pUr5PjM$yDX}J@#O;<d{GO}G?Qm45%%kExX=Sb9qeNa!zFsJ
zdR0_C@6Sw;Lu~7@HMUw~lUq*@62p|<cUOpHm^|Y`BD?ogg2#^REB$!2RsAoUs5&d-
zQC4Q2xN~WB3Q<v+QU{2;Rx1ME5duECv^2#Yh!5}6QS5Qjt=y%TV}@}CidAkZ9C6g=
z$Gs&Qb4Je9MOu|DH9VL5gV%lEpAvssMu@C%<iFLQkMe%y^NB?%Pn(nbQ`K{Gzejp9
zW6A~DbPd2ZM%#si+gSn037Ix7!_sAwSLi)e^Rv}fiMHf^!C^7>uamczL%Tnu*Z0`b
zaf0F8X$LNi`)_qT>-1C1Q@C#XvUvKZW@H`aYiF`(5f!KCkWf#}q^`c7<>WOAGP&ga
z>UBXZ(XzPQ*7aj9iX%Jxz6$U9>q&C7txzoWI(ihcQj{1!A&Ba|EDF^z))5ZPEiu8~
z;;{6KR5IP1tS?g`D6_9PYFP~J2sFm>G?Jh`XHT5%{`7T#AQ{Ot1349*Ng=cp!gMQ5
zy|!&~N+=<ja}&wcqI&qa%%hh?Y%GhB9~I;A_-NmZI9=KKl_$GZnE6dWxqto2DeL5c
z+2gcAtwzd0D<lw8?akxNZ{=lnkN_hGUDE`PnJPn%50ew!a<LaO{vVM{$4>WaSSWDa
z#n=sHw6Ev}j$$?|Gp@eK30905fx(pR@mrjx3YCY-EitU}#!-4O?vCJ^ko!1?CIL%1
zB<8eE&Cy(w&JKJvyECmMob{BsaGfTu77s%=#$4^trtn@?t!#XI0h;^sO3C?5Pc=w%
zNIYH9F7(c(Z9+Ml6~Pw|V^_gWb<q|@7EjRHWwBwoC1L7Yq=^|E0!m-odnqsIP|^J-
zBaZiaS~mRQl`pcIFLw4=#$)}&e6>)b4zJ`x>pV~2lBa6eLW)RwVpTbPEQSoD&&$)x
z-e$g^xt(0W4}3&C!@D%v=yV>t{A`GdfCtXsWP3q9rXSLBOxL7LUHk$?qguRk+DJTf
z6J<@Z@U@uMWWszMFT5dJN5|*c=X$gEJ{J6!6bb~wPK<83?r>Vd`E1%ofZ@TGivo2h
z6n!wA3FaY5NG$-Z4v62^JL#m}G!4FG*!U89g{t9ieQIb{IngMqQnqn;fyMIAd#3*V
zxP&@lJnd?hS=x!w5&GAm?=8$~#~kVye=hlQ##v_tZ=QZ5#u>Jju&I5X>hO6b#+vTi
zMQ=Gr{(_JUIX)Vyl+RkF2~h{bR6n!^YSrWwiIU97;026N;JHkdo@wHGk1hrMEVFcM
zrK?D1&35b&PPx`9vEx#&H6Z!8W6s1)L)X@i3pH_W-ZrIZkCE+ek2#fl>s<7Fr234a
z_S$-c$iCw#e~Dm^+-&`LGFo%SG}-+;hCEEAzOawbo|(-%92x}w$lF})GFcJdOe93C
z6Cj(P+KFV{Zn;x?#E>XO&;(^Vr0)S;ooc>U*q>Nb`^=q)hBaBFIHgz2rS^#{qsM$n
z*dL_yv`SG&@3*CaTiNoPEEHDK6F&)!m)Pf4Y)3xbLi!f2lkB;STq<KE5x%#wuBB)2
z1-56dx0yzM?J{G);(AGl-6mL+7*%DZoQ+H`xf-m_*RbCw)9PZ2j?)y!k2qW{Dnv-#
z-<vY=+N+ZlM+vZ))VH{zUp;vlJihU{V!FoDOgPAVM;0L}V8|-DE>%LWjY=tkG0i1C
zDGk-uF`w?CG0bBq;BJ2}#MH~J{GMiBr%)>u*n~dhj;W{u<!)lRx=a1>duDyd{7gV1
zKGi*DCD@lQQ7k0Yi%&e;bPBmO4g@1qqf2xe8Bumtq5k7jG&{ME;F&T4clcLa)yA1g
zZA7S+g&nn$5+3K}V>}65H4)`|RvKc^V57-SD%m?E@i|#wh3f0+n7$-Um{#fepu@J@
zagsAfp#)6hU4yu+IP2x^{0VO3xqv8-g#{HWfq5#&$lTlXtVzF!@U-~IB#V}Yd>c}_
zgs?^|ga;cCm!%|y$_YOOMyW>IXPiQ*!KM90TRWg-f5t8;StaY7upauEaE3GZBHC#y
zh-m9xRF;_Z=Fh=#!JXIhXi8D7NA55R4Nn}|SVj`puSP=;-UwxbMJm`8x2fU?K2_^7
zd(++*N4Z0$LYkV>vY1+C9H(sI_-7Q1J8?vI{Kg7%!Rn41$W#;g!<IWc&AEFlvIVIX
z7VXWu!ycc;cY>#zzH+MxLRW8k=Md%r_SEIGnz`X4)uo$NNb5;Ee$h;#1@j`?{BLEF
zj73qJCv_Ij4DBF-;6C|GoJR9q0TxF=leyHb;dr4+i3e&41kHMGz7Aix2G90x1?Swo
zVOe1vZ!b-^pT>Ac#hFks%i(Tm<}9m~#7<})I_#{}j5>*&^PYh}jFzZ1mkDb+YzWmT
z*3myFYP0Z_t$ouu@9#L@LwmNb#uxNP?X!dbbZ1lpZWJl6tN#RbU9OI)rk$BqWT8_b
zrX_VMR3cq@hRT7afLsvz7B5{QiQeN5^WBd#db|cf-C5&HrjIO-C~baD{v-gLVKt_F
zBALwenWWAcI4{(^^Q_dURKxCegkE9i*Yy1~AksS%mP~fcbVx5(c&7GKJxfztyYZJx
zo=?nh)Vl3<UQN+_9N%rpM2LyDZ`yfSlCt2Fl@Q6ov|GqzUQ{yP5q=kc9D2X9<s9o6
zr_*^BOz_jMjAYMuj?))WTY;(=f1|h2lDafjtgWw#<ltL$8CLmXUfWQ}KA~FW+Bg7q
zWqujx>Erfsh39=pV@F-gdDtx&CyRkqWnZ}r+^S@Q>s*klQ`<*B$f__jy-x)2Jf;<{
zM>uHnyy=*AjrTAS@iU%G)EW+%8;tv)p%SLrpJQkn4+K-iCUAvN63E|GjH7(a`6x4v
zx~TbKs-bP;>~j|-QdvVAs+lZ2Uniw23fe6FU!dflt*a(GF0X(3!O8YhzqRe^AD4ZL
zAkE77>YF$;-N8bQa6S1F4aw4z9n5hb$-X^VX>K6s{|47)^_)JFmq7)l%|7L1{yj;x
z_){@+R{N$J@4=_!p@QmdODEod`=u7~bxQqb+P<zM6dG++pD?~l-iAxSJH7^FlXH4u
zl-4@_RPUWh;lnDwpUusT^6bTg$h@LJ&kKyHhMHI<hbKIIv3X|TYg9|tv~pk5Ej+Sx
z*T-_wZ*RNl-j{yJyD*5d4I-s5CjF>q&ZdW;`?X4|S`QvF%I?q(+sJHDR#xYsOp1EB
z*7loSrBbE=+pY0}wr*C`*CLJ4f@oXg2>U%?Ax|#Z4EGJ)>6?ux+#1*;>ak{v^>^7W
zVn4WFRz4i8kca9wTFNHQMQXnwhgllxOdO*kw<)Nf+IrlgoQa_!$*D7-vL23U!kYEx
zc~NjPcwWk%7}3`Cs;J19E^|Xe(M1Axx&*~fkl=7viKLc_aX?y;U6tT77yeuO-F7t-
zu~lwx-o*tLqd{UHz7|&+VU+%XpRyHUO{gwQuMct|z$m7I-`8)9v1b9<@n9>Ij^=g)
zpVoSpYXgOjFl&InPspJ?)XfT2Bhm{0Ch*BcKm)enNy`rJU!%#B^@r$&5y0GM!{n{-
zm<)EM6uomRjts-2k0p#McI=CcP93>*XQBB1!xMY#)1mcsMyXfSL*fbCVy%9Z4m3FP
z4^|&%IX#}<&pb2DvXb|Grd-D8Kv8g>P^~a29lw<`i;AkRNTNPT8$`}VAbK}vbd&gt
zTdb%Z`lDMF<xSAR{I2_sjR;d}^3L+MvLA^s!-!;mzhd@#CZ2OSypgIjrq0ShBzW@l
zLvY?%x=yjk5Z)0%lWU2ViJhHdyD>N#vP3*?`k-mzb^=T~AYJ{3f|a&`#+=OP+L!nl
z(^sc-4#?T^wy}{_)raz7_cejh)9+y^R^gK>y|z-bsEcCMa$I%aNDV>4;Z9)nwhS9z
z&qEZ*X>tZ0pTsy4!_IqD5VsD}{r7JVe=p<dokrd?n2`jFG^bwUb*AP0?Asn0zGPkI
zTnBk3gFx<8lFgf+vU8|y+mBdcJecev9IMqGXFX|VMRxYQzR5)7&LrO+)ZmasN52m&
zUM7JPahRrF6gHl}Esr&8bl#wM>P5WxW;J1iI{h47G65@U4Q^(wMe#ZQU;{ki7L1lo
z!O1hZw3&|WjR)V`Qj}&;Zr2zU$r*7x>-0#xxE6kM%j=+MX8LQ*20N9NumI$NNX?V(
zJ6p$JQhO6V9CLi#(oGT6a$&DFDAUdZmm{sg@fE9q0m>uONpV~@iZ&Wl;wAR8c8d`1
z@=T0_7-;&;^cEZ+uSPZBas6AAM&YZIK$d*3nR}xj8ZBBPf#{^H{?x}88n2ZZqp!RO
zANev*T2ybbv)b^|;40LCtF6$3?c<tXPkSFLHMy<EAaUuMIGde(;ez?^nz|{UW~J-S
zlKV#OK2^;Wudhg}ftAWQReXzNk;jU3WZef=bmAPMW?G#bsk*moM>`#@UET<JQ%EjU
zzTB6nbW$iDu}`2irs6!7XCZSdH1c3~XdP*C%vfr(rhJZ4)C?Z7!A#+l-a}pV%qDtQ
zAv%{Ah_JRdrgjdYgtEuSaAM}xiUwX>nyiYgxPz$^tebP|&?-l(ow?R=Nt^@Rj0eF0
z-!Wedmley&F1ArTib@Z_T^p|}w=8X?+RZ*&ZMs_-aP8}M<lnvmx8|CfT|HdGlYgrT
z*5clkSa5}KCHSmQ1>^R`VXpX#TzX$g7a3SJv-&kz9$?Trm<EOS;w=4LuKtzs7%$J8
zYL~x?1=XCtH?AAD`UZd7f7yS-Fq4%c!BCtWgx{&PjC{?3ycR1@-;+}RheH`mkL99l
z-|fRY4toKUzYoqc4-G}N&TaxLFnUlk7lO9J=_<YxOrE&=4r=Q5v~UBFANI_*w@u)&
zp$L-~#TqH=C}bG(<0+^0{prLbH6qU$qA%W(z)@j`zN2w9^y8=OlFfxip@+bn!HOC2
zJ1nsf1|-wxn=cR4J~>pc&Fp;6{JdIMsc%qZ<wxc?Gp`%@vxK}%qC`D}(rt_Hp(r=r
z58f#iiYY?aT{|X5K4kDotDgC-AkT>9n5y*XtsRXoez>Hr(mf4T_?(2*IHmKrgf2HR
zX8AG|qwHf#-OW45p&X=-#)SB1r;liQ*Kv-_FNqF$E;@3@(`U;^w;ml1XRB1|w?@M0
zLyjy~?fI{sHylJ>*W3FbDkkj4pNfRO6SP;;=1j<c5f3SBW|@(BR6LHdOS|^Cow~yX
z@ro&NsMtQPmT0bDH0QQCHL^sPK!4UbG7P0YTu9>*3~@C}-d<R0da06mXXd@l1m3!c
zLPyISO@a3qt!xBRWKE|pNe$e;Mf*mU=CZMhyJ_9NSa5FIDElJ4ce?B?byJVmvRpGN
z%|kCClDJn!BNy(buUV>HmYDY;!sLNlF<i;1)`29ET6@(|gdwRwwsyV7b5Wvg3MKHk
zmWM01gG|unX?ygOp@4(riX34IH`Y0GmABfZnka!Z9+hmm4~2{of+-@Kt-Ig(w{+mE
z!=>`M8tfksL?0Izjm|1X(bW=LSzp6^rl%7^DY8}_=7x7ItH7{Z>`n_pyW}f`pR~$Z
zw;DHX4hARWW0Vv0%16o)nR*?!G#$|?0>m#qgw4JcaEqzuIGNfde0$2YBOeNQi)VHp
za%_2}wKHVYJy;%vd*7yht)$dQPk=k4S1~Ft@iP*hufZK@)Uhcai%Kqj!-QOB=}VyI
zu5?~^Y@6~j4W6VTIS7-GAPF4%X@d|nJR<a3hcaGRmrU4tY?mTRb&ETGYHsTERYXv+
z$PE%?1StJz=80S-OKadd<IpG}P+!T*6<`OunK?AG-3PV-L3_PJ>Y4}lgIv#8rWJ-%
zc`aYtHd>*|r<(}FPaepBR@R1atku^@jE34Gp{>O+HG-|iAu1^*g+d(6Dz@@#*0fA2
zib>b%;Eh%L4S1lz&y_}m<yplB+&K_=xxe7))wPs3d!O?uqYR2QEiO}8(F1JSiIM#3
z8%33mm6^-5D=&&m&UjQye38CdMAH4Wr;iw8cCPn!>Q&tjn!UlwVS*Cf?2l$tyR^#1
zTHKkZ+J?snpVEvUV})8M0gl2|^#VQP`r?REt<C+442#G@oxa6@$MucJi}oSg)J@-o
z(zBuWz-}8@R-XQE6hx`2CBirWw+B!9NV(CW!zw+qfz#nF6@7HsBwJ%!e5lR0G{*gn
z*QO~f85yL<!V|ZwN}o0%%q=@PvCIh*eGYZUXmip8GL#BM2q{mSZ6G*R5lA~*c7Ajq
zlm1b`NTnY84wn<RxCneiy4kjda=TLIt)V$zY*V9(k{Lh35q)s^=2pGLt%8uivzf7a
zY8sPX2a5O*)U|GV+G`#~n}D@|Pbm|8ov}l8%l%HWqEk_Vp7YIqJ}Y|ri8a1m(<9%Q
zmNK?+bQ2tyaEo5n8&Q6LL%}}8$I4q~-(pgcG8cZs#84`6PNl}n6znMI+hX7t;H**_
zQn}j<>!0aqrjnGD*zvpc>%k8Gzcl<$@f);Gt{Nbkl|R8uhkkv*LDYBJMIn=2eUBMO
z-$PtIY-bN6MYgNb2hC=nnX3#BM6YViW_wneCpIvhP1i26ERT2?uooAdWw$%l>%{TZ
z@miDYeqN*!#{ruN|0`)#ZzI*Vn@0}hD9CaRB0oajI$O`LPs#Pzw9X~jT9?ePmRqc|
zY_ja+(;*<pmmLavpAhBI`t?I;wM!<6>xWKo5ptIOF@M|b8u88V6}5=>j<7kUdJao-
zYNy@g;uk)$g~dbj?w!1qy{it8Fg`nQq*33~$pIx^7ETl86TT@ez7;Dg>*0uWC><;I
zbgEz7>O`G{iFS{%B)nAaMbqA;M2PMvzKDm_TEVzA3lVd2h_S8aj1O}?HT$gsN;E-o
z_KW@@@P?D#?hJgm+p!|yXKKZw&HUb1PNn_E-TVSkiqcZxZ$P5D)V_8EZk}iRJYbK)
zICaML3+d@r%ZJGV7sbu#Iyai_txbF4MhRuZz=@%J9ku~;W*U=<)H_F(U=NZN*$6V)
zU3OB7f%xUZMy9^l(1FD5NU&q=%)Cb8i=nbNat9l=7Yox)i@k#11ANh5_n)oYa(1&e
zuk=t||4MgTE4R<*$kjSr_bP=l(y%`fY2eh-Ae@{I_%}+f8)awG4385<4@1I?$Y~O-
zVB|A1fZJm&te@rW(!~=`;)t1Nbp1suWvN!`uhZ@4iB0vR!rx?i)QhSVD*0@U+9)dv
zE+i*w-F=~)K2NRPS!u@TpmL?0;;lvel;;+&0qL{N)GB|g*lD^9QaxM!ni&E5llu`<
z<p`}^%$%~&*&yD4mrIjm%R}V?eTp9T^hA&CB}#a?Iu*33N(Fb|ss(BZgy2$^KE_r?
zrphOU4_S+1SS?FJcLLYXtv5RgwA<SQs8+(g*f{TNxvgeJ%Ordde*L1SMPQ0<A=_=&
zWb*SzE#tSU2}ELz<C?L@X#<I&DwZm>sCZ|#pu7csYQhE$9g1fQ(=3ga&ctNYg%h(9
z;<H>w;GDR|QH9fIH4GOldb<=<bf{hx6Lf0f`g)J;Dcp-NVzHYi;(u`W)lqFd?Yg0~
zv{+kQ3vG~M2`yHnKm~VqDeg{isX%ck8l2)JSkd5aB~T~?4esv2p*O!z);Z_Ox6Zoj
zuKUL~f9+(?%$_}azt0|d-{&a?a&YV(530>Id05n4o%TghR6XZJjv9i5kA(aOXAP7Y
z2K@t+a6^2&PQqGO?1uMk>x!;hsc>Y9&0C>;-(%Eyc5vCs8Z0CHuBv8RZf^5FVcnMg
zTj|aPL!K0%<;$3u5`eI8BF3s8H_*@`P(9|wiG%TYtn>Q71%PL?@*LCAr2Ge9Xd`~0
zHuX^S#s`=6!@sPbU*VvQ(s~Law`87Z`G`nXa#{lOR&sSo^6(=PEYi~3c4~bto5sa6
z@vWqI!HCPW^h~zPkGn&dCErU8cCP9>meXE3YK|h_7|<|Ln+@TwvKTjz2>n&}Gw#gu
zEG~zw_(RxLiXKuq$fELCAQ}Dj!nfUCt))0QI~+kM(48+`WTLq%9eSgBn&DkN2v^Rp
z!NbM-s$CKW&aTefq-`O=^HbG!B5L+)AB`dG>N{0y@RyF66tg%V(Y-i__;q|?(Z5g8
zTuJq;cML;jB8>4PtlUuU10chW#GhTic{CIjx|Ow9hLjg4>@M4I)h>KO_K1Z@aDPut
zFBx)jjAwcc9$Hsuz3oeX^arq&Xpa-;5M=x-xR3FQ5gXNhTWSCxs{P1Eb1>tj<zc~D
z0;fsC$hMIIoHCqfGVXk}8rV3=6=xq^evn31N1?sJXIBucTkFVS&N18Sf)$q$-lffd
zQtv+9Oj}B_<&Rwm4&Ty3CG%7BFou9dVCqiN-aX^-a*?0B9cCxG7&9^OS>^T!w5CSE
zOjYqW?39l5p2a0Z6+cU*;5XL<OCL-nJ)zTlQNHN~!|u1$D!w>rv24>BdsN)?FJDlu
zzB**&mOb^-)3~K=(H_<AyWK%HeHL~l?MCc`d|ha9d6?f<wKLsPuEyMK*ShTW<O>%|
zSJ~^y6G4e|sjlkk#NHH*q)=Bi#ToqBFUu2Rb&NR#DoMlBq!O}ks}o`=gCyi+Nl2*f
z0KpOl4{$D`SLv*k2zJou#cDgo!<Jb8@lxB<XKQDVvfDbZJR}G84Q&z01cr{lxleS9
zHTUXjPv%lpGnbv%zoINgViWY8wl4fG`)HQBTZy!pYfDq~(?TZiCl^<qoCzp5_e&z}
zx~r^7ZF{EUNg4)Cb2;<F6gLsCT{c`+``@#ceF`y{m0hI{MYQ|@pvyms*(_|V(&nc6
z4J<!ID|DFd(HPbJjv{kns(z71y#0gop^53f5Z5cccyC^cRnhmw9og@`I1CplpHwc6
z8YJ&3+9YqOWVjt2?zLvUaHvTs@vb^*hR&D_u-vfM)n-06!65y|e5CPnQ}&(+O+9Ak
z;=+tu*cL8nc*y|yMZj*LyeK&!Ijz<?7d4^xTGgwR!RetWg}$PRVkqRd#pD>xE^Iur
znfo%$GXSWu0Q&<Nzv;*;n?aZe_7qu=a=hq^OGqV*LJqo3olG4!SS0DjL|N|T-S#-)
zhmQ<Mkk`L!Gng4FFrfNO!xXIOG2Rm%4jOo;j{d3|HwX(mpsgRIYp77Rk&=~GgdMTk
z_+NSLUUDakzoX`5(lC+8B3_XR6W}FXYMI)9c78-BU(yszg)1Lb$|;M93&knk51OhR
z+nHUh6mW1h`nIhL?avpw3VZaMZo|$qfKuH^)3ip>;(SHzNvW=yzUG)?U524U?Uq7q
zvz#zNp`i<2!$l*yGu-=NDl4%xYtrz}B7Cipn!~*bG@!@+GN09dkaV)>eXj0FDrc#H
z#L=OED{#Cx%t3E7=beW4^SYHdHkcE4=qshkp(>B5sqA<<#s_kIG&<$(RSXwB=u{DW
z`I@xba*{;kXLSFKu12HFXUi}J8{fRI&-PRCyJ1pzq<*}_k62Rd?9binj&<KMG*H5;
zUa6OR9EE}=oJAQ~Z^_r*9nf*o__zDUJglPAt~(eduty6t3j~MAr^xvfzRRn?Gy7qF
z-tTA0-Y^mc)tAk;9`0+S@#}jPL>M2HU5CNtrWUs`iy}CK=_jVeF{dp*glpL1bY~|(
z6iqJc$trUC`1;=dNFi|GreSAOEvly0g0Bbcd@}2JG?n7B%k%SjQxj0*l}puwvOH5M
z+7b981gU7>mI*yA<_D@bidZ5`>vPh>^s8Ub%4s4M*(C~9h$dk}OT<P;UlwT`08>e0
zxE$#Bm4oi{?nL}i4@bUemvp3=ERE5I$uW0^QK)ee0jHX2qbTxf2y_(+>sCCb&z-j9
zX|-~8f8E;j-+rGFu}YdE`nJmvzWNHMf<)yd&pko7<_=@?vhQ3Mt%5riE*(>z89uSZ
zB^2#O740EI%0fo0&CQ~8V5wGzcrTj3i8hrL)tK@Z)CQ8B75%Zvt9X(g_|8)#v2lL@
zheBk30L2taZAN88ud#vLeMLID#vLWGCNpOlU6G?kDu!ih>_~BU`X5ESQzpr_4mGq%
z-Q6yIVT$U6vL`>4@}wk-LNoe`JHah)_L?&)IcCFtfV=@PbsfWP{Wh5Az^2lV%#27!
zn0o(-rH>wKluQ;uUNR~nHnx<|L3VERla)S3|JNS?xC)~hV|-U!2*R;OZ00zWE}yhw
z@D2$<V|Pb6EZ(k=S60I{rP;OlDUxRR8YCSJ4mBG(t$7?$z;%`qmI6d8yR$}XeI*x+
ze*mHK-<!q2hHr><D;+AeW~W0IHWul9mF@{7rZhcms&HLtm#?vA%~MVqd`EV?BbQgn
z*PxM>kSedphO&5=qiXR?<R|H=zE!5@cH`c|Z7=>SqrK^E3@FKa7KQ?q^nAz*4=xVE
z5Kc`1$M<{`tG8p})J{4cK4BAO5t<-&4G6oOrcz`=_3xs|59!}8oPz7t_DpJ3*;$S6
zQ`tI_Q1SB3jHl3C5XNSyZruHK<GOWoVgJfE(x=z_O3xarGBN_lKloQ)e6`kUZ4CT8
z5@&3tE(uC$dU7phxV{+!7!W-BQvLA(?{MT$EM*HeOrbuzYHE5E(J|1UKNnK0t2k9*
zI^v>CR{8}prozo-UavLK&o5l{IC9#bWHtLSgUtyDZ?L7->lJ88y*yc9I^-e07_;?@
zWLqvldtOQ|>kWAY>6j<g)tyAupP2<lLC-+7(^#%e7OaEG)B94V$vP<pU0gAi4<pjY
zZQ;mB5%#S;W(CBxeHn)$PYjqZ8KZ~O9>t@oxuhuYS|v4%dAyK(s<OJ|qyt!EpGGB8
z-q}RGbl|h|1WGw>a45Gh^+~5IEvKi{RQQ>xVsH4)nAz-*q5HHv0}H-_Cfmt-+sy8l
zJ&APPi)l)<2~`j|UfLFq<*JN^CF0MKHK0c&Dz;>`yu>q8#noS))9D{*)}WJavGJaS
z+*6I&>cpxxRGPp|A5nos*Kum=w8(m;%1b}f_n|JUa)tDzM=rL!kQw578;Bqz+)%7s
z$)D$g#zI%o+i%fq0n8h_Kf{j{F7n5pICO|rsz5OJFvZI3m{ZTlLy5^W&w2l<=_bF%
zkVyZ4xtVH^@-GIv-b~ECV$^EoLP4C*fM4sy*Vg^k1j%1$?vl`2#XkTBhBwE{e~t2R
zRv#`XTZuciEl(ul=`nl;&VnD1lXcY{KI1W{RC$yo>f0`0k9aEgcb1QE!ZvJ~f6Ig$
z9azSF#TbesOta;m-qpYoAKfNvD<$0(fvA(L%qfs-68cyP@`4=VaiB$!AJq8%>ILS@
z`lIa&ks!q@fxU}kK=nkJ?Hfd`yag`e{b-hWL_a+kI#spw;3#4Se;H=Cy*ji+#tCW`
z*g5~AGuyywI|!B{8;G9njTh%{WklC~U3k8wF_pS|u-`iHB=P`DL)AYR)0>@CU(-#x
zJQOT|K0VsvQeHN?4sZ0r98Fp!eE6CYK=U(pRh?y1lqiO;$5P6VJg$nvCl(4eXrnW+
zNQlSOSnRUNlFH*ce~U|iwlgPwaoUhD4ALW^Qhkti++w$gY@?O4zsRjnI%qtO?6iMY
z{gRH#BA?lSw0=_aolEy~HPDz~p+-Ip6&WwNy?`CM6+Z<HJYE|ee1@ulB7e!@TfkWQ
zhA*{}^lpC=%4x*!0+(Kd6+<+q8*BYUE^r^GkuMtX2jV4N73#AkVLCZF2Fdj}ox?hl
z+^F!R--NFGj?Z{jGS(n-E@dgeyIl@NVZ?D+o2}&fltlH(Cv-)&f|DUA@Z?HulSPB3
z>8}D_d%<OiTyJ9cgz0VUcQ4@trl)3^qxzW6d|A|J8u25ZjMFoyNTpOyGsNRporOli
zsIzWvPqvcl_m#vZ^ATMI^ZF6GkjBPd#rq63yS`rFu3FFY9M+q&H#U+B(JFTRNQN=2
z7<&xQlNfiS>V^RNXJd7$P;i><lKt2#8ro|b5*tXGKPfRji}g0XsJkp>=_O+N=g{kk
zkvY2o2KBa*6sH=K-3Ph-dU{<FrXHlo(h9Z#xU`HT7%26e7@Vb0FkJ%Ich$V2nz7C}
zx3iBv#0FoA8T#7`bWaO``RKn`ahTWSzcgKi(lPqm4v!?Ib#5Ic<vVL*X$r3Rw&-PR
zx1pc+71N!vhMMeh7s9$T#Lf1Ey9xyz6rm|ZrcpQ@Yhg|%4{3PXp+<Q}SZA36Aqx9_
z?$<{c^iS_Ecjt&tfBBp`TdWI5DgL@O;|U8F6!o9xYb+aThli;@+zl2p+*L9(U9!0B
zx|6{=TgQeW&{OOiO6Nx<#6xQ2_{vJJo1CB_h8M3`J-2PDDLk=kqCLK!v&3T6&@@>>
zd_E=k<PD|58q;QjN`(s1E4nTQYG6u5wr|KXfu&Hy8M)2eL}-8WE0-bekoho&`M{6>
zV_qgSSin{X&FPA8wJ3h%BCy}h=>ZF)N_x(#*JN|R?#3Pm_MvE%_6Wfv7SrgV`|_mG
zCCt{da6VqZpfIZTr$k)C)D&-P!F;VmP5J!%irw=Ci`<;%(+UA8%cMizyCksfUXSKv
zjo3_V=yRAG)s@!j@iE>>1Cwu@XGBR~4M_fL$eiRBImQ1qt`SK%s3{#^*zGmF<ovzQ
zK;q93C<r;&?xS7_D}R^aONids!r+MM`U!8XGb7b_?xb5J<<A-Q++FX@+s-(nw@7O8
z1=t(OZ#0Sqh_1$i<lk^y?ty3(nlipi=&x3&+me5J^WSJ2nI@IT@I8gPaxZ{)KV36<
zUS@-BnSTh(@xq;lcOeiQhrXPj`0R`a7fgh-O-(J{SLh*<WBUG~KwJx+xL#JM%cq_@
zHF;&7*y^l>IwVaK9M@t>>#rl1a+LmZM<RrDI>SlL%2D6<oG{HuYf~mJZz%MN>{-7J
zDU)32wrh@j<R$5f_e%?&{3;;t4(!p3D^7LC>gu{+S`8`ogjp!(@qE&DIV>{LcXB&E
zeah^K8J&x(0BcXvo@LQ*$9sC(5!gXdN6yHSstHm8-#z?u&QC>Qa*YPndHF|zznh1%
zdkuIq{s0hi8a0*r$yEoAu-_&#Q|tMoM*f!|p7oACjqcZLmrw?-9i=Fmb)xEb^RDDx
zH%vj2O+K4#eP81pk?Tz>mUfQfT(@kYerb-H`3kK9DHiC5+Z+kAlxLj^&AN%ur=_)P
z1t}g&m?4RxMtr4b4hYa=TkZr%U2rxx3!xtQXTH=^$vyqj@+3nqSk!h^XC9Ap>wPP$
z-FkKXk?p!!=bjjLJ6m!z{F9=CqQDYD$9oIXvi$35UxdH4KvK?x;wbM3WOF?5MZJBi
z$Cc5Hy2tG7n&i<$B2mpMWg<A@qmHpu<xKIj+G0=A8J`M|i8PzJ3}VeP_BB|H^!vh*
zR|GF5J}1U$xr{G8uJj=jC%$mqj(KPo+=}0?ANRwU4ZNSVrQO3r;<6+sqBbDvD}!{8
zY``2-Q?(5U7`<S+Te$iKjG>tX1@UsYy$~nNQ-+IvO)?uT?2k47ogU9>P-tE3qYyHl
zU;k_(31Os{gdoo^Liw1oo@mZ4FSagNAf3>k{T;H%Hc;t&oy9_D=DMfEu{J;bSx1jR
zX>y<5#BJ-ZRN4F_hx<t<{f04D*~s%bk#Ksd(T!J5kJk>)c)QHIv4z&w+Bp&057kq7
zh9{%YetEoA5s%}1NHe+&IBi}FZK@U5`OsiCGMT_Fw=D+#o;tHt3TONC&~CB9S%EhV
z(pAs7WJDkIqw6ACl>_)knf4Y7Uw&+)s#m(m;yudh!BM5M0gx1>0m%OwMuW|AN2W#;
zK-E`Im5k%hl1<`kcm4h0TM`n$c6k}rsW*bF#>RO3;As2gBhT;3^XiT!74AO(+L6G8
z{7RsM5tiT<m(}#+zq%5!n%@g1<;kX;d_CKe4wDqhW+aT}G8AVT^iAaTkx-Zj&yl9@
z5}x`Cy!|H5NyhhC`t@KUo?6?)@TN@);m+tmm~tY%D=X4HN|ZT_6Ee!2f4B18<AhI%
zuSKSA;uM&hHYUTq>T0${kX|4%4N@s4>=tY4Wy(gF>qe$D!XfXt_(Vx=!@9p`nvX2W
zS~cXL@Xfb3e{`+~>jx*6NKW2F*4PqxTFQ13q6rQu?WO4W%V*^H*A;{HlS&2i!y!ER
zK;J#8ySy~E+U-ntvUCq@Tv~>Ep2SS)<qdPXZv}S;@ol#fIyGqMviUf)8s*KMQD|j7
z%=y*KM3~_?&dyKWQyurj^6iP7wp253<#r}!)=|i9f12n>)<608oc~#xNs^vRFN3#@
zL&3M8)P&^-)i|{v656C(9_GBrfxb`^&Hh%&pWH)P@Gr-|rp8<<(wztcU(kI0ks;Y>
zV$$sNWQ&LM1O-+E^%F-0IjmmBYFO%<+m}7jTJd=HCQfc6pXGGCBPsa%Py$WPK~rLV
zB}DeBsrN(N>&7a(ZSj5i$&Y>mi95^_JMUK>8H)E`IJy1-XgLy#8eBd4_^7qaQm<W;
z%$kq!V20Fa%f3gf3cltb=0x4bm~au&Mz^XW?-3%UJTqbwyzz<tlfcv0KX5gj2yZ*9
z_CSdzzWp_lU%;E4*Zn4PJD1C)cB;9oJ0b3>V3}_#&m|F*Rf?9vwj>yVJaE>%)$g;R
z9mDs%uIcSXs$N`~6(~$t>!bR~&IW^3x<pzO@bK%YD3;v}FrDy;t)W9#!?))L3-6vA
zszrsWJnOJ><+IdRFsM3N61z>I=@y^-12CFD+oX2DByB?wqiu|tN!@4xR7(@fY#txF
zXYTtf`3CQCsSw<vSE@k!sn<vqjdhPiLML1*@d*#D&DltZ+JsDy@Djd$u`Z|5LI0I6
zg7xP6+}HhnO3m_qiGG&#q#gWvm>ZKQ#d%HS^9qZ+DSy0>I5a%_+4)*tN{Y~v@maZQ
zJ<7yb8jB|oW%8+tR~-7LhZ>dMxDo_f^k`quH9I|98-1L%oz^Ld*%PDTkTAhlc;!6)
z&2j}x8DTngb{^?<SlN?!ys>Wsgk{qnIp`V=dVYTL)q47Q^()0N#6*aJUwhmT;l2pZ
z!7^7ee;g-+-H_2mN~77VXDLT2B2jeURR#RzRQ8Bb^ax_sa=+DFLct?6!Ii>Z_&4)e
zyEmv|?P1!AUhLvab9t^hhmRLdeF1pyI&CFn6Hof<$E1tqJSj}KQ+ERjjT?+4uN2fN
zWbIOuv}sg+FWFb3oIWeWJt%5Q6y%maLh!66&Do91xaT?uknN5Sjz2_GBQw*JOjgsZ
zSL5yt85Ws>$r_b}n+Ubm3|ZjmwkvECl0NDvTZ6s<C5KbeEE4YvwI|Hkwz@XO@Ra<c
z+Bw~eV2zoG<#D9X{wtgV->T}8efwYS`US1J94)MiulX0s*g(C+KjC#dQd5O(a3x;7
zq5L|~zN?*i)o_wJWWn708PW7W1cSZRBGM#9)q<~XBv;1fk#sE0Ng5AR3o@u+)?!Vo
zHp9V2jdB{5$lN#Zkum*O7gx(-B5sG|;Jq(ABiSgJAUiiNL_sWsm4h+De%!`$l_6dM
zuBPKY{*_gIvqpA!Ni-+%NxjmEez+>z#~ZrPY9MgsJy_qOm(!{8056?^rmQf|zAnXC
z;Be`Oxeq2I4MyWZW78Qr!V=f^ivihGzd&A^S^ugCb@2ZFmL|;G54$+U(KNtYX`x|D
zn)$ugWs3K?xP*rWbXF}&>Gf4Hc8<)?HZagbp-Ew1TWLt*)=GbhX``X5agHb0(cRPJ
zD&hYDB8V@6^H%Np@=PNN>N*Az3Jjn!`JLZt6U?OwP>b185|;5i)Y=@FHioSmsr0U*
z19Qc%sGsIbIb9y_+2UD1R9gjIkdh3qG^TUQX>u8+b3KyK&_-;~GrbKm3D=TmI^M=C
zlf-0=6FrhELZQN)rWfz*PShj?eR%h0?1V~wgSk~BFb&MKo7LgaRUEov3NM9v_FT1J
zY_2=8Q)^ZZIh{%u36a1iKt1K#l|A)#$sV(onvw#xPK06olBQD;EGr3xyQzaD%E~cv
zMM5v_`9|=2Gg*hLw;lJ53%Ls)TU~W*8l`aopQmuY(6PuqV}US%hn^z{h@y9x^fhk2
zr7*;@_bmVD!S9Z{`UL}3BiZ1Z8IX0#X3`*mX^k3~lq_=|HULz3fta?6H~gY)x&}c;
z=90!z$EY+AGM?hv5f<WMF;F{dDp~9N3%&~`Hd<D%Auu<u&}cvhn9|s~$rp*z^2Fso
z*54pH?;aEnZp_iTm96|$5M{nu@q0vHTe7)FC~)YdoD}R_q$gfY{`7gDC9*V-+sNP?
z%IVZ<=(gHwG{D<(oL>obkRn;D5=?$X2Vxs&!K#GLod!3uz_j>{hFfh(*r=+ig;ev0
z%yJ?$lg3OG`n;Xq#~nu}g`Jy8ylOya2(##$&3VK5Hi-s4sUXMN1;}eDl`1&sYfWXM
zSfFHZ1DjKNhw<_F-!i&vkvJ&EdnRQuX0?QIkwf_{hx{WG6`DHii?eVq7ZgqmH~uSg
z(J3HIc^VdHYaxg3jeHufRPPd(VkWR7JnNd`*f5h!oi8aV?lJ+ES0$!u40RyzD3+2-
zXo+=+3&#1hn~hf9NK<kiwYazGE+35hQt(yeu8!cmL**9-^@5*8Xvg1&cM{^5Yu$Y-
z>Ns6aQ~v;-X%T2x<P*b%xP+CKF;BvJc<*EV&vpL*2A@9un?Kn9aZvG}*W>;G#wbny
z049zOupF}s01+9$mkasjr+B4Tq>nWfi>@GGpJIvP4DY-Ur!w_^wN>C3WuGF97^%im
zlx*ne(CdOU(BpX3SZB}!$ri2UEYASsd78{81*P3#Z<kXUt7a6)p;Yjh5OcMEZL|S7
zIa@@PH><gSo)6%K|AIkWO9g!&8_H5T8Rb}Lq^DawB39OF+zp1-Uiq*s4mxbQxxSb>
zQ4Zz5I7r{r2#@Yl2!yJ$`nk@Qne28uG*o9PbR(7PhBYQk>av1R)$!%-O3X;bM|Fl*
zt_ws>tEB#NutbA;6vxV@M*c!HwfKv9UiE%|Ey&Yxghr`GAyqcJR4!@&BWB7`2=}=T
znE7<J8hZ`golq78|5{a+RG8>|;njP=h}cn(K;L@8bH0<ijoU}l2J&vMo*VRN4`FgH
z=D8mwi-x*l`<glC{gU3<1DN|yhrb-ct}iiLKremBC(bLgV?PNd@!;1190XX-H`$B2
zQ;0wCd9-}17t6oBdtI;i1l3JU7oce9Yg3Fmh%lt^7f)D5ZGNm{{e>4mO3#JgyWXc=
z)hX^(uhVxtb023h`q%S<Kn`Qk`t#4R-?9CQqgJz6mR@Hw$%D4C+x>O@W>jNKva60U
z9=%C8HLm&7P3g9!?elOWhWqXHQt3pvuHP^P4XryW^aA4-`23rM24n}G_Uc;9e}N`8
z)&mOMS{_1-1A#f^_of?<raFgddmM&%xJJ#rkf#Efui0w$vwOIeQlqUoFY9;o^}~&0
z%0BE^eL2^&fHI0t*@asMMmV$_pLaYzY(OWB)|T(r4Lxm-*W4$FV5Q0Z<!D%muwV7O
z&rM`rXYc0|4>72Y5PK9OpZ7a@wCJ5}K6~>?Did74v==Kcp@%diR0r5QK?y{CW|1wp
z=rd8UCh2<ihjD6EM-DzzKF_9Rzh<ID;TUX0Es>o)mcZU~xhakp=u$&MvsvVa0~b(m
zRZsKf?r<J-wB@F6UE;^(rh<q+)<IN+XTOc>j?z4nC*E-HHxAVrkBb7STXlPo+1N<s
z_!G)P+qFJgq3y-UO~r)lhki+sEwbe5wj9!4w#&mm*-Q&u5kPgzHP<9(W0VV}q&s6-
zmd`=enVe!c0S$8x^Ug)Qoq7}+?W=ygc+aDw>t+JM;2n0IZc|EbE4ij=2|w|$mz0VH
zlBI%P+7f}^(ajIf$WfZ>7f5+)7#6|PCPp`_0w0?w-A)<TNSj5wRY>#G0e2fHO?VEY
zAov(UG#Dd9l#?qPp;FYUBjTtCSZhyFjgcz)>F)tSI+iI)K`B<>6m_@?Yo%yXP+2ZW
zS-3EDyqO*fadXWj10Edl22F-447*s~G!&1~zMhUMlbtD3mn(PSl}_p=&C57Z`XZ5s
zvW2jHh5K2_7pO(+azjPuJdMPCzgUZ==!^b9TdpaU5S2`EBaD2Kx*GPFW{V?LXIfkc
z7EOf=Fk@9Y297xKB%6F*dwX?*3|WsenEPCA+=(>~BP@-^6u+fOyC@5ZaJ*=a>|OxB
zmaXSF(|TOJ9#xjbuny1Y)~m1^Usg9AQ-oWzEo1C}>$&_~3w*V(qi>7Z8%Jq^mqj8h
z;mP`o4|m<#cjP+lZEColq<h_0sn}zy@umEBTvMXc1OHNOLl|Z%>tj%0rMjCsGf(kJ
z$3vd-%7WtJ!Qszop?lsEw%p~t`3Lvin<&%aYgBBFuU`@p9KQ_m40vd)b}Su=AQ!R3
z<k!m;{X*pF)_F@s=|ntP8oM+|jiSu!5(>iZ&P%_d;TOnx$=HB)OggwK*8&FItV>t<
zR_a>^43&P@E*pu8s=ji2Y(xE%wMMm<zEZnT`fFDTly40_VQ$SA#tXC8-tAv)n|73w
zV<-J>TvNQ)CwRz^Pn+iEe|<ZbCg<JXGTe&4O|=N`?F1!Jdf*#Cm?;?NZ!T1dWb@Qk
z#H2cFl5a0q(0pIhtfQ1)oTRws^5wisQqDtAWN2s%3LO|+9#HKI@b-ae)`|&SO}>s>
zy1MW;nGe$1yvnd6=cW$1=bUFirP%ANZIg5(G;Pn}cYjHr+$R`3Yq1ea8@{x_wamVT
zU;nU1{p(Qa)%f@vG2}IQdV=LK8%1dj>9O+1<dDyW;S-L;WA-Ujvox1-;9wQF*YK*p
zMxtnkzWX15A0O1%yLrVCMXjfiS0@E=SchgAD}KKc;N^C_7Oju!Zx5~e#A(Gfpx5?5
zvk0RjV3sUb@{|+&OA6E%B5X!H<5K)`tGFrim5HK8Wf?z7wX}EekN(nBSSn~g3F&qA
z#db(JE6bz#<}Vcdze>0yY3fLA*ZEnGCD}iK%S7di&hqas2Z@ENGkikpPd6uy{XQyD
z?r(eeyC|u}Nf|r{y#at)N1YNUm?EpQ#9As(w{hY;o%Xk#L>skcG@mQ6Agij-DBQT6
z<s&Rt%27xBxREPP6*tQTH!>yw{egnS_k5p?k4p8**ZWOCnf&3o-mQ$Z^mM@7cOV?A
zZ*WmK6V<u<l<IbCr)9sqU{2!1|F%5u-t2Wtk+9ewfC^#bg};is|Lu1w&!fxl8&@}P
zz^(l)X1V)VfoDalK#q%>5b61?%QuLBoHmyEXNmt*i<8p|4&+u7J%6yBLKiZ-eeuyw
z->s=-#j&_*>CUXs1^4jZj_h8=VE3ZyQvBZN=NujxXVcOgG2yr?!h<Egi0!kH846f7
zcRF7VG&ckI>-gu4>qQv@i|z<%UvW<O|Gg20^T@5NyHBux<CnDdH+z{&EuMa(+-@#j
zw77_`^KbHRxcx|r3|us77L}P1+czG-)xRn``bUdgJ!9$MJ2)ljzU6Ujv)WczPUofo
zCKtoZ%8NETL`=>N-c}>P?p2sb$n5SKAiAZ?SUELLW>;K#=mf`Y(eb9mI4~O-o$Jt{
z=jE8<=_dXi5Bc0@zTs!qlR_=G#@G_7p<8m^kh&bDVQ`vnFRMxIbG4ib;9>>GvC4==
zoLNIbybgk+1>C{uakXEKinD>aFhubu)GZWH+oublQ>x0omI!rnfM96@_YsWx9_j9{
zJrnvZo)GZF;KfjA5;$Bqb;ujiypk-E%pkZk=mP7r3Ce#e&$4enCOqfuQ(#|vWERX(
zY_22pGWc_4@egy=ytx`eQ<!wTq-7EE$EtFnrjuNro}#cimv9Zx_xEZ7?yNttd!K`>
z-i3?Cc_QgNZsQe`fVWYjcnOixWzlTXO>Zc5vtbRqMn|A=XfR*D`bv-3=(P-3)10SW
zV@qvid`&j_TBD54H!2^Rsg{&OLA5;b=#iwbWQ{UceRy|NzOGgG0J!~X`m_^mng*2_
zYpnewXK*RXs#zGpZ@cSXKU8a;+_I)_Y-y0|HG;Mmght6q6;L6uKyxxXkm>KQ_^r4K
z%BtSavd=v0+;>(3BCnyb4p5RFmmb0{<Q<$yR60CB3<|g)L~qk8*Szo)rIs`-tqvM7
z6V)mv;*PhvE?0b$`bn1zIExvbVTp8N2VLG1%p#B=*;tui{qK4Qu>F>iq%W*%LIh?G
zB{e`-p}c`8S(f|S!mA+{aCE!L;%-Vv!c5o*>#)6@-WRLi^C=%@*wR-8W(%b9-ni<;
zI>$T8YUo7KZNGaQENF`r=;nFm2&<eE1vLq>^tFglT`>4CIC-oMBzKbX(7d&u`8Hx`
zkfyfu`S;UVSwUO>t#zppA6~6uUadvjrHFTwzaU9-VvRREK*VRbaX!{ZJ7Z>gF}SWG
zN(iT9$}8~0gwt!O3~v*NkwdO+yGN;-%?qWxRfoPq9r@@++Kgu>eA{7jGY&F(F`num
z;VFKRmRYba)X}&0V0I#Z=*LcXg;OMZpQG`?K1nIOIl~J*zN?u!hA<3WA*cauzM{}F
zQ&b2&B#RWT+ppdfSR~mi6^<nWYi$Rd7~V4aG`<#Vb`)u|W<s4q5eS{2AxOjeDf{27
z?9O((zgXTLqwaO<3&OIs3;1(X+*`;Y21i`&Kw0yWxa$mx@FElm=MMQD@ZPym0gFB`
z$J>^>SlIaoi2P;8#7=vSeU6MaHoLn6gb>6*d8&WF-_Z)+GmsnpFs60v<YQwi6!jm)
zzWfcQ!27EhEG86h;{2C<WB30CTz_Fk&_AlhtDL?po+!`80t+@b7D8HRu0v<U7HDTO
z=1pH|PGS*3YO8+p5XUj1YN7Q>khN7`WE(<ay7oE5FzYeRxDz859d=Q;x?HuA#`m^M
zgZ>7lf5)PSF--)~nys4#^#yZ{z0&i$yVoDVt;nqmNHw;>bMt+;x`Bxvn2Dl|t2V>H
zMDLrr>z?&M?Q$&C7_S;OPkHS#IbSsgz94pA*r{EuLSdT>7;ot*?;Ky%55~gzyv(Dq
zcCy~h00eQz{^{w_*(bN<MW-xuTTy64ERKAUvhGETGwa+6#u$Ni&HgFpzT`KwbkXLp
zH(NEd`J+W<u~&u8!+sBg{;|}ryk*ODDk^kQ_wfSaL?2mc-v)YtAeYB(3|2>%p_Qf<
z=YS5zk3tF|;pTlR|LQY{QyLOXm$c(Mh&QAk_=x8)$a9WoZg4q2SU+I0^YSh-yn-P{
zSyEz6V=nj}!bpN`%cZ*EqR^#AFyj~~DnpcEAK&SF_Q1o@e4qH`yV24~#7)}g6Izy0
z>lx>aLTxsFonzYwk9$glT~O1X?;Wdac$Q>s<VKq$J`ja}6blTmGd43bqMJLH5A!%7
zENkzQx2&)>tld@2@gX(}HvEhc%OyTJ;CXJ=QYklV<5II%7fc!_1^7)_Ua@XJ-d)l)
zSJzl6yzIp2eCH^+p4Yjf;F{(*E1%^oD<6$GrTYaNrLW^42SQxfwf(dBDejwopI}iX
z&<OYWcr;lbnMpDCn@%eCqsvj{oO;Q2Gd^OUu;m`k3>&O3Uu-0IJWXU8Q)e8jl&=t7
zC>-egtN?D!V@k9HF}ET`{-C*<G^~&soni?(bHB=5D=<~Es0d4a@0I2^V%9R&`aWBA
ze5b)7!m`Xw0cDt%ll=@LJgcPdGE=HC6;w{{5N_{j{QYDfKZgAx<d@j5I%?S@Kfdy%
z+Gk9R&Mg}O69_t+>5%42CQrr&{(H~X^<;ZPPwM>H<7%X3<U}9Ksp14j=33Hq*z|)9
zoCCSM+^Sfo;8R{8%lDO_>6$6*6yDbFfjTSFQB$(-jNTglEniVjd-!Q=j#`eG@MB2V
zzSf;}M2Atm3j!ZBV=`6$RcnuY;w5k(Z%MP8f)^IYrbU1CIg4_;L@Euqk<jJM*T6Jd
z*z3}_pRd#_Z#i;SP5QI5q$TuGs+Z2xonK(D0(16(q7R14k?UVc-H3L#iq)80e>)1f
z7T=$(rO@DS`2AwmptPa)O)k&7G5nshxT?2oNwbaIPkr(kt(}6^P!CiCdRXb%Nlbgf
z<>DR2mjqiWEr`g%-$SE|=9!(WeKwoR3M7lZ?<WlztnTi2E1(t!zhsnr);WI<^%>Q6
znXk)`(3o~dn>g>%c1unAi12LJP!^nHhDZ~o0)%PR!z~L6;~Wy)cvGGm2yeger|Ymq
zf<N$@RY`BDC+&b5DtsTuyf}aF`m@@8k^nZ|ysOG&gb&vw!LQnU5`m|2X=~12sal<a
z4>4DKX3Oz#&0ed$L=Cxz97zAL$2T=*hf7xSn%y$YV+u1K7T!$Y1Ln9c$R7F-Wk?i5
znn<x)ojY%r&pIxvry+rjzX|2}=5N^lHtSo0l+_xL-d^6a@yzMy{<_GARX2wKiQ39_
zaba(pKG4qy)k9-aw<qsv@(qdV#>&;IUNTbBHp0&Mq&JU7pDp=_lBVx4*E!U#cstT2
zB*qcB;<7-RcFuJKKlfV;C=I=)?_}3xouHbX-Z7VDllwMRtrjlnai0G+v~EjVAz1GH
zAd8K%%=B-#vVeq!2C`nHsbPT8baML#lt)KAEh5CGW#n45&{WK^jN6a5bXzV~jHRm^
zyDq<&L0j%R<4GjYACA$YxaXweu*FiTG{RP$qV$YZtzas8{L^T}kd>pCN7T6l`Srk>
z2)Hi9WCelaLXkH)hTcksL+&)2Etnc|iJE@x8VeMt2{ewX@J2>|MRopIQ&t-oEy%{a
zf;xOnbpYkpjwHmAu&~KuIrQEEfuIMWbtF_OxWk9~M7sv&om_7^?8Y<4q<(i5^Q=}l
zS}XBQ_e(z9%#+L2P>Iu1WZp7Es+VwoPV=u#my64b{YL9`E~@)#Xl{V(+_I7L9P#ly
z!CS4X3O2qIX>FRY(!?VY-CNv<H&gObC;Irua-+{Q8PkttN5;tPEoP4=avFO0&97aL
z@HW?xzp^50PN+!U=8Fw*1+S~)T|e`iwR9up*$z$mYFvS{)00>qt{LvCTf<2_i)g-6
z>`+E5oi~%9iQR<~rB9PB8je>@N`Hs$<yNj}=nD^RRH1G1Qmf|~d<fww|5(eQb?|wk
z`zrHOB){Hnr(DgX%>_C9eQn3Hmtty$E;;fKWnR(BogrjxfT~F_Y>>}{k%5L;7I%H)
zL8dd|y@SR(J5k93)oJ<!T&ct%?=xJ7ivzX>gPl4hs7I|DBH}plj!UZ(1r=)*V>4k|
zU+H%n9hwEtad({sg1j_th1(UUVuH)M4Zg-z17kJA&8O-z<8|+QHzbfCnk1HD8-lH?
zn1I7v5gD~;r;L`iKB-B?w3MkK(Jx7n-4@*Rb~L_&dpPJYIIuSgf&!&EHmjGv4^}H3
zMwfJZYiwYwDO`QXo;$x}4N(h<ix72qb~m2{bhur1#GUGlfJMr?iHcDake~+^^Ovjo
zM|(ta()%}l9Lj-Ou9NcXkOAN^t;*eE7a4>thkB426?+KG#llVAJ`cm`SigIoaC(!d
zPjQS4(>*pCdr}Gsck@Ch6xF+!%u!UR6ryZZ_8w7r5K}tnv4EzkBOGUH7G7`j$-9sm
zr#luwK_FTm$q>Wq$k20z3&R1hgNILB$4;$@;#kDlvd*B{jIutZpEzSNFh)*cIYqo%
zW+cC|KjSUeXtzjrA=%UW5kGb#*DbQzmwz}HKE!TmhEe#<zd^zlHO28^?pDT4rO3FA
z6paFfqM<@QIq$DUC-u=tZ*%k7sGbBp8fV<K0HTKj3R->9dM?@|RiF==J6s)RH75v;
zYQp^5{<U0PflV73%Yd)TQ5*=B07>2ReW8*3719}5c7r$5RR53|0gPvL*i99NO`Y`<
zLtQQAI@1EoZtoh$mWfz?dOIuGt6#$jqNhvEkY{1P$9@jmfX6Nh<a|5%sy-Z-B=P1U
zzuKqtl0GlN<oF=34Cst@_S(3Z0cY|nPD0M{m2>DATBo`n#2Tia_dAa-tSY7CcS@)W
z7T8urY2W6Jl@)?AE^sU79u@%11oQ;-7uG1u#_4CNHA7LwaaUPtRZg^a)#2CF5aANf
z@8aOh3$ow+>ehN_d6?WFOIa!n0nwuJmxnL{LxF6XlyA>_0_gkV^xywpm5duniG6Zv
zz#_2C%~w;qZ6}*;trnNV!wU%TL>u@$X)<e@*<EZgKYAamP03<0e$^u{w$)HV2TX<M
z+u?DR8=^KbEBsR7wn?iYFD`8p*kh;E$+d;OTxFAQLAkBg;u1PP`CjxZM6o8N%H|>8
zGqDa!Nf}B>>Cr%FtN|_mYW(z{=mMa)ymeJH<%_3>z$3E95{eh|;@~&CY~R0tWRQkL
z$WNz?su%Si`9<5eQ^nq1GY|AFW~?1K?C?)(2iJ}>o8+6iK4jr24yRmK1eQyOFJ&k|
zu!{#C^>k`)8xBXmNy*MTu;NXc`y_oQl5$vyT|z&rTG+?FV^*q2;|Z{@HzOM^{@lI(
z$RnndyD83GJ$ZfCwGOY0b<)PXy~;hj9S@&CCc0+6akrZ42MpWRFFsj%ogPrWl53$w
zEwZ=v+!D%YU)*z<-OeHt`>rE`p(xDFHc^6Skv){my(qTeEek@W(-sw1UDStIUY!V;
zdRw&%wePIlxzZad+}!goc2ot`F(>`3KF@fVLJ6gVRHD8~4QS9jS06Pb8cOCc1Z@w6
zzCI4!I*z^0oKqzeJNc;|qp#kLv&{JR5aa6;`O0dkksHdeb2MdVK)J~dO!@<Obc_oM
zti0&aI+HML=!3(mK0p8V`g{fZIAkdczlYb#f0TrME{*r%&~=$Wtk9UQIx@b%2Aki>
zUGgK$0<NNf<|$=^#R#vbL+TTXBA-Zq+RZ1ki2oIvuk!Cp0!jG@Gv=*2)Cb%#+HEU-
z$2+06-fL7fQCx=&)hlV_sa>hXQtAQB4fsZt^e}d-%d5*MAP1Jm(E)*O1n(d|6V)(@
zHM=h&75Sd2+83pj`kl_58P~DX{$&KDKA5D`EUTi%#vpk|eRLN5cnyE}gqLY?VhCL#
zpZ}YJcebjuEzg<)+A#}$&5yci`a*aKfpYDdZ#Uf;9&2lH5I*-a`k|leW?Xt^b$EjJ
z<H|!r(bA5&9Kwp~)v!aClD^#BI(}#^-d*G53-+br%yB(d<|#n!i#%Jn<us^Vo@cz(
z5=<|jRol(orhI<b*jm;^u)n1KIydqso`hdtjM4}S=jnZfc>Fg2;4={X1}B!0hiSQ-
zbRxzn0WgdeD`V*!3QKDK>L-Kq^Z^gks2bB~>}FR}2rA$WAWpI@NU~gv{<^uvc^V%$
z44MGEjis?RmoogS2Q9(DN;k!c3fbyyu1r9aaj=uBDAaaYaJge}xs})l{C|ZJ|MDF5
zx9Ow|Z79=n$4-~)aRwBt*Lmt?%jj$dr<Kosd*%$z5zmS}rEC&;U)C6gd|f*?HE%gk
zibz57Fg>MQoZv2e#An1aI#k_kes{gzBj>`n({!)a|5~%Kd~|oWi|}R^-IAqb*h?Xq
zeMWtEE~0Tvcv?r-CL5|lA^-<_405*zq{{(@|Cc5vDLnvi7;*tpVRrgk|M@(Q-7c1}
z6=*My-DF~x^808H&Z`C=-(mGMIA^(VI=rAyno(td@iKucCt?!r92j1SWd(})ohW<L
zx`@$7#z2V$z27wZ3(u+CNKSYcjTPFYZ1@QtqUt>$9|Z{@-3)NmtS4Sa0L_+M5Lq))
zsbe%V6?5@u1IO1m6ScgcvoR<Qi^GavD*pmBd<~6hZB)k`ALEE(<2naRerH2@fTKly
zSB)Z9ZNDn5ux?N8Xe3n5^|g1$#3CceJ1aiA71>7&+`00rJYv%b7CP--6Nj#CZPSkH
zyDhg)lu^=oW5yDei_U2dxu4^rt{$bh?b5W|tO`0(7QzlE5Qx_5%k%4xLjDm_abgBU
z_^FmqjX5k?d@FRI27B=rRFacLZ~nI~a5#U7eBr#ImMZX6V(;x0Sp$s4bJhBN{4bda
zMl!rooVWIIIS@Yu*P6%z{CxUa;!xM=Z;Cskjk4ayeMVt+0QIZAPOc8e=LQY;1V`kx
zhWh;czDjrDw=QlU@pM)_x~)0S@4dztEl%C?Iea3sU3K)6GjIEsivWProxffHoWGk8
z3(OW^Z6IiCfi)HAx1n!$-aIy!1CdKuOqxp>RdA)f{fE4p$=fG%fzHwMHOkt_II4Lu
z@3HoT<?@R{W$_gbzrlHS*W5fP7RtmNOJg_mei(;U!mo`KKxFHpsWH3UIzYb16t-8W
zzUK3I-r&uQRKwnLquG&(C9LpK6rrK}GI|{F4S*Ggas~0PukIe59Rp<U@PQ!1e>2-P
zi2yYe)*!`fl)rx@`JKB*aus*+k>Vpeb6yQlmV}W92xK~f8mHr^l8{(SW>W#dcL863
zVt9I3Vp6QTjGj04X8U4N>{bDMTi8QfMgNw6-9LbPoL7>8dgY;QPxO(VH{t`Tr^`26
zM&>!B0X#3s0t1be0e^}AP&=S`t7xT$ukjI_)T%~G3~Sq91g=+BQly*-sx5$cRvwbF
zRbvZbX3DpXaTXyCszoH!RdC=tIAy<N9#s8w@FWRs3$k}jjZCxpO6rQNeoip9FLvJP
z_vR0P&iN%zw)9&`!>N!exc^-}Xv3ymueoZf_fLCJrw6-i(6VwdSiaCB;A3AVioU_v
zyM5{=mA>9Y1ogg)NWW7XgqGN@`mK~JqR!FGQG|5k;E<1!v^o7yH_fO^ChOvmIvzw*
zQrTTApLBiQ#EV%Yvh#YCCg{dlrOfVC9-NZ%UEgDN*JqhNtv>~CreR~kjxZY-LJDP2
z2tx^xE><r8`cVTES8~j+VuBDCFGTGh4j}^#aArUSbjC-BD%Z}0;$O>Aq13xZkt+r-
zk*kTI)LP(N;)QU9f=#~$i#ThPq(-+uSfcl!z%)tIjCm`IVr71<-)KquuwvxJg8|TT
zlvh;Ekafbn?m>?a?Au?0>I_9|N=tCtH(_!xcaLm277nOxhSH%28TC(6lXj)+pc-~e
zUb!LUjE?&w{W*1brr+Ck|Calmhzs}qb7Zyr&GCB1#AoAZ>qf4J{GuY2^J0N8oARtM
zzjxrCRda3N?ohMuRr_jB8}<#s?x)vp!ZI@oi;x~heb9v@tT>tvL4<qVQ|x<lPxi|6
z`1;MiZ<x-V_S?F04)3GYal#6ZbBBX9oa3!yR}ulI!|4a)6vuCmz``DO*>;=MlPA-S
z=hNFxPgMl)kFjEAuN|^v9smGj<&N2ofVheGqfmxGD(Kt9nvpBlslz?&sN(K+=etGp
z{D~*N{a*dS>e(6&Ln`A34x`O3+IEM1UFiWld6s}~%2UCoLsdS(^nFdKS!5G^LLcTf
z(vA$y3k`2sQ=zV&*AM&7#o%*D1`*D^W986j8oP3|n!}u!|J|@_dH=?V=%r?^Ri!2(
zOeFPR5A*+7!X-tk5o)Zi+^+3)VdE<?!uWpK6K&L+UoD22pg>eTgbMoU+J`2^8J8oS
z!sG11ehQhYtMoR-!=_Md2s&EM@IeD<duE)6?)2#B7^s;jib#I=dcumnWnMkB3iPPe
zkg)&G6wslvz#o0N<f3JPCElKS<D?{mI7ArN_v#)1aqNG4bN5x>`=9QJNo|)~2B*m|
ze-nT&;r9RwyZ=l$Fe=_*0P1f)arF35m}9%`I~>M7g)Z)QfUIrzz)z9CEbq=krkv((
zWivY#2y<55jY;F>ys<3otokGY*clX}v(K^c2eH?D){X|CC^RNzb9ef#rp9D=vokoB
z#PtqK(!`*4j9h3_{3o$#Ute$GZsTZMM6K-M;tBh?XzKRVkkZ_|xN!o0%T%GqKxuEL
zmi#dXNVG*<J1d#0qc7_bbn*+&aNkhzFPwlRB&go1GX~fm%lLx57XSzkZ8cmCwkdyT
zH1m<=?A+6B?SJ>Rz4I!f`r{o{Mx8rv0Jti+#_9jOd#oL^3^aYGGKGfrUEq*#g{BK*
zrVZZ4+QE)TSMTf}U(g59P&Qyby()}BZjNt%SC0TmC~}6`^{$;@ys`Y;Wjn1dV3V($
zj#<;D$UdGORo!uTWb6E6T-NXZaf$yQNB=SFY)o3>_BbZ;um9N6EFWq7G^{o5I`-}K
z+cy9J_0Q@1mG4=uWQq<ZVsrv-UXgBa<qJ3pn(Uyay{c?P=`Urh2v#CHgVn<*R<iBe
zOMBB%rn2Xs7wU4|aYDI&`~lP=xb`Zff3i+a2M>k((3-ZX^{{og@5t`_;b|1tsM3Nf
zmTzRe=E+^`LJ`Wwyt>dU<wNybrIi*x|A4mj#wxcCU{4}AIQq)yMK8flO1MLKgsyj|
zB*(IWBtsk(srk0RG?Y|oF-xM>QLvQ=!=L-^rL>`Ez`Bns`wxBAlou@-z50ph34ZVI
znTJX+OWc<bo@TJj*n1vz?6zpEsN45pb9SH7luC0g_yKZ30{$XO?Sth;Xv8M~mVaE|
zMYs@+{uEg@o93o;(qhE^O|U}Ef@SYlOmRQ&b!~H^svCM95!gq+*!R~H<4{IZWi8Qo
z9+Zc4J<MfdLY=Jd!u+zEDJZz0Q)AGT2g?JP<#GO`uGzZBfRfM{Hz@DKAbO2jQ)2?%
zF!%cP+c%RO(tlO);@_C|&Fp&2NW)J8JGEQ|Tx2D*3p{tdqu#k_1EYE*{&SUN>Nagk
zTl`U$^Ibp8Q7pu6g+kq;G0_v<8p@6x=3)n9G;Tmqp9>;59Neu@yZfE<ig71E#p#;+
z-1w;)A3<mhYUmS*&k2oYO)RgbvHZt4-kqK0A>^@ldw*DOw7y773m{{TB|A3u2piR@
z%3rCUlT$E@_RyiN@p=o9gS?_S@_;2i(6~YR(ipxUUE!t3kAtbE@;#d(crWKXfMIqb
z9f(kycsq!rX%JE20Y=R`B3$NeBP=Ahhj(JWv}0#}WoBu^=2f6C1DF$=*e7c%FE-&Z
zZ}~w`UA1X_pQNAyR9;G4Ltjs>ezhxbZhX5We6=yr>xTHFG|9LBr)K|>u}0~gnG|v~
zoQL~xD)g>IE|7`0hP``z!3r5lL858@mG^`8;t8{y3$sNbv3zSa@)M)gQK)O8!B@Ww
zrsZkF0Dr<IoYVbd!QwIrFzcp|dgqVog*#K8yBGVdo{w;&;V9)3wUb!yA$6wZ)xR`E
zKbdGI2Y2Rk%gGE)6n<~3)zIf~*sl8l32og7JEpNa#43MNX?j5xU1L0ee*ig*dlBbv
z_`Pyl2d}b~8x8|@nvFYfoQ|+R6R%*@e*l(}tE1&tVE^wV;Vo%St(^m3;62<^u@{6`
zZa$o1M=^Z&;|5YqncgcNZg5t|oh^<X%+CO`S;6^@FihYByqqJ<<s<jyz?|n}*H399
zkP{oCs3>HfGUD_NKUH*q!12=CF(#^=zU!<%fMH@u<(e_sGo@FTK>yZCoT4W6i9CkQ
zo<9Hy<zj3jslxwjP5&qvj3q_D;nx(HZ|6Q?n$?;2EKF|E1dOfh9|x@X7Bk*FzQMm}
zOEMx{$(PiK3|z~M1D;0`MCo&s8D5a>r8}KST($<`YNz@2Hn_C@CfmAwOhH|YToOfL
z^M;+?_7TGTZ?jzX@~|1N0?l#NvkGT?|LT}QvfZ0+WU8Lrsrp-@qfxh^S=t95ZMn%A
zwxs@Q@LxA5bm`~i$~24hf1z<XUB;fZuf1i1tygFL*ADq-;pQyyxBqO=U6c{qKWhDj
ztLjylHCa9@CI0(SW6a3iG0#VzuL+vRV#cdZm~TI@KtErw>uNI5_}6=%FrD8%9V%<~
zCoE39q2dnHv#r27HuE3RFg{`}z?18T(HT$h9qlDG&$Llo71CdD#h@=2aA#S?Hkuh_
z?)d+<LX4GTaT8DiC%XBalrFTnI+{!tJ;WXD4VCltq3YNkuGwBhA9^0YpgFf_7W^OV
zy=7EeUAr!v8f~FKafcQy?rtp}+=2uv?ouE?(YCmkAVCTghXjY<l;T<_790u`4es8P
z=h^$+<JmvXH}?09{o{OR?EG76&NXda*UXyNyytzNw%H5~+`KRIoGNY&3<7%J<2O$5
z;##<LRsk<cg?w309J;Wt;O4F(qtgPko_VO<NeDB~o#WHl4MsJ3k!Bj>Ocn8OEIz6q
zmux-1wr$DZ=70(MUZ62f<<@46zdKj&T-go=?q<05#X}fixOYKxwzMYcWbep00k*3F
zHxiQE*=aKHghtn`g`v~s$-#IZ{CKW<@07UW_Pddl@t+E>wfiR-j#vcuf4w&1e#Dun
zEo~H9WR8Ja_t8BqM7w{RMPdmCUh+U+DV_V<oOMNmLtnS6qQjLHm*4@d0Bh}TsqNlv
z@7wdS*CvW?ry9lQdcs#M*ZcW6#Vz~Gr<VkK`Oc6noH6hf1cTPrz)-2}o~hlFBiP?H
z2<m4;=SEz=o@)y)z@$bn;mGF39)Nu1Pk&{)Xzi7{NIz@v_Bir?LD_ogZhYM&BRLs+
zZkl`d9M|kG`X&AaIhozHljHl_+#?pJfo6mO|H33zZUw&CU*-NgQJP`HTbus?AvY%R
zZx&&={8zhz|1%#U3>ZGnU1RvzJAmH!UW6eHWSe{Wo?sB>IJNMeVpG-K&c8?l1ToNE
z7Uf5L(_|CWJ0EN%M*K%v73Uf2|50gi_ewUwztD?8Y(W#+?q6-Vf98KH_1mw6D!h|R
zYCiBek&caHHg}xmg3|2PH&fG!{SWNIwrjAgAOJ+E`<hRhIs6Zx_JQ`C>o3?9!Ct9r
z(<#}(hxE3!pp0`C%rlPPOx4@=KR9q!Y!(-{xo1OiVS+Hs1i^Wyf7#})TtVDsyZN*2
z?1RG}K=PgIJdF*@6<7FT?j6BBh~EwL?iFIPxA`5;%|r`YW5DL29z0+tn|E)Ae*GxO
zb7}a2ewe?920Xd24gZ$r@ol>5tyn67_JZ-Z^cyKkiCFxr{Or_*?CzmQ%`HD$K-1{~
zO&ZtqtB1Z3^0yZkf!%UVPG$lcQEuD4SkCUk3e`*7<A`kA`zyxR-}HmRD=LgXr*Zia
zAx!7qdl#PaEmUV|TQ3I+lyDJVQ9GxgLVWjF7D+yDF%G5$Hnz?jG~o2ML?Z)JP8Lv|
ztm#`rPZSDfO`!lmuCbHP_e#8A`5$mqYFilr4wwbPiwBss0uVq6v-ed3Of;*Ey~g^V
z%>O?h4Q##qJ?_W=c+)k;drS+GJ+LO#V`qrzE>;EhJ>b&`jA5TZc+KhNGPw3X1=V=c
zO!h;3T;Nfdw%l=15pBfjp5%M4XXR^mNAylJ^`SS!WS6mj>x5)Dzp`mWt{Qxi<Yp6T
zj@IbxvX9CQX<6ZuBgJ~ZBVcOR!hfAYASzF{?^N@-^pivG`<3m&yaD)O1)XtU0xcg7
z4`%5PZh!c<h+M>EAO2fEp8gLi>@oEpfU((`^z(ZotGL+!J!71#^rI=5ZR+(k7Mb9T
zpL5-!6zgEbi|ELNGden$>F^=QU0$6PHZcDlQ1;})i>&_04QoM1ihXK76(g2xb(;8x
z9-lq+-?d(EQ-AUCH2`B!GIW9-uWb$k>02<Jn|(Td=(ss$@~A$u%nelDLUf?ikal$O
zEg{25IIzGtLFgv%d$43cgO}UR@<^>c;eu_i6nnszdz-K<^5q@|W<uWk->Y4dbaYp?
z`dXaD12<G3Nn}nX;|dOV9z_-D92kq-Qj|#$+sFj>f%PO|Z?%UbEoOzf(-efK7`KEn
z+mz~A!<iPI_6|Ryq<$54oKAXOIHJ^w^IkCrAo*X774u(weRRBCE#w_L7V(PFk#+Le
z#}-R4wg3{6;=LBneh<q+#`@(lSgMh_QcOto7BnVh7RdbC2u{bRuIAqHX_TDXIzPHN
zSDnjW<8B0vzE8r!d}*OU*zU(}ww*Jie?A^_&$HSu#p<X%`bFF9C5_p|ulvi^_?*DC
zV7)g%$dSMl00J#PX!gsY_r_p<>-A$)bjry0$@q(^rLhl0r+I7j2clMioZcs2IVZi~
ziKqG2%?ILEcU*oaU&&0_*MZh7oNIR;bs8&ELCW`?SkNki;lF(QZ@`$MK<?$rdH=le
z=Coj~^FY$-?yuq={4?b-3I@afewuiJ#MFvk#?3|d5zTVw*j8}QrA0Tu{=M4`zM*uq
z$?dBrQnf7&fsaeGyUXvA0RKz@T#Z#Pm$zxnv6!6-VoIqusUI2q_0R10(cS(N|9s@p
zRMcfL=CP%z;5scBMD+L7fZkV3Bdwbek2$LM51&3h90Vw$lK*O2@<ro5rchHhjej<k
z4L+CB{?A^$xEecJzM!oRB5l6Xy!X$P|9yW6PV?8=FoSyd2XKJtGJpZCH0b@ytN+IM
z{IB<skh1@MH}UW4@yi%Emp;z19C|Dj0O-K#Mt=YgQ2Oavm;Tz{>oIRbs@A`rPJ*@u
z-|X%XYDPRfOj}82%Jeu%6-jqFY`ADQuCnswA!nAx(XegFueH1i4-oMmfS=ZjM@TC)
z2V&gsF(?0c6WokM^#f^gqH1M@$qo(j;!G6GWwNHn05H^3ac{}}11R0CHV(>VUt+Qf
zF2}gZv^vsA?p*Paba7nZTH%hW0|Qtk64o$|!G;5ujx`!pRw~Hs?n_m3lm$L{C8gn6
z9xF+^|AEesR__nsp#XP}x5mY5Kz3;&d(bCr{Hy{`3SYdoulIBNh4gVWE=T6(MIFU{
zOUE_6#g1m8FyTJVs=GNVCptkJNoovWkjpl(74&7dY9p{Ih{iVce(mpO_N}$xlnT5_
zJ=I$lJ$`^ON4j;SM5X)q?GKDSP|gRgA{J3Ajmu#$%WbH}#UKjMU?@4r_x|ld&W1y7
z5Y6lFWH_%x?YNIKYWxoivwKDEF(s+ik#W$}NWWK>j<dU8QAWjCfC0SOJ_a2__ZNIB
z(*F}FFaY3NxS+uRp!G@<llvb~F7tkOqVe^g03kB;sYw0j4V(&+W37Lp=D&z1qY9D{
zSKfa<$aL;P2R1d9et`O$nt*>20q|)+-DoiKEe2J8@eOk%>*@bvBmN>f=K1>5U%OE2
zyCMnG>g1Apk5S1fd!^LXe*iv8(s2g&rD{w6B4$N5W=eU~bZEvG_-{A<X6nO+e^&5c
z<b~<nzfBAO-b_?Y!_~`w2YsF5%9O@x5|)xg{#wK3ueHzqZM%-GW*0w6(NmnkXvIPM
zpf3vv7gggIYk;Q0Ze3Q1+_lonUz2fxU`q_cnhv!S2S~vhS<eiL)(S*|jy{utX5osc
z;1rLIPjv*B_Sn*KuZ`NA&nir*p1Mob+*D^@4j7n4VuIrV9-}^{xPC`Qt`~58V|{_6
z(vLEvzIGz7_Hh?@lQeT>l%5E1GwoN`%Vtn%;B}w`4Sj}>GyMUqJ8bT+eSOe3XQ=o+
zgZYs%%)&@;jxf^zp6OnKJ=jJ4rY=vTl#Lqf3+FQ`oF6)Yg0P82dTJyWqr~N#;Ska6
ziXv$6Km}(*V9nG*ixKc}+=Hshkg2zM$q-hosdL(37{N*2bQmUWo=0ABKGc#WYM7+B
z!9alV1A~6?E*VA%WH*@3jy+T<PKZogqCZyu$ZsZ^+S1MhOBrt$>5jAi1Mt=)>8#P`
zOwKL^NVKBw{4cgmRb~^8W^DlRWt)Ehxa7N+&X}N%4`mka9G-Ht+W(a!&<K+QEC(pA
zzL3B3PhAf*T7(^e^8pQC0KlMZEaB|aDWm}A?9tTOHUM;bUt`&iDu~m*?bmn6<VXDp
z0_o6sN9l8WwM)(q54N9VcdQHEfpvn~|36(k40t_w&>EYak~DH){HXb6BEFa~2xofk
z(h=ja4hukVdXKvkOE?fbF5T{SsEbZzy;>SP=!~OFnHV`R;9QzAv3YW}6n)SeC!I1;
zbFusO0UF~%?eMS5+PsThS%CKY0Dg?KEhi6dCbSY}7=nk0vrLEn07mj?RQi3j0*zjO
zCpc|I9E`^;r6g$_NZbRiEi2vy_#Bw3pd!p}Z9j0bbJ>*pEX0TY{`RV1<hJ(%C#@0k
z4?z0acQWX89@!s&;V*1_P9;f5@3VH=wwpfyZ55PWB!Dx^&_G|pW#}a)J_O;tAV0_F
z_s6Q+2#gP#dG4HmQD!97m^sYyZQDp2PxxAw@_Wqg4IDEwTE|btrhSsW$(7-bn&Nf=
zKTf4H;&AOWecRT=p>gmQa5ZBYNO}C|b~Kl+*XhIGX*PeQNguqGQ*kwWu88WVJD5wA
z-iyRpLv}wvTg6|v4AGs-eh8w02UY^A2m0N6ZaRWI#k?<G0kX@|Dmuf=rJG2^Z-@a>
zbvA8y6d`w;x=RPD_vnn+&BCcW{s27L?EUpp@10g7FWwDtpUV{f0epJ<^lB#KMr57n
z`s>&qKsz^P9;vemq^`la{mE+f>%;xPI!`xDynz&Rk)Y#00O08#KrG&i$Ee}>&J=TL
z?k(xRvxEdvdpy1jQpUWL*Z5^Qbm7$PcV=U4?av?HI&<Dp9X|}Zd|hyp^7-CrEfUiS
zvhz28RsG>H>gUvQGmqH<R^}Zx;Kk8!`*F{FJLX^bF#PciDNX=pi1O|zkM9Rv%7q?*
ze?9q+5wi#i24Dx)i7bW9V+O!3=-2mqr&T>SLYQ9U$^C^j{IWRV<$9j%1xl6+9zcN8
zyEfcL%y5iKw|{t;V#q*td1mwMrXltjh8WWT4b&jj!Az=DP>4Zb{8cJI*U&c3xT=N+
zGifA=m~ocB#U4+o(<!>SI=nf#E`3vk7%wY*i_;CmEMiVAXD(f5?Qgt_A*f$^hLATw
z*-{d8DOUx5bq_n5YjesBAtX+R3EH!@c@D?CrDeJ>W5`WIVbG_qPp=lWE-ZUxF}xR8
zf=*%B*_wCI9cKn3UOF(o#{}ylyF6Hht#ymvkOA}zgD`By6QN5vFA562H)YhWKk-O9
z^v30;4s-ni``cIG4cDsqA3!tMyB`B4O@TYIcQr?-?7L?lSpK)Xrg26O#TQcs;WW%#
zI%47@W6AcGzX<-$NO!rn>Y9F5gb|U9WQ@|U0`6%{+@lM<alNVkc=s&oG3qY9<4*X_
z7-LF+^_t+ct>~Z~!{U?Gi{0QCk1EQxT2=2#4K})lMsEDaPgZ;nxc#VcDx(wJUOX~0
znw5Sg^4n~0CF5vz9gu%?6pUfC1cuSam^tFncj6S2pJp3_aYbw2O4@!^M!^<K<{zW>
z;yane?o4z7RSYyWPzQfq1sacQtRg?$9~@9g;ae>UYK}4+bV$8FTfLop+!N!&jdq1w
z`1x_loYCVop<AdSYsmWN`LsfomrV8oS-XD#Y79RI)qtj_&&I4LtgJ{c=ILZR{2A(b
zKE)drXtI$pS<gkSAr_FcKV}TX7Ho=)%i6|=lWy{RLOSgPU2;+nb9n|i^P!ki01vj`
z{-;$&LEDt^T}?$xFXa-~laj?c_oA_1#TfSliIRbVHAfMVu7$t%?y=+~1VcBAu`kmh
z^iy#KhYq||G}cDe))&*n=`j}+_Hz?~@A4g(?cgoKMIZjQmjh<Cs1k5O@8=(CLEcMz
zNPA~`VDg{p{lDpg+T-?J(Dcn~?7s2B@QOLdo3_^jc@>^@gMmwK3QgXa?eK=6S9NVn
zvvtPbF;+}d;a=fKkD&8T^&6g-pPtpCxrLw0e_K23Bj_{v3ZQiC<97ZEVB-`)Py0IV
zj#Og&WURpIP2VEILd!$I5RZv?4Kr`r@n8Izs4395+Z5flbdSMLHuus99OOsBQ(M2i
zBOuWmXePa5pZZC>?@!1niQJ}~yj8B>o@-b7`QJV6tb$C;as>nwM-wL#fF;ERmUH>z
z4^*4i`FES&+#Uqbd1fz=4Fg>B_7bmiFTdlS1dIiFnu*-|7T=_?7rg-x9z6%%j{5(?
zx{}(GZsPv(KSlh{Sp1)hi+kezQp*35jPJep@8;6~+wuzN${X)s$Ir?QgWgLTBaR-)
zz6JoEZtvq_4sR}aj*qVs_<+xk4j)7wt^^cv0aJ#b+&kDmx^LSp-VVTym(lmm%l>T5
zRkpEF(-!gKWP3`()rG#`eD>D&{Jz#%_r^uc|HZcA|C+O(t`=s3kN1h#m&E66xr?ba
zV0rOHm3(U7-oJkbI9;yLvU}>Z+v@U!m@r7V>*=vs1hoKq%1kI*#5T~h9)R4Q8p3hD
z`<l2`WZIN=6TURdm*LuklDjP80(_c6AR-XX*GCVraYA!anh3V~qSubH5w-8iMx7xl
zyP!*NmI{FS*7>`QPS|HPtBsVQv-4n_W=)}})j-vER9HJ4w4L+ZE%?uORfadpbb0C9
zytKfdGHm4R4wrSXXYwYS6s@U)jpEdp!EYENvH5ui(&n;F0^~7q$wji?&Uj?Ds1EOu
zSvW65d8wsY&65X3@s|?IRc%iWAenP0Ihk>liZ};OM0Cg#TKRK~7-#Z+xk8b8A-cMz
z-(YM7)jBL+F>v~C{7Fk}QpwdEr{fl+sD(4|7%ZvRlz*AhD%D*MBb=W^gF}@%x#3@Y
zNTTG}IJ2}QTsSEQqXh>t+d=VCxtivx2+#)Slcd9-{H8yEYUvuT_<+PJ8h4?%N<Z{O
z>(fd2ZcjsOOuRvC)ViCK!5Bn(2u_4O^L`6{V;?kVwU&wfkhNiK%$;x-6pHIB%aZQ|
zF&v|gK{!Gf6H~kS#s*cO?$w61XS%b~mEuo0<o#c{I|sQEGS@8NRzy;uZAFFde(V{c
zg#vXwSRv+kOg2$2>_(|%QOJ4WSeYJU#d3YCC0zu4z0hR5Lf-4oCj$j)vvWKNSeZ{@
zj6=}Eyvk6K6~dqg0=Vp9wzUhntQ-!rDhK|H&`Q;Btj@k1E!Yf|H%^M`;Rep80&~j)
z9fhUAdHFlQh|j<HjA}jgkrCzUiz`zjG9A-#oSZh=1$g1Q0g~1Pd+ER)pn_-hpb(`P
z!|GYzaDS;|*2WBF6x<;o!7tu%68+^_6If)~be0wWiOVzHI@Z)kwoO+$JE0RluxjgU
z1~O_=`+>lf6>&}XrTe0DSmMDK!{^uGyG$PfLbhhH#=6epf8S;q##+09EQBZmDJy<m
zMjbose0&ee{|<%sWJHU_W7eToEiB#oK{z0F_}i$S6ibU8N$e^ziZU)HZNr*=NfvFZ
z^g#qh2Ly(9!<>XV>JQoHW961C?SAB3&_##@KwH{PMSZ(CMu&SGa5E@PKH~lCx{+;k
zQAA@m6erBfGe-hNNyK8ZJr>go_z4P{)QHL{TwdBp@n5qky>%$6<%+fvNEwnQx<53v
zQbWj3qL`T6Ih>#jlHXO%@47J{bSt=Bt0bYJWv7@uvk$PAGyP$YB?vadnp)LFy<_|N
zd$yWg!%8W!glzFe?sfAE<S6ys4>M0o5Cl?hjM}%6w{_TKw1wT<DIJ{8cvI5w(zw`k
zcS*<9vX|X-%0hyJuhl_U7CwbYI%Cx&T>F6+H33qAJiJ-!i9%I;8`D6$)-gXiW6r-8
z-r~#4O;fwq*?TnB0C77$fOtRsv7*ow8RK?B@dq%-v^6OjIUNty1)iiSJvS{&Gz@Pg
zq%!@DE{}DX)lTb~cqQ_QnHxfL$f6Bl(@rMWR-Pyv7Q=VB_(}1!IPvKt!}8zqk;Ree
zxm5R@4LUfiJ8oiE7SPFG-u!@QP?dl<gj~jhEi6c1EWpI8;36>l@pAs6pT*Dm5UR3X
z^YEG)5(*@-ol>CAx+>o?Y;eTF*E9w#HAYUFS#V~q)No&-kPAG>TRSPGg4@KwL1!GR
zVt5(##s2*(S7ihvN{hb7;+sdMm4K%rFJixpRaM{GO-)aV3LNeDEK(h4UC<dz>J6G?
zRT%P8{V>(7dk%@5bwE1fo9O*$N|l+R;&;Ge!+4aasWFGx?!^Xcmk!=p>p^bL_`!NB
z>^!gV`o8AM+iVup107Tn`&lD8;agT>R&x&>e+E#TZMB>B3TW6i{k{vc_N2j~k?kTk
zfnp@LQclT;fF4e*nBx4Wz90pgch)Eq{lzbR$~_%TU3&HWV@)h@doD)`<@T;aUZWCr
z7OBk=z2ssU=xzoV;(_7ubHk@}6?b(}Xha=-39Pj$)j4(N<-mQep|2LqE(n7<pE9el
zNgRA`ob7{#0JLiGyYlM8F=w!2vMYh}#mPz{y9SB1vQ?Jk1{Dp#*`qwWPt8T)N?UcL
zPoqDz<%mS_@zV{DGrp4kBAb?<|0&<MwyU(LjZyfAU$eOB<d0n>WpP}6U2U6SBohx*
zzGE0JRB~CG=&Be>%dylfMm^$`4&8|)s&3*pfk3n%E+q*u3sLA<V2o&4i?jdlMT*^=
zWhaRz&5lX$eG(Z+Ij(SELq;38=Gu9^a8u%taauT6q;?~9{G-If4GSdiS@hS~7vA`X
zIPG`+p14mnSv*NNb0S1Kn4z2Q<HMtCmbe<IDST<%m5apK)L_M8H8r7B<P(8{c!Gnm
z`>OdFv?qSc$VKGONfo8URJc1WM<5GC@Rg2hNOGF}CSTT7;>1Kqspk;68%o+@D^?*F
zfuiizRbfaL->91%`Voh_HizSQ$=&sAS^phYQ7ljKxrCEI!~Wnv<z#v@?MeFe@mIsz
zWL*Xw<$AoB*+L*Uk!leGPu^CR6(r+e=j*2Ub9*C7<*jwR_&oDeG0D`_d|{=Ulc)ez
z_L4}R6A?!y(i%ZluB-61LzaZfWy#SLLmTRt3CBqZv>K65xd#5&iaD7cmBjWS1y8L3
zKX0Mr_<Rxd-efh#=je;GAHm+D1Wi|wesKCz74sXDuMLCI4RgxF1sj=HBt;6HQJgYE
zenf>U0cje6E57Rv0zD<reK8q&)zozRGIhl;+n3KARxkn08EE%!6x@oc&uG6liN}8h
z1`xED4q42GupMo#Bf{ZCXjNyL{SV2cT1AB`ok`~i4q*kym!*#Mq9zF|v&6-gGXqHb
z)YgmGvC^V9p1FE!8!tL`QPu>vQ9oqDrsQ@o2O=rqyT+3o0TnKw=1gj`nxlPo#wnBs
z`=syfV8pw_d*tTeF7Yur2PbS3_}V>?S80>-;)g+K0SKEa@x(>u=as>QnwX^bg`4?^
z?U1bL7((Fou|h(=wLLj#+Fl?*D_7^MkORk>Gn<Ms#KA=?!IFbi*>6?_F^^(=6cG_M
zX*M(Nw8sZc>O==G89Nyy3O>xbfj*qDEfwlWHlRI#R<5ucl~#!h`qHzZXD_oVDyw6&
zR2@77<TK}%#5{@T?AmoOp8&D?1UnU!VjWd)E4bZmuf_=WV6SzX^%tZ)Acs=v*{kV3
z6DC2nhfJMP!{m^3N;BIFO0G{eabsjBH8)hcn;;si52*}C+`~akUov(eeh$5Bi^nr(
zu&f=(DgE{zfKUiaRqghQ>sB;+ZwDHC>Amrp`InK|v6<hwnQtgQ_zJpES6EkZ3Pv@%
zE(@)V^RIq7S2td@N<SU$_Dyrs<Yl^%F(sR*;ySSqxPmDBaG6&#ALRO6SF$foYw~uP
zJ|_?QJd!OQZs@Hsb|ZYyR-^th&))nh@MCW3xI0pnexCo1kSz(@(`HaUsiI8hC<Rqy
zO+!l9Z598Gqe-&VS0!KGp%M0L-04mfn%>zTyc9u&$A}#@atGuVrhuPyC7uWQoEvBn
zlusJ8X0;~km`+vCp|A0s{Z}l7NLG_I#Ol}IZR*RDJ?ESjQZdLh>B8ajJE0(5g|gvk
zuPk~d0W$~$?<<=Il9mdc#Wr4jJ>c$g#J9u8w=|)C^OKr}`c1y7xkD-V%5Eh`p;)OY
zf1t*Y1IaR4xV|i#er|N88nZ5EAc`bu%&y1*&YOv=lr@tz)orC{4AyBJH}LaV<UfLe
zBBvRLL|i<j#xIhKi{obN#p>(UT5}DAQ#n|3teOr-nA|(yZ&yhi3l%5^y4eRl|5)@!
z&fz+ZWZR69-RD%Cx2rAq0#wi*;9G66{@L{NTzYvbq&2w;E4r$7ILrD(l8T_ccl3qU
zLO!^Tx@eLx@|Gs2%K<1;Ix?UbAicCCQhcH6N#|p*tLT_FDq*Q6Nb>?XwWWCweaK*C
zqU5-)9nCFHX=2lQ=1hk4tR;_8T8NVAABdZnGy)H_ZcXP{*}%y85CQX-n0=N-%|iuJ
zX%kN|MpczY?r5o&wQw?gx4Hl}Ik5FAP)+Sh8GS0AS;xX-O21QA#4~zwn7@VE?-C2o
zallGHn8^BM{#HaphB@X{q9E4hk#DkQE*E_uYmZwI4L@Z(zaZOe?)cj=Z3&hxrQKv8
zi)bY7!IypEpF^luUpmG|c^%<<#Ra|&)Fh=HslkGKK#hVXN?E7Dmt6)l&4m!^*a?GZ
z5&}p9s;6*9bWZgECS1j-`qgCSpv;`F5;_PM=f~5J)^~2vHy$XPSg=k8Ia7+2$ZN_O
zOdx?#d53jIM%aG8#D4BBRg&VK$(C<2OPpn_d9spG2C2V$h9ymh;u#9{J2G(B1tlo7
zBPN(^P;VHdJQfx|U$aQpD~N*1<xYE2*?bj~x=TYt&7!DOZVJ`5linL{j%A%x*+AJ4
zYrnd}XSeeuJy9sR$tIgHV<B1dMe`~hX418V3=X*;p>*W=X8I>=F&r|@+*EqM)d+D-
za5ca9CP2m#^<CH#1V!RSOdjlR*GU03N9wH^<cT%x*lNSVUjEuNph$ImCYNmUFr#q9
ze1C{usd17dm{&QT|J7D<tgEl11!DFl3stfkK%}}t(L%#bMO3CXX`PkrB#)4d$M&dl
z(%pfi@$|?kV65KC;79_a>UF}i3VkoQd7RH7CUVr|c=kXYwhk+~+irzehbwuxpUDj5
znOGeg8R1)$v#(h3n6b-pqPB|#!>cJ;3$u6zAZ~-~CajA2$r*3ywa71cSSphz5gDCA
zc6S+~jS6*8n!;f4yvoFagnP+)JDK%8N><sDNn6XJ!$C^xMH6KcC30{pfEDDK$sUUd
zBs;GPo)tIONik&AMn0)l+R!O5p{w?Pf**pID^(u{4cevmp~W_BH>}S{Erl4Z^zU=4
z4z4KJZTT#KV<XtUy)Yxo9Ccpc5Pd+h-(>w@0nFL2tmZ=a1(~C{7@-ppC-<0s!16Nr
z?3sV(SzqDiUcxR@ev?-eMcB*DJFpO$s>Bc#KZ9^sdS3$zZcp)@A+K62FB1VmKT^}M
z;3LIV4T7&PC29&Od6*}1TJi;7Ha0>pU^2WeXFpoU%D`iw3_c?C*rjjRO0S|>2wN8A
zU`m=>e*$MMryY&1cPt3b^#49{>DTBLJmKaL>ZP+Xt>d)mGj3LArTqn2OlZZGU{Jrg
zou?iL65hyh`Oa_8U4WXrf>YBOwkGk=Ri5h1tjdpHsU@8SMaHe$d1!RGb|>#Jut#Ky
zB*Qaaa%!6f7TtFz4)SoS=4I%xL_N0pE$@KTub`AcZ0VU%Onv>CYe}5nnGpsBY9tnS
zL6rj0_e}QogSj@<ZS5+Wss_?$r-s)@44=3A%YSqCh8fk+yp!g4J9KCZ)1^##F<y>1
zTmEXyyQe?4^~0e`6+V_$KED37Uy{Rqk2vIDOX-xJj;gibcO^tz!&fh16m92rpjysb
z(%#A0$y;D!hntTaI=9O2g+@|pb7WS`NSQJ$!_7bgUS<#EOL{xVeL1adPzb_|d*g!=
zeKB7QL%+4P@Ua^aRCzfv%*L&R70vX16Y85atR5<;2qC%6P2_<u_EM|onGYenGUJd@
z52Kkgd_XHk+*%H`11d?g$(#eST4Y$BkKl~O>vN+M&Ya9r3!6BKckPL@_@mT&!ii^|
zfxH=`!+5FPcoi$psJM+GzJgftr~05@YLn(_G9?`7JdWR@5pJt|QcZG}rcvQ~yrGsV
zpQ}viLzSarwKMRL+E$-{{SynXps^WI&@HY_?Y*!~8mMrMp6CxfLohsSInCnxT-fTM
zscAX2a#u^2e3!%7V{<_p%>JBlm(iz8aJ(OggY6-@B)wZK?;)J{Di1WZQLh;)$_gU|
zn_4IXPdLKc%Xn3OGM+RS099WltuMbfVOWjy7n+q&;w{1b`nqdUMAvI9cTkG41?-@c
z!3E`gA?Hsv=%pTVE~^JWTf&ucXQu8vWW%AcUD0CI$TCc6tYH_~dV-f`7}iR#8$ZW>
zt=CG>Ph?0PQ7@}P?c7yXr|WOqJ*x&Ynw2=sRO!0D%Cv7)%~>59aIBl4Dlel9S(CC7
z<Lyd*<XM*sXG|@8w-o2x)SjUXj6ZpqS4I;^$4@Esy$U^M`a{JeKDe^WVQk_ZEq1sj
zxK^1ldYfGuPtYVGXSj4Vk7Jy^s)~xvq>6HiKG$=x#%P?1okUMrzL%A4AnDi|`Yc)N
z?RB$FZa!_l5l5LpKsshz;YcHEd0pYpmF`z1D5hkg<|vVy1$XNL7v<7iQ)W<VQdDV_
z1dGF~q|TGOw6v!w^^xnD^pN_eN*>Dz@yC&o%l#c)4N)YbFPNlgyvq9@CVQB0K%I(J
zg+NU6-5xaeUNDH0SRhDg;Yd)Oc-4!|@&u@7K;X<6bT76%Dt=cFhm^UA%kI_>O}`%P
zw{91#)-BnHXW~ZRU+T=vroBgADutYzzDZrv00HY3v==~BxhhQL6Ateb8<Zv$4-~KT
zGe+x|XOn<M9SLU^=UoK~o|3A-kJmxIKt*;zaB{566tWJXC5YUKfc0(u7F$o3T#MLj
z*ss~0jcbY3*Vn0)GpcAd(${xEQ%$dg0R{_~3+Z*8zkr5_>Zx%^F(&ZH$Ak`0n?|GP
z*TvGxF4R5~oeA{X8+9F)HPo>iCIT-jCW#a7oOnE0pG>!5cfhZk##UC8sl}k`VpNQ%
ztg9~wTSDhUP~{v)HJX5-anF$o$HcUF0;GPRi{iIOlBsxQ9YA$lKF-jWn$9q-c$uUN
zk_-}=P!-3@gCM0afr#%MN^>U0y-p;#OOS=UsvX}RPV;WAqXfO&)h<dS(gK;oJh_If
z617Z)SDIV9+O(;S&_W9Ncp5Ir7gxj_LAk9go9Kg^l+p6f>ux)+8n6Y0KW?<8rin_I
z0UM_pGF2BlL>WAu^BwU`0{A=2=~0pusTg~qxKO8KR&n&$Hm;+p7JHI`N7cKIeyK8b
zt64=TT<HEllRJ)$xUHBW<+KSD6K0W3kOBgd8=4;xoTl?KqA(iD91LwJNWJvQaD7Y4
ztIxB}`S`gbNcnfhUSha6^|gW|X@*Jxje4G*@2_|bmrSfLiekpy{IoBr!>Q;tu4UXC
zyu4Ls84`PN3)-HlZHWU0EyDCxUI)v2(Zok>@gSTLJ?Z|UcDl&ICTo@zRXHmxWv}v>
zvra`e@5dxkUZmt*@9bNfbj>v#@rz0krd#=$&Qj)M2##sHjV8-_rGO}2_-P~@Gs1F$
z&-+39bwY<fhK+WetQ)){Iy$1mDLB{^=v5o-Q<oL%SHVQ(PgM0yOD7$PEedqDetb6?
zfN_1bCI|znOJAj{227oWeWz_dxqY<d;C&x>EzVCd@QYfF?wK^c#WNaS6K5A1F7t`C
z{MU)_p7dx<V2Eu~CoJS*I2!(Wwp>>%Gnk#yC^njBhppDwuC^~Zl2jvqCld&bW~-V`
zfSMm#&*4@=xMwj!m%`$9WExUEsM3S5)G6JDa__sp+R%b5`781Op@5Ir<KN2Jc6*t>
zMOJDK>PE?9C&61TZ3dPEssM2^P-TG|;$nZDv$qNDAZ|Ncl0JKiB=S^RN8-VbuF`IT
zHKxkz5(GD2m9${IB;m7sR$X6NeK83=S=6Mw5u7Hcx60u&PvDDKdBZMv=V!lsy*CMT
zA_><@Ugusij=5q?$kG^xvWkS#;U9=>N1K-$-hk5i?)kYl_CB1(4P2z$jps?Kw)M`b
z2Mp6ws<XauWJXkst{o>^J!R?~`<DPJXi6vUT`5|22Ort<@Ql8|Pt0B{CwEH?(Air(
z(WFy}K8W>xs<t9D)J&*lK6Izgtx?kZC6esu{71gk$ESNv)xCZR<Wp>rdPR^ko2sSq
zGs+Xyx3%Fziol8D{t5Gbc0y(Aq1`3Q2QLL%t7Rwk8;Q(7glH=vy4j3Sk*TmWqF$UO
z4g*+vdYy<CEX?$%<V<kGbG}~gQ;EfGFt9OTDA$NMxLj?sjCFEj{Q`DLnKj?&9+XDz
z?_YApRcN<ZwacCt<&`oMnrnAu&Kth5)04<Y=wm<l_})syNySj1!@GjtIdHukofM6_
zm>=&?lEd-A?5R^;IKV@+`!sDmCNBK7BL~>Ev5bq_uBN|C|Fc!^+df=>{h$&P8v1Rd
z^CJuP(g~32^PP0~u4=c$4#p>IdPa?Ok}7$&^9ps;t;~%&_bh4-L>Z8#pXQotVuQ}}
zi|QOkQAR|>_T6gBB|4{KV%M@h*s`t1dwM32do6AO4-04VB*USVK`bkbn|(z4fV~MA
zOO0h;;uY;xpEM;6t$r<y&SRYI3|X%;tM!z3Bmha$<iZORZK?{S@R%c`X_P#XZ!8P+
z^a6=tb~7S~2etk_&qCsMwOZ5J%PXpxP(9Xl-^~jxc`{f?#M@X6U>WJ_+=3BxoKgbM
zq?=vP<;$Q}(W@v&m)rz9E#=B$OaS6L!Ivy6LkVA(e&xq!e`~J|Ae<7oWo0NV_Tl1s
zV$2dr-)$rLrBLZ7X%QZE(c1RKsePoqT!tcthL=P4E_KspPq}S_XCr?^0Z)k1w8^B<
zWDswHqh12GgyC@eAW>Zn(+nYtt(8mD_6%QBPE=V8de$&Tw6+|a*3ojj;kv9M!t0~=
zVz^6`yvj~9*{8G7$!8*B7v##uJEpJYo*`7%a4>o^d(7GYE_-6y-`ni)LHlCg2@TeG
zktYpjj-5#V?3=GDy<YJMa+|>BOKN2W2LEljk4hMgi%HTZr;npcQC5jF2HJPd@TJR4
z-_ENj-+oA+mXtj*ibhpgV0Ah<@-K<&ERhe00~+Ts?=)D*$a4@jGBheLPLZ#ZJ)1HR
zMYP9Mn@<iVN+=CWc1MD!y}xSik^6XfES@J@BWP}<yCRQ&4;IdNT9ot_GPkPNgS|ZC
z*Rj@Gal@t485L-BZp4ehgk6*t-OzOJ&<~d8N??JKp>e8w;I^7nOloLD(UXo8&wPij
ziPD<wpp<Sn<14onF|mQt)%5PLzBt!P&k^A$tAj}~W<rPw^0EF4KC#iaLkel<OrMJz
zw+Q~0xx98^DEC}mdM`T!#DQDYi|O?&Rjki)^oi=GnIXDv2d+S(2%pt{N0}3OW-TN?
zhWHjq;Pg7fZxn=Oyzg$BzLMKHW~iuAp1u9bU<ZGNJLEQ{@-B{ty62<yi7Zwd9hE{b
z*elLq+ABvJEd3iqhGbbDEh=1<g`T6n>wROV!dVm-LOa6gFI++$$gkY%^=-HU89nn@
z)jZ)Csq5|c^K8FQQI&cQ#6qQ}{brsH2&7;Ns&P7?9_8YGpn}gsnFDQ697g_d)2~qJ
zpg9?)HWohJ;^L&F!ArS)Iq=1gG9vQ1rKsQxFz!@V&lJKgD-CC|oTLjUrTK~VvfEcJ
zv}jsPC{{J4bAmh-A`q{YJHAFGNmw8@Gg&BABg~4&*fCf~)*|#E>#TD^KUYvogMQp<
zj|W=xGc$z7$jN(HEr0*_XXZRP1K@qaj@>aZwY;O?<ZRT5oWAKyw(vB>#kL0HA288Y
zTqHP)-GT|J)&X-{Hb^k+@?#>hnzXhcbb;n`UFC9;kn}oazZUH;Bw%Bee7<&&K&8QX
zqg2^jV?*<K6sPvy19>;gAH$ZO-O|5ii{2D>dA{OtgO6g4p(%6h%(h!daca!#nZL}N
zuQ5rfu4b;-uu}-4R`0>=GSEd(_qlDE!S1}dh3^rJKx7vK<`#2g6*)5=ASYTAYf#9s
z<9f>35xvJCz6lht>am2roMQ?WM`OfxXVfC2>1N0+vjTPD1Y6UN1A~2Gxh#Y#<X2%4
zydaTpm^qYIX=J%EOfusgaSCjq=z@|xmQvgC0W--Tz*9kGT+!WdT2c=1$nPz1W|Bdv
zzD&XO(V$mEH&A^_A54oJ_q-E^Mu&|*{`L*k6|Q&Fc?5jluvu<dvWVW7LSr+-XGX6%
z{mia@7|o#07b5DU+z)%c_cnz=cv-}~kiH*%3_tHERN<|4;n3?g$<1<FZ+9m3op<B*
z6HU7~@)Et=exJzEQ26t_6Z|+jkfFhQ2IFDPB9>hCCov7xo5G==yf}x1b}M=E#RE;g
zGzwWWByg!%VPQP#4wWY1<~s4zb6JC^F1=iMP+-W$yO7_Z@26W&dljvIrvn4Mu^gVs
zg;ZK{n0W(>-6n^o6dVx*S~QO9kYndqDm(QjmU{T3d}cwm9b1^?h6x|STe)+#?$zSN
z#grJ=K&?XFaHi#`Q7E4ISX#u2#SexgCsusME2gi0ItQB_frAFYiUwDf8roMcV@*HC
zzI~G`t~BW+u2Ed@(#{><S4p;XL|Gw{X+JKe+@D#DP;r;Cd`e#SKwwg<cMJlT<%vF|
zvZhn<TZcw-i&EmZT+i#u$5j8c5Trj-nqx?&DX~VzE$`@es-i4iz=xC@!=|q`)Tg?9
zT|`iuLOai1H&;Znm7ad?4h~O#qP?QS?Msvl6j;-T>*H_^$K-x!R8@`(x#+TI4*Yl{
zX9p(eyn5TAo&GrZ!;58uCZWQiNvgcYZ8e#=hL6*Eq%x_74HNH68f|pjxI$HxKCj3*
zc_srD|5Cgr7tL}aTXaOjwt0CS5qQNyjTE0rkysYbMjG=&C5M!p)53Al59&PwFu~Cg
zM3^uC++0hvQ+_~!2yX`VqrAkqRcU=85~X%O8KhL++u?4+KveXA^+<p`qT8#?dNd=y
zUehJ*e%R{47mc{;!O={gE0)y+ue^h6%w`A4?qd*(=NjgJV7k9awnAl4TuCs7|E2Lw
zXp*tOC1dFQ3eDtnu)A7!@pX$rrvW>SjaIxq@8%L<oJ2#Z+Na!<lsGbSP8BSWx57jT
z_7x~<EuA=$M`2{e&tt+?(_*{&jtqW+y43|oiPc0;jtOEYB!hj{mxw*0U02Yl@%}ny
zMP}YVc5SDk7Cx<U6vxBky+(&CQ{xh=Tfo5VXepp!)fs|@<Lo4F6B=GLlZJ~e)9zL&
zV@@qhO2tFG#^?L-GGAf%?rm8-Ff}d7uv<5DpHm4z(3WRX;whP(EBZxVID~?VSgd%v
zVqND42+<Lb<*_g~<0Wj>`Lb*0*ee$S(HG}Zpb@H4nvD*bnQU|V0$-SWjhN=YKhrrQ
zz?K|$$8S=1j1&vaQAu;77j!}8U&p>&hwt^f=%@yk&>R_Y-@%jd?ym9PR#<|kYQB*d
z+sEm@NqF}gJj`Y-f3aR4nqNgYAXA|(TX;Rus$IY-GR?~P845CcO1ViH7glbOoAq@J
zsRM*Ij0NIRL`KaiIUyzA7BK4tazN!g^O@p{FP_uO=%2y&MhT`A7f1+PY{jkXo~{b9
zTG>9TC>>eJaw$$-GV&vgIxaL#4Szpd&tav*Yjc%v6u7DB0yfvnAwuP2Q%BhHhZJ-5
zq)}l+FUV!ZZjtI}E8VPyQqnGgY)55n5I6qgJDDL8!Iz?82^PPc1ZfeMo<H~mhLYsL
zL&~3qy?9MavkX(;30fd`U7JtLY3l_)ZdM&q!SPC4C%OH>`687?Fg=7sf8avGY+uP;
zfs$lOjrKvapz~}oq9SlitOsQ*G>nooq;PK%3PbVeeKf91tzy%Qv<&#+G7!w%iL{+a
zRxLN0$jj4kCYgaKWEm69JLtIfdYpQh(COYLxN%pJ=+BHO2;#CT+hH+FrWH;uL}W)J
z8kBVin#@zv>+$rDMSWw{AMS`ms#~SMQ93YKOIE^jV5gomZQG)(C=m5`Gzoqk8AGKM
zPHV^r&1BVzWQ;MSz-aYS&mMPp_(7t|C)^9aRdd-DI~9HAc+lne2f#ez-*Sj1tua^b
zp$hAW$Hd$yG9gd&)MP(ohZj#6y@MB!wkn^SX1>a+r~VO%`Ba7CvPpl+kN<2s!6(9&
zwtKGth)d1)VF;f+ib1B+spD{s+3niCpOvjtS@%78RRR6L8G>;5Bgw%jatEE8GxO{b
zDT3~?`Nbsi3%C@B)%4zWhLC!tXP4e=k%FWwVmiI4QL?zMPN6JqvRSBe#;|_Kx*%AX
zdDJ(BXfCzQZ;WRzpp}Tm(UUbLhH}2}6_i!bSCC~sJ?XJIQ-MO7#+_9_jq=O|mh7%#
z%{Rc=^I|)8xti%CjhUH?VaI#seHWS2(a+y*`Jxe8tUpbpkak4L20EHWD6zIhLtfkj
zBO*iE3CPdwJAqPr2Rx20O?v{~T<zt_BEuI@ZHVvyBV57#4bZj|=0WpVdp4?<9pY?V
zRkyBlbn3Za2p$BDN2Ru=lZ#}#|H8x5KcvO=CSp%&nh-y#3HMr4IEXeVn6#(e)^SMK
zqYxx@c+MvM<6tdk6JzC`PUYs$F&<J!4INj9{XX!pck19v*X%u-Hmnu-%%Eiy{=%ZE
zmW|5pu)vH^kCU^uN4Wqaf+>BbS7s!8<ClXlZmsO7VzF5E7bY#N!O@O<o!3>&F=yIe
zkQpFmGVaJX9Wfo7YJeuHbDWJ=O7HtEFE6Vu@!jm&f%E2l&n!u>>8vzccdo@A<Cs&4
zE?iwwmlf(P=DcPY?xIwr@lo-RqLMbR2Di_BIfxG!SGX8^74=i<ZnIu7D0~MMQw!mb
zWNgi#i74b@CWtqK=n%Xu1__qQcb;4%%NKXV>s$GFpd~iZqImoiB#AV67J?P>a$C27
z$U<_+_8g<dY6#TYG=D|_28D~=SFUsCQ^|iUWN1+7$7pL+3Ua6cVZ_B<a(x&pgoCd$
zJ#!@b^0UKQu|Y}*IyCJfM<-Sg{<}FalfHDKZ>)rgO`>q4@g*U7!d0_6bk6Tot~e+&
z#t}R7F@?pnZb*u<Du<eBxC)Sk_XwySuAAWmRhA-bA}W9yzC6PV+u;3)B#G0TQQy>F
zpqLo&DD>0t@eIk!J1Z_Bt26X*B&$`V7o~SZJLV;j(D)VnAafMStEb5*N08Fk+RKzU
z8}sLN8+lTH=MpQ~oyc?!721AvOE@F)J5eR_R!wOXc)=`t%RFA|eBsp;Nq-qhXT$9e
zVFQg`V`Q;0>6X$wQWm?(QU)luV_9ib#P5Jn?F}UceSx2eyjltGmTVFhV<!{f+uL5z
z2y39D!}jc8b<cTyk4YSEqR&vSMbF!XGC8EnQvpn@rtQ2Vm~OmM>dr$2>nK~*<L#7`
z$ig(PgkIcNuFDko$7*3StV9Hw<`d#1pCiEq)evS-EX+_*<25iTZUfF+=||!#uMZcb
zC@t4jDOm)rMa;1FzM0z9)-yN9EFOqzl{EAm=uJRZs7X1MMU;SZyFYNw)}V)jC4W({
zHUAk{P*tUhbxLQFjagE*h~Gu6$zBT)H^DTN=aX-sHcpe&sBIm@&(kh~{sSmslD(zd
z!cmkC@A`JHmta%d^ara^a0o94SZw@|_GzT1a;X>xdJD~fhih_}Et2T`sd^}6O3u?+
zp$fhxXPT&m@U;xENa}BBShF_KPH-AfNPshvz&s}yV-i4PeG}m`2;}5Itil3F3wNT1
z|D!%hbIQ&4d2;6{(U`=DBo2aZhaaJokrB!8WE=S{l`8geXB$?`wlB{%(YC8T`KS3O
zZc6I|NrkCV3Rgd+NAx0?Hu-(Qqn_n#${v+n!wGJx-x!@O7|C^EnwBeIE3poc#ZVJ<
z3!D4+0#9NULs$4C=mH|Qv2qk7o)~E+jE^_9qJt9C=**ooW32AmP|csSvTQQ6p6$|a
zl6H#ONr)&nNXLkf)s*k8j;`ley=*!;y}ohjpGC=b42IM2R7XdQ?Xl^JjM?!zMP+U&
zgnRUrf!HQ23dT#EwX#whj~$pb^AuDnfPB_CbiTFcdHZi0X`a}7c|YvG7UL&Zwp*f9
zHMiB|#lnp@c6kvTzOkZBqZG^viWVfzPyF$rT61l=E9NrpM`h{;#ukF=2}+=mip0DR
zq|`8JU&Zd(qmlnRQw@q>-s)rJN#u^?ns)wBiM-aA?CY|g9CXaE-l)MkBpm2px7ZuP
z@7fvC=pMi#Glp<Tg*uwS9yzQedMKK+6(JS#M_Cp1vPD(+H!XGgemD42tw_;b3-<lm
zj>|gHV#>$TT>4aan8}RC?7|e!&ixwlaumc2Eu-~fq)^fZ*X?HW5_}$&j_mU)qLwfD
z-XRjkNntZqf|tg#MUr5pILz2V@8H+`QgD=p^bepT#-aHWWy{V^vryC0<>cULcB(kc
zPF@%bQSUzu+d+5G3Tko|pClQ9W}cN-7*vdoD}JQjR&AFt$+8-4Da_E=j(&a3zH9X1
zHrzgNL`<vLZdhmLGIdWcl$!{*4O_o52KCZudvdEZqPEdDKwILsVZ5hjPt98yzh>G=
zkf7E;rp@T7neymi@_k%)W?l$0<0=%wMhGEySHW3Wh5L1`s#H@M&N6URxHH$4A6wxT
zrRv1^xF-1fwDOHd8s@_-A*3`R&9sjad`-xhMqh3<6j_e5hbrgJc^2pumNZ*9JI}~J
zj8%OjSHk@Vz|VJZ$+_d9Ho9gutEiB)%U<)WYxHKQhgnr-$daFdHEaQo6sPN5$sa)4
zkZztzzpMiYqkmNQ#&mgZtMJ02)&>2D@*7S|oCu6#!&3r~VSI}7sMNkHmIM=BsFO?k
zXUdCmGcSXMLWJH=<YbUCTk8E8pz&-;<mGCIAGZKZA&6CkMYmu@wq6x|SZYzn>1C~R
zu((}pR3L@)7#)|7TCFDr8)=TeI?q?tAz1wb5D@u!6|{@|BpLH+ahQBDC+E`SVMp&z
zTen2J{dU0?cJWBUXU0bIp}OZ3zShbXkR7n)`k1r1j60u}WfOfJrD*@AGwGyfreJJ|
z<~v4J@V0o4uAVhpHS&qJ^7?pT<^$wk;#c%szSM|ezfXGcTvMOThGQMoIzD!_iKoe>
zdYADOiZ@Ph|E+AHOeebTps8Gb)K;%Q{WwNz7p#rI>vkc9ryR@wteol$`h}lq+L07X
zjJo~0IYwz49-RBjc6nZd&^dL_C4(Wn^^Qc;MBQMV#NcSmISm+pHHq0s@l;KXGmiJ%
zvax+W@M|k|+8m>PLA~C*EMunjsmYAa0}COTrZ1v?Xq<K{%n&)p$nQ93)Q*$Z1m2cp
z%+*T~8#8e*w>m_mMZmie%8TcdczI633RduXXBvUTtcD{8MM_q>Scwp+LKR5SW8d3v
z2d>pLIIj;jY8g0llnlnN-i3c{ytrQzW{X}*>`E%-xZ=Q;S8#Bh*JEc2SVrSiJ6_Xx
z(<5BBouBKT#0y88;E>35a~0fy#C|Jx)t*pjyjOxQz|g6|cyx~NDnTM6LmhBGG>v!8
z&}tp__^8Or_)-vKL!o!5U?Vx}DK?=h(CF&d(GpDx3XVL6S2O8wY%9ZK#u7{l=hXVX
z9Qs6SFMR4O8t;!03%GdyvzF6r+~sBBcrC|cswc9e$?g_!nq;#Il(dMR*ATtkba&*Y
ztm0>9+fLWX13@1O$r<dW81@QA<0eoqS#71b5(Ag*Up1&rdQD`SBwJbM$)v^#?y0=g
z@znLRP9vm$y_>;ZzUGoEx09f^Jz*N@q`;Zm(3|R{`*lMGW1@ohz3}Z2ytRUX$l4+^
zQ_1T<ZQoYbK9y*nP38r#gof0_q^m2&mS-hbw#x)Z18*>pK2>E#AAhsrk(wQ)ne5MZ
zjLt7!f7JJigI~DkfWVZ?12>9wUL=%(F5;|re*kzu6GC>XN6gwLO|yFz+`)FHdHGL=
z-spx$XGhY|&x$W`B$vFO=DR!JuXdIXso^oNvh-pnV}>{Dm^eGay$~)WRbNyoJDi@w
zt>8DqlJ483D@?wRwo&BzRrimwy1#YTy1!`DjkeL&$HXz$4aCuAPEa&jDyhHvdQ!AJ
z7=H3J;$2U|SNlD@I7_Xyg91LzM*^{$c{;z1bRZ@eBe)He&Du-33`4V1RJ+FtBlY}*
zi*3?Yxd&Olc5!(JURj;>>66_rH_&H8c8GgS<--e1Ib9y+XW1R)a%xTqF8+?@nbJ}X
zKfKo2xV&7Qb~H<LalG-&I@eZW6$y9xjvPp^ljj#Y(#7iwUgB!JZy2>_;cW^mSCm<W
zE4D*&cJRA`ajpf-w`rryzET7bw|_~8uk=`0kG5)m)Qa1((@tV4WHNkJQ{xi)l>VFd
z0;xIqRI6papgZ#HCg7>YYT8VxOltc@=+B(N@f5}5#`g^zN^y}w3AYP27{N;%^wKLG
zrmLnhyXi-|09JVKW`+h|^;N`d=?5I$8htBS(w;46Sj=MI6P}1(G;98LWRG($Ty(Ct
zbI!jYDdCY*>;t<bHMpthw8ixkS>Zkyx#QvaU(|hNR2;#!?m$9F2ofMbaDpWa?k+)t
zI|LbA2iM?|KyV#ggANRCLvRT0?(Xicd6RSQefO+)-@5nD{qbI}H8VZcUDZ{!t9MoJ
z-uwGFK*S4(fYj*=*$3*09koM!x2JR;?_1_=)!6i!do<=Pp~E>>*XY9<&Q&X4$<*zN
z>ck7<JCeJ5PyL}5)`!UF3!ZyEsz(`@N9Ebzc|EI&>B*_5T8%50#;bO#uY5_9mRmpt
z*Rp}_5wP0pFihha3?mn-H^tB`RoL74y8htZSB*P65YB`RiQKquODh@U5{I$75Zh{D
z`csPbjNvaHb=e6PFJ6SN^bV?27+0`aEaiz8v^o#a)#IoNbGhI<+C*pPeuOmdK+w7c
z!$=KtV#;bSEM&eCD)g=4bzzp)Gw_Bwet!Cj41PUEoHbbN`ww&1;M8YOe%>g7%2ho%
z(fs$<7BA4(x=_c?l-;&{YQP#SM@>SPU(vyCC7vQ?D@L0<Cps|M>cX<FKvuQofC={>
ziBrRS!Aw!Xh*;hBM6`S6H4^Mxi#e~vQEG5UGkw13Po|i?UmlqPbH;^4=dTn@%{){X
zq!m5>06@<8%|w#DLsBL_Tq%3+t`v$S%-;A=Sy%0zA~+kV8#gRuNolQz7EKMs!GB#N
z14#;Ci&OkM75F2AKUg-m@vQ1bo2$GfJJ8J7Wpd{ZjJ&&p@ofy}dATGqS)o5wkM68g
zbH^2GfTaNg;gC?r7@BWF$%&smG*93D+~LOBRYj56A;OIMxmqwM{GF_HkyLRE_%^SU
ziK<M@W2iGnUcS{-Y=MwEG6?T5Jhqo<(>5omXp@g1t<xzEEmux=uOFKF#fc`N=7kx0
zcfDs^4y4k0fXD8OU}a}z4PO1-u4i`EC}~ZZ4drk2sadWed8UC24STp5vbIb8L@~QB
zz8w)VbD`$dhc;LRu+U;GJ-<pr0xjK!Nt_@~RXcctV3E>D&Pk~6iOfxGmaQdAK#U)%
zWrN<;d2*4en-J|9z5foJsUycX!2Ug38m@<7lts_6XQXDVuJK$_><hvrqN`@_N?h`W
zD;caU4(Vi5Du<l1TNa7Ndzs|ODs;tPPi7XvflXs0Mq-~Iznfmpl}>;5j)iGy7ePoP
z;Fu;ARv(TRwVGSb&80dkJ5#3;f|H`0m#!i@k2Og7W*ozJRc!l=UBsZJ;wdqbanh%n
z35jF9$6PCWsI`m4;%^xR#9~Im8E{{t`~iGAxiDu3D5!oY5Ka?MEvCLJ_K@qCfVMDj
z$=tUiJf(|2^Bsz7EYI>$q;x>sq(>c8VWlu1E4~ixj*lm1kF)Rl8QxALnee11RCY{&
zK<XeX;Fi$`t0qk_8Q!+cPBmMq^Y&EP_n6`^WK%X<vIOQu*Rdkif`ChRbPYzZSX_`P
zH@8~9X^ry${|Pzemt)QHe6%axl6xU#=*2Mo&z)bJH(|I4su}(gLw6Ffqc<?%oTFqS
zY{C?ZqebC6Rq9M_oVAUi4lS&%QXA7PWlQ<~gVG2-rXurcWDG(tjIRiHChnT_YJ3@e
ziHKI8#wK-koseL6sf<Fe7{{&T4!F*+WF_Y0NV6)Oba-Esk8$q?f_IJ#X!}qts@*$p
zkJ@@@TPf=}E=I^X;^a}sP!i3=kl&_vg?sH@&FAj)MI4ex?B0=^y$$p;=vk%7Wq&TO
zx@|uv5f)HT0nXTDm5X4fzAlkJx8jP<uG8YGNXFHm<Sjf(#~;u%E0+gyQEP`4VTLJ7
z$HUqq&|I|V!p99<izB-v#1C-D&5I`*1=VflO<A5*pKZ?d2NCXchcPp13XK5sUn$eh
z1Xn<3u`5#6Yc|-qUt5R9+wZCu*+#+`CTg7qN;p5RBtvjh;}M|`TM2sPN`jI0+!dVq
zvkj8*r1X6jn6%Nfrl-I)f_0)TTu4Wx7lV25%6JQVV!_1bwqYIJn++VD+HalCv9f|G
zBMaTkS=EUeV{vfq(ogN9tWFju%T!0wa&ZL`G2$*v(xVNiZGJbhPX*4l<I4ROO*xtz
zYFr|-(*ZnC-;E;X=}3c9X7-5%)6zLyQ~h$g*K3<Ubu0FI>WV6oCA6dT^Jptxd>a4q
zEtgXuWRfL6jd^0q|4e5NE`~hzm{pylIGW0GA@GV3lxRj`_BIKU(moigEy#H+quUYi
z<;=mrb37jpVdOeEY*165lA?6|AXNbyf5+xeR9{tEBpw%BRitF{*{O(N$vE4yor<%9
z=xLD9#8cVD#jz$ruWHu)@Ox!dP*Q{9vm?f=s*_0nP;OFgEVH?Z1DlfhN}HtPJdSIN
zOEAs5UCx(4$tLs^_cZg@s&;<&bO+OZ-hn_qlu_}kXp_eb{fPW>8z2|04-F!U2Ye43
z{;`>?UHxjVnDCb>41TD^`YHz7rg(yHFp{1~%Rj6F9mpYP7Wd1A_eW1|c)IkNB~xfh
z#?j4R*Dj4GJmdJrxAWv5lE~dbb&CC<(Xd}R$vj`6_A6%iQx!xs?ZKRV@*a%Gs#PDy
zncOa6AniDH85iZ0{wm^hv}_43hPYn?vAfX~r_U4}e;G;1-+^M=7;~#o`z)yl`&Z3I
zDEqs!YO&fna8Gt$eUA8eQ(0kINICP021k$jGjrm>;;SR0>@u)@fb`9bCZ+_8;rB$g
zGF?=r7Q6`}%d^Wv&r-!091%RTj|ZBoIh3Nhj8P__QWWT4!<^&)%6jWpd$QD24b~u2
zs;I>)n0lMSNJR?cu*3(k-Is~hWaduLO&CLm^999>UPsf42qT#6<b???QGR;wj=qQz
z^fSAV&E73m|8THypvaz6jLRCl_@(0u69#B<X9k>%Q@)o=7h265td3W}XK0?){iA0f
z`A`Qh!h;vLU?#RE|D{`K6Q(FFxAY4QSa__oB``H_Ou1@wo+0+gq#<dC4ux=*9#6ac
zQXN)k$}Q6t-lVqA!y9mE)sJ6AOm+-g%?nYkPbydgsgO#f#=4ZqKy=N#q~kJ4!Z_<w
zD`S8hWm^g4GIPF{YQ?@8P47SDiFD7bp}eA=7xxcm<pA<q4*$6GP+6UVYAH$+v^o83
zmZ!l%lc~o|e>LJl(y6gmxRAo0PcD@wWLKjF9XA27RmAiXnvv4KGIlN4kfPQb1<g^#
zkVMu4DN}g#c`|D;@ebX7XKTPAbrK=Y3)q&J5?RLgr)hcJMY3Sz^~O%}EcZZhF&WD#
zUloTcnR$f$^^ttFiTH{1l;seaMpSk-#(Y^KPsrWOf%c_!-w>bVvwZK>vSO-va=7MI
zi3e3IZ6z<ZZICq7%ER3HVBVm!8x&@ik3(QirLN&>iMKN-A#EyQR#KzRrdAViu<ve7
zTf$fU%dIo19advI6oZGpmi|dG{ZqYhDa(&J393AuZt7TR8(|oonY7|>Hu9~Ej9K;4
z&y{?lS~NwnTx{YRxhU~kUDV>9X$2jDv-(=<%JqrrI%Z!Z*_vpK>QWZ7Wji$X7WGS`
zCv|N04y|;&2MO$Jao~#+r`(u@iJVTB^!SUC_tgdqSVP&y2X59Qn+tDmF3nAQ6D4^n
z)@qS@UwiLTJIq-mtapy)A#8r@@+3ZF(e|zk5uanP%w(z=I6)zDS`5<P^m*d(RkqTp
zE5iddou%aN`$P|jF-kqN_o#?U)x>h&)-NT4RH0mEOnld06=k@Dt!UJ(nL5qh=ju&@
z5QTY!9=f=Go`;o0%%0veUvoD2&CG+!SyS06%Tk!IjpVia!{j;$fSl6YAilv8xmx3K
z!%uL*(PY8B2A26Cd}Iihcp%q=I`00bsPnSynYUUcv^!fmIMm<E54!k&v(OL1gx&&;
z3o0s5gh-Wvtj2ZG?|T@+a&qU_1*4Oh=lG~nI@A?*G?zIszw3B~c1cgE02Lux&Q3YG
z<0k72^{rOedOM3S_A1HjI(>vwmkyDO=?9#v(FLu-e!;p>0?g5mZl?MJm>9~e_@GXa
zt^$sia2ieQl3(0jp&NILNX~8?u8h}vm=jy-DllscY46((hv;3bmU_F<<8qFK!0Hy>
z@41g@I$?Pn`cYf<RRwNAFC<I2v0Hz~TU2u^uk=l3%XPwN5cVp?Kt@bzH;8wbLB%Lu
ziC*Ta_E2K>S~xcvlz^C#i^eAk^C!g7g-fOwvd4CqDxKJoXhqCC!=Nxb?DS>Eqi<6(
zyRdf4&3ZwmeIdon_T;y<S~s@?aJo(Qsm_EcOcxO}5Eh$kHxMu98rMlS1%@3=ZEz$K
z%AsO^aOZ>PCWH0Sz0)jz8C5__-i%e`qNz>c7VM&;_Zks`*H_r0SY?t6TLZaYWbY<c
zf+}O)^J)>-cxu6lTx8SBx<<D*%YRTO<ux0TN@9?CFr;o`@3Zy(5b?~aM;4RtwV0Hx
zRmZbUCh|X6oGYxxT<b<H-`g9Qy!X_pC9TrV$hXVR(Q~w{0RpM|!`Kq}R`UllUwg+w
zaKt5PYF}OSe8`AgKb_GQLXeI7^b?`C@}b&BU-<f2{36FussO$CWW3tLN=*ZH?i-L!
zmM-KA2&24D(_{no5B3jUz2Zh0+efKrczODz$zdczxYNZ;yBgeS1TLcDR!oa1<lE{B
zK^CYK3&CV~*S9oTP%{1FM*9&GbJw@%$|?5;AT^`_R*B_Y3sXFydek1QzGP;WQnUhj
z+*k-<X>ofWr1^B)L|U3Fas;m)KP7E@9Q3_tIVKkM?D=K2GldKp8nXNx_24~v$;fj*
zknNQwDBvxk^MSl<d{C*<ePK`-=lW8L3K0)P7$+o@bK!)P@U?}>YHFb$ZD<J`^>l1W
z^JtU&iBXVldkgs~{knx-r?uLBW)Oi{lYz(6m|$9WR%yVn5Q9B^)b#lI?($sgM*N$e
zEZgj@QDzsCa$yZg#!JQNl9b`!0{T7o#us8&dpJF**r>aUzT_3k<~p1pw;19xp_#{*
zDzmAR-ag;=rv?>QDY-YiEAHOBW)FpCj_(LHSx-FSQ29vBlV<^{xa+)1F3I)aVBc5R
zEmVh5BYs+Gnq^3r`({vtkH|lYLP_rQIQj6mB&^%@&Q%1Wo_AGu|3eoIe5>y6E6(7P
z?LlIh|9_N`Ir6Q)(_7FH2)5I?0b-Tua0&+mFUMSU?}!99Og(`-0w)&WC+h!^A?j}i
z&$By31@9hF5zWi|E$;iF5z#)Xw=Fq{N)NuxM3h1v;PwX)i{G+CA<G_mxgs@PloIt@
zmi%uDfMJM!+VAetN%=<F_)K*Ai5bbvLH%TZO=`OE?@7Y3T-`eW3CSb%v)QczX>&zR
z)F*fJ%=_4y{&hbj#^p}@!CT1X&KC3}dU+$g=s}OB;bBq;A76p|Z2p2=?+2Lgl&+<h
zUw;GY@Y+eA|G}0aVIh6r<)VL0s0n@S3Fi;sQKjZd+b}}szIXqE@o$GwoTsO#{s9QI
zGt)p3?>zBQS{l<44k9@YX14kgDv%*t%a&I~4{<eN>v<m;A2%-Va4$|J{s4FdzHaaK
zM3pgfK!<VDnQ2PEr+b@6idgSaAfX<Qug#8iz5=(mG6FxO6_1nO?g)T3t>zG+>z5=Q
zlk&B)H@DUmsUO~6-n}JrWS1{{bl!kI4r-INoW8rgZOmql`UdkF6^NKc(0vWrO%(DX
zc;(MKQ8HtvYh?cb0`4xxBmHX`##;_A>jK0<PwpPu`+K?q8(Yx`bg@roXF6QTz9)X-
zwR$g#mwasW#Z}76s5pcJQ4ue6OZ3+az4=Z*-SN1?pugAmdj_#qKm7sp7HQt3`DH?*
zU<ibJ#p#F?+&=(8LUM>k;_0W`M^Vb_o4n@FfmQCewOcK3+&c%!H9qz}q*S+r$muxE
z`tY9hGjq^D9B*M>WoIAnPMu`S&{g9=OIM7{Q460Km$zC}OnqK$DNcu^=>Jap_6Oj^
zmDwxtS|{}8(fZTnnpxrrl;ZyJU{1dy@9cHE9kgub*$|l}*2Bfdt>A&SFe3>PYMQ_p
z0ZykDO1r%G{+!_5d5+VWRev<{>^3e((38iQo9i4r>XaMH7Vz@!v`R*|LPHY)P#0k(
z(sIL0P6zlXu)dXeQpS8{G&&v?8|u~IhgC+*ks0XT0^tfer66FU-^f(S{6~^y3&b_*
z^vxS@x4PU1`oV{^YTw{iJbVWy8U5`Nh6niOPu~Z0Mjs(?Ch~m4pirJ^n)T=to|)C6
ztu}dwA9!$krKA@2ZHUVP@i13f+<tUjYvGFnsIsf^nm?`6pN>iq{vFrSTO@W9%{Rw@
zkj@?Ca(yf6B!~VCJoZF`45ZyWuY6!$lQyT~&<~_TecCD!jtwPkV{BeND?6#p64sNn
zXhB0DewQe&DKidw3k6g<DBkLeD)JKk-0LJ1@&u1R(Yzhovt4h|*}&o@%t9pceVy0H
zJpJevW8J`Z+=O@u?sl8~@UaIWKSao=NP+_0{>2BxK*09R_`dzLb*LB{PUfR=lgPxV
zf8H$iuOlSsh5q6nK!W0QObP~;u2M{yM#5y_*=Vy8^KRwR^rDSj=7hKV-+H|Iun8|w
z)QGc7BZ!?zu+>_QILf3HY78Isc}xC~QsW!a%2c}lAq)Z7r6O%k&tdxHew3LoQ_~U(
zNH%rp_bxmCbh(M>l(f}r&%dd%>sTnq{rkS6Pm!4V%+Yt}PSQ7Wnj1Am4;msNLyM2V
zc<iwLP62qCPuu0TSzOC7yFO!~6*&TeaC}5`{_l-WRJ?SUCgf?ilGZ()mEyP=A}w@Y
z!)SErnI?)NzL?3Wruf?Q?D)g}*dVV{M?&4@oNG(30<Kwy`srOE+e~35smWvg(Z{<>
zc7)n8Ka-hx1n7&T)DCK9Xt`%YF88)_PGY+^0?Y0I*8fl`<a(tDHI44>y`%jH@bj-e
zzg+oOZI0wCMeP32{Z9pI_ji7H&(LU{zRcOu>JtfU_N5k;<-09pKJO0o(Gi*vzI}t>
zDY`L1XhoMsJj7&ErF!!@;+EXnC*WfhPnq7cf1NuL^yqgln0MCggp>bS3)VjXszcuq
zi5r*Ujen?PEWWzJy^i1~&%fFi+pf-9SaP$)`{}ci`_!Oby|)VCxi)@F_mNL_^5oD?
z*lLNLN>rIDXHY**D)TiJB3SNFb%1Y9$MGuyIvzLOdfF;$P1hDRi*R<T845*2AMQ4K
zH@?=(Mlu=Nw_+N6bfQPp_lxc+&G&<`;ToUMFJ2TcZ#rJ~czetiKTxyZzPIwKeiRwf
z;Z(=-ZEWF>)tdar94kw5cX0m&q5du!Q6S&)AEv*Ouil>l8Fse{5IT-7iZjcKpsYhk
zgfZEs$Tujx3_)7Rdz-J3vZf+1m0qmvdtcoe8X;@5?LGUIBw(AK19)#OBYa?E^`*&Y
zGczYC!^i5LkO|+RO31@^n9;-Py29$(_l92aW1f6nJa1a#Fc$*oEkioy(#m=s1)80F
z=HV9HsLH9RAM0+O_{?HzGiAGecQ{BdU)wRlXK^Ope`;~oYqGD-VK5zJWGbH3;Xlre
zSpwv6PP1TSC8Qn4$f{DcfqIQMW-qb7lNf~SUC7T@or~9>ux*H|i_2>Hq(9Ql`uCXp
zE8NQK&w~2;P97eQP0wD}H0^rT>N7Lm9_kr1LD>&R!1uTjZODEA04r-PGdIt83~8c!
z@}Q(1tXhVN+0gSBL#hJQeuQH2t;KHB`LjSbcYV4zJ_JUnoxBYw{s+}XXR)OMJ7de_
z7ZF-7)kgA+<A<rISp?>yAOqghimPzvC`#G%0QMgMXA|7%B=1|x0#5hC2-5%Y|CtK{
zQ7_f<g#|oyW1Gk!hl>EC#8ID1!HhmhUnJggb|Q=^V-Wi6HB04w6qy8HbX*0{P=l$j
zJuGE66=vA=ZsiAdsXCjE>D2x+qjsSW%fkwiKw9a3)2p8}<bSzpI}wCJNVmy{d?#^m
z_-3Yso}=Sqz?XGnrSf*vD%22slpx7W{~38uvK*YT@1fvrfuj|trv^wN_Z&M#lz`1E
zaZ|{&c|TM+*-u!EB+dx57(vL{r@(6Td<&Sk#b63Sa$veD?EH?CNR;zc-%S$?;-=o2
z=<}2lfwD4Hsi(eAVo5TimRyi?8|3P@_=GfDju4MSzFd$(5L1>?BGL1g*qO<E&fRli
zm?S(Ubhm7qz22YA1S}4+?1b+7#pJ3Hm@lyb$<ag&TT;IveHhQU>U?_><P4++<<G--
zL)#Vznoc7zB=*dn&2SWpbTl{)mdRJ|kIt)05!im@xGRRzmCn;aXGr<W_ZajpFS-&%
zoiTDn$WtY!GheQQk5pVC2$IEj<PmU*x_}$mkY2cTD2?p%6WDsXLM~~Q-9-H>&ichN
zt|DcpuI*lZMLyMG+v1p(UjZ>l&5}d9CM{Li(E^F<Q$V$~sS0L7$M$bLslMu~suaqG
z6G!*cSRV;c+JzV?<2tl$SVPpM&-B30Uk-?c20~cpY;tL+QJq@<WwgFX*26uTS-7+A
z@3t<TnHBBmH|wZn<$q;ITzL|r*Upc5@9J$E<yIx-Pbtt)AEW_6g0C>j0|mH=3@3eH
z)pWm?0N_W&@Ct(W-x;1D$er`nQG8oRP#d+D2z+K`jXnwLG}~ig6E82F%620&?ywFD
z@0b+}7ksI##j4JQr1|P@0EU?cgwbuMx){XZCw7+jcr#z#K29si1EZ22s34BrzqQp>
ztagkbSU3#Cl~C6OR~x)dLRP*f=+ZSQ?ep$gh<2?_zY8qFBjv@J!-iBMX^67E8R9v-
zhLR)cC~N18sxjbK|AD|<+<bBUT<p36Jnwx69Z)8MRua(ow<n%q@GEoenjmlynLy77
z%IMxjCGoyZfVoa-mpApOdr8JE+&pqp=6T4tZz!Gg&8Ep3i2|ps;lLUqqoqvC`9zg<
z-Z!lW8T|~~y2PTD#C~;M^$ek0#FM8r;gUtbuAh^%({%7jzS++CyUr|Ch;{)hrn;V`
z(jnMs5}e$c&ni@WFsrnyR$yD95c)#oUxo88a{7OE_kRO&9qmmhcpjj|Y9qitYm0i=
zHf!KO`{gAxMUbMNv9uKp9QiNjpF)$w{PvlrONh<>)$zXl3#okQJVIfqTqzL$JpoBj
zqP;dU?cveLPWQ&tJA>HYfiIc*QmDx-lS}}WXzobXel^yvmIb_Do0nV0%Bvvw3dBUZ
zwoVICfi%iDARTg2jmxWcd#W(C^-a#yIc#=cze+GSb$!n9xlN!OF(kgSUe;if<y$Vo
zKNhe}3~KLYIc&}VOQu#e_F8n3Pu~2G_A}C$jKVv^bX>lrol*Ov&YX$t&C}2hk3go;
z+THUc*g(hK0k1ljX$Q3GJuDzN%#r>4&6<fI=`^83Csz&iLy8m8Ixrw?*XgDfcUYD3
zchCz)KHDbs<meEz*GW4{*)o!;UJa|iow@jkNinp$Sz#048myHd+5Au8kT;fNZ*pq-
zwz`Apn0k7)DRa_dh%v|_9f8>}zP-@eIjC!Qc;?doXlwu5!j$&qg~r1h=g3Sq7@^uO
zuN<J-4zxkU4mB}(7TR)S)gL2((#eE`gwf+<H+I&aZJo%vIS)KZN<7IXr8ozj)bUEe
z+4Q=hp`pPAwUpUFAdoT}vE_{_JF2Pa#^Z5i`3w937!Hxm&IVUbTYO2dfmQtuw&K)r
zZNdqtVcgGNx13BVv^PpF>>By8pqz0Q9Y8TK7E5^s%wX%htW%FEUTc)rjB%R99&_Q_
zRPj+CeP6QLWV)4~-KN>DJF{szQ&*#`xR#i&YHH+KS$#`rQ4qQ`^*r|_D>g}T4LK$;
z$W*={%yh{mIhQ&%BYUKJvpZ2=s9Eq1i&(hyaT;!A+W(FDwF4+Y*FVN;ev?`A4d=<C
zEj|MswG>pCS?O(Mb&UI6gCG9&=euAeOL2*w%cDDyw|@V(ZUr)qZ5L*yilFH9+-E8y
zpeN7x1;qh&R7c+b_N|ZYM>m}ipvzZx=jrM9w-DrPHYNlavH}HKujQG0vjK`uuTTqQ
z9q<_dD2@YJGP4pCBiUDPX}bp_@evpxb1?k|C}mS_==qd-g+-=ZTRgvU+0kKBG_>5f
zbZ`8rlHPFzpcu&*XXZu1v*~oF9h(a52@8U3{9W1r02nTRDeqqgR5H)(Zaj%RKZZ6U
z@$Vj9k`DmvptTdeEJ)26%gPU}FYY|OrOpAw|2&kwc1ZEeT@b;?ym96CZ)tWP{+^D^
z9W%QJP%?vT=zr)iKoj|Ujep;ehzJb-vL{Lg&gl6)_ah=OK!Oec3W=<`2WJmwAN(*y
zIi?W|E<ylI8II|w)cG|42(cI-#Fg|&wwxbM{OT6sebvIO;N|6?#_)|{I-y8y7*^x@
z;hGzL<Kg?uFa#W|?X1<oaqV&lsK&0g6cV#=+9UGs7f1Y(V^T5=iy!MolW4?A(F|YY
z+}yg?T!xn*2*01JDKb7$Z0@)Ia=eVt;Y2WGA;@=ev|Vl+n;w5NUb|g8ZejR5&|z)k
zZ`(ZzGm_setg!ri_yc$#EM{mYF`p+*C>H7>;VnPaYb%<&xu+~KF;dU6Th<6vzSBcq
zW0Xm{6rHQ5n?$s$aPngX><tOz)nt1rVTi-T`~iGT<WZ)%D`7}@7tqW72Ov|5;_GMd
ziZO+s7yaU{l?{P|JXYkdzw#~=F8lrXM%_P&DzV76$a;kg&CnH7Cmt9V76}C#Y>~}=
zv63p)rT51{hQ;n`)Nn2343tXy%R^NK#|vWh)V?U?2(w}v)r^WS6?xi&rsktL7drj`
zUW=L8zFuQI4oE8eaXVU3L50cWF_`Q*hj)NeWjt9jpR$~G$yXwqGRNxBrM6*d%@g-I
z{QP9IX0$JeQd~3HeejtuuhJks#<mpm3@Et~xlU&HccWIV_AkcXA@~G17gg!}gq~}$
zThUCsdW<PmyB0IcPPh7d?CrRf+H8gN^}G$%uNTBs<Ez{U(>?emb^;|vo;=qf?;t_S
zyC0kL4I-k9LGXwQdv(M7>;|?hojO+0KKFT81j-Jx?i<UC+wz^W54E4;mE~PA%o63E
zQR;BRqgQDsNN|hHh>=-Z$PNVPf?S8HEC}#<wecg;G!GTR=VVrsjV0h^_p5uIm`fSe
zqd=EB8U-k@dEt(ecQ4?M(Iw>zQQvkLSo6DBpnK~;B=MG=S-KOh(<_s;$6Z@vGKz}$
z@R#+ojSer8HGKqo?KvGaD?FmKOS1En#42%B3ddID<79TpL;D8CL~LIt$U@P}is74Q
zS8Y0`{md_pN#6YSgub%gwp9A4Fz04!KjVy66J%W{o+nqJ`f*6f#?l$4y{oo+{!?}G
z?V4CJEk9#pIV*d$)NkM{w954&GvFrGNAF8_#DuhTkn_rs9^)P3sq@TnDWA#vhP@gG
z5W@G%dPtgQID9^~4At&qiO@h|No?r89Rs?=cq3P1XoI$QC|vp69Bol%Oc@rpUG7n+
z?b0&Kc!!?$Dfd%k<ztx5Laa(@tqNd<*<4{pH|S6*cM31wOFf?JvBu(LzZQ<Zd_m}{
zz@YuME-IfAjQbI1Azu2|E^9tN5uWI3dC|3Ztz0@+CUVx72(Bzl{uM7(8_6lJD(jv$
z51PUxS<iR$H4M{V+|Er5VJDT-HAg&gcTFW=or$@yTBlmlQ*&O~ZyVuKKhvZN!>5*9
z^1)zNeqa<SPU}OY<M+_1sb^cm^fT-3-9`OXCswRHbfMF_Iz9|qHM?9>q-*L`U#E>n
z-^-`U1u~UhGDXqVTWh%hy|i%O_N&l#SXz0P3+u@zHrSo{*TA<s`(+ti;r#QH0{B1Y
zYVqc_SZ3iiC75c&zq7vNkBy4CPIXvq&c^D*h$j`63>an1%hLsN?C)uh%)AqrBd<Ph
zYiKLp+agA1cd7bT8<refW%n)5Xp361@9D*oHK<njS6qmMs_jl4D>x8T3|;8uxLM*>
z$^8^l>|~X-GcP*vVw#tunZ+EJ4U&*1-k+CrL~1~K>i^5oWu^1mBW9v?lbK<J$8hdM
zajN0GYlu>j@{U!u&KQO2q;%fI-Wv)MrwAOAAH-s?XUJ|vt5(*1Z4b^J;Z2J?o@s>K
zT?9shtKT(A`nU)Tof{nPOM4rEToQWnv|f!4{zCHDb4_ew<)*!Pq6rr$9On}^$8^fk
zsHRu3H<GD*u3)_ybn<3f*=~}Gy-4}|_1`2Y`6EXN)QI;R@O5lfoZT{A{LL(|+SKm{
z!+C4irEC|8i$px%Yy_s)D>ji`vM-G+kIC;=J?7!~^s_gY@wc@!hA=tLb?gFHm+pXS
zNWxw+w71T7p6jS!LG4|cfOQRD4p%1-<6Y)yUkR8I$k7*Wz*Ui@Y4KGaC(22OGey1A
zW<jmq+j05VNA;KS9yE?&8o?8HAfqUCXcHxhLZZsIdt3^&_d*ZSF?wL}uezok46XZ5
z)=vHj50M6wp5#ulp$S(wtAR*UhD=r3ypDUE^%L-A*G%g<jZIpP`_Q9$AY)hT0lm*c
zz8cK7D~!uAr&jShmzq7NVZ8w7gTE>>L7_&o=DD%+tny|lUm5E9s(g!DQoC7P+Ma3N
zhc~nrF-t-*JtY(S7a>;W*=hwNnGQV%Nb&drRO!hzUQtWQCslTGqAe_c0O+<>_#>Sp
zH1HB_T;A{Pp<7fMc;Yn>=he+YH=XK@F(A%-biuR)<zCU;B**VUsca2!4Vt>bTM@x&
zGp;4;BpahnInm$f)!Od<wMXF`?H6kjZ664VQ)#DjghfEzpW&}UbsO|?ncY=@K08y{
zJPsQoLFAn(_Zi${@6ms~uo3~FZoghdSbh=`sap1*1-G1ryhFNfdDj1tP;MYYi#<qZ
zKtt2<Fsi2LE2(w8M2luuF>brADom-6t^6UcFK-k^Fjxgjl!HnqnA6=Tr5mx2p)Wb)
z@p!gkz$%ChhkA6CvnEUKqs(p93t5U-)52)=NLE45UKYhfZ*yC6li^jEO&3%~n@BFe
z<Rx);#+}m^IH_A7x?Hag(#=hpoyLFQGwL-sZ5|WYaIF1!uFFjoE?w4|4J>4ea9q9Y
zRf-@Z0q3<DO=V}Hsv40v7S3}PX+Y`VFjezz1B|T)9Jw11Z`EwSN*ns;;=y06uW1tF
z(CI?73l%`3P}R1EjHE~3m);}Zo2k3%)f2>3n7eW*@(DI!WPSkx=bO~2&kO_>%;_0!
zaohSz1<k7PKfaUr!BIDAaX1A&nAmdxZu)pSl0oM+@+$1wigc$<J2`_PH0+YKg`9-F
z{%GC`Vsff<=lPP)G#<oGrmlQOVCQr<4hhyXM(8gg>7k3RRambXc<^`z8g1_L<U$*s
zwu^3x6UNuy%odFooKMV+Bv()f*mBH}e(&23yslg(XXcX#3Yl22>uVCNQwo#+{&nrK
z#(LtUu8>_WB~sn9tF8VB99T!*6`qo)`Ku&Uy-v=sLlLI2Dk|Q+hE!o|a8C8(*1lg4
zxK(-VTIgh`fNiMW9v<KO&MQ>DQ@u<};}B^-{+8>T$rPQM_C#m|U1>%nls>xfJY))A
z{Y8>}28A%W{4e*Nw!xeRNe4SF9t@e5f?t+ib2rKoIjW~@VG^}RS9!DVY;5K@T@>@?
zuW4pwO@qL3AqpZooW1lOM9D5&>wGmVeq0MDJ`eA8BMpYs;a&&78yF6{r3VC!-xwPR
zXdmLVmv63W+X=W0*bKOu)Rs!SK^1?E+I299XL+uqgz@%{cn~Fqr~>tvx+9S46Znjq
zu{KlDnB^S18%1kbs$P{}#kr*ZzNycH(a^?VC?Y^=O%hUlY((=^GHDsvA|&!U)3r(_
z^F?#VC=?3iBJf5y)w&chu*^1N?58Yv<iE}4blt?&$wr`^N7&5VxR@>2<OYum1*g`Z
zUtbLnaE_ule9?FmiHmoRyZGQD^K4*6^M`R2KMs?JOHG)Yak5vlL_rhb_ok!QD&s8&
z9@33_+Ae$J2^h1gO|#)`jLgo}ML3mtY(a7GS#4smRvW7yGx0$;-3D`Cpn;n4b-`{9
zK1wSKaGOJm$*gI{hJ^1b78!x6xswKb7g`P(pHk#Tny%8t!UcXTTXv10Hu(L^s`mf!
zU7qbb1TtsEjUSGe+isRGZ8oiz-v0dz4rDxt^yTI1Lt}uWzCdZA^&f!E&+l6tsYr+i
zAdd^7Qj&Xt#^;O9%Naw0v)|mn+E$ap$qrlEuhJ8)maB|5D{PUYA_L6ZgVO4sqQRyv
z;={ryi&xLBy7heYs5JWK`6~g#dL>v5i3o-bn`;G6UjspE(1Y`t{n^<TSy4az_4Z<|
z0ln}y>5+T4{r2I*Rzr48AQZjCWwxQXt!QaUb(;0|lhm`N^3$YStE+G^A+LS~qRX$*
zn?`yIT$i;yqzN;GSu19S6B5I01h0Iy7P1K}6Mqi)_nNA`@Cf*ym5GnfJqJ5|QvbO1
z&|96ak;2BNv$hkgU*8>q17B=*%I-Btq+6>T9w@-Z8GGaE^E}iDIe;!I93F9KVgpQ>
zIG*SOQKWSk?-6VLTypRsrbX!Y1F7%|PAmxzo*W}-YZQOlyi*kVXRDzuw$CRft<c0$
zlE14%^4c;3$;?PP+#J@~3@fPE1k5^|Qn7cAVt3AQ)_9xf#;MItQEb7|m*7V3-T`&<
zb(v-oUhS}~im_9w`qxR_tdYj*p%EH2%GlZokI1OE&HUB>v)%IFmFA&A-5Z02S+*O!
z?7{F0u0Br7P%u;yf}P7Mma&gjelH_i+}K|z8^Jdh3!jQBL8HNb_qf1__SQ>2h(am|
zljGgwocPBPql$|65`j2bOg+(SVjlG57zgTBGaDLpuo6cgFzdkul{Xf5$SS_nI;}rt
z3zn0z@$FKIQ>8C7G*Los)Jg0-um|6uVjKTl%a%Rt-F(=ujNc6XB?0WgZV-39lJ`mF
zleF~Gwi_jC(Fr-y?z`w*;-4+AbzLfDBk81`{9AAs-|Xa|da_Jexrs><3H+`}`qJ=H
zwz4dUMT0nNVwRazDY)l^P*^U%ZxD{|I$L2#ZDJz1aaG60=bptN6%2Db)6mG%D^a7-
zNLTU9gTf%=_9R!IB#aZ#-A>;ZB^9~W>A03!6y$eAWE3by*N|`9A33@#GzeupFVuG2
z^7%>Etr8-n*NT$E&OXtKA6J@v@zDC(;J@nm?`%v4FP9W8kiUji`{DPR>E1jTpoN%D
zw@(3z!=s`RS%FK+F85vruR}B%dC9u~C476@WEl&7%uzw?(Qzdp?TjXU#V@$AOzt+r
z%q?Xw(8ptPYVul$AFjG4c)5-5W+6)%#OOwjp(u8F?|SL)<9WF0Q?lR?7fAIj7)_fq
zh8%f%==q@A2O8P3NDCd6nl*E1$Maqu*ln3!V2-|dZfji`lNRvdp-D#Gd_K!V&_EVS
zT;~0VyQA9ls9F3%$0qZZ{(=g@T>P0pi}#j!B$_~CrU390NtsS!dUp*_kT|F3hoZ9!
zkR~Ljc)ky1`Om$<@AvO%(+UP?^5P(mTOL2^IH{gi{MzvF3n8K;<Vr1<lLLVIU&3T0
z3IOsR-h|g(Bf<irUiNHz!n;MWdvgnAOgYir)tds7VG=xh*Vy69^NeQ}@CQIIQSwv@
zLNHTC`ebA!_@cQ?nSx-APUm-|V?-2)3qL1}abTxpIv-yLe4_i%D=qv!2Em-;hd2Vz
zvA<x}&nYl<JzNM*-}idJ9a~wG)Tm5D7#%dPJ1<_g%nl(ULK}_~eC?gOS`isuv}C-E
zIuCQIeacLTdbG8Fi*zfwnaX~13`sMFS0N=die?7>YBTRF`wc(VHf35W9?0vg|Hw&f
z)Hv?EpC1Ft7EmX`)6P+}-&7^nk?&Qlm>3A)I0N)4K8z{M`|e&e3)kg8^SuHBgtuZ{
z=MjdOywk@)7GlJNq4ezHdS)>4OHnc?@s_LG+M!{~=mq_n9v*C^*^chSl4Cahu5D41
zgq+j|<3*-SPN!yQdw3hCB2+x`vnhWGPK0DJ%wC7-;NkvBWUBl3@J!d2I+F?4Ow;;t
zn*oJ(J`#(%vYLVz^gR79AhF=2>PAB&2D9)j-5FD)q>F~m$)8J^LFk>ZCYl6mvTr%6
z)bg!UswQU3*6y>)euQPMse0iE_XDX;D@m8OL^!z8SWv1>phJPrrup%nGLt7EB_c;?
zN@o(bS>2~6uXj+z_%g)oInm*$hCJ#QR}JgKuEUDDeW;=1ej@t?Uz2NASaLx(A0$d0
z#?=X7Yd8wrbW$vuP^IIP6K|o%UCfzt*iNh6y9Vj*UDC#j=IN@g_}^2m<fE_VgH94L
zPuO1^^4UZM&Zt|C9u$#H=4MHUeEvB~na!Xvu^*}%GM-DM$yKDDb`TnGR7h*80R^gy
znj~ZT-wYkXpFHzNqpoz;KRhs`0Up)tMV}f^poST5)*?wne2RkIY$|<PTZ;K~P1rV4
zBDT#pUqI~VPN-@r;(jkD>Owz7j*-cvs4U3nO8*=P$LfL5O_J2F={d19H&k=KVP)ge
zJC;*Obug2~w)s{B;(-X!7{F>$@@&<97*8?D#D#s^_aI!n47CP!jxEw=X%~&IsM8OM
zhfrXR=wNYCy0N5+m4}vrH?*&-D2?P5mXcs7h@kVlPAlj;8(?f!i9LNH7pgHnl}4r<
zz6lN11RdT3?@yRD*OvHqzUbSyouAGSGlFl>Y6z`+eH+Qll#^Y;19QR3(fe^*8s*~W
zMy@pKW&Dgd@yN#&M?6k82Q0bb`TAFoOC8tz37y!IH3VlmO|6#)yZ8(-(P7stmfyW{
zx#L8GLk_lsV@G;i7!|o)D!vfGP?cfFFR&)4#A;%zvq9Bm_F8thE%J&;c@&T1Ya!07
z!bwW^YqejnaMTTG+yYIO@e!D9)Sp{Ivg9+|p@@ZtGoThv+wYf+rez$1REL%0P|<`%
zshW`*-Qi@G>OLFwcp~5P9gC!KAOuI28m@u5A6RK+sG;pDvY0gZb%lxEID<;DYu{r8
ze#A4=q+D(;p4U+YR<BPNyrZ=4#~E_3W78|`wKW{}aFVgifAwQViw6pL826%B-X5}f
z{s*8p@dwaaB!_gDW9{<f{4}_Hg>9Pp^wEX|bpxr|R_v53!u+?crfw81lPmitO8&!-
zBm2x%k+toJnR7fldfk?WUNpF8ESrjn8n8HRa{R3#zPQzt8I#?zn(3^I0o2tvrKMJE
zZ}#2J=xu{ul&Q{5iM+(R^+b`&%xV5&Z8Pmup-azK`PnaJ8j#NsT7^sjhYf;vx_6uQ
z7^+;y*6lS{irZZ|iPrp$R@ZRcrC)}Y#;$LN8>TTOwq>eH`xX~1g>6AqyK(i5ltmiZ
zi*USAGES95xqyy4oPsmzt|%4VE#hYvuTJDon-98dB9n||8unx9&+NJJPF0_OjLoOd
zOGX?2*_XYf(TTA=FxYK8LdU(~A@QDx{2ULBlqse6t%V%5-`Mi+++ijPvqUiMM}zQ`
zg^86e1#)xL^0hHrt#rDEx;{btqA<<0B%7z~uXwgc4zBZ+aZnO^+adT`rs%#(^iTFr
z70hiK{j>&bE`hg|!;50LrTm#J@M042cCYu9M}dZqO{S}XUF~PZ;LGkWc@@C5x>DJx
zC65@d@jgbCAgOHSy^A@uJ9P~Ya{)hAlvNCC-l$5PI-@l+9lHndI>)GFLZ<_?2+E^^
z>dQ7GeXd<q<}W<HZW=oSqlLAdbxwHr@|(tjT!Cb!e4rQ=qv8-3G#a)R^=EZH&!M*-
zRSr)bUB2r<SgquFZUDdo0?H!uY;5D}qk#j#6Dt*f?bV97Z@d`4@4=4>;k8)`j75a~
z-pv7i$vIDM&NyS;Ew6_$ot)~<94pP1o<C{%jz+~H{$sUj#d=>3yqd<dqdhjk=7=Qd
zYwLcd-@U^?Sk&xowNf1la^3eCR)=02+7>*uevOO9!f1`N&v)$6^V|`iSR+iE6BZKW
zlW7$8UdRYg*@WmQbLE@i^mGiMG?!zR`MvuT#LB44(<YmUWA|F}LSkM$_R;tav%B+M
zL<kj@_e>oPEZ(k`eYx87exbUgL!ce2)YZ?;J2L~dRKaGTQA-nZB*;PXejsny7FHDO
z>XI4iVB^L1doKK%1cE<Xn>TAOljERaQ}2=V-Cozp_HMFDqnMNPj)t>d#;An?gXd~@
zLyVDMplVUQ^&oGZ>RtPHmqzZI3bNEOy{iO@1s+pS$<$H5>&owW_-a+3fF?bM^i=Y6
z;VE!p4%Ci6UU!(kkD-?98+Pf(y)wfIuBv8k>#1}c4NtTgB>V0*!yD2U;ks9A(njAL
zVSsn&t}}J6X36=BGeuSP(2BA`DqBg$259o~gpOk><J%pP77yz7Bpmua$+$|>LXzH;
z_$5>S*;NGtzxit`=l(DpllilwSx4q#{#rXUi0X394*npaTu0`Y7@}8FIo%E{^QLV0
z_Z_tfFJ_35yhzdWjrBlWo`-V3DpFZF2EaK+ru5`FGiutUM74oKi|?K#_1$Vu1Z63-
zhfEue05^l)Mtz-=m<SDna^)Slh-S7Bbo%vm2ZU5YEFveZ?2sAcU`9+YmCAKqDQq;B
z$#HbZ?%h-!F-UY0Ri@;PfO~8qh6_^xJ6xbDGHWXbm-<8Os}FkID(N&_O1dnpWtH2l
z=fB_GC_SFna2%&(aH^GhhSG4Ou#~;Ee7hh}b1>%>sQNJ;_&94Jnyy=Cs}tg|TxWD2
z=Biv6=RqS1kttHqE9F!Nvk=|fNO>}`SypkAv8cpA<aAvR;1xy-Ed)M>C{8$xUJ{Mv
zdWYpe{2~ls?jv=MZnc%;P5xprYzewlo>^F023{NzEmO<FR<)q8v0MQeP4>E5FLfO-
zmrm*xs@mIeT$6ZmaRiar?`RC(8z3+}Lpj6?&3;(!Hws>9rdn20Jt&n<l4>W8#ZX}U
zMy#*?f7$S{E<4>-hZC8~L1DG+c0nvy$+sSL?*3e~%cRW$t$V9wzGmO-h8BJ;Aa8dN
zn28WtczB6~IXC-I_<s`g_ORN1Qs%a8Y7{mQ`3Vs!vvJ<u#>Vy4XdVWL`~*-(iLMUZ
zW<nE_BxFlHagE$<PEo`qO|y2ywVhT>3oqS0J9PARUDTU9du^sxz}L0GH3Ovz7t><p
zr$tduj99;(6i7{Y-`h&mjPh<aT=%+~U=AFhG2i~)dZq7@%%3@Fr_^8l#6Cg(AcKB^
z?)T3wUmJ?GU+qhftD{(FIp1iPD2hy#FEp8Z9a*FWbrIb9wBFynhoOIEaCqOD_H;<y
zbr7#xX92e#8fbSt^<7BKPj}UIX=tYZrSHsd$D$eVY9fgvShn=7OUg{>56rjIv!U~6
zy@FJgy|qSX>UxchNYzSx!ln=7TdArgt!R<;^Nfj#>9^lHlF>UNFA}GJ^<$JwYnYgQ
zwy92;%At-)j#ZH~&mufoku->`pXEKgSUyv7!KHQIYGt^cO^OKSbuFGQB^E}n_;%a|
zTr{h*QDaQl*5i`p=-yWtrBdfM8_KqwG=9sMI=Xr@7BbkgD<cht!3SNvgqlBNeYf!&
zCZj&s>-j-%R)tpH6%{&d`dy`hqDG{SR((}R2|-_!hd+~PuetE^CnAbgd3|D%#zx$g
z>%IwF7${)mjxd`ad#Bzwo}&)9k{yzC$`>Rv#V6~6+UfOQm%1f&dK*@~ZD5?chj%(c
ztj0@1PMOeTN~Ll3_=`N6fbq<?rE)HE`F!xW(GrI_lhmtIJ<^qd&p-d;PH_F2Opx2-
zN7QWc!Rp9{ovId#FIA8gCni9-91BU7G_4|U9v<kL8Vr;qS>wzD>p*pXZGs@X@i-pI
z+6ncEVs@W+aRw6!RWaZ;#U*{W%Z^7g4VvJ?o5L3)#(cX?c>E&(O-h#(H=ov5rD=To
z(R?AHb<3^jBm+nKj)CqAoMsN~Y$v5D8HDq3^By_M0hdgpL0qN^d7$FTfYxmBuM4}U
zD|FYSj%rDz+@tXCVqsIhnDE1W7S{u&kXriFyMs%B(tot(zuq%uYVQ?NIawC?*v!Dt
zXegNDI$G}EFc<HjWPJw`Dp72tpFUg^FVT%LW9b8dD^zp(V&s!KfxDSM&cjP20}09N
zPEo7P(oZ$sCAaZJ$$r_Y4KSynF)=&Ylx*SWlum-ggfSB-XfvdJ4tq61lOxkF{vAWd
zOz1f$*e9+wTzaLLMlsV=MGAAw6wHz>9)aolJnf?sdKg}rwI-ZiJ5oxpf@~VdWN}S-
z$YvqP;?DD5x8^?`CF-$>I`MZvH6ybx|Nb$_91?HlxYxNrT~mSl<JUWg7TlI@nsYmt
z+dD+0lJQ<4C}=uu|C?7YejyW^vB1LaIner32#SpL)Cwt=@XKi{rzSed6K{4lZ=%5x
zwI!kA=zT9z8xPY?l*V}Mgk3@+k4u(e)B2l(Ve=3k9l_?*2kr;At9GgeTdw%6uN+Cd
zelraf=45Qg21Xd&d?=d{(>gb`m^1ASJ}9+X#dX>R`Ma(6W|4atsR}`I;S=+bDM8B_
z;|#=S6aAP&wGST0bk%%vgLLeS=sn^6<paMGlWx%Wp_4wTK0*{R$~`Ugn?6SrZN34x
z)xKm``WqQz1NfkiH%?kD^f&hwhG_*inIa{+uICzu;QkKcP|t($0!_RM2AC<+6ttF1
zES|~`fj=e_xi%F@InSYUV>{cBse=^j6=5ttpY)KjX+>T(Q~iY-&Tc@4XJj*tvk*ax
zhs@dLp;evosFlLO3jyM+jT5KxZv2j-MR-gaL<gBPW^-mZTK&@hF?jEK_pF9Wdsd&{
zi@)``8psdrKgsI9O~B-sQ3fh)jzhb_x`d~hfm~EOAKt~|sLC3Ne@)dW)KO$xn|DBM
z`>dRM^KR3hjHB3%o$XZrg^?Zu&k2>7Bb@r<h+fQl31bLXGx6LsXKxBGwp>$0aJQJJ
z@`0yrcTKNrU8#f==U*QobqV<$E2lJ)yv<Z*#7os&gC8#p^YMkY@Fe5b;qoQ&KT90J
z1Q^|DR(oo+sqBt&pDWe{O06stM$EHi743`_EkHb$M1ch8>rFsoFRh?;IoZL^oN24^
z8Z<BIWT|55FUJW{*dtR6sIX?-c&2W%pop;(TfFQcE)Pi1uJdkJyBF6$pN!~t!s6iI
zn0Hae$}W1zyv(NJGhUB**c3JQaVO_k3S_cPZ1Xif<Sgcedj_hJh;Kk;o>gm|mjQw~
zd;WNu^C#M;g0N(N=60cdZyD~}lQY0@tV93tCEc=SZ3ptKj<CQ(&yvtMa>V+f@Mq+&
z@pA?*CvNw=L@?YeCrf|6fX?|u0|thZlaLA%8r{zUoI>Llengp;eniQR>=e&b4_ThM
z-OB+c?&Xm9`wp+p90A;u^LrGCaGmM(YCC$6%dKCNi(zx=vNC{p4<Lys4gwJQHz^mn
zOHBD9jBiD`N0)y4K@$kmTnK;~$%TGRVtRh#89#s7LZOfMA3(E|@_AWOc&lj{ddd7g
z^Q|$=>{}7vjynDp{G&=~pn6G)SX}oeg605XsYHLN9e_$(4)f|(>X&bG=;>Nb{wxy<
z^f+GJ?BsErGeRzc9}8(5xSw0;umLA5)Ou8bs+;PCg#Ll}nF{&k8+Dgx0fqRu$~o=+
zc!{}#3H!y}q5TR1=fRdWgpBpmmn{=ezjFBXRyQ7n8u;DUsyQ->@m*CaRj>;&8Z=x}
z7tK^sRfv=9`R!YJGS~!Nr7?DMhZYn3z8EJ%27lBJgI--Yh;u3JZnhC<o&x=fewZ1B
zNiuX`s%r6MQ-u_Ml|W>qH&iUy*3Y#DuHMwggVJMXjg+Y5t5B;^@(MPzW9}%C+2*k(
zSQb~RCzv`{rl_L#aPWNVqbzk*yF8lpddlr(>s!EPgx?;ndiOT=XGdtCu`zlyrkw8|
zATvEal<nFc^zmd0`9xE~vQ6eGz4$8`3Oy1H5yntID+UBRM96Y26X^py&R+z1Ps>5S
zLL);=c;kEr01<_>t7Bi0BVTwmiVRjY#i%w2>sKh8CSsbGk@_gIyw>MXS0ER(TIRU7
z1<C~9p0H3B#m<lWNUvYs#3MD;{roF*B)()9zYpG^=ImQPj7wL9;~M|o5wq>dbGxGc
z!vz3D>lT3BSUU3Dj;wya1HegmnFKH|P4E1P{5SIi;)C00I%V+U^ePD%fZ=2Re9s%Y
z3*g)UrTM-2dj?{or^!SXR{VmzOHM_BawaINgt9@ehdR7|Cz8tFOol&ua>R2s5K=W=
zwD}}Lj)f&0la}V^8=pP7UUpOa)!`4onZ1j+w#^X{eU+bRhYnc@a{kc+cw;6_r4l8|
zWPZ5E)GQ8|{viDH#>L-%rl&Z?gdL?c;wRyJBHbPPjA!7H48*|j(zFJ`%<8qMW>Le7
zYp$ZJ0xD7YXv2vXJ<GG$=`??o)sgM|zo>h!sHVDiZxmk@l`2TDf`HO{M^q4`g_=+j
zigXAeNJpA4(yLNJ=tw7l&<P!s-jNbo=)Lz|^zi>~_BYNr<J<e(owKgyNang&BXiBn
z=lQjo?J#*-r9NaZd5RtBtXE}*MhWaD*MMSKwY8tAwO;w9$@sDemBV|PYF(6aeKH)T
zpODzUuoQJ`%*lz9s8~3h@AKGKdK6o@0%_IxS*SVjNy`EL*e2pV<ze;WocRG|fVE-?
zgZgJ}!&-ZNT!%Q4S%jlP)3}PppQb437tKLKg{eGa?G21f`GMn2$4sLIt_CrS{j6VP
z75S`0Iy~3tGNcBaZqi4{*7s6}nKHRlQlN35j{2Nt?p7_`)g(cMt7&T{)$d_#a^-qk
zi+sbc%3FqNm0A>VYnQO#pGu{o)@a>49Ln{@7<V;;Z)e<%>Rm_3d2(b)&D_pgE@`i}
zZ-ATE5vm#+LC!@$AoWc!;QZ9XWprX_P^iIU8h?q|Uztt2*YPz6So1B}?Pb@s0Qjyk
zJbM*O-AL=STgw8x3$MQ;4u0|FX1$}sIW*neiJZ1w|D9=znW^t)p!VXciU7Gfh8Uo#
z(bn>%n%<-R3K;>wV1}$GtD?w8HTOJe8i<zAQi}U4_ewn*J1}L{MsDoab7R)x#YIy(
z<mJqy?MfB@UxK9tjf(=W7h6UlZvrF9j}+$r2oU%L%cN`Nv{|I91`s7{Mj+V5<uj?n
zv~#l5^&-?MEZOoI)Y8B+V%VTqjG9s6QPd!h*;cAHD{Fs<_{TpcLaa2-4XoR4sb2`o
zrn^dwlftzzAE0iuj@`d!pde1M8r?;MmlJnYHV>eARI8_!2SMj%7rUaGHY%X)&uH|u
zQE1|6`2?L)BSTvYiiL7c3gK;B>loanV*=UwVx?d5<dhgXeE!a+l>VG3?gs3CM1|G)
z?WVyG%c@RdT)m(*>g|(bq-A!I<{GWxV;V6KNS4C=bE?<JQ)13X<SxL;?{OZj`5A|B
zCH#__-%-6es9eyE6-dDhspX_y&;=O95nlcz?~rw@Ii8h>`cg4Z|7Ou@;n8+h<yR`1
zn1Qqqy)6n$&44#Ari_DSNIDtgDXpXpi7kciaIYawSa&oV0)w1%Cw3f9%%4$<%Xx8M
zgR~&y5evIkyZSyI*4AvCDUxfgSiP3r+V)uipbDBOJtX39(h`!C%;p|J+MxjdvE7@c
zE0{Y~ia&W@35fMDj<d+t_%5!}pV%{7AkF3h>iZ`Bx~pG27hH0k^YR(y+CynOG~v;6
zDJ^krSDJ%nlL-smG;7C%uo0&E7xgCejXi&ItDlpE!Kx18Mfbno04nUfIwB|b>|9Xn
z4axywUbc~l-VnAq{nZQ<sS*nG$@{WyOVjvl);kcOS7Db}jK0P`)ag`9mk)n3#izgd
zuRn_J+$i7J$fz^*Ya~B9n5E%gx@e=^p^xWDsT!z7oDJ*=tndVf5;n768&bd#zaa4E
zg-B?`R~bC-)rsYl9_#&;oV@W~DaM04Ir)@=_L<P5Jr#l*8@d0!Rl0my_3d)W==?8%
zO``pm<=!g-|FLBCnyhJQP@gzTi~SW%r`ac)xv%dbx!N1Gu8yOY6dcWC;Yr=U>+1;K
z$T^W_XHqxbS}UVNy7LEmA9(wQ)=$tRM_uG_G1|_I;_Q0`PnHR61An&Pb!fT%5>$H9
zzF7(t{6Rjn?yxGcoVyF0mL%Kytjqf{J*^)A4CVyHkM-sYnjmEEW5jZFDw%s9^ja*H
zH&&H1P22N=wbg~*e8i-jgUWq&_Pg_NHaxUV5rf(;%k1AfUZ1UW6?|}djG-IUiFV|E
zKZ9lpVaXYcs(iVq8OY9*AjuXAQDNBdvOhOLX-*~od*%+9)H112c7sK96luAr1I8c$
z(1kULYFC3LC2_YS^U+x6;f?qiN5V1Bk_b<YT+G^&kLHk{34aL$_2F+OlZI|SEneI|
zDZOLprU`}|{{b7v+(j9avHvCL<oHPVPVVuyqe5G!+wbEqZy+4^8p`e;XC2H`5<AXJ
zZpOjNhS56KGg7LdNvKF<yHwxHwEPU=aTwmCKlrtsI6L3T&(@S%t#*C;1mG4<!gYYk
z=zz(KyxHq19W5l%$#zHt_qc+g#sT5W^7C5Pyezjf-2iK0QZswgYK7tmWzKJ}FI%-4
z<rP(c+Mu5BZ>)Xv>uRMqNIDGVq`T{!cb_R{O_d3&lcUCL%UhV?S5%Bs645;_Iu<ix
zcw#aPG%m9`Pv7oDEsr6~ZN>n|7(xP$I>+ccVtK?;f{yMJ8osN<x60!4?0%Pjp>i&w
z7)3^@Mb+mue75<(utn$eb>bnvOQ!jBUe>=!K9o7az~ot?2|)318Jx7nIC`n%ZE;WF
zu-qhF%Jgx?^!gNQQn5^*<93wCpqNQb%)RN_DU;%9nhwqBA9#^HpJU3Mb}$85IWSYF
zluB9X6g}TC!Uoy;KKr@Wf3bXtMOuvf_xUeYhn&e@!FM9)51na#p2L34U4K^Y<rxHj
zxw$s(`m#WHhdeIHbP+i9E&QhEHQ!T=ii&F5UZ#M4D5Y9^LpPFr6c^UrSJ>TsZ@A@g
zW$V*=tT7JOUutbLTUBB_W_IQ(73ea1-YCh7O(Zo0$)qcMbCQXKa!lIQI&J5J7^->F
zwTjiDv2dM*U&T0k@iPG;81No?jd(FLD7|JclIJEky0!(<htqG+!=fdtmm?B(>R0pJ
z>4@!95~REwLp6TWHz((Euj;LhIWh?}4yJP_SM>(1Z2|gXe#l5-pIN}UU~bu~j&uz%
zlRlBiam%DmdJv;0DP}u{xyf8On5T7o$~_gm;gSxlA9d|+w|5kS8(~}Of*moodjVrt
zGM^Yfc~K15$7`&%{0UEPa4y}IGI(qD=jL9+<Ny3G{{Q2;A5%I(d;9JD2T@&OBe5nE
zad<}h7pcJxtb^~dajam&22t3#dPY6cKt%K+a+Mx9eiwDIKUtVbefiJi%7g90<2#n4
z6W$Z~RLXv$q}fs4*|kLw^7Vf`yOu|z`0?lW>#*JBna5O=;6r)qHdFZknklZ1!Ku%I
zhJc0*P8->mysLjig_9ZFox`8HDB>4Sh>y27JPbanecAZZb^7Tq0bkx!U#S{1RsD%P
z%iAx9p4QjLgrAk~9Wz;V!UX&F7)<Hfq9&T|lO2q-=BSm<BK`WrpW2Ig8;dc_`mKY|
zv+mx{bh1AvR;+gGIWJXg53&&wmK%@<=OhyC+N@knAL(TD4=&y!xgAZU3O<^<%a9h3
zv6J+pqea<sKt*jq$4`JEL_48SH9afk9goLDp4@)O`iqzI9U_F5XYgJ7p0%+^_8OPf
z`kmtU4aG~ATLFQ!!GLJkQr@>^GOq@attu`aYAI?S-jT3n$xo1Ka1$FPfPzaURJ>hD
zv;oJTzKe9hRNlsmKQiD+{p@lZnyPV^CF;EW#M$6}DGFCFPEO|hbuZrjO)%~81pjN4
zet)NRC#upgdt{Hh*FRVt{e~>-o_JEJGD@|8lh>KnF>!nKl+gC@B_&_ui?-`FDrd@t
zg6VPoU^uR%sA35%A2zlnP83QpRn0vz@2sl87}LfpihB<=4bw#HJhk24bSf<nvY4=5
zo4`MUc0d@>7p(t!_1G+6DR9~NlQi4*1bjj-7yE>@s_Ci(K5O(PkcT1@z*$kM3WeqM
z=KqSnmr@0F+br&sLit@5co>g&M;;y3%8{vfWUY#d)1M49YaKY;bK<>EUvaq~qtCNs
zt=rqiL*3+iJzZZoUfl7%6nu*E^fP@QfdL_xu9pi$;2)y4_iFJ?&(vhjdbfR>U;7<D
zYT1>;zl`5S(46M)y&e;|ZlgGSypkUL^+HEI-<U4?Q+wF5^UC;qv_Y|UVHKl?Sj}Wu
zJ;tVpS$9=Nhg)dNWd!>y1bgk<T({C>n<Nae_cFR8-QzaK`ukTI<t_~Zo%<p>Xm;9-
zIr4%0wsZ;0R^b)Q(Z;zGNP2#m&~Z0!QhTfM{AuG0*{d;4-u1IbU$AMrKYL9Fi^z%d
zmPlKoyb{l>g*sQ7%0l!hz;57bnmng!<=pOIL-d3|YW7(8KYX1f-eR5LhzP{HTFnOj
z4HZQhf%4?Jvwh(8d|S(N=$En&t;9WvgK}9*EMztcBmE$kn<m#6{uX$568<*4%MLbP
zxv@|`O&{!C1ZsBdQPPJxGI+1HNfEPKJEud{JD2bd&`jc+J~(pKVQ<+|630rsZxyX3
zmI5*z7^q5d>!f9XK{FmqyZlf!KiZgy7GR3f_)N+0DGKk5w5;vRwTb7X*bs&Lc=zqQ
z&@Xi>I|)6KxiqX0WlM}8^~V2ds4$r#oX(z+@Z;)0w|k7&uH#QL;-<3)Gc|(nI73}i
zbDh=FP6i{HLy4or=}#;5*=BJ!J|%@D3Py8O@WHOKbMf<L0rcgsdaQx7bwAIykm_$f
z1`-_0Zs%JQCM-zV$Brq?CP*kon&&slry1Q=g;vUFPq8Zc+QP!IW~RRrR_D|^i@%ZS
zX9i_?SxoUq`u)b-825U(l);DUH2G2hhpy9!a-#sEA4_V~LtJB#ZXK9`EH|RVI`Zi~
z;|iGvMKc$tFG;H~HP#Jhr)#98LJd@(t{D#2dOP~L8wzXY7%liKRT^PMt@DUwpwCx)
z$$PlFNCD}nI-HYSXz^M3N>zKU?$aCY;t$EpXN(c()RI)`lNP4hS3=fRmaRfPbMMN#
zD~c4#og>AkGOT;upyN~v4;c_3Kn@74Y?DPeU01@8u#!@8>4uy_)j+4K)X}AcqQ%#Z
zIBFL<wcV?$=^KfgUpQacgIW>xf!13@N<ZPinx?=R{rlz6@Ja2Wm`8P`Qn1drWr@aI
z(_*(i$6Y3mjzea8FeKwR$WDpSse-kt@>PTCuFutRFaNDcyR0oKoCj3oj@zpJ>}Occ
zK2b#FQqo!Q&gCRaN3Hl;Y*k|CDF33XZtcXXkE;>Rli^ES(!8)=NI<Zu>?;vy?P*I$
z)ZL{qV%B`JTYzR+oE|@A$=LnnXYsJVa656f+6p{OGlQ;vzl<$d?fZVhz{wz!iZm|L
zvy>VwO2_+q)QT@-8-;dK$F0!ID64h3geV4{R(oj28q9&P;q6yHcHNpggW~<bpw?wh
z6EU<e&#l^zkdfSqwtY&lN&S7s>nDemq^>U(U#<_aeDZH>NF^zEB6ofE#qe|j<ql?w
zNUw2O%BKU`554(-=?n!HsBmS>#?x5Ym{^|_*8m6+XiT+U9Qx{$(rS~GA>$WE(&?=D
z*GY-Y0>=aL93_#Ux?!OmW6`S3F1MxNi$cPLeo&o*Fgp*ppaVQIRur;OsJ&ZTK`Lxo
zln;(&Hmp9b$e#Ddza%+o84CG0Q6d>)nb0W^VNn7@+U!lW_I9XuY)%G?r^&g%Qd|;M
zlvE>4+`B)mb3YOB2FN2lhrWk|`wOfTJmKd?N3NvqFZ(xO@7v<ZWsH0~<O*KMB-wlR
zw;WK+NEa=+QeVr=VT?hWR16R3MR|qjtL$SrjEU<|G4Bytg%E!(kczw12%D;X0sW`Z
zXr-naXTp%na(qYpjxHV<d$R{|gLya*R7?bmZ_^yg<TD^du8VNIhldYWda%WxL1TkR
zEyZ(w36a@!o^&>w{(@9%N3nQU#Iuu?WKQhvfTM3;*X{&z65n7uK0|`o<!R?FBB|JC
zWXdweIu@?3nwWAflIjoqXwR<Ze>z@J;;{vapt85wP90E87hxPpfvT53Sy74|HaBu_
zf&7{J#5Te->ji9(u9y!OZ)9w}(iC0$OK`W><c^^4(xfJH#$ZXfFil21R9I7q+AL@N
za=Ec0Mpoc=_DofKeP49KYKlhR=hJqU%Sz8yV2{A>9s9_{v8jt{B%fPxzHO7NI(;_e
zRqkB;bBVfVuOxQ6oW}F^W)7DmRw{o?a8nuS=2Tsw`qG@|)#1e?ZMMWxlTTvhNNYu(
zB{QPkGlS1ou0TBQ`pIwuczq;74L>NX5l9pcZ%x-tyg@Q}0y=d^?H$!VkM{v={_Nfr
zE+L--zs18gSMl-}oFDpG4EJqI(?pBMwoa6xdYG$I6fo52N1ri6wM3rNOc{5z7o~;j
zyPQ+IS>laZZ_y!<+*&tg^#A0jXxC8LWf$Y-?CP+_VBYz{a(2jVO8n-1hn#e=HrEJ-
zi+!*H6QR#J<Hm5m^leSHh6dOj&IAqAtmdq9a8Jm(B&&eJ?!2(C0X|$lvJ07({YJUQ
z{MlFHtPFYwn#8D{i3KeK3&7h96zHJ6@ubaM1DDgwyY_r=KzP@!At{``x3gLYJ(u|r
zz_u(?5{DWC2YW+pkPIkb01e<EkCF7?hq_ktGwOb%mAdfnEVR)^5gG|npZ8`vAlUnV
zdmd&W-^H=Nd2(#UfGb&k)gzK%{qwh($r(psV*Hj4H^+NFWnZa{`+H6;D2k#*Ry=RW
zRrAJGP1W;9(Z%}q5~u7%W~HEyI)!dBu&oYgxj{yUe7tJzD0_oUjvi_#X<K_?bn{@j
zMZjuGvzmp^#>8*^DITihjNUe)+|2kQ(DDg7oNPwyYCA_g57Ln6!!ldWxxyLvY78?!
zeLSt!%vl`8yx)AL?H%nhE28shhELg<wGhXxub#O3E?!#5q2h%byk2=N)#vEK$iKnk
zeyuF8FB?<pu4>a(5<m3!{>_tqW{G5NOk1n3q0$q$shIbNw&LEhl2m_R^c{%xk90P`
zW4=)F$IPdti}sDNx|7Lg>&heZP@BSJ>-FX31+rK#OU{;xW0KUr1WJQsEHC~NOlgYo
z{w1)oqfecR_Q31Y^fboKonu^(4$~6;szRD8^#U?8o%ScthfNBtP)_AiBlOVsTw$_w
zKtZR!agP>1e?^-*_*+P5Fx)O}rs(=$33XT|R32tZGiQJd^E4?%++?obml%H?&8D+w
ztl#D!!7PB6&@HH!Y2u2NPQkQCaA1=@Up#|v=(`@_?cwZ1^a2SE^4p&#n=@Fjgb=LN
zZgS5cPR;sG!*X2OU_dFfbtj8|ZcfOH=Z6?{)Dh($AF|^ZJd$-vAfPR^En&`4+?|<U
zJQq`<E>8cUv{RRQ9T2BfU4h-J974o@ln<_QfR1Z=7*C2hjEYHiDDzwF0y}r)1j|u?
zS*c7vfwcFob6Jq0bd6B5_=oC_C?s`<em4FvS1K%`u*4#sX^Usd01gpduWVWVWUh-?
zv|C+D<_dPTBvh!p4)mgS)r_Q!j`5VI9bTPQxA?L49jaN;q5DjfU7FQM%?4jQ1EOVA
zi`NR&HVwkO6oM@WNbLtDC*8#^FbN4<WmzcGERnKB%%VIY6`D%&Ii$1N%W#P$2IN0k
zL<-EIdU?PX9Vp%PvK#tC%U<iwcnYTp6R-Q*U;q7|Yy<x_=alJfp+>HuL6d&kRrVNn
zVsxK&0Ox~X0kLkNeT7AHP%Q$RCvt7)%u&TnXwR+-1jqgvXG{!`4J)K{)gBS6nuQ|y
zg`f1UzziwYe<(zfg3uG}@{BV$eY7-%3($CIH6!doS1;DQ4noOP+^#Im@6q(#bgfR%
z?aqvJEh2P1D}nFWrcLa!5SzVrpMH--JqGv9xz!QfBrz4F8mVD@6-B7Bkst$(7S8XF
zyCUfygSxAN6TYhHoeLfkoavwW4@*|2o6&9TE4%p$hFc#h)5Heq$E9PsVtQ0J&s$Uf
z{l2~x?{Zb1Y*VLCR7defF~wb7ar$kiR3t;+r~_y!`I2^Yiy~c+hN_6JoK3+YtNy@x
zv3VC%?_cDsB$_Zk{j%G%<o?1i-UaB*aed+Q`Cle?mf+Qe2c|6YCcYZRGuNt){XG3N
z>>tngz#^6q8eBV0B-(eniEm2S;*DkiU(+FEgh?_t56M{%AA1#d*QHVj!9%}Z=N1^8
z-Y4;AH2kL4JX9|`N)~?TRyFvtz1D+YvofIEz?zd^T=JisH?LZTXA}iVpZCd#Nvo(5
zWkL`zvZ+IMrP8-Sn4lrWq0MG>@@0l`hIX;_3I%2NGEQ^_!>}0{<HE1-KbUy$qPZx;
znDJ<yQ-8ELCc(92wVOXMe4LO+o4)Q~Z7*w%>O+mn+OSOx$5MUOegZg3aVAwgEg_@i
zE$Tz0c<HhJQ{P8J(a-IMSU1ouj3oB7k0*ruT*xYN5l}Ub3XN6=M8>IRD&R9uZe&`-
z_w~(6Q5PLpQ-Wv4QqF`uU5HNQ82FA@pflcRWQg%D??vt`EB5j=0=1Tc>Ax@KpJnY?
z{`xlYU#s!|I4VrAo<MmXp4_9BvO-mcx*Q2|T(YJS;`47KTh8TjWddV?@d^p)<ho)h
zaoO^#6tT#pTeb<aPsqjPhPE@skz-C5EmvNn1lYRf#ssnvsu-@(1Lm(bukJ<U{?^<p
zjDDFNXJuvmv%621wzL*MzRzhE@U@iiYZe#R0~?I+ii){_#q3wO8zr+n?<l!OSVH1>
zoKa%Pm&Z|#bv${s)^3OO`}1yqYmr)y!PR=tf(i2kk{jCwvZEK!#ew4p56r&Th%@_X
zx~<Zk$E_N0T4A$l1H|oak1?=$Zv#}G&Y0p7^Bm>0K~f=T!qm3YokkL|2AZDrE;X4V
z>@btNXv6}6+rG*H;1-QJc4j40@Vsx*+sT59*Vt`iIK701MrRjYO%^GlVPkzA&Ax=0
ztpfWgms8aIj!1yg11Af570d355PmtluS4(Dr=}+F{l^(Bs&@o^Ei!fAg!5vL>C&&0
z<RPC`+WQKK4xu1fNr`DV#Pl0!kp-#EU{HN9`#o`Yaeb7sdC&X$2CX>mn8xX8dv*1o
zP$O`5UOcDz?(#wltT)nUwOj)s5!9njUfu}nu||<ncGuKPrmBa{v5WGn^76Zh3ixUS
zNVsKcH#bOmnUxr&vtrg8$T5w_)0a;Rr)XU#1|}Ro)SAy7<`2hAM7Zpp!zH+nxO;W<
z0XPMFfPjL=_s{_uvHWko`3?@&8TMT^W;M>5%S(vA1ff#oq2~g;7N=8c6@0{M6oZ%<
zP%w%_B$#Ez%h^&SG#?0!pB)jGKxlqrce~`_TCQlK$<?*>H<gon?X+N?Xfy9ioXk@s
zOI(qhN@Co0#W++0YL{WI{E(~rqLOONg_Fwc@YhnmL@(fWdO+<P8~?%gezPEZVl((4
zqtpDChZo60y)mvLs~NI{pyPi%<u*k&B4RpL&1Vi6v*$hDPjTxmSDNJ1@W2eRdO2#f
zdELTKiR6~;JEs$o)84Y5>)EHAzLY+wzik8<Ir><Iny0H~*~G6IcZP3)2~5>$8h@Bm
zfD4fzrIt`L)1w~}$%mQ;|HoJ6zrOrui9Gr9Pue(dR_Eh?1X)-b>#u45EnqPElk0tL
z2PUY-()Ul@c&}zp?3K)R+yQV87CZ5IiG5A{aA>2)1-JuLk3PPQ*@{shAguNAb{+ku
z5B;QWlc)@mI?l`ZT&m4|&dK>FLZ5PxqWM(*jupk3i`g<pl~181a3+-+5lTF%y73f`
zw39WQMN3;!tZ|`!5=J}dzO{Z5(t^oJca><DrQIP)el14l!5Rt_?Kqdo(2q=%EEoD5
zAn~xA#^GCjB)sc;5)5lA0Ja*-RY_PeK2ks5dw+FqJMi?SiT~~DZTZMC;=XOh8$h?Y
z=@(-+nb-W+Mp8~s&@t>h2@yps7iwKWg(E8}wY@1&sjz}Bmt2j6;CEV4oLro-v{2LA
z-c+Cc<(inEmW;m=i+9ya7BpW2rqztTIcr+*zuZzjvwv>vcB5#*JEI@b3>+_|`@ZK|
zLzJtwLFVAkm+Z&$F<CT=H1PFhPH)J)b<sxxX^;DDn10{cU%<PyJ|F=_N_OibT$pNh
z{$CC_RTgRZOjSo=X8yQ+NhKTd*>J;Xhw~ZJ=NtbY<BB#Jf|kl%vp<foqH7*UrSr+w
z9oT#mA9?4mJK%u}vGt7Dms=2?+^ZS2Nk?w)f4IAKlXAzi>vd?_=>eT8jK13_hcN<d
z_3_RwJ!E&jPWp6+<M>gupYNY<@P}=^j51i<@xbh+)x&zD?K@cXBz_Txde!WVm<}H(
z^5@R{S8nn@7A^z~e=c+x{JzfYd7NC@js7Lb-@aG<Otyez|LD;wuCnX3VQQtT1l{M<
zndj9o^^x1Q2=_ycJV<Rk<wC06pv6rE?-@V9osd8B>tu22F!uyW@+WGBa^@!~ZI9UM
zNZg`I0<AJHt9Nr)M|$Eidm7#T66~xHd>JTpr2CVok?;qwT=18`_Br8@JQ;SHseLyv
z*Je8B+ob*1Nw+nEu`*>+^f<YmV2U@iyS;m|_}1R@-W;-HqX0pw&&e%Tk$;^6phAa6
zUo6%!cDx-YZmq??GAis0J@c{N+B!^YwBFm+n;dn~hN^)iX@%#)R>GX{w|q(`;=F>7
zIhZ0`N%_0jK;=(B-YGe9vqexGE#;Z|?ju`?e=Z~qZBsOYmQBr%F^)X}Mgx0{Kv*%F
zF*Z%+&r>5?9!}HPq-h}cOZtii0a<LcD#BK5Ghc2;JWtE|Qk1`6w@^`uFt6q|Ib}AE
z{saYac{wIjDLSRpQ8-%YN&RXLNU!&5uUfZ8AmT$*RWF?QJ_W!1_bun8u@_#^!&WoU
z{PB-W(4Y+h@`+5VKO@Eh)5-Cv_eI8O!-fTx6&y*N+VnXu`uhbsw!~h=7>|!!DWTQJ
z7`|F7)xpp?SF-w5ykg;$qYT9<c`}b#<D2nL>-COzsFQWG$d}fhpX6hBqJvD@!!GlD
zGoez!$x4@xHqs^yQWcA=kDc{#MNRLDz+TasKcn6hPm{|UM2#N(_>D&Ny74fN>?k8e
z=Q=dY!h9+cdPg3uh1xnwPIpy0*roq-GeE;@^MS&-hTn4*935^*og8J|-q0N2tiwx{
zoWPqg@f`<WDji|WHTJ+bP9ws-ezW5WV~zc%n)ws-?GHCQ_(&_Dri_W&>f~wZfF4uJ
z{H3p(>Sm48h^dgLWtX5}CDn_-q-d5r-CB^Ew9PUx>Q=4MZYfX1uXEEFsQh!}89~MT
zH%1eQe{scpkJ`LB2~Yd;MdqcDrg?a(P3^LV`@6iI`zVtMYZupOiA~6bQ>Zr2kaB^@
zVp>nMQZ58&cuE|CQp@kZe-@h@>=s5x&-(!55l<6H{$#$j!95(#trD&B%@#e9Tg|<a
zrKW2X61MTV9h*nbsm4Kn^YqbTrnNEcnO@~fdlS9g&V{8IuybqRCqPvRhkS_n{G_CR
zdtpxvFQ&~e`4#+i8(bh+@4XSfd_vYq*)^Qf`1u>1+fbIL!45EedCM3|P{oAP>K+ZX
zIcY=KnEs+gu;McZNh;)CQesm-N-x()f=LS49S5!Q{HSab%dT!Jk-swZ-3md+KydM{
zWhia>$)BdXq__b6r2D7JliN3&x2Y{V?0u3q+S4D&ySD9FkJ~Hv{|xZmC75<0P6%$x
zMfU#~+>rbwFq#p$yAAmE(Yozo^<M&H!K?WlK<r-vCfbYpGX8`wj|&bZZ&PdQ{t{40
z`w@IRZrz-c4wG_Bx=d9ovuV!+|Llk5L#l4t)UXTPX=x&z5gcNW*uCpS5($79M3Qp1
z*RM8cM(y2beAsthjKNwPdX5G!{z<r?0^bTA9PF2)Qii<3deGZEx^EZHbIK=M?W{OY
zEz%Go;!ueePpOs|t;cvLNSahbk<^+d{_ZVi`QyM2h4*(QIFQSx*YUT^epRUi{)0!=
z)c!6_|3EcmWMUd`I}wYiP%|e=Io;voZRTeSX7+^!n`t@+ni?4lq_Kp1>_$2s(%z6^
z4R4Z9Tc*K?VC3Tg{{25b+!xDukgc`7G0<D}dZZWTmQwZlF)oVcUT)MSMPXbH`<fNP
zzWYx1^6CFU_afw3QR*ZB^H4Hh<1sHWIp`ClU42G-XUTB*slPMQfzknCNDUSzDHP7p
zF)csTBVQ8Ma$$Q`S|dZPpq6{XwJI^O^_C=X=}m}w_1)<iS2*5;Nswbl>IAS;sb~CZ
zW!yqJ53~5?I8v)Yo!`)wQ%-qgQA0O#sCB0bM{7pD#%slW)zv3_CL|~*eU5*TT91(L
z4-PZ>-cQJFV6I>Oh8lhupFhG6wanmE18qjoHYTR!P1CqHs7Eq9l#v<8&3-78Zf^uM
z8g1(MA@^wKUZpKNopYrF*W~L2Ux}3Z6OGq->Zx9)73QM<<b)Z1bJg^&^PQ{A5i0=E
z*XO|@!y9fBt6JWH4)1>8n>PT^?_)@UoBBpBKFksva@JtuvefhnVFN4foWWQQQqGQ?
z=j!g^p}YB`5ew0pDd}s&>;p}7rQrC&eeSxkjj=7`>MkA=aI}CO{npgP;uq_E{?Amu
z<Qi>g`J`}Wog{Yz=&V5fbc08;Z4*Lw-9hZ9@tFXq5j7;ldej3_y-ike+M<^Ku8uAb
zV$l9_OUHV`zKb~`)#H|r)X<}BT~H)dRh0VldO48-bddk6Q!2jfu2WYA$ulr8FaNl(
zUAClERg7ONRnIFAL-2wNJZp4XN{D>;D)+HFiFUd8vRu6ls)aXiweD9wwU@N%d7>Ys
zmS!&w<L_S+#JhW2m<TfE)tD6+dCrPEb?L{ji=CB(&V;jZOYPV9cS4-C?l7{qP}E|;
zB525R*S&Ba2i7ac^xOBHk<WVMnQ%CS0|7;Z1tF)kv%)WPl$4Wos?<9d8KTdg59Ibc
zbKB%ljX$!&a(ZddX6PL_aWshJS^E#y6;g=af7fPkxaa?BoZH3T&@sm7U&N=GBi6P=
z*e2g#-LAU@=T9KC!YRd%bK?^?$zaWldvcsjz=kP3q5fT51~wX8m!0~)NbT27`@(I!
zVofP|Fni?vrw0{Ok<g8DCb0WC&ktPlY^jH0G)D@&WWc+CJ*f^pQe^k<qbZLUUb<3@
zm2EH1@&Tf<$Q_f0q+j)jgEPu`DH>e3?wx(TzbTT2T_bCLOm=#|Tu-&|&fzN`JL8ZN
zu3hSbQbK4&^UR0K$Fk>`+$Xzsz^2)d!z9@8btR|;{=T5k>c4u1J_?Y<DxqieS4DQ~
zlhS=~OD2=4d#^=i&=yU*s@Ql4(n0qbbqaM6Y$RJbTt|wUIJ=o}=+FH>Y`UBa8Gb^!
zTh0+p?>gjI+Bt?3Hn}WNw2pc1kVsTTZ<S;Ymvur|MY^qTj!bP>{IJ)IW)kVv+y6RZ
z|9^eN<nCYFvh$^hKwjwLANM^0u*>P?`x<e)Uvv6T3WvT*Z4h=&f}l^+FXb!$nRD5l
zW8MF{$b275*!;*?n$TM_dzv;r4v0J?EXR91Yf5TNjZMt;NEEsH(IMPNF5c&4X5N;9
zFKN{5+<9`RE8qJlTPf^WFszh)P1|WH_=j`b8N^qVewn;Xr|yY%9&lyL5eyMd=E2={
z7exYPtjESFEkMIM9ZW`fMgz}q4f#l@zCF;3O6oi5<Ytk^N3<>vFbUpcpHn1E-6qD*
zt)uPf!*}f~VYmvp4~wGs6_vAoO#JEW*AKl-H@pWKU`@eAZ>PgLxjLdTMSOZi4fD(B
z`}8_!TXdz4-M1^<xoa%Cor@n+_so_&<ul2I1$P+ag@ln`vm6^NbFI{N?Xit3@y?BJ
zQ4z_pl4?Vl3?}$2-KEL1wqC9;I6_JKszXL^WOGO$$F1d6wlWVwhEyl;uoA(6)O^+C
z75H)Jv4IuLF{|#Avf$zaLovDe)e@@hE@HTi@i(<(&)@dmpdv3HU=eSYx~Z6UL>_D~
zo-;vb@9}WPW-P@fpn;(<*~~lNA@f0*xMYJwp{*Gl&bX3l^)RU&Px};pXKs~fUoMkl
z5Y-O@hMLz;jq)NMbU5o-4sRshih#VY^H3t3CbZ%rGD*NBJ*!UB(BPqALC$35)~IOf
z9HNSzK)3NFqYHbuQhuEtP?>#f4<Ek5c^!9Bn)<-z)@7}T`zokfW_z6)p_E^?9{ww8
zT4?oFCGGgIOq_C(Whr)%&@syX1FCtBpXD%?GAyO@`t69)%=-#~(rSSc$idd^>~nb?
zY_sEb<vD!t;jUDFh|!)#ztvHuB^c;nVGI-&2(X(?kntM=yysfix74V4RfD(_#3;5n
zf<e7#fXH%E>7h*SOsz|9Yj`?>wz!#Qv7M3DS_{_PeT$~z@uwUq7bNGU72nE#A)Dws
zIt`Rz#(~qaU()3%RPV=O43T$s*@>`n>ls>6v@WTtBd^Fq2z<HA!<UMAWau^(rr0fS
zw<(5qna%%+t*DElf`F;Oq{8)nHR_h64q|tf(i{(z6shWAb{fJ@&~6WuBf_GpsZzKg
z3Y1FCtM$(6UKeYmK8>W)#zgr8Bt}Y_)H1nk^cD38YE;#Q3xe4;_`(YQ3c-fdS}g4T
zZJm{)w3qC}Yq%3@AV_t`JJFPT-X|~|S#kfhyA-ymnk74pRkzni;+DBCC||(AZ#*Fx
z(5kqCO=EXP$^h1w+1)GEM$}%IV0`u-_1@4`sKV6xXtpTRc?`E~>*F*p6UFA%8!?Zo
zbz&*#&nVm}h{D(X{unvVmdo8MVGZO?@FnQC83kn1VWMjd>?{zr1p5rDX2es?+^`uj
z2mRe9;rg3z>|2rWG1V=q=@TNqhk#)ywaT-!EY;5-23?D>xPB~r%(TX7afP(l_c;mH
zJj2{WYn`byRWs1I#GLHOT<vg3^%X)yEM*|x+dP`v(~Y6J5-67KNArHNDj6sf@HVbb
zbmLUIYO}v5n6107&BSNErDVa~>6dy-5wfH+iY5<dEzMah-G1X-Rv77ylUk;FHTmx?
zo{T5WC^`PqY{Obt%-7fBOV~d9-|<<XX&m-MlX<Q#kiwswV!`I2qtW#G`$0sH-4y+)
zsf2;Kd1Jq2x&{^wy-hZ2hpCh<xkE^K&*cgo$JXZxbld7zuqaBuu@b{?I$H@&Zv1U%
zOx)a*Bi0DA9Orp|JA;mSHPKh0FJE71slOELx#opSR{ydk44b?YIOD^I`BE^1bsF_-
zW9B1wD3|@&Y};q`wr)2b*h-oEbzPY0a%UoNt08s(G>7^^d@;%Bv06#~CqIcbM8B~}
z=689?t203E?f$#+-E05pAMCdtQ@LD+2?G+7p@iQ?RZ7Jr^J1B!nK@p%L+$zF)V7_E
zdpxqwA&Ccm-FEjE+tt@lt#dzqo7tZh&W*7?xfY;**kNFc6FElWEhbzRt0F$UQ!dg2
zw|NR*Kh-`a=`~-QX)aL1@~Q1?42kQ42p`xOiRH`kC`}z6;~PpAoXj1~SHo?LsCJ~Q
zVST#zwrO$t27jc*Z^tVD*I<D*#Ij>lIT+R&hGTdwF5EwPD5l5-97ge~g{w_`<BJt+
z0@3=_!3Ob;KF5rtG}t%hxVm6@YO)as%}}0Rp?(v~RDWfa3^T(6D2V-=CTeMr4)<b|
zJ7E$~O-A)tC3Wf0QrSMA-oA=0m+gTnKhK|x6H~J(6Fm00#9lv_f(vcr<OFP;Sk#L%
z7`eu|Nc}xhf4-$W1cGpo7^pS}6-?0iZR>6qvCwrP|0x2dig5PmX9=L5O#CI#E)=hw
z+oPBOWI#mc8tT@O9=|Oup>sYG$`*qkM1_%zc&SU-I$lXaq9rQhH<cT;siEh&(;jU2
z@s&%gh@Cbjc!p%yp}a%|xZyTxEtE@{Zfv8UAd>}-^eL6)L~<Tg(A-GwUOBb!rc)n?
zmZpAPo|=yy57sZd+0taS@m|ftQ>1~fcuXXkY`_L$8^cd+-j7l`fWX1;RMd1$#0r2`
z%GIROju|Hv1rbx9ZBN<8eRn09Gl^?nJkTf~ZnB&P)K4c_P3!onl;3-WBaH4~@Rqm|
z)pduBq_^>pFc9vYI+zAPvd1x3(L78M;vE`=Iz^wfhu_1qG|-Cvnex1Qiifu6R*K)R
ziJS%U#yC5;3!YYR(lYR@MIvhQx6eISVf#xYu8ZSTu>2`$rzW-P*o2+ruG++ttzeJK
z<{iAuc|+p$zI*}CL?@@Q=F-Elp!xz|&@*1q9A#yUqGRHS3T+*K2JTi9yE@m*TB}*=
zRNVxSZDdaySaG0o&8aLEHDc{&A+~_O0ejUbWi5ehACyFxcte0i-sja@62Z#bD7;OQ
z>x_GEutn5Z__JI!gc@ui<p#jNojc_Kxr^PgDD_4kpDa&2^O%`gnRs*-y(({Na!gc&
zp%K5!Fc_}oi7!^`ckoyXFwDnxXmVV=R&W%lz89)}J%r8G$Q!hV$yKHLtb{%t%9~xS
zDf)iK_uc9XfrQTjX+z>o&2I-P<JUD?)X-l|E)!P96XMW1lSCBG8|P8OnY4wJfgw;(
zMT%^NsOWLF_Lx@6DQ%CP|H1L@UP%A*DTmk_ilH<<P@Bf{WT$zLPS)-Y8Gh@2mG-Mv
zb~D=GDIak4RVxb7(3zw5J$y{MA}c?Y>seJT+O^ZX%|iF-kq;>9w)!^LEimWumd2vr
zPf+g{HIJi2yz~b+#jiNQXg0N^_~h?j5j_5^2(+Rh#pcmB!?JaH8bRA4>V23>*k1y(
zkrR`zuW(7?SMF>FJY_FBlY&L6d2)d;`w*p2`r9D;uAd$aFWv(T6!dGryHFnXe5)>y
z*2n+B1OD$*?p;oL0S<)YFg2C7r>p|!4|ztqdYdU$JrG}OKweqsP`E|<2hqt=vbGX^
zq8zo7(c8W};~_m;Ael8yDPxC`YhJ#Va=EeTth;XsNsLOLZ3uvF_7HYqO=!+G@l<I~
z|0x!}$JagkPNmn?gu(bVFmXJ8EGT2^dmuZz`-G6Sx{U{cBdh@?H*X$;hpDtOdA0SM
zR`KshwO#9x4GsiQ==*iOZQ-A>4=_N4V96K(;ba@Rpxwgn^9T_TShim)7T}U#r~b{_
zz#s+vXO`Xf-83VaK#xfrHUah$D-e}#g(l*To%4@$HoNe#upq@T$nrp_)w9bzC%!!s
z>{;t=CN*77H%hyrinCvR;r>uCF2Db{9=$q<7mTy1%O89`(r2gqu_F1?g^H}SsAT0|
z0=k3x+kzz<_SNRpC`mHX{!sRm;<v?1I0#XxnEqV8-sa&<(d>n#e!eK=b4O1s$U<SY
z_w!w5qJxWdipz2J?nhhJHj2?5OLLJg7N0xYmPTvKkPO{nrxT_F8mIdcVLz_(4g)4}
z$yyKpbo_!fjap9#dz?+}BNH(`{u$U*)XZJ|?JHZm#|ut`%l(oD*6~NY2FC)z5c@#u
z9-EAWlvPfm-ugZ#R?sr+(k$k)jWBsxrIYgDPUI@KwcFpmZ2YzdE3}9KSAS^7@8yJw
zGo%;zn33z}C2Z$V%+;8eK65N;Iz_*hC}r5IEYRD4wSA*`?TiL`Ohx{HQzX8(a;0Id
zJvK|Q`wVZ~lKL5T;H2U8^U5Sxq|#4T-Q5L}KZiu*QP=0`qf}$XJ9_esU5GIMDE1z9
z@5<z`ci<$=Tn<d;SoHB7RIHh6LhF`%Qpt;SHXm3ezAn!3FWWp?HEnTUEFm!uVKGM^
zf|&)%gAhsM>6xH<J`^>PvDI|B&4=23>1d6^!Vy841nCM3IczmQ57%wF-<_`|x0ee3
z<)IS^T!Xj0pLh^(8j=+3s8L-4;i*eVY3$|q2-J-#92&tfn2hy{H3ID1HnmL@6Q-h-
zlmF4MU8>r4erDr2pJ*(}rN2BU8|!_wH|;L^i=*8qJZ0~9)Priw4*18a|5yhP*UQx{
zLSD!zm>(SZwgpQ2TKY;RH3R=Ud(fEJJ65M?6y;8@GYpCsu^3yfWJ<Wzl=o+S3r~FY
zzrT9_{g~__IGLzM9i@ZTNQ-bceb8(D?yK`S;6iV?j#7`~nAdSFx|7-Y!LlaXn3I!4
zRdKk|t%$4h3fHP_p4rrZs>aXM#4wEIwbaotgn{^!UA*5vS6O_z)FElxXaEROq%f}#
zllpGKRV7vID~a7o>GuJZsNXsAPot4aN`S(`d`w;N)Mx@<s*UboDiS7uu;}?hLgrKQ
z#@~W@s)+{!DbX#ctMi%#t#D)n5{YsI<)+&GhKnW0RAWq%MoO_}Rr)ChL7fjqG-Z*c
zi=R<b3{l>0axVi>WWKX9Je~28`&D34`OhdK_M${a8(k(qEZs@>f$^IhIoe<}Uy=5v
zV0@<4t1?D<sJL{3it}`I*jz8vMT~R&eKjs(<xg*RQtF_`ZzladxF`#w(Rd=uxZNEX
zOTCop(ptv>Y=ojw?bjxp0;_pShaSB=GE^6(Dv-W;ckje~m)j`6f2$03S+kP7KXlU7
zsE9$^$^jlh6<%;xvbXuy82$=@0NP5ld429g<9nlmv!Z>Il&wl#M!uTU#Q3UTVV^X)
zz?>49%E^hUPmG$!y{WdB&thlknP%-jZ+dx$eaEBIGb(s6I8S%6v*zWiy2Ul}JNo*8
z_4Xe|vdJ%pStoZ1+fBq8tcQuje+kZCovvSWS(4w5E|&hGMSnRR!(vLM{$Sp$`vi8}
zAt-@@S_hMFE`DGA^Os;QIG_BY@9ysF1hd1KPR5N%r@sXMg6?wTnj<W8XoW?pqrAR(
zSU9#3AvG~3rw@SWr-hh4D3W<?u`@^|WM<PfFE0nqLuGYSGtr!v@n<hwFuVT6a5|%+
zw`8lgNZV+&41)Ht-aMdSNZ&%O1Fi8cTv}A7G%frE_ls8>C`$WXr2y`GDbY?F;|<7>
z$(RQ5@FI7Hot;-txzIXsa(22XiZZ&DkZzUlBYZ056!m4pUw|L@ajw;|a(;o~v}-gI
zlXVA2J0)Q1aY>beVOv?M6)^<mjlQpo_<Phla9H#tH6f>hPkcXLE%|WQw8T%a-j5Hh
zSMOVaa%KwOidVm5x?59Jz0f!dfQkl>1Uyx)6P?R_<@VWVQnHb5$9RxP310d(&FYpR
zM;wr9jbYH|&DyS}xlOaNC6nU>-fCNlaJ<}cr6&3Lk!GBeHXJe)hOOy4m5?v=FjnAn
zf7vNOtnM4<2k%LSH%33MVaG=ZAMR_-2lW}uJCj#pIdNv&cO0w0etQY@?B2QWn@D%+
zo_eaP4EBGwOSe<n?Lw^%zEsH+SaU{mz%%nQ(?;3W>%`Z&Co1-RRVp;KRzuls<a?U=
z`95H|JrUnb16?5E`M_HVjF#>|=twZCKpQ)!sQW_2e2gDOVJ*^LxR(Wx?{iSiZI-a~
zC5?1*3g4;~W%_MfHuo~S6Gnt!|Ix|eMef5$^9)Z(ywum<*BNxCgiJw1vw$t0rP4gG
zi*k|n5mhVFD{`Scl>yFhPd&-s{+aZ(sPIHdzmA$I+J~i!z4~BDywoxle3lcdeOA5{
zy7AxdOm{Qxfl0bgMos)o7OZEStko}$1paj>^~n_O_pa(4v_W<)ST8eDqdW>o;Q*4x
zeCwf2+7|1H)do)^E6DVWOP`aGnY=Q^4MV(Op@7|GY!Pj-pZ5g~YmQr;!!@V-WrJHk
z!dK>0tNLpmjt^O}v;F1XqDV2mui1d|Ll!S}Iol||6vJfYZu)>$MoM8{GJWh>*sA&}
z0_h}_Auo4Jk^#HvUW14f1V}+dI@@^lP1#InUZY%RCv#l23(pCk>pQq#G4bTCO_q**
zY`i6kq`rigYh>Z)*7TN@6#$(py~g2M?<*V!B>XOJWP2l*aS837K`GTOu2uqBEsY5i
zPTfBF>246o*WaEaOUbzKg-))hS6b|^9}sqnt@X`Z(Z_z00ZGQ%8Gr1$ylG|Mo9c^o
z^h++iZycpEYZeq6Snd_e{0&}_u%U!*&epS{r#M`&iGL%I*A9g6cwX7A&MC|XeTbrJ
zUzX4~+V-yT()d+lEg5N8<Y)w7+q%VR`R{nlxp30Q$C7sz#qLf6lsV}I)MAoc&~Y3R
z#{yD$)Fwsqoquv$`f!d%+b=3T1ZHnDq?9ave;Rt?GL(uZ?5N^C+@xYuoBp-U6R#S;
zoY+&t=<$LsF{m5)yhlqbS+<jF%*=UWP3E|SZ_vzfx5e*he%yoqk_S9>D|#rriZJ#q
za{`oO$N5&~|2fQgBBaH*A@1WugK5AH%lgmzw+Ms_l<U_r=HibWBvuK(?|NhmC{TfL
zk`@|Bj%;xX6i>`&6Ho=uTB_*K77${Vp*$9UmxoETn6>4-*=M@%q#yX%NS^*^xjo%|
zqJQx*aUnl(xa%9I1d?5knfOo^7Kqx<-~Xd^`k$35k003m@htR9j|cK?-q8y9`1j-4
z3|{y)CPu7f{Pww?G~^AraWy`Df^^R8PZ@yTW{9c&_>l);z@zSy_3p-kpA9lU8H#s~
zvMW=x9+N}$?F4Ch3qL~&rt8dG6_bZ$aQ8?k7iUWJ%!~5hg9gGs6f~k7xeJUv-XmUx
zccqV!qJA<HOlQW_1}FMuBv*7ry(*<H4ZE-nI+gu8aEEy2>)&>l_+$YBXuRz~h3-Te
zxGwHUFnyr{g4V}fvuyFD+Zx&g^u?*=bs>rFg|O^fQKKr>S4mxAM%eb_4xhE|PQ>&b
z7T2fTSH)4|)C6xuvp)#KzHV|H$xEd)b<K&}p;ns_QT!&IMR<F5U=}f7LNH~e?Lfcy
z-dyrs0*rRhu^2FVD9Je$)>f>PJ-#gY8>a&U7U>w)J!iLsmu<(+yo$;U=Z8ABeLo9-
zhk;6^9@PYwo)UH#zVo5sSQKEeprB`Mi6k0spso+9QZ@PYt^miZrHtTyHS3}~T90@Y
z33%KesckYdu~KyOsEPkWmnnn0!_-m8_O3~vBXY5uMf%kB^v1&5v4+<_VIS}RHs!<R
ze^?FQ60BWa4cv$2;6IL+zyIgw_H`1X_x=SNTjZPwpbu`_;^JN;J#gbRN#yKWZeBK^
zh8ZVBHU}(ctoH>9g01`B=8V5M(M=I#C@3Fss%f0gzUvVYiho)2^SHx8Z(4*t{3V!p
zb5N}P=Kuc5J<f%z9jooB%de(R{$`7Yk$&TcwX&Xx{OMEtJ$}!gVP=Xxk`@Fcq*;f5
z2oBdU1QBN^-3|`4P%Z^3LaHC}?vL45Yzo8s0-o|v*}H8XB6Xe5a`VnC>VN5OesrfF
zdbDk+Uc$fb@ns>kq_$dSIfm!N87@`LkXS?2P_70}%n|2SzmrHM7xwpe5SkXcL6z!M
zb0OomA`fW--{rr{(RnpiTl_0u*Pg02`d$7g_wO9G+Jl9+A#K7IN_T}+SIj7^Dk$rT
z>b>8_Q#(mhqZz)6c}s&1E~Vzngd94UFb^xnPK-`tp?Ibb)#fQfx_qqowftI*9?;^!
zbhYZFCWtYaot;ar-7_`DYWUu6viuY(j_E5<&^trxfF(ZQNSK|Of%Isg&eMWZH`S%m
z>5YWalj&%Nt%mKBV9i{C0zD=9{4pA|&Wka$uu)=CWuC|Zvq#5nkSwS0HkDw%UB?X4
z&=dP|$$dz_alZlUhmR;*n=yFpqP7Ntj%1HGZTX3ZuBkf1YsctxEZNjV>_bMg-_k3z
zU6T#BHV}2U5&sj(GF^7CG>qMm<(dsUsiUx?iOy0Ec7nJW4h_W$xE^i-p>D7Enk0b9
zaEVp}cEurX_$#uKwQXA)71w3uVC8{M@TWe8n8cvU548>_$yBLl!PEk&y@5_k0F+2T
z7?8+yFH=<@UAb=IS4y^OU(98e%)Lk{d7`zpNduxg1$)HFWadK+{Ikku$&WA_gKWy#
z&JUvWp>;3xL^ws|YeaZP;?yQ}ffa`IfxT-QN2PI*6kZMpxyNmdWR)x38ojx^>>)GY
z^bToeHgTN>0sF{QH<ME?Mt7ZMAUtBg)XgqvM(T($VpQV?S^vMn)M@EKc)rTTFF0o>
zp*nCDP#9hxhAOeJ4)mB$GH{!i2{U1Ig}bzEr_5)~R=X>6D7e(Q>o|b&-dS)c6mljB
z#WRr?te9D84mqacQX@8d3C*}KY`tg4uM_iz{D6Xd3N7%f0qm(7($G6-)#fijeTjkn
zN$6d!rP8}CYig=)a*cxb-&fq_hWYMo)Gn9dt+}iYZ7#axoO+rf!{|!Gsa;A<&k>s2
z9dx35M0M{Y?{`_XXV1I!Ldd*wrTX_20_ZX0aege9D7E_>hM?ean#YUjfq}I~aHYzF
zQDO|S_pdTX6_J-}p^nouc>>}z^Vv%AK*U|i_*w)>CZ=qfE&;50Is&{2H4*23YccX?
z@_zX~l~cpSGr^s2i?=;~7W{bUd4s$To`a6NxcZX5`f`zl+HGCKz%U(iQTG3fy0?yM
zqif%VsnG{0@Zd#T+@UyyLIqE-1Omm2Lm<IjD!4nrDNcf0a47CjoCIlchvMFTdEWEA
zzwdm%KhAmoI%}O-nar%2duH!s?fafRdtdi;A>S9PW}IlQ^oGyJAj#$|<K=802qf~b
z^I&qu%bF-fwThGPcrxYlCb0(P68K3fp~Hl};tPI##`>)C(L-DYqYXRpUnCgxho?Tc
zYDlE>IX}n-4$gKnEUcradz*WQ(GaV7s=+(|n|$Sb10K|987jU@ykPcC@Ghq|QZ032
zAo@1*icm>yoc%B5XBLBS)#HJ<NRi?X2ih;vA_CsqLYj9FXYZzRv=0EaFD3k-2)W#8
zUb_U>nUZPBegqJ`6-Pbf0wUuosx8dRnQp0nGh+_b2s{B{2C@vD`_Esj5hg^X*uKfr
zBb^Xbo?BW(H}CXxXNcFsr9zr!*grn5Amj)0Lg=Rb$Fnt@o0%wU@&<v6tZ;6g2Q%)D
z)3>yl#i#V?3`J`dTrVwprFG6T)f-<(JQc7b?2qFR{yMP>a`5r*DqNVE4sd&arc?+h
zyqvOoiT(<)6@&Vzr;{OaUur(c&!W!qt`^cAC^vBTT_^q9QD?IR+psU!6F!}QP;pn#
zB6f`bR`ueLoL|baO!L+2*k+ySQr+E@B=RD~T4WiAi|7t9AWtg_7$p$z7As@$?Ndi{
z(Cyo?<PZ686PTt7FbxU;uHorb$(p+~#NPET#<P@9g-^E@5@3_CW@7~%A*LICBW-JK
zgNpon*ED!|B!M`zfaPyFks~L}g}a0Brrj#6?*oW)KG_f4Zro-|s=1Z)avM4JnXy@Y
zyMDE0A#6Z3sN6u92@v5%Qt<5IJse){DgzdVQ`#a9#_6Io4B?V1*zWC7<NE)zRbMbE
z!#nO{3pDXJq~h0Pj(~Q7Ij8}Yon26?yrj2ABX%GSm>+$s%i5F)-UR{@74bop5f2@$
z1<zd#Uww;Y?=qyR{Mxh@nruw<erEKU{;-N?yv86r0IJ2UU5#LM;$#8!yd<YIIFzb$
z!t;A`Z^nLVZI(B6vCHEHLTrxNY}GAfRrWQ0s>fvJ1>gsgmYuLfUqX#XL42=<S3o$M
z%Du{VmTOufhkr|8$WUU*^qY`Y$kG#dcd1lUcC#(!#b@Wiqr$bx;<<o`lx2TxZ6Mkf
z)MGZ+Bt(~|RHEQy!!8+7$p39Q`9#Hyj~~YMB-qNws$Bjg$-)#)x1q%U)GKs`nt;Rn
zy^3u?(@3U5*IdGOEmpn>Rz{8!JL@kPUn~S6|Gw+B257;++iO#zJPdB)j}r$r#`x0K
z2p?Sm27W&tcT`LbtKq*6I??#LpxX+i@^&j7(HA2zkAvT2tyn!NcG^wipuQ9OJ2P!^
zxLyVvhfkpRV0iNZ%S-vOk<SmF)cAZBhl^uHy#tb<4|E?lo#J4M$X65Zlf)-m%K(}8
z&+49KbkfIwjhY=NWhSC0K|7Dwlhfb5=?y0K6q%=eA^kU3Ln?G#47)gMughvE-i$bS
zVJ@?)_!aU&4eBQ>ZwIaOZ+DnVQjH%Hcnj%LE0dhS!qz3%0#Cm^=~^ws`&nr9YjnAz
zrDNGGDe|d_Yo+SUUded?VY@>73VC&eX5>`w%)W4)x<PoHas@amNIy@NV-sA|l`^UE
z^@$+pmZmIJ)SyQ<fL81!Zkn7`e>&1-nBQzPf-3_&%Pk6s=I%dQLU?{h`RPF3MFtOW
zo285waJ;d|rgV3*u%oTUKrJr*t{w^&dsRI*tuN+2+JTZ0fTY?OKu`}L;<icNBGjt$
z?)G%?WcsC-Ya1LKY7jT?ygFTuX3gaKPiK6&_ND#!a)LeyWXDkmBA17U(w);rNiI74
zyG3L?L3(IiokCqex_G+l64ef11XafJq<B*6p4eZ_F;#R1pu$lrW!irvLH}=AoTmS!
zIr94yS7C|@YZ5xgpX?2rnJ2^dSA5+>c8;&EKRc_IhWXyKNPM!e93}&~|4Y!c$?nu6
zgy^*8Gfsj;`Sj{~sGp_acRHfwzKrNkLm7Wf?bZ{+JL%7*g^Mz>jZad^;<ie(=pr<B
zZL72uj;})a7#1?;-S27+hHT`jq;Uf{gEuKmw<E?TcZ#`TWPBHH4dp<dvsnZ8)MW<u
z5h~%;vG!(<Z^0>9)Ejs!VjS7=i#bpI-XL{z*0spz13s3=55+XUtB0O7LA()q8A~ha
z@T{g0=5XL7Rh|=-7|DP*hFuc4KV)4Iql4djC5n|xwBPwDmK4=s6y*s;@1oa*bKJ3R
z^c(X9MxK6-<@8Dh^#Z&8c{)+rxkH88gmV1N!n~j&MW9Pg)2T;nRg#j^!C3Ht-H6}4
z{(^m>{{61h8&1w-f!7$%)D+$oVs9kjP$#A+G;WJ>7n*jaQ^Il5xG$!?@1xM0r=(Yh
zU#0^sd44L|u~nbJmn*&Nhq%JdS2zfl)%)Gndv2JrcIse1x_%g?$Iid3&J#BBk-&6F
zx<{r^*OYj#RsGsoPYD+(qJ43({<cxDk(4LEq2&ExU$C20vj9zK9IqO*VAgMcIqnIw
zb-;%M&b^d@a<F@-rSZ>(uT~YjAr3!i%i{p01E@|oQrt&O-bj26(sV-%s@%Vzt1b6@
z5qAKC{<u)DXZ2Gn@&O=zzz<sAHmMK7a_78i=Q(Fc%=jrnRGP!`^*!6o>)4Ad^`2%b
zhj-|Hb$z%RlN+WhR#<PZ;-THHoD$l568}|op|k$X9^1Ho#bNwm=D~;j{Es6<FmLtz
zAlVtQ0^*CRXli{OEB)O**D0gE3KBVJs@I<3^&viHq9!h`MYCr?Yhb-E>Zzb!FA?Ll
zTYKmnX=oiGzi_>f1Th_>Go`0{+2t^o$D%NVCy#7%sb0ULGqPx$&bEr}nwerb9(Fc5
zwEUA;s)b7?bza!H;*%)f;Lk_dFw0djT^(+HS0AZw9oHw+cg>+r`RdLJV-Q8k4qO~2
z+#fWR(N>$Z1Q)AV6>G9gQD1KeV^K!#1&tQ1*2f#f$&RC$QD-v2u3S%VII0YKo65(o
zv)d(h;nKpZ=~_rr1<*B=_dqHE!LrK3dzx%#m#nVMh{^?xIBi3{dcJ=c>$o=z$2U!W
zBQ!YCy<pZk{1VkYJUlr0!eK$KmhTE|dlvbk8Vj}XykIm-+Z`XlilV1ZRJ7#L4C?2{
z)CvqZ$o=sVzBT&S(u~#fJ@wJDZPUf%#UrZ)Np=CF`+GIJwasbRa)pZ#rO;H@@O?Vf
zK_h@*ROO!AYZ9%IgyE){^BX8j7YL+tOW}XBmU-v!c(CQ^*e1JWLRj`?Ui89{#>rOe
zC5^4eRa5Y!mCxHChlhSEI&SF3BDv|^!XLhMd+K85gc@Td1r-UUnlKggy-o*9_wh%x
zc7O2R7Mf9P>O3{H71TrA3ob~~l7=te6zT<#FZ~sVnERZ=&a)EPyNc57GU{<X58KZ1
ztN$ii5hCMCg~@f%zPb!Biu|?0G-hZrA5$wHR?kBH!nCZ_RHdtwZ-}jh)IEQlB`D;)
zizpqoATc=ukt%YTDp)#K=#pzU1-MhGxeie6Op_QY%8cgqMwDgZLI|0w8v1DaNtyj=
zvoBd`jhO-|<u7`=7b{jG^GO^<^O6Z^kB<h`{AZC+n-B=vs3!stV>+MYNvM}$5R3qX
z>3>q1sE1g@t8;Su(_dVak14eMered~)ZV|4*4wt1p<WBSxginQh2B<`H`!4|7$4P?
zBn~CdH6VpUN5#~`jPeFjRgvF(<IBTUUkcRb0TNmOfOtb_guk@z$FYxDUNGkUFS#gK
zn}JpcN%ccosG?E9*EIzQHrYf2&;MhyYDFH#Z&~q!j<=JJ3;==H^FZ%y?s~{M5?^c6
zI~N6We)6~tQnAT44A}$hOGS9;{_M3!MU6j;+!fLZi&F<eIkRyjM0jwAF#31L5`lA?
zrFPwX7|Y+EgL(4{t4Ma$xv^BCb#uTE`L#N&vH{x2891^!RYaQy(yWHmUeEq!#ezW6
z(ACX@cT!P2g3q$)VLWa#2Pc62K(Bx*7lUQq%}JrBrYuDwBu*wzTpoMQP1;fRaZPj2
zl*v)OsH4=2N2((=I-CtDMV{v=zkX^fz)o5~N<NXLHe+-FEUT~+tT>R{@zL9*Qp^4Q
zw*bXn#qe**r4+&VA@EWyN_$>-se!(ByP?iU$MHbmIWy@LN3}NM09aTgxes6j6i!22
z)BKpPVbxQBGYFVU4D`}hEA#g&pI8&8B)(${hY#mNh6yX{eV&#+JnnTqJ3#O87gdpR
zx#Lwdw1O>fGojPm;-z-_Zh!Cu)?0U`^U4Rqc+{ZTd6aB9ZhC?paAE4Ov~mv(hI0Wg
ze)RDri8bgIk#+1jz{PkKR*;-AVi-biN=K*n?pJ58C|c3PSy6?Bh2<x2{8q~z;^=Y5
zd4>+Rr_!>{-etEPw&4C;?amA&ey5!7bk5bA&+|Y(CBlXW$Ks+g^Sru~CrnkB(e0F7
zkeeS01d(mgD_Il0a%YT6;+t8za_aHr=)QI6*t_3WMTY~mq<pqAu4kv3(@y2r5=hIe
zcv}MD;m)66r$Y_8+lzL@9XQXq`O!AWaog|Xq!C7a?v*$MTvS)E-YCByud0=MRZ~Jc
zHav0`<}qxy^Ev!t+N{RTbfHPcO6za)nf><=n?s!<9d4sXrjNZ<vQ?}erj(6ulbMMl
z82+J`!KDN{lzSBq-rlK)Un+88TtYmr;v?HE1c{FqtyV@pM;xK%RyhT!3vvJvszIPe
z{-^HK1?~&fE{{lZOd<pr741&FDA<8{xo;_Rqu8gb)509GpzTa@Wx>}^&(br<`_HxX
zhmB1I=z{8llZ`4YYjiS|zfP6wmsIbPeRxyURRluX7$U1JltEixv}}Mj`ZFskY>I6@
zG_C$d7NmBAOis=hSs4^xt%&$8#!cod9fy!w<`704pky$Sr<D$n1KFnXCho$|)G3_e
zK8{5j+g>hGq`6jQ$9BeVFI)uFeESdjZp1>utaV+I4HtFcot;nKmAx0XAFAHzHe@ap
z0aSgM<RPlDPnk>b_?~L!15QTs5YK$5Y!sOYFicr1r2GP?6RdChzSQx1>{{X+(tT~%
zC3>qx^mL*!n5=YAF{67@#Pxi1h9|{a$IXQ0lGuE?i3(uL+@*_&)_F<gMt-N$k!H|K
z)%sl76mvO2QaG%QBCnz+RES;W<F=uQV`foM;>nBq&b&A?B9J~Z+Skom_xwNjN&mBn
z4OKnz9JDNIk1*&qqGylGFHP!B1dIwWXQe$OvC3k+`$>G~J$AkR(btRzBRt<IaRtNY
zImcRMQwbYYUx<5!bx1G<h3mi&4&$XeCqj9{MmgHu(#O+XHLde80<Rj%2WnQ0$a%k0
z2=M;8n+PFrPB-|=i8B$Fe9?5W+y3fK<mtUR&Z;hJb!%s9%l-Q!zECSq5z=V#@}UYU
z0bf`vk4I*62o06gMu`Fvb5>$z1pC{{i1O0)5<*|t8oE_FT{w`2Ol)_7ZG+4u4GLAV
z*#-xG)}PfsH{*qhZPsAb7m3p@&1;%^zxi0rbd<t|k{Tn#7Fy=1qgp@A90BrDurd%V
zm{ZS*bsXN5iF|t@POR8Ur>0P7IqDr6R^lFi!?#U7F#BT_{d4Rv_<64Suqr#ak=m*9
z2EvraX~^B1NLD-?^KJ?3c{K)loA@ev!{mW;W>qtb<aCPa+d|VUD$sFF5|+{UysR5F
zln>rqOin6+k5hpmOvEaenrx85UNbedbBg<ELm|RnZ*8Kaj)jX6!2PBo#`^ks!D89C
z6#I<5I;=<VsIh1J2g?c}@x&&*=En>$=pE{;Pq~mqm`N4OAwDOT;=#|sCr-Xlz*er^
zDF>s8Rxo&2%g|i?v$<MfXmjSxbY4ef?PPGXp8Ho*V+b63HyzvjU=hnm?z?kyp}<tH
zS7>IjqfS1HDSrzUqb>ub+fBk(E?`e^i6n}uk2AGK3hIjVIE|bhm&*7}qPacQ=K4Y<
zuTLWcc^kJP+-5$k$fz_LpCDtatSavB1(~F?VJ!<kU?daNO7v+S1{JJ*H=Bws;Z`n;
zFGz=b9oJo)wa}y0N6^~bXeTp94Pmd3pUphurovBBBj3a5W{;QxvyW<ZG8zf*RTruz
z2Z<gbRlv##sM=Tj#~SSI#1Zx_y0YvF^`y@zud>AlWxo$+Tc?$TCzckQKa!xO?HFto
z2=X(eUc}#yvQ;dBtV#<o0N5w>b0@ZDK@9GRtMO$65H47O8gDOUNIJL#1dnht{)3lv
zGT++)p>hVg)gFAu{Lg;!zu$i_T@hY8NSx%U%e?~=*+SQK=zJ;CRIg1<+)-_eOI)?G
zP#4z_4#3SCCTD`JyZzgqC42@|fVfY%)|0bJCM`C|qIG0GsHwmGOhtD(QvTk-S!^I(
zXQ<(~ZGXjQP6;~3xL=leJ!U8)vE@#edDkU`e3OGH%h2yh`$_8vQFO}_&k{YYIkWEa
z$$4I@(GWOS9;gLg@mCZm|CLqka3-@lZ+cW3y$pd=Of}Q|@2)ib@3NnvY465wgE<l1
z&M7g(lh3<RW-8$sxqzunqx-guyyQf5o28=i=4a(SBV_8eTP1+GW*W0CN@%w&&47Gd
zR`kZufbc{0iHD-DVz|^-6bsKr^6(AZ;@n41FR44B8+HDCQGDu5lUtjj5X?Q53L;1Y
z7jV}PU;-Unqj3@pwm0t<)DJ}%4bJUv-wGAgSVg7hL8)VYwXwy*yi1G$N&0y@N$KyW
zbq3LFB!g;TQfs0`gW|&#BM&=!xiP1<5Zja$blpJH@NFa{5s|pZ{uS6AQX)`ARWTEG
zU6+aFpR`gj|M|FT{)u<hms{CGr?SJJF&{r3{j9~Yg|r0R|JWxPaL3y92d@Vm^!tkE
z$1N|1!42tq^XW*MjrJX}_S?Vz;H`Gp;?(E8@NtSz(N2%2K1RTZ?v5etmpZCl>ZX;Q
z_)p9g{@?|jHf0PB6G>H0i;Jm$o#azZnNQr(QtQLUTz{bkB<7Pz{8HmqIT3kg;g?mL
zsW~Yg`~p8=CviYZOz!FU%2c7PGnu(kLV~YU<X-%{&(!AEn^k%|v#9Pdy-~H$5WS-$
zLK52Uv+yS^o;5FQYxqBtyt1l`Gt2D=qS4#8-d1Q)0CKFUZMJ++)D)V1QBippXWNWF
zW6;VFv9l>rj7}f`M3H3Jj=tDSn=CD^XN+*@MD?3mWz)ifhZh7NPqw4+zqDq3Rr(xt
zWWl`$?$<xQq<P{lR%*aF3U6UcjSFF-tRtt?DX-;DyImJB79m?YB$pm%uGlr@r(YSi
zIG0UQcjbQMmaUUDlT;)`X1{KsD>RnoqRT0l5cJ}_tZp2cnr?U>gZT0uvH}@?BZ#_N
z>r*_x^QU9QxLy%l>c6SI)+5&TcrsGsy_y8E+@Dgm%GU7sS+t^<-A*ARVQcWi#{0F$
z<@1iJ8)d0Pg1%vMslev)A7}o2d+RHc&GI!9B#g-giUFlpp3?-ECdkG7PCBB_OP}iW
zQi*FI*fP_r!Di5crtss`<iUCGv_WC*o=R|A|9Quky#!W!c`M#0z2QSbt4g+L*1l*g
zt!c3STjoMN+u8zj8jJuyk6ZtaXUh)_4fS0~Nr{nQ5}lf|nG_KrGgC@QNeR6iU+oJ&
z^L;;_+wGF85J7g}vLf%3anTAHg+L%ObKMr=VzJBTj&q^CLQ#rUQBb`yJ9-N{ecbKB
zEaa<cJV+ZH)7#m1%hJ8b8p_As6CIndUttIqbe(QIKDDzSw|B0=d241&B#DY%vtJYU
zxO)$6-gDhtYk&K1>)rnUX7J4<{)%fs*SfULkw)Dgc*^Ny)U&^E{}wU$gZ{zWGp94W
zc1rED%l|44cmBugdQS?2gch`0Kb(umAGki7Lm$3(_D=;&ehjWU8D&+{thJdK&UXHL
zAJ9&O8?*ecGXAkfjNacK8}JWsGx{WuaSPI#*39fWI6uK_XFt8-@BbghbOZmdg__k7
ztK6O}!?)?l=*2)hPPx6;Z$ArUXpImo>V6<9`smDaORymO6IW{3kv6EoFaQ4TZOKNq
zrj2a*bKsG}<t@S1jmG?Z!+oCeU5`82bqjfg9K7i^_=uPNbv%9<EM&-q%S!M4@F37g
z7g1{cPPu^z)s-Gb=eU~li$%6;#q4gjf>oM=eMD!e)ri8%y(&-;fBYciL5=K8yF&z8
zJ!q|xZdSbv7<aULeBUzrkw&)8l1&Ujl0|9jy3^~whO6Yrha*oX<!WJ>o+q<lfk$1n
zjekL^1@z9Qhp$b$<M}u{<<VAZ$#m~kRGxvT%784h6-Gp}6(8JY4I___=ChWbxtItN
zaLkwZ0jeXyzNZ)@-n=HSXnJ*}vXaa(wS8XT9N#&6Lp8|5sPvWHc7i*EKq?LLwa}q-
z0kiI0^5xuJpfN?2D!^)>W=o9L4O+;Fam*-s)ue4XR5WFKI0GnQDj_&_7Hu*^Zpx6)
z8xTYnNq`;ZTM(+hGx$F$XaMQf^sfRt4SPN18)U-f_&xgCUK}Y0!L`a$9C@^N{M6lA
zztrhN41d<V55ae-)JF=jryS2x6>i`tF=ASVqL3g#i}u;(HLd#Je?Q80Uk>K8sCi9o
zDU|~b_#)vZj=>eAKXH1$(8&MR<}WW5nYrPGzt?7M_J;IwYJoQC20?9snfPXv&62df
z!?MpnPJ+ZZ=6;?jK#u^sjw?5!wL+!BAl}5Zo<mC@oY5Fbh*U&sW$ULTFI<)}^QxaX
zNMU1Ff|Pg+URBuHH*UMWI#>_i?ZLwFFmwCX#&byBFzvJ+q`5}AmtquC<b-h+s<jK&
zSB!N<u;$o~r=BJM&>zc{^WzA8Y0X<^-}JLtda_j_9b#4_>(P2uao|qS!{3vdRn%8u
zl4umdWTM{2%rzNy^jed)x`OS@B%bPX{?K#0c$I9H`#iapzPG<09{hJFy>BNGlY581
z(~MWIesHRN8Q0_&vURUCLu)wm|L??lzCUs9g4mLv!t4rWJetgozr$Z?miZt!;EPy4
z8UAn1{olR+%J+mm{Mg3>=VizBz1;`TJSsNs`2)AwAK?z)cx>IjO_K4aVRd@*&+RIt
zDH<w@=3|N<T?x8X{2H2qKXb$doRA((_`j}m``MNamY`ZcKfLD;$Rav!SkU|@9_@~G
z<2?E7_M6~<pKSg=6?!B1W3Bz22Wjo6v>%J^OHF6=A&dJP_sRyAOx{UAFvquVj4W0+
z5#3Rwm3MeJuO9xyE%nUw{2vGZ^!<kP`~B^nSG^%~{-0>nxi{WWHUyq!r*Gbe8LQNJ
zmJhi(?*O$}@S26^tEARz+wPI~hdLEdmpyZDThb&%*NatUz_OLf4Qw7=NH8uKr=51u
z!uG;Cd?&*x59YozE9KB;v#j{PTk+4ugJP;?R8=MC3D7qW?IPgp<NCen9<*GMO*d?)
zq#~C_5R@%+<mLE)lUfe(tD~F((Fry&)ENlM(Tzo`9~S<>lU=oa!yN`|CR)JIWIu9%
z|G^stfb5WndBeo}JNr}j0v+yeYLv<e-L)G+-46fY(bg`?M2w2ul~)`4PE!nZJ02Qb
z(Z>;GUKmW#?Okdj)}`h#pip<74AY?qdZA}e98pt+^i|_HVd_@L9L0uy&x<LLx47DL
zb&}+`R1#C5uZ@UN(hXQ4E_ngJmczkwA14gxqLh+UJd+O$umz{ZVu}Z(d*=uA#U|wz
zm%mwh{`j)l9FvQ(Zx5B7TJbu6)+=C>IWOJdTd_jQaO`z!tp0e1qPcT{#;<vDgAmCr
zj3aWoR?iB!^6lg=4@dlFYl64^kbQoV_W=)&aL6^nxUFmV#0y_nPX{v}4nI14TsB?4
z3;xXh_invRt3vWBcl?5W9W-4%e$B0{H;(xUZl(xj2^T)Gb>kMFLT9|?nhGNGFU_Q`
z0R?WIUaP756PfXGDkj9Hr3KEF*%@!a!|qee#gSE*Cxtil&(=Zs)*vgsHZPLLSvrH!
zt=VLexVdEGQSHGkq(4_t#C3t`?2oggE6BUMS07DIb#b82*1?Ng4b4AzJ1(+kIKa%%
z->)x4U!R8IEXST5-C6EV|CTB?Zr}LzJA5Uw{hXn~;NzWA@%2Y!Un}tAaX0C2V$wU-
z5zqn9eVP05n~HI}pue%))=H?i$VBsx`CEi<JN<qKE^QfYHK{En<hr;R_1NB0Of_YS
zRNoF}zSh{UyLslHkdnCXBV#;Sy}#VGFD8bmDDg9L-!;$^?C-D4HK_^hPeZrmVRwzs
z+fZc{mUZVC4bj)1=J2#*(1NeKKM;K*B{gg)yRo8!O0c`fI84r5v*fxt7ZD~9gN@{(
zB^ZT^+8KjnSLobqY?9&p#FF{>1BAlAcl?A4bfT-LntaaymX(XN6Q@_oCvg)LRAsJ^
zAVCr~`QTZZD3|G`;tD8FZlhoi-75lj_XM(>H)f4=F;Qf(sinoJs2T4%u_Q7G`GvN^
zPnTu25Kstpjs}9xmtY)&Aav!#CXraE5EpD6=MI?KS;>1~^%*TUlF^GW6@AM5;>bUE
zX|Mh|u381Ec0>pY2~zzsAL<Lf6#A(cxNh+tlq0X$9YRfy-H^q}8ddd|IkV*Epk?Bo
z&UU#N_ti1<RrPPklmqKVoOy8S;M)I)8>xO@i)MZE@AoXGR!DOhjxT_iquynH%DX^w
zUQIY1Ql@2lMi|y|q-gWC3&EaBK7zpq&2#NH*m(}*@3en;t*#!SR}55DR|Z2}CgMZa
z)6;NmEj2kwQEuP3hU{mjFBez8$&7TZi<{~1x*8WrmJ#4vA7;hW&xCA1cgL095aJ_&
zCS-BnONQO%(?S&gEPqjD%<#)^4`y=~XEGIQ1Qv=G&QEuP@4Z5va=D@GQNt|JXl$(W
zw!MnsR(Dr)?R@XaUmVG+HpV;2@vkRi6iNsCJ5Z_<Q0G%jk{NKyykG5ITy&NGW|y*v
z6SKH6Pq!V@22|d?S$`cijzWGLK-$C^2@Ir=Mi1?FSlHljFc5k7T2bfLCRA~RdqEP2
z#&(bOgv8~cfeo5?{lflHv+Y8i0owTCh5@l{e6hQ%<oB`^32BIdzG~8#G%5Lh2<P_#
zBXlhwVq!;pwMsxr#&{exg-YPAE_j2QDNgzzy47||Bva_hO<La5r<!I`JJph}c$EHw
z21dABdYCGHayl?41kURdaZqTQ5U}Ne>E{r7h9-ZNtqNFrBw&|OJc(+|9aYz_EbIzj
zv04QGs^)lZX5>!HUNQa5GQ=U>?lEl6c~OPZa*ADh{Y@dVD@AoWwnIK?jGtfRB@<<7
zMvGTt@u9LKdqIwT43ntR$0w4SrefnPG9F@4x&mu2E+Yf`<ipS9g1cvYme$Pa!s-k+
zdBln{ARyv{b5DygUN3?n;sV5zygEWfI_xA$WLH<C*|j8VqrGA$&vDk-xPHG+3HnN^
z?uU+Tnw{XO(1*_aaZ?Z%r9MpHNB#u&xYrKlU=oKEV4C}63X>E(hZNLIjNxQ7l`)*@
zQP%o<x%)lbg=wah9))8B<SEg?1Ta19l2-Bjo>#?CrrQzPUpgbNptCDvnARU2j`Ydp
zagy$Q9k?z7FY(f+<cXA$ao$01O+&-=AhHNZIe2@Tb|CgTw@+JAi(P~RU=knHXMwS`
z*xl0NT5{ruzf+N7gy-=xxo0b(G*p1R6IO8bYzYxwfy$hGn=mG`hcummmCG-yjN_OV
zmElTFBidS5D<gxrYJdtYmleWb%1-8TA!OG>s6|DY{!OEPT(@)4S#^x{ec5=~7Rj{T
z8Pt*_r87<LjHmmuw*LuolYEHV-D38AG&a-C#x~a=5H4Ap^AHLm`-3N%7<i^`8t=Nz
z-4wVW?~^k)08~s*@s@+O?)3jnOU@e?X_s|o2cetN<&UFMSFXZ^mYx`DhNrM<DK=d+
zJ^4;3!e^Qe0g(rmLyN;w-BYy-b-wnZKszs>Qv;zAEZVm5t78G&)%xiYtibTqsvmD9
zgT78phT}3QRkD!6d}@Pg{i+IySxG4g>7Ul@ZS@+2*f*Xx>PIW~;1tNhhzDoI`eUFx
zbs%>hY6fMel)2bkVLpk1d2;%+^Ou5lbt4dH!MGN?;rC6_P4Rz66+AreGl7HJT{qfO
zZ?Q24$j)hvmzWeppiV6tRK(zM3pvilXG2(YiL?E|6L|%uv!DNXMs?8mOKE##+)kLh
z`ebq{Eh8~;MaYtqV~wMrz$%_i7E3oYtQgo2nqU*$)x)%6fBnHTfG>>x<pLhhm8IlP
zAv5c0w<Tn-aR?)YV9RIC4a==c+5x*PdKdxge+8F+Rqnwovfiyo52cX{I`Hu>(J!Fi
zMuEBOQ1&|oqAUGe@!@{5d<e@f$%(0yip1b_aSK}Ejoo0CXGLNPrTv4zDNb9lmoEvr
zgR2v2z!c!)1Qj~+-oWY;AyeTet`U!@`D6zUW&<P0>dg#)hKTV~v;OQ%{Y)*&51*iv
z{QiTSsie-RytmPCMAsi-_g^&|a(FfDZWPBnGiMq+b8^^HTRX&8n_v*#v@*g4Lw`@=
zx6=>fQlU$<khc==mF2|{QamGPQ>u$yu0mb39%|b??G>XM^~ornsVhdUYs)J@SWvhF
z+Vb+aH{)s+4k)T59`{{Rd37-zK=6@J((MWc`$!wZ_B(L(^+g^IceFxS@dpq5tm~>w
z@DH9l_}T4C+;);wMxk}DIjKhsZZy;5H<c9*k1#2K7^-ZShs)DC6vwE)0wTz+CQk|X
zxr?tesms*N_FV6zoYhJewnJZx`;K~w#3vBhdLcb`JD+%hbO)lH_EwKI{xi`3|JRWI
zkMBwD+&O)m`-A8GPs#UzpW{LsbHB2k%@*HTu95M>wzpb;M)}LQ?$Fgw+<ZHntr%OU
zNJ3bLcz80n&5rd|Y|ehC)t~pUY~FTwK{`;_@-bHXu(kT`>K}cId!^-z9|KY{lA5eO
z5-#=j#vJjvo$Rdde{=3IkVdGS@b7PaWUcafAi2K#Y`RZwcyUYNE{Zy6zOR(kl58pS
zhakgc(vR*!k3O=589^Y=M7JIIXO+VcJj_#X7Q$mmzKRx*vIq2#)T?9-rle{hCpkq=
zRpQu0^W-91u<@2Tzd&TMbaN|xUsS54$;&YpYOS!<LLwn>%HFnkIfdDL)jrti4B4j_
zpxVYIQF5B1SVvojpDO!GIdsvC^~sAs@mOL0nt8L{H|JJUB{aCF5jsCoxq-VJ1idT?
zm5@=tYB$c4%IDuPBn*=KS}^}TA4%}CbUx8rBZiEe10`n*P$s1fqNf&i3T|9!LkG_S
zDXI`65<x>!n*~NsySzkPK#n{14N;RTHKqpI*D}<Z_b34_wgGRF-t$Z=Mvq1W@FNL;
zYxvsmA$4rqe)wIg=Qwvgdx<NU^E;gIkr}H9(#=~2EE2mpxhe0*3zfGO$}~_%*QU}z
zro5j%pz*?+`;iotCRr14iAYM1pyNm3i)xvFYxrf{?yuPB7@fdEthu8-*wd<#{#Idt
zx(px}Z<{zkaR3efFkuD!G~vw|qtsPVDcPtrH<HwnYR$=1(;ie07l%~Y2@$qT`+yyQ
zy&*bP%SmUhJ<Yq^IbiOY0+##%5lHADcMqskJDfp2!30&t<tE~Mc!>G+_|w=@r%YzL
z;vu@qnz6F;XRbM+X=f3j$DXHUXNKyBPSIq#mL{ui0+q=A&h@gm#nj6>Z%S064tdI+
zn!Gc|XK_7PS6A7$T`S{PPZ^VG{kv}rsEuo!%;^v8IP$2gpxu2Ln%Umi)3p4@266X+
z$pupB@bA0ILCl3e%&0o3OVlOeLzn`$(NSFDH0k*4ij*kVt$jd%l=L`zkJm%Ag<>o=
z)i1<Acu`Bewobg37M@0FJE@LGY=3z8@W$fNiTBJQ$BQv)`@-EMJ%~pcFE5<{$}$gM
zes-2ai!vPc7S+<ECIe=4v)9U*0+A5eaZR1)%G)N>Kj8J48Of+N)_a^<<4#PsSylc0
zH$z|7Q`ZvNRWBL{#}BjOMHrabu%6QH^EOo&B(NErsfp?>&cJ{R1mu969^hSvp${^v
ztQfXVzR#A06NBXhQx4wAbb2mu<diMVnKyrzbAd;YKWqSEd7a3kM#qwuo2Qwu5+$Pc
zA<fq0$)cOORHf8V&51c0f`+XWzFIlQ(>d}laM@t)ZuINx26o0df>OW&JJV~J?_a_D
z-uIWeQAy5*?hC0LU$@@?VGgdl2`q0IuZCqG>!qfC%P<+5VUgA|>+T(vY<bak;J=u$
zDBRN1?4!AXqlW{bNEI@=rP4Jwn%1}l-z5)$$%7hc*T~a#Kx9-f-+D{?lYT~}0a?Wt
z4i+;S<--(19V_L`dgwAB`BAu-?0Lx~JbkCpacZkn+tT2)uQ-v&n|waaeUHwMgNU-W
zWerJWaL^yeZYO+HH!gCl*6hm<YYfo-_!vBaiO7f$5yb>x@A2|^&Syu$hxhMXmfZbq
zGpczIYK@jtEMT6*ca{6EG~!mqCara#x-8H-pg!-I*5rKzbc%$6Qj$#l2-}tgoGp}N
zp4WGF`gvw~$k}igp*v$z9Rd5H={UL<6O{OhvHUxqs#0SWAWtrxh!i>F$4Pvk?>j&~
zxbB=OZr^7!S$x_guX4t-Dga(pdADLGFjeibd1P?%DCV*LO~KyUoR5vUyRVbbQ;|U3
za}gum-%m%nN!@4&XS*U*5%2QfS0WP!4G9a3`K5k0Y}2scC`XP+rn^%OqIR%meCurj
zfAE5rJa(-^g+2ZM*6FS0*LeN@Wsxg`oX9CvTrm#}<8O;zc)KTFzC;{HQ|BFHZ$Nv{
z;T`1Ou3ZGOrM!T6X&Gi}m#A7jlA;%Y@Jl)kV6@nTyR26G?UPqLW)icG6qMj5@xF%}
zBqN(4*1sX*QOiW{U#HoNhkV5T+IeJDXGg|4Qh!L#(*!HXE7Pm@WA5dpd<lLEcv}3T
zYgCKpD0Tr_OE4@xK~|plH2z!mBV7xtzs#sa{$4kYW;GWL7hRu1ikC~6iKh2<g>fY(
z^{O^2W&-Mw2nJ<uuGA1WDl7WJvh6C%GJYhU$Lb?@`E@Nk@y}+3<Mu+*2a7$W`d!n@
z)9{Wk-JGg`sc?%J?<E<{dVbI3HE5lb@0i!A8$D428`zAuw?#3Nk(8ZSVY#n;{p^i*
z^n^?pEiE=o4=t$rZJs8mKHS}pQRj}7=0*IbLaJYGU@BocNdm-MLAXw?;p@ak*|un)
zu1$?efq-h_w>ZY48BCmT^@XR)RX~N8&M$j$oAu-^q>0#&7MY8Z*GP_9{RdqaEsuBI
z=ldvJra&l#I`*r=SA&^#@%%j%BFpgk#70OO)N{i6EjUB$yN8d}@1d54y~ZjhC6ePs
z&2HP|8D2BPJ&iKOZt3_;Cu!vpVBwp*I_0k~)pIlx=)&%2s)cszY)D9bZS68d9+sYw
z!5dlTqfH*ok>0BXS}-Z2v+FGZBLdHKPzZG<g><#rie{cY6h=2lI11#u4#;WMDx{Bu
z*eppFar6~Imepbi%_l76xDEjE@bBK#WBVaj_?UX#Zo3e{WcF#<pFPrg)C#M*-8U0?
zp+wm>4l0JtZYVdE(5d&e*qlK(G555>Lv8HVTw+}%mj1caWA97Oc?khN<;KH#p1YAO
z0<a+MaglE2>|aOG>GG9p92OyL14VsSRaUwJf7KRI?#wGlmQ)Cl1#p-qZ~>G*?QJ~2
ze3|}2P?fPmmA3v@bU_g10!ceEB|@(41-1Yf5U%S6C#uX>wI-|B4G(gfIeaQsw0&e)
z%;JNLW0;?aVMz%P#S^B;Gz|OM>vP;;I0zRWZbdfO8A)3@XjGNNZq*w~z?2N!lTW1z
z6chBm*p1^aBn}k}z}CtDY~stzw#lOxV=)2@QlnR2M&CITe4>~9#lmD<y-l)X)*oeD
zx^Qe|j9`l>#39dkt6!Pr$@;0yFxQlnOy<RqQ@$^K9!4x2&-``~$7}e%UGVXo@2X08
zuH3P_%Dr}U`+c+*kZjv&q4Fe>jix>16eoP>c(t7`ZQ<oOcY}*^+Ww<(HHBH^I4x2s
zqk9FmO~tKBp2yqjt#tdYcCmn%0^%wXYNIcf+*#)>#uO$dp^~%m61r;1RpYunAn`3d
zt!HvVEyY(~yGd2%bSrkLPeIu+hc|7*J2@<;l>cov_%aOuZ@3Puof=+{)S__KNPM55
zxhfYL9YT^xC|pI)cpVxy4g#Udaf2<Jm#&)U|7n8YzyFA*;UhCWEk)K!Ql1o9d%6um
zS*ej%UUo!jRnIXJ1xz89ug_S!_1if0cgx2Muu1xXvkVO-;CU2^-|BIc7P#9AAX2Oy
z^ci8ibRu$^ntDBaZ_6_$`02-}_a5G+;yhSwgF_M_@_7_iqAbTrPX1b~3qF%u&Im}A
z6+RUcz2Oy+!$lIZY&mdl%WxsB7oMF8ZwDPFIRcfZjjq`HfuWYMl5?1@jGpWJE3^v5
zFS;^9d8NgZ3{yOgdx#aSwB%V&5JSfsJ=|EB{y_3;g%rnR2H$URc5%XZ>Vdu?8#!9J
zY07Dth_tzrC&~x34WffQ6pGBc2V2R1qJBilIL#mNRF}By5F5qxCx8LFcf@bkC|l8Y
zO_$M+;pgFdtGLPbQldP|6QTHuOC6cVe-~dO)DuD^M1Ays!q;-YA4q$1s(=v6WqsaR
zt~a^G8ePtWVYAWFtt@Ks>|_v61Lw_Ol`_2%u?r0f1CaVyqPB^I@KjiRB?{ZogPn^3
zdsl(W^=1-ngtn}8%O*jJ$!Uu+u5%PQI!Jb4x;j!WN-C|vp>kq8W7?fRA!bP0B%9=U
zN2_%QIYv*aPS_8>K5dBlvy@zPtGWK%3U=Sc((cC*Q}Rl!9fCafU~6n=x^K<Z!eoU0
zy7c_Xvc*U14Xtx`Q-is;iMPB(rae`l;=+38QSIx@4!0$-<M1glU4fzUW*#?a{p%hK
zEOgr+@O*NWU5Gn&Ssw77@1+N<UB{U5qT!|2RKFrbJiydpDncwpO}=}`j4uRAeY&}b
zP-^DeX)a3ulwj(EXtz8b)xYd6s2#eHg|cX;!N_);IQlf+Nz59VWS-zZux)z`AAGMX
zQa6BLvf~wnv^hwakM&#=QSbgG7T*F9Bh+~|dC=4M?xPk=-S9E5<M1rIIvHQ4TZ%Z-
zV6^%%CZ2&O;0wm6)5n^?@tiVgiZTB?UF_-Lh|xxxnS%3r`Bc7o`9d8=O?c9o=WH-?
zdsKW9MuD_M&nq*g{@Vfm!z~j>t5SrM=C1hddfNieK1ZHX${Hl6kt;$g_>`L{y57L6
zRyH|XKg}SrDX>9B3|$+S{|B!+Q-&^8GY$zv)8|cYvRi*e^Su0?Ne7l@?XuUY8g|Jk
zS*~M>@?j`zVogJ9ufz~k{p|g8<}57A6`NJ2t;A|3^nM+R;$bZ<vDDUXn=%T{yZAk5
z->}9if^ONPZLIpzu$yAllr}^LHvrUPlq;rrU2Vevup9Czkmv%yoG>~8oES2*vtUpR
zrdM(k-QT||jIl9#rnWUt|NYd3`E3SrlBde*#Wcm=%G2al35@1LVhL0Xy1L2A0^TS^
z=s@EETs4f?MaC$w_HFPBm>HFGl7vB&ct6=Xc3?Y|Yhstgw0JIj<4qCpwGr9tcjv{S
z&7gTwG+TsgL~WhW_z8B=ZI)>zh*h1vyY2a-AkI`oMw`fe`vc5kS6BH#TC!p#Pf?^t
zw1f$zt`q)t)2`YK>qI#`=tl92y<8YU@~XiII%V2AicG~p+ED=cbSA%C^l7v>P}hlZ
zzo!TQ4{>x4tW@|1K7yvo?amv9IF7*tBkYiL&TwZ@l?*v@)<WACbsI^sIhwSpDT-pl
zsafe33g&)xzX=^Pe1Td$VtUp(oI4b1sTM8gCEF9>bB@b(v>Z`;8;mv{E=APR0>!04
z#NgfeUb8B)U~_rP;-kFU`6r5|wW32$YUAZpt0!#%^tQF^T(DZU_`v-5cou%5VzSk~
zv=}-os3@*G9$LjZtg7@6V4ydA{a=6qUHJ%V)`iXP(2JLc&LtTr6v~H*-hiMia7LU#
zT#S~u6e<*D4j!@7pBF0*O;#+3qhOP-(p-7@ziy^L&WFcu9^qL-Bu!fG*=JDn-+S|~
z3+~qz8Otc_-ZzovjA6U}^-5;l9(*{@blP>jQcxJ+ht*|77)p{R>TTs4QQFo@#s36|
zFLKoY7&eor-_|Y*J&)b4-PqWU80sr;sg^<}FvNcNtpF=ond}%WxX>zCc{Ff;v()Fh
zd~!G5T%!marxZQ?a*I-5SU<A3pIu6$o_W17C!rYjmp!U{trfl5SAJl*+qVJ9Y>8dw
z@Z%@KAL^BAJyXeFUhmRt*e0riPKB!ufOl*A6OmOrg1D^ksh5-~0gS!j(x4?dq|QmX
zc#S>z{IPl?xAf?YCXx)c>%WA4E&xp2g@&orZy%_-<^!Fnn20DZ%K^Tz7#U_&rME51
z%n$1&l9g4M0wL|e#t-YGJ=AZQh^Ss5h>i|PN-otR(J+qNuA9tcOo_pntCU(<F9WC9
zXn+)J6-v~*mXjzyK(}Bh3=~?xmlB&OwL{(HsQw49sHkjK*g$m5%Fb;K=>wl%Fe{I!
z<h7PzI)rZ}yy?N-iG_uxj6@lAU*Lw<3m3MvsP(^o-@f46;;>M57OB6leZmH7ISDK{
z<d45{IKiBPH~TvKpe*(agGTi!dz`}u&*pYJZ&v@}R-fP!|IKg{2YWB(P6fzTG=p@&
zsucCtr*gj)%VIRbK!LY4Kmm=TB6_jnY?IHhquuSZ;X|rf#X2WyY=s`pUA<z9{cdj7
z9mg`doQrR2>&)9elVC1^xu2a`0m<HLtfa&u(hrxnPW3}CsI{Kw8Z0l?4-VCsQc+kl
zAhHZ}n>|FH)b|B!)=cbIG#cf~L`ymmY({OC*-4WjtG6CP9T&84k2x$t2F_r{x|L=s
z=An|@Ur+Nbvc`e2)t>)WYG|x<^uW^Pxt{ufg?DZxD+vklfPTyh#OHXoWT06vEs*iF
z@xY_mMRQ0L>!6(2gmR<KR*T%3NE(pa1?Xh5V9$ZpKWeNNF0DV}p^dp$T7A{$kWzMC
zpZdA+bd+V=o;@YMXl-@Axg`pW7pOq2-h=bBlLm#|$55EP|6e=JJt{KyuXv$hm3Y;T
z*!tC!d#EM6M+slNp5KSohZw=<6a#(xKS2xEwGY5=W`<*1;=i}F_<s;iMO8e~)<`>R
zC?=+6XB)|5Igpjr$bYgZeo<&BSU>(`dG8v>A?ru_2aiDD@OSypQ>cSH&>D6F8-uSO
zh<Z5M<KY*i8|E#|3!G=H&)R2Z%cFFi??PR@jrD6P?#B>st83bjBdCcR{Q+hoe&l*#
z_<S_x15Wh|swgo4H#g}tEBQC<N0KU%;hqTU$?nou4YmlQIA2ZMda*L6@q!wuRJ~kC
zo`k@YE$6&4nABuxL#BteRR{sBG`^Ob9f50co<G)}81C<hQTH=k(m|R&f)ZBy5N#zM
z#`4zP4bI131gcqn8Q!*SJ>k2gaawR-#zcCl2+K$I{w}m0Fk;M91B?w@D6G?D3(h0{
zhD+|2RpR2jEet(TP-X{W=LRTG6>tND0V>?A;KV$(0yZcpUU|6Yy1JTUZ`x%w>-OBB
z-BI+acXtlG0!!#*Xj^s-fg6VJO*hIS8!-M_kxR9QdoDrDkJRdrZsPH$JYPPHg)vdO
zSUQv27Aa@usYwB3&a%5o0vrRG`E$75sc_>y0OhcMttw<>WNbK~)jvz1w!)C}{iHl&
z_Dw+t2g!5_vB4`r5+;+T_e?Sy*LRPPeET+J9w7X#mM(^}rsi9PABnEP`#o>;C+GWF
z_Ms-;0tT;rFHasP)8pFFtItbM#P^@f)qVC+6e`5@?y2?o>Q&jD_dW%^sAgp8zQ}S7
zqYP)9^<IqBwz&CO4?Xdl@U}dZ6=`KiUiGY1_K9R)O6~I**c;nl%ontJLIC;pRO9i7
z1v&K+Gt8|ClLwi<*I{K5l+v4+j;&O12<og!4T<F<AJ;1*825QUewNECEnYL7vMN9I
z%1Z}PGPOfxt9(4Qr9IOX>zi&W^@KLHv9NVOh@0^)x6PT=u=o^SqGy%6%UidNt*M;E
zb0$h`ivI9xm*yz=w2|3l#wb6WeMiTn!7j)}Ib#^fb3uqL7&|Inr+WNMMScdQtBf7O
z$u#;f{(hdOd>%b1S{~&jOVB~dqut9;p`(-YR{vsyI7THz;POSn31R74y(O7hBa1O%
zTYp+P+~YJj-c=&xmlbtd`Q#2HEH#A!tVxgWbgcfVEHyedt6Z7f(aes8f8IKmWnV$A
zQjg=8TlHA#LeU3~4+1UDeL(-!o($={C9!b+7M;fpiI_Tp^nwW03&2GB>mGxxp}Zrc
zMr-#)hU|kGtk12wYgsN#*p^;#3iizAya3qIkmeo2`daIEywB}aOL<|wZM$3PH^hEs
zY13llQ+te6&P|HOn@g+q;9@r<s?MFV2b|6;O-I=R@O~RY{z}v*oq$DI-dl*ue$k_G
z28*(8#|;LQa$pLGXudAPnOE3vXM_M%>obUtNww<Wj4Him+D~bE^R@a?Y7`Xi!!NVM
z5yR1Z@W6RiSHwv=HU_&&>jN47fQ&Q=b{u8!e;g3{yq|;qc2>u!A?Vb4xb(+{%gIpL
zJ)7^~$7Z;We(rhb;dseGBtafl2QL8p8JJP5E?z{f&u5wKT9eTS6c>3K+qz_IM8*(@
zFwkSDm#;3gu^8_&uw|<^gylOh(;x72soz9!%&Q+qMtJp6Cx6d$15YL9M=8qtfy|lC
zrp*m6^hT$lya-5{?o0d76017Yra-qm#4t`nmY4QXJ)2UbvNLe6?_CBgvFt)~f^v6|
z17pkMq{8YnOPM_Xmg#sV7w6*^a*oUL;ix&yA7uTrG6uxa-g!@_U6zWcAc|GR<t}Kk
zViYPgP*)7;!6POR=B6#*R2C=DJPjeE@GgtpctY7HFvSqij}=4kDBkPs5qAt4H0GL@
zjP(GXn;quMrh5mtkE=&(>l~yjsG;;0*4c_~P}LuD=#4opA}J!z12CLeVVdgSiJY!c
zG=ItWmZf)!844tNy3H(a%IRk+M#(0${oYZqWrzjGSGN9Gz*bR9y{@;mM*YEiFd;zb
z{_1OIu6TZ8T?hz~eSrY*tq7XpTaj8~*oMpMT2~EBY7L+UAh+_&${{$Jpx~9)w-y5D
zw)C=L+D-<^&Yp#>LPjbY@;&-@adtcn;ZjA@5P+KYwzVX(PpdeC<zr8@Yg<f_%cM8B
zw`@W&K{O~|d{QTR!7ZvvQ4q#HX)C*9W(HW;H|5@a6+dAsS#A6BU{^Fg9?58W-gfxd
zpvECVliN-{2UI-tY)K~pSTw0)%6!hvXklpz<g*B>psB&3PyEMf!;pSP9{RiIIX|s;
zsaXf3_-owxg4t?;PI>OqWi1%ViII8(F?0c^UlDn}WRyp2ZJM@xt}sKKslu0i<?p%g
z{7X!bRMElvx;StKw2{oscv?veW-SH+K`5hK3=wv518%@N5O@o%-Q=J!pyiC{qx~P<
z$_J!gM;M5R6i!jT*LtpxYtC`TfCF>FSe@C0fcFVL^KM8UWu}KMaJBJwe}Co0cM}hU
zP}i&s>c-HSupG%9;IfR+Plz9ESp%=x!}2a0a;~*M8+;l(5t8E?$@%RRWM-Q0f{gAu
zQz*V^YX3one9el@^=Vy3oe%5`+vZuD?WqMHQP>g30v@azR!2Oq;T1cW#vjvT&0lLb
zuYeETSqW4e{jOM+jY3Pmn5#)@%}QB~^kjBU$_x|Ld+H^hv#?z_39YsnHkOQm;-p&~
zS*BM;iVAbJy}LJl0{vNc0@z{0rrhye0<1lEA+@#vu8sSGd3&TW+d#w4$zMmJckkkB
zw7`x<xthOP&1zWj%|sJzy^@mjwrM?7--gi9slK)QDGv!E+xNKG)@j!2N};QQ<PG~2
zyd%p0(2=sY+~dkIqA>(iaG|HFu(ri0cnb|q>%0W*->E5R2v*}CH48nqI?!=2IM}E#
z)(sl5H9YeP-)`cWEUACwjN~Od6jLsSpw1-L#<KZWC;hGj8<ykPrRR|>g3IS~xuD7)
zbsUIdN$(`}obmsMxwim{>xtI|2~h|VG<g1Kg6p6mB)H2kgG(5Eu;3062n_Bn!5Ib{
z+(~fv!6jIL!2=8qZ|>gv_TKw$?Y_6Kwzl?EP0gt~-RGR{Uw5DW+4r#nN<fJMu}bmD
zMU%FFV?K2gRu<%M6qt1+Cw(E5`cC)89%`&J1J+KKtJW5reM+P@GNh26mXeej&R}K5
zxvmgNg2HRH?hYpz@m+3S+~P>#n_IfaS)$)Myp<bLxG`V-bVa<6jh3%z!8qUJG@n{x
zIXRD=cUti7R`h1>P}uy6(Vd>9Ej9rQCZ>Eo;G(TYE`QUUqTw4YMkmI2U2JMd9t*rf
zDThOVSwQd=MI1Re2pLBw_jM*;L{46!%3e;*0@77iBB@28RZ<U9$<^DZtaguW>i?Hm
zrY1WMHk9mqar#u7=l~AY$^78We#Fi_PAUEVJx)%(hxzQx?bOw?dz>D?(|eroh9hIP
zdz{)kY|L0ob>QV2^YmYlcb?$MyZ%KLEJ+L%NJ8=g8QA23#pDX}<Gja_f}H%k%?g}l
z8NDL`Nu7wBA>Ula`OZ#?IK4bs_3*gI`F((G-dRqo!s0znXZvlJ$z=BVs8Zk+%a2Tr
zNvUh-N7L)9;Cmd!Ey6@$n8oqU#isntppf4OyXQ--&}lU_k@V@NL$;j)>t%AoGS$qs
z@UJtV>SBt2e#ZyvN3Iny_o$JI^L!aiTpmhu3fVg><dP_`X|TO$OrRFO$sJ48bs(HH
z%^ZJSyhFmFUWke#Zss~{;q>p}b(q5#jMZMtV;<74f_qY+p2Wi5?1Lhd>aSZ%tA_KX
z+|#}ZI{soF+_l3PB<~v{%F3(_!T#$kmkyq9#Pol))U|rm2Q~*`HU@AHDT_JkXv;I2
zZXu29N<YexreL8MFfei~6XJ1P>Es6A5f}2h%c{vA2JHCK?Wog}%sQkDz;W~EY0>OD
z?h?o%B}ETTEhQSLnjKfR-8@S~YqOe~9BSGVvr2=z>L;T7gKI*)Ou%@I()OS0Ci&?|
zO5u74(j4Ql4{(@ow>->@uffI_giT~8T=S=V$WJ;A(Hv9)6vuqkj$(hK1cxeX!;}6d
zC~>Ryg}}`^!l`$N2t;Ec;#ZZc$Wb43ZlnD)_BRDKWK#@7#4*Tj0pO)u%v!>TM3T2N
zOh@bN(YNrFEttMtAEgDd`sr@Of?m9ZLBdOJ!!Kw{@hXsa;w~ChyD>#mh+0=I=uf7}
zG)&ne%_cI^DP#><j<_-uPM&*ZA^FP8nEhH|yZV7S)EC`g(5_!xg;Wa>0WejMvCjq|
zb%MDr;cqJ0RzHLT7#|>Sm)}xVd`2=>aRCCFyyRnd0R&QC?aq8a7Rkj*gCF#5c8Tar
z`j59P3o=O<yX{sR_d}@B!yb*rw_u${(!t~5kc$xiwVt7y2VJUa%?A>g>!w|=9bm)3
zZdHYFvuWw@+n;OU&wTvVf1Lcy+GZB$QY@p*Bv5o-?+jbuH?Nv_yR~qQ)lz8p_4EA%
zgWn{yh=mlN-o8?v*sif<_wg5&$F*6(lhsifP!XVe<DakF6-m9l<Tbt~8%;eLPIBnE
zX%pb%QIv;fIg^G$JY1{*RxIA|;)+NH?4*$>E?6ODY8<r61;Bj)k=Ue%>0CCwUNsFX
z%2>t_dhhScwoVF*J99Fj+NEyr6{kW64nX|`^D47nd|P%<rb_5SPL~}u$=LF=-1Oe^
zpGImjEqBx6O%*Bd3KpZa2@yA^7$pam1%VxEV<~u&RSG7d#)2yQ(%Zk@<IXc!b~i0J
z*40x1HL=uypgNrz)+yGh{9!e?S{V^5Cli2CNuQ)Z*1Cv)-;VB&SXpS;wxcVU=F~1R
zN{6E$BNWqEg0;lOVtHb6NVLFO6ch%<QP@Yxf#hUCizQ)U$&1B>ul`C@l~}0@rl+;c
zOGRlR@iGR7yVxW39Lxsms_UHEOTkX0aSO>DhvC5#_c*REqMZ|J#)_9k^N#<jLhj7z
z=P{o#zrGaFKi>A0;X4g{Nzg8*$v5?M?e*Gs@SO2o^v82Eu&b?F=V^E0^?dV6EJ7=5
zkL*^>3vug>UZl^nURBkF@e3jdvsB+-Zn4BEa#aoZZYpt5>IY{;#a;`k58Os@2d>{r
z{j_@F7J*6qSRgOP@g1lqW@F1+&i;0;U*@@14lnyxUFD>NT%m$arGKX6WvwiKZ0PX=
zNfq?8NS|j_E>P<o^pBGOl4xY>a$9MNOQwu1<eCa1c185{OPAbaUFvp9LmQb89=RNQ
zhI(zH!#IzVEP7*#;yU-*iVxCBNe7FaRo^Np4=2O$wQ6=)ZMEuJQE{Mnfw8{P*rVJ|
zc`ij~#(W(XMJPrLPD)dfD;c&S>r0`fMYm8e4m_7*5<|Gj$)OCz2f{5EEbF?f3RP0Q
zoD-&AoAaI=c_0E2`q(h-Y4rDZ)7Nf+4G(9}?|zRH-vvXT>bhy&Es87O<J3bmuCyL4
z%=!T(Wc_^ku&eavUf#2;uF;E^h73-5QkcH+^E=vAjmxX|HzkL6{hU(gC3)E3R(T<=
z*@Y?vJWS%;#f}R^o*lx*pwy9`G>|iqH%Gz^=K{?(?zmt6%_^q%$rS5qrXG-a^D&RT
zhKByrA}hmA@)aj$12fuVGJtAjm^i8<=@d<)<S)=IA&a#;E~(5_mdziYELauLXP*S_
z>&V&fE|JwLh|%?+OIQylmdLj|((~~QhPrOCipbO)Op$Mat>2SHHYJ4mKo>LQlUr-X
zB%;%AX?P&8?tZ~RCCM6T&$R+^XcWU+cDSw}bnyN4)&H$#+xJLky=X?3TpqkQJ8l5$
zZVpk>I2OX&$O)en$q;mu2{azGmdRTyX(eUqG|E`=Mq4pLN(wL?sKqkL9KpRgW!Jzd
zufn_s*O`JENofw}0V!G=Gk4&T%8v|oCrIQ=(YiWdb*(?x*^onIG0aeiwhgXVYNT`9
zY^(DcsTcygur_T-HTDQstF~mB>w3%QJxcPqgTsZrW{XYb$08uw!IoAg?J?$R%->vf
z{$b~YmxQX92i*c%NJ)&E#!|URKQ0L-@d^*p<RHLoJ9z@`<E;4ZY)5@aMncbt54ElQ
z2MaT3A|cz=&@4+;Zb$&RoUP>oh;Q3IfpX~~f4@%*I}I9$eD~H(Z_I2!f*tKro<|c)
zPeju?X5PKzotzpo2xM+r*n2n6V2062*7r^ZsyIw%e1H=PuaPI_*k~2l3IJ96`g(tP
zB9F2zoT;{*85@USg!?*@(GY}DZ*5<cUD}3DW4SCgf*B+yQ~#`ZLjk*G#;z8W+r^E!
zxrNGMskwy{wd`+{6ADOyaR96>1ZxB_NK3HJAt9kGwEh_m3k@wP4h_TVN1-rxtlmXJ
z1^^p}0vCv@S(7&8z2O9JgBk|+L__r5YboJ|j>1ssvD~#pR{3}y7IQ7eHnJrm&R0at
zbY*^`kvfu5s)^-xv2-&`8^y(8>3uWIv2@sNJaGF@2?vL>oNA&(xT~?@i_Ny>OlJOU
zcbcSvwgnXfcb=-LaNsUuF@e3}ky+CflBv_YEDqhcjjB2@WM?Ru7VEF2*;GH3(>6GK
z>MKnBb%eb12oeC{=V=X17B6ohgCw9nOs_3q%J{DwU$?layqo(RI>iZzN-TJ8BrZ$U
z&3~FP|Md@uKESXYBt>RSxIH7y=k_qD#xj_F77~}>=S%ARcgPKz2yvvtk|$|QqbG^N
zH4z$|;9-03xzN@iz`3{8&m;>`u2Z6~NPsJa1Qp1x_SwnFbAXb)kWb!66&W}2uNZ-<
zJ^5eB!M+?-WVEG+GbAQwSof98rq(R(V4Qi)E`j(pcRS0pR;u$9<J^-8&m}fmAj~n{
z=E9Dx-w<}8Jk}5;-w}q94efKLC!#Fr_SM@GppkUfL(+woSyi{IxF-UVS(;*vra*Eo
zOikNUU|C&#o;o#z#jry|3wF!-|I5R9^sM_H$NU)Qb++x<3;WI14>));h~%R2jlVF5
zYrmH3>o2&UP3CdWJ`pcys+a!vHcT5j@uI!f<NE1fLPK^=zR6_u`dXAt=hRVpQoS=@
zw4t6wU3=<3EuA)d8D5_2r?(MA-xmJ3F>Zy(-2p~Jg4;F&)V1F}jOhFCH~bI(ebHZz
zam-$ez9Uv2B-k+pWN3FqwA5RzW%i1eh|phu(^U(^M<<h`if4FoO!F6jX*CFdPk2bX
zR{0cn@2)GQY(1LO1fltg_h<AIK?};KMwJJfo6C9OXC?N91(PIEdS5&t;8u#BhB%XI
z4(+wMTC4NWB(>>nfDo&Nva)vHgY?0zCzbJJ_Mp`OX0ymI0cK73F-nzAw9;Q(^z(JD
zBbud|>Yzug<9)>=JyDxn@dRPGvzA&uJdxrL{=P}2?T>g&8zYdDOvUpiPQ`@{Zx!)e
zpQRqC4CIazWGVXU`~1r%9A8OJ4r=opjbKpj=E-ArFo#<b+s|zEjzLoLK9_aTki7nI
zHo$|T^**!dv2_ihH9ChbRvREFPT*nXBWxZcWsp&5!;AmWXa9?*4}q%b4)9)rJ_psR
zBj9F-5yubBL+Y8Hr^5)%%;rRJxOC-m7wuZMUKkjEuQ+R0HCx4zuERaJ^G#HPeXqm*
z%=6vljC&lSNLrhUx%L;f<?+${-O7#f9T!&3PhvGgJ}6^(MCT*%)!$MfxYfRi&Hl2~
z5&VtrRoLweDc@qN;x#<VO0!G(M%!=Z9j3<dsC4M&ChyP3()l4kTnyV9a#*3)P12gH
zS!)JbkWnXVNb3+C6Jqhd*tq|p7_Q4)r<MQG(ocLG9M)>%^yZs@_t?|_rw6f73n%YF
zQsU&N9^0agF^7RAD+%JmQ(`HVLgkD}4L7JtCRr3;f>=X-#ivns*)oEFOj3DzV6|dl
zx9@64@ze$jL^cF5FN59R;~O7ZRdKq{V|}2zO!fMt+wj<%6Xzro9V-bfx^2?qJV1GC
zqFJZwIOVh<N8O-hYb5>AsNFkWheyMJT8%DMflgn>x+6@Hs`6evk9+BSeOB>wUPHiv
z4peF9RY37B@6pnBx|1`z$=_&nMeX_OVP1u`<nt-F)bGb5fC@0n?i9WX`&p0X!ao-P
z?|PM<u(O!LK~)cIuGQE+1#RHET|9HeyvQWI9nG4T5Ib6p2WY~Qzu3`}px-t}&rQhC
zQ{+IuSDk3zUU)-6kU**;J=J`DlbN^CI2-PIP){6E#1#D6bAykUr__1M;Fa@VTo-hA
zB1<^T=iNspW6HraN8=mUtE8VL$tvu&wv*9Lx<ikkWvIHj_N3i}2S5Jjg2DCmZEe6=
zt$$a|!D2UU;wUBZ4`0Pbo0zBbJh;KwDqwl(hv`%c8A^LY9qo&wzcRc7`SEOlqwD}N
z@mWZ_N#5tu{qXMn<M+ZbvT}W&lMYQ@F3m_(!2bbO1fyDaaV_h4>}Hr^-rxXq35i2Y
z$xduu*8MazX_0uR`MC1od)D~RxS#*$FZr(}9wn189pdsRBKok|iT#kZXVKAS*S_vJ
zAn>4Ph36EKt6f`g{}PAEu-Q3-I~%L^%Q&<Ix_??K+OAOLjgjr={-wL7-OThE=dy>_
zjDgO4gvKOADm*EX>8dm0!;j&*g9V*tY5{nykLreF(`2!3hb=}&F&^I>-_eLV0x7^e
z)Mr^w`lOKkmvONPaqX&cV!|NX?mTVPNRdJ2r{>Q}$xH$7Ea!zTIbtOd|Kuv9g5VS8
ztai>Ec|Zind)s0ev2aQa^(ox|H5u*^%_tm&dazjw;vX#H4Rf!5T7CTP9w(m%)fGWD
zR7bu7(j{Mr655!TG(cE=Xk_P~q%kLiH`<j<6!4u8wyE`avm|t>@Tep|zF67LX#8=e
zN;E(Edv&j-qdA;NAUl$>n`5b5ccr;Rn38y5b|H)c5w;fBVV*uTT|Nu^hc1|J!8uPp
z-c1@tUxjrVO<MREot(~{UVW>e3Qt=qrS-_n{MyL;%lF)RTDn~0SmMQrb>!5}^Z_?S
zEt{*q@$^O_N1*sG4NUKIL0!r;qS{<hCJb8U-yK!Ye!x9*Kp2S$u)oLo$(^C^Q!jA)
zIRSuF4+=mpzl;Q>1srzSb*R{Klr4FHG=oaO(&qWOk<@Hj?6D$@ozjNV303YhL^y{p
ze&BF5d==R2T$4(<ZTjx*G@mhvPU`pA8@1u;k>4$6$gUK?NI6RTIc8rPy>(hHNj3QS
zhw%RxN)I30k!qY&+bzvb2CZB<n5J9_UKj*6HKo_T#fxY^@#YRGJUzaf0TW-ZV)c}f
zSx;AskdvA!)4}gg43@QLdIESueV#AK+@gFu7P4ZXxBZTT>&I`K<*_W|&#{)QWtzZy
zoENrO5bD|iRDD%D0F3ZJf__YY?Cf|fjnIOU?zoFu-AS2&tfROtc=<WRTB*14njFNx
zIfoeA`qZt(1(;s=Pi}Cu&CI`wf%gr7EUcK<LD8&I+)rG9;xgksd|_RY+<bj85N4)M
z3wh?*eZ2_ti4iDAfJf4LZ{qkeYcw|vSBuvtmLxhGl2`a#0E7H+us$hG?LHVAs8S2u
zE=eb!abg#HG>n<MsWn3_g?lGqU8=r67c5Juf7w|SQ$)zehI$efP|bK?IWr$z|Cu#I
z2Mw0%cO-rSWoLri)65fS?B8g?pCM51(rvTnC8Ps7(m-9OsxNa>^v;zG-j}q~n>STW
z<$r475>zddL*<zn$acMz(h2RiA@`DOp|LL&^Ok?-P)pr#yf#vrP?XQNOx!JQ(3!B)
zRL2x4pEI#(s9d02MW=s0RmhSOnSvO#Q~#-meDyX8+Qhxa{IT?naaPGt?;ys3Bz-?k
zf{QPn*IPF{vN3!}rh+~g-1GchXWOw(frVCKt5Flr?s&r{>49+>{>#YsReWedeG_*1
zU^a=VHKR1V-3fn!X6XP_z)Zual7VUXXE(DvHW^j9*I(_t!ij!g8r<AG(cU39c%I7z
z<%AxuV*~*=8vHDoH8);bjKk25;>~>m^*yc6HSQXDt93fg6oIGO(?z(-9tZ|~CXY!n
z_op(Zer`_$@9JVhC`N-H)_k4LOMeQw00;lx)Iuf`ar+iYT(%AK{A1qj63@`;R;$>$
zohSi8HFA=~3b<1W&PLj=xX)kh5VzgnCFW1r$mx35vmn)r1ZpZJVWENKczc9l(5XnA
zZ4FsbJB2thr0rD5$k^*m4pbN@LE(LJR@e|PxO3{t)!M8hL=g9bCG_1!uQ0h1yr>rP
zD)`}tPrKxoHVy5RYd*mc7UAAhd$9{G3(G&bBKk+at9*2zT=E^dO|PKQ-{x5tXFsLw
z%kHRsc_<)3AzWcO$#J?E*y0?{)tFSVS|(aHn0P){pjeAkKPy=UQB&YQx^(w(X!l)h
z9hw9e*(H6Os&0I9#6F}VEADRP&`P|L5@#*fwrtWXt}$^l9l7oGO8JVTa1CBi0?g_3
z!w&x!KL*V4LV+l1FH&tV$CEett$zuNrc;Fz8s}gsF~$s5@{g6Jg8Bv8b0t~FK5d9E
zqI8KN&rUD=gE})}mbiiGFg5es_rAHEU&I*sVASTyrBYr&J<PE&Yi$n&vRAovckBL)
z+UfYW8Xz>S5rU%5)Y>*E>V!sU)Mjj0xO1@5*Arq|9`3@Vx=7zJZ5v>mRZyz?)1#8Y
zNp#i(3I@#xN|ZMi(hg(F$J@~(`vl~)yZBHR8dFG30wXE&ouA2bkEPp?t4_M>!+eRP
zdGaL5?-}xAt24nj@KNqN+C;=_AlR{*0!!F{O;9u@`!XwtK(+mpYHr(^@?W?nD*&Hc
z_gY*d4vMCK09z-{WeSelUmaKc$)?+BG4GMm*%8+WlC+Nl2HrTY47N@f{=LUJ?$?)c
z-edI5ZVLPk_;%2?51;jIH^0?2gU@*c=IQD(W2U(@j$SW#MDMr%N|W7pUz+#G9?$o*
zt?+m7cS4o<JJ=t(8f7M>LPwsU2=EiJ0izZb5!C%AkBk7J=h{ilwfKoMkyOzEipegj
zKfsnlr2;4&?lPHpMeX55{9?l?D%awHB0~zATB{uGB7Z4N^F(3LZH!yUQ$-<vO0ygH
zqDaO`mRzj`U5lDBT3j<f3H--AkLA?~@Lsp6159%8-5J}`;!xrdw%!}52<3joL1on|
zS_P^=@*SoKw9s&XOMH+Ga$VTcGVz!VI(bs=*7-xuHMf?j&fs|tN4`4@_c{IZ>SRB?
zQ5f-@>ut3uRUOAf<+-Rn&#*9SW=#*{WSPX*RKsH$f1hyulF8S@If{s`d?8e7s7Itj
ziKI@fY5pSRkT9fSlTAEldL5MCtbEgMb&mt}Bi>&ZgVit>`M+2-9fp{}(}sl|WtQ1&
zC;tk#x{PtW*vM0FJzEqsjB(q`*cW$6*NM0ove*DcGg9i1Vf`6-v>)-yLs$DCrL3Xo
z=R9*4F5{~`@&ZCsNVC=>V~WaH*4G8*EN?3Fq@#JJprvdGb7AHrhIMbXp2Kntt_H$r
zXc1HNYs#aNfi@`UdtzHRPqlH5OtN8-ysJOdg@vZ6usakzrQS?knBh*?P#~Ng(q#El
z>UXvBmE8ECRgw=SK@Y{A6sPzG3MRs=QLONd5kS#YCNG~7r$}PXUtDIP+gj?H7P`4_
zLR}gV7*bqY@I59|yKq2VHYl}@PH)U$du+bmur}6x9%Y0o4y!ZFnf<UUb55wl4&E(d
zVIU<>moIakS?C~GpM;VYyUk}><uS#C@Rz6TL-@6Ks`7cVI#4O$43@aK5XS0o_2hBE
z=C+2xYe_Nnyq)W#hG2uZf=ScJ8hR1#CpiFiW=?!}!_R*fm?^BRestIm$rnZ-AK@o-
z^ti@xP0@RWHNmM{A;C1}Pj`+dkj%ZxCY&K8V`g39BxWWRgphh5ZAPiw4My#Vx&&6z
zt1pZ$mwU^o+6m(0v^Ji3U(ZSQW+q8uWm=s~(+XJ7N{Xa>8rAN<1k7oqwuxB~;NIo2
zZ05!c!KxZkZ4#36`OHfLzb8d()&%d1$rqzoerRZ(dWUdc4CLXuJZDzyNjdj0L%*fk
z51Wy8Q*;;K!_S~{)IeQYD9q{OGckA;aqMo-l3G1_RuK%)&x|kp;PzS!X)x@mKGYb7
zQeYB)l=nk&Ax>PxE(R^~O)bK6bxyx!z=~(2!e6WMH5KEc(?t0|LRzFx0wRoSf%3Rs
zsXxx5<vX__f>5Q`Jf{-^kY6_#gMtcDee;e%KQBB__VIU;AFE8Z=U&PDB7uL}dNWa3
zU^@v@iG28CoWm!EQ5w=!s&x#+kbpDu^`V9ayDXGB>kOkP{T0`4DoF9g@;#1xNu8OQ
zYL}bhShGi)y5nQ4M^V_oD*I|ayks^WAMS4{`O5Bbuj`*fyFap3kFAW{w>>gb(k$1*
z$GPXbrIRN*nzkq|11Hh&BE={T35{PFVnrkQjY3lkL`#l!17z#!3DN&xL<=U|IQO*n
ziT7U_7fj|&+aW|`e({&M0ISntxBu>jd|Dhz8<vL6)mhU+yi>h<-dE=eMD<GQBCV}&
zd{m?P!-NXH@u}u}4JRe+{Xs{3p=6P$vb?)~+ZWo)BxW{YL3N>gou^}|B?WWZy>{kF
z&w|yeSg~@1JI}z$n<CBev^5JS&}aC5csri_F?ph?m1~O7xA~29!@=?UrAeb`Wz?%j
z0m-Xa&tU$PWQ=6#*6SU&ow-JRI;$O9cq6-QYO1AG1amj^XEY7j+Ap2%J0yZ|7zDeT
z7#X_r=T{y2Fd-k!_$!)*T%^0uu8K|<8Qtm~X>3*7H)Gt$J?=7Sc6NyOl=bv^C$8h-
zQ9EolT&<z9WT4|MdvLJZWo7L5`bQj50F{*n&KO;K?acD(GsWbwmZ=0GlQpd~`zsNV
z8_uJ~$y#MtzBtck#?`m{0M#Cf8`}uxb-Wakr7$>h^AkCRvco$wB+xkFB759Me9!N-
zr2n1I$gMr<*7%*-8#0h>=|^<Ljp*r|vN0fGS`wxa)tBgL^45yeeOkcxy0myh18rs*
zx6*G<Ki3V_*V9fWxajR^F@Y~8XVI;YY?!gi7pR%MoWw^OH*kBtRR0_b%L?wXN`V~@
zyMW}x&XdZ37IAkPT?+HxktRLjS{P5*=SolkPjHwGiP^fCzJ4Lr2ozg1Su$)tHJHcM
zoKA7#m>e1zl>CK=R$B;cqDJtM-B#J*A&qba2*xkMC_1CfB>TN)+rQN2??5R8zz{a|
zJ=$UE23lfPXF)e*-0WT*7^E-Nwq|C<VV=cmuKp5*6qk|s`{^EsFdMFY$a%GuulxZ^
zL{Lu@ZUs*>Li=dx9jCH~NT({7sT57K-)?;2&&3y6-OYJxTev7m`||hgD>M72;TA_3
zZ_+A}Z0sSNg3_G{sk0n#BB=D+Gxt{_yaEW;IcYuAGqSiBRu+Z+z&{YBGJ&-;6Us~n
z;R5|)@Xk1tBiSuda-3R3LJ+V2DyPSST}8&1`=$PTr76B(Dn%+|&&rcfxR<^g#ENd~
zAyq}oq`{Oy;+#VHG9UX<adJVoOz5UM1p`aynSOm_o)#7$PV=`BDX&BfyxIcEdf_Cr
zP*`5}`&GB;^0EJGd`KkTM?Z;NnwFa~8lP<Gl+<V`=|T{*unxpW=uOq9H|}tKQD;)2
zVPSgwglbJj*<d0G1`Lz;nfcs#b>Ywy3%{C<C)UCM6T}x<pz0HJ{#^?ZATye+NKb@O
zXlunTiSW<04!hK9ksoM$Ys_5KA5PLcX6Wu&C@czC5%X@|-wnw0)6EwXUoY1aiKMX~
zGyx2QW7gw2YnuS_ON}2(8pZ6g#p6Ih8a_rB6Gh6Wo@)+2xTtf=xXk=yKHHL}coOdj
zz<_Yo)hjpQ!HJ!Vs1Qlq$0pny>a>PWvKb}0wF^{P-x-%BCuiu-s8kWO=mC%XDn977
z6~QV`ZxT(cSoSvYp^u)q#%5HcM6!zD9VXd#DR8t2*=YTvDp==(Sb1KhfXJGWTzE#D
z7Sf{Xy>EHv(c^}7R1snBamFSj`1kCv5H61b*J-|C(FYg3c;W-+?{Toi8q50aHwgfN
zIl;QErx-8GugK5m%g@);|IsPYt94&vb{|wOv+F3R$NgQuhn65etJX^xs!pa~fXa-N
zy%7d2CMn?qSS?H~{i|BPIM>&VB<0t@LXgW8oi3ye!qMN2HOiDwf7q7z)rzOLAgZ;0
z`5+(mAOdj02xzl0Jsx_as_xcmJSiFO^^&MNcLWF}8_xy(v7Qhr)%TE*SNOt?SBYY3
zq%TD!yOhjNyBK=MS6r6()!ySMV?^(?^h}GQiUN<bYv5R19w#ib)Ug#a7MaI<HgKKw
z{XLFvV)Pv`>s{)1)wO*Bq(_Vcd;b$&J<7)mA#Tk)n?L<Tl9gK6#<xIhAjb8@>r8j5
zk#Um%amI4VQ3e4+ciCe}>P;FSz(1p{;x91PUGE072Cf2s=|z)jA>U%zdh=DapLoD{
zC3(4V4c<jr?yz%G#P5YrnAc%JQNgYxW*BF)T%Ib?GhvQi0k3OCDrq<?vmhf8P?9oh
zTU#UOymhZkHbpN|Pq|E<(~4$F&Tt9Q)HC_+1qGzy!<=grK%5?!ClBX$`#7Lq3@lPA
zQn%!VEM-mhLybiA{cD;grYD0!o+ZsE1Va<4U-db4y2vZ%?K|gPDhFkho|Y=hi|Y=m
zt8}C#`&2*CS?>0u+Vd0PWSH^=8(?uo#JGyXJY>|yA-~3lAuLp}jIv5;%FL^+W6Q}U
z%DHpHAQxi_E&Ikog{ZzHaotY}X2HQJp5}us2X8qC4!GA;Z4z}k01BM7zS23N03Qoz
zSCHgE7`#H|8%0TJM@(=>0h?-8YQDmX#_6?;4JJFf{e`_k9-G`3>4LReo{vMEltSs2
z1+?mMWrJpo><tr|nZ!$RP&(?+(fH4miO4NVa&Cnto|5C=^Fz>gAuNcJ8MWTR)qzZ`
z5>gS0%pGX+;S%n4kua^+<%d&T^_7N{?&b>rp{xtA-;}JuP#mV;99WZ4m+yXcWHs`r
z;S%p0{Jvs#P15AT2#Q%4HlDKUf!P3GQy6zI=s4742{dEg4E18a$=dNNu!pU86DCu5
zdyJvmG^Aq;L44el=@0QI(8tH`*a|^wf+XU$<C{j9cj1yU*=6d3*YSqna<H9tysO}7
zIM1aGcTxzA`BVe*11v;mBm0nc_%YZ3Dp5j4Ivf)l{Q$udAun5%KgC`wP?<fyYyvdV
zAs3dGZ)Mo@sm*pvtUi&t<~Z^EEW##Rq5AVrIA?t@M9>Q!jKUi$sFkTLt_^!@SGd#y
z|7$hz>EaJ~f1s=BJDtht6}VZuaQYwA8XX%|Q5pKY!pv96SH$<DjjrNpyA6~wvPPkx
zi6uez<Bc&VFaP5+{Fa-9Q>dJJ2tU4wCn*aaM}7pgz1I8G4ETVW;*e60;xUtI9{Wnw
z9!cb>L)e>mDq+XQ(4&nhodV^Ib7cP$0|Gs@)Gs20qruD-lqQ_n8W|R%*hmoz9Vno*
zC1CzShMKqjSeI*N*|BiVgK6PxsC~7SG(@E{jc!(TR`;9xjduN7BhT*lQ}>}*4r(Ly
zM)LKxan?R9LRsIEnMtYd-!n_}q?BiUrN{WPsNjmWugOQJ*pB6-A9=2EFvDl%3#emk
zj^`*U89SnCdV~dJzn|&!Mp~*Aw4|$hDL+;rR)>U4S^t*L>HF}FNc#C~oVF4}!Go}!
zBu+5;4+dSP(<o>zO7A`UpgL^o?VEwSv37(BukW@>!`3Ub$U&kATHW?MkfPGsZ)chP
zAnG0`lygP64N%lZ`nqO200(2eCbuu6r)pl%ZWX;viGPzbG=6HjGukv9s2Ur{^J{iX
zciGp;19hHogSfTEvLS`T?;5Y7JIA7MvHa;o>)FHTtP`KkK#qOJJN>(MVE&4>ZcSU!
zRo+GW%*kBr$<U^itEizCCUn_XP#_pq`}W1p2+rcgx<5iIHGruf!w7~~gV1=`4C|71
zE(iV3+j^yDhN~<CVOE|B^*rOV`j5knCG6kyEU(S`-Dq<io?SEh3TU<>YG#62`R`(m
zgUUGIU-&jZR8_xVnIwJh_lK-t-17AVK$srr5;3|1uzNAwuco2JLd}mG&s9yz8=`YZ
zpc{(RiA63hM}KOSW@U&}gYn~*Nj$AFU-J#0ppb_^St#s1u?r0r7C$b8l8(K2knn_9
zUQR*9-{ADfYJXRFB;CU@M6_Yo&gxwz;d~_9b~6WP-nh)9fy!ARuh+*CrkZC`7cOY>
zq5fckv)(?<?m$Fp5gqcrc+|`P>9GM^*UshCfooezAu;l$YOj6GK2szoKt#7l8v-zs
z4JUfSn4a}xI{QJCf^Rr{%0~Os{7#^E(mpAtl)BU0-$(g~Kc5abN(XgwY9Z38X)q$X
z8k>EPt3gFv2)-kj0q!`*tR_yI_-d|Q7wn5yxn;{BXeHpmw=pR)Iue_v+E|i{4J-sh
z-Mn5t9=0+B%nAcg>MhFnB=~Z>Au~Z~?g9wCMYHR;$G_C$85<LojUAC%XGLod_iDem
znel!V+R71iIjWJ%KTQldAa&JZv#lxF*U9D&eRBemOL5KX{IRCY!VYf~<F<*}aV4KC
z)?C5`I!Y@Wo(oAtm!P&#p9<MsmqNEC^P2T4=lf1b<Bn5b3lceLkwj(0rn`wcL*~j;
zEW8|D-m!>|n})XLiZ-W(Ydf~_m>uXiDn}TBy9W)Qe^MY$Pc|y<$(?tn)BhPEQC}kB
zMzzTTKP^ebe{_{;!tJb;agGmBUuCmsf56qc#A{_lk&8dNy8De^I2rDOtPFTw3fc3A
z89&M9rVOa&{N@%6D{-k&3T;hpRT}_VS#GlAX-{wl@_mJ?%>)yfK?G432XGjx+aM^m
z(|Jv}0F+duwV^ehD@<OEcC)1-VM|%U0!RIWJZ7pBF2N9~FORr#H7fK8|FS!0o2%-~
z7+t(<FM)(@*QA$#ukC>Qh^&Elw<YiZt8}79G^j!Srb5bl{>XBwbv~&(fF3XqsUELk
zltR?we^Gk&YF^}+yfe*ZZ2HYO_iJgF6uIij8ZS|o3K4MrOb+nP>E)$ue@a}!==Q!i
zJ$?oM6Wzssx(a*K>}&Ej2F+9jxI~?^_xvbwY!}&K`TXQP>@c7nU!k!CH<1ZMYf_N^
zM;o1RSM(ur+EG6WwjhR_BZSBXb-PlO>l7K)aW3h~ncLk>(a4!>RG@4@10#}&!l}Sg
zkhS*C1~yhX;+J$VhYT{8j6Y(%MpbA|_}8i8c>7n$oc&dWe8p=rUzU-6;Le0=yo<QF
zKD<$%+SWTY5tggl9a>GYDs7=RNXmOwvCpt?lU+9BVg1`C$_m1hPy@r~>hg+N?Jk4L
zz$lq#2O%s_DF28~-S92Dg$kQe>?vV|{_HRh{}wNyu=<AMA>3?p%Zp>tnhJ@#0$0mj
zBv5YJQCk`wm;5`9hrWed>ZiCRtkmyIe@>V|facmZ1hQR!s}~uW0QR(h7{B=K{{+%*
zA8kX!RnC^9iAAjnc-HC8uJ3(iuPEf*t7ZJI27asT4rw*+LN|C44T8Y7f60J%S7+y`
zFz*uUb6bU%5an{tqV3VAYO=BE>3gIINnGaNy$Y%wO8MUoUpO%s0D&^MhLWVjXX82T
zRkC}t7tR+GH!t@DVP^ehhI6gGPh5ExdEI`jaW!CW3`urd2^h~5W1W6|ZN6e*cV{VW
z9k(WBo%+p(W~BwZUeEA;TNy~%qD%XtBL=>$t#CN5T_`ziPVfN+3=h?p)r<NblLIPb
zBO{Ve`NmOZ>JdQsg`J-y>RSaT5r<rGvaf`8I?Hw?znmFAu1Jl=b^3!U6`|0mX{_OU
z#b8FZ{S5YQJ3>wY+U070FlH|@6y<h+wi30p6UQX6gLoEOX6Os`8&H0Rp60{LFKiR9
z$|^kOQ1)^N?X&_Dig2CUCj`wt<s}*D{Z1!TA&+#veElT%3h?e91MV?7N+*RYlQC|)
zXM<X9sTvkbt`AnP3aF!VRn;af(5KLiEUNf4zfDqQBqKlnlLD5wxYH|3W##8)ETXln
zpGTR-_5&%0WlGJ4j5&cYLFtIfO^%ZoHZWT%(OV^3xMKCOSw4Gl)wo{o&+z$oLS7^$
z<zV|qKBsC!PeyxYhxW{u1nHerhc?z(8zev;tdS6v>Ix&S;w8-9RkuOBRU*fH<C-RW
z;%?za-Z<0~yyUe=;<(}yB2L(fn^9s6Oq(+46j96?G3mT|JU+A>Z6nS;8I^h5mCn3I
zAoryHL-*er<&2hujD2Toj+zt{b(23+vZNHdV(!=uvUnxKx20PdV#~J4mQCBK6CUY?
z2;bG*74!<hH{$?-?K9E~tairPe!SyS1ZF1SeSaT9V+A396ECRkyZ|_u-}~HWN5mrJ
z6gl$h+lqsw_5*$yfkyFR%8%T8a=&PKn!`{u4hu$kY{tFvl{;~ccEL@LQ6c1pxD}3z
zB(^n|C@5`(Yz33CG&6bPMcS4xlK85I@gfoB)|qe5np-An6EBZ?%O8!XxuNF<$Q6xp
z(HLQv(u0)>#q$)p$u8oQi_{_kg1oOa&6%|`pBM-!^<egs&#7*_5FaKkpy66wRJtV7
zzcnUUnL%qZX_3zjAOr~_stSsN0xv9be@4yP`{{9%Mn#8C1vRDJA_Vp;M@3UoGYMc5
zPJZTyW><&_&6?W6H&4JOI|)s;*Y$FN!MP=Vv$hq14RR7BSM57)Q%lV<qXAA;^r`0W
zWdr4NUFlwYwUNQMc|A|#@#V`#s!~S-;<=V{H(Lnacs8%h&4^2de6jW>ld4=Kg(bbt
z4~<dfj+woU-|AKJ4-Ij3Xtn2Q=(@oKhTqK8!I08!>1@rk{2?_guxHvHql(&xND`{)
z`dTILKE7P>nb2(&3Uo#2CO<PW`SL;n+>R=p4BT~>p&!L;-{X9GiafmimHX-ze~MUd
zJMXr^E%6@5u`ByD7q7nZBHttQ0%LZMv-=>*bXO9VeZ*7uc11pLKQ{01Mx-PA<n+rW
zc;z-uC-4xA)%JuU>==I+1Ws3oVrzT}%06;Ad!-jh6^$|dad<=9lYNr;Xm_UC^ls2u
z_jXD39%uN^aLv;m*J^p>JMubOElx4XpME;Vk|8|l&-unnF8~0WR1yb!xz6j6yic9>
zmt~EKgSf-q4#xtb7B$<ltpQbcBvgd~3TpE*zxoA9Y9bW!6l)t{vLocsHma@!cBsAQ
zYm6~&cIa+|T|EYkYF%(_@m0|N-Q6)|5QjtvvKrgxF$LOIP5u>hW=4wUtt$?c6rJ<-
zTyhTFL7Ym|`-~`uSQ!SJjzh@GhH?xhi#^l|9A|i%i(XV*Lq;Jpxsz!*3*&{;TRhD!
z+7L0Lc=fQg)sBEX@b)-^z_~c0pp|NMXVk&oGfxS<|LT#Mcwrj)?-~9rv`8awf^V7@
zp|b`qD$Le|%Y9+wnfK>_3fdkOLTZ_ZWD?`2-SpsPdjt|D86ml7H5Y>XVV%5(&j;TT
zWRyO`BF|4w{DpJl(|q_m)8o^V@*NfN2LLYBi{*-hE<B@Fp3_JpEXK5sre+p4Mh<O8
zrFZOCOzf+Y1T9RfL9$hHrZR)Yee253<lu%$j~l>lpq-@sBGpbz<m^$hiH~#cj|q$O
zbJgWp9~O)}V!};(FDH+XZ0&Kb7JSk@!k_Pt?uL~WnzS?;S3faAQi>DQ<*wo4R*F)u
z7J)HDg!9e2vW{D3Rklif7qSbjiR9)FbqhA>60ijb($Lj1hE~O<5)sImFIP>*KveqQ
z?*myxNh>#i`;Z>pxt+3^{TyGhV^c`&QzBRQxVJOw1cv%l;p)g;;iW;g3@qS}Nxe$>
zw5=<;I9H}nZw8g4>-dEy*6@dsunM;zU>-V^)(m(`rijt4R;aln!W<W`MLxGZmq}p)
z7|%lNUMPauH4FU*Jd~v!vzbMB+0a4WEgwP<F6za<GGpM{cH;~eglSN&_|3HH_K>Jg
z^cu1;7X0!rx@vU>S#HEcE7|CYYlA6;-i({IaeS~46YJN5KN!n<AN?jYTF+Cl_-wCE
zw1<EwXWzc>i81F(wd=s&zPh*7L36ROocJaD%&4_#bc3-DXi&00k?OA&)7PQ&d8je!
z(e)_sqbO88T<NCum-zc)U#e?GTXE&skWb4oufC&a2jwb%tmEm8TjVJ!t?dZ^EEj%M
zs;C~K1;hWsQ(B!wOD)DXCD#m=JOk(bn(%m*BbS<)nU{asD$us24dG|16H;~q`<1jY
z8%2GcQEPq4ZK<bfR1MIa)yAq^qcr@5m<D9U%gPe@XUh!=OqnX3%89~+!aL%P#_rZ6
zE=+Dpoc<)q4uT9~e1$_PnS9aOKs70|z!&XMn%?T`N><#|&y_lS!O_K{6XZjGI1|%S
z((>}-#y8T?h6U@=_9Ai);A4nK=x$f?XLua<)m%=?L+%J*nHaxriMFv~Mxi*JxSyhZ
zh~WJ75coXOqfr-ZIgSk*(g-9MJgIA^6g)9#bh(V*i#BK&w9li1j;%xLK9;77W5Vqg
zNO%mx07Z7#>HPhhLLzQgL}zHZv1kYng)(8l<9deAHGR-T?P9s&n^ny7Uu`2m@-kN2
zdX2LPLLHCjjdi(SSAdG?ZT^xoo2e!XgJBco^G6`-990Mvk_PS1Oz<n3lV3fxL>A7}
z>9?3)qwnOg<5Covr>c2WN1TLEDS$8=0?U;KMuZKF<4G2{xw+7#&onIJ4d66Z-b233
zRmPPRbTMm^$amOG2tl&XGmmj4RY+W&Qy1!~Zqy1?+FM+P&vAC~q?26M7ZT?m|8h$Z
z-lQ>xN~DCjMB?dGYpw^an|R;j^jc1^3Kvl764;o^PT3PLH2$+X($RRC+<=IMKj8*u
z0}Cf4JU_^p=hhRjkh{j;s{cV6U~ApuybR;=B7{l8jtPc!wE#u#1rkKlD_wJ5W6VQn
z=F(zR_H<Ps4abuBWtAPJ>xXzkxL#&e9<{SibHs%ML#8L&Xh5@oTat}?m}R_r_cidK
zF0H)36kOKiKN;fK&v6?`gk+1PY}9qM6R_$llGKD>b1rZ`*-ln(>r2sCAb$9<B;av3
zzg029Fht`EFFvcZvST~os~tHMv2&E4kB%3LQ0-Gu@50*mHROx>n;t1@G7bwEXdeq%
zYIo_WD5B_TP$jV?6NDi`apBa5t{=NYUK1g>0X(splw2)hsgFIK*!s$YX2-R*x0-0W
zVE_8L`d<y;`v0=r!>EdSQ;gIyHY5nweNCZ=*EG4Mt4=u?%bME!<q5~Xzsf_s*0u@F
z49k91QmuM6&|3%RWKbe4yQ-a8VpC$eMZFSKDOq08RyOaqRqWu?i%Cb{_zL6YU}!D}
z$DRcJ?;GmAfn~r56O+Tx*qRWjq=vWpBs}R_gI7(6ZK??9z|h=daLCGL@W|<`d&$1(
zHU_(BptP-wz4ng<heRX({XPC)ID1gqqbJN}AZ9afX=|=Z@N+TQenC0e%e>hQFU40i
zzbl;wgIE@(ty@ELEoo;$Bqd9*OvmWQB)t}5)FdeCnua2+(|VOa*kfPiD}pi`d22aP
z7*~Nc@z@)lE<S#vjlH2KnniH3(o~0ctfVI73aZ@fF-wG2^Dz?=8PPsSlI6(GG}465
z*>YX-&Mq=RElcayY!hmy7n&(&>A`|{ddN8216hbFM_x0at%NDplsmO|n0fjA|7OcY
z_eV=V|8LcO8q`=^+{oShb;(6Km#JMAp{2`C7>|euE5&;OX5?mfh%d3o79pK0dsMB&
zG94$Wa=oLyt5Tv3b@(2esh%C5ZU0hCFIGn#Lf0Xqn=7Ult>;s+VV783Up3!L&_$k-
z9L-a%lkdg<5fRH%rdaAq(U)qhZGdSqn&il?vYME~rnxBl?fTM|)2X{8Lfop-N@u5b
zji1~u!p5)IGJ;*EBh@-#Bq3LFtoZKd8P?8k9C#|W18dsXn2j(o;})2R*QdtH9dReM
zVE|+3-;`Vz1Pv{mjFWE`kDVrxsaIIM89XlGro&`Rvpy+GD?^&X)Q;c!q?KROPs}#e
z$pgknWLy%@CC_;+KhzD<G1yK^J(g(<T61Hb$x8FYIIs4W`mIXN46JCt92_ZIe45$<
z6;kk>6V@~;G&F{AYKj1>KU;3z`-NLo@fi6xyNx}tn5hutRl;Fk(Qqoh`}olRovGu|
zBd6p*#1@XoU_0vMGwtRI`tm2?-@`M%-%l<*ui8F6%*Vns>W<+2*HT2Wl6MsuI4l2p
z-3|61>i9L^nu<P{c>RBMOO5~Wi7y;u{n=!3qGF<BUlwiSfd6l*@`zuyeR>ohJa4&!
zLE{*`CeF$aeBHf^qoaIRKVkWo3b!KoR>|A5O&UiX-L})Nbn_FxzxI`&lJ{u6`shEn
z<uGITW`!UVdY{M7+aE`Kw-k$+<(BofA=}fF=;L#ktz0MQ<hQIf{4X<lQhN3DOYpFh
zLpsVHW<QCtZ~5ETIL@hVyHzA5-2O73kZIF>P>SGcp=vYU%t!+iYqxOnlIPC&x8d!x
zZE)}EYw4}7J4eYZ<M+uL@9JHKqJoXB;!d)M_ctE69UdICFSIjFEST!&hep@jWM;o0
zudxyMlYbTP`LVcXa>XAU`GP?c$<NQ0c_wh=|N6csY0Emn9O8&e|4T<dRm?JSKlwj}
z^lu&h@?ZM(Ukm)3^#1iioAG~;)PI%af0p-u5zc?U{BPn8{`Y}`?*lwIjd?JU?0N8+
z!YSG8KaHYiBBZA3K{!sA&j%~|-sAV3?Q;7@8^1Hvd>Erpq3^k~UyS=#HMB;MS7QWc
zZAcfbJa*OW*V6i75aEw`D7qj;5X~(-p45K!!HeZ=GR={_%g1(QIv^)$?_{Cg@ozIg
zyJ=k(`{_tz<z!|hp~fDs%#Ng#aKw8fQrQP6%~oVfJqFnxL2l~$S*C{MI+{5S(neNk
zw8z!Tv<e30(Hd^`reop7+!K~ZPm9c=M*M<)w`!FfkZr7BIqZ!r|7-{v)EpTSZ=YFZ
z*_@|4mqZxgTh<k#j$#~Iz9}6uPd!*H)U%SXwf!D5sK|t6!(x|y5x0?0z#^NF3{(sa
zFAja+hF$s|Q1EUh^Z5U+Th3$}yb6Bx<8*85j9}$L+Xr2dwsmy&_*dn>l#*<{TYUI^
zv1y2k@F*@YGu`wa$0P$AxHM_^aC>ys_6|Mobcbs$b+fE^x|}67kIuQr@nbi=Tp&8M
zj}f@0@#f_g^_Sr`5qI8vT``De_<|XeyT_p^li)tN$9V>bs7St>Us^(Ydqm#j%u-@O
z>|2EHaqj+^T)W0@q%y_QHLeGrVDtZiJG8a}L#(f>+yw7&+{~kIY0R1bx^>@8&*t52
z#bQh5<JS?5{~oW?wp9I5+u;6&3a*z{%kXz4Q?qAL(Rqu~Le_v0&-Qz`8nBc5pJDlU
zt<f%d4@lqwu_`Cub$)3NaqQ>mjhM~NPehb<G)o9iF;kMuEmCF_35k8O%oyX~_6mV8
z4F~gP;d;|$NmmM7UEb#&C*w^12+HL6wr5(hSomPro}(Hqv@6-7*i&i**J0PwH6M!S
zAbd1ej~|D~LN^tCkp5E`Eqm?}Nr`1i9|!Z$lmWTk#QoE0g#Z@@zC_knscFpTQR!?J
z+BtU25<ed3SI1h}aLK8|qHRI?Q&{Q)(i0;4rjgM*V@VpWy^)pzVRfL8*H|HQS3-Q+
z#)N!K!4aXb4yPq$o=%CpAHLD|rcMjFua&K9Z>?)9e5Aj|=3oKJqF^*adYb9QGw-<G
zIvQR$tTJ0oL@SpYdu@H4<E<{&v!BYW-I+>Lqq!`aL_dP@kEKkhNRU)O0l>HJ`H6}M
z2y3%;s8vJlSBS5pP}C&IS!2~M2fowf|4L!@L>)B%@`@`(42_J8nq9(E9c;9Wkk-2?
zkS~(#q|7>#ReDxM?9>2rMNHg?`O@#mBHLZkpTVS7)zTfn<Jp-u69bw$@5q3XJYxj6
zW;8xCtuM0n8IN2u6E^8T<b_znAOq*7AT(=q;T3(glmDMB(sIs29?=k^mWs|s^0NZ<
z5-Zj#E?S+a{ZXIgNtB}P0yGb35z7(ti>}(xP0_{~-?Q0=a-xTDlDFz<P0~N(e}a!N
z!Ifo##I`a7liOaNe`_qEP;Gg>tv{S(L-}Ji0z2Xuuw{V>;)w@SKDk}ddSkh~Pc*q1
zI*WQj{#ktKn)jlHX0+IctVj`w!<UH`%DX?R3`mF6hTE&;lKo_>b6%sHK0r83-v*Fn
zpET`DVKi1NZXS0#pEMs~vv%AzvHmUc2E<*=@AV~yv6VeiL+qTuy5cW0RrfyJ<D9?4
zPBxphdmI!?NGTT90IGeDbCTPBWvzX#dyCC?y`y@4OC?j*M!-ddtn*qS)w6sPke6m8
zAJ5-$ZM|$w%=ev4St}dJEN5AhU!&xh7FwT_CuJMJbwZMRtvbdi>(fM69$MM*hQM_7
zd+(Z&9p9<!+BPz4(-`5VGFJ?{U>Mo6`^2>#%e9>n=RazUhBK-_yaB~YSMamwXLHX?
zhG=h<r9p*Am1mT5a2Biikv)fM1A;+N@q%;Du+rE$WzBAGM098S^K-Gb=@eU%fzTV)
z$h1>AkOJUMg%4m+TffXbncFINUsg&(<4t&8<8I+VBDZ$j^#JSS*Ez66qUQc`pSIv!
zW;~gopj2_$vxck|d`sETmLOSKxhYv$qo)O}wc}*-_3B9~fqUMXMh&HSVDKT9s$i`#
zOC$FX)7lYuUPydrB*WZ^2+5|YzfhhvHL12qeK+d(N^ota+ywcRu=-tL2f33FO4?pk
zx^p07izF}deLe4UhCGts9GeOEjqev_)5EWf?s4!J4b{svcIxN0NM-kJ#8rNgQRv$$
zM#-A5aN`QobWK$4yU$;yWfH$okq?~#%p~VD&H<tn+(+kev51jGlszuosPr6^KF>v@
z>No>tRI$sa(58<J=u8vtcI&E`4(8sWm2oBwu92|y=o@O7i@L7<F1T+fCHzuVemZ$g
zz1FE`ea(Kr($H-<H(`X*OuLO|H+0dmc?wn@>oV2%m)OdWUlU`60bwR4BHgoA!&%X1
ztwdVut(0`OK0WkqMz3)l)iHq<H>MHP^T_h)P&6?bWl!#X<Mk)tiTDIj__jd!{A4ya
zC#+B}EXr=PNUQc`N4R}$T)LVeM?oyx-=?PSLBgKD8msIO`bO3%ExvOJFK_+wP9H;9
z^aJg~bpK|In%0T4lEsE<j~{2Vjb;H$q{DT?Iww%+Qu)8p_Lfm?ZQ;9bcin)N;#$18
zL-7_55F`*hP%MxHhXjY+jTLt<P#hvyaBb1z8e9q#_hJPK`(*#m7<Y_&Ki+Z9$9%|G
zYpl7}eCK@T^Zp*1E+8tW{pp^lK}fI8D&((K7@a@#J-jOe;HUr!WC-OZp^H2Xzjt@_
zN?cKvqi;E_j!n;yfr4BkU2X2J@4b}Q-&Zebs?y6LhKS*>fN5pazCED~uWP(tHqXPK
zU|zy&-6H9cP*d*!9h+;`)~;ur-dw382c`#`1`Idzl*K5yv!U6<CUIJ?J9o8%`M%G;
zS84(B<D0f$F~~2yoc(fWX9*3m+vcjXD5J^}jC3n(dBkS4R635J`ZBYpQr$l#T+WW2
z{J^ziu@n|bq@+gR$c5c3;@yVjA|rp3#9ltNS;#<n=!t*8fIap(%0MEcj8)>C)kMuy
zC(mj6){k@`=nV0={VXv=Hty*3-#c#YwbM7>VuDUh7;gDRgX#=p+xKfmE^os0{aMI1
z^*4z0xyP=Wx6*qolY*W4H#5G(s@U3aef!Z`be4`tYC)K8-n0?xo=hLJ<xJ6^UB?_E
z&xGU~u}{1{{V3N(@B#vBCIalZmrtQ#fyJNxy)(D=ENDTGIDcH?AJ5Bkk8b={H`z+!
z4w1cKqO#!;+oUi>^Yywo$`|6!uc^=hAb%T$Qz=qv<HO||?4(an$J*=x1x>82)NBlm
z#*ny#>fikjIkFuk-QZgNn#6upo>xY0JI|l@DCng34J(5~=1x{Bezn@OTcJzj)Z<({
zk{6!)YshH9U;ZdOipsT3yPsk&V9MU`GO#3$<7*@JUDlOJl^OY|TVaLW+u^8ITd`bX
z?0a}YCnjgevNI-VzlJQ-VSa(g15mXYK?t7IiI&_}dez^CCCtXzSYGIF5nW3(lgzgQ
zYu`MxZl{0O-~Jl<_s-9}XtIZi1Nd^1oHXT~vl7RtjS`XH35=pVr*sevkWIav6@30f
zAu#f!tDkBuQc1bBtE{_>h`o!xhW$oQW0IF)kp`{Oc*FQK2&I(+=B`G86z>CT>E<ZY
zt*EE>9UP4s=_WnSkSTPN;A)oLcKca?@ozf2<D#N9IpOEmdSf3tKkXYN6{9)va~NAN
zSh*E%!P%hz`Xl;>ai%>{jBNGzME?pB`R|=&c@|<=E!y%1<TvnZhu9<ZrtjZ7M>Wi+
zB(f4e=TGUq5C<bPsW;>aijikzf{9Eqs>PkfYN+=czQ+#BZs}T;b4wC)PIx3hJd<9L
z#H<uRJ>TAWUZhuc{pwn)r%Y#fl1(J9uv9BbPMpb^mxVLJ!A0FELYSN_v&&=Qr!|kH
z$*>s&Y=oY({`l}=a;4~p_n{9kcf-~l-#CU5yiOyX1WJj7v1(R`M>VUnsNl-8fA2`+
z8QYIA4}#VL2N`Mqh$$?7EEu^SiBs4Dx}NtCi`(2Bo><|=6*d7%J2J1GGD&u;^?9e9
z2%1smwFe7K&A6<@C@CIup6-XOUKW$+ZUJlZLT)!W#U+SUUBfViyChY)Fk{DpWxons
zIvRlYY&0(z>M0?V&Md`%i$(^}64t5(+6B7y?9d-ldf#WpCxon2L(q|*&c(shR#&F{
zX~B{Od{W$^BvnNlFny7e7nWJ9-(o~1N%<sMP51pJrOJmyz(lk}WDXgp{13v3+v>)k
z);hRy5KZ|6R3^az-)nW`>E@&`(WzK!aWt!hPEmR(hqABhe^G39igc40m(~7(GhjB~
ziOQZMes>YEv2Mf6ML*gR{15xAYg1E$A9}o0*DHM+so^f`AT)8}FeCc%YoZ39!%^b%
z$*yiXoH+Juf4iD|bY75e+0dYd+N^L`4efA3*<G^R<&g(a%^*IYCovvri&+MV{+7N7
zMjT0WW?DG3xX^oyp~Cl1er~N2;(XBTpr;m<w8EBN6r9q8ne<$&-bT4QXW}HDIekvU
zGJrcXuqKq(|7261!w2##UDu7!V<1<w=>BG|dxg#kZLTe|br6D2<`XZ|0L@T?5`#LR
z+Q*^No)@8>l$q|4NFl2hd@TDo<;qC{E+*MmR}gOVtpuQI5Xa9CvX$u5y0sBYDB@?*
z?i;YVGMy~b%f|}(_)A&(M3H>T98N6zF2}ys8#c`j4wjzoiCi9gdBm_DMJADD<3b<w
z;#_t^<^{i0+qUWSFx86qnv;RAIDYN?W+>vVooNAZ3&P$o9iv~|6I9^FO{DV?65d3I
zE6Rm$rSOUFH|Fn2ntVK{^;d{Gjv>eJW2dicHIhP)%|lK}P&~Ca(UUF4zC&O{t#f2)
zsh^()UYD!J$RZ~-2*xj^T@2LhsxEGM5*aEj#x?5!bZ|jn8;kZzVvmB_v!zm*uP+$d
z&qBNPZh|71iCYfhi|ny=O?1DfzgruPyEa)_{<X_|vaa409z0Xe=Ja9?qSHp9ZCw#-
zjwo#?STtP-jg_RQzKVDz>ZP*a?sUd?CfB*h^?b6Bbn})7{jh{u+5tYxM=mhWfaduo
zhj5Mo%;T$8T~jVw%`jKA`Vu$NS$%Dajs)y1J$_$maL0Gz*-I4vta0q@UYhI%<J~_Z
zZELRYkKgPIm!Q<`ij6{Qv+DU$=nWX%ZaHSc{=Ea%><?Kiqx+#5<$~l-tYQ^|JzeH-
zmJX@LS%~Of`;~0*K088>Vi0{5#DehCg)VwCFw|NN3FQ|zorai~g|2poW!6pEq_PF6
zs)3LE5|&2a=Gjh-3Yt5Pb2^~AQXM_bY;(`v<IFFL(nOPx;~%D4cn3g??-K)Mgtt?m
zqy9<K!SPoysPqeRdEp&{t;AP1`YL$|#|#v(*$A?4!+0i>T~f9+AV8|G&7xrcCAUtN
zy{Cyu>F9LQ<m4cy>UnCat@0FHg26!N@awk(HYLm7!Ys8*4_EH30_JRVvp=<u#Lwm4
z;-L5by<-Y}G`H+haY1_h`rkVXbVQWd*mEUJCh><8gjf{j5kaRhB~b&(mhS#<9Q`AF
zsLq^i-e&L+MCB%Qn$p>vr?#NV8)&^79gF1b7j|W|IB7A}gquv=n8Wsv*jO9gU-5tc
z;eJ>dk$p44Jo%G{+VE%YHF@>ZTimwY4Ls6q)NIyI#c*Eow&k=`+6>vOA^Ug~MNA^6
zbhlfsTzKQJDx>Xh5uU!+ws4WM8!oZ*)=E6+#nB>q;P=N5dR0#+%Ph_f5I$4s<^4_k
zZ;(diPZ>;=4lBDA!K&FzvPL<u5_8R17M_!)u&A08K;EdEoc<jfX%ua@U_yA<F~d0>
zR%4WR4DRT!;m@N2m(Aj?eLPXQ(iOq~)DX_Dj<s3p!I-Y=^aWew5%(nDHK(`}Hh+@y
zmK5gcXexhBc6^}P<2i`2#(MLIY_eHw#m`Nl^Lg@dF?Q3naj>?rJ%=uH=~>NC!_d4C
zV>fOR?H<1OMY^_A=EWmPtdyib8a7t%jhLLTrM>)`t@p*aX4i17S^uustjIz-LH^4p
zW+VDNr@h?0=8r->T<2Xqlub+G-7m~9{mH^gc!Ubg#QC36_=ZZP-6N@HbS|B<ZyD&H
zbbC&)PE#@YRLSz*@zLObv`qz~aEN7XuI5iz3@ov_rty)<naP$eu^hJ!+tQqn<rrza
zGro$8xU0cX(=}CKv56>PWs}v0N_ab2q1V%dj7PF#N(t(9>mUD_$qt;ZjC%z3^c!!Z
zt>5b(GNBUxMi&zOswOkkt(6hY&=Wh-xZS-LW%o%xzL1YdAySc_u;?ZCH5#jXFe+Fr
zRtPU<?@iYsov6xrtD&~}HjqQ3nT>t#JupfaC5LU36JSuAd@aZ8nUTBm@|f+byA&y(
z(nrrUYBD39nRfLtdO<b-kxx-H-nwFEPkz79xQvUj5){W1oo-l2&DuYxbc0fD<twq3
zrRpyHMH@qs7!^R<)Jj!&mT3O1<(UTMPraII!Cjq7-{xZ^B598IZW|G)MfIhYIw~vH
zGo)PSzMuc-&A)f-IdrxlCX&T(vdr8~!F0<Heo43gzkLKubN^o-0ppn3wd*$T?P7$&
zw(=_%t7G;3m&bdZMDzPShW07HR|-EMjy`SC#LtOMZTYuLsKVy<y<eF2Yj!Nr{*I2j
zh0#nYeIxcm<1C0v)9HIp6n?Dl$XDJ9Gw-#(U;Zr_qp&IcWV!;{ev`Udaov>h@13ln
zEC;gaO=yf>3-^L~5!)k?Z)wJ_`XP&`)gmJ*%X*f?p9U}0YMdfl6I8QUiI9#Lk+p_5
zXr)3Hz*HV<n!c88LzCZpt~GI$tVR8H&#ef@rKQP7$4~vJ@5c;M{XCR!dxv1E(JptF
z!e}<u^K%YEa-4K&U23T0n<_9aBOEuC>u^;lR%A8ziWw&^4vNwz8rnC{LDk3XKU#;h
z##oiJKYa(d>@t#a;s}vY`K7KXC8*(mKVQpnaTX~2m4Fod+ZE%y6KJ$Eh1ix{zF3R%
z;`en2c1A{~1ige8X-R1}xx#MWo~a_XKawkLZ>{S*Gi=4&lvf2jy9?~^jVog0S%ZgC
zVI;O<O8Qnp$_P9Am_f8;or$S4O_E7lUo;>h=UZYW3%1KacY2Q4wCm|p5b3yR%Ai{J
zF??GWTC|eS=YZe6YieSbw}=F>`}}UrsV*;+Lw-7&wE1ks^+eRFei_zEr_H@?oe^we
zFtCrR0Ogxb2RP?C3Gns6M43v!l%&o6X6Z4$kY^>@P@q7G=bRxUXVgI#_vsE1Jm5JO
zV%++>CzJlq+t#GbJS!4G;99oOHfJ^muYys&x)#BV?%w8WrOagY)2jzM&qiS+nQ8vq
zjT$QUFSSRmle=05Ckpz9G*zQ^=qxxVpz$lxMPV!Bt^`ss<9NH`0;M#=Mix1%E<%jd
z-9TqMt5!x*_Q9IuH*y$eQaTZK$=cR8b0-OoJ3k&8lzPv9#^p^HTO+7Rg-i{_=^xlM
z`pAX3?jp@lCk@OEumS8s@|*~Ce7kP(ed#S%n3$I|H!oYx7`JfRVTE;t1MuE9o0k6X
zXYbnvCmCt_^o4--FLZ3a<MQy7z41w9qh*!bN&E>#Kpnm(jc6=B|NM6~vgs%R1lftA
zY-1~8ppPm|hm+V)u9Y?I8K0TJ-Wr5q=O$+9_?|TEGSPLXhd*H^WoRC2O5W167u?m#
zY&P8+a3gh`=&O#;*P}9W?Zb;)*I~Z{-{$fdhO%FO_xC!lfY;AqwAFtafX6p@k!iuZ
zRZE0_?}SXVB}H=2LYUA3aYse%OZ#%NAiYm*8mU4uKZq{cNeSYn=m>vL;VLfj-d5Zi
z_3d|0;!517CivewxmIiwOhcAusYEHfsr27Fe>A>$t-+_JwchnZd95~dex-!Q9|ubq
z7d?&R5rYpB1F*v3;$AulYBk)I{8zGbj`PJQ6@VBXj~BUU{nAr9KZC0j19a7n8mWSL
zYwS&w>r#emAm@a0#<9Qu%RDR_lkaomYEh9y@Y8{<>d_L6hdw3u)+<+T3Leilm3FIY
z;oPD!dKvzM*xbgh6!l}-34p0?PN%pO1AiDd7QG6OVv)i{ey{p4z&^3~)R6wQqQB(e
zP>+ZCsdbm_8$NZHh&gtk>REnD^5J8Qh=bFh^Y9Y$I(#-=WsT~Vk+GfWyfAN-&6K4z
zxGsb|xflfJzkJ(K62HQp?l}%e%1?cs;97~w@0Ch_^v*j?lVJR`HCyvlkp{;@H<Kg?
z9g`$Zi;u#07HRb%#3l>*X&_(NKah!bqBU^9sQcGjcL=e==dMzv;{rRRw7W<HZ=xTC
zMwd2I``RGbyj5K7{=L&7QCsPI)GLD#p@yV;2WvLC<zMjR3q;Me^Kwa&z%d7z2qW<|
zFo)TAF`#ypVm=dl(RfPJJ0X@mGl<HXdqg+knef|^P}TZMZ8CsbZ?}D{tO<I*bN5Y6
z<O6BaEaJW(aI2aE&5(_D{h)|ZHSsmblG`YtK9zH!;`;l;j=3*Wj$N5PI3x4VU`e}D
zoZ;tvxea&GO04&ojP{&0oa2YoL8^007kKjTg)s#U7hQ!B*h2jOuk+6T%YP!3Zd?3+
zA~K2B;!@fZYI3{(KUdA3n6StbPds0GOhIEimy6$Iz^D1y?fm{0g?Z*ME);-$7Y8n~
zw(+uYG<bSaSP!%YOCU=%<0mR(9+uDs+z(ImbxC6>mNMQ)u+BpdEhHYzN`P%gD5}-|
zmQ1C$-XF19q=qa<PuFpk{ZW9n5!!*_o_GzrNKI{Xxb){V>+8>Ow+jF}-t&8ua=hHt
zp^@qINYT>stmwc0k3JJ|iO%K68of>U!}oRNj#ZGUeV#^sDHz0JvCFu=)he=}Ax#fR
zwHAS(jB=I&I~SOx=oGgrS(gufceDMTO9cjLp<wQ}6*dYTBQLCQFO0-Wv?fRd>V;Ll
z9t-!cG=8QQR&yLo&gQOdiq16Dr!W|&=z8kGnay`r)SBd>YxT}oqgtlm+b%&=XWWC>
z>p|Arskd$-Zs^3pm0@7pU(II{njB1N7trw#&q!-^{we8k<x1m-$|tO<!UtK^;Wow<
zDDyH+hS$o0Tt}_Y`DBBjOro`G;19)uQCy}uy4z5ZQuucQxszhpuRhSa%7+@l=A2{l
zYM|{0EQe9=Oz*P~@VlD)9yHc;^0WnoKZ8)orbX^~ce7ZNMClS_9zPQmeZr<n43g?J
zQUFg?9m0z=Oyhv1V0#(m<l?@;5iKkyqD!5e$b1ll4%9=gqEQ&<H1qN>1)<yy((l@0
z*1(id2{{=dQGx5LA0M5}s+c-THJz@CRE8r(HG$Gcr9hBDn;jKJTH8uW?q3TruWuQd
z#+pRGz0GA_b)aEs2zmnI+(mk+TZd+S2GDXMdQ4z019_-pWKbBh!;n+vci-W-Vl1UO
zT3zp_CD^1tdQb{r8^bc&v2R_D5nrBf2G`H+gX{0L8xQ69zxnFO-{KbhVsFo-WERCQ
zOo{kM-&oT%Z-b>#f?tmJ{bZW<KWb~QXaqMh#uD-y_y|oYNM2Q6nJ!b4UT+sQV?ecb
zmp9sH@=6~rYO&6Wf1VmAQ6?--OfxW<)l|keOWxuP=VeN&5Vd7|G5}%(ks%L2e}x?u
z`jM<QGTJygEHnNr3)?i}Y@}6QFRk{G-cLk0eZi>4)#mvRn?*(98UkIV7|!-ox6Z;n
zMiFi90e}@QoWV95wF-t@TP*GebG(k}aLRX^x>hT4>AaYGrevznWvbcZDufUkrU_-C
zE8T%|m|ULPX<%4XQ!jjU27VOUnao93%gGq-bQrbELBL0Ra>Wdc5xe9D(l%&>ocz|%
z-$~?h(ez)_V#IE|v+=b-=)bQ6Q<||^2VP`Ce(x*w`O@{=)+D&+W|p<9`R%vdb|%=Y
z7(2o<-B-#daC0yYWa!v-_qYK%f!;_*_rnm6h(osL==jxg-VOQc;#=6=r~>on{T|2S
zzLX%bVE9CP+IP&G)3Uk2O()4&wLJoRAsd!`Z?S?J#AG_HBhi%}nH~D(1l*7A1<<N#
z)`<*s>=#%Nz0oyt#=e77w{@c8?I4qS^L-oitowX+qW{EdsLW&6)a*EM-r+J$EO=EV
zC1lsFNGY$=V0%|gPk?UnL`)4(#`)=)V_a0ZShj)Ob#a^(P-ct%t~QLC>01Y1XQJro
zoTr!GF*(umV~r6>tZqtdTM@)|NB-Prp5rX=sNR!RdJeOHV!czt!_PwC$c}m{3uPZn
zvw8-CWEX07?dRGNbjC9q%hhZ>glppN{K^RHj-2`CY=aaTlL<?4q)1N(SU4>s5hU8a
zulxwprLJCvk^V_~-iN>=%>uiQC9x|bMLM!xrcvWr*N1T#fPjEECuiMG(7a3A+=dBj
zjm`{9pFK{6@~O|kWn}mp7`CYolbf`PHPZMtS%Wn+v}2!wJ}+ATa(?V*y!~?U7H2Us
z;o#g{k!k|<=jHW7xO;qNP%2vqdp3SNhe3k)>0<UpWEZ%7oLECOF}kJIORRZD2)yKk
z^lI^!nC}j8O{?`kS%xF(11d^fta0Mk(myU4JPI9*!O7BLEJ>Z@yUG<fkjn|+D!{Ph
zFJ{35&|LNHzS?J_o_%UVRz)cKJWYqW!&I+hX1s_hMFsgD`lC-J*B1!5ZjtYxfO(VQ
zuZ<RO3I};5W{azr`EJ`YaV3q&zSb0%<=?X|Ye=5O#`09`mDy-zJ~qu9CkH?q#*g1D
zX1$MTaCJB_w4F?|T+;CTFu&}^)I2JPACA8%I2N~j;;@)(oZ`#zl#?}EV;u+s^=XIe
z#urOjsR3X);|eX4cDQGH_2K0=E@O5#pCQp(4V<bL1!g9NX_;AoHx|2s^X>J<!ZyWt
z@X_AzFmYW6o|fUj%X_7eRAp!kBm0o1>1^t~Jhc`>dEE^Nr;&~_&M;yub8%+J`~_((
z#LxR@X<~p3pLoe!Q<48)8`1&i?<OAMov4B>#trzyNtIp_*v7drisk%<BEy;*d_R@r
zNNm@e-vmS%fv<*}+kr{2%LXL?jSKw{uwQpkwyN-p2vni9|GL>vvP&H}=>@;BM+m@d
z2V`;xS^O9uyFMq8evC~F8U41<;Whz{Y$>Gy>vAY?eq_q!{E{ZOST?GcPgmA0_PhEg
z2F$FUI^a6}TYltr&{yX?CA(|#hEC~AS9*tgO<#ZgX02V3@M}VW_WR@udv@%I<C%YS
z;8-0rpUrR=mv*-H-0S_<qe2@aDmPj^Vw6mKlcrzCE-{oXA5j7$1M^d6VzN(WZ9XPv
z;~LF|8<)4g_H(`kkB~f_k}2ptLKBhe7F=#A)cF?=ah3`M?J11Q@99YD9fvL7Ac8OR
z?)g9GhT|1@X)BI#M_D<tD|qeC{6HmR(^?{(B{wP1?$9VX^};XEk7Xu^8R5O$IeS|!
z%UJkaA+}xkPs9Y~(z3~Kx^=*>{l@Q5F5Ldc78rCjF3>h)IZ<yWAmAj?vTnqIe8!p;
zYhdnH*xyBioSj<zU}aFvwKiV?TAPen_JW^M?#!onE9X)vC;#E_L_1tO2^!lYBEzzZ
zDqt`z9hm}kZ+G_RY@(Vem|H(nQ_VzvYl>471r%IE>wMo(e?tb`EF(UtDbQxzL#N!C
z&cq(VvsR)<pwf~zTjU5>5x2{JQiYtZ2}jpQ4?*fjKJGQ@%w?r6Y?)0vjeI<1aK?bJ
z=s`LXx4z$CJ@H!I()VCRY5EJ&0tf=Bmk5{7o?@8ZHv<K}_ybh%lYsHgR#${?Wom>g
zhRa33@L*GGT{H$9QeB+{;HIztR6V|!tUap_&McpD*=cfAz&y5zWRgt%oRx>2>qv<0
z^S-?hJUG~ARobU7UG&cE!_+!wtlMl}V&{dEwK6SG`{6M)pFIi}s8uSXnvw?@c>J^6
zqnmIV(P>5p>+Yd!LA@~0*j2Ju#-c`$;qtPgZ2h~Oo-4ca_)cUrxWL=knu90Y9L1bt
zx{ieDKhbpz$p(O)GrMQ(zrL<HCEG>L-$0wigMiZ~Rj4GPo+sI<B~y#?Nw3@^q#iAD
z$gN~Y_C)<fIc!{c2B~=P@15d-l+?>y)Bc|EKW+BZWbohF|J0pJfSpV^*7U0E)w4sv
zHa?VB5W3|GO*{Ejt#ak937y6=gg~O1yX0q!Kc13fZx1U@kH7E*?QKSL|7KPnEa)sZ
zIQp5xoua{^c>+)=(lJ1vmOdl5vRkj^lskx!jteersvLiHHeG`cBF_0&^5Vrk=rJlg
zyVu&?@ezl*a!Xt-iXZufu0<gi_V(gWhufT|H0^Rb@G^C4oIa{u#YY7++v)pS&5L#0
zULZEbj-n(kb*y@+O33E#T30=&OWB-lzG3`OHn~|G-$ftO#Ar|z>1VkwNBYMHaUWJB
zM_Hyywgixf7yKd~%iE*2YlHck?9tsnc68oNM$82kO#=-GBfZBKDMIBWXb#ruok(py
zENu<IM1G0dhIY(&WO@#NX7TD^-ZphawCd7vj^$$`u4C(Mdx+tZ*d(?;)S^l`h6hS%
z7;~nb*Y^fL8o2fy+p(zBqnth4@1ZkI?__!3*B&TM)026>SKm~&cd!Wm5-_n%cso=u
za%;zNP^-=>%!A9^vpdU2!ES71#3oTjkdmYoe;v*4$#f)IagZ-ree8!g^uzf}KY$lp
zxkb)?R<H_*S3=Kzny*OYt61&s{O33B9uxO;cNJZ@0EV2CGC^owl7#$0V5Y-`>3X3y
z!E-0s4pa1ZpUayoHe_qB(IJk^?*Z&k+xTumA(rIPqjnO*H3G*^h+!8(C&xHB^7A66
zM{V-O<C{$Og=E{1R!r^6PCGw>>(eSkVC6ZtaYh~?8wA)k9BAXw1fWI}I6o+*hhtvV
z*-hgk7i|fe{1Y<f2ULG;4m`qhgw(}~uY}@a#Gg@}tJw8Z^cm}0NG<^%u43lS?zwJj
zgFpB`#gTi!wz$6keyjr+YaJI^Hm||w@9F0Syk2FbT~nQyji~J0#-b(Cqz->Q-oY{D
znx|a&wDmu7%vsY<HIAB<1PQ{(<bZZ3<}#M?U8gXfeSRCh=qokjXToGvF)*L>@dWXe
zT^XaRnU~^0pv5}WH6v@!P}}9pug~>cD!qu5mPz`lu2Q^a1FLFBF(AyYT_;@Mi9bbS
zyLfnXbvU+S<CE}oS<WXhXB;GVEN*3cpQNbeQP^or$cGR0R)b(J^1$SsR8AA_^^Z@T
zp@7?C-b+u5?wb~@nPg4CJbHK2_DK3X!xhSa8PVerM6>V9^IB=KKOvo>YR(hkfh+H4
zVKWoxbI|sl3tl5nJgPz)ro#*k`HfLw;v+SJR}?a_i}4%e3+R9=+x<1GY0CR0@5E`@
z2mhG*r?#yV+$PGJl)k3y)2#Ue3;t*)I(5-#;Dl8%b&57iKNC@4J=VgS5&LI#p<bL=
z?V+?Eexxym$rYeNUh-!zZFfZWGE$8P_g2sK1Tht)2LzN!arJye>>tgeFp+y9GvWSL
z0c&K;tB`a*<TQ35k=GmiJS)z;o(IG&E@8%<<}YO%wxQzlgj<ibP1zD%Iezxui~8Oz
z{q4Bv`)@O5j8@zJq&rkXUqDqHj1vuIcEwX!CW*lc?0C?>cT&6d*sv|$0a$*obC+gh
zHX^46y<LA_ZwL8QYi*}Y68JeiAw%8ZG=i)?C2~vHPx5@bd<Z3g!KP{LWj#-wF{$nR
zqh5s_RI20Kg9gsGJISC<>kDN<sTI_3NVkYH6Uf4KwU$S?VeiS<R6A7)N6b32;`Pu$
zLwVBbpt%0G^b+zPSL5$AdMGNUTnQ!|*5|+a^z#g$Fpq6PW^i`R{mhiec<UWeVyU??
zT<Y8B@jrc<Ew5U{TV73|l``259BWFDdD6fa_rzkL9CQ6PX3A+A?3`vs)6~?$2Famv
z8gn<LtaH0T4`D2@W=K>*sH+!U+2{^AchbC?j*rmo4u=Yodpq53@o|%CmPM(+*(lMB
z2^$0`=CATNw%s*F`X*`P)cO2PD+{nGAnoU;=$_C^o94HeyR!1*O!!)6)GmLyZ>Ed(
z47r9gTw7-6rlzn9<}jDM4?VABm8Jk1p|-J$DE`zw&+3^04tf!Z7~cd9+{E63mHE?d
zX%a(nLa3z0uT=v!aRHIIg6K79k;u;TD@g{`k|V>yt%5B|RYf_00tKzi$qTgO-egF>
z5~@WkJt_9m`+M)Kc*dL(3WLoYs-6|_`{zDX>wT~=>CI<v-8GSId$<Z*u3l@e&K)92
z6KT@{m>povAL1{B-dS%KZy+f=D$maQj{FrHa?}d+rH^F&5p0-SCHg}?ROOFAYuW3*
zt{nfixa-HDkt}4DFXoy`(Rdgi#5*+ism|$q-D3hcvGkXxYKPjC_~+BGwBmJaIYH}Y
z&F*4mASOs;Dxg*oXO7gAq%kYj<re2zROzw;6WTtBmAYuB)|^(&A$){qTqnk*=Y11g
z8bWsRFOLDmT2ttne%UlAJ_R0<EuzQTjt{|$o(QdfKqc_)s|R2W5p4tLd0gt0T%Ryw
zV)gNc^)3o@WZfHq&K#Z$=~WIy+j+rj9kZs{C{n#bEb+rv%{T<--@^)WkQQfYMvRKx
zAaC$kvta|=4Ms7waO~HkbRCt!pN#nudp%zqKrc{%e0tc%_D58%xU!C^!rB+xc25J!
zPyps7aHMrd(OI!`lMC;>in>d-9<y)5-qNDo&xA)VqZe&kNA%RUC+<#i*|Bph8|`mG
zi`HKwCTdOQwYqI-d=K`g!A*zGcx;^}F1hcP6}n93+Dvx%S+kj+!6pW=vYzc>X~Z5@
zS-inz{KeG3gmBe4<UOy)niQj7_KPZxE{#h5mu7MNfZ1?Kfw#glu-mXe#|%$c;sI{$
zxtemvq-R#l!7mKBi&DxID935DqrAs4<?Q@aHKz2Ic!Ot*7`t<vMxLNXi2+X%{NFps
zVgOKQ2<0^WDe;#pt5(mX-2~+3vm{(%8@tOf+OtIEtw?*nhflqVhyaX<ODtbdY}`7J
zX{7zb8#CH&QmzGm{xJKPzRq)~`@7E*qbQ}s^YE)IQeNd;Pq)pL-hJkR&(x8^8*Cpx
z&cE-VkdA$gl~$v`POj>2Cyr!h{wH==ogX*9ao$@&MNcD>Oawiq#IhwmWD>L$6WFiN
zLzBfn00XF++hJxWq0?|O$W3!CJ@#iawoYr$mW}45D%6`YjJjB0`Mg|I3GYZJTDu!&
z;K<!@1%>T34e*-T>t0Bc9`m0MTo!vmf9I6H+)|rlWccv>eUYq3SdCw(8Rwo-6pvOB
z$V@lmLW(0Rf|T8`aB$k5x+)y=WI70X&7U}%^Bp(EzC}^CGxXlNFh%<ApcWq$s>)9S
zW7-sduxJnIORe_Ge))B=O^RYrg4g7?c-T?Z4szdrFzfC|a&1ZpM>*C9oO!14+LR(&
zXdMwPI@_YOsIT$3pQ*n<OV^+FEt1>GavgwVKN!>sqpj{elq6Y!!C+08vh)o5l0=F{
zMNr1G#dQV2;CY!P`wwhca}tBoDBn6uOI@+vU^fm6jR;@>(7t4uoS{iAVJ{D%6ihBt
zyNVhn0M)hYx6PXRS`Jb_c$@@F*>`AG&>d?`_n-r!pp5r1Nvtd49cgnivFinM#PI{C
z07|F<M&<8d5II%(Qqv}Zl!KTw>5+VXO6yh$*fX-Nf%_7l|L8=3&^0{-dj-LG%f8YS
zo<D~!{|Oy9ogyHg;7<0O>fXw2c7W~4XSGk<zfEJLAZV;@e8NTGqTLE=teqgL%f13v
znu6A!O}se!PwGISswT~wohJJ@9jZVv!RdNBy_qq)7Qv>a`l}ipp>O-wbYbrBvj^N_
zW|4SZQRO#r{BjhJbXAp|am5#(b*w!5_83dDAAI1+NjP3$cs(cy$Sej7H6?tSDv+B8
zmweGLl994>{a8*&PY<7ysR61Yqz4d!VJq4ED-0$^AC(K63i0?CYu#Z<<dA576L0Mp
z@?&z*y{2f00l^HH#D@1tnv?sx5!9)6Y)|ckbuBhI-u2A9mzcjo(lrsPhvWYF-led-
zQK{qrJL3T|IBoI`ESv~0qb~5|%F}Wx*>iiQHuG0<$tulce<&mdJ92Bragf~8;W>`&
zf0tfkreY5r9y#00H2&41AIa@4pS&|<1tkYo{TSdGc)!1To<Qg$9X-pq3es3aR!d)`
zLh7YNhU!=~JN_^htq}}&LmdsuH`f3i01!QpE=f8?Cw*Je#fbOhB8e0KguUp!mODp=
z6iaQEy2t#L$2rG@7k>D8T5qOuXejObjPma*Gny(yvBTqpE(E9O4q@a!5IO%iGXX`e
zo=27LD-$4yl9&}a?iF);{Q%C$oqvWplcS`6W+v^ioD)f5A=e7E0n0ch#w^G2UH>>d
zGzv~-Uj5N}%<UEJ{sOXa9(-NLk*1wM|HzqU0OIb%RBvvT<o?rlDWnknpk2>mO_VVi
zI?Meswci4nYqCW@d#x@%Hd9>|EauuTra@pM(^~bZDL0r(<4JBbO%X_;B#IqA**<Z8
zSU|yt9gY>O(Jp6E+JzfRxZp#k|I=(IXrdt$nfPJ~^r%I7fAw8@x-(vU7K(BobKb>e
zqf<u&GwcwK8@e8$NOKhG*|^&(&aBqy)SQ5_wIa~~nHH57P^G4kXF)M;eHty!Hrz`I
zv|(bRto{pPbrn(*n^jBcH(8|=TAGFF#!yCyc4H)IPL1>`!3KN6UNC}I4XJitOcwM?
zjc~{uzVXEohPN-@Dglh`>4$;pql@xYKi3nTC0qObnq@d5zbh0iV$QNFUN)MDG%Z}n
zPV9deAk{9Nit(c>HQoo5ED<zPSVJ0q%UFDJ(JE4o6S1biWCH8@`<??FzUMb{cZd>t
z=CyZH*698^cNn*?Dbjff@<fmZle;|o4lsb{<`7a-NI1~1UKSPIEN=6Q&MVaZeS?~9
z!c{N2*jyinN#}ha(v2y*iEqZS^|y})iML1l-h7DV^6Wx61A@DWbG=#4E1s+mA_bo_
zBwMsz<*#)^_T;UX6Wo%cnK1*L^o^gf^%`Q7Rl1E7%Nl4(B}+yv5yKVFhnGdGh%>n_
z0m+_Kdt>*W4#Gs37{keEy0PC10#!^*@hPXA`pj~N+y{w^|FFTwvm+O8^0CTzr0A_b
zV<DehEJWAt$wWC)xOzC_qS4Z;-^r#>VC3q1=B|%-<~;uuYMB#W_tj2~bDZzf%&c;6
zrc%Q{gsz(gy`#X27lo3D^2kL4!eLIHQ`v9?_zJb%q1B<I9w_8Um%(K*vea*Cbe<L*
zS+F(C#Q15Dkt*gX{_@NY#uHZ|QF=*+)bY&N9|I*w&GXOBq80lVi;ZZupOJmA*Pn-&
zOyPidfz%6trFXxc-MA#|##T<HWl#>%w2ZOC|7tXy8%kwZ89Uh7csf6GX8h51vMd?B
zv;3|9=Y%!HlrJNGJgJV;-05}dpjvdlS(UjctHit;dlA=(*v;NZ;WD%-m7xeG!1$D5
zi8~>u6qFch_-Bf4V4~n_YPw1IbxpnX?YK2lf#VHymw$Zr<ij7NV($xk=o`e_%Ac$q
z!9sZV+k>X#^i8~;jCy7ty1QxJX5vJpF6fQPyC(i-CEBfqOEv7G0kwdJ65o#-RXRwq
zxL$&mSy)si&KY2m+)!TPD+YV+U`qdU;{^5#^4B#JEN)IU*sLp4A~?)i-}MHPtntq@
zZ*GbjU@KAAelFh1@AC-aS!EpQv3QpKVpF5}EuUo}R{S5%bz4kquKG#hfcXCwJd&)N
zp6<%jLNT>7jVA8Z3s<Vf5{2u%k|-6#3s>k|Z)R)#yQX!MT3WIXgl=Kr)XMl9LVX4o
zf62icb&U^3We3g*!WqiWbSbp_+-Z=b0NHE#?MJh{l%ON$L|WBw6D)B@1LT}z)HyM6
z8Q_G2r42Zum>1V;*{}pw4^z?t90^Oi+1adu(c?8b35k+HgRLmB>gtKOgim-^qS)d5
z{RJl@ddO=r`*W9!^fq!z)7IZl<l05VK6y|vJnU^Jptm1Lke(@C99zYQh^%rLz_C}g
z5^Oz-z2^0s8RcLqcc8=Hr9JAb&YH0tlj3`=53XOqn&+4w4z9H_c67vj={G8I{RzfR
zPi075rpSY`WrSkV0JJd<jG-VSkf1~6$UZ$fGtIKII}_9~BT_<|WQu&xxvHU<4cF*9
zKe~QA#-?buJXbw3{gHgX4q(@2zwEwJQ*h@*KWxal?&aL?LxUP79D%qDZx8n3OUYJ^
zcI)>D;Am^wm+gDRfg{%+`nh|B;or$V)ZNA_de6UmyTln+^LM|#@N*uBeC5U%nmGgD
zmkng&t1cX|67+7LRH(7HzY-a8_F<yx7Q<R;Xmm3|seltNixU1v^Dw4sEjjZ&yivgM
z)sM)=)^5-{x2gZ?j1>2O-EHbDukRwwPA9JFsVbE6eLr6umDz2WlkeQ>rpKe~teiML
zxL6_<EJ`S9TG|aiBq6ze4BOP6A<ZuIe{}D0S0@tC^f?)W{vG#g`(P<ahECCmRpUB3
zK((-o`rLV+8j^}kJlCP<nltNotBs?JGC~6Il`LPQUUK|wb}&tPkwwRtj~Z;+krK!(
zvl|etGrm_tfB%F#<h)IksCdMRKVv!syHS`V3Zb@qrel{H&{M6NK69)Hv2IT!oNptl
z8x4A%_>X4Sz|UQhvlGoG?rvZ5aS}gO@5P!8YyY=%`e<O0cA(Yk2$KT+-Em0BvfZQF
zmQW&d<=;C7OT!{v;W@?Cm#GY0Vt@2ilgr0bMZRleJKmXmf<(KViA}=Pn~u(MF*lx%
zGuz9Y=Vf?jw(BJKPrm0^3WE;54XTC4#S5CTn$sX>qOGLFCIly86juw@N=U!VBs~Dn
z18ttcBK-nEFl`>M<LK(OVE#zw=G-f-w~t6+T#w#=i|sVdw}3-acXBvQXxBrz^gb8$
zTtrw7tf2%h%4oJG_bqW`dKW+hSFq7(Z?Nm41u6m*0OTqpnoBy~CG1-vl!HlSkgTYo
zDeYlifu=Hhh@|<U1nFkd6StA)*afpnCoQKedrQv{BmPUBA7?Z5mVd7u_G(Nu*@#Q~
z#udh{AgKI{8>nJ4c2o#X><+{r3Wo}YRS1)HabNvQD~+UWBy?gU;1_^1<0*v^d(AXa
z00}RnObvn#9;ALhe?VTv^@&?VJkz(^geQd9r%aSU^3UO$l4qS$C`>tFz(o6R>$DKI
zcv+?{#QZZs^@eNm{m$!D4BT8Vvm^DzJf*K#3Cl3-aVf;}vQL8<rmQ0H+~|9|xm7JZ
zaew3PpSL#!r*Z!rwaIn53dxkxv_<Gtj8pkhV6m$cJ0POuVv^x9V2yg2`f7eoai6(#
zx41f|0kd6ke~c&TG-8tga?cW{)3#$#XIBrXd4ypuy<KeDpHjhN(}3S?T}V&D!)Efr
z&jmQ<Qe(yXYLCf>>s+$nN)7M5%6hk`ov26xV^1POscc2fYIZ{GaC)4Y(Qv&ia!Uk~
zNG+nZf|oz&p26Io20?heWyp<$YVA1K{07lr*E4N6aqLZER8haSCAnx{_e|O&$iP=I
zQ`LUb8M<j0eb|`^5Sdcer(wX6%7$d+YuRg&N|!4TXsXy-tv9KM7(oQ;gwuSyRpw>)
zG~0HndH2>%#CIirmYdernRS#ZhoR5Uw85-}<4=JrbO4sd_$3VuWrxFrm!4QLVkJ1F
zqW>crP+Vof_{BQXpzWTi1$+KD=9G}$9re)4mX}<JM+Ct{wjQ@vz}M0xKlPskf?c$$
zohIndd48?Aob7uOP8hdVg<EUO7fCS;&zbW6=$`a)&K#PysfE}!nK*_SIQ<!&l@xy%
z;Woav0CmrB1re<T9#I{<y*{oyl^MA&;SyG|ePw#fPO46_yHNY9tSIeB`A|b{4+H61
zP?T)^I4VIeyJ(o5`aC`nt_bBT?6o5wnhu{Ml6TJZP3njfhFmkhFsZYJ1)QqhPV^x~
zmz*uZ%bt$jTGT`oUu&MGiA!T6|JA~d`&>y>g5o68@1{{9Mi~KC5E;ez&YjrRK+2tz
zEp7IJ6&9X-0P|m`bdOe$`v76ChrSH|X(GQyH&%A7&k)x%m5HLrtV;{+Nq0znpL^qW
z{r2IVYPLM7vlx`z=px+Nlt?i4xQIzp5<oik%~6<wweiwJJ?x5w`WurQe3dx|9GrZQ
zhx_9TPZQ`}464Tm>lYNo1w(35i!J52e+ln+{1TgW((x(P)>I!jk0sc<NYci$Aesiy
z!{cnoY`bcRV>6kDYUKA9L>VMNU1*u|sth?9M)1RCO?|He?pV7|9mUjnEm4a%!}Lsh
zFA&r~RD$!ksRiZN1@|JeQnliDP5}>oUe9AvxC=3PcAK7}TXSx{uZiAt$RFXG64!lF
z{3jlf-%q$tdBCc8`AT7UcXXm>pZWw|^C2}+24Ova!93eJ<=njYp9-R*0h_<8J){Zv
z8UV2Y4!%$QC$@ymY_9jyjP6<X;WPB2SJBjf?O;FnB8&bWsiWV(i*hqHnwfe&P0&h9
z`J=gxtb!tF1Y(@iaHqU}jjuaqT=MR-UA1UHY|k;Z#x?)p+nfk1b4gIo+X9jmI+1s#
zJv<e9M~mE?`RC*VpmGC_?L=;2X7Ek)gUrm31nV6cd^q;T$aG*oGvxHCZOEAh0>V-D
z%R_YH8A;RO9>?o8S%@k`Ja@K}=<&!Lux&E$38Ve3czLJ$vvQSd(+E-haBfj<cDm-7
zk3FaC-m)xTAgI^)nhE^QUX-v@{W<HH7syYle_b3v;bqhm>V=>Wq;DHHRroej!m#?7
zhPH-}Ieh1LW@WW^85BC6p_~w>mu&(I`cD8Ex3J){dsbubaAUYDaiL>4UbFZQFQs$<
zC-BTJtL(qgW;YEilwX4LzV6s(o^RI6{7*n?Dnr((s<?Sb(^&rFZSQA2bOyKK(^Mea
zdgW{uspZep(6_<uc}o7XVLC2Fp6Lo?R}wuGJ-sNMFcHJ_!}K7@t&Zj^jTHyuCt)WG
z!zgvfKfO%=?B0AzB?xMdkD5_z3M|deItxnCv73UAC-WHGSO-|MXG>eKPxgPahf!CD
zrJ)K07VTX%i+*R{{|0QQecAg*AMff#y0SJoKXqLu1E^o^>!|a6R4jTIq5|_+*ex!S
zRE&9Y79{#DzE!>UJ(mw)zc?v~OvfRB{D&5x(A@0agL+aclQ$k`M*42j!z1i1fJPsQ
z>z0c3Q9t|h3ElFgu~cV#MM~PVC%?##K^4P<{E1uxlerLZut363nK!1}B*)YiT#ow&
z1RFfTd@tNv_)zY*Lh2qz^D*WX&for{jR^^eseB+R6#Of`pEvCR=j`83*QHGcWZ?$5
z_f8hV@>0(H5ZZ=A)roA2G5M%t)r`|%V#NDD^E!c?UB^5^+fHb0zmR{LGSo{AH(FO+
zs$*vI)hD2abm!XBk?||*71$iVPl$0E>o^$*n$I>NUR@xz`smtWDo*sFAyc6obT`P1
ziSge%MBxK6C5=oAXr^TyddN4Q&L1iyyjtK75#CLN33IHN`a^nfJB*4eWpp=vjqsCN
z=Pa${Ro^#cRKGNfB(j!~jIk4Bm4>-ujDPdeY4i%Y`Uw$8#Rcx4l)a6CzuLNL1%VQ{
zX8Qd?NXMw$QFm1=2u<;|%Q?8T4%el$$}k?RrXII;8Q<x=O%rvC`nix3|6TS$phE=h
zRXd>rRq>x}lyZQo7co;3Qs6N*)9rZ<sjDi^vYUWU`ihn59-(V=ijgSGh>1~)J@tZ)
zMYGNP`Vy8Q1v95^|Cd|9XIu8P+q;Y!4u^k<r2bKnVbLpl=VToK8@JfkOf$*A`4k2V
zqi_2ZFfP!hgBwYufhzeoMOHjo8bc-%PdJ{6Jr=ldkgI|CgrB?albNIeT~;BENfA?n
zRr`&7sm{XO-g?2wgQ&8_t$KL9Mgg#+D2oG7(0*5TI;acrPpxQxfI)s@P<oSKh~J@v
zUtxtOtaBz<TYjX0yWv`0*EQ=6DfYW=!le1TTB(ts(vGYszMTC75uyWj2hcN3JcJA-
z0KroOJ+9>`uEoM=eqsbn<S=^;tQiP6f6+`s8a92~MzqtSkIkU-vM1lK+hFIZ>la+!
zJ{nzrVD(kiDR4uVr^uM6S`0jxSC|UIo3pDuRq(+=KZACR!L@TSo^!pC&8Nm$%`AX7
z#uxLR?q2WGn0H6ey{+|RuW9ZXlm~@Yk#o_ddC9uFI2}+xy4q%O1Gyo$gJ+Kpo-M8s
zkG!$n+iQCt<rcvFK>q^)fe=Yd*tAj?w)^1MrQ;oS*V-x#+RQgRY{<XpPn11|w}0`W
zoih8G+$%K0I=)G3OA(yut*6}&(Ri7sxf1gQ;^1Pv<Q=)rawFTr-)E+D3gGAodbD9h
zQdUgGjIZCh#DpL|jIPjnh+VPs#kP3UuFz@Z=q?7FZW{HI3k_7SG}JJ|2OmscMs-gk
z^WmaKXAlMY$J?UM*I?V30WT+Zg?ZoLs{#)67iVV*9dxO|MCBv5f;P9fFJ<bIS#c<E
za<J}a&tk2>XO}N)NX{gmsbbwXvF|$23H)=>H1b<*;c|<mqZx~DmLE4g%Js0UbK?H3
z)Oq?~Z@#bs#xoQcT43z?CTYewL(s5{1BJOXph(Wlym$)<QBKX+w?-{@4H?e1qjxKB
z_$-NOPpqce&JB3jQ(u|&AAW`HEiyIXWVNVf<Scp-5``pgoC&|OYEO84#!HuKT1fNP
zQY;`JpGN35B*{E2$~2C1RRG0lOWbr|jnXi&prXVUyO&A%&kFRz+thDH9LQ{k`2EBI
zD;fJ7ot^PNtY<R|i`ahgj1zLI%3c%}0IvrOL1tBOkwD98<Z^k_wk=`gQYbJ%=s+q;
z?qS$AAwZhpagzK6@4+@kWD6Hw+FA&-YoXrdG#rQ+8Y-V!D7P(HdPEqGh!u-NTeemB
z-m-*>Cap1@1nghek6*WfzaU1hcMF}pOJcUi1xvOR^tRixW-n$fHQFYa#}AL}$jFp$
zLvG(+E^iV-3+Jx@m(%NQyD&h7cl%G#oAx7=VelVF$@Y@)_(SHe?P-C&c-)_fWx44^
zj@H%NwS(4n;)%Gp!fu0f7P8%N$o$erckWxN?B;|Ey!_FqTYnYR*9_&4uAc|F%HG!5
zPS2D9N?0mfdBEobh0ZB_gf(pyQL&N&&CoW7TNej1C_mS=xVBXWBS+^KrMM_NbdqP>
zcl55I^|wNm#9+xo0TUDF?(dhEfqiv>O$gkCk-Qdj?AF^2ic>g!31#wDJj>wqLVa_2
z&46_iQ$8v|JaN<#+h``hh2E%e4AU1uxS7_qNsd;n#u&Dg;(1S1^H-EZICeZ~mp0{B
z`<u=4WZ~Qj_hZ5=5ju59p5JWe#5XFqm<+}v9@t!9Yz$|Qr7s&8hijUb2ZCuiH;Mv&
zQCpEMa{3tu#2qV`)Ol^!om>hBCJHS*&PYlwm-b;DdFS6n=9zL4cJ8{0(w_E$4;djG
z|2uonRp~FE_|vq>IoKdL_^RHm4d-GesgxQDzMRq5m!!wmY7`q9j%$a@H`B!H!bhjC
zu&I_ypmyDb1=BxjI)Ni}%IB_KHPPIKpFE9`@I{jJFqJn*g(17t&zF~T2tz&>uhhM@
z>mhk7DyjnPklJn_hHKl3!Cs|jTvVKnbx7I*<Yc>gJxS&j<QQ+I1~J@i+n<acA$SgN
zJC^&Wi8aHkkx7F}MOI;A<5g2BNJFQm$m6H7d_10+kcVZ)6}qdk6{y`wsQ}iJd88QA
zPnF(9Cz(X5O^tbrj#0HLi_Dw+BmdjCV}pD=SBx?*wXJ+oh;m1o<Yi%ubdo`AbQ*v=
z8y1Ab_Q6%BfeulWnGv&#%P~=&d7(#O4HBNZoqvG6iJ|gcDM=Tf0uQNVyP5@V3B+DX
zUaI{%Ai-?Y_PWQoF7}6nvn#>7A{ezHPT;?9jx;j?L07;ZmUwCzKaq3R<LJO}rd6{t
zRk2;M9D-U~|1;0&9U=@?hT&R%&F}{!i%4Ssb?G?eq!U%%<>%Z61M8uVRHE^a<KH{w
z(@8NEMCaiXbkhJXE#Au92+$)|iy2F`vK@zt6>WSWIz1dn!ZvSXyl%5poU;8-2H0?o
ziSr4G*s|3WrANo8?AX)!@eGZEDLa+$(4bOIN)!vQ1lu~(uyx{<`fUfd^+vTF*RUG+
z;zBj_Ce(Ak1pAD;Q0u{?$L*xUBa^U7M&m9`VD?3(d2cfi(SIG%gCXa-n!i?=c18ii
zqpbt@R`}QqBUCU=`AL;uzQ&%WCho7mwL~rclRKJgQN`tI2=~L>1)8CUugn@ILM3{{
z3>^LPRowYG&0BcNJr87DjCiPsNj?Xf1aE0cP_c%S9K}T(l4Japwo6T%A}m}vz4`sj
zZB>D}`y4`;pOi#E_5>!{o~~ZNbK0oUl4~P5^L3xncJl+D!CqDwH6&wa(Z8L<ng_*O
zFkr;Z%V-N-6#A|y{H!X*$KjYGAE7iASaBAgi@8eOuU_|KeB0aFN=rL{;)nd^ju>^@
zcyMAXqYiB)7oe<P-7Bwm8FzNLzuc52#}m|9H1(ySg61Q`#Z-CpXG8&VxYr@~=-)f%
zA1CDFD$f&-%o*tZh(@eTT7wo1x-cyAxMJ`Q1#$jL{?jIDT_y+2P>0PmTaYV7q_&~G
zwOJ-eNjoVc<{0$|Vma}iRK4ota#;G_v|w?v|HaySM>W~>U8CL>6i|@fK}tYElM;GS
zq=qIT2@smJ2%&eR$gP0%UP4DYB!nK2u2ktlAoO0OE7BB%`{a4Qwa$5-_q=O;=dAVp
zJL{U6-!*g1zNYQHAp*{eZo5Tw|J*1M-Yeb9UHPo|!&+Ho)jm7c#LvlAY94r5O1fUX
zsojMQTMpor+$6xult$!JIo49{kR%o)M-$`j;Z4O}lkc9Tc>6@2=)@oK$JaX9Wk>#~
z+~Wh{%Jc&FqjlBN&^d0cBfjo|NMFe(Rfi+_SkuPAy>AV`$#kA#En>wGgB|8!x>QSl
z#H{+Hvu2ltf57OJg@fx5X|MFVJb6ssLuJ(`26ST1{2hhs3p!n&sz(~D9uHeXJ?*1Q
zl<cA&G#VYY1TFTI4;iq8vn;nb@-_x2{gDqYlFEMP-k12K=l;Wel^92fDa`Y-Bh^8D
z;naE9S2)t_=XQITnZRzuoQm=9)YHXRda}T9Sbou13nnFXCnLFDkn9<c>VY*b_g_05
zBP)$MKO_L}Gkdz7O}bmz=tU?sSvC<F5svfcxF)~?pe1$hL?Io`TCMZC``z$1ay_x6
z4)0$%V0bDiT`a$}7riqo-Ng{hwa#wzw)UoubZh9*9jp69T78waL4bag5@)V_GJ!Sm
z(w|&hjIJ(bw!VS1Uu+gAP`3jUMD<GK!&43zAteH0O2W0KvpxI#1=3QY<D~V+DQDJ5
z%d9Kfh}C#H%%FamsyeWrx(Y6Sq111)X#618%DN5sK7AH9&+s}GjD0NV7o@HB{@b%)
zazSxhExzKY?fvGqxj=k)$OAFQAlvXCOJYS&KgDSELRpphN`#3*CUio2RYoPdo<icc
ze)gY<6X7T(7fR9zxgOF<e%cB?DYf~wBIS5;7CqIzl`kg=5T7NC|G0duHTIUt@T$d0
zvf=%Ku_-8d$gVk+(iTMISahDBc3~avD=R-&r%3XE<ie>fQDzNui^@v+l+m(Zi=Uv(
zO+_A$pqK21|G8o0)8krs0kA02x1QU4MLh|jg+7KM;#h+psU5s5-jr~aN}YHXgIStj
za{G%E6MV03bV_IDtcPS_GHdGih4m_Sp^c>+c(&VGm++w0dQW-}b6{LT@LU>f)rG*?
z_=0mz5GCLQxynz4F8l=&+2VJ?y(jp<Z@ItJ(W-6k&WCCHlqAu`6$B~M(&8W_-cs2$
zQ^%x_962zxq4CB!DlDhrq@|~Kx4-gVw{9pc^3$rYM0SmT)zhT25%5kruZI5K_)E25
z+lEj^K7qNc=&T<Ll|JYtNu9!3zk+ma(}H?#LNRoywzjj^R@L{zZVcIX8;Mu~Fc>xE
z{&1E>Iz=U{p$Kn>$oRc?6C%~K&vGxIz3p|mXX0rt(s$gZm~oM4CSx(+XX>0eb1t+7
zN?c@^p0Z`+nU$P!f(@bB_=hs`Km5{CP;dnwFfNw#4zo?kUm<eU%cT8{btOD~;I>a}
z9t~g@r*+ZXXA^b={f_~E$cQ-FkY}XJ5$~DscW6_fI0Ky(F3)14GE_=%yr@Z1Fi{mD
zywGn?{j+heXTLzjFl(T4PkJXXFFT#kcPlI7tq}QXJP|4N_~B^<(an+N;QfgG0db&m
zqh;1PJ2r30w5LwkOn;}{>7z?R+*jv4-zIP3hD;d4)=0EcyIvZ+LNQH0i9(L;MWOO4
z+xMDr*WGno5sXQ-X7OeSgAI_3sf3RKV*{Q7zSYx?%#CfC6r%eo_|uV78hiOzNL-v6
zAA>p2a8iH~>n?o*JRc~7J>S@xEpFH&1$eP+l_#Mhca93NX5K{KA`Z-a330%C&Q^z7
zkta6HCjQaPN848~yqI3Jda(X%6jngc{^(n9Y({e|n}xfL^WeIAV7LaYib<Zjl9Dp|
z)XK?pqJ2y7TA$VX{)VbMUCdZEWBVUK!zaFMRW_KxK^uHVHW<!Xst4N@5)%9Eth{6m
z;9D1Fmft_HLDgHdkVMdpq*mYVP62~~hNt#f6=Ptd&18TIgmqY7GeYEULx}4E+&77>
za!(ZN+Gu9-g(&_?G~%7VUgS)-@5?^<(%EhuiIsMz0EF*`PgTAfA-`OCg5%9GpF|mC
zOhaGZ4Cm$@t(~l!P<kQr=~Rz@eV?OfPT0{DQ6<^qog>2yQvOy4Mj&xA6cF|vLV*OI
zW!KH#10KG|Zh=`-POMc>CA1k<=W7K6aG1cOhA!3C)$fZSr?F{I{;^17`hdNB(GRuT
zWNJtQ-)#iF?p`c8x11DS)@=AOCc@x^Si#Nc1SoM=8Wt|N6@p?YbTcR7E^@2wx;!M<
z(nA;I^yJvHm3oFihh-htgw6=N<$Bj>frfnXl1y`JDsa6<u?Q<8d7NsHI(Aszf?w(C
z7p5{++sCz~`&`DiX=>WODj?x6$3jM#o+srZ-ARI2S~$v{KNsdaC(tU+RbB(h7sgfx
zI{8*$8%JmC+H(rF26}j}z3LE?o%`6)0mlIM6}u^<ewHuI&d)!SN-I{v*~szb<C@Y&
z%`&FyJ?8{{#es&_Z-f_j#wv@Xu__nK^IxtzNKW7(`%jHIubwh6YDqeToYOfATrClF
ze;uzuyKX4g?$7$n`VPDnYawy9e{5WOFpcHC2}VMpTRG!K+xc7)I=ru|CD&geHkZGz
z<%RtG>%p7s1547?!>%<?@vQo94s$#I-0-cTzE-`h?sjLfbBr*S<h92*Vz7j(vO1i&
zJ8-Xe=-uzDm<y(8&&AaS>>k62%2Io!)!`uZ_OmxkoorVf_O1RHk9nW(bIP^4gy4O|
za{7yqXOz_p#}JFo;nSmwgi0f+<v_-!%~Y<hs=KW)_4!WIK$2$CX*4qXyAAPAUFgL<
z3;*&~|IXy)SH8A;^dmA$y|~}@cTXO+99e;w^u>;!{l~M$sPa~0ROi=Vo6OS78GNVk
zhwFDtmO@*iIaXVt|NHFz)fp`wAz9S_5mq*X`h2UYLhAp%%!8H%i&odmU;mMRuQ#Z7
zzW4gS2QTh><mstn^6!N{YW1zEJD>UYB$4Up-!V-wt#vgQy^?z8wN7tLpIzM+t*C`M
z$@td!ncgJ|X1TS;nEfQRIk(SR2kKuh`tOZR`_GLR@oLv2$L&+DO+UFzN*z1SMBIiR
zt9*zdJ#LHG>uz@Wmb{QqX-v|c3OMWBOr`!Bc5mx@%dz?O<MjV}{`4S3pi`DKH|HhH
z5B>#fQx^O&G|8w=3a8q6Z}$7(Ul*Xrn+Gcu8=YJQ21Jo8Tf!fQ`OVF?ie(lJ74-1!
z))4emz@EEs3^Mgs;8f#K{h!xHL)SVzcL=|Y(kDsZswo4hTifd_%?9somsRIbWHstF
z=@AXedh$t8EeD+SZ^o&dJZ<)wcQOr1IpZ>^a_l(8rTnynlD^ePbSGk5d<L}!L{OI9
z&juWUlsu}O(eoEbG{3hgNqZv%q=raKduDZt&K1~v+4>1cqF28f0{rs%)cBfj#6~AP
zLd)3GO`gk`YnC?X(<_nbUsn{$b#25Xu0Mx6yQ`OM6N{oMv$B>(dy`i;>93HcIQ=Z^
z9GLG;z&vYuP6J5U{NgJ1ZhrpIt|rkVYvad7x^&R}>pKUII}6FY8Royz%et%mB5x|!
zFH8$qJozwH=p!O#rr5#Qa(GnERD5E<dHWXj)*zYFkkLhQ<%vkhE^*}R@NYD0nZ$|H
z!i@wR8>r;kTR)%vpBsy~w4ushQ_W}*oniIrzZ3`wgei*a26JM*BCn4;*88;Gxf)M6
zI5{^OH1|(>eIuw5c-(33i!?ed*lHUxs1V$1v+w&=uRMMD5eggA2Mk~O{|)Up)ol_R
zTx!+27(0F$@(mJp%6NL@;4Zox`sB}o?y0EaCo`o>*H*97;~{H9yFE~!)oInsW~QzV
z7u8K7Ld*D(psC#$2rG6m|5WpIx))!@#WNP=HQ9VaNMA=1UgaMMHaMMb@BB$~dj9uN
zzJ=#^A>pL0^4~Mwq|UPl7*8-*Xncv<`K&liVzbu*NP<a1*K6D4KbKPTljSmQPjxtS
z_!)#>442m_WRbY<&6MtKO}SFnk4wptOwiw4nonKYO$GLtM<kl(d<VFKzuh{K6mxyb
zWc2Qr)!Cl4%K!RQJbx~^Q?G@Qh3K6Zdh93{{W}hbWcCL9`1fJ)R>k1F{ojWb&!T^I
zR{BowwZp%UtABymzu@QJvxfh#!TRU_6Z(+Ge@MoE4C*#xk#`5z=fY=x$p2oGq4yfe
z!{dLb3*9m6_5DB6r=)=~P5*)hdBgueM$${flrjDO@ZYK&Q76RwXTX5o|DHkup&$Qm
zs3l<~371Fyy{Z0h1p4n#EB-&n|3Am$**SLpAFo7e)-)?`eEXa3f9p9K(N|{arq0W*
za<53gi>>_cmec?GT|8_ep*^UTaLuT_t@&Tz7-;o#l!@z~8>$bR-0ojaD+kIP*9b==
z(cdj+|J1)DG5iaI{uUjjJo;7r<-<?0EDPQC9kc#|W1-RPf^+n-bjxe4zb$f@I`e(C
zwCzZg>wj)|HqO3wlMmVcM9MY&ho|sMspFf|gM6jU8zhaT(wn3%Ed1kYKolYYbD-_d
zzWfUkMv)vg2P%FD>@kf1{nN@^^Qo|XIrck8Hq?+pyh?V}8a40l(r4&jgN4my3zqb6
zg6Fq2t=wL)481!Z3lZsP;xdIh(vTcsp1|!ZP%({*%B()D8%5RCZ{Jdi+Kov?iZZ)R
z9Q|`cqQxc6_Y_l@$F9ts23`K_IfPc~bobw3kso2wD82%3$p*aXT5m|JK5Zc;2mJOv
ziSb#A&J$ejkGFbgKLluGSbjf4{uE!Gbvp*N9*1bLA1WS9BeNaGE4NBOU+G<VS0vG+
zXFRArWGB{xq1vVZJ4YUo<FTryP<+>aP~}C6LWe|&W9z&ujoJf-in(@oK7aTxD!nyZ
z*WGSm?cVl0hE^2{_p@I;^Y{hJZi-NNBj=pY-o(1^Z5AKcNX{bd{98yx6`vH2A`Kv}
zm$8F@?tTUHDqp=}ldY$G+3AQ0c4J;ckkJ*^-&vb|HFYF{%vCL^v$46I^%U)W1wq0M
zYuB;niPpl<C>le<IM2eoG6GEbO~v13#XHH8X)H}+M&<*iEDi)ODrdRU=0j(4U$w}s
z+VZM^A{gsr%PiIEeluP)o9$JVA&zoxzIyiDzCBcWO1jf&Q7Ry=)8oaiu>XLHL8T#T
zybxNEVmjGkQg8d@G9dlZzWgKsKVp*<sZ?@8Kzx{ToN58-aKMliQz*3)JUH`3TGXts
zmxR$zLtF63(_l12>Rv5(%W}!!YmJ1>QobMhjA_|#S>Y(l7+wm+hGBtE31mW#@LW_<
z%s6K)57uV61Sj*U0YmEKgPK~gSl!vEEK1%7`}5Tg#lRc7{Cc!s>qK_jr@i3}w>WA#
zuyRr@cgWZ;@XuaUBqAs6Q?(M<9UgA#26`&`nQLYS-dSTMqd%(~2ZYcYKmO$!Zjcb0
z{-(*QV!&{RR#38N!AwxJJ`EWA+ELCfK}P&0oT4wbCq|s}KBJRbWMAvCRJDNIi}5qw
zl*H7`yzGRV+K&bYKE+5%vX~hq2PV_s6A3bjCW6L1#WCf5%|pq+JOf)$LV$J0@IKB|
zHp;<*kzrHygVpe~h|t6`7BQ{FGV9+O$#4$`a{zrwL^XWl`-G<T_ST6U%!}YxOdrr~
zo16NRx5#ioC3sdjaT<7^xWgPBVOU|^PuPP3Q*$j-6u-!dJs8pm%GJncwb8e)DSgz3
zj@%Z{i;bfY762-<W#!C0NzG3DWP*YOz-{zhZ>MT#re%lHTG~*iyV`%5sj&qmaks(@
ziGhaB<&AX(UpYEw9}G*|?92uX_n92bn!F7mN!nO>Hbv0Sj6hu}@Dx58%PoR30K8_3
zVB;aV?<R8A69gfa^=6uV<bkgP^EnNgMb!iAxZa;P=*ON|gNobfcWlZiA6Sn72MO8S
z(a6FML^j|BI;{foO3=r2=7ssLGi!02X{|l*So{NFr?!}<cv#Ew*Uaz_oaD;fsZU18
z!__Z>3WnWA;<oa+$QHhy$4T}`7S>zrXVQ|J>tGO%j?BeCq=#`mEQs&@H6Xjm?+NS@
ztn`<HKz*n?@+L@G3YA;pI_KvlEx(5~cv=xZl~+@?=uM%9POflFkSFGBki{8tR1E4+
zRLEU1bar=ZOzuvCqc3f@!P!Or^@fFmaFkD$(B45AK2YuN0e&07041<5J!l^;+-_-6
zPpuX^U?M~VFpHG8#;K=h$Y_}Ki8wO%wZPfKhU^r4>VTCKgB?hEfLSwWW1&31!~(y?
zb=;h$R1Uj8X<w$2nf^ID`gcop8@Tb%F+&Bdqk7P*-5gZo?z&MY1m8^a0HK~%KPw%^
zeyV_N8D%FRDw|$lyQ)Si;NV(mJLcfUuZ$VN$@UG;N~~TzV6ep?xr$a7grLQr`sVa;
z<27Lq0Bed~;OZ|hp~-0?zaW9`1z)MX-3Q$kzSx+OVq3xo@R!TfdvrAInz2uolB}Yh
zTuYP88je#W@#L=~AmxyHkigfC3L{#Nl8yD^dR_^(wK_gyaav($5L_cjs~x-4_sXwJ
zpj#2#7yI{|^x4Bzl0IoP6U`PcmD6t1lk1^o*S#nVue1ig)D>KvEL11gIEGA;uqqBz
z%W#olRQou-Ay}5@Lh*b3m+e}c&|9TqA(*gJ)Q>INQvTRlbr3KkxfNVU;8MQ0rVo;n
zg|U+Ml~NreE7LQg$mkz`_jY(;(-W3p%f4~0oLg~9(MX>5RwpsHy#LRcs*E+Spg4%*
zQE~1w4w|noyMy4hm^0p-EgsOQlZ8F>`S))xmqT7w?^=Hc9$b_J$Jo?fm*t%cab5>s
z1Ur>HKQH*M({r_aK7Kg;E9ra40*SHGKKaiLx<5vJq*HXxCSA>l&R?r46U$@e%i~r@
zEhML#lk8({E|QhmhRUi3iYxcpYyR5ex8}zXY>32pzSp_Ise#MPb^RF;?DgltwZ&nI
zRYpi?YRWlhmsm&O@`=Zz%j#B?@AdI~2y6!B+g`RkMf~zhCA1(sBR5R$UNg?zv^)El
z^3boIEfV(LWmv62loHr>z;BWL-f#`ZTC~t`VgGE<JM&?wI46oX=T%-5pmY&3<Ma58
zYwdN+k37&>Yzh2{>*MZ^jVaU`%t-B1T}edgbj5IPsClo4Em+w%O;SdfWHh?|DT>j;
zrr5+ej{ML@nwKGj>FLQ_xY!+C9q5LDBZ1N~FFH}6#CA}_L_f6*R7*?W&E0O@JRFEP
zam8TGBb!k3+w;Lc)*l`jzhXFlwHh4pdQE*^h~`JBof)DFzlRR*=jUUqKz{?<M7cdH
zpIl0K@J^;gc>xCED3sxZJVYcg{gJKJVEKGKY2w)w<H0~lY4~zA*7hdIiUziG0yck4
zp!ZMQT<OWL7zPhY2>VATu5^1tgN!kt$ZvKbZ?gFWFeVhf@LYh5iqy`9)4T{kgcYTb
zHIQoD{OFUNRy2{{73z|ei7DhLo;5>RNK23YDMrZP@mBSREo2YJD^BNac8VrlF5kx}
zHT&F=qrj<SHKHJ8Ixa$BG^-ubQd^L}n1;n^SH+p4m-k_yo?{{Y8HV8G9S{%7%`BU$
zpqs!0?f$fW(BZWmFc?rLdD7k)oYLrh%dihqSMvqG=8P*Twfn)(oR*DF5R$x!CQmBa
zvQ~S`NjZKb`lZGOU(4Mrigucd$?JY}3)i2W%QdHih9T>DO;bNJyHk~Tm}j_<Jl4k9
z92OF<xlO4wJA`G&M{bvy!BlsfDxm%81A@J3*?_9J4f#g%^M7vC5mQp0FMX94>Ob0o
zK+xb`2~XK~l6i3v)c2?7Y2QW`lI-HecSJNRM)hiI8Ca7(sJ^LZ?`!YiO4t^Pa`$~y
zRGVAe)2l3OD$QdtnwJ6})dn>K3p<dvAER(OyGjFDz{0)Yl^>gu;<fc=QM2L0OH4MK
zv6XzE*o^Ky8Srm?X#tmvzpQI<>C{FMTz85l7?hO?tWD^)>5O!BzJ!g8aSTSKW~ERm
zow6NC5~O3+w*APr$JBnc-dVfdm!(;`?rf4FMl*hN%k)qX1jz}upN!g>M8}x*m?T>k
z{e_&0kH64`4;FT%MG^z3pP3-r7IX}&)>wTY20sV+@S=HmZNW^C<&^cOu;c+r3dUA@
zyW;%6u5!oYGEhBNhOF2b6>>_T|NI-$;d`cYll3)V>0_J?)OA??qj4B9ZOCu2p~dSP
z)z6yUZ&Z_IP>g$pcUG#-npD(S?V7TTRX9yQn<s=dWzVf^{a-g?2OQDr2uM!`JT(01
zK8}r!z!r;zVgRhJC@#0>JrHfqT9A+sr>ieqnk$CNlHI_6-?4TbROQSNB(^}<V#AL1
zcFVqd%ljfbKRF7-&Ec3_PiOQb<3PLlaZB>oH8nd~Ll)U-C9?_~OB_tylVt5xdshe+
zTd{xK<%IgtxC<}kTQk25W$|E*K$H#^5qI#dqdMXa^m8Mkfx=KnJi?M|TzJk0lKsLy
z(;eke+8_L$U3Ja?dete_3MdtGjJ1aq!T?UpW)wOEcWI__ExpfqWg}WI3Y0~_C=FNB
zBLK2^>qz*@69qvbcvptOy>DsRwc?-wmv_~eJw;hRH4MD2@K2ti2wk5lTj_hhZCKEk
z?J0d*wq_|Vib`0HSt4=md1U04n$ui)bh@tnLGh;rr=H-lK<V%1il6^LI)&$@tHa|h
z&RsGi&h5vY=Ww5g>A?-tqTo}iB>A-GYVr`6Q7R_qV?}}ixK!RDR%7KLq$WP`2`AH;
z-dX8~HD{MF_XE#nK+L>)k+BDWH3?oRZ>v$eu_BGuFV`O$g>rCuK;RVMXR)>Xut$;z
zV3C34j%$K({siM+*-0hPGJAjxAsvN+Xdsc6rYvT5#+CkGt(Dn1v_gmq4PHK`7sQd_
zVPB$itCbvP_9$bYuA5cCkBrBho)XcH9A^-WG=S|fQawWWoi{(U8-(!kl$nIO=oJ$T
z0W(L-*`Ua=foQ@|;<9FNoOf=-iPC{k;h(@5fn2?A`nfdAT=RS_ao#$xCe3QgDEs+r
z<cp+n3n73w?5&*Cwr&(7JGp<7byxHmpTMP=c7x!s&fC4oXm7p*70%lYj#|JCRBsGG
zNkm~*e6$|I2iNJKSF(39KQA)hD#WHxv(eVa<mgcN-gaK=`;%zX$TcTzAx3M>DZ9w1
zFECg3G;)Zkl<&od#mTILtVIe6fl25ARtMMYVN|GZbmCB3prZ8Cyy|;kki&CT=4dOe
zH<Plr3}zH);aGw{!dfasyLuJa1+`U+Qtg@|&l28L{<s^~H8fZRJobE0s;7Y_(R~Vs
zpQlG)?@l3jdQC9eC+2rdKA#G*m}+_L|5-!o7YWxe$7SdLJy32~LQ4r+vqZ4lCp<_X
zh7oJW%v$<Xf5@;NxO6rz_d4a@tpzW1?PVn;h|K$^Sa$E#CB0mrf{Fc+LF=m19+g9Q
z=cizkHr<*0H(x^SfOh_Mmht-1-53Lb26*BV2CE;-U&*M8*MoSWy^|G-9P$<66*I|k
z3Xwawly3jh`}1Yn$%I;zr%PHK=WW68jI{TOJQ`tNvZ^9`Oy8BCzIKsbr%3GumfXx0
zr)o?TSl1fvN{+7;DCT(ZXhfOrrJ6~cQ_hH1{Irufu7xKXLY;WmP7mJ3V*a47jVU2_
za-EUdCY>&K2>yvlq^qY*xmEaibrV=aI?d2qwZ)OX2veBl1N1O}a^^ny^M)yw?Tu8w
z_US)2jw~n0ZuWaLkbUH0TV=OPtq|drF8^puOOo3)vXI@sdl;9GP}~33+~Tx|3|{0{
z03{qAU5dE&@mHI*=aruZfNiNq9AUs$ZR&->om;e$62UF1;C_di;4OOP>G}iiy<I1i
zQ$2}I)G=AK@z*R}A)5g%FzamgD0@hMq{*qtWy<|!+^X(}dFfuB0?Us=K7ur~87CL&
ztZAGU8YOb4L5w!fV~-E>Z%Y?(M4stjCrBqEEew^@K<^6?&g_Kp6dzU`!WWUNmzWmL
z0_PjvKiM#msH^)ykm*ZcWS8c5_2gtKHg${|eh>{Pd8Cn*nV^*wL1t-5x85H|Gom9w
z|JcP_E5TyK7G~_pap#A(dp%uE1UHiiHb&O#jdk$onXNcQ%}8@(diZC1<J{t*`kGNi
zaFKUldcJPCoPHV)eIM_Lu2<~wIA6m&#*C_jBkqs>kuP0e$9QT>--DiCfs7NTMeLG%
z7o{vXBd5tN{b#tf@}v?D1#o6|y7S~?j}^6;zy57n6eRTjcUq*x`_1q$1>%DLNFJY&
zgm>fiTjE&h5MFLn$36R$UKKbv)Scb6gLmn7e4W_3#~{5o{eUV#K1s_nS1Yj>?g^yh
z#Q;EOueA8rWg}o`elwC<Ey`U>{g&hYfpoTQY!O1I0r82;+6B`dO0-EP7zsD^29nm7
zQAHa0&g)s?h<<mJfas0`4Mb{EW~kd(-6<p2W@3}EUZuE=Nsd>$Vsv2WX<>2CLUGvS
zaKoJltLzvWUd>SQ9|gj)`p%j~3yUVMnk^$xBjCKB`^8o7gDhqO(mFTEZw;^tX|8#@
z>r-Jf56f=Z%o|Z*a1kUdhO2&1Er$VQ$nN`X>3T#~@k_xu6ORhgi47;`KYsLl^tyKm
zTLB*mRgXXErG=-(6$}%_AHX$Y!04b+>dK*^IwjQ-lIJ(&7C+DPw|8oJK{kDnUtWxl
zc_BIm2~x2KnwnFu!9h8bCzIOIh&qJ6rSN3;c$%PbyD8Z%E}AcWAj?e+45+-yj0c#9
zOZig_JL<MkX18Z!a763VFEn-BSvbr2bq?w@E?77NO!$rTn>Xw+I3dfk)Ga6rr)^(g
zt0wH(?*d<#KE9F^(beg_xNspTrr$XG-7D|-`4}2&3;y6pkz$ID(yq{k3d+7uy5cG!
z>`6%Yn5yR{d%fBjN6~MMK08kz><{Rf3|zr7{jO<bjxia(|DLJhd1mRbzF{fAayF9Y
z<*@ef<%a*nxXn<{eqos{OO36t07_VcA2-h7MQy4!vh@`g>zivO$ddMjC}0bxn)74m
zHSG?1+HHgF)F$zPrlJ_Ybt0Fe2~HM_aU)JdRP|U3_xSFFNOs3^zGcp$&waY1?UT_%
zN33eci1H#RlS6X5dXAEOMbX2w)QHbRpS4#AK;;>L%;lwmXiovz4=j@qC9n;9?jZ>8
zn6@g`5Cu8e@iQXwbVn)3><)`Xi;qTa+N{Pt3&nr}pbh><b^;Yw*aFoER-r<5ElKy1
zQhOwzu*Y~<g5FqX<Y~~WVmqBidKH!|ojv93z$!(qPy2EMn|tZbjixd~e!1h(892_m
z7Ed6mO#6eJ$KXdG@gQHnLe6dew51!`rd&5yZ49hORiF*aHOoq99j&Gdg_ZXAetAtC
zH71H{WS1k1CrU!7Tf()wnB|3|bz2Pi3!r8?p?b*`#4UR>NQIVGhZn7}m40FhI)#Ke
z<NUC%WwJ-?TT%pQ0F3iF#mMTuilA(Y7XvWcR#+;ge4CPEeC=sk1kKB*aSYsF2KgMu
zx#$pl8g)aIw&vgLIeoeJ3B9_xFE?&a0bfX5sdmSTl)%3Agkf@qCCa|C*E!Z<3`&)h
zr3Stuyl0}&d}~JY89zCa_hpOMM}j_mwafzi7{kM<3GZbyyh$Ei$tdO`CVg$4z6@<G
z<~n5Y+{q-1i4!eSnA{hREUIL-Xc+n+J{<OCJgqXM!vo12be*tk0#EIUqDhsNpr7%u
zB>E<l)!iI??yZTctqvG^j%>x7BtVUY%zSiNoFOBn-gUo?J}qYd8cK3D$2&hZcI5&5
zL4I@&GHA3nImyoWE_4CJPPU|Fopy?5O8iOnCW;ya(FY&|VNlukr7A2EfUDbQ$6rU^
zNs;IoTqwl{2juVenB$e$?pG>O>i$b(hVr5QL<moJQgW`0khwkh{=$8QyGw<TPLaH9
z(*YUK*jp-KSLn_N&03Jz^C*o^AZIpcda~W*)$aJluEb$VxcGOkk%NMsR1>QAJJY_7
z;`3ea6AKFql}6M<8eaAD$n_6M2V&S>NPC)C!C38=`Gc)kpQwf!U>2Bew`?wz;3Icn
zahlss=(x-f16`pO<M!7u%Djb6PP7#&g)>DrR6<}QP~r0Vwt|7>812%4<s@^=1usY~
zTm<k|{7Mz6pRM=Sppe%pd`kTGxuZeBwEIR&e__|3rTcjPiU?D(oZnySQ@kGXhUiOu
z&Zxe(E}NzP$ulIQJIiZrs{Ds_PXfnyX<fd7SGWVgGi6LePhm0VcLp71dw{%Vwf5kf
zW$SCn>d3Q&1+gLGP>xrl7Yia5BPrn;X^h*^s@eV7;Ed@<Nrp?e%}o@E69!&ZX%pLd
zGLBtFEBUdyF{1QU6x6HpADl&?Z(|xV4?K!?t77?hichqlQPB@5{AjeFUFiURcOQi#
zZogBHn(d1E4Q;pdb!E%FscpAa^eo}7SC&IB;Rlyq!S+_Ym}T_R0EfhE^G8oZ3$)YW
zNU4spG{$1!J9T>a14AyRhWFFrixo35b>HbEJSgcTMY34Rv$GUKV#}uxa#JSNBeghS
znniY9GwphHzf2KKEPHU*6=1o4t{kk7CYJ(Bien^7TcJ_`A@h54%M^}%wiFK4YKa`T
z)B<FplkC2=j{>Bv9>PoP3Jr^SCw9gSlkA^dQN?dhK;2%WQ?ujikH=z!$Zi2ZZ7FK3
z^5xbVP;oNe00kQ2JeHOTm9#7%L<`jIbMEn7tQFwjGOR_CRQ%regp;$;jG3hG_8Q&t
zz`ddCZH;pvEha1<)NA=Pc{s|3_zK<6a{no4h977W>ni5Jgy#<K!FK>Kh~c!kYf4!H
zPJC>BBJ1fVmp}Zdj0ZFG_nO!{0@yM`nP?4D7}_fiDD|;I*%5HRV?U6fZm|9if5q=W
z3x$7fn1O5r;(CgvrssI>SP1$sl<)!6D=uTlHn%&!&P<${E)6Xv1+PNxUVOXwSuD10
zVT5}2^Y(reW~u*QlR>4qfd<T<&YBSV%0Sj*?0QOi@Hc<!t63KW+RR6BP|~!t*<}HS
zQ3ERZFi^;Gz$q%c0Q?qG@!npZg5E`Y|J-Q2CM|pbXP`59_eU=_YxZZ6k0A(U7uRsL
zo>UahCVoecQCWku=(7vke?1piQWi4lMJ{ub;&T@1f>+N}%mr1G117;ZX#%N|n_B`c
z-q>0rGQwR^8c0ioJ&*B(FbJK%O2@4+9ZpgslW9e8?YYcno+?$%D@X!P2d&cy!*3xb
zm0PzcNMiO;vVr&0Pv2aeqd71xWGf%Wn~Ks_PKF+D&dBuU^y)JpRvN4gE6#@H!}Vq>
zTh};y-w?8d|G9C;zkXJ7PT7+$KCU*j?1Qzl4i_wSLFxCt@d#<P%Ce`1;C*DAOvU>F
zz6RFSCO8?<5kMvr6S3$(6&)}$ZRgdavK_kHUCyXQ5Vvvx&wS<UF>loKP|n*zO-pGL
zSrbOdagkqYH#6@f??=&C#vEs*FW=vo6fv&3Se#W)fc_a((sj)n5iQd0X^F!bRJ9R2
zQrs2n%P1?**pps^oqX4sB%}9@2>#V}e=R8maZ3ZCUtSLKh+1meDWzI(AIU<0FW1Z{
z84-nWd(^SinbJ5sYjAgMr{%s}1A2e81xE!}usrqlCwcU3JvCT~Dy@Y7woiIn1SW{n
zc569O@BzcLks8n`t-Z&oGhXo#)s-9RCb3X84PioMweozOj{SR)Rg07?-PfbuJK|oT
z+`xcS{9%ie=MIcwG3nJBeO4T5zZ{P878uPl&OP#iW>!x*N@bRjg>y!z#u4yP-x7+<
z73JCM>5pPe(<kahxH~$kafNX=C$~WXQiKnqI$Fki87TUOa8`sWq~HO62+s#diIrYT
zZqD294ENF~!dypw#-=66TwI8cw?nc8BKJ&c4_&!?zw&RETJJp%a+K5lrYtqAHU6m8
zXz3e251)TcK_`Ub*}M5-^~f*&RRfm%-(+5118mbN*5SHGVr=ey53A%|EHfi#EOam`
z`q?Dc?_AC#pE<NH5a7FnA&3SjMh@lLn;VrgDYgApHSZzV=j?i4uQ;xgZ;Ks0>}pZ&
z&!E=#H@>kR1k%k)&hGlv;tFA(ya$p>&C4w!3etSrnUa%h(cOQHLsuICcBQg%K=CuA
zEjYkd$(36ehPnPZ1H}k`b=WRL0QAJ)!+MNR&|~7}O*KI;nCY1QEbgYTT%<m#1ER2^
z`_#9vv);E&=Dd}ow$l+lqbt+=a})jN!{V&u(80K#i+@3!T&geLby_T%-Otb{!roM>
z35U@+?5Wmyw}7Z`oM%#V9C0;rLW`7Wp+POx7ALR&R!VD+64&DevuY76*z^Oy#d=B~
zb#6M~>KNpGxysdT1~saUzp=n_Y;BD`3@b}z1Cs=4UaBBa`Cy#>=f8JUBrAzhZ%gEZ
zddZ@~w`y{!kgOsaxJX8+dxS(o2(LZgYGoS$-seSA3%B5BJ-M!u>S+*A;<+3P)w5kA
zKHU80hWLy@tcgNQhPJy}Ux50Ru7-eh`?AgR>WL~W5bc74pJ4NnbWsggqAp>-;fS(|
z<1&z8k=&z+tXMDOl{Y~AAC@@pl)^vu@5DPO`@Cm<G5S&IwUVzz^K&QY=M)>4?&82X
zth)R28`dL3Ymc3Hey6@q5lhc=R;X-HGuS_SlTtCUopxalPlJ|f+9T-qrOLnPX-iAV
zWomiu&B!C^rJ}bDk}YRh-i+|58i%kMp!btVTSPU!Uk=23boNLK|Do>7iRw&E;k~VW
z(!zhCzN&9euC&Pb@m^yUs+Wg@u_YSE&?H1Z1=$+Z-8FM6SC_E?7%-(@zS_C<myL?`
zn7y!ML~RF6)f$*SJ$l}(`@~sW17rMnG->?)sj?3qp&PsZ`XaFS+rq41aW?v0-Lb24
zgNxQ+tcrq$R6tvLT0^*!oKE?mwq9=!rI4H?ujEa*J<pn!VF8JVY@!>=!t+c$hPplw
zzgpPd<cgdO{nuK0Q}*blkI}8#tW4bq(IQCssy@%tGQx^xfcegG$s9xE5qEbK2bBuY
zyllZG@@U4zP4VZ~c?Za3YJSG{p^c@H7Bi!(t0If3sk!ZYetsm9r|y1RTN~>_tk(E?
zU*uMII2pZk3!h0XJMlSbf-1)vUA6_!_>~&aAnWX{l_Z$5Z_7O8wR8^dbu#J0GmLuW
zl*gfXFqlFf^Kk)W$IG(}XF`)We7<`zM|L)H{0PzHYz*`iQ>5tM7K1F|S70leNOD7w
z@@<cb$Jmyp_-BHSW+Ev_H5B*t`}T?x$Nh=}81(b0uHJ$A-!*Nz9Nh4-LDd3BN=>@^
z+4hz;U>jlW@N$@qTy@wEzclM}sJdn=VmH#8aKqR;wCHz0YFqqG>{w$7tEH5t-Z<9T
z#7Xo^B}%dujRMN?V5;!M?Q{ukNvFew)>Fz->-3@M61)g;vOCQwRP=W;rio^n)-IxG
z1PjNp-Zy>l9nd#!q`Rd|bzHmPJi?28q7?Pd4eB=8t&z1cr%Z4qqa2U4es=K)<Dk3(
zb9t4W_G>|kDGgy6Q^Ys0aGUP_msp?F<8wzHU8C}vojJKE(M=s;lM~<{H5-f_F34n}
z2C*@hTiW7*Dmav*o2AZP@Ex(|QZX|YH3p*7@Q|jeA!-T$(Sad)az1(~^DwRKd&`~5
z`9>JpdXDyCFC-}<eSX)s#Kc4Eqm5xvAJ(W!rTgst+0@t41Sut@{mp$JuyZaIR>-Kx
zM%b22qX@u#Gt-FV2bMp;kNR#O&>D$#C-M~^@?)Mm0qnBkOEvPM81+%;D-%ienbP_w
zYg-q9uCR8qe<FWzMO6%$@T>m64=2o)GKFIV1qFZ`&RP(OT`VmSqiM(AC@lLsf37*z
ztX?6Ly?244d)9+zO}MV9TIf!6Y;jj9(KV2>nQA4$ML&J-s8CIBj7<3Aw%qC6B)yU;
z;1gi|WJZmJp||nOo}QG;Uzbl`;<#;0#0ys=g^UW;WRt(b6({2a9vZC^Y^Lm2>ex-j
z(^EL5+BFp0>N!9%f+FGq!Xz!b$d1_wOQy`;U-A_iNnU2khh47P)t(aUrkS!M@w0x-
zFq~VPc=g}??GSd^Bo>-IOz#uw#DntRd9%mwq)WlaO!e6kGJt`wV^Ow%ZtM&Z=gFY4
zQa0&y8W=8>*-#<fgRyXNe+9Fbxw+qf_0eJ&lvvf$y=xVJ0I5Fln-iF2Q|p+>xTr0Q
zV-7UfZba+!y>LtuRtH2n(f|6{Fq?_;P7{CiNM9BEEGN;qr>&St*ANS#O`-q2)&IU;
zID#QL*)rvGfJToAXM#wW(+pLoxVUbRKdqtNW|0J+uUZ15irp_)w72YBLga?sWTiuF
z9J$GXI=oV=Bc=h<JBP*7{e}5;?y(uriO#wC>9zZwzeU>4xM((7+@#iH0d}FD4{h6=
zt;Llyob#E%uOLN`1WuBG$6bRDY1WKa^4oeE7T#?Uvt|!3xdQnd8Fe(J+C>NrnZ+(V
z&k_uJ=giP?S|twsWD`57vvUi%6n8vD;5IokK8V!f%1V<uANi)XNx1fuAO<}6xlHV6
z=ca+SyhdR!PS<*75(r=ziPP^+a31arC==&Y6f{Y3O?7VN#SDFlqUD9~k-r(GP}h36
zMax^1&E%Y-_oZ*txpE%en-Z{kI*(kPs}#x_=`qu9LN3xbp9Q;CyRnRXSa|Ybk}Vt>
zP9}m#;r6%?Y{>kvteu2&Dca&{^qg1=G}sBKcv}6Hv~AxUMWC6$X4uqon-`A^!8iCY
zjx$il6gt|`jMM~iMtx5<o4rrD=?M=YBTl*BgPDuFUMkAm^wf9NK*FCdlmY}ci@?Lr
zgDM?rRLoi4i*J94wY7QfTZLV1$7GL#Tn0*x8}8cVfwm;c$GBxn;u&29mwfkuE~vN8
zHo`4@zZ!-2aP*W3)M7&V<IUSLy5Om1pJtu{4Kuwd+b>ZDQF4RyztUudE_TWkRrco(
zHvc}citX%^ytjSb8ul0_dkkR!HZq64K*vPdZy%OV9EniXBZ?;-X^K(azTN40njQMW
zvIOyx`4{Lo!6&(%T`3vMYH0zdk{1kym0)P;iUhj}?eX__X9LRik2I6A)6+6SC1&c5
zf#15ITID!f%Nom>Sj$5f*TD`KcOi#(S{K+56&>8;VzCfp19el74%fg;@A+=Jjui8Z
zCXwNsCI}MW^lJTPMI{yj)8A^P!%OV58V2`KUXdkQ*bC~JZk-F338s2}$us4NQHiAS
zUi}Y$)P+>5paR7^6AaoE`@7TJy4q8v#0LP>3HXvnUIbXAaQ8ThY2aB|Z<bRT_rU8n
zDK$iaBHI>ZvipIAi0f_o<R}?IGJ2Eo4;raq$yIiy{s9Ee?;rs?=N@MLbB79=@vL4g
zC#G8Ok&n$KeHZc~!qxkjVIMyh;4YAtwz?R$wyKooKEP{DSB1^Z21jvtqnt~)jVM;J
ztCM~RUO`ZccBXJZmm7(qHO_b=4?#@7s`0ik_%xwafAnbeo3=(DY}XlZz#*nt+)=N>
zP)~SaIB>_Y&LoeQ%RZc2q2_8kz`7?bQd=m(+r=^t7IOtw_QShK=qq%Bm66tfm-v_6
zKKNGux_cFW3=0m#^U;dAuZn=JPCr-Unse&hoXxN{;hwFLvL!LUw5=_t<C*8C`Bf0i
zxe?Uk*rs5KIh(}~c;vKGl3BGhp0Itk*}O)u35lD0n$3%EF(8E0Z)GUGJ1P3xThH>t
zO0yz>+o0QD^Y*%;=F@_;xppy4hBkp{>oKG46qrexe9T_zMbImwoXN~Q3{1h%fB968
z@R0lBEHvCBopoaKW_4uoH=qYfyvP8GN*xev#(1{{EZ|h1eJ#>^j&{B$D99=akq6e%
zlLs3&UBqd&`g#0#bnxr{sN2i`p~CL6|8v9K>P+R~jT?D@egkdasN~HMz{)-P8#n%v
z&vy(@`TNn0%@eZak2fe1-iOa2jyC((e#VyiSyDPTZXO-p_MF6@aNS~ZNndizx_RgS
zP*DF*F@L?yN(fy{y}eb4&Wa%!!}VcB>vj9UoU-B(4ZFD97quoW+vllfi*qiD7<k|`
zy#lhpIHqu?$`~*Z&iVXk1<Kp6tHBdYpY}(`d#;5m`|Fhx)l2!4uqCo`hiiWRtduGp
z8|Kv4yX>f2{H1V66BU{hD*|9U8IjH>$;X(uujvk{N7oi=oHXc>t=jd})CyOgA|S%R
z2G%aEI=)J6smyTc+1~e~6ENF4izBlfQ0%#Q6}<Mq6(b`)a}Wkq4En5V_USUd08(W3
z>+}2bQWoOwiNt$YmOcM@6N+(DHTdzjzRY_Vah^$wH#9?0{Nh^Ak(X3WZF4PhMZ=CN
zMX`bJPTuOIrRG(<j5T`J>-`(}PQ#R?c|pY}9iyW&%GYK>M=8Q8h2`$w=T(iFZC=dL
zP#!gKB045Ic}JIU+VE+;05<#{405RP2{dspd!&XA(n1_~DQHOZG-r)?`6P)(XEJLI
zC4^$r{@`sJ8EiKz;$}`QKmJHRDO#FDfO-x&F%gh_jgT@MfFX$~dH?wR?nik;e$7CG
z1|NYQzdf)@)KdBhwrJS4D0=k9_=tR3?6%P)`R`pjMcp)VQ^)P#9Ic5MkYz@?*1qoS
z{L>h!e{S^jDycmEs=?8Hig9s8)FKkZb+g0eDCm8X_HC4D7uo3QQ*BR;-(a)a=!bnB
ze;zm?JH=-5(Sd{Y=8XrjMj!BQM4|Nl=7DOXB1u@;RPw1f%mha9TG`Dg%Xt(k_~1le
zu2AM2niszvs-5QYU`L4DQ9hho9|NliRwtc*Qq<|<<dH%fTXsbDIqm+mOcRHJvStP5
zAGuY;Ypey++)Q$~SX8+WEB%7Q#Lv^mf;WX7gZ1awAi}au5p&}E#5O5|$$mhZXLD5B
z2j?3@mU_o5@0(L=YtihY+1UFf=8y_0HvBxfFG-LakncWQmtv^LIa&(U6MpuQ#KBu5
zWJsj<c)1hDv?eamrOQ1r;4IJaH^6aA?oTZ$#fC?AtadDOaV!~eK+lm9;!fHZAz=N~
zKr_@Nh5pME8{fJky4ol@80(dB6Vb#>RqDQ=W!W0JJl+P0%-Q}PPG)b{dbjem$<L|-
z1wj)JVWOs23hdxY3*BPDBw$Jyv`EX{Z(e3`?mQIAn!vj?`{oSB1&qjpMVXx%L`hPc
z<WJHNWiZE{OrvtcwV|DrwpTRY22xxO`VA-BZ|$VFet7WM`U+@`vEZ*a>8^2lV(&QB
zlY_AHC2r#>*t>@L!X|U3R&^$AdP8T2#TAgxr#0(M<Ing-yCpKZLgga9A^8FJWmdpn
zNhN{Gi&!BLnPcSWY!df%*=Hw<%HAtrW8JI~$1MAdDLOV1Nfwn6q|iHzY;z4@xcRNA
z=<D}a!aiRu<Gcw>IvJM(1#{lf5wV)~$PJf)a_1bLofm6w?(KhTpk8ArUUrCMgG$*)
z{ea&tgrGn`J$>P;I~i~BjOxXDjh?q<onGMdhkQLOC!H-EsC-LK*{6no=yDXaa8g!1
ztNH>161*P1)P3sAf=}6p)juJ!Ewslr2<(9v59-Io@C;NZuf}+P1vcEeneul=iq8_G
z#m8*oQu|=NEe8$5n77E6Q-^UIH{%LeBq#p`@H^(2f4)H*X$PWQ<fpOaRj$-kRB*QU
z`xcF^1{8;Hc>`-6*E##sl*6)1JVZFhWE`iLlaz6=TA3;paTW6FRi|mZh1u(L(C44B
zOL4yes-)`@ciTVZ9N1lQ&RgEH!Q-x5*x@KSo~ovu4sCng?~@$)fs9Y+UipHIYrF4z
z7%+*&yj$r;EwN}9TUD9H8)g<yyDDgGCcSuxyR?jv!Uzx5uk_n3q<>rz1DcA2e{}vh
z-Ktcg)Fu^oAoq90oJ#LUl{+s48x~*&y@_S+>q5A^C<TKOV#SM+LJ|FH6|uwWQ1gr^
z8n!#NfhI3}Y(QBDAu%t5X${>0mP$6c5RATNf<c_Vx8Sn(elopdlxc)gdsv&mS>z5J
zB)aK}&Gp618r8KvmkzLwAFmXuch9GH)7kASGBtV2^>77RdB^Nv3NX2DhQ8AnDM__j
zNL<UgqyDi(xe%!UQ6xzj_-F4t?1WuP2I{Ph+3#Ch^qHuJM%3QE3VhJpEh+Ia2`>PC
zmKp;sor2rkG!oc3YTx8Fdep5yA++ji$YaUbh|S($-n(UVbRZ4MG|b6K2^UnHJyT2G
zSI|^*ZhIZ9>HDLw$q25Hn_@WkW#PcvDr)~=_U=0cneREOollx*YxOV7*FY5p4XDJU
zJH9cAAG*q5<EAJ*Mclk=<ZrOqHh<1hgq2gF<f4nvujgaIWlD|W=Xg=BZdpHOYhA90
z68Cl_5--5@ps`^#Dk`c%K&yj(Y%2{BrrcYzQ0->-#cg_S(^Yfhr5@wNllvPd2hv55
z+$a?hDmbq|@U&rU1Orb|=FHP~2lk$(K53Bos$fcozTsHFoStgeAQ{l=V}x~IS2Uhr
zT%;6%Y0Q38d^JrKLtzUu2=ZoKM@}NQlct$^4LR(Z@^+wWe&ouM`<(~Yu~%uk%pFCc
zHCxO*;vT7KJmA+EX?Wx3X%QNv+b_3DRD{Y23klJ?VQO*n@<KkQ)P}4rgPn^B-mi5h
zZnHX5q@J9lI3^)&x=gwJ+SD`c7?pT@f7k-D5)rs54cwGlP{M=y7<*r;eS)QUHH8Ny
zts9{)Z1ty8Z=@4G=Uiyl*ROVBH0itNa>35}(lv@xWjzm69j$uwCRFpn?eMJVth4WW
z%H`8RpUEqV)#YuTHCS1QhMBQjM~L>s-zje{f8)SKA%52hLxD}!GymME);tI_*gci4
zp^DoHQ2sOPSbs{sJaX0Fc>dk*8oQu$T)*GGbf0Veynk<RFXR@;CHK#p@!XV;;thHA
z;jl_*kGWpWau>pWUZ)TrfyJ{Si$|7$Z$-V*41#<QJdpeL^G7Lky$mdT==`1JofBlj
z?wQruK)>bSglymQyVgMhk=E;SEm9koM?y|iLD=m?)D!!8!QU48d6UlY$-+w4SsvVK
z_avOZP9cGFdo->1Uc;9u!HckD0{6)Ccge~J4pCCuQ~^B)D;j3_momo2#!l5ixbJ2P
z(k|BeBYToGl{vy4ZkLB-d<!SoQmoiRn5&$`?Lj?3Xu2(7bg^KZ<?*OQz0X84;qau`
zEFQph+yA`bJ6`(9SxMDu!1a^v88>R(RfO&)r#=;{i(w7_BYl0oeXV*yXgQRm6d9Qn
zx1awYD;rTYe7$h@Z^4f@A#e1v6_n%$!-fSj;&b;wGrwM>(}Hrt9t&d1n=I^L9lf<j
z?FFsu?vf5m0<$J%(XzSGXCjtgYO=d8f)%KzUp48X^;}nR3-;2E=I{|$!%;cd?0uG6
z?ovgI^bo@x8Dh2Enf=+?1a;jZZMF4%?ywryM@c=*aWl!tSF{q1JLQu4rCJc;ZR#Q5
z<S*sJM8_Y6A*K(G6f}FT0$Oke1N{iWqV>7h5^#}Wg?4XBvNofH&SIcUx7Jb7US|_J
z`XJb0R5vGXN^S5Qw(mL?8*1G++Q!uUMDxBFCGpn>^N-nxUKSsx-%utY@!iuQ%MrYt
z(RyyXW?|PkD8m;@E}W8NEYxQ|4EkcGG2{}^I%hT~Lw{*t+hVEGJ1KwXW(l~<lUd__
z9=nTroj6~`X~2OE{V=p*bZj_~us&UCtu_@qW0A_tS~GV&K6drz!Moxo96UA&`q?=-
zZ^aGcAl{}+{di~6*-P@MZQ#8ZnX^Rz=8aa^+-UHk6OWI8@;r^_KQ|hvIhgx%z`a4%
zbyd>Bfp2>BSbIGO)gSH<_veOZ>K}kE;k63=Btii<?fsGzwm8Z0x^TCwtMUDB2j|84
z$ac=2aU%Wak36k9QISBCs8_X=LZkWGm>hC)|F1NT_$Zb)U2JyT@*FF^Sz~YN7qZHi
znfvA*71~A*Il(A$;SdG86>VRLli<vmI~U7A$m!1`t12wZm+w<fJf?J#Z7eJUap8#z
ze-&@;QB?)j9W^>K0Y*G)&D!H*TaN1vqXzOt6)**m9TNSJDSpk5*Hu*~EMXeH_A$Qm
z^4SjI;ibAGkC%2vVUl{mRGFhU2C7YQ(*Ph><4C(Fc9WO3NJc?&uE%Ciofuz9ye7j-
zD0Y*r$|*9bg{vlJW@V$p#l@c_m0pU@T1Z^^fdZM`|1a*|GAxd!?G_~@kN^RK69^F8
z-8}?%9bAILV8I=d;4;D8f-{55;4VpU9bAGt!5y+Y@4L_Wp7-0=b<X**uk&O7ny%`e
zs;;Va-(9`B>RxFuM1Q^IAzbT6g}cYg%~6rpk!{@6;ly1KZOEjrMBOjdZ*vv1B@KD}
zHuH79D!8jf4K@g{C{3!C70tkIu~0(dej5(o^t$cA_s@Yi$_&2ZJKZ69L%d%&2&`h>
zrIrm1YB5LmGHXQdA<R5#IcE1M4HA+NA>=+Z45FDWMAhcJh~LvL*KBNK@aM!jMJLZL
zv@4mp1^TtcUc&k>G3E7aG|IF+CdR3ZhQ7=Q{Wh5@wWE#WvQB8N(7WLr3?;QJ>a8@?
z0YO9UmtIcTU3oUyHNw!ED-GhWQ{4&Fch3iGGr=LlY&1@UU%I?9y*73Q{vgTi-R2fO
zuuj+Q7rxOfw3k*coM0BHp{1ia6pBo+RZ7;EMooG5B_1lu&Ce4-sGo$&V?1p}b$LRy
z$85S+-QLsP=%RzB#&{i|AksSo7o^j`_pTUi2y7i5QW38^NW-Go_3KGr(x%_Bb8>I4
z02B{@U0G)Ec?b0#^cj0ieobB)jRZBq2l}vPJUW$;v#3}}(xfGMdJ*>XkauZXKvc!j
z??pdNl-AQ(;+li6C?r=R9Fxwu^Ob1x(XpP6A8Nz}(@<+}fC)Uzq@haP5K(@y?O~QI
z4{bd&8oh>U;n!XcQ+8~T`%V7V5pJYmC!#L-#u=s4mRn5f86+&P-Qy2To2qiVq{8M0
z+HRYx#sf7M8$cu9O>5>p6E5q0q(DDLorFy@UoPX#<1g?*^c(LbKm&;+rmQRjFs2OL
zmrs*dS#Bkxu!Kc0+zY<O+N#QdL|JU{h;6BAX@OUYb)m&2j{up}1T@v8)C}#?UjsCY
zwh@V6mQ2PMWFsHhHRl@xAWyN?ZR_N9tMdmZX-x;-Lnt)s!+(0#=6N`n^CsWy!J$99
z14OO`^z?^N`PQZd4O2H`rVPS}P6&^CBbTUxFUg9clwixCImFEvqqJCmngH~~V#GN$
zgv~IGrW<EcY(HaA?9OKYM0(;9Qib(${DN6_o<F<jLnbF66LK>jETU3-=He#g)+S_Z
z=4@GdObdjpVD#vP%yxxf)VabbtLwo&kqzIJ1P#ly_PGht7l|+AZ=ly>-^yYvv1k^p
zzxz_|4+!2uToI&Y+d@B6K!!sHzaB|6IPB0qw!C1_qzBogg2J>_NCEvImq>Sg2-Ej^
zXmX#JztU<oA>|aERADWZZ6Ac3v)_5rG&Ft5L0rVSoiX!=z$lS9IJ&j2P?C7f=aDfx
zG5%0PMcefi!fJ1l>okI#MtAuTm1kkK^DwV3YK}LnK(_tJ#8m@&4CQHdV8U3F3Q~$q
zDDI9R(t$^ys~Eqj<L{laGnf~w8Xra28I+Y;<>_GtI%9OD3^-Do%zUQ=asK9f1@>H8
zbu|VNYh6`MeUr>`oi3W*><iXlk=?+TJEa#GGV|Zg-ckFSG(*%{d#fB8NN~ut>c<~;
zJe2*_&61Nb%qB4%zsx%oi0w%|I6*_ZdrYmw!Xk}D<b;dgVnS3K`qMu@tK@tAu&V5+
zuc&Dn5E*wTsd=ADWmrkMEm}VypMvxCN#?Yjd?{Ri8^*U+YM}^&R$JCO^}-KZ0_MWk
zYp4wB6qM0fB53A*28dEU^YLYu+lgK97cT#yu(Y%AUPzkR8BHbFgad{khWnJOl#|7-
z%D}YfnK*O1rw?-bveF{BabhnPw=eDeKz6*D+u%GqLG3-?En^JJ!e5HjhFgkVAt)8@
z@T98LX<j1>sc9`R(;N<a^{eS+`Y#uWy6z>2UaO~^195~U!=RVt_gp(#XSkEBtfh)C
zQpQOp<h7${kL>couVbKb9fS|OvawOB9*{c2Sm2v}g6jl1%OS8po%xAjN-ucH4ZOs*
zVB%gQinL;_#4%eQoiuQ<fI~&aG<oRe0q@usp5OXO1gS(nt8>f{^R1K)a$iUq1Y49?
zj4MKLTa=O%<ER*07s?Vt^%5J1`SqRKlNeVVY!URQ%V8kM`*ICtm*Z8L)O--*h3>0-
zyGu@U&8=1C?l2<tp+JkPE;}=@Xyp_L;i^55{g9DjzgRYuzqnNcHs)zrT(IM<&TS^m
zT*xb`p-$4+f%=z}*!6`5ynNI=?!e`^!(${qO9h7*VEy*hxGj1I6%GBNlBHH>H^zLN
zJIXN4jH`m5#&emLsJ>$P_}S4-rG9H~aJPu%Uc0V8s;Zv0+qGF}Jh3jBzULxSl%Ja!
z8^2(xwH%zqt(*!hp*ACqZShhoGQq{R!=e!zFP|}A@Tkm#E{~>yGt}Zry32K_q+>Ia
zvO<4GQO>o~lzGUSQ<T!pfCb0K$zUb*pqjib>q(NT8VKZ?KT4PNqe_VegFr{9=N8fQ
zUY%JuDAJPZ4-y<!pQVtiAXvRA$&|?8Ze7qN=xp(wQJKf-hj;lU;w<LL+V0|$uOF>M
ziUt)%rar2$^fWfVSMi1GF*Yso>n4;9464wy%5gqSi+D49ED|Oj%ezz!w8n|4O;}3u
zv1jU|gUWrqj{OF*ovfQ3Hw9l-t*R<YCx>7Ei(cYl`ZvAAn+|p3uuEBa&di+V@#78s
z4`1^)oTp9Bret>xW+;R)CcfANQqZaL$V{ICIgr-mL8|80mtpyZO|-a%wq&?T<02t8
zdAAq^m3_|@n^ndixyF}FWN!sx3)c+=K|Pvhbvs^Z^qaKci8xwn*#y1x``%`y_ZA>O
zT3}s)sGG8fAX`{5gAYni9my~koL#qECPpzUB|8XR=;tLTm13{G3>RHNK1B&^N*3Hy
zQQs%cEs@rWRPQSL$*VABp}A7Op}b5RzM&6KUxFX`*wxLtSjfaS@$_k}j(*aJFZrpU
zq9usw2-UB*&P|bWQ!M}&h1bOd^CU^d#X=Du-Hi#%WZst~V5p|Ypuf>o4&jTB+bq(?
zEdYTI%)9-^SvH0)r>XK3*Mw*l!gxt|(0nqR2c;qG&;?2J%o*m~ONcLThe9BpVP>?H
zC3n~5pp4oXcBuE)8t0rh6?VlKdr1oAR3YxdA1YQ#L_rPAe3Qdd3G%hJa6E@bc<~1B
z+N62ucM^-~LjLyomhvK}-h735`|)I!arZ!Eo7)7rzKM7jS2(&bCzV!V8ogE=ta_qU
z*f2e5+RygI$CU-eUX}jeHPmp!?1~-|?4E1=7M03=#6BJ#&rTemM8Q>K6f7Nu%+g$-
zGNAni*nB0Z2|66cy)hzUoEsV_&&aZGFk{pEz3^qD?odPy^bJ-XFnYyEgpch#gQyMY
zx)KGY`)i~d^J>6Xds5apD5hUV(2Ppm)SdZDj2Wjgr=v_{6Hm0iDbgq<$Ye>XODNW!
zaIj6F$Tr=pf0KHZ7@L-mZv0)ll3av*z>MNH51%%mRn$Xxj<^<cU~TW&BY9Q~Co8Mk
zmGOi~rS(GX620ZN9(2BjMjU~wSLRVmbS^kuTLy)+#vM=%R>*K;+HLV;JNuHb-Q^()
zmUzP5n8ycSd)z0pH9y<;F|W~-s)O>9C-c}Ii4*0e&!e4F7b-%2zr9l}Ug;H=Umi>6
zF5T(C`;hrux<H%W=B#=TpUw<*Dpx9Bt8nk*Ycwzcbk0bw+A;%+t+gd<u4z7UMz8-D
zvA85Y3#n0Zr5pl0smV1~WXkn&zW)+SJ{9}Zg{L_y7VZ2DtbSCgnf54+>~_>j!T#m_
z2bR_jGto(px#X$AiiH-#a@}odX?JdiVp?HbE&`)M#4oS1pyZmkwwFs>3O@X#qDr@`
z2PeguvIpx0HbU~<Q$DmVrN|Jiy!!X;2_rz{LQs_f2~(kvi;Q)BXsp4Uz7W@5^Pz|K
znZ0$eMKv5}O7JEPD(^WoILu-zHdrq^SejI>G7qp#Eqd-9%9na-yB_Z@>3$K8+B=o3
zSG-=UU|*iWROqZUU9a;!8Dw?S(qz8y^{A$^vwR0HnQZk>Ic$V*fRjre@J@wfv}+xd
z?^fm|7Z&6mCvL@)$G2FzX6b^Il6n&ll=>Hi2KKF&_<>c${qH|*Lbd5PAC_r^3aopj
zzQnn&b|b>4_w!p7pV>r4eJ;ge=DE$%a1oIwp1?qNT@t1VBWG#V5fw)BQk{`b0n_gx
zYO~xx@D}wtX6qvo2coV!P1Nf5r)>vdg2D)%A>|OQDuQ>8m2~Bb@S_D#x?g^y4Wh(a
z#f5A|NU~r`d*gaYH#B8pTx6j==hFHzzR!5@db#@RM+gS?8al)dSkNqRnhJpv$-j@!
zC><}wPWzJBzx(3xeD+-4OG)Ru8`}VjU60P5V7C0%Drs?YjY25z>!yAIF8B*~LV#%U
z)sYTk<z<QXbV6qq=CznytOpD0mDiTLnS8Q8o;~}#AW%(s*PF@Ky9jEBHC!YjVl+O7
zW7NGV?k<KK73(Chp&8U8aOIT*d8X62{jXAY-nhy{3?ou%M2j98+&&-5hH@Mc_Fbz!
zH@y)NG!A&1l-lD^&SxbuxO>L48N_*0sGW=UfrhcTO;`YkX9{uJ*Ip{UPQ{aU^Lr{T
ztX8ZA)~u$t#8vq>-6S^zDI4j2MsK$-DfOTLwKS|uyD-XY;i<~8B`Xb01_D0l2ARF@
z!T8eVtU<GST3)}YMXR$r2CdPBlJ&i>;BlS<<2iDuXPAz}+3gb69Tl{S&xOsG65p{*
za~fGcUtQ!5H(#4~5K0rqz^yOY;Tp58Ro}{$O1iJeRrQeyzX{``2k$@-GJEh+cazB<
z>0(9eH(<YV1rHzj#GRA;Znv10ebMpN8qy5BhM?Ch;l0{>`PzF8AW=<gx&cFyxpHyB
zJz`LJYD$1$G!9Fud#@SEp$?9s^w*1t3sJ2c-{cESsU~pqR^ke&AdMvn`#YsHA;*AG
zICedq;3X`j%VqY*OH(DGAsq4!^6Uey`5q&_0|~4+EK7QE9_SQ~2D<t;7OXBx*0sJv
z0zmm$=jJ0J#8zM2U;TXUHh2$VK=-p>^k_OX-j1tkny95SD5Guj_I7@yd%-{$QRigI
zhk3}95v=K~^Xi1H+x5BG3XY9UPv87-UP;v9g7SPMZ29_dZw;b`u<2$nrgILDRI13#
z=C)bL9&DfW+Md%!<xT0DHAgq2TMwTN;FXw7gL?|8&w3{%D-`5xx(oSG$tX;9|2Ujr
zTE=G^*1Nr^045;Ja_oieh2m=Q`kL0HC^Y`_<GAjY+Vpn6Ur}3)x?2@hRGtW34J?n=
z7F&hl9W2yy@=bRS{(j}u=#_8Wuz%rA99(MK&vV0fDCRH3%rn`*pv4yvHeVvtaBtg$
z<reOguaDQ7a35p)8$4ursH)xV?q?eIafCc3*cQQLpUA>9Uuvq<4ILU(mWklKDSZF!
zA_DVt$X={(Yyq}y$2(}Ly3^?+R`d(aCmPP@;p3h}d`#WvcGa>k>Gr}YbSi9C;!CoY
zPboZZ&`ptvf^D9%`uEfN7&Ef4@Y%wMRHzw(1%vaBF5@}&j19kP#*^EPy}iN_=E(W-
zpMoJ&zR?Bit%IWIIHfXeo4`jVE?-C6r*p0^*I3wtIkK^OalPVOusZlTBS`h#hR{;o
zZhTxMQtm5CRYPW37(Lp2<(#QJzPjg>S`}}bw-KtJoS!E?$mDCpM^H2D3fRYA8Oc_U
z=aA|>w7#gAnf58()l#Ig^c!tExR|>M^%XphJm^|vZ>@~l3iC9Rmc`vduwSUXDU~VJ
zDv#2h_IRe*P%@=!sp3dhh%RYyU`=y9lqOR=HD0$6&~%yl*AGn*PeZ<Px?^Ck-i8oi
zD6GnRD+$aeSmgC9AW8{5fN!Ol=l6j=Za!1J^vu+JM3(57x%n5s=KA^0M)I6bP**LO
z+v3K?Ygn*C(3;AnXsKCmN^-I`e^m%tbNHOLAREG?HY=+v0@3<-pxVirtGlH1LrJO6
z(748t#F#>@YByapbTc`UIk6eVYNK2xbJvg%Zsi%%MzxhZxfNj>Gxh^AUKNs1>bkHz
zyu92jduiQ?=OfzfCH!m#nvj~DAF-ygMyF#WHrJd8VS3NfLYSc&n{s9IahJmclMuac
zCW1(xI-_LzdW(OYfQU?+avDq^RFLVV!C`@mpVRE2i1o}ieAwsmFYi>5>cW~;YE}O;
z%H4R|LXnmbRicWJ&*XS*x1~}T%(~#bv{P>KT-wN8(-9=8)d&NjC&_(Ys~*7>F_E*_
z#hsZD*v1qkIK(b4lRcy;&~29Q<ME-zDkY3k%oCEuF1WMZWmk~fMP;Iuaq+RPk637d
zx4K*JI_4W>-PfY5zGlIbcEEaX1$H#5)x<NZ+qPV=8v@($-=+er(~=b8fq&Q$=0+{t
zg2uI<9o@Y+7v0NZ{5|UZPBX(LoWuxg(B-={vM^yGIvFx)b%H@13_+f*WTV{1T$sLw
zc~P0C=y3F{<@Gm;USav=QGM=0n~s;iGN0=gXwbvXj===eX3w?qddsw^r_5`ZCigWo
znq%{RxKe@@;Y;cKMuz)Me$2!+Rw2=)wFJ9+h5Fi7R`_-FU5R>PeYpA<Z#fMX6s00O
zdBZDO5bVBsTwD?fEYP_$`oRG5&AhHgHpQB$o?^CbM)$kUUB$*?vUNYhl&Q56rYM!d
zNu|6gN868UUaf>YA>@Lh6CD#m+uG;*MI>9<=Y<h<<q?dDFi~rKzVNwbcN}`$G`2($
zafP0V$Ag43?gSkOvwRRaza1yu?RA_xFXbL*yMJ)iRK=@ra;v)xXLwu_7(7+$!yf)E
z4@WL$?zayKlsylebZ@5js>LM@e4F<g;lq_dC&`&BThw^m6u#?Xf(Aq?3|c@O$2RmS
zk1pvBRgV}4b%#*WqM}M`Ef~#0)sFLq=sI_fdcXJj``Gc^I*&6+iU_eJH`@F+jCqUV
zpLzVX^H*Q%&B-%`n5II&nk?C{QAD1Qtwm{keDhF@S`=nhVk^%<8n)h>2W#Q!>l}&e
zK>QgP!OcLMnXy$qF1}oqWwK&bXjS2_!92-BOEamOdJMznQ4etaz)#Cj7QKLauG&w!
zMky`tEA<=8Ga2jk;X4;CDT^06FwWO0O!^jD2*ohf;>-j?c)<vZz@cNI5T7hGN8ub^
z5LuB@7ZY`!Vh~q6$u#@D?6&nhgpp1oi<#{BLzk{b1V6czB_;I-xMqC}!4>F1yN15*
z;~}mI<~U;=qwe@$`e{NR-xi`WPpv)@P^U6nW(8k_-s(=((8&;=q5ayj^yKu~jJ!T^
zGY=J0J7O_xKN7x$%Qh_a^o}zoF~{d!C}=ThfcdoM`he|<n$81;P3-3uM%07aMZ$CW
z;(bpQ)7i%_F~^svfmgH=jbmf7!qtAXy+jeE+B1zhsIE(u`3OsSfU%Jk6&HyRy*V<l
zUD`;SmrGRHXpE1n5{UZ(P45H0^eH&-o*nzf5@U`<(=$z-hFzEFdqGlawJ$hQkI)pD
zr47+C#O#Wu?h^KKm#KW$b--Emu_jEr9AB~w#Bl8oo^W7tkSY6U-W%*LSOtCApY2Tb
zJ<=u1ZT*L3`S;1QGYYet!?){YwlQFri5d`#Q<O6~dx`wd0*koB>|S`pz|dk{x{?%x
zjP`?URs^~c_4g>!YI9+^IEu2(D+2NibclT6#9OHdLKLPfTIC=@wn{r({Ai$qovHSC
z%MQtZ%5Cq#%cnsIFl}sSD1#S|pjw}jVy}{FnfZ59D8f&+y<*`M3hY-4-Bb=zZ)Lu~
z2z3+kR;ad+XHd78GBTr-^S|m3=ART<&luDhFQdIK+f8`Bs>B&w;O3bxn8v~yR?_IA
zV&zEAs0WSEr{&X6V<)%#utuzwY{tT<EfrmrD;YK~I3I0pKXqQHICq7q1;e`_hQl@l
z;AnHu+33(3vJ5I^SwGL}^a)3gd?Ot7<`Ermo)O5EkGPiAhEF*MtQSijU`Q%6pKI`{
zP>D{(Fg20HP+!Xy7m_@0{oUA#>P=vtZGAI{Io|<ooc4^bMH^yLp>0>ou_}>T4g&Mq
z9DlI8<jB{u0%QQfp4Yjq%yFJ%&iedzy};qk+-$CFqovKDXJlshAvlzj62$UDY<|T4
zg+uU+(9ejnk||IPmGPyNq#;Qdz((nZfE}C{JFNtp@L=)a3a*7~%Q4Dbt{GMbU_~aQ
zKdyGIg1jQ{5e_(CjCxO`a&4sus)y#|%j1x@{XRJ%gl?u8Af(8!UJ`cNW!cL>X0+i1
z_EV1ay;48V@gA}`TIEKg;RC14YeH0ctYSYaT$<Fyr*z=c-}`mQP+13^bjte`T}1RB
zNT5>j_byJ~iDN)t?uoL^3*Im6W;k51R>Wb*+-kNa(QAk&^E6#2!m@;*DA5rl8KR8h
z=x>xAhfip>D2F)XcnK<|Tyyh{U|TI9?SUvu;U}CZwXdiZn+&Iagy3Ri{-vXIx0vd8
zpk?9vh2t7`YBS>4_pc3$RyS`fR?dffhottBAf5)EA!7IdZ~1E|%YNEaJ5M@Y{XoV<
zAf6(4K|?+)-D^(&mfyg5FKQ2!=3eJ~IHKw>5Oa8ETZZd2H8cv8J1m<V78!5$vQ1w`
zB-o~kkc}z=lUg#2W4rs1vMqgQ$wROKd`l=hw~r||(myI$K4XHb0yYMd8rP6~^=s9Z
zu`?>A`-;%v#F>XhSsF00S#{2~gib}9b^EDy3vKjIHM23UZtjN>L}t<C6KH2CwjDNS
z)v9GF2}a9@gI>fjImv~(<jZ~4FtA@Soo>j`8TKQHq~{Rh>eijeVQ1wR59Z{3jxRSb
z;!#S^Ef}^}lVoe_UGNzf-!d6Oq8__`$!RKmka;vhO<_&n$`t=<uWkUs#=U}tsqo(M
zJ&T-vTY-RU{0Z&ywa?DhKD?S>Xxc8LEW%BznV70G=J``*0ThUr{8zqZLGuTzU)?Wd
zYdM)B<SInN6!Nw9>Ooz|YoC!}YAoTdDp_nG(IPs0dFO&m_olJ-$&6CfaZIWJhFjwa
z8RYNjVz8Y%nNyx2hmie7?Ob8iu!spC4V5<Kv?8S;`Lf{vk!Gc?5U%T3VH-|U;AP4d
z^RPu{65r^_>Hw{Aie5_D^tx=J>VVeU?&uoPM7j95IkM;oD8tO2;E<F@Y5G3XoVqKw
zZQ4up25$Dx#HF8fjI^uY)t^|6z~<c3hg;UrDl|WUGU%356hR6!J1Cqrtoni<H@GVo
z>hj2*k2}VKeHbU`uqQ-G6lC&2T}l1GayN)F9bpbHH_e5g0JkkOiLvDeQSDKHf{Jij
z^&wxb9RFrU!m1F#O~?W;PzCaG)gQnfV{-&76=Czd2h+;)!!FK}SWP;|zAnRw2VJJ2
zX7)q2KMdlxy9W*;FUZ5(gv0i}X<7U%&??!pDoO{Xsg#hSnHB9rPxPu6cMY6C>HP5(
z$F@NytP|~NZgx>_=0=@9B83y2{$d0|KDlyMGz!sm44O5yKW{*p6~l@&(h&^b*&Wyk
zn!mcIf2}EoRV@c`Jie~bEUiiM8S3rh6Z#PXXwtP1f7Ml|)Tg&*`Y}IF5U4ocAT{z`
zmpdqFUtG}Jxi=GT0@{z9hI8_X!pX~x>r|4XLbEf>P8*zbBa4};RiTse8uSa!X7VSg
z>~TE-5<JSS4$WUPc4QSn<tV$d=({@DnzrsfBX|*@C|Z?rn)yjO?eh09HFKH?xuF4N
zCdp)&3<RSlHMFf&z-MpAXUMabQ=2TQFA+VIA~S|uUq5foqc4Va?(*QuyAcm^aHlO5
zcALXC-J%zDYq+mx4~>paIv^fQmg6R}*iytk2_iXKfc*&bcwao`L~jpTm?Ui}7eeIo
zx@t<-*@Ey4Z2*M==^c8W_{2fFU4iL4@Zs0&KpZ0m+doJ$1c!uXhB%)}ru-zxA)MUm
zxwWtA3sI)@hah01w=55iE;`yQy}<(NE%Kg~W><U66tInpEgaw@<nx(A&SwzqN?vZd
z{vxiI6MxAks7+0!&L!8_no(KwHs7e~RD^HgxC>dfQrfKd@!*3syFirdl81XOFnP&<
z+%CntS?HuE*-a_r<Aj0?3lY1I?35Eb8}X}*BQY0+^_9Uwg9mrTB9aG(1!-h-t3J{s
z4cf6qAn3;p=6;bq*I$UuwrLUBLNP}L>$>2+>5x*g^65vuCN7cv*QSCZwvtirbJEfh
zQ=-U?O~(s`+2+Vt9>@R9KsC9z&z3qwhRK`~RO7MViC2#g`Scj-d?y9sJoY!TnCF|Q
z=qctViLIr`rZAb(1%b!6z4kHkIa0S=g)nD~c1iS(NPJNai4JcHa%@G(`DFzvQGfXm
zu)=Sx-pU_{44hd%Y^9)jyd-9VGxeT#nBvrq6VhLh70wRpz<J~Rgi&n1M10<7)-Rsg
zt971M5y`O>sLA?T;Q8J)=RpZrE0)(S<+N9C=VthI$pPi37^wqZm!0o3XQAUZd*6V?
z-ex-bzEWMLZw-0@m6G*N7{;Pv5{vjgQ|zwB5#A%QC;SAsepW2XKgRyDPDz*G%jiC_
zA>GsBnByd@&n8e1N5La5gf@I4Z<p8wgkFwYNoYl)NpC?Xx*NBnalE=O1F%j!4mOSf
z`DvR1+bDA8%~I&1GL9}P&)VK&N>^L7kR!~xQqq&{$QHUIM~3zIu2|E*PTSdUDCS+X
zSSNqK$Cw+8<j-a{^`JLvyU=SZBdWUe7Xkv5j=H4bratsZG!gK<Hkcm9Raa%ydNG94
zMmhV)c6M9i=w>Hd_fX{Dcph{p8O)$7Oti{E8n$6W98XT0oZe$!Uhl}WC}b&<9Dg!h
ztX(2;<sPIFh2txr6fxfuc6;HT{+pt+nKAztQ{s|JmO6AqCNo}-sK#^2mrqjn*oxrb
zfOb{bME|KKKeIc?r7cNpyiV0>i@<@h=Ht=#K4u`2Z8hVyv)j~Bv*C~!sBv-bj43h4
zg&d+L92sWtVXgkhUU{NNs%PPyHGa$&5$w6B`2=ELR-F?u%C2i)%(d5LnzBjxDiXcr
zFrjCyX!29p7J5ztd-&H`wRv>;EjqncMc?yl#ZAaRkT{AZ#3n#hM&-HQThR>h_j{21
z_LDP;hd;Cu_roI6!FjBg<uMX-qcu(_m-gBx8ciWykcftjXx&2n#PmUd;1>1owq^4g
z`w5hSidRjJW78`AMn3N?y(Tu^jt!FG8gPA_*~p$EpVQr+&&vtgz5ao*$|8!5s-jI#
zLlK&xYqyo7AEf%3qrYRA$sM(nk}nJG!zCk{<8}JpIYQ<S5*G4!620sROdSeGkKTuU
z+{~4h|Lsx7ljLSyZF>3YZTVt#=GSQ*+BjDG-<uvW_SId7g&pc0u==-a${`m7x4Cyg
z7s5srE6p8RfnLjdzSY0uS1Rn*#Ey5Dx36;|ch?jA%!?6U!>%7vL@TR1YI4rEZ%IQx
zR!}`=2g(F;J|^&0{C0he0#+c1jxKgEK{d;NQ%o>cdLQ07T|@*7);`vnq)^x_rxZ{O
z1ajq+$OowZeu!zj>$m_S&mZ|01LTu=;PWQQsVT%G|Gvrj!hrqu93(h{k0BK*+1+x5
z!~XNS76fNPq*SA655`Oy*X8L@b~`3}Y@&vK0V^<15HNp_qx3zf9cQfBMGO3KC~;^W
z{Q#ps_Ln4H9pVhj`_(Vhn;FD02`&%g#e$RMkP8<9QIJcf-T-@5-DrYln{o9{c*h4}
z<r|&)B1JZjm<pJDJ^}7<ftqo%va6P5l7V_|;@G<v92z5XBRY1}$vesQp66tFzLXzj
zfoFc0=Ln;jHluFRO942{qmYf9f^EVP|Jq%lq4~Fb2|=I3^opQz$8CZ&*)+*c`y8Bo
z6`SVL>~i|Y@}SmB)v)6L{BGh<1N>=PX70FZW;>M_c5)5H#WZw-)8=w93X7e<@}*KL
zaTD{nCW^io9wWIQlks>-Od$&Lr?}&^xiyMQaPpsVz982-FPs!TA{vC2qJv_N&D>_~
z;7dLiQ}^8vT_Yb89R$f`M|SV)3(^Pevm}8e{_=T(#uBxW<-BpKMC7P<(YNwDM?X)(
za=5(~_{L(MU8Y^8_GM2AG>1ex`#;Avz1c&3V?~mj)Z+l<)8Hu(%91Hbz?mo(SA+a!
zCMhH&DHA9Qo0ANCFs`r7x}7CLo;)VtbIef5%*qTA6saAJ5T5mbgjc;C+~fXBIHgyS
z9L!ftgGQ#M_C3aoq69GN6iFNooO%y5`v%E>HGG_i2*t4_#7H)Gv);uQPCjm0pQ0h?
zvYRf-lHc`A-1$xF5=YvX*GnTSS#b9CPIazyp+~NLNEfT{6Gt+37wFbOlVEghcCrzY
zG@~k+RiaUnm|8P7ou&g&PTKEs6?;R!e%Wa&vzZ@%Gb~ek3<O<am_4dQu2zSFXx0ZO
ze@$kfD=rDshM<~(Eh{~p*a;r;+8quJ1W94jb0rO+!|C@m!eK4qRjflOJ$4nxZLB|>
z?7rLx4WGM7&!FczOXc?_y-boQRR49I@FJh1WOL15?5yJaYgq%DSmI^$<Rk56!UUch
zB(h{mh(Ns#-HJl~jYtBZAO&aRIk4d@*T~FHJzN7AT&c&7+S=E!?-lU-e3TD2Z?I+b
zS;fGPj>sdjnTOWG<O_9QXO&$8OTy67A#%z9_0(LBaA}Y^>#u5Ie|<cJV$Crd6_q+{
zZHOrKnQNfP(__=!`35JPDJBr2J}7FkMeLDtQs(W_q;Emn6D|t)a=r+o9Nb+8+j7q>
z$V>=vbzd_hPw#RO0J6Mq9=uaOBZ$724zo;oIqCab0s4T!7l@G3CuF7fjxJkX9$aHf
z*d#hsGuC@aw({Xsw3|z)>K1qS_O-%%O~SS|mD4U(`)wL`I4?stDjE4y<hXV%EgP_b
z#py?Id+!E|J%685A%&yYjz%*^R!8X(u)ImA*1D7}9Ey{7{z{K4a*}4Z@$#;8pDo*M
z*!c3(gukzXS8|G6zBcDGbe`5>q0oc~r7!cWBr;KKJ80s(g1qH~#_!+gE*11l*85)I
zL?g`o_Ft?<C{aix_MCF#;b}}brTbDVdl<DOgRNL5#~-37dumNy4?Cph6yjbPz}D#n
zNRevhGRwZNsaQIdS^mJ#-|w9xTug_fj(c=Vx2b|^trqGHr6v<}Yxmr#l)?!Y^<4a{
zT(?{)VjW2BGaUFASYJHS2x*NOM^#lsRvn>^KNv%TR}rT2T6RjkcA9AlaF1f+=vqZ;
z!^EUk!Sn(fjuu8Q+gZtw%5prtql+!3G00Owoju|&F6Fw(X(i1@;~y06Qa<#hBz|UC
z9)%}G;9YPTReIFUTr^GB=;5)v)*dV@PdsQ}4K*s8M|f=1bJT=U1<6(zk#93~s6(hK
zHOd(G(xGb5sWX4}sU=%6W^`eHJcm8V!zOdl{D%p51Epq#zGjt*pPEp2KNGbO_3Yva
z4NJy}aZ`DPpZT=7E+SI{lDaO@K9}QbEk+}?aQldZAzq1;9v&fgR1F^62p@YQi4YQV
zVA<1*bOeDmz14TF+i?PMMUrHxfh{su%3u4NL0s3rBOCAB*F8kW*p&2IVGV^=5XY7<
zXzrvnB9opxLQz>^Fh?Yh$3kjcF^%$2YzcvNY}0TeG_f4X;~8}BB-1G*R5DRziKP-k
zDXJM3k(|rRPKU4;n6zUyPD$21l78@yexY{Fi4D>@#|f*(1Q!~lm1%!d(%M>6?hAQh
z;Y;ik*`VEXNNfC{wN$pvPQY#Z>007_=Nr6kL#+Be0wF|4FtMxEDlpM9WU$tQPL#o(
zVfZ||g?2))aHcH#itFxsB5--kV;eA&Nu1o#cEsX1%O2x7k9a09wY61biXqj~o&;Jn
z(B0QGmh{XB#`Mgg)^w%57>@WcbEWn-Sj#-RZ5amhQoa?NmluR7s1b67<ArNd5ac9E
zM%5rfK7eL=Q-J!;d=1Q7Hsh)rd8iF<0Jsv14^c!SB(2bOEao*2HXY6d4!7oMMl-VA
za73gf?STLjNC@h?n8ci(i<+v*V#_s<gKMO`b3CAxhKtVaknp?t=q&P<@=FJSDCY-1
z(%>p1aoKd9jM#2stDuk#t76Kvq|CI0o^z;zQeE#vXD`RMvLPF;=m?7H5OgB&w#UcW
zL0(aIt;>KOcl|6GC<)v!jHFPMU!3p_ZS28nZ)R!)$ED<SqAQ$aCvw_Oa0DUM4BN8D
z47;=x$7T_x>nswdbLabB=g=aQ!{^Q6S|m|#GgeYft1NKH{buuL3|fOl#>=`<CHT2A
zA7m<J*iX?gwf8!*BZBT1Fn8_cg_XNlJ!|i1=}awEmCb>mJ_ldN2@^Zhz4>XBzx*K6
zV07e;@L#89=tpA|BzAON?N=Bakn}yQ*i!7aAZ)rVflUAHNiC`7@;E$d!@HQSzyZ<0
zk24j5UVo5MXHQEn(hD6GaaMY`z79o}ZEuTAKHCgl4<2+BE8v}AwOds1OfCdlUm&Mk
z*V`5)DHUt{+F6l4c{^!gNkb_Wg-z&DO({RvlM;HJo(d^Hd^2nFZ)&hsoh2-{NHptX
zZ?D%GR)tRb&Pkq(tA5>sc9~Ao9;7wBJPy!Iht1B><;p<R@|1Ih3LD7XnUraNWYFEZ
zann()mQOiR%I#`|q$jin$=nb^xqgZ14yyDfm%J<CEO*oM{;64RURhF2DP!*ew=X#a
zxvCQmL^(wHjRy7x$GvSwhvkWp<}wvc$@1#MGBkFg+_vn3tgbn|6>2ImG)_OqG)1Ne
zO*Ap+@G#Vp_ee(20qT&1r6eMGRz0@sKuEt=3mcQ=E#r(vq~Y3;>?Ic~eJM&8d$eX@
z-1!h<iIj2;7=G%>H_1N#ItJd?aud#G5{I04bMw`4nL9oQzP#hS=^M@L-wH*uIfgP9
z%NCg3XT-G4UJjC66z?^SQSmYB@5|@y!QePRl=QYF{8>AE7Ib4K(;_O<5Hqa-cWRsi
zyZRoO7Gua8n40ZO2y(8IN}8&m($8qkyzeE=_mZ(a2Y5S9MV8;MVk^@cWpRc--0OsL
z+DD^#Ll;@YZ5qF-=sVS2=|#Y374(bQk*%x!7}|dmvXoMOKgg4e0G4OOn3ru_6)Gr@
zDddQ&L|*lTsKuwphhO`&<x>LT)t<Md|NIxjWKf21tf3v(q<(Tu|7&5><f(JUi<64w
zA4;8w{cF*PX+lP|UEY+cjhzdj?-WK_8TDx@rTC;&G?PWc8FFSkYhg@0`H}mD<dSSh
z0y+7*cPC1dsIec}csfC%<U7*}8Ia>->TUwGeqyg&`a&9KR&o~zKQ~=1@fV(F>7S83
zX!P7;P+7?5q7Ny1t!i~968(`@2YB^UbUy2q_r)lb^lN?Vj;tX|l&j(9D9Y`L0!XGy
zCqEjy3&R;OY_ExMz?y=ik@?+_73y01(<K50IM!0%Ansa~9Vqfix3Xh~auefk%|fNZ
zEE$$rTEWn4jrGjNc-`F>%`)0q8iC?G)?fF8*e9?sCupfu6mqq?ydqP=H)z1}WEONq
zP)(NfAY^A+bHyQ+R@okzzQ1%6tQ2-xVcl3-7tXuKs}cnePQ(QA45#XOol}RoJ*j3r
zuB(67Jk3XW?DGO)F*V-^BF$DN9cEwW=zq5tZ?0CwsU|6OP^n`~@{@rXsAwKWl4(mZ
z@|x#PIAmC@I4$YtcABxkvQ(s{9W?uYCSHC>gasuC%O6r7t{0Bl+S5{gc4nn`Vd}@u
z3`!{<?l<*TD76A3&;*=U<pWHGwM(_Kj$8sxaCu(K%W0JR4b+Pa$%h27%#uI-!Xq2@
zkynOQU&g9XTpZWvv}$d%H|1dSSDz~xQFRDAJBqMwI>atd%1x2$&|I(VBH%-q+$i|J
zHnRTzarA_MH0yq54~aI;K4LV*s1yA^DlG_amK3X%xX<TbYQH*jVILyE{GaC&R#!Hj
zjYfKYLBh4L+?#1JUhcsmAR2ZUSrqt!f+BZkc6Kj=)cPNV{hyQ&*uMQs8eD(EoB5I4
z_TF*Oq2ZH%rk4x#v-M<O&=cfOe;bSJ253zYsTz2FLBeM6M6RVtp`LmWf7Q*L0@f30
zfBQu~hg`_l<?L*YsR@tu!C_@7R!BHQIako7-xB<@p}M}%=SP{@OA>_2a)`CA4$8sl
z?j^w6fdmsGLkE}N_Mp3?(YBkw7|!M0Ay1c1vofP`MnW>2O>`-=8o^qM+B<R9Gk7lZ
z2m_UT^cE6kJfldMzaFE%&57;H=9GkEd_}75P7$HK<qwB)|9gi<$Txj)o-^r&*x~h}
zSc!>3uROo&3q$5mG87(*`Hu`|na=A;$|cuxqgSSwJoC8y@U<gW1>;){+8h5DZSFHJ
zSzrhpJ^5B{=JVB48Nc|b`Pf|ww%S}ipOe*|7#&L4(4*MRV%uSwKS(1oJ6nNQCemD!
zC4Z21)m_<yY%4N5O~f%)R<{gUi>6HGKEOwEXcwe(QXZ#qKAwJ-zoRGogG6yNY=lzm
zfRMRi4Eu&0H1GY6{Vc5sJ7&d!+C)O*%iN*yLHY4b^3t-6=B28Q_~)po)8>?W6{ftR
zN&TpP-$eg}dF)?iYxfm}fYv~`vh2qTRj9bdY0&lS#;AuGO)&+|&+cSPGt4nH-tX=)
z4aJkDM%O-2Ro|7*L}J(Y0+ppV1~Nl)dOMYMim}-rKHkwMVs|2XZj6*3>7JB48U=RK
zA7rbEKh=5i2dR0_X;2!ul&pEl8@t96C~lD=Y3lddJGQ*ji`!Sest%`rkOW(TPMb4M
z$K_s>Yb|F78sJ^3>`;9ESLVF}%9OE@7tW<B1I6dkFyJyk`=H<(FHuV6SkC`l-pZO9
zNz~xXj#&TF(_{XR0os!t5%01x8XNip3hC>PLJ&djXMnas9f0{?oglNL(a9|hUsW6F
zQPk2uNJ(|S*VO1>MT@PEa;*y3wiiUDsCLcnub0NeBLqIPZhjxMJZEGrJ~tB&3F7Tx
z>oqz*6~CHNMrv+Q%*(69N<wcRK*2s{|45YlyT~umBo?)_VNOI{3@r9Ho73dJ5};Uc
zKeH}v<!Dsa1WN38HUB}{bP&Ey{glnWpQbY>`YqtO9dH|ozoQl%-qmkqcVHOi)c89E
zJZHlmnD;z+ixKF}y7;^JXZ*akQw|Ghqd2qeO7ABvvx+`dlU#9&qtDlW^PnBQ-;87}
zYjaF2DJ)JuUz0Sd0%IAV&-jGk>yl%u8Zej=_b(U!J(zU)H{OB!VRIti0#GXeK|A-U
zmH5b~>h}pfI|UC3&Ch>1{<^Oy4#a)p%kS(I{~&iq??%zN8F*Tsb*}^r2!q6-NUr+_
zgdZTG<{;2<&4Hs|($YFEW`2-0j-8CD13wV1&pB0>QnuT+b7|{$wA~I28Bn!FP#6>Q
z8?f(J>^uv6kj6JiXE!`36t~z*K?LyvQe+++0Sd1_qYv8laR8=a1q0OUDEi-XIwYOm
zB+nV|wgd!7kxxc?YQLOn&A!j`3-}fOKBva{=`p*;_SH7t=pQ5-?~s(4u%e|lU7Smm
z4;LTWWKSm;1CO6#CBE`iF$L0+Dt8m?CrHR*{X?pqt$&aP<yLx;p3ezHT@&UY1}-2o
zkK2R8^GN;)saam^;-}O6C(nN(kx!;tKDr>Vd3TBbAl<G!nW6Y}NB^+{8O}Z^*Pe5m
zbMpxR_VZ(A&HLJ@`}as?^;w#?$HvebOCpoUZztHKt|nC!7JC`jM5QV5C&7VaFGQS1
zP3=po@pEcimG-;RM>m7<09;xi4|e$hrr9jvUyrFBephi#(d*Z#KWM3?(qveAkgYEM
zly@B%DM0utIsCnKpGf=q_s2jU3BWS4ytqw(;Q|!7P$+%&ZABbrD1qF-&p-&qO?Uq5
zRP>0zrRQ@ZxsRxSaW=KG_JJ+G(dQdrw&I<LK)|N%m7c0Eicx6f&3gm*{Fj*ls$A-y
zz+!;Si^u-czLCM=DJIAN6FC$Ed)6l$n6jt)k2ufkZdcfz2kBt{tupaN$wog&q~oHq
zK6?y6#9to*80m`{b}VoIM`@FPg4RLzaC(1mO~OIs8{pXfx(5J@&t!m!`u^ye>YOkD
z(If1EJ-`r%)V+Nkp8qokQ+jj%oEmU8z^1Pe$6r$&H%EW1?ezQ!FIl$3pvRr}y+T0$
zCiXEe&_Tc%JS2oCw$?3R3zGLQXZ|4lwVF<xuf|53==1*4)PIo55T2u+nvV*UpR^ai
z#g$|AK**-eAA8&eT-&)u$NVvk@I2efMAS;2qm?D6Fx-xs&yq8bHCIoF{8yC3BO{-J
z77ns=cquF%!>(UfKI;7SL&;m&p4Y$#=8OeC1^q$#YtjtZq&N5D0Ma*~4Qg}^0K;Sn
z_ycqZNHH&02QW(aUrSw;kA#2!%}mw@T?woL3=q3>sSDUV`$+R-uEnQe=IH-&b~>6J
z5lGE{-5+3;--z(K0LKZ~slqdD7%f6`W(Rl%9R5N2>z@sjA8DUbALOVBRavcx(=Gr>
z<b_cGpB?XfRZCDtQ61R__DoI$h(MYS2^BiPfPOOB=5!9wnP}ISKtgP-aOKo|cDnf0
zH(=D~F@deBOdnJ5BEo+sYfpAgqM`ZNZ1>Da%lr@0ss-|bt5Vb9(VEx*W*g$mmChyz
zFK6xOEh}QItUg;UCK<!<(YHR@<euR%;1{LdA%B65ZzJo+*B4I9JauuG59Rm|Bw(a<
z!MamqG#OobZBbsX<+g8eX`Z|9R`<^$vJrPCg(`{D`6-iO1KSoH^jtb4rmWzHNt;y&
zZU|wBTZjD{->;X+#YIe_Tv-~1mnkanMMVcYbtM-|gMk&(zKmc6nmED@7jSv<%w`Vs
z!g~Mpm#KUG{_cn6e4nc(rH2=ATt=b&%G|Y_GAkSS4$BZ3)l8V8m(jr9W8RW%{}#r!
z!=1$XNNGBpq7%*`PcdUeRC9i=zHsJ)<UV6K;OEJwAcoI^VbcLCHBqIwo1UGSLi2z8
zRF*UR`7G$O-0GF*ex_f$$V%V&XMMwA(}!9Cd}A4>$@9P{lv5<O_%qcznh(z2R4eI$
z7xquwlG3r~2w&XfnFO+aeDX0VZKS5lghzP$`uK_VrBw@u$yMRWy%$nR3^$qem(8U$
zogXK^B1@YZw%wjSMg%r@1up;1p{h7<fOG=Vd?rU#CRa(oKhR5?*w;UZ@1)$75j|}B
z1(GA12p2Y5*J0n%KJKIZLGm`C`0>SPWpM9P;5hr0F8kfU^`i&@forotsPDUf7W8_-
z%8&i*|G6^SrTg}h=MR4$HDCT9wN}QgceZq-AOL;8|GFU}&(qs)6z7T$(-@DZ?cy=W
zM~^02|15UwA19xIjQRsffeD6G$I^$6-^}aR$7=s*3(`L>MFQNbRD*t%`4!|i@ZRkc
zAp7@Ue|m!Iox*0K4m^JP*uR_wPS&2^29Ku9RQyu28=CVk^<t@HCh88nZVnmk#V_>E
zen{|QFqUD@S$Dm>XtZ2zec~F(Exx16{s6mv<Ux{owCWf;*6p~(1S-;Pe}_U>HgZ9Y
zA`FiKo|JEp7<sn>p=EfFlR%0z)ZhIK1+{hjg7=G<hg0SHCloR|+vc>+--rC$CmlbH
z)lle;z&`mO0HtA5|CB#S1RT#u=U{tbpNltzk2|ive8PVdJ$<2h%yM685(0QK@$e`1
z8(TkMX`f~PAVseLS&T?(F0OV6#m+#H(yNVlBp~+gHZ0{+vuA(DA0&CqZ!xcjM-Q!#
zi5@uvbMDJNJx$1nUvpc0_P7%Gm`cUOHkwNR-PQR{?^8!R`yZr<KI!&O+32n9%i1U6
zek*^F3en0^8ieiDM6CJ6;i+hMTYr!q7}z2#)~oQ=1sfmAJB@(vrs?d!BP%KPTUmkC
zWJS)WWosuOog9l?*Oq}N9)gW5*YfpIOgviAMB=%paS2Aj7I-v+nAoTTYElf$^9TkP
z#*mtmwo)$TI!{k!I>f(zB9c3kn7P0`lWZSduZTsi5gBpMMkjit1$vYMiJ85^yj+)c
zD+EZ9Bd>TYf5`a<mOuRm{JIEqTui=q{sY?h{{eiaij5A)r+>l|&41U4^z%Ql^S=~<
z``;1!-^~Z3<;Gv_KkCW*U+j{-<R92Yb^N_~@$MhMM*06>NUSsey_WxeAB6kKhECa|
z<^SOj*dEfHZ~nug{=+<x{?S)R|F|P0YQrHS?UxjRktTrG_&%g`d;9QNzLHzIzm-tO
zx!R9DCS}PTHK0Aq^UF9p&l`8UXE)iwN-Oj|rJ7eC;9^TJ!qtpchGuTbF#>Jd#Ubxr
zfrqv=ICMtIZ?><0J&`ZvUXuP|xf$jUOmE)2MWOS~-4nrKXCXcKgLHj^mcuUQzbGo9
z(v&}${e1rql9x$(VD}RVQ}D29Kz%g2&c(RX#yzOd?;Vz4Q-+Zj=ciz!qLH_+Z$)R>
zW5T?w*RX^~tgbWk=T%JPZK-*3f9gisiVB?QJ3dSM`1q~Ya9m@_|MQ{l>%ir!Hleey
zaGp|^EEheUC7+S_i;_cmEnpc{VZFT;HMh2p(oUp%gFM?053MhX;oOp;31iiE6$qn-
zwo|2gBiHsmWB<;()Q$gmQ+@AMZuC>!Nq24b=;B&<=TqFk-_!K99Sh;k3n_X?`tJM1
zA6c&)wq78qS_d|~T5n*hd8&CRCeq#CW7Ub>Vk88|5uCp}K$Be{%30%-@cNI^|4&Ld
zBmTu#Jf1(H(cfRx4==TIHTYNPHC0UEiwm@@8-uA=4A0Qf#MzoZKE6}gHxXY+eT#%7
z@48(qCT2=xHFz-CON6U%IM({35(qXesO~{KBgW`d`IDr@VcWsPP(?E_1Q&BrxbITv
zy?lBEO(R=b_|HMl&$j(We$T0i=qhQ>=d}#jBjzwa+2avYq^`*muAR{u>toX9GPE6x
zV;Ecagkgy^+PsG?f!>!Rv_`)L|8MR7r#LN|X^NGWt1_}MX5I8ThA2&%7KwWNyvf^q
z2Qf1-)Tg$P>sO?Ta#JnmRMb4DV)Q-AiVe!%``=zm8<?~tosP1cb$lu6`FUM~cZWNf
zgUlv$#>%_iQb}I!4SwnrDu3KQ)KDa!+&7n0$d#}mJ-E<_k>I-2kH;FZ=A1adGxPr3
zwr6<e<g+5(0QMzqrDMp<&iN;nbDO((cE{r;$5t&j;85Ig)oE!X=l#BL_|Eth4Uw_2
z{5i?d`c96-c>a+bTWf2=4)4;Baj{@hIq}()h7`39&v^eA6`=Y2JX!0$Gn5BGdc`#(
zYiGmelN0)U*~M5RZ*TfByOx+de{>(h$Bj0$n_w=XzXvF1NHpnAM{;)pMA$N4R$BUr
z*0ew>l_zAl!eZ>HD)lC7yeW|he`__UWT&|$GFtS<+Y#dN??;(QiT6isX}qN$E6#N~
zQXDhkFe#;EN5p1FJRgFe5e>a<u0Vho_%gPnVg!kPe0vqbmBl;O+ADpy^TgMl;cv50
zfGgDh7X3n=@2dgyEY8oblL)P76iWXP;n%^=@$8)Sq2CAZ42eEZ!i8PkuhRf#4<f_G
zi1Ufsdg87enV)C>2gywTa!3FE{fcM5n6UIrsmZ9#PTbrcd#VH=A#d@wTL~YQ<<Hq_
z8{@{QYu9dq_f2`T`sDvsZ@jbyv{y`6vFu3H^N{*qJ@>*Q8$#jo?zv1%)7xi9)JK{-
zZ&9QYb3`$Q-m$NoGw(#TjJ>mb6V<+SgP&m`UD+iT&{rvLhP+bVftj+UW-EVTyKy;&
zE-RD$2T5lL>1%{ouL%dT5Kjs>yM*5_#bVMcD}%8Xj4>2as90sf8j{SKuc-CeC(?+6
zi<}W8$Kt9dLiAN_&yY&p$wC9(Qz2=|**|7sY#^zuTmKA4>HmAzNNzR<lM5Yc&w$^V
z@zYlgN2jW}<<`z*B!35t9YBpl+0p-OjI!n0UqF<!e;@*sr-f)d{?&0nzqtLoW1pby
z#2QAaasoYDvDx4unI~)2Zgvw?o^l;kNg?t`4RkxJNZ%P!b5Wnxvk457FNQwnlmIZJ
zPF`#(N2>nqq`$z!C_55I&JO8l0vg%<`X*tVTA62c>*iB~f&RW@b2S;dFb04Q+znFJ
zOPW?F`?n{kfE`F(9?KK;A~pOSe++jcSpHv_zRK#gyDZ#ckBUWJGm!O+4Ua<fVgMsu
zXoM>)Dh=Go;eU?l{d1YGqmC==rHw$tv%DbI&imyXZgK|~L_?HptvW+r|9w-1m1Qqf
zrB)W{$5{WO7NhL10Riw|75jp51sH}20E3`Brz5<b|4QcC9pRn$Nd)|+QT+rolD_k6
zNM$3tlcQlvFr_o+!j_GyN63!}KvY}kMLU9aYi6=i7r8xzEQhEx44dQW*E_}~ZZ394
z!t5U>v7~m%J5jYwLQ9UnF)V*u1%<Z0?U?M5sWz^dmutdRS+1X!e!M=%;j8Q!OkN?|
zk!77Jbuh|Cl?mbc-|G7QUklUdsMEQCT}MBeZfz5tK%4-w7}fJvcG(@^jX<Y&0}h^n
zCLN?fH9hbCSrP_oc4b-1t4gurz1NGG7dpvP`!?+V`Z$2bJrnVTw&va6%W3NNH}}t8
zG%Pht_C7sW?`hZq_;Ek6Z;AfZ-jDx-ytj;sYx(v?8%VMfG<bl75Hxtu;1;~m#sW<P
z!D(pRJ%SV5HMl!;<0QdpB)By0?rwRV{l7=<8)x5l-xzPa59h8AYt^W#RaL9jl&V?3
znng!K?*E^Q{O=V0>xE%C=grFl31YMlqT|uEjdfz5OGeZ!-2KzT2U}|gSJG{6oKcVT
zJdmJcx2un2_h<@PTSj|OeJC9(_Ksy$8S#Mni#LHfXgfLYBjzV}w@d=Vf~e#Ri><X<
z;sVT#qGQ_K%gYe&NwZjJ#mnkXxh1Nfe423^yu9ow99iyjk!%;7+FO#PkLJY$L;8q)
ztgDWaZhR+r`5eVwm~|(+M#w^3(B_>%KM7Y=BfC2j9=rl=lXA?Z-+<4~hHrC%-j~nh
zc-{(l-}awA{{QyBbmWia%w5k9LHO#cUWlba18a6Ta=j(nhE{PMq;#q9P8Ri@#r9u#
z-{6MVyB0{iNp-yR*ppsOwQI9Ng6bCa&TsrU?d&O~7iw&8t}nd>$jHjG#hLHKUmp9<
zw4j79Q3(grD%|lEu`8G-I~M}m+RZqLU4U9B?eRI_HD9I<!F$ZN@iPEBR>bP1KVi{%
zX6|=9)!I8o0Mw^o>kWEzD6Zg7nQ3*c1Q@$f9Wi=TFyk8wP&p1Oy%!J1pz5A6dOT@h
z^zU85gw8LI@xNV4lP?XH2c+#OrrS`Zzx&ywyaH#fKe^x$4Z1)Oly;*cXUacOYbwLq
zt~kn?BU@a6V7!+&e%y1r^HI&-Xz%wY31^Lm`gg>edVGmGFV!fA%3Dpmn+Q~GD{>$C
ziv9)&dk6EKljlD2u{fpkzIhB-5IZm&1lVPM05Ik5pZPO+H{qg3{Nj}h9EZ&66fbEW
z_@gTZq@%x|=3gJmCpLPT^dH~zKgBkej{^Sh>Hk3-e?0&G&z<(Kyv;uOfb=t0bltbj
zK4GW{Kt^oS<SBraf$$grk!p<W2dw7lC8V3}obLd%0V}<CFvE$GRe)KExlr9teFYBj
zv+2T@)Hepx=c8+c{N+OhxJ`)LXpEQ-5`UWh|J#M3ct379k=bE{U$KUs^mV5FBsUjd
z@I}x)3#rp-5NQ>LEg886tNo|;Jk9wPen+iN9`rKsH$bUbveW1+X2VHta$#tYyClU)
zbI7NnrhTLGc~#~Ac<I5J)7tpj?$saCzTX$8k@yzxw0+`8*!1>t??KL9*wE3lZ~u)~
zRJZRO{HA^bJOP=J2FBj(m84|p*E{8Hf+wQ}#_YH3);`2$dzE<3Jfje2vE#|)t-0x=
zX9<lA=MX`5IK$P^<+s;kw%-BITm4ZN55e8}>Gx%VPl^-UbXtFbUIYH3nZXV|6JyHS
z`hMj{)0Y=Z-)P8-y7}ays_VenF#7KNe3y{WjRR=6+%g(zSkLE;`dQ}09Gm-cw^YfR
zhm-!jA9-A)|9TLDgTi7KA8CUlYR~#+18sL^a%3{xf%P$YoFzrxGaO#h@QnRGsPD&g
zbYsCi{q@vA<rxOwANc?M_eJ#@ymb@$r!-L!yKqZV?C31sO}^qg+E{Ln`U8n+sW7!v
zO6eSM_b^Qke5q_<Y_#~4=H~Ov<6ddmSUNg(`JOX6x>ZxN?>)f$fyHPE0N|;tTt|3#
zll&P#{_hI@7q<?BaZ*YkN8W{w0c~RTue_4BeUGj0rvQ^s?Ai~{`@hnA3pj|2CvJ(+
zm?M4t*w0m2kJh3GbvGd!#~7|C^A4?}X+(zaUlh=7PVs)MNc@_ewvSzCQLW$l=J}m~
z%f!e?d_m$e#z;l4n03>VDK-GldHq47Lz}A~CLhPhssDtEFYuJAb4G%**E@0Tq#p{&
z+oB==ACU%8<Oj!w-*{*@CNN0J&gYQZZ^LR^XgC+vCclj}Y)f#Y-IO8@5c>RIm-jg3
zaW1Yne?j%n*Z~kRw6gsQtZkuxxMMWWF#?di_VSTC0Z?u2odI6Sf}5{yR{@Bk^CZk{
z-Dn4P??t5xFmnIP!Fd#UdE+rF`i)cnHoutY=8Rv|KAs+*&>l(SFoZkXBgKx-{QoWy
z(=g_W#oOkydz1XbE6$jLuAV<%6NOVC&Tx(Qr((>zd*CVkqiP3LZrq-RBx2pS5U<vz
z$MdR|WVeZ<J4Y;&1`-@$BLy88zyWW%xr3#>y&*Nm`x}=9yN?4NnuU+1HMdVgjpuFV
zV0=P#LTP!m@wP={w9fe?@ThFwqSffycxbFTt3SDe)l7VLAD(q{Y3uXTroGrv1HIpX
z(8i<fN!2vqK_fUr_YA?Bpw1G_$5;%527%Z$^)#%VQ$A*d@Nm=rOu0rN<okQVK<r9L
zDV*!CMJK45mE<4b6zY_%@r6@wBavISOyS$I{<>CKA771oHLAA^4fgx+dAZVg@`2c(
zI6rUlvpLu_FDHdLku|v%G*cp<#pX*RoR-l=`SJ)oNS^Y5y-CCy&nh5^+s~3!MJ)U(
z$owq!P&*y7LNOh`NTFEjD=2?#YeY90GYixDbu^Kwpu@xpk%_{GuK5q=`e0jlEju>1
zItHl@s~zM*MT}uo_skv>^0q;QY~kTmbaVMHa=VvTh7RZed$eyK{1l9L=6W6cu<7>T
zs6m%4G7}5KMP}N0Ikow1yuTdnE<g=X7-u$Fkr<3w{zn;M@GNH-I+mU86jYQ<JZLI&
zQ8f%)2%%b=GSa^Sx9y!U`Op|*x`oyH8@6#Ydp|mI{m8knngiP7Q~)$Op^5Wf#KLfF
z(A;(8IEK#E>+2sUH@UTlrU1n@jAk@FWBC3C{7n}A5bli{%dyC4G@-b8vUp1thLz&@
zB0P4fdO6zAT;g?fta%xi($n-EOPAuMm@J?cP?}&i6pZPqa_}BATb;sbv=8&9W+zz~
z%`m`=E9bUX&3e&Jla&u|E<FFxI}JzLZQ>@HN&M0D>Eeg;cbq{z9uQF>!7y6dN{4Q@
zJlP$GA&~?CE0G9bs94ADd2eVdfiI<1WIbUp<IfRu1B|rC><EC))b5T(`tbnFa?D@t
z1UyTCBvlAcT7zG{!D2A_UkH@{GHFFInTXhQw^x0c7dp#C-#X?@S<5rW&|FJDo<*@K
zR~}HF0uWJ{)9xiqp4FtKT{L$Ibq!d%w_4f&NJ3nHL{sQH6VolrYK2-wUlDZG`wJcw
zpB9DEOBZRZvFpWXP}}H?d6FG%CZy|Tge-fKW3KebfL)SAg1`Njt@+=GWHy{(y&m_L
z1#I8z6Wr?ebGe$F9qhagq&3)ScN16pTiO2=Q8K~Ii*2(#;8>uV17eO&PgXD1GVdzT
zC_17*ua)NyChc$;%UKB(b*~Dc<W)&u2V;cv!%!{Nb^sP0%fx->%5p77oJig?mTt$o
zV3pHSR)?-pp`ko2%dRE1uR>l$S;^F)+yGBTGIwd=msQZ*PRgLPRK}0maa%mp3};f!
zZ0ob?$%K@}E{cR8x^HJ=iHu&%Vl}jM+*q6m)U;Cqc`%I%5xy37>+9hXxGHGhTBrrs
z@R|q3O*bE=mD88}WR%?*^khH!uS4y)!th`MCv?v|4ARP?aP{_+gz|W8ESN69Zd~VV
z5|)<n>G`agm$x^=Z_`x$`}UNP&PbQ~TOkRLS3Q6F^Z92>>UoJB^biGuW!;;1f0fRa
z(QTvl>1&q1&8u%7F8;~&-|@oe*4YWW@4or<U~8xe<!@*{d)-eI`|Dnu$<5CQ8MD=9
z3%R{=cwn8Ck?xros}6Gp?pI$-J$?NjeZl9CYLsmrFrB)wQL^TE(aK7&P8pa0wpUFO
zX0D`*z%qaS8*rO%TZ>-XZF|RIPVi@_fhB4}17MzKcfQmYFH$EjP$uXH6xJWFRIm&9
zQ_t33et&8%({Xk}n0);>{5x@lVR!oqAt}jIMY_F{=;*89p77hBAB!9>dyp#t*(0}1
z9ZWt;pP(%2pjjjA;GQcl4=>z5?R!8O+@x9`Jm&i)IthS&ccmx%D^q)ev@9T2Q~+%U
zWwaOr6nNK-F8MB*8#dB!?En-gUeZmL4C*8MAZB9#>wNV2%@o(pYiptp8eJwMUA-_K
zWpj+)*Rkka<a%O1p`%+HA1XvvguEmSaLnGYL-@S}C?u1M+vRF%f!Qz075of?nHi$*
zhCS4B6Y~^kDICitCaYnHaWSJjD6N76?L3G`-{XLtA<K1JZj5_jS!ScHtm#LEc6;;r
z%Dgj|j#a<DMXSnI*5aVUzA(;{q9+C4KT{63Uk<OnRmy%R5YD?YJiHuO9;2A<VVl=Y
zR--pz(JwS(!&T(Ev$}%J#2o8V*FIm`aA|>?RidTnz~tseyMc&+%U6J`T!otzg}0~u
z!76uob-u2{(vsa%v_My}sL%uDk>}By^3{?$onHwBYPo|Msu$QUY51H`<dx|2{lM>>
zh7)dkKZuEdFCC0Ov|#sT&(|uBCbAS4YL)qNtVavYg(<|0i*DqPe&y=Pr}C<<1BW_P
z__MGGx0g3d*3=PYIEV7FBq1i5HMgCgvBYGHUSWq(6s{kKX;Gi#FUXfOb;@#wrXV}D
zoYvaQC5{74A+5$d;*w-JMugGlSLmNkJLb^3iCD2YLB?RMCf)fe{oKXtQj#yA6dggk
z#z3zzZ*?q#_3(Y_pySX(1j_j<j@(yi3lAKHJhTHQI=WxlEaj+a;nu{ReiH(tR%-`0
zL6qQ|otF+1`>w|w!0hXe%XsYmq@!1_i%y%$H9ciSao<@5Z$UF3ph72B%!;YlmHKTc
z631#ML8Ax*z?1*rh5s^SOa;vjPeVEa&)r-ZJT7dKut&$IjBp))vOkl<<5TeT{2q)z
z2I)+NWO8C9Srp<%3!1m_gyt+-O!!gUKr!5yv{lNM9r{n{-e5POr;&z!nB8KG4{62}
zFb$!MW2ZYMWNzWt{SRXUZhHO&-uy3H@$ZpDvSvMx3&Uy~;|Z_4#Ef`c?0Dyp6{wGc
z0R<}Q9?;BiJec%y(P=|*>kO-yE8CuGALEnWrg9u(R7aY?_Z@2Qz6S`(SfwCe6WE<Z
zp=Nfc-XJj=CbjI59n<$~sS;&J^eZl^Fi}&5MRN!ov-WdBH4}_K*mvmp#+W6YsYWi+
zjzS<eMs2S02PIJ+k%G*=S4xEx9UnP<H7)yNCpHqLVJ$mTREtIs@QC4LE`HLYT?CH$
z;rh}TsojmshBSZsUSLSBTi>dgY3WV{Vw`R>%d>?C9>@NCni9YqFiu$@6>+jBeIe->
z{W6zv&I3NO#a_*uNm)h6Uv|ECg8G45T)Ysl7k|*J(dI^lN=}|~?#_E1Giuvd%|Smy
zwsgpj1Cmne8ES6j`c!UpK5adJlNji;g;sVPN?L?M$ghR9Jnw?_$fW(o3S+YAIM51f
zd#x~&y6o_fV+^$tb6bX3s{N_@1?Poal#od;k0c_}6~EDm$8ofSmq}yN@KYJ?O;3H&
z2*WsW=KkAxq)&!?$y%LK#_BmXz3Bz(P<1-dK46JEL}=UWo@=u)5&uL~?BnAG^9*5~
zzyrLddYIT{s7N=C_DYX^eL`9FO2}A4LZMU`Xoub1#a~*cRblCet9yo!v>OE|6{PCk
z5S<W8HyIthX=*kh0oI#9iak%X%kt!9g3vCS>TOL<c(;DS=yW|mHcgiQm3-0mCiSg|
zw(NdN$`~^D(`aTpukG~>z^RQ|*E?Py<JDz}=>P}Kh_5p5WlN(!IN<gkaWdoZHcl_0
zjkD{)ICTHY8+)Y8W7ELbSL#}VrNq71%|?3Tj1@C`jsx)R-%0Qv16XQgtZ2iJod6*b
zYO)J?%&J(oG<kI6iUAA}C1|QZ)noWDW!)O7_(JSKN7Q+RBvtgMFoSnjkNqj2(_cK7
zv>AT`%4=&*(u-`8Z|6S{`{}pp96bUC{9}dxkYK_J)gE9~wYq<zJN4GZ1fR($o2^tX
zEI5jzlr0tig{{dhc)Y}8hHWLC8ptuSTJv5|<<)Co0ONBP2Gbbz*Lc=ERExPwt+-|<
zFRB8+X0lCbLoYQxKg$f!oRLv1u**-^E`2{?2Hhz5kybdiZkqp~ffv4le7%pL%W{BP
zt+K^rdw&dqeZF~avS<6_*1p60-9Uucl9?Q~gJR?d$%5%RSLjoyX75CR&}uBVVo6cT
ziBq493>(j{GF*7OURl9&Cl5aDm8BEpKFlME^Do99J?2pR4PIWGRdWpz=rh;OVQTg;
zFU{A3dhyu}dkqsio$sNfqD2I>v`j2Enpv<d4iP8aqSv${YYQzrZSg;<^Yf4p=iQa9
z(Ng|tS4V{I$0GZU?(2BZOZ$^BKkD5dbV9w>XXD2Q51@Km-M`EY(s|FqoG(2%4&}e)
zj>#0DsGGxM{-_nR9f0OP^!YzWW5(ZQx0UJqF<_Z>2l#QJp$04Gr)P-A_jLH8R~}4*
zi+j%ifE`AA9@u9J8syzw(ye>_41x^@Kc$E>nUWk-KAxdCssJzK(#%cUUW0-Zd0+P~
zW{%T0TSc^&$Wwk>8r0_xahIuEcR1in3h`cDF|RobV)#G}DP^vCo!H8bV(Z}vF|1#^
z&(4X|h>!-Bk3aL-h}j;M48f$O^zqP-HSg$k1bdV(M7s05wKp|1$?p|`>nBY+n>9Tt
zy%lSxbSAf@I&LxP?$DPg>4|<-Uy1T8(X(|iM`&lV4~<3b=Kr##j4Pmmi9Fs}uoznE
z)ujF!-x%1L4tLiL=2Lou%zw7b&|*4aGUEmjyZq6Jm+GRn(f=vA(oE9vN$U-}DCxju
zEZQA$ur8%Cdl}<9iA`HR8@mBGvUtsq_lD#s;#BA%sqN$u4&Ad@NUG};`57i{v#}2c
zQcG*&g^YX-!`P3{e4VNa$&9g}&~32QcOLQTHmEEyjQ2;;g-n}Le2~C&5V7K9w>Mrr
zS^j?b+h#gu^*ngtG0|9qCPhq&dr|zu>UgrPj`#+KIwJo*HOzQcRZg|A422csL3>FA
z*3Pd;6mm-p)39+q*=E7+z$}H<4J%Sb`?UMUlR*0ddF&S4RHLPAT6ccXh}Hkq{|_x-
z^zeTIs<b%oKWyD={sDM08PR;3Wm8sDk0t7YHng8uP0=!xrL6f;9sQqVVFb|Ei8^nu
zR7A;C!xrM3zE2w3>0Vp}ly#T&o}xhzqk%2g-o}Q)j&0wczGezczrOx5cw4spn1sO2
zwnEKi;y2oWKza!BIHDC7=z_WS$7_5i*+Lq$mbZvDxPCQ+A?zTDJ9YqTJcJf~^Wf@$
z2%^*mwzL|Ad+D0#R#Iq7Y0=f2`4YeR$JCvF?{7dF*@yA*vmpbSg+ohCZqC*1zt9NW
zOK<EZ--%P65$<>Ie*+#zT4|dd7l&7^-Y~jY!UK(+Vx9XC%<i9Qa?(;|XQAFKlyf=a
z>~&R2CCbHP%Wy3pvO6&ri0dm_mE}?4y`C5PvOj#v&Ra4DY2Jgaxqj8)7hCh`b#ga|
ztvY@$)DK)xWsDdb(~eiO#Zcr4P6@9c0WCi*!VVZNO;LX2`sOZeZRx(B@>GAqK=bX1
zgZZKXeU0n#!yq-yB#n|&PZtN*yP-#pMx&~8q@&=1HilK{4>Ei>&b=xr2~H(u34@3`
zS+Zur{=s!}o@BEDcDFb6y}<&#dT=h+{(LnS_s<SP4NU5b%5dmGf6yZt$w}=FEAD4j
z8<aM^EDqJ`M}&?8cBYWtrA&p54<pG>4x=635mq{x+M00QGPl?&iMLphvhhGuo%P~%
zo%LoP6&Q{Q#1fO`Jn72i2AK%05D<|;=zt*7Vx=Bg_5<GrcPtc~-#CJH*`QOT-Z+A0
zK9ON1xGo&0BP9=XS`UpWc6)vuz`xOEB9Y2b#9*uFD=l!C&GTZsK97IaH70iW0iUUs
z%nt87p3M2%<<?o{yn0BIQoRwG!^?Ih(V8Qv5C>yw(~9l8P6y+G+yTG2!b2NjjIK43
zk$mmGm_3zm!6(dp5o&m~eVchg$ha5z{pG@Y?+)#bhH$x<sB28zU=-ikylM_m%1wGV
zWjcjrsLL>gyzWAfSw5e37DT|VGYQVZHc(KHs@dY&t}RfgYep?Yo~LOT3G;nv5_a8y
zz5XeLZ|@KURYt`N<@11ymO024#V94p5mverX5WVz6OvWTgA8)I0#|2%hq?UimeHbI
z5VP&HEuI%1k3$QSotBfmuV)Er+(mDcoOImRgZNNvPQyMuo#7@(WvQcBljSvCpktGh
zhK+DJLy5CGORV8Nw3PdVl+tC~RwcPjFvWO7Cgx-yTM+RLHUR*LktdgapYqT|-gB;V
z5}UKU`1;xPVXxmTi?#EbLV%q{MoNm!1J=)VETvz#lnc1crBAmUW+s;&d4MB)z7OB9
za`6wj%eX0DB_;A$IImjMlax$USWj!iDf@~fJ&X6)QevEiUI?YqrmUf%F{4M6^#@_0
z;N5rwjoNEx2KdNK5UCUJs^W*>*(`TaqH+e6?J-8u?#uW<ioLLOr92%wFf{h7yF-$0
zod7v|iK%7@;mRE8j+Atp>o1pLEj|#-=|vGy;l3y%>r+NP4%P#j9>h+R=*ds)nTzmb
zMwLgVNCw{g^5+NZGv1ck);*sotN1(Wpp(R8Fzg(qY^36{Vq--~(H5_;+0{V9qh#tx
zgW~!HRZ6-6gT~9xdD_0;_bMg_zD%$hpHJJ^*73a4@Gste&5PYRtXEYA-J;11vO|`}
z+hz)(H{OP*8)O2*ltbr_z4Zv;gAwN2$*7n+uV_2(hP;bNtEscPFySOfW>}6%{&+8*
zPE{igADO)>MV|AMLdbF9sji!dkQNAzB1~>BM6BiEn{}E%9G}jauD8e9jjQb<rj1F#
z<^7GTO(f6#aQ+hV2<hb;sJ<ZPJ_|;$r)qr+6Bu?K-{kcy-P8mgX&a6iDD%%X8A*HO
z?`R6{YHr7csL>zr*K36g<T$7@OG{Tf&9rQzQQ*;?iPaaj<-;91PEV5Wvf@Redz(<U
zO@S09KZ0z9(#rx*dtr<+wCVONiI7~Lm0?;co3R+wnNV)BP;Vk6ZmW=6saoZUugJ&E
zRH~Z0%;wTPl@7cx@Vg_vTc@J8D&>7Y!-3^v3|jk-&lHt&s0cS1R4e)wxA|CQ&)V~P
zi{*01=c5_tizKAqj}QTq5|39K8P^U`3q$LPQ~0x&l2npT)}||P6}XY9?6C@&IhG&~
zMJA-vRPi;Ru=Zf_<i<?xYiDROFJdz^kvqg$yF}|f<+zA!2wA!+SP{KL1gJM|#|LQ^
zbQy{g(?0*PSCXV`%jdZ^36m98>C%1{B2*yPp<``3Mb54)T!9RzAJ<k&vZn{mgH-Z6
z)fv0L`y?vMnmI}@`Xi}7aIe*_P#kOdYoCz|@p_l8M-uGm5}y$Jd?>3p$o?1-@6MkY
zrl9=9wm6a{G_m+sS_ypDHT46cn3I9a`74!kXks-CRa1H=nWy^sB>pZ>u%KWAKQB)C
zo8vKpB`nPKq`q)9)!1i1oUlAJm$K>pDp$q0=;2fa1D*(KuOeU2R>a!DR5_teOSst+
zb$skHne9DL;WVv0*d-5tylyVbw7$LCsWkOrymH)(A`-sRS7S?LUK4&kED@v(2IVOS
zu2LD~x?Zmqqfwt3aRc(eFEWI7HAdGJ5N}*eQ=aGNr+Z(KD!5CohFp!yNCWak2xW!s
za=F7kFs6$3s2jZ07}gi^h<8f1)3M^fGb5??`ZHI)T<};<)a~>reP(F4d2&TpFTzi(
zsjGeLmU^Z$3YJ)7zkz*(Y2a70;kc~S6J@tE|IKE2xA?_@M#c?-JJwCh&@)G=YKeye
zzJyGCoyct$&WJOdbfl3e13WxBR4xB8`KyaK6_0<-#3yZ?&Lm<AH8b25(i^QE62#w(
z>~e{eEEH`%;Wgjko!P?+OxpU2>s-6dYx!%mLWW|DUpjja$f($4x7>zvAoIc{Q=cH`
zQV=Y2;^%Ho$JfMtGVY^!BuDNz$(X}3)yg_buGs*IgiX%FZLjfEs$Q}8mwf-sIlCP|
zZTizuk4{zJ?ZsxQ6|*2m*gAp4VStlQivHu7qME;OD#6qaJUE9ZM5{|p<Yxuu`X`&U
z!ex9m9xU|7<==ysDKnCtzj&n`iN)J>jK{LKld>hdOn4MtpG`)ru=x`DzF97RxLeSr
z^zKK$RHOaGUA|7`x+woaP4RJI;seGF9lRpTF!e9+1?w)gxY*oyBJG%AX;Ud$^Agy7
zQQKylno%T^fE%L3X8rvemdnia#|`|*&Nd6LnH+-z)1A<~(4+&Qh^6oGo`$f|;wV#b
zT8*lnsMD0{_w$N#(3FvcymZLL^o<$pz8T|qX8$cr_|V7cE&+_gGyT2cBr|@uYam3b
zh|5v6qJc71rBJ^VT)iE&sLw~)5&8mYrx@3%G#9|-uuZZi&|Vgu@&pIM1N5xO7;MhS
z_c=fOKxW5<#av$%k?cMg@Au%1XCJb26pft!ajl$*>B*(SE>kP8+;ZYvwe9>VQ!c}v
z9#OIvD|Bildwj;no~TdgtuFGTV(`ao?4C~dK*oeN?V769&Xv_qnFWE*ayd&l^)9p<
z#?cz+Q?-lhq>E+1!9mzXZ*Hw~luS=e%|GiNap~A4LSQU=ZS>Q#$#y&!b@LZ9YYA-O
zYa?W`+vELC{aV~@%J8F=r-je_Y;{-@A3PqDKG1hy*K0KVijk6Tly>wvr1=Cd=85&N
ztOcRE2O77-1Tk1wf?LifS9Cm&Gg3ySm$=Qhnz51<$#-~y<F;L>^`O+e$HqOG^0|H?
zwY|q5VpVzrh|d_cJuMY^UP;a`{u%lI<Mr&zOtE3fTJGg<z<3}-Vt!}P$)o$mEsTiZ
z%u+gp)cM&*V(6vIp`Zl0@da~-A*c8tmZ$q|_0Go+e%2c+duzWQw{@f=ZEss;qImUm
zkB%`CY+BD|)E@c&V~zhgp-9}uK#i^~K$YRI@Gs;Zp5c#Yhf98tiaj~zbDX(x*SmD6
z`wd`x|Jh?KR;HFsn3L_oY*<%=Q$d;Kqhz<&ebxKIqTo%3i&*@$_JdTmEZe50#u;T<
zY2JB$pkc;dlhBmr)DF{u>G9I{e&QRga}TXqjVPygW?3BUii#Ga9QCYodf~sUw+dyy
z6!x6nIIT<wYHcEh4or7iZ+e^P5%|n5mo{MA#=R*Y*`a|5yp<P4AzEy3gNarV1Ng&K
zQ2ol(M4j^HDAYwpBF%N+)fMTp%GAJd-x`V(koiFhUH#T+s1>;Qnd2P!C$KmIzh;iz
ztnYT*Yn1iUgY#pkE(PmMUFu}Se!mz@4+)`m$ai`9S7qf!pY7a-xeL0Ev{2q{Gf`sN
z=$MxdSMeSyg%btlMeSn_EYqAB^VEK=?a^fu3pKuTZHSMXf_mLHv!-;G44F~ShnG9w
zJ-4r+c%NT_&xzfi?6^9b58*P8&$3rv9Z!o^FOZJb%Tcb|w{~$dm4O-$>bo#cP=bOB
zS7na&u~Ui*T?>BVOLq6LGqmt-ojeQ*76nHBsH=MDn=^{=AAzaP_BrhY$!GdbOecQ~
zNnnq5w@A`;Uf%(|Sz#6cYv+#8vs*@L)fb@PYi+iY3Gn{3*R~VIQM0MTDf;@d8B$~c
zWvog&6~UJlUI#60CL~vOeFn}>H-VboGb@4aRdzTV%c<4n&-rWYr&qhO)!9vfzR#Ie
zB+cJFrQu_k8hoy(C77Q`Nek9QB);m*YN68tT_DGFdFFiJsP=R2=olN8!F2-dgD@=a
z%W;YBey)#X8T+J$G4Eq0X!A~DZk)CfnR}&X^%<8)+*00d4Du>0D7o+sr>JbRK(t($
z#H&1aIJ3sQuh!f6(B3J9z04<urY*CK9Y!e1OsJVEx8KKFbodgJKKTVMrcz-EQARL3
zeG>(z<FOQ>XIsXR<n|+8Q*G3%k5gl})+WY9lu^~<mfFNK#|}IX;Q=czwjvWw4c%ty
zN+QOL^(Nz0Xz?|XzBtuTPza}{b^w<Ov~(;f7Q!nNmoS2MdpE7YQN3d?9A5(zc4^)*
zN*_8?&#tmD$9h;EF!iiO#+X7QIbOr8j^$xTr$FSD|F)I+JHE%tdAME0?MZ<|fg~e1
zPn;d|#zB*++`j=94IXx9bBL7umDWCXlI#Pno+j_a0S=YM1LmjXJ!k9^?c7?>i?dC6
zs{hQQ{}{#uCVNPbe<vS3&#ORK6k>QhkIWQ*dwhQ7Z?{u^Kpe^W87RIjrnWS6@kkGD
zM3F(VY?ZmBr*lgJm(_rl@6A*BNX8O=`^Vz{Apv9T_Nrmn)J`zMx<D?Tt-ZovV3xkA
z#oHag_x_NDS0rx(&#+nq33@zKI`Uq_lG1x7c@uDR5zvs#=8Tkmf1KlLB*wb%Ow~cD
z|Ak5S_5|M#p@noY2F6~&@OMeK<a&aHYA0b%-@JTsKx0sa=p?Ly34S7X$tL5qQ63HG
zx?)>R82cn&gYEa9B0{#ijq%_zROOaBO6yGRo0?&6Un<L1?1;A3`_h50v>*Bymq(Yr
z8?z`8$Q}kcfQ6dGZ6Go4Wh~8a$R<61RBvl@uo*urbDY>qbTC-0W2tw+giPA=&Q9hk
zMT0bZp&$#zxNESFDMGt|B*(D)`P~z1n|XwW7mn1s<2_bmr9|!9WU@8FD}wqQaHQB7
z2Wy2;-4P#qH(4IDTv1AR|78S|=`P|5w$SSqH(J2u@wL10Wk6!0{$JU~XJ`!CkK(;(
z>lfCzA7<<LZHfz2{m-u3H~1Icb*g&L6>|?rLNy*)u-A%Mw7vsVCJs2N2p`x1JszMC
zZ1)r%6?@HnWb<>SZoY%K#O*scoVD@@S34d-`z8~kJiN`KqMJ|9X#7;hUB{?u{$dUM
z0?=m#1daT?S-ghBhG|U7mt2lIZs0}>&mkfR&%Y~!bXQ`gH;GLoZ-w9byv3XqujZlS
zca~dW${!{D@m3#pr;CpId?-CbZMfs?$(D(^MF~^%E!TlxpAF?O;f98Im(@)y*m_Tj
zAZWx6sx2%zoZkRNboRUN@-Ju|$jJ+NpOD)cRNcX%#Y5p=$cE{@6%LN(l?sZ*t+*&^
zPD`(;$Jc-g#tikc;TEpzslp8#Ef{r1|8GeXTjxCVzRr-)U#Kln{5q>&_4*F-8*ml;
zsypWH-e~VBCT>q~|0N+=!79z89KmyVF3b^S@g-H1rG|y5IySmQDm7v{BD+0m*>bY{
z)dao>(sinY*big<LW?M2)ruLUHYZj&;-+X4XaviX${Jx$;bWQqHku83U90D<2w_RY
z2wagW7+tIk9)8ss(FYo3llxH1J8G%Gz^p*{G&B*2m3^c3hLs(?)VnUa)MZn?b?Nny
zbEc@=p#Gzx3rw>#X^f&&In3ebY>?WPVE~iEgP-4!N?)4uzXlwTrMYVM^}sb<b&uEP
zah=>nDYT|16Kw#UDe$52&RN||c!j=X`1{YnZmjr{sgIU&K*`20Fd9GnVxL&5p~g)0
z^LU!Lf^9#$HC}-aytH3#ec@EePs&ZlPH-iRtV9-eKFYXJvfrDFY0}|F)UG^KaGl3B
zZ@tf7<ghE9v5EtFW@*<OMiM2v^?vE)BP`<4kEMpLmSMZ_Jsw%5N9L+Oqw<qVrOd)+
zm!cZz4dLb?TJEs|?}<|b<voRa)P1U7Sak@*a+qdwky{vu`Nx;EMRuOJJ&7?*CAI@!
za*k2+o!z}8wA$ez5Zf`hA@w@Gdkud`5<6k~Y2~`B6}A*Vh`d{sQJz%(4Ilt`9Q_S%
zppTT$tf!l5&;46Y{h#FGCd;b4#nWoCz%mg$gR1bETztar%)_DInwcw8H%xB3fS4lr
z`gf;b0#$@|<_h{~7g)jR_Yxl^MLZmiAAx-&0Z%v>^>(@kX;;Hw^MDCXTPwo#OokpY
zh6Cc~kr!`1+<1SEA%iB)XHcNJ$6iEEJ=HUMWb9QV<3*kwB6;<b>4>xVXk7<p05`RW
zG72Q-NAB8GyOzGM_fz{ET<1;EkN*Uj3Hte5J)gUw3Vm8Kn42C{5=H6CTVPcI45QR=
zWIoOw3Sv%}|FAsAEl?cXus<}!_OX0?D3Y4^Q5PkHY(sc0%1*=sT$$s(XK+y@q7yw_
z=Kd%ytuNh$HE)@;8KpLo%|EuLjVp%Zv-axIhnSPry=fOjJ(Z|gyKsW*$UveSwFO(4
zr%z$$-RH|Zh4LAfC+w=ebB%=wd$OdqakCRqUz9@=HOG=bBq&=Bw!oNePibaFT3u}v
z_tnBGp05bDZW*=iNtqHuqUv8%Z;ou=PJC&B>F*I{#!ahxXw1_^vlSOj4Ax^Yx02*0
ztn5PihvZEmqIOS`WLs>7gU_!)1{z`y==eF^McY5j{07`ec^K*1&mp^5ExlVOb~|9I
zV<rkGI~B9fAcxEpoJHopiag0GjQ5FXhE#-TQ-k|xdG3b_)qHT;Wj<2u7O&0&!?RR!
zc&d$K(aUTJrb~4*1$qotMX%beut6@gJW^f{lw!@ATE%G4bumAk`W$y#Gn$>nF8evL
zlst%{mMjDod}(b1A0fU|1jW!2YjK1Pi8EY%R$>bnjfPzs2u|g`ksjr-bT1Hg5JIDu
zE`dgq27Hz>MdiS9q4JkDUNgyGvAm4{7!?Nrm}@>Qp&s=@1mOucQ`~cA>Uck$%g$C@
zU*=Z=H<9kk#mp(@gqzcy8ZFE*d}9Q;?t)6w;Wj2c*EaE;uGqWQ1KTqz$;=DB9Lr_L
zC3VDFKYaPe4fCuN+?E*P<P`ltK$EW*)ovB2UljFY;q`};AFT_-^o6^`1H9tJz74x^
z6R3|bC7HQAYabSb@&qREJk!%k_f}{;K72_sR~11hex2NQrUOyiY9S!-#tF9)r+hV-
zoXz>vP9%<O`K`;SbM3KJdz(S|(HCE6*xX9ZJ|6LTuO^MA2s0#F^OhS3?4U#+b@DdY
zP9sf0AjPKHwxjW5M^WAr#n@MAd_uVey`AjV+5Mq<#IiFgd3)FOYi&*D2Em46J}$HR
zhod8?(Q}>3Y=$@cwCfL9RC$Z?kjQJ^(MptddLtf3C*!Cmivp3A^H^@Jg2rMrp(E|<
z@J)eb+oLNi+1aix^Ou1R@x99Y%1s*WVV=?jgg;bU&We^zF~KPg>T&KiZ;OtB3Us26
z?VyYvW=B95Bl8LvS%;^}81tpv*JrSBE&Vq0L=3B#tt6k|${hJrl}v>TEu9xEvAVXn
zNm}pmhfB%_lVwXl=5n!keYGPV+!VZUUFe#%%~B@mm5aTR;=$l=z!#k=Y!bZ;2X>&7
z5QR>bmeuKWeTfjLobTO7e#H;6(#T^hdrm4}=Y&CF`-rs=wH&#O6~aV@vu5R{JZXc*
zeKp$`G2u(^sU)kdf+z~wh0H)-t&Q&z_2T9p)xrcdvb328wO~7;pPA1=31;Q7`UQiZ
zlE1=7ov5sch(=A(XTNI7wC$#QFh%hx9P*J5XN*7{HWfp{UAc!;{yNNo^jodcMNXp%
z92S0}4Q5y(zJ#oS^Ialm$<!%JYu~;)PkyS>>!9Nle-#Kjy!s|axLB=`{ISZ8tpt|K
z-!Q&wmUcYyNip43mJMQVE09ggS307gavVP79Na*NCvJZCt!Hoqk}H5!L(Um<bVB{2
zM3W9#oIPF57Ulp?Z?J2s+8hL$e+4o{ehY0^pV2^DMZc7Vdb{w~SZzS8;=Afx7wiby
zgdH9c9U`r?7t%JtY|CUi=XQ$`*dLU;K4qHDPDizrWBYV#<}_T_LW3G$EhFqjc$9C+
z?l3U~%WsrZ&EwNGQSe@yc>WWaInyOm3FFVkN9OoJOZ@92{Fco8q%9GX9AqXBbx4?>
znJ`}^@Q^>>9uEO|@KSK=8NWNyS`lSD*<%=tR-9*d4@)Q;{e(&|BsrZTsJ>LL(m2<L
z%L7N37;^&FTJ95*;7VcjJ(K0lvJGeHt^WOX0od8>5%tqEO{q9fnC{dk6FhHC*ca|W
zpZ5&#z1VB24j5NwoVVBnRVrG#t%)^jPLjkeAh60eZcAYrQ<2)!MGnMP%*rdGoq&bW
zB`Q2k*pn~QB^g@+d2C@2J|0nlj5==VN1a~sAt1Qqx`s0%ZXd)@FAedW8B4^ybR$fn
zg>8(`a1AWR2q;ctndE@(R4E(+(A#LRR~Rf`^V_M@BbNv%tShL8#T6~8uUyU^B&3j~
zH8y|p|6AMr(+!J|N)xHWJ>z*-wiLp&Rkajlhw!|@dR`ek5Jzt1DeV1SHNrsH@%mUW
zR3*hp>=*eU($-KFztllH!)(mErODqGe!6mWq<mAe=n&oz?LU(y_PWJ1FIFKjw)WwL
z-O|K-IhZ&4N&+0vPyiMMdWG(uMg0N3BWd5NR^9<P($vs6zRfq}D$MJjoo{d-jz%W<
z2LGJ|&i+bJdG`nnzE|bSI)C}?J4B!$>DLXxFC&gLjsAbVB{NmXE7pGcjO={o?<dh#
zr^qi$v;Odh!>%lN1PMWfIQgi&z(D<;?+|k1Ec8++fD+Ik^zy?s?i1(b49mA#Q@(Sv
z->sRbo6d<F%ZJdI=0DhH)!;cD3*-@?(akNppyTW4F1mem?{>u{a0*CHXs}fRy@UH7
z4)d1<22J^X#zeO(1rMEvvbXc`=y*SlAH*MnUe?>6B^$<(C*-v5`%PeCUlTqkE}k{;
z`CdUTuC0}3%C;vs1R-pDz-NWRvQ?{qR7?xhl!xl;omIG?C_r!yq35dyKTFki+;7~z
zmx9*cw)?V6rL#yYuxF(1Hr85q2Mmb%t|{MgHvPOu9HHt5J148#%_qbA6uF}VaAWf1
z)nA3$1cp>=^zr7kbl+n>4i9;)5~5%!EiLoSWzt==f`I!fb~jr0eNOPDSCFgrV?Fh<
zl?E=3t%7ccY+>IKsAg9@-v!xj(6hh>RT2SQ>gs|%GdNg3F@?~xRp$V;3A}dSpIgj-
z1H;Qt?jI;~Y~Giw22-y6>=U`F?U`x|kGF~w;?tTcXU_KI)=O0NS0xVCOV9tw(jL|S
zHDrY7ic(e!QFnmIy)?9oR@Zuaa=CYSl{7J!ZYa1p!&wuvI4D!}`8;oV?j%rYy82C^
z;;L*ql<E1(TlUZ!q*;)NC!UJ)=J1G>o(jolZoRKi+QQ&uCZdwM8nC;|q)T|k<-5mB
zj>J;p4gFd{<!OR5?q6g!muB=l0$!nPFx!EwFW@GNV%Wy?Oi+Gpnl(F|GE^v=TLjzB
z&=TVSz7RFj#GhemABCPP{4%5!i;#Avh>3zKd_)6H>1HycWcu9br#cB<RDzOO^c&5e
z&V>aRrwKV#>&K}P!zKpMr%27dB}+6@XB;ItDH!3Sz%~frizc=O?yM4$3)7y=n7%Xj
z9eFpXHSW8Z!TF(PN;K(~$*dRQl0U;h8>xXliknZaL_Ub}TiqBlsgR6J*UC6ENZ*H2
z{=;QNtk+(`S;9o?7_w<yXeiX-uKG=8Ks|fp`yX(i2$|~(bL|}@K4(Wgmu-!5vt{L)
zVxi7n{bU{=2RM|!K+31lam&&2jBF(6x#D!_L<l`RmMx;ODqf*U>^!arMyoQBrfaZ4
z?2rPp63(|MN|i0%4v&ka1(8@Z$CKqgs&ln~qbhcger_S>cNtzn^F5u`l(sX2d|F4J
z*1F==OefdJEVpuFwOFK!`+K$Ck~rpx3c-}Oe~#T8aJn~&38jSh(QbIH7MY&4Ws#Ed
zCqLJXE|P_0N5x2Hipu-<RUEG)4o_|p<!cuuG=WZ#oem5MOlO`|*$0h`<5Iksv*~{r
znrI8YWemvOv=}o277#X;x1aB?V{UbKyJh@nL`xD3G+Udz+ZCfxM<-pjzF`=(YZv$D
z=3bZJNVh$pL~}mi-{c4;T3B=i`p9B#rv5E6{1Gx!2%0~F^1&v$q(6L}^y_L9EpBAD
z&Y$6+H^y5WQWc|Rf#jcxF#m`JhSA8;A^Sf<=}%!jn6ZuMW+|I0Z=xj_TB?vRIY2V1
zF`OZZh7a$;zI@^k>3?Jm?@8moWI6j@8m93J+wpPTqJ(m~o&NPTd2nz8Kzb$RK;tXx
zS>4Msal&n^m^1J2I-4KA0cF}ue&w|lm&`V|pX~P6#?Lu^Cch#1C!zk2@h;?+nAak~
z7FvEzas8cWA1l?sN{nK}N5!QKZJww}5D&3!y@Adf_>97JDyJG&uRAo~EKVfPYBG{M
z%WB@;yVrX9dw549r)V2zCI5I|CyOWndp&`C@Q7}h^3UgjU)tqWmZ_k8c?xEarGTG6
zaWuhUg|%dhqi@?K!)Jx^`)5uvG2m5yk>xZVB{UUAs{ix|l=ZaA2IX-oyDS>z1VLbD
zZ$DPO08a3womdpVsNI_+%HClZ{_r}yxp7-y?$&DQNX_yh*)xpO-E7)Ed?v&?z)JVm
z{AdUq!959y%}Pka9a&bod3?4<={mS3vzD;7V<~2_B&c^W-U+uehtT3PYT5@f3|URN
z?gfPsev4bLygx_}e~U}D|F~*mMM+eQR1d<umNKW9@~&!DegD-1UffZx>KYs9D*=?S
zL)EJ@foy%bwDxD!k5hOP)^7?l-;!2f&08r*o)J0`I=0la>@MBhN^U<Rt<>Np!OIf#
z9lE!Ghw4*Ax7xF7yy`3%-z7ogK-sTI^yavle2SyB3|>c3MZeddYyU8yIbCqBmq3Wx
zJKEP;sCv71Ni*}z@gINXKP3F;jUYqdFfPc{g&W$RRMIC}L*b*-S^Qz_$6^MKoLtF-
z{(P{h>=_Qx;wER#>n26dt?RX-1Q0_nD3MmQPR6r{F&!hvkVMXGa&kCNh-^QEc8-?d
zqk!#DTMPHm5q~UZQYI?>RZkOr&Fl$~e^%wOssdYZcrVA~rgCA}^=jdWui7K@9HZ={
zC&Cm@);zY~@hz~htqaB*U*M5RjCX$YXjKBwyP3($HX1xl5y2Pw=^0CUh||*eo;6U{
zijlNezCKFL;L8%(HT~xC1zLZB+i)O{#hzm%>E}&A$;q^jkHE#J1)!qEd<vzMg0rne
z29#7(fxh!lNEwt)FZ?n&Iix*iW9>rb?5}w;Jv)yM@HEd<VHhIRWk6K!t6xk@oAFDn
z;|a7|B=sTPYT=x%`%4)hICWTZ_npzdbuMQ1vZ-9Vviq^PeBWwlQTkrO*zp4ozJuei
z1z3ZdkL2sWT?fDB{I~r7?Vk8TZTg1i;`9ChY~*6}P6g2-MBnqaji&ED4+v%$a9e9C
zTg*LODzJF@y2$&Wn`2Sp8@dlLJ+`u3yj~(+@f+;4<?dM0sHqa|(e!$JEN|JU5No4J
zK+wYn<`zqPTF1g7@d$z{Cf<a$b(AEyDcy-;l#0!^0Qq#;c~-UYg*?5bUQ$zDcb`h#
z(<5-3s;@k-s==6-Mr6ieg*6IZRvTHraA|)7D9<Lvy!-Y_CfftsMc#!6;+*HN$6BM$
zDrWiF(VVj8IXnOd$6G5!hLYhr3%?{RB&X9z(a;LDK3xA<*!Ss9WcD4qjG(2hl%Lj@
z^&?6P;1H}8emIZIxsI~tAKuH+)$e3WG-~Z(>slJ5H>s;&#j|Cdxb&4~m7UF-xG<iu
zX!k9W%i<4!qi2W`w<&RfDA^JXyG;clyf1JSQ_Q^Rxs-ZH9C+DO;ce=~*CAHo%mTLD
z(Uc~xhXsQOe5|Izt(B2%VWKqE{X~^rIciCw=L#eXz7SJ9h3Ya(82Hf5i8)0W2j5ZH
zas7b9ebkZ95WY`EN~&O=E0wERJfs&d0D8UO!k_&9vwVrLeEG<>Jz3v``K=CVHc8k*
z<xAj<wI6JArY`DM$JJMEQ_u&g#a(2rwf$b+S6ru`nm&KNyuDa9RW=VP7^_0c=){Ih
z>2S<zVGkr5RW}@$E({K-;Xqui;)R_HF16AwwRYgPEVbO2&cc0P4ZooxDRL3!Sj<z*
zpN(SS-uJq@RLcWpN}2I7f{C(QUc~dO2+RhC#}@Z+9u39ER=hD5{*Kmtw%>Odh90n4
z{uGh<S@c$nX~_jNCrhK};-%iVD_@J+D>~4iD3u94&6RRFCuV?U$=j^*6y0)9eBPWV
zS9xe+F1!}Te5yq9GYsuFuDF)1s|`}7QA%ZA;G~o{GSl8v&ZU}3nbf{77wSmQ`_s^Y
zJ|ez@U)EYr{`)k7<%s1CcYC5dpQrbXvK}MW60Tl4OWuo~cuJToKQ*{ixnL-Gi#R&J
zJq}g1CTAbgWe9_nhA40YpiJ>>&$}p91IkiLHH$x}(2d%6)mhIwSh;`YN7mE5k4<v|
z>=s6kQtr5Bq|kNKn!a7CR#w(aOB%^x(ut{!qN?!+;J2>VaPX7aR&RtSTX#%Qd-ir&
zV=h;*+RDr_*PfTy>l^bKq!m1ow=IxOU#`=0%9k5f4*ZGsAoRlHei}=jEz9`P<iTH{
zhG35mEsice7KTj3cF_3oAUAW;8c<Q2s&65DI#H`G7i2PS{%h1hHVhMnwAOiD(eW94
z6hg8E)C?gI=2kPk0q}p>E1VHdNo%eOXy7k3HDrRTyatnJh|IrlCq++sCE7QMCe|F=
zBj(ISn0?q?lR_N6$>pjW6*YdY${))7)d`F5fHLmg`3T%fp1pkh(q#UPg@N{88yXB+
zuP4^ss!LdK!>pLpE@;hu1FU}8KA6{@V@ZCm9lZ!Smzi9N57U|sSk@xjK7)#M)`oB`
zm`H5-v@BumIaRU0e`XTqMaAX1)-NiSlA?)>S;(I~v|(BErBac0tl=hQp}$m{NoS8q
z<Y(t-p(fH?De|}J)_9OsqON7vUhy(v(F69?i^ZuSELF-*w1!H33#L2OqQf$F`{KtH
z6X{GDWu5~#LbrB%0F+8GtyY%m+g@qtx=_u2+7m0Ns>;$lNrw!yG-$lS`qlhKgrqCY
zsGi@r;PGK1;gMjN!Bw}F4tC}JKm{b=9djM-BTWLAh>%BC8ll#_*PoV(b*n=jVykD#
zs?W~h-5c__q<o;X`t?Ksr1Y(wPBBb|zK@mOO5KH-QC5*7hwc?Ap~f34t7IF>u4!s=
zGM%IB%rDOpK|fhUV=Xn9mLWm|<-n?}=V?vGqJ(P$jTC{!BGR3uV^jrwz>}vdAk75L
z81UDagCX%xSgcJ-PMLKk^wU9bJN@Q0{LD1()LHlo($r5MEvt9+RoidwA6#eKKD_K_
zew@xh3jM1AWL4t&0z_;cqJFCzl)6QoTCF^XYOsNPZ=e4K%TAT%uU<3X?a7D0mCae7
zl%Nqa<A{T%-vIJ?C|(2Xm*pfM?PmOP4Ws%dOcpq@=|IR|xSdZ$QWF<E{2~GVX?h|3
z{Nm1T9WM)RYPxt#5&gW`OiCFfYFiOVM$CItT=AqelvMVpLNVE(+5Cr8n0J=2z1B8M
z&`5G);g;nrk8aYpkxaT@{1rB@`U72HRB{%|SIMZiR0o&DWjn>6R?%|}4OnsK#2tQ%
zLZ);=YjnyS_mb&aXF9b=aD)uK<yTcTp3%&(3xTWALw6o7mlZ}G)_Pzvt~`9Hdft%z
zSv)vlEkq{IB)&R@Zcl^CHBU2I%%@Bh+8{({?b1aF-q}Pc;ynotPCrKk9jx@fh&+VZ
zUp%$1F{Res^;WI64Hs=L9CeT{)~;KTNw=H)R>w)SZxyAV2rT06=-EufTQ`$?n<81?
zURED7r;eF7toB0eFwc=9xpaQ*g1tUHfr{VuHH6|rr_lI{{ZhOjF0x3U7kb0LyfU^p
zVk&^-bIehJ{~EhSn<7+hv18M*b?TQjrK{gl#|DC3Pew!By{r~nac$iDHF1%}vA|kV
z;iu11p^E`ch9OUiy+cp0<F;RLoq`=8RvOLf#VND4+<3X=IuAIY+3O9cvb4P_jA7B3
z2JV@MALpN*a~bD5+4K$gF0~yR5+1PDP{2a*6y7VS*K=NJo0*FeeRN>{q7l1qyPEQj
zcW-^ca%+%K4rnUtqQUA|xl>W#aenZ4hhsXh(<cwbRWZx-Lj>57D)ckjsLgFQ!K_TE
zNgg3O)EzvYoTruiG;Wr(E01QQ5rp?ZJv@z$o54~kwrLf&^ioa$9O90wwj(VKXRYEW
zm1y23rzY<UgU&cx#W&s@Esu|_A}I2uVACp$EZVVtIZ}<>qFUMV{e@z~m2U~S{=z=j
zTvPekiNn2I$q9BGPRFdWhG<T1@ijUHm-K5`C1Xl5Y;L>cF?29=<e3K$rpWhW1lr-)
zVO1e1iy8nXkL-h<9%&&B$!%v^T$&z@mFVw-p7DaQ&1RB8EAp>#o0A4lIePg~4)VGT
zh7Tq1;;PyoTE9PwVm-hz=3Nt8@35&>sG*U`yzQ$;Fvt;1MPicdoo0qnLO0FXazIK`
z<L$f|Oa8pg(~-${m8taHH=4aJXjkrxCZtkf!ISAf<eCyWG(A}>bk-?*^B3||UON^Q
z`_wk~f9Z=TFsYZVhIwv2Sy}0sVuB9Zoc~|seFs=m+qUkaiGqT3D<Fa(7-9t#2}O`5
zAYBL$b&E(BkPZo=C<2PKL{JDI9RUlyBq%6VL8=fSbOfX%ln_Y5Tfu$yX5V|xz30An
z-uJ!lynG)dYt6N08S@`w%{9iDW26t8x6*)5m(#D+HQC;F=c;rl4PP-UwmbFM_zgqZ
zATp;|_ol9mriqop^Xz@sm6eyzXDaSeKCyRxZwY?dte9f&dO*RqfIieP(bBK^#a4E#
zklihEy0_qJ?~{D^TxFP`V$QSM?*bjlT}!#{=AJp@6sjA2ey_sfo4v|=EI9aNTEg>7
zZTmkEPyCmQ-E9&p+337x6hma>QZmAC_XmYlXl>s8gU6o*v2L$usQUC1+4qB;j>~DM
z@!WWJ)7lqPT9TqbGN`J8$+)AR5>KoL?BR|2be?Rhhuy0)4GYqmot<R)qA<S(2Hq%m
z;k?g368CH`)ZZ`5R@d_MUXKymHzFlOWuo2mI8)2F<)bIfP6m#>6leh6-61y*T?+kf
z8!W6bI^gwgSNl(K;fb8l_Fdd>ssVjl$EtV_N$#6N$*LT&A%4KG9nY+mXQg8RVS^oG
z;C8lO2mA7e080TS;Am6WO^orYA*(5HdxlT*7*Ku7$*yRI+g*3MZ#Wh{-BnxJJyIl-
zZ|72x(W6l+d*(ccRMWQe*2Rt=9fh=8<d2;(ahN})a9!J-z0Mx)bt_ILvH>d=PVn(l
zgcOqcu`s^^+LySQn5U-Y+){x;+P)3#w{G5a&F;v8XgBMhuT*-mx=bNMZM-TSIxnnW
zGaF6lj~kuixK3NSMij}&*Z1f>>S4wYG0t)7Us8IT=WUucEUNr@fUoI9)zncm@p%(b
zK2+{oH=|^-_SA)r+NACoS2?r|eAiw`E93S%A55OK^!I!3^Y+g<5xHM=-$o~ycVy~i
z8qW#K#y91A$QCH>%P@5)FuY}F_Jcg#Caa$rXrADwtJhcIWg)5OjyTvuZ1yeg@@#&D
z>G(Wpd&=hCy5@<1b9*6Qgs<Oule45dAvYYV&|gbC>uA&RnZpV_A^FWId33NT%<#bZ
zc%6jzuIHavPL9?4jiCD<H&~k!BovhsY<v9*zAmpOmRKIK35<N_(*4G*%CPYu?EWK#
zpvJ%+iQkf9;jx;<IWcl(!a42`Lf`4ckYNqoG1<esE#b{i4{;56=vgh8XPEcUVF_7}
zM16#<F|gRezR8|96Fk2`loP*Sh~z7t+5oH}srRmG%|RZc)JW#`U(!CL7JFKmX}6=l
z)#S9KD_)5ba)ZQcC8A?-?UCr1G|Q4<EH^!%311rk<xOG`h0fdP_jM&I9Z$Df1^uHA
z=m|(5(>k4O&iE;$_l55`di$%6vE#!q1MRl%F~`CgF4*JFJWb60?*)T7F4Z2Z3v$-^
zxOLJhg>sdW6kV=Wzh7^U=icPk?S11nZX{oyjz3N4mYY+^eZqR*rm)=8LfZ7X-qV*H
z2G0|NU`ZVHFN1P}&JM<wjXkNGI@!=vXfs(jd#Dj+2eJMR570R&wIiG%W3nep@$vWD
zjm4uqa&H_mG_a--J1AG5z%%>fghxM=<tU#r;j+HjdE-SC<fy8coWb{FB=dcg47OWH
z@jSuZ@pVB?Sw!uU4&`@}Wy2?bDi~jmtt;kN?pAnG-I-C_*78ZybkL4XlDwGjeib7%
zpR~O@zR=w!{&~l-z*EIjZE2O)36P8GX2PWkhbnU7_dSR|I(nUY()!^8ep6D8qeQ-w
zk;sKEZ?6HD72I{}yqM+Yj{JdUNem1($QT^nNoamJqM8AQpUb~c1td0B!oUE*Zv53B
zXHKo+?(8KSo>>#(<MRu4Bvt&JYsPes1;4vvD9Mxpjeu}t4GU8ou&8bd32yK|ldYez
z<i6Z}TdSp*<)(gu4`Q_F2`C*Pl?;{rPeAFwzRk<8vgH{h_R-U>#Xi7N#8I5JkMO&c
zVFxQ0XKi@F2M7X^F<`7rsML8MY$<RMjA0xxWe_@eO+OC>EwsJBKP=f&Jj6jg;CGa+
zuMz0$A6b0-pzBO!XEO|3?PioETjM7)Nnq5x*;s+7`U0TQH|6&x{^>8a+;j73-E)_t
z?Rf@DvX|u^juei2ivEb(w@dVVMpvF$Vu6VDf<s*HO^v{Fr=|tOZXMJ9vFZk!-ft+h
zo6U-s{V>#b5fjPs<xBa~=bB+>^^EgJ&0blhspDO8Rz{a^tC%$1JbE(cN47$H#_77S
z0q(D-;}7xdGj=4Nk%FXD4?nn;d3^Mi`MbPJA81hgTUd_n=jL`7gQRPHUX|}$&dZeu
z?0PHf>?nEe>8GbN`DGn$7ak~-Y7=yy)Jnd@hoXtc%Gn&8-{<<jTMmIun`E`iN`)%0
z7-V!EKF9OQ@c9cXwDI+?VN+5K?W5;)Z(Zb(l5^;UaOucIJ}||>yL$<b_=;+<54zqF
z_Ln@fbK-XLbH6V-_K8+%)1%z0^QH$<CNK~8pUY@J)IGJd*W-ZeUhCs_SB`q-e49Ai
zeDcLPeSulX(KAly;o7SDxgE3Fg!|mQKdmdZdbxdiNZk=`-;N2t-8Oz}BCXd%R`C1l
zZ;2IhIo`jSepWGgYBJ5SFMmX1#9Cpm$t85W5VXfud}@$uT*NVm`=ff3tKA~;5pu17
zdzRN4b3|%*SFwWY?mkoVHy%^PKUS(LI=KQ)H)-VE9j#5(IhtAfZfwbzzx1ol*-Y=r
z9t5PeprYT|rnKi~ypID8@v~xOr}-hP-i+pI51Y{kLVj+tY(gstl7rWNb$!b;#hk)Q
zSdwAx71?cij=Ri4-BMzY>Rr}H>%YKY3VaAi7*HoKJljnWim#L(tm$w6iTOmk`Uwtu
z`RVoGA<H7dlWZT)qV!DczQU_!+je<}m7g?jvXIb+T~@fccd3pLFwLQ1gBQcp+aIiH
zRQOWW=u-Z*;8spqaY7_-(K|#8!6P1y=qyRqbi8zEMETZ5nFtLRPuZx}!jF$RzM^mT
zx+=58pY7Xsi!Y<r2<~Uqy}SkwR?v}W+z9@sG5+hqKQ*0m)~M--6NnkAFKR+OSV<eW
zfB57{dCtzDN1e#u0LSR~%FxKzRjuCdH38)lKhO_R3v=J=FUuw<ra#utAMA|rwyW3A
z+Mlau*d^+1r+X;1di?c|c_~SaCp&i)0DGPi>Zq4`cVRhsF`bs3DChFovA~p_M^;5-
zPl2jUfz%(CPYfP)`u7$M1mW&Ml2(rdGJ@=?_D6@d_gGi1_3eVwr1~$T1dqLZ^|QOS
zp6h*hZ(WPo!`!55px--{a~FiG)<e!qr$x&`JN#y>b=^^e^k;?qr=&#KL~biwt0fqm
zd34><`+{=qBl|w|a-qm8om5WId)kk5Z@0&GHpD$o&*wE6t!t?>t)_T#zRwp;el^Rh
zIhcH)bis7>r|5~LAhEp;6P;D5qr`6S{+ud7V?CWH>;`bWqb>CbT2oWg!pVcM&er-=
z6{IQ{Z#y|ewpV_~Y32^mpaHvq+3c{H*^=3p6e#R9*QH=UM}&X8b$o`;LW<Dt%JxXi
z{AG1u;JswqQo<C*S4zy;78UmC|A){yWQdpih0~L@_&54_PV$z-TY+UE1>3B=qQwN-
zss;xxo=YQo<le*}L`y?r8MX*8fC2$|?H-fH?9?-Dj6+JZcUMF1Dxc{v5G{B=JG11o
z<8F0fAl*lL*I?iW?KbJhw~-!;J7NYdh|~$~6(u=!+cmd=5mNG8K&R*ruVDj-&iChH
zrORmhr4R5GqOrTFQv+k29;$>&Mf+xZLTf|97#J$mxbzL%@#*!~t%HvAF0Nojp<~Mn
zh*JiQE&~gv>dv-2yQvG=SuD<WH%ea*BhDTy-hS!kEo;5#;KN?$9a30RfUt;r!lB=-
zf!#N?T3E!nZ?)^_y6;f8F}m$E05Mc=D~*gW>A(oUfSp34U2jXNEw}6J#%BaiUKe9a
zu8>lB7)exkQ1AM~JMYW!vNv)~4xt&;g`{&k&q=;CYSi65|6y=I91KnUYFL6Vo%s3k
z13AMI!_KuLD{^PCTKc-*ZJ<e5C#QERtJebV_p@w4cci)p5#}ARDt#%}XzDVdF>C6~
znRL^nJITr8+Kom}&tE~0X55@0<(!8keHgf?ZkVVIh8gTEK79D(sm^lIObsALS9@6k
z;_DxwPQLRK&}IdXBg4u>D>eYT)+m@bS6EotZZH+C?CHdq_<e?^3DKgDG854mHR}y5
zS6CHI3RXg^Ghg%xS9m+M8VmvLx)WA(VB6H*X}cT>39RnE3zxF~c`U>WX8RjpR$=8j
zRf~R9SF{)`S$0QW1Gq17@+24`?Mdf~h;m?7Ul8sNdo-VDx++V0Bp6>2a~SLqpqa%A
zs6~sN9B-~{2vXev7=24=wHbt&tQNPi7)9^BZC3+GNI<xsu`+@5k3Tn59NpcULyCH6
zej~CduN(C$h97t;SJ*&{br9>>+x8?r6qNV4fW5T^GzKx}DVTJ6M*b383?aU-TMs|v
zj*c1c3=H<bIrPkv8%FXkG`}tvE4?j{=$j>abTEN1=iX@*jX^Ze*09d`?5b`ija+4e
zdl<KOWbw~T_HQ%t?6M)8%j(Jh1l<0|UHnzTRccsS8Kok=x7#3)+&gG6yz)%IMo}X6
z8)XgmO#WTrip5tVzYjN_(6Y^Sd4Rl{V|i!1R#hPkT{@2{)V}yAhmU-kytM4?<<B+&
zC&;Umf0iV5wzRC*i1c~si)IwQC@ib->amsatZw&hiSoO7(~7?|%D4j^-q~rRnJSiT
zhOoh1dsiL>H5sv5Mpqfy?we}8I6e;R;AbR3?^S!v__*6=dPdLG=RI~>Bu%qAK}8Gi
z7{10-+4mMYU`pR|2S8eKiOLKur2&hgb-Q`K+gjB*(GFQ`r%p-ylvfK}oy5=d+&TMH
z&#P;1V%$yp`o{^$75;{gUaETh*5Y-4C_@aP%xkRd)mhZGx(M01qjj`ms`8US=cH^A
z75$)CO6v=}PFLk;gr?X_#+9SPST9$(VKJCAaeNlm0l52^d}sUQKT{bk86z*Ls^@No
zfW{mW|Le=|KbV2+*h0$yl+0cWxX1FkP|Wsn(rWP6cje9rN0y1*BX1NQyW!f?(Yf+-
zu)#arwNDWUJxse5xSvA4)$l^Z5Z|$|NBazC995=y3EdngYv1o)uH9+00oW)idxyS}
za2rc$3{Lq6E&ms%LnI7YApo0%@)-$`DxG0Vu=bwUEaIH7Z9ta0RH{S6GW2M$^xfhf
z0|`jJ^Lnf+vRX#cCPDn@fR1Q|aK%iaAI<g}dyH|ih6eELzi8+07X1-dysN(q+-tV&
zMwG4Zx}SeSh`Pjp*;%ii-QLfx0DF!TyxtpAz9RPw55nwXvR^1EQ7@GVJU+|mdCB;O
zb~)Eevh0~-O4$ahN8IIPuj^%%tUXnHqI~3VI@*yZJ9zQM(Uscku6~fXp*eZ}<IoFd
zZ({vdzhKtKZHKJo`#bOY=Id74Yv%Y?7AKz3T(JhVsgHZ#JaJSfxQASvep-$0^&0N3
z@pf{$#WrGKU*bN|!ScpG^2?{k&tpU9+Ge=~7WW1xW*AirNTgq(IbMo<Gx~N|`P=@?
zxEJYhiLupHmQJORZze81l~2D8-RSt>-|+R;yF;<Be|zJp_fY5n7@eTsE!NXG+sksP
z@m<EZy(#LMidVW5WZ{sLGe65N#|%sfKM}rdkbV&bt9Ww~U$o!hmc@$ubtw~m_Yq3)
zC-(-oBMr4&8f9eCHqc=q=SX7Ek$b8xPA|Uc#@x2+d-XF-@a9pyse|rE9W1*l2F$5H
z2gY_P{6sbxTz_w$=+xsP8kC=7p1%R0+OZ|?Fd@b0B&lIhzJ|jZ`<mON(qCqEIhrz_
z?4}T0joU^$k|ZIs9PolxSjl|{ceQOAcBx^wYP(TNKM&jNf{IG4EfRM_I*V8TB}S@|
zzd14gUVY{(ZxiCw5nCLNB;t{==y9uZg#abgiC%LC*q%FYu6>ejfP2@<gbEQMNL#5z
zO5{mmj?z0smN=a3aUf}36)<>tcX(=eil2&5IYzs8M6)FdYhf7>;3y&c)3g`wvD_8>
zD>qAu?cc@Nzr)D`o^*GA@~X;iC?qAet@Yk*2^)bkWumUu%9Wo;1ZEl*K%g)-Dlf|7
z8hXCD2SQWZCd-;OTGN{M5SO;s`n?7*soGSau6V?XJlXYK<A|O!x(D}#W~Qe4{oug8
z)-c(&RZobLFQ1=sdE~&U@1?_ug(x`7>m!4qclT@nu?d=mHASTSZ9)FBnEeJeqcNN$
zSBBJFUxHU)_(0B}H}o2JT^0V;j`Pzf>x}mIbBZ#8n(o7OEn#JA+tP?O-qH^Txrq}F
z8$exEB!%bHzjj7$z#XknM$Ngaf&0O*D<{cG;W1E4R{4OlG=Z5PJMVrl@~z;t{WzXO
z8>a<p5^z5vCMzdh2nM=6wQq_i3E$c0;RS>vKPwsM?EBr(!~W+hYN{+=<2}PO&{>7B
zk1TI5P(8_uXQz#XtE+?cyDyWADR=M$%GrrD`8TJnjF%^Cg3({s865jC4*uF)oAG^F
za=bC=!J7eofz3$8Y@2~#H?IIXEA`q{w`KI%#sL56;wNJ(H2iKBTr$xpcz@$86?U9`
zni=W%pS}O@H^vp#QZzR{$to+cN(pQ#1;ffGW<Pd|K}`Lyz$KGc<~nx?RN18e>M48R
zF0bLjxb2m&W^&7`A*URF-e{Wr6q`zpj}=Yi^f0c;hRl*)(z&Va-Ofx{qPo2^BF3=g
znTUaus{G?(XSHZgFuTPM)^RQ)F&^A20RH=I^D&9?UylCo_kIhhSQTTzsQg)_ov7n1
zV9@3NecD8mPl8Lwcb3$k%HXfL5(6ix@orgF)s?~D+_@x_{V;b}*G#RZt?vTm%e5z`
zj#W0Wd_oTk=uSg;u{>Wzk)m(5S$Nx?HyB1k92j;H2B%RyLKsIG8rD_)2))vWR<ZQf
z3=)%YRxtQqi}e4dHV(OUO_Wt(Sj|D}X1^Iki>3E(^dAU@t4NcCi$|}CiqaiU2TcH&
ztf|@K+9&;1Q8od5V0`?9=`TMnaH?c__ju0ps)NtX2NwA+2Z`*?%h;d##J%33?y!0`
zKT%!rF^}U-;Pc4I7sK*)mPYyg9A9y)6BD&rFPuNPiRDk^ScDvXlHOKl*_A+&HEi(9
zY?N{T-h4Z*@Ii<Kg#BGmsSo9TgVIA$!EVK)&HmH%2a?8-gy$urjb+DlB?C%E!<$;Y
zrc5dOvosSszKBzz%f(CMGatg!?!Wa@UI<*U;67pZy3Jlo82WPkR@CSHIa<}Eq*1KB
z73e6DM2ax~`m8jyB+~}+!>#pV!Dk5xHAkyp1wlLXZznr_XBC=vmOM<bOP4B1bSR41
z?^4pH5qD^ZgzoMb$qY#y9X`I@tUmytXxa-BR9TJM(=<L>nao(vpxT_fAHk>e&Sp3>
zADi7`XK`9DmyvGGu*Qt>_-<wy6PmtNkRN1?yB;B-7*y>|SCduV9-%(kYA}0lduK)N
zfJZOVr;lsrva;dxKB<_-H@jsdi+Z9C39mHYmR!Z|;<jac9jsMReH>%&RVshsETM&c
zGwI2`f8%%gYoLxFVOyCor5!QuOuJ`<eqhaQD^=TIy}o|_p^<A|B0;LP+S_BaYPWTK
z4i>6NPENjO;pF_dw0=d88WNqd$pfbTcNnET_2YZ>CK(o)2yx@nT(OxIdqqU8Rw64_
zDrCn?`T>ZS%KbtIhe?Ji+)~A?46-su|FG?rKS^2Xj1V7xuhi=IeH|`?Z%+vS_~iWB
z+~vSIyqm1PX#c0HUFYLpIGMU$IvSgM;;<r@)W!a>{D$1;F-G;q-^$+WpwoYg;mNZ}
z>^;9GB-|9y^7F`t<|FIxkMkZ72)v5%3LAG5lk@UF2LBi;KZrF;uT;%A8UE7uRM9hc
zi$ht7x}6%AT#{dIOUpBx&WC4eN7Rewc}>_~ao_#lF!%qgc&GNP`B8miG941;wL||@
z;9!H_9dUhZ+kptku6J1-TuRe+Jdh)O=EG9kUP(qz`?DCGDeTYvjc1pYzol+aU19dR
zZs^sv=8KS#Zr+rgxi%=I?d{4rkFOcZfkql-7eZBEPZ8*B^%FBJ<E1KDDDz|Tb|%;|
zMd2hIW#$4HStG^Q3*7Pz2bYzekvU!ZDpR{Y)Ru%=EIVz-KjZX}t{uxdXp)O6@|r`v
zOYQQfqQ2kM;JCcLMvT5W(b>vqx8SYpcg##KL+$n@7TA@ZMu_ECa|OEQtn0vOdM0$i
z9o+=lj-zHK;}`Q8M{<zi0CrC<<6Cz#Ju`<Ettv1ogy_@l8F9q^2wBbYM(K%sQhL93
z)%TOdN=)%;V%C~?NK+-BMY-UXGu%)PveBRS1@j5{Bc74M3umRANM{U}RK>UaCN0cJ
ztghNGm1b`_x%C3J&$9TU_`<p_m2aAGSNQ7HJeOt^x8Q1*39|#q8EeNM93_Qk&7Scj
zY3eOq;m}yVvH{F0yFJ<fLS!{bpV+hs^Oq|-MnW}aua>Ppu{t^YHQIP)9P~%J7gLa!
zpNShW_XBNZI@8u*E{x!VQdgI{AHm5&qM24;@@yH_^Nw#(;P{iHOsKyB2hYG)V%{?w
zm0{`V(qpunqXa+_>t*wl&`P5J9B~R~q#4RV4)Hh_l?I5OFr%mEBah@T!Z}JY{5<+#
z$BR(2XY|Iy)mOo&NMUNDXAphoi6f4fA0Jtmm7MQI6?Lh^$gl?9$XnO3qUnRP1dtN3
zvk)hevfh&FL9b|-qOts4|6Xs@V<Bk7Cq30xpU*@owt7b}%A}*5HTB?;S2IxIG2nHU
z!AyEr1U)xfIMK6*`n-nUB!kgyrq_HKyLKX2GOMu@UpI{rn92Mq4-Wt5H4Y?<VgpOD
zsi=90IN-r*=~gN)OpTqpp8@7{H8ga6a|@HT3_$&0Kgw&W%<W<OYBA)u(?I2UD(_tk
zomZw9Ru)sd$oGC4(5ld7Jc?iMvzQ3xz1zjOCr2qN`~noOMBR;~*8e(hDRARz)Lp~1
zeTMUvDhU8jnjJVNO)8R>T3KKoR<H(H#{Go#c8dvB!GoHk<J-KG+A^>&@mD%Id!XuA
zv1frf%L35#&ptmUfT)A9>u@6#DVQJ4D7Y)1p1zC-;XQyxIbF_m|C!mzi`W1ziA$BI
ztpz$8J~0&nX_Pz3=<4KUTxK0;9L$hDl4hN~3})1IbQDVqE}2PNZaKiDwmnHxn;{bn
zalwUSJ+OOok?q|m0n-gYQ$)_CH53=*vKQ_e2w9ZQ$_0^t-z?dE|4+CAydWd}UZBA;
ziY$-XatvJfV{Uf?TcE!M?rfh2JVX^{ws4M`3%F$_6ije|Ncr;?we5daCCI{gBpb1U
z201f>{7IZc^n@urqW~$BgNy{SlXJynmEB01hD%pYKf!%6M5Id2T42k%UP;WRn-*;V
z-&E0ai*nMA^*QnUMS=KPWxWy774a#@4WP(tlmArtI?TXmG3oJz(&a)b5ua3f67$z}
zENO<uAU>z2iAC|)x@G7$)f{};)}gX)rNh>7BgCTGN+y!=>O7jJ_}PP?Z~%Ei`tf#B
zZWdEifKb22!6~1q`8mRRa$Vm9On<Hkn2Z)f`*gv)Sw+hTH3xyH&wwvoZzZO56`j8(
z_L|$(0ZgwxurC_q*<nAK_}`SzOn&5BAg=|(6a6O#qNPk<CY^cJqZgf!B%EvO3FT)M
z(R~Xoh_@OgT*?*U9rRz|spyAkYQcjz7IKP>85Oa{P@R|RtK~b8Vz4qye&Y;rs)`ir
zDsfSdc+2Ya;gK?v*sNWbTm4UXdwZO4P4Xd(n3gQh4ZaLikZ_*AC|e1E&YVwpc(!A1
zwo5JbHK{_~Qr|m|=V{h6Z;oea@Lior;idZr&m2xyH+`~nHfhxOayY_Ge!n-Lr`h~T
z-`D0JI~VI-t3WAsu6)?%(;7FP#f@a{DGWIwTxYofnE1kTS931`W77P`srIjYz9diH
z8gEs%#pz2Yr`{dm(Lt1)4~s0^Z`!cBP?lZa+IX4;#`qb9=Z2FXMy=guRc3tO0FVlW
z<3jtF+StKZKR8;xVbgiVsT5B9%>MbNPTH$i$V6aEK^o$g9u^wqD+=bFcf-q>ke@zT
zyInfZzD|3~pt&HLWEH1U1D45b#PKM~s3XJ32(QmS^=|5pGUMu>ErFZij&D8j%`GT6
z;Li%c#+$|FKX=j6HXHwcu|Py0(;5uY;1~+O-+~+t(ttIHVaeIFP>h$!-T*R#MHxS_
ze~>wdZixt5!H@3}TWVt?cEwYOPK+zYDwhE*e5GXhpzvE%#Xgm(l<gGiZ}inZe4MFU
zhbk`$twMZ~N7cMU2PX||>HfHHUp?c;X6@3%uDHO0bx>drFze*d4Cu{j`dMPwh)BoM
z-mqBU;TNbj-n|<@X(zS=5Sahkwmwy=BHE7E0}SzU8dMu$@Of7nSJ+-76fj>5qk;>I
zdeheSV1fCP?)9-0<he{_91t*DjHAkkJ$1ll0HW(IZa7`e(Oml})$J6bqli)>k>+)o
zF2f>9gQ(yF0!RbUr8jKEC+XI60ls)b1!C$NR}j@oB}id%w8>)X+IFS(HhCYj5?iZT
zxurMk0W0tIl6qZ4)=@h2<1A`K1tZVOd=GNJAed<#4IEWksh9V8l4}bOk3^8$@KN5^
z(&uOP;OCmURHQzE1kr%cna-<%+eW|ltJKHm)o%dFvWkV&QbM~7I4M6}RYpH|fdZmc
zXn#~JwTMZ+Kfwx4)c46Ia<L)(8wgH5VOpSZ%@**b-J2)Zuak;VGTB6M8n;d{%m2eA
z2=9fh90kJcPX+}3$r6x>-|@zeI*F~Py<&h|zy1zMgSa&@p}B*yY!^fUBM-!#MlrEz
zZb1wxq^fCyllK>-D>E#`{w7brE=2!L1^}iOzYt5qYy#FUz_K0$Y*eG|CSYOoV;cZ2
z&yB<c2M3RBiMzskr2>M4q)f}#YTY(4y#uWqm`S#{rHQ@Tax>dwRR8J3^}>&?8kW0s
z95j0C)MBjdoYdwLTod@FwQ|IE&lHrv!;~DiF7mfop+g|EXkyCQpXrY&W~q$QnV!+r
z2R8lk=tceHgF|D?Oq$XcVVc<(jlL2N2gvE+R-1sL@u@u?gVdkJ<wUn<Tt<6+mf5Vb
zt4c$@j{z)FV?AXexc(0SSM<uKNidn`zi}?>*5ywdz;7f#ZW?C2%UY=Qu}I<gp1~KM
zmnH@hWY+?NhbtS_A}VeDsYnCX$^2r)LM7)?pVs47PH#>DSJcIm9mcag(c3W^@X3z^
z(_QDLnr)PEcW}+K#cPudLVPj)73C{fA5778BO5&O>yFmMI~+>&^|rXLf#)1%8=TBe
zmdq)gA|Ojbcx|>v7A?BZ2tm9|;j817oq~zlOP%E&f~*qqwu8^ugEw3JSN&s)kQF?a
zqkBX%>ZpuS&W>lbhB1P6sfjwB$6{_-Gh+>v0cSHbUZoewnkg9^xNr%|D?M!!bd8JN
zkdBeHIg2OB5EEJ9%G9}j!}8OGeNCu&8R84pxP!t^3+`6<%*)^tStg5K_3wku>MK$n
z7`6l-?&p%8)jYNNTa@y{{%iDQ8iIFGoU=^OurJ(xemgb+P_9WscABX8YS0>3<~Rga
z(h}A`N>8Zr%_AQM{#e3)S^(9yKmain*#U+(6RD*&um;VywI+!^N6iMRv;>O|fv-o7
zN3jtj$eRF#_3^sVW%{tdKifp;x}n~$9#yZ2%ql}kn^jVNAd0rMOs0rpvevM?ixPHT
zqolP3_H|+^Mpp9_^WyM-Vko?KWU$48t%H0j;6Uza(6cj<3y-PK*tQVo@F1OcB^k4d
z6Pr-w4OoKq`6EVFQ3-%ynV`zwCloJ8s8!rALj5s{?UZUY*j-19ksqk%fL3)A=PX2;
z<P2>AL@AK@KG}0Kgv%kQ^<cKafy#9?O_~MgBCwrO>>4!xO+`qRR>R8ky(B8B`l1Tt
z2+Ic0Si?7O^&RFtzI_Ae3rGDJ;tHAu-ys9;^42ZQe+%ZDH|SdLE!_YFOz@Y$#`c@c
z-x#Fn>C^WBlvj>HKjFm34)|Aq=#q<JbwmIOqPTR8ZSYAB<*gf1Dw7cjVExSt2fyG~
z6JM+y2h5XRYi3nf@FqtY(D$%dRKu^8!{hkDeJbFS{Z+6mrO-u0&if%^_A>9c8iDy+
zgmq#XQuf~3S=OLgLPpblP3TGi?*zCs@E}g7wsSXL4nZ(SIm@|%j$f@1VV3y+K_t%V
zHL<@<wRAt1*dG8LGExD7R3-602GTz{!Pv5;YE&n4T9xP9@D5X`h+zhZS-d(uD}w=j
z1R`RP&b<W%Fx%1gC2&~`UUeRixMEnnw%6lT2RFTtdFdALi2jQ@evyexD2Pn<t?b4!
zkNiiZ`;)h~_5gnocGv&<22i}oxmkm=?9^`#T8tib5McL9T*mL5KX{-Sm%z?Y0EC6?
z`&^yc;R@@Ci^Q}Ji-C2SOn~(6IDuCxI#d72?J8Y_cA8D4&Nl<s7=KMZ4GEJM={m%d
zXI#@`3m#|~H?H6R4COqYMC>^}*l2<7OP`q{8Tl`A?rPcfp+~#X*WPZgm@)zRGHQO0
zHril2?{0?m8wIt5%Fq=*>QTty4L~71oQIW+50+bZmMdy)+)=j{L)~$6rgemyGd{37
zFImvaemragn34#I$~Tk;6)<ou$_wy?F<8tBZL<Nemqg7g;&lL1OS9%ihfD#h3+GTL
zl)E;7NI-6_`{b(io3wWN*!&d0g)6SYBe`6lg9eOeM{;dfT6pt4p~R3EdX!jvI`?}?
zwi)YY%24L7t2w*NG4mhvGu9i4?ZDOx0Wqry1fD|yGcS$O24KZ4f7A&CIVB)zf_;dp
zB)T}T)^n1+db9}dr?jTZvRq)dzF^{jB>OA{|4W-;QL&l>6%icf*v;-3c2@_Bh2()F
zdPKqj*nAYAhZgEXRlTIG!jTJx-L=7pkV-vD_yz!ziZ&($UwCtkZ>B)l>ZKo;&I;^k
zN{Ju5nY8#Z&ye!s6G;b9vvYzrj)$73qDG<^F+kyrZH?FmyPWke^X+Wc4$s&$Q^9L5
zZvZ4kkQQp=ZqWLgg<k|@wS)!@WzASNPYyBZPnlx9xBPm15!;#Bv3SuJcT{}GTK7Ae
z^_$!-D9v1J3OwuqV_mKuE5-+CgYN*~umA?P3u=&z(glh9lVyBazC@)+7fQWisW8tr
zV*~ah=si6Xd?{cV_W`14GqZ)WCgA%Ithk?*wg-;tn?ODX?l(;USZWUeJi>r4F=AGO
z0Frwa->*+U4Xkx9ur1g)gU_vmwQH<;a?b(8B=gV#vj(PUY+|rrjzLRMMNRZ$EEXW$
zup#TeQL3=*!F<bau89p*^w4JyFQu>tIelkag%soc8|ac?{VQPo5EF7;AmiMjn+RrS
z@C>eel|ch+z4+!-HKPUlG&g{-a_R!xjBWD<pps*jQAi(|ocve=*2S_aTZH!o3yVuw
zmJ6?O4&}Bp%j(|Qx4MNf>Ae9A<#y1=rUH#w_yn$rZJm5UzD<;w{X*HBUobV_gemv`
zJple|ac<aM32w^&V3xJXb^;&?SWrpKqmI=vKzasI!#W=7oq!sVM+UN}z<F4ZT(i(Y
zIoc}VcfIEDWGEWs0u>mLLaxLtdw#4m*#t4SbUsau3FLHtw2)Qe1~3Rh`Jmse;5?Au
zL3$Q+(tm*BMSS~OH<=c=0FrltvU%h|-Q-_oCjWxaT`&SE8?Y2SUM2)?vS&iK6Dp|%
zN*=J&Ak#Hpt%^<;`6R=_8Q%iHeBn+T!0q_8t6me_j!h*9!ArUy^R704pYQRh(Tj%}
zPBTCdtiF~Yw=0L|iZk3X-q~0*Nj80H-(}>ztXu(WMY|T?FM4or+uN-yTX#pR)`Vc?
z%?O5ntV0DA*G943f%xO2_=U`wtVBpW%T>_kwOR{C_VXHUZJy+FabEH9Wz}vijY7#I
z_JlvKm4|!NShU+f$A0Z4Wi{ieYpSg2oAxZ=-`O!^Xry(G7$$Pzo_S()UX@<2QPn;6
zOM!Z&s%9FEzSjIVHM0bI3f~*L-QW{4r`lCMS({P=X|!W@rR}WO9aR?6qZ%RM%vgyI
z^-&`2zT8a@*})kMw4dMcutB1G8%~Wc_h~h`g7$&k{|?OB>yQ~V(WR+ryRzTjnaR#j
zJr|AA&^W<_UnX}M_M3r`Q4RjVl7GzNzvKs)BeDfF+|$9O<UR~4g>j9|cz#K&1#}o-
z+@}VEnHBlgc50gHrz9~0N6GBl>v>q5eS6;I$0253SxK4y*6m>jTJ=xUKc~f}v%|Kp
zay_3^nZ6S0kl7gfk3-BW4sD(QohJxCMyC%wu(8t|4p_|`GGoT6a%5P7e!;Kru>{Ue
zSFV@({AAy(^XpWt5_xFThd^-CAwz1#rS<jdriVu8W=pL9$y;OYKf7CyK;MN2OfBH{
z_P{vofXG*_>=&lHOeos)KBT+qpy9wPiUYM~*BHG|#T)yQWq*UblN;qW19atxTDr6W
zlpJ1F-L|IkZ9n6C)G%;3-H9$#I2*>;?acP;XEtLutN*3e*eBQxKtpTc9;oB-U<(ra
z@ni#V)>;VOwS-Djq`wvWxGTl&6`c>yymd>>I|kLrR=n${1BbSek+WizhF<|kY=zcb
z$6`o+%+TIw_cPKI%|&pFZlfSy6}G}KJ#^D^;t(^S*CtJ%-ff$%DE7nMzkY0z<iP^M
zFne?I%*b0Pa*DK6_8+H|m#6GMQt^<|ngQzgX0<7sbI0cvSOTf6VNR*^G!70;)=r{P
z`-N`X*M-ryLE3-3m<MxRKeb-no||K*m)~QL^n=bnJIr*PVU^tgWM@l>*EXBI{ZIPU
zuzK}8v;GxR=y-2pNYdR|5VQN$zz|_zds6KygDeY^wM{bOkx}_76%wm0EH3rmotAqN
z+zN*)`V8shFqS$StLIUr*=kzLi-)C0?*<DS;4c~y%#S*dJFkm;US<8(UqVLqRbDbg
z=K;8L0Jm-><nDr5y_(JjU}l+#sh45V!gCwWClAtm^uXu5D9r_mi|`?hgNjG?^@Sn^
zb#}h~F222n7b%P-Dp#-0G1ms5(h3_`0w!K#YA!i(>n8K)Uy4hBGT=xj6E=XSIh*>;
zSwkEgMd~`Wl)^>~%Aqv7G0tx)L7D1Nm*3T+yg@5jL<oL$I$`Y!%RH`geH?t5J7}V!
z3n>5NW{8eKTbjv7{H+=G^}nNEO|+j<yQ568!Uq=dGBvUxKPrWXqW;gD{mbaGt>>J6
z4V42%bCH<zB;cP;p)!KS3z@26g<+~(+^Y>BK3EP^^oq&-uIL58gRKMkOO1chD~d_{
zGn>Q*N%XN}eX<mP5L83x0m`dZsK6D<C{s1O<9}S)RaWKIK1ATk{Ad$+KPO&JZ$52c
z69@oh)FEgIj*mJD4q+=OKy7pe<(n(xLIwj31dxhwYY<|690(A;`KE81`fy`VAFfR8
z+f-8bfl6vlypq;@W<M=JmwpVOOx~t15wKAX2xhgjVz?Kcq;aP1dVNb(0?Z*!{B4Ma
zf5F#3rk7XVKf+gc4!QhKv~}lrZaMm=u(x9q)@>@AsMNyEm{P6(ME{J%pXlG(J82@b
zN2-%q^z)gXKf%y}o`CX7A2BeFxBfJU!KwP7Dqo)v5SCTdWWw48kjJzY{0o@jKVay~
z?>v|F%Z%W&X;WYiAQu_Nceska+yFi?EeTxJl)gtzztdb7_$$r-6HotLpdThCrC879
zfqsX~S?A&bjS-Ir(2XKN$QT5<h9d*fecte56$G?Loc36qc*fRh;uzOgck@ez=^@j!
zqk8Igsu)pcFG%q|^BCl*r$vtHVeRQL%R1<X8aE`;e%b_2hW}-ukh<?~9>@XIyC*{X
z(TsNga*Ltll-JJ8YfO`mXasElyPMx>nvRp$`KD(3PUj%YSkG+(*L?tsG3<NQ2EZ5m
zZycKG9HR{X6DiI<lbNnn8-;C=Dzr0gh4St;<R5K*h`ZM4K}f1n?J-uV=s_r8#jTpK
z%}&0=2`<p_4$yJvnfiG`c)hADg|lYq=g^Ks#M5ip5$dO4@7h4J?QY8N;!}k5)Q)iJ
zy=he^JD#pGlCNH2Uuau(st0#46bmS2dXGD;eHXeNJ7m3+mX-3QKcV1uQsW(`t`aAO
zbC{X~R?S^6?p-a1ayy84y$Fk+d2+{PP1Aj?`qBg1B?p(YW`^C*9iR|hJ+WO{saVzn
zkCVy?oo<F1t+=0v7NpWGMGn%_!YGRo(zfCEC*I}T8|6WbueFQ0J`#SU_p;OH>b<py
z@T2@mM(6FSZV$+(S`(m_3uUJ7a$RN$Mo)$?@$s+woX{#}9jSz@u5PQmF;$kI-c&xn
zk}IW}+nBMQ7H1W=_^`j@se|ITz2jwz+s$(>KFq)f?4o-uE=y#zIM>u2v-r_&_t>X9
z(Y<>#IsdGeBW_`)pj4!Q5j)`rMSah<%Nfxm2f|jfgex|HB|~km?he`PDwOVkAN&c;
zTR1*wB+6EjcnP(tb7h&PteRcAHfK8;KE4>fEzn(p3L}yljcAjIt6n{<%i|Mw;{8V)
z5~4>S3xS4=DaR!IRE^^l4Oi_i{B^OeKxMJ6x|!bIr2<2U3z=0$8?VnUBSNbl;V)~$
zd_(=IbB<Y4+c+xE!V8W*r%Z-TGJF+NDFNfTj6ew9FW!94(2Sa8ty@jKp+a09r)rtf
zRz)mT;?FNg6Eh`2Rvbjd#Ut9;%<uzbL<;?az!#$0d<Q>r1DIE7I-eS7_%K(RSE{v5
zwlu_7Hj9zLiMQ77V3diVlW<g7tXSj@Fz|x<Z0W+<<fPkk%FJWV4Zw$jNgT}$jEcw|
z)EQpV6mgu>Sfq-BWe0HCBAD@mCI}iMjTxVTPY7bCx5kTn9!=Q*{1`^y5M?Di40+So
z%_r~bS6Q?-E=dy$B<Wx$5piaQtq=21m1B7I+{{_i1^lP<fTg;pU_T2Fq87rls3hkX
z^5{84>IFE%jZt)P@-gREMipLW8jIzYa5g0)QX!xt$%BB!Iuw4Unfx?|N)pPhG!sFi
zH1;c|ts+K3VKkD+{MrKTRSqL?*WuUzBfs$_@%dvK>x`<Zo`V)+&%{K_X-|+3zPJoy
zd%5lXZ|L&+y;iHfQow%6;-E?Nk4x4!_qvl=K#J>6l`BXm6BCROdy=vaXjDfvdzdn`
zT7sP<5KC#3=(>eLQwy9Vqj|TR$p&!O%OD6p78N_eI-~d~Ezpfzj=(!>fyUem++a8J
zX6T4fAooi{gzFO3ek6T;r7r3t$0$|HgtjX1`!sH)5D~9?CfY~BWdJ!#c!amsSqGD>
z<7INO*j?t}#55uwkld*FwQuu1n==e;lFgc|{WbOfs30}nbnuC-H)2Vf@{6WX^l#uq
zL8298r9<dDycY}=QE6KE3tAwZz?MQuFkyqEGfUjLReXWSeXoe<!{GecBI3LZTSH^R
z{pmY<&t?_2<q?x5M3hz*K}&U@m-^3PhW1|bEGgUoUdeyIrXwmKDZj@T1hYNZ>C)&8
zpyGV$2GF#Vk57TmETY_`qCpFPKz%fy>X(eCk-A_y?7{}y<p;T{SIBzK3miT-SQc@G
zygv)h7v@pZ;YXIr`HW7LCjML!U7H`>04gKOTEH9{qj|_!(l$5*0}`HU(mIQ|$8nb!
z<nw7@qHG$U!4gs!Na?Ye@xh`;av5>J{qn;5lYRrlI5W5QV8Mfp{&ea<BBdWZs=<qv
zUQL<f?nxN5upbw?R&942(mf9W4g33_d962W0CP<<w6r`%BAdJ%!%_N!4&{R{Nd+ll
z;j=pO`nIjxLH+?7x5*;+=5k-&K*eZxxuzM3;hF;34;w)9*>LN4UuiGeYTUt3{!SlB
zm7o#7Td!+0ShJW6Uxl|t1zHpmLkUzZOZxHcX7e#C@)l3}5OYu8u3ZCz9awjDP=j24
z+$Wxyr;K}h3^UF;GO%O>-?0}kr=H+${ff%`dT_z>Q^%U}4K<Jf4m=mMdddnm8rdlJ
zdQ`iDB$q9nQgxhi8#V~Ca~*jHCZQV-7!-^1a%&H%g)8+cBh%OC>v-}eNKmPIGrA-w
zaO7WJz)?hLC)n|GbL7L*U@QeMM7wHEUbRT;iYV<ZH)5^2sw#PP`lWxlZv)yu)Y3oI
zb&%_&iN{(L@@k}YA)z4x%tf0Ux&g=tS#|a0%#c-cU%XD5U6%pj0PZsCVG5AN5dyY;
z;$fgk&(wk_jU-+cPxnJI#fL<iyXGP_)dX~(hbk3*BpS_k@HI2$=Rw6hsnHZnP;)vE
zj6b_`9vyHF?q`8_)`nTo$53-mLGi(iGO^1^nw*u@7KQvCy{cM;l9@(fxxqLz`x<FE
zQ%j<BckyOM;XFi17nE){S3wDeCaOzhyw^Y&hJZpW_IDwM(iU!PhG3l|yf*-7XwTH5
zDE)-=JY$jeDwpw0Sd-u;xGqzPiWm5V)P4nuV5W3K#d#rGBW{i)%r+wLdHGn1F!<LT
zWB+7Q>||Y+8zm4|hgaADK<h&OVPdkR7@cTqgI{I}^b&)kwKwQHMw0cGz>+^1E-X92
z9yvpQ_~l2nr#cqt8$qG_ZqjL6V$NG5rS3-Z?*cB6Dc}NO%AQo=pSatP-2`nbBOI$p
z2a1hPur8@bgkLT=pT#CGi5g2D9FoZMY<9GZk*fBMvG^=a8m!a%zC2qli$pDYc5ZuT
zzvdRyYgZrDT=u#!P<N(of9E=?{-DWH)N}6F#PKC{Jo0-KNL&0-*-#Qbo5S2yZ_R&J
znph%<+zl0LI**;KN!+CNUj%<}w<e=WS{Z-nPm*L!er%{TQ;HY)p*K!{L41_8@I7F;
z4vSykl<f;$Zo*V(o8709W)GT|Pc-;M10@akMAI#tV#$pthJ%k0M)Hn{`BWx+Ko(_?
zRVZ3!ba7C^SaJ|(IQ$Le)e-Q4IZ4$qr;Edn5He`U`%9p#3*y8dxCDwiFeim9F@1YZ
zP_^rk_b5#F1|WEpG72K}0e43!BgZibK|i1e>UPy-rB~brt1hSnjZ6EEoO3^>Uai;Y
z(Paqk40Y{C4@@~%mIiItAAPJ<@H`?=WIZDn`HZ7d#H(U_5h_8dd#?=3F0D7mkD_xx
zo6$F15D-4r-$^!z#luDuyM`YhP<+cnfi0GUk*rr41>3lrDoHyKZYWzP$mzB@$DFm(
zD8lQS9PkC}V%Tz5P@<`QpsiGnSAVfvyB;BIziaLsF$RM1o}Do9Mp?zZnB;<5g-(}K
zZZ3^@>1U30PZvA5Y1)Nz;2vn&IVQ%sgc!f{O*E{Lsw#HKjr<G;BZrP+Ge0WxQ>9`<
ze271M?zQavh>5ar?(=*^Hw{mvd9l3WJg~sMU1+5na>&*~R+{t`fM2K{YAPS5@225k
zr~}`QC6~@jy%BDeFOvyem?o;od9;qAZ-l@~t@&yDLCc6FD0+ijpGyQT3=&jmwxB6y
z`&$2kmJKqqdl1jGc!cy`6<Dqq!VC=(=<}e>W4ugYx$+uzXr!O<O?oMX0|Ecc#o%#i
zo#j_y>cSU8i1+h+)-}KzYl7UaLRQb0qKb6rGQis48~B13DDrN{g1V07>T9J#(*v7T
zu+8{?aQo0zNl!BbH7ec9f7B4nn?J*W=QN*Rd90EFh{3zLkv#dtS$?KXsPwfk?I0Zg
ztN&ZtITleYzqO7B%m_DITTs8ZrR`K^#G9?HFe6-vZ*56%Ybj`h2>5{d&{e)Zs5Nz{
zyA4)aB?9&WTTP@(o+r2ki@^uE89e!|gY%%ZEXd7Q9wd3c3e9E%Xjc{P!Z-_95Uy93
zAg_{Dbh!OlNp_TN0yS@@`=F8*D46hCNMZK2U1HQhIKCPS4OeBo-#oWd?1ECMA?R~~
zT6)I8iFqju^YwB?-Hu+<W`&go3c`FnT|tec8c0#r*yhX~Vgj`hp6izj!3QYa_3EKj
zg%Z%G6(l~382l+GmP&wQ`|ObD?Pv~ZG>r%5>*9)X1QR{HJ%%oRH`5~A3I>@(O**-P
zCYXx}S+D9(3kr%_E*zr{7njp>-9T@`wcb0@g}WfN8^8}wls;gx{B}rfu~x}x{c<bl
z!j&JvLZx2@?Qo}5^g+vE1WPk|FWeeMjco0>IJRT$`-0ely$i~LX>Kdbk7X(DnYIJ@
z1~k9j!o2}eq`=15>9T3W<L|XRxlkqshy6MO+66E3iUmv0+^z{3e`wyil7Pwx$)FHA
zh0s<D$I7oKy<Z<RoKkbad@=ErJyv?{x<{>&JrlX19GaK|yz*C_T<(CP$xkc;kULf}
z6zx$Pz;xsf08MtVYk4!iAJoSm2@}Ec^kBKQ!36{eQX~bn6rE(guzcl@86HE%oB_dm
zhHWbXm?zl(H&Cw5T%2#^fmGFhDSPlYez*|)ml{*H_L3R;Z7WNtfv*_a+t&u~Fd1|c
zT_x=FEj*<s;jFRsdi4<G-+(_RRxb&%_McRP5GU||QVrw_+cKJ(3VdP{$tWh#k3D|N
z;$rFZLDQ-Fk#O%!E0Y@lYXjQ_8+MJC=SpHUYl_0=EqIV8a;P60$G8|Nd(Nz0C{+o#
zfDjDkkjC`yh57opVj0lGTUg?kQ~^1rgiGwgxl-PyI?z<hn@?ch062wvslPR-Jk@3m
zwX0|A`Eb?A8H+w01u`V8By#(y&V`RXr{<rCGtQ}7D!|3BOjs`L)ju10-6hk(*wFZ_
z{N4qK-o<nKqN7*#q%2*|zw%?H<NMUb6b(&<D8qWauIHxRf_Q<D?j^hZyZfx-E}yEp
zg?qeD9G$UOvRD9m3?~K+Q@ZghV&3m>-mJ~`d2^_YDMF!`g&R2%vH|G#C{ZV{+~Z`9
zxxmYAx3rlT${G?xup=ihD6ZAOE4wvw9BpiD1}ehy?B2~2TSM=$hJY444bzINyKQKp
zws5s0{t3^+&>E_`ztlV7+%><q_HXyaTF(=`xzRczz^;gG!r!AOIOi?2)K}U@yP-Au
zj{G1FCH*Dk%j@jy7ZzuREA3ln?p&&v%JG1}ux@d1s}-9()Ej0n2=-+^DxC#CPGK|H
zx9pUHP;%?zJa}$FCF(ulc>>~=!@_8H&GzD+vSJQ6Jn=jKB@L|{r-I~s4O27IcYQs}
zZ{&2mqVK4fV?}Y-AIXnfcYx~?Zr}SB`qi`i?{~!g)>iuR?c00RyCl;S4((2c=wVOx
zJiH&RAYF)D@J3ua<zT>Z>(Iv`*IT+HM2Rt7&Eks?ZgqJ!JurZWZ8;C_NI>1ccH(Xz
z93A-Tnn_|d1f)BtkKOl-w<j}nYU0S!9OBvfo}S>irR~G@27yWug|CfP&Jl9Pvp%yD
z4N8UMG;=Ul_!QiTL|lK#0h11c3&DABMiclJ0ZW*D7DRRTf9@xkQ88;`wSmRC)W4<q
zJ>eo?ts|Y_MMYlWQ2p!0a=KyVn9`Rlwrc@FT8n_RG3=76UX~9M!LQ0pIRWUUxtIcH
zV#EF7=B0}`sL-r0u4=C<&ydM1ScGA@b87;b<-KFvG-!8Z-85tZ%-z}7^vGs2Nnu3!
z1sNjvQeyEH+>eZ(SeXM1$N{PXoBO``|Jsee9f?!Lo-xEZVmfRu1dAYfS%*oqA+F%R
zl`_9sr0N<XSsHYN)2;4|fm<Z>^=wtS`$Vq$Z)f=&qoH4@PLKNR&tg?#Y9jyQlr{0=
z0d<3lCEr3b39|e^sR$|T1-S$naEb?>yWG_h#ihEn>i_G%S>x$x_<W1HYdi(xWy|>!
zVwJ~5zWMVU3Jj5&D<2Om_o58)v~z9khJC7|#LtOxY1@>MlleExyIyILu%&mJ@$Uzd
zRZvmE=YG4kd3EwWF}GbXqf-@kzHR{Gswsf9!Fe&BLA>h+XnY53f5iyrw;BZQTHL?C
zqXFF=^_`}=V|KO&(=N>;OjrR^P01O{Td%DvSF*zgO3VoFeMHN^EwToW>MOBzQXTRE
zFmW*Wm2xr9Yp%%FhgFr_=;VQ_fJ;UKN-iUnZaq88Sb6n*#aN72zMNwB5V<X#hktLN
z?l%5;UCkU!M0015suUD<tI-M4O@tkkITJ+q>OY|<y&Wbwp72C7Yx!B49LDzLg;BI<
z=Riz;<sxY&SV{WL9?Bbm<4qE6xblb`m9o-oDKh5eL7HCoko}vRTc`j3_&3YrgE`X!
zQJ8n@dAJQgf*hX;(5;+)@Knqk$HQ<B6^~AL#reUoD&kW4*!_sS<rXrRGWQ7UOEav6
zn~J!=IB5pMakd#&s4XIOEPLtj5^yZ|SD&_bmidqY1N`yLL)PsL%X1TeQ@tT(41k}J
z!$76U!AA#S9ZSJ-XZW%VV8lJF-vJ=!h=xBFC59zDXPv7~+)GvE2cGC9S1Ezt;D;5o
zg98Q?WZ+E4bME<7`vm~H%yg59qyq9%S-i8_bbpqU)?F8V;*IelzTloh`ze;22HW8#
zjs_LW0Rir;=qDQ1bU<%GkJJh+Om*(IZ@$N((u9CZ6YElTAu0Vpc=iijeI0qc)^3q`
za_Q$fH;nJjRlmaavAap3vokDM$>lYk*rMuc)l*|QtnoV|F{mmxsd;#E`7QXga>^2C
zpBHmjQx&KxXk!apB~4CK0U`eNyHuoG7>m+4adr|g%W<EZoCKp~f#b3{GXf-OmnR2r
zQix06E{g|@Q}zlrm#+lI*W!<?XkoFX?lZLYcUTIB1n-+_V7Y^-D<(Vj%=Q6=zGdz`
zCLO`7s^HC?;NRv)B*s4)ICoT8Zxk@sJqb|`W4WQ0P=f1#x`pX^QfX@FFV8=tjGc81
zjc!yeF7>T6q%>}>WBzvk|NZYR0?w|Iku)Jjf2P<&`DPgP(oUgV3Ddg5=>P`m1Hkp5
z>5n^B@I@)DuqN0*FZ9g8ba2P}Xp=fOMV05mzE855eMwoSDV_Uueq@O^M?u(w9%j#-
z*PJ}g8FTnB=wJoHf)M@<Ht41kS>QB4Lg=O0<8PmJH?@bft1KO6tP}Fa`8>KxzT7mW
z((a9Is{wCd|KUH?E!E!B&k4D!BFz(oQzbZMA(1viQE`Z809Vma68Gbo>Ag7DiZS#u
zyBOPe)}5S07yR-64_0#@rrM#1=|FIqj=>gb;^qn1N5JVGn+rH~?!KP6`71gOsWYwF
z!W2E{!EFzdgl1<`Fj_X%UU$?ZBbOyQ2+#91HQwc~ecx84wA^#0BrxBJm4iLOA<NHE
z0y5v|yj>7rpAJ+i=ZJ*`sHr&L)M||qyV|;tmh0RMH2@8JcN;qW#f&`X*iFseNp&1s
zu@Cs!oL|#5<0f{XX{@FKb8FNcwT?SPT3XYRQ&?qtm&{8!P<`uvarf3?ZFSwEFBNE^
zNb%w=THM`QtUw?*lmsah3lKcGw8h;CQe1;;g45y<C@oqDQoOi3^u2lCZ=d_!bDn2E
zd+$H*nSafdHP>1rbIm!XtTBFr610Q00=nDCWnU-wkS*JlCV%-c=SKqeMgNm7Nps&l
zFqwD2kj6?_qvRi+Tj;s0^yh~B%yKlI0;O(onW|BT(0K7Y={E`Og%EwlZDKIo)fdA=
zvFJ!FXGV-f+^hRUI2lrgR@BwCF%7eN^EY$w|4gI?WHXC6-O|1_p`GvUcZHWdA5)mY
z=>1aOl*5TL&_}Z!96d5|krzwHO~;^Uc4O#!rIhzvf8n10p<St@HP^6pFmp8qjN6Sf
zxZ*=MGt^`Gctd#`n?v))^wZZKz5nv}^X_l?Ipej{>7QR<q=nt}KFwMoe&LQqy1Uv$
zLyagszBF^5dW$sAQ2vZIB~*!gp2}RvER^C#&yMyOX=RrEh4+<$pq_hDT^WKjiCX|=
z-7gAbxN}VbMvN3;iR<6tkz9YuV(zVfxksrZcJPR}P&-uZqZmRrX_6~IiZ$b;xU;IC
zxhHIRZ8+h0>=Ug;*k@QD)5ClM;kB1?f^h%AcN_M#TM{)NjZ)~2zq+k7Lv5V6>=~8B
zcf^`rdkjz7Xs*2gQW!_*nQ{RN-9IJRBv0PE_}7T>f7a!H99weR7c>G*k}tgCn5p+2
z%N(w7p8YEQQy(wh{}7DH?$CO^Y4rK=RF|ij-!-~<rR*{n^_*}0=!RaK6Jnzhs0Eez
zqO`QW-ZikMjr)8gqNDjEe_IVB!&{`o<lto@wKjEglCkRy)}u+f1v@@{A3@hN-jm+^
z`AP}k+#-F~8DAPfE!TRTa=^qOL7I~%8M`{6>3{c5YNsX}GrT+iM?HObkn2^miKB8s
zB(YQ7`hoNR^FItaeIZ|NyM}SH_c?mPiJoB183l0cg|Shi_9k&Y87Go@`UJYLwcWJt
zIIR$#$$hJ4_@Mi*%c;pa3r}x<S?lOs`oq_6?w*AB9zB1&yIWS);Vem-B~H(p_Ss`|
z{roD1pi_>C_-x*COzy^*r#)<yA^B4iJ#}lH%5r<{bw<EI;8Mlckg$N67Vh(Pxu}SE
zHn$?Q&-R0ZZM1{1QWzM@WmcrrRc*LqhazlP+6+LM1(xa<EkA5-MDSQCJ{wi?B}<k)
zV@)0Kj#q~yG7+1g<KNo)8T+?l>LgQcN<f~yg+kuwcfWcv6#4<gunT>-mUMNAo#z6o
zP_bC6e6W1#voDVqIcejo->EI-vCc>}HaAIYFrOW~)l||Z+a{#dv1>+Nk3T>t%{NIO
zbY}Oa)GREZ=p_m)hs$n188Kwy7tm=gKm=dNRrfHm`{TS%7-v1rCyGJBxGoTYUT)tZ
zO{zNWRA59Hw}@X-`R>89sEKcge%Y@8RR?!vkBwp)r?mXE-UDTTM|5sl5AvC$FQ89@
zru(x_2Eaa^|LNhzXot0Kp;QYg;@HU{5ZB_;QPq`GZ&bw%lnI&2=K?)dmnFOeatBnb
zMYdJnxbllOHSCOi8|Yx{P@6$jyw+Ioy5uXbx>H>gT7d|iBz~;C^5OEIUQ{Xu&o|u<
ztgKgFwZ2Z>=RuoJObp_6D%llm%R4_OXS*rxWK%QzOagbdMU_1+Yty}jw)md$?ToZZ
zet2UJt7*t(dXD$upVn}{KqoDGasCxKwmG5Oo38%+?K3y`4u4U2qHS#tM-)4?9b1~m
zsR%N|!@^sM<@r00#jnynOEoU9-M6(m+??^D+eCl7dR1g^J2G<b-|dsnnUtDOiu9iF
zH8xJRe<MF%*K>-Zsp|03iM`!-1WAyIcxUFBi5llT{4U!dqbuyntw3Jhwltk0_T-ql
zbQMSCXxrp(p)MTI;e`qJ3(VR@&d+eTaSVgMpKi;?+za5yzy3h|f!>FsQ2p59n7SVa
zZ8*X*Lm<f0CTM7r;|H3(QUcgS4#?AgO<I2SZ@K<kxo5+3@J3I$k&&kjt|d<QSDfS(
z+XR`}86};T-FZ|ibxbpyX1I5!Hye-X<=si<-#kY^)ST23zP>y)w#UP<@it2vd6u-P
zG24?_=4zpmD5cR7^EH-C<XM?sy^VUpu9Zly-EO{9=S>tq^92DoTK)0JheNTYFv+l)
z5U+L6EbfT%5F^WIKb$#OKW}PvAZ1&#e^u&CrYIjZZ`Z0a=Q6(XU8S*TqSs)^k@znh
zs|VdYDB+n*e~IB+nPeMjeA^XrYQ@2>8au(LiTs;6<mS9eRAH4sI7sT#J1z=?j3j@H
z4c%&PW*Ys@GvGn)Q^%5*I5=>Pli9|5|I?cKr#dNW_9*Y{J6UjGUcBqYUV^w-Q^`LU
z@@h%~JA({GI5>=4EhdE+5#Q{uF8a$lWhr(-<uFi{IzaoE9^x>=B+{K-JwVk)#v5BF
ztvgZbPeU>?K8Q*T#5T-M%-xIiQqenqv8(~eNw%NNMb+_3iSJSt67kw59QCK4m9!`=
z$e#8yyAHaqN$qA-ivWW0H>QInbIjWA+%6x?L*z&d1y^D@mioU#%V#TkcghtXNo|!Z
zPi=7!x?HEN4s%$G66=v8GP~`jez48zZ|H^;#d5S-F-uZHJ<J+zz-v$Xer^3s*mC1D
zN!gS5#z;MXy3W)xl_r?6Yt$4~HcJ*6^;xm9`URVcnqC<@lMnul;g~7j$K&&c<)*Ip
zvVrkRbMNgMK96L|>)IqtI}4;`HZ3BR-Y@`oCi$<02}Me2p$#0Y>>jFGRlH>doU5`L
zL`5znOo#Ep=?0MZz(<RxCa-#;=hFLhKe~P|7H1dUZU3aZI;f!Uq?k*o;y4oeJds&=
z5bX3kB^f6m0f*FZXxiZmJqLUY104H&MzgxQQh~aoScKO~G@d!|?tcrx`beLje%kM-
zsmzl6Gky37PWSgaB!1mUUUQ#sG{6!pY9@C2kO1dCOcqO=lU1rWHHz>4LbKD`r-W3+
z|A*=*@P9K43vEPQ%G7ijVo-0Jz&Z+oOjT#PESDN~&P7CEu>R`)C!II2caJBPY)o3O
zf?jfo8jyPKV{UK?S!4l2>j!uR1;&0iE%*P1Vm&-LK4!iSd@be$HMsHrdjA8~H{6V&
z#Q&x}XQ3M`_c%-P;%tQ4C#PM)?fZ@IpVxg^dcIs$CqO#6yy*u{$KnnANd@vqgW}lr
zsVP82AdoTt?&)5j;g<9HYA-|VVbGI5;U?Wd^_jO?`*xG}DgVOhZ#$VY*zWhkd+eKi
z5@3933o$xC|5+=Xqk`3FroL_Fle*pDY>d(rBUks;7;3PpLK<6_jN9v`RLexfj6{5y
z;UM?t4oLIVT+nT*qTS$IxuH<wdsAup%r(qluA_#5SF#!5YF}>6JNdjUVtT%84!Hu*
z=T>~d*DlmGvyr1tSiB~0->G^YH_jz2)Dc6mQ_!IpH=p&XyQ(j6uFEcl|Km`WqX9l*
zlMr4m;~FB>Uu0vi?j1$qGPeKS|KV&`a7w$&h)i^vEZnfePK?KT?o^F(Yu%`sK~INj
zP4zDvB2q9b@4%wmV|71`5KNt?+fFZo#5b{}yDa0E(b%XzKU>6pwbtDq&R?27OAc?O
z@q!KZwYll882qm1dpUV`zE^77or#X=ui%Zu^II#KXnl+VwtM>ow=}CYvhe;aOmYCr
zyE{!TBh9M8a>c&YtG&9)?R&I|>J^i}lR74?&;0SZasdgGhP2}J@b)2r)GP$y_0{J2
zX>%`mC*|TWuE`mJsN6<>=4h9?kE1?+6p+Yt0$R2<mlCQfJ{wW$uc;L5+QOv{2EQ3+
zy->K-@fQZRu37{Gaz)<0_)J0&%_^}8&)ScTXtr>jl>igN-E_>B>^6G~S3uh$L%Tdl
zXL41V`pEQLt)!H}xRoS;R<~-0|JN=vR!iu|U{4|}{<hHb`>q4Qx56%POLaw3UY}6x
zn-V35fGa(AH4q(?Rk<}}+evCAX+jMgrLIBb>C8khV}nbuo~zopG#~>+l}#8BHw66r
zsqC&_^~{v)!aQX`X(bU5k`h=MNq9>Q64J#3rx_w0g3`{AhIqDarnxZ{d-r;jElUiK
zfUyy2r*G!MO-mcP?r${NlZcNu0b6-zvl)P~y!~fjwUlA)W7W~(3Pon_#Cs<ry?bKA
z`$q1@Y4bUjqn^KtET<tko^mww4!({UyG2biS^SWhicyZfmqW>VNfShEoN7>3o7Hkf
zh%nV3u<zM1p1eEJvfysp7a362Ctgd<si>JVuHA6ksjO}^5H{+hP%)U(C#Cqf70b9G
z<B$X<9cQFP@XJ{307Yi2tdw51I_qmO+!B3Cs0WeN&nOpj@v07<>X0QPn8;)&ScK%*
z1}_t^%dnA%Cb@^N2q`v=y4hxHlKfT6-*wDH8qDkTUDHpL%s86RqS%*o%P3Vwu`#Pb
zhzZ?wAtLLLDsIp8(iu*CS&AebPS!^mL;NA64Y@LJ6IH$fI7s>E#rmTokE5NII|V^!
z+{ATmnyy?X_OLegGK{tZr6li*fw!go9&1uzlbXd2KiKg!8>BNwaEAIR3|(u4o@GFn
zJsEx&PI^^%5%vlprtoP~)Rh@YY0wo08)YcwiDi+U{ffw9P8J=&I^5K4<k<uF?wMH|
zgFhcb-=Ai^5GG)0j+!Q(&pDry-RqMo5R<X{9-6^=3OqWLm53rdSyz;}a{jhKt<X<j
zzmkoL5>qM#vSu`@sYy<PlJwg5n12;ZL2^~fg+QkXn*wux2x9T-)iMC3jL8D98<8xX
z`mW@W@9SM$m#g6NxtTzlWg~WDYMBT-<z)2>D*bfU=#ti*syY>=twfVpTBTrivh7(w
zg_e54sYbJH0On32m_ah`u;BddI@U_Oa(P7NRflU-o#QvRTvvm`xRJ62u~az#yBZ8x
z9hQ`7`fAo?8okKA6A9f&3pT=0?(#GaekqAR2S^z$N{4FqwpXoids%Tk+}QH6M(Ky#
z5ga$#15x7*jK5GEPFf-J$RXgDH0WpJDmFmeb|gnBU9%Gqzy4~2*w7CdyDn@?Mfat5
zWWrM4K`h%eeNSo{m8TTg;$xbeu5R!m%XjVrP~R5yH0{S|O4_i9t}6p~i2ZlN;@t}C
zvEk`@@YlfSy=jb>2GV?d<?X)Kd!T+3`grXCB&6w9wHut1>ZAs!U{(s}`W@xEG9WgR
zqgTTXBeE}cqt=l;+sJs5VL2)`{yE#~0N-Vk&J|6S_g!F>Dwfz{$~RdPVPe+?;?f6Q
z$-c3P>kVfWS?4n`CXy`B-HZ`Y6Bd?b7PefnT#)eSdJ)*-c$I0p%ZE`;tgNvN90#uo
zde^Y`mTWMIG}P%4M`<=0myK0w+wYCOM*GT8F}g8lF~?~M{n?(=Jr$bK86@g1+|wDF
z(qr)o{6Xh-TW&J@zCPpgy5)8u#=zDdbGnlwn#&t8eG|i6a5FMvE&6Oc?No#|n?;0m
zM2iu5dvAJI<?w^2^|GOf%X?u|j;O{*%hh@ODISTuzNYOa!q*7pVl@evc*$&{_PSyE
z2q#3}HYamJGiykfPLx(loqu^|WEW{DR%A^smH}=#iMKS@t<reV$W6__+~sQt<5P~N
z%4xG$49n-8oR%V&h%yVMfo4u#Pinqk-~3ABW-7-d|H?9%Ook^m1*z$4O1s!1=uT!i
zQ7TOk&}Xw83%hY`5?Mw%_mu-6X|^h6VaV)vBiXN$l1KDs2dX=Njfg3p$6CL!JEs}#
zBdl7X^9taKmZWIQ3D*f`4)+8#anNyqqRmm8WE`)L!p4fCt`i*9RLUW?CBY<`6`4yD
za?)bfy$oD)#+jO_KZ!C4D`sbxyIYm_;-AIr@$ucCDOonD*xFWDH;d6h#Bx>kI9Mi)
zJ9#C(s85GK?T%O{B_)j&I&a;C-8^W^I3Mf8+g(A=2~)@8H)DE`)fmt)E6C#wH>KD%
zwdA0ttXipHYe7Vh4ujq>Fq>94rKm|)l~lAk`WEeY{-A)S@-R}&c3ny|i)yP6)O6*w
zcwWVI{PF#=b%)SdGlasvmz|?qy13k2qpwYHaQtZ(S(#MiLkLJeBvT_)Nq;HH`t(@i
z<DT(745ne#4#}KD{I2QBRom5iP{>~7hME+bfy@TBP|-N6b3Himh%`*4p_RIosOD0q
zn+l+T^R)26ziqXhH_PrgP>oksSUEp9xZ$Q3WZ2Pr>S~UABv>heLY&%H8J3!4Fn`p^
zSB8atiCY!a&MZQL#B*T`Ox-T0YWCcoj;P!b+MP@JHfK-YbtYTc@Q)dB1LSpe!1xG>
zWRvo7HjL=(Zza00#`#M!Bi&HS-`}d$iJU5w4_MEZzT456t(g5V$#}PT-(uS4NRQF6
z?Gdh^MY(%(kl*}qcFC(2Ac8@6MN|v2(QfTrEk&`NDV9|9g1y*I2|wW1t{L%rKI)A}
zL{%o{Un{0VBfO!0dcz2Zh4cAdVBz$;(`ak0-h^23ND{GC2la@Q(4rxwS0p9aiQj*c
z1o4}|W?8(PHx+2Ivy-<vQ^z=hKOQ$)bck-tN27)|?(^NuO>h69lB~FS$08rzyTX5N
zKI@++&cQTsq!@>5lDxdsHo2P1_*@^&OnAy7z+6=|>iG8WndX0;?&)eN?myrIl#0yG
z)cCCB2X<yl9&oJQU$VI=MLYAiREAB4CHO&+fwUNfNRi!&s&25IE7z7Vsecmcdx5TD
z>T?uDD(iSrM$wQ-Sk|rPffnJl_QI8S&{v??*_oJQ1v9sUPvh%PKj^9%BTBzGFFRN0
zU6M|eH!au5*_Qo5aaniyLQzlnW!rfG3!W}!x4UU;S^hrrXKToA!%|;~J+{Z7=>vrx
zRt)4dqK~sTpDuzmF>_>6jg>Sr-XmUX)3+*FrUC6Na%p2NwhJyXIsiklNH3IEj<8^t
zqo+zO?OL+UQ2UNVK1{EOi`OtUG{u(h?c_|RPX4)-Ay}w{^IRV!)5G6f#6$#Bi{lMT
z+8G!KI4}2aua~<?$<CWK8hZeX3VdRkyt;@ywH~Xq*uwx#iFYm2=Vl==yF_|ndxQpL
zi%6JV{}F?RkztCQ>f$7J@2)d0+51%`n<P$2QZx7VT{r^R9ymAE%UaguW)rZcP%?R-
z;5<9M`-gL2WM~5rC)~m@ybg2>d#_~6*9;aj83H1Ey?<E#Y=ko<vnxgc9P4{{d1p*r
zl`znJfKDQ7?<MY)*Clu#E{KH+fE|xtTbWHo%YE8sx2!O{U4K=zAOW@hRz3H_{B@?u
zk#ENHL~Uy>HZm)fMaU@OoNjak8Acj~VOz_B$Gf>m!_(%M)-e(k_%nD9&0VWSb<!rp
ziCV;jPot|buoEMNs3`(Vx(pgrN;ug_>_pslmJS=JXb{*bEqSy)04bao#?O$!9oCGS
zx{LQ(7eg0m`V2ODR~HX8&E57VIJJ<4E5`t`bzzOJFbyPBNL@>{enyes7v?b*<$Q?g
zhBZ`pAJ)awc{wdAR!|O#3}fE5*?zZA289w|HM^V_GtS(Z`o*dzz9_0XXXfeB*DHU0
z3QerCF3>RB@1Q6#d6|j^ogwXJnj&3vtI-OpANq8IU9~qv<=_S%*&ZT`lv9SWU2yPJ
z7siHcM4-r0i<N}px&lv^&kk337p%7eB@_I|hTkh`M$U&6gP6SDl8&~z&zs$Ea)STE
zQva{}5Vt~W%q7g{74<9nPb*3Rc~8pZ@Jw@G-tX-ss=OgT{vb5f@F0%T5^}!MSaZqg
z80kMH0iL80p!l!v{J)jB*9RwHc%4g~O(xPW9vMK`ygTF&AeG2n7(mS8-6TudhVc9#
z<(?}IoZ*Sv<g|WOAOaaqv(N>iR7+p<Pn@x}85~lj_ZI!OMmS}-u9&={t-Pf6oFcAC
ztz1^ah7d~SixxCQJ7z0n6WC|E$x(r$r;3Ly*2VLObV^+&TJ4QXHP>lA){A|wDx@)(
zb%0pJv2NT?d|RS@q;O}(q_ZuioXc1xM0mdSMn74|Puti@{kbrmD<5Zr?#Q$+>v$@j
zLxXd=)Q_yC$L=<P0KwO9#iQRAyG;t>iQW%)HLyc6z>V6$`F+ihktsw%h47Y|8!K)1
z);P;>U>Zusp4%CzmH)(W3*cy4VKufv!k@0$;qweFM{CVO>B{HC_NK`_+e)G9Fm~89
zTCZ%Bi}37+_SbnqS}irTv^S>V<0@WfN_HDZR0m|m2Mw|<=52w?fdEun;Dm_B?9q1P
zZ%Wm$hG^YTD*yK&%_kBk&y>P6?c_LhkSiMF{;HSHRz}OqTCOnrgzwGTKyG6ed-V0^
z?`I0l9}nyt#{(<1ZQkVZ-bmUZ=E(Wqj;K|>2pZYHiN`$wrNzYxJ<gb%ETE4nj|{l_
z+wf8OWVS)TiiEQ6J+X^C$$pGpUsivxg~eG0Ah6DD!HD(z@5Q;ClvmnmBtL?4QRjCU
zNadtPq6j%oQONwzjf?pN6>n<(+@a0O+ptaW<qY94G|&4i61m3(PUO0nN)Db;R+x<Z
zE@hX@)-oJnCQ91Kx(L`!+ljR+awtOt+y(d!-ODsd-|XU^z3bjMa_0V-t=7wJftnzw
zBdxlmW~kHE#UFFqcL+2V*DVCH(!9(}8CL7vh9Z>S6hCtL-&*8<<KUk9pC9Ja6)OUd
zjOtgryV8@e{sK2XOmfqCxmoG3>7ZAhF*rE(O(+0OU8#6WGJ2zTzS3lD?rc6VY>|1R
z`lhJ!#NE;VX{jopT4zm`2I9tLe>m>WCeZY}99Jx!QvJxwiWD2_2K?Y1eYVxjgJa1L
z=Z5CBWZz3cNNAx(OwYj^i-Pj`sU(ZWNhffeuQq%emRS^1trRIl(cqYpX3RdPplu9;
zvn_Sy5+A}E92pJk+9ajr1y^z04m?oyD;>+~#eDMt*E{DUpsp5}A^S)kpt|<t5is5$
zE>sLP4<pRX0ys0SjC}eX4b75nJw3w{m7%@MY#ErH_~gB(Fgj(HKBY%ibFCRT(SQ(d
z<&T(D6`2#(znaL-rPV0YVV$%p^j>}Al2XLR*ACwB{Y}TP=xJISk=Eop)CrxFowb`k
z9n$v;Tc~+$S*%*AyU|!;t&mHsP7m2Ig{qk7wa=kUAlT0dI&y#VCNA(SOu-L48!@xi
z@j8f(26sEtK>e5(;d<$}<z`P-ABO9r?6h;dN7Drt(wASmrXUz4?_Kg|*?Y&@;H^Kj
zvaT`A{J>M?Fa4weg~23@iH>@?0L89q!fqb)cW%nbEvX&uZZP^)jqmA?qC3y$o_I7l
z2>w|N{tL$;ee4PEgfXRzm$&o5qip&Q|Fn!GS&0NLb2gmQ$<sAvv!Q@Itj&uSBiqlw
zhRa?w+e%L4;lskc=&_?`dF}s=mh}J6A`;twTL)oj;5a!|PvVU)?AULrjb|n2$*?36
zP1uzvuF{aR1AqUQPrO>M4C+lt(I2Bx|DX;jX?99=sBJ5i0Or0mOs`(2W!m{}gFf*d
zcr>Mb`S_aEpRmp2>@gyf|H<_yt+Z?_EZAh&dUlXpy6V{UbNl65-ZY0(qVq;PD>gFU
z_<-K(&rBJ;Ge0|VyzANW#d8hhMmEP!hRDO9py93bz}@K;+8}o|)RnJt&=J2O*IJwN
zK-Z|oS<KT3lJ4n&cGYTB(rAu^yD>tiaHD~gRO;q}cV^bW=A%Z{c-1wO7*dZM{r$+G
zqDe3D>8R$gh5~eww?QNma}jK@o8j0x)T!Y@JMk6#=zi_DJQv<MN4mM4=G|=~wChsp
zO2`~H%>E420?go2)GMXivCIj!r%+uW=dCNJXp#ul>Ojsq<U2y<ERtpOjQQd|v!evV
zpOU}nm62__*h;IL&YxuFTMMVk9HQ3;U!=Bhx$4_RyjG59XEuqz1iVOM@W>k$4;=+Z
zyE&Hipp(*@yi;xI4$D!v6D)wpXuS`23e1C@vs~;&;-$tko(YkTd#j6G0)87kv5_sT
zg;q1T?xL|j->Z*!%B`LSzv_LGdD{PWyKOFC0nrZ@iFR5jnqo}EPZ!`6To!6^2g5lh
zAdLd9?;FFW=vxs<0r%qh33N9Q21g{ehQmDg%JM`4D2nPd`wLvSU61_|Dk(!8$~fHs
z=~FKenN%$9$pTdw4wH&O{Sb^I#V>%*Q$#e9_q+5QVwxN=L)C8u*fC$EAK}y@4JzKC
zEWTGvRMU~oR~tI}#B?jE7u)P03=rcRQ#J-`DH+ae7uR_H7Md!Mk|ZVJIw8}Db{Tgw
zKXpiP#H`cT%F0XQEB#sl_U^pprf9SHaPc2=p6Ex!kpvwsI?lfhync0yRtfwN^(bL7
z@r=n(46BZO{5nZbk5%i*!KmkT@WCJIF#FPi+aa+x+o<dwt9N0{2IhVAU(k7V%t}$&
zD2*07!#$!@g&3=KGg$_N5vB8dGFyQ%kdk-mg*xOR^-f#S$q5724{v%BYXaiV)BNJ~
zS3=Hm<CM?x1v+mFMnm<#_)eKsxmOmx$Y=t0LxDe2#o7!ic;M^>yPJlyQYSIB6Qtu9
zHDlq0v1+>J2P8Lgf)|{ZkmVP1>a$fY^EK;q5>#w`3Vi@6(K=VKXXq2sb+DIpfPh^}
zYmv$`O~zbT?U$6>;NOmzBEAN2Nv~^}gB~lUdka-#{@kaFU}d7(JWJEl##n?dsm{j0
z`Nm8+_Wl;CNR|wxWV#T2-q*qmlhW{d8_=S&YA}r)MP=z^>^m2w*yuyWN{7vQ7qWa_
z)k6X1B=w`=C}rQD5=o4`{(yXij)U?&&nH73)dZm`^%F<$HICB8cReAs1#hBKgi`YR
zkj^}WbvQVLZ})0OH7AmZ;2B*bxRIj6&>N@a2E*b=1FMKNof;mU7Q46aosCu!50A2Z
zVRKQdbW0C&O7m4zl&o;Ny*};KqxnxA6OCyO`%R=!M@E8@@aiS;0QTKpU$D3dhOf88
zbP5ej7-bgqX{pmQY2DD_@(OU;fFUT=N)0|t2))(tO1U*+X8T%;<lLPa+z_;MqFhxz
zNV^sBqI5AlL>{!ubpECg<K_Mo>oRBHeKyA$Nmf-*jCyOyrp{`c%vd?nm;_=^C=w}V
zJ*tGyHtBi#>3!N?xS6*XY%=_4G>2Y(ZoO&4djP6bLCcv`l{26t#LmPV4wt7Gf{2o1
zFQbj7ifPG>n>j_f<%!-+PNU;~tW@i+(Al!TQR||wYS?rQf@34>@-rKY6)v-QQ){8A
z9uIVfXs0^)p|7(sP7`o|n`A$TG3E9VSB<8gir<u%ZF!}0kjUk-Dr<ArI|w8tnk>IA
zsM`(*I80ip92ib>*}~<5TcNX<HMj!UPa60D10)z&xOg&Gg@xJly5&g8G~uvP=0j8z
zD5Vsiw{LlX?f4N#_pf?LIz0Cd{xirJu&Lb3<?p~;u@ZMmHUnrq+E~ihr;Gcu9a$Pj
zskvf3+UMc)5}v=6S6JQU?(E;S+d0T+=8{82wT#pk0E*V^`_kkd#C4+%2&N!?0-9Z1
zb*3$OQ(VccLxv}D8LE&Ji6(ka8s5jX)eVU<2V)bzi4@QMj5XJ^x{)KDMgkE0##(oZ
zmlB_DrWvG*oEAb~SA(eb1?Lhh71XcAlI=&c`Lf@@tXA?T)`z0TQC2S@?mN{j4-%K3
zj(+?k?I@&{F|xj_yc3;4(pdXvj^7pG!cYO`!v?aiZkmRS^_iuhN+%9z&hN(F7NB)m
zjb834#TSKXmMGe@A^lq(So%L6`PREv;q5W?(WaE=vJkI|C|%1!dEFLc82DT_i#;i9
zMNkn=9Pg%Q#`>pdc#DkA6}#t~Vx{z|!g064sTBh)grtDO`F=81(id6)(lx`+{2mcQ
zsmq!J>`jmzTw8^6+6d2^$=3J!hI}ulM3wYad7CI(9-GI~y;L_mj)_3?x|Lr#;UzQT
zJDJz=KV#n`3Xgs(9eSC1;P7lO&Cb^U=lS0Ea%lqPFQEg)**3Yn`j^9Ch)%<^IjD`2
zgIXp)Q_>NRsk>K4yg(@3bIbQLBZp3$xUx;RyAg8vUNexW$?iKRau1I_(mT~|DB;xJ
zfL-RTMh80}+Gq3WxZx^VFIx2F$QUoUyvA$fnb6@or2IO7<``zX86-sI_w2ymw5e@l
zyq1?xMepq92@w<w+9>-$5yj9TIKzo9n`To<)K0k>)VI+*1_-L{UCfx2&F^I{*%%#M
zJ%Gdv?2Zout?NhWUU~r;OZvY&z6hW?@$K%WF0|<53s8w>*KbrT6IBUu94gjl1;Dl|
z4TQ%59$DT?fU08awAL%Wg<+D<U(i%TmG#T^wP7mMR@XhX@4lD{!N?AR!*%Vo1h$F7
zzg6hBQ1h)M@~&lG6w3?bDoom2Ytpbr?dgxPqPS0HGrgSYNA-}zlJD1dx+%uKfdmMa
zi#5>;Q_vM6(6?9z155>&?B#jYfEEHH$HFW<)}N5+;d`21**Xy=BBbUxZI#44UVO9X
zl)`wjtSr{Jr$77XIofcz!-|^mO-sDUi4GIQu6kd#Aec3W+OT|<NYJ`R!>=3nJSf8`
zCb}X#@7~$yryKryt*|Wl`sLV7_zu~AKaXp-L*Pj-c!%tz$Ifw+PTBO+ow(i)Mih|m
z2$B9Qb?sg|L68gPl7G;vf~&Pt#>-_$^70}W@5WDngQ#8@`NRhAo!bv9%v*WwUcmH>
zCI+v!cmRpe+i|TUqwHeN({I&fo>;Ah5)|#K*~Ot%{X5Acgg3zr(3{R*m`4s93>J3>
z4PByVHzfm&eD>=dAE?(VY?@ra9`>W4MW10eM)5KjR}3dPClCR8IsfV8i7>M)Y=z6~
z<duc(S1>xz^yt0SS&fUO4eq-?&*$I7{H)OO?ebafD0DMoh@{GYT+r`}Ff2lUMzZ9q
zEAk*tpp7(1x=Ow{!fK7l)_A7Kw0FGWAD(5i|LVc$5mtMnWD_=M#Q}vSA*bGtG^G4}
zbqfnJ^`~CTg_34Liv;sI8xOy09H8A!9@z-#rSJ$1mVgB?JQB*xZDIpGAjb%ixlptq
zd4<8F5`dfIP(u!cs20wYupN_W(bzctI4Mbv-GiaxKc9$O(%EOw45B5^&Aj0?IbSX0
z0^bpObn{PU@dZ}@g`-U_s*JXub?-F#w#YE-HjlMZa>7wkQbzmw%X>&f^Rb(D@sH!K
zb}rKFX(jdxZM<2#*CaIN>mQ8=V*H>FE*>_h4$pgL@;Rzaj)5<z?j6yG|6G!PpSh<C
z(BosUE+a)-nOy!n)+SMVc?@U##%lf&adt_HNb+)A4Y+d^)x?-eKV60z4Ns1|u&0Q5
z8y+U<E+ZVTs|YCR2$GKVy6B*?Exw?6x7_9;jZ^G#ATp!iAa0@{o}vV{b*P>^T<^T4
zD2u#K79Cs5;GZelH&xg=42Q%?MfE>;XKDeJSp_1vQ5d~wXdj!CSo4em3ImFmOl6}h
z*COM?RPHkm^#o-eotag7{)NM>EqoMfh!mTZJm}$8B!F@!jcVpPs`W&TTP_$dx4xJ{
zh`;}_u9zjW)rXga4F{!jN!+00(;1S@%JaRo_`t8)sEeDQO`mGL^KOBE{yL0uZIz`3
znAcyV?Xvn9sw7H@%f8&p20n|<LHu}rZQQW!rPEsII<C`v5YZR*^?)(<_xs2hC8Wbx
z_CZ@Hyp07~Z+3pXkjdqn{&J+;#g<<h`^7h!#=O~41eA_8tcMuw(LLeTpZP7g1E-ua
z=%MR>I2=GTrC9o=g;NhgsU|W3Y2HZ#YFHG}yilzKm-y)zNL*g0{?1$~cFG7C3A<F}
zIrVv*Pv#+HX6eY9^O+FE%k%nbZ2i04*E4BkbR^2A2gKVDLa2qs=Ip|kJ-9RojMQQF
zsj?mwYUrFSfQCEA=Uk~7`x=HcknF{G)LUj@cMo*-Rd7BXn|o@4iO3gjVSe#o<T=TE
zf6*~#^3*89-Itn+M$uE8c5W6XO?JQmJzR=a<#5%sHJDzeZQhZa;Z5`$7mH(L4%V}U
zZpmiR?~y62`CAF;J3RB<oBrj#dHOmYE?z&f=iQO!t(9U0O<FA$Dk*D&a|P+g2&T1`
zCqDL}456I%e2FrTLk*Mx)`4n0mbMN{Je8eFV5^Gdp4eMbifJYR<AmWw5olB|ZJKYn
zxJG5|dgG2KQM`@L%c-bSg48}MK)6OcRh>rRi!J<O!-`Uqn&~yMv^clf(0emO!D`gc
zG_zqVVR}bk$VMNbQICB!scEBc$gW(KE`33HvKo5NFqJ);K10lM6vgp6Uzi<kSmIAZ
zF(kOi%LI9uINpeQS}_yk<CN9VY8}uP|M61(vOyOCHZJ-SWz|+t6Y|z8K(Bg)q0IDY
z5q9XFZ7Z5CXwhQNQBW5gl#=nSI0WmIhzupAB&kWg7yEW)9@=oQ-jm_wFJ}|<`344a
zr8P<G<z$}u_?Qx!36E&xmHe1C>)uq;nDEc?_TL@!Kgr?Ru}sESMBM-VAi?BubnQLZ
zLn{;CFOfNZNUyt2IowEpu(4<9`uz<1*$*HKAp2i9)s;1u+>Y=;Y`o>Jpmd>kSZlWb
zVcJ#+GHOYb*{Vs~*J~gvI8GxoZDp#wS3d9&xXjJv)8}j+@_9hpFi?5fv^+^3ORx@l
z{$R-4o-fNsaafHsgK6sc&^xAyg-tkol#^YZze(0$?3;<|y=2WaX`*B9Pm<7X^bJ{U
zSZKP5PcNEpva61T;;5U}d3f^X(`vp#xmxLjrQt#r9uafqTCs9OI9;mXn)tX{N`pu~
zYUG*pVIP#1t<cS}r39=G(rj_m*-4Lf(kb&$cLW)5r2<Q4YMXQ*r@x<O*|16ozUL*X
zqYTS}l#Ht>sL2rU^&RanI|H!`F09)riywePp?;5i`=Z7<1M*Yf+t)Y3wM|rMbs)3+
zb{O&`<3zn($fUMtq?0JoZU+Iob3)NZw3n#7cd%0W6_s_%KNZf1f(@*t`SY4g)EN$p
zBwCRjbnb}Znb**q#+Z(XT7DRrb+-m5*HBlLRkdlACMVfP=lwuXMx_enXvj>Y(k7Es
zHI+;Ig+P<$z0Rh-CVXz&GduJ-p8BNEL*AX)QK(eQJEth3_LS=_zG)e-CgTmOMwYvJ
zIreLf5&I;3TvBp7dW>vxC5!b%0V}DFQ6Tta)ERk$bi8tU7Cu;QD})N|BxTEDvPu1F
zo}99~iJ_EY=lWR(jk;^I!lZ{qDVwbV8bQ{EtV*K959bueAwZE&HWa?-)AO_v%xm_u
z)rM-EJKVniVe7&9cgXD9mY^wOORW3;z``%Hu;)jl{MiS{t@*b0dvh;d6RcTddM0|z
z#qx;WBDG$}$ND*We%lb+_`NMUFWt%O*gi|r)qa}Ei*l+4G1;?JyIMph<WI0Pqpj5B
zDS)ksR3D%+IQ{IiJL79TqiHR3?NtO*poJDYZ*BwMM3cNp*75SwAt9_`^BX=~6}p(=
zuxr|c72R7e?QJ!6R!(-`NUM&EJK>!RJgRn*5z{_;lfex+U>}cSM;&dx!UzFkGBf+c
z!_cWb<{25Ux<AT7B^#U}1=ay7Ol}DPlhr`6Y6>t4Iu{(#mcv4qZpt8=8}IeUaiS-w
zBaj{*^M>-#M2ZN9V6(VBoiKCOti)kUnlJ!}FiB~qhh@1CZvM(PoZl$xGV(*ZXrMai
zGu+z(8tSG2O6@Gxs033CB~3bfN5$$gz*MT9nEH2nVXAVU2N>J<`!#*noIyG`>>cgT
z<sH@J${nPSh}8r(-BObO!ezP5{0rx94~ugZ_Pdmox(xpL<~i1Db1X7!B!NWR_mj_W
zv$!rJpUC>1&9<4O`?5{c75_SUf4LXS2j-{|-XZ0!6Th$XPxKY-|BIo>us|#$FQ@|t
zO9j*i_k!U3f-%+QRdDAhtgRxj3q7+%wuC(R+nEywe?n9A_$s91!0F(rdgF4btyBAJ
z)VEu2@N?altPx|0_RQt%Q?>LGuE&S#Ct*J83pq=9qA{2>&_)|e--Vcs>`U~JN)#wg
z^@*1?Y&4K_>&Ef1CqEXw1k7+BktdS>UDx~RR2RCG&=+~@u(Q#{Ts^gvvnsP16PEag
z3Q4y@Ol#>mmYqjcOaQf<MxWj8-9PjdF|zJ2mStWpG!=nR^2Pi0Sx!=_24<r7_Z!`r
ztQ*XZBZ9xvS*?R5o>6D{Kdu$r#6Q?EYqMWG9~EldwlvI>xVxXv&cZiU1`DIOHvgzo
zs87jEh_n_PAL`dsQVC{bqRT|Ja`4@@YR%J{daFF!d_BkAX#N5tAv7k($7U+V$KG9O
zDRbkbU_V(-Ip8`8Im8roNaarSD>ee4t;H&>+n*Oo@&^~hwpw~Le`doL`hN|2g$^l2
zJ38jqLI<T%q^LxemC2}%wrePh^m0FZ%{ZJ=8L_etu<^|xa?$!p^K(`qg~R5Xs#-m>
z*7Qth7zk2Q>NKF+5~Oci6WH;d4^V({Hs=o38zTtIHAR<vAI<PtPX@x}zBW|2BAOx-
zGjuVfhCG0cV4o)ZzRFgTX9%6R@>`n}t5Sl``yswAV(06r&Fcg0bj8aXHiy*Yhhdlj
z@EeNVj7xLa+~EPO>D=CcTC|2TZ9b9c!)o^fw89J3u(|1x-{YF7f<fzPhS$7!3S82b
z&ISuf<~=-Bt1!GsaVKjRXVE`YaiBW#2mIUYY;Q$q#MnNKVxU7B+Z4m=RZiw)+=7UP
z9~HvJp?pQ@`)YCa)>`Bd>M<C>qMz_}RNNTm3q#gnZEbf$NJheJTO+<_LG`Y9H)gql
zo6J(ZL{zU3xhJ-Za>h8}>WY+@e_B(&<uX8PiIqNI9Sq=6*5TA*<DhHk%2_CHZykFl
zR`KOrEXYP2e|OiWTSgwuwh%tE*MaxD$l#sD{xgY`-ZBv_E>T1I<~ugIQY{}qux_Xz
zk!qxv^oUfFuEc<YEi{Yw6ur@QlVFNGdSE-6kq)4KPzhVliB89q<hmI-wu$wjT+T(W
z%b_iAGQ?I9%F<5h1_bW~ws=Q4fmPUT1p-)nSSo@erFA{iD*L8k;RNi;;NRi<%ievN
z#Yq|&{+@-8)I-^O7JM@wOKihU^4-e!aN^g_8hzeklYBjvdgApLjzlfiYyJT-CgAwU
z%O=Rm)BUjZYYCvrEXHDwiHvtnRyK7g*9aaw)6^8DyCuMksz@rG+!8BRr++qD43y{H
z071E2?Vdk+<xkmuzAPu<IN2l9sMcDRGU77)8J$a99nI(Q^YnUM<$JklD!Tq_8=`R*
zD9&unXUhD`;J!)3Z}+|(4}xg^WP1_P^&t-cv{`r|hNc%S_h6w}=bdH0(9bm3PVB`U
z)04ygX@pU}UmMH}cHwUZR}0p#o?$EPFXl{-NIKE%=X5I=*pf2BsIBU_VRHHQIy+6i
z7H@n9e7#Exq(~#OEu><WS4TV9^E%kXf;-j(w#ypT1KtyIUXlmM+@2<3EH$Iuw5!0w
zEetAJqK=b#tXaQJ)dyaRXGTUfJ4pSoxd{)LBjcGBnVgz!rD=u`<VkDl>t^aD>mPNI
zAL+&fNVb-$aXJ`qE3nXe!Y*ojcRT0*WE|i{UW!JfWLI`l#xgOwC=)Maz=oR2Z|u7)
zir0T^pSBCub0e*V>ImGPx~7tC^tG*+WI0B5U+9KwnZB+9bBOzpXD+47Hgso#jOYC7
zZqF8e^ZCh+`jLGep8f2P342Gh&$(G8hqcYPXqE4vcT#TU;!vB35P_DG=`H6kv_<`@
z<hPqyT&eG{%S@ih-zv?!B7Oxuh@@uJK006SJZ^}B(Ld|OWG;lOB>KPWJ8W}885XOw
zKt}b;^sK?KV3RaT?NU{H6a2%3NPI{yZz*Onj5e(+pkWEQfYFP05g9uSJ?FL(W;KRZ
z*|u?=`33Vx3_z9)tsHcR8~J9Tm>D+1CL>ppn;+I&YRy%{qg-UxhEb+srOd*Puf<Z(
zoUVAj%AM=L+fFaMiMAJYB=oM)Htf0ww|!mOxltE!kF?9hi9qFJibmCfe4kG|>=z6=
zCYf4cQG&0tMMjaMZsx8kb41`Gq}Nz@nP}Ld!U3JzTwg#LRflf_paA%uf>$b8r@UcB
z?=vja&^dxzPc{8S-Mzr=Me*pk?5JKxTN#%r1>L>NSJb#<O?APtqEc*T8OEjjM86Xg
z<&vd%)oF8`$tpJ(Awm|dC+w@hq%7FOj*aP5HmhWo7inzOX-lXY`KSI`Uu~d7yJl#c
zja=eieJIL&$RRvpYE{-x<L32fD?dYLmR!KE%hJ#nxsCxt`)^&u36IZw)w|m&MACD$
zYaP5RaMfa^zBg`>+WpyPCQ=$wFP7JFYad?I|EW-{RN9^38jCUnKA)^N4+@uQT_h$p
zZl{B0QRNVWn=7(F>zFL={Vs=zeGG<5(2s&!b1uF^ucS#YnJ;aE((o3PG^Wuca=SfQ
z+?tX#@Z-RV)Y%UTU|)!Ka2s{*JYP+E`%Lp^|2Ly8Msh=$@AcLih2w6UN<A9ZD1kq+
zEqqox0l!7Opc2V!KT3efLK@{B8M}j?!uzTopVbc7R8&9)S`*pFfX#(LhU~+_$q=Z-
zENzoRg#qSuz-PJY$$DNAl-_FTQ~L$-B*)8_LTt-!E!aR)p&ng71Tl}}e^9jt@xS{n
zgsJP8l&!}wSBv@P9UoXGGM&}Yt6dLTjCNFEu_ra@qA7<pChYF{_0}iCEDq#b4l{N5
zeNSA@cKP|?m5{clblQ6iMT0G#Fq`fS(k7YH-R7lo=wlX8>JRXfG+#29@FfqzL>d>5
zM22)%q2ZTfe!Ob=e2SF*FrTz#e4L<e`NS`Y9=9Tego$cq{V@HslsQeMm5AhgCPqA*
zpG%AlV7N@}@}whZV8pWau#K2~EL<bldjWQ-{nEXhk_bqlD(tJ=#Jk0s%2_m^7t)xR
z!5D&bF~Zh2FFfEci>xz$RO1jtDzJSRydT@(4LZXs_mCP9a&<&p>d^=*mMM-_H64po
zv8R&KJcfU};^Hqg_R5os`_I>vf|N~4O<ug=Xutp&iTMT5)A&gnoB<`~J7so%4_?uO
zD-k}Xm&C1Sk~W~U56Y_!<&Ia^^aiElCU<Qg+%==e`Dc3E=&d;_u?Rc~qUa${xAldd
zu+cOggwzpV&@&>qDdta?x!lcjT_RThL3}3~*jG2sc|I)bWCVAovVYm@#^(c3@5PmD
z^M|7&4p-SfT26AcjND7`@x@?Q*_?*^#@;na0pu$x%U;)=?`V#+*sf!{_XMyR{ub%T
z^PThkd)A9B$4Jvzf0Y$K*>}970dgTu)ZPy}Z@|;5E^T+_0)&^PNq7Ho{c)Gq(uVkx
zuBj88J8;z+;zU;pO9C_+e{klaC0{!)_nq2wYX`Q|8{4z@oZ5WdT;w|W$Hn!Bz-s)B
zE(NbAU6_#va<W(c!dM{=&QEa0qI?1-e*Noy*gXGa_T*Mcu6z+LY9WAk-MMUD%eL~5
z3zZr7)+q~Cmi8~gs~fo<bdd-5uA2*LM9>rR5%YW#_pV!fODE&v+aGPO|LZY-i)Ydf
z-+uGvXW(KUQ$^BsKp;!uuDtWxCpE`cTmccRjO{*NLC>F5_bPYE=0J|G;g2Jt6&}3A
zR4wH4F1Sf1bPYW|UqvttRcC)aKO3BR;<PZlR9CSvRhm@QNJ9tnO`?rWAv0$X*Ny2}
z;Jy-_sL^I|&ORA)_B3a)Z_BA)DH&uIT7Gdd*`~Rkcz7VIwff|2+q9FjZQEm0W;VKV
z-Io3GAbxs5w!C=43RIif^j_VzG?;Ot)rm_sR|zqhzW7E9Sg2WDP{CWZmJ*_<0pon>
zqt3i}j;>aeX|eXdG9S2D31jf|-$i#<?));S$mZEt8NEwyTYUz1nU0$>_<bigU9WS0
zr9#IDP|hs12(lki78EMdv#5JhI{r3k-bQyD;19MA<c7OJ&zO0qhgvUAwt$~Dvq*$o
zKly23H{55HQB^zsyY0$fdo524JM67J(?|EDX*kCbwQQQHeD>s^kI>ITh=I8C-5?qZ
zlx_yTO9<d%aoPlRxDz;`VsZOI9N&kqAn!86hP_)zmPAu_#}c0=S0|v3nsJ=Za#pW~
zX&!8A61QPZdxPBd4_`Oy@SOL^GFvMBWk{sASZ&0`Gr?I`st+BOA$Sk`mVlhI*h-$a
z7kRJy-65ctH6~VxYBuc$^d{b9Kr@Gi^V>rRtWv3a2MW8c<Lv2em(L=`G^g!m6X~t%
zZ-uNgmv3#u4e!*+tEEeB>b(()vEQbE+|vi0ru}mZ4oitq(C~%>N2zc9uE*$iE4Y47
zk5QpfHIyw<jaS1T$vV_KhQpy#oZyN;o<|iaKc^J38FRKRfG;&i6+uPQRnYqVv$gkY
zzDg38eo4?)@0m9P*SwAGbfXVo+o^S?oxeqVn<^(RQ>Ad194F^}n}{_eTvz7_4XPmZ
zKYn~xb7Dh0Q5U7)`0}$lkjA<%BC?8usFtl<9mMfNPdPU{b|UF3DK9^>E@YjlKqP|)
z-Z6O5#7shzZ3U$i-#rBIb9+0E`K%nm$qw%yc^(%2G7Tc+CNZ8^(RG&i>Dbqa5s5sg
zZ+V(;VDu`c7uM;B7@z?}yL?YJ+k!J#+QRCKn-3RmNOjuaHT;oh!fY?kmKXL$^d2K}
zYXZUNCe3QhYY&^wH=!BejiQeQe^XKn>dcnUa?vOE_}XQr<ct`Yf-1J|<vu(y&zBCJ
zdndfUb9Oe?>_i<@v2ib#`y$}adzUs1Jbyj({y#na-oJDSyMN8X#Z_$Y1(q{JczG4D
z4_xs>W4g3yj!sa+y-pnMc>cbd=XKJNyJsKjkIc1-J08#T9pVS<1d((N{;!*R1^iL`
zUrGu-??kPWJAl<HNmqZXgW9`}SAe!p{B9Vm_B)58@`#9f_Rns=0BUeguajGPa+$PA
zroQryx}Ua^H=F*!J{Y(YaQ#c1d@*q0i#Zd2$I!f%kDu4}!?D<#y<^k}R%M;}e=)i6
z=RbLkEnkz4pG$Fk^RHipS8~kVyvS#F?pd!Mo(=`F{)d47D)6_ai!c30asI6UIoB_^
z7Sv$=UC*Aln@juis8U<F;_t`{+`lHTA>Y35cCwSpsQjICAPI!5c>y1OMcBC8%`4DA
zo;({>uQg#0mRkhgy`uS_L;def66HT?OQ_t<fLU`0Vqws`NxR1%M0F22>G`YTC80Y5
zJ0F_p7j`)W`a{&@<(AVrzK{qmS7Zd(&8n*H9M&zK%P~auT(;g?;b01Ri+)(&;NmB<
zIBy%gBO{E6l+HOr__dO{OuUAhbF<&V!vh*t(Z7g9+C_b#3Pi~boqytq6>E7bHTs{O
zY={k~^?FT_5OF^(Y+WWTCQ-M~5Oox*A{vj|-)M`PP#E$s-I`IWoqvLNLJ~}Ec5k`!
z<?G%X8bc_Uh}R*qB6KGwmNqlfeR`KuJz$GuMB&-*A06Y!t>XSkl~C4{WHpPy*1@)9
zC%Y`+MI!oY!X68t^{s%(qpn&*?0(S^*OyQFe(IDM$Rus~_YDKy%H|ow+0R5S!k0gj
z%3-Dmmt5CW&W9&_q%>P)<GVXHf8hXv>pCeJ>1PDErNH9n@v@MD?d}or=HFCUdy~XB
z@g~K^ZH70Vhx!EV_&kGgjV6Xo_6=m8=aKs8V@UOj4wKiUN~Ttwqiy03?EQ(HpG!t{
zM+2x1>$}moGnPvt1OWqhg#heYSP3R9C+v(sjNT-6_LdEU(vv0|Uz}$vBs3-_#C23b
z+p2T?o#)H5Cd-iE&(cAZNLS<8kOmj2(Xw^7`1H9|7-w22+)7*EsN#-rBCl#Yr8)2d
z5~^L2AK~E2FXde87p-S|zL7lo8Gf=5&^EK(u5fj71^eZ7==T2tP5zr3pjhr()So7E
z(hpKyaQ^GO5v}V#X7X1N_svCtRFCXJA98=BIg*w5Z88Rw@;Sw^@2l_6`}SaO1Vfir
zk;rzxBI<$-k@19*>?tz0;=9Af%><Ck;E>UhLMl|SuXQdy=F@T>X&~gZ3(<H;(VMd{
zK(;todCB=EKR}t4<6A}AlA8ZJLRN2TE|$SKEAr{j{iQQD6IvVOK?7@D{6AgRP^Wb*
z&)Trgw3zPbUL>cq)K>Z<4foQ{6TP^%<{@I)4E5loZ60*Z8{hg}Pg)k&l?crD$;Kp2
zDgMaBga@+CxYjkqF5-`NsX==Dv(47;!9rkNO#-EI(FV<rob1Uf$r@p%yb~)RJO2>V
zD3Y?~xWM}T^Rq{ueV2C<*Lz?Rs%xJchpCNUR+jFEK%>=fTaF(oKFQPjk6klqUGzv=
zLl2ZUo&O*5-a8(yu3Z=&5-AALBL&f;M-UN=h!VXMZ3a=Imqaf^gh+J4=%d#lI-^I=
z5TZ@=P8hv2jBm?*pYz;#p7XuuJm>d(@ArGpyZ_j;*WPQ^y87DdT6<kr);g&X@X;ij
zvc*U@Z;9~gMZo{Y&%bEUDlPq^77y6Tkq5SU%7df2NW0>3Bn-rzq9X6IJ2}QM{9h0$
zsN*QVF844<`bRx#6Lf_5cv_1m#OCDOO6ADB&4@0Bc<j?eSg^PZ>DyJJ93h(;#TU5U
zTwFr+1@tXC-a+}!eVHJ9m4dK3u{)E-6h7L$y@;TjUJYsX+k2pLbWpITDPhpz3EfRh
zF_dv;D)(zPwBWJcSY?ry?*mzqn@7SPZe9{WCp%R>KDsAQJ}HmSjrS<a-Cp73rK4CB
zky4SZN{odb)=bz>y`E|L<oDy$*TdT`K473&@GzuUiE;cWFg+`!ZfA#MzTUpKyO&|^
z?)IuhGmsZph<2h{EM(Y_;yT~#VOHa_8e6Lq&96CwjN;LZYECW!xfp&A=1!EArcU3x
zIA<c|)N2MCq!-`uTwfRI{#3K{Kybvil;>t}H0KR(YXf^NOVZC<32=fu)q+aoc-V0E
zptGill6xijTVYIlw0e5>Npahgk39*VB-_w3x(uBLhf28PtdNiaS7+$+_7FAr%?j5l
zYTZ1I^uFt3(sXZnq-E<>-E&wp&8;#YgVXdwd4fIYbw(@7_<KgP;36)D8#YSS4*77U
zMqBK1ZP<9)lWKF}(<>8<IYk$bb{FOH*u5<uewEqbU{n(-9rS-^=ri6tZcu8ZMb8!Q
z(|ppX${E8M8~m1<s?k`3b0{z_T%$m;YL=CW!u&q1f-yWr&dOF`F39$AE=%!KHNT(n
z-@23`I~#rnKwu4dl=)^!fwfd;Sm}zf0d<=TKXn`pOW||YWS2;NUpvmUsL|WD!?E+0
zij7oDda5@!eoS7ORI~MH)w8u1I4VK%8Rx$XQ*0)$Ba^7GDs=mT(3Fq<xEv<wBKmP%
z@V=62g7@|m<oNb0O~1(xVE^UlPXgakG#~iG1D-Vrn7KqGJQY@Ysml1RvgC5xw@=j-
z3U6xF@;pEH*n5=Rqgb&^Pe1v&Z2acu^%hHowK)ERI7Ot<u$4pLT2q~8$;-pwK%IJl
z<ru78`S-k>h-jAfcoGHkc+Hiht5@s{MY66AXMOoNX8(CbGlhT5eQz1-|2)6y%@g=2
zw_R=$aW9Ge=ZB<L1M?{OJC{8mLXyWwRH@#q+nz|ht-v-#nh-pzKiOT0<>E?``ZZY(
zOd%F8S}f7?a;agN9yU2t*v%EkK6k1P*Upe{C)Tdk(k=RuJ=`B_Rq`|9x%>+aw0aPN
zGMQiMLkK#fa5w_XT{lZiBGYFWd}v=z^ze~dPs~&(O*XL1S<G|st5}BZ*dQ|Gan`8e
zmP|uhQhw`7i0d+;;_+igLk+&{jnC^x%(0<Et*G9)nO53tM&`(^#VpS7J$XTT@twYa
z%S!8FY4LZ;Ib#Z}e)ia1ROQ@XgJe7@Qd_BxwEdFXU!eKkULs$8sny4dE?5L1^mxmx
z!~LEkJl?S)@*e9GwD#p9>Vf){z*3vCc=%ICs8x7=OmUsJ0XWh3PV$!|vo8kj6^8YE
zuGb5RRm#3=CO+|)DdTkDxyf&}c{6PFSU=iG<ishSHa-&0$RE=XI4@wOjCnuQd2@ZS
zrTJCYzTljaD_^{0Tlb4DYIcL6t1%zOL#`vf>(gO0P4f+v%}9djvg`+KOY^%NM>M82
z3am1d0?1l(TFtH)Gk8wQd)*?=U2wfq$o0(x$33U*Sn{nYPs6%Wa42*Sy`aur5EgS+
z!xDYbgL8;uysJW9aofb1siQ!_AK&_FtZIrkcj0ILffO!(v+$Qb#9d7%z0bKOsgY(y
z1$M)CpB?aQmU5zu=I;1>tJb$%%G08f6%SOf6XZx%wY>IF)t@=uXw-4z5_0{JDNZ10
zawjiQ;=y~RW)rqV`^(V=8kNT;b@H~IW;IS7>ANJdI+C+u>(R<=ZXdq|3%s{0M1%P(
zHY^r<6mxC9f62R{8_P-cS_Z?<)LXY<b@EZ$er!F5D#Kg=mC!i56}q1$RCOzR;BLoz
z#hkPb%~tyxK8o*hMWV}VO)78M-s@J@c|hbor?n91tXT5wJ!7=}n9vYM$<3ddMB6bi
z=`Cb{WX`p(dIREds)Lp3-7Ngek80~i(wZJ<+<Ds(HvT$yP^mJuOCI}TLXJ&i;M493
z$XK1H4Z|peUI`qYe-S#ef%K*yuX1qKu79MxykVTlC8$@w<C(7!8czA9M~m!RoM-8?
zxT&At8(!HNIIUx<=^nmZZ(c24yVmwq+hk$tM-}{2+W`CSnD@ATbE4ntgk%%AjA_RA
zuT=!R%`S>aQ&T~5u1@D?3R^<7VcT@95#`ZEUd;owkDHyFI4Uq^1~o<Bcx=nU?S5!!
za|&>YwgvZF>SAK<DCK4hrlShS)#?426P{`$ysakf4nKagc(+XL4yJeUTGCHor+$@e
z*4XaN*CNu?!T5K0Via;Z?3izQi(k0b{*Hv%Tph58ch1-EBckM0qZk=3mD2icYV&IZ
z{}_YKnbXfv8-K>ndLOfl<{q*&&iZ~+uwg+xk9&EDav&%SA;KcI97X4yOHQ9KZiXzi
z)XxTOO#QHKw$er^%$a3nypIc&Z$T{*JV6Gs8`yQ!7H4YZS9L$$&d_qtQ@{H(R%LMZ
zwyD&^(Wm$C^d;*DO=|YGwBxidn+w^PD*qU%|L6t&YKZXB-eUDP`c|_-W`_!lArIC_
zTKEo0ghmZ~{K>y0Yn)r$W8iX?-#}2{E%4=AcfjS)QL!>>9<1^vi}0;DcEP~A{nqR?
zpHS^TqPibvcUP<S20s${xEqTI!BAva-ps;(>P_6PI$U6Ddspfb?IyPQU1i?=pRsh@
zwzt0p-Zz^K;Cb`_epIE2HSfz+O=k}itk8;~Udg{TsF@ml=ci4RZH&yMl?ann3^U2v
zLa{NmR%viKwT#F1k$Qn~w+>&a@=cj04{27~ADbkv<uxA-9Xi8v9%(DaTV|Hs{TVLn
z0Ov-hXhu&mN6__@2fv`Bv}NoVFP}R-3a(nG({buM6~lUjTJ`5q5qi68jx~tQb|6Uy
zO+W7nCIQC3?eo}t?ou;z>FnjHefa8hOHXz@L`KJv=NM^Y?O2OM*M6O=NNY;e(Ht|3
zx9KoeTS9ifFjrG_REg*M7(-@9LjiIDvyV8XF3@m47mQ7WOczXO?-P#*)G~%UpmoFA
zAO)9Vl_~gh9#AQSjjgZ=sNd9!l9En~yFj0)RQb?Yd%#Nx6~;PC!Zmf{l^QDa#l#0n
zBKb)9+!Bae!RLaW^jCYdj%ebsXoB+BA(gltE3jXkfyCESlg>1QSQZVEA|Wl^N=)S^
zlt>Mo+<H-tOm07obZ+@}Vqag4!-T81($AOrftfE&C9dgl&|Mi;cwVf1yJY8X;%%`K
zjG%4|+a_+T;e||wKzy(8-1nP~Z-U;@KDE1T4(A|#_d(?PJMHyA^Q~%rzhtG5A6JuB
z6O6Lut<j4Nx;NbYUfH?}WqEwe2R=rR9V$`Q60Ec;);y+We*7qlyNmoalS2Lw_swTf
zT2yLwDWG*9nQi;P_$w&oI}83-^L4n*Q7xTidj49q=4B#P;!5&*i{GN}I>wtkxZ=1M
zUq=1dbl?i3ypoK0MzUA)nD77V?B)`aP%h0%yd4<l<#@YZNK_>bFYJdLkKdh9cjDud
zx~IiC*?mrfQQ~uWcF3u%Nyp)obruZ1S(VcorjbSEc3C5t<-mKVLtiI@4PY^kl%&$&
zb9tmAFW}O(MU^$agWukd4GB8(qvCc}svU4TJ)YDPh<AT1!4&CvNa=U}w)&qRkGN>d
z({&)o<7yhXD(m3X47^ADF|nPwr^Piyyv@k;u#NSNcqe_TRR-l7*(`ZV=>tpZ)=D#>
zYZ~==ZLF`~hsp)zhajV7zG^s5u*O@GWO!Q%P^qK`+b&@pW;*;LyX=!XVrOClXNSvx
z@NI=MWgFAxrqE~pF1TCjN3IKHpTcz-9ge%&x<%_8a>&{57g>|zQomIec`%o$-%3<W
z`j`-!6M3Ae8%b~be)lMpR*cZ=dB6UohG+I+7!Kl!S$jb&U>o2P6D^V>_vPg0BCr**
zvK$+(^?BW2V@qZsZTsiE(1K(lBwNz?VC3}4@`udI-N4S!-Ga3X9Tu=iw>@;)z5?82
zUtL2NRnZYBU+wU1wn<?!)UCIeuY7#ZM%N?8QN!-V!4Xfv-DoY{uUW!92^C`5-|3jU
z?kf`|7FKiUa@7uq8@Sm*xs!S_)W>Ygtn|O0j@&`^(y15PY#9YVzvnVBG{%9T+slA_
zFK{^#ezlODA?>f5Q$l-K1~JYtM)uh5RhzCDD10i-T+wgksL|4Jjy*wQ$Nabwd8&}v
z7+Xwu)Am+rVnB>&o|=;<tV#&ma?|CG{*71SigR}H&-0&F-11x3GOkLGUK_~YGS@Kq
z7P7;<P3v6WqEMJ;sisBUSzD@L)5UL?q>;r~q*panYY~VmE^qg&CHg4<UsV6RBl;+%
zgm1+KtMa_0fZVJ@kw(FB)WfX3Chxm;fBXZZ*!)+Id*aCxCWaVwX<#XnmA0C_jI_L6
zi^`i1ncQIFUz$D4lvH$9t_^JveVO_A{y=F$(=5aIJ(VtZTBUC2wHWSd%EK<;ojmN*
z;7+f|cAC)MW@l|;QtZN%=&%i^&h0|@RNi+FX+87=6014(75y+|+(q<ebV$QEQh%LZ
z{*>{Rvb27?U1{}(@Z5)MxMVEO(7qKbPHJ~2T|IB;_QH$ZRZ;n2#gS{D1NEe!6;>R{
zhJB}xrG(@pV-w6*x>$~*o|WSRnSW!GDi@w!a0+<GZ5(!abS15yhHFI-S-uSUY4A|c
zQ0Xq5BW4sX)f>_&GzRRHAK;b=8V=?A?!{89H^Ra8PJ67kwvOI&f`0}4u#cbDHZw}o
z8<GN2GM)2TJ>KGECVMq|L2zLr;b<Mh|M^Pn@djj*;6GPE5Qn_0FDg0TfSJto$j6Ao
zP(!qa<<xcjQ?Ldje$g{1<Gl1dzRh8tSCo{V<?<|^aM6&8Ql)UI#wJycc`5|iiDCsj
zCFof{7n8(>TKN&aDIa9_f5(%zo#gT*!R4`0ET2(~Nr8BPTstveBVR+Fepcl`1&6fw
zNhODN>u9oAPIsJ>&5fal%<Z;wtg$;U<-4QIjfibmqerd=V3|LK=$CwMq1;eZ&Y2yW
zlBwh1ouo|Ax;Lu$RS+)$pQ%NfU$eY!JWyBRTlGqWX}To2|AhK5LGtHYOkBxDZiWHE
zFFaombop;?A`c27j`=PVB`P(UTFIZrex~sdeP5xH*JHZ-#+F=t`}XT$)9lczEaYDq
z^*xF_n?DlY`}{rNho!MGO3U4Of7M!YygX7r3FWw`6=|<-q8_ap)xznR4UJQie_AM1
zkw)auAHyE3Q77iHak;6iDzXl_M$q(i61u}_9y4S%#4M<h-RsYgS0$a{AY<E}Zs3yq
zGQs`z*K(H>=V<kjNNPn_>h9cLCRc}R&THU{Z0v&~<5!&Ph8EN3zOYxYW_32VRL}Ku
zAHQXtPm0b$#gKTG5I$IxlZq*M`m~67TwUZLO<ly}7L!Af5&mXT?81SlyJNgg(}Iy0
ziz<b`(87HxeaVg7JH@)?Lk0-F{2{n@&LH!^%Bi5ln7N6$QV2X~luh$vcctFxa#S&#
zvq*cYjyAEI72@D3ctg(B$VK$|pv}`3h2g6$h5AF;SKE)JqQbc45iBBd*;>o%?2|cz
zGu4B+MfP@|#vF&oLlvAF=0|+$@CssJ&6G_a41GL4Gl@MaZ3@QE5|@e>hDjFiWRSMa
zi{?<#QY=@od>(tx+8w`aras{;6{VkDfhnx)37#o4@2)=ZxNxslaG%_?$~*?G-ex0e
zvehQ)`bi>QF*(GoP}8o=vE5(Qkw-aHGH6s(FM6*2mNdWmWU{JPWX;@5Tlpt_=0J$Y
zfA08Oz(^Pmbni(VV(+A`afk1c#$ocz=H-`xz96fCs)u+biJY__P>ZAE`D4zvQZT3j
z#pm(4B`-q@U-5?EBUmcUwQidwFPU?MU$Y>Q{3sRIa8EiSgH_-Ane*-?SZR177xlNI
zi$7|7n-Ux1U34E-Fmlb`e#HGq<ej?yGm+4G!@!tZyQ4#laz(d)sz;3VAV*S`v!pBl
z$F*IL7hb<uT_khQ2x&^^_ZyL{PvZoNLW2sq55^~uwhOrt1EyKUQsf`g<rTY_&CSwi
zKMWzg`~o&EkR-FxN$!yB1{zwl@7ySpG_<wN?MjqM>Tej2vPchp+E>oR#58;BXafTy
zuJ*U99X%-DTpTKf9Q3rvRoFf@kLM|YRC7kh<*4Urc5>xQ<wcvz>D!c|xbMoVe@uTc
zEytu5Nu2*k$2K|~rt9@2cNk(NxK?*dcbT^4fG%fWM96C}pVH=<Pr;piOW~H#D~@t5
ztM6&IZEHSf)X55@$kE=vTFtd_E9p{={?9s(Lw|>DZUO=h<*ccp9pcfE?6y&az*0xJ
zo9L^x&4sm*j_*s?aER9%O2ZZ3pP0mjj;J84Wqaed2Bk+oR5p~d8i@3^(!I0+U8sKa
zrqRE3cUsCoP~2i<`{)u_Th41p=P_&R-1f#r>Satzd~878iAhnZz8D-iA`Z&?Rt)ZZ
zKni=;O-YxqIWGjS^n8N<Mj_t1Q~F2c=+#43i|$&HQ!x{tOsOAruI_et^fnM!P`jVo
zKPPDa2i1NKB$m>Np(ndiOH(5uZf>}EwS*=LUm^U#faKEB@e`*XaOCPH7N2<U$g{Vc
zjC*@N>D)UEVk-l}@;X*d#a+(PA^BGCE&5hrqc;rWZhIkei}&I7rNJ&Qn1e77MAyi|
zwr1-BL<A11)HY~VR}4Db<VW}JyyTzPQY@3MdN({hOlr5a`zX3gF-0m-<M!C(!ne>H
zyK8f44!8TroXZr4>^nHKzP@*sT6B>GKIT;8<Z<`oKaI7t@9?T;p`M&uW(cbCgck+?
z+sNNF6Ymn!{yp)DT_>@uLUyy#%dckdY(*Oo*YGNQt<|Vm4-U+*+)ybppvvQn*De^m
z79a9?cO|6C5YD}AcIgCe7*TKNXr)Sr_xQF8B*)LxmDk6U<uVGzx3_^tJik!$oEH9x
z0wospNw`Urznm|n-O$|oOD&`;vqCoWkwLy7*@pnFA3b^ItILW{0#W_9^<uM4K7W51
z*dKLaHTCa-{I8ZTkn3wu6c0}q*VLC>X>>NO+HtShduj9nd6h<T<P4;4@c5K%VMWZ#
z*TP(Tw7`d+yM|5Ng4<NjdRW5nL+n><ozUK{Fp5X!!gPbd$;Os&KI0&2@S@^?IH%cE
z)O`KY4b)TqrM0!(YlV%RhuOPFw~YkUtr(|$MzQV^bJ~SN8jT(AbnUo~K=T{_3CLG@
zw$au2+xo9KxO;>V^x&>;VxIbK#X50M!mfMU2Rj!YIu%(JDlf1(i=|Al#a+`KRvAiw
zCOA*|Je6cZD>G<XjegVL#;&PL^s|&o1~|^79nd8KUz%`7k>-}~UMJI{uy0m)8$-jy
zYtMJFS9D-)X^%}u90RqF9#g%iyb_*JUvCJBxZr?_1eOd#hhAaU(1<EG{6fPu9SgAi
zGkgKvcv^%#?0f4~G*l%Y{;az?G9FL%+03Tc>$S<%b<lHd<kpV6XcM8te}d@0JvQK;
zcq}+G)iD-#+q%$4NPinu7+}^}K95(cy__<sUuXC1cEC%l{atU62&TQ`?(LH%>G32T
zFpK(6$~n2|(eUGF-ryxXv}5eGIQhvIVvU6sDnlOnLwf6ia0{{5OC3S+OD!(iv&3<~
zXK4So?*Bjg{o86=E&Rf8V_eayPd(8YsOJf3@y48|f4MK{f}d8FhV{p<0XNH3-f3lO
z8O!BpF2~&H?)Fp7=`nL$luRU-yltW>rzl)$D(g;UIUJkG%_?Jb?fHGw3*2k|%b~6i
z8zH7yX~uvjbsY8v&kG_IrE|H3wcM4Q&3DnCbaI~A5zuKYFN^9uj>;>WDM&XLSc$%8
z#{qS4%bO&lW1n87n)&v<Cwe6N%h%Rn`ACOaYdy02OxLWH?%j%_5Bq@PyPdDTkf8L^
zm@kKaPDV=Mb<jt5sc+aLZ!yp2rgBlny6ZUurZYIneeK1{ogz)v!)^sgYtu0a&1e)x
zL&Y^GuSC}PUBpR1B*g%Qe8h3TjJZXcwdV=;c<XXeZ)XbHwwy+hFtGn|KwXGwz=yYb
z1+?;Cj=L0FlO%#rUEV<t_YAWrN}|fi1i1?20@Y&jI;x<K_q+veJu@N!8Ph%IOZe7I
zOj5c`O>*<c$LxDTr?s-iT%`lHcbRv)!pUBjtwblub>?cC<~f;l>LocEL`eW4mqE{+
zZ^kxEc5H>L9=SqMyZIBsY(c^i(Y5ymmX6`%JNrSoHFiBhta1rs9bwb%ysA^V=OXec
zJn*evJc17QO?<qzo$3LyFp`4DzzJi|K<SYsvLoh(b@5mK=3L2Ait%N?1;P^*T;r9n
zup?T$0@4hx<OR1H{m;$#b|DN_#EYd5!V`htQP2>t_j$rqqigdXImM@$QiguIZQUem
zr(nq2KJte2JuQMxglv1y9gkmA3dNakySG8WBmvMDUqOqOrk!ylx1iVIa#5Z88@&;m
zi?607S9_j_&|C0WtiouL6mss=Qe++;f;_C7s=bWhmDY%+n~k>nY7at>grke8ZO1Kn
zZtpsU-x1k!K)*|JdZLBy4UOhTD}yPNG^KY{C`u+tk;9*A=C;_Qr1!cF*V$g93~#6P
z_0G6(`j)8oLcwAwv7|Vb2lK++z5q=BvIFnEQXWbH*nm;CQeV6-vYf`zF6n(U!TH*$
z(-b=Oeiy!i?46tU#0=c6TM7o&it%W~a%=Z_JIu#iZ^v}WAZ%opxWx>;q1P`;Ar~eK
zU<h=;H^v2~_27l=Nb76_JIw$Pso+B%J2rjM+tgy$deumJ1odV<v2yL$P8bt4#XY;b
zwa>?E=ypI&#X^8(K6d6mY+xWLzG1&!?aS0oHVQ$cg0|kez;k|rM2+ucEgyqOZ^&AD
z`1o?e-yE_3@dRIQOg4n6ydEq;{WVeT&r2d+0y@61{@qj1lesHvyefSJrJHJhT;Zdd
zy*|2qd8{z73o6}&viw<3Qsnh%zqiq4Wdu^?6KHv0I1}sX0%H3Qfgnl#1xqD_X#P*I
zblrHP!v6s6^Isa}s-x-p6C(2A(ce+H@LSV(w%89ym)?CyDQ`#<xpF0}lvkx2RGh6s
zA86oR?F`~zi}JPvi@#f2#6y!8eXIr0#p{Q=X8Oi&HyUj3t>3w8-STv2>qz3mhe~Aj
z*ea^l-r&Q*z$9d1)bG{Ie4{=^T}uC910S#crEO1zF(8-K!|1Us-&@OTd-UBG8(DUb
zBdq0Iu<oz%VB)j}nZr=$oOrM>O;IVvPv=?mLUF+oz6Ej;X2vkiyD+okr(mT^i^=nl
zpu82t)A(@PmnlouzHP_%4##63-XuIR#>dY48+~{{(c5gtmnqwRJ-8n#jF0$(Leve-
z|IU<ur~h9_^*82m6_il^!775na)8DWf3O_SY`Wp;ALJtd1mRDaVE9uWM*hYNYJfu-
ze^k68tDG`;+(eQLevNnzQps5lYNGf93ecaz@TVZ?^Ac?TdHFj`+y5v)cwGN0o&qx4
zh~$Ax+HGF^7+b+jt#_ob1k#y%#ydiGEDj_$RLnlwc<G%G>C2m~;c;3*o$Rr^BF;dR
zb&vkF>{QDVBl)r?R2($j4TONxV=o^-kVf}GqQxa+$T4=vaWhx5VBGz{=j~oFwRT;y
zy_G|1v7~yU+C61&i{gl6?=NeHBdh#Dz;_`3JCA$NQO@HbFB#CvH+@R^Vsk!J0iO=i
zYgv&Q@#)QEJ;P(wb7=4fKJjlWkqW%iS0lS+$HF6IYS5^mdyjrYg7QAOfa$u7aoo=b
zgZ(Wpjaj|3W24@HAKkbRAtl{ezWoaqe;~rP^oS3A=`n|Y$QJ?c0rUxKVPTQhAclNw
zKwzZGuBV{IW!fpQ38+ob8l_l^LVzDS%sJK2M6Hg@0m(ST6f34V$+&`Tnk9aK2&-3Q
zN4>|-Kp!nQIUJNd!A9x=gpt4Y@!|d7F-?5Js<+}r^nP>KVPk)djCxoR&D`57Zs-6c
zA*4m{FZ90Seq0k&g=DlbtbH)%<B2Di+Z!Lsq$b3`G^Xb98DOW|_Qu}(*hz?Qifn87
zOpEelF?66wDKBbDbfX29GsdUU`##jB3s239j%`>rEu?s*>mi3BkP@?;Y1Rw1ynUsH
zdx&2tcz@&K%VZUeX~TeSePowuk6TgG3r_GYp~910gKAyp-Zo4bk<|}F{GZv990c`}
z)zd>W6G}pZUlOTH81I#y_`K%`!;+*VBxnbag~`-BuTn}q7J9NRDA5^q3vvVYv{gTv
ze7wGRC6+;}po)bxTFM<6BTxB^L@Gcfqwq)fQ`<P(n^z|E1Iu)C5B-yhooZZiACL0?
z61acaY1zfGv_!Xi;W5dJzK!S9rUfLkg~LfMY0{CpR@rUlftt>(3XY3UHJt@46;AIZ
zzD-wNT%zFHY=l7wq$@GFpN&j8-}g<Rim6d<LsBc#!IJ4xhM9Upf&(gvfxb`UlH+Mg
zB45X$rPpRGdRG^K{J=q1r)%u7Cz6#WkjtoB_e9P>)35u_K(if~K|;rvvP-)>&%eux
z!H)V<AP_Bzf$FPORyMN;{N~lB`s0#>4?g<Z>Co*t{McTzy{$vjG#`b*;6e2^16&)X
zsOrs#x0lb2Tv^<rmDw4H<4Ed`zH4i4C3JGa4H$~_wZQ$pZ%;#r#x|QJPZIKcqA}E`
zlfWJ(`WN{7ygSorI3>;SxhRJ7;t8K>BdZlO)AzRm#tg@_Y6O`IB&3<|QP_>RW7F8`
z{bzgec|K7DS|htsyH9`@2>TWSqHXi6<7Y$%?reR3t#c7$guOviHS;Fm=z;nvGak^v
z!Ad(J>HLtwu0K958NH*D9=XrE5#B;@88JDbv5T;4V&IOW>6m%rfAmn}$TW_ml3bsU
z*RBqlMhlx>P^>p&Qsv;KM7n!qn?YY(+B@OtgRybI5I9xBsvof`q@hJ}Hec#RwmbqR
zvT8I=8Lp6;Ut2mfO7~IHgvs$;^bWa&-fVn&l9-G00%31FES~Xscl1#0<PdM4&mQ{a
zlyuuJ<7gA8_U~i^vcdlaXORDhpXYMR4R&a>7XM+UU)E5i&Q=M4)geGbFl#=HK4gU(
zTNgc7-o;Xi+aCzy>#JC=T!Cc<UH;Kn4<n`ocjY6N8Ls{vQ7#>x@b<x2SD*;|1tiY(
z=q~*W!eicwHVmv$->G0*oYr0nvQ3%N{TN$7V>_@twr6q%nnYB+G)s{~K$4o(Y+75%
z7lLDJj$R6edOu!T0g1PZGs3y_Do9aMg<i+*O{67}K_}p)l9vek!(e^h=mnq#ggF*z
zQh`{?Km)T+qb9Qb2aXR5sJL3T`wOzR_<SZ(302H}*qu#6qkQyin;^wh;1Ce@=8ddF
zm6G1D930y?OCD|q__ULJVnkP%GNp%DW2JSBrk8N`^dixmT}_0iz<r&txI~i!Yr<JH
zkm`P5U<q$-X4up_lvr!#3?w>6a}d*p!5QybE`c=l1@3DUQ`p!6A2J&oKy1Nr4dM)<
zw&?r&54Kb`5ij7lyAFtxAm8hr4@#tZ!`OAb+#Isey~PdpJbDKK4t-L_wCVHJeav4z
z2FIi3KwA6vw;1-&RE5`IgCBs-);sZ7%5IUlSqD?Ur-yvh8#xxfPQBbL=948|Msqzp
zg<CJ{)RwRmMhkDh&@AS)AH9E{cg=^Idkp%ckK+uK#6UBTAPso;qotuYg=cKkg{oE?
z96Eg${NkK*szhCOn!R&gM6YM;!3k0NeD-#AMrrPa`S4{BneALv0Xx9%V(oJ!%ZFxO
zr>HHK`c2V(EBvl7m-V?fe|!92D+GH_F(0m-3*t}lc7nA-qck7I!A803{{N=Wfcp8V
z*(Oj?k7R8u5BHOr|4~V45H%aY27`Q{vahn;-`m?{T7^!Yx&a0xG2bW3NYV#r3?017
z#Cvi(5B@3BilFfyHSwq^zm9xT0b_j(P@))8XP9O^0gwlUvy(6lFY)*Aq4^1mcINE|
zkpczWuL6B9V0i#t=IuKJv95p-cvI-~^Q!;;6!R(QuNnhr?%!;R$jHv9{((Egzugm5
z=l$@EXwUQ<^r^s`^z#!cyQz0oel#|Ki3RApkU7f|SH$4xUhVs;3a34qy*Tv?1(b9s
z)O;_l?lt}y$Z3!1Z~@OVYO?nD$v)2pqJ{fn<9H#Ewr1fi{z){l3ws=)&$Cgw5K2i7
zRCeYaXoA*u0WDS?(dD%V!)o=_bD@-T_@pI%LCMd1Y!R`;1vG7ZO_ZGGbS#lCC;3M6
z*9~j*<UG`myWezGj%-~Hjw!oLV;~>Z&h^~t0aM>(ra@>V;?9*W;+izadD*FA5C_N6
z+WHKtX%(mKv=x_BsOLIVb*-DZm1XiF7M|!dZ=hpXG$HaJcBt`*h)3K(j<$X;MY#^k
z>(<#@0&?;KEDYU>r=EoqQK#59g~O-L+dGdL*I)vc=!_N{npeQc0hf2)wm0r*H$JXr
zLyW<+@3v|6CIq51zJxHlVVP#@EuarWp4-r5b{ZT*KMV6r(Pmo8$odTngo4u??e3Fp
z&#M>y>{#xTZos(h4b}LE@5%HK3ey>`EeXG}r?LG8JEi6z9ejcyL9QI(_m}VrWbN13
zry+1UYtWwSspW2o2}PeO`hZv)T-tfC^`}Kh3Gux60(780_cqUoaRpB1qQaP1YHkUy
zprN~cRWMMGy6X(o)z({B@Ajs&HECdP;p-&R8ORW8y=$DT)i6}$a|TM~;Cj}>1~KHG
zJp+B+tAuT<sd0C7@Tm+j2_X8H?JaS&!&5&S*;BpMAIP2Z#2y?#DuKYxAn`ofOdMRn
z6I-kgOZTPM4C#j{WYwRd3md;nTxdNn6(V6rzv_|_Id_ug%UFW5`X_N1ptaP?e>M(t
z#@O!~U08`sP+5Lm)rj1yI%db)1Hjou>Q7C0jGf*99cP+H9~*LStk>J(LO}lrM8>}X
zkpW`*vmNJi4DENJ5qRJ~!OA(_0H8cc#WRty6xqhu!Eri;WoMu$UpBy73to1qf+6>6
z&byZFq2I05Z=CQLKc&RG*J|_Kd0}>fmxpI8C+7^5?z9^SYm_H!$gN4m!4qF$i}a_`
zL6?I@&P#45aaDF+B|&&{v&t`lVIn%~Ds*^u?K5<WU*ZavT+h`VRLnK-kzqyn>`Of8
z5YvOH{<Wvu-r8&7-EsKhxksrhs0vVBb&aii4Id~6z=U_%x#|prT&p-1c7uw4q&!uB
z1<)UenFTNaCJT}s;sp3J_hiRpuXwv$f)GVs_7n}jdC&d6%Mkc<1jNZD*S!QWSf;bl
zFYldaAb@HnbpsnZvN*AyMu7=#alW!yV^)9SVq5Ujf&F<<RO_d&&I{iz?+X`Ocojw9
z`{e0H;o!nPHrk~)-HV`tm6*6Q(6UqbVw*MFZaTny_b?osRS=-GKK*WQ1n4+x5qxa$
z9Jf*lUiKwaA>Uydg5Q|*(k=zdlM<-#M)~c@cO!Odp6<-n!?&ODT?DXUIpz7Rfch}(
z+PQo#KVO~ACKH$1tfd<Fdlx|JZjte}!e*oJVjnI&J&%wkN`Uo(LvTG=1a+b0VU(9I
zWeT&o=l&z-v;%vSi6ThyO|V@P!R4T%41kce6Gc2@0n7dOPRtpz20UEZ?}^Iv+*>{a
zJ#DLmC_<`&aWY2lZCC0s4o^&YEQ7;`w`>uvLmrdhFj_}~-nBWj8{0}SAQ*1v95}5x
zX6vS%a;{e;%@=x!0$5vL=VJh^0!tC}h)8Ii%Z#?6?bLd`ollU7u9l9V!X`APbmziv
zqLml*I@MckNA{Fwc;f(41(v31)<G@iULK+sfc0bS&OpwXSG#JwUEboU;rTKz&p=tj
zle{N@ZZWHQ@uK<us%;q{fLh-HUKakTZvUKa(eEP1e(4sa=szkK0AJ-6Y{6sfi1FP!
z_+%3uvw8Tjq3yuAX#jN78NL3S>Z+crE*KD0z}9dMPjyj>0<fz5!fEX{oPI-{k~ClZ
z9K+EGz#1(5Oufn_+e5a4sz*YP469UTg8*p@PM}|QuL2_%>W>*fZlMXJKq<T{*kuEC
z!+#nOuLly)ljN5ixGEakDBRJT!<Ht>SQhhlkVlT`WG<&Vh4#Veq$J30s#Em`;8UKD
z;Po17+)GhWyvy%t(^gPfmk|)2G0|Gsa|&@<eY7w8apq$5u>lZNXT9DA7rygm$_d)u
z<o%n%5fu}xqgDVld+-Fyg;cHkQDZ~j-bFOz2OVO2;O+i8lin__!tx|j0&iB^{53$1
z7=hkX7n@UX<Dl@u0YTO*#6qd+>La`Cgiu0Y3ZJr&BxN-q$s~5F0O@BM_=GoKinf>~
zsJcL5?G|dUv8LB^`T=c*q4vkGrJfP-=E&bM!>e(=a&?PEL0^d?tm00G>hKmbkhZR$
z6?X=rTR^$2rLI<|{$>7$&%wARp*lh%KeL>(;^AJge?NG%t6nR=V9sp2lx>`9XZA(3
z`7_`T(M&uhjOnWcQpwN4mE2eKmgi}8bhUVGWA&kRl<}5L?cePaK;CUR4s3#rd=mJc
z0z4u+YpWEYXr9}`eq$}+t+yYbX8*{Oc>l@&Tt9^4&@nJVmf&{r*s=qb&_k_K_2@5?
zHF!A}W+r+gu`qV1KyOM<9(ol>$PFBt27(V(7EWDts_w>o>rC7S%tm8js{aFVlXOO#
ziz|ALvtGf~4gTaPuTrIg9-p6tAs;Q7(CPz_GmyP-P@yb`9?H`R({w${O3y>u+NNq7
z6U4rL^Dq4U*Ixp!bgCDTtgfAS_Kme@qdz-<D~sPc@jb_h0>(=}R%ArvJ=O$7oZ<=g
z+PD2PYsf{dnS{;i^VNC04xWYCKmYk3{IGgy)PdH42N#i$u_j8;4-j+(C|!o!_L4ZS
zu8_`MwTHfb;qfXe6en@1)TY6;s?e7bkH%Hk5d<Z&==VLZ1)>92<Guc0bn~my|JD<s
z+VLD>`2`{3OWp0-#`>RxflWaF1pPc$2O4Kw+>v|p!ZEL$k9<<Z9~YKWckj)<h#?<g
zU;qL0ul%L#WS`MJ;xiD13j~W?B$V+}7ynUr1|swQ=Qq0oo&FzjkUYWx(YbsWE+V3p
zgy}y@ZPf25Pd}pJVe9C=rDG=t4`NVN(E$w2(FLFJYOiL%b@z<sd?8S_yL_-_xtRNs
zmJs)9Ew61Au=W)_kh{0@k`d~m-#-0nu)D=9^7<X6?l3o}s*fI_S#sSxy!y{A)f%4X
zbyduuNvseBG8yaGEXKzUvt9h=cMC6K!t9tsrSv|K+M^mQ9q0usD<C^fw`I~yq^q73
ziP;XVyk?OT%G;~FVW&{O%>J>}RX-x|j_rV4RrYMZsR!+4L;K4{zW-9H;|+gmT@3qz
zQn!hL?AGRIyH3c|DraW@$fQyoahb|0(@vP*c|`X00pLfpH1_y3?&^M%jJf+#%XG*Z
zXH!R%JjW%b3`o`(JJW8qNAw7~*O!#w_<$zdlJ)Xtzh6*Cu~{V=rOJf2gm(U_VkvY5
ze)?_p3BsbjQUS=Y<jYlT3AvKFdr<M7)q8kVoyX&qNZRE$Tb0`*PRfsR3}11bM&G7D
z<FfAa0bx{by2HGmqX+Ly=3%LBEG$f_dn#_K*60Z*z}8r}8bms;={Ovn@W=um5Zc$B
zDD8aDV*#?Z`dU4Hmrq)1=6^@qg#~vn6Lvf9nykZ~N*>s>g8;X)#knQTb~rE9;I5A#
z%U!3-JdGx%!S=Nx+62PqQZ;J?b>Io!vLk%WuzF&-6OCs$5lUTm_nblRzf_B$Uf)s4
z#j@VuJxF^G{$NSp!C<HZbn)9Q2?btWf#dK}<2OLkiW|{~*^wu&@lmKA@r>0tqsGZV
z=5yHYT!Hs;A6dYDAPd?=2k!1G@<BJAlEWzb!C~jK(!HTByN$nZgk|ESaXSK8H@2v;
zujpX0@t7P&1-!9^$-FmPSVXA(c26RA2to%__~;!idK25-nsElQTBh#+Vsh9(cE>RA
zCV*A|T?Pn%d9`NweeP0hMn=ERJU5vxI!M<UNO=nBs<MCM95DznL@$fU<GQl1#5+{n
z=K^>mZ%lS<Eh;+qRLet?xiW4#FE;DttWHX}Jnqqhy{k58u?By&i3uk*0NIi=(0CX#
zjy;Q}<2ezlL9w0A7`hV~j4H<42i|5kWXZvl{je|s*5ZT2Me!~lqSK$Vva<;)Tb;2A
zy8NGq7N_{dy8?l`CMvTDvRi$*v0YdH#Pn~}^8E)UrDDQck@>aa_x9*bTL^%}z4tk{
zl$dcGox!hvz_*U0wcOk-HqRg7cJH%oDKQ=XRH!}jRz0UH00i+?e{XB_wt60Y$8CJn
zYlEY#{w}zosi2Ni`oTTmUzZwo967hNm=5PH$tS!o6;uP!QgQ%Xm$cyTSq%q~;H@SL
zJJZsIW;e!#<;~0GEtkt@%uJN)cSrBTwu(cfhNtZDSBiCx&E{sGtnq!pUwAPq2RNiD
z84oc=@F{DLlP?urwVi<Uyw!sZ@-X}Cq|$U4P{vk{*Y594nE(J%c+Z%qY$$j&h|J)r
z6&X&tUl@hHopKY7@NVoLKG@pmzBpufWVbbjTmYN45E6%=NVk{{>T5f?a6EC>y`xSz
zPUua*`-%34yFeW4qDp-92>+Mf=5;2Rb!B6CGc^|8#yze7O6Bnhk9b!s&7ScvGt8<I
z_T>W244O9%SleYhzl|d)p~o7Qo%o!uCchhb@MXR3aVu=irT0*RExd(SC8$ci@}5|B
zwVuIe3_0l;C`}(c1Nc!^^<QBrikM>u6uB`ib<cR#Ka8G%t`K1ADE5{3$~N7D0H1xI
zZDUjesCYGgMr9gH4lW;qb);LYH0X|5_&Mu1K~pyL@kcG+j{bDH#7x7xL*#ID#91#H
zY+>KnRIRV*5M9*L8g&M0<K6+1r>9K#O`y}^YU5&ZSXSYu=obLtZqjW1h#%vG_94SI
zFv;%$U4C*9`JM%bRwJD)d;|5ayHcn*-as2eJwQx{DJ@}c93AD|>f@=zxUF3oG92Cj
zSo6X^A>Q+AYaw9Qnm^h{m0tDU->l)Pswk5zE;aV|uyQeHp~31r!#yHUn4n?i$sgU{
zLCwn9@->Img2@$5^E2tsK(qu6^Npc_#X?8Xf7N|IA)6}B@z}-SS;YOn-Vfl(FX}$g
zSouvKY*^rbZ?f?I3;K~p1(315ZD?_D>zMR6D|L^1;+Gxyn|Aqs8QIMmf<P=6CV%Ui
z{2%NY@x&BrC141*<JBpaeS;r*aOb@kAlHDRER_YJ4d=8p-X(ZH1K9B3AQ>>7KY`G$
zjCy0}!imYcFM>J;lF_3U{)XD-Ec95dXS_Im>Tj57II3ZIY*Fw|%g_}pYaEZB{m75l
zHyE87fq<9N-Mbv$a&n9v?^9LjkD=eJja0Q-@iK2M6QV4$uOK*e$N{&+pop=TbF$a$
zSPFJwQ{Adq9-$88CMBfjlqEme^-$UB?PN^P&uVz&Za-6w4DJ&3fu9?#s>xp_>kM@J
zYC-eYfI$ePQPSrx#;IvU@J<t;M>N>)t>J9~h&ey5Sj{;Yq!yLCE{6uR<1ZE~bI6rc
z^Y@{^>0)UPslah0V6D$Bej}wiEKF+geex%+Ix5dHdSIwJ{&XULq9%uXL{xn3#9JmG
zP&ouJs)60BW6sACC+ew&VQNi%dv)=y6ynoG6X<L{c&jgK@VqNPm^NQ*Jfr|qo(~MI
zy>Hx48(skN4-meZ%V2d}nWuN_W(4fSMx9fOih^sSv9NAcc0{NdTu-dbyB`$`qk4HN
zVfbs*RI?RmekvyRbJnHt5Ue~fu-<u13&}>`*jJaF^@FQOJCKbYE3K&Fx>|5|6NBjN
za@dcaUb;5+kbEf5x&Pjkf$D<-2@?MO*ctLG-Kz%j{5&i;w!UD<oxNUOiay07=eccx
z5MWE`x;N^|j<DO3y|$gfqc7M>Gm-a>Vz%-J=D-FXh?6ztAr%dM*kJy8ku+;Qv5<lo
zwLQgqX1ffxE_Cl;m{o;MtqRQNsO_y!*_$$80<75Z$sEUl#6e^y3hy=>VnbPuCokWF
zqD}53AU>Ehs9vUW#fjir@O8nMEsj2w)|I!&&8apaN9={LPZHgvt9;bS*@&{XTL$6K
zRXu1s?dCCW?|}VvwT&|n-|6#fGCQ+bbx=wsnpr~SGf=m3<6dC;G7edZYdj|LLvQR~
zSn-PsZ{og%%^zz_195Ny|C(^SCd+%0uMc?om_V93`^FxdtYOe2^{U3FM-o(2*f{Yz
zq6|1r(-tsTQCCyu_o}*{h@XKT5F8Q%gSImfY*(V!xeoXBa$|S1&jt1~(1hgF>2*~x
zoH@}Ztt@VZ`V92b_xFdOgQGa!CNp|$u0G&i1kwFnYp(56EElF0jT-Sj!q(I5aX}!@
zlWv-6{_xX5Ahz{)5GgGF&B-M-z(*C(c+7yevBL-T{R%!%1-w>Lp#Ofq|9qDSN&udy
zyYCUCEXO}6HZ}k_Mjsxiruw{PzxQ~D&-*#8H25S>RRJdnsw}|XC)lqhlLGT_^bz4C
zSbgqzpW-<_95xPL+Z3PVWxe;kcP=X;wPpju2lgYRPo!jjw>2=Z_5+l;(cJ_;YX2VI
z^?{N!mKJpH@d3~>jV$;icdYFU#7r;`pC=^2b>FwnzVN*l@YcI5TLUEbo0%In$^qC7
zYvR6r2HMU8ytE*)J&A7cG<oFd5I5kjzV7B9=L7_z<Ip|2uRe10oM97F^z6!U+T!iT
zBk({swnTeyV0f}26;qg+vl1MKSoZXV9ch<0XoA>9Wx64C{5v320nTGTyc?9dMSuTP
z*EI)0`lTn`z*D@hi4yRkDHQOOPl<DnLKA;EHlLn><}r%fr#jsH;FCVU;oY6Hiv$+C
zkqbFTX*@|9V!nX0^WG^>dE<JGCGMpv;EU%6yo8E4j0=3RiDFL{aHZCs*B{C|O1&7%
zWUd0;u2X+j>r>!T)iyhQ^*DV|;TVz&4AC*QFXlgha!V-q*?M`))d4r&P&tf}5A(@S
z*M#~-Zwg2ig$fqh=(^ZMMa^I?d^COMO5=8kJ)#YSrwcv$fN8R#g)Qj$!nR3_h_ZYZ
zZx!Re3;<e{$2$Xcf~Wv#q=yiEZ$!pi*kuH%0CMj<eJ;#nOMnV7{PSY;|AXZI@9$C^
zfdN$c{kAXL)TEKercmhVAUl29JKw)k6eA+5#xfG=O$Wvx3mozz;G2o?N<%{!3TcLo
zLxP6<hPMT|Umxdv4?$J$X2o02Kj7IRL1MaTyiXP{7hik{{V^Lri4$>>xb8Y>q;N8q
zn={{fOp3blw)dS2%%=Xylc^g2l_umPV6jeSiuZ`2;g0hb0bqa10FNdmY(2{k7i41S
z=C*)+189`2_C+;tX`2FU{tb-w1Cc`G&P}MLZ#fm>7}H0wJdD_K0SV^>#lvUiPzxaJ
zwdb|=WFC26!K$Q^EpY+y+mgc}jrx}?nZ!e%x5>!Ij7M{DL>LHy-=;5g<E{T-`rtgS
zlx^7dS{Jun^q6D6NLR~4Un_U$P(?1Iu%5LkU)JOv_~ui)EGcCfv-VhdTRE|ZVfGl_
z$!YA%)W6QU39rKnxEhb(U!nIyI2MBtJs7WnAuhu9CZH05RG|Q7z>w%GW97kwr=h3c
zvs57L5F=)VTaZl0Wv|}%!QkFS1r|SctjW%a4Ez6FzYEdjoM(j!g(3)-jl0xGBd(@N
zhi^1Zy=ql@Cz;eV>@RN4U+LQg9G45T2I6XVaH*-PT9)4VO9@BS!Jck~et;jw8Qs&w
zZBJFrp}hXCxGL3n6s_RANJv^6G5Tth5Lv-<TR?mV!z)XlY2ghQT~CU3I$GEVoHU<s
zXunO~%9SP5SSKH%Ey298rO|w{wpo<8Ga6geo3pS=At0>-OYB(Np1o$UzZ2cF$jd=w
z=1tq#Vq+9!J*D?-pufpkPF*9PT8}y#E>7xBOCn_$Jf6Dh>VF2JZ`)MhH<eD;iWiy5
z+<Pz>ZT+31sXM4USn0~{!mvqUj8hk3C34Idk#+@;hJW30iCHMG6n=N%NCtZKp&{;#
zcG7#4dSZP?yp<f|wiWB?sVT8}Q+CyNh8`Bw=W2%;RHwnA<FvsQb*O+ca(zoQJg6dP
z6y~|(igg~#=kBxU_^z{FS7=SiaZ9?3AkX<FWQvSnNE;Jx-#9aLB9!DTW&J)%p%0SM
z<27jvBx;Fl*AzlV$jA<*fw&JdSXvy#Uq_xaX*y-M?MjMjrI@~rJOk-L#r?d;f<LV@
ze>87cU^{(}N2TQwJ=={~C3@%R^(B4G2^w|=%E`%XWZpaec$f*><tN%!f#xC|RG@j{
zT;CGxs8fs@nHQ?&8_6-;g{D>Eg9vJib)vDxI74B!;z1+^g`V?_P>bjNL7gIrhO)bP
z`DEKCuWcPq4Z4+q9Uyk}*P&5PW82X8&V?mnhI@1zz2hEV&Oa&1(Dl&qezqkeKK<sr
zn0XFCl-hHiXAHkOR~!EE9s^YG!Zq<EQJ_}@V_?q50OY65Ua%X;uZdjAud0V3F;nWI
zy2Ujr>rz@;qo(0h)_M1M8J|gJ#0c@GQ}Qs$$e2qHVaa()LHPycsp+K#R|$Ud#$7RB
z4?X=Bur^45+=B8qQroX<V56w0G>nB7Gu;Z<dai<n{0hA*AvMn%Xm*Y`NGZc{flppf
zr!=aapVYmy-pu>{g7M!_>Ja;O_t*2nhLMBX6;Z8F=W%ZDyFPN59d&4w$LTLDs}Xec
zG)B;(O1I}J)R?#}z@_uklt%d~W{-Q|jmN3nmbuS-eXXbDNi=#z%c)pl?=YO&9zf?|
z|6ShXgXq~c?=Fu9qq{73(r2IXKJjWI%DPi_`a$gGu{&Xy=8x=(*x?#+uBq{y1MoAB
z{56&#m;5|6zNvAn`^CFrL@+1Fo1+Sl$fbrQ6}wXjvOd8B20}00qRgB=s07G{A!wIz
zW+LY+KDHQ+KP5Re&V16g4Pr{OUsvb@@^@VMc?%nHA=Q3eg2b<}251F0uU~{Jtft14
zVvPG99xz^ieeJXp54d8u99(q(X2%!RMO2>yd1cESw*am=mLR3JURO|IqX#~5M^4bZ
zbKBBZ*~bUAjsPO`wGjzW=x^j;IVXqg2U%dj_LnI9DU2K)e+o|4@WRSQ?v}RC((o_x
z-9+`*uE>74Km3ODA`~1Ib)tTHa2!tg;*|K8V5OfEROLh2f%xY?m<G>s9Rwb)U->0h
z3g_bfm)?Ih$qa}}#4nbs1ER$E2Y06HFY=UTch@!?+(hm7Vqkwt-ZN<}!)$wRKynP_
z<lM<6Zw_mYm@b*bNWJXbW1NSTFfis9_4k|nl06>r)Vq%NZ()<CWZQQ9?D9&q2v4*c
zzMOX9IVX2as{{)Uf5gq#*Gs17+pz33Ox9fZ(2Q#qaVMJFCOBf73LmpLm#ICbHYUJT
z6Y)y{f1TWZ>21!%z2o5XQ~2)!t2PT5g1*jQ*#0_+s)LVd3C^i3{U4HAC-}j^xt@q}
zFYRd2%mX5ipt4ve!zPQX5f*86=}Z&ey4;e%v`vk;=Z7HL>ud;8j75Oz5@^~%3)!5~
zO%bw-0|O_iL@6ATN!O#cY0p3m7s5_V$*DFK{G3lt5$wE%Yz70CogL;uH3#|eZ%49c
zcrO=!j3j&gJV)22oLks4z@EKce%E0kH7HhV<LEAYN?$zsQ+MzO)BmRNzG{k!ZThdo
zI{3%-|2^5wFjo!IZg}UCjMuG>+KJM+qPGaDBNe#nS6io7jfho}-q64P8Gj*;O9<9O
zLR~iomT4Gu@1TR`MCD9%aIGgccZ!LGLxub;y_fB8`5Gdvq<h1Xn%y8@Yxx_RBvMl*
z#U<BBXkCjz66XyPR+BOBE?N}Rwah<AD2T#x+f--wk!m~;6IdvAV2LE-^{!mR|6|1{
z6o2!tApDQIo-5b`rg<+;e}X{8N&;f9xrxFOww6#m#YSyFqUd2v%vx+ugU~FY^Q<?J
zJ$ifR^n!NRLp_vj1HHp}U__$#%*!30b}_rej|{!4SPtCZ#Mte*ADyIE>B9nhBN%Fl
zcX_<)q#^8Z!zBKTgjaWtJl}4vFA`-Y6ZaZA-*z3Al2Xu5kr+C#zh@m6g2*Wx<D?x(
z+lw-hg3@JhyZtov){6|&&f0&C0LMLM`+QHtNZgoOx@m{SDFc{yExN2-uijppY0^^Q
zNNreK^pN`3^2P*eWmWzjKFd>Ed9oXmr@~JGa~_zc>1b*nq#su6RT9CL4;&G}KdBp1
ze4Au@2&ceLGUN0^=wdy5lNx<|L=Th9=mI8<APti&{$fO`1kqw{arl1^B4t7la&Oxa
zBy;3KUsdG*${1qxQ3=7bHhcu)A;&6wU)%QOV4<-~(cnc$04Y6wJ2)oQoUvX;gzvqx
zejYVMFittw`81jouHRfo(0yZ2+=!{O$FfiAj34hFb`ou^Ybn<~6J&3EkyAV@A^K`8
zR>5we=E<kt+D`m&oo~ZOZ1Fb3hDb!!bCUTZ;1exXYw;qw$q5PdheKm4ym6o@9sAq{
zqYLhasR}uiJ4|kSSUa(V<rTcwmE+ZuxJ<Hr`qZY(74Mv7zKa_LRfAAIlFGeB<>Pwf
zX{|500<i56zi{^eUgeT{byCIB(+lEo!L9Zgus6?MT;fuuSNN~3A^(ou|3ltaMzzuP
z`$7fUQcCdxg%;PO#VOk2F2PEIv=k{WB{&p{yNBYmNN|Th&=x1Shu{vuUHjZT=iGJA
zIcvS|S?~RFKb)HnlgUVC&z`+!{`22Ikw5Z!OR*1cMeTOrvzeKBzR<dPOZ4(lujiMQ
zg@QX(S}#Z<@&C&n@e(AOgpBO@A-ElSn1@eat>GChXpHQ(;djP+)>W@US~>Bi3+K(-
zEo6dAm^;i;Zp(OdVtSixG4j3{vJ0q<QA;O;AfF<~|0%-%lbQcu%*;ciw&qQx(D!|?
z*9Az=U~QYm8!6wHyRV1;Q$Zf-d9vIx%zvcbAL2PWtPqJCS705KT31+K=4(Zxu9dKe
zR@V>fQGS2#a!0Xtk7?bM3zJ&bakn^2(i{zcxlnD<x#E-U|BlVV`zlaY7TrW<*%sRP
zAf}&*&APZ#SNyF4hB|ARH~W9Cg%#D~KeVV~oviB-K&wy5!|kQ+@MhU}t>RS<MFlS-
zjJ=V!HLrE+1{#q_9G|m28PWC7%MwxU_grAC+(gemV;?z>xc3NGxkGWTS2)&?oa{`K
zgwzM+2$_}si>!27glZiVS{(nmO1YXB$@=4D=x(QEb^<@&3d(rTl!*xlu^amEa!7l;
zQ?|+g-=#l7d(pxZdVKcEo;j+}JBg~d+L`?Y-D#|#52i!mv}+ZohEc3SM~c$|urV!G
zFOTW}lu2uW7_!bgw`D4A(wlR)aMk`m{Tns-A$j6-aNh1{?6imyujbeq#}_@|%=C|S
zL5bn=5q(`lJQ4_A4Iz;*X5yWQvh5i25w62svATVo;%}<@`GTaHJXZjFFXD^FCn60F
zUH05&8B|*|ad?v5;C{&`821P3mO!zYG>1j6XKsX)c_vA9>)M>LwxUr?60QJGFTNZm
zZy{aP#Ey_>zD%sOts^48jhNqtHX?57FAgw=ghS4y88?L+_WsVZm+!l<FK7!~U9O<O
z%?$H#oF6NtbSceN;sum#K89wqyf23SoK_|(E=6EjL+<6M>e0pV9-+;D8)f_reEB7v
zZpST6Q4>E^wyb(lEIob#TwVt3P7by@>iv5Ux-d5V^2+~>YP-@!`}ZdGTE_%pt6Ewg
zCZjoboktnUa-br(wmO{gw@jIo(8sFth^~)tddE~J*^F_HUZ<&^Vjw!9?Phw(hneQ+
zICS!ib1BNxNST;xzwU*n9G79AY@T$Xy!0wB{c7_QWu38s`aX&zC=Jas%hs_ElB>8P
z5czQXu*<Y$CcAz8ViMGd@=|ZR-}diC`^QsLDeLgi?>Nf&#M0Mfecd1U3Pbg^#{>tT
zT=6&iM6&c5R$j47;AJwTogJ=Vn>Ki^nx$UegKRW9r|V|s`vmQL(2mR*|MTi4>6_ho
zu)(RP?^`-=Nh$jV$Q{3XG;v?>-W5b3!mJIEH;WHglfT~C1NeBik+)<X+yzagSgkK%
zr^a#AP>e&X6G1ac$crqK?~jC06}JkwESpzwi!YAa$?jgWF1LZF-D31--toTJ#`N0<
z^ZWpknBGl15KdIYn7{~$_Y!w*lw>jfJ`Bw&D=+W=*4Ik$sJ*zvBd7w79a-V%q24%S
z5Rp4-W=4PJeEc4%`J^U0yvoI=!~Ge-pY#ZI(JYgGJcletJR5#sJN63CGK+Y@`Jud#
z=cPOD!FYDCe}$KFe&cxr#KUP_=S{|2>jlbto_Tq>`I(TshzU^+>Egwy_8OtqoRe%n
z!NiuzsAi$5O!lJYWQX-RquU9exiYxYQ)}F<Rbqg&8`3GHh=-!c8-0uw_hxc7yf?W9
zVafM2kILCm&>}-PJ}skN%LaYZ{$pzf90xsIymvh0(@Y(F+cacV`Q=x_<M;IiXTCHR
zZUp6G$d&I!K6fT_j9=?J>cR8Hjt1%Z;xBoHY-1>+82^O9U^#Yh3b_NB(q?Gv{!n{4
z3Gs$0)uFW~4Gq)@AfAU=!)`C$nfLabn;b<qbgL;gwtO|kM|fVCubto^9R2EM6<@i&
zyGMe$iIT2CDA!i?I-0`npL)fE4UnR)0@Wbmr=K+~2L#2alHbfHqS-W6X+&yL;}!V(
zX5D@M{1CA-Zfba;RRcs4-&%qZr*Nq`G4KiE^SN51mdFuP`nI5NW~X{Gr<Qq5f5G~V
z5>LHtk2xY06ot@C%I%TlM&8;FeZV9`&){<;BG&v^Ue&qAbKrG3Y&2te_mv_oPRWoY
z-biW6_$}nLwFi%zHOt`O$Q$gPZ87=h1*>7RyEBk9?g+oMlA85;ZNXR8Kl|VNKs-Ow
zF!JgGiz6p@&)^~(PilG%3?jvkYXMj#6|;^fjKUel_|dIRdqSP3T1)U20D{LpqBcGF
z1kE0flJoO>&0^4Te6dw9G&2MI1y9qO|CJ;%ndCKolegf?k8QjtPh{=X+k1E{F=}2Y
z+&AtyS!EFCdbs}KQ7`y-xM}V13m!3;1{ZKu;El_*H=5+2X7R{EFT6kC7et_ThcKG=
z$WY}t7`)il(Y4zqo&y{){^21r^Xi9(ct-EG!LHuM3oq|nkrWT()#Fva#A&$bC&a`y
z_-O;Ycqy9$FA|Numx_hJrs_1ZOo;EP-C}VmogNzQe}gyvdky?cQ~U$Nw)GfbgLWM!
zb-tNXF|nicc)8w~-H82TwMHwdAhtqd9U4KV+;?g36KGNd@A^N{`5iu*xF1o0ZEOtH
z@)LC|t%b11Cvltwm@CLKgpbkK-Iaij`BTbhub|7Hi2G5%W(E?1<F)Jv`e5O!3X!g1
z++kat>{6iJWw|!|v&VEZAFHolBTmVQUdxE2xOIvvHqaG_y{jy-)u{z?7nMzL5!G^~
zC0@}>W}ixfUcNO)J5P>1Qh6){f6k|@+6Oqd?37F^pTcvC{4ZhLzjX5dd!zr*|M~)G
zJvrvRkITE`d%wtCzQymS&%<p*_=}%3RlIpks4Xk}O>`%goUp{&i6_0>g!&#WH(4$#
zK}}npLhP1W$pixvUzAFoe2S#>i45;#4inx7rnQ5-BM0NsA%t^jGSAiqlbM1ZO3;v^
z4c&`aEd|oq6(Y)tf~b(TJz6wXTabKZ^$<Ay-Lgjim6U^QNANFKsVVJPpC2jU+#Zhy
zV&zW{c*#n?mp0_D#AgJ(kJ0Dj8$V|pWo>9XQI3(uv*>2`l5zf$1$n#wym+NHIb4fb
zXa-9mLjF8&9yU6Atq1gCs+o}7$`-XO(%MS`1kVkRqoSGR=t`3>?jqmfPIaZJ`N8cA
zse!}CyF<}mH|J-MTm3|0lbQj0rL>X^PPk!GxY(cEAz<qV58evg_V{{%W|_BMmyhSP
zjQVElz6q@QGzFK}irD?#Gt^lgCHKDNp9=fNTJ`qrof=p;laxJPg!eG@--r4i`MDo(
z>ttl8-Q7=A>2`E4wRX|lizK<Y$(+u149JNXzcVH?Weh%ED!;Qz<ds)ihvBcYsUa=2
zLdDp(zN~}nqJmmJ_`k)~%)KqFB>l3<n8V*5s4OSj0P(BP<axwF#Y;^aa)iGaI<ohj
z&qPC1&sa@(6pdm1nYE1VAHcPwuMhicSK(jR0B3Z}AhcNUeP}f8*Ax(7{L~S9eD^QB
zD6e0caemf>ONs1Wj5J;n4@T36G=*ycDl*wRDls1!n!G#+8ElSZFYii3EmGE|0+OjR
zje)3OwJe)YPt+>TI!k89*BKR)9JGty8@-Lb)f4axh-`y8#$Dm;D~?rZSbFVWwvgi7
zNs1<QQEvISDN#)gCmGJ76O|$6Licy-+Eoqf9JLv^zh89$4JQXYmm7-uXX0|S*B}(g
zAV~^Q7cadZdLL^f=akMD-arX6wep$=$dWk0*fP9+HPH<O*O7I_mhQHMveBb=vZeC=
z+5voWY+Bjr9H(r|lGqB9Sw0gvPU1I6GaMp@$)xow+ysz)61WmfT()^08>!QtXDq=k
zt5@Y<^I?zNO>Hl-pi@^gS?LG1VHq<6b*3(V>7y8#C~M{!2g~)z{S-FXT<@U4VEp+B
z&rBe#t(*C{g?s>IB4>9l%BNstMp_Z)ns?kMqm~^SIiLapfnSi;xcxqxPj~QgARGkn
zIQ&AL74rp|nPxH=O%(Z_$AIk8MrX=(e$rCe73BWP(`D^58?6%HNH-Vdh>Ve;JtZ8k
zoeFbe{9Wb@T#zL!G&(<x`=HHGYC;XIwH#&M8q?YH+y(k;IJlpy_}YODMG7at>=Ah#
zJt0h7Wpg)GDrSYgV!QRvZJ5a_&dIhy4BAoiqb;1AE@&g=k^0RkmV4rvc0hAl+v#4%
zmL>K0#$f*qAYHV-mDc(!wVZX5{cvGv!o;Hi`yXUUD&I!aWt5zaq`gL#<)zb|i1Qr8
z15+mxNmCPLowvLk-s%(we+$>9oS3xw{If5-fAbVN6q@?!U{xS_rO!IsK2d?WrEeJf
zigm?moFj22QC|L)k=wV<S|~(6iL+=o&r!?XntP}(asuqH6YxI2fr%U&#T&e#ovRd%
zb9lA0RkQ=9qM?uMI`YiSs0Z7Q7xC_f33+qbTITDpb_I}$Swk=R7OVwa!P08VP8;)J
zTC0c3ssr`_PJqbwyX$%u;ve#v#`CmFo`c(MMoS+gI&Z;Bk8NfJA7IMpl(&{ZT-7YK
z$Spd{mQfc2T8$+t|0BYQQ-I)j&F)l4F8*;E=GkjFaWt0`AsG*ZQ9ChLMh)WzIa|&H
z&^qLc+6X>Vb|00O_aWu>k2?)IcOXP)meCRqsYmG}xCM~AT%Qy@S(Rfn!?gWb;p8nL
z&r<GhtcV7&FR`j2e}a5TX<)&R>h3qfHV{00l*${8Hsf9vvpE%@qKTPXbxMX9wMhHA
zUd1d>K`SkVyxIo%b4Kj-XW{444<UU@w71YouV3wj3siQr^Kxi?Wa*4i*{VGJ=W>DG
zR&0Sy;I)-U*ym!U#0d$lOsmgM@UfROPPMF`LGflEzn1F}lJ$3ZOLe^GE}RywE%ek*
zbDVL87CLF|J&QtS2-@`ucvRC!UnKtm^%~7YHt3SSDmS|J-wQUAD?DH#K;l%O)(`|D
zALihW@-=(jI_-^~8!031rOlA$(o|i01@+QNPg<R;T#r4vB4!pc=9B`IY+M$;ePhA|
zF!Q;O(A_Q44O!T4bGi8b@7wBEFQT_!n13Gg@<4z=7$>$&mRwFHzGZolotx^z))$&K
zgqclVEh`uD5Y7tegitqtFfnz&vjjIoJza*)qPKNe>ifn(whmQHbQsweoxb9M_Mc<g
z&x&|WAP=KQXOvx7skzKxYltj{#=evVpT;BfDfVME7PQ=wqk<dk46POTVV1*m#4=PK
z-0UO69S<j-ZTQlnw}fyad8pI`P(BH-Q3iC&6!-m56savZ>8q8@zGYLgDW;_Q*y}O<
zifX$Rlo*_T!F3`2*tlwYZffR<stRnbKuy$^c9+3iHVx8!2cS+7mOEw<4+)`6p1fov
z$c=~D#e*zlE#%FZYfEACkd_t*es{eCk=|(VuKZR5hu$iIcFWGpE;vgubVfx@4P;Hm
z%qNS}2((!$w8_t9dZT|WS$7c-X0<RWaY_1Rr?>^(_)(}@CVsc*l&g`3h<SJvu?1SI
zNg|Ddg&Q#Bnmy4~0pB>IVVJUoSW&@RI!1ScnlGHmREE4?pBBG|SgRoIym}kwzv<|z
z<xG^qS1;n~=h{{%5I+?;2dj=96r-y_X+rbN%+dqSMLQ_i-d}jZdvA{*0(??IVxmVe
z7Hj1X%D2~<(Rvel8})tX4aPsYpx__Q<LN~cFZi>w8dPPh);M_tP-Q0omfopZKEWsx
zS2yEE|9LLxgzd!E;KY{xeX2a=kM?z^%J=o|^*k;O5NJxaP>5Z%7$1pC2!rH)=HoLU
zh50XRz`S)B5Dv@hRC7s66k?x9OKgIxpAPU@>8R8#6zjYPsgCA0Ke{h33mcc0msj~5
z85v_NPgLnl*ebqFYh_<58_3FLk+Ue*?zeKQf)E2*C97-`8}ebJlWVV+fR>N?GJL1j
zKkm_+M^9ygc3*f;DA>`OttXBRqD;RvZ0E=qwqP-%EWE^I&kb-&<Uhy1K)A1<$FjiH
z!?}qYwVc%0_Eig1fn(?izjP?d*wfljf`E}I46I+ah5fPV38jI0NdIYJ@GJPIaSwiI
zMp5|i(PB(v6Yq=L-8f<s_Rn{pYqq}Eh;4ag@9lB<Ko(n8;@k$OQE2PGt-a$Yy9_%R
zS`6}gRHqT*@d1|%BC7EF+YKE+J7oOa&u&Oc=j0+WCFS@|>_0vxyY^b))-;~qe_D(0
z#I<T{=lk;Y>P$i10*dFx+uq{=*kx;~rWu@a6)HL3Kr4r2r&LKTw+pN&jntSqk1r8x
z{?M#nz3(Lkbj=%3w09|ThC|Nv#7$~a`w5Dk8k*X&s(w4(RQ(<!$R;|eUQxpn+NnAr
z_M1(;PBFqR9cN|HU#>pgUTw@Y&vx)Ae><#mFFg{{N$B<fCuplU=2QjA1(TA(&hIeu
zOQ?R8S&~sb1z{W89kNmqb?dB~%ClURvIJGn<~I-@pY%t@E*i5y-j@eWvrV-6g2`A>
zb;a=xHhZmmIt{U`+6=G((MCs&{07d~`AYWu`*PqFk*%h|m0szwuJu|{TiD+I5%mBa
zR1Rr^vJ5}Ji9XRQBa$8t9FRXo?nX$bANAI8QR}uVB0e-NrR1hr7+Kps+bh>8K+r;!
z`ITVlk0U~h8r=3!KD!&O*6_l$!aXhQNx>=BqE^0sGd@a=w%TFh<zykBPSnHCrg<uH
zK=MBMBe&5ptN<zOk{v`+jw%oH94gl)%6!TK(tB$)$Z&S;ei?E@1e(gL<$c&1Xb2qx
zjUlwOtxfuHsL0He(`tLn#sZ4{QaDum$SJ9TQ^`Rh`G(CP#aZewP96ZAicWP}HW6T&
zzxD=H?-fF)8^>C=sP8X@Oh(B|FzmGM*cg}c#5-}iu`byUTZ5E5?Kwj!t#di;<4-3D
zg7uRAw2j|oH3}Pwwwi0rh!I&)g4jvWtmk&c&KSqAiG}(K!1OzxbjOau#s)a*e{{ie
zIT|>wegdx63vAkVG9ak>bp<%bjP_s(WBw@oIXDLTKw6~K#vp}EGw=|c^b=&LW8STD
zjNMkrA7An<w<r9TYy8Z{aRL{LIMI<53zJdgNBb)`?RrMJ<-OZ0Eq7+rH7Pg=<s12Y
zXmpZ=A=8LuH-2*)^D#}slXg5r(~6IudAvUEO0@C!`8345@48CdB0oWbWDCc)%=dwA
z>m_Y_zWDyo4q~8(xk9A8|L-Y#85%WhPyct{V2MN}*ZENb4V6z9y-6^bxA%YJ1`6&U
zNh)mMnJ8_8uiielGAS)+Exc4&$1^Xtc=#9IU@^H$HDw+S-3~Fsk3aZNo$Ozlb-{|F
zANI`;m+=1lNMQSVtzW4Ufci~;DVta6Vnmg801Oo<$|DVrfO3P1l`^Zi%q?;#os4~Y
zqRDTW(86G=4*US|8F#@n#tIbIC9K1X4sg0Hw-Ivb)n(H+Qp8Qm>+H-qC08DjOzi5f
zYsh(S8>&rx2YyH3TNH1|lHw+k6Bj)wEBX6L5lETYD+on?U@$B9p8g26zOAn~r!5f5
zDAmearrfe%W5vfli7`r1@f0l8gUjPgTdaYKp0dH|Nmpa4E$xaRDguni-AV<s=%Iec
z_Sv3~!BlO()8A(g7HhpeT5sb0eVP2c2toe{%8+v>S+S{{kdqP|iR3BVMbJ3qOU^uZ
zmLX>BVC|4YC;Rbp>LGW-@r9s`!C?w2g{(DD<(fwxA_0pkw(E+PKI}b-8LQZKBRRs*
zj4*c()hL-1troinq(D7wruhWEwVz!vh98YcS+`||)McP-k@IF@J<2|Wqlh0j0$6H+
z6;xEq?R$ouBvvI>mR=<-(mW+Tfr6MY8X8AFTu@^8*wVz_Ab&S)49E@p3-7I`%ydeK
z)4sgp+V%Pf?uEyf%c9a;%vmc9IUii7F3l?F0JOH=Q-D0w?(uj!K9kDqnD`m6q}!WR
zty;c<{fg(+kU~n6vMJZ!RYsC*_<bS36>17-hA!vRL!etIX~q$osnTl2qFzYRtR0Vb
zk)YxjJ!tAc6NUyQmVcAAY*YEUPdK{L8v}r?%J~%X&M5)l-<_V+EK`8<?CQ~lMA8xA
zI!UxThKl_`_y4aO-~Sm^pO~$8EVp>sdh^w|rA9_7QTBMKtO<YS-CMj*A=W}8H(cnb
zBuorLC4N8yuy81;I8fvhfihjWsWFodSYY(C&Y6m@veihswlj8u>b+OT%e@~pWw!OI
zNCJW?NTnN%Yl3l&+E0kX9$m!KGOtoK!+P%c(E^R1OE7W+epqVIBVx3C2qWA85;xh7
zLkKm~Tz;Oo9?a8CX5D-&hfUp6i>RU58f-Cl6f-MR*pGDn`JrFv^~!_nk;UQHSr4|T
zcB38{_T^3%*7UisEz1Qd<jOAa8#WAf#^wqIasP<D>Ph0)H5>sTXT5#XNAM@qV#G`;
zR&39xRwUHk6M=psDn?h~Gx~xFZnG4TO4+jleQN;92MJmL9c43Nv=g_XRU-N%U~@7-
zK;&APmBD|i>Ec4_nU%|I@j!Hcv44N4W&EJCT)XY$L77Tj8KtS3QqEx>Mz}UjL&|Y;
z1RLSdnVF*GH}Ob$Zm#~b_Ow$kk6PMS2R})4WxMt1L^|w{nacxFjsBI9o$@iU*o)S2
zcipEyz`f(!VVgW7x?;hgxPu*u7Sn+%KnC79tttoO1e+d<q(0vB;{eqx8vc2eYiWOj
zI9{9UMum!d?Kw*BJHU(m?_BKfqmNYUiflM|v?)svt)m)~<yl(P_9)-{g{iMl1pB|q
zsQ-PJ!j0-hs0mC>;zsol&AXd-YR6C4iEHCA_;<c}iFjz#h-}QJ<Kz9$*ST{V`ZsvP
zAs%~}&8DRByb~Y9BY(aLUtB=4RA33yo!R;_WlnfbUy{Za(N9|F3Uxg_yLO~Pb$=zE
z743h@f%kAqLkh3<hHBC89u}VE2mV9kC@?nE@#^dSihE$}th_1L%<RNd(j%hdd$7Fj
z<gdJM92T4KOI+v5-X!&`DEzUK?gqY?#2ikHzhUI!E=RPsNBK`@5w6If7$gdi$5!D=
zx1FN18dleyIe+1WdCux$>TWTQk0WEp+1uHds?YD}p1(;!Xc!M2lKt4d$$mU*2<}7U
z_0p7F83t(jjv4)h7xazd_(H5m3A?K0iAVeAf*L?-)-EteE$q}at(3f@l2c>h%p^(R
z)!p1yD<MAO=C+1CBA`Ej+^x1r8>qilJ_<^HJ~Nx5T6q|K(eY*)N4mgJtc~x{{&@jD
z+Olty{XzaHNRkx{sd~gsmObl=PPELl)|vWQ*yBLG*kpHo5s0;9*yrcC>^P7;@(H}u
z>zLI={SjIJNE7sXR<iz8pM?EUy)AN})or$}6L2;34wx)o{W+<R5J|?ABA=mDzp9$g
zsR&ijbfW8HnzK!-TRSi=ucvu+VDe=*{BjtLW{h;N&!0p5F@6=|*o-}@W_9L5uzpW{
zuH0P6wp<`?aVm{aiqziHwsy$U;Crs@6`y!g<*PF7X{1HUkNxWRdd;&o#+M8>&eLu_
zFE3>*0;q<tC0z>5b=tqD{1(JdRz%^6nVd~wCx+klmCw2bb*h%I#l3F;7Y&z+si$ji
zoup7I0V{>Q$XSkCIu2SL#j?_3?rU(K8Iw#L3R_B?N20$B1XF<)8Y?EZjRBKII*qVq
zhjwQh7v$_IzeXtpgeCgqEYWE<zF4W8x^3(mCv->u!7|svh20eWd~;iGep_kwv*H>$
zOEc~qN;uz{nWOOr^56wou^>yQGWOgqPb2)c*=>^Pt@QACcrMN1C!Fh-r+Nv9Yi!y!
zfg?=`oUDr&#kSgA^*kF#ct0b;2#Z8OKXpyp9+|e7R#GjMOjwQ_9N)9&gq_h&pIka;
zda|-by>BTHWbMqAja+eB(X<m|7!4xIW|hz?YQc_D4Un@y_E*{gg-aO;z49;l1Ck=D
zf8pHM68bHAGx?UGCOns+<xKOfJSuxX)Jqggv;o-Yg%ICq{ZtN}?fZAu;JN@*Qg~^K
zzOz5sBEot&D~Z-XIgH^`L{4gxZG(6H26od#K&^06e)tc0QxdzWvBSL8)Khgt0E>JS
z$ey2>uHsmJmX(y|hh0h7Dvc9hbua6sHTZ_wyx#4Yv5cTUBhF(4qml3&F?TsUaJm-Q
zK>eGW&z{CS=l;P@KEwJ~MQRz&6C9^>^Sd!Rsjj(}A0`;39JFj|XM9H9``#z&<P=b`
z&J0M?pDK?=*=mqT%(c7D`mbse)O|7kIo!8?zrTqKCIQ!O;D&=+xhkyB)$AKxbtd#m
zb&5coI!FL(X56t|XvJ1yfz~J4=svY4J;JNvgf4HN;xxoJ$4|zG{Hfr$o<^ie;Mb6c
z{35(Wx9id6lUZ*bg<yEJbKW~!9Gww1K_qtxY-^^kuvc_78gZH6jDu@$xlK%YnZ6mS
zb8Ja5&f=P`vlD&Z*idAavZim5BT_cd=&)fAYwNqeu^4814F@MS7+81qkD9mA17t!Z
z2fDe~SOuN+m<*pnSh|ZR3*ri0;|F|S$A8bM=7t$|vCtGz)TJsZ=t1LrZ>n1ct9s}|
zX6!Qhy`~#0`zA8JezFzsw3K)86f!R1`CKwnq9t*%eBV6nTkamao*k=7q5F6m;gip!
z8$F(03E%7OF9ZxI*C)egyQoV_&8pyEV<f8ycwC|Gy!Ae5;?L^%V`{%r9&XSb6Mx58
z$Am@HXgu8*rTjR&|DbfLAa)O%nnT;*Vklgk80MHeo?Ug@>M*knTWh9Gc;ooUqxu4q
z_0@1Md;5vA^W69S`#F5moT;|ypCs9qg@*be9R$1HDs9()kWh|FI3t`3TFkZ@KuLkJ
z_@P%YqgA9^0_-T#Wr}jrE~>Grv0n9XIb)%BPp<QUPSo8s+ig8RB_VZvY=I%&RSMyo
zZ8=CI0H)p>mBt(*0Yee*4Rqk*`E$|~nVI({VRe&%DYQ{s@2s+!40N?U1B^@K=$`Su
zusKt#`W59ReH<4lETs2(&);cj63o{goWNO!)&y`<?UaH03g-27wpFKN23&o(R3CXD
z-j*7g+jQo<6Z=h}7`MgCS65XA852M*bVJYCln}9^hc!7Yg5?A@!Zhy&%4B$3sW87v
z44bSeh_EEDz#UwTM@#4l6?^jqb0ZBctI4%Thc&=P@DjLyqj*HQ!C~^{8&QBtH?wpB
zyCR~+!xnvtStV{e8@Iem%2!=w$x9RY3z>-dN@o6^Du}74K3m=iEmX6kQKo5kUiJVB
zSP9U~8NpH#Zc$Q}BoI{TT;9tQ4GujTuex9xA9ywSwO()@&A+(2b2OAM|9VXDGNq&c
zV|R&HKP!{&sgDWO-s?96x)F}!tBRbkWw2|SVH|KZy9JW8g@`z{A}i*IFQ<wicjj_^
zd}DhxJ*L*<`qUx<P*Lp^Q{S3w=o@&;)m>c{w(++4fGa~MTP<5Z+1!+vw(453NatA`
znU0kX#B>V&J59sd3S<<=X3$94llCs(*YZqxaUx{to_XU-ffCmN*+`*I6_CGHM!4>*
zW`t2UnEBZ8XtUl-{>OrYrYSEKQT!GI8%8--V)^8f@7b-9ygj#K-E`POD%Tfv@20Lb
z72}&P8_C8nVGZ;`CFG8){jw1z{JUjLN_l6|?woaZ^(yFafnw`C38ChfngeC+$2sdv
zPCQP~g|7anPG;N~9eX$%J*MomRvu=UmgIp+e)1QdpMZO>J^4npK$=or;%WffMGn#3
zijCJImCmiHhG38pcl6+IF9lNI9TlpG!y?BbbNvPuom|i?d}MUCuNsig(pSqBo3b6(
z)A)wz!g8wJPxvXib@9}^8>i$q+YtO~$FQqOzk4S0nIRQDurg9`=0($Ykd2ZxWbAei
zqQg;3+zV9{m`{gKZf}R(hbPc^%9M#W*{}^1FNSTmYPtZ!7Tb+I8ap8646e<VAykKd
z;XU0052%}SKdDe&@|<gLWP_a`k~DI)xe*z6JTgQDn4?iH#ikuKj2+tG=S5+unMS%*
zCa=n$ytq&BvifV{mM^jV=<N69F6GKk6h+hIF{hza>x}Q~!`{Pn^+CB@dqL93Qc>h4
zMWbtyv#dpmR*VtCqA$0z7-X%8BPJg!`)G_0@GneJ5)A$!Q$4dMjzCA3%Bcn_WGJ^I
z+aCyWJ|XAZqcwxY|5=WgmkoQ!PUh9Sz_e%~`(eE9^b$v1e|H>1xQcLnu(Y+EGezPw
z0mc@>3hFgvw=H`vsS7B91x*ACl|`=YP2Vi&tZs#c-k2w+@4Q2uDcji^s$nAB^+R-~
z<Lzxh4cn(Qn^B}i{kdKCQL3_RWX>a^YNm+|V&%YT*tXIU7e*@I*cWf_Y^s$|AK%qz
z&DhwhiZhxfrnb`L4zO6L>I!(>GE+pBBc-kCW=|_;KUrf(+su%sk*GDN<7QN#ZOM0B
zY4Y)Lp^_FQef548#J@^ji$xY=S2=ZWD&{<sbLrNE5@5~!EXnPI&ti#Klev+eoSbG6
zSs^r6rdVh|61l<|rya!fJoE^5oLcowMfRJ+^&ZOiPl9JuekS@kn$l0?vTjH`g{k4t
zO@L~h?n8-99twP1?W9gjO=Ux<o}exRm>0G!2ImiWB;`v_FmmQe3&Rq52JU5yjh5-n
zA<A|$7#)(LW{!%hf@gZzOsZ*u4NUq8%*wFs&+J5Jy{s4efjo^~mF7m7=WuPnRZvy9
zE<tc-J#1YPHOAf<#^%IX(-K5e0~-<^oO8H(x<%Di^36`R*UpBwS}0&8Lxy5=3O)R!
zEl<a3ey0n;toLKUV|${T#*KW4-$A}WUW?C(hNH?v@a24TyhCu3mMg1&)px&tntthE
zn=tb9uA$bN7fC6XEPldy>K4{4mc07dMP}@WdPdL$9A0xPH_l03vVSyJzlN_58w+o(
zl@jiHD>swz7aps~_O<pI_6~6Jy!<EC+Y#v@=`jQbvvpU@5HVLNc7A#Uz|a1t2mdcC
zNH_^^;-jEr<?5c+;E*RL7pK+y#JYNGwZKs8^12Yxf(x<QPh8Ut#`)t#C*s`gJdGx`
zp1S0nxNLns`1?qx{=I3PRytn+nL?E)Qm4hlWb`Mv6Fq-{e7?ZEIMvP3kY_%|-h@J}
zQ4aaWc|NrGTtgfINZh@NdEmf0HV$W7%KWrxQJacjsM!I3qZQ;CP||8|z0#$9d?Gcj
zh*IZdUKFYXHo#3to=B`~iHo3hUD=e@=8$N#`utk;X9il!GIUp}q>dJJw34BtikdA!
zx3idcMleGS33N{a`=qe_yanb#bT(tLaiqXPH5n_|sU-hkMSniokqQ?!-?e${m6QqW
z_7`nvgSkDjRgCQzuJ@<i>nl}Jxha*`7r{u6&`$7grK_K%Gu2eJKNNkI1;x~VDwU1d
zBncND>LO>^woY*A+A+I$gPgjNctYCJx-wVvm$)c7lx3OzkiFMn?vh=Mi*3|QW1zv8
zipo3N4Se0Hp_9j~_H)i*N_J_3x*#;XFO4>9SSzDx$7?ctjMSb(VEWa>?qw!BZ*W>V
z)~>7Jc{0i&ME|xut*B1yFT5qf?3Gp8c(mizmxm!ed{St8@U&mK)}iEpc&eRHzW%BO
zdf}_d6GWD0QuLI%;TiYLGmL44z44@lMnxV<Pa!Pv1XP?fqXxJ6T`oasrL7_OR<y5Z
zOEs{JSAt9`im~UGStStb5TZB#>4oJPD!t6&9C%`jTDRD1;wO!)@4J=XKo)COg7VwX
z1cEe@NE@9CRw-2rnq4Yj;ubY)9W7Dtd0A!Xo9|8-R}E$`#z=q4a$%S$zXRdswqx$b
zN<(a{yvKc&V7Ai5Ve*XKMfg5S;Q{Y`Rh{LJ69y>`%|br3$o||~Msib;an|ZbVX}71
z=6q~|Y)Y&hf@FP&76V&E=D<9I+y!)Ft*x8E89R-tDgXTK%wlmJ$UySJf^@{FWTkfK
zkyqUF25SLW;hvEh)^WYvc|}wQ*+Hi_Rhz3EqvV_)cG~x&G#~YIdEskbLV5mTSatcF
zdgro!KzV}e>(l9H>>kTcXg`^=(Sdfn4T_c{D4*>aRgjLpjdKS5{0$erlcAXk8{_h+
zzZJD3^5M?`YRTz%&|*tkK052F-B4p~pk53^m4D$0T_nf2>qdys#7uM;y~iXq4wr~F
zFBhdzhUrI@E$4Oq9LJ^Hu4;7)wbw?s1(J*Jt(565cZNvHl=XY61Uwj5c=d$F3~_*N
zT(#aTK=ir+|H51FEGjyZ;pPTV{M_fQz4Cs1dUC?~^m(+m;$vRB_}y7QV!!{=$^Xd@
zeAfP^Y@$V?-fPg=k*6Q^NY72n{9K#bEi}1*1$B#~rHrje#Y(n1gsIlpwn#V@lQ&j^
z^=~NZy#7d&i{FY?dF#;d|KH-k@bI9*k|?1g{`56i>s>CIP!nRsQy!K^mtj=cmvuug
zJlaxs%16w7dXA004d8d1-Y4Q5)hAMs_-oz*Vd`>yY+7^X79sEZ<+=s)SO&OchHY7z
z#=-<rcwjRpV4VQg+@2^d<(k)oeU7hw&2~4J2t;U$r|EzB<fiFhs8?l+77`-*KnYxk
z&}i=mSS$pl$%pP*tgCg({6Kjo=Ch)H3>QGnD+MpfvDB=_l|mnDPO)wD+6VBeE%P<w
z>FHg+dMBpU470`4O2;LMQf;I)={EFIM@+Of_-SGEM7~Uzdf=&+DvcQHbm9f?cx{|O
zyILMG3(2r^4}b?HTRWv<yzuHTJQo}|=N8KQdnXCd9`9gGb63%J81B{1+;>QoYi<VS
z;#Z=|mrJ%uY>2aglQEB9MbtY3JCGK2O9%iO+Qv4qaD<w&yQF!1kRTNCY47f^^I@!N
zn*JKNA#-rItFtDeCFS*2Ek?AxP@j3mNng6CsybP^z=#q}&eYSuTnGB>VlAk2p`uy{
zX!=EGRyVs}jkAk=)Ascq+;ZHIDR|z_KGLFlw##sQO^5d`)yi^`DMpq%KX;`04N4vs
zTome)>Ye;JXQ0W<aI{Fh&N-iv!+rUOxeA(mY3XoiktE)>vlSPdnYm6}Ueob@dB%OQ
zmDFTiW4>@HgDmGNP9lo$mmysp{|Hf1MG*J7cqDE@xqK`_D95!vPIAB{gB})~Qj<m4
zUBZpUh3)2v2ZH`d`~9bj|8r9LFT9B-rhnm09UYuj-s{Jod3%p$Q{MZ8S%Rn@zfcWp
z**=r7M5$I;zG-KeVu#iap%7isOI%G47yLpGoVu}Kl`d9sYL8-cL88c|bodqohB<<B
zvt*EW%j&$n2Wp=N3I)~$<~@*B;#5|QPmH4ql$C1>4Xt4U-ZOkaN$MJqd#zw$j!cLb
z@6td15?sE#XAGE?$S~@dl?-^ERWz%k>{1Y6&q^a_fG`Zx=PZt_CpRb?H45Ky33`(l
zU)wohSdffh@Nu4-&+gg}m<z@5A3}J(r$BdxQASwUtW;E3-%nFwLehC7<^8g`xI)rg
zxx(Hu#g51FzCZ8<5C#^sPilUDVkKAz6Zrbz)O&Xndb;@dPZk4VPbWbxokvBp2jifS
z{8`6hzZ_<Yz9Hf)|K7ThPF+UKlau!1-!cM4w#A{-tW2Lu+Ff<nMV;+l32KB}gXYu^
z@#`cA5CkjPBYU)v;8URnW%wlLKCPSDw5~|xK<E;@7>5rkRzIW;VMrcbTnBigM!lO_
zW&>Oyq<kCgrHE4o!P75i8*6)O(@dAUIvWm|bvAZ&{iI5I)h}NCIiI>cXQX!Ryj7|c
z-aj-fpCyqL`uk$D;GBk6nM&jIKflye+A4n{&AOX8jDxNH|Bx5{58Y8Xu^5rbt902G
zx8=IMgS!P6&e)yH+eqPjo6J2axm&-n;AZe9;JEHK+wbp6BLkm8olYLqEwsLR{Wvlv
zM$&TQo{kR6qT}G6GxcY7nJM*_A$Sk#=+s>I?mT{q<8aL}cJ-dzAoeSfTfqAPB(hu}
zh3wtwu+y5nVFM(R)RkI7<cZ=#OS9&=S>fH5xfVIK!l|we)20_jzP0@xTBqy#)I{g6
zOMjkilU}~j{uVS-kOu4WBv7<|xRg}xX1RBdeeam}?b4)I4Q+V6Cjt3IvP5w^47Jlk
z)1GQ|{q$r^lJn}<?q0LIk4!c+%0$e$wLtLr0AMI+HY7-#lGWBtXLvHEM(mon>m$}2
zl%0)*bwOE;$t6cW@hPP)f8x`$Fa&)r!`#bJhZoYo{kJB{HSfAT|5XttLK;~j5)Y#P
zeZBjLmFkd&3tfl<ujCE&$wU=easl<xsW<Q7S^h;Sx`rv1zW&dm)Tv$%f09+t&3I90
zkj$d9m!-8Yupm_{@i@j9vX8V#_u6UHj>36Cn5LOBplk~_97NLO9f^dfg+vY9N50i3
zJE_(dPHRNm4?Z>@URqWglWxizH&A1P^VRGq2c|(UMnUo6{Yqux%Q_A<RV|XwLsQP-
zGiNO;mJP;4EpJ)tY{jA>@6@@bbu5<|Y`K742{ejA6SOkUJR3?{lR>n4qb~1q2k08%
z*}seOj{0u2x01yc;&O+kmKG*&qEF39m{J6%lVOoQ#Y+@73PC|m1|=&N?30JU9n-_&
zTH5})e4t7ki_8zqp+OietmB(iZoBr8v18&5{Bdj|zw2CBrvBuDIf{E9)e}r~w2%R4
zzP90_yfN$b>Gl_UQzb>aRf2)3vzQ1$Ur{=lUPLG+k0|-l^EE3Xb0)XrFoq^jiNPGS
zP%R&1f7H}4sUg-RmMbq|-{wmnwl&lGF^9Vxjr`!A!QO1N;hUZrPREUSqr=eiPp-|h
zg+Ynp?0Fyi!^Incy2enl2nRj*+srwu!Z_0&sLKV{H!bFx(OKD~JYGs&YSX;Zq(4?u
zd<VBJUuA4tAI=T^8Ckc1$$N^--}UHqp9sFW*jn4zpjFE>reTwnY+hXLlvRtMA+>wR
zk^rN}aguvNBXpIzY~?+a7v;+Gt0gpI+_MKN@bK!kXk&43w3O`7)@qFx!cwpA$~oKz
zA!L37eu|m#^@eiAXmnFo(olvvaY^!|c1p(uz8{O!>w3vg(@niX!j|>EWmmK$|I<lS
z{lL2h6u-<v13Bc+ytAF{#rcwPSx=g+g$4AX6$bTvc>-cmX5{8X0Q*C$mgS_<XQ9R`
z#6zQ9f+6BV&~t5B0SYxIB=;+np~$eU#m}5k2GA>;M_?s_LKU-dWsiF;@<lVV<QBJK
z4&Q~AQG@L0XLJ4{5)sojS^+v0mF$mkuN5O{8W5u3(^e;l%S@6$Do3BT8XfO~s^urR
zOyo7N8>E?^n3yzElw#KSr)?s+oxd|eavqHic+bmw@oM2aJjI&6+h%lObK(t>fbRWy
z*c@fO;uh#gDGUKc$b&{v1yJahPdX)mc2iQS$7plH*|QvjDXQleJi6D9b~%aXge=sL
zzS_*Hp{&+fY7KJc>RAbXjtwZW^w>@2)~u+=^vD!~iMf?kBZGgTTL}PSjTFw#kVa%@
zXP(s$s>bOdUyg~k2%{AV=tg%~O5_{9Ug4sHeaw^}ui#RVK$1CxZ9i3Ua!}GZRJvwD
ziQN_{ov0-FQ>!LJu_0A^`fX<~dCTnX_0?VLtMz*yB|kd{k>FR34hQ&t)Z_R_U+Q4&
zTKn!Y?pfj3Yz~O+bsb!LRW3^-OQS$843syX-DheinnNbgQ_eUUZ_0%pK-z0OKEJDN
zXmz0f<vvWMsy4Eb{`fjvgwi>!Ts<tv?RL&-rkj_;WtCq#CcNG^@m*=V%edwbZ6`vK
z4LQg7DdcH42$4&5w379;Z*h+#z`vGVl%^SZY?3H`ipnv5k4o%|dc;)8m!a}wte6tE
zK3S$>QR!f0`MEp7O1FqYw6tcs@SL9EIpxG+pGLGo9FyoqFATYnHI}(^-TZ}6fq&?x
zcge*}UM1KrF^zACg{MC>8OP%Yqmv%*sV`-VA_~j81CU1^5gf63{t$;soZ(o3mp>KA
z=IE0{lJ|%y#OWUF_F!mg*R4aNrcc8P><lFzDt4yL&W)wO&Sk$zh<Gzd%oe9|IT2tl
z)*|&ROdP$)c^jXu1BsJ@Ty@*MXJcv-%ajXYr*AIz<%4I&1H~krQ|%uU=+!Wy<KNq<
zNb6Vwsq&^8huWdP3eLdCNr)jko$WJzv-yLzs0<Hgi%r~Z#e!*0PnVmrQs~$sN^nzK
z{V{iZg!6T)X5h9GB_}Dz8F4e25*LMX#=A*+$4g<R(mE%@#$>@EUylG%hNpU(Ur?~L
z1Sb75=ekH8{th=GL8bf|3eF3D+oeZ54=H9k;aWxGg@?PTk#$_Qy5x-CQHQW+h7MH9
zVE%XUAh}i;#A5*`;NLQm)WTuMxI|E&v;*uGCb^k<QjuhS>%1+r9f}%#(5M|kyxDMh
zNQFJSV2E41<oMI(<7eTy_x}BXNrPJZ*74@6CpN7OuF#HW9S0&&?9897_F$LG6Znu@
zuY4%;+>y2C<vY-3!Q*3hnF>fo<$@T?iq)6Siuva?`=puM(+!*iNmRjLTDSTa`ibVD
z1u6TpBoP;{;*(x|iZ-IKP#D(TpU#_AGgKaK>L}+Jb-*BePnRFd?EKPtcaSfX`paMs
zXfAmRQIvQvsg^-H@mZncgn52xqB}Nwo<&Yab^NKk!bu#gMa|<I;GE0to7~^g!aw{F
zFxC~qs`$p{+?S1;u67iICf4>Y6(75F0Ko5Fm`Jvi`uM(O!=AGZy3Tp6H@-rp^nqNK
z4VznDKPayB8gYmBq0MWSOT|?i-LE|!KWt3%HPJb0WsWp2%l)7`K&svR=Fg_nrQ>T(
z=~2rLzcfmRO1CS>nEmhJf9>7!Axp~5qMVNUZEv#%^W*+3`H<r*GMpFBF+M3z*n;m4
zQf<A$OZWT*Ml5a>z*9f7Y<8JU=Tp<v7i$oTR2{{tv#OO6F-fd)a!E&g>pA7StCV!T
ze}Lt;x~`-DdDClloA+5_R3#qaX`?)&uUOksOFPx7tOe-g4xg+NEcinfInhDjg<SR)
zd*%cos_OpsEGz2VhcCt9{r4|+eMsy`siz=SvS=-9Ql)y%$fA7cfDq5jsWVP|Njmpp
zB@nido+OUd#CZ2KR$-$DzdHF=tsGtE)j!kv_4plzutqu1eIeLABG7;qW!KAAkS^q`
zmZRRXk130L;vKHMO{$(JQ_rm{&pZ$U6y0MV?8;S3qaQ<LD+VUG=f%0g_4kdN!m6`e
zhk}mKt=4<Iw;BD+{gU^0AE0{)U#B2T)>~RNTOB7G9CV9ab8V!3Vw8B?0|?YeLwT*q
z!{tQbTBdU90&uF{giL<Q$9Vvn6rN;R70Zo&M02tfIgjttIP!*vIa?hg-_!2h1QvgK
zs<pl?`NtR>D2E;@Q*b>>QbN2E_ggFKYT`(2giX{}jRfm($Iq5_%3jIY$MN;&3G=0a
zf~%FJNKfL(dJ-{F^@-xB$d)m6#&un+>mP&DPn7OK6@QvP7)|i#q8CzaHULurqYN2T
z9rfIlt21*<llFp2{@xHEux5ghh~Gk7>^BCPW*4Aru9L}jK9%HJwxHgKNR?l(jJjI+
zvP;KtJof!Yvpe0U7wGgst9W|KpJ#l7ff-XeI>XhjH{i?KYW&mC4s$q-C<0*;xOn;m
zyOp9eMAL=MYC*vbsy}5?Tc5=CMi8m6Jrj$8SXQSbiw~)_a_s5viEFjkK8lKS>!{^A
z+K$#=GJ5*dVd#fmYGJM9V96UzQ!azg2*#!>s^zC<YR&ECNiFiA+?~5UIrcvuH5_v@
z*1)!-<+(Hwv&@S>CBuK)aVkeFrh?*?r~x9|0^`<eSs8yIz6P^<lbTicVLMT5rwRT7
zn2Eyr0J4ZDYX$RePP6-{n77W^yArQQ=h)(?NE!PQ^CU+2mNT`oqpVoSWaU(DuXkOR
z)5l`<W>5|8uP83OJq&lZGV?8r7y3BkbK%C`ekj~9$4u*!7V}Yn$Wux(blJpAGZWsw
zBrqT_6-$J%a^l2F36B!EuBP!LZ>4M;A<2rhd@lu@$-tX(H-YzXZ6Zl!73}zu%LPtl
z;G0{{7Bu@+GWAN#4|;k@V~O?!e>CQxw&2DIEPsRbgw=ZOZblpvsCNFto6>)`kY0A?
zUP}z)<)jH<HqV0R_L+jKpP=pmCzWMV`YPg@Vx4K$fZFq*Wmdh#*|w?q3Qs&MxWv<(
zgi}_b<8b+8xbf?T?=SDx*P)2n!i89&MHG+F)}?#fg=%j6xFLJ{I-Qc$FJf>_4NDE+
zK@Ahr8r%{O$$IJiD_<cv5#zGDGgdzbcoq^m{TqW?1T*nOvZa^mcs~fcS4;Zhj-<rT
zlkB+#5?s@7tIDhMtYsRio3{feQ@oM=R+HJh>bWg4=VNAErE8+LII_T}tD|<%e{@J1
z4@y$5-I;#n3?S~W)?334g}eB$7skE#NJC1qWnYjob2jj*wM<TCgMp%sbO!2E(kN&z
zZtJ_{@AoX(Zbsyj+bq@G*ZDP21bqp(NTxm`XNo=sfD}kI6iZDoNC4DhLGPV9<5+Fj
zwH;iLPX}~J?r&Jyrh?Ryqdd+JX~SvLC^tQd3s~dL&tmIzX>w->X}S_;;%fkB^Nyot
zMvQPjE=*$#N!@N^x9?Ec>1^eNh?itlc?u8zfKimQ#kbwgS+>HyVsPzqjjk>S(T@Wu
zN?8hyMHPcXa241unL;H2c`h7fb1$l08S#)k%6z(@!45$Z<LU}3Vp}r-wiu*bJrDfv
zF8M!LcPHt#sI^oow2j^f6?3a)@#^;LE0UXzr-W8`PrY%x+1R`K#G#xEGhPUTtJG)Q
zk;f4F!bf}x(busd*4BYZnfd$S<bCcl%(sf^3v(BcHn}-_&ei5yOtnz#o$slb{Avd?
zS8s|0Et(I3zb76EWQI6tDxyFuKd?WN4IEz+JoP3@+=_eN!yWUx-&%wwWh*a54cKx|
zVQzy4WdK6sl(boI$Lj6XPcP}>B7Omq85r=Uu75k(M>ofy!$m6*W5S>c!)%vAE0dKd
zVLIRS6N0Tc1hb4+j#GxQl1;r~CxM}}A5_p>0Z8~>DL>bpKBybOrFRsz;8F8MLdRQm
zKrG+F)U07&1zrR<dFguJ%)EJ_?3=yplD)A6=0+r2{L@Vg=8WxC7j~`m_^leV3H?`-
zO<xAJ{YXmy&VSHwsEOhCzolL99*u*cDOk)b-$V-*67pJ80Pl{M$fZh!$l7evyZ^6}
zFA`1SA=A(K_tjrI4PT3RdNuyhc%5P-vavQ)+Vb*mJ|+3a7$){_r>>`X86+`oQm>O$
z8+H!wY6B%Q-&O$)z1o0gMyo`X`tjeY0q(T5OdD+x>d(cIkxUo?yCheOV(;U<3Qj&;
zb~1!sKq@Q71V2|-7P*nMm!*oZApRt@wV6J;YHp;4d4J@add~&h-B=vD?$=sK?bGsd
z$0T4IY;Hq;ee4zi-wLw5ycc$3zmIjM#OOB-z5Ye~MVzHM!=Q9VDXEF-w`~DLC}`k|
z!Nq9A`euylf{qtuK8Mq8?*VfIft_tQMen1`YYt~&WlY%T@ogoqMmXTw{G^`G<?Q})
zdPdsQs!f-^Pa?l3wYJ_X@3PO;tAohB-?fWOHdc(8XdK`9!BzA-O-=z6HNuf8NG7NW
zIW<XY`OxtfUVvJ)rBI^~-)_vbTN=0IWObR!HQ8>&&#53g6WH2`hW1Oy@zw-+fpT4L
zf583qT;wml-d-p5AJ+OT3}vMb-`2KCf_LW78W|-GyN!cv&`Qv+S`Gux&QsGgS0fY&
z=`QV@eZ|BGUji_Azgb!Q5<25)C}Pj1Y<>6v$eVp0Up{^XCjP-2>~Daq$RgWK^9Wv@
zvZs5p^K)cH<k)G139kKPFywxa<PpX^v5|EhBWw(&wM}ZNWo5C@(xSFD*`Ep?wn<*>
z*jXDm8%!tc?@USC0R{zAQ!y=P6g$rdTucqbcykpi<c)C0T9ak^c9+kRa;}c2#Ep*e
zFT}8(AoCcby2i61fX2N{yc7DoMxtYFc2Tum2{e5SRTutxnU)~)(}O>U5Em7;o6K1X
zkfk1HEsLi8qerWTHNoh7qq->UyEt1zCG%AO@xm_&j5v`cxbcTemNReO5}CX=rf<lZ
zpRs{z^WQ|cq)gkMe#|%F``FhZl%U&DmxD~C&656Jk(wAFHO?rsB@~EqI)-@H-xV;h
zKzD{sb`7Mf^cp)<13cG1QSjs2FOS<=cPZ~$Fa0m_-U2F)<=Yz_NFWIW2m}cdNFW4<
z1eX9wu;4bhyAJM72oAv+G!T5S!3QTnf((-2?mD<Td7X3a`Tpnp?YsBAyWV==eZAI5
zRhR6&>$i7zbyw|O!k<#Zs-*HRL%KqzKBq}F)KUu{pOo;O4TsFfIt+FhOF>G`yp8>C
z$c!Ee$Cq6M3cnX%oSSTCyvsIW+7kphI7Q#8EgaVhg_=H}ayjH4YZxN5>EE|_lpb{I
zNm!RR?6G1vTB@aD@xszpCQ`gsFRW!mDx|@S@74<WyOX<+nEq`ltV*4a$sgL%x=gde
zy{F2MOi3PL#E<kp-B}Qnk{l_$TWjw!OBD7B`{QDOTCkWvXm|Oeg3<YJ?vk}T&<&?}
zUQvYJGh~JvH_uc_0)=P4O;4j6(nWhK5#DRsgncK$Ruj7pQe;x^uUVU6Gz~>p)~!dy
zYk3C*N@Ne`@+>7hy5Y{d7cxV$%Q*>^hI{C>bT-%t)s^S7CVh=BNp_b-mI`U^7dPs(
zZnCdt>n8R}9pkP=Nu?^ZTYYQ6s8VC8JC=HOEc~uWd4;#o`>ir_{lOBfKOz5a0lSUW
zTbWQPDVcNpf{=l3m>*5fn*&W@C&0moYcxEdF^+@gj6WtBs)yr@52*~{dNo;6R>aBk
z+?n1J%q_X}7~y~39866sjc=`$;=z(?`+zBWBQ#U8zT~2HgDBNNMWA_Bn)v3~IAZ(R
z783Q4v8bo$b&-HzN(tp-=P|sG!`r*qhg@LpGSfpd2EHfNsGK5?olb0_1{2TIWKr+j
zm&!0jy~<r?bNHZX*GFlvS@Gm!sJbXL`BbR<?e|r&WC-}2(e29JSkPm0rM-VDm8I+u
zpIcwi#bQ;9z_!>on+9v6#-uwE0gm$}Sk>d4j6<++JX}}aJYePHOqnTnst<K3wv4#=
z`Yo$9(pF_<M88S+Tk6SazP69{w_OhHUa~Iehm{JCU?t=?aoghRH7cf~$qp&gN6O`l
zxsV}77g?c7Mj4n8gUupNpS~({MI1}Vu%&7)!9`>7_=pBOouX3UKml0~bVJCCzjm{<
zIlq2B@#}TA#$~@waJpcf8!|YdkdjCy=UqhwGEn6%R~R_yWLvWS`1y~-R2M=mvhKFS
z(%h85KFS!yMw_X<i5N_TM$lAu0a^FR_DUJ)>Xdt;jCtQE((8vw3#+Q9+Ba51qKCK(
z8a3~WYOX875KULA<)xZAPO7CTxs(BDiUnZJ4|$C~xukd$#_a}D-lJM7ckURtsFk-p
zZ+O8z29^0XQ9RDu61-oI(NH>KxQMXHQTCo`Z>U(z`3VYl$yKU+8(+uikwYr?97jsZ
zsANRKUD7-ML2|VqbDit)@_o-UxhdYmz$!B|hauyYTlY84Qtb?Om}guC7i7}CHHg_^
zM=9xW+5Ev(1nU`G;i)Xdk-FR9<a1XHOZsXW?=`_NiUtV%+2%`e`8D2p#bk|7y(zk$
zo79AkVt}cxyqba6K&QI29X#xEnXt)DVLWCnkWwO_JI&l+t)>LS&lPeaJLU$5C`dVT
ztMHATLTcShc9OL|&cTyfo9;aDVw>`qiV0SsK$|%B0$=G|I9YoZ&Tuy}Y}YPHS)^PP
z#9J%Ytnc=PG)+W&o@TXSpOf;k47<cESgr9jJ~-@sygaT|LG{4JT2p-uaFEe%Wk%>|
zk`i7+I~DHSK+6F&SIGM6@>#(jGDru89R+U^2#%MGqK-8i2sJr*Xu(+bA6W(73uXK6
zECLblV%{?vb*Y%EiZWwa3RHM`d8=d{OnTzNIIR3GoZb5#YSP@5h*C!`RSCfu*^@n%
z`@Qgpu_4|j5Pv+KAnP)gIK)3g%VU$HUW>V;q#g*%)BC2*1*TW$B~I%3Jy*GG7x%^c
z3}ZoMrbbkhurhpvF|qm{-nWk(FOtKOcBe-$P3m_J?Wkb^?(0P+a#o|h^`-MpMus1e
z)7!?K7R29~zbvhiPMwrKFvWzZY`v<5oy8bWAEK7k1MSr^S&$wx#o!{sZ@Gi0J|WBE
z&iwlBWcIvzBj=h8g3I&cLDy8u)qD^=Up_LC(`0fLCY`NlAB?-BMXX|o_28&bpt^v`
z5=Wz8hfI-ax+jcUR$GiED{%@tYKYw?uAqkt-*(WIa9EF))OatETWUO^F~_8VEt#(`
zNKb!8jn)bmj_f4Zh`+-EcSYez$6jR1v|F>*#b9a&UKX(jD<g{VL}CwrWU6Y1C8jVI
zq+Xp&`CNyt<3(>1kE_(g@pnQS`;B}A3|5UeFr?eb)S7#1)Q4D#S-(c0wY3&0Gqy)?
z!)J)XWTfjou}w8uEgTFJObxR5v*c={juoxb(8^24DLyYXTf2Z4ebnt(NBy0axO@iM
z3RNh_ppFVG<CSY#73-UY;$m%fa`_4x+*)B5Qy#`=z0HYH*_w$tPaRnrn;tGyQ6eht
zJm(AQvFWzbV|3fJRNU5v<WBK6NX6JFm-gTwWg(X(T<<d<LLZ|WqZrP*d(FP#gxXS*
zrqqjFJ$%WtqjTwbD7xS0x^qY3ih?DHuWMqAg9pA^*ipk=JKp4}xS1&8#Qdh3&7C8w
z)x$SJkcGJTB93sG+?F%%2kYBs^urd)4D2s6O#-atbjpLd$$2@%pj#LSnweU7sHy;=
zwD?SdcD9zaifLY@T+*`1Xz9Hw)GNGIx1Hkj3~mA8JFdQfXMj6hH8?9YR1RX5w)YMW
zleMHWj18NL>}C%wM~}JJ5%9Y|^hcR6*b_HKw=E-;A`ao~rupiTTuTOG^$W34uQ4ec
zy({FcY%sSKzX@h*F_%Wvt59}^(Uk9CadW6-i@~%;uo~%q*u1m+JO+TM-|sSdJVplg
zru1yeqc-#bfhMVcQbeRigV8k7)c8jg75*Ok=P0NOF5(uOQ`-Gtu^i<s3&5?;%ctGN
z(_oYsH$i?+?QHOKCODt!DId<mwU-qV;@Bx$(q6AFrWW}F%GjQhYaLYT_Hvn7sJ@)$
z4G?KwQqXSObU^irSy3`pd4{n_t`!lvbuxOZMnaVH-#BLNOKE>oJ&4RHrG7ABqm7_#
zwrYA)QQ-Garh~s{wX2Sxk6hK9>Ws(x-sHSto&5<)Li?B`c`YeuYOLvgesROb7kaRN
zIXm}w?WN>GdsDL=xjXA@a|Q-4QbH~Bx=EjJR~CkZB0Jb;+BKb5dTXf=b~wp9+Lv`_
zyKA40s{`ru7(6{*TMsy_w&=06)EQ$VQf^TT>6@u=fK}ucd8EtyQc7=3&W{&)l)Uxm
z+A&6S`t>Njp**ptrjj12%=I?w`M#--L!ApfVJ(WglwR=-i>d$gg{<_(R=##d-s~GX
z@@h6D+g|WfC7PKpPu49d`6Brzw<YLK;IkR_KuKcPLyw1_T}U$T6>j$rq~9z#zp6Sp
zI7FZDO}5=wiIH1yy>$2UX_j7FZ85}o9K3<%`h57{oUT2xZK<<+?hrJ(+&C1BGu1sU
zH2XxW&=h98JFsA@#DiCiocfOGd`4Y(JgZWxW9m!NkI?`%g3JxlPqPl8-xQxXOn&l|
zLMV^x&Rl&78JR5AZ4$gYEpJwQWA}@srvk7ow8EC&7IYrviG4p&*F%$O=$W_pqNkwQ
zp+69*Y*n7E>{szG8du85?IRky+KYxJeZ7&ye)sJoI(s2Kn|WeiLb=`A>37)tQ%ct}
zUO?+^cw-%cK5cdl*G&yyd`js*ksMB}OFj#4=|NLeyHwqg_S!{(F+@c#cKH;e?Awcx
z@R7Xw(`!n7k_GGfL(~-ES4!Ul*<@?`N$G^GwuQ)WS@?p`V+6Ej7d!NFbT~^ko04*I
z+he;9obp&Vm$IA>pT06%(!l9V$qUkvX5Q`cmiRqvz>M-n`0a#so$G4ToXnNAnN7*F
zJSq1<)qUQ>B3ddAsWi<vm@{RNL_Q8)(EEo%*J;Y#)-5)4x`zf(4^v1*Tlj=Sq1u?`
zwGvC$4+|b@BSuFj@{|D0JRRaDIb%YW{8hl2i()2^didkf6#;1<15w3pX-O3%*T8N|
z9sexHYJ)E%@esT_ZII7KY&z4MV~|e5l&Ffk;!C)9(r4<-hFdPQ)<RD}tqD$v*hnA<
zQzC0-u3yoOkxS>7nKWif>5WnLo9fPQhAEPY5-%1OYM$di6qw?gY-Ha()AK(}GoRr5
z#FYB-T&IurFv7{J&(QLv8s6sFS+51575Z$er%!()eL|rL5B8#Jj<RN{`mm*d3U64r
zRL&37yCP1dEG>uBBEA<#q2;QhDrb-ja`Dn(nVLB>pDFib?#b=MV%@$PWy1xZI|>^-
zkN5G2A=2#|&f&3fmh9WaapLuvvJPD0RYx8Ut65^2_jvF(a7sqOmyw?ipT{v)tC{MY
zjHgdHS2`dE_a{DeG#yN>d(JdLdp%SKuK2!a$>UFdq3F;|k_aN?H`$d)>T^lAoE??s
zbneURhsqjya(UX6hOn=XIJboh<{L@##1UsAr6aDjw?gYU5JZ!qpS-&fFAx>EC7puT
z8;h_Leh%EFy#Q-r(n@U1iiGf}%#VURBtv$Z@m&|m1f)gXxpWV#gQ&DTo-1Vy7n|On
zrV%Z`t&#dzO84@y)f`J|<;>_ob<c2xY3f5$trjN%+R8gmct#abyqzD<!<$$xf!IHF
z{zawP9`ZF5E;?9~dBWfXon)zS6>UvVXODEtI<CodOD8ffDI;3S4B5Q2kbNSCYip!`
z!YlE?$0;Oo=KGY$z93b5c1rBkHC)Z?YRpQ71jeY>jY$ZlnstJlG1zxY_H`Z=Mw+l3
z9wFex=9`iq{EV`Yq;2*(g4;_i%doNJ;SW7~F(~H3jyThm)ug=cE{X7iWqo#C@4J$m
zDunzUNN)BSPZ4jt=U=5!!20|4H>BKe*Agl=L*uLySH@oOJ~Y-Ixj$ekzS1lH(vZlM
z%VohwMK1UGw+gawoR>yZiO;YCUT8*=3ge*SC%q|yu|Cr{X6Aj)dvG`@DxNzyrc2y|
zOk6q3{ccuq?pk;q!@e`_`XpLxZiSQ<>T#|kXah#We`vag!T5~y)crMdKCEf#)d_>{
zf^+hv$LoxuSl1cJUuFa&-5M&iD@XakhQt?<bZ2vA%w8aLSJ-A7<kL)_J_lIJfrxu%
zKSAtXLL$zBeneC?r@D_J`{bcp#~J+?l`W}1Fn=>met-WL<q5KEIByb~AB;mC*`b-p
zZ@cFnE}hT;jc};}fk);Hei`9kM*aFX^Y=w#u5aTKzg8z7jzGL%L#@nYbZ$vV%-$Z=
zThaBkd1gapVTsXv$w(*Itj%Bcj4JwGX=;pG9Toz+cwlM}eP=#&7IB8tx_fd6y7rxs
z7iKRRK0)KsX_|3e`PCC6HNeC1syILv(9wE0ibr`Iz5mAcP3@-vjjMU)-PNPFuF_>q
zw{MT=t?I6{lo}IYH(q6G@wsNX(%#V6KlxRH9JY1t^MY7sM-Tg{YKj9|Vd_rPxgV#l
zl+!^*jh~!n{Qy*LJ}sWEsQ9|+HMfo$y#AV3yIGQ&^_BcoG_ZB#c1veQ3WcP_wXE1k
zUhd2Y+*UZ^PR1CL$qlk<5@m`;YG6Yjb3>G4{}be(+SxHn;dAXy*d6^ZuhReO1a95h
z9Xy+9tsYaevcj){Sh-${rn>u3ysb%R@yx#G8T=cKb|>?Q^#`jX&wM$C47&@#G@MsJ
zAR>j3!QsoicW)!m{N*uiO>JyzEapc|ti2{CRQpm@+pSe<Y^L56Sv_D0--aF4a)Zfe
z-%Q|X5v-P@$;yVvK84wEU(Ve#{!x)QV~5p9=bbjuILQIEZcx_{X}dd~;iR5a?WB#6
z<(Azw)VP=T!&24CsY^M`<kWsDIybHw(@T9km>O^wQ^mVPAJu&5aD|xK?IOpX$3S1J
zwiCI>QZ$$}L$;l1&;b>%vtW^r1Y4$R=V=X}Ilp5WHmKc_><WJgO_nL>uWeg?Z=Vap
zqT#=<2pO3mOfMxhMA+R7H(T(!Mka~b5AQP5)Uvuc&}vlQODMGBM7|$vpv-Ge*2)=V
zDi+BX3R|J3l@eoj<_o6C88BFhdtwlGcMq@iqK0`>n1$I06I?7Kh=4gEiHY@5gL|KL
z8rBm7U&bpxN*nsVT_4%41N)j&@j)bkzg5=5sKB7R`Z*oLMS#OB-wJX8su3`Da9u`c
z?xN{3X%fsg_KkD6vO0+KpkVrJCYk(V<MR}B`}s;nFbPZH^}WsKiw%WiVsbI)+3Xp_
z6Q`iU2Sy5Nj{_z0+~Q(1!%gHgyp#h^j`gjjtC1!_gb+O>Q!Rs+s%|-aQ!1Z&=WC;4
zFb0>zR%%n155v=xgqLz<g}Fuc5~6OdI6|CBaZZt#pRpt*{Su#nk!HTyXmCo31+&28
zxKzzZXyXfUwL{5ZVlVI%w=VrGi8!e$YPiVd-dPC)K_Q~!M<AkiStKZ^FMftlAT|^E
zENQ1VC2#PJ@q#bQmu}X+;$TQ*fvI6?>UEL(yicP`6Js6L*B+O<(^me0FA9nBV(?}>
zC!yv_CECKDoC-qiIeR?8m-U&O{dva7WCW=+6J6kmNiF#^65L;bvj0oR-&}96YV%I>
zh5rQUO?3{p6X5e~Pqnm}vDVzG%6fmlnubLN_(&&Pjp{6ehFiHtsLbwsNcPUU03MOu
zfylI`I&X?^qy&*=vyroKm0)E_XME4%)NtniIz}c%86Q0A0DD&9?r7wveU@)u<4>6j
zyIUAq?Gk{m<(@+UiB8^griERLO-m#rvu59183x?D!T5GMN|*fDXKXW4JR_!R)~B2$
z{mlCO81aQX_I(}4Vrm6uLGuIgC&<2U1{F*dckWT~&+JF24_FxK^0fso9*|b<#kQx-
zh%?vk!w^_#@(|boL9#cwjaVFhn{<d%2*D<TRnG}?=wt*Fr^MzvB*hRM25Az{{l2lT
zx<IBss9!E068F5DC2`AbJ26B`K|3~<<!(s@!tYhqk1v-3o|vw=2d==T^*Z2{$2PFU
z{jIQKaNxO{TPZTnQw7Z?Hl4QF!{Q<+aCsKrCCk^HNG-#~vk`4A=nF_52M-aHL7zOs
zYvvpMxX~+clI?!OnBQ&9<Og9jVhZ5O;gGAAM@WJm*vuXV?X2c04xAS80#B~QY4>*X
zb)z5a!B{Ox>{QaTqlM4z?iai#ccJ~8Ei9kh?a`>Lb$vrRFy{YD2>3s!Fj}YZ7|C^1
zJ1E`LZ4;g!jL*`>q=kItFVHf(d#?nD9#?qxaC5uM3J6#VtW!mM71bEEhy&~P(f7kJ
zm1uh_#K7WyE@bcl+>c7;gH*Plusj>vyslmY4pO}9?}mlPy1ti&5^zxOeh&xyNl2UM
zs;QJc8}RpfjIATRx_}&S7wz*^86DLt1*T$FI0$!dVv~xurz|Xu&T{n|{ecwcl<Dbh
zpKynFWUps@=C^zU);(7AC`L{-YCCe>6VNAwhR#~q_8Uq<ZDVv|3!TSQMP4mM`28;b
z@8L1#zy6_pb^(DfQ}%?Ofj}=)KT$-Ol5?KsW;o4%`%yfw+?;Bw#bM6XAadJq@?ptl
zz0t9}6ScO=^U~)%qMGg8#M^+;T8ixYpcq%*I<Ay)D@V_uL12@5<79uERH=H=e!Ptr
zCQPl+itwge2Dz-!O$i>7CKjtihVZ6VWAka0)~IKWF|aPkae2$8RhY>YmQ6h<QcBO!
z%C)10;7Dp0IvA)J<SNAn#3Qn4yWnUE`nTS)ie-X87VsA!(2(w(c0bT9S$99sABVZl
z-~RuR_fLP@gNCb6tJ?CiP7UQ0jVTifK87Hp3R3e{ysBnjTh>gT<733<9(I&a#W5PY
z9>#(4$EriNH<un?QsGV9z}<iTj2!jQC~)U9&)Tuv9DaIsnx>K%47UYhe)-96tjwHD
zEUN<x9AAgVPC3jduLhQJ{v#Raxd#gWkjpr&76;5wW5=Z7@2{~=v8%p3ub6r!pW&}z
zw}?@rx}x6_q)CNGSJPa^&^i76?Ah^H1{L5^bWA9^F+d3~D7RCIoqcU%2VJkFenzXa
zEqy6=<8XFNmWz66d29v!3Msx>xXHm<vKtkp*D*4RS)JaVt&-Sy|7>#-^9_rIptRCZ
zgDn5`YwR1|owSRpBTtub@5a-QHOI^9%Hf*T92`r?qKx*Z+;mbcbzK%~WZws~S-0io
zZG3c_Ub~*dKdJ8Ri$ufV1yRb2-aZiiD)di$9c1yI(;l@a-cPWfzZcR5p}%qd)SO9&
z(!Gajd3KhtF4k3rZkcxp_wf5CGWrMk00HoZ^W_B1P`CK%eCt0IRi2>6qBS$(rlz_u
z>kl?LV&^n2J?+pLb(G(mEj28yYon(DuBI~Fsi^3vZ0T<?lfP)$GzjzEEIhq>Py-@;
z`A<6d-)7-(t^P-ufA13OpEvZkE8whF?bsoqcP$XI9lE1<Pnh&kXiM{q<EFlV3_a++
z7Q}KQ)p(m1DKc)tSmlDrxv-+sD1t0yYv-AS88p)k)-w=pNX6YB;v?j5_(3gR{G2RI
z#-tpxgcMIQ5Ak^ZWn^U=q(to=+>M;CrtPgQ)5Ks~PDj;tuSydRVH#O4gn;xp2Ig+g
z94?Ienz2F0ATJTdPWuv~O+X<%Jo&*g4*p{t8N1%{INiHaXTA^jdvP`Ra3Ch4)ugne
zah<zKnWAiYd&=e*R>@j!-ReD=zz)ZC%FZh3Tp2QcevcDIc>Io&d;cAz?)IlpdxUgW
zzR;f2Lw#{1v(bb|B8BAFH0zcZ0}oikH;JFk*)#Hjk4M;pi4&bdstSIrbl1}pZ;dfx
z?>oAWPUI81EmxT5yi<~G8)w;mHB}C^j!r!zd2bvsW|p_@R#ME1lzYgs*_@BQSrF5i
zR~Pz_r3#F1D;D|G+<Oy}Pf?G;sevR2-4?4;GNpE(U^`w)5~MLa>Gd-Gs+CH}9c*?{
zJhN6s!zD0+ot#htd$Gsg_9-jw&i!>Mi_Wa}v>==yCgve<Qq>7frMSpoJxMx0PT&_@
zuls0i-pzMVCyP4MZd7(Xj*HEn4!ae6%+OU_7D4ig3_-r?ye6+3)~_Tk152{0%{Q+P
zuG747$9XzgR~GMrB#HAg!)m18%R~%_tgUVxIot+|)?4o{tyST-Md`8NK?^vLE>n4P
z3Sf(~XJ%iJ!EnDGfn&z!=5-gQSeRb`Y+=0nEyN?J@t@=De?Ttp+Qs5eQ2Ci3k-4Tf
z0#{JSBp;8rZ}caq?UU9i%i$gxKV@!Sem<T8|At9Z?D@reJpW{uM}wvzz>1Dv2P<j*
z+okq!fBSha&ZU{mB~j-w<^)IWN|{}#!2A2iptK`up_@5jw?XPPZ>Wl#_A5Sg|4g~G
z>ai%j?v-Kb?dpmdU?~HmVg`=%+pV!Aq#{SfjN6nx5GG`GL!xs!zGo?mUVSl6h%E6K
zo$nOtN;6y^%*m9M&s;IEvQo(*d-H6wRJ*^&O%O@Po*QXJLo?TB!@zvRK5v;n<#l=H
zLSiR+!!?71m-+59d@2uk@*C0rU0;S;mXcl+jr8ZwciL%|+S1c+*ScIftjc+t-gjza
zU&kN&Omb_evf#`G-)>B4d+0TFeg@lw=XR&hVQe-2DZZ!Y1N@SaT|MZIePYE}hFX3v
z#?suu{xyO^tf2PfQ=<SEnsp4X*3b!W5j#6DhjvD5zJjhC@op+we6}CPsFYJXM6HxD
zkU=dx-hqZ2qz7MP=1%Iyg1A@*kD}3FF3=o59O<g2XCO<R+>Ix<l(yL<LI0v~qfQPe
z-&=5PdlnR{LFF0`gfIGAqx(QhARxBUzpDKAJ>B*LVMTjT+4~dWL%ciYS$m^F&#J~k
z@NFA%Fhzk7+_pBAps=E9mNBhpKYEZ{`pV{cv+#@G0!X15FI1ZS8l=1IEW8(<AmTrU
z3hiGI{`Y<VJ(vH%`>%*h<Qv<^Ao2~6+&>6uar@s(=TDtPgV5aKCC2jO?<P~zp`!@J
z)`p@KNZ|fmZ$lb`M?rPJ6{XOyqPEa>w-=^E=%p3|w+=B0GtmBrNd{pTR(mfkRtIW(
z`l<<y^%`*N_sZC<|KRcerQcj%OZcYz;)^?K4l_SNa+Nr9N!Xj0XB`K3Q`J~b4n#M8
zbEgM-%-b*b+95onSNI~$6mh$RStVOs8ueMll37@$v4m7UU#zdv)`R1_O}MU7{5!4L
z74SqPN4B)%l9ZWh@BRCddzR$_x9RY$k0$p&S!ZcNy7oiH9l>>syvfF2X=XV>(KAFF
z_PXPOMspHl)crNwrr1X2y!%vbK@OkxybiyHBORcrkcRWc4l4r9DNC5KL8bG3#z1gD
zdyYgREGf!Ok=XW0-|c)B_uIg85Rnfh-!auI9+po%x#aypT?CaOk5G?Ojwl$pvTEm#
z2)`sL3UQ6da@Z|-G9nVm=Udp9JE;Ic$v|aBkLZX#`?3F8G>Ao<zI-m&objS(8L<@W
zx@L0j77$7*Pysk2lNOGef-jQlZIyZ|x<e)#@QG}DDX#gbY@c>}!`jTB0JVwobuq<O
zmIf*_%<S4qQW#H7DnItW`ig1nkwLSy2JdC0&W*FQpRg@pAOoQ($&1hk+o1EYhJ}vm
zzuy{)QhUrfl*C59(!w-*iS|D~urSqX;X(G|Qxd~d`a7dKqGpHaHae)#<J%}>O_z0#
z47F_CflO)Yk~1CV)T`~z9kk8T{7$&;jwrB+%NQx3^=ag_60qJFU0T;<n0m@^7ek_t
z|3L8mReo=~?8lm67115pkh?R7_y@aNY{G6<ZiSGQ4lh2m&lxzpG3(-bB)-5p(bo0k
zIa=vLw$8+5(WBeI^NU9?lZs46YW6Xu;V8Ms%|2OdNhT`9qgsr5LX_kqKA9uPR&3=T
zFU{zVd7i2c$pwgiBNSE*u$flz@Ny@B-DhuS;GlOZjI)2%Ehs18P|up-)ZMi7bl{${
z29bjtV~y;n?yjy{xw55ee*0ENKzp*VrtXU_k5pBEN*B|80%yU=YT{~96}g8;lURz1
zL46MCG6~e=NQk+O6wMKnvh=eOO>9!lauPL8B<CzeeNR#QK<(l$;QjX<k3LK`DThjj
zLS0RI!ylwIlw;KNgA-}u!cX(%IHYh+$!QDXU^_XzN2{9l=vms|2+2pozsJhu$`6MZ
ze!h#WOctJ6J9zY9nepVa!>MX2T*EW_Zd9c%yFA?A-D{p@Ok0j0PWBXKqU`5&OkE(t
zF_xGC|5Wb`Lg)#RCK9<SM#(b1aQX>qG0bS+ho)WmX<4sos*FkAL<O1r1bKfO8s7aT
ziZP;lv=|P3bxJI^nR6^3<eefNp?A56R@6rlm_V{*gCM+I&-E1-xjrR7%;Zw@a5YVp
zXtp#vGS-_@Y~j77;+kJbMAR*_nf$0_KM8o^TZ19ade!7(G)K#E!OkNn-2Jl7<+`N+
zz4~G7oJU?i_qle4SA5I*_*^s1jU8r|FT3-i;QruGP_~iJS=9L#GT2r3#xdpDhVPH^
z7B7jbc^vSK@zF&<9YeY@4Nzj^+77eOmz^z)?&D7ob3<pdl-ZHB?HvA35I{>yAAtEM
zJ5pxguFjvJh;MAyWmv<q;l6~@c64)(*)F(wc0+!G($ESsdOrc1N<Tuc$;nUKefvQp
z@~iUl1*hFIH<6K-i^6N5`p|FJGi>gX)A4ibwR6qczW47I>vT0+oY2fY=DA=M-UVhr
zY3P;>!G+#ZL-s?$*)*FIOUzU^hl}W_+)8KYS08ezrhR4Mtfg~)7hyhl@jbwUEYDQ}
z%^VTW1-tOc8E`{*OEc|SXz5QG=qNfOgw{>lpR&*j($t35C7Sv5{^0WVh|QWu{<!CI
zp0D_~>wT&-eOWN_n2xL73ux;<!DFZNA3R#-AJXXp{QNJuprd>}hjs)Hl!gBo41mDp
z%VogDvAEdvUgz5D6kv4be4*2^bqs+&K|qK=z$wN{<f2oPzN|LfX|su7me$1K*I*jn
zTz>Q^js9Y}50)p~G_%*xu7a<R=L3V%c}@WWb|&9oZvgv&uvP#rI<Ai?qJDw|;7P2~
zEbt>%qD~p%OKt`#TuKgdxZaJ$cEQ8a3MQuaDIUJu?Z0|d^yLQ?6b&zpNtWZu-&{nV
z57=a4=<W;+-1G)t?g+2F{=`(xj@%d{XmQz?`%!i&+^txfVc=782pWK9jC)qg-$Vpo
zqJ%T9#xcDa-JfgmunaC=R)T(lw#I#%mz>Sb`ep>YbsBFq-*|c<PF+m}#{;e(<0V^M
zjgi0+$GQ|Chk-tK+s#G6F@GRG<49Y+|8NJL^|}#Kj5v@_ybAN4@V_R<`<QW}ZQF0I
zGp{psu!=aF&e>GJzUd2X`w41Z^7si-Ir~zxd#D@tC37==k;Th}ZMW+uC>^A!@>O=|
zDwQNyEwYn+XK64o2vp;W0f2@r1#H`o0_D#E32x!t0-!K@7ULGp91dq)UxeP-6jj<2
zE*PXRbj<(FG<Zk2uiV$0<3J$up_=#L%(dB~@@`m)*Gk9r0Wgl1g<EI&!#O*Q29C*u
zfZhUN<^Yw{jL7R{o_0&$0N3}0sD6KEWwzn8!*dgqM6(otKck<ZZ@+Xl_k{n--6P5?
zeqg<8df=uj>I<L@hp$<)00$Y}yjR|L{)E<Bio&U^bx~eUe`x%@f3AD5I}XbHdE&xX
zrhLBJej~ETl5V1ht2viTzp(cU=7)Y(iX463?7JIx{-pA^RjF+cXS3&M|AygTsLmhu
z{0+3fSL9MYj~hY>?gLu-%f?*2VB```xgNco`wm!@j2|%o_DDbKU4FspCxxr&Az<>&
z{I*BhzH`7!GhR&R$894*ISHLIV-djcG$#pqR&9W&pC5}N!x;3M8Eub@eYXLp8$sxO
zH@0GwXRwwZF+V|w&YMSM;*NVg8*PA`z;o+B)hPhE_c62dY=ic9!&0wFsV=>I7yO+c
zhQ;iDO#Tl$gqALPV=M<4wZEzREfwKpL7`lsbBmTA_ATDxH;)LWeGgIq>y?;#O+tP;
zYzs8vTi>!q_Z_gSv%-Bb4dLI60MdOhhNof=r1N|N=^T}@)4Fn35+1dj`pN1_-#Gew
z@Lc;P*!JsFl0!c(+j1h`1)lXcGfwZ<xT(xYf!h2S!*V?RQ;-a2=u&Uy4vOD*`vU<m
z{!3R;Xhcn>Ew>p?H$+&S-;n|8Sl&$I_N6Z_%;Oqs$m9D@5afv~G)ipfCrC&feR_9n
z%i*DiqA!pM7(mgpdc(6TzgNm*8+&_f32WlQmnQN;qI!Mbk$OKt!V}Af2Y8e%4q3Vr
z)HeO+CH3=mc{)eg5D)NEQhY~-ANL#Uk1p_ugS_N_Wa5O!US#F%Y^Qptupd`=WE1BH
z){bv!Sob`)Z`Hz5#^XFR7j`4&$a$|NrQ|$g7jH4<hNxy=q_*AN-0cm66&Ix2G@Zi|
zbne$DtaF@&C=qU6OZ^1dwaohgJSnwoTix76gC8?pw<4F`p}U<hoid*L;R=Rb_W+@7
zqYPUd;7OpRarku6Dv0<0h?8i)u{F)y%;`t>-$lQ8%P{;D`*%UzjQ+jo{%bkYY-dmI
z{R@ds7oB3AH@&_?+2OOI_X4f6T(3^87i7Pswxe<gqVOcTOz`^h|MioRzA(KVlB)-|
z_kSmwMh~TSUE=!{#C4PFJ2`gs;MU(~{85quV7gZEN)+_jY>#M9?W7P`tfFTCfE0up
z)s&S`?|oAc(Xt-cjxL!@Y8l_$`VOgEN``+R$i?ywZPM@0ce6sTZiDL`#p+!%nJdyM
z?T&yI3~TwU{1ydJ`Eyxd^W}u+`@QdUauQQrY`;Z#1HyI-SF`Y5vG86xu541+$S-QG
zT-ZXjxG&HXEKSYw(yBLy`Z=elsdX<Q^*5l8{X$AxWJ^$o{o5c$>apYvIN&rZ$wYOE
z;A9Gk!V>pa`aMp%AKRaa`XBWl`_9E;^-7`h5&D)EWcrRCuIAoD7f(pioe6l6EFOIE
z_L8mE43F3AA$ehZTfjFsg#dqp$EA-y$5Kns&Tpa886{(!EzR086p!!M5xGu06igj*
zV-S|G8+1PeftJbND8?k)I*w)7^)OPSwrcwIBjshe=2>D-YxI|zNJGh+1m3y2Zo49N
zUcG-`lWg%}hka5@u6`%8ibXzoT}L1LFL?NWF8bq!kNs7<n1-G8+l4q??R2VF+*Va#
z-a55#{`ZhhqpZ)AZ#C?{bwyX#NGe5(CXmM9g9OdX9Pa_XKX90-sh7pfwQ|Iksq_38
z<?t4y<iQKoA#vNX*|zu>urJtwA8K{AvGJJSvu-NA*SmU!rePpXO2RBu(L!r3+%b>U
zoU3j&fbuoj1oo{CQ!Vb@La7C1xDo3uSKm2nL=#r(<+}V*DNI|S*D-%*xDv-_&w8NS
zgJd1_m&Cmbs6optQR+FH$TbMtd9m42D0kfUG|==$8n~9dM#qD>kY6uow8S0>e+T{R
z1VWc-_5W5yJeUnR2@_@?+bP@IYY^zq)ciH^ilt3?z6iZn(d98*g#9PzJBaXFYHy$j
zuQ=;KbC~Tr=;1MQn_eS1$ANC(1qj{e&s?C{AG5akHZt)X=muYa(EhUtmNug`{fYdW
zm36&L5S7w2_hoCPFD;NjvuRZ8^_I39KOg6|_z9}}B~{j<Y88oL1Wxmw_1te_W^DvV
z!A!}n=O;2^j{y1tP#Qt`{a-5au8|5Y;a><T2@S&P7LMb6cT@webU~SWaoADYWvEWI
z#No`0pCCMtg4>%w#!G9ivXPZ4^E>-PUx*U|9*MTpwA~k0OYBDg_ie!-7Rr123x}<Z
z0>UiGHdH@BI3O)m%=y;31tAKm#O|7?rMySzKjyjbQN5x8yzGphbndI8$9;MAK*&uj
zD0rnD;=7oMA*^s8SIA8`NaALJ#TH1@LAS7l+ysK|tdwml1AZT$+`X84!L-K*OnIVJ
z;rs8<j+^Jz<{hyCA2ua=i_sWj&?_8Pb-^Io<+5$S=h?z*EF@^}b)5k*8A)tMH4<>W
z4`O1z`&IP1lwhIUo!QgqtQgImmnuMVC6L!aSDJB(EiCCZcHsM3p^WO*EkQQgR2yiA
z97u<mycK+1N~(^rT)7RrR9N_>5xijSvdMN90TxCy-@QEylMl7p+JMa1Xkl;J&Vf<S
zIWrmwu2Hm@uu^DwjRL71Gq$m>1fTp_C%ytY@%w-R3i3yS4y6?|=ca1A+=P7uVi^lN
zkNPw2b==@de@5orzd#-ImlXPpvSHz06h%kPvhXj>1)y|-eu94M4etna>+e#(2;u%;
zg#4R|pgjamM8wChmG`6kV<wLT?>7LKo8Ppex6WWY?wkcssNc*0t13n^kh)<5$^Qn>
z;@<&!XSf09X75*m>X%XYuP3zNM2AX3xwVkf08=07?=8Qq#h(EGOSQfKK>y#J`Bh9K
zBX*eWm)X(?YH=erTdu}FYXn^+8o&<An1NZM!>^MV!b-&)S5y;vu~mSD!$?>lCg!_X
zr5Nvb_7afugYUw=f#wDMlU~v&`C`3TSyZh*s*$Z~a^9ZJqp~=U=t5h7CN{RPK$DnB
z+Ma8~&~-lJAVy}2n4G6Sw04`?7Jc7l)3Z)y5$A2eBb%3MJy!c6M%sC9#-?%-?@?pn
zN_X3cI!mzTNkXMO%~5Vn7#1F5ojWzV)xGUUIVSmsgcz;tY$!RVtrV|sZ?#$mip~N)
z78lH`o*b%bE>fw5Ohc{FS@_HqHn;FCrbTsH-@{AvFvGX$5d_#ePU24&`LVx;S3e!o
zDD^shFH{}z_`0YbyE{}pO*fK_`a9@i^o46i9aObtGdUDMf0Fw)sG>Xon73H&L<!Lv
zT3ort_%Sh;%<ub7rUEnAgQ!r6{h8HUf^*nc-)lAWThUh}>xAo{(@=G~+|ha?tcCl)
zXLvee1`pr2^~>#RYxTE!Cwa>YKS4{sa>{@lH?d0<2VbXGVn0DxU+_wz?zUK#%I^e+
zv@EEj9hWbVY^!u%f8Qs$VggBwc6<=7=Ak-$+b+LX47wOE3qHwPx>;8by&>|W(jtCv
zGd%4(gN-lzkL@PP-4&{NMSWce&_)TYAq~`5{A=5PclOR`=LfxNHM&!gcD>)a+6zA^
zT)tU%3;(SvN_?LY?cbRyyR~W5$~FmXUiiGocBTr7Dha>qYiT3A6Bv^58wUn>+p?Y4
z-^(BV#zET$s%jFF(>K33kfsyn0$xQjoxSo}KPCB{D|~$R*Ak;m0drNq<p7;4NQ^cF
z%z(qATMj})=4dDN9%=dbSnZ9YJJbb)u87wHE92<ZLwp~P>SNv1P&~OVsBNjorS=Wf
z$C76n|D$E4v(hpQVa-+CV<lGkozM_qWGz4puS)bceu5^$p_kc)YTd5U=m9;=xKvaQ
zHnux4p&1Jy&NXv@(2D|mtJNe<rQ3J+y3nXJOK~>x>(u0-C}`*;@I37a23^7{av{o9
z9Up7i_JcDxJ4pGsK!0r8Wa@hOrSugiNbxvtzV0wrCu2s0u%+!aK!zOF&G5{xBv4-g
zNCOuNez9@ibA=B1y<s%Tm4~<S(#cT5taOXXKS9SRK&IHYqS6UAgs<`qp=aix_dU3n
z(Ce(a#<S$GJNvVypn{5E|BjwO2f*2Au^cet_ajPF+!K3M;L3I?**<i(`vAivzLAg>
ztOdQyF*G|f1l4LNd^KDVtHbUNIZM8I<j2Sio9}kWo^F|;c_`fW07DO8(`N|4EfvAQ
zQX>ejL=CfgY}zVH@3^BKzV`)9!^8*^nLxJ0c9pgG2B2}~0JOU~A_eT!<<%}Yr2=4M
zbhh&a!zAH~iWM9lb@|zlu4A_tP0=L$E_MR-;>|kmFU&GBR~X<s<N%<(^WZ0_wVY*c
zHPhQV2~Y^&VSp2b2%r&x#8g_W_%|ck|52ulSnf8Z`d-M{WrF?YO|5!}9?>Ys`>#rb
ze_Z}&(8>QhEZ1Bw{T07|*VeyP$bYLe7>Hdet$(8^|D>a+bjD@jD?nENphxH#B>{H;
z8^kL7yJbKZs@|f!&Ih1QOzr#=RA08)g;(Nvr^RB2?RUik3QSY}Ci*Nxo-_5wkaKH{
zmJ&0}V=m7~drL2wn*z6}_TFO-Af)n~&39nDM{#|I03kob2L@sgq~NahHTx&1YK*dF
zGdz@vcKS8{CVK{e0-y)^oybtV1wWt%8B*a(*<VAY4LroS4<aH~DKoDnA_GiPZQKos
zAKj+<*MZ6K_{OBDP?LRK5VmE{r~B~jg437h_H++@+bH#M*$-mg%hA5mAy`!c$4SK<
zOSLc3NQH~gP&bA|1?#b+K;F_gc3Ew$jLef5{%^$yGL(vkZOA?yq0*I1q!6l}`31u$
zBF5gyNGa=Xxhp!#1eOtV(fAZ4(b!w-nSYsR<WiOL-6X!WeFjFHBkjQ4_lB!hIuUx&
z(kv${t#dB`$D*vy>PSggZ@hIOf=0MH1x1L&z8jt+j62hst~mKzEiYb!Kt(DfUytn$
zA|{G1njCj-1EEx!<s(!oOk0HgX#an+#`6<}N8Rapa|{R4r(aiCADlrKf4r<2{c(3k
z8lpu{;}6qGlMl8tjkKkR6VlHye>j5t*m(1L4#7(G;x)Q*UZeZn%_){ls$1Jb-}RI8
zj#^UC;DXW_RKm8YWh@oQN;X`|S8tT$Wu=P{@<u~be#qDpoh!CMz2dKA8I(e$KJXRd
z?nK=NL0^5P|J*vbzrf&slejgEQd@}<FBV4>o<_o@;l0|Xi;9dyC~c`m=<`}x5fQ?%
z6`q$XNT=>v7$kvFB<P)aVm38ZIa6@aVTB^25}3bzh967J$>Lnoc&k1B`4`J(fh^g#
zerNaJF^f5=IElW4h&gU0XY>)c$Puc33Fl$K6-X!;e@WA7tj09jE8gb#0|nPkHNKXP
z)jm8de8CXC0=85d&mWqX@X+2kOrwupX#Y5w9>WOEzu>0nG{|A-{N!#9$s5cqk(4Bb
zt|F10sZp4!s<$nYlS7IF*x1-`_Om-9iMqP%GFyYjM8k<`M##FXtv;PHaupO71Pjb|
zSv$RpvgWn6tlRaBDK^rCI;O2L(rAXJ6cz&E**r{yiC@&i`B~T#LQYuZa`~1e0_F2(
z;u$LQ-q{^NLdt36oX_!?&z|O`Y7To5#8K68e0XUc-)K?S%NCMLYcjrGzG`ZoKu@su
z6u~1JNyc3Hw9QDFM{vhV*(0MHH(xI7q%BY(HWVKfY4o)3i|6dh_-J)GQNPM`NLZn#
zvs}4U*6TQx;$ee92D@mvUKn`J=Jwr4ve@?yGTl;^0(ZGe=xOU0A2#OX>&O5Xbz&ts
z8y`c$w20#Nt!Y=w*9r0d#HNNEF+@G%3Qwx{4f)7MoRdbhs7s2&e8KO^=F5K=i5k;e
zOE*80sE)AI9dn#67|k6aK*3Tzl}L6_aKeXal<3t1OjHXiaoXGSlx)IS#6QzdnsnpF
zn=7m~(H>d^rMtIURFjcHI*GC#<dSUm8@y7MOAHrKXj3`A$9s1ql7~;cpa=OzqN!}W
z+^93rnBM#>%9wkziR;s9HR~ER+u_@Sp`~_fdX7Q+tmccd1=-i3_@iPDsPZT^mXxp}
zV=ZdNEs+T(q6^vvd5;o171%H%*vNvN(K*ywnPH3JRH5yjhxJG~X}0?*xk;8*D9>Yi
zoNN`?GhZmpQ$gkAU|QMO)B-s%iDCvxQu86lmjvXG38>rr(PF5ZNO4L%M~Zb+x%!<`
z!t#&AA~A1sg;O%~zDR|b$LNmRO%)dmH_QT`1bo)I!mQ_(Zdk0e1~c7)gheiwm7B*r
z<o2T*Ra&u8Dxa2Q^1avaaqo97v(!Y3BZw3GSn?O>yK18bGg{A5psX7`pB|SE<vv<{
zIi!3P_93l758qjROd<&ye2?PPCy`4#Y^56?JRhJCo#gS>Ux0y<I;&75fCH0e1fR7q
z!f1L$RAvMc3wgETu#DIap;fN4a!U0`vE`vCB%1u7DMR(hDIW52$@z{^`4Im*jhZpZ
zLFVXeDS2m2UkssSr6L*6T*eGR<&<G3J(v6*8M$5;8L&kAmdWS#(1H#}#etC#OCZps
z6JNH3KXd<}At(KVoyf6B{OIIjM5Xhm638ghxM{S<Ed}?fx*E7+sKj$}sE!dwLc3F6
z!G=uM(8!+LK2JNow(8u0O4Mb>J1FrYTJ~vY%G-D;``f-#if!s2qtGf<JTPpa%x5Kj
z-C>c89^$vM1>v6hHJrV76=sNfgYmb6KB^TNG<AhVI^+}H+hBke^I3FjgoKQERNQW-
zhl^)rF(s7Eu83#V6%<)f7t$VzbgS(YnGS{ZZssjyO)2EOqc2}oLq!+-(2IxE?u(m>
ziLLhuJejl@lZ<DPBZt&h$e{>LuBs-*G%HCDj>2@rizo@XA_*|KrA$3IdFfZn4Vx*7
z>EXtlp5q;!DUqX&TAp?I?liqfBo`G2<^2YVZPof$s&6gpd+H0;KUy*s1G^%cV$|$g
z5xWi4+diUJs(Ur87B9)953;9@Q934Py)52jy3r&2U?{1D9O+Y5-uTQ>&m7azTc2`+
zi;m3g7Dwrw!@%?d>UM@;SX{-6#da&_gy=ag^oP1fwkXwVm9FU#X2aWCX@hkYLk>f@
zmR5XaG_7GJVp=-gI6GkJ73ZEc>V-xKlO(=Q{f-Al+4M;LG#x8hBNiiV$@{mTL{)}!
z#AO02pybLh^HE4#2v-b29Q-kRs747t2mMi4uExQ7rv!xkt&R70Eqv|h8PbZftYS`&
z>aGqCE{YNd+O53oKB=B;Ra4X3^@1vo%9ljme*{<mD8la>1=s72RS&KZA8(q?%<}EZ
zp%6B{j#qk}*So5)5jg&Sb83f|<Maf9m)18gO^NWMsdS##rxHy+*j$F!ReIPA&ilxU
zBu_sL&9rdj%XWEo8cm$c1UJ12oIKGfW*w>=9o~MrsilL6e!lX&x@po<*GO4SuPhRm
zJFoiKxq(MOO-p4GQN2|&p38%?YpTf<HZ$g;<<b@xi&~&wC>b&@N(#cMYA%4mBqLz=
zOysKMU|Aa6Hr&$eT=`lunQB(D5hy|85>@N^^70v(d^!1=TUj)TWa1j?FO|sJ#Rb))
zM$~AuOPUKOC>@kkG~+mw4p5gx1h3o_RMw|-G-D`ctaR9{CO5ouFMW2ZZAOjXRoh4r
z^-<EmOtns~HE~Rm;4I|pmLgSa%b%gDQ9+5*N~$?t)HJL#A9KZv6*I7i4C9n?+9Ntb
zF&|5bTZ-+gkBm4juv-$+i5lr#sPu`6YA4vJKU5lew<ay~I>|}}zu%vX0&H2RgO9}N
z7gdq@K_WmQ@cyQ}gpawk-j+#oGMV;$0wwKNy<EkdmPgZCVowtp*|YgKF*j{E<)32N
z=0MD6XWxd@ZPYqG7&Q=$diPn?vh1nomf^R&!L|I)BH@K&CyLZkU&6kL4m;`<tVmea
zl~yMZ_PS5-nbABqUCEtY9<?&+47}G58R3d4ryK#-GwKkGiOF`xWNM$xjN$Y|!u7JJ
z<4=rB$tw$p=$%6#GLSr72bx&ez&is1EmKj~#MVT)-24f9$!GilYKLG8E~!^8JtG^{
zSYYG=w*W1Q-I3ts9NK^v_cv>XB)csBhZ$Hj<oIO8YHgm1Hl5Tqx7!<$#LueSIPOa@
z5$#b|FyXI<tXS5Qr-dv_aH$j`L>*X)3*1t6s28r`yf7%&zWG!&jX@o)eCf1PnCMaj
zDlsvKQ0hb1ZYnMVf?%qNO1zr8LMmdgM7Sa&$evbLZ+1i7qfR;D-Nz4=oEI`zTlHdw
zTzYsliO_JoDoeh?RVynliV?8%=MFL-agFP=_T`+ri4vCLNdfmRiaZ5WcCCtQv{;I3
zl=~D}$%y*UxU?K9rE3|&?Y}81X30gW_#u*2H~XF&-=X@LPp0<O{`LY4Yd(vH>Z`D9
z!C)S-9PFupY^s&|snkGEXIzpMCDds2b~~@VB3x#C1(KGW1?$6nMH=)>((QY8n+o_{
zv>`hakut5QCsAKL>u|Hy0GdX7FCT&tPsxV-qW80kQEu+LJXISH7S_yGJqx_D{LgYJ
z?TN96tlWp<HG^`TRlJc3ImtKiABj4MBu(XpUd;Gp(-G-LuP&P(lh&Ds<Lm_ElTzyH
zKBFt5=IZs}LX~)Gk7+0>%!eyPLp@rf1nDrDtV@H5P4Dv{H1L(SJb!dMfx-2rVPVOX
z-S0}2Sh1Wen=tdy5t2!AX;Qh=%3O?2Ste^VjrMtMT0b(L=df?$rpgs%yXTXshQE!~
zOAw$DkWCJ!Rg6oCPV+B_$)z({?CJfw{j`QjKplSqms!wF-dNnKoa=KFT)Mie==woc
zYZOkozF08X!`&i+hf@UP#f2jyo<y284os>&HDIvz_hZXE27`25uM`!l2U{M3N(9|?
z@sqh@Z}SI27NemA(@73%98UvNtwcURwcbWzONAi7;=XW(5B555H3^r49<_5+#7vg@
zb1{g)WRT*Pe43?!l=mwaL@oJBEsH1$K3e7tE1ze9Vc&Jv74w%riyBVsEy}(&nyhI`
z>lBsXvM#6r%fz3coR!km{E1<b>BKOZltE+Qh9|?uiPV~Hn=+i0G2Ku^>N-4<w~7fS
zE=+8xYAJ5WnX^^p&IfZX<~<lvFqH0n9Wu8VOm<(j;u%lgNmGe~t;oga_F$sO(0f~<
zI2*-X(bHq)-5mf)+G#&j@MO}%c{OB`8Pk>8Q*qN30-lb?>B}_~;6HDRZyg@m&+Uj&
zfSetW+A~5{bs@OiRWNKRu(-&n$({{csz0WsGS;Fb%p$}(l8u8+1oq9J^1gWXLPr1@
zMbYPuVEn?5#^QW^jw(cj1qk7oV}5I%dJ3rz5xfx&)mvWEm`xuoC;A+9@QF&rx)l|P
zNN5EYpLJC(eoMy`QPu~7H`h8+Gv{<biKQ((UL_5Z&Y7I;R#fpL`&c0LlY><Zl8U4^
zx6NPXW{W|S0*P<8-z|UA=}$)fSv-L3VM2_7*oPylg^r+;x|pX;1*?2SrJ`W4aCxcK
z<XCy1C*`ms<%k4VpMy@RRqXm|9T9LSD%vDd2GcxxkwUZEkq~<3O$Ihgi+au(-ZS&7
z=W6alyowbuR?b7fD{}2rsMPl=7^5V8L~zX8ue|-7y!M$36!OIPRoR*8%L=&EqZdU&
zh8?LzJWYt;Mw11rf`ui3Acz#@7oC-o#+Y;zqpea3$GC=m^-kOq7EZ8UP{+t>UdM1m
zoDKneXC*il*MHz2D#5P>JpaZjQ9vwcfF6?>uq9K#aWpjSFPG$vKG*enYIm0Rbe6A|
zI0DwjQF}yGwXQHMEG(p;paAB083TwFIPd}1!opJ+8I22I<?`vo>;?phviXE%>(g8W
z_bQqol^VqZPSKK#iTed3dy>yIllO)cmpA?Q3%Rswg$c72xe}1UsIFm4+L4~-4P#N6
z-HM5fw)1M%ldGP$r`1gdx~B^^Ll#F88I}Ta%@^H4k07?M4kv4dSh(Y~KGrXG(?e>$
z$7zmMKKuO4QZWWUX+P-YzHXP@_<9UAUvV%_q*~LiN4DlSu7Xm>v=#B{5-}AOQBf7;
z%LK)OQim*R>cYa37<>*D742W&I>DLZ9anUz)USBm(t{6HklqD948Ju()|MrsFGDRt
z_Qx%=@^!t}z=@L<Rj^J9C4~<u5h3o+w22}z--c2O5ww5T{~B~wo-LL+V$d3hMXs_>
zE&6{E_ufHGuI<0?S}t`d0veh$Q96X80YViEB|s<|AOS*AS|ACbgEV!al+Zh&E1e`H
zNJ)SoD!q3?31Fdj3kVA8d*b`<-*4|dXXfnl$C)`9X2^s|c%JLNpC|X{`@Qb#!hbh#
z6|cop!`Vnf;~%y*fZvR!L0#vk=%rWIv|to0YK1xMPf4Q~;7KGBoxQ&H*MD1K9%m5?
zEWVpQD0sp3rb^&Of|;%Fhc5J;zAT{WYK&`la=*ad$CnLi6UJ7_@^U7*A^T?ltEFw(
z*-r&rP%es^JYPPEzfuhOyx1z<{`54TTA*>QOJ{%`w}|js>n?=G55Mqhx;2#@CebNS
zQo!58X5nto?->0?&PhWXEjR$c5BCZ)<FsvORnIiAoTO}nfQf5!2^`N(>^?WkS269E
zQWaCQ-J5@;zgykMRX2e3!>B+lQDNub<+jZ*ES%IP-|}~+*VTPm5OzVKVyz;tp5W!R
z`|Juy0M7Up*H)3%T8FSOWfK0eBrtHM#!EBktX!JdGFGs!>3sNnvm|n=(?+YmSwaFx
z+yw*6)ss%ZEo*N`19Q)ohT$M($G-Os+!FB%S<04qDVIvX0FmUTC_t_NBlmnqR%TvN
zPPDwF5=cwq+OOf0)zyL-ynv!AQeyOpbN6`IEMo&t@}IpZc~oK$dp0>vKvG=a#ql$H
z#<*G~{GKLSq!O`8D4RQ}Z0__7iFQL^>iNp_D=ev`OORqEPn?eupt!SOpD-Dd@5hv`
zm_*~w*`-MXRet6UO2==Ek9lAOT`%<Zy)g(3ZSGfsB9Z4DR%^)u*U>1wnB?%hPMc<c
z!zxYHTK{{oHTrYW%w3P=#m+F2il}@QKy1YTZjBZG8juCxNAxyJ5Vya^FNnRZG_JPU
z%sjdzBWt(>hj!BCGuPi<JNC=@99NI%M1`TRs$Qa|P41^K=8WI^p-q07B}RR<!0=_M
zTX+}ih}^5qOd}*|fG^gw=ckT*2xpvttOtk4;%j&UnJXo{dW!U$J2_5j7^Vn_3Yya!
zjM*cCoAL)UM#4%hcdapw0sd_8jHQnGDR5JvPp$KVf7D}{MyXqg6P}G=VdC9TPW4JB
z`@8Rl9_i!V%ZNSJF@yOyAaXDr;<~@~i7Hz@aKpbcR$+2O)7$kP_@+@6)D$!(niw<P
z^BX?d04&^TBhlQSWJ;R(f5gwJzI0S&A=Sb99O^7n8Oj8{t&pCYfkT>aD9N0?aZolP
zs~ldtltImdkI}t-Y61_9_d(w$kK7@YYw4M!h#e@-c0EqvPG*YMwHLFv2W3K@ei8~o
zfm6X$vNyrHc}V#B+2G&|K=G~EKp(WQq`1mfEtpN7)BaAIAz|DSwi;$kD_p|c@up2V
zXMe(pr+${eJuCil`dl4c;F+@ihI6AB`drAI%xI%MG!xR0#_p7E#JLDHIGIOwK3Y(Q
zkBOYB2Mb&O$UK#J&^#EGR^k3Z-!>U5wpx3sdbPG6U&U<X6WS0W9Qg00*8>2g6*nO~
z$SKLtzy?#h8%zS009n1Hf;;l}7)oSINvqxbCrP+lkD$)A>aVS0jqv-lhOOTC$-IQI
zTz}}S^Py$CfH$$i(Kj89L*45$FJods&y<WStbz3xJ}8Ey%GX(`j^&=JanpcA$lLlk
z%@}`6ov9U_bh#lZOnv~qn0~$UWj$;U^nRe$A<?KOzSrCa2rPw-iB8%8!TtDz@ZS=S
zuhQ{BPm9z|Zsto|c(wy70%d0ZsMWN(H$`+RJMM6{5}LlGn4h5UwiO5E!ay||6Y-DD
zO}Q62)a3lge83?fr)^=T91ZBUohVlu1pir}nQ>Pf{22^$tYX?=O4$FQ;)+Gzke0oY
zhSIgwiu(K$*X5{KADp|fkAI0HOu5<7%b$Ag)nr3n;!mdkC@veNNiS~og};X(Gq+5n
z8S@^%JKg%b>e14s`jB#QptKx5G0}Izla@EJ*ikx;3Gz%8p6+p1*TaN<vi*vv9kW;9
zP$;Gi8$z1dOM651<p>tTAmPFrxp<!Suhy8JByJ8ZD=?(m2+R*twKkop_i;MF%@d7j
z$}RqD4m|Ew89r?KEY~+n&MYmctLRJji`#QvbbC85Yfdm}-AxMfuDnf!S#nHy(lAgn
zMFRq|P}~2uWVMoEd@2INDHZ=>T<XS{5QRt`!eXVD&-R&69tY<Ua<;c?wk7qGs*+z(
z8&PYREexdwJBzC|Zt_V`8P|+s?6&yNSbqt<YvntI-nYtwC6%Q5ag<AddgWG+Pkyd{
zWiB%0v%8O<vhsxYk3*2U!eoR=P)x=-^Gfp(sZ$9xvr_I^PcPQl78Y2H+_Sj{J8EPG
zrLC?=2R&THINftPDtzhb8}t=NA+xj?EJk0r)3oW1-p=a~mW`{w6U`gLH8?0Mf^nt)
zCeJFWc9wj_a<lm(sGxnEUM9;?jfjS3rmIMT^0EYKA%QXod9bz>yJh4*{h3jd38mn8
z{Gp!Lq?{?B+D1alHa-b%05`N&JWE8E`45=A$$<bRr%rnMrIDITw49rJnbwANN>o4c
z;nyT|uZ#3V$eh3=(X0SJM*hlj7dF5z)6{)l3_;91F1Zyz^EK?Dsix43)|U|d?EF~r
zSG*rvPz;K%7QC}v(p51b>rvc}rx=xKresC(S3e?^(nPCv;D-N$JZ-@Bs{Z4<{#bDB
zQhU^z`az4OB#T7o-^lu+HF=IY!&ub0G2J9LEhQiYl-%>=m&?K@4(Q{qH&J^lb;Prm
zr3te^W)!zY`_M`*DZ&MhA76#d5h+6HU&Gw>2p}FFNbHGHSyf4)Pw<Z}hHh-=ELw+t
z^Ak)HX1cVdV3gRcydqVvsxTvdGbfhj@C{(cU$rH%NiJ^gO|C1MMI2Z=#>Xa77nLHh
z=H}|FtvfowC3JxN1%*Z~nKZDUqkU4NGzi19^r}MCk8n3Ot{#U)ca}rS<PwfHzF`5(
z_litlq&;y!UGjPL(2Xozoj)aNh1c3`$44bPjFYf4Yaqw!$#6If(Dz7lVq9fpdsSB2
zeNh3{+dFZ=Sr<vFN@j+$cHq|5?0~^S#9U(i?GM#>h;u;<$2M~fhx;TxtufN5`N4xQ
z1k~YgDI*Q!+dIayTCb9pKLRlN-#j*G!qz<uxcpwjY?-TzV#AN>P_3w#<4uEYlz2y6
z)iFD_D9F*hj*@H(dT*xMKuF@au0swt1_|*K%!$=q0p9uJhx%=ZT2*db->xN9<je4L
zrjWjct~te_A<W?*|KZ6k1%J_vL}NWMoQYq#e4Xrt5+vuio9n)mFI|RYs^Xs;i=_z}
zCx_dCC5@}h2Xsg0tHbU74V_vkhS)Jt(zFE4i|85y*3)y`;<&zDb_S0pbE^#s`WUbU
zh}w_XG=5dn3Ur=Whfxj8TR1SR^#u72)zV$0T!3hlJ!DZRH51x9k{S9U`-ewTE+?a-
zRfM$u7TrxK87(8M4)@-%Lb;e&yRB~Qr3UGsO=d&J>#wyK<JiC-c`0?Vz9H(s09DV6
z6k#1KPz{|zV8uo;GcP1MlFm*u<Ie8Haj{60VDb!!S)YcaJ7n=+RMSO~<83|ks!YnJ
zK6CVy!kvj{2<D^(HRwY<;5_cUxQg>Y-e5K#y#O~UJvENObt>Y=*%V$~#M=cw>uHrP
z=yBl=)W|SUh^P&QppdG_HP8UZbs>*-d{VeW7eWgrIt>mm)c?LXiT?C&=zQPho;JET
zbSvgy0bpg962{((7f&!?0F}FRsrfeLQ}??`u(eLogbe+Te$~Xi-`ts-XEHOsm|y$O
zYcdO<R4GtsJm6^Dvor)XHy$+1EXvNw$p2@@_)Gn7^t}BrEl8p*ampZO3#d$HCAb^X
zJ@z}|xD`Zr_#lQc`z;ds0Zu)#0?j0fe}JL2?u|!^A8<e~63&W85sHJH+cnHT@o-iz
zR$VmP+B5iiRh=yT#{-s~umgH(bde8|C6(B4xl!rqXRqZms+KwSfFbChzLzbp$|STP
z1n@x@oGZZTX-SRc5gCIa5448-mDqs0^3Wpx>~1Viu{5`&C?))|#-Ay!8PeFib^L1c
zMQK9=Yuzb+JuavyG)db^bz^S{q|Ui+Dt~viBXj(6lCo5`#%!9fzJmVXjsB^LAti2g
znz(_-H_&^dtQ27+iqkGMH>CE8nQ2eV@7JDujb$GJ3{Wd>7!9Dh8T(+JdT7XHBZ=al
zUf~j;TJdH!3QpH<`)j4Rrm@Fn2yO)Kt)ln?OW2Q@{ZP%;s840Rv8uejzodFoT7<3b
zG?HF+=8W6=nco!C=bpT~b+vAjZ;qmw$>vW>fpnPA#nkpT+y3ive(#+X&4oEVy^=W8
zfLsVExfdni%iM~ZiCKsnq7a}~3LE#?ybBjt`>ce>nbx9K_0op?X+weYr<<vmS>!51
z0ti%w#uE(t?mTusC{vllZ^_O)Y`EUBy}5N&Wzz)5n7~VmchBpw`&w|QQDmU(L}tOy
zJ++^TzQJt2`H8m<#&PW+bW>c&PCd1~$ynAhqM2%dAOV5wlkcr!i*B&rEGodJ+>(+C
zXnbx-#l#w>+e|`ll2vQUoxW)8X@zOU)gci^3{OjoGN?39Lg2VN@$z3vMj0}vyGfJ&
z_kUPYtc@fy&NH(}Ow*E*vhpl`Rl`}Ykq<vugz!i1;^C~f->-&$YD702DJ0VeytOwO
zGI@FhLDOfAe*>uHydOpadLb%}2!*9vrmLz-h%a$ri&krYqm`s>cS-YbK=u$8NnjIA
z6`t*=k=+J=t*}v$4Wp(csxdwvZsRY$HwAinw*1DRZPT;ZI=8+cx@3>bmqw?Dhi~oF
z2NdfT3@x;`sPb~S25D2|q=|_z8$A`xcoW}u#0jedIzJWbxt16v@I=DhG<|270gDFV
z<$!8e;rKPJ{v<364hNk7fJ5Z0K>mw6J*1fqY~D05Ccxo+Z2HwbaOiy@(O0Ogf45jK
zdA~VBy1LXbPCykCW{|5QHt&w#R*7cp4O6%pSbQ|1JexVU1wj?AdIUnW!1d0M7w_Lh
zo)#(TP7+o~OD)!m>I_&k#2<oFKh%=enqh44Hn>R}#<KBq$W@T}9~j3tvcu9lMFYUu
zM+Z=@EN$y5P$Uk;nO7{DN8hcwlR9X8?aIRb*#MMj>LM~lK+5*(-<Q+G53RFj-7}SX
z9N7F#gc#g><a&DXyD^}m;ssSIJDU*oZGxAVPrFR;!j*Q+D=FKt5-43%?(R(dT4#7(
zH~2pYo|L8SWkeU%j!HC6@dqcm&Ht^t>px4u+5_Oj>}UYsHd5?1idL0l@B4|&imvjB
z`*lQK4FARa=+mNse)eC=`>bF~bx#&69sSPSLE<0rRaXygcM6%vd+r5Toj1NKU4Fka
zCnp8)u7jIX&^*F8C3qD?_21P$TKqS0@^8m-;i(bE$q-PIIqA0yVDOgh0#}WHzQail
zShB2YigRUdW*~6-72~@Me!YmsvmCy2N!(IvZa3=m1nTw8DF#tQq-Tc8PvT3-1o<yx
z9<rah#nUtYnT!Y}GX71R@ZQNl+t1~>Y#Cx03v*gFcQ<Hdo?nTztb~L+-+q>!0N6t5
z$AbU-=+8R9iU1r|>VUHVe|;9_Yo6hWLYW#T;+qmn=L`>JqD|(IdLds?Xw!B<ZoR-D
z>g?=!wGPx4@m=T?%Yuxwe4HjNsbiTNEp3sJ_pBh9YOIoO`59H=svx*tV|@3Dij-}9
zuUCW-1{7r5qwDn>HM<giqh=D1sG!;ZKut!emEA`+X7=L=)(nyiQ`b8wE+@cmQmD*+
z`fIKusZSOzmdjp`)1Xf`oXcMw<wkR0zc|X+;(7|z)Jt$>N?q$wRRdo=UQp%A2Mh{{
z&WxYh7Cf~QLnE+FmGsdm$vJLegA~wJZeIS+Y|TS-e=0n0YM~$h_&P8us$-mO^l6Sc
z26<^(U31Z8RgvhfVpO){OUaNjg!dkBEVSv-pDUC7rqJ0xl^R9d`?I_QG1e!wzcdP=
z17w>z1DiHqo%6Jcs^o8Dmp*yq6|G(h&+#f$)(bBGy&lU-&0^!`s=jIKPUKZ?DL9r_
zqge!1fvJkU?)k}mrdCE$FOrKN-~z$-ni)Ez@qa|nH(PkV?_QTa2k`y1UJy8d`qtnL
zOjB9Ts2rtg8#?(Jd5#k9X=Z*;zr1d5m6VaZAC3n<&nyAU-uqbPoM>W9IAl=8s8-V9
zkoLp8XqiXQd$ceuyYuv1LhSj5FmV4A7ON$^Vqh=darzO&{b#|;<+6}$8_eyTr~F4m
zYzsUkpqgMxsV3+T*zAUTPsS`j)@xupjPP+77m<0CseN4or8zO8!QTwOSo)$_S@qK4
z-F5^UG{yYn)#&wiK2H$Y8pQ+XxIB%cY(QLFXJunNfOl;6&l=ZDLZjuyUHBRHMd1}_
zhyh&&B*9x)2?Js-PIhrTsOv%@m}Qq6Y=6_G83q?l#;=w%!>HD#3ow3MdXaYI<C#Xc
z%3|bn7R7%kxvf8qC-Po7!X3m{fXhhV4+|tJH?(PQnTrhQ*~^OJMCJxeF7HpiSQRhT
za|>%dT@Q^Ekagt-LkYT)9PD8zO3mfLoRqmCQ!4foLTZ6igQ}U<oo8pY2nx~_D-~js
zQvvslqWOn#W8mMGuK}aP(Q?|_d`;zj3&q|f=ZiNvJ=&`$dow5gXB*-v1ZCQ5WK4H$
zfjv?Y3p(=+gQLl#joMWt7B%G?FkL7qTk9}5lq_(LN6IsBR*1NsTHG9j{8pV6%|ALr
z{VBI5Awe~6{GX_~<4*#QE$K%kTbq)=E?QNjxV#b%m#)?bJA(fT7~@D^7+&UUXr|0D
z8921>=y!@D)l$CBjil{l<t<&VZ7j=0OXnQom`Ja@MIEz0#f(=%rq4!)-`B1*FG%%z
zpwG^(X|t=kmM+1z(ZW>ikjWUvv}rGIdIoekn+=-?JEv61_QYx<r~c)bHO$)qEbGAD
z3)oa+dSmqSO?LU27O(K)B{A`j@idE)Z-?=LV&KI`8^%dXH<Xq@FI|h@tw%T~XFMqP
z3A2)eGp2pMgW0%w=soiGv0~-W$nk3&D_Hk6qb*@JxHZ#hq=C7kqnXz@`I|T0gLGF>
zPWBCK^L0u8X$={NvlttiVorI}YH)6^CV`DLBG4U!N}OrYyXO2S6=~psu-X$1cx*&=
z1x((0mdEPS9qCHGaYr?gDAv$Pw`epl5X$5qwYM#U%PZ#?G5*5%^ZIiq&tN9;>+)a@
zo=%7YD+Pg-gbx`ZH^NU~5NOf5iEA=|c>Gtk+v(=QI51Mw(19cV69Gksiz{j9Rx#a)
z5PN~aIGp^qL<fGByU>PW@k8Oa$xNuVhZveW;*=)KI;!S<ZrUf&%z52=)g1`}Df-oL
z=i6cWNj&*xg+;L(k{Nc@@!xF$_FS9qCX@V2NT@hh(3;c=oRXWLO$bPXQ&PnEr2Qpk
z;Y7d?oNdKgDHIGm+N3M@j+oP^!qNlWI%@(3EsuyC^+It{tV^0BH!Z?U&2N}%&sFqK
z5zaPVXkB2IjXWtgCc{eVPB{@^<wgnT>9MEFrBx*bq@)9K((Jg|UhgZh7O|n=l14i!
zZkK5=e>OKXo4iummgDM3TDukBOEI#$2zO-Ly+K(5{f%`cAFl^JzEG`za1Gol`P#6<
zcv_nVs^UM6$uP=UHWCfZUeORJRo2-`TpD|aK3{%%yuf)`8d5Q3@{1HK+!=)D&MUjC
z5M+Wf4E%f0ifA=4Ds-vZb>Y;Q`*XvX847!88wmE-6#^t%*R#kyVqs9b4Q`Y_!byTA
zmY4$K;AN6X_+>968l1h-mtgf*5hF6cOe!S5V_Ye;n59~J1JT!)U90Yr0GLlueS>K2
zwt>nP@#NfwrrD-xlUGlWTqY8hLBD{}%H_s6hfs=kST1;$eznC~T1KiuN0ywHrX!T&
zyBO?)XcqqIXrthOC%bFi?}=0y9(Pic`?O1$ehNsltIlo-%Aw-MHv(J7*F&Ihr9Q0C
zbOXuB$866TGKZU%r?q4!L5pIFERL0_zsb|oyYz!PP2I%r=2;o{&c_ik0-a2|;9<*8
zJHh30Yr08!sb$u(((~W29<{YAD~ozIApdaKG-kOf-naW1+vn`aDa?#yYx2(Bxlhlo
zvkezEb5;{;aYaira!i&|&mB~a51hzkS!ZeIeAQO#{R$1=bY{Ldvdcf)v!q>d57+G|
zrCExbRHCgnR3+?r5cl?*bJF4R@{GV`DVHubZ6(R8a{UI0YGYo`SbqFBZ|n@!1q&Jq
z@MaoneSkc@)|qfZQFF8~(98Ia7eogR7Z`CNBiQzmJKJFk(Dkl~`-Z($)mLWYce=lX
zS)A{#oNL4I%w#26g=R0SSvs2)3}$o9>^2;`ZJB`Uw!AB?m4G*y>Fpt;awU%7!Wh4k
zKMEeYKh=<iR1KLzAiASPFG}BJ$8pG%L2ii9#27WMIrIQH@4tytLYoi0p$W3XdHSKA
z>*ME2w7@xsrRh6t-&cAJ7A(J`#-J%Tp8z8g?#9}b58ZCzE;mhO>Yb$bVZ7#(R3r+M
zkIr@~ou(HcO*}2l4%6N4rD@1&+{sOly{Vuot?_4X@0WUCi2<`Vf0}4g!U<J70g)^!
zHcf2O@wX;)ej`>u6s4{xn3%w|5o{21(3)(@+6jjvm)W@p`~mut7)?s_e&{$c%L`JA
z4wAbHyv}+E*?Hbx$gQWT6bF+U!5BP9jo_Q06r;CH&a<-GzR`o7IOnqkIs^RcIgMY)
zxO--4#f5xY_yM}o9SvklK|3R^1<d)%FP%iIV%^cjNvYya7xF!Bsmas2$>vpT#L^<=
zkHPp3HB~>r#SMs7?jTs@9k$J>`ks;NTVasOa&)mafem#_89gkQ?sjtTqAy{fqvcY)
zJs-?W2cbqu3g8oJ4TCRj+HwdvQEcFZM6I|P!=86yUGx&T)+rGhv>Mkm<g%fm{?W{V
zfvbytIQs`S7%rOZ%35x5m<Gc|nPy`!IKQg?*vjHA-<vEM&TX?7;^xuf>lxw$FO}B6
zlt7agbJK*)9+1ui#c^s!H|!}FbE_@#MVGTTJ%*kXy_kr)(w;+T56QcJ=F_b3$q+T+
zGF#B|F3b&4TbK}V)gTs-jGZ~qLLQK`NL5vppBY3m*Mgo&H{0K@X?wH%);v1fr^Lrk
z|0Y5<5-K(q9gCUr%%H4otYp-(^V7r(0yc6|J9<qMUk(g?RNu0omo?4dEmIrhfzTuw
z$aQ}y%~ThIr@|s145JS96F4LtdQ%rc{nH)z9*N&zja+R84Eg^XJSqNKfFZUm%4U61
ze6k@M8vlcy(AW`&?&}inJ;D$8UhGsNk@-QDS55u6)ff5D_TJ)f2}^-!K)DHwC%r^P
zMN)j3^5z?J2Rj=D(w{jf%{lmRRG+untUL|wldJll-tpyj8WIUCnJfdqysGNc*Q<y$
zejT%{^CiSi>C>Bu<19<8RXjBA>}`S1)l=>a6R97Pu%ajjvwbm@B>!f?B>z^2BuELc
z*@5ha0jt^om*cSSB#(Eq*=745PAvtjc`XU>7>8DU{;Yrg^25Um1DvX%bIC9FhO1KN
z%+=%LvzL;UO!O{9<z<&jl_d`1LsdUP)S`Mi<Dx(Yn!c>HPW;g+c%rLd{F?4=U<sIl
zg*mX1h7D0{=6nyODD@}dptpzpOTa`eWfGfx4GiLf*p4eN*y`OfyX&@ET!xsplxCui
zj5QIsE(uBOzyy8gx%|Md!q%OgSL^6!f}^wZY7Np=uZl!-D6&cORxFR2=a>+?vyj=p
zjZJ?QQ8x3&A9k``q~C<+Uy6)p8;b!2Q{WMHGA%}d?Q60R%W^lMfDOEC?1tB|Q)<-|
zk?oAk-pr$5=^53vk@QZvYtC;4_H#2(&>&7o9FGE+|Fw)+Y=OSwwSIxEVTuwxOUb}!
z8T1KtEh|zOgXM6sL!&)ml7nVO#qcX{B2S!S6Y5w=XO@@d-C}+0Gaku#wz+Df1*;-G
z<^^WXu(NxE&R+Ztp`Z+3(Y4kA(iDxz=1qT+1ehHw7(i8j0mFYrEyt1+WRiw%SXa~n
z;`E*$`+m?*>^JxB%~ICOchTXJO0v-2!{g3Mg5~hE%TJY*RJA~-EiIg`Pfm71+6md&
z(L8J%ud!F*W9J*Jr@81NhFpc$wEWq|bL~er8Eq*x_o_?$&N5DPwF3W}#D6>B2KwEj
z%4Dal<R2}_<>&LYuG(fM8Mmsfwm?g-JY~-k*al?!oJAgIws})o^{5*~O$GEb2&n7C
zhB)C?U^z62KY@o2n3*`3Xq-A0gpg7<*Na-K6^&;G!0ZEQyMDKFvvV;Fq|3<;B!e1w
z=EWk7jGaZWv!>|71BEHN#EYt<kI=)A|64V=|Gng%@d*Qla@Va>|A+MP|Mk!8Cb=g9
zZ*=Sn{#kr*`S}~J^Dlm${=@TO?E1C8^tW%`{zv$-^AYDCqsQk*@1L2la1P)8U*G(H
ze13A>eK7Ig@BSlFnLyk7<A1A=d+N;3(8n7_pHthZoxcu$5p9+^y5Ic%%s;<y`VDgb
z?OybzYO%r@<a6Hl9yeOKRxD^G&c;!A?0Jb>dB)w1ZHwvw{Ct1enptk7Y4j5R*%2XC
z9@SwXsW)xqTOZ(Fu==ffPuQN)1E=^%tzI4=-RYDozg|qGkR5#=4~G^$Gjgdsr63&`
zfPSPUj6Cx!YB^hAjx4Lzc;$hPoQ@CO!6SPy(skufed>}a08=L#2kpTA;_r)kuz+zz
zo5pg_JzoqGU<U7IR=_8U^G#TbLuEI@cuM+BmHz4foVL}olad?Z*da&sdL6BPwI@ej
z`2R(P&klszx7#{bR6ZvNMn^XF6oeVFTXVsKnA-_Qli`2YbJ+#IW5`+Ba8scEWN&9_
zm*~6JCUW*hzzpkWrzY56BZarq6FFaecqKkN(-IpjpzvYA{i4|A#fJo}%cTV0M($Dz
z?^X||Oic~xT06rVwcO7T|HPN`l%Mu@W?s2n)n&Tex$ak1JazZ)T7-gHLU4vv%{)I6
z#}RE9sZ6$`AY=_IEYWIBoXXalSr(<>De*Uv9I6(^1Qn|xqr?tu=-LGqU&#rQp;MNx
zB|2-9%i<ZbcxvMlX&=K*sS8U2QlLZ##$CJ8tjfS|J%hOS;pJdefCOVZ6jq2kaInWh
z6FGf6B)u$#cAm|YLfRtUt|AMcHs$)-PF9<gA-{Z@6Y}|6#tzKhe2JS-h_>pO*KYAl
zEP;EEHa*W2;+MhTfO(bgq9)PT-lqPIs>?07-sF*gMto@sUJJaT=&X43(|;!JK^Og_
z0<4lupBplw+yWXNdb7Un1J8{#MBiAE2i$x2Hu1PmTIkPN-`_LwzaJ4duW1nmcCB9=
zp@6G*gsy!&>3hJtj&G?1?8ZCRTM+IB0qyznOsu8AGR^jxnTj9A<i^&R72a37oLd-_
zKiZ_XKDc?=HIVO&Z8WXKO&`N5fFiUOg+FVrtykcxMwhcMw|LSEntjLXU@`{AS1a_4
z3*+y+pfA{7n(I4QaLoF?(u@uk_-l1|5FvwblxrwZ>14PML}}1aVcD59Q~GlkL$SHY
zP^J?!Zk0%Pv{=lcn=tpK7sumLRo?&iS^U=%-Qf2vJ31zTGP~rGxAX#r!Y3tDcg9al
zzL8^BAlzi4d@fMR+N}rf)C!Zitf>|jYFuS%dR`vC%WJ9@xqa)uK4#Oxzn+<;`z*xu
zcDKX3#X$$r4*R?x70y}Qs|55rLPEgDyrI{Dj@B$5bWSMNky&w3CKBa9J5dPC>ryJ4
z;J>s%8OS&rcO@c0y2Y0MaEH6)s?KFut=P!KLyODyqq{BbS?ACDc@M#b-&I+s8OPmq
zs5X3oJEO|;)vyyHx4&Vz@_eMsUt37h_n+RJe;?Fe)0pY#^4k>T6w`mb(Q}%APrKZ-
z&XXU>RDXti3ADy29;esdp>6mt*K5C4&+;DM4BE^@TJgMl3aPXpLG|yJ+&I1;e&N4P
z@c*yh|6M7UCoJgk)yKiDPoxij%{R~CPnIg*q8PINlp9EULc54L@%vV9f}TQt-v{_K
z<39MBtQ&O@{u1Ze%-tE{<J)_$Z?AgS<l-gHeSnQZi35)t@{`Ka=HBrGd^~;V;?;ZH
z%_q*AgLgKk350)$;d3#^xZOcP3D$?A#1VIUK3|#cfv-KW2ZJJ;Bd?WBvvFJZ!@v7m
z=ldHV<VQbLE*wb^2Wu)Vp9Q@ZLXHGB@))3i9^Q{ZSW1mgO-)DV@uS|v`6aryZ-M~z
zdPEfxY+BUO&5@O0Ug*L4wg2I@ahzeL^#!8!c-7=M(cni{X{G+MDO)hm-n>tA-#X|e
zWe-q4a>dx+`se@CVR#g58=fk<?|ORjoC`5N<j+KIiIu@^Mf=yv)q0IvlVM(os*&+q
z<Y{kAR@n0iv`dO|ifJyiR691`+n~#8eR4vd+U5Crd`k6zaNykG)R&WR>cfzn=P#+r
zOMj2tKZ5Xdz~ci~V3;hkD@UBwi8P?aRqu65)i7=EBLu7GVm@rv9IjcTtzr6R=1aJ7
zI#`Rv<@{nH{`q7DMs<9w_>D<uSh{&<0*{MA5R80+r%d48*Q*{J!l+3N#UbU90+~+5
zdeph|pDMX<y-m5{?}w#XO^;<BB8jJYHOe22kjHlj6~k%T`Y>0pZ5EeLUnpVDy~HX3
z-B~vJcx~zS3_b0iz>$(fKl~;3y6yb9GFK;)(0|L0STVky^W>RjQD8%zR)Uzesi~pu
zUDyyi#U3|)sFM&+oB&6F@MufVu6!a4<FC>O=<UUVG^B-3N`+6$fSnFX3#u5z!t5z<
zS|8!**CGeGA;bw63(%aciNZOrm(WJ#b6<2-d1k97#pv&$8$h|Z)4JLUU!Gar^Bk~b
zPoGG|Sk^tyluoPJ4wEqM@$5f4cxU=$u5>kN*BF$kJmBcc%naOLzGN&9QmIcdzL52<
z3t+p9@o?9E{bc`YQdI~w(&pJya}!_?)CuG6Y%Rnip%=6D+EQ5GFq~4`Xqtc7qr)S*
zG;!9T=laz*WRCf`1e>f%v0S%avJEMa<QimcQ?=({-8b$&kR()dAwe<DyE{~-9{S)x
zCgk3ghlKsuST$JXkkAINKxaX8>@`ZTk{*TU8E|8^d*Bt}Zk8b-3p-Q$Y|gC1+#uYG
z32#l^j?pc(SQHg^ESa!BJ5m27>8bL}7!RLxnAz<(k4|fA7fkK!Y(SmxukEcFg@Woc
z!%iG7-S>A26bVnK5|6vKhIP;L@WM$=T{hg*BYw%QD|*qV&7PWfSeH@8F;<Ydvq4PZ
zQC7YU(+B(5=c7lfO89m3Wk{3qgNufpGetJ%+siwM&N+2F77sN9Mio?bU(Qu5O7WQh
z1N&>?$<y(PvDvK?H7xDXU#cCN!*LFz+m-4NJ*lONOQ#>!mXsJ(T9~R@3WN9uSluql
zXg9{TzLtntuvgolm^nzFLuJ8jE)Dc#vC8@QVBqU<3E8-;MECrKo%o%Ivf)DI2!u81
zly*h-+1}**nta@qQCfNDkFf*rcsF1)w&BLvafEil-KR00Dos?rA5e!3l4{jaHxC6?
zb}SF+jH-os$d->wY*l)kbNs$J`*5vnUe~hAR)wfvMn6|sv1qk$s&=tV5Sq>uvddwN
z@$-v#-gTl3XxM4w1kC!pcn$vUH^i)s7s{i*U|H~4QA8FAR+yR2SYVJVX-dw!ql8$M
z*}LC|L91G4gnDLY(o>=ul&=3h>VvECPWlV|?p6>*?7q+-Tm&#;^MVO+%yQ5Y$d{?x
zXJp7!F!wC-;*-K|8`^qJ{`CH-p+6C(I0jK}w|#}Ew~fBiCubQj>QuK!N#S*JInfd8
z>Df)Hs9SC9Q&o9Ji$$HfqbdZ42(Mr!o*Q+3>72XqK|`7Rw%)Wne6C!-Nti6#jhO(Z
zU3Yt%|5veR$v#=acXaVA@a@L=7u5V0SL=TjIHokj3KRc6Pvz|7Q+p^;uvzf#wRMF&
z%spelPJrhCI;+Y04BqZynkv(sCdY^Q$*!DonenRkZv^aVX_3vkjV~O_apqvB8xYOl
z8UP(vKN6X|liI?os+3hd<d_qiH|O8>fE$>j5QyY!n4L)jR$41`mxg{=amuSK9h)ob
zH{&V0OYYnvr_*`9cSC0_r&x14jg2Gp?5NffHC4SgN1FMaV_OYBC1HMQ9SWL{q=9gG
zf6WI1R+7yb=W`R2lQZCPQf`;5Z9pzJ?R0dmQX6iUa>Ad2t}h@OISkX&sxmA&?a<wW
z28a0ihV^Yj7Z53Nd!=Myx-ZCbwvaV<zOHV62zA1lmXSG!TEHAOeh$7g;Y<4R8q3!o
z%A(j2f}|G{zoN{YbTYVWtjmSeC%9gEW%SF&okd=pzT9!h;x%#_M&9iftTzE|U6ALm
zJ_U+(bA*(?B448k=EOu=VqM<2c`9(9Z?UR?yO6U0>)rNW7p~8l*{ai$Jk?l1T<IgZ
zS>(E4pxK~Q3(sF}l~;hY&1tMoL*mE;j^Nm1d!XUv*)v$_Qgm&;nbxRCz>%qM%I@x{
z)Ps0%mM|NadGkBj0V1XRmsU%^J%Kkh7KYXb7-YCn_;g3UJw0|>f56f|x&G^EsY!vM
z*kf)N;!&%Kg0);`j(J8-*$oTfw)pS(J^@*;Wg`$k@g4JFPok%fks#qD6U!Sk<8Ss<
z=gw25KiyGt*H-X-neod3uq3883HP@7^8Gp0x;uOUXv}@Es@xQlG-T1=<kNT9{Fd<3
zBD4QFf=C|>6TbyVv)aXc%d9qawOz}j#dQHqm1m8mj{fz>r!NP2V~^|I15Cbh$qd?D
zF<oR>2xgQ$GBJK=x0?hlsP4%>jH$aW)>hRwb16T|kxE(~u(>%ip?<2yH5AGz%q#75
z6w&#^<XKp*SDhqEd_Y~SgzQLdj5@AB48(2aU!>n&;)cx+#@Dk;UZ0n}0Ejvj`}Rxi
z=~DQD4&1ir!99l4qH@wG#^zGv;v?0iOUmgCv4qkG<?MDormSa%V&j8je+9aNoN^ol
zwj*e-hu25SFftNm0-5iEr&C1Uf?rEM+o6zWG3~9#290~3_x^AZ<!s8>TlY`=F(&tG
zYq#D!_jU6l^xQ!n#k;(9mz%nPZ9x$~(1*4|x(8YFIj8RSgBwxxck1LqqPf2vZs4sp
zEP%M+hh;y{&F@bozo<+#v)nAYR<-FY2dlG??y%&#T>LFG+}1N?mYPy7uOu_Lcn0tp
zXUbu2$rEP~hrkxYR<JKBbw3QI{QbiJM)B2IS`cl_sTs4@QGpX%yzZxZD?u+-=D_N?
zwfNxi<b-?HB~>2SC=D<a%s`&|N>Fg`Jfe0lWXvRGZmV@;!`11`7du_S@2ok7N`l>g
z!buRIBJjlfVR<UdskC8YO><(bc=fHH`41_I{5Q$AfBkVG1${~A*x2=w7-AnSMiVpQ
z`jA%+tk`5q@8?kR^GWL79JgY=jkVZ5j0ZNa6FnZ~``onA50Pkjg3iiCZb{n-goUy1
zanM1j4uq&yP|S{pq&<l-#~X$UYV<;M&3CM}uvZv+{kUcqn)_Un*-))t#g|UI%;vQ~
ziBAk5S+jb|%I0Bj(efm*%%b7gYkvV4CV@oIj-P(|anL>5Q=gsqxkzB#dP*u1L3ORl
zd1Bar<b+l;CJJwrOr(vus<?3ICw)`_P1GSso069wGu=M7f;^{rmykOu#rm;Q-$Rj5
zlP<el`ar9s)bgY|gl4;g+BEdoTne0jaMShIKmr*!>(P2`4@@8WbKa1=lr$7-$?tJ5
zD6jBZb$?P^%p}t9rbG%bjh{E3<})v|rI6S--qwwLX63CGbM$fNdu8k8V?%TX_tHa=
zZn9DFgc=sozIUg3;xaT3TAeJ@&2PKo*0R`b^Tq@^8DFPfCt0G1z*M40uh7Lm#BJL`
zP{CRDZY?rP(=wGG${GT5Vs}5g$=7Ye!@yN3JJK!-^gb}+Uw>5fR7@q|LrVvLVO$mw
zHn{m4SWZ^^oF2<h!&}O@g8A^vIO&06JTVi~0=9bu^xIV1jUEw}$*v1b9p?gqgF9wS
z=X`Fz8ev>W`KKXPBmT<bnH<(@C#<|+n3AQ{JWwUjlc&F4DK{#P-^IjeI+}b05Nvex
zY7H%?2)Mj-jnB0yX-|vDvs_-`M|s$vkj0%=9C%xpS35xVtgf%gdl#9Rlr;yIT_MX&
z{SsJpnkm(Z_w_M-TcZ_B`Djms4X4<WWXuKgGoblqF_rzG%N9|c(9u`tkR$jFBl>Qx
zFQO=L4?ZsNM}k#Stjwofk&4?7k+&!Ny7J2V6wj6gl1mH`wGvm?A2CWNbi{X*_&?-T
z0NPddF&+0+f<?FYA4t7)OF3Vk!he~23?Lj&XJz+xOid(33%wdd_<vRvs9W=KCWn~#
zS$_HjW+Xi0nYO6T%)9i|8>a9|m8Vjm9Mn+gf9lp!e!Zx4kNVWCIx9Wu@Lzw-49xZ1
z(msC>*2YrrvC66AmMOX8Qnfd_5rR>In20k!8#+;BXU*vt9s_u0!IN5mU|t&k;0R)^
zGO)13%sIX`%c9XF`c6dRJTP2b`}*Gfu9vhMNThND*oFMKI_jsx)K1+;)!6!{9*GWr
z^aXP`3*1+V_x~I=n(|QN0d=lbVQ;-+#e=uTG(KTbg}wJOLe`kl-zxfSUA)dX^NxC*
zeU*K}_nMst<(x_tKAwe==&Pp;iV|W#YF=tb4$#c}GT8!_6IVxr$*Z@p5&R3dEV15%
zi+^`IDlsnoM&HLvqIcOlFiwqm20_}rek6FLJHL=9^E9sK*P3k#O9KEd{j?5wFd*yL
zlK6+m*8Sk%+Si(zYw<JUJ|FzARqQdYOjdW!3l=WIpH$T1?rwYx`C<2|a?pLEN<Rl}
z`sJNn1FkC5M6LX$wQe^1HuKm6$Fu%7Bu~3*uc=*3s(sQfSC}mS%}H@T!?mm0y3#gP
zCSQX~QsYHl6lvMjYX>!_Q1nj^OSC&DLC^+kJ)^c}JSmea;OKg%jt}nY{#*|p0gfRt
zr(?FM&9*b|q_s}dUL*ho&&Fu>V2vL0dKGAwjQRwQ`zWLr>%-f)lrw2o(*i`Eq8|^f
zi{iS)uf3b6XIz%hEd??EHem#eC*N%Um=_mYi&7>c(C3S??X((A%(W8EqNE+xs?RjD
zhb)}ZWzlm#&urz~*-^Q@4X*lz%|Ib^X4HNhmqguBZ!aF7;vK4#zFKQn`zfM+BX0KD
zpp2w*P-_(B1MO>KQ&O-kjtRP5l_iA)A5)1|jT1K%^8$&Wl?s9k$k_e+)t7TEy?)aI
zkkVt_=QW)Hvo%IX7so74cdE_|Wo?U>&P&H^DQ2{%Rg~LXmEK)0leJe;$w)N*Wa)C!
zAeLWv5<j2?P!38ESho^IfFR;zliL-F<<NRqVYY47X?^?)8B|GkRE<jq+%+r39JS!p
zEaKqaBQ5wa_Iucy1fE|sA^4&D{o#u&n>ymOVV6u1TNkU73k*-X>t;l{*70e$o*Ug2
z8YAdF#dQvsJz7?LlRL0EPCjP<dSubHZ)5aOgn#3lunW|*!ddZU2Qh2V#FAvJ?Q(*g
z%^8%H7gg)5Q1D_=A&10h)Uk_|a-F#L^>n^kl1bh}VS?nf#NZa5>t%^&0BI?8Sg?eC
zFu6Lf&z+7c4(WQojl0kS3gFw#w~Tz+N25AGTDK0yQ7&b{?iG=C&7>e<5ERX_r_!zv
z$6a}M7+X~Fg=I=tRg%$Ri_yRSh$GaW7JXwYdEUD6a|=dIR5E6rdAh9~RMKX`aukL_
z(Bg(BQ)RX$V(cIS0N_~<yQ{8MK}n}wXIZV5!JHh(w*0CQb3GgDoM}b*xSU#Na{lZu
zf`S;qd!={2yQg0aFt$y;k!DA`xZ4GsYTTe%xvVx%^-Vq>cQGvRqA>a|(0`yS!@$L7
zplQYB@{;Td8WV9I<(oOludOm|swt30__J)=B!q!oXy3QJ3-E0?96Kcn*#g6Z5E~K6
zE4}RzFEb0yWba(*A83ywHJXR`=JeP-sHjH_v~tO#e{rji%P4z-FRp8XcLTsP6HW_V
zAu%80H16u}dY2KUN&}utj!BiM+MIx02K-8tdChm(9nZe<1gt^RJ^}<`7`{ncMU@^=
z`kcOdIG;}?KCnjxYs1gqV%_o@kf`QBl)BO-YF52okqnHK9kb`WHOx8_3qOF{VhJl8
z?`RIdub4pwnomC?$Wf{1(5YO%kUM|roPWcTr$>*IN4b4cQ&i(itX?`KgIz9r22!p7
z>B@1b)vhtoMu?Y8R!oTqNEu|M5>43^dVR;N95>jB3)O~(0o8M>51{2Iq?p=|ko?(-
zotcwoJXh)wQnmo`db<ZQ_32y9Tw0%Qw#~CHdbnC#a6HWV>kPY>R8#4F2Lkf`V(EEv
zc%yy`d*j50K04m@e2d(MpF3ahY$yS7uxKAd%T;_m3Q2#>^ARO#Ra*Tp25t3ifoW#=
zq!!MRH}egmCdmDaJX*Q&1ud*H+$^qV7l%jlNYGe`E!Fhe35M8Hzm*$}M$d90O-(|9
zlpEgKbtOHzR|d@6lK7EthroqZJsA2%P@M=4<s&)8E#cmHE$4BpdYu*CX6o<7u0j`J
z<%9Xp8^o8p;T*3GXv5EckUDd!Zv*GoWtBESW!L5(rTJXjLR(?J-7GGELED7yk6EZu
z89H(~`X-te`4IOMPN8J%q0e379g)vBLW7&hVmj{9l8SP2tJSvnDaka)0D4oAo(nmq
zmP;8XSPutYLh9K`=vWF5S{R#|wi&$|sOpLXFox?k%co|AYu*29I~<AOUnN}qmJvVo
zi%)^e%ge6LJA6Ye0%nj$qJ-)COR?IlmO3gsvwze_O_X)OR*XWuJu|hVh2&n%qmMre
z9pJ{5qxOlB*X9pY<L-6MTm*bL(vkHov$Hd@#stoMsU!YJtwAV4wz?3_f^#i!oGV%I
zre4!xLcH%!WEdByE^xCUbGg5xN=3@x=?NVP1BAY@XH78K2Ht2&tUn#{<!^6S8Oz$G
zbDvPtiHVl2UUr%G((r-d<yd2L^lWOO=O&K_T&4qc(_mn3APsQPyD#ZA*gLTFxE`U~
zailcyDpBoee2IBSN0Sw=DHHQCEjw^~$z6P4up?}szjUT{5o?H%1ne+8B9aGP^9)uV
zpS7u)+>hB{MJ24Pq$?UkbGazYEKjrd^-k8?=*Uywj!>RxO}-gid>^B*M0*``t{to0
z;*=XIv(UMekJPmC8c5GL*C{bkJh9670N>=>^UcX=PjAzv$s0pcYbavOXZXwY-GtU>
zff=QVmvj;q7ad?`<^AKiunvHmRH1;?g|SUo`t{#pjlgkrW)Oyasd^u)GQ|j*Gh&H#
zhxiDNZUxDyt@idBrSLe9ddvy{)ul2APj9lrMs*CKFdS7FyQ%{Xl^$vlZEp2K*HW&k
z51w>AHB+slV`!uM*|)7Y?XDl9ry2W82UrOs-p1OyB-zO=T$gQJ$>*YtF&eZ}2)}yr
ztlPu*4id(8GeDgt-)!%3RVYCVpl%r?-k~3Be=Q<U5}ajbQWpfia+H<2*zIh;uQqNj
zReW_c4nI*Ee+`a`hI(DRGoqyP!YhP$n&+5?fcPsn7#c!YoG%T%?HEIEoo!I3+pFi@
zCo~btKos<C2jN;?lFz$TESd%aZ!k-rAsa9a4B3R9vYzr`D|dK?;Ie#kQaR^GVrSD6
zwaa5ExSQS!C(t9a$Gq!ne-67e&193|?R)8|n@!$I+9vnMvEyJNTAcbuRaK8v!<hsw
zMgsL%Wh1;OYwS^6?IYf96)@=~FkH%gW$i;};LAPNuBn{NgwvObONB{!A5AG=Oo`bP
z(i~`0D62n6pR=yZYdpV8_T#ZKy?uA<;C-agAxHZDSHxL=1>VgU{kMw`v`>oa2eC`1
zG$nD46#weTL$?Y4HPKL!8u;|LU~{{DCO-9WTz;20bi86AN_K}y+Y9higKe))KYG1B
zAl<yvs^A<(z1j|w+ef{L*dp;*$z=f2Pu}Q1(P^8&0&0O@TiPk*CjLy%zfcd(zCMcZ
zy9h??JlPdCZ8xRP#N_MntXKoPf+ndV>OqS0F7XQzcKULBq~`03<?@*fFP>RsX0&AU
zjkiv$qw@z*qvbbGcy9Xe_S6ixvINBA`@6;BOf!Elo<TcR0=T^fR$V2!jhXUpZc6se
zq(eur>enLtzGM6+Z}}tyMK6uLXUt$e1A+9b2k3-M@W5>2J6+)-<P*{h5!rn&KgLY&
zWL>mCtL)sL$u*-_nY(cD3VRZ8K{gdvrselL>WgP)B8~ELKHA72B&dE<TY<bd`jW<#
zrOYhpeR`~Y0Qv13;m2bi>x&2=3zcy*%u5cXdq!YIRhdmN=&M~8$=t<D9v(H5zMXK&
z?ym*8z4&!je;oTO!sck8vS=gr?<qg3Xyt^Q7a?M@)%)ZRgWTlwZIcjx@Yx4IUF6p5
zhTLe>oZ6KAa*(9!R9C^n{wdI6IWs@Ozdv}i`PxvzqiF?M9+K)Be{VuX3e0ZM`$?eP
znYdjyyD+7-B9DsoffDIFgwwYlw;#R;1n*ofo#%?#$Aa-olk0-RO8(DTOb%(8zuX)R
zo~2@+cSm%(C0u+fBQA@=KP$DH?(e~`2S8@>PflU#nZ5pIks4-RWtW5Uy*jdi=C>0W
zLl6x(BX!}Zrcr6_ZE*K?V(0Zq+Ynj9etsIR-l4&Ji%JYF{D`Mj0g^K;OD9~$`%jR^
zvF1&O$snn>W~dpY{Q{U`qE^=G_$}-hy52L}$cL7ol=a2WpEZNsDxM&k8kXIl047xU
zh5B-%lDi>KXKKYEs8Xh5^;UU$&qp-jFB`v?(L7Gk-0;tC9x_X2;gbXIV?x3w<-8_s
zz^sAlE{40?Na(tVh`OfBFF2<8Nk)MPW+F`K<DCw5#mgbz(hx-ry^YK7EF4~K@W0vt
zm`lx7kN&v3BAQv<Z=`1~?4WU-{Q1H`^_WIOb#mOwzPswZ6yCpwmNc1}lra}eST*%_
z2WRS`dql_rgG^L?NRp{R#c>5p4}4Ptm2MkGE-qf2F75}><O1mf;nfo;@4Dn?wygCe
z?+U96Uws~WedDFyag9-4ttclnehI}P*;!`q7sG8MKd$j$M)@jtWw-g~ZmjaZ`RZp;
zt~7>ytL|q3y22ExkTo|C4E>?=leBNX&6F+A(KgSmdGiRA+XnLNGbrgKm*(n4$LHRt
z2ljpnKV?YnJ1}M}lqj0#<u0hM|LYHcD?TS^QqVsIiUlU+9TF=B>`)7LY_;=-f?LOt
z$}|euZ84nM^Ds3eGHXCm1TFN2j_cWafMcSqt4nJ?OEjx%qpkZ5kziXAN8<`i7`H@_
zk-cSUpIQ(3PH&LP%=rAQ^D4_8a%0>!A3}2!8bmDhJBRT<dIm=_#rJcPa;he!WZcvL
zx-6{H?b1Prqb?&hfncTFpDf`w%x!?A1A$rxPm^6;UNSAK?l+UD;}ooBGc{YW8Np@f
zluv#`&2Ytse9ql=Y9~LFLoDQUm3lkBB=Jw$yppm0U=HtH8L_TX^BwaAHj4W_98BZO
zq>H&1<wqoy+R1xtf9%+r`&<9fu<*?u+W-?Q+vHQG7hDDT8R4D{a#^0-X|V?BY5v{u
zeRz3}t1oZi3-f4jJJQyg>!*gE(g^b;GG8uBR7>_vM#U?H%HR;)v5y~B(q({!Cr^2o
zPJH$2-9hyxWoQW4`5!)>)oP~8jks`Og%mEDpF%*S9Q-%jQT=s_Py@9-|98&}@VP7Y
zrPE2C!3yb6t!~RQd79dsQ{0bRy3f@@cZABT6%qq+Vl`q3#$!w!&SC9(vxh);VEPP!
zk10VH{kto%_DXm;(%fE16kjVK34UxiJ)oxhbW2l#ds3d(ZCZyx(S*ZzOpQ!p6#wq+
zHKjT1W4(TjJ4w%Zr4K|5C*s^^f^yCJR{i7fLKgIGh;P3PfHgI^?37ceE$y(&3hD_R
zAra~1zUm=OBMZ`4Tvi3rHNX5yzWX`T_<;JiZI4(iw(c-WxpQJez=!uujk)X5szg-(
zjmvS?SwHBHQevgb=Qpwk+YAzGH!_UBAxRVrh&52-_Q_m%y@lUZ5q5%N7%lPLI55;i
zR&j92eKGhG%A=$e5Z?oAw7jmP{hc!5m#&*FqF6XUg^N26qj5<4Z=@l1#qp^k$5On8
z6zgUazu`o3qDfc+tnBvZbW@(Nz)Y2(n}3)24OvxMJZx5DK2bZav`f0;d8A~aE|w5|
zZRb)Ad1ApV|FHl6BJM4t+UmM?Vd^auS}3k9?i6>Z;BE;J2owvN;BKW9x8e@PA;AN|
zp~aoz8d^eecc&-4&-0x3IcJRXy?@RaU;bod@0Gddnk)Ot+H=iyUAC4))}(+@Q}}z?
z$RUNaKR6ZQ^m?bj@YgN6k4=O$x41s`O--;s08Gn_9=-{#K_xe)f+eMm5iTfw^4vMc
z$`~QhfT5M0um%g%{zfD{!T`ld$>46pq7VXN52Bqc(nbVdkp$Zy$3`c~;~a<j5K3ui
zZENU5Mefrpk}Ef>7jM0AI=n=&%uMA3j0KYi65A7EK&l9>I0X{W?}h8!YA8}8c30nH
zP%lc3Id6BqtlTSEiITbN97(PYEO}nC7n_pwB?uqFl75%Fra7OLmj_>OMP}M&);<ja
zv4rdn_6yznJ^r||g7-8?^4c6l)5!!Lo+m>O35C`G52TLjp?`3Q;uYcAFGZ|T>^0_n
z>o1WFoZ0VN82TKirChH}H*}wDAm=#fHu&##W=93{QoWA1u3~pmD;g2CI_DtJDDt+J
zm!fa4x#6>Zh2GtH#VC?$J1aS!ry{M{KEoaJ9*)AZ{GF6rHGh{~?sL|WPG))#f{>7K
zI*^vKi8Iq}_q9RIqYp;TjtCKv>2UHpFp~(8w{Ae_22fj<l$D#w`uVR};3V@`?Ys$$
zj->#y>`e)k1-<R25PkuZB2)oNr|z5J|D4A%^SQT_)x6w!zdKobXT4D-FbUpKWD<%T
zQ;iS};}BF}y9-4+jPd(UdEITmt}ujFK)mXV-*4c(pBW)CV_$)#KFrvlI6pjKni9Br
zB@1kiOZvFuVT%ty%)RrC+j6Pi6C&un9k%6p?35Yc0gXaM5MWxeE#&ima%<Ql%2jDx
zYt*xj!av@KRo%rab$P~{oA*9?@$|sOD>n5y?KHXP+*=9cG3NHAGH(@QvY;DgI?2Y9
zov6>&qtprnz5tp6M!dd<RXjN}ZJE`#Po43>lh62SLWxiL^~=1tdaPY@|J%@5lYAL0
zhb6Y$QobM~b(X)O3}r7tCvk%Dv#L5Gd>y^mp0wY7+@;*TTsaErV(dZCPr;`c`|}EO
zJJ(-81UYl+f2xJ?{gz1FR1>SUG24cGlqQUUpsClDWj&q|shbJu?ngiDn858|tSq%|
zkegtuIp-VaN={YmX7o@biI$U<6QoVeWg4zorol7wnqu!H@gx_Rj4z(FWMN9;diw>#
zNq&%V$}v+pskZ}AMr~OLg%8la-JgEOH=o?}h^S1EgRs=+S-bbUKR96av!?6<_$jKC
z{Q>1BKy*E^zKZrEe4A|ze`YkV$4OcarDu&)wmwfTjMP@@5i=d@7W&#KzV6YX=&?sU
znl1Riq7XatBQHb%i^L#c!<7J<F=q#$0Eu%Fv8*%60K&PZ-YA<_u|vcyStwGm5sDkD
zo~8hs4vr^sx$yMp?a8H=3?rhd1)EPPb0G!Adw3o>Tgfg(rF6LK3p{<Ln(Bf{db!{h
zZizu~sjF5QG})Yp-BE?f@?PK%u<H~DF+FQN>h<>J#xfQuOE;zQT`T*hRF+2_`vz@}
zpQCFHg+IM6c$h!y!$2wUaDw%i0@nxgHZs0Up)w9)ME^cK%0>o)C{3=KXxt5~T6x!?
zpPd|_tnay5Cg*gwHMhIv!{AVN)Dy$SAy;tm>6XxzMWf*Y6+WM4AnKGxPk<wQ?29`-
zX@6}A$)sE}?2Lg1lKwt`pxAM0(iN5TeaPZANL66o#@^+}XU0CJTYugtHDFyt6#yb@
z%9&_Pq|5GPOOlxZNX#x7QXME&uz{C`R_0b24U{9w=gXoo+An1i3&&T?3~H0&5(3Bo
zuIY1c0|=<LcQ#5VtM^ofc-e0WxX0LZ1=Vth0s@Uzd&`H?$&<qmp8O<o`?|ax7YA2L
zP;<dmJD?nrDN~rR?DWsI@y7&gS8>V4eTmTM0XO3s*5u2;2qH^&a+LV+w)cFi>%a&L
zlyzowQuhy`FLeB*93Ql5>1Ym!PK1mBqa=OFmZ*VBH0ZEzl>lJ3Rh}qcv&denHdj8r
z0lArWV~Q&tEa;DVjMmzUct>@R3=$^JXJgD%Rn#QJvbAI^NQ&Q04`p?wC^JtgnqrGb
zQbtYrn{bHVm4M2j1v6(m+^|G3228_SYVAIrqV=zLDkwn`%wW5;gIAJKjl0Tlusc*>
zq@CxHfmKvl1i6VbTIVU9xamTtID3l|u>>G`8~JsNTgl>Gr6vf1EdWv!^YNs)2NfHS
zaCc>O#O2fU&nfjiPrc&7aN<S1z7g`L7Rv*-dql^)oSAtiz{(-6oj0tgl&^;nD<Pol
ziSG2lK6DE0R?qdQtY6~wR0C`be|r$u1hQ$yrAC&O@D3|zFq_DR!52qP#5E(#7aOR&
zR~Z;*-2E*6;Lvc|;!ZLC)XgHmKuav(d2S2a+T(79KmaTjM!oiXicuay0c3tJs@B@E
zjq1KcTXYhAsR(Ff?sQq8DNv=rY!2k{#RSy^1Uk~_kNrrJ=&-JlkXJ1k5xajMSZdYB
z!9im3PF~PdlCo2$Wq8q)jrBxA8eI67$TLT-q)#~yUECET%tU*i*!yVGrZW_6sqd<_
zRX<6_F7+(hw{n2L!K>3wt!_Wn!_jnEjuJhMu@=c+NXsq5^V-Xwgb@|Vhp9!dT@)!r
zaCD`{FpCdM)(0;_A<8@rAm?j&-?XaRpyP^w%HvZ3Uw3k=J#rH`Kl6URwZ5*-mJY?G
zO9X@&=(uqgcX05%LRDh9f?d-j>6BQ(=VO*=*7NLV0Lu^KL|W;f!V#5Hu|Rqm1tO$F
zo-B|bsp`%^-dH9p3+N*uo;?%u#I3DzXBW8$)_ZH)Bge9MZf{(-?F!uENMd3VzQ3v!
z%Ayf3guv3C)b(D<Gs^nr@}QvaF?+~z(XZ>{Y9Q{Onam@N{ZpcDV)!KZkZ(X(XqG}D
ze0_k2bDGzzOU8H{pB4>cb@QeigKX7nwaI@`fMpbqLzpb$IIcu&N$KeY=?FD7qcB@s
zQTB%~D*46rn9{eOcB$Cc?KgO6L%>>_y1A_6*~^JR+B9_t4N6u*>;12To<Hf5pH3t(
zRkp_X%H=7tv0t9-$ew@p7Pq#115z1XB+Tq4<Yi{dee)&1IwuIzrzmDh1IiTn9V4(&
zANHUnvIZc}L33Ga(7G}G^g&Lubx4&jZmnjZs$HTKbo6hDO{^&csgZ!dMDUkU!^RPq
zu9|q%M^x1IUSLe>>+WX8Z+NCzi`?h~&ZN?lh2*l1hv90Y2kb-j9JG?$1)9a})-3i~
zHotq+@iB(+7N|agG6IGb`)Z;1eN-^E4#PRW)RxGLl)QtVJeo<1Nl114s0^z92t&zP
zVZu7yq1TVZKH7UR%$Gbdej#9yTuCYSb%Iw2uk6PL^z+8Wsv={PS(n1=#1A_8tGTw=
zQYHHn4SII-I@k2(@56(cpD)@5P)_?Dq(}<b5;8>|Htu*=vOB`;1$8knTGsj5E{Nta
zrmLX4aLADfT&&&vVJ>BsURtT{N%t6P@)$ACKR@tB%;D6l+e_#hGNoFGxm?Xe@9_C%
ziVZezqp#4V<JG)oi_P1V;A^QViFL@DZgt8bnOw(5=lDjMU$@r}ry3yM6`|ySByVFg
z?|R1O<_?apTIASt$O(cys`R2q7_kX3l3H;{{zBgh(eL(!8Utdt1Fam7nG=xLE&|h}
zmY2vJrN`zP=cEdB)uJ>K3A+3QAXEmO3)OIhUR-S}7OIea^_3hw4vE1`d|m3}ln%45
z`l<NERmqXrCgCJ0QLZ0_-5yBF{s%|z#$^KLro8~t?hsTl_zba{aQo)7t=~t!ZP#fY
z@J!1{s9CvF#|f)OrfgZ+mDHCRs-HhVV3XC^{A0e}YnOuODfgA*lbR}KS*5bQID|$W
zT>(XRKEF~jCchImL}e|?MNt!_Yz@gz^RC;O-f8$cVny?);~PjKy^FBwo@2q?`clNw
zdP1MNLy^*Cj!H#yD}>#x_@Y<r%L(_fF4wrOG|(ztHU;?zQv}%3$fsbgY3NfPP(w+m
z%?T)4@`?+X=>mUbx*K!WrxMQHRBvt!$?4l|i}z<D-`Emb>!Iz81TBonCQVaS&eGMY
z4rpVvn9YGqfIM!;DcjROI5PbVOee3kfXQ+5u$<w7tGm`)^0&it7M<kF3J35XtN<>m
zDThwv`tkH6BncrI_Bu-Io-t={W#Si7V3b%zR2ih@Rh1@?p56O}^0{19)a$`a_Rv1}
zp>JjVBQ4Bk`OVVqK@3_SPvjAX;kRQpB<XgehIPFgMm^w&hOqu<2~+S^I+atWS6{tD
zoN5KEaJW+upughM;0bbt3@Zz`Xc@y;8U*@tsLG|5C3$NCZG9C$-W2Bn-ey*2bK-I$
z!njs*wnXv>P2S(s?<M+&9x^e5*#<jK`fVDC-LchDv2hSTusDrQ6UkoNdi6Ol#Kjt$
z-1I8aD1RSkUNiY>%pnN`MLyHp5@n+iOpCOsZOI_JEf-Q<g9!E6eeE=yH3`O_PgO#d
zxxvKMZpx;1$D1qq3)tC4nGJd#QoMJnJOLM(dtmiXbCOaW($AGZ@%!?HSamOdYkW#?
z;bqmA(*nt~wRhhWnRvM}r)_dEomdU0Cn6Ys>qOzlel2D7jkY?`wG}^}IfU|kNZS=E
z7%gB`N%XO1J3k^EvoT?s`+`wEg#sHuI~G%Osrvj<%2n}Z!@?HK>phK0Fs=dKIq2To
z%9pdWhX=T>`V+%B6R_m1taau{7gxU7Dy_ov#(cKpGMNpbq#+K$i!}_@j#;j>wHS9q
zzKaEaC__)XtBSU;t`eGviK#45HzD_JyuQ1X1PtAO@g#b*1W_E?HDbWi_hUGQm-ca5
zMOS%8yczpzw*8fsDQhHqd2*P*dfNI_ohBgz-$3TkFVCHt^j8NV!o~*1Q(-U=zOTAp
zsj&{7Uwwc^{+{>tW#ZNN89%yUO3YVs$FNltvxrUKyeQmglxyS}xE7ZzsnOmZza^J$
zcj@vnzO=o4_6LXYIxh0Ewe8sX(*7d3#YR4G_p&Wp`wDN%tu6v;G$7?G@PYb$w11xQ
zmyQJg75@zH=@wNX(BjpFu1QJhwCztcgDq$dWq(9#Zzk0DRIA^jD#m32ZQzEO#9PG5
zI#*+Pi5{pS$bbFVIMXJXnCOq(p~+|cS)wg=Um2#Z5S^SSKm=5}ZtE1DvzJ{Qb1frN
zx9T@-pndqF0F$cJIJNOCzoiO3?xHZ}w>#Y(HD9a$ig!+wq%k~FR(5J2`@INJ+b1U5
zecQ?vy~{{1A4PlDRJ(o-i=`nU?!T#<y2t_g#fyC@Rg}D$ngpcd#XbzvFIi<Oi`o&#
zt&?feb~iq>ss<ROI%KTdgiZoTbZN~xpUKvbN4?@WXAYG9EOE4dcG~n~59x<DhW|{_
zSugzlEn^q)T&-XYh+3(JMM-JQcbL696q~ns$O4Vmx0tfuY`73JAMIx9o)i<9#HCLE
zg<Xe8irLeKCOgUa^jJ;U8ur*Smru4T$Ig~Tr*rK26n~Inn$@pDafMVEgpe_!g*MCZ
zFh#nbfXF0hk)mCGo1#}jhtthe!NlO4V+akiYC(EoP{EVqpO^<bczG38XqFG#x_Nph
zp~q>=_X0zTGC>8ciLVn$O>tKghl1`Lm~kpd?Pw(Hgh$KhB4!bn+^woP3sj)*I+U4G
z^?eW(RRCJ_2~#1FNyG#5m6!Aly;GY_HJ|?Aj6;*aYf7lLZOy?}8>Fx$h8e&$C~Zkl
zBEn#&KdA$r3B94ucaIV~>$DkCROyBz0de*6@)bXx6pg=nT5=foxjGSrS=hF$<UvJ$
zZpo|>H2zL0pWgX{qn3sdSS5jEo$+G=0<4{9>>rQ)5<PmF`YAFRsbf9iLjwSID5y^v
zjrA>oj7Njnx>M5B5Hf#LHpGhhk4pchY`i<!HWS=)dG%I+Zq6E|TdVNjC^;FF^7cUa
zia&eIQkE4o9Irm5S)-X-fP1{f=0wT3rLjzE!8E~0P$^M4YFXwQ`{yy*y`hm6Ifb2^
zFrM(tSE!zo%{@|kesCJY0AMgV{nhvY-&nUuDp`2%cmv+7pKzut8{QJf=SGmPYAXro
zGoO?U%eCVGem@6HgpYsVbF_(%cT0+lT{`8>84!$^qZbsccrp#G9texThg0D@TcjE;
z8m1H_rMnYQwl@Z4nvGenH#)xRw=jfH*zn4frpK~7!kxzY1{l^!OO)P?c0GRQL9la0
zc;&oK_vZ3`-ue-pnrX2}wDNOPP=cFQ<!?|%nlgYLkDt9P&2ECxc?v7ht2ZPw;ijuS
znWl{`x$UZNgdpXEUS6<PJG~}Hz9<WLVPr&*FQaGErK1Xs@EPo~>z(a6X)+&w>X6Kv
z)WtSvP{-~OK)(#G@|uPdo5g@T@sb+e9ay{c9nZixivhTf3d}huXfMn2rFsmo*5~GQ
zaBS-Ncu0*9|Cri?Fpa2S+c*#4{m)AI@g?K%OF5FfF&1%sd71FaQ5MF$e8Fj>F-qPG
zhS(PiAij=TPL4pf1SRALloyce?LqqAbWIq)Mr!KcbWQP!o~B;;08=CEwSA54hbd>r
zGtu+u2#-7VCh;D!?t|5t$Wcd(%~xRDBprU~ykFIF6}u+;3w_N6fwZbh^+Atm@>$1+
z!AZfR$B0>~`Lo%*_eR5)qNC-4nO#4sq$mPHmnL70I}GYyFgF%O^5Wa*R;Jr62t-CO
zXeH^wc^hK^`$s^qFHJu5xf!Qw6o+fSA-i)vOE%+%yL^^8IULJ-=~N}SgrF{~(R(H{
zA~ZW|)cwu#@$7}nch0XLASBDa75E-y5YRYVU2ZX1W|vSjJiU&@##{a_K&t2=)!<b&
zH!a+h;R%cIn3W)tl5}G01W2YMO@DkhULfqVu_zfnCT*elhUw%9ATGnDSxi+-b%2+B
zty7O@ryw_1&}gfca%<J?-W^R#huvp6nl_40kP{cW?I{YpM4XJ?g$!M)3ZbFVs(Mkw
zeVOBqWw$QZbl1nGyS7zFFC0GoY|l3d+6yDO`qZOpPyB+YG~LU&&yc8CF8!_@%6ymb
zII1_{sze&)C@0n8E#^mWG)&3=0~S&FkTt7Y<AC5rJ8>fyI8Fq}xsYdR@v5OJ&d7p=
zJa&n`Vk0!tMm%8psclXq6x@|~+4-5$Hh)u)^+;=g^lPspDY2|9tQlLH)Z?gJ7M0DH
z7i&xR^spnjS+w&CjN9o=_VD%GSR08#Gt5(<7Cd#6r&wUfq83Y3LCWzX8bPYr?IvKg
z&6U|DFrdKq)(1SE#htf&BVcVzy|gyi8W<@M63NR&G44PLw4Ugn^zzE1-$-?JM&A`1
zP<gLc3E18K>4t_nBWKhy`Q<yvy%NR#=4mL?8Th^|#01B#SiE{09_dBTFc)JwJEe+U
zz<~>yT+>Q^bzhWTPCPKtcpW9gwGt_k>N;y(jLGBFd11YvTL!Cu5ucimn=3%9HCNxW
zsVaQA8ft3|F>*C8K<om;+K93thBUb?lqQsew?a@eJgYxA510B87L^HBhdU)2`zu1J
z5;KHjjwO`1s!Cl7nwn(_k&|A6JZ5l_khkZ%1)&^?6nez9N!g`3YZj9yq`X^qZQLY0
zlvX(ffej{|FeCr~s#E?AUHYug(VlMnWWRm3aYr|Lz0TpVKgTeHV84kv)b&!fPPJ%?
zfFn|u!lHLgGlu1>fvV8l99kQi^H7l|p#HRqDYR-zpp$qqeACyR6eHFdnD<^!GfAmP
zFC#NiuQCx}BBMvRgjhnm()|r#U5%$zwN9F0hKr3{LXjetY@r;f#recaBbNbuQ@B-S
z*EG(rDn9bPZ%z`}%-vW<8yIMzSh7$%7sQ#}WmW7^GR_{C-_8=(s5F)Aqu<>qgWd4W
zqotMLg%!!mDHT7@_U0UaeS@1WsL`Xyubwx~SV)*6#Xl3f!ua(~c%*-8A<|m`Ht8C#
z&xumb<j?#eB^@IwOi7KaYe@-_02<|Yf)!$oPSN6m)thjzFjc|iq2(v79qfQ}6>y;0
zh$$-b>=((x1hlh?>N~_Y#jp}D=k4*$HcD}d6GO~~k1mG3qzCdC#Ak<<&Afptt9Bw`
zyWPa|3}QozJ2D2sUccQz0>V-tPUc*WpLh00N$k>wgUe(9ORjU)A;Gr<exqIw>Gc5f
z;emlY_M`U7pfWIPN(GHpbc|~G+f;4VK1k9yJQkzU_M@Q4z+nN+C=z6qi+xZy?>G=W
zH7qmX)Rv1aHcQMznDnFhROy!<|5)OjxZ$$_g6U<>CMQh0!iU^vT@Bpz?OYmDzn1!P
zXcY~~1}JS86csRZO=9yk+rN53umZ@ltBqv_%qP7Ibkt7L!R~HPmPvBVe%M4U3Ij<n
z__fx=+izPb3$IqnL4sc3M?I7DOMdLHyi3@5qS$5JXOwAPHI7Hceeqx0lXKnlBN>y6
zsgl$;MJ0ytwh~w9pzo4FX=*Mdz24=BRDQ^<HSJQg_J|-L5*3%9O;%pmrSVb|d>5ks
zC0VYtbB8^T?GEh@kkdL<;fD!BM%|>n)hT-FhAnGZJEPn~it5el5q?qE&d9g9Chs_@
zrFsk*TJ~H=s{jO(qgJ4l0Q>23fL3~lS_E%@oYfEJ3}V?tN~M(-iYBLEf%@#GG^=Zc
zn>60x5SOY?R{ZCaCxa$=D(*qWgK6!K1qH@{PI3@Ju}BDFVoCkwrI-^8S1z*5>0NNS
zSFNj^D-YhfGWA#$Hw|3vO(A1}P?KVInw~4E<OUCmG!HM0-tds%;+LEzK%J^ZwK2a<
z^Ogl~sY2hbMdeT2jl|8umy0P}{_`j(JM_f0fwI6+hwCD8?3WloX=|Tk*+~^elp0ko
zHdRxAS>We8FWBCSu=04Oc1hpukz_HZ=l^u$-cumQi>S}*0~~7caOJMFl<k!8vwU)e
z^=LWzCIAAC73SF?6)aK(>?bi4Ds(6y<0m{ftgUQ*W0|EHAEBrOlNoizhOqiM@0Gc-
zX85x>vml<2TxZU&l3A1eXr~#bvDt-kTf(#zHlpV@6#SPuayDj9%CI%fUW@OB8n}94
zVn1LKN>_B}(HljxM@oJp&kC3D8P?Az%_!A)J8Idf3rHJON~Q{4wjA-fS{>RYhhHQq
zx5|m9icNuD%zeowfv>;L7_M1Y$6Ka+$ggn1&ICzTO(|6SmKG++L3_VcK9`TNi5@W*
zshbZpBC6PwosxqyWJFY&+j$L&TX8KobgoduQXJRGOxJAwf~B25KPO8D`w}c#OczO)
z40Cp@3ewp|d3?0!*x6!kle=PufG<;hz7((z!pxNe%)mG2qivcQ_M!E6oYtl#7nSrR
zdLSybe{izFS4Fz##nFq%1JiK~L8}|P&6+=fk+HI2Woh@S#vh!K>OVL<OdrPl3UR<9
zp6mtJdnc;~5}gnQ?5U>b*OSxA7^{yB{_Or)L61@ldvB=@$c7wxd2T-4H3+%Y_UqUR
z9`rX7!tlB}HZb|$zYB-S$nH&Dk=!`UMqLYGgja|w>->DC^Y^){9n_EfPyV*IU#cT)
zV=V|O>lxC;2*!W%Ps`qTanb?d-=D?moh8KGeyH<Li$H!l>5vGO=z#E4Q1Mc?p&6?$
z?oKVQA`ZwbA8d7Ims9a(2M@|$=OrH9aZSB$4V9|5>$;`lr8|t%zc#<8pFOZ1aiEE{
z%JZLAqagk~xxVt#_$ebD5QbaL{)^ApYuzZiIT(k0SK!39j!u~ky}?S^Vy$G{Mc>+R
z593XTNgE28N#m!)T*z~*;<;Z`On@`UDhU6}WvA8;b%I9q1<8L;l<QVrE+NUz?K=JY
z75!fqOj#aYvP}3o%tl@dnM$uvSEMMrU!2(ADIHlke*1qq7`Oea#UlpWIsd<%%TGLd
z&`|MDEHMAoKixr6H9y;vf8y|;W;rvir#Jul`!{O;3Z~Po>Hl;%J=W_F&gaj*4h<1l
zH?aDqjryPd#u#hYr;AL&w$r}fx=jS{?~%8wRJ@i0ln2z;$()O{#=&o-mG;m5H{({2
zgS2MPt}h*_o0BClSdm!*RQ&e_V=qfTX~bIW+vQu+!TMzgW3qc|*tM_aL1>hW!QN#A
z%d-pB%llZ9*TPk8M9=E39s4grHIf>QZl(s754o`pMO_PG#M;mjXi4nu$y8(5PV&iB
zWLX@GINUN>oF_@vB1NV1ma>Y@sO?}txT2<(S1G&`jU2$_-#Da(T??8DuP}~eC>z9t
zR}3!!n&O9E>>aXf1q{ATzVSMs%Kj;FC>=9jV|2V#F*6LD%c|R*k)Qm7!~5vg*CTed
zk)4h{BBqxa&iuRO+;uV{$$-pt?Ch}7^NZeD3RbdQs#Ng^EJ(L>ip!lLQGHR77<SnV
zd3v5DfKj}FHHP``e0jl5>u>uE(fT3g`jVI7u)5$->^{4Mdk<CzSB)t<Zl}c0C9u34
zSDnW3s)Ew-^>s2-ceF(OFwA^uf9s`zgiFszhZ^%41^7gZ;SbKy&sSJ!Xd1^e^~+<7
z<w}s@$i#v5PN&$G1M5F6rI@p?DocH;ah@RxC2uJl`43x>m%6<V20mTfgsao=?qyG=
zzreeh0k$dMCDPl70ecLDh1_o2`WIY()zMi@KDhGfZ!J$bIk)J<-uG@KSUuE))GLR(
zw^@56e{f{wC#Pu#t}h%p|KOl~EmywW-C6P{xqPK>)-9D3tTz<qpE&6(Z5Wfll~_l*
zGmltuz0`~AY4)EsMhG!^?)K+wY<P?ebv6Y;IuDuVziT(xV*n>UgHN2gGZYW6HcEUD
zQj570>JN@bCj4Udyk4C65AynF8!>v8hsW$kEE52YNz~yb5}KZt$M5{=IXiAC4_FS<
z{P#QS&On*&&wU*{{|>kGiojlmf&0mTNj`8-_V|+i@|;LNXi(FsyHx(VApI!Z0C}NX
zLFD%r4=4}l4wLMM@BY23zIX-?>Mm6f4&00lUcCAD9n9G_f+RdYocf*Q7o~vb#&W8=
ze4zF3?+)xm8xMU=+B*WA|AQRQ{sVPz1^<OQ-F|;jCt7##AJ7TJ0-WU%lJE}>>x$2}
zf4BKJGWp2-(<p3TORC=R??HycIOD<lO(Hx-j9AG0duvLsFt(-O{RId0;>#uHzc>=}
z7f0~^_UeDfoUO83v%f(1|9#DCbEc;2NvVG>!hrS8_}|`rhuN+j&Hqotz#<mKKZqr2
z{MsPx;>O?4^xsJIZ<u`$pPhCiJNw=1fbKux@UP(YzktN<e}&?|04i`$$aLiT>Mt%E
z_sWp2wz{7${Ea=`jIYB*IJUpSD+1sDvHJ1|KLxoG<q_%I<uo)s-u#9e9hzl*r2Fda
z>I>eGTVIiwm9rs%n5$(hpp}rt3Ab+C#wE~x;3xH8n+{{T-DL4uWriD_j<mSrIUPDH
zDZfsQR@^wT@5~&NeB5$4$WPBr=}6Cin7TDT$Hvz>Sn+eIDa&G{{26caZCQPs|8AC}
zPrFlF(B7G#i}7~R#nf>x>4zgq!>S4*X|~lrIFGdqt`iQ_)8L7uCj3;>Q7K<!=l9a=
zhiP0*bD3{+exhzk<%#q^4Ju|++jRy`m+pH0Hp;ui$u0`%dV$%wj_B*TZqG`e84q1s
zX=K?y=4@*2Cf(oN@;gY3k3z1XEw+CAE}}ic@rOp%hx?|XN$6tHzHe>q_`sZ=J&Tmf
zr1b88aH73plih`Vt|i(H%B8_g{prrHw*7j#e^gu?n35Lt82g;i_yk>6+*?AgY(CiX
zy}oT8%H{R&tZ*|R8~02#OkOyn`T3Ew7PeK-o-Ap7afYYGev+!Dnsvx91>x&}^?3~S
zJPD3VN?+s`$a%sgPh@kO(ZDm!{RfAEyLPJ2?s92gTt_tWKN%M8SU)Na+PDqMxoByr
zAvY+8AGI`qIok~E+FpaH{xUJpXb<1o?;~I71G^1p%^1dfmw(;B8^>$XykpZ#(aHBO
zVy%*b;@=$9QAWBBbZ?Z5am|_Li-cC?>uB@YMt+ZCTK#tI>+GBnc??Ai)6kc=33`|g
zGh^_3o0RYJU4<NM0yvyH3v?JouJL#Xc7Q4~krR4QkChSEVW=P_{N+Fc_gs}+a9z%w
z%B&~n#0|a8rIfxsQL|;=wuxJ8TngFQI{NWTTuld#MV$=Z7sxd1dQ5V#M^F|$;aF#f
z%|CG00;SAHjoTzz<u6R*d~=(iiJ9%1r3*csB%{MJk*<V!4VH+0KHm#=h6C_$cm2X%
zW`RXkR0EYu=g7LGh;k-Y=T7P!f4{0TOng(kKs=*9N%U~96_<OKT5dR)X0d^GaA#Sg
z)$u0}DnohXDK@Wx;4bhmb;B(zBKTRXwc9YB;gsV5)-%zAYkc5dO}Gxv{0=qLaYj&`
zfJn5Ey5V6Kste8E>+8Rk{G-+UjqE=M=WqP~>qVEjI$H67%$ga>|1+I)aH)`S<!)-o
z`AW3=F5~R+RnH4vP_pa$uZAZ@o5bfQm#!>#w<daB<z`&jl4<cFMHF>{%Gz{_M1%M{
zhu)v3x$mqnaCne-Ddr6|I~Z_#FnpdiA8^xYn7VFCCU6*@L2gxBX<L3P7*3Y=E0Igs
z;nSJ35Hk!_Hkp0>s=2nu?pc46eb!X-!RoX~LwWz=E6l={H<!n~R730UTm1K5Z}62-
z5g`@ZB{Uj18}Wxi9rK1_wK~O7cD8ObMuv#;0<AnPTkd|Zk(_y6Bb_&oHcJxd4ht>3
zZ=z>du$sB3-g^ML$F|=OhF2#ecT84I4}W^Ew9Z#SRjOyn#kB$>riWk7>+;vBHiV^g
zd5k0yJSYhw8xoykNNBnZpU{U*&ZPXFV&@W7w@6~>O?YT|O0a#|jZXEJ-nlMYeOMPl
z_K;qyN||U+Uyz!k+>i`EAFjHj8OWV9F36EKfn*QQ7}v~TD=iNV2I=Rg#=nf5cmw(H
zUn7Bo<MLDbV!C}NOH{sFWx*6<u7-MDz6etlnk{M_^cp(y(q-xTs!u9N(i4XPWjZgM
zutP;>yE}ue5~f+E*ZAPB?51&#mWR?Lrq^QZqU|@%1kYENqt2(p>K?Ul(tJ<H&D1W|
z<h(Sz;FebI7-Xw9E!TFjp%D|eWG)Z|Grh}qv;@0nLUJ^~TGf?8FTcm!MM#omhTyJr
zOAK0UJ*dDfAf-{0qS4+sOJbQ|O?+ZkbGWnBjVX|=gU*nMxm_;n<j?W*z(tiUtP3YL
zV~NAU@8LA5x+$&UnJ=<waAx!x9D4%MYrB!#?A<=%C90)MsFyvtQvH^73h;pdMZ}L?
z8(xJ^Cq`5CB`#^dAFgy3%RyXeEu_j`<F87**r_Iu%~+6Xb>c~#Z_M*Gi#CQno*pK_
zSPTHYirg`3xXcx}iwI5F#`x6l&6Sa$lg5^&C+Tr2B>;|3(+$>8ogdCSUO^Dl!{7f~
z$Z>GQ3bGa)oA~<Te9#cFj=*<tqjUVEQdl>hRHlnuy3kyv^HMjpp}Q+C(|3Y&jr<9V
z(R`f^-n5eLKtSPuA7z$}#CdetGt{G%HKe7qXx$uMoxkNOCuD>Y)(osq9~6`n+M)Le
z-atu)sik76`dkd9tE^c=Z1s;@s7&|><qPFg0gj*5%2nVx6US?BZeVZz^^3R@VEi})
zBJ|<3{5KNs@E~bE5Et5n863h-jXcqxX>&B38mieTu{V(u)H}+3n{A`-J}JAooaiHy
z(?ZEG5xSr8u}gK0owK#mYV6xOOJHJUvdBDB)jThGa;kR^Po3aBlKrH`Dro(=sOL{z
zNuZ!<qDPf++=;2=d@w1s)x^m=YXyt-cowM=u~&{7v+4N{7ry@WjJL10M=&Kh4nIR5
zD`i3;i(<mK5$Xxdm8GFqRFd^kvj|;#anm1cCzLwuQA`_aDlrainjsY0>ES!?IJGh+
z>Wz2__oXFvK8m&tcDqTuDutU#Ssx^wbo;z7{(<Bi2iR8UD;NE+-nGakH)bcvhQ5W7
zAkF9PnhTm}7ivm~Z5W<$P0xQyUUcmWxGelvtp0y{yBF<W&T~Z<ii4xpEU-DPfpY&@
zi7ygsJMxw%t_)>JWc{PT^=NwXYv@)vdY)^`K}!HPvgXHCgfUxW?&8J}bm8MC21)he
zAb?D@y~bPIubHmX%_q~9%Akd~Da@UuEI+2-_YKaulAB4AiPQ<?dy27TGqA<?9C1y8
zLwt*=%zn~bUlH$_9De&;FD@>0Yo9RH6%L;HHvYl~3!U&KLKxWOtd*h$?vbU9Do#Xx
zaWEPOyK+5kNGck1s&fzamhg1^P0+vn(3<O)n;EHW)7YxgFMdTQ2Gv>_wV)zkn)_*Z
z6Y%P7_MQK?0VY2|IWr_j3`jr?+=3-Pv`iake(XiP)F^A#s@kdzs}mF_g5efup~F-P
zrY%(C!A0Zr(wdoy7C{sAZ%j7qdtl<qj#3?Pt3YKJF#$V8gHg9wt8+J_`)noQH@f#R
zAYan^rC4Qy6^I#tdE<+@$(s}(56%FJgLwk}<#%?%qNC&y=hf%&1VckYa=<p()J*T6
z_;F4r`sCy4VffVRk#@}@yWwk9rCu$?UBEm;cxaS?Ijh(H&Da08Vfb&X-5pHjUFz{O
z@0pDt;^6+POIU0wVYU5At@UcJF~4SjvnEmGAGwMrT2r}-6wwMAxjhjC_a8el$YBM&
zzkdJRBNIW&DbMh#11p74;UnLoS!#ym4P_TBh8KT!PU~(Mb`mU;#nMyC*n+}bofGXs
zL$R!(eJ=zBcSi`MKB#f6vJY6|9w44n@%}c^aU-6hGnLHttPjaEth=A8Qqtsv)Y3AH
zp_ww5noT-VX`C-7G63oLd56|07Z}p1_}Y@L4=9KjMG<h46SB`lJTw(!;-mHqv3c|9
z;xbbR1gJ7=JN_QLUVI*X!zs@#zuadkA*mG|SF^c8>1Q$IJkQ0X6Pp&j%FBJmNKS~6
z!HN$#$D(W?@Q@AzG-FX#8X3Nb)6(qxjHHiWt4+o!`~EVt>Ee8kT~e&U{mSVuK<nyq
z?Wa?F4a2VooSdOCA=&fu#dqv_V^VRrINTBII9mTr_&<%SQ8#@g!nVQ-(^5vL-G0FP
zZ0_?B=Yh4rNiyyD=`Wl{92{0nmcHK?EsY=h%iUBag^GU0^snUY{(cyrHCi-?lRuVK
zO@Fk0_jL0Q&fU?C3VauwOy7&20XPpEopJszM4Vs#8@K+RzE7IrGTx@1aOWlVLkY8t
zk8k9iUd44kBOC&yr=jO3Cm*-=et~fuArvkUpHS7XEL-%LYYKnR7eRXq_@wk79DH=a
zdcEbBex|7l0|AwW5^$}39ZvxIPBqrv|9YJ}61w?_Pg@p?lp^%iXfx8KV=$L;;G2XI
z8<wP7<?L_NNJ(|0_X<hMP@EVwM)rb3ji)D=4`Fs+p5u(JbwCI4finZ;%2;W!MD3_Y
zWK{-4ys$M^r)kRjGvsX0dDIGNyq-nTJm8%~8j4p`FO2ZiIC$a8Nn$15`PDc5J25}+
zI^)<~O$$<V463?CPpYkd^I_<SpzX<IdhD6*2NBb)V6m;$Qvi`*PKwq&w3$OFy_Me&
zX(Gt(6D>mWFL0z4K0Z?ApZ(HY52gB6ure(G9MW3ti1BkPyP{}(V0{H^Q*Gd4<0E?r
zW^N-aVjd@eiur<m!bKm?CNc+s;O!6<i$0e5dn*^Bv#*fpFKa#A<p6vYAv5;A?FAF8
zL#F1n(-j%&XF}x(g%$WBf&M^lZ|oKlxRyfZLwq3TBYh<XVjWp1N|O_w<l4;r3_2%Y
zL-qg#v5>nR5)y1}*e^To$(eDZep33jzDk=2p{j6J_Q4>+v*i8+JPy;Jtqb?aM?vQy
zsfG<q-F{3=t4Xi{dQg7}>eqF{5uomo-NT7NGY;I2&OvLGwQkwqk)BDSVOE`o##Esy
zA8sb>SPwSfks2XLnzN@=Ea|Fa8~a<wz^h0#_eTyAH(UMdIp7bD{Cofsy5&vAODgwc
zp;)F=$4pqElX9ga3h0Xfluo3&<)YhkDl(N9Yc{*yN$3+X&z@WmjHJ>ER2OS5SckE_
zp-S}-za^?gXI#JeYHlxeWF>7#<3*kSAy`t|x@OBXI}*9!kSJtC_cQ5)o{_^VB#c`d
zGhAZeI`>pbvdL4BaLcz}kbT=Gp=t0b@?816B|?OxYMJ*u`IAUaXo_g)u{h!m!>Hpu
z+&zQjXZAJ{Ufp6QH}s`l+*>s^F5x?pShg9vehb;+_*t1bpN7}@1v4X!bKkBno$!|K
zLtph5*i%6oqD8}pP;8h}lCE~nr(c(M`AtzGv-Qx>+N;B{tPHsVla`-OhAM_<`Cd%i
zMNp7VismwPV{K<Y+_q$NHWNDi=D_p!x4p+kW7X*4LQgZxxG)R#WToQ4PTQUr9Gq{f
zXzJMsziKi7IBkJ@8~bWc!)rlI%!%)dK|-=(BwAm)Xc-@mGt#qv+)*2n#`P9&c-g+7
z-#!RUHyT!@2ZV@<dKR1E&uW(%o`b_#-JUA(E=)I{{YIZf6V|Ed%mS#m!2nk`=CKXp
zd8HN`Z8(X0=(EqO^h5j6&LzQhLVk|93nNTjpGwCTI)ykiHg!p9^-$=@ZS8u!Mu+)(
z8zkv&<`QCIa>}j=;8Wiq`p^}Mq+Q%ZF9S$p-bdf954%4Hf3DaHrEmr4f9{_DgF}6o
z7<qP56_gBUi4ax=p2L?o)8{IO+TL_ZNHvy+NA^!5RZ$!#dafL(SsRxK7}D0e<%dSB
zGYMq{&a<&>EAW8hZ0h|Hk|oROEw_Ss8}mOnd}mx`F7p6=tc1W^{kk5H*CcD=<pWsK
zH7Jv^POhXYLbCv1J(j0dG#uGkAA@|Aw4gv2pCxy?u{O^sn$zLq-px5{NydjFrS0fs
zJ8}>wJW`CDmJJLQexUu7N7<{Za8W=x_Y|&KFz>+A-%@LcXxtHUnR#AfufuvkS1R*y
z?&CHor^<eeTcrkkwmu%R=1jB?Lpg5&6Zs!68J$W!mCMPPHej6*4XVK{;6+p?svEv5
zq=7BwFOoe_)R}<)sz(R{>5{%m*&{96@66F4oyn}K$}@C27ks{XkW7qKNpsC$38Ho(
zR0TaLzisA8NsRWL%8c7>hm&emyeNJwM7)ayL0Ox*8n<28#c*!y_|9TPtkbBUwU>5&
zSaJSIOK!##^QLVZ-k2&<ot3-r9F#Hv${=BbjdSSq0C%{X6n(BC@p?gG>8$be93?LW
z#`c3Hm?Pi!pK=&ckj}2@)or98%u3*L;;V}5IuW&lQjL?XE84jG!vXpMTD9n=brou>
z<QBAHlmToGu2=$cY2?I1MerdN?rBotPRz&FD3N$m?w_>e&KBl$z$fz6J8cA=srK+R
z8+Rn;-F)fMWmoxbTZwQGoe66#m#(fxp`06|M@`DXBu(X`xkK+dU0dM5wYj*|aTm2&
zTiF)nFWE=Ld`0P=(*>T-^#j<Xwn3vQo%Hdc<ou46K~j#3D0DtOiDGyDblKJ?BI5$(
zdnY2ZYuOSn=9Ci+C5DXd4}OVo+PAm_j9LR|EZmtXO=(E#dz&hEH<a%u-Avc5D4O&h
zfuQw?n@=~n22_}|5E_PhlmHI}KSG$5!x_I`{1?th<4HsO+Wi1j>9u3GEx6vXN@~B8
zVxN9u3)mItnjJY&dAWcz8w!z2bTF`#X<0+ZZk2VHzRCF>wp&l0hCCHNSH6g>h7S5c
z0@A30xq99Oxr0a~REB-4-gJV>m@5by_KSGmc2YYR8M&}&n=~&3FD*E%3zBKn`17Y_
z*d;ycuSTHOY`gpH`uaNzZ99FSZN)S8-t<w?6lz0U-;Ej#{ixWi;0YiauM5>Kx1}g8
zS)uoUbfe{J_NPA}y>9V_nTpL^B{ACYZ+?{&jy=s%L)>>i+?qKQ=EAdRmW^mZ^)c4H
z?s;sHW4mc8Hfh$0av+1S5QE(2s|l|iNo8T{+ED(lQi@*uVK1dwTPo8P+(MTMCn7Ir
z$Q`o{*OzWWtRUVbRzjKa=Bx44Py@N?9PM3El{dNbC^YV6fLyNDhw20*L8fXDB{d{Y
zwYvYg%p3e+p%g>U!<J?mpKS1jc=_t;?vU&BXV2#iwz_g4o0LA|L9PMXrq8D*+Rw4m
zcYD0-<S>#X+ZRkUa)<DNcqWo3d}?6mS*FK=23@ioTM`Q?iz%;}iE&V5mmhc~A}HMN
z9PdTMi|}8?-Zz|b9Cg>OB<H%ijjdNVNB+^zd}!Cbk+-bpz4u(DAEL`@pUm+lCqfEZ
ziVNb~N|+hX(~x%~XGlT50}_nz8<l3g1M=2_#h3P1QF||PZ!Lv=2hZB}uXXOo2lZu^
zuC8=g8xhVlcnl3&I#BkNm?NL=rbT52l3#?^gKb6{)|S;wHk`{vQ7193KQ{Of<7^2a
zK`}7+JX(kGP5$_+mRF9yWT9V$#P>nGNDAbgYQOPL6_2h#weG<;@zEk9im|??D_5?j
zvX~q0WZgnRmBJS1MF40BI_E1|TUP&iT2vGB%2vj0UMVv2S%k;?*5@X6O+}-J<4Jm?
zEo&L2GPRlYMUz`MG)6N6U71>fvN>SED^Er{!{n9zeAQ$Nkf12O_RDu4(bMklihrzX
z6btK7KL1f=2|b;{Tiu4w3J-fb+k!rA%X3e?UPXivIjjzzz3(<E>`Z=Noh;v;Arhvc
z<5@mdqN2Ss<4g@P{Uzht{gdzBuu0Y!SB*F`Z&?JeA)=wE9k21H4rPlikAAM~jbSyS
z#=0<H;70k$fwlTOcGd5s{8bq66oS~R>454W!fQQz+v9Nd-tp*6UF{X_X^REii1A19
zfx{8a>vjwFnL%2DS`DXsbb$lFY|tyvWu8C(RY@YXbQ{AsdpuVvKh?pd%O==wrdU!8
za0lR(LV4_Au8$m2)8|HKS5?oaMt)m4lw5(W>R+`)KI^uVYj=Yc6$2#jVWB6zlOawT
zgobXB;k5wbB6jP9MlW{P`<+5c=5Y>BU#xJzPN%l&*b-W@^i}GQm{K@bmYGFNNzlNa
zfuvu}{12jaN;!*XQ@j;(Or{72n{CJ}<VFJIlc9@evS=^BlVVb^P&vGy`J_qG)i#Nh
z=Ecx1HFT&~a1ypm3f(O|`4BYfJxLt}TjnyLv)_bokv**+o~F%1hNFz=b(i>^#fc=+
z(oxqDVc~1<am^FIF@BFLd^6wi-F&_vMFnG=toHUO*{dmBcc(V^i1*Q5$F$agVGPNn
zVlDe~j1iEQVcNh^p$S2J2{6ibm?u+}qY}Q~=u5j{vstdYRvx|l0k(~UBcQOqSapb$
zZ!e9sh)d<x)-gCd@5-=k$`iPpi(yuy)Px%F^~wZmRI^!6zu?jBd&ZJyvR;LL%Qeo#
z_ee?NNoh04z17CqKUL|31e;6#sHYr~L=z=thA)L~0rZ5gU{+vAl??-}+-Y6tlIpuf
zSPn(^Pn#e|-Hic3c|>F4olAKfF~OP_mVO{q>MvJep49wYLx4AKOek=ed5pHkwgd$O
zfk*A+5!y3e*@g#{ftAZT>xr8?ZMrgU2C?+<x+x3R8Y}0`D4dI76_nKW;_M$BqbN@c
z0}h}ZohdO$XO53otVpA=;#!fF1gO<z5;|S}N_RJWhb_EAaeedkdfZ0B*{o(4YiL*#
z%}Y;oF_nC!tp!SJPh|wNWoexJ<ovXp)FhHUk2vj=%QVO&Ol|%YKyN!*L?R{}(Vh)%
zK=h~=Gox!+PtqFMCKVlX#w-IaoFHaUsTtoAy@{c88_cd)2Vl=ZIVRXsukXMJ)g}70
z4EwkT(=Ho)&DZAm&XAQ$A>riRX)$>69XgdOaeQM(kcga)sjcM79%Z9c$7?_cf#UYN
zcn!&kV~3~3VOcWI8i*iyNDU`x(JU9$0u%rcs!^0O@SsevpUMWOT;gsb4YcYN>pd6s
zNAB$!jm}NJqfXM<Ph1`%>2!+*jc}I}*)!Eq)*fj3m<Rn)O5H8ceic2po3^mBF^oFm
zht0&}nh2Yt5CSl9ag9MClR(8Ae#<2s421E+YR|&9DUi?<e=o=*%4XK>KEf2cFdb3e
zVzxN-J2hoAc%s3{N*GWjl?={nV^0->iYZTA94z=f?bD>u%Io#44Y>Gbce?)xxe{-u
z)j6o!x$u3}Oy~qR>%I|+^o2>1<SH4NBgc~s@&;!^QnRGhm*~^lN$(Mu9ml&E)dof+
zW$1*>2}mc2Yi82sDdpt52DN+Z`qke@o><ax;x4)$3EEn=LMU?6rJpmYo^|OXS<iiK
zIxMz=0bbe3^IE_1fDZj5x#Kfn@MgRR^j*QXB02Wb4PQzElYm`us%H9(E%RkQ5laX}
zn6B#CYYK%<y#8@ym0Dml8F)18Wbk$EW}+M$kLg1)!)Y2|p3}Tj>CBNKVyFf)9B;Ew
zQ;f78(J&|zl`kLtcBAaf5CaqfVwc6ED86)dji?3&q*!>FX3k796*2;uM@!-NH;>_9
zB7syw2JCM5a6)a?D!N395?4h>2JZK4F4)OzqL9Fj`@ANqNKnJUo15&Ge-6=ta2H)7
z&~<9_9%)bGqMr=z+^in;SroA8$~akFAwmZ`!NkXjxdjXk@@x4lxJk<H8VDr4iFBDY
zS)w!##4q>h`>C@%Hb)9Ada{*u+v1QeEO52DRP*Z{REuRnmrM9bRa~0~eX?SB``Q%D
z>Ud)UME#IDQy2a{ydSMVk-zUC!C=>qk+Ue3_`LR$(0rNe8AU!#MS#ef2YV*nb13z@
zD3xUPun$I)GhY}KJ1ON}^@i#w`?WZViya4es}*RkOmPs~j1k)3Ig`d)qLK%AE%8;O
zB6m30OZr#Yd|kr7F)mlF92Y+@f@cn$_(_QKl=NvyE}b7t`xR#o8Xf!Cqn{m672;U2
zd<=?F>Xwk=tRUr~{h;>iLuG#D^9ufM{>DbKYlBT6J<nT7wO<GR6zh_^{^63()JLxg
zzTCOA^Wb<qXBf#n%rZ+a=Q-+_dGonVYox+c<T7n`6c^nhdE~2pTXA1|IC1&*_f^xL
z^F@63&G6a#6MIq|tF7oOpit=2uvb+=*2u7@RFNCA-NEzBsuBuB52;nw;f{|<sr_8S
zfTvy{W$+}&f(VOzkes&W25Z>pTcNpNgOYnQN64Bso8ie)ldg<*{o&Q_%#Wv;CIE5g
zC;Jb9uP0tx-r=HrYyHOQms_Vry7NzG2Ss?$XlOFAhoI7ssZ1(wOK(n{e2&&f^s*(l
zgQGmeJun@*)+FhX8cu%WHN|zA#bwe`6?QhM*5p26Ri6$noD;QgXUCs)%mkr;Cw&U0
z=+;Eu(=B^Y(#gDodJK^&?IPsG2WvckDC4{%it4pf*;DZ@yTPAlU<X7k!kb4_uNXYV
z6ablHOBqoe19`X$#Xp(%Lc=Vil1@&-oVNfjIp<N++<km&_PzHe`JW|~HG1=vke>3I
zeR3{In)mYkPD;Y?V&~eWfmy>8%nx$2BPZeLz^kHcQ~5g04&`U-eQk>+nB2_i%9?<f
zmos(>JpAQ{$GuS6y_R&?X&AFcxh2Pfkqime;x&6*l~vp+LRK(hV=(=o%?_O;d%PmW
zsV6K8OK=C^Vo6dFCh5Vh++2NoajGyCtCNQxcanq9DV$ydZbJ*tTlw~)W?DWbm+6Yn
zINAVso|1K@kKn=563fKz9I@!8Lr3a4y(V$X_yYLK-sJ+AG=D2QWzy%#7pS`N{T|1N
z^^zCtOCrP4GX@|L!U<2iSF?F%{@}br?GucZ8q^rtbziG?fmH97u{_7L$Li#c+at1C
z2og2@C=&@cD)mj3s}k(ECIuiLAK5kY4vs=BIqMz2%+A6F)x=g#p+CLNj+Lt+k4;zp
z-~<tsht_zJ61ONsIVx1{;@+z*J}$B$d1PHWk<3)jt(H>XXRT!<7*K{%al3@-7OVvS
z?jtbVd)f=J6ibch5zBnH(9LnW(#p>i?j}c`L}#FbvNL?1yC_K2pNEUd)-ON{)4<w<
zDA$3K_^E(23(0;o;410Iy<*RScCO}BjuH)L!VH^L@7K9`ZosrKx$ezHiNzBPM^XhX
zSWOUrL8khE=vNGMLEBWTNI3*L!G4@pojdj$E+~@naf$EY$SodhQKf&CpJsM;KI2-Z
zc$xDmy0@u3vUwT*_wW90j=p-k6y?TpJK<JXPqvflhN@q0kPA40D(yELnWW$q%#|p|
zp!8eG4~)i}^JhVuQF$S0q66&{YgUJM_0`&|y8(JndhAP&%SYga4CiN)qQDJb*>KE;
zTKhqpul#QE{~+$IquThwcF|B;C=_YYqHQVggFtZC7I#WWa40Up-6@m;#ft=&;t<>t
z+@*MN2~xaBC{mo#o8R}H?_1~Izs^1Du6y<$GkdQ!dnWVF&g?zU`@YYk<Q1WQ+&%!X
ztA~HuY1CV{{fD_I4Qj2n&@8!5rSpW0l%9mwkTSIO&rd3{zkvAF&{dn(T{r*Oo80zz
zFUe^vfkvAL&Q0<aH`~M%->aCS1})J}4gNn~mw8KJ*?a?^Ha^Vs?u&-xY%#RyZOMDy
z_Fm5ErZ%r)<2hn~1xQ`D8Qy^}AvksaH>IKf#zp`CtJHj`;xI8NdH;s1AGy+%?J*Ob
zkbWb@n`tuKszpBWgHR!arR=)t`tc{jf1=Rs5E7fgimS9%auE9IQ$V)onadNyekZ&Z
z->XN}KTv>)5U-01?|_!ikr~J9gSo?9G<L;Bn9~QKUseY@_)<rHp4Xu^lBp{j1S~YR
zM$xE=Y7{SJDV7JIw`axSQLj#h=f22NbjIo&UbUfbOh?mU4)Q}??Mc>ekhBFA?dPk6
z^9z{G<?Eam`;Ax58(6Orp`^<{Jj0mLqS5!wC^oB+(m#wclxuzMKPKLN`~TF2xM$sq
z{^4n{3TV|&P_h0Hf3V$TZ9Oe45lG3Q(~3t<G3140myg!+gOn2U%+T<qe4gB{>>Knw
zFG^X~j$+QevAk#=(3APFpi4W=?5abTT_!hqj}u0c%WhPxHX#>Hi27tqMaS}TOiDao
z)<GoBpN`mQpmu*t&c1RokX{?c<58;`PbK3;fj7rN;$fRuyGZbC3eF6xP?S|*5`{iP
zlv_dN)JuK_NRKYA2LEr}`eS~k4Qn~$I{+EooByp7;1_JTJ4uEwd>R2&&KuVr4J_F;
zlsSoP{N$esN6RS9G*&7otAG`X*uH&^84bFpaMzD=nO@zgjj5i@y2PrB+Z$@7=!%XT
zzBZVVhe9E8axGgQ800DK&c1y{VD+v(f5vK@2mh9nQ=5f)y}GVAT(99CkruqzKN?{=
zLOv}Q`=t+DM;-`3G8jT81?r$G6y^d=2lkGQ_W8>)Uy=lwzjaP>Z`ed1XWJ~ApjZ_9
zpL>;y9a@B%3u*kz51=mkWuKSoHJY14kP28w%T>P|Q?D%yWVKQKO|4!!VsWik&qrQ&
zoS3OCAIYdAL1iyS$Rup?nsx?ixXA5^R2WugEY(cfyH>nz6-P#+DpGe76C3Tin$ItB
zG{oQ~4|Y!gAp1GfCn`E^Pf`Z}U;oR*&&poBe)%x?-)5|;vET+ETeace#=nJFsW~SY
z@Ia7j>US_*|B}g<N8;PnTnwFW9w^JBCs63%2jgD}hiv8yZhgC`39vpIvEVI@mQ!p;
zk1aq8+`j8><u5?#!G_07<JLptbhp#a?_0>}=pQ+fS>h5I`#9PeCqqFUnQ{7W1LeE_
znV$)iFMVPdb@c~=+?Jf#w^p1VMDL!$aLWThvq|FCB?JS>Yv=$4lQkuZ&5T<~Q|sYY
zo#mv1!KMZnSN$et=FySrWuPB8^u0|5o<G!a<qWDB7HwTM>)Ut@Fcz(TR99VP>HJSa
zd+RlzljZc<@i%R%ouAU9>D<IW1P@kku{WHp)=cx|fiesm_{H|1PT%wRJQ%NTe{&z8
zapNDT@ULmwnE^24qP>^6V9d*h!PqYU7M#+5w|MQwrvL6~?|xvnp?qA{fLN^EhOoLp
z=*}h_dlOd6KvTj<oB6@b>tM=Xspanrh5?CMH1;FIfG>~wI7;u<gi&8`zNxgtkD5Ly
zv^MGp#-}xpS(5q-_|oA%gOlYD`~jXn_ZOgx{drf4(wgsjg_yk2Eq?)E#ESD^(Ybh`
z&>QTPAFpzWF}5&9v>rD|zk-=QUIhtCDn+N4#xk3@GrfHD?;!wCQzbcO_AG4}vzFTL
zrU^p)D=5xB^%t<Ccyi%=_Yu^o@fR?YemW6!*FEtc;cPE*6(eYzq0c-H*IR}nqYIQ!
zXVT6HEx!G)@(8&yQ!Dvr%U_g*4ARAzXPZVi42-{wiJmuv<#P;tj?O4~;ukvMLz0?l
zSy`~)bD#y@kLI|{qMlmP7TlTgIgmv(+^%7rIm5;@u5{9&414YrMpn<iu5IBQ%&V`!
z^cqBU1u^dEIZqV9jfWlT(X>G^dPXD>cFyJjQ}czw8q&VJ?u1l8UV_WU#9xu9muW=c
z@fpKeQd01<%w@?IvC^Fx3AD26os6<1c{?wxX2_O&Q!@pz(w(p5kjpa3VMTI<=?MuR
z-3`z1ceU10I}g@NcSu=%4b01UGr^PHt*I{%$vS9PM4-$@pXwYc!ULF9BEV{`mOstY
z@grLEElsn!b*DBzb8$H$-}B@u^d;k?Vlbz;J-g3kHeqEmp$6whioonK4Bc#*#J4?p
zoJ)*}mZ`b}vRBr-#kD5hv>ZKX8j2u*`dl>DINz7xd=w+@cD;Sz(#xL)ifsNc;BQX3
z9(IMY(K(DB2o)?g+S`LRImttY{U-!YrtOncQ^pKdD-~31z8`!kJCAa&&<JDL!ntyV
zFt(q@XV8Yu1v;s7JUg9;=hs+lN;61LeQ?%2`o^P-(rcb_AUUF7u5ykg_07Hj*u&8d
zM1*tm#Ye;2g8Hb-$e|;@e&fVh-b(B+_C>(*&*x1i)r9s$^;5L(?0;saB|B$4g{`a#
z6i&$CDs?`xC-}Kgf=uNiJNRu2U2>u{Xc~$ph}4Q%ZmKO8x~HB0nNn$!=LwXt{?0X$
z5@A%RJY5*7UoyR`2xeBWv%~$}ymV_n(B=hoq|aAzEFnKn+_A1_;AY`)KnM*yT}c&W
zw`o&Q*7g*WQ>61&-!YOg)`)uh9QWjIG4E4@gZC0kLKZ`_?FrWt?Pu<Gski2*c@t)Q
zU!fI<a5gkS+Yj}aIsv6f)_HaPGEUK(%+WH(Cd)9kZ=}m&#$u0Ta)C)=6#5>`HLE3R
zB}!%`dCDOyVG0QZSLG_{UL2EiNCq=op_Jj|r9@l)m-s)n@t!TFU6W+1zE|)n<{k{w
zaA}n6%`>qzXByF7m7I=RYRV}oA)mpyeecL~?@jWZ`-056{+@}u3xCtS-W|s=YtrT~
zfYsNpeF4%+P3_JedSQG`&}<)kcS~00KRJ9}(x0)Ec&J~G@qp&_e3|@RsTv4xP;p62
z!o7T2FvTwPhc)%>*Xfs9wjL@TKzZ7Dl}N^1)xxQ=+KdQ|;%QSk$I_awZ$FRln)Ss@
zJ1E3WJJc=F$$;Mk>E%3UH9QO;T`E)Q6DNtJ{WNZUdTzP-=yX(|p-;J~K(n=aS*RIy
zvd$O26t!)=R>!&2s6SIuh%j<5_l&gn@|4l8!l~yD#ZPbcnh*UGnc9kQEA$m_<9BAH
z7qttLW$v$ry&X05xL8t|+Sq8icr9k#=UL#Z*%Cg+MU#$m-Twoc?z+n@+9Vo*YV3uU
ztt?iI9x*Padd`*TiLEGX4oW}^wKy&ba|2UfGw2v={7eN4o0tud7-k;v6`#Kixs5Sn
zr7+uhWELA1>Z1;us)!Nn<<N<>S`;cG<71k(^*YBRf5-lPQAX8wo*+ee(QQpTad*`9
zdfFt&C?ZFXR`kljSGl+J$`eliYZCM`IZHj^^I6@>2i+h1wBo4PW@!D9+S{5W7bgKt
zja=JytCbVmv*dNLD93Krhegti46mAA255JM4RT}%vF#>-z(6wQ#l}p0;*;S7Ro_cE
zRjj+AFcBJWqf~p-{<|R4PLXj+o8QQd#}r*dsd~bZJRiVpy2o~LJJwfjNj?7gu3-Z(
z9{VXJERx`AAw;?1AOi=ZwwHB7YTx&%GB&;wY5`KBolW=;oE1>pVv1ZdNpj`jl;66{
zgC!wHG@{2?0_bh)B!@-s_Q~*TL~pcpuG-hGyv?wjaPH?+=Un_eo3h0l)-RMBxb$b4
z70n*2JKd}Q1%M1Hwoq(0QZdAz>H^n!9W2MF#NKqcXC`Dm;m}WN*Oo1W^0RxH71L}M
zAyKwpEVL?rrcwzh)|D=_JHLr%Nc&>P1>uYo>TiBjX;00<mrJ0XjEA9`g{?G$h9C9Y
zLTzW!8lCb*bCrH8twNM4YZd55Jx-1MSz9Qrd%X7U<UYmxrfRWz@IlSCjBmt76Ovvg
zhP1!lasF%t8T}VfH~n0CU9JZIprl^<dRbih!-6tg`s#B{*NeE~S<>AcM~1%un-u5H
z^UyvX{KMP-XY}jYhhywCmQ((5?U1A={_1M<8?J5jL;mZrIh)&;u2}zXxyKj$UFt?}
zMyG%g4BPkx--;&zB@K$zvQ{NQ0g`B76k@S}5<ugV*!suORI;Heo|rG?eTBDJXZW%!
zt0ddKls0LHH_P|7ec5`#De@@7>rU*I)^YJfv-9E@KRrU%31|YrMp}PHIeOWHUJb{}
zl)F!4Eh{pEHjBUKPyT$G5(Q<<W%$Cgx9)1iK~h_7C4Z8Z5!?REr#?Azkc9gQ?LpkA
zb-8k5?UQF1PGARyhF@nYCuk{`>@>u4xHzfuL$pvBqXoxo>gFh}*2F_q`YH!rG`_`p
z7@z-7<k{;`pRovGdA2qGKC7ycI>hrunKHyHnnAZR)%0-9-cZ3lkCT@=<yvW;#50!2
zV;b$_tq=PX))V(c*42Ljh~_|m|6uU-C$If7n!kWS--hR8Tn$hjegaH+`p<918VPSj
zd_uiVl2ttFL089B1iR+y6+DZHR+3IyIy)V}1zf3B@|DO3Hw%&34-zyZ_zG+WyX}uc
zfoj(8{Su^B2Jg6`DtVf&GwiOclJ+W1r3Xn2>TNE-`uvQsNN%AYw3I&VaJys-s@W)l
z#y@7h;%)#8BiOtt6mmO{gwDJUIoHhZc8@E%$pv}v;hqlz057sTBZ!?K1)HoBnazDe
zmrS+cK9^2%24AWvfN+$@dZWhIl;xG>6S#a!S@@}NtTCBI&K^~`X&`r(bDMbr=2h<{
zxQDe#->}UT#ICHIM42>*U@_*ZmN7{CboN`;OstH}Eg+QXZIk~G=NrJe>e5|p&D3(L
zkGbS7%<NCW11KxeLGeXX85MS#cl4{q3z3Q7LK&w&NN0q?+O>-(*?Cu!BVdAlpJ*Kf
zB>9ZIjBpoUE5`~6_<q^q<zhw}BvC<Z41)yHe&uo1-FYv8|6Q4YjqK)T3Pr+LY_$r%
zb-l<e37?D?mUkARQ63S6FtdK^&1%sYD4c$(YRwrvNbSn_t8PavS<X&wckeenLy@3{
zYaeIXH^i)A)CQ$iPAD#m$)h-}nxoZWFX>&K5@*@;7!i1^XUtPnm1xWWoOP`yMzql(
z<L<Kne#+!fVY8RXWuV0xn8qUs8gH{?%x<Tt%tqYBPnmiL8dJbGk25|&EJb>mg>&mS
zwz6UhhV8;y-f{3O$E|>qUR}Ju<lD$of%wff8peIhmZ^SI0{tN}v!EsxEKL&S>2x<N
zWxX+vp{;Ld!}2SMG8H)M6f|v;$hIAaW3!~Dzcd}a>LcTzy45Jn@Q9gg`=}F9ihG`U
z9A3pXAJjd%yYl6WV{Z#jJiZfKY?<M(lb|yVPgG$Kv0>?|fFKKjpwnw@khk`?;JY0O
z2DR5H*Lp`1xr)}h&<9&zwNFS+`+On86_#oxKKG>SjIYewn}s0Wj~89`E+sSNGby=Q
z)PeTT-ZhA*{G7o*jAjxiTL|JTrb(hw<x)(dm9_PaB$5adIwP&yfS@Z{mXgu2P4HK?
z;KafVwQo3A($;Px@!{7=qpS>lF{5lqDv^!;RIFq9`>0O>#{QyH)|#IPsq9#+=R{~=
zhnz|l`ozM{r2GsZZF6{NHP@blNpCtxwIyGd>EM3^-w&gmFP+&@E}2~D%gD~7YPV9Z
z9Q{0+uJw~8K)wLjgtSocUJgx5jAfq4b0Hm!8}Hz+uS*i9q|Gh^ce>SQ5)oG{3&FL>
z$~;Pd4G~s@NeUIyFbeyRxke4nbULj#(?PU5+DA_`HcFZ`s;07XX&{R~h|}CA)-$d5
zmx>tTv(*|*i(w8KaLF6We%(fD4m{_KS*8*O7ufyP%VfGuqAIGwKHr^{{5Q3nSNw~v
z3t$$$=@tGrj?}feu%rINHOSX>9o6FxL~ic!sv*3tZnxlW2cGWofz1fj;`VnHsx=&X
z?hLo8PxJdi$O@#4fL6Z3Ndg%G``b>_#o8HOS~!W0)zaf7OX~OCYS<5+zQ7~ET@mup
z2Ps7x)a}dC+&>CcSj%<cuS;gXINjJ8`IC6{>o0&-_Z{v9Y|%a>MZ-wQV+%#VDegy!
zQqHsLB!FQl3^wZ-hO_s)^gW{M{xrj^d)12NVL#grpu^`T;4g4#sjoMYY;RvR;4Cg%
z_TCY+A5WoVM=29Np}a`m?H$cmiAnJp>Ue%2X8VnJNWdkjUpf8D7L0)fN=2%j3^J6C
zTcyn`F!B|(jRqxm(WV<^y0y`*qzbS8V|2~fT^@8f@MI&KWUgH|iHIiDRv*eXLv!}e
zbQzb4uHn^(2z7MOPi3RQ{pY$=yP@w4?2cN-)FBJ(pYYmm4er=ul$H!rvcN;<CCL(!
z-FMIXuBrSE<{1_Pm-Vuvh#Jk-=Eh%s-xvD5B+pL@vN!NRtdy(8YQa3oj;#bDJ;f~e
zbbc2jI77~UJ>-Iv-dZ~B6jXelIqcZ1&@aPs4n#!>GQpk(s3tGC(jROrx*WU|@;#pI
z%(e@bKK#@3>kn488PHoDLV0f(bMT>dc=RtI=9?+W_Lb|0^Qy)9zkrCvFMk1{O=5rp
z=MbDvUKn>;_menES|S2(E_$=MN9$v?E-En*Hd;S6+Pfng4~iLpGT-jruSl24w;n|;
zPHtKYSVJ1>Za@1F*GRl9C&7DlXg6`r7%0W?rgY&o(_MIaLVBXsKkH*7&hK3j&sr?G
zzotLaxydqDHLMo!`)#jLQu-njKh#VoRRQh!rTI&0)7d=opDCRx33J_=Csb^n4_)~F
zxII!(4a|KKNdtRBI(gP3c<&Y-)Hydbj!i717XbwMbr7zdot%c<=UHP#adbFM0YR-_
z|NTt<KUp@*&-?!ZEdKm_;4wee$L&m6LrX)|uD~*>gZ@rtkN{JhFX5J!dV_@4l}L8b
z34IjZw+^Zb?5_J3*%>!(>f3O!!yS9J_dmkg@qOGLUU)t`zY7Wx`2|@yKEtcd^d<XX
zl%pXwvMMD=oqPxWr}kiNRhYmDwU|04DfQ-Xpvzpkl*BYSz;m>HesIY7dvr~7dFhKg
z6{2&Z;Q!e$IA{OW!Zm$4p)OHq;FX^9$#^L}f##g5WQM8gVE3RuP&gF)C+gM+8E>O)
zn_}TMn32VataRY%u_q-5ek;|&Dk#~ieqFgJ27U+81%3JB=N1{2<Y}OGGg%bjr%yJV
zmOe_uQpTa>h^mavo|eotwRvF|H<|T%F(j3$o;rr=eBNscm=@x&4sW(!j4Xk}8a<a!
zI0V@$D55eeWy44Wxz-S)aP}UbkvzMUMeypX0{J=9``D!6>#Kmv1%|V)(7=_v$paJ4
zLvGJ97oZjKIg7JA`w305)(TiaCfhRm=Um|(IaqBvkLKyt{*T!gP8Tx_h((UEaeOmY
zHJ(V%JLb*`It3K)DDtCVfR<OSFQghVOV;VoZV1cOE}EpoOCn2I?_rkDSoYo45=JrF
zIgY|bN;vQ8uTSyjJw(Y_Cd>E`#7Wv))22!pldlU{i`pqmm>ZOKK02O@d{jPB{zrC0
z#+clh&J}r&Ty$F?WfS->v;R+f9FGkEfUin!rwu25^ceHBr!9I})V8tq!4fvKan6=E
zBQ;dW$li1>5@@WG-?PP4uP+vgsb2o}O<_NY62<7+&q6nL&<3>0Wo-FIgQJ3q24nbs
z)1-c513o#QXXqj-PawPX4SYjMNeQ4}AX;*cZ<>W4##r++x}C%1+fJN`j&F+G4Il&!
z%9qT(;B`<{_B@+Mf4-G=KNI`4^5^lBmf}dD%#?K!ZoVPT*+CZQYr17$7IYl%g=}3F
z5Rao7w&?y<59%Nl>1r5>%7ZX>j4@TDEGK6KTt>!w=qbu!rI5uB3<;Sv&Tn5Hl&AlA
zQiK`42M_{=o1ZzfyZyqKJ;9ewvxr!{8o8fno;x2Cr>MsX_lb|9un%YNlCOP}Z+1v?
zW)u(ohh8jnnuBp`CL?kK^QxGl2eQ|qK_l!{Z0y<U9W4HX?SB{*w_-A7_{>uqY;31;
zv+|QTCGO?u#D0-i1$H3h6e`cF<E7Bcjg9&-u^d&p864OjHa7oJu2^MjM#CJ%`S||d
zx>p1R(B3F~as9^=M<%D5$r4mOocdK39b|gYTfgjp>4VcMl~tAvE}%)1zcU)*`_!r2
zEs0T}$PjW-kbK7Mvvrx3078O4zIpLXGT(%<d|+3$4uwa@T&WIonhGLBOH$eG@lK!1
zmEO-<9cx~&20T>#g$f`0feBYUD_W;C_p;-0>l*8r2w>{rY}2jM4~u@DtC2t;sA@7F
z>7-NKu+7_)hH{}T9<8iZ4(_I;@{0vS;6*8yYq0a*`}g#WPv&)nK5MiVwO=ezU1T**
z$K|+w@^gNpRC>g2xf0N_e}UJYLzd%v##?{+ippAGTj$z>w3fC$?Eu)FYaF!lb-AzA
zpv>G{-761=hf6F_BWEeM)UW}ip{#K7i6zt#MV3mX=mom^wVd02;fx3I3NWuciSvCv
zOvP_*3~kwYWu9Z?;&<O}&)FNP=~V7vtRqNUZ&=DiM5Px-%(qXr=2IYVR(hCtN$dES
zN!bjl<=nC#ZSQ!qgm&LG^?vMhfldW=Ha!UbpZ-w)%T*q`hOoJ;QL98HPZUv2RGMN_
zLwS71NQsGpg+0k@@R7w*meH)-8KgE@2eDP!J`&hj#~|it@Jf2FWaGI2(RF(8N@~cU
z)HJ|nVOA{gts6R2jJ{ucn~UhTTw-5XPu9}Y%`ZDFsZ_2-NOWbT>ngLvCy_ArhpNiC
z7wXdpGUDWb-!WTBK(dOQF9FLHzFy<s`ux{>vywfmd0DN@<-(rDq<yUZxZCS*aPCeY
z*-s+ye1!6qLKx+DbI#{wB(F_s%mtr~5Hblu&H5CYtK1oEn1H|{ILW(v{^Y_L`R&68
z?<0&_hlMA;Jy^as_E`?q%EgubOb7;KYZ`lEod(-px4xo)*V<Cn)_`CD4zsmUFfPB8
z^?wqVy!_XS5BCF0_8#a{!(P#bIQ-kR`ZgN;>_3q6|1>gAHL7|Oprb0)SxB(C4)37m
zdX{j?L%r#75M{NmM9~z+SZES@iX%mTeYAf^QnYJuIUXAl?uGfld}VKq9SFbj-W`ot
z@0oK<Wgr=gJGT6ahy@8Tgiq>T&(g}0KX;RjFG~lB^iU{y1T$6XbDM~ZEt%bKO>nK+
zi$$9>a}i}3QnV1)%at2Jm+ET9!<cfF^WK>%A?AQ?mJ01_a}tlAF)RB_rP6CVvY{RA
zS5#6A7s>7}O)fO~)@&rvcTKt6UB=~x@7*Qkn#!>pHk<<eK*4fIh+GpwXXizaEF`K@
zfndh0H#u@XOt7y?jp^i53tEFGIh{r0l}!r}<yeojcj`2EuG{HIF*!`-4*611ZY^vN
zx{jrS*QPrZ>-qSHp5v9nTAgi2dkSZNMo=ZZy3veOhzQV0eYMyKk&oIWEUIK6A1__8
zDNu<;9GEOR?hkWia%cRx6H`rh1sl&t1~6iRKw_<GGQeGU6|(WZb-LA8Biu+W=JSIC
zL)iPoz@V@&?wpd|ruLX8@T=;ud!ykug$Xn{pA>AC0b8sbhz+@*cxNS&mV>S&nam%S
zR+!TexRZTGezj4D1lPMFr@K#h266JpUjoO!4*0Mhp(pPI5S6M3rrl=UpAbCcPVj#|
z>i-%JJX9T10P?bn+`>nwUoQdk8gKAg&H#c32RM#s?4jA34PAQwcF`zV3i?8&@eo`6
z??M=<U?j!!XhYJ24`8CaZ~Vtp?4YuLtGx7X;Gto$-A1m!?#KsUbNy6>z%{mqhP%yP
zo))5_IN$50VS6Jht*HB_<IiE0)6KerdhNapQ*;0vbG5YN%}y>uj6y6;$%}+nmrqBC
zc!nu9t0fMiR2>DT?X=@WHWI_X!$B|=u=N|*T4rFo4>q4;*J#LO!03^N<H=MkXkYL;
z4yt3x@qh<u@E)SfM_8E31U_Kg5rbDk{fd+ks=s&+1X`d0Nr>PR563Tb7W}U@VLRi+
ziw>qWKzO0gnD&#?GIb8VxjckSiz55UzH3e)c%L1+$UyFc*jkH|rVEvS(YLVbV##<I
zRf2jY^L@k#md3%zCF|8pw_eRyS=7m@^2Vc7XkzlgcgV}Wbd9+19Ia<`@!BvpJ669z
zcjHGda3>EjHJP#(Ht)T5$Q*0ClmshYSIz39Fcp`k4E%$KsqcMf>S?i9;oD+`2of@#
zB$jW>MQBf<FRApgcnx;o7~<lZH8KZ@_qCV`<z9}uuz-csFd)cjv8A#llu}I4I#sZt
zmC=YjVWtoSRB14Tt@x%As7qk!kCk3ZE(0EoRKIjteNW;Ske$7<yGMh>wE3>V&`!#U
zhOr_2CFT0g9}06*G4bW-wCF}*{h@=7+Y3WeiJFCfAO5^WkNYCV+h)_ca^Rj~5=PA^
zxE+p(j-Mm0Rk9d>p|Yuxl*+#Ia*c)1JS!Z@lro0vWxG$>L4LCa!nPXNAhrWVnm_08
zck^mGVM%D#`bu0{Cj*cdrt46?Vb{!Bd`V|N!>#E&PpfpgTp$y<6e{FEYdL}#ZYtw}
z@#y8gN}kA5X)fhd4L)}%QN%rmcE^^B3!U7OF7>tXk00s7geKp8HG+DTHw{|or?TW}
zlqxu>$_Fk{2f`xU7x@-dvx&ZW%d^6nn2T5APRR~bmDNrsIzgRp>{QoPFe_aSQ|m<>
zHUp~8`wE;3s%>S|#pb*@Gl_>IRIRLIiSCD^ytN#X1lrXG+PXM>?OcW(pYO{32Qjw>
z+0L`DV#M{^uNM=zq@-tNK^soun~%I{dFzBUYYfI?BZTJ2Mmx9$Yl1KX{!6lgWPZ7J
zs^#MLK67-7HO1oyCSsN{t&%mJ<VF=4t?!+b?sA(h{<-1klqU$a(F_~Lcnwix7+<RI
z_S{`5FT=U0Y`Jo2;Jf4H?Y8ig$+wk47q59Dm_-?Qi>EDW7M0ve6RksM??_h=qCwr_
zbkBu?cBsi5W7sQJuIhXyf>EG9F1&A%TuYVn4u(RGyWt7tP-m*VNsd}eeXsLwhiO(w
z8a5%rb(wcBrm6ska~#*TJug#(ud+UsYraZ!lZ$l`R?*dnAU|$>!elhF`zy=5&7^QR
z)z&~KjD&C(nj7z`3F9Z{atm$^S}qc2>NLzIOC{d@MmosNweU|==3D;ki09n4eSYOp
zG`Rx<d8!EnsSL1;^IaFEDif4LL#0L~f~82ot=(&`umvTd5viEXKFgV|mG+_p8~MZ{
zQ(|!Y_>9zHidLH<)2lXxkP-yv2vnOcdwvr3`%0Z%PML+hX>#dpX=1io<M+nzI3@2x
z_U^8p{{<YrA6WY%)(;4Zz18Z>PH)iu^D({oFW@uLTR@O<CQ)b3pZfYg$!Yd~0kiyo
zpy@8s?xoGY0M1zNKmPl<$0x`57~_UsEG+(}b3r*{;_7oJ;6*f}_Q`$;JnQ@RJl~9;
zna;ilZtRtDjMh(JVyWZuqX^+jszblEATYR5G1eqr#VdhjD^KC`nXwl0XagV!-Rv+f
zWKi!MB6MWwLQU;3viA6hkX&%SO60sHK)X?CQ5Z<!zfVy{JY3I3`(*4}*!z4VP*=7}
zI+ZNgQ01NKyB8_KpDuHaSSOIv8SNh)a2Z&geDc8p0YN5}<@akjQSj=|l>R~M?c5{7
z`$%g6<A+Q9x>BzooL7{#aY`k+8%vzJWn|k^z<10oc!wQCI}j`MEdeK*_O5UUCDjVu
z#QcGnLoBlf9qA!nPV|P*Z3Xc6)U;}|KlN{#*yT`U;GgG&?#0DZCuDE#24Dg)I^Rg#
z$J5=w<eCEx**j8n)9mhqMz+x%HR8`KCSs5)D=?Lx=d_hgeNE&GOzg?nG2iRzC^Zs=
zdtvwOesZ+f<`#f)2<S30tOt4wXE&3T`!CizsX$cM2<$7nqFFZ^T&dPleyhOo3LHkh
zlqLefW{q+c@@la22YF!BK4dBRtAO!wr&r+*AsZ(Nq9HtE?pAk7yC!n)Wxd?k)Y)tu
zr)Jx(zGjMaA;<cNWWQx{V@8y2qis@E`uwLB2qIf5Q>;@JHEb`q1RYcVi0^zy5L2|$
zK&L9)jOBO$@69#d|H%v%fr2h>D2&(}tY)~0AY}uxWvYqnUl2((_t6gu=w}J`8|d=0
zoPbtLvaVeiR+j*-z#Oj?dvdpRZ%cbVy}nkb-KLrms^d$#iaRPK1|Nv*cT>mP{|dz{
zHSR_#>H)1(a``x;Jl&Q)MOgI<VJUPAw(uQ)#pC@2nCW~E;A=cV<G7q!Qyc2lJH2hK
zB+?#sVud)~5bGrA$AN!*D_OXd^<Jd{Ln$@5B=*(l2or=JA`7Pw>I$jEY6`e#J7}$X
z>!C&cY*@H1C%u1U{BO$j*~&<Q@@CZ%EhjYbNL{(Rj0K4+VXksHwCrc=Uc(Zx)Rw3k
zB{L#rd)jHXy}K~q?KzLMeC{Key}prZ^Eng7K0860V_WYkL4ySgFhUcq1?MKh9I94z
z6)DJ0jcEvKpsHApGBsjxQ1sF21@B&JuO~fw%}*IDEuxNkKjF2+eLtdR`fpTY*Ndtg
zEXcL%a)@QaGh>DP1#B?&0)iwA^-LwzmXaExC_YXOoNZ=CJpMf$m@abtK{;+G+nCck
z!zDT{a|ssE^`fRq?qnX{S*_TziUpVAc&E~^vu-5x45LCUbb;K_siK&2sS9Q5h+<%g
z%7oMAV@|`~sXI1ltr@etV=exgwjcR26ZdTKf5nY8hN9z`R{#r@@9Gj&J?9!IUiLxz
zY)Xdm_Z!9Il}9{Z(@l+{if)*P7*p(&ey_4xg0gER7i+no;iU@d(!^I7oE!5);8JI*
z<5yAEt_)5I)dqH&w4N-%`kDzk7pH0$`I5D#N14Ew*&1CqQ!K<y;NatAXTH`mZ!qjP
zUwL<NHdww9Qp)rrfA%#W9>X07gShr7+FGrxM-kLAPWCN34bpXVmu}25j#VpEEB3?)
z^tRR|MLr|hFVmXm%EUWr;Z`-6j#O!|QkB=I+UF2l%F^RC`U<SUBfle|FIs7;Ztyx7
zQ%!DmSVux?5)Z7<pV2E}uPIzJQu#?S{`y6g%$jZI(cEO^)08wSP>C{%x$UBcxmTHA
zy$gD=Iy~i)nbKFh8J+it?cX@)Uk`Yq55&e@_xjGb?GtiM!op@8pO|1YHQ6tM`Q6v4
z{sKg`rE?Dqu`l0&0mfiyiC>SLudl9U{sP|BV=3y&sdxa=#@_)rm84;B>*o$9sKyC&
z=P16zsD=2p3HSZ!9MMkkPj_{yMkRt0F{IR$2st?xF!Ri7^q@v(L2tN*CZgGtH4-a_
zRRH<i6MT0sM2i8#>)wm~K6I`A&f%RyVE(A+T~CEiedX++csMi(TBgn6aa%kc`z{?X
ziS?+iN({tTsS&T0M2V=R8HrnzSBdi`M0^ka1Fng%pBuSz>Ah-#o9|8lzo8iNd#WM%
zt9py3X0y}wZ?mvs7MW9{o%YD>S(_m%jbvZmBcQyf!?0}agf5*HK;ZLn5fWn>?grv`
z|B8mYJ>T4f42|mJd@~qhiyC)x<XGBmB-akLGi#3)z1ZaST!I*U8mzkgco^Q-8q{nQ
zVvU$8R|{1s8_vO>mUuGCJH~Jlt#=;NsAe{h;q>m<;JZ_I{yOiPuDg%8|7PI*8GjOs
zko_z*=gp^4k-D!C1XR{MZB`>wyThTmA4DvMw8R-dMxVC35BAe%?Mmf7T1@e@AuYeQ
zY6TMYyXT{hICQqFH)10g^EewF#t%zYuRb9xB#7(l_hDMO8jAqY4*a=U37l?K4r-*I
zHR)5w7><fnpPC~<Q_5NYIsH7t)IjUqMo_E5Rj;5f0v0Qqud{yJkK&?Jcz8W3>N(uR
zrT3H=s`Ty^<ju0M?f49RrNin-_DSqQbEJ;;rJT4<0Y5=a3TWhk{I3}5hA6po9nWgV
z?|b9w6%1U>@6;6bQ8u4}A&!$dLCw&NQK84nxF;*aNPxN5clzCRp$AI|+q1P#_<Tl4
zvZ2`o1ulD9Tw){V%Wv9Q$g_pzRT@AI%erA^y9Ar@Dg~n+Q#T1sHPSk3kB3tUBWork
zJPhnv7_*gn)i{c}Oyj}k6huxvJ_-7-rC8N#rfpwK`F|zGm7qv`jrLAqP=%+b)pKp$
ztesFaa`k>6Z40*q{BaF_5w;Wb7cjDz{r<vWNBmx-`X&CMW4gnUS={*Hr^$MY7-SI8
z#K$VaONgw$W!>}-O0LCe*q&O=K4_^JYFxZbEuW|66qkX~NY%)elDGRL>s3TKpQzR1
zosMfl#DWg2ZFP#Xby!Z`QDoMlD=?wp2+3mVmzYuP)$=T={#MoFmmVS1t}}y}b$apu
z<qKY|YF5gY=_^%5=(2xz^qGWFiihJcliH07hwRp~UNjz|M{rzKDeKH2*f|9HX%w+?
zXpVf0cKJH!<uDvOrqp*Gv!*^}$}_x@B)vbP8m}3ix}~hT=O`W=<eqL)Wv)ZFvj#ZO
zW7jdN&}}&R3m_pS1|r*4JOrS>p6g;*LCr@{<&=W)OQ}{Zwv)t0McG8Vvt;y%I#)A&
z*Pk@6hBtmCbrW{<Y};+O%}>V=8E>hIH)Bi%uU~=wN$v4i`Z*O4sq$J1Ksw}ZPJG~K
zb~J~{y|h|0DnRRO5q;A;-iI#zQD9+;*HCv|6nH$LS*GbHA>x3r0*lN2PR+E<a(!4t
z^_6sSXSt7#@VIPQz)<0`fh%Jy|M&AtW7g$n+u4@vnm=02B-1oarrNB-aCyJ{8BwRV
zPFS9%5hu>DLDfjJX{8t8P97gdqBFTKR5v^C?2O-#E=LoxO=6Zoy@o+A9w&1=7%!0j
zj#BO4OTmbd(^Bj1tNjz9`x3ACo(3sxUQB&>)vd-{3xvXmv(wZ9fOwoUdz#Xeg5}nO
z4G3>zrf!7P$Tya$D!Q1mOZqewM=Mb;S&&FWKfBg8P9^Fis@|KaRrc&j`s4J`;`-0$
zoK!W(7e9L{9;bU@!$6FWS9)EH-c|eW*YsCI^NhTSg-Ko%E((;`j?I!gOB^xrcr!>9
zE_PeJbeLV|&~^iTiH-FcOc|x@mB1O<zE1aBCDprfjPTJl*B640c)zLRO(k^|x~Ntw
zMtl6EjH;wfa?<C2US$k-OX=ZgjACY=#jVi!LH+~Q|07B>lGLH&#o(99c|_B*_@^(a
zNh3h5<{&M*%|*OS-T;w1lCd{*GCr`i@rJG!7eVXZ_f?A}0JHh7C~Tn2fM@5zw)KYr
zOCEf;JCV+td4~gR4%y+i58cZQ0GBrW2i!h%it!r%_-qHn-24V|iTyOVg!%buHsCaH
zq2KSyW3UT;b1D80-RwU|M@hQ}?gPW$cFr@M{{=)=C<B6$R-ds*?x*y}`~@hOHX8b<
z%an42ca~1MiuEFfqy>DOa@vh<f?s5HhB5>!Eqp+3vYclQ|HwD^&3<IjUgudJUPc(J
z3afUvLQ$8_tw_4>lMY6uH{K>1v#u5@XtxW>R6|2Gqav(StQv7{jtx>Vv4O=+t&C?4
zznBdL2QirmIz|18O-atn33+czhfoJmg=%S#j4i)FjZDcKwX=B1Qs`s8thf}GLBn8w
z)Bgru{?Al4a}T2L2R8u9Ic<G%$dQi?xx}Qh-y2WuW=H0Dw#N57TF_XO;O4<vsheLg
z0Km9}uM5D@cgp$Xm#|mjeR?uu=9##k0%^ZKhX?E1T9s(oo!%94Q2j)$qOJpyDX?77
zU+pi>9%jUhXvoi(pm365Olz>CPW_Kckw)qUjO|^B${}(NceIN=0V<?U$AZr-fBTm>
zO+s}Fy(}MQ{|29#V@(oy{Rd}JJ@_wRJ75(5mZfS<Dp~B6`vE>STjRWmW{U>ju-+n*
zo^(h7Z#6w^lwqBAtT&U1G)5z(Avr1*>^?X-T?(Of1(D2h!0lK+H(Y^e*ymFW_hl`M
zzR)%pc95^ypV5J<jnbJH>EfQRgbHo#ZYp(H&K>}RihDWuS$GgEZ5i|`i3^}h+vDA3
zviyo6E>M7Q@Tk;?af6RyLviPab!P0Vm<YL}OUSN1SBw9US0j_9C5HCxsCCA+PLfVg
zrlOxV(F~(qI?uj9l*53Sb@HUDePEwP{XseIP9{X?ZhOSK{p4+zpJashj*WmY|I@XS
zA62x&c98;~pj3`vx@I7zt8yqBQZ9O2mQ3hKn+Hjr^3&CRhN97i`!(^FX&+~t*#)!b
ztzOSAK8e@k2syw0viV2L3~Em|!**umu;f?V@3sOJDf&(FPhRoN4mw`-!}&uIbS|JI
zmBSm?W;EaW66dTx1es&zcHpbnR&#{0>#Ms)YH?7AH5kv#RE}gMI*Far{wQi{$ala`
zsar7wTZl8gK9p)}hIUAUS~~*XhH`oxC-b}jip%gTClL*pZG#t(JYo}}+BLs!jrX(1
zT+FRbxaRVVPB^JRdcQB}!nCSc>slPiY+{*!Hi~xSS*l<$g~r)mSVD1qZXdBhjpO3l
z2eWN87e7XwMn?O#(HsjA-jkTLD@=e&6H|YFPldV@zu=G-BJG1>(HQH3`(RW``gewu
zUZ-LT)EJg>eueu2-x9k7WnV7}i-B@>TE5`C%5JcKia{7cRacus4$8)Q51`G&udWVJ
z`d<ybIipU8N(@ehw4xpXImYu}^Y5stg+KqCVq9)-nIn9rf%!?wU2oXKS9cbhK5%N4
zuGa8AopV&sjdrD`@=m@KLc*{Mb;$N#9g=MVjab*dts470QvFVgy{wkw7te1tw`ctQ
zivgHuC-254=ka?Jwy)n<7mN$1y!5vfxU65BWd-Q4Yy`%dbvh_lbvi%mZ7EW9GJcy8
z`phcBQr4VlX2glFcy1#)AX&bpbf(n@>!HiUK6=;44ss;pnmeeR+6E8I+bXqC0BvAP
zWm)ZySFW0@2;j@rD^|$bZg<foSxMq502dCs$@Ar(CcG&`s5#o5i}tNQv;P=?a@p>5
zCXP-M^#vTmF>3x}%tmk<8<4k?k~@xz4ig$_K^=1wopKJ_D=SS<Xtd|G|J?#LS<HQ}
zkL+)R3ImhHwuYW9$fd=$MFPve(XnijaIbgZtcU%IX=?SFaxfC?E%(dwFrhYfy#JW3
zSTId1%q+FD$G*GfYM7AM9fPj_XUtE6%|Vj=B11=O{un)Q@Oz=d_obkOG$A>y`wRb-
zWXH@~?voJ#DoXe}KNf8=p!O~}Z^Qv!7<l7Im3+>(md=6)wgB(0P|_CE%tg!-=d&a_
z|LPo^wvp^Q(wK05e1OYu7++iGpB_Wr$N9w?>8VA9!RzX`F*3&6YxZMIX!Qb_7pY>Q
z2D+%~yzFM>3T^tx_sR`pd?=tpscTWceBGks)U~nh52866Tp=Ifgu9jv6C2l!mhC6R
z#{*FraLqn6Ox>v&oXMWy_+SoMHs1bXMLcn<VmLy%@a>*~!<LdL*=jtVKpD}%>IL${
z64dY#SI<zmM`Oa8=bhzd5Wa)ypppCtQ_;9vw7<5KL0ijgw_K4Fk`=P|h;UZq9IS<G
zDDnV;49LNBL9fBZ%`~&bK;r;P(GnN_YApPYM}tx5rYonYsF3X73HsIR|L2aspOdYR
zwR4a-(~~JxtzuBd+6OMMFR<6JpFzne7v(5q58lv>;KEgji+CS3(>@(9apmoi-CO3*
zD>Sa(`t~d`Q=%G+?2A7>5OwnV2^@bZ?mbmLy2>KcuM)?=muh$7h>&X{19Ely)yZLG
z{AidZpU*Usz1IR?JE=7v?VkLq+miCVA86>HD4wworN@*zv)0$a<E}&%dmLuwj)Yt-
zDA~bmDOI49=Tlr9ZyVRglapo2rc0Q1ft&6db1>EwF?C6WB@L4kyGl9r)(h{m#UOg*
zA@HBnY4P_>Ba#K{;wBY_Gd`zYQoHDWvv+%J@(Q^QM(+8;b}b&}a>!Uc(<uq#-_ImQ
zB3&(!I$@p(ArD^=b`r2`#F3lnusGa~u~M!)ul6_a3QN&T8KraNZueaAI8a9P302X7
zQ-PLjl%n6%N?r=AlPzfDooS@oapu171T{O1x#3M;#YZ7YNY{%NqZjhr!G{47wV<JS
z;}!RLW(t@UNu;_(36^-TPz5|A-!kem`f5>|rDaqkrk&*?!^9ve+Hq*5<2n=~dQYMA
zdI$F@7h2W&VvOEvEVkTg!b?!|*=i<Vv}*DQOh9O(EICSGAz7Fu<?u5v&|AEbhF2#Q
zj?+$|JWo=Y%2~JfcJVIKNBIhrmet)gOy%H$PsYcaF4m|O9TeiQksQ|UIA0|^Spma}
zGmAl6wNXI0=g-W*sAK|<{hnB)`RustNoLT|NN+IzlLi~1(>hRMOyF3y)KqkkYInI#
zx{F*Rt7cF-S;M);&pgA}TBZO?kvr>*OUtb?$5*PlGD@J8O)V=*sAGNMl~v090WQGp
z4p_n8r^eT<lrA?c#bezmAOg4Iby>T55ur9ujsK|pVPjvaVsn+!T)X`KK;ok3&r>t`
zFikQ01WRcMOtnhhdUw9J@RpQ6d4k|{t(BkKep^xM`fXoD%m&MecF8121+&XqdftIC
zZ)D4Rw=vI-dW09LQ=|=?rdWQi%PcDB2i}pGb|xpH-4cG=h}J-!*?JZg4P>e4pg(JY
zp2&PlhO;R1t0Wet(?S$1vBg>ePyqm#lfsaTqt%B+4YnNpcn8m;W2i^C+9MWNtdycw
zCXC~z!@Fnqr!9q9GQEP`qWkG`ev2kr4b7(USh52JLc)Rne1hi=Lm$Oh_GL$1hZJwG
zs;CsUc7*a_#8!lsU2bHmRHC&S-U~DFlIjDKm}1p}PamglYAx!EXn**1`%jvX<`9>k
z%WElLTco?@$64P)voY6<I$-QdZAL!S5m?kzLfdR7OlKeXEx{zOZDgCv(2bf{t-+0K
z@3k?t(dp6d>G|mxPB49skCqO4;nvNn)h0*X)m~CJtXvzeTyE_3De^l=el+-IV>Ce}
zQXx+V!B?vE1s#KWkI=~L2~-jk2oti?`tDYv+7t8@tvB!ZgJen~ZR1{+E3*!wmn=ST
zXtdOy+^7-w%zUIwSgT;Fq(9%M_p`OU8(gTJifKY$oo`8}K`7K*ZDNNq>&b8xmo<1c
zM6Dx#{enuTmNk@s5?7vSJN`i$6zI!T^mNV!&)?pzZz#^%-pbOkH0ImL<MK+!%mDv8
zA%~L%Vta=VglB}j)LbHpdc8z78&aY++T<*AWl42qpG*#+^DwUXSDZ}k`pr>c+KFRi
zrSM{T$<N2dLq8MO$1;f<k_rMnr*LPFM|z5zIqAC%d|ovOztk&Ax)LtJO41VVLb44#
zO21`;nq2qmONANma*tMs$r$a(_mgQ`!ye(45Jz~cKg2_O<1sH%bui4Fm^Zslo0?&m
zX8bO*J=$)xRAF=0=||mj1m{kP8_gOki9-p_VsK<K^IV<MhMe&S>0A<j?RFf*BiSD7
z!O?&Y=WwLEL+^mY?2kDH9fl$uUbCfG88{>Ad!_u;R9W({`KI=AA{rJ%L5=W)W!CD<
zMEn?B6H@#qJ~duvrH@8@u@RZFHUamh;wio!-VxpQ=D8BJ5%*vAQH-QeNvh-U5J6)k
z&Z{ArGN|PwUbn<>{gTn(rJTJ?h%cDDET)_bh2>oz<T43}Rln%Ld_h=iiY($Lyc?3T
zs!7RCIrK{U%v0&`wm`jT?fF1Ppk0f&nw|$frYYKUkYZk8svujRODS>GgEm$f-299)
z+FqaB=7aG)>zAbu%(o6z0jryTYD3_l%<H)>JzXGd<fg<N9t|m8q!85JlmtC8$rM}6
z7sMf(`53d*<~m5QUG&&)8zw7JolDem;7gji*6}Rd#6?q5v~yTUT3>zLI2SI?XFO10
zoa^70{rYLS*3TJOo1q4}b~v8VVir+5r$v%W3sP`jc?}G_m>Rf2k0Tr*>LdPV<9oGY
zbx!zS4{wh@*81CO*d8C?I~ZrCtwp5q`(^tDEC;qBJWD_s#`Q{JKvl{KdT6X7auw2u
zqAct&ZwA@>DMk;t$qaEoXJPk0P+8Ih?nFGPJo6?HH&?iW^@Nwe!YobllgSD$CkI1=
zm8P;IdRQ!PnLw-GPWkMtQU<)Id_=?}!$Tj`)a#yOKvnom$R@*Z`Vzg1C8@s-JT8Ji
z;Z2EqFBD-JPwpn={pW3lReRuy)AB!R#Y1rRQeGCFpWX~8<!sh9$b5#mW3No6XhzVr
zNR#tF!Kk&2-PVeG72<S$?bzdEa*njkwCBI|yxZtGdsKVL(())~gdgn0IvJkYJ9MkR
zZBTN^pFvv@$Q6IzCr(wKAsbQmnWQ$9@<xdtFObAIm@er)fmwt1eIvs{V@;#jn(|!C
zDx$%lo|QooH3TKea)r2FG>D=_xn?HUdulMUX=Gf$6*VrK{ff+^5(+W=^`lk%>e>yD
z!}z$8Z^hJt#LLBo&czLbs=7mNfchiSgx}QtGtyc`)H4r4_jqPvQJPOte6I~D<yB((
zp&H5wtbBHHDy+{_h4{6$?fo3rhv(MD?_XYxjjd~26hHHi46>%SWVNb^eqeuUHx&OF
zdQ%yc_3j6xh$Y2J6xbXgutZnS;AqTDYdIv~ZBy-9oUJSdk0MUwIg@@a8>i3gFVXB(
zXp*)a@oOE^>yN&WPg0UigPR}>9fs9ax~x|N&)DV*VHLV;zust1@NXK%_xo827qnf{
z)=R162w(3`SrXHqkUKE4nFKx6d9V}S{w3zyq!TCU$k=Sy@4GS?J7|DoMZMc%TD*?&
z5(Jy*7yCJ`xAi&V+ou)YZC=#^v718LnaXi&vNO6HG&nfp+G*sRqkoy{?{K+Pu0NrE
z?yEO@7;inz5#7M)X;R|8S(LyOYBTqA#r)AL4?-(*N4}`dOWd5F6hCVWVy;^W>34G=
zZf2~9S8?s6;Kg$kKTurW9TbzU(_?5ufr(M-A@tmRpTKlp0Dg_W6l}LJ`Cef#dLwXS
ztYa}ctgcglrwX~QeP{o1f0<ojGmtFB@HL9d=ni?gx{L}H==u-jfpnbTO84yW*!@qC
z>%=|2dFfw(7URN)BmBV^SFMx3?ruZlKK-2kgR+yp+^_u$$e4+r9Q}~$-tnKk|Kk+@
zOB-8Rl&<%^-GL?3oIzfAokr1Rk$z`iocx@BUyd|e;OX@hZeZ^;6!yMO_fYdFz6s}m
zx^WM--k<KZ!Ej;_>agg=Qg7VmbP!iu#INxcO*O`a%?Q!Y*?8BOk6n=xLWY1cRCRW6
zr&lThg;(?Epo-G;)AnYEA-5Y`u#wVn@sXNNTa~z~b=(Az|2PB5Tu#$0krtG7E6hR3
ziO*;+l{Uiu-M(8MKetP6vABNI`gqa{c~J$s*(CKn+Q_-UfVWFyaK>jKcurf_%&A>k
z+r#F%_>2Y2?7XvnssJKtPdA{2LKrri>G<@B){$ug>x&PhDS^dkA!0^h)tZrFYND|q
zq=hskSGrBFl<&#W#1{V=syh9Il*mf_mYZj&^S9cM^(9~X>brf{?Dw})8mqXbBdZhi
zkeXv@i*b!we+nlNp&Z8bDg_>s<2EO;k7K;-0)%+v--{;_3sb5j_5|kTsa_KpMJ6tr
zGg~hQ?tHIU+D~1sy4s!(=COIvV>4!AaTGmVUq5Z7!)?v|fD=~#J3S2!DTSJjR#t`@
zW}dv#+F(jj&)V`xD$i3aU4}nbc1=ntEW1DpAx!SYb)A*6^=ModwUE|*OZo==62OWd
zLL63^ydtki62!@BGRb}@zyI7*t^(IEw-T#lN|(=o7h?w?A3NMYfJW`>f@ef=1D48e
zXN59A+iAo)_!LX3Gi#M<WyLg||30}jf|?ioEMu?c66g5O)lduKc>d+OV)@)o%jncs
ztzF3Dy_rG6Y^~DS5|#n>-<OF?tLkIq;j>8losPC33v#&hW0X{rs(fV3b^Jniyohf~
zL>%XE(Y1QB)mXK`#og{xN1X%65JXWzzw~5ILkt?_*-oR8LF?zhmB6Zdx%Pi?cUD1d
zMS;4eMq7#$w?bRAc=6&z3j}Bh!QC}b+@TbACqSSSC&5Y}3GPzdf(9v2++Ew=oO@=@
z!+p8uJl&bQpZ3F=J$uhyzW)FFPg@#)(sNsri0aXQd+``iO9^R0me&#=+Btd&mdh%Z
zk2@smIBA2iF`168^TvMh{KN`n@&*}ST_5gw=f$mk<PEKT|KDl`|5fP0I6vSoN-y>w
z-WQ1X-ODnP*PSwZ)r7$-MAU!XGGt<kan`-<e|WbsxZ|@Z1sQ?iOw;ur=(Z!JlYe+S
z79Vbu?lglJa*c5tsFJwEB{9iuwh$RCvF_jVDVeKS^;G!GpJ?!ay2xrvB`1+W7Wg0D
z^Gc&M2cxBs#e?WW%3R^d8+K93@rLmOK2O6EKgPUGs`xMbOUx;EgCu4Zh9;%ldP!tK
z@sV7VDy)gmbdjNlKmpfZUO&#E^YZ(`a}|qLfc!<@>=rgvK65cpsjAY{hYLYL#2C{~
z<At{^71vAnx*}qJ&>HcLL$4+5hA}fgGB^V@bDa6d<w&tO_|Jy8{6D-cKC<BRy1&8O
zjEuOTsB`+iLMi|7wl^8{n!@QmckfB9&G{Jy_gVyhmEsCM<e#9%38;Kh|KU{@-h7B`
zyYyGW!B<$?ZhR{fENfb?yNuS>SPw%p!^LBMHx<xF!LZs&Y~x(6a|IL8JRhUd0r@k(
ziYN4aTI7UgVV-PObMJBie^_qcZOK2d=bBxU_hDA^lS-+SPwPBQT2||>t3T0hWa)O8
z(l(}&0}rQFePYjjs@Hn_F)Jn8gPn<6F=Cfb*%ieJA3K9Y!|NStskF!0SKuF#Gp+RX
zOGI>xNo(>8y3=yP5~lx<W>L|JQF?l&@}B*onpTxk$Tyq|NB!x{2&8`rdM3`D<q=8p
zHNc|5@lYAoKF+b6%o%pB&nH>S_9jX1W5-rWUfa{gVVPxDlG+iLy9;dzT&CWye|R**
zfB~J?CC-<ARBtell8=jnFg`BhP!$gv?q<J*GjH4dxYzqcJoJdvUN@7SyR7RT0s2*d
zgje&T^kE{wPl*3I!`npvHnj>iHvP}FE>H>5H=54=R2v10X-|u9!gazwC2GbvlL3qS
zd|juJN*N=^Yzkr=4o0@Z&hk9<N-5XmaYrKj<BNWdgq9WjhMhiNRf4A|;OEjdxgv48
zVilQT{Si2^EVDy-@g{3770`yE0MHvvq&YZyV{DLRkQOPUl9yiA-~IP#wDM3q@m92#
zm>&Gcs>nqEd!dEA>BI#R*oP#cl=<&9EF3rk$F~^5_XOASMJ6n(4R-s{(&GJP`PKGF
z9!h7iRu<1@WsrPLQvcWOG|8PeSc75r7PjS^TrPg0>&^xMrlk&U2$N01V}X()4c@U?
z?mQp!Nc{#k*Mk!*ZO#{R{8~rD<7j<^5@9qX*<V$%$ZEVg1f|rFGO)j4A4;8bBpc&}
zwW+Y)`|Cl0h7zPpXe>JZb|2|&*!(kNqWN05_HSP&^u~mNidm`K{|06^6nZdOC(7}O
zLnxC{OqgbZlu^|*&Hh`OD8t5ZC7qxtWlCx(--S<AfUot@HEF{sa<6=3&a`#gL*wAF
za*d<24$XusqY}L}?kta5aEi`KIYHau;ze}i2eA`o;n!P2UL0OiL8TxYvIQeu2WLTi
z?r3g_)yn$**Hg#vgF&7xsNbJq@TkzJMQJH-+Re11SCOtjZNxQy(;;C4X3faA5KXDy
zz@7c4Lm))EOAM_2#^Z9E47e{0v>vh+PRO#$qT`A`QwCI$P1mt;@*s?6hI!DhV$ods
z0oPe()`*M+eV%;v5|9^5KE6K|7FZ;%GNU0;gjmaW9F@aE?|N4Sa9dWbjh3T&#|K4U
zYph`O$E&bF9{=TQ9ad3sivEn+83|2-B=wvRGyta@>$0f{E}tCM3}^v(E#-U(I^eCv
zPqPwst<f3tSIAP)SL9Eg9LtYMut*lA!ygMdsxqoSZ5qOL@Ng4fRBLqQnNZgoH#G>I
zb*<)v?Mzp&A)$n(NViQSg+0RcF`tB>tNimLXYqATo|RAVsS#}(LRRkObO5Ke^a~Zv
zEvO7;dq{^YQhA=;6zCz^<?Hw7D(gu<?-(KFXgbnFf;Zi;Ye>6hVxFJ;Des!o?!^ga
zbuDL0Af;OiJ|St3X687Tj}{;dd*NGS2uwo1tpb4nR?Y8jTyA2mPuhr0Ka$tGu#DBp
z{_Ts*+gvZ3-Ceun{=!#mxXf`cVzC_FBpLup68W!z<NtDFBzwyDjwH*MJ^%7$%{}Ub
zvnv<Adm7wr({U{gfv*TR_WtTL*dYfig|YRCO?RJ1F2>AZDLf6}MgQfz5M5#I)lyHT
zpLleg`<uhWk`^eIUu=B=nK~u~mL;E48cPIB(;A$G`JNq73GO=7ogOOcOFUStj~fiK
zuT+y@yKjF)=gFg;Bba;VcC@^Ffp3Kha=IJ*I}+Dsa77*BRVVKM*}OO1&jiSfSf7c~
zb&PujM=Or%EN}>xs3!qLXbmG737r#!Y_Q;ma$*98;I+j}X?2v$<61z%t>p17k$|f%
zeByY%=|eF?-7vw*Ww}9RXRLO@xlIV5`bCso=@6`g^);~sub@UT@+~?^<Ey@4>%#uX
zpy$Ei%>$ZN!YUvtYF+pK@eU+U8IoK72OE`^-AX?<vwlc-UZrKGR47bU`w8%4m$DvH
zU%OE>%ug=iJ5}_XqbR#Kq$Y^Z?10vCnql#sR_ui~PP^50JCe#9>x!j*fzL@XANMpK
zhJId>CwId>4ilr(l1O1{jaWACem4&S7@7(5_gA_+zn}uk#@0sM);mIoNAK1_r3+aK
z^;IedYA(<CviQW$$|*{kZKB3=0w91?ngx3jI&;jYj>ek|*y7P(`R65~UgdKh#o1>7
z)9)=o%Y%zK7gdY>4btU3K2DG014kZjUR*yBpN(BxZG&&OQ*GN;pl|)iXKMEwTQ#Bo
z@CZz<3kps)TJAzNAHVvKT9m@`QI1rn-xD7yVcJYC$veD$>AfM(L|t4A;kIvPZ82*3
z?6+S!lK91Yagbx1%mKnf6^PV;>;8bRRJLBmP3sk7XM{_2zI{z+3<VEo>A*+7PYE0x
z;c)U;YSsODXDvJL?gx<78W*k)Z)pS7D~0KIREA}ZeX<;leprw_E1s?ePqNSvaj1Dp
z20nkl(!h5la2g+_o2DA!eQZgBI1L|5;EstWc~-8;Z%9?eux3y+VPB1;jFUl%eo8CW
zr5HeBR}U&0=?#j;dAjf2(5^djU9qQ1KJIEpudfb-n44?RjMEol&+M!;0GYS3pzM~u
z&zQ8d{%T`y9?fm-ri?bpg<QP~^>h)r%N&*x9ti95?On*g(1!kz={p+As5zG*N+`iy
zrxWmdlWe0M8Roa`M2aVF*JhZisH4Q1>Ngr*ldsD&urJkHI$r+vHHhgQBvV<~CAYHc
zZIGAm_K`?!euSvD*RP-9=hikZ&x%Gpv0go`RWVdlYw>jCBO*>(aQ6eb{1xHxRU%lo
za=m`z-e&0;%dSqEqQqE9=u18GUvMy@TOwPEpd~{>>+v?P3dr6Y#lF@{kXx$`q*B&G
zyk6-!jDi%+-q2PGyvc~yF8!1$l2I!}_GM|=LNxR=1j}Nt&Hbst*)H=CdOH%fet)e8
zsR>a=?$JBAt*MmFUa81spEZn`4X4tH^U{x0?C^5I%A7<No|z<_mHV1`>AK*n^LNQW
z1#0@Z8a#@}xM*qv7v4!}qG6X(T?EsGQdC1dg91lYoa)ZsKTvVb7ih9KF%VNBUIJXl
zj~0|47_~i)HM1F)%JG+*AHv}e?*t3XQ{Kvxb8BnV9jk_TPX!(slT~`#bKfSb%UNC3
zMir}V7dwvsu$Sf9|9P*z#OFuJVP8PH?SltLH<BFQ7Xs_<Z)vow6kr+-SsLlytBmA}
zsCXn}pvoV;QaL}QUf2R`loUxCF7|r#%iI?sg1O*dgJ94=>sZcr8$N1owO6Gp8ANO8
zkRX#&yZS+=8O|E;4$EYa{}>19C!hn_r)8>Dgw<MPhvfU0MePqxx+{F%(3@?n@c_>)
zN#3e$0gpvI#|Ghj66ylPDSt5g`e;fT>@uK_$g0rnFD>7>v>C~GsD0#Vy;n%^k5u8j
zW6~fznHN&oK2IaBOGY;|5o<9OQ+Zu+;64d@qH3W}oT`O0r0*Ny@e`XeAK<L8NBA&G
z^qtMaNJ~Imw7T%xr}f@#OL2UU+R9(aVDa_}%?4P53UXV&u{}F&y~)8CWF*_=B-)lg
zHht>$3O6D6*$rqNja&E1`J23uYomed>6v1tTnfMQ0=MegCG=}a+HZ!=k7KBMUxVLj
zw|sra0S2vVm($SOEsOj)j(wNA>Owz^rIg4c*-0N{>!{mNicm!x6dR}SHte|JS|4aI
zsF0^TJMQt;KRjF#(;ukwC!S}ju>e%1p-!wSY9Hzqi7zD(`yRdZS+7L1tS-h$WIp<1
z@x=DHrECa3mwI0`X+)-Voya`!y`w$qwxhjc7_MFFP80^omBAB?r;1j~|JdXp)CZ@r
zj1$UcmgZ0jkMU>(Jga(NzfEysk^G_CfPho`-5CvzoZsZN5u%Q6ISeWK#(Kqx#gT#5
zAPNQ6K{GvYEx~jGhRuk~^RLpEoW)h~L($KkS69e1uXOkzfu>dSpl|`jJIsP7oyvPF
zUE9l$jvQ`4%l<9Yj&;s{9sR_9(n7udXrr?Uyr<TCTO}w!)*y05*8@Lyf3{b4x>l^t
zyC?DO4Ehl`p=Lm?u5@Q6qhv(bzQ-}B^L*vtfmt9#rE7t`6YKD()XIQD;T=u{BKffF
z=R%Nt;xG)fLza`QRXM$rVxz+~l~ubd96jlfx<2SbWFYd(hxg*!SdLotMj!j-RBniV
zS|f7{A4Ns^^do(mkMZ8*JOT?64Leh<!@%-L(V@(t96(Gd{%pe=bnJF0sBE$<eWKb(
z6ps!ue#hAM<QA{ld8lLWZTCE7t!d_tZJ=T;G%O>qSSRU(ULFC*Odt1-1itajtRPR>
z1=(#G(3(SDY(K{;Ja$2t|MIkLqIG>u%@%k}Ha=ML!oOh3VmBUaj0v#o#V#o{s6s%Y
zYRU3Y&6+Z|Qiw+MOuXMtwMvMHmLP3i7073R{=u0?`_S{w1(Z2H!!MDib7xJ@2!6n)
zOXe;VOO?SY(W-@7bc?c$r*JoIrBZ3R-=JO3?9d4Z<iOC}d3IexI>$t@_~T9My3vlL
z`RZ(j@4lTwO5}PzvNA;c(3h?x&ntY9IX0*_NPV5fKBHvm%yYR^j>k;79IgwhxlzZ^
zxN=Z=G@i`+<_z<|xv#YE25g=!*Q|86evah6A`QPRGVqv&gRW+*L&8y&FxS3NO-m))
zE}I&a%mw|5WMPi5{ENR2!tlm-opDzD2RTL_BRagc)USmc`Ka~-+Gk$09mdgRd1Gn7
z0s4CJoLL&qN4d(d<?}{{TrW*N_${SGcM*K8U<&3LvnUt0PNzWgV`rlM8r9Adg4H_-
zt*gbcKKkySpx+ukeOp(<Q{H|MXVHm`SgC(AbMV~s%5Q;_2pJ)&{^yfhfGToa4MtB*
zYFGozR5}I#vGN-CmV%$Mn`?|omH6&=IRy+LizCh#Bv5WO<||S<*2~d#Tp}flD=w?w
zgpJc__9k}hK#}=Oi2<KpIJkW*@VvD8k+K~31OM~gOUYQ92evjNvF<z#nMGI+HO&%>
zX%6F)@AGJ(XQ-5Ho*7^WXTlB_#p;{Ez}~`omp-Ota|IPs0X$zRgdv&0KJu2xj@!i3
zw_ZKdNuWe~<=2G-7V{IS4=O_{mKa{(wj5j<_n2FbF4R>XX*#E5S2ESOfEKfWsB+LM
z+~_S=yB%$!^wWenZsrC~p#_;vLeZi08cLgb-4fC@5r%PQKoqh(U$nBqs%ph^zDJhm
zEJ%Wwfv%J`a=J*N48x3a8zQM$Cu#pX?;9@~&LQ^Ak=V^QA}VvfinAr?qV_nL)TH>1
zl6#+?>uHr?!UFV@PogW2vG@7+1%<Q3rpd;MSN#vpoPXNLBMxg5c||hx1ty(j<PqQ*
zEA8F%`K~>Ti=;w}16WBh=x>{&Q*Tsu&#1w%az&+DPb8oSm=E0F4T7qmVHCJe5#Bxe
zhLYaYbpZFJihl$0l#~66{hFrc9U2HfZQHt#Tm9Kqy22gn0o#HXe*gl8=P3)*N`&5&
zU>|PMKEd-M@5r7D&)$d3c?|{8Z}f)Rj`79;db3#%9r<ld(we53&uB!1MM7WCxpUY9
zKL1Xa6D(7+4Wi3tX*7NGo-hj>I>^Xn%V*Q-c+wwo^yXYmr)Q8ZQnvP$?ye?yM>0E%
z{$jkIg~`*w{)HY9rAVT%z2~YB41*dEU?;=F7@F&bHq@bCOMDc#&UWh7*VFybsU}{$
z#tq2McUqll^3c)qJMZ9@uPIX`i@x`W)4;f%7stS$nA@fkSWIcd%zb)r_8*aF`00km
z0i;b21a+sTG6TO+=O#jiRGQe@44XgSzlsKyni=;QXPl-uAhA>}PLqzF0o-{&R}JYu
z6;suwwX7J^j}5k#?*aWDM0y7@kNTV`@2`pzIzc})tThPFCRG94Um_Ovo5WDV1{&-;
zSLhG0$=8i(XYmXjs)hXmLiw60C(+;WmEAOd*>Ad#A3jMZJXqxk&v?Q6wSIK!%`~f5
zVjNA&=c1X`Kb$c!^atisb>WC~f%QSHhAL`~lNZ*#JOsNBk3}ADWu*sH{z?EGQ4=HP
zAhEoZvEh2w3eL&0MZJqST$?HeM(LmN8<tJV)Z%3{C2FlxWpFlZJ>@XT((mrd2`*`0
z+Zg54dzstZwko1uOKvCVyQ0H3L9SLg*}qH<pV;V5nv#oj#|m4a)-VA0Y=u&!0eWZe
zwL_$4)OhnyS!^mU=~7&B|BC+lq%*b^t=%uRXl8@UQ3Vxo&Y-IRsO3&0aN(qpuoNTZ
zDC4PDNO(Hr6{<+qxI?jJ+gT^iUYo>s`joHa8J>sxGf7Gw6ivK)U%nBoUEL3dq~GaE
z5dYN#iC+WzaEGK75wvM&1iFNIX_AIX_+8}IbhtVMssiQZ#(dSI|Fx$w`zXYp+0N5Q
zvC=5{cv956hX_QErdcY0$GgmVVbFzYP;+x(e63S9VA(`*Pn7qgNl;38%%K~hEiT#u
zNQ683mMBE&a`~WhjkD&;2mU_N+L}_+>nS%FF4iC8rCTs`i)XW@rl>Hia<kPbN$&X)
zHjazKp?6&!?;D$ZdgWYn4~FVkehjhKc#q09%dm5&wqCJ*`Lx2(g@yz2@Ya%wXSW1_
zDIa}$_)IAR7l7nP07{cEsOWH(!%8v#C}rXMCG!38)<e9z7I))cm0-`F<_Q}u%8CW)
z4quT7cFPO?RZ19>m+1Ttwl?rmX0+=<8heSXZ><+aslDf$@?_HGDdxsvL|a%WPFd<H
zNsg!y<?Vwkfa~6%dt$xKg}4Q}%XxpKW+_&WZy1tpi9X`Eq;ZA#v1?712wi>b%cA^U
zE1t(DgwhNAAtS}`oe{s+p#pHy=IwOe=2YR{>CodGFn~|FsyIkO%<tD#GU(g?{LKgV
znFZ(ew(YZw*zS?dbQ455&@+0%;DiFW!}9n8>y=C^nHQ(lRM=H$Bn?`W<oi{#HPDq=
z{0d;&WkYK}y&Sz-N3k~C&uHY8V+|uJ$&yWTePP|2j$(QJ1kWkH=0S&%B`XcuV)L;S
zb!Bpl&$5NRN$1{kgt<!PN-MQVQAuv$vN!9BzRmqZ=+caA{eGK9so`gElj;T-ngTUy
z*b&;HRL#WyRi7F;jp8w%`V+1@{0@Gh!KSaP^U_2pozGPmU~roM5zR~BmZ4zt6{-RQ
z#5poaO?VrVsN=oNz;~1+3#@p@7wyeJjrVp3GaNyI*3;7QV$!cqj(8VnoF|`x5O+r#
zN5+UEOEtewXQ@sy*3!j`Y!xb|L^8%2LmsGbJg~2RlxnRJAYwh{?$0bH!c!W7x$`%{
zr6=tT1ToJ;&Zkt%&0(uel-KEnKWrB=0^X$CiB8e-Lx}fbYMul7`B$!Fb39|hjYElF
z^N@$+g5-%l$qp+A&tx1|iG^0v55aAMFTl!fOWzm0hrL=pc!5L>MVVCG@}$L=<2Jj@
zC$#`qDsLa}Jay_w2B3LvtPrZ!T4iZv0>E&xu`JS8{0ytAgI^~!L<&@6zsAb*;^Xxt
zOcQ1Mm?BNtXPkA9$$so-8L3-9gUzYf6-yZ`OlU1CTgZy>quq0<`?jO1CTxIQ+E|nI
z{3hww7#{4`l22@tj@%PF7+k-=72DyDMQ7ZZ$I9IMJVAt&4lzk58|ia;I8;KvvN~st
zk)mzfqZN(+6MmKIh~~`M?ay%0WNIAHxuX<gF0m@Um)4BSc$n+y+jXIHh1dt#IHbC+
z2GN*!G(WH9(8~h$G-WumH`q^+Mouj_^bA(Qn|S6>=Zj3_wcLZUH$(6RVA9SZ-2)!6
z`|!u1v!}nV_rK1q)VXtKl{dW|#n0=)G9JJ_@vCGz<FY!%W8MyuR%TB#E=`Gh3O|)U
zC&+D3Xs|Lmt0!Hwzdq;WR<_qKK7VcI6uI(dZ)<3JG9cbu>4CLIw1aL9!T4}^Vr+UQ
z4&$$0%Mq`i#{H@fAd=4-Q>>QO9sRR%8Fi}uUe;rz&rdDv+izXr2Mn!*y*+Vfd(A`m
z*8C(QI;fYQcq%kHux>+0#@ZUHSigaw2DKK;8CGEdRvr#F3XgU$Kp`^d{=#^{l@i{b
zOV~al_8mR2BH;Tcw^Zjj-AD5H2hCRHOIyu{+3G16J?~;>;+3)Vd-HHmc4$Hyc#1&Y
zY0EVwBCVB#<zDZR`Gv+NiFIyij-L{EAIr83ji;u10;;f@gG{kqID-tSpfRM*k3cVq
zSy?p<c^5%n<06q_3Tp!{q0jJaAJ|7cv>>*<2-+m!qxnb-l_5KZ@?9lHW=L}O0HvtX
zIoBKP$Z(^z1Esx!>-M-VD~5bxi5Z;cfc*RlKUso7l0L9Z|LwefI`_z0dSV!+`AsKa
zV7<~rYyZv+b->weC#)vcyk8a@su=b)U8+Z$>+^HeoYZdtG^5Pd`7iW)8@0R_L14sG
z!kkrGjvG&5!mO^}O1w!iqd2Qte#u_GT5(qc$M}M0#3CFkOueeRt>U@?y;Ox#OxG2~
zwZF{RkoIBsxBQ&<oQrCj)+{$Ki~cnnQzkOaYCg?p&Qnf+1}p=&)9H%KV<1z>?uG(I
zii?-3FQU>3zyf|RqlW#8Y4S0)_Zmq;No(E3phAesgjvTNE`y00_*HIP-O$I{z1^VO
zd{fUDCL;s~Rb@9c8(gvvoQUN(MZXDR%Vrof0qhGM|8(Q18!&lb^ZY;dwf^5-_xfaJ
zsQ)B(<E8TM3$L1JRKJgEK6HY#1!uB~LpmpcJbvAtkh7%$)B=Ihm;ZcCm2a)Um1<Y-
zC(+L9BMUQ`uRdus527;-EskdKzb$=gGj5+2bsq2>($OIQhc}WSaQP2!Rx`+?qyL$x
z^RVb|Xvpk-$fJLF@*B92R)onRpCVQ463$&0zqoE(etUD)>@L{*R{n#_GvI#XvF%Hg
zqm>(xs)8Xqwd;NLQTJ|zQj5M^DP30L@sPErvUCh;%EG>*H50{FLn&ZPQM`ECoaZ_I
zIt^(^y(>hG<;M@xXFPE2YRi>%qzdXyS|f;N%wUV-6E>zHg`Bn|`ucQPe$p23IqfVO
z$l&Q9%-u(yuIU5;=!=97=~Om}k7!RtW3Z}AiC+9ppefCIrioWdl^FZ8t%E5^Icv=b
z`dlV~ytwc)4avI0qV$qu(u~aA-z~2ADSwGXD8>bA>K@zUkKZ}R6I^y1no@4e>jlKz
z5@B+Xj7_`g=YpUTCRL+jeF{e<Y$SNnc(TYsq6$_E$S$)7J6$~R95U<Jlj?I4Z2keI
z;Bk;zwlMXeOK)-=bNp;M8X}$tjM>voStEO1M;^s4I`0w9E3^tkVbWRR6m<$t)t5;U
z_96ILCeIi&D!*jy%)^pmjGe(B2;U+uDi*u~*o$+neo)$2fXW#qhU#l|6$QX;`XP)(
zCBxDJf23-IeS1IIH-SAi*M3o4mAuf`!I8*TpsVT9`}x2saWF&Uc?Z=an}>dBZ16gx
zA6hm*+cbM{b6>pH=Ibm?Mxf|y(gcIkL|xI(az}SV&yyE{eg`7@IQD9!$oX~6r9@gI
zQks0v@ExCiOWBs6AeREca7z6pB@;*vNLVP{-RK`kUvcLADz2ZCM_$T$#cFY8jArxb
z{ZaN9(#uqvpO@i&%J6P<=AS5sYNJY7Yhidi<7^35sF9M>`h$0`aX7&9rhIITKCFu?
z=O!io9QNS_JQ|Y84_OferiNoJML?G72$k=P!uC2{#Jct{HI;#6b4z0wZ_a5VLfJ5(
z*I_#&V!BrI%$bKS+lNv;x@~ht=0!)E%d^?Dx%ag?jZ_v#p}+%<^KXm^-_^fy9q%@O
z{@up<F)MjL?20-fQw%Z5rlE&YuW_2<+k1lG$3KOS+dp3cImc$7e)Zgmw`h{e(BdqX
zTFvMp@Y1THqfeMrKgfToJW!b~FerSeeR7OlY^J;LvC1TI(!ZEiJw@8e$Ac{y&~YQ?
zLRds{5#72@8vM~m(uzwqRkM^m2w40)PizOUtJ(BaILpr73PKpJps2x-Z5rbN|3>}Z
z|7Cb|wv$|LZ7dUw?tRBPuTx*od1@5XRraEJb{f}5V(?&Pxt~Q!q3k3>i;K9h{RRUd
zmpSbTd{v!0p@7X7e9m@uEK>P10r+;5<yy0YyU9GhMWt}UiBJj&LpETx;(q*@@@8P-
zh9VBiTd{U-fjVpTMIboBs@1^+5nV)v!!4<9`<qHbs%J~8f@m@fD*XL<uf0Zx6Ny=&
znNFNTj|wBL1IKQ}6?^oM2LudQ&#`1CC!mA6TwtEhyvI$?qf09EcBMu+IWKGJbP~WO
zx~aO_8_#PJJYsyxfz&eBFy_TQ1H-$Ao7$iBK)LdWx19yas)NUO^(3^o;PJ{ar53VT
zx<#0pN9C<VVpFJv>d!_t<hJitMM@0G+h$}8&&@nGZAZhOez~HwGPltb`C^?)CgAJ4
zLK$wq(#Vwtm4@s1%_8mjE3Z!BK7qR7DotRT!@d$YXiDqPiIU&<JqWV^mY0>I;ouy9
z;o7rjp=s^lg;q^0_e>Du@~o%-fP=qx5F_umdSTVjz{hHf*8DWFB0ND{yh2%+o!@%M
zD3&$wK-~0+cLC?2#;(h#3(uT)sOmpFe1$sfGvq>w!7I4#F)p)2ZNl_n3h(M#N#nKR
z+Dy{U;%8~(O0@!5-{<GT1QTG$(wU<V5T?Rv&-4?yh?y$6fEigiCo|?;xjk&D#b4Lc
zCJ;CpBpIGIaCJa{O_Ck$3$<dzAIbOvuIOmXuD<06)3~Qo&%Cw&up)^hz{h~&cv&Op
z%q!Fp3!YEy>qLr}1v2oP`sD!rIGhBlg8`|XB5b#4Huop-{FT}4sn5#6s3d)dKhPgj
z<{Q;V7s#}i(p4qAcCi>OFTU%+AraYFM()g6Nl?h~43h@uqDwCI`P!c+d;>nlZ-<H@
zqQkE_`==k8Y0!>w2>sL<VK5k|cNjBAbYoGwxyrHPYmJTqpTe5Niw(;Av&1g*fEcAq
zrTFQIv~>J>6B=IKqdP=aSk@A1JE=0qzhl9!u42}^=MA)YW=ChZpP)@Ylvr)+HzsQo
z3t?y@D^~<gqrd0CdIwU%vj~e(tj&iR|L|fGLl$yNXwbTSeE1U2YclG0zE!R|YrZ8u
zDwkXZhBthwT6|~8=@`q_&#~-!sZ+bCLYkczN%~#)SBcr*7xRY9M_Qz1DC>rgZTxf_
zJ>u!>-aSR#4y*M>AyJe&>ioj0HNtBM!w}V2Dngx%V-&{g`8=5l6P~3^*N3ibDLx`M
z#YBuU|MtvzLo~V)zTQt-B&h9`jiGm2&+qZFc=F8^Zmk`ivPzZ__yF-j<kCAX3@|?!
zgkG<$=xFdVbrj%z@vLL#Hxi0~&sti7SQ-nbG}<_^rQ)Zt(&8{jU|QjZjg1Yb=jXsL
z_Z|-pc0~pg1+)`AbK5rqdc|F7WtBF#&c6V3_V)k7YsKewRBiqWn@DEsO2wZ>5N;#J
z>P`6W%YH1cAIMtJmQ@=V*Z7|HthN~cs!HpX6cOk$zt@Xk?W#O!x(g8V+frkt59Il3
zUJE}lWGq`;YBaQ|^j=p5T6;HQeIACMXzy5BY+A+FxRBkzvFXM*^n%$o@=GS855;rI
zzI4;hM(aDU232I9P19HvjeOP8Pdk3iE3pOuPTTk;D8VZaFaBJGuC)@HXdO-77cnx6
zU>oPjHuQ<lR1H{AZ#{gZXg+Xmw3BLLld6%dxiui$UMy?Qtr$B&4^A!{#Tm!^cB|j|
zSBHhC4hauU554efsv>wK<`elvWM#FAevy&4OWO~$Ba}IW*{%ekRRJFJE#mU=YB3rf
z8c?-w2QX&x_~uK~rt(lms5t+`NuSIkG{w$kk6a+I((D&XY{i2!TuyaeHC3*5I<wBr
zO+Lc8-bpb@jb*_Aqn^*PB)~Mm3X$6G{DYm8N#ed@pb?CGE5H!ECeDp@-%0eaRPT2g
zZk|$!Xi5f21aaG}097kc8kV2wT&Ef{hZ{4{Q_c2*t6b@tykBmpBZ$Iv(mQHZLu1I_
z98vq2_loKb8%|Be^kY^O8g$v9S}~Z#QCbD>c~v5{;tAPGmO_;*2WFk^JY!t5li%AD
z>8{w?qeOkKnyk1&v8{<H+f5g@fs)yKQs8q~b=gwHSuxoS%aFQQ8JEYP!dcUxCc75<
zMQS$RH&caqB5M&Na9M5*7txzrX`EuXl;S80%63rQxWr|Eu|l&UJjKei&N(oRg_<0?
z<#s}(JynKikpx+m@3W!+4?FsAzw@NCI7iRhbh_ipECLIo+_u5J@m}2)&;y1)&ORNa
zwN$?gL94x$wX}dmc{f9`GzcRTI4glqj#k@i7p&1LYRdHhwEX1HqnA!5eN(TZiE<tI
zlUbj#90r1T0@tos46=4`zq+p~q`8SmBrU7>DYJ03@srp@<-#4lcdKr2o)(`{Y(>ZQ
z&TN)hXTao@F>O-KF>0W|f+z%x*}aNn!Za}F3<%vUR)&e3?fTA*fr9viqqJ?-`roCD
z{ni~u0ER2Ylhz+^@vETAKs2PVG*v~B#E+4*q)eS99r5@qc9r>W6ug*w?Mt$)pU43-
z#?p#e6cmgelZk2Jtmn&37lRq?mZoH<PZ6ayJaQ{p!KPy1L8AU39Zu2k;rmd6bgJ8i
za>^{e(J~`SKZbG*m~kotY$)k2(sno^!%A`$1EDJ9M>ZCxTprEURzl{#OhP-o=0j>7
zrG0}tI7M`hRpK=J2nBoXkWpv`w!wP4icZA(7~Iv-Rok?B(mqj!bZ%**b%DxIKc{T3
zt<jja0yKAmY@OS0+0PVd>78%hIcOiOYS2ph-576Z!zx`Ytu?Uh2kQ3?DwSfAhp#$%
zZc-Fiet^N0a6O9!)1sSN{od{R51Of?cccICkea?;RF1CfiMoFjR!Y%1Aj5sMVW2ai
zuG7;_{|DrfuW=#-)?<fHRau$#pErQNYpIP&dSE4?t{;cgMZPu;i{Zjr>dz^FrK3@p
zB9?FVFpXjYjpG=iPE_?uJTj(9M6lY*NGLl{m7*EBI(T-msbB%{xM|@WhJ);o@1lJP
z)O^zh!jvagTE|&?Us|K-g|d~+DLvDvyCRJ`!j>PDY~~uxX?l%s4tfE3j3!sL#qTqG
zp4K!6Ipv5{z#^hXe~G{*HdKT%!kn8|&z3YZ^|f`q>|jTrmYK?*#CZ!~pXvzT2o18h
zX77i~pU-)%Drzj$z#}gvf_Oi`k*l^-xOyHSl5J=+KbHR5e$hakwIFIxo|P)clNQvt
zv;!62AdXm1w2@X3@Q%HLBwTy9H`x-`0I1fc&6m*X{PUqFrSI!`X)e_*a*1Fm%ZaL`
zqtyzrn#sgD7@yrBK{mTZIpsRy+y7jgiAO7S-4dij=PTtPRt7@zVNQ=xJKP!}V9op?
zi-6`2ZOr{SLzzcV--Z;uB@2Z#9=_&*cr;ZJZBfwP2xs>xqN;SnLiKDROxY>onl`&f
zP7=D(<!(%rm=XJSi`Pn&KbEM<0txRe*Q6{9rsp_$Uch1p6!^TxRGz}c`W?X}Hlqmg
z*Go_YX)A&ZE2nQY785&Ve*6!Yw<IG?ovyWf3o6*T+Ee1Uh?~^#D@d?8n=bd@12S{C
z0jr*9`e-(xZUGjO<|*qZG!Zmkt{$ytm;cj3rQvxab+)@=@#0L>pv3tE?TgeJa+fZ)
z!>Iky?#JH7u5SRNtL}kJz?P>y*)FzEx!hyY6jYf61z5Q<AMhA*o!tAj*Je0ZgcM2L
ze7bisKfA0@TghcH*Er*<vsK4sA;orCXu%-j#0B@1O_yth&9JiN0!u$QWu^3Rm<W}|
zW5@)`ol=Fhk3W{GqK96-Z$V}3xy~VJn;$|F_5Ir?5f-@|&D>p_syhjW>yOw93x`Wq
z)~R5`S)byH=^aBrY26#Sg7W*Q6%Hm&W6^!%=zwuusk!RI9)>{4cmemrZ-iwwwVkAM
z&}_oQrF4R3Iu!G|+i(xT1W$J5_q!&P$h(hqCD3@xy!rOuos~s5s4A58&DL)Pm!3S}
zOlvZ3qK3U0{7v}R$AbMI-h5A|+wXK+dJ^Tqrk#)R2Bg%)494AWB)CMm^>TBn@gtv|
zU_(=xySo<vKZ%Vq>*yie-xbR-`;BajON<x4)imjtnKonJ2Yvl0(ZqmeN0nkS#O1l+
zn~vDLnBoTS<yU5t))z~+!B!*0=BQEid|$qviidfjRoXm77Q$%o8L=*K-PsA^IEB?g
zO!1e0bwt>Bq#pR&{4SG+)iiK&5g)0ZFK!U^yIDgQvTcLL;A=rZms@gDCN)1ErepY+
zef;>LIOq)5CRAED+|A|c&6ORCd_RqR$0ijWH1A8RWlSLBI3>8)hA*+ZZ7W@7LTr{I
z+Kz>t;v-7dHRNw-TE0CNIQFn0iVVLpj5x}MW8SXB$y9cC4N9LRt7WNID6(--WaEOZ
z^qs{?78@t0@Tb<EF46B+?RVZbqNjqCdcAEHofI1P9m1<!oq(e~ea^-p-#UY-Cf!<Q
zK(Y3=_PdcDsB)R2-((VYtfbd4d08!jOaU7sC`(J#fM9cII%(BeXbp6IY#2ZlCo%Mm
zbQvqR&;5Qd=<Pb1Pt=(i>r)z%A;O^v_dsnqFVs5w2pbS_DqJuLehPxfml(!@7$Zqb
zri%o2Qc^uqbXUON7vB93dBYU%u+At{uG|azZVoiPD+_7?Y-Mji6;4`gXP`DwR2G-^
zAGSkFb@$ATmIQakKIZHVIC9usXW!i@L#F{ARffA@9!v$CoSaMr#p%>SAmcJv3q^bH
z7?f!cc}i);J68ve^>paea|GhZ2HLNcN_m@-krg9(WMbJF^JMdw08m(p+As&xOQcK;
zKaY;_M7pHBF-(<f_EL3ps&tce;lie`;%RhL($tW4lTl@pt`H(`5MNHoc$>^b75Gp=
zQ6U=NfRMd_7xRHX-V@?yDz%X>0W9-`T^G735GVqj+`F2j9-b+a6(O4;oVWekwDxvO
z8pX}I`v4#&D!9O<a}}+WbYT?AC!<^3uNEA}ZFC+gD#OJ?tvR+$SG~8UXnAgIJ>^xI
zF=+jBJ1CFU#?o>|ny}8N;x0R9(eAT_6#9eOqHy*UpyaF};iKQB=+eITv7-E$>zVD1
zW<>BZ<+Hz{IA-9v?6&FWZ#XanN84Yj_`hE+tK9<+W`Zd|G9O5`mFSE!W5WO8(JKd^
zzrTEX_!o#%zpdCUZ!t0{_>!6!`Kn3L$$x0vC8IOs6DeepqQW7w#{6g*V^kO;KK$(4
zZO?>cf9;RfJF>A}Evb8R#Dox{iB|+QF?G3To;7kBSAiNi!SdJN`nuc@%rE+oP8(A2
z8DXzEgWrmZXbSEGOLjLze8e@X-nqn?5S=^hB$P!|18QDOYm^GFAB=3FYuuO3qKHk$
z8JREb;wK#^pF6N?sR=UadyLK>sz~?K_!B5~{hi>5aFx^rbSJ8Id@k14>uWq-DXAF7
zr<4|IvBW9*Ej?fBtb)qOJ}OmEgT^WKT8GDnNtA1m_#<YC!ATBMwEQ3s#f&2T7_=3(
zQ21nW*rIZgbCx{W0G^;sKWXcT&w7eKUeME>J=|j2LK~Gf!-+yN;_S<uDzsZ=N$em_
zzT{$hn4i*}QgJgNsYf7lGAeBJI<#-nht8iG$F2q|b-xRKqy%*@1LbJxrB+R^TRTCV
z7VhH3q4zXLr#)zcMx42eD6N|Pay7U`M?W0KGcp)Zit8}AIcT_*&Dy29laY<HvRWuK
zZW0sfjpwC0B?PXZn3GL@KBxN~Hu{;7F(~Q3HPTUQR^<e=yTYr`+r!>R#|MYMb?_%{
zC;EnJ8U3tH1p#YtTm-O<f!&9Hc!rLUjK*W>rDNwdzI$EPRZW*AL1OzG=FA<XV?W4+
z3{EA@5W*)^FVXog58@IBxSY|0eP0g_j+UJUS{;wi)dm9CJhKH3)9aGAKE%wXZR9GU
z_HNCRY>X1sJx*vL%x~Qm{Sh};DcGBx>#W3nV$-*YJYvqB61Ifg1FVgn{`jNyRR-nH
zGmZV$8o0dvNEhmqPwx`vYd=M>3<XYt`6R);R>=%uuZvh8k%EWCm+MN1n^}<1ODee?
z4uiPK&!#^TlQSw8riJRtZw{yG)-Z~*7G&4h^x<UKJ`RF|VI_4?Et(`8tyYvv2x{X$
z;r%Y$)aPQ%7g8m{!ohWf3*zF>+G^rlNJyE9Xw#$}&GNKUvX}BXU45ETaAbIS^1ACN
zwECuEZ~V8*{UUu`d{%wUPa~{4pR1OAe-C!Urnu&A3@`R<Xf5O*()mj0mVWbPsz5N3
zV-;MxY;xqe;kC**Yov+XO53%WpQFx-wTN`?id88bw=d<*RWI)PI-zI7qC9?dU1DIL
zfn~|(Vb^$}Hq?1-$mM!)#UAa{VAX2qhrJLNgY5LmtoF{((+gASsMN@FzjF`J<kV^9
zLmHM2ImBzbX}7R?KKq8`@R;Xj87S9&t(9lU-Og+@xgCc&na&q5ZaCbHu@TgG(?snx
zX(RRzueyA|@76z{(CFQ{#R~yuX}L!r_2N|~-iMFro!6(0vEsYUNn-!4%7!aMkbKdz
zdGGud@`ZXZ*>UtelhY~Mt;!^l&U7~@i6bXBFS5#De`z=3knbO!9$TEKg(3Id-R1h}
zZ{of9uO!KrM`uTP4jr}l8NFr~`}=`IQ?qHHoq7&kq`+h(iHc_HAA(q=?ZPSYCqroF
z8gbkjYrY@BN|z-7Ow<x2&%M~YHrxWgf5hL0+@&~-BcChGiPP$G%s}GU!_LH04mvI4
z&Vz1`oFhSArK(>S<)FgpuV+&-owai5_uu5W_+Ua=sfu$!jn$>ZQcn1ccjOgQHq63V
zJblkn`7_&sErt6XED>8m60y&#XlGbWh!}1y9-eE_<bCqoend)^<BAb-Sdh!`4^yz1
z#|o_XwIv*+(E|yH?3>>#aNFc3#YzaAxxeuVO;hvhQqxB|48MMpMMu(<TsULF|0c<~
z4meIlOXnS3Y*&>O7x#0~<NNI62Yj~t%ny&C&l&YAZ)b@1parUNKSz5VGuB1&*|k(i
zX`zP=uw+u&Jf_$0ybq5}-v_cfNV5gRmSfn7d6&MSRt6JvS4rru^Ikd^%Ea*m?VTL_
z5%Vz}q6EtHfBc766ueJZeau>4;Is7C>~!sa_bDol)x<%I1w57CM(XdAMBdYq>|pad
zeye~|zGpvlbs5OSbHQfxw>{)N9w-sTC(~{Ato^@ghq%iB35-p)={DVLa>7X>txHJ$
z;dMrRA(^lLuk!yN+&q5q56{T-Lgx8>wewZmJ`)q?z^_b(ENYeLKHKN-zyIeKzUVT;
z!@KVnyK#Y0-#y=H50S0NSOosVD-dKW#oDjQ`f-eF--TvSA7fK~>+EZs)->H8RBN@w
zIowzeiCcpAdC0FHM&CIfa<xUP9T5w%L8YlmbX>g(9jvwN*oekCDd!VDRkLgRGbLfG
zh70D;>5u|V9~&60Q(R{_#>&$&YaBkZw#>S+G_X#w7QN|SpXrhgxR*(oxIcwc6y7z>
zzbiEwN}ya(KkBJbH(@uwU|CEn7x{%eC_Bn_uJ5)(Mivy*YWVrFD3<^V8-#MBc2x}M
zeL3F`LOmOOLbn?<=Lgf&zt4sCIaRMWR)4lod%^mHS(i6QuLQVkOT_tiA4Bdk^!*bj
zBHIQ850lS?sBBsyx~Ef4R#GbkN-0TEqY?#|Oy@n|VERyyTDT>aktAX!ZhqZgbOgiQ
zT)9!O4)HNx;ucebNjWMWL%$XpB6vSo2@E7?i|0`xfpxs@VbNsUd{Fn?I44+LwL=tc
zv2;zohOXL6nuEN9`ii((m-3;%y%`r24!?qVA)Q9OI=U8w8Y~t?rKa4JBepfaQyBQY
zP-GJ?hH9v1=^C-E_5orvIXVFjox+~gfid4?z-{+4{f+>L$=K9ao78sGICZiRK`ZzW
z#xutYGM6CorlhkT)oSGfM37k*74t^?=@Edc6%Bkm73yx!03W74)pUDMndPnJCMj6q
zW1yCp|J}rnm$k=k;12`mq=_F{gZ=SiW6vRa(fv-HgMRM6M@uu@_1OlFrR3<}$$k>z
z*LeJ=zp^X2<{nwFG@#@4y66AlRdHn8w6z%cCWsy!Uq2xKMrHfq{KhG6>!XUb>mqjE
zzFGYn+pkN7z-HaHQk^0{S9%46VWhDuzVaO(DCGjkod{&v;CqamtVqb$`@k!HEZrfy
z52d*Hxp(YuE3JEkx*f)+H_zbsl|hW9@r<G}Va6=fae}rl;=Ey%s$b64(<XP$u8(M9
zIhirriOu$(C(gWd6ha5H%U2p?aRC~09R>SCqbBMXU@Oj|zubBv8n0zOuTrsoGE?`$
z1pTh`DJH9R<$rxKtFM1$Y;n7q3an>zmNZCL{jtaCD(%!`gkVisI6YaMiD<V;?EEsf
z=()jXlbdMT?(Y@ob<!`~r_{PlRmsGio$gi}{t3)Fdl@esexJ)(1qv#hz~&pP_gHhK
zvI(Pkg=`j5cI~t^UWjjd`Gp=xaG<V#CkoCfcS87U-@fw>rH!a@@_UpUhq7UtnXSvk
zMFC-=Fj+I6Ez(sb-?g_O=5m~aJpAm71Wx?|+bJu>nsa^{M%c<p2ixVYmCISeCs(p!
z8)&b-wFBS<$mzH|&}jaqlHcD#_|So$*EridAjYamHF<T4+&bagT$hANllLMdy|>Ok
zY3^sl5cO{{is4}<PVlg==h=0r^Mc=q&c_QiGn7sRcaN$2k5eKC9s^&`u3IG5D^Mp=
z!-Qeww@$v4dy}&LjQw+?f=^^c5{xg<+N{I91@t`1p3z`<*1<=@=ELAeEgjPfzgL~R
z?iSY@_giWuqUYp$;&?)xrgk+J*uC{rU5V4i<pNdH?bsfc)Zi1%6;Q;;Q*4#N(|Sb5
zg$%NzJzEf)-vT*pGB!V(C${(hsBTnxu<1`0BRu&SujT=R6I|(uV!h5up^9;qMnkgK
zlzj`^8kxx(&s(K%L)d$}Le+}#qF${8<*z3OW73a)^)^R}bnLqV35WTH+e}o>>4F*o
z*u-FL(a*uiw`ZfHwtK`wSN5oIhr*Pha?|Iq39AVn!akjCPx~QMaqWG@nMm@5A~w|&
zB{0PdPOl_TY?m%R$83Bq@`2u;^&(VU%zA#ewwlzg!LPI2UF4mXUQd8BdXLKl`vc}l
zP$}0WzMt#V;lA)llratUdIV6>_TtZ2GA$x$B1<c5J4ML|JLgQ~D9Hh1S(-FX-0X9c
zTCp1xcAfM%WwqMzJGWCOy>u|{QlI^1qVIyQ)BJ`zc9qx-<<!$~ln^hy-W^b9qUrio
z7^xa}x6|U+?-s%v8K_Y3J|VnKp!#FXI`6MU+7zu5X3dRS2DDXW5hv4r>$=g+&3&);
z`+xd_WY(Dp_H(+|S`xpCC3VqDM_a7s)<#d6`VH$iQVom2H5W6^b&Ap^H`?pz!m<wK
zj4~XA`bYY|-dU5YHy`7UzCCLXL^6$!VnibJq1x4V$i2F`L%o|{oSqB1rHi_MMy+ZD
z3LVAVuga&_TQowYxa)x1I-UYGru7He$FH_Y4P1+mtdj}F+!z#K#+2{%H}yIi%?Er-
zB*bQ)f!QrvG=`8k>aJPU1#ru6L+=rKAi$G|^fgOy4=^f8!GICQ^g4g&gz;H{s0gpM
zOY|LQ0(W`zZb^Nv+qc{$jqmuvl0siutwy<`|76#kG^<}E`;Wz_9&VP}MP&fI?_^O_
zx&~TeuI(9+i{TYU#$tnGDb?u`$Le6a@3-yuNun8ZZkk(6emvg1&2HLQn{`|vTcGXj
z=5qU$oHn#B3`}-gII9O*y^p6lf9KFJoK}?1urr+O)^j{HF+7oR&ZIR__|^v1{~qjV
ziq%4_&Udk_4g<t9nbzU!V%krRw;`g9XoZwF>#yHpYk&25tvP?+fLZewO|GSwYQFCK
zlp+kFL(HIpUy|gNjCtI?<nt$Y`PO^aGcuUo;5)v$Oj<4OOeq?uWmt%DoH3!&8BymA
zc(>Ph%d1r-A9iP84js61rv(hMlEu|w(7dKH5rY9q7NZ^gSQ1DM)u4ogbZvR?1Y$Qq
z->qTyBKJ{_HVA40F+vlUoXi(h6-yO$7wJ^?EA5Gqa>hhqGXK1&udd`;r2GkwHS;eG
zlSQTSih61BN+wusl9l;}_2rDHm6)H+`e<=2t4lAVSTSokvm@HCP_EW9TocgP`&syz
zcj?7!yAu>x8-eLLCzJ_e)*d$7<rF24*0WBH*-((AA5v|PZ54lQs)=Ze^2T}XA8{-m
zKkGb?Su}xtG%0O%ZfCdb@%K4r$9l@t&=WwU^?wC$8AB*N&I{ktL{`6SR3KlkCXsc(
zy9|DozpC+bH@fnfu4~&<<)(gN6;-Kc{iAur`%0RkOP6SUx0}%BYucfR7wd)=RR)5T
zF(wCp)PIa}PlG!=`%&AfRZ|u~eS_73%pNIpme`Z1QgPjW-lDMqsw%eJNuH{#4TvGz
z;+)UT-V%umB$p@Q`lwTM!{ps&nvp?QObd5TnI^wUn~fGH>6+BbB0r&~pj&aU+lz?j
zj;a&gp3e2DY&vaU3T+pIU(!3bE$uct(*7_cd5f@FCOltkp!vEe!Q$g<XXHWWun!|&
ztl5ND9yoET830ZxtWZ+N6xVfSJ<S=bN?OOt1qW{&i+6pblUGzEdtdBR%(2I#nZrNS
z@HAyR`Rz~g6Eh{NYK#9xmXoO(6}`5a&no$lDy+;Td`9?g6=;=?CsNQ6Uz+uBh>Cse
zNcWMTdbkfXhFZt~sT65zpe=STo-Rj8C$Ks7Qa-u_{iq<C^jRRdjiZzzEKeyt?_Q+H
zb;0H(KK%M!;LB4p{Z8487{7_ud6x|6usKmZlB%VtxNlmtn;vb!pj55=quK%4I+0g3
zu`dymuVl_447;h|l<cRt0DfIXit(UQlzBM?7kSy<W#*Mj{Uo|~_mWV0?iM<vJ8WZ#
zN;jr)iK4}Z$EWv`a_<c+rp;n@-6CLGCh6h}=xTjjhIJ{hoV!6;OK}ZYMr!uKJcAig
zTy_D22%{FNnsByKr^k;wb4R2{2PYpKVL7f9Z3(K?l#w+iS^le(!=XBVD6u+yXc)*L
zSAdl*aK)ulkyl%z_ERotgqbXqQ(WKrAosVlI-*v~Z+<s-k{i--ueO)9UsaOPp$?LM
zq=gtrD9$SPIOA@e1neh5c<GWao?t+_KF&9W)_=|V@<N8Ja?@sD7dF*oX@MU~n*)Re
zW(NDP8L2F2V#G4!t>tJ(vq#bX8kUGTqDpy)F6w+oasv1)a`CeYHSw?_{+h&l@1UAH
ztV>4O%_4tfK$F%XaYJ>vCRAVB=WsCh@B8W7_8w2wxt92$JhoHA&m64PBvYTeh|dLh
ztzqAh(*KLNw+w5mY1>ArQJ_fCLXZ}R;K8-U9ZDcrkYa%Zcc(&;phbc^1PE@yDaGC0
zix#J}P@t4PJNNy($KK!1_xrW;V;$>QS!-mixn|~?Ij{4q|3UHq<KN%g_LxWR8I5Ia
zAw$(LKiy``@jdjc-TH!7nEw9L%{jqB+GC&RDDgtv9s+6)P(l&Qkr=%9sZi%v9!ZPG
zRRa$PT}2x|4;g=YM9;Bx+X=mKwPqsHRx1}AkmPiWrip==fEY{qKWa0Aos5Ht*!bo7
zc6gULP@n0Qe>PdR%)}s1hKIr@wMMMN4Zhn|Xct<i(0=LQA!`bTK6SXzOi8jbPdIrJ
zuHErE!njoG==|GAyijuzOTW8UAHRGhzgVM&iN4I-vs_ZbtDmo72LD|M4he1p)sw_8
zxGYt0H}O|^^wVdomu}+=tze0$yO619Sgh1W(TZ`29{P1L_)xS(&`uA7eaPZ1mv;Lk
zjMSPEC`|G>(d2DWt|JMRLu2%@e9SpDy-M@_q)7Jfqz}{In-{-1mS#2%6%sb>cH3j^
z#thHo>-6BrMr2?rQG+KzhAL4DEkoq{9n4dE%)OMR2Dl~G1s+NY*Ydgi&tO>{ACdaB
zA;gE;xc3y&`I0e)SkpP{LP({7l7Zp{#q+(z)9?TF)&AE<z$53w<nSPl;Kb8b<rexQ
z0K3~B%Kq6?RDSxQa-Z_t#5mr@LZ#(`!0k4<t5f}JYhjmDk!=o3AMJ1A)NV8o({!W(
zuS}sn_uL^_3<uWnb=B?RT>Y75UCM<yn^LBW$>s8zUPiP^5<2s-KKN5wc*7W!->CKP
z;paflbz-t>nKZiE#*IO{u=FfQ8)(C4FK$nH$Y@}tUT`5eo8_*~B~6^#0Oc}s@N7&T
zbU?+yO@08%(=Jx=7H4kGh^{M|aGw?Pz6wsfJ(-ZMeiBNtf{P4HF<W!)T1qx~nLjo6
zQ9?R4J)mX^=`)5tv!3tG92y;-I(o*|rae1zFjzUs*By|psBDlvx18ImF3<kKPRBsg
z^@cMwPk3_^Dc}hsrL4H&lxPk>7(M@B39hi$X?+Ib+uEE^*ew1K$||WpcKUQD-kM*%
zo-?-E%pE5SG+YJ21<sl-=C`<w7S#GcQ-UigYWW`sfWEbe{0`F)#u`mxC7HM}&LThC
zVg-Q65dRy#v-;0h>n)aofhQ*!V=+tnU6`Ne()1p#F@8|JVo6|-nR@ea=^VZli;xty
z3hBovb{NVpZOYfU{4t%||6z){OQM>cX!D9l=KwE?a<dmO?I)B%f)ATpD``$jF495U
z0^gl<*%%Ai@E4ra8heDWsbj2=GaDG*^#E<A$zn)JeiNZ53;2|PZ%}OVR_XJrfblnu
zm)c<#<lak}e{}QA$m`s*_gs9BQZIp-Ks#X4rDlS5<(e$Fm`+`5M{Hz6Xf`Nic{s2U
ztL{ypw<?N8-NPoYH3ag)=dCR3J-gmBUA3!yI}`GH3956tP&$86*<8a|ecc^7B*$Ck
zVSYg7HUgHFSe^R}{ds;6u%QPLW16VHLg1T;p+uv-v+5LkO)HTfo7kmh(j!wLMQtu~
zZRM0!K+pa9+yGmoC_ZL1lT%_>h~$@!=EKIP2$7Toz7F4#2!49dn8z7&!e-w8+8qA9
zkqLzaKSxH0|0%8R02nv|4-)`dzBD7}8?Z3heoL#p*;Igk2}LTsJC@zS6kJrAtYmXl
z(;(qqE5{HDpl9ihwkT4JCSo<cc4z*a&%e+UIlzKF4uqKQwNO_f%78%bKi-ce>nO(&
z<z`EgsiI>LRnehGhfz`K1qFM!!(PPR&-$MX?$y-7=FauV#jYshm}%aX+GO$Itw&3~
zjo~&525Bek&mLLiB{%j)WCtgcQ<(ZMFv_z#g5&QVhp@HsPQ_9FvaB=|Y>ac`0F*2C
zEy_KVLO&p|`Yo#biFMXu@w<t%PRWQ?BNePbj8`NoCUP|=Dnh4$*|KU@Xr}?IfHlfV
zF?Ciy5|9LErN67Tr8$(eQi-`n?BQ0xhF6+A@@uW1!V*sftWRh5pMTRjjW;@~)||G(
z(Bbi>!@qpV3|Q>&^8>8eoI=jq1pRya=WF)H(pe%i`-O#Ff#Y{f6D0r#FR!*4!yPe)
zjN8X0+Pa0p24!dtO)mT#*1P8pWXqWf785xG8I;H(GaK&vjm=WlGAX2m2t%PTWk$2|
z)@L?KsZsHX?->OaXL*EhvfVAzpZ~;Z>lZC{;%#Nw6lH;v5w@zsX!Y1={q%M$i0HPc
zPTCZHVi37keJ-Ek)Z5Kc@Jk~-`}Y^dRAa>mId+S_bW6xcWmW8%C%a%X4Yll#DQN7J
zl(I&1fG8kSAA3MlqUq}G*tA`p`l{hSc-IB+w`%LR@>4QR*R0(MZE7#mH3n?|%9g%@
zHd(N{>eV&BZBTMj>CPLeE99m$iK!m#F}v#uy&w$V??_%}LEf>Jij5z$2)&tp(dnFM
zctqi`(Os1p@Fg+@l4PAO8mdz=y{TM0^P;zSCa-4I;q?m@7z3AkhB|ZUx%GgelVFb;
z3r=8f8LFU`n4IAD9Atk;d1<m#h64Mzp!IU%DPTS!Auy4h<DF3>o~==BZ7n%A|HyNT
zvi$_>-XPT$$hb$@>Qxc#Q}ib2VY}F?P{K-EF2S^U`S5;p#-Q7B<h&3io1-2%r8Prk
z_t|Ao&kd>BbxGo#`l@c~O7o-C*mlZqZ;|w2>ks7Q(*J}gy0zU|3KCzJq^`O=d)t&o
zl)7<2AGUFL^Ui)ccxYAEzJk!j@!*2>-`k#TF1$>TMxNc<`kbpE{-kqd^qxTH{=X-9
zxTnka@Wh|A6XrhG{XAh|^WP^rt0A23vb!f6r4%CrV{$3<$g`I&CwTS)Lvl_3O_cb?
zvhD0AF3*#_`<L_6zLZNx_n{fTgYmZlCsx{C(O8Im|Bd<g?|ZpaKW_i2NYY9>t|GTH
z0-8Z;tNh>T&UL9Krt}E*Z^>3xnhAa13MAgCt;(HQjz0Qs@Z*O2VW`|0*OTwc7CR#o
zR)(9Gxg$UETEG9aL=UKL9R7Y~-zBrKue0fU>;EpK^4IsyiBp}!KsiAgsPlQ9-(|B4
zXZ`$Ji6{Ku1QG=%v!B#-JiNd5*YT3=J_|=LTeQW)M{>|~9XjLDZh{db&{fRIK~=7E
zgg8LcJgAeay=`IU*v_^4D5&Xd1yz|Ur#V5uM3T8Tjt}$yjNZCT&Cg{8^Wv*3sg4*V
zwq4Wt52XZ=6xgqguL!SfPLmO$*8X(ty0fVSSjVx`(J`U>ry9Lk`PVxF<C|Y~cn`$E
zg1+w=HtT-u{lJEnfLkSSR<ap(@FF>v@}J3_PBOpCCB*NoHmDKjm8e&ZV9}N)sRf=E
z8`;*(#1&_>rDQwLI>e7b;d;uk9OfM$VGkw|F%NO*`T<RHsnk!9iG#EDp!78#+?w48
zBu34U0Drw+4yBkg@;EDj(4+g66%u#vm09mfgG3snpZjC4>xcO+<g~d`^cSkCo0@5v
zdx*!a<`L~LDyaM0`ILJuUt;_^NJD`;=xji~o^ofXXn;&c6!+fI-YQ`|tIWj%{WARo
zgPG_x6AZV#?P}9R_UA&L%GPz;C*AkD9ruz<hQ|ZlTH}FD;1TB;BI<pdN0yz0A4{}{
zIBi0B#DQqGseEw|m3Y16g8c%&86vO296q=}cKyWgZE#SkJ!OZnK30=mVk4qKbGPEl
zVhTjVoP%NI!^G#Z@s))~ZTbd2bFWKYG8uTne4~d%^D$K4cAA(w7M?0&ioz2{nsR#t
z1QP6teG+y9>CgL_@V<_1n)lIH;^-W!!PU!+7vG#rA4+MwcB!g#)0NPaS9kBNCGUSB
z)l;3$OqwgiyKA7-Bzn{pc+wPtIx(m#So2uxxmxvw1demfhIG0LwvkF74|yJe)Y&Y#
z9w|Xuzm52s@4Q!M5#D*FjN1@X3O)M$%G}6WAmC2mtuD*Iru{{DD@XJ4f2_4nCiU=N
zOMD>5y^Evq|Hr;_iT~z(G7k#2<a#1Ntt*hhwaSl+`xd17*V1uz%641&-z&sdK>y&s
zN~gG{et)DdNKZ~Jm%KTbv&rwF^7dPjoquin{-1zvI5S;;I1yjcgZ`gfM@CNn&GgYk
z)#uOuwdZ@y=6|cx@cx@H<v)e_9ov2O>*qbsyWjH7=3k|b-eUfz50P*F#GV@UJ6LK@
zoZ|k-z~--;YjWJv>u+!tEztZ=3Pz*2>g;raNf;vze%;{NcL=OqSe7L;2N&Hx>I#-C
zbo-m@{K}FqdW>)G5B}y=z~48|c0MD|^_qj%JJSd&?-u+mzZSaPzH@{kfPMkKR|&0E
zUjSoUX6g<WH9%*`cO5a^>dDrRxb-#i-(UFs-B*U>%G*X;A4yC{wz4jio6EGiUB8Pe
zM>SOiul(h3_X)~ran5^Y*XJ8>G()U{PFxjeT^aUudoBM%_;|agEq)MhuH3e0aZ$Fh
z#977Sly6ZKH~Ho>3u%}<ml(l1TJIUGnZ=AC?;y>!xgT!pp4Q7JH(u}&#}GPRarVgs
zlc3)fvx}Wn@*PJ_d`J#>HmjrVKfaUq^Z_B$qoLG!=CmLDc4@OmEB)ie`$Biv(;sGS
z&74sFoioc1^a&)sKPI9+EwFY+_B#0d9gXUz==bJdyNT=jBhXEcTehrE0*6$(fp&KL
z;SHF9Kia^klGh*B*;-YrTA${P!_1t%x#5&SnmKLM6sCaF-O=|qQkJJ1FqGVeSbp9j
zfYJBCDdr~1g68j&gYjTG=<bk9c_6yJQ}z$&Rj@yN&u;MMiY~fdK%n?;J?F)Fe96I2
zHzB$G9LaAzGVdYm9N=7G97W*F!Z~qeRaEp{QL88P<Xh{PpFOx;zk3{Jy1vM#ku4mJ
z8hPiC;WfZg!CaMQTqTZ7be72JaLl8{I1`YmRyfV=Zk`D)g8+pYU~GUO0;dk3?GQhf
zNyI11|J&1Ls)E16K_4#-Lo{$4O&O)LWy0VeBEN%T2<r9gXX@Qe1bDc)dG1xm3JD7{
z9*5q4$RWpOT<s*`?3oqWjHF<ob&aMz4kMyuq+nBvQ&Ef{iL44-zXFO@IO>^wlhgx5
zmX+H;WgNeM+*w=1=zqD(l7xJ&5A~VeRkJ#p$_O4Wx5yh66`Red#5$NJ2gI%zHBK@2
zf!b0!KqDXGKJE&RE_<cbr(TNR`wMto+AOF4{8kShO6H&vElxzG@;X;oU8B80ZU0+$
zQrUbt<#dAyurObgMGU3$-XhInidhLMAW9_C7EEb?qd=K0nEK8H$nBePv#3Cd6SDLq
z)R{9<5GjNLMAQKLl~<AL2ao2BdV<st18uwF_)D=>>E6X<|G^uJ8>LPD!)a|pQMN#m
zuhhGu-WKrjaBv<+xMc1~wV#X}wBu)|D6c84HF~jZH8{#`tGXw$VHIz^#;Gk%vhTV)
zEiw*5-e;qevO~NjGBnpO#WuP&gYyQh2yYgNE)lIHQZj#F-Sd%)wd7?XLH)@Msq*3V
zi<v5i!#38eg<5L8X^7zw*f%NZ<--^;$rwkJwd$3F#))BD{k*bpeS&tL<lAM+FxM>@
z0R%yA|GHRMqf_woK#5j-F2X1n+eV(TCid{~qY=j2FvI^hNk}F?1Xwjgk*+uj;hr;f
ze)%o8L7?6xF*w6LRIpesDJ7Qg^@p2LX(Wa;Ev;uUN`pQjp+JAHn0>bp<f3<GD-|n-
zAf_AQ?EP@1h~yUL6cs)>W{%lUTTVksiP$g2sd;*M67=IV*$q1Fg$Uy#XH_{5BWUP1
zh<IEo0?N(iue<^Xe$;u*nEO4qf4M_(g{{|-e(9zmx7hw4Jo2djV`o~zn<$8=fpE60
zg4SEgT!#S%;>~;=GrCAkKxtwI|75S27{NLc!7m*r#36jrJaXKOL#j<;*}mMwvf{aM
zY?QsKoD?)7bnq@SD=wY^Uc^Y$W3U=t{ZyLPW<(!>!l2DzJ6jF-^=ab~f(1;Y+I#Qv
z|MZ;hO`4sY(0$>J8tgvuq=dLHoIeMg(l-90+q^5+g8nQg^1!O*8}OhIt3C1ZdC*yP
z?`8Md=P1#`_T+YWfgbQwh@03)`c=p%K=s$hi0HC3J#Md6fp?hNGQVmH$lg@q53VJ`
z!;e@w9fik^uQ3MSzRnbPEHiLD-O1;bo|qUC{hm$P?H(di@Mb~Z?vUx6DWsaQhI<-3
zIcG_8o-ZHdP@7wh|Byf+maM@@cFHXJPK&lIPUULfyZ7gtfBd=!2CILs71Ab>MmKCM
zpnO>!R$<$U-OF?Msk$wWO?(UZl@OO&7}QrfZZuj~fYQ!suu7Wx;Xs!@uNXE+lq)#m
zm)jqa8ve`+&UY6jkZB9P{8^!(FclXVN*ArS$F{S^Ze9j99V9O{Y6`R@%^4gwPA}Y&
z#QfGH#cn%GLTakG%O~0Qy$(u67I6_e)*`(lEWK)Vi9d;|9bYN{r9fD|7IL3VzOOp`
z(6rkRi+Zt*i9?3#yjCaZx6{xA@X@Dja}hPjTTfdI8?+wdGVGBli()4g65|bLbZ&y0
zQvN7W^W!fmjPZ|gBY~Hpv>1_&#xCB=QT+qaYNe`)SEXuA?d>Zs8WAS@RZeQ;`Mow-
z3#_R%BBn(|Yb@!;D9b~j2dKU#@ybvP_Jrk^i{-cQDq~kI6w|^#c*W;NObj9Em_5fo
ztZmzmeuhT*`*&A7=?OgT?Wxe#c~%=}<l>kjYFchw(1{=FKX`6x9XzfT5WKn;kgKK7
z<4%z$AI<Fo$kltcL_!+!EbT4hzA6D+oEH9sYV%kTS(CA~pG*|YfuhmTQ3+t^Qwbw$
z%(M(Jt`{-VFNRICn(hlzlILMdKTY4+_bvx!6<KGXNHbMR#)s4l-twzA&rp&?)9_N;
z(lsG^{#HYh)EPg`@yFQ!b|Ld(J<sXKNB?O6czBPEghR7<DF*kR%%0CMIZ`8>3n|J;
zSJAJUO6wdKNTz%sEC%*q=Kv}i?X)m3Mx|)h0s(>Ls}(3C=qC&ilNsu~cP;sR1lwm0
z48GijYmDmZI1X<0fifK}gOYsxf8;!h;3oddSQV@N>k%Fv8V^WZC@Y`zOS<#TIzt`l
z1&E9%)mE7bX;E**d_i9Xa53gR;U{J(5&!hGn8<7Jr16S)L*$|!H6!)+JS?D`f`?N0
zAdM_)_K)9}W558BHisQPDkcpQ(Oj;TKWk~Ntq`l*%cGIt(~Z_iqKy(kCST9_Y-t2Y
zQj&fio?kdbsGV!GygA5bJb1wHm2hd0nr@UNMu?v*vs9x8Nn{#ePhwc&Q2UL2Q@P#P
zjl+6KPJ~10Z|$e``OZt0+<awB9aa@>$gB`V;BD#@tHD4U6&J*ztErG~E8fzHp0Kel
zG^-qfQ0ucEZui8`$i0nfE_w4Nj$z#x8)YOaH{P>yAE&t0da6;7#g#m4K;_D(G{fbi
zHI!(s|N5X|QA@290tB(lV5W-3QH&8XNsl*)RpV=w$?&a+w8lIswlrdh%$ZWTVvT1!
zlSu3JEUrLQ1cGVt@uH$a6jTjG6<jm=LROkxNb=4;aOn`3jpNXpE?i@9uq&u{v{%{O
z*exv;QpeJA=CSl?vzFmHfnm{Y`CCh+xmlF}0jGpXRC^VWW^DD{szkXX)Z>r&O5;UC
zEhPm)SH<_E`qiWQFiHoJz$g8NwoVEdOKgvSqGkkNDx-lP4k`oB4e$TxWusVZAd8VL
z;!cyp3CAsiq7^DOWB7PTd4yy6w99R1sp(c#mqcGy#PdYJ4@j<DZOCKPYNd9c>(Car
zi8GPj>mok{PoK_t&8}A5DX?JQhMn&o#D@}Tn!F9G<+L#uo>w+EyzZ>uZQ$g^4%4^2
z67w3ZbCMcU&M+}+#Fj)Es~n~zZRG5>2~%)d8fI_kM^gr9`rJNaB;IjJw5X7`oBp)%
zOJH;4cm^F1ET1#BK2yW97^MRF`4$B^l28Q*0XmD#u@VxdmSlzwbqR&0lx2F{-lW5a
z6^Utq(&Fy_!3%Hqj3|^hbAA7ulO4dKc!h+UqWB;_9dD%aZie8cvG#{*B=L_;m-?*v
z>zu|?oJOqhUEE+(13_dh`W}nELH1O`IdCB!13<enrD&kVKH95VQniK40~-x4sa(i-
zISc!Ol0tzVDHsjQ@;H>!J3txc8~lri^fAU!NNNl@?PV;hUSdqt4X}g5Q!h3LJ^Nx=
zq+-N)pyVhuJxoCIKFpNNO7PbU^j4HsB`jX(j6z=<YA(tuAJx}j;veY?2@`s>QOrsW
z5tEGz60beIUA&ySef4io@NaLR@bk7o7q8LVDx2f3KY|~Kh5bk{Dz?JDhl3ZyI)e-y
zYV*}f2M|F?@d|k}+>w|W^K!@$gkZ>8rh!_D-6GFv#`4&xKBAr};jYBp?3Ce4Y_a;_
z<mQm}R+hSHrc;MF+2BN9JgFhiZoUr|5VN|1=16B}XXc}}R{pS?5zEdSpAoU?x579j
z$Sy~9Vq()5Ew*2C@B>sccm9o~%aM`1^a|1So)n-IUTfVSxn|h*yHOeG;JChjdU3w^
z*4MY#uEf`S!M8}ksI4ljD&wFktXh`9dy7X}B7u4RE^~ckd120@@z?kQ_QYFeth{2r
z4OYi;gC6O5^zS<m+HA;l@OkER`bFw9n>h#$S=FuX&s6V<;qgIaU^Y_WV#(+mUKQa?
zqOm6JUkD>_uc2_39uBx3TUn<`fkOIHu3aw7$^1mL$e))_Elg6k+hDPVQ)ph$wpC7_
zbP6&Ki6CNwzKS<1V-p{qf?pvic)gL501ACu`|U5^&ECdg7HN<)A$F771~%C52O@L9
z1v;Zk6sIa{$Ef!e2^@vCv3B{QYYoO)zU{|HqjwWos~K;q?9O;RLLF9{bBTVS^qZVh
zfn=`_${K3{XLZ{g4^So)B^VGhG@<zAHd(VBP;qAoUEoZJ(=dgPQA~S1E0OUpC}RW{
znxd_^qZsHkvP}|4OSRdJi1m1C4Xx_jzK*bT#~Vc7e{N4wT9e;nWPZ47dHf|U%7(^D
zL*+$gYABD8p6yD(t5l;3+K{uCqFl4qy`H(h)A<*qO-6oaH4YE@8VAt+_=?v{GhJET
ztXco7#L3DWFjQ%?ZdW}xWxKp637;i@ygG<gk1U=nu@H6m?&u8+Y$D#3=&N<rtNc9|
z@YRP0y>-#-+Jkwy7H{Wj9b|vP&TZIFd3nIduwPEv7fVzf#{(KqyK0_xqe0{gC?+9=
zn0Y|Mwyk39rX{lAgc5xyb?&_OKm?QF^K!5EXdsLLDWDz33BX2jnl5O9(K#{4E97@s
z&&K*3S(U5@f~PGWa*-FR8de{K4}e%}>tC;|Wix!9tISN5h6<fi5_~s2#wdjXPD5$_
zH1bdtjd4C#f}uiIDpnx3*=c&58p`>v%T1`&of=AJk7?*kS5HKmHYf73*@ouerVoDA
z_DIYy>7i}3lnX|cd$1<leeWv6Yc0}zN+`;;(Q+~lPYtnr?3T$Nu8Kr#{o+-czC16k
zfGS1{2t{A`ztk+!>UE?S=^r2I3#;~?Q_fCR_xnlE-!TNdu5)KVR+ROl7wm{B`m@AF
zb*PLNmp+Q~YPONOQ-qHNELAD7H_fO(<N8vGoQ%{#+;dT~8Y0889!Y#8Iz@(Wbc(g%
z7<VS;m%Kv19`;EI^-F}_voSS5@N@N06C@vE<wUcW`W#JYi2caZ%1IF#0fxMUhe5~B
z<ScwttSwy3&qC?qEZ<H8XMdl7bgDJeoMu1%p^6uvYtn|5m7C#$2KL>plpHrzK3=GH
zM-B7@ux6A7Yx%wZ{r`19|8Gz9|HBD`cSv*R@!JkO-P~8-n_u8cK)JL1BO=7|1#wJE
zWK?v`{h5l^+klRaNSfPJUy?hoyypYG?eB7lQ`?_%;c%edUzL|iQ^t&9-&f?Ub&#j=
zrbz|+oQjJ72k+GH=;sfH$+z)c8kzbXr2S4$wo@xY<W@)5Ey~!R>=JAoMQ?7WOm|+V
z_I$Mua;K9~4tTHB%5}*Tu#MI~N(T0dA6%L$8?sVj=4e5g`d}6<ByO6^B0Yvtb1!*a
zgyFEY#p@bp_Y=Kka@Y8k5IpZuuHAI#vDYT$4C-yI3gll-Xa0s0g1YNaDD5i-n0bvk
zDi~`}A@l$t<nKWKsG!u{Xm&Bs_cLBUNthLTJ#+6%u=OHZ7~l=l59~#Hw-z8FL=^?>
z4!YXGb8HQ4`*_>161H@c@xF|;==5Xi+H;RkYD7R<W<fj?L?xEpuGOP-&K6A#FpGUT
z#ysdSdLg`J`c9}eJcnd`pn|t;YrR1WJ?V&2gWw<PS*>4@^!dLoRI^)8FMvgumui+!
z^0nmWqXdfw6BE}a-_f<^z2-{6(xkhrcx40=g{p-2NhPTk^;ka=%f<&0HlH&q#)NP8
z#|;xCu4prJse2<}u*a`XoKqx4>+7dHPU!Nj?ySWSE{!l@qYj34{$~d|D3xj0U`DaC
zJx4Z4Sci+?tgya2!I1j-$8;H3-ec$U?FTg*pfuS?WAf-(QL{ArD6u}BiS~W9O#ZCa
zZ2YS3LY>I;NpqN)zBd}6zJHoO>!}g><+=FKUu}=O1Q|w~dp~AiZPuGvj)_<wVTFKt
zgQ`mEw)an=Wg6UbNKmoM=O`})jsOef=a?}<zFoxQ9^urwHjbOi-chQ)_Q6F|Tk+)=
zLSc5R-rw4rZcr`#IjKsODmmmi*H2a<1BXKcfaAW4HG_r}dtS8rT*7FAq<zn?BSCAj
z?i2ZPJF2jR5Nw(gt!aD>3Q}4K=AQu%w*O6nI7N55R=w&>;EP%jtFNsOWL}I<)2dr)
zVq(@EAq50h22QT$BFshWq^y&N<{fEnzZcEUj|7$~>~Sy9t5<fMh(6|P_g`KQ{Qdb0
zESED-><8pihPMulp&;o{4<WXJJ!5T1_x*O0H%vINc-9m(@mf+8DP*b#NMHoFQh?_x
zdqim)kD3`b{Zn$`{g%$8b%#_dl@iI7O;77YDxWAsPU9Lf9BiPX!3qL_1o3HwpMCx9
zls)$7M88)?br_4)Y%BL1FE_8Sn$xL`&{+WTb2}!0%KFOra{<sF$@K=QH1oH^8V2>k
z>Uh#tNwMjYOoP&*l*8ke6<Hz`kfn1|>=F(6Y_(Jth%M1-ytNFGWE{ZkfH?O^Oi3<b
z5M_rRzIbdcbj5E>m!Tz7YNhk$d*evos2oOS6(fc{SBx-2C(lE@OMfmRppFvth3IdD
zNcY*5lf;<B4&uT28X>?!sCn&NSu1ci`A|NCQFDE}?6?i$+}u#^hg`ll7d-C%s22&e
z9Z{S~#lyFE>m;h#e&nR-N$i#}+$4BKR!P7+!<VTri|#9YU0OqHc<Y`wv)MR9To^Rg
zm(CfyK-I(mPAL?AP;z6w6p_T6X~O}u*|T&jt%%X>_W>rHlYY*)^C*VQABpQ<p%nv2
zRGcE+lTW&shu{W2nfhweQpD-%Y`mo7+v}D!uEN*m?qB<5t#+|dwD#TwAV;6r1ziMM
zGx?ZvhJ(1F;ef7IiZ~g4?MbWe!P*KPYtGhvE%fc3G_j9`T`KiSkaO2@xrx|1`;w@g
zkuwi!0SGY>J2BfHjEM?kFN?ttD<zJAEVO)~PRZjpxM)75H)O4_CYX-)&+;f^$(Z!0
z?(NUjtB<`)YL<)3KeTAZszRR;OQYlIN@ifHNH7P_C*zDMZRcqPq`JU6EZdfCw)gz2
zvwW2|h1VrYq485*9K(qk^4PX!2QUP~%@O|ru15#Z8XO#Xm)A>b-AHtd8N<ob_A=-X
zdw<I8ZsQzqL>+N2$41E2N^0uIX_uBQCh59prH-l2v^S2ql5+-4C`<z_14`Va$&*e}
zTc`#DlVH1{{J;1aj1IN^+HYHB$h-yU+`U-1;?3q1QHBz$#A6@T6o}{RL9*La&qKy=
zY9dJz%_NZ~E(038D-lGsa@qIOeWniP;1}HWEYn9Khksb&4U46E_Z<->n&DUKjI-S8
z&f2A8PxNS+o^z$Eb@9f(ku6O5)4nou5(KPctmu4IXmEs*j&kxXzl1|t3)aga8wpbK
zoqCzfTb$;4Vw(FaQ-tbIpR-L=jjfhDbJRK=OO4@_(c4a<DRxr5p$VnFc0SV~?AA^X
z^^4js%BY&-HQ?@<&TKuyP`If&T%9NK!MvwU&T5TQThB2=p$##EA!4GcH(;nXr<NCZ
z%Er%QXwIKCrCc>s(y{$?&EVPAb@}v|V^)!zEvz`@4QX8MCWq5za-x|-OWvo@okq}7
zuuKlaUrJbktki`tUhpk{n@gQ)2C+%~cPI%Nqs@kR?>EdW*Hy?vOP@Y&uu1rz-Rnyi
zGgbJ-e2|53F_=SQg=23!a0gu!n`5}0>-IU=d4m5>z4UF`pGv#`;E7%PN)y#T3-bx*
zHJ+K9E<^f)ZDuo&T+u4u^mE<igmkhN#C8?EQxolet-Z6>nfPwsC}EOnwf^)6TDMZ`
zd}l%#aMX6%mN;+MD^SFd-h@DjA7EnJP5oNxipO3)_SL~7UR|UOwAnPzSE>=*n$26H
zfi#h{2IDU@!V?@WG_*LPy_+ZBC^{F@I<uK2W*Y<<z^s9m!S(nuKOJl9?MZ$4eO-gP
zcN6<t&X9aw!mbEu5T9nFr`@kGsTS8Ze5X0_+sAB`eI=)%eStrI8#8eNGpQOKcKT0{
zZ{MgGRb=p8>a1Sz|7=bI{%FBys;6JfE@S&IZk}2iwp+$I_$mlQe0gmvK<k#$KM*AF
z+0B;%<ATGIt26<{&<h$c+l;{aeXPoF?u5)tlK!GTXe7&;f5g)*V(dR7)59S{#UBgx
zXHQDLo*GpQ%`1oLH}9E)4}`0LV}n~q`YVbc+hRL}0frCVE0#O^y3Q#cLg6hexJV|M
z#gMwYExV}n7as;2l)fU_F@hVh>?3h}cO$?)^1e2mWaKP{9G$mlIva@e`qU8u&}zn$
z?IHD;_|R;gdK<{o*b%R7h`%C+I41IrFcaq|f;J|7{B2P<+c9sD3UuGliU425anDI;
zYy}bYHu`?Y=d9AjAF-*N@EB=j+Xcm{p+E{dZJLcPnPuxuzN;q9MJ%COm|7i06`0)=
zbD~qycXcV+ykUOhy0<Cq8H*(mboO0GoCMCC<}O-?!G)^vx;~Yn4fJ+$wW8dTD9Pfa
z$sH~K1!>}U3|jNH6TCMoJ$r6F!q=+st+iazybLTT*Hd2eOz*JNS({oTUJ@UfJXAED
zyb+`1kIkTh>!s7;0zOn1@9IZb)56=0Ml8%Kd0+PGqjd^ueH(Xvqg|-QisPd40S||>
z()ksY`n8sVwl5G+73cKR^iRGI){pPfm!xZj(|znumwFz3!3}Rrg@EnOyYCB3E{@)&
zY=J)RD8i+vB2W`X$RyGc0W?roIcGXg&qO~CjzKEH^(Yemk#SW~Q5}tmf1hxf@J<k$
z=K6TS&H`4egN<24rV?C?%C{f{Ag3DROlwB`&p9^BH1&Ya3bjNFrTSZ1nWD-a{2I%g
z&-F60MCo$!g7Z6KGsU-e3%(8;@K!I19MiV0muNO0cwNUSH6Kv}wSFPIezzZup;5z^
zro4m4?W^0RUcJRQZm6#`kKGP$?7N0fp|*DXC63qqkA}O$ziS`8I-~?6dWtLePJH%r
z4$xbrt{gGbtbUtv;FuBgf%wGh$KxPq0&#=WU3Y8FlU<L21S_i(i-S!6iGsmu$mt_l
zw^{h?q1n*7q)8G`L(0#qR2x&&tc{qz;q=!Kez3P>x#XI}kiESp470!IHt$>>WBmJa
z92dVfU6pYMr_K$&vU!H}F1=XJTKGDXj&dJBc+iKBD0Ok>J9<}tVD$dNxX|nL5{FLh
zYX3fb^{%Wk`z=(oO<}*dKym7sRH)lb5%{o{5-7D2B9$A)y@%X0Or;lfWmYO+ed1(+
zcX+D$D`?l%-i2v3%@U<aV;2EiOO+BJ+Ldx1fVMECS}1d{Qx#K{t+u$<R}ko}K(+b!
zFmQ3LScEWMz+X2X8(U8~Q7Gc;n3%<}6QqR?5I7qhuv^jNp7|kDHAH>P=hK)4WsH`T
z^&U4kj$&2+;Y}t|o@;cJj1#8|-oiat?yI+JkH_o8j*|?H)oDF4{PW<-r%fyTd?;GG
z6tYvfH1Wz>BTmts<TOmc;dJkY{1a0=$~>usbjXK)o%SrX;z^-NT5V9hr*gGL)hq&=
zF}!I(9nP~rE%)S79UMII{xzoGZnrN|yD_;jWm@UO<h_KNf88;Bu?n-ST}3kC4wfIa
z9^XR6!j{y!TFnpCTKHCF8{wj5&TD3!uFHHrJ7IcZIxvHPHPa+^gJqO@2_?{slbY^|
zW;&Mh<o8(vO%@Iqme#cxt(l=%WhL374yUV-H0bkAU*_#CN;dJE_L#!VP`aX!_f2^#
z+SUDZ!YjhBpJw$c+dyPo^dDPE63!bG)$9Tc3rX43CcnMW-_6&m!$CA^S;VB7i&Q#J
zA}>4&JHfJpZH71OOnEZ6t<e9gAF($~z|K(pr%`%4_t;-CE2qptrW!~ad4+V_a;pnK
z-^;+N<C|V*C51Uhwq%skOtCpNeC~my-m2uobJd{PVT!~DES=oxH>0By^KZx=dqTzT
zzPT*$@zh0YH&;fdpyVRlUONgdmV(Qv0*dUJXSuPopx4_J`J+)sF=|$`+@xYyw!wQJ
zqMAa5YqzQcWCL<c%+Qh?8g}Yf2l10{s<pNf5r&8KKGR`_h{V!}&gSOe$I+EVNwC;;
zJkrChoS~S|trH`jxxjclz5u0)tks5*5<HEegX0w5n}&dF09N<oHqGJJCgVb3!#r@A
z#qJ!DmfdKSE;mOLqo`(Oqq*<OZg7Y5<E8ragwaO(Ga`61PX^B;iHHcV%<jw({rwU{
z%oGantq3=<aS#>To3`yw))CgJp0i1UZ8jDp7lO6b)l*DcU~M<M-gVHpU|TZph99->
z=0nUJ=rltkyfmsTo!eg)TRAa^wbDsve7uWQ`#4G<0?34TCYB}p6b#tWPtR_L7f<fp
z@Uaz`K$cGHeitm?xNRIY&}LO%i)1oX@q{b)|24o|l<W+dm2ivgZu`FGs|T-%Yj?WA
z=M3Qo@tD)=RuhgUg3Prw7J7VBk8RW{-mn})S9ehJCM5f#5Aibyx7b~_TC8R)M__Dz
z(;?(h-G=}itGag4j&{m+b*o)j>MZE!SXZt?eq@xxw1!92&s)DcsCzRwKkC;b=;lz9
zE0K(8g~|?^LtFb}$z)01xRr_~%uJOo7|YI~`PI}wxB6s|)}#1Zh#mk>)od~aHEnw^
z2JHkQ^*RN(=j$E8LkQ<K|KD{OTxkLN&kOdaLswbWsOT!YMk%_@J{u>;fV#9~v4X>0
z`C{7ON~l;YQG{w_HX#sf-3@Qx+fCFg3SPT11F@Qpet8c5ziQ%mcnt7QTeB=0@1>-L
zUyR56kxXC{OSKt}W=9LNNT2Y=lGfewKwTPft{D-|CsVPNe4`r62Ou78H|k1DjV6^L
zn9Qm`5{YmYsp(mDl*QR>c2ov3L`5yI0Qn|I78heZnPvl-=4_UClpQm63_Q)E6AKE;
zfczTKHfC9yc2GV)@^q-)G4+=$V0TG85%UnVktE)AdBh+ptk1nC?WQaNpF6Y8TG?7A
zf^1W&hp(Yt<V9p4loo!bogrV)+vE35<N7~%&t9{trsA;Fy%NRMhQ+<<$pO77aW|4n
zI4~{Eg9Xzka~Y#P&IS;ffXt6>SOC1fAxl=RL`O6}d?{U$5bk#pLC!~xq@faPozP@W
zG9Av+Bg<sxY&9i6C~5MK2Q#UrgcJeP{&eI$A=1^m^~cNf)Jm;L-91SBW;+f4;gdRi
z|F%(ptb`(zMeYq>k3dAK22|v^(Q{XqFYG5+(Dtw895etvN@8*QzxXtSlec7(4_n=J
zg<1m~OYi_gftH}ko?pLy{dg}rlWWo8##wlF8T>-oqd1_&7nP{i;*U4Ya@NG-y&&Ab
z>O~~GiCCSt>!T`JSqN@}VqFOEx0bmZG(V&R;|(KS8j7TrG~R<0WqA;qv1e!hjASo1
zw!d{yOIOzS9-W$aB)pe=mA_`V{=n3OPfGj1EZWVCJhx)@h@GQ^=sr9l2R@t?u|U_j
z#p+pLr>-yytJgVCD{C<UX}5d+5Ll^dq~43?Hm1)C^q9oDA1<VN>p2C?X>_{wMME)D
zG8TnKp-kDbu9BJ|(ut-4yv7F{Ip9yIhExirsuw96HP7D`fj1`)r~)4GCB<swO&?q!
zY&}g^#ken;W4b<0Xl!Q8JF8eVxK^s5UCwX=VVB|EN{8IXja}3)@1swxeA22<`Aqxb
z-$K0{yC<5h${x9_Hop)87eQF~)J@aUNi_^-knH(D^?gl+!Q?QuL4h9r`k!C5s_R1n
z`O}v`=R<ysj|!ZkmPTH+(|1_eeO?VHVY0?Cl^Xz2WjLKBeVwGS=CuxxWAYqx6G3vg
znUyomp{B3GB<SoY&R2B7f6=t~wa3y+2DdXIhaL@O0UE)ov|Ye>^`yZig^ZNP*O6jz
zCUcgoVs7k%Xd<uDa|RThi{MITsUcD)uJ!kIw4HPI01VcGk0lBQHsLdxCP8BU5TQV5
zy7U_RuEbFIbrOr3!cqd`4%hUy4zn8AKc$rpS`2qewddOqXkAkuiKsS*h6q>3#s-R2
z-@YHoK)(|z^<?q~vnkY{21YFueEu0nUk={Cz#J>c^C3WB0Gx|0=Yafax@9UTyXMPr
zwcP?Qm3jliqwZ)ZS;1_R4hDoGE3TLm0g2qAO;cPAirF9ck#sO?=Ut=OfKo$_!8Zcc
zP)o*vC2hbtKjWI>ixRirqYd}XeOa!THz?}VpS?}pyJu%!{x$mCEV+~@68YZ=#Q(w!
z@B|VZlwK)q54_Ab)$iE*{R&uSLWwy3eoMn%)GGSS9lrKukTB<X?wOqLSYWlMMBvMp
z!yEEA1_GX@|KOF^Kc;tu6}Urh{){m$GrYqSRa-HMSVVu)I@(=vu6nqFf)>oxS;=>4
z?JpasFoZxf34k7~-AJGInE*_$J`0D8os?~|N~(G>#eB-1M8N|)8N7iCzR`Jd<K~S}
zBRk+l&!~L9LN)ZVRjUy#(#T#H!4Scf2~MRz-<t#SDG+~b;$TZzh0jmhvQ#D*Ik(<*
z#l?0M+1EXmd#Hl*VI`Bc%yWKMt_6&||M4<erE>aju>XT6IgeKG*A<0nZVn1>VrAFJ
z^zU>gii<8MuN~*AvIxw6S}Lt<eTJt#b)hq^0ONp>3Qf>bg`#Qh*E_R7Z>g9Rt1s{I
z=`}(Oh7x!98d;TNhG~w@+8Y{Y;Nau}76xA<m!-FsX(BjQ0&|MS^OLZ;sy1aL$~o&+
zl*a?%Ru$HG)lu<gpnvU7`OrF99hOhZTl{*PU%XmTCyZ{F=_k7CmtEj#U74j^plc5r
ze+$PusANVEt-!Klp>x`ONDV%4sUR|l^`03w?-Y#wt)E)vcGi^P`8YoKGt+u!<jwaj
zM!S@as@cw8i3_7t`U!2!+uMDWvi2wIpf>2(P%YnHyT8l$)#DW0=i=-vZ9$F%9OK1|
zRO^c6l}5Mz6~nK!ptNx0UzPYG^L)7Yxq;(T9+SjoM^T<#5v8_ml$mrbO3kWVRMY0B
zQrqHLaF2@j{wKAy(w;^RFvpqxVSI1)W&HQN$&+VPaC-WK2#Vn$twD#nbXJ~-l;F{0
zwsWx;>WlGJtk!SA@dbMDq|pd~m0n`)VGZXSQy7@aKcqA8m9;u<86Z7^`z&TDOI?_}
zp70)1Ez6*WkJ^n&`7?{3p86D{USfqQA{@#eb_|s~&C_DD=~#M3BX$$Jx_LArG-1Dz
z6bk3h<9@`GbCqrQp@12~r$kvdaeOzL#rnmPEGVa$XY7Ns0c?I#i@G{HEffLDC$c50
z+9%}bAk|n$2;smavjL~Ape!k`y_7ELrm?WDvl!U)2|Xz#Z10U2)vlp)O*)NFJ`JyN
z^im_SgvE3ht%f4L)eTT7i%VVblT{yPOf;jGvy3!+!!Iqe>|2|W#c?Q!#a`Ekc?IjG
zCW{|h5fnk{2ERWV|9mqR_eZj(Sa3EiaZ4SpH53gL9)f~c2kji;d*%wu-e}AdjZcNF
zM;8R==xzD*lvtXFy#zJeBcrb(ul4f!KVIPG1hh4C)nPQ0%m2iAh#m*(jTv~b<e2Rh
zCYbK1iVjwKnnjlG;Gnt%JEfRJ+~JKT*9w!sYPvIXAMw=JVs2v}7o+NM7offb8EKM-
z0nap_V`Hjr#ZaU6oTOGB)b03KBXUe-Jc;o(z1MVho||;w!>)=+?S-Yb+)A^rPR>Z0
zrjp+kfi-Y1<@@bHQv5WV|FRiYirP9>T$q)pBF2%UOw{;tln208=|QFDmC{-*O5vTr
zLC^DuCBT<>|FaE|O=}Ko;Vhc%`<;BPRpq0^^VQQ~83}hoddn5X_xy=-_!TCaifL*d
zQ3tTbXAV^`17Y8+ihF~<sMPz3yKuyIC%f9(OdEwW1(HT>0-DHqrX;4dhEiT`GJ`Um
znx5*<x8prU(*^mSM-V9|`~_}Vg~_T=k%W_bF;;J~0aGtUFh!4YFgjx@PAEW_maU2c
zQoqu`nl*n|qmJz#_+(7Hq$J%~>Zjz4bWbaRnp!dL0xN)H8Fr}`m2yIlPieZKoR%JU
z0dXR9AXW}CE{pX{8H@?<Z)G#8qi4A2au)B#8|}LzF-C^cvk5asE)tEDTh0Nf!B@!;
zDO1}HMhMq*Wei!%iE%k{>x^o!wo2;Q&{oWS2U3s{PnP%7GW;}3aK~pe#`h$@;Ezy|
zeJ%anvv}@TAL^2#i5Qen(fE}`MB3IMl@qI9<kr&%=eZ;Y6bxCAyT_#_#Q}egzyS#b
zf7D+WghYhAiAz`Q`*6h*Xe_vUp#Pc$!3u#ekKw=*y3vGX@t}nI0YtIsuc&Fnr<!GD
zi-Eoo!v`qMiAUAeR^?R*8ZM3oyJl4WlVb*L#VhAaLM$D6#m%eBw&|a?7ce5eC9jmK
z3U4)UpTqT+;(T=ExFgsj1<kGP>XLbVUm3?2j*rXXsK2#yQsa=~YGuEk7Ao6T%$VPI
zp`ty9&m>i(rg=TL`kKx1T4<EEGgVHcQUyY74;QX<l3emByp@|1g-mY0GBS}?(}7sG
zRKyJGQPcVb7q#g+&`X{Y*H4`?aQl+K*kd_pb6zNqc-vk3W>LtQyE&M~4pD?@?E5mO
z!;;=fWHV%tA?>zQB;zt<YUh>Esk&sp1S&j{)t+uR5vHvt)^Zq5P{GvYq`Am3r}es*
z3kvZ%P*7m*S84|p*@iG48)JyPr#7U<)YNLQ*8jm{dO6KT{P;Opcf7XAo)HX|$7`2k
zo5Yq0yrF3oQ4VVNiX7275;*)Aqo(N#lsXxx684ifVz@p1+4Ms=k=ST#k8yK#h@9Wa
zu>cz=WIrMmkg2R-N0SZT?Au_GZAEcAM0l7%q=ZIPRfWp9y8=Xzj459_!#nsb$zGRP
zOqzytEmQYiXFTMZt#~B&`=Q1vnZGJ;1K~}^I9CC#5gQg7%JQdk%X~7c`fziRW^3&B
z2T)l;aF(6<KC~uGlS)lZb}X&DOT8E6Q&8NtshIL9!}E=^ik2$!i8=2@XlbQUuo3@X
zaxtA+O@ZdZ=*2Fr%CB~pA9=w>rd{8FpQ@$qt}lZZtfuI?xesEU%FYsiMLji#wPhl0
zekTS11nQ~~+S|J17n1arI97a<k==@c`U7l7Bs8M7LcM>`#%EaTHD=K}c@HFWdtOx6
zSB9oxpfiq%=cF-oDMCe9fU^d%JZb_5lMN?f2m`5DFosl=$s=V^6Odd>C(h(;Q&vX=
z-Y6G1)A3R%w-#fQU^-4gGD08x-Yye&{pcSj5zLVqv~epa6fpcnI^2^O+BCP~wf3t$
zqfbf^BUYB&+Tn;_Xo7-IDClWehvMUaO@~R0;&gQ(QnCXofA}xnBHxT{{t-Wce<)v8
zDgJ__D2XVYjpDN)P97EGZ!cO+4wsBJPkc`*#mj}FLo!XpC`UCk1o%9T89g|yW!d*k
zT9z);&|~eu-1mT4Bh06)T0D@Xya;_emJaCe%tr~{khi++(92AEyHHRXG_!)Frv6bV
zN)p0D`INg_>%*LOIE#@3mX?xDpT}pn@4g78V&-TT(4g+MoD73$M!5=Y3vy_e3?ha8
zgO|E&G6mzGZ2Tld6I<I`#BHV+zL$RA>PIb$O3B2C)~hgaW&`wSUnu`ghPR523?0U}
z`aM1n#AzpV|KX^_B=hYz(k&t%u0nMqZaqc{jQ~IklAJstuh(XcRD|HzF()P#pNEl@
zRd3vR#w0F~VWExC<yyt~v9sY%h|`t|$!bc+#Jgud)8G+esQZ^lKXbo3Z*gA>n)1p?
zxZK4)?=n@HwKEKHd5F~rwiX46T2o8LdpUZ*sg@Ulm#*I&0LpbI@1Rd?GR#erin1g|
zjgUv8Kw<>@bCnBUJqJ{{RH5|EnDQlN2wVPHBb%K^fStyLH^x6L=Y*Na2cbHvSrWz8
z?xZR3QX?%jy+Ef@PQj3P1y=0n$t^0%i@|@^Ca-SncDGb((-(PN8NOIy6R0*GMk+p5
z^^)M3;I*h+m>(?APUnrA?md@)WMi+}9~elh^*oCZvp^=>Y~T!GMYL&?dB&_gOreS0
zY`!Rh`S4cXPDO$J<*rC)kE@E`o@t@dlt3fy-Jd4FLA8aWn@f^XN^sgKW?qqi&wNkl
zj1GTL?UP0gk{^D;h(<cc-$B31M^%NA{oFqI+@=vkSD{>p*ky?pBCoVPHm>K3n)N=h
z%`8?QQHi-gbR?dprFhg|vaYU3t`P3tKL1`QBB2c1jw3SHb0wDk3_r|NEvt!ORge4Q
zn}yUbG=xtivQP8OYD80TW1|fEr&y>KPaAC;e~X*lloVytZ!=QZXzazGpD>A1Qf}TP
zr{*L@?)8PyFsM1lbIY=dl9=QBWWVI2pqfqIc`QZc?^of0XfuhuLXkFU^?J2v^s(60
zgDGPc!pO;S@uX#yNFrWwlgUE)$T%FLJOM{lsV`DjCE*k9oHG)q;9`8OvPQHuP|>gB
z^65xI^o=@#qBOE96H&-=*~S<KiGNjUXw$rM5gm}K(&Cb?t%)8@kPlNYpDBy~6}^<Z
zXr#e3<yG1|H3xvjHCXW<%qtB@BvNh{-C_oJa=h_l27kB%!P7#3eNQZR%OQm2$=9(j
z5qj~G?84k+Uuwrw93+5N4Gp18Sc?D!gQ6ixy2T<#UV*?2LuIFDeTc&RR}n<kc7yw_
zUNj2Pq^z}9tlcyN?OYxxPgu2G$QwajLVltw8qC#BM%eBncleLB)z!%H7hms?uI8f3
zA`v@1acfh~RI~XF%#F+bppiPaP7+3jZJKegm+VP!p;KVYL0vs*dQT0y)O}iH6yK4j
z@FmH%lcIi+2J0Frm0wlST=1INu4mRSUh40@@Nuk`u|938%H^;!zl;hir`b&-kMS-x
zEoruei&BW1eB1^Mvot5CWHYz1GHZdU;@8`JY}^z3D!WT@)pL|wBn&9%p~%xIo_m4e
zgK$wFD$A!`0{DEAe<aQxdhMSyt3xTa%ajP+b_Z_?(Ihn6*(BRi#rLP#iDVy8Ot*>m
zp^y~Bz3sC_-8^WZ$ib3*A1b#*M-exS&9<vzA+;}h)%2g*bVZy2NkWp4n8eEf-W7W`
zm}o+^lIL?$R$MrR3}>OWd_1AyMEI|QCDUI=LZgAT+{i$Gx91Opp&7g_HZBZv$C+>W
zd@6p=1}T-SKRpG`1I=PJd#)RziN<=+D-Cu9TUEE$0s1^izHRuKy0~3KMBDe$W8Kmj
z>Flt#q+=}b$BM12)8Y5}C?LO6P&^dV<bu6p6tW;y0&wnQ3YF|$$fyoVa<4UA_iubp
z7<o-;Oj<@dW>|elF<zR&!qjmdLL#E)u$W3diTFWo`_<~kzR&3=M>-Rqh7}{rA{d;=
z_>o=hyyM&~>Huvk;l`pIwWr{?L55xs_y~`muYd`OHPx-qI;8uSVjipXLR0oT`1a@2
zD0C#d%7wf%yTo~Yso5izgQ=Ra>Q2obHVtF`y#QO;GlbFR`@*PR<2111^4UY(12GfE
zvqfu^TE4;7u$ICJg|!Yl2dS*f3xopl|Do-zqT24l?@wy9P`p5KikDzTf=i2A&|txe
zL$CtD?QJRU?od1l4hb$T?iO51ffm=|+R1O`KWo<9%$kdt+~gv1l5?`o`Q~}{{%peI
z6)_qO0>uk$h8t`d4=<06zbtFH-dH=jD60ZMuAng#WNGR0#IoC~szQIHC7}+Aj8jl{
zjyN;5MY9Aucq}CDAN?u~I##3<`Is|Q1iOI7ExnE{bCXx6&FgJj!|cp(UaUBj<^++~
z%L6L#CIDxd9u9h<0*h{uFH)B0wH{y&|8U-mUM}=Aj(nu^f2@t+OR<538<i>=MQJb<
zC}n1*rXdHU4#FAm#U*n_+D1}-Q@u(`CNUdG$bB|-vT<AXd;?!mxTvmLIY8c0-pIGv
z@IX;r1%ysmmo@`1m($zITP;x1<`S{cMa{7!?%NGsggQ$NhmHA5=W6#lpt;J(T$*%j
zEgpy#S%=e|r+Vz}v=`fUbYg{uAv3Xp`fLWfhpy5MHvI*<j%X_PH!LIVS%JwyMF%p*
zS*bs!{-R=dlrln}kZt%)wWmPjm1}NoRn&fj`nI%Naue^p!B}V7I|opuQX^p~9}nKU
z5szY3+$X8wWTjWhK)f`uzd{IO?&)DjOF6}%_qqGjXz&iqr`#JLR%6lUPuJ6vGX8=^
zNVt+?7ZH)oD)_{~=0e`frZ$qS++bgspEG)U8RWJX*7)C9wzLyxg>=`NuqOlzM5wVl
zdZ^$FT}nS+t1ndTIXaUg@VB>LtS^RZGk=H<#rlrsPf1IcDrU;+vYCoihL+oWNSS(^
zy!~^2!{ikLzCLqjzUjSj?Qx&@ha-?b8S1Vd_FG9faMBLa*KN$_RJ);FpxR!opkV2h
zkkPy-xL@k^*E^Ts>9``Pe4ivv3qz$=@;mq(&*NHx9TCr1k3u?!y2|Xjm@-Cn)f6wb
znmwIgs_;ob40p*SbmpxcCO!}1I-!x+p3h(5L|h!D#tL}(WEG)D{dPx8p@%$ODVh(r
zQZY}l+i<GZ@_j`AkBks5t(1tF5E%|G+MD|2c>~Hxw^SZ9gte<Tq<TU9ebm?xRKtR1
zuEorjVosk0f-O+flMyIZwY^QG&733{an9+IEVIjF(5j8<a$l)4?iLSYTlTqSzWDH=
zpa;uANIl1oq$M^&{l=PmkOzu0k#<@IvdzYs;HqVz*@7NFl}21mi!<GQo>-y5{cA>$
z4-2VqnVA<?S{JIFPWD@~Er7JhGK!KRUMo7%+$fYsGqG%Fmc;;zGn&fB)yY@S3h)^T
zt>k~%&5xVL>0?JN@>0(_%?!kO1XrqTaR$CY*S+$n__9Er3R9jel25ZPrWCMXBop$n
z?<=q9x4PIpjy@@wKjl}#zp&i!oJ%t#9x(Q4b)Rv$Apo>9r7H8mRx7pMWgXNUgyZoq
zyUin8s%&XbyeE)f)YBUL*_7_y9kVch=rZhH9LCW`Rr3?kntnm(z#uWcoO2Cix5Z-k
zaRP<wfO@@2n%O>8DF90ns|ah|3h4u}U|CcP`W)aAxA#iaH_p98$C7)RQsHaV(H=d`
zGXbcoQLtL7@sZ$*tnzzXHK!h@twqz%*~&HRot|r7RcG=83|*%R)?Fx<H%FpozTLx+
zN^<#RwS*{LUJ3UmupdHa_ZUN|EvA{wP^nU^%AVm}Y=(Il%KC6&+)KVI%@fOH+<g_9
zoLG~`TU@j@pudJ`p8=3f78f;Rz%3;*tJe2_RmJ;4%LVaSbN2TsV2b{_k9Kfl<BM5F
zFxfQv?kX0-1sWOkSi+qPrq<uGmW<AawDoOnjy{X2msqWZtY=nB`a#PWJByXHM-69m
zG9V-W>&~$7+x2JV)>5!s*kp!1<pW@8Al!YPLB>)`8G)W^c!MNv#&6A>S}!}(WeY0w
zkRHB$ye;?gm+7+|A~r5{Q8jAb)Z;js3WEh@RnId!ew`oZB1$EenqECX#&bQHQ-`k^
zm(ztEXM?#uB%S6Kv(9{$h?^?z!s0#}I7X&uhI7XOn0ZiXYK@;ls1nKJ`&MSA@6rqE
zWp<W7S2ykjS+e<ECfz7(QxpEWHk_euGD3{Kz=GiVB@!K5xA)FC+X3TT)Lfq47`NQO
z&Ul3l<)lXNOtY4<&W~J`kBRZrtDwwjP^bGlDwF!aABHSS5dCh5Kita52b3K2D1_zR
zj^D_P9w+#Q*g~~OG`Rsj-W$L%ENeKy+=;d_el`F<-^Ute>#4q^gORf}cNZifYW`^e
zeR_(BYeV;k6B-XQOI0O(0?MMqL&bIk5?KWb<AFYqXs}zJuXs!ekk-dDoa$OFddih)
z_P{GvPBsRN)wHu_9OtiGgXnR2bO5>n5eI`F7)MR{gs!b65*wN|jz)L$us-r1V2Jt=
z$_`Bw2a}25639E}3AK{OTB3y8;Stm2>jb*tIz==(fC~H&PFN;_QZbyg?%e(uZTotT
zAGW521gXpZFvM?+wEko1<q8L~XUzcsTZdDy>z}=~w01v?A|u>VtDqf5ajmb87NnLM
z!woMp#1+Bygz4%!gPT$&sV}B!U6y*N$)IP5OK>S@-mAoP7hP^)Dvb6+mn%0?moGeQ
z=~;_La3Op1nX?&+_5y_;7<}@TAi!J<Wz1RrSJ-M1(gX#G)hQ1vkWI=-EBfIlAZ4ja
zmupU{130@Tp4}~00j#zSc0os2a~ym;v~7Z2@prOh8%;vP%270F6qV}oPSn1>TjJaz
zBh9>n_T`iiRa=OpQ_^FFl>K!v?c5@>ZRf~5x6_)-_c-<znRiY7qg5++n>N>O3yFH2
z$CqSF^*4cCQ0X^I*UBX;qQ3mS6$9C}JO}dM+Syc$p<r;ws`3}3jDI)^G;C8brUXF@
zG%u&rEPFyZ!uf?MJYDN&4SKuvcm&6E_=RxMojq66csJ2Z%teVT+#*mUnq<d|`J2eA
zoqS5I8TfX69q?)D?EJzul_lLs$c#6O-J2f<kfF+1wN2abAIK5w9MwVRh4kS#xxqH~
z#)xykOV~p84qU$bP9a&Rm8ULa>9#J*i)VS3MO&M?$E7P&`7;IX*8*<ANts?XMUtAT
z-0elxl5We1v)0f*P~NQ1$hyfBcB)V=MWh#TxR0hal<T<XrWg8C(sv;O8K-9$e=%Nv
z?>j@lCMXy8=dwafh74aWaPlB2AEuHSXzjhL;54P#H3PmU_}cf!k=^RFECcEV7AT50
z3WbYN^RC6w8I*%@*$%JJs^@7#_6-yi2Zyw&!?R`Yi@CnUM7b1)X1Xvc!AOh5a`|_n
zl^{ZCPI|UHz2rOiG0GwbHuhR(_484_Y2%Rm4=M1mQU}k^&bF&YQkBt;72Tpxw5eai
z+0Ki}kHCL8-DJ)K3DyNXmx|euxUcV|qHFFpYyJ6vlHA`V|J(5Vm8rST;8eZu+3VM<
zTI{F?0hOt@*HkUktNKUJ=tyCL*oOUkQ%ZQ;S<K*WP#zV|e>j>MTq^Dvoyr|kqPP(?
z7HheL7VQkb7O0KDP?fHVQ7!&Z{ieDD7p?orL<IQ7L?X?+wmq+-E{`2dyj(UCRQNHg
zXSTlVMCib_jFMHY!az+fr*wG85o?rwZC{y@<5+JX>`dXqu^pQ4wAh1nNn?%Urxiw2
z$*+kR>S)|YqCYi4I}I+kA}G;eq0BAkG{0P4Hy@e|BH*bisT*&@;9s3M|0c{`r(eJ(
zMqKTA-mQVihS#YK^?Bdg1RhlNl}gtVsr}^^G3e1y(#@0t!Ge9A*=$I4+=;EXcc}g4
z!)o$><~T+!h19IbkxGMRa0j@vxQncc>kDs}&rnGafv&iiX`?vIy~E<bs5z#xBF~D|
zH!n74xpIxy)?y)kZQlu8I=9PXH04;iJU)pZ>SLmxl;z$u!ydMLn$+^Rj0~e1K2fTz
zB-G@TJr+!qFQoSRgP&=iA7{L2c%<bTUJI;zF99)DB9{*dCCfBYYRJ1D3Mpx2MX<J@
z?O66ID59|?M&VNn0{La5aL$a-q_{+9RXU$Y!1my~8h`4hzYT_i`1mr$HFKTTOODVj
z$7&<5B4^&x{qgR$cRl!N#2S5+C+ug4h0&>Bs|1F7#k{$4UMSCJBZ<%zm7MgX)#w!K
zz9@Ua9Y0S}nB76JwX2OvHFQKX`rUGsNcs<=u4!t>Mj88CQvN4)#y~CWFKWV1#e3;(
zW}+VE(1x{%16jFQ=7uK+EDpE~@x>09T24)5iL47NQqFqAdX}vFOD%e4EG=`c9qs99
z?2Jvf_B>P%>>iQ+{=A+k$`fuF7%o&HR<f=zsf$zrm)Z-pq%u>0-f1<k6e>Upe@9O4
zkT+>r<lL<2U47!5FwGX0)VrZcFNy(S1%{)qE%5MS)S@`}C4W*DtE$Q>YXbz`>udk$
zTNsnE6budv)3q5bG@X)8L1Z=Pd?OKIYTu)&C0nFi6cr=THz~Ht19A!BS{ad~i-Wjy
z%Ic8_7*w2F!D9)ZmTJ^z(xgw$vi$(McSPIuXoOu5E04;_+OHbXkY3}Vgr5N|pE$Nz
zD<cX4Sc_RucDRp01m$_Nq0k+Bf&Xc<4GkeX-1w%!T&3`VwgvdTuFhTD{i1U+Esj6s
ztcazQ*2DhO+K{-wBAt(iVK$XTORkQ>YmSj11+|chXopy1(KCP8ib=VA<ZFEUpRLBF
zvJsV+zm3YonMJNLYJNM^Pq3{oSxFsQEMU={e<h+5oEv-;az^a4)6_1XyS?P5j<;zU
zLNvJrD+7AOf492loI}<c3-##=7iW;p3(6(k2)*KoDhcT*4IRXkd8D{{!}5QtMfXL4
zU)-v`9J3N~TP(YV!O(?UEeh<linO7J{J-DKuxE0B5Q>RT_hz!tSP2mZeGLlPbmRQ#
zoXk)np|-cYPTg(g&m6qcqXA>x7!+b(!dyOjVL+o5ZEdI44KqBZrT93k7%qrU%d7yS
z`kUZN7^mMpreJ9~-US&2T^rPobCL-xM4;NF(X|pU`6+Sj{?Oa)I=bsyTimZZ*6j5B
z$?AQzGAng5y1MP^kkr|Ow>Xu2Qyq12d5oNRpiinS&0=!UBmk{LI$rf&Sz5{0wD;fW
zsU?O+zKvK`0G69Z%3KF%4*i1eO}xQ&iID;qIp&i~ejAI=$uxRXA7Uhk$Ga2L;267w
zh}^{obAE~`L4zclKeW#qT-&5Ca{N2t!RHQnRSFk8TSeq2ON2FcJO=t8H-B923aTM?
zDxkA1?y*F~z9a+`Dc9|fpK?%s7)NiMlDjmPtmSoHbQQW_&gb^4eSHX__!9w~Wpp{R
z8(}IM(ThVNlrU-d{d)V}r4^8ys#tiW`a|N!-X0)<Zp0R?;iZWtIJX_&-6(~6)wf8u
za|T0l3r>x8Fu2}~bW2sO*DPMU&fx?X59xermtxyXHtlQ%-;WV|O#!AIu)dM~hojdW
zQU3|eUi1$K+&gnX>l2x$jXmZ+d<lLr6dg1xY;Dd}x0rX&92-St1nDXL2rFN$Dbg?Q
z=dcwRnkxG!*Iu%qhevzRHtk9-zooV0rpFdv5FAEuL6>w<(>+?%Vn(-rLDu_MRG+o4
z5V|f}M^d=^-G322m~lo0+_8sLS*>3T^Y)8PCXs7`95%`KO&K(}3#BY%;TX$l(sc9l
zSC-9<?#!EgW767(g99t3p7M%&BFQ!Ng3Z#+0wVuiv4wi~hB#TSWV<v!e^rAF)JS~a
zgCE8;03UJRXw_9rYf63DK#$>aFMr&w3Np$#TOi-18{}ul(3SSA-^s#2rHckAqD1}+
zzV55I$`CHJ@Q8(^1s|t`cDMVyEqMj5&UU%fv)U5)<PM2v@umYNn)Rw)s}xD{rth=x
z0(^1eCvz8uN<3D%bWBBU`p+9KfY4k|=wsj0G5=Y~IqN@;o0(Yg2Q>pDG&XGeli8*;
zO}DLS?b5S-lb&A}%)ipuv<nQJDpb+XhIU)>?SYgqRhTC^4NAURL8)8|$r%YL9JlDR
z!9XXNG{bJmiUj_O8x@UsN5E`ikNc#%>M`Uu7ZepLGsRZJy|C9(HtQU1Dt%xD<s?r$
zy;vq0KGNzCAe$hAnWSifWqaI3C^VVFv_x}qmn4%$r+|mdjEiB~oxL?!UVrt<@<&8W
z?6U(fOtivkezO^lhzX3l#fGrij)ulG;IeM8AtKKR6kB+V`xc5ILPPiB`3`efNnna&
z3v|X<2j94n8@FPUaNu~I=S9PW4&l1)({U%M#&*OzCA2Wh_xa?PAE9Gp@U<uf%W(a?
zN)V>B)KO?Ba!_TL8@EL2b*&A+;=C!V)(RVK0WT+HD!25?YnwECg#3jEK&8r-^JssT
z5##~#;p@}+IHko6XfuJN^aV=mM)=uu4y^RZ%+DQ7TRMh0pe4M*GZ3$lS3dw>c0FG>
zrWuYN-H3k_QXanB_t_rnz5M1*Y1_ln;N5<~p=1JRI9v}LGx+o2LMA(UKM;7}a4(hS
zRE+ZaZu0V;E~=h*c1pe@9+*t>;FYjspNyy9_<Q(<_RwI~+*fnkE6Ji!<=vV0wcISL
zA{W+oa5=ln29K{;T5g{+wvt+XXkKiu8s8mhAj-%a<L69VLBiU0Hp>K?EyKTf93N_6
zxy1>;f9h0UT}r(6U7U}3VC3w33xbeyybT5RX&_UY^^ra;6zjIGZ`-VIa}OTv2a9|Q
zB(cv{<3HK&U=J-3CAU9@Yt<}6)K!`Fj@u+vV<WW#T?!Y@|C`U-L@tPFOb#v;^1rT=
zvI;>Z6Th}pmkZBS4hVr0WjdCwh%bE2x^y^CY{51dxSpHpqQ<yE;CpH;G-dBPXNe+|
zT76uhGKIjr4${bt?4a(9Y~=lj^qyjxJP>!(ue^3cgCr=j6+Km3b>M1snT<OnJ9Cer
zK)?2&e4=cPRMGhmkuwwsi*SXF?q3>WZ1Jl*5N1iepN@066zej4C%4K4H7hI@jW?4?
zI?#MvnyG0aW`U|+k~2V?;$lQS|7PjO;frJvQGUu)hYTxfH4A8Q*87xbO~}_?ch+?Q
za+FY7PF%wMD4RK%kW!2G8(C}7`(ne|`2i=7{r-)xtK&HSQ2uhmqOYw$x1haYdVJ`F
z&#0H);jLe~PRUt2onOq&3^D$pKf}0}4hY~v7hSl)dc+1u?I@C!kPH8Z^KzM_*7{SN
z;064LP#`z>4RG+Y{S;U=<#)x;Qk063FHP9H;<-KNuZ>EIp~|Iz1wAHcq1ex7Vr9sx
z3P`M=(z!~G)U)0EY>-0MXChXKh3Wu%p@m2>J+I9cTg%nafaE4kA6aVg#MD+f`bs3l
ziI?S{Q%_KLYcL>1{s&exX?`~Aps60oIo|;!Z3Ti1-j3@WiB4<+8uXOU7WHuThAC@Z
zb?#ZmyLCaws8-|21O0;cMIYN_ls`uzAr|>Y5mdk?7l+e?UrJx*S7PkuNYAdfJz#HZ
zf3bcUlF!P4|0q#RLX2MP?DXTe5~<F9{E(B0<#2Q;;?jh`Z?eYalvOXpj4OIl!LfJ~
zn~VJyD@7LFUv|QJKNG`Kj_}Cty6F@e4)%XLo%B|;8i`n^w3Dr+8@153?4$q-a-qlj
z!<PiytUpg!0APbH<zj=khd{h_FFL^2RHbl^6<zbt4*K2<ON8!(H)@Pv$~h<&&xj@y
zS25v6a^*u+@J*DIH<9JCJnF@y4!NQG@s`8sXh1iRWkgL@xpvK`O-fn=Q}*YVW!p>L
z0#)^vgShQ_PC+94rJGq*4NP(SN$f1NeWFMfY`GHSk2X9gaQ4qzZFesB`-|E=c6kZe
zFPN$_c4=1+6mUQz3EiJih$}$ynHV2k>Cpth|E{;?5efXIG$9B`EZ2d8X(~0sjgm4-
zXSQeXhV=AvMF=fG%3z06QZ^3jO$!N8w?Cy8@*xz_*_QmH%U%F8%csk2c+K_|@jQaK
z<UMWnE=qE-j|FMT$o0hpJec`MJ@kYN3-e$>z9aUDYxJ9?cYOh%+GCwR!?w(8JoV|z
z%DqEUMbVAgvw^!-F@D8r?1JaUT>blt&t!TXDknDvk`HZuc!*5SSRQo@y72tNIRjpq
zZif*4GU15eKj2dJv7(+qC5P6S95_DHzVAQHW$aCkH;X3HXg{18iW$f5xbrn@{wu;s
z2M(ewHS4_f7H8v{K1p7c+R4T07S}h{?L<Kce0_xA;w{>umX>`cnxmTYQC7{=E0e=9
z$u0<&%kM!0`_WvZ?Ow^9553GiZCUFVI?|;)Yo|hjy}C{DyAj#w>^3CP(!<f(%;#gC
zc)#PosS{1l;X60Xji>CrQopp4zs9Y!1>sZmHOB3#oErTsnUc7kL;xeUKs+pSK@Aa7
z<!covxpv{`2}V|4qxxGe9)5S8^V-!bW3#Gi_HCc^>2X7b;y0;u5MeUBD&!&8>8=me
zm&I9ghV*S;({vv8$1u8RY&r^cpgQ`-COfyI3Harzn}avg{D<R_`231m`E=#YIWggu
zdWo^n?mZ_Tg<^;zRxqcw+v;z?Nz|917ctw5V2&m!>gexK0L6-P%lO%wA8+)F9zB(v
z3Q9ZqNk1X{C_5<yp^Ky{eTp}Pnh|!hs6Z?FDZ1B&@0{V`%_@Y|X+visbWhS)Me|dO
z+YCvIeDXLjHtflIp3jM<INUrf=?oKCJ|y@T4A)5tnN1N;)s&R6j46Jt^oB0@+I?^-
zPyR`9;BRzdEYrnUZZY}~2Rh$;GY}+V{qSs(X*RQ?oL~jBp%|k$hPHEZtS7ASPjaTo
z-O!erPj(t@0dNcaHIG(G+KZJn&zt>+Lomp#9kxdmzR<tOg0s_Ln@AWR8D5;PkRDCR
zxAzZ6yPib5R^v^{_q;4+{`Ni!$TN0*mie0Xaak)<dKqQ|jYMk>o8prOf(i>;^=54j
z=3>m^Y)LhR^(Nx42XQE|QEqw3svs%6(TM|4^xA4k_yp}nG}YibOw*~}Qn54?s@7cT
zm=?|1jLJHAj3&$q1(a)~O=dAhij#*~3#{4Oh9<sEDilf|M*bnH4s+Bqk%-{Lz@q*T
z6^N;nPnG>@dHP4Xgo*wyCdos~{HCBUJ)ZD*9q1)iZUe5X?)bz0r>w+g%X*_bGbKBO
z$P!}tGV!?ReN0b)Xu@(Pmsvkcq7UbO(ck!O@#b9PaYig;kE^NE^N-6U=sW{I5I6Vr
z%Y)F#baVB+FtppGy=n?9W#qq+e77Am=Xpm`Km&r5D0c&0=#|Le$m{R@{)~st62$v4
zOE}$8Ega=6sGS)b4IRyM2HNZ-kvHFbkyYvaEO2FZ59H!wD(882T6@1&lc2BB``^XY
zCU17-z}isp7+vb+;?kRLXK_;tZdia1hU8;;E2}^0N{VQFtame$Nc~*256F)<=Xz*C
zB(vOr%_w(~38h5Kl|E~4q0qJJdqGDV#@0`Sf$sMONYo?k+saQeO>r=fLt(Q!7x$sz
zJiYZiAz+7uav<)gPzIyx8SO!8VN^$+5NLmRfABxVmA`y4X|3y)Pn2aQO*%u}jqtOc
znPvm6omxxb;<yYrmInkij3yVy_B&&vA4|s6S4;5EI2jai%kg}zr4;Yi^T9O5#+Lb4
zB(pR~uqD}6vT%{IZ=lPcoPEfQC3>3jMOerDtUWO=?dxb#16i<mI49$-B3E#KA185I
zeuV37Eeadf?DgyDo__7sY76$N!%-cp1%V=h$Ms(EvXu1Ke}AGu7=O%c7vdD=d72+A
zag-=(EPaTuD54PEPJ!91q{}Z$bzK_i4jMcAk9Db8mc1`aGM3v(I#UhNK-~dh`EG#?
zwpR8}<~`oG;CIK|Rk{VBDGexkQzEuiBvYXB9c;Z7esW-se>jqTo(`%pL*=tHUDghE
zwZ|K!)!;Ko-v%L<<-U>lk&mk5W})s9Dg1EwXDK6#hCZ?+d4)6S<V`)RS6u@$K&(8l
z;*{~Tmy7K?4vHF@`w)C*bd<0{&{Wdzd38;3$=zJ3O>Fy0Dico!QVBVaCvCF#pcPUM
z(J$@Jq@K;UL6>cGM597YCN&a{gA(8U?(y}jBCW<azRP}W$=cX4Jfz+oa1O9Jo}u^)
z^yTv@oqi($A@50<_8~RAa>MqdD}!P=BkX4Y+<R2lela<|23+%V7n%{h%DMzf?Aj%T
zyGmcq>N6>68D^B2Ije%d9K?v&=svI4Ov_HwmeKs4ql~KZw<t4lbJbnyrR2l+99N7`
zTVdbP+PiW=4ujb6gr=*YMTc%F^7&<*(LpbYRILI@9Q0G4*Jn;KP*zMq$E_f<>a8z&
z3PCz{MbH&fF`C>c9fczU0Cy!oht~ZIvLpNgh8*cBu~e6Nw^tV{!zIWYg+%tii>JjR
zKpg{4u}q7F#`o@)%+)8f8pH`@R^R{MXutn|e~#$xy<717KO8^oNO-$j!(q8Uf4Yh6
zFXdjJ|9V0Ge+^GFbg;u4Pg7^bk;tAMbKrRI*)i079pV1tLFSpz>N+;bjrLJ$XS%wN
z!HO;9uR+E(N?JOQK)JK8-JM&$K2_Yz)+e0$zyIs~aB#$rGWQW58QkBs*b;2Y5ZU__
zgbO04qxLRl0vlfw?4EHd2&9VPPA$1Lu1hPH&Zd%Xm$06*-~@^ge46U}(I!>@uAGo(
z$hqL7#d)(Ts^Zo%VLTdR?{!Zpz^v;~vp9W871Krbl{7Q+L{L!h-ShVU+u;A_2ayCp
zLgcH3F~Aln-kRZ;BGgcXW~7#L8M%C_=oDMsUt^=<h*)K2?er&7I1KcZ&tewrS8^R2
z@|#^&iwmT`b~EF6SS;$di>rBjG)9T)EY1}slO;bEJjhH7v-i!|cH=0>44`=ZMBw*F
zVvGNC%>T!G8jk&g#V0vA21`%Ed;%jf?)^;lZTv}342ZcX?-hh(L#2ls5TcoGT9sRI
zKz4}-Hc+K}y!OY5zl<V_n#zBdW}ZKMktz#Jz0ZGVSKG%kh5LHag}%s&lYsOKQYZD=
z0G#Ia^jxtzjBU%-FtKd?`vLo#u~E4o6oqq<gdR-;c<1MnDO>@>4sWhve)ZT0hD8sM
zz^u}vOrN~=J}p0h9nkR7+*X_%s3wjLi82uMfAkPIw*LS4js8D>kf9(#cT=Z=b{(M+
z=msTUlKqp9tLxv+>X5KqZ><>EzOLiF;ENK?{f&?q<WU~}@4gMl^VoxQf{opxK?pVD
zlFGx6H#dXb?$2ImTeJ*CBd=UM^oOEt2DKiHE<4C$ek;4^chFUI&;MSWt9!XT#~}n(
zkUJe44D<5P%Uqr_POROdsTu)U2m^-+c?>3%SyQ0K#M5sC%(isj<n0?JvnjjAwI01q
zfoeqZwU|e6)II!%GT7fR=v|bWjAQJYPckQ8GkOd3^FVS566yN|9+&<?v)dJQO~(_G
zR<{9-oxtL`1<StCC%D3CW%`bo<iETLaK(T&>(B7*gl~o2?z<nXs%DJKb{p<Xy{E7V
z<Dafrz4l8SrqO$RhOg}M`;X3#>Ks_Ezw<HBdrnc}U=z5w;V9`1oSfI*NSBj!ooC<<
zRjuN^p5}{3U++Z}78>p(Zf>=bSOi(-%tdnWu>fiC!p1MXmZY_68IJxkV*PJ+M#H%c
z-q(=poB3Q4Q@?90{n+mie0-DGlHYQkKJmn8b$GkeoftOK)?-&-A3VJH3}mfv0#>WU
zT>R22lxUYxrf$8y4vY{`a{cxXC*`I0RysZ6tl?wiv#L_1gXMttfOcnV@Lrbn2QuJh
z&`+0dfbuJj@%2_;GWodrx$TS7^G6is%{-TT-M+Xg(<tfZ<B^}vKD+GkOu8;NZw*)H
zLRe7kZ`3-!{KLs^>~t;Ah`6nG<;{BJg3X5BYq?|{q;K&Rz569^aNHSWE}I|nGw@GY
zPteZ<;7v0JTW7EMUjVa`Yw_9je?#S0zHAT0heeIEPmJzf-%UwBZ|EF&a(?c*98`7y
zC*BjUrD47GkFH)?%Ah>dpqukHb$a{c<Nx}NgL8j&#;-H4<C<2suAS%&v_KrX6CaVn
zo2zEVmL;W&KNrh`B5;gEIT9ZK_dhS8aN(+7+IVc<S#`_g0YhOT$c7uPZ+w>gJbPkC
z#oi7t4dJ%zUFs|(rb|CaeS3j}Lt=xl$NAr#Lyvdfpf<+WAKSL+4$toFZ2jQyl*+mL
zrl92FHJUEJH<3HF|K?qzf?Bh41D3?CnW&Pbd^XY6;&}F7{hUO#=L2noRQKWE*Ave`
zu&tk+?jpL?0+_s(^rHNVmlb<X#{1~>ak~N<j`&x+lH$Yfc~;U4hNXssNvsuac>+Su
z{!`F<7;T-~2g~QaI6sBHS1>S}U2PcN)QC7>IlcR?)|th}{%2EhS0d0FJB?E0F4*}r
z=OCdOU)%1KjFg&U!DvW|-L=2hoJMB5j|(*xI<NIAeaDnH<};SfKD*a%U-2BaHEU?}
zRY?+>jvuX_oLN16q;2jG@L4+QOrNIQGCIF$C5AuVI|Th~)AeOnJ~(6O6eXp&qr5##
z(<J;FK<M|N8Lh>)*Km~I*9&3!3^kD$j!z+2S@IZ<;x%1m7E*FGcO9QOfKwi=wT5GY
z^=Vjs?_s&$rguk;mXMzk{kuU+-(@QHz#+?<ib1kBzK?XxEz6?r8&dLegide$egB&>
zm$>qri^S3TUu%zprzvo8as58}?CN|4)$xpvHCh%^ok1#Et#Y|o#_)f{0!!s%L&oH6
zTKpW^KbiLb!#TPx`Ag6+JLoz`xWfClN^Oo{Y%i7sN6O{6{vhT(y8HsPCtGTAk&98u
zmTPjpnHIL2LY`$szl;@ry6sNZU>?tuig+nI6{yNXo&x`G`sR9J4&`dq0WkGhs^j(W
zxlAdqs|RfLaEGm;?M(bu6PYkU{#Sb9#s_Tew7@H*)o|G~m7H_UY4Q?l9nNJW1KF$w
z)T`a#g;)6ROF`dH@MSFnSF>4ZRtvi5v2H}e8f$5vR{;-F5Gbx&Rpd*XXj`JJN||mV
zz?SFfoE_*D_+f9}V(o>K=de_XcDG5WO+B?s831j-#n~Ox{^vr0EFIEhn<(VYCPlgs
zCnBIyLdbcJVU+s$R8a-n(=sT&UUK~;9q2XRYAw7s9W~iYt6y8sW%u=(TgnWbm*!%%
z$v2@g-FoP(<OWg=XK;MhbgFAys$2B-+>}8sw9TNkjaI^j_U$}vk|bgM-~04Eaf*Bk
z2uT9Qx$bLKi7}#*Q#|y?Com!Fy-p)IAeGsA0Ut7rc46^Em9RdyeUJ03j6{{?{O%u)
z=+KHp1ILZXR<qZt^AB4A5z|j+n<0Gh;f+K!!!r~zV`^6?9S&Je&$;dYXZ0QW*QlHH
z9|9V@R_0y~fHKMG#RfT(it7oJ8r_n@b({0NWJj`PU#r5WqGZ#o)3R*1w3GDUx8YYG
zfl#xtpBNa#T7<`4$hWl1P`Q?+9qL0<z^NKG0BX(I98&{bC`WpCFr#_lp5YoA%Ay#h
zSES34+=Im)=v2~JWgs*_`B2X!Q9?w}$beP8B|DBXyo}*_Pkg4g`vD^{<xZ<Kt;9uc
z9u{q^d5@sHquV){A4~%sshmO$LAHyt$#ZPQ7=*lmH5lWye#lg0iZSLjqXbskX#rdd
z<Tqfu0iM0W#VjQTOMKRll)f)A#8To8eW84h4XEb;=NT~M@O)L&fP1;Z4<DP^S$Ti^
zj4##3@S$!n7Fjx?-#Zu9f|vy{zqjSWLX^6at6&iIiej;hsYqvo9`8=D9nEcFVFOfO
zR_7@gJw+y=4cX#~ZF2x;>EWh=3bhb5Z>OR1Z>Jaam{W<Q#52{)WNk<b-j{ukSmI}r
zjihgwkEIj@)pWm=0||d+<K&B~PfkWt4>sR?ogO(m+wU<#G%pOCu-k5KmNIPDzY-KA
z&AFP~=!@EtkEY>TA!AwIY99KdCn3-$@QRv<mX?T47>#uW6;bs8Noi(*@K{GkM8s?%
zGxo+7D|L|xIU5OgMh|M(*0`y4-voWE;$qO?#lT-c^Hr*{=RZFu=7;5!^~*BRisXhE
zl?;}nnua;5@~K-=O7x5AP=QSsTIHZy`P-*ETMATbw2|y6NA(_(WAKc^cxFyn(YxtB
z_8-uNj4Bu`keGB}>_d{Df=dZYT`@E3SLxmr^C=<9k1O0C_f^#Vi*;K0d5Ee8WD3cL
zsU>(gwC1dw=2)jwdU$856cpm|>hbQq_cSezYm~=Q6PcL>1rfH&5fNxX!Pw7Ov!aRm
z$Ti6W<BuQO98v|T`U6QVJbHSAq-4AaRmjpn>;^Sk<9c7$`41;qQ8;1PmSAzg_q!w(
zH7XtgYS{Mi__!oI#ffS<sOj3(a?nmTMhxn_^s~{aHxfJdYuAun$pj-o=~`L0e~Vjo
z9~5>p>GjSsi=7<exv+xH61i9F;+1+BT{n6%X4Mf51tUB|<^6A}$^*r1DJd1K)_5w)
zBsSmV!f{_w{p_3I6qp9^@FsHN(zQ|Jar(l-3E`280<=0pO2#5R(XY+84dYtZ<^2n#
zuozKO$xmuLhUR`gYKFBo`p!_vTZaB#0=2%`0|%1DOX^!hG>z1OY68+vcYV(#Gs1vX
zEi)tPgS<(U2dHIegJz@p;HUIdDNjKVg`sde3pU%Rhbjbb&E)ZJ+uNhdZJ77^3l#N$
zgFyKr8sh^^rAfIOD5tL3%-fWz^|Wk6o(0HgWM6zn3@N02y%rOXm90t|n%+;dVO<iz
zTg>l55cydozZiycWkkR$TMeMGMHkNa_XCt3&cji;xFyewZ2O|7^^qj=IU}m))gKM2
z{B3J60wgGQPoni=ojswQ7t2AcSNy;GJj)iHrvrj6Osl3pZJ~m`?4}GNB8|RIH9Sp-
zxwfBtBMlU<j^v!qLM5*fj=B#|b0*ij3u@*1xRq=&cSX&0`)9Q$_17iXebV5@7MBvK
z?-|M!p;DE8!a}GNhe#iIoC3&4kffkD?00?E1@M1tilyv#uXj<qapUUp;XE7eQqP5{
zqE$*+;!h=C*K|JPsuD2K{DRa&hTIA9m`l^_+AUZ2R;ij9ESeMae;GAQr6*jATxbkl
zV5g3DJCLe2+cCnk55BKBPp?NA^nAzcyRX#n3*EhS_z|s9tO)*x^D8m0gI&yv?@!%u
zA7Qo8pT3uiu{0e&O=<;Sm(MzJo)it2eeaQ2r8-;$;;;Kg^0Z*;?ZK+180M&=4`Pq&
z_GzXK4D(eY*1Uq3%2hp=U+rt}Uw^TGHR&)m6-NF;Xm{V&zV^+?h(#9e^CC?!XG|pC
zi2oZf;1w*wW2d4rBhk9&tF5$|yP&~$6e#=;phde6s4~D4mqHV&BU!{S+Z>aZgB<A!
zp~nhj4Ss|W*-X^6GM6)>wU6f!O^W<$Y?1MzM!_F{vQM&@2(z<7&iyt1lngqg6Jmiu
zN-T6u8j_e2qDqh_2n<IHC+Hyb>=Te$v(rr#P4SUbjMVG&^y_zGai5CsAhyh1GroXP
zfXl0&(@rBPw!O;yON9G|e?8QdhQ?K;R$_kV*#E6vyyDNBIgrxNWP_BBO(98)8B~L?
zgqR)?W}dZPXroaSyd}{Bi6v5y0t<Q_nfbLx6+nh})G=z+^T3!3n=gCiqODf=hXCF;
zy4p-z0cNq_vT5FOT9VC1Mgv;smjA&Q_&)rXU~kH!Sm>)s+IYTqV$y}$S7O;eFF{Hs
z8|(LxH{Xc}ZEGzoj;pE<Yo8ENr^AeaW3XaKiF3j&$gy~lnOn%RsAX;cQY)VAHMPY;
z2C{B?{@AXdDAgICTQpg~V_b1FB9Zp7p8jlI*VxTMz~zta?jd2P)8t%}Ys=*)t$gL~
zo26s5Y$hNj1amBGLoevMdXDQ@m#Hy8vDtBHwq1PbTRSO8P_@X!lPV$G`9+l5UHS=M
z4`z{O!sT;&ZjE*S@A)8vYF2uWmP33rU}snHN$J{+r*!taZ5O*X-Q5{+)QnEyz`*60
zrp})tb*Z6R9lqmj0a;3pUY|x9J*s-aq*Ibr*6r>2NPOFB8h0W4{TAM02Ya%KUmx@r
zn7T6j4I^!L4|XEvy{#=*qCMC0SdGEiUA^tw$(2imi<Lc4XW>c}cwMQrQeZkZ1GNm?
zPN}9C`K|=jGZ0tL8D^@~E6E?ZcO^(Nd16M19Mk3}d>mv}ri>g=48TK(4cZm0$%m6h
zr9&4idcWS=YyawCeK8X7Hr}dLL57I~oSa!U&9(+o8T$}`7wI#q)E$k_Y%~fsy<PlW
zpMmukYY#+aWnW31XI<dfjoVAh#;uoY)^S7f-ps;bE!mY#bFEE0(G$mUo;_F&UIP_!
zS?J}LFg<Szrm)=@?VV_2HE8CmZ?_y*sRI~P;{L(@%mHP}pcwRlDW%(;0&h@Jtop8#
z{ulge$}HpDV}*=0SD;88&Yf)4!cxMlF^aBM8tpp}AN{sHxt)Qb``Gc5>-X<<O~nu*
z=E5>p6{JAg)i(B&?TvpdNgh?9SS%k(w2dHa>3miHPI#oh8n2%G;O>3FhvrHNK@whd
z?d%VThVY#k{F2zOFxAjfkHRKZ7bi&I`a!3@*ULe7q3Q1@MzK<a3k1zuGb?WlV4hc6
z7Wip2aCMz>;IfEX*E@hf983)1p#CA2<4?Kt5p@Hs%_xf&prf8)*}(lQ^WdX2otLbC
zFtbiPC*&d&@jma%V5Eb}PE_)*9(hAIW);kv4r9~u<i$bsX9`1jrJtbQc7*}+VYK%m
z<>)=s#0$73S8CD_yCR2J^5$tb48sCTp`k`$=h9_raM^@#s$?>2krfN9b3CD#dte}d
zt#B4O?_R6cLoWVx1%Ot7tLi!EiM>5x+hT>1@h=E+itsSJ$|fWCAziCu6{#Kg?V;`B
zH3ii-^MXj`EK_-qVdFN{M&3vD1iY=WtTiH|o&T-hTy2$#j7kIR>I6T_*<5P$nsAaf
zncLKb>zNRW?&wCp9n4loY{ZgT&stHRId$q<M6hx86sK)`VisP@GIUD5@Hp{N@ymU^
zP)*igMn{c|$x9#u6d5m8v`v@B?~<^sD>T%v4arBptI0xET9sOf!J|os+AS~K48CXe
zlL0QJ>8MgXT^y0|;bgVXjuiKm<e`E_x<ExB<qg6m>Sy!laGlqlRfolC+n>n*3w?9f
z5!srWwACVGr=KM2%c395xn2@mHh%C66n+H*sLx+YGoh}VHu%i&ERnZl)?jnkE#8L-
zv*h&S^U2Mav&V}XjD0k!-Sdv58{#SF+Golqh>XRorf9NWPlcpY?G^2u^hlL%zOWc4
zWtc)F2NCOA_nO5#dXeb2GHYq|roDUoVxWStaTVP+vRKZAuoOU~SsDF?YpwQ8m%lAX
z;v5(@Q`lSuQ_vMZC&pv61wo!Jnrg><Lg)gtS--<ENGb)5%*O%LF;$XHx|Yf<Iq?8i
z_@QV|#(2X9UmY5?OCHL}aXMvujJ~Kz&iE|y_U;jv2p0|cNX2vmF!M7%5C5ZsXk08k
z2tgQ$D?}CRgvAFnSBG7o_ua)vHQhTkTO_e$=!n#agmvIZB;i?uRJ_$BWK*gtQ@^<i
z8xM(1-oRpn0GyGd;S2wDs|wGxbz2-7+s?ghhKvNhBy-d!j5H`KjmqF#{JYRYE0#K`
zDnCvdG+59cWK0e#wCX=(-SV^PC~|hb<XE~@oUihqFH`i%R&_tSzUQ-VDF`j*XJAUo
zG&+wbs(yCm(r}i)|I5Ee{9cxUu3V%1Tm1w4y={?U)~I&{O2)YC$5gDB5^;wRw}@GY
zRxBGEC8K4GkvHpEp_hSIQSq;eeH52iFv4%eqyhZbU;PtwVdV8!!So8oc8AJYr{U3<
z_Wo9QbtR}4g5KfL=em4GBfF$yq7FqfrwZh=6dJqI3h0AGm%>uZt^-qQM-3EL!-A4d
z(6|nNfs5oIHu=KwZm2kGsF~+qp+%=N_GXH5N)0dfrz1RG&5>lU^bR~owS9#`RMM*)
z#==#z!xC-Sr@|KCfCeWAdlgDa9}RHbZf_{w#EOkVeHykRD57bZsurzg8%dM>Sc+2e
zsm^>cKam8u!68=ZZ7e3UtYOho<WUpi?Lb&qGCk=)rkeEvnNrt_<YYJ7RT)aCE$f*|
zIH)H#l7VY;9n`_x2g7Wd{E68KSJaEz9`oxljH))^s+Vv|C}^Ph;3O>k(l8+#k7_zB
zVX_55NEL~UQG7_TWJii{8Di!v&(RuP^i4~O$1W3bqYPbz#@ET8Wg9u|PXJN^sg9eD
zL{^I=hFH-JMZ-2ZO&e04Oin(dqUF>rveXnpaj~&fR}PHrBtEja^MFK?`IBl+qm1~q
z5u>e*dD%JT+ePH^SZ#b(WF_?&eA>D|V}vWOoQ8#JyJg(UB-e0Ks7>4~!|ec|hrm}Q
zz`uPn+znFJT2r&6#vQ7y!JHMo522XIw3c_&8(3mSGK{X#VGYl(ZS8Ja4gcZ9a`P}I
zUzz>GL4~7(?++8P*cH*Xhg`{jICbp|!AtA^aK^wGtL5uou}hf*S+=@QoTz^?`LAn`
znWo(rm#Ew3G6Z(P0h!7Ln)xi}FTmwVFh%E^w#H7~cy+|Er>)9|jBD@xgLrrgFD5VQ
z>~qAQ`_0dxwYfB9M)Ln%!;{W>GL&DT@f_@fXVo=y{%lB>uszGK;$AMM6amVu8cxLn
z&AT$b#IxS34225VKCRwU1fHA&*FN~K+!zsA94M2TH4D=z0rr8y3qKmT_Khq~Gik?!
z^=ldCy@_0n>mXHJs)W4Ht|^AG;aerTF1z(E;Ro9W;+A0gz!s@hmDU!TP}C?kltkA`
zw_qgmLUS>CKIG1opDXYijIm^yPv4f~+vK+r-g<r!{_vvziQwV_>`J*O5|uC{)s`3$
zVo*jyylrT`Q@2f);ohux+~-C$M5)0)dZK>q>xFatnU!^M7-rB*SP-Adh}ljgv(>Sg
zpyo%OgF}Ht(D8I-Co6_o{;|<#xCiq)=q36Z1IitraGK2#TM{K(n{a5eyIBTj<beB1
zCDv<#@Qbr^nyFS;(6e`Ugl#!TRk~y=t1pg(B>eWyH#Q}QPqsJSM+^DKzcbmSj#ilr
z6OJUqWv&oTm+fx?jUts5zKwE={k~gN{jzwL<D>IRy<7#exuyR!MtZcCsaxsRmS8Lj
zRRAg0W6~71Y*&Lys~Xv|?@*8%k{XK<LtfwHaW4+`SQYWx6!wMVF;XlJ&FHBg#pgY5
z#&$_ZIIz%o1V78oEe0##^kF$T6!?d;_};>7_w_^OrbfB0%nFOlDwZs-Nade~%dzj}
zBQIL=6&vgq(Il1Ur=ZacMYXf&B<l5~v#viw-5Cz9$P(p8H~cvdZMY?HMU6mrn0SOj
z5Bu(<-c(<W1yLKy(bV6Y$KL02*l?bQH{LAttE|~}b$c4R6;>`I1i$QN_`e1h%FhHS
zqYi95W*VdKVXpU|h4yqxs?UWgtngVKtj#emQE;v@o0-b_tW@P!Df5JCg~Gr@(${Af
zTwOGsE+)Em9L8dwhAdJqv+kT4cfDV#yGpDJmylHx%{OQ?(gRd(Lxm5UR-0FXe-(;Z
zRFwVn4a+cU-ec^RAkwM%3KF^1vTVY;r=Qq-j8N5OBpbPm8o5jHL+i0J1NiQ>(hH~7
zri_S1<W2N&X_<Nh<P{$-a<*&%Dxse!#8!8gK?a}6#M%9lCM@e9<_)g*p{Kep1*+y(
zhnN3&Xj9%39mjLObj0U%PH!@`l?8Yfgj1SsSVE3pdSAz_qL7B1ABy?}WwlaykO42Y
zA%l3UMw}4RjD4jw|42YnU*dslbb0M86dlEdQkOHx7{9qmwbf+++ZC49HES~p@{^7#
zxbLF1`4ZRs{1ekl)$Mn4-;Ryx%WAl&8s!f^)@NfTU_WdrqMFwG*_??w{!b=XB#Zd=
z^oh)QSXbgiez*7if;by?PjPsTU2>oA7l+xMi{qrh=}P}_@;S?zoq!LX8rm|)b|k?Y
z$)w?HL2rT&e_#ymFS>)z7zr5X3RW)~y8hu@N?zU_NL$`^!LeG2IQ@S(a}8VnaQ;gQ
zckX<CbTK(~^nf$fxkLVH&TPv3mZ>}c@~7QD96`6F+;3uzJ#C9liRHhlh#SuC6wB8r
zahKC(nwGa=o7H%f+)I=v$<TIo|HQRqw7rlKXED1v+-+>lHqPA!NR=nJ+gb0G`mQQ0
zu>N=LrZ7w_fS)ZE_E`-%y~4*}bEa(g1(&Iyt3oc6cRn$-&g0c_ny1Bwcv8;lTxG3%
z3M1E#Rsk|vJvZ;7l}idcf_Ql2`S~6(%sU9}hzx?njZDrJe{#KPW@5>iz1VqaG_aH*
zwXo2b)i<a?I4>qNI%{0Z+Xerat;^jRNU6Z^66i&xXXbQ&yGP(sRtczNZ={sJ#;*oD
zpj9h|hVR+bmgO}9H-8>oqE^=yB7{!k_!)A}cuKfABmTJU{W9o$TaB#@Bt=rN{!-)m
zbirLzSI;5AL`mMpTBbUxs`5QCyqxoBkB^8pb7vo~G?JS|HK{A3@UvD@AQL0!k-(>n
zL~RL!&U<5xOe4dZBH*~UGZ~$7b<|$ZzV;#5r1#@*=fqp3dMg_{k>|Cz+5Vg11lwM}
zZvp|yo<Fe=3pv}-Y{4Fvm-z%GrPGzBN*XSxm0H6JP^fVufSeLRdA4Oc^@=dP<wG3v
zt-U{=y-WMIHrKGy<A+Fo8(9ROpoo}af#o8rPwD`P)Oxl`Z-Ra;lk#P{vUnv68y`Ot
zI1S$6RmMB;s+@lRqLQ=aU4K)SE~QRc!I&igM(tfk&Y9T)!{}%qO~><VxT`5<?oR>K
z6rwH^(*vG{KH9y4M0DaF=XT~U*<ug7c*4$KxV*%c53kG*v#WyF<H`TwOi5A(@u6gH
zpQd1WZQ{WXR$zhQ`o`<k1*}xW;t(RUZaG8`yvY8C14~D}evm*_{=?DO>#u*uar@%O
z>J?A;Kb%d3TvPYtR%#y2dZWFF=i^4`3K(kI@3fqP48>n--1}q&-ac>PfpFD8VOf>p
zR<82-nsB+m?RsVvhdx#%&%Rt)EWGMr*G8jA0imq__7nqaLG!vp3P1>Db%m6?g#6@~
z!AO-sF*uR6gKwt&@dC|Ldrj;$D%BkJ&xnF((f1JGV<FVYTem-(OdbCdgEF5`BpWqj
zt*p(|PEl<&kDNne0atq3Izx`SnMb&M+@kT)wXi>B6s`enb7ez`TJiR?7WI)CHiUng
z%9cWuVgb&$m){h0ql7Wc3M(EWkE)`3IiYfF^y5fB_B$gXbm*UA2O7N}XN2s!C2O*2
z(U8QrJI<($Zbu#3QjWqOpS8b%tyo?eS#%XnnZCPsS#+i+wG^qIk6zdy8t>p9V?=0Z
zZAgvz+gR}f$S;t{W@@}D81(>Q%aRdB7mM4jkgRBb+U#Dx_RMMM8`x!;x&LJd{c+To
zUB9gkcX^$n@RUQ^nFouvIz#Uw{vpH({Lc8?dToXhx})yX^AE?9I+l7d*)*=i!T)&v
zq8g#5?N)_Tx_wGxz&QUQw_3a4BlSW{j!9;K>lY+X$*3YsVJ6u`Q|3{9qJ)?$EWofL
zH8@n=qFvPf;>jk$GvXX+f^_U4&lxT!DfkY>?oB$#1WHR<PmkzB60@4@iib(~OhhJT
z{ub?(jfyAcUF11%o~n@Sw%`JXyY6{?{rv@f7)5B~Zlzo8FD#`BZmD=1v@M$mYO)t(
z1JAjRd8KrEmMT@r$uUZ`H?+r<wx`tsDT6}va(es(`&@Hv=@8S@FHijx6cv~hvc4HG
z5B&(?V32853J0pgzlQ=AQ(8G*Eq{|Q>xXH&+t;gnIysA(rXKXs(Oa~XHJT^U`*|RU
zE6`lC#YRgy7S-oiYtjE@OWu&!P;IffHX}0frVKk{@h-t|tW4@bimRyUZ}6?+V)gwh
zG4`~FWcf|;-Iii^`2*}Z{_SM6y}y?QrC4B$_YWtp6U}mO+&g}6T}F(~k1)TDlS-61
z3VQhuXVR&5W|ZI`PJ@18@Bv!p`hKg|3vD4qOD}w9BZJ2aaRg@TUq`0VNUbU*wom9e
zts)7B1&&JF>~$zXV-4#w%<zd~1nVI}N;y35mfDK*fZ>4-V%-Nrz>|P6GK^}vD3y8r
z1J=F#NhBDb8v{%nHBg(eQdN|92LS4Z(quSPPy>3o!WKyD9-(L+(PasI=&R~-O5N2`
zMj)XRydj{~QiScNOQ1BS0pL|UOzHVJ;r@Rk?k!{DioSkfD$te!E$+}VxE8kpEk3x+
zAcGZ$!C{~jEv2}-Lvdzs7~G1x4Gb>DokG!Kz4L#cJa2N7o7^u?a?giza*~rXD{E)=
z*?X;@WN0*OLR^pB0w_ty%hNzxDEc3?T=kQLBuS_lgny_I3cyTx-At`43*-qRx=2SB
zn&G@jDH-doPzL4ByaA)kI1rqO&UB35K}=iQSmfT!XdScf+Bggxot2+i(VSJ|gKPy_
z1KXc6CHg#57BkR<V_Pocn~FWN!hl<p@pb6(7=*RY^Yl<Vonb%i7^>4o=FOta;&5x7
zKgqQ~^=D{q5`eMyFYe?UojmKHXwka(^eA1An2;Ud6%biu=;aH6zX&0wE94^ZUA)tW
zg_*=p3I#hyMjn2la*W+~Pv2%pPV(vw(XgT@(qI@IP-#QNL@*96DgVyi-_qk{w%7vn
zKzowDmG4{E_L8Y_zwJS@R%MV$kUd2fB&;)tek!~B($^c=W0*l|s(5LX$6RaL8`=Z1
zS76vW>-td3Doi)}vStIWy#_2@dtf&;@>(j9A`B^-vIHLsKuC!Kxq7HihAi3F5=SZm
zY}oK~@mrOFET0SMgGN2_l`qTkwYIE^-U7{3-bS*afi^CAd{3t-*$qlv+fT786kTSX
zZt)TSx{Pw|a?9N20P8$VPHh%Y+(@RShT%s?0JMtwQoK{qKEiU*@zsJf@Ul@)m!)Vl
zXhv4nAa|)e7Ho?yPix{!)9_r49mCUvbCq3Ozq1j<&B9F`Czd4KL+(bB(+W5R5IXb7
z-`WF*2fTG{QBP~dMV$Jzr=p=wsU^coAm-IoA}CZqZu{s~onz8ow{P;8*hj7<J^`1A
z=jx3lBLzkf1cyebI3S)7T{|8n@59k?6FMmMKx$rznKIs0`4-$^wq%$7#g(_?`KxzV
z<XN9ClmB6roy6Qv>Hou82TgtQ%@yec{WJ`)3L)JU6ctcTp8aXiR;pHq-wm~-J83Mj
zwu6xBMU#^#6&uzyg9i5Zza+lOFdh|Q&{kZTykYD_Qz~koBkJR9D$52GpptkOL}?mI
zZBR9ST^8x|LPfO?;W>kZ{Vk`xI!xqG&Db-aM+fq9$)*GKMhqlLmOLeoI7lO!Cbj|6
z-e2-4{7)=hr!4g6<(g|n8A8a6SvTb08qRF!stio7mT15G(-lV-LlmHpyrG=*m`mwE
zt`}5!U3<iUJoloqMqV`vTr@1_^m;4R-E61Z+^~dbL_K@{OGQu#-X8auQ_pqEGlNh%
zOoFOV0=vSVG8KjVtj)x>TqfKsO{4T}xGg7jRB)yI(!2YQYooFC>*kk6nx*P0$_2v<
z=r96#;OO_c*V*v0zMeG2Q^0-`5l6-|BMhyl<`fyjMaFH%XA+ZY?xm3jw(9RLt@vF9
zwTqx!C@&AKk0NmQny8p$eveu$9=OtZNwQEJ)<ip0*d|#i6Z<=Kdh4w(`y~Wb@pj@e
z3+WcYDoF9inZe9Uf|zhfh*M_}ebS3Da#JhU-uWftZ939@B^sKbl~rpdq*chpr15>+
zGx7WOtqlFz-}lPjOJRGBoZ*)PDx*MIojNmR^`TB5TfGV6k_z({c}9`+@Cc$O>z%Iq
zvaZ}LJ`}C?ji=8TucLxKb3rXXRf$s;Ydhc+iK4uN%uL9Rm^T8X!)v+Df=5?I&$LXr
zxva*UOCu-qI0zS<n&=~#9~Z8bORV{rD8&XaG=KPsIs-TSbS&N$33)ZW!{$axF)G*K
zcmp`OE0eQ>#7T@+8<kTr>bA^ej|+YYJA5mpni_fCk$}!20Mj)bS`@#LPPHKWJ#CLt
zlXXBlEO2Oz+Gz>_>A7C{eZ`^kxbYV=5R#htW&Kd&DJI@@?fMU^g<r;h%k~L|aNoFe
zVU^yoFiiYe^B>mEjpmle{|?CAyp!=?jw<rMZFuIt`}=uj`u_or+j-^u4kw51B_~+3
zayb`9wPxr3GvVu`C6^(2n{wI4h}bD1r4tJzAy#!%<$~`cC1LTJZ~dLIIpZbGrA^7y
z%(QK&m~0d_JRRbwprE-ogh=BmG<X*^$Twp+>AF8&8|clp<c1@B@7k{k+4_f7*IUjf
zw1nApkfsPfqbE|EqVn<-CL&OC5sK=NmyDi`KjM~AA@5SFFpIite(^gw>|QBz3y6K<
zM<`+iA*7^V%;{w(x6FNIx$8DquKRs(Ks~gHXRk}drH-U#+23&3(y(c()Uxf*e&5%u
z^6=_|TuaWa;7uoAdG}9WWc0!?vLccI+T#BPu>8NhZj@g1{=+iA-^Iepda35)Ex^m&
zFrPyGgGR*bws%}g?Za10d}^)k2eTUWI?GkWaNlvdWv^QpWetXrX2{hN)Tuauv7CcR
z(un>!djT7l$ng1CX1w~BOQBaKMF@%l^_cT(EKH25mzNqfWJ!zFDMQLGG@ZBLCm0*L
zSz?JM$3FsS`6g1W9kW(?HxGF|UPNxA9~%qb?1<;@hu+yAC>vE3`_KnMwh&Mvj~2u9
z13_UAzyRsWeAwX+H$Jjvzp_6*xa3EPME;K^@=*vU`+QhjB++;;fVisvsnUFh!<}%c
zSs_E+1+aI2u)E#Su5Y8xF6keZo8!xyQU(A15_j8k%pH^VN`Bwo9h6q`eP3Gas*Oub
zy+5)-@+beoyWVCXq#g&Z4L?s?pg)&-q<A@f1NtJG_MARV+ZmZ<M&Ht3JS0^Yq!B{J
zV_8bYC@&j2dritOFG3Zg8)RIpalcajSim+^+dmDw0qw)+R=VWODU`@1d$a&#Rdqk=
zG(e8*z*O+mN%Y3}hyT%Z{NG+ifUT8=Q>RigiLmz^6)Jyh$(KuA@Fj;LP*Ob3vU*Cv
zrFdaa&%N(Pp4MU0$w_FP45;ZZTd>sgeW~ocKn<gTWQMndG}4i(O-%`%n=pf7%o#hh
z08~eMk#4khq7_gCH$%6DjmukERT~!!GM(C8)k*>}%d}9k*#g$uErHjOlAsT0Fa0*B
zSkuz-l!rZtieGI@cY7+Q7p%_Zd%y&#zvUZP*_I2V58Q;?_@aA7!d`qt4r!^qOEz&q
zm)3}BUiMr@PstO<1e~W7SoWkstUQ!*=~C{Aej;H$TaxU336Zr(JYMCJhU!qn@o0iA
zF7rUu%2LepA2w4@*TN&%kzzcJ<259DT&Z_M<@`1-8Ks16iJNylq%+D^RY`;QNpd4Q
zh0|1u7yCR;TNlx_bTEyEgkfX44&$!`VmS<J71_(hr*T-E)V<56H)3DO68G%-XDp0=
zVAp!Kz_?cvOcEa{C?#xL`MLi6NpcQ9K97Cu1-iJiJjbGC_eEa*{YsD|B}H`M@fA?0
zu3L7FC1&yYZ{t^%LIvB~FFy&AXgkRJE51Hc$}#83#M;wbJoWkYOeuObVh2k>B^o1Z
z_6Sv6I{WjH1iNqKSHdYk#{^b(l7tB*`50DoVoMt%<_`0QF)SGTlJD<NLd?T6URbnj
zpZ#{f5+t^Cz&<<|_Q|2|J;MSNasU4Km3>ch!JoVj^Z5Q0%0%%ID&=$WE|!?q@%<zg
zEuXmhI+o#H!j%P9c9sPE;wwwBI)LxuP9D4FM5`i=_TlTBy|@LxuTK+7Bx`xtu-I8N
zjBKS?u@8<7_ZAPNQnnvk5Bf?roYRW?!?6Ez!U}pnpPYX4!P?Kbusx3cKf!2V(6Y}T
z0b8>xjoVkMu-ncr5BC0kAViV-m+oMR$@*f#V35Sbo`gMU{Pjr@PqUOd3aj`($@=W}
zpVZx85u<+GjIUph{Qt>{*_`<K-<kRUa`C?<Ci*{dKPgi1Z*azpi^rXb$YacF?y_T_
zo|hb5w>v+@q~anwXXrl}ln}XF{`KO&2I2W$+$ttBpex%g9w$A_C{#{ce8LQqBaNir
zuP24-p9?U<!7_Jo9f~Ea+%i@22z8ZGxq}51ixR#44E6d-%kJIw=J{U2QRN6$c49N~
zENAd1)*jwp+oPRF4!7Z`a`oCVtW74!+0N-#9wrD`$o{JkBgbcC=0THmA?V$s*}>DW
zwVj`o1M}C&O3us9f(wi^%I~LXm}+lExR*+6AH*v_IM4C#DtBX&M<pFw>v2^9X1>P@
zy{3eIc%|HQviho0i=oxtTYjH`G0uCVe)l3D$cSVBPMZ#D!ML%TEq^@eTbvXBJ`+)J
zUmEOz-|!|m1y_z>;)m8MQ*k&*U8cBZk4imUB^g3bgyl;S8WBMD^~Hat8#!4}&9~q|
z#=7KIAWL^kOmC0XtyC*~7oNjWSc7f;2}7LV5*t?+%Hz6Xy0gc7>G?5@w&~0vk;>zZ
zX%^va2&wBRq|^l2j4zaaq|^p8y@2M>5daLP$pB1REAX08zJSlXFM>BEwB-u!xoH&k
z`E!Z<!XKy}CP=dIY-{ZB*PE_kBlLw|C^<!iHPh2A%EjWsEUe5_Oue$?8NrI2iQe1$
zAicPCF%nS)9n;bNC;}a0vC3kB`M_;nHIl1zAEqenA97(DDS-r3Ju&?3yIu*XD#Fm6
zG6mkMB1-(wt;oMR(stnA(BhPNK;6}&Gs}RvFTsuFVPw4l>Ed+D1q<Ys<z6~$GxO`9
za7N-FqN^$^&}LtVOOtW&PgS*FKYp6D2vChT+qkA2Q=$-cuhuWeFIiE{!N`EHYMobI
z7xuus_aC*h$|n)$Y?O$~a~+zy{#Jz_W{O0yd3}ZTS^V#6kPDIBzJ2+C;|lxL?H8gf
zT=!H0D0gZyEyh9?>%xLlnzB~5FY|1nkze-)V$(`$<0bOrnBVr<iR;C#mF6k8jhQb*
z9<EEVi|nP3G+840;bvsQ8>C?&muNnoC~7qHoCyuJDw;X>)VXNG9~4}ShYLQ}>@3oN
zPwgl#Yo(dmNKa?~u(^_1&_ETvN$S>n^3;qI2+oL_D-dOec?>Brh?Ygr6>3G#fo?zr
z&qO8G9rFdbcyM;(7zD?lJk{(MH-b27jQ7TcEN?kpG>?VR@l#cFoYasAfWxQ&mwYtL
zl7Ly}I#tM}=c@kKg_OGOe^}$#bjEC^^&sb@-BYT9kIi-@?4(xMwI97=f<#P=<fwe8
zd^^EqAm+hdgZ&jPj_GtL<Frj3%dA?(#B#~K$7Ndgg7Bs6C0n-_R#xD57oF7V42C|R
z!(ImYT+O)DQjorQq>F>Z6zYsvU>-<m^Sa894xSG`OHjsvc<E4E!7p%>-FF^ObEM0m
z9AwBlBeY|?5-Ts>Q{bgX;Xv%FzwLtliiSEb6VUYNMUY<TH?q&cxhq2+1C|D24gX<T
zFA`(~G8!2CF-3Y5It-QkYWo<Wg=$4i$A57H=o}?9bG5AJyyfb1K~8U(%1sd!Lwntq
zZ-W*{9dX5Q;PLe8o415q!s_k4H}024NeJnre%DD3QUp!rdUX}U)c&9%!t`&d8ce6j
z^3w*ptgTXymC$q_h>@9gLDG>%a2plg*CnRXIrkud9VV~yv1m*-q`VTBL*fQ+sEw+h
zBOAHb8^zfp2Xjx1dU+1e>_gA_hd0`V*OO9{RIUH5`>i70$YxWn#lzuOi<|o)MOE8s
zeUtWMv1;mmq-_0;LZUsexP5A}n4eCD8yOShnec4Aj6SruDw?(X!i#{pXteD65kVn<
z4pN9?ON*E@KD@N+Rcx>;T>q#TTV&c6XSqs5BagvoCFI1HD1FR{OHav&PYsUyG5R5V
zAeKSK)rf8`3hYqcAYSJfEeiq(+>Z${Ox71Wv<wvyS}qs}Kag8i<Xcr-_PlhT)^V^E
zXp#OK6WZA1Hns1PSn{%AO<s=ZXAv5-*I8;b$~W?YIB$ASIpK2lW22e({L49V6_HtB
zl}YfaM-r1=)DP3<3|fG0kg!4bq;ucPf{`IPsESY_FX_~EG*lIXM}b+loxR&y9~xui
zu3Y~+;3bb*q0HAvszp&aGPXh^51%pWy&%9le}Tx9?~$9<(?}ouBGg7x?z1>UWhCfy
zL65s}7ZlrZVb0s}x9k}m+qD>GLqj)dazzj+_2)d;Ujf;0v+|#U+-%1mo4(a2>tPH0
z8j)50SL*th$z8J=@W5ejdfA|Catn+i@n60+c}OuME^YjLKS>etFeUX5OBhJavQvCl
z@TKn`mRY;0)O!3eJ*RK>Kdi+wjDXGc>SXafe_r-OpUI&NssFEKw+H+5cP+mf{l#y|
z{pW*1=}&o6PFGejwL+90M!kYa#OWbxkMV-1Hs$s?!ha2PbvG%)>sw<W;^~l#0aZ*_
zyEtwX%_sYro~l#W?^`Y!l~gM??CAwKnsW{7BF$#4m4()()^yD<HPmW$z5>2s!6syA
zG^L<Kf9g?Xz14RYwPcA1ftR7qM%<1V#e(Eloup08f|5Fn%Svwe=yzP~N_A8@Jdfz<
ztRy4_+W7ngkQD0Pn{U+mpd4EZ1KBN;$?U~z%dBR5Vg)^-K&2z4QVs&BfNaQPlJho*
z8E2`v#EgZIyxm5Zy@DtqK$2LO1yQs-Uq71t+G6o><$L$ujPw|l0)un0$WpW5do4~#
z-`i!@;(??OH4e{*wN?;1$7~LDq%owXm!ITZZhC<(wy#-K%-7q41q7pzTLtW7G0v9E
zUXC7x8cBgVB{GRoWXgCbT7Ck2r8Y=-=$vaj$pDX!O%_toh1sm02J<6hBg#L+7Vy-=
z;GSAh!us$L=@8tf6wxW)s;TB`r`Yhv0$AA5%{@`H15u6$Q3^CnFMLkLd$H0#km>8J
zcRe@M&B<SD1)7!0QJ-SH@G^@gj&3c`4#bUUf2)(1aC@wlKhjj;S%2&LQgKJCzO-t(
zcqMX=_q_7d1jR&{8X|K;-klJ;Dwq%9Tz0IFKXoEX<JC>hA2fwOA0`-0$qrB|=z~iV
z{ln6WuY<r~iuU2XVY|Pr5uB*Obrp4Sd@Wj6C`nmGi+$WY=ik)Y$qet(H8C^OD#F$7
z1?7IT0}fr?a%xIm>7ZI_8*V38%Cz(XJRweniba7I+=Hfln_>fEZVswKmyVDlVNiDJ
zsa{Zo5c8W_y@q}Cv)m$;W(TU!1DjH?UifOAZ}Yn(nT7uOO6lxLG1kPiUXGx>3;k9F
zg@Pe5X!+x`P$paFQ@|=4XSKO(YG-2vzNCC13m1V{SD>JTWPe#^reekEJ@NFX-7^}q
zO!81OzU+k4!nHw_5m`mou^1L2QwvQ7$x%CYSM5Qvr+U#)(TTN>t{&A?(~Iiw(xN*@
zD%K>1qbShcllBL63q^|At~+DlOAF^n1%nA2vR3TYFmmi(hzo0kLV>o{9lVD2<6AbH
zOm})pq;1}vOW}y6($7?a04c*BQvF8P`$%M~^@rk*FrDsHICz_?U#4h^5mFH)E-p~@
z>t}Gk5qN(_yw-i_o`XP}nH~LQfG^(lWa-|zt>pSxemRyY6%idn$39<`E7Ov>r8)~j
zpLLGa@c1~4dU#z18G>HGsJ)5yb$!bfJfvy;Xm$2`m(*h-b$j`eKn5yBF7r_u<u(@z
zz@oukL*y2b7P589#6XI8|NaTq?&tY~`Jyz<Ku!1>P>IX<OKj$VOTq3n&kNa~=P%%c
z@@)s)Nwd@Jr<N3z9tf=>O)#Wt<)L|(O3k%0%q3Yg27AJCpOCpQm!o34Ck+8I0B42}
zFyy=6PrexzVwmTyeLXYV%tQ&nbSSdUBp&k3Ozt3a`4A%v96{A4<Keg~=e+cF0)c-s
zJ)!>sg8s}Wt)cNc&m(f0r85R2>=2Anzp%rVTJsZ)sq2V%b!&3V`44M}qW{5S@)SL8
zopYbL@%z>R17r{@qkk@FT0a)mu7_q3O18`^Yu`pV+&gO=x2yyzVCs8Z02=NA-6Ntq
zZScNNiL=?FXOs*=?=zPiYj}$z0KGTl3zPr6eM?A8>aFfa(?~!TY*$x+QVaQEI%Fvw
z^8v0lWp{Sg_&AE5V0>-bSIi3UODI}}S0kYZ1P8+`5b~B{MrFXO{hxPs1D_-W(aT|5
zO#XnLn5-BaD@Q!mKhCB)%=+Ru?pgxEf>9|sfkaBCmrgBp=CZn+BrTUiWoyyO4vR#_
zth3HLoQau2fwCO&M)#5<UAr@GLN{qD-*u8)MO9Sa>_t~0#d@ZPL7!0@RztHWfcFL{
znro^}{dH%Tz@27c*oZc-B(g-r)wm@hZ<bz5p^Qb5iL2I5EWti>R*12=Q!y4@F>B{;
z<Txr<yv9nTR?-*SJpn%q{7sPMQz%*!_IP2CdWDSMavlSclQIB1CWD(Pe$3}1`Ek$6
zGe)3QRm6-U<`$;R_%e<Ap!bs+dU=a=P3P7J*!KFTy0*RD6vXf(AEj@IUMoS-&yml5
zEG9PBx<9XBQg8)i^<S1=+VtMwOcXuofyhr?a#n7rxKX8WXiU<|Gv2F7L%e9h(WNf&
zwbW!PA2Q;*YF_{VWVT4rU==X)<}w!<32RjJs{ZaRt#-8sYw4h-bC8%?3ees&)`f<w
zD}eGOi+p*@;CQr{_hfyVa-r8=Fg>^|C4`i2Rus{@Yi#F#Tg+MZ538in^dA<M)8t$^
z0j5{R0&Dw+^?7~uMj_-CZcoQ?aCgon!s)?M%;8XyjW%EK#k}1=tY!Qo%(qa}4O1#D
z%O;GeYo9f_Rav!ky6upqzv1c&(S^A@taxEQugG=vn~%yYU3xC1=6o%Jm5AYz5Qon4
zRY#TV^)Rk}=uAhIN41_innO01@CRt)Jyej;@i*lza9RqtzP&uJ0~#s8(64|{Q8eBD
znQUfsYNFenx~?P_DcjsMX~ksLEi1c-2y@pw!h&;!3iyjtYQ`T%iBx(WLY37}dA;CY
zQ5N!*Y1KIP{6ViC|H+7G%I-Aqm>rp&P7hFRwuT4=GfsW#&2UsKnbOoR_uk2!tn|=}
zaWZgYHs+~8m?lX_DW***eVWEN8tVj~GtnzTEIJ{~{Yhf&;Zj9lHiT>&v>~Xd2%wC9
ztE|cy8L}n#V*jaP92Bhhn$>wF!KP;FtV&#wmvh&;jaP`>ff^6ZOUm%H{q0@NQ_+;6
zq3O5Pm0x#bQJhYmfw(S&HGh8lEGE?6JDW<5IZXB&xMJwSr4AuO7knGNG+dn-^+&}+
z6<Vj_k19n3@MN=sFNE0ttfLTw`c1Z6(?e+{xugtr@-4R0*Y*g*G?LvjNSP*-F!xqn
zU5sVwSv6S-^PZ;y`36Fe$_ii7?ZqBZ5U*A@(RGm)V<JUu{G^Fe7A$Xc3i?v~t(1IP
zS`6Wvt{{H0C!5w=dHr{IfY7Ryq<4p+;ZZ^f`~v;*A)0xkHBWCh8H6}8UK{~KfyA{j
z&uXG@EMw==$6!Kj44vNk<v7bK85XyI#;lZu8Q?yt{kswDVHlbN?892a1HI@Jn{~`=
z=3L~v`q7u9jaW`*GPdl0>%w)7H=PlrCs01@nDe|u%gd>hKBPVqxp?&G`+3sa;wIbo
z2Ki7hgO<E9#X2s&SO$+SY)JiQ#RNO=ry=0vDPAI#p->-Or5!v)#9Hp4WhFNK7V3Vb
zlB)gk(@hisH=z&2-O<d7Qx7>WI&iwm89VcGi{(w-B*Vy{oX=;DPcrth76$CY-PUFY
zzqvvxmU-sL#IJ^JtS-F9>y26|ZMkymD^ggiQk;brD3z#`2z5klr6JPPN{V4GloUfd
z#@m@K8R!5QUe;~q%}Wk<9Zk9mx6B(E>y7&K9S-!kt{CtQq7-e2(Pai3z?bw%(_o3|
zV+@SzL&Ie#rZ^(YPu1pm4Eh*0Rb5%~THNuI(^vZfhh6>;`w2u+m2FZN3g0AGHhAs}
za7d4H{r9ph_UriXQT+F3{^zD4mhYMH=o-t1g*hnCx<Tv&Z!Gi4;3Q&ERvb#gYp4T3
z#m~M0Mw$DnxGEgFua9(o$SBb|_(}M&B6S{aCClzOdF%M#hNyp7wH55H7!U!+r@{HZ
z`V^SG5uKyFmIOKB)Yd|ST#RlvIAla#mv?Z^bknfT-DH!>zfd71vD*`s9X@y{r5%LB
zXjQy#AyQ2($V9oHYy`C3voI$ll``BvZT2~{E<B}%%FnClk4M1fJ2aOnfAwbBn!d*<
z-LS#z5WnjS7f;?ahH=#|7H-thxHVHlATqJrL-~qIFUG=we|eDY^bv<qhk}8=S<8Y?
zZd?jm-3`mlS^vbCF<#JxfdjSDQlXee>*_ue(jVvs-XA!5t1hF<;l4)S9Z*bN;J^Vo
z&BJtj^gI&SI8-cT4Z70OQ+#lEH3vdFDwxbH`C909Iqc^9hg(@qZZ7QqKGc%_!&<X?
z9m9Rc_z!E299ijyffl4y!~eEV`riuuyhlsjZ6{k*I{M~7+AhJ{sH6{p?D-l+2G`!6
z;&cG3gV4OW)b5amtZ~sEa-s;OdRB0FL4yb}uW9q2_Loq=?&q~L!?+*bpKPpL^LcS}
zAdZD$Huji~Uli<;p;II|NGZjx<8Ueo`aQ7#XJ<h!dcZa#PJ1!kJSjpn-(H>IB`x8}
zFU3&~y%bOo3pg$#)vNeVS4tomVAf5qV#!CU^K^ua%*0#do!v{ehSA1jre8~$Y7M_z
zxb~6HubAlGkEpdev6Je9tP2?OjvgDqT_-c4jiLRjpc$w$E^x4DSjNx)I&pWFd~Fr_
z-YsIg>RvNdm!_-Bom1S%BC!Wzc&<pH79vvuWITJzS<9}$K?2{F%-(H3kELhhtBNOC
zA{ReBr!_1Z`^By*M6-!qqLrl?AkJE?Hy^6%EAQxt2mW5681d-O)axVwz*_<qI_hBK
zB-`!(eJdlAv9~3Jd)~Jz;$gJ|Dq?`aaI~IkE^5JX+hiLKbQulI;w&{RX4ssDiV+)D
z*kZ?NIt2+U!b)L6iEvh>9cqWq5hs}w?fjjIqb#-@6l+7tv`M`a&dV>K)mO98Wl3qr
z7WS2@%_80v6s>ioDVc9u*;Fy^2#Qk<Ot26Um0C`YU#KI6SPRjBi4xf09C0<<SXKs{
z6R9H~_?WZXa!u(2OwSP}m?|%m=A)-O^L@5!rlx(c>eBu!%5zi7I+$)-a<Vhg+|Hw|
zFp$JM`|HvAePVBu)6&ecWbE<=vyYw5kykDi_P~9cVw{q!(r3kG=N<M+%GV?#Icd#Q
z6TYt&01MVcyUlgcSznEG%gvh>ROJ#y+;@|dlGx$=V*8A2o)~l_Q{D(_ZF<LajMKJU
zR~u!-m{&F=8}wU6HGI3u7(2Fd2&GjOk3yzuzb13AKp8*nX?e@Im_2V>7&@aO_LM~$
z9RiE0Txif5`=jsO?rL4P%0b~Dc6Chtshv{^u_&-R7iemT498z{{&jy-vJc&^%gE9E
zj-IuzVA6*lG;qWXg>bigG^Dx^_2bZx9{($l5^}zCs9Sgk{9d)q-R~(xk_EBirp7D@
z7ovzVqN>*A6j0FLDM_imbY*iW^<)RN98&0O+Y7Q0?V^lDWl{K`OlFstMt8g*<$^nN
z(Qf?7q68B!?$c>6(0=_C(ZF>iFwNZ_<n=RhrFZfd2cY|&kWf{(K6xwf5uY$JkhT7J
zbm+n{Ao3!jdA>x)xn`-2efC?`&2@&Krmc<yyM#1szqkMS@5ZV93A<#~PB7d*R5PFJ
z+BR>=CI)bDRa2CVjKC;JHWj7k)NzSP7Y10^)al#On1Fr^P7)KXT~l~?xps&WL5z*%
zQ+V&`B6wJM)ho3vnf&)7^~ytNt@k2`$PLZ|Lu?wQ=uJH-9%IbBZECfCN2sP`gb5z=
z$7BWJs!^LsoSIi#L$1rBSm+ql_aSIEamw#@wf<`YsI6%sS53=4C*O2Bi>f}jhA#bd
zo`jv594Ue8L5}YCS=yHuYF@hv6Fgnvj(81qPwyxy(7rNqjWt^OQrZ0M*T%l1Cn@om
zKFBpoqeSl8rfb}n=_!hWH!y`i&|&pXxqn!a9K<~X^&@!N-4VavHt`lf$VUrZbjt4#
zpMr2(Wb+L{UP)Flpqi6?6!O?JNz9NfX6)+)*{?{1kzA9CK#7~qwC&Tyv0U7PS9KMy
zdnvpZSk!9k%B1v*6sM8hG5aKc|L$8HmPH;GdK1ss38bPDTz<)<q*~kyv8xU><0%w)
z;``_+ZS#()i%+7Jq}{OBM6_yaY6&q_CRvJ!PyIneoUHNje5fhBF#H(Z>QDX9uFz;G
zw0+b)NfYlIK|P(bUELb3>9(2Gmh~&%VXWVo>h(W<1uYtL8sx3j&3YIVOTGGCLXz6+
zvGn>VJyK}@{D8g95jE#Be+axH%7LMwVXm|ih+4#r)3vdg1I6&3H?hq>EX6th7ho&i
zd#bE^yzE{9sGus-DKlMaX2f4XAqgEF@q(h#>HUHwc|HgukAGSLaIn;Tq#YQxn<b$*
zYZAO^N1{Iz*}{5;Sq=*Xs;8v|R5FPTs9u#EriAuH5NHXs3L)BP=W47tWp%n05gz_l
zGqZS=+dEI*E#NOkUK+)uE^y^KO;jtqswVnXFkiRdwgW7z#%Xb#-!e=gpk4gXxZUg5
zL8x=yoH+De|L9|7f3do)D#vv8^h)my1Y-*4qS8zKc(xiRlBE>P9!%8JXq6#qSPaxC
zW?om(-UHVacgD~~Xcc{kBFLg-ceebb)|c$lE5<p}GC8=Z_XK$TnM4zQm|m&`oc%HM
zGf*=wCk9tXK)+EVF0P2l+nTpl1JG6xz0x^J|DsY}!HzCGVky&qyk7m!pw%qTd*cxt
zN{zX^526G0@L6|)4nN)GJoWIYqX@bfNuyz-96>t=ygX6~mpKP$hDHKzdQV!Jm$>}H
zSh_u`*EL7FQ%tBg>3PB#8~1+MC|RyrH*!0#(t%&6hZuEY3)=I<x68azNVCV3BR(Vt
z(Lotm3E$T=MLE~gCPoQ^%cp)7kl7Wqik9&1nDbnk4i9#D7Xc*pOF|<<5YtAU(H25Y
zM>EBpy;T*4YGcfv{H1b8`sym<y{Iqq9;3AQ?|TWGFImjW0VHX4yHGxl_Z!`%(51k~
z`IAE~us%l!AILFD!IvJ?LGoTssmts21Wv$Csm3Xecz$p0EMls{M<n66*z`TU_;(YX
z9NFnG?$qp3?bOtynQt?rvW4pG0Et<!*ZmKDw(w~7K8kug#30JbK$wl^1>xwoID{Z8
zm@zr9uJY=|7Y{zB%u;;jHy8r_RQGoM#ScrTE~~Cghu-D{9eG-##8NT?U3VA75SbXc
z)<0K9!+$v95wF+?b0>Z*Pi}XWfh`)BAle!TVrG$Py$TVN<=h2)1?7$)I^LfmeBg^a
zb@|MmcP4Vh`T5ff>J!KmHI&lSuVG$VG6{LJ2%@K7C`D9Yw*CFd0goypD9epGl|`Qs
zcz>LNUMPmKyyYhV4hm9@e4FEkiA<}_{|3v~ykt(llndRG@0OSN4YC8PmbxSer1}1Q
z&rz}=$+kWk<6yeU&Ny_=zSuzf_D>VW2Xt9hTu7Xs<YG~$geN~tsYpXx%t+eqP(FHL
z$sMq<@@$eU9u^g`<;?xMv+-MLSpN9WVue^sZV&n`O>+r~UGzZp8vrZHb#x|5SXfiv
zlW*c}PXy_}4BNa|1t52lt+Et^aCQB|`E7S3vU-iDRS*ud&(|EP)imSEAfS1Qq4;Z-
z%Yg+sPi%ivrbcAjADNMUBxn%3>mG{Q<9%g|DpEy9OdM3gg5@pj_E<`$^R(jg*OMqI
zDeu$nK)cTG`fHEBo=n%K?)iv^S(dz`eRWMnN{=mHEQ1~-*VWPI3=f4@Ymx)Xvd7QV
zbNck!zwDO$ou^1}QgE0_L-6sfX07--z0bKv6)FA0nl^lmL2jSE@wBeJdO)?POI`V4
zs?_TudiN@(54(66ysB-14-{vmdp&Leppk=(5lGHZ=s=fYYT)}1pZXV8dpqvKy)nz7
zB`VAYI|BxF|2&T&wOCQyJ}k>)UWA@vejp@2D?3g2hlPJ){=i?}IgPQj;JBYnrG7Xu
zue@OaNdy&}TsC6NBbF0${&qc@8;$B1>mKa5LHl2fJICmGyZSeet4UpB$^t9trnd_i
z{|dWS%*osT9~QG{0xz0)PQK8Xw2s?g&OAYBfv^nM_bO7LiQDCx=z#1#jGIVY3S<?c
zp)299wN^<hIqF9+6{{a1u=8dIVrEV7D{Me@ScFfIQwB%8@gZfJDIw%w++{}HR?iXQ
zK@u5uLn#gp-uks$#hBT1s>t()zby@dot8Zky0Jp~t8LGG^@0=h6StJiF43SUOsA-5
zyL>SoI_G!<u@o|#*>m18JTH;kh<_ed(35^(v7RfIcy}@hR%$uCIC_;|W*WV#7l=Q!
zhR+D>so?XAlVTA-!f^}O-fInt5NmvrbbQ~&>?(Hox}rjFf?^A*U{={w{q%*3bsz_5
zL3SzgTq%hNU`d8ILQ!?UF3^%Uz^U8qjiRRKeUVc>&s>V;DJdeoqi$W*OMK7R{)af+
z-Lug*I@LP-v#;?)?eSlmJTP~<gslyD%3<=yw32xAa<arcu(HW{aZtC}HUJzYBp{q#
zJhL3Q`l^>j+A}ye>fII6_;q|#7|c8TB#2C@&LFQ7c9%ux@ez?2UU1h_X)~i)jZNj6
zxnKz#JTB#Bp?a!!=ra<?ECKxSUYiLgP5c`|(*U4fJYAGGG&CyurY9|kQhUX6jgJ{o
zHk~=bdO-KsUms3bX$OAWr(>C5$L;a1mPKEpkL|O>l!6lpFZC<##h1_RpMn8-`%T9g
z_3`(}Q8RotcEeTgOaJCAmnE<za)jdK7jtG<kS;>YOnI-Nx+1{EFXMbU;Fp#yaYdJ&
zIkBb-nP87&V(7jEbDVr;2-qjFXCeQSpQXVSQz2?|;W~j3Fj6-BgEO&U(?)UxWyM{z
zX2=1C^^7o<j1P<gwKn4cT2^LORz$-EIx@-~)k#Ob>+iv`=6Yk}26bUCDt2=MRa1s5
zU+@XLt-G(Oq$tFt8fvC|fyC~}xiW@{cspvqlhJJk0`C=lHfGrmmx<xg)D}HM?C>N-
zM+vgLMZ}@-I=lcODR3E3uqMX1^kNqx(Dly7J@n6M1{DoP@Vi764TYV<1YrVU>hZn|
z#ttQ0VUK;8w*d7t`ntUDh39WV_?0`SNGVI4zV9E&O}UPEgDYT>trV^&0Db%VC{kt<
zd<Sdk!*xM$HKA%)|A^bl&qpCy?Kxw~qcIna7$jj?HbL~UA5F*c$i(Vj$OjA#-BDOV
zzo_!Ere}HD|L;TM7o+bvW4S{LH*w*sr>EgEf2CfpUYXnGJREhd-iUXxFbDfDOyOZF
z#wKTI*+DgKa~z6aLiDpP+7!PbbabZZ8iW?dmG2<q+9X`5eCCb8d3@SiR_8_03q+w<
z2K_I(J9uP%@MR&|hy0t1Rh$cx)U#h+_%g(1UNvf5LSf3^<93RPq~)y&#j4(Nq(@)b
zz7U0%bif5a{R}-kGM^(=Din(VP!)nR0s(J%<}c5*38sTy)`t(gu3NUOZ?%bCYv5cu
zmx&1UT>Sp`d-W$XM_iNo2gW&4U=q8p<Uu_Xy90ya!uy6dpbh@Fsz|5@Nnd94Uq`Zo
z5-oR(WQfR5kzx7NVDlFnOX6A>1w5^%Wh0|><=Ai3Is?S9Yl-4a@M&)O-ao776l}M5
zWeW!zrx_+#O2>&m&SfGzeOqH?v~ZGnt)?Rj6&rLnkAocJ1{a7jbnH>T6I!+!$2wgo
z#p&)k90ohG*c<e%^SG9<gVA4_LOr>({g6A?IdpM+DF#iP?2#9Zqe5K2cEU$f)Y&a&
z7H_4BmxRTSY!6@LmIG}5sF@R*Et3MLC_`;v9T!G&L^0A<l>PCso74~RFmWE?x5_<Q
zs8PG^d`;?w;m$PuMgo7ms_*Aqw0r)<lXkrNOlm!Z=4F{n{`<fD5mqyPafFj8H}zdU
zB9ogKiJG_dw(iE;9!|3J#qa0)y5A>ml@jvyVYoThq+-6M5H)gxv+a6#YDnlfimAy8
zf@H;G8%&aNXYVBhQ@)<}kYEl?yJ(K6hQ~4~*=>>69tWNpXTwT^1^ho2fQl@c?TB7R
zAhb?sV&9EXgotC;E%ZuTF$g^^ZU%B%Bn8nPMu=<ZLw!maMn^t=f9Nu&(N*sf@$;(|
z=6@7N$wkUGL3ge@uHH;MqXl_MbR4!IhsG?}I7~k(^o5oO_lv&E|J7r;uvPtMEvhM{
z1%BL<ka~L=ql&apaR8=(Ypbd%L~0zLZ`vps1nRbNx8$W8Q5I}T*k|l3L=+K6{LWb)
zF4)(3gNlAM2`00w{;}IaFgA*ja(CsV6x-&xs}W$gR~E$-{6JI^47To-b2X*&`7LtM
zGdD5w>p{FMAIi*Llt5>7cT8oH+~UU~I8?8Ko5fuR<E*x=Ot%~?Q!uVNB=gfg8R7*0
z0qp!gKOXh#9z;C!-(mo%|8)S9Gyh(G_6Uve%@8^JFJ0BGpWTi&xS=DV!=>vFnxuqa
za{vBI9A=MX`<15JdFM~=Xj}8AoLps}rc&U1%Q_P=pSAZ7S3uuU7F7AI2P|~5@+c+C
z>5_qQEj}vqr10=Ggr~#V>d%P1hS`oyUvsM>#Zt<bKOvWdy46grTBS>#g`Q^iw^5h~
zpd4BiJyWq;LQGsijluc7>|Xe|f|MmQ<#KLGW4hcI9npW%@F@Q~^#7TwO<0h)gInLq
zWT{%1&n>Erjh{EDtPx$-Sg%SpSEg3Gp9<#=Axfjmr{h%&an?dykXhF<_2dq^<}zdm
z_T>%u+TPyA_!F#JUDJ&<#wb!)sS|4R%0`E6^P<ISQj>GR@vsYFydowqCdv0sZc!Xs
zK+pJ;=K80U?JKyVC%Jj&H!(#%j(2tdg`5Z$I%y$>5MT_)@jIsV0@kyY^_*GWtbDT%
zv-WB{FbM^Hnxc8iucHqPzbc+?aPzWo=`JOi8dI3)eGjce^b+7RXqDYU3r(2W6HLX0
zEZDt^pST*PR&6Poi6tNKuDCGH-^ol0;RY6DvF;uy)CxXF!hPJLPTGlLv;!X1YQj&d
z@u6yof9JHu!>|(wnKA^b)38rXg#f9MFnNa3ggT1wk{eo_I(-YW5y>Q_oRaz{D^>H9
zDA#H$>KEg(ZmDs!_!JfLO_1<X%eu%{mVsSrJxiV2#LrHEhk1(vm(`LtdHTnyKg`^9
z;E?ciWc2ryBJG`p#A|J)AC1VKv`Kya6N3S6z`kRK)XPl%JYS-{XF$GYlZu%lXB`T-
z*>eGSB#U}NfE1B5xy`&yt;T`o`UeZSK{Gx)1z|Uma{{udrH8jYASjKBC<_MxT<Wpv
z6kdsZ=rtSVLi17Zo>e|tSkN%ERpr<Afq2S_zUzF1-k&ie9^8aEJ4b#a$1GRy9w?nL
z-=k8s2<A?eBnTF*rew%zS5pXJvRc;&{K%vP@w3@|B2TnnGe!1f(r3$90>^*?y{-)a
zZxy~BSBSG?WU&^vag8%q?v6(6fvocLWmI)b>ncJcesh`{hXubw8@J)=C%CZ1k4Dei
z#E}s*Ao_w{Vw^c=o8b3Wb(5tvi^rxNj1Pp8Oxwf!*SaZ#qRP3Nb3OW>1+3YH>H*Vs
z?z_EC^TJ#U3hISuu5{TKO70*c%Rh0Nk*aboR}+q6C`z|oe`L&BVTb@>u8K=x7%j)l
zDaj0BO1^xMe0mp3;O`eSVqHn%#b}r)PPHJ>E3={eB1m}vq470WaC6Cpn=kJ-*p@^D
zuN=oKxo%Xv#y(G)d6DUOU*xsX_r#F}4%23%D36+-@?1;F3hdi(%V9e2%K(!v;=}m=
zuu@r@T$(x5=iNBxN{Wrp2w$A9|FGIKPytZZZ7>^?eh#fFGtx<p*>*hPPlja(eo>~u
zz-~y*Lr}r}k8^a8fTomjdN*HVv<aa?n0`WBkXC9EgpLL=t2YDeOG^qNU!hmCuSJ{J
z*4pdJs#-z?eH&t2WDTi<_9=q#3Pg9~)#J@A^Lk-qvRCl=FLL0}K3O++BEB|GbaNe{
zeS{mOfmG`o8pW|}3mXbTes+=}YSIkH!-~vX{@2V);A!^7f{KFV`6BSqJ|hbSY_f$3
zJuW@(1QuFETmzaO&^aMi^p~y^N7_mx_pCa+Dx+f6(uMVL;=u**ScdxQ_w&Mi`i-Qv
zjZv_t9xZyYV7exb^t9~i$>VeumkOqoAN@au24teRglVlz)#aiu%aBYMiknkPFj`?@
zsCm$#DGg_V1zB`Aux*?WRrP1ySvMQ}*Fhg6IH9|?2s0YGgY*PRf~h#b^C7$)t^=tt
z>k`CpgQAVmbRy;bkQ(9I%@AZ?vaURuQ&sf)R>I2}msi_0pgBRk`$BF(Nn%UinfFVX
z5{iRjJ7O#jWn-g+6i1c~RmiAF^H26-NdrXh5U;NO>Pvl&pgf_p4eoR8Iie(BSZ<4F
z5Y`#?g_{=|2wonV3Hh9EuXi>z>4rJh%5)7qFpwP3-6pt!mt3P@`W}DGIvzJj|LB2c
zBbTA>dp4;XN9`jzag;W`{gj6x^*mLAFXz4Grla0!3&iR4o{_$HQtDDGQa7C!Q=xc=
z5KvLjfgM;`?k80GD5~ey;C|Fowaa*SRVJ%vGwZ1qCD0}sV|ONC8_rlK8q20lN)m(P
z%GsT{%de%i_d@Ub^(571hAu~4lBQmmJzBfYK&8IOWd2F#Ej974t%_oorqAmtcK9wl
z8EF|-nL~__E3WPhUS`d=vzr|PNgtKvDCf$m?8VNC{)#M6yG^e2`Hg^9p+5d$RYvM$
z<Daet@aU;YWO8B)mC@@Xy+M$Cl$PoVGYew|ezb$;WQ$IZTigU>R9V7sl^x|M9`lv;
z4=K<KMwL*^t{SVyTSU)2BZ#Wyr_GayT}L1|pE@Gyh{nL}Iws!^`xZ?@y+2exBj$w9
z8#yk>1z1fvCQ6wbk1d@sB77a1fwf{1t}#xvRD3}1-7g^pBE>N2bR|EE%ar<%YHad(
zUEEJg3d$&aZHk^=kzm&_G9AA_HbsjJCfiDk+?<Q$zV=w;Ra0Afw;M%hPN}{w-~?t`
z2rS1)%eiWkeDQ*6Z?35dF;X@ryd*SJP-O2HJTV1m61rD}w{rUOLvdy~Kg|DjT*G$7
z(JwSa^z5UTm0Vhmk5ns0r0YLT6c)mESYZbNZnUrioHW+hP`KrDTh;O@7mFCYOO8um
zp|>vh^j-)a1;yCm@MV3-4kXKwPc9SlMW>}_wygD9^xTDiP<)|om%iQ1&}N2U7Y391
zt1F4grA{8U(Y<fu##e>9?7S7NE;3VQBuNT+w7XsUR&G}Jawt4J%mK9M)_K^tNE|9I
zqZ`eZldQlM11EAut9tje66}etU?}MiOt<jkE$-D*da|q37@|`X>eC#Glqnb;_pIuT
z-=bu{K#axA?&Nftn^vFy)T}v4%-<j)otVE)=SJ^px(?eGawUG#{k>-hTd=z{N@=du
zHL3F`#)*KX@D~VH6!fxV^Z$t%#&mW{5+pi09!$&2D{;t*t&dgf>grnL+Um3(_@QL;
znbPw|hMfztcvV>=dihFi>}uRMt+RG5ct^nGO;BEXI4$!p*}fI_{b;34ZuT^7`vhB9
zrhfjdZ+Kh8?gRRK^vyV$cS{{d92@gr5qY<FE(+;VuW#bfjlAlG^~%eA;%uIxn#tE<
zkX0N5>+Cb|kR`%$1oP+4ZTQYZNd*wnS)o_*>0w-RxA|M?q0<%=%{TYW2*kny+-~WH
zZbY?A_hCYrn1qM&uWg0{$1BGL3FIP~;B-VbXCj*Q06|^5T0XT)&>I`5wxLor5<Ig~
zstwoeq+>bCKP*<h1od2Vyh5EO$x!l>PC!XhYx&Q^yks_4nKbMZKFNgeNW72BAS6fX
zHXZ6z?#Uyp{DLC|0(lE4aSJwk`S5vd&zvpO2l!v5D#FQddv%+}gI+*N;C2jLmrK`H
zQ>u*=VX-s@RVOF{osFgP6+Kt<pQg{ff9`8FOdSo;H_sw^!esMw^E;j=$SD^x&2zRo
zHJ%yaY}mN?4PWwKlID4XKE2nBgr>_Pe(lbAlS|=s$HjL4N%C~lSd=eBOB|hUv%9JV
zU%1Oncr$l<yPFE$T)Easy#g-iW&9-lf@|MKWn3`7;Hu84H|2t1CJ*!T^33P!<FJcz
zVbhi&B0jSlaAf7q7b1`E$NQcdiS&wdq3v82Ka+9H*{2H04byEEG%r(4Ry&#{G59)i
z;TV9i5x))8dK{_a7NdI<;+PyI4^(cn7wPBgvksdIZkS`>sff?y!{{n59tz1~DdhlQ
z6po;DTUg0hoISF2^MIa!nDzBWfo1rt=oAkIq!JwMpJ+2q-QPb=iRkH-Bo+K~VH*zp
z5J-E5@BL{=aijq3z19VCOLoyKM}uU|Fn&c^d9V(fnH3-Tfp$~CuSCvL?PD_I?px@^
z>FwA<ilfd8_q~fbXf)_X@NpB~T0O{Bo9O-LLU%{8q?&o51QWCSIH_07uaIM5>w64^
z1<mcmsUhJ96fSI9BCblM*rup0TQ~gltC9<S;rtD*kE1Z|esI_eQo18c*^@TFUbx%i
zb+XAf8e0fXcS*?S6^60i6hEYA0-m04#v55>;c7VrJ0tfL#_Cx79BMh9*2AslsOU4k
z9g`nY;AY1b&%GI6KiW<?X@{Y2rc5&kut!O)&TWA=iVF0!9`*3jQN-lAyY52Ahd+{I
zRG&(0y0SvIPgRgnL_s<Z#vVC1X>@@7KM{P_9BhxFj?@N^v^6UghbmPtLO9+}6dLP=
zgu0#WCO31Z*5%SOQuIAFvuH!(Cj|)y2@iv}bdNyj;qcV)@#63FVWL7K7gs5BX&U*u
z6t(b6`)Sbel5Eb$iuTW{IkL?pX|QJ12EA`xlwiI2$Rflm>dDxtF{Wa&qPV3>%hm3X
zSKiXQ26(m*e<k-ae!)=YQ#_=SKsH4jTzQy@_%*1c%$o(XsAQ1U4^|WskDH*LZ)T__
z9hi2#8K>K;`HZt3b|--oz#&1PMw76}Pv#{+-g5;PAy$gOk3XeamWwP?7Ilnjp4S;n
zI*he${yW;I`>yugpPEycrb&-H_iSbnq%PgvHIRG-eG8k1vyv(mHxX|Umo#YI1^Q*t
zEF>SCF0w#u8!~q*kwu}F0{BHEPI<OeOJjhfE=?PaHJzf(T<vWZf03?pl@AwTEI99{
zMI^o})o!In#Sk0C8W6f0UfB6o2b?k*8U?}if(7y!%|Hf170Dr_lWDJ(i?<7=bO34i
zW3xb**D~DFwLw4@+#T(F2e4Njda$gvegv>&Un%7cuDga=?F%zAfp*b@a_m9CGOB%n
zA@_1n5aGI5TrO9C4il~|exn@|m!kSaj`i<u0RCR7jOQyQ?Q}?m26J^3w^=?_3|Ei^
z%MZn7d^NA^*xb?HvSHciX!~?IU9qQva$faAQF11#8?@r;rar2=Kcu^B5@Ppi@xINc
zCY<I<vWq>XSQrbQx6TKXG1L!i$;q~#35!C;z^AE^&1*O!Eh(jAIRN4^1bA*sEI_*u
zv|A0<WWk#+n%*bS-c|{Z8lPUlcy5$#w4z`eXC|sQVkwi%$>#z2K&1lcbp6YoglP&v
z+b}C&e~v|Z9g!7hnMn28Pw$J`;DJfPmdxVTcjdO%F$9_iac_iln6#8g77OPtLI*f7
zkm1~Fnwio`t~y9~D<1APOP6uaAwhqcRcBF67<AkY_f+h%`*M4xTDnrLbXcy?d73pW
zPCI~LGK+}jDx8oTWHZI|Zg|MRL2{o+S*qj)aK};Ammr{@-^QS_7v2U;?Wf?uUoSat
zq>#@M#K|FR0Vcs^6N31O8nDk^Z!N#8N~1i<Z3aN1bp&+=ci$@%-vr@yK=Nwwfg>0g
zY*ulN_&#rLJHI6}!v}i>-kLFa(cMu?sOSLNfB|7;FaAyRM_aWW&bt^=<DcT|#Eoge
zWCU-T2vQ`Y#L`GVXTgDuvs76^DxvX~ou*&_lrfFaT=EW9e^AD^KE2?lV#|Un{WD#*
ze~#|yrcxvxbrgn76sQq_U2|%^X1cI>!9cUA?;*+YMmF6dUBQQ}Cv;BDtGd!lkftpp
zciKod55->jJ(gCxZ~#P$U7=e~T+H-lq?y>b1nsR??GCwsa>Q|$WZ*lv@OORhxqU*!
z`?yD5{cw=Arqa6tB^x$ZZV)ASzL>Ok4r}A%C$kR*48d~aH{T8a7MMa9I5e+gGFmmq
zf0EeMSo&;DMV(w#<eM5Z0E3!32P%4sWl+XfmT7H4UnN`U>gw&5eccQCJ{N6Hx;yP!
zhp1Pp2i@mE{}*L%8P(?2ZVh)=H`+ph;w=urC1`;bCs+w2K(G=ZKyYn~?v~>2?v@0C
z1cDaY;#M@c6)jp^igbV3&wIvso^!_d#(3W!`IB*vdtLWh>sm6`Tyu)i0zO8$btbFz
z>FM-KvJ;B%aL&57@%r0@6hmYyLj@`l6<Ob$r7&x}|8~9}k9X(4efUR;(cUG{K}Y0<
zVtS7SoWtl|-%+XK>{1sT<XuKLOIJs2Sj;)t+a_q_@U3WyUm{i2%LlwY!RW{#gX&Rp
zie7U-ba_&lMH@$uuwvB#KqCmP?g&WDB*VlZ1Sx7g@HLuv%%xUg?CguD)u+`5V!c%$
z^_@U?$g&<hy7>eox^~p9saam5E%i7GvM#G~LQq}Ova_Y8XkT39WbEUv`rydEqOdR+
z_~@<fJt28jiG}d&O))AvF!lF-Bu6T9OHncK&5tP+h>p3l>71O4Y~`>6X;e}D!xNY$
z)o2!xKbp})Tp}_BU5>jPJ2);q6=qWFZ~0mG-S+ZmFUsP3BX|%^BMJY65#%=46x0kr
z$2Jd8GCTL=IF~`^5G4h8R0Pi<KLyytqSK3%xC>0RMKpkl+pMBB;{NK7J|U4(bCQxl
zohxbdpVgQV+k&xD#WbfhH{Gi^d4AkDfer?RTM;P{>J{{Q;W4eLv$OS(1y2)Zo`XDh
zMQTw!3b(z%e3Mbr&#qUJ17?^&z~uUHbX4g|*(<^|Jx*YAP&2*Xjwzw&GrOGd^JK}@
zZ)`*`;hS|;yB+EGYpQ5>CqW4~$!4P|=?b$tDTw=F*^4k4>IjW&G!Dv%n1aiaakBv8
zb7-p<H=TW3>DP52J1iWttVLkVINwkQ^{cE`C{0aSex&~VrgLUS*dK5XD1@rNGom5}
zq5{zZloDkOeZL8A1;1sAzW6kBdvScBy@IFHOan=4;mv*zV4VugGj1BCE0g6o=29x_
zg$QJ12uWyyIl4{b#2L;XrKTJ`R=rT7jIT>8`gBKBo;u?Wvvmd6M-D>%SEoUwKeLu)
z*(h93Da~I)JC0gEsWP@#IEtdioPvHcTAe=$M|+usA^;|=;<H#zFgl>(UP|AZnj8)u
zMu+O`6UUSYDk~1!-YI_lepHmBP@4(zAf-l}_4rQkTflF9EYeZBEPN6hy?e>cGlB6A
zw{P4YMPm|l%D&a*t7HHRgM6Pmh{JI^ACsl89K`hlOvj%t+l6|W7v`{SF&b&pZg+P7
z*$y1kxu5ow#9;jyfgQ|gtt$x@Ajr<Y`C?-mg4(PIrW#aj7VS@!JrDJ|c~!MS3Z%Ml
z(KShdB^O7n;tCi!^6L@eRuyVBW8c)vsb+g8JlJQal;tvTL(coD^NlNKvONQ*3lKkY
zEQ6G|*UksDR`IAUIzrU?1j3ji@6A`eniq{MEBcza%$uk(nvjt;D1Ucz)>2#6;)|?U
z7N)ykz*_c%Q36@_shr2US4@EZ5c-_YxTa*e$HPZbO;$p}+;N91&L2LzdN9dUVPvI+
zLvo4fj|-;xtA6b5pQ>q&4WDd<<zdd%D!+x&;B1$lz@bP_sSt*sl;nZH*ltcpIn5LR
zI>|{W2lkd|<98ZJgz~h}uoINnJPfrq9-i<jI$5N_9F)BBX!dvPl}aE(qwC8{`W~2J
z;@CIWM><v{r2cD+&nM~P<KI88#W&k>nNYjX!V#P_p|in799g&7n?@2M!()wjiVqMu
zJT*|B;^m7*U$8C<NWxw>BTSKzf7!>QW`oL<Vr<Z{vRWS#xz^X;68+h(Od)MZ@2ECn
zCL$s&DzpCgQ$Qc5fl%{=LdeUc{n|%ZIW`M<!xCBxe*@bm?8T~^d}FfjV$@^zRR$^*
zre3sWc**dT?Mc0Er5>I_xDhJ(wKZxS!HKUh;?zM2R<b`i?d*6M;ZSDRw-M?#MW;m_
zyE81K1}1}Qa`W{2k5#{3K+)m({bIgc@;2VKa8)ll)$``);>aBpM|PR!m<*zPHg+O>
z0w*|5Cv*4U{UL|UuZ%jk`v^k^8ag2)a0vndjdsdRT&}7+nR=5R=`&_P8+b=7oF|mH
z*RXwie5o`IOv>e&l9l^@2EUo7{buHrfj4~PCq*V*j}4JAw^fth4T^JUzh~ixM1V|J
zWL@sTTsY|NAfV!~c;+}y{Yb28vv1Wc2h1Br_&d;DBzeV0nRM|4D1zq1cpu(RT+93)
z3~36A#O!b-xaM|nv?$ldTE1f&D4zEWO9Pw+Y^KFjInpW|RdqDjV|3bgy9q}aLmx-a
z$6QsklX6^l=e)kDtYSCS0$!YHkI;nG<)+!z+39yCy#>P*o?<5_v0S>rhOtv)CPHUg
zR~7Hj(#YurTv4IV6ar&X*=Er@#w`-_L1b>Hn$wEFF6jX^T3A){x<6}c5lLLCxYO-U
z^ffmvAz2>Hd~JxLs-1~D)qawdVhp{KxRz?HYR|D*!c~IW)0DBMG{rq0d9eJELT<7U
zH$GQQ?FofBtjI@^y5^O+U7{q(-?ysoJ(!771;bQJw*Xa92=1B?4st;uKT6H9d>pNA
z4lLJeR(oKA$D)Q!HhJ1PWV*;AA$#4IpNi$XujKgM4Y^J#`<hJ;!iI9`;PBaj9s8h+
z`R(TwoM9*;dPLPHQfcI94%8RSHE)|W$J71pl_Z&&u6-Z~@r#%%qS-R`LcN*jW1T6H
zSOATE`e~3v=S?jbXfSaXGohu^>BFHdBtB+g__49WS#I;8wIw3b6)%+NPI(8t(zpBc
zcEkss8c%>2O(U&+EtSpAv`=)RZd!VFqK&kxt^47mn5t*ymQM{EM#V^KP$DbV`a2_Y
zj)g@YY4I;i1*6B87?z_wYk0rzV1=O{TNlJbpvX98LdDP*I|>VLb(zViKd18i6)_zW
zcI;6d^{$1z19_+tMS1S?BlmKlkuO;{)Cso7xCECF<J-W+4~8vDCC(xGh*MC&*S(xw
zpvoEa>su|_v6B!;x2U@iZ$wz3=I{s?-=lRlT#^j<JXI|XttZPrF@`3q^`s{_PRyRG
zs2;Y`Z&|o%(1K!Q;W5HGPSGE9>jfAZYbnFaMM)nZu>OA%d3!p}HXhiq$)T+?iZVG<
z_Xl=+1tCVynQI6_v~Xab+lU$Rra240ss`W^qj)J~kX!Jw(Jb|H$g1S57|1|TuK{C*
zMUPpuoe7@lmH0v7<esV+YuQi=#G|4yWPB+n@6p5?J=qdDAn&Fq+n}CCF<rZ+m!HLy
ziV=SbEh@tMCWhC6(pr7YD+BFo!<b1ZMLd&9D76%dPrS?PPW26{6j^D+zl%bp;*aH}
z5cg;Nm1imRyWfiAnr}WCLr6PPXDi4$mO$${Yd7=#{iw`Sp#^PD@qJq!3Vn#c8WT=)
z#U{gP;pnMzkTpmV=U*CI>hShwN%R!+;7U5x%&UU_(XrfgU)tN8Qc5PiAJ0DtO#RSP
zFB*GdI}L<YO_^J+j=Cfvyh5n!`XaSS?bZd#TBGhGIftG1E3?%(@ZZ}ByEhYp&d$!>
zt=Q~6#X0_PKLwuepQL5X0+$7#5kZ>nMmMYbChocezC#XNAf5Xe?lgqf{Q?VVk65-I
z$)d-czn*X5<s;=|?0-*EH>;&%R-04F&B&gTe*J4UfD&4rLeuGXNVXyi7$k58EZ7Ed
z!bRy5{Pf;>v>50&f8%oyvr|vtjZQ26g8uE)ADmlLTBzltu!hqhxD?Yj-;}iq?w6ON
zQd^JC7SgZG+#lFelti8j%vTefPp!JTTwcu~zcgnyTWjk0Wa;NRqFOA1U7D!QhDf+V
zSMgkw^(AN*$>I{g^+RVxn{6Xpir?)~^xL6==!|GS+Q@AC<CAwt<zjQ~oa7WIdZ)D9
zT#awhNq}?tq0kB_I|uY!Z)Y?xe#0x651Nr)@}(3jiGbZPy^r8*SJgNuR51dyk;_9|
z?QB$9_wf%CoJHsS!kg}PetKSQWouTxw|FLg?ptg*Rh!!SGNjfLrSD6xZ*S-q5+O2J
zK>h8!SY+$Dqf>?iXKG-;uJx{z${G4=dJ8Q?BE<0PsM{n@NZ4{;?<+3A!NziD3CJ*D
z{jd}O7W#PN)!)85+k=O7`m|2={bJM|t6PlQCR!E!3r5;{qASbk(3rSGVJb>ZSqO}W
z7bWOtPC-<gF|e5`cs<ft`}6VMhpAM4odGVF59`=S+UN4xrle~zo?|(x3WPc#r38St
z5AZWz=7$tdbOSq(YI-x8V3fqA`+P^oH)@W5CO2!hw%A|(y+b<g4gCx3_Y48pW%Lc;
z9=E_Jljfh#iYk(9*MIO3#rPRiZM>puVC!CQYOD&C9dp()#d`D)Z>W8+qPihE{v}XB
znP*?q{W~m(*J?fGF|7y702!zQj0hstQn`8?>K5%W&-H}=qV4I|CGY4@hV4KEBpa$a
zG6F0UoiY7Jrj_Or!B))y)FGCkjeqic^TWZVCFd+2%8s$KFlQIPbE2{5^!?yH`NF0f
zc0yf>QM)+x+kR|FVj4%7&2}|hsdSvz`~Hx?U>8_(FM@XBO}iJ(2Q2c1W5RZop=;O{
zkM`sAhmT$sekm9*G@;87R;XDS>iTQV>s>@6KLvd+6_q|(AeM+ELAF!-gcoCSvowY0
zf7&hzNox&2`c!;X*S%mp_xr0}Hll2@-noI(48Ro)N|_$>hpjduZX~63>T1VXJuQ=p
zjykt`{eW8@3ej8BsjjpGFFWc$E08jW5eN7AOPqH3`px8B!kUEB(>FTCIrycD21n<-
zT>Bk*wH|tR{PdB?)**`=kniU~k*pjQ*Oi=NogW)^p(+zOBZ$}b`thTkq%%7M0o(ba
zk)c~MnNh!)8e;+^>v>c<cltwjqalBi5JA(g+5-D9pu^nZA~uJTZO8yIyHg0Neb`&V
zeO?|7Yb~q;DvjkRgMit>KqIu#+EewU4l$tgBPO+S?Qs*TTKDa-1lF3lKiHef|3LM~
zz#`YQavORskvU)A=DNJUAIz^v2_w%o%!rbwyR%pe9$77iGI6Z{fwY&X<^eCQ_3C~F
z)#EGwjVt~}$8?8x9xc4XOO+Pk=CY&Z<K>grLB44><F;j;0KBF9kAA5HSFT?wH?oQ^
zMf<2`w=x_W=C_3FHbU2K@=oTW>$KR+GdgjNCh6`Y<pv%i=<&ySxyvwa4qug|U{N6D
z!<No+1xRMi(yZ7Wcx8ERg=^gbH+NLD2wt2@E!id-^kJBzq(IdRp>z)VNJxZn%>4^O
z3jbUnPzl|L@a}|q`m<ky`s)!jrniU?QZUq-5}6G1^JXfe1{i`)5;3!NOwkkiCF6pT
zb@Z`pkR*rV9+ak5!pJ~BnAzRA=(EMp=TH}>@z_uip%go>TGNhk<2F!{6k`m3NHNd)
z8dR;xuoxJ>u`<b3sHPEJRxewguMBMl)KKZ8)=w!V#<S08I!~Px$sU4ql}o$cQ`Cy`
ztSb3iOoA<JAtQ4_XLwjm3h|U;v|+YhTqk*O+maz^AeL&iBKnhsS3`{$0kP_l=R(*5
z`f?Jzziccbv%shp?JO}F8nBqJ%Ab{t-!vtA6V-xby!MMsoGG`0p4xYVv8+vk!TdG6
z2gqZw>b0U_9P_77E{g)iyyK6)%Rt2%^lJ*aGNZ6Lao$#ptA|2N#YS+gdN7x+eyvcN
zjk>B0mr-kGO!!8PMPA`KwD3puR}0%~qNB~1=U-M?No$f#;e|s_;4H&7QnD=(6G&SG
zOz<-6<m9tK+1I-5N6R{H8H)+55?k7J%bBJDoe8|Zvo{?N>N*$=Ms?%7;>1fyPg$**
ztN=n>F`)M!(Z_BMhXCBG(mXy=Bp33|Dvq=hD#`!7!*f)q0=eeP^Vs1m@O`+WgvMJJ
zUE5?#G$9J{^i>SSVWI_pLGizkB!0XLR$b<?2fAA5kB!RHZYQ#@POtrL38RGJ^^dKF
zI9o^)B;yp4C<eBNi_FQQwDQEX`3xLJCk!PEv{EQJBShoJht;gQ(8brKW2uzEMZ_yS
zOzKC+qkwo3Nx7~I*?S;~`vYpIuN~4K3RJ`7=}Z#XaTc6&oLM3l<WaQMIaF5l9HXf9
zUAm{}HvG%NN%ogcE!%`lXU!4D)9`7b+@xTMO&#AN$X;j-Bz*lJ-vJCeHpWxbwcW4t
zTB6HIrz7>~y&cHqODF$Op9@5tmR}!0t&g1d22k7RMz9X0unP1f-O5g3i4bG3Jr%6a
z^Cr=XygdBtGf4%1bhRXT*J0goI469if+x(K(}e!5LJli)HfWtyxAd2>dyaYN=3tuF
z*jFbbW$Qv<_^CC-+n^E{WbYdyhBYj0AlFoNRw$BVpRPo!I~LZ15dwF^H;OSukgbO%
zuzx@$RRe_SV-cg2Wt*8P<iXQN#^&S5xhd+5DE826HHc}k3L1?XZULrQ#Lo0di0xF=
zt;Qrm>qkl$wN}-5ip3wUm?^&uV;L7HGy)JpwOb=7j>$`h2Q};E3h=`kE|KwIj`Ua$
zpc%0Rsdr}3Bq`YmN!idMIwUbcjH4=|jSoy{VNo{CQm`qU#Z*msfL42M7d`FP3A}GT
zx-i5jMVqUNuDtJ^DywzEo5@+H5;h^$4a`e{G0+5Otq{jvao+NSj~i$Y{Jn7>cI#5k
zXSx-fUz2+{SQyP~CK+WMx4e)o5Eh$euH7(b;gyDA!Jta;4PBz4ja~ca4fmhTRZxq?
zExfa*!%h)P6~FIHWBJyX3MFqtSy~IC4NTz2aY`1kJaz}RAEwzWa}J2^ohW5T*ds)7
zftpIvMTkmSJ=UZcoJTuiYJl!FSYW`aH$S%HF3=Ky)<96_${LYt$R?`()MkqxNJNpq
zr@GE|DU;VPVi*|9ooz!Xh9{cYF_zFN6^m7)YsrIx8ingl)BD|?JZCmG7H_avq>zdL
zv{Ey(T?nv7s+1eSup=631F|n=T&>>Ot=~Z_yU5NOItt^Zdk&J?8!i7_$<`s$A)W|x
zvvj2N@5-ou-*%$YQAp6vM2VRnnwt9+NXnIrJuU>sK{---rvN6tE!t-hs9}P`kq{A4
z8uRf1>%+M~a;c2kJz@7OVoVIx`SmY*Bhh}aO{bz5lk86idG`+;X=cATu2?rZh8nH^
zo|ny@eGYJH5SW9OyR!q8Y#gYIJ3m~aI|)sqD}4ds+wPO;?>f2g{yw8av>~x|A%7ic
zzw1rjl<r|z_UC6TASFYv_qKNen;5k<d)-(sh**+Ik|RUBIW-epCcH;JeBHwb*6_-k
zJH`l8Pp#%HmZA8{>4}W-bY)^nlvsTBi8?<udk4XY!;H9XW}oeDX?k;zb|}Rxx_|G$
zVYhHLnP>z=<7D>SCxD5DpO&#sa-S?U>!n$y4M}P=eUe>uM)#Tq4xwswJ{GzV-%nHa
z__SOrD8%K-Xl0aFRI!Y+s1atI{^>oqlgaIjr+<#Gzb3Gc@E)gVc<3XnCRY_Km<LJ{
z$yUW^)Ri44_N-F`TD?2kMrcXb1EbUe8I#g%p=b_0e;-3@`4LkYamqV|_}EmW{K)7w
zv0I?lwakS*72cH$vFxS!x2#f1tA8PDL!CXDo``B~4)86}Dj6*Wq>h<!10eMLDD5y>
z-_6y-Wsvv1E`h4}Uk(L=tXf-kS}?`YiUnODohjoAGT=54YC+`DtTD)*!Ux+^BIxc)
z?}s@SOLd3FPFE=}O=Ie=<-#0kd)=5MF3qAH>IYhPC=ybs&FME|94EqCCKxf<<g7MV
z?NXhehBu5z+^BWId$Y>)X>NUmGT;VLK$$VNlbaV0s?yXeHFQSW@VPRKN07RxT=a4$
zuk;~`xkxXK_WroAwrVA__hfZY)6;_VpA1nJx1(Pc4KACI&_5oE(oak<*5_V^BG_Lz
zno8CHv+lEFYM9D%#zz|!b;mD?9)o$Q+;)qLWwSf-Wf9ER_EjHag6V{6t(r|@JpR6A
z<j9}IHH*-bYuafOOG=qH9F#C?H9JMqdoxZ1G~13EmyPm*TMw6@UHOH1hNZMHFT=X1
z_j)~&d`A>BIwBn(=X9|fzkU5A?IR8J7#+V{^5kL%t@_N&8JkE25nvFl5?t#jS{znY
zW7XFk$4@EcN}b$i$5GTlu(4Cou@hz7_q2{ye@tS9tYixT^i3yk^vmFC*qh1NuLe&l
zGKnw}>&7-`d#gZKGoC3?aQyK*@`I0}JTeS<)gHfu0W>fgzaffsQExdtDxRtG3crb3
z{sxc%bsD68^T1MTpr|AR;FI4+G^_nIYwrE0L%{A?!L_u=dF^6Ijc!V@L>a7e+}PLg
zp;kwcu2lr1)2A`$A%$4@xq=6Ug$pMz&GzLuh1!=l_7@Hm5Y)PdZ`vOYxHk;&2y%JG
z_xv1`9?UzF{~$@jVLx&18wu1*>&y7roqO~R9wsbatrbncS&BN)WuJnnWBTDkA^i5$
zNmS-|KCSG2m2$&olun-kz&G2cL<4~bHq3G?#*RG`q^$Zby{dVkTx!t|`!j(k7fB(m
zM$4=!6;e{RIWb#~IWoKeN*^q!Cz_PL)YLcz$Inl1XU6(=tHKd6J{>c&N?kS8x{Pra
z@3|-gIT_HJUa3(GBlK$-Q8VSK91K*w6|)iLc{HM@wRx_(96$0zD!)h1W5%%?rqwQ{
z!wY0$zth|&DZEIMHPElOPBZUOGN^biSWB(F-TE{ZKlW<i<4Oz*_N@6@0i`f_$%9CA
zXx1)Y=?jBrO0ryIh&U2BzgV`o6(nBTx_uAoet7^@#X1qsueYcm77K=-tGwJx`3Aki
zQzKZz1ohws8+HLJG`q(8lf--8oC@C#*h+Hd07SZ9COUEcXX#jLnbA)rH`bYMX&IMt
zGgmWu?2C?&m#sfF&L+4@cO*;pJ(x>*M%kxIT3$IrBZ;<(%MC2O`~BV?uU07}&&^t6
z1G3PI1lX&J&9kz|GPjSd!gnB|o&b-3Fr?8UgYtBj_?ahK+Pa#e@hiAiYT0g$N>6iv
zZcd;^<i<Oq+|VcgdoNQe2F8~S^F*O(r4@gUE(uN6OBA2<cd6r;Ax1!3s_+FvRdvkL
zNue;b9^u^Hsr`Fp?>y81*T6dGGTafiU#C^GpH_*C+Cltia3-oArcpPj<g3G}N*l{`
zwZIl{^PE#7f|aA(S%01lp3xNg3A|<&jou=XEmrBLgP*i_P;uHFT$ixZ)bS?6@DRid
zt3{dYW&#yq@~eh!Xm>fh>GB*7^fyw2nPl~xMqJu-(OG29$U2sdNf3b~(sM~9%(52R
zM$U-@s2e%5V;1V`-ze<W?Ks{eDhV8Y?O+<BgxQ&ucapW0J4tsPdrLW~L}z-_Y!{y}
z&`3YIgR?;mxOZD3<QyEAOD6cgKQ#7xas>e2U=DR9I`G=oiopfTu)8rx8bS!WMdh!a
zA;W;K*D4{R^6b-s)<G2-F@6&Rmk?A&UTg8qGj03JJguCx(ywSiB@cn}g6?$xW+e?b
zKSMvWUVTV^`K1#ob<Xn#lknNM6}pkg-y-^DwtwvvFNd%Wq3c_yk0B!`wawh~R5X!E
z-m{QT;m$ZiCFz0aCeS<G=V{cVlEIc~SIDa;`ZviAFh$hZ;hr8t3!|av-DG<RDKnM@
zM^UH+F2Zs0Xc6pEWd{W{zj>O)k1;B#s`?PkY+a)2@_@yBk(~^YDx`4m_r5Oc{BJ55
zd#zi?<c1y!<bvk9UukJ+T`L;!al&{eCp9*wqs!2XSmWHL!15)-p|qv-<$bdPzb}`0
zUV&>@A=-`pe%%3A!adiHo!3TP%O|X&s}H@Ne%;mCeWP)P`}-D2Ugm0MD%o!$!Bbi8
z;lO6%MBL!#63el+srvKGk8v$mc_$uE?a;GlizheEx_!;pv*0nqp*ThMb6-Q!ydO_9
zcFxB2xmq3${C!J!VuBF-#Io*uqV?3<PXd3Mx@d%;obrD;OX0dofpjrmpNne0#OG++
zmQ(KOj0_G3Jo?E1g6Pcd$u61kFv_yg*z~x7EwtkJx+I+Q(|-M)ZR<U$vSlod+ai2&
zeAVsToqVk<$0(>2BdBh2uTUGDPZT{I7#`;IN(Uun|B$z;0DdH>_~0;}_AMH<j7n)0
zYUw3kO9h1_vbpQY2Q`zUleE))%To>tko<CYRa9C}(`;*hwT(O1OUKmyzUknXWQFRx
zl@bMM01dK>ZB=gs(Fqf}6*;__AG#=GED_qaR76hFzPdZ*_1*g7Yqs^@UB5deERmlC
zV?-$oG-$P%9+OV3*;kj?@QS7Q=!$vd><_SVYDOkU<u}Ft#?nv9Gv}erq7r^J>x~sp
z#{Jmx0H3_`O*G_O)%h<p$*N~JA;$4pmpXgdb~6UxRCA8a$TD!f{=Lj{8j<8NrQh(x
zZbj~l9X@@oEF~zjChla~@E}?%%{1g?V&^!-ZC3i5+Pg%ndUKv{E`s$pR(lz{T{>sR
zJ=50r!G~UFBEjJz9|iQN0|QfpH6jG5xsX+$bF~*~**Q*X91bxtv&|f4m601rlkgvb
zXVLC!{y`df=s|=0adRH9*9WCwkzb*$XntYQS^hjqYP>6pDJC3^C|r``f!3!uPj)(e
zT<yXK$Kz^9SCOAX49V4HrtrVG=gKA55j(Bej>t|@LDE*LX^PS@t%`oL?ll><Egc~-
zCaW^m^PrmHYI(BRDAm&(Hp7zyX!~QDPo9SDf~6YEGqQE`BWS{NDI=$yQUoAT&qkxF
zO5rwf+;KD&=qXk;j-0%xOgw3o`Kx@a&cT-xubD)X;hWeKBSUBY&C8)_%l2cm&7UYN
z62vh&+l#3(Y}it@*m2sJ)e87gB&@sYG~WcpPGx=mX(~Q{CAkviu7A$^XfrYC*)}l!
zL?%SAqnttUX=hC7SZK^cX-9ny;$fA=um|hlr7<D~-I(b~r`1BHU-M+6mM|@W>Z3-B
zf<J6odcT*F{-(jlJ|!vaTyYwXs&K&oepSZ0Cp&qDnVZG2QRbfHWh9=z70nZw&>;V7
zEwE*|!+T-Q#I)Q7Cpfv|6tH)H?pOFPp23_qe-y0iCb@}f)H}=lg)s_b(`nezg7j$v
zK4ki25%nNnWmy%;Pya0SnA`~>>7FzCN}LwK*K%&eUhETw&eTI~Db$3N+U)>m4bara
zNKn~rCi~-ZKyRNEE0&LHC^n*ULx4t2NaKgId_XaagzonWE!N<Ffl1_b#qHXvxlw=o
zG)dQ_f>qDl<}lFrP@Od0>+P#lA|gpYcm(L^UT@fLtmJ=wr}t^OGQ(l&*%Zo&w)~xu
zvkitz7qh_X74j)lMxIOPi%mvigNf6EbDeBk#?`A$sNT%Jj&zZ#811hL9saNqx06oU
z!L?|BQ|H*^!KqmNS(`3-=e)G0&}DGREH?J{Xmuic5u~4R=5xD?%02Z-e5Yvfa&KpL
zglt!87{h4#uy;-zAToiuTgH2gRNkc%*JL1}m7Y1Y|B=W**Zf8Yu-y<cKmi*syaAK3
z3*gO8-x%r^{w3NNC-ZHYlV;l*XfUmg(02o>zQ~UseEGE%YMrNubm^Tl=WC$IMy>?k
z`;!{^65mDU%&rl~1o)_}$fpoB3rz?5O#BPDa(0YyN(56w7|3Y=pRY{tw^P!wfy*qG
z^f0>Twi}n0hZR`;(wz4C0m9en^Q^w~w|tk7k`3{a9B9Q(n1sMh2E33$MmU5P&&+h|
zv0`ckS5K3Se+!@R3Z#|0eqDCH+<>1(BNC1`EI#qNOi?xL!_A15Q4`j}sZw>(ozxU0
zpo!!0**mI7vvlLE&quv-EaZJnA3*ZQUr%z|vi0`lI>DTx&mF13H1ERQm}9eyeX(&=
zwDhs9^Qhd7uM?Z;MdV}rhqf7f)MY}OKU2e~MomaeSp`^#WN70b?8Oq^6X?u2_u%=l
z3(dC~Y5twoln;9PRcj6G$o4b4@)wA#Mm)?M0dsIE!3CD+zDF8bbtJb28NwyDbe>&s
ztv@XBrMecX9!jm#D!Sp)aqw+X&FF4K=2^yQrnP(h>ZPXQF|PEg<3pvOs5o#e6$wx8
zGq(7>I`IJz?}4y%LL!-;%kTjy)8rX-Y9FH1LNEGA^nnW$L^cxO?^^&FeQ9*=?kB|9
zS@-qbuhF)C!c(0OYK+Ksos>K%=mJ=aG{V=HoX5+{W3*A)3HQyW$(dJOIMJ^P6|P8>
zV|#3o1`fX*N<KZlqt*-_6N=-)P$9=Io-(dD@!UtVn&WCPk5IG`&1N#axWNd9cy{f2
z2!{e0dzX`xyzz9Lin{P;w>`*d|2;a~g<x6$G>t>w|Gqq2lB|rYpvU+h{Ah~;jx0*S
z8nK(hq7ofDViX?d@>uDLt_(c^CNn8nzjcmCTlg^FT*^0TWw_2h)Qqg(<}|cYtzGYq
zkU1HTm~pr6Pwae0>Rwv>e3p8p4R8PBAisTjN4@bOllnL?Fo%i$`=D@Sk-_roORr8o
z`o_=*4kMxkt=vw09??H_WAYA%EVGoVBh^o9+J3?g)^V1u08}x7cIpsw&(J{+_ezRm
zM5i~gNtsbX)3I<61~{x?zm5bHL7XD%n{YzR2=G~3^>%w3ZQ7uk?VFS|M{`H`-AM9u
zZOlr+mvVGE_-PKyFQUC-X!!o84$9{KSF16Scs`Pgy=Z&w_KOyp<#dCGSz%%BEFK}0
zE}Rk{f0cp(o`z@I3XjdsI{bYGiiw--#b_gb{UMd+GgyXX?a8eFaeDlM92cMDrafjY
zv{r<8@0cTnNNF<KF2t~7xOPKpj1)tFk#si$ZX*Us18nX8pj00`)C8tr@wn3Agg;ZS
zbY^rV=lZ$5z3Ze`9@T=ucq;0+J)Oq~q3zyCAUkGHa)nn2J<bWB{4md;<7Xx71EuM2
zcbxn*;j3(%`7gLH%}{H4dv~fXK+OUHzis8f!};f7eo<-}3vC}I5&SYd*G}<vb?zR;
zGk>u-uz$S6H!#VAS{_}8kWe<c)VIez=j!Ja>?EYt)gxkPa4*rk6Dmn!BmT<CiB8|3
z(CivoY&z;iqA`kK1Y{qqCsNZeGZ3oTNr*~+fBoQ|wsAS71*eS)B<$Xyegh7GkGul>
zRK-^!Z*1Oz=?mUiq=um0$>TElb2sx%yLYvXD;4>0SV*CPPVZogEpB}55%>*pJcc><
zknoy@Yi366<#JzhxRu<YSRp-;2d*%rCxcjEm3?;KXPqPm6|&ewrKc&wQ+^0Yb{X?g
zSj2eXnJd+>72ZhYS+OGG2+qAL%-NZlm6~c)0+LH1ichogC(uEtAgG-}K*^qC7Wc}K
zgF6RBbpv$`RR!MiDB%0_lbCJo34Xe3fM1WJDV}j(RaV8Wq%V$Q_|bggG1XE)sg9#V
z8eMXMC8G!{b6B+d5P3UQX&&1HTKJ$`82hmD$LT}P<tct@uim6r-<u9*Xpi<XjR!w&
zMiZHcj8#ZLlN`_}*IO*f5G~25>aCC?_k3E{qo&53PmoSl*@zv0XRSVk^hTnhyj1Zq
zZ>yiQBf6g3_XyB`C9K~c3a>Ekik9tAO4IA|?Kob^Ie#;{=aFzP!pG@tmA1`xD<`73
zt@8~l=+W`FdL%6HlY`ztS^K5jXlz2=d*M>h4cPk5Yi-Nu*9II`q+WH5(N5xF+q5XK
zD8j@n^4D8W*F>k5RUD1Fl9Ouqv@5j8s8ZsgJs<?pqFVHsLi39)jCeJzO%5r#xG9*M
z9SqpS!u#>8n+~)VM4FdJNecK}{4i6Su*!XRy@9!=kn)c{q!0aaz^bp3?6*}-pUjA4
zfB?ECJ9iM|yCzRtHm4ki8}4}br-6gm;-YX$Ce;@lOn3Z#jMqDZ%=Gxu)#nZ`I%>xg
zvDO>UK!)3oY)jc+SkHH#p-FyfAS+eW0Wv<uyY~U*s`<;!>xjFOYH`}7y1GhXR1o@$
z_-9SK>KOLQN*veGH!&bXpQML)M_q@^tBRJka)db$-T1}~qo<<fl4Sb@o@S@!AhOn*
zr4_oR#=S<EiA5As+Y*U#q!yPWDFLvvjWgb}fDy^}Nc!_zV1BP%-9f$3%_Dcr+rkoq
z_tX?>aK*NlHrey4VRT{xFCwTg;Z>o%`m7O%CWESGp0QNXuEu6s_8M`1R#h7w1YGj&
zNBW13-$iq>CIvXvfvrWObieqsV-~~her5l)@!q*CAX_1(!+>l4$XwrTMSb+kOQbo>
zzgq#oQXoSzvzK9;*MYVgu>D;QVHW#2Ay8)N@`9|SwpKK9r96x>f#@m>nV6U=p!@Xp
z7T}6`CRDM$>ENIxV!Q1_nZ#%Ie|~{I+*T5Q-@-&I9Gy(K7V9Z3NSk=#3M(zafjz9R
z_4X>uA6z*btkYd5UVC1qKt9p>O$4@j@C#;3|C;hPwy*;PiLjTRe9p7@*b<aB0UEHQ
ze2#=Iq;ls3kruPL43SMm7)^igac!G>at>j0?UN5IZV5^;YIRcRxi1%RHvPF7{UEGo
zTg7mg|6=#4budfYm6PRyVQ~9m5K!Oix8boR>WZ0u?|csadz<*ayn*eaT{r87CFiIs
zOgve`uK>e<SaP4My_z6z@E=}_OOUI$ny;}-0rkQ^F)xIc)=%}%q($fU>$`3yH15q^
zw_*NW?yY|UR{!tdo;`6ry8MsAZtwlJg#0_Azsx`XkFu$^uF)Hx|Ksgs|CFxxfB%C2
z75qc8)7~^)lATk4d^Y{ogwOQlIi~QQxv5QL)RY)8XxW3mZ?V)%wg!=wY~Qv`o!h^n
z`j0Og|KNM`pP#ap`<d8&_hoO+jF(!_L3c?jkwh)}KIbEjUANVt%&UzWjTT>ITrVu-
zxc;WN*-g(9%~c!4;`=j-X>d>LLQ#b7pRd>b{m<j*yc@{$n&lKhPw!Qe+v?e`s0LqQ
zkH^H)$M@@dZYI<>9+`i7GF%mx@lS09UJ1Aol^i|@{^F@3{8MMe0kO<J+szf$M@{;|
z;U5KOuFHR%bL@Ip_jxq!Dl49Br}f7(_GU5J{>b2|O?ui~sYR5HIgETCw3|*j`p)+T
z6xbvAn(X(|G|^HvhCufBt=GR_FwO;mEL(=w6|IXz6dO`m^UhC>h$1ia6fT0Q2RC1L
zw+5d;ke!NW?e0N8U+`yLe-?iK)Vf?Gs6mpb?ouz_2GzJX(@G5Eg{hqQeWF;HPKEO7
z6gX?(>g8XPu194ScW+|nzy6+`4vH@Q{rWyG2;dv=V0v%x=^i%hqThAnR`a*tmMTxJ
z8$^QY>z}UjeNMma$a7RyHo`je>Fd*n-n&Zs=W4FKy%dPK3(uodz5FW{nfEs>HCi{-
zIY$z=cjqULuh1q=KYD&P??qZ(oE#Hl|EuNkm7-!~qNz;H#jklfwxCsSqFdMW9~G&;
zX+C=x=hgJZWisRpz|;P|rH*>~kb8gNBoyDkbuNgV%3D<E;eMVcSJ*XvBUy69d7ns5
zixA4L%o9D6u9@1$bPIE&a9n&b&zfxu^5P+;Pt}rb%vE#T!HYCvmq-7*kGH+%8vkYR
zEwDIsh}>(Z#mQMNw)kJ|h{7&rcznK8A?BL*c8im_VkMen*`+C};g`okX4$v@?0Nq)
ziS9NB0W4bsKNBPQ?=1XZeO2&(Wo5IAE9NHbKhjcfeW2C%<lj;L?@at(iS&mL5yip$
z|BnJBUW<fd%l~}9wCv~{(dz%;-(~)vYG`%)k3t*%=ZD?i`hOWq{;7s{RRq+Tn{^pv
z&U$AN6W;Jou1jFP^)~#tO1}B;G3?Qoe-C%}E3?L1T$cW$?nnYPcTWFl!OFZ-6YF<x
zwk<_yAy2Lpv|JO^r&oJo^sYUpOJN5JnI)YqUJJ`>e|XI=K`aZ5rdN7gS_-><&w+o1
z|7#S~n_lj*Y%6E}T#LGH4hy+lGp+q_x}yHCe_NiOdp}G!J<_v8wMYwFaJA2L#a{X|
z6?PMQSHU%v!t?t$Snqtx-?uy-+9@bJ`%UrHT3C0+bG)jgcHA<Xd_meS<=4$~(!ie3
z*RRX~sba&;Xi-)d(c!t?y8<$`KScSaJXGU1vg@#YYHgmOR?XiJhj*_B-xp|V(ZL3z
zLerQcxZb#b@O^iB8odyNw;8(fZ?tX6HMzt4OQrp);lY7gu=V2MceTd+2^X2fnNpo|
zB{nGotX39-M#`voTS@g14Q{h{q*h|v7b&bD`EzHN@*C;@P-m#mn@aheT4hO<v&fsP
zi@W{(U-}boP9F$<@3`+=PhmqQ`Nr1Z(FSZPlmyNaSi)W7F@N#YtYj{sGBdg7X`tEc
zUWxmNON(@=P8!<`o-u_oOsSl(fm$OV!P%yvZKs>ohmX(VMkW{y$iGzRi=Hu*f!Y&G
zz+Uh5vM{Q}pf66+Zm$h?44Y$Eo2Fj`Q^a$XAZba7%BOcVz_@ByjLIpLcVv;IOI(zv
z2&7FEK3<wt&o1h=!41kohCGY3)Plvie{_1LjB-etC#(pj971~xb|2n5=Ac{f5rmy}
zF?X)W%Q6!*`7IzlwlD%H+eRSeX8=;C)kfSywgbiC`~&5E<!h3#FLLco_0R*Z*i}1A
ziPnwbdf6^3+J3D6UG~Ywh=cd`r|SwMpuwUIo8itOpHq03p(h)G56>&5r4mL(*At_0
z_55{){azGM&H~lpSoO1}bp7=U-yi`aqMdby#f8BmBXPxPnGfkYX)U*7LehD{Cc7Dy
zZ*Jwt+EyVAOiIT`*<U7Z7;>SR#*p5=!4W(d@64=Y0G^MrH&0Bt+vVxEUH3Tm+{7Br
zGvS`bRBYrSm#T4+x9{NiMyWX^HUDd^m#TGqntTXhJAMv6%DdSWireeol|sYE7rgq-
zWn*|qb3loyWxZb}cS}mFB+0qU=PUj2u!xF_SctzyY|Dx~WY{wsQ^v$}!11RzM;Prq
zR}cT~h0bh+#w<-k=egLY(PHzdi7Rl3g8`!fkdehF+of?`s8ucPc8%6MwG)z>1@Tww
zUl^6UX9c@xuI-AhH4WOQQ`6T?kU2t%Hs?ydcx%m}7rqLx0=*11V3OC?rZ}~@+xS0@
zX@k)d4mAC1(MV?tkVOH9NO*#6sa4T5W}=04X-_$dVhO`+?Bb;1qT<Oq#m@Jhh!6xb
zB*}P_M)~B*B69m{9X()4Cb(w{XI#2Od(tT3;*W#xfaD5NlLXq#DO&F}GwNB`H(P#-
zAl0p<DAt^31EA*GqJ>9B<%y5-NYx){>x+^X&nrwc*DTS|)N^S#`*%1cBxY~hJi!B@
z6mdFgXF=^QUOphG=<|=s1DTtIU%ns9OLEZ8hIJ0~alSwUlX8*GpQ_VqDk|D7Ydv;9
zp+^izf?#yZ^0Al9`5$WM3xP%(LMxqeU9ZNAZHy}4ITx&y^_Wzu!0xfibI*kc{sJ%S
zulA*JP5GLXDLM!WU8zECk+V9Tl_;%osMK1cZ_XcegW}2>J2xKVXc4mwzd>k`RX1v=
zD+i+<Oi4q3JZkmwW4E6a+3P{MU|UQ6mt{E9`U#T^RIJy#z7pHhfE5JSiVW6B#)PCf
z-l{yidb!rb^>`uQROH}+1DUbLkj#Q{6A>FJUL@`(+FL#h7XTd7W}PJ_`djvK9j*Of
zj+88|bLo6qkr}D7xBVq26*R7VXU=D}V~V)B!>p44^{?KxJt8O(;;L$#)Y&CJo+&I9
zhttx%&Z2rnM^=dSa9vEbLX6&If3}JDX7~A=7Ax^arKeBZ8G%3xC70<mTOLgxReMi*
zb0rEp?}H68dzZY!X10BNT2Aqm3uZBtRk^;_shu+se@Xl0w*0V{r`zu8lO7N1#&%WA
zrO1644k9mVu*e_7&V*?fc2zH%lVckt-VSxk7$tWWf|pFu?@A;G0`fW)CSmHGZNtVv
zGnl}*56iE$rRbC-LcriG)3C^uwjBLz^cPL^#dM@RH;1{XJVOGQmhyHpnH4^=r>Icw
z%UazVm4t53^bPWF`V$Tt-A)c~6V{>GgqZs@4}9EAIs1|gZMe(5OPA~bm<3uYxu(bq
zPvvfAq0Vg1rnLz+)3+tB7o|Pj1`9q@3%YPoo;M&Rzrl1^wJ#vNqTA6g(7isn*Ztl~
z9AYB_iEJQSeyib^-*@$1eWh<9vj%DNX`GsIgm1ph&AMGcDdF(v1nE6pXgLq5FU_ZS
zCgHdQA}WYOFq5F|hd)#~@g7)U!<U57y)QB27@DRq6`NKN$VEx+d$}n?wcRt7w8%|e
zKX%a?#@<*c|8gs8P1Tb*4S5>JmO(}&onO3k|6WQmFCFJGo2I}C6;<=!)vGXtG+!Cr
zbD6H)({yLo!eg0)au-;dy;DHG&)=Y32UVuFDP`m*S>jlQ>*Whu3u`{tO-h!m4*Kb|
zpEP&zkZD>mIQVmMK+4ARbmGgQJ0jNV!A4jkgt_b%=46#!)n;Cr@;#*qHq}N=HHxXn
z&4R4tHOQ=sJ%Ke44nq{s6`m8|1KqI+t6Kd7GFY#=iXv)i<)n{TB$1x>)3<zr7VCmy
z0*bl+WwJC}Izi4dNSklO)!35s7(ifbRnP=i?f{k@MN-lae`?9Lg<AR~L2c7=Vz<o@
z=3-?OQ|O6IXHfR>DDxG|o<SJ}hy8<@@tHf!_LKE>&4u9V>>M=^;z^snBcW$=Oi2J{
z?veSU!)7;YMcGD;AO12cCAM3;_Z#S&au`Fa{pfF+LThrz;z)>x(c&`!-)Lh|tCrf#
z<k*yHtW#%&m#VS8J}s<4T{*nV*xjl#6f@C(8uz_#Epn;#`}~t`Uc^({-EICpfF$$6
z69`_TL?}CiuXA(<Kc_ocFi-v3Yv}n_=}d-Qqod_c$Yqtzc=pSSs;$eC7wM(T*IY%<
zS<Ld_S7*G%`rEa&5fW=^vc7qG6%*Su(*}U<nsB@^ZGDmHwc=OwZl9uE1>0@+gdYIl
zY<4m2Z;TQGVL<~7jl{$ksimX_vkW$4{=S7?0yLPnxK@{>*}|kf9p)yvf}QmRcErX@
zlndr0Tbp#FGz&(y9x-ym4Yrauo$-tU2qsPHSL3#r7%daDBRFm&#ite4MVpz+-sx>}
zx$aXd!&m?*pWrSWA^;xOJ~Zmi$jprW(G^3-ktQWGV58#bL{QOt2Lge5L0=LR(}MZ=
z_(|ez*YlN~ctxy41tyh4JJs;0tOS`#WSQD=7w(F8?nS<VS&y7hfo_|9bOU8r)Y~_b
z_o^ZYv!*W6Tly^6{^Ch=$~GhP_@`R&f&^$6;5fzwDi+sE11SGIdoV+EV42fV_*VN^
zXY|wuX7_No)899gtn=uY)}IF`xq;99-S?+B?y=YQ2Ymb-olk<IZhEE-?<+e?@%uE=
zP6xYIMHRv^GZc<uabYs|sN#`{!Zxi+%n}fbE^ApDWB2b1X4nw%`{Hcn*_)aaL!N0*
z>fZ!YB{AxZ>R~M34y4a(<9UaEj#VJBT@EApfNU*;5Dm;EFSNq9zku%ti}#WrSkQ>D
za-QX#v_&yAH>F5tE;}^FZFBJR{%GLi3O%vo^bfijCaA&UF+9UQ&I8EUX+o3k$k;7@
zY)rh5H#xKGi!2K8=HEU1xU7xi+=I2gO8rnD!bA7Gj?U-JXIp9k21FMD0)d>r(Gr~E
z(%!Fz(kx3dy>@vzXKeHlYJTTo_1W-Wna(Xp(^sbNwaa>z4639ZbdSe*h7woDF)1d)
zn&=6P7$*Vg$u*S3uAB!v3>E8=ebJd>W=W(f31~$)dh!DeDj*0JQ7Q2!5c=)~r(Ut6
z1EL?8ri_nLa#BhdDHME{;q9wssYRMuTZK3dVV#ox7YPX|c6PU6=Hb0gE)y2Qe)E`%
z1dGZDI*uG_j`?$$fw3qu1oy4wMc=B{e1cZ_6HcPpw_2qt)-*-y!$=VMs+48HnEhkO
zf6WJijb3jn)&4affc|Sfu=v#(_|JUsd-$#2Ol5(_SGQ_z#wfiQHpAr7$%2)}%2EOk
zK!n;QP?9Z9Gki)tH*y+H8uo`dN%d@(_2_U=xTgKj2)5`l_fCe)BFy-==xN*`M?Y2`
z%}UPuaQAz?W)d%|R|mSI(>7!TSP+OvD44u6tnF!IiQxXj>8iN~xBoS&{E3ik?(_bJ
z^+oa(+pl($`Q!8T+fwGbB&IQ@bS&A|5_0SZ=t2|2;t9g4z*jARm9xTH?(o&iD-(yO
zv$}DAp&Ga2K5Thp%7%c0RRU>BC)c>SO&}>n{n~>1r%tM7VJw4TD#^zD{G^(U7fGZi
z;Zq$>!gszl%N4$Um-%Ue8-W(3uG^%{1buvGJQ#{Q<c5ncj9Gy6jU%^E;Xg-5A^H_K
zyng>a_Qyl^4Etg84+|mb{)#c~v{K#7b^9PKbm0nZl$ep42iL5Wseqi&ZBBIfPM${k
zgwjGeM#9p^euAbuHuYLuNnxZJdgj~d(&;PKnc(0EoRO?m%7zawS18}6Y_PkJSBs#E
zN0^hat2ou|J{T}uOB8T*&rLC|p<ODAtaf?Y4c6o`PjCtxHnA;T2E5Eo^2AnbnIx-v
zQcWV;9w!{IZ#UpW9z*Kex1T@xlu!GrNG{!fy!CYhZ)q(8^Eq@YJ_BJpF5BGhwm-kl
z^o@%g-R&HH)JO97EiR%=Q|y`Cq4PMrUjS-9pz`RDBfsJkXxqRY)M3n<SAb10)7$5b
z_HAYt<AMpW*#3RXM5S4d(6_r2(xL;YTv693!jBm=#H%~HLLw7ca_z$-(E?L5icz+;
z<tf)?-&u7I+PSH0SH8Trni(AQ!0Ky(5fl7xxB`{Q)ic}{p4@|~G~+#L_|j`;4kHqi
z2c@IcO2;|Ama0UM#O2PKj&mo}twvc}n{&f8vsKOf@AnPxFKhK{fVV~pRibWZ5Us2_
zf_d~3K3vp1o%!-HE{>7Y1l?Hs*u6+S;-|J;sC^?id)Ay(GZ7qq8bw7l9Ki^eOThZS
z3qLfws3~k>Bux5MbmOa+@$s{hJZA29h<y{T<|`nWpE6hegu_~aMYTYsU@p1*!>~2^
zT7xSXaAK@(SU6C<F~5)P#b@Y(q#L3+E0yR?HErvp$J<6a{z<9!#o_e08C;(FS9dpV
zo*4a@SLh)FM|V*(9>Ncsven|^Yi8BkNfg)^j(gRKKuXzqrb3R1pU~F4VELkE&fj8*
zJyfO0m~stCveyIcae!Gr;*GFsli8PENgEJIjalVH7Mwo;6CUAZ79Cs7`qnz@4@vlI
zI`i6um|x#<LKS9c_>h9y=M@VHro!*FKdeq=KB=|t&7C#+EU>3ojk@qUMj6+?Rxe@E
z-7Zg}nX;`F6t`&{?3Pn3wsaKa*phlVedh)3r2<*X*G!qHFpgK-5|Pd>mehlv$zK89
z`sAi)&J3z5YR69FfDY-k5&Y_j#!1>AF8GB>s@^~TvYGk3@Rhr{HNEEG1iRsmbB@hC
zN%~P$LTJ));&DyuA`Ks}=rlG;sJmm28@}CP#DSkjon1}7JM*_4wzqg{I!UQGR!IXG
zDj&qo7?mVI!(O%tGNeckc2+c~qX~SFe#hkP`M1>uMkNsY30bEphq@-mNCSs>&Yr`4
zyI}`&o+}ABER}Qg)4C|QiR->nt5Q{84{C(}6&HKk^Qa#sTxER6I+^hibXGQHoIgqs
z`C>>9GI#|^Q@DOmE&X?2N|}WBT&5u`@?-o(Af3K0Cx$+wPNWz>DT12D9@+bi=$-0r
zr4o7?M$y0U*2y!_wRQ<{pJaCvUxOHFznY^FrHhL8+!S|-4CEiIMG|`3&X<kQ6}6gW
zbE(dCv-(D5M%htr+?aI}%m==cmFREs^=o+k{(zDItowf5spb928s|GCRuQ`oh4s^l
zwZCuD_Q{}@GStg-<d18Cpmj+97GI{PwE<KI=bGv@*xwZ)_Cv|*>Gv;}a>cJ@kl9#S
zmm)<{%!P1<t<G42RYSFny?uPfZLm2NJ5H=#R@E4FVhz*}k5%_lXY^<CkB3w2gFjbi
z9(HFp7-EEFc(H?0jrJ2v>=~lx%s$$YTtXX#1R;=|2eUWK;Mu~S{<FH&l`)6p%^0>1
zy9#%%!6!GNDC$N}hLHpA<PuxsUd2p!rp*a9(B-ZqKE8m!y=8DjX!_CN`1r<1RMr3b
zO?i>7&NnsR&gWuP`2tVnWvG2JfY5xEaLoRj_&8F|Y$+Lt*`~<Gv(EI7u@KAc8Hsq~
zR3~1y+3@i_2B~S8#Oy*i7YF?U<nLQ?^fj)AR3`N<PF9#Po-sk#4&~zn#}hJeuv(FU
zky+?6w+drP|K5O5+AA$@kaF^NFqMty_e}R|6|dvc3Q@WD)9J42ZzanFwixj|cQuCw
zbSL=^8L=_Cs7ZkM__%Xf53)ig0%XAV3bCFtmW0hrji95^n{3adZ<n_C|A>3fw<P;N
z{{O10EG;cHEpe1vag+mBYUMyh(Lg}OQR2o?Dw?U)RZdiL;vOji0`5IC_cjM^tu!@f
zX^zzL`uw<#@2%e-@IAgaIS+1}yn*+5o{!gayyQc^IIOY!XFv9~5-UMyhfD&&Y(>sh
z-*}jRIEa{ZHi<~bsn1GCI?Yb0gGl*Nb$|UW7<f1k_kPaicgeOuSRm$i{i1_b#$c_>
zyqbSH`g7xS;G{s`E>=+O(sj7Hj6B6sM`7b|+j>Iug@nG0<SaU`W;ludepp##koTGw
z&)^wz6NA^tFtbwi>v!jORd($ZC*S=s;(5D`jWc?HXFPCO-sWsfY@N99&;8LMu1A94
zp14eU2}a$cL~-qBEYy7Rf$aB*H%ZV8)B8~8=B5ArWqXpkeR{#uUp>w4?ojSI4A_3x
zhR><ILN|H>P$%|U>ax};=;_8&Ib7xthl_cZHPx_qR1VP?fKC(h29o$_<uizAI9js0
z)qIU)GM%N5)^DGMwhbqn8HQ6o>sUm=Ba}pPRHuI?^YB)JF7hP9Rg?<dmZer%H8cam
zknYfnyF5Zc7FFYY6`XAPyfExFV5Z?OVBEC`ff1o@qsZdu-Vs&?xUsKet?G@D#jeP2
z?a|B-Zo3{S(_Nv6t>X_RIYBM~j)82@_l%Ue;e+7gdLM08mCHm^ZGa&Ti{z42-Qb^K
zxQ!?4lMTqY>Rw3`b*DiZvQW{oKhrt>-eJsP%Gk~7RXNhq_N@;c%wswqo5)QHe5X1G
zqDQz5oLTP=_b5!Iv-;=d9%d6XY3fQpGAnLD`t-b{BpT&fZ=Pq_xX)3|#*04Y5V(B2
z-NXkGV)qHEN3k5&_&cW3Qi)O1?Ot~~+*|@1C9LGq9^JnG@g?i`uzk5<CwQy7sp|T6
z<6CyV4=9p`)6*HIvE}JcghGGHk5JA9e$LfD@A1Dex#{n}^^$SkcNG-(H+_i}%P&8Q
zOld(5-3y(9)&3@Z$t*0@MPnx)#mW-<j#%>#;s8>Tb-7k}A<jLL-!>J6%CP>l$MM8w
z>rL_2XqlJAVt1xy{#RNIV#oo@xXqI|_q)=P)c?Tgba6-QUY$vXY@OC{8b-YecUP?O
z7t+wjz!QPi=>s!t>o|cq_V29Llh<ncB^`u<2~rhCbi+VSb0UH|;Juc-K5@~amB&jH
zdRGMDmgBfOIm=Amln?*2tr%=hHOtPRGHK&}d;7=7GP^Er5^gTy${4sG-1Slhz`+%*
zt>OF5DE!bcf-erwIphzs9RB(sLWx8E;?83CYZ1XO-C=wLJcqmt0c8-yBbbOj^1ZR*
zIlYW(pcfStQJT^_z6>}&x%xzlFc(#~s{S}zYslPey%t(#T~m5VD+}Oe>?b=}r4Fi6
zULt|vIcQrf$#<%t$I(F!Ag+`&XME!Q<{sZ?cOi8>V!@K#-G6+yo(otj@5miI2eha)
zfcV+y`WcM(MC5C;+_>m&MxKHxm38ZO*S(^^&`Re8FY+*?ZG7BUs?g2zTBKq0xA*1_
z^3~6+y#o&N2P8V1NbNqJ&Ulf#=sa=>xIpw*a9-F!f@Xs!Wu+jqTYi>F*#&;;vdz0z
zvRN>Wy7Q<GROaQ7n!_(V?mLeX?s(uc^&0wqGv2Vagn?e58Wm+B`5)rpz*5BNGufl0
z-$wUY8cd>V!Q<KHujMrs7sj3vAc@vyyip#=;2Msqdm@RZ6b#VYaKa;__^yRBwggpX
znHdX=PmTWeAWCY+Viuo&ndju3_B~FZ-$+vaS+WxRwTj$v@m6PZ^L#bG*`QHU{29p^
zu0Z8l2M6>ywa%(h?&0VPmA?;^t6Utzf^;5{x|)j5Pj4+QpdxVHQAoTmx+uS3A=$jF
zOJlGtcL9=0RDpRFs0jZ{T<8Yq;n~-3?^iZl;BOKfr#iuW1%1z%Sx7%DG>n%9WLO70
zn9#6PU8C{s@q8y{Awn6cDoY(q*J`5%zJcv~1FX9j2qFlYnP{V$fM;_liJAV{cm9?Z
zYqm%dNfagLW)@Abx{xw%uyw|>J(J?wtfU4wa>=<~Ig>@#uHh^3vcG1UG+SsGqL*k<
z(CP;hX&&WVe!{<k-7S>`!cu=`K&)?*rfqj^#!IvavPRwYcfpN@I1(T)Yl6nR-&nlm
z>c-*vQdlvcD|~G$pcN6plN*Dt5Ce8TGC}C%-N}All&soF^ovR4d&L~vyGoH<1igw8
zx^AY2n*{}dGYNH(Mo)i&Dy*cQ^EF#{nd;>QEtM+k`@m&2(qaS9E|`2z;Jblc`fVE~
zP*Nn~$*7b$UbMG#()m7p-rFz`VO&$cEBz5;QTXCrOEd1<_9cCMxSYSL_RZI%_m`-5
zN-rg6C8jb8s}{S6E`Is<bYL`#4`~F>iVSfMjCtx&aToxkE8+1xD505g!9AY+%S<hk
zY@-gRR|qB0`83Lx-rL#>O~=ZyU6#3Xt7Xft2GfuzPC^cOxz_yfY&y-bt8<j{e(JjW
zA=IbNe7WQIq#UO?Mw>Xk!KBBrD-4t7`A=a5mdwTjo`Mcu+Bi>?jk+MhW|T7SJQyZj
zaw0dDsxmUHmA}sW7ptN}t~2Gv=*`<a5>kVGL;Zt#q0I+sWSIjnbey_^F-Xh&r_|}d
zE|K*Fkr{Cob3JQ`<NR&jG1sD+XpvoUwMo+Kl$=;cbdUI=(Q9s%D8OX{qmE!P@+kW7
zy<l#SQ=M1{pWj^7rZ6M@Aj~==o^N&qGgvw;^6Pg0X#}i}+uM#0c;3~Cmm@4Ha2p+3
zKVj;VCY5uG?h9~Emlz;uwy#IIp60hSS5voTN&PS1Rt#SoUC_!Cx2G;y4Bgqj%WaAj
zMpF8j&W{cLsS$!&CZ~q+!)fXqSe|}x$ZtbyC6~FAE9crWCEh&_9>h)nNoWE<zR=_L
zvmAmX&sozRJe%C?zun+-Rt*)Bi;aRw;@k>upv{c**JHf1QCuW*>ZI)6I>`O%Mhd-9
z6S0w_Oya{=pAd9CFA>zv9l;-Nc|2X|4(R<^(CghTac5~Z;F34iWUqh7tj}y<xX;We
zB|9aa0}J;-shT{`u*nD_#Gb`CXiKwFe0DGuiK5`h!_Fw~k`SRO5T6o{pjgFD0qba|
zQ$Fix(V>A8YZ-Ok8hWO$(rbqNfSP<T7|iT~>-Dw|GrtMf6~Nys26&bCEa`*6xPB2n
zE19>`(qiU}&_%TRi`J*VBimX|3WHs0PzA7C`KIw4sht={if-K6_^#n$(uWipTH)1+
zY+;{?#^tov54*1&A7XTe`I2rvK^HJ-t?35LQzh4pO8>n*E^t@8I?rwQqg-OCJMTFp
zMGR{4uah30d*jywO81*e^zcxCi|ae1E%<S*P_y-&i~vtZQUI(~yyNEiACmkf!3vMK
zcOMDwqH?&u;<jBT6hg$J;3o0}rH!B78-J%MPp=(7h6ztcmT9i$Vx}JAUVp~<Un(>P
z7h94{97Bak%{Ns9GVxYkWQW#=M0=!RtRx2&?ky|kIWzQY!$prreZP@+A|Uza`lsP~
zFj@ukk-3SK@wWY4*_oOg#(n#blKt7UgL2+Uhp;aVR;qqpYJBZ+5*^x?%N6G3+kT5J
zbTsQ!F9G>lJ!bb0Y1KuX8k<QHEu*#I<%YeL#y=ZmxMw96iH^gA04WWY&aI*!Gg+R*
zrIuSuTDiBoCr)yuJ8RPWM94Ob!S`4{mu)9UJ=dyMGlAzp@=>>CWzQs#omcYPQ;A`0
z=UB}bKY%WZdElhiP_oH~gv3;zLUs>#4=!GAq}@uT>$jHvGS;%LC1Ny3#WY#-uomb?
z-J#-=ZT4FCYP--{;GkV#86;Mpw-^$4_ZjzgbdtR^W$eqO>)0lupTOgpS%1p4R4};?
zH9=F`@jT=vknYN1=pkk7M->$P@|6+DLnfuMyI^uBSCklU*?!<Yl0iG`g8`qL4`){Y
zVVf!6xYl>f-qUW78V^t{`Eon%tNfWLrNQPKpoXB5c!gI1MSTX&?8mMRK1W_LN51);
zZw`hY4?ZBU3!h9+c2*5Kf5{7bkl3u;O8Z4VeeT4s1$Xx!RU<@Ab(zl=&Q!8zT#!<^
zZmCw&YRZ_0%{~k0c~^mCCGJT*Q9+Q0B1G4)+b?zgAWz@<x(a6GJNAoQ9Npj6vvK)W
zp?EWeK5XU4SUiMbU`xmiOFvwFyvuA~bdvvGWv#P`W?#_GjDSvv@!^X$=K8e1e^mf_
zQ0Dejgt#V;!!=UzwrgR(>HR2vEwR;N<MMoiAu;_Bp<-odDk1aeJW;-H^>0-TM|5Y@
zZLe+}8^0g5atEe5gSvGB#S0!y^I=X8*Z6f>5TyHInis>&Vqs;dME<03Ls;Wah)IAO
z<URAmXq+on^+|!5t!PBn{n2nuei)Fst~IBm0R5>NP*Cf16#>+ELzUH0QBo{T&x%zr
zPr%vqXaTWGDDH1Zul|DFnY>l9q=m!28MaCV4w)L^Wh(FR$bW2+llbx>`85ykjeW4b
zm+@|5c_N_h5i@8I$r?pjx8}^>`PelSfLYeYv<KBzoBDk!u;UN4Q%_$gSn=-nm@FDE
z)*=o_J^{9<^qBK~wlFQ6$2cn038}Z1UHFgSom;!aAEG(tex#^-p&N$Xr4uzzynPg9
z-<*s)3#cZkj2*?Gi0tKKIe#yV>F~Q)m&VS@+q-V~GOM2((49B}LFufPDTbL>)vfzs
zBFO4rh4gOg7uC{z2I+ofKBq5Cik?*k1bEf;_HLf}7O7czvx@hx+UHhE^e=5JUaDx4
z{sYKNdsc^^I-h4tu<-ks5d@|1)yFz?R0&5$<+RZIvXRpCjeuI&<#v)R?Glki%H88x
zgygrG<&Lnr$mou}B3ryD>ei@m_<Bvl#VYxd4+ZzL&?Nd%Ok?l&$o_Mg1H}x9#J70m
z3Lc;?&x0N9>!g0E^8~AS;|#i)ZuPp`kR({U*tNsP3i|$LQ(<*VK?R?i*ydp@|Jo#$
z;kEK?evg=92l^;-)~+F-i$1jdpi+@euCgj(?i^BC=`)x6YUIjG0e+&xtoGZ|l({R=
ze+K+{4*2!vW2jdL5*JjP;~%-JLp2zzGdE_INI&iaaDrQnQWz$tkR=;P__pW{<J!^w
z4Am%&f|S{N8{Nx$1$W;0k6JAdG%Cm!UHohd(|p^aTh%0n(_DU;T{Ast_Yn_R;)bQn
zy8B=}=Z&MEdv_A~whV%`E)mi6=n`9C2nEe@V7a<Doh_fbo&FrwfSq*Uf2i}_c0#V(
zTbbl<8uoP$D6D|-2E;!ZTZXJtwQ`gA0J&2JspBJ0l2TjS$4Zmr8Xt}E1IJKOM6!uh
z#S7CtNOlR=x%m>Q%RNmWQkjXYv>h6>J#1#6BEadRS*yKW3z-P`yw$O9Vx|tJjx1x$
z3{xl{eg$z$<#9=LWRxDI#T-3+r|(~SFqDv<T5CJ?U4N=)OBg)-(s+kgH^|ufZah<~
zctZ#{Z<;w*?;VqjGwuFZk5T=j$FL1}wumUG+@j+es{3{}5dArdUYe%6(U+1sWw~FB
zzAMpP7o3>_C~Azh=$Sj@i|B!XG&2U|{mUp@iczS>r*DX)C=rw@_(5jaSs%1^@qFQY
zbCALo#g;=N2-f_Ff008`mHaQ#HjHn_MxcskcV&rXEVbS?Qm;viDUN1Z>H9CXu&&3E
zhnfNDrJh2c8oa*GJb2xtCU=8o)7A{`Htia;jqtJb-2Y?3u<@t)V$dpberr6pDm<qT
zo0I81a&KrN1~V!mF?-5ZBjW`!L;AD2<1eRA?jKSU+kR?%yYvMs_|!SLa6--PlF1Z#
zb*a1LenJ>5N{Oc|8o1Lk3><e6Yk$pgPGjUVd*dD~QT`>@`BC!CzYJ+$vc-h!B&GN2
zSLJ+u*v@R9RX6UulS|=B&TL&>^pwspimn;pSiYaVxdraydR8W@kry~px+U3#jyKzG
zd-<gV66Q6~Oer3tVAX4Q>b&i0w*!mD`s{TAVqBD0ifSt{WV@7k_w&5JsxNKtQMdmZ
zS#$(~^kU<7LQ6@q_5LCdCbnGWDIg}E`0isb(gUd#O3I|oj4wfl;ADphyqv|GwW>>k
z$4Y_GPhZ~23IJm-PbialZJdV3#O|ks&Q10wO5f^EOMeXzQ}`hu+L0a|>z<fRaHXQ3
zs<bus9twQETDFC3^bJu)0&w+Jn7Q;cH0eR5q;dF$P;}z~V2<>GF%MNX>+$G!K`jTr
z*3uHfTauuWKS|eXaA8M<>;L<UZ}I5J|H9ZERIDWBUH5AMEkZ(?BDa%2s5BHei#vb^
zx<XQa6&uHU&F3f(WrIq+JWSSucE*552ZHylb@Md(W>bdXtvrI6hXYJ;J?AV|T}>B(
z!+}?<$DmTI1AnJnWUf-;dPM;AwMR<7H-DA|X>@&urW#`=Q$33WW^hMeoAuVZ#Mu>Q
zL2ZkfgXOnJ2~bE~vBH$;w;kS<1Mc4JCZThgnVGq)Du~>S^z5AMbOK0I69kaxWHOnZ
zM&R^}jP$r~UGYjHa-aEC-OF&s6++)yhW&|F3p<M)ETvMuQbN9-lVe8|XNj}M(`lyX
zjcyQku@fMAvZjdO3FGGG=5qhn|M8j0tp%57SE};Uvh@bcWhY8cgro3=SiV4-BSL1M
zQF*)h_oum5kK7fDvJu5#ePIx0lG>B=6{oqN0Gf%bSQ@i)b$;dHZfhEp`~Yp%fkw9T
z;2@D=?F?%)jv|(tn@uwO^ZhlVqw3|c0<>t%4HM`GdY)c+O3-@*_ei+OTN6H=Aoh|?
zrXxwCGD0V61Lpd!51}(xi<dZhxN7^o%5Sgyw{*{ICHA^l`R-Qqr5`M_z=R(utGEB9
zO|g8A%P+2<3z2-LwKbT<Y|l3eY}C+?g-AsjbUw``_-A&2$3B3s#w4MZo_1|jEE^(m
zt?qn%iB{TY{Yr<WaVg@iNVaM+zM35CA~SX)im!{(Rn+fz>@fq;z7AhwZU{A+3^1{L
zHoe}VWM!5;f^hFnX*$lW!m-0&RDV*SW6T;AC3>G~kV%lX@Ct0hYaGF(?2>t5l+Uy8
zFAD^jBk;m&3&!$Q+(mVMa%mA+Ta3lk0OemVUCQ`?Qv|NgPIUz%&j3?45}0i7OLY5$
zI^8t`I{$w2xlvZ0cf!->5^pm$N_BzSB%8{m8U;x0*>|@zJ<7U&V%pJ&X#2@pXweRZ
zVr4x`?*5PFuqCoj1iWMK<*(aN8eCdTUK3c{Y4Wl=|8KZU!Xwr~PqpFWEF+hFy}6!U
zVHOOvCFR+;5O^RGGL{OzZ0dOveyjJqR$;%*Z%=qxef<-lL6E*;q~5jMBXh$oF^=t>
z%h!vWR$VXq>g3cF{5VaoZs33|!^=&SGO?P6K8Q`fx$1QvBP=3~o0Ah}<6Fxv^L&Pj
zyYbZ&6{?15bypBW;Iar3?$<U>;b5XgI<Z5U$PP4pJSLIfYpQj*#hhl7<z>|tK4s+p
ztpGI2hm4Dy088`pwln_-8iwhSyR+kE&#HF18sy>G<on;d0=t4WfByVAGEwvmsKmno
zRGYF=bQ`YfsS1;`eNci%UbyKEe4{AnHKdj*W4(9Fb~+2&?K^Q7m|<4$3TNC|9a$~u
zxi0m&-h;1EI<Db+?6pH6p+d7v`ta3Z0qd67-ocMVieXlcV{6fqJY^r!X^W6%l#&3O
z4>p6%53J1*&Q?Q+fK!Hh5qG727n`If`QTaomy6vRm$;np?hifvDtNS8Pd#DDA-52c
znT%)pQNklj6*~@e`Wp8lIIOsZWSz)eW@hd9F-$fo3$st2x5me4{O_;0)FC!KoxYB_
z%&&Fs*VwPsrql@(Ol_b_TFe?GKIjiOo{V};l>F`6b@N<pT~<ilG-=mid<Y$(Zq{+b
z2z&%6frY>Sji;$L=733D-rJrx@0k;e5v%J@DSUCxY9%O09ldjwHEqtVOE?T}R~~=*
z^&_ht$RkM6ydR9wzIaubQ<27|ZQ_1)RcOe?>UL8Xv9avD?UtdJE&n#r=b_hcy(rUK
zZ(D~%^LMtX+g#Uzy+xu73Oli!%0yFYr#>wI{%g)iC&Rc7caKzCy{kzO<ETM;4Vvme
zM|cL%!xvB$Ryi3oB+!44TZ&5O-K)q`QrA>h5%?G@g28AT{in6a{`KEJhO`JC82FI#
znhk}ztohu-{o$AUug?>rI4<9Sg7qUf2?qadXn@HY)<(vCMdP0K>i<tT`v2_3SH?lt
z|Ne41UHk70*jGxU%B&PgirUA4f%w`f(<gH+*A=pYcFu8d)Qr^M)ODfut9X`rRAD|Z
zbAVrKtNxf`N<CY7^6>t<|0^G0ozPWPCvhF6SQU>PtX#B<dyLtnJaT#2{VqqbdVAfN
zR)KyUV5(&ntt4$Vii+zpJSprg8_ZIweCc0_!m9i#QJH;Wu^-CEYovNA?{v{xOcL<q
zilR^~>{lP#L#m6IvDxUC!e!0qvHdBNF1#F<`8n&Sl3z3(2P%1C+}h^4=i2r_1|{GQ
z<`mLag)2bX_CjS`)zbnsIVt-YO+qLxX0p>d-ZS?XrD|fpzq+m^8H+K@3Rv9U(~t0*
zdDhirQ~iA)uz*O<v#(R+2#rBv6F~|JR3tyif+!~O#(Q$oTdM#v7Az1P`P!k_5xaep
z*Kz%5pMH3g);sAwp^3zrjnh;5(*2WoD>n{X@BkF{vdjo^&>yyCT4O2R`%Pkn&XcUH
zQf{Lpt9e$wn0YacNw_WiHj1BjtEjQ!#y8L;!d4N2>$wtW?{(T9`)R5(TdTiY0Btov
zw8lS0*C_xM?D+tau4hk-11@5Ni}wtEUpg{Y2l#63nZ$=+o_OYYXuCvh99doSc*Tz@
zKL`ec)G8p^1c`N<V}Zzxp47K-$xf3(;N@+rnjoNvv^XjIidlAQk+EYVu6Vf+Wt5s8
zIZfla?5Z%sofbnYo)}V$$$q<l3B11eugbKQ;`Jp$iCK%+rEU(7G@TM_eFg>fSPb#J
zC{{^n!SSoNazRfJzV2bjw)ES~#1l*QrFoOrzQS_+i-6{Z3p@{3`F)bD5>aYCV3(1C
z`_?1;;$li7<xYl?je~yJ7C-F*l6fM-#LSWoA6Z>?olaPlELd>t!f2L(TaoD6Nlz<D
zvyqK%UN2_v+w&Mcmcr05)lQ>)mwznq-9cU8$X6r`S$G#&_%R=0#SI5035cf|(-F7s
zzVAQ4SQesX`V7BccH2g7Un$g9L9|UznckpsFUE{Z=1;g!T&(m-&T-D(ssY<nWU>0{
zCvFY*o0?T2x2uiik&X7(yVo;y+fqE%gL7_QF*dR2XG}U^FJTIeO?h~~w4h6)mroxS
z=)~gevZ*dg@#o9S?pTMQ+q2y&W$L4%p7B;wHv@BC)XkHbIR(gkD_OaO^pD*^6KrMX
z5jr_N$)w#Z&=YAy9=%-mZr2H@v+_pIS*vg5!_wqRKVjV}C{Aq{QkgF;lBY3k9?K=i
zEhRyrNG{pOp4wT7*?7^G`Ot=H4dSeX=i~?-QDMho59_s+DDKBy&hC(9&&<{Wc{Lq*
zTGqqOe>uFi`_$}y^jhb=5%BBSQg2-!0V3+Hpe-U4RyLDv#f?SgQY{TBYqoDGE;AEE
zAgMg~!)vlD(%18TJUC=5QaC=m`Zf8oY0T64!NKDBU_*0x(u&w%Z<q1U#OwqY(R&W(
z5Jr_B4c;>my??a8=i09cOd{O(1puM@Y6g*g>73DtfL>N?HPgz#&t|*NZ~8@*Q04KP
z<J8JaoV%a7w*NpYVyezA%3+s%OjDLh;BzSxx}+u^GpkUHoWDyxY5kdTo&!x65X2Q<
zO?<m=`N+<a)C`5U2Sqr$dc-2hjl5DV?Y<+AROEt2NGerv>p{SMX3Ju2f5zuSIaMw5
zTgiXk*NC4bNzYc^YZ??!z3I&<#42C5o0!Dv^k&R~dXLr4F5i^4dDKkBaQ)@-Z)f}5
z7}aN}nmUgf434#2u?+x3087*oHYf;}YF`PaJ1)bATh(;^@g3(%F__%oK4Dvs%Be82
zvNM^4OOCF1!3L+Dtx+&z7!X6N0foiY%h}DFJeRAlAC8c*yya;6flpzlX$`7BFU_#c
z+KLPMJ{WJ1L>cP*=h&(z&C4K!ykro;x08kzPezX3YWG_Ssc6Up_iR;s5%KVj(Me0!
zR;=|!O}?|rYFK01=FogPs8h4lk7RYk<NW^UMH53U+QVot_1fDQOL9X?qs!8X`q*${
zwLU7RbZz`P@@aqr**rij_+`S8e7p)Ybg=qs;gWoiQf7H77LHaf0120!&!bNno^pH=
zF2I-s!HNC276|$rE+c}?MX1;r7kOmD-S`%}DWMVXGiNi9xZd!pkIk~O#I|hy`STA0
zavXALP=y%DHh9f+5lD;yiB(Yf+Lq!kJ@}xK*-73{;)!^Y#wYNUg+Frcc_f?hT$^=B
z#)Ua_BTfvKO-`Ik93-k1O@_g5SlX;xh3xLY#G7YOfK9n%WZ)x}W0|0PmlumhS7i_$
z5ND5c?=TUGN*!T-Ui;F?yY}Br#arPvyec18ugGfp3^@y8veL6ft9M;FGZd*>lLTB9
zXmYJ`Q@#UJ^yeN=#^pYL8*01v_k>As`u6uE3(I>$qm+z&a{f$A^`|)B5;rz8tH=GU
zhPAPqXWr(8i5=}qiuP=OF*5YJFxcY{f-!<h8(B%8rH(OdlFrRq2a0G)pOHaml`Cu`
z!JhBN^jlZbOx9?eVaxyiir^A)@C!(lr{>HTH+&h&52wB)o-6yi#wL5JX~i?gdHZM1
z*DBX1At}!|J-Zrp<1KK`PkwB$%G9MiI!v{kAwUm(VbpGcV^>khmA$Be)}IrM-|?#s
zKc-DT)V+UcBRae3`DnorJ3A0MQeG<Dp?h%5DTPY=(Gt}XQmJ$IRp53A=Zj~pr%&n#
z49{1kTE8@o!}}spRkn&#dFk2Ls(f8thyVuHL?W(2b4n(W@%BS-@!&n4oXdFL9Ljy$
zQ&A09U)u%M-KjKiX5iTOs9VUDYNvS5bmjBG^4Y<iTG}RoYWe)RebIgv@pA(`NL-WD
zs_hjdIyx~gWW{UhPO4C6S?Y`6L^;b)5=Pl(T;{-9A+FfEXvnDQXi|`Gs&yy-oAN19
zyJ-a#|8(v+sNK_+LuSg}TmiF0o!uJ5>Y%9{vhA_TFtaoz@E*6>^n-k3^Tl10JnrKY
z$T(Lru0zL+=AkZt+)X(Zfy)z=T}zLy;3Z{6)SJPt&p*EZPjqi$a&=&(W<V`abV$Dg
zArIov=HOFCXK=WRV|t?G<H23g-5a3qQ~p)p<<wfEBBw*cN<yY*Dj{f<_smu(vSE7z
z;7wO>{f2qB)gN%4cXfi-?CsMH?yInvd{~wY-WYyjqbTxkq`{*)(kpkU>e?Hkj{Ybm
zQHhPKkhyUQnU`2-km}4%j~Te1VpES7L93(!2TbpW;ryfurB*^%vSWoOY>S~LN*q27
zhRHOu`(fXAFoT7LI1QO!om+E7hm_K0qb8T{D}yKeXI<*f{%mqXu;Ni}AG!HBx@4dl
zr0fp_gJ-ao4<+X(I9$XCYULX&*04U1X!;J8ui^1Dab$d2?14-QIf%6tkP8(Qtt2a0
zdGPj@H>{l3pUNcFXjOfA@5%{8`veA`i{-z3Lw?H<%A}HN_~_fu(_=~z2!nq&B+n9}
zB*lzJ)^}tp6J2kDc+NFy-dK-LSkwzePWq?Wk9^U?eDhRlxGCGClb4}9*tHdB)DzcY
zt&tO|Zur_qY^6b*nO#J|n+7|_yig_bkod~3#~kQ+60RFILlRc$I;Iv%qd%&1b8xWx
zf&8PB$x0$T9pCdZ=vqVw9@vL0f>xfB%^(|(RpiZfM=6{svkyV@df{{a_m^7LYu1Dz
z+vH?|147bBusI_oCCH7DxPkqpd~-dTIe~52$nf^d^Jvq)(DtW(I`L`G-&+8FLF+qo
z5U|?R;ggR(ZW0K3)(?fCd4UGhMw|FVjomCtvehhE@<Wh_?wFL?*jq6|Ebevm#tjDd
zpedxUo55UQANQ4X&Od}F*p;(JFohRAiw2bV=a3eYg)k2syVQ8m&xamA&G0;OF7L#X
z00uFQqe9W&G<^<+{B<wHWoi7po38z!7*ne1QpL5)%dN;J>&sAgc?p8`lOM2zSd#oH
z;~TM-aeWj{#d8{F-z$@)tRKiQOAs%6K73+2i~v+S;2L+&tXuFd#wkJZn*3k_G*)It
z(>;05m-_7{XN8Mc3qQsWqupuR@~Sv`#Z<~@X`-B;tMsEG4Y|gnvbQ@Xh#Bl*nfnjh
zBBviF_D>TmMmY%>8#NGDucOFS$K1acSDM0)?zzF`g*2-Z;XZTKMbPF2U_wlnE9oh7
zoSVlXVfWhX1eyd3d9ku7W=qLoX2Kmxy3?olJa~B#yQyj=!%NyQG>$<h(!=r_{Q}06
zm6<b!acd73^^vbdrRC&@Q0gc3@`jBIR=?J9jW77Y=aq?a91uOEm_&2&HADovSQlPI
zimqwbTB^$Gz1e7`^#?q;rsljX4S2GdcFos9?#MF)bAAy-CkzFkoRDdj#Y3(_*bulR
zT?ZKn?bMF~Xa%c1v=@|705Lz#)f>%r%&MGO>_>68o!E9EE`#*{Ic8Z>!17rx5-y)b
z<Ode-8z=dS5O6`en9{2G#*cx4y}ONX<L@9H0x=|bNA+>vI<?RQpFR{*A=YwZQarmP
z29I+B%qJ=v+<g*d?eFZU5bxz!DmaJR1b0@hE3Khwsv!E&lu=pki~Xvs(8Qh`DFI2j
zr`Z8rzDp#Lw9cd4*0|mCkN7Z|g+ouct*98LC!a@`6NUQCX@>RnSbI6~i%;;{pt~uy
zS7Z_rlF5Aomg_&ah)F5o)5>Ezr3qi!pX^kmA#J5L={cbHfmT}wolV0Vop*l#tsnV2
z)U~vs8cPRa$o%E>h`^;X^I)<S18Rv3%tX-k3R@-jyeT6-8YN7LYO!CrB8RPqY3(UM
zU4f29Ocyhg>n5cSw#IC*9y3v}pjeU@`}<4=NXxnVVG6f7gk;>mHS|uB=NTDG{+a3@
zCfFM?j^{z<>Dk%w{I0Ir;4Z7~c!5nGd8=8VRMn>yzZq`DCaxc7R89j1T{{%-4i|gq
zeOK)>37y))!GTCe*am*0bQi*8?~E<YKP?!im4DTLYpzj~Z1ygtMh<!Nog648SF9LW
zao#5o1g^59S)hOi>=D>O3+9GD%+2ka69YY2>_IEqb?QRZlvuS)Vx^4IAp`f5Qqz)&
z;nO^o38JQ^4_ppzeR$1T{;u6+!U`p%1|l@lXLwceCr*>*G!!$pvES<bVk1E4(;ilS
zIfG~dlQWWHt=M32K0NQ%sI?yFak8_H=vO_SWZy0yvSj{s4DJ2QdnPUz7a^5FDRz>E
zlKUXzw4Vu1#)Yyr0y0-uSZ@+9v|Elk%bLb!6tCE$$P-p7?O)e={u=BiDBZj?jseH>
zOSq2;P0x;sbp^i{UvMwV{1`$UaGs?+kgcA+7L4P7D$6}wudauAFpuIAlVOkB9qUKM
zRUsgmEXT+!55azw1rw}ms1ce%p-8(@++3F$wVCrl<CaRFT|zzZ(Qjh`a$C`vqhdNA
z0}^>Q8bo}#1V#0oWxT;1k=-3mmQL0ziwUg{5FPDQQ<Bi6I-S1opg7l2&fg{pPcSxv
z%xK`DjTZUKc}?gi<T2@s`{O|kT*Iw%k7i3HE|}hf9u1BC2w%GrUa<_aKR;;g6WO{U
zxWV8<H*9HhtJ;#5gLUqdD2co^BGri73&r9d9W2xh_P%gPgqFeSIIT`Kw=qxE<Lb=9
z>%Bwx>Ww%z$D!z5i7^t^%q_>I9g;-S`=~Fzq}3;>2{O-4DKhR3t=@_oWbZSH4&1&4
zg>*ST4;pL)dbY=O>(m)SUCE;l_v*{@yCZlmqezQ}PKySZtxydDs)gaHkUfELfZ*8X
zX4I>aM(T^yU20N7fu#As7HpcY?LknYoTMVrw!j=ihR_T_ll*TI*+N|<Dn0ea%vF=&
z2z25hn6aI<t#3;~xYu~AI<arqJLvhP!IKC8!^gvx-&U3&=8!IMZ_NM)#zEc~#$Aft
z(D5p2$mPCGl0Ppea*p}_2BkecmB6KavDnMo9mcAj6E*N&qI)WBdnQW`UHySvakIS<
z=5CMTTn9(_+Dx_7*;OkJybMMa@eN>CD3F3F_dWR7TmTG+YpBdOk(&2*9oUpm;21TW
z$`tltL^5!pMtZpfv(ksal#>MjS-!+1<^_tV&K8`bbH-t+MjT?!nw|1Alwx%SeI`r`
zj#K7~oA-+cQ|6_5QnJ$sa-cK6P~f!m$Tg*W9d1|vMn5Ed6R1G+D#8`C&v;aQz4XJ@
zmYa3&0jziQbU1_t=J=!2B$XBHK^(%tiINQ|v)Ph5sDQ%0C`rmE_j!*_o-x~A*i(_;
zoiC?y$LbrytlN*0%`|E&?w5_P(m<JZ^B$6^MV3@Yt#iymoO7~BIjKD)zXfpEghcg6
zMfbUX{d=6ULS243p@sG{SJlLOM0>|1-s%ot#%aVgBcZMh-0(HpiymJkX-+S*#}z(a
z%AEt4te|PD9HoApQ7`1k#(E9isZ?z<fg%}Y9W-MbS6d~upjY!CRCN8aTucSm#ZJ|1
zn+RYPLfc3$%j-a_OrqF}89oPCRntBn-5X~_ayX<^38nfiKS%22Ng|WdsR+r%p+0Db
zwWac?4{MV`8lrUjV_gu2eU^^QKXNlU3G-K5x0@~7Ln|C>2+^qU3`+$@X)K^PhAe`Z
z4ed9h761+5`Vw`t@l^wd-*SpE`r!tUQN!<Mj@!1|eDOOAmKVM%$S-B<M|wVe_kli2
zUFvIu2VY}P=S{E?5bf&-t`%ZdUmThUd1WW{X&v`SfU@@M{OC9G24e88MRG3DKkv56
z`3ks%wC3B$6>Ot8tK;JbcHz3T?9AY!w&puM^LOj-WC_SO>fUnfV&anPO8|RM8Xyc|
zjyh<A>;+lw+_>YTh`@L=no|8w)kMx@)K7L3D_Xg^MpveK=l;CgYerSQ=e5z(UqAaN
zG)Lw%C65WZSmNF3MtP0$@lPIm<9>;t^v(9S`=$y+?HC4xidq&Q2aY9B3#CF7KcSR0
z9Sygf^wT4X38=C)0*4TI#_~XktLbs~aA?8`OPWl<q;j~lUb*w_s{1JRsmC*9E1+lr
zeho5sVVwvTK+v;OO?V%hyp)?8EE=LBOjL6t9fo^ej-R|7^AXHyzwrxVopsHTwl<c`
ziwjP%F@hKoM3uOF%q75WDd7jW=0O?X{F$}PEkgv0SOCobZrhh#3ji9P1?vU#Vw8mj
z**cBmyj<lq2vGO8raEs?rMB08*Q$SE$#Nmx&p4GnUMbQ@b8JOWQ46(-Qp*icKZc1V
zooWup7{~m8a$9J5-!yFmSA8<s5o!JaFzNxwcglAK>vR%@WTW%@p~lK3rZ}k~HS>TQ
zGc=qvi&?;x^)dJMAMKNgS7wW0vdCi)b<+7UQzBJDW8k*1Q2C_1T(SDUX&|4ulB!Rk
z&(8Y&oorX>kfe0|jPkLg<BK1_Fs<oC-#OR@lqviUK_35RL8ieslxKH6l-Cj8=EdAo
zJ)<XM#U#0pyAw_eRlkulFUu<xdP`!o?eB-d(inQ=2@K@D4N=xuMfvBRdxFd3)|v=5
zKdd0%BL+j@7q*0W4;cIA4n;T%cghnbX`YR{KA?rMuiz2;5WV&!6f_yVJmH_O&mt8y
zD!MuG4elR8$Iyr--aFP#-b>_sgL;yEHJx*qNqyYbAUY1L4w?wG<U$y6*sAs~pm{jW
zWrZ()>m9+W(sU(B1mHTa0f^RYNV2m*v>h~@*rDZDz)45)>FMO2tmNESNp7dMxKCYs
zxMP{`2H3j}4#GK}1^-4oxiac2RICYH(Q+WmvE6mezn^X-Cva!lS5ys%)npVWd^N<l
zdFuG-@cyN&o>E+F$2U-R+nU$e5!&SkOMYU7oowP`V^yXWWr-AK;*v9T(Og<?hVc^k
zi+YXVq>x#x(K6cP@Xfexx@`4{ZnA3Bqi)rH#82k>&w!vL*0e=;Jt&C*V!vhjz|X)Y
zXw=c(jz&M|KhEnhXq$H8>>;S=kxa&=q(lR`3kJkTM{jM`O1o}9YXAkzPC4T<MYjds
z^uM&*vofudX^3t&wc?izj60q%#qx~|_Sle^K-dgk3!xXK48YGQeXSgbtdXx1G!HkT
z<t@8UnPpmxY%RcFfvtxt!rbKGV?T*DG5TMW_Qu949swLK`+PcwL+QnG|8S*)d{!E1
zr+kAd4?4?NzD#)}O=$R)g_;DKktRfS#bIUs0V4DBlLZPtl0<Ibwo&arL_BI-UYR;y
zye0Ya2piWmyInlEO0;+VWDo6^^A8QgwZlCIiYNGPlB5RTdd_v1ST>zznvLhC?)P!U
zUotuRUKh8x`0JxVL1p>zGqUD^TEE(hvD5J8^M;vTaRr8xldX<ila-&}MuhLI3fQDQ
z^iZzuF_)0Yv~dm&>^p;zzxJ}}^S`wf1=!io-F5YYicr7S%wHuY0<mo&_xN)wGD}oE
zG;vj$6Av@%(@gEXuZBXrluZz624tX-vAIAt|FdT})c$C0_I_P{eLy<;PESCyz?G}-
z^&~y2pAA$1Z&M`ZoQ0KG$fB{o_NA{Vc4O|7n8Pnqxq3^kC70BI;AA;V8`ZcZhH|_^
zmv}KKE{+*;Z;Jo2g6HR>t3RK==nu53q(K{Gj-R#PwPIgS3ti)r%M>mh?9M2*Aosjf
z&_z^cC}RlI6)U>quCocYrHr7sedT|lk{-zr049rU;c*v$Qc79ialpWdA}nSR8dp7U
z>`QsvW_|URn6j$%!%wHqehYXP{iLW!cMYsiCy-AOox$a?g}>A194ukg;hEWU*Z#*)
ziGQMkSIR;5tKL<373@*|f@Z`AhBRNPAogw3%LruSjUv{J671)Bz&HL?NOSk984#DY
zoS|^Z%6KC_1bVU67W`>M%XNg6?_xZfo3MGLm4VFlhLEI<@$$VY{)4Z%kX1@>?P@o9
zgR}@PzS(6cO3G^R-{4iQjvGnZ0-4(YI90@f=R+93i<n$~yA`V*x-lMW<M8dS%l9UW
z;1BXuU(`_BS|zb*^GdZlrK2VDaK-)_wrLixaI^EHRX1wKH1%ZwZKD)sUHFNdu4z1P
zBY7o^C3#ZHM8<T9)p+v8V6NyGN}i<wo6ybRAV|X)T9;NO*PpslzU7<fTVH)#SK&PW
zXF0H-#=cC6_rM0m`&%YGI5tYMz+_~7Y=6djbOFSFnT0Ekqx7TLN3HCK?OmeS1h3z<
z!6)v?KkMa(R%r%<JIz7|Rw81qS8JUUbO81E_0TjQx_+h%7JUqZ=oKI^prug?asw_X
z?mlUn`kkS?AaU&ambIg<6kspgDUk@07<uvg@$hs;7kR7zQi+y?V9ym-cYaM$Ep|H`
zeQW}0_uJgq_~4Qm8}R63$>VI<Xv*9@+x$L+o4}c*q4abw!{GaoNAb=QELuL{xzy?B
z64K(Y+xF(h-(k<aE3!JwE_G~h0jvo~EMNC_Y(Rs;I1I%Ql!j%e3?#~a%CzZmUeHa$
zov3~<tlwM&+#l1`9HWQ^%8l_$7A-X!9BX1j=o3bIq2q;x2V*FE)MXj@tuKW}8!u!3
zB{$Mes)HrvuJn<03F2&d5{VeF$#a~azHx>3?7aicf#Pk%`;hpEui)XhL|3tYX_!<s
z))iVw@G5p!HUany*#;<DF?ZHd<@czr6+xi0U7=Us^|-3sEiUY<9fN>wa&O7AZ|hLE
zZ!9+0iT;Bl<ICRzlH%1><cuhWDPDya_r|YeW<_@rWcf};SH*PYz8^r{6cTyGcd^q~
zE=;)<An5z-0>^75k-f=EbKDr6a_fY}ng9$)o;9w`@X{|ox))~s=pT*e_uIV$E7y)*
zM5$HM`|Gv=m!8x?$iOCa(LB;W27kLR=<N%i5;*PN`=1Ai3ghM*P}$oyPNDv>xI{eN
z`lHa4TZN(wqc9V}Kwj6!EiYLvOM3X32#n;nN521Ikwhd$(Yt4_v577CFn%-XP%XSD
z%aNKAAug&gd&pDH-N$j=8z2$X$Z9wEw|*33U`zj(s<`#Ga;bFiPn1*SExUADtR|)-
zWb^9xV{Xa$I(r|GO#q}bU3D549dC*VE)0ZvS$gc`lYCRI>kC&ZgSd-MtHN3sFA8w4
zBG-fNv3a;%TijAHhyZW`38MVa%A`9KDbjGa^R-xM?zKcIic?Qu)xybA<D{ur(6*^j
zy<&xq5}dX;8@jwdp1J;is6|7<3_E?r3rrppyuqxLg6h{ydq$>;T(M3a7H}k1K&!mu
z)G{*-%RI-fZpc=nF~R}+9g%4ZFP1zMn>Tv11F)aa@2~EE0(MzP2`&L@<-3R;^i}cF
zA+5kNs~nUT_NDirv%1mS@)^#U17|f*fHkwqQoN70e+)#Vv}onr=7K}u0KL@I@R)=z
z9jq2-B|fBg0CLF6V|?znOX<gKc#BUqxKoI}NSFw9nCOq_3z#Qr)C3?riL=EK{YA`G
zT4t4kP==q7ZE2Rb8ev6bgU9Yd)6VCb&()<@-@)&^2aIzmkQPtQ>!OK9WC%FP1pPwS
zT<~#!|9zuAopp}OQZ;c#Is?w8^jx{Sq_``vtc<+e4#av!U+<`tl&hE;u#*6gBaVvc
zj>16MD@a>0$zQ?~Rq}I>uE_s}f2b@f-yTrwt}Q~nAV0AjtNxI0t2U&lz#j&FlZM*F
z<4wtWcn&3vudyszr{IaA-_x7&4Df>57B4~e_k+pLuCo5UjT7cB=qeDP2JNGI3J!L(
zo&-4!4V!_}V@f>H6=GUL29+ZAva+{D<rKuF*(-l0`R%w!K2|<q#X>CBUcPA7NZ;rP
z@2;q0otr&kt!1Fn-%a7G@BH<Ek127$;`utH55spJgin-{R%F<fTD+v9Xb`o{iw}Bp
zbP;%q%&XyorA1>Ysp0k!<PevThY#0rvkKDC8|`{TBNkv9mv{_LG*-r)gf?rM$Xd&+
zU)K~h9e5DM2GATvIPq*SG8U@<DjzKUxmPJqyVO)W2U#tmODz=v1>JGK3}U$_WAMel
z8irZxHsN>o*oZVPb1l=JgTS8oLM+xi!kB#jaa?*Bl+ukAy5W^XfJl)HT`0DavJi<|
zHLyJH$>a4Iddk)cIQZ_Wo7*Zy+D!^$gE50b4gWK7^Ko-O;#W253`pd)8F1Dg%GZlJ
zdrjj`dcxv^$lizPpBJ4qyunKNz7hk2<w)(knM+mXh$*_V^7HW0p+wofvR;JNXX9mR
zj{5v_G|IScawDZ`B)Tp>HT!~V{%_CU$oAJ<XpyX4(k;F4>8pzp&pA<TPCTr5_mjv&
zPi{cQr@rbPEB7X&;oz2ORC<{|@w;CwXrjLk`DnA2zUmCgHxENmhsjJQkYXM@N-9rR
zuF#d4?k`~X95%n8${;<D;-B4D)|n8AWPOAfl&f~iVDoaZ)q@O{X~We)MWKvC`ywHh
zP0lAKb7Tor<~<5uOVj@-(#?VxomkC8z^V?TVlY_FZeqEJ9^<qcR%zvWT46`BtG-;|
z2vux)f}~CKE10W7Mop+lnu<UkG*4*C8qI!`EM6r_f+LcAtA9ugz;y=@FREH-!2RPB
zq$$+8+b_p5Ig|Ra-;i*toKi@A%yJbv0M!gjSy>(|SN<Ww-b#UZU8U`vd(U43;4r1X
zFO55PzvdTmR{~7Wjx4nY>k7^V(K81_V12mRsL<~N@P+r@o~eCJG$pxO^4I&tvr{@;
zeyAd3R`4R#gq2MK@2ja@>?#^}>5ICR<$Fmj*GAF<6{FY)W3VbFt;kkuRf)8T7-0yw
z!lcNQ0oNv&XMs2Z_59AVs0Emq8OEUv^0#A#2nN!fpBFN#*=IcPnnh58g8Etc>dxz*
z;Vkdq0*9)!&42X5|Nqbns?pX9;+)Q%*`GDO99{iyug|HM#6KPVK48AYGPX+P>(#IE
zPSauZk(6!w$z)*O!?;cgXLt4ktud%E3~Mxf`N8(9ZKaqZWSmMgt|ykx6nbEmLqd7@
z!wOO8d{dffXG(5TI@m;Mlg$+^Z>+>!=2;_B>!;`z)wlNfZ_CvcYGL_=mV=MEwrqy=
z&K%!OHth?w5;DdA5e%IYKYKpKjDnoOPVKvuzOFrucjmZD`U4g)pHlFMOOzRUewQT!
z_namHp$?iVDzBHH7~`GiN*Lx4(Y0bFJY!nrDqIP5oo+Vj%gGkV364i4qa>K&`{yUa
zRL)?Owu}>|)EMc@%g>$v#O0WqyXM_HlMJ_fnse6GL19wJ-J)+WSgc^?*$m8_<QO>^
z7yh{Aw>Kt|f8`aYSMmKSsv_CTUe$fV8L>8=jTgI=gpoaAY$eCw6-_UUsaDQEccC;2
zgfRGjTmJ?-HESZ<HCj!B=q-_(N!vsQ>|4Qq?7}`V-%RZqE=rnn<G=JMgNYC;8+=Zs
zwnGkSkk2?Lt!Vt|wyt?FY?H^qm6s*4us}7$tTGB36ifWGG8G|IkBTQ3pN<Rf94@uT
zIaJZEe;2kd8P9A`04|Bx14|dgDlE#zEfth<-|nVp4FPi=dipNA{{a(h`y?e!AH3Nu
z1&)4_B7c?Vwaw{BFNj&Evg|9r^u3mosm>ISD%10$EX9fZbAdjZ+osaziSA&@)c#x(
z)|2!mTxp$XVr^N;Yr-id=_FRHLF14U?=TWaGET(WBljz*8(1H$xv(saO3k#v2P91G
zrIjkKojq=aRcystoy%DetdXplxd3$GDV`57q4hZ@Yb8I-IKRD$a2<?dmm<0s9Oh~s
z7974Fa{%PFP5*$T=kYqmWSeMI?PFQw!eq!GA^Y_`skM>_S@w;}y`&K-(V6vAXDRxE
zV(i!b;`3oe{v&i;_C0aYduPPdE;Cu^B4JEby)A~odMH1AntP})>gV@E_mmLC9WFp9
zUv-SsuvV2fE_Y)E`Prnlw9iM$CIPQK7Xg{2E@66OewlJT;4RM3GJyQXu$W!l2ef{7
z<L%CeN<Q(s!uppmCLSI%4ZdKfd^I1T>Mh-FpGzoJnPO)ArG4EkWNnxJFW^4f{-p_X
zqca5%p1h<Vv+lmO61kR;Q^MI#@JDS0dgP8T)eBVpkkWK-nZVz;l9qMh@)P1Hb&lTP
zU4J08A#Fgv%Y71|F`-D3ExVqwb<q>osPGjLxDRs=KK`E2LRQ@*W6wHzi+QMUFAZuZ
z|C~@IBQgE&k<u#y0HNxMcl~XzJkKX{&vs4K6lkqA*#4B6zRAA0{&7&IS{8^~Y7e~<
z#L`&cr#n%M#ea+5LapR(Fwzbmp1jvtY5H45-_3I8_njLaKA4;gNd2Nc^LOwn*0Nm8
zuKL8=A(5ke%&|Xu-89~;3?*9Yx=6~;J!bhUh|(Q6f3m`@wYLo_c${|(JzcL!-89}X
zgEnS=K3R!osRFmG;<;_T<D^m%3o=i7y9_4PtmM!1rWu|BC(aV0R|LfEs2^w${d8xh
z2JgicCs)yqSoTCq<dD5<2Z0ZtQ~b+0^v7R*aw&Gj^AoWHx6?VD7LlV*AEtRtdZIe`
z*eD1YMZ~QME7JsjAkUmef05<m0c8l!`&1RT)+XA`<aP1c>rnKg!rqbUv_iWFaUR!Z
zW3gk(d@OSDS^?%&-I?XO$dXzp8q*ZimCp0ol`v!el(|?e7CfH69l0ttG_qESn@Qh6
zImmmsb{cj5Q|4?K`5(LpG4={HVL>0Ao;#Yzfo`n(<Y`N~mvl{q10FjeF<G(^z`_?!
zWAvn@pty_*h)}(!bT0nn%@7v-Oj9C0`j3vFN0>>YNFEKb?-EBpQhQk;%b`aF8HchN
zWG9CSeiFS;m}`rR=?4E>(O5Jw`PnOeqtubkFBx(6d|-xxWtal@=wtm1BZi|vNPy7c
zhE4<H739;g$c5V(iv7a~%KU*))8m}1@bM?c)%8|pSvL4>P1{(*K+k=ORM6IXxci%J
zuz%dX^Gs4<Lo-%f;hvF6aNyWs`WLulbq-73FzgTp@X-<B!$>sEi2`0E?R*ovza{wv
zsO6Ycxs5L_?dSFO`qA<Karf47Z9ZM%XrRy*FJ35x7A@{lBv6XG1q)KVxE0q54#i2)
z6ennlJG8h{+}(piumpP3_uYN>yZia=-giHD@BMu4K7S-n=9xKjW{y0QGv~~C{ZrqW
z0BFs0Kb}>^f8MOpYOMEMf9ONSJ4UqeA`!iJ$LYMLo&3cib2C<m_8eG0NL|hAyhPM=
zch-X7f4S^;cN5Tg*|A2;Bt(1E*n0UQP;l&B8Zw>E-Cg#`_3eP_jsEELcF;;2&D62!
zdF<)(h$M@DlO!Jsy8Ii!art=(%MS_NN+#^l(_KPPf(>^a^<2tnJD1?s2xeW_p>bQI
zU%!MOk<Y$gfU11et=HlT+~hX^?pRaR#?GCpFyN$<y-0>?GQ%$XG+Lpk(*RNNGe|8H
z1}Dqd*^?C(kj*CIPh%G+ZSem}w^U5nI$;mZ5LF(RORDanjpxtXcQojBCwr{z^5Vpv
zDELz071Y03vDZcH;UnTFJl6(ZGE?$4IOsfClSKF$Jmbt}qqc4n-KvdGw!j>*GSNFP
zE`I94Tdu#;OOQ<6F^N;WaeJMFnwWA#hVjfiu`+LQnk-PvnV)=;9r>}6Rj0uX<*V|d
zkkDU%lw0}M*s{6|j7pMi5BwbS?zTXcNMQn&#y3XbcfVJcc~xC!)0UVi71zxTi_R;A
zmQNn?xr8__sKlwbM$j|=OvL_+B>tNID8uyI(`1pc(Z%~%ib=kWWWr*`gHE|~DSCnV
zja)YMZwU2W@Ysch%!dZHD8V8i`vvmJg~>F|ZxglI6YRs<D`OCy0DdNAmyKem>c>Yz
z>`J09hElfjtb@YxVmNl`lo7VnhY2ShdXFb}tGC81ZT5~nT%AsA4GyXaX;HU-#BNDY
z8B&px(Jh8|1$v@=EQ*<*=9o2jS0;(RYY{ROxB3!w-vI5c_Ef*uo_Oxi4eseOvyd`n
zhgJu)mBfk7qf0!z4|pD9n44$rD%U+X6hMY<#cF!`gY7akOF*ldb2_5Ya*>R_$(|Lw
zl|n-S0+)kRwgL(MP<i05H0NJW32g}8&MQ8SdZ<yb7Z2AK=vPtHRhBCrL~rw|s7o*m
z$OSo4=w>ou^naGyQ+{2g=~RLeEF}+drI}|bNm67>T6pQH)Dg=w6L~Sv5kbNJoVir4
zpu}7*^1T1juL}76eX8FNhiU`d4*v%oO+ok$R#SxwO0{LRd39={i<(8zI&93TZdPLJ
zBx&yn2%*Z2wx5?GIEhRN#wZd_mXb?$Bi}`!|CLug%8RbyWvkJX_8(JF0tzVFf`5cy
zXJ$|i%Da<lU53MpH=W0GHFiaO)lY9YKH<Jg>-G=wg2*!h=ja9LNdFiwHY|wU@A~P;
zx?3`5nFk%O`AJVYxby=eD9>*Kgp&Do>MbS~V7-OE*00L=7$T#zB1dEiUy`cC3&<%V
zEi_MP{eoKx&M$LNV)t;~{st&LhTxes*rvjI$s#Qc<oaKyW=71bcB7Z3k+<F_LEuA#
z^1h5DZN+<@5MK`jb=Q44^iEpbYOA%FIfpzSsfSq&=sR_ag_}w(UEq&qvY(}`W{4-a
zE+)M>t6NX!M?jXEj-vOR>y!&6z{M(MlSZFY)%(1YMN2eNUmk2H$)Ums7?V;NTbw_;
zZt;wnFlBX%9JWbBe3wv7a4CcfvY_o26EAT@NJtzb(S8=ZS|)0SSj}R*k|m!ez-5y`
zZpjJ-qZ?-0tb8tcjL$gaUnK{BR&bdTknfHxo6_T4CO`6B)Dz}fFRQPq(xcZu3fH4x
zU%N1HS-b57*YvyBG0hvrfQ)s^H<{zrziR7oY8J=O@<+($Yr_tl_+lQ1|6&KBtrmqJ
zZi@r5&%Huzy@j!MP07p#xn&*3S>GA@8wmCF33+;2=L^JvDYQ6b6wbyIwYjJK+DfAl
z%1rU%fv{W@m0?z_cCpzm&g^x5XcehuI=Q*5zD6o0bpx0fcJo|1w6l_B*Uni?IrOn{
z*bKKBSl=`xrHre;11U<HG;cnY`+P$C#i%ia-aS5t`E^CJYPnGZH&kJ7Bw+JoT$)>B
zZNWH#cZs2RZQoQcy<$yiqr<v*GGf!BV7q|oK3S7F+Up{;%k9-iTxz@Tu;T7b{}IMQ
z?L1o&t@AyH>X$F6<TzgXyj6e^l9CVm)}W0&vqsxuf<}{SxofloEzp@Nf8)oA;j{W3
z?2R`qY^NocszabCJ&;_Tep-%E@ux0+U8~#x1TpncZMs@>!s|s5i@EhV<@;tga^yd6
zOuLedBwi(aRLdQ+Tvrs*m6pRa#IMGRV7|-~d0T7u^3a6MyZP?3i>#WPb+%;a?wNL#
zt1+#$IlFnO!Y|YGEQCT)+CXb>xEEhfC8xMx6{nHJ_ZJ$D>1;i9FpB&_r}2a}+jDl$
z_ZX2@w23&}B5+ug`g^MIQqs@$m9}{aDY+9Pzq~)R>34zVx|ur#A0q57p23Ev>;)C`
z?FJ{<l3BwO-N&nNyT3<KTU&D|{30*42OWXrax8r;Gcsy`T845;cubp&E7B>+?+JH$
zd7klpTa`xpT7aVM4gYwl*cHkjwj_Z?E7A)#Tgn&mSFJqZ&FqNuf~|732m_tDp$-=v
z%gGXuft}O0c^kG|EeNyG$JcLPHJ%Q=P<p56cNcRIdms@8BP+(HRGDebl+PZwv?WXP
zZ?+|S&!AG!|1%*jG5B+WdjfuG>0^!Oxf$CGkLMfo_&c`V*#>VjTeC2KL)$C{C%mip
ziL+)JPb_-EcfRL{qHR4nGVU30w9&#L+-vybtK7MBQO+8v!k+w5$2nEe5SaowrAR^1
zc*W2`T!)7nHJp~6BCfNaf3ezW3e9;Xp{K1pnflX#Nwy>irXrJ3G_lQ$jzn>F_-<Wj
zW6^4DXU<kbJsCkOt`Ki(DKmw3hT#(*!+NYW`evZr{Po8_MhvsME3==UIxU5SF2TGp
za{1HdkFSdy!e>i#i`0sB+yXw0?UA!4*pY6Ek<>c%rSfj}CO;pV>(?F=M@G5&5>uSU
zO4F~1ak3wj9B4Cfa<i}(?AzQo)WTx5_2!6u4#UNI=ooUJ)S14dgz$-B_aew4!bAM&
z4c9M}UMr&tyNdgo1lsH5d--+?XxI&XK2!||q|=*tY}Wq?Au8j!7(&k;2aV3TRqd@1
za1i+!9v$oI08`Bhm2dR#(O>O}!4-e`#g(QnTr-`PM(2*X+hUBtoar#F_gtQ7+>X|B
zOK|X=Fg-iXq(xd2d2w2CW`FAA)7<^o**d8bEqJxV*HsB%g>2E!60O4TB-QFaVcM28
zeE7Ahpk&6(7}k_5hlymC0PJFwdcwgGzqd$joxFnG@i~WXZ9eVODiZC4UknPK?j%(j
zP)u_yWrdtQ*%0F6MJw4!w^>bf?2uf13``D@(Qh?;H2S<$6r^)AbAu5Cy%!RaWpmT?
zZ0@hyAe=SQ&5i>fel_gImk3Qx$`sE-2x!f9T-mBcH_-deb`Cj?3I;uym^$GV2^1me
zYDEcoS)s<x`5-nj(R`sAbao;JQ0_?IBx~X~auR_|Y3QFVXFJ>mnu#cG{`)j<?Y0^O
zXCSgFu|i)UABDMTb*-ORC=dRK;-)1%Dxtu)7DXTVU754j^4VP%?Mz@a95tKH=IOP%
zxoT&6p{sq|xX@f`pAzD2s7C<gG~?gV6RgsSGW<IA1~%6eQ((nag3CD5?^^dro#&$=
zS*BJ%NIB9Bsk&yzWJ54r-4|7y#PLe0NRwTs@5>lFr)J({K7)der8RMaid%bn;+Gh1
za_X%HvrdnUC@XOxZMk|1X*)w-b^P)Ct&Oc`L;XD#=I|frgUe?|=_^k_f~#E?Ci6M@
zzqlg3bbkZBmF{a*I%SA-J_*F0v{;%D4XN{(_mHV?^l%FSeXg&eww$P3=PDMcKjbV_
ztOFb5%Qxl#8AfNEiKlr&0^=ijCY8)%!(Wi&*p<<iyeL&H4v<!3uuvLm*Q?oA_-0;H
zVczgrImejnba$6kDJf&{)A@aszTsE?&4bO#7p`AB?P?i?XEi;HCUwAUoOMvbcKqbc
z*X*S5WQUeF%rvX2)*IoAq76p1FWx$}XcknTxoIl<HVF15o9BCvbM%9FVKMD)V2*w_
zQK)Sl={^CY?Y4@qDWtG(ms?>b_i^vox~Mf&y+nuR=3VLd%#K=mn0oLGiT%r2iDNCt
z(WiVf-2Tj63eN6&Cp%x{@mK4*qoi|swao_wpLfS_W0MFdye--Y{hn13H4CSIX_osd
zazmcoBcAL4fOx(k_5Vsl)_)F^m;<r^fjjE~H2zu#i!mxs4-r<83mMp@@dD3#9HIS@
zfh*8BH~;p()YFNr<(BMxgZK9`eDw`KtAK`t=%FnzUsAM1*)jsZXRv{3q<T;29WIub
z&Y`$ea$tS9hyLVE?xa@ad167fx^&1`^@oHU>5*)%b#%b=+zqTSlXRy<*1ewgHNn8!
zT5L9~8f)DKUaeStC6gy1Vbti%O17188&d)ziUJ#9ymXp--x3XA<-X$&c9zgd<!N4p
zKL=T;16!}dfQF1E&-lPgESaK3Gi5Qv=6fYg0w4>nh_ZwYtKfGl7y5r0%KwVn2Tx7Q
zxsh~7QvW)dttWu3Cm4)!{^7DZs>Kh<NaVBOgR;u`wV3bJotk}i3!owpFFd!Fqx1VB
z?p@Uo0zN_a)vkOWvkL~($qdDOBadQBKU{KAy>M;E#J7dLS?(J><38%oIkL+(*&|Yk
zAC1|ZtK4V=;4w6;b)e2?^g^!R%+<wYQ}?%6=}$(g?ox2rMS#qI#Atw{lY~vIdi!Hp
zbD5m&l=&4Q0B{b<`;<1!Gx%Sz{U7<p_?5A>Rem@i9?ceOZ!p1L`W+Lr*CpG(DE*G4
zfPSiAbgSz`orF&~BW=^5dtq1I<KdgSK$2*TiQXnvUd}ZJFPt(=L5loedUhK|HRG%B
zLc+T@qVVSJl(ACz)pC-##ovJS4~6p`p_v)&L(>UcwObJ*v0&W0KHcXm2j9|W@48oD
z?q_k-52_oHT;8fq5DCeyoA)Q&JeVKnEP^QV^i>_{fGl7bk_tQ7u?C2@doDdMW=y)G
zeW7x6^h2WJ>JWDJ-ov}0HR!wgtWaAJHbvWdaS<N#w(koh^Yqb5&NRvcbdO;_-4yyE
z%JgJX1B1?T>JF%{ZGSB>Tl|>l38I%`9DHdS5~ACaxENVDd~Z$1PnCN}#0>OJzT#j=
zyCkbQpdL_D1=@5dPnmB3GO<)o-R}J?u{h~ckGPR?ogb)J_yeH=w=7oBy6`NR4ar=$
z1w|Cq5~&5P4QX!x0i1rb#v3Wu31C&{4hdjBMMF>>NIe5n5ad}A)0(Ldt+WlK+C{LU
zvkuWr0$l`4vWhvc%Z+^=`KeSam9o|1_30PW^>ac##3%=mws4=yup*fuyY4@mEcy;x
z)J?hPaqJx_)Yumt(AlO^mlyRx4wpfZ2=!GwtCd#i<;|e<%NG&_`XQ4co((~@lj_iI
zJZ|hG9VDJ>qqo&Ml8km-MMiPY8)TOnFw%(7l2mIW_USx=Ps){p1JGw=r<sm*$yQH4
z@;xq|<T~=SYZD<bYK!3hxqC7Qt+WrP`_KeFY0<utQf47T+>ld@D_{Kv3>VLwf%ee(
zrm87gf6fZCRk#Q4sL%4ZwPEXytos%ohg2S$GEukvL5*cUZ8_*el*P%A_LcPG$`e!d
zY^LfdB>(uHwnN~{V*eE#yW)9+0ATP%+Onl@fRbZukD&T0Kt~5`&s{hgTj>~Zhx_+3
z431{K@f|HR@eY~wwH<8lN<o0yqdMA3mVYBYj@4Q#R64M)=6X7WqG!Fha9B!5jji@j
zy{@wYJ$GT9-wLKt?mE})8xxI|N9qj^s{8-Y+Nf<lt#J6hW`;=?)Ew9sb9=@9hgcw>
z3s-RvpU0Y(p9(LY^f(@;)^x?9Z#L?y_55K*?JaI#Um!gl2UkrH^`tubC>$``zO2@}
z7E|<GaCTa3hFoNOAz`}LI@IOHOFsUpcE$TcBKueI7s|X0$7Z%3XT_6&f#zNZe84^@
zMub^F_I5Go510^{Rs|j9jSxx84idcgt0B;<{xa0_hw`&Lomq5e6;(IC0SB4jFy!HP
z&aBB=1DlSvPT~s{UhY`bzQciNTNhj1aR^pG0!n%Mu2#r-ovp26)fIqe8^{E?o|*00
zgiZbie8Pj$0o2g|+kS+kU-eh)AYF-@7h2#*is_|v?h81VWk4dn^0X-|0J?FhR<?I#
zzQ>jyuZP-b>#@;$IB5(eyZJb079oG<aq-@VEm;fwrlJIC@-L<=Hx}s@J6@q_+(9CG
z+&aep8vwFf?s$~xflNk&ltWP(2E`P&{lwfH88}pQ<mohN=ByE)^sN}iVn;HId_+%t
z5Dh5MMaIN_^;{sP(X!zf8Cz)){?uE~ZWm4QKo{y-Fk4~8XvzK;w0-YGlj%v9_BQ68
zsYSf+krtYCzX8)Bo^|Vyg`@YhqvA!6S^j-J_W>6qv+(6g`83KE%?6(Sx4_M|O4}e0
zGf;WZ0UFf3*d{(@+kBLq;i~HQt2J296k0G-27buT3bY|czZJ}&+fB&Gn%qeb_!;<D
zo&MqP3=g!rFh!Ubb__LR>5SN&^@$>0U8p(_OP&?i`UEcMTzQ?H`n|n3Qr^0PX7e<c
zpXn9@heO)iJQT)trpo?6%iBco*|DaB7dlsspgMF5!}&-oOEC_u;I>W3z;vJOpdwhR
zb)ZxAZ)n}Kyn;r;UFYr*MV~{me~TyVLG{=zTH!fBoUgv?NV4eHn&1*srd|DEL%z1^
z={=BxkHmDb{YePq!!>J}PR${#Pg^pA&T{N-6Tw#Nfu_)2&`*+qDzQt4W2Qvp^44uX
zk!{Lz4>RL((}Nm~-++sdNfW%y3+iS~^s~{g7is3tgEFm33MB~3Kn#56h-)kgHK8L}
zHAv&;A5EejJ5lhjszj#-k${qP&X%+se<X>GCJb6T1RDr*q9{Auk}1Z;egi7d;w@+b
zJ40n#?cz8EDzu4_2PBGJ9&0-Wioj57v)<eZQQX+4be6++D6)CUm_4Y7fj4(t-L>h~
z;Tb$a(sBw#&*@SyGkvRHOnf8U*hycb?R#m!eS8h0vz$Ys+x(fopBPSDUWtM2=;_bD
z^PuR-ssTO2jZ9x$3y0DJ(dy#r1zO;rf&fI7j)8%=oIyPxseAvXR&c}9r<N%HyjaFV
zN#`HjErMRTuVr@Ws1CP%$d;cKY>SPZNSOm}OxszVycgG=-jPSYs+6-xz@Y(IM-$5H
z&mLxs!?(>VD(!;;A08}A1m>bhF(7m?cUk$e)jf}hQb@tFn-w|@$17M|w+Vx<`6L$D
zYVf{If{RP4zO~cX2&I^t<H+fo*-#RFEYtet>k16Qi#R#M8ACK}Z#h!=S$ocN&cLHz
zMprqB!HTWIY>-jjeYNW{u4zLG`CT!J`RXsDOrxL8d#F#)Sme?#Fn&{7v%>-}w}nB1
zXu`Ml%4*NxqezEH6Um~HuES`dw@n9g<7srBRU+42{L{ue-i1X5iF$D1DXN;D)Hx6c
z<qn`>sN)Cve`g>d-l()GLGD6DFqd15Mqb=eu&pAk3Yqs=$b0G+68r(=<e913NPm+m
z(pzgfo}lGk;5Y1_Rs2tp9B*nw8_f2BBj?B1=d(T$ocJf9Q%B=OUvC72r6?0^tVu8Q
ze>gxB+JjAzqv$3lXkMXoY!eJoN3CF1_&&b9GOGOzc=k2$9R~2+_42sRG4RzZ)NvMD
zt108XrO!ab+GNf3g9m5b++HR{IG4Gx(2x@6Q^E(^GFA13kOd-Nk?SkNKM+;|-<jOC
z=={3^u||ZH#LoKms@**5<NNDh{Oh|3h@}TDH2x?~a8R6!r05IUbJw%u8ppd=?Wmp9
z6LqGW_eb({l6bny8#Hvoq5j$fhkvckl!;mj|7!t-*O~9LzrrZ-+i0bz|7(d&mpEka
zuWkMS5kk)VJ4}tP{^MZy*Ro7q<XiC3sLn&tsEVLb<vlyDa|(Pl{u>Yq@Eh$J`MFph
zbV&H1cg1@i7<aGQ^|FMbto*=1L^6}GagZ!C_Y2X?+GIrqLYGD|ew4nfzQR@75&tLU
zy!mEn&>=3dzZ#?yo6B!(@PP@3;cq}(yrk%xhhp+PPxY(pA;?dmLC8_S0e-vdIK{!|
zqmvN4v!6DI)5Y~2Jha0KX$hqx+I-@|gMQcfzF-xT<XO~jvA17-^9r@&F<r!&nfWD$
zVf-<t$<g)lPs!We2ZVQx$6I*vHr5<hw|Zr!hXF)qoq)n6CkyZ0s41vP+A7Lam?fBE
zcxVu!%WZILbbKReLw6Qf1<>+zvJ(9sICQoOr^^GpxJ*rr?I1|K9bveOL$^1v4#M}e
zxfG4Pk-}6O80<0OH7NWINUZ-2$a`4Vzu*opKgU824z}s?%t99qC3o+?pgrT~y)vtk
zY}g1I0}%P7enIgVge`b`_w+sEU+e*{NNjCc9Jg5oxlaWt02B=xAn-0E_fS|Vy1UZS
zg+9dEZRS?T%v})j8HRyDL)>i&w;|907<1YRXp9o$wHl#3(3;Nr5<oOGv<@(37vov3
z?8Y-43OfesNm1es4^o<N&&!!!Vf^9$HlM58E8{xJXRzfAfO-pl0E!D-c(<~<Oy40n
zpO1z78;KjPAE@6w#thi8*j{y7ki2|wZ^Pwd=fC3~z+dHJB)O>-Ef0U?d#Qi-+@ra5
z=db)o_3zo9=nvcX1%CDWg=VnHy6Xq`&a$Cor=~^!B~zn;aR1kAPvJ#&iu^a}A{q|K
z>|UY+Y5WDz(UC}>fOa&pE&6{%@&}r4@UqZx<*jxP(S>axP4=g_MBBlCQx0fs{-G_%
zpnp{ZDp!RGw^EncXjReV(zZZDl-xl^eDoSE)t=iMPbIJ3Jot-L(FUX2BL78&oZZbm
z^?({-yqbF%yw{cf235<5798vfU#d{_fxg=fb~`rhEvv6rjML6rs{lsq@;rvFT_s=>
zI3eA()FaQ(H?gRT1GiIsjo5XsQ~<-0=bJXXi(f(YxIz7^!K`A!Jem3(j+9;#cbV7i
zSWnZxn3S5=Mz!q)sG(NwMIJ!Dj3ZU7YMN>~dhg3_{u<+<@!Xt}C)+O_NJX!Q&|z)#
zqtzW>d~cS<gYQKnVp4}Ate1oh-@XeUy6<hPrX|QF`7q`{rFtJu>>SYG`~cdHSA1Ll
z%yc5-HEfJMIh~u>NlZz?E?!mjd2zDnmvtKonMgl7Z51WPdPh;!w;^dbmNM3s{I}&|
znr;K;;2@w+1&=3|!(acuVki+IWFd$LC;l6}RxQ`_cg8>T`h`#bQH7k+I_+$_t*Wok
zi#K^akYcl^6?K70RAsEtfSciHW&AZM;JeQq0~uxTKbrh6yLmwK8(?I5A^8Ynd5Wg)
zAA@_)O$Pvw$XN}P%E}lwy*FTa&FPcXEO0GHJ}q0@S?hIu4}DTejQtmf{Gz11#4kb7
z_*8ibcB9R*lN5!W^^|%~@hZYMoHh?lzlbq?!_Er3=l-NoMztYT^d>&rwqt4NQLGl1
ze1*n%${q+@Twwxz4a{ZyN9X?&H}^=MXj8H1L6U7@Idu^V)1?vp1l88En*xj@y;z{i
z$BQ2caBLrB+gX~<WTnM>AbdU6F!<G!gScTEF-es4>YOhTP`+~vv9ro$59HWD4W*Of
zyG6JDxmWSWT+2y#X<o)6R!u#-gX^&Bcsm){LKm{CU%7`07&h;k@J0cJx$+Bf7=H=?
zJHIX&0q@=-69A$D0eyg{)WD5k!WcR?sRweg4bSfVrxh7oTRVf%A4htFfRN`cg7jSX
z<qHAW)Rbnv4*{d-PNZ-c3oE4eR{-@24maT=^Ydz7vCjYR`@s;4*eoOxp@3Jy%^I^t
zq93a5`Qfc!wG-fdYrz7%H4fQ%hOMYXl$CbzR!aHWmFwNx{R{XxBDm32U4T`9;q_)m
zyo;nzuyWa0-@E06>t!3G0pCA35fImZZpV6Od&Dh22*?}#EUs*FOSF3(OmWWj^BnLb
zf5s#j%c<}ihUh;A@VCwY>I5F+*J1!l0f(EYk&6C%<)vzSVl9#i81*05Eai+FZ=Z3c
zb3-u|AMSBWVbo`h7@$_b7360OuIMRL{`?;%vVQiIp65TG(|?=%hYJo32~q^z$QG%G
z0>VA2;u(@j7|zoB+^+zCnPI$$*VabpsqF=n{HJvKH<&mJiXLtN`G1fV^VCqCpB{Js
zG<BA`7nh=8q!OhK0R4?i%q|98M$~ARIs@Opo9A9SL5v#VD^~N?exX~B)^I&0(z+rW
zAGU5JJI&kE{|5M5Gb6UorC&O=I5eixkoEpg$j3ho14HTVm+4f=8qWhmzQZGB{hqPk
zfY&`yt8EJnG%`yK2gWVdsZ6cU_Itnn6=Z`oo@(loVM>E%=aQ@^7Z>7(>Q^tl3rzrx
z0)MJ84vu^Y4^PLye7{Ge@iMd>Kxl+7v=t1ab1(@CZ~xz~ZI^->qX!}$7DXj|zy@G2
z2T^1#W|<KKh2gQ`&(K={*oHfXo+s1wH!V$ajvlQ(R!+8U<$%1mq!pZ^%_u*a+WDj7
zf#3%dhOXR-dpIr53D->THK`!v#89)WW-%Fb#4GvB95Ar;4cg}7EVNfenz%u?zh7!I
z8h$?Q$;i)tEjI(1R8tJ%ihG(kLtZ$+&NCOJlH6_VFHrX)L!#+bpA~$2y4_<M9F2JR
zSE~L$zwa!cNMuYg8T<{fax-VJT5fUFB;UQl-kK1;-rSrHzka;@p>5*H0jKzFV1W0<
z@iU&<#_bx;KH%<dAe%dZTtL_C!3jnQ^RkKYomEIlQG>@9%JRl*8gx$GwX1dtvNCT<
z?&t@`9tckA@YZ1LDg&rj;x<68a4s+o>u<>0Pfh?b)vIrxJp>4j$NkHAK1$$e1JoM-
zZCUWA9BVJ2a2Fs596$ep>6JVO$UHl|KwDv%;xGM8;+PJa1}Z}O+rI(;ud*`b0;TAS
z_tj7r0e1#u)JVKAZ~V(L2xMuKxoRD@V-if|I^r~W6+l{FZBM`1vxbp&dNjWVbrV%P
zG9AT!%HmbAgE4Io-x{#gmH}7>9+?i`>vV>CRcr&)6&K+*02{=>5dvMU6bH)Hthb_e
z?S=Gd&o_kF`bwF$#Txbf%P9WC7lFsd-%ITNd|`T^2NKNUW8C%#qqTy&&IQwJq-pvr
zm|ZCx8T$>x2X{}B=UzgHefu@SE0`UiALlDjzKDN|1sG5p3^_6rVoZT<f7g&yt#Cki
zMdzRs1TU{UG<;<2kotQc|6hIP2feL;gH9gbE4g-&Jh9;knT@?%E_02x5^C_t@p<m;
zc2k_0L;dV8=xF8sV%a&5E<Z$8xY?Y16Dg&wG&+1c5SAnhYIFg~j@Vg8If35~KKXkS
z3`WDOwB058vV@DaXKEom5iXm{g_6#3%H;-Jem<r)raan%_HP(bcP?tBIP&w9&@~?H
zmA~}%zv>pFM-Lu#*Ml6z+U+oD2{3dgxt{lEasKw`1pqfz6|b&hzPiPN)vp2%9f^95
z%^-z70&}_~))qcL$5eO4RPTa?`{3Ce4*vvVbK?RL9${I#bk~9*uC5I%6Y=rGs$_~6
zJ~~_@Fcmt1-jetTy`sBU$@>lIP4)tc4o&GJa`JPWRz7)PReKM`fi=j6bm#V<Xbc!?
zk>*aKLn*HwOWBlK{>yY@U|IUcd(egg{#Bs=UhV;>;x|hfY|(W@`aStxNARt+!GkAN
z7eRmCXj`omW5QN|%t+P9?)TtsqSCGjPkr*KKks>Wgmu9JW8(_pk8E34Rzosn>F}gv
zMubv=6^x3?h40U`V(A-)`Sovo%n+tqSxC;R49B>WIhNL0VBaHw&+zNT%5?e~Gc`CW
z)!Ha0YK9S$5;GO_DJ0S-YT7~!wG#4W-5~^0=(XX0<HDCa&@kDxnyHZt@mPCqKO<|a
zl@vsg<GgKk9S~k?Tp;|FZV!iVwY_>2*-~>FN%~={;oxYsV{{v4B*4|C8SqTCw&ir)
zVfvtnaiNgPZf}n(aIxJa(Aa(P@_V;d4RD3qc=26+eu*0P$l8sR$FwQmC;$+%Hmw@a
z+#KxDvm%^0XG+(J)y8yM*WI}%-;2O_cX%|g1~3*WOe2Q5t`B@6vop@$ea18nO}M?E
zXEYdWEQYMQB2c)d0yKJgZJsb<k4O5%Hyh_+-rgynKj}Mo#)(8F{b_>{q>d>((2Vf}
z{payu2Oym!mvwgCPDY&S#)B<WHLX2?x}p4vvj~SSPh=&2YiW@(x<mo@bd(vk+uI?V
zXUH2;m}eZtjv)P(C03)?rOok^KCU*%8s-wpKN-lNb-=BN!Igo6><C9yw>;xs-e`OE
z2MZKrU@O%(0du!YygSc7R)bRQ5Lws>m{J28pdL;%DUV@IZL}&Q6Yf_81Y`8tucm2F
zmqUfeuxghjCE83$C(HLWf6@r26Q+mPXPWCHSVz@hB-rC_<1fY;$9-)qXf7;N0*GFc
zb{qJ72w)Xpoyw3_nFt}q+)BT6IO=mAu-zlH_F{|4mMiUE`JNh33qg>Ew#r{a@4;6K
z-;m5x3f6I)8C5Jqwk(Cke_ap6Hqp+}+Hbb%V9zO$SdYboylyV4+T({ZPhGwu!Q)d?
z35zILa2Xx2*Il>MO=BFy>&b_|Kf-+%e$H_2dLw}0Bz!z@6r>*EurRG=v$=QG?$8o1
zE__4B+l8|KS&CVWi|Q7eig`1~UfhvgXI$R6&I8dc;^&f4VAWQw&keYEHsFD?IPNAE
zYSZzW(5rvC?#cd^-(rhN&SJ^|?m=y_W{K|NkX*q68Cx?G^_mWh+&f0CP%XNOUXY)S
z{XqOzDsh^wVLae)ermioZKS9g@tzAhm_>2D7@)AM=UmkKhRY~u6TFfeVBDI6nb_)!
z3RdQ<tq3%gUZtXY##gS>r`?5f?+pN;_@Wmvy$jH54ahQes?bMU4Kq*odFV|8njbD7
zaHC(RBW<HaDT#-t!N(?dRp%_;e+bzY1exk22}d@6)ojjMPF%;B8bIMJaT&hnUfvwL
zAwVaZ<2H7sM!!!LaNceI3wzw&nVLXcpKO+A__s*?y&0xy!X+YAtnYx^=}wBf&%YHM
z@Nh}pdEf*?$ye=2A5+!RGN9270A}FZ03>L_D<rwOr7&Ag$3Hh-0$6-R?a`Y}Z9zsC
zzKP9QV^^&_FD9D1ri_#)u%=Yg#ZLcpFJ=7lntL>k63v1)&;QQE7=g;u-b-9#7=fj^
z-c1U)3V4N&003?#Lx_)`9bkjW@LG9Z1AEQXIj`09D~7pJ_`GgrnlFNguo3RUH&XQo
z@iRk(Yfvv8jy^x=zVd%V-~6pA#!lTKi`g)RHvN<<8?u^%T4gtD`zOpqn4E3lBVUlz
z29zX<t&EMGyZ^+~*zc>OoqLX>lha||AjV0TjrNNO20?2Z%<u$XI@EdT(u0AEWs^4#
zIrcUm^+s3mzh{;+jK9?>T{7H7I?lIa9C~_Mx*@|s^(m84G4&vCBM@eQ##^DuTL)b@
zDJglfLPF8RT!7+rEVCGn_6TdOj;c^Hr*%^N7szG`0_ms`P^$6Y9p1mzC#yIB-P)q$
z0DFhv5tPy3OHYeiwAr7LG?rjlQo}P24fw&x2{-w->H7zw$%x3hp|Y^rfTs{f9Vhn2
zn)_i&(enzz6!OgOLrl|(Sa!<zq!ksakLIE=K5}RrrwJL(Os0i2SuS72#wLFuZ0wJJ
zUvKKZtTX-8)_OLZTtPD;aaog2t~}Pms<Q^*TLWEvwLcA0jeS5eH%N$;QJ63e;?V!W
z-njI&uBBjaV=^l;RA0JgT3hLgI4c-`!Rkz4)@KNmO8SWJENV)KV=iI-bp&h{F_X+B
zzfAv&boIFvFzc`}zK)Nd1xCd<S&;1#aD^T@jaPC(Yi!1CW4Gsa5AcZ)OXU8>*R|KW
z=Y@W$l1HD6VLb-oj%^SkfDi9Cbaf0-7YT1Vf``gmDMS2?AH3f}fsGy)IGKy9_J?Z^
zFC@R(=OKfQ#86)wPSjT96XU!Nu(GlO{C_f`BwoEhnxCQMUU2{GeE+ZeCVY8U-?kt4
z3@bHtM}C4H?W-#MzVnLMehWWQ5bPanX=!-6Uie5>7VNkT0ZSc@w?XOv+ks%;LG)7U
zZx;^d%N<{jRNvND6i6<}LKOTA+Y~EGP5j+De08!T{xbS5AH11-dOi~9ryVC4IRA5&
z*6P%D+0w$+r&*2)iqVJT>~tjN;o_<HhAL<j`NiTl;54I|6q!bV)zb1*y6p<mPPIz~
zz<?v3l+WfVS(EFo=4e|>7QUhgFoY5+=h<XGaZrpKFPXTa*dx;30G#z2($H&>CFjd1
zDJ59iYN^yPG0FZ=<Qwhn9gL2Sjt!42O4p%LM6d;Lu~*q?n@ffMRCVr8XZrFz^_9DJ
zMK9P<C;={58vdS~V&TX7PMvIn+w*D88w7Rz{QacnO?ue<g3zRj8s*}T&rcmDCzJ)?
z*}J{+oQa!f$H#-TR+IVYG<fXq!lM!UqtVw07cQH2vvRizGUMjf;AvqqF543u{uyM=
zoi($mAhI$s%wY5~J`A0h?JQNf+R+ZLZnL(FEBF_sJRxIf*-QeXz^}LKU1L;fHJYbv
z#+;LN2qY;-!|C3<S2}e(92UQ4xVy6rr&09FG0+RB^;llg?w)|___EUq>KAM8y2i*P
z1m+5HP^ilHG42-{zJj{GaI<GRfC7zIHl(4rxm2(?rtWm|^P?(eh6BaVdx5)ksfKd~
zUfI@C<ubL%?ent-+z?VMH!jrW$0@6&ln`J}nm9aXDGX#7m!!?4i$(xf>+Tn8=y{%^
zAiLHgy}kDIdxCp2eVue9np7@phN-m8(a}+_8TR;Ata{=wbNT4`fijd2J_PR8=1h<(
zCYt7a*D=U{Q78clp4em!dYjZ~^r=@1f+S%coScv%5>;ZN$n%XeSz>?VeLWco;+;{^
zvD0`ap%v6+d!Uj^&0dukPkhxZwCqvS7bMGg!%VWC_M(-D1Yn%3@ex~WoLM$J6aO-=
z*Xq-boQzFL$tzn}ynT7v@fKj3!J+$ypSE4O`7!0w;?$2`4;fd7-V+h9O-Q<|O}3nB
zuud<^oyV}np&~fFYcwD(I$s-dRI6Uc8%<2ME9ZEuetSamGjidDt68<RqihT<zjZ~e
zfV>s{*U2T@S5%i@=yR6l-~T8v5@5q~Yyy{#?GiLsy;jrNd$jVrE8fo5QESPHT{q*M
zVrnFzzhXQK&qR@yK7?u`VK$I|sCA$SsN^`6R68X<mrV7rIHGQK4D)6xbj;2$f>n_^
zDW8g%M4^VCU*chSROw*2j2j#M3E2B%yLS1!$#>qp0JhcaiETM1n(%iwH9&^~9G1ph
zPjU1~3rk_`m!pPs64;>1do#*uzIP-6*0Lv$nI`%BUgfHPPzznRp0__2>W$!7lZa|6
z)FSmWhubQ;Y=E=doDQIpug2Vkllke#O<JWtNc~;*|CvvWaL3|EK6lOy)5<1Z_S*x9
znjYaDM%%~+zH9q$07GG~=?hX{Z`5boV+LS5cXp@2@#%a1Mwj7Rk8hQ^qc!IM&=9fE
zfr;W2c!1{W3#GM!mL2`+1xEU^P1^qP`jk~JKd?<$gSzhUpTP;!E4pzI$8r^M#zoS5
z&ef(tH3VzW6u%1t62W(;A*P9+TfYkW<+xX(2y|)(%(@;>Ues95<8f^ci>W{25N|U8
zO=xGm7@yBx(%%izEZ%+3I93_Usc9*`PIUMp-OEERnV6V>*)GBnSzE!?BJE%r>+X3l
z+au8}Rj~iCmr=IqCX1^z$JO$q+l84Jj472wAzpp-oPKoh3TnzbxWn^H#MJA<ff}l+
zPwb4IbJD4aJ4kRoNW5X;MhJ8NW8aoBukS#2G_YWK*u)cm*Mt2XeA;Zn)<MTgn|Uk}
ze4I#};4^cw-D?z}t0?nPX+nvSn2<$~)SPqb)|**zpdRA3EXF)BAjVegbG)9M)bEx^
z$iN!t-<x!`$0tn(<r06f$G_SL$rl~4)6MT=n}!m2<CDABj3Eyek53w%w%n!xkf#l<
zS=Kswf<k#+J`KE?i`&0?&7AmJ1h2s5?5fIIoUh9p7j)Kp$`<fhyro<feK3Nz8Juaq
zA;WAN?9EASfu`RlM(VV&hNqX`WZ9oxyVj1}dqE}oDZ)pre?KKok);ytYgNKn7#|ev
z%SxL@k@1fI`eS)3hMs4aQR2A3VZ#MX7W3Xt46so3<MRX2hi7&?*{!=&E<$MuNHr1p
z30uW@1V*~CCpfa0emQU?rf#!7f;Xjaw;-=NA;G^-QIF~QpSQtAUTXXq126v}{9?_n
zuYq&XnF;Rd%k*+Ov8B;6Q#e+756-{co?r;2=Ncn(T@7Nan}&be=!#YUa@`rGli&X(
zMp2VN=WFj5xh!0U=)sUyIiu)C35Dyp$XaDYRq24^a!`bwj<9)|qARl@ZCJ%Uu5$6l
zTg=M?=3*CklUchUl?#__$3V~g_C3*}hVsdCgqC7JYaC~?5a;*@RBH+shGuUQfb72z
zu=_2>Qk-qxf3KldIQK@G1?z7{GP3$)`;PcvcEugv8~<R!S5)zPQh1%dDPQqO)a8;l
zZh)G=O#eHn0l7iaY25E(P2~B%0YAW&#O^w)_J|eyFS!3jr~k)CJzKvV)2=<{ygupH
z;#<FJYFnzw43zCuqKRT_E%7Yv3KS%%aE^b;%ON@T(=*c`1!6(1khG`o?x))uP2hX>
zY9)7n^ru6O^%HVVs`_IuUFtV#oXb<ojUqAY8iX5F4x+a`GM?`&3iI-0Ri)GEx*43A
zxT}?4!Nlg`9t$786JoLAwv)rQzW3;iQloRm>ps;w6MgU3^4>;bnb6JqvN3Y?OoE&t
z-SJ}sZUL$`>sWc|4%xM1$^H!V+jU|e9UaoI#L~qt0<2#Xp2>%R77Sv;r+BA#FVelr
z8Th$Ap`x<hi!Nt%oL}b3vPBgas;1OFR!;Brjyrj7Ol-?P!gcs&24gqgg80cPl@&{$
zTI0NB;A_JK`7M5OA?8CPjBd9oMO$Jc&q8%aQtcK3C)Mar{c13J9S_zq<z0!5oq`zI
z<HVi_fOtG@vvLF-2_KwFo_+}(=(@SMMnqP8Ku#|6i(I#$pt81aV=Ni?Tu0|up?8J~
zsb6<YmHJPnM6m)Rg>#DCIzctXW$0J&3NGds;`qp%`Fv$-0;M*&!lC#^mCkgX_axc{
zRlgV)d`O=7Sm&-JC=qF|ND+x7Gg&GrMJV!3#4yXr$@lgSypKoFG3)_!{*RRYFA8z>
zE+X%0Oo@X<TOdR?L0>-X_sow^h1;BlO+3v70G`s#kwQ71%>GL88|S?m5E;&V{T5c6
zS%TAjyVQEbI#jQttQ+!FzI7i~k(bQ0k*)eO6k^%*gOv%ROvtF#%6i%Hc0_)7cq)DF
z>>J;%!I!iP%#D<Bm|7jo%35?+DOZPEE-sHXbh~FKr7EeT#y5dty;vl=eip~i$6pB5
z{3A=n*Z!Fpo!dt*3x9`Z{S*^QbZ}70PgWu9_pc!DCas2(7SJJB2&(TUMM@AR1T}=&
zVUwwn<*7lu3o*HFe}m1sRZB%j!qHW1=>E<~x6H-Tsx|w>TPn{&axo(I<AW!S!`)Bl
zM@#xV8p~diejWM9CN7*Va`df+L6_F<H6d}KckFBG2TJ^A-_$k-x+XP;4!b#I$~S{}
zbk}1S1UYbG^uJu-nV+e!cUys#vsFA?SrdvCq`#O!`gTS*14!ti^PhEY!JHtDb-qkL
z;#A}2^@@<MIifm<T!9lN1U7VlZNI5OLR=6|Z^G`Rw^K15M6iuklU^mtH=h-PALXI+
z;k1zwfAp<kM%@fm8W|bP&?lfiqK@&z6cW@CGfFFgIF{5K87aM$HU-jE&I&mxmq7yF
zn`e~mIwS6K{DUgb{d3t;YHIC;TZA^xPBeVt|7_*?XP;uhr%z7soUSbIH$QB`s<mr=
zg&7Qhax+}4uy{_ud5BcK-N?BeE<R{pbZuWnWBiN0`Y#yi|F4f+NB)*j=>)(wZVc*3
zMWOCH+Y3uXoDUMiAPnQfobjIdD1abUO6bE0*+xuk^e`1E8ZRj;ie$myv>RtEAM7Il
z2BWXfZ$Of)@>O7aqJ(nQx2X4C;SpBrU3!8W4}mg<K1A!}LdDB1K)@886O82SWy01-
zT4}R0uvJt$htBNUgC-8`58JW(#`E<XW+-^ctf@-gpXE}(5a;1<z?v!1dk{x0yu6Zs
znD4cT`a`H)z-3h;du>@=aFhi+dC;R9@i8=E>W8xLWT*<HNf@2?Wg=fa(wX;uTu8pE
zUm<IVh+NVRGwQ<WW`7`b)O?(63BQ$piv@4z>ZlqIc%AM}dm@^Mc?@c^G?~ofe^cH<
z-k46zrp+{&E6iNoPiN)+M%%oz@Td1ARTWchb}|_J`S5MCa}`vVs$^ELYCtqy(mC3&
zlGj4i!)U_FvrPI^Gi~nj;3T8DajILDmcs^DwocJY*4zR~MV2qV{QR4>8=v1#-Ec*Z
z6&FM2)EXSlwwLf7Ey}wOiqe(c*(2&cd-8wRA86ILg_L!u{lq68K>QGrx%MO_H8x*{
zfu>a<Y6Z-W<{9H8#eF3MiJYnGWd*s<3HI{GW}P;hn!K2d-iO@g%!&;{Kl(*oza0wH
za}SF4j{oG<%|JiaSf(Nkn_%oRs@H@VBs(vVwY}Tcgp>Im5R$8g=Cc;CbOumYSKG-w
zZ0xN)9DI_vO~i$W$#0PUX`fEyE*6stE?5u>X^7F>s}~bON7Hx7HF8+PT^-X-p>AR&
zq4OXaDlUPPsf84+)qJaC8D&AoE;kb(n7`3M=Kg0S#)Xx6mJfTws`=PRftg$X1;HAE
zFa~Tw)d#Lvla(vcX?x6`vganN%i997+)1EtD%}uSuX)J^PD_Q?7}dMax8;2U8;DhH
zFE6)wwJ&T_@{(OVbtvyFmF;Y{=`@;L(3e#&R{d%`^PP*2D{Abr#m=SZ0#vZKRyGzR
z?^MTz??&&sof^H-him$@(~A7*O3M<1ryRB$c4e%fp9KfA`*rnz@C;JAmbYei67{^j
znZ9{11w@}}&E~g1zE@GBmNLPn$~yjeodWT)zOt|lqvB~MW20yIQTajYR+Ct*@5gqF
zr}so?37i~w9z5by3Ft(1?f3T#zRq7or#`V+Yq56hG3-dI0d*`{3+z@hm6J;|F_FN9
zCJov5JB!L8K!|4K+y>vOta2d{LEpe<Bfi-;)>S`w=9nT39?Y&ZjbqrEY_63{n5G=!
ztsWeZvw9zpEehBTb%C0mvD%`~kK}jIC=`6{I;QO~Q7D_-*!t1sbfVV2d$#athGSEc
z0N(V7O?H%j%N5tr+7ia>m!?#jo2>#h>l3b#{AGlF<W#FYd-klyKzSl<3uZb&LrtDa
zsbAFkV2=I9hEP^YDnW-MaYA2KvbOadydda}1*MIe=><BDV$$Dw)+^|Mwx@(|xF2t<
zqGYK=p<!FsYFWfjs(jW;-h4S(M+PRad+A7cR<}j0v&Bial{X;W*R!4g%9(c#t58hv
zI^*|QC3h4WEFfLJc9{)|^UauZvOBZ8eJbcz9u2p0n@b58aP=Fmh6&iJJ@4KX6dSIg
zuA$F0a(!~1?)zTioQYO6EIr-5cAlGN|22AB<3l#6WYr$j{B}+_t@!g>JDq7e!6%2)
zEl%WyI(V>&Cqq5?(oj>ULY;i+<mOmvpWce*FdA)+eAc(oC6@5%?c)c-La)&aty#i3
zN8OpSYV%uY6d5v209sZ@X+KU^r1R8+Oeu4~hK%~WGik018bL-atHnscR_s|K)x47V
ze0othk0~dAeAWM^E$huOkcvxiP440Fth@JW{-*0(jU6GQ#lc?x3&FH}rnDHfVO=bp
zox7cabmo^sR)soeSREUCsu2n4N&-HeMH7|J`OZx_Do#1U=z!-UN3{~2v&5|EBh!}}
zr50j3bBk4wXlB?Vn?1q4m+i^zk*3wynl4O7Nu{#jsuYq_lvP5uZP}))#l$fu3vVtN
zV3_cK0goj<P89TeS}Mju&Px%c#MoO?9Jp?rJV)6|{_DFb)?2&csc|JJIl1VkD<@9A
zX7@tIe0xP&_@Q$bVu2v1w$x@Vxyt_eOjDX7*-j0zd?`ewhNpWjV=bOyqZ8im+(3~>
zRic=1JN&8bNj*E`@?LT7uR^A_(CRM(+QQuA0}>5ocYw*)HlKTp<{hJc?lhnyOZYRJ
z+vlqXU41=9KGmCr=?SWRN}A(VPwEm~kH>}`!zdQ1-N1S+N+95q)2!?Qfxa;^t;bsD
zttD0cw<hg_P6|Rm8+999*M0WH`&H}y7IN}%4KDpJZ66;Wh{E(6VIy4cHB@WZhEmRp
zle9dsJ;(xPRh1&!NZAtIo~Me>_1n`Q&s(WBH|}Z}%qdni*>#WZ&rI%@1c~rgrdKU3
z=NS%ZQgz_wZI~w`wLHTx;FVSek=pUh{Gx)Lw{qk!N$oUXefd3Bwk#>Dsa)Yv_DuC>
zw-7EP4*X{a95dipzZa*sx^vixG$?D)?WSYu3I`*vP1=)k9mAvpTt{niV&rGc+^J$>
zw%ss;qIOT1OsR}7L((i!y%N9Wpp3_)`+%pxwv+Di7*RwWu-9l~!Qg6kGo^H7NpnTM
zrDFcom2kgp42`c}T|QK=$Pgma>zZ5<)t#xAtCTi?mCuv}RLz!CeE$90<k1hh6gU2l
zHIs&W>L#_MTLFf}sBS_D6wxTlMYFgU;$%T)cEBsaX>rGL-ZtgwK}VBky4C6Tcw|nS
zR6mG<kg~y>S;B3pV5S@vmER|lF?oRdwPTX{W7UhdZXBMutb|;mIdzmC<4Y1urFAW*
zmB(w^)>k2nIX$eQvjfbAtjyDJ>$^2yZMn8@!oo3<6Fbjd{`1SHdykqxcn0YGRCE}}
z4BG1zBb%=G3F=O&M_h91k#Pif(!&V*GjK+-o8&X>>%uNVhD?*7xci7h^~em5Gp-TG
zKq;IWfAKOzgTuXA&c=1OA6bP(SjtS{==I+}b@?bz(|jU6<>xv);yt-XFq2VHC4s7#
z>d;lK6h<_Bnm6Q&jfB_^IZbtl0|x?5+_*BIYHjC)IiL?coh~j3A(zCJseDg<L}?~&
z@@_O?yF+#Thqf5k&~3oP)b_k{4?A}kDFc1%!LQkMZLJi&gEN&~3VX(hQrt0N#jdn!
z*SAY*r^*M_{Lg#86MHUK&{m{?`C?2cMm#EO2bk&>eNafMk5oH9GG@UdHFT`P0uC~r
z>G%5}`USt(C+1>W6vRH15av=0S;K%9mh<nf>#(dX)LdCw2Lt1Mo#Kls%co@LQnpte
z=L5m_3;g{q4%8F$dlJD+?!u`){jK~>8OTj^9>Pd17bK2JJjGNk!J@9tO#jO`TR9Dt
zzS>C~7^7dYpn_)6guha)tV!eTuv{IOI~IrlVh59pyUjXR_Yj;~y)$3zi!yK@w6x=H
z2{CqOYJP`UqTSUhn>&}YmgInW%n{7W_<l5u!hQ9ss6DPOVOe3F?M38gWz^;6hMu5d
z?a5*$YWU#l)mxEk?1E?YPi8-KbWa#JrL;cdRt<tl2%ewtR?nq39Pp&FUC(pb!X4_}
zdiT-Eg~{<N_g%i7oJL`%C3n5qB*nFw*UYOfTzlGQkYji;M4~bkpWSdD>!w2xhch*w
zOPFj<D<KTOpPLXh+NMpBDn8qO=i{Oll1}q$%Yxb|L$+I<c~PL~LgM)RF@*<vPhLj^
zt~S2Xi+zx;YYOswilNgyV=gV)upWZ%NS)GH;n4_>AYFxw10{_HOE%Kk9`7bcZX#@r
zXX7g@QgDW1F%|@JmcDPa_!yl1Vm-@RXzUE5I|(OxC9<C%x(N&c&NgR-x8h8iGfn*l
z)S!~RDZ`8dK1bU4QJx8A;-CJUjY))`=v&BD6@?k~lq$S`Ip^Q~nd7Zn`QskV$h<*S
zzInScwY(~|LQa?mIP`TmIKnE2#t%6`i|})J0JWJNF%rQ}rwJ^urjH;e2itAdSKv5;
zFQ1O!JQZ_cuANkL8mB55P|`P7E@UbiF3bHA!atS@W{P*JTY>^xgON-9(OagoML%g?
zf8$+AyIHMV_hu*hj_al*=42GFrCy_0(Q7BX$D|?rX1D-E^ELuTvXYU<7%8`F_QYzl
z5#@UYVRyeeYa8h8QyP$c(bPR%0fi;aP7N)E(33SoYq*`h@kHHBL;G-pXZ|TU1w43$
z5g@u_7|hK8fo@?sn-YoN_0SsJ<+TniqkpHqo%Q<-Ff3J<{0*Q!ZPCrIItx!n-YMXx
zM*ef)|58gd7fLD0@p+qb*qBzQ>QRewiPyu!LtrAnu4YAN$47P;=WYO=USL1pgkY%>
zII}O$a(U2lWij*7&+Jh9G#T{$5ScXsW8oE?wS(GE7boF-ek#=Yl7lqCRx77*?f+u$
zz2l;4wl(1f1r-E|A|T)k0+Nv|QIQ-a=P0qsV3U)9BmxqfoO5c4O>UH&B@0c?IXBSI
z;B0-*y(fJ4esgBN`}=(}XQuz^+Eu%1*Q&MFu3fe2d8k|z^k3DmJhsZ2T{0zXFq7$N
zE=HQYM%I5HSaWr{o31YuWHTKU<jL@%rX;VY+E64)D)b|xx}PwORg!XuaPxyvozJVv
zU3Mg^ecqbcws0y3ddJNGw?$e~{<Kg%wrT+hQ!BdV1ok1jdmMO!lw({Hel=U<j^`rc
z6~UjWV1@+>#hC?KOnH+upIuGjvY9p0qou(7cuw0^EFDu5$+P;Y5cdHbN1+<3iKR*T
z(7ZCR+CGbdUUFOxFE8)dkdcy$gj4a^`nF>gZoF4wGfXc1!F|2m2ReSkgcH!$`j$M?
z5yr(;PJNP<S1!H!mTD;p$#$EXaoGxCKYHgre-OD?l7H!=_H6oeJsFK^cUpB6Ip#4~
zBzAbwDVnr6fxqhZq^milf7vj|N*8jU2sw@~Fb`D^b0T+59L>#-)i0Y^&ySZ$xFDy+
zGas`Oil=_^2&~63Y0`24xpzyHVhcD97)*9KBusfS5I>yMmy9)GH{QravVF+W$GK3*
z(AlIAO{J{HRzeT`V}w~J7njznChx-^Ir70BJ>=RwG5Rx&Z&%!PtgHonBm`m%Qz+f4
z!cg|^)sEr%%Exg{;JVy;p2eb9<~;Qk8?kG>2C=zSYIy<|z3g3d(F(;Q>s^Ag4i1~H
z2CSS$kc=5>pZ{U)`)h#Slq&}(d&lOvPvg)iEfyw|a5(VMkR-5krGEjUin5X%d~HWN
zU!uc48LOhhlUF6#mTPYOXnNblF3aeA$gQrwS4y~F%$s@izMCb%ifDYEvcaa4N~B`0
zsY4A3z>E3%TX36NcDq^OaL$(xwS9NFr}Zrp4CF}+Z1cY-ASLBu;vR>Fh6Wo#mEcq4
z9zSSkruD2Xy-_{5W-L>E1)5vWT;5tp7`MZmKg!P#6={aK<UW0s$x8At?qj=cvFRc;
zN104bvNBWrgRs(Y`*?9mOr439++F8$v4rxzWJkjanWd?~35MPZa9w2RgIK)|?Jb+e
zmD#v0CASb7|4rGC*#*2xE;*l+_=gMo$E?6(U^KFu`lj20`TJMRRk}87g9O=L{L|a-
z2A(;NxZEJ!pbQwOp-onZkZ*iKp;56J*ZA&BpS+X@qfi?X*0iL#wW8tsIv`gs@rI+Y
zvN{boR;6?_T`Ad5`=)Rl<(*5N#lcTAg)rs|s4k=i>e_hIC_UZoZ~=c?J6&UYXP=Z!
zPk;&OE8b#gFZz0WeRp<kv7PsD|Ih@-fUyocSs`>yZuAN?bancUbE0@xmR?W5<y@E6
z_pe83<XzUheZcgq6<p(ii&>`!-+yf`E&QKzIOYi}?^wNUR|tWSFD6)xz0oPbYhJt@
zi_$U4wkQK$NKT=5|5R#wj4a}}AI6reIlrPwTVh2H&iTH;e)mR<9|b}CiHc?$aXA^y
z&Ylj{5j}Zay*-I6sgrKm6XW8xBZmvs49F-D`8F>v7iFCCbthkX-(n|kggNv%`ifpG
z5q8<%<hr1(BEIfTIcX*{@KOCFm(jU*RL@?L)_;7LX!zx~a-lem%u7ZX+hKvxw`9v6
z{$%7}Y>2YusmuNENh~~k;-69v1qM)dS+N{KIbtKPZg}Gv{vc<y$BW>6NwOKc*pZ?V
zPcCn0OT@@M6Y3<ZFOEIs&Ktpz_jrKSL7qqKu5;sg9I<AxWR6C`IQSl%3WtZR_`@dW
zlb9JB=PAQ5t~%@1&wU(acDy{pEZks|C{CK*%rT4F@m<Dg!9+vlug#HDj$cf^8PeCS
zPEQ(J9}cXh6=KdSbEOOmUkQ6Ut5KP?#nVR-<uz&+N;`FW-{f1CDJ!ODDqme65L$I2
zGAI4U`Bk}m;#9?+?b3$R(seuZho>jfiQfsp1JQ#}`xnSIt`FPY<rOQ=T{2M5l6vPl
zOfRBZWXCM5eol~Blr}jT<&lmVnHX>~Yhsd(D>!vrRT{I(dloxG%b}W-F_sV-hml?*
z=_^fTM@Ya7UX%>I!#&b%=P1=grEhn4rAXVo*ihj7xYn^oW<7bJoh=Vl%kX?g?r&J{
zId@I@zB!$OqsCSHRB=X2nAg=XPyc-U1Aj_E<|e_NoM$fN90e9x56Rm2!9mZ)se|gK
zW1#n5%2QHS_<S9kO%fre>)Huk^`Np(%byzi0uWVy&=jt;Z)Dmt1EAqF6{{X;Z}2_o
z(wdD>&DG8t{Mc{)670Y-tN*G@uDGmr|3`v~<mBwgmA$(NDdqtlp}@#y?Hcbc>!2~q
zban5H$UFWj@BRWEmOI57V|g-(YK#*!HhDZMLJ7HVe0CW-_H2tj*;#$8pT5*sv@qZr
zb+n2ugjeniZdYFW2w5(=)zA`k8daj+?7Zm&Tk&XvPrTiYOE0aCzxE{v$Fjod?z3+^
z-DMXHE{}ba|I9<V5N@X>*B@6m-s6-boyEjAuBCOb-DMk(Try8#)@wLOp|^zIwg($a
zbN3NFfCklCD2a61T>~ec(|5n1u<tc1>`o~w#fI>>ywXv1SB!c_tiktkD66I}o1w-X
z=021W;y>s?DfgqGvVi6&#DC30_}MJN0V6Xw=u|vx((>5g9$5oTgk6;wzt{b-XDhK_
zjnQJ4&AUZ~_*%kgGPN6V=X*`LqPA?&#DcJQ7>=geWjFl#55?*S^KC4B18twcwkZnx
z73^T+2XbWva!<iJ0=V-N9Ldhw*PmjB$YRwZh})lpB{z+gR{PdaNjFy_lXrR?)<;OA
z3i9n%3WHIFT<`Sne~>bEEaThT*~8IPX+dz?WYlEW-y&{eBZ!-De^8V}QBGNruvT$j
zOH8Kdsjd_GQ)=R3vfda22EOu);E(>fhg?zh<H0?%RC?I<DqiUI$*)+Q2NRS^(=ELl
z!cO#-yNPSwV&YTx2KpKwr-_WL?WHZ1cg*(dZx2}~HTUz8x-bYxXVLRn5ZmT|NT8O#
zpQcajx&^X{U4#g`9LHDUzHkHYH5<}zkC$4udfdL!$KEl*yq$&J^^SHObJsVM>Qz*2
zmdv%9zqdNG`Y@Hptf=mFINvmTm`FiGVux<Rn5zNeM(0Y0LV%|F%e}j3`}&4b>LF`!
zImf$C4PRHRT6*9xywd&Tu!)sy`CcloH?KvRDrRQONRq6M92Xv;G!S{I;>-8Mr3LMb
zMjhYsnb;yNdUENfp9G~{I!4_ois4G1sna7BDQKr^TMuwh0kclcG`ZXDl=pE6(z)FW
zjs`+2JmY=FL=_3S*It&kxf^<n1PP1`%lK!Jq-bYo#geeH2CqD5X6tW~o6dHds-SSF
zcS0U45e7OKe#^LxZ8c~(V4_#rU#q@x&ad3&sp0Yx&$*?$sI%UVOrAoknP&Pfb^N52
zM(iKFE0XGS4g);|Jj)zUQeEgx#ClV;)Hjozm7%DGr1xxd_>w(7WJeA1yyB7YNaYOI
z*iX<NN{Y{&u>k?Qo2F*gUvJNePVHK&*YKVVg&1@Oyzv<1wu1#=d7vhAP`Ky%_Zv3!
zp~k6x^sLzqb9esVwyOVThX3m?R%lu=V+q=RukrB7PtcX@I}^%;z`L&P3>(xO!VPsO
zp9iGgFlxxbGh856o8!4fvlpaknM1U%G>~(35#AUO-{j#aw8f#g1ZsrTIqY3JuwyBL
zRrS2O@9y-tJDS8C?ElF7!uJo2{>Mq2_MFlyI-v(TtSe&(@$1^ubJ^SvUY=eM2Cnd4
z+-UaDQ~`{#u|W0cCTg}~ialpMv#of9bu2Y!q&Ti2mY}a-%=YN`mitrTn;YeM%5q^m
zB9RqLB1HyHJT<kId{4R!cl%&5@#N~)dTvsPV$4AFUK7mn*_7)@X{~9<R?JFW^UO1N
z3WNHbUed;l=HB6!Oc<w>{KAXI@`hOp#>wFMsDRz&v}SGCEe^EQx6)S~-4oIg9};Ma
z9Gl(eA=Hx7@8$io2!n(8v3ZamrfG`?{Vmout8EOrM#fk<sz(Ip)IS(E?9Z4G;!pB%
zUj-#KOhYqpi-m&XN*{mgztcAwkC%$CO8hj8wQ$_geWBO&M&+a0qPtA?p(C$bW`<l7
zV{sqnPf`>><e!pr)n3i`bnIrByrw}<)$;n8?&3|gUpu<E5gzJqcw^dZ-60zH3R&T!
z((vfW@2{0}$5UML)8hah$~cVdkQWZ+nV#gQk%b5;#cWH<A6l*ry(RvQwI}&HO@Cl3
zCf&(7&3StN&Ve}a@_CoRWnG&i_p;v0tePF8J~agW#EIA#c4Xr`?jwAFgi7O?$0dL@
zd$c1cyL2@oHkLUBj@jj+7^|2Wy3L>|=W_H<1y++EB$wMK4P5K8jf;$03)r%rOmjub
z7fOBY+bb-r2;&qj7*>cp8Hu(Z8v*m$75F9l_E6<H?v;*NlB<ah3tGC#jT?mxg>bZl
zI0_g!-~*BNOZjZ8I(!rCA}^oTG4)T47gfvPF>==9h3G{|!lW{knT;LO^z&8yg(K9`
z-DxEG(vxx|k?J@ThOzw8N6UVj@EvLN6;B`R2GM<7t+XzOO0UVUDWjAXvNyO`1n;vM
zO_Ys?Xfcl3#k;FPMz<W)EyT6(e2!Qu%X(AwQ=TS!+31^-4B!UKjod*YC!^zf3?8iK
z$-TUM&g=5W9S*XbhT_|kKD+SS<*y$TZ?>sC_JvG7M?0J+%ghR_r|Y@&1&A=2u$?wC
zq=-0I(epm#fxk8p97h{;Xxk{y)QAjse{@lKm8)gR|3}n)m(@lljd7hx7BnG8N3yy>
zr%$KF@`TR06W+L@iPu~WZ87ieQytz%G1e=)L?3JWz~${+Sj4saMwtZvv6=nDRV;lq
z80q5dZucFZa$~l}dLdXFaQwO`8cRh|qdaKp5Q}2c#TWm}vI2*`j+adU-QbPf?fqk(
z0U$Ie_`dmD3T-xduAd;y-B`5L@_@fwU{7M^KeX`QP2MO(KBpL7c`01NY13<LKECly
zq#COj@m66oUCS+UQr(UEg&Qm-(a8SsP_G{)NHiq4)JjjXS39KraP4@p?ZCeWKmjt^
zCb_zFn0(|=6TjudS6XjsBrJTV_75{Anhx7L^zw{%_1`3XkCAt#S5TLdG7PQ0574W>
zvW;KN;Rz)XurcJkzx=e~9)%ucRw(ajzM1H_%VcmEKilXu{*bAl-xmvd<q#`*2ZMoT
z8)RgbQIXMol5W)TmQZSFqza|)i=0|Zu!F)uzw0_>EPrttc<M_AJ;OPxu|8W&Gd1in
z3Kc0m1!eQfyvz_h$l_i^hS<~guXC|Y@3nn;hS;@Z`f5$WA=t}k+*u?Mj_M{Yu4V~$
z2?mZoL`@WFQCLuSYr2~-!~3Md{TsR|bHW$S<`fEwh}_}UyGnhJ<v$)dPS8qH>fcee
z0hh8k=E<vTDsB=faM~{Mtm|>X)Aw4ClFaWsP>Tnmr-m3BLy@oSE2a|=nCRM?23>_J
zPQF6)b9N2?6g9Agm!@|a!%RKtxVYhulZj!SXn>9}TEHoqcGKy}pq&ZW-Y7IYK{4G&
zd`?S!7TL1awft{qr~bQx{FegQ@mJZ1GM#o)!t&L2R9zM2Y!uWh;6fCnnz6s_?EV?+
z))BPxIf20`LuN6xTJ6@1mUfH>&#^pV=t_vvBnRIUE-l)(L~&L8`E07fN69a2-tj*P
z4*}87v$-gqQOmtmQgImevbX;{<P)m@Is5CKBF0B2+L4BQWKTjb!lb<EJV3ZaqWEe1
zrLsJ@U|xoPFdu{Xosoh3?JS60|E2#ki-vO4IplIrmeB5~zMd#^a{HTG{s{BV`~JJh
zNxaK%geT~9IN%RoQ1{P#Dd!{S|KSK$0IQKky-`#p`QB=<?0<dCMJbh}UTCSq=>}fe
z)XZH7ejvu1&#O`Frww^DQ9xFvnO7JZA(JzPTJBe#ajPA~KU$r;hhUFd^oZU(!rlFt
zw6}lmvg+(P+4uJB=!Ei)WkbL>lq**;_LGbAoe0R)>yTK_w+``$G_HBnt%~1P{tIS7
zJ{n3m*!=9$`-8OfP~`T`_81Tqeusm9xf_^WPTwXYXV?uNI(&eA@Xrp_|90h;OMXXv
zTy20rMWRt}X;VFqPcLwdYIrqU^z66y@KmFSbNe3s?v_e23Sb-wZNMbo@o{P}=+N=-
z`^rks!MC$qK>KK1t9c*Jw)&Yth3agBVsNG$i#Dh)gMsJ#t@?J2>vSlB=x%o(%tRo4
z{s}!R=eFlO>@$BDGNs6>ijhs6N;h$TQ`lZ5s|UAtrE;gQM`Ib9sNzG|vjc>Eu<e=S
z{{;1BUrKOREBdUW8-hM2_|&i2&bK!^^0o{0Cwd_jE3+r2E8?NtUmKbfB_8bHW$%kp
zwB%(Ck>wihb{cwYLfk&Z#=Ozrv1BEy;j^j>*KnhjO!%HKh&f)5Z`i}dalJ|$x>NPX
z4>##vRNrTP8T*AF)u&|-MLQB}y4F2waFgGK%PO1KztF53FZ{mvN5WF`GwwWkR-*QZ
zbt?65lhY{_FcT4HW+P8DrSVkOyA;DitDSxU5eG@c6K%2PPl;NB_y^l;3F=F%VP;R&
z{bcF)l0(4>g^)N&K~pM|^qozq!-nm<0#dQ`2G%i{mNS$V%LmaYv-GL7_d2JhUnuY^
z@4Oo9G(O}a%AM6Fn6=c(uwN3PQC}V|jvRbTt2i7l5W$!jU#z7T{LtJ&!=Gn>ili^y
z2(r=ib-}%MMbU9?tf@x*5c%L>aJbTRadj52EQFG<O`e1ee!-yCpsyu!R?iwmZ9le0
zWpRg7po7uqX5m3<!lKN$yz?tZpO{9x)k+?ncnm*cqq1zfa;z_(R1Xfpw%W29-hShQ
z7a>=w@i<0lB#e8N=aMk@Qx9jB{AAz_m4{3{k;vxH&%O=Dn=ninSGZArNJ?){nsmTj
zFP@k(6!s&&Oh>rAN{T11S6kHp^TpKVjTCbzOA>8<NxBQ*%)Ea^Ew~s7TH_ME6mp!w
zo6fQNF|o(;I*0@X-w%IHpxwKfkcXSYjIr$faOt`(t332_W2%@ec;1Dh-~4IWlQ*d^
ztTw{e{hH5euYE9_1kT*lI8A@hgc3U{HLp(I+!ejDWJyUbFm5v;PdF;6zR$`tv67G@
z8G2OWJ`@>~604i@3{RSStANSj*>yxmO5Rkm+g!rDzqsDslK7Qp7V@rGeG_B<#gbOo
zi*`lawy6PVp`a1muw5S|rnX}4ppJa6ZEmSDmqV_m!RjQY9oYBfp(R;9m1LUad=d?A
z4nEUtYAlxsWaZwMa~)lTY*5wSNUt<g^6uA9^*vomTy6cVS)G1LJzz_ir<&tdgso4$
z)bibWT<FWkr9HF~CZRbJ?+XR}XO5+eY39;z6UHvmb_U#rVI55WsQm)B6DyJ{c(rlT
z{_;!lkPheVYR1ljE{n2>ViZz4yC7Q~E;i_(JfNxkBe!sii2$1;VN&VbNZ{b42E77h
zWzFxX*ihN!M%q?}2q(#XrODHpVeZEzSo1#&w;kQmRst(RU!NS+b8M?u=;sdR3AY_=
zRkL_l(r0V-#&Zy8>oZE*cE%~Hn{>SS5eB8s-_M|dD|NM<)+f?5A}O{F6dd}wy4V|L
z-|X-V1{>zFYEv1-%p0+CAf+QT!Hp!tMLN1y`r+acE}hce1Lc&|WW_1R$?^5~Gq(q9
zNv1g4YoJj(jkjOx2W3d$84z+L3O#kRU#O`NA!ey=|DMt%U;x=7S5Xpx-bo2A{V0_K
zFHm?h?P$23lt#$69-$hmByhdN810@vXF}`a)8uA6b+8>$u;Wf@KG7lCzv%Dav|CKA
z%!TKNOjW$1xEn4xP@oc}>)4j$J?hs{Zm2c=Ny9aHT$?dFr)t;0Jq~V;jMLMg@*C^3
zjS;)CnojO7=jN2EIB7MLT4Bn68KH^cY&fiiobSdeYDqWh@lUtl%H@<-X6xfIXbIKM
zI;hcT7wOcG)bPx5nA}6|5p+;8Fr_fa=Uxbej`)sPjNBgaTk(!Q`+P%X=#CTpv$lZ=
z`%^n3v}<el<o%r;`l1;Pr|E$*0+Ouhy}k}#uCgTUhZ><z3cbUh$8yA%yi18-P@Nb9
zP3edx3ABB>SmnB3ne*{VRMCmdAx>j?((KD9N>OGbuP;ikdT-jmF<HHSV%M9DrM*jT
z#Hv%NkWe_-vhKCWB->g#r2`5#D@pm={_Pa=6bfAZFHnx!32#?~a{U&+sF1LO!z8I}
zm~JeSQi#JQ@qK^+B43lSY&u>NSu`5=(NE#!V3$j6Zmgxw+B>9Il5{sp#EH<pDlMq~
z!w?SBM9+klvt`MM+cWoS@Ljnw^Jf3WTITPzYK|vVD_V@LS@Ol%mF+#<1u~9OxKojB
z`-KzJ-D!}K;(Z#(;eqheh$PGDfXXhrIqRy`SkYp+!APui7Q0g8(|A1n-nGoJ5-py(
z;y3FGPrD089wX15s`@8!FN-M`j})`Z>i>WQIll67bUA=Io=&CKF10)~Ok^-yA;pSz
z>@R1&!<jtRyT-k|`qfS#<s-6+sU<jt4IHIj+E$R+Yry`^J9A8{{;4wGGB|WhISq1*
zYbZ*a3SV~AsMR(##$Cg1Ei=n!nTFy%Md(%AZ8|ihGO_*0@z5?D5)DqM&LiEdT2eGU
ztE)b`douRQ9Pu@_+sc{eZK7BXMX;4K=ixU6LT(-VB;5dw`sLa-K~`HGe<tNZ8qx3D
zA0c-;N=Gf;1WU!kS!x~+sZj*Do}J&QG<aou>6X7s{#B=vhJGt5-ZQS0g)Oh{p1#YD
z9xvg(L0#1T@N^&N*jV0o_>}xhNPPYu+%A22VkifGati%31A`#=jH5{isa0=4q~|Bq
zH^LIMhgPC)*<uDPVkE+TZJfC2P@gv!eJ?({{qwUOTn2%H`{k>xmz@3b63S3jE-3O9
z?z2T^6q${Yp-6L!GjqBzf;S=8dFi>pq4b{S`Sg`jMW_C~A>^2;;%=o!{!dW;koYs*
zrJ{$PZri{XFh!+_$2Z_}B*eWBPo>2qY7?u9@0u=zZ?$Ux|B=mmc{&L>VrmDdih;)V
zmPTFr@#!9xwTsljMUhC@4l0pk@Y+eGg;vUuBA{dq`?{QE1))}FbPY?qwY8M4h@uXf
z-$Doe#$psCox538mwoQnx#)4xVoF!y4}*>4{{GMPH)Z{QRiW-dx_Y{Ytv5g5xN;HE
ze(F86j%6Jnm`z*<;L>ryK^$ovB4yCdW#>kvJ0AHf#@nbvmKpJp)6<LRMgj4@_DQIw
zSXshbMQ@$#CngU2wliF)LuPJHsNgjX^RPaq(91}?Q@U!H;&=i)Yd+4E$t-Gn0eC(D
zOh7`1Ol1Ypp4NT_h5Kg`({zxFv?QEI*QFoO6<?~)&iZQp+0*r3@REY~i1A}A2@~A2
z^~+cQ)kzm$;-kZqebZ_5ZxP^M)SJ}%BmVfExC0;$Hv)Gw0fGse$p9K&_>Df92K25U
z2a1M%CEWmyu(>{BD&`#5+Ba-wm!p!B`r*3ddrQ4a3_rcuzv~2rC9dcMgLbgnEmqsH
zq2G6K6eg*3+6#Yc%-(Jj;V-SJaWmG;r*?%(8YO)?Bj|EDKMr?c8miX}Q(JGzk<Vjq
zh^L^D(;OOI?sBeNU!NG<Z4nh9DeZmcI`6_dr!eo?Bi-FSryhFg8NFpR=YoGUHEA#i
zvzF!3aI=i_wd)(;gwxKJFX6y(1ZMS}<TOt?MZilrvtvxGU?(lOQsG#k{V20=TuH}r
zzu|kynBmU|6<QjW_*sX(3~J_^%NhPQW-vD0#+pXv^Jm~yR+$%lq|<Exg~NSXljv=Q
z5Sa+YLM<@A=ixR#)oWK<h}>sp%{-z06*0;lVuETpjw6j;s8I^S-|&2&BUXZauPB2n
z>#(V9g~F!hD*02cRSp$Yp#zNR*KYIJH;}!Xq7kRmtY4dkmMMRknMDZ8ikFq<?@g1*
z(hFTlh}QwzD)^}g50gv%D5+ZY9nkY&GK?)RwPdj0yKUZL?&?L-{xzJlB*2*HFlMlU
zCUdxfAz4!aLUIr!;A#MIi0zJ|&%wzPh~*F9BMGgR|A^>FK+T?Ag;}mcXGQ%BMt1tS
z`_s6r9myBx$j8jg{H-|HDZdr_(S$ef+461q4S+kd^BcjFBZ~m=-`1Dcs(DT5!dRMh
zDzVcM1%12Nlf2q&|7*n>k-miD75F|JTF@0XBIJcY-ZOugBD^uRsMNwa5s<WKoX~9S
z%E98QC{^_eQfQk*ykWJzwf)jfJ-*s%?g)R8f;6S9Wp>7Mjj9;A-*UVw!g}=VlMueo
za;?4Ft5rtIkbW_J&9H<kRt*h<Y8wb%d-%uVvHH73oy6*;T63>;)#hBtF8stFd)%GX
zA4y}<BRXi5tMyoxVSL@3mEIq1Sbb^DD_}>KU{^a+&#v*H`C}N-=J%~=#vb)YJ%#ew
z$u^K)vF9TL=J-lwa8c9o1||c6{UZ}tThn&2n-83|aJuw^g>cCIeFwJST@+Kent5;m
zk~2ahW;AwGVMcwVK;4pWN{=>7Y-9`%I@{&S<=6NV<U=XeWBhRKVyL3rE(*ZoQf_(5
zbY9w7+*Y+gyddakmd7N@$Mca!SPC~Lj3>9r&fh<($1*REp!On8g{TZHu$9C)C&Dq{
ziRAmDWf<-5ICy10(?YGX9CJGgj+}V3oyUevPj8JFS_0)s#qzFLu41!Jalu1)n!oLG
zGjFY=gAtM*q$iiofFD%ZwnS0?_Jc!nMXRT(VShS2M9F1K?jCd11oqG`W7k9j4&>-`
zAC!z~@^5o)yfWdwH{Pf-x%Fq6!D~BLsSBXUUwzwI-3;eXkWT;2U%u(jI{jgL7ZKm~
zQNFE!--x-ApHa0gZA}1xaQ~nV+Xpou<3$lJfH%8=I*{H&^cOf^Dylj;n({-@+Sln8
zdw|RROkq@<aCk{v=hw;~>&klehZtQz$tIRZ09P?-osm@l#US>cQFq$*3A2kBu)Cf~
z3_*MPbuRg}a0BjE;NYrZPG%7RZ3;w`0C1r_Gpe!3rn}cIPnumN|MY_aCTD&?i9ltC
zt#rjxppx`20Qe}fa>~zpT2P};)rp&A5OHWQIV`i~1FALJGr1{+V*QR6AvGj&ShwC9
z*vPAXq)`ypnmWl9W#m-Rqu?E$bm<MUSA-doHnqsp?;m}}H*vU(*5X*Hhw+B9FxHhl
zv)y1=z_~!3@N`t&tp2v6&e^Z+dnFKGfoXBxE{ES>dj8M9SpQN|xF^j(c(Yw8&8nrf
zVq=7pZ3NxtC^&T*fldE8Osg?rwaqRv;$?29L2$yzm1G_dpSChsLB30)$wEKzhS5(D
zjtK+hm+nE_wS=JoaYDnompV;{?<VVaZ9FJt6}HJb*%ctHZwhn!%AEvhEsWpVX(!ho
zD1OwW@xq4T?KA0R3_9xb4B4)}$K^7#t}8Qq-Dsynz|@mpT)-2pm}J3E_!u3LXDg7?
z{4`I-p-qfncLH7bLHW}CY;|IHwslICvw91~fOX&seWT9!om1HT!H@53aq=s1gLzun
z2AUJBWSh3~Nla4DlpAT57<ibewS!IJwZ@HDK|eva?JUYrH9Bk)VsyAAnqWtv1+h`1
zChm)P=WbR%`n?d>kwuL?HZw@jlcZeD$HlSoDavV2KC*}rcc#0v#529|PAvY~-*&Dw
zCPLrWXK0wRXe3<WG;AwCX5VVOTd`*Wge8f*NpKWy%+q64LRVyRC*A9{969Xzu1YU{
znfpCKsdl9(9n)}r_I@exUZpBHoyg&|cfzqs{#9DCFpX0F%y~}YrDV!d0;KS=PRC8Z
zqKxHm#++)i6ge`fnyo>k2Df}E1j}+H8LM(op(ghiQS<Y>VcuLz&Ezm!LTR6`S6%Xs
zKPd8CSr_VMns#w%_vHIF4@33Gh<d%~2OsUoycgTwEh<IkJyR-t8NpF|ro2k#DDQw&
z^EvbV`mU6J03JjcN39YB-G)6HQyP}AegU3<yLjWRXpXDrkIHk~j@m&=yk8!$!pPi_
z>T9BjDot}+^OD*MPwm4Sos1TG1O`$ikOdLwqyj4go4rQ{o_F4!b}-cq)WqA1na4%>
zqCUpjgkeJ_f(dhYL=|4jLrz#NrNh)MsTHKu?c>ZW7RTY|F1S5Xe?eRLr#1Vpi=Jhx
zh=`8FdDL~2pVIS3C%0?^i2y(-;FTepbxNm(jtae@(a&C=vx5Ikxu1#7I5K~(jUfL4
zfcNmj%^)?3EKl=Lds;BPSnUTy@yYe(VBs^PsPYfT#UDmAr5tM&@FA1h0__ZAH6U|d
zr^iv`Q}`LTJTow>X-=1x!%hsVATyEHI(|QMy=aC@bztce6On&K?_}8Id42?^`z-hv
z$H2q;dIE`h1+G}6T;luGHao>QB^xw&2@ms1?keHDXcy)qd3Ns=cY0S*wqbIb5dlDZ
zyG{aTZhZ6^8Bsejv{1%MQ35%`{qF7nCmg7D@R3O?s6VY~mWDqfev~5rGF#)d27}GJ
zg%|<z%pS%*H44jEe`BA?wanuZ+3&xgc<4RyeK_hhpmYfxHDk0KC(h?Bj=Rsi+7{<y
zR=XNSW5JxJaCv#M`dbC)uxPhSSgNJd8);<e$0j(v*Nve_z}ei~u<ri!Z6igyH4sMu
z2sf!Y{!~NG8Q$gm=&G7Zm5vA~5;F>emP)tmwgmp`M;`q<hY*f$jK9(Clas@pod?q^
z$PQo|)MW_|w-p5Z-oDT~UNiKh-M4D;U1q5=Urn}iseS(b`=`DPp5?;}X~kmtogekS
zqHn1VMgt8*)^1sQ*fe<^Y}+uuO|s3h8>+ib&q^J>e`dgGIJ(<J)wcUXJe=2yD<JA?
z<>5fHZ*_OU=Ea-+73eh=T>-!iQtXiy^MlZP^^Cg<nanmNJ$X7Ne9I&e%ykGe7=v9i
z@^%aL2i~-i?ylYz6S7;~Jr2AnWeDIaGzMa2*}+VVQ~8fJhX#Maz7821`L(>d8OXwy
zd!DRr<=fvK^$JJdwzZBZxyWb#rj3mEW>-S9x4T{cD?WN)X}4-yg5~%rIcEYcw)=*C
zOnX`G{WTL86VCG#l{Vi&ORBwMlLs;-(Rf;U&fP@{@U0jYh+ucBGEe4!_IS2(?pT^x
zcaZ^?RI)>74E!@Z!&#U^aZojgz>-f}da^1OyMko3Ag}PAIXU4OIL%4R!ZD9_D=$&&
z&cPkR?NK?V{1}odlWAg`BnzXNb1$T)SOI&=W8V3cCX89a@az}309A!X>tpG7=nibr
zH@HNgBTKJ9lUyrDrCi0bN3)B5bZax*wO_%qKf|!>YXSeWx-aIWjB3OJ{tMGH;i?(c
zLtz<lL!p@wV@Kbs#UPwHlj(WgsO_;p()?8Xw;-A;e<Q#=D2b0x(XAw)h>=Mnd?UWe
z=*=mZ7zmr5Q%V|UW&^S@yg=J^pX`j)ZO;6Xa92iv4(+@Uwik%XyOqe6VNJ?konf`_
z`#)gh{nDcU{AaETR)=UCk)H`V(nFK9EwInKYY49Kk8p*}pj)@a?nOf&w&QDDI;Nz4
z6Eh7f^~?dJah7A2{kU_rEM5A-O)LI$QO1Evr~69rqyah&UQQaJ)hHBH!f}CfaP_?W
zCx|Y}J5?(o1m@Ir5~x&ps9F#*fRE3KkR#?Xa5NlU%TSU24`UFpsviHtLq853`b?}v
zYOD9?0zg9Uz1aXe$V1PP!X&fhDWKK(mL^L>G4>a*NW+iaCp_GoJ5AXAafbLWM>9-f
zd!|9lu`;sd>24CQ(a-%)x9KnLBEOIL`kOdDmuk5dPjlFRX3V#EKg}rlV0pG1PuFE#
z+xvLE%XltlmlLP$Gu%YHMJBp8MR<El|J=`;Ei$hQ8bER(#W?vfJ`XIv@3J`Fe}QQ+
zt;fF>U35N|lF_#@<Y(C9^uFk5m`1+s5;Rl~-whK>4M)QIR*BLkmggPI{5no^g@zA!
zN2^Ta$X3g9FTUIGij7`HB5n<rr)P4Ei>i_x#Z`^VmNy)1O~L|+Dt=Fr$FQh_rcIU5
ze!_h&YQz-3zK4-8!%{h_e-PGRlK2Luf1iQn#qzyoTS5n}OU!Gqr<{tudd32wb6B}n
zjl;%VQ({yk@xwB#d$HWJ4Kg4ttKV6=V)UzFe_T{^<LU~SZKAk4I7R)7VPK<-Wtku7
z8=*Y^Q2*bz_MfEQ{nF?akZQq*<snc6J8di>nUOJT%MkzVDby^Gm856#5zcQXGlmiq
z0{aydNC<(Xq+3Hn8|y&-s22Z88-KkFK!1^6fm;|r5GM}A$>md7WFkKDYRjx(<&@(_
zc>}6WL<t1qKCD%>dgeghsz8=10As%AW)MhzdML=@nL_jJJyWdLNUD{sK@Hw}3srtu
zSTmt8fuee1*t1$=7n&O*pIWz>`gUtH?`jqc#sDY7Wpqyi&0J$`jI@KUt_2G{VN9Rj
zg{IWD7mq0?YS5H2W+z;6+$*zK4&IR#qnViSay1fZY%%Ro87-|huJ7u;Z_TqD<Yx0}
zUI-C9XfTmlJo7C~T1$2oY@`+gefGGTyFECwELYsZA<t1DkCH*-$+%ItFMRLN*u?a;
z$Jg$un#9>8FVzlvmi__Zi9BOy`O}`J&5AOE1e{@_jnHhVJno&avDJE-f^?T4q0p6<
zXo6&KNrk%moqmu>-s(oB+D^*=Yq2CI9LGjFGG1b5?z#t-&~OqBNXhAD9V57dzy9RH
ztJ^<J?y%!g-eAH9F6X!z_Zcra`9Au+lUs7Ny!O}qIMxrIQrppfKQBPdtx@4En|0DC
zE2LUuamUy)zUlK8)uX;%8(-7F7zK5zY#%J<$o7!X=#ig8CPm&`a^J|fL9D)+Fx$G&
zGJoUr_!0uuP0w&B8tfaK(30wY2=?}{6!-HZ#+Y;NlwHJTJe*8ED-^8*4ryS^_18ha
z(`R!ko`LEz)G`7d8OC}ub7b0<7qo>~2bda9C><~Nh2F|<prGI0MStb8D<7Y2+qg|e
zI*Z&qz9mj;{3sAd?l+PdsOhf?^e^^LX#Ha8C#d2)khH+b%XydUv*N)@{eem1<7#nS
zE9n9!0x5%ho6r(}${$3&L%n<{L`1&?fc@qaEt+it%W!7p;ZQWDyA=hEDzzlHuN^{J
z&VGJEee+;(2%)N-A(ymX6@TS)dLtnOxw$~?m?~O@`^CVA%>MO+@+wrI%-2qKZ}Vwr
z9oqkGoxtQgMsfYku}Xb1*r;dzK!kGPx(Uj2mPi&;!_!|gk>^1jM@1jCL>2A_L4>%N
z#6nAw4jx+l{dlm<yhzXgkaW72fFSw{Wb5A-_~%ntVSYuAjcAQOU6k*<;5nW;&^QW1
zq(N6tfhFX^ND*%cu`BO#o8<0y%v@xHqL;JgybmfPsaw6zND0$%81Uq#lvJfPqJNp=
z+obW3qxcNSWgq=etF-QrQy47-tcci03mJT0DwDDt3gH!hE{H;$va?88W`SyLyK8;{
zMqWQ+q2Am@(e;Cz3gBY_+*mE?^J<t@+zTH+*<;ZETo!{Luv{?*a)PSS!E-^3P(9mi
zkjj5pVFnl6^MAC(G{x%X$DlMhG;Jo{NHzy#8_VK|^B3SNNDezUSux3O`@FR;y05}v
z^Wf3R*5dA}cqXffzn=%w5hxR&37{2Mxz)Ay{Z6``$9mP4O4xGyxt9<cl<T}Dp$*VJ
z{W)?3a`owfgwHN*ekHMgLBN83!_eON1^WsD`qz?!4kS1^y@*-^363JSBoaaMs!oO1
zrZ@g4vH*nIAEsFB4SlM**dG61E4wzmx&Qd$TJ;kMfcHEB@ZcLhMNNPj6ggW!_=Rc|
zUhYsLoD(h;#`dU0>^)#iWI#)bkjEp9;m9Q})kmz1{pI$Jp6P85^+z#%07uwQP`NVH
zmi;FPK`3@@<Z$>$?B{=8xBn|j;;>yx2k65>TWJjYoUTOrpZx@(pU?Z}b^bxj;SVk*
z9Qk#YsB_T&^p(Ys%`P4r<=eIxX|$Bh$M2nNOghcJ8rA7GXG`TtZc;Wg@w6g6FC$(S
z!8sB<l(Jb_<Ln-BulP`$Hdz8O7hW_-3eQH<m-?cTef{wbS0m15xe**1vAT!&wH6bJ
zGTZX{d2RCwau)(8k7If!jf^}aKd;wUt~*8Y7WFcj3bemsH=s99HqW_*#f|EQz_XGv
zK0Wl(D7z3D&fuerwF7_Y*>~Z86dFi#*`4ssFkvO((1$Wls^OR+N#T7EYf%(H)ueH@
z?LnXEO+D(c0l`rUdjaW4xOhxGPW8dh<Mp5|-A%0I7<$g<K!<CWJYsiu+cI)3-On(d
z)*U=`LM|GOjPob3xN|SXRZ64zRr&(*AxA@}&4ihd&kq9us8NqsZgY3&c~@aap#W@M
z33d&MTs5cV_R;CV$z6WK5(q%9KEr#AQ}GKtCU&_?8Q)0gc9JDT3z^9Hn?IrBUm22I
zVv^!2svQ)&v%`xPHzSQAQ_UD7EuyrsEIQ(H+$%TNQ*#Wrm~M8P|MG8#aTdXbCp_yg
zJ=B5t{F7w^&c5uO+LhU4=RKUN<#tylJ?t_e*HM_pdFbUus1<s*Wt+J}<D=6nSv4EO
znz55zaqNAAa-<&DSKga=1fK3cR{OZ?hQqvU>JIF1VfGr?(|qiT?_Ij9=sVJ{?n8hR
z=S86@<WqNrfe1hLm}86FJn8KI;6*mw?ed`{HMRTOi*tVK=EgY7+Bgf^-0Fw0yL+OX
zC9^_{h7C{16!Q`*ibpm`Y^v)@$7mxc(Ko+}Fo|?4JXGU?JkH^`t6OfDUOby*(ALB{
z>u43{@?UG>cV9Wk<`~8+_+ulU!|hNjUmLmJrxugbLk!x9nI;2gJgmr=XoUt^mhO+L
z2+1qVFi0v5so#y@=kh75+=ktt{9;oVoFLEIw!IKuqn!urEJZe%p)OG3k1RF7Ch?<7
zuFak8I0q`U^D=%Dd3L~NbuT6KJ1A#vj)BW)M!4k@0J<=iXHlQw6n5sTkJ`9NQxvAK
z8bUF-Qe(0#l(z+XRU$#ERA<(VD!9fUN4s%dQiLS7)~c{Lny5x$PfgXOTDB3ua*n1^
z0jlT+t+h_n=d7OlM=WQtA8%c|pAA{#o0lEQV3iECSQ;Nz$R4=(x^M1V!C=>)Zs?@=
z_BJmqh1^5Wm{2)SVjQ%Z0QyOZew=ZoDu>Dzf8Yiei73qOeuOZhbJpO*8NfBJf^N|s
zx`-S&`T8r5Be=2qRcCb;H1;XT_$AOY#uqN57>DsX_kbS@Rz2wsweCaK*x3~>5-LJY
zzIH!B?;%dZts@K^X|QWGpcV;@pP;nn!vdXiWQl*|%__%6jz)2SQ^inqG?;px>L+Mn
z*5DXKPHkW2{}WWN+N7k0d1wrf#%Y-hw+_P2Qr5J9c7^PM&k&pv%Z_D3XNcP7R+G{A
z-71MC<mxz@bf1D0fqT}q@~aw?t&0g35ehQlFLP&HXM);S4{sNaUz1DiYoVj}Ny!3k
zQ6N^vbxur~epM{nB>D<NJvaZWsX*{gkXMUEe0bRAQmJ5bXrte*mg2R_bCMo7-)caG
zPLr>aH~>hOahY0tZHn*iY}_P&77^CfN3j3?C&&%r)DOCPD7t@CcNX@M%DK&F-gp$w
znzy?6-8axcV2ah`V{w7MiMijBdBDQl8RmB#R<6e;MK6BTWsL`))d$o!z&92w)k-j|
zVe<jW#DDXw&Wwnrbc7P-jtD6kWHE%Ip5mHd93cxTgc`7^iCkR%>i@}4P+9XKj8Ocv
zcrM`bA%R|Pk2=~Kzz*YlV&lCC<mjAF1H#9mLUN3AHS%P5Cgz+A0YEuS#fTiNG?RV7
z{=!m2we7a>nZBPO+~S4{UlS8ua;lW^;<QyjpCk^Pe6E$X0%WnTBcSJ3_8R-lq_^u9
zJ9hK~j;pjK&`&X@Ry`S8Q+@jjvQ7O8s_5Cdj!ED}!XGgFJVA9$)g+7A)|}OAad*qu
z09+byWXig=he%LdgFg{5A<`JSx6cE}P|Fd9(Y+hvzIz&TuFbfjZwou0DLKB{!S8YC
z5A)rFFD878&751&PmA_|Cd`ylU=r~(_88ano%B4wiq*T*h>f|K!0_5jR|DoQ2}L5!
zl)9*w)tuRrQQzR;)x!=@iwIXd+H!v4NUXo1Pf|o-6Cja8&z5-q1SN?!c^+ba)<hNp
za-{N*pb6M>F9X<RK3<4CC*FgMwqd}AR{>{YI;fu@9u$g**t>Rk4>Fuz05-zu+IKGx
zIC7+;D8(HJRjr3h%$1Fz#dNNy(a)Q9GYq=wJ#+{-MDp~vd*@27BQi<ZT&K1j6865F
zvF=<BHz1gce}ZyAHMN!l948gGWE+R#Qy>`u&C`fJGPE0(orm*+SbNjZu6;^%fBmMz
z_|UatE3U8K9X5BgxHkU+@a)I+x>pbIwDyg!bf$lJgsjXOURggN&=YJN*4fZbQ6y5m
zUT|XgC0+C7OJs&KIdS7`0FI=>V7uFD2pquls4s?Xv??3hE}u4d$NTiQ=d6Z!6s(LJ
zoS0G-HTNtgi66ND^RB_I(_;gzfEaEZnS#@%n4chSZE>aycEOf=SN-EHx(k|Pb1mW9
z$rmK<0?gM0`vE=0MZRjS<4Ju(d;g~4bS8h=zoz3V7}&!cd!bvG1$!ve+>bbLEV~dQ
zrh*RW7tP84sLLq0E7-yhAWi}nv4*4RP^pOs*i&sXRKZZs)5ZR_G#Vp6xwG&vAar74
z?i}FX#NC?D!DO-{_uyzFUwI%0!`J4soI!U82OK~vD=$at@l`kA75F*0=gvl9fb&Ag
zVp$dEl{uVgNs29*M<P&l%_o!4<_0*|9N^RZitKizwfeeC<~W(Re}W8rmQ^v-C2m)H
zxCm)<z_)3jPk;B6SZd*wlWWBpB!6RA^MGL)I~+62(LV7f*mvql2IR~i&@}%PFm3#>
z#pc~C7m33ftyREIUb|zlhc#1(g#{D<cKD~GxmM8%;Km(Q^Wge;)!V#m?sC-Tv+xr%
zGpln9l90U@)g#-aq{<@(^ye4}Q9u8IWyQ)-wp5%!q|Nk%*aa|TOv0?riYYdO8n34D
z>IIQICo{^d`7l4jN6XO<vEt0U*`f$IV$ZnC_71wQ9s6j0t*eK3k;&dyxBkupHz4Mm
zlJs=zGA4u1IuYlB+L%^qDXU8p#IJ$PXj2kwc?O`#Q)@YzG`D_?=<Q8$8*+K^<43kW
zpfq#n_3pA{;LX<)*qc8tt?1lhDSmh;d*I;bucY-8B!f+F<l$KsD#A>rh}m^&JJLr5
z3<r*<_}Cdp;6sL5v$Tz$K+#k41uHcrbdO(Sv(noe#QA*hc${%%?{cG8JI&(C|H)0m
zO5_0EsI7T$Q2B!;Lc}?v{_amuGapubY~%?i(^&h;hC-VnTC5AK?Jq%g9Bkl^#epdI
zf5=Ymu*7S5kxyeHcik{v25}J=XKq`?g3?^_P7V0+wD?wuie6|POjC^LPR8(YK6dZ0
z_*4f0?BA~ZgRc-k&)neatt#T>P@k(Q-QwP&=izYfv|QaIK42(U{pFJ2X|MpjR`hGd
zqzIC{=TVR@sC2XdRjDFJWjuLOzrb=u<V)8F%%5eZ8rq5`HJ^mVSWe!HZ&~AlnDEjU
z_m^V>#{<}=ip1NBGZClKIBUn0Z(;s59|gdD)g^Z+x-xI1PKcR~c1q1L>h<!j81KCV
zEZnye<Q5wtq>?>WK{z}LpY3J_!~in9s5FyZBpk+g-8yc)eNnk6yGZgK<8|Y>_0~8P
zPvUk$$Z|JbI!JXq7BAqod(?8b#IHm+-cZ7$X_LrBfCpGY`yo5#k$ymA*6+k!j!OU%
zMdIN%%mx<9^xHh*Qsc0nYcP;b`+un%{HPM#`&R(aRYs(l{~OT77c-~%tD^HYiuLDE
zWvF5_%Iw?FH=Kg=E#N}VR$EZU2Yl7TF`yrh4#V;K(jQI#?k|^LU{nb$xj&JEL>Iu>
zEd=t@Jt#SNZ+k5qf^G#>MB%9>{GtD+m=s)6@U&o<0|zdCM`DahAg7<MbJf9HplA^K
zI}Y=wM0LYMyb%lp|2Xr{Vfgn%mDvlob`jabncq^hXU5R$OZ<{N=%Q-znOnOekV7{Z
zi1X=(if_fQu>%GYYgwEWDk@biK?{&G6O&Vz(P@8T<n#=;p|}LHGt~hSBE@r=s-Ga*
zub2apZNGHERsW{$iyAZLMcP4(4>0=bv7P-4T*qBi0d%8%%br-W<88MQw@;mxF9Oz|
z6Qf+dVXub?Sg+zp09qsJ_VsANGuFyqqTu=sC)Vrqdl}g6DZGbQ633Z%;nWAAXmhb-
ze+A5qpP;!w&IEq5KQ@1YMC~MuhdfOEJplzuWyCP({RBmsUjIscV#3f*(?vbH26|X+
zCb)S2cb_uTCT?sEf0+NE`0{XtmhRf~ccwPIKCWN-9>8J@(j5q53P63UZW{knv$b`*
zf@J%M%DB3m5yGI7za2xme^6O=aox9Ix=TRzS<d364-C$Jg24XIfN{C8^)(0)a9XgO
zXp#o{|2+rWh{8PcZ#6aNKYYAj3U%{uwd=;;_MKs?Bgo_fwc>B3Vc6>Wl>jtbVh}{9
zEy{}dt=QMjAa;vyv8$ol(B(Ry=LVuS-`??Z6_aNbQ6{ACwN8e3C;$>2_y7QH|58jq
z{hk0tqQ22&ouA%Xe}1n-h0LF5_g>c*31N_*zH@5*5>2Mb?#aDhZXopu<_}<aO#yFV
zGM;h=h#vvmN*%y~ZJOsf`k|XED74ng%kGOJKvYdFbCu+RerWx)>$W=bqxNIH+_)?m
zE66z&F|iBpmColI%Sq}bqjpEUSlT&55gPf5{Y-zbKHgBtJBQVVDb<24SHK}83i1VB
zxRyn>Ff4JqF4!aYjaR_R!N2`(EVoKeZLcP>^TaUEib?eM*BoQa9a~nlF&dmJ3;4^=
zfr8j;x%!XmPB;58z?L2D%gA21>*}{l(zeBf?=KSa%T%KX#2Q024H)c`to`c?pZp?I
z<KNoCcbdbC{e@T?_^(Uoi1U7FE4-PGC_IMs=;mgZifDiM%7s*mPM<j|=DOAKK<$(^
z{maBaQI~{jC_@IY2D-h(IZ#`>JM|m97DVFhyk0=y+dLOdg5GJKn4Q7A6-7Ibsxap#
zdz@W-n5y0OS%EgAT~Rl|etv0wX|rk5_!hHW9<;t*GbDze!W%>ME+L_fYC7pYv1P5f
zHbk}x^|?_h`q8xMpFyGF(ZpW(&u<0-!|`AD!^&@NiRB8#qM``My>`AKxqp)JYfJw+
z77b9yB1Qs)u(CWZRM$a#zzkL%vw#~kzbTOl5Y}EX9|!*5d;YHs{vUc!z2dp=-_*u=
z@I7K8<xO|v?HPrrfo{b&88|D1tVNm~el_BFi1LsJ*8)Me$dRt|zMs;whUXa8R7Mtr
zWLM`SP4AJN(&@UY>fP7MMQ8AOPiHq+uN#tsguvf+_|>*b#VUsSY>-`tqe`MqlaCB8
z;ufP8w%RdKaEWlxy&{zd{)DBf-Iq@1UqFACMSlOjS(0lT@&RHds1c(ND*b{WP{iz)
zM)k0Hwy?T@n73XB&%eK9I8vFv<^-vZbrJL>&`mL3uWZnDEZiFFBNRVi&*)LOb?AwO
z;%SR&!e>5Ge~U2&`6+!_-pB62#AOCh1k!1+3thHW1~5MPU;d^=yWN|9O3pLr#j9vE
z^Yd#?AbfyM<DmeCZi+6({U_+q84LBa&FSl}T&JHP3ecm2HiD=>1s1Sc@E}3uTCmnM
zqbQh~f?d24|EdT16V&nToqxWr%h_9tYlT-5;jP3dZlDI#ztzAZ%Gn02jENxWj%4Q0
zww<8dgQ~kj>9jw(R;xMYms)T7DST;1FV2Re=^O*@<G}ZU3(jo^5OaT7WIzYNk6Un8
z;+8h^iXR-KdmqT-ZbdH+grI30euD0U7`Bvy=HG|o9|=rfvw_5Oe;Aw#enBuez?Qz;
z0J1XP8{9Y&Kwh(hWXEFhfbZwT05s+A8|l<VS;alAL)NdCc414cl}znlST3T?bQVVM
z^<8x!B{<lzB-BOOFjPP2qTPoMb@ha>{c{s_4Zlo(#eAwod~78}F?}0Z4L+q<aj3y)
zgEER_D6R?ZCAzl(9RzMo<;KwHW<;OPADUvJSlaG=9S=j({cf3gODovqeIvz@_{=ps
zKDK|Jr}epL1>!Ns5Fm>yMd#52pV!xNNJIxP=fpo4LkwPYVPfFd3NE*dJlF1ABrd`i
zuDda@Z~)#skn~?dDgO%$P5<%*pvHp#sT%*Ij&rxYUJsd&U5Q6v3_-bH8VT^D#sE1-
z!~GR(K%Pdy1=k+65-afl@82h^TS1970qZ`m{eeN@;KNcKgWW~kO5j=vs>kSp0OjpF
z@G!rYMP_sa-noch%vd<;#C!(a5_W4;kufEmsAxsNpJV&)YHa|hHQv2D&ei9%)IR{S
z2^m={1KA=vU~MqZH>wnCH%^~`aOOj@141_M)n<=Nmj-r}N{HeIsIgkvB<cMd#LKaf
z>;o$7DU2L4V6~+5%zNyAH!;9dyI9WH40y6UGqzBsb*Lqc`hgen=vR@#Pr&zge#5C(
zuoiiDBh1d_@HgQrU>KG~hr6R;u9VjUv78h=?(g1aYG%}ONwoh92T8Eg-O}{iIR9Wc
ztnnGw%C7<JjZM#tkL1QLtwT3@*X;c_?%p%3$*pbIbvi9m1e7WSMCl~d1VWRd^bQFm
zbd)LyMXA!niS!a6Kq#RJNq~f2f`Fh?dat3Q(xoa$5dnQyeAoKc_pN>GAA9Y+er^5)
z^5huLDE9#Fb6n4LUtVYq+WX}FC=FRqsGe?y{FE~AcHzO*&cU<y5G|AY5%(`#V6_|Q
zzh3*mnh&2DgQA@reEG%*au^6z?#mp0zWD0o=x9%>{q&ydpZb}*v(+4D`v;e8SXlNY
zLmE~}3KfSY5fIxaZ!O=AvKM;ioi#|yj6Z!l{OsXPGqi#KM~GxFT1+Sd=cAAPae?*s
z;{BidVs8D^Ozr=5%f7-&F*e(XyIwlyx?|(cd<v^1$arkL-&V2{<$o;zxmDQ)?wQq+
z&r8S>26nPtdrH$BBFdqNA(aeLVx$c^4YR1BaR-x|W)d_REFX)RkzA}!((A$KZ{Jve
zduyvp9dSR(hO_m2cVfP4`vImw;TGs!RtK*-HjV^~YoAq9=Y;`9KnAZd0g8~AS7w-7
zw`Hd#l#Aw8KM4aQ`U?blLB@=tAtZQ@giPrKdA}e>tw?#lxHI~SkQ?nV-dtE6*eIxz
zJkGg2(z*;lHITi5K>eJcO0`0|wjiA}PJ9dsP)?#<13XtQlmrBZhX<m?9}O7j_Zvcl
zI%%}d$93%2f^L4|l(TH)l~)?pdN_swFt&=-t_yTTTx3+}!T{uS+M^lR<;#n;5?QvU
zG-JApo4@)JY-k9Cma2SCVO~0Niv=j>IlcC!(xOZ)t_GMfXjl8$LZt8Mp|^@2%tNX#
z?JV{^6J&0Ru$YmPqC(#AGi)#jTeGPzZWDq7ElzIW3(2%SHPO`4-T@86&g{H=JgGf1
zGlEGr%zccTDLywZcGYm%Q%M29+gf$E*l18EL^JxLLpeV{&hm{&9j}<r!uPg${1l~X
zRdu->-}-tKHD1m!_@skDQs7UQ!>vl2gQ5vkO%9Y)7ihPcHk-&@Z!`MRcDP7{N>ctR
z7I!Xy)h%~oNFwn?8kS9M9qixZ1|S4<+(_{Jgd~6MXrQ1lgB-0>E2gIt`5@}MAmB3v
zO+~1~-P~+UF6rkq$FbIp+E#Kpxk)JA<d@&Tj|p`7Io7_mFl_G4vM{EEU~Z`fLz+Q8
z?wMauH5cDm=asTCV70q)Mas>yORRkj`w}lQx&DB&(*hlyx4V*ZJsVaiV}HRD2R&?r
z#v6G`jFJ1CNe$5t2zo?OlgV-`X;%1VnimGZ+i=U+bj19T!2r=ZZq#*2!m2_KTig*f
z-?oeX<yj@zF)_B(-)C7ikq$%jG1wr<f?z%nH90hmmkaiZzLe0Bz;fiEM_#i4c^9`Z
zy;e0{qF1>v^Eo<m7a)*hL}?5B{t6Dn;)K^%bdBPVuGKC>p7aKkaEsp7Z#lp;VAd*u
z>g+|1r^jOCfw8AUtkrSlvMBEH20VwnZ?$=jXDe5_EOS<B>HBPaQmXKhb_%y+yRSwh
zvykOhZL)88)WT|B)YTzVV?%v|4kS@Or)iJ$3hc1z%(72VN}a0?#dmKjLGxD8_K2HD
z1EDJ1DsL!WV_rrhi_jsffY%P_^e4-eIBRmtpg&Ja%K}4ylM$)0!^I=GvjK&=+*)rC
zlp<WcQLTRh&t<Jmkn>5A`5?fqXXZ5NaiV#1Y_n!X_r>Fj?D|_6Wg7%mP!*P%y-~K2
zVviK4(%tJ@W?<5R7@BYU$8W|8=FPAM0F|<bJy^fe`_KK=0`0b@GfT$L-f(HEx6dqB
z%R?l-xxH5~MdXbsJ}dY6l)1*7yJDI~8+>w`Jc8lr<}>u>F*?i_jEcs?U2v!quF&ht
zcJ^lHr3{X9UqmU#EF$kDI%qkklbRo1m9xtAG@HAwyM_K`$O_QH%pO0T+|gCy7<zYX
zV%j_L>a0s65U%mizUC~rFAh=@sw+@#J=YV;UA5|gtDI^nmP{8`QLW8qBsx&90_DA3
zYovB5^(=i~%_WoI^&XQ~Ue#?g-h=k~Q(BwF;wHX1Lsp~Zjf!R&<w<EDJaK1MDLNs#
z@5g1fvft>Qd}5doG}gpM!#acDZ#1!q_c+<z_lQeFv{hl(Pf%9IhE8|i-MHF^442$7
zPIYXP?*;_zi!GL0F3Z;&l<9gxlQvip*YwMVqNbtA2^(gDTA_~A5ofz2G?e$4iu|JW
zP}<+ELt;ur@rY-dZ!ck_bRfg5c%;SW;_T}seSAs&b1)YUS*GQ<snc;o1vc+vl55nm
zB<MrU2}c6npS<QC>M313myWBoJ8!5PHtKVv2%Ef7kQSS%S(Km!qwN}K?(jZ<9G5ts
zAkQ^{u2?4fs^j8+UC=r*$Ty1Vodo^>g>hk7*%j+2%!K*hf~doKo?$_$a~la$&Apw7
zz;e6BI^aK4zeJq)mQh)tiF1w4e$M5nzFf}9&UuZcbY{qc2v)$HJ%L^>z2>!1u3;64
ziQFN?iqU=b>)epgcK1<pFsju#!Y7#xmzT5|tdwUM<KYhsZjJ#vy}i_cCMH!V?&mUb
z3KnVj6-)`V*<c2ju@z%mPr{VFkhx~@*UC<kum!l>FnQ1x;p1#jEpK;uovsoKhi}`k
z{d8_v^$bQM<YbVO*}8&e7fcC#?I1AMcwqh^Idhu7$po?WbOPATx@{252%wGIc~0pI
z6Gwj34FA=gJtedt!4n16kxVpo;Y#1D8M;{M5HHN^IL32PXdVbp_s(>Arjalsm6YVE
zs~MMMDSq2)1cAzEh}p{5(a4+WUvKw6Tx~7ZMhb|J0|vEXQji3EmCgs7WW%`)TdHg@
zuKb{<zk5)P(c!VdzBz@19M75G_2o%QP)~di{ScBvzd&>)94Ort=1ka6;lPePIh6AL
z%C@ND_wxC`>ElSS1Iziy#}|^n(wFAi{yezhx1)Nn@D7@u@%>%Eck;vFoqHx7A0Gv_
zv^A)r1d$}fy@vxJ|I-vl^88uSPQxeXx$WcU)o(iEXOHvqjEzG*=qkI@a<o+Ir!|KT
zUpC8br>DrFKw|7dsY1Xqg5=$7<_*CcW$_DCTf~s{l1p<8w;tsg8Uz$S9rqTsU&{L<
z1zavFu(-gMSnH`Ay4X<|AM^BsTWjB~K_OkkI-f%_uyH``-lHym&TgcHMKp_5vX1h2
zNUc7pLY}sjvE@e)%)I`%)Rn{u0>k)f#+@XfAz(Y2Z{x<Kufw!vI(mIX;u$U2cPHt^
z`|WOtt5`n&T+5rgp)o2-%z`AZ_$P3XIZyHd=2uxv0i2T<CP$p;e4!@%KB(t%PbRKg
z(Ui?mDQCtsMN~T89csWH+;sEPY$rzY3Rzs7RYgSW=Z*Ml)y@x3?aFpAy|hiLyjHw%
znI4V%T5oSPHa9ERMS<cW=%nl*q~(m0)oF$yWVLGlY4O8moB>=y0(NtFZ`5zGuCULf
zdgqze@BT<D%B_-tuXL9qid7DsgR`tf=QZ$~Ha=}X55*GCo=zpp#U`x@jx3&EP1?6h
zs%Y;w8eXc9wyWwl93m?EWFl3%q;$}X+)<I7?1)H~-T2wgST;+<E@HW$a*T}SzCuEl
zm@juecJDQx_*RF#bHmzp4vbyy@XM)~R<LHVt(Y|Y!2WIOE!uZB%=|;Swbj?C*Uy3R
z$>yKPbKj3%KQ*5%7yz%{yFA-~k7^nkM5{DaH*<Tl^|~27vv{dFM1GhPFH-<~Uq4C?
zddK9%C~lr2)_20bqpwpo`kA2DmB@EVsTYpZM_I?EVFmjK-@A11SKR8#N@1$DY>&fy
zPAi+(dYaFdt14>tcW>h5ePYR@p>|IsHN{_4>+h2!Pn|kJz6yk{phFMHy0?5u%5K9|
zV1UfQw?FvyvQ2<rGOe9t*F<(Rqv*06DaN?614vF1gye5?0g`IQ{wBau)@be`t%voo
z{*J_oho&}-Y-8(6UF^i{co0?PUR3!JCB*oFr$Tq8b-$OIo3#kPyo9J$<Xm8qryhE}
zSoQTU>$xL_T%HeGP>NAhH^C{8jmoWL^&uArtKK1W$Nbb4x5EThQOc16X$4*0Zjn;2
zF<_22x;%Szn*`<UK?Td#(zzk8jeJhF7M^FAme5U<y=&h6eSq?x8h^{sC;CtL<rhp&
z5~8r@ci2b8t<crX>|*=CfK!VjKyfd4zWSo<Dq2Rk#Fjz!g@sshR8ri+3?-*<R#a6O
z_IJA6k3Q?o$4Z0x@k(*#G8}`ps$|YTKQY-&_Vp&eu@u%)wOJhQHj@NZByddTod4PA
z;ltX#2CHJUt`nQI>P;K)bp~lGldLh48NJEda4z#CV!mtl2TxiwNq4nDk{xBpZhjY8
zGmNSIB|tD%mKCPZByPu14*OyUy2DyK9+^&TPO#S6UsO>{6_n?-ln@*J#I!L@9Wq_Y
z?UcLr*H>I&6>PP{k%0Ji5TFEf-_fP=b%k+_&9amoD^;V*&8;cPhQ-ANOAiJcA6DUw
z4327(P9FP=QZDWPQ)%`@nZqGnj?fz@6d4aI9mF#0gMqOtDxR%%XicfQwrL(4!!8eh
z&_@7y2nN2*SAM^uw?ZhNh}=Bzr{v*#A{fvYjrB|)L_no#Nrd;=QCLUWh6n(=n**<o
z(hCiwH!|ds5*e5i+-KD?z~F|IQ;K7O@@+iXS<&!YdWYEYyG>b}47HoAD=ylER_v`V
z&r^(w7OYDKb}v#*d?aXMM-}$9D1@FAxj{Lz#QZaICX2gEZpE8)qb{1qL!<Y#MY_5)
z(LL5s@ntU1`qjoeJzR^{IW$oaL8Sj8gqIArdwjY!CM-%MV@YoOjS>nSmhTge2{Xr_
zZ757#{xPfLtm3}#WAQ_4{cI2>QCb#(jo@R$vGmTQ7Wa7SfT*zEWV9simQ&GJxz@f%
zx4pBNUu$<uC{)1E5(*-PmL&T_8w@1u@e5llgTK-f?Tik5*wyTS?nxSfkHdn55DU20
zqK(c~*P~3VytN6>u}IG@r3tz5|LX$pwLW0eQ#X3iU!gFu(Hxf7hSVFhtn94ZNM<*r
z3U^M>knZ~wX%6hGE~lL^#hQh6)$-;c-{LRkpK-k3T^du&a`nsC-$)(eD<S54Xu19Z
zu$39;rbw5MQD3Y)J7onf(4YK`b7KFK*XtY~kY=1gTyOA)dz#9R%6s)Efy!dz2zU-G
zGm`GjZC)cKmY-UXo)a@uci`)nMr*1&fK1#hn5hwbA9n(CY_5*^427BYVdD2_t?qRP
zTrsg1oi6FNV=h8%G}zt2j?FeFK8|XR)v+we+8#5<1pKOng|_DAQ`s94R6Ut3|3T{-
zklb97)xqY`U$=}wcHDceb@41Bnpu3BNd}3)Se|f|7+zaxNQ2@WEm~N2dvg}p@45Ox
znE;gKH%rB5rxjrHBRG-12EIt~MPmcw{vH>_ZxisbrOrt3MOki6Mx+|A4s_STbT*ok
zJ;-e3L0s>sGZ^cr9d+DEe@BO~)EUc?P#MZMsSDmkc;L&*C<zH+-N>+SrdN#O%>6vF
zxcvCv9$NbdJFDH%#c>)vV;|Tj_Jp*?T0M_7Yb<c333o3Tq&ASWTpYSHN0afQ!lcha
zhnl2o&H7C;C9ZCynk(Wd(z4RGF^LzQ98^GtOF9t%ws}@|D{p_r)P?EV{z+|hmXq@9
z?V(HBof%v{{B48rOh*ZRd_A>@YUi2FC_jg30qT&i02Ky<aonKPr!<N4XA*V0#5s=d
zc-@dWQ;-BtIk?i&<^xJ^o~EvN^XuDiMgx%P!b0>)@FC+8Vl@Jt5q!LQ;8&gWXmY9@
zU2acg>hDcm)(jDxWl5ng%c?H4x6)Ujxah3X&ta;$ivD9VeJU#b5K_r#(ijZposa`^
zi|tD<%u6M=qm)_(obus)4Gno79UVR<fY<tU5A1ll>n-04VV?k+SkS5kEsPT$jun>f
z&kVexrX+N=?C&_yVp7MTxUC~1n!xgoY*+4L^woKfjnfffZ_(NG+^(F0F&I+bdQGkT
zs?heWalX*%#hpF4_zeSGLZFcQ7~bax>z!t@XH5s=tA*gt1-=^eIuf}CnJbBKvnU}N
zK?s|Dl=K%%iSvWZ%t%frCntK+diQ?JNiCr};V_ngk;`msB%V`n;di3n{52F?sg2N5
z4<7aad~O?Z=9={TxTZs`MB5MhW%f={Sr7;W3yW0q9f{aN-V6)ju82S7bA0lZ{bTss
zqi=BU{W5A!TW!g7)=uQcfBT3P&>YY-lA)uDs!GX598gPlhg^9LRiJ}(3klT<F-=cP
zj{u;3eLXHLlY_0Tt%JNVU|Cr(Ag$Nz?)|79I*JgVXJ(gg{_1BlX>qO?RL-aq&&)2N
z`yi3?_XV+lgwSQJxbK42%sf0YRiXdxskUEUSzH3;Iix5%iIMnEK7z75#Ww-;@AXA<
zfN&byM2!g3h-<jM9A*DLh;0fhuikFcd5(0`yaZ(KLx8&md+`x1Dd`9vVv=K&jxVE8
z=>8OPr(sF0sOHVh&8ESujP&%3EEl~4b930TW`wl#MV=L%q+`vP9@QcAFPi%(iJtfJ
z{hnh@I?N!b)+{}UW}=*6ojOCPsY@Nsh_9iOfT`l<{<wtFVRG<BAI6)&kfJ^0<^VcE
zXJk5>ll@zrYD#HnHX(^zCm}JIcyb?3_%pknYYf0=n$snN-vGk@`ACup6^`LoLX>Zm
z>dR5=l^rBLXprFl%oONo=)nKv6bdHy^-=7A!Kfe!?J%pcVhFFiK=T(ZZS(y_r!WO5
zfW5Vmzi7}%C)DwM#KlT}PC2U-TX|kFvvuB+6yMb;DrL<M%V8Q`;bwTDa?7QE@&(u4
zT&zwh#23VY?po3+F~6&;pJs91ci>vjDGr3E>1=-~x`Y;hqFqz~IDN`A8j_K48&YIl
zZXqbt8ZCtrmW7oK_xOSu8HpK0#R*w34zQ;8)KPUSj)JKJo=&i_eoJ*S8q$rFl)l`u
z?>wmO!X=ZP<3T4-L(V(_T2p>l*|`uXn~IWRGo@w&`4Lrj_7u5w-S(;2p^<$pnrb~P
zKatY3pd7}q5bukeXHxz)J|GUa@#<mMiV_<b+y)4Yci%1<Le^eMl8e<K^wwfUDU(SY
z@xY8(d98%ro1dVn_lzRMrFyy(nXlglh_j42L3j`j$Hl)~8tGi6md;hC-6#CO2sk_4
z3Q|WUh=Q>MX+|^LTULct`r!bR7HCkFUTL%CfkIX;JRQ{L@9c1E-e2+dA0`#kU(uKO
z(r<JV>zNhk27R*59VshpLST;MoK^q2hpsW-^L?&a#+z+G2aPQ<iyd1eZEsGMhCcBl
znr%+t(8Nz~47+l(zoV}GCCldX^CR+7aQ24`(PQ|YQ+IfU;uGKSq=i*tRI5IDlN=1R
zvT3UrTnK`H+dKL=<GCxS<Huj`AUMm^GHm>~xsnd361R}Q0_$D`PI4E`H5dMKuYf9i
z;~pROiBng)4MlCiE#o|vyL7?>;@T3!t?@C~-Blk&1a{%g(jngxjuRS+qRF-8iiJ8S
zXUwp*Pgi(DDnqX>1GMEC_2V#+7kN=z(aH`;o|W!U{s$h*)G-023a8(UaU`Qtjv*?r
zzt>Xipk7*6DeW6h37Gv=2WofZ$p!wz(o1KI{TK@m7{W6yRJaG9^xB#>iS*v_Z%;&~
zX5o`4l|9er8<f#5YHmg{<VmG{r}Pnkg|VFX*CrH1!&H+`DRysIsEf)uE>m@<av5F!
zsw>+|`o{R<ZRjn-f3&Q|=7mkNa&oL>U_(~5^rTA8*P~e~_PdFxzWJ-OAq`)ojy>;`
z#BaVb3j6Z5{GQcUXK23bq&IbHz(qA=ez*!I4SGqIk8!IJ*rpWArdt{{*1`v?Xe>;)
z2D%XmLzX2@7>xHhw)HuZH|BpL7cG&oO!r=W#^T<iuc-EbYG+GW@QL1hk`pV^pDEBD
zuq70^z5<cZgIDH}dkqa+=p+$btob*VQ+$>|96#x)KzSjA3yg;+)g5jPfx;cRHp(v=
z5P(o_eyYv1mX(>H>M6#O%KOva!V=2iO@0CPbDNexT8N-6Gu0iLNG4d|O=p^v=l#Yq
znZmpkX+3Wq2!;Ar(mkN3PU@*(QI|@IRdD-*A`)308?%=z-fdn>Zqj6Q;<bcj;ld8P
z)R|ytMBfZ6{l9{&uvL_eO9L$kUXdazefi*qv!lF3kj&)9NrlI|;t-N)iH2uL#yL*y
z&t)MIodSFW@Y+72Dn!FmKeofUSe)y-APb^@N~t#!O$mDAtn@>_%~e0zZQb%u@7bD_
zjAkFnB-8rs@!~E8);|4oJ{%%*n`LOSs5qD=T<?R@IsO)xa_A$K+z=fp?E2X!=FjdR
zxQVxVOaIWa9oV+W&}fjTpks{+P9SDfaE*<#-C=pu9$Fme>30-hL&<|cDXJ!dOXg@=
zf0UKyu7D=C5SJU*`%GuMEmfe?&$^=Q!(##AN!ITcO0U(Pe5-enWi|es?Fndb9tr4v
z^v-V?T6y2iW3Wg{nWl@Tip8aE>%oyLo@BHJJ5DSvDd~j>_j{`mfDS(E{fsW#a$h{x
z(-o6{I^+PAA#!<K)qeci_*jFC+^|-~YL6VIeJmyvzfMKKU`mfE8U417Z%Ht0DMUPP
z#-t!OE5p+g%3?R;ogL4_w_aS_44{-HB)l<!*5_4}ex%@YWx+F|&#LlZlQ8Y-_(0rq
zGd8i0ViY_c0k`flN>90)t?>QYisSN=P=do(dC`9qw=c^ZpA1}HUJ&X}Bs3Q6n8u2;
zbB?qi_$3BhJ|NOF^h*teIxi=ua%^7~)Ud+05ummFR>Vn8kU1!0E8!oSRUiL7N%<PM
zNFCUT*lLw{!`|vA9j&01rzdT-NAg-rXF*Z9Bj6^WJB0}6Ffks9N8<i4pSU>-G>Agj
zRA^?vlC^wOJ^iEl-u5ZI87G7|CS6Xht$%DE`P4z8tFolx4O*&$%eU6SGqPkH(oftP
zC$5hp+~5w5uH))%wP}}JV=HIX5$aaN@k~|=I#%)RnYIF|-;)%nP2;O}uSs)~P@6Ef
z$Dd|RPJ2_SY2PLDR$SqSX!GMhr>9OkPS-K@&~3WKxxs<ihhUJcv4L-vc}Cn9<tRz7
zq7{wI46?qUP)-)7$a}?+%U!sLq^HTE?LME2Ix&J**~~x<T=l81nkz7z$lGu@;^5|z
z>W+)<dtGX>pUs-KufUcL*@)aIxJl(^h`ifenec<VFSxPY4~wXb+h7q8<XzQ95^x+a
z4s*xOX20?b4~S8aWsGZ$JS|5uGM!U+Ab~T0!6>_P8Lz}LLqROJ99-IBpKY0ApH;-&
zoor1P_|*y}RNh@tLd7rKi2!Fxq-P|j)%yi`Fe10>#_PqIYT*t}s&uZx&b57QCoZ*E
zwU(0y|I^-5pqlpxonjQqZ$Y8;hv_Bl@^=&z;xfYYo*}nu|2tJl_~2Gd8iRk=?41vO
z+y7yWX$w=7t#Z3%bBst~k3as%KX=q?WTa{IwkgG+#DXM4%K-DoX0%Ub!6$Z_+=Y~*
z#SdylSh|Ab1DO<VEhTA2gJfAN3$ju>Gi-3ofhy_C-GBErL8$sDt9{ppd=5_6-ME=X
z^EofMyB~##5`tGLa;JHsx+Nohw7Ycz?O`Zj8Qaqw4#0GU!oPuvmh;S-6GaJd+%n}0
zI!R2t*R0kcT8s-^fA+_2NG>)vJHe|fq0yc-WLf344I0i*D`z2sWUFKjyxtn;8rn(~
zfuQnr($gJy$he72>T}C$?q|GCcsr%+v?pZ8K~aawUWo5?e(glikA9aZ+8Z}V&6F!b
zde>P5YOzJCLpq&K)dwci?in+E(y$6RQLvnc<E{x!SlC%YK^#95&#dD91gSLS3q5V2
z(9?ns><5{@yy|)a&HJm-E9)!^G?L1WKW+Ey*IdQKuqIEN>}C0UvvjVHkAqdkwZ5E%
z_I=VvYwqn|;qJ^fC~inX#BOK1Rm{kg+&1cD&pCx^3$JVkf~O2f(vpx=XY`x8gc_;<
z`VDNrRlUk2M2B$oBk&-te*AR!@e@(-6*Ti3uZrQRKR*t+eSqdx`nTB2s)n=`80F4Q
zin>p$x$iM7l-I=Mhu|E#as{w(*Jh^l3dJ2jG+L!(^uWLnBk9Fz=cUyENON)Sbyo(B
z$AC)OE0!=kD+{{1$tR<)5blcVH>ZHL#fF{!TzgBV=D&f;or~i346@uD_RNBIJp{Xi
z*nQ(=2F9pohNe@StmYCr&2*dsRC3?drSf+9i;d;5l&w>gD3w}%2o9-)P}LjIS&!yT
zpYp(u%1g!_+Bp;K%5BW2uhF!=#{kV*7k1R1Xm7%gT3{?e#aQ`m!hgV&Zt)C;FI*I(
zF|Z_OWo$Y8dBQMUJY%!p7!DJ0T{0f17>qjS4sN>LS`j2G6KZ9=>$;+mH018Xwf5_W
zb+i5kBOZ&-%F9fwoE_}%js;&7A$(vu#lC9Z`$T4`a;Dm~RD0kXEi6<GY6}*!ZK+EB
zwvkZl=S9f7l7HpvS1j))8j?F~G5p74pi5ixtb|I>mWf`&%Dc1)Op}(ojnR1{rdU$K
z3FK1gabu%HFJMmPz~Kjg%5&MtARpqBy4)GT@~!@x#mu*QrdR;Gm9#%swN6(oD-VvZ
zM<CpPSg>OpwrntL={}w=gJYFw9fBKN3NlzH%whSplY5g8&UktnDm5*bF08FOo{l$a
zu_TXy4pqE57$!wtjj3NG>$kZ>0sd1=I%y$7Gky*jVFKJD%_j>P2{)qBl7d`qu!vVQ
zw~eEIR}Bm@HK9h79DoAi19S|Op}aM2tEK!8axxPzes(o08qQ-?+Em<)iL9Kg+z4hj
zxC%GfeP}uo{F|w0yG}xjx?(<ShI%L9hkQw6eyV(=LJ1s<!UlvUO|LDdBTDrGg*9^F
z-8>9T8jOwX?jaJrzU3G_3_lKAH1*JnS-x1VQpoky0p2^N;hlA9-(=Y-Xxtc?WtJ7K
zKxZmX<wXlhM+~~rYK((<tCdm`t%I#dkQyDA++h|(@b-WtNBUCYYZfZ#Kt(YoRvLWy
zODy10NJA9YXff1t!^cr8+heJ%JRQGI!1`LKMVbzFKctL0B7i}CF-IU*1EJu<t)%O3
z6CsQzvfNs6>5hgelHI~u86B8+07GWQft{vU8=F*O+4+@ljgTx=B_nd*&;-R|vpJq2
zS}oi`F7GhvFz@leP~WhD5rnt1Vxgi_F;zr(rNhOYCHn+s^_wrwL;~15#mrgv9eX19
zcmy9{F7L)O$ca_zb|f+ga#*v{5g#33^iNY^LkMln&%Os=)AU%^?&s%vHnrz7XpV`r
zXC+NtLiq^m_!#wKK=&qTs(f0{Qhx&#{_Z@DTpKts=T^;{oNx*1hc#;RHp`}Rph2Jm
z<3iqQ$;B>n)_ULioAUq%HybA=;XT4YPt>t?B!KM`vxEY5#}B`XIJe3!W&Dk(YHu^J
zTba^P%kt#rPCT<9FP-)D7zt({H*-{ZV0kCCwY4lP)bj6ugE%B(T(!F@)YAcm#&ic%
z)iHzM*qPvKKBYMU%?w^s`Cj_A-uwm{qeL=DE{E%=A&7<K+?|;d$?4$cAfd;yw$kA*
zw&&dLTPHzZCqXBT#A~~L0-B*V$EsBx_=JSmFW{ZH?#`I?wiFMCp4#4Z)nqL{yJ;wb
zcUqvB+Dpey35m{Yl#zAa@nt;n31lYAtlrB@98j|_#ot*Tpl6P*P8)A1!-%8{z{Ddm
z3A00^Zn+PcYORYt0N=9BV0l<t)Viu-HoU#-9i~bAHtR~1Wcnd0qTJv1PE=F<bts}d
z-u<t#7D}RZsX*~gUGd<@@BQ<p6>qY6SshCshH^;gSi4eN4u%&Ry5~?O;~qXecDAcA
z^BSyZ8=6Z75n})mkmak1W{?P~S$rSMDn}PFFtCdm`Rmm>b^1bh6`#%|eC`az0lJ|a
z!@k7=APWE4u%Fem(y$J-tcuPnFdi_eD&I;hovcur1>8#emp@Mn)f0Q7uglLq&4M`Z
zbj7nu$u1pybJb5>wsBGk1XN2-7Y0ujHR8QMK0WLi{2~MabvL68U1L_?{m-IubjW6{
zZv}R9D=B$;MLa@UQu>eNL_U?EeUxPM6-F-O-#t^R_|1EKp&X{|Zom!}zTT(%x9NWh
zlRdb)5u0EB4O8f)^*Dxq_YW;Vp%e?)oks0+Jo1ua@5h;PK0VDhDeb~4Ki3GD)Qe|e
z6vIeqHKz;frOmSxcRRl3!}5Q=IU&jKt=Lc`Qhd~zk+){x3(q&n$SEu{9sDO9i%pB*
zMEeH%(#0tT{RW2jJ@pRQWWgVMW$r$3UhYmf?ueOI?M*!MacYlXpi?gHD~z0K5>~Yi
zGf#^&VkX1%umY;E3THsV=eB<O^wFslhiUGF-*np6qz02n3V!Def<jgDSu{swEI8LZ
zwg9n_TIh*aB!gtYEM%=O*xgMfXt(KLn9il5CFxuW5ECQUp_iIRLHNUn6jw;TIA359
ztn85;OqgF#Se$9B#O<@F2<qwDnyc$#Egewk#`T&NQAgsAsTA_bsE-;T8^hr6@aE{s
zpbq`YqX*@B+89m=Z5_H=#VFfvS?k}+Px%a@FF*NExzV|XGtxNMx@zSL@0k1H)GZJJ
zQBJ4g0rHoS1C~NV`4C&MQ_ZZ8?MTbKyL2P$#wF7DRzOQszDS};VfiGQWmcTuh}3UH
zAJ+t!b98NWmBinblaqYZm-?zBOjk;8Qr$+S^O_Y{#EtMiGwWqnJU}HVDe?*%&$%fb
z_Tjt13eCe>##<q`Z>cCLG=ow%GLPv<#i7!|P9AlbASG=@H6VUYi*6|?npm9rlqKAb
zdd({b8K>!KOu4qCLk?`-#>v&kZq^6jj}s5$wTOC}dJ~g2Z?APW%z0)AhdL6in>XiN
zxUJniSaxDkFdP`mwo2_OUx8nCjY1vu*rfse>5Y1GaDF+1gH5?0pNge`IId<%BY7hw
zkOG>mMAU)kCQ?&vIbm<-XgLAfg#elUg`$Z*oN1+HwE>4f`Zb6UlqE6?R%VpK3|0po
z3qqc#y!6!^f?z)LYM6FxG$?-*BKMEU?DCNxR*Xm>``}^?KUykHd>-9Tl;sw75qi{v
z&$fb$IvZ*0nHAJ5Y9w+T$1)9@j^lH2*a%j1pqqn<4QI87ER?Q6`YYtCr+gr`xH)yU
zl4aL9AP+y{XOo4{dT9FkfSSHtSI!rWUXQ#`g<kD^+mzw)g3~i$2niRIS<L~Zn3N)k
z1JA)LayzlK36MxOs>e>QF<Pv(>P;~Q=osP>G9@xrrQXqivV7n<Y4awsmO~&%Vl!@Z
zZEc<}yvQUA_j2Qa#qs`2tN_)iJD!;hT5wE#p=N$_RCJ{dsne{RpA%nEk(EL2GwKN^
zAimNfuiRU{2gCS%aH}eZ*)?h*R7+NdnC&G_opEno^U73B5#+Wd?pZc{%?*h+LjAbD
zu~N5snzvDVb9?)Qv9VO&yL$0eM%Y!1naoFlt^iKh9`zzM9=-3_g<;ORPZ~jjz<2-g
zZVjrymfp|F2~(8#>Ek5N+Q{?5e3DJ}tdxFaw1;#H_onldDi3?G5}3h;L226!>eL;|
zmF&Apv$P*bvpbsiQ0CTc1L^A2*MlPzT<r*NBsP&#AnSlRtd^rOX;b$~mK9EnP*-YW
zSnY4mgZ8PjiJQuQ@#5x2hrX3g0nXE!Fz6QpGz2psKtZwBEH@gC@JT8D+%EZK^EFc2
zvbbZXm076CxPWE|m4XR)mu%3cN!Ixu65y7u4ldiskHYDgpcUM@wEWJ*Sc=tQlyyRW
zIVE9L+oFXpaAU~C)-xd#;W2T$20uztmuv~;xL5qMafAJPY3sl@Koo7RQbXvma7!n>
z&=of`&w!ik7vU>0cJLRsn3hz`lO8X__mTnC%alBBtMoDQj5h{WW!jF7=4AY<9xJq&
zHet?)i{*23VAPM*)RRaAU*Rc8N$7lbWNLvr?3K(Wo9a(2Dhp_MqBhG|Q|&p1{N}IK
z3clr`Z`z(_uH-3XZGXHQr8J&epMtx-?vfCI%VE6P3StrDoi&s6G_C(~4O^66DdW4U
zO`IgX9dvrWZmqDOVIyjfs&_bx7yM~wNWkzf_UW;mbo}9L1muT8(tIbHvaU~y!PD71
z7=&DK5waFoTqMBt(fdS>yPBcJ6=E;}^j2+u;fSDZF>P6mNxs^!tUG3SN(cyuA};7?
zPI-jt);UP%Lh-{M0h25(!)S-CShcO8WNQg}68Q1tXXLTJ{)m(uw<VI_=J+r;jYrpD
zg~g$qR`i;GRpGRF&ehNcwns}dWjW{axV*xP;&ZdS>ZoFvC_;Qy=j3b4WJbZwd`lHj
zuXhQ~mTR^yOTGJ6^vE-wqr$oH6xP3!jM^eWGP||NG#Aw9a<lcdg9+bnNpXk9uFki?
ztUL~@-smX`MI^Acn=1Ppm%HCrG4mtty~ofNn#;F%k{mCe-?@5amE5W0iw1J&CnUr(
zxYBA$`x;tG@<fJoqFG*;i2TEKa9J}wF{#Exs@;ro1H-H#BGM6nXH?oq+hm$xOn;6v
zfYnR?JyVm<U4z)ZCeUF?fbG8HjdsFXSYb*KDC(^TIxVs;qNK;1tf~mET2==MZ6{Tm
zRdh@58hY*+zBZeiS2g3n|8-j?8|r#iz7oE6w$m93U-)j=sbbsHs`()yPn~yPoi}zR
zTzWj{9Y$yKwQ}I=+87xT#qN(ke=@#rs=gI&{pC~63mutVhj%|-ygln_KKs4~_Y2?r
z8omkr^5Na%`Ll)|*Xo(hN!m|`bc3CDa^#84ws)wR(DR*!La$kLU|UwJcI;f>bE4V=
z{1m@4)mZ;TzR~O3yRdp??euI3-c~CTIm4oW>cctNWK&vYU}xcP?29$ej4Mygv^T#N
zHg8Twb$Djz!7{<}SG$`)LU|Idbkz}SqM3_(+Rc6WJLP6hrkYt+UfE%}pt`i=9G9!w
zQ8U9YQ}hfu0jf6DUOUmqEDOEl5(^?o@@1vxV?7hYVIkoVP_2MOt6p4J2(+ITfa{&4
z?vyfo9yYO%h-Yzf$VquGu4b`rDZ=hi$|lJZuGPC+UF>sGF-fWAP3^z^)7ZNKS0sLN
z#ii=cQ<<q0@GrRtS!iU9Jh&^Z&L>A&Nj9dY%lpkmxQb3-QUnx`B?x79kX`!M1*;h!
zyPoDkGJu>DzctC1dGd5C0PUqLge+fzgoGnu`&@~Xs+ifhnuTLR9i76m__mFRdbqul
zyfvuA!6V97le|~jnf8=qO07OaFUA5K<>o)LYEg_eY#IAR_*_>UyGY}xsfC3F8zcHd
z4Xlgjwvi%>dd`fJaeKsbT8oLHzLSlxNwfl3tk&s)ei`eezO~gDX-bZ(CQU_NnYAmL
zm9Md$H-116UTU%Q7AhZZP{hD<CWdu!du80=VvKxOAG)X&H}wuN6iC3`rmR<1l$9nl
zALL}b5qUK!axhqOWsbwLSi#IK$KMH)|AfZ(_rXXH+QW=8>7Hg6x4rLDw~!(4S-J7G
z5r?1i0sd4`AQ%g-aC6~|1PSL}3FK6_V!iINrB<MQm0o*(Gy`K$osY<78XJ5EAA8Zw
zK~DdfmR?0P_=;e5QbO0WgAq&jjNYLvG7@ER@+AgqdnKd@2;?5k`wXtN{f$+uy#4rS
zAqoh)9s$X@vVy?5ArZSL2B;~C9l!S6x>We>m;hV!*=vu6@yvEx^D&bEsjxB<e@4rA
zzB6gl))0dq*J{{4nTfr=>?vqwU?R~$T>vAjZmy^0#GyDCBp<jsEA2)?(gNs5Vz>y7
z(PDXC?~(e@qBg)9Qu))Qu(}{T@~hnDtp1PZ&op;e=`x=OyF9Lb_pENj(dQ3q_rZfS
zzKQ2+_W5ismigScU-&~#>lJFbq};%laVZsKT^UY(CN(+j-tc|$l)!+dDPgH@884)q
z5)}#ax%2M+m|J?Sy%H`-l8_f1k79A^iex6=Z|8bhZxRpqlPcKma-^IuD{)fu@Km6~
z6A`8mo>eTRyFcv2z@)4^U?J`w&z(uF@^4);FtNF3bNGh0hM?o3<&iE=X&g2h|J*SD
zBX3}6-XO)E>QjCF3!0az7DvquKYoQo!7aJJ3+fw%V3a+xQ$Q-2`N^lY!8#MWO%qXB
zH;Gvo%cnm>uhUk>e4wn=u^y1hmz#Lwz}$O!<N<l7>mtG*_%w>HIH0dP4|K)yF?2@J
z60Zmf)LW-`WSx;jg+4Dr&^l<FK%_u3|6spvwCo>SJ>q;xn9CzEYtD1)_Yeb=^>So?
zLTu@DeOvk)n{ugWzx5R9<2B!lc%xHj=q(AfHj=f_k!&PluUur40)Z4HoIz|va~?Ex
z45?sjeHhfmi#O5@X3}!RZ97^Z;~e1_6|B9mW$|LGKBB}#=&yZOf<(nRMqZ_&{XtG7
zzyZ`wKhx%;r?m}w-steK|B6=3sG-&<JTd97y`!1w*A0=J_XK3=1DUU&z2Th}9)nqa
zaBt4|=Xx4ltMbHI3Se1A<s902?vNO4ww2L5Nb*3Lbd){FKG&B_uWXvF%tEHwa7L=G
z=G=;%+&BtThvyP3$q10JgNA@c0SU3{My3MAs+^sKfX&>!W^SZq357ZVh@6O$)p0W$
zqPP4MJ^`6XrO%qteSBsTezZimxFhAFI1E9dQw*iik8!IOZ|una7IO(9!nP_uAg^d8
zW97}|14Ep(#;Nq;DF|CGPf&otrIHa{LE}hB(_mR0kJ@$D<6NJs$9??k4Q<S*z}Kef
z<uHFsVn`dP=#O*=hqA-f&V$1ro9^*Jeyo0V8BKK)GJHAoq&y{N_17)n51)N0UL1zi
z!R?(^R`$!I-(;mCiEFfAA+I;JSl46b@Z&U;L!syp;O#6UP;U}#NmxF<J&A%UCYhGd
zHeY)I0@RAGnCpj2%*Z<R2g)_2S=;cX2;57C92gp6D)nL9C&C(qW<=9r_*R{4Mt{CC
zS_P+(Z<dLn@4kEDETe9&qC)G|%uYzOgX|p!x@(UootW75CeZ~Kh1FN&xU<blgqP{F
zB+J9e=nm%Lq%SjC3CpkL)QsgirYKc=;HuEY#T4KHYIeWR7~<LQWhfu)_q#+h@j>Cm
zK*$V>$opKD$sQcDi#5m8Ck#k-w`nl$t9-0=z;&-qVH8CAQQwBS5nV!O5Us9-H!e&q
z%#2VHR2BVu--4F6nnx4q$;{2ji7*J`jz`5w)0xj=+a5t9)JfAXo6c3-n@Sn)C$chn
z>S{y+7<Jnd86+JFndu^vg^GA)A%#J~oPR5QeoIQq=w~~z6V7gT{&ee3k>`_`VO<(@
zB*TQeTTrImIPT-|ZC1gzwCIMj8*&Yb6GM{5O>=1tV&lcKN1cNWoq3(XO1&&3w>4+&
zOgt_p>V%DT;wIk)Y*9CYqcZV|99N~kVofo$I_9`J1+Bz!W8Qg|BZ<yj=!}Q1IpSKS
zW1eEA6z4h0G<3+q&u-Gf7X$(o7mA0f2j5NJPEyZpZIn5ak}S;$1=Jk9wh5pDc)hx(
z9~z7<bEX@N77Gh&p<pc7)MHP{E1PvHVm_}nLJfRgbvp~RCtlZ`X%rHp-~0}<kYW;1
z+@XgG8w^F*tW)TBC@zd@s=f{#_NW#9qw4o!O#yq?o6B82ux{ST+I=uT?LJeAw9E-$
z9vv`fS(7?`(;GOsKx=5s9M4}<B$2=A^8*E8WrJz@%a30jg*tYI>9jlZZB&J5ddgGn
zS;lx<DxIz3sw~TK0E|{km0p&5upLd?mm&y*;`-_JPpMzZYguPMxu#_5j!fyfz8$n`
zZXWQ@f!KS+?ay>4sRexh6pZpJu|{#MDQ*up5sMl}agrV|VW2F8e`n+L#)L)1UlX#*
zcgK@j8(PJ%>mq{mr-3*%A}(GMk<hCpXvgb&Dw-8(dK9wK)ihv?r4o$A;+)f|Rhnft
z_6Dn&fa({W+N28eyzlg*ZV#KOK~VEBLQ6{vfw0gTJZ;E(zZfRwOJVbIzCq8qyTd(a
z$G^0wxIbO$LYekufy}?9b_R{jYJ7bDJu?Q6V({rAU4!1LBngX@u9Ju^t|g+6AB)`W
z*YE3p5HAZ=fPz_ST-QjXMe06bP&eUNgP!}}VU^IOCVAwoYX9uT^3y?;gjG^B1KVv@
z2|D%R?}+IxFJ$@spbs5WjnHugO`F~^^Ut)Tud7q<;H6__Ih6yI4z-r~tU3C`32^R>
z$21BK_>Cv=N})+4=c2F0?1ZZ_%^fQ!w;B)O^~$uhO*Al_@jK<5;E>YsEj1sLPurx;
z`4c;TY80tQfQ|B!qLT7%4H^%d7!SgW2KxIA20G&ToSam29z<MDN#xjdwp%^7QY0`A
zYDZtZB2c88!Y6q5edLvDI>C^mYp&_)3my^P6FLa$?FPqH(M1?EgoWk~)X1iNOIU#L
z6uHkjKrnOP`&5S54hYS}8WG`(IuDvf0(407wl^0$lIJW|#$8$`nD|9Vl|Vu8qQ#u6
zLA_ZzacuKo5>3)@;w47Q=K}dcid3Q8y}gieU*DSLL8C$K5Q5|TNTz)OC0HezGV?QT
zf6?H5B3PELscGd@7INQThOl#)&1C(CY~t2T7^8*HN`QDgg@0K>CG?*`bMMe5*Q68k
zxXSS%B5m5)={6xyB=5;rp35p5(BQS3AB805_LEcSg-X|3xPG=8M$|T5_b&%0-{?{_
zJ}CRHFUz`gO1(DW6;+0{2h9<QltCe=A8h*?Uy`K=wJP>o?<6@uE3TnGBU+nor-8uc
zW##G_Y;O^@v*Ouc_O5lx({=NXhfbAaxGpFf{jM^f;~`s#iq)@g#pUBEID<ca2-p7j
z=d#1rgt{kx`03-5M}K{{$vShs`hDU~#|71C(k<l-)_+~-p83Ufu`m!QJX~t}^T(Zo
zzpuWLOPM!#@wN9~7xFHs&Jx4tCSJVIKHtcq7k+xhw=?#y3tp&q?}~r<o2kwce*KS&
ze|#Cbb3B!Mb}CJ`9SL;**M-K0PoKtKC|?Mz*ZoN?`SmN2?&9(Rue)2R@cr=N@ZWqp
z@ay@=e}7)%`^>*@Y4hak#=Cs_9sfTx`k$S<@ZBxt%)|KS&sztt=*_>F|9{=Q`v2_n
zE^Hv*2LJHnoIU`-4=oD2ZL;6bO^5#${eb`PAA0h8VE+&LFh|HzrpmqBKcSCT_l?Dq
zdGzD2^zT_WauI*<eSsO~J7hlGjWQih`0IOM?3tgzxicHx-{*TsjNP_eVYdw3wddB`
ze;zEbvwpTA{sPCZ(LK5@sP37)@ZP+WRFu`dmr#}){^)kceJN2}VmKuB$D*#lj+yRT
zz}4NgJNtRkBMDdEiib5%yRZG4Z9hM`j{Tba@n;($TrTp*-k+atx1ELRoTIKDEM9$k
zFZB2rb25J8^FKGj0>AI>^6V4ujQq`aQu(h7^K%L-iVq7fnEq$8bf;MS%jduQK^9#t
z>rjth)YhIoxa{y{^j{aeQ4f=x!h_6A-}pbA4|%xqS7H7I)BiH`%`TOHzN0Z`UPg2`
zs9OicFQ5N!$H9Lt)Xup2|A^85OBkg5XBhm?xTc4$;D3fL>dyat<o@q8Zu*9uexqk!
zH$LTGFTJVzdq4djh`#?4SeL&)p~oNCuYBS<-3uq4?lW}uzihMr(%Juv#>-!8{+G>2
zQSvJ{r`!A7p>NGgp|5qM0zw>LtnAaRf=WNk@%vT4-*IW3@BQ&*mo4iJdv!O9y2^#Z
zXJUn~+28WgO%whqcz6ht%9YMm8q{+#M}^4l^t#%7Y3V2nCb@_3ZDYuKsy8Aw!fr(D
zv)m4#zD>vrIdQw*asR(Q1<Y{sMY+Tq+aGVSxcOgCe!*YXzWR^Lf2hR$*k7A@UhHKt
zoB0v=&rLCq?)<I7Wtsm!y9!%wC;sX=8@*H**urvf7!a4Y8(>@z`uEn>=g&-MzaBOL
z?!4|<{knR~<n8VA#+R4>Jy(~tTx_<hVs{*TaN4#u(<T}0n2Y{uA$<8o{<C2x7fGU;
z6Xp{QKW=u|`?u9E=2`$|c00R`JUbTL2N2aUL6h-KFs&-9rYvN6R|jQBSzw#4`;7CN
za45@b(KaI$3wlcvTv{$l20-udHw-Vd3r{NV@7?358R78pV1KI4ThP^N(psxi>ghb>
z^r09jjZO+7!I?O6!_$E+p}vInmj=mjA(e0NMvOl6%U5?Tjb@fx&$<_?r?f5VDBSOF
z{{0^d{(s~({CSXS{Z+N^fi!WE#ZJ##)#C!HIabyYt=?z9$hzTKqBl@ntv|;sn$XGS
z>z0oV?xif`$e~@FOYf~9@KIuu7fUDdvC0aksyzj>Y6BV7uC3ZHYt~aYjzH}j4$gm$
z?IqvA7JgqXv#3dP%(2!F@=L2@mB=bJGH#UHp7JSYrmLZat_=2gim|j9>y4GDrKAwi
z=`>UsYhV1*q#Ifm>lkQw<+ARlLY6P|vqeOhn<i4jWO#s6NWRUn*6#Sr5dOTWy}Pw~
z{mxa1d*SVbk4>d7v-4vcG01v@cte7@lwtQ!g_&6W77k#{TxuR+8V@VH{UaSn+dLrj
zXeM{X3a|3x&Tou1LP#554Ns+dCVOQk*gZLjjQk0d8sO7cx@4VFGmmkoelPr{TZzkB
z@vv4_W?Wy8&-&V<x1icuk-N1>!O-CN&2)|mg@;6g7B&$|1{$YkzrV14-!r&RaeObI
zd(g$QkyT9)@x?`<nJ2)5w^4htM>w4M*x&biF5`vn1)aCP9d@-c9c*52-fj_Xr!xf+
z<pBUm5ldr;BqYb?{hYRlXOHb7E%-<goTtXD=h>7}GZ%htBzKi0c^xIIYZc0bvOzW{
zJ;fj+4YSLa&c!8-B9&CH)<?32s1zs^;-jUeUsS1G${W-&U5eU`wb=`g1V8_i-K^Za
zzA0RY1R+Q($n>H5BUf5@MLm@GA;MDZ3hF;#`&L<VtMc%XVRKivC!#aE5nQ$g9px`>
z`iP9DXi~Q=<%GuEEq?4;ZcIYk;FM9XxerckS6MH`^v$r7t9cT=u8hf3EANOt#~<Pz
zj2`D5`Vd=9B@ubA903akI1`&#X>lkb@sGpVO*zku)}38txP(QbEEA%()OEfr#UiAc
zrD{iBv4}2r6tyktRde87(?QhSRYUFb4^TJx7!g+0(F}4ruaN-K{I%^+SI-HDw!F|k
zWS?c6Z!9K1@_D&8U_27b^-S$sy6@<&Vb!nd>Ce=M9f|D|r6PrTO?$}^+Sw)I99v=J
zYx$NjnNBk2F%~<SImurqKB22SogcJ38RblEu#c+wk|9*fES~x5eW|hKhLsZ>dyUqL
z(5Nz!SnZ5~P>ee~NWgT)mo+XT923k@Jw!$CFl5Lb-BXUPY|C3?<wx9ejMmCw`mc3e
z@)pkZvxGvz7Rp)K1Salk`V#eaEIqVcvy&}ailKhwhHW8S>ypbm_HX~X@Ur1tnhNGS
zC|la0+z$MjwD#tH%w%<3BX!c!sK2ciy?`hiNY^8nU^u`)2>iS>Cxzcg$7g!}$H-1>
ziG=hkd$q+6Vk<!9)n9xa<1@vKJ5{_mY<_v?bC2st_q1|JX&4#q$~R?9yQ`+$Rrs<Z
z%hR$$uNE+4`W=bgqy3U<=TN>35vl<8^`<dLvjz%T!H4v%f1t|p35)Nu;}U8Xo4L!c
zO@~l#Avo-^-DV%w6v_pIhmNX~v$y?R+xI*ARg1z#OU-u9l8>9WpIAsK^E7PP?u>i4
z0vk8J)bhc{S7Yrd@&`<rXG|(}3%3uC`_J$DOYdHL)jy1pLHwyNcY37J91AjY^O0m@
z>K)I$h=sW<3^)y%`vf?lY9C&sq}t4iQb+1k_vTMiU>gIPGlVP4&FD3_cRav^k+^de
zqJs{UXn*U((BVYIZjO&mmaY9XsD;7g1*&uqjSZgPTMn9WLXWC0w72s}{WKA^o{pHR
zmM6|NJn<2esk#CJtI`@g=ke7mPns>_N(4i5m#c+eGi@^Qc%_{>?iiO=@a(LWWTDu2
zd!3XqEKj2Y|4ut`FvwT8_}-t_1(A4j#z3;cNPZde%y5|f^<xf-<Rgj}H{0&)0v!Pb
z(DqGjVpm$E!jBB%dh5B>=%YY0?%UPIL97ST(};u*7CTe~i}z~($@&bvtcxtTnPYHw
zK8ZWF%60vuPCQ%G-*HzB>w^3;-&oN#WZfq7Wiiq$JU7@ey3!E+r`)_v?d{D9ntdKE
z9%hZ^NzzL@xJ^F(e7rGswc%}?SoYOF%nqZCD<=FSb-R34mRr>Blup>2l#YUprR2n5
zWGl*Rnav&}<k}quo;B{vqt-H?jletqjk~vuYolx1ciY~D7AR1H6mM~Nf=i18ha$nH
zI0Px~(iWE#4K77OAi-S<w77e4Deh3DICSTJ{=faa@7|yHr=4$^HCeMJS=VINIp=j;
z5d=IH=NkC|N}sD3(WVtZP#hO6R6xU$?Mxt{pwMSB`8vF5I1MpSMOVON!H|_e+P7)J
zKuQX3U0k55QF$q_`N_v2+}HghHm6bfME8|9W|<j)gxgJ8FN_}ywtd<lPbwXx-78kg
zLc+XqZY-3;XelZ?kE!#0iwwOhf;wm?^oi%<l=K_OEpMFJOD6UkDWxcrLP9s#JaKm1
zPaMw03M^D*P`bq~6zqxTbK#|m?ill~tqkO?uIM<$mZ8y4(g1H+gtpyD;Twfy6u%z_
zlCR)rXr&R0Fn5_HeA6bU3b<aSNeXpFih3SJ#!S;dji{^E-&~c%>g8A>)35@^Du$#B
z7RV07$x7T!&fcKE89RZkl(NarI~}wb6-Q)cGMuiotQC!l)XGAI0^eLdYe3Y8N4Fw=
zJ3sB5mnyJqwmKs3IzwlKsv0TfX~Xjeh9+_u@Dt+-*9;Hz6D9-_My>RD$>Wboh(_*N
zpgVCO@{q?<dy}yA3{@ZP)0gmXa4`#Kfdv3{hdekk3BV6^0BZG>xldnnXbMPm*85vK
zbJ$)?{ZM8$%K3>$5JYJ#ys5SCcbjBbGgl(lWM4m|tgSQSh{8Vba<9rMR5wo9!<TUi
z6`Lq%Jk9pxx?-smuG9V6n(APvT)c}5i)D}J3**7>*IEd*ul0-7=d(NW^f^<{aWZX{
zab*2ybw~24@-_c<Qx$;0yuh&h-igQVp)Sta&c$w$WR|dKNJIW>9r+or;nq-Rij`n!
z^=T4hQ~t}LhZ`ReX}z9_)4o5Mm&!Wr=!^hk>43_Qmh&oWa;Z##Ws><n>qnZi(cS+v
zwxTZ&BkEI|mL>F{gk6C{m$E%~>b75fDq`eMa;GmIE*i%XbpjshOSD?9a4r{tstGg(
zsB}KSCEoCmSVX8qIQaJwYZvfnq2}(3_wxPvlE2i1QT5@UH81JWynE^y-Y09MWvW{1
z501&XUU&5pLxCEucZ2nZty}8adPXaBUVho|gL@|xVWUo`+&Lm>`a_e;ld_cED?;;*
zPD)xj?ed<Afv|4d7A=^^Xl>yHby0)0IeK<#YpqVUm|1I&;GLUTt~uY>jft;XaBwA0
zJz1<|Dw~*U!UXASDl~&7``1^Q@RsKjJ1}w=b4_cMi%p_0IzU0jgQfKkD;{?|1^-MP
z=I$bo!!T}j;sJ`9y|D-*g9<2r!kA00+;KG1GQO01Om3PGRxA0k)w<EY4Qgwy(gfEl
zm^l4$f>QCgiocn)jlKA43H#!`ged~hR{V;ofxFwU%)9HeO3{1jB9%BPwO#Z1#JoN)
z!>H-mFj^x5K@)~Jy5W}+JPR}PAqvY=<Smb8Y6DfeYx^a5BG`4JL-?G;lGM{<1VptR
zH&_PvlT6AxqaIf#j(XTt(kAdb3Jd>vU6Rj<W$?pe?>pQ^QcS|&dX|`1H=EKzfQo<w
zD|92W29Ov%!;!KBb(ok#=5KgbIC}MP{@9+Rap?u{@9GWyiE%KvQXWn)LYM*li^hwG
zDjIy2NH^7@u!lNTkWbST)JuPUfGeld8ng`#WjTixWxFp({d`z3WWKN*AK~c`d5DQa
zdWt;=MMZ9`t%}KG=P?n;Pq9n#(G7K@Ec~c#)cm`p1X8F)ea2c?WWAlrxA}6!;jLtA
zoZd!7Q+8)MkuMo{?U_jHDAF7GblKI0dZ#U|H<&Xj|1bc#x9jHnHm0iSD?I8Y9oO3U
z8B&a9p1Z&7AuxMY_w%cs6#XRow5WQEJ>Q{UcT{-&BBlQyj)8$XWb70Lavog5BV&<s
zj*Y2TcB@UDz^u)8|IZ=%km!qKkzh;2$<|}Unx7LTte_bLNc$YUVV`^xgVZ8VB15ms
zhhW3>%Lbd3CeU+2veDO#uGut6_#Q3feXjJ^;Zc>=dQVpS2jndTV|~-3r-tR@5Z!T<
zxNo5h>M0J!yHWOfg*s7eq{dU>WgrwOEfP7s-sxb|xj48T40LrxJmP-QJf}+~Vqj+w
zQ(y|)??IO$TVzTWC)ovXQ|-*1MwLW9*64+XGiEEoo9j(f#xqo$*ha>00sg*r-gBL@
zM76$Oiw)(Ar`Igmx9A*u$O_EoZfu)$Hg*B0zo#X&NylvhH!I>}tMn5caLPP@R<@|D
z*xxj4J4t|bM|I58PUjeGaBaU5(CTh#gZ#eqc@mG`-dn`5fTos!r(W$eSiz*k3Oryl
zP#U`E%RRu<{yFXVADjy6cktzg?w{NZw!HVFa$cFPU?QQkPOwq=Sp7bM;@1u}0EJu1
z!i$Pum1lidvPBPb%`P2MW6iEedyc1lYq){f-h`|3I&6+=EE);O!fs|I>n$??7JrF8
zOp)7f881_8yQ!&-8rXMX6YEl@fW4|nB35c3GU<X~v7x>uPsktKQ<Yb!B0AaJ%;!ju
zjv@olb9ZYKENBSfZ8eA(40jC${-i>2)u&LKoR><4$>o{M0MDFeETkRvg*Iz+jhp4m
zLT5<H_!-e9+KfFqklL_zis{_=6h@JC{L6;9XD7I0IiJpPLOsl4D<6_rPVlk4iHFqH
z%m-*GvkT>{Z2@QpKnV|@W)DevLsMeSm;nvFVh2dU5yUsH4~;z5+GUW{N4#3Qf%&D2
z%7(Fs61<Dg8yOyzNtcHuq{5=g`vm7A?P37Lu#74bFAlIL7+jJSnAyagoqUk;p~$Q%
zB`)kzoyq*b{}7eZ3i-(q{`tL0f7E5!s)|^k#b(_OX;T<&l#8~7qUcmwcvn3m=tEBQ
zAb|FI;m`;kD~zC#-hHO*1JbHth;Y?$6bEE;L^2g_?EOwm3biAlNX|+@OtB}%qtB9?
zBIErju0uUqzP0WXFy!LCZWi>9KvWhH7QAQYW6B?yTo6lGQ)=T>vgI6Bc5}_n#3W`#
z?#rZEnw4Rt_=yr56n`}dI=lXLZHp)1+tI7)-QhpTnKDhYCOM8><1lul((uv_Jm5h-
zjV6fS2U4oFhmtl6vcQD0**5B)p!nbzH~?E$a|4yaWUx(eSIj#x>M`A~9=6fETgeEH
zD^Q$(Mv3a!jtB<9L?#HS0|k~Rrseb`&ox#~yjT=!8ZWe7OdvW0acEX}=ev%Cx;is;
z!<r9551&wHgRB$2@k88#s5|xNe0ah1)uj9qLM?_=Frq5clzRIrA?bc*9gEF%cJgOm
zW|N%F>%3o9juP0_!{<yIN{AdtqG1*}sl@DHp)#Wed(x<JlV$ad=rAFY)aUIwU!l(<
z2Bht{-e`KDidva=sR#zA;C4Q}q&g1CO@3WC_K-pfr3OCpYQLwLWk)GnDB8s9ZB}}C
z?p*YGWtvbNFyeO|L5$cH5%X^C6@Lw3ivP;vak>=)b16m<KI)-UJsFF?k38`|)Y$*p
z+P0)RlTCB9to%4su2~wZ)#Iqpen_e2eBPB|Vv*@=l4`Bnpb(-{G{Fl&sC<^<#3>cv
zW)dq9Bf(g=3seFN^U#5x#!}Zj4=@!$poDhqkMvwdQfwUbc^;b^^MKRfs|PH!nUr~Q
z`TdhcYkdkCy#Q@K>({g<N2yV?T809v943ILEbLWcOVJ)gu^c`rt@_D1okK7rc#rt>
zAWrnXmDkUV^e9RyOMV)54k{wbbA`GD+9}WR3OAb;v&~+fT1VEwk$=pc)*Uf*yj$>H
zEb&V5kO<}cAH`aG!ZewdV~6t_dyY3SGAs~=f+yx!@^W)D1lKX`Na(ycA=fCjU~L-a
zDlbj7E2~64k$~%r#qnzWz&x5Q9`f?n&)g7=F7ogK-NZ0)TvLgu=i2d`MXU8T)Hvyp
zoUw6^A}484L`%7yr$i+>xJ=yEV=H`Wh1ymO&3IWm+)x%KY*UUrQ4B6mR<F~oYBao;
z157#uSt==}<?|7|K5TdOCOOXZCgMk6ue9>fUbV2gAXa<|03KqJ-R@uxlD-=Hj-R3~
z>nd@$miUTs0G+g<?_u>_5gE$HdW)scu1fsH&0`5ec+s+W_Fp(3XgkupBE-qG3Cy1-
zRx$4iMv3C24$8&EeUJ}mSyyQtGb4LS{x!VJeO1;CMT`V~w7s*J1ec8b%<Xgv5q+PP
z5jvYxN7kk2PRxg)t3eAzi_aa#siE!(G#(1<HqP#r=A?Pf2D*W*IyDDPr0k?+AIFhA
zy$$w+3PGOSNdS_DG~XK|p60<sa#ts@Gc~a_5?JFrnN8vqL*(8|_Sr8j;1oyDpw2!}
zIkbfVSQa}=TvN4WF2tdaq0Pue5S0*>MP3w?L;Vv+OqZ431=MC0wxE$$fU!5NhR8C#
zj?f-aLnIUo!PqW}qf6?D%?l5ue|{ZZG|Is@KTsZk<bf(cC1L|Sce>gG(ZlZmDHBGz
z&)_gI$HECk?a;Xjxr~14=fU56E4pLm+JLkcIYVqkI#9AqO0{w?%Afr>^j$EIr*2Jr
zH%%1f#mM4ruW~<R6OEW{&4{pLCw-ahi@kbr3oZmw53LxvSskQNzj$#htkac6Udz2G
z-tGheW?Ocb8rPdlew?sYX|N|+>nYK-b1xY)OQHud%cYRp=%i%%?i@(ms=-z45!A}>
zgST&S!Av_o2>j`-If~O{zkA9O!G5o*Gs0BFlr<)a_8)q=F#MBj0^&qsGcqdbKXM)*
zdQ?&)n8JH_avN@r(qbIS8u||OL-M_*#9OU1EmL3zQg7(70|%n~G0Wi0E1F-^+X#Xx
z)Uy-p?JiEF4i}1;r1Igk(f<5)n|Nz9TehBiK+U;~;DzWc8^MdF#1HS{hwLVqx}$Fg
zWwjV92d4LePV9+K)Gxfdo27d7bu+`x+i?q`D!o`R6W-Qkizti**Q-V@2cg`R8TE`-
z6%@hQcgDL!^Bcb2WyR>9_cM9}BwiwD@@QQQV^Dn`Cmp`@1caayrKm(exFdJ8YTg%g
zISMr=_B8GFu=7!008Om)fC6fbXX2@p+`QHmyp1YWhO>c+t$YgqRM`@g;g*?3XcSG9
zq{<(f>m4GoWDuJ0u|1#&$raczOG&_Hvg8aQAG;9O6~Qd)RNWqFNY|abhy~AnwT>KO
zn>~+xRj5iD!(Pc&`KzZGRJ1yzqFSmU6Cc9I$HqhE#aNo9tK1m}<ick3w6qv{XVJ{y
zuNlQQzIWx-t_1U7Rz<VcDSNc6Hw}N8J*(#1%o7tx>x*+_?utf7Zpr#5L?Wj}9nol8
zmSut*kKZhU*MU8bvTbfH##JtzWTi37M+IZYoC+vicCY(ZlPJxSjv{Ri5n*MxMkYN6
zfiNu>2#!rP!p0h;!j)dyHe^3}A~Z&jBnT`#q#eUtPnl5qqfghUMU~S~=67s7sv(kK
zghFyklr*g#)u?I*%uyezKmJB(sM~?NG=!)RYFfQ2iDijP%lvKHmZkD+(;j9ZIv1`o
zqVt+&48G3Kuz-`m_C9SV8k4mz#XyOwJDhhM&oPGAqzx5XjyzlSYz<Gjcw+&M;$Nua
z@WyAC1jMKz+xdwP`-{<bLYo#<*kVRs*L3cs(uLl!DKS=CLHMsJ4)zO`nWH?G3Y-`%
zl>7>p@Sd$H`L(JW#yl~!K~d_G;6$ohCFArD1eEIU?UdCpj12N=()dKy%OBVP^M}_s
zIgIla+n$OxCns%w`;QJ-&ydqg-yOiJnGdXoY|UyIEijm<<t!8Sd~|EgiY4eRHT}*%
z=)T%RqvwF;$PaN`^1csh(u~onJ~Dg?+C7hndv51pP%f;6l8IkqL7UfO;0b3AMu)YN
zx!SG8(nN<?y9IA=<i^8EJ2Nzsy2z`u)sl;bW{O2jwUTf-vxAa!qHAgZ<Rb`SoERQD
zipuF1u^pwhvJM&37|mhl#j`vKgK$14vex(6-{g}J3Q&NAs8Z8a5`m+VQY8LCI$8P?
zD%fjMWbsa^nhv-b$)G2hslu-$E_BV-W8C(u>=kYqr)qm@gmV269`dnKGC3AxtuD@Q
zIv}XV(Xau_?KKpmxfZgvF)5~EpbEV)EZ-vUqB!rSFO(;?D8@8u37$+8@g|pPm@jgt
zm9PUEy>yk`s$QLae-&AIfM{0HVHYeB$w)%m+a=M|-r+0hf}#lPqNm?Q9n;XS_8%C9
zAg_Wl0lFjyH<P#INmcY6DIc5Bj0g*_xDad$hL5-=JHz@s-{m3PD6nzb`#IlAcfmRF
z5w57OSG(@k^(UP%&dcrgo`(2v*X2t|XVt-!apc%8cW=RxkB0NO83ZWfK-ljl4^Abw
z#`6-R^<}j;dnZ}>_t8ryTuhafcjQSR5gx)_cG;KUy)=lIt}Q^5cM--EU;L+^J$cd+
zBLoJTI$?4$O%6ZBCPdE2>((8Gck3Y#9u9d!njt5SFs~9(Sttt^k0dSwqYc7hh@;5b
z;zxLym7hl~P|knc6h(@JHs_1F88<o^&0V@|tFU=e1AegE5YFj`u>XBjePrX18md=j
z$j?T0JGofW4K^IFDOc{W>QhaLirg(?7w>@)A6F|Wp$cicn>{;dN9LVi6H+1xT^Pa$
zHD4s_Xi6u*<9aWD7;L~k=UtDbjsn5-U+D83=F}8l<Tq$htCY2tz+bRv^`sdx0bpe6
zFBT9RTbgGlmNKaan;1kQ>2Ibj3y1P*2*RqkX{pbW$I7|uMejzipl6olkfb|zDmvnQ
z2S;@eRKCW++jBwJd9BP?$XYj7a=>IWY`UB2Wm?4B62bFY9X&pT74HI{iPQF?ZG-6|
z?b1B$MdH{-cw8kDo?`O+RJlmLg?<t_5*ks+hN4r4&lDDPSoxW9l&O|3<fECEa4Gxi
ziSw%Yo5iW8g;>fx!1=~1U-$$e-KG#Itu7NgM7|U;Gy?jXP%xy_Z3#;vxz_cvoAWM!
z(Q_n#i(_@bI>+U+P8kjL{`}h?%yo?|KGcQzE{M#?c2Tw({*hadN@t%%=r@KqD<c$&
z=e#5*wG$-g%f=NF7ujA?d}=&lY}=?+jH5`sK_4(3Qhjbs_drEe^++QuqO<913iiH|
zz&>MiKp1WayMRv|9(uJq4jW~VH(0I4R9vGP08*2SO;~wcmg@c*z{n#qF^!9|ER|+Z
zF6fE>EWl+D@;-7e)b96((&uj4ry~wDJY0f*V&>Ikbs(J%o|=0V@^TK%(5gX8W(g~4
ziYmkVA(#SB>V2J`c??-tyGAv{8@Min@7aQc8tU}NS2zmRnaf4Hyiu}(%ce4GVkpHT
z8xzxOftWJMhp7BG`va^mm2Gi5ji>4FI~&`${QS7N3dihE6Bt+Ia}kW9`pH>p#xVff
z`i$^REB%+|7mAl&NX%bkw}(Rhf$V4YKT0=UXnF~Xc?YXip=>^3N(PI<Db|_DxhM)y
z0gw)9QsG`#no&P2by;s@?Zsp}-SLy^BbVZjAt}`A;fm*sbGlyjUW%0;Oso!6^aD|a
zEuZkgM9WsHD=n_^Okgf@P|;9#uv`RACslLt{GxDtyZVA@g2qx&@Vl8dtlNW~i$7zC
zLA_ANvLjACH#6hwZVwnW#VixL<2oCkjm_AZb?zn_E2qG7WbglI+VucRX$Ymwe!?W>
z43Y3G?oIpB%P&aoG8>;_s^w(7;r_~idYYc86xsAzdbYQk>eVR`ct~(@Sp1WI)N2y9
zSxd5E=YE)RtDchISW=LriHXvy5$gF!1T%zNSxFuY)P}LH^|9H1)m+b9);PRA3@qS*
zC~>_~u{%wGmbmijO!LO+8k?!)Pi9iUuJj?)iHckdM0&acd1bT?%Ua^rnvqvD89mc}
zvJW&b?Uq|US<I{$q``ZY3&k9)|32czfPBjggf%!#7nT{}K+S#uLT&BZ*4#3b)ilcS
zsL#=Kv{i!#aN4)^*nE=Q<Qa+OvR!N83@04D^xaCIU<|CT293}&WLNqUV^gVx*481F
z7j+$=Q(^<CMIAgl^ey1AE>UB9J3+%h1%3@Ja&a_NUd1&5<j=n79IFy%(F%lIMSrO@
zA#&hvJBSC^)5gZlbCNoU>!u01n)Hsd5*Dax{bJ&R2@M5V(n6jz+0wkJY<caSm^m$X
zsh@GPPcUOsUN@t+ECf6#?h1&|pIvp{vd+TkQZA=k@`6IpN!Howf8taTnyi29fXpn?
z0ju{Y`hO-Ehhx_{C+E@z`<xtnr&B_iOob1R57i3VA=C;5&24e^K5eO7p~bKPPo>7@
zY(>m}8ZBIsL4bRgdBOobjmc<m8wYR{$fdfQmVx+okXwHrc}2QlIXqU<g`@a(tJZXE
z;0!+uJ$xgF9p1>q;O~;!FfoyqgZ&@{hUYr1Xd#{sukHi$UWp4!0unO=`^?lViTcv-
zK$!ia&h0>+>zq5w3~e(zEPP8Y%yKn*64%P=M><UiOi0IGZjWTbGLuy$`a+}fV|{tF
zWSR@DP$%Wv{PlrsS)#Qa;4XQia?-nEW!a<%P$Ht4dP=#TIj=i{w`y~F!-|zC??QQo
zQwOu|`F-A$e<vHqKGk$1F--rIg57<SzP^A@7@Hv<!>X~;cS=pjgw%j{Qr+hd9g50o
zOlJ=Q=V0DewmjsY!{(BUJ_OO3rV0t_e%u3EjKr~@A>`_|XD>)wC^PBd{<HP1T7uJi
z^-C67zk0shJ>}yTKO8iU0vOm%NG-LEH@=$xsO)1&^#Rw5wP?bwSvtOB@;#qb7<aDJ
zO>8+s7uqM`!#lXHL*j9g0i+8>L3;MSfYVC>^233GfKh{eoEr!A!B|>ar4$q&#|PAh
zz7O9EusiuyNS>(0f9SBp$H9IcRl!9BF2Ns0>vA_Ek;3`nMi;a;EaJZ3^41=Vk=L%3
zoRnTK4O1kVX_oX`IQI$~m?$?|A!ytX4-^lcGekCiCB#K*4&nfNs^&X8T6?ZLZyrH=
zkXm`%;2d$EDJZTr{veTOPXIT}>;;#q;acL3SDxPk8QBH1B<Y?IucP@)$gG(cC}I3N
zl?%)}YT;Mhj^0h9R?^QV>Z7W-MY<|ENTQH}IQkyx2Cw+@1VpjLdO`$5DULuYlUwO?
zC>wGpIP$9=LxO%}WO={yfJqK8E6rMp49^BPD&*A&y+Rk&$hKFj{af&Yk_qMaW=Kf_
zvEFyiLFeicqG*ndJ8Nxi3h4PcVY9L(qJ(xn!}I6ExS?a(gwV5uF2Ow#W1hG9^+v?~
zlN*pIVP(g8Rml(9>~-lnpEkdabnl8)(bzh4L|C;~)tZIw>GR<6M}cElD4lA~l9aEX
zSLF|0xYOk1Gpr8|%FU?BPpc2~!?KE)F7g~Hhsj*r!~EBa{xLRyRJ+WtHgJl8*;ks&
z@utzGJ(*7P$@4&i?Vafq;zdCaCG36V?a7iw@mV#cp=taA51SgIC8yUck~=Ig_nbJQ
z)*8zDkccF`Q8NN)s1B_*GgD<3w&7dU32VX}DWX$&3e}n!-Y8gF7HpW;RcQ`et^tYm
z_ga~Dyh$uy7>S#}N8bPHA5!V-uoB%Znbm&2Yp?HO;lPh2WI+iu?bb!gyP?3Ony5sZ
zOq(1K?SQF@OYkGtw&f1-UgA<cS0ZnwZik75ree>hOCkk+%YqhRM$JxQSYGB`-xf*w
zw3MBg*+8i@S5?hEu~kgRxewDfrk!Hm7oU3lw11w$nyAI2(!(+}U|m|(TAq<Aaz7h`
z35RbN>%<DcN#SwXaMiBV1H}6NpcvkomnQWU?|pYcsi<^lr6&vFRK9X|cD&)2o+u^&
zr-Ah)GHYW)YWU>aV3(0hnuW9>9v2aHROyrg_p|!YE2-<IVQklvc3NJy!$F#oa17p{
z!FGQ7zSm=A`Vf*ymTF~0c0Rp5URFw1uKA{H9!~F(l7P_z33$&mn(H^j_POqvE==zf
zJYMBF{bP3wBCjCW1$n<T7Yo$Vg#+JxTlH;L)lX^+uh0#xb|=0XNf#!CrsRI<i=hOW
zXbD&V%iJkv<IaO5$8tEk$(!p;5hL%-m8LA;Mm?ckT)YzvuP4c>ZE4Y!DN+bW>-}^q
zD%C4O);UKbr)K>bL?scMGi>BUUqs3v<|c!YX&<c=8k`6IU`OD#D3qWLi!~CP1aDe{
z<_e~5D;@i-8SiD%3Eq{kD^)8gt3DX|gY3y7;H|>DEJ`8}(-JnB?yG1n!*njg&m3If
z5YB6RJJ$X=U$%kJQr5m5Zm=WHTnYpQ#^mPb`(I*d+Wg)k&DUhfeZsgTq*Q&f!hV&&
zw0IsxG!uGN`K`;{abtKat)YDvSUB=31fKoEb6DsDUlRUtK-m~4wOY5s$ayU*4An_g
zjD~>QO-uW8RYNr4n!Jz_MZ{nn#fH$GvM0z<NH<9qVDb(yx*z@}nAqeI=J~xL-}qrL
zzP*J_LkO?QKLltSHj!XlRc!_htqnTmA-RM&Vd#d9#XvJ7P@j*Xe^B0?vmylFd1}9$
zYs{3le9iCBM%|x~Ul*a1nVo9YwfQb#0arp#5*pp-QC#XMCx5Tl7YE_)Ww&tx!tSb#
zd?Zt{zMFIXu~{v%x#~E0(P1q19YvDqm8t_rm!p_}HXNeD;c6vlJCZ#(=;>Z7zg4k2
zQs)YY<GSQ5*XLR&vd(%nQ$%f~)_Y?#ks_Yok)2wwZo*(0V5Ia}D-0V8gf9%RfMnZ%
zNbV@?qndmKDmQJ-`5`1YO~t(H^$ZxVhJ_GA3txE2_|BEvXT=N;mZ6EAWTZ#uFVWF2
zJE#1-Oa)?iIei~7tmV+n@a@j<O^f^{C|I`$F0%h?#x|4aq=o#1V5*C{WRlDOjFb=e
ztbVr+Z=LwtC}NB|rn$nn;AbXw?w6!DvZ;V<F#&)@M>;p7ccu|vVcWpVy|%@j^%J!V
zpAIpJA3o{`jA%m%59hUc)ro#yt?&N4J1Is|KCf@mW=&Zh8ppIQmlEJ>_x00N^>6w%
zK-Ku04nGtXI%D~hYqoNoUh$9IPUvT8o!sm%vo&>O-HvP}ja94l3`R=+Mnn`uL_0IB
z#ryf9{TWAG;XX9hh9a&J7jQ^#rS)`K4D(b2@^sBe(15RSYNwW!WbTwf38oOoEK8^o
zyI<N$7yjC_j;S}@shbVwN;U>rjXuze+D~+Z5@^*#I)DvyRe`Z-R&^k_fav3dIRZv~
z$JklTj07VBPF>*6hND6q(cpriN6RMLDIft{9>c-H!9z3z#voyTSP7#@`(`-6!x**D
z5Sap&Glp{RJ-qM=+BW{9+!dfaVDV<CLal^%;P_5gwl9kQ+d)x=k>Za{!BoJ&1WYJ`
zWL^i9Tsk3_cljWDVsY^iR?x4#X0FCxYVYUxJ#%pg7?J|$4q@*ja5rweFW2RmN1}BZ
zVD(@Kw}J-!t;D+bkbE@kg#hwuG%~UV&9T?Cx6vG=;V|bje>*%|q58TSb>TVpuE@Yk
z&JXH5SE0pr-0n1PV4jX<H!gNH@!eu`{ywSJcIMzZZRk2h$6T~v*q0c*J^fh{JJmFQ
zYKw|!j13KOJoowuYUcRXKk_aua#qcKwQig=elb=-&JsKjSpi{3dQ-X)7cnt^<1>td
za?ui#4iry$D60S4iiFC$nP`$?%rYfKv@fp&n+~g7ZSju%BZ&CnVxMg-m|&iIbnn_u
zK*n>){(hT9;5}FK?#wT)=KU~w65;tcTI1sJ98Wu^%Q<@$SLuM0(KW=uew2$kZ)YO8
zk?Kp#!h@`fM2R)Y$Njl1EJ7Guh^ZVWWy8qnPs_nHM=4H2@%vX#<gDE`KZyc8pI?OX
zMQ~3wP}%Xo;7({q%M(YmxP-v3u!@ZOdzVNBh<<`R`i}!CW%O2lS^0+Q<j1>i>!Xn9
zrodif-kL+BC+Xf6QUn&BVGbCS8oQ@drh4Ns9&^J-O%@%N$>O9v-Va?RQx<)x{7)&8
zEESx3^Sk43sH}ruco;<OUa=_6i}rZhX}gyAsL-sw(9jN{BOtNAn!3S$gFxHTDapjq
z>OM7SOgybJ;vx_n`p(j$E5zudswel+q3UTUqzt=3HPu5%RmYx`?Kwg0wh*4Bx>8SS
z=uWvMy=3M131x-5d28Pk_+L-p)Fs;kfvy*?hQPIN57ou(z!IdlW6{1-^=->8iNhoS
z0`tqRvs)&KK7?TLXX^dTux1aq_SoSkk0L1048c{|m*uzawRZ57hIpW_vTvn#U&-ER
z{veSMfR{i<1b>h>;As`Hr}^XjDEe7EW9MEwCsK0)ckMbDVgsqFc6xLv0+RUBaj17(
z=$nXQ(05$lV$Ct=Val$>3z~p3?ke@%G%K8cErKsFE6x)ttpBKk$Z2-I*8nQ;w1Q+w
zh7(@4F|ucfO#9^W>F<f9PCu;ss75EV+A^`>pd>OVsAwAL2yaZ>11mGN>Y;u{Pecy)
zliXO%aix;2mtGOZ1&v-k#N-bH?Isl)qfOyJ*`Kr3=_W{(@n0&cXa=QzIlb-(h9_>$
zGP<0d3Bf2U-+nttdpBu|dcP&5#PpkvKl7YPzhpB@Cf5g4+N?*du)887q52gO1cIQN
z=xXAb5S{7uH_L?KE~mp;)VRHldf!j`-ORVBP2BeN)Cd+@T(Mh>i+=<ym1$+#WMeqZ
zU!p)6r#3naVnqnFd99kzdgy9ns8BoVVPYkU_KQdNw^jzbQM}b*w*HwDPf9GQbOB?`
zU4fx0`EO+H^|qLog#x8c(W;ETRu$#f5C-0(asB8nNYml|^s}+AT%G#WQO3F1-{g`I
zYGM*z$|D+GY3>+QoML=BSt>SFb{P@bNOV`-S({3w`uIWW&~T87YYoqpFsNhS2;hp#
z6yIHA+Wdie`P<}TLlyT>3j|WE%Y`y+H40BBF~m~lbPrKVL%`d*Z2QwjCFv^CmIwUw
zIx`3-FXC$9F>nWgMt|JiE<Ol<W00Y`TLo#IF_USY3S(C7y-X?_k&&oQYJl^6kO>v{
z?5IUo<g`q1>v>27!c~{Bbc}7BDNlWhf_avR<uw^~jjm|{i_pXP?+-(APOF2hR4A;=
ze_esda#f0h?pdyk#oTT27ksx=<)!r|vmb3IU9p+mLhOxahWkxY{)|V=04o%Yl_keR
zrw_@L8NhqL3;T7DwVr$w2rC7ZQ!4uisc3)G2Rc8%<npJJpqNVM*ArIZbQQy^k*C`F
zUn9Z`OJ2@_y3R9(CRB&CtHKPOz6O+r+PqbC<4ZVM!cQP0rrs$%Z-br#qg}#zie`}R
z$s*NmF57ct)X9tI2AGOE6jAzqu~4<t@L~*R#H@_8-a=H{K$Eg5ZXO<wiV?k*Y2xL1
zVt=J(Ye8ywTD3oh@(%Sbi)PT-cdOLiixt<c<DyEeTsUP&RiH^JapphEeSB%4&`Cph
zAvhTXxbT~Vu_)=!tGv`<<rk!uSA}wgTWzo)%hOPboTW8+gHy<4UvOw&jL(S2KJBzk
zkDS?rzk^bvZb58iVK+f4Mig8$XktAKI9D8lk0+kADCD<2#Yv0Y)L4}ZEk@O8+eaTJ
zwh)~0BvMaHLPh>Qa@m~zQ@Pl2+cBa>Iasl*2ev}N<}NlcHx}@vD<elOWd?k$BBpC5
zEezjlI&sKWYd!j7Q_zCxvLE%u*Rk)CG`|c96713S>$s$WHhl6_1by0)1L>H%Mm~Io
zi4$uZ+hyb3Sq_|>?YXA1$cor*P(iD0{xH#+BXQJ{0ZMJ$Qn^gXSwl+(;wBh0EurMi
z=gBa+nO+=)Ge0=LPmeS}dDhEa{5;rbJ)28W5r5AAyN6)*vhEqxVXRy+`q(YoUdybW
z^+Jkap{C?Or*wgR7b*GPuf$8UPylDo{k=&XnfB3;)743OFiNW9VeE=wE>{J;=pSc)
zl>rBp?3`AGZvHY?qI8nUpNGl5Tve_mulj&?AW^dEuQ!Yk>Pw&!1_)S6nUtma(_Y1Y
zzf~dUM}2kKfYT>Xf}RfH4%NLoE|!nOe)1!P>3yoZ;jXVeq7)pk2z1b;1RBPH{n+Wn
z;o{U@yjz5HK4(#Bu(akLvx-dJ;|@_FARyafnj2ML0Qlv1TD!lK$&J4q?p6AHEIL~*
z=jT9-kX#1_N5Wb(Sl&GIp|$0{{#@-)w7`)elo#vOYi)K9YS(7~bx?P5G?tVyV+bn{
zX_+_JUAX4A;XUl%4f-dMTFHku1<#_%(xEOkIKnPTJ}u@6=0P6&zj7c>vv#4mP_}?D
zCA+-GVrfs!0v{dwR7A<+ey`S%Y2BiHI9eQ7Z)W|#yuFS08!|E&U8&Vwt}lG(w$-?0
z_7Po1?p^|>^4WE@e4AuFn`&H=Xq0+`idBOG#3=7I3Cl*OdW&1X+#}9|4}QkFZoQen
z7a*J3sFI>YjPL&b4U<aJOb-LH=_t*{Wm+4y$V3o4>1ffv(ky7Tu&SJ2net^$b)m7P
z)wTzC*q!eXS+NwT_87+6M{;_IB;F@!Q4S8MHH|9R6mL~=Wr*rAZ1|H4$*w!Ub&d@z
zwE+eFvl8#h|M`P97sycz+U?z{mg=bETmrUbS2Jz&w6DWz;++0g#J_u_=K6>K)QT>&
zb|CvK_+y2mB)Pxl__@8t_~C1+iS-YADa=_&@mE?q$vdujy4H4M)>`sYu?i+0dR6CF
z`!o#I2Bi*p95h768@8KvA)vYgECFMvK{{N)N%?6kn+GOk0pxDN1+X3$*GWrE2IUvp
zKp-4wFpY;(;)B1FRDYJm{IfrG_fBsARsc*4)W5&+zt1$Sxu3qg{ljtDbfot^OR^<!
z{N%N-=L+qm#^T>cBqx1AOzS~Fp6+(x*?;C&Ky8E4tG0J^6RvRW8_)ZH?s9MFR10+c
z33qzWN6!d{f`z*TX?yD);`{v(Ew?lFb3+%c;f34_xvFAf<(7XR?Hq5^ko3Jye)xDS
zovx7nx#IZmqZBX6xE7`1eaLN=>7-c=`n#lwe~aG8f%zObdhSWR>rqlVi;LKOu?0Ao
za4#h`2B2p1+c}kN7O|uW=_{7>n<dMfy0mIF_|({k5C%3(3=YKClT0ePp9QwHBG26q
z6);#1S%unCOO=q2lEw`6;zbj+g2{YPj6p^L2xdSw!SQFh+mb@#Yr?enW>&8lfUk;Z
zMAv;ML;+oj#bq2~kXF>|sVos|;E7O3kDabJq#m^JLh-S&mVxz$rlu6~^TqZSm>x9h
zEiLOMZI4NKAt!^`MvplmG$H1CT$#~K*SbxIY)QyEs?fcv!naTx^VT^sW1F?0qV+Q?
zx%h}+Z6{d6&9}4Q5ABZkq_?~u?)=|Q;8W2|ZM#Y_A#~YJcwx+8nkiu%Y$}(mv6+EJ
zfU$hpuT{a?RM;=PI3<>d%I{e=?Y%)FA+$Or#L4y<s?W6meDDG$kcBhe5`eK6f(}nS
z6>xIbY+`R3GTMOZ3>{4O;%79*#Aji5Edqx$rkt;uo$pMN0>&Xi3^4KPOu&lYaP7|X
zFm5LdCn9NGogE7B-zMW^d)Xh6<^!i1`95h{aB+N*^do38L#acg{gJjSm_Pm9&zCNX
z8`K7LDu_|J+z8V$Fd5@_b~Q-T<I}cd%}-NN01Fb+xhlg-O{nF{h9TpRKLI#$#gh(1
zB`52{XcI}CshvrpJVw%F*k|OWs`5Y+y~9txIr6PNl%C;UaW)GEv?R-SeKd5K9ZTU?
zVSUtJZ6`wKu(||F3Yg*E*DAjsR%kt;UCVxNk!@=phF0;)`NMxu&}(CfCPRqQa~7F|
zFA%LuGE`!&%8NI27k1<L-j6N!XycLKjGXv=@^Vutl7iKnFwxriaN3UAV&427<QkH3
z(3VO-HZwp`H93InXFy9P?`0dKi0NxMl*B^SJ#rHxR$`PuI}&?{|J?iZtPmQ@pV`H#
zqQj2gLnp}q)y|Z_K_cpMh5RZaB9jEQM1QIByLfA*1Jd5vSq0yfuNKf%pQ0*jdXq^I
zJB?qiGYBCkmc3+aL~4=NSz8hv#HUTbq2vh{U_y(m8SRbMPv`vpSnHKRw@ds4<LUP#
zYLp6lYY?Exq#j>#d&rM@Ky5+)7LG>oFCT1;{8ZGq92p;RlnWrE2byfpb9RtQ5Qo!%
zQ1XdT!8*34Vc+Lf1#b@uI#V5pO(nk9{24bFHxQT9kyjCQuJ3k6ONJQ~jD&wIH}?2}
z%H-A!dJQgJp!`e_<!kI+S7Qt*aMp~*(!8|SEh05e_L@z2)3E-1NwarZp);^Z@3(^6
zee8YvInTkLZ}aL<*Rp?<NO8}h{zN+Ks;e2yR5@@l(=w?Vt>Fkht?nsbcx<18-LOXz
z`OHRdm&TOes8aLmuB$=PAw4#24@Le2YsPQDq)F|{BH-q)i;sHzSYkG`$co=cObYyb
zROOJ@knvaaX>bxX?+<Ec_$qiQiY!p$rH$_Xbh5m0GKHfF`DeGq^H4d|k`TPrFtKqe
z`aA1{HhU7q3nlAmhLWqp+S_cxi`GfnSCh-Rd_l=Bj@fRqgnFC2T{vvU3LFefdf_)h
z<b~r!DK@6rS}0|{ij4T!D?9=}{<eSe+URBcvvG3f$J;1#JIG}+-<5R1VnkWU0#7WI
z<YLGSv><*K5ZZ$Ex2ciH__ey%WjOW!Is%EP8~NwOD+z@u%16Em+5CfjbWm&9=oG2V
zi8#!_pY&L%|Kpdl)EZvz1lLxRiE}ci@r%OO;?TiOZrsE!<IRb@1-urrmbmk3i@s%h
zpuZs)3;N^{n7?Qg;gKsD=~DVa7PkNxfyo7OOCUuX8=005`ezNxQfO_hF);2W{$BT*
z0PM|$_gDfLg{nC<RCb@q%vO+E;tmI8i@ytFdLNMaO*ShtCxU+}6`?dBU28>Y<>T~S
zkx+N9J8-rPTmv)Ae7f<9Re4ZIqmtu^t3^n-H#5C?H;O;TV}`5Q0XR<s^1q~UMlMY4
z)%|oDcMvdO^?W|eHh13W44+<f%n%w2tJvbHtxw#cw3sYPs7(*sL{@veFfsV7S)`)J
zv}p~g{2&9@-pHDp^QD-NEoE)3EmdInFscePN0FL-1v>;DelB)TbHEvA)jm#*5>-zL
zAU}o$C?+2H7ct1o|EmMaxqDgwNlkEK{NZKRCeH4tU&pq>yuOvai;kTtD;(!|&xl=_
z+r8#J4DVMXYazozG1|V|a#33FvEObLiXm*Z2}udNe?+S?rSSU<Twm+8=C%XwpV(E%
zGz*ST({~!`W|V(8U{T9MGqaw*;1E!OiZb^KaVn9YrOnGgI9p)Bvg`eWayk#JYGRkM
z9#k<W*sSH#g$S2BQRQrq#9|AxE&GeFv#zA1C8Y7Yu6cELnhB;^)J0>y8iPO<EdHOV
zS+wnk0b*rl<VIhv96p%*@dE<zbrW=&oH%5XWXg2dVC=2f={mVjOk_wR)>Mlpzl|hz
zMtbQe9&<UdtZ!vh!8jC~KTLd;&ik2@T41Rf7~`tOj-^HNoRn2?@6gQ)!Gas~qI78E
zJEmk4vw6J*HBM6+XY{{zK>f%|FBQB=X4XdM5QR(+mSL@(^Ie>XAMP0cC7j9pZR8G5
zWS@oy(4SO)yVDY*8CmKvPW^qvCxzeRQ>DQxvKw^~^v2pS)ox9SZ_Vl2L7UyXjd^kp
zvmQBab@0XqV98(!V<EJ^jYO#ycgJy4<4n?l;um7OVRWHFrw>y*H75!eem}+d#z$VD
z%mcZ5kgL=<p}gJdg5fkBe1(=etRg|V%2_c?WK!a=I5LcbtWaT44$8$t`WyT;>IpI0
zya9#2f4nXR_o~xQ#Ylgt#b~6!7!{>g{%l!@uqdSku|-{S8pg0^1u7V?@YkFHwf0+^
zwEd2F!bL@eQ`y$vSUQO9`P0^XDak>Pxuby%usT?3bb;ni(uM(a1|$sJEEO~HFZC1H
z1$}ui2$<U!&2XXIaFMMBwkX$n-Y&NnX<-bJeClt?`eie%MBLwRve(%+UE8R3hLK4y
ztMb3_5)z>6f#O#Y4O{6$<@_$%k~jb$m}+8dsfWE-Kvt=qEZ74W#luB_hmPK=8|Zc?
zH7s`k8(Hh-Z`ecqeE1En8hNOyTUEQVih2>cM}~){xiq)8qy~YJV0KJZFazHdUdb&=
z&mvk3(s{7Fc7DR%sry@$!sQ_01INP%2@)TW&KD-wcOC6XpGikE<}yv-@d}ib6v9I~
zmUS15e-vJUAw*^4f7L~Dq@kNx-fv6hRMmDZPI>9@PVh?%`3xy&=p^*bCKz87B(Uy9
z?u&9qA|qQ!-!HjUgrvOKH!K0gna@UM^MA&e-3ROJ6s-Xp+S$>QF`guFv6P9^3Ajl3
z)S4Ek@Y|GJ!b9MOEuD%k;~o*^pAa7H)GU9)$SN`twju>LAabkD0#c_*WExfWaVjc(
zQGaX+V8}_fVDkGlJ`#y34<V}yM+0_JjnROMMSI)uwijCgWfJn0&d6m3Ybk58xmG8m
zFBJf;>u4>j0r)tol3ie30m|{zvOE6vgh6bQQz<Z<##PB_JJU(oFzD~2wUut$xsjGI
zS0I4~?rt>ok}cu1y;en9r;lXfL3;6Ik@XuX@zwR6DhjMBOQ1y)2JK&}z=4s+pr#l%
zR<3I)Tt>#3^MTTX&0x+T5^D8F$+R?tCV@6qL5mXYhJjtmyKCz9A*0e8!^JiShPna(
zJ)BZ?oT@@aOM1(sYowV@I(uZTWG^TKf4a|u{g@~{=B92)`hVxk6_3Q;MY&igr_-cN
zpWBUo4;8Ekr4;Gn7g133Ataqz9qK7sjkOIs!En8kMKo>c`^Fs-fozR#@dZ7pApq;&
zO!-1jMJ8X;NnsL#8DuO_3k%ANn34cw73!aMoZ_TaaudH`9)))}8k8tj(ls+=%rs%7
zKO)mI%SC8&fpHTzyy?Sl7x$)+xs<^kt`*@UT@EsJvh^Im8gNyyaJey68;N#upnqsM
za`=*|T7-5n%~mPaVbQWrT<ZwYDt`BwTcxau-%9^T<VI?7H{#^oML_Qj<3!5S0&^(W
zk5Wq~>h%By<v6X*S!H}Nd=-FsMm8mg^UmJUQ<x7|nR@DgWmx_*8wU3JH1bO8BnKLP
zRJvOBk0qrP!#D62qH=$;#IqR;GLL705MpZF3XCyZ3gMY4hxf{Rh^Q+%1t>b{AsKR}
zz-~@F+0->%J+TE_;?G@1#;e#ax9ETUxwJO~Fl&}frKuT{X1r*u?F`COD^SVA$4#uC
zTqnx^z7G*Bj`xF@QG5IYWt|tDbdEk*rc<iRGw1Y-;HdlPiplYvSJd8n&oI<%N@909
zmM`bX+|)PI8;Y)es1@gmWpf9XazwK6o8-1m*$Y%B-lMW+9Vi7~0V6Mg74DkG#i=0i
z#_?X3YsTR=J$GRxb%_s$OJTM1!yzu*nz}5S(DbQmb-3UNCa-;HlMBH?QZc2KkLKxG
z(PYB<9{eHrWre+lts1Yt#L#LfUZJXJ%!y+WZztGKO-f6*IH~&JGr-xfK$r*Mmd%Ar
zi%C%N;Y+HM4`3~KMx{*edxcj@33mDnC!rynheFwaOi3O-N=8MRq{Z{_E8gW<*PLxr
zPNAlDxF+?KQC_Ccc@C6q(aQ*AU0$3iiuG`zWWTXFEA`U!?<1+dj}jJ4mYb+YNz>wL
zi_zMHU)a%dskd`kH*qE+Zj+x%oGz~jtl>bdJ(O-sg9F&@%q1fLZ+MYTYmsOD<W_&3
zEjsnE+hOnst#f|nvzlA5@i?HKd8uChDuqB&wj8B@hTaPZz88pmW#GetANGvE;H(5p
zIH*&cdsm7Vd@7N;qrJwj@-SNByXXG*kuvYAzmHh@SMWo+tUO1GA;AwsIRU`|cOR^E
zZxh7hbBPk}X{1j=S5n^MUHY<&<pL7!sU$E66r6U(#3%c1^1n4;5mAZ{(Cn)Y@7;IP
z3wNM~OQX$8S|)}YA@6mZ75Jm+hy8%I^{&ET(8!&+`1v==m8FlV0~ah^8yI%W=zlg<
zyD(j|AOlEzvXZuid@%$a;&x>`dPfz!Qw_5cT9!9WYX7Huhr?GJIqjNjTG9o!LA=>F
z?_U|+mYueDEBto^G6o_nfc=W6`&~ZCaXWJ>u`EH!tTW*GsQ*2r$G|4xKA)`ryPXVk
z@QSJb?w^P?y|F^~-(x7^JFA`kzfR|W&iE~`y?KS>-;sy>eN@Bp@b^)U8Fl>s>acw~
zD3K}Yi!Y@D(kk-bvAf|t3}xTmxc+@KpsK|I^UC)nBu+x$)Wi)+*}r}WUdLl9Yn)&=
z5;u!JR42P<fA7MWE1D!7ta^9-m?!%vdmv`t20oyiacSJ-cd3?s+V{lze&yPl?K9^4
z{uX2y`tV?=$osq9xPzT}s>$VB-)iC9qPXqwRP)vASg#1~?}HwuIM&<2;@p2YSlxmP
zY#tucZ{A@w98(VVyD#T9{6KXx+9STI-<b<*D8ynVedf*`%=Np&Vb!1@tFd*MTc6bU
z`zYu!sqns2n{Ku5@fLWE@U^JBbl~7sv$<RS?Mlar7OJA^W51?U%AaTIE`J{#M47d1
z-c>}}+$BidbE7^!ksQ*yk%(K-YM1(_bNvtdCGSmcFs5*JvW>!h<B!nWv;b+a&-r%2
z>5Hq=tyRC9kn-xMG{O<jXwTX%a<9j*Xt(W$dok>m{C3QnXX;81{hg=H$J?KXj(okE
z?wgKBT%KH+jcybxH0>U`-(uvHtj8X|Fkk$UpXb(AYtcI5r1Xbb%lgUSn17a;NBiN|
zS3C!iKOEQ(R~i^cmgZpoSvfWCx9*6)k6PbwP5K4s{e7ep>YrtTdRV;Neq=)I=CiT~
zx7MKF*bm#lOx2Oko=IPVGygs+XXz5$IBQ?n2=c(LcF{P~zNc3Qj6U=QUrrU?sejS~
zdYgF@a!$no%+6KY1^WXe`SI3>J~q!R1-0~hv<)>Jn}MshyCt4QH5<d#?yot{KTGRf
z2C6yGJ5HZ|97A;FJj7()KG<$|?{l`<5NV6UkB4$ef^Yb}kFRqJaUBe#Q-?pE&@%sh
zbZ&LrLoBWMxHNYl+lS=pzhUr?iLc+i(uoqw-$yq83A>y(z?swv%!n`{Eq)SuJn?B5
z69^p_|1;TftbXOl>#@)8q&qJB`S0e!7zannnR)0bEzAEM<XGYVC+b-kQooK)-oMYi
zQA|;P4lHe`HtXG!P*ym1#Ehw3`Bgk7w0Ig%*j|s<w3DmfvtnaDw(&phkAsg8|0@!c
zpSCmpF^m3R12jwYyZ@hHe{1sbsG;w_f%~7`Z<5w9)7tjo->U;=2t56gL52Ub2P>w-
zoBwBjh@bz{;hX<-h-vS-^9nw`Uar`J2JuqvPdFZ7PCobR=pp7(Idb*g3cLB#?D3kL
zM)KU>N9{$JsS@;bp<(y8JevrA+{1ACTu;E{-$yI)uL6&9g#XlNZTM$l!Zic4srtd-
zkx|hKrj@ns$p5zYZcd^vc2MJWc2RcN-FTk;l*wD0>lLQaA1y}#P!ZM4-l#;xu=MrY
zOrIM>#n?nM=Ax1Q8+^_F<TGRYL4StluGq(2>W==~?Gq-nkf?asMU@miW(GK74mBAp
zMW&9k0Zfg{_i<)debz+XRHQI!F!|!P(Wke+n1{nGU2_4(oq2lyRqIXb47=HJlkJ{x
zoSJ{_WPIUt+cMeljcunuN~Aib^t|HC!*yct?a?Ev@9(p?q%#;@q<2cBv(H_(uvAmw
zY?xB1VZnj}^!$5C3~^>gh{5%<C)ZR8Lbt@NfaUZ*jUz>}B+7#(`ASAHj9~6i$lCN;
zotmE)9OKL&awrcg4>cqrBC-=n5_f8Y)Yscr_ix^&2cPg5#UH)h<p1Az9IRy8YBICC
zzP`Jjpyj+Ujn%d(?`$-wf5YYMtMDKH?)g~EU24QUFsk>%n@5kp<%LSCn+NOj7;Eam
zZR6$-q6&U)6T`DN_1z{dhgI)bZ!OO5K&(9fe(?X7F9Okr#`c4tH&`ZJwkvbX0b<_^
zB+$gEv`RS-zZ5io^Q2x&xq25loZdftO3sPc{N;D?PvUMn=fvV$-HhKe3pFj3RQ|Rz
z!BTGu-#;QHf`9_U1}m<gf~)ggO<LMr6ncx>9|4cJ{Z9Y*K5YQbIkmO^`utAQzNKrl
zF|tSZW{$-V`J|ekUs7X0VX#Fb+1jwA#%o0Ma5md{h_cmbNBzRQI-#f#!KrHgQA>)F
z?qwF&niRGnLW^Owc)?4!No?P)wYh0d@-q%kv5m#6Tc$0UUn~4GQAG>+sRp^aS$L^*
z@cWj}IphD|{&wU;H3DHUW7$zh2br!RU3}A5^4d?0t4mVun<BlsO{>Iy7p!r08=)46
z*#0l(-ZCzZCrTHdKtd7-BoHJ(fP^5yU4j$bb&$acIxvH~1VV6khu|={41>E%&>3Wq
z;6a0HNZ$GH?t6dt?(W_FazAW+>FMg~>Z()IU3Jc>^E}63HAVkc)q<Wxt!!loTqzoh
zr!L2qwXrA_vEMPV=-m4X`svj=8<8zR=+`ZJOKYoIDy^qFg{5_;F12qY6EmVU(vmDH
z1YdGl6|Hqe{0tkaj(u7I)P<7`Y;na9=jK@QD-li=?-636!w3|O3s^`okU4xu;rKP=
zgyt-3jg%PFZs_c%Wc#pU<-yYDbpLPV@ZZ%YIKv=Q(w8je8e-NHM1GS<Kjg$Z7)i>b
z=q(+(te-$(8KRCba<WqRNL2KGC~VBRE0Ds9m)zD<Et-pq^!1Y$W_%_@tTxs1j(Jj_
zGz_o$d@=_32!Y?VrIpH3+iMng+A)ZjFbvFS7z$nr+xehN_YcabVLZ`x^Uct>Ea+s1
zE5+WPWYCwf-+HkJ^{Fxw>lOhm(V!*iH%aG8ri~Q?C2Z;?yo$Sh9;w`hj-)HHus$?2
z`)0)V^EUD&;YK#GE%uYix5`Q*{S}yiB)!5Jv$63H`fLx*66@kj`B^>k8;SY6h2ABM
z%V-VmHnrmEFl5pz_VId0ZKPdCCnuCDDG$25Ev{dQ%ae9Vu25duRHB858L#ZpiwN2x
zk(;y4@u%FPkIU5FqCca>ixA92-d;~|vrxZBak+VZfL>sYoX>@dbwVc})z7Gkw7<b)
zEH24&yYu}W@Z!Ib{1<NSwb4mH9Hy8d{bg$IPk@f%yG_%l005bS<y`OjEg-#wccXK3
zcT?wqVR&0Zy~61Z%73XV+y_>r&@ATL8UdK1AT#p@kS1duNtSA2euPCoCvw-KbjE}g
zUq;6B=Ai>4V(U5p|NQbcfUv{pPj%J<;oZXv0HYV<+^O$BPx|kZY@wfT01>7fZi0ne
z*<ZjLtXp8-{_N)$*a^Y5o19bzu}f`$b!Cm0a_Dp0{cV3qk*nLC-E$nrx`zciAjl_i
z^k`E%8SyDO)Xw*kL*!TeXk~FW-1|0rit>M-6JA*7k%x7@ZaGr!9<aD7tiEUcy;DsX
zyqL*fR9O$?H<@sk*dma?G~+>EsrDO9QxzDJp|5y`sg`b*nUDOGA{6K06Z<OZ;d;v<
z1ACcpQemh#gVfR66mkf(3B&qhffFnYY~yT+&hs;RMPH9KQ(py!)N6W2>x>)I&lDLF
z<HcBLI+g8r{}$OTi{G01q)dmvx^1WX9%RH|DXd(VT_~#5#MtDv)kQHf_nTIw0YMK0
zhTq)vd)CD$B}9l~1%>dpGqE@+YBXlBA<Q+)sM)g*DI-c1V-^*dlq?k307EFbHun31
z>&NaIObeBXcGfgcQu&2Ru90`ph2AlP!fCR%35g8T9juCcxRGLcMB-5*vj*$)O@1?u
zQ}}Kfip1$tUH9)!ZL<t<CvuD&uXvi>n!j)N8cGK2*UqpNn=R?|J7brSc3Q3Qh7>AZ
zk?uF{J#og@$#SN#7)#GpO%+QN%2iJlJlm+RdUHsAbYZtM`O?6zjs5Arpb~OPdj8vA
z!R?)p{?C!^i7blDn46k@A=OuU5H)6c;zGGh-DU@+qMusKFDfaiQH65p2F+C;YE6~n
z(p;}2%%8oVhQY&K7e$!hPrA(IKE*6IUiMho@w-E5OzUbg0X0O!y(1ko+R(*x?Gvl(
zQll%X8I^2<)(uU|R@=q62z&wxQ;@btmBNcccEfa!WEG-U!D`s%>i^cPe>=Q_qR%qg
ze1>I5e22HySW|aqCnzd#Q=SYqRI7PkT5cUF>H3SVWhXnmTRD1TfG{c%U#7HLtI+a*
zxiZ1RH~c?5+|Yj4pVUQ@RsDO`YWef-w)meE-<-dIdvO5GnMS7DUqDIPU%;*YsSjrT
z^&9YcToy9-CkLZ`>{<_F{I@g5*k;dhVj;#eD&GvG_nSyED@JqJ*0e<s=0^7vsM{FR
zN1;>8jrVgQvx_nY%Wd2}<=h)45g)6O=_ikw_lr9>{OHmgMEb1Bu{Zy`=vbyIyP;eR
zjcj%wLD@dXJ9eI{t#PjXSpO&A%dOp4w#dz_9jntM@mCNbXUt)X#q~YNKHcxx$WjRH
zdHX|#f&nS}Z}^By^6gKQ_2G5(OzH7ski!d^fhgttb>G&Nv#VxXkl5i6qPF*?{jG4$
zl|2Z*7Odr~x{N42KOB+4#iH}fjUG`tD^yH4Cb#%MCH)}w5H^nvzkw|DXZ`l>w#!|y
zVl@*3tfn^7Ec<U5T#+@AiIQeGGr`3jk6KjR(Xi|K{?2tEOjI`KYH=uR&;JRwOl26_
z{}N$!1~%3>-AapE$a!n}ozX~lod3GsKktK0!~IgQT#vFu^KQl6TJ14(v8Ako!!U1p
z$2BQp^Re-X38^2n*I(Kr;GQYq-x~MerVH{Djf6+i-|6ufq+~U0HyooM9wYUrb%4ni
z_@j`GBPL^V$L_n!jlY1`9+%IJ1R!$Dly$q+z|Z~Q^M9QDzX*fY8|G`)2C?1zV)gDL
zy^0so6Wh8s`q97k$!fK5LCp@MtxGK6`78Iv@`(0_+jl|M8kPIz=@oh!87E*q((D&4
zqNkq<4I<Ua-va;?-q$(?nrEx^Rt51^+`2I(Uk*Rfv5p9bKe<geJ)*taaPIp2V9J)o
zP`4Mos~zjC<+W5)^l;Zwe3=bw$-CAi_{|T|ZLuhvby$#woe_!t&@|*ZjERutTtLCo
zwQh6Pbg1v5C1DFtU`f{rD@*Nxx1U@|DWUuctU_UiVkX~8%HF>iu;}`Kj^6)}v&U2>
zRUCITza)r2@pfZPsr@H$k(LJbhN&CJ>U^Rf+|Fn(d|SMFhW*Ox&8w^Z_ph=31rwBn
zZU{bSEFvm0U>MQZxYe1`C6|k?PZfvA92>Q>FK6rF`?Y;RR~<`Wc{h9gp}oBRx{Bys
zZ}*kd*>jsQrt(|Sx^R!mZ`LxWU1r<pZFRu&z}6jagUxW=2rHZd<D)^TYzkboG&A#l
zpAx9|jZquO!@916s;;2bF5g}b1UEwHvlDc3bVF?BGy-EMSGTVCcQ40(FWxy-)Bm!N
zxENddtLHKNjSB}1RZG`A5tmy81Y8?z-Td{&y{#SbilGvvXwN*g64aJ9nAr2=-8AEb
zS1IS7YFn6QvCxtgwZW+3^LxfSeOqza<XE#MYso^}=h#4&_+tX>)IgaM^Tmf3Al|-q
zho@rZ51C!zA6Yl$C&_VGHa=RA_%if_?Z8V1nK5{Ct`wV%$+j+uyr71*0_F>k7O<$0
zGjXfljba4XQ7jRj**oJZFyF|g97{$R3>NxRNPIkS`S`!=+W+f|*wapIp<8C-7<_vc
z_TfZ^;g+;msu0xKuZ{~`9|-rTN<x;~oS~*V+D0)2UYd}=NB-jefQ9JwYqe0?exJ7=
zgN+ON!inBx3^}gV<+1OyV_@a={$W4FfV89XVK7>nL3MCTt_Ui>`Fmoz9cUJ)%iQfv
zOUnSp@E<aOOSZS_(KxlC&(4gXtzk_X;r!U(diKGD*2Vr7Hf^SK&SYxgGN5}ZAR4jB
zz5lEt)@?4ssevLTn67g&cU_~f{k3_l<sy;=-xmAfI&=An|MRF32p-8K4EeY_NZqGE
z#y;!o@4nHz9`L@5x0FsVBzZ5)byf15c_~CRB`v1_X;Fk$DdYPzKC`{`tzh+0DXKzm
zQm@aYvkvKobT^z|E<iro<d94$_`O3lqjbx-`~(F5Vc5aRWTv#op*DVoF;QSmrIT?!
zDq``xYY89@006L_4K=6X7dn(6W<PU2Xv<lR0<5UnV0k1iO5mK3>0t>ZqE7-ShYv3=
zFDC)3Y7U4#fCFR0Kfg-<m_AGhyLpAVX&NAV>*FVP2EY^qsGh1-Z(es1V(M~W<b3#x
z%oaeA>QE5d2Ox|;1xQvzb`Ajw=jM%7{5sSh{z}Qd55SYa4Y!iOy_#7uUhcxZc!2qS
zVHK8+4guN#O)5-=O*ecO$3L*E8dD=E2azYgTWkPD8b)fmgJ%H2(T`JAB_+lgWCId1
z%Lqs#i~}FlIm65Pi>2WYM;O)u|LxAndp<&s8!<;R;C#w<=f{PLJ68^gM~%Pl&d$}M
zBdzKV=g^2ag3TMoxyk6$15WZ|HSwBN_B!#{ZLABF^sf>GdL7Fo;nokTjrGS@AmBZw
zgRN>C*Igt;6BPk3r#-v?1c^-#bZ&ifmb-9nCGa;df1Bd`FL7LVVKM?JF)MAFF~Gk{
z`Co+lWJS$k;2%*w5FS2!gY6|+bmD9JPbMNY4kOV?Y(LSCf}4PsZ8<CLq8<L2!Endq
z8x`#W1R@&({;Bf~D&ZDDoN;4gD0FcHc-dhT0%)}E8*6prkjGMxHO^LRdtiLy7Jugr
zAX0pHbQ-`6&<VEjcB3hM;Kw|Jk*!<_5x}XKJX5Z^G59>x$N|Tx4gj*uNBsrpLeI{Z
zTn5N}oAh-DF0V;|Kxm2Uz!2tL9*Ydm310LCP^UYuvjthnx*MUrF@zCTPsZoUk+()~
z8;6=RQpL2LY1!+=B%Tw3HyB>7zSlfhQmBO{6&Zy7^54$`jwN3S*D#7z9!<|^?gRYj
znH3|o)8DtsUjXDWPhb<O#8ZR*%9Tt$SFS!Wc{ohwy!2WyVkPMp)n`rd*o}*@#`_Nj
z7YsTRmkaeAoL+=An42Dy{sjbI<c_AY>c6Tvg@;@WH&E^i%~6d*m-&P&Qq8|d2MW3n
zL~;-@y^nuAec;?)PG*z}R)*Lq@fm0uZ76F}L3Y9cV22;`_TDv@C02GXstmii^b=EF
z*H2YO9Nsx-?5O4|^kG&}ncC37-a)&`xFwih+qp1!tIGKKR7r0&17?T0VlYDv>=s-j
zaj$%Q$8rPd@nPnFZI%M~febd2E|D(cvjL^qTB{#kX#J=QJ1H10mMEwtHRGSLedqki
zp=A&5G)KCgRFAEs_N1g6;X}y&jC(9*l(mt+O{Sw-_n^(WOPLiqU4KxbrCEfk!qwr?
zw7qC)hui5}AU|eo)qBeY(o#_rup{h&=_OrR%l>ZK1Q(uMt!(@$2Yq%lrOoijVE2J1
zpCHu8*7w)8a{bT!n}RJr;WypPye<<~ZnPm`6@z9vFCZW9&H)uO!)A(K?<<bSl4h%i
zZ@hf-AgQv#$DVtl_GIai9oH*|n6v<|V277W4bLC_bbn+~MjwpZJ3BYBd36@n_;ldI
zU^uaf)BPwjLS@}ejp~Ftf-&a7Ia{&;s0Q0Eh1}eVW$UsR!^IpJ>~SC>Qq?`6D3dH#
zp<1DWFf)uEEv&a#6<A`lEejS+(K#4CF7li<`e;t6AT^2&znL8m4v(*>++eV;<qvdU
zCS`bv)~Je12sD>17;ObRTFf!=5_3sKJL*7gdFm3GtErF-epc9G?C1eC%C8|P9AbAr
zPxTK>i+q`e?87TBe)!w69!U0wIF$U<C8aaCR>2t$$iyZ%8dHQNgMhV>9x@mcPyvYK
ziS}Q>%e*oA*-a^^!azjH7AH=RURgwwQ{cRdm-!Kgampo?Sqk-}wc%Cep_{gTkviKV
z)ra;Q9@tA7Zbc<6DZ+w=R7=&gb5G8+cy6g@f;fdYLC2Gg{%idUCXJjdSSOfWB(Ar-
z!Tmv>iP+j0YL9~7bPgvyTy|#G-FJnPC%m}mH=SfGQvkJ({%m11!k%?>=rtauT5z1&
zeT7Vp>NP!N$iICz%U<qWYq37g960TW(Y@HjKP?fJRg^~+O<_>KGeAfuTOO*ZAZ!RL
z!pZ>a%qvL8%G7x+ocI~Ahvg6Mj)AUUN0~~)NzLhoS}~BxO>olA{jggu1Kn`Dg>cle
zP2QuIO#OzmoucP$40}Fd-P6AKD?q==(nrkxwdereXHssz%+X}MAP=MSQFeZ6Lm3;8
z2Ru-G#dVQwAJk0$I-(NF+vI<U)>hf`QL8)(Q4k>S;;R2~co;RhGQ=u9cDNdQ?H1J(
zsy2Y;AYf~SiV;j7L}ZOvj0!coYVSSQed)bQs)QjE?|0^Y%Zz=ttW_zd*3k15-t>4+
zE3slq{{wQA6mPqn;Jt?m{L`o4Op43KQh^3<Wu>}Qm1K4cUfHU_JmCJUyEXj~l*y(9
z#Y{MiE149eQO1N#W1z@?W`Dlmqv_l~@LRHSwdp5J%gyE*`^$M+Y6`7%I8&jpYP^R+
zx>mmayno-hD7B&dRq}*<m3EmL2`E&mLQ3@}Fks-fzyTt=FKxm?jxyesvNE9@vfdwd
z*r%MCA+io}QztKGn&mr2+x&X?c1~PVYpmqL6sO;SzY)1f1GfJ#`@3;nMm0X70c5@}
z!0=$boq;ViiWn@V+Bk|#oApF0Znb)7yc0TiRjbg`6-8pjlsTqXwJRoL>{=2d$>zo5
zr18LtVX3OY!WnDrRmc~~whd0&R;F2QRkjAoaw$b63&e{c{dgu@Hst_hE~^+Bsiccm
zTGq=^%hUj`rpBWsm%dRcyvorprwB)5k`kE;V_<2LJc)3O5hH~iiQue_+Njh8*$>S<
zqKv3e(lI%d^kTpX`NnbErO%cFAxj!C5T)~Ghjt5IGE{6R;7$cR+!x0ImG)^aB&&Th
zXR|LSq3r&dfYxG1R<?5I*(f?F5ZmQfM9T-<5sdu>;EL1wNI5~5M#sCss|%n($`u#w
z3X!~Z*M0wXQ*#o3%}i*njlrwPkKP=TUfRn7pu#C=rujm&L4pR8HU`f5M`>??;A0M3
z@nG^e{9L_6ldf{_bRVng*N~^brX4cvo)k?oT~gD`nYJ}uuoxfrEp8U|T+OL^;q=Rf
z+Ne7`nzH6&Pjptu(=qE*lc#bYwhkeTQyw<AYz_AO6{44zL=nBL5h|Z&{%F^@vATjP
zR6Ap>mYQ$FTT{b9r{ZUwWfiY`Ig*lZ-x$}(9WrW&u&p)9FoiS)=k?JK%qBNB6xo1?
zF;wIjmMR5J<>|sQ2H9l7baHZX>HypT>LOn~eg;MZK0FqpigKkF6^ku-RpquOq+8Sn
zjbOVH)mW_(Se|5cZ>eT63}5lJ;#WR}u(kedQjtdd=QT6o*gwSeQsl)8(agE`9D*``
zo&dyYE8E@uicBH?6PDTLAk<;9&7_kvu_9Hatp-!dR<mqaIniWXX+s7THLH-NkWjAI
z&Ul)WzWs{ZAinUvX4LbkOw&ts?o|JS;GyU>-z^XJxgxZ2H(aI4NWkjpT4leJPO;B%
zi~dOq99nBXhG958XXBky<JS6q4rz?OpQQ_XhBVvOJ4JuxyTCvSrM5lq#mgJ<jZjKu
zpf(c?w$xoJQg#<U-Hog;(8jft(Y{fW4m%H1jC0C$=k=h1C|TCa)1*8PPW*j);QJR~
zAkzL9(05+}Sg^K2p`)S?^5?bc_p;OQV4WTt_)`|2RIH}DUNe@=Kt)tx8ir#l0)4U0
zsB{>x)cr8kwpEO0O0Qy>Z^SjCE`K)BPK*5mZ$^?{muJ+6WAte<sBJ^DR*FII##C>b
zfmY}jp3Zg6328l!aB&iY@4k?Q-P2AoGq3;ypIS4FhcYU-iWvMgCx$WrvYQe$ZffjU
zqcc3lZ^%Uy86hYa#`!uMvaGE}VVbn5yA-Nfh};FfKQ8CL=#|MCJ>T1EHwsj%nA&?L
zN|g6BG|m?AiOddX*|66BOKT1E``~Z3!6>QHPnON@hNz`1>O!*%RT6HJJgtK1y=S^4
z-s@BecbbLbpr}pWE^{A6Tg3R4o<4slV4q5>CvIU!dMNm@C@|B6a+i32pY#_ic`qwg
zu3~TfAj|1y&2Hm`=Fg^Y(@gW>8t~mTKZ#hEsr#i?W)ETd2RHFdV>6ieKUXkN{<+D5
z?%Ic~J1~w_8M%7CW$0_iNTR9?(Pc+8NjQm}+6#N4`~_EVbhl2Tr>^aH12-Rq!LrHp
zSXE6(3z8q`><VJui&98_;1V#hGDQU==Ung#)!x73Yf9Q8EpcIRGP@*MI6q$&gokS~
zeO;IUPc~?<<0%9^S+Oj;5*-IN9_-ndasI%79wowFyi>&11E+@ErJ$;;&8NBPV_sIQ
z$F1vEYuAE$iwwWZFuGdqKm;3@$Jx^HtS{u=(VqJ?X5Y(jugLBFzz^cLI_w{kMRI3D
zw=S=$7X?RUirNjTeh_eKnu7F-sKZH=;q09s%`^3yAa<xAlB>A%MAo|uID5=YEmgDa
zEC<b3^bae*gUXai8eCng?=rO%tzXX5Uxj9f7&O?C7`Wf%nHq_O92S@^@#+)wlo?#H
zzg>oBJi@hfetl^eB=jf&xpkyGe`cynRWzIb;gQ5AoS?OittuJ)ZK!YfMl@|m>M++H
z?`!RvDxvz?Ch*xTD3}Np(z`V%|Fc@b{+wewfA&gZLCYFvE-DV91~0$JDqm+)R9Us`
zEm7@zM$<<@dvKszyAC9VNM&n%V+P758{}81#$`3zXIvod2TR#t_OM=($C(k|L~xd3
zcO(yq{{qZA-haNJl6XGG(v51ue4XvTD^#HyTjd+s?9A8t(O)*3*>RJE_;3P07HsX4
zF_VCi;A(Ym5nLnYbS&XBFv-(8Ogt+K`%v6IgdZM9`E47~%Q2+-%misi#_@_Hvbic;
zBI{nv|J}@9k|BLBxb<X72`X%**8IkhEZQH7R1I;ItT=>tGpeQ1DeqsVf^ox0c*t3)
zeO--<F1SvY>|(S{-PS+Db{BK3$fP7ie$2F8r`MZ2t3;!y-=WV8OU9bBEfY@^AI-_t
zgGtJuJSy<TCT0T9+6T2m*x$Tv7u^yL(=Q+K-;XzPNYoiNldxq{bme!ax|xT(8TlH$
z3t7|=rQ*=u><TO5?U&tx`6Sv*Z7Bwl!X348OC_Na6z$7@D8K7|n?q9c3>}Fqp}Lj4
z8c@(N`QCkorY0s(csM;&)6}F!QK^)UZOGas(LN*TOWNfwlZ0ijHOE&({b!rxeTnWx
zBmdLU9Fe8zcUq7r?!hKv5G%_b;!RXUY-m%E1FF2y&$;kyU{=wHpp$#575X%mmJes4
zv_*+7uqFWe6ze#@Gq*CwVgD2Zm=41+>7pq}4LbY14aqgxe>L%&tpz}1xTpZHrf+4b
zpio)OP`doKJmC)K8WZkr#mc1E?LUxa_D%?&hTK7JqiHi$@;^FR%#Ol-RT=OXaE}xB
zRI*poQVGdZENF-bL{$v6Qv^D02GQ6GdX|Ml(6IQhG%w>ePyn=yIVqFC(7-?Hps>Da
zC)`it#X^&vO^MCdmLISBj~iUdQ!sW~jOD${(&44ed}1ha1nRh<aw2nGG;D2%U(UiY
zQD2QVS17KwS#D_Pm>Mkv>e;IGd$WO3E@Z>ii)XhVBz9s^{)R2A{smYN(0uM6{f58N
zh0*2y=Ow3!$ViTT|2Ky9j||<G55#2VN_c*8WmQ%VdLMQ4Z3ZOL&U{?q+1Wfzbx4y2
zwxfnJh3`_2o^a2oF|_&kJd$a@J4_{bQ)mNtn2UY*a9j{zl>3U~_3sD&^fZUUZGZ2j
z@DZYvu8BrtzqjcxAWmjnps0R3w~5aeJ&^6fk;uE}YaeC5^B3URdn6`uzHV$>w;7zB
z$`+Z@hlbWK5{%kou{qzT0dvKxpubc*QIj_G=<WT}G~w&i%B~5$fAsc$P0v0I*E@L3
z(mWr_F+JXAPJ0>Q71p0VV>|EmmDCb^(`u<@+NxE?5Gxhd?5v?y!gbrQJ3;?&R=zTw
z=_=(bVM<GoVfh+rB21(hYXg{?Hb&ZW;L(e)ZvSMy91UH6RMApFKLhq(TWYw3acZdt
ztfw?38rZJHD7%1e1vCp2W<l?d;B`iKBm%#GKe*^FG}!+C8-_VzSKRcD#}ICMu<NB}
zyml=GGBte)PTI4$3*Mt*mVwsYCAVyGqYJu<HP?S0G$-SZ8$XjBnngcLgvB#$ziZep
z!``1a3*%MR>&<-m^00UZ?qH6=yERUdg~i8^D72_d$b}k|(E9Qigp!(@t6CHe@>MK&
zrmUPI+&K9b7R}(F`4)80c=Z#sO>!zdJ2ehGTs0%^bSf>7jCP$fhNCErwo0Kh_d|2~
zL6wr}gmIjem~%dbLT#P!kUbRm<8Z@-Q98Eesi=HxLs<XAfPXsnq)Hl{+%g&lh>>aZ
zk$7l=cVzSp&(V$x;bMnl=kE1sMW$wsgR`dn?lTuzpOXKr^BsII6<>Ly-B%rQW)*Tk
znGlBjV69f#QI&mt_E}d`M}h9!<Trj`!H7SEA)Q-OJ)QfTRQy}E_(WU3>v5qyby+aC
zf*_noe%4ARk39W^5V7HVdbE7;IDz<|=Gc|5y@tFWLya9Y{g7qcicv`>-R$H_*MSUd
z{jzU%=>p=zJ~Gm$O;MIj&-Q-r@G1Ar)f|3JByCyh)Y<D%*_~~O+s1qy{3T7w-|!J}
zIil*-yoaUDw}#@kBsy!2Ov&KT2u=jG7FQcss~ua>4avW<hu@#>@LILm`7u%k2T$b9
z(w)_FYhxLWST9JbZBJ(`iG3u{*@4o{gO7>Vd#QvnYe6efz4%YXp1Qo$;G5_=9^Mmx
zMJTQ3u;=zm#&oBFF-c|(<N55bsN^-5WP*2bOpo=-yBN2bbD!=KEk%;6`ZCOGr?!Zy
z(ACqvd2Ao>PpAK5dN2#~o3c~;{(9)RO7Y>as^w|?)3hlT;$5QGDJaHe4q@^}Py&MV
zPK92zWkeJDwjpCSK>;kT$VUL@cr@d*OCnEH@yYlmZT6-~^YGJB*qjj0k3wF%<Z)S8
zGoj&jKJz}!_u*!}*n<Tg_Ui~(*nn)v4y@=Xqjbw{depU1$M`WsZaa7KKQ`>Y%)zO_
zAO_yAcZ@A<{XnU6#=bw{p?Cd6Km_twev;YAqKy}k=e9B*?lG^8$G+NF+UPdU0OvN0
z)PMXc-HT{jxr~oUM2i0^<zkZ{pXMt`qgStE!NK!Eo+VHwk9awEIG#gLm`)mdwzk}?
zT-P<&+`3(C$>9_c9Wv8M{szqkm*XZc>~4ELP#Ggwn|i?C3Du9Qj<RhkT(<z@2SU3M
zEPNi!r!$Ub5Tc}wShYahY^r9rQU$M@@tUdDw-DaeIsIF_OsT@L=|t|yv9Ph$XQSa>
zu>s%f9JCjF8?i80T%zwe<<@Wy$p-J!ZfF~v?$}=(Fr^;)&>TMFO^K5Jrd_2->Aqz(
z(#I0v6E(Bz*7ER6Y1XK5PKX7wQv(_=E}0w7hLhZUFVcFpF-FcYRK;U-6FD3LJlTIi
zUWkc}j0JF(5FMdx2#tfKp=)giBQo9<5!xS!tU5_^((8@i*RG_}qtcf@t~yDBMVd#D
zqgJcNBEACK1gK~u{?#`Rc$6;YUi>RV-G4bu|Ni15W=-{EHn8H0uNf(Da(X63B_NZc
z=u;*I!u0fXLPD_bBWX;ScqxV=mar1D5e8`RzdQ|jItx|u>9DE|Y#*sAC3i$Rr9|n)
zEuX$Z53{0jS+YD&td{_A7G5~Qa&+>MIQe~Z5%W9ICqw>GNK*K_r&Bw&Bp*JwH&^PN
zX};7vj{iPfsTXg?6}H2!BBKp7N)>!Nt~lUc$dpEClCjqtjek@JG16()WxJu$lKlN}
z_lnLqpC-fQUiYVB>_?t+FNxPKqIl|$5=A}lUObOE{ous;sNTf7<R>>4Wb5Qi=KbuQ
z+VrD0|GxA8@%)iAoa}Sb1IHv;!=Y9<#;AT{@<aO(`&(IAnTc3}AM|ahV5bh_&oo+P
zALa%t1}!(99qyyq$sIm(ClBjNP8~%W=iGO-6k$ai^4+K(kus$$Gnc66l4vama;GSK
zq-FX!qCqaAsI-Rqk@;(**BS=h=<9p*$^JWLkwq_?XSh(w9*q@`<)%}@UE;8zZIiTT
z#4TLU%=dIU2d+UaB1rhkU%HvSvhh>hOsr5Le=s?&cgO3)x$iL?GcY$1-XRqlLZ(@-
z+A=seIOdYc(6KC;JRZr;OFl6TN&gwiF0GxFLhd)PWIzWcWTJUu3o1>wYCUYNyfMPs
zniKXP^x!UQU=$MEkGf2uG9Q>IAANFQ7ud2`u2(=;UCLW-va~0iew>xIJY_w-CXx|_
z0+sMjX=lVwD@rz?2ZHIO!Bu^A-i*d{Eeyzs${|anY`g2S#>MXPVQ*)r2G+@FKL=t2
z-GO?~qv5nXGdi((9gqbr-*RBd|E<JqLZvoo_(#lmG0@#2eB9x4j8~kj^`Od=iI9@P
zsgjY0-*<D1W@1)z{>VrW>g@X%I2Zo~0NLqFNcs}I-S3>Rjog}oATnc3m(JqG0U~6b
z75BLEV;I66X4=ECiqu*EI37#oDIlcP!OIb!Hva&i{vQYbKg)Ugj$2gzfoH7d>;pf>
zGLKa8xThT}CSpxwnw5ey823%_Ir=9?^VjU_C*=2Aspp<0z58O~U^Mq@?7OPObHfJE
z;9tPnXlu2}U0&{K4F91@S@IpGc!p8?f){<vZp=r(<C4{@`}E)xqw%n`P-jgxJcjWC
zwI--fN;_YMA49ZUv7BLg{f=^6A@Arj?%?-h6JCnTABjpxlS=WWtM2n#)f(r`gBOBY
zX)(K<45(Eg%$wype*qJ~zW`-^7Ba;Q$HlBHA#+20=*Qgx2Vs>!=@PZ(td9lCrP_zC
zBH8ne8+Ji8ysJ1`Z*<^gJj9m)Ry<t#uohc&B(SrV<Ml~TfzABMxDq3yin0g8NU4VH
zh?h^vQSkPyhX9^VlFwHem8G*r<IhS#!B9_ON|G}wyIBt9xOL03P&Y+h_>8%E4v>e7
z{Qw$dVq52!o^^LY8DISh1Hsk)6jsEpIMZV5moIuEc~TzI+UR12U1@Z+ysN*MSzjE<
z^t8oJD=lTk;@EtCH*T?<XH&_!NUN+TU?ghEJx{CU{lw8*L^<g%`7ijLmbcheV>vry
z4|t{mQ&LRp_Pu-B4p^!%{CA(nE+YLarZs_-&fA4e{(Q34)46<dxoSyYEAdqCtTtF6
z{SOx3JS6`K3M=1k71|_tIl8lddGt0rQ@|qovzLAPk&O0Z(wxd&2X|M--IZ;;*3WxN
z9HX5F&pG-3`M-$sJ;%>D$mJ2eQLfEW?kPey%L`Xg%yk}lm#laWhRsGKn5XB;ZG^>|
zO~<(JDfr>;-fidq9DZXN*_HB5>(geC076QY*DYcI{6n3QSLU{ksd0sHfsaBdhB^!6
z3`ByI=*xaK=KcVkV7qP{t__y9G#n8MTBlsB`kMwyf|P~_L$dviMExaGIsElkJ>2ev
zyKnBVomV%<(>`A{x^$j3Yg0T1`#$*#I8C>vyic5^xh`w$BJVy9?aV?nVBgy?iJ!z;
z0K|pW9;M#g+;IK{Xq~ZOw7M990REigxf2#*NLGra8Ub56iRUZ*a15QCqj-d-?M!Yc
zR;6}r#VpeCXtsLLOkPDZo8MA)qXiR6J>3hfx0v3`e(=4x+h35P>wr`=2&#ReN-+2H
z!h;K^$fR-IG^)IL$t3#A3$`tZ<fGAKFHa6`-s@E9h?cmeX3wyox^)e7$p<WH&=;60
zfM4?aPLO1Mq>U8`24A;UiuiOVG&KtkmWBMq926^H#qxu1+<ewtRVXphAb)It;yyPl
zNg-M*S*AoYUaCMf1_%T8D@UtqE4K4KYqWnk-y9=M%{B%>aQsFa%&%E*7@#WsJc0IW
z$>vLl3gKHc7$OE-lWeooZMy*qKtT_iW|h9+R)8sX1RT)Zsp@|LL|TMheA&l!jfz_Z
zObwpR_q_(0ul77=GAmXLnfmtVH!$&mA|Hy9(bgtphiWU9M2TjTP+Wbuetw@7P0JPV
zbAAciCG|--o^)~7z+6yJ-D6ftLoOQLloTsylx-tZ8k@yh(V%T>kgiIu-B_y7an-W+
zNC5_Y`9QogJ(axlTt4FK();r{3O9$cokDsWDG7(#>@mHX%f|<XO&Wo4=5feU+Zvqm
zk@Ku)CT+5_AxVW?`#w~|zIpCr8Ly93q%n)RjCqj{YE7<i3Ny=)R(>R+RQg_0h<$oC
z#INFXXOH6t86zk$PG<59iYB9)3)h>9_HfaHmK^#xh2}P$P!d-a!Yn=jM<P$RY3ao%
zkwQ*}WWnCS(nhKoKsBluoIfP1*gRIR$KH64Ax;>1iIsI_$TuYh!sq(SK>AA~0V)Hz
zy1hMhyD5G9mM!_dPIvBUhqw*DA2q8&Pov+UVZ4ubR%2hOWKhMnN>*iwMtjWKv-WNA
z5U*(;R4BlOY?Of;ykiDm20P!|D>GuO^?i8!Wy-Kli3&b;);)Z8NVSV(Trkph5Bp+g
zD94$}K*AmJ;}3MV#(dgxgWxd=Mg}3kD?XsWd1s(JSJ9@NCjnY=+S0NWxNmmqDQtc}
z!?tI@nVW!i5Ne=%-Uo|4Dnpv`NjZOV3n$75RTc92E|*rOrL93>C0tR7Allfzl04SQ
zUU!_dZYg;4c#kiY<Kbf~U_r_Ver(!^EWP=PISkdnw<<U%-7nlT4o0^#bVJ;%97n$A
zsRUawEEZ@JA9yo%jzUk%CSrz_VDfwl@0}!mPxCwqzVNNxwx+BBvGoh7_;7gchYOle
z9DEHe<zpi8W$fJ7DG(r*-B9wDp>U~x|Fn!V>@3?m`>ni^CaO{MB(P(pKljn-yX(0G
z)>q?J-f!0lc+LH-SYI_&%GH;7hP&vI@&#I$I4M?!{*v<^N&VOgRhou0Ge4RE!E}xo
z*)Df<DD(L26~Hm~R&6+0ndWukb_YKfKcur3&bwtGyF=5w3>x+YeYBfa&5gWtjNi?Z
zR?(@KC8yWsOPKqqE=RlFEoN$fiKS-?r<QUw7MO&?u`iV8=_5VF1P;81L@~w!q@^zV
zyKV-1FQnHrO{&;8)XQZLe&U(JSq4R?T^?q<q6PIfZ;Mbvf5SR&sZp(%_hZjm12BsC
zN#1dYc&=2#<9n|+pZhP;0QWq;G#ooz2bif<pZ;^NLCu5o7mqg!)+{iAkF7M~s0ohG
zC*pPhZmu6&o<cWvi})!wH?!aSb$+)Tz>44M+AF{79_8Olx18EbG9Vl}hT3J^Hrp!a
z@5haQ-0=6q3YCQ0@efbWe^V<Phs`ckN{7l|SNXrJ>dUIN?>06%seYexR~f}Uu(-8r
zZg{Qly2NKNO<&@M=KQwA8_{0rZb)RcRHa$&>izX6(a$;jkJZe?8F)Mez)ji`TGlt*
ztv|#+0nxg>V}G`e%0b>L$>z&xgpuwtPkUEn*gL~{+mzv4r1mZv8F5He1`4wt?VhJ>
zZ?DzFOO%+9tOST_$J~qap~tr~pF)3g)^xoxo9C#7RyOJqpL|UAUq?;^fAAi4!<tX@
zm}46vv_q&U{^$e4ozT#Mx_5KgLXQA@PL<_7YgJ1ek`-_z_eY|sklkN~Pg0-ZU+^2+
zKOFfpkmWwcZgZdv?&P8fbv0d5Vh-?Zm50wbnTh79RET=7N8MaSmy?DnH0w3TpzEm2
zoa`g@Zyy?%myS-LGArY?x_(c(z%!y87grnD_xdCH!Y`CMIrR{7z4kBa_GZFguqx3a
z+=@}%OWY;rKF4K~rEz-=+u-SZ?yo@T2l!nvm<EP|5z#%kB{iM<f?ZdK9KGjw&Xo`!
z`Kmnn^srH{rS?yDM5<f5cCfob#M&&3A3s(;B}$}3<Bgi{$ZCl(fAS6+XI_sA)D84;
zN0<&Dz~d7*zfmg==GzZ_<rH>Su<ozr1e$NJ@gcDtBmO{wHm{Y<U&o=+*-P3?@Ad+Y
zGH;Y<a$(K$spiS2Tugu!fr?2ENde*-ZIe4<C;jbT^=5Li+Um8w8P7-7kL(Vh6|{wd
zYFU~bben5}Zz~72i?uf@R?OSKMy49V<dS+LK;V=-wi03=7^nDqm8=A3ZvCO2^h|}t
z&OLIulh)|pYF0hKo|ds;7=n2O9#MB#z`%CwKANgOJ?GqPC9g7FO2yGUvZ4ihBXAH|
zr`Dag=n?3_7GHn6*M*j&fVsnEJgr^Kas=14ML-Sgnt^XIOr59Q&wy>4M!-yn8+CB&
z7m^uM+5+a3rR#I=cG}Vzp7jT(RB<bK{7&wER=*zA2YFiuJz-nCe1$B8?e_Z8)o@gN
zEHD>Mbkwp`uT~Z<Th?{cdP<N@&@Pip|DxQPb@4>qE`FDc*!%VHY*@HFB`IX+k1O?*
z+9k_+=wE=wEa10CU0#3y6nw@NuwY#ab}IES7F~zOe|60`m>qJ?{b+P=91Iz$0+L5G
zD8wribd`>j0Y7o(yQ>JXrclx|^(X}1+D*1SUrrz%Ba07B2D^(Gre!n7EV&HcG}Y?{
zdTQ+#sTw>7+VZQsP1Dm^=5wP@mV~jHk5>-ahUf*f%-o7IOVFf485dIrRC!Fo-1$1)
zhoi$k=#259!5SPbB5u=Gavhw)L}20w2wEv(e}+ss0qH$wRf{7WJ7OrHIb-K_W{>(N
z4t?Jo4u@3a%Z{heO;{ycs5`F@nEe{b;)k2(+G%l)zlwJm52sSisq8P};k!Pd9KdMS
ztMXPpUC#$hcy7YSw&(2cv`=W)qt)A;I(c3dK-d0|tZng2#mqoTbsxEl)&@7JSi|XU
z&9fyCi@8`yGe@ps!h9pyQ$~Kb?K-uZ2{^x-U5|PA{Xp{^^7&TQL?^2xC*Mn?O@C2t
zJg*kgp;}0}@Y`T4<O;cLy-bFFESsI?E`h6e6fITjM3*q{6{<&kas}B?=tF83=zR`+
z$Nf0)(p&fDgM&O1C4F8T<B=54Wagv5uJR<7GyD8SLlHiq56CWQCWP$x7F<7*25FU`
z*vfP0#X^AVd71(D&)Ue@iaZ*pFc=i;8#4YX6KSujm&3XZWm>>$`7+@9#*t_uc*v_@
zkYlu(xL&zItRmbF`Jvp|uh)5jWHiMtB3{o$f|v8ZcPJ!X&WNn88&4Vj*+7XKChmE>
zj=XxgcbP+E5IyAnZ5$pv3p#K8jLRzQF7%YOD9x29pp5z*H8+a~n<CsvKK3Zv*`+G5
zb3k6?MRtz6n8D!_uNg%3nA2GH6nF;Q6mC>(9=u@&ns;&sev?YofOGD}hhB_GGdJML
zDzwYW^m~19$(~>;OP(&C%2N%LEPPO;y1bk?vOu3zd>k>=w13rzy`KCH_5KG0>4Lyp
zeaU}!VSBE45$iHiB0%v-ak~zDem<D0gCUpV(aQ(2E%ubn-jF<fK^xfkIS~I3&>pi_
zzwYt#&h@QHoL{E@=oSCtlHg8vf9^%+3wAFoH&NvqyLtMAp-f00NlD7d*%opzwQzt6
z0py3QrmWIEV!r!C;c)zoiQ->-@c^?s(0{3KZLPjCz}?4&gSzSgc}Xj}O-9;7tQlJm
z(ptJQHlQ>u(|2$HVZ*Cmnh)_`-cV0=K0*6qd=0`hH4hs`2TbqkrF;r8Y}WacN;ScL
z7ul$zizc3F?&tBJIz-Ia^^c2B2bLV$f2MhbxRWr*fOlH#rWHlrbIvQIzkN!pb-FEY
zy#L}ypY5&-9osWG%R!x_G>T?q8+%2=5*eR6J75#>!Sl>~@$|Kf-3y2v_iC`}`}g_@
z%(c5}s3IV3pI@DTo=nd&uJO^=bF1XC7vzBpsf{K~lv<$nT|MHgkM$=s_}_ZURt13(
zjm-ug!33#7il+5Ih@%y<6bN!s1JMc{@>kqV895b>ngzMF{}A<g*J|xZDcamOHlTdG
z@$3YO>~QWX|5k<0KMw`VHS8CGSDftAJFbxO&0u)YNkmdYJTc65h@sR*bLr<2Z=;uc
zomHk-D)7cHFv(ZHPtl*qEcrehY4dWg0gW@&#J)Sb2m%3RT@C!x8H#`FN#8|tJDWhc
zGp&o98gFG7{cKf>xnp$MDzbc<cXoTzxX^W<!2@K6pSsI8$Q@hptAFzp=^<KmL}e4h
z2+*-=sUH@FfnUeIZXtlM0l~u2cB1f$h7&<)i-p3AcU38ME&C~UDAT>msn{nREH}6V
z*ZLsD@!TsHwe)nZ9}1;wvJW*(fm;UDyLHaHlTe|og^EOA`eayfy>vlO1&};!(@SeZ
zDw;vKYL~ri!rHBM%4M@7;D%bqmqEGela=V&kv>-^M|#!>4!gOVcP@rm;Zt%fPA>R+
zl}-^YK0(gC6~2lag*hzIP+G@Y`=^F&bX#FkcqQG{2n*4I@sGWUuY>S%?3xYMh0=xn
zUo;fO8%8%~i%I-dR`vpGW-3DXTk63dG>c7mwg+O_-WT_$QhqDy%_@N@GJBB=hU1#U
zTQIHhAES9HPWMVG$@klq^RHSS8J#{mO56%%o^l{t7Db}7L4w`v1=ZWeUClCh@+?e!
zTFDGwK2lj<@GChzvtu+?GoiE5=;ifo@X%sC??Tt;AEjgdwJ})>j$3yMvup$r%QOs_
zQg~M)=6g=Jn{q+)WmFZougj;hRZNGAFGLp+6>cz2pH1ncz;M1(|HZ>?(1x!zan11u
zmERt6@T)%l1P;?0RqfmDqi5>*3#5JCyz;dQstUQ2B)xo1V+`EBmTV|y;<6{QMQ=;D
zWjT!OalnnxLyzH<Ss!zR)~MUAW~>j0(0r_;=F5F?a(#H?SN)imtZL~pABfga?_qI;
zGQe(#wjHldvEUNpc0F4kr}I5@zvahybOohh&7j2vC5M0*vl)Etvq9QcgtOz3_P%VM
zOvqt2!K1e={U6uT`Yo?_%lN2;3-iD9o}kt#yo`jjllZqDOS;gksafbg=Ayl-;f#Ie
z@;oy_m69!`+8#K%Tp(F(#>aFa7_!AhbW*Y8TFmnPYk?LN$W)v2t(~4gHe4s6F^e~F
zA^$j>-$8aLx-93m<1LDbe``yxeBqoG-<gWp-HgFp6z6j|@c2GUyjX|+JlT`2nV^!P
z`X~WnGs~RKUr*aK%T&AMY*Iz_TS(oNdGxsDRP&MnjZziZ>gD%xhC|bjH)d)Xi4J_c
zv?+{)y#5pzH8E)mC*(QtDL`;iE#<$>us@QsK)zLh3EQ~|RCPK+4D5_Zw|+qku99;N
ziwrNb=VXeO+{Xt$=2DbNS5LAyrj=>3@J1vFP+?rn+PkNp#)7zcs<;?<#@nW6gX054
zjsbuvnw~H;eLz!iz^l7@t>=}m%{@qdrVUyfq;8!68zPo^=0|%MF`JqXDKZVpwDrrE
zh_XuMFkEs&y!6vn{G%5*%y?0wEdKc>-u^2c3YQG4;;X%dTJ2tMkF&U63eAnOgI<Ps
z)dss(&GoR-TXh`aBQz1Jokub_Gho#u%9t_(_mnLey8Kk4*f-y~TU_f3i$EXQjh0O&
ze=>f>j&SFHasAPUsn&p~3S^WW>gg4#r{HCDrQ){pcE(0$i;=w69fD!H@S|8#{50V3
zgEu8<|5sV*kYv-Iyb;>t0D*sXV%QqlI#|>h^L?ZJRqI;7ra7uq`|Xcnz;-H)O;WRJ
z-Nn_%`u8ZmbY>Pj5E(adW*vg4c?w#@{$^*=lnAZQq$o$2b-U{?{%pYJ(gb2Q%8smH
ztX>y;P_>D$L-FZ|J}0bcxsO#2^j*-}u4^V9<|&v413~t;^pKET_gGNz_jm&R!DMo)
zzIo^M8X&?$=a+Hais)lLVRKHF_rem+7jXyr4OI~kDNY@{Ii&EbIUP9DwUcGP?fy7i
zz_V4aVrGx?9bH9p$d*h9ot({9A3dY%GXGY}2yE-xAiG5dN1&Far)y}7g(m+8HIA9A
zs1?7|e6(b5HOw+l3bDRcxQmR8An7=i4b&)`!|gOI?(^4NGB>pIYeH)CbgFt?A2r?}
z9MDT`SDq(FIQsnIhqfc^W4@e?PS{*m^w!J!yLeZD!K9Xa8J7&k(3)Q8sE{T8LL6;R
zNw-#IJwF_jt5PVJ&>*ATuK?2jVA)cHeA$|_wb8FDc5K#Njf=pqtSc~X8muo7Ymy6?
zdsZ5y<pf*rjW9s1n@>2=lNO)7FOxG@mf`g0v?-W{)l*4Qn9C-u6)|cEvMXkycIRGq
z*R2@ykOX;^ci-_)9h2S^tW%S%CiMZ?_ERGEQbsY}??hY+YoetV2q>4HU8+p8q@LRM
zX|ynf7Jf@Gol`lGtlfEew{`Q`+Z{xY#E8Gf393A~jivXQ^>=F1!(k%RWt7f?$vh+z
zDq4q-Xj^^nVxdmcg@rRy)Vc?;3r{E(#NbkSvnH`6o#tC)`QCN+4E+Us%}jnax6xmj
zJf_n9+i%2J#CYa=ybW4#*>)gOx~ZIzYJqiXZEja{^MhKcyk%&N5EK_k70x2b-4{v?
zQZGum>R9>k>lm@>$wj^f)hU~0owl#5Kr;3g&Aj&v11=`3=Fz7)+RiRQ=?q>ns%EoY
zyiKcFmx($L=QaizSQe+OV}R4V_Z@}dF~+2gvd6{w7s-LI<YJpEy;0_LxXZsnP#5g+
zA!Y7#Rtp)f`w8o%a-~3W%h~stxuhYk%}neGB3G2$pyCVkdgsGk$)v+%A6xE9xc^G)
z%df0Pf<6&?ev|B%-Ai)NMAGBpnaxskv`O&?`O;Zi_^7wku))(($v9QHxL~<#(s%t$
z-2270z6j{a#wptMu^*3C#_vbN&}U26S7}QUg=;8EP$lCaGxD9=F7WBEga~}nV8yUd
z7+=elOy9Yqt@ivcOyH^_9ettTy2nJdgdLUUajPSDvrs*#J1fLmoRl*M9oM+;nXcT(
z)<pdCWL_h4CK11-QloTg{LAhlv|gidn2wx5i7UCiP}RNj4Tcfu`~2^eLz#SJ2cos7
zQohf?oKma$8D*d1C-)UP11D+|+3@dLnW;jVpd!ka1~)tR674TDMN<lHp|J(TuXv6p
zUMo7QC7<f@ZGRa5eiXALSv~#Q|IMNgvsz<Rg(XU=TnX~-K~3J=YyAZEkSVh%;-X5q
ze9ff7Xhn^5Lb|B2p|C#8!j$uA<^-+}eY)=;MTaq`0UlSqKvOUTWrC<8f5oDh-wf`r
z>haRpBUS_&S8G~9Mm4_NYLD*cX<BIxZwhv6tDsO{Q~ppQ{dHD1pOHWo(RnFG1(ago
z`NvQ_y_taPzW|5zBbp^XU%Ej&E}eTYkRtW%Ao%4^xR5psPOasbRG4{UJyNWF`))Tg
zpmAudRN&=8e;6u+_#k&=h0z3=Vnb>aos8D`*}U)kV!u+$rU67*w?~9pauvF3$HRX=
zJ%itL23v+SFoEgPXcUH0W<;_M-y$585}DHSTW|TPXMapr<N7JiwXiNOliZf8$rp)S
zP#+XdGm7GV>m~R=;i{7pRVhHcZYbI{Q5~~<*DhzoFw^<wKh-&~&!QH9gOD78&()|8
z-W3Nxu0Q$z&}=R}j}brdC|JF@dQ62og_!~{Yx7muL`9Lp^Qp1VsV*Jg{?qFI@0m7I
z2<@=njN@@ix{Ea{3aI)`+jrxC=*XEECp?pKa7<o3Z|*oMf^`seEEozPOVBhrmpKqY
ztjc<IPuP}wr2P~N8GXM_pfuue&*&q<s1jJhgOe&(Fl{cZUWR0LDH2+;ab%kO)X8np
zOz<{sD&1*ECuFjSl*Fx0so10ApxwiAjG}UzrDtYXCDIpJv(J*o7wa`vK3Oa@{_$te
zzBeDu^BgONTkBT2iaAo)y=@3->?|{9rAxk=wf>8}YOblKrYMy=0jSX*Ux(2E;m>vT
zJtQOUGZj)WQ(j7)yU{=9WwW=@#yF(wq0Q_*C&g3K4m!G)OWNts{<7Ip4Seo>V%#W@
zj9Xd}gT3C%H<Ia05tXa$JiidH8>`>APV`c!{uyoKcDR2jsOKL07howkS$!X0V`XAx
zv}L!pGHjS=f<Stw+by7U<Q-LaTB?M`8cnC)^ItrETAEA&2Nm&cy-l0^CK+=c{#-C$
zkrHC4eEzP#x|z+x#DUUsE9Zxq?B#LYu`K-aa?<OP&0K!+>RvR3W`Ut#nK_zux>U6x
zS;!Wqxyw<jS;?cAG|DcDzn>!tXZ}d3`CFy2K_3Y@N?*JhS-5o@3>ftv6RrQCx|mCF
zUi70?;370a&z3Y>;Duul2?X_%EVvW5`0RK?d*1)yND)rGvXLxZ6N`7cwzsmtq9QZr
z8q=kMm$uG!WCHZ`_;Cm1-Qsu>ekv&+!R}b}g0}y5n<|Kvm8sVD&`>YYJBtd}A_7R_
z{G&fmA!0qQFXF;IPpLqF-Zo5&?EP>csN9b2*S`H3;;1b!c%GUB^=gtTfBAZ<zE|X0
zJYbG^s6xxBxYJCF9{n>^-Q<i)Q`u^}AFM3HRgnyrDnl_;KGl3Ho7O(9lW?=M+_{<U
zfb*T)e~jRTRpOJ1HrKAq_2{rPlh4zj3Ikmk{Qr-)w~A`3d)K~c`;@k5@d5=N++B+o
zD=s0yf)okGt+-Zj_uv#KKyV1|E-fxefa31OrEm5=-0y#n@t%C&I>;Dn&Ww?@#+ozp
zUi1Fl*LCj6RZMOjwypx{LvZBkex)bqg-vY>(gujz^D3m%K#@i-qdb7SUjFYdb!7kd
ziSVC#d|B9o=0=HvPKb~GVtv1OvLW{2A!Rl0o<1L=1qK*A1$JJV$o$1})d)I$8Mb9u
zhkM^he)afr+3Z1G`7f68N+j=HQhDd~r*ECxtl3Nxtas)kf3cQZFd9-swsoi=o7{i@
zVhO(ReR@OWb#FF;wa#@zk@aG<vSSOeaFyZjG`i(-S~Dim2B}k!7?7BD!1$n9{4Av^
zb@z~%nzRy6?hZJg$SKt^Et=I!PL%XdIMI5b-62`Dl#`wa9LyjID^#apr)o?Sx4=sk
z$ludBjgg`@t`MQX$j}CFYsr3r8VKFaP!+Mqx^<$-{$jzzHKJ-BIl5a^`#BY1eGXMy
z4(>))vs<<aDG<mka;nOWiM2f_R;YY)i+m`%p;qGji=TFN7_p>kNEA^=p%=KFns#S2
z;PU~;_w9TyRHeIZlJlQ3X7S?k(@D+H<vj`)mCw1d^$*u3yWc<1-kDA+{JJ*oaq&}L
zfvT$2%F|B&W?q63hasiGuQ4HPuY=J)eX7wL3ZGG!vfhoHmF`>LUXRKV-Pd)u?{qve
zQT@IHg-U^=h0A5qsyEGGv4O78RK=|D&0j2hpSsKBqI8NiB6LcAnm6;6bjOc10oElG
z!2pJlSj3z7-ckWwr}~D*`gC><0jUpTu-#XC#s{<FCCs9lrOJa5+!FLsyi@>BfHiJP
z0%WVmnvy`W&09;+L0QLODB6wFcm`N=Y@?I+A%sBaW%c^|$hi(u11(JNn~hd80!A`}
z%Vt6QjIzk|{H&V$uz&y4$ojw9h?OAcPsn&z2CKdXuo&aD%4X{5c%q6ln`t#^-n%A)
zL-Pfs09MvYA2u21hl{@$PfANlo3^6_wN%DZdWF%I)*5x0g05p7mU3_3t0v^5^r+E&
z^%=ezcrh$iwgLh=F49zcHw3LMzkI$`VW--kR<jD*m~SIhPo26}c5*!OP16GMh^LT=
zs?!mvmu>FTsYOFe<@S|7DMmd54c*4xIly-P8~vrB>X$s4pC5AeNf^j0saD=3*qY2|
za}qoIhGg~X)tA|^0Hl7Fye}OsS5dT3_r&-xG#@5f3hbHb<k>MYG8K7@+B1N?2ANk8
z#|f#2wOft90&P?M4-+fLYC+A@2SF&0Bs4+ChWQc?Im0@Bw?+)v2(r2dDhr#~)rBZp
z17NuDnIjx4)fiPOI+i{B$q{#NH|+y`t3L-V!LC(!t<$ShzGTFgK&|ssbW&VZ;9br`
zoFB-g1nXp)ttrXFoR&>FD%0L;g^VEdR}<;&cnaMTm7?TAGZjoTcDgnKgb9aq=hpN_
zmGTXl=aO9kLs9O(WVjIvK;v04kIvSkV)cnd6*a<IRhA-w>BGeOWhTifW2LrHlUX7<
zl4}KD%J_&Cw`_{a?Utd&^Ci0tIgdEEPOl};F~3OeQ(6z+<v*YigmLqml1@!PWU-xC
zG_0s2NfmIK8Ut$wQhR#boiaGBj2$U)S%^@7#HK-~)kR%_M8>M&$~mI?l_2uk_f&D7
zLT~tytjxf&B=qB_lIeeI@0;Uj_cbVqXTIH51vFoAH?pnVxt*@>5B195`M69~Txi0B
z7bV8Yxy;AOwkLJ^s6g1+>LtwKEwCGn0Xh3ukYs@*KhGlRtTTX-H0niVGNg8RSbE7(
zKtH)+cXM5`@hxifS$W;G-VgRrXLXt8RfKZ;!LS>^dW7RopK@~Vi4{8m1C2#G2Smgu
z)3sh2n(2{jedz%WNW4V0^M*zE?r(PX9r(e7xTgF^zupAJ+eCDBLPxW2QoD-!==AA$
zKS4L15j?!h&o{Gk-Mcy9dGcSSyZ<l8(|S4@i9kLYk2O=39jlLu?X^g=r^Pl=ed}s6
zGjXW$U93->pO6isiq$45cU_$i7C8bl!(zpzZMAfk7-j-XSOlf^dChtX?|O5^LKDxV
z4(=cdn_9IT@T#<9RZpLA*-N`4*+~r!Lq0eZ_qZ_)SaQvcWyILwMc|16-BeokgMxmD
zeCp!D?rE!CC5!rc&HZVL%R#JngjD=_v-4zrzK|zlzb%`maSBt5N{aaRzUCD1;FNb9
z4hR#UE7Bsomdnvr>&^8p2XQ|Px1TcpgOXt2z6LCML5=UkiOf`mDzrFRrs>t&g<zNV
ztCS+de+x5M0#$;M8OeXVYg0#VDf~@lehfj%!ZYU~*B-4>muQi+qZ2O?6^p>pw=@HL
znjGsq6l>W*Z+J5l1JbUgbjoKBZt+xU=vyqbFEndZ^IJKpxF)BhCZ;D)Tuwqxa1}vn
z#rB(L7O)d~khxm%OkRJf?)E?+=?+Oj0}ttfTZS&Ng?i-eTMM8e7u;be_KbXboEuZj
zMz8eYeer|mFWA9r6k%A3@jKmm$E;TM$|~Qh=ds8Yjk;iiwY&bcGn<39D)KL=q)mZR
zo7nH*`pEAtZ@oMPh-IC$QxfCVR`XKZ-qE&2w?@|s95Vj<yr`;b!gRSNm$P#AcWy1D
z|8wJn_u%|5R8?5v<G=fu1ofiT1$YC8{X@kdCjL)zIdjIVJb>ROKHTKrOFFywto}FS
zE+ZX>zk+37+9DvFVIXw10Tw8I1N|bZsKfRMDV+ZZv|WfWpI-UwWty{EGE+mV0Cc9?
z^X%&<*6o{l5r4vTE^4D&v@OF%Hc~I{vFWB<?Rreeo@iZ<x?&cI!6~t_P;<YIB_{n-
z_SCfZCN^Ju#b*aZ2_PL+*QNE44g892MEi|li3gA1a2+M^a}Z6|U0(aE?jYJ=!1!Z7
z`Hs6WXRG}=TZdq%-X?>G+UsK`i4Wd;lQ>YRJ5BUgfOvpXPQTGkbhGvrgZMhn_jynL
z!@HitYUuORFp)Vhzu)s3UR(!N$MbOFGGVaAP0vgIJ&dF(n^1-xy?h|5X=C25E(^+%
zpaxqVXB}iuF6%2))0t*sdh&&rz#wkm(wsIdZ}6up(FJ8vv+hj8A-9_8xLW;_g{r5t
z_0u<oKQCI8axzlV@MPd(ikE$}K?J7XhSLLvD^dPQNUM%6Y@2%)&O$g%IOSfB8dIw-
zf-i6mBA$9);$_YF5?(U#P1-sQq@u+Y6(KYo`4Il>{N%(iC^t8W@IpqQ^R{%|``HBn
zk|>_5^N?edc|GO6Z_*6vFrRJRYHN}lM`3;3b1zramrIW71GBiNnpx9yD=4b^e=v-n
zp4r>PDKc!_=f`N63%{&-a<j7itm^>z;oGclyF4d?oMIcN%f2u$eo3TLwL42+YzUCt
zxOZvFsAsvT63~e7rE{rKYJ7Na`?;W`-$T9M1wXLn0}oiX(RuATLvuuvN%~m!%7L1v
zaCxAFMyE@e!VTBX{-BiwhzN-E@ZiprV%RW3kY1{&XT&TSNll1moETr8S?h|W)tC-R
ze?BUyu{#Y3ouxj0=Gq9O$|k>t$=56ekI#Nwliab5I{ulzNs!;)>$G%cUbbn(MnUc6
z2>6|(Q=;+Z#gxBD+}U@C`~!3>sBeG`i+}#N6pT4mLRJUn%!%W|uaNDXq3gBwgl(v0
zn58};SEsa9mNM#OludZAV;C%wr|fgNd)esb;@j$;mz<HAAeRu9hN=3p(@CJD4@S?E
zwcG*=n6gqW3rsedqVZksamzC%3LNDJjF??@7}4{A>ih&LRw^m<#aSsLnNi-1uKH~5
zJ4b9#mOP&CDp0V%$K-vXQK#b71paBtC%0d}$gl+AT5Q%I;ny|Q;2?n?I7D&?erq)N
zNf4_6@V(3p@%5+Dt)^1e&KZqn?<}ZmvIIQp2OZ{;j}u-~r+V|w;D$z2gK;<YspH4g
z5@8wQ54Gsf<l(zzQ>GQt;HF_-MIOn>S%wB?s5?^&6oFcHm}+=(6R?RBgS|GMs(*M}
zU?L^Znifr305p`6(fbbUE4L<XY)XLcC9Jr<RUVBa7QhNT!XoVZWZ$b%*I;HVKqUu$
zzg;C<Nu{lj{3p&b<mkM7qRCloKg#vPToGk1307b~wbYIach+~SZ>S>CiD~ZpADd+n
zn)YAawQUUqA>H9CPZ?(M_>fw13^UDGGJLtXMC{y+??MC^81>$6ZmJ4!<E!3AE$(9A
zXAgGGl9@yv1zG&7HY&nx+6{@nSkF(R{$d&2TVj1Oz$|feF@(YV_qQ}ZWcG)jJ@Qs`
zz1YqCeG~L9?$gTcFy|xH(igm;FEY$@xU4r93pbCAJ{ZNZh&zKP9WTBUekhr|e7oWO
zgD+LH<QYvl3A78Hq@9~yVGTZRf#dop1lQ5Dd*@4fQq2t4@?mqdiKLD3(l@K86(_nG
zj1Mb8)_Frw#3#<gCh=2RFrH={j#Lxt5ecj}S0}XUk8tYuQMwZ+y-)B)%K<H?obuE4
z4N{UNNUXL00lv`8oX|VQk&Zn2@Jh#NdDR@_!(*XZh^vCIy0UiKD>KDol%Hm0rr9Ev
z$nj5;PH?Nh*Njmkq)z^Nvl}Yb^P!{u2O7ZXk;T`-K2yZ}N&qk}xKV8toK6*8&&Vhx
zOBa^<)X?iElVvfk`SKfME+a8$;M&i(i{3t-s#Tf|mSBqUX>XlU&y-$<?P8HbRkTHv
zS+*8T!nCA*V^6{rW6F)(_~6Kb{oCwZGKYls&%rQ8rR{3)*V>O4U9Qi%g;8IlD7Z3z
zk)Dz~t>HPzq-9cpTK{VexN563bhSNT5qddgk046`l*!PNS}{l}nIspl^Cj1hXd_ja
zuy0#gYhA;Aj21sbvr@N;Q!JOfV&cLU!os|BWs5aAm%FRLP|bhM&Mmd|9*okTw`>w)
z3`cZrJ;eQyU{?(SwIVpwQnc+iRh%QmiNNV|FZgOit0=$kaJIF+J%|d8hp<#HYpBty
zr62cMRB1er)PqnkOCJ`?d8~ygMg+<>aFFgPc!JAK^|=*!G!n5Z7Jn@4H;^#J0C|jk
zhX;fgFUM;OawSL{an#;G!+=WpVwyF2xe;Qz+HKqt0dVcGbr}OVU_&IILDyl((eVdf
z{DJ*Rq+m&+u}4pZ@SwU?`}vck(c-f_qlK@QYj9uCp_}<qJ2-xvWW|$gd0wGel&s#<
zx3CWEHezCR;Ct}WgUWWQ^ujLLFDLcIP7;FgX2MTVW}zyQ?%{*>vva}|x1B$?N63S3
zJ+SI@qO`uzo@$icMn*{M_v;uJfZ5uIll)oR_>Gd=Vl{cjHLQn(FLdF{^QB`7$eQ|;
zw8Ns}I4Me)pX>8{UljwC?{I*Ei3#Je*lL0cqfP=5c`(+4+-F0%D&r3PW!7o#!9rn>
zRKg_3hqYndL}p!^bhDy)&T48o6*Ui#IE{FLrmhV$ci#x$wNe8SmW<#-D1~sIjidkK
zKDYx!)bhkyZ=jHQ_lzonT)5Zm9S5oo(n^29Tzs#2Qm$PjVh>Q!6bdE|pHaekDDI-m
zy$YjuLppWV`Tmm2SMK;IoS~K~2Tq*)RfzO}{1)b+Dpp{2-y~kG(D_6rt`@<Pd{G0m
zO`)2&VIs6Up)RGs^OLdVo~DPrYT&ry0hN-5B~L{%Ug0;Z4c^A|O%vvJ0mJK2#q+?%
zq_RZxkiF?Hcl&SQ#s7=m$T#iTbF`?d{>Aji&o`*q<F?Vzo)k%=ew`v-BVOa9So}+1
zrr}hjvW_u55Ha?JE@8lkV<t;xT*Pc2F-^3leUKH@lhLqFRarO~MU~{Z#*Gc>>DSSO
zy|j?vTUB?m%K`wVZQbYcnbN<=cm=~aRe>qQiT@>AARq)|F-lB?z%Zf(g^;e%2n^U9
zM?l-^q&y&%GdH9AX~lyT@BE2LC9$a^00-SvnZ@_jF0*FzLN8rByUII>koQ7#4xqi#
z6T9=Qx#J(v8&F;6<Ir1-c|sIamb(4R|I4eu|L@;fFTd#ji)Hz+^VGNP(PgD`^X^;z
zZ(~1UhT}cyLq8=7;nA2BA7@R^-#F?QQc4M2ssG^o*EKSY%I`^cyj4(aan@Mec#_ku
zU@EK7cwu^itJ&^z+j9RKvD7>7@}d_oA2#^Iu@)s0(#0))aUSq5V#%4VS~wC_2)@Z1
zmWK@MIlZIf_GW>SK1hDx;#$f<xz;-EABlOLB650bTCsvgj?+xv2DU`qh+_&AKK$^1
ztYHYEZXIdWA01_sT>nQX_N1wX<GK&Co_rdRJ#0zd(fcWFAzm5P91ft4UD8<)eNHu-
zuauPnXHOoN8X29n8E-An7>^zewyH1eN?%E7b8>7=J}CRWtby@P>@p^*J-ic`sD<B}
zaBoc5y4u$Ah>t0&*MM4h*!ic|g)=Vf?5xzYrStDg#x;{(8d=tk+N^!98|Cr$yT0fv
zY14M0_y_)0xyq6m9-QXvxYjVKa)q<qXZJkCS*Hq|7_;s$&fyriR;;@=Q%<DI9a>0F
zMp2kS<H)HRKBXiS*OJfSeQ|-E@<_NLSK#;;3wfii|6F{4ts_x@L5gmm)y?&UXZIfn
z(uneuzv$1(=F%VR4WDMS4=MNEym8&#V~YAuF>9<|agd!q;gBUBveB&b@EjXXi%o^7
zrA!mrg{4|hO9?T|`;LSx8=Y-@Dk11LE{#MT{Jb%KGu(?_MpREbTKl`&+h7~<OCJ|*
ziC;HER`q1SA{WvyogWJ+cM&apI5DZ49wQJDgi4ZAvsPmzps|vCzkaPyN<$0~F=$a0
z;L%6@1rjIY$8}jm(|Ad;M*ffG(ync#)U;x?EBS1G4%foPn=;JJdD>noG3qR~mQKOk
zjNbb-D?~Ppbzlm9rEEpA^h^R!y3z;G#Fdav7k9y))^F(&R5d*xo}Rvy?}8cqu{F#|
zgFDnJc0k%<c`1(zhpsbx*p1Q<T14}sTV87B0Sw7U!{{&&NTZlW`Gh6@dnrTq%VMVn
z(<Z>8_ay&xV!RP?bk<tF&O=Rl4#-1qcqNxi+rD=dr>|_g&f({nBNUXeacfO5^A^F+
z_}aWLX>ELl;Gq?hbFZHvVqZ=~e5gJEhd#&OY>XodTu-*qx$YU4(1<|ePgb|SJQ;m{
zvNqSF&RWZjdVzO<WU}{NsD)=(`9iRvO0nI1Xp3(nD$SZ5F(#$znd^dfZc7OQ1q?tK
zp?K^33>AkXd(XbUm07TGyyQ3}GIEaEz_7N7UHjCM(9wIh#OOk_9C65te;O&p6Gb*)
zoA6^v<YHZ}>0wYo@Av!G)w^b*#trdr*31V`R2!ZD$sEGC3oo9p9f(TXj9?cKT(ff~
zq=aw?atlcTKlN$`m;kAbt-!ou5>SWYkExXAB2<e_0rsLHZ~PeQlpA*V*rWB!bsfqC
zXGfO0*13{1zt!lc-IweMt4>dEtAR&Ya{<bUvNq|per#zn5<SU3ooMDOaE(@LU2yKl
zAK6zgY7FwbQ=C!4q0YB>_B<f*;C=fdlGI4&3!Y`iok~H~Bx1bl7=D3&L@mG8mV|JB
z*9&DnupNW1fPZwuOuSk3dbm`p%bj~8A2gTBQh63Vvt1WRXduHn1v7_gIn5d{fnGR+
zx>v%qGYJN~2Z2|hW>^8N-T57y6i5AUlYpGH^CBKkSNd^@plCpH{bYrteV>}qQbbvD
z!Ords#*jUyn><2|DC0d-1e0P_m6_S=PcB$kcDSpVim`L!#x(KalktDDmXS<%yOG_>
z?vI6OwJO5HOeE-6g@uol$3^U_aUS&vPimdiY}~CNx?Bw`yQIY>IZw@rNC#lm;AD*D
zrR*eoBHFzzIU)ThkvuuI^PliwP($<SVqLgU^+~+T-28w^CyM=~_^LpusiIB3{_Mea
znf$)L1Fd>kr#IGc-S3vb?V4?^Y6-M0VR;+-bt_xM#0ap45AJ#BZij??YkD7&%4ICj
z>(A4>PeSlR-b7qD&em}5;_0m2(}YPG3N%LlM)=)8VWSctV(#rbrJ}45F?*d{TEw2h
zQs7&^yYt01D)|E=L;>Vik2TT*3Vv0*^Ttd!e?34GAEplBD?M39UBm1^hhhhEpOnB(
zVFagTy(Nh+6UTgb=wC7M%B}ozHZ!Y@OYqiuBdRb)fTqb<pm!?<z@ts8tIY_HlDm@A
zbW<FS(=g2bwEa*Bw}!Ni^*a?eBZeUnlO0nbagEX`Bc1uKWlnA%u~8Q7>aFh~%^uzM
z&u|?j0-2N(Pzd;lFUQ4;e5OoAz28lRd=G)^aPb1&DUOS><l$qnhSC+yLdAhfphp79
z)sjnKz?)xL4zK!xUas?}ZD!xI4>$yLSCr^GfuFsq?#1Uj_5RUMP#)^>a&Ei?qmA-{
zRUv%5f!30&JV!-RvyNjglh`5jt!sh629n74&nhByMufeu>gDJs9;Vgt?$Cv}l#2bO
zYcGFP9d~)`uPTUM6-Sui51=IvSB7Oxksd0drPx~`kYlO|@OmuGSzsT(DLUU3j9Sa#
z$bqiZWB!bEoLj~?=WTMHFI*P$B5zFdEK*oWPM#8QpggFU_I=}g$KbLceAhi1F7Fg3
zxx(o(VJdgZs#+LNrO~+4F!%`6xB@ouIzVM#-hO{W$@tTRy89j}k)3I269_()bJTx(
ziX9*~Gp5n(*i)fLT~~+}R*m)zAtEz$NinH)Vj8n95&&fzrD(r%B(-&4RE0Dmf1oj7
zTkoQ619rKs1q-ezrT-W$vpG!A73ErKZ}hFrv1eO@<3o#%4F#Nmhu#QBZpk%7Gm1)5
z*JUnQ*|a4G1QNl>6-r)zJ~i$xj|(MHEI~JiFSa{?ELlu=IzeSRWqdOTD)lrn2ibhx
z;t#sfFhY9Xf%tu;#?&xkH9@&VI9@&eFM)RG+gGClILJ4C-wYJPnoMkp7Yle43C0_2
z{@66U(gHAKr8jsCgr^IqpQSdG8Q+$lv#0=)oh94VUCUE18z1i$1A}V6t8t0f(?f!K
zZYq&)toqbMko_ZyfO3N$6O6*@1u8K?hD4QA(WJTzd8H2Zo2vAxs&61K<umsk*@VkY
zH9K-Y2C()}CAe-U*&qFsN)xo3W=~Vo$$j5iWoM~t9NTXfFf5`dSHip;F`%5KQKW3b
zzyY0V>P^&V5LpkkW&{kS-d=ro=VIuV<m&PlYCQS=@&N4C-Rq~BI8Gd4ivp$zl?lE!
zH0PglmJ6W?pJp0U)ffqeX^7u%qdgYYJk)eT^|x1Y<IG;$Y}fBUKcq<k2MUm7a5YX_
zthQHWsO2yjtJgUW5NS4A&DukWxj9=TNWGP&U2U%fSo+JxzZ2ngE56T3Pt~^n>>br>
z0fC0~Li=doCzH!UR@3YiCzI%Y;B2o6q1&9JVu?N@k2NnJ1lOj++$p-w0T&ZTjpPa>
zc}pd$bJuet`zU~885&9_b=<d9C5>yODTRv>460u7`2epAEn*@OmS6Zed76dSY(88)
z{v3<vj$08ZjECi*dbhEthAAR^diP+YnbF97@9T8w<)J`1H=*X&f6z|Ak#8S6X~@`}
zn0FQSP&4aLz0n3wR{ed8QYnd^zgXS(Vz7ht;<fv7jyY3iBeBu2KV^!4%E<UCx7cQz
zwceNNjI60!%`O$IID$t@+WCjF>6vG0HHm>rr|p;Cytt2#Uf7n6?&dLh(V0vFy&)7A
zQX}y6a7x2QN41Qc8E776)dUyF5ZH68eH`97f<r-*ZHpP1hRjwUio<>cko@b%Q|nWw
ziEiV*uP^}lxuPDeUU`KKx-WV7^=!gZ*a<|{LJyS+ix(oaO7i-Xz1?^CkbsmanQ5`)
zitl|xHvc((-^1_&2nOrhvU}C%#9XOvcTQ?f3U)wpIwyr__4W);IT|~0WBH@TB*($T
zTB8Z;6w1OsbSuO)dU4OECiha<rheJ+Pn~`$wvhd7u%RbBkxOP!I%~&i1P3bGJ7#K<
zkE+$|T&i{3PdddpusQm2Y?Q?+<>!}i6i+u;UK4!$&lmN-w#tYfzZyu7YB+4qlySW8
zMAz>N(%3hBNZI2dFu4dKzv7th+pGLkV~QW=0=#$2HV6maN)0`T`2V6ToLLlMtP0HK
zYlz)kT&ZjQ#k~Aw2pyesZRQ%YLXKolHO!1{N-pU;IjnP!hl?vDM<Ya=G;@kInrr~L
z-x8C4Yg6hkCZvv+gQ)35y`3Aw$WMCtMX>php3Afbo<(V!m3^c}WFz=j&g(SN6&DfK
z25y$}EimD7H)`Aj7Dbg3MvXw?-GO8{x(vlNtDh5Orl`r-Ip`J;JB&n#3>_w=w)xZl
zSYDJLZWEi?(mQO{ZLhGGzG-3{ut8e{3~D!Vl=tBd_=t3nQGas;*hc^WsL&);1t8c#
zLDSC@m9wFK362-4M-rt=7A-##{rvg%rxh+m_0i4OaD%AS7A0QeH=5@!wA>!bEFSdn
zoaXik-+z5lQU*1p1nnl2^Kt`dVS_$y%;F6)gWy^%{7msPY<Ee1DMD)=5(<Xf)03+(
zS91|zBJ!3?I>B4T7t2+6F9!U7dh>abzl1#EWpr?_NJtBu&6djtuWM*EgY$H%&$cR&
zGZ)-*fR>n1C!Mln8=}PaPim<%ObJ<uDw`${%p8-v+K%TtuX*8YDu0FxmB6L<sqc9!
zH$uR%8lu^t*hhV<Nvl-~)=ybJv@;xUK7B4E78dR<0oyjVcy35RQzMeAV!`CeI7eMv
zjnX+h8<jrzixq^|-4v+L`Hps=&-h)*i0It8-&nuGV9Bz<;&<REi-y}#VAft?e!9<+
z$c*0J^qzY&@!a~K{jY1KeF3xAqipVD13ddBzGhNoZK-o-+c~O)zwSssS3$EBPNM7h
z)81Z~Yw}ob5pMdfw(+fFp9~UIUl^p*M(FlLj@e89gd2Ohr@Jt+AcOyazwIZYlZRp-
zR&}G^T~~@NpE~MrJ1(+LP?N2B1qalbk-03d^}baaxpa4qmgs~JH;xZl&z&K(p-Pq7
zW&9~rirlUcG`=FENaqm1f}21#F_gJCf#&+KCBoi$BmLkU)FfX1JJl2#BEF6!`$j4u
zLw@+SX2MlEB$6_q$fDtCxgu=RW`680Cz~bOCTluAEenWzPUUDsCg2`Q0ti%SppI1o
zhQX&Jow0YmefeLf+<)vSQc;~M<YqS_Pg4Ew{RthLLKrKIor6)}zYZxpAjqyDqikMS
z$7lnjsHoV&!C@U4m4W#n<4wfyFZsuOiuFog<qEf?e}XuoREjBApL31v{@HOEn`y1w
zQz^g6PQ8Y)GnL+CD2HD#Fwm_C>XrSJGZ6F;51t`&duOixb8IlU(#d8sjGQr}QgNZN
z#YxPNm-iKN;`jT|eGYp8k~jN2Gp$BZkY(GySPd+#;1cVYp%YGF1OCSnPrvMc#Dkm0
zj^Q2wn}5{Xf}PpI5h<;0zT|CTw*vP(Z0<&3wHxdaw^BTwmiP1)Tb|R&H!5k!m@wtp
zPxe$KD`lw4a*qlq*ijr*G7KvfyBPZ}$p0!gNf)a3)+Qe1;XXbOs#It}Io9!y9+&M%
zdog2#6j70JyU%t8|LEfdwT}?qYia$(+WS1!@fQnIFgjWnbT3H#ARu*N{1?mL<9Lva
zaO&ADXQ<5b=8H#G49s#Nc6D?7Xp)K1Tu58l&7oC@AYOJ;FIi5apI-L3A%JgZ=>>y-
zimxR%<OD00C4b5za-+W^K07;$Hy>5)`p(V!O4wW$8a%{j-niFNy(({S(AQG72v=w+
zJJh|&x$1N4U_@UvJ8id*Y`^m~G2R4$5M*4jmaJ~IF%dqE2&ZDlgC??ZccBtgAPgc!
z=8;M0dL>Pou`jqBr-BjfsYUxX&8mLow@$7F$<4$E2wt{M=GH1P^wiW4)_+^g;wG!c
zSj4Z}2&xxYE_uJ6)-wvBlKZD0Orn@-vEQB7Uu^M8%w}83$Xbm^DzhN+D8Z@>Ke3zk
zLEEmKjy8G6URb$q)|uF`CTG|OyM0mIv1ViKoeuReHT07`HsoA-GJmu$gS*q&G)_Ua
zJUP>}hNo3Aic@!z|I>KMbgm#kRPT?4LSw$9eQdvG@w83WU#y$p!6|Plc?!KvAxERW
zH(}4aGe6sa9CSV;=jjy}Ipmhv<udEQWEJc$GO^7qR24J7MW@B77b{{k*J(}*70L!m
zRQuo~!^0pNd~zovIcp|#g;*B-Kos_~6cAUXb%j!|TRA1Igi(Uo(X-yTssA{K{%7OU
zdfKoO=K*ZmbQ6){hja>A;gM!TzsHiX&u`%Z4e23!kY;GKos?R{o`Joa>qJv-$r$fS
zWh2$EpmWxgtCdZ3DA7IDIYU?U<mZ{wz`^YW1OMe<{~KpITFG}F_=0v9b4T~#8yzZs
zEdKLOKRI`obyoS`O>VAm7Mj)e`>VK*YU1dM1k@@EN*^8y&9wY#`@lWVi{IAPjuWv`
zJJ+EyPr`2aY}new4gO`p*prsp=*aYnInK^_-ib+nCplyD%D7VBM%(W5SowKgD?^g1
zg#v=Ic#5rKyDNs0jk)oB&*#d*%eeU8Hl`5RekQ#K@n>N1TF2T^BCZ@*UpR`@&m?fQ
zkMXk{15<qP&->QL6+z2ucEPXmI^@A+AikmI&%xa%2RuC+CUOqzP!r$ec}`+!y5Okq
zlhM~41BeVccWz-xj3#@6M?fvGKzd4A{0vcB5)?1?`1&W$dr9ZvSW}*BXbAWvh(1fN
z52ZQgkX>o|dSfY0+xWZ6r;?a;P%3;+@*0@jOrP8zSVDi9?dZe~(RZG=k&yZ1I62Hh
zwwNsWuIC_sYp=i7QH*F;h*J;UQ?8VvQZf*`C{jF67PZUfDc{0AT+5tvR@{a#NnZac
zR`j+*H8RujeD5)wC#ud3Wv#FIP%M4c95Z)#-pDXJU|h6$MOlpzfX@j5HM)APv>YmG
zmkQW_W=q^_un$A;G0t)ZMri~hv|^;0K9>?HWwLyT5~L&s4sZW~H}c{?IIPrnN*imG
zuqJr-T-*5OwAIg>!ln5e7ILIaAJYxOUdkH3r;}!0LhavR^*k2Ms(GF4Ctg9Z7XIhx
z`(MtV0Hhb%ek(35=R^g4qZm#1ku9%L`>nJQfp&#FB6gpq4=kQO>Sbb_Z(ml?f%?UQ
zwKDF3UfF2U*Og2VRM&l2*CQV`T!N0FCIV%7iC!=2Qk336)C3|i`i}GIE1n7VV1gNA
z^YC9SC3LpD0<HyZD0_KT1-<B`uJA8N<0p~XO~UNJm?Xpf+*kb#%vtzlB4Mxh767_D
z4T_ORu_9Krvn6^-3<Au|N+EVBZ&Ed>-LyX*cXefpo66S__rVA|`XmRa%=5OFe7KnR
zr8D+-Aq4Cy5FT;sh9qBS-c?RrzU7?u(JNUeA(~4R)cAd6PzYg|&}b%Pd?})EvcYrn
zij;4Zf@`nN+BPs-r@$Dm{#Lslj_cU0BDQgcl6s(2ItuI4wm#>YcgVcr@uu_{*J&2t
z>A^^=|8$vYpibftfpe%Mgk4n)k2^{z%94vHOUK5>)NZUII-!GCFBbKydd_#xo?7S9
zW%%p<CZxJKkv=m0&M)Oz)NIe-%lxG<D9H@to!TM1C+N56-OjyKP22PCzaG{LCC@l3
zQJlu4O<dBm>?jOps?q*q5Vm*~O&+n=NNb_kwpOyJKh9s^noDUyRjM!xNGnORYjqlj
zgU@K~ghAWH>j>%E7aQ->{YyU`TWmSPULT7%XW<OsNi-Q(nPa~*KS0*2X<mjs_lsUK
z_~hR}#l#sIq4S-dzx?ukZdyQ6<qV#=wiFH{fvG(cYt#?*sXtn(VFE*?i69iOomujF
zqEz}+$1zxfVdy$nQF;n>A_|jgi<7O;O0zu;Cc}?X%_Ubw%Y0smcI>xfFy#8ur6;vT
zA=PRoXh?9_$6AFRwWht-C5OjZEu<zT;e<Jh>ZA*#TN~(Uz+d#+kf*#t79!@h<%*cH
z(@k19Ny>;hKw&#FI=L^dtc)7fHwgl2!pW#0-RZ_Y6_Yh2w08BWYprUHMY=niPiK1U
zo`XV%Z)Zse3&tv1E)2J6>`FY1p}^Zuib@IU=B2F^t?*{A?{9P7ENi-w@KKU<-}jos
z>j14x5NZ-0@gsU&7?DY~PN6vq>jZ2jD*sfSZA^+sQGBu45ln+TTi}UIl7#}B$~f#&
z6tg+_1r!vq)BQ*;mn~8U_WxpG7%`W!q}T<;m>HBw<{e&{sAVIWN@nO-rj*{&ilZq$
z*vnP7q`Pqb$W9$;T=vfpS!$U}cKpUC7|~mCyr@G#lt;Eny8$@-oyVx2{^rL4{c7>v
z*1Gp;sp}@kD-9V%nO`qwt#7XENqDClEAE`;8`vu+XBlQ|)&VQo?rS7^``mhabq2*<
z;V;zv$<4O6H+mnm#;P96#FnwpB|Sezu_s#hc`tda&zNHt6A?RU%*rjci^*MWB>iKG
z^$aEp&g`!Hq!l_b=8Fv@#%i@iVQ1h<#WsK|hX)ic-iA}4Gm`QN_*RxG$wBudt|kmp
z=&#nRZdP51H}%_;eW^d(ZB46>-MvDsrsfG!2#3sQ4dEp5?fEcyr$)0a(C8o4r5eN2
ze^@{m<ShwYB3;&|`kMxU5?oa?@>6#6^SAnieU<|1Z8ak;OEpIQB6hVn%}49gaNtMz
zq7a)RWGS|&neQdKb{ACR%{WY`jDS&ommQ~B_LI_%ce-}vYR`^MUx<wRe=Oai+C^%F
zduENAJ+JVWZgc7Vr%d=-m;K_cQlB3R9ZrEi`KgXMne>ilqBe33u?NL*FPU-T%$MSb
zVDZlk+_g)_wr&&fr)_nmfS4o&BL6_`T=-apn;Zkp;_6diKrrokt*a;=K}D?`x%H%(
z=y;Tdmp4k@rZjWJkOW-zOo=<@&Z36ba)8s~k#WI5{*$Ewo>gy?h@$60D-X_ih5OA6
z$FEk;rEZLF_1a9b&3F3SXx|j(TZ@O8D83}x0##peGZ(@sFC27{x9lChZS1;~II6e#
zX1vF*hwH*{qGXosjblWbSL8ey?R3*lynV_$5ApAMp8u|nQ#IbRgF*{|BX~p41#sPl
z7M`7*_TlmVitXn*onS(fMrC51vwxX)k4nNn22SI6l=}eE-16i#P*%XQcnaA*Nx^oT
z;~5x9qU)wC3}A=BqBcr;O7p9fBe924R`3hN$ngI4UCf6(d*XO+pP=NM<6WyCvJa{=
z2QGg}8?MsjOSDoLb^;oPZYpt66*Gf&+v)kKtPB$}eTZe}5zIGiKQZ{z!MttbI`sLm
zzlFyQCs7@HT2yLQi6hs$5|f7cG781NSUzK+k&g5cTCkC@c3k4b7WeXR#@3D~(_DgV
z{B7GWf!%3p&^C#YKyCQjMsvv*><eDsy#LLTXvm1SG2V4?M1gjk<~V-Qv#GZn2HHXN
zIj0c-*hNP4ZWIG}4=a$(hxsoSkCRU4qQsZn({ej38IS50YZ47tH~V0f&c(@$xa)&<
zEV`Oy%$1C&C3RkMLCzo99bdoL++MZ7=(7y+1?{^+bFb6uv8a!(h3I}t4S7A{u2VeL
zTn}LV{4W4ghKlZ`<;@=n*`WEv7x!j{UkIl~yU!j!)BftbNYuX<ZpETZzzS5ACf7RO
zuQ`8kIJ?3)VO&-}3y{Hfz8Um+joFhL!@S<i=~I4rlXHSGPab)dLU;L52?MbB1F$SR
zyD)~@Rhkpd7gb$a*ZL9v5{QPq`io_z{v`-5A{Nw(i}xgv_o+?Pf&W(4pEaLhSVFF8
z$TtJ|Jo_cs?xyqv_U2Lj^r3j~p~yL*NMjCUhIt^{wlCwY?cAJF7_}?5tr4kG3ez|T
zBO`_YdUxk~X|&ccVti^Ec<C4wyE1A`Z{{JXIO55!lOjwSo=gcM^w;5fdBr45yeoP{
ztnP(=L61qXw6kQY6qxE5UWqplsC>;iX!UkT3xRy(W|$n))L>Y9lr*}W2D)3sWJkdb
z$T1O<xzeeD@LkKV%&;rTa#I*PejU%I8@d9--e8xpHC&+MXaiuJ`u+9m!SnIQ#qpv$
z`N|iyd1>kc$m_dwb%jz>P?R4we-QD>`Ky!5Mb^B2-+*qdBI6MGADHwKDrO3%n&c`B
za2*MdhH45{8DEI$lzj}$x|tk)gw*I;GtZ~Ay%^WyN8Z#Rsy(N^ezdm8ju54#wxGF)
zG6*0}5mCSUd0w~Gm8)7}NxEmE8RnU#$b(PLJoD$;m9$KV^k2f!IBt~kW#mB~@?Y<U
z7-2l)_S4pv+vscam<9fH&b{l3u@<=R$XfAk-QGpyp=zsiPP21Y0s1m~MfJ)k`>03&
z-E1N3rcLfr_Qg>qz}xv+&)e&rw_|+kgEyQd^&EIjF~<uI?;SBQ-x|gn&IbNK9-+PP
z1xTWRVIKvR8jC>GR(UK1%D{r`*Z9hpx(BBNG3AE4b$ok2Ycn)y$}nC&;fs-Nd6Q<v
z=0B-c?RZ4?MU6cwNpQYu<(h@M!TqJ{3z{Ox#OpB?acmfaipX{@!dYCWcf_9W9Z6J`
z!=pGBaugIW{?j_i3JQu{;Za|S+^_<za`1D{kK8a75Mgi9|6;vvs=#`P{z4Z)z96$*
z>Bxl1kJH+8uBV=01@@+M=S?2@f87*8HrZD13*m;bE)dCA0yyyU)p6jo@NbtDIN2iZ
zUN2FBZ+8H6QOqd`d2=S1aj44oVq>&X)Jo~2h{~4dYEip<>t4E36VKJ&jl&qzo6kuK
z6*@T{9%alhgXVwt`f<q)im5>&4XP;&xjL|RCAXs={U+G{DleZ2+_;}i3g(71@hbPC
z9Y+I6w*-mdNJdh-K%w!r@+N0c)7x;X@Y#y@^*eeOb{Ml8<7kChC5;-E{y1$PW`0+r
z_+}h`{Ggg+8o^BjqtxpoUpPW<;9zjAxBiZmpn4SFhLvEs_FzOk{hVDS1g_<)TXX1#
zJ*lW#ZU{F<fNa%!*+HWn7tPs$8b+1XGI|<OnL=!>9A}&#H@indc-Hz=XEh5V#qPYP
zUhnGiNE){?=vsm-C$kb$H`<*H|4>PFT7u7Vtf^D%0ve-2HAFI{8$_sO104cM_0KQr
z#vcNWIxl2Lub=&KI|&TR8n4@bu%yI1b%gA*^SF~R5x|<}PstMtja7{XJLgSZ{`sZJ
z2m_)Z`3UBH&m@L}((z^|E~;{^CD$^@blAnK&#P0m=Y`EYotAZq&2K#AkMs%vUxqF}
z-1GZSLXvn=i?J<>!H&i==r<nr1N9F6CfSh<l6<X@$z3)l=9Xn9g}9F^+Es=#M!Jpw
zs9{O79cVuxPd>#YUApB^vsL@qbFflPxg=PwJh8E0D`U<rCG5I2Q0A+*8ZboDH^jU*
z>LUO$uODWhqh6|<1cVuRlG-KyY;~1laARG%xXsuG?TIp8J#&#ouXKg}#o}o6`ir&r
zENtb}ZReTS|0@Fjzq}d!f7DD)c+u~pd;0W1TeXRf!1zIi_2cMg^~-~^n^Uao-4b+a
zi+_i#dR5RC*lcqdN9ij|62r{3mq&Q4BACVh>c4#(Rs4uPPgf}J>O(N+|LlAk;)$py
zRxpkO=$E&)K|r2GDIIM$PZNIFQzJNI?FE#R))cKQTk6jM`9yCvzU5Z+>(A}}d12Z3
zlsw&QjNj9`>UGd<g0rW~;Y-9k_@b>6S?u$I5}(tb=Pf}$uBnppWvF>SPJuzmX3%Sb
zI6oPUxOMyq^I7}k%*d~L<wS^<erXe<*hu<ea5lBpPiO!5B;lmWj;=*kCa{`YK{mQ|
z@%lLQx<9NiVs2a{;4hYgR?CiiYZw6!Ab1i@irW~YozpKDzA0S@NT5~NGKvu=5)|S5
zS-l4oL!;o+{o?CR2tw8`nQ8)`GmHl;D)j{(qGlWn`l6}p=}kq^`|KQ52KCpv9Ai3A
z&5!_D-MK^4hNljx)G%FG;zk*TvWicS2d5)HRjXJ2fmv0hK6Z}JKLh2sjHV<x0f+wm
zrsT_nldjobv*aPJX0rx!;|&6k9=-7kJCTeFjh*^XDfFqZ{FjYp4Ph47w<4)-?(ff%
zT4hMA1Bx<%ZqL4^n+@2Jt!5mMU+wyugU%J}xH8=?3YQT9f#o0s$LrmSzQ%*?@>wgT
z1ok=Pm7KA`pGys+cIJQNW2j2^ude^J1*n==bq%JCsYh(745HC-xODl-@db?15w$fL
z;7GIz)wR3YZxunQC6dxAHB{_zZoCM6J?3}3Y#dwNSF<IE-B!}})Ny9r=G$B{E<r`6
zWz(Fbl+$dp%Clp_9R&!}60xzSL{!W>cRgL^#-G4fP69UGhZU)cEhT!!*N{DYy2f!U
zOO8beB)krZTy$H%hNR(v`s^U8N4;;kTlYXip4^6_4dql*m6PndFyQ{2-H}DEiQ^uL
zZa_a&qqt=Q$NkOeCi5+{;Q+eggmCy3#MNClA3`CY4PL0r5b6p4v%L$Fq=hT9PyuFU
zRQ0$Q&+G(&-)1KIymv6-cx$QYjDYCG1QO1Cu(Rik=80x+#ongo&qbod@P)8+Jc1gI
z#$}2zL?sSnJ91760mz@Rg>EitIne-~TQn0mu=8T-HTe3%51aP2sv9G0TpQT7mIBd_
z5}w`*Z@!(xml}j6;TNLbT@D!P-S2neIZ!rw(phlUE3g3dG^biPppG2r96Gh7lbL;-
zJUead?imps#UPQKQFgGF3`G&FMts+IEQiKz7r%;6!p2nf96z|;*r5kK{PDuBt6aN4
ze>s!&HMfz%zMW(!V84F9Cka@tSc*)^FiLi{t8YPjyZsok9avJzEYtC{YBa6d(bd44
zuxPJCJnIn}o_hnS{A_?p4JD2h7>gdUXe0LYAbQMEJ~a_<V5Y@$>rH9Dz{0*M#6No_
z+-fJUi#*p1o7H^XDfz6#H3K|;t-E+W>oqVvdSRsl7uE1yiimUCAHBEG^@=hPk|I4l
zJwC_5|5cs-${ELHvvFuD*!(axbSi&hDj8XyTo~OF9UdMX4Bm8N5BzQWyf$P2g^>pR
zL2ve43)P&bW))Ox<cj-GwrTcVVbN7dc;RhT(d`{I$+bKYGNmR==AaEcq=_$^tXSMx
z*3)NC1;&|p+}zTJ`;=&?MZAp2gvP)(bbp)Tid-#djO3hBQgm?jIoV&)8A~%K*rGqd
zBJo3Iv9ODlDTN6jYLHY~kmU!TOUL^q^O^Kf+Xim_f*F@-3R_3Xy+kFA#^zef52~qF
zsvcVtAKD5NzNSaVZHsa^ZOX9K8!lqDlDvewzco=iO~ZE$5Fcy6*&eeFQFgiH$Osj;
z4$v#IU63eFE^RW}L&MiL*?Lj8SVgj(On0!-49or%;_KGrD)WJx;6pH1OJa1(_`#3`
zu?}iJ&(n9zq$7I?eG}WVf7~8<u``cqX_tEOV<ewp-vwx9d?a#PGBm06FYzQr&;GAg
zqeMyHpMwT{O1{qA80*(WxP<zxp#Fe~rCMEv!!u{<7{(sKMeGvw#ALayWXjy7cd=Ki
zqO)^riKZqmwVlC(?)7&T3<C4@O{pD$mLa+q&d5%(@plGP_)7~R&n&8f`y-qB<2Z9|
zAeK_3m>rD@rTDR)`du?%WQcT=M@v-S6d*hm^6<3Wc@8}}RiFU#D`#}e(Ny7aYFV0-
zUBXB2Z{lYvihX&s<-?YWZ8F^@LtgT)-s$ajiKzU=qP_xo5nGT*y-LdLTcu>WPHn>m
z5u*Ynba@_H>oy3LM~)`Oaz80~aw(2t3r*BS(&7Tm0X0w88rWyA4&MlF>&^oqEmBJy
zN#rXn@!IhvbSaKNyU9hp7_nWhVOFZBj=Sq^kF8Fo_4?K}@3Xtd!C!~QzKePFM?IB>
zGyA>V&9M4^f0h9%xn$ZHkq8>SWms4IrnH+~s?;iFX7PN^^f>2_4EGU-?<%4UZ;O~U
zntV-i5jpmoIrl@h&Y2aZkX+Wl*Oju>t&s1|i~0dhTaD!yr#J(|lHFb&l2Ee}WY%nh
zUkkgSYRAqN?~Xp$D?*6FIWoGXDElx12J0&A9!)fbCrv|9!hU>YeF&%ey$_Xzijm|;
zE%*MnEnht5P5lswHBF0pFeqc=PSAr&xB$E*|K#S`hiFG`fBt)Vk?1DF)DNnV61tmN
zm1V|h4PT72@+pWU@Yl*{CwJzDp>t_scwR(0FOok#04@^KDc&Bo5M+Sh#(e~e+=2jQ
zGdP*3<Qwc{ZlWa9dR9%l#6SHaGZ_UhC>w|h$HGftfi1^p*mG%5>cg1dC3Kh3imi57
zo`gb@fO$2BFbQBEu4#l${aopAU;xKau^xvuD{59;KAX!o*tf!_7@4jcv23esRJ*n9
zj_Td^?{uBl6B-?-71a?zU&}SC_eM7;-{wQk**w^Y%db-uBk%9@93=GWj`L4iEJx#7
zsX(vy9pnopITYD4%4qKr#DR>fwnQWr1~hXgQ+jc5LN@-HPaQ`V_7WLH6#pQ{iB9$=
zqU$LtP5&K9ox*D|p@c3lyRF6~5SyqT$0Yl6w;t{(h>aw%R7sCx7>~8xLy_gkWnU&A
z&RM2U1)uVP)_k)-g$`TCRtJP}$E%e@GS4dCv_B_$o{5ml^`U-2HiPfsjj=TAo>=ph
z<V1F+UNizrDQ=6x5oM)aaByj*YRFLppfIacxTywTb2=BT8^hed4l#xzXem%v^s_6t
z)V7*BuKJunhpgX6!bk&n^ovNn{w428mn}bi588=5RJP!jzVM_Hu-xsr&aQQ6T0Tqh
zg+cVVbFXbxD88cFN)Udww*j@_0lNLrn0FDB;xF6IQ859><>kE09(SGHD4K21EYR8E
z`Bkb0c4*u`O5#Vk+S;a~eWsV>m5pYSbH4t;i*Xblj-70g+BO)`<POl$39Dm2ImYqd
zS3RlSCk27x_mdImB2%C^Q?qsnJ(RR>L1Q?kOa<2!*9c(G!DHJR()M|2i{So02dad>
zf~<MotwY8e{a)?pv+!GF_sZu+dymRyD^*(Hqx6c@^K`khbaOTGM@qEHhawn~v~q$q
zLNt9+dLy{m6MMPa5)Ffjc<rSqch(XP-bL@%XVd-8Fmb|u()IbZJ?R0vTw+;wS{mG(
z?K9CTkhzdvqc?^oA-`5gPw7Lyl|hVOR-7B}hk7q#_?WF8l%3z8Y%s;cqGC`@sZ^sM
z`k9hbTaSOjjB;WYuGe5R>zDyoR8XV8eNA|7<Q4p%@92MR#ed>?F>KrUs>G^inPF$)
zb|7)9%|m6I@O8>)JtAsDx=ZwSUuHk$;WWa%{cEHNqB*zUKkEQ7-tn<oaq%_3+$4s%
z<}PwkGKcXlp4Z^qH@IBfEr#Y=06j+Q|200_um6ih8T|(cw2=2yvolA$Jc}_Y&#0n4
zF?lq+d4_fOq*3KyevDSlq3w1nAj<h&bM%0Um*cM}sOCFUz%h=}TjS)>3^;j_4G5kQ
zV@qV#^s(!^7R=b=Ku*IN1_-pBVGq;FQ;i-ONoQ^&zf{=%oypyDh0P#RxaK5VsnOgj
z;&T3}A^q#+L+-chMW_Ur#L4(c103wz(^Mm!mv*aY1QHS2mwsy}wJ0#1pDmIr-O(3u
zC@f-?!N3dymSj3BrylQToOOKsbsWl4t2D;NY}9F)<sfx!93j4{Da5ENIIp3hI4Ad=
zHJxpo02@MLDWb?f%7;L6iT6a(>bW5l;HIEAWQKo`>Z6TQ4cth<6o!JH(xC_~&vJ1=
zT@GgfZQBa&&#WDNdT5V#K(!UA-Bcg$gSK{E7wOzMO8Z6q`cXWZwo|v+Q36Y@h~yL*
ziZ?sZ=Y<!6P&FqG)6)@YenSr%eJK_pv0a4}dQoEYMwFGCC3`-AP}FlnlUhgrQ{d9w
zlX9)^WS<7BFKZ8FYuxfa>mA_*FLWG%|GaW5ujMfkc7hR-u<BOFHuXDky6y@aeZPc@
z90*jnvt<MFo1e@B&#jQ7UX4qcvyNx$L$p<f9p(qhC+Vk_WHpP!sZOBtapL{*!M=zX
zwW}hv?@_`A6mp!O&32HsO^wnlfIYqTfLZ3VyspwR_`j2yE6HAtZQU(>2S-*%>DJ^7
zvw~XC1#;H#yTZ&CJ%00j^6l^>%%(>SLs@HZex~(gk;)EQn376`T~LSUw3)(s2IYaw
zMmHOjHcH{LW-6V9<r<}ObdQ8`GE)5bxMT-)UiOlrMYnuJKG!c-aI>=cKg@k~R2)yc
zCJ~b04nczx1`X~>V31*OX0QPgT!z8jA_RAL_kqFPlHjg`TX2GV2m!M5`|jO6d(VD%
z_us8Qdg^p_Rdx00_wA~h_j#U%%Mrr@?NVK3GWvX{pOB|JKm@NqMW&KhJv|VXx7ro=
z+p2drOs*~9{eNpB|F%GygLKtfgwFJF%9tbr(1(5BFL=%gA=NdaQQR~&+%jK?keHlI
z{3re{*M^O?zdyqNiUPB2d)eifR~^d)tkhiV4kjtNc4F$t<(;iQaON`ZtA|#=V>SDc
zVPRn(=-&=5_?}~;L`2jqMqHqp2Bp}0VHie-ylu=OJ3VNO+oWe$X~eB%hG{*EWi^t0
zxRo00k$LQ5^o%|Kt@dde>ntEXPPNZ(=bgo47)oLziix$pWb6+29uu55u}yu2p~`Vw
zdHCJazhwAbz>k_jVCRK|${)tw66xyP24`7?$6Y2pwk)P{N+JEB=H38v*TuKfEyGtb
ziOHE|kR3*ebS>v^Y1!#1pDXOhb(&3j!mhZ|n(Lvw;c2vwp(kfu%@%(Oo%BT_z4ON#
zU-a#}Pt%l%=D<A{%zBGu{q6(v4O&EQ<YYw?TjJ1{f3aRej2O*^MOhP;BqOWco@6d`
zN5zpFg7gaq+)v$n^ai+CBp*J7lRRIU<glu^)7b~i6RwSD{v@05@fw7RIE$3&FRFr+
zhvlaoPHOAIagL`E)0)Tz@K(J#VL3ym{X4A#NtVA@-_qgNRsN0F#*<_BmjuioF5sab
zZni(XzHlJ?i*@uF1B|4=@TOdR?yUNN)677%wz>(e-^1)GKmLn_*P@4^PRTBKeIMz>
z@fXXm??37U{m(i9F#NtBey5?-eZ64Xr~Fq4oBm`ek;TGAp(4f!>)5Qk$w=O@goN(g
z#S`+q<l8BY_$N#)`Mfbp#=-Y-?HR7ipDqvwF1c+^Mbb<lO_Pze(EeCJN?I9%dp#^Q
zzyO(_tEAjK<u-Wx$AslI8vWbrXWbXH@d$=>a*_@ah^%jCTQ|_5$(K)c6~9)s!ywq1
zln1LPYpz`NLShQdf0R|;nOk)RRT#AldK6>1t4=a{CU&L3<CmM|C%#_3uC-7K1=(G5
z_3SsQZF(mhYE?u54e*2tYR-gBn}q{?dBbe}oSK)7{W)T|^6dH$r(9|C=$3JVTHmXs
z4mi?rD%-hmPCid7-g0*CwmYq>@cEBo?tr4h&#O`5E6?P>c{MZ6WAD)TzgX-=-M-@H
ztEJ)^wJbdYZtn<HYj$VZ=0z#Ovkc1@aYwt5_%PG%hL=g1Ux_Zmpt#vInN$Trg<LUb
zda@J%v&B$xLG#j|&dE3B!5NMtif1e|X5|(I1{X6NeV$p3;ZBd&F<bc13N$-F&?rrO
z)Vz&K(Pxp1^Qk_JkBxgyAa1x$+>?JTPgs%`T&VprE@qQc=(mAu2H)Bzjt4QhaD5Wv
zO;eU^iWzxdy|f92=Jq4Q*#!^vs=G%$L=!V=oZCV&T^Cp_PU-?4(D~B`!s&hMq7?J{
zd>Kz#J#YBN?aQj3)OtqtxO)%nJSbn03BRC49$+#oZbl!4)4Tt(c$!M{_5(4Me%aTp
zQ%w0oxM;|-uE`%?|M$GLo|k8sbG-T2_lz_nMt`yRLY8!`$~zp{|Fiw9;KSu%yz=Gb
z{jcx;Yso)7$F}$dVb7?&R9kPAwmAPd16*`Hd6)BFt}AJY-njnQ@_)7+NmO<B1RFCn
zA;y0J#jqeC%}a032Z=7)ACk<EUU*b)Kd1<;r^|Zb)a1AFHRVa;FQZKmOWhsADkIT!
zGl}Hv&r2@WSmy`uzgU#MjefVFNBSIdzl|;Mhq%O3Kz_BUzP+wgiF$5MjN$!$@~gu*
ztx-0$U#A`F{qUKtMeQ%d*wUV~L{o9Q90}fJ*H>iT6;{STS}fBO3K_^~E{9fh<GlCu
z@<SD;i)j_+S&n&)7S3gIMUVH(I}4Lr`AV~;B;x`)+c?;x6{dn_sF@`0N-0oa=5>oP
z*@Ial*B_6^P}$n=KH!kfmsR38Cx;R1IrURovaOPb1?=*yn;IEs!*xn|czn|#`8jCP
z-oIE+$p<GRRRI$-KDd@~!=nli+a^wH(~aiP?_qv9cloj<swGe-z(go>ChJd4U4e9s
zqE@J?W>_87z+BDG6=oN0&U$_oPM=auU*D%jOcg!(Ddp_L$zMbR=1M>R(li!x-*-X`
zS`3D{DJN1E3#@z(s|S>CzMQtMD(1zt)!Y=NgvF-Wn{fh@a8$(t>A=T}vqnFTcvs8V
zkGkws1{(azZS8L2pHq8kBZdZC`{=+YG{r^?w><Zmh*WlN?+v2am;Fh#a)<$CjYhlz
z1c#!Ly}*7T_1sXADgRY?sMv={PcM;Vox<)?bZMgG%5we^CAu+1F5V`YhZLn1tW^l%
zJVx1pRQmfB&O8>J5IkHOlaiL`nysPesKcKY+A}-{7ZyS1J$sf|`eIwQtDgC~p^%j6
zS~Imw&$nSc6b0?lP@m5$rn$D0MiyLZ8)SQjQuTY2B$KxfNPhu=F#EA&qS8VgShoB9
zJ-p?v&rHYG$<oq0n@4@BZ)%6eu*Rs5v~phJ-1q_KROtEj7nWFEu6dc)532yo?BB3C
z<Ya*F!FdP>%?pzoeC(+(uN;gGVa$ra@>D&z+uAR9m!Aq_uK})q@SoH;$$CKN9?v+C
zG*yi0`~DDK?3?fp_@6deyN<EY*CZ_!V<8w5x>eX~#rl^;S~WImQ0a>uSN*5`)c(uD
zxoT(Li4nM;UcAF}_#V?6ZP>wO^5`R@Hb1IeQRv(w{WeSsL|Iv@uNg7se7>K}lkaAf
zwX*z1)$(*7wc%!z{m(sKM^3wAVQ#3U;v3D#fAt~%(Ss$Xhg^(zcpbg|aG&FaR0r#W
zH~%>gX-n8OxJ)4tgc=Rh?p(+p^D(yEVabmmng!p*`GE@*d_N*fvQ_u=k}_#D-JLF9
z4DCIVSfsM_ZaA*F#a}!4%_A^vpZpg~$|L%gIZI$^LY&uq^|bwXC$+ZK>`fp%_h<J|
zD3whzDc@_xC*?5Fy-16%4h=I>RZCk(KEJ<Mjt}3{GQU*>Cx*C@XCQxtwx~5YQwI=M
zE2{$+@to?@GX-Mu1h5cNSlvm>vs$D585mZYTt9hM4kOp~)}&RCx5@>U3CFi)O$ML~
zV|Meft&jJ?f<t<2%;3szcK5|r>Gb2~x6RK>8o`&Hy98FZFC)}t1>}O7AvxSz>k5wN
z1VMT*J595A0S?vtK2&eK(||2!Td*jLC@&!Ll1Zgta!t$}BnTk*a$+!elm6qL%H%Aa
zJ=uKiFw<7zvk*hH24c@oK%xD`H}q=6S}NlRTjh5_9(?)vWN^K-|DW#{q9kK=?G}gU
zI6gbg&c-~(4ZlyHhWpXoHM{>->bdy+j>~228~Nn3L9^`-<2N5Vo)59kdoV91jU|3R
zcoc*6`0|BJB}euqnYZ2Q>e@Nh^)D}Mm8MEh#s_By=HG^?hI!&837mFtPWGv*?u+Uq
zC5C(O1Xel_$(usT?Cd_#&f0tYuuLjH<xue=O#znRNd>>+X9@bY$iv3q;pbsUPmy%e
zGAdV}Nj15HBEZC|yR6McIjGNor(nM#4Q;{PlfRxo`|65TlxhO#{Mm_?r@Jt+S0C*C
zM6qTf_8@z60oyn;F{APQ{o_VQS2jo0u2%L)S=7?Ah=d|R=zeI)UgeB_WRo%`7POEv
ziB3DJ?u*2JnI1h-eUU1Ao!aB1GPs+PmA|`zA(Atxpf_$RLMfW5PO3YI;cde{pU@g-
zqk&sJXCtivBrq4FBVRBa^eHGp@5IYpz|3O%@VBqIDBilEKyj;f3g$zw9z-y1rRqE4
z&@M31iFejpSKxX~-Em`>NAtoD?6iqeYwdXI%Rk06FZ?~@VViSnpg6o!<@Gmi`9?0C
zB4!g5%0~OAklounLs!MamUGdirha+LvrN65e(!kA@})$HOj`)W1rdb&OHSO={*xEx
z9V2a3>zVbD^Zo-JxN9=p^Zlqs_-aF7+Xdd|knpHakovOHPn<5sV0lO*9BKeW!E|oL
z&uDoFVBJrESkD`Wqk0Gu_3K-0XRoBp<Dc^SZ2u4zcQmG4f`{Yr`h3<x*Eefb&N`-j
zHAC;aB@<XC!=WI7y-|@MeZy&C`Y$3+2T$U^@5~K+PBbub9_m*kN82WNP3&oyn;OAM
zU*#r@gq@l5pQ8^<+K<x=3mVR}YNFZks{L_T=a!h29tmjV$`*Bnsucsn)r-GHbiX8K
zQnck3(^#(yV(t#5_#)D6J&__yZ4#srlxp(~QOAbE%qT`h(dw{#K{H+Eu2J$!y3%QK
zN8^ard96qmr2$796j(8#ODWRiL7_r7g}nI$J9(p@BqUGSdfj?o*z_h9q&!z|@};E@
z`6WH)kqBmLfEXY%w2_}2Of97hbW)q_T?;JC7Me;GYGJY}=!8DoFCoh`Ey4l?Om#W<
zEkClJzQ2c^^v-<u#qRQv!{N^xh?Wp8dq0<;=Za?=7FMW*a4tVkc4c3*+lKDDNhI1<
zOCKe<=~QcELQ4AMCCr1D9?hI%Ck{7lcMKUbSi&AvwbC7|ZU^?}uDfA$i}VTx626FF
zSd^Jjgus30%9bf33d;aMCN!!c%~`59!=|1UUsd$Sh@F)EhWvO&=whFABg~IKqt~~;
z)!r2)J{enqc6mOHS*AsV5(2#wHaTc{Bb`P7j!946^UGoJOY(?NyvB3>^~KZebCVm`
zZJz&NSfK8uyG2K84GuPM!JSMXE*S=B7Fj2*QKrgC0SG9ZQcd_%1m~A2<7KruWqX@4
z$O=rS-;K*}GHAH9At+6rcWkl^HZX|En+=1QWAzkPV;EJ<3zMA}XCupEeA~w>!qu8V
z&4`vIJRLdhlFL2v>q}*69kOB#O;O%#jc^7-i!_LmSHjEwi}44E{7?9+$xOEg+hAWS
z4Yj&6!hf)Bg?L7~=~nt|AE6eB`pXS}Attor4VvJW`Z!siy>oSt!-*1R1z*VGUN`~c
zX5TzfJcJpq*6@To57m|O?^2?2-{vn|jo@a@*FOw*80S3wDP*fc8L2F?HdQ*oA{Q}V
zst}|vtEyO(Lm0mO$bx;;5WVhZMtx2ee@oqS0IBTEpDGg_*>Y&tsD!*<yqLT2=C%=t
z6;41Q5xlI{nsKU08~C~GR+#E@Ka!rVvBOSU(BD;Xs}Ki#o#E7B32=q(dQ&9D!_{bA
z6r=)>4%sbpnidMJAH)Qv|E3LFNff?J`ht2!g#3t6gQC^@UVUuI{{9v3QLmKMfWnh?
znU|~etU7#LE87jLsYXpm>tLFP?C63b7m(LQS1{$Pm+9(zDk27s=m*P7Or7`w-?)YA
z_&uJKBN^yT`~twgC`W1#C4<4TjW))D%nl(_W}Nn=Dfhy>;6Fv{Um;#rmBddcv7Y}P
zellp3F^}wCXb`mI;<y$JpDQ*FP%N6^OCMbI&NVK1)A(~(>t(h10B}aDp%mR;y`}&X
zfkg;?NzF?8C~5XtjQ)!kqH|yS@d^94#;R6aZcAIILs~*~Sc%SEM~`KrRx=<{I}f8D
zG-;&^r%=w{wCD1M?93RD)-lR_NL|Xb_#l3j_^@k40Y|TC$7*4g#(v*@<*Q9u>68r>
zz$SiopuEeXaU<O&EE!VrmV-c<_cLr@`Pv|cAv19~Tbye^UhTMiOOa>J;u8h<hJZcQ
zYqaX+k=ANcjz)oMqGG6)c3gq$@YE*6N}8G#v}9PUGW@IQ0d?N|aa_he6hX5MuUQNd
zJJzWxy;O!~VmLDLTOqObQ)SMk;4--)RI}y{rPV+H!A~3kowJPw@1%@3zf`I>&$-3<
z>$qkJp(kp+--ZNso2fhvDYpVujRewlnwJZwN-A`=1vbtI28If?IIC1UHL1coO(efQ
z)qYg-o+#6_f!IVrQm?+8zjs);`jrbIugIcpyQ!&av0OU>(4yy<6;XP1*YhMF9CM*K
z8A`nB{OT(q7%>pBq23To`h*s4Oz@-|_p;wZWhrElWDk&V(bJW;es79eL^w#*>gS+n
z2>CDs3Z@N(FXWy`)2HyTD8$mG$<iuOOAisMI2Xt5%gBc&+CAu^!)w@)NTqP+<}#%a
zD0Kqa&DagKROpg1Z?&@bfx|4hhvF(BiIlIHxjkSJQwT=F5iE&MX&E2QYbDWph}Xiu
ztvl@;jVRcs2^&2ZP7^)G^Gm_7m|fi-{-3jnl#lkE)Rc=S_p6zdUDMrFJPd6L8S&l5
zZ)uD|L@vGd!w7*qAi-<{L$SnR*s$=>TLef@a2JstP6bHXSJJ<$RVpG=0D@Ra-B=#+
zJ!kpo$&pkpU0|Z{t}5@__i~ZkV`QquS`SL8ZWq7b>m?`jY0e~b4{)Jw13`!`BWtrP
z)s9hQ_Z<)*FUBX54AV0TZdwfXNFp`uDOCK*zerpni1L{kX?hMpvJE((7k8~DZJj0?
z&==we->2zp1?ViAm`4TAxoT)F?F~<5rg}372i%{TBmw4XidF0Q$_$qLDg0NHwF<^^
z)%`O$wKKtL%K6g(p<gPCIu|w$EV3);kNvxko_j`CdstZVzWY_dxo#|fLVDgUrO*Oz
ztUvY8ISk&M;q|H6E(MkA?oWgM<<p=@m3R?Y!57k1FI(267AyftTZ5U)NI1-0HD4+C
zO+EJwHp5KAOQG!%q#x-@<IjmwAp_ul_%xg9|Kih#s6VXDyugfBuj-X@Pi-}zyD9&=
z(0}kNDvf8uW|sMw^wyx#eu*C`<Q@OGuF!nA-$4WndKupae2*_+>`!3X4Le$M-bR_r
zj25FAQZ9B(^;7brQ#>3vXJ?AZr`!tz!e*AMDQOHfY57gGSb!!tD(?C~#5UxKf^%6c
ze#YkKPa=zh?uL$)kC>)YBWYSSYfvS_%N~NnW8SnL7N;8GzcOc1!i0JGIwbBiHc<9F
z@v=)!%8wd8;=65B4Dh7=PVWI1)4lq|-ozO;*XWZ;d^B`()M%XPI2*OpXhWHZw~V8d
zHfZ#VcOhq@WI_lb8Dle%zHYt3FPZ*DPcFz4C#WHa=>2lCTvr=^m5n^G)-K&%AWs9e
z08;YTe^b<}wXFRL6PZOsO3>pgs^fi%r5&Ck{^k15d)<>_+Ny>Ao^s)q{34z%({=h<
zrB;M@Eds@-2x$?vXYB{a63WshJ~TB@X(FSsH0PL07*fl@mxHYkjCPlq{)Q8m&f`^W
z=~8E;g}V*~6Q6b-R=1gFYWF5PPVRNzjM5K$u74SaR4FsMc`Nj#CgdgDkLwYm=Xnf%
z;<o#@u5v(MMy&}0%UTi1%Q3tIKzkV>Tj6__rC#d+Q*_Qm`$*A*sW2#_M?6Hste#X#
z1ZENhLp%mP%^tj0J}ZtJB$$A2sXr{I!)-h^Xp^{`1y$MDYpu<~g{~FV0yRrSHbb8r
zq>M*wh|~z}P8{v}exuL%&!yP^Uq27tV>Tw=UA<npHQA{nD=Mrg+!Wl{rEX7}F~$J-
zzGGL?ON#x*hFSQFdwe=PT;a<a$H(X_S2m6_Gh3NAzH@w&gT>mEO63_I2RM7YpPQAL
z=GE%kzHYFa+_89la<utOkO%J*e0hHLJS?QqaAxWH>;}tlNbP+q<(~%tbad$c;?|RS
zZPrG@A3_2$%VIY&_8s4O!!r1ivu8bs;be`ke^_R-5n}M5HLvD5iZX-?dk(jXGNd0R
zJ-jcU`8_YiV|UYW;tqcQnyzh|_N%by?;Gox)0pZHk;<Djmp>b}J%g!<)VlGdtE-KN
zn@;vgQ2^{qQ=^M!2smr9#m0$C2=|a3u51axO%#FK_r`UVXr35|#^fn4YSAe}-oQOV
zjd2M3Bi^)|FQef;FRF?j57-YTo0(MbF|`U$x{W3}>6u>ES4!N;kd&j9LUB8S%o_a2
z&1=4hB*deB3^T}!SZ~f+5hO!u&a%G<?c-PQQAsu&mOubh(tkX8nG;GnuI(45CILh}
z5rt{_TJMXp6{2?b%C*xyM=YOqVL(9DudO;J!|nur4QG*zx{#Ja&=Y2+LG(*g2a;Pi
zTng*?lVxN%^}NYWok1HF;_C8Uilvh3qVy^psfZnz-V2(+;DEB4#+SS+6nI1rHdj3e
zw(QV{)E+uQh8!W6NNLhqf@@)sL-!(Wn+r*2-=^bIyz;4%-Exu7>~k4IV7EpF#zXg!
zyW~{Fyi-u)HpNzhJsX54LbiR@Plt$%*VbrEGgK^>UV@D`nK{t`h7wfz^z7ytMy+))
zyeDs4J*I}`WvP#|;irj%Sx+GtmkT1SkzHu1$s**rdPrlWma6wY>W2<ffAzX<*?Fe+
z`F<P%naRA<um-hNM>zwuO6nGLTS#X%=dIn%#*~~dXMXr4GfQGOI2~Mce=!rmHHjQ^
zH6z!XWY#=Ye0~f{32=jrs74*=A1#(6>6w@EAnd(fP6g21@xgM<gr);^ljLl<%4lK>
zzUL<{__Ys$s2cD$C^A7`9mhxzdx0p=1!&PSRHhbKT4Sdiw!R_JT{3E)$}836%rYEu
zLmNwh(GkF7<R`Gc=9I+nI9*}?x@7L*iWOQYuq_9R?DVD=`DG=WzTP02_X|<cw#S(u
zu4_o_$#hbZ8xcy%)-KU5_(6TqJxtcAV$peXgln2bZ$NcKwuBafxed!SmpCGm>TXzg
zgogF-3T5VNCK+>sDAc?~<p60Ho@ei0Gk^)CJ_>KNN|8!OQwdMAZw8odWTrk*DK5y&
zwAru#uOKz{gO0;QJcOTYHm8NGv_>?fN7{?blX{yigW^%EO@dT$JQ?~T^NZ$aZr`H!
zh3!F3`#l7@h(QN5ujJXe|AUs2C?k*d@9cj5$Zo|KesknCi8P@qPQPYV3%f;iET0Zu
zCD>&Q9!+yO;PLp5sP`z5A22qR%}STF@wL93sVlY>rKd-7UqCa<4QAz>*D*Yf3{g`A
zKqpcI7FY#;Bee!KRXwU-B(mCJ!F9Iit6IaZkdODHOKJvD4GaZ#DrF+cd`%<mX=;>W
z$ikM=?!=UmcGMAvkL$-?e;hmiK{c$M#pfcCdd{!R&v7D)eSJ<X>wP;jgAX}~GxNSW
zH7Yi+z%<zV5CcNbNSX5d+zV3F;%=CQ@&wF2E!u5bz=+P9lT_XhR&Y*NEKt7KP_0@n
zbD2KSN4Fs6pFD!5L7fOEEN4n77GR`mM$CftbPFmNAb>;T1Zh$-@l)LsZ32N$Rr`#t
zK9ud@qN=DQIy~fb_6L$vr(<6XMM;KXce5LNO8G$bp%}<nD17t;U?JXr^~WY`I0ZHA
zWf?f2_kvRKi)eMJ_t8!xo|wx@fho=FuQikOZru#VnT;4;wh`#nM$cFj9@DH&(8R`=
zk!$a-q&!)rm6AmgeFdbZlfh&hFMA%dAeK3ifS27jyMy;U%OB@{Sx!}RJ*^f<RhK|9
z&1?N=`brs3Su0zivG3ZK4}Do5W<Fzz=5({QOJc3b6G7W6_yaxObWuL><V|O<diuPa
z+o7I+pJKn!p4K=4c0U>5ZMv-1jF>ut+2>3f2k_iLY~^_t@`YN;J!|K}wbGfA--Qi;
zISE`N#h!YcgKJngQ@@T9zH@00w>M3DRm`KRuWB;L!?i&1WK_$JYqZpF+;25k9D%k#
zqoE7kV|1ss?P7&*c5}1Vn813}@&f3HmgslJU&S6#HaXNh<$SK6eeiYx1m~jwzw%5z
z&R?bBSk)}h!{KTVGovVYQ=i&sI%LzR7EiDZOEgLeIc7yKcz_CoV2eCCo0gU4I$E`n
z=_-#$gZS?tGUZ;*_LB~TgHs*P8xPenQ)e`vEnvS+UIRdC3s<!*WG$qH>anqt<Z%{0
zRyO!HNKKjXxX~jS_8oou!`z3<>#XZ3f4H!ioV{s%*5c7<{-5Fv*TtY2WZQ%QvOV8z
zV)JbiA}uZ^Y4Q&m8Nj0|r&_IPYVLZ$9PpULmlzSFdGxf*e$s_W0<V&+q@K-81p+&b
zZ*Q{`WT%g%+|~xZ!-~8)k3QNj+eVSUv6=#;iVdlO_#4;>EI#1yM6h;UPf|#z@_t?_
z6)ba8qy~DFi3<CdG5fi1m3IuOHXJI~YoxL+$g(sPZRFw?;J>xx`2KlbGR0RJwceTn
z550cnVK^<G*cy6(+XD#92O+2TNa+$Kyfk=mjg~?A71~EUvn-3lT8xFZg0{5YSvl^-
zOp?S6o(G#PTkA6K_>+nyDJ|s^RgHN~BWE=+%n{Ne_G`nz%89~F5Q&HEwAQD6B@i0l
z`A}t8JkLyA0x`o#-BWWJh<n`XbG8ubp!9GOt|lDufh#Tk?X!YKV%C-|Q+gNYwh`Hx
zCRP;$S+0=vDUQ%=Xubi#lynaxe#T^JY{^B7AM*?FdP4sz&eO=Pi%QQnvBO^TWyk)=
zJ_(Wdj#}mw9prfZ_=E8DXtN`o3cq|beTgvc6~AywxHq;)_LES;{T$)IBT?m?ish)y
zGPfbn&3+W%QBZMcYi3X~Yv`n`mJ#)*m;>1NTfV*w>1M}#YCB0Y(yHx~k6Qv;wUxLk
zwj+cng3do65;itiv=E9uTtH$zmEa-|2*PjsQvPvP?czkSL1wwpY*tuitS_9DY>Gaq
zRbr&q<8Gf?Vy$^SgbTUQ!uzsnQBX(#?ej5d%A1KdmQ>~AaFmj}<L0UrB2i*s&>df+
zYAtP^zpQA-8Mq(zr;R?cCz;o0vgLMR3seNLY(j7LM=Ax?3d<HN3i3Z4x=KAL9txTc
ze?(BV^Q!(9iUAP9yaw72e?HfCf?ZuI%d43`o-*%4N;@tzLcEBf^sLrE&k1vpqcZUJ
zyT>2<DT2*39gn`(o~(zI#-aYOPvy+lB~EAAn*aJqdB`qIgM)sD2ss!?*c3xuPlZGb
zsf!h+=_7DiiJ$Wge+}TPu@ZWo@3ek75#SN^!l=;?G&nuFLTt*WJkSWZEG49!+6%UA
ze^@VD1TJkV<jrwas_T8{WdC}cSuWg79vrZL)_`^J9nLA5q{hSj*ouh%{$4_I7W&I-
znz$TMhJGoityu&qGL0NlZ`vo#h~#12<BX$r{+OxCGf4|NdBoIE1mRDb8XV!`m{^~+
zZIV>(94omo)8Tq@*CR3^p`8#c<z7!JBmh<29Yw*D8E6S|H`oNJ2?}{WN{rP`nUAWu
zn*qNr3?5c{Y43?8A(ib1FMs(#9SpkrT?IUr2JLj&YH<l<9z!Jx<2Bzi^-+REjrCbW
zV8}!06EN`K&ek79wo<(mhj!hb4!0iIHUB8gb<;f23iq1i<#HfE0EF$+J#?!dk-V(m
zPa{yq<_vX$Cwj|aWSo!@Eury@<?H=kcBwf&w#~(dVy1SIuL4dflU<_?JL8?>YP9v7
z%Oc=Y7viCymsQLIgS|z=W59&aCkEW*ZA!%1IQlpzOyVJ7y?fW26-1pig5{+klR@r6
zSvsXylhL6m3&s%vZk^_8W&uei5f<aCZg1)SO{iism2E03L!nf}za7qhe-dA1d&rDk
z!zDE%gLq7aPqp{?++y-o)j%0GrNu9m#)A!R_HLR6wuNxic!5Epe(jQET<t4i;lA>3
zr$`$?>b$6ZE2cN?vjQWFE0+H8g0J;M?T*clSBR9l%SEbA!@jjS){^r`uw4eHe6pWL
zNREy-tuf~Un-B|VC`J%xk3_fVkAR*NP)($vK+%l{6n0LkZ?U~AT$BAH#;(Hos6^Kq
znlY<ClV|{TnC8z~=o35CbE8Z(h-~p~rT_~Rd%afQ9pSbIHE#~v6vVy~ufH#GP;33P
zVhEh0t+l<ZFu!THGkIK2k{SmejP+U*f1fCuZbRX(O1YN1?8KQ;2q*?N0X!~rc_sZR
z44}q^(ybN1WF)XPAg`oj-f}lRISRDhLHm8-E*B=@d|Eg5Xcr^=nfuQFWY^=4gVbi!
z<ZyDN@SWv$-h@5PV^Iq2SS3`=#S51TZ=~^;0}8fTovwCf)z8Wy$Woto_9IJVUk`|!
zbIr|o$#2sZQ*oQi8=XXX+>a<>O%m?ztPtah5tP1rot|b1N(U}<p;?$|j9Xg84lJJx
zOP0vi45e`s?ssOXs{=3VtVkx?9*kx4VZFa;{Sj#whny{f+K{*iTWE2BV8<A41+n}U
zWI8NptxK@UZtt+Oq{VtBTGw2OJ)|zip7MH5&?jy<&_dm6l#+pHOAhDfb2A%o7Bub8
zA-GhR7v9k3lp4~qHSaphriRi50HV6=>YQ1N$e-=1D4V~0MI1yHT(Qr`{b^`aLH?{h
zyPU_E*s=`AS}fnXmTfSQV7@F_xfum$jf_{+tF2g&96!Rd&8f?+8!!@qJ0W>Qmb?q}
zC*o8hnSC4l?j<45DT+K_(6YQYQd`o3rrxykIFw!;qy@dwSZGL5P6D5rs}`<p;-Q=t
zd0=)+O}=o)OT$Dl*m{78RE5P0UVWZUjJvf6$2s(6>1uHe9FbsaW{@p_0-{A01-Nu6
zZ1b7joHr{55lOM^)rKXbS^J2FF_k0(1V&Hb$A51a|E}agUn(JwENcl_eysrc?r~Qt
zDcIUus-(t1TY#lwa6ri|dEQ{bt-fHqjuWIwO&zoKiAJc-T9EYA^ldaBC7m@%uDCuW
zzO}kyS2)2T#D!VT$KXmF88*ud005|cXh|x}7NX4-%6$w&P%egHQSex22n53gG@xk*
zPtM@3wD|f+P!Qhwd5+{uiq<hYEt^wq?J&U3fJ>YLFKQ=P_}qvz{n^O86LViA+XuUI
z{JJR&BQ2}(WY2=fl+NpXU{}+I@{Mc{`q{)5=7p_h;H}W%jAgfYw*Xf3O}IVq@=*AA
zsrx;o>N2<4ep+72KmAKpJA*}SUAeSEv6^snP`no8C&>EH99XBos>h#l0iWU!rngW@
z_?@TWQ#f5rWPdcK_+T%f-&cm-V`p179&K^Bgw0(l;Su}C`%Sw0;l{aonkUbcLwF>(
zcq;kTpzkw{WX+`JOYE+&Zi|tOb>hW%OU|V6QmqU`(Redo+KLvfEToo|HT;j|9u(&a
z)Rmo`nHj8>U>+zP@1nV_&%#1*97?FBTAg)~ln_kHty^P^D}eWo<OohO=3GhslK*|V
z@hJXFwP+GE1y?b<;Kd$i79&dLdNN}`zQL^cFXudFBTdp-bf#zW>*${^x}7ao&)hxy
zT8G|PKj6+F>&L2bRA+oR@LlR7`9E*N!~RySA2{Z!As=xu!3~H7tR&!E2v&pqLNXQ>
z9{Z5RH>}3-f7N?nm+`N+hdZ4o8S62o5vE*sm+i-BHLfA)%q}&n#bG@28ny<f;m;t-
z45^z3iSVS~*GB#UvIl%`9y^FKJ9c!wDM73GO3d$KjbmEE2F6QE5|!hZ#fn!FTG%KA
zSG0W6+OL;5Qw<Jo;|YWQHa8X__|QAL+JOLvjZ`F1GjNLS{*q=kA}!J=jJLlo7$Z$P
z?T(+2TtB{WaJ8rLP>~Y7JKpf|rZ*lq0#%u@%v>JA-<SRBx+uF@`{X<`;bW+2&oN&z
z#R<yqQ<c81vwi)RZk@3OA`{BXGsOZfL2x33bpsE_hML-!8e%i);@@xkAS6kLdrN1#
zC1hUc?o(0Fu1MRb$|c<nZi%q*!iAsiK>$Cc$$HilVK1YWtTfIbwr;&MMc;XRkL}+a
z65vKsF6Y3mMx?b2i(wE0=);WCXwTuExA|{PGfA`x>>H>0#ZJ>jGk3?;#M|1@$JX*j
zi}8t-!(quQrOOVv>#qdy_-E<(Z`uPrzgtZl4?Wz83b{;oZ<q>;`imvICeEM&a>|I-
zD1@xF01s&Dc?~5ZgpXZGHInIMe=M|$@HCt)?%OMHmn>pmIZafuSM=E(XjODZzapcq
zDVxi<NC*2hN&poT*C7D60SUrWHoS<#&H%4CkrlV@(oT(Hn}zx5L|hS5$1*{8Xox4`
zg9$e0pmeihSG^XG4`CtsPnX>TNk<rob-4J{`nQsPi{?{X;_NsZ3zu6&Z{w^s+gJRl
zHn&-$K6hRN`9}Z6`XPzmjU}^gg)Z8gTD*A9Yae^GBK3*}L1g1QZ{y+-=$8Xr2pd+J
zu6}!<RjOaZ5=du%lI$yAF{)-Y`BN}>|NWEE^vNGa{^=^cc9Kzl2EF!e2f96p7PsN9
zOWK<8o(svUrsN3x1w_!^QDIVlTp&p-wQT165T#sAe*8-}_7gJvwES7FSGQW#zJ`e)
zUFf-15EvY&ZscuC4aBul!bmXHS(EoyaK|;j^s`nTQYuMrG`RO^;PYTC$ucr`PZ{^l
z3jAh{!Fb3^7SFxsv+s}x2JEk$eoC+CmOnO(uaNf%J-AH4<Nf8u?0W>ku3`7NKD_?O
z6K_%T0}G42%##r-B(z%M7z>6)sfLAJS%-xM=6HPe=3mB$hsj}SVEtoiArC*Y#UtW=
zV8?&<!d!{d{U4_HKdo1d@t*=>bH(A0X>AyA(!R6n!@c4U{vS*;CpT8SKT*&(Yl_1>
z^v%`27y%zoL+h8m`WuRy3-NYd5K}=vq^|h%{&XiMfDT>q$<$oZkT^3xK&BtH?-c?K
zvOYp#Y(txdA#EQ;IFRQ?JB;CKj~|pK&e|S|O%CZqkL~mL`L_KKJy>**|8v-Jq~Pf8
zPnl}hIP@1QOS12tM2#AOO1wJeT>g}UMKg|A86=c2@F;q8=rSNoC~qZ@3h+`hMT9GZ
zS46{`{X}QkxATy6A_<pTs+4%s%yEOUJd*_-!&j*SIi!#$??n7}eql9)TWq8IhVfdM
z(9cWcqj7)LzVj=p7#41)iCQQVaa&dtzMd8-RF8vW{z6g=L+4&-=SbapfaG{*mRn^$
z`2X;IYHv~RX!M6&aKOO;4d@v#D*3TEP&{ER^fS@6ZmZf^!@Mme;Tj*5%%-YnF}n=r
zE*~FK6GNeylv#}Y8rr`H{>6G^fNtW62-<$&c0+F(atMB<MHjJbnk<phxEKuAUT<sq
znMohu$0HLcw3oIDRx4C3pIoC+{^i&5C%|?X-h03`)Ie=?-4#jI(;n>UWTwSBQvt8o
zE)i5O(#}<|o6T+73)Nx=X_66hp2i(4d?}_0CrLJT!;!i#1geI9jcg}f<z<}KWvo+?
zU397&Sw?!M+UOw4Be&kV!SPr{0;dV)FpQAZ!I_r+FDo*6xJ+i|o$SZ_754k>?UHP)
zwQa+ETA!@m23Rx6mwH^{*d1j_-QP-m+Q@eJl$E7M@0Ts=p64r3v)U|zJOoy8YiocK
zexKa|uifq%joW+_be)f9)M&+>=f9#E%$>2Y;-rb9THKPY!;)1Rj*PLge2IdsG%4HE
zcgZYmvg!9424sF*lr@q%?sdB3i7sPgJoW9-<9`y+MvYLyC8M`ZMUxi1xNx#uqyY=(
z73ob12Tr2YgqV#eN#*f`V*^LGU5Y<_m)p%b8$CfSSUw*O2<6^g*vGELW8tG`YyrrH
zR82D18^^4mI?7PDr`uZ=^?uuy#_x=8GTW>j>n{<gvIX&$g+}ngRm(eDhP?8gtz%i{
zK08`zVN4v@AC~JA{~=eCf>vZj)ajt{>Yq^EzgRabj{b1)U##VX;J;Xen9N1j^5M`?
ze;a}Sg#(w)uE@gA*<<JR0-X{3V>Wt$yWvz0H`G?HG}CV67MsLg9?LCF5q;$Z$=6Jc
zGV(7je?;zWR5D`}fi<m~9lst@tXd-XirnR4`sDX1+N7r#o(`7pk>)K|n)+nCuXBfD
z59LlrO~I*NK%?-b2d9EEDDP&U=j=+M9g9~NA4}@kPS)SOVVf2lc>7ZX!&V+S@2qpM
z2qP&6CmlB=ic`^-IV@k@S#CDYs$_*hLBcKc61nWf%21i2b<U}H8Lp?K-V+38_OhY1
zXA(&yJqfmkfzFjduP6ZFP>@^U3;<$~5XoB*?|hVzsE_*vjhW<irzXIQ3TKRkfRVhj
zioqxxioVk~rP|w;EXrW@0Tos7eu0-YnFjOln}d~@b@y?T_6epvASCF{y%D+>>6RIL
zY}TiK_^-*r|8^iGP%Bl<o7#q!2|pJ~gGvXpsh6U+`aWUc8hI=3Dg$vDwnCP)>YYwI
z8FJRV08QfiP!Oo}cM7SNz!X_+{43fg23lJ9pwQHiF_-}PGh&15YEx-vI8*(o^JYj9
zGcXuCDx8vqNzRljo0KK3_f={As81pdT6GT@60E(nKorvp$IX$<G%OL-mkTp-z1*h4
ziMwGf^rOiWZ+9mp;m^2cV6P|7JOt<cknUf#{A!)tUkFNjIz_u_hz^-g+MS~27M0c6
zXSH02^mwDQvsh0_@wLHoa3g~<)(5+1&W3qeQ?Gu|Vxqr6r}Cz#|Ib!Q8LIzKm9!Pu
zY@<=uuw|O)s0pwuQkAF$mH?5HNRru#gAnL5qSvI_fbm~OCOfp!MDousZtrCL6UASt
zBb7$AZAoJ($<YzFbvEJy)~NfJE~HcV!8G{5mHr5Lyf*{*i4b*&UF1V<u@h7FI<YF_
zxtNvdxozpN?Pl?MC=uXw#OD9@{Qs3DA~9}yDS~3ibuyx8k#BUEG|Ol~xwmTtD&|iy
z7lLF9Fm&@z^E(6S0R~zQ81b$FWe`oh!0aE47<i<qw?Qf>CKCxMW@e?9@RQJyI=EP}
zkcR~FaPkmxd!AB|K&f$QGfKi5YB+mnq^PacqhHaGw>?vx38fKIkPReg70xEvn|alI
z=$uE#rFK@ww;(7=AL+K}k=f4nlVoESC9%c3o-cE&sG6!$3&|f$#h%TVN)%?QlPRp(
zQ*9(oB#GH6AK&+M_qu3%tBBM<I;Ik`D!W-`i>hP=6AF@?n969m<KqMga8bM<7!8W_
zp`e?Fz*3UZ6EnZWZf;v_{}4Bv=}n3xlwxzy83l&ojG30{axn8uIBU;8-OB!FQu<HA
z4i(6kjk;)#dmc(DY3ue|zzDO+B=mV&P3jyehiHgfFGPVrRAd5lF*!ZaBr)%0F~iek
zVe0(S2Mi_IuPzVj)|DP=M#)2$zPP0ImSt_i4U%0lN@nudFQGM<Sm{~mJN*~_B9&VN
zEEnHzvL`U8i~HUE7pLaO7#-W!*{c{t&~JMfMg$i_HLc`xQzLVWcsEF+`18{pLnQ{f
z5mk)|I#<K^s87dnA8s3YFx)?D9}2vX2+T%?g|iKXgrAP2eLCT5)pQz{mtHUB!Yu52
zoX~8Oa*^eWTA}8_#Q8kr@vncR#bv~X%xTxe3pg2ehs?QXFb{RR3hN3p%*dP3mFo(R
zC^G9qE5OU-+rgCN5M9}j{z&<n7TSdi`5K-?17t`81cs=|?W;5t!ZSz&G`q~cyxJng
z)oQK}yHulKP9)&cWWr6<D0$7m>RM{6OlsfuSt|j=A}GK#pjHI2J$Ed<Nb^0^@|Qrq
zi7QrC%uJ6!$-Bgn*VTzt=K+K%1|v-8NubVhukybo1?^6e<buCjB%Yc9&0pHJ%nRp9
z8_X>047jO`=~h=;;hsv98Fs0FyKd#=g(jgCbF+!f?s*STa0Iq6=Ks+*d1)DlZq{s(
z5wPr}L7GbV<>?cGo@>y1Q3?}7P>_*OSPiP1T%wTLQWFKws3Ohh;;xRR*>3ich6|5C
z2Q=<&Br-+vq7D3H6}*@h>-5E&$drbVdGITe7iDhpz?lZ?;_;`E*-Kaq`QE13!NKHB
zm&0=HC|cZbULQzFz0c%kzQIMau;RvG(ujz(Y_kZyVsSeu{=BLjVp_>rN?e_()?N~+
zc&Jge!rVjV#CdHQV9>37Vv#LQsT>H-DjVrc2&Tk;9<Rw(dcw{kpdVYrwk9K)5|d6T
z_1xZBhKpo%NK~4QiJ-c;Xv;jO`6o`b<u(&o&Q%_s5wmbTW(M7xk?!v{vVhZ-B-t*E
zB!)is2-PTpIPTrm9Pjx_XqiAb=5<vG_X3~Trh$Kk3P5pPuUM3|@#F@g&c=V%)io(C
zX;L4x2cC*hMnBXgTU$7yOHKV!y;Zo64hv+K>VE9qr8mxqX4_hq|AKjL+)4-+l3f&Z
zMqXA~y&wJjI*zO+zC7qqu|vxef9Uz8syTca`&gl6FdT^tw`ec-$+=~@8PjRwOXnc6
z!G8C(lGdOvC!^s#h5bpBjmQY&4hUQpjgN$uI_73Hjt>o5zE+)*NXpbA&O=zfs|0pk
z9)Kn4mc)T&>$qWWL-}jSmF*67+}PH*XelKI``<>ts1DkVP>S`DR^Xb8a^{Nri!~xF
z&dGw;n3<LoA5b1~Sd`zH99XemroSNXLS4Qo4?!F?z}S`bYfhhZc88gVtuPp01^!V8
zl^%p3+moFqD$>kt(_gI*d1o^F5!&rWjM=izIU5FTl*WXdt+y2oG!t<8RIIhh2X3$R
zo7!T-3C9$Y%{Qq5hXC(bQA*8I)9>bJaoaCAo@2?r)+B2Jl6qB{l=ynqTdtycAka_F
zYx_tRaV|-mN<I5<&R4jWw5BWuTR}VDeHNk>+M{%k`z(gj^%=<*`_cSOkE~{j7d_m?
ze%zs)F#8M0%!eD@L%g8|MJ@XTAe+()WTrj$uDvAnmAObLzDQx!xmrPFBXR!SpmC&7
zBrGNwIpO$3q-=_VW@iv$<Ehj7nn|b^=*!V!?5rir-R54yEf%x1t4S9b(pL?on+D$N
zT@JFATMdoqa5_({Az773_DEyQ<e)%jmbLDP*?K543h1J>ThdBFL_BKaJiyc~&}(|a
z^eaPpX2FWe@ve_De=~uJ3O6DD2t*xy=pS-XNe~rErXV2~)LXdp>qnc?jIdbqXi*-s
z3sfwgWNin)0!Ldw2S@;MfU>1fhCM&$PcASI^M}ZVS!~j+!gQoL5`tTfJN(ECM|NgH
zrC2GRZdsNNp=_72qO220O~T3UPo-4|eR<Bsp<6UISu{~{nOF)MRvONyLq>XkOAE&3
zMV1+c9Krzgnv}c6Kfk3Ur&_^z81TC#7wbMz#Ok}&EE>~hEY53l9m|bG{}BK9uP?;^
zM+ZXj`=+l4{)~@r?;ZDJSR<}usbr3r0qn}iW|w}F%EqBBJh!u>|Iot%2%xVWc~M7n
z(#;<DdGmrJ9(z04Ft5+q_K%D#u^yrb6iY9&Uo+t^(<G@bwxS8VK4^KLuX|@}z#M+X
zEee3yPv8WN9|tn}%sW9@U(&&Jnr()zD+N6QN0-T#1>$UH?!LDoh<QYIM`yhta6fdO
z;3If-sN=o*QSEE*=1Lph;5QqyV;&{Lt=HjX*UPN}t=9=-V&7x_?c+QVl`pe1M$<&1
zykK=)4fEwfL4cXV5yhg}U8n-BS1C`eX8!bsO0k*^DYeCgC_u18ZbW2HDFcD4^n%`u
z^S4o$Wl6K+H>Dz!Xp{_rNf|o|Pj*G*&^2|}5YJft+k?%Gtuo$IdYkJeq5zUC6Kl6*
zz0Qo+7Bd+LV8fD4?_oEpKGz&9JD#Tn0T!fE+hYAOAz=;WWLDM!f<PpgWDNuwJ_~6g
zp?b*p_~XAF!vFQ8kox>$-BWQfR0AG?9ll)O;^<y&1-?R$ltlSBfM<M=UUs1Vr&y+w
z2G4AiFtg-6iwdgGGM{ez?1ku}y3%>%Ucg_hH}_K=gDYd6f^!>`KOf=k))G!sCj{Yq
zhjJ?4mLb8C554pCx`;U}^<u5o-yE_tsg1n48HYAR+(B&TwTnH8l%5Y37qqz1_<xlY
zREdNj--L3IFuJP;J~Z$<NkPu>_@(S>Ompuqm%15_96bGfEgABE9nybSXtF$z1~BsQ
zy#6Q@Y7YEolY>l>r?s-~f1!N-yNdtk`3HN?17$1#-#Y*aJF}Fzs&%~jCl;culM1PX
z;UYk*#T7OVzvOhe&=uwQwm5V|6VYOq);Py2`iEMUv}Rycou67$#F>72RV-s~0x4<B
zOIj*}c7*Lp8VnPOwz7TJltzI;Tky5PVd*GUln-P3h_j145b|2;Ald4juz2Q(U+W8o
z>Z1Mb8Red)wRpD+%}HY}kg2?E#%SH1A=y&6h-)dqp7(J_`-YY)^Pj@&&pv#+c=G<E
z!oEiN%g@(tc|zGv)hOZubc-eOYWcUUidS*%@w6<xrFfK1CTRY2K^uDHN^j`<n!Xa7
z>3btA+r?kA-uTIK6>bzfHti0?p(j|K?li0DX(`@27W`%y;!7&&8pwsw5w5wKS}cER
z?B``z^FV7WfEhiM`aq<Zu;yK68=?65YtD}gOTThWfI#2ZF^?SVegTU8!Y~nV|L^pG
zlUZ>5Te#n%*8v}?Qu700=!bBbXX!W(_wwxDVDqvj7E3?7y<dqDtap|lO3U+bxrTWP
zy|C|iXZFbG&YN-f&)pkvB@x@r94S%x9hUgQP^}CD)maPGvpOEEoCk07jU}>sAA^r!
z*RXAfeDTAF7_zz|tc}Xh)ED`+iyK&aS}IMKSe*aw`}|*5U|a8O|Na=3+~{|6I=q?F
zlzQcxL&0ZH>+9jRWUTBaBNg&+)zs_cPf}s>p4G|&(g8&4&8eVY?QPu<qQ$R~V*SH?
zwffBy{aenBQ*CfuC9J_tt|cFcg{$+WDUr#BjR|P;k3%R0u>2#&h)IGh|0GO?|4Hq9
z$DC%UyNvX#;867xKlf&~pj%^R={g~)s>zl`K5YT#7C2#~2hkm@{R#0^%S{alv_iBT
zwrvjJ#;b?h4QR)Id%@)ZS6j}gpo7E2V5Akp9^RadB)5<OzeBC&5{jU7wd0my809h*
zRhztXqqkY4M)~5OAXFa{P?}mzlPB#EwlcfY`e`$<;TN2j{~~EdPvqgYf2t!$+iA^+
zI;MF~&%$b|P%2-Up;TDE%5Ko?6edmnyXdOeJ}0`st^II)T-tVSuBVR3#*$s*Rl2=`
z>26tPnSD`@_fS8B0pQ#w1*(#?_+_8_-7h*(HSS;er_sWL2-(w0b+}g%AJP?y1bkeu
zBC5LT?@Z^N@5&IZFL755x13nh;DKu`7AqEN>_WVKFj^hBuFp#3xCN$ccvM1`+`3?6
zJbf9kut!U8u&|T0OW}uy(|t`A&PR)vYMi(8h^Lgy9_Fq03I4`bt6c-dud<${7S&j_
za-+8FrgeUtVW4Q3WYfL|k+;na(I@!)vy@<@$Sgq-GV^5}7H{}{MyG%#t{`rJOAu7b
z1>%wuO0|_c;`?z&an@wDz;2S{lB2qNFG%uBrz}_dv~_|mq@)Fj0>Zd-7AZtB4Xz@-
z%w6K+J=ohcx0+!RhnYKzuzCy9p`h|WXJ%AISn@)6)sg-+k2o>2;7$GZ(oos0=lh+%
zSnp4FyKjx9^C&c5=NLSfU;T@vaoYD6>)_GAhn^v9%t~&N{KLsI24G*_2RS$SaVR%L
zNVGB^-oI_(w2&O`De9neyn7KISF2R4%awH3s_I{C{K<Ny$q7_#J99a{3=nE|MP1iF
zt1nd%mUk=-78Vv;Ss!0rN4x;*Equx+vZkllJ#)a`2+Ihf3zY=NW~LQS?Loy$oD+Vg
zO+O`9Tc-vjGEK$ie9x$}v7Y2NmGSDhilNlvTQeoVmg~A07@WcD7XInoBVjBw#nhV`
zK*&1$oLeu}Vr#qf#OCz9qKm$#nlpyi;m}T0IPI=XptveshU&`GIaqqhw3N3!;`s|t
zKQ-P~6uKcaxt`z#2$;@xpN<<s2teog3)FIEjATWT+KB1d7rGJCD~OKld#0?>YnL}P
z@FdG;iOTLceHZFK<r$*-q{po(p1D+Sa=F?Fhy3a`@Z}6~zNDrUU#O|M6q#BpwqWg(
zoLgY%b!i&Hy_!~ZuUCdFme$78pd31ct8x{R(_CPd1-hN#^Sw39GD1{63iZq^+QNfc
z8u12bC0QkOVaWievX()qLjHJ2NK<O+uKcj;D-uJ@dcAqDBm_MI)nx?&N1z_;=L<Vw
z1xw_BGjrZxSiy`Y$(BMAQdOB={#60va6)%L(0TVzzkq<YTAEHg3k!&v=pdsA%{hbv
zASJ8ejpz=?=3p}$;?`3G7bt-kXekby4aO`MN~=`_SjCd{oTy1?8lYxr`v8ReG_>Z-
zaEw2?f$w#<*>z-E@PXaEzvXoFC0zLH&FW&;p+h6z{nYW?`TdXa<NGgrin>CY_e%yC
zcCLo?cQ@Kti*vqOk_s4;c(6~*=fhiSkcgP#-8ap5dMli;fRB3zu&|}BZ&M>j8hY8Q
zMIny(J`|bIrXyFTTrciy!m!=d6bQ8J!*vpiRBZpWE0|`AKU9!uua~55s)F%R8T>3O
zC0UpS8}ELQ<X~e~aBUMW^^BeO^TdzFnMj5I59Z!7tc|X18?Co!ix)2r!L>+%76=j`
zKu7`vX>kbBqQ!f+xNC4J5+otGQ=rAYNN{(jP^1*;{pEh%<9(j}?!Etf-?4w}`7tw@
zV`a_CT-REeb*<|>&m{z-Nx4&LeKTZIe+UndxYQz`#Y(r``~9SbcHDL;jY5g}RjgF^
z!y)iFQYlp{)p}Zj=pgME^$+)IY6jQ6AXTo%n}wEg5YjI-yzy&YK6w~1>ZjW7>M>P-
z-AYUyuy?~xRmn0B=*!k&Vfx)SJ__kV2(E->PUx<T95UShe!B@*Z<0SRQLi;N0SeU*
zhX1va9AY)!`?xoznq5qsv%v<DVz*VS4}hknXiV6j;QQ*>`8NoWw)IU!WTojH8zyNS
zuxbWYQZe}Si2G^0^_o&FnWsu1MZ96lcZ){sruWBd9=;W93UiBNr?%31*(h$C#Jgc9
z!vv#L&<PDqQnwE3%f9y1-hx|&%yOmA>m@)E#YaAPD5OBJszykdtPffw${m(c)UO3r
zRd)ostqSmYxv)f8>cPIlcWWtNuqUh0Ecco{HaV33>VhH_2(dF{qvvy@U3}A_=E8dx
zSP{a%aLIgf(R!V2GTYwXrN|heq3%-u#3ne`UJOE;NRS9tRkgCbSnTQW&FyqZqlkga
z$0BR@2|+K1^mU*^dNE;(h)9OYQ0VE=5H`ZW`5#ptfK_oxrDh=43m|g&^PCHX!YXcz
zSHvWtlY4U#qr2n6l*VTqpU1AUm1^y3CA=)s4tM@i{%Mxl@8Nl`K2lPBIHgb>jJIJV
z=6wG$4i)>HzT#P7jp1;xFgeE<HTR$5c$%gtt@+|mYQrPz+1OYHYFhsIu<c5EGR&6M
zWPWyW#FmhHO4FfoKUxC;so~>S8Uuk6tO$R$f{+42CL))Lm6er^s#?k8H7FcY+lwP`
z#SB3Sn=t?&WLMls*cAZ;PqaqefZ7{McQvD?@Rj3^=aX`O-+JcmXMi;~Gx)it+q`Om
zPq)ZPqy@dpE&|n+?@boHtrU}#{m}yOiiTKBS5KJuQ}2O-PQ5EF+21!GOtl_tyBer1
zwzqaqznxWh|DbP;Owe8l5zc^;lZsnW4Ql9zxMD+j#ZbI)mRh5BkW>pe7osld!a)*I
zT*dil6#j5yxX448va{6QRwHqmn+7cSoP6Z9L|>ThS>{xL5hc@?{dILRH}fIw&YZA>
zpICYit~Xed$^0w#=!6hIdhIO1xeaxnwzP-4EyPcni;uadoXf|#a9yoo>K$#qB|A(v
z26?75odu__SXSs+VXFcBi{ry30(*S3LZLjfGDG=)-vVNOB;XD9q!{y?sRC>c;D4e$
zzgNP9Hnbhbr-}B@{u+B|Rk?Q<ZUs!cZZ~3IoP?=zhQ3KL*{^`l30qQ9H(Z=7Upi5?
z4+>e*@;9U;Y6TSQ*+BVF&I!z7Xxl0NQ9<AIYOyJTxIpoAc?cB^ZC?5fSuf7yRAOv8
zwSoHb`jU*Ni7H;g<08pd%;Ner)r0Yhvdq)Bf{G6i&YEzEd7N_&INAWM{uO)GY+H2X
zDfm}B8lIu+NB~}uJg2;C!)-ryM!#gJ?Bww3;xGg31eR1Lnq0d7J(Tt!loBpsGcX8r
zuWM=#)Gtzmv#Wi=X@tqAD?*swe+#W!^s5v*`xceFnL#VES_Z$l2~DL5f9{cEdd%Rj
zwf&2RcDh&_4NcU+0<*MI8z8etB@R5>K*;VK$V$ySUmtnVwSN%IE4>H4Wc${Tr%6Ef
zh1UcJ7uPm@_@dSCdnOiu6TIUJv$0y!9Qx|$CG5uoOpbjZ$FO0jXLuzeCld`2chBSs
z+jXVzC7NN7xW<<{sk_MW9-$X6TqN0Rj}}W^IpWJLQ!zOun6#%mM}98CPbvf1U|PT~
zxyj8={@sm1Zhq~ug%p1q56?z@zcoa-T9i)li>sS^p5YWLolfeJ5BQx_LIx<(><c{I
zjsfkd4)?Br{pL|@uL@R5<w2um<R1wPb5_<n_}_;9Xmit<0ZlO`g)#n#QOj;uS&2?t
z1qrWs_Ld*-KiSFV%0s{D@$BlmQH>|k^k=MT4FUb^MwzjKPPQ#DNBoT-(x?>4fe6E^
zXsyTQ6g9*?CMW12NXnupb4>GxOGk&Quw!ia6^)Xy=ykU42=wigDF9JZC?i24a@=^^
zg&15=K;Zp$_pt^cAn*X29C=|5dEB9X<pxK4QYPL&^3R?)CF7exq)eHbr%nK&kJmx<
z!{e6;Dpw>NN`7|m19LGBW=iH!8aY*VY??b1I;qkGaC!0d4^`3;G%|0dfu&gW^E?zV
zAE6aXgQ--<j_>N|bJ>?iLK%YDHag;qSVBO?ZeY3mQyzO7grkgIErYd^67{~qESPgY
zX|PnLK~B!dwkMRVa%Ji29qM57U3+RuOu6HuBTHGTS_@9z(y)q!>vB~YSj_q!-<H}|
z=GWVM(Th@%6I+6C=M0y*+~lDxT-{z$*HpSar)rT&exV-z$*hKoO|2SdJUx95C2R2|
zGm=>b6*>lpD(G|-MX8N%&VhX=U=*cUh8hA}fV@0Ob_^yD9dj=oWw<AhC_-yZxhE8#
zTj~zxCt{vuw@Je~rR`h(9Ck(F$a2Bb1-Ug`nuy6f(37C}5)dTGDhbWUA;~*PpllsH
zrNOynMkDz%H&knJKm(cnIu#YI379Kk=lT*87J;cFyxF*_sLW!D0&^Fnx`y?^a}C?i
ziS(_!vQ)M`eEbbRYc4Dv8x`GA`4q|Ve#QOqp<Hv%`&XAV#6Uzs1tnIqcpQ#obrL`k
z@uf+o2Kstd$Zt|B%rgHxVB3@WZL$(5E-YGpc=E*Z+<78=x+|1UZ>j%AO%67n-0!{e
zyGOK?7OypESxe$LJQP(i;v%;Bc^T%@nCz$CC~_VuC=mf0_nZ``3osr-F9?s%#p*Te
z9E=O>-Z9s85a-r?qZNWl0y)6w7L(d=EPV`oqZ?||ma{pUFeW}w@eXko8`@D2)LHw8
zUrt6XU&F409KZZ6za!0E17R|-xrgFPE$;pLQ7`QyGZ%-p0-W(#ec3FqQ?gjr@at30
z`$`t~86+I#6-7Dj-z4#+rPl6Dle_m<GW-tVlrenND{6Kj;3lWqp&n6Yf^`b|;|X>p
zEbRIa#@Y!V>78GdtXJ4m0-t)w&$i!5JNr03g81B_071V;Pw}FZ1CvdTjjpa*A|rq(
zB*qQHqhh4jB0!8zHLq!C1RW4j2|-+(NP;Of*u}-!*$!DdcgjB0PV!qaY_(k2-j&1=
zEV?E4xIZ5;iJlmC#jvvzJZ9!G!-!Iz-DWnXI4p6(8fFqt!NYC}d(Tc^{n<3Rc5*PB
zwszoUqhZ)_Px9HFYH+DI?JGWeDNY%1A{XfwkW*&{n2(=Js%l{ZFrF(3kTi?A(A_=7
zpa7hno)zZ!0-J)@49w{*1>8eQLEBN)(@pz8r_70#%#`zeqBQ#?H+11MElDh9<0CB-
z)S<d>fNWUBB!%lc*Ncnm@yuJ2{lkZNze~qoL)pCQoiEVfZl>VDEE~m8uSDg-p<HP3
zlMTb<jQ7gq7F<Nwh-vd~;l!pX+WAhzeU>(ezRH4ZGi~VV+x*=Jf3?K7Wcj1rC4<b(
zv&09MW9^D5&!)X{i1FN?p<GD4<yD4(5(>r;<l{nE_H?R0??E)QgNA<BJ^wXM5*)32
zu+ls--Jw1`e%At-SX9xm+aUNoiLmhf^lc#JT~jVVqM*?QAB!mskNk%5oq&71$25ZE
z;%|0k<d%{IE7M4a4I14Y=;+K>{&3<iY2$;YyLI;&M|Zy4sTYkbrO0X*c@4Vw-cZZQ
zZj^oB&UkL69HsVouRHxF^d_%{@rT=V=zX=$RD^-)VYeZ(<*F<R*Nne?+QmD9g1ucF
zLweIFHVNb2+8iI3-(~V$y!{gP&8xgRm<MVVb*;&+4`@mth}q1L6So{Dr~7Q#^lPd5
zCmcFx<&)>Lbl-X$x@mlO#|ZV6$?Pk`hHI|l=&@&K`81006N*Km3B|${z_8MyKy4qQ
z_d1c8U$Sm)Q=1Q%*YH?z6t0*#Y8HP4kKZ}{jVwJR?u70G`L#16X-P}>k?x*7R$n>|
z>7vQ0fzH^4$7gwtgV%nq)I!5}Q6Nsc?<v91glI0x7u>bSY*KPC%(>(e+5~T$Cnv<+
zpm+jE`|u}y(Be9(s!diQ@9$g3%clftv@gDQ|Gw2z`{Ptc=|iLw)y>a|4;K*qUGu+h
zAzS<BD;!S>%oh?ZEf@JMxveXaB@Y+jR9TgdHK}zNjG9)w+Z2T=AI$BSO=)1A<17@O
zfqDnO8TAA#SS*&Pxiy>hv`$?bpQmPbPVB$R&t(313Jc`oR#fQLfMew?CK&;V?@A<o
z=?L&~{-S@*g=mzaKMd3^y24sY#sgd{8Fv@B?3T!v4IyF)9vGCt6X>e=0$=tIq4cVS
zBAcDCJ`#lRS*8BM2Ad?s=$dE51pI5*#2llgr<#y|4!vOHo7HIh;8lWPjOT<)Gi6Db
zCwmDknN6DmqDI*J7^9U|tGIY@X9RuKqR*2Zvhnr6s-(&aLLbQWBf51yU_>Iv-c(Mt
z(@Rx}<6SpnzQ!^WzeH(NVnd8-mH`&8C>P03A-VDQE%D;XLV_es(*8=1wPb?A88Oq|
zCYyLQ1K+ptr>sO$pnQ6)qP!$+xo$|rVn0a0+ulxLdpbPv>Jeqdm0b#4z$1F5aU?#*
z<i2?{_QQor+D>=+B=@S*jg^(LP(qq^hPBam-}FZJrdmiNsADs?6I-eGd0hBvGfN$}
zFjcz%%{SPs>7|GmDH3N%$8enH<a<?B5@Z!fb8Np5u?34{GnHQ{*(;C&%M7q_H#-$J
z>l=J0si@;M4HZ=JkolmZq#PUY^QbjXNtK&E1-w?`B2hz;{Qa2wj{Sk`8lVOdtm>)r
zV8kSwz}erD-o#`WpR1nFNK^T&BP-L2cU<31k4-I{Ec%zB-L%7fbN{<`!`=M>!+zqb
zesx|Jiz3r>o_C?V6TZWXnE9y5B2=v|7E%nJm%>?OsiFs=NcWQ7fXiB;B!h#wPm_sI
zkB*{tw4yrhVyXIJJ?HnDKm4z4-`bbbe{$iXQ?EK1g90+M0GvVfn^j|Os0aG|l=|Z1
z?;rwr`uEBm6A{WK6S7Wb)cK4*M7P+6dp~%-7}XSER=l+Th(sCDWB6Bq86Rm0JcbLa
zz~K9q1He#XMpVW8sr!CSzv@R8qfr_UdZLkh$Jpu)sdwV_iTh6n#RIl_8V)nw&zAxV
z#)-7Ma#u~u+4{+T+<Bp98lb}97lEhSqvoy3_EBr;$1$IM`K^#L|J&g}Jhvq4_D(&s
z-O%C*Wr}*q)#lTzjOu)-2-=JhDM6GPpWgefklVahOF6z6_^DgzdJOEtXhddSDs}bi
z&Z?)ywV*~>VLpLn#8dw$cYRut$X0;9TM|>|#@577D?r-_Tl_&wiD~*}w)a?>#l*pS
zg+L#fboU`aM<evjht9G$YWSsg;hQn(xxa5wK27|SZHGU7T=Vwo5t<-X3XQ!UH4$xG
z-dwdFNdB_EWTer_bFE5l`|cr$3#K<}b!udw!62kY%f5y<Gt^!6#`xhe<$4M{Ks2ox
zFcnUNZ(I$RUO->#RG|4m>Es==b{)DmtClLW>RlJ)Z8XITUJ~7_CD%j<qZF}2Xkz!Y
zMaly(lQGA$iF{Ry`Z}|zuM&@SQ#&ro+(>og^im_jwXuD*20UsAKZcren@H5^*w$JT
zpxyyJJJDLSv+&iAOwI>+dqsG4qc39m<6plyI#5|ku&mJOk_;{FBC*}%{%6zYGzYMJ
zInG%wD$0YoJAGXdbEAtK#_<mBDQ&v;oAbU0DJy!Pb`0gncIb2(a$64`)Q5|NguE*l
zpAP$_WkCvjnV4-Q=m89WP!zT|E%hq?Y4K!tT&C?=jAi;T?&=kc8{mdFVSCn+kG8@{
zr+@A=&JzQJTsgTs$}ty|5R`QP6xT#$uJDc;d%1%j?b6S{HJ&Ts@|=s*uX;O6T~jxs
z4Xgx1c=`UxvVZ#XwXN_u73}X@?P=eC7lv6p7Ln1fsIFe2Uj2iplxloG=Lvr{f2Ov3
zmO*LCKsHJN4}j+`;#Uj%8+hL{bJg^g%vG#uA7frrn>;M2`_>azSWvwX&A1auVMle;
z{fz<Io^&Y)X$jd8i(4;nA*J7mP_e@8c*;Mpx5?E<QaAPp$bd4vD&+XKd1peD+*QKK
zSc(3W;#NuoOpB0(ti`J7DM{IB5tSBwO;p^*hYc>(aaYgP0$C^>R8lNo<A|m#8mVBD
zuGY0n_lr1GXZxbw4f(!h(8@5bS9tPb7KV4Qe=63VNl0cBma_e2CHv-k*V62Lrit+U
z8*-TxZJVmZ=+q7E?Wty_PSx`Ja2cxK&~u+F>E%5^>{Aq-YiXCDeivm@%84>9{m+@T
z*yhLUE5lB%mT6y9QK1w3tBH2=N4rR5BCK#elswn?^Z1VlAlalKsYXSa;VH6H=G`9r
zT8gh45q|%odI}aTCj8bvtEJH0?cj!l+RhT(#M6G{B2dPfiGx`7b8K;Lme@0xDdmfY
zZXrp-5$z2TC|V!Ti>E&&d5^h2D@b%oj~ygM-VrVnExeE0_8)tiL$pf&OT9BQm|tW%
zIhdBIyFh(}UWS;s0H*{pugI$$ioT5tIdz~ZKR|3ADkm|0dpTrg`0lwV^`)=-r>-cu
z&!&<9aUtmwYR&NNKRh0~YwBL%GkU!B5};ZkfqKeNuAVSFDwo)OBqtQ%Fj_MlT2oy>
zMz1y4<nhj#r$leVc5zaSODVOtrItOHmQkozuc$<?U~}yi68okr>Oy^sL#U;+eWKgK
zL{6=&ib3gmSoPkZIxhwn$ZU)izh7oDKdny|Nk%wM=A$Uj<!<X^^m!|CN+yvad=k^e
zJ6uqJ8^_mR`l`+PjeeIxdbKmKbixmox}?O6oYYW04t`NNN&49eYxFZ<YHGxfjnD`G
zKq1yAF*)^=oYDgB>Xn3?GsvNzn&f{40`cH8fl+LYT+CgVu}?Lq9I`|ZN{ZGChyL~J
zjv~JBmG{NrJ2Az-R*u9JD*NPhiXoe<Jo@pOMcN&N%!bn%dyO#PHWQ){<%s!F{WD1^
z`alNK4oc2P75U1Zy>S*|9CVHQ29kc%Kl3k!b^pES^#6yGwVb`kKRqdNbl6WiV4Uw6
zg>dV9>4PsdE1!*>P8t~K_qN1I@{C(3Zr}O1PJTqlax*y{*B9F=XSJZ8@tvXYG!{y_
ztMGy#tk=>I5uQ9T=ig#Q47``zf;ijxxy?T3U;ldJokqMvo`+)O#2y8WBWAx<q4lqc
zzi+**Y9e*$yZZZ9mBoCwnObgw3A+0gJyx4Ue;JX6wCynCNPghlznz8~m3C~M|DA3z
z)9cCe?ZQKIU|5l&Avk^$ykkVEqpx~%C(IWkwv1d>`x+S%sFy&P<2TK=HChPHV^11$
z^g7@CfBKY3rB{RoGGmsL{dT5tA%-Gnjec`p{&LiIXJ`vB|D-<UWwX#;-DEjDCd0H9
z;GRa5rBpl^1nm<sE9iO-(oz@Q$SoS3b+kSxI^ZcFWBqQ|OlwchtbZ3@QnRadKqC_H
zAl}`c7}}YH4|+UUGVKh(k-bqX>Q}4{e$hMX%jIYn&q&E(x=cMpx9lTtLF1F9>fQps
z=QaMMuuTlpt=gSxW$2PMt(F}wbl|BPmR=TwbFQ6@h9V0sL*Y7#|3{C4eZPNE?iTfJ
z!Fr_G6m<O-z*OM#f6?Fm?^SpABaFbQH?|8!W+;eiW51?p6|3<u+ZF{Y$i-V%kBijh
zIH1j^D8|Irob!FX30V{%;hn^cShnZ<@_;}6NRZ9F`5JAwVs)%AP37k{2*+!(%lM~L
z$d)oooUJ4R6B$-9!@#62N`+v8D=^M^XxqqRrnpA+AO|qbyu3m@RNsK~{mhTV{mjjf
za1P<8(2b95UwDh4^$Ts4?0bKZ<x|e-FujF=$sY~6*4$)5RC4L=j_&dT?h+?4xZ(aC
zojHzKacx6Y%R4*#tR0=wW;ULx`7G)4T%U?}fA_naO@f76M~*h6{9BUL_?xA)rI@ti
z4bRuT6j(>IiQCU;Va4w6HFP?HX-O>qSi;h3c!wqkFWJtXFK8x_PKL<M17zN&H?+@g
zyhI?j;%w_W%Lggj?meTlUUZmPloHa$TCVN;mB|{Uvq0hLGQM+iO8@{-*bZxINGUIE
z)optXepcnG_Y1f=a`TLYhpVkgN1^(B^){8OKQR~-u%lopob<H`J20eB;ME_n-@^C%
ztC6>N5&nOC$ayN<^yk;sZR?edoUdNr{HK1kNs|aaE>9j^yZ`I;;`Oq?7rNemU%&s8
zS%y&LGvQ*N8u+!x%{Cd#v|rZss{F@GWchuLZmM6vf7P}F`D3hc{E2Wq;2o_Q$rSa%
z>VK6o@)Ayn{qK7&x61ew|EJVi3MMAAb?<6~@QwBSZj!^c7DAJ~yf-#F-8Tr0`K#*a
zH5W%Wr3wD^zp6bvQ@KbSPg(fja~I|%9P;7zzs~OreDh2CPm7NSCe&q!uJWe(6E33e
zr(j<X*sFmPHw*sk{}ouIBfq|P<9&<&IO%4_NB`t6p6*A#`CrdpFB=2)fshyGTtol-
zoI8j)P=r#z4$t<=w#a>zOJDt+@wHp9ah|j^I`bcNBloKgD&G~**?XV<CVd3g`FuI>
z3wX@(7jYBR^U7N%^z&C5sy^Wv*Ijwp>(B%Iy}E^|y#STdovXa_-P>D}f4)S7pPtA6
zmb`pvJ9U`mv$d?Fm9rtQ03P8wsPz-vO>UYit)rnTIXH`>Q>gk))tbGe)3v=eQ)Ec;
z^LvP~^uaV)XNn<R#|O=}b^WoT=CPNLo+r+lW%&a;iPo&Xd$houcGBz?bmUhR6^HK&
z@~x-23Cf<z`&)k%ZuSy9Ek3wteRmhFQ<*Y+D<(~Tsp**@iZBs<a7pqp%_rb5HjQ2#
zy_EXTw^>B8$Aq2N?&JaXAN=+o^(NUN_)Nh5pL_o@4Pn`YMEJkGL(<^WKPLWh6FYoo
zpkr^N=HJ$1`;SwtjQwAIDT;34gYU^d24?$jgYxbS5FFsP^>QodKd$p%{=tC#<`?=e
z_xtCjdjB!|zi-6#Z$It*f41)*myUg^vp{FNPw>vy##Y6R7gYZ*m9W1sSM)SX`lk<k
z8Q;|>v>kX=b%#7l(fxJvv-|%4>ItT><yPe)LLWMPfH`{2)XJqGmeX|6_*+h{_mPH%
z&O!=WQQ@<(u&~4>ifUul%wIwIwzbv-Nw5ElGhq$ZF}2M*Oa(6zroz1kBPVZtNl7q?
z%}LaW(nOd0XNp&yb=~exVdEZ6<879Hx2v$o^?hKPXl?C$V4#z}Yv!WL4B=Eu`^A}m
zu#w({Kf#+-&!^$8E4amb8e5OPm;vuayd(_Qj}|oOi^IazhJfGHHpE80kKLlDY@M#J
zSv6iF(cx~|*}SFx&!GN@KvNQAnl<y=-}6of(de(ov9z_j2S1U@{LP;LMsJ72RF;ze
zzV%t=?i)hkyI@x0kLP)-U?t98BZ*Eg>DjActm)*eO-o>mboEHosS+hp{?T`(WkhU4
zP7u@c?)KKvxZK*;l$_nSL%O-jwpz1d7`oBF8vDaiDkr|%rPML&+*a|8FSSBcFK$dv
z)>%Ys$#@e&m1)QFl~<x#QLSGxxmN}}_9B(nrv5y;d^oIf&U{`4Y`Y-qcIZCZ!L+iT
zDG&bAxzHZ{`_|O~A$W8&7ZCD7D(K)>A6I=!lwrlsrLdo(J@h~3*f%S2A-D|qe&MWC
z=0IU0APif!+T`pb-v&;j3}<jmPXahq9e8?x(6r172)=hVa<IsmHk+J|SRIrp_wqat
z%_RI-?ShWnU`P5>=_Sr0S`tzEwE+Ao2P7#LuaQ-v{W4mS_8y<%3Cu<xy<u&smiB3<
zF^n`Zdii2b+fYe0S}!>lpJr65)vuP46v}_ce5}G^P}!Y(P)l1r?u{%HV<|BBj)3}i
zcx`&A!<EW8fpT4wpRUl@Rn27;&I>^~OjbYkq2+><cyODBLW|5%lpkqh?S{(4*?6I9
z{QTM-ZZEA>e2mW`wCc$ogFHeMsgC8wZb(2r{^%7G4r!D(8NF@_@h}+F62C#fgWsqM
zoFUC}C>a_#>VR0Sf<bvY*`%a>T8=cLc=ivffg<=r=5M3iNZn+FO}*@H#v2u6l>|Na
z*Wf!a4oUcdyGkgfHA8WXBk#UiZMmAHXrdKpuqX(a9%WttJ@)zaO)%?1&@g7+7(U`O
z&ENP&J4aS9i&9HNt6sC*LVHFF4Q|5HFji)KN(&V=an8}seW7o`T{K60%}n!$9`WoO
z)mSR`$Jmy_%K0cYRqSMDUl3aBYjR4oNi?wpU8{3eaPhcX!l^esQhcY{E1UndxTK!N
z*O;o-+dPylMFK#>pxx2~##<2f0&dOT{9Kh{j!6!#(Zg^edRn9n74vhEnZlk>HnO8P
zbw<Ns@j0`kUoYhFFYZ<BN-tDnS=~r1H^oVK2O1Chfj)<MLvnwB^%R$pcv&D~w9c^!
zMZDW!8Z-WmSbstj;yz{*D;2I2sOytYf=`#a)&?-Bc0w8+&ibk`NOUrYbe;5RV8G+W
z)gx47?U?RRK!4DF77}NMD%RVHBFyszXhCr6W&W+tI)j6n%Ba|NnE%7CdHpJwWx#@G
zidh_Cm2;)lCH@(nU6UivQ^!!(da-w0NCoL2{}WkfiPnqtDO5h^cA_aKq!S6k+`68p
zy%&*YHV(91{W6W8u1T#|k)4yIPuBi9A$5l#5S(ZhsELK_>9SW+$_g<$mrF4{qMkWo
ztDgMu_O%V*ZM7$=pe>KoSHh`)h$wxL{39(=U>~XC{czxAx%salEtwN4-ICGG^}=;-
z4wYyce1AdGqZ2A|$hhdz<}7cIXk{HsIOS)&wR0jfC$czhj;Jzo6T7iFN7Sf%Fg;bp
zDG%qcd8UT9G~a!So}RhB9>TKjPC9%zYpV}+xoP}xo)gfM^XozAhf^Miym4Oa`RbU?
zO+Dq`x5iQ*Vwd`SZ5{_nIxcE>%nIGxf<p{+#St?Fl~gQ9Yn8^?L=H$W1=do8+K#TK
zJG`ba#X(+2X=DJxm{2~|5&3v_{OCJYcOoihjOF7Bzjd4DYdK1(m<{<q2^YHim1~_?
zyG0+Hz}uI@MorD!Fq$7B+Zxv=^`ncK8=IqW8b;ctV{vHEBs}k2pLgrSU5HsdG1J=E
zWT-&Js2)Y{nsX+olf;&&;_$kQf6PT+;99!6%y@UFnE(Cnn5CNSasJwHBCZl5$n3}a
zXLW%PA7Fy68YaxosJ+1%7gedi+%7@yk&au9*x-K*;UWMKrtnzt5x$?gC`1g@Vbg{!
zKQr1Cj%k`yu0A8$@v#az06E=cmIf7c?hpLDdBT;eAYowZI2z@e_gb=d;@7wB3g^MD
zwQTqE42uHW6tm8~^W_Qwi@3Ea_VEo<8n1@Xu)(ni4bJ!nA9Z}m4*>TXia~~j^~MC@
zt=#%)Rs<m`bgia38WLMWv7Zyf-r*!iv@+|=*~w2-8q8or!wr(6Z~u{LXkefPEi?+p
z!gxh9v6S}s-%}j(+b-nIc48o8i&0mTa)~5dQpxj0RCX0My=Sn?@$^!8vWnxGvJt%@
zIns#jd?cYaVw_gSxY|u7b*NzGPhaxw)%hsOn&T@bm+ab90WGvzBKKh~fwbP@=J#5S
zuQ>sV_-lqAQ|&vpO||+h@eWX}d$}V68tnvF3DiG0OmRr(9{o_&Qyq{c5&db!a&|aI
z;=@NP38!ElcBNoov8A8sr*}Wh<>3w+Hs+nz*{UC9ASuzJ#1_=$mZA*z3J@55ZZ3J3
zP%cq2U-hR;M5D-<0?_0hjaO}d5O?Fau03?O^zaLxhkBHo|E9z|?_&q19+gQ;R*I^d
zW_HhVyUlO=^1a!x-QSG^Lh_cr4T&FX8NqGtfL*L{Wlj!pxjV!uMTmsF$lCZk%Xt_#
z0`QoY{B(@LeL$m?)8|}gGDC7cjQ<%}3X9=74;i~o4z4NBFTFJ=SU`^$l8sC9md!MD
z9bGS>LczyfsL?)Xqqs74sI~n#<!gN3xCal%K;sC3c6g=Zr<QMSI1_!y>S~lcqEz69
zg|pN?$2<Hs<{X$BDpa|i2YQK$=BFi9AB?=rQ~Sc*LeYB#xMAfnvKS0mG!e^NgZrG&
zgh>^eRWxG2ROWve%_&PH=%=%MjFLdg@3VzXohS25;H&(#h@?Q3-OLP7KEM=qS-|Lz
z1&6EJ$Xj;Rte~@K|1dXHeYkxkZ`VTxmq?4uk&CM226UIm>7}&01NN`?iVXoVeUI?{
z`6WMHE8<7E*x=Tmza@t9i9-Bs0=$o{KC}_9%>GH|8$nL@x5x`B6&3)jPzuQv%dPp(
z`5sVePjJJ<Ee8thDw%CN-AmPE**mT7B?l5w3LCiu-YS>>bfxZE5C7_C+v2Q6(j$wz
z-<%Ncsi|Uwnv>6HA}b1`^(O#%m!j2c%g_+}_e{nSBMnrlAqOnA0EzAr3Tp^sivD>w
z(U?5RG4;9e*nV=nh~NWM0W+%x-c++#?{kv?v4=AktwnXno|m&ENZ`OMCNa^y0pGtM
zIX*fg$M!(r;c$1b^tM>GJS<Op+nCk7w+2}=-^EZ#ORV)O%~kRk%_z}lzoW7F)sfF0
z6-+K?+25lsV5pZlRlwauDa$?Fj5t}v<)po8tb{K*3w$y9LV)6oKQPw`(=Wj-!KNkY
zcN>t$C@gsFfH5G^r~$>I1fwj`4+Q0ENIW-quJ_YndSGTjzTk99<J;Nu?vH_Xt}c~k
zvZ*Q2hbQ0;{7$%$Rbd}cXR@zp0_3Wi8gC9zO-f45ELNIyPmo*)9rG0b6S!g@k#x@i
z<K|+lQ&4#@6_Qw)Msf26zHjJS!3K=QtWShf#C>k+qumKM$#)bkvCSvvm1&>VN!5Da
zT&d5zJyL#2Q{rd>zmYPnN19X;d(@nD7&hPrYLnNYc5lZpH3PC!1&AaE08#trhR*z$
zGpU><L5V-lezt!1D$LJiwT@=6`xzN{`d*%u4DUcg&=E$GL8?K~hi7no&Jq+zKusA{
ztelycQ3l&`pgP+LN-eJAWEo?^e=iYf4Oz9Me<WSRI>oOK1K_!JUL~i*-eE*<6f%|z
zRj&4Ylt>?aR>~W<!Ii72lxS0M-=P-Hq!_<O%VN0{SRP9NGF&J}z26kJ;-UY+jU&Xa
zNXxm#Hz(|e#~C8ITQpG=VW(I2*h%fy4b_#3ACce;2SvSrVA=U5zQwh6@pA9NN1d}W
z*I@Uz(MPS4UoYMh5^k0ejJsZMX?5qs?||(Sb)t#0o)b%`S4$d<m_suJwK}B(2<M^v
z<hjDo0&U+@U;Q>AEQj%kwlhjm6APXYfgzlpfyXydvEgLq`KByX=>&v0SoTh5ESHQR
zxzQ*s$+DL6e%ejI>$=nVBrMiSLCkb==w;DJneETNZ@nAAxb$u6DMTB04ayad+tj2`
zj&D1g&})53k`0;XY|wBaf|m#-1~9vqG^y2<J;<9tw=Cv@Y!d8i>D8NJjl6eAX~H)O
zKg%Nsc!*c2K^MqUw{*;0ZGW=<EW$oDR6K9D6yJQdJSZE^LaMqzag-0^WLNev>kr$M
zp_2esZH7Fm(y3}@Id8G<As0<#7H8#nBog~*zj)@u05C<R%KuXAfR5bqxoVl2+q1qe
z2Nx9-`pxMj^(Nj90V|D|xI#6pfq;cPW6_06)}nHU1EwqL9cAh_PONrKv{K!vV=Z(V
z46w5AQ*CTvefv5z<M>NzqA(~R$v!wKuT^i1)WLk5z`nn%Vn_qXyn%DbiwC*CMT|v1
zP{J9N%cf|d`=!D|H=gxf&A~|nA=LKuqT}xvI>RO6DaA0Q-eFO`9?M??g(l`sxY8V>
zlbCSJ!zPvs;&0!-YZ?c#pgq59)n%4XZb5nAdx@)^845-0X&Nv@ktl|SS$~5)QU#f#
zVLc&ER(UK3zSNH_91>c?@pcxx{47ofH<jlHrWgu4$tsy4%#?D3X(fof)%EpDxcF|L
z(XS_e+Sv`<v}NfGXYR2(N@jJY?MO<$Yx7#VX2-dobNnpcvfEFsJL<glN7(keX2V!@
z5!9i&2!x5#Sd@Gu7Q&*6q_1IomUiA(=jZHSkRfB|;Rp;3Ljk%5{0+~$o|KcMhnh86
zav6rdDmULeLA`icQJ$Vi(rBsZRLfrb3OAtd)Rou&d2WjOdu_l_!06&|y6gG-f-8Y5
zT`1>Kr=K`K2dHD>vgUG>Pm<V!9Z*PLtY0aa!54DO!iox;g&KMow8s_qzEYEwxMm0j
zr`^Ritle8P;fv-L{c|^v)<5x>eGsfM@ak7dm}ns)`|n#1UWB;t6cn*b+AvgYZ;x2Y
z_D%~0-jTa5r0~*pHkM4Rs!mH{tm<kg(P3xqfwea!_LjU#RY~wJk=6_E`#iz-pf3kh
zxW39XBNs8xyIDiga-BMLln5$~FsPv`Y?5L_4~trktse(rY2g<4Ck!tPdx&tVkoTsF
zO0u-{A4!ZPli4{?^aGmb5k_8%JCPza_#`h-RBE_(SP4b!VoV{__fZ_KZaCW)hbl|k
zJP|N{CFSEpgnbzrY)~8Sy9#T&1Ksv~z&W%1dDVlGV+RR@_21|IiLN`a;C~cKzdOH2
z>+hNNI5!NF)E$+|d~fV~wQWsXwN~b!pZUv92~vJYYOpXXr8!~_BMr`FU}&PwK<d}X
z)Qb_uI<-i0rPoge#tYZ>x-|?R1H*hjkgr9N#aM_Y&GgCi2R)EEh%W{eFa*&y4Tn{*
zX?Fl4@T!uQ3L3|`Rf7jCIR$$U7PO*PJZ-DW-&;-`yRUMQ@AcD?TV|bwCS@p<IDe}<
z1fVTKonh-;t?aOahy_qS*cliOa+ZV2;#$}QG)uVpz=G4m!1Jk<gM<5#HQI&5LNUeE
zQKGu}JLODvSC1`>X6LiO7Ff-s_+l7mZg<CV0cMP{s#+&F%$Awr<fjL(CM7ARjs($*
zi(X5&{)V3ihHPT>6UC2gDp~-QG91Kk&Ig)3;k49W45VD*t@BmFI^8oRkgp<d_r`?3
z2Vh|hxkmw>7Rw?0D~q&SyKR0J2a2>JZA{MFcsWp{agF6l=_w-4zVs{1B2lU@$F;S7
z+&ICFoO$YArd0kZ-<!A<x%t<fDYFBtZ}xQ}Z!$}!Kit<^-Z%b>w-QNHa_Nuc=1=jk
zc!AeQiqHCJmV)kZl#C0Cqm*YcN5n?X3TW2rPc}Un-C!wV19w+$0isIAu~gcDf)6CH
z_CGXilvO<|a=3qo@;BQ$>}pk=JJmYY?=6lEREyx{zf9Fuqx$nDOq{x~awHW>J!O3f
z`_MVz)pKHFS|EHqT#`iCl}E35!Y69G7B6*j%47V?hU2kzAoIihHfy9%)GlkN^&K3#
zPdf2}<o=YXr(*f|a=Vo+EO)uo+n~H}Z0#kyymjEajs|F;yQpzw1^@K#TQ1R*OhjR2
zI_+ckhCIpBxi^m1RRRr5`p1iZx@2y=ZEh-2Pe}|HnIgnZpclG%EeA1STd$x8hzVbY
z5F;`1<=^^MZ_}CB2j{QHrz(}u=D;TI`MR7h5BH~kaZQ-z!uWIx^I91G;D$UywNs$l
z5N#GLu`wzlSc)^vO^<%61U>%NPPnts0W8be-NchJN+w4HTII!uQDeTJCNR5l3B`ho
z`@w1{9<oAAaews(;fa7r!SXyTBv$Cc{BMD`M~%qs&Yj6ed_@X5(5rN->Jv{BwA1`L
z>l>Q%s;NYmJ41R3Y=*mizJ+XO<>vw|lLgcr3CWDG5O>e(R&k2@YWdX4<Wh-lxyXm-
zA2!ZiQSRDxsoSPs^2PyR<Iz1yY3;thZpK|KfSx%fDHi&NwlHU-I+UbG8a2=1Q@7lC
z=D#``KB&G9)u|pP8!O}gdOGNIz)?`*4pJd5)%Mg~9qO+tqE&*+t!OD1=!>+-+T=WZ
zGB}os0&UpXSBdPB7OJI0gO7G*`VZ4xP&$vtcPn=c*$GK7XPW4-;S5LV$lH|`qD(zn
zJvq_j;sq;%!p;4f$_J*mEh`UVr@z=8Ig0!ub2K=nWB2|vNEV%@rTFRn{aC=!7hKoZ
zCjKM~Qzd!Hhz*pTA&P*Lg~bi%?bL2pZ;U<b(}JV{w++~R6)L~xM0j#_8HTVtJr5Zz
zi>@S(IOgoxX!t}%847hAV$XR|B1nnkQfNVRLTF~)GoQB#Ovp;q(9eVkMOE;e&5@h&
z{D9Tm=FmxW|3g}4z<GQxQO^6hw)Oy)n?vcwo{R$o>QhaMrqN?07dPA1a(!+_j=|0?
z9E4l}-!h$B?;h5<U=T$%{KAs6rsd%@9*{K%`C)w-q*TswVQl$ZG<g>Jm)~<!%q_03
z0%?*Xa-G?$3N29n#Zf+YkMEpY<X*kl5sHL=eFu51DyvmAG;vgupxu=P5(Fnz33O5$
zic?Wo4)zo&gR+E*G*eTfDfR@R?hy*lH@-(x4NAzxY#wJ3BhAVVd=7{u`0OuQGR%GP
zL_rZ%ZaS6>y^b-5?H2buMT^sqD21@TR+b(USp5_2a5<;y+T?4jg+vrrlN(RibaaTp
zCa>}_fe-B|P`i9o^v-43h;>KtIJf>Tka);~ibp%^P7##P_U+4(F|$${knngKU)GbA
zw{QG!?`Ud0HKP?Tu&X9Dk1_~-m*m_tU&$?zUIZ1HWGTYG=|*LI6z89tVpsdnqB#iC
zD5WmZ0)ds|e;KxhMGR`_j-3mCHgEi;&+T%L!1>~tHxw|bnnEr?ztcNn(i^u#&Rvji
zn3XX@$zA`ZH!fM&Za#XgCy*+L`UPFH&^+~^f)up+p|kOXFNXA5Io@8gT9`jOg$4$D
zB@I8rYB%A6CkuMwIuVugzf>@ejZysxL(o(N%iaS5bsVjUgKiJhM@W7&`3`Cn4JqPM
zOHJ+&!HV!F0kr0=5;cNSOD4IC&un^?B8fTe75nX<*nIx#_f#j*Fxu_K&)WTC%|MOx
z^a2CqF@M4l`(WkC0N}SqW0=zcEIp1ZcW|K8R<d`x^;l%v*+-_2wm)(<_Th6`>FRW?
zDhp5MHQ=aGoAcloGuk1_P+t>q`H^zyyUUZ@!GM{3_^u8mMiPT+fz%$V>LW0{i3%`)
zHGt}S6UM%2t5}ih;NF`B3VTbURYxNA_>LqeZaqNt88MmI!)AT7iy(3C9cH+Xl3p?@
z4pjo0*wHZ}l75F?9)6h_OU_7X|8$9SL`tw^^@Y4%JXWhKD=k-h;9KBx=F8(Hj_3CW
zgda!%zo=I}Q%O!GM8Af8b6Xi{2$L)yy37}vgU-_nXoG-ZM~cqfNg8YYJ6tC2gXXvq
z6yG}sY$cCct_9bGXM5rBCj)Pq1Hs6<k8ylw;-8t4*XDmBVoG2#PvmHFnfM4pvH1^c
zBu~vB#dgTk9xYjwV9ktg)Rz&j9J}NjLd*n}Nym6cH_@B4VQq2W!ox)ckqCnm&Y<n2
zR2_n}Et{5?8KBsp23tT>`cSR7t8fJ)UdT3K`DzlbYQ$4VdObMK){=dieX-f6-rve&
zj>C$rJJ+a*S6)Oe@9VnW*wKpaUkoOj3uxh2f(X7NrakA15VuR2K*Wuopd5u4Co=5h
zRRY!|P+YO@Ipwoe^v3SZF@du8W`j;In@IzBXqY)CQ|?}YwyRZBLzOS;%#$$t06{>l
zuwrPz)@yQ8z|i$%Bf?sVS&(jt-M3WM!-871$u&hUCGDf2z>lSz5iJI$^34X;CkC9A
zTr!cOygMb~2H!FxiI~n(UcX174EaU>)MVx756B1-Til(#hy|BjNK6fFgnPGsAAV**
zoMg~-qFlK6K2m`bW9+z)qRKO0+Z;|dMw$RE$Saq%(AVMQ=u7}IdT>q#AGCGpKh~`%
z43^KVSY%MS4t)H5NO#*s&OO5qB*{TER>0@%w`*$4w8Q(vq#Cg0xr?RUD|hRW&ne>G
zeOkouT?cTrzD}z>R9IvaE-tPY%OJ}sd{b*z?&ImHt+O)k{`|$9(9!c-b)_2|+{syI
zQIdYo2ZQ{WN*CV?4w=+&E!Egd=SNQmzL_$v1wiLm5?_UMGH6#5n=(rpyH>5yJl|*$
z?Hf(TtNPOHUmd(HAmi|5-pK)bWXgZytW=G(SPLhn0ibkn-#QYHDw1z^7{+8@c9u>A
zxaPBPb{-%sNPqTCa2iv>nP^`+eu1}2=#X^F%I51yph844OE=(x3SW{^tKFz5(68l)
zCK$f#Cs`#ENJXg$1X-GuBh5x$nCXZOpB7BCc7&T;!VA$}+OGMXH()B0i$KrW^4a(P
zr%h#C_=C&ZZkT3>Y0d*HXn!3p1Cn*<CaA{&;8)X>dE_$lV81A#aBOPOl*;~u=1T3#
z(~d*kt3N)ZxYo-fU5G{F1s@+I0bPogB6GGNwGM<~Cb?U7e&j^NXD8Yd4OvN8(E?;%
z|L_9_@lErnOxHzgrDmrmyO-iuJBw95(puKU?7-!l#(w^+xROv8gcn(tMwB(D@V{BI
z>K?vE&N?L1;{09@`kw{QaDEd|ZFJhxBoSGOFhO=*g~MW>eEESmGr9~E9vf<sASp!M
zA!x&G|FRcZpQ+g;lTFaWmQN^ryt2O1exqnjCa6&Usm3()*fab1@F^O)XV^xqk`c)f
zDpDHzMYmXDN80s!=yROBs9I)PW_%%dG@0Yg%k*MxuxHJMYj)OX@_p<M33)*+q)iUo
zP96R`jIw6tUdVKH?W3kdH<vu@s5yd!+q{Z^dMosR9fsbItF;sWdybiAz6W;!%1CT^
zIHhkQMz+OWC?x}7;a*Vt$0kX>S(iYIz5G`t!?pHOUBqV<&P~8tmvM*W4bN%ou$~7J
z_6ZKrQ}uiwzVBBbD65~JB=-1_ji1MexcfwriqcW^b?0e<JJ8=D=EHq@Tbo>5dsMZ^
z;D?lYTgaNmCVxm4+)usrm{#Bs^p}2Z;jh>u>fx|eh^W~U9De5h$w5NcOmSSY1$BRW
zq&QJLs~|CgbUta2&pXToy{4J>To31#&*`ANpm@vWRjKpWOq%Wx3-;_`p%oz}3twGE
zy^;pOrAZHlT@?vFO&5?BbRs+`J<~;1okOztOz)a?Oo1}s_87~HukjpB)))Nrt5KZ4
zQ`k8QnM3&#@$c1KISj+LvD}q36gU+FE4)_B>?nZGFmDXdi?@LBCD2e?I9h=F+tct`
zQy0Fbrr2s^jxo#AfLt3Dub*pel*Xj}kz6SpebH<M@+``apUruP&bR=aCQY3{%FzT6
zznvahLn1A1p11N@CxJ~(tv#A(KayC0r-@13g{_gPmq4PXCc5v+k$R4iDN`Z93pvuK
zBoXUsk_5ovKODv|fnoyIkbsqJ1XvX|`Rbi}?wb~l0N(#mn4g4eiQXw<&{oZQ-3~}%
zAnSAM<#17Bi@IMhz86>SIiJ?Lfq<1;Dwhqq^t^ChkErEV{s6v-Zz;TKw!uJB-PEYl
z&$Hgm7Z^gSq5JmyqIzhE`JbfyReM@jn;e$E&cDLzz+}cx)QY}27*CI_Nn*(s!^WiU
zjY>#U@Hn_prh<I<O(~gtOjKW`rh0xT;jKucPYf#q8kVIbTG;~1YC2#ZiDXw?CxO`j
zu=_hg2D!I=*R|Uax7KuLGjhIp_i21^Eq%?LKC#uXg&=re%6jT;Qw;ITBBkT4?NR!+
z&!YuR$;`#NW*Hpw$<18{X}N;8z}p1IaAi;fzuGMOc!3r_45MdVQ742-NiVg03i%Pu
zr#cm=(=j&v)5b=zxg$I3T1HfnD3T&!g7#(TkJvMCR3%s>IMT{aI-ta;(@Psb`SV1N
zXJ?d^9Fe}2Z=p3aTJ&1olD@1WJ+U-sW5VE@L35S%QRwd_xy)NEiQ&?r_AA^hOHa`O
zForxMLl66TI@}c8SG(W8&2el44ye-LF32c*PR`LiK%|hMyue6Cnd!o(xY$08td2Ss
zVr`f9)IK@e*D)#_vHPjjS~CQ~J%6AIrWvKvD~qKMyNF1yn88t;JHYgR1E|Jo8tMC!
z_EiG;*aTTzZ2<dLL|)F3G?R<P-MI9$M#u2<+~UJt(&x>dIZmWKrp1jx1u|kI?E|<l
zX-Q|DNf>!zKqizX`c)xUyNMA+1^6q#Lwi+IZ_(#hqhLm7GA0eI-)7fR=c3IxBRrF%
z*4le1k>yCdD4gG!+9Rpqpi~St$4C6wxv%Pkz(tb_$#*(P@(wg{(7!}a*dg&rM((zT
z62>~|W85;OI}FW4ulcpp9rz$WZgkCucZ(Xoh5FVG6=-<6&=FQU5-Cp%4AXzjRLj@D
z7~tphZf`EBG7Z(Qt7?h$z?c+5?S14SS-~mLbW5B->nI-VESvFqAFfiC`!jH6Gx1JK
z(ZZuqn_RIMXah|jnYfHV3ZdSFMCsO8V*A*)=czmIRn(sQm_E-6?sj{e)czEMWJFbR
z9c?&E@NiFdhH9s|O9JPMPo0w56+cZmNj~H>T=2+HSHVU~&|K<(G;~1wBn~9+YPXyB
z*cYYd(%wPGW~qJHRCg1AV`<nn#mHLN34{Mv2`k@ZGGUdPogvM#d~RezGo6@ZMkUiW
z^$gH0UJ-(1$LwzGvfA>w+dRPz&QOdVQ+pZIA^UusB#m?K`FSO44+L<#xHUT0zqcBY
znykjLt<THY_8?@9UHA|_h>)kI6sbm(VZE95TZ|s)n`-A=a?U;t4&j{3yCtoWMcSmW
z>0hGuP4LnEpN$bXQw>LB^+<|(O+f9wL+bgy8`CBR0LVRLdVQO_BAX`}3RJ)FsU{xi
zI^9Xm`;~_Q2yLi9he)4!3I5W+S1CrFXOuB4)}F|{!_#vpPL@mw(VKrtwTmoK*w)+q
z?l4)DWUAcO7Kvu#a{v3*M}7fkEZ~y6P*eUxwhby$6sh7<&gz++XkE`KvFh_-rr);a
z+nGJgxU`Nv=5&MHA_qL7bFu<_O=Z3E^YcChE+v@bk1i=8Kir1-BVjqP@guALhevI{
z)Ysl#t6S(?5>T%r3ZrfN>;&cPV2I)g@LkvNd#Bl92muHyr|szD7cDqgYpvAY>H1fK
zxugayH|;E)HAAA{m;`#q;G6|kNja9Zn(6f21=szQD?t|}IhS=Cy6!$?a;?;oBzkX9
z@v;Q1ExP(bsG4L}px}ja<Q~+1r&pjlnTbg*yx?fahDxs-IXuDMIWg%L|EHnrh<zp8
zlx&AsJLXT`$B=@eGs=bhf|9Xuk=#a4{vs>(@1RDQo-dzg-MLEL;VENsu-hEexII`u
zQm@XLfmwDVf70k(1nDjf;;tmfU%M!!DIt<Ohb>Lr#TY$OnuH_KJL1j<6+@8Kr=Xw$
zdn8UQR}z#R?`yEgwM(floRpU2%jj{CtJUF0Cn?l)0M*QD3lx1|VMdLA8twPROJtJ`
z2)ub5NE1#GPhlWvAlkcbM&hP5PhmZHXy)rxmkgV8&9xp8t4^CI|7xEAL3hcjL{T@{
zze7@wiOmP=7nlRHknDxys>zd^+|VqE*8LP#R=3Klna{G1SC;m1oSXeq7=-DlLYze8
z9m;%5&ixnO>KU<u&u@x<RyTAv8XBp!&<r_KqinaWxRqt|)qq_mKDJ}`pa%UfJ5ocU
z3ABm_hTsCN4l8wUqj)xbYwKneEfkW1?u;MR^Wnw+gbqxx{@<VjC;t;VU@5fGy$={D
z3K|VdZHs=NCHMGfcA`-z<dCGQv`a~a*kbRuvT385=KfeNo`-h}zKh@WH6>Nk5C%Dc
z85p=4vZo+N4@Aa3tf+;EfbZmLTWIMUpvPFX^EEn?gT<w0J<}mYvK31*EC^HiMJ2F}
z_2qGz=POrnqut5!a&CU=*~c}UXYlgp<f=J>;$rce7IhqHWR@!)DZ%bRJuMz<z>v;r
zgXdiPDbVTTo7y?S{HnD>+Xvqu46kq+7E7H*I`o{iV$LrdM0XwwSJoknBNT`<Qh$3t
zo)(|o<)0&KqUf~SYGjC{M29@9r1_98fY@WJly3atpzo9Wqu&0#8X3@aHI1L$)Z>j#
zqaVWz(6yWssBb1;>kB^$v?EIk@#gCjVC<k<YIHH1ZQy?&<O^zh+tb7^vV!NG^bfX}
zt0$GUo`Itj>fbd=RV!fMj#_m3L~Gc@H4z|+RV7PW{(=56mG#vg?`iC}bmF&DxtaeL
zbMGC^_WSq^cWFN;irTAo&6=@_+G5rSA@&GCY}!(^ReSGEORR{Qfmp3syF$#OD79&g
z8s&56`@MhXdCqg5`>*G(C;2NmIoEn$xz2UHuGbK`2ZCWP_^e4T*NkbroHFevKD17(
zMN&@}S@rH?Rq|))qzFMmn-h7;^5@fonptn}x=rNU5I`bqLuQA`4_t_#Nd|?NzbXnu
zIByyBIdf51=o#G!*2s_l*`Hh=f0oKETSGyo4z7*;qcOWcjWRHwyJ{MmYiKy_u67zz
zatO9#sr<SNVax*jv>i>7eL0-Qogbz`4=VySm#$KM7A`Xni%#QS@Mhl{bfsp}wtw<V
zg}a-RU^kQHr4{G*u<riEMygEid44o^uXL)A7~;Fa<Ju?+zLAj^q3uOR0M?4`tjrkN
z{r5wtSFeIX3ode&rRai0Y0KhX39*p$g`=GB6C?Tgsg5WmE#l*U@&P1eS3h|jztD``
zC{=Ks{WHhJdyhqawk)po3L|V)K3|jYsx&4pr*GUZbh$g^J!a0*woSK;h#<TU->g{1
zmBV%{$c(%^;%gxSv$(wwTk&Zv)=J~-J(YUS63R{L!m#$k5_FV@rKa7$>=RADhVok(
z27iZH^gdaia8@cQ@lymWZb^fK8y^kzjv)SNss`7CkXf$=@okmm0*+EpJ<*W}qZLEi
zDM<{<0Eo!1;DxLEJE=Mv#2y6Xy`^|)sXmhzqa?;a7>PaY{R0EOdgbn0p8inR+d|8e
zg?Lqd3Gggn26w;SaB-Fd_dI(mP3N#OYL^oL5|Z-PBdIUdTk7)7Qb2H|So5<sQN4C0
zqk7#s>lv*^?sPx%jdEaZqghxL=fbJkX+`B_P;bBW{nN;a<bSFSK6Rs+OZEIS1QjT9
zl(+>=Doh2L2Nh9Qre)f_QYJf6W|)f(o?5Eysr=W>rnLOroo)kxk#Bh2^MrEI_bfdc
zm@gvvczSZ<$R}2`<L*8k;RmN5x_%!S)N(<AP0A}=zvb*sj3uTqu(Vd#HUdZdhEK`X
z%o>;xZHdfE(cscOap{_HWy#^|p;Fsqv)Qmw_E)+3HSEwLE4ptQK-1Vvs0!1T+&|4a
z7eE5_ZI=;XzeAy4f<%JAYGcFds2tTy<>qK!9936&8ZRPUE0pcR`X^FAH^%ERh#&QQ
zlNjWrZe1>1VgiPP0vu7h2$rnq8F7z9FutlLzZ4pl_)_8G1Bu}H|Cbk8<dvVD`QK-4
z52-^=cQ(#XW~Ky&eVbDV(<iqzc2%F*%2zBYL+_Qe*i(lE+#$94R#-<{&$v5lNycEc
zFWumB@uj#!x|r_Bc-6T2&!78yv=&3LaUFC!=$?`r+WD~Mj~7}U2ZNxeZ%u=gDU+k>
z=?)HPr}gVjaQefkLaR0}a~zXgTH&j2HdXuqLEnR4)U<^9ua-wJ?BO?9d+I0mwxe0>
zwcKlG<#^b%jKnOS#*2CLt5I%#tzb8!QsvO9@lIn&n?^RD`TsuiV#4(-H3g&>@(hwC
z_|M8~J?}yu?{pp(ytZMm(Q314+nKC6S{-<Qx%AXmC~+WV^_P`~pR)!G;q3!!hdTTe
z$=-(`5AON=p98D@KR+?|%BV{p^z!q0dat>lDMv%Ugc#-zu8s3l7K6cy@P`z2LgURa
zGL>>**nGQCrbiz<zi^8-*Ss27c@{s%*V6cBm@AyUc$}_t>?}o>*>Sd_ARcxhoRW+X
zPV>8C-Jx)lIReN{myxYh=KtQGU9FLmBrt#o)x*Ki241djW=#}hbSCX&;*$8>C$CpH
z)d7!O1!US>7Q|S%2a5kNw&4t4$R4tv_L&o}8y!)y7CSW=9wXh(j6M?gZZ?*%Z!<30
zcpf%HP}a8_=BUS5^nnS;39nF(N`F;@6#k0{^~t0BN?|2qMMK>DH~uHz)AC*$Oc;?v
z(@EgjD5q9cXBrLrnvLZ9Qjp^?->#mgBo)*sbL$j{e~N|{J<ibQ;9ry$Ps=LG?~hKN
zF(Npz)>H)w6I~nVT&MDFPPF)mJLJ5mLX1`Npi>!-2cOY6$SyIw`~bNF{#0l2mpAeC
z2+pw~^oGNF&Oh+Ev*s`p(OBY^<@(`YJu_k+$1Q;_mVC%Kep2u-bHSy&rYcNhQfI{z
zI01;4yiWhHRa$0~@XmKs=Ya-7^HciNSrC)1#vUE$sJ1U4Im^p_1zat{VbzQd?)lnc
zZ;X~tcrdZ)-SfZ|VqW}K04-LdBK28!j0lb+haD1ln!W4Yhkr>I6toX+L<+Tf_O<tc
zP&Rkl9yx+qMtg*pDi`I`C~Qq=EQB)ClY8@mm!A;^gZSFoS7VzYQX4KRaCauHL(F9P
zU-Ht;lib`8MyZ<K3KcAc7FIs<*ya6?ZEcr?4Fp6A*;Gw~wyJs@dlKBW)pXg$G*c9H
zxM<xuXkHj3W!p#qO<3G^7VeLs1UVt(k>e+$o~J%@o<0VE9|heH21M64A}M}P0*3+g
zb0_RZIRRQJ{^pX5M+MZm!HXIGOWtCejqpl7MN_M0_ULzT;cx!xjP-A_m@laISkBxC
z$21=k;pK5*62`0Uv)W{=(vSTb3_66_|Glx%lANMuf&VaK@;tE1VZL;v6jA6s;eg=R
z=F^HDH*%;|7;*Y$tk*N%<Ld%3eLc(w%hDN9q-h&87%fFGq`pg;29r}opuCX^wnbP=
z(bmH;YN)?z_Wfrgwa!$BQYV$*0MRc#3z-W&w2wv#UxaNvL+3+GUFT~Dyh@dAH3!H&
zc_Cgx<U~Cl4(3yruP7RKL+KSb{m3eFiv=}rPOs#m&07OkC%(!bFN3RXJt4+%%0^y3
zuYwf#vn>3XT+*_I`Lx`i2VTzY&;PPT3;v?yD|jB5@nbf<KDlW$?jWG)cYF;Cx3<ML
zA3X57c7ABlrarJ=Wqh$c_oPI0@VV5U=5RJGcyb4~NXm!$j&!0rP~@eWpUM$V)-it>
zyii~5pPn2?!N+RqhD=O|-8aUnu1p%9`c9CU=}FCli~Aa3#1Ia#$Bl<x9jxr;C}V)(
zM0PkQe|2kuc=8@?@#c@%u5!EK9ZvKoR#-|>BOk@>!(-t9l9A|0=%lbuet?BFU9a}3
zKJQHiE;!^gt_C;U@Wp7t=5)i%ZNtti1^Z3()8z$4{V;Uk&)uw7s|+fZhJJmhLGKrd
zhS+?!vSjlpe)s|Fl#gIwO5gw}EJL|?og^Ja;brstOb!G2dKeNl&f7b9IJ^b0!m=uK
zdY74cNn-fMj&qyFj_t!s8KcpQ<;q!B&3@xgEulwW0#D704t^8tcF(i2;1z5v!}M4$
z=AGbE3*zq4^fezYOlYnKx6FAOzeXH*JN2}4g_={pxuLErZp>o4`Rc$Vg-4lUvuQ7D
z@<Fc>+`Ihew|49F9COWi6RO_412Mj^gww9m+m)f>?)@1E1;fWiYr0ua;riVodb!=t
z9~-<IN@Ah~WrqoJud<>`73sbf+~mKSi#^T*7ojcaqy(cjm-T=BIthj*(M3|^&XRo6
zALhDh9eJ$3mF0gjduo)Cqp1_R)Y*w&Kog0!p3T?-gVRK)pxa%V*%q{2_fw``_Zb|&
zRu8UPreON2LYe#=&@t?vr8ZEpKo0LuU&Occ&xf<rsWVi>Np3xHu7|Q1dlZJn`I;|D
z<$gBkIa3EbhHKl!P`7VYcGsE;sr=Y-$6D(<$_|=?ukbptRSl+GF4gq5#b${1SM)l^
zajXbMXW;||Q?GZQUr(0Pu_{_Xow2$}Ik}d?qMa2Lr4bA#O-+B<pq#Vqujln+<?|e$
zA9DTUyRs{fhBXHqZRhMAn0uElYU37JBw4*Vn#wV_sm|Q#u6Hmo!m4t~Ue>M>_ikmB
zZ!lj5;pviGxNt9hrp}PT%1fEMg8+-d^M7x=@^S~w<3|*>mxi0bz9v=pE7A=T`I(|L
z0R?|FC2VTqo>~nv&WMpoV^KNB#y9dJIJ7*=`AOQ(+FI1kZWCN34@>DG?tt@vw5ZwF
zeeb7s(vyT@Urem&l-n~Ah0L?8ech#PUacGA{wX$Ss^k7gKO~<jinZl&OiU>ACGX%R
z7}U522;eaI9{4$M`~DHHzT+n7q)N;KS}ER8`~@4YuFr=zo7nsj7(RnFw?LcjUx^9A
zSjAIZw`f;P9f=V#Npj>6IzK{K3cl&~#FyIa+?km@R8`fYEo9LrCvwgQ8^Lv&ThE@R
z_L-ft$G8gVOEgWc4YQZE5ALk0jG!mu>aC63R37^i|3{y2-L^KmAC7Ch2%op>|7|yz
zG=WC^`a7*~U82N<Fn-nJ+wkv=^@s_V?(I>dAuRY$SJ-u$5(Kh%NfQWfvzrSQr()U?
z7f)ksNpnBkvd<`&vqTpK{`5Jx{`W@cj^1!^SWYW{ZGHXxlf&)ndL}H5XR^xfWSzqv
zW82#n)j0jWA;T^6#1rAmP9Y#dU_8%TGpKSBFvt6HgQGhN2R~HPu@c#fjc{vBY4l3L
zM#*ta3Ccvty`8x;?zly|hHMyNf0p&IPM=t&$h*wgs%r(x{@ci!q&=Y9(S*cvmKO~m
z_5}MUeuO+_Hg^xQALTacwbU#QEOFjX$@bvkb%ll+hA(X+QAFZw^YCV<#A)6J4NNvJ
zo~eWKo`et8d=CVd;?Z@tCUhBODQpC)+Nr59Gg8M<iE$&!KAJKeX4b#AxfdVm8v1xB
zRLIGTC8k`meN4_<^rP{#y{z=JsAqu*2qZaQ`d%bOy(RE{EQt~WiX*e``pA5+s;Z%F
zt)75jVJ-Y&Z6>5|R1hheIw!w{iBEGt*#-raub(jBh(u!0=^}uOl)r_0!5Qqcx)>D&
z8oAXis+0D5Wf=T!)FLZlX$t4iv_5LztUT_Qmpv%I_8w37<G5OEbcRq2liaO$E^$<h
zz#5VDo(=f`19JJ#LK(a6<4VM$9A@$W{S#Y3_OAm+tgrYTS}jT&LhnS<ijngKH-|KK
z#UMIF`zNlf)$^!hD5^3rKZ?QQRVpTqNRsa2RJVU{gg4_N2@Vgg-^SHh+D+_E%v0qi
zmC4ftu7%N)A39T)LgYzH3~@p@r~KS3O+brer0tC*((`aM!1;3)I`ZSQjkU^s3$3!Z
z32s{vUR=;Yd)HcliNH{ygbhiABJN&=^F0vsal@wZn~S6ubQ9TXd@JVLqp`Dw9{`R5
z7OwFp`GG2Z)nr`sLo)DW9_Q_cL@2*(X0txXc`|G<!>B+9!2)EdTfhX82YIwDGvwzd
zc#m?R@Y6o}fTX>_0eA^ZRLYyI!gfdTM+Olzdo_nOQ{U7&gc=E>eJX8a52&BExfjXt
z7SVbOjmkYVMkK@=1hdVF9G+K(8g3%R)1jD3rO48rKG{TtbIr~NZ*8JC2<z1u<Z=g%
zk_<v$HkWfn7-`ysHFnbm87@wJqG3!1eun~&;B*|f7=U$^!$z0G+fT9}Bbv#5F>?St
zMXQdalQx5st|+nMXThOA-6ZEuhv94z$6nH(m8co)w(xtI3q+C5e{WO+U5D*H<Wd98
zbF(L!5PGtKv1**ZDc`ju*M@@JS?%ZA^s!V-L!^xj{$UbhjD_AvpHL<^jj>S}E&irS
z;^%gfhd$DS&p5=t*LVsA?HR{JB$Ol_{$M=@-kmIB^zeo(=1HIpecCjqbWZCcZ(7=W
z*F*E$&HKH>f-jwmwU1Q8*6V7scLh5ItsY(iLYw(jAJi+pwyJL1z2Kjpg!g56(~lNv
z?+C@BP290f<(1#7VtlW1yQsF^Fo0;^v8;{2X5}beC-QbR(QWqoEKk5u!kKzk+VL~E
zkM#6rwk%M{qYMgSM$DWj=zRh$%e>Nb%Yk9zAylxq=j-7vYoyD{V2417*=uxZ=u{zj
zxz@XM$uxWsuGE?rn}(mW2T66M?M`o4Dk?y_M@9$@m$y-0m<vbek3&7eW@63CCn<<s
z-UTe(A?BW5vw`c%UabmjF^6pF(Zs{lj}y!8R}a*_is$>)#g*?&_&sIkWsIT9`ZBd)
z>NFni7wz5>HMAWnC`^^}=LM)pI5s1#P5YDX5DJ_?RXqJ!K5e#(lJJXuo$b+f0+i^=
z#sX!dT&&zcq)%i&WC+LW^BJr(gY0XH>zY=&<cnz;<y{4(&agR(*;W!2dqq(sGe;q&
zhoV~TAN!EvwiFo<!MOyUMZr8hT66uoIh1B*dM3V7T+af@&YVm!Gavs_mOs95sdAic
z>CfRTy$}_p5i^T12FnU<N4U)WS<PcIC`&iw*+C)LM72nPSvyLTvpq&f%n&+a)%~95
zZV-c9mxuX5phFHe>pPF<hspVP$`}~MYtwnVqW_Ey7VM&+_oVZRhU;iHZWXyW7&g){
zqVhpMStF&kd#(IoqEmw&268`ia3O+)J^}Fyc3V3=sMp^op~)XNQ}mfHwNpLD>$*RK
zlS3k@8yO8Tn^}y5H6{Y~e1m!YLnuS96@Z5#V0kB5c`mBPsP|=Xifv$<D~EY{6krS5
z3m%P}x4*!|qu$qJyW$*i|K1piz7xQ(BIpdY_d8-a6c4uWZTg-Uc}B9B#~^>$=(h9f
z)O0~@KoqyJ0?gryTUH)U--R{c@|qa_ZoSyV#4zQFCo3Qz!8_qN6rqdvs+w3`OJpO7
z>GTYql8HSX^4yuKw$4CU$?rmhw<?B3yFUYC^q+yNMuQ@0ARuL$*p3eZ?x2fNh8DX!
z!!VG-e29T!JssTu5E%1Gw}xEk7aeN{3ks#v!CC`%o%hmt$6TF<a{K(6yPtWwf^4G9
zsgr4zIg2y(YB)BfL(k~~D<r33K2z*H4_FGO4ASiLn;!YPIxma(V-PcB$m^)>pv>(&
zLPh<GDF0HLbesElC}W*qa&p&k{@V#RdOXVu7Co#8Z-Q|Uwm22lNVAVuQ+ubaAyY4&
zxH%P%{>VpGJ;^65a+7PGkD(|lN~$592OEDJ-55l>-M<5YfSQ{;?c<8+m!wwu`JAXZ
zRWrXyR;xNL{jtqG(Lx!vS$l-yIzqyz25T{GK4G|Zaq`aSnZwi^a^Y2yak0Zew{}W=
zLo!1j&1AU|@GOJ+`Oa9ncS<>w?eTV>l*NZFivb%)GG{*gh$vrt!$(m~D82`ZLk4x0
z?=eXmS55KJlYfd&H$$7oF?iB4H_}-@)fJ)zXU_0~Hb-5((O4h2D|Yk)T&YaD*{~LF
z^QE~#K;#9-tt4Ab*KaT5%lozLida8nGh!A5vZ~Wl4Xc654`I^cx7TF-3o}C&ol<9q
zB{K2l>V;`U(t#Yv(n1(yaV#rNo#s)WvEW^<sjWfn*oMjpaX?*ND?nA_FLNTfOqd`b
zryegY@1{~)dh0Lv>QKAEP=T7mcA-MYJi9mz(bAPtCgHk1zLKZU@uNa!N3{2gWU{dB
zFknqYU`rlj;lTV{%IE8(Tcm$t{;TR%hc8-1y830+<~pJZFLZWK{=HG*=>qZC@u<aZ
z?<=;9!-Ohzytp<K@w}xt&gP#cQf(8>&FjSgKyK%g`=fKgpyiis-VWydwMy3CTMMHT
zGH9Q-x_=<J5}8;iL5=t*mD~z~c^$bdZOJVzMu|m!k=gi2^}wJkxCd4rfp={nh?Fmw
zkdF6_1iDUEEijP9iBFc~5>!)Dx3DAV?pC2ZrsoZ3nHNQ&dizV;KwPuxuWhybd9guQ
zcr9|r?$kEAiP?BXVNF2DshHI<^EF98EB05bmlDwHRB<Y>STAi9vfomTy9`)_wYh~&
zuZn-jql9x&!Mq5?0vfyC3V-?C&s52r@7s(jOjRyNHQ*lgdZh)Yl)DxvZu1LxSIp@c
z2uLJ0ut_yllO_st4LMg`YN<$&Wvxt!dtQE#MV^t3&@BpcVLE?(1`%4$hIDTzJ6UX^
zwWrB;ZyYLV+8bMo%4uypRbJhP`YPR)EMr}_mS?d^h&(fnZ7COs@0*BCTSn>hT&gq%
zE9<-dfhQ-$AgKrpw9Pik{nNk1-0&G(d@|0Fa|kiM%|AE#O2ULW<p-FJ@AV=za)1Mq
zB%R*Wg~-pj$D4%F{MQ9;7LHB(xjDK!RNSKOx_iN=;vJ^eveam>*TY<EL*X)0U{Hih
z+O)Qj3_%SXJA;ymZ@@Ba4s3gRdf=yRUrSA{@U9s-OW%$JQ{59`wk638ceSw+7qY<?
z{ENTKOgZtw6V--OT*;$N-tFH&x`Gx9E}m9`n!hAZbPTwn!UVPOq6F%+6&SM2I^mr}
zmJhuKqW6<O;XVnN;$`(~@*$-X6$I8eqr^v@5tyLpgxWckX}hkK4%LyFo#PPRhIVob
zqK8zIh}1krclf4&6OPFvHM@k0M?|(B*~pyW=_xR&G>0BBa&1EAu-lD@srP+aD5a*K
zm-21}94nz8tm+}LtOno#hP)^9KZZq22Kni}Ndf5$T2u1yR1+vMi7lU-oBhg8{I?io
z>eb3r&=f%I<ILlPyE67K|AY4wNL4kYb>xpe)0*zYTMr)?SlR@$_GE<|@u&Mu4eA&m
z(4{EGev8(mDR4ElaMQ)iw66DDLwB%2J2%RA>@M`MMC`b|$<eS(^&yfNVjh|t%xh=n
zWJ&~LcXEV_LS#JQ#`qjaskjudNm_47Z~QxAU_#a1xJMx8!ydJBlJKWV1`Rsx&%dSU
z6HB!}(@nJm>?m1R5G>d#=i=QK#pZAy8Yc#YSnL|oIctJW7Ph?DDy<Shan<0kGmrZj
z9ww#TRux2hj<2CY?nkVHWG;`&4raF@7&x{*2%GaHZ)+3>l`_mi8I3GLs2ik?D;=9q
znO|_#e}cI`7nuB6j}0z*H=`i{s*7g*CAXhl*-CiaJRgrEXk!s6AMHjJ*&_FDR-jl)
zfrV;Jl%M-xvBZ<vhtr@{-D`~~k=!gi<}3rVF%mkwxt9SMQ{OF+Xw1NXVdU3~1M+#Z
zs+3c|==l~@E+Sg3O@0-L&UXHJo|y(2&^5s-$>f3K+`XqQJl4Zp`l8ivf1Jlc(30-!
z{c+iUd7FfpY*`ObvVP{jC{By$gvFG+KavDGX)6ofomf5ZTV*(yS~}y3kTqZ8K+siH
z1ht|Yx-Q0Z-aMNybt;;qUG>}rqL~uOzxj284r`fNj^`Qg>BS0aSOw;Rzw~BCc%ZaJ
za+9ZxEx(J&4)o_{o6&SpcUJ<S_kV!$T@8<f$AhxtI9O9Hv~P)bRv_!$vpaTijcKDF
z^VUZC)uguMBtYN9J`L~RQR<(;7)9!JNYdTMwgo6i;>t6Q;M!n;HFgbI#5exO*_!Ff
z_e`!bF}I)W36)pFSqz<G&(t4=ka^&+c_A7ie$3nYkCdCP*XYU&JP_gY!%RzQj@v&5
zmY+CnQ8KIEUYVJ3`~pC180igNqBj^p3<3<B^?Q)Jq(m7@fm6@WJq^-fCV0j9^y^4?
z^>@RroJ6h$K{MZ_)m|-2;j_rs#_>Nw!?9i_-zv*v0X+xw<8|w>DbcTn<&x8SCamc?
zb0wZ7J*P#xJ*9KXym~P07c5!@eaD63n7l;ZThaYel}4y>(QPA*UXZ22SBvyn;O)iy
z6y5D>Vn@F0HDy_m#K1t-83vsBk9mJj#fb4`kMNUx5SX_7BcJ#v;W;hs(l1o~48#Kh
z^2zbRW$)39RE!zWqkrYEc8k*VK7J7_L}=*CS7tr$Wu|i)L|R)GP331Kk7H(G1~SCr
zV-O&#y2l^l!I%HM?1_3PXQ=ehrP&tS0|$k6N6B{|CnO9`#o<PFaU|+_%;1dg2@4;i
zshDZ{tYPf@S_Pw{0B5b0OU!I5PtD4CaC+mE!A;GKnx2ZLsx2f!*#8p==b?>bH1&5Q
z;XH38>MT&D(q<;&Dbv%2kSCB+C_H;afWU6t`&71Bg63|+Z^O5a*w?yM(|)%uibqa4
zT40&C-VghV!Kk?j{)n7WPd$^Cx$L%nP35+|v|UN=-Rek*3Xu$g!3H3lzxS8K5d!Lq
zVZH4#KV-hHzf>Se(P0VRx70&$_BnKG>eO6nS!_Q^b|2TL#k@7EGFyGVLLg|X37U!s
zE&>`Epxn=PiOHEcBnK{=>1iD$S8PD*$zCPI2+fOj`r2KCTl-u{+pFEE{-Xsx+vwiB
z^yss#>=2na6>yLS8CJtW&3!zo@2(Fmr%Q#F@<>U3=j;z3Yzik5|N3YZ4A9GSx2UH9
zky1_&n(lhk#nexzD<kDcEtDyp?IT1dUYSRGW|}zV1lYM`H6*axat;zIKAPr%tTRV?
z@r_5x8M;EZ=y^PO3&bSr>SnO01?xYM^(}}54g*z%!F5=+PW$hDFJ`Z~_YrI+Al??X
z%Ah9e-)Y~L>qy%%ibJsJP%ahpAL!}wpG}4qEYwVXQN}1Ix`bu8%Dj>mM6jTITeh9)
zKgA9f%HE^7hFoz(sm0Dneys_L2>H`BR+y~A*X6vwS=Fcnnz~kIhA9G4S1qg<Q>e4f
z_4zpC2Y(s`UtetXnwtlrn%(9uNkkpb9nU&L;uxT04?0pjDm43qV&k*ugI<s`yZ*H~
z|JQ>=jW_1+8adH<VN!Z!-hFaLev2ao+GFRBN)<X`OGfC~uGt~|`5}mI27SI-{-=L$
zd<CCb-v{VBwtyp#o@}u{efFnyhp}1rkA2|Zw5ll&gW#p#>!7GPL3cuUhSTP6eumVl
z!|kpDsjx9{MqS3P@}SEEtt<)iF`8V;{xdZ-WxP5)JRj@=R~p$Tgc>u{)!5&Y;H94L
z-icpmUe~HyYDrNTD#&2~l$pRznGBm<xiLD1%7&;Sna88O`MO^c2UbF%q5MOkk7s<9
zS&i98)`Hh44DBk;;|f(N-ZW8{kLl?l7`vmuHJ3x$!19n@aH^85lPvR<tNC>Y3ye-Y
zwOkl2346Y!Z<w3?CNV~SYiwbOTQj!KQqkJp%o2@F;|N{aa{f_Z0Qj<D3_ayv*<iP2
zrhj`DCHO!T#)R>RqB!q;Z1Bu6{!0D5?~;~wBm*hL>XFcIGj~%tcA$?BZR$~%bA)3a
zjYm6Mf~5IUx|@Xl;g~>F%mS}i)Q6b@`Ps}?>y!?~r`9LjU)d9+8rmL4GBB{z&xhBV
z|1A=L*#T0J=G(7@`GiZ2b@6^qongNd$RUz+_n?Vhh<?0y1rCP}Ua%A?il3{&sLB**
zA6ozJS8^lwrJ?yg&-~G>N3NP-c!sWgC`-J+!<tENTTH<iSKgGnFCC0#d2CJCQeYsm
z%9%&1nqE4EO6y&cIz{6EG;o$oa-5N{$rcMTa*CmTckVViBXS&+1_Xoj&`#K-Qoez9
z(Oy~q5mE6zC7IB_3EC~&0;;8Bpi(m{eJeeLwf{4h$%=_BYEvUOEI2tQ#{!9(CPB*!
zW)R4Um0}9U2fptNc;&n<c&pB)^Ly>(bL0xPBetq`;>p4_AE0z*C&!QhACh17oTt;L
zVZC8z*h+RHpAQE#$v4+JC~>jxxh~pME#2DV86_>`cl$E>8N5U@^Ir4bYNO^A!p!rl
z30n9I@RTdmC1Qi#Hz?12>3x%;JGRGw`o+P!6|Hi+{&Q*1hwf^e7{Ko10Zmn>iHY+-
ze&k%RVMeZ&_GOcjz3r2_YGgM4PnlcYOSDPJNL$|sRa-RkO{Ym$J4@vc`3%sb-qsA$
z6HJv*uvN#cxiqfQjJ0Y2!RVmU_fNWYc8B|E{ntBPXp3xT_c6<dI(>&9R<^|gj^g4q
z75Q`&Q5nKbOKa3+<{o@BFZ8wkl!No<dR);*MVxE@<gQWU+UV3WXB7QC%t$EERz+po
zX>f!`gTs@ZnbD47i}?!0fM3m(5_1nnYKb`_k>a`1q2_xb{AE@l-x2mRyI?7kpoOoV
zyG;wyRenGCO(6N%?6Mfm@{ih)rUr?@Gjjol>lL9|o53;V3Pqyhj`GC#tPr&+Ff+go
zO?}J611&S%2RzHo&fiRsNTri%bY+<n!P1UvPTSkAEy|8KwCl^otgOs}?hdPECT18{
zcUq1Aseb84>`2!g(44Z|UdGsZ+6^`giEB&y<hSp+hEfGcYxcBvSz+3>{g1t6Tg}@V
z6>Tme+IKEC8#(0=8aVsJF19qUHic3+aUac(c@orHEwtr^>5TpmstT)N-psatpw%bg
zTO+ZPXPaW^$|HL4v&A|g_FHmgPf-=_UYlZFQylC|M~wrY>+Up9w;M(RAP_Fuef&-H
z9PmZ)1G=gyF&Rx2wV!ch3UnW)py;yYZ_~Q2#Urn5PJ(i1&qofU8jl3$^837+E&M0n
z%fa<2k(t&mHSq<pawEDRSqr!|)z=v26!G^{(S%`8|Jvh1_+4)aEG?^fhRS)UayW~5
zhl0D$?JaASTqLtMKq9?rv@`DB9epDA7tti|N?mP0*ja_zq!5G9Acsv;Y7nDv4>~P>
z|AbR&64n2(muRChB7650X12i*F^Q~yv$=>ukr$vIrY0s5_poB~iWkj8rv{U*(vq(#
zN40GlQY0|L^f%z~;Ue>ii?omF>%JLeNM%o=1vGxa=y<KiBzE5WQH73*0ViCnw>5qS
z7aqeXHS%o+@+Bg{arRM>B599zQGMxf0KZzp_h(55G-)nVJFZ!waJx}~vQm#_dnLiv
z!=j+s0%}u{v23pbq<@@L(`bC-0vYH4TB-&1&E^SDSVlt^z@Nii+xZOqhO-^~Qf?xx
zY`PQTgK3>O#vSzQ)O$ULWUSGrM8gZujI~CdC5=5<ZUeShjVe8Y9*+s_A(#HK@)_7P
zu)4H^Yl{<*kCXwoEK`^?it1@Ty(gwJ6o4f7Ze$x~)b*t&=X;UGRi#Z^-|EFRyA01M
zv&98oNyq#mbBg`;<Sq<@Oq_~0z@yGYIQUmZWnvrBOK7=k_QK3#FYRW`91L#vvrjo!
zkU8K0-DuytaWu>g1=N0~{EVQPIG;g;R<DRkO7KAWmsbT^UhP-PNIcKe9ep!z`JwFj
zBmr#|@Rr6sX`K*#f<Vegb1sXRw|-+{pTBOW<N)Qf1pc$~#|}lVvvIoe+sn^A%nfXV
zLg{Yxf`OdEa+nug?hJw!Re{#SANlaaYMJ?h@`)XWN214tEamS=w4%DyY2ik$Ikdf{
z6?kDzcC@dHyodF$hQUvoa@|#jQaHU~`GJoTg`r_0u$|4R^s2;fs`h84)MWN=8+f(1
zSr#{jgspLR>wU@8*4XjB^OwKsZ1|4DuEUjr`PM<-OG_FjoiB1aztN&;LH=oiW`PAz
zCgx}0y=*H|?bCM{;uCcfC|KJ}6!$DnWAiFN9)VNcsl5CuAC-90vK#4G`+&gQR5~8?
z@0l`&EgFcW)qFD!T^lqP-Pa)v3AF+QZH!`bW(^jnYdZq4RKq#+h5o!j7Y2$oeudUz
zX6n=3{26+t288lQqXI52&TAC{2sxTn12r|TpnN!8GYA2IeI-y<KnzQA({*>qxoEjX
zT_Wd6&+pi&4aB+Q`NS0g4g0Na<YzU9`i5#B>e2D^2tKr8GP53?CjRFQ-wPG!u@ugU
z5=%V3IKE(QN{gmexLYrr?o&xGe*DohNz!+e0Q?%nS2fWIl;Zt*tirn|Q}X>^5}T-e
zXey$IXqX_)ed|S&A;(ZFb7T1e7kcJ8j|V9Sv4of~z{AXEgTe(dnN#w=9{qbm)0N1=
z^7W)kF*ZlL0}>i;m9-aY&+9>1yfLbb_WEYYi@UxF^j`i5+VR%W8jOS4yf>}>Em)l?
z6#SX);g*Cuu;i93ZQiEJ;|bW71&RPOif>>utNvm(ttgEF3n|G`SUDkN188@scs$D(
z_NH-lXrBN~fJw)awUe_Yq~R?i09jYp>`FCGoNk)Gz;$FVcvUnod#<QB19+UoB4c*f
z<?ag|EL+yrx=B9|pxgh+`fhLB=B=i;a}v*@Y)kBLp|RLMXZm_nLShCnbM5wLIunPT
z^nJ{E_`vG1HtLo6#h@ndiCxs)QGA+VUKJm@tU@Bb!BlbhRG)!c3+#Rs7=7tc58mpW
z2#`KuA!q3J<>uSnMV0pH_3EHx@=~ZNdbMk1tJa-VN%N!`TbI+?gp*y(h93IdZS-L(
zx6Wk%Cq=z<4PVua*I)whG@O7EIElz_`;&REhTY0s=SB%chX2AsTVmQkUam)&7g1Kp
zn740zHGwl`EQx-}%-!`>O}7j?i$^pdX$Zxj^tuG`4Dn20U-B#^=Phk8jLL~LL7JNw
zknafYL=U98?v6o)@f6!ly9R>Ppk7~|u0Ad6iIXNcB>vbLYcfL&LgDvZ7rIUQ3nK<b
zm_B`ZW!b%br`@qng#P}TCEqG&SwNZNa(=yHq1Jpknb=I|oZ_FN&`6Ry{H71h{3XR@
zdE9trP6&jRQE_k4F%}5R#<sALt=d>oXVBl&*X^lRr6CAxC#Ah!^YfZ4hsZPyl1)-&
z=Yk)f#HKol1TodWd@+9)+GbZ~VjJ~AA5deKRw>%!@rws_r!Y6>rW1m{QbRX^#@>p#
z9{OLHt;B26`P$WgILFIz;QZ9+p3T?8^nR3GlRK?AN<B^hMdU`{s|<kQvzvK-tIAU?
zEBlDE-1D%Hwj4c#>1+w=BwmocdQ_4hzsj47ULn8~k6FRf6Q@xr;E@zO4;+kFv#R>%
zs;Dfu_`Hvy((#)zxDH?|u&AGwBtPe-2^L~;<iC`SPmD$BL`Z1_wFpI`iqAB2=L!sl
z9R+NKS{=KGje>=3q}*xug<8^H8Ybx3gov3o_;V(3&dC+)r5Tv<WuK)N@%!^~>0-{$
z90?eakAH*8AG5ktIWe~YvhE^=M7?O97{ojWV_Aea`zCk$q{p%~f!MgZ^t64HwnwRw
zRcBf^>&x6e=Jbqb;jEE-^MyShL?=BOdGXXJ=Ek6Ox-aTFi;mlKPIB;rtQ@J<rBSGh
zzhl;j`ACNujj#6ax6zWdKsvE1x!^}?Ic6fo7dX*O{ycy9JJ*cFy`eayY)Ox%E1=|k
zDVM5pVW7r6V@3DRiUPUyT0IhefgaxkXC6)FS;TRlYqO$O@3vz8jF~M+F5olMkqHX-
z&|X1_@h8!#%lo46IF9$e9cHvrCg`zB0BG>>=CJr|=aI6AKviS<Lsw|2fvCa9HY=-|
ztdk6SE6Rmm$t-K5WRclOBZx%BK+)^Z?LTGzL!<gliZif?%Cl`-)JSt$5IBkv!}F9r
zB53PE2CX|O3<h3>RjN|WyP>C}{Zz8nF|DfcWsX~gs^atPubH0c#K@g*JgP8Z|6|U*
z>XGr`674}L6`P)>1bUyE<ydy9eNlM}hI3|FeRcooU!}XnbF%>{Ru$&UQk&sBf}Akg
zN2tU*!T8TppglHZ^7PoWO#z%_4dpDwls&ZJ<jL&%uK9<s*f?y#qq%w}>B=bcGRM%M
zx^~c@+TycRW}nG%0<gG7oP1R=V7LCms|Xm?6SF{j*|;*t#~=z>e<U{dVTDDhy7kli
zQQR%v-YPTJ<6QG8aPcWI=C3X(OO$-K<B`+P?}c)l9n@M_CbhiGoIjQ$qBXV$lA*E@
z9<!1}S-=9?7M`QQrKAwNzE=0ut4HZgpiSEffNR9#bE{8TJqb-auGR*?2QzgO-kFp_
z1P*|Yb{<>XO#1N+#hz8gz^-alX+mD0#pA6IdA34hYs__h4__;tYxX~Qs}Y@qd{)(O
z)fE;KyPa&Q_UDd=+8>DK9>_vOR`kGr^S~hO*YqQ6cD*~oo~G*17fV*AuL(=Ghs>#$
zg1~3`O1REuSa$tk?eG0M!s3U#xVecp)BW-mhh=2H==cZYkMBRxqRV@=j2WaYshfY%
zM~E*)7~o6^`HpR3vq<-pF#X)9H1CfZbxQ-+qfMPt`GzR>Qy(Us&%PulhUx@D%@eH{
z+t)89YhL;%r&9_tRsucQ7(~XB(d1GmPQ!dEeS-}4G@~*ZNvjU553x{ED3^`_$aSl8
z|GZL<mre$;JJi3&cFWPG)=JsXP*Y1ce}1sWEyx*aG50rQUM*o}>EZK3%6}R*4AY_y
zR@t!D<(Hu#mckw;<u-VcGR|>9d(FFg_$^(C)@!OS+s3jF{@CN>8WLlgaJPHp$Baq*
zhIxTUeP0`Ts*mI?sBWgC;<Rf(VZQ`ET5j*Ke8Kd-ofZAZy_&@-xU0;BBmy%O(<YGF
zh2+u2P-LeE{xn8F&ACW85L)W@ReB@Ae{;%i*&bqRdAwk$RJn*NhD(g7j27Ix?}OZ@
zap>no$H_%mPo~_>3R%u<7OFGdtOT1k7Z$+zsXzASsO#!<^S|c(uI#$&xmzi;2Z@)=
z<|ynxV&-nKc*}3?GKIax5jI&@P=$ye#^1Nbaz6sFK7q|6bdl><$^{XzSBnC$yW%Mb
zp^SztFJ!H{zi7?JPB;NZ&i#2WXks)cZ`v|~8i~a%1kJu)cHN&5{R|haU!$Ukd^@WU
zJ=|@rTx5}cWqsa>jH?mYFYwiXRnSv>Gij=dQL~=*cGnjq058orZTE_Ht$24!^gqe6
zD07dlMfIak&IP;kdL(==%+lkY=q1l4z7}4$g=V{Y=n*||_grU7%St`xN{Y%j9%~WU
z`<(9U$taqdejn`f%zL`kmvms9>x4BSq|fIUMNOF4*4BFRqmS530d-J%sLz!2;@^j#
z+x0K7S7xuxz7;H8l!qpq$X%2KUQ4dwNFuAiuxs>0p!Kocqm6%W?8I_?-i7Q2@bAH{
z={k<MMjd=M?~!gI{&*=+J_v_h314S<?wn$gs`${HV>lk)`Rjo5bmvq#KDZtKoM2<i
zaji3FvO<iI|2q!B6a|*FUE80B)z*ke&j)w@ZSEv#&`bJiFJ9w@_p0V%=B=Lfv##I1
zE8$F)@g5thn8BalaS#q-0NE~5tmICj3fW*T1SbYEci%#p{L$n5mwd<9G+qZTFjbd}
zikp?dwN{jG?zF}OquvDRn2*mH3o*cW^SeFH+!n&uCkLu|B3VP5l$4E{+PyPvuQ1LC
zcf3YUO@9I~M$g&?Ca*gikX*Rev6Vps0tI54-KuA;zmNdZMXzf=%ug?EhhVSeS!bu2
zL&1A~+`B-GaI-UgwcO&7zIXnjIC@{PR55OPMU05=BBo;)x6KX>PS)2u`-^*ia}rzZ
zS)7DAc*h)K9QFQYF{G!@T2sA3*A`ZW{vJ`3A>lp_imd<Mu$)84SN*uM-n*th!W5K)
z=T%B7fa;|h$Z~a4Q9|$Q%W_7#o}@VsUWkxneZ=bYG#qCYToO8$fpL?#)Rl<-+ZQ50
z=g`lt(<I6L_i1pJdHcao*`%G*Y_qpU73ucKYXOdxv)A;F)9KVtF26#?hKw(;&Aqdn
zgK#<)vIr$sNY+T+rHzSOe4h8Q=R+rGLPXZXvK3&>V1Y`rjx*4`!Vg3v8?(7wii+F1
z0<?2~3#1Cs6Xi4eO{a}Yx9+?fY&Jw_0%5O`iFRABf;^T~Dk{plqA%v=6dN&Uc->R4
z&4B{9+cmGkyPs<1_^lY;en@LmzC2Sa{1htEY<u8Ynq74yHkRex&?Y&j_cFl`nE#%s
z8?Lb!K5+dhla!xj$YMcaogvP@G128DZRe_Cs$cWp!hhMVG?7@h6`b}LC+&XNR`Kmy
zLD@jJGCEMBlA9>VMU{zQl(Xi*x{GDyxXWNDAoN^VU*aNRouP)+I~dZa`(L~VQL0M9
zd*(g1D{2=P!wU2EM=ZX2wYQ~Uyu?3ysYa8KX#d&H@8<Pvr{#Q0&D<0GbrFfZxTLyd
zH4D@C&8J%cbGOf$ihFyo-4n6h;c}?^yd{H9zeG{g=2^ngtN5pie40U~PGw}S%0nF5
zRf}Mi@5#DLCohYTr9IH{#f^NWeiO-QF6$q9p+a@Q^Smd)E<>aI<#RW;=&^y!lrMs(
ziMzjv*ZiH}^Ld))81^QHJjH$68Wz9qgx_$+In2J6jrp!d`&MI#_f`c1uw^hocy24b
z1rf$N9YN;S_w=5Jau&%o66u49j)mc$=MmaZ=@L|}>P)kfZr{_N{@AVebae=zXsp?r
zIiL)sm7V6o%W~Im>2$$%t{%ZERLFbNHYG^>-~@+ZJp;bUZJ+FbjtwUF+T8+v4tldQ
zpLucIf<TYJIpr9&GkPHFl2Qs&sPI{zlj=z2=p7OJysiG?eU+fQ9cV#~B&ic?@@g?-
zFBNwpfBJl3OTClMAaHFyKs{Bt7&97oY$6t0x>^Mewxk&Wv~uPSJq2Rx953L0BZ2P~
zwlf?MXe6a*e1>1B`TH~wcg-c>^2sfyoVBNt+_Cn2^P!baMv|^p98d_&q8sl}DZ(}C
zrfpqn5K#%RU4sq4&d{?g9Y!{`2`H5H8DKzCqfsw1|4L?f^r!iM>h<iuqnG#XRJJh$
zSK$Q20XO1BEO~vi^`AulaBV4hhe`f@DSyrSu;b~b{hckSvjP1X6d)c6iN9cHr+;D*
zL)sPcqgGC_*n6w-p%qHOgmf6l<@>33a)SwHDuMUPED5tN`QA_2j2cb#xfrfMs0B6?
zxx_a+k9fu8kqxHmQqm?+M%MqZr($=psJyJfeb)fN0t3v7fU^*e@gHm7M$@GXyA<!y
zHp#nQNK;;mZH#mBan)&S)9*~AQ}L3IDIXrVp{)`QJJR}!5vjNzR65I0l^f4Uh5TI6
zgk{yje95|#V3b}09!qtZI&#gb0W`G96ePx{PF!r&B8ykM8U8pW$^q*;K}|;0!+6CZ
zobZ4b2l$u6HUbnwZXZ5wy+hRo_o{1A9F=<?l=RAV#7N@%oxgRC@L_3kpZvZofi8MF
zpMM&WV+0O;z5J0Ed!<3yOEcylyZM={Xb_4Ff5o;GArmS$2{v#Ub9WQJI=0mM-m;!g
z>BN5me;PH~G@*)uMKS`19p?ubTpdKZcHzqW@iOO<F@98b(Ib*~;h5G<0!(Tf{T&tf
zzABE^kza*J19WXSak}a@AbvKP_buYnQ>FoVfcKrRJkEESW@mF4V}!SRVRS;Y2cF(V
ze$obEr(&)$M_&u71eB6-&=|(KYSlYm28ORh9irVQ^;k6~z6mGnJZJ#O^tKprSnwf%
zn$^-Y#EmU@@#_NTQ2=NXD4LdSi<Tr%bb!-Jw|Aa&N}PZ^Gg)O}U+Juf9xh30uK$I1
zeqiKx$a!i=q#D*P5GM+Xtw$0etozQE??;p56%t>f-<vj-M-g9vE#q=F*nu^Yah9Q?
zAU^OPLyOsv(jIY0oL@GyRo^^&-Zu{Zaf%c*!b>`BzBn^7`8Az#C@vRQNs%RaHgM0e
z(EK7m;tpy-=%7~y$l8V|klD2tM~h`z^}U&kdVABEkA#Ie)?nFY@G}^hQGrH_@@9%p
z$(i$oC(4b}q}KHV6{?!H7#zzLHv3{zC$@V-)0&CXL3LhP(SNQz8tJKb*`FnOtY4RK
zUNr4__QldH%I7zKKnHyNZZRLdlQb`^k*uw`>e}~Vl){t>D9y|Lo092kLDVnk;~vqx
zz#WNwp4zVk@|(b$lW>H21#B-b@J<}VBfO$M0`xwOM1~13)eOVtG(DWdmQ?B**U-_8
z4eE`e_=52oR=E$>L}*XFo22b6At3y8NBXI3B>t%oI}}&v8UO8tki0YhaCn<#sRB21
z-iwbkby+H!WbN;ZI<+(LRT~(@m&7<tZR-?tpnW$*l0_N8qoy*x@V|U4%9}DTL`Vmm
zU6fGP`!psNNxRf(fl^hI{Am)Pv|Hqyf0~+~lb4%w9Xs6CkitscH0>nN>`FT_-Mn7o
zxxF+4@iT+zW%U&ZhusPb+#6}Hec=8Z+5g#=xvU-2|GqVPf@K~C6w*u@HTZ&64&vc@
z=*pMwQ7vKwj_=JaiR5ccBVi<3xn9(hCN@(bZyXNUK?{ew%m*in+Kt`OoyIp&u99jY
zI(5!0hFWjb0d%77B{O07+~(4n%A%fgw$;^y&qAHg5?-}W>4==TXIA`ZiQXc&{@T5s
zQKJ`5mmbNNU7@iD;Q!0xJ{v)8xidP6HG)m)@mti*xla5*@#R^2CBEu?(`d#_0x4M>
zPCaS4do4)oPBIR8@=jwO;Nc%)WawM?4@_3>iD)Aons%r+g(Wp>b9oqoswIsP-fkKX
z2iWut|9e9Y42fbciI)<}3n<{CI{bayRa`E0Ei8WmtWjL3wlOkW2y-L5-zxiXEiuL-
zL*FAX8<EHN<yz6f0PxK<>CpSZ>xJswK5z|<WAk_ug+V6}JJenzHVMxO1XCq%x#WXK
z-!*<ffSe@%gX~ngdQ1)*l$&n@sd&XyN|c_ag0Awltbpzr<W%;^E6oVEn)<3e$TDcU
zX(uA*fLuv(b}!{6@7rD1@jH9Xk{L`5GZY0a9QCVYh7TE5IeuzkOss6~n6Ds1U~XU`
zWOSZn*5h0N@YJx|Ivls9aae|MKhVzCB-{~`Ge=HHh~MpN>E{qTyK8NjCAbn#rj8p5
z-B?Jduoa#g&_%RWK%Pi6(DC%?N`mmoJGJ-bsi9b#(uo<0PdR#6>mq}|ZqTY0m|Rui
zF36crwo;NuRGes+XxVK2Wf8gzo_!(9jLx@dU9V^V^&JVO@>PDH>-b_B^>wbPk&XiC
z1HrtKz56k9i#bWM>PjRLfw$z=RgZ8P8a7qUo@e;vKmVMtk#zh*Rhh1UH%T(-9`%^<
zr*~;<L&G592d-uPkz^AN*k5c^J5>n`aj53$q2YrO6U#WYo-F*-W5X229D{<Q7_$8-
zFWuBft-GFuN&aP^FFCfLVLv2P9UV|WRWrB!5iCC(K5y~w4d8oAp7QD1Ldn=4F9X}S
z2eUjJm5iWR<8L<c<zIeSLhDNBANpuG8p?O{)^9gA(=<}QIllM7BU<&7>C1FWI`!UJ
z|Fp;3o%~5AS#fvEC0PVkaHgWn@Hn=`@14A)?S4`W?FGKpztaBLtlYppq0x&~*^Vy}
z$dFo;7n@EHeRmvqSGoxlJEH{3-zVAFG}WZFeVrpK%KlAA*FS$>EgwS5mEf4f^8WDW
zM_oMhaqK0PPNh}u87}zcdiL;3<kbr<@r;=gEhp=$e{YN#y!=~eZ1jnYFbC7VDppz=
z+FL~L{<y$cHb*=l(`p)fdnp>wX?!&_w{=!<dT;}DD!SJrM#FYM3VRJ;%sIS4!eSTV
z#Il3O?Ev=ot~Y9pb8jh_p5nsb3)>bDG5EWmNph|K-e|2cJ9{MC%5!8sUU!)lmf#6F
z91rEXEWn0qzM1|}tP467+cWMF+dEhAeEmtb3%|@0-q%-i=GyV^4MBKBE@ge^G|3HF
zGb|y&@^Cox+GzgY8xr(ZFE0eY{y)B=a`QVj{&2~}Sih3}=Wh}A+JSYL@aU2WP`{G@
zKcDy4HcQk<R1*6hv4e9Z&%d8!iv}j{&9-7A?$31n*sDLfG4p~;^MdwNT<pIMQX)0T
z(AAC;cr3VUW8D=ES^gb7bN&WD-0b#_rZK?dnx#HsaKV1#X5GF@+%cEaD)Fz>(!Vz{
za9ty}twR4k_&>Jq|J@$3|GPaoo!?T!5(+F2NB^gzR4@FF)m8fXkgtDbpOSjjN$OPx
zZRppM|5oqwzpF?6@9H;*f2HmIr+V9OVf33ivn|-J`!l$bv;THBr>AF=R8%!-XuAaA
zKDiWiU8iOPsvj;HPb*i_PyZH@N*8GU1NouS*YSV-h&*Wh$A1P*tfWV4{ahMOvehP=
zi9fGaby}VIE4>tv5xji(#`a}>eoAA2=gFGb&bfkD&h^CGpq!Alv1OA}@!D_{wBzaN
zVnf8`GufPEo87L5q|sd9iVhtZ_gb*$iM`v*_e=Af=V9@1O=u14{_kr%v6*HyO!*t;
z<ju652s))+Y*9#Ux2-hD89siMRq{+VKIB=BogMI4tauo25>_W?L&`q@xjr9XMW)i%
zHIoYZ!X7vveE!NSgs1E5^Nq)sOnxDagcdspb$rBMua1sK|6DR9UUTdy`mG$?N|^ak
zmqR1|;ofW~kcGv$!yGeH%xRKRlD;CQd~@y3ov_5)Vlo3El{Ej}U^4esu@yHxJ=*5S
z|LU@L;oWsq>M?j#b-35#p|C@W@Z})exOG-PS@+`6siW>+W#j!BDj{}dCac#a@<A<O
zi9+l_1Jwi7?jerlk2<*4P{&6%gfE$mt~rHOyyOr}|K6~DbZ@px`uodYW@1Vb+o|?f
z0V10>C|$Q{_8RAF$8)AmlkKnkN``J!;q0{j7jf?$)YR6ufqE=>KtU-%I)~7t1f&Eh
z0!j^pVgd;enuH>RqI4<3QRxtRuL7YZbO=@8NQck~9YheMBTBE{yx;rYd%yYaH*^2F
zGdGz@GP5VyJ8S=*XRrONwVr3IM}G8FR9-Id=eKKSpp;7apWUAVpj;7_7dHP)R_~e$
zQo_YryA~c^&9Rp&Z#y-y`g<CTkNjc&_KVMD{6!p1r0oQwh3uS%!!+`3t0l+kwMp|C
z!8<%Kr^fo>ovP{`W&RQuL9!bYrIOD_m4YqN-{&r*EpD$6ck)i=uSFZqXKVec{xIY2
z#C^K|CEF1*Eq<AQoiNqBw31H&3ne7oaqgGf)OE+^n7p^QDTPQWoT>Vi)$$uA|0+%W
z|3Y2;zfm_yLEV4QbD4Aa?^<20Qu?rMHv4~p&QT#I?_QNF1#|x|MNg^l|C>ChY8w8F
zC;kVb{!Iel|B%2!&**>l<@j%a|L2(hColYOkpF*RPN{2<aANl#?xJdLIQusR{hySy
zl`~#J0mc7DLI15Wc>hf|bM;q66r56*ernj(&s9eMx1HVkFQ&ThCkA@e0p3}w2f6?8
z9~SsRrjXELLS6m9t;ws+H*0}^|L^#=p^yNG^5X$oD>?cU^tw6)u!$>zSASsUrt~@_
zEkfQFs;ST4FnLpvN+HoU3&}5`S8Jf0B^I_Ri$DK!`l`4cOzV2YhUEXAT3t;xlT!cJ
zfzNIo{)=EhZ{Pl<Yuj?A6{v8XQY8VbKlByro%RkyR<3VVnEH<gzJob6Q^YUVydPKn
zQ}uUSZf-suJpa_<ckB&#8^s0;7-w^VU$u_pii|hIG+ICY+<JG%aVAt>dGcp<W%f{f
zPXS>JQeJ2NsVzmud*5X~e7({0A1=Aelzx+k89l7jOMBjXSCS3K@h&*}(|VQ7v~n$v
zg+PuJ(1rJ_S%s|>e@^e~kD3xf#plDD77F!C2N4cTF|IV4O%p<Z>#yd2U-J-t%i1H@
zmtcvlD3JEZDEr|;-?-p2*T{(W`1}0tk^gREa6K`gVeQQLn`s`A;Rx#Q{XVz%Uz??u
z_KK(NTzG8+_<~+l3FKgZ<~SwdUR7WE-8OZ-i4zjh`^A*^MLRxk$}o*9L}}^jK0Oe+
z6lmYOmqlP;8oCxq+aI>-xHMkXF}tb>Bs#6ASLfx;_D6&wFY|J%LC`<i!_r5VI|&_r
zE+wTs12EJ|GlC;U-LOem*u~I5qNnn<&Cn;wdXeF1pNOd|8BNWEpFBsTI$!e-OZhu2
zb5|i;Bfqa*_m!Y5l~n$H&7_n2(o^Afu%Doe>(?&rIFrii@Sv(KY3jSAe8g_LGn5|l
zTA*j-+mKxVa6+uaLXp!ZC4bNWpV%3cr>d8Gz~6pY{^+RpD7VM0rkW*2*LX-YSc3y1
zHD(yzXeJ>fbzS^yrFpfovw-Vfa~HH)^d|aQIJG(Tw#Q>R!iw|)=T~1fhy?CYivW{@
zr{=R5W)p#L9q#Z!KkLELH!XRAGvLD?*%UJ^>zEP3_#Qjk#!*76h=)$QD8(|QHIeO;
zMH_Iiv`XD0;~8#Z)X<1kEisLCcU?nu+*dY6FJ(;NdRl1>5%;va0i}pyYh<h~Y%lNa
z%)vCmVBi8%fLL*}OL0*=c@99H{a`zVRKVHI{qaWK<V`rObKhBEQ#Q@P8E=}-fTzTV
zSBo4^I%|~hYPIYIYQ;-sIb4h2I!jvG%rgXO6lD2F7h|6$b0)W5+k+XOqm2ypeOK`B
zYa%iE4@iA@V()NZQWv5(31k#V)55<dgxb+3laV6Te)_8Cvhr+8@^cIeHLgsW%1_Q8
zk300yY}CyTU~RwDe=Qv8hd&Q)AQt?oq6>wSwY1^!6WXw}IDR3Z6g;XoENo!-T}(VA
zVhxOS%v=-Zg%#uRgm@s(?q9Fw2)z!r@Ds4xG~v6ugQwf*rT#r(BiaP9)N%eXbpys9
z7dC13yO3v$+UpN7B?Wm}0mXCOwQ>A{6|*s2DmS%2jORk*Tb2Yy?QxMk0ZJEVoD~SQ
z`b2x3SVlYL>y>mf(P;6+oR+N~AwT*tR7Q+{%*^N2SwBbvF_P3+HRy5ayazT*1Levg
zK#jf4fBuggDZzU2tMMPx`&VB_eq04@YW`f$zig<bbTXYg+YSo<m~W$gEb<Dyt5de!
zKKCfo{AW^RU*vn8UI(KlZUcp1>SFWFaU9bNv5p01PR3KhIEKUsXECg4Qd3abRK~%}
zCEJ3@7iIE>wGPN=a>tovZf9ZpJ5Gazwt_caObhUc^>Bv0GC7^_Fd?mb2u|>0MINeM
zWHtMz8vSWkOsd{3H}i;^R2S!}`6ElqU#C;5Dd`%UYj0fV1Rs^p&~dN~^>3Dtu?6hf
zcIfY&PBl6o4b74wS*Kmzg*?V#d~%wiA?;e&?jFq-oiQzZ38rmjXMkHwyW!9zmqjhT
zU!yz;al3)GF}815#OJoOM}*;J!5=(mHj_m^cBQEuYqWR2Hh-X5nu=va0#+WaYUjdC
zvOJvFvbhj7=F}Vax+Sh+xeZ6DxcvW&*vuA#Fm2S;T)LX;_{vn5V$3J83fJ#4*_N*}
zW*|7y=WaXgknZ|t+jLy3q=uPbU(Y5#5l*_QtAFs#IKiu7hcZy%#)vPdVOuXXEUBH!
zgeRsyq4Dt8mF&>X|6X7FB^kZZrq`r7Ca>X-vVD19GpPgX+o-?+(r|ou;u=ot6l=+x
zz%K#@(}rP9Ho}Vur6)0dElcknZM`mM%!7epRjare22VJt?0GsOEz8{IiqDBMA$19l
zr&!tdI``4jgNLTQbwhTi2a689S+()^`M*<TF+aPkfeAO_3i~q0{BnF_^ieUtucgv^
zrk?qo{H?3`lCn6w+AzVQ1SD}oV;`2g>+hgQKhU1yyC2dwlc;}=&JG*7kP%Drhy)lM
zRvk7tdsB6GI=PQ`SxNN{QO(AznXTQseD!c~?`(wg`Vpmv=Q$J<xR>sC6^$Hz?cjJ>
z?rL=T`Ofcaib+bGUw=uCn~GE_I@bk>sV`u~m2kLlefdn2X6uA#R##MC2&;XfZ;r+o
zV;`UDG<SF;1tpFb0aJ2APV!<mZKM$h4T;yz%s^sEBY}Z;(S~HQYmP7aW-mc6{YukN
zP2FR{+$jC(rSx0{pc?26<cU(Iu8iE=s4+WO_S{DDDz(t;{Zn?wVn<(J=qP$7fQfs@
zZwE#p?$_QIeK@J9fRV*>*?+i?Hgm2#H5BLCD<Ul|{#*q**`}dJe;UvOC)pQZ@q~R4
zn?l{#aoF*nqk11aaq+nHq%l_97{o+mFYtCTdGgzi^wWsNvkJ!BOQ~aGO}4u4P}#PK
zI^q-xgO09OSJO3AS4XJpZyzN%dFC&L5(CMDCHi&3rcU14?IxaW51rCUZ%>6*U-#Hf
zPWt80RvxWZ%d}hEFC~Ta)NUn5$-28fON|3X5UFmS-Lx;QdYwS4fldPQ>d5pl>edDo
zRL=vsdm<If6|q9HMGRwvt4<Hd;jG<XrQ=($Z=c&?ycz^8Lpxi%gtZdC1X@W{tJgJ$
z{)f4@BEtMEi4vq6?U5m&AD5aimO0MP+egw!a^!|-)LDo!E8Re0WAIpr9tJ1?2@6Z(
z`jAQ+dj7Pkd@^^IHD$I8(zbn6UYum=792rn#W1;jN-IF}v?oEZ<_&#GQwADY=2q_`
zTSUJqKj*{EY(#yE(fG3XWw3#!>Uo(xVfxIQc5+b@D`;5!;-KT#I7!uiCF;#y5qf;X
zk-qjYC#9$r*j>w%77Fsz+ywr<26&}<vRnb8GP>^|z;uAusT2JZ%%2s0=gSxqQeq^7
zO$px(5s$)7T%w}Go7OzMT9=64R7dx9+u5_1gT2jUjb5yZ(>Xps$SeH(Yd&E+a-U|W
z1kg)yaoyi0BFrCskLPrC<b87}xndD^B-kZ=;eHYZwMOYmq9b*?P7-vc^R#2Y%p498
zW-Wnq)@-Hn{>GJQbnJ@)Gi}~R5?IWUgNtDDhIS5HLT>2Tck615q73vdL_IBJ3Y>5+
zA*+zoB!yZv0;(JP9wAy^Iv;V4^T553s|-)5;2rw2DP6kcq9b=lmijP&){lp$>Dd_)
z$Fi3t4`VHF@u`k0huxg_8Fst`tEYxlann_WuCfp=QyGk<Xel`gYV)yQpM?y0A<QQ9
z1xF5KR7YDMI5HxbQEtF}e-!ROv%)WX)sj&pzeuo}^a|u$yE*5eFB>vhH=`XJEMJQo
z1o6ADJx`L&<6}t9Q_(qu#nGpZ)0g|Pt>I)E2Q?L8LmFTNy+s*DBonJM!4|Pfi~~UG
z)eve;d7uQoroE<>l(o3<vdm_)y-@qXQ%bqMv}*YgY=mTlKTJA={g^or0z;!?lnAig
z$<?k<wE%60O(JGs)Cz%$<Jnnr5%!U8dH(Z(TT9Vpq-#YD73XJ`yv7K7miCyawU--%
z*&Xo{Ebzp}86P=Ru+m-o{k<j8ek9+rzGi}ltWJ%PyYSJ>fWhROds+vdd=wC|NwJxQ
z6!Z6IF*ACtck*;p{EcI+<x1a|LjraHe^mqJhK3YG8-*4xtVv#$HFu!L+#GwRcO@`1
zEQW8o0%v3n2ZghP*<ySXVECtTu>95}khY7iu$dePq;0B~t6?X<`0k(pm&K-geiT#n
z9rQWj4n45TOwO_d{$ij}G&w6NF*R63krqZ@5iCdbM36{dVI3#xLqxt)Emz$pNJn_;
zhNYJs&&V>;p#%3H#<8r&rhLe4ELR@{&`Q+JJN;_EOn*>#&2Eu@>Q={xU%O8rS@e;@
z{KLYSQoqUhB8^6SZzWD>iL$^dChQlQJkzk@sf|VFTsMm_ghZ<sAYc&3-JHtDwcTPz
zrMV8gDreq>;}V}}ESOWU3@uy<H0B})Y&bFsp^^~(wZ9`++y4TL6=Fz{69VyqC8(I@
z?FX@s1vaKUm7&wjS@|1H0)p+ZZ(~_pACGYPb{vNZ8Gm;VJWGr9Q1@tmX6U3bpc>8~
zDkC?NKH((W67STKx%I0c!jDU0j*)PmUYoG2q((4WMJCU=NG`|}crV@XzPV0YyA|Cu
z#?x2a%n54Pisk{uBbUh<`KPdtr3xZJiuE2oK`VCVa{R%UxwuE$nRA&_O&DZ2)?7Cw
z&3$@OVmF4yBn~lPjav1boy^M<eBtV}t(_dP1!gIQz+Ub-nM}uy0vI^7@MujxdH@2(
zvK}!fOu0TI<U_ws1S+5&pIs~$UBzHxJOhL+cZ;#VuSI^MxW9G@uSK3ydh^8dz#kLP
z-`50;d%g&a^B_J6N6ADgCbo9KKusC;Mtp;LD@R0=!cUeH8pZ!a@^DW`t5&DNU7n<T
zJ&&-4su{a-bCo<?6PS3XT)08aTKsm@){a-jLf@cW_|eRRS?%np!WvjC5EvLh2f(sX
zhVVa!bvG(to6D&-E1o+#jaXn5U1!y`nxsQXJ|V({kY5hNPqLuM^i-#^b}~^#hsX^m
zsYnQg#lBUKiPX2*CamRln5u;fOz%zFMwt)$-}FPZ4HSF5msdBIh^NW77JV2_Jv3MX
zFm6e&;PXuK;m)4yep0k{M{?Lsn+oY(WZthH(3=+06E2N$PjS+k>OdeYx99fl<XjBD
z3Tw?VI;3zInWm>8s5h5{Xr2w3h+K;Ld2A%8$cxxT3P`tn^z}*?BA&1morx$)&PqG@
zD{V-3yefJ6_iQGv!-^ke;gZ4{{dXy3yJ!a1CzjsAKB~mpW&F-@yW>rr3`>l_W?)-;
zNR5vFvYH9`?_<$V8jgO=OFfwU-BRN@;{MR9>Fhbq-`DoBt${4d6(?#+Zv&$>E02F)
zdvQcLWlK&xzpqVp%m;oQ)?PfHc&U8CumAg6QBJ`0Y4zFfYrQUl*4=z&aw#33d<v~O
zuctuEz_!ndfu@f=nOAd+ga=W%z>??D``zhfzLb?jh5Av0D6#`zL&9NC-6FA${V2TM
zSD~Fe$h7Rtsr$k=aeh8>Yg}^Z{h8>cumyD9W;M5uq-HHHs29<jVr-sBMWlC`(4QN%
zi!Q1(*B~l$--jS`Ctj!q1Nf#7+;pF%`ddtjEYR;DuIi)^jS42m%hsg7=y|pF;%fT$
zw{v@57}QXUv5`@kJ|fMJHfge7Rbj$q45$LvUzIlX*uBmFrXGBPI#-rtpP$qjw4c1|
zRoEMwq1n0rXOF8I_WE7@U%5nms?Va1wdYX?_!b?f>$c^p+d!`bLT<CAZyxV_Mi~$M
zmz@ITLFDEAYGdPnH>!b^TjHl200x600L<$p@p*+^>dRiSWZuu}OM1f*(Dw$$q~uNL
zd+6a^PqACQ4fGG_@_?J%j>;^h-&?Yh+!3*3o)cabhv`aYQ<(3B@1E{6MVGHwe_xxk
zS3U(Nbe|}94;IzdJ6Lcn5Uh-2I|-}Arp+;OeZ+82dW*o3fHg>Cb7a-7?3u6L?y-t<
zHB>o7L%HTBFC;0)*3#Aqs^O&8{*%Y*IG<(0pBeN@)6F0e;EycKH=AJ<j*>O1w~q<H
z2PCP+sK{40JsZ0jHI_$a+vr3fYEl+NU{-o2g!C_>#=^VI(4EKBQb1lQNg$9sa6(V-
zE~S*m^OKd~Q~=Wm6xgr*{;~C*BI<+X@YOhK2k(?sfx=H<CTF@m>%V1D<ztO9`{gRn
zQMLO@Yc1S)o^9lLjm^<LozU6e*E(DE^7W2fj-D4yG5`@cKHSP?I3PL{AS~HH7ku3l
zs2aEx%J{|PfqC?$uIKFbs9xE+q$tbDA_P`Yw)-<rJyINB{QXEmcwavlztU#_Gqu%8
zl2T0~TK~S*#5faq_0g8XMJOj54vU_gb6s&q@$=O}?#ugMtx;OPue}0s1<t;ny5z;D
z_0LgXY-duW)Y3!E-`DU5uKWAbF_#l(cYj~&@Ugt;_c-RFOo*91W)A#PoO6MsClB8i
zxC_a6<U~!M>YRdqQxLgd)bM^3UQqAZsH?i0vz@8K=KruWa2+<M|NK~d`GW-8V`3p+
zC^K(%pAPSP+tu{d{<{xg8Pe-cUnf%2VgLrG%Z=XIQXlQWO$u@DFnY8hCdT*oHK}-8
ztGo(Dh|??g=chSAYZto&gUf_z$KTifs517kT#~!mnC-a|sxrFxeeJ{W&=Dmw>P7tG
zi8=+y1DkdTuh)SJ-*|JjA*8ko1C({q6*+J>>FDp@*Lsu^lsA%I9O?Z*`ND|;B~I_5
z<-SeT0{!|2lCEn_pTBd}fF9<jrHIeNI)`cc6i1+LKEzw{kVVk0(M{l6CCD_p>A*?7
zk<k<n#e|wx=;S>htBm`8D}WTrPiPm7zai<oRXGay=y%JWq&lqtZSp1PIxuQYoOqT@
zuH%tbBG>|u6#b--^mmO8CSq*j`_oez59-6K9HiB_Adha<uj=%y$UNT%EJF@n&Ec3g
z?thXx@}J`tM5tYF%<M)=%=JNax{<>@J9%<{bw`~jA99&kcX)syapI-es^py8f`2%l
zouUi9C6MPs%pmegJ+i#36zO<yhf3G*n8P3`4N4g7gzl<&KMw+^)O(IO5H%PIH@))G
zJ*M>qk^-4%f#ecWko2t&d%Hoe%}NEe#`F3h<9OL<rgZb-85U6dyO7rhvZd}>yX3Pc
zD`cFyruHOk8xr~G>RoXq@0D%jra=R~h`x_m3f&cRU2D`0VO(xpk(1?)09CNrG0B$o
z^ARb&(t>n*!*rDq<ihC&BQ#3Z)pm4rbR1;LwSb#F_><2;8`*GxmxjEK-ER(mytuBr
zaYrm}jl(28=6kn5*P9$Rq(Bv&cqh0NW}f9$yMS&p$Rv;pnKz5$1O@XoqrXd;z&R8H
zhS&pjWQfIKeytigJkaD&LEZSI*u<dZP>|qtrdq1N1e!a>oro`WP30JTQFF{c%9Yo^
zb9>2Bti06KLK>f&q`SH+?2PC)kORvO<F?;A&aZv@atxn!r4bB7R&-a{kIiT?!X1**
zwjn09%c+4I^-CuCf4(8jGW&51KXXEPMeKcAMVVVHcm--na63HBCvLl80*&+{!(M*u
zw49)E(LnLZY!QK~2S&BGe8>arGzrw3(<ko(IGVTXEM4O!GSdtVhh=vXf>x&2U9%_p
z5D$12gsZ36g-vdgvH|71uw4hFG>1ytyf%Y=TL2p7x8Gn*^5nCSUcKe%1%g!H4V%#U
z`1F~5@|bEV*R;H0WTCkIpx`&EZhw@asH#Lr<(%27@$0=^fMst<++S4)KW4HiKSLE8
zgP4J*P>Ps~!LYP*=v+u8f-1dCV?bUvVZ1*qeJHCq5dV$XpVL{%EMl9!v-ed(hG%+v
za>fO%VI;*Liq$mhsTjD?%??g2t8;1Q-^#h?A}pdKJqsb%o4aA2&%!kdM8~v19$Xb8
zJDv@u&~#AeaK!Z8jwX%IfSzu@xTv6hJ^n)dR^Bi8u*s_G%{4h>FBxGL^QE&oQaM!Q
z*TV_7p;6)LFfx>ojMjL<q^7%xKL@&k&-aX9`8umUoC)nn-n^?oAzfzg#C|Ke(qh9%
zN=6(WEMSe@S3?DDD^!44Eiv^Q`)@D_QfFExEh7R2wd46uo^RMk#g(pqe)wA8(3$GV
zE=Lz)V`CcR)uWM|(n((H(k%;>xPB_k6?gP1Gvq?tJXSNWJGsJdZfWsQl4(atk4aG2
z#c`2`mAz@(CzJEz6!9@#wMeE&dx>I%qjV|po5FLPQnj;~5vJ~z7uMNe4z~9bPwaSJ
zFt)^Zupyvh)dKZ{uivXNzwXF*u41npE(Bt5*iSe3ry_=i-7~Y_KbyBmA7N*fb0j2h
zQ#-%nc{RX7&zTkT+B&?~jpZ@Zpqj5L#5kZd$~}1s%b+bS<VCsll{WqnN*mrXtoCT>
zYs>T@8<l#M-me_)k7x=@z8yFO*lD&%#yCya(bB_v4}dcLO^9hNpwwuuIP1nIHHEH|
zx@BrqO0B^CT$1_@(5mGDJWYvgSWCTG$J6P)aSMJtb<yJRa{|GzlWJSXRdVE+;D47L
zp5f;S_iB4onR2=W?Sp)t+}&xm^RCiO=po%${9b3l6m$CYQQ&2&*X;}K?>FD-UQmv6
zffw8MlP6qx<2zRW>AA-5Yn2g}R{+TGYlDD}P1)7t!1AKc`=96Na`cZ?e%4lll1nW*
z$TCy9j~BfjylBbt5V!?yxf`A9Zd&ES4#s*cxt7+?NeMD8%PWu7pKxDjOSA&h(vC5l
zFJ~DkAu9}x!{m3TVq!g>{VD^x@*55qgdSL?k<rDYt}M$Lamm)reHCvnpTyVgIH>(~
z^A=pA&-;8rCP751gjVX({XRPvOnEnWP^L)rb*KZTSc$G+4u$X^jgs}I%8zSe7}8gT
zAM1UK9=Icxuloh|iB#8p%LoB4(gqzg!>e8%2y3Ug)+sPzTaw`UCt0I53|VFGC7yIv
z)Y3jv0Qp#Y6m<G``63+=3XiMHD1HjsI_>5Nn{uAy9eOw~Cu94D@3U=bdN8Xsittx&
zsyxjWO3>jE*#-yk8?{}hq$+V{Eo-&T<mGu2#C>eeCO&bS*;V7EDKiLuD6H|$q;>nN
zs$xAD6%Us~s28Sg+{b|IH6pO}4&_s|{J7(&%=$OQJr=cT1%Z2?PZGXj4oZhVjZ|Lc
z92mTNJ)V9wc6;Rn$~EdxCBSd#Ihl+{!pltGm)Ic5M?Bv>L=`XDM}m=XUa-L0qkz8%
z`oMy$A7C~$vWJZtMK^?!{AvkxYd0`C5__hB{r94Bwp&{Rm#zQMOuiCp+r|F#``SrD
z$Kt^jq-V`fF0eAa;)-b~a8O#6`S{ABG6u}WP{;q(PJWiu@~HJ!ql=)6f^*Yhk5l=_
zRbpa%Q*j}JpjW0^q4raLhdD#KAD|}zd3XhSm>ssituR+AhZ}IryVqV%y>VNU=r!W%
zOqT?;6)WEK6vNNhO?V8{h98*F3?!*!vDX(34?g|mLvi>*A5{{0B{?*Fm%T!vo!KJ*
z8nB&--Nv}P)Ex=2xI+HJNz1#U1Ke8$=+Ds$kP)l412vbrazgopyU~l@^FO@wG~X7H
zL|#2{28l7kCE6YhRn};2s%>dTbf(akhZIiXnJsTQ!0(O)f|sv&zsyKjwQ`@BThs3#
z2}>_H*(7;kUIeaG`EAcSInkBb3H9otPo}wwtVCE~KBmreW+Gyfl7T|rF6q--dqMY9
z(lV*P&#d82^2MW$<`Rh%YC`L+I}4OWv5(VT(<H1Gju{UllbL+5vW06>48<h((WbO;
z{|u*u^)lXbh{GH-UOCC(fwplyBWzN{x;0$t`LR5Q@wOjY1<Mtw6Uq;bVYo3pMQWAC
zX+oQ}{;(os1&Zdqh~f_go_8_7(~C>fATT(@`kB;K!C-?<WY5S)-7U<+8l7ssrk1E)
zAC6tS$-dqzlb<4zj~8S-eY31TzvGH=9OtqFYiu{Wuw`n~R=m<oJVVY4yhI9in<I-x
zpa@`CmtJX@)Wpr)wu`14BZGnZXUO&@P)7~?V59VzgcO5sg-lOZQa-FCJ`NA|%`|>K
zQ69r!p@9;ySyg}tDJV0*b?x}atgJh}h<E;aqkPX$WIuM1iz9<&8a^_Q4VMCQ@xCBE
zb<FZWD~Zfjp-o6MchKq@iYCQ7u9XIiPyRA5mLV7q2b~lI`KXZwSD%+?61iQHv_75}
zL2-k=NiuJ1z*`E8t}Zwook+2JO=BW53tR+5q=d(}aC#h5V6pKrO#r(^A2{@@nQf43
zd3tS{YI0JfamYsBY-p425UC^cojCFxU1QLmO@+gCQPvo@U|vLaLIU;=fp}?eeAfXO
zm|J)WB=Mut-~~&*O!E+IvYSfq?b=Z4-JoMqqpdJlHoe?2ZMI4D)59vZ)sQkwKz8I#
z_{_H$6C_D7GO96c_y}azOfRnUGS|eA#IWJE<x8sn)>k7=cwc6UYw=vHR~X|cC=aE}
z0`y9YIE1j4gWkW_vdFIt=u{e)mmjA}s-mMB)vO7)JF?4?`vQgaeRY3~UDMXLfC1os
z&%kCykX5szHd)na){u9g+o?;1qP1`Ag#Vx^t6FL@{D)ZjI24DU4rND3h+Zt9?q<r#
zC8c@bZhGKW-o)iLVZ5yRE6<rbFx7W&szsHQSih#t6ysUQVda1`X=~YXiw&lwEYt+m
zWrzv(r`*To_}mL`CF{B?7QFok>$5eNc92`#vy9pj$D9y*TPZBakoVP}^VM;ciat}J
zcHXHl=LlE$liL42aQy%G=WjW;nFA*;o;I#fPvy}1<P_<DggOlxpnWWNX(A1#%ptW>
zOQc?N3#+7R)Fh%wz|U_gFic2Mo;d!=S-kDlUr+y&V(`}={6#q-S=p}4SFDRMe+%l;
zDp?<1s93E3U^+N?uXDI!mgo;@VD-+avIp|lm6LUn+|^gIK%dz#wiT8yjl8{rC&-}f
zNauO){WZ_UPk@I9U$Z%j=GKlBHBc#oGkbbmoc!-Jjg-tSEQS+b1q?QyH(b2C{`ddK
z&37Z;n{H4i++4PI=mmk*uA|@^`va`heac#)8N$d~vp-yTz$n+;E%*6m?dFr&AI-2d
zyiPH+NEuipCmaVitLJcqZUdxd3A0>r@AB*?HMLK&8o2oBUeStir-k!BPqyREiVxR|
z(flCsJRv>%o&NEk6^ZXa*D)MITAmQMviFlj#*X}T$~~YCBEIp#JqQRPur%!Vjh*?<
zq#F~m+lG`z#V4zWoh1U5cykGP%7<!3!uSdoaT42j=Q(h<Bl1@y+e)=KWnJRSg`<P6
zJ`BIxY{SNw5V%RFh^K`zzn&~KwxaIiihFcs>ZvhaRvNCp;iM4eD68G3-X-f501|Es
z#Z&{D`1NjBDS?#;(DYyh9tG;ErLOnu@#D{}G_29>;k0I*npJ~M;)!~FBtx6;IKdZn
zXgx0r=~D7-LR?(Icb+CDdp#t<3>!{6gdvzk+V;WR<rPEDAA0#%(VjXi2e8}v_SF8E
zV<XZn*X}|9(7~Bn)P&+tGN+1^d$|vcd5g}UuTbq}y41NH+3fZU3H>_d+98<^VkUn?
zZDW{-towb4W+BqSK+;3tX-;gKOR0~2MGEe7^rV`Lc4b~i;BB6zUu@F<)Hcpwu*!5=
zB)z_F8y|-dwG5-rQC-bkWdDwIeB{NT;c}?Egb_{UG5g-pMtdf*j>qDSpYK2FVjILw
zu|YH~`i9##Jv9fq@dDNKAOSPT4@WKo5wLVTm`ztpS^ev_RrqTv0h~1lCo(k`jk!xI
z5(4-jpN32DO0G-TE-*6+SZV&fCdx8uHOnVhH;f=|m#lgUJ7Lipb{X&#oB60Y29AMR
zbzeyl`axfCe~-j^`XJ^i1-RlVc-zZjh5MQWaHeaUlVqNTMyqG;sGr+y6_9t2*Ty&r
zqnk39>aMdT^9dhIS7DnIEOZ_}&4Cy=`8^aD%D5Zly5qXTj*9{zJ{y12CeEt{6QYtr
zk@>p)FdouSwYD!MR%JRSv4AuH{%cr^r1Q8Ys>|pzMt<7}?2k^$F9oROq~PI|7nz9{
zd7UOcsjt%9mqd+Wf2kpHkRLBMg8|#7L^n6QiBriehtoe74+mV>r(Q?0i#ECZwbaso
z$aA(+g7nShx~0BJi>3{Tt<NgZDH_uHTx#%Q$T2r@Mre8+8dQgMCXnCUfw1(bpA_a`
zU(<Ry(xTH11g*xm7@n`@4QN;A^GfJuIkL)Kw=;*MH85!Nvpuuo-g4w-3){*s`4+A;
zi&!j_(65>un`*3w1pzE3)r=)f*ck8~e%xok4(qWYo!OKobGpW_7K2gIw)yuzF4x_%
z{W8*#sZuz;nOOW-oh#q7FN-S&6;X_U6l2O>ldaeU)ZuBbrvOlGeIz;9m1|_|0a3G-
zE~#>$n{JFb!SPN0-9epxJsr4Uf&ECq@<6W!<X4>8sjy9{hsU42F&G(PFhUtFC#g(@
z{riy@cebrIRwXGmJES*uMv8i0_^4XS6DTvv+mh^uBzaYc2pv1UC1q28kM0na=$Ti-
zEt`iRnmoIa3mA2zv}D<sY4V_a?ig&7Y38joa*d?w@H0EC@9<S!K41D<?|ke->jnB{
z@eP+~3DI4qPACr;iYg5flC8RI_Z+bVgGv-Z;2UtMg0F;>>M8Y*^m3ga=inw<Ui!WO
zbJMus=kYta1Ad=gsyQmyLk>dD+x;J#7as!!h&6*U++3u|VVyKv?W1C}l0_xMMEm)d
z5@)>{`~zCLT=)JIfteD+5`YgmxZb3~a5>QjAWREHO@?jwXTBfhA?NDx1i0dc9kB8n
z(eFUP2WdD=H|V1_!ged3+#f^)-ep0Un3&86`AjxxWC^jN+mniiK0BHNtO#~o<>*GX
zvI{&HRr?!^>;8c-A5q?ouX)vy>&|zMV`yhMjmVO)I+nhvJp4Li60E+Nf#jHXDJ3Ci
zL^^WX(FVn&mP+QKlS*EhffhVaAUA}asQrUGhO@yuP5IC~@`fAJnFA3Oq@r3OJFF>W
z=gii3M?ZYMM~6*&WDp&A-@b=ge~G3`FLm<?CB6G@lzokK&mdH<3AU4ym=P2^iP1wz
z5_swaYMBdL@r@4-L>gQ|!wWw4O_hvb;3K@a0TziYK$(q4$cYxc!o%1p6P`{#cLY?<
zt#O0Zy*C8*E6%*3+^O*ewkFtft~!A$*wxV9pj_@1LP$bDAeW1Aqik|Bs-UirU%Wu4
zZakwaw{(0@GuA<MmRDEazGz$s^}OfM$%-P>b`=}&V;Bw^L=&>Gf7s>?sG3kyKPD{x
zfn>vJD%P3i!`zkQKYV!9%7W!Wf6z=R>`n`smRNr3?=1?^^;Il&qcovtjINELnuV8Q
zq{OKLNyFtU#pSKn3UmYT(S%{MW(P8{;W-_a4=t;qQbe5es1e}F#6i0Z>UtWLY;A3M
zdFhXXb`=AoE*S@9xh4^vUm{(zptp6<e@eO?hSjc1pA6-E&<~ej*5fcN)*1qrT<0Tq
zq*lg75CF6NE7gLEfiz1O&-?5>&unNi_!a=a2BX&pA$n0!PDYQbBIz~{LORVU(kdNt
zU}?3U-^Z+<`U{3+M>p4?_a+GW)8gf`G|QCTpQunXg*OF$jmG$cfYq!%v-KxN9~Y0+
zcQXnW9P`)4va>y!3p;ZK5`T8-vkI`H(ns1P-syfgb(;(nzZkB%Xw}v?rA*QgDLTrk
zcX;LX@@0wl7W!RK<#*=ewqBv<LfMzoev2>GhjuRK>bAkn_En&n^G;`T)4+1jkEDi~
z*gh)5eQ%lZ;_s?dpaEotwK}1_cEzZ0UIL&VY2kI_(90UYP@}CfE;nng=#m-qDNnXn
zfczdD-{@glbSYsEzh!5n-mqKp;UoIbG1%6ljS-=^_yDk6!lRWO0ER&-VuEs&AeqdF
zN*hM(*tq38r(fOcFwXroDsqtKo3Sznh#!hx-`O^0HLilSL>b#RZP7+FO<ge&V-!<F
z^H+O>m`vjrNXdZbmH0?+1Q1!S{g%eu#T0?Uv4m3b^-P%J5uB<~ov_Np%5S>q`UoEF
z6V-i>+`_Df8+idKhGQX*ioa2xiDo*`m<?*&4SLYmH=uDGHo}7o%FN)Hj!USetU_oy
z=LnA*e%8UnLCH?MMU_E$VtB1}ByZ%)0XwZd*^a%cfWpF$MUxojo*#VbvSAdQsco`P
zDMwd|O?4?0&VLk`%tMOQ>>5`K9o#g0iAd0~WkRir$)^cpwLQw#Ok8(P@9=uRj!nEe
zic-u!SjbIRyS%Nzrb1YvQHSG*KBsK{EyB`5=?Mk4E0fN<IYon=kVp{`ajF4769P4%
zq_iAyrM*ktKAmoNcfl76f~Tdt+Z{o7`*)brU03tB7-V3wcdmD!-yH<FtPV}8ZpghM
zG2gIN=`XRGw0QId8$5|XuV?s>%g#8`KNZLCMqq^_)iR7-`zN$Zge3XkWwWB-+{Qnn
zb^mdI*op~FG%h#kjt#|m^jRczvFUb6O*gghYo+9BlkxX)AG<$jC1vL1q{KVg@Z0`X
z`p_Ur?qtV3w^ZGOG-^?%0d8AV>8Lf65@JvZ)(_T8jFdcWi2bWcBy}S-_Ty)NgmD#6
z!DQs?GjyIDyPbehQ$?@q`AZFLcdfVijIQV6N7(|F&LbN~V`AD9DjQe{X<>tL8-tC`
z39dHX@R2Uxd~l~oSCfI<okdqQExyw_>}AoxVIGJ4wqX8SzRA~xoNrYq2JL1Peipq|
ztf!h1ECsGm)J?}$uyK?&n`M>f@d<1a0SKdCTW{|DxhzL6My|UNsmbsAJW``hL)xl>
zDm-Mpg4x)eCc+N&#{<dr2D@(<Q?@xsV%%!&pQx*Z-nfR2U=Z@1P;?BiNZ<}Yn6xTT
z1frW*Gqht*;0C&O;<f6uFP;AsyD`<II;JN#Mp+~iB^?@8B%#5dnv@uaMP`Qz(V+QC
zqm=sLG}YP`NUG959q|(o$g2Lbdy3SNJy{;f73VC2<)=|mj{{QNtoXye^Fhid!?wEo
z7QGd~m&mVSo{Q{Wb}RZ4(=n*70T)tIR+w>&tArL7)r3D3se5wvfl1g<o|C^GRbXa#
zBJ|KO-CtIh9cBuaX=hq0>wr&eZ{OH4-=WRh7Uy}tN0%s)>Ud{tS+6TLzfXG<O(Llk
z4`~+bHO6EX>I?`E=;`!CU%nYq^HA5wM}?6ugj-`?+xiB8;T2C4vvX1*x>zZ|{=;J&
zi_QmX08fv#vf4u@v`I+<{MfW)W}f|XsRU`4-PHKm2d`kIjt)mffW{6%FY(~`D;$O@
zBbUqPSvMP`$h%#*)f<->a%`+N*&Ng}%;e^&+8>I^^%&P^85a_;RfCw9^l1FI0!~Ef
zkfa62z!r{=-@%QN)*NfQIyVi`(k4Bwc~7dogP0vQlJgHq&?<Vq@#~;Thrn0zt0fOg
zQZZ_bjhD7hpa&y)vwvTkViVOYo3Aqx=8x~q((QVite{*)y1yG`)a742V^qlVB_X}U
z{DHc5(VHsyD!DM2!yqJfLVh)v`i|N|n01_HME_$}t#n<D0`A7jSH$Dm<e17qUaGGP
zqDIBCb|1t}QR+Y>Jd`4vSn$0W*AJ%NH!S`+KcyMN1}udPd(}+PJKN~9me?=eL<$yS
zJfA2EK30@D2$zy#;2TpC8O5UIdiz2o04&}02SkQmZ(iC*lii=^Xvpms()CueNZJ@p
z?5o-23B|rwDtV(z@e$I04h_YLp=WIp%O)W4`$ZM)uY2-cErdH+zR^#*dWn?ul%bzI
zQ)DhM!4K5k;T;~<G&HWEHxZ|~OVzR&?h#DGE(^XVQZ&A@0f&<}{i(G3wA!7h9lS)6
zVy&c3j3@juSJkfelYFv(ZFSn)8(r+nyPc8eBFrF?zhw$1dbhcpm1sIq;z^4Bntchc
zK_t#A<VmY3EBU9aG=+^$TZy=;jYBlT#RIL{oPIcL^M<ZBn9^<p-XE*H;w<PLwXf}E
z@Obb~(bz`y-YTEXlDr~N^rdP_N-!NGv#4<#piJFwehYsRv>R2)so{pnRR~!r6cn@r
z4(e+9Yt{NYg#9p%L(8#0n#|8W7D-|wt0DU6B^6d*qGY~s&&rgAzJ#ZKoR!gSRIPPJ
zaUfb(P~dv!ao}jNM=>dW>ld<1K8;3H#)#)Q$G{1<;Mqe9Rce|NLFj0qO3}hAE8190
zn_)4M0Qhq-WV&WTyL8@RGiU|bZX>H3F7=p8+PjWPMwK!#E6UM{yf3F^;=^j<AOz9<
zAR;JEg+$#!ASw%1m+F{Ubd5wZ%R*_f7?C7oUpoG~^_S)8!y3Kq29bJ1S0ah3<h2<w
zBI~nxNt;nC{8FRJB>hsONO7gcv$o$nO$Pdl1F9=C3hEFT30jZeCQ8lu&Em1TMWY+O
zcAah>vxZRxGFd`&(0ihCCakP)+^XaECiz)^AfNuL6%jw&C(6{_GwTt1Z_Y$VF9zO>
zOYC(<LH6m<nppZ#L2(9Aqf$tWsZO5xuv?>3Sc+NP-RQB6O8eSbPp%Zs`x~{X5<<1>
z*mUx|@gwVcW<C%02-A=d-r6}Wa2Zq@ht-B{zMPn8Yz#kbEZJx%k9jOd36{c0#TBoQ
zGtzUiNgMIP$~4u&jul~L8hPcA@VK?grRuw~uXBm4bd2nD?C>K;7P2EMrD=vOEDc|-
z0cM>ZtpS7&8m9GgaRc2Y<K#+i8Mi$i6{-(Ec%tU*knbZ-FB&%HKyE0~yFt`HT0g1@
z7dtiDs9<O^-!m6Gf%1lWn%^2Qb_u~Jn$8&G7wYTIv)gnT*LTV_3z_3I^W%DVC*9Up
zO!FvQpzta!-gl5I<MrrfEzN+&+UQo|4B5Bzf#EVLL${>Mi)lStsGx4sG!AfVg355E
zV?f#k7ob=@Ei}9^X$pd{xTp=ES%31U?aU5({0SqxDvirR3bx}87|B<)o{e@)$CD>9
z1ta?>mAO5^%r}wwOBx|3Fys1~rA|!Mj;SyxbHdP6Sj<N#nqX~xHRva~T3>O;T4)jg
zTqTMmW+Z<hXc8>%#6ioS7xm!b7*U}5<o+FBrjx3b+LUz90uU|%TSkS;d5Qlr7cc>s
zMd*bi$HjQF3Xl%_wfM|L6H%pUcaO8ES}m;^<LFRL8#SBGTKm}S$}c_n7WgUfq~V^{
zTjsCST49XvX*&%j%0`3xz&gG!wX!1?)3JFn`@$*Kvs3(|@Qrj~)@v=LBy0L@Qu94=
zUp2pwfi_UYWt#VM{&H1U$K<Y2?G)9DF}_Cn^<Nn=@B+OQ>ecSp&Y-$qiKIbUEtK!v
z)AEVR7O%+gV#X>5;=BT~$jF|^0t#6tpAms>v?>`Cp(c&_PG0{I^m7btl~rd*c}Uoi
zVm;>c(+aNk0!NNo5$H+GN{JBy|IxqV<lvC1_koKs2|vY)R`E-vyv04>S07j}=W@!J
znG1d}Q4s#dC#jijs<l-6QgWlVTo!CPIzvgxVKZ=dNRk~KY~oil{mCyqH(03SJp<+o
z?@rEwoKpt(sxONqlDW#PbaV^mpBZ&bhh#j4l)=CAt&6@>*3Evu&gd3rc|DFw%Btu6
zu>^@m^l_f>W(y&0sajs@pTxdCR8TZWd~#-d44oa)Do$Wx&$``Bi9;j*JI4gFk>0=r
zp9$03Tn=F|HXaQ<Mlz#=>*@A7&c7tDGHlji=mzWncF9}VCo6Jd|0K|%*-fKr^JRaT
zxEs<l9FSv18lcR}qSWCJPui#vyrUBQ*IjE4d~0j|Qj6i0?QV>Up-y;||Bw6ZR3kmG
z-Ov4sefrldR$jY|S4y{-RE^PaWBZ(z5>~FE9)gc3b*mcm1>%7e$!4{7<BEQ}u9x}7
zGN>xjSTi|Nthdut(y*kiYeJ{w%e^6-c9w1iB=Nut|5$VyYCxWinvM`${uJ+`GROzz
zD^uF-&YlVwV^iw@II&_q$}B3De79L136p&qwK?Q`rX}~E3%;lpS#^N#(?Z<gmAry4
zStHA6i}7U$^`98K=-0Yz0TxBjjxjg)U(xQX->q65EUAY>PV|sCt<WljRm;w$u32XD
zz5)WvA`;vf)?pfRyDD#}$nwO|o0DT#LA&%&<q1v6hzPkkJ7&BEns(joiAIY?0Xco0
zpZMuOW^DId;G5&Y(aoNQHVtEZ3U-g*&5?E)FW$y4RwGD<5*^6-Zn)8;b8Oewz$m%(
zgK6Wm(da1F(%yBxbix-r_zOh@W|$ZC6d($Ch7G<qi+w=~x3E5_#L}Uet2^Ae&v;cr
z=|xut!~F_IeJ9?XOIM30&#M=phi==B4R?G}&r`$yQ(!I~ku(GJkWK3MJHQiYiHHXL
z{lv7|Q~j>j^ASXTI@UUzXnfvLd8{J-GFQ4NMDUSataOfP=QKLtNoMCN&<lyBv3P><
zACT*ljq%TtM9GU%r#}BC!OmJ-j^>S{2%0gY2Z0~#8TU3leDC6dE||M)<Wlw1jH;?q
z*I}q|28EfIRYRBsLj0=jz&6#B)io+9g%^s!4f`*~s(}B*Aq%H0>KrA7<8tMY`6UzM
zb{Ui*yqPRfb)}8E_;}s4(&Ld<Hs3i-lunN?e`i$JtbC_?swsnNe0oteftL<0>eZ*0
z&ww#!Rrr@|BdLm(Lo!!m3JDo$=bm_gAiWkyp2$y=%eP+nP4ts(*rcejk@2fff*qc7
zrFECcVQGWqw1Am~_<6#jq^pSd^bL$v(!SVY4yLc6IO$HFVUP1BuHWhOfcW@dk_{t+
z9bBlQOe<QSAx{S4DFiME@WbUe3_fxAaId|+NqbEFC9~B!SPs)EY<59WY^V%Z|I&p*
zb@Oo!V{y%p=Yn>*M2d^AVbrQS0ES=e&HbvwAF{^zJa?4eVLbZp)0dj>PeT=7sChhY
zWy2Ydvg6lOLRK#yH^WJ_XLV@F=D&TaYD#1SQ=`Cn%Sl-(jd@?B(*bU}Xv&|G7B;tt
zWG)|1L?wx7<D{S!h@}ELr8?Tip%Pv8aA*6y9aaZ-$QfI?9$J@Wx#WV^tQemEd<`Rg
z0+-0wf#(lt>37Hby<D|h;(I(~fXgRj&{ly9ro#>}E#;T;yBv9F2V*#R)}JjA#~L!I
z@_}^}HK`-XVGN3hq~Uw)$R|5$VvB$(y0s)Jy-m>xTD>D=#P7RHD!gsVV&Z0Ypf;DZ
zSw`d8Cv=uX&-l%7T;B(PUw{SGrv2pA;+GtE`B59c@t9UinQttx)-FSO7+M7L$idOn
zi5HB$oGj5xD29%G>`2P-_++A~)8P@U1P8X*Qf5THY#EsFEiWe~DTjP`Us=bJiUmK*
zCmGlY{B{1xzS5GD#6Y811?GseR&yrKPMOm)Z`4gG)=12#kX&3GkX`pAreV*zP>77@
zIfyzQBikt;7*dSjm0<8MUQ@%5WtzrYYgiQ|Z28<w?j1=i9vJ?4U|u*Vw2pEsdp_zV
zS^V|1f4jCe3#BMet0YxqTLA2g0Z|F-_YHx|f|*4%iB`mDgxi*ILF`rf5%WiMnA@J?
zK`-DTJjP$kOVh&+C)nwTI(}RgPKz*`%oD3FA1Hm3S#Ogfe5w-X-B0qOolAuQFYUs)
zet3P5`I_w6?M$37%guusi2w{gmXOLO4EwqTXf?q@iSxi$?&s14MT3tV$~hV65`PXF
zuyl096szS)=3mdaCG-O?qf5N~?4nge*Q^Fq)R%vb2n(N<g0Xu$x;}pt!rs1{?s+{y
zT*qVcnbX)m9^PJkVpYsE3U!eT3+dgF&ZdEEUrtf(yzx$@c8BLGdN!|J7P`E*nF&!Q
zyA<Vyo-SyjzX_T|Ru|?(C+Y#zJ%qKhf5!Wdy-r`pRvTpOFgxr+&lp2ZgYyR54L?`@
zWZ(EH;#~B!uv`1XhHjR#;%M8WxJ+az0mE24)Mu4<-f?(Xx?pPAF8hi}lt7^9HtBtQ
zHc7_2QD3?Qm>Ie?jwGxWYmFaOlm&i;nkt;?G_2kLrD+t*ADA^*Dc+_j@y<MZ3ttaa
zNij=aK))4!5|(%>Nmq}+Hw1UZ2&+OKZrM_$7&>YvYDr!xKHR9+PRC^6hBq`>4PB!a
zXi#?W1g?CloLgh?@i?pQv|wtXUvAFs*6JT<ca5=ramMNRa1RS@I_8g7Z`|+OGxah}
zyT1x$u!`km?E~zkfA4T_wvPh<%G~3?2*xM7zUAm&fW#-W^M-foB~m$^&;?zNqed<x
z!bvkzd`26IMlY2O+_KvrCN-JTI2hpC{vz+|DYP!7Xf<eQ1$0L#<{eDsMo<*-eA>Sd
zm2br=lxDNE^>W0aSUQ)t@q4SLQuHrZ1cAI`7EH28>*+xx#bL<Y=#(%merNg3o&bUD
zgGuzU7rA~DG8{b-ddp%xq8s;cWUT*KZYHm6$~h*G)N1tTj!@lOi_fGwHrVUB;qv+F
zS2Qg&p#oLHqFc%4VhQRANY8&MH~f;E1tS}e$RUfWH|GVgAjSw+$Db{N4&IP6+p<2i
zt{}Sc!qT(|o(to{1O9|RvZi2GcSyfI-kKuZ@QBLjk_CD5b7)7w`7MA=vZLu`t*D$@
zhh&vavEhmIjPW>KdL-4IyvB-D583Bg0Hi1+deZo{l@z~P4Wr0{(W=k;nZ@{d@}h*R
z$Rp!h7_{d;&6jr=^#+3HojlQQ=S{QOER9<r&?tf28`9}OjK~ZE!3MvX)l<8a@Xf-$
zvw@E?##=-8E}TU2&3PwZXLH9Gq?L9TLZEQor*HLKZsKYOW?Ox;9<$d*Ue|!L#{qHD
zPRyHC05ifVao|o+72h`upr5Zy!!?#wboyt`qbb~<ufpkLM|m_t7Lcu?nt~fG^Y!M*
z=Xy76zuo%P@q!pUedonk<?Ls>3U~$B8HPjfwX0i{J7sCRDj4O!)5b`$+Zo|<0y7@A
z4M7;+6RtRqjOYVm*kWXDWZ0SnlFvV}OzzNR-7Pl1Nm~oZSK5k8tSzO52r&!baUk1C
z^Os_pTch@dnF@S~o1%%r{_*PJocv##cvC6PueEQNTY1ZO4BIyAOn}^f^l+y8unc4a
z5fU#s+_qjW3jmQ;V<m^s!YWuZ?@$_D_#f%!LZNK9dFc}^wTJXGQ8V=b8aO+zE0i)>
zbB0a%h;Z~_)Pg1PbCc58+gEq3!ddncBoo7hP|_xE=r<YfWn_j;4H*bn6Y>Yt{f!(o
z+;}sK$&OCpCrMcDud^2y+F9riBFf6Fl`{->BHv-VHq3mox`v^gQEJv~SSVP=+@&!q
z;j<HC`#`f*S@VYzkllFcU2yG$F{0UF%%T|b4vGk9gf}`;bxdd_tcmv!h!IKZU(!vR
zE1<3*7iowy7ojKh_`!w(!0>15lX6|O4#Gr>sbumc4@o60Q`=byo&=0z6}pZxZ~dVk
zmzebO<*gJOX2M~1xqDkc0;F)HWKYO29O5Q@>M~WC6jXs!-T9`L8i}vsI4=KWUj+os
zH#R!kR0|@0dfRqC{kcY8?r+*PUu)vs&1>OiHPE<=&(hmNzqQbzv0;;bYy;M9<WW&c
zEi`!1@k|1cnBh8t;g`jrT0Xz0TugD&p1~Wk;<p@7EuniYG)!-p(x2c882COV3Z*rM
zuWtvP;te}H;`!#uowhr?im*|$FiKk*bZ@51eFm4DMr0Y>(;Ctrq~Ihi>=fs3B+1SL
zHjDjM6;?vW&YoZ(P_W;W)!M!-UisSe-j10_`xXX}6MCd8!Q_biKWKZ;sHWC8it}Dn
zyda>I&_OAo1VRr82wW)v5^6~3LFo{R5Q+$jh2DDyks6ZFAyg5C(0hjnD7`BnAR_pm
z+?n|>UuUhE^YN^#th2IC&U>Ey{`TGrI91tgB*uY1;6y9_K+|9#&P!T-w|`y#*@(Gw
zXL-0?TF-nl^V+b!WiMwNk-s^{<A--)C92$thWFV{vuyGbzXY)8-&b=I0(DIhbO{Y+
zC(FCf3{?t%dMTOZQ)LShLZY|pGqWOV%r&C|0;%rzkQu3sN7Ea&qnQpO10x0V4KL6w
zIng;bb>H%KFKUq0Ax&!qs1EKNn;O)5H5yh1Y_WJ?869k!#l9Dlz?Z2HlhblD3RgRd
zRIBg6$++;B*c-@F$0NP&`QQQx{o=`u^=;|b0!FFf^K$jIg;YI~Yy4}MhviHhyU+~l
zsj`JU8Q19g8swLTqPtWUl~+~f7bZ`4uNs!DL1Q=k^%9}Okcw(7fdHw<XTI}YC&`&|
z?&WOF8Q?gv_IVZ+=JELqUG#5sE0cB+<%}c(WcpIlgSD9iD)Z;)hq16qqzU|@_X@{v
z75Ge(0>0m|Ae@(X*pScP1cQw_qIXY0r`~1MTb{eff;bATK;nv<ZV5c}v}zrQo>*06
zyYDk`1Q4i8(8fHA&o3pZBg{38SRcdm;s2&*yz0R=F*G)ce0ohs*zn7!U6HNS@ynRi
zZX;~^)9x(%-QaNATCF;`Gx#FHs`VS?z;3f;_ZT$SC5-YQ?S{uz*_B=hEqT^P;gG=o
z^j%Bmazr!Iuqm7cfle+B@d`njxGS^)gbcD1p=mleeS|yKq_p;Tk^U5Za=diC%)SlC
zcLYd19=F$YAcrhhIY{Jhp)I;x{B2%|YzmDGujGg|u$mSR>&9-U<xV(4LyC2x_Y8_X
z)V}#;_QwPaFf+tVxJ?fQCg?YB*YV&cDv<qb7z5h$)!Yb!avcoYKMDAmmfgGrKxB?)
zzAx2#2qodID=zvp)RfHL7T}G0{fAFAy&ZxB?YF0&lNT3VhYg^?vQ-W$l5R=S>mJtA
z7E(l0$*6))5lq*`Ozx4m=ou6>!^s^~JEmzo295HZEfz$hup<VZ{qtiD?6z~ea&l2o
zqa$hJ3pM>JJ?C(&*_ilGJ2^v&J`lVi?xL?R9DmNt@i!&NZi#=^$9ZM2YswDj^j=Id
zD~e9o=Ns{Hw}Tedid;-ssMP-N`*W(s{tAp$2-+epMU<kxVLQ+iBeX3s;L*hvrrHG#
zLT;mC{Vxn*D!?g1a+w*f$7kAbAOaF41WRntOs>}&R$0R+h=T-8zf~k@NfBVfeW^K_
z`5DpM@`#kL!$VMvU>jQyGY>lhv?Rs4+6fBjPt<{zyWvyz|4Ck8=ULIlsPIBncqq*c
z*=I!C8B*pT^;OW@F>-nSM}yIH&G@&t5b;1psZOt1anc2!VQj}SZf7R;!7N5$V7P3r
zY8JV#UpJpjL3{O%c%^`E8R69@kyp(fggrJ&gdncmQ&-85Fkm=^VM!6LJ=V%edHb5h
z+#9{e6yH}O^luysPezi*1TR&`@x*^MQLrf4aT|BJM_m(q-|0kgGeF7t@9)>XO^Q5u
zKUhXND@onP`;V~o8?Lz9G@q9FTV_Rw#~zdge<roVc8Xy(2fdXx&3tl({LQfBmeD#O
z15!K<sQt|Mi}Oj-aRSc9C$)Z0hou48oskO_C^pi@%~nA4=|==ZZ`rp=CLpqu=z}$}
ztPb_UqS~2xuj%D^Sa?>5+0kdpR0U^vx{_|eDo4D+D?vhuMxk<;1I@#Jovfspz&JRg
zE99URlG{f3zepPnd(ZTRiAau)k{EV&b4eS_&)RP=@E5m5ncZe${q|E}yGatNc&4~n
zZaa+`D}MS<-G*F@>=%0@h138XGr_nnDCx5X<t6*LgryBLRcD@Y<BZMH(5^BQ8W^{t
zErQ{Pw9^XHVz=3%P4k#stQGX6{l#imh3p7@B&r(L3g-N(;=f*bnaPuym7kg&29e<V
zPsDM#r|oGeYn_<whZ+vCUe|?K6LN~nb)OzdFz!EpTwDEKhD;0wF4#X}a}PY7^D@TT
z$Pkh41H$mvz<D`A11)3W==6$XEB{W1Ro!<o)#7AMA!cfuHvr#P;_yI2gBAxR3@P4P
z>o*>%Zsf`~$d2H?R><h59Y-&8pJXguVYa?oZ>F#tPCKG+r0GwW^OQ;|6=|GjZ#d?z
z>wF=s+qoM0;G84J7H^YP9{~<=l1-?r=s@I8P#|f-5Wg&KR5BLGOYOT(oYCCKq4Pj0
zr&`LT0M@<RN`zBmK+?|~?^1WG-PP{d=YC-7Rk5tGs2vGL*JlK4PnR+AzCURf>dNsH
ztH~_;H*vN!=uWX<cWj@BtUPJrIU=v*IkWEDOuU)TNq*@`$*geW@clH{xv_a`>}JU(
ze-GM1S;Gt^uj(Ejq!elOx;rH6AK%0Y+?OD)pk@QiX5(&+vBM3kHZ@Fv&lIZw#i>O;
z*;E1<>v?uTXUNsg+*mm~QAG-Ny|$KbRpMy|>z(H($w&DbGEO}yGk02$W^vJ<nq)-|
zkRk_K<XOI1PO>&2Mi1_N7>?iYCQeivixMW?X9l15ND)B;l;@LL`Y~>V6DMQ@H*b*J
zM&zc(NpA$>j;n|ld`T*9mVtF4Cq=T=L5-Ee2QH-;c-KL#X~a&AX|EJny}-J&Kj+Ts
zh6-H35PgyNDsTvp{Nwh?9-qnS2z-N9cm(dV<!?$S<u2^Lb<vx;vviN_tXn8SeVq!^
zEoZmdvP}mw++(Cu{?CEQjdEt^E$7Du@(6iTG$S)*m<lchrj#slJuD1EmYM*CT{k8r
z9uIdtAMA+XX8qcut(USqHb*Zs-;7z`pgv;)CHSyZmsV>|_f9q{u?Z;6TzgSuS~gQ|
z1Blx7eb-+xQ5v))g;**b#TkeS^k1Rs4btDud?JG%+^qC}R+{fy91p+7x&7nLsQq&|
zSV9R%Q}v&KLsqsrN5?O+*9Y2RXqp~Kk8fDQ){R%pt6jg%x}{fY30`=)z<4j?R<&Y(
z1reDm{q>AyBR4FKBlqeSaJ$WHW-FFA8VZ42ww-{O`D<Bw#*ETE<GYHKY8Y5^CrxHM
zz}3Ff6vu;#y@H8qiBq0Ks_ej%DYXfGEYM<$+Qy^<4P1xR%5+z--lU-`U}d@jjzVN<
zfb}B&2C;Z1F`2vCLGvW;+T9XyB;hH$AZ@e@q1fiABK!6~D+6DK-O`n<pYMK(k^X+?
zk<8YaEGmwLuedk*v$AE+qp6pZnMn0d*`D=})jcY619>-j7`Nw(oF=`4lrmJlRO=Ia
zV%WFJrz8~IAc2d;)9$H>+QH7BtRlmuG_w`A+B}?%fMv}qD)N)PDGJJua(cMsse9wl
zPG$Y<q7fwiu(Sl4pr4zU@-{owkE+~!Exq61zrZ=_XJ4xA-_BeEB~rf6RIltx0n?)d
z(8&>|X%R95Ej~gj5|#_mKC4|$NJ~?JcB|U89*khCjce898S+;(8hbNsu^f#eewJ$C
z@K|x4@wM3_d7yIHerF<MBR*_EZ=n7AsEZOxs4N5N?gQkeI5)TH^$YqZduYBV>Mu`n
zKbeB?YrZbM5<nfj;LkyLIcP$Qq+wz7)}_=(vB*f>B-<}NA~Vos^YOxWe`;X}+RzDu
z+ZwMm0(8sE;=~l(F`mOgx4_(z9K*_Euj;q}_5><-({vzEH>a)BZl;RnSn}gZk8gEZ
zj`Fs7meAG44l+3aRpxHA8dVQfkzF!SIjm%$p*<2-2A*j-F`j12S`4T+W_y=$81;3#
z(~#R_yh9U@nh2Gtczh9oF#@RLA;w_>Tt8BMgnkc;F|&jk`6f6D{qHYoH+LnrdJg3D
z5*@-|;`@M;<#gARJDtn2I)R4wsSZXn_H0-AV82+)hKL<Ofm=>5Tby2kth5UPrt80i
zu`jcYeh@@FUc!pWdvkeIX%}@UU%=&H%~pgWy6{sYEam==GfY`xhpk?i<s}?uH_q)a
zh3w$AfLVUS$2%Whx-Y~kfa2+$*!!Us)cxmuJ-Bg7nuTN_3{JWtkhS#;XP|_(OE!kP
zk+toc4kqGqCw-%etRm>YCl>4SsDr-VH;SaTp-%Lvq0LK{<zKw-2b`KET04K{oDO_F
z?TEW;Hul`^_-Z-lQOviSAHMi>90lQ80)j2{ztwc)au~0n)wf!QPx#n>EQvD3DIArJ
z(1nKlp+D*{rlqENc1m3SJl<xT|B@4G_$z7SMe`<$S?-6eUnJdz{SejPQ@;&E<Sha&
zfl;!?ZJ{q=ug!k)e_2v}3VmY#9i8-qINuW$I<@q5&%@(oRasTRFDsvcS*zn1+P(6P
zPC7A8GurbE{epnLcf-{~Iv5mP&OX8}HK2`z(nE@1{Kgk_&BRUhAL+s69^$Au{63RS
zx#eP0)`;E=Nk`eJucJc~#nX??-nD5mK0H1WweNyG&(BpB`hY->9UugYySxb`z}(C=
z`jvgDfoy+zTY5)PKDr;^SKFg{qi>Qtfls~Iff+H(+qq32nw}0WcV;g}ZwM)#4q?}a
zz5n@qA8*@;N4N!UK-*Q;qEA5?a*vV*!@Jxv2sUu9otH#vuFGC{HJz2{*QCjbmyOU|
z1!-df;^a{hVP3t^IzPgM05>b!@y$_Irn%kTmGqx@!#`Jl^VBVQvG=nF%J06!ygY+L
z(71@i^t0e}cT1x%8sVpXn6B`xz19Rh)1tDBg{ldQf~0_<YE=F2^dW1K)5aWwTecNu
z(Sqh%?hJPnr~w4g{E980No~s`Rv~v&SuXI?{mh2o2cOk=o$l&<@IZ9l)Tcb2G|wTj
z0Ioz=2QD93mQC5{CcmwCm=eZrox6a~nH`j^XGu+-<u)`N?$rt5y>hEId|@N_5`Bwt
z{BXF@U4>DFQPKKAKu!Yvf09jN?E;+xs3(S*vXrl=L{)6eB$kYGH+fvE4I0oPW_3I=
z^S3ao5eNEBj~5#E-n6Q3#Yur14DquPGVYAc&>Azr{Ge0SHx074Yy;H2jr}u^^Wb{A
z8E#_~N2500cLpmjxOTYf!rfRi++){g6br%n3A7Mv3<3&u2pBEI8TRVcIg8iJ*v^vm
z+!g{QEE6qSzZBYjoHu3})heu>a3;tnmo11hJ#Za{jATWzNSbSEQ<<9p8<qzaxSfvW
zW*h4-k3LGae9g3wgxDwkwQ?%TC<&PZDG64!Nyb)IxY`os8DkT4ODmBe#D%+|*-DYS
zxEeDnoq}Q-B=^0j-i1jY1x?nEQ%H$%^|cd(Jtv8S(hf^kbYai)*2o!|(<zx5ukkm2
z$S%(8lV%*G8Zwx{7mYARAtvhnM2rx9p>Q3w;paVvwx6n7MK3ufs3v-^m6K{OeKMJ8
zi0e9<bD7oklZ}@5wcx`Zxj85bAjPNK3_icDGi_Tl3N3&ptfsdRF=;+iMXWw+9NG)z
za^~)<tC`h8^VSc}aA&V4pxm#+8lRO?mN~i+Q)Yg(;&O70ZuN(oxYe^M-`rmmr|AIy
zpw6oPI#fkxSAs!OZ{=5pyV0(Nh|2LY*R5i$D7qiakL2dsctv)*y^8uAC41L1z0Mc+
z&FLO{vu^Y;iDgrJhfqqlYga#qRtd&n5d0PWuZ=_7&GXR2mMM}IiEa={AwfOof~|~L
zY?B+C^gNTpFZ%Af-{f-FWiwv0ePNeNZYjqz!2p_ds!Zl=<Y5*6bb9|ZyMg;GliR7s
zh+5zR{sy5kat7NvSp8_J)O07E>mP+S?NVi~f0$yo<;F9<IYANtN5uq5U22SEwZ!si
zdzKQoGKcSRFF0%ApJ7^I+xd?8XC4jG@Aco?mUsA}1FQ>{$rtVE!4Q?@a*!5sD3yVY
z7X}e2`1`{f32iSOFF3Hx!`F`T@z3m^i8jTq=}OkuCwO#si*@8e6QW`go358SY52A<
z7e7*}G%~p<!uQ#Dbi|;&G5%Dy7ppi7B|QivQo(TB!fXP{=d$!b)7`$3JZv!{v|1@<
z@>OeA&v%|Y6GqR&GGdUM)wX0ru{UB{`F7KOqb7pW)Z<hA;ql;`4Y@hePW4DABmcDq
z;c4CkK1w{k!8~F(2lzvo<*nV1K|xdD)YOwv+1~qkdYrj6@xBRm4jnPXDaJhF%w`S^
zFZrbVo@S<}Wo2dr%{ANPdzz6kQ~-iN9{h<q4tAD<CJBKJ%HeVdsA(LNwKaE#6<zF4
zHS3v>x2IGF=KJ)<=Z`zpK+ocZt$2=uIjW+Uu_w6S`6PC@J#w$rbw*mTe3mGm4$o4D
zn|KE$kM9i%LPQ?8SWC}0rSczleVG3<bzg|dv^EnkxfGfZC36|i?aZA47S()CN5|zq
z69%UH!InL-*zeY)Lltf2Leh_5kqI(j^FWsplkQYMvxXEjzl;5mNXk)hr;-|#Q6d3e
zwoM!9KHZmKpX~ZAIME$KAHz+lmsItRbF)@F0S*KLnLan;k9LsO1{NY}BG&V-jpO(w
z;AsLsrz2Cn+g69CLtm4FB1v%`r<G>WTdOl3=dnZ$n=0g56RQvsL{iR9OZ$;xL&GZu
zrs7fD!mdqhVm=dNNJs9^g3q3(r&p;MnQOvsIj^smt-uC9I<#8R+Fj#eViO@|jGEUa
zc$nI5*oa^V{!-r;7#jlrw0p<x9u8(i*|tA{$2JF|0j(&HB^D13Kp!H_vTDXj5Hd~n
z*D%2kr-9dXoih<@gCe<J^{`uY{>ENUFt*CjQu|$z>?44B+{1D`Q#{KZC#q7TH@SmN
zE_D)KQ33M{TWDpbzox&%$0D>UVHx|gbq;%OI~!Naus=M!KFgJw0HG2?!^5uHdAd4D
z{6f^1({;~52=954Uk3Mo*bS?YfZ{1l62?R<u%qJ!0ZsF{WJ-)p#e+V;bE@Jb$iNJr
zY-RIJ<v2HnUY_~0*=Wmd@&ZpH*af5Bh}<=siOtZgG0T3w1S6^`Au0^XP+|R?o$Vu$
z@=g9^O0UqGZXAnHO1uZNZN)4}GG<GLdN01FwZ6u}OBZB9jjj4j;T1E`^O?^=*+3GY
zWiwPx37^QF+1rr=2#&uX!Npo$N|hk<KqcyC8FBmhX^I}-^fLsU;z(k^_rI&}7i%I?
z5lI2U)^`qgYnqiv@$r?7>D9S)l3Z4LcbuKyHNlyP_VA%@Pca!vb#d!p@=9Hm4wkSW
zkFp)3i777N*NHx_lX)LktQ}IUHpGIZDweie(ib>PD+)Pa8!Zd`6q+ND-cD<v`?!-2
zTCA1A?`*o{G(v^=%m>rM<V+>BrqgxF0SGp-oBW;4BC@%GX8b_3FJf@B*KtxvkV^4_
zN+$SnL1=k6|3^RWO%<NheLqwp=8GZ$L+~Au`|QVqL?_zFS5bOIz7}r?e#Fg+q3#1>
z4bBKZCkrCX%3&B!Ixpboy5974YU%qiUGMAy{yJPU4S&~t>#C)LNWjXcFI*`&=;v^U
zmy+Yx+t+K4<sOtSa(v8wv*FBD1MLZ8-p7(gutHg={wA}J_qK_;<APmDHyQN0<7kBh
zyMndz+Xc0-YDw^60=in88iI;uh*?q`p*?MqQdQh3Z?J5Vq5$luK~pSj$3io1klp98
zD7)N?wU*&<d6hybZ3|@g$b)p15_d8cB-rB<tTKH^<?<-n)B4;b_7b3Z*=QG3MX~@|
zAWx5lVR#%MhqJ=(hdE6xQ#j8gvZFU!n{|G)<^;*79&L!&ms&lmkagY^P}!cZb*vj-
z?Z%Q!XG3^5t1{S!Qi}`zu2<CcT<Z>*94Eu?f|`IG!Il;}G$)M`hr7)UsoPI=s?1|-
z-<b<GgLYB?Sn&)uS&`ZpNew=u+VuBB3AxGD{SLRiq%o3a535$<AC~iKnY~+8(Se28
zl=r2*Q)JXku}DiW#_RR^b8Cx@L@0^Ma$ncgjA0U#6D0EYSq-en1<QOy{%W=Q^#B5-
z@rJovU#twI8l)cPWX8^q3zGO`Hi<<&j&{+TMDG}XWdC=lp#4EkoFe#hWKhh*IRE*q
z8QKz0bkq745W>{_kl1GnOHhO-7si1Z1`}5)zNV#eM?k2XFKE2a?yy&+`s=V3v?s`+
zggYhL=M-vQr9*%>QPmoxQ*jg3Hzs1Bm2hL?UXO>}G*z&nCN)CLgoa($ok3VX<gata
zoYed<7CQ`jciqWN*19T&aj=i7BVq_r#FQY=-M@#bb?W2N2g|VSS)MoQeA+s(pL9SZ
zA5(|XJRif-qBG}^v~L^O0COUL>UzsWStmS3Hk!q`Ck~|j^ih0w!fzq{KjUqfDTOli
z!qNl5RbI4FnSlycGf#ut?uMeB@A#LSUtUV%h6V<D3OqUM9(U&@&<K;oWF5Mq0JW=R
zk4~1bFrN=xG29T^IG|yY3^|JX0ei>M7dimnLf}Lpi;%rU?wkq_@2v8ONB)eTHU2a6
zPRYRk*UrWNcSgwnJFd9v+vpdfD^I(01&HUbztsws7qt=>FZ%ENH51k)Ke+7Z#~S@i
zqT(y`pH*+EkM)kL1L#R=P-Z1H`>C5~Ayai>k6J9#NzR779T6$rN4|GA>kk9!35jIj
z=El0O!P}%KGC8Nn)yYR^Y@f^k=PeMoT3N6+G=-%+)&KGllb2JTGB-<~{?k!dy<Ghc
z<J6s$!QAbDJ;QTM=Og^%mT3%f0|rnB>rB;TL{Nm=9r{Y!xXJa!<?WAk%!>sHyQ)H%
zYih=($8;<Mj<u!q^vLuokmh7Yk-GvOB#1y}u*bqugHuzB<%Rw`#(Q+k3Ce_u|Bfak
zas&X7*P&+BDPA*^SOYE`F0kfH>JR@JdmxyU{l3c4X`}jNsHcATmj$rh$hH-*ZgazX
za!u9RTdfY6W$}dJ+ixgv@IWxlB=j}b=9|J!f!}`9-Kpk=&ZhEkpX!(w<+7z{PWO<%
z;U#X?1_W|e7#u&M$N=&q3?cGj2X6Kd2Op*ef{M*c5DTixV5Tw{UUuJj=wd}a{3tU)
z7G=!*6@%pd!&{}srxx|D)qVs{kksq%H-NTo4+M?H*2|XV@!5A}KW}u0cvt~07riMD
zNpTO(_+EmaIW6SH|MIJL0JEb?UO2oJR1WOLEEk&#s>0gvC#vOV-HRe5NlIM&2r|&4
zVV5NFI{F4=7svv1mlZWTm$_ZPRoB_DDTs47gZ-Il;m|+1l+~amMb$=fd2vhZ2GEJ^
z=a&pmi?i_>#_s_J%s<d*#~5X%rI(X)_PI<dpvjh#KU4^gBS428J?gG_38jO^A^2wd
z$PA`B=hc8BR3}U!^{BiQKW<NU(70&XqvG$=Au9u=+L#8Jc9F9{vnH%_-F;uPDzk6H
z)}0=qS4VGlOGO@<-fb~r^}j$jd-3)<j$l8N#J$+=BR2bzM2Kb?R@G8`5CC+?Bv}6-
z<v<y7@UYByN4qq9gooCC;C8=^A%B3q<fV$M_ph9`GK<<6_G=hYOmk(>q4TOvn5Z}{
z^kLm*f!MEsBv-e!Yo9&_J!>3YXn@dF3@IaXUFtPs6gCxk>_6%*4A9uIa?$fm5Pk}+
zNmy<ZdRO<HrEKJHMyfhArLB$%J$@@TKmRB?z;m$O!GUqkaS_R*a)enO;bYo(*Xm`9
z0zD5#I|K5bWfA>*{!RU5Sk}0{y}9m>?>hL>B$9sO5%(fmEmOvuqpwBxj>Donu(-)V
z5~5o(1%eFn?BohnAP8W>6wt+CRId$Mex_$wIxOZv<9m=5Lj;)ODjVIS_R<ePju-h?
zf+eCxZ)2j_VZkm&54iM8lcK*4KeP%u;+*$QmAq~_e#(0g<W}SLIj25SbY-R{?%eC;
zisAR`1oJ?kRWmro|6jF9W#p~u1i2`bI|(cyvWyfT^S|w#n-i25ph0)taU+t2eEJ*^
z>$55Q&OoxtkABYGGBQQB;N7&_)HsOc2^L~q?Vy|G>NZrYmOE^lk9!P~W0}#Xqj;{V
z>RVG+_w*r+-0Q=lpJHSDZKO(OhdeWfTE-onXHDuPSQXz{J<})emSmZ8>|gdWxRTow
z-@>%pu{T`0+qaRT^aVv7d?w$Y8<P%?22P^BSUXK-S4z%X=e?!oMMFQ#aeH*Nm@R<X
z%b$yiQPja_X`k+M&`)ln_-JwRLeXv{`Y>}>5gok~22l(0a8^MSBat2}92WvDDVtF-
zE2#u)m`qBRb5@Y8aszJ?Pz2_c^=*bYAcU5F*9%jREs++i39m-_i}avT4P)csFLTr0
zL#y-n<rcCZh)`l0pRm+vTtCviUHkhEsO959LWrHBT8}s7XGnl&H(zk}MyoBsY|V`L
zPlc&Qj~nK-z;TMv9}kabNs;O;Ay3?mp)ErTLJJ|c>?K3OeWifi2uomgWSAD2ZLq*D
z`j;A4Q-LC-Oda7OIUQGqn_4U>CExd;cfE+Bi0-KD6c66O=9Joom{d%H<mBWhFma;F
z5K;c?jgvanl|dfT+(JA|Aobg)mDAWgtt<UZ7jz(Lo}*XXb9+SXb>wV9U+5zaNCGUv
zo1&P(9}v1$ZGbr?Z7Q>8=$6Zh_-}oV;nvO|aYeCbat3sR$znrgA9^j*Zm-;SKe$7s
z6s*fu7BygG`IOxN!t{QJN|y1f3r4$Z4x>gR+7Z&yj1$FO;Yo4KRn(WIX`5E&7e%ED
z;)99T$f?L*Rxt|$F=C#tB_!yMHigtgHWsU*=(&}@V<FYp5>pg2AW*PSVBS3<DC^B{
zB^i&~zJZ_G4@ZBO+GaNk5?#M2UK!1|iVvfFp?ml#yBugGG)7(3Y-h5>FuK_i=A`oi
zM}a7=B11jS-#9XeCtl%wM}eP$660bIPYxt*bfs0TT%?y-XUMY{|1+^6l0y>=lXMTN
z$;=63x0APGHs^JF5_zHvG1ZXJYNSMn{x#`@Ni{?Mnl}UJYG>rXX0dbNaX`8_Xjyc{
z&URFsdwWm_h4^Y6gY>>9;o_OYpBzVJ*J-QXWQ)h%uY=bSEz?i8Ou9b{uC&`zs*v9K
z+-ZLgy)k6DK8qL2;ZH$^%9Z0&pAr1$cpxJ3<!?tqxvkA6suK@28sjt7s1F9(kyS1R
z38n~e4Dn=8GPrxgopYD1Tf7#x`B>d|TyS)tJ2sixqW;5#dr#6_@LVK+8!$LycAZ>`
zphk0GeKEHbe{tRaVUVep#+D`P*mjFQIprXN*`^xrA9-q!TkfKmxE?x1xsgE}`&*)Z
z#iTkc6!<0LX=(qczV{G0_ZA{Wo2dTzI&MxMuUUX>k`kHsTA*(xk|14fyI@V<5AnaH
z-TyS~e=ei2r??*qO}mRTK{z<Z-|2XYTGWzs4$&H-j7MO`hSa)^4bXI-wx0mds6cG~
z`%jBLkB+(|pXW_V-G2U-qfT4}E+b|x%QILQ@?J?Q0}8q{L@#ug#ZCnRSPFNGB_A-y
z=3Z=cD*d(&L|KkF``rzExeCSjwBNE-q2+ck=OP?tvE+F2)N~k@>5y%#yU5mj-5zo=
z%+!_R=?(tbHt^un^Dh;wUQ;WkZvXoWrchR-;VA9N`|p*M+ZsjpmfK4Xhs`oLz~0s}
zU5~kZmy*0PVg&65JlBZ?u?@B_F$ob_{TFsS1o6~pe`Fb{P@X?HWa5OutaOP-znr(~
zYs{s-5f1bbH!Z%WaMqpGR(apBFf+9<jNgiRMBmMK1#a*$>Pl-rd4ZM#=O9Jr1^(QS
zcy9MT^rLFCaE@HpMZ*sZ-S_sW9l)`7aU@?Aa>t|wrbF~w)%p1dr;GXWZn1J9J&%6s
z748chJe+3Xuzpo3o79o<^0pek!|e8&-n5sx`xN_tR+i@}(}VN_C2N%30G)pSi8!T0
z^wtW+e_D5%pn#im^#!=t3v^qwT^AQi@7yRXb-!sqB4B5?(=4^<rixql=-pH7q1rLd
z2`sE4)QcV)?Teg#O=}Mn(1*)?o6)7Fr^;(IJ^9<az1ulb<g1aL&xiH)UH{Urs167F
za^BA*)x_GFX6IGsh>3rT$MMXYNdAa_o5kq^Mq{$*!78_sREoXID%&LX!KKz}tdyVB
zNPGS4rY!CBJhTAvhzZmKP=?vY?Pe^BN1txG?y}b1gUdOTAq${N=mX0sGe$>d+q0sB
z0!dRBm&PZiB_{YOR#SmLhBUB30UF^TytKGUYQw!x;d32UG<Ja5{<hkALaQMDY$PS<
z#S448QG+V`KR^lxFzoqfbFWV<D~(dS4}OQP6~kWMk@_{<N0`|4gcjb|vWF*to7QV7
zbNpITr3p;JV;?&jkEHl}`KP(j<ZMJUoE#KNv@P!~H{WiUJf#xToI&1kR9{g3NvU4_
z4+f_o*~hKpRM#du&pfemM*GT(qkq)Qhn+X20yeB~1Rw14d}dW){95+L=-b?2MPac7
zk9op>BaU{vZCss6|KaY=*2ll`>cM9SS2Kx)_UZ$g1sjtzk&-}j2@bXI5emu*=HAv|
z%x-q;&-dKuDj$J3$2+xMyM^uMDXuNedkQM;2O&{9Df~IG3e(Wh5VhY9+jrr?035YJ
z^f6pBLIRqaQaUc1pvm@o$M2Xeh%xh>fe~t_I>x-)vOY=r*?m+~6K{kjjECC0^)`at
z8*BZtGH3|e(#+%#J^uwd=2vXc6DO-|_-0#QmR4aX<41fHn2*ISxIm8Dp^OF*wrd>-
zfikx9KE)3Y4GwxJJD!LM%yPTb-lQzwbsrnN5bY11e{t~KuJnkG?`r)S`t>7ZDj{fA
z9cbwrG6q$gxA9WCVAi0f^E(VSQf9@}=h(I_(SvDB>4EnrUO+Jn@3nOm?{0Y|iPO+S
zEKt?CrR3=82Eme5$k<bQ_b3?4;iR>nu4#eBB^|RkWSb#ma+~;#FrW0|pq{r&uxh3W
zE*Q*LltM)`ocwM^W%o5=m-Dg~8&)nGN6J^yV;}5my!E$=4Fk1eR(iQ-HfO$L)QrAk
z(+raQzXWB-w#8$78wuLET~tsgSTx1_>fKe^izo$;m|BhcD<j&}#;TadlJ93FBUGhh
zfRmv#2A^=VvQSX!Yv99z^kPM<H(2Ply(LpdoLrLrJOZu;F-y2av?%EXn^)`2`0n2v
zKRk&etxOfr9ZT=qRr<TT(?7T9p$^3E_FA692M-UT=L0Q`;uf@CEfM4)Z#vB%;2mcl
zf%Q&;Rp(P)wWLL`$4TSJVhZ{{!4Vdug9xy!kj5%iZQeot{HM4(q{mT5DpJ0H+uEY<
zfX#7iTRpompp-zi+S#=C5?t|gw2onXv00)a4I?8^!SxWwo~ZAve)WegR#*OkffEij
z?tr$zll@GA(mx)NM5MAyS2&3tteFFvduc(-x+z>d1pi`6#51uj_FPG#&6`E4Dlv2Q
zN~~wO>xxHpSc&LowB%peTEUz23ZDx(_`2iuDm1&>47Wh9sd*4*HE|AL3r2!@1fPG)
z%nxFR2}YRn`*hi0g?{9vLhr@FL0Nc5x>~VYP!xrO;Wu>UKId#$iCeAseQ<Wkts!cp
zaJm=-)FG04sbQWV9_uXOH<A};q%Q(jR~WX$);=E6Okz^$*-MNEsc3%kLj);?T%Udf
z&#Q*~fIyNf)Kc%1FVR@Ls2IVU!0N)4QX_S<f*stwmAcB`Jo1POs$<}su$&c9%bdu#
zH$f#rH1{U$A@ZC#b>Bo~W+f^Hb-F#k&EFNwMqf=%Vf!84lvew4p_aVgI<LXuuDgTm
z)EEAnhjnBX`Gh2&c(9;>QQvb0X4X4qT(Kpx{u2h3LPg^Rav2=OlSSl;;wH4IQ}#~!
zA3UHh*>GGUaq1)^pfNVCqB4B?1xk3WN{x8MnFMw<EKy~96oh>^LjB_r;{O2a8clxj
z2TD%Y$Z^q)NsT=Jj$SxkAU=lnGKlqXCp;Y18rq;|xahFUW7*ANAw8u`dJ;U6cLU;7
zoF_Q<&J3;lu!fla^p=xWq_&BBY-sD5tImgs(6Qwi?!SwFtA-#~^;(2DSGe3=RuX_n
z5#V}hd5P&)XmT-X#iBfq9|3KpN#i)CT)w^MDNcXMPR<KcC($)V8SepK-f-l*yZWor
zz2uk3jC%%WlGQFRySk*MS5An-2wVlh-6ZU8F|5>_7AI1jAhO}9^iWUcwB*?$g7TeY
zCUa)^==$+eeLH`x8K+dmC?!DBpk?Pg!aB}gkI=ZT%)0a8S8X1n7g42lN?OoRaqWY*
zU+K7$xby&SWScwnWnuV>*sG)y*GH*+ZHTkZiP;CA*n%o62kl|DqueP8(M9?lbt~|Z
zP*i*!%Dh$hXPv#_<lz-i6eKfYMelOc?OYTBhjHsy5A>9pC_J$)MTuC~*MV5?Y?UbL
zI~3PFy1gu({_B}uMQmb}tg7t7L)*o^1zsl$hvnhm>M>r{HbM(dG0S`ysx(|;E_-js
z7=plArDJaST^-CT>{5^`=E4s{Yfw}iE*j%qXj+A2rstY|Y~nr>bLDvU?W{KKET^yW
znSlU}w=vX`=K4t@yMF(ZjNIxmBpUpyTqp-efH>J3Xr-p5l@93Y8Vu>dC4+V(<V|x^
zGJkZiY#&SAC^hK8tMghP*dG>;oxxHB1~uO#{d^_@g`ZVBP(cm{2{7YPVa4b~vK70i
z&anpdC3Yb%<VQgwhd1470x9h23sD{l$rsj2op6U}R|6ZKuB^Ag{cQ-&Y6P%;fLN`l
z-ka<+K4k-P@|x<7vkN+G>Wq4~+g>U+a+F*ss&zqZ{d5XGN)uPOL+_OvuBtc+yS3yp
ztl_o~MINbvQtMotSB53s81$qQ=xS#CC~+(E>aT6nr_3cQ4G|!@6MO~AKNAm4OJahl
z^yDV{3rEOp3UkIWSBB&ePG_O5h%+;{qT>hqRzzHwS?2ps8>>wI%t`k}WAEQ@|3G{-
zO2L!c`BxhNzDnAzT(H_Uaf_WpqtaNbnH^^}Qx!yP0H^j*Nw4+t#mYIJI+L{-W{<-g
zpC<Vv-uGgG>v>j60jLAnuGwAnONM&GLTmrwzP2_m;qqCu5JvJzVA8@~pmDS9RpH?k
zHhkDJgP;RTZ!i@jN@mdAKQ3r{2y4dFh6Ktt?*Fz7zG$8~;0o~oUS+wJk1#B*lw-v|
z8!9@f!V2SA*885a)jIzuGV3@r4o>@aqsM1zp@op1=e_lAxlxhg@QYNKqt}vGE%5Y3
zo%)MzQ_%{IjQzd?iy}|dyPX>6yB@Sg@W;_I#Y+FIRlLkL47HMGdin$8tmp!iLj>aW
zE5->-&kZyzqG$rMcW!vGSAnGsGR=`MOZc{bvzZ;)stLFq*?%BxK`_XSeSx#@mgw-H
zeZ3<A`0&HLH_;0NuIjv@lpl!i&z~;eCd~cdYR*$_T3l=~Id=D2r!ejh$H}w(xZzja
zsf$@^+Lct}UdwYS->HIUPx<X6DI2+<%Je@F4_fan7NNGM5V14zXW|r@OI3v%$k%1u
zk@>s1*z9*708jsMn@w%pnVa5CtrO@e8=5mu6QaGgB(22QAbW2XCv&Pi5`{NOefz39
zIq#3#>4Ad7In%p`U8(1$Y0*2d46MCSmmbRz_h#R7+<U0s#9c;I^e;=ZlnHc@tiAd~
z(W%F$SB)r77hN^~?5V*Tn$f1pA>AdX-uuo7GmeclZr<a#sX?&_<y^KHAm@6&qW%jm
z^|O_hnkS)PR4H-0d0`hT*)}k20is&6S}W}qoDm{?MLh{pb?x`n>W4lQF>!b-XwJ>d
z?k`j(81tS5X8a8PBNsV$R+*8No^cQm5CS(Efh5h$23EH{^dkinczm|<@@us5$Kvt^
zlv@XxLgjE_q<jqi>w{Y`KG;%Y1<B?^2kqCcO9SF+L3u#CT-XG)YYN&z-Ehs5-Fs#{
z;IxpzWWAF(R;D36{W7XI+h6>aCdKp2Z;D+h<s&v%JEyWQgJAAjy4|=F>kaNowbZQp
z2B5^MX9-#3nlurZYan+f)`>&MPoAHFIDt>cYI63KCi}(_b=BNuIz)1DJ*p14VXu=L
z-+!_jZ=K%kB~CNN?CdD%mb_dG7=<e*?#~=9{^>_sPUXG-vmZ7ZGGCQ0sr$(`5U<&?
z4;Q7S)XH)LdR~lM;Vc0J)PTq5pI(CCAqzuT8KMOro_%Ls2KKIXqGcs@Hnghbt0Dbw
zrQaUgAbzGPJC43wwT$etRqj9cLxsATcY=b8f92~6D8q@!im5t00xlsTVUUZ;PGSNQ
zz;hli;buQKG4mxnZtd(~<6Fj)60%&fKyjG9_?d3Oo8<Kp&Frk#5V{8D5WN(k?gsAl
zKSiFz+`_<lt?-tjHz6vu#~*un)!x^3oxcuUE!}BK_~kw0<&h|?Qn03iRzb$9xhUul
z_Z0x#QkZ9I)71x4*IY`uI8VQ?-i<%ZR68pm&!=5?u;vGM|G5A3=Fre=?SYX|4LpG#
z5hYEFVGv*7tSkG%VkeKLQohV+CP4|?XuRW@zq~=kqzrF(f;rT?;Qjz5?Tdsejc{XN
zz3}HH#?c5sRWRfL>4!p$>a;mMpX=iKWbml3GRDT-h;Aj}XPi>_^X`?a7rz~EgM)jP
z>|A1K$H*0Fn9<2?Uy)%Q6=6dBKgS~CsrI;XsFw%qt!Ku@qsUOTPbw;BQ3Q*lZ<5{f
ziZL9}pwt4Usu;qtg0<h)rbLjn0-eYxQt01TI~C()<$~08;())O*>;v7J6KqVQ_(??
z!hjp%Y&(?vDUW*DC21irJ4WuIQ`N(o*xIk8?|NSTP{=4L#cmqaR+MTGPWcH;N?Cln
zkd;AO<AH8^YgsGxx%Gf>NvA;A==3p_8C}4ci;-fTW*2gd2xg+aDfIs0h7du?Otkan
z_$piK?p&j@T}HIpJB1~5(X<oc#CkMSfr@<2xKGL8#7{w<e?Wv0edGR-kA7bb1`UUF
zqWJ9`ZhfIw5Um&D-E3w1aR~s+iiRClmM(JEPa)=m6KEXlRQZ?`{yD@SAFh6iW86-R
zV&S1|swHas*~ZRjBUN40<etgam<OIY9mw@vj;$t2ePx;q`T2Kk;`iW5bV>M$r%m5)
zmppgDID5+3hC6dbCK}7vaC0dyF~y>#0OS-nsCD|(8NQr<Sw`Fme04><O80+KY%BEU
zBxS!YN+B*mtg=dx6pyR5OW2)4Wf`x6IuH*hFo50Xn8v=A$t@>*yj4rJrO$c}Eo0eA
z7baMVWBaEb>b;IB!BqD0fIUm0{==R@1hQ(VJ5x^%F?mFKJgDQ@uIe6b>R974J?oG1
zt3JinphCOV7pPqX`BGy4`|D}`)c^hpFIl|1HnK1OzrW^n!e3tTxrP1juPWDsmv0gz
zUw${Mpf(c(Q598Oq0f-fU;-h?G@F5+Txm{8iq6JG&Z^(Dc?v#?0`NKydtq+uc_9dl
z`UD%Q^A#AXLtET~kmKO$&j{pub!_gI9HHo<HFtqJ>pD~KRZ;9cx40Q1WFIdOTSv=7
zO&R<nmhrJoIyw8@jLSwLJpxY~m}L}2?Zy@djKt1QnsUkYcsGg)rQDwHzEGS*DK^6@
zkS7Uh2;9IJeAq~jTFSE~=WWNP?Qiu|GdsRkP*I4Qj1wE%(~+O4Xna@F_~k3wir>$V
zTXPJSwr&Y$KaILC8^`hOj;YwuD!RD`=VTkuE;5`K%UNVtw3I6G6RddcgdmZ==P*$`
zv%6OS4tuR{V0Fw`F{DNK4LHZ{_~`@##xo0L6-E0jm6+Dtf6b><G}4cRW(hp1{@7&+
z^Dcr$7?zC~Xyq4r&RCPwfWiwwvMeqEj_@vB9go}s%#<B3nX?37Pe(^@XcU-1%W4Br
z22*!ujZ9Cu<L}EGCCM@-M%}O<f_;hu*lq4=&V*FX=qNTiXO=WhF(^h_+BeD#fq`OP
z<$3x7O`kROQo{x4gp0%lVF+#Np|0BTJ^Cj(dP`kcSNZ#Y1VG1bwO{^5Tk9y*8$?(_
zp^*E<f)M5^Z77MEnVJF;ovE$87@Z@?K?}nOM04qnh4#hoI7qo33xq)=M>d}DiK6bl
zmw}CynM*vq(QcfjVn`k-K@{6e@0Lv%c~21y@zP8+#K0V%-UJ<+X*hwy5Zo)Ffst9x
zxq3@2Kw^zJ(LZ2LEb})mTjC!Y)GAurU0hF(ki*H-+5u!w@m3e<(`AD0GdW@`Rpj$v
zyplm2`?Jal(@n^|CgCItoKLU(gdMn$Y$=z<dS|6=XjR2KMS@<^?T~Y;O$9~{qv6$$
zc9#Ggv1<lizklEfu#}f<pQ(GHA^+?ZrLYMbr#+a+vF`n6(3{p4bs@Qret)DxETg8I
zLTKs?Ds*(bFWdJYJa~nKJG3&k{hkdd(5_m}7&vKj$kXqO16@yAoPZ{DBzHvMwodxu
zI^Owg=J<yh;Ead964;po1g4HtFbFfm1X>`fq9u;iK=R7Gknd#HLL%+t0-$c~?Zt~{
zSKx|%ITY<dC9=t!_b~-wIG+*j2SL&V!|7D#dt}dlYL}-IR3SjY(Ox@wgvuc7VS*CZ
z;mYR|(lELdo0OR+D!PE}dfsYz3rV2mvSC2jRkj0K5Z~fr9jQR{P}6hs{AbKtXby#o
zD&KSbi?bEp)i>QkNn9m&1y3f*tk3fNi{cxLVz`qrmi1LE0jZMl+lTH7{dZOdOe^Xq
z-p*51D!nqe{?tmZrOy%{aMnSHhecw1-<zZRINLnT;N1YI?S=|kK++i_JH@=brIV8z
zVv6Gal@rwyUp0%Be`E@<o>t{mvkN_X_2i$iYrc6hT)?g&rE5Ob)E=tO_0*4xY*u^@
z^23<d&@b`?^0TbH0HLucxd<lDCUH{&D%arV*8NFGh*TM^%A%-Kea02vrHn$@YyT&>
zVYKZet#o$~C_$;1{z}A}7%RcT27BIPq%QxFHQD{pWtVy5D>*+-LP%Af30-4Z6DA?c
zTxG5q+&mfltYn55q?sqKTRw&ov0v3!j!r#hnJ6!jyC$9y?H#}EWBF5&$3Z8&C%BxT
zBE(I63Op`I@v#2dfPI)Y1p4uA2HajgGx*`rK?q^KMf_Ntl*<?z72Z?;YJFHWiBp)a
z0H~m$l7ttqLU87~PI~vm7cV9Ocd4S^ny4Mph=zMxhi1hfN*s_Ny(3biOEhz_AJLn3
zCo#BM<>RKEM1jla@^Wd|iO0i(!=FNmdc&Ag>{~~$px<|Chji0Fotq=Puj@a{@l@8%
z%*jgZ(~6^g!0kj0lyTGw7Rvc=b<%cB;hhsC^~c7#SB&;oPBb)T;?mD*v)Fn14Q-Uz
zo}((zWo{N9N{m;^I+;5CCl%f&{{1;&ky&PPt7^qvW%|Uw*<h+v127P;OSQb<e!Fic
z{27wVbqbaLG<p`7I^KA8JF_J8snMVO?jDrb$*%VID=qz`5kA9&SiBr(r|I#x>C%6H
z&ih)?2<v^{O?B&JF4u^OMC6zm0618?@ai5*j$q*QIYW~Ctt$)sw;jI5U(wT47|YuE
zDlNCh!*Xe;VZg)UJ44DA?O=l&L^W}{4O^4VV?Esjz6d$x?g{L)>4G<<X}+5fH^~h|
zUKAioKw$Me<LIHJ!Svs<C5_GO;f38At`{9DF}M-BIbsHvwX;Y{*%i;s9srX*tZc;{
zprXY!fdc`!ehoy`GrX@XZ4{f5H<MkS_j%kM!!AUbZ8am!cUddU-szD`O?J@D&h8Fq
zb$1WCv&16lHi1twdG{%dF1q@;UXq^`6A!cP76?*7sE7o;=Y7x6`@D0V2&lZocD?_%
zOQIT=tJ=GPv&l|!pRK%@-V3&4{YsJodjl2Qnb?|lupne`C9SK%WJ#-30VKS|8yX<~
z?e^-+Qr=so2@k$rY$hE=^7J-7Gh`m$F+&+hTP0WZ!>7NlNj_8J)>N2Le?aE+OD=OQ
z-MlOHv()@k?rUc2_}HFDHg4Pr=Ee5h=~WJSYVe$x=*>-mp)qRXW&Al)IYOXG-oDq7
zciWqqmrCWo3ENLFU`o^Oji^k9-cN^Vdkc{{uY>A1Xzo_qjmgIw&p#Ei#O%q-C7F^O
zRF*c|d`;LE^D@XLoK{8(?H1PV+w$qfo*@m)J03be?nWdiL46b=!b;qkcq9H!k<osj
z0}ll`eszCqUBi`Fr09$h+eVhRl`ym0?WgMG1QZFrKeCckedJQOX7LD$Z;Gwk;?1zo
zHvp=co;wIP07Pc*KCak_(4)Kf5!)7XQLp&^F7N2jD_SMRI;YW%F*}$n&$WI1!+Q=2
z!4cQ)$Q(c{Vj&=WpjCZ-W>1H@!=ify<6=4<V$Fa_&p0oTD>8{P&DCX*SZn;yOYtiU
zQFGtBwRWi`Eh`(b3P;v1i(1|H;Kmvy*uI&U1!2dz=hW?|@%@MZ&idN4=Df_zB=EJ1
z^YO92CtX;xdxM~%Ny2AxFY4|Y>je^t^uog>83TWRZxdp1;3@v(sT&&oB)UK83T$>G
zM}#8VC@sh-@8<M{tI2#54j=|0atnx2bHF(7<&&P-iYnxmQG)(6iV^oC^N#r*@qxtd
zh@gd;1%-viNP)cMjR6-O^QYqpYQJTED}-n}5|;CQq?>5tWS&RLTETjO)X}Ga#=V|(
zEIlhL?tZV)YNeWaiFUpxZC{dBwb_c{zXdTo<_|cHqeP{5-U}tkjxv-EuE3fSh^J!H
zqm$N92iWy>NZbuUwS#aXeW5B?A$ex7*KiQ(uha!yO7cT9|L-p;ifWcx)o*w$BGlPe
zmnI`jQRKG%t4ihp!z%vH=Z;>p2VZ<Az(a3OTf~v)l4EIBrDpu%A*UELcNzPa(&>)D
zx{4!FBr&5#>dbEPW!=mAa<$9}JLX9i$4ae)6@zm1sS!J%>g})G@?wpx7iA)0hWR<!
zG1BzVK<~SbL~4wm1b3OL#P{qsfe+qSsq6mmJ)qM44@<Sl0B*TSC%lpC`Z?4}XcV<+
zj&We1i|Iei?(LW;(h4N4YDF;D3TW?_u0W22pbKWsmo~i_4ZlH(qXL#stGSkE)tPP1
zL$C{$5#A^;oREU({_}RkCe4i$Js~Q~LKtDr9pT%k*UQqa_&0qa7rk=aOqqo+`BC5o
z0k)9niu={Cqh+g%seQEb7`(V;$>`v!(P+!;$HJ62ZaZOjrLhmtX9>Y=*H0SB9D))|
zAKeZbd}^WGS#MMGsJYjZU-gDz#{b5E<>`eJ8;glwshHbr4>kONg%H|e?3m{MMRv|{
z<o_MN_5bgunUw#ubnZWU_IGfL@OfZXR@C3D>bI8OK29&ZHht0hf-kC{jCIL0{2h8}
zhM`G$XUtR;a#{5DkMh$O5C6Ncf`BQzV!@R~U{%Nyl<luKr9?wGk=Q&cx4Asj`SSK(
zx3uOGqsGwwKY|~6XAbVv&vqsy3pK_Mk5%-a3Iu_GX~7_?hkP%Un4iAD-qexzYt)cd
z2DM0rIC9-)UN=9Y$+4~+x3@pzSyZ~1g1@`}8O24CFBTrm9rAup;W4|&8<ZF3xF)MB
zv|OK-zYiEKE5t6(i;EltrnxOKEu3!u%qT)OfX0dOWDIuxry`L!+UWbpo!svQUHJa(
z(3Fc+qfI*F;(T58@&C8KGSl15c15p^7nh38!w(8xWC{$m3;8F1vs65GEAueJn~n6r
zoaIXW65j=AI+s&K7{d~q*>OtauWnkcd1>N4q(Z)Wy=YJ`edKQe&wS6Y?RD2GluX*4
zKL1O(r*Spi`G0>|NhFN~n<Qx-tihzC>0Y#B506_#8&?x;4-p%e^+jC4`24GO2tYOV
zw)+NW=O|+R$FhQPYHDcmA-Ku&!q3ij=q8CaSr<1ATcZwz-7uvi*;n<;3(v~`_g9SZ
zf@W#fuK`>B5K7`B07Ex1cN$5-Z=KzBW=VduC1iKmTz0odlVzyqg?c!tv+*tQ@LwBK
zanYgc9&LfYf0kXQUmLgk^WpAa38;mwf)96Dj4=SwLm~8KQ@Q)j&%e6W{`c42dPbcc
zjx`U~GZS}>*P!nkB<b^T)r%9&U1@>;=PBCa<I{`V^<ny#bQUkc_?P5+5RR;x@fYbJ
z<FByxUtaJ(0)JP$?+6XsFG9SeL46s*JUaS^QD<ZC<gPPq+Br-09~;E~{)+w=BQ5p*
z=EB3$Ki7$!nbV!Ccevw*=C;)b8AT6))axqy-(PoLP21UZK0GtQYrN+E{`CCkf9HHo
z|If3U>Vp7@Lm}JC=JLXxC)tcTzsu&Po(yIG7DE<(%<wy=6#QX&4SDu6H0y<WS(%qk
zIrAm&oyx-hgS)p5Yja!s27|Y>(BjfUDHLx)i?&Fp;O-U(g(AhR!L_(M!Ciy3Sa4{8
z;_d`@cehTu_d9d;{@yur&O7tYT=RX;AGy|AS<jMNe$UeTzN*Qi_yvG8WTdt4tgTBM
zn_btPqKt)do*2?5YpZ&V7SEQ8p2+Yxe3C*&X<r;1Q6}sm;WMIj^~0Mw!2F76pnE3S
zj)2^6f|8Jq<&lZV3BPmiVnjDt`??y<C+|h$7LO;&ymG7y9usM0{=bRzdAFe;5A9y`
z0$)c}cXIAsX_@a|Y1dR@|345$z#s8<`qH$FxW!Qag}e)Y&;-q+1~iXk&JE{ILBF{|
zgihA=qF!5R<2v_L{LKWPn;2^Kc`4+DmFzD7Ub86&hi(SCHu&fl)ij|4IdqzLb8m5j
zAU|px6WoRK>f%G_hihc-tES6Y60qdfl1pB5K5AR#;$-pd4%i?4mb}mr-ZB!WY+q@U
zgFbr7<;m$QMg?Ku3CVJmjKJl*s716Xhpn%L(5K2_Y(uj-_jV5T2CtY#)C?E0oBM>i
zX;vO*i?e@Hc)R(6*5x8Re71Qo%Dr>531g2U?iawI>2e%QuS21qc^)@C+zS?QiylpT
z{=n|=CJVixrlLf0D*t$^`OJd@#c^91!Zows)C;&`4!*{#yKo;rXJc|LojJSs0`A<b
z{ra}Q`?Tr%hkiJGn&?zs`}z*Y;Z5njhT<5?RZ80ADDT#0>u3>6cOLl@2kDTAE?DRK
z4*OL)-z~nsj=|eM-uN>bIA5@GsJJLU3D%t#n%nj4KYz}`yq(6rPkXj<^<=&pvuw&)
z$C2)~>D;~j90W)!a}I9~3GZGN{_Y9<EjQ@a2xD>zbHt;)iU4YxX076d@_KDA17r?k
zuPAQ?J}wh?SDnc%pZn;Sw8^10LKuw~Vi~JVPY;e!jH{%k2#YdivJJK6xDgg1$ET`C
zx(Z;36{>=3OE}|svhw)4Dq1;z0le`0QgU2}8zG;44&>^^v|SrdpH-y~JMT-C>FzEj
zd)!zB=|DfFvL4pxPQJ;Z5F{>+M7zCh;Jc=M;IG|iI09weM5`YAJ|jf<%u43MV*fw1
zk&L+%+pH<>xcnDj9-C_S9JQRK=;VC!SH&gbnosr6`%~r9hyD+h|4;e;FBJSAfb$>d
z@;43n16}?h{+Ol$dt)Eb=;HiUtnR`}n&TGmLFYHVFrd?#FnRjBgN)C?Sc*;${!}z!
zV8Pt{I*EFk!-m7~CPZySO3(H?{@{gGkd4PlEw&00eh)SaxndbvT=N=D;XiIHsNERP
zps5uxo)9z8Su}2(adpF~$xC!o=UlYUQj)2>3{bz$a~W!sdN6>VL)UD}vBPyS|3Co-
z$e$?i{eWD4RbKU+xu<YD15Fa1*}2QA+%z=u%!^3hc9(}q<mejWNB#kZ%J-Jiy<j>m
zD+guae?wK8mQirl_FpJUp*S6l_Wb>ql>bv$<<Lz>qhM+g5gKWS&`2vIbB5~Q_WsQ_
zeD^<4QwNQj9^SX#B=Z~SZ-m4r-{rVtv*+FJZ1^4w+`-K}yU1&M=sguE1r%O;qH|r7
z4R3!-<mw7MqAqMu6c|ZA%|Zh^*)IUko`i6i?*U~a`jVUhbv<fTeo`#!WZt)Pwc(Th
z!Qc=Eq?e+#Iwe2m{V%|4_OxEdf$(1d-$)EWcj0%3_GyZ$-PpVCS_k!?bQxbX`=kyd
zpiOw#TWdFNJ7RK8HJPPrn3L*f?}Ju{&+|vRvDor*Yl%xQP0aFy94yL*hx^l9N~#jY
z*!Is;Vw*$S7YX%>wQj~gE%=sOiDb+u0mI2^Ii@nRWnUe48H(CX=`eN1pkxz9q+W8C
zD+QVZjn;lVBUiL4H|Ev_(^9#j?^&p!c06*s)0CrCdW`ZG<Ln8EV2^`O&onQoTaGEN
z7O6QM9T1uY!ec;lJbB@9%WV<6v6=L_);IiHrCjX@cgy}A;C2JU<yJU}_IdH@$qC?j
za?v9D4y&fE+2v&_OWnbnkGcvNLu;h{%E55b{bjkusOpnnfW-8MH+Px2{KubwTrXpT
zbf1Jc*8o5y35_6|!<|(%=CpPO{~`>><lgq}`0oDr1A!Gsbd4|6<cx-kLmHi81JhCg
zszXNYH&`5VwXIG402(H9j{(k`jo)GB>|Xy%T-~c%a>iRnzy_Dz$j|>h!uLnydb72(
z{|<c43Lq|-^yK({ld}!S(GRab%bQJ>zTQVq#D16k|4A6LBpQ93Z=Dij1KldXIjnz~
zUuGQk&674+>iXb!XRIy>0lxrOlGo>$&N5>j=^rTah&==W-gZ0<iu>5th}NiIfZ2FV
zU!f%WD-9=i^*Av`hGlj5XiGe1OsaWz!piY=<TtLBf_x0<tvPs!W9=7U;DA;5aMT6n
z$-=YRUy$7Bu&EKROX>|!$m6nUhlSLq8RZ4G9e@E9maA6?UY+o+dU=kU4Nn%na^KMh
zAmbGe(tC?s0L9@o3Iw6Hu<IgxUj~B)Y;&2u(sVY<>TZQvgpIw)p6!!Yd0qv-<qGmY
zB+uACIs6Jts_ESfjBBK}b0O<r)HJ2oJA_8JlK1wkpNMsfr}StF2hJt)Pv8@5Z^A^n
zUtU1l%9gIm0g6jh6Stn-9D79&j+<zJq(mQ5NBH@yiDQMw6*izKwJ!lq>152X>p~v;
z3s4b(mEDI1`iTo;4lwY^bp|*U(A4Ar`j8J#A*`|;UVW~4-e=@_&!>L%q&%+i28@`?
z1+dNasb3O8JZ=!T@c^qe_8UB4)9t8-41n(__SQFUPK@m~9{cxq)oP8oIJs(L8Eu{d
zkYiB41S^L>c&VSa;d!qBYAtNgrz6-Nq?<zt8~5a3bgJDl16?|}DW4oCN8tQ`S~yR2
z&&7Y^TZsUadhiObYq}rn>u%89MT_XIH~bfx$tE7zzeQPi!1i}+hy7?C{ip7&*B`>L
zkBJtHlb7g>Gy3XlGQTDJ=D{B^-L*GFf00;B*`5j5OZbhzXrM<bcH=L=o4?3;Gu}UB
z{$GXNw}^iM1OcBDzM;$LKFj-;vD9c@8gullpe~H@=yn>A`WMN*n}&UKh1RR48!Uj(
ztJsVA{@B{(ribV*p(PqGkGM$?<Yst+eVkDI3-GoX?VsL6I1RurGj>w$2~L*0zVLiD
z-D{;xiL<8m41RmuOg?gIF1@2~c%#$KKpa_G;Ou<7M_9YnN}7FS;Bi+Ma>yg(fDdMR
zT`hh^8?h-H?UxF<Au%X%30MZyH6-p<%-=n|%?K3z1>jJ?BycL-v;qjOx1w!}$HrrN
zL(S2L{>F4&H(L9n6fg%%`DkPf*ZPv>E*KwTPL}ee$gE6XpQq~{Mgz2zCtEox{NZFC
z+P5rui#(psABU_Q45h{OExmRuJ|Gdkf=1xv(e{NcjszrpVtf`d<SP)28*4IR#{AG^
z;R+P0m8SR*Yx)Jn!@cg!B$t*}XVFrbnj)9m`LbmpV~N6scZhG2Hcb=-qNdfBD*DB%
zGtY!*CLiR0o{FK8{3M^}j-D*IRvA@6`KBH;5d3ZG2+aNkP*?OgVmLh^<WZ@$>J=oT
z5DxWnkkgh9FYMxTs{uh9n!Vnj-@%Scd}3mf9_*5b)6fj2xqYh)W}UqQP_Gvg5rckN
zSw6g8qDBNEf>-NYt{Oel&B+*cg8L6BG54{cJ2@!xmw1enE1y@)2oF;<T0>+JlOy<5
zWV)6*m3TXs^;?L5t>PZr@7a^YW~ytzpe3u;NGqihBkf|o`y;h&L`rw78sl8pTyPeO
zg+ARz=1j}yN<IeqEN{iKKG*Yyh{gFUMZvTBzlC{Diwo#_GsszJrLNm7sI0Ndn|m)!
zmT_&|)42^DB7_@%4K*?*l+cSWKe-bXK|gG)X0q!=cJxA3@pcJccJO0Sy@47(_G=%F
z{GP?{4bihzM9a!Vw&kAVyS%H!X&|PLN4j<*t@nQLSjQD2X37&qx<V^OYn84HPQW`w
zYM^_GyS3@WN$y;giK0y9G(o-iDvq~8=YCi37D72pNC?rMN@c!>X6_Iwebhrd({iq4
zIpIg8U}_Z#OK4cC$qZ&^7gQjC^oEK9i)CW=_UK!_)IOBogpu!G@p=#p8k?SUSXy|H
z$nZLwd@J31!KjEn$$G!*HIZJgXfUn5sbK=U8aIog*2Gusk-qASb(fgGvHCwnMy0tY
zYAVpzp$C#hdt!n>{Ht|zj*Tb!MTN99D=yRaM&vb4Jzk0kWs~C#Lp!w9()9|<AeTNA
z;cE$B)9I>E9kpZXQ2q?XRJFYcp@;KLM)GF;M#(QTCzLOR2KFl4?;3<e>EAi##{ty@
zquzPH1hVpQ@n9V$Z0krCD=}$wl|uwxYUHS8N9DW(@&MVNSAwhbrE>WPf4)vDttBuc
zo=?nd3*NOzN)Bh0(?oJU(=hkjbM%&rk4^X(r=v!~n9HbtdOwixnfU2_>HVmrw(fVO
z{;!{O(SG1~{zHd!<FkMv9~pkEdZ_7GZo4}_yY-H?P>RcVDOHSvgTebOaBm&0@=ny6
zl!!e|FsrY}J-!dMBXKOlw~5CV?+0_3ObOl<Rdo3*$7r?s)N9z#Jf;2?Xt++g`Iu@I
zO6s(3RnX>5&^74@3MPuoV54@IN-om8|HOC8PPH?@(i?54NUzPj!nm?`CkaS@(J^f0
z(^L=<OUKhE-Ow>sNHbIHlO%A!%uszdfG<u^RFs3JoXT3k5`vi0&PSU&NiowpK&)4+
zlWk_35HEGs<jB-mO5HeBN2MB3BL+;O$ZJFS_23!859dsZfIAr3+I;T^r`2WGkHko)
zLfOJq%~{IID!UrAo=1(Ds%%JnO&9f*^2Lmm&K*vnCV4_rW=CS3IH{R`$I0E#8u%Vh
z(KvTYRsxoI?8rsJxw+bkcBc)RC;(z|i`DFWZb^Ijl9uyjSICYgLD#pZ)qU|7yDky`
zNtpiso<PXoq^)}@JnPBE1$M0j3iQ;;BXuK0M0$BAtvV^bSR;Fga+6Q)%^Cjbx$`b)
zw)ivW)^L1qi(5%>?G=D?xG228G@Hn(M)}9=8#E$HrZ$MKJ|zM+xT=s5nHsiMo3s?l
zHO>lUu~mt7+BLfD)%j^-6!y)EP#H0>vr8cfrn%be@48XtcSQ519yXF^f120s2rqrd
zInO&YOACEm?Ho(q72HzNXl%fXT|O{BW)OgMpok9&iZB7*>7K2ms{}DQn`+<Fxl==g
zUkK|qYTd_2sw#u^AO5$J{=USHoI#2lbTp52+~z4iToL4DMIGGtNj|=kuo3XHhs7dV
zLTJXNCV`5`!5TbsPg&?e{r%#rXt$k}sFQZvBUha=wB45R4C|$x_IQd^ZH{qH-?C80
z?C|o^rLAFP56(*SQbCL@H{9x*?U5Vc)#bGue}Z|q?2xdf_z?j3h~Q`-xVr<xDocDi
z0%I_1R|2p;c=inN`O{StfS=o~6F@@^UiKsmePa6tPb|EK4(p$0WXKdZZaV_dka$?9
zy&>P~?gV>Ins>_%frQ?0%q$1)(<PYKP@pr^9w<38-7bJVDRvL#)7O)(d@cbxPEO1N
zpCvXWDIx-J_O9q%@1k45xN<s={ZFgMc;>U!98h2HiBy?D_zs!iYD<K7h!`FKYD-3y
zoS(snC?bMaF$n`{(kOZWpCz`>8$V$E*+S<3p$mDL@($^q<e_(04s4HL&Yq;lf8@qK
zM$<spavm#7tafAP;If&_lGz_5wARu^E(Zv0J$2Rlz(+E<C7k>`889txT!bMn64oEQ
z3Q(;MYX*)7MgAt{9l1ppA>*S!gZWem+(j4gi!y6@I&m}|pT7COOXhzN<j(rw<bi9m
zimGf7|DObfLM4xJhn-Dop??rJ1G~jAW25zs>L9P{fcPrr%pX*iz4`8pQ+uEYTUkzf
z`@50v{vGcAqgwn<pOfqWX=ZGkkr%T~rrP}ds5GQP`f`pB>{-pp^>l-1KIYn+urdx#
z3D3vIlW*xx=;yWePcH>3u|BOdn!z0gx^H`#`SzW!<!Ylf3N_za55MClsBO9_cB>#F
zA~8l=@9qXKzP2pf7>DAD;c))9s@uOTOsT0t{RD9yf-$CDTQ@Xjb6t63MOzwod2U(r
z1i3t~)52OgbK`i>-9wX1{nx^9AXs5Tl@X21mxEjF{=!ytKUNGDmoL{WegPD0xgS;k
z&Kq+Cg1WyG(BIjg!V{90>(dVIdh{H=gfu<F!~tM1Y0-Y6p8?%Ded?^B5?;k2-1^~O
zGrlY<msBV#4}-E)NI)-toPVuxYb6LTHxZ6C4;&>@BvT|GmtL;w1to3Q<g_*L+Oqvi
zc*1!1$f8^`<Idaaa+{WSO3_+90ik9C!A?e!^gQg!ob*7)heI`eqhod{h{WPr#BWD$
z)TnJ~L~%4bqU4aC9*VYxTZ?i&fEs>e{3`|iCkZbVp0xb}e7sr5ge~NimUb=^PtFZH
zcsMe<C=FcJI5}66sC9pyL)N)@G3NSdKTsN}?N^=`UBPJ>w37zHZ8f=6aj@P10C+DZ
z?heaa=|)J;mCEhENe6(Q1=OsIk2xQ1?`|H3@v)V3BP^vYmij%c@5<9iD}io3I_K8O
z)EQ-B9Wq@VSI!hnuaE&}<rFbKW~UP#A;vSABzI2sj^mO~9*4*;$n6#%ys_8)kl@j$
z`;y-zZ~U(Eu$=A*m(<C@(ky#b{?f*{Q>x8>@2~)ui8nhnWFC*h*JhE}@!%Jrrk1Um
z!s`3hJmBXD5B$xf4e(CwRV9d@f>!6J)@cNloG`&y<0m+DyL{R1{xOKT1vQ3SlMG`o
zx^Bf<ZGauU>Q>i00z+q-@8P=k^k5iDaUI$NSpTz}V6z-EaGov#(&1B3PaH_Ak?0g4
z|8LpItXOLMm(H8Du3bmLSWY_W(he7;Bg7**w+iHYTaCBAI?N8G>lkUIhr0P20{2j;
zdR*6<y-QboSt)RXVCt0TWSjF!<lFqpMOW8(B1ih2mG;J4A1mexoU6;dGNQ)RqtiW%
zv<JeKw+yaz&BiUA?p4fgLXBA$o;r>kc_S%{J9Owp5#5ZprR8HdtaIvC`G7xuX4OBD
zfCQ@|3r^Fro3r=+Y$U;nlUto9nN^ymYuYQ!;^Q+o(i&5u<AM?sEjNFEu%9p-QesX%
zzdXNee1p@wd<t(n@O<{SJY+dq)?nR-n3wYJcflVB->K#~QDUCv<Tkt^<PNXUJ@G&E
z<eVvIbeRwjXe-!na}Hd0JMFD9US1lZhi-lD4u`An<4Uyst;qk^BRK3=q8?hM0nIep
z-NrZkJ;GOXrTu$LPrB%>_Kvm+ckoO^|I!8wseX|(WQ-`v>Pe=<j9>}jV;E7Fb@0ud
zTkw1Ko22F1&Z~12on6QECG1JMiSm!D;hu*qzu!;IS{q$Hy8dD%n{;&z{(v4N%GJZx
zRu&n{s`|FQFu!EGeRH~H06_mz*Z*3<$99X0-IsF#6wjN%k1@D<$cU`DhM}d*gdP%I
zC7<9`x7Vi*hd2yxqL0qDP-FCtwn&2Ciwlxe5kUdzrf#{wXiqx8_}V=;457F6WKoSH
z72rl4w`Bechx=?Cn<^^O2rJ#No%-*EW3#lcqq~SqZXK%o476;<Gku^|;a@M?pK(v7
zmCW4gTzO44e9=9L>Up_UQsZCbTwMFsKDBXLM*SCn_}mpERqyXeg%RM=ZEg>kK*<*s
z$EhY^k<R>-K8<{jP~>p_EMPD8q@s2QX+1-<G@^E5Bn0y*$DEBq4N>MvaZtj_n1;~T
zqP*a?oQlN3k&#qR_bUMs|4+DzMf^yGEKAdQulC!W2J&K7U1kqUPl}SJkB=1OPe>hz
z9L_E-E-ugGi3G~lR?0e;KJV5(UQ$eZ@d#9NzAIO+?(|csCe4^50<X>0ZcA7>%*e<&
zo<dJgYUIk(nbeb&hsb&rtE8+aFMIdo25@H%+*hJ=)AR;Y)_6qAx%r_qxDj0X3e#C`
z9@{ES&&9=!4)?M?b>`BO!A++z6Cih4SKl}dAUN6uh+WpXi2$q$*8zO7Q-DyJi-dTL
zg4C30PxJ?$S$VBs6OgiEf8}BicoyfbfaQASNz_;Zo^k>Ztk`=#+PrkD20ZH80o*lS
zIlTIWjpUK5E*>ATDnyQ8&fM6cHsvlGYH#X1A&D2;M;Fskx+d1U7>3i0_Q*{7=*HB;
z(_KtZ6UT$<Ine%-Qh4z#{FrqP`~!2+u{3YMs!%st{v(E^**GxO|K|}!{C(c8pAV)T
zmx9Ypt<51&Sz^+aLDrxm)#d(bM2O<XL)PQvCepM>kbzHJNrzsXrp20$bs%=E^hNPD
z@AWT$$E3aE0p~@0GO!m+qVTO0j}1nh5~-V{l`F1l6o(Lkk}1ul*217kO1T9WjMOrk
zccmh0o}Q+LT|1zlTXh9Bd(5dLGH|QDYp18DwsuP6@>UwyO=Js~(F<k-N85Bjc+ds&
zAo$sF%IsIF19^_>IdJBM`v<CnL;2#4Iq;rtAC4do%I167_EyPxVI^=%D8*cTx2xP5
z@_txd2Fc2~MTY4&9L9+%lY7qWj2pKlY12DW@QH!zDMw1}$`C5Hug>6zuUao9`DNi^
zDycGPQx-qaVZ?kg(u&O$A10~F_L+rb+R3k=k}Vt7YLgf&5Pa~(hFv-UV#2*?K5WIf
z4}~vO<b9Hr+8OS^E`yFWULT!rvtmX`ac;d}j6Ut2S8a4Yvj<cnOT!F*aDv|uIe?^F
zN|3}j001ToJbdhlZ6sc*>xfx=;sAKGHjS2<y3U<e0J>}9!WtNjC5xwfXlZIpvjpJ&
zAyX+wyC>K4xEJUfN!D1dw`mq{NcWzYc;XtlHWSEK+|b<t07wep3;@-}(ma$u`pc&P
z_)}~FX2Nuse50sMI_$kPc<<&az$(o+9#2>2#_kz>N?X7a*0QN0C3$k`2-t_#`5n5v
zAp?NuE*k)Z@5n~nt>Q7PG0>~hV=-DA1uy{TWI5@{j}S>vk7}}el0NFfXAT)#vy1dw
z#+CB@uPHc2FtDIb<13fJDy^O+<5F>KAYTp^X{mE%q-?m~w@)N&F&m#<V5lIKNpo$o
zAa3K22!7%g*>YC2(`KpyNhyg~3M1Wnd$^L%U1}3NCaZD|L@hsZ2mhGP)IR=+Ab1+M
zqr5fJQK@0kV~B)W6i9>uD>^I)Ykl~orL?sqVjQ-{1B9~C3%Fw<v!3_aPa&4KWq76I
zCEd_5ekn^xn70*MgKFrzjNBTH+#$kpNW8R)ln@JoGmA#7kQG>+f4q*~t?0bR8?X87
zSf_6ip4Ptn;KN9r(;<OjUq@ZEQ1?uKY@H{yIQ?D^MmkJafU30YP#|_1+AH@|q2yj-
z|MyL7{u@(YGg6XLmG~4$u0n!-GK_iYt+X#wes>Am`;K>-@BJWMGm>7LSHk5r{sQk0
zEBz;nejsoL%_h<I^=tyT**i&>wMj2{Z=kBwF3R|^YzAb{Rzz&fUk$G{4(3R`BQ=Uj
z+DUt~dvp6{*E9XU#vA~?`f1Q@r7n>te@)!tP+fdDdgrmBe|Rw$u=D$?&M%OXl2!9+
za?2p|v{#NB67Lu|IOjZFF_g!yeluIe!(cj5&T4R?pFWi=`4th}`#!VXv6FMtrdyfr
zMKH+^h9371N;Q*rCw7vyn<woE+~*f9>7)aF;&Sx7H8TRO)#<&}vU)-e$ZwC$zPDH~
zXNH6;)4k@h-?kw8ey6b5Xjrex$l?+2%i~c<6P{JAJzEOxUGl<MCSMVx8TKcOXF2>e
zst6Zgq${6d0To7!;1`RS4)?*_QSs(vWqw|^_ZDqx`&=Xo^c`&ofyi^2cNQDAX@R;T
zqQv!vTSi%GECuwhL)#R0NlI#as1z6&fYO1K=uW@t=Xke!I=*`D=d@NyK1aBHH2HIW
z@ts=PECYJq1$_j|ZG*_8JmnWa35FihLBwMop<j};-aGuV)+XRN?DfR5F@qy9@1C=#
zVFrbJW7h2((b`gW=e*rJqyX%MPa<x$*FWfIR9`_h@dzKEY+d5?gC8}{$NaB8J|Y+H
zRnMT<b~>>J2~F7t>?9!0tycQ3F_=q@HB%ev8+1|?iKiXtA%1gEq2@2+k!H(J)lh0P
zn|DZwjSDvK#7ScUNLosvdo0FvVQzB@Z!o1dPWN!G0L~{T2iR3Rlb8U)w-prkyv&-9
z!FZ*Z<EdGpb`T@D%+O?Ul?HW_#q8Tu15Uq2BQ-)!W~2_pKab?l`of=EsP1{(LfJ0>
z&Xsq$0Ctq2SB)2Vx@BBQZHF$rdj;>0G1Eh;Why+M>RpI)f5s@ttgKGF%A)HTAA%?G
z=DV`5Smuz)!GKvLf1XN#F!g=P5V}yyj+iT1-l$P&Iy{k`(84R8J@y7Q#+uZ}<tioN
z)GACHE5YQgVmZ=`$0p&p1qyKDaG`OEOez`v;2X}6z$|Ds(@w&EfA_PdbODO6*V)~T
zTIN%rr%Pj;dZ(lsXEaib^u91CoRnXbuv96HN?A`dZ$vGRp-VgdlS-9JSIpGXvpL3N
z-8%|yrh|3GE|(MOvvIO5VzoQ$nX!aO%jviqF&?d8t?Ush>HSUasSrMzP|!2K7%7Xf
zxa>62RJ3_MIYIdD1I@Pj4?nJPTdw@cUPfJ2y%;Af*64s+&+_@I2~qn>8MZ65Ff*CQ
z5<82}+1wWk8`6$sJjsZowrR^@Ch9=sCwU#RPEK`Ma9|cE_o^IW_>pd7*5|#=)tV^h
zUkNAfDv>WxolgG|h*@&^NIJYp-DJv%4kfo7H~rr4%exWv2#zIt7-w|crYvD)I92%f
zWEO9`;mJZ2vx`1XgJ`VHP&Q+1xk|b+G#<v^%VW`@_ajZtpNaM9>PVhM+?xkno|<Ld
zV3PtEexHsF3jZLJfDR?Ar{UnN2+~yf=q4rNVb3~X2+EOS#g_`=M?745dL$JX`@BO`
zRp#7sX~Px1o4(D`t!`TC;onO#+)6|HQz*v?tNUbNB|5~Zsr$s&vE5D4SgSHw3JJF!
z%1-ZyX)_9!iw48UhDI83G~=b#TD&&C+OyyJhQcYwFsE7Znke;mtTT-o=&rneRnNi~
z3gU)8+E6AoN04shd+}ur8WUOtcS*XHi)L(av<W)6Uwh@02nf(L-OmCBaqki7rBuM(
z@g%Dc=8ZTkUzOlh%VwjLQiSei0*|;Ot?&uH+EAE4`={wkn8sSoS66aJYKT}i4nL_r
z)+?Jv46ibXl<+8#QfIdp$ffs-=4LDMy?#DqV4z68;SuYFD&##bKOmmSf{qakTh5Nb
z+6Z{?amG#53dYH5AYUq3?<7bT4?XM79bupzsQ5f+Q2RxT+?NW7FR9~sp||SKAsQ~x
zwfQZmSxir-YS`hMcXzcu{Y|6=;#tD&*3XRR!}Kx+&Hil8U9W48tUuPh{25KdG!O5(
zB1x0JT7IB`So_FADx;7+pQ6lFO+Z4)4uAfL6Yjg&26>4(Rh?9hnPzjVPEzC1xolA_
zh)OXPbhbOtidou{7McA8z;J%!mPGVYm|#C>0`9z6`3qpUkduO8f#lz6+6@AxS-!W~
z*$u4WDg0sqPVO$*EzC5uy}Ncl+y;N(7^{F+ZUmU(82p*OyX@#&oYu*}ai%YE=p#z=
zF|^@~f;|dLnwQE5MTU_T!Lv=93D0Zz6E4ZJ*r<RW!n3C)T9PJ;znv_xgrJ{*j*gC=
z-+_=2bm(_b6k_xsP`RAdGuGUy)^<{^t&kilIH^|uXl&z$QQ@G_wIWa|n)~7<`*6M)
zw}dE9$B8LqD<Ui+-Y5rS@>v(-*PQZ&t#RNlfX%+v?Qn5QRYT87if-SoT7mjA6S;S>
z(y^gqcLfMPvh$_}@|<qTi3BqGE9K4^iTI6m4P{qLD#-+SBRTYO!rB7aN&2>tGcEaQ
z6?HRLQ;b@qTiQkE-0kd2mPUmNsgannM77c&6}N$qqrq_krXw@0&ttmW(jXIe-mL(z
znWs}$cZta-GP5jD^|0T0@4IxBL9vn%>%-05>5xH0n5e0l0$aY%W?EjCGed8OVS#y>
zWm)m*Str{}`qelcp3c4t@1$W5S+DD4JVV&IW)M=1T7m_+gWFP{e+HBqfhpyDf2*dE
zC*ioryWBzYk^3SstM*v)j>YIB>466oka3-F4x;w3VjG1L2^sLRdd>hl>C}jtL77WQ
zyS*r<N-o4L#}rPbHMDp?BqoXz!*1fgGz`+`Fpb>rM!0Wow_2T5S-jR|YWzS08)ta5
z-=ryilC#ranx-35gF!s&@!&~i_2Wuw$t$~goL7ZlY@H@L5)WHf55uEZTSE8b%FjJ5
zAeJ)M4$S=>6*w$P74zD+bWF2*;kGBu73dkkzwuyZXO9?&0QS!+SNzX4%_SOMf=?@7
zT#f`RQbpT9Rl8tKl9umllU^#brC&+9L$bvz-Uq-}`Lz{(1bFs&Ky;0+>=qihdM=>U
zMonyvy&9r5zOVg$2<)!Fi~_1=i1+-0jz{ynGE*cF^hx}P1siZsd=FBVpC!VC^938i
z!MwkJImpp<T@Qry5vuu(_Ippx;2twxnrIWG#`3U)&+@f-4{~eY5Ov~*==&Kz;-FQ`
z_FP|YJqi-9G)f>-ns?6o;{p|Ca$I}13%&8b_pF-7^ObO1A$eOj&M58856Ytw(@^rG
zaBHzzxz%FT!lC3EQ(`kks!(c?3RRSC>al&6xX1${OA7Rj_ICRlT9>LB*K^057>Mb)
z^;VP3aj+HjSI*r^kig+o+ci&-Ca<rQD$`CK+D$Ph5*U;X?5FaAB0Au-j38-)C8HP_
zn^G#~hc>EoF=DQ+oj=D_Wye)ea~sIW%_=7I!|))A_q<?QBV}C6tv;1-5-;;1D^nX)
zQ#(E+330@9KZ$8MtLaE$sFbtWp2=mfr8OWRPCAdEa|@o<x{Wo0j95q?$~@OHHQ_|x
z^rE6oq-_uraeG@ac<8G6^ZHJ+#J<5twHJE3lrFhLsk!a-APJSJ*tao7NsxWkyDCBT
z9UIoY_>r_&k)0wY8!*YpyEP8CS72R|)`e4pvbNy6>LcSX9A7)v`>>k%OSax~LIlfZ
z$*xgIBc$_GhI|qttsZlV<$V{=@H{>gJmqQ<FI*X?navqx<%u`@?y6-lmp#gQ@@aWN
zg94)<M<P!$s`C~4i$9gjxNcx1BBm7>wY_xM&7Oz7G*i)v=Z%<@=?r7x6^Z73dAlm5
zoHunGSDC*?Yh+-`HT>0Ar}6Io-o|s)rSFzCmUA2KMiy55Hwk{ntb9iFI%5sN=53nv
zqTe0H3*vU3=E+n_gN4+h$$UGM9CHQOU?&u$6(yKd4SQ9Snnb|Qz0Z@5y0?Q|tq70I
zV$5}F9+iC^=6ogO2<4<verS=XI!8q29KF&!kt3du#36GCIe!}Tu`JCxSS?BOVeq(a
zlTG$wfuT+#-ptxt`$s|V1Gy7rx&uEsN3I(fju^PRX93@mQGA#j_a9z#Jdl!-e}CQ_
z1mBYl9xreWhd#qLEoqzb{QEJWf1;H*l8SIA7h3=nGz8FP7^wtkGbWHHH<KH%pYTvw
zFXW{u;6keOX{%KOT|N5c)Jw*yTIasK+MY`*X*Q@(_$P(_lVob8bF(1%vGC}TC#b>~
z4&Sqz3_i(GvLw&!o|`c2ac%+3Cw6UZ6JqPhei#^Gy`<#xmUoje=UW`y+q#@anhnvj
zO=LgK%Ybk)#kwm$R)QRuSn@^*!DfI`uonN)i6_p@uDvUJ+~h6eaZsByf^Uga0hjb6
zY}GIaC6K*|(zI0E2wMyH21q)q7gmu}+=>tjeEKsu+oF;oAPi!B1?#(T^xhujPvH-i
zPf#oD_m|EY&Z^^d6U$-^ctO{yhr4u`D-*M%nEBJNB5}t3eR0Lqosr7y>M+5XI;(Eq
zQ@g#@`>tbE)#lyEQtk;J^P`nGp1uN!FZuax5PNNM^o67pY{W6e3gjEP?=_yQQOS^|
zCbbSX73Ek7gsb0q_7u7-b4?LW6oC)447C|z9ldfdmMu{ze3&&-z>_GrKFY7El=3pW
zBM_XT=(Bae{4VClz_F8JD$GAoNevYw85yE!WpP8tL%i>47yCsu1hwt)z3>tB{?L3O
z>h)^X-n53V)gnQq5HA5J*m`25*5#}<o=5#DM7Ee?JkCNbThs>F;?@qa$dnrILZQIL
zSD%w`N7SH#^OjFXNW{LrgyWYYY6C}^l8hF4lI%ugk|cpkEZqFl8>UQy`^lUJop;Zr
z2C4&AXDpLB%TVpc1Nl`8y%YndA#YoY1b#rEZd*ZoqZ$6Z7A%3P?1+mRmeZU%EZ3M%
ztgyuGju%wQNeCts<c7gdOyXULkBDm7n#<mN+)7I#=(vV3QA0qD*J{yisyFzfu`zIT
zt3)Q#8$V=>1@$#8zZa6-qOf75YVgKiCF<&@Y`cCMlXMhm_VHcxQk;YKVG}{Yfo#1y
zVC#tw-IDagcq&{{5+eO$3Lm8<6B8SCK-h<NA_-J7QKX_)5zmO+6wlaX;n$sC{5XsM
z!Ls#v`3&W7FeS{#enI`fh#l4=RbPx^N?;N(J1xsg@Ml}dS4JgYN_3x4`dxL6REQzy
zUvPPOI<0rEny^($;X>`$y<dIT&8$Lhe!fS1ovBthNMDutZVePL?W0N|FVz!)@P0|-
zC(WHbH!32AXXg1&l;@u$|6|`nw>#2Ax96?L`E`Bxne03Dt|!p?Dy)h(04(ApJ^NIY
z3+LfP@Q-M5)7u_icGYPA8oM8Al9N35`fdvzDQYBM-^sR}%NR&aAlfMAsy%rJ*)K4s
zmXj_i=RHnxPyY$+E%q(j+d_Z_EJ}t`dRFaTvYwCU@IA1~YnX14dEn1s3nQNkGzaIg
zjI!;JRhIcpX@AT_<qq4hz8E?J$=JWPNCWe36?wsL@~OInl3r7`nfGeF`Y7ADqL<9}
zObRE)O5i=ROH?bf7sNiwXCm`?V156t2r~M9R)KB-r+>oSEOu@2k!*s^!kOIsXXy_m
z6NP(bvy@ArZ9$Z;jTYQmeBdecC=SP{t>_)Z4kuE=GP^59w2d<dBqYWble^5FP*S!f
z<}aCxM2dMG6XXcox1CAJ)&6;(*|5HJ?cIpo9PXo40es4{LWn>v#(D2ciY0CiBKEkr
z8iK0fXy`8hc|iv_@&3@W5`kt{kB^qa)26K<Vy!cYsP}qpm3Dh^RiiYpc7B$xHis?V
zSqgi#YzlFcaP?OPty94Rk&7p<!ZWjdDlXFuJSQqJ-o@r&t{u2B@HJhqCazS?fnM9<
zg}L>)&!&YJdOBSX17|wp)O+nQ+qh5A7iaA7wiJTvY_}MPYd1XnTJe&8ufza|jf_Ne
zQs+A8>FE&%TW46xU?U@7Brl^ElZbXmNPWGSqN;sn=9vM$m=}K<;p2NcJRrO&#%RGK
z2(&xV8qlvi9zm}OtRVhl+h707{y&kg|M`=b(zb$c+ph+2#EO+fcQWp0nXf=g(qr-i
z!u^RJkE+TX;^4IhOd0;icG2DOxCU*(9}o-V+9G--3&*%Fk6=f>uoJaHC2O{3#&mm2
z(>%stIRxv0Dg>9#f?rb5>K3)U>?W|RG5*}CT-xw`Qs@hMiTPEsyYMdnOOj_TdguR5
zHe5sa9NsdoiaKj5NG1A0t?Z~3-}yOZD?+<K$I;7L53FXBdzr6hGjx`noI8(Tr2QtD
zOAl04h%k$9V@d#9m>^ukQli}XnI0SS6G)Cj>GsBG9EDx*xSn{%1VQgE?D%^qCLtGo
zc!f?#2gl;Kox(^0Qp4sE+Q;|4sGnC~p@yF+Z#|5pf{)m!%bLUm@k209rcca0GIXyV
zfwlu6XKe=|>a_PKE7_9(9InbqQu(v{B?3a7swk;oP`Omz(4$CVL4D_Gla>#u_A#6E
zU}RKGMAvH$O9AlNQfwmxV{i^TNeCur)TMe5RU`#}nylPKW-p?}$eX|9tExshj);OB
zP*6Ix(@IW`>l+Pa?-5I87W^~_h#DloDoQ17)aSV0^>J68SNMbLnvwQ-le-pQ9Lw9>
zqOr;p(~3kU+Cfw7&R0@ls1?<k70!(+bGFF4qmjI23MB4x57M4Qo;2Nc!L$%8+~#Vo
zZ_ps1CoA=#^!HMe1<EkAUBjkGZ1G`wV}VliaE&6;u90ZQCr8?CrOHD0@ojjaj)DoA
zy_Om;1o#Brak%RDk+Xaxd@xlh1!TKvl}1FHEmUSN^N3Xy7s7&oT0HtvCft4uoWqsW
zMFHlKbm?#WB$i@TES+}c1RW43i}ybnHGyJ;(JN{eW8#*NqenR&$z$0)+`&z7SZK3c
zLrjxi=8zknU=BC!P;8Q^ugIi{M@p3pZj#1~YA1ohRX(=sbxH>qXBL)S?KCYM;<mbz
zfCkOqbi|1;Kqj6-V=b2A4EMP82-2*FAKN637Fd;}appNsk2u}sPBzVC8I%f^NgNJ{
zmM>6S><cT2<Cn=&8kciV{yv%#?|xrQ@>7l3x0k|#<oCDRjrb_HH*wBf>fEMz7BeSo
zEvM{w=A2{rRl%<G_BpcppN-g6H_Or8MAX@lQL8u_1*|;Y0@l@TZghX*ghVy?X&O&1
zB~kx0;EZ<DzopiM>%gHER?q7s>e6cV5jDH;;Evo;1=z7n4ZTdJiiV||P>$NOcD>fm
zHUljnew2?b=%lf=0eQyUN}Xk)*M0Imu?e62dv`-k5MLaDWLpx+-R;fBrwi9PKkD+U
zimv(VQ|l75I;GI|j|w5?BL$$(t;Zkr?Q)m|{Rn=lS5~19J(*)Ld<RdmpU|fmA)4D>
zk6#_LDpLIniUZBsa%8K`8k&UL2w>C(^V*eZI7v8e<XOZ%{UKSkOhz|aqS#SrQGbR-
z3%+Cjd=VDty7Qm|=LDu|39$m5UE?=I658#t(bTA9&+)~PFH)N*YTGO~2inMFpvc7a
zUaST4adPePrpBIWL%jxczKyMZEN!~mO}IbD8lT6GZ8J`4-H&UXJKn}3->EN}BG>v@
zprN>~?`v!eeVh&sT&ZL|K1@75XE#PNBKwi;q%Cf{=F9z$_6~+p``@oEtEpSBKliAS
z&Qz5!y=e=+dI(}El4%XLHRH)wJuLqeodKC;psk-`dsr|kaiUHv^OMb|*2m}7Rl=C`
zr2UNhoLtF$3aFaw%>v_2T@N|#+(ftoq)vzvk9P^FZaG=W(;RM-QVDbUp&QU)mg_Uq
zAA;;<TSau#(gv-FrTN=;v3}0W><q@2)_UTRk(Au)2t;w%7-O-W65iFr=iCb#*?jNo
zPuXNNUZNSBDB4X;oR}@_C|anx`qF98phv{E<R^j{LD;TcfcAfkpHK-T?*81*QR`ZU
zr(91yy}62Cz`=_Gm8<F$=uoCaCM8_LVl|6YXbVU6G)KCBRQe-*L+J}&LWLEwHPvBk
z*uhFNrIfHlnw}tCnGtG>*FLcvGwuRPa|-(6pO|p<)lBQ3KYriMM70!p1P`P{=}(da
z_W~)Mz%uL!$sBU>J1F==i+sP9ZQf1LCK<sLW@5q;>GA8A1xgD&Cf*@CQv_d1_HdpU
z3dwV*CbL(4T|LOrPm0@7ti}DntWP$s+)B<NK3|*Y)~RmTm87<!#9WKrQnoWXn3z%h
zBh6BfrCUo|1igbjY^>?%o#B4RU2u_!;lob{TbhuQyYS)!_@f_zgyY%TC9z68yd$lL
zjTVJ{Kf|U=B4ttGYR@lYMM>LOZ7A*M3!Y{_1?7hcJU{C82(7o8tHBuFNjO3LY!$x4
z{J6>k%9)z8($70u#Y*`R<g!clT2#w9!+vG69@W>OM8VntWBW;=O6o&BQ&LCcdrkRi
z81lBla`yQ#UxZuY`9|Qeb0%vwJb<)6zlw+@8NJGW)+(W#E~F8pm6UHkR+&%@V#<D6
z$(S%emm?O-A1b0Mn=Ipa{;KZvrpA-9FIx==-#w2fTx*|n)K}y-@<k(5rEqHOwJNp9
zN!~}SNd(HZ=MJ;6)IklRS^^8y5x83XnHr>*pzpH5PJWX%_C~blXoKU(1A^%fUS^8<
zW2zx_q4qmGygO12iuNmolGUVx-6blD8m4T@<k+8o8p8R-my-(CPuCbCNmPrG*p)Zx
zV7Ja{S3cO`2CwGtoveTRp$wYJ)+z)EpX&tTY?OVt&VF;>y5+_Y{O-h5_vJ-%k|XS_
z*733L;gOdd+XXxVqg_{+uj_+h)6_2jtGn7U3uf_l)s%)h*P*9n>n<JszpVEztWmPu
z;_M7?2f5O{3CN~e7yBTp%L93ga1W07qCS-vP@l$3`&b?-B$`i<bhX!H*`En%V$gWL
zr(!+)q^et_S#V3Y`<rdM&xC_jZ_AZEQZva?(;?RKo3Ua^t`<Ud0wSZVRky2^jGh}V
zh5b|>v<Nso82M0PkZqD5RI{9_<Lm%03JF}q?GW@^%xRJ+(JKy`B^sG@=TRGL-Nvn%
z%LV&Km^eR=k4l7VAWcIvOe|K_%)wjI@3ak1$4qCZvUtsRXDo$nqApIhR`$<QiX#bW
zyL<Y@qS<6thPSl^zt}OpvXn`@HkWP5_vigM<zC39g?i1I%*kN?zOPW%=Yj4}+?;t0
z%>0P4d~VEP#a+|(+p;}%>$@uG%FQD1{>qX;Yyz!c4p%VW_z#AG3L$Dqv-TZ{OTWZ|
zNqUHF-aW=~5KZ@YITeeE!ta6>9woD4H4s`(pk)8(Mtrb3sg3+YU<iJTRBxJ-&V+@b
zs=BNgqYj)*ZrA(mnUYzMJr9_NI!O{iG92%DqdC$i;;58P1U#A-ebp0g@o|zQ_I};+
z=4wx>hDBlceXE{F)kTTEDp^n@(3HS=v}Al?>M%uFwbtkC!?+2LX)$dEZjPKgLy&oH
zrnqGNVCKn~<k8*}-KKEK>K@)MG*Fi_Mr_wC#-4#>Z4aDL`C4#QExVuFS3!jzv}4WQ
zP8Dr1UuH#b5wQ&FCOs9i6aubg1Ep-*s8%@Ts7im{K6yL-V6G_S!S;ODNnqtYT04`(
z>t&=$8<|#h&K~b<i{?pUGBp7WObx-ch;U{tTa-Fu@|FlD^tMn;um!#@59T>bh}Z&>
z2$<5(sM~^cthXGct&YKlRVZU7UE5-b=4AD5!v-$G#(ZS!HFr%BwXA+2f0}Z9kCug0
zk`^*3hgW-@OKloy63f!Vzsy*Tuv_lt01kVw6}UK}gi>z(Vu?HQ<N9W-CWoy0XM$tv
zsk`LlULpp*qnz{J_-cL+5w8I`6vj^FeAg&a@5kR9OifToQHi)5${m>Go|Xs*Cw^@j
zr}36m1Y1Q!tn|@2=;T0`Bs-PBl!Sqm@&~GfkswtMk!{Hyr1?G$mTP9C(;IgbOus1^
z%C`LrzzAHablRV+4n8~N*ELtM&N3Twx5-`VZPW(}vUJ61BZ84Hka%Vjky2l~F2aI}
z))ifqWYyeWvs*gU$g5dN)zfTN_^-Xq6?1<qLX-y<wq@!$Z0g@T7H~_}Nl9WZF(;}u
zUYVoj?DwLa@jqsZMfp<z!$a-s<kU9Dw((5OA+hV7wT7>SXVORxK9MnT7quX2$X0b=
zY%qV*Np)awjZ3+7fn4u)NH5P^ViA0Rof58LIp6BaS!4D}*G^q!Gc8TW$TMJ1!Hh>e
z&L|I|%u<#~THt3?&{tJpdTzxM%z`p+bgA7O+q*Ze!`$3S^n`?AQ#d?eDkqXcpv2vz
zy{Vmx6$uP>t=6>IRN?$ojF+)7;z3E|z(LS}>Q0AbZbLl(B_5$OV8HO2yUSMqI#Nja
z>ShX0a@`yA>ubU3>$xsKcWRHy?t&Nt(UBUn1T>}H98z$BYmQWvwvZ=9w(&W8Gk$#Y
z@6_>+Jlv`KW7F%0rO}L$39R+*t#}ccLIiBa1dBj^n5pRuCzzkgR!GK^rmSPmTRytb
z`_7KGS$|K`F95%0blVjORcBRKO2z@nsRh>XFJ`<+-NBv67~Z=7d{8Xx*x3b^I8vLU
zY1syegnE-%yf<cBZYWLNrNAXRV)ceSz-_W1)U?_zn}!PsGdbnOiUis%%osKB=HT)q
zsn#t_R!4puVH}~ED|Ts#R?eJLlhw*UihU(vPx?AK?Y$jsv}I6m*BNSMq2rOYh2CV2
z`;JVRwfMaZqpHj9KI{QY{)XOAdpBR{_oJ@*dA;GH?E0yQ@>txbN?mgY&gI)=aw$4_
z#hc$G4+vAa<nC8>REaF5oZi@5P9h!o<=xKNP|RkOIM9}R>~}!yABnZA*ldc;EgT+N
znda<Zp4elB4ukm1vd120jV_WY#5-FMioVG23acS_N&NbenMKO5MfHlI=t|jaqj&Rq
z+81EBMy0LZt$i7L@Hh^18)@6%I`~nuR}~?rLSm&@YFx5qxm?LgOuEoS9iW;dWdVIA
zrp<C-pvLrA-^x|pMkba>J?ISL;NcDt2P1Fp%k1<|`bKN#?+8rxemQ6MjTJB>D5rpa
z)o$AqOj_M9se=jItzy(kAqPjdcnU@lGXCqFeM~o?&mv!=4LfEYbtrpF)Burrz87;#
z75OzB+!uM%)AH<(u&az#uqZ7AzkdkBpC%QQFK3j=*NP1qW7Ajd#}DT$kdKy%*y1)l
z>F1BxtQT{Q??K3@k92ACR2eKV8v5GXOV8LYCtveLqfbg|^w?$VTluJ0&5_!sXtoBI
z4R;vHFskdRvD(Ls>-{VguVO1kDX)n$l0=ex*4OE$&KelYd+RI<n#!pmzFW5Sa$2m4
z@v=Midz!1%zOMcIIStcb;<@hYp64aszqC4om=|`|?1uL)&DXc1B@j{tDyFKGsvB&I
zpZocAycI&Oa&uTSg+`h0+#s)R1csl@K2V+tkGl+05~MNjGn~x|IxFFdqnPi-pU&0C
zeF|bn8p`=UBH3ZcdXEsgTQ%XNf^hmKYdDg@W)r`z)D4A**3!K&o4+xC)a~b12JxND
z>(e<2I&b)dXmW=*n!ziem37AV%@H~cpZoKAB0gowgsa0r-<0)5Wl#0x^OSiaKDpob
zmy}04Nwuuf-WkWf@LoVURyi(MwiLN^`oT_}OLo2o8QNZ_V3wOS*ITmHYw%og%F=&h
zRmdA>l21(p4!_NBkVFhVlXXvBbav|~uxOE};RtfZIu`x`X+FTa_jUJDZ3A>r1~f|<
z$&QkguIzfw7F{5>!6;B*|E}{{6nYhko}P@C4(#&}?f-W|IUb)%w^R=(v5GWdzW|8)
zEx!Pzy>q_+@m_}kX*^fX8Z>F74~;XZB25j%GtgEuoccYnlwH7+p%pGagn!-(yD|(8
z#<ka6C~h*FP#an1<BZX0feA2j6<TptyuQ0}+=XC|dMNj7&N$nvHI^)kj<<vUStJp;
zk7&ljpnwKzJeey0MfXP^sw9sXg3g<KKWkgGxI7>*jlC+zw<J}Di#D1JF26VAX7jHd
zxhxjCFB$#Rm$j30!^FzY?RkMgWD@_CZNur#VqjC)o1bgmOUaMF8Z)rJ3Uyv_U^<nP
z2D-?TrdF2R7y6c*-jO@Dg(n`YXgP1MgFjh0bhIn#Ml_OOJqyb;kMbW0C06(NGAA_N
zX@QL3TI42>tEVq|?(n=C897rzDSuZ*P^9@Wto+3sg&AopJzpc*$vN8{pK0gA1l}-+
zf|hO0<J5R7-hdk{shRgC01HtQ?Fh>&IaHmAaf58lI;o3aa`ACZ18${!sA5p@F;#|c
zzN9#5LiX`OvuHo)wVQG$DXos#+T@t~?9WFvmh%4g_DF(jjHM^K!n(<o=hY9>TpS^m
zJtyT53v4B_qRuB9)E8$JM9M~^ItJ3E_s7@Ltc5(ye>40aC&b{;hUC`9nX#p>SU~>n
zo;%3tMmn#zjFT<>qs!alMWRb-ez&*0)A%uNI-)aqL~*${TYVRpM5k?Q=ON}4p8t;Y
z{Lh#N#~Y}v8_T#!q&BM#KyUYseCO(iMPLfu^GdC~sJOU&qf)yXXa%|X<8lGIgBbiU
zmnRX@dm94ZG0{YMhCYGOk2W5xq9Qv#CTv-f7eAfup=$7v%Htxy==D=3`^up??rH(X
zuZqkkIPGortGsLS(e}RLAaF$BwRFE;o(J(+!VD!RjN{cLT{n~Key$OLk^jB*_-JFn
z#a!hu8v*<pi&1_a-k{!Z*3A$y7k9*vyhv}Jv|bzY2_FIa?5Q9EVJE%T#(%s>z->1<
zzoQTBaedWhey}~+uLfLT<ijowYNS`&zg;Wo4^&ho4ie?p#wXY^Z<3X%;6)kurE4FX
zmKmE3bR|k^#jpW)`d^s)epMEtEUpkXrhT+j^6IDIw2ZHf&MS-0<C?kjaOy+%p)<kw
zTB%1O@qx}c#Mvk^ocCzMvpqpafhKG=y=w9hZ*|um1a0##z#Y@so`g2Ig&_%_j5_FE
z<%Gg=omjARy-M{&g1o7q{YVX$MP0GV)sN>Yrj2o#MnewN{~LGT9TnBm<$FMaf`A}F
zKm<`lKtPcsSwxU5IW$QnheoobZUo5+lCvPG$t^k4<cy>yHaVvz=j1!yduQf;vu3UN
zzV+T)Z@oExRXSC>_TE*es?M(c6J@vVwI!`fIfK<Jh#|3j<2{M_$k^|SVO*Vs(D;_Z
zk1IL6mTyHZ%eqQ9G%BoX2V_<`MV@(Q7b!itKdY7dWa0k6e03`!{QR*7bfY>orIgvg
zXuDrTn8JzW7qGFt1(+g?;{zd2?RiY95yL&cCflwa^v*)cz1enKXp;_SB%~py%{Ro1
zMhwEt`m|9J8e9`7JA*<O-n&Nq<$AEY_vd}l54ZZns%K-gk8a+Q39E7Zf~&BueBExd
zRJsD)NxlD}h|BUGm7IL>S8pDRY^CvCwjQz7j5|-*vKCCs2O<Is)kLdZP7%bURR??g
zOI`c|+9x4hn<Vct8PsQn9(AiV1uN}*@9;je<r&n_6_H^XysZYuuPl)ZluuJnd#@+=
ziIzN3PID~m&1%x_vFmy5^Fyt*lJ-4jHNz}P(=-*ReB?1#R|vYJr^$v(-u|}MZTa1P
zs?rR5ErxQ6M#?A=bMVJxDV5g9$a`MJHj|Wdp6*(T5}ud2#sa+`Ikm|GI%Uil7NdRJ
zLcTrkPI4ePvgCXh=KLN1Vx^IFU84xI7IS>+Bdtu(aP>=>k4+^1=RlI_(ZY)GzF@n&
zQT=P^fU4m1HSE^iA8tQQF;Fc(ov2oQW>uZ?9}NR6jHJ<Gz}wHQ3e^K^uhU~DEyjUq
zLKoLh_e;3FN5rJ0%%6%sA&+f|F*vImCI~XP?zAW0<&2(0>M*suczNX(x3yx<L|LL+
zY6x$_=WR90LY>@iNpb>x{+49jWjpL&+|?jPXa}3)A-ddG_D_kOgbDVdglZoIDr9M{
z^W9SK6lK0%m7$9eQhUNQXyJHxJ4Pj|Cvvy1+x-0-p7U%~23@*0e$jL2aM{?dn75`e
z23JvHo+npcq6pUNmdp>wZQwr_W1D?vS<p4l{iPGp_HQXO?cj`24%&Vvn`IG6%>7N-
zQD!&%dUWLl^mWzegdoPdV_okxP$8bS6!P#U=$e*{N%PY=LqY|=M($_PJJgBC>+QY-
zN30DGFKd$_0e7Qb4!@j;SuZf^w`vfP2{uDnM@n~BJ=p4E!+ZIiX6-|6=tJ2sa*M-E
zV`*I)QR1SDnEW%)z>heB<)CwuXvVpJf)UZ_AFQKTlNAZ6n@!_ct_pj6nrmQLZlV`%
zU_-2WZ&W+F+u+OG<c>3RIdZOwAk-P-bne!EE1^dTdjEs;v%xX>2oYiRq1i7ee`*+;
zsk4Z5#!LTKwU&|b8Vw<Z$qqAXJ>|A4ceP7gH!~C;)V^=uF@eh8(@umz-1fw!<qam%
z#)c<}hEkq~f`_q@K~5B5N&$n`tv-qw&8CzO!^SKg+LXnHG}+v}DoalBD!>cD9A@}^
zWZd(V>TQh9+ehUyvTji-{;HNDqq+PGO`fC5br_zs3Tgj&!SuS$_JMf|o<5vKaNhmr
zpSR+cu9iAZ;+55fH%Ck*xFesUI5;o$T}`jsN^#Q4x{$|2NHXtj<34b`;D?UpcGnfK
zi9>&|wdz>v9(qV922bxyj@Z9{^stm`7ne)kbGYcjTIb43BBdJn<W{_~$p!(Zri?OU
zlWs5O)UDH#8~=>K6@l*dW1e*kXLj-Iw(7nEZ$;9;JW`AF7XaPnl3kuEFHgGE3n?`1
zy<2JibBG8}#Oa3!wD#5KbBGza%_`|@?w~7Ha^<Ow6>kwIb@Bd`%1P9Dlaeumu$=cD
zEp#D^#yc7)?7{Mnz!-+Cn6cg_s7sG+4~b*3R<vHvhGwR2rB7Fgl~s%%=};kcgt<cF
z`fSYA#hxVDT+O-q3eLj?VJm&gz}H!Pw+5D^K1ye+9buvv@Q_1sexDm`UR5&GI*ZSn
z9A1d>#%x(L)h9HZn(q0<TOHXP`0|acwyR|#G-ILaS`2CLds@Diuz|>8(ry~nY6TUG
z9fF2Xg@kf5Mt?mr`%&XHwW1N6&uCq6M2{3cNK|QNiLN^Ne6qbqQVN%zUfS!iwh<p5
zshY}NkFhs%poU#9{)C=5tTEe0ZxY=VAF_H%rS8V)9DmwI<OyR(J@xlH+vYhd9ov^1
zi%4+3GiR`iHH)Z$6M7~p6%PAwx(ygZYMup&T>5uA6KBtl<$IHq70dg&zU`3FVo)^;
z(z;e+(je`Qv`p>${CS`Go*1AcdUC?r-u!_4P0LwG+(eHvt9560zJCPKBVw%PV#C|F
z8rzH1xd9`TCa<FOtJt(XpZL}biy`cDG8fq+=hl?^4DF7kx{U>+BVPyBIxo+bb0)Ec
zpNo&lhK%NZrq&d-!RxtQGcdee=b(`&-h?rr|GGWrqY}&+*q_;DL;PUjabQ1lk3ET4
zjC}Bu<Gjct*2iP~JkfgCz24dR*@{fL!t=yOxNT*i$>+8qRYaF6r?u`5ExB_vBHL+K
zRNA{he`Impvkr2j+15wTJ1jEK+8Y`H69{$EZZ19+CGRT4#-?u`;c|thvuwedBhc)d
zn@YudT}QXZ#`Ggaa_5HM?)k<jjJ#<kPNR85-)bY>9jH|>V(yTXF<|ACqcKgLOq8I4
z01@IA-Q!fE>S7Y5bj)3g9;mJb-Q>PrRIXP9UcPCwNfP6c(sA>i-J)J+tG0caavU9S
zpUSbv*wa|bZ9UVEc}GM20y!1DP)RS(oh-b)yKHmvz|8J34OazyG`(c*yBSemxgR!b
zwNB#)Z>hKx#{Bt9vbvw!D`ely=>BReK*vGeks%~CKzD!NVva+DN<|ITC4x<cw+BV9
zBZ-f;#QI)#pY|>+ux~2et_XX2UqWBfan94|4O}&&jCr|e63TJqRkUJ2QvH8_dJN7u
zH1z9x9x23L#Eb2963nJacR^p5CF1KEYn&39`ut`OgRk$96s<M$JvxDPPNh51K)1wZ
z1(5z)Cm*){^&3p-(|*-W9JD-|n>+UypIkH5H;sH#!je_GcFgveU{;VsH&g3}{MGeH
z#JF3Wc9Sr5wvuaUhGI2=Q@&gT)i7EnH37nPS61ackEqjaZh5XtmpS1wI++hD2UT3w
zhcVX0?F<2aR@%O}SVV74gpv-5-Tx;24pE0M0m<uFxyesePHFGdk+iZpwxu#zFB6UZ
z66J>4r)QOrD5MupX&J_nz^Z8Q^E278;^NaNtEUOpBZJ>7uiW4|-a)8664p44k{(Wf
zmC>5r79+?0c_bebouQJ`<Ey4|CrWw>G59U3>x~U(e8I)YY*!w#IN%^y>rIIB4tuZA
zu+!m?nPj=6vW13~cc_nx+anRd)O#svd#O@i^w<Xy1C^~4@5TFN+o=y)I%e;DHH}<4
z+1BW>PW+JHZ?A9!v9Mp>Gr-om-YVP{SVdQnnv|vWnMZZFh%(0>QasO`a7cPFu>Qm1
zDCDtJuvYZi)xcGa2d@KdMZ)h3maC_7UiXV$3vSagb=MGy$!m8<`U0yeWqH=0su;zB
z?$bw*R_PFl-uN`=z$p>A6MtI8D<?OEmNNF7%QO$D9>Dwp-rGJ`{^~y?Wyc{GXx(N|
z6`hbHM^#X#sPOsX9ecDryWP(T*X?l<${I%D&d?}^vILb~&!E<L0_L|-7{j=C&$M+~
z?PuU*$lL<t1AM-*hmy;p8s!BQPHS&<_&zLLEn4K2w@K5n7|TGmg$ArI*t!nfQ*;S9
zrrj1*mRzo{7CIAZBPn-r^SSozhK6N?&S>p7w=Z-V%DK-SNX@O8P57;KcBt0{W6X;7
zWFyE&M%BIu`If9*GtHckOVDAl;Emy+=$MB~5jNCN*@!if8tk8Kw=#I3)a9n?V1+5~
z=9^>#2FKXCcw1N=Y>2oNe9xez-j!2C+?wd?Eg6%pa_IBZjaDo_{@_$XuWa1Y@+@+r
zBHz?B!zhz3F76uwSu14t6iY{F`p<v8f0v|0VyP}H>+S(g5nL@9?vMR3p){e{6;ij|
z8^-$Z7&{GV0lxR6AxQFZ-*VMJEvKuMgYmIZX7_}Q19QK3+GkGC>eGeCP+e<FaOD(~
zvdNa!T-|$0Mp{>^H!gt_m{gTBIY$3&Z&3fSoS}thX^R9`8WX{d8sf+>V;LSIF3WCv
zm+F#YwsZr1&kFMgZWH<N+Y!0i^kLppt4+3{q%&3rv(L~E`^qGi$b)hc*!bL98T+y5
zd?|V7g*dykiKS853^}_4iaQfF_;qgEV&m7V7fBhENY!^2NegQ<9v(CLV!K%=!c(32
zZEkl*%Joa7Dy`DeTIyB#xv;`Tj}%X~2&bbXn0fYMA`Cx6PB~+^{Zo&ww4yo;rVDSb
z&!5Y}$)Y||+U{D1?)zLGGgP0a6Bd5?&9`e@qEoN+Dv<B*AI!K>xUM3u?0wECmd`CS
zI6-(Z^{S!O<yn6FUW!2XhCZQakF&;0`@?};O3m-L|3^NLXfLyn?8?<&gA&~HyRR(4
z)OKAz^0V5z9DB=$Z9m-_^!sKotz{lu`c!JsV{5W$#Eb1uA20YXv*VoJ4;nIcj*+`8
zySsRk#hpk;yu}4Cc^){~?MHjtPRE2#(6wQQIG@r)3^_5xeD2sgPJA&tjYZl%;-23w
zPx_i&%3ZW<;(TqV{<;TfjKoHtRqgtEC25IVwb#Y_<h%a4{A{($=+C}+9;0GQzwShg
zX()xYNc)1V$cK-Sc@9nMnkHF`cOH#*Rk(Pn8YE5}-W&d|ifGkid*$?Cut<wKjY7dG
zn}SI&Jwu9s4S5m$jXcxg{V7zZIIe`_%EM<4m5M=%N8F?~9}b8IHp}|fG#=nTcf`p3
zjC)qub3brph;AaFAa6i*?5j8LLz-$N5@A<DEyok=M@!wsPqpZCHTJ6}>{g$e;!8PK
zzRmn5t?;!>%}t@E&)>@?vo+aCtMAUs$zCt*YcUS*B1!ie=4iT=J#+Am{h3W;_C7)G
zNqF7PK;mbLhjNqo0~TM})P+0c#%zdSHq2Sg?Illxx8vFKj(G$~hE<K%>U)*iuZIrs
z!;Hyrb}xIL#kvG*qOC@#OzzfpP(4zpc_*Et7Zi=$lVQ7`<-BLfd9!bT!a~;ePNAq)
zHP;=3XFdfO_d|zwutDoM)gOd79GWZ?dmn=ioi;WSpXaDsP-+aAn?1P+k?J(z$~%#t
zjEpMkZ=P^@-v{M4mE-7~CU%;+{&9gG7b#L?+{;-BZ+^Bfm$CjyThVH?oI!HJS$_YS
z20Nuwn_cI-J-WB@{Q@K<W$WEM<s|HeQezLpCEf@_rP~z-Y-oGLUhD+pt{>~jm+(w#
z<}22XqH?3$6cwA%TxZO611n*xXx%i8o9e>QcToPb?Q#Qt3VGf<bl_1}z|*M2?(arR
zZuYBzw5DxxPovxk$ybSdSVn{8sobbEg|p*cR+_7#qmHAXK5CYt(n<Hjjr~XE^yJ1>
z`%;;i__P|n1Xepad&(eZm3k-BD;$>$mo*XUS<bEO7<)H)&#IvEZ&9TFDjE9w+I&%W
zAEl<TPb)Hu%DsT`%@y`a59nt%p+X>>abkzZ2NT9kref{gLq^gjR?lX}$&%=ie3c<#
zW_tXbaZ<s|`SheePW|l*gg8hC)6_F6=XsI?8s{W}2}qd3luXs5B54-sxl83diYfV@
znObgPPW%#7*=V5D_i(M3@><!c(Q8R|0;i*?v!xxT8bqAW!1H}8ou9dU%yfx4Owj|8
z<<|&$rFWiTSq;gikN7QkX^UQ=10jK&ZUOC}f3NtFjmXiXxZD0VcY2p>o=BcF(T83y
z)_QuAojU5yU>L-HB~zjLNxycC)R9HtwzhZYyVuE4)GooYUk${{`bO3))Pk=xyfbK1
z=$$@;|8TGrt#x{%ZcW&9_!t{hi<_a9kGNsX2c2yhGtr9okv08X^)|A6-EPu*z0`D|
zEV{Jf6FmWDPIsj20EHfFz_XB4^-u^87u#cIe$1A?G|b>t(!Dxgi~9B`V?N6E;n;Ku
zCQa}|@N-Mc6`GQ2{ZFmsmZfFcH`z)p`7nxzvPr7k)TpO06NX0(dG$HPq;$7D8bTAL
z#u#TAOfv_wj0j8*zr86H0HDmKzRsP+cZVh_pcoyf1rMhiK3$->Lhe(}FCwNf>J{ly
z(MupU-|OU0)5VoArqOcKJKcF##w=nUR*-lrD(9(lP`#)s6rh0OwFx6BxCP!6<7)&}
zrIw)0K{PFG)41oUjW!j$@$T;V-|dYRBABkLrWuP_s?0KVjORHia2e=KmydhJa#XS>
ztA7&Fc^Bx@+G#WWUMXptE4DLS>8#@Areu}pRI=t(lBANE@C`Lnm{k4)SnA^gf?|5&
zL%F3GGpKo)@7$DVPZMo)v}r*Vl@_T|TcxE_XVsel?mWc`J$<)F<5Vp1mfsNi`#r-t
zZWfvH_PPQBbn+n|<C0QGG_UFCQ1(Q1#g03&4-*Z(XC#-q_}=UEq~8#sSe{Y;(Y7hP
zNJTnt#40@7F$eF{%D&q~7TJ&VngPnMd^+1gyX4e8jh{`%37sLG!P>3t=yQ$+0Yp(+
z*cD<Mp}0!NA*Py=baZg^mI-gU0q54OQ1dn?V*aVV`iV9gtkA{Dr{m;z<K?*E-PYZv
zuJNI=&CI@Xy)vsKs#3MI{!Kjoag1F{|Ju!aXbQgxo0i+2H|A)nQm+_@dSe(IOo|^I
zcWaEr>eWMzE@-H}LrPBZLrvA8mPw40bbOlSLK;|z^SmlWk8>JzdiT9h0<kO`(#Y)a
z?s#4c&6VT~iV1;pXY4LP_dTY6;Uj&80-`3bX@7{UthOUm>QU-L(3F2w$9$`Q=f@0J
zl`+KKU+wDIyQ);R-XY(qlqmo6%1A-^Z4-p@-81tmF)?~-E7d=RA`oA3xfU@yN@daL
zqz?z@u*np8i0RMU!gb$Xe$%^7V$%bVtI+T%)ft=X`>rt2)l4p>NpeX^C>pTmChe&@
za=F=IadLZ9IE#f1L#8Epnb(%F8xmCIr9<{*F6(Oo@;|&R7tfXU*4w|J4?T<^+<V-J
ze(66v=y?_G7%HbsDjQLB)ppH9Uwljnt!~FJ0&y;QQTgo?T)zO@$JdV$F%wW4mE`gi
z!nK9EWK|(C(An+)UNtIa=<0;n08akJh%03rt-~phm4#BgeDPvX|2qD_VMX5=Ss(6d
zi%6@VM?g2Rqhjux?=d`&S^AyX$y2BoRy_SzaP<WdalpUbuwix5_uug_QZ(IB4y}jg
z{<x?<0fucU*Q5KtA2O!YgJt7sbd-YcX{O8chWNMnC;0!oa<q<oL0%czLEr*u?HXa^
zkx8`tN!#Ac9?XYHzPeTE^sJ1BLs265+6gZU3zguL*`MmFuIgiGtq3Yxk=TKV+kFi}
zd)J_H*-@v<3m(*!#mb^OomKA_-K8DGO(z7E{ROTF+`9$wb(Bu!d#@}>|5iml;!cF^
zAXBXdBBN7%ttfklD8g}-C+^dNmZ3oKtI4&j#HX0`7@oM7y>)n64t(JoW?g=_+fyJT
z_*+uN7S2^`OXs@y&vKOxO1TJ9YO)kw!FZ(x@(%{*93KfOWX*?vwaIuPY(G)sie=9q
zv?+VPkW6-ev+B4ueiYwhk(=e%C1yi5&bFr}U&~IsEdye2+?|@8+qqfE;n*S*#w8*%
zp{ByV@u5?_$=F|Y?q;3rv?kKH$fnn%V(P?3L!oR(z)5<oKy%Cq<5_uPdy>RHM*Q`W
zdAn8IX9KQb>!9qnB?&mrCptrPpY|-OHP(C7qEvR)JSdE(6dSJ2e=kVTxO=g9KRjvG
z$^>QGHXk0VURoXIoO622u-8jUGWjgBlR(5sG>R*VV6)ju6`yA^LeF)yB-GKO{3XGR
zP5h_KTjBerjL@$e@LN)5J&4;jo^;_d=??}MPp&m$63jMx1*xY5xxOB}yOMuy)a+i7
z&$czhR1kBg+<w~TiBkNAc1ShLnp<3mm0<e=6CqQnR9g-Gj^QDVlVjSVNe?PcC{b`Z
z$9e5zizvl!>iUT5Z|mF(3o11lhYS6wxaNzjYp4{N$1>IC`c00%XnMYc-F?_;xfZB)
zf834jX;LVAp_<wjdLT6`zQa+WXsoGC_~e{8T%XFB%QaQ^=ybuo?2WUU-ZRz4>+)@h
z<Ud;l#<*wm5wEMJKH85^PsoL5Td>Ha7wdkPPZy@JQI-GPkg*@OXshIzS;e<?e%rdk
zzQ@JRFm$Uq`<tn<v)$W*$IFbSg~LyD-JNbrR@*4NZS3@%RF{kwj*8vR)r>;fX{t9~
zgVsRn0!n4`nI_{a$huVo&pP?Y$+>N2HMC}j2Rp38JBHa84DkvHNp}p8=0hAV&aB-v
zMmfC+VjhC%k%x{!@-}i|Vz=>cM+qxIs`F*;-!08}kijNI9aDEPtrrFRR5y`7<yPRz
zuTdDAr<GwE${CB~;L=G-&qkO{Ixc0ThPu28rhW2OIh#HFT-TSYa&Glu1bbl=OU`Fi
zIrbaIg5SO_>%$hy!}n_;b54rqQ8tau_Zcz*##W0tt@s*uXrpr9=4!dsawWHr3fidV
zufwc*tA;a6P}-8wmXl6czNa?p^$^Z#0jD$8u;_^_laMcw-M7--aB^)`Y^{DqTe^_+
zxa!*{MQK(PSYj3*<f^FZLF!m(+e)OgU+H%Ey$;u%bR3>5Jym&>VB0w6)YSIaS?a6V
z<jyI9>_qbY_4>T<ms6gIU7Sa`rS>z)eA6rH1Yw-v0N#pO`{#Q7o0g;If6YU&zo#oB
zu0*vS0&u?ios}{v*0c`{E3qvJDF3(s_#%Q_zmWa7Aox9}^+A*0oB~#?lza3!8;|LF
z=00uaDmtEA0E(tlBd4xkTAYCqTM8GlifaXlt#dVXlczXNm>!`;djX9xEEau5r^apF
zRmsb8VZ~#A2CvGx&KO4mC$q-O03-Vc9r6guJxLdhQ6s*DW++GTumm_O?yI_xm3YOk
zHaH*534o!h=}DKb6IyoeF2Jy2d+vnK`?FUB<wyDpzca1iPE#-5pVh)2lh|%nw?i3S
zA_<DPnl%<jw%U_^4jy<ise|^&P?r;$t0glMs;^3Un25t^0!o;fm|5d<kOuhWVic5~
zdXBpm-%roI)E;4YCWf1uTQBEMLy(hwY`=g8Zi$3*v0B$F59Yl7g)b&pbR@SCf9!dP
zu-5iGhoFT|j7G3oLz9PPfxcj<POT~WT_sTua=b80xx4(0+UlO?XkM$GzYx21%xbsr
z!1tDJIfyPLp`2O~8I)grw)oRCqVdA{XLjlR7A029<89;I)h_zJvY$I|2~p9dZE6@0
zKW4k2IgV8iKsyT1_R{P)<`cP&C4VV{R)`%GA&eM$qZ0ckFcJ<uu`+uWiIBjmrpTc4
zfTuyeCPPrX+jK3>-KP_kL%NIISO!*UgtD$h-wHa=foX6RyWi<(NPp1IsE9EsY@>NP
zA28<Fv!uKe9`o^}z4HVFEqSkZH8zN(0X!3nvG1D|S3{ID>r;D@R*PkovE)$6U0<`Y
zOjFFNcg-kE#~E%d^E{TQ{8@gP_Uy2?j&h~|N6Fi<>yS%5zn#4BDa9%mQ(AAjV&zzA
zzGhb6T48h<;;tgvsNGgwSWzX{^$7*$a=pBF!Rw5)SuyBVM3LFk%K3sySBOBgVPRw*
z9h9(}J#ke#q=#rAcS@}z;gD|i>*{CqIDfkrpEq(m;z-o*x54CSgM3+1*)bg04w~fm
zVx{F;?>u`;xzP2wNWv##!JlrO#aMkq=oD|tUGR9EDw;-VM9E~I#XLDRj(#v`qRN35
z<cpU&Oo+tTvO|<#8Y=kHGw<!J!?`6qFlMmc$&w1g*maJZdtD{jg1J7!3}R&iKjgXt
z)%2N<Bh)fP<q;3&Bc9|~OV9RCK%DrAn_1*{%IF@XKMqOA;Edoq#?41U@N{~Wc*f#-
zI$QY^I+q+a?yCxM98&81?6<{-7)N4?*$Y&b)Yj1M(cvA;ph3u$&4m5oeuD2r?`%rU
zS+L(MNu%}LAA8)NXn<`0a9})4n{7${L=ip7zK}m_!}Xo|v;b3Gd%Z-RrEf*^i*)LK
z`(o9GTa?q`9j7L}w6Bm3eFF78nj*iVrIz(QQ!dCB9(psd!sbOOCuWe!vgm8D-%PM;
zKeP&Kxsufo_-rLN4@K%e{^HZRwa}j8-Bjz(?PPfyGVO<48=GTT-%8y!x};>yW)7ZN
z)LQAYGEQkH%_((A>U!1)TTd(a8({E$)t(AV*+kNcO$IlMtn}`yNNCmihEUS%P3wm4
zZ)of4nF7kiA{c{2!RYaKwQ8uq6S|pf$Tn#V`==hECk-$VY`KzI&iL-nu&APro0cpr
zO`@vMx|E(e18cdZTSf-tZ-zP@7xk#6Z;u`3Td2qAxy9xr5&4rQJ$bF~T9#zbR@d0$
zTQN?Zd8D3xzt_?0jOkw*=(tfAsi=IdW+T$##hH@)`IE!>+U2!}-=>V_SNBhjeTpTn
z@}NIgzv>zQ&}r10%^OwQ0T!ce?)$@sI{YbbS_m4m^g{&aH7{sxV_k0fE1DMbH7S@f
zB&pdgUyqiyc-e#_WgPOgoum|EN5=23tE*?zoMEm@DJL18%EBV6VksvZ78VwB<=#Cm
z0_L0teQ&vKUv!xn-|?XK9x_z%O;cK&<lj{;54)2R9PTjt<Cf|+*EK`!Xk&%Tk@YXn
zS^Yssqms!F2Z0y)XIzOlU9tr3KOGF*W}Y+kl;nu<!{|6Kpu{B5O*TCztm1KN9S_>C
z--*l_n4hyj#pa<EGes&NKRP4r`r4utAzzRP4I|H}8kGB*soy_*(QiW}r^iWvoXKrF
zA{dF(RbVH%6Z0aov`-E6QpTHg-O^N1647FLTeDP*wW5g=5qby6b}Zto{@A4}=5umU
z$5e+db3Z|F>2rRZXMBF<`hlin`7Ted-_UYY@s`pxtCWZ7(cuel5EBFI?$)Y#B>xNG
zVb8IfU9_zb*<ScVhWIT<CnX2#yXJypU9FeVCli#k=58dZlUz94Tf0PYJD~4f0LW&v
zHh>3!7PSA8*#AE&OvqqZz<tKJC2<o$;#S(nom7s+yKIZw8&v#YcZj9E9Bv+S$B0zT
zT*Yo~EMmK!e{h&D2d`)R`Cp6R<9pyH0#~db^biI}qzo%Vx9Z!vlR&7fpLI_KkkN6p
z`Wa(8A=-v5Bl<RyndmLhVCf}f%%d#=+wD6Hi)XlrG58UDZZ)D;Nrm~aaZEdYTN-3K
zz=v(YhG_f8=l|gM^K{?~93JY8@u*X=**96n1;>r8tYVA&y)3RY=fqO?SihTg*;KA<
zpPPC$f=Kvz{pn47De6b1f+Y`^JQU8CfWK~h{-aC?Xv#*d54)uY6^I-2e!O|`7;`M_
zOZ8a`F;N+kjlTOc;k}l1uBYF(@>Tvpa!L%vRH&w#jmjt=Yl8Qp918_xVV5b}ALz)0
zY!j04lz!CKiK$^rFO_OKXDM3I)F{1nO;S&gS6lJrPDik$+B_AVPrI|Hv}*fA6u<md
zMFyv-DIw-}NB@O~fR#V}p7sYIjnlElHvL2(@3`i?+B<WK!E8lrN#JeT0i^|E9MO;9
z<`z$sH0%Y@$086{7GOW<-A{X04+J}vdC#1^#$%>p_T&*G^Y1>JWYYqA8`-5#w-#Y0
zKC|^iaZ9gD)^rbI@~-~^h<*W1S{<+phY|v>|4bvEX6jNF5T8H01Mtx|4Z%_f(muE$
z(ZG-`SpZJX6yuF6Dm9{7Yk=j0b%ebaPu`65nmaLOZVwoO@3|od#VG`C7St7_a3)*n
z3-6)o!)0kS>r|QO3N$d#MV0T)G^7&x2GCxpu@~C*F}{@4s233st#qv(NAAlItKZ=_
z<xu6f07Ny>P&~p=5Z7C-Cb70;a|_UCp?!Z(R6aSm`YT|96l219zgUlBnJ<?*>Acuv
zHpzR)LESU<JP$)d^*|GG2O)_ZTu{yFW4s@)(ck66d8jDMqbhLJmuknxL+&t>TXhD0
z@z3r$Xv9=uUC%$7dDr?@IO*hd<ViQ4p>ujAGlU<Ns*3g_gs@_nW#O+vCgONCmxMZR
zWUCrDJt!sD;5TfVU*GT6p<$Vg-`XmNY`B`0X)en=;T+7OUF~%;IBpzunCngUzhy_H
zPN_#6Fe@^^&L>;PNuDa#-a&d6_29+*G<)H^>Z+0=6a+s%+yPS_y|}rMr~o<<WY8Xc
z;y;{8{U(Eev-sT@Zva}@`3spE^wGl(F)pEEwIY^RPVx~vYMp8W8A9t@Y*4|p$K_4p
zN)_8`m{;9sX|v{TB74;}%v=<$D?s&OFqcyxv<uNS4$NhJMw@+4<_^G-z;#jnQ6Hlu
zu-QAK{z|Ow&JP;>j^&u?zPzt5r(ly3^6Xn6+(7ng67?+uj#!+msXMs`fWG?7H;HRG
zQvh*Jf;DhYMF8N8Fe#b=g?AN5!4GrJsHf;?5b>e}N)NqQBD{eIpxSuReVGA*n)TO<
zCLO&9QXTeB0lF_swf2COMt1cws9=x=G4X$}eIX9uGF!s6-8??n;lh!ll<ALizJI9(
zn7ErNU%Z?Py`%$8i`T9{CQz36&wBe$#cyhG!x0?-B>?dMQ|142AvVKtj?*CzX<*&J
zAoyInEe+ddHI1%VV748z+B`r#cGi9kFaYDq#h9yQ5sT3)U6&789OL6Eg28|M1XpA=
z;3S0v@jm*bD`}$J`pgz{@_Y{i?1~bu$v<Z!D{IRC=&hG#<J6iI=aPTL?B_Tgr|DoA
zl&$*R6^BaFB(YrIW2E`>Nq-69##{#90#W_gzz}p$s%!_AJcF;wjfCjif<1Gld)3l)
z>=LZL;R_;;t+fWH$=bezJ^GvQDC}^oS>#y(W~7Ptlc-R#X0RxE_TIG3JS!7uJ^f1E
z<&>`?6rIG((*^q68$Q`SV-iX$oHsQge8;b2M?=gl0!w*|M_f+6_?Gf)s#xB~pVCdr
zN$pnV(`#E!xvFvfW$$hx<-(p_<%`?ATS9(g7_+^9dNf@fAe51uk1ErrTZoC&kJUsc
zuAAwuubQiSH)TQfv>|?~3&}z3b5&dcs5uR_2`6N$;&oX?{HU@hrCdQNI~uni!t{mZ
zCMNLv`}DtnCviztYQb^4-Q9y#PueI<s+@3B1fKOWK<mp)YxM7`W`{mY9~plT7sFE9
zN(a5Cn1Yj7e<k?}x$4DPjF+HtdHtoh?emk+8fD(z4royAxtd$XAd{WN>SW~3&V~6g
z=#6dP?UP<!bjR%k>PzCfvxddideDdMcXP9`%=jY^y?~W|YQk%6;{+ao`D#1AdvQ8W
zH1s_^RC2^G-rdcQ-qAi|cx>vz&3p9s8bo$=sJ+?|eX#35&W*C%t_`0&!+Ci2cJEyJ
zA@_e)tp73be~0g3^+)z1njgTv(HwiHuSk3PV=zdP4-N;{tJVTr1mBlA_ihp?e~csF
z29F>tdE-*qGc8&~Hm*{Vy(NtV@OB3=1N64#1wh1twC>&_S_a!yVZxuoO;SB1gz3JS
z9Y;&+FptFD<X+s^q-YNc3NGd_tr<C-@1}Gi*~6FC3kWp~F842gLhhWCQ77_r-RNFc
zh3nZ1x8>P-x(@8R%G1x!B*KP+xSlX|SB5dr`s`aidovAj31yh*;d`nW$GNw*{MH$-
zB5HA-!BK_w+gqp!BbQ~WDp|=Heo(aAL9IN+_&T^%5ssU?E&YF1b}js9DT-|E>%|Mf
z1A8N99HK0%vunURU7>(@8Phq(6Ch~<=6cYaFSjnn;KlEPG`KcA__;JX6NZEuOw=Q8
zq4gAcLybrRq4+*xLdhh`dG!I&Qj+7*y5@RLxEHlgIlcZndU0Qsu5*d%!Hn>G8nKUQ
z4_1ebH^ZD5+lHo_R&gQmr>qS((j-TMG<ITgmvN&u(@*PoP9-Sgg++~s!1Ej+sc93C
zy;nYT?nP7%!rTDTvo2orzGyQn2(w^l_+$09t=~~zHx7|q(^7^j-(!3_M(~{J20&}o
z{hP2qkF7`sHy-0{a&%xPfQaWY@UH+bt;GlyI||T|`5V-wFgX24`LBDSci;Z0czUgb
z!X<dHU#U6gSzB9GA8oj_e{6<2WR=|Dehv&>?$xyVjv9F5uI;(Re<4=nox!2ES%<OL
zGEFLjp41(1E%Q-<ZtEsprjcQu<%|ml9y5B!r=Fbv*rwsVdkeV3qpxrWJMWjJi8URb
z0JM;?pxI}pZ@<&<PHn&1-zrbZs+C#k!--`2uSxvBmHHn(gsi%HFZkP%502_Qm3a)A
z0#T^h7(<EQ-zI>td1skRkT!0e6yf*b%(u*dWAwl@a-s{i|NVLXiy<lFn6zVzyz}ko
zq15p(s@5dQfvE6OX29wqV2YSa#!Qc+k_Q`H`K?nuEo|N*bC!TzJt?K=v{bc|fqjX+
zSp87oa~1zxdc~6P`(=^o0X|4XiU<jPgK3wBY2M@u#^R4j)=h)q>+EimH9HpLFWKcu
zvcA|YJcCLjl<dkQwrSEo#j#=)ADCs)Hg+oYZloAnj_D(#Zxa`8*NLR_$LAYF$XlGH
z)|3rasd9{nG{Q<X!&900YPsH`i|6ts7&<`S;&UkNfbQ^_xf#X7kGZC_2r1Zkzk-+~
zd*TLZ#NinmahW~Qs5U5eW+_AhRJZk&jWvoX4;OD`MGD=+6ZvYTJE+Hj?v*x8m9Ep`
z$~V+aAyw3V`fhuNBbeurFNdopX`ITOpasP5{@8f@JKoO!Evf#$uJ=j^{PzYj0QeIp
z!eZNLzPD%b)T2q@LbJ;()CJQ_30e*11(YlFVn$o3v~<XYR&gXo0isB7!f`+2qqUq`
zkKy9SBVR?5c%<eSjODoC;^VD<-=0s2t}=_NAJliHF?{1HW>lMB>S70aXv*Si#C-P?
z%=UbHM{6ybn&vJ3nf&*GLVi)*?<NrY0_s&Y_Sd&8@!W#$=4O7GkLPHaP$em)rBjr0
zyRrBwYb3SWbUV(}^X!c?<6AC9lf7GS=eQX4D?Is&LxxUknMvlQV7{u^DwC-Lai*Ew
zZ0rJ^PeXavWpz7*TgO&(Nfe%i1mv_-={#Mo(bp@793f*6b_uhM|7WGpVv2qhu>kq-
zaD&cM(k-?Y>is@NB2I9eHD5i0-7)p?PDxa}A{!(jTlMnVolJIDDvC4gS!~(osyRAY
z_Q{IR?Ew{iRt&mM$1%`an2~yUICrdi(P(w8j>_tm^XzhamIe%7Q@T8Um$D>Xvp1>s
zy;Ym?4iU1T7PW8(X84NOQ=~fh!K8!oX;Ot73QJ}9Q}3=bPA&`5-Q}X7&NsrcyzSg}
z0{^maSP*_n_8L#>()Q&UWr9Y;vP6YA`LMVoPR{hwKC0o>iWO_q1hLN!gW;ktngAuk
z7@Dj@Mw55h2cmb7523(vKgF^%<_?m_iMHl+kt;@2`4bDu^wa1MwZ@gNIZ%O?sP7;7
zAm<&c2Y0x2E>a3ES#mD@uBeE|3GMtX0e|D<)Uz#km^cZy$`H$`wf~qdAVD8m-RVYG
z>F{0i7jRQUTng8+{3QI6QB^j+4z(Xrrhhp<^gD6c7;-LVysu$9ZrKYzj!?h2L4Z6;
zSP~LS#wr@sl;nhY*m+fDpO5UG<Od9-8y*(q=<8WTkwpvA7TH1e<XxAPk5lzlqr_W^
z;W9wB%}Q<OYs;h&<*b-9LtfMo9`3~8$cps&PuM;2UQ)v)ui+dxkJH?ZO5uq&U9!$O
z_akG{$C>wVqNFCux67yd7Oa?#Bh5hTU(s{m$B@CLMeMN5KIo)<{LUpn%x?4Yd0yzs
zU)fCmkwg8v5x%esaN~aHkMhFpXc2|^=rH#6{sPS1k8!#oEjL|R=2<&A{;UY^rM}Mz
z;*@p^3;x{3BgiS}az&uO7}E%D{sJsrj+x{k7DA{?ujNAo$_uEFh5HpwjZ`AwAG2sQ
zY^}r3wH8Z~+I|7wr!Uc4+gDmI9EhU3xN%*KV;E_|E<hA|94FWEZtvBTeX^5r2h`~o
z(cp^fdu*yxTm)U`;wK479<H0-oO=?lp%pJr2Y&$-4#!MHq%RWTYL3-kNUbh5y{_)W
z2w_}MQrtb68JKQhCxaT5UE7sUv^?F^4&-GpPH<-(r(fKzWxBSL_6vBSwW+ZO+~GHL
zz@R-1*$#mDa;_rwo$!qVMllf}Zg2mf%Z154T;BvlH-q^cK4$-L95&B}J#ck(KB<p8
z-9Y6mZZj?HE$^P|)n|x3L6T6q#lVsNh-$B!Q0xVpD7W1y*%EHJQnCt1PwgJS>MC=F
z?{;up3aCStFHH*=(YXt<UL4FWN+d|71?2n$?wIqKO>?d+<BAg8%5bSgs%~FtI}oNn
z@2rH%6J1<M1p6$2MKh_Zk3Ub9g~y7)q$fe)T2I4h_tT*rAn<b7MgH^*Af++-Jn~!C
zF5fRe&Rto}h)>Hu1ifEr-6;znMXp|cJ=z8=^O4nC(A^|s_={KEmtIGmV;`>$KZlPF
zko^K^;0xBoQ1-xixNoPTs#9NT=Sc3W-Z=Qu=lXByw({$N+(&udzkrEYT!Bod7jb7l
zWluZ4Zmkc=LoL?#=?CE{nXuicldoLhU|2q6n~pU+ygog}gL;a{yEZ>NY>@5<dYMnj
zyxfgA`FfXu7rF8*xn|+g99YeOKkvR+X5Et@lzj-^jAH{eC9w0nkAA3gt!(=&>i(i-
z(i;V4vEo1$mgT3mG3MkD<Jv7aS<(gR^A8={50QVUEMT0(0+&QxWNgjgSl7_Bhb|?r
zn4wCl&u`q--1sAi+U3VD(;AyW$x~4<&Yt(&Gu-#D`~_Iy+Ox{yH?=Hm{{oh{Vb;Ka
z>~1LQ8F>1efMd16>a4uai^P&$T63b_mzI4Q6Ym>9m587G;QgX8wj)WxQ!-1Zl6?B&
z0<~&HRW_A0(p_f49oHOjHZHSN-=Vyxgj&!sP^Jz$$(P8AnM&$9qm1jEPk-hi%-lYR
z3Y9Xl(y}yziH|8Atn+*^UCn+GZ8Aym{1>q46@+2gj9y5$?csWFy7~)nZRft|qa`^1
zv|3A~<C18B4+3a2q0rhXOW7hMf%sjj{0!U|$r4vu6>`FFr1IvAkR@25&o0rmA~hd}
zAyr-Z(U@m(FUCLwkx1beG_(~3UG~$CTjLf*lg_0|>4K+JmZ};xT~61?(R0_N)3?gx
zq<7eaF&#Xl$l*5*1_VlVxuc2&{z12#Wky}{OCt6e6hKXoq|mPPWf#2}80(RCnYmvU
zI+yLmjiB7_FtiGYW=G{$W9N@^?mEs9`vz9L_duTAKabG|G0Ts=a1Iojl2I7$-Gb@M
zCxFb5FezvplZWi=f*-fzFsh#moc#id;2btv^GsLPlKsvvegR`b@L7}Eb7LSm$N9GB
zFF?mID*6|&^A+bnsVNk-&aqoGeaQoeWskl$Jn}SwX+nMhtxQ)>5woc2i-;45Eu0qk
z$BOYT)V8O$KN|m4^x7%nP33MgG3##8$8#f=e*PR>Sa$4hszIHc)W33E|62B~CIchK
zVV5||&~x!iCXf=PJDwL~$x*!AgY`ELWSXW?xdOYXzNa5>VOgV|M|);4wZ30~GN3#}
zU4M354BFYQz!6@oYyL(TXXA{KXyrjKD$m!}PZ!<B4I@ho2fc1?UABm6C1t`l!-G%X
z<AUAPo`z{678&UG`>sr5Lys0MVJiJ7ML-#aVccpRI!{f5#RKl=MPlbRhMnx<t0~ux
zj*Kskb*CeF;e!%naA)Od-~5voG8eqS-!mwwnY8^o_e**^>{Z}OU~TL4Tb=0hwBNOb
z!VkHplMp8|FE0dfev9SZqU(5aq@(Si1QpehaWY5au&6s1-Lqdnmjspjana>jerP9=
z%&OP5qb*Y!;j*!F_^XS@IG_oQibFNles_Y`{udDK^-%b<Rav+b&CS^+%^hNn>(k6}
zwQ_1W&=snC7vAMXejxROIBClqrlumU45%GQgJ*K61mWFp{-+;sz}VLBNa|ma|HuLv
zT>(GZF}wVe(PDp(5vSUKt_)xP5W9Tz3+QxXT%EagnyM`hJM>@Vy9t^PMBmkx={zh^
z?y&DCyVL`6*FL`WJk^CqMuLJM`^|(5#EC4(1mHA9+xO>wN!P`|Z)V915j1hR<Y%17
z|7H-QItTIAc>j4uDl8H3e4WRQvXa@24g`e<%c8!!Tg%DaDaXB9PXzwo&~EG8D;0mZ
zGM(vl2V|=)ibHKT<jbi!&emz`+3sICtka93NAs4KDk`9qMK4Ex0WeJ@DSSTY3&@b)
z5-oeSTa&N$3vg5i1wsao7N7o-d-78Hg7>%W!yh{G*W1Kk9MAo-u<krxxl>4BHTC1x
zVVjs%RQ8DsSen}=#r>kNE9ozVBAzy4oyLZEAUV`QZKnd&t?I$h50Ff2IEr3tHwYv%
z2DIoP{H=DKf2tjg(!;-^SkOHKJJ`Ua;O{25T%!IWD0M>}B<OFI_Co_ptqD<G|E{_;
zb@if`+F&2~k&C|@z4DK}*J1o4gK@tP@9eAycFt+rNeAjUA*z~nt8w`Jd)g(a{C~?J
z`aNLA<J|X3`R?^~O*7D|@+?Uk2-nx&{IlzS(uMQ8TlUMYx_>J5Uo8EDM=3<L7h~cu
z>heVruu*)2OY7oA7YO0y3L9f4z|Jw=Bc^3i0+Tzqf63-WRC_Xx|8uom{LTPM9?-Qv
zDpnmw=N64I-+$Qs6ZMLjx$h*do0<EWml$DmU_blKolOPWFocYz*!1d)o%?(&5tEVq
zv^VqeTYFz{^0HX!dq^o>oO6!Q?-5$~{(WnJKblf9W^s3VKi~c2>wK9kXjw(>Ugoy3
zy+E7>ZWnRHlHA&rIiRnv2Am$-9ap@^*H9c=-06KRXfc8`G&1uXo@^%F6kVLUy}M8Y
z!VtP9b3dVJf6$DX%iY7v3aGr3eosToQb$4tLz6uBB;|iDJRvoiPmb8Wha{JIcKem}
z-DEN-kWd{VR*V-GhfsnQUp}WIXYl+Mc;=nNyTs?d0pO_6u9#$I^v2wclB(j|sOHCl
z*o6T5h-%*f_TC7_t0%8q{=Is+j(1oJ?_jLK8L;dBo;5caoH0Kup7Gls<3`A7be9f(
zO(Ca{m*l*P!}NEd{_lt9+XLxzRpXwFeT}@cdu=JfG2B#_Ha$@;;B(i<ICD2ZaB4vn
zT-?mU_Crrnu1o|OVm-L0!axgEz&%X<0_l4;Aim^9X3qvM-A1{|ocH01wq?LZo?ooe
zgX^%0edPFxc<nuyIdF6*>P(;T2waE&>_-ox&UBP7mg)D52rGG!<DKxWhtmmwXh#NY
z7@U$^rO!ukHeYfa7A2o`3M|!=Z;G-W79{-+dy(0#ikCWWO{t(j0zzdkauKlIFDEH@
z?xC;59CK{9ZlU4<(blxfAvO4imfzI0q`}aGpu5Q5;ilBf!2xi6+I-%QE83U}8&pO8
z8Ls~h8yJAoya3Z9%CiW62EICnknJ%MTa{sl?jL3GI?<3ij>?QoH3VeC3cj%o8_m0q
zl+_(&B?rx18od@>#~8Y!m0R8i&lyQcDf(9J81jg&L&V+p4O_mVcpm<ad#*!dKtIc9
z!#_%0m8rEd+kyoj7tt(9+}tu4KbuO0arm5gU74yaJ*t&D*C5;jt{&xECtvicz^NL4
z)4K*?bT0uLLGb{gZqBGFEZL4acQ?Q~l~H(9Ep@I%yvIaXEB|iGIq<jutd&ADAK8vw
z<4Q7|)8k^iG!j}zR#a|D4}&dTf-QX{SAiuo+}Y=}tsQ5bs}PT+By&APPA%TtQD^K1
z?oD}SnILMbsFT+e#rNOix%ZS@9%dN^8NosTFkJU%ICL1q3=eg64xQ>D+<Ey&xy}-?
z!-tH4;6gLa>&~wArHmD%tx1XZ1K}@#6?mQdpodv}CmXsxw#f$;$1g6*19v2q!Vdnp
z;HG>AX5>Y_^`QG~uw4A6#Cbi=7IXL6z`uq6p8g+UT+c<FX}9RDxewxTfav1bFTid1
z7clpMF&2>3DGPFiP~z^YGQo^+3~PxMDj~cS;5L&97Q$eIYNpVP%F)ZOB8Qj0hCc>!
zr!ork^2Z+}ujzqIy`K*;LTx$+FJ(${U(^$0-ndL96(TQ``zzqbJk-f2A<WF{tZn%d
zAH@zhlWR}8K?bG;4`a@siyhb?$B6`fGnAOwS{`RhbNnJD?GMr7MdjPju;bJ37;}kV
z8sUf?U{T{l;xQAWZs2|aTmg-j!_Cd7gYSaGiXCFN#@WT^M&WUQa%~s<FeL1_jB!bV
z%)JaWDq*;{<0#(kb?1n4Z}~+!jMM+*E1o;19A%#M3kZI5X$%~d+rs))K!4&*=j}LK
z(5Z{mw97wAF_c2%Xc-7Xj6J$0z`AMeb96ByP;euTk8&&P#p2%rH|Zed_!{!;*3+$%
z9v@}L%NnTbj#f}XGq4b5VF-uIvMq<_$LHh2l`-^gUych1m##v%wlp;M@uG6AJh%FT
zFY<g&zE(%rb31i8jXC|`&>RuFbF?z$X3W^_og2o_d}KIl8%NYRSSBBZl0MfV-os^1
zD=O>W>Mbxzb}Jjfg8b2A>?Zm67vmm*V*>AN1njoCIoP|qZg6efNwv3GNBK6Fef?k;
zCC}?`B)c6>WASf;#J3yXyAQaJmJ>%2d^W@yb4U3A8e6857vIxdJ-dBRz5q9Wvm*^;
zn(yziHd%Z2Qs$45PNZQCu{^WHzja>HZWsqZju2MK!p?p$U2rs9vnAt>#~sM)dJea>
z!gO6$g(_>c-EII){uu8|*lGC>->ce>cKVS&e^0YS$hMJ(^J?!&a$wl|dt&$Xo)KFe
z>=H6)TlfcyY8dP0hz)*MQ@&K9#?N>IrAK1(<X?kSv%7TMmS@SYxGnG)PJD;+cGH>S
z%gfY8?!Z*y_g%sVQ)IW7@MiW<1#fD<<-41Iy_C9<GMCF;@eo<oME@9k|5b7WiC22j
zf%nTU;tf=@owzrUS9IZ>z39jP23}$(gen9ywefxMyRL6cgSlSd67`yop$|lO=f3g`
z!1!8$?woFHI*=I-0H@wYBQwK4&t-2O1BB?C1So<sm$_HT^?=3!VMA|2Y0_~B<2G>f
z;A+EkLv8YLI2fVwl-!1I+nHV7^FId`D{dZKZ)9xf`Tl1fHQ_9|eN(Q?&LglT_S3*O
zZ1zoEuTbOKV3nHM9KQHnWclLF{UA+M+I+sUi~fg5t&6PSp$jjHt~NN{SbgrAuy67Q
z{_?q4R6TCQ0b}EI4Ts^5g6#~zx+d=1(%D!hP9=Q_;2bbD{sL-*-|eV_S&XXJxz3vY
zs~lF)mGCd%Lv-(41!zMHh^1@4ISRDC<On^#fp_t0yKRel{9K3e7~JrMz44!7DLyU`
z-<IsKf-<LWLO$pf<<U0|IsCx-fU^jA{6iQ(x#0O~ySY+3x&vOS0-jj;G>SYCBYX=H
z4@!dr&$jSOjW(%slYFh#F`FRCg!-=GU7nTi%e#v{aQn*qY$y%A$oevdm^j3dH}lq3
z@Yp^bI#**{BB^}%P=vJc>9T)W_Aj7R0eC2Ktr6d$srt9FWCGNoIFD`gw&w=K4>P>*
z31iM_v_yDI$hu@gcf1~o)Xv21HF1cXmzd95Ve!QwKd)Y1e>B}(SZ9QHM}J%5_#4d_
zDlqN5{_5%E5Akh7)5`}TfG2d6$#D6PFueqr{^GCn<oszo;`8;5sp(UB`X!2X6iE)w
z;xr!FA9cL*XfR$ki>{Y7na7il{$1-k575VEb%Tn@7G760t<K^?nhX=Z2}mC&Z<D-4
z$uD4A1P;QZ{-7be*+qf{r(%ow^LpZr`tG~R<Fk{QhFB5X*hfh&kQriMkQ?8$JuPX@
zl<AL|!055MULLX5j)>t5z)ej$`^NqUr)|p<xEF!-*3!NtEbUnT$k*zU?aMj9tBm#`
zuG_|jw)_@|=c~unTco4s8uZ72>Ca1BslTfRvG7Ul4X_02uM&6Pn85{7$+8J(JZCYC
zzeR|jbUqazmVQhB$ln^q4%SZXvggKh`W5^hFcco=2y6p7(>*IbYtMbBKRCC|Wlh)P
z{OeU@1=rfnzWE0QR@eVo&hfh)_`kKodBj=61yW-87bQ0eAAvo8{`a?VtjjMSA&WW|
z{!TDaaHY+xJuF?&Krc_&7&V?X-9XHYl@I@u7q=guHJ8Du+Ago^V*cu@;J*(DJ`_|>
z?A_W6KgNG<ynNg%<V<d{`vrL8L4J>xvOI>yMXUTkYcOeB9UkgZLX<ds-nG)G_tfTo
z<mhk*IXShj+!wU2XFX)li2XG34U+rA;V}3G28S@ycEa_cpRHw~7r7wY@d-a^8x=US
z6v45-$M>UL!pD_*8}jeIq4qNcwk4+}UVnJvYq|Ue++q#MocMfw#BpUi3~u8;T=;wZ
zcz6*L)qC90!n_&Ef$6G=3*t8Jo_YTy58oabP@y$9&XIp0t?#GnC3-qkoNvV(pCg*_
zJOhJn=?vq0pRGdY*A%{pOpjn`=6^XLcIId0G!!3QdMD>8v767B#NQX-I2Mcx{=Cb5
z(Wt7yxw~IBXPq%We*shn?KIN$Aex4*)pn)RDINr~kDXk?IkG}%rXobEH~trQZyD9*
z)AfzgQm6vOTeQUqR<t-RR-7Wig0vKOcStFvI25<y9tac&!AXm|dywMpE}`dY|Mz{b
z=d5+kde{4Wc+R^{zFb*XCL??H?ERbBGWUQgaDe19`j@rdPXHl?i%C5_{+QW5NJ{@V
zbp7`Je{JLcSKt4+H)9bCixgA(MbD8FrP|AVtu51}Sl_exK&kg4f{xDS&V&1-7i8JG
z0&xG^{ww&uu!zg(e(8IjTuLLy`kbcfJ4v}BKkK`2`e5p1b>qjJ>b9pQu-^XnJgENL
z#05O55rkTuu#Y)8dkZ@|ogWY@83wj@>}V|?mVP6_YuR}=q4MDV`jskrl;WCDyX^5l
zL3|)kY5Vzn9u6er(zgaDtxW`;I7)KtdsCJR)^<>psJI;<pZG#XX75)m%d@vY=Wa3R
zrGKV_$80_<qUSUp>-nx|<XYA!aK5dHN&@{8)c<D)k4?+etmpJTa<Cv?;5_^D=AX}f
zI^N|!uGpIL6H!E5iK{oS>JHHh_VA{L$3Peg-81BkJH&o(%kEwDtAGCmc-|tr_HhE9
zw<q=?AL-`Z<$IEF^{%{n8a}A0*`*tC?$&<#^xE~{<N<8U$jnXp61IJKC+1$1k@4vs
z!DvyH$OyFH2Ptj=DR`^(R<0r)9VC4>?$&>8;J>u>|J7GAh?)mFWa9cK6SRJ0<*8d+
zg;_#%3Re$&@#09F!rkpbDzV7X;w_pqX%n9l9K}6N&8c^fPIzBPSJ3@282{+Jo}F!q
z=12i~94#yp5E_;Z+FHp-F4S(hq&oS_N7lyct$5|(N6F<&TU0rLCG*~TiL$7nIobRa
zcMzYhOuuHfmNMC%PcD3p*>td*%@iaT0Mkhgnezt?@A??bMpMfMrG~E(L)<IDcx}0Z
zRv&9w_z834%w)07^f}Rv6&g?0)LGp)2iXzz;4+1<+ffR22e)acz{~^R7M}*)3U#Be
zvg|w9P_01|F2Fx`xk&}T$D=G(c=RZ%B^f`5jQ!H3W@O_ir;_v0x_uRhsxt<nSiT|?
z*F0k^?Vh3sBx=_}&f3cBky!F@6WS#nzj9JSh>2!ArkmxOKlooI*m#I1AMxA?Vk|`<
zgm)OhjZCv`Ym0H5j8j|Ad!U}=v)CQMkm3@v(DAjUH9cZ{G?VM&znB1<Vj^L3HsOh<
z`>YnuK}PuZ>1ud{lQ#QR^CWyR9)(+JXCTn00w(*FIkq77EG<YQXDmyxU?6mZIC|xc
zYhE-f?yE<EK!WsIU%m0MJ#($1GsYZ`L6!+H+)<q&+ms2=ZhY%wUes2}9616j8k=!l
z(dLpdQz^9JS1Or(pl5e8fR5aia(+{F*Bq&6UN6<VYwz&p0cYWI^RkXBTNA%K(FY)f
ziUfsbI3FU~!Khc#H<8jcxgQ<;WsxlY`4Qt|?7W1G&xz$UN2OJE!@OG`q|%QO&C+3&
zc-*TkS;RdfGW9M-f?zwek~D`O$r-Hg=Nn@E&(Tk?UnCiY>F7qBg6WAIJ^<-5^jf+v
z$yV?$TOf7*K6<1s2zW<hU(HTRpi1<MDWgsMybx;*PD147+X}v~w6FiOhHiy>?M`#V
zp7z<5+s;tFcB*_+h91)k#IKgi2p!+{9Ib0)f((K*H%d6=IYJ!FI=uK2w50RrEG^Wm
zVzErJW5M|oMVj3!6gFOIG?l7>;5^OuHHq1ptW{6?dIJR5y2dy=5=mDx-nS;_*$T#!
zV`F*#r|JJwzuW~jiTXyT5|3{yh1Z?}?)-OTWTpU}Yqzf{Q)FG_=C~}g?Sl)pK~@~}
zSe3~qvC<D-){FQyv?wGhUwgRzxvmN8mMGuWgpojopt+x6^&KE3&9a%HI<Xbq84CqX
zT3$yYwP>Ti3ks*lPJ0(9Pf`$Ky>i}r>yP;UANkyylRxB<5W-?s_QYbA;xsYl^mJq#
z2{lCk&J2HoxRZsS9@67jIj9Ex4YK?b?r1zE$Q<+MnE&;kNbfczW^tqKo;Xz}1{VjX
zBN~YPCjJ`%4*0!Eyy_eBDD^I@MsbN;zoyI>1unV2#AUyMuKE6?rafU^=Q$6=_=31v
z@xz;zG4C<vdYLZJ(TvQNlM|D3-Pc+hi44h}pZaf%_)%@I{xtnT+x_q|^N*L^BSn<6
znyBA8ZqTG(0NGY;@#6B;_1`kav+n#pL*ki#LPfbbQ9T)6uKzCe)Ayg(FY5M<R15!B
z!`VPSviY}?hh>LlDZYQJc<(wlb|iCos}Abq=TE9N-(0<?8VX`zUH^?$B=I_M*0p=1
zo=@@CP~uERiTTM1?B05P{LU)b-mf$d{~>oabT2vd&Q>ciaNwZsS1wVq`Qz`X2A0TA
z>}9Gi<kAc8&y1Whuk)U_oH*?E&p*`18`y--VQyWb;-URUh5sTH_KAbh(NB?rGv;+d
zH8-n^`W4@Q231^t;rkZvPjD-_red*nbNAo=6Y3XX;*t6v|4z{?G;vJ)M+Dz0#|6l>
z`&Wdb0Y8T~{~`+m7S`)za$+93y<b={OFdDs!M9!h#LYIj&G7-xC%68Yx<7Se{UP8B
zY^iOD*B}3(c-~IqKNRLfwko6TzbDF|ICB2{9&Xj>HtXZJ^0%1T<_z+0HSO{J-5AWT
zzw?mmoM`O6HqraH*BrCgO@;r^{68aMLb30){{8jdi9y84-)6s5ozlV1pZ%?se<Y(|
z^zZK!8~>h0Zr6%FauJwWY7EBYZ*_RKwVV*Qf9BA5-uR&UpE)+dulc7#;LcPv_x;&w
zkH1wy4*$&rTunV2|1=rCmA(7#xBMfY{Qq+J8zW<`e>KiA@#*qkJmPKl?}E~<r*!{%
z`>qr@`tR3}L;d=uzccp#!f!aomwy{*-ulmo_>s%Cwp0J~<&*Fg;lCJ`Tg|>rq=PP&
zhto`Eh5Plh{}IvfrxAQMFRHUJVA-nu#`<=72dt@h=Hq;q{e*t0m#7;^m-PVa*77p-
z)$4uHA+ry6D_lS7WF+f*)qR+LqWgKgw?1J7CRDfGWk3;45pO@3nSyqLJI|biJm$DE
z3!^sV)hgA@d}S`vr0`7Qc^|CaoHwKjMVc1jqYXMO#r|1mq=IJ}+Tmrzp{J<MT*_Z7
zyc&4lIy&UZw`*<sH_?t`Lo7L?49a0<!|-nXHP$ffs#=GAO}8JbG{aeQBM!xQhgVWH
z3zcelyh-6QXF1LupFcCU(-!NJMc`r@Z416BRDn5Hs|!yOB(_tmg$v#aqY)2(UQg>*
zerk^n6PyZ6SlJ8U8-7<xrT_C`o;D`UqMp;*<vgT*Q85pa7+*!CnYf5(vn#fet=7Em
z{2s5P{Cc?+%ia7Z8eJKde5a#9q6hcoZ2Z<4qfVa(#Eos#I|fCkYude33p-dpZjk$)
zUfiv3#Su;8`nuYYb0BnC!6Zc@!ocMVv6K2tkOO=RL0fECcd%AQ?fvS1v9}v4ZuUd}
z{B|9;)Sw@<B4A2@n^N=LYOK||gVwT8xQV@?ZwFkrfn%#vT-b3&KrkIbH0P*ZYVp08
z+&PAc8#Gfgt!|0o`@zHZD`J+Mf2YDP$9>M7)NQw(Q)zOk%?t=>@M@GqIcjeaPgXFf
zz{ZQgx<s2`y|{_->XrU9#tU^ry>FglaM>Jei~n$@A62T-Z>)4F1TiMHw)8hvRM_#Y
zD<JP8a0Maf8`PwKLdJPg#Ybus78b=T4;3I+A7HJ27z3|F(O2gn6L@N0g`S!mbkF&Y
zbNa9bI%{N)t|ys(aW&JIEKCa2sxEj_YM}j%2Q1l}3oabw@cg>>l%FCnA2;?X-BeMA
za#2lyhfEI*$;@Q6TAqSaWk^k)oBrcMG6Vbq9cX*P&O(TGo+P8N>6o79nHApKoc>}%
z{x1ryT4GH4a<MDlag{0wf93D^^{i5iItrC=x3Ieu46z1km+2+l3RdS*dYvyRrQl{1
zrW?OyHBfBe{47SEOM_k004#(H-?S0xd%i}AO>rgQmbRK<nlGVtV1LVQvtHnEi*LN#
z`_AUredUE!uIc@J0kERgc2>ho(sygqM|m>UdvXpB)(}kMwIsFToI8?fq6AYA*5s5H
z{gstUoTE3NiSp!I8tlqh&f1#BUgc!N<us2uLfa}0?C1}TD;=g?*T)o%7ox~ClZLij
zLbkb?q?EH({3qZ2drJHsPksN5#UWq{o%oIQRAPy23CnNs!u>ZE+YK(XWcMlb2kaU4
z_N#vv|3CGKFLYCb+V*>fBWs>zU0SzQbREDF;#vi_))jvjvnZ|S*I6}EiE5DP0^8Tx
z`ic6>A4#xAFtzT!dX)(b>9%5&<%Qq@q0_Il3zPdiD_O2&b#5nHb99{O(vzxlyXt4J
zJ+E67kLJe*<)0=J9lHe=hQ-XIK0aS!Kmzz22gEyg?jWDh$ukClah#g^)_W~JBphpM
zQM^J9m(7ypSbqPX3c$S;MqQjLo4I}X<5H|cCQG*Nz?^9FaL0x>v(I5f(b&i04JgM^
z`+ntywqni;gs9_JX=hr#@&@_2KX#lseLp{$MavQ0=^e!IW|1<CjZYhl$7|B~tY~}|
z6egeFs@TmR=V8PD)PNF?aF-`mtuRJkX{_i=Rjhbo5LN6?r)5ousmY#Z?XTHsD588$
z!jOqQcVL(KwM`=qc~I=fv}}iJH{%VPDvcP%2IS5@p6#n?{psKy!6ZcP;gEA?k?_W3
zVPk{8eqk!P*m}d^Vyd{OP7$oSv2xZ1>1Ysk+!H@$D>?nHRJWzg_)xwew^5vJYMQCZ
zr-rOgdGI<ma#>ov`83UzOIviwcrC?;<dQF>2mMv3r;0!3T7wH`bERJJWn>%56@JFy
z8*7j^KN1yo4ZUV0=)yeUy+k!`8f~6#Rz~)oA8a+yjM(Por_Gnqk|TDDr0IWS4atl<
zF<CU8UXdTemLvA7XQC>M5Shnw{jxmFhO%If+&3+iZJw78`}rbKRLJQm=e^@;B(Gb(
zu1N!_7i5n!$I@>2_5@V$yc6!XSx&C1Y5x_#FNTcLfDbQnyGfjxQA7zD!K~~I4eK5R
zBDJ)m3cfXHqYG7rJWMwYH1(qiNu4Hxnij@~8#Ch}rTpS~DKlEBk21kjo{6v>yBf>j
zHOU2*oaj8I1`9pUL%Kz=vdlUedJ;>z4Cz*)$1uLB5K!Ek9vK;HGKPlYB`K%PDX2oF
z*ObR<5Zq-zuV}%Swts9rmNTYMDy`tPO|L+e;n#d72tJ6Z0z_A=HGy78KAkT+6GeB_
zr%-9oO-PMb%=4_>yl)v4>AS&o0jF@{7eL%M!)qNNVqA%01uDdfb#Gi_FO`ME+KBMo
z=dYhS5B$ggD!6p$BV)Z#K>!wznR_Wuhg95uBIfZYFr2lN%IucpeoPrC#-RNQolgTd
z-VYNj4n(h9VO6ugvF6q$g*2p6SeiTpQWvAO<|tN26e*;BWBDB2H~`aQSfT}a#XD_!
zqxVRKixFJ*{9v7L)czW8y@(zD6Vdh?D}Z?QHx@Qq9{kx}>)4$YjGosfosaw_U{RmK
z@Cjy}Pi6GpUU!!%27G!=Uv@Kht(#wSeE5xZ*7No#@W!7pqxbk^HH^g54_!6T+jV$!
zua0KO&gkl{`Vh3sy~SwN{F>eJHsCUvl4D11KA2H<Bt`*E2qOXX-Pjk5M2XjD0|JX$
z-kYG|t15#-C&147Q1}(<imA|cu4HPt_MAu;Gw|kka4~Ahtx@m$2evE>IoHj?C<?b#
zH5uC6|5ak1?>FGD=zK})3%yw9)Xi+e^Y$ILUNMpQaTA7`bLG084B1!DPe24fPn9JA
z(vFO`p(?P`;jWkKjv_`Uy0I(2u?7iVLw}rxNzD1ayzb0ko|!+_9R^~hcje!rC|fSF
z?jr7r$?EyK2xbfzU0O{@?A*DZg&||USvYRnGf_S6x*_8Wyt%NVw7-6f8k*J>{f)H-
zSY$5LjQ!o?b<4sspXUzV0F?uM=_qJh)mFF!yhnYiXmP$2t<NW`PF$rAY~-|Yb>R7#
zSn`Qvj}2T>D^%m@WRN(<4>x_aPiHPH;MxR?`Rd}RnNz4T4XdA^^%1twFLxQY-&pSv
z^}n%Vi7{N!8T)0qK&)g%Xkh)q{BJCV;;riedw_D$zp<+CT|607-51I3`Hl5`+O%d9
zwNN6_-2N3cFD5x-Pq0v2Js7vPCsr`P=aDOOQ^H?%^Nw69g?%=lC~>tj0r1S@2Zrc>
zbFoQdb7?eMQhAaN_gIZ0IsT2MrFZ?K8Q>1fYTRL<H<R;zzp<A1Qgnui@?wcc%Q|#Z
zMdHOxn+L`{>yBhX&2MI(q0X=+etFA4_XHAuWBr(4yM`#<kRAsm-n5=@LQ84<0=BLO
z#8E5RZCuaCArhD$OhE|ko1-FwEqvy#dAi?NfxgeVZkA6N428GPb>&2F)`|_EX0;x1
zofwn})RhmcQ2@+5&7#P4ERe7=e@U0^X?+KvB0xSTuQ1et)CD2~a7Gr(7<nBsAE9&o
z0|k&#7rITYpn5MDb@8_^0l+}Tx=*?C;laQvs2uk>dL43ORmr%y{+drLy9#yla}=1~
zPrtE_8&&v+kzd`7tztY{j4<-FM&n-4{8$ncCDMqP{?7zu0HeIIQ7!sIY>9a>8ByIM
znLni@{*<DkKca*BH<H3;9c($EzgjP2uh}>53mAt+T?_*K)i|aHNTCKj%xNPAj!`Ht
zbk19@2`u0|H(%8EW{J7*{HW@WYJ)64$a!)dr{|%r-cI!K`^a719F1Z4bn?zuP`zO{
zcU^REj##0qtvxr|qZte4av}#}@)8UD#6$Jo6RlBYdtzSueKaf=_}>Blj$Ns1e2YrV
zn)+|7zL1mqc@fW<Z<fcn1R!I@m0v|~%n@N<EXt*SW1-!kyoa|foPps9C*Rn}oYz1D
z)1I4U412|2oLV)>wVrTbptT1#DeP~rZ6;ck94M9X*J1P90*Uh$886qyn}JzOd;S}%
z$tnVMW$+v8P-E!9NXGWL^W}QE$bhE#5g;%8`$!Gv2fNK<0hdbVbG>0A5QKPCz@~v8
z4H$<9o6<6;j;v*v`JTk&a6-oJZE>+imlwp@^6_V<ng>{RRyWC$FTwX0WrpXX1)%bq
zl<v6_U-G?&c;}I8<%6d=JGT+ctwr>-Ruc~$X{2BU`iK+*y+<+fo3?Vc?do}a74!<h
zaj{~4_oEYXF3gSzvc1Bed?m)?k>9h*p_f-`_a_alN1KBTNbM3A7M&G0a^!~9Af#+%
zsj@IG)0ekNbX1jhoNdBPRos=fZMiEGMMt-H-`^?zl%o=1zX*=!6eD=d!cF4GxKzeH
z+*&dmV)gPtnx-wt5synFa)VMh*+Q5(Z-=7KqKAl4h_22qR^q<dhK8N$_&UXOnLovp
z+n|2o1PDr`JKnpC8+IJPl}qtpz)d-!NM1%>?^`jce!J}0xQ&iNZ3#<XeM-g@q}uW8
z!X`pfvyP^lpKf{pnMz?(VqbUI6IX{OWAiWvP9=k%qz9{kTvOoPO4W2Sw`Y~IokZ`E
zv^I2`^)tYvLmsYwwUi_+shej5LB#9!IN+uSs&x)87U4x`1=Bss<eJv@>ZqL$oN3ge
z-b6D7#4|FPW5a0V&hG3CseLV>N4Bf|?Vme>)3Qy>sn<Fj-XKDaO_mnnu5<<DTykdP
z8B>qL!EExIDWPfcfH45ZMAEuygcW35q^inU7vLP98!naB)m*V$$wEeM!l^WP_X?%N
zgihNw#r1`2B3x?%WvYlRt_t>@gF;hInSAKhX4IrnfGB$C`t><P85w(YdEjVLbE5e?
z*;Ar3v$#b<l@h~MLq=eU9JW2)TO%Or4Zk7u&UqZB$||$$Q@c^X`Mfk3S-}M$h%JV?
z=Sbwmu1bUFp=d5%LA{KR6B`e&V`zWY+Tc(O%<-5HF0WFMYrL~o+Hl#f&UG?K3??n+
zzC{(9k=N^@m6L+(??x9;a*>wz0c#9QeOb81aBs7R$zu7-igC70oe*bealJ^^6u&L@
zfYj%BjEi7){BCBPWy!PP0axMlG~91~hB>rKxd`6u&KPw^x&>Bp=@3>fBKoSu{L01G
zU(ST5FRa<h7MmnQmo~!WZjSic7lrb3m_C#fmV62x4SKpiRCMV^_mW|Q_3GKv&6^Y@
z^=pUu)w`!mfe6icOv}dB+<JFY>DEHW4Y^A&sx4#<`U^?qW#x1IAj|^SJ4{N&;y~hi
zHLUl##PPgM`4IpdiUNw}93@810bsI$ry;5;n*Mli#YA##3VXmGdYIh``1OZpwEkFD
z##@KiLZyn*#Y&-v5u6DwpE9muM@5NfFG1?wubk&fnNvhnuPp{zZ64m-<<qUaq*U5N
zs2=YPFFT4T>b(NCxh5w#D!%DJJOax@iHjW$*FG{9+AEuZ)jL2gg(`_N{JcC=-AR-F
ztP@j$%SY5z%qXe%!f&i>)u~&5bsrSpXt7cv32?r`zT^PBACCuaZ~?g;0S%0J2FM@1
zFAxb$u48$QOx@j^6e?AT)7i4<7=nGV?18fkWutR$KUFCsKxi86C1hm?O0378zB-x(
zKvBBUPNf9fZK(ea8MNWj3i>nT$m0)ik^H0AaRBK?upF@MzlJL2)5$m=`vd46qX3<k
zy6-Cjz!AU|TkVDn0YC;2?V8jjAyDZ8AhFZV#{i^d@kcd^r!*0$V${eL^<JxDx~y1%
z%q_G-ifq$3=@_M^I(AjgMg`cws%VQ5P$?jT<8Q2gV_h02|HV!QNeS5vzp;vqFt5$G
zKbp5(T5?Mrv_O*pXn;Kp>Vjq+qmF3p_LlDF`93SVSsua6XOWLJI<o-00Z437jg<n-
zOY0-5t3SG?7_&A!=FU_MXd_Kk0A2t`{HKhCJE&700CRlLsKCyk1&|9T&XwH&5c>CO
z(fAKV0AHbQ^9w-9%Kgzdw0kMgZrZ^b0D~?=uM5ki4l0=s-T=warRvzKuO8*ie<@LA
z+e!p9A2#M_z+E?EeX@A#H`YMV@sM%ik{R?3G0=}n%&Pj(%hKW-d)mKMCS(F=bQ%D+
zoXiqayCVx`Q6#EH!00Pm`jzhA+(l@615D?8`KJ%VmPT@HfFmTwYmqAeDUSNy{SV&a
z@OvjccoPb2@Etbw+^c?LZIEBHZQR8ijuN+%cZI&0z@*=l3@rwwmrkrcvIJB={9HJ3
z7Itzj+HDDu6M^QATywa2`Nd0$cjEgz44RM%y?ZlV0i^;^H}*6u8bgp*aQ!6XW#Dl8
zS3>9%hm3c<2y_{kZQfr#;&oJNM_@hr6YiZ!u5;=kqpSXWXcEAg;hYJ4{?JoZH5Kik
zn_xhf|1$01(<xdluDXbkcnkIK0~l`8{32Rnm=}O#0rkYAqJQ{37XuaA6_+qzVk65N
z)3smu5R@5kjlW)Lek`1r&{K}N_T?m2q>!8s*#tQ5H<pYa{nq(_xuL~@4xr6*{oYvv
zM6QAY%ZbUdj=As3Ph1M1#6!%nCUM$A*-d7R?)iYWp+&id;U(I9YyFYT;A!X>bn*{s
z0X7gyAplZ<U}8tf+bN)%cPDnzGXSno?XbA+900~<e$|%;1Zoj)8ts?oa-9vl1=5dY
ztP`JHt`%HblxrUhOYktC)BGVpo~QI>)PF5R;_Fun2rN;}U%sxbUca%DF-EU40G=<d
zysMsA8wE5CEad0VCa<82UV#5Hk9N`<p*?Rx8${UT0G{P;H3F9M&bdI`x3NEkQ41dc
zeoW6!e*w(-+s`z9m2>I3xoxUqrYN;ky8Iv)E!0(z`j{s_zzQZt`<1!Ws9~=l068vs
z|Hi^Aed~EW5Q$pJ1BQl2Tb4x%VD+9qcv?GDS{0-2BUh(8Oa!14W6SFrb_t_FdGjlA
z1}}_DQo&j=AC!{fvVBCV4g?4xMDdw~`!)ZTB{{H9n+;hTX{WusXQ(o$e<fgInj)$s
zjtW$NuP*@`7f-hnH^jEgY9uM0p;YeUt&k8o6Y{%KDor>J0d_NQ_Ye}t?iUy7>bN6f
znTWg7$+SG=28vmF`UiTW3Z^Q&^`VS*!XrbSanpz)(;_RqM2n?}%pIb>^1+Ni#$VRi
z`3(q<!EAvITeHTpIuI8L`h(I4P4qjlsGG07GFr9d>Y{>bPmqc?`Nc%ndl}IcxZRjs
zQuftBJGQ$d?=&n0N8Z1VQZJOt<z0+RMSi?$ds3|SJ|*Hse#ci!D<9eLsdMQg>9|d)
zd@uW<hkVvz%wh@=Wab3mZ0R5_G-rb+Yp`WW#gJhL=7avg7(U&UR(E9Y4oxs$Zq22-
zoowC%sz$G*ap@#E-Pfs!UYkfWO7e56C=B=Wd^s2pdDr`cK#FGLf9l|Ep#s`>SpR?*
zh;DovntC=XS{inCmp7>%3rj3m_$?L%n}>~OaV<v)OE--&3w|R}U}z%$N@Pvw6)#CM
zjPOMt%(dv<AcKr8ZC_j$e^~o!S&wwk+a6a%v-z(hTn!!aB#%tbsh-lceEYe^Oy&G0
zL{qSu*Hfr+oB8{6l;-DT%|e~BaNgZ>gveyq?E2mqJKCOL7cr#cnHWlH_^V)syPsbr
zMCt2C3FX1NWd_KIa_IB~W0?2bqC}Ytk3raY+lcLWHothZVfEHmqx`WAQbE-`o9=+O
z{*31pI*v?<6t42(A2@GF_s;{sO6>bD1+=GJrvsfIm(KJrQRzA_cLrK&bJGQwPxE!M
zB^(W9^6!!LXUz?RDR`#8)W+-u>H8w@FWFYlIGWp63HqF3TeDgDt5$Jz<CH0At_<(g
zae(}l!74ez<?3MZD4b9ieofGcJQ2)Qw>#=-C2+A`cdW2Ah9TLB=Id4@C_tsLv5C0h
z+t)?vkyxb=jUDfLZe)cTHrWAeEr?>d;)PlDb0h<GsukMGQS)kqh;>MRt|qNC7MYjK
zJ<c*>>CW#;T4zv$o7stmRwta>)rRvm{8~*ZwpU{i#r$;AbBhNFJRMAPee=A1Qxk30
z8W9xdrdDS*qrGvE^Sz7GmJiKfiDlkmLJ&9-Zr!HHdrKfkq00V|cq44|$}TYlhEpwF
zZs#LbGDaJ9J41rZw%(YZt^=;a_}uoY+zAr*(5lL+Lm$>BD>j73W`;$Zh34Xao}cl4
zeE$9E;1L%OqGqH|Fa~0WSj3ZatbUxWM;5?W#H1jf{PYR-Ve<g$a|F^Ihj%26ouWl_
zf!B7SiO8&nJIOTYbYo?W%{4Z2ij3^VoB+H?Mz|3;Adl}Z8U`{eK32)oD;l*LsUx5$
z>`&yL$Q8V6iwyJtPtEiy@15M90?U@^ji;d`!*J^(G^sePA9J^mTYdlaD$0506rQ1u
zw00@sE;oSHa>E36is_F0=)bZ>f!U`t#I`lCnopm2$WmH<7ypf=+<fZnNOlyj41>SK
z35Wp7`b?h?M=>&r2WVx^2pn3ukR>dE!WeeEBFZ734A>k%d}o?fSQ3ixFiH(L9Info
zYj#UuhpD+OOm-#hi^o<FpI5+9k=xue^=BHeh%tCeAuHIp(jG*rmFj@bQ`fKFD3t(L
z&6Q{~Lo_xkKAzgbQ?>TFO!D;%+GjL^Pd5B!NAjm@r~^ns<E&g;Ldg*x;(~`%DOXM^
z{w8RpdLq{&4;apC*dw91b+u<;JJ3#}G|c~^ZYl9|BC`JFWIUZX<x<ti=fPs$-1xkJ
zr@G~=$5GBnu6o%dAcPlxN9crTF3*=sUDB}9eCY$><9l(d8TwAxvpnbpv7Oki8^$;J
zOLmOCy^NKzMLe@B0Z-%0H(b#EN)(SnY-AdX+)ipkPOEGdpK^XXku1>7ZcQHSOS1+~
zYX(<*#DLsvY)O7Z6kI?MM`9ljnSS`cO$1H0K9VQ}u3M{oa6l79LwRyVvvR-)6WQ%B
z_Ic2VUq;O#TVtC7r$CN+D16%ye@3x*t_jhpJgc-PI&`&PZlO;7xhV^poNUxZ(Q!Te
zZvPlFQ=~U`MM3)DwM3|nueW^OdAMGmeAAc66R{9+ZWKomK2LKBHa5xLLq!WQX7e3;
ze2`!{(ZzuBRTAZrz87pkfCN4C!kWY3#nF8YOO28%g@JP0Ua^|8o&rMJHoWc?z!$?r
z?^upUMc=Xlln|HdL6CDU0nt6}G#jyOx?_4+pPGs$-A~vkoJss5JjAUqSH-2L$quTt
zsN@uhzw}kRq$H=#QiCamDI#o(PEU;Q$i2krk&~|BWo^HHi8WCD`S}9Pn49I?_lmGM
zYp4E3UM&o6s?(zBJai;~yyP2kb*Yto<n(CgI7S?}#&fx~Zih!nF`v&X7v-%rds~Lc
zR3+KDbOGebj>k)k>}bB-%L;e}z2U2c*IUMgY&ijJ{Czs{4RU>$!0An>CS`NDaK73C
zA6osX`g>J*0VTmMc-3=DUhFMTgZ%=5rWNE>fqZMPX;3nn2*)6IRtFO4=2|k-HKgO*
z2K!nxfhn^`p9k1#d6B>$&&PsGhFQm$QsRXtVMvW<z1YXvi3m_~5+*{h5#h{MMG=_%
zws1L?TW9Rc?)baG;wl^AHg>KSA4wq+Se1AlezDDj)f3;4ahFRar1?m}ptYTvX6<8!
z5hf~A|BDCa%pw=*)7W7<ANp;}IRxRMz0;Zy(&8*KQ-5P3TWdG4mMu-<E)AvLxNluf
z2ve1Ydk0uvI6Ug<Of1LhROFYBJ7y}#*&X3d8vFvR!_;Ki?`8LgT(r6Pri*`myWJFC
zsu=N@vSieEw$+EDzLb|WIKP>TSh)6Q)Zi8wFYZ$QcL}5IWiueag@-=zkwPEM`%)Qn
zjpY{?rwpOTsm2B^&RY3(!2&yp=EJ!PpF`VbUB~9b<g!L4I2nwdesu;p7a@Y<AH=<o
z;Az<M(5hFjR1K~Yd}`Z}(#=HFT@1oY3`$bAALLY$D{9rxtClxFha{5TE_2e{^<+4t
z@DmH8>B+3CG~;1$$}sQ9_20J<+y8vJWlD6O(}{8>4%NQJN(gdt7~wm&RIafkSEg2>
zQ?!WQgQ2tMzZX@i=jLv@1&-Nkl2bQJd*xWDzoN~uYfY;<e}lY>%lY85wU;=FnZ|vJ
z+{%cgm%GWMnfx@RMeIWdckwsXy;x7`aT~6M2-B^s@0U!CwZi60K{lRbI~~rqVyKo^
z_i!qh6MEJA1SBR5vKC572f^U<QIa>&4*Fvu_Ejp#l6r*<CM5wYchh-;_}c-mKOfgb
z<J*kN7xfpY=cp9QU?1W1PXy{#@$%OAL!v)DN&lAdZc38@xGgsPc^>+(v{ik0tEix>
zp<E9omEk^K$}j5>`xabrl4dP@y11m*E8NY%Lg3QP{J9><#!%fDtKX-ra@CuDTYpa`
z7Pyp$BT{Ru#R1aKaW7S;<<UDf!)Rw-h_o4a*L%z}F)!0DYb4%I@Tw<ekmA~$fDGQ?
z)7Nh4INleb6=d;2<%45A8C=J8{a{g>4&o6Jk|?$$6`>n~uYVqG)vLb6tykC)CT!zo
z#<)6}$e5eA;(w*}|6s%Ra-%aM5#?^V+9I-Z*Ed7ApoQ0(ILT8~uHbKXgc&wZbjOFy
zqzfqhRlnvktB*;`ls&$?KpQ#P<88)FQT~CZoGJIj5Yw(C-N4Pcn|=lJm@PAO-5A>p
zut^DF7!@q**VG}@lTuswX)G$&TS=Xe@`@Ho)Bh+#NB40Tr$I&AyP@amPg`aLkljbA
zG<3$j^^TI|iecc^;%P+L=c5zr^!-?HVU23th6XTwWQ3K9`LDu%P|smmC8uQEsb);B
z?ce2H9lmJ)25ITxeOd~#-O&6|LXxSIU?Xh1yhD*naa!`ydxo!mGM5*sKLjF=H)gjM
zNCt@+qjXqK_=(GBBXRu>U&rV_nA%G${=ly57~nSIP&!gHv89=$owfpLguzF5Q)1&o
zuRm+Dan*$SctsdofZwq4=^*Op`JYy<w&#~K`E<}=Q+V6CZM`e)WWx=x&z_OxHmf9A
z=LZ1^)|bExBZZ@0?79#0nF$Z{im0uM)Yx6Xka+wEmHZ9^r%Tu{3}Gd5oD?^G(qr#`
zvA<OCxdjP*;cl||nQ9E|;FqZF%RcO|$Atu!eE>w%R?yJT9L+I>GGplzxSd_r^vQu?
zVev14WJr<ZJp$+Mvp`1VT~2pfo-Fqd%c8wH2dfoztCUMCS+?WJur`$l1uMoAJN2Cg
z`E1n#ApD!yH#Hy3Bf4D0>5C7WHVR=N7&<d)AOF^{W7o4)Q~Q~tyOJBfCt5jk$?u*%
z>Al#}60SPf;|V6ZAs!LOSf|N}r?Q1&riMemG$u>SRSTQbl(T0vy>@UD?iyDfq$@q1
z8j7Eh?K2b~&EqzkjQ-G;2SEe`uZ#u1guk~6vs%EUJ}?RH?W-we?NWbh*DwlF2Jvex
zmPil<$rSa4-(98rM*Be9TEKj76LI~bUB*=6kx}}gtti`%I%UF?2;m0_$anGH#MYYf
z=K;<W6V%H2IUs74x#~RCxkLqoMTmx43ie(t*ZwyoVmXL#8)5!#uea+RUhg(5f7lz!
z61h`5CfH=$Hk7({KZh95MXp})#-i$q`|k~jID2GsW1NWevN{XBe|^HOe6r!G&P`6~
z81~rqxK{@Es!d6<p)jP7CwOE;u6B32(Jm!<2<;_hqr~Ez%*u|uTDM%(BRUF^=Xm9}
zuIv7P_f?Q>tfC=vqt|&->AeCqy$+Loq$E~6$^qFE0B=%LT!XE)bGyArjnh7njoH@a
z#nMgX^qO`)MD5ne43KTbgxF1w3cARR7UwTu>&rpx-3~P@H`-5m15;e5R^D3Tc&b*h
z#*2c4c2Et$<}?MyJuR*j>Ea>Xj?Dt;D$!|?NQBKT#9Rv(;K<&a&ox;qI?ex}KW`D2
zVeHwo$LnN>LL&Y^wgGvY^3A6ZQ{%oo(suYJwc%F!mZr7gcBS!9p{mxhx&^<8`u%>V
zB{|BT|EX=?f2*B;J6+sKZ{R<?gnpGqy6qlfe(L%bt()n4WTe)-hlX$l*P5h@{<OhV
zQH6|s1j5n*=hwraP9nV%hm+fAKflZvanXUYL?h@hPf<eaY_&Vrclwg!(XeH5CnNrg
zPZf0X2@Oh@VUg)Zc0=)CCyol~xmu(8Qmmyu9y_W)-ySYDL%sFsURn4|agX^|cEIBD
zTv~jxH`x;pk+gi5<-awq|NTEvxSYS`)$&M-6Mhdjc);=PRnf6`NA0kfSg=XstjpXi
z$IIGivTt8VId_mfYwxUGA38pInZMz%GS=w`KSY&HY~m}-(XIN3>0|t6-ms~mkH8q@
z#q&fVV&B9BU!xxfd*H9r)vlb(b?Fp^K#pTZ27S#uhgk)A1>RrM=9#P8V-#zRo$cJP
zHK8eby%2YyZltx|XrDeTLb?x3$mvib1D3X{oT=rfNOXd2OIzE)ihmM<MWdV|ERBZW
z=lekLPonS3Gv;e(HCo!R;zC+0j6w;EaP{To8JSgtPYAG5hfBkjK@SzvJI;l*e9L5o
z^*HJBbG5{@rX!`3wW`i~!2L4HYkSSjf=zrC%FQj{^_cFIzEHrqqr6XlNaHdWtVjnR
zLQyra#TbQdF7<|AznK=JiF%&+p%Vzny(sa-Lmv4c&gJN~Y22B~)N^iAjfvaod1-;@
zT9RXv0{FQINjXEiLfqRvZRIaJ6)hc5lihjqGYX%*!NLme!_w=+60X_ZA4Ev8I^7-S
zGM&~z?aE*!|E5{JH-wcFq*oWhBL5WrrZXl+y=rr<;8a-mNk^-9D&E&2@Lq7?I5)?&
zig)~z)_Hdj9hGlsMUGyn_e_x)KWRU>v1A;%>v}<V*tqSj1d+W-qwIJ#>7F~88{^zg
zuXhskRG>G!%hI_`E-sdCG%kjRXIacsZ8$QVmDrkIeHa%%HuViX*&QDlB^#|;XvJ!7
z_a^+XVj4X_7uTGGbRuu)c@Nvxd5|dgbt5(fHpc<}9J(u@h!L|=Jbd0zQ@POX&svGm
z;#(hRiTkuQAH|4FQ%S3@vu;saIn+RGmXru>YF@`HD1hj*lslDQeM|5X;u(wLAYu?H
zT`i6V3uIk5zE|q*2;q9k#mz=Tv?=Y+i_37Y*9mf-)n;vk5)D<5)YQxEk>WhM2zILF
z9jLdWvz@H&$NVJKFkV?zw?EsM{n6xnI_OB%XAl~;j|$nF)LK+I(5tl@vPz)^axSTI
ziUB!=?CTo8NuZs+7RMKj1sXy(?@ghHI|?ybq~tPHoL~xJW++PP>Spbof_zGx-u|3~
zs_wy!5$mRSSmUXFyiJ2h!x{BGW78Hlt;7DC;K1Ev)>SXhN75m#U0SD?9u(t>Xxnp+
z>r0JDSn54qdc(%+4m!4YHQNK}i%XB)2dnwNvFNs3$*6uQa(D~TmN-3eQAW<x`^~a;
zZfKTO4K|>dCBpIfL(1)H_cz3`b6vuAVhf5_Z(Gm031g-(3c}_ho@Z|U^`*L{kO#_b
zoBFD%fEj}I`gJuUXJr#4WP=>Nx^%R8FQE&!z6oO@48F2H`Z(@R>wjmnYFixRae03I
z33LeFRf$X4O3ar_IIi>QdKuq4pepZ^ojp_6zj*(IR+l!eX8-WubQH4RQ>(pcYS4~3
zHOkTzD<D4F>RCR)qpO`_@^qS-druQwl;UKNykCJ)s542gd+M-vA>8h)TzznrsuP-^
z`pI=2@)BHzy|5yY31l@la>|)0s{YlKlc!x>etBoVTEC}cr<%TC53tVQOc-kKUR{UQ
z`HH7W#9`<o42s*dQ&yIfeMqcJuVWMDAqg|zMohfB>DD%nLP}PQsugUhzFm!XP3V)&
zIKj%x(3grqxnYYG^<%asD7>rMWzrqh9Y}JQ^l|NJDE&lY^LZonhiF)l>40bxnq<Lw
zBx2+uT3qC8`+$?UM{V8wFw&$?pw#>DQpfP>!KJ|*laClJ;Ul(cI{04t01c(!vZjo|
zXOaprm(k-@uT8_)r<KTTkmt<}F6oXUT^Z!Mp{K~@+elQk(Q`tu5SP}AQV6_G8SU4G
zn@5-)w0?T@WLk(gw2Z%}3_H3uo70bh>XS;oc}KUeVVbP#w|dT^S*Ylz($X>T06gKG
zc*%ly1iDCn@u=xNDOj+8Bh_g#g8@hy)HwAuZ>u#b$bMvJgS4C+<-6S!7OiM@lLjSp
zINhb((auN`befwp^<DO}4;z(wn_X{EySFypU(HDqF*b0vG}nMp{&Y<fG@3fV<IttB
zKASlet4xvpd67x@XR*qNJf^x+IC;md^~{Ex_R4BAj&e^E!_9%(b~Bpt?T!^}CIwEK
za%t_{wB|?r%!Xw3621t(?SXJsh<NzDd-^S<F2UBBTWICDRIcNZl%th;go4(N<NC&;
znVEOtd&Z8!U`wJo_pP$qp!ukW=ugZf`}5ALT@GJmT5HJ~eZuO$b{8;WK**|C-!seu
zkE1jmll8bzzNhh4jS7lyEfIAl<^`G{;#@)Hju3n*ms&dM#BXZPktvdKIpS3OJRlGn
z>IQM!S5BMC?*UDkv$2U2>{@d#1}HT;BKPuFre|P?-cGksnCHX+&bewHzQA&wmsW}`
zVF#rydybOEUT)Na9ZLiwQD@2&9sE%G+>?v_XDo}~qkjEpg#)iI#@rPKy2_TjvE1;&
z669`%P15~HTMo*Zm`dIe|D?K&mQskXjc`xdyl2$RkcWs+V6%dZ?0(M|`nC=$1ODn2
z3Db$pa+`U!;AXs*GmiDOT`?_{O{Hi8*Pt}DXvSt<a%LvQ3|(i%bC#{rhF%;6w@N#0
ztMRm{H$`Y|ncL2Y=@pdzmlNSKh4kiX>JR~vxU6vlg|J`A(a1#?h@xYqLO+P(JP$Yk
z8Y8xwTj&+KuVzmc)Op!Zclx+_mYi)*)gT19F+XEln5ymOjq+O@b&uWsj4kdb1PDgq
z4Ax6o_QnztAT7tf&z06nOFi$^a~L~IQ9FsI%F36$qUpuBY_;6@QA3qG;*D|=nAfU+
zMMbB}9<a`heosVg`@oW{7?uw~X0eq%hw@29`_;Qn8qY@cwzeeflYqpGE`2^iPIYVF
z7sr<>otEos^?H8gv89YaZC2AYE>5<{wwf_LKIe8@$&Dt;8yS>7%^)SVU&kSb)&tcg
z!B#ze9P_BvTEBLxW2Ejpd?VSt-MwkBGx+M_6ps$HT3Br7=W{s+6>e44M>6XNDK!lr
zKU(?eb0g$?f4aUUaH33-l(F?-BBPpHFJ)qO0;)@4T~!M)gAydGMhN<}D;uI8gq$Q>
z|7iGONmW1QbJLAC=5}p*7zj`BlW^K9lTa<tS5Q82nlNH2IQ%#cmEbC;7CO6vnt<pW
zjks#Yy=+m5lO8|Mt~}dEG$LD9!w!y_El_<KLCtOQ&?su(y4r(72eS2g-q~wKpK~KF
zb!*UGk86p+%jj{lH#0dLKJ8NXW=d;4SmCZJeock77y3DQqfPgdFV4NfIp<eUp-DTQ
z=$37DyTUL7KZ$QLnloq*7?6Z<rk?C<Ipg&bXovpDXOWZ!sodo4GbG2#WY7S}72j82
z=$*Y<hg`Ch<V1Vzext|T^=%rq@jTUKnQzuIVtj8D>))%^J&N8yO`v}?hAW~zc^#)S
z?S@p-ErQq!j}>|hm(GuQs0t&MgHibP{p7VT%^J4Tb&aBSorV}{H)RbCakfUr`69dy
zmQyvp^d1VHjMRF7ZnoAnoixr&CwknA(I1ETJ2RR?>aLn-sF!z$6VoxsDMZXyS-qG>
zX*yV5G@%$+?D2TLAI41dy=a8AfmeElL?_RRn?gO(q~1riLT|%$_mwJet&o{vX@7Y<
zWHnzzb2)5IzCk|%$~^qsTg5Ps=^)u6-g<uO@Y3U<_5M+H>wraZ3kUKdp|6G0>in}n
zH0@9AS}-vH4_r6RpfBSd!fgwIG@YNw`ql6CfS<ZK3^sjug!Y~lsqylps54J-o*E_`
z-s(^vL1!4XvKbaPf?wtx9reAf25tt2tiQE?>Uwf!xq5mOv2}QInf)1w&c7}}u|ax~
zN4d6z;g?bM1FO3$y;`lBPPSFs^<0NMg<)01rQS-_+aOP+HM8ttt!$-hmbz5Ed%ZDb
z<gP&@jrl3N8%56I204R%qc4UzBXd7+#J|%vjW0%BcB&Zf7+lP9*eBNcrcoD16}+zk
z6LAo|r*1ix`4EfRvb$%eB<@YU{=dHk`0xGW7cXG0W%`YkD|0@Iwe}I~{w61sIZ1h|
zat(<4qfMFrG9q82JAnIktHO-cE$<B|wun!MnkFQ|VaU<#EZMn-*}aJ(f24Rk7G&>I
zNB1$_o-J?anfv??-#XhaK@kLg*ZYLj`ohV?k7vqSvR_mcsyQJ>ghc{ZA|3nsRv8+I
zUGRDUuca%8{w#CK9JX77e#t0&hNX4uQv~Z<YZ6hhMmE0WPAL^mYK27G(fprrU*@Gs
z1Pc$7EJw3ezge1nQjCD(TE38=Gu}^h`$bMP5IBVCs8|C0BV9<jV*G&&UyapwFMT*8
z@0MZAofZIVbo3Cv#+R7}|4VsOccJqz?DuH_89o*@x$`wyb|$y$li7Bc{1RmYIJfF{
zL_wuS#0;<~X@9xQ4!<JdgoJ3^3>AM>g;_=AH>RaHuP^Q`d<%n89*G4V9z73^Wl1RI
z`p{QIPg8DIcsTJk7jky<{FO6FIZ@8@B;Hy5PLj<kx0Qq=&8cFI9YMtFm>iZZ?31|W
zv}%!C`yvnF>?F-f3#N-dn!_k&qpbACsjDoF;<lXPB_q3&1?37V-^Z$xz8F6a%`K9?
zu0u@mtc}hm3g{=!VjmfyeTnHycSzesTFfI>Obw<iThHCb8PpaRR)k$?`YO2~G3l1A
zY?AXi1KESXDUtDJ;M=khi?2mH!xo6EPaI5<z`Zy!Ax{WKw11BFMtJj4Qdih|8;rdK
zO+)~<i9exs?DbT1Qq24GGQu05<7IxAqk@~Mq#ezm&1HmwbvNX2tvBXq%W<Wq>ehvp
zt<sHW`%uQI`|1&(P^cua{PJ(`(^VayY?L6L497_~8M>%o-LU3|Tr>x6mzwtq+>Q8W
zl=JBhva5CBjGvYKLafp8FazANoR|^Sd<{B}9vdNW(KrbFnBQg}y?^^hCHDw}ZZHx~
z0j4-_-(sV*@lmX`jBe0Up7eP6^1g{KkKLUL%LoCuO5UQ3VLV02(HqTJ*#Ol%`8bSt
zfPB<Ff_e|UM5;tqw89Sg%SbArTBVvr3n&hem691}vf7xKf-6j!%cV0)-O>ckr;=&>
zuz9C9z9(EnpL6DEqFO?|2uKGUIO%MwE-ZP*oig`qDo)h0eyco5KE-z!2Dlw$y@kcZ
zNpJP)@zoTM)jUeV=1hMS>5SytKJSv6=oej8y!QevR5RsiSW(V@1q*N<<?7KUSSP1E
z-3fdrSu)H%9m<NVlEk@H^4cr$LH=1CSTT)gv)EPhQc2-0P1eHe)uRu{L8^jjot%X0
zj`#TVYrMF7TV{+%OpavjMm2oxvpfCphIU}G4h+nG32}vSpu|l<$2qb^nrDRs8rDJ5
z>kodxJBn_NN8PUH;ossCtqtk914|7lKN#WaU!=0q8?B}kmH+9D{L=RH99|?lBNh8$
z5$DF0Sx-TmZ^OlHeuEhQrR3dNA0T=8;r7{YEC@T`02ruag0)%j38%8<_K|0DNymqq
z_b)al+q{iH<6o#G%WbR=10F0_WVFs070t2b&z=b5=_d1e9iuttTD%TfUw{S+nhVD4
ztXx~skH#&I<;y=}DA@HgSBxH(+u4aOL>lZ{<8wzAm?*b+mD1}7x?4|G%NVy8Y86gA
zDz+<sTg{3X18;1WphqotNacCb!<TWRpRVd^z-w?RkRQ*5QW>4ieT_>kDsP>r{S1GN
z0udN^<V29|z6ZJ7v(xm^;uqt$?ECO^k&FiI@Ln?H!5s=$;Xt|SqOZm>0k~-tY#bF|
z#l#?v0(drsKN=*u=LpR}Qz(uGB7`N<wfx;m0y;5V!)hZ}mT8k8mj7(}#k2Q#DhIVz
zP5h90Hh#s^JC~RqjUw=>J&FlZh<((9c1_=|w|zd4Hk;B=wSYGtFdngFqKmN3PV;R>
zY$Tel_@G&fhF9mSV6MsFGPSI}Vzfn7k$v^<j@gIE!3iei*Gs;vuV^!wKS{8@WNO<I
z2-FRu$x<j-ij+X7KNrehX-<|e)LsOyO2v{%lC=ympbXPTgZr~=8iJ@4L)7)vUr|Jz
z8@csWTke?WtHZ7RLToh%Ksqm3HpVebI25<KyjEM6>lqeCa@r&9>k?8GqbPmphNtf<
z@JPB~TkPFYeYC&0HAP8L?D8HH>jWm?6&!i83BFPf)Jb=o3NoQ&_gIm=lT7_kw(8i{
zB{g_nYg5tV1p6rsCDO!B8tfz=Cu|_AuPEPb_{<D+nwQng;yS$-S^nkRl4F@LrVe68
zESUeLOWL+Z`*pGM`34U*&CP>dL7R$hv4bS40rUE&4lVfQ{oyNlA56PSq_fAjVmLxc
zC8>>bLddINvpf@JRSGq*NbeaW{fx3)fR4YcOo!Z;g;zDOLLWSTB-)l@5`s`QPk8q|
z&$I3)<q{>FU8LTTn0+Lvx{S||&&Lu`gg-;ph(A)MX`^`+IY-}@+tG3r5vUk;yFjK)
zbY@Lg-y=5bR)KbaX)?tw1lev%0T#|*W`k1Nc|}6@lI~bDS&xdoD<}=K*D%>-WwHni
zaonbi6tc2B*d$T;8aH>TgqJG%vc$8-qDC!RA2Gbv9c8;neGqou<tb^p8xk_|y)MFZ
zT7ZWE889C+Vg!CT{zh^m&PCpsA)>>`GQJ&~YVIzX<cma}ry0GRR^_21xkiMCQIak9
z1U2mRkAq#q{KVN7^JMok2p&h1Injil{0!n#5=cme7DR4Z#zVSEJ02~K7rKN*f5Z?v
zL*l=Y_D9}Y6wv3s>B*Tpw+|u1NJfZLtEWA1sl!t~cY2a*DA14`E)Jt!u~POVO*Q@%
zIHZ~j{=m(|+r^`<d3T{KT*p!EXRkq<C|!xX(lnBEQS2ayR6h8<ng(aihZ!LL&nB)7
z62-}56szbu{9aW!U*2T;n{MvtJHI)I7^?eW^xo++(}Ei&_RPEXKd4_$v$v_M417D4
z&1Q3|qG!5DGyV$4gH*k7P&u=^jgXItlDD!rrl%L<qaVF>?ybvDp-gv#uo_!$5|hlE
zIerjfb$`=ONIno8!jwZUNzZ!_Y%4l&oA=!3Uhl1|Rm7G(0^M5^r0a<g9)TAopd)Hx
z^r$3<&ElV10;7>N$xVJr^98F0fKvpmM%H9Dc0SuX#C_@S?Gyq===3LBT+yfOGCfY3
zRy5=~u6J#|@n=i&(@LTp*{AWIPr-smJ&uZrtXx05Cjt?qe8k8JJhmzS;hV1e{KqDo
z=_iuO3_||Bf3P*&wVf5AMccTU$haqGwXPg>yGjUul;zZpwgp~oG0cL<$s2)z^15na
znj>WspbA6mOhZ!69(|mS?JtX^t~-zilL%*t1Qt!}O|56>V=c@0Zwk%J+VtvN)-)H<
z%2F(4yb1*q295CUK0|&<xO5>pQr+@s0|@tRQ1#EZ4`58$Dr2WKJe-iv2Q%@Sn5pEc
z^|&#<O>qe2@X82&J9`$s1s5#!Cn|hWpi;r6)T?OD{g@tH6#dN>)_|zP_4Ii;^MCPn
zmS1hOQM>O`p@kwPxEC+(E^UDXhXzP+C=S7jJJj&v5`q?o1eXMYwZ+{D-UoMgeGc#W
z6V6)aoe%SE*36oFt(krAeO<rHuR*IRLAL654!2mz0xN~k`eqUH!8-TkpW8kTCIgRE
zW!!Cw(|G=}tZy8R8*DHAt32MLE0Qls@qz4J^cn*$K5j;1`HLw7trJcrQokA`C|DSb
zk{r!3vi@1)UY5Lg#mwu7<~^hw|7@XAy?FZ8X4(AN%;?nQd%Z32yTujRMuCS7p-mEb
zO?RyLAZ??w`T3wF(c{Js`G1g;hl`iqNzx3cNVH>3(wNTvj(9#Fg|x|ZoU&fnD0hOC
z_87MpwfnnUaTjUA8>6L4V+!ascVVLu@&inh5-TB}Aaxzs)$f~F`GrL$@={5;J(BeB
zQE+<(ScKhuP2gjf_=*<-3<otk7{2>+1DGfu&*XsV(Z$gkwC_^n&x>}~v)UHAYIRUL
zezNBD0~d_X#jAqru^)pWbmHX$l?#U3{d_#q`pz7^ef^Rt1Z^>pzW*L!8JF7ZjXbwg
zCYoSXdTA8gh;ac}3T=+fkfuEIfFA57GLD6;2vPm(4s0@qsxsYu3ni1ultoL0M%TBj
zpopkGF2JfNYyz_O-?ludO$k?a$*R!%%5C^TZs_7oX~%~s&S)y7N4I6(-$vZmK5k?P
zjR@-p{%JbKkx(G@DW8~nPo&p=%AGk|{e#_{@E7gNls0vKQN23JP`9j_{)d|l6~y1f
zp?f_TNHW7}J7lqqsO}-w@}T+Dr!-efdAt`2hn(;g^`@G4a5}4b^1e4FqQYCbISV&?
zv&AQB9dGpNNz=Rkw>`@L%2h9)|DI&>muIFD)my~}Z(I92w119KXDK|${z&>s1gEge
zdI~fB7nROl(F~E@ZT_M%@;{H|r`SN&Uv(y+JjSueWpi@#$aEpo_hAx6l|x^1eD{l=
z!RYF}$71O`!DRyJrmfvvx29ZRU_K2u(v*}eOuwzj^w$zZV_ZGcHc<C}k`Auwscf5H
z(3X}bTDNJJa1|eYd3s~f`Bt4Mr%&pi|E*aZURo<T>&BUTpNI5j^3vq;lEdUf!z~Fn
zx>u4qbX)G+0l}DYU_UI}+d5qal`<uTn^)r_qcG+eQ}>sN<OvU|3)wdsjT|Q3voVw+
zY%s5lMMLJI3{?-^XfPYLCr<ZYAu$<t^`wl_30r2ho*s2jW+vjtsuP!=gEz0rsHGHy
zl(vfnc>t-^6{|{4&O&PUj&mKdXlhSzeC+L0SAb#J93mrP!U^bTBLh(Iz8&nkn|n}l
zrg-(6c$b@HB8~@^rHYI0kBM}cjr_!4zSZ}|iXMKLLbMS%Vkp<vWjm}VQabLdJnnvG
z`bIUSFAVStLXZj=&CFCujL*@I2I%BRdC{-S2j@#XGp3}kgJAnr+C7ZoQ<z90{npnp
zc_B92-f>)GUb5y|-*h@X1eY{Gf$emR(6maJ%mRD*s;8W<52(XG57|QH&KzH73f(ny
z9wV++iV?*a!;-xS0sx}IqkMVz&EZlGd4QQu+arNj-b@{2I-?Jphm<MXLdqi+h+J2&
zr747R`(l`ls=ejB(Xt`{dLg$*dKzh0kT}@$KKhm@`67Zay0`pGMI45s+K)Ld=I425
zoZFmuN{-7cdF1p#Bo#z>zKZD!qpT5%eY|jc5L#5yqcETUt~jYe-R-pW*{{SJaB3Uy
zzemK)iUqtf(be)YWtM&QGVi6u+EeHi3k*C?z$>Fgx#%ak<6HpVw18akS~<z88IIl{
z19(#kazZ*|?|t~Ei96z?u+rp*c%}E9DTaoG;#-ohX5{c&^5k8-d6x2I=9d(w431b9
zkU2%S!IPh*<o*Ow{?s*rD-@gJMnMjcfEo-%W4~0ND&1y5XS{kG-b*do3=1xWEH`{k
z%JMZ5^&8NQ-IlSIL26kd#cVd|ua^=u$tIl28PRML#rcy}Opr9V2kX!<Ff0-v+_*&)
z>)vh&t2>8Ai$3`-8lWEU*(t448Ikdr4oTtH!y9`sfC{Nl^#&@}kdwP3h00wzfchr8
z=@m$F8i)LP-9ix|d^V1>$ST7`r4U&1Z}rGe_57+F(G88n$a1_!DgJu<F$$IAiaYI+
zj^yl^f959pB{yBdb|03An`O^*qxoO)WE@&EN%&P>UcNE+bd@oMq*#^Fs=tj>dcVO=
zuvJyt3nAgr5)uE=*Fq3+Qy|CE=XN2-L%UM$P188`WN}O+zkRV`NKpOHWVMkv@Im=M
z4sdm8_D+DvCC@88XcnHRgx;n|EoqTG1fR{F&A|_$Qt8Rz9lzhD^!-N!j#~<qu+mWb
z{!&}jU8vLxX5IAQPr3_*KBZL5PWo}Ml;+eSp8prQ3TvsR=viMt58#Rp>bKDpogdKL
z8)I+r$2r&&eT;KRZS80st*xz9gPt8Gf&yXz2@{}@%@dF`DBr>O*_c}v<%oqIbD7l4
zv*2bUXApozv!_#U=rFrvQ`?wVMIdYsBd@k4q3ktCzM@v~vNE_a^0_!$!49d;Gqo(n
z9Ui#{<Kaxi1hm8Tel7}E@6CSaxj4z$y*xf<)+M_%7xG}Mnd><8E{C_Z?Ho-9+>^K3
z`KaQ|DDNO<?#^uje#h%_<GY(!tYLxEP>zApQ9?`#=2P33ZEbrB$f%d9q@Kge{~qmB
zoygzAdoXdj>;CR5d#g*`;M0eXY-f{?`cyvyfx2Pv%q8DFy@!15=LXIQzw}N=*XN(G
z*-@c`Zms1Tt6ZA8C3OvJV|u6*zo#JZR^9Y!agO<%WS$})DRvtWEE6pxC-Ce8ryGSO
zOM7yX<t&|3wXId#8{sNC&#7pd%@~*UJY*9a<gFc%#dbDtdEbYv52NF{+NSmo770W6
z=Yiz`klsC|qVuG0c{<!ZkLIkz>2St5nHIvIl{xPdXVA7Sx?a#miSkhUSp_pI*qP|J
z44Xg<zLyZwFoY0wGq54jlLKbG-rfpsWvGyWagRq<`gA{bK=ZBH05BGo8G|IU>Ux9C
zCb1c!DUF(%to+bCQ_3Y8dxsy~WtmwK6Yj0vZZ^3nv`kg<x9c(5%EQL3Ks0jx2!rmW
z&$aa+gh^)A{-tGADS78G2`Z3s=ijP>dEw1SB&d`i)r`yL0EGa5$^C{g<G+)gh0g4@
z=5wI&3i;n&5qdEkA6DcEfv)zKiP;+;zPvOfwR%6<Oj7{ko5bjuBwpxVanF+fPRm9B
z3O{p~eWQtSZx8E@3rL6~C52Sw<>Ts8A;zqxPD!%kF4(hK(_Tvx3`LyZ*y_kWO05=*
z*(Xq%NMzGR)-@8uaElaKs|vFh2f=lM=#HlP_tN8+32!R31M8t#?2zIPRRqD00-9cx
zUG)M3(&<>sJZEJ)I|)a#SY`H6t%Bjc!Le;Jy{$`i0SR^?N+|5*v8Uun7IOC@??FCg
z5{+C(N>Pc*+F@h)tix#B6{99H9KLg1&<k%gj6Yq87UmC#goaK_E^jLUJ(ba9q#mBi
zth9?4az+zJdPi&$Jd<T%aZwlinqkotJH50M9`G!*K!UDV*i@nTGUa#)I4K4jie7=?
zG$-q7)Fb5SQ<D+rpJ*M$rsJ&A&l~x1Lf`?t)pcavrO8SlaF9N#6E-stVA9P}G*{mf
zt9b(^bJLw<d;Md~)*4uvjdkd*G<#ObC-=nRp9j~LI!lm1+gH2h^0crLxG5T=-nQ(P
zgKEQnW76ZnQn$W^zAbUjuUfx?Wyf3*iT7vBk2lm#{AkGeCr$4AfU?3jApuGrVNXeL
zuH2UEx)tLF%ZXkaYswA)AU671`8%$7GtdBRQT=zdW!DixdBx2=Urynj$*Q9fT{mb%
z<L7m-x5G5H^`=Y-0$fL%v&~|*=bUf17`IEy;2IZBt#_#>l4|7%O(Qj?GDZGm8ulkh
zd1;rrVU?xqW)d&XeyzZ^C>HUTkR!7mXYJ+xd99f#!!;~?Xj5W)*FDox5+femjo?GM
zpr2$Ah296AvLkOCJ!ial9HxeWb<29;bT}Qp^LaIT5Z2zfm(S0G+LCo<R?%H4tWj*&
zUk;;1FbS<Nu56El=%2A?cDhV8B5@CEo`}{v(#E7J6;vv+UI*L+F6NfcTywVjbqCA)
z7#F**2Pr<aFrXq9vv6*Us;@I{urq8a(Yx8luBIpf`DcUQ-xLe3B|=L&QhSLqzbeEe
zews4Tc}qqypl@B3&63p)$i9hS=h77+N0=%_R2MUEZ;Mq;f?0SBD^A1Kl>BRTZw&9d
zXj2P@mSKch-l8hxe(A~7kQYDPt1bGQ%+Gk0)T^_K@-5{=tSE+ty6tL@!e)&B%!7Y^
zIL;5MxUI7>65945uY<N6DixXtb9>AIrNy-Ko#w@K(!;r?JoOb^U)Qr5C7KcCa-$Qf
zIp|8YwrG=+w>BX@{kn`b&hde3a*2oZmqVYUR-rY{qd(EybLQX~!wA%dv|--AvlRXy
zeLT&8jslpY)sTYI*pRT9&1jxD9UaYs@|<=$nux*+3=XODWP~`Rda6p!rrKI7X`$^(
zy&mQ{rDvq1Ir2~D+?aV>>q{~qV4dD>QB1Z8RX8${2Rx;0>go(dL0p|9GF&i=aMSL8
zyiOb@M7ZE9<%609t?dJJk>jT8*Ve2Be3Aze%|>n$A#9w#W6GzyLMfOZ(lBgD6L_26
zn&p+lLq|olq?pF-e;CWtLrwX!3MATF&!9W?*xsyT({4D-$W)b!1EgGR#}FXpaU9y#
z^f1Hw+vdY&ye}L#-}V+V7$iUMSO2v8J)$J(*VWdqh=#LmC15FHXL5~#xp9sxReugn
ze{pe8AF-MJ0`a%0Ej2)bCa3qj*IP)2Am3I-%phW&@ZZxY2NUvbH<d_@GCjUY0EDsD
z2suH-JMvRoA@Y-f6I@!+(MBmzLTAP+M+F-gtyls4W5ANux8Oo<?u9^jFbUxi&hq65
z@I40b6j}Z;$&(5}Dw_gkk;N9Dh(kl93luwiYl7K1k|+Th{r%Mg-)Pye>obuNqrB6v
zmRXAOb<dE7`p!wU=zSA!(#10-U+*c2+Jm3-lB7tjWuSZZYGX#nu*$&ocMgZx&FyGK
z<CwZyc|3<@S%A2}c~bn>lt`rLZ0}rspjf_oE4`#cI_u9r0eN|beAOm?6SAZ~RlCs{
zkMo}|rUZz$a39T(9#Qd}S1N@eqp_QF7u4so`U!MQY?Ea|Wfm2zLoK_zkT{bh?P3Uy
z7@`)NF7(^Tx?1;hzGbO;`ur(}Xi9MDnDD_LQ3Q$CrnVZyl=WAI8qVQ^HH@6Iuilm&
z0ic(P^V>_X0DU35v$Q1m4T07=n$6Wyh@4TT);VbWa!|?4hp-IwO)dye^=t630Ke_;
zeZQZjx8=+G?GID}dUvzTI9V}c3&JTH8ATCl3O*AN)qfMWNr!`+yPekxuxC2>oU?Kp
zv~2j^JKb>vFR85Zm8{ty|HcNSM>VD5<Yb;zTS^(?k*f?VHDwTxs~|Hcsd3z#Vm$>r
z58rP%IUi?39ZIjf{;bd*3>e89kAfyXpb@3k{$=wK8h|!d+^IbLM(8uaV5fZ<$dhFh
zX-~O_mup$9roKdC{S>s)->GF>L^?pyz5Da@V8ZUmaw_XI`A+IZ6`Oeph9@y6G5lW+
zgcXGmVx+M6)r@cRBw6XA8YX3+PVPBGbR$M6n|f^VX?q>*K&LbUVncMI6K~Y%)L#TE
z<rA9aNng{8etF8xUN*<gH@0Ensrc=%f@5@_uN>;|!45{gF`4n$g*B)0`NGyv-1Ax`
zcUCCfZV7r&XZzz(BXM2K8lY}mcVjB>rIxL~!>?$wrP<5l*;q(Wdd{qAM6?sGiFY!X
zCd<=B18W;~FcuZ<wVH9zE9Eky6qTG-pcc$FMJtRe45lwKTIu+a<ySN(EWfZP)4fqW
zR&DA1p~z<WfpSsZ9*2LhKk1l>X*r%dUT?N{RtCb;!aDD%I=)fTIXe#pisxNyw0&7F
zm>Myq?det7O2aAgk*74zYSb?q(USW!G1999XG&3FpZetS&ESb-cw)A#C8T6sA~F2R
z#~=g<O9#=}k0fgA45%MgGP%mzI}ypPNAde|eQDrI4)}&IxE*babM+;W@zUF3J2o%h
zf`O6vM~exaNkinlB+crofL59W%5>rqdjaZY6Xi&^=K5fa2skAvz>rHo_<UOpWyjh+
z_01Y+$wNEE4~vS34ou$dN~?;c`DszcE(R)4?-|mX;jmH{>$yrcd?Tb^-vl@9C+%CU
zVOj8{7L;CI;p0_z*xb>dWk1idQ0HgDuWSBJ7rC<tI`Z22xm(MG$C@<_3Z<a)^szpd
zpSpL}=SS|&IcE+vNmVm9LD?kZ51h@$=$01v>cwW~{_-ILZ=*dDk5yHn*t3LNe>dEf
z@@st-8{BxR4M`J<XFNN6p?z+6VM=s(S?gZ>OmY$P*AqS1M$%Yi{;lVz5^%y2<7b!}
zC$e*4Iuz8o*Wsl;`&Zfp-#FybRF!gGhNHsJk7E!SZX1+l>@aWE{IN&iwqENYE6qa}
zMkfD!;RGhF);UNP`<K0c-TUws<MHJ6uxMk{#MKIv1#z!CAs3gO+1xg=+KbBWXya!R
ziG3)_tEmi~nlJWD@1_o0T5a@^Ec7-Fg`3F*gUS`@l$jc*yZrw>Dvq6*U%0(AFKZ^k
zh(Fm~$QyAoFkI+Dxr=F?w2WZ{p>c*k=EL^XH8GTb_9wc7coe<TcLvKvFD7CW+3+UN
z1IgEVh92eCnoY-T>ivbqixrhG%>NN=5~WUp-|GK~o}EQkU0;_P>QqZ7iBovf>T9@2
zHR{0M*}Nzrw`o2?J6Lq`)ukRCyuw*|V)-WhHiiXS_ixr1vQqePU(TKfGT)^S>Jc!t
z1*&;v?Twe#hNal(Hh7q;xvMS=1^V-Ou3$N>TcvjOTJNSV1sL&01riPkJAG9U4t1Os
zd(U~Nf&v7+`ms$KUyPim3TK7v!JHG+U+1A@e1(0bIP20C$x)j3f{WXlYY3iqPO_=z
z=5gW1&IobFGqkQP`%AHDIaUz^Wd|{*b%kH`Q7buGF$`%nZh*PCe<)~Eia!-e>qJUW
z2qgy0M@98&vltUZ)gxt!RAY>-QCY`k5<7)>iR4C=F;!X1MwE#Q#x9&*T~pZviX~vu
z+^&iL>5j9i^{NW9vcC2R^tQ$v?Jv~KFoPiiu4Mh9_q&^Rm5&ij=6fXv<8UU`*arZB
z>GzBqdnkJNtgCo9g!MYDZBj1eDpUKOqov!Ke4~A-((uYR?RrJac)YZ)Y1NvF510J4
z6h|R%sLg`Tm5+FUAwh@M0!-z$B@*-eD6<ixoSJ)t0};`t8pCE!jod_VU#JYz?kHLI
z<|B~J!q?8ND<_$E5WRX_2;JcUq&iI&e<msXH1Kutx~_b~z?9ixMO}xX_aN4tI^1(A
zZ_p{{>*-``xU~`l^mcn>EBW$3k65bkih-8gIvZ_8Bt4m97WeJc>b;37G{xiJ;bCvr
zIDyBkR8Us-TrqG$!1Lg=1k0AH2jo*O#@#gF*>_P1wreGN^|<W^iNSag9WifNl#<hn
zYGygHw5_y?rnJp+Q*VXuQKvSUYfEQM`fPMuwGe(_c=G%Lv{VwkLY(zD(tX0Dgr%0N
z=uFM12%4|x?#N#f?T2;TkW6ZJ63u5`@$iR^hdpzWE)xG;o|Y@Mx3^BmzCZXK$8DVi
z81`4t1I|&D4Dle=@bebPEkh$&Lxp%%VJHJ+kJ)(=7ex<(`&aB#8HoS%JM#X8C5hnW
zv&bvW&e=tx<qmP03S9%sLQUlOV7EPnuT9J)FUvH)d0Q10&8~s%ACN*F776d#4NN4L
zn&$s%C`u@x7ZZqlBLq1<V!ZMz^-Wbp;7{9kE2XBvtfOK)n25HL4P2GQ#PhtmX_kyC
zURHo|Cu!~+X{Se6+uJOp-dsqPrF}(Zr3^w_P}jLFNSq7g>hbnoA&0epf`^<hXMeZC
zQl*^e4x=(9Sw;%n#pKes#pk{|wz?;$qx9Mxf;EO0Ca;K?|JX`bwKsfF6Im|nSsR_R
z4G$NaqwtCPt($6BLZWt$wd`24n{>|qO>Jr9n6qxYBX$^e^a)Hh^0E(<X~te%5_uEX
zC$u`GAJy#ZX=BJ@sx!)6{E$U~Qx{i0nN<9EG+<^IcFtiSAO1)il2v#CcU>;F%KgqH
zb!uU;sT>n>T$$VE8{Ty_<L8t4eaav}JIqA4lvWP}GA#we9%{@dth@B39A`}ltpw5S
z;jzOC@?y<C$1B;BW@Sd+b^Ss9h#(B#i<_QC(lf{Ih7eg50n3USW7&!(-@T==uuk=d
z!d6VPsb)BKD1Mn`HLCv9dPfb4wTByv!Hl+z*p0H&84HwbO@`xFKyRHhQ&&s>jv>tl
z*7Y4j99<sbsMySu6Jd_7q_l)T8fYAf8TT!15q9g1UTcePJXz`78*7J{5p1bUXa&}d
zUsioitHsZQ0;6x^b+j;C$ybP*gdsPqA!g7T_m<lWhGs3swV>O4R&s5Ic~{{GtVtoy
zQ`%VgE2JuPh{YCL@vXo<HUbFc3Vl5zCJDZfhz>8KiSSGu<maG?{{Fg6x!S>vL!XQr
z2@I!UOB5ATW~~p=ySR?DnICjk2VxDiErXT6E^m&Xoa6EpyNJ@A-u7nQk|vC|BvzNz
zp8+&Yg%d_+8c9)!4DY0)onh+owZap3z;1X!fEF7kONAP`e<l6;)1v<Z|AcyZ%FG^S
zxNI+Bd7&{+yu4F-3^HA8%sSB@{sc@zb4Hq4+^#V4X}jrbW$3w0N>lQ>6R5oG#?EFK
zb_g-f^E;3l(91~beBPg3q<nPu*ie^!R_r11ppCHgt>eL}JJy|Tr_>z>uMY5vXc9hI
z!a-+&!D?-y{W(mc@`x`DLXk7@j#+WCC3gJTwL$2j-(<<epm>a?jlB|f5av)YN{~Sn
z^t71baKTO0;;U}-(h7AO#powFfNh%@V%Pd#7O^ttEJm12(siSs0j#8&RZjO(mjiv^
zo^KSqPt{%@XsZ8BIsg4$k!4ZSvh?0JzGrRSv@(0w)N_THmMIvGdHJf3?YTj5t)1<P
z<bxc6heJjS=d>&3>dnx4Jd&hm`23$zoslIT^KWviy;|v|1(e?^$NVAV{4&<6x6eQS
z>>b%J<&@H~1jdLj*ctX*4u#|Uo%2o%@@_ERYf^0fC4}``(HjG7OG;wXQj#L4+rOf<
zM`FToN}9-sYWg@mEb7AsXy&f=zej?nYF}1wBwHw7E5+WX2PH(huZTwFm*=X(0P`hz
z<d@s=)~V_KJaPM6N-GfytjtVBSv(&dBDBWjAL;8rdBv4W#(&Dw{oC%5zyh@VWl+T!
zkMKbPK|96|jq#{YN~4r<)k@TnU#p{Y^}NhB0V1Pxl=RAllku4e;Dn5_gRy8Lh|YCi
zbhX<C+)3g$Y)U4T-qwor@1G_DE}TP3KutD!D$R$8Yb=iTWs<){e21zQhM6lh3|E?4
zm->A@|L-iQZ23AE-zGqr+`u5Ll4-u^wQ9Fzqp4Q<_i9r{`j(kFWlL#_T<*ezAuqVz
zDf&g;($>HO$vFl|S_%+s**9Vv;LInj9UBEl60lhVL5kjG>PkwGDNvDq<FQBG8@RYM
zCH2yRB@i}OzU<`4JywZL^j6xI(^`hbsh3I5;<GKMP^@Yg>8So?xzS6-`ss3RI9yUv
zvUCW4T16%{1e_R^{gBOlpY@HcRs>K6fz+_+?V>A?$iufRV!~>6)JnNv+bK&0OU#v*
zwDVLzo;nDj^1%w*lm5VjtibM0DYfd$S7(-CQCycB9v^pnUS6n+5(MG-^9Zv_;BVZ`
z7Ic?#6|-UvU%AcuX#kRbJG|#Xq#eajTRnGR!tb@zAD_{T77>#A=QutOD?+nhOY#+0
zzprVdTwihqTaSfkJ>Of|eLYVSvHnK>{_6a9YD$`FMEo3nUy@)3g;5z$>5gE{`67?W
zX=W`cT32c#HpWyU<+tsf>h7k55>?5ke>J2NS20jGw+C7djrf>adLp}Mz_ny8>qLYv
zI#V>%a5RvYtuEau^!?e@^zjK(8n118DDjh@d1dllLk4jw(HS-BoFeC^>R%lEmEs~_
z4^P;!wXvnZ<A|*1i1q{Wy)k<av^xlj#L~IVEf&Y5D7;YxCtq>1(~O!OR?v;<Og?$x
zrUcC8Qguqs<9}*5E!Tnv7*K@&h#$?Wdm#iV^#Wnp#N~{l14k(!w3zA0j<X;09Q^~z
zU*q)I-W*x5ruLRigXufmPu#`tZ53)dlLL4dk8{%r15k@_qBi%BBpQ42p^rtWwkW!S
zx}$YIZZy4acF5W=*Tr+<(P3d7N5rXH-LXzp1t$pFk|r0b4C%VPwU`?O1=<e&GDy;f
zN6dIvd(oFf!~y;JPOwh8vyx28r}O$>kF?n0z?n&z?L%=h@4okFr^d#w=~!eT8^scL
zJWWk@ORX3_A-9a38|<3Km<dLPwYXK^^u(E+@Q>~{BqF=Z)6m%2TT<s*+rb;DT{*4k
zm|l~TsFLX5`+FI{90?2LDUU9$AGg6zo86Ls6(GSsa20&6n8m_k6LMRbIKA9cN(E|J
z<N#KL4r|>>dIwP7rvZi_Ks~In+^y_G;^Rusz3(`oqdvp~y|$Pw<$Jhpy+e0=_I;{#
zkPvOlF^?$#bvoYfU2tz!aO5(d+jcA_YoQy|<-g>siY_kwj2N*FaZ;@GBy@}96dnCh
zm3g`<nM|uurtOxnYf(R9HZ*8(KPf=EjqNXsLPn)72K1GF`*d3odD!(PH8sG|>&-}p
zn8<94F-3(z#QC}TpHxs`d9|wx9xtL&G|SjH<}|e3Cx`&d6rT)R!aej17k|`NM8Ip-
zJZD^jF8(f9O$gW`*DmH7C9X41wG}&?llbM<P8_%CQ%%PHGRzI}?6=9CAN*?~fr+n>
zQ!WmHW=4>~e9}A;r_==)(Ozq*cVXHN*=aEehi25~(r|vTG1$*jKCnCYu9za^cajuz
zRr1iJx$;mBx4R1;BkU&WEtcnl9JG{A8E403mm3$}KllH?)ad_(xi~Qy)9TjWEpW(r
zm+mF}&w|gljFQPOvf4dnU$5%P*bdCauBTtUc|;^7E36R@J!)saRLs~Rp3N}93MT|Y
zwx+*&qHR<XNS++o%1JZ0FQou@w2>O>YPaZJJM*9pW+Wf&zYRM8HX%IG&PQVAhj!?~
z>4&xHG$^(zv_ZM@)BLR8GO&r1aNH~ZP<6hYH#v{ANM)JQ>v!qV#DFq%8l0OFI@l0J
z-~$?c-h5rk>-0|~Q^8p}=mTSKc4&fpq_>d|kiGN)j<WTE2ff#>fC94Da<Db@r)Z*d
zV4rf>#ZUNE!i9M{9CTx(I*3ur@=u)}A{CoBmO7e^f6M+{%4V1)-`K`!9tB&(5*wk*
zA7?OdV2UPpCb8t)PLu#(o<56$#FT?`jc&A>Hb1ZiX?ErL`PTkC*31ALfw&*ay80r#
zcXi3<Ya31PBFmlSpc5PJsM&tq%KDHWrwUzT>Tt+jA7ulZh3)avI*h~*u-AMCIXqS2
zd&s0-!E~p+1rz35Xv}*Z+I%{}bBO8lE{`(iu80Oi&2zw6EKzXr=O&)%T%gz)E_d?_
zsD`48uugt_9)|>E%>HBWhZ(ne!9O<x9M23c=REbw0%@C%!io4~4O&W0xIC}RGr}$y
z)<F4YI2&m%7orDblrzS8uoRl#0{Oz98E(P0F5GbU`pK!O>#v-d)f4CDA(%(}j#e(e
z=#JM95@Q$f2i^aQqZSD*G_;^sRYzm&K{-W}@L5_cz*jn$1Vm$Hii68)%^n0<#EKgn
zj6Niy2SobLrE18R=@gleOWX8=z7|)~3*?Qn59cVAjcwH9J-yNyi!7_#;p-6-ju$_l
zQB5TLXXYkp&hAl7{||~=uitB?7g<*>mSVr%3?+;w4XK<}>nHPdccz2X^5@95;$ebJ
zHld197^9Fb!vyATI^lOr@k0~C8=1~D|2?t`c71#HrQ`MUUGrKHudj^cdB~rLG3sbC
zwi&%>n$&(J;dW{=i@JUl{sQzYAQ=uOeI)kka3!L(I74z+VWLuYh?L2YbmUM*2Um5<
z7vRbN>-k_HUe0w;UH_d*AIUue@fi=$%tI(pVlg$l+Y4Q7hf^Y=r(B*|hPF2H$%YmE
zG9xjf4$^}^+fa$pQ*8Qlzxr{(M%@PCw)S4N5KKAys8_3_?E#wj`bM{g!rIpTa3Iyn
zInxlh710n9oeHAY7fjx3l3%aiFYU)L3lX=3O>Un)=aF>Osjl!rAoW(uaXYSWNmahC
zdS^Uo_Dem4hznRR4r8jHi)JXEFnvkYob-KP<$RO#%4MD;Z**HfNDXdK-kvAM=PSD6
zEWgI_#jr&3Ha3jjcZAVhBURlaknroS<8*mV1~=ko{(3)4(|Z(6MRR#S;21NHc95Ke
zn`mPQ{x(~H^7XbUd<}G++A9^@GfzgY*&nFrLdE4_95sby2tmQQrS|_G;ox>{*E3cZ
zx7_WopF0FTZ(v)*+j~^HX*_mh_2eVs453^8@>)S<xGJ_JpZ`RD1n=}SNf*Y&d%VnT
zRw$d^K6tCeajXBHbUA1z{3`Gdl-nIqvlpN+T%qS2;q!Ix+`0n6a^Jf0OD1fmk&VMf
zW4=_vrM-lrUMcTa7~WKX+$~XAZCmK2Rn*mLH?L;L+|W^NdGrwKyQCsVQn;^|vuZ=B
z5ciPLr260kQDY;JC2hVY>dbo)9<GwaMa(a1)1m0n^Ta3)UzwXmtCxQ?P0F=*onGe<
zL(HpzU6R4E*FN6Dy5h(Fy<k^3Y2>Dex`UlI@H<A%E~w#<Mm%nflX7Fl7%hKMZ!6iU
zL%K#^X1vkePxaDbg^s-KR=M)Y*+p}$+aCtvTRZ9S`oEihS|%y`2T2ts-(T+Rk;m6T
z3+8+?Eqp>CuO*t%M#-8!KCi9p*R~V?k*HXfc#H3C4Zfzhe+8w(QH99b1%E|Cc<qa}
zx_%q;nMZ;lOD#)**TI%GA~Ed5Mx{Q%FD=ceD*Qx`e@eV1c1`!1nh<MNa+^)P4b=5o
zpVdh`4YR`50Y+(Tr>Tz){myx;=bWCErd_5j#WqZarPY6xAWGqYJ6PCQ5nI$6x1)wK
z^c+x;+g>YH3lyqa<1R+z*ES$$Y28oEMY&3rx`ZlO480%i`e_4SnI_YC{n_cJJ!+S^
zDsR`|Zol?`IQCsN?p$&K9t@DH+L7u%ZgCmE;aO7<wLtDbsDTY`bpXyNLL_<8^ZtvB
z{P}Yy@fd_`^UA%hcqz_p2T-L?{b`PRci?UA%D%pf-(-I{kxQTEp8C&uZ<|Kz_Ev-9
ziNARmS`>b&1H`IGju~tob)d}rD3Gi@vYop>a%SyEwI$!2Ki4`HA<82HKKi?Cn3o${
zZXe6zIrkU&yJ(R_*tnx*R%Qd6xBOW?=wC&vy<CNkfwQ$dj6odnafhQmo|_NJmN27D
z3Y2TwM&?!d51@u9cA1LXQ*@ajl23@085f&9s0ABKJ2W+Vbcg3k*E-X_1KMhHHCm9L
zsdb8W$M~3H7z|Gz&>8207>x&rIdoeGa%eTarPrWStPx~F8`~bs^z`#v{M%8>WLfut
zA<OEDY&iG{wHuB~NR5E$V_s{T@3)Y7Dr@T&^7=PJh3Q!Qt``FcLpqz^uSw>}>@ZjF
z-R15(2p>S&9i7RK_Aa0No}XHyJVNi&fmS~c>MUSOyd2$IG~$iAOGX{z*>t(%nOK&c
z`oD2q?!%4f!H?d)yxmyyzv?w}6guFZj-gq1DSOqC5!UCeJD6c#`C+;Ds9$0KdkS+Q
zcZ$Ce*&BUlnoiNag+gd97Hz0ASn$vF@~+IR-?tmpW5w?j*A#mSA%YF2;n1Ks1h0_E
zW|`@E{gsQ(42Hvu_=B`I9S<vM!B`fyoM2wBukf;7b<XQ*&FzC@^5r|5#Rp4H0AQxl
z3|`-<ROkE>nk3)mVM@iWLjm}U*-UO2(kO-Oi1|l9;7S~-Sn5_IJ=j|(e0R|p1~)F%
zv?s?|vqVd*c3)R#A+dg+d%R&nx`f9^0F+|2;MZSiV4gI`Pcz;`Vo}@&Y89mB$v0mI
zxl>tiVDqMykt}D$CQGp61X6bE)f#vP_J%39wM(|c_OE!8j39|HVzh9YJkYSq$W-Sv
zPqBT+3}xSu5L(O4-_@QG)}~f6Peo{BLRCn>A`n9Rvm!~s`-|HMU9@6xudSWN^Oc1x
z4cbbZ6^WVbUvsI+ceItE24_)d#Su16#W!_4Q?<l!e(C!uC4v;LS4h3y7}~TfEjXpM
zS3~<=`McnuIZ==ImK&_hjRph!OjFBKTbNd2Lnr;dx1jJ6gD=kD<vHpYm|nhUAqTtG
zbW15kA8(P$V6iM1u*|n9Y}w;U7<{RAl~L`!Q2pN{F~N$cReLI}k9;LFzU|zkr!IQ@
z84pSz(N1?Z>K&sHaNg1_&!4qYb!bdyxQfmz$1Dy)heyGm%t$H-<_)&cBFs}~Ezh=8
zYTPQ4Vjx$l&Xd4`xdv7O?>EC)a_3|AQ3K)SzRA8W^7b&GB}$_>F=nd_T_sL*__wOB
z_6egq>yKgz7^WhVwKGsyV}jk$Ico5m-v|D}&cr-c=g_`yk$g?hc)<Kd789CzIu^3_
zPY&dt%FEB<*VneET%AQ69wapi;^_DM>BFa$qRwjWI8c^mUIfN?8(LS!V$_uCkZ&B3
zghlHGw9uA|$e$89oq->!tCR_aNN9w_TPShuWjr+%feNiIQpE*Ee5Ll_B7M{(r%Eeb
zHjm~v<_W=3t}SAALg7&C@~fY3pEs*CGj+%+wYgqtdIDUd$F`c;gs|UF1?DGvCyUb;
zT+c*LT{Lwhz=tn2$qqdq1=rm;3%2G)2NM^`8z+~zuTQ+8G5$rU<=mhtdo%d{&r3MI
zkU|TWrm&hwH0699@?gl#K|EhUDN>8hx?63UDJ^WEx%s2Fv&d3vS)Hx;ePdNSz7X~%
zfP}}VJ_n3(&}ZOSYT)ZzHIa_f%<>xJnZQdi>G)Pxe)S?pJ!$3Db`5K%o?k^*au40!
ze~$(<chB$BP1f$ke>2^;|0Mp}ei#tQfbWa&|2TeQ#h=c|ZLzjwR>glqaM0acJ|n@-
ze}d*tw1+>#PD&rPtI^dPrEruJo61}1xq1TT(}ppvwqvJHd`u#2@`-AEl4YOwLrKiV
z%GCeda@CB3st@|V7P?Mx<A5zXD7}42F;xm<=rjI>647Nxj(IN62wuhDmu{(F-9W|A
z=gy!17j`xOueYH03VsOr32Awxdgif)+v`#KTkEoq{?e_uE5aB2u5YW8)cNEgeC&S?
zf?nx@8?Vy~A3{{w>XIve;O}=MJj;nJ@<vzvy?>8?Sg%3U|BeP+1%3f%hb2u6#9>y}
z_BKcXn-3$wVv?}M=6_ZkZEaLb5RS!>syOY_LWdpF)S-Ju^^kW!!L#Cb5A5f>kj*{f
z4<4WY@BRM2pZ+Oq3ngrs5A2I{7$)3*u=brHs#uEbo8`0oXg9h%u2Se*aXIFIja`$^
z?Kr-sTX~wpS#S6obhFw$i)O{31BnDoXHr<0{=MYa617kHT|SCS&3`ntUJsueG08nw
zxodWtSf?KKk|5HxWl`_$N-=lK4CIumFw)CpPeXjI00Kb+GCK518Yzu@Bm2cw!H<(W
zji4pVv_G1q5P?*x7gr#lrCbhTuv68~w<IbmYV%D+fP1gEafrLn0VRJXe=Tvo!$fu=
zj@u)x5SWSk<IdyCt{#8<s`6W_CcCP_j}&a};y8FD&W4C+DMoOMiz}^FwVqH8Sgc&G
zJ)XItXnpc;W7S$=o?IKMI?#@#3&zcpmR_2O%$h&%9}6;nXuzm8%a(mH?MN{YHKf5+
zS2M75SC87-TGU^84}On|pKtqvDz`DFZYnX=C(_Bg^UgIvs+Y-M_4yZ^UU7=`;`yJp
z3ThR0piGHm!V>vJS=Qwj$E#w=3UyvD^##-Asg3-rUXaOc!fgU?tTCi5+iiOO@MA~!
znv$BV8J8a*jzAo*JH+t2CvA7c-b_+g2l!|yWu+PHQ1q8oDCdtF+uZkifyG%^Xx_|V
zT=c5R$OY)3E!-(Gv0kAHaV~ShC=v4#DXpYWZ*HAdkP7fons*jmJ)?(lzJto8B=C?(
z5MRFiI8ahXA?<OJMmza{V+5xVh&?G^O(EXtlt7q8UEJn%eKbQ{c5DxFI%=Jx2+3j{
z?sk{7IjI4OcAQn{IK%IoSff=VqD-Rn7b${*ZNa7#BTYbPJLGGp&2$W;WzyC00xl~#
zX!Ki_yaej($H?hEPaz5Iq;z0QE+HOXJSr_`H<7kKdbK?x!Ux!2PU-QanvPQ>cNTZ3
zKma>Nf({5rtU0O~VA_U8Nh8(}xkcc<#u)iPe{Oq>BX_$#i?a;F9$A4}Y@Dc<gq*Vf
z@3Kh_<_A-ZxN<<9YQ}l>Y|gpL2d~V*p1q7$xKMq~LH@ioZ%A;}ovmUwdtctRu56Gy
z5vbO8HEFf3WU^anzaTDACHz|9%lE_Yg)Ec|@IR+7^bCaelEXD@SQO_G|In4hkG=fK
z&Ew_l<&MI^8udli)E%cIPh64mjZiP(5v52m!-w>}DVB<wCR3KehSi8nzzofkUxd!h
zLX?r>lg--)&<$gvsymKcE~7#B)a}S{>a=xiueVImDDxVXiZcpGlK=~0m#g-C9x7@Y
z;UL-eR3{f)GL2;$AjlGiP_PxEg!c?~Be{rN_YaFcp2r%z!IqN3Z7NEF9M7r8&dixc
zj~IzN$#`_iEzo_#DtfHk((ZB`+yza!-NDnbsj*eNXLm*wwUu>OXx9?(xJLc{gt%(r
z0??C-m@Qi0Fm<c|v?#nBL^hBe-mPHNkeWXo{L8o1lEIGVqkUz-8pqa%5Mc&IU)8?@
zy&zS?80fC6qX6T6eMQxD6PSeJR^ib4IrxxK_8U?B`Wm-~BrIslI?&HyS}Q$SS>U|s
z#qq<)Urb3QR-Juf3t%=Wd);t+dDq~+&7_%A?`=qxJ*1B{=EJQA{t2&cHTLy!ojH_`
z`Hq^}O%J)<gKCGL@fHG)7+Yrp;%1hd3;Q0M&5lii{<h{Yg_wnxK1`_ZC+oei!H8LP
zbjXz_+tJmGZ;reB`$7V_0Bv(l2EQAxW9Cz{-U`GV&c%!_phYW`9mh6<gRb>V|NXsF
z=fKF5US^p@f!Xn%2+$fS>M;Uftj5+9%T2tz5rzrnwW)jDnW5<t{z{_ua_Tu?Ipj8Y
zfevqJs?nH>rnX)+tVrc@FObI2hug^%)bXQS+4L>se0Ptj)|s2kFv@Y1I+`Iix%iiV
zk@Ghcz}AAU9&EP$yQSO{#<J?zp2n-!SW6LmSs9H<?%o!2Y&~zVQRY%J&R!qM(_8(8
zDH+vS%86Dg1~0eUkdCOnm)4fd+cGhVLrann_b&1ByvtWnwC3p7Wm{m_1=U*(c6}ih
zo#+56H7x){mG287wcr?~NhP_LKZFQ}m^<pzf6Ob(S?O^g{-IxUf1e8faVdP5o44A9
zrXz)r0uDS<bheQ`a4V32nXg375B55~>w<NsdHW1#)j_m#ajG7&)}8OV?wU{aGW0r9
zD{nGaDTUGEbv6gqciUX)xW|{OE|som>eNOTr3??fN_lpiRHGZ!v6qG${PRud8Ply=
z&q$sW1riyt>R9e6i+g{zHESEs#&wvlbU`p66{*^24KZ-_SMQ7wpn|w7E5s0?$|7Q+
zj$-a6H34xAQuZ>K{ug}PYi+6(M$?V9qm?^hUp{=Fh%(tu=8?Y6$Da1URO_zi-N%Ym
z4a-`|%9&vkXN=fC^;iErI+E-3aQDy^uBU&io0i`<^gC)@<4<Z@$(U_`B-pjA12jA!
zB<A7B!p%~ooVg5pn$%M8=@!4|6^Avdt_MOBX;3vfVyT>xSgO7w7oC)e;`(RDvj#<K
z;!P$x(=jtzJ5J9H4Pxc~P*G+T11gx{mne6HMH3KlGX%@jaP0-;8gso;Gl}St$b!*Z
z^VLaCUEh<E7;kO>X|2aW^wu}k(Q8%nQfbFyj=Cipa^pfa6s2@JgslapD%t@SJ6!<@
zeQ`?nWCPW<;-lWtJ3~p~p}D~FuGCNw(Hejtm{}oB-&;SL_HgtFK$lq-Higo`nNPy=
zZ8wOI2kV&Z3V&y1KB@X?NVU%}=JGOIfqCp-@D5*5r=N0=!G*DC*KD|$O8)7jnp1&9
zD@JZOdF)zZ;>!(ICZo%raa=Yvy4o3dGSOEzktH-3=v%!;zujbfWvcVuN^1Cfr0BXr
zR1|vBU}Qdh5UAZZ_KfG{mIQFTVTZV<J*8yax@cR_9FI_vjIs#z7hJ3*4q&z9{^g`x
zvstkxLmh#Jv~`|qSr5OU?e84U-VLh91&=v6Ez}Ag4|pf;nMsZD!r6eOCT|Lui34g4
zQBaG4!cmse`da$j@^J+mtKJn=omM<gb#>=kr8MchSv~#MjK_^SPJh(7wI(6ak3npG
zD5x3qX`O)wnH#tVnE=MXgdl<H?<E1T1DrAiKeXAeqQSMJ4#MEq=eyP^66W>kPAyq;
z8p?6RY;Oo_^1K5m%!++JQAu7xGIp&U`6a(3&T3*T7n7rI$YC)q6A}MCifU3T>?;!^
zX#PH$FQRT+s2<LqR+Fo0qcOBe^Y_k>O_q*s6V{S9!1<Cvr!ECAU#ECLO)(uFUn*Cf
zxfN0R&uF2m2!OZU0^~ulMNt6>L*4`@P5WxACKJ8J{r$dCrDQ1O&}ax(Zw-DQYgt1s
z7E!gr`LXL^ifXiGi%RgSH;PWDE{;jwArqK$6l@n5ElJq%1%6u4yTQ)?oF^<QBwK8m
zu~<kEFA<(iRM+UThg_R%ewx~%jg?hV^U!~l&+s|!$9BW`%uDhWgCvrqj??WxnC$Lp
z($rHQtsba@Ak>|1seDN(vv}ye!F8&sZ}n&@P>zHOfBZ0~u7dv0r+<ZEU-V*Ulk1Xa
zdF`E2GK(g%YlfBc3P9>w{YtwOxI3b-SO^q}5FQkF7kA&i>B1cz%~$?v5HgI0g`1XF
zj8vE4M{5<C#PlSoID667F#K@OvE|NoE7=81tIdd^mR@Zy#)_I2X$mcMPv}wtzlHzQ
z>P$@a(m@HLXf2KCB{cHheO-+y#UIVZnDI=~?>TaA8K1wTT8W|BY(>vdS@a0;2o?N0
z2vQ|*dOGn%@bbsEa#|K1(xKfvzO;yv2_VGOk+aWZ+{va`hK?U;c<?8X?@Kra(xk09
z={lfEnW5BR`5Wyfx~slE(?F)8D8x-l?;O>%D_QVnaO(Adg9EE*&c1!C54HWU=7^__
zF$$Pn+?^4vT>8%7$f{S`(k1z-&uJCovLR15Be(_4%PX9tWRnnc2@Z~TxHM+me2~i*
z=vt_2_V@B>DOki?{7Q4;GF<!emPcqh-;lH-tTznFEwB^TUH@TnKS^-GPEnq{-ON9M
zl&TmL_TZq&7c$s4Cc{-mjw@+g4~O~uZE{Vx2&~ZAqJqL|HXKd#rt4AuD_y%{KXF`S
zvs|xWo>7IvkLfP~GxEe;DKwpR+l2Q3a{og{p*Sne9{U}7?H&!K<m{wBX9$F91OG$}
zA0r_uhnSbs&aOHnbY{kz`DR*y{>A0r=lvRgoctWhfa)n}pU;ssqSPlu;Q+e1mI`HV
zZkAm|wi|0fAA(OJB9xH)?`Xa!BzxQ0#3PR5CFiGBk$9$Nrfc51i<w14j%=-~l3=T{
zF>5vTH%p~3xscsL!wv%!Fl9TN8KpVJQBW-h(lN92=5(MmGh<-P5Std80`88<+9#w>
z-c!lV!3C@X-i?(^yV8Sg8~}jRY27+}!;+DYV#dUxe<tk;Z|uZ(uA3yC!cd-Ap=N$t
zCP}L7cj5sv({KHki4Dt9iS>c9r*u$;T08KXmLAF_MkC`xmeAH}$sqHy4-qqbQ}Yt0
zzL$(!u|Ms%V+t1pagL|b;sV*>d>pueiM$C|?W>>d;UZ#`gm~3eIMbnsO*c{+qMjH>
z3U7m5Goh@}))=n3s1{u!l~&pQ3_K8dmi&_yc{Xa^ND9a>5XL6nLSGSFQ@)$oVntLL
zn7J3*zb|QsPJMZ&?ZW#OCaUIn9n+*xsK18_D^`N*)Kzec=|nOld6V{sovK!)!W^dP
zk{CLeR6?l^a8-+M&)qwxy={XVW-h((3MOpTH3vb-Sqa~S<BJPf+Ytdt#yZ<^HBxa<
zISLBiwm}f`-*O3y^og=8t{NA(yE<!rzxtnygbYuLz=8?Mcq1GOYnzVvb%muIGS#H+
zxIjIqc<oo<gQD*nzaD3Qr2Vbx%h-e8%MSbrwF|6rDh-U`zk;4XtU<z}V$G5R)&>D0
z)*#loV1FchBbZZp-DcsjtUdp|q>!b8_o}|4e>FtFkp%_OyWYo}aXZo%qGk%CAQtUE
zpy17bm1ONnY?1MRGf_XA7K)TpM@dh%UeqtYC830vzWaTdUQgsqivZ@Ne=fa4eE(dQ
zs`uX`4p9{sGu2=ByJBj$7i3G@GKq%|R>3!9_}T6YH9&;+rOshqab(WHi-+L3RMVu?
zTl8#PqjC+u_|An~uU2T8Z@o%tq|PR^g?&AJUhSvue1g1D9tui%)aj-wVnORAF=Iv2
zH;FVrQE3Y$ksU|#aq$5Jb|Ue_-jA-^uGE`)sU*-5(~E{2v3iYCh$Zit*FCcp6Ha;8
zUgP4~TsPJg$4D<WVXXqt3@0!n$alV`-H@8ipd18FJF*ib{uWP3rvvXNB+%+uEC1?O
z82_@<@uP0!iivpv_+wRTXa^N|UO3*p8<l~{0{32+$<1)gmpd|K|7ubs^U(d#AOJsG
z6x(k1l5T>wVABV-Aih(fji^brorx*7hn#X+`WIuXg<thSn8>hU&?(E|yy{!vF9TPA
zWr;hWrmsgUC*43W<JtJh5a?a1%8(^T?Q(|)b;az|j*Roc&4dGOt4PJtt{A1it*k_+
zs+3pMg_0_ll0Goalg+(6;fG_2KpTXLhq)@)Mg)>Hn7~~pd=_Ttmv6bx(Wq1eJ!m{5
z*8xlO$vfIDpP(lDG7VkZ49WmnCjOr8&JjoC<FRZbb;PBcH=uNt^B)MLWqJ{35K97q
zt!Uj>dA2{l4O3@^;7KW;&#P<3O3JNODw<N+r{_>6-T{E+dHh@z067genVsEoe~su+
z{_C>fzUxv(7fT+<el!Dk$ch$?_EyqstFUEyUNczN#G2B)zJ~1x4L7K^(&Z6TJ0n0m
zrjWNfm@O!F$&g<Q*K|$;uyubAM~bGS_?WfOvy`Bs)lYZLG)7&s1Y@hIF+zp!&uA$?
z-ddUomYtCRb7(=CjeI^cKqiYc#$ec5vTAa+t<(q&&ApQB3J(=U<u^9Fg;^uGph=18
z>8=(;@%ayRjYm;|9rGVh^>+O9g`z(PT2s$h^3a&AfG;b(al!MFi%eh@!Gj7RTK?Hz
zsf`vsj(dF$FU}9jt5bb$W%>T6ErMkheFp*l4Y0``437$Ut0CnnQ()g;U%jIi`RCLV
z>oymVw%B9)vE~LYGj!bI){c*yZZdR_)BF~OUK39uu9|_;CvF#)yA_L{;3wWFGc@cY
zWxABq*)YwvHv4z-A9)?49g_Wn$t%=8aP1-xkr5oWwmEOsKL4<5f^}G>Zzwv}pH0U(
zPa1pmbtCJ5v4XKQStVMQY8stpL5Dxl(R`1g^pp9HZ!wz6`LolpK$rZK3S+Z*jsA~C
zBT>QfD-^lkePV}T<t<?sXPVHB)|n}CKi&_-Ci)A}-01#O>LWyPDW9i6b9;%n-ZpuP
zy)-<k?mfw~rL9b$#@LyBzHw1|xZrkR7I=$R=4=*+Hr+#b{3XR;#WpdZ3N>%zmk#e6
zu$afuGWpV>(s0R*-%h7{VsWAPUJ&EC`msGH@s?6#R8&M%X*#v@>{;qd0X_uEJ3377
z``nb=|KaUDxSHJBFig);@dE`9B`94wgkAy&0xG>rNCE@`QbGw$AoOCT*U&-;MM{7K
zLhnV76zL!kLRaY>MXFNHWM+QB%vv))AS)|ty>Irj_rC7?y7Zi=hC^>A{a`K`jST6j
z=qVG{E%v^ZEMI2jYGHJvD)9sTKYL^KOycD$BOGJ8<YCo7vQ}Phscvpuu6tZSrd3^4
zVtD~c(4qDI#@zl?+j>Cp9P4fU=NAD&gW5OJOR0nh5PrkD%Z&TGIlC?_3{uAqodcDL
zA1VNzIIjB}GIGD+LrgZmW?#Bk-cA7!@_1k0V-&kH?g&LT+dF<CXg*7w*x=B5Gy8e-
zpPgU-EZ_Z}wS;4O>vnj<X*eY5K>a%16ia3ZaRXZFXxs*8Ft~EyNj*0m`_tbXF`gJr
zw#_tMapJakSChroJr!>LAd^R*#ohQ-DhIE?aAk&YWa-CE*UiR-ol>`OWtr?T%wFw`
zfBYFNn2ySYV2_Yi>lxlw|H%2hsE?G6*=#mxw|lRXQrMk8tF)`fXmG~h)stgc{_Dc$
zjbi5R2D}KevdmTx1&-aQ6KWV9Hn7g`HFp{qd4dbe44A11&^Bo{{&Ha~6DP}rcRO~!
zk9HZ-&eV3clKXB{=WWxVr*vode>Lza?~3%JS+Bqf+bodEhswE9*5zpaB`=2`_X`~C
z07KSWK+~lG6A_O@z4Vq)Z|>ODNs=@{9L{t!?)sj($riubCW?fV4w=@H5iX5RT6)^3
zm+3595lyv|z+-t~-bM&P8|qJ-tX%_T0=U<^LyLKvuUuv=<9FwRad#t2T&sI*&#$Cz
z9V?ure~h}k+lNRPxR<>0yx<mcZx=dTL93npB?So4>?L3arh8%_5oG_;@A&=v^4X9;
zv8+S8s=%<>xR+De^;a9I_E($bAElZg-@V)^)jl)HX(ALvMMS<mYqLkK+6PpmHOJea
zAjRJaf*Qg}DoNXIK#gjVJl7e`#;tU7W#yni_l<V14pZUIT)d{`#G>c#@X+!>(v%_S
zexqifgkp{Z=7oaE(?U*x*p%{-yAr0{zmoxL4HI74-?JbaX_0-~x1V5jhpxbzlaBSg
z1$tfK^rZvIqubBA%spoBXuPo&in}>!;1tr!%wFNEH$3d?6;XGYP8~Y88vh-;Fx_i(
z&zJ6^XD9A!eq*e1RCh21=1QcA>2l-W*ce>)1oAjaL?Br8C(B<xC8zb{q?)Sp4eG91
zt1fC<&t4es<E8o{X??tgr8;2r!pi*K`%jf;-zH;TYAW>BYZN1G*|<jMQ=fTrnbw=N
zxTPok54}t~FPh6^T<kN@w-!EmhB8cVLlXJin&+D}xj%F$v24c2Ma&|1vJkeHigj|?
zX4PCGGcw7U_l5tC6tQysk<4)WSkaRBW3*{sSg!ihfqMWD-Re7G`_FJj(@Qs}DDbF}
z59DymgfEv#<(XBTzi76aDyOO#Y+Y$NNrct5P0YO7urD4|$Sl<CM6MGM_!1l2@TKLY
z);KF{fn{rS;3ar&F<=_ECDnN``_Hk=FQZ$VFIo$vG*eseZWtGJ+h2JxTBe;O9X3&B
z;$LdPz7*8H)S8s}#nYVG=^Z%pbqan`((>ELjt1EN`DSGH2XAV2dD{PS{l(p^Mj@K>
z5^&RIP4f|U<A-k@WWISdX7Sf+$4NPV<&c1G)s+Z!rrq8I!<)#r%V2Ry2t4oE^ZqkV
zyCeWLLl@2oaSlzrkP@T{F75Fy4|Wt|(!mWO85wcdt?b0nuG^E*70ZH{;KK84_T<46
zP&Z^|vl^QM$wCj&jNA{}o8qVJJfq2V2Z&R(!sh1t-V%qZ#@{DORD0vzPytmV46o)!
z6e$b$oa9LX@Y-GjLJEfth@_(=kVNO}B!B2hq>l*w(>EJ-$3XwN{#io3(>8QFcg#zX
z+~J&s*_~y&>$cMlon#jL9Ql&Cm_t5h`8s{_x_VfrS3U#Qxr-^8{NCf)CRNzx;`_s|
zXUa%v_bp?h<J_K3Ai1iPeXGs`7UhTuf;Wn1sO2vdu+%@PoV2&KKAOwDl`g{FGGVRQ
zHlOrpj?E?0t!N4;Xn^B$a}g@prfjt#OzpT=6IVM7++&$%F#Ot4fcU<~kx-yZ@_<($
zNq|7c+80tvw+`i}IK&Q+rE7xSDSs=9O4#*v``c#u0i`K*ghQRW)g~pixoI0G?q6Z`
z^GcKaHH+gftt+2-=OlOu)8t~7l}!?df*kH0zOyalw&84d1;wTXY_on`_cZCl;1|Tx
z38u`3@M`~(HNfOV4Zz%LY>@QdHEf*(ZG*#N8p-8YHAQg=LC!2{sAqbzzLLx7cI_}x
zvV%KHfNAnX$kx}Bvv4RGm*rP`4>=uoRFV8`c0SWM)p>*OU|w=$8F3!Zj${!sjJRnk
zdpu-+O6S$9QSJ9oz|a9LhEi~|7!^t>Nt<^!OebQab>d1rBM6Iv&vW?`mnX$DI-Z#8
z_2A=^u%#9bBi-pye_PZ?Z%(6nJZu<~b!;9EHn;Ir{x)HE;Nkt=fryiBcsRTD2aR{=
z!aBA|UK|;?>V+HRxt4(1p86kujZOMM5^f5yZFP(u>m>Gf{F>;f2Fstv`|gHzeu@EY
z1Kr6Zn*OSan!q><0)YS|XhE_w=^F;NLUNKGDljq#)zwQ&@joAt4SS@DNy2h|WHWc7
ztRYD+U_umq2^OHF&TI9*k)x<5cu$p@CEXa(x;Cunih$x<jPj*RY#3o8L~}%QJS9EG
zB=PEsy2>!gYdIEkae^g3RAdRY=lAtQpYJGGzl>t*TDPn#@L&1rvpEaSQ%9?2a1H)=
zZx7O)|LT`u*c}IAvr|}U=mh&Qw!v2>x%`h6@EV-3uxO{owkLgT!~8|KJP{);&xOhn
z>2x)%EdP%?v!rw3|E@{B+E6|ggx8Dxj#1jze<mn%AdrXbngK?3mISLXJ%YaGWHc=l
zw@jw!7=(mA?X{j6)<b}blmPE{WDazqaX*HeWL%0xI21ZOX8pTz)8S_YCf%*B9$&E|
ztF)~ScT2_5`n-<A9mDd^k<PQ`C0qy26j*6dyIh&6Dt+l`O67fJZBFRKhzOlhwYvaF
z+NJ*XA<#q#ra0taG^p)=xya6TtB$+lo?K<Pn0|6>zqjdlKvtiFFH@SLZ}G#eTyc|6
z&&s|}93z1iweN>oopf~@1&Xw@qfdL|amC95#emIKM*#sM1wp`t{j2jl(B^PF%R}zs
zVfRI3_}+Qdze4+u-!Do(uAn>^NF0PTtr$VtZg%SsOM%HZ8oOeY<b<WmQ|2^AXUf>L
zKetS4oh{QC_z!Gv*#mLrVdwg_$klbEu<HX2DU3b6!<fOyki>u2DrwOKk4*2_eEQN{
z^=O*IRmI6*#t<rg)bWJnPUUWf(zcefdyCwPVW0D(a}^>B*b(kzdGrv*g&}&L-$E{Z
zek_~fLI*nx$iDc{9&3qkDU+T6OT~d7K!2!_Ocr0}vf18#DAgT$U2)!E{cFGxi*@q5
z25s^_Z&7cil&qzC@ljH-Emxwzr1QMO5Bb#6g>SwD@O!YkOv*l4A9bQ$>zchI7f_@V
zyMX|i$#t{saSXUw-g4SCk6s!xrcIvrYNl!4g7a;Qx<t>eoo)4M1E`-MhJ%rqf4CW<
z-)H;IOSpZ1xM*k~v-M4*Y88j=^i52Fg?fKzA;%n7!;Bd~!}CV7l5l9+ltcWkUe-#j
zy^e;FnXpUI^jSm7D|CG!FhYj6SziHgLV%ls1OM*DQ}k;&I2|IExr0>M48)^{KHShP
z=KSv(6%X*v*LqXylB`HjQ+vls_m}{jEOilftb7<bS^utPNv$rNy+q$%HVl&3=K+jN
z95?Zmp!M;^gxfU%5Qb3Gz7FXGr4oYBJeb7Z=oi6Y6Fq*_<lTAG46S#9Kac5&@g)ZQ
z574F#Yy@UuD<++RFPPc?)NHId7&kVXl=s9Zemo%u%*DVwAT5+of^H^HK=GvL`qn`f
zwS7(S!}UC`xw)JL%tA8TEL~P&|B*CzOGwD^)~}RtEb=SX_{VqL+Rj$Z1)+&_*hbnw
zn^tckFN2xiO3X|>9(I&HiKZ&>%C>u|$BKEa!U*8otQ8qP>;YZqIp2&)D;Ij9IYFVi
z^siIbt1^a-;vK_aBTIZ^WdaV^@u}g$ha=YP>N74kE{_91?YbiZeIAWfwDkB#LSLtd
z7@N#NP-f4!%%P^{zl$jFB9AbNE=`y$Frd0at|+e?-3@@o#!`!`5&|?Z^*%R$5&YNG
zXv8coS2`}1t&)w@Y-9W`c&y!XoT%h%*nSkM`Gh~?Xo{mQNZEEuTJ=2ab4E#|azmuR
zgz!E67OvahRWEvhPRtYY^g=V6o>@tn3YNGu+KK;0Emc=5`tu_DuoL%=L{{IHo6Flb
zJ1P?FmbG5IH!e7Zc9vC9B5i$dMJ4$EMhS_D^(hUm`DdE@hmw?qw7&VBH(Yn+6<DV7
zAiLhC8Ukt^mjk9uYYG=ZCvH#<PqS)PjJrI_nB(CiT;4OK2Q4LrqAMgg52g7|@Mq~S
z`8jU)4GAfvIn(e<aEjP=E+sLy=g@XGP=a0Fs1r~FnD{fqZ0=COO(oZBhB3J7>3~A_
z)NW1Iu!(bU=5`qjjhH;n^hfag8Mx^v7f=kBuPTb{bPmIZl`dFs{C7<TolA20?;7Am
z{In>pzKB-up5^SCYf%y9v4~!mfXoFMe~dn$jcpW48`;DlR<{_$k{Gt372M{^<F@`9
zQQqD_Di(2{(kD?oicKv_`6PcQ!B!LG7;DyfJU}pkc|pn?q$^dnRE|XJ9-f^nm@(7W
zx-{KIt}iu-IwsprJ?P!z7oaW2x54L31fvYMWSS(Ry+ad=P2UjU9gdD8=Xl!7T&E_Z
z2$KV$Gp;|7Q3kmmc7KgA`IaBt4`Oytwv_U)3mYj!y~?@Ci=~TxC$5bWNe{5~rr0|t
z;=@<!Oi=~`((kt>xDKka6$6Uh4q!a|UpGL?^d#b3Z(ahsqX(R7;%Y$%6Z4)MLDsF8
zxuGSPcsa(Yik}DW!b4!*Y#lmA75w_2JjDY=j!>hYEjyK@(th(-D+bVZ&VBOL&q9lR
z5ti!Ettom_?)W6TMr&ffMYWW!_AbtE%>aGMWgvwzPvbIur;X@!W{6<Q|NPXGo)ewe
zexG7uCI74LCp<}e44S%BOV)w533x%(|DGlO>DVb?9xoFub86kUX%liAFKUGbB1WOm
z0Tnn^+v11#kS=nQtc>uv;frE55tgYncNwj!*)hTRu<m~7Hb@)yzK~ko7fZ1C(t4LQ
zQt4mN{R9a<)}>9Fr&hTuMNRU|4B?ED+e-*GH6qtvYO?*Z%C@<0pmmy!=3Eet&e=_(
zgDcoAR8={4A<#B2frv-Rl-<7Wc0SJfl;=yOvhY+N58dU`ulkkGD!$5=NEe!wDF`TH
zn{Bl=fNF~}I|*M!?~H`v?sA;OkR9)c7UAaB*+|3YKo(X?R(C#T_jtlotw;+^tFiIt
zVx%qK5*BBivGXARs}ZINajOL0x`J3|e#)ANkn6Gr+`}F^DQShOPQcr3l+v<saXyu#
zsy3`&g|GE}!3I)v&$$4g8wEaC?u<E7mFy*(*d&{^IGan&ttwW41EKVz?~2psds_~Z
zJthcrn$7xPXrj~Z#5P`nl_j%aGfAmkQ~4%Y59-jB8N(ZDDlgn<tuigK{4h45s>f`}
zKZW~~zu$c`vjl-Z;SX~Ze7TPWp4tH`IzKt;p*bGL8tx(e-#=td(z8G69G#0TwLHOZ
z@JZbF#|>38`Lc>+jVgR&GxD1X^*TjV@_-be<|T4UcL0qXJ^D6^o1-A^1sDmol@{!#
zy7)ec!)d|djr-J~GzJ)Km-$dlHL27<aB((}RXU0t2yApH(I4|Da2`4qX$n%iOCUdZ
zzRfOyCqQK-v-pWd1B2SuTV){56Wup6HHDqb@JoF!gpuFiET85~i1f|hksAooO{F~7
zf7@jDvlo@Jh7u6ci>P{$trQlyM_cKnC}Zt+q9rd*lcJLh<>l6|ro~w>dgIO<#Ms9F
zO<4GQCCQ!?>Eh81;=*x&O{{ht2b@%1Fc)ah*BybIXY%`468mf7=W3zZkYUeTDaKJN
z`J@z15AY3IjdT6G1ZP?}R6zVELvkNiuZ0j$#QoQorA}#&D&e3+Sk_!_#$ohZy*AY!
z?<wwfZQkXHi7!2|jXb~`m2~ab9o>?m=<X=`^9%o<*Q_4G-*=%#MZ1lazNY8ukfOdg
z)<1~H9!0u|r*a2?wzdO6|H3i?TE6?^lUEwgR#!)XEr%BEUx(8qi23I!^Hk`(57o)2
z`%exE^FWzB-L#AtHzuI98h2X}O@5!Gd9DA?F>gi1mIz%sda9Tqv6Ht@@2y$+(7*|R
zuGmrAa#$z6mdmAS&#1@~Iu~}2={ZN@&EiK9!XR)Ov8UX~#8}AS*3MHWuyxIN71Y1-
zqPj6sL|N&{_fl3TukMtiU8~yi=^hE2n`@SCo>Uv(&s&*FbLHX+#p&M4?ZO(;)Rmly
zApg`;Y-51OKTA0-f8n~PTDOG^Xm?_4^Cbd8q9a<-soXPiQ245qW$2EAVq{&}?a-LV
zia=?8((<g~8W2Wg04LE2-N2z*YxtPQ&Bq)C2>Gih>v?#2R%>kBF-x3f`F^2za<6xo
zm%{mtK9}-ClLi+T*B0Xk`04x~?we|ViPA~B|Ic?>=6GEE=d#4d-xX_|mjNc&X#~>r
zvX4+U!QC+&*yVHl!qu_f+|pCbUmB$~1~M^HI`?!DM18&S{EFe};JZYz8X`Nm_SXNb
zTcyj0JP$=0p|-)mhhXqw71pF<n}X6^;t)F)m^7c}L+0wm*Q#q7e31iqht@OU;ZY1v
z>Z?5_4EfwSs+`*dZHSB`D1mxKBO0)*K3TS)8RLe&xT&B=FtiXbn^^eWn1O0^v8;V_
z(gs>+`=R#`%9l3pqj6$$j|%%a?Hp)b%{?^(S}<-DS602;o~7J_4ISGQZlxP4CsmZV
ze`N7`hx_1DQ7lwpF&%f;Y+038m3*$V(x#EJ&2mJ_=+RU14n@u-Vln#vv?Dz#tEbzp
zG8{cxEhKoS3n{V$nWt8!#R;p96d~br<spME9sgpJO8q!SZVJS^Y6{^qfm>ud+9j`F
zca0uzFRjkxq?)+}3tEzG8r_3~Ui=%F1--{trsz@>11&xBbDVaR_oDHB;_JdJFzSAL
z_wm71@=<%CJ3Tjg!I!=BX1)AR(zEO>Y;8(>bAy}Ql-2LxELzSvb``4AzucwLt{V!*
z<#gGP>qD|X>@fc_^%tH847*IdW@iwb<-K=MBsEu_KwI{2ta#s>lD)-Ut9O(1qgBL?
z7jbM9!jf;gz5WR~5r_l1OLP5tuG&R)KT{XE`oVN_E?f|S?VW+g#UU$Gx9k2X<^BBM
zwLRiH+Bjg_C<r>yF~-0XT6RpO3bnhfqKVqEqz#MGZ#S!(eBOMlx}&pRGbPiuNs13O
zGa)f*hyGmKJGkihh?<=ZK8=WUO(u|$I>ytF#$~)NCfnDM-41WtHYT$37i2z$oIcf)
zaoppLI>Fmt>3W)7URhjFuW*MZvsR}s>zmItvS}>m;m1EIWr6|xhqFy%Nz|R?_S3xe
z>yQ6GhdTeSFJNNxyYkMDKyZWqyB1mq@Ynv@8UA{N4X_E9(9+<2D@~3xi*?)yTx8W$
zM4bG-c=CAW+FiSLvV_Z3qqhy`aUpmgulno#)Sd@vj@<21sM{hncR|nAHpN~+_Y~GM
z7+{(QG#1C?8{#9?MwRLm%m3xWjZVwlI9F)(?c8Z0*4It9`Kw`7d*U<gx9u4a8j-5s
z58-N(oAf1vnGBKA^Pbp4sdi|iAjF5D{AF_IWObI=w$0<3$5Pbqg*nSq)WFhbZWm=G
zYiFGML+M9}{FsnsDV$c3CeUbX7)5xetgvU}yZR6W=6gY6Z#D)~&pUW+NV1RU#MXn{
zG{%iSVB8%8W!B#+W1Z?Ol;*ly0mF}bW$s})GJ6l&Z5Z8eflDV>n<&k2KOKgz$=mTZ
zv?C!uc8}}IHV-e@ObR)g$TrP^E}Le~)qiv6(LqtBLfyhEgiku(9DZXYFNB1-&qE0U
zs(I14v;5=I9Ge~ElX6Qg-)Ym-TU+0<7Cz2!K>n5F9|{B%)#N_>q(-;UU7@(oiETUH
z5Q9yrlfp%J0P>DN5lTs6F-GpBNZY}z<>S2sjyGQKw8v6z{xIqEytTfv#BlZq7dzDS
zyw*p>TBmD!UmB^!?j8DBXn{J)HNFPhzxcS=jDQYCO=^|TqPuZ6<qnnc3tP|xmyKrP
zgm@qsXw$Sdz!Jq0V_acbF1LnFp;z;`g1vwjU@IhKe|iMw6>6)Q4&=qp__#R&e#d0A
z=&RJYqr*x!2>lfh>-S3Ux{4-$q(r>fOm+HP+!e>0Nyq)a@8RG0@e)cgUgg8i348HM
z{00qjN>?P?nVmdtNHMIShB$!&+}ggE<bIFs*Q?Kh_7VZ6wI;P*564UprcfmibRIr0
z4xO{&t2(jt+lX$}8T!nG#Q#q`#;`k=t)^TFP&XOpRyyjE!Yp23U#T+4<qhpdLvsuK
zY|{s5<Gfu3#I*$>lzB~)15us@U6F9l8h<E9mdMd|=hz<UK3J7|LVDD_tDbW9X2nc?
z#}_hBR-H?0T6mffRj6HpE9w82ucm2K#4f3RjK6kOafjvpK$aygA_i|^CVyzf@5E2d
zb9c^c+VL=?r63K&83YYu6BA!?CI<VURTq8k2BJ%o^hnSt-oxRp5)qsk_R)-}Td<G`
zz88QsDSGjx?nyN|%MG^QUiO|(=_j8UXihlH-GS=Xh<ajBf@CY^G0l^|m55eQkeE2)
zRUAUd=u9+Epu^_#1e6=TnZ9?i=rWxt4)8Mh_xv8fZ-+PGcwuc&>faH<W0JvZ9Vd>T
z>mOFFj}zr0*UAT?Op6!TNDcBWFHRx?Enw#$@l#=zVv*Ed>Rx2pt&Nnd-8C<*nWV&1
z@~yj~%gL5MLMJe$J(b+O!y0h1QBQdtuOr(ZDxdvPkUNwfa!fLUsC|7!LA5LXPMaxO
zKpK#Syr1*BD&jWhVwz<*WVeStGMo(~R=kReM%kac&@JJUY)%HIrs>;`{R)uQklDNK
zA_j7IDzG#dn*b~@aVY=g@xrb4S@(m;xA}~vJ?$~bQU*J#PuVWzqw9xq4fv1*Qj7`2
z5N{<M!`4_uHtouizF$2MXW2r^jG3ugHvXe9_+HJjWIjXH;yAde*pkQd4)>XN)@5j8
zqK&4dZV`05KP5W*3_E39I+CBowl*0Z{jQ_Dvdx6(oXJV=l5Q@n3*UUsG<(K^HXOMp
z{d?EDUY|H&`;t>O(^H`AZ^+G#cUGqYu%>kJ(u`$ru>d$#*FZr)K~R?N$M?Ja>0NX)
zhuMj5Txm;D7Cxp~YBGHR)w5H;`eK?x>JM97U{wCPkmbfg+KZw;nWo~#)~A4COQLb}
zCPS6<9gm5(40$$1Dyz*NZ1G~e2btCM?VhDv&!gU^S8@?&f*aw*Ux?X0@0t-#H%V4`
z*}w?t8-EwiQI9|5AfImrHR{IQY)RexfoO>cfvpSNfk2(7hVw;VZ8c4r;MGj(esj_L
z`xpGt8L;n_+=tdO?UTaFVEC}*h`;MmMH2z$asi6vX5*EA)*LbZ7GE28Hvt^t@8^$#
z+?cwW*PbEgGhQh}e#-hrN?Gzhd^1F_E3hp7rlO3a4_b`{xq;u8nj|C!LYsP<RT*4l
zyB5LH;SWD7k7)`05@d1TZ(D|{Gp63(`Z+33pKQsl_&xGv%}IBgPdH!ccTeNHpH+M7
z4FaoK8MaT5mw}70%)X#f7DS#=^AD>QoKgaN$Awd&#dcFMdOJ_|#-C#bkTIjVUtS}t
z8}(Vu#LCcTa|MNXD`?DFdSKW}xfPXLm2LLWM>T^pfN?Oh!W`!_Rn_j&42v-?`IJ>5
zU8akHsj&UZz{%()(-LXLUV5{O2J<rYPvfEj3|`dycZ~xG8|8IsVnCdNL~OjQ&RX#1
zESLSQOHG3wZ0C`iC6+r>b?PK@0ZtoL8UIIY3ZB<Wmwm>EO{X?dBBD9UlY$>U?!72-
zN}|=WMKaEiv;dOoJvNaRwQHlry_KEp?NL@k&#CuJW0Z6h^IOYN1txLgn+F~Zx|U|w
zcdWn6BO7H<BL!}2>G!HneIzspFUFmFKUFe!b4IkCh<=3}m+YN-9O0i9NUwLg>g4J?
zBiJzZBRZzMm{A8(lqq*g>$6vL=bv89{FTVz*rPJ9?ceF#-`!#K)_Z#IqV2i*l&MH{
z7>}j4rti*K^UJ3J5mIG-15G6`FE#LJ&UE_=aV%qKwXb#w#W0TJn&)C;fva#*qlEB%
zOeVgnjCZDWU7g|2r}p1xgIQnsoER&3-{TFJGC3Mlj^epDk5wDX0`(H#=3W;1S>j8j
z;tRVn{^i3YE3bXjRO3H1&^6_^QOV0n4D6aO9316XPU>h&WB>&WdQ^XQ-0RwX@@~S-
zGOrdEUA8iCz`?RIP&QQHy{Opo16Zx&c_;PR;ZXB`*Syg{Q@y<li(cEvdXE(Fb9=Th
zwYxtzZcCc^I}KRu!46JYUi|5+cW6dw3Dt`jtId5NNx{8Tb6YZ0dnx5Ql6RmRmcpm|
zriT_uv-W8LiCXwK%6O`<;YkrZGaGG_MmBzZ6H0R}$j@i&pksh$gF3R`Lco?co4C_x
zqSB-XcW1j@SET#;?eu3N%I-%b79ne~xDW;}D#C}ZvO#5%TuECvyxbL~ZVS*9BALt_
z9M{wWuU!Kr{(N)6H^@G9+dD7&&PUVc$~!wtjSX$w_bP-ImA>?~t1?Uj@s8g>E;ySE
zwbFMhX_L>tF*OTFNR32N+vcG+=Rkf#*AK0Ga#`nY*U_F*gw6qK-AF{GPdgIHk{9x2
z`(73Y1jxvl6KZXJzYIxn0DI4I*C!6Gf$ODndRanrwa7-{*}@N)G?cyQ=C0%Mkp<CK
zhxqUjr_PeHt*_!S%2^WjSY$1jXmK|br)rYD2HjTo?gTh{>s5@EnpCEjHXAy$G!h@X
z+Bj9ey@~f+bCl+D-1n2+-yU)?T{10Om}<dx%goJuN~8p*2@dNNx>!wcCk4*2*%WE)
zD9F)my?yGa&IBg~dfelllK1`h(r7Qi<y+leMd)ZlIpsWXs$mn2cbRS8V#Y@q7TO%$
zXWl}iK=NGTMsngugJ`QSg}UzIUa`Z96D$T#{<&^!Pp|oj+Vg0`9&qQ}*)GbM>@?qW
zWQBY62{4Ar1b^0*f<u5h%@Gd6xNTHgr5<4(eMY?uoi0$tavl_qZ`QfV09II_8VA?)
zFRqVLDM_i#BP%gyMzLo<($Cw9Ri$))cS-#Dj1r;`(2uoMX^cPl#|czy-JDEZ-UG4*
zH19c&MY4*;Ti7Ocd%b76xfh2w3mkRoY-HIaI_(G6;-VR1JjTTuZUGzGRq+-Xe!`>z
zZ#X(fmv4#ux7J?pUSR!xM6-^OY<z%ssG!Bf(F;vI&08K00)ZAwodZcKXO|2sqghJx
zHJd?ZE%JDh&6z|DL%JC`pLPB?t1-pCic*m(KIN&~`ksikZd8?Gi1DmK2z)~f1pM`F
zvHo0U&#RI{yshq7fNa`O3!kmj=~?*eBPpqa<W(Jy<_s_`VBqO#F<*l)4YzMjp~w~J
zBJPlsCrRK0KkZTqscBDI;NOdEnM&!g_5NH6xWx8I(q_=k`9)K`5;QNk+uKAOM3**6
znGHw~*h!9qmGWCd2g8YMzthC!cc`6#IC+Esi4u*#-}ozi-bWE18U7$>Zy~q0XIKC8
zg_0lniNx+ERXFOMiqsQS&bgVuZeRO1#!3%etOibhi=ztkq}7FrD4%*T`L%wJlQ5G5
z33}B%>hX&hvW8XUh#yLqt1{ZEy!rXU;?w6JR#{1~JobH4&aYGLDphU`M`1<Taupv|
z&P#g~DWYsHC4~jt>8+EJcjqnX_69+dS(3m8Z_*>+VyEwS^(1fG42B-vH4D-a>H{Bm
zD>GgleCHjJC|mXkBfvJ3QfSiAT~)Vj0;GkX#mTi?fB6U0D{%VU-eIV@y6PRy8*W**
z4Yo*Xd1*+pwfMQ8xQaz(oSU()d>M`K9%$L^A+-5hHky=Ykr!gWR6ybz;5}`5^-r$%
z+I(Q6n;r7F5n051sKrv+4Lc9))C5SJ`8rW+tMc@`1>Otm_lP&d2t#)In6;zqEN=iu
zHvQ7#t3U7s-d?=^x6;#H&QUtbu4|*uuug;TU4~yN@(q^BV(S`L#Zz1R%fiA?_6=xx
zu%5_w<B;sQkgulgK>o!V(?U-h350Iz3#c%p=t;B0zmcs}WS+;;VgbsEU5BZ_(>f5;
ztX02v8<qA+mtaHftv624dH-w+lTM=L&=t$gA80uz4wUaIy-`@a7WV&_(f<F<-uHU)
z%PN_3zSwKmAhx%`8&&yUbe5pH-Np;=|0ztHiM%O-U!MNo`AiAJ20`f+`Cpr|>x<?g
zMKS@O4DvhSeD=W;!p)1yemAzmQ!27@{I}&K@fM*QVY^>nl_zF6>)J&`KFQQU0gKLT
z*6;FJ3#vk&JTe{?TqC};ia>PQeHiER@~m5H_WN;sbi*P+|DS|9A@Jk?PaXIB^;iFI
ze$fB(&s;bB!2q&p`JJ(%p%)z#;y-z{g{M8b_L;jnI}$mo-FCG$()KqRj%byze-Zzr
ze_X=F5XSx6d`vIW3y5>MMfS2@mD*Sz2sO*;>pvK?_&;L-^d#M|ve-FFC^Op|L8NNI
zyn7k*_<z@Ycz7rWc1i9nrRnsC%r)9a23uRfUjARVZ5P<yD)5lKroh(J1XkgCnQ*fA
z7VSz$0RI-9XWrcyT(h=TOIGE6m)oN!Xq72$)!EQ-k7j*QUjh}@{q`XFsvRcHC|LE{
za6;u{I7}_Yt3|2}-TEiu<OQ2bsw5K5A!pq#CH#i!EbGxu8+YsNa>eJ^Cyes~8`f;q
zA=n7cmd!x46-)YERFfa8xdmdP)?DCWl8&W<jpN>r(?;(&o>Byd@wj`s>~lv|$$5_!
z|EAc)x1fiAk!I555ILawXPIB-_7dZewpEUPg)vr9CKatdCUyI<$9^wn53U!P*{rX(
zF)Mk=HZ2+)t3>5M;Jz!}`!~6a;EGk_xY(`il{>@cR0cit>4y9>hwd0Jq}6Z|2Zmq0
zwK)641LT9Y;bSiQDXmxpvDYR!<n$2CF))Tf)u#^oR<TWX13#DAxM4IC%<Z{j6agzF
zZ=TATH~reG0Q#u|@*X0^5#0LN<!8;LrE{@#1#l4rQsqJ^d?}G;%T&!Cln;xbms*5q
zI7dEc$mJ@N;shTp7Udf8Ckh|g#0oxN@Ee5a>3IQ78{#+(Lj?p7$jg)44sA$HBJANP
z^-|PF9Sel+@vhKcayg-lL%9yZGgS9^;#89y&2%zNM<wrjLz_M#S?%lFK3Yq8j(&CR
zV?-{0y_2b$#Q&yheySztrjp=jVG)trg-XfRhfftD0%F*`I51b`+InSjlQIsA(iRO;
zrAjd08x28A+s7gvcc)z5YoYZs){Q$c$g)Px0&=Qgw_~rPOU-enba{*m`na%Ci!jWx
zt;<1)@WZJwkxerakKWP+3bD7=3iANIm2$0USqcf;PgawXhrIt<wGN8X@M-mv8tk@c
zd+Jt9y7AV!ats}tku_2;W%ed6)K`%5^8H|7hGLLqsg$4I{iU-b%A48(`OJqdLMm>H
z0ek#Cu`@{!6EX||07!-L|N2aWX9dSmJ29wWE;d7$8am4}q5RM%Tz_y6vRl+0eI{o`
z6Wa(1pdJEv!o<HOesoZO+7k!-XfGYz3Xjh2y_VUzAI^)B@z$4DXR;VHWOQHoLQwVS
zEdTv-&wuYP{fXS8=zqFG(^Wi}%zE$<4&WhZ`H2~pnxS)PlJ=&rlX(Z3cuduvLfP-U
ztSzWda@DQ_-0?^ZSo3ueFVjjxG6Mi&N4@a_N!ESJqGo#j_p4)a9@7wJ8kI3xALn1c
ztdyDC9S6aRs@vzx@|F%U=h(X8*Q!4$=ZByS_yEELju(QwVY_$Vt}j}Cw%V2ytuthM
zcSlW>c?9Adg3BJp<MnB0vsa(*&=pVJ6OQls8v5<vpU?7frvLuYe9MuXANMRJZsU8S
zIf4!bNBm0j04Oy#rSL#w%X1df)-V&0{Me2DmSc}H{-c5I=H|(2rj@mGd0NNmhySkG
ztL<sv-Z4=33dZbds?&~6ZotE8OPAsWyRif>X@`=lK**#kC%Vh#^Fj*5{aB`GZUULn
zEB^8_lF9m@PKMey$JV(jmQvZ)*?wz0h*X?DHlmTo|0S9*?l&e8@@=f5u;9OIw`h$J
z6&@_L-O^ae{-46ji@-Sa9IyN~+yq8+uZ2Xxi*&!q8>I`GllmTJ4ME%D!v>y0s;-kh
zOr8aLds{Fr55Ov|AlPAxvK;D8;Is8_42lgwYw0YKqEDf=m~pN!9{fpA^<WpS%Ax#+
zu5N)o>#ra`0bO*vfEMoE&|chZXG+sa?ZU9f#-UI!J|p~-<DFZ3q>fzKwyOBGu^44f
zRWwo28qIn9ly+~bR=1+K<BQ5r7HT)ca9kxBA)|$qJFt;=JJ2R2IZE<f4Hcu0-@;1=
zhV-G%#kx)cCGKt}Z)QtD+f#9L$ke$be;?ARdP0utHbU%sXeo1##W<G)uarnA$}#Y%
zADd`Tbj`YessiC_rJkU`6aR%NJ!?Suw1~ARI#3A7^xw6Qhgv=^sw+(-bfxU4Z0KMK
z!y=e1R!-tH4~&tI(<=j;T6h61Tgeomsr7!O$Plksw{E>fPufp2_ap(Z@C#dGF0;JX
z3RdLd!SkMnZb|WVY4Wx63e4ls_*Eb8GtW1lfjhv5qm39d`7d-#B(vq%;3{vhVk0nx
z4*7sa32nM7tI6H`c|X;hYAM979d7wykOI*yS`TE<Q2BDO(si$UJE{H|itOpotI_x_
zV=x2BKT|IXR|$AYsk-ObGE2NG$GJovEj=nJ7H<D*Z&H!P_@(3{5Gd+0sp46ggJcJl
zKL^{x{NST#9n6oc|5~V)*-|)8Mx=j^6kFau`~g*l(W`#k!#l~Akr|75{%lReZS!On
zUi;l*+kAa)T42^3Q_4d>bn$Zm3(+F9yGsG<&!5OCJtYE!>#Z7ojv_$oTZaggGup;Z
zU?V|g^0wBtH$)llD<zKd3T25+JmQ@=u^>;ZHkg^d<_v}1F;FG=e6?nIB<8_5F5=Lq
z1MA#eMRmva>Oks9(jmO<CvP$M$kZ}HCAg!233$uk=9llip_<Uzob{qcQv7<`Rz-Lf
z&PU!X%fqEM0}=BOSGl~XfoK<HaXTCLokueaxw3@L^+>p2Voo43EH8aeN^AlD^Bi@Q
zE0jUAeA!Mdh$<^YM_FV}_%uecdh=w*fTRoyy6WW0kRKtSi!?uBIDtc;`KVGnSb~K$
zzoLRnZw~Orp6t_?_I={K*5gxE`wIgE2S*3gP2Tg=6*h&Yu{?uuRg%#`bzH}i6_<We
zRS`R`fpXuz1j^3cqqCE0Q+lw%#A04;6aag4d5X`pE?(Aw5iMgxBs42DpQXD@rNhf~
zXwhutehg2Vf-o+Heqx=c<l`KyLMpLzov0GTVm2Z2&El<dILAZMqmBlSSnr6OREM$0
zyMn5x(Vuk?WbH>Ox&d)Bq5q_%L<=Lg146Y4x`>PosabXxMqNd#x*s0?d-f8njZTs|
zKm1u}wCuXd(q8*0_ifA!M)oelVx{~!hNnoUq3S;9Fhp&H8&oQJ7tIx*Nfxj7!EmZz
zHX|}}PV~l0;6|>K`T(;RR_;6^P88>2L-^ccQ6T1ALU`QVJQ(OwMsPUSUAYg?NxzEL
zZ}a7?$-2S8(c&+*BN3o!%scTLKD*m2I3;VzRr^sMZzo?C-*Oxi-!<Rme?&q6zZ@_`
zsY}Aq70N7#Qph~2G<xIFqOQkr05=x@L~O!DC{D`vQuPC0*L}TNEkia*J{uPyf`@6~
z6s$K2kYE`H>)yp-b>3AlR%4yT(G*k_C1Be2!di7ypjTr&RE_KiS2P4XN!yK*GD9nO
z`%U77t;6Xv^*@qF64-^tb^_|Wiv%Rh61@83=kYo|jZ(rte-__doUR;z04r=mP@%GR
zjxx%8bvBqB!8&JR+)k@cvxA8ZjjB2Ya-O8d{C91DU%*CmiKJ@(7B6R^b2K$Pz<cMj
z(gk<8@sn*gwcE1EO7NW`))cIK&X=~>IL=ewRn^19+;0c<Z`o)6yJq;ai&O?gDElE>
zRd&Z-?^g=aYxaJS6rypg4!MS{8v*D`gR6=K4-Q?}PAJE5!LL91F@Y|gRoW(TUeojK
z-7f{phUbyTXEwWm2_j;`8w$YaD)B$xe^!n)M~%^Ihx(m6V^ZiQ!nIuzj$RdvyxWml
z3mU3m*=axPeD+K7pOJ_M&s{dMXf+5Dq$^xTU^o^w{X@R7d5ut}6m0$=wMc7*<B4l{
z!fSkRLSRm8N^)s_cFRj>8TL~|Jnz^~^<)P%u~gpxoqp`Xu8cvzg=g!pS=^o(>b_Xp
z<1U%dS)Lf=?DF8ny8W+Hul{qsch-zD4oyaU^3%nnWg{2YZj*DDKKruo*kY`dy3?-n
zdUtQzQ`FiVcq<t-9|lZ6HQFfHF&>jYDSEfzP|3lpu)FPFGo>VRoN>mGutsm8-kU<V
zm(i5CNEFeYZGD$S6tS-}(f9$M`bL5wI@%TaruItrsl{)su!DGO8WS1?qG#1OYOKP&
z7+yMozDq1~Je^m`1frzsuQ>D*a;o$Th=Bs%G<2H)dDMfyFGPlPBhC?dG`8KSSoLU4
zDKsFQM^U<3nRjd=>qu1%>$8@2uWjQcIx3>wlepL^ulhUIDzS3ve!*<HnaD^}PHdiE
z|A_~~Rpz@ha<ob<fT`_-+rEjsYupg}kRKmd&l28G6j=9r4gzZY*!_c849+hYh;@Il
zGZWbUBmx6GBF^3j`qe{=j(!T=xXfO}fBar{{kL1@J#z_ke7AH#CZ%M9JLyXF>06Gf
z<azRc*OD1J@tTdxUVKePGMUDk$l|~JwO+~ceax@t*R;m4&zrTeqGp1^Er`cADg3(6
zsjWtZkix<NIcs5!taW|Z5+3+?3-(IRnUQyS?uxAkqbc<o49T3%xgw`E8hX0bEHx=)
z^;5EA0J~IM%&?1FvX0v8(r^$TmqV!1gZEmeXA`QEVzVzPzvI!nBI^6maRGi^VfWV@
zU0pS%Ye^lq!AGHfVwgRAv<|`)C}&|^1kKI&DH#De7m09u@ZKSEDjUEtsXth^6`MX;
zmuHIq&u?=vNEf}(rbTI-nj2#2&uXOp_G+3e-Fy2eU0iIIeq$K8t;C8Htkm&aBPH$b
zx2_$|Aj^cUgFYK)t)B|t|Ca3$35aucg3kL{H<g6uLEqV8NPq;eA%CY$k((J9yo-a(
zjATUluxv9DoHN<qe)*&K*~;Glt}(&R)8)GkR5<g78y2bqQn_bUzC0iK?;1R<o=XH|
zg~@O7BHx6asIR=1%?tW_MUum&=L@o=WC!E2ioCxSElIh5K=|+4Q{yKC4i%X{b8uBY
z+5<-T5Lu01?_CWE329=-%9Xj`ddanMkC7gXNOpWwvtH6;;>2RckcWF*k~EhexSM2{
z=c47UXgD2%JLh(s6Rf%6squTjJv5J#{49mnP7(Gbek)|NLV?L4j=8_eQDTvC5Z5@2
zaQbygD9==}`T+1)Ebc8en?&+IO=4nQmaARE&Gf!ma|D8RoMY$bHdkv?y|7d-dtkf7
zp<293+b}*dCU&dmcGo-s_N>LHokprRx_&uz91vOXvg9#f{0Jz1B1K&#j7;)kPF`0G
zZVAqb(t~A(-TbQwBMyVM?nLAsEs<bX%oOSc7E)^*$&5~MT8Jm3x^4GxyL)#hot|GZ
zI&u=cFoXvb;T@2O(Qr|Bm~fGSDjWggL869RIm@Svs@-tBw~=fCMzh3xP))k9erDWV
zf)ISGHRjI8Hd{sY3DfnQ7xKGn>vofJxVv*)+IRhmN;)J}`bg44(}42!V3HKAUdK7O
z>D2>Okyyp}ey=*cBwK%G3rQlwj*l${Al0+7Yx0R(wZb7#cHM8iLD2plA?g{?e2?K#
zc_PF%HV4!JElb7$j#S=JQ4_|6eGy(ccr%@GZyc`RUQ?h##iyPb8@&nt@vV%|#FCCn
zu;U%QF8!~^Ds{VixqSj{a2SiEr2>2u;qgu4#kt&4sNx83XJV<)dkPTDYF<*}&=cLC
zAUzurxp<jw7azy_$ZR5c#9T5-Yx>1UkE-z60);FkjyWqiPn#^SPjL_uqszO!%oj$L
z&-TDX@tlUZ;}G+eW)|a6e^ZrKCKhY=38CWP#K!fW{83J>E>%#u<HJfP-a<IGsLk7a
z0Gt-X;2kO?^5Hta(UWlf6jV>B2r04?0u4ynEZW-857-Se>#mXQMiT2mn&{*)-yF2H
zHB5Z~a6p4bYDrWO)Rq(%CcqoFM5pNGqIbPyA6K|55~7+DLcEvQu(fW+qlV@U*fn6e
zjSfm-_QbF>wfdS3%(Vw1J)3!z_na8*6Ml+>NGYjVUMVi+e2X0V&3~fap0#FIk?|;t
z>R=%<T59gnjUrv>rwNPvy&VVuv>2s%DsDWDV^UoHr&;-#hbeyUr>0q$7-j1h9M<=&
z%;mX}hTokz!^$4mQ%}pM6?);i#rf++t(>b6qg`n2hTZrtg>+8UE_0;tik@>uPkkx-
z9c68_l6KiAC+_uH?luzHPa3A9;q(3166l+ceei$R5(7xWu-A^~gM~J76(B3Hd&72;
zE=bqJE~L-T$&+<=@;#s7;ie$6n3AL&nj(HO)tdn?dBh(hUSBtPEN)F4Xf5}v<&fR{
zXGJdcc_V^;w0>D#J|+S=B*Ck;Tc}Mk(qx@>6<QgXTz@b)Xw-{~K2z_|H|JxQ$iZ+<
z!L(3>Zyw&A!=f%(r=^a%9{$%O#_inynDoVCZQH_jc`G$Jh<%J@vl^;R2YuiPBvC8{
zVVoG#`t!~^L(`NRg1keOr#oB-FVHxvo*WjY<B{&yY+X)lci0|zFxo|QdZefJm-&}c
zU_HaMvJ;*$4CDuNE+_RuMuH0-q@Hf;Rl72|_O-7cr#$ec{Lt!VmJg&r=DV>rjBhS@
zM1yKy%8%^y>~iTJO=Hj6>V9`><d>hrOB)q#`jtZ-wX>7*@Fz|mKP*>bAG%9*sJVI$
zDIO^GtrRu>IC3xRS2d<w8f+Y@1-G|~E=E8u$n^{d`5p?~2xL-lGUD5@x%t_-h9loR
z3)SbQHneV|?)=`^sYQza`SjYkDDD?$?l-?}Na&}WG)P+~un1)~PkW054dJm3lVBsD
z{NB#~cTm2K)CP5rxExtTt*2%wteH+7%%TUQ$nh%TPh!i;4BgA+R~BKOR-L!2@x&*R
z-%`EA#jVLQLSM8|QsaI3A4Xb({m#zjab_bnvN+q%!vsgd%5%p^PDECeYJxxPz0EO^
z5^{C2qbPy`v+CgWstAOCu{_)Ey$c|Xn4r_N1|Ok5Q1yq?1G_K_pSwx@ogrsuzrKfm
z0^C5W^Ov`x5H8{Bt%>Pb?ped+jl5<NTkl*W_Nmu#XY>9~YGjl9+GK``8l}}$0ZOyQ
zqa-P&p#bJDyM0A<Engg)%9Ix|QdGP<Jvbx$&a0+x)3Bj`FV+0q-!RL!W(?o{WC?e?
zo$VT}$l&#F)`*7mVpqAIWjn8^#ZTKO(=HII&o5sgBMqlG*#3b@0ozvCo?`)?7g=@x
zo}HJ1oIltZ1jJhRd0OE9wzzFLlg|1r!aUkpemU}XVn>#$wv^^g19Lu}01a1?RC{^~
z)3MFP2_S?|Mr%$#G4$XdE3AcOa#K3u%0%z*C9Uj(MF`YMvw9c%tmayVTypWNej8l<
z?^^1PC{um^9`ztc|NPn8$@3eR+!5DzGUWz-d&yb;cP)T&H1_gO@-0gZTS)p<IRvBm
zJ2LXCf_cUt>A#liY9wT{r4l<ibDfxcd8uu{JMYIL{vb4n+}o8UEd|>RQwC4md-{)d
z_^?TC3Gpwy1-r=zt4274Lc5i69>~*)licUf^-v=InNh79R*x$WDBF<)M(El}cEd@@
z)&c(3e3@BX1zATn-tv9T*QG#nrPA{p^WC~acJ~#M*m+9gqsDZfk`(`d>~M^e#r?T0
zB^&XR9t|9rWW_SHd+V1H-3XJ)FIp3kBiU44aEfN}9K{O#XH9?x^@8n?VSp3{>0*)h
za*h(_PD93@{3?`MG2Zjr#Jw>$NjY@|P5vaJY=OF7L77n(E*q<pIAhLVv23L4r(g}f
z<0fa`b~2?WKiW7f;o!ou%&zvWX`Ov3lU9n|?oE0JxBZ{uy5)mxaWnohpu$(r3<}Uw
zNOIK}Cu~wTWJ1<0hHXP<2d{udOPoS`-t}sK!{Oi?`sFG9=HvO)s*?MXP-+~DU?mkq
z#AhnoErt}1ld?$_qB<pmeJLOiCpQQtH`p(`P@a;+L9OTS_O8#Zh72g-=H_kut+N&0
za!B__6kEiBN#94RO<KLv;=Rbv%(d%n>o$K38$vvifNKTNYAS6i6kCz?;Y3@wM4(Uf
zFy$~gQ_K-w?dH$t&;oY6NOfmXUpWK^&55D+`m;3f6pMX#N|>}$;0pV?7=MzOmtO3U
z%leA2Tqk58@^60P=haGVsex*N3}7p#aZ;|hHw<2$;N}f6;@_D!UQ$zV@g54nCx}=#
z70GW?bPc>te@;2)ohmLo%~i4!Vxj1MP(_8`I@&aJyu+yGK48xr$L?66v8hkO!jwAc
z?UGEI%_b5&NI{-U6Fp5IVjJK&)w@eZ`mI!N{k38p*w^XC<E-i5d=F+2Z<QlH<rzwY
zbmK3$RX*9sUlPAItF4z)-u1VV;hFXNMcjSLPVM5SK(i@3=Uj8W7=uu)`8Xkt-+GNF
zG6V(P%1?VKOfIdlKuWcLS~QW}kz1U;)I=}a%Cmju+eH5zs8O3xV6nM>b3&N1!@%!c
zbPI;R_qAbR|5%fY`yT9R61z9z*O&L89hcVrl<=S;V6hFub~$NsYaVovJl3j2ag5q9
zOaH6g|GsqMu*Ghiw-`P}YL1}41G(oQHk^j_X7Fm2HxjO@&K-^88VV4K6fNygw@5vK
ze#+4AtS|jpENhDzW^aT#)Uo|<V%!>_v~AE=@Rsqnbpk!WGfC4&7Z6AP(Jj#ojH+Ke
zxMQrxl+HZh97gN@I!T>&>7a;<jhU_MXBkrt9#+qp;TE`ub;Xy~1(4m+ANFnT^>#D2
zp0abNPf7zt+rXCI3nmn2#jI@;XnEBW+{!MVn7UHMN)n}aH`{XSdt#-=@jX-+^}58<
zLx#z_K;{PvIN-yGczr}V*E5WDu#d-kCGJVbCN91Tx6t_Etc@clve<GClK^pIdq(E=
zXEgoh0K=m-@mh92Xsu&d?xbTIX9*-6;54;a1@&v?`Ga}3_1XC(!Np9eMGv*4v*low
z<^2!bGD_s5k)&q5@VGmD^}lZ|6Aa2g<`(*;dhU0iz==%LcM(Empn#tJhI&m8w=u1m
z+qx}tsA@y<7}OWtD-sA(LuKmwD<tsKYJDar?`BP-gX62QS<zduA4_@l6N316)cM^d
z04v7)bQpEL^fHImkl$8c7R}0ZUW+sM-F6bnZFrcZ`ze}~lDzt+hZ={piNx|8jIddp
z|7Fc<Ac|NBPW+8KW^J^6sc~y2f}h!ALy$MQTaJTu-pOE+AvO^XFkm*<Z2leik544E
zJvnU5VeH?qh)h{E9(1E4V&BxP(sD1W_ev2m-)dUBEB_SJP{zpqc3iw`!P5r+u}*0I
z2u%QL=-O->VEn!s1l@HCvr^|P;)o6Ni4`l)XUs?B!seCcHj>U&LBx`B@2d%OQr2qL
ztAsQ^I~&s~S4VoV{{a5~VD7!Y;cUb9-#lqg8VQ0ZA<+#(lp$(@QAf#`q7$9bqIaGY
zy$?nw+AuSQD5DG_iQa?JTl8KMJwm?ouD$<(y?=Q3zJ9;gIoG+?ecfxW^Ef`ox&{}-
zWmTPNR+Fa=RyIw}R|ufv_+au&FU|2d{c?u^yyIo`pW7epZ#9ul-h;n0ygo!-Ix(c~
zeEH+Tud-be7R&Ny^lT!hm&JswKTy^D+*5Y%IR&fTuu;){7%G_X`kM#$$*HfXMZ=E)
zuv&w^XzreQnLQC~aiA!*P_h%l!IRliF7Mg^ZEf-DO|alFj<2w+lm6pL4TLo#XvL<j
zri-w-HYMrL@mtv$Rcu#{&4Y3c)-akeRsa**y`cJh`zOwfDz-OAkg;+7T;tlq#u>Y6
zt${v9gr;P2p}|o}Yx>1~hZ|18g_07?r^XX-O!0P<@1o`oE_c>uzfw|Kq|ysmG$9a8
zV060GU`MVI^tEZCLX@<5ZEZ>FePbq%)rz+K-+fiBwF%(7aNnuPjTnrES1B%uZ$6G0
zj@1cfU5d1rq>@EcELa{jLZ44cV|NcKj1}KS#qV3;LvQjH%FZtjDowi@VS`s;wR`FU
zljlCl(2}?xW9-itYkm~h?NYolPS7S?TZ*hcw2gscBIU7*;?$sczjp<3;F%avp1Nu0
z8GMsl@QK{NJ9m1m1=U@nz?c@0W=74a-?rGDpNFv-o(m-$rgv8k{39%lYfUN)mOAUV
zmXetMoSMY9%9%V7VY6WOILM~SV+Pvdfxe86jB?|&v*^WFFkY$_EkdERaTUCj;`A;?
z0}rumUPIgotl)itD;pW5d7-Jh5bOT7rjqyHxhKN~JsK2N%H_X3>XFAzMvSnzBI}N{
z@D&JmK3;^<Vk^ln!*BL&qE6k@UH@U*p-e^tv&`Qsop)UCTDDZXPWS34+t?SFA-F7f
z8X{%muCY}?<A<0Lb=f3$iMy@{ZET&#P_!A)`;I@kFW~(>h?XT{WAJH+7~`s`JBdE)
zO-gEs@^;hhZSG=IZVvs2i@HMuv~!sFq)~Z?E4_>n<M-K80+YD}2gNC=zp6k6A3bVv
zbm_w<FoY3^V3n+10eIok6t7`z$Arslz=3QmX%}>s3eoscr}E6MI#sEslMFQ~(*B?N
zZX8)H>G-_$v}|U$hpOYj!MCgw1mS>b#H}R9{_=drPF6lUu-VH2?hdsI<_*}FSg-LZ
zZg4uM`b0<uC>-s|s1<_Z!S`zGoPSgOZ_-!BN-W;VG~QX&+It9?$(y7*?7*)zQLO>O
zmleCR6N-3jE!8xm{1I;_`(y#2r$W0#4;k{{D1wZdmwWVQQuT2Lo1VjlP1G27e}~L=
z*=I_o%=Hmnp57!@v+v{tbN16?hA9c;{pvOW_t0@}rEj0OxYL}~dK-FX4w@$j7=MnM
z?|;<#HsbB6l0uU(h6gWz`xh&5s=dv!m7{><xDlxY`3Q`*)PVaFH)6n3V9`%m8P*j0
zuY$n64~0L#0eNEEKHL`DQ=g3S_zNaHmo&%(nluL1A)uGDnd^^I4wyKy=T^WF>oJ$h
zBb}zR6WrwES}zvKveUUE=?wo;(w6jn`~%zP!_#rV8j~b$+0tLg6sm4aZA(ifhap;0
zpa#lwdf(!lXRy0-=rKT<y^mj~8zJ-n%X>6E-D;q#8r|+CbQ~FPJ++Fj?ZH&8h79c&
zYHiTEy7D~2-Ly;7@e=5!nav{cq91}aq9dm(`{5qu>y3%qc3)>t9uM9@;8<Q4iwQyn
zh6{MXjoNX1#w;9UR(LLNUqZzF2ghA5vxP|6o+wzJk7M<b7T@7uX9!Y0W)da&{AcC1
zhU;gdqH2RkAJ(8>cA%p{KYswBhIi?sTAv9>i5KjtxHPhOUH#$oRK~&1+R8RaM2kA9
zOhzKjyo-5rY~SAY^m^o!w{8I&Ix<c<G@5MFOrPq{%y9dC2sSNfmU4Qddv4n{p=Y}y
zZdGfs!zW#=odWS5n*uk8cF*tu??4EAaVP9NPYxCI-)Z@m_QD>U4hP9N>d3DbR=&S;
zSI)>`h@M?a#VPx&i-^LDvH99%lsm5a$bpCW>}sGz4-=Sp45a6twi+gJoP&6bci`g)
zI0V1q^|{wgcw5K|YnA)Lcst{<?&=q<YzMloL-!N?jBVFi)4pzxSEltSv%;R{s@%c#
z&o{F@rXGe11s0Hde;cpAVaLstwp-kNddAVeo;NIPAu_5vW*2eeY1MjW&v-j(8ZFbk
z<~<wq*j#F*U56~<<nI-om0EA`IN<7r@R#%>O&Is+O7(ZXyUuw1Y|t)RBXbF_7sc{u
zyIHDGBPxOik9(D@`6vZv*!^rHQaG%%D%STWVJLD3=i*dQoGKD<l_(@fOOnw(k%`c`
zv=;mKlk=_ksPbx2@WriJmyM3~wB}lS3zvTe<7{ns@b2>wujFsLN_Gur!NjdU_sKfe
zkMzqFOnFzI3b|<VK2;?X@6|lEJXTE*A>{x+-<>*@U@R3Yo*Pwlny6eVkYTneOtvaM
z#V*ay4GmYg+*davm`;3h+<VW6$9S4f8nX7oO^2W!%MI+>10%eY<EJZ9PIS2@4`-BP
zm&f&|;!1J;zu!~CLOz%fpSEwN)K_cKsAI|T!cQfD*QQD|?y?F?fiu%GoNvl<apx|F
z-Q2MfLDfPwvHoTc*4m#HyqH|Nt&liV_MQnk#6AYtBv~*Gcqj89dd_L{on5}35B$9{
z=<{&aXU2ZI?oIJ#@qFM|SDfVO!=je~+^?23M?Ui7^~f47VWPpBN(3)-j^1Pl<rJH&
zJp{ktvXbM}FK<sxDd{0Whh~n!8VS@4IqxG&zVWGX?+VXx^Fc{e7)%3A5_AwV;nj$c
zVD+&zCccY27Gk+=j>DHn5`LQmF2|%MV0Iq0Wd$-;Pp3P5rUYleZ<uT5D`$Vr*VaQc
z(2<=RT!5T696hWaD~ZC!^CiX;FU{@@l?cd+wxSU?$?OzEv69&v<EsC4Pvn#d_?S*L
zXL6_6|MdTRMHjP)CYk!1WuiUJZn7JSYaAQt3u_{UxWK(`TL_nvX>^Vbzh1#eW14{*
zfok2d64C8Bb6@K0blkGh9-Hwgr|eJEP0t2ru)LXoUmk}#-lA%Isg(Nf(30Nmfcr{;
zy8M5ya3e$K$-V`+z5DiFf3Hj&zqnY>(L?^aIAg0y$fLO&KhyG!rTKrZdHR3##PirU
z8FC5Jd{VhK9EE`Ba_uU}TB3~g%hIEz3P-25rupuJTbQ1i9rWqt-z%^51+M(t!};V2
z$ITYh)QJu3t+z=E8~SRLCS3M{BhM(((O{Zq32xnO(`_0X$I59a%G2^h0GUceJ%a+N
zg4fz5S9Fl^TeTAC68bJWfYr;Pe-{&0pea|J|K<l50AL%0ne~3-zn~-!i4YkQH$C3J
zUbqu8`18@nvm=v#pG7yDq1MY6%-NCUNDes06+lXGrZ|;#nKurSkQtDGKsO=SoHzY#
zd)#Ir<)1^fX&3DG5e*d`sSmZ&F2w>aHBX`hlpmmZ@^5R>>!&2qYeZ7lKZgHz-;S~h
zdaB(b1&iluQ>*=8`cC?e<El4P-zaHib3ww>Qyr?7l8Ay@NSir47Z<V!d<8#V?Q)kZ
zGa8hJn)V5M+B_GZ>O@UYzJrp6fOnOORihZ?ioeQA8Z%%0apm~s|5SPSeA-Q_Hphjw
zX7k~6Tzv|$j?IV~8<r1tG=7*Yi34!+(owU+9$gy-mWi~=iQ^RO>=$NZ9LMn3g;Buq
zXrbG8d@8rM2CPIeSR{|hwAEWjs)Sah+znU5N|NM%sjyD1!4anRtxU8|Xn2<pNV5%K
z98_52(XC$}3vPx~=p-K4^;YzjDVK=7N<1L4)GSqK7>`Y6J}lH+zjE^Z-`Fy!z&MS<
zZ=H}N?Q;EMdM$)_<yD|=9uK-eU-J}Z93s({N+-y(gyMqIrO{srSSbwmbB|bgQYQbR
z${C$bY7i!}>4&+bl(3th$0WiX9+183Tf#|lO(=_R(2So~0zX~3vM<d`aGLWfaWDw0
zQzEaUAEvgbHwYYV?g*}YOi2Mw1?$U~#tZV0(C}KgBd<VYS4zNJrqL27BQLB2(wp?M
zX5F1qT&Y%DO4(!-OvMQ(r}R@*&BAIFYFrt-NIY3G|GIsFM&y0P|7rM@D?hz6A2ThL
zv}n{j9Flhcx*M8Uq~lLO%GvegIn5V$JViTvy-rJt9E~~Y7q1)zzPS6W^FE`X&rS9T
zAMrJys}3U1!Q{9H`+Tu(#<UpfxF4@PWAV*62ZYl))sj$razBV#<6r!&4=^c2=|tqT
znrJeRu^%(DRX$f_`YMEZI#e97(3h}Z1KgvWZnXW=Q%e*=_F=IJ_R~@qB^r?MO-JM?
zQ<gHectn{^n&BpOC;D%uU20gniA%~4a=bL@{@K2A<=WUTL2B?b4wXN5TwLcjgj|p%
zD~u`^&q$KU8YVWIxFf%6F2Qduoev;VI^YN<w8E%BEG?qNWqK?}M1gkf<B$tpSedF^
zsoQwtK&dY(2M8Sg_M|Hi*AfQ?NWm%$uk};6acvd%HJelQOIGD`XgH3XSS-ja?rdUO
zejVq;&r~#~g+V;T>Cyq!^ff3|yc~GoOHQr->G1!5*T0V~{znJ9_wve>f1nxOJTdw^
z;}HYmk&QD1+Xd&#|B4*`=eNVPMuyGfzoxqUNgIsh)uJ@`rX@$lxum)A?sYxf|K+(f
zJpsxK>8p3Yj>!xzCwnS*t#Zx%z0&Tq4Y7FrJe4Qo3ns^#+mdqRDalFO@=<LPd0e%l
zeZ1RZ_5&~OYptf6>Du|a=9lC<wj2BZOTBWmSb#9C%Ge6FE;UC;qi_AaLad*Lo<wKZ
zf17XV?M!0!SsrX@UHaklRuAge80246zcmInvtseEwfSrUeJ<K5zR>hXJu{7N{87^5
zHu=__((g;UH3cTFwFQe|`I_8XF;Of|C1Oz-{*GswQMV@^U~Tz+C>q<D%hB@)>~oR6
zFDga0lVuXDQ`xik>Z^X*@nDPBY`MW{St)U}R>t!DLRLy~G+pcNk5*UXt&@at&GhH4
zb`=6=E3MP$Fpbtei>y1t&6ZL#kJ2P5#hJkG(>5CM%Q&**RlFmYN`IF)N|}oCask`f
zD6$7418;3yyFIpLxCQ>j@Q83-rz7rFT*NY?)!3)0c0?5GJr*8B&<ZFXVJP<62^>b_
z%>{Cf7|x*!gzNvjR}{v~PC;fxTB?KzbWfHLY*1Pqg=~ymf3I9;i70MVQ-HD`8E}{3
zZErHFP5ksh4Q*2XE!d@>C@lyEj&<KO-UB_GQ=n{g!t2o^8E+O2^zCw@o~p$`;&h2G
z(+vpRuvtN97@b8h!dX+dAHT-6Vy}WApFB=VfENsI>X1=+oX{79^qtoYM5h*vNL2yX
z9<o7ZyE4YCC_VWDLe6e54>kYlvck~Ycf=KP)3#Dj0t;grSqC(N1fdU?Xke@0QPF2x
z(HjP2aE?aJZ}(H*^;YdSKRE=^#2>i%eU^kzlwFSD>*YM}>6O)}LRbcjGEmCcu`Gt8
zM2KS$6L*L6NQb;NwWfW6uN0y2W{MM2EAI{cU@Bplid#spnTU-O6QBZ_?N}t3=n0lR
z%~_TWW05#p$i$gWM)oGHfY?pNCwe+WFD}_fLB#?iFSCC?057NO-$gEFhi#J7gCd8E
zSzp<qpDvS;#}7NEBr{9llHN}l9{o5<ptmxs(s8Q1A$x6ZP2AW=Pas)SWK@l(uX&n|
zs4he9DQP*pu@h66t*(t7`nl|2D9u!LkHry(BV0mYu^l9W#8q#ok??8n?;Ie<RJ;iY
zQ48C?Rc3OKAqW+E=&$40qz6)Di_-K#lg*YfAM7b1Nu%qhp$sF|vRVNVrsz0g>6zc*
z<T~JlZzcRw<~Z?J|Kr?)=3kFu?jlwebhHfCCH4&|K{AovfFK(y0HC0js~r*HRkh<L
zsj_gI7mp0#Z1*o6SsWL+AyvUx44a2+@a7h!_9YGx>87-5KtJaV9wxaz{_p!>wW8v1
z_Mh8Pc&|D3Z=aSV<_W1Xe&}5<9=Ks5?$^(xN<J+6OnM|Ge0p}k#C9DBI@e#H2TM+t
z>XQVup)rd}0kBfZhICJ<cp}`!IHiv_NYKbfb|84Bl7oV*tb?%{DD)zm2qU|)Gw+iX
zJsUehSDkBeG}E_vE{DP&1QLrHDFLN@ofCa@?@Cz}TH!N-;VW!{+SqSMl2k`V{EmfR
zc4lyLO=d;__hO?}2nIw)ROGXcb&otujz*<-G7bo@7~|N?#>kMK-r5aOZob|r_A&yq
zJ{wG0q<Yp@1`)(pg3VFs;j}g3oBex5<a0!9rS#9q1I5%2v~=bScXxLnsk{{^oEsf;
zQDErHyHr_}+dY8A;iYxx>1XTBfGj<W9W&FvSL*ct3<uo*);aYmz`+a~a26TN7~4?b
z@%uOMg7!nTRQaDv4Uf{)Ep$(VgKtijQR#FVf}|g3_;DO3&dgK@!Ku8-juR2;l8__{
ziGB#9tKCAG=oY)+G+!*{B_<Oz`bJ8AsAMK*eOQ;E|7jd=qv57K(ZbYF>xrQ2&g2~D
zmyy=VN-KpewWa2yL{AJKJV-U6XIEQ|#!Lstt9JL3fqfg6ogc16dQAfuOVXVirzS?A
zSn(0fh9{7>dI(GB8XYb(tM`Tpe7Sw28%+n_wYx`gPP0mcH{^cgvD1g@A-hjwJaJlY
z8F%5S@e}w@P_tkSPx@V_xi*W19XfnuWHde_C$ZrCrISW0rtMHb;^%@G(eT&SU8GzR
zK6CFaOSv_BX~QNS>$(^uAs*-<!4i8+1G)3bo^~g;&WPmaA?ST4hbXXSwpHA#F!^dj
z+**t^!xgV$ht1CTr;Z2x_)XS~5qf0hx!mOR{b9fx)C9Vijb8gpWY_nVLNOzO^)mRV
zSFo?5a4Qrl(uoHl=(n5`9=FQ|69&86v6ekStPh*i$gKTh#%I`8-;1|p;IMzjS~kH5
z$a2Kj<L9_n$B8Qq-D-7AN4l4gc=94<E1|zNrunf@d_BsfT=qPCiC`S(3e+-=gFrLR
z)%6Z=%*qqFi0D`+%i*V$M6g;dyKGUkIGZs>jYLOuf{Hsq>BPbEoXUx#UbvU>I}Ql}
zqg>zse8X;#me=BE?wL9RKts<OfH>NNZ>omW+;ut{Or06E;<(k4nhk72I!D{)i;p(B
z;-V4e`SeS5R*124975b6T8!i4_JlL(@y?1LK1ZBaREg5h;-GmtLIW|dT$~C6u=n}b
zOVI0opPj7S=l1bww5qH`>^DA-nUWd{uH_4k8D>g*YRRs|k-$dN1J4r`1&ZB7$xXx~
zwZ%+Fb?jt};{p^ifWKxvZ%+14NvCHYgVYPGZbO~>Wb$<ZL{XaUMRX<rA6M+sUOHyi
zGm~tX{7*5F9KhjH|D8gToksb3YwSqH9h!w7I#gnDTTJ;znc<0cfDwne$tSby>-v4#
z#glYU0wc8Nv@{b>u2c7J<=D__Nywtrb^K%__>AlL`^GSS#c7etC?O+Y(pp&fTxuGM
zCyyb(@cl{gK{Z($^Hu!Vn_BRl*$NK^b?=4IuI0Qy#P|}gK%Bk{v@WlBtWeVIuJ@xz
zPy_q#+qc|rJ5OKs^E+(OZYb+ta}uOCF5w&KP^%P<JHa{VB0N{-;;XHDw`MQ0#<l|p
z>81G6mam1|p%z1jeYVAAGMQe>;3ygG41>TX!^lWVe5dqtv7{(M`TU~Wh6gYS#&2PN
zF}8xQAJEuuZ|v?jOM1g-PWOC!7PQgi{V>5Blk63xR>!_k2id8sPf!Ytd91W^LPdQR
zY<j{F=&d`pMw8ps#TeOM3j-l;Z#Uafl{u%9-=5HF$M@73^D`5nB_H(ft_HlZwCyzB
zNo17DmiOIRr}1<&2byOrR3EgPw{Gcd<+lFU&oSA$497c9LmgkOGXt3z?o-g+l*j&Q
zyP2to@u-x;;Q*;R%_<3wKZ!#m@tpk3+=wC+>=_TfT1Y|KVB2v0C%=`RhE99b?(f+w
zOR$Tqxka_Q8B5vTB3xwzq`*ZfsVRgEn>q@J#%W+Q1yRUG^wplgn!_96>zgqQHHN9T
zMsCwj%COEa^quCqN+@`*ak6httoctJLUc)pTA&;i-W8zT;I+DF>~5<T$5Yg=;zrz1
z2d$8km3WJi9W-;pt-B<x!#_Io-j%-FquYif0T@V+l9TeQuH~N1eI?xfl#Y+J;d^NX
za`SC8>cyxpI9_8^>pU79v@2Q|Dd0(|iLw0w_Cgl&<}h9q<{qk!N1hInxljd0Y?ZoS
zz8@zp0xObaJ5)YehGp+0nbcO?Md|GBm=pJ+ip&kNdj|@kIx@n6xg}GnvS6u4u%dC5
zq55LC5|yJ_Jup29C{U-FBsB7DiPui8tfnY6{L}bM@kdCN5%-o8_v48I!jMQBl1gGa
zkTu7wZUiXmaMs3LmE`PM!-ZxDw@xy$DX?0JcN*Rrb#0PND5bK#kEqNlJXY~_Ng)@d
zA&*|scD%$ur2OZu!^o`@+t3@`0V<U8DSS*OmkCe5sME=BsuX1CQ||-(gicg?aCFVB
zH+aS-nWj=^A|DMTbmlDeZUBm3El#39F{O~aNv|u|60X{t@S>v)?id7%5HfT{$s0MM
zs8m^5Nk3Tv$6fBtEpT{z_W;BMf$WKH6n{}32T%Vv?r?NwsGXjmp7;I*?R5k^f@R;E
zNXu0KAeE1qG?e0_BWo4UvrP{d?L0+$wJuu9rj6(DM#s}keb0}#lCxu8g}3|Yc++aj
z*l4-f{3$WysWJ9aLn#>a?jIuF*&HGww)_GzpqZsF%k@-q<CyHIT8e9nT`DC)mZ0d;
zmQiKv(E*dKlDiyAMb0iSiUOKBsucup=CCI4?<VH=Gz=NBDEzdC{$%2YLQUuprDL;<
zctX;S%>0yK)Y(?uK7P8z`L5bPLYRb8H4g2PF)EV)Pcy>~J0?hrNld#47sv_<xR*y#
zN|-jIjOqO@`J!tOIzu}8Bt#NHXOs(R`$MZtCjG&WqMGLnbbO#ant>gT6R=u&8;W;C
z+V)&lc{og6qh`uXJRlXQXDedIGwvwL%7aLt`?Vp8FFnQ<LW!PbG2Wg@(ocK4Fhj!1
zU6S(JF;)x2;$SAe{yHqYXeJd!1Z@VmtV`sJF#!a(<s(djN@RbJ?S@ZX-q9Ggv+XSI
zXYScR`XtFn8>JT!`Sj%rEtS$Do^|Ckh0G2i%!#8Op|p52MmBFyA&_5peF}*a64iJM
zchPdaKKxnIA@MQI>#v3inJAEVChm6<f)0v+1^0z!N34wkQvC%~JOCt_QRz&lz(SYJ
z4|TTn(^>}KB&C+HA6R}$VoKTq`K}oO`Gxp8g))SIH_bR2gwXe-TF2nCM0_-keC90A
z72s%(UA5<+Sg#g#qz6gO^?CWlM-x@yWD<S_W!{Shta<#wMyuE*1GQr!((2Z?SPU}>
zHa~Y)26LBAM@2(|`OAF|=WG~wr=gzQXjXma7<aNRucer#Y5HlPUkeahDirK0!#~$<
zB__j%X^UWo;Y)CL!-n0JB;QLW=sMn2p0MUni-PS~4Q;u<7zfbDI!%f%gR8N{R7nTb
z3@5BM*^|-0`09A|+VJtyub=Wq*t-zM63o9FG}B~~c&Q^99Pn`+vQpG5{17>BRbb<L
ztp0(KRiXr5V3XOE;F*-wh3_fXe`PB?Y(I=Ql{Fv(H09bTsb1I|J+mdT&DwJby>`|2
zV)5L<_{%gE8^JU#5|rV>m3x@x*w*G(<@QE`gRCi!#_*{w19nKXf}ZE5(R53hjpFds
zxLMtxV^q_!eG8{VYtVFEIMp6hi^Hgtx4fQ0qKhhZYnHo%qi89G6b{M$G$WHrE1Wm~
zZBj|QA;gi(4R$h?An6%V6AZ~Lfn}a`rScU5Mm$|yn+}XSTQx*Jq*?P0C2?_EauLV4
zN0D1;g+B)px9WF8kgnb)yjL@iqO?;n2B;p@F?C^^$XYLr&DH9o%Pejv-Ms`$j@cR2
z+D{!Z5gG5lu)&_u@6@|5jJV>4KdNWI1zet{S2-Q_)X2O-Kp_rF3W}VY=!a^X=u{!4
zP1(O!s<pG*%~2&`m69EtY1xr$pVjphK1y*BFzmPX*}LMyd;y>&rmh%5u6@n^ao+0F
zkteX9I(KNSB+H~KlZl2J)wEVdmE214qt;xL^n12PX>tWQ@&i~3yWha5Y?cjNz$;ep
zn=ClI7~D$Gg$!2eCq1mQ)TkJQIrG^^@g%hUcsBgH!)Wg#ra-e~GE-=X@tRt4JfX-T
z6@Na|uT?Zcm%H1!dfd}$AY-_;9b_-+u-toCK>Wj?5nw{J&HxpOUzgQkWREv`bl*VM
z(m1bhj*FL+*-v~&p{U!{Ro$!HVQ$wZzE<OZ%~)Y+G!7y~Yt@e%5RfZy>(M9p)%?Ks
zqV36ycWw8W;|Q_I;l^c=sjX3=AxVd}dc(Bh`PMR1eTOM4w^LS?Q{UcvsNvoslBL-N
zDL(;yALAZD9IIXc@EJRQvi~jW!QN~I1&IN*Qgbl)&%s+9l7^oam5k1OAz-i*Yw4kr
zfl=D>Xw}mN4@iv(n};a8-mD7u5HdX2$kdj36zATRMtd!JF{!p4nIp?;8t`y2p~zP|
zOP@z_+NPOCHutclFjFM$^~fT8sl5c@n3#7}zZghUX~okx%KnMFHGsq=VZLicNHQ$<
zDBiO^^)gG?3J^=r4%2&2n-9UMYK-VJh<HgAd8vXEZgSHh5=N^{qcNVwX%jXIf3Mt1
zpj*%k@Ft}QKPy2Lp_@=eN4=ni`#4i1RiiFOZ>C9zIxJ^R?dvS~5h$V@z#};0_=5Gi
zoR7067)c^_3r{3@(|4zBkg6+p*UnYRtYqPGdbExGtocy}#WOWD3egu)R%g1qf-pM~
zi_4KewekrN%`(4e5$;8GCpnmOx(;nH%`}V@uzSM|^i8Am?epuh{q>n_$tlGGO7t3i
zU41^%;8SHiRY_SQQ;vcJ)Wz7*4qMnKmY7pkb-9D1iA8|H5zRj2^TflDWU=g3^kz+F
z_Dsqh-{pO#>rI08pQUPRLj4SrIO&K8L{w!Xo%pH^W@!Iy1>P(fT=`Q7FvfVKo3Ae@
z8E(Mz<IO~RCRi=MQp8rO;l`<>x1+7lPyzI*8ujdc9Eb}9wx*Z%Q*eeoR?q{O1l<u=
zk1X2)5A_r+(-^L#`b0gjwZ@%|f|o-I*|9aLCij#xoNxO*GhO$j3A8GE<PuT7?mmr=
zE~S{*+ajtaPDAW!XO?Oe7K$Oao>kKT`<jeK*Y{w%w`(Z}Ssr@DnwkYVB6lkbRZE7J
zdYvhhlEr*|c3ZU}*QAUDz*qBuLv3k;Fw=@rt3}Oj-&&rUTi;WwzAs5UcP}VYR}}UY
z)8BU5rohSgZTUdPjZK^(w!P?;V`8cJQ06*xsdC5b+qVFt?>}l;L5beGrBvp*@wyd#
z23z}g<Fs!IAPET&$B6_qj{*8qg1QvBEA^s$HnAERQ{9zPCTQnh3ir}HW_woVy$7~}
z#?EH_cT5lDYtmjo4P0|umf3Z@XELyJcQX9;T1tx5@NQ%ad*1U?mqlmpA1ibfkE(?i
zg7I%Izia-zV(LAbrywW!7<OD~@jL51ol9s)9()q*b9=rs;FN0Zv{OClgEyj$bzYoT
z?X01*9BGp!P~Hu?(|M_img9e~sK(pGnBA&uq`#|o0)2HR``P_M*A#tn`EZ`(W76x=
zz<-UUK%_jB5Nfp6bVfvq+jbZ~1l!mYh%2h8nI_bk^MsfW{M&(=*;|-zUFeOs8*XGj
z2(z;D1^8n7?s+%c7)u`l9wTXQ%qEN)=;Xl)6&LZ6kq>1Pw0lN8j7SCGdZ^|=CPKjp
zA^?@Bx7CRm#mtl40sBiQTfSd@UkbX&b_(D9)}-7s;9tcVncf<SRBbe`6@1j9O3EMK
zuR1GKFK)gX6-m-WJo11-{7;wLHmS*(QZ9-SDQXcKT?GYzf`SvQ1MVyE+*7B0(9Im1
zjmGP0*nvOr-oq+29k__w4@v5>&O)5GE^Q_S&~xP61%4RC(LionnX^cwbILXfHhGSb
zSN>?olvDD?bOP1Nd_ybBZFqwOXd;8G2&z7|RJOP{U!RB37wrhfrzaryc9wX4GiFyv
ze^7|r+p%U+qn>l$$JgHd@BKpETXA+UMgf=r(*;uwZ{705p%y_a3f3gl*3%C^NT3_n
zN_f1}iPeFb;MTjfxY=4>e3E?hvBN#wcNP}o-EZpNiFEQFXCkgW_P1~EG&90=BJJ@r
z#wkJC9pj3=)61Xh78y433>N)&lhq41Lsxe=&wl~$4wh8L-<X<nix{Mmh@^e~UdgzS
zu4d29@&0?|zh;r3_@%v|o2e9+>Mck9*1@h|XKd(pc}D2P*vO7c+KEhP(yz^eQW7Aw
z<;42M-zzyCI=L(yvVnsBH%;%=r-`r!N*<*7T^={UT!THg2G;gPOnfXFU209x6`hVr
zO?fWxjwY-z?E<hH_5J7CHFg*0&Q`*x1FYsIhr4T6+{;o8I|!gezK*H7kiu%JHUlMw
zNr5#)H8%cLmW;KPPsQ78?H>hOId$e>tT7K43s8lHt>MI==Pe}{A$PXQ4-p8&@$t3c
z3yV9wA$m}Zt;dHrwQs5MxTzizayR0}o~qLV!2G)v7m4r(5QX+(ZLYhckfl{dN)^Uf
zn*VE-x?bK9@Vk^NL$+8~S|9A%{#Gq^%9h?>e~2E8wZdsCW8Tkx_p2$XF1ygRLre^^
z4UQGrn&xdj65~TuI3a{H55yRs606458qS^ly)JWbUR|&kL-O3*Iyvd6IagsBpUnmz
znZ7iXFhnFEx^&j2U*cI(;2iT9pV<iq?;@xh(YO~__3^@dq25UkcSgEB$P`wNbx2^9
zy)CR9{te2(?)4P&Af>49+$EfMquwJ&NWDl)DXFAeSwnpIJCt>-Ae151IeM)*6s{59
zn7cwQU*59g#=*rNZnx{=#OPf^Y^Q#5x0w*@#G(Sl*%%H<abUFqHxv90FkOCyA7JwQ
z_)WRbm4!J4lSgS0K1UqQnr-G8PyuwQU%)6*z6VO8*W?On0QS75-g!j7&;XO@SU3lJ
zq;BPP%9;*aX;?YVD9C76Ao-@nxXUW}SXj#^?pq&OiT#e%$n5>2S=Lcdd?1H2YOk75
zQA?-gS)oN%0~IuiwLafll&%h=OQ}_Ls<s|6=Th<g*T^6`J-3rcOmB^S4w1Y1DdNtX
zJ)wGAQ>`VlS8p0<>r5JOyV66=I>~Tt2h@8em>rg{zPnMx6v6JAQ%kX<evLm!GHQJs
zTd)dJH_mH6!uiF<3DRTBy>2^v-~O~%B#;nYX+}du>5U5+b{5DO*H-<8&~M-i^^4V7
z5ASGWSW47wG$aI{+QczE1j7R#9fP9&HDfeBDP9ypc9quFg35juhqPCGFQ3*T%FdF3
znMJI#EvCitrPUjZ#!k4TJI?&c?RAg51&08<PaWnQ+lVAVz)7<zFY55D$4u+GO!w_|
zI$nETYiyHDB%S`C4VWzA+dMp=3ifnfR#gD<ib?C{P@`j)J|lgUCHQ#xo&pP$VrZa-
ztL6TD?K<%;Q#d@{E||^HtXV)Zz|QWM`}_KH&+qviVN9nf+WMQGiNn&S0r?zo)x$e)
zx^m!t9%*dFFoG6c^T*Ali(LEC(cAT>oEo4q&{O#fhm=DvZw=&Uf)UeoWRZy(^f6sr
z0B;iF)%J(@Y+HcGtJh2H{JMnxo=T$4w8_n%rg0_>+tH(jz6b;*e8puHA>*z0oK4x)
zd2;Gl@zKyT&-?qLx66#DL}Y`rte#Dl<IJq6P>MUjR#sf9ys1)oicZkNT3xzE-fmiQ
z@w)0Zk$#<Qp$^K-<<BMRS()m=Qi2smjo%jttt@k^g^5Xs{=)^1-s)F>dt-=w1tQL1
z*ij&qa<rT5mCVLAw7kC^CT<+1lAIm!%IgD`h_!#M5_88`5^inSrg1NU8OmUU7-H<9
z3gE8MOOSjPIW#`S@FWY%1!Or{(#&7o##+z;Jm)KroZ%O1yj=$y12yT;9!Om_T7Rj&
zxvkC3i{xz_62`5DKFkqWd@LN|%VDJuU>soPu#2&I?lE^kn?qx^CxAaCzN2=Lj{tj~
z=DVMNCLybwojX*_srFOwjghEUj1VKBEOvJcpw-vcN5Wn0>X7BedUJ`??bmK(VdfRO
z2W)oPGnY~&I9wpgQ2_=U$l?(vaG@!~8qBd9wS~&=YruzG)BRkMZkIPyHMXZ#7_4Q4
z+4R2H5nPbkkiOU23dJWmKA421;Dq}GfawP3MKY0&H}-Rh7owVBK4jojnKouK1es82
z6U@2Dy!c4`!6s@mXy#o{d*r&*Nt#lCx{X?4Dtr5Xk;GcU&{ihk>4Cg?Tjpobl8E3E
z$S31YltH`DujS91k^xQ$FRo^YESAd1bUm0))C^wbc1zaur=Vv>f;G7n{ykbi2@CUw
z3#41!qiA+{$$B*_@GXKq&5I{`SFl*tZqS&PcSjORBctZqBeyEWR09Z++z@eHpuYQv
zTdMUK_Y*>TmqSN$SA1|1m0HXtl!}1f*Aw5k>vf9Bo#!uvo;%Xfb(J>m1efNxzPQ7`
zk9Y6dWj$yN7DFSeo+Ck3E6J>Th#&1cOQTmCU}C-BY=6{uUMoCN(8<Qde9%42;j-or
z)j1H`@_)>x)$Qo^N2sb=vKn%ghj|EN&e>SusHugxSiL9!vCjLc6?#UJ@#K<RktBQ0
zLUo79!wheXx>>4M>|r`8#PKZe!p*gRU>xz5NW3Kdu1)!2?!WivYA;4%<?~!65w+5g
zux-FnUw)^C{2mmb-iWD@L)hyiu#i;!^4r6sV!bvs@mdtf>2;MXj`cAJr*)H1Ja)^o
z<}4PFvK*$D_WmXhOHm1B9_F&7&~*xVc2f4JCvtCdh<#?b{cEOH>5o1^i*IiP^;2)7
zHU|OyoiX$eH4%A<&L2#~B8kz7Q<S*-MzEt<eCf;x&d}C_dXInkrOg3;N-~b;G{+Ek
zu4c(}vP8VfaU9D>E`l)!JS8RxT-r>DzcRUvq)?KW{ewSx|I{bnA1l2dd!~WN^qK{Y
zJV@dKr`b@XRSe6lQ(4zQ%sf7jwtcO8eOffAB;$7-eX2!+J%i%cRgqPe!<8m<Pn)~F
zoh%5pp86YJ@P1(*AsLy<rs~4j*d@k2)bw$%B|$&RF;!$HRmz63tw07eW((Zt&%exB
z%B&$?M=R@pKzl2w|F^#MPp`w`Th{kLcmhJBdZ+|rC3zEa>>5LRPq965fskZ8*K1_D
z(wMzMwu#8KZVD7j*h~Z@ORh_{W0Gx>U%!k!?GH3z5;h;>o{J=JRw8O$hFB?cB}T6T
zN>tXy%sx+T{3#qTJ-5R*y5{`7a_uXvccqWngs;!5ir%XW_$h<9Y6X1G&4q$)xN)46
z6aPMPB<fViv2Bx*zFX|9X^V1Pr6hBvA+Z_;xC3}*88sB@c9+u~`6j#BBvu@W(bIaW
z&P%#>z^ZsiZ@UC;UwTkA3b5(ZlsQvBPUL$k1KsgCoVEu|7G_@9H8AyL^?e(Wnl<$W
z*6Gz%%cPWX_A%;a(4@!C4reB46Z-jvH8<?kGiX+qaIIX%LwAfMOK4EDBG<|*gL4XD
zYm5D_|G2<ELOKG|W=6R~;W3$@N~$g~8du50&Da&0Mf*i-W7xcqtJ~S^!v-m`?88`D
zM-k%=f*I8s-|3^V(y!09P4FHJ9drw>tbf19wc0vUvod|Vigb0~@eSbwl)(bJi#L-Q
zd)PWO<+oVLD*xWawRWcc`B|q*3|}TM!7rp==%x4ozI#-xTmZj#($$>sR#56R7NcsN
z)bBJp=%pIrrQ_!iH(uogta`x_s#AeZ$|(7<4T7`^ikE9tKYQBh<pcmYk;<$MsR_@A
z%{|L5yOf0y=fc&@gB}r(QYXZGs=z$$o~!~YZr7G-E_cu+nR#%dvFzy@rXPB_V+!D!
zvx*t5P?SPzG5h0dT91^{wd_hH)Y8J_1m%;(6D>OtMjGyxD(}{zjMTfA3H&x@oe1F|
ze0lWEQ~fkogeMIh1`l#H$}p@N5qo^QT(dNBkt}2_#@HX35FS(Mn&1K0Rs=;oinC~I
zf;WrVuE$P$Pm_;u7Pn*6{YS-A@B_FUMA1l*Joqey#U`x7zb!f3rC1B-cb_*OqFi9f
zs!JDSPbn7ka*%YWX3%jmdvTThzE42shr#ayQ>+c_9m|oms&gTa1M*oLqNt?Y#NjN^
zE8OSOF8!Yz%gOPT`a*1XxRW?```5?0b4W1BS~d7kE$o4-ocL@_Wv_5OV}b;+hsxq{
z!vBmDQLr+$%%>@y<e|tsW=P=WDfwfrBvq$9vwrzti)X&txv1FKuJD6<$CK6vc8dyo
z$*BfHmQFHe1X|(epJtX_w&#ZMrPQ%z;4cF`u~l4RQKiVkLW9MZ5j0U^&mv=3EDQLm
zCr;hz@GFoO#F*(XI6`&-L;P45Zzy`N)eIxTFe%Q`_ZYaT-5a8#g}a(+r~GDf_95zy
z)iaWYl_;42(dnK6j=fUP7wBqEj`w;^ewp}nm{A_0aAQJ^&1lapEds<*RGR2r&}q;B
z;+CuG?<yyMK2I?J(M*+hyxgp>D%NftcDJ-Doh>Xv%l&|iOcNRPAJVAd$*evNAgbxQ
z+6VWY?-4A%%YT@sfeRc3(1QFx7bz}nX=$9IHFYzXn!##8{(;_HwGW>^u$<SwF7PIk
zKni^BVg<nO!6CCA&8HmoF&AEbL!%(1ykQa|%3JL?!GHiYkmDkvPr4uMiC9$kN3y)>
zy#%eS&U_-u#ZPoz%O<?5GsuI+85Ea$KgBB^DWnt2CFVqW1-x$uYx2r`RWG62EVC$N
zF*S>QwksZGmWA!NJWAOve$x9j?xvirox&2rN}+4rHMY)rhGGUsAomrLKhz}4eE78D
zMx$J$HQ0+pTbo-6`z029@$!58PKGYB^e`fMh%q#Z`pT>ZNsAc8wEZc!B38#fvK9*$
zVe=d3&{6b2l66871O>src@fx4)&htYec;YzX)6v=N*T5ZF)ET5Qy*63$Y@Q)f}LyZ
zYvr5T81?+nS>Z90+Fua8My2?1kVhcGcvo<-UhW}cAK_!%tmXNbgAIAF?N(R$dyo*r
zg~vUJUFIY}fqvV=>^X~kv(&5^i1&rNVTRHdIf;j>bO0nb{0Yci*H!Hoe7W?eDfUy{
zoL4W*LuLI<>=7*7Ms`E{*xC@#&03x)bj1GRCKs6zaJ4<f3cpMVJsS8hZTs7}U9}&W
zGZ;wmx|ph-7o1Q|02&0si~}SrwPm6+@O8gy0@Hxy-Xn1XacgXf&q=wChW5~Qh~3!7
z(N9weup*Gi=+9c~u~ArJ#(nLin|fc~)wrS`^!YNyftJ1m_NoTR6g%K~{i%HLh6zL$
zx<Uiy96p+CLqOneS3aG!7TPOYYi6zMesH9gSQ7;s7G~9!O-ZRKGS*jDjS^V{=qSqa
z&aJze8ET9baDnhUb+Bsxps6FDAFls=_ZSbc9i$a&yaznyjO+_IJfpMgja*Lcv2A!$
zq?)ATJ+TWkV}GY_3e@Nt4r=y{KfFF;6w6;pVtb_b9Z+34ExB52VfLwDvykJH-IL~|
z%yAII;Z%3EB<auEzZ+QIfA4i$27G2mmyOu`IR9K-U^MixFx7A(k)u=Dpb`;^G|;vE
z1Dodb9(z-;_)WWJrQ>T)lTdrHa)Ss#X?<^|VuZs*M=R=vJ6F{-J&$CuBy29Y)4H~p
z;kH(SUuDEao+V_QSEzo3anPE%OUMGqFFh?xR<+s^Gq0RmlVvlFTnQg5G%t-l&pvW_
z{`U%{Ayb#3747H9PiMr0x~i<vHC~)s=FA6bE-1kBs0*b=d85qkslw9T@0tt(UFCma
zmdeS+&zH)d+bzXEU+f#JV>iON0@YCZlwyH?_0nxr8yIX2r6Yo!p!8j1Bw=cQxwws~
z@?Q)X44w&KnHuyV<C_N%1~LfK6^{nQ00MpiDXT-6j=M2TNVmWgTSe8~%HEiXx^g-N
z1-272XFcmcs&H@%=k#05&WD?>JJ+yu?tJ@HmkIPtm;P46MhU0Jp=izUyxOM`E24)I
zLs*j8X5$!@7wI1;zmmZH_sU$<9E+KWO>b`_+H|%r=nPz!S&=dKK;VBzBY3q{^TJ%e
zX3=JllS^;yl-z+ZFp`??j7ZbEyY%u3KzHx)p7e`E+o?&^>Gs(>$)$YvBNXr#d1dSM
zy&6UNf0#Bfuiq;1emQfR;abgd>AR+&_X=EPcugTFM|I*BW^b-Z>FT?p;7b2X?knd1
zJ3IRSpbER^g?s1(H7?};s*T>=T;w{KPki#~%6-Ftt6mEgm7H~kX#=wW@A?M<JPn-o
z7F}FBx^kaiSN-;#@aFUi3+P>@IoZGA0HjFeHMmqAKIIVr{4ed}Cgze++kVk1UXNfm
z)ucJC|EI`nHJR^^dnQ4GhPX$XU%!`zZJcr+(Rmp|Vs7y@l@Hu@Qh}6yTDChh4xx~H
zt9vW%^ZmzM)}$B&>sd>d0SY9j^7K|$jF*(&cG?-_UCslYZZ~{Q)Gmv?y#GzH`M&;I
zJ|<$aHKNQ@Tq2_qc}_pxlJjf-`Rr}>sGyS-x4Y)ETQkV$X4}$Bp8kNcKKdW8vIuat
zXmGG93q87<+5JgmFCi+SD@LWYn!)i^f>ME6E`mG&E*!Da0nDTNWD!FM*GN46$VVg8
z3Xb~zqpVD}SO<MvXGt62_{qhlx7rz5+ucA%$MCkzMMLi_A7$UxcLd^e%*CP<<tcUn
zKgG`PvLRR}9LB_=><VFsGB5;1DvcZ~@!;b8Iw^$N9rKuVtuVn?FdB$dRe(${?sjb1
z7_%mK=8_#ktbYjEnP(JY&QrzfJPKwN!_7oSAwnMWm2PuUhetpv5`T|3O1CO3n9LHq
zIvi}OKQ)}Z&>H2j$=NM-Ro`VNh>~5_I#K7e)YGdTh&*YuU}Bhy3EC^mICxQeukVn6
zWG`EuX#(jOh+n7p-3xewRF3Yzi%JS?=-Y+johl?iEE#9og9+IlLYa|qC{P^Ah=}5b
z;=U4Hl5j3v5jP97PZYVc*JNC&QHBE^D_Pn2azrl*UvVulZJO)cYhw}@^ghZorn8xO
zC8;OMiHlj*uJ7|ieVU_pTAe|<)y(PdfPm9XRa5ktMeB0lz{5T*QU0zJQ<tetnIk~%
z?&tQgVCxy<YbYv5D@ZCbl7Q1kEJx`jc+ZX|=x?md|B2a3o7{WY>s$7R&3>`PbI;Ko
zAg-``c{v-Xz`dR|d~RT8a$-TuY8(iwgz2AG4D)YWfNE{?OPjd`A~6WE-^zmXu^+RU
z%<o``mTkmyoZt%?8Yk)K-@E$<BGrDs1!3;iNgN;02h0{mK9A<e^KKF2g!0+WYnd^D
z{uj;n^DUWp%?(+ir9z%r_CxVHJJ#@2_{V_&Ax~?mgG!sPiG5Vl!L4Ktu#DTMExRI?
z;#P4e%Y#2<DjJDxhmKP;<_t?CC0R7p4<_lp&tJfiY9H3!Tt+J*=wbtasUjHD@`%mv
zLX7uh3{r^0i?#O-V{%VLJ-sLGEBAA|VtdWm7MlA#Y?D(x<Uhau*-9s&5Iq|7x*d18
znX_c{R;#!xNgFGz`QNN!WNR$UT8yTSL+f^m>_+C;B8K$bE=cm}pHmpk@H1N`#yeE>
zjf{Z%lJ=aHs2k%@;pTC<=O`PxkOfbHfLNn~M)kJZA7KbFs3gEq#5b?$A4-@@ET`@5
zPfvs}Y|u2*-4mGsL}PYi;+cv|%_JRNF>mYQLO%09=UdD!qoToE8}&cOYNvyxOn$Un
zB{68N7APL99Q?5sqU&zqo&W8!S?au$MZ=P`o;&a)-mI78hOtZa0IQld2M3`i>2>nu
zL7E;Tx-XPy!O_bXzkc;CdF)mZe@)@)atCjS@Vm?3Iz10isE?WQgtsWNd$#!Ici<a=
zZufTjegDWXqTX*GTyPPR<NK$BV&e{rE+-;w%MsIg^3AUpy*3nCOTJtC)`mWKW%vHq
z*+sxrIk|%Qc;5pqVqE|>1Z9HEF>Ogs;B>u8*sJuIbUMmRDUD+XEe0KJ9v|G3>CH)f
zZrAW9<Vk7Njb1&y!7g_~n(ZIrFPW@B`?)OX6YmV<PoK0Rdt%xtn`-T{`L9$4)EYU2
zn!?1zHTS3Sl1VS4E8cY`nS9w`vS|2xdke?g9N5)NqJjsV8yQ4`BwBz)KKSBM=H9I&
zV`JwSy?5ZCuA1YcyHoaP?-@&%2Ih`=y$t2Tnu>;BZ+@$bq8jB+L_(OBl}#LrJh&ES
zg6H%i<q#&5p35=SwnW#ZJ8h<WRzF!~-W;U$SPL<lH4p5zl)}MP_;M4>T$JnLL(FI&
zI!JPpMI!;Sf_AlN>Ao@TRW+>;nRWaI=+}5XAc-fTB5_3{Cf8TjM0C7u7|(a8`|ERD
zv*Zu+gyfey%xM*r$Z6oOBWeIE^+ZRsb@Pu>|7{NhzfY=6l`)}ah=yAtUerXn^wxcz
z?s?EEu&jAN$4o_|TCp0&w8OrJM6lM$XEq;O6b~rF(c+87%*Z0XnF07@>K~uequC@Y
zukdM0<FXii#n1B1B=+yrskqVeAW!kNLTDga#p>^sn~T>5=Zp&LHHF~F^<y0x5!XER
zk#g8bO5CN%@;k~%IXQ@1#2Obe2DK~_e#&2#QyqM1YZNpTL!jNvW^~Ctn$MVeVP~jk
zu>l`VtMcf6sqSh}Aj?&ITx(iqIQ6G5k|gmxLfd#00@!IOG4c15dmRGH$k`}8YwbG{
z|2S0$iHi4&o$i0G7REDMZ`<wM4;TVNY*)u}*)_qo6d41LUyzg4yzAmwd(>RN`k<3N
zg{R`_?k#oP+#>TuDi#bKJS#9j&DOgxc#dlM&}wJtu#jX%!b?01K!?;-&${Vln}1qm
z_y&!gb#a8UI0O{+#olY2lzZcu1^Ht7vUgVqQv37sd{bEuDNVQ0^-$sJaoqyqH6Xhr
zw-jb9C34L&)eW>%)@fAM0(TF<y_Bmz3zD6&Y!)*Xh;jsQ@knBlD-<5QQ({T-UI*I+
zk#=d#U-x-a{bZ8qN-3sdJ@GfDF(HqbIHu`$93Oq0%BwJAyvYkMq0qVjuGdbNK(T2F
zxa()xsZSj?=%+{wMe|O7YMGZd#u@H|&cip`<f<xzt!PG!nq1GLtu`5ZF-xT{U}Xs2
zdOMT{yaAven}jRInyfEX#j#mpzZ>5s+6;np9e;e5)@eH4%dX?liOA5><f+;J9;31{
zxU_mV5bM()1QQJ(eC1xGrTKy#Z4*kUo(nDKsTs**7^zFRFUXn}9+P+UX{VNoA&4_?
zY^sSjkrBN}n>F|o_Uoc4FgHp-DudhLEJh)_JH?%H9;~3UrK6BpLpFYc?J_@i+kK<R
zc0bK?NEC__q5SCx8Kb7906WZ0$o2Z&X0|0Mq|?sByskA1JB?i5=L1qxXS=!NqoAqv
zkfb6L_vxU|9E4WXjzkD?JbbZR)js!wF(>WyQ2qfKIfygD^u%_((V%N7z{SLG-acX?
zRn3KxNxdi964|Y+i7!J`5swQB_VzD6|9eoT;8emmSVqhvCNwYF*@?w4?i<|k_<JRA
zrVROP`X<k5f0#BxTmp2*vQ95eGrD(FZ4jHv`dHSvv@%ow6*wY<6@&d=us_9jvsLn*
zicvef`+3z-^nq1XKJy$W{k1VG+vH?gn;Q_7Aje4U=p*u)mn*@vG}j){F%ex6v0-Rc
z2%@!0)d=G6F8JbP$2Or9)O52gHpaL1-mQ3R280NvdE5IeeY-}rF+Tt_K20E$`Uj|*
z?$#ch`35KMAA!l`mxAL-y&?&^e5L_0qoZvdNUr(AvpuN{XE@@sf$HHxXv7U55ZiaO
zP+@RC@$3vlJLDJe=du%c%+EGa{~@2h;_89>i*)n<!`_?6L%II{<2TYKm4wnlN=h+p
z+Qg{PA|^>r4W>>>_OdUtNT;$?qB2U1$R5)X#yX=An#yjnPFZ8DGX}HzT|>*+-{-Tu
z-;dAx_kGVF+}C|y%WHYQUf1isw!5|kR;8P9t}?=S6p_djrs!Tg^jUGx`C)33<=qZ7
z?*g-a_$_N1kJ?4iJ9&|QVvK!9x5lfN7L?^dQmg8*s9if0sP8**+oS7vLM-*+^6}f_
z_~tZ5XE{ts9HpJ%B?vLRr0eUeYx)$Evqwkw;*s_vu9KlX(hgIQgY6cRgl6mTmCiWH
z;XV^5&k3LKqbfRwskH~^rxutSxl?rt=4mAulpFhbz(NAA?|#3dmhz(^w$)>h72-j~
zS2X8og!}Y5k*r8X6$g!F7;zXSC25^i=xbTXRFrNv(xsQh+WjCo#RV0FzO7?u(9fP9
zCRe~`qQW2hrsY#x<3CP8X8j6Q<}Ngxmg1dH{e%-c4D<P`(8a|3%e`**R5R7-59+!p
zrFq<kVOKPKl-@O3Bb0CZCi3Fq1~@G)+coW-ny@JS7JAmR+nFDu_ch;mLG%+W)XmSc
zW+I*y0*~EDft~2jXz^%d-dm^C$~E+O97XFK-O)(AsbM}TSeljCOhW}&qwXP`M`$OT
zlL_0;1UU&;pgp-43>gL$bydk^onu9-ulN~IEMG8!WMni+<gUzwy6B;Y5s&Q>7#V#N
z0;R=WjwYT}7wdU$ACNhXMXY0Z%I$Qz>#NgFY9l8sc~U*~F4;Zpg~yukEzYN!USQFC
zlAd4`WyWO|FswQ4<78oHc;5qjfD_`8pb~q~X<NLO%<7wOsnnhx%kp(QQ}c9mSL}Pc
zX%K6v>x15d^ob0eGbEAi$hOSBcsywt>9TK^NB7fP)wQW^YhSrw^pPoatxM6~4@WfD
zR_oTAHc5z?*;sbb9;LrLmF;S=ZkdU1lY$mWBiOH{>CMW+9~$GgG`8K{FsjuWm~+S=
zxG`JK3*{l4JrEzO=H>q{dcj(w`Xt|%)%vMx?J{BeyN9Wh9t|(oo`qJud@I|GZ9N#^
zaaS2^K8#U}$#?BvOKag6T1AJ`T8d7zVT+RpvWiXN5EPBs?{fU8?gyZA8GqQni0*%V
zGYj`pEK%d|c7OG4b%mi}fg#wVQOB1CL*j?u=FLqG$&%<DY$%qsMY6z0OT>Nb`j|mE
zLaSn>?abh)#fzcv`a*tDeqo#jzQhH|yDDW8q~lvL>1Ef_K9@w9MXwr|t7*^$3p{@K
zU&Pqf!7M*pPuaoc#oeDgguKRXn#)ma!HpkoA}++FG~v}h?A+^i_T5Ipk(j3SI;Uc?
zW%iZ7x>VbL_^tE-M!WRtnh@Xgz0yB>pEdq;_!ciPsUf&OrMfL@*?r84487!%gJD)D
zc4%c?yl{P3W<#xh?3`=xIos{$t-h`#@wNb)vKdb>@fSORURjj99yjlnG>rdLnaz?_
z<d#}aa%~*B&0>>#{19xxd64_(HJOti7NIoB(6tda&sYzG;wfljhZ%0IZZmViB+;vn
zIZNr<!)z(8f<)nm`}8;Eq#xd=I-RsCXwcPr$_Fi|z1N&G1-U2Uw#@~nZOFn0RrvYs
zmbtz#F6sch$2BJw9CCXWH*wT~wZ_S%Bj20Ub-gjCw1qAVzc&SKf0#Wt8@{wy;B&b&
z5x$)@FC~WO5$3^cs@liuCUv;37;(#qj)-2jPgZxI@$P#otKOxlmKI4E|A^8;VtM(a
zYagc-+yM*ExF0487QO6ihP$+;*QaLjSM5GzoKog-WmIwPohMCOE$wx0M;R2}3Db9A
ztYajptbSLsUN!FQEf14Yc9QX-@bHM3pC-3G_d7h~{`Lj<e#6RqN`e(8CH#txPlBa3
z_!NW+ZlP&W^U|(XpWQ)A>{T+CY&&x<tTi|}nB=<MT*sSm!tLDb>c&Pp=g{j!<*b_0
zA60c``Le1b!nalPZaU&#n;B*x^QArTIYtPJ(@p}Kx-UJfVQk#H{I>lWtyQk?>w+In
zCLx0tom*YAMCWMG^&T7}#6&ZL3BRS#e8|wqW(rEP1EY#;tv4xtj#X8auG@t$a4adO
zPl>NfIB@NQd6xN{SFI&;qzYvEP&~r}#4Kz`TQ-}bcmWaa!nmT)#5z2W9G*7?4ZTcZ
zPdaYYxS?$TK3n4&X&={JvSdTECqc_?;N#AintT{aP=wU)Jr+WOM|7RQg^b^Iaynx|
z7kqfJ<5P^qB-%Nm#3}A6fg4tGXPDDvdKet+DK)q}Ub?8r7A_b!Kxnk<8$))II}aqN
zpWa!24!oIf;LS!F$fc}Gy3yT~v92bov&*7d-Ound#s2Ln*<h@N92TXb>A&NrJkOK&
zdpWmYk5JS=o+Mbnntp2sOmp1{-Geficz3HQ$Tl73p}q|>XOhw<B`iBa5-@-Ca-vp;
zK1XFChmq1~%E?E&s*(wvE8?ago{lqpA_HDBgc!xuuKw&!P`nS%M3$6>KDWrS@+A~K
z+m665*{GyacvZNyc}r1&eM7F$Twy|Ol47^$$4r3CWj~wMEG0ZgWc@JK*j_RVZ=D-z
zM0=NH`$IDz{Z&O#f)kF@IL3HO>EJwtb;l=7LHm}EHI=IlZPNIuKz|ClG(xEgtncJ3
zfWL+t@aSizpb#ZCY_hOVvJrX0Ed1mOrV;w6@TF#rrc#HzJhOfZ>b-3;UeD7m1#Jry
za5@f7l=0KDjPV}SsBi7;DsE`W-fI|FrL*`;p;n8)_Q`yW?TFLMaJR6xcIm#Dg5*oN
zVVIa5SNZks-JAdbn04_3O9q#~cm~(oQK=fcZt`j>h1>SRT8K!JVBpGwaj#l=YOQtz
zh9Y%Q_vOwSMDt3R@N<1tBx*FQ#zi2)iae|y7m{-qHvy1R@{TJrK5sl2GyLi56x6L4
zp~5?Ii6o=~wFJkYM!jkm3Avu#LZZu!!Gbr!mxeV5G<1f}1&u2-C`KG&x=!A}y*4sn
z;2n@AxCi=U*>7U1&gdZobyLs?ZztXPy`<7v;-xTLYonr)@We*7(HMVn4ObID@yswk
zQ|&D-q~|U?A`D1o%e4<yS;s=I6K^DSmHNWpZwKnrv@Ec1#<4Qg$}~omI(Sc)f_-W?
z%;#P*9bd3Bc(k^z#6!t*ZB3DHq*~U+Te}R?VN=lF4zwE#;Aq~?ZQ(`i4W5%rV@fmt
zY<wo2QJf=j{8SlC+}OG7^NoX<^alnc;KQK32M~hnqfJv#!7|kP`w=A?@j9M&R@k)#
z?Ld8608R+D4#5pH5GPZ&5=O+H!g#A-!fm!(Rn*z?OF)#Z5i9V*6C0BHRnc+8tWzPR
zL(itE5LK*r5CtkW;ogU*7ckX%vsV~^^I%dewP=DXpQoTbeDIzODrbVd?x%1a6^$lH
z>_I*%D`*Nj|8xp^eLR11(eraz7Q+_Olzf9_<l{g{>GD$$v+TT3{gw@eFct#d*AsBB
zFO!$!jm}|f4)!{pi^DO)j==AiA}+mCP!rO@!SH3K$C6$Qg~^$5^G_hBAnb8q0%io{
z{E%}OYqo3d>{-qb<}D598i3jd6_UOb!FM=e2w6e=tc<aggX}l$lJK)>mk8#h8>lvw
zdwiA28~2*E4GKB|*~xvMUi2t5@eU(BHg3WRCbz>7$vW3p7%a#|=3W38I_Z0gp!DVd
z;1nAIA81GzN-(6>%%|J7DQ7gEkR``fCDmsG_=T{Mk!RC=LzZk*rM_bZ0X%C?L79z>
zbROfDE1+NZP2XX2<0&X|7coWYZ2YB~5pCNA?!%?Qhf1x^xH1M;AnY_063c0;9GMAa
zlyz0<4<82&a6)(2)VNjAqnJi9WqKc^3E=ZH>yQ!gKQ-Rg>TPj0H)zRaYzlsLU%&J7
zO=CM}-Nd?-gt$6%MaYS@+m+t!8PxVy?e5r1P(?nDG|W3!f=*Ly?mg{z2Ytv&#=<_K
z0rtEnV%+_;O#DH|F1Gp!shA*{fm@uLYc(Iex$S^E(PGoEgz+e`&JG>eb#Ku&SitC(
z*Y7YM`Ro%%H+ivpRl_fOG`MPnJ5{c4!iA-b$~^OR)Ir3a4-Ul`?)7@Qz~@x0960|n
zB~AL4Z9&D666CA<4-6U|6w{(&(jFh0g6x{2Q(x}hhQIpxdCJmy3t>j1vt8e%7iRMV
z?*#H2D<WR3z}2?iVT8`vSJ=4AVRrHPcZq5zU#xi$lJlX&(K7Eo{%6J7{FcbFXmpib
zR%Qcs<r_bQwqdTN-CTRy!!@rS-Ct9dLXsW1szD=Kx^2m%%$Z|XCz(slRoEQvyQC)2
z{?e`dP>;`*rV6zyaxl~g#)JTT>*xTD|K7U(#7EN+Z3Mg0mL>MFXu`<HX*m)jkaG6k
zv%=K;jqV%=de8#lS#IELOFp?<@?)pdAr0Kaku4NvIL(Etw>Yc!8ZLYJJKTxlr-W0v
zcbfR7{C1OS(h+=do(l!vCZ+nZeelo`LC&np{oFX_pr(wE$AbPO=>p$eD#4S%s-i@b
z9_JLUT$Oy}jPaqX>-B9?jbyx&3+BGxC|~qwL{Z9#HoC7)cj2kpvD~^)!<SlP&O2YM
z>)5_8Jk#{^#d6R5<!(xKo14r-HTxn%9oHWUm<(*(^x)jjgKl8S$>akW)9t(0%kMh3
zN?9vSt~zd~Vo@_VTUSmowcuHr-;nlF;Y-1j4MrE<ZJOV<&1v3gwZ&)GRmUAQdK9jY
zmb!{xxGL$1SI(lWm^)939|ZK-cls^&UTm2jen^dd{NW-E3-t^qipTp01x>7#rMI`P
zXy3qSetRgG(4VAQyUwE20GUIxa-4Wl%9G&f-?!W_dB*CeX9h|5qyyb_Vx(JAR?IED
z!=~GvzHSe&+T(=x<f<Pg><!{mr=YfiLLx2tf#aHH&6ty$H=`~_hM91T>T$bM2G_)^
z;f4lx!h}>yiVpe06r}JB9i?=r%7vRJTXNAx)Gg8Be%@!}hDw7isz#;NvTodhMKOZt
zM!4qo3n<%XnAdkHLTuH;GoErz%Mk+VxhE5t6B~7~W?LzvcWQiz+(8FzTh0YUYWl>*
z=u(93JLUCNl~YjXY{5L>C=Bo03U_WPTZhRFJ-nSUV6o`XR?x2+YRjv-I6H7I7{Q_`
zPF;W~;4e-+wV}|P2(*g$0U0MbIB);;vhGrQTjD{Ea6e@<wC2VHZQsT)*&cY=@Zpsk
zrXap<)f1(&H!qO}XUlW2$zz`sn&E~@@2pWmPfAZbAGPD!6hsc4d`Z|pK6m_<VEB{E
z#E~TQE8Hzl;8-7l5hgi`V2u#wyTA25b_&ATMo=FJ0#?N<W&LnkP-_SBVMA!=(;TLt
z_oVJAXwyf9rs2bIUatvfTo4>`ZgvgUy5pP$VaZ0i`m}RhhMY{pjUtCf$D{@(o=Q^>
zkp!&`sCAowG7Y$&KbQ$>H&V7PAOi7v)6DnOc3@6YIwnGMYvGe#1G>fIH7?DL8-<_8
z=#lW(2Mx7x{k^xhslBww7zOd-#BDI)eA7&@1UsD%Ey_0>Dcfvhh~n`PLXUkR0~4VW
z8f#yB4Kg?A{ItS=Dy%(sgtJv8nlB^jzFQjeFdzoq*zQ+8S!5@{Nv8An29<=$jL-E9
zxzc#^dW_?tMDUHV>K!q)zM52aBc&rZh?ThrgfDXxcimzg(fA2z;7)7Zyji!i)Ys)5
zt$iE0))ywUA2I^b3@yW2>Je8)C&oP>gn!bb=fxljsO8?JB%l`Xdf{s)P+D^vpelr>
z<2y7Af)h$V`KSZmI_v6#DM&4I_{jL=ph@0czj-@Dls4g{*h4_)Z+yjhhb|S^=Zk`w
zLR?5A*m6JzPI1xmCrFSFujKc`g?Ppk6r;1YA^8pO@FmKR2I)ysW-2b7d3!<Z+YbM9
zxZ7?Vvu%S2f@Jn{5T!Bi9hT*vcSnJ6^cm<juPA&rAIH=#nt$l}{=jPkvQ8`EyvPz2
zW{2@w6kJ9D^nafzj0Q4}t*)L_{t<M?S=aelgh|l(oWO$MuL$1nupoPY`}71>=OC!o
zimYMhse-Fdjv*QUgs>f4gVRujb3V3_kMvlK&-1+H(Gq#Cw=@`4asJ5QDp8oLDhflV
zpm00@YCu|)-AWou9##?l;Yb!ZOJi`hf*N?6TS|`<ZpFRU2KcH1eS2^R#t7h556A+}
zKTCgt`k>mZ0m4Lj+v*s#$@0$y!pg7FD%oz%6ogS4y@tJR@I2~L3a)v9CY&E%hjqe;
zq7qMlY=-9)R1}JO2rpwCR+xfD54iGQQlu99&EFZ~fy}kUaSkUsFf=6i=d)1!2FhfH
z8i*;|@9s7o7rl$b_|3<HSn)Qz-%AQ$z<5*7<WC^NXjACsG6<tc7F(%f;_1SrIADDy
zxnbk;a&Ce6^9rteeH2p)1bzTJbiRdbKWY?RyGZd6U?x`QrxruqiRVRDWkE~nc)%kF
zSB5c8vZ38qbf2PtSDAu_pWrwjs~+wjSl6u8;C${Xu6y-;+y(%x;EB=<T$!nEDq_^H
zcFz0*IMz#oN^{N<r&ax=EhQ!chZ7VCpX=eT!7?2`{R}Umh3pPyOT*+Kxhj&gGeHU1
z)&~kK4Dn%Yv(5zR2E0uj&%7e1^K=YV#Lfocy`?_q1MmO}g<h}k5QM(;xR|4}i;H3#
zY@08pK3hxb+?}<iGL*g{`QBmE&k62Z6t8qYI?QfUsUNG+xwaLPIOZn4qD^Cy<-@?h
z{JVn6=Q7kht++ih+vzWTqHJH}><sL3w9D6@ZKV}GyeK3!igs$ncB=M;R7al|n=al?
zxhu84kl#P%*Sm~ZZ(uy=T^`(Z#&7Ytw1glR<N*BXjTA?C+VDP(YF}u$<(Vb-!<0^r
zr6e4^b5DKSs7mpKqOp!OI4sRo=bc5@WKDa-p)%JC$@JR#0yk~dc+85rER4zR`3q3!
z6qEjEV@#5n)rzRIubdlS{@7-!pk+zBa{o#Hy(YbX(a?*y#=Jw#cC1T=+i^7KdA)wn
zj_HQy(X7uoI6{*T3En~BD&mr}A`-#5J~{eTXV=~>*ni?5jIT*lkD3hN&z-xomVR8N
zJW7}Pe2AC1bmNUaeY|EXsb4Fw#%9H)gJr7^R#(^c)K>TH$X?-aG;5m#>&eYeS+*CC
z<^J4okN+%T+q&m(?mtS@a8cFFzrQ?!<P-7qXnpo4;+8pmGgSVIWb{vHrZbIQ2Hr^o
z;P}CC4UeHtR$!Wg8xkWX?hnPN^}tN$g?kfI10*H77DYny*49}ww~Ym+__>}$DeJ{T
z8%@Jh4Q-!XLL=^qn^hEEABEQa6G;3URf<KQ?0M{1;3Iy0L&aF#`|}%0k-ZNZG+GPy
zUurYOmR7wRy`<&rJe*Ij+<)lkCl95eW5O#=^d8FbdEJDCYF$V0L_HGjk>CjSXbejA
zM8*R{-=T*-T9db~%5Ji<C~{-Lu9)t%O-!lg-z<5*o;VTRGoNcRjH*^wlQSPZ$q6`e
z0dr_H$eB+mq-YtpV9!3YI@uImK<kGd-Tu^1IWA#dsH;3X%^OmjNtzbw=t&!nuu;j%
zZ$l#OlO(Gi3HF<CJY@K&-JAUC0@^GJm##>(yhrrBIU17ouCjxo9dOM9bENL=Mb-2o
zY6k@txncDBxYCeM$c0|OgjpTYdgH0+W_&da=`y;i&G(7vz8AWRqy260t|e_k2lgs2
z{9YSD`Rf_dD@$~V2no(X3*LK{q<#^?cu}W(YlBZ4RzdPy+PiiKu+ji}RM<*}>q(gJ
zq2F%&D3uyDR0Q~#)u7)rib@u5lD!Pz@r$3hx=K$0T2B@X#T*w~7F2{TmOdq7Pb{^4
zbs8J8Iw_)ZR=vw2bFTk)8Cdi~_PcJT3iK4rcm4pKG{{6=EEbcI{SlfUSa4iSYeh)|
zHg<Nrx39lq`lnIodzx4e{b^Nlia|d^vOvJ!DwjJG;+O-i2lf0_xfykc$Ei3%$A7`t
zKlxI8$HNWTYq#wfp#?dljA}BMSD{MaTF~^-hM;M&y;tfe2&Y+Oi>goysDj$Nwr6f~
zq2B|sf4Dqv&g<<;LW^5IQ_$&KreKRFyLiq=-q9&2@jjg=R1)6(ATiPRPv-5PUHz2?
zZpRe#vvBNtLt^U<ElUl+8pfs29C_TbZDvWbb{7tHk5^x8YUf-Y>D%uUUK3&Tb3{mW
zv*Fdt-^=~y>c1y}%gc^Tj<G$QK?m|K*{Ii9TdZoSs3sGF+A|nyY}QPz>W5ks2pwA_
zKh7k7_ZNMBD;S#eE+?=f%akS;Swh%FSwyZK7!RU+nu7YI{o&ZVzbo`_b>kOjnSP~v
z(S2P#2$GWB{Te-{GC6-T&u%m(VEt(;bMW!cEmnsZPE$|<Nzh>+33C1C_dk+=L|_$1
z{an~cO~AI{b;}~EY(uSzyMskls)dbqwyv+IYMy<t$=i^0a_Mcb?l&2wEe=7Fhj~A;
z_=Zs7pESMb)<rm<xQr!yu`GTH5@Yx;dN~f>XpJ{gCirl{n&)$dYyYI=zsv1k1cqr=
zZSK89-QhJ;(1wV{?D5OJ?S$vb92Tu|Krd>iy4P6jRXxC#1b&Hh-Uc-@({aL@&9_bF
zWl8TD$6Js}6`Edpu}7hB+az3B#-o6|C4BPEqN`<5uQuQ5!Iu^Umg!ktVB6dqw22?f
zxhq-x?;@c8vvi7;oE>|xqu|4s;sR=itjo<dvA0W?*k)Y+fTH%vETLI%;`xY60_sg3
zYYkTew{0PsnlCzXsQjwOmG1t|jBE0z5>IFFbGY3O0e24i{o|Lc&gG$Ci0t*gDd^d2
zFf>=}`%z*9lYpx`>LLK6Hek(*qZn_@|NpGK{uW`o`@d}$P%}}bKjG?=jo$Qiflx#A
zlYPP+V+Fa7QT$hihAxa}dhmQD>}_JkhTCY7>TBED>q5`<TT1n6zW@1w;0!7Mijk(n
zL}%{O&-V44@!M7_E;JRdJbEML>Skg0%>`7RL?<yPz)JIRo$64lR9NT_HK94$U___q
z?hdKq>E{0P=f5ujiR;S^Iz5&!8ezgqLYE(2sW4Ch#PA-19COYNN7*NR#N&9Q7H{=!
zJMg;b$pvxo_>1oz23&m(8WdY+h@NY3@>o6C4>I=Y38SUtZV-AX;!{lZmt{V|f2y9z
z*XO68H>06&#@1l?$m(2(uVjZBtBnF9&m~h(?FtAhQ2y_;^nan`_jN!jA8iv8nlhrf
zw^Dn8{9r3k5B#+ps8=r;B<WgSUE!3M-r(}Wxq83Dlb9lFf=eh~<`hjoM|AY?3RmEW
z#71uxFFAu(+RAT#B2UE!V=((q$0QUQJPG3#k*!C6GSW;<uQ8)B#=GV(c*_}Gp|9Tr
z_UxQOzP@<D5<yJP&-GJ!cFx08M($K)f$wnD)3G_?$rm5Fw_C2&Ca(5X9_ytZ7@G)=
zuw=ETyy5u`6)9^TtnbjLaTVc@^Os&auh;c|L74yEG%TF8UFeiLp{|7Egy-LoZ?5db
zwB#oGUG3T(bG_>MjZCAA_v&IXiN2RLSMIz1L|iot_NbTO+qQqbZQ}5@Pfs4&73=Gh
zbIRU-4ne<^EUw@!uyZpoQTB(#)}Oh&#ejTi_G7JnnjYDcp0yadw^s8<mAHYgpBnj3
zxb|5*y|d2P+qDi6-(7EO4)e3knc(rC%{xjmGS=#BUg(_JK0YG3j}8A(!RB6q$l%`Y
z1VevyCPf}>yKWg5>?t<o^6WG7LoVxyXFUwVnU&Wi#xK-27miz7Jl`#MqVb{DnbA?N
zY(dpqg%{aQ4>H-DfMitHVah~-5+?%eEdBS!?*9eqSMOBf=%%&s%cr2I<009*QC+c2
z-<bSy$?(RGc{IPY(9?_oebsEQ?mUg!<oT!aFW2cgIH89xpAd~j-_zK=OohHfQvc2>
zLtE;*syADRRlX`tqfM>zXTMZSI<7hq@YJolOZWZkmD`6+l=J_D;s5`?-~24JC#vhJ
z$aunB15Tpd``7BsvbVwX2VFM2VM9Di(nnsAuvYZ&yTZKwg6zc*jdnv|Q7c&kozNp=
z(dpbhIx8(GtS>R4PbN<#Vc3Q#xmGV7?o>Q^rL)zlsJ->I6@O2nkIg%8H^$lB<|klo
zYIYT!dq$NWzh1(v#N_e!h$mlvdN=`WC9YvJ$2P2t{ZGg9Cs6$Rw-+ur08>g~+nd5!
zhQ4>S%njP}+H_8|2_dFp#0|?ao;4Bn^X=2lAz(>n8KvOqEQhy&d42uzEpiq%0=-Y&
zL)?S9L%6eAIvxxu)pW4y7byH`7$fCMT^co+`ar|x4UhLsT=;=cmyv*9--lJ7TNm^A
za8@TZlGKEL;1bd1e}8(@^il=?Kc-o1^cJw`2c4^Njpvu*KI&W+o7AnP^d;;9pXa(c
z1wGvf3AlQHT=d`N?#!-gzTP~>00VZ{xYY~xR<0j4D~o2eV3JLc5bh6w^&^{U`c&78
zgxrm$mu&mSZfiLQb$87-^angjS{mIbT#!9>t1GW%2hvA;H*-Ld>arMgMPUlEy1o&Q
zlw@wIXB>`_lb<+t^;N3;{XRQdXh>cO>1MKF52L00hIPSJ*IgXNU4o%<p`FCtKhwm&
zi#Y!e&^c%JR=b_?C69je5sGIoM|HS_7O(sid&G_?Tm2KVWnF#%;**lJH@}#Wu)U2v
z>xvHIMdXDD6K&B1BBZXm+y|VFviwxs#OSDmH=AZWms{Z8Td$bnVBH@ziMlwlD99o4
zkq5`OupL`{EP`fE9vT$tt`ytI`>{G}h$XG2McHEYCqnptV2<MlI8C7hl`Wiz^ls6;
zEA7F$EQKs{o+GQVHze?(3f;F$tgpo66^+_B4?SGw)Hdz1ATV$H#BD%!UH(9a7=ySP
zw?#EQqAFSPZEjzAq-NP0{#Cbww1c|vpLKL}+FB2-ss+Z#JA%_hGv|b+%E;-SR$A3#
zF%oJ$6j#p|bQ)VZAC4j~=U?@?vtWzWt?gj#)Zq`BK(+t*{f{KDaFZmzX-n1BjWUyc
zjrvcnBkF@65~ZWk@s|7Fu2rEClZ&v;^OMt@T4;5&$hBr`Y0r=<2_&&6Bf~Vrea(l=
zY@wcL%vH7`ki%LFb2%JkZg#QEfKef*JlWU6wBI)NhOg@>VZ0zPG<21>CkJ6-b=5B}
zHSsi3ceO%G=~Bkqz)5N3kEs{uO;65j-S7P;+$b%3i{n^t?G;?ydPpcigOsifZW2B+
z!wpa3xChTJqCfqUa{nf`#mX%hA;NNAH6-SC@F=#$*f)z&7ZHEwb1zPMkjx|I6zA-j
zW<_k>8KwpzGB~F>$Jih2UW!~_I;5u;o32KD*;!7swsyI7uF|EU#3N^tFIJMWE|{<>
zFTum}p#m(aP~SBz)0nbpm?zFcrE3ZA(}YfI@5=fAS0p4kuD0sW3rG%j<osc6MoH;~
zdeZ}x;u9lCkVVYB<7>uw?k#A4IeAX(*BdJ7<ik-Gcli9fVtPyZ@H$amka~BSpa1mp
zC7+|rAyAW1l9IF#skc;2&UBYxro0#?r9?Dgy#kUwt;@Sdxjq4bzzI??j-9iz03Rsk
zUKLRd(%>a?o_i6=A3O^rv7SR|PH&2er9L_5P1g^$<OCJzNyQCRxy1frX@LiJ`l*QU
z&KEAq5^4Tl(%=7yjX;_~*^bjU(j4o*KlpOBv&SD?q3ka=|L)P>z9GrV5yVdCmK^>l
z61DjFs7Rc@`~~OGos0IL%dFw3m75vqrXM~hXMQHlsSlkC3tbJ70g6sts~#I%=4%Bj
zUiy$x`npCV9LzI8VkdkUw0f&TZ9}Wf77}s6r1%21jyr39(1mQi9*5(1Ss{%bsjB2?
zjSDRPvuNi3fw)p0{?5GnQ0W29vZ3QgSUQNTqE&7BxMe{<Z+@&$6~jjOOQs$RC3Q%?
z2s67`x)^<IOm3TgE;!2vnibafHe2QdPrlgRhI-Y3+VbI5MS+{Oj=R5%i_#)kwO(54
z!wF3NqEPD}`MbsA+Ui40I;*Q5G8yCb=RBUJVeQ<FgGgSd=Ey(QKfVs{U@KsHEO%<4
zS!W*qK1+K@bnuTDEIPe&tOE+}GD`G;REm}~zW*q$vn!!UZ#fM~_kw0wsK;FaliJw@
z{<1cBtk1=WQOM3ZHR2@VxP+D4SU1gE+(CUf)0NQ!t$nbIPU(&mXIu#KAhTPXi=gF5
zXC(Ir=nY@CsspmxPQCu3iJdDR|188Qtx(K^bFsjp;I3Fk(faqpx#DTHpYvL@{kK2M
zi$j~6@6(g^C>nUo@mFkYc;AnwW}qO6(-v5szj~3&g9qBrpiSfI&P8>X#nK<(bs|{g
zvvX5JvbPrvdrW#=fE=7(IJHecr!u_EGbhyPw968^Mhse`Ai4MX-37KQ$NlGa8#JeN
zP~fo=R!OrUV-ND%TuiZym50HxXAu59BYHfMI`RRk)va`J>^&<<>n7knh{-VuXfAct
z(4iJ5nhpt)qgg&M-E)PdH#_@Ll!86oTU<S^g>CbSO27?3sa>rMn^#~V@kz)Si^3%(
zi*Fi=$o5~}h*Y`!0m`S^m?(?yG1VYZZbD)!Nk#W;3L!-rHGqk*WacoOAluWeSZF@c
zs>zps9HLq8qQ9u~iWDPaQqV(D<XG_{FO<`yw;QB26mhbRQ7Yyk`tKmfglw%`jU&MB
zmf^Qz9u0fAr@h@sN1;I86>F41zW<I;W8Sv5i4V~<$i|XX=}MUeY0-4}JJmuzh<OY=
zHn+Ig#pX;ZixCMo?5gS^TDK$D*nc)<)-a>@GZZZN9AVAte$}=?>{&1+Z5>I6ywYD9
z&2b!|NaJ&gC9E)*-I!uY%;2l0q}A}V5~zq%*+f=LL$RDv8EEvl1S+Qf)6amO#S*A>
zLsg8sP-q7M9$GKvQHP5=)5amjN~m{{kD!OK`CU*^5)Abm3hn&l(hAXx5?gGs`Nzeo
zT4Qi}vuH?<erQvJQ+_)cnw{TUaGiwxDD9AF={175X{jfJ@{6;nN{=B!^U}*KUTTbZ
z1&XC-n>iy*G22l9Xe2NTQc$d5U0LSa>(K={Fc}44$r%40B|XL-%Acro?QOUZS&SUZ
z^ViIGWAgk}^3t;9&>Ad(zqC8Do5j+bWs%{@c-2ohE=fb^B00T;3NfuPiycN@whhG!
zC~E!`Sp8{?9XjnZOs$73dS7U=4JNxkLYs&VE!5#LNb7~cCu%n?QapZ=i{)^={nzLC
z)og3ZHoMVU2uWYa?+$MkuFUU%@`YA(?(khG!@i~5i8*3S8m@q7nd?#OMw+2Vl@~HE
zM(svhToz}H7ICcGE%atJw%`-n$0)R7s8QcMhiBuwdIOYSNy2$~P2?(zWwbU@qlqto
zES}ZBlcjjIREue`%yU{?KGBrLQ0`XA1?eUIP(E+la1ZpTA_QYeixl64cP{Is`b(T)
zlw|15%gA%GWFb{-UeUKm(t4d5KfNEX$`wm+wI3vRSGrMcH;uMJ@AEGsof#95mEV~@
zl9wj@>@15W*PJS%B+q}YfE>l&rqA#eU)unqz?CIw)aY^B=VEfSX8ilUm@17}sb?V<
zGBfPRO<ZoWI3r<#g9dsL3E`szDh<c*m&PrU;@9=ILk{e;DJTPyJ27!;@ZENbClZpz
zrn@}i`%7at(!Ab$oR{A1ROtZoU=0B?<LzWohoCp9ut!6k6lc1gWTVL{-UEy1oFowm
z>oU1$e_7=F#(Wz>OtB<wa)3k!O9!p?jEt~a-U7Y4o~+!t9%q#UopeRjT^weS^~8%5
zG%-y+z#beD87<fEb)xet02l9DBRJikp?m>v=jolJEjUPQbx@hPH%jv~u^1|Fv*Ooy
z4s|zH&!Um*Kafl^{O1M-c_3~I3;LnPeB`>S0$8B9Lw7jR*6Rb*I67=hSS*3mrq{Xf
zV`qo;RlU?;0Y|%sgDI9p-2{_vmj3ge5f%-MfJ3j|>B%YOa7K)yef?F6fZG$P&_vgN
zF43ye1?jIiK5RwN;O|iuXYkWm^e*U8i<`A2nbZ{XuF!v$*NXI<i)2(TFbZ-#v#pag
z0%<k!bFYj-Rv!61!BG)!qCFy^Q2k>=X^kVZ&EMo$LsG;S*p{S{r;toPo?fCuI4*Ho
z!Lu!Iq{;a%)OgbTLm}&}m?;m2_1i1b^fOXt_HLxjeX(@+93vMgVx8_YfRQnR$0V3C
zv$P0Bot;IqB%+DvL7XNezL%DX%+n<A0<5~S8CxVRdc&>u`O8=rwWmVK!^5y(jKXK=
zQ6t@N^bH*-ubVLryf#Uji_w!=ogBfmRst4@bn6@kj*o*L?koU~h1C*1lv9;nyxbz4
z+4L5oAz`Z;0T$V02fQoBUk+7bTaH(juu`yx1&B<O<4Y$3Hx=L^GV?=DosJLO1jH`?
z+4=N+6LF5B@=RdJ4W6onxU&*g*y7pwBNb5N`@RzG;qHm&V#Mf)%lZC_45tX|U6fMb
zw1;vq#R|yiLLFInXr#CUjODAAKt3x)#LIQ0xWrV?_P{&QsGBgpxe{jtV4H*$3(}fg
zkMUv`0?$iB7U@Z2kpkV=+4<Jv-S6p;gw^&IYImVnw2e`VZz|{&sScK6rj6Nk9z<%c
zm`9k!jZ@_Id_bjsM?tO{4)n>yb&wV_o8(mVLQo8h13ladjwuOGRcYpcp2NXvx62rG
z{<MWNwbywiB)2-SV2@iilgaLY9MU<cI^yx!*?e!+bZ5R0nn$F15k*dbyld<ucwB>l
zq;1~0P%bh&S^?yA{E9`%2ji!A33MFAMmo0o4IaPJq|@7_#FoYr5iXHV1*W^#8<oQ~
z81HtsuDn_?Cs}8|g?Pu&dTJU1k!-6~iM(Jn!DP?!9Azq8qAe((`L9I0wJIw<cCXAQ
z1DyD{ccjD8H?$=(qDqzO>+xJ%XZPJye5GF9jYMBhufEZF=z$>YKH{i^wN1-gq@(o;
zu?_#PuYvm29D(7X6!xf0YH);c6J${6>-wp?Ue4nL&u!?rGWFm#Ir-1QbC*K1`pOUL
z==_M<6o;x_yg_p<@n9D=oTD{%XV*}^6GD~KQn5IL@{?`p<RuJ6IlYZ*y3^ME)9CTz
zJLhNBXKM~$>+kK;{S?*Y2u=;L$f#P^vrzS*_Y+Dz@{z(SpNGEtwJaVDBh_Isrx=m2
zq=3aJ{yU}iby}GQ`bcjS74!1=qX~nPZyMQ<%!DNqv1PX0n>;Hg8m?6AE(nBO6zs$t
zQ?i%d9vkO|_Q+GI?{uH8=o@4!y;WwoSd7eEN@i2sQT}v<OS15+JZUIo`nK3?t{+BE
z^yd%@n#rd@;@5r3%4uUE?z|+&&`tR-ZMu61C#1W^uNM^7_0MZZ*r8F&qMazk(j9}!
z65a)eq51-&TD{q#liI|-)Pcu8!syMkMe_RXCBCH)vBIMwM?}P0bb$c1JI7=rV@#P2
z_6}RTmls`2JVoHrdZpa*M>6Rj=l01CCcT%pma5#GH*o>!hCs#knYN6gKiPaFAKm=e
zjP}`t<HaRk&b)#^J(<n%;`zL4t)qROH{JfZIY`54_N1JLcW?u|Lb871gh(4@6|K|2
z2XMqv0MS2*uBLB^Q(vO*+gl~{DTCI!;!h01j!20b*)Pf#AbL2RZH;50LdCDML75{c
zRYkm=?_tRYJL7JP>YugNM~AX3XNZMJo{@^YwY42vLUa>Z@J=^2x@#O?HX78}!k-Z!
zQhmT+rUpRHhH05@E>x&A6dju+Pkm{C516i28ML%UM9+0mlggre`$aU)7Uh$INivUX
zY1WS2J?#w(zxtJI|F609rm4;xu@<MBetkH@lZfHiFHcdb5+c?mzY0ji`R}UuMWzkY
zWCLO=rp5ECs+7M{bNBa|{)*(^Q)R6}|EdQgn~&4V5mC7D*R@;+x5V{li8z^7)9)#o
zX6`VM&o_!H^}fpIFQxc<l2E_r?t5KdW&K@H-B~%-QZX;T3O?|6qWNC<J0agx@lDEq
zl<N<a{+--!68c}L>klyUmxTQnoBT7K+y!7)Hi`Q|<{R?ARV?;Zp5Gw)U3h;fmT#>5
zl@`C%CI1z_e<1!Am({DffmP|QH2i9XC%ysiFQos~UgmH93We`B`(1tC==^S$e<1ST
z!;SL?xxTUUM{eQ|n*1i3KO*cKXMZW?nOy2l2J+-@(Eo#$RA=>DJ+b*1JMH9t<wE|F
z?cXtEX1^gDe1RP};Qv7-zjU78Y5jZr{(-uF2hq5auMT!bM+cgKI}$k|(M8x7i47RY
z=hXZiaDOwJuXvK``g^4Qy<ERz$R?trWnxCeC!2#cFxRG;dEluMg1q$AlYWEacLW7^
za7szl(-w>b{nY>5fd7L0zvv<UpldvU#C9}!o_3zL!W{Dt!EtY*Opk}ng_b_5C=e7F
zkXDj)=haFCGFlgf5>14bWkVboKY`x%^Y)ID$AsYB#BYEs$diwSVK_RzJ3~Xe>hO_H
zX(XN0kdd~k%@gfB(*sF$cd7`h83QNx-EEqui_YpUd8^{sf~LV2OBH4h$@ysHT}_hq
zU`}}Yc(CW82c3t|e(C*lmoBmDP4b4VqbM)RWe?BliRhXim>Y4k%4(6x&_11m>UkP<
zHW~Qd2yE<f`J~i>y?I<QImZZ>RlI#Kk3`(~F|KX#lF%F&y!6L_kzWpJ|KeJAYb4l1
zUwFQX`}W6B{+tX5KBw72umhYMKnE|-V>gPGD2IQM{on9X;tnnY##0J}AWM~+^=P)H
zWv&PJreb<)P$re+>{qn^=&exWy@4Su6+7xr7EPoQtv)`zwB*neMk4u%95|b^NN;;R
zyG~C(^UTDou*>yT2WTc|0x~@H^9NhbngoA56LYgTm+iBUu;pFLqm~`-T<LT=TLOFl
zqqmr|I))JSp}8W*_+er_1y6e_!&@a4YJKC3T}cq$g|rZC*|%pzFyh)dT_GjU#>MB3
zb)cw&gQez1T@;el1u5Ao&V!ELxZP|83w*~i`wMBAyxj@-k`Adrp8gbch7fqe+?kX=
z_HaV{33v^VTpx#cH&`L%GOMG=_Tvv`%@`&4TT25*-OKG`lPw<EUe3lbn;w>*yd%_Y
zqipeMc~AZPkgpKrG+_cCLB6Q!K<-&}151hTTxi6cjDsZwd3dWORleseoqRdL1z+Ej
z|3q+w)kvil?qg#QC#>g8aK@@b8h7B0-#oTp`wSki_eh_D($^-LP7qK_IM*KFH5x4`
z-O?u66n1=~MGj#L$$BBvi+j*4-9=!x3hW<wN<ODmAi&irMxy{TNA}rv4OI*E$YTao
z_g;)PD#oC?ag3w+*kOxW6^He`UbaldXa^kQ$mq4<abbm)xf_zBA(Sxp#(`r-;;(eh
zp)-Vw?TLqnpEyyeW<7!X&hiA5%mpMOI7TJT*c|LPQze}6bsH4&dZpr=DZ%JwEsLl2
z=%wzxErJoN-J)9Jj179b2z|1hlFitAoS*WUOhk;mmLLCKR}Nx@t8$l;DS_C##?>m%
z7p+K9=_eYy)Q2w9tz+{)MZ7O$Yie0IV~r;re(?SxFBc{ADM;U*Xlz^W{R%L;`}8hY
zfR)9^t1{kRQ;;-ye*EhN9LAvF?&->(6<#j2XSTe9*Gw*T?`Lt0dgOU8J~?O$bR~W%
zu5BMv-cyBP9#-Y=s7(FUTQCWIu*l7iu}Xc?=G`3`lnU>cBC(#$yfOuWqgoEz7vQ7F
z@vj$U6{)f$?}Z8x+=J$=g>CF_C2>q#t-Kd(7v|6{B|BfE9;x-%l{%XTIQ;9;W&2tY
zCbrbj$K-Rats(eX=IlfF%v&1!+9?O(a3*w3!qy-z{&t}gOEIFYKV;NScMC>H`6T%)
zk67MQ+u!{&`r-Jxr=AbF!n+hH?|y#Gg8t!5j#?#6ed8+OJr31OYCZd%w|z^EL7q>%
z&|)%l+3>iaBH|<26A?*T$~+`6>)B^ZX_;6fdigZHP^CHrX<^_Nr63>t2>(^DfX-3P
zyiDOMpj#2QCV1E`BN%K4H@uRcCbS3Ztw7m^F)7Eqi5$9&%TRhjv=IN+yBt@}!hi}%
zA-)_bR}{-aph5o0`=ux0sqma;VPVwh2QAkR<kK7n^vzoec3eb{0(=MvQn6qsSgtvm
zFSv>LVB2R=tvbm)t$WbihognnT$`jzA0}UM{YbVG26<Q{lKMrG<O(l?)4jdIwf44a
znZqyg%}@Rc6M8e4_yd^y^43hQWRVOJeKHpN$n(Y!*I%cOpw>-6Mg#lK79eVITfjLg
zUcw_<PI(r>`TXc}ovV8L;E~&#xVSF9VFnN8NfhRHO;+JZL;9Ia79z#*u>zmwgOjJ5
zoPVt?MBEB!#cZZ*z3{0gl$)ZvwcR?L6Hr;^)QV?PQZxiD-r8N1HJz)v`_@hdZjQpg
zuVl%01OYpFK6-~ahJU;K3ThoU#1)|ijG*?f)5EYM=$CjoCL)=%j_Ey9T;608FgB^_
z^3%^RbdZk-?@nj92Hrk@S>Rim_gC#XQy!4*t*keGW^)&PBMFwLQMoS$cmtza)E1jE
z^u7MSl|EkxY-YBlMWt?B>%Gd(*V@!y6`y`Ac`ulw#yv3oeA>o-k^i^pneycYwi!M%
zEPoXlEcd%eej!mH5v9`o7h0$7VI33q0id^xB_4NW5^Z`?<%*HF&Y@dgeAE_X40dje
z*IOh$UYIhv_O-U4F#nQ#?}#}wmjv5^t2{~gA_zBdTpdl?`Up_p6CF5i0-Q3b)1--G
zlAkv8ex~fN!@EpSb_a*>!I{%3X>Jq^8<f6L%IaxIT+`gtJ)xSXcCY-Fm&m(NAGzvt
zlN-8X5Wf)f-$lY#w*NZ+FQ|<j{igiiq9=$Y4~Y2U%(emau>V&2D^pzrQ|BXHqn&fW
zxCsX^`FGR6AVgl*q7q*F>-JDrep{c*P!f>Zv};|EOT@P#!+rLB@zGR(o%&y=F=f+2
z{+c};WAI-}|D)=E+Z>K@{l8TIZ|MQ;;o}9^iZ30^@4?R}_l+tR7AZ|kemBRz*8A^0
z$`|waEe8HkdfGuvdl-?9nfuODT51?l9`4-1E~}G$MDD%LRp8*CxAG?NMc26>jerwk
zH`a=BO{{U=G{TJwxvbB`wGx(nb6oePWrb)z<6}0+P|#dHlG!o^sS6y}6Nhavn-L}`
z55Z?!ewwpVY8%fpqRGibVAJ%facmVZwb;VicB$@1VX}F+_nb8|HuV#gfN#zD@1S?4
z(+6lraX<Y6ZW=TuFPXFyp(eQbxAFE{K>RlS8$A9NU%z?$TX_6e)BnTj|C`09J=zdE
z6HQHxsuLOslfU;A<gP#T%zx&=eudNs1_x7*3|6}*ESU+ves3M$x|ZK|75__Vj?Dz<
zoyTyAxc#C&TxGICqr8^N(zth%kH>(3WI9ru_CgVViNoJ`weE>&r?)i;Rf?05B=X|W
zMi)fEHT}el`MBiR^^+<hg#8@tIyTyT-fuhmJA^E1Eg{yDq{=yst2JMrPR{YP{%H(s
zCvLyVSK<>>%=koeMDW7L1-K$vER|iM?yBEw%dfsecn-=v-A2d3@_*ff{<TZ_%dX@f
zxsl&KpNV^%&6)a~zY0XIpne4gy!Hfon5jwJTtFD%eGiuYYmrG66-;bv`2a*+`==p_
zIy4;{eF<)Si2*9yKkl;1eSD2N>+~{#zn*lk{B<w(TQBmzlomDFBc%pn$g%%fxAkX!
zX(qP$)+<l92RiLP%l@^O{;l8n!tP9Qk~&8PT?TsDiqSjZ8-uOWGiXS)MI23T=yK5u
z&uJf@q<Z@#V_ZFm??^)j+S={d@+L}5E4JF#kkhA0mhterUgqaC^#|buCt@4=tGKc=
zahollvQlv1awBO|np4lUq+#KhF$b6zmc$vIA6;3+hNV9%;_a5Ho4_aNwzrmOVt-@P
ze-`;nM>>{32Iw|Yu00cUE_0?sz;l})yN*o$J-z>3IewM%j2c`yO*+mRWwPKc7QEC<
zchw!_Q>cevOd`Ine;+{TI^ccofo!@LfS2}Jy_HUDVJX0lBMV-)>|>%%%8MM3Kr?sI
zcQNNo%XQsT2ihq5>47HMM<bQkg?Vz?1q0xuLpfOim#celk+d}~MfYhA($wW;Bg;T7
zk>x{nljolj^fS<lk?9vSHi9=oe7)t|l22VRxY88rM~%GgT@Q!b3-+18wg+}5UO>CU
z5l`QqJXWVcGFk5G+|`L8@3c~JCfgC#v#d9{Xf$2K+s$v(r;KG?8LS^uu0moRBF0YU
zYg+Ce&L3HneJ^yCQ0?({TF@PRqYF2eL-Tr{EbprdwPg}2)>BhehaGc;l`LrNqYaI?
z@SRTfEb^s-bixYuS-seBFzK%j=9*#>ZJT<T^D-*}3;5bEFK6q*J^Jfz5>Q(@k?T=Z
z2AZ$YF&}X?8SIbhVJIG}NlnWXYG2j+@Sn4M=BxxB7JTqxvG_4|U1e-w1O@Z*#4)hU
z9;Km^#~fPjs-9|7;KXerMQ+sHOE4wW?s#~5R~+W5c!1H>u<>P@mr_cly&sM<7)zQ|
zE|X%Fj~{XCrB!vycsjPd=#u)$pvubQBc=zZaPRq*)C}qws(V+^=x~RWZPKbmM7)=C
zd>HIOoo73GvzUBRta_8?!hN-Efun?0CHzgo{^$!<&U1sZH3<uIJ}k}ZEk<mg+}Qgz
zcY)?xzMQNJ_C1oUdh&hCKIa%rH`7!J6<JCO^|C<7IFo$tr6_D)qx<8V0-2Lpq#c?{
zh?>DhSg2+GftE6}O6^JFI`)UlkNrM#rMkHi!wukjf{*mAap0I*vv6CI)14*l!)(4S
zYYvjcUEjWZW6{ZxPX*<{6OY~eDAg!<1uWPH)rk)~y4T!`dkpNWWKpR@qG{}nblqXp
zdj*qu47O%kt~6^1jFGv2jG>q&53NpNujrRzA~x}z!MmsR{9RSKuu2_lLcQ0bliV;e
zhO!tH?xWlwO5J9Aas&AHDrd*%`X<#!vwbiHSyD`cu}Nh5HPcodv;JLG1vJ+6x&U8+
zVcI=cK2`$FE5hxz+O0;cf0Xet{^>LkU7%Uf{J42-8WD9wpu+H)+=KNJ%?-sT=z=4b
zmpR}b2)#at=F9Kg=Ya*O=hB519Qrv>=HMS>5Ww|Nd?JhdO0*i_+U2Ug`ceu^%N!Lf
zntDI6tLk|`A3vpsVwzgG)P>+h?3GHI#pdG8TW517BsprtcZSo<h*-NS-xB6m1HUpY
z6E!q%I$dqAeKQqvm~P>c46EFCcRFRrEx;JCBx(3?%#0{p(n$5FEml50YBSk?li{n}
z_A+O_bL7>~Jw8;u57x^(Ieo*lR!Kw{6M;%@@RDS7el)Ko;D&BlIXh25+5;CGn-NFR
zA*c{qX{ot!dorHYJhu-)sT-Rs(yZxpR?A!%;ls5V>41fI0)LQU!cYSogwf3<7rQ6X
zdGSjoT0BNMAI|XR9K`_ty48d7MpV0K;%C*y3E}4w<Tok0rw)%rI=FdF3;owhvG{T0
zD)pjUC8xcKkLZ1$Me89B9ggoUuVV#a_e`XQ{95KSYF_TTc1$4Nw#z}yZAMGKs-Jvx
zui3Dm<_3v_>cxl_3oQJV(v6e95D-l2>>zO#PxG=s#O}feR>ElPHYs*E-%J^CjCh_a
z&L^+fm&1}s)Y&UAV958?s`vs)gtj0eHKXo%jycme3zgm9jMw!@B}sW}VULW?8Y`!N
zR?Y|C2?=j6((4PQRPT51OP}Rl-oQk^RxqyoOuvrrm~)Kiq<aq5CbSo;QxXlE_rIgo
z!S#+5VuW#0Nw{lfjFDM0wT@2i>U0SY@I5~~wwUAsXPudI3?mF!(pN4#h>g7y$!~Xw
z>Xzc*!s&5*^nRVK6WqD5#xzm04J;}5k7Q#mXaE=c6Xi_0M51mw26Y&fek!S6M%3yq
z(j}|izMy&|@0t3oc27YSF7*o|d4gFAV68hxW7djtdOz&0X0%Wm5zg1^doY;vR-yZI
z2tM3ux-s%uP%$d$txVnOA}lc^?0Riq3I9IAIOZ&nX;lknk7S+4_=s`Lk?!Z9R?dX$
zcz>4y!g}5d2K6jJ2p4N|z^0(Fk=fQph*e7jmbzaGn|)@EaXU6Z#v=VB;zlvTxP*`v
z`gXaiwGV$F#{Z6pwt0<*DT@ys!#1N|Vpk^<Rt`;z^DM=SlRQr(@B{SpQ&&Z72h51c
zy`0`g?csoh=q|PGWM%hq8Zqs&RFW58o>&JK9)+fLac;?(d8O{^V|~${dm|rPeCMGx
zXu<uH%WPt$u5!$@R7~K^6>K0$K&fOMMJQ!PgZr=gh192&{q?!_BsPL8{5W0DD^x+H
zuK#o;8g(VOK6}jmRojZPI01swlw(Cem~2Kr{O(n`T}k1)J_DMFU7SEoe3h-6%ZMIA
zq;e?^KomXn15lksC2N_LHeZpT3bg}DOwm`}`#@rHMvFkfGt3L@vu0FjGCZk)vLr)y
zBDrKl<f}gV$9m2NAa3ChYS()OM53{Y;LX=2quZrX;o{S9v)Tu*w8x0BgxXb}RH1;#
zNyF~s06byIE}@9Kngj=cox9WAYes9u0x5r1p269)bP~U`&M=>Ol<(dscsav!Ur7z+
zvxwTdccXU9>$UuXD4hS=L^wREha=wC*-B5u9u{d5wTqN2bE0Xm>o%$QD`Zif{PZR^
z&{G33nS4-+om$Tg*lXiw$C9kODLjhn*8^^f;6jSOtDeXhBztmw9VHSieip3$8AkOW
zVRZQ~r*U^xn;e@qtww39y$;{abgXp`H<$m)rV?F){RP<ePMe>}qiS$LdpFoWt&H{}
zNQAciBE(K>C`pMU%>}ZzCKF>dqv<1OC?=!;>CWLv@)|jGL6|Ni3#<u#7uGF23ageC
z0rdK><ONtnRk9Xa?MjNr)qK%<lJTV(3CF;@NS`LYLMxPiL$H6<RyLZj^?IGF?=FGd
zH{=ql7e;ck*_bu%!`3}XmEU=DFYjTY=^Id?;HYr)OOdYt<~y89zuEc)CSR~EK)1KY
zQV1Lk2B?Y0X8HLB^o=*Tln+dEDc@JcZxjJ6t)=BIo?%|(0Zg0AW@E&#aE=B~{0qOn
znQFxgBL!KUTq*Y|&4H*f5q<r=VAwuDU{9hBfBqt}P>u#F3F!|6NdOaCYIkMKXxBWw
z2PVb$p6{$G!2j&~s*fN|PoS)7j};k@p@PYy1~&EMm(pK_6RnlP{^Bs%7m1ay&(nTj
zzdKhLF$n-G@&xoLND_R7=}W2(InvfchJaW8h9Y|PVuTVdS-8dEm|&a1<cDbm1@zUO
zAt1ScntqP%c}eTl=`M9Wb40F(uB^hoz=b^!kzQ|IG>eS7fOsHWt?XW&gd*X#faE0j
zFukcK*JH9m#&j22_fkUUh-J<_5j^5KbCHv{m-d3W)4;OY#nsZ?Bc2geSTtM8sS(8@
zQ?{6isJ`KAO&pu{+Xl>Kq@1c#Nv$$<n?T<q>NzlC(<(gYGUM;RbTgu^Xo-N{^VOZY
zcTPcTD92~ofl>Jp^iPD?FNj(}G5Us%NlfuHm_c9lON@!0BX*IV0RyKjNbiZAHgXU#
zq|d-Xz$d#zf)2-`52pv3U`x68VtWK)(^mcy<&-`8{ue1a<GOO_TSQqT>k>)BxE-KF
z6!j%7eOx31U-cGbn-&WUoUovpCgOS8hlpVQjYI@d<OCLCiRYz~w7qurb)*1JL_YK<
ziWKe<edhW)rDc6m!j~TQZ2slxNmBiV)2<wAeH%>Her*b{l*Tk6eFelB0%~)?%T-%>
zz>{zrTwc-P(6n2bhaF}i$E-;rpUC#}bCvHa5Rv$`<n??vFl$*b<IHdTf8@PsSX0Ti
zFkA_P0)hexjRVjOqSyo*q(y;%1E`IFfJh4@vAYEXlu-<X0MTv*MMWHlAP__}3`%T~
zDG7mYut9?a4Hz+uLYTrlCSmBW0`|H0oOAE<ec$u`c%S#Z`v(s@yLQ#CRcorXR;?A%
z9^AJGmem7qYC=R@nj}j4UVj^o-79yzJ(js?QmQ=^;ymALx@be{gft8XjJa@`n;Y`y
z>+!#}hx02G)~5tRG=70hH|h(ElN5XQX2p@dC>(1Ypg;R>Ric-&!Z9=kx!x-h%CoKk
z&Jn%4T>mk83B&5h`pK8`+PaKi1<88xCFDc1uQfTQ$Cjmw1{4-*E-_lfKce)+3-=IP
zc9Wa-FN5X^?L$H2=he%i-*CR_?Uzwtd$<_!=uvFD&#zT=Vt0S<QfYf?FV~LMnM3?0
z|3j0sXWD)5iHSuMk#yQCs?hvDApS-4LPZLSgFML17AW7!eVO92;*n^NeR_2qWI<$b
zuKFr3|95;Sb}>&nW2IiTzrfiPD_Ux7HPq;#!j6XyTf1dVPQB=b?47#KL0-=W%|g9}
zeJ8DGC!<Bgs&&k7@`>d+j~7i|9osOS*-@1p$Yz|3@1{1TIxF;EHpVlPUf0$bbig7{
z+j$eOH7RNtkrma8*B9yHaoOXa5}ByN^a2`r(=X%$C#KUf;v4w9ISlr>{h~KKd9&NE
z!W7o2J)OkqNgMw8Ru53%Ude&ENFz<uc~9tM)K4daM|dUV;A#hEQt_>pQrebmlAMEs
zZzTsyWM;L=ySi$6`YTs1`YE`os%kJD3)vH^TRc3xPc!COy1&>QEt%9>$jLdW(LO-Y
z-8W^s{n`WdiZ2oV+y1-qlrsMlGCO_~pI-PTIq_FaQd_oGz@XW5{5HK#h}g5TWTg2J
zddp{2;MGd=-qmJPnr4%+dO4zJ!<BylsV(L6Y0ML)+s5NSq(%Pml3yZoA5ajQtaa}m
z999o+|0YcwCg&CC6Qw<yzK~%Vp;bx#hI?XG+@DqL-cH>Tz3_|A*u7F>wH_}4zc$&<
zqm(!$^4WZ{)>m~gD)1A_CBr`Q-8lxiIlsiJQQes+JLEt6xbAHoZzQ|NW$mVuafeI!
znG3-I?&ofcpK%Eh&Wj9_&Yy2N>jdaL>QxF0q(*Xk!VSv{JvrT{g0&)=Dw&^8GgOJ5
zrkA4pe5%qK1(fC<lbUIX3%k8nh4q?2lR42|X&J8+J%!b@MD6K_&B7h>Y}{hPQkdZg
zUpj7fH%I@pDJUeXc4nJJ3Z7E-!=}aUzo~lc<uhujxx=O*$28GL`l<{>cgJ?TT%B0e
zz&bMFox+TLA|p*RKd<caa#7bJn$ZyGTi>9N|NEQFDCMK#wEN1ABOX}hVV<*{g{5)M
zOq(<HQ`{{SZFlsIo{^m!t2*$MkC}oCq~fXvO)L9ZB`a~|x{w_&<uVsKejTIW%QBQG
z>nG<^SVf)3w2DPktn3&iRC6ZG1zWkiKW3fMz)ul$J3P-w-;|$Z#Ijs=Dxv03G3O?`
zdmW_1ZDiQGGnY26Mh1p@IB$uo9g3w6B8Mg-XdMpOP6t1;;(sj4AknW@c<jl!lTX$+
zL*!lk%_qzD$NX7}_9+YXq+&khBWY_+7|!-NA}_;}hKxjWr!Pjr$ri?9als?tWSeUY
zcW`=B|3^DpPg$(sxaeL8`A6XhZyK=pyRB)(Jyp+sJ)ha}97VL|4J1u$F~ZXZ5EN6Z
zg*%@e#B+qDP@|!xvAkt-iOdWP=4f6v8X%X7JV$Y~`Vh*G2wTs&*to^uP5r`IU7f;f
zkqms+``yjR#*TbJ%#E378%DLURrq5@ReYk<Xdne%7NdA-)6mEafUu2Mp?B<V;SVAF
z<9Acrq<}^?E`_+1e`)(+)Ft0)Xw){i%OvxaX3fB(WedvM7J7%{j2oU)ERLmDo-*$Y
zW-==e9cdRzfdB8-OEtNpGmF5yD=YrT_jtk*NfW!N{7HvS{62VBO~y}3vIngk`OR|h
ztf%R1>PI;Js%$x4Li(cWKl<Pe>~$XdL5U_e<wC1pZlHjVH!V=P!{ALLf~(%J0%n94
ztpp!Wp_c#GUmnqsHYuTE&OsPHm1|#M<-sm+O~}zD-0@|wBHQEdg>+DGw+5{$<L%V`
z)9-i-;yd)Cl|C5a*kdbSw9-YFCQg8jrczZ`rn2(&@O@d2C66hWvbN>6N|9|X{47q<
zyX8iDFr>^_Ol`}Y9NwwvQ}um+`;6%0&E-$GAd5X|h~-wsJIM3sppccX(L^is&wg2a
z*{S*Ec>?<QB<@+lGFU^|-6>Xr4rD!p`3|123HejDCHQ;FaZKBvk{AxZ$_{me;m7{r
z#^l%xvCEZ(NoXHg$cEtPd}R}NqKeds%CSxAM?^}$IB(?!X)4yl5~{*iU?i+qhSm=n
z!iB=VUseE7G)udbv;5V7&KE_DITBY^2h+*UT4?&uAo>DjSLyNwp=tvcY$Bvf4Hq6x
zUQDeMenEqOz)lGQIqtynFsPA{8Rsv!V@#W*r4x((Y1y7CMK0l#sI*T#u=z#Qm|*5h
z*Z<Vn%8JPsmj==(jUSc~fG75Ri{?$O!1N35eE1s~V6WXnWC;OI?}IF(irlhl8OlL>
z;$N^Xh#-2x1(tr~ml>a1*wL3sSXq?6TO<CdPkGiK)1<#2NdSb=&<Ys;{^IJtf4V5e
z-ZRT{l!XLf%&*k^H(&~50>!2+>ff#a5rz*dim4+*uuMqGJOPy91h`$2l;vky<1V((
zm%nz6h%h3|mCk(CnqrYU3v&hGsu>vBIcXGa#GANi;Y}=k60<MgkT>vI{%OYU&bs92
zO@huLakVl;QoT7G+h65H-)?qZGXK5oi+IyaRu7S}W|fIgRoCKSIKU}YPx0A|d+3_w
zMe~*}R4n217vx!}Y+V7_-!vr!uds@;yk?6)N|r0nS7;`41*I|Ze}nyR&@6(#p#@Rr
zzgXXYCTc0V{YSe~C<edk+S!%mj&!;)b8zL~r;JZ5traZ(dwHaEP<3gh_uuf$W9kcT
zVFtU;uh1X@@)t%F6^HtH1zREhed$=b{8BK>(Z_Kv`OAYL|6?Lf{_)1*&aM?4@czP6
zp8v<Bq=d%VCG1+k1A9!v%)#j`MgcJdQW<6hw+kL%vns!c62_~<4u&L3;s4K&q5l34
zS{#C1K)S7Y>kd!FuOKdMt%byAnLy=<>YU-fSE(>=TC%VL`-;Vg`B;TcR1NZ^<~(}G
zPx#xhah>bbJ(zoii5*8G=ZSXSMzWLpce9^Yvy=wOv}!1{E?S!krhb!Rm_lUL_?*f6
zzkp5}WWj{_m~iA&i#Mz^Ed3~?e~C9Gp&1VU+=Lh5`Bq)}fshSLISb{>n*BXRDV0Gs
zlWr=dfaiw)1<(GQ%AmaTzzM1d)D?a!#}mgC{~IGHnw8`%u3DyKy=7LSv4XBr*zL!&
zpHI&YN$eo`c;l(Oz7+tHJ(k`qSe1^5A#s0@qmKrPsVyr{t`&xK<93O0D@!C-WjHO)
zTV4o8I_u1riG2CuQRp-!VW1E>yRzQ@yjK6gO;)(nKV$KVXaWV0DI~p*kU|$SyuT4q
z+W(_}Al|;J_JwSzyn8^LmA2Fg3Ay`Ub@n&R{=w~_M)42f3I3)62}oxz6!J*RD6?S$
zd+`CDwjzp^C@T`^gN)6G{`5tg*cFgUC421{`A3p>h>J~o?-;d=|5;}5Qvn5SEk79(
zTS*rqau2pWFsm$HIkTzxyC=IZ@6eHtSh%#jgiMy_GFw6le~(4y!u9Uxd?VYwU^VO8
zw~QbN6s5H+{GWmKd}}dn!3ICPLS60u##C3x|6i#4Y1JE;6oi85kqV4p<_5HO0C5@Z
z5tJSHMb^*D@Px`q(6kZEJ+N2_FbwyG%<0*Smjg||kvN7Kz`{&lo!|#jX-&BwFh2I|
zs%K)Nbwlb2|A#|yb)06!0>*Le{UlmnzHduYDQ+y;W#5f&zbl$9>ujjS4dYGghGGs2
zEJ;WgH<8hR>Ni#S*453^+`MndtZ|r0)ZqGjbokVis;lKWe)n@VoSZ#PC0;W$Ns?a5
z@vpmYP+xY{z)O=>#{c-MJ;o#a8IPNQ*|yy#vWN3_QdyeLP3KEWpjUET|IJy8|NE<|
z$Wf;3!+MxEwSW4oH|+D`MgJsdm~o8^jQ=X}s$bE(QMLNRGq}iVaX_^wptV6fr6!DA
z@-8^X3dj@<J3T;b938?COL*{v#j%!>r;O51{HRFg`fC6g6)`4g7pU80lunJxT*(m4
z%Hvzf>yBu!Fwyu6i=Z%g)Ee5ly0C7ZdB47mxyLG4jo#}2>E|Q2^+ILw3A$lMS+{z1
z{sQK<*QHy^$4>bj30KeIWUtff%fI^k=5<W#1^;xI;lq!*;{oyA*(x2iiErp4!XKS)
z%BGLcnB%uPJ5s%uDt<W`878{a*}&4I&vdBzRQ-|aY~wo+)`klZ$0YIjiJ2kwpA2tm
zWl{64d?I5HdwJ25I8%tX^`(Fs^|BdiYQ2CO?>20^mDLfdSo+psiTV-i6w9nWc-)&G
z+S$!X_RA<2OGudA#>V3n&Jrf8z1kO>**gaJ5{2zB!CY&fi<FP+(-;h2UUFnbzQGG&
zmweF!QwoCRMe+B1Bg+cRFd32xhsRzMBGKgVH0kT_+qq0&p&F^9^EIrUqE^z2-QQe-
z?%YaJFxTrJob4F`yOYZJuVKy;XR%gfiX@hGONRJzY*=Ea$|6K+v#?v}<oDng+=T=_
z?e)a$MHIPo>265jz-*6DX0I5ddM+kZa?8<dc=AHDC5h(*zZSYEWxU>gh8|T?`fEm(
z@Kc*B^r-%b2)-I5h_m^jkXRlW!K0{DZ{(Y>#B-|4+Jat%>e(`>uC~y5c!P(!u&i*@
zh*()!$sjQz_T%4!JcRQ^L{EF?LZ$@B=b&IQ?-*j2yd;hitSdbO?*Kx`dOfXz;n^;5
z8%^BBWwM;@6rp5+T9j1tR_M^14(gkoBIy(9?9Y-lpc6F-pg0+U+PLD?g9Ocx(yU{$
zW2)K-HGKYk1QSCS17%02*plH%k^!V<m8EQWR$s+h99^vjfW8~r!A@a?$SrscgnuRX
zjSF3?rKjm|mPk7QAlLfW2LE=yxmuNmB1R4gsAr_F{QtO=&-eZJt?~c*NZ*Z%iCQiQ
zcux9AYp}NZUz_}YySEVk_b<LF(X2|8xg1`PaNTg>YhVBTIZa(-qoITJ(`91XA9Z)i
zs$RRn0t?<V+Z}ebyqGKU-|W&I;@v#^TSF*md33oWC&u3F-$U;-MC?s34@Ga)e|m$n
z>osDlpMOz?TAVYh<DDujn)3E_Cxf&d9rx#V5BA!9DfdA7A8j_X>|d_GCL*a*(Os;@
zQ+aQ#=$hzNJkHY}RX_1CX-`HTT-<D3wW^HPKdOlikdy}tqqc7Z2qS4E%SHT!MDm+K
zhDCohl6$i<T{W7pP+j(>Ex53qQH$(49ec&N?^?z_ucPI{h6Xk*<>qcZ!MVglMrk(8
z8GIVkX~{h?P{<ijbhucXr!%$w((u{o>08&*_C!62vxf|qqW=4fXNMW@vA0qBH@+!6
zM?QFb%&8(eLol(SAhXTRoO|_>;S}a-S+B~Dgc>G?-=}_#M_Wt!hJrL7Uf1V8KDSi2
zSy|(QKb!ov8I%IVH)&HP3qUT=xT9LWAas)51uW=Q7Qpu{y=asemkKK!E?09z7U+$D
z4XWo-8v|2EpCe8?k*8n{MBg?R9+zdO#Pfuw?|Mu7fSR9wLv<LZ!g3wNK>LGOdSw~d
znEyLaGEyl88VB1*aw*^G0h5&P3`uIQf49B`TAWvdz_J9KDUk6D3~tM?zAyqD!#r><
z38-S=;eGDo5iK(6xp9PJxULkZx8ggpaJ$NpA`)12V+z*!lNo})$~Vah36jntT?OCD
z8K%h`IFlJA<P~MhGiQ2eb)Ud(S2CBznE>*>Ev<D7n#*AXlgPmk=vCyK34P$i_a>H8
z!X9DIgywz3)*LQ{Jp>fI<^~@|YOtKqFq6&U>(eOv7Aoqw?9ie5dSI?6M6p<n*|KW&
z^U;B)6iZNB7D(&I0vcZSP|a+BdcH<x!Uu3;BuX`Muy3Rn{9?vSec~f|sV9O7Lz-Z!
zT9`gcU(lY1#4}cl1ASQW!PT;=snqJ3MY6WsC!&O;&ZGV@(tJ1Wmu7@#nQh2nWacj0
zQZ!fJCMtSyBWYtw&_{nxAujj@aLDf&zc*;>`5xSG&kr-#>&YLBm>mVqlSkP&6qVlx
zob%Bvj3j;=^5(`suT<(osZ9~F=}({4B2APm2)Qu9C3|W&$)<A(6P_YY%>L>SOPCx5
z`(U9_o0wh0!%Z_i{lEvTEiV}b4jI$tJ=D0Y{lLQHyl{4>Y*4<Y^6Hu98NtX9a0^5k
zcu}IE-}5cZ@09n_#8OzDbPKuuw2$S|fQWcXv0PKVxssve23zV0Y8RL`Ow#Jk(Vht?
zQs5>{Dng)XO?hc7#Iy;}qRLgnc*QOm2*neZJ6jh>h;7#^Qt*M`#w3sN5@8Tv-rXRc
zCsrf!UQW*9b+pl+pg3YkCDcxtsQMDK8Htsh@)AKK*M~#Cg>cmgzT7XWUZ2iR3!9yt
zB*_Of*UvMe#z%pY^CL4<`_V*tCHk0bipG_L{vCt~DEQ9!l-R@<!7`>w$$!|khc98r
zWd{!N`3*~<a(gFs4fyxllF4ttji?DjjSRLsI#x0+jv>LFa>7VIkm?M|lSBHm)1*xT
zYWyiAo}FA+#eI?$0o8+>%w3a&a?+-@)zke$BGUTwd7IeSHh{#1*u-KgPbpZCJ*mz*
zT3;b=^z=}JS)YUg;4Bu;$KCXrBW-V<;XOqd<Zus|L<u=h5w1Hf4vt3&sWx#Thg6#6
zj@U$YKa7t24DtZjmv+YIJTGN#CPoLy?M>0)YsOvXTnyu}KY|;PpY0#24*4B=^0)(Q
zCl~XoCK+8VCx1HD4;!~xG}%7|4DQ;`>zV0ww@ijjlJw&fP<zKsciWx=YTu)KOf1mP
ztcY|V&#KXfCsl3i)<cyFZ0c{e9~l4&v9D@pY{?0~;lZZGwmkuD0O~3MBa=YBcxbpc
zHe0c5UJ1K#adVvQG2~w7F4^?yp)r*8yftYx+@Muf-35kW&C{rjLAAgli@A&22tp=M
zoA*|$J=R2;b#pMWM1VBYYffCW77Z3%AxKC_SJen!?`RLlx#PKPNLX_ccC!nI^Liu?
znZwPZ*4?}kuL=wrO4`vcY(uUiLZTScy+@iwn!$8PUrbyJ!G%KQKL_14PUm^f#MR<)
z)G5X5^-0dwF>8~bE>`Lyw9_@N*+d6X>+%o}cO<Ki@>0ZBcI$|%Iqsu9wUzN(vlWg&
zCEF^RWaPYeD9$Mg7?IV11D&jbj<vl7^hTOXhIpx&>0s59$NV}MGUGkSeQ9&PqnoN6
zheA`1At9zn=nYF~+%}}sK_8`LnmATh9@fbn0@=;}HF*dub^Lr!5z{iDnqD3>2-KcM
znva4LP-((g>ZtcnPgLy${<TSE*vehpWY0zr=r#BOgmyoi9OB^)Bk!2GsBDex1DT)o
z^n2=Xo)AMey?Pj1rm_?l-z&D;u&w@OpC6`w0v5ePJ~D|Rwt@9y1~2=FqKWRQ1(JeA
z$_2_Tgj`Mu54V*{En6F$y}bkFrJdzIJ3~V0P#Or3_%t&8;@n{hNc{fKv6+6S6u`d0
z4}|FHC0s6IA4QGlW)n5h4bj49OdLh5UTvdvHEn!`v`wE^<~G#PKM_6(zQ23WD~y+b
zL>u;l?~h(*X2t&i4i!DIZ*5p|iXvu1wc6B^!fp`^fqhzO;pox5!a=Z}xLPzZY_sPU
zQrSo85|gRaFCwk;XI~s3#MO(2d`RlmOKa7*RTMm07cpS}!M7+qFgByIaxeILs{^s6
z&B1?U&>Wl~CW~!Ot>!sd?FlMk#-K5F>iI#Ox3t%vz|Bg_MIL$yYbUFnKIGufd&1l`
zB0<~92Jvd^{{jVOP4m{(iY59@N!a#Cv!;q@(Lxtuz^6eRFfER$R#`<=7w|dGO`svM
zJZ%o9i*EuqdQ*ZL<}J)8r)GD88`kyn4UAZ88vuEq5USIPFL?+0yo6N$_y@(!;F9Z@
zmezO=H5c&%0J$?wGd7k}qvFLx#8#rx0WM~BGUZR8d4dV^u-=m0n8>Uot?oPYkfLq9
zi0RZt9)8*D<8m{7MjUrcRy8u~aeGlN$W+2a&PbS<amVE3b9inRIykhpv#x^a<@^N`
z>nbCBHo_?9e1Ahyn#4{9E$@eE6Q?ZLDLVNmIcu{vvv$g{h({5w+8KM3DeMlw#I%us
zUQWotiS4dwn#ieH#654rPY)uh;`M%?wEg<xpY1HAad)<NRd0U$;A&CxjtkW2M9Lo>
zW#zNKHHO})!H#lr&CAfXr|hJ1T=5QvaezlQML$BdHV&R+5Wi8p<$&ppjdLPg!R9gV
z?Z)$?wx2u`TQ~gkQE%$5L4;hrXf~8fd7E8WV%C#68yoyTKD#%}pTC$~3&4)WoAkNp
z@=$S1bl6u*El-)m2#%m8*pBK%Y;zb3+3;K!02lQs&Fn;&ERWc-4|$3BDJEj=rvCSU
z5XyaCGC7k-Zd^Fg&cA5*B!PBcW6m+GVf{6gEe94Z8!djbWTtw-IE+r%@Ew&vbOD{8
zjOs1PKYLd7%<XC1zlMWx_*hU@ZWtf<6GL>O`ZFzWp-T62I}hiw&I^60_gj_IZ?hD$
zE|Yd8ov-lqeK<Sb&mZ;z+oc%rUtY>-dB~vM*TB}|s2ORtNq!xSG*gB(e_(d*^_){H
z>9e&CzM!pnHuaBE(*LO^b(;=cd>i5j54Ri0{>SgP_4^LSvf7uP!$Q*(+&WocbU<%(
zr?dgm5nGfE|4}g1ioB1=pXlLGCP|ym^9spLfFwH;(L*!6ivS9Gyo*2=HdpC_mGb<2
zm|aS*YAc|K%Zw)b@08m+aL~U7=AsuTx%{Lnx9QJA&9RkT2$~z$P7WxR!PFX%N72-Q
z9b$q6`v4SM?V`T`6ob+X&I6}Utp+Y8Ym$?-i|i0t%%gytGCK*@%zQqDNX=M-*qe8K
zJ}L+*ou2w)u$}}Iqy`xv^&9JQZqlPGB-JwJ=u?Wfi#7&5B%4I*kY6{0+=hmRc{^2_
z=P~JU3T7sj+Lcvy2&#$8#Tk+=!Zk$ejMqTCQ7!mI9|cEuhDCu~m!}70DPh76pgPE-
zt_=|YjPsK$q6@aj4w*R3!<;ug87bLY0Zb-&Sb5(f=8XOjw-DPT8&GphHn)M<wHoa&
z7->dGhs?Ab?94_;>qe4v;jKCi)wK-l95blC5GDT!0T6})Zo+TT)o{D8nFPU+77pwH
z$Tx0>%8tAru^MTPasW?LF}_fd*zQ(>9?b@KiZagg$Z;79>A=vKiy=N1NFu?JZ5)iF
zTX*Y79VZ)k$!4wFG>|9E1dM!$g%bqot#xlOBn5pvlt1%O4gVIfA5TnNkVK0C%Ai3s
z6S)o<O_t$SHG|9HJj~j>X&wZ1I%M!7cpHn`y;c`_$7HWaoAn5|fxL=9UmC^PK3_sw
zchJ8cm~k(X43e>%a*4afrp2~*5e6}zG-1H6nYf&^U2nu@*9c4n>$2r=l6@qZ%jUI#
zyIGmO`LpAbVM}(hs;ZO?G0{g6q4h~=<4sA<KFo{;#Qww3bf+!X*Hg#%01BfLC7iP7
zFNaO9BM#}K9tqLJwJr6CRh+EKu2p4aQOMjO;_7KGa|SAcJCU9(H1&~@5%~F9LaMSg
zNuIfDSkS*y?r=XRSq~i17#t#=7xKvHy)n_%GVNwf*P%`tIVPA|hs?qS77;~fK>prs
zqA5MDixDpVmbB*hVfMw)?v~a8n7X^{$wb!WPHHJh@%Eu2Y7In8V&*ADj|Rz7T-6WA
zTpje=HA}j(c@1oL0ZnC9-A4v#J$Z|xlM6*<Z1-DQFdgh@GqG}UF<K|9y3xhCSpt1=
zg;0-fZ(q9~{eD<%4Q)dy1}xQ?NmRTJ&T|r)-1JR$0XMJ6g=%?fOLYnFW3_4#vl^x&
zxA74P*#dAxYrMaKK5e3u605r!?MF<;-9>m=MfUmUUzm^R0_Aa?+mM!Nq1|;Cr<zF`
zoGa;B)Jhu8G_jmyXw1XJSn!;Ss%5nT<Ig(8Vz4e>&k?F&834tr4mcOjTcFdE%rA!q
zzbTYWZoDq?mfXQddR+b>P<9{$wYGHDB*)i++ny-vNWG;xej~^yGGQzs$pjcwknfT-
zlbnl+fC4yqawyJO$8xi&H|+w>?U&xT10vGZs*1Q(sb%rY#v9d%yyKrT2PZ5)pLPBV
z<d6KE@fni1eNEX-=F&M?r$Wk_XT`!$D@(ECbL<|x={iUpnMW(MZk^N&^wCuFe|RP{
zW(M`{&tAu8?ZGFoAX~J9f<ym&<iUmD*5}Hjgqgv*dQ=fRn1$)OyZGA~!kqZBb3m}&
z|J=%GJrrq8jL!gU(D>CWEvWWoPrsITL90fq$>g}Kj=K5Cx4ExXEg@%?B3ZeR)2lom
z-aQ2c26kgWgZS#%vDmB6RohH=g3U+%0=?UDgShzsRs1<Vqawoiob$9O3z5^Z4b@$@
z;FTY@Ta?jnX!uZ78RwB7%S7m1$avAAq;yCI60vx({5t>28CK?k`Ba<Ao?LdnsyjQ`
zTf8xFDBm}Ldo>&#=c<fteR783erbUfXJ2e`n9VO&?if5jDM<x-ot58Z%k`3bUD;P}
zuKDn~<|pDl2Qi);U-o`?LGMSF4$JP2t7iPI-}3|=`JM5XF70MBVHW6l%edA{cs9RZ
zAY99<;|hW#;Cubi$uwd&o@y({eupq1%xl)~YQZFxofAoP4bDI)$m|{I8K|<8C&B;W
zWtJ!GW#D7KTOe%?ws)GpFvCRL+LxyJEi_Q7dZ-8Jt<P<?rv(Izw9pW_fzQfKB%Nog
zJ55{|WbDJdQs)@_59;}mVUuCpH(EKyXOEVSjEgThd=ifM@FGj2NU)5?e}83K2DQMf
z*pQbOWyJyc!P&^5KCmY!A0NF2iX5a=r2%$G$Fxu$;}%>#+0@5HnORcFF1?HnhzWGk
z+V8U3s?{<rdg`7tBqU|^IWPPff9i!w!r$`fh>qGQWZ>tMqmmQWn(ma{_>#-XC;Gsx
z=lawq-wOp3u)>E+h|bqg_5$ERbqzCdOxEwIv6Iz0Ljc;HI%W4kb9rbNmbGXnV{^1&
zSrh_;<O2d)C=#es&4>E{39=ECZ2UUl=l_rrm%;2(w3B||k+evStXMm#(F`CXyQyVT
z?L{&1t${DqW)Ha$R}XUG6SYI^1XEnu<Pac98*U&PhV@hLyjgxG{Q>9l=@#<xOiVQT
znB3+hvU*QZr5!>m^>H(lfi4tWupI;+MC`re+(OLCSkn7{T-w<HVJNLhir9KO)=Qf?
zvr|r2L7$72>xV}C7|T#@|Maprhg}$WrA5vG<C7_au%yYd)Lt{w=Mxhh?ZRw8czHB%
z8sjwq;-G_fZM=>)Ef^@SVOWySER&U+dl7wH4+>Ida*3(!pCGm%=Dc|67T_cxdALJm
zWIMSMRn>9ZM~xsC4l}MR=cSQQ=Wb%<=tNv<KLtZZ*OUv39`}v`hxF9S_!qX@$7gkA
zoH|RLD<!0j+Uc|5++wV?V=?x*Y~D+uIj;=*V={f~U-B9u?7g25U)maRvytarHFcnn
zU3&`}!|bW1yP*y2v2_&8bk?4QiKzhFZCttuqG+Jfh;|zWM~tuzUb8x7>ts`m@_h*9
zS^*!x#^Z#ntQ+W18N6FZFDAo<L1_ZPsn>e)$%V8JSdy$ZU6rCu&@|awnTRlWX3@=Y
z_WuY!DeMG&N^f}BHY5?~^)Ci90kV&>J>KX(!Zkf{UJ0`&(CY%UB`JIG^C_q;xX9_R
zfCC|Q&K$j^HjZ`|A@AT%KSWdK6+)pcUIYj#{wB4xCV_SgG_<g>T-0W9DA<?Fen-0_
z1{;P>W0)4a)XS3#e&A8G1A%O_Hz^NsB7u{VHXBquUn0mu-eh&4ouVWxJ6X7zgVY~q
z@nSpG8qP?~Ah27;3gnTk{#1W-7hoS~LWFG`h)sY}*Q1lByoZ>L7<@RZ6)_c%hiuiD
zWkzzzDn&)QinqO)N5$ixy<)<FH$1~j+8D~KvRk3Tyn6Qj5ww?N-iieESb%(VBG<mz
zi$Jf2cS?gcjtxFBE>97<N{;;#h}y*}WUF*nxv3;NXTfyRx`eUxJ$dO?@{I#dpECaf
z+P6N(add8Heb{qT7u@Ofv4S?SzNUkpD#qXN(PgmmE4+3b8KeaaK3ucMG*}@fPtjFK
zy+1#{@0la+)t=BXIh#eojmfr^P2^6wF6j&Z#r1!`r+L*DToKF-IA#arj+i&S$He_<
z*6a9)j6b}Zd!<rx#$>8&N1L=-B-A^i0y+*&KoR)by=#?c9INAb=bQEj2y2<k&<&;g
z8vQ2>-z27NI7?oUgjzY>v>!E}{&=D>e9TcViu!JSziC9*f!&_xUg(_@OzBmq_ntV6
zW@&e{|0r?4Mk8at|M?{QH?!%f#l&6O86~JLcDPtD<6@=hZQ6V-Lo(&`__BWS)eDgc
zIb-%8S<%#Fg7Il@esI7mbo|dZb0fI$(LCXgR(O{E5`$<d=$U#~=XU<H4@>tr`{Z|_
zUB^9~61Mn06gK~4?8UAiJg0ZeYo-K1uP*6Ii`cMqtk(<T>qW1THx-?zTPv~%5pxcN
z?Wr$#@Xq=>vO+8~W|2iQw!KM|%x}n@F*)C~hm5g=B5u||#j4<*+EGr5L{6HdH!UCA
z9nksdYsPncw}p_9`)w#2Nwj05{GIqAR->k4Y|Pa*m-x3M9r8@LLE5%U$Me0sC(g<^
zyk~k73~TH|NVqU|%l#o)d5yx;{L})e1*j-vJtFc!-Di1^XP6J8Lw{ad|7?$r_FJ1(
zw9xb$eLpsqzT#YFrY<BM@Lz|_`83DzCrGO*)L&!3q7%QNC_4xw)Qhcu0ZR2V>+AFe
zwiB`P@rXcqZfvHkNuSn9EFNn25;X0CJ{)5krL2UOkzqSU1%KYq67DY2WNQBrue{`x
z$cZj_*Ludn<c?;>VWsn*e@#i5MCRlS{})sPD`<T;!)ri5dCZ(11FNskpmU@g<+jaQ
zjK$Odpp>HJ>MMevIoiqSBax&WyH`kS0%0a=$tR<$H>hfnO=8w!vB6!k!CCZxI?6G`
zE9>YmZXV+Bzm3Jr+|}i?788@qmu`8|vL+%4<yjeGVJ~p6eTnYCtpkUssMu@;gK>Sd
zPvq#3sp&eION0{ilD*WrfU;dx??nMn$_;l5&`sbvn_LU`;8peZ6F_%a_{c<Ju7ERy
zpk}Gy_VY@^2eBK#&CPp+FkEZPp}OLP5NO|qmAhq}{adB-9(nX-*Qr$xaKTcEtKPhj
z{<pw%-Ih3GAgNYm`W9s<@t*>8dnCF_AO%Y1TBq|}AfN_gy?79We?G*63_TG@?FGAR
zEQY`#-G5A4a~&aQkXH}=N#;C^<t$;Lvg5g&QG-M@^@UwS(*l~+MuJ*|(ShzxHk5eF
z@5-iWMS#^MyqoD7*@!vq$}aTJS27bgxI2T07?H`9bhuH~K4()wT18d1QiA4as69#d
z;m&mJTq4}Ob({t_+ao1*%gV`ehsfMC4Z65I5adIcP?W7D!d`(8xcNTRjK7mNZ?_JE
zVaCi1cM)M!pMtrWN&Bu5LEY0zfK0|%22kUe(_s0ShjbjWiqry~8&n6kOC7PN6pzl>
z*lDU3CA89>BCt$$@qOeS8>O;j_GBn(Tu#yCJTzqtQllqCjA$UA*|AqxM>errpEAg%
zgq%V;UQEq!NmjuYgYTWZv{Zv8gC<DEo<A0{YbYndw!R#5)T)JydD4c|%RDwFN;{z$
z>#}>N?2eI=eyZ8<F!T@&1e!O^j6r+zZUWpE(<^P?z{NR$CQR8xpNV@en-Y|t7Q}Ym
zi~kC_){PsJY@?K#!GJE&F|7eFyl|kqzIxqhuYn`HdVk><IHbJ$PBl9%qQB-TLcy6n
zJI;NMz>e8)`+PZtF1ZN#o2#G)HV?6Cyh2lZFBz(PXC9qMAMl^T<AKrs?uS|_q2?pD
z5BN-kfp@MSF@~fTpx0Ewb?^BUO9+D7f9LMmTgdZh)+Xj*24OL-7Hll(!o-^%0|$N7
za=2Ak*f(gpONdk@)KB-F2C4!1XZ(nc&bU=C5mvSKRP$cnp?+#DI0N}%oA1XgqGTGX
zcJ~UBlacz<K_eYNdvL-<9c9EJ<8}g$`;j4d)MM|A_&o0e2&!dc-Z#$(n`cIb#6T_f
z3L#*YK3L~mY#U;U{KXt?!>6+OiNMcGBV)RW(_?~$%E`1|Gre4s-PsQu5?Whm4_+>7
z#(CFg3?qglR!gC>Pa7IaOss5Jc5;V%{X0h{NISs3aecE%35z6WP~<gf)?bqry_d9p
z8|SJ{aMf;IIbFqiE;Lr!#A0uO`Vu`sb@?Etn9~9B8WLZcteNiV6QYZX>=dK4!nw(4
z0>v*A1Fhw}^BiYuG}-_ZUM1?G0zSv~e*l;C%}^Ut=~LXd)1+;-he83!<33cKB|#IV
zX3Fg2eo+JB81hZBdp}LtVoc^TJCaX)7U#RFej#B3p&oC5yPRwq5`Wt_K9}0_ZSE3B
zj2p-ZY3-SG|CI1;El~5VPA<b8OROjeEz8>&fo}Y%^QW+nkTZm={d#uQnt5)byx%j6
zp4NU<)lD(Jb=7J|^t+3o4jF@AU39#W2RHN;m)0gITc$E&Vk{NY`iZHnp*{C?j~!pm
zXnabkdYKyk{g3F+pJ5_`iDX;hx7TdPwadYRP5^HD{e!<iJS-0MeiMjW)cyCj{!jWv
zNvE%1^3mt7HGi5jt(NJFa~J*ool!Uuv0cw55)Bg*$c7hqy{^igzqCupqq*dIkCqA-
zFRoaRkukeWb8Bw!hDjuV_#<CuN;c;&(1%UFa%Xn(dWD%cRv2uHEHiYTs(tD8aBt+T
z9+)MW=TU`K==?d&L)Da#@_PT;#yDE=bGHAid*GNyFGWS+WsgVI5VjX)q`uP<z4_NN
z+6flzE~>)#WPXLG1=fl-*=Fz2QR=KvbHUptgqingbc+0TT0W@6HQHop5AjIMSqvkX
zQc>p}&z{sNYe0?6I^|dHp>iG@qk@9)!PqiFsAZCKIz?Y-8uGJ>Ldk`aIL8l;_3urJ
zUAi&(zZ0h0Q~o>|Nv80t`1dP12Gafl0q6i;C%&@+Kl9$({Mt*XxDCw??JTMIjUOFZ
z*Z)(7FEuSvM}Kx$*3r1+g7N9fbJ((9Zk|3(`09_GoYsB&OSo?4-6KPq9_QlbCnf10
zytfA3TL|`Qo_Bd~=1AKxwW;pYDGTzodyl_29~tT0;U=7rh%_;6vFZap8&7!q5$PgT
zpTf9!-JSbiQn5z-`g6}%L3O84gUxmV`cTf<?r|Xr^>}X>`llRyHDUUw+#mXAn(Z-i
z#EVi#L7(E>+NVpn`%cFjM-P@IbDPwhgrDN}D{U2cj5=j@WhF_*4mD}?3Z})|wrKjz
zJ;L2`G%O`h|KJ&N#~7R+x2Vo7_5RV3!F@cAs#;}IfkQ26>*v5Tl84?aP;}P(Sv6M~
zC)nCinIJF~XnJ0<TIi(Th#%9ND=@9iZJpJ(C6gYgM=pd+%gMikwM>?D%MAPS(+RQr
z&hpJsd$qGL_AOm`%rddDiv6=chRj7vrw+2%opBuH0Y#@(Ep^Stzv2A{6!Ayt*Zn=*
z!!O}IT1N!EFau0v=kpD@cYQ>f{wnCu69s<x<GXm?j7xgN@BehVkWm>|4eO^#BALIq
z8X(a;7Ex&{w0*xRD0&a{3eI)_65M<Zwmhknh%c9Vwto&NDjiO`Bo{*_hwasx)YpKi
zZI0ePY(6@9?Yf7cL!k|3eQDUp<oC&cW>nIDamAmEA#<5!SYx*Uv%Y-=qdLeP_r9wj
zHs`)S#OblkaFpg}x#!_|hU;8BJ2Xj{ubCdBo0?XsOg_4&uvW)vp$Mk!A$?QI5|{?C
zdBd{1lI__)k)W~&FBqJokMyrAl1;r_mnLMB;n^d_eOc5Z3MyOi#=-dCW-e-`236la
z5;>Df4>PfrT>ReU;-tQ*%i23Vh;E|9jS><}JEj5+n`$r9H1R4#m*lcZ(kgl^Au?@%
zxMceSf{NOkQb;!44xQ3A9S41)BX22n{(jb-Uc^a6Y{?)#{s^?RB6MZkZIs?kK*=dQ
zk<j`idv$kR@|^_R2ghW=j`pVVahp-8W&1|EB?a|pL|lSulEcXq3Ix1iAG`JA&GTsV
zSne9(?&O>l9`qB)+1r1#BFrxxTk%q%bj7JqF3A%`LK=EwAkd5sYFp8()O@Kj(ByR9
z#DbT63i-^Prn2SQ=Q!-D0pKRp-e4+oJ9#R{od>0z;#Ih9hr6Mem_R6^u9tXjiPki+
z93aMp%5C*+&a|4@36i2jp{C9x1-HW$1lzmS-hmUJ)agDP+_&TK1j}cho5|WT+#NgF
z)T=fYljHrZRGhM%Dw7E9!uz{Gp)}5Pwi$6GFWKqvLOae<Z7?QeyIIpT^3KCF)wo_p
z3^6M%)^5er1k_Gr4$&Ui$sG*P7n+%tOic~XkPMoA);e29Yo`bG385hDB1B%8vh{@U
zuXD1Bvy%%j@3p7p0UHZ*PumKK-TE}Dv+U_O92KhBBiYCpdI>mGnhlVZHBHyAm(K?X
zz_Gup$ey^3GNO5&4CnRil!Lar*pbNAval2B0X6PsU0AGx90W3SSag8sm$QAcmG1|v
z=rgmk&5FBFhBOlbEF6{y^v!x@u3&`KT0#yb1n);$<%wo_;bG_puwJqGy*Q%HjI`N<
z7e{}k1S@;VkEs}(CWxa5cBwAjq7*t|uzw1%FSot|FPw}3Zgz^-%Tq#!P&%ejAOO|j
zBMLVSE;_YJMXrb$MNykN1Maw{3^seIl4W=;A7Du*E%NVz8j5FG$WB=;vTHycKY`W^
zemO!m1K&HkqbrrjXH;D^18i4a)e7J|<s`1Z!R@E2#QId~%3pp`=@ri!mP7-6&InYm
z5C(@D=oOT*P#_HlJ_l4AmPrU<;6Nm)oemi8=@-?5BLl<|(Lyx>(h$>8svX<XK0p*7
z@_!fr$$v&{2aP2F`7)Jxc-m~V8R25y^8u9KZfeHb$sMA(4)#GLn9w&v6Gah2QxMrS
zCUVPXrbY!Lw4HK_pZ$u8I)le$!_kJY-QFRHlSAU@=n9gtUJpH96J{$ycgYULJn?NG
zFAwW!ZLbGi2fcRe=N032hm_qyoE7PqSd~B(u;Sr15m}pM&9Iin(iB}2aA?>+hJkBE
zEgS0<#=Vb>3JD&puE2*jFVN>m#;&0@E(yw3tXPPFyYx*<yD;t9&?mXe<>Z3b>>Nq4
z`$?vk1cnb=bU$H=Q?qW`oKj?C+z5-Mba9fMjK`VNzU;K|E}W@KwJb&3fRMJBPlSF!
zo82XSUV*WdBn6QX?N-T9_Xsr)v2Or|@<BWO%!Ozhdi4g4+??(wbYUGP3T)RiH?c73
z=vXzS6QXkq7^Igpi$hI!BF0(GC4xGP<=jHX8qm=s(%jN+y_l_-bWxjmxW3t+ofZx^
zKK4aV4*FQC;%CKi=;v~V`Py^Q6Jt#h(gq`UbMxV*WM>C(PgCEV>s+~aDXZ}rP_4rf
z;9kTDt-_3zLRSyfBdE!fl!%Waj-;rLcM<Ew6T{-jow5P7uYJ}`PtT4|jSK3^K?m2Z
z&o^InR`WMtY>=8hDMI6HALW5wAIp>!VT%C%k2sqsCVS%iNQy1Zgo!3PJYR(T2+_#+
z$bfHWH85*&u>W*Q>8f*P>{5HSB2<_g5~t^%*?5sm(uVA7u0rE@J!D=8Hy1n%6#+(9
zQy<TH4c2h;fX{g|)Qjes>2Wbht#VC?P@mfa_S}vN^m!?$D@({gY^8YWkIrX8!}5#n
z_U1U}=YYfM<WEX>h2^voz3!Te5=#O}qZ&qm#^y&anS<ltMUI}05@|E)p#6)}$B4=G
z@#|D~Lv3wxc22z2Q}TVB`9>FLtGTyy_0PpDTSJ??h|P8@1YuY^zgMj38JcuE1jF}?
z-+AJDo5K|owgQQDjbihQqPLW?&f0Z;ge>Y4rYPu12D1{kE$`~-7g)Y|$yga9zJ9&b
ziTz#1`?ib!FGc>=1dqfw_*UF=O4IuEr}wp(b)HAreP4h2E(d$J&AvHot@^g`&>oe+
zu;$`U@8S^0Cy5E_TU6=~?!WW3a^8K{TRJxyokOquc7M$$U$V@=`j(c0^5Wzl924GU
zKiryI{-8ke@SVWQ&}-w#Uo<vT1<i(5k5qMk(>s3o;T0b*yT$*7n%&D_`rrLWV{Otk
z6xXFznj}CrPc-->J#>0aiCN_sap8A4s(DX4n?tKN?u2PILN2pTMbj6P8erBv;y%c}
z9QpiSIU()-vW?~I4d>D??NNFOy$XNihhE&VxZ^Ky4v89#A>P%txI1urRBz^{1$}hM
zu%`yaxaoM-9@kyTtMaYL=;QKhutvSJq8gJ{K3>z5ZMrF<;+}VZS7s;f*=WSNOWfxS
ztJq3CV}5XOf4UJ$bNYax!zFHlZSarJUa7U+?y&2zyh*fvJUJh2-~C(sy{IR%G@r6w
z>zJ$k1GXW-*`+7L82C9VC$l`ga*Pua#~0D=HFu{)==U`5st9D<>|8J2=!Xv+{$~5$
zP}`q0S_^``#bZ<az5uh=__KWp8v<<B(AoUYs%|5t)5lrRW*u$vcC$TMAEpedC7i|Z
zO?inXT3Y0_!7|?7unOth3h!zSXg>P+s$1(eUAyBx>;pZ$f@W+=?x$3r`O;vQo!HIQ
zDn=!0A138`7xQBxZ<LALpLNkM_0asw%I^B+<Bg}hqZV1ZtlQ1=^xxy%_~GZ;izy?w
z_9!@4*6xeR84Yd>Vn6G+G+{tb;=kGaPBs5kga^NcdG}3)x=CR@Ep#3p6S*|kA;2n-
zDb3f4x)L*ZgVV-aXGUzlj^CErd$nUMXW>@ki@(6`+p%X|?AvG|J5SnHz)XKns~W2M
zsLkgwZ)T?y$a5Daa5hz}?q`bvnEb{1>OjemZBoz3l3khDDypCpCc#7f>^;whIeHH5
zk6pU0dX8P=-Dg3o)LeM7h;n-pJw6j8H2rC#v%<h|)!Pd}Q-<7BpL?2I)aN(pqqg5Y
z*;j147{pZfe>}I*9=oH*-zdD`tmT`K1Kg`b*7FlZ)0HFX##=k>$7YO8N=?I20bP}$
zmco5gpAQY$*f1~gecx@)a1W0PcD`#|Bsh*$`YonR<HlADHtwN?W*}j)=<)~5$nm~3
zD)zKoN1DGNy#2fOZ{!%6#ilN-k>7o;zRbJr->dF%p@4s_;8*Wq<AM1_++jMUV|xDX
z`PilGVX*|s&4WQMF86Qx7CYvBvj&o>p<(_)+&(WdE;-cnMkW4yX5`BakM~ZXF__O5
zWp6J8_bJt8Mr+m^7YzHo9ZD5VXVV^y%_!u*s!!tW0`F?B(C03v#THJAViU_sy$Qlm
zw=-sAldBV3I`U!CHr&TQD&2Vde=%{~=eqgYgLCo94?7D!;FOcXU#DiqJ+mb@>@G+;
ze}bRxJvd>Gw}_|;zqRw}@xF7~N%d<s$v9o0g^$SG7rb`AhP*MtHM2QCH2ZCB?+cpD
zx!BV7IK!WfmQDYwcEV}+Cl9h>c=+UMM7F#uuky}MBb9fm`kw<2kCJC)ww)uPvL4?2
zUG&}x{MqjKb@D|S(V}gL^B<4spd9R*Pun>bz4aL9%o@9>n?aik5xm>Hn&Za1I>B=Z
zZj#BOhw7P}=V1*%Af<xW7?->gQ4rVfym%CvkM|AMCvj-DyH8}xQ-`<>AgIRqStPJi
zH+-|?STg|qXue%Zan>=#0F~hmOETZN=4~_h1{R^FpDjmp_mDqyL2r<3%3v)?hL<t}
zHd~HMz&e>;pk_})#b&jE4Z&09BEuRc!bQI;E>^0QA>e5SzXdh%s84AmJ983P(2A|j
zwND%ulT4;c;|+l6dAk1~6S2WfOv3@Wz6W#tu*&c8Dy(%mFvZ06N{33DoAaI0r>G;1
zhx*go-D}Fh2N&5A-i0_ka60eN*|?y(9(<@IU+VWAKbp$NFW0!z5bN!o^B$gGCr?t<
z+|h2*9&)YsxPwFd>LlmQ<pGnDYQ1((GWzN9)2Dsw_V+v#$CIBbx;+k}qO>t<1;9<F
zkYqSj!i=+1JR-I6t_C2CT@gjn8|`4+=^0?WmU=f1%UGYiR@hNG;~izV0Mv$(vKnph
z{MlpjMdXqx;4>!yN}VA^UGf%XSYez-70QtV%T5b%2KH5QFNoFZnBDD~oElX5otk1=
zSz+-YQs^6F!bImHT!YDuB*oH_`6}WFV*q3$45?cjv5nTGby6v%Hv1{R3^f{$A<1$5
zcZdLl;$pVKc_$`*hPMtW#s!80vn7;wG|+T$6otXU^Qo`TQIrZqVB<iV$^6J9AC^Ye
z(`(hV8u`e?lTankQcpO&NfosrBFkFObyxY#lT6;7>#0h>U-5Z!)xqZ_GHB~U@nFmq
zm^9SRTQc0|P#OO<p!;XNWyu2iIoA8b(_*XSmrBHusx#*1eL!_UmX75$Xr)bof~zHG
z?bc^To$;mFSe1yX$ylX9!AsIMq<_3}RnSOhJb;@Rco`&eFfgbO0O&G@o8TaL!qpD>
zSeu!Q9oYkPxeRI)IOyQuUqjZ(2A<~4<yo0T)`4cJ|Kc_;4+e-aqlFF3OlXo6b`j<#
zMb(;arwLi<I?&C38;1GQ3Z<sZUra`qV*R{`U8_4j$F{}7SZTp0klReq(hBMz+Sn-`
zPI@{$6$lCw?$SBu+0Koa?A7hlDOzb40@%>7{K{G~?}+mZ2%@b64V?Ww`9b|s-S2jn
zS4zAuEiR+dPN7}TRnt{Cq$``7jB%kclHbA5`KY0yXXl&5b3ordQs*v*_O{LmD38k1
z_FagKfV;%&B~ABgo+mNNK7j0LbBhQ)Zh~4r^cEip<GRfU`$M254S+R?^~vvPMtCd}
z27qZ$t647B5uvac*ps;vnIr`>u9JFE`ph8QrHhQ{UV(>-G8VcNce}Xj^*oVAK-lLm
zWGiK;WjZEv*tp(4Ky~MyFCT1RKTU>V>u2IuF`Qyz8Kia3)=Z-twJ}jKUm;gtFEJZ#
z;yjlu(18KR&D^}Zp1~>Od;}Ww;FN&;FdITL?PK*i;^8|_E>}LOq_S@zo%TisdPVS@
zZK~{$hZA3#2~Bp5*qY`cHoIJ%Xz%Gav2b}0?s&TD6k=TRV2H-yh8bS|R**H-<&gOv
z9T(RGPgR8Lp=G^hbmGz@rbs(P#-EAx<x^eod@37P(d)D4LUdH*S4gL=zA)Bdcxndd
zW4Bzfr~7`w-qdchTQ6QB$pWC1h}uesUaYLtRXn<+QJFCD6|x(~e-HQf3#f14K~3#*
zE-5n+TuiK;QcMCoa3cp4rh-Nq5GQhvG*wV+4>f|@`g-UW?Y!cvGX$Vobot3>rf(hY
z>qY2jA;(>_-98@<R=+iRn01+x$_n00lCdF(f{N%mGQ-nhVBdtWr|5ExlVe%qsnup&
z!KhiE9SnjbL)R0LgSDP7pIroyh5UyN{JmJ5^mc^PxJrFiWR&1dbo5`gfp4l>xMy9-
zSUPfV={Dv1pLgbM1+gkd!xQ4|BM0uoP&OOK{kA)=`qHV1y2qe<F`&7eY+^}~;l5_`
zU;g;PhkS1G==;TaLYHEB-uy9`pyz(0lk-{^TIwq~qu+@Z=uw-j_-*g~8YcV6Ix1bT
zl61=KtA5dS#J~@QN%v1)ykHo!{@X{YPcydHJivb2P#%(TpkiS5a^az$uYT<=yw`dw
zVAyi9KK=S1ltBSH_t14d%|tm*c~@yl)afYI%dD{<iyAlYk#w70u=b<OM8C4!z;~A&
zRJ+6C-SjFwv8bu~t^UKX9s8_2@X7<V-R9rPvePsAdb0g$e(!zzC%Qk`&rhe%(R}xt
z2_=oO>mGF)g%*>Kjy?N$&i%~&8z*y{&b%zxQs(^Fb?|Pw;nr^g&##_%_?YW^iSYAe
zA4AW(+F!~^{^ioKe&yr+N86g`MVsCJBp!S*PE#JS45WnPEH8$;owM>v+F-3#0keAM
zg`V_o+<5ZZ@0C4jz0b<uc@YjgE6uhwp{y2$7}(@#wyZ*>1;iCQz4lY_DQf#R$l?dJ
zZ6z6m+K*vFyUvIb(06u(KXmwQV12>m{TkZdE#G~8;`9b=VFc#bPtAR<c_*^B(8kdh
ze0jf-cUe^Cd`t=9T>bT=Zy+tGt2S+7b6{p{&)e1mX$K29RFmHA*D}o&pQgWZ@Hleh
z`|mA!B3o~tdUL#Bi_;abh$8*V{QeKS-5jwP7XNX<P65WwxS(?k=bkNmgCG*uI#=Va
z1jOm;$k=Y%dM2~_q8$vzQdOPo=J$?>cWHSX=TlUcbgV8ZIltgF8+V5Dn~^qa;5nr&
zr2O*-tJa7v*}T*&B!TYZRrP9B@2h6tn+c8|EL*R@`LX6y;(N;0N6EXYHs9e`->3+Q
z{G1;mTh=hCZ1jO3PU3$+aYQz33E4jt>!5PD8Yl(nH8m5hVyxuT8P8$|)1SmH`l+OL
z%j3t*VL<@R$&rMqqV1NKF9mpwspY+UaQR^5w|%u#_MApe+E}IK=cDOt7xnRn6n=i?
zw2}K%#=b`jhpvRqD5<x6{r!uy#J_-t)o*DdKgZ>TX}4T&zVhLKr`e`^s(J?-c`t@1
zzSl6%`AM<e?-yNhg3Gb?!o4Y{axDY%X1{e)XB^rw<1^cN@);{I?WAYv4(9|~fTZr%
zk-!|mX_w2~u4uyJjGIo^N=xU5F^?Z>Jg#}Ox-4Xu=*cU|18+ev`az@lWnOFbE7q%V
z%wp@@!uO#cvV5s3v|GL&CIuDa$3I?Wyh3lrC&}M24GB57O6!?N$wCZ$@y(h<NyU_g
zu=r+YXMlwVe*50YNI72el`)fFJfci7=gVZghEfIbI@vHtL+`wKufvZ;%*wbN_U{hR
z1%ADtU__(pitlIMejeoqfk$x!zqqcZA>r#6=6Wb@ckgU-m=N`>LJ}t^z2L*jaMARn
zn*@bBDJs{5H=R*OU1J<q@-qqi&H3@`^1}C<+@HE`JL`OB%csg8XPe5jo^ZK(Kk@gc
zraBqs2`(e8FZ?;Id(`1blymZyQ+7#O&vngr*WEw=|8Vz~VQu~Uwr`-NLV*@9Uc7kG
z;%#x4kl<3hI0S-2sp9S$9Fic#f|s_qTY%#3?%tmHUu*4q_H&<e@B4l7W+q{h8Hj#=
zV~o!b5qtL2_AF0cSH@YXY`l_|zmRox_F=`^v)BG@a$j&_^JG3>44b5z%Q(!8g%g9f
zVtz;Y=5pp~TC4bmL{Chtu879MV~I(*<~IETz87)k7K{ekGCtTkP8r+JjCG0E;WMp0
zsAx|!AZ49GPuy)j%Sfo5E#hTAOs;s8b*cw*Zd*Z*MNlf2(FM+av_f=cBto6$8lMpT
zOK@R27wqR*^#IVqhHx(Id*XYuA!94anmO8jO@R4_6`X6^IAVK50Qe}XlI$i(=F^H`
zP8Z((!SXYv-&%C#Str6ym&8f-PEyAGDR0^+O}W;<g~D5>H#Y%qQ+cZ7*23W}+(TxA
zVb%u~9Weeg1}R|x$tJ!%P9w1De>8A59dKzA7TU>^?A*VFOlIljE?)@?&zffZX$-&Z
zz7qAR^Edoz##K0TQF^ME^$ulba`xTM!O^64wQ@!ka5%wo&-Rx$37uC5u7?9&?#&6R
zpDp^|VD{#n<!fG)RbMInu_SHS++IUE`1s2hc(KhEV~*rGHWK?ri+qxYSkFzIde3Dh
ziFLS8d(cT2wGKdiuYd2O_fxu{v5tvdm?jLkvj<i0n=&_uO@30=>ulw@UVqZPCa3It
z#cZ7%XC;rcH}+ru(J@RZWIQtCwVM3I6;J^deEocasdEdCjNl6oo`86O`d}pdp}<n$
zn|vd)$dV7*I(9xB6QKvoy&7+mrqC_cn41vOq3g#(T$vt~#u$L#Z>r)k6prZO*$Ek6
zL)X5a-jP{b1dw`ftBh)#o8f7xCf1Wy$1Ojy0O0vacu7ka=F-b(Xu0-0e(=3f7hpVM
z&_Wb=)))<zXE*Fjw?s#dn#;u4^D3^6t`NUcKUQRajF@mXwbHouS}_oMULV7yBf(c_
z#8+FuSyodPUggy>e%vqlK!_^otK!-je?qr&l4%oVTN3cHHBR1qPcn=Nkn@Fak8M#m
zh*4}L_iKAP9-B&5ieO>9svaxF#adR@eb(SP2_l*M%oLl}v&li+Xhi%6|2gwur@e0J
zED-X3r^c%n-CgYiledq8VE_5>c<7RXYv7FT>27Hc40;ebgY;<Le;<RJ=hz6_Mh7uf
zkh=_)OFtuIGJs9CdYL`B2Svl-dyVn7x+@xhX$6NDqi$6T%=WI+>v8+f2xqkg57v7F
zy%s?3180-t)1ZG1ydYfLme$7Y4bZPyTFfA;8R~qs(Rt|Ztpj4zHUtN?Jq!NxJm?<k
zV7^@h#EUAMDUEXhL9A=n1LNP_Q404TzbOG_-wS}Z;Utybyu1+w{iBvW2zQ&s;vHZD
zfFpQ&!((s_c%0__f@oYCKi)PRfVeR=$IH%M#z7R*dQn6F^+BojXU5<6DyS#7-o2aD
z=JvaZ8%Vbx{(a^TiaDQk7rtlrpeW2>%Wh-o!o6H~kT9U80`pa(L|LlydeZ9m{-%o{
zK`W?yO}Sse?hLZRzw#oHq<SJ)`WCdJR#)yrE4<oTrE|7_=Xok>y0g`lkFCLN>0Zcx
z9UQQKbIwty?gC3TK%}))6*aZpT2@+z<5cs~b#U4BF66szrPH;ME3Q2|ZNv6kP^sC@
zzA9(n0zn>BPa>De$gA@lWLJurd#Onn1}aW!*qy@cWBvkP12MPTo4uLI-hd+iAYEhz
z=QVJ|nj4A*0<?dkr(^&78sPD?cQr2eGdu(M119yms(ilt*C0XU?Y?Fm4hn<Wp1ToP
z0D1b~N-sPtMmF5`x+ELb1_3J_wbJ#V7sxF2$r{mfuOiUh%JQV%y(=@%Aw1nOUDd<M
zYLG(dg|&bDC_Dj2q%>aYa_SE!s5r{$Xey~Y;59L^J7!G7=w;~yBI*Ed434=l%Lke0
zaq^7$mHVB}FMtNnalL~)>0TgVcmr>HH;)rw>@WX;lMgj2Ym*=JU7cNTk$w)UHL^-R
zyadKZBZ>XV1yDBV4t}^FIv$=pvD}A44t+W6Qe`pS(7hHqlZ_u%9M%}?7}{zBw2I>9
z&fD%Z@x;X`Sa&^wFMLOULCs!45B~LVK%4x-Kls4`V~R3Gvz4jV-7Mtd=KjvnyMNr@
z8vkH(ajO~dJdf9Rduy9;-T*-|9Pg0u<P6YqC;Ho;s0F~x4usgUny)~NlB%f)hVb6q
z-=HtaW^ZxhcqM(@Up>Rkd-YLHzOdo+Xy^y-n1Vl?m-^H*)$15UQ2nYwaCfn{#SGAn
zQqjArz&q0F-s%h}yEmF-miZV+KWQj0lo|Vpg17{v+2WUcT^Emno?oaaA1+SjxGf)n
z?)p!R8@e9b_2Ed884}x-OYe&L!1sV(to30z5!@cSYH9+J#f4N~NY=aUu46%ocEBX5
z8EyO60;EVHa2raAQ1`inpAF%#9(dKkMkiLWINBSfk4*hNc6UHvzfPn>S(+zBLlDa;
z4@unWI55!6t;qSF9*ssb@L_`PZ6+0(Ou>$WIEIbu4z`W}Z*&2hV!2m&5XjesQv2-e
za20?wkjW#dvcB}HiyiOmf;915N5Op2Rkbm=!cu+@5*vYm>=QOxXIpAkLUr>bh$X?s
zm~?h9IUj_*Aig2pT`$+;JqBsXxAm&~oB}!Q2Klx+yH<7hX#g<W)om$Hl$&@gfj&t|
z7W#X`8FU-mj?RLZJuZ_;wpJ&mi*fT7fQgR2P`+~<l>eQiW@pLNa=~Oy>+W_-tGdwg
zk=-^(r=9&b5JJk8T9Wb$Z!yi=+Lku`hW8DG!`jx>sD|T4ODO;Qy(sP^z&;M*aw+dW
z3gQJlYq+rMT;Y`t!pmG<?o<bI!T&@GTC9BCch>w*06&^;vtr*E(r&Dr`PdV5zg35M
z=#qiRw(lnG3Pg-7328W8UfI0@-C<h9TIT>+`>O78^P_DL#ExffphY?eR|gmX3;)R$
zb20Ij1pxQ%U(8@oA_L*9n12R{EC}?d`x^X@5L9N{`dZYa23GL~uyo(NmYbhxxzjs6
zzmjUU$#KzV_p0iT-W5B|nfFb?Su>!D-{(1MuyF%S$M?Pe^`1`&j%Sh<rvlox|1smH
znvniF$CFf6u<g{aZ7r#DrS3KiNj0;ut#|Ju1E>>KX8L1Q&}Em`_QSPazwUQ5CzC%$
zY7kl_lW=)``AJSNU7y;CfA`8pPfyyIIVokFB^{i<GXt(emZUY0*V@k1xwCz}G7MDn
zZ4;*9)e3Y@9hjob$jnZjZ@HPa9NP`oS?Y;pn!{|78s#spCWK0K8#^!%6|3twHfrDd
zStuDQ6vpkFs7}Z@)wUH;dim&1A&4NuI^MBtR5IP1jsaJoKmAef!-J%Gq#hTUA8(|?
zVk!78bz(O-u);WDrZ=gl6q?lVMMd=}Lv`#vjgvFAzN~tj)@#o@f8OmoMW5U(y<Cxg
z^cQp{1bVd=b@ux&NOOQWP<El^FUa!s|E>Gnr*Qdfj;Zw5%V|OMX*m@Qy)#@iOi@-(
zNI)0~B7gf(IKt1w=w&r1!67(Nj&18)B4Tyu&eAn6{@=q<$8VdL$uMrOh<W`g<7GXj
zbl5z1?;*7Vzc}Zp*KI+@h%z<p6ppvRgf0yQA2NBO{}B9f#(bVRD`a1zm})wziBHT>
z8I!qgKt9a1_{url|IUXjg0>R9<Ci;4Uf1y{Iy70G3e8FTP<mU-;5(}V6mQ`b_@?D8
z1&OuGy%*bjb)+`ngTJ7M-%wLu>?e~-Ss!gY8_jiTxIxBc>BMu6HME@<j5|}GDsKwg
z-+Ja~mqH7Ak(!CDL(*;k`(yZj<hf<^X*=+&yxqX{SS%0B%HhkbH^Y)sxiXK+0>rdI
z@cN8Ecv`_;NLZtYfu}p3xEp}U`5thuJdOu5{cGXF>opo2*1o3fMlurH+~^Ok*m~EU
z6@gICLWj&`XMfL+1kq#VxU0p*$Zf|7r+<0-M4@~~cXOF0bWi?*0%Mb%p+*MUZ#v9h
z-EH(>ItfkO&uj^|JzB|<2zcZlD{3}Tp1P90y*Fg;$#e*XZV&c-`5A*}<UEL&F4yu`
zq!1+k@S9yERHZ#byXxRiKt~^vOsQ7bKxiGsEo<~Hkx4+kR@g#d)`?HDxg(~CB;K@8
zvWVzylR?<D(2j8M7Q8xsXT;9#sF&^;w^4~m8w6%BC<k?LDJf2_SRR`(`1)<o{PNab
zCay|+_M?WPl15l0M3IZ62Ni+}K|3&&*$%q;uJo9U2F}(}+}Hi4vd|VDl{Bkzf?k}N
zw#o5?Y;wf)I;qyNX!q;PhUk#}de~i5q~Vl*YPzx7mXP74m=K%L4t|=Jsh$|NAvEcW
zrTQ(P#f&yyZ;f|s&&Dv-nhYp$1Q?-8qa>vuKNIbc^-Y3evYA=Ygt6?)I!vn4HY`$}
zd2~YGTFM~fV`KWr8>P^CqK>{;dNKo@f~J{wG0V-IkYKeKtoQJ=PWkt(!CfP}YGixy
zB`cd{+f<pJaH*o$fNn*pJ!kzMv{INl1hp=v+ZBqqD_J_=DAK$S6OL4r+JjA&+Ot(<
zpzDF3qZ)~c>KoH{N~?kgEzot>Us!`9)Xv4WD_(x#>9K_K#;UWWeUBVdQ<qm!sb%yo
zia_FxuvWP&KITjMA-V-C=^^`0)aGV?XrBKLa+0U%*Qcb??-4d$!XqDy{tK!l-@r}=
zI7U+%yM1cm4TTzO%AP;hmUtSQ4o-pz@NQ^WT5v<ckZH0zr(-QhVjaP5W1sZi=AthN
zYEJcTU)%D`)w_yAvd!Yr>P&Mb^<ooLA!8ycPik$xFGonnn<r5mv$b{{(Wq*0yc*Q_
zDf%59vhJGkm8L{}$=<VJQ2SZ8T&Q-!&eVj&ob^cWeQ3GGUy%OVyg?FFkBRR@-J<U$
zt8!9VzpmA@5u2U3s=+Ai2fBgrDsu^knXt}n;ZnVq+EZO=XlHm!bPBWAqLe0WX@*wS
zwy2^)v~D7_?SZgdqtoc}P`F{`nwm%q20zP?HbJzThXYKOfaFw0+&v?1C);`R7=KWU
zv$8HuQt3&yHnU1Ir(146Pr+hql4|pQM7^xsAJ*@^JS`|aQZ1X}L#_i(ScTmabNNV$
z+}KyxgH)V`VdL}q;!*{XtrcCMZNagYrzKjS;nAbt9rYJz%QwFaT(z~sL)>{EXgeA3
z)A_hfLG0tVWf*)WdBmuSBN^R|<Gv`$SxYK>Xn*u*N%Z$eHqu}jR14{wuoz1%BUbWL
zeSY%NP&Bo6q=%C3*R?ix1e&k};x`vrnI6cTEfD$mi8T6I5BIx(63+NSa{(}u@PNgZ
zf}U_$U;`Hk^?QTzD>(Tiee)-=yxu^ISybnb9g9kSdTV0BM%@oC*onOi(Y@y*sSbsZ
zr=PT7#7K#-v^HCd9HUV93P&69Uyw8BkCcf!C6!Avd<NMvK*p?;8DdRA_Ar^rOqaHm
zj^a*MD12vZlN3rmg;7QH2fG!Lw+9~dy)uqPwI>*<$tH&1I&vA14}2n~qO2N+XpEKU
z+pj9*?Z=*ne&<jRXUhR6QWHFIn@tvQoJb2NVbR}3POPZJUB&bg2#i3Mqs6O5rk)h=
z$i)fLGewWP>yOtD`b~B41OJR@znMr&1Hwr6taCB#QU`70bxz@8R8gGR!6N@KQ!`_$
zC3Fzj#la8@)e{?l0NaN89*o>{(m-NMwF6_{Oqn59zXB0C94XCNs0}-bft+26*g02X
zg@=uq3^tr=wTx&9u1p^EH5WXRNi&INnfgpeYth;BozgMtrgSIl+is~xQplOQnVzu4
z55=ORdFCH^b~=jS!&x30iV+%eCT`ZTsP$hC!&B%|VIrD>*xEl+?PHLJxo-D*GU`N<
z&HUHp)52*cKV?xbQXy0AL^&j-%Ej_#ca=VAe>@3^OVy5z6>S}01`7x>Gc#_Bg$R<k
z^RKL1FL!IGwCU>z#A{i+PW?bG(i(}&Wkf+bdTBad&Smr6{hoy`jDQ2FG-P9)SHbi;
z<pX(aw!+s&CYf-;0V^ek8rlg?`be*(fv5VEiLt@wnWs+EdP%HeV4GjH4&?V1A$y&h
zcu%*uwbR|C$dZ?7yFx^gwpHPi+?7AYkGd<as&}N!(4(j^dlQNx?LsZ{Uy+7caf$vu
zhLERMw+RLtKwXZ3<@NKhC<sKM)i0~m!hs=O$}DHpb~B;j3s^(Lvx0nxBW{#`66d%g
z<=e3NYdsYIu<+6^OM{Xqza{<%MX|2%jg*44otn3ari9JmqGQ41tUtkFsl(5qQ`7HC
z%1HWIu$^|X)TshOw*7M2St)%U$}DxoYg<N!;b#Zmqul<yj87>xCtxsZ1B1);-3m87
zF5*f0L#DT79pU<}xzhBgvxv_ZQNeLi)Tod^1;$Qq!G~%hdQX<bp={QDyCiq9qcl{E
z<p<3}@dgLKcYr0^nzE`Is5ZsC#;+Uo)+VbOrg>2#@1gQW4ag++Vr&{_8~qv<z;hI>
z_0hI&x>WV^VF3wqQ?HANgG{7ExH4ivw;0+ds2!=)<`H^D2foAD7Rjs~wCNM#x>RdO
zGcyL2w^Eqx^D|yUnech&oY|50XfPK2L=CO4qC7SuojaU$ndH6@KSf22!3><kGZTp&
zi!qhCDIQa!FH-u>dN!PS)V;IpERBbUYB&nxc@c1PljnszDl^I+#9PrA744)g2G=^`
zVo$_mZanV$JJ<4kYQ2@_F+xs)4`bnA=Ir{Dn>Q4Fa`KSz(0e4PT8Ro&qrlc*Q1us>
zGa^C=kCjnmgP2qN&Byo`u~!k1_;|8Jc@#{u)TIu2%`Ipa#nLlmW(|q1gO1{NxusFG
zobx+7gDr)3bF<RRqg~?f@wOtrhJ}SiSRWhx1>y60sh`aUa(cE(@a*~lYY#<R#+X@2
zourXtWj;PLL{*pROy>#PHca)P_muCwG>r)D<8IHnzaYZy@!2A&2PdZp+u9Q2Qut&A
zb^2v_+t2rIQtP|3*#nh3s_b4fbg(wsnbqirho1dEj8!{ng})tsda~)Zt*gOc;w2=`
z)9J3m2C*Gs?9G=?bE;(}<%BduJP%<XIZ1tyFVd?)qe2Zasl^!Vw7^FuIGM_nl1(4h
z_{IAi+?Y@0R8@h>CePS-)=6=>r+>xS%bQNe`?ev+Hu<u<l1Y$6L1P9Un_YrD7lP`c
zq0Fo_cNL4%Q1x>+B0Jk|MAM5RFU-G6n6K7{Ka~nF(vJ?NbbM($ot#*tV}X8yRxm%$
z?KZY6rDD$y<&+GmvTR7omrZCrED##q75%(6Guq-cF~lJ7gL`o`T&T)L#nZH+<*}Qu
zGhJV(X7jTCY1`N=reF}}pp}X$;I0|E7)_YmFt^*JcGIS3DLE&AIyC_(0g7oK`u!-;
z%TK8etvzL2%NQmjeUo5_@)HXv?{=kq_6TDSvSkT-u?;7b-kMMPQ~GU`yRIe;;!C2H
z_dEnWVf?^9{z=tp{LS-U#v+}LN_|AMlXb7cHw()N`SF`|ed%hvYxN3q%QZY>Uo7!B
z*h7Wir3G6F(p7hA^RzR^j0k)2_U7)YayYrr4o0g?tyh6r4_(YFs}LW+<ql;PGTBC0
zYUntsFD%n$EuNYcKdUcHjpG|lU>=*XjRBCb%^{Nf=;LH>qb|QX@!?TQWKFs)8>uDP
z45-cVw|M|M0y3Q&Z#`JS$yukT7wSGiA4;FB6tLkzB{YtRIoT*hjv|(ZlQzkUQiRH3
z_GR`>fg*!8etx}y4zrEPs#^imw8!y@5jDE519P1bjukIDsv_W?8~U~y^;?fSW4*@c
z67fmhkk-^J`NX^?3S_WYo~~hE<p-J?ip_aYM|5k;t4GeJ&tD$TItR=;OY5YT7S`YC
zZudpluNeuZ*nW8_rJK<2OjvJ_sE5=u$yX9MwEu_{IpwSnkTj6<JD7``ewZ||RS`M<
z=SlGxq+3UL+fDa;&_!slbKiY$(yy<~^w=_(-_3#7yXc+88hu|p8>tshjudQ}U!0xl
zg*`uuFh!ASu|Pa)8KDN5f!tKgmVu6gN+LPAfboTQz{}}pVU4g#CspfJCv+$F<fhS<
zoznml@2_*T9XYPS33VN>yv4oFn;u!MGOj_zuD&Adu*Vvfh{M4W_=yy<1e(v$A6Um}
z)qHsOo`+g%tQ?t@w~OIHOhTO+Z+tcHE_u3cZZc#0s(Gun_ePtp8s!@HTHx*H-n@f_
zW;{iDuJ2z`+4r$AS>7Xmo5ej{7w*5nW6FAmf<t$S6KpXBPt9MfQ1`8_Qy=C8hfEtX
z$TRsLMJz!^x16l3ghGmIe}``dgdzxU;?DcVTX-4V_xF!96FN>?R5%OgTZUFde{qd^
z2EgkmpNt>6Eg;xmz08|_^W!|v-N>J0)+Zy<9%X}erqyU(a(1Zf{UrR#%}tT+MNfB`
zy<7vG{Z>(Kb@%g-5g1Zyz(j@`lEonwyByaL(V4h7^Dv6vfr6L%7kmnPx8b@iQr!Hg
z*BSzG937!;93$smv66?F6noFQ)4bggBz~snM#q^>MLh9=U+f{95i0o>*00xT&uKH}
zE;wV>N`SR)eOe~sM8MbKaPP7{|8*k7=8KDNyh#44I29|nkWEZrGM;!OA2}+>tW4!k
zihZxZI6N->OW=KqMFgWsj?&rC>WFh5Tjc?TklDi5XuW#!!P-rXs#DR?kWlc4L*vnW
z*EzO=akgo~VG}+MIuojI(>Kl@FJ&#$Q7!GwBEPSpvz`VQe$Lpk?W48q$+H?d1Qp7d
zl2wD1DHbQwgS9(Tc>K4E_n$}a=7j62SVX>*oX`uZ7h$<bkE$CW5vwYBV#JIpU0uBj
z>1ipxe?c~654hdulk6KBO6By#w2|&z_JT)bNTtzJ4y4b+=Y)D4Y3UJNFU$*oL4$-j
zwitbz(10pa3PUIBZc4~hO1Y{Z;w!v!b(YIqep7H=f%22fl^SJ^*NP_MplFDFKF@-<
z!#w0i?;_)C692i1lv<)){c~u^x=2d7Zmn=bC7Gh!eigaS;)43fSwy4F>p+3rN)~xJ
z7rPf1j!<Z&A@iaL{g_<C_vt;)vdIRg^p04H7CU1yiL78TwqiKi6OCJ=oRp9iCOSs#
zaAYCX;7bxpl$w3<JfyR%(As8&jj2boM@?D%XAgG?_Ceg}lZnqt+J+M~x|1fzfa%B5
z2^sOlsxlcdX&)pws2YX}t@C7~?=iVq>D#Y04%|eW^RmfLSze|$_ojB8{?1}u$*v~`
zEK0J%swfoOidv$YRmQ}7UXwqHxl0&M^DMl@+n-_HKYg`coIiHbN2a^Z)=FYWH>yKW
zZal1;afC$FBt;YnEOAp<djTfp!;xH<`K<QC$})#QyA#I4bc7TYf8~MmmlTOSfeU{5
zpW)fqbcZMF^By+4+o2~s>Ww6yx9A}aP2;*>3NC~$qI)gfY_16lI0cx@B-WiKb6G39
zy^n>xm|QA%OgY0b*At~(-j21qce>eJ%Q$!Qk|MgyzrHY`z~Vt2JRYf(GxjubS}-^Z
zCO&>r{IjSl7N`Dk^~h?g+}9^H*-80Z0>#A{CCP(a^G>DRV(U$&*B7~_)5A{ww?5M!
z_I~H2t9@H2<S;LLXi+AlKhDNwd8ea;TK^natc9lkoTs(Gd8y=>+eqaAndDY*%vXtW
zZrakCd_MTG7X#>dnWfgVM|(`J(~UegQcl+h+P<XG=B!DXM*VMsbZLO)<<aEX(>SyS
zQU{?|m(v+%*x`#-!XRsq37+|Rr%GF>+DAn$rwVbH6D?*J!2X$dKL6gJw59|ZH@DIF
zhXLMP%Q)G`(eC)Xb4MzB3G>kG48yiKJ4gk7=LJVVBlU9!Wyl-0?c(Va3au5)w!jO$
zPN74yw<-`%;|pfaFa$w4!c2Y*#*}-CFwJ7el+;43fBOc|&p(|e>1iH6Y!Q;luy%j-
zycJXK%}RT+%gjhvb<qe!P-DG`wEq7Vw8Vn`>!(Qn1e`=oY<OaPA5QjvUHO0B;?4oS
za;DrJlc0aGsQ=G9fk4(qnWL=V{YSm;y#D{%<^Od(Yg6)4Hg3T=$_xHp>yA`N(pvrP
zABr_V)cN-b5<hg^@Z*Us^50M%Aewvh{uxdoYBagL7{^VLP|i49(*VR09*S_z%ErTI
zmeMS)_y69)myH8fJNLe9`V7iZ0RXfTugQ`HmW$rM09~dg)8B%~DR(0Jx_UJv;XFY2
zTTVqvoxQW=TSGkvUyymR)r@gF$G7?jbU$Ykxg}p#RG4w;_XMY~#i1=6oh*F}6rVAF
zaL753zURw;d*uTIVtT~eWsNg~>SD0V=C=<tNk+X)=fr(L<gxzD%acgE+13!;$obEP
zo1LbnnlAt{%ifo=28VLJ_}@Xqx!&|G_oKe#T;9HYFt?aPk><5sM1OF1^T!E~(p7lW
zSH(?0m{C!Mz6*>L+BycYn`oLEQa;vpTLaY@*I)-tTY%JxvP0)yG(fb;G6`t>5wH8z
zI_F?ihZmT+hwxaNY>}?aml7ay7q_9hSoc@XZ*jOxJa`Q^=K&E(5I{~7nej0oHb5#$
zH#1!quF`Y!FF@s=e%m7Hff*e>2!!Ejt`NL<o$9{V@F5(qckC%i&HnK%2p=M6NE}dG
ztLF_;83jw<Z@XDD3BvzAAuHv5%p*RSis*ZNFYWZ5;l9J58ITk#K3Sds;5Isq<>d4Y
z?u%H{SwJ{K;4WlYPax0ElJY_Clz9j)j#vC8BpW~+(!NE}23eT_w!1%Y#FNsf0DOb@
zq4fN#Y#i~uq%a=%gr-XZjwA34A<quiksf%BTP;kP5<LPcsiRMOXXtu%2BZSDa6d_n
zEDoAVn+OB!gmThXzqfleUyguC9(Nf!l-G8EiptkO5JLq0`-)L@QJ;E0H;_tGhH7Y1
z!xn<CB8YP|_!qAtlG2<q2E@#aZEBKed`$)u6j%V@aXP$ZyRXR|niK*u!dr@OeeJq%
zl<B|P7LhPdM87A!dVr(6JqUQzmqDa`P5Ges(NOq4L-s%R)PFOX-r@&QK+ihvKWYh8
zc^rHBcnLInGaN{HNpA|wCoF&(tgP8#seFq#Y8e(qRCTTt)eu?!M>OUX7zIpE2p%8T
zJz|RUnI@5GXaAqt{><1qDxcG>y{Bd)NyfweJS{04QjoIaedeuQYE|K;__)G9fZ<j2
z=?Ks9YhU<nu;)GQDe@VKb}pvjw(xsfA6Es|DqR|3v7Uxrcw)1JUBYPJW}*PLB61ZK
zh@)D_vEQajH%oqg7oqf7>;+xDc_)L{A+zn+;VIg}xSR@c`m8Cl)MV1He9%2$o+5u>
z92xQyx(qfkDwonUxptDVCeSHP5~k+#mBYy?^TycNdfCb^@2iS<Kk_o45sT0I<hVrZ
zNenZW+AHktq=DvyLdZ30>?<=vCN(@iuF9~Eszz*^?r9;xgla5=gE#b<pQm{RC#oLf
zA5*4DV*1)L-qyS9nCI%2Vcd%7cQ9)@JgR@nWpflc@xFwPx7X4x>M&J2oeNj8zGRC^
zGm8FR(2y+%6g~@df9t$bvww9fP+$5Nv@-*G-&g6#|0kY0`Y%YW%;Ej@OAx~`cRIoQ
z{Eoc8pi!>0`9J+X?%(vXwK8n-7ykvF^sLqX1y#A?{13h<K7|8A%+t<*qCb`OFmcv`
zUp5+6RmRIi)S6(`(j-yle1!(;bnQLvaJW@Ib=+I8B3_zj7|T`8Z|%K725dhZ1a`hI
zYmAnwB<838xyTpqj?L*|-3~c&-bRH~fx$&kWpwMy@y|SeVwK*+P(Az4oBIFU&3)Rx
zAY-$$S5I)gk9H<DFFyCW(D?5j0-zOfhr&!>yWdRfiMxJ#%@-dDP6ef9at(+@KK|or
z*T~H9;9FhaAi04x{F#>gF;Nvv_&EBN^N(8E;|A-hvS)U!rngdW{C9i)WdBQ=X+_$*
zT2Wd&iuSyEA3q?l&BS>)J6LjqpRUs^&6@P+gCC4XfiL-n{1hJIKHPPa$5kQm-5$RU
z1i_5WHik%ZMfiCClnnTy{v|@+&d!3AOwTwlPEVLgSh>eNYD429z5Lm%H6Jmt<Y%7e
zt|7H?zK>a3y<5r5IYLH7ZCW5XYHF=YZn>juOFbpJB3szdNwMNvXg><+xd(ZL+x*3K
zjr|M$y=QqkTL;<oW_VR2LxmFTykoZS#&l8pHDzzjtP_)+&DVxA_NOV@Dd=!N$bT@%
z8$xP5ar#vVUN*trfBK2IS(&Hv4If0rnC2Q1%M$D=Raj(;ntCvlXp=+FGIO(|uWh@w
zEM-RTY{4q+Mj`1{zU8mD#angOXkR)|a4n_!rpRty^~6p=1W#JQU?(o&b6AY^Vu6fu
zSJykluW=T6g&I;8Ro0TPKlNjlbb|G!`T)OW2fWX(>GYMG&8gdj*{dFLMFkW4OySzf
z248hzy+89rGx5TE5~W~!zt2Z#-EMXfRMJ%Dr?a%!oiAPV<Tmf|`5nj(OT$WkIvTv>
zF->>W56xInbq=td@r{^-hP4u;87G~J4ltPujxc7fKA)BSEGN71J*#>?+?8OQvh}Hg
z5Iz)KpcbEo)SbqQD*lz=I7)@D3+5_|F{0`ZpO#Z?HcTsps3Jr?aB?DEMXmAMyl8(U
zETIvDY>zJ-p`qsHpzKtT52d!Zj?iu3S^&lpQsFpi2MLq1Zqk0owC)w!S$+LbHye$O
z1~<~XM-!fNx~31_mRb}a?HF1ssMbu@P_F=gskF$+J$y!4^D&OBG@}9T=v~cYkJ|!H
z%$>~a)}P=?Q5Wh>ARkb&Okkp(9E@N45|^Rj$0t5~^HRm1=S3CUR6Eu3meg%JgC4SZ
zJ*F>gSpUSxWgr(F8Nr!F=2zT(Gv0ALQKiu&l3uNP(^sFdx;4bq=k0pUbPfzUJ#qTE
zjvuCoZ3I54(@z&Kd88Rj?XFZ%l*?^QzM+m7gsY9@#noI}DkB61UyY7bV~T3bs!lyw
zMDlkH6Td8B=dg0;p-AO|U5onSkdXcD`M{e>Ik_7ra&_*(s=bWOP<p<GQT9VeY8cnd
z`J0cG_+12w4SNtXFMCw!IcisDOvKcRWM$wFsXisf&ZXh@XQDA93c8WP0}!MoG-`RN
zd!hF^De+o8*(hJvS8tr+f2y$fwltH-Z_Ai$#dJrJ?k}kQEPe!%6~Vs?(^QNT5Nfgw
zwO&O%Eh1|T>clPxuu!~7N=Quc<;rgNc<^LN%IM0xw#r{<*JSGDiB$SYz^f=%xrjNj
zEJU$+8sx;=J22=nxKe}pT4Tz>UQ$GS;PXU>E`NB2qS(e?&``gg@swbY=br)p`Bfu^
z$^w<puL->!4PkJNKFpBohAp^ui!8~WpIFWDN_fP%*B8k~Zc4P+>9(=y5j(167=3y|
z6TVh`m`ljMajmS)WbrTzk5@_~-VEX80#5$I&nHYxR6V*+iYXfCZH?q0Hz|t<Cn0In
zj+9#*WuyO)lgY9m>GR=tLbq@8n}nWB5~waghOLZ08N~74ESd;r%C*?XPg<oYtCUu}
z5!K@4xwTAq$;4|S;AgJ4I?aE~?2+bzN)%VED0<~6TViygFLI2w8d`i^JXUrufA9^p
ziVj%|sYgQZKu%!cndncl?)nIq0+pToha0|9x!<MeO`T~hemEIQR`2{8sHq?}tXO5W
z%4NaRtv8oXQ7Wxga)O+>RDb71skHvoCyEx>G>CB5J2$BNhCq{7#+dOu`UcA_ED;Nd
zS3;HQqRzx?$Mop_RG(?DXpH^zzTjK~>DLmnlEoK6ozrHSgwSi{5S_byy^-6+!+RyZ
zts@U%c`X#O97V-b_S96?Nka%tU61R6v?TWMFq%k*I6=D?6$RF;wwAJm-M<vKTN6G#
zi#mrVIl%QVoWC~LZs|8{t(#39QPU6K6yPLKn)i6LHT>Mk>+4MZdO4X=JkW2mq>lEp
zN-?VM_oJU1G$5fzF)>+q*_8GMu2k813hqo7&yqV7-Q&{UJ}086GECShD%M$;9xk6E
zONkOyRpmlX?qQ3Q?>Lh3uVFV|Y~B0lkyvgiw6B_HXY}T&pymjS#A8Q?i`q0if{qz4
z#z69-O4}}$N#2r}cLV<5g^hLM*6JrdF%l9^OixzIu-wgFv`|Hu&>KMsdizMaD3m!y
zt=&WKP4FI4dR*48EM&a<>{}kQA=cl{v+?@DXz%VU>`K836VUOi`{9?+W0ZF&Zfj<q
zoLgDwxdz3LTycA!T0<2bX(P5gBMnv#*5}iv=|w+xg6WgFKAud~{8pxUTq+Px()h5e
z%+b*CA?Jxfv3;96bZbLvE@T`Hg=!P)*n98SG-c|-YI80i56Zt7q?SRS>11?r(Z&x%
z%bv4u3zPR_Qy>s{QV%+5wAS0a;+%kIhp|m%PMn{T->0T6-E#{8TdLV-f-Duy!8wgj
zw6hJqh;@dAs^@8V3Tkh0I2J02T)|KHr^5-O4K)3fmwrDj48JAHNc-6!kv+z-7yqkT
zPe8@gIhjw8@K_mYRcjSrVf!{;&i`qYzMA4zVO8QG-A)MoAeO3xL+?|V!6UL|!+uUr
z+Wx@8@)VC61sm<1pe--^@^f7k`JJ)YsZ8`3?YRMJTySGrvhsc4?@L_t>M<>&-oe1m
z_c8+Qz(p>@MH{JZ#ZXVpkT%~)-dbh1j$92-VJSW#jIp~yQk-|ltXuwH5CHzllWl8o
zbe8^EmV_R)8+u0+rxNjLmQ|1%foDcxY5}$`zW4@aQRLrPt2NTyF{&sNtLD#BS$!lM
z^`oxh7=pYkuBqd<&{=;vS#Mz1)%Ba$dX14Qr7Nt%GeQ#^60OjT^x}VUTOH2q8HHmp
zuFYfb$+6Tvuhwpq+o)Kq(NuO1HKuCYUX~%0Pd06^95uyplwq+1KZoSkCBDMNJsHj%
z-F2zjOLg$Z1Q($4)datA#D1Hj{v^T7p<E=bYg#~FNxoV44gEaUcoS)D@YK;!2nz|O
zmT>W!6-N|BJ|1~N_+CCv%T1aK{kr`kOqfklrR|2J4!d&R!z~o4X4)Q(D~|DH&M4EX
zG2_EA3`G^!+vGZWwT62nHE7tlo2mH)_$uWrrDLgRHiZiwlO}XUa6HbgE5JGzN3vM=
zb;0RtcUeRE?~;Fc`9a9;e)Y#l{g@Y{#w&_7s4|x(g@_RJO(QXPQrE60-I-hVV%Kgo
zN5tVy>m*H0^%&fQP?sjNIPAA;N|e2skP)iP!2y}n7#IUl;h_&-KBX?NCqu!xw-Q5D
zGj)-EnS>#MuOJy~WqnBX4Sx&&T7n5Sov5ak&sB8qLy0ZqLdgv~xR&HXkq?Wz9QC8z
zYZxnDshYRSkA20e3Kx$TDw;D^RR$ALG%00fh{*0T)mz8~=4q4fhwRLp8}7l@R-v%9
zqKfQlK0byN^J%A6K4fkV1~a*D_Bzn>J6X*gY8N`p<-<zYF8@{FOEW6>R#i#@6oJ?f
zXC~bA_e3N`wP9#4n;GPu-Ht9pDqDV1(yo9YALJ=ZGZj7$CsRSPxpvQ(qR0laayme*
zXLb9hde>JP>$UaZ6hmtxf%?owZgi#>5#Q|T)s485zzJ%o(7PI~F#828_m22oO=1lj
zxr%Smgf*!N_tYm87#xx$KPET15R%^b=g(c^V-pmd`A1(n8dLsJY^E1(QzM<@{C0q6
zx|R^Bn#u^X9GY&~B;PXerF$_8X!~SS&1ns3OSKK%{pq~Q&oUE^66vgACL+1MOhs9P
zq(<cfy(>u<J_xP@Q3GQ&x2bYM`TC{WucS4@TefxY>MJ6fNA{GTg(I}ee>c0xvKdHn
zBkwa<j<d2P)-UTfe;TeQPHSlwpI(XE5fD)}{0jmH<Z;h09^d<J<h;UtL)RasPSMg&
z5lEMs-C~1r)rcFlTX<i&>E-jMudkjkm(xrNopg<77Uxd_)OVH`z4cUHnCnwV_}~NL
z7Q{t7%dXu2b-r%J9dgcdf?9-l1NU{{LC#1qnXWIjI%lTC_!LvFc%Mv$-MeIQNRb+r
z*4_B3n_Ti2s<a$o0<Y?KGMUtQ-Ib*%ketz^(ZWYyK6vl#F{9nQXn(jKoqcZMUr;9=
zHNyLPs$fKqqbpoqJo|IQt+RzGY9M)3v}h{KQQpR>IWrOCh%tFSKjZc2*KP5POW&lA
zz{Fv>f3*IXVP=EJkv2BbtwmtSH^H^9TDwO)%!1Zr{>w5=0|VU~QtspCiKO?eJUhk6
zdTqbbti6C4wN$I?pUFRWoPHv``)e#V+X#Brj%kX?wwir6-Jg{Cx{E@ot$7lL-%*Ul
zuyt}~Dtsc>j%R{vCoh$$;)hd{dE+guzfy&%)vjjguKQQcK%vI7SBYMw=U$v{{4EQ+
z+7&Mur(Fe$s&whbpmi_eV&&SB%_6k|ocTEesb++{&kC?JLev)W#e;czL_Er8sWPjK
zDH{eGBT8rtyHq+Rce*rn%@%z{dvi0gC-t!r{%ug=sUHQL@A+jHx~W7v_t&Q7bnL)V
zac5{K#P+FGT1Y6xDfD8ZpDLmDd%M(xzO@yMj?&)NWAA8je8K4<hVNZwO<ZJ}XkMzS
zbTdA*`@P`{Ct9LJ3RMli0=R`dzg?~_BS*4sL2j&~r?q3g;0S9f{}M@GZI^lJz;)H(
zR0izuzOE#7Q!4hTr7Cx-4SgjP-y!*1cg=E~Q4?96FxHPwhE8kQ^0Zw{inp^&2dvGx
z_I5BBeU@($!c2Br*fuV1K#Xj@2F&s7N)}p*$k{WMT_5%)cXpZ&XeFHr3`HXZ&xX`W
zJ>>VzlF2WeW@9=6#~sH_e84@&`-_c5G~|b-&6GskV#O@!*ifh(i9NI{-+}qCi#St;
z5$$G$4SneMaauXr(@$z+N=MmOJ}C}eYxY>nYH107MflHyqe?RrSqT)QmDlcanmXHO
zQ6&V+${3B3O8t?-WMh-Is?u*N$tlljc&k`x;c_C)&Dvnf!X`$JQ8X~|FLhtJ6#RXj
zG|fl!?&Y32_iyj($n$<8x=uf1@08^Up{I!|%Y<07VATQ$#R?la?L6CD%Cc(Ob$$)c
z?FdV9qMgJPB{kN;9qUK~jsE1C$!X7$$SF&Nwu|<7s8_<)o)XC&5l(5Q4J{d0a#Upo
zpSMH--=W{sfx+<{I_=dP-gzVt>&1R-;kF*)Yu6}&qu!ejxaKAgLkUfstnHh1&tW-Z
z5#JV0KYX}$=4XC<Fg{l}b>>gNT{^}=hOg}5F{QFC6f)QyPWVw!&c;^`sT>hGKE>*P
zmnHhmZX~mX${PQf)UVTwH&XmM$SMle9tA{(Uv9s;Oj5*!mKi<Q-bIvyZr3?iqLa08
zG;|?CSn-$iSk06*5%})s&mH99Qur$BB(&8JVi#(LpLX$7ndgcLH;UvkLp^PZCeO~m
zE}P!$AIF}0uC}@_)Z8Jr)aC!4=$O$OcZ(CUo56%k5Sh^tYUf5cuGf9xCz2Ui-g06f
zXrh@Uf(5gOS8y?7ktZ1jTCK5Gu8zCP->#nv9-6L1?(K)#U@V~qHV(|W{NyDKD?UXv
zLN(5|_(NvQ%o18T%HxabOUibXQVb+AB+T(O4#Xyt@5!@j9ba;ajZ1zvve?n=7cO;~
z!iYS5ODNsoQLG(rHFq|~JKby;<H$(#Vpb+QDXq@dqp__X(v#3FkYvJ{E|Xx#H#kDp
zQbTmVSOe;I?^$PbqSB#pA_{I8=p0W%UsYo|-rMG@@@wo5g~F$YG1FJt<En_!4{QOM
z{7o6viPX`bn@?lRTLXBD)RU6YGO!xq9CTqyxFC(;=G2==66WMn``prq+P8Ce%}q_Y
zy??uZ@=SQd*%PV-XfAIGd1aOaMe?ibh1YJ(Co+^JtMmISwMF8To%6#HMrh$J3>&he
zRCkM1*f7IWHkcY_!)`~-^%gTUJoslMmvi2CJqs)GSUykZ#i=+QTPzomip-t43)n%d
z_P1GLZ8IwvD%XURC)9y8o_VW_mn{JC{HE7x{IK6&u-5qTKD05>i+v{xK<j`NtQ045
z5czd}Ug2qgj5j!f5Z<2&Pa0)=(!`=jm>D(M%(8i~%PFC_0&BrzAMCL)vF`g&KkGj2
zLeT8*ey{xco+nXFwd|c2$AXiWEiN>JJi1d3A_JzC#~Sr$p)}(SvZ=gw-8wE3<bH<<
zZ^$ZilBd0Bc*X{*tK%Y}o#Zz4d|<3sGHxr`WV<ogv%kA;q<%n7krO4dft?BUn<VuT
z+?WjS6(QWoItUlB*Ag(d=X<b2Yb%d-p2I{6l<Os{A?53N6?eHJxi~*hKK6i%Tdnze
zZ+R@IA5QtFZwhNC3v~zjQ6_ugEy&f3Y)+VTh8<L`Sv1k*`-vfbbPGPSxb}>+%Mi}I
zJIN7Y3U;}+;dywm<)1KO8ZFjgtQy6tEc2ayGff_&J1}`^w!TITB_}7>)^-m6{uiXf
zacO^`cht;<=X@P5E!^8ENcF1&pY|$qHw+%$&`i%7?4`RPdK@v?S0PL<eM(Y%=rZ9T
zFQB&0{QS2K3p**f*muv#3ISuxxzlQu^8pG~Ydq#T?-ONqclrsxJI--8qCW$E9Tw30
z^}@7&-6&!-`unX&s^4N3n_TOqaE}h}^QpobIGCM!i@Q_Sn)U20@)vkHzw1O^qn4?B
zOwubd)wj^EYX6ec%EeBMrmP&j=*rvjs+Q@gN$M<L1;u$U@5A~4Zugv@R3IE5cBF66
zn*FR4e$fAuNduXirdOSSG)9w*eo!ms9s!M?E`h3V|AodN1$cW!&+z;*B(6bZdt6oH
zKzDNt<!sYGE2?ajZXl!m8IXx+&_5Ud`w6!c6ZB$zHEuny9wb=S)Pw<XnPisx?H|wf
z;Z#Ea-R3_#18v{0pH65-?P30$0e2t<X9{+D4HAlypI^u({O>mioc9g-_Z0+yV<Ir<
z;S>x6@pV!44dN8PCiW?N{h!_c!B>IYwvT|DVQWm(vGsbQaOipMe>>BWI8b|VK3{GU
z6uELVu}n_xxpm3u51Q93|J~lgdbsUk{6V67_tE+8;zTjCsj*nrm(Jp=KSASam!^Q0
zVL-gOxp(<%3t;)WYbF=7`<eJfTwQEVV_A0{G1Cn<BF}*a?PmI^aD=Ck=lTW?cVm|(
zHa*tg5opo&k5IQ|8{ajqx;?(@!d`9Ecp30UqJ7FVRyg@muN=UZkDPpM*t0+0Ke`(_
z1=}v4!724vpUEaT%V>U0X(D!72b$0bL}=knyt@O`^8V{YvXKtK(bo3-&&7Xje1H7s
z5{{t|w(fAbRby`m^ap5a<t={!5C+o@Mor5Cn*sWxNrq(6Z%}?$_5KBb_d}3^z=}n<
zJn;R|s4qwu(CIidX!^YhMB+;Yzm)`0s`b@Kc~fKv?Jo8KJU*L1XGw3n*@>OE4_WQ4
zJA>(|VMC3tNo)dPYnj69z0oBgrnxoWSeTU1#6;`sJE`vywg5V%nG3!2Y!FV;CIh|y
zV*up&H5*@5eHQ?ItSN5t0yR;ods(0TrPuqXS3#uBMYu-{M6}*PoO8VQ{KYm0yWZlH
zHyA&tyVMNAg~Tlu<(a;{KmE=(0Ml?Ru-IEwRfj`^0zig4i#OT$b;Pmg-RRUapYUM-
zs+~rR5JIHKo^JtU4dtwTl5HMU)7?Oz-y|zIx3+1@9i;58r=67{ws{=%Tsx}(>B?Ig
zs0E4`ZPgGz+}}G5V%A_k@Q&TGo}E02!QJ!`4=*2E+XaEGtjqR^ntirsi+a~UZ|n_Y
zlNt^M_pV;W;BYJ`<^Y9{c0Pe`@0g4Zg`23F(ls}~e}Nm%EJbKyTa<F!a0%jKV!0f=
ze1L;5z>Muhe>hLpMLfs#Ib4%7Ja{$f31Ua<@ZL%R60|aJfQ|1NP)f%j-A;*sY;R4B
zfU^585Yp5Px_d>vEdcTJ`^eVG6;Qa5x=&^IWAMW{+&}w);7~1mLUw&}EE_ilctRMv
zl;GSiJG3{3^E?W6!3F7c`d!nTTm^AXL}mK_LpZk&$PWyB{m7>5p5b;m1G^g~a1<b@
zQ3F)ZCKzTF44Rr`uC}g&l(($QdAup#n@N}YewTiRAOE8c!Rs&f&baYp7o^>9{HWZ_
zOnL!F&UbsFImaJB%(M6+q~NDPT&ynkh5&tZZv|9fH03s`?&InA1UGKYR0~f<w;|aa
z5O#q8DF8qNaf0-<sTdfZ(-O(c4>sjrGQ{A|)-Hr$mXlh0r(6DexNPX(Wg*yq%0fWw
zUo+~?wvQfWo~{0xfAKoo_Y?N9zAtWmk!iE!6wqVQo9Q%hSi6Tq`D^I<jo!&D&ixtR
z^GJ4~+Gqc}8eQt;lNWnG+SA3GgS`u^3`Jw#&T1JK8jq`=JA-+9H5Q7*t(@PK8jWqZ
zZwiY7gJqtDm16q5imoy%E8A=TBCbpD56M(>OK;|+j8|X2{MM$}Y~;hYj#WkKq99?y
z7U!i(b<QlqW2VJpY@?en2a;9QU_~1P9S*|gC9;rjb`kQQf6)C$aQ)9~`2Us--ltnP
z!=o|rD;!TA%pT}uOG-8LXobnl*8ry8_(nH9KYjv?jaHY|PT%H=8#m%K1s*Z5zh;l6
za-7$*dxrEu;T|o(gBmBW+^Vy?bNryV-qvcdvkeFnKAgZ(oB|O@B66jZPt45^`OY_}
zU?!AF*8g;>lT6?LDK#TyrnD^A?E|>DR@sNjrqHto*R(4zf&H_K&ofyZ$Cs&GPe6E&
zuI23p-{?34PTlPBboSpL-Yphl)n-vg_0P`=#24eeWFn=TrNW%VE=wWr+>Iia<C11P
zB=d!eo+Q@AnQrFSWqtadW|TATrdMSeIw@PA0bS#$;5f9DiT$`4GL*yp@IoVGLrqc2
zpk&2jlKef``0_6(H^cJvudGpD12tK_ggjNv{F$bw$A{p%O-eiAno%)+`bn0wB@P*B
zJt4bB!Y;9h^vW8AdRNRYT8k4qwaW(2gz>U!#EJ_mAXqE&`(zU7trY2^Hc3L2wR3Fj
z!0PNdu`wHM2??hXPn1*uD_FPs52RAQok1}_09}Npd|pwC$*n1O8rJS_1Q>H+$;|_Q
z<cqto8}9BWoAo}MRAC<5+5J5V!Th*f0doQlvXKu5En@*cCTvn$$6o)J&97rM|42(T
zawu+`-GcsW{E}Szx9&~khG8t%>S)ZaR}G6$V8RPOSd*HVqJ1u>k$8kDt@HQ$P^pCE
zzo6IwZm7Jvv8tihmf9Ez_))Bv*+t+b>9>M+8p_9bLe-n}9kd!T1;jVZe?d<S;MC}(
z78C=o|LJn^d=1&l_n-Uj;9ZkG8fKj45~cx%goU=k&-T;?H_z<T&ok~v=CKaH&mWsG
z=}pVaUu-7#oEJ7DEjECrxnoS6*rb$|VYDZdbV_0J&AQqMb3`k4u)I9<vG1lW_+@n#
zfN48vP53|<o><oX+lbWJz$#sYO;?I5m(_#Mc2~5XxR->*uuXm~mSqhp;xwN$%1Fb0
z{4$w<Yw|DXi-*HL)r*0S5+eZ@5p=a**ls=;&E4g>ijX29R|=G|cVWB$TZ)dB$ia#7
zYqbhE*yJZ#>`EDyw>r}K@2cpMm17qm*<g#VzV4BA<{rh*lMh}Pm!vzgCOPNpyI7`e
z^y_j#p%@7C+^h;%h)uX9G{1lnC@RFH#RRC3-Be?5svj1+CHYF)Fc-|7MD4yKHs><^
z?V`xOsmNKQEgl|1S?yG6q<CN&`pK<mqOv_vIkenN^y`XV-t=krpKTd5gsO<=Y|<#$
zz&BDYTbtUARY2AriH-caD3naOBhFG{X48*nG1jJCkbe@M-)$XJq@8xwo*5J8Uu~zP
zDljglU7l=(?OXYkd*%pJ{Waks57DG5lM61&<X=?sP>|2~aU~s5gCD7G-awnNa=g?U
zaq%SMEVtd!%<B^$Ate+jW0n|hh=iKuIL)=HHq6w)ex{>g(VV4nKhvYs1rCh}OV>E?
zIGbH&JZCks-z`(5?5cz_#oV+L3t+=O3;L*M_sd`Y>D14Pry!CZHqn3)WH>sh)KDGL
z{ti5pAOgcwY;>vXbxb8z3vZVQBnn#^+Z1w}4M)?bFrN&P!h=pimEt2sXK3{F6Q8@L
zRZ!4Y>mQ9iDWt_3=ZJ7O<}`PO6>ogox5-6WM|BK#K=&P~iY$HIvVP=D#UpvfIJtGp
z?5)ICiL#5^*>YYZWZaI>s8YORn1H3?$HTsjksayGT5!%7YA;kMQK@}Yl=Zhty|Df#
zkxONdq@q4)-9Q?P5>e91<5G(rzY-f^rnA9_aT)n)TmF8C1hc|OVYI_jT{H65wiBd^
zWdT^Sm7P8NO?rG_F1NwEqcV5Pk~a2jS$)UP>vFo=Zxk!~VT-o5{=Va|Gn6I92Z`<O
zGZzk-4GexazTV;XwW#NRk@w#3aE0yL?~q8*5~74iBx;Np2~mSIy|*ETArU1;i_WMi
z7%jTdi8hQDb&Q@M%IIzM-g_IJ?DahFe!u(uVgChtuOAr9v5tE!>z;M5>pIW#lV)4?
z9#)>)TsWBA<Eez5h<yy?Mf&M*R6OF^h-<Y!m}b?Gt?zc{eG|z^_}fg1FNG?f*Opr}
zAGy6s@`kxe$sw?7TxW|}>C#;Y1)Ce;0oYKTd>H%D=Hu1*yH*zWIC<`7HH9mucglW@
zZ;v(5&3>h*=Exnzg|>ZpKd<1IsoY1lprYUg)3MqZ>2hInaI#GtY9x1VSm7snZ6x!1
zyOIMj=$>p57mRf4)*%wX^Zg>j2(@YXVM%E7lItHRw}K?AmatiE@?lZZBRI6bwQQeh
zZFsIKlCd@}@3lFxe|^p7!D+d6L$O}nVy*eXkEAy^%UhN^yuT%(rzvF>MKzH+Eh2?`
z)Wt!rEF_ZaNgY0H0X;r@IF&|ayx=%pv{~!b<s-UP0vSq*I=Aw99y>1X8SGLe(q1T(
zT|%$+#yi}3fW`X{=UeO4F0`%%-;;rXHk$<Xz2Wct)yAzlzbrCWb*_s(;TbG}+;SXv
zU(fmO1pNw))mdVpVbp<lo=?Hoh9|XIB{_2PjTTnLDpi5ce*wC=5$)Z~wpk-<CK0*b
zpZWa^;;ZUOUzZJ@3>Pm(4j8NruXDme3UV#(?%T$3!)&IO!$;n9l|YhR;=n?J>$cx^
zufrrPvi~@yLz&WSqbFTiXJaZ;mE#SsuIYYVYby5)XVD5$^r3D3yymzoJ@q`X-|~f`
z)(4+J5u;x<+z5WX9>w5R4r|U>d`fm;md&+OGu8~#G0HYok&^fF=N+(Z8$<p(zlcWx
zbuN)itbeS>u!WACQLW<whChPb4F~U)j^?!#j=M>-CVe&}#q*0jg677eYzFyU9^3If
z)``sd6VVyK^hv1@qnW96ch@S&y}wMDKKkk8H48n9C*{r8v(qR|uPVF=r`a4KHlbQe
zR#TFPRKp=(9$X!)y_xqZEKE@38bfET1*D9l@ketICFlK%9O)m{BX}RQlXN<sIl@<R
z(|A83vW;lMZPKto!&zD3tZ}S!_Nb4WxA$2<y!Fon1pH;g#GjY<0A>pP^n?D{=78VT
z2aDNQSsdffN7Xjj`DHFoQ?!m8IoW*2b?4$V_%Sc<r}tfN{NvL4Oc(*~Rp>2R{;fa{
zht(MWRdaW_#8i7PJg-3qcIj;U7%;owFOaJmX~$A5QH|98n`C2W{-fzF_Y}y?)SD4b
zOs5VUIg^#m#0_et7I%>~XAhxrtM;AvB5f`P(RPitaIVxJXu~&MV%)p)zJViwf4#W7
zs_uSF4dsZ9QB57-M_*o>-QqEHlM%AG9}ImRB3kCujM1MPtql9M-j@jD3$NNlq+?@s
zGa6vgYDzi}Gq{+|6j6if#l-2S9$mOlu1Yc+iNGo|jBO`Gel;$*uk%4gkni;`sadNm
zC+_>q<{L>>7w8n68SR<75uMJVv?CP@_m0Y40n}NR5e-#S^$QD_uVIsVm;9Z;d)r3;
zKq~r64G=1Qd+PHaZvItKA1zW9`oq7f&QnAvam>i!1Nv0!Zy9eTzo0+>-VHOg+A<C8
z`)WHmc3`=DR1ND^(wC)?^l{x677K;-b9(b?{g02NziH=H&E0z&qJ7Vw@rX$gDBUS#
zMP>GeTM^kGya#Tw_gu#|#tE))=8y{2{tw)12lpO`=#BjPe7Tbi*H0npA93UO^fr{a
zAB{R%*>|J+<fA!CQPcy8Vkr@xo6*WmNiHfLHUx-)?H**^efy%;=6(3@*hVj_&QzQI
zq)Pt#g|L*$<i=tCayBlw9@Vd*cV8U!(T3I7a>uwR?{y^qvz0;dJ1ji&ovv5Wcg5mY
zhFzd^5{3{TU$%3UrHNf?Poi-jO<H_$z%Q$=hTwQ#jfCKM327SoB{BE0UtG#QsE#Oq
zDI24M80Bm{OT{C>&LgQ|%f8zLed$^Oud-z`!wo=-`KwpsaXco$w&XaoOu^ikcVJEX
z^ZMf>k>A4f2z%vNh?;t9Ikrps86ovWkT&ZPQ;-Jc-6r3n<Q$Q&wLeP$p=1lC6E<`q
zJoU*d@8E_E;XYM%(ei6P9=-sl$z}ciz!hMslNSl|rxaG?;83w?H81BXhumB+3cOd=
zz1k%>{oAK4ABTvgU$@ymwmpkkE!NB6z;4O4i*CriO@Q86?5yvXp|d{!TKp0aenmMz
zw?@$pdhOi=vn`c)mT?k-82jA-*~02AqxNR{&%`^&Y{w);hrf>)-$-$5r^bEac78S;
zOW^?J{=+k_-K%0?e|w9!Bo>jv7v`YHmK;J$J^ZAJ&gy_D0VB%Ae_bgGq6l74`xHV6
zsk!^3h6JT!>TtO;wwI>wbF1m@*A&P}{5z97jd{9{niTkxAcX;8$co%NY`DKAzN^f!
zI=g#VX^aq^FDy{`mtQ)_8L{a46VX3E@HAsG0xBh&2XFUlmK|%yC2+>xBqFR-wUeGE
z1ob^hwY7iTUPcOC@;gAYvkv7#ng*iiO9%Ef!r-ZaBU-aMHI~gsi^)!rk&m)&tuy`_
z>S@<%OUT=#v`8*@-r!a#^6@5rVvxt@Y!0E99^zF$RJG2!7zPAkB@J(l)}Lu<!HU|c
z6&IL}{u*LzC`?JM`D(7*7eOP|E|VeUT#{o`>4qXOSp&Vor$@SFs~-n$)~A4FA_e)n
z_B6)OrUylx&#sRurz!mrVH)PsWmD0Ja8ajX{I$%xxWpm!^J$Qou4yMY^L|N}HI&h{
z*yy>HSP-u?7CS_(z;SAW{L7w}92*&A?p!CsRUoD$bOeVwKOc6ocXq^rd3TZ+=DMx+
z=KXQ8f|Mnx$3G27)|CO+LWuM<&C=%BkVyLju6=M`&Hc3x0<@2#_J!5r+OOvooX{qH
z=(_*WOG7!K_ydw{(-@ZUJc0+V=49GMkXQ6BgL20&#G4G9Ca$O<Cz;v;KRaA|`8(b*
zn>;WGycg4+aN;v&d|wXRZWRw(r5Fw^^kPei?&qk;1;E86Z&Q-PY})DzRLWR$^$*$J
z*Ntjq>+%n-w^G_C`8=2G`jEpG$%x)W>YmUcqAXQZZ23N?<HTc@@?6R|g*HU}%ki63
zubXv~IcbB8S`L&FM(A~pJ@05?coN9fcIl$2DD7Ms#7k-XZyGJ(`=V$PQH9<l4cS#g
z`O{nZnh5tCM&qD>wf?>6JKC|q6x!O<ie;)FGyWh5DsrAY8FHvF>&Raztc2*KghTJ(
zyCEU*;V}vQ4)Qbk@CX6IR=(|s7==UY!l3#1Q_hM4^R6gsq|qmB8X7t63FomlG2JIo
zsv)fxg)4VWXjhH7>REF-`tQF`jLC)OmBPZl2)-05hC03uK(#$H<*=4!9bSi>=6R;!
zHY!jKkVk#_oO4Qf3EM7ts_g?9i@wQn5Anz2zVFVexkSMk&!BpvmFp>cr!fiE(<739
zfu%zYU2cs>O1Q4B>bXpV;GH(%!>P#t;gY<G`XY@;X0%{{PY!i0H{+8^iKTI;;8=UZ
z=Ax^<Wcm!`TB{RA@`(vGQKM4-K&B6tXSPZQ4Dap9sBt<ONmc39-KnXaD?TgsmWR<L
zB*Z1AJz)F2jz)gFGHxekv=dS(^++-?c;=TH(vd@t=lA6NP`mL@!*jm*-Roa=LW<6m
z|BRcI?Cm!!5Acbq6&ihk;jytn0&>{{D?7V-eXcfRp@Crn@_{T_2gS8(LFJ;c0tl|G
z&hkv8Thi6#XuG%`w)FS0>pmh0)B-Jrdm@xry$AO`sLd+YL+%wV>1g1Tzx9uA>+%i8
zD?9}<^yM+&7f)K==35D|t&Ew_b5BO)jF0={%_q-j0yt_JrBY4nB+-f9IU{HG4R*$(
z{ncwx8{*Kr8n1IZO0h(<@wv6C0ah5L%5&?0C1lbAKg7gtE`ytSHr0;}eSMtUu4-9N
zoS~bo3L~UO#;}eCx0oBZ7M}2+-iNms_bl`nPWL3QbadKcb8>jn2Ct>{BuV_UQGKn`
zzIaZQOd-VuRXcLLrZ6YOA{w%{$)Y!I0q{`Y=TUT?UTWB*%oa>-;XvMgJw>Agh=pi3
zS?2^YH~9^`bYqu21&M~Mt=el|1GpiDawHQ*&_^xXy{qi{_r=kA)kof(kCPnhB0UV#
z0?9nx3xjTxbqr%|_f9rJHUN~0r+-eXt%rd5Ob4liCs!^`cj9xB{d&o*Ma0f8K{t!0
zWP(A~xS(sW$S?NJhG374CuA{5$V3cTDRmqE2QLS9A<1KBnh^1LdPC5p7lL6tIQuHr
z^UHrw9ei{3h06FKujOwf=}OiX>4I!@|KRLPzZ<ftY<cCzv`mj6PQL&M+bf2^-N}*a
z+a|rwgUn=wmJ#Fp(`h5QF$#+t44jdk$Bbm@<TL-fE=J~WYH3}}kLe;)E*H3^E^Xw@
zB42at$}@?o6AS`yr7`wtu5}8do13{)SQB7Rn0q$QilerS3ir9t<E`W4zpUs-A1dtj
z6jd?Oqk7(4QCxtQGLq$EZeL?u^^v;tyH;&4cYFb&aR&MjfDK8L;tH+pzFGtDXKd1I
zie?fq&?5CORTOM=w?UE4P7xL2Oh}{hlW6?k<NH6q(`la#y{Lar%P1#T`n505NHsj5
z+J#ou+EGq*(|aZiI&9foCof{Lr53X07o|l?OG^mN*KrhdZKm*CWs%2=0#;4g>gZa9
z;kPl?9NTYRrQgW6XZ%I}g%J|$_l>B|b{hB~8!eq;Zd?DV93f-mc`FSU<H)IFA8fN;
z#Sr|Tkw6fI;ty(_iC<hC2Nu~}ZTD5oJuPYurxygh{jbRzsoF9!bNw&K@_)z8tHUv1
zOm4tvZ__w<rfSFw(p45(9{uU*dAFm2+45d7{2^ByJQ0CB++AnU6TCMLVX>nyTpSu7
zHZyvmqLm^y-Q);f{%xa@VQY9NkKN?aliWNsMTwo_qTP8V7Q0+xUpm?NzuGlm6K;@$
zfxQ`2^^jw0e)#?Wi<9hXT=mZKDoF3gq!gx@eErh6k>VOMHJu72t)kvYfpn6YdQXK|
zC!VpRN6s5>aB_csDTWhxMV~1o!zOa6fkKhJ{wxDK0`YSrOm0-U9;R8+0!Y~ENQ8M)
zcyU}J$!c^0ltsEgo0#;1@O3*b16|9#Adyc<I>dWQRQb#+DtY<L#T@_E;_}9;s~I@_
zXJS0iO^{KT?s;=*yhCBwP9$_5lN>>>6GM-g0+(b8J^Zf!Oh`?pO1m$zvG4aVM0>~x
zRW2BkAASWyJqWYjb2tET1z9#lvPS~}=nlPY1QWdmQ9mj#Xj>JtJ`SL6(hLwaG+fiE
zg7vmiKVL;;&Kc>f2%nULv=PJ~$+sF()0mSAK*Dn{z^23k#3^#i4O<q~)>OBVNv>Ee
zw{Q~y+$LC<V+n-IvqyNT%$*z@Zh-Vs#8(M$mlJV6u44y|!4nVsIRIxJlxuyneIk>(
zEl!RM?JSobct}yop>_Dl2uV8eV|J&r8Uh$htmrWk%jcJZ<RI#L+QikL4$_k4y=1Jb
zIe~*l&zG0i0B9BckWq!O0ChFq?_s%sI07Na3&=u_KLbQ}2cg-b<kLpl`zhXa@H9mL
z5ep)N$&7b`z6le<@86nxN}(vjgWMHfTt5U(7Lak*^X8>~I@zb_V$L<vq4&cZ9Lp2(
zK*t7XnhdDv3>&#DE(t1<7fye-T;|-6_C|uZNCV;Y13`}4L;%w!EAJ&i7YU$S|I?UZ
zZu;V~E3~}{lDX-INbPb@Gj8AoT}3Ev4+5b0D_iqG2aH4va+(9tA&}PT<hlTUcI`je
zi^dvb5#;F<vDSI(OM8gxRK+e0fi4~Yw{7YlXrb_QTdJ<^{|rd?&2V^W8Q!jeU~9eo
z-}ytjp3Bu^^2qTKoxoI$E;A}fO0tKD$#MYamrbOi_b4(sMcC~vvNziRv!#f*Ln*18
z!fAeT_lm1|&m6hTNbNrmT^RY!e|Po&#V`L8BnATI8?&jv#pw<|ZyX*6*ruvjFXgq$
zNPE4$2D56-H%ZquoEA7d`;Vao2oK|etpnG>7e-q<M9|U=hC3J<WWDik!4%wi74QzS
zrd;0$vwxs$G|&e?f^*>O6S`w5KnlVa-may=^5Z~On=a;KJ;GmtgSpDyI5DWlzM!tI
zD(E6<-{h<J4Wrwjt1NVJfo*`^$!`UERHn~+H^hB{d2{r;CNErHL~E=IQ81)eLUR&(
zJ?N&ZQqJ?g<8P9AA`o0$s6p6j?%~iKwNve;@^-M?ox8eRR;9Pk_zZXsHNMa!b|b~K
zxx_dtpxr}t8zFHDnDwzMeWJX18bkT--lU0W<;fofjkaul_#4LBWml6%v#IM?QK*t*
zZEb;d<)?WnM)NC>Hrij0n{f>OBW9>q7rk}|%qnI$1^#Bv^RcLnXOcTND(qsoU7;^Z
z;fww*qbCEA(-ge;dcPvR+i#_;D3JEv)QOpT%h<2@VK{0~woHfL=un>Z&jA~6;Os1x
zI!Z|^y;S&-a+YEw>~E|76I}&4r&QFU<F-zJ=}AlhHG$@f<A=E0uswk@$<4xH2<4&>
z6!ue30?N8z8Ynfuk$$#a-Z7VzU==VNH~O0HA*wCZ*N6A7d*)r;thM7G*i-*EZ}hGv
zC%1m2ByU8=A?vAFBOaAtb4kPcU~BLa?j@RfrdHxRU$m~*OARHyn{B&G7Vi&BI9*F}
zV4<Pd<6_S~w!u;nM!LLT#eyD6jmf{>xrwsQEauiAdTRNhunX5Xl3u1RA0uU_?w>^4
zI_aa-53zE}2DSllabp_mW=1V7cZrN2C1YYu=%FUsxjNbJ+uK#2Jl4EV4m0Lx`QEJf
z<fGA2olERi%|wlzCCyB&*mu4GXDo#TR~9DoZ$g4!q^aB{N2dr2qf3T#w|u^K+wf&v
zj#7nsq(u||UH({jszD>}7r581CudTR!mi9PeCzvI5FzaWRphAqMA{<~uJ7r7SmpP}
z5lTV=QxupgXY241dGP_gP^@@qK7IZ(<pLmSUp9o20n%+!t{}<EC0fF?h2W3XGAf$F
z8J;$<ybl%pQm&K5@<dqGTom?GF;`XX?Z-J;8^tZZL(V=92^ZU<O4)WXSgYEh(?(r;
zz{|xK*LKSoKhFUkNu$BFS)#gmj}an-Lk+xApv|JyH?L|(@0;*Qf%mprB%f^dV_Ycd
zFfLU(<2gRiIS94m&rV;NCY#e7WAn{?fooihl6vs_jctU+6YhAZVhM-A*oj$!p`h-H
z!mDaX@eGpoac(cC>d-Qa*!GVQ&19u_k><-e4@IUHMhemLhfil0LVN7?UyK-j7uWit
zS76P4<B5v(txflwSTUu_yvV$g3FjzT%)-G9m8dNK#3^q(@l5S-xGwTwssdIp$9G(z
zys3vQ)(p)r7J2!M2_ckC<Z@PkU`0<va#3*=1t{00Fxf0kuAe%&gQJp0p<HD^VnL}f
zy%`#6^IX;`rJ~zqsAJJ^(&?}!D&$q|iija>Ela0}3W>jY>S7$GkLp~rm8Z_SXWtS_
zcThFZCGO9{2UbEGd@_?aX%b|W%5uM>9XYSur6Olz<&oh2<J$vjJqfn9h>Vi9V|zpf
zzwwX)H;(=1GXZ{yYfs7t@9GHc&K5lW#aUAFNG@PO5C7w1q*j-1HP|AgwxvbE)VO;r
z+k5Tw^N&Jh-bEhM6^0i}YuT_>_t=K&NlFo=rv~N<zf~*8v}$&!W+HDN#T1v<CvkJJ
zXsDXge~3sgxZ$J|MK28HY0@1%`TmJEL7#iiu^B37n30;0c8KhglQ;j^(gcq-m9uf{
z=;-e5Xv+ucw6)21JHKF)7;~EyI1q)s?GNsLFup%ik)+VHAyA%M2~Bxxoq^F@(M<AZ
zEE9sny`5xLa<ov?vTutyP4-?q=Vi)#5LW=jR;+~%8RImQvAUKVuF=BLCDYFyaBQi$
z<;Sv4hFNRT<p!N<*!JZfRMSWrqvbIK9lpIn`PVTqU)yfK3#!m6kgZ9^;|VTKMOZ98
z?L|v%{*0@{91p$e5AM;GDnT2%d~8jii#>bV?Of8GTpe(`%gC;YRnRITTLGhLm4|Yn
zsbc{vIcTwO_Bs@So1S~gz5c1kIk5V!$RH|q^_`k&*2c9r;S_g^JfXeYlorp$B=5x+
zecV_dz13Fue&QJlF$kVCO}OTY3COuGCZB+;W72*xeT_2}H6-7B;kc6;hWde!MlF;s
zxpH->2oIuAJvw^{Yk=3f-eIl{HJqvQQ3om*mBC|Rp1P&%AymC;red=0-Czt%k+T;l
z|JkPbrus`n<x`{&*9ZS^p`x}QZ<UDl|C~CsMM@~-j}xh{ff+_o%0aeyHZrmFaH~JO
zcfIRc6Fvvo7V^A3<EsGsGa+GlN@>a2gV`KU3=CFTP*DS#>?Lr@=iCaFo9%5OmLe{2
z=a&H(m7HO-%mGBdPzB%382Q%d<gaaeyRZ;XAZv?_$Qv-Z^dk3+xza<ch+#tDy|vEZ
z#1f`LE`Nf^s6Ar!Hx7OCyW!p%t<H}}g)b?=@j8^8+G_W5K923D_EsCGMtdcDnZLju
zTi+beOeeg`!()Tfus><3ER5D6cFs_zXvCJZ{*_lqFgiu)@B<2hd<o|^?NpV;h;&pF
z@*f`OKN+m%$kb-#urQ}hQ2C;O`JJu9A&(j0i<sA0-FZ}s+I{pUpvfUaHnhnsZX}!~
zAjv$=P+4>QlQO+*6vu;jX9av8hxB$3J^v&<mjV{Q9YojdrGN+6$xc&?PRaQ$fNFbR
zx`CEWTyyr%_rIlVJfdu(B6J02q9Uw+flmY=HV-w3Fhl)k^Pv0o-elG&?r@4WbY7<W
zN#@7PPdlCx;+X+H+e>%I<}}~W#_k1^>{1u6aaX{Kj2K)Q)xJGT^jeySLcld^e4#J;
zb6|;19_88Gb>>|<rm&-B+f=bJ+8voD?RQAwc5b0F$gercn1QZ5%$+8+n~JdLrrcxd
zjKb{}pO+x#>zp-$d$wgo#(X^;sjQv9a^|JRo^QO47G14^DDcycZx3hn-yR)aB0N_8
z<2r*)vX$eBGUt_}i@GgmwAgTq8(yDu=b_6#kodhDqUM%pCf<)eFBdNob9jbn!9`R#
z1Jw|I4%^NR#myEAwZ9>DUQGOVD23I8@}kBaRZzB9=Yxd?t@1sKnzz@RnjcIGnx=N;
zv1k^URYDAHgCSQY#26vlg?+j7*yo;67#FRe106dFbg~4^M3gn%yJ!whr~9edF$$?+
z7}>cfo(66Lu_3#=@#x%cR4wwndFBE6ViNOHR8avNLFKvP@muQ}2jcA6r0H~qEH1B0
zT83@OV(6Y4gu%i@^(d8($_i3fyG%#7o)$j7kkV#^CoS6ROqY3=^?)VTng~~@garbx
zdmD&<bkNxf5seyhEO{#lhb3sJLHGl~k&q`r&pPhdmTu?|SF-c;V(2ZRo+XhwhH*`V
zAj$O)f4k+oLot>_Y8xB&&Te+)I-4gFMX#%D!Ls~6@ODK>+L!?y#n1}XjLsnNTP45m
zzbe@sY>S;^vJykIv^qkF)9*#tcPe-O4itPWYOj^ZmyhWB@~(QS37kY{COJ`Q%OY0i
z|0d;rP>#jCJx(Ecjx`D)ra2p3y-jy_QLfr}!f_o1$6KaOCOFK`Sl$u5q_Jx~XHS!@
z<Ma3!<~<azN|~&IEgW^6qY=ov+Rtt^sAb3cz1kwzL3WltGZ)wYZD58e-f`S}yFyAL
zV{E3oO9*ZCVX>%1GM3Wc(%`U8*de|I_U&!DW*)<mh%nc~%vYFKfh=X(!$m;>tgj0Z
z%8H#IMWDNyrtC5=oL#*Z%5H9Z=AAdr{@^!-gTo}4Tp2!845`)eqeZ%_x?{m<0;g97
ztYWYT3uong)fDOMXVfV?2`(4It^8VA!qce-njhJJe`{mx3moPV+GKCFeqCeD;*{V7
zEqWpJ1x)q^{po|8d@-crPG-QoKvC*qk0fk)>*tfIMUg5#?+k54x$4O$wSqSIlLhp)
z?i5YI;*$wj`Pz)Jz1`MsXhm6<7|HDTo=T>c$MoWSo$zAS{KN3@D_gUzwN|j9kz7<l
z(dBmPKM+cx=$uGsSa76I5l?%@I4)4a$Xxf<S9_@jT?~~*lFDpV6)LRt<n~|X7#(J}
zkgeGw{0&fpW47{lRwo^To=qx8Wu;wFLjQRa1XhZSxD4WJ?{HoyzYuk_c!Sgy{@9_`
z(d>0I!M%5QBk0<55rt=~=WThhN9;Wk8qv8O3K6-=jmgo80l#tp#f;6nXFu^zmSh7C
z^zh6`Fc@;){pCYyJKMJpJe;pvRK%{Vc21TjEb3%n^Yjb!tQZX1g>|&8oRwqn497)J
zGl%@Sqj^+OeLo{vJmqUJ(og1;WnuR%S#)#q`ew3fT``uFrlz!tD^7Qwz8l#cTTppf
z?wGk%O_b{)GSpG7jD<l14I-L^B9<S?EADtI$c71$3ki(V8|!h#U1fe|n_FOt*s3US
zg=EKsQeiEGCc6wPJL8@eja4qfJ{O5Yt#&y&Gde6=x`Hbnp*zyFT>_}bLO4{wE&-kW
zV>tI)H0ZzY67_GY&L}b}<S->Cn;jqwfByD*i453u4x!{sT=@lAmsq9gQozEsoTTa@
zx0*^yr0SmqX;>Zvz|{JmE{_~XOqCX?8Fz5rT`-fi_Epj2gkvWn_Nyl2*sgcRiryDw
z{X92OZ-5cnXK06hN*v59Y!ecT*2sHq+Zj}!!TmVVg$vOKL%dD8IbZwnOAd$HhbMA6
z8L~eLoL$>hW!PB!&HG+_qH=d|(c#t?(TCdjiau!^QDFUT#-{cPGt^IIG(yKSBeI*_
zM{ql&+lYK*El~O~YCX+ODP!Z^D2X*EIBMr_gGl8hZ=`&h?_}wUyjk5rd9`81+t3^(
zqeN{t=NV+gVa-ME!NjrP_X&MUsyX2@uZaqt6ai^-75*eKN;rUP{s%I%t|(3v=p}uB
zX|YMy6h<p}fRxgEi;7I*e~_;g##`Jpnf7!=x$__Bi($S^XdI9QoEIqC@}Q$6_j!%!
zQ2JXt8jc`&)F9aFZWmWn+}MYhY{}H5?k7iabDhH#eDW63beg%ki2Z%2a=}KIe5N#G
zU(PbdwBFR>xjq<M&!re#<T&YA--U`B->UZd?$0A{oT3)<g_E0)=9l@R`G->KyPwBU
zs9~XdTDftmn>YkEZ&Oi&^MDzb>&so4;rqv}cs%j~fipIYcIdPzAHW&Q_@=8>6RL;R
z#%2D2-muek)*C+c6?^y0eTMeM*yZ`t`tyKJszg8N&B2N;(-)iVmYHd4NU4$T=Vw+9
zv`_b#<c$s(?|2_B+g%X%dIbFNaW=RQOj<j79fU0k!yjYr@W4Lq-xY;NSQpAatGhYf
zk=jw;a&@O8MJ}VvFjGUzx>jP;(HV;7Z8dIgQ$*5<$H;xxPDrOw`qDQccito^_^&lm
zvOQrV(W_Rdnvdw+BparLMC&1N*IOSSwD+wJ$(CuE(?(hR+R76VE!50>s9>JBjF-Y)
z*!!Y}RQ~8Ujx#^4F+P-^spol{#tEcO>YiQOm*YskfaF||<4~1e5}UHQt=h(4|AF4}
zc{9YtsEcMJqLo%YMl!8B%vib!`p`n!I`Kx&$E!yP6p(|eReCLOnpR8F*Sa8qmIM~h
zPSr3q{8BgO=^2C@B3j;h4B-Y3Xg~V9Sh{KKSyO2!BUOq=t|uhs?pPfJdx51{sg(go
z3QN2u-wtO56N7l<I8SD42+Q|J0e8*Iqgbu7xGVc*V<KG`p-u-X7;kiqy+B?R>u8xh
zfU~I)@O`me(roBGr7a`#qjxlA19!<af}5z7sue>Uwuohysb5N8dTp>$<$pOtUg5V0
zPG%If+1nHI+%TanQ!gFSsq&th`XHj?6x}Kq<av7VMvQTE{o#<?E#1?vZnw()2aO-!
zTsdok=2c9=1=|?ym~Yw2);P9%ZTh0_1sxu@Tu|Yi2t7ViRpikMA5P^4)d>3+Uu+!w
zu0?jFdhYLl;4V`^#P}<U2XOo?nPhL&lYq(%_{Al~HOpdw8&zrS3VSX=*eSN^<sPe8
zXgz_$0)OH&-cd0>gt4phPW__um9^eed&0gUVCw1@w9F`Pf)-cLvdo)NSD-Gu9|qM&
zpi$MRFLe5*tolXG?NX=fLZz=s7muv%)z_zI(RoSJcIMv9q^a=qC$-FN-w(_I%Qn;W
z#6M8S>DTDjF6;nO`N_}-HW|HM#ni=Pbnz(P{K8N?{Yjpr0|a(T;pAJM#5_=}=2wY~
ze31NCi_yMqOj=7|kx<wYEi7Y7$A+t3uJV%rB=?WWBlDaF%gNyYNcca}{Rk=y!~W}9
zc=jq;0RXHGTR1%P@E{XOYAw3(e>19s;6qCYMD6KTPP6~i_8F<G{z8B9=)0i+;SU8A
zKzu6ErE!3uRbJBNmzttsaKg0t=$}LQ0;_8IIuYQ<;E>DHI1g6#52w_If)sl?SDeqN
zMfAVMttByPelqV2{o0YFrCHsrw!V8D@kndEPH+V@T>V+QA4L-G?HMPgS2mb%hUu;4
z6P1324Y_)~S*r<b^q2iHa6rQ;UL3*C!LeXGrKJz8iLPR#W7&HYI+1U7Mqa95by6Rh
zWy!d6_SKE5VSKy_uD2oA(=eCz>tWxBPb6b9b0<ThYIVh)1B)KkuEF(vmRY|<d;P_Q
z5m}Zm!`^km_yU!QRiQ=EKTt4}8f<;2C=nf@!edlye<utMgw+(KG)HyfcWHFxsvl{s
zB{}hPI|<LgO$6=V<@Ajcxh7lOV<`&DI~u-v_z4w3L9$S}cH5m@t^;Bq{T1ZNuKG(;
z&}5%081(oG<w1A*B+(c|%^&OG1_G#LC%NP%K0Bg`aotDj1zpxJ!h7bl&AgRg%A|>P
zgUi02gjqUp&K{z1@jw>q`T^?}*HOh_`9RsYhTzjdV`aEY-36IFDblJqU4*%clj!sQ
z2p_q(c1Sd_C3}}w_N)lk@Y80nh?`KOk^Vk-(Qr9xs|^G9_>wgJ=?ZdOLf~){WDOWL
z=0GW!hgXbO%cRHm=Vw6k{7NjO1UmNO@xkZA(^WX#k?dY4+X=yxqp2CQ;p#mSpvHe0
z?7@%+bMb0~A`KzcdrJmaXk3bWT9?wEUWb(F`x53r>C_S{Nx~+tuCro`d&xko3ScL7
z5Us4EAY|v7-~Ur4W^sxXg0H6W`w17s&E;0A>`j5XYpPBkUU^97!jrj1tUX4G({Mt4
zmimAbzg|@!a(|lmv=~0>GCU(p@DQ1tCO#CwZdMxOZJ&qFFZq|5JTOiLlGv!jd~TxQ
zUL9^jW7k|Jn`83T?_+w5dvU!3CV2PFhT+97mGZiWcB?H??EnPoe{Z+!AO80xc~<_L
z0S$Y*9g2rkEuB5N2LoM=Wcv2(k6b|lUIJnji){xG$(^SZTu4_C*kr?fy^}Pc?&Zf2
zi9muN%)cQ>w7oB3QE-8*3YJF8_WT6d$PlG^%;LD<o#Nx|4d5K-m8_<VAyW^UK~Ym8
z4v+_SY?7g5S50ipU@Yw98=G@pw~QZ>S3oEl>t%Yc^2=G(w5qd(cY^Y`n@LMBH>Zab
zTndCAfx`eoda4wpX9c<MUwrE^`82~(<cav|D#+-}n77r~a`*E|eZ*C!051vNd1=pA
zbXbi|{4O!=mfg(NT`h!>#LlK*1Q`{i4ro-hfnr>CW*b4AB20nft?EUkq|=&4hR6rl
z2^pBjh0N!xQL0repJlun&|l8(l1HiM5lzG*$Uc|%R>dXKSxs*qfLK``)cN<+Y&oC2
zVz7UF;&$k`e+YWmMiNj7N|Rx$olKp4McEmObG`@{KJIySmA;C_y4Co8GV)RrL<q=%
zJ*aix-`qdC4OmZR<eJ0Za%>RC+lPT8uJ%$zbyj(LB0#Y$Ui3trDFJ{a^f*Q#k6nfW
zBmF|U%I}ep^I#~RE&+{;0dX!MBds#LvR7t{EB$V2tr=Bt$nfp~jhSNwLL@t_(Q{Hf
zNERU@qh92l2>e&lq+f_5K|mLTHy?f9<J^xjs!-fD0rbeICm>m4ehHLU-yqfjDme2L
zPxEjE@oCY;)p~kOi0rStx*^cdxy}h-0=E~J2RcA}a&35~RQ3C>sNMsJVfDR1GH28Q
zL^_a=uWq79PJOU;tk76Ra$&cW>80R;IV=G#HGx|z2cT;yGIWP~q?1O*ZKeR^`Vznw
z|4Mv2`;*0S`(;QQD7`0_5cHX7+(;&Mvbr=#2G*;6zTkD{Ow0!e$UyH$VGNx%|CHht
z6<Ath@H$o8`iiDkc$@ESI<X_c#=YTP-{=2H%Tvhh{F@Y2;eqv%!LVt}s;iK6##)~>
zOTS*bc_MlCwnut`>_3-%lz_%-MjJFmGv*6Bx9i1&*MFBS@+xX&E@sk?^TVw?k=F)z
z7Jno6ud)m8#pjj0h@cOz6nxoi`t|=8t)JY1pLDA|Z;53!tNYZMMOgHxGwy+}{cWAy
zLmIFD<tQX~2H3(8HQwQi^y^;Hb?#o<hyU$T&kX(pt$Gz7l>Y-Q-xk^gawEB*C+`Y@
zc8|lTv`(HY?;?xF2JeY+p6}PV;Rd(MygLgX3{FB@hsS?yPwg+(kt~o<KhX6~rW66>
zNnC1pnpWk8$yVV7^KAJ8X$$0PF^zO>X0#I?ygcixmU^uM4Y^rketeSby}$8Dxww8|
zw-b{s^s=(|to~6lv3e!X{`npu&?CRXwP<i}SGIm)BY!F|clyw!FzC>>HSlPN>3-`c
zcnn_G`fHG;{JCU`=tqk|RB={ljBWKdfRUgalE8BJ^3+z@C;TO^Zg!V^WUiuuzc8;D
z*;vVYh3}gdvZgsLh+~KLTiv+>X5c+x=jxbRiT0Z-&$x54=Q*A{?ll#IZg}b=Bp7vS
zQWc6n7&E_n$7!ndtB0y^L=QlPCO2hiTK!!Pel7XE6{E6J4K4J=c6hYL$Y&m_==nE0
z>08HoDd@%*5)pK3E=u)XJiI*cLMB0Tnn!PkbyYY1R^;aY7EOMx%*{(Pz`Zi4^jMqW
z#JJCVF+Dqh#uSL{85bEl;;*&A)bUwZHult>`FuF}lzJ0?#|{0U$R}|L5BK3GHfHK<
zK9d^~UtBSI6HCzPy;fFLY;>4BOqy*zLHJD8moQ6JolEW+UWKX!JKauGIz@>tFQj@L
zZ!M<ov6YLR*ZJ7nE4sAHRG!#XN$}aLZ-<YMzxH?J=_LNxz7^Q@+8GrRNnK+I%qCRj
z1jr9`#m*M})%~LJF6c$2ABYm&<z|2C?nl*b-}#@pdH+q<16jwy<c>Hd=sx>*y(ON(
z&&CT-VxZ8vl~*c`UtJNS;wlSmb;*R?k5<Ce&fl^x#c`Bps%N(<u5(O(HEeR+n<qSU
z{VV1A^H-<h3eU&CR+V?HD5~SetgeoKQex)HzR4BoMm`CJ(s}C+^$D15*ZVN7dWO@u
z!CWsctR~{rCRM^=6<ceJu(7J#y|!)v?I6RiBuNbU8*MACG6-9sfs=E)Ay!x^a)rls
zdzv7FcmEkp8*ebPzwzS_)f$9jJVB)#al}QLn136J9LyKX7NvNVB9`cLpJyfka)q0*
z$Did|M?|-Owb8XmLRyY>6TbsI<?MvIy)}d~_L_hYnqG?1>+Ys(Amu%Q)(|G)8$9tp
znrP?^%>$B@H3NEMviZ2OxnyBCutie|aIZQkd{U`R<&X+^b9u2WYP)T|pz7qUwSAqC
z31i)b^5mv*_1JeDap^=}B{$*obZ3?5@H85P4Qszs%bu=lorqJtxKUK!GGwChHLepq
zUYw?(S>w9S%h09eRlLU`F}jqpZnihN!#AOCSy1mFV%YP$=nRv=40q>+JMnX9eLj{1
z?wVTpi?+}$L&Nzao*Mk;vvK>M4t&Df+cZi4Kr@(I{KpGPl`PN(y<=uT-95EcZ2t-j
ztU&uy>)E6g9V5aePEo1_=yowN<k5xrrluo3P#eRHZ?-YMOt4A5Q*eH#iuvYsV$(2B
z87xlq{L=TKu3|5VXrdW>mR)GCsrQsNeZE`P5tra)3{YzSdVw>h+<&606W5N_T+|&6
zb!-i`bRR<1M%sVu+seqvBv2}AJ)-4?JdNg<LvG$bwZWaaks@9oX4xSZUe&wORD{+J
zAO`d}&;7`Jm$c65H;-IzYTsvf!+p-{x6w7`I{=<KIpU7Cr;^ZV+-J)vKUl}M4!2}>
zJ*Mk){(*F-rBiF<FaCij)E5uB1o5}$^NJT<MqJeXERz~Ya&^zq8z)rNxN+>6^l#U9
zBMfwqE-c84+P}PCU)1n>a3C=bSti?H%aUo|JPn1B52xbz7n^$`X@cCRg>}kmU`V0Y
z7qd%y6*{B53cQ2d8QsrA_D}sHHN%$@Zg=&IHtM;qVZm67kQyH&A0nhQgJV<etZeq^
z7lkE710*K*lF!cjAssds8;pkT6E+a5XzA~t;QnJ0Pm;xUNPPG5Hdx9eJV=LEx6mJ`
z*TtN?PBq1}y-j$Vc}fKjb-JfPnx8gaK0BW^gNK;UG??T*7V15qsBvtF`g-lPwqodU
zSOu=j^$`S>o2;?`OSlcrk83YmY%Q8Y_oo76;HTqpe{(Kn5{jEV7~OcRcb4ibOq4yT
zP)-{C$@+*RFUNfjyeN<LZ>prgVe`!un!F2~myB6W@kPwGNkbd!ty~g`-2A+Yq`^>1
zW9jSS!Lekht$4CeI&-6HpQ6<$HW60Lt)aRGmj|1l!pmHaNHu{Gat&SLOu8vA2y=S@
zwghYafg*hWpJ>d3OwtV;Vp0^madPEH=|K3Wa23mD2I`zi^f>Y1JK(66DB}!W`pE$w
zBglwrhnsG9dvWAr<h>{fW*JZM#}L|Ay8m4hd<<&0hLEbkM|)dqGYi~2KI+r)@)i0u
z;l0d?sOs8_k5OF+ucNOSoHB30i$;lQ&R4hc%jQ^y9G9(KrpO-l)O3JFh%<T3Kt=TF
z5Ypa+;sS*YD$jzP+3%@b$d|cej1I}{&TMuoeKlVH_bHa@R~@-eHf~kztX}Ml#$tDU
zc}7fU#Duw-8Q^R%9gS#zf@HoFH^?v1JHKqKv=Ul-8KHs~1&@kbq_}Ndy;g?iKfetg
z7vVznL{i}Jw$l|pJ^_rt^UL~RRZN1@KzNpwQP4TxDlylbC^Fi*Iy_gBHn=dAo8q1b
zl^simD>xeC6NvWD*CSX;PwZxUY7*-3>8wOe3Iz!_j$`v(K(IK(`DR1#6&Nq9hw(1n
zzp(%h2F6tMU~1CTCwA1eIQ#M?d0ZHmLNltTjXO3!k>hB?gpr*?Oy!q<!!=xsmd|^@
ze*b)Hw0c4kVCv#TD4H|12QS`67F5H%$zJ$RQ4VKE>P<=WHZmkWUu>N}mDSruOp>Vw
z7`3&THu^Ew_kyUfK%pVAy+x{VIAeJTAmd4{qV+BsJF?>@XgDSXn{Q}}62WFczTw6o
z8f(l`5nw4D7Lu`H=m(N><WWwi&L5wiB~iQx0;DfrJ6*eLWrnZXF}ovppK4VV^}PF=
zS%v)iTv}uxu;eDTvaa;%&0n?vfK%&}9I>(PZ3;fajegbB#@O>LIPlK!`edZa_a87o
z_h+p=4wYPv;B}C`oP3S#esN41-qHRfv`t6Xehv;sPwt<Q_oatdT~s$jyi_hM$i-C}
zO1=I>cpa+(T!VnbhJ+9eQ-P1VV<zVHm*cGF^puni*N5S$-p`q#HTJkfZ<_x|F=_t;
zpbo%L5NftQW%Zj073xxOaN`zqfqS>YgXyGY(hoJ+Iox_-Hqn8VKb)R~u1qod>%tt&
z40l?aY<ZEt34wXku`40~aL&&Z^ZDwj`|h!H{9ARfkWnBph^k=gCHZsxNe`K{k9V}(
z(}&&Y7195hsmD#gPeeCgqu}p)SldzW2GjN85$Xhe@HXfbFoA9gh|ITH<++eJ3d^*Q
zmmc5MA9p-n03{wR_SgqBQYhr;kFV}tWmP<hsI<n9$MZdB)+b{Os(GmWN!p#o<0Len
z=>+V)d39@R1-XWd^CuW@zBaHu+X%A9YuV&xdwEJ<-;ebAkXsuD1T)B2I}O&H*41W$
zhtzT+S88=?BmXC_VhI=vuvJAZwyL?zVk~`cQ(S|6eoE0Vt$MH^3-d(!-6^V${46vO
z9z$k}JV6~039WM=rBAY69_$oWWAN0pP%_Hd#&ER0Wm>T&1^3i{<cq|w;ct5nNFLJF
zepdz$l+qqZ?(CyWKoE9^)hSSzm`iW%*!j#2VsTtDfj|hyxs%_?@gD<7(+mku4v69T
z8YnV8qE0UfQnvs31o&mb&Vsl1se_>1FA8MpR`u@Y{%RZ9ea|{e%pcO;+6FKh`3%S=
zRo56OYp5vMdAZm)?IrEZRfL&5T%jn%_%lTlCo%u+a$`fx@%bxCl`mPi3W&^<Pq8Ms
zip~;}C1JHGF8SE&ymsb>$nI=g-Fr&K@;S?z(Iyi?5QXQ2N?G`WX`mdy>-i%EiGa^g
zQmKhd4*Vhs&u=b)lpYfj))a^s&cnkskkYP@fV<NGZlnefcM46;7aOPAihDTEZZ?zg
z>DAtXdnBB{x}i^_q65eX9S4jh*IOgVHRa`omX1lpu5|Km1MZ~Ro^Tn-eSl7Ud2?sM
z40d#Q`ihY*Va+U+3+8lkCIMiCVcRT@GINWYon!_~)$Cg50xz+fhX<hRZ6rt?RYO_(
z9_V360$M=Kd3zmrSf=1IgGKJlS-_F(fK-uEc)>Ey3Bn5Es^Ab{>^zWR3vIlemEd-C
z1fZQkpsQ%1J@j8ttdor7?#u89@+_yP6qY}C_D}qnj>01yiGk5&<0$e}<nBOmIPNK^
zXQ-(Y<VVp>$w3UfKj}3g?b`v0hTOxns>{4tEcQtObfQwz?7sJ3F)~N1OM3X`?O*hg
zr^<U95!8L&Qy^KGuMFZ>G5YrQouD7!-ipBx4KX)}?_F}nqFS#h^>i}cjF7%n&}#rI
zeE$9h2*ezDn7w@52}IX-kci<)ZP%%`z0O|*|5xlt&U3kV{67it$_jr)r#Fs@f76O>
z?4Iy#zKxj!-8^~Q^Y9gLkN_~6yHieOp@z(bw^m+Zd3(?SF9Kzhl6c2-Z2x)A7%uIj
zRZDVe_bgtWxFIu7thJKJ#*=Fuwk#iqlDPvr`2XhS&=V3OxbE}NsEk71unxIitl0sn
zTsP2h2Xjh4mr-a8Cp3m=#F%RdI*%n_-7NBbBOVJ0EjNu<0di~$Z>Rr0>gMqnxy<IV
zP~(~TEzRYw(ybP|W6`*W8XF$_GTa!##v5O2i{^Iwz17!>y&zf|)qkLPz#*F*>3h<A
zmrw~NB^oa2{R7dvoePWa)(^cLIk$M$g!kRMJ%HPj`v)ROMSJ_$$5=YL75}DY1{NVq
z!*zSQ6O_6zv<Kx-<@l)f^k?@YouBKQKhwzS^~d6NWqbOkDN5rAztuSiZ;7~(7CVBy
zdUS8#eI;p)H^M{|o=3TkNKq!B*v9+<Y-f0MIa6w8zIgoGSKB=H5+AGGdRHsaL2eHO
zsnrLJWFAP3_RZZdw>bJ;1_C{f4w?HFX=}bL!0hJ_(kvs?4<>MnNAq^~8<(=Vj7{v`
zG7sx&Y%HT4?Q!s#TuTchSBz`tr5`zVl2ON13HOk!182BMsCf@Ut6<GiDM7ip2^oem
zXL~)CR5{)=H;U?tt55s{4;`m*c8j=X5O7`uVwqE|c24}B{?7&Y*hEl#z->=Xcl$ac
zPzh*NlbV{+67yJYNI9+aQi;V+4;-zB(|9TNjK>$YPl%p()aZQWaIF6aO8Beo_VpB-
zS0O6+1@4(ymBB}ob-dcTGRT;Rw@I)aLr*po)~>w>+b(TyNHe}iX#Tc7yScefW1Ybr
z{leUrwv_Y~UyMZGq!@&q_(T}~16d6(q}k&sgNo+%*L=ls4f{vg(Zq+)-pJj7-n97R
zAQ#noL5&<^CZpe6J#J@{(x#K*n3J{oq_Z!d6Xsio%>ir)S`AfCV~ZaC2hxP|wAGx2
z_~m(5%I_NyiAz-N3Ra)ssndj9z+RV7ct|>X$E<Mozo+40TX;MUNFic#$s8E;&RDeG
zR9Q+IXw6#m{)Q@N3S1XV>N%jX)m1yw8Xli+g;ahEF1MO8Va9z*(tnbigcGRp7KIGB
zsr~5rx_vtC+x9EVNydDHXK3ojagp~hboE+iGTLr&aeD8Bi6I<+K;_)U7H9l8;gqRc
zg<o&=x4Xv$<&DrDbz?1Pq3XWw;%@!#q8fp;@Wz5fgT3DKdC__+PuLNTcBB59<5V1c
z)swf&ZQ~8CWJsd(`ugH7RRP`rs$=0=cPVuqQK@!fUSr8!y;0wN@FbBc+#b`l*$g+Y
zHnEI7Sw3^4Gc>E)Zbf6`+CqEY9FQm%MsP!uv$Kr@Mm3_-mT>oo(&xa*Zr?2aOaMGD
z*c?y&xP$t6btq28V0+#^odpWrZn#=^{tXI+)SD}*V<yKphb4qe&2a^1Ke(Y#{t3u#
znE4c{Zgqzx$I4k<B8F8J?;oiJ#SsvKv~cgk@OqU7{-5Sb%Re9FcxPjZf7FkKaaLh%
zrV=ig=i7pbCgI`idG_cl6bOMK`f*b6uG;j&9~EWw>xMHmm*z?W8fw%Rb+vO#VXDVY
zd-MA|M7QE?T<78=DHowVYAnBw#Zbk`=Us_ABK59n=#KR|H@n3N$KF&pW^<RS=S-me
zPVnX7+LB$vPTC6K{ncFLmfV|>>*;2j2rhyLO}0=^jf%CnAMXfe*Lg4}U(`Q@8WmMu
zu6<q)A4za4zPGqjed_fmX-kmL(;F1x{lj?Ke!UYZXe=Xknhebh_x;PWC?^Lji`$LS
zgp@AsZRd((n}!E)lP@hW>d>Bk6G2?FXhKGXn`5P3yDi>n2b_UfQ^7E-BU~nJbZ6=&
z|A9Q+Srs^rcAoG_saA)N4>GmuxDnify&3viU>ZR#Xk4>Up6fW$aA32nk?9(1Zb%wP
zvs8FM<JCY}MXbP|m9z)-Z>H92ZjFye%!AqpK%LsM&0iKe;#VtcCJWmmn%K>DqH<B_
zQBP7E{bR5KBb}$?UF1~Pv`_uRWoB}Y4c7KMQRL+qV_w4iUD;v7qDDB!8pi=;zJ32-
z&*oZs@#ix|6|k*$?PXREI&H*`-K}_u<>KWOjxET|*Y#YuYC>lI@Z{~muCDG}ip{z(
zxCmZY?RWw<${G_EtXLey?9p)dK?OZG>IV8dYN{@P>_t(*MDNn&_liO@N(1A#;&`;-
z)N0CrkMTVn%PWe7YN&>L#~ZkCdN-hRE10apyMM|jvVD<eQlf+_-texSXz8MRvaoe=
z=storEoxD>IyhxW^3gs9t`U=9-g`MJO9{4%YgPjIDXO&W?SrcwKpk%XrriO&eW?Cg
zEqY)LzRRnM|2#X?v&^bexUrcG)q!w9)tM*pPP5Z(;S2`AlJyrJS^i9l&I3EWlI+97
zjHbd?&`Fgwcvs$;3)y(7(F@*?EA^Qqq+H%_>0;C|aI>qQ{&N3j)(KO+6a9&;uDTMs
z1tO<X)GGMHhtCG@_*h#dlH}^?5|U`phHLOX#~X(KOfs<~So)ZdN87(y^3hqKO4*v`
z2WRY;E~nkGC7kImro@d$>H*!}pD+{_*G(Wzk^rSN{iw-i!-`%Ey9uPb!TufgKrP9g
zR6fT)sbPu3efnEgMIv7L)MB`$JE57y;sD`ZXIj}PG@r-&I=CQ4ozyFt7*b2uSyjLW
z9q7^D$gS~CY;Z`yP&Jq|JU7x((15+X)~Dr4Cg6?28Un5+My{&$K_k`ED#|dUSRH}<
z*>dK9)I}8oSByy&zp_j3X`Rm)Dim5;)T8AaR(<RpjH4<i9j=|e>AJKzD_w79*kad0
z1LNWz_?SdG_4u07v*s=9dmH}`#34w<+Ye+Q>HKtMPFJ&rsLAg10L(S<bS^Z9dOgsQ
zM<Yoq0V3jpn2GxMB8eQ-Y!6yLX;5H@`52v8<gp{W*bw+r#IVJ-OQ(BvC22_9sB%F;
z6PBnN(!Lk`V-8(!U?_92N?sxM!qUgcNh8X<wxW-Jr0Vn~unrM5fV#tF6muetYVQQ7
zq&5CjdY;B^kEwziE+vTxvSakX)x-5}Q$c8BsSa6}?bLpOiRl5N3Hi6z9#rE6EegfR
z>4Jzker1hFhdmMve0U<|XjMvGZ}C*`=s>m7r_o^)YTK|}+g}*9u;O(feVr%Y6}&vX
zchN2FvbA1x^+v($dUNijl-uVkl11v$D^_eF^^+%LTHSS}yB&Xn_2kW)x}S%uU0Qj1
ztI8az-@E+3xO?xfr2qK;+jd!2T54)qxfht*Tn+b51s9@af&+1noK5#iO>vuBFf<pa
zxXLY#%!!D5<KFvy@%fx{UDx+-`0^8npWwwg&)4g5f86i)n_7i-WO09s8pyKW3EeJo
zw6_E+!9k3tj3j-y&AnaA18=`piMt1uv2z;8h`9zBnI3mHvPfZ7^4z_xAdDT7f#Y?o
z$T;M)ICmFje6zy`Jt=_Vr&H~;xO@FF!4QGwIzlxR?iZOk=m<%L1_?Y`nV^c@!~{sZ
zN_dF^J~=i$DPG>+1&mW_3LiBN{;J$Vr9+GDY}bhGbIik#0{cb|Zl5F;i=_Gt6<M$z
zc=I^$S)yeanzBSVyQ#w86`x@GqyU?A=CpEla*F`NuTje6HlNwY%)~>`k~Ef<pgGtX
z-9U6&bJk$zcClh}Yw-8+N-=EntjS}pV{KSQSi=X9AQHklV`H{r@CVt@V|t!;Ul-uq
zhw$|)y3`Ggcwr|RUrl7FOCP5-n%8?Zh0R&w#4i4cR5^MH2^ip;CJc1<mg5DSX6rtE
z1?%sw3^#VnU9w*kyJ_t>KV5mjxv_U*6}wf<6YF9MkUt6f<Xi_?Bw**Dh<Sv^;n^od
zDHmU%vSlFmq7ydgq_8O*8?cIQ$DV(U!Vn?@SgCP8<c1F*YjiKPSPFm4gAhjh4I7fd
zmc@<4IKv-A)FbC64eo7fM<++&Xi(5a3c4OxgRR&T#$aolcvMF8%O0XsK|7@Q-H>D)
zc!gU1KX!)@+WFRCK5<e#Urnoaqo1oO=sB9R)ZkW<ItpksNOo9<kXTE0Zu}%j;7dn7
z@QMsrsR%~YtWOlXq!2LtK48l=7kxG+TUbH8jj%Ap=D_Y(|0T^<P-9%MA2&Jk&0F`5
zF}q|A4KiOhN_WKF<&yob6?FeN+LoBvcUQ_tdPa?0TBNfO-gf>mVSPb(d+ers$|K2<
zQkp};6|j*>nOCE&_Ej4uIM1wKg!G@fFA`QQFMa#}o-OGzC%9B-qcuzI*1(I)Zjx)y
z)oC%tIX_o`uP=c%lhQsJ7^KTr>$AG-=eAhdQ~2}3Wn0yp8NY4dnp|cBZG_ia-}9BD
zwi_*u!?r&@4(r~1e8pB>OSu77s4RcN_3?(5>r}w;?&><7!EYtMBg8hXX^+X0M&nqe
zE{+sk%$;r4^Y^bMj()tq?p65d_{IHGp1*z59sn8w&0(HSD}c$A^G>KKu+O0A!VMbn
zOM<#s^5cdnZ~$mW)bp?1qchA#NlXP9*w9%5u2<2T{PvbflceT<bcIIxskATLIk$G$
z#Mrd^w^GU1V`P9f$tDb`m2!)jZ_~ZxTCexp`^a1EsBS?myc20Hcrnpl-OruQaz4B0
z$jhFN_qzq-Hcgt?+oKbbP<C7-QRZp&o{Vvf?lXQ{+`xKi<$&0a7ZYi?xJTFY+1<{q
z1?kkW*K4Z=ta@KuH#y)rqGQs)M1I=j7URzddU1J!4bdF3MLS0?W<GaPYl<Z&eY_B7
zug$YfvkoxPzEyLaL*K|Q-NSomRxZCRJJVu(9o@?mu8fAUkHSDl+Q5~<M%@l*pjAaW
zpc$8&Q6F6X(jf|JceqEo)J7<!Z!S(8Sf%^L{vs%UyX>x{GeDkDE99ZG1oE|1{;lGf
z?!&*@{U|h~QIf8M*Do#1a(VN=rU(;tRcK*-NjfbkDWuGAH}-~}OkJhP5v+9gi=lT6
zBCY6dDt@H99EaFC`pCRYXo-bto)F~rj_K~ImvE|$sWsO%G36-9xdGvUXLjTTI-~f(
zBLq#gP<-~mlvw?d$*<6=+@$UqSK%_<^<Aw!`;x}UF1mb;`k}&fI{nnGmBS`x!Gp+G
z+gYv`NB>VY+<m{4Pfhc+Xszw%PflOln`^xC|0f)gjU>u3YATB}lw<$>>G%EYE;m|+
zQ%SA4kKCe1%?iEs>3@D2FwRYq&-Z<x^0)>QjgvYwpGE%pNt>v$9WCyh(e%TKnP)As
zZ|lMT#i7PuA4-b#{c>j*y#f1XOjzC1^joK?TkDKC)s3Ee4J41td4q78EHy`?$wm-P
zy+1Yt)K>fXexdcqWhFJembv!fwfh9!Ma`%du<Fj`I9FwP9|zKw8>6XI*U?(zqnH<8
zlswk~S5&Kxu)k>+d}>HqPQqXB%Gt4Mxdk}_)d11C#qw3{%)V_<ZEoILB=E{P%U;bK
zgt`-uD{=~;xx=a@=q?b$KR%^3%M>Kz(p!eybK{+u-*LFf4<Kpw&vp9ZzJf8Ftq_jl
zcMry{3J0ySOukvrOB2k+ofJil`(#XLd3#(~0)FARPIS1Yom^8bcC)Y5!{KT6ODD$+
zwFmQ#s_(cuB8JsXlpdG;O5)DtWVCF9d*L<Zol7-m?5ppoo9VfQ2quNZWO-_0I@zvx
zCHku&rVIBp7i!i6KybgXh@Dk)c&lD+Rh?;$#ZgjIF%!8te6Xu7qKb+be0@^cxoS*_
z(~kKrE9PY<>}V0Ta+@1(R3)0u1JIau;J9NCRJpBIw6wuyDxA3F;r=W8pJV)4yn0fA
zs%^I?DAZP6rB404lHxgi<|o;5>{nz{FyI;axoEIXfl1#{Y+<*P#;dpCVE(rWE@oRP
zW?qWI<;*p<gy^8U8=fqA;t-x^-&S?fik&vxX->LF?ORv1Q)pU5OQ&{RN$QH#+Ykz~
zPaz*^%7IZ9TPNR9Ug_ngT(4OZ@Xl0qA^j(AyDz09x|7kPe6EmLYvdE^ky+STS7eZR
zqD2g4w(he_xT~^KV~4aSDJwo@P(R2{h+&V_%~$cOP_XJay4CwVH~-{qYNp!vM@o0_
z3&(+1B?C^8BhsVcGbj~-cqBXHqulxD$!~ipi`zO|GdFW-@KmzR)+n7Y#hK39#+CMF
zvPMRrg>Lzo-G`!!IV2V$GVLSV|9@w?A`H+O76z2|h~DP->X+r~x@!;kk1q5xaz3@R
z3Aej)MsewjfDJ7$b)}HrPmtVd3e8QHo%R>P8uA;$E#=ry)(le2PQ`)gxJe@RjEdp6
zkNhxgSx#M}gxCLiR@<wtTYd(&QP!21-)pl<bW$&X93UB_YU{no-b22XwY6zd5h6W3
z-&RKsvibER;Wy4PE1nV(|M+#Xf~<Vkh@IYxzq+s^KaUPm6q;)(@a1MI`$Bb52?WK^
zYEMq=n!m6%;~i;&`NyqS^lzN+RI21w`$Uw3b+ds+ZN&u7t&$jML<+3Th+IMKwhkrE
zA`ZD$zHf5_l|Bfjfh}wDSJMyqY7f87@ip{m?S01XW8-A`KRDyenAaaUTp7zE?Nl9a
zU!skBQ22jzVmTLm?)*HrPL}SMk)6tVtv}6bzG4-t@sGc@WEIudw@Tmk|6bSoVsXCe
zX=kO|pogrtst+=Q7tZz66ZjlI2n%D6&iAjeeQ*qV%aK+(QSY1RJ)ha9`l{m@D@1g|
zCA+e`oEXo2*5$`1cz)xNbE@mk^_n$Gd?{pqkuqU2Z(A!2>}gaxfrFCQ*^<yo(z9zn
zGFEVUSye~l7{Wcj^~O?;2Pb_cGIotuY^HZoM7s3QDF=w~>H&Y|_OMXflD~cz9ZFu`
zNOq7(1jm}I>`7H`+h!YbeQk!53iP*S!s|V13tA@9PO5hnLOHip^&O__oz_9(_u`c*
z%@<qOFuO=v8%&oz^Baa{o?GibRZ41{D9q%T>HPM>QTGiUB%m`WpZ_~mT4vNyvF(^o
z@gE2~0FaGYDJ0tnA0Y8WWt$^&rs2t|fRk0HMWNhRQM+r??LVt#OWP+01pwjx6jo<%
zqBNx<NINAMQg~}$z?=-<c-SjxtHo^;u5OW(XMxB*!>u1MymuY!z*kQEl_@c)-`)h_
zBNnIQVJR}OirI?Nlo(28N_D&C_-nUwZ^e22y)G=`hTET-P?jo|O<$7o`Om}60p$em
zm*3X0%H>*zSiLbnN<Q|04a6$z^EXt=eI5Z(fYm=68=l4cR=V4HFE*x52A&-2rQ86#
zr;?jHBx*1am!v?uI;dCD2_wtC9(ssS*K3RvJmif!F(@V%v8?wsH?HIB8)(^fUw1tU
zwd`KR+MJaP3#|jBe>X&{inVt_fgz}vpY|X>X*-H1!YRS|G69t%OWWH|bqbK%(3(@@
zVn#z&=`+OvDQiIFuHX_VfJxnMjWkT3HUt<?J4t-+Dr}e7PdQLCU<!U(z7R7l<`ier
zI;5MlKZ#qO%tFr=OUtng-L7&Q=9C+Z3`#IA0rSSQOt%;Od#s5z2uOR$K5lHtT{r!N
zudYJti6ZD#{*d7_KldI*Wah+Mr?BG835fS*d7#{!mp?Cw+ywYxrX=vNd=iQ^T@^${
zx4+X&w90w*vPjyhDo>WH;NOoLZtXhDV-u^0BRQy~*c0`yf<1gc2asz)RC6OzDe&fO
zC4zfC2qL!Zbeb^d!*D}tXWM6U?H)2Af`aT@&eCKS#g7lq_l`V67NYXb-#X*L6OH+e
z#$ogM?5MW^Spu&7I<`6lzfCJbU6YaLjOABQ163iZ^>E>KteuYd`&Lw!H*?4D(`!H9
z1r*bUgipxvrGkgO5d%MzgwtCv>9LFwzD!U@M4yW942Y%cax9zJ*SJ-{qiH9F@R<>|
zFgMj(GGku+yt|(@)ISWdCCP}yLu10d3ovD23;5%9@94qN!mqQs8j$9{%eQl1bC0n2
z<=!@Ndy6Qo(Wx<uWHVx)p#GdkWE|f!^!gd<W+uxO%elx7^09hbX!*8z2r1O|uSzNx
zc(q^<{zU?CtK<F6el%kbqUz@&kB#i1Wwd*)sZy`1+Jf((v`2PXS89n#PEE;JL{S7f
zNXK&`>moygl5Jpm%UBY)qE5`SZyjN(ko{IulGdwCNRC+7@tjB9bn6LpLS&gxfA&g{
zP3ocuUi0H>6B3vrd=F=%-Q~knbxi(0%1B?yPE4NLs&Zo%`D5F^T=yPkW<_J_97Ia6
zddbiG<KR@Tm*wu^M@O;yf>NoKQ+|k2o|y~mHS<9?1Je%^nGE6>kEPf(Fs|FjW)ZFW
zTpu8pcz`y!BYxI@Q;Ywckxs9M1cY@%gG`RL=O;wbRm?ZBk9%NEhi(sQCr&opbHPSO
zUL)pQ5*xLG-h9QbRw$`Ih2;HKZZG$$XPU<7Nd3a6Xr?UwZ|Zmkh93v7GlfmQTDI>G
z*a`^mvB^V~>3a+Y`QK`5>*VB!nr%%)g)%S?Bfz~-4BAEQT>JZKKL0Dnd?mKW+FCEd
z#U<G~Fr%;EN?RUaxS@tcX|bwJ0iMlK9n%^2Sbk;xI`bx$5}ri_9KC*2pk=(fcxq*=
z|0m(NSR2_RH8`lF9iO8te`mCH)Tr?jwL;E=sS_(-#@}HOD_q9LYFL(5(mE1)Y_dkF
z9gih8%f-Y+7H#j1-T$X=TtE}MC1g#_&1y@|WFf*p>SWUFY8WuE?pll_ulwf7sR2cd
zk3X;lxI4CJu|J-rkF#|XNpwzQ+Hag30kjU{M=cJr2AWt-D9)OZCn?0#;{X)4bzyW;
z0&p`{nTrCo?&3Ql)SFDmjHeGFCu;7L#{O-#B1M;t!pd*ndOEyV{c9yV5;Y+@KConx
z5v^O=iccDnk2SDd=^+}X@GBPHPPw4L!ia0;#q;&Ve0GzML^|!y6`_bh<0UE9?XYlv
zCxyxQTf=6)D_xnl;90kGff+6{LwP-JAdNX$Ae3Mc&94TW!WSrxz1DxO`STU(L52)j
zb~b^@hY1zd0wT8^*=13@{e0Nf^nKTGraGRus=C{MA9_*u3#^bqhQPFcr~7{V(Dx|w
z4{3eUhhmm8_tS)de#gq+AjtB<*k+N^4*V>pA@BrgE3E1zx(o26PLrE##bk1jQv{~7
z&dDtI41uwml>$Hxsi#c2v23BG^(>ia`<yYEU}jMwVS|{xZ?bE_fY)<tJ2t&hke={5
zbN$=$w{LSP^oF~-1Y7}bQem-iZMkiMc7Xd#CXx3%pomL<%zGCU$U9qhV*Xs>6?ZqK
z(e4TQgy|7@r+0v$zEBOY`CF?puXZ`v=zg$p9Q`|ny2SQLNRM3n1jg=zHF&PB-*;OA
z)+d&xUq*aJy4y1TXC5!$qsJrl9_cL&2T7F0NMeQ+7ln(0qNPdEmiHV$UR{NKIAJ5l
z?+M6)=+4RmbUCL}C$Ll=-42x1u^wXm5VE6PjQfv{Z}%xBle5nMtpuZ!*};Sj&Uq?`
zm1OoQsI8goj5TgSM9f(dKC4MU?X$79dbZJ5F;Mv*E1VWalUkgb$_fPqB|3<l)jpfW
zpa$P)BpDA8fl(%d{25x7nV`k5?T1WLbnXnI?{|Ia#=-L>c9$6Q<;U&^fI^bCzkQJ$
zqOCa4EA%OAI(A7^D9bo*uKo{ow%xhA(X01}IrUCJ7|oo$9@M}!_Su$l;#}X6#`<|u
z4Eoh_$TSuj5asi3!r6^Xzti)*yL`;ou)d7rvfFx^YbG~bJ^G0@-ZdV7y;!?oz_E;N
zLIqjbBh1^C7ZE2Dnrkvm`}>UE)s0~j_*7L%jdsk_o{NFO8o*W(Z12d&tV4L97S=>G
z0%#t^r8vUK&82m3fe{hm#Kr^4&X~I`I(4l!=fGv&na0?TL~DOnRaLhu#HRzirXp%w
zH;Hy3kX|X(F+6U<F$?(9ddtUi#OEU#{X57p>sA97e+e}Q%5+lgLy`H;L9?sa0hK7Z
z&w$+tJ<1b#V<`pR_5DWs<cd;>{_>J$fB_1H0=}XNc`~b{wv$(Y@`iHYVEJS-mJn&j
z+&oKYE`7DwH!LjQ3DYT<hmvIjdmKq0BFL+egvM1+*n%BVpM5Xo_$*3ietR@i1`j1w
zdFU1QfLSO$Vm<PPm!7A%VipS;E6Vx&>U`sIn+-C6Awg<3mH;DI={oWbN>Dzu`9O=`
zM>p#v6=_8{@Ml`Dkk@{O#IcOvCO+`ulY;y$mi>X_XUdjmch}LCd0fkJ%cZE$+t!+q
z8Sz9!3B=R~<-R&oaI_AL^SUd)f*U6XD4Eh4cL=}bcZ{Za@cnCR6Ko$~_yvM;hLdF0
zeulFap1LKxM9kt;vc3!0cf{Ao90UG{d`JlhS~8+}_ze1XZ%ob2BFLRVjjtvZ6It7z
z8x#)ajg4exHYS5C|Dy{JFgizBKSc)?p*F0e>u)>RckrSYeT=@cY)5s6)BL9^a6gu>
z7Bs(#7GuhCc6h2?#>o5*rSq8(O4x!l|69qwK?EUG%;FlrKXxTH#bVvq{aRw$h{F3U
z_@IE%dRHrEcf+q6gw_{|_$kZ-GAkDV<r>@aj>cIA>1w;^Ys1P7m-3NV9l$<zBXjNK
zbYvcXbL7)5G(D3^nSG&M9F`C@FvUkssOta;)M(-CO`{5jUR5KC91JpQkA`MnqW_AB
zNd*h1j!!;_1W}~g0_mmh#aX1gll`HLilz$3n<UtAL+^TBZx3>0>{B>I^eN<0ne0s`
z5#iG!8)xz6Um~wJyoL<vuiR;!(M}P*VKmp~TA0c!1fXHS@|^Q}uF-1Bg@sEMPZdzp
zeV@QYPn>2_xpi?$TsekVx2hn##n1892b}M085z@$)_Yn@;o{VD!sVx+^(t$GU)|50
zkjhsWSM3{eLT}>M>Ts*Z2@yc=s&iSIzUI*-*Y&rAuds8i2tyP02OC=IPq376)vlFr
zPb>SRvkesY_sS|ZM#e4WH(_A}A9zyvdZ$Yw&a&qpiMzl*=>)e$RB&Ff2}L9!Jc_?7
z^8E2{<|BE#JT<*C5K&2h+_lNzM!P<@A9L4#mI#he;~Son^MJb1GI-@~VH_DEU*iwe
z^1zTuph6<Ew~q_%bRC`1JsC1jq)xW4y%XH8#9*xktEkKv-+#r@2=7o(u`vUGqOq{Y
zvKwqrXhpk6TRU6VYpF{+T0JvXMN&dij%GN`YQ=lEfaVeJZR=uma#L%NWZS8FG4Gzm
zO#a~#Ey>x!Ctr0o3?7lOZOoYTjf$*+p=ZGo<umpxBS{ToSBCEKL99r^ds%~yhd_uC
z*`0{oSr;5XFB}9=aOaI8DC5jPqKjh3(*4pZo^6LB45kRBcjD`0*@$dQjfkvbbzkpu
ziBV}C2VT`!I#xF!mv$}h?2J3GEGMD{sPEcM9YH7Xi4x1b!WaM-B@a}bXwEy9VSIl@
zSJqC?7r-+uS1|i)rt`+3wgiP|uvHIt2x*jttk)0l)|V1*>j=c&<n|y25uu=!5N!`U
zs{Ti^mQmYy<#-|@vpL@iT+HG$*2AQB5#hDM#G@cy{^iNYU=hX<a=ZPT4lqtwLh6Z@
z1iMLzW)UWZZ)0+vP{Syfs=-b6+ODdgDlH<9d9=28Mv9`U5ksZ+`1&|GZcNFb0T#by
zOh3M#<I=}7U6a$ItM%tr^{l?o)H&z#$oKhgr{_C8%&HL&#|27tl28WTY7$>bQn*a(
z80D`&#@@ClulRiX7Y_ecOJRJ!HJV1-N$=>TJF7u37}vh(#x-2~ADK6B0r6+QIwu~@
zS~<8nnB9x=VK$)#Tm8D72+l4L6;^{9I!Lpiz3>mwDyoMCAJAPU#UKygN9gJV2@)Xj
za2uzSgn&U0AdsNeBI+Cghpe5iF>VZ+G=5x9Un%yfG`YSyYW}n<^X6Q7LUVP}Wa`ea
zK|?@EepMx`qxQ%@AZLxd5T5GNBj*EKbVX6>Wt{4_AzncceTg)~ZLQ2YB6*l^yS$`(
z481v%nRbzXf~7A45?ET}f4VK#h143E%FGB*@FTQ=o$*w6>+XLtYpT;_KFGclMC<qf
zbYRzY_v|7z5$rKr{vVw(|G%ZAuH}(SiB~0(w)nJ^1Jid`CD-OVR7v;`w`%qA!Lv9D
z@SGEl$gYz7BlAn<?ZR-{cT53z>n3|iL?#yrH7?(p?_(?kliJWu3_@KDz8JQ~lCNyJ
zKbb8UMlwP@=@fJ*kWJXcMeFb#kQ?@E4_vw?6y?A!&RWh5cCYOx9_U{%#Qe&h?NzH#
z`s|60H|jKTY-Mrpd^id6gBg8ssSzJe9hzi?8(F$0WIL4Pd&wL^VxQxS>T%XS5WTEz
zY^v_f-i6ENzEQ;diPpsspB9Ikg}wY>rhFar3)keI5Ilo;D_}*hS@nAm08}WSzpJw}
z@s`AB9s8u4R-N!49j<5Mk%}g!sLtJh8QfQ|yf}_xvxTcCOa-O~iA6Cj2XXv6vYTO*
zR%`HgaVqbWv)ErSm+NKY3YQd%oSh%zcW9LXpg#OrD7Q>BWd4A%IYw>)IOu^_zZ{)E
z@ycj-i4FOW?vi|($7r3B_u&EKOvxiAtvGMCK;=*FJ{%h$gP86ETVxSB@gy$D{JZCE
zmfE<ajIprjx#F4d_C~)eVAWbk=Fm22=0<Q9=N&>QF7ua;tNVfT;pATwWYrNOQkprL
zG!p=w@GmShUtTMMAmjW|p(BYQ+C{}Mav9h-a}~GfRi;&5zF_^Rn;R0bO(cFOL_6a<
z>wtJNDm3NtzrJ|g)6XkARq;ppsAWp#q`22Pw65QHV99dJ=VmH+=}1itR3GP2mFPFy
z9i!%RdAhG;20O4izra?$y<>o83~zjX#bCTB)qL;KkMkGhZhZu?=T*UR&nrA<KcJdd
zOlLMcD9^3C1)y-S`f9oLXKXdndNR&eoHZM5d4tzcQknJ$l&d7a%PpxK*z!&b)<vP>
zf!C9Wjr;tXD7}Mnff1Jw=!i~+Xy>rM`*h}^T#XSif54>n$y5Q4{vYb}O8!+n2J!XO
z0PlRo2tQlJmBN?CV!Y%e)qb?9EmA;Veh4&;n2XHHuPJb-ct4^>Ucyl(GEMtt$G;%8
zrW}+XDqZ9On^4QC6^%ZVga1k}egY5!-In9CPP?}u+|XfU$=;d|nMG;9)neHHwx<}>
zlz}T!B3lEW;^WI7Jhb?<g}3N?Vo-68o`{6Fp;*1dc#Ziz<Wag0$cnWu`p_M335aEg
zC*gWbJhrJGHZjf;r$1|LI3@1`)AZBeoa_XZtzzf+72~V9GpI{3dS3nu)xEke4=eL4
zoqsvL3V(%hqMc70^LP=L=>S~nJCx0tM`LqH@2p^2#;K|XiBV=}-itKC1JoS>>~Ec{
z@upZb3|8%)Jct2-CnBdBpJ)7ZVD9M}(LH~5C?NHC%uVrMkX3ei^~M&k*afEZ9*cLH
z-TU95%jfZdk%chPNKaLhNk>fJ`K4^vDhWQf`DP=aKuLy*{d_rJq71j#hApWZtK-tB
zZ7yM9U_3$`8ltJ(!;p!EwSQi<N8famr1n;Xm5?rwZ$IA9UD;OBjJoX!$q{<|`v27-
z=`QB&0M9(vHwG4I-p|F-pj0|wA<86UqW%Rl9UwK*N6+iIKsp`w$om@i4H^JWIgn!Y
zYI?2o?&Z8HC54a7Vx_Tm49b#wTXZku9&vg}I}|d=?>@bhwB>+y9f`%8U98s(t30t%
z^jEfeb6whnCAj;z?yRnc?xBIcKp`i{f~CQkhKhTCF<xqicWVWpySGLMN*L@5wR<GK
zb1X$?WPSPMd=cMAtfP~hkeH?}(0$J=P|IednU0hjWuIQ=<An#(R)t!5VSN=R6TLfw
zO?ST9(r9GLH&}B~9dtiCV&}Gx<eOLwLN5W8YESN7%z3#v(re9{cR@e2IaK)t{W;^3
zHgzsQ&Y|y<$4K1;!<5WCEhWWTIzb75)WEcFz`4HF*vHF$)RNRVM~%6)f#i#xM%9}@
zcJCD{J-SO0BcBk2_D}}7ttq-MO0$6Lefj~OQ*>B9!s7noFkQZuG*?}19i{Qb#hm$q
zx>#D+^pS(Z_G!S2`|C@#fsodvk&hQOhtPX*T&b>N);{kqCJgu!QVG>t`(jToVwD+P
z2P^+Z)z(K2OwfHXiI}lqN%cE?a4|t?16a$NP;|jC;C^OZ_wJr6-Ak4&MEBl|<;{<D
zey&<9={i>RcjCTXluHe(S@QD@YhsX=<o=(X@^wRp=IRUoYlkmQzMy!kFIIBOk5+UV
zEw?=R<6@<wx@}<Ui63q2c*iCA0!#yaB!?hR(Ah)lEw+^dX(g@SuE1d>g`0%Rzl1=V
zH=CGw0MJ=ppnJqLqNkIkM0cOjE3euiduunTaO)eLB?Hv*PB$Hs7r<+eP6VJ?9YCvN
z&Y^=yE<pWO0^8d^(k1k(E$x21R*>oyAN%)z8dk0KQ(4kKGSQWn;+{MIkb|D;49|R-
z3sO1!DEr|^*w^3xG(qHy?>Sqv!2SnX+*bAfpuPEC`d{*-ft|HaR)+xR_^HBgx66Ls
z?z4SX0z)eg{PGXih_G_K^B;xbMSc2E%i{)Xvl<>lF?>)DO}T$zOq!5w|Hk*o$?8K`
z6Dw_R`G4o@O6(5U*Kq4@WSKu|iarsed%|2=I={h&Ob>6x?NZ-7uHI3S0#2LKg!a;>
zM(;HL=PC&ENK8TP#<_EV2twW6fWu|9>-5OPjS3NxN1xAQspp1S$-vB_%*UTz(78HV
z&#^rkE<D){TZ<ge|6kr(GrPUz`v5Na(_A)^#P}M8OjxT9jdLmO*^VmR#JA7f{PLJi
z`BF?xO4BM~;6Xy<VV1+&lRbN{YfFYD?BN<ICYJoV<H7;_<RIO6<d9S_mVv|G;%}m0
zR7YsiTPN1=vVg9p=3VoZifq-dow1(5C8I)&MnuMyg(?ho*j?IWC9D|nWY#Q}_3gM2
zFpfIpAn&M(7`~|6Vp1!aOeRIsiXb!Mz)T&Ou!1@W?~Y!QlXz`~SA^V9_%6iiu+<5g
z_WDN`w#+C9rr+;f3@Iw#^?COy9>zJOU@(MdsNw@W&1mLL${_7W_j`ZC>_)Tw-7H`)
z0^57ulDznEqZ)T_nm80VVZt%pikK-g;(1<D@e`dk#Bx+>DD@=o)}>IY{L50CZZ(5^
z@YWS?UI$YHd0}OMBtFGoffdIwQUPA|ieE-Z0y*xzX?W*q@U&lK=c(g)-lQQmbgHfN
znT{iZPcJBMVB-~Xw<nTW`x!tzP*bMU7RnT^uF>TAfoW=0@5?NaIO8Zq;j%%-JrcW(
z{E5hwV9!zYkmfqcwVYF{uK+203UNzfZM)0Q@7fC2FBJnY{>p7lFfD%D5*FjZzUR+o
z_l$jB58vmz^-!nx{Qk7Fz36Ev<y*0<v(6u4ik~{ca_g$oG3dJ5vvx-F455A2JL*FJ
zEI|o|<s>Gi9K&LxZEgK1H$FQ$@a`9VE#C+zS1vt@UHuV~rC3^$^DVDTC-K$06PT$+
z2yHF{L(E-f6S0DC>z+rWA8`LXjIY=WS4%^shm*>+IgBg8YdJM*tL0zkO(!~SyY}%q
zWs=AR$m%hr-{6;%s8O4aGcjP=vN{-U*hyfzHRF*Z!(P$_8-9e;&r%H+kz1=RmxPJ9
zc(tCy-ty;jxOUfg4lc|b00_o_%@E@H%|l~pY|!M-!=idF#A$`9cYBOhzsBMo4&rKa
zO9uCn`1XG^eZP^i<Q^|Qwi_)WM(6{h<Hfvym%-jSRq*llSD;~#o?S8fqAfW^gs86s
zF;8tO=HNnN^h_r38_CQ<TS8v5Z<*QsguMKMxr|`gFM<nD>1T^~R~Uq3K*1s^6Wr8u
zxj*2C*CK@vwF}v;Y;rYXZ(Auk=-Iu9I{fxl-SLl_9*kgv=hGQKfZuE#>~;%YeqWLl
z_jX|A7q{+9&3(i)?!MVK8=gdo3hNuB0_{wG*)jzr*Ok|9CNqCTS^uQJ>#h&#{l0rM
zC;MyV6_EY?qcSLMEyiA?tK;mT=5^=k+-H6sz<$!eZKZ;vlhp9lzR1O=b0xejL)~H@
z=T`1cw|zsk-aB7=o&1GaI=kgS%0!vH#j$EN^mv9>wyK}K2c|NTh?IZ3Y*E+B6s(Dz
zsj1PZeK@eRO=U8<r9Nq{G|j<WqHgZI@oxwOa>SUMt{v%(d7vzZDchS3={=O-_rGsV
zYX54yzkC~lde|J(mDdnqsGJlhqw396m>yP;9FMf80vf&y)<HD=uC!Yy=5@_Dd@r<l
z$S3zx(+qW6PbMghq0g@w!EZ_a$?UP(1M^6ZJ-h93|E!L+Kv5uSMZD;45<E6+zP~qI
zrMStnayEI8kaY>l9%sm&+FL6=JUU`_5JmG^u}pg-u>pn9!LBM`jpwLl!?d+T{s8yb
zQ<4Q@omoC_x82N}r+`u~IKWF;`aat&p(!o8dx}vH6Nh=ICqe~{{#iN4X%*w5SyHaV
z&4P1vMe<U}#6KY;Rhw0XKXOY@jyw~^n2=&Wt3Syen)jnQhq+tgi-sLy&|QPvwfAh<
zTMo!CBMyqevDWV47|fcfNb+D$d+Rp2iU=@|X+tz#TNW=KmRP!ParWDJ!iT(DD_s?6
zw3eFoO}O0NMUe_+QZz?lekDGc=Z7hUjw4lPwKccv5UU$`x0t`8%^IAoXrNU;8+{}9
zSBYbK2K@_#KlaNv3#ZG9F&Ny}HPD&3<?u7neatyO^8oH@YjQ|-Xg)kkfXlE#oNvJ{
zy@P4!x>2czjj{>P_;AD8yBu-pjFMqx;?E()1`C;T*oh=5^mtM<f??4BJ-B=Tebqw{
zakD9VSvOV*e$w?Bt%cO!yQTAqK2SY{uT^fax}UXRB`T~jSCgw(j;xV<+u#v+!NF~L
zE64b)zE)>zIE!-%Q$e(*Xd1zbgE7uV<qI=*$b(y{2z^}gf&a$9#GQ7tF8`}W*}Z?V
zlC)Y;u@b}Ow{~pzG8kb4hlwR$ZqrD7<rFNa#~b$AYwF-59SxFu;`fD7L1mvjrvS&A
za3;o>eKM>TcQ0f}sXTqutZfcO4LIomgyKbBj|LX)KG<Vr;JaHW?~&=yHdl3_V_j-C
zy%?EPkB8N>#&i!%w><bzQztfFLc+FZc-c~ASX!A7k&Y>lnj%B&l9vX%T^pw()5>5q
zRx_}A&LvG$nVP}qvZSdrEqyQc&Yf|#f!_Uosxj};%c_c8^FI%@T6Kks72fm`*qRrz
zpIiPCC7(}%;a9e-!k>ufRLTdmqwi~{&ZY*KK;crTbVm*s`EU#hL_=l(gE)Tgt;X0D
ziuu2i_>T?=mjwD46%RwbNS<AZGI;~8z)1P|g3?^f*a7lkrmtsex1`J*AD#uN^NAbq
zN@9fl)qU<NW914;U+IOxIzciAYPEUoGS-$A$Wk>Mov5gsJI*ZX*sh>E&2{&!CFZ`b
zy`C@>=HR)`2BzQjMux2$b2*@dH#E0YQkb%YwVg5C*0U8-RyPPSd(+<}wpF5E(9u0m
z8heh<65>!0*G@o<wKj%7*h#LfQf7dxJo^Gt!Hx($6?Qu8c_vNLbT)4)8srvA*<Akl
zB2mbd%h9-=o+aLt2T~tUORliHL+8cRy_vNM@QU=yicaD0yL$OWHHOiltQf_@Xi|JX
z=@Cj;QJoUFuq>7Mp5XW`lvnPH(U|@B9Hd<~rwRXD`Q8LWy3o$-;PU2*VRVp)yr^$n
zPm54h8dIvuYHFGx)88tY8Qs4*e0&<UF!XI{H8nhE`r#$3^!@jmB-zdd$)WApH>CF&
zX`-I^he3Spus2@P()J+j3i?-E*+gswdx4ST-{E}zY_6=UDz53=4kJCyQiYo16>7`I
zA%Q1=l`j+Ci1#neb;krE63X{B4DL!SCs)Vu=|+C7CNNif(a#7iOM`M=uPBRMYjfP)
zk8Fee<NX~iE*>>$pZSV&T*PQ!XJP~&YpDTs-je`lJ)AR3xzg1uuu#k-=tWKHZ@%ss
zF3rwa>6Tz#R8j%?q!Q6BO&nHm6uL)tYpvGi#Rao5W`35&Tdmaxyo^+AI@BQDYjUcB
z9dJea$Z%84V{uD(y|_;-B9&b@L1VPES>DM8<!%lJ0($p#C4${NFg{V{rAX_nC#*p#
zS`B&vme?iF4|^F6FQDX3o}NBF<rwoy=9mTU^ILBj+;pbOH~pM!nco5D$=80jkXoud
zSXN%p_`(FDkC@YeT~6UwC1u9&Rt_>3rIgHVf^jXAfpZGd6%|hUbV#OFMgJzE#ov0@
zLN@mUNP;fm*TLPG3t)ItI4DIus6yM48!6Tv`}!;O-UFGQ(X>^eR(9yibQ6;7`Z~G7
zg(JWpyFHq5vTsIYaL(7lJ1}dXIXn5m*u(WRXu4s6oStf-u}~Q_P1&jNyPq_dsEJkJ
z;+?5%+@M{D>U5UZxk(?KxS&tyR)AeYUcJ5eYM^IAV4CITEGOiKM2caSPNE-;o+e_v
zWrCE(Tc_N+gM5EocX=R=?Z)>(X>G5O+*(rMp2||&o_i12*IF#Xefk&Y<@!3|)zoP!
zHe)a8=c4)d#g*&%tr)B1=Q967_S6vC^z2Y<fr!TAaMg;pK<qy)7iX^2M3ea>%e7s6
zvtJ@hTaS#_z{P_KfWpN0aE1ZkiF3GzacWY$Zqn2Ux(wsS_~eEpuW|_A6JTEKD|5lH
z<GyI7m1wK#!XQYHLL?qm!}VL(lU`35W`PDuD`#C4*|F10J#ou<70R6doF9a~Bg7!i
z8S=W&%6BLqN7hcN;0CBDg|~Y2)h;`)b<U=K8#WN1xL;uWboPPBdx0b;hIY|4sCV4~
z`hZjA^3poNS>2DP3vPop9xE_`neeoG(K{GJLfN^;wG#BJ6Zp8|ig1=EiN<qx>gIC)
zJn33+7GLS_I_vwA@|EbVU|r5i|F5FVuPx;wv&D}=uTH<4UWsjmLWP|L!$_en-^%q3
zRwlSb`ULCAyo>Zo$E{j8L1vU8KK(&W8C6_%NWP_+{5HcnMX+-v*u+2FG{VqcYEUo_
zjJTHseQ0J#1PdOd(=@uDZy8Na@h&}140d_%tx_Bv^LqZY>F$`FUljiF+V&l0Nw32B
zSFkd&;FV2=&}&Y~m!9F#>nl*h^8y{|7(Ea)%aeqHD6Y%6b*<a~g2){T!Qq$aT(0}z
zIIQmmI46Rl)CZHxIViegYF129l!9TQ*hi6lpAVX0{|2Lx=$oeO`PGMoQ0Hrt2T@Ub
zGZx26VGf1_8DVa>-sKOmmRD?T146)n*<bf-sZKaN)B=L6L_5n>F}?oE8?9~hJ(o9M
zNQ(`vZmX&ahpWB__xdGG?G^Z1T0s`B&H(@Ba8~Dr8?zMM%PJT_MqS6VdcMdT<Yp^C
z%IgJ|q)P5xcV^^Pxqn!};o#o^nVY%)c}!dua~yd7;_#}4qEI&;E+8fMJkgvD+RegZ
zZPp(6_oqB>+9vuJqh6Yuw&3^UV0{#S#>Kp3ZBL@F=F0%iS?$DHTMg-Z3S?|b;CYoK
zgzegx3eMy5zSWG?{jrp&KLJ)&TsC!f`iW7vU|~*pU%^mLmyYBT_`w1W#-C3w+;3zI
zRd9^EneQ&aDf5t<)<uZ2e#OQq+Bp?E$f9h=8n#b-z{=lZ$CTRQz|8UhAGp}4f1t+<
zCR%g~a>o`UV(ezGCnYu;!=CVy?7+UrBl7GF;ZuKf+6ibCP$(_xUh*dB?u^=<OHWg0
zz91|PXMMQ=GQp9|crfpZ+nTn(E8}FFIL-pThSP8{rsT9+={jOc{e>1d9Sk|*eStZx
zLD2Tdc}rr-!hNu1$9oTFzSiS$1nC}1P+V*F!vx9}mduN8#AV&v-`}}Nb?SV^y*P=Z
z?h6a^`(e<n9}o|d*AgOTJ1S?J2i8vGBy{YmT<{@Uvsrz-zuYH?ge=bg=y-0Cwf9%&
z+Tn@6_5M(Ms0j}wFId<p?2h006NWVbCiir6%R7w6dj>`Zf}KI5&KikE-xE~w85rb<
zF%BFwXBWbS7COBQDg4Yv?ejs$Dv10k`z>c&MD1=<_lQpNH^K51^!r;5E4?DTv{4|`
z>iX)xUXp)0srfHcW{y36tkX7Oz;hi#gI6k|+Tk1F47s8rw4~Ebrk<8ai5!YeUn5p0
zXtzpu4TmFRASg9{KIh2SNW2k>L$}LwbmJaEOK##QU#(7EMaUb&@9*t4mf`thAVQ?$
zU9(g~KjVLNGBB%m^aK0a$K%DlaYsJpTE2v9-0K_LwL-=U2LXVQRgAL>A8%D&byA+X
z)kl!ta_t1#8Mi6W)C_mW_0Lc8Rg=t{bf}khe1fb%_TSl@2RBPE01<zPXzuC!GSd(W
z%hFfW;~^^>$z`~|6QeNia8}f`OC{Wma1SU;N+cx2_DPhM61^G%fT5Ezu1Joi=t4QI
zFtgtur3m}6PSX=2;HUBUn~Z{CKTMUj{$YvkbKhr#I)Yx2B9adtz9+;AiE-OH0IEMz
zxY;aGyq4I;;2|;MRCw*YUS_~ExA}ku_)DHPS=NH|>wV#y+ykYJ`hG`OeXzqcwwu3?
zi`ud`vuVgWJt0vPLm2Qv#Y)~u)AC-JBf^7+RR8F#`FI217iU7*v-N&I<Sodl_ihPQ
z5`hYxL=W!v{53V#W~l}7wWEx_|Is!LuVNnEP%xCPyN9e<r8!{hYgP~8Sz@<v8Cso<
zU#AmAJK-1miOMo>3MA$h@0Bs-GDW|Yj($xqO<!-0sT=fTmI13Bd$Fsr9iQUquW+jt
zIXA_<Ha0{(N$>~+fOR0q^BT84qThV#GwZ&Ra~frpEx9k#<>$lC^C~PAE@6Z0Ykjb{
zygG2GO}NuT8*JafkS}B(#4;^xy(6KAiEd>+7+%=omo#A49IAa0?M*=uv-;+pH6i&p
z*InI4nA(Q8Lrf48S6@0karo*xUQ944V6x_OIiPr^jYx^U4e+=>{h|gPJ(wfw-5!C@
zCQp{!S3d^TOXy}ye7^Ro<a-XhM0Yh3v98)MKj}4^?$I$YRul~`FOK7^lCw>s3>`o^
zoaQ%!%Tp?OIAQZ~C)e&tOvY^6D7~(5x@Ge<EWWhJb-B2>7aq-5?fNj*G*-{yILE=*
zMg|SSai*}Sgd3=PS%jf_f*!q-A5Un`3OtD+Ep9l(sFwe>n^k6*RffBgZ*O^ZKdQ3|
zeFI)J@ScYp0KzSx<zv}Rp8%ru=9XcZj|W59jl<)cBY34&OyD^b7NXPnWiUBl_{2Zj
zsg4zzf5b*qS1tX?+szN#T)g3S)yme5X}$m=`+e*yTLNHgO@7=x3-q>a>iOLq?Z9wk
zzD^*&zQi;@Xv(gmhQh2+?KDw;uE0X>Y$Eqv^oUH1>enq%A@(XxmjHWl$>G5ItDV)e
z;DvuVy>7!a8v!Hlcv5Zy6(~hMW2<WT^0Fwg31(3zRwFr@RPMg9P>>*1HKKr=T=wjm
z(6D0%1f)_I_NOFLd~zVIV6c+%SF7_I%UMke*-CZ`N53cYrHxztRw#2y8h4$S_L))c
z6Ygkm#5lq+4p_3sKlv3C?C9@iSXVSSs@ZP@dsf=yAb<&r>PRg3vMV^aIJDw^zuH40
z-la&d>AP+BuGJs?OS4MsqY2_k%k=k54>OC3!U(vm-=HicFhjvGpsstsHh-pD-pYu*
zH6`L(g1PxscKBh8!L%csQ2ihWjSOz8<@BD+wmf_rozzWEZPfc=WNKPfr>+?8#r{>>
zdrLvAHC&CE6jh?lF3U1@YmZwJ;Q#nPI+&#$SC)cd!Hg)U396v8Ju@AtTtrI#_P7ZZ
zF6m{<9EKeUy>XV>GN>WUSDI1JaA(Yb{$<wp!a+f^BovdM`~CFd!aNqMhni)%DxV`_
zBBIoWJjg$zhe2w+Bx|5JMHTlT@T#NAn#{Irq?&W>XA7UO8a6&Pw4;1=$)Lw*<ETd}
zx^@i;u1Zvt6^47i%o>(>sQw?_E%mCNbw32(sZ0iA8^e!7=1(=>a$${(rjQ>a#WErz
zWFOC@cd~YW3d_Uql<+RcD5A*mMOzte6)Fhrk=iaJ#NwKhc=oDG)Kkv|>E?g#dpB+f
z=Lze~Ww-e(xV>xhe=zdr69((a8$h#>_MQnHqj-ki=jxh?U7hh&6bl#N9(dQ=B|84B
z<6BiedkcbsIpC&pW5f+vCR^cTnhr=ne(Y*{S0;1RoO%&R&|(2+LmLvjC(7O!VY}w5
zkPSG@4tiS=rtR|MmGP5;2}$F%+Ryqi0gUr<R8q5MHA|=lwvX7^hqWja((sZ`Vcs$R
zC%J<i{&dx$Xz-m^(weEXm!YB0qsXtjCm-FA+PV618R0<EgL}pdDH)D0LdEIz82gi$
zo-3@>U<QZ5fve$x4Xgl;w-RTmicSBX-tH|clL(hzh<5F+T&G5)u>k7b)$qf%_$j74
zZSpt5RASV8Z~9)<l=|7JZfi}-y8};0=ALqjM`66!A9iDv{yklhU@{hqRN!Iqy=U~W
z8Hs6Eu}mM~F_m)Xi~Lf#6YLhk8yYePTl}G`E9#jPEzZC}FW*#uqQGxIr;|e!=69%E
zRWnseex}k>HS|4V2W)YLM3c4OM;j7W)|-Oi{E;)FgM3o+FApG=b~CN<<QTxf(y(pb
zk1Fnw5o%6=+1W$m$`p~S_xHHa7N-$Lue8bvg!|IXN<YUOc#A_wBGWaDUL|2g#dtA5
zbpAtep_q45LPuHd=-&nT1`}}04kPEUztT7iZR&f}8!Oy}xoJ%w6h_&M$8|Dmu+NjV
zD<HBp!&c<o&Fye2k)Vnw6G|=2MpFDv8q>o<{cl~VXbDpM;K+`pscrVPei@6ThS4AX
ze9JfQ8c)}YJg`W#isfHk4iI^bE~~2k%NTYBMun`|8%l$v<KDI)#Euh1{f_Z#0$++M
z3_yj$i}suR_5a}YVzl}Z;*XYLlj76^X(JKm$l~MCQdO{{EsT17OW{sg-yQ7g>t>P9
zoINng-2s_6a5&l}ZqrYy;C>*lc|UY0#f_7<me?sJQq&G9W4K)rEFyXM;ciD;!O(IX
z<)XX3f|pghw7ZAtyJ=N}5qh2vI-|86IMRkNUr=JWkUJP7H7>N`dCK3Kl>8C@ZfvcH
zFt#teQeSeW=PoSlV(GUa>v(=)%2|Z)aX2tHjP$APfjZ2*;#Wu%xY2c7+b1L3s3>2o
za(q~4S~G^M9(EjW9VjxyGoH*!+F^h9P~q%pW_tFUZV{IAdDB+ohM7(pK{qe&iiv9Y
z6=DK6LcktB_%1<m<UyhCq229a<J+BGgKnH|S=8<O4nDe~wG$nqTdA>#M;~VNpdj6G
z0=+8-f(t7Cfyw605Y=Ud-`dS2Z&vpoR`lHu;R)e9lx}>BJy?UO#jE2R){qS-6Otz~
zRe4l86=O-`8bxomJqiNk>;EllWB(XYktoXdY=g=@oIkN#1E;oxwDF}Ma8s~^^g&g*
zrxWR}Rqj=195l_*waQr}jxUeu$^xy>u95!6|8Q8k|KV!tkRx}A<~&$c!Z5%j+uQ^;
z<ktr~dJ53-D|PFMy&G<19jX3#cX2FATS7yZ>V#%+q0IpRuO~}mmg$aF+F80NGi0<;
zMpE3z!w57(9DlCFEm>7ipJ9og$k08zOJruDL#fUiorm2X)16)IY5tHvAC*rx`YoeJ
zQ^$Tn(Gy839F=Q7pMg!F=#-EWaIwspa+N%kJvMavfsOa$VUaB;e|=Yef3{pc(YPQ_
zNRk1d{)%}sY7kPm@vITP#;-lQ_=ZKRvq!fX_34?URVdqE*VVi{Zz`{E%3()zXi-@0
z-~Z@na2Y3@A6#XON8{e8(qWZA?yJGnSnlsI8`Wa9)X!^Hy_f4`g<BCr#zq3mjudaw
z-OZX0W6dFQvr+r%x43T4EIhK5J?X1>k?1?X^Ae^SR-z3%dOZ#o9e$^Utl-bmadh#2
zb!e$Ueb)znX!_|FR)tN0dFoV6GryKyFQ49b$KYN1Zpea+2n+8u)vxn<-aJ+Kxgh;J
z``?=78d<vDwEaa4YUi!WwkBFs;@^yRBbZk`%oZg0G6%tkL}@Ui$x5KHd`%@0S|9v=
z=S$V<n@XX>cr$;u{_%>*MBWt#^q_+<NbojaHSsM|9$yOwQfTIRNP;B5lk0g7fbCsv
zlHhN%OpRM$o4-c&2C1iG_xb%x2~DXk2o6Z4hPM+4<Wc1V)Ooinh0hzmsn=YMzu9Nw
zXM5|#vFI*~5qE1C38fw_y57Yc&y}Ne++$K@nN;f}|ExS$z6SjJG;37OCG3!7XMXUV
zeZF_KmH|du)%?QcG)j#?1Nm>pkp;qqEkG`8&T%R0+guC#=zf{_va2ZS&@um=;z9I1
zCQu8isPj)S)8#(c6-w^(n6Oe;E7UaTJBadYkF`QM@#$@je%S5MFgKV)nt7<S;?h4!
zmx$57CuTin4&P<P+1Id8m>#|?H2~07$J5pUX3xB51IP}0?ZCty2~Kb{u^CFxi{HLK
zd#g`HSiq)w`)-w#flN%Ckl~L$&-T}jqd|mFel`KCd;=r6jkt}(t=(%){-f>YmMwn#
zwcK-<A&9jkUJ7%@%BWN89B!D1)~Pa#i56kxmM#I+@?c?VAW~paij#8n*n79xj>x(o
zZJ)6N57cL1OQU7%h81lux;448{mj9oZ;KUQc6GH;8%TsNLZ*jj1pt*y*M~^$6-Ua3
ztRZ{2lcSQoT|F}BQL6P<_H?ze6kLh2?o%5~OD=JkX$Y{SM;OpARIL%(ZyA5xA|$j0
zc6A#AX+ON;#^Vg~aIizOtCwOX4LBAu;i;)zVX%az-X7gzJfJ3|bFb&w?ue`US!kGd
zzJ&XKbZKl=r3Gi5H8v7|XfzW?X`*J~ZZx|VgI2K|T+3*v9xOBe+kZ%guVyZ?RYJlJ
z;jl1bsYrsF{<cVxoyru1zE_Bpmd~e0N3n3f85BJH0R1%6r<y+^bO)EhlsEK)0CeJt
zv9zs?2uS*tuhjozzLelh32V?Pli1iJlr+A3Z04I1b)!}HaxUDvs|Dg%`l*T_P22}Q
zxxfD)yB%lK>J3(hY3megI}r2o!$pU3wWSm;g6u|pc`M!8igC#K`L;<b1?`ML){Z6h
zKBbZerj->eWiw;M5Z1|^rwW&9g^fuu)${;hPlK2c^VQr6r!h}hIanZk%m!|1Z0Pqg
zUg)Q~Wmx}^vb_FE!O6}})y5`Kafcp%c=~>`fy(}E%{SN+iQJ5KRk2g3>RkT4%93Vp
zzPDm4baTt>n>C)TbsSMM6aUcIYE{YZee{h&6G%5EDMU*D<ZFpucyrvHzorbX|62Zt
z+#YDwm8n{m@rty4y}ZAMb?WDqzM7&=W6I?y%DLtK`u-Tt{*ohxjU!eO;X8yKLTj`w
zS}mnr?f!Ofee;Q`yK#Sxdd%tw(=MLvZ2^jJ@Nt?>EV`1<0Pz2CcNShvH}JpbF;FR`
z1q76oQBtE(VKf8j90<}dy30qA?i?c}25c~p*nm+=NQ{tXbayv`K6k%+?z!jw4R?RR
zcFy<gZ12ze{d&FRP&8Ubf>OCswm?df=H^~^bv{yzP=<bJ($FnwNXgSQ<+25-s9Q4F
z*uV<U-QCKpUQpRiXPTc{w&g9&rjCosDol#5`OUZ4t;v91r;T!B7v=t{PgO1Xew2ZC
zcD5wEzj-<mlW8ez1>B1(g*nmllD_(ljA2xwC#yG$Z#=ad;Um@K7%>A&OqNVAXjsQQ
zR<^TM`K<+N!!O*()>@n_#JQJ?6laFE6O;!ToL_z5^P6O__kQfD`Yg9;!uF5RQBmk8
zb(juj-#m63Zm5rO$WG&+9W>Y9u54wNEeF4ub#mccP^$=??~SRGtJ8iosggR&*czNB
zSGV|WsJL=xnVJDtaQD~w_*Fh)&tc&Mi)7USi?pY?ZCrCgY~F%tfH3({yKhwpyLc1o
zyIjgr-dGkz&!WNX11dq484xOdaiYu+HF`Ow)gU-yMyFR{ObHq1%MJzdF;C#uq9(gk
z*h5>2tn6nI?H*N{b{Ji1<If&Ca5NFLMvfOuvFF$fPo10A{>}~766-N`r2Nqk!|w|&
z99Qv6$9W~#jnHu3w6`^|Zlx|)##w?vkZXVMZLy{}A~2VzlH;3V`;#Nu4UeZAQg~n@
z9sQjJ%e@^bm7oPyi7>L3zgIqt;d=*2cOI12&>!z#rzshT^|j2s>`W4jdPGb~#ZQi6
zptD{CpWo26cE}?4bPmveUKaiH0+Z%8ptbz?i*WMp?}9u4`o^Ggse~eJL2vJd;EzA+
zv9la5s_@h0d+`4I7wr0^bH)s!KsSaaGIA~%pQP%gzpt`w1vaQHy;=Nu(PrLcu3!r?
zRF%DkwkPbxgj`>lJLlc64-sz3Fmh)JW1WRrbWK0i@tFIyf7M8^Ky;=!*RYvKj#kic
zi2PHPDlW>ICM3Wd4~Ht*gTB02h)|bz=2?gQ37}?Jg!~P;%wb=3^Et`M+ymCv1V1~*
zKmLZNsy+ij;64@^Zjtl@@Usg*4GiQaCW<h4hYKkmV^|B=g=X4{mSqxTJ{w6GSIG)a
zTHq_2yiTuf1w$*~C3%SY2JUxqd;;c!m6cu73G0~+u!oJIYCj)XpfV5~t&9{$ZL;Y<
zApp+-Nqyt56HeQDg$RUN{6XBjIMSj#Qzz+O+x{OfQRse@tP#-T1GoojCiNj#js|y7
z(NV~d<ofIV96mw6a>D*)3PS;vgd&}iMW2u(p5o<bL!O?w%E@@DT!nX^VCW?wwJddi
zwQE#b-)$Hdf@Qr@)N1rtu`3%DEk~xJN4zZo2|8uS{t8<w?LR%-FNHv61s@woa8>>j
zZig6bt{-SN`>1W`e056nVEMcNfBNXRKPL>oRY(uL6>Ol_$u8fz`D3MKep7XNLpjSt
z9An3-zp^KK&C_&DBTg1#{u=iQG+p^86ySs1aDQrVHa--)vV8zZg9(V7T&J&^Qj8LJ
zAMK25+Gx7b>X}q8B>k*^-WS>GtJSMk*)AiRcJyn{pfTx7g0eHjZ(;I^IL{PXKM-xI
z`j3Qd^<mh<NWJ|EMvhArH<$XXv14=5fD~WB-TY$K$Httd7YOwkIT@e8p|6W>tlCB6
z{R5w!Z1vdFojs;-dHff=AMxn_NF-hm@nGbO{00JYh3hcK3&uBKWc<_6Sn*)Mb`z3<
zhh^eRk0+yGf<96$9^81Y4c9FLN6$5+mq8+$n4E#XK>Q3_4TMzRDfEx+HadHuil<l;
z*WT_2>4yn6@{8_}TAcnwkG=Xg!9Fny)3V5ub-^hq_Pq%3?FGu6*wnn}Ws)7_{gRL>
zpD$(o`3g)X?=AXt9^4@(b=mXY+E*k09h&yCKvhDu$*n9WPbWc7-MN9UCOh?p8>#LI
zlW;BCh|Q^+?+n^ouI($J(>fqVXZfcs<}ZPhcJ409G}}>;>HP=HHWP7dayNC4+@;^y
zPG*&8nAUf@Md?OnewcCsoq*v&wViIA{zPU>q&Ax&3XV8g47qUVn4iB*ZdfhoOdOIs
z+zNGZOdQjaAPnS68{T|w`qZ6-QbcW)R@^%}DGnCXfILrjmo)RcE1b?~#eJv(?J7g4
zpg0wj%?lLxLneeXHR3uGPB^MRd(7@O1Z9;}UKw4#sOQovUSY62jfsk>)K&(<d5$}z
zjm`qoEIcW5sPilG<R7EkQuqTY__~4N63BZNDj7m?A=NHd4t9k3kXjC6i#w(+BM(x#
z2n<CAj*g!CG-GFrEY8L?wAa48sn+K-d{L_Sa6DZu>m)}L$tkM&rDzKLfc_Ncc{Z2N
zy2(9)Do6ehozI-R+-UcvW>LFXvu&_PbZU08oPZ}{r;jb-=TMg9%kt9nH^@?}sjexA
zCo8|W4x7Rgf@h2v+l^Oz`t6HS#6>B@y@B7{9T{P`o4+Vew0IG#j7M3uFu1EXwst?h
zQ8qqMuB+k%6yQe}0xb*3qgjJXceueA=Owqt{rQp%!JxSZTp$a{Me&-|hk9{(7-+-j
zEVerHKN9+fEx_NX^KAFlhG_!b5~k{JiP{Y`z(Z&uq*-Te<(0cxUxPR}w!r<K_TGWZ
zcD~iaF`jFspL~H{ox%LHV%|vxG2{_e-@b8V8D={AlB)d<(&`@%z$k`Q(_Lvr+@bg!
z(=FTvqbXqo0JjV1S1|#{N6Yoyr`J$?<VSG@<8Vvbd{cJ|@V8e<EGKn20`&`7Fi%~m
zreXY>&L(%g^+}hy>`(WF4A>(S%{cis^~!94UExsX0+m7uNr<<$w3YvWmF)TajY+Ri
zkYA<iIX)!hJ-$?8!>yxp&X$ObM1m7cm7#{BkRGUER$+ZZ;V420IujRx>vT@m5M0a2
z`4SXv)-VHGfQQY=ys0=LT%I3GV2-T)a%?5}njW+q^@y5gcjj`p156{^s-l72<9|$1
zY2##hqC~$P{F#gemP|bAgF@zc01QQ^k`9wUo~=V@VK1qP76H$7ygZZC;O%TCshRiP
zsA;@zJzvT$qo4x50AR7BEUMoo%w}6&Ti6x+)lVg!8(L@!Gl-nPAxaWlT&GdZ0abWc
z!K=mL{17eORjV(-X<APj-TNus?|LH*5+WzE=h^mfHtVBXWizkH7KjDnVngMyvR|Es
z^uns1KLKv&=tEU;g$!Cv_0PPE1#CBaZK0P<6ewGE*=gl^NuheJ{-t>SQY!YT1D4o@
zQcS=Yq10%K*h&-+Ol*AwTYFJz;iVD8K4DPI1uuBlUFWrt*{0TXJUbj2tU2ti`NO^O
z`d}>KYdsj=<eLPKYkyRAYxHgVa=y%Ip^`9G__o~4z@^k_%nX;uSE(@GSOXxBq4vNR
z=_hKspWBbB6meMINFKLN2xeE~`IGmCOzLPO0^7f9i43?=8p+RgwiseLEYcqdaZ`e0
zwFP@8n3auW=-TNkMx1rmM|X4$97Z(aH@1;vis|MYt_J7Be!;Nj$Zk!>s+P+<j~UH+
zoqsND>`t`hKX{Js><?AzL-~2SuoF)Ynh*9+K5+>uMHpVAStV-I6a|F*<mA9pk$GQr
zy+Q00HGN-?o6FFm@(MMyuJZD%z^eV17^O{3Cb@(7RZHeKmHu*T?)JEb5SN2uX4$Nd
zYG`v$A=T|EiVh_(pjazOkeaKu6T)Ao*==H{!zY-Fb&T0Owrp@1z~P(uQRjW$GdaP7
zuI}yR-hksO)SfFvzokRnKamWTO<ER>7?;@|%?77m%QE0TRFU*3U?V$FqE)aAMQe@p
z#7Q(p`VXY(!WM7rei#=_bCgaD5^aLl=SvRW4o0H_Rr_K;rjFAD{l#Mj-<8#X+A90c
zZ1vq`soFWD`|R^P^l4UOczDI5@(~8B=yxcs+8wwa#BRg9AAy?>rib`lqMPh?Lv!4m
zma)K``$Xp0s(l?T!5*ZMo|~RqYFj@$Aosb)W%;wyWS&NjC;7aFK?XuyR9<TP>)Nof
zAXd|TF3)_!8n{_IE4F!Yode%YddcK8uwg1vLE-tcv&O2<1$dw6i-|1*6|2A;dOs<b
zYeo2txoE_XoyB11N|`EZq?;%a8A_Z!gXk4`DDjYvy9zgMzmvJsyuR))Gij#+>SGW@
zCO!RVRL`(mVZrY27ROM-cK8xpZBUPVg09dGDZzp?3?doochFkMkKl{xuS34F(0tuu
zhj1w--Y%q=M`~SKyn+hXVgNeCVB6#91npL%^R3K?EPuz}p+>ii|H)Pjr+$o>z2PD#
z{jIe&g0g~aVzxb^;g9T<xpm26s%_p}%hYP^PtIePPMW@ily?=aQ3S^UHeQp;2|#+B
zH+7yH+|FFm2c<ZG$apbR!qipbqkf#~h-l_(JP==t-a9-<)4O4G+b^&<Z1(#Ra#QVQ
zA@6?(46CnWiomqgm|oZ8P)DyeA01dQ)E`eGOq<JU$XLr9AV4VaOfv(C;vqiTSM&_4
zE#T)<06m&6Ba6TbGfP8MuF(x+;?UoZED3gV&2{+pk|mcy%OkYclzDTNnVW@ZV2;3`
z2}6Pki1~uo`VX8NN8x7;&xH>v+ozcN6rH{Fvb7z|U<O6o&i32or^m>QZ?qW^MmJX%
zsiFD*kyQOX1>8siJJuSU=|!6v|8YZQu+R!VEQbfaKmT)|j^a&`qGYV-&dB!0&P?S9
zwfF?(O#FQp=C|BUH8v~*J}J7AfGW*Fn%P!AB5raNFf&^fvt1{9v}`v}lQoDA^z7gG
z)pCb$8T$>z@Mk0`%UY6#PpZtILXS@>)HRiB!4oOYAE~UYg^t<<)S$MVzaq7|<8Q{d
z4!E5&P<9zCp|c@IeMgYyx<lS$0FUv>YNsH<e9DhcOmo@Q)|jNmXuY#iqGLdBz>ajV
zkp){Npj%_B?|Aj`poqRno{kKyL05#XqgVH(UR;Qd{Y>y!DIv^xnoYD#EH6TLS5Qq=
zVYY3vzLRG+OiR7?!Ct%d{zPP91Rm3|t~Xn*9+u)Udk&n9@gXp3afz=JZxwvy=J|ZO
z@v{T!c3nu_&9^DcwstJ;j;02+*8zaizisxIYTsk*9@@?66w<{0d`24M!kZ(sHzVl%
z>e7NBnXwe}z<L>+M@e7#rdfvmj1114uKo3^Zi>JNwXajD7D_RF+z8P)*5^j>|Br<7
zYLTmLpus=12(?&1<&Rr$fHoYQbcN9c7&E<TJeK)YBXTK3i0hxIiKZn)*sr*~G)IBa
zx=DN&#i|uU8r$h`k(XP6UFkFN+y=~FBDmeYjs|_k=!*|mF)ezP`B^jC@12@g5eb^@
zT{9Y4=YR&CPbS3!NqVJ*HY(-%WgWgv)I3UA{kA_dm@X330;8tQSfO2Paf&hO@)Tb{
zmiu{A%6NsUmcfcjI5CzUeoe&^ILmqO>Dy)MHLAaz10y!jLxaX6zxn?8RgS(!c&om)
zUrn&lsDA7-ax$V&OrvrfK=sC#;s+O_qO)eIt@vaGvjJAK)eE$gZ&fjLBbha*3rRJb
zTFtk()fJWDcMGrxG{cFQlX=YKrQddqZ8tWBnO)dn0xdQ*8n%bKa0gNh>k6w;WQMQi
zkz+Gq!8&RnCZU5ue%As?@^|B#+n3B2n5kx2htpw&bFpstpK{ZOeG~J_ofdd|m97q!
zgDuPquKZ#ie}F6NEgU{B|FuId0~PfanF+mMg%*YWoRc5iAAO5ZWsXf;T;9wdZemnm
zX)<q;J9MpUO#v&*F$<Plzv%Bq<j;hC@aBxV(if)jHZ^H=jh-@yE!Lj%n?q?i?oV0=
zAVkgNz&&Pqk8}hZBsi~FKcwhwDS19?*(dEIg;q6X8d=nFN18PIzbO%F6`ikluH`Z;
z9?1@+&MU8a)~L(inH~Cb)>ln}MP-H3iK{&Rh=?2wYYQ3V%mBI%AACp?ZF>UL7sba&
zB}H5cJ=o;gL@h(iD%6|ZSj>g}L2-OZo(byZ2HTgQFGUnH{9GAQh6XvFbR#|`%aTWt
zX*h!Sm3>S@!qhUOd$m=qL+>H!P(o1vZT%D}|MW6!OR|JqG(2FqHV=HPX_I0YhSZMl
ziV<p{j-k=Z`l&<n1S8vm;2MrBPFSS3Xmr=Ug*hJsgzpUtYqPW0ESB1w#jRc-6?p8n
zR`*XH<gbI2ao)Sm?h+waLikNzZJ<`096IGoLXmk$tzF$BHRqB+HuN}4pcD6n@0R^X
zkh2RWR<`<iW|!>eV8H>m9KDbPGO$Rw*G3^GwLppY=)TH<AtdIs*5StP16iy4@qGIR
z9JRRFboVHx7M9oI7MH0xMdhgkF#S|<OYZ)-29<Ac(unYJ4y5a)mn<A39@dk8@LeVI
ze0TXd!wIz1Jrc#}$DVJs`2&1?hhKA9jJavR1y95p!WI}T8BG0EIISk8875$Mzjh>9
z2d$8}NTP;gym2=T=n69k;-(Z2AK^ZlGey*MYnthKQXBh0=C0l6Gs8+V2yX4%zf1A<
zvHAAj&fUqX1sN&VO$yg{d7E$pp#>^PjWYRYsn8D^5H{5qTxo6wm4LIdb+@YJq1vzx
zH!QP186lpqPm0m+_A)SRTiCx94xPQQdxl)Da~K;=qUD%<ebN11=vVX1X2*@vqtk!F
zvtZ7;`HcXE1ySo(%jq~C80&c5i+B7<Ka(g4o^!fB$@5OI^rSBfmozUo_G_(VtLe^%
z6Y-b1-{yVmGH9d1)|VSL8~cw><z-b|v2ClG4b`aR7bJI1bo5R<?!qg44tmS~uQ`N-
zgueJwLGtYqnb!Hrd-CWhBJ<Pa&d~C8b13H3z3ikqrTyI%Wqe)K`#W@Ifilhi-dVCM
zzq1Tyaw#nd86GB%v=Z4LChcqLSXF1;$>3F-DquH0`InjLkeVSEBlL`@hydjmUzr8g
zPbC#ty(3XY!Ys3QLYhwYuHVpnye}uy=5e3<DER!;sZ1f{&znbo%O{Wwt&kRZ3~|pj
zeokUoTJrhi_HW+n^nad7F|<4`a6{~#1iqm!=^b~N3~9wRk=zUZX*bZy9MTfnNb<am
z`jC1nG~1=@<X`fWz{gg}|B#0G8-(TtOZhP>e7|qzwq3|n;I11S{FEeZ$=<?nYX6jk
zH^!;ZW5d>=hv;J*RH?EcnnE6bAki+U6&oLhQMFaaeIt55aJv&dmP_|#ORb!Pb}X8c
zCXPu+Ei^5ex6Hhk@sKz4%+Ya*oA}vxclC{v@k$2@?YcW83`ilf;~kQR8P#GO3UW6A
z+eAR8F|815kh6VAn3dF>We=ec%&F7FRPR6Pj`ktMIUC1{#djK8+n{rrQ2F^bk9+K_
z5=6!byM3$Wj0@rWKcse5U(E7za62JVw^2Rs?kbJ9Bh}fdT4h>WNgf)Dimf=5{TQpL
zvodzPqY0+ylP|HWHnwVbb&oBg(CK?`QuU_-E27VX?0*y?v>f}d?yC!b37sKIP81$|
zum5%1l9Y9)gsn8_x@EIx>g%1QJK^JT+GF#=FJO<$hz628-<`22j*HXtfB8hmENiF0
z!QY>dpm`_CD5Y-8&Pe#Vsdwim<}By_r8*IwK3eVUE9)mI))$g{c>e3kdw*(+<dKE1
zLJ(8n`S`!|Y^BO&p70%uKaKygl&V%;vdt?b-)~!Jll#;B=H96=9|;#8P85#W{_p7?
z=Xmt@|GnIMMcnOa{J(O{i_2>s<KG&g4qwMU*8SD{ZL0l$Od(lBQ^?pSc1PHt{=d)P
z$y;v`*JdsG7U!#L7pIAgP=dQ<%RjD7TOsjh^7587z0E>7`r_@?U-=ZCk*3N0&{LRh
z`HzJ9U*Z25sr`TXqLTjG*uEF`{9ZvCH~^LmI&7`AJW_YMl)2ooAwn3l=PLg#ar~c8
zZlsbrmK2uldmU=#_0+&?QOLOM;Lv>kK5y5+jTTfCq)o!m;<G0ol4m1W+fLUIw)EyJ
ziIgGdUz?P|u<y<Pk=Pfg^_OiAhFN`+iVsD8G(U;?II%(uJH8Z<7x5X}cn&gF)AnJ*
zYB?{?;)U=Y{yi5jd~8V`bkwwyd~;#m`Wa9WiB0&u&QPEyBReI?CZYV>x|zvM%A>$n
zUckmY0gFj<(J&}^K5mwLLPq+ci%vQdO@+}bZj*DAuiHM=<({Wzz9{0%i2Ge@2r?X`
zHl)_u3A-Om_0Hww0P38^sqgwyM2|-MjVlyvKksWu!0tph35_d3LgTOBf46YW$Yt=<
zAxGy70@=h$!(CWyL>DjFvi5e7nD5c*p7)~HvUK%6k9{^%cfo5P-{?4Ts!16!#OFkL
zLM;>S)I_`(9DkwKvzO{;PuB_@qvb~t^XZEu+@ecvv%wDA=d<8Xo*_Tw+&DYwP{V7}
z5u;w@vUf3;QjZf9rrtypY&&y3b%fyv$olEohaPtycu?K9gnePS-)`Met)3)WJ3%H=
zbC8}D?b|riua}9DDt&TV?i$~lOw?;0MX8C9@^?*lxKaFlS*0_@gII2(Xi?v)pntCz
zoV}?OnF(^SKs)0WyUJnHeybGA6k_3AE4|jS0(lHZHUO0bGHpit=^EzPN%>1-{tUFS
z1gaucNIhiG4U4uWCxp)wWX4k}8$rm}xTfit=`AbnyYU!kh{5PVJG<`1W|SI+^1_8L
zb<do|<~`E$)coBEmoyCzZWNrzX(~y<xHzM{+)P+(f)ObDwvF6Q*DVf(WHQ0^IdE*i
z_qHzzUggwe=A3;5xjOVC6gBSE4|LJjtLf!mT&j8l?Wg67c!E!HwoSjv*FN6=x^B7K
zU&S`-lLHuPbVWr}4Q!=fu;pgQXGP8hCzd_`r8AC-&S{dK%J<Zed^k!dlyMq&lIK~;
z3Q{WB<_HS1FSaN%d_F#tM!oQDo-#$EfSVGv?#kr<w&mUj>hgq{xcYrc9px)CZxTbF
zk9;rgD5mHkewMbr?{rpiS?Qc`T~mvWzN!PqIO6GtjQJ1B%XGX=wVQw+_o8tEnnuff
z4R&c!@z#5?sMOA1{uK>>dT}8$bEF0s*~dje?Ju@A5ZN>0T<2nfs`W<{rV!wbhTe{^
zG+ky|*V6>R@%C|kN*2M)qQYKJj~`=}LVZ+{l88x}>1#ZqQ0Kxi{<=5Mq2i5+dGi%5
zM0d+m@vM%XsjTxdl7sk9$<+ztZAdhMB7H5Atfvv5GX3an^~v2(%0hSCTFk(H*UkLB
z)`;=Vc8(^6G@U4(K!hGc*Hb!dgNc4lbP9*Ypgd=+h<xR-+<CCpbZQ<mmu=Z{gSwv6
zNi6ZFghZ-`J*RRM`4#h3&4=c54y(Ekv^FpP{2da1<ZKjIE_M;c5QHP3($7XsM{J}|
zMV>3!bpMqUN&YHgh{C$6whIjSL8t1ruS62AL4!c%`gu$&<-#a3!wkXqDEiNbh0*Qv
z9f7{cA3*o^C;dJTo>`t0M96zCNC4H;gQEykb+mNW5bmwow$1TZUHdN_))B@3k(g7d
z*vQCezqUsEM9lIpsZ2N*0R>1~v0hEa;|t@Wqtkj#|DHvt)h9j=H@5v^awO0*_|^t%
zL({EN<l-k&^J3n;a(bq!HOJXI!Ftv-diF*74f9Evw9I(>E33y94U|vwJX13?V*gxw
zt0P)g0ZI{Unx2VVB_(66tlFF@M}OxjS7ajpxzHcYGg6c_e^+KhHl7~efvI7ka<)s4
zHRpMOnbIDm>Dn&@6cx#%AxQ6|`pyV0k;VG_IIc2lkSSA(EI23U-yQ);<;=`%)G+AZ
zOJILO+@q=pn*KP{roBp_*H}bAIHsM)gt<byK)Ev~iu~oV3wFM~!%@U`c5>hH7$vX%
z+B+tKNBqizk<>KhEV0Y8i%(-w<QB)%L;st-PAI*i;$4#DyH6&0A&-!kGF}`y!+3pe
zCeb`xD}w-I&;GI%kFv8gwZO-)%vW|CW&7ksC^{w+FXEcURAm6Kres^ep`pKl^340k
zh(_|PjT|C$X@4m(X=tdhoM;5Cj%aogeL>pYuyv6~*}?&HQz`!ST&46)!hKH>vaGR}
zHe%4PeIizx_Yl-*a7m1_7<<fTP6HRq9-Q6nke~C{EtAY=C*65TH1@nL65~t}!B+G`
zMt{qL`r{42c3wps&_8nycXE_2@<+I2TubW1KP>mfsCK4@Q{A=viLDW)WSe#z53As8
zE#GvJ>4h1Zvpo5FuJ{<k_ga@KiM16#5`t&;<jDyScpGcdVEVR6wc}Q{)sqbM*%atf
zX6iSuy_%J9?xCvIEdf1bzjwIv0gO3qkGyekYrLz@et(m1GC#D2TvFCT%v+=E1Mrpc
z=|3j5N7ogTvc~rTqm5j@wqfhr9$ls5DI5TVa@m3(rqG4^=j0$&oL1HVFh{a?3~zgS
zeYY-*DTC@~q3h`P$oG2C8N-eXFV{2O*>LUBALY_o4x}|1hV?&R4u})a9pW0L&qm$P
z@1ftRuW;vZtmiduE>PwXhj)%<(`l)5)U}7W`#gK~!g3gEL$YmbMGaHAF0soqL_h24
zoQ_ME94)m~%AvnB4I~rB7@dImH4O4)chIoU4omrQ6A_oktCM4=YBN@*J?KwNZ9`OG
zFX^m_(QmxKMt4xN^F;=S9?#sgmHT`KU@Ce>SVICi>vmJDTWs($SxAa!(6Trx<F_^b
zkKyEkU>O^_xjOQ@yV5tt7gF@8m7lDYUC4JYE_fE1zF^P;@4G`>iUpB~Pzy#uKKSvr
z!Sz}hyW3nHbNW6h92__VhgTZ%=s@YE(;j|1pI>Pg{#Wda)OQlaxIlo;JT<JOp2YEE
zeVkC!Eg>BHm=@T)lnB|r{_%ZIzqE1UhpLQ0MRtu12sMIkq@S@xxiXYYmh1D;ps}`Q
zQxdFEGhh+zH<^*928n5T3>TKT15#~K_hl*9$stO;g-+L?qH)o|*fHl{Tk+;k2SH^M
zXqsxZ<adMw=KW(~#?qGaw|sVjhVMaAD02mo7czWK?eFbnov`nPDC+i(2Qw9xf2Xex
zTfc{n8P=8QpZUIX@r|xDKLe~;CK;f`)qTs)OZa3OnjQ(Up?9m9?U5lT=G?XOcJr{z
zp_OXEI2ir>O*V@w%?$Z_+q6fXk7*8BbFQ-n7Q|MmxQt!KP6cfUiI5RzpD$9gQ1kli
z%KrP{X_vC`k#}iE7Sa)`5`4nQb@wt)Ssbs0`<Y&V@8z~sFqL;r-P6-3%xf;wmHv&v
zxzY6#dzKZpr+PKotCiMkTVBGMX6;#Vz%uz*Fhc=rgoK?z=?IFp6rqtue-A0l%)&_l
z6LLC#n+@#?(rGBfuz$hs)?V;$6e<2{#0QZwwXO1VY9<^mUhGVcMq{oVB^+%-B-B%)
zXCadWSpJcZ(hnHn5ll;>U*(AdjC#K@t1u~R+fKye9YC@~3WHN&Hq~GB^a<iF)GX)F
za`am3(iC+z<aAvV9OE$o!l7JL=DemsFGcM#LlX@I>h?_ssz%H;pEz|pukOQmlGq<{
z%&z}4Yrj@C{=%TjEKcI-4v1cFP5pD!`z{+X?fQm1UW1ZUwc1Sx&8*+d)G)f>xyvVv
zcDU!>huOYQETGDlspVI9vr(ObD=XI6(yiDuGZ;-w>{-)1vwBsAJh#G4V&B$BlF^|Y
zJc}|8A_s$+zs=HlWbU{YLPFS9>7;mC9q?ChGX@u$5am?Z1_QWs;#OZRG*)<!$;SEu
zDn(@OxBufPrn#FZ@db_~<&dD2J6xTLJtS2Ga@yJpz4h~7@_TH+cU){k@AqYr!=}FU
z7eo%fbK-=<D8q%Be3IxzVAEyaw0EGuKCA`#Tf^ReJN+a1wMsjBc_=gz$he&N2QnHF
za$RcsGMiUfI~i@KQ0~qZki$_@!1YehU1L+>x>rz^Of&?af2Vw4-1$d~xKRsvUY_dJ
zb-b+oUgiQ2oGVznD0=w&cE2Dcf}QN{tIG`tR@<g)P8O+MG<mGBTlXzH)F$_I1o5Up
z2u_n>!NdK^$xAkMRY^-p_hkdzck-G~5#tV{vXu44{GN9VRef)-_)Y(1u!09wvx^z_
zwneXdmnJ601Qs1v21>E2o~(-gt1t(e$b<GJ=xwJwa~4now`3+ed8(wDVOU6Q5zC3J
z235|GO@+?7jrnoa44xDpb?nx`oOn+HDWQq9wlV5mJm!T;xW3-QK_0sq(I$?Krs|Bp
zId2OcB6PA?N{?qmQY%qLl`K>Aul%~h-B&_8!-2*LC}bRc&^i>5rc<VxIQv&p9w#7)
zgmM+Pj=b2E&@Qz%q^5N!V5hN?^_{X1o^PIll0LS)_QBO+l&6bxE(C(+qx8QmEni#9
zsb!wg^%XaR%iQ>YaZupi18v#9{FyZV5(k84*YXY{Xdwb*@w?Vkcqdr$vzVj3<U@}r
zNfY^m0QFcYW9KnXqfdUz7+z{edJQp;l_|c{Y}DAcn*rPr(U_`rH+xM_-`SjNWgy*F
zNr(_eK9%&HCA_hCQC_CCO^}o^-ydXNEq?7rBu_3}&UaQNBQef)WyyH_{tm4__qFft
z0=?jTQ_%U6wRy+m!S&8Gc^K1|E?JJ!b&kKyOZu;h2Frg2Oc#JrHgx0f@+;nW@@Qyf
z+cK_cD5iDu<n3#MRkCt!n}07TRkl$j4$M*D?G=xbp7Br0H~HI>i(qF>3c_sYdo&^-
z=$=m=8ez($iwJFG9rsr9>Ki3}1iD!BWQW$u{6HWzr56jrE&6r;_Fbb;+W+D_Ty&)+
z%V%w#^0NMSmvb9cU!%?Fic;i3?J={u-h6&^0;X-ZQi8w8yvt?oZ6@wvdq?xha&|Qf
zy{N(N?h$Zf=0Qp&^$caE%cZc?x>^EHgjHED{7sW2v$9z?ZdrV{UK1B1VXF}(n)WjJ
zG%};GVU@OKVZo{UNx$%p!A3j#y!*0`B74R;?=#}C5@M9uiSGpZ^qOiE+n?{r!h<Em
z`)2?_?wk$m>F1t8R|9dBNJm3I?v_<lX^pYj9H+T$@LVp@WV{qs6KH^EqqRL0HgSQ4
z&=#o_S^s63O)ejA`-Ii!sS^|^G}O#Q-SBb1O&T3&F&gpR%!5-_T!j!?DA%`&>=WhF
zKW%`=^^L{cL4ntS6h+(_hQ|E1`K$r%(n3ufru2F{Eq<5j2z!S@*jI;aE5VA7vm1uE
zkIt9Vu{nWO7gCH6C%iK;q2S5Uulj}VpSQy%r<u(`fO!BHD1ytZe6mzshLQb?@Uc*c
zoG|nj=-|f2Loe9lO!Teit0aYyD_BA0_CJLA0}Jgf@5|D8mfDNtHGg28?BE|HWbh;C
zHLKPt6%+LNDCz)$d^?oY!~y57dP!GSm(5!wKu07k0H4`eD-7_|6WUdn#_<uMfd{^w
zZxKfCQz>G2(!0w+$^H68ddj%ZV~Tq(#M)fixpWF%zX@gqz_g3>`Ocg{57TjA0G*jH
zFz27hqVJ#X(7sHdp?g`$P4P31Sr%H`N%?}Zew1d0MqT~Mir2h*@KOKd5+%UWXJolx
zwvY1{Eb|)7Ujf;xndcXj`yF`xtOA-N1|n>)e&hLTrI}OOFQ)z#|E-nTY2*A+Yk+OO
zY$)*RgcwFB;I>`+%Es0Uku>kSl1~BdMSdWAHs9h!b?4bcAcmNE&XhM5IF9<-Frxx^
z$DXu2nvqu3NVNmcfaA$E9@gtoixRM@GfCRaFPGW2ojQH4D4K){sDj<{A<wn;TEM8C
zH0B~b0LFpIcVoN$dV&{j&B{M4sS|zMT|4Hq(!i!k27RPM&u5#uV*4t3AC7`*le>7%
zHSs1~WrWDd%ny&>;x67o?<KiAxJa`oo${u(_zq?KM-uCAUO8ioWixYI)LNPdTCX;C
zBl{z;GJr5*iV_85o6SDC`KUY1aFZsVIsReMQ}fzW^quk~<Tb+)IBd!C2DxGFO&z1;
zQ}_6%!l46=i$CL{C{W4#ub_{_Be;RN97XR|#!Hd@;?DFz&&tvdn}kQ{S7M-jiEw<G
z#fgpr#~k(+oAYG9v{^Vr@yK3;?7GX$*C;01I1}O(>TM;SeQW%D5;<k~f!aemwEIp4
z<#^$A1`l8e?`G@&F6wDWF4T#;q_<j#)eWB|)4J5Smxw<{q-zNc{}q@SpVJ;7mC~<}
z;1{)9rH(z9B!9J&2SLte?u$--dIyTGG(a8k-8U-G2M#Gc%<cvXgSy0BU^0{VkDZ5b
znw#JFD$BBWem`|zX1!Qea7DLP_J^2pwJ1iwl0qJps@mty#YjBU?V{ih-G71YH*hzP
zO<+*DxYUY_X^>2|Cw_G#poh5{gso`R&#zcC_9+RcX<FZx8LVrJ=}!n>uEOI4mW((7
z8PSY_b!?Dfnqt@GK_9e}Ix01Vv!tk%hU2TCN^#Q49^&==`jP}gWdlo%f_65!10|D!
z`;ra^>fXw!uC`rTKkRZI9lJb13;bFzvJ>5D72MdRH^Ht23-uzqN($Yj)U?!Uo%Nqt
z*SRm{?8*m(XZFq8ov)u>Q=NLrQTP2d5gp;FmL%~IrhJuZ7slmC4>yvpxf2CIkg-&~
zW)SG3PS$Sg6qeaVV=>HnD}%~C(cPD2+qDGXdB<Ou;P4t~$!+%03w`cYmnP%RDFtkJ
zrqmBI^_TSuTsB3s)$Fkr%s0^fQjqmrmU=v%I(<mI-?zKI6U^rX*2bG=kAEo{h`!3l
zWtLUt_V~9Z6oG=AB57CT2NFKdOhj%<_ihSctYd5qYXxD3jBdMoGXDFHy-$`Mv7(sd
z`NvPK>Q#!pOvW*FxzVvmpVkaN6nyTh@U3bo`+p?ih54<xZpZVhD3y)ZTH|9z+fGCi
zW;2?4dw#(mpeik`?e&WoJ$|2c7nf)l3S*dkI*45r$h?25Wz%YJ=<zcT*baiVcmh0~
zUhR&iQ0a23OQ`y3d_Zpuf?#@i-P8!8Z09$in_i1|4^sl@rh%RqGECZN_JeDvZQjI)
z(SQ?U?t80rtGIBo!D~M1g_S7FSRvwtHM;XhOh{9VGNNRkZ*wI`^V35JKf~{*>6%k<
zNaz~I(N`J+FT5a(&iRk!nan$6+g*edk9gn?y8D$cQOglp2YAJ)L<2Hkw+zxIFM|n9
zWNDVGkgaqQ=u3=NpLt9QD=9v^(2Pouan)7x#n(JRo0!<jxj&G!lM24q@1nPRfAJQH
z8j*qD`uGMZW<^kODMv6cb98zYHOLlJm*nzC{YTQV5a-dAC#t8BC(X~ArkDB-F3^L2
zlO%H<`3L0OTEjrjM9rF$g_^$(J!D6-s8V6OZJDQQ1%FFazd<ZZ*r(h{Q2NA{b4Nfd
z#gm4~DSEI_hQ$?|XV`+E38RTWXD<l1FGJ<ygPaObLxVblLz8PCy5PA9G&XW-d#1|9
zO~)3SoN1=B`}yXs**z5LCBbjE+;TM1VJVqVpdZUpe}^rp4C|&Zbsr@Xjf#$r-K`cY
zF>S?v37OsRE`JYoLz<MBP6YD|S6CjtD`}<*nbnSRRZP<>4#4v(=GauOOvgPRw@-D_
z6xF`iPO@Ioa@>2cm)Kv=qZB7gRwjEVGbL*ZOR4DP$Dx!3GO$=GlHWG~8TeK0=5v+^
zytvB}4DQvpkv~EZluNaAE_A#h3`Rz}trbEOntsuhAF4l&?&nZO)fn>!JaD(WOY=%5
z=KU<N^r`;0wlz0Fq5D7qKVr@Q-MUt|kt9w5-XHJeX_k(xYIJWrxX%8|lycVm=!2^v
zxw@1$o#sSYsR-lQ=lqL-HgUC-4+33tQo4xAUAu4wI@Jyn$LK-Bzt0jUZ7jGWi<W{G
z*b=nTBUH@fc4Q#QS-l>k6qRc638@^FYC<+3)kTOQs$oz|Y|p*rpi+i#nzNfnTiUK&
z!|&g!-QlUJpP&k-8BJXseO{>Iq(Ozsw8N!`u%!>2E!oRj?L!rlyMh8ox<3C1Cv$l+
z>iEP3hjniKFx&|(Lh&AR)wVGc7fqAXzQu&hI(e-TpbR`DpvLxP+PS~3WmB>BkF?k`
zA93^bSxl7y0iX2imVHjH0U-4{?&^<OICFJuKzJ;@$wVABSS`csWxV8S;!k^*KJz^>
z9$OfDpdebFFIC+P!KHNi2jfhQoGu`LAH&#Y{`lR|cxu1?a+4{bMvwpmbU?GKqA0Ty
z47D0LP>UCT($=jFUrl^`s?BN7FOhw=q6L0^{?qoQM#VIivudN}z^luss0eie0$t3v
zU9mgONt=F3{I&urS;?ed@H=c|v=DEgs$Ey8;0DY8<83yV!r8qhKG3VCH(qM>7Tx-A
zAZj`IsZA3>kXQLfjgi0tTX#Kt+bmDIe=8NqKqqeEO@1FwsPU!$wMb1ZkyRdxm5Z7Q
z3g_4(&BDY`nbkz@=!wyIqbH(Rf2}Bw$2`DjN`H)CrHrlkJ-Y1z$ea8zpxP*+Z+HHH
z`k?2T-&!NFSC(R$ddY|YN4Oen4{7>>tSI}`qoZ<7kIg0#1+?QqX>`kduEJ&TTY2qT
z4SkW@mPaZQA2pw&Rqok~AZw`eN>My(&KztVofDfY5E>0*`@?G8AE&yjoOQ<1Ts$wN
z`<nn28QJ|61v{++Dz2}34;XAH6XT~>45j@eIzc!Wh6Q%>^O8~s#;sWgnhLjc%leN*
zo%MI^yf0tk&g|U#=*jf#tHiz=h?7P$r<q%#lkTLV?iqn^r)3R?d$zXnQ&m7(RmBD~
z2>W3EdnV2wMKGl4F_lj{GY7#-RC^{&EdPjO)VIwHDm#E&mDScXBlf98#{AZSF^9;I
z@;C@*Zx1ADmSuz~c9srYpU)p5pLh(-!WHxE>6u8^P0tG_m+=Ym#Y&#3{j7Qn8HE~L
z+aqu)jP6;DgxZJ#_lP&LD*KZq@@@H>a)AyE*SR5soa9>rudaQ=^4+2$$ZI{oh9981
z$kePf{o6+B^026xAeEv(yc1pc$3U>5>jNwRz!@y=Ra9J+UGdu$5r1U&LAW83Vl|(~
z_*Etzxxya+_0qIC3Q8VsMgS@KI->;jX`J-dRky}Q<Z2~pamv+q#(tm4CwZcU>8%Ci
zb-K=DWS*Jbevzs#7jtvXAu!nGRZ2uQaY;2m{5W=+#jeHcisvYeI<pF%j32df_%)Em
z$TPEQ%T*Y;JyWAO@R2M5)mBWSm0dD-GPOxxUntv=3z+;!(=<Bos-&wC)@&ilX7hlp
zmIxlGilzwTgYKE+6<E<|KSa-*n^%!@`Fl@eD^eU*67m|cGsuyYz(0<%0N9DxZJ&rG
zhVwn*5Cyi|J)#+$GHA{aGn@%ZB9y5%KA>~NrkPhPYY6I<(IE7Rk-M92O>wJr8yiY~
z=Wvg+GNuMNvz_mFRCnjd|DMf+)3kErMIvh_TCOX*lL5gaFDSvg%B^Nii0n{Xlj^rH
zGuMwS)-WjIT#-pKt~PcVnU%Jz7@p*rj_vi_++croA<`eafw5@fn7Pn&ll`DxIC6CR
zpkJ4-Hoea=(IxMWHrd<AD$OF>33C(&3JO6*hX&T4oG~^unA}!r{?%`|<elyPBbGZu
zK44k*K(KjgM2gqq$LXAPml+SWwGoF+M2|=`o!sj>Az56RkVXS%9Zk4KhOVb}Me4iI
zf4By&JgMQ$iGOicS>g|lZ27aS6XE;A7b5~)D=Mrr8O~?;&q*$zF3p+Q240@R1Y>#)
zgnYk?MBmYT1_R7aALe#8;Le@7Ry$fR!4(m-cY0M;Rv^BJ<(dC0zgr;C#IrZXS)C(N
z1DGc)F2AzzqnvGkwi^A|3+dLpc~oc!#r=_~h7K35-uq%X?K%fQ2wsbC&J<gmmLC~?
z!P)mOH+CfOYeXJq9KzUHd|^d!g9ahoOjvkRcfbAPO@E=QI2eWJS5@@%)F0}r$Bp%p
z2N8ErUvknbQey<wC7&BPdiTM<+0~W^3W=*3q0vrKZL<Cj+si*(OJ4J}n0D_?uLF;N
z{)x=Qy1mSxKpS}TtJhad<uW}~CMg=<7$xLuPnmnH>6XhFaIn3Y{WTBpp4j2S&Z!p)
z(Hu(P-7R4ccC70zxvwOW2awB=kXI^#(l>6?UqLxW&DKRCeO;PW74$Lm_!ESvX!_h{
z72~DVmlX@oth03e<)6lHIwU%=ri{)6|GkDpQ*X=nqj+vNvB<WkMsg#b)WzZEIJY@Z
z=gF|bLig8xOCJpJcDjoJz44pB;+K^)=<Lg$y&TMJA)RutY~G(gda$LXq5`t(?Tv5A
zQzG(s$BH|p#RTHl<?woAGDbDOUscey2DyCQ1jo$%JL}csscmBPA$EMcS!2rd0*;E%
z(Y-zwqt=g{fV7A<%=QTE87$6kPA=Xh^kHr5vFuKd#nF=4f%Q@|(MhSv%Y3{&7@mi1
z!IOV2$s2|_n2Vsa7BY-qhHIX!VAxo^Jj$MJf0-4);;jaKa*g9>Xd;upq-K^4@-2TK
zX!-pJ6Wk`L^STgwhnf380Ixr%b4Pe5EdN7{uKrp@v%{h=ozPtEYWDe*U_n7tL|~$%
z?`8ei^T0RdrjkDaOM@v+rcdX6-y<^N4kq3G3Hz~(lnVRBqt-o2eR)Q|^OozznZ<`b
zt?L_`RA#H;RO2tAqjSQ{14`G2=AhrBLfz4ISFACCUM*gDt?8cJN5K0%Q4=i+;SpB4
zlYD+~l|1*};<sflHqiFeIxk5Ze-0WZ5GsXay$BzFKFNsl(d02lpD-P09<K!~-re~G
z&*kmN-Fi<Yv+VcMI-$aGC6Z}{fo;r_Y!3a7qqI15v<e`<N&4xa#h#iQYiMVF>U{qn
zZP6HgG9{e6poZCYjo^~PLc?xXC$MWC{#3Qz!G$V>|2iU0bnC|)*5=Phh>^!qNn`ij
zC)GLC49BGO1$TDs$T;C|5ItX~cJzvPl?ObgBOA#0PxNY09D=`hFG;)d=`Bf@4XbPV
zo0*3?Zxluo-mhwt;jz<LG)|d`u1gRfOE3s%dqEi?-m!hI%Hn&kl~q(RZ&6PTFKE8#
z<#D)+UVWY;<MEi}w4zqb!N3*2telo8K4WD3?X&B{=_pm3Wi1VyfSClqXX`-yk0DZz
zR>!G{>P3g-qZ00*1BM7g=MzJ##6?JMQ(_qV<3bdsIwigce}oo!&3B&Lx?D1cqmKbr
zfPZW{mdkQbTpkGxG*JU3xN<dElCw*iPDgNPS`Pgxt~g8Pd04fd)pNJ@_j}(7rxC`K
z%>mYv?>qn_P5bJ}J4Wk%DIfIVC|wlYdD^$qH|a%Y1Q^pevHV;x!Q#nPl7CDc4`H~}
z1P~+%#)2(^lcS{vkGAvIi1?LTPEn!O0EU*E$KztzhlYn<yxF1nu-!6Nm!jXLi}LI0
zs=9&0v*YBxgI=Kiz9Ztc+K-&WzG+^lAgpfD=Jz{`Xw1nL)A583d*sC7Id4$)@cyLo
z=-cyxuUbv0E8cO4R3PfM5~5^R)tWyO?8zvORV!fo`&s}?Rhc8T*;7T@u0VUvQEf1N
znAPK?rEp`AjeRzO`a@<=<eb8`Z5Jafr@o?9c&+=$-qu#71fHO@xTKV^)HB7GF`vlv
z-Ye_Wtu(XZAVxe#R3Wigs~M$*S3Fy*wW6I$a;%=t`;SDg$f`O@49WIXISh8!4k8rJ
zLsT@(F~xMYdli@eM`9Q%j<f6J2$~UU`-G%G26^$4*Q<m~t-F6t?zH+M&-6R2Yd_Ja
z>5(ZI0F|iac1MH?kI&4Da&rFaJ@<U36XUHm1JReq4o3s^rl)F}0dVR5Ox8KfEf908
z^X%BG3wtQ1(-p5&H0IV{(&SIiCIO~nmY!JEPzFY5+3P*E&ZL1^iRxE3BA(*&>}q)y
zQF~xm+mVZBX3KMgT~x<J<59biXPWEsAC)G*@dI<4FBmcPl37NH(s}#kBA#bpCi{B;
zRB_{g!@JB?!61g^EhWwMDNEpkEh{0qM_r@5$_WBF@*8P9vZ9g{cqP3@8<Xm~u)Cta
zhATmM8+&BIJqQ-^$CgqkM2s220f)(@$dl#Vc?R=-VZ!~IfzIfe8MV8jDZ)*3l%Zs|
z$7f)ynpNVLfjQzj?B*?2;pv>4Ss<u$wh61DpE+p(l}!*Ci0@b2YP6#&&-Brb3<VU^
zh-*g%&pcTMBNrzZ710L`!~VUu;}<#(#|()h7IT}7&rTfTuNPXYQ4^s!O0N!VUMs)^
z{I0-P%UsjKIRb#DVpLbMj+LXLc+He4{vN+f<;9Abjeu9G%>*m`;A;vd_~%%N{giv!
zoVe=3;3gmqr>S6Dw5wO>uJ3<&W|D-rmf(3>{;F)=%%*X~=b?H5lh2^A?;;EzrGcp3
zuW2B45fh8MV*j>sF<{YXi1%kf&%uA~c)^j~Da=i3ZT5um8M$o?Lx5V0j9iaZ_Hngu
z?_qg*>jd?7ItXV*uy4L;ZY<h|RAj9VS1)eaRQ%xV8)6@+b@AS`JS-h^r||8$)xbSX
z2{IzsU@S_^vE!L)t03fn(uwsT+OSl;d}4AeI!y7`(C<r%-m?nVE7~gD0yM^BIiDSJ
z>V)A({j3+#h+x*IEkE7OTxY8gg#iJ5y`$SFrAk3WVS_2njBnJF5+>@B-d;NkIB~{0
zdJ__IN-86^T*vu>5ZxdW_3XHSKR`FSv7M3ZAfu<fikYe1<1C%r8TE%=8>}>#J(PR6
zlagMNonw@%^Kbc@r3U5u7KYGVCdR=kbrFrf1Cb`*DxM0*^1OFr@hkAM^zj|v@pSYo
z|Hc-l0adxNoddFZPBE)&XOGwE#Sd};&~>wC2UoXm8xPD+_|pXnwHg7^4)_J;_pORM
zQEUaC+BJ@Xqln4p-Jzl%I&-fD-DH{I4>fxohrLTQjO<Wqo)wK&J-E5QZ+0ySLEO)=
zCJujfntrAJK538LHQ$PQnlF)o5d^itiWVhO;~=ZL6_zq7i6HZ$mm-TxU**U40z1}r
ztKY^p*7jX8fs14dFSj<nHZ$fg(lz{Isvpue+uK`O4`%V24efL@Y>zYf=rNPX_w!c+
z-}$3o5P-57U!w{bTrxq;5F&|JuDy9kykl^Pq(kD+skZIhZHt+6co*M~pO6lpQ{eoY
zcSYDGXE~2gypP;l$lHWmhN+N-^o<Ffg%k1zD|P_b*vs)oZ}sM|GsLSAT8Dpj($_MR
z_>?rVXTvUc))XK)GHd7tOKPge>l~3c4Gel-_Z7e&X>YB<44aLr%=DIxJX!l9!}ZA^
zmP<))%dV32M>7GaJn9^`=1ROVcP%8)E(|@KrLex7y+L)+xreq37Bb51k6@ebp-^vh
zF|H3%Z>5i&Ag6Hr^{iwjb%NcK?B9K_rBZ`2?bn}A%16%LVn<Uwv4a6s5JJdBb(U3B
z>b|@?SFp{uYN0@-j%Ora_k0GsMOo1incZmLoB0mEvJv<G#YWxiCL!sQ4YM!_W9K7e
zljyfq3@=wXp*+nV@#4`K<c|jtyE_F5Vt;((vNzDNyfK-(Vlh&Sh=)7$54}HHr4ib7
z{@IGgbpz(=^1`Dh=YLp_k`;;yn9w-Wl}l)>^W+Zrq$MaJ)$AhsWnJdu?66L*R}N?2
zWO|mLT4hB=gi(iSdWw+=NFKkY+0%Rgkc5Uu9*?hZS-~I=n2$*jpn6kte&~3!Rk^Bt
zW&{vBnW4t15q-umsp%i&&Kep@k!Vj-xBPtHuLjWM?$3_}4(HzOGG<ii+Ql_{wha!N
z)p99h=pdd|vm=cxVgSsE5_&K-gZoBV+e=cXt}SS_T)z`Sk8s|_w7)A*k>F70O?F|v
z4p_j4u?&T+Rmke*1=~dY0m^&;w~MgVkWv89?%kUDpEROK;X1|d_;PlR0R{{Sod!r6
zqozt2NV_rgN6CGBviS!m$q3b4;(I}1+!&}ws&?}ER?cO@xwpIy@GMk&0$f_bg4<w0
zt}d3;HKchup}IV?aRj`%LSxS%n=o;7e11kc&!X!GSn><>BKZkEvX6*r3Pdf>hmAFR
zh%ObIQJWPKGU7Z%%TJN#59>=*k%Prk`!cQ0Y414FkwO{}h2ScwHF3B#d;Xf{?H|Ss
z3W+NdNz8mkUK-&|KP%ikg&``y5~#wwU?}6%bieuY0Dgm7bj^(XcvhEMZ&wl^xHES2
zzqmWgwl><RT~jYmXz`+L(GuJpS{#A~f(0qA!GlAE;_d`35`qN}Zp9&3u>gVMP~4sN
z*}TU-_WL7tzDz!4l9|js$y)2Y&Y6Ob*`yLc)o~!xJb>GdDAG&ET9beuS$oQnt`p|9
zNJ9qbT#!KhhJ~LEM31z&NVHED8Dj%SApZH=MT6y74zhJptHvFJ$a)K573;zuk;M_Y
z44)>=>?umr<|<bdtR_inrb)Td$oLy5%?^MEqxFzW;m8xW+E<PSZz4Tw3cG|0SXm}1
zA7Qn-|0GdUY%7S>C$6cNu#E7JSvTXBhW-$<eTSy07J1ygqf6be$aIx)s-!qFd+em5
zPsWU78Mgz|o>Dkc3F$E@{2Z*}+62_()fN0%NKFMDpE(#5rve|j-xtzk64dwyC*j<P
zWpD)3GnP34l;7@oGZA;-h*5%c)|$W}P1^`7e`&=cu^On%beFu2n{z5LeZl_m*M5JA
zI0`WCtbQcqboWBY`ADFofgT-BtKl>&B$o1muioW8GI2adE(%P28!DtpZm$D)<&#N}
z4Ol{I5krcn`xzS^1zh6S0Cr7Cy#G4o=$kF%>ejgPxp>&Zy(_=lr)PDoh3Z#Jh}ZYw
zn;BGwHJX1*g|Xlw3#@c_52~h<tc-hIfM0dLElPhSC7*L&Qw|TMtJEj(x{N+>M5WHc
zqplD6AJmgcCwk@Ir<tY!pIK2t{}>=NP+E>Y*5OzN23AJil^x@Y(F)h3`>Bl~FBY57
zrm2P!l(I7*b<dX&3XBbAnYb7m3<{Po8=g<Hc)VbhQiFT{%58O_8LFm%uIiIYGAS^v
zyVh-ir<M;Up_=703RS*?Cf1wPrC)rPc$;{--<YOnR7^UU$Y%KWJ}eR3?qJVVMQH6z
zx&L%5)%jS2?|IP1&sMa>-h8I0>F#0nqgQ;CZJrC7lT>b18f5MzlxH;d%0V;}wp#9v
zM565>2!!<^x}`p#qT?a0|Gw}n%66;ZRgGBqm92-7D81g`1?RZ)mi;fMO)8}pP9yLj
z)Yt)SDPKs=+?h$E-d?D~EUuvd6<A`+)LVT~6Q_RdcYgb*bq+p<N@$(OTAIzpW27n9
zh%V*p$($z+GP@$y&<0cKL@Hk7j|b#X)CzI$woL4wDm5Z<vYv;>z_6dwenGcai9}jo
zkL1AFM&zik6)kFTp|2Z6h&<>v(AH8uAbpTKu7QMT(m6rxW9#?o2RzI(@>U2g;}>Tt
zuO?85%Zofb5NUpJ3SkaxJ591~MdRiD^psAr$|X^Xvw&LZV4?U9gL~F`4QSz5iQc|}
z&Yo2s{f-r-93B;8(sjAD=q%x<cKR%1beb~$pO%N|?b!fv>w&Rb)q?MUcIruV)#j=;
zbz@PCjXgh;N=BR<Bmb&gLYe0W$G~ORj!CEO-r{X(5`lNHdX-%j3E;Qwdt?C9GKfus
z_;IVCMNaN38|k}cv?80qB%7b=A#Q7w_Qh^PIrQcX?@T@GgeS7Kr@3y85L!hJa}GA&
zN3BCDHI!B35?+WnxZh`Wdb^Gdw2;o3t<>KNU!SbGj4XZH{A{t@{`TG}-Ix+j7dSd2
zw3>2$Q0i9Sl_%oeX4RA3p`Gnqp<tSztEQSYSeb=TiM(}>q~`yL;L2mltR#KoI6ocR
zxRpEn)7p~1$!*RyW8r15)5DJU)0qWij+7D`1}()^9J=!%tTBvIG)YgeV0B+Ob3|gy
zIe;@nQHYG$iV-KchUxl8{m57CkM+gT80)C9ai=I+#3}QXftQkAT(UIES;0LHHf1l^
z{Od!h;M`%q>`pl#TT0UJaWk3*+-Sw0TCXpEc78bKWh@-S=gx$t63lt|7x9wi`gTC5
z%Rt{@A;lwq%{oHKnRz?G;j~$)sKw1EIm+)aq`N415)-8lQB2iMw&S6-h8=e(M~w;1
z68$`RQio)uRkTa-uLOACw@AGvR`?hXYfC*vz51AOdzjS}J$9f=U!j~%b3s4jKnnSk
z!UQ`3|GPBP1)heKdPDb<<4ybewnInH`-S-A)CL!+lA1w%7x*JpEzJU%c+^1)-SnUg
zdyPj}ys^P6W?0*%+By5bn8E`i!8L66<l+{d7G86Vr}T(Vzz>33pKHRZ2eM7x)aNj}
zASmqI`wNr$gw)FjyMPwI#*?c~a%%V~o=D?vx0%+%`vZfdm+T|}eq})Xk8JD518v7j
z9`t&)Kd|`b_5;~$wg_ya-_-a6X(G2(z{|2hY(Fmi;ru-Svl;mL?ZdWxL2N{>l$6Qu
zIG$tz&Df5|12=JT4Sje?DmfV>RBra&W7rcsdV79uw#t`m56|$J8+TN>r2_n#uIpB2
z@Z!E%`MGXXnr8X<-M?{=QoqI_&zq>a2W~7#y{zW*sRyz#Cd~=EMIrmg$9O8F_w6^0
zK%?!u9|Tc3h<^XdTForH<pO^G4P^NPSuj$%z3Pc7lT-k%gJ-1Y-ooWB-CnWeV(@Uz
z@h7YC%h<3n%>K!jCl~!lU;BBJoF&nRHB!oF%K@#Sc(LLMxsNuGQhxV*3`o;NIsBYk
z+H}c9^kKHTq}Ur5@1(S>2P~!GnA+Hmw+}bN^4!L+9B1$vN_NL$TUwi2{PC!SRyxr;
z(v>{CPuiI)Mh_11uZ4C_@c=?3sTl{kW<(#z;m*ffvw>l_Q#=|6dB4=q0NDQS+Fi&1
z9t+$<%CGKzxaWfRQ7od;&`vSX=tRnI_zSU8=hb;`f9GAsDIU2IHFeI#k9(R2?*aR*
z#mN`ufq~oe6->jIhc#HA^DX3~6^zys+s^1)Jj!7-qkz5u4;$&a0`3HlNk?VJx(5?i
z?Y@pzcm>b3|MK>MA=B44&N&u%#*B>1rdm%V7Bs`SfTmr$Qc}^KWc}VIuwASM3Xk$C
zCp5snC29b#%wOZF5iB*QVr*jvPuJ`@dK?qm>^Sz}37Q&y%q9Kl0b5?$=a<S4brpIr
z4Vp4iAIS0W2)EZi;Nuw@Cgc)#ikcRer9J$=3a*HTpFsrw>)?7l@!9qc@gLq)mdo7Q
ziOzf&Uk@H>{`eVIo2CIc-+3>C!Xy4zN0pR!pTcy0(@$7YTIOi$@&Pj5P`7T{e;1s%
zTNYS05wsjT!L(A@kY!3xlKy|~TfI+~|Mx2S|Ln<we*N93fBm=Pcn@6bnD4)WZCQR%
zxK=mR^vqmJ<3^j`I>E~b_(l4<S_0V@hQL4>_E$at-++{lcz6$eeSJ^QedTeBwYHV6
zbfRu2m#=MeASat&yb|8z#xWeqoP<5mJ;HDYy{Wp_`_Td7t_d{#2yf5QG5_I(1W2=H
zW+tKfN#f9&=N+0oq7@xaP6i3qNBk$JxjP=^%s0KUeROfKEOc)Y{|!Wbj-zOPEhc#t
z-#486%0}Mkz86<!^(^WhMNv?BKCNnuAI`|lR$3sg(5>#9`)Tt_b3$Iz9HUfJ(3X&z
z0<bGQ72Lv*s($IP@w=$q|LJVXQ&g_c#B36MkInPwRvUri%LpOJ;v2jiDDAqDgH1MY
zh54GA__x3Q=!dMN@ztL(1mX=(L=4U4t$?ORX;7$=m{FNEL>x1)(@-l|xQ<p*-4@`Y
zd}I=Ztx+!b5Spwn5#Lv0G^uyDfjc!weUDhw4Gg%IQkoGDw5qvsYCl_Q`Ga(#(C0h{
zElE>i(zrI;=a8X0!VBy3#cTZJUFGR+wDH<+bge|2VCx3}RU={2TF`_MyOsS^^p%}u
zV}>aFOU_WU;g{=huDI)(vQO=t!o41p<oOLCnkKq+2)A9-v3Nn;yOGb)p)pIurUHl#
zZg**MIlnH8Lgk&}t(~3xojIp6e;-;rF54r&y79Kh@&W$-lEKLGxoUmGK(VXrN^}Bu
zi|7r`IxX}re%HqB<d#-Q%&u4Mm6rwk$ssLlrWpN9yM2l<GG`4&^SEDUc;2Ffl0Tbq
z?bjhJPUP2&0Q}d1aQAGp9;s4<La{77fWoVEShe}6XxWh!YA|sa-s8RpuCWUHZvP1X
zW50oPTS$Sz9tg!m`S|^O@p>?SYCs*>`7DgSJh18Yn}NPx;SlM4f>_Pk5!)qEC=ZbX
zI=_~C6SWg!6vdb=4yj!%tiR<#lz3j==Wp+@sM3zU>tj88S&xh0`aQdyroel)0uDxS
zy_$J{o(^lmml-Rrwp~`4X5RWCs|=(#vf!=}tI8M<c;TS~9&b?;8*yi!AtK&p{46i{
z=SFZM27r8<tkP68zUzv}<zo_Hp<!BMCi1L>!4^p+-;cQG7uxcTD9ASqA2Xh`)X&(`
zKHv6Hw_U7rDqAkEq^VpIVc}L_lx#c&XtOTB__uY&OA$K3YEkk(bl`5+_&O8J$~sWa
zfuMF;DD9k!LSxE#B#OVFaM<L*CsojT0l)Qxc-0#MjYlc>1kJ`01M~Wn_1^3yv#67T
z9#;nyg>31q$~Fhlhr-Dn(hP~NYLJa{i4gwDw(wS;<E9BwX>)7@Ma|JNP)@-kuhQ<a
zlyA$mI^+AENxPx8Z(~KRiY-sg>QB&XkKWAp<aCD2RT>ALW_ZoyBrXowT8av-Mpvw4
znHK&G_5@`4!pz{_@l*-${yv4J+IY@ZnGiaKVxR(CgE&=K*;CVeqBEbSG!#Ewh5K%!
z9T}r*38Dk6G7IVgP41OQ-YgYi8N))-z!oolGnT&fA{tktM(Du%gDhuwr0IK<`q}zi
zZL2daK1khvaOopFfO)q+nDtZb9FIzT!gwnGs-agqX^(~c-V$Y$EfWQ-L!-2mtHC;Z
zc+IBIyY62M#rW?k>d;NJrGuD30+$ikrVR~CDUB@gpVQXE%0*?peC-HjkY`{mWDl9=
zg|k6oPi9?$0?$H{Dp{UR5GmHVfd(cNO72K*X#}&A!}j8es73m}&HgI=MNySY5);0D
z=^&hsj%>z2^#O*IyNT|1QmmMfj`<{c68J+V#PYBI@X8~j-K%2*_Ejh-1~Q6EdlhuS
zgC#JzPptQlk<xj@zq2sqBnxl0&I9fxP5EkrA6H&BM=1uEX>|Ub>p-<9mXt&m&Q!@8
zG!eWhL53-1=1r=CcKK@F4tBnZx!+YjZui)?K>I^+#~W7QXIo6wm5waGsu(s#D3xo+
zz8d8BGI+hc-7D1DzoJDudXu4U95{C)v}TL>!_l{<cZHz+{_XQ-$BL?n5)!MznPDVp
zj`i%z<o)&@t^$9}BSqm!I?bOPNe@qWLaBPG2>1^}gw{tma;)eSLrjvueFVKCKcIuk
z#Ul;&<TNQG+>$=_#y3K&k>9?BF(i=U$U25;Gi^RQm?YJvSykvAM^XS_r!k^!sm2UH
zHp&QBKo0li)63miG3%k@=0ec~EVmHau2wt3ph<21I4$$o&)8;EejQa}r8^dB*1<d&
zz|}t<@Iqm4@aQwZ5Q7P_tu;eCQ#dS-2!CeR;h_9Je`98G-nk<stLivKdt6-kJ~G8;
zx=am(RQkb%AE#8kJ+*7@R0BWFREL-djD>yXUJj76G<80uu9gwODh@*Pst#yOA1tPw
zPA#;p{#7JW#}F?x9{>{azp@6qHc!SA67xX{&CleH6WwQ8tP%V)=1r3^Y%G&G2S+ZQ
zDF}Pj2|t7p^&7oVKgV@$n}E8t9#@=%N0lJ;kW5H_#(g@sWEW?P{s}f%*QwHS)QXGq
zA-Z^+(JX=QU<kR&axxuky{J#W#(i0AnBp)(i8LvkK5UkB5=C^v)sSk1<BRztEHi~C
zI|vkJ+WT^&hxQ~f`g~<3@q`3UlADee5TNsd{t`ZJFnbv;6y0>k;zWzNU|ABC$+<(=
zK)8i6Rks>ZQwu0V(n^2IUBt%{l@N8*gafAlMcP~NGD`_95}>M_&Ck!suj0If;7Z)z
z_SkZ|+(LgudExdJ_eSR9$cT9F8QdIw7a&edpORTSz>y64f-S*Hf^i{*I1eA^P)3{_
zlHbIN)Q4-NMkd#<qLoES*9Q`<R!W3Ov^^l_mu5g$Aa-^MFges~OLpM|nfo~WVj|na
z<XXFMxt&FiNHx8v!ne_%=BqZU4pBUH7YTOCEU7}=FQ@PF6m4vHE9Bh)ww}P_8`pDL
zngpl0d3ARm>0>u4n}dlg+Fm~!y4l$j&jkanSUn*m{|0&|b#YQFLB}^$5U9N)IG=Ui
zVuVzU660T`P`L82M5r-`F|J2Bp=QjuumjZANFOaRl|G4s{;t-HWmfA`p>S{X^oM#}
z5y%N&7{M6YYl<XgZl1QN<mq3lewaqIg&Yt8Q$NKKFnK312yte5Tsz(1)*+R&=Ye+@
ze_uBGsPT-ZRVVT9GiD0(J9OwgrFs%-oP1{8;{U6#ISl_h=cR$LZW!I9YjpVes%$Nb
zJ9IAHr0lmB)2H$#QZGd-!zMT8pGl7%?nhdMPpq|3Tr48%8_Sk#WNF>RGa)@T+2l&4
zY6Am3etVjv)PZ!Lx+<2`zU7Q_WsBB+!qpW{xtS3$RZ{V$&)Xt${`Nb_2kxXieODN>
zblfX<+-OaYB3A2r`(t1@+?~%G)?e8!r!o>@paVB4$0?kQ&)92&n~s8QF6gyN_|0dw
z6D_)n7<<!Ch1JWzY!fO8ZW>Lb%?K6eU6yXd+og}fh1C1*nr6AU#7F1PKMD_NO{lzA
z1u3h7<dX{%<yBNj?Bo*}8Sgg=d9KT%sRv+4&|<*P;*L;```2?P{B~SIJ|6;-^&nCS
zblV=c)(D92CD{^*DMpFcsCZ>~3QTIqInAvSpE_Z2DT&5|ikMNoz0~UVv?!1ht;Vo*
zRKv3aBd(hLpjouJVoARQ+}{0>dS<-YvHi;=b#WE$_d+C(67;<=6}#t6QabOjCcgC3
z^M?uKP?~`te(N0afiqll69L7qFfern`TMl<inS!@Ne1c#E{f05@6kbO8;LTTT6DZ6
z7lbQ~t3BY-`6<d@vi``zdFaLc4?Piwc4N0@Ec~l8;6BdZ;y1cT&mh~rJ8)`pdlFPX
zA7*Sr^I|B?05L_e#+y8EC7h;{=N>F<2s(<a>B!FNTNyR0`Slf^-kuOJHT|?Cg5rf#
z&{EA_IJ9c_5PZO1LjZ*Sbac%$P0VneSrr%L#C%L+nRLk+ZNE)>H#PRWjM?z33waF^
zZT8+P8n;Eo4>gSf+*B7Q(8G-zo$O%DVA|53-+yweo$Dm6X|XEBhP`oqk>p!6Q)tJe
z?}fAWTv@_$Eng6lb<NMA0Ha^oDEC`CSDe|=M0Lhjt{0S}WvU2lL?L-2i358A5r|sM
z|4%HkK=1@2NLI4M#H&|YvC!rP5yBngmQ5S}^?@$zp~AP!)yLvzweEA#qB6l=o~2;v
z!5(A%?~#nv;>Hth9%-_9qTT#+VVc(Ojn>QWDGSc{wB!nH&^(SRuwj3`g+h|h6T<j=
za0`p+`-ek!NS^=Al8~+?%92p*L9R+Ig-l_DbM5A*v4Ci!gN3Vz@YGZqt(^wXNInrQ
z4x6Ro#bJFwm8C|3N@atlLiN)Q*<!uCX$x!89r(?n&lS&M9rsA)p}(Ps({5+CesDnp
zmY3GxSl|t>0RI&*R2a1%>cu5P$;Gkt#I7Oa;aH|)o2SmcLtBX#$c1*|Hic$AM!Rx?
z6(X>Go@!98&7?P<t!4uy0gJbJXp;yWEgc=)z$JQDXxIBmJSuD7P&j2V7AiNo0XoZZ
zJ}4-pS|p9XeL7PVskcpjwLP_=i<lq)eNvj#w;}~)>sl!px=@%U218I8Ipi;l%dPqb
z2M>6x3#h7J^3MtYKNH>OA3<exp7^pffk+%DoW%o*Z4C>i?Rf^(`7M=W5mFSPpb3Z@
zfDb*_U2%r1bw0*IqCL$TUFsAkagOm$i%`OvO^CjiH6MnMk8uk5(Kr$%K32<UqGbWv
zdy+3e;2fq^&wk2Sa>EqMtxcC^)7H9Md5}1ilWp^3UapH|9vOZGC1xF=i`PCR>=2s^
zjsg~7S?208q6ZCZKm8^(D<V&|>9vL`G;%ZNT_rG=K+($I+)FiD=jR>49pFpceM@<H
z=PBbAr*IEfqOOx9nwSoiml{way3I!V+RbVM6FB%StW;FBG$UoQ>4U?gpv{TRMwGUP
z%M5zaGn_jOmy73>^htQGBt6cCFaAEnSREdG-P)Nhea^L=UYf=wa;!K}A7AQ0!UC)y
z*Q4;mk(gf<qPb-=DklwW@`x^T<=WVFM%>#!(=YKtA5i{>*X<VxY808y<|7VZ#ghK{
zF`#-&MBtF0ub)0_PE(R$)k;;_A)wz@+46a;qvzyLrBQ}MW@M;UfbvkKI{zuPD&;6x
z0b<pj6_iuJ+?JxZjV4_p8{}93#-?Z^`S|HoUkucIQhV=SF!ZsO56RRz9}`B$xcv%^
zl;g+75kKV(FD_YxX+WF`!5w2cdb@4v4_r@{N6fzc&58aTV%)iVNLvrJ(L40}ldr$0
z>J+hCZ{Pq01vYVottllj#CRH3m64&EN4SeaEsV*%@-P-32Y~o4aA_<4o&J!@+TUv-
z4_T?rP3@~$Z;Fc~-1bzTQmBeY)vR-Ik;?<^J+x}K$&<;#qwTff_C?LlO9EQ8eb0}N
zpN`Dh`pAU)c$k~jpqi(%Subj#ndn>9b^8|!!Yx<|N(ysKQ1g)VBe0)%5xsQ&yfgtz
zzb{Ks1{GN%25Wq6O!I2XQt*OX>5k*l<74zx)W#iKb=9}uh59yz!r2;s%HCECH|e38
zA-~wiahkEh8RsAl{;9A!^CZ|Pz*?DG*SMnPmvq^-Yo>n2#(L=Qx!3dFDX4tZ?_x~d
z-A8w42DyCArfuv$2K{n-Hw*oBwJ3Ej#?q0ix)ZlQJb68RJ_~mDN&Vj7R{pXK)M0n%
z!Ua?p6ZAJT5)R8X13Kxv>oFVNGzqIbP>WlqXyNMcW8JMV&H3teywx99a=p1nH@}v;
zn3Ym~Wo~0<=QH>*DQd2(rJ80EHQ#7!*8D!Y%YJ9EO*>9UbcFwmT758vx=Uj>bRAZB
zrc$3T;W_8oL6{)o1DnYG@sx5^Av;FY)^ZgUq6K#7?P*aA_A^4W<*9txP+e#$Dc9Yu
zlzHl*h6!CeDn#(~Bu+5It~5?Y{8ot-0W69lpI}f!_a`rCNk@T`q=(Mn$U^w5P3_BL
zkH>CrNW5G&Yfd#7SNhn^_aB0P9umC*Q1}Xpm^la#%h{B;{Nl#fgU!jM?}-UL{gW)-
z%XeNoktdnfKAPlZe2_L<Z_@JITKMv)`ocf^own;EA`gM|7mBIB_sL$OhB<Q;Q$Se<
zr*8QzVzxTRL8%|>=Zf<h^M#P!D{o%xR7@Kmdak6A-_Xp|Kwv2p(3}7%;r;rxeOezE
z!m-zP3~FvY97!+BL<9XTxG6SL&zf9EElhGDHI7sa+~)KSDEm=X8l5=x1}mUfu8_TB
za00#I9~CLVh<jTjG#3JwP+z_k-i6SgFi#G*(9iH%aC*Z7(_X_G(i;z+E~?J^K6eAK
z6rNABR5Yg+T3M2a_}qQ1FkKR69XX1-kY<!L<Zz&+HiyKasV^b2nt~!Nqck($+3{6h
z<%ZhivkuB5{v=I7g&Ey6<3->d`+6>Aq^C87;!gc7`;b&@;40UeeP(EkbnIFN?9491
zaq=c1#SFXRboa0G?%InV%3f1XsORC4$d#WCjTUv|&x;d+{2Om9)|1CYW`Y9A9ZXPK
z4uE(!+FZ3<%CBl=I)Ec)<?vhgnfVCf!U1$@P>0MbXm%~e$>s@Epa*zyf0v`QEOaF8
zZm`hH?Zy4;51OpmnJfaz+wAXHCMKf?%4moNO=x-A6QhS_wv<}G>4A&GTlXC=^Ub>c
z{lGL6#<m4>+<@k|W-d2UzIuMbdx|G~#OO%bXHa^~K24Ko*Z9=~%;22;(0Fm`r$V)a
z+hinW5x-emH*{)+g%#x~uqY)_u;;?ZJ>M5^#WXsk+>8v}VELh#nI6JD;xktKL189(
zlL<}9VZ099SoZA4vDc2|{kd>Z%@xu%q|;__5X3j&vy`d*KKOd6tj6*jHbYuXG3Z#b
zI9#w#)Vc6WzrY+3M|_DeC2tYSSQyY3LUX4)pGA@z2t8dJ^^9o9F~6k!^)xwS-jGYm
z$~0;s=sTZRY~hk=`RF*3eKN%$G2UWgN6TilnHd&ALp$?ANS|4Rk3#5zsUrK|_AiaB
zs=h=vnS#1zn0q!DIr(lxDMSa^lNK8NJ$^;O(z1D=GW1e-@Nc0z!2M6EXs53PjKBN+
zTw+n$x1<HOB<j=qn0i31D%0nJKWG7&kQ?Mr0p{ZH>s$S&qaXgmBfs+p{E467aFZlg
zcDP`gHr}?2(Y0tXnwc$zZx+|1TgfO?)jmfbJcBDrePVei-pnWj4lj4=1~<>3F91g~
z=pm0c>~pl9{Py%}gMWUk5NX#*YUc`z>(4hb)2RQ8%bSY${Cv>5(T}uFfXh3W{^^^V
z&4i3<^Fo93bi%o^xr*il7T9Nn^V)htXOn(m(fzesCA&pZ+SC^ae%te%ic*dVw_O;z
z2CFHk!4@{st19cYY~@zqtodOvz40%jrF4;Jkkrsg_-~*M7Dr7Q_gVh^z1OTz#_VBV
zic5`i5@JS>wLeDBu|>#VLP4)5VY-%LU_ZPV&NUhNCaUncfLmaZpfyQ+zWC$1s9zR4
zO$LiT9u=@fN3(O5(<SS9q;d5Y{|u{+*UD1*U_{IeU2Xn~3HN+S(?=w|9`beR%%bR;
z<S;o3Eh)dJ-b=#u6f(bCR45e{ENE*q5m$qfcxjFtVzEa1pr608wSjOvHLt1FhEUQd
zx{INL%FYwFT_*blx{hb!OZ-GH(qKUy#P;^Y-4iDy7DGP`j^Bz~fG7)1M<^)vm1L*g
z8T03`>XCuD{IpswocnXI=sGLlV>i<dv!08;d-`q+oAn!7#d?diT1GS_?Kq4o+`8Yg
zM8yj@F~Zp+5<|q*OY-t@=gsqiKS5}Y*t6cKs=X}0qYmEqZqMy2oypVb_w2qt+6H4&
zRkLSG4)t$#BFem(v^52Kjcjr!LKqyM|7jt{ve|G{{T-5Wb^tMFB_}6aPp8LJ>?MHH
zpo?Tz;DKmxbxJ_ddeLvpjJkz&WI&NaqBEh!a>!>rL#qaW2+_)e3l5M0O8`qP$sj$Q
z`d>}h_u{(Wj65H=X;r~5+zJO#Z>!&MUXh)+y+1khE5!^>GVz^grOpW%$N#cm0u5G_
zG-<+?>1hX+7vBIseb-Y^P8858b<3F2`0bpBKav@pHi``|?ll<!LV0bI7k(~CFjQNT
zoy=uqs_q5{dkgh#bX%;#2%>~5<y1&2B#nYK1#=>1Veh2~Z2Gq|Q-~eN9i7JW>B}c#
z84@mvx(yuZ?G8{tH4l}dn#^!sgrXOD@+po8<&jFN01+Es-4!P*-|YRq%UF=$GLucM
z=kRaTAo$htougWAOmHde*_7S%HM{rlhtz$F5l0I{ek(<XT{H#VKzN1jw62&hW+cYZ
zQNg`ys9wtnGgM;Ex*uZFUQLNsN&*%n@qzEBxmJyIhsCYA5N(_(jIxDl@zDB7@i#B*
zf*<O%!C4&rR3=iKlOB=wBKi`4m^X<Pdn7pp$m^qKXm+I~G4r#Z^NI&}&r?Qc<Y>Ts
zlkMY4$uabHt0~|pIcw01#h?x$9#SKfs^#VpC8mm*`iHyn3LfYAV%jv*@&Zd<<|v70
z?kb)u>TMm2PolnnZ!M_v5G8|-!45;#*U#;1gmwX{xdHJiLO~@hd2r&~Nj|zh=80?x
z(y8Aw{pFVI;@`cc2BO1oQ|3)5kF5GS;hp-+LDE~|8BMKa^ENt-aVvw0U|`p&yB0}t
z=~g+wQ?ax_o5F?UW^CF43UL?btBTKwN&p1d8U}_4MKrXBC(!@d70g)vlMgJ0s;cA^
zrX)}^gdH_fzjV1ZQ4FS1OL7UJd13QUPDM%jv7&+lK$um-gKk~PXsPHDSfo_imhwaG
z<1e*VF}P0^T}pFp`fSw_{-|ZXoQaiEJ^*pJ);gTPIXry(5{Ije-(}>wgmnz;<v_%D
zIT(^6XA)K!sLU?-Db-tG^HfisguoL>t#s{cn3M#h<#XWB#OP}_olP0bmtYgwiU|%*
zb#EO2$fLPM2+l*FU_7;1uEYq~H;(J<NiM>^2WLo+4VTm=s++!y<GxY{A9y6_rwJH4
zeY33}oV&DG68kiwfDEr!@S5mwto;KM^$O8I64XV6PWf{2)**_f(!ls7T2{Qf!^-9d
zPk~I6EzQth0l%|$Jp-BThHAq~Ct#|~(QP0_Aw5$)KUoJ-v7PE4J??GAqtoNL<uhfb
zBm0nS+VP+yvX6p*zSz1H`ATJlBE20l-Jxh2PlJ=BDjP8OfCe<9xCFoQ`(|0fh+qDk
z*2q7d(?HA3e`($e=`e?-DTvKXiEL4S#cxll_u9@Ph`Q5JKqPLdR5DzCLg)RMPdpOB
zXj&wmP$)&G>**+?iI@RQseZpZD_2Rcnev?(ZU6Xs6a8+anI1>XG`*BZIaEAt0-<mI
zHIrB}4@7E$?FCDFr4=>o6OSyXt?2=C(|TnoCj+}2v$!Ax5HT&Ix7Mi6)=;*#wR0U-
zSzmb#{11;s3iJ<I73%<$Wqj?v2>w`s{r(poS&!#b)k~{5CAQ^nBVDV|*%BD(!JnJ>
zS9%Mkxj1P53g>p>@w-(u5h6%gw3_&BYD!-0HQX$2;Pu2@*mB*v_bYytFFv`MgdVS1
zEMwmrx>n+bvekM3`<WW~wXTwPIIrkPI_83Wca{gbm+Jr(pFx$>6FOCOd=-5F>(b+G
z^7c1xN}>YWT4-Fmtx(?V@UDSrGP1Wuo=%z;(r>f%nhe^l$GJ;ignKI#xC&{*5JQNr
z_+D0|dZ0#zWmdyG63)~bCdS34!@&S$Co*&<6s1XAWBsK*NNqc%<y6%7w9C<^>HC?#
zVw4tYsC)j&wh9t$+`>{zDbbfP!jI5SQ^@Nw+=*=UX2$5at%!9v!OdW`tL*_k`=)RI
z6j5KXQ=-Od3HI_=*Sn+&t}<%+Y}{n?_^uslT)^^4T16|K*Q~2|K}jIpt@L&_E)C@b
z(?pNg-gjy-twY!Qa}B5;H(|k*b^qbL=HzJB8<5Hq#N-bSKc!GE#?2BLw_9pjSYs6%
zlCOc?h7>9&L~~E6nvHR;mELRL8HXv#tZ!k-jlW(Iq@AAw14V<ka!5u;TZhfP9C<CB
z+LfR?y<>sq<eA)xJP^uOd<xYwu|{8QV~pNEhd6w~Ort^|qH&|lTh6>;3E&~Pfk4zw
zCMZ@kz*=2?XSIA?<Rm;odpVjIC}~u#4VNqZG`2$MBrx1|r5IAsoc?k)7$Jkbp6R*B
zjyw2HW%F)Y&cRZ<NZuaWm)3ggsNHM99WZI(yeo0@NQhR#>x=p31l_ChwSVNZ1}Rns
zByLaJEOBG2CXe^8FFhUT0J^;<<OY#f8rsa++I$et5vzy<1m8p>^gV_E=lk#XCv0Z7
zKnOTHA)venu39`z!0P%(XM;8Kq3#ykf}3pB{E&;!$X6YqYPzbMud3dy3KOCx+0X4>
z+gMyVeyYrG-mv!eI=j`?jE9nJc167>bGyZ+uVHoZO2S?Hy~_d@Wbf^4imTX!Tok?g
zI~n?>ia$e(e?}nT9=<{zx+DBoH%R8?c7=m&G5Bvw_0v^9YLb(km5sIjKP24nd}zBd
z1sU^c+l5Vwvb{_Lj#rQq37wpK2ao!_)aoee@h#oOeUU0-Z|L#OzH#w1$IQ`>5w+(D
zMj)2K*oCt}V}2NMh48<k*o0asR+SvcwM|(oMEaJ^)Mgm|eBM4KJAcZtE6A9m;d_&)
z&th{+SA_o@MGJSes`I9TieOmch<1LQP{d2^QqIf-X3-*(PZ66FX%NuS*-`@sbcC}d
zK$zX<EZddWW#z3^F29<^riO5%)zDaWxWQb)&7xGxzuv-buRQ%m-kD&SgufQxRf+4M
zPT4DVA)4&PkIn4yiA|$g&fRGw4HWo&(*{k@^(^nRtoShWtSM#cD7BXYzbVh$b!eB9
zzQZFHPjmlSed$WpIFm(Lv!b5Tyjf|{F<|9~(*$bQ3_-GRp+t0PWLZ6~X@1E}<}rY)
zax8YZcCG`}r*UDrFXeBaw`RTOXkAIk3wg@BQ(~oEb?_fvuN|OJy?5?BxKgEY{g>T=
zLUdz1SN#v|#NEsZ(lvBoS3Fv<u87I0mE^vV=Hh4@wD%45SBhWg<Vcq`fl9K)L5H8Y
zxGXoy%Q(EBm1B>rI4%+%CjSlUT>qVO-Z(U(Cr{B%<drJpRESr<S@-0o94iAwk3Hz(
z!>W}}LZOP=u-VVYnQrM`a^K?uPDPUK><P7*<2+*&jY!~ZCc5&zHvi$hOqa;30_$j6
z3CpGD^xTU*mRYJby(9+?72^jd7vT6@jhH{sDd)W8L=_dT^+erlkd>u|E|9CDVSr|6
zUo#;*lTVUKO?W;o$UODkA~m=T73EpsSTxYEKEu=%n?<ijwkYTeVybUE-l_A;B7hF0
zd!&$;l3!A**~TpHWBhSr!_-OLmH3y;)ZcpV-gFLEng?0B^}nh%y{X@)m|QWxO<kZ>
z%!TAKaLI0?rr0FJY7mgBtR&sy)acIwa-NCDoD)jfrqn(xsQMXAmV1Uph1rEU*<LUs
zJTA8N#o1Jx#%Q~7T)Dj3Y5ig<aQ;qTCL7g5<6+SdLKU#0F^d!qoIXh*+@S{!vs9-#
zEY5M4c_{XEH#z&~hdBsDcsh{rBF1+p%RpuC%O=V<?MVR*(AHuh0C*CA=llKG`8|4a
z4e|O-h|PH9)=qESM%PD|TC?@jW#(Tp1w#B-+U#}1^c-48Xa!R_e~!h3!M)QSzQ$$5
z?}ob6>^j?EIek`AZ&7wOQPHB_u7(vrxAqkB?b8C!I@J>N$e6pM+x9tq*Z7q7RHBrw
zAQsX{Bd`}fFQ+s6Cj2$7_Om^-EWp6WMWcsi1oQnWk6<BN=SMZ8!!WXfF+O#!Wd2f?
z)&WhNrw;4w%4C%HRc9WKkc_%XvMBQ1gIMn#1GpND&$55a6|@AflwVk5-WzhE580`K
zKW~$FFArF->;lDSCE;*>Y7=aT<}xb7NqOoyLb2XlHPFLG-MjDA+&gFI1kR}3IIPg&
zy9JjQph3r%?zXk_U1;_R)BIV3Ft_nfMa)zta=g$o%DLd@OUGWxq3Toe*p-9CWXDCU
z!Xq#Dc4Of=$^EUIqrlC%qw*T0?9c)IyeP?6-^jjzo^jW{iMj&wg%=Vfd<xD-2wYY1
z##+jN`j~Mu1opgVHIhzmDTywI%lZA4H3|2^g=_s`&N*(W%|&mlLaEwE+Fx7%$iwgZ
zZGI&eVq={dYZ&TYH}l2--I&8}nx|Fas|?q#82N_g@=QNjMiV3{BRg8yc~{R4-=Q1r
zXuxW<OMx@nJXV`&vyp|uTiPWqsfWyPxAC);jb#B?^NkIhb!jiZvYYFce8#AS@)8qf
zYLhNa{R^P@U|lh1j%HnLIMA&gr%AyDY}no}43CO{qvREkhA|>G5P^maxbWIf0y_I^
z*DhMG(w|bKrugQ6DG?MS%2^f)5vsw$;x`yyB)z?{_*iQg4I)K&>2nUSVYwyB|GM-#
zga6aUVkE<h_Fq}U1+-P(uiRcTQs<YeHSRm<cF(_v{0UJ9pc!jb1OJ5}sut#{)f1FU
zwQXgnv6k=PB&ML~vyJ(PmPDu1h7{C5sF8~qA)D=bIA2MVXSea74G(E);iRCGOug&8
zHKlITgjz-n_4Pq1VjSLH_|CsNPIl^YAWAz%0c?D;>oPojB0mD{-?(hbmXz%V+g37B
zG0?1T<na)j4u?LKQg7zm-c=jCu(}=n2BIk%HH~0n)!LAv>chzAs$z<{Cb`L2Tb%d)
z1>WgoM{iuL+Z)!7*-?H9pz}LmwGxrJ4(KHe<9T6*_OE^6mijXvSO!|H1;ZLoO6%sH
z4&4WQvHYYFRB=R-51;|Y7mIe|(2V2?wN%XL-#h8{4<iS(S6Xv?{d<PpYV;UIf=zE4
z?qRv+KQrAS&xn8L{j7(VX*e*YWYi&J{q4VMKpoOpfV!TZ>OA_&3eaCVoYyXE*aKmK
z9tjwxtFvuiIELM8tF8j*i{>(7;V1{OPG%aVOeGqBFCNo)QfX5_R3qVXh=qK;zTWej
zw+xebQs!GdNIV`O(nt7YvM(bWYi?3U$QtNJ<Sb}LkEqPEn<i<B2Tc8k=K)BbM@wC<
zu>;(~64R6;B!A2|*b>ikKAHa<nVw{6y+(LkL6jYOQSP6r0a&WfN402Fon;h?TM82R
z(QQ-0yd~`iRc!lA1Y#)8Ge9a#ul4eU-2dJj9SplK)!qD?`{o7#cwD8zVVC+TdwOzC
zHS^i9Uch|1e6sL2&`g2Aa&qDDSi<Z0%d1Da9nWN-!kRUZ1b+_zJmjWYtD0My`*+H$
zYM!}zzdH``v3b58hfww^d&2BQjfvtd6k6T-^)`O<WY9_h71>neg}%ChHpnEnuk5;1
zVs_iXl+_QH3qfVc{gmgz4QgNYk*)Pv<}M#gdTgAO$T5S!@{)nEx?1}wmcFS1xd_!n
zD}Ffldh_0Li_SN$%A>0J>?O6k{UChH)4xaYRb_Nu7W;30wv}J2%cY)PwW;RXEeQ#;
zM`g=JMO1NJ?)MusS_;b#t(nzJ#!&z@GeBp!KvmUPodK*b*UfT=3_9F;>8IX6%=8yu
zrS{sdF}h9rYwa-ShC^R6?&qUIZhWERJ7h)Lx#{j8P`cMyBXatn#7-}S8(`<ItCa8A
zjpbm|yFE{rAX#B_8t5S^pyn)8=^XL{%>GS-V0H>VL2mcaj3)?gIBSac+)=jAVDc1l
z($oQ|3bjcdDxG-OBlIJ8>3#2F@5k}z+$)f;QCIY{?uEBwqjZ*|&o-6t_ilu~27)F-
zDF7+r8>NPxo6y4m@)KX(YhL;?A$99*Ee5$_n;%Wm+&8NwD2@bgml}w}<;iUJ?T8R)
zr)+GjJI*Z`mF4fWM~@T>ZFTMEf&2DlXb>Buf7WVsnEekgybgv-T~sXvM=G#JIB1uY
zeHxihQQjKPvhm0@V$@j|wfo4X<4}RyC@tG|v`!t_?u=}>IX>LoaSX11*Q9gDL7x%J
z`r?uiR4T?SSBWN2onvNWs8op+cX@|k%6BaKD7kseQX#RKE$ZU#<9Fvbob4*}!9$RF
zER24`ebxFXu5x|Ktpy_S8(R~EyZs3aE8~9S==krf4>VAcqg;|Mm_FB>w(0|+1r?5j
z03?Kr>3hO;=U#RvhMzFR32=f&IHijsQrHJvmptn{^l70s*-HF-rW;T9gTK6feiZFx
zYl>4nA^lq!91kzC?0af@>=yi@I94)!)YJ*BP=ZTCh0y>*Qw_3RA=8Z6%J(eN0BXD>
zKDB=G{es496An|+ru#Z#4!a`pNyD(F$$TTzZa*ycH8}mQ*)VyJtC%ZGqIjz9-c-FO
zig3oA>#yB2M<)EsUP+q41XW=ZclNN?iMID4g7L{QO*~GO8392d@@=P!pQ~qH#F`U}
z?*GPpC3CPo{f>2}T5sV#-xaS73GtH+VnUy<fF<MbQ4VB@aBCsMxWx({QWyNQGeF9d
za|xg|)+8~}d48~cFlT@6F5ItT;qI4_ukcZ-dmwyRE=<34&OqdQx|C^CzLsT?wY4tI
zH$I~X9TRcJ81^Kt(A`)yXXz51Md~~P)#M~bTkDPEXFjJ<D_nuzqGFyKwbsJMwN&?7
zzZi1+Mt@)j>nsI<I(Ju?yPOH2TJKA!&O3riM!J5>38Z>1@e_c~VAf3xlvmcOqT^-0
zZGWW85n!YJFP8ss`F8JT(}lHF3S}xMYfgFRniP(M;`qz5%-V?7PEyHHNgbhBY1qSb
zy<SQ(A~jzc{&NSg_Qhu%7!!}lBHL2Tz^6&kM)T-fBCi?K79FSIS1oRFtYW1_i@a{0
zM;96U8VzA+%#Aujiyy0%^02g4w#{Llh{rG1_W4g^h_OsHonaspV4~2>a`6wVo;AsJ
z&4T5%lCEBHc4|?0(6CRVVQ~l?()G-eV8{+3U750)B3Jmwnl%lK$2p&66coNcU_Sf3
zutNgLnKlC}U|hZ}H8wOXzBsmu;uFzu;_nXi4Q@F~T<_dTWu0)KB9|0yNkwuSQ9RiB
zezD(BCNaWWgap5j=_-|=SF0=DN&Dq#FZhR2Ua#mDNo&)_Cit$5>Kl-tK(mLaM0<yq
ztqbNGQiyTd`k40ow>9=4bfW&wJ~hI+HFzq~rtnnTt`Lshn>%0Ih!@eJ`%u4w-!~OP
zSprfA>|6nuRNL?c_z!6;<`%DEt1rji*rs}+mYz*YT>tAVKVJ#rf;4(ZU6lb2yIYme
z`}os4*6JhxA%(6aGT9hPlv??(P1(erMl$ks{Y)4f%A%O4$-7(8Kr3ZM=zL&!k+{iF
z@YUD8n+{g|tM~${aad;2i?b=k)ok}__>p`4Ing`GF>(M?u~hbC9;3EY1-5=2AVF(-
zMUJ%<kb1VaR+-{``GYf#MDLBJ&l@dmc3veJP1%M~k&uFt6Ec!*I};NH1eWCkzDA)X
zKju>fE1<C)nKJyiXi*lp!R$|!#)CcE1poOlBLjVRlQYCPLTg19cI{p=g6uV69~RGv
zDljc(^ze`URtl)V94yxQai1+^c)N1z)t{Ng)CU|cyR0vi5em${xJ{6B&I^0J6gs(r
z>@PWCuSFPPd!NcbleEZ`s$cI;bhRxXl}w#j(<qFv?$-XUpahBQF5FhO8%&@Z`2Fw2
zrn!Jv3g_1HC>yv28QB=>6Z5@sKa6s?B{6bAH?d?q%aH$jRej+E0f&J7P8y9B{+@UN
zFh(?quEmQh%?0zhGM*o2a#VIA<~{vKSXj?BlkYmJm!q^A_yaeSEkL{pal8MoDJ)4z
zpRE(<aq~Qy{nShLhMCj|YUBI^t!=5#(o#F(>vQerelPP3J=hRn%31hr5vs+XeLgrh
z2W#P*Z;{KklJQ&N%DF9VVXvQt)s^lPgT}i_v$FZ;j2g(h3_NodwM$rY81q>%i#$8<
zD<lsCGjF!5kA?(2U+T|N!={X?YMap0BP*pQ-dqY-c5nQe+Zbo%B7^`@eOA)N2+QXE
zgx{C~^TekZ-Pw{gOjszyCm0|1_$Xn|85a|AT{q}VWI_DIN;ZbzKfI*ZWr8~~{7Zfj
zrr)L<gv!4qrmhTtjf&D7NVXG<(=#Oe3ms*nv`rjsbL?wfOy+{mIJ4xP;R(A)ndE=5
zc_Hfii`n(KzUwQqpza&u=biWb!!76}AmGE6%MsGMpZ{K!@&6uAl({sAKZ^|Eg;0iC
zeYod(#>|{Q5VW=aAkei&N(o5`t?f_$co_5IoJ(}a=onkpdUfOf<yAi8c=OG_506RS
zJmPLQ{q#i9FDI;9Q}m&xuOtj>8XzHt_cp0NE~`aCviTAZATT~~8ANiIlk)>_gxSWV
z16tJHCJ|)%_TEBjB8b(PycX{ywV#U8VPwRbrMJI*^7;M45Jy~hWf)KLUo;-b2u_(b
ziRm!o`Seh18#}gq;d^twwSbpn+lgj8^;7%X_USpJ5OuD!v6us1`S3N#9S@sKeB(Xb
zlg7Xt1iCoKdpi~2?Q@qa)8O;4KuWSrVIB_&5)in}F@o*j6(%QQWiPI!J`td`?el%S
z94jY2JeM}gl`g#B&8@%q7Yb^sCzP#X9;+r_o(t~gCv{%md?3%c$R$k6On)$tXyjNC
zYR2aeJGmE@sO~IV=jF7cq(OMv6&VMqp~hn-KE9o#iWQF)kfyC-zG8TUXGa?+yF23#
z9xEzZ7qv=wPQ{hu=`~YnyHDokYCX=$qPXbAXegTM>|b}jvM!<eIo`Be(JFl}%vn`b
z!lZU!c9!wpa0%Y)aoWV&KRuFNw#uP#FfsOj@!+GIK6jgDOB_Sz<KhOjt79o2`>9V)
zJD$lgNr{r2aGTchgg^Ljf7UsaY?=1Ax@%j8@Pe6Gymz}NJ0G|<z@%Qr(FaO?cr<n8
z&Lx^F!;5F;Q4wlP`0Jjlsb!oBk2C#e`n&4MTYSc;vy1Tu`lQ-aq)bwT1Gdw1$6q*p
zGfFLq{Mxws^2%5mJ77H<*nGY{`+-1h7|kRF<6&!ve<1Eygg_DzVx}?i55HkeB%9Cw
zn0z9P#L6;m<hku0TpZ%*G}{<s!d|jw|NE`=gr#%f7B7Gxy6#HRc=*c85p#|AKNQcp
z-|?iz*xQF6D0StP@*KAwg$#t(1Fw?bFrB>Vd92<}c%*~DEv>`Nw%M=xl5S--cm~#j
z#@;06Oo_!Z0ySzUTNT>ZJKpLAs3=Vzd7i#tRh?S8saz!>{CdpgABq(mgbu%YkB7g*
z%@FjP(Z%>q)QaF&_ZX%_Bj{_|)=ERlh3~f%H|_J|e;*-9I$mat>@-hQZY@sl6!BXB
z-}UtW-IntW71iIk9>T<jcodY)Z|Dg(_RgQzH(SH5-Ois#hRi=Jm-?U2{5Ur)myfK)
zrc=xj>FXN|tsD&a`0!idj%{c773|i7WJZ1jEU-XKB;S!<xI6Xj-`~fKn-M>QA`#A0
zhjRV=6_2d~<%23@c>8)9{LB*Pw1d3FD7eeU1P|3vJ09ct*xW~Cl9PTmxe|Wt{Y_Ng
zaGP0;S;=J@&oIR@Q4;yww|izC@0_VmmayHvPYYz&J^}Kr77o$SKiNJp*_S6<XUFSq
zGi}P43(2?+Ty5nP<-gJC6*TE>lK1CXC$)<WtLsY4iY2zh>+jRhk_zqj$&nH%Y<D(4
znoWc$q;%HdB}$AA`MeVdnO1Ze;&Y*gucOrAcgqZBpK7|%--15dM9uLNoOTI`w4lM_
zp70H2=%9?8kWs2y4e(u8@DqkVyzf35#45=fXu<I<B(8JH9XPN;e>Ix-#{@93FR;w*
z{xM7);Ax*L>&9dE_qr;u*D3e4GPTlH;#Urd_Tx*%UEGq48n6_VSYyp#jUSEh?c#sQ
z0eam5lzq{&Bovg)Py51Wd30whzsD*{f#x`VIk`5IQ}sC6{`__|J~6%~PDW2j<1*!k
zHpc9kfPs1I!{OJl98&y1L-Enue@Kq3vYF$GZP#UEG1q_E$@-?4F&7B`Vp`|mP8vm=
zg{jn=_FHjmvmeh);IS4vEXnGAt+oe0PbB{~HGHNTA*mT!WO;)qvSm&gOUTMf@;~((
z9O+Nd=eb;b$}Hq%HWKvg@o;qXi4Ys}2J5*Fk8f%SUuo{5a@8u6$LKZ`ne;=8s|<v%
z!{rzh0i8@lopt7cRzT*YS9da$!Stc(#+oZMLaPBj(S1=2Hf)YU43blhhL&o*5e)au
zM(Y55siCYC|8jSXDrT^uCM7MH3tw0^?c0FXLH!6d0WI}<8n=Y(4`6xMebJbm)bVUH
zEAC<axiAyjhf_QSF=L@?Y`3XoKh?OZhtqsVpgN<d)QWOYf;+uV9D&2F%JrG@hiU}%
zfC?U9ii!yCxV?Na3er&z3U`O19>HxxeMPHD#1Cl)T5wg7$wS&rb0O!*S+}g;(=oap
z2}am_uD#+CPgCHOL=e=JD9D5|iM2%y+E+3iLH@@D{8soxH!U+vst?S~;mp#Du(w{%
zlSh+=4=PoVa@uqSP!6jJKL#sAa1R_q9s&?4l+5IQa*^vb0FL?6RG`s-^$|-hMbgXE
zFW2eHB?RV2=FO7Kkb*=yeUZA*$jIX2enaQthT+z~Z7!pw-Kdd#?Ts1x!A0cs`lPvX
zNtX&ubONz|d4;ak4mu~g`P-G_Ey7{Hfd7T^tF>X_#|j!VCe``%kEY33BA(Di`{gql
znoiGVCicDI;UN_i3s0OlOb<zB5BnMV=P1GDUjv0vrHyLI4OqC6Z0zNx*atkkYlW?^
zui3(<!_)3%!ry%y{q^5w1c0hxbCiogML*v$hmJ4YAcWNtp>=7Uc8kKlZpq_kYUY0f
zI~kS?Yn(8V3lh@Hb^E0og+)vX;7WtMNr2u-f$YMu=#~PUV#lre70ziDy}C%}0ew28
zWv7w2yY2GwhiHnL*#F_~z2n(#<M+|f>Y(F6@z9pmrV=fswMS7~?Gm9?o7h`~R@JIK
zQWUjCN?Ie4m@RD)t5y(u6PqAbPx^en=bYF1@BIDy{FCvzlh6IR_w^pv)p|7NPd3fw
z>MO@O-67SGkYE#w_F9tkzi~JZUK1f>b%Qwdcpm?chN*o|UsgQgbW1ht%v|4`&$x&a
z;7{Q%s^JA*ihIq>VaQ>gRCfvGSEnGc-uhm#S)X%?T_Ha(N9N~b>G2AtY=|b-7KDg>
zV(>)vvxs&q*@Ia&>R)5mCvSDMQsNA0{|P-TF(d3>#B#^*s$NP-OJ-*A<sHzw6dhL<
z_bx&+L*1vEMi56to#V-Xcdq9?4$CTE3=kVHnNe@fVtSKrR%!4<*rYQ`|3|mYlNj{m
z4Y1y7Dn<Yc8^3TubS3_1bTism9nY$8H#MpS>3LVK^_K&)5LUaS>R_*nVM&dXjwcbI
z+MVr9o%{;Z-O((ew!P!lYg^=h<^*q3UyaeYBl9(N!Id;RLEXEe3JU^Phw^p!nNbPx
z8yOAiE~C)&>TYjx=kfd8Qu`O&=U>)r`g#5VT)h92fXvKFD0O{~>1O<7nG*N*C{!v&
zOMqoiBMqM-VVJ!7^k@0xN)`w<<WVc**nf^}Enyez$Gq=3K<MpxfPE7RzZ}}Y`qB+E
z(_u!K4~8s71$x9VTSi1n-_8qZb+mp5oPZiu3a@eImXvUzS~D_C92|mc>l}&0(wPE7
zhR~c;d1~d_sd;>wBmNXBZ=V@4q7o$z(&wIm^H{Ll(kdR6j?Z)_Ro&KF23NKh4a_#p
z`tkyY9y?1cs*>j$0)%?@M~4odjFW2^8Hie`s<`E}xY#&J&k{p&&4_8RT>;c?&R&o~
zC5D-Y|8ZK~8^81vvlJs=Lzf4z&8A~0bQnD9jTL7(l|2gnc%;sos#>aE)VES|qyS>l
zvb`;hxF)5TcQuXUn}e8eyd;;L)@htu&07vbr)#Do^%x(01(;^g-6WM{PEH;c2??dU
zTMrD~uonx*KDb6@T!|ed@cyiUUJl?NGP$<{&Gt`Lh1V)LI3Ux_os;=w6tyzTx>rQK
zf-`gRxrLUU*v1NMn77B=ixA87t@piURH~w(B%arE7cX!Ev|D%)D7%hzo%jll^aI!2
z%eWPw{e;<dFgVm!QsS~;TElp7nzI;Ye<dr}<PM^~B~88897A@jcDM4L%+Rs#m3kJK
zrE&Y^LZb*0Ly2*GRzhQt6-{*QL#A*_vBw!3d<$=_K&m-tKLpAqZ!>s<1?F%8X!E(*
z;!*5xLpWmy%AG4vm;!2Q7J*HE{h00#WBuiAv%mN#z;BKy*}r*3s#RANdesJVH8Vt_
zgC-eGa_!sTf2ihQ|C`PFO0~xmnkm;<Vpg2o0i0{i1q18fxZRb`xH;3r^AIC;ld-!h
z(@U;W%9(US#}8{<iGOSo{XkY<KFmbQd9(~-oKAE<bG08O@Nc#wXXc2cm<>@&vk2&5
zv6CT}nKrM4k)EJNI`gWw0COfMrx)M6tpK_;>U^(9R3+u^He-q9&Xi?24)S?g`q~i6
z;ITooY1zw)9MGrZqfOu5&ftX7oRE;2n+ZT2DKA<w6+(BIT6|Cn2m2#T1I3dmv7yqL
z8n3w#Y9-%_jPGJHY>ijG@_mb7D>_Aa6{tCXSd&hff0w1LQ>~KjcR2P30L;ej>ef}h
z;$Pa?9<6o^6Fb#m+4^;DJGZ6&K8k$}<;whJ^0=TA>v5BJtLRB3Dd0AjGqVysQoHQo
zo(0e&ADd~Er2jLGz%mES6q8uxb@bJ(*0K3Tl*gXvWYtPErW;R}T32~h9|&eEt!HBf
zJ2n8LfdbmiE?eY)&sj|AV?)3Ex=S93l*^eOf8Runn?sG#Rn%3Ho8FNi%sR^i#I|Pr
zS{~ilkfC=Eq%&>_0oU4Zj^&j5s9)6HUp4r?MnZyUG=O-CKt%UIVEumpOe2_4ihk-Z
z_qln!=?nC=ej$gh7d*e#cNocgZlt|DEz1sH^w>OH!vD|_X_D!ODRHf6?V5jIc~ln<
z@eAC=S2QZR+Alttx|ab#`05s!fWStycxuv4tltHD?<k#eLV-S?z>QAb%x#<C9I<$Q
z_=BrZSq-(2lS^t%{@Yj!?~!XVGZSwDY&To0pnQ$v30>-w7v1}>Et@<)p15Lfo^W^Q
z>O1Vv@7g(Vw19AmnucxR5X&U$gF>4F8AFEOIed3#0VRG|UY>mz8SxzIZ}3?qT`$k#
zv&Eo@)RL`kLLEAvZHh$?lAa9$UJA`D3##ygwv|N_l5eLh<pyn#TmIFnhjHHiXiD7`
zy}?!K=q9C;vX`b@Q_TwhRy1(rcw;GU;6=G}ftZFK`2GEKPU$hD90L_oL-9NW!AZfL
zC*oqV(+T%au@!9<PyTz7k9O76hf1HyC%*Pv<s@#B2%k^ft|ZPY*<-BzgND%fp37E;
zGd755lTK5?x=AtHiQmzE&*L>etS$E#-7zm}=0*U2O<HKB-%25`JM|~}s{5PG!Kw+r
zTqyYVD^cda)vMhG!Kq%qolTES;u+N?YW1T`OWS$p5i|M@YK}Tb`(5^I#xE2DJYHO5
z7?V10x~|Lho#0SD)&U({ejMgv<Z`t_RG?U|D?-cmiGvQWA=fRf<YZ1KTe3LMyUo1&
zp`Lc!k>KwQ5;&J0Vx^eeSJ$CmqYiRIvE79pL!y<x5>j=p$UMu@hlAf&r9Q0bpL{5-
zVfVJ^y{iS(t!fmGJ=baIc5E~(av1|e!6!uQMzut5b7gA^b)*{BAUgB7l(aJrhM}q1
zZSf24D@ZbG0_kgoa0w8F$fa?AA7u$TAg`}-66AmwOR6;~$eMa2s7&qr;&?H~Z{^xc
z>zfq^^Kv3f#8;N9K`Y{)$Afh$MKfHz4n7)v^jpZRo9!B`mg3A@F(v9In=*^1pX?A?
zySV=V=x>SIJ+C?P)$9}58g&<Y{A`3KFZ4B2>uv_$&zHceRv>?jFtq6IB`x=S{uEwj
zLsRq6R4H<<W}c@f=$_?R{YvHU%6fQ=kKvSr@m~3<D1$>5HO1u3%BR6s9q*S~EK|#~
zBqt0r@SN`p)1_%)l@l#uS87CoL+kQ4zmse~clOVj>cCrSMcCLgB3b%NXe-TEfkJ^~
z;dm^IU4Z~|az+7Jd;pl!as+gbP{TQUJ0aM(5!Quzf&&t%mZfdw0|LdR&yKay>3s}-
zB<VE_cu4ElR6VK4Z(FY6Z%;4*Z<D&%ewWpR_-C@jdGtHi$SNpQzGDzgqz6i5U(T5`
zS7*5*_#uw7aX|$9zUJ+waIc<jmpXX;BHGMz>N-d+HZ@?JT_6X7Ulw-8-|l{najKO)
zy?sT~3d5@NS?}$sD*f}cxfBcGDz4=spvNQpY`n%W*mCu{?>{^b8^64wMeXYb)x|`P
zgof(AOGx0a$M;|>MGV19zaI!z7^Lm$k=sbB9bDTJH;1bPDNO;rjzp03#i}t3B+quh
zSiK4g_U|jq2QL?cZ&*SvWJPYI=s8E~#IwtYb%^lUJZ-R3vs_ktBDw%+gdRnVzNI19
zKzs#P>T~C-rpUCTY(3#YlIbPG5=^96jHnT_K&eqACo^-F=KJ;<1x+92cy?AqRSXMD
zqu4?u^VX0ds4VugS$Xkp)n?ah9Q}D;wZscI>|DwV)Zev7$Kpf_6xf(erAM{~k5&5%
zI!%t$t38ZLfwdCIv`(PJ`Pzro0>E0$I5tGuSUkUr3ffOY9qFQGU-z{{kkv?ja4^t0
zOG2zG@Jkjl#qySBf~_dTM@_x=-eqSk&ZnQ9okg?)drm=d^p;nI(Jo`l%Skh}t8sSt
zd(~AoG40;te?v#CScKDJ*ha2^9*PbOjwG1cJ`IR_uTcg}*Lj(K)?JdL&4ye}9c*(g
zsJi({THY?4@>KDh?GLw$jhLa38NTC3jnaG{hx{<0r+hOyP4^sbi_;BU>AcmN|K$1F
zQ=wH2b+i3=mrV^Ev#C4wUhpVZU^w5(sO*M?)RB*lEd%t(PRL~|(+9F&VDj;$F(0PU
zAP8fvYKmO7(siv(XcrIfh<Qh7gE*zE$Fm8{$^aeJ-Vlo(m4gp+hfJy*)8EUUub-2!
z-|64Xe3ub*|M30<Hy3Eo&eX%H4YEw+1y<{%6KX4vLxJX3NbnemMqOt;V^<?uT&F2=
z9TgJvPdVrd((75pf=kA274Dtd#aCjPAKZj=By3-%Y+J>#Ly~7P)}jT)8Go@=NJ>X`
z7AhG*AgCpDe4rtm%t+tJ`^$fCDQaPuY>Hpk5uGRUf{sfx8cnj9*-jfrk|rh`L>nDV
zz9`&#l1_H^lY@>z4Ta-k3sPlo*3@c3HC7}G-|d)aYcYG>uFr1cURf)Cno&G_TfhX-
z_6P7yKxt#3`|V0*?hxA--YVG#?G<=YmvU$qnd^7x@tB6U7)2y?aOq|3Q$=qrL3^ow
zYny5lUnhfz%DB%lr$xI@UW^G;yIHQ?a&@1|?|YTWA0I0q$^cB|&xqkuW9QUb8EUM|
zc~UkYY0O%wnVKpo>1DJs*8PhnOG3a>euPXmF~)4kJrWlZ;r$UQBqA<yIc$Ht=92f;
z&e^=+SDcSJ7P-K_vg+`()cGWGeK+BrT{iUV%|`RsEov~&Z^|YamHXApayWhJMm_}7
zNWb(hb?5<zQ9<!5r_h0>Lq50aKb4I%5lNYaIo*yQ>+_2xc7>J`HAR6{+hK?N9A8&I
zrn`A|5p?uhHI2Qic6FP#bXB5m=PjzclHmUBbKv37Ovf*4uBwBW_2-+y@%z9r0g3md
zpH}hn1BDOzGkGS+`P<V{kECRjn7;~@IxH1PiH)TCJSedq#O_0GZ?hpY+8`&@Dc%wf
z_i<I;-c@=5+tOy||Gnxx<(kb<QI9K!0BO;-sRj?{t_z4u?7BJ+=cqg$r~nG4*UY(U
z<vSUfI2+wa)lPBH<|8KVx)hq;4&bddG;GJ1;Wz|biqmh!Wr@0(jEC8xtYe85?PV1v
zCU{X%;BP~36O#^O+f8{0C{Z*2iMnizcFFHg(zSKbCS?zh+br1YrWFqg9uoL#^Vo#5
z`3$!uACx8YaBjqLKYae%{IEbH_V6orbS7Vcnolm-AsRujecmPf?9<n~Es@u*JTx_l
zVt5~QmCeUg;{n&8=~#J8?rulndR*@|LsgaiFr0t5=cMBJ@Z<16d*Wxx{?4)l>O?!c
z*+8p0DB_!$n=9|4A9e$*F;d6l_{4c^f)Qcj)i_nVJdw?|#9GpDDMNi+W?HYOgB<d4
ze!f8Ikyufe6C?Q#;L$4<F1bd-P=m#+<fL3|Q+ubr3k?o7&?I4yX9l!PPjnP(kFnEU
zmYAAy;lK-{7Z?ZV_mU^^7x)}1q#izv9ndZ6RUPcO75r<rEu^6B`Bz@GM9zI;3};Ua
z^NG~t!M%^TP49z|g~$t68l=A=A{fupcrUI9i@809HVwMoEqqQwU&vqhCuNu8GsoW2
zX{pDI`^CF~UJ=6}SJCc9i%9P6nsrpAg0!W}jYW|5ZHc<B%gd@sF<-`zKhnmXom2<k
zw%ZH)i5R8m^mV<fO=p=94B$|RFUfjYGGbrzkO0gt$#T97#biPG-;tn~c(wm|n=Eii
zi^YhRjn2z^m(A2aqSea3u{O|{y8x0wxwcfO`S4?`UaQt?%MwLr1tSk-<kM8PR;_KV
zib4rRemmQO#h4I*qvoj>3jW+TW<q8Ct|1Ltv)LK+N1{!R-3zbvfBKo@(IT*y%cq{v
zbzVt3s=55clIi_J9ZADGX%3FX!S20{kGAKwE;b+dU4fW-G;l{5ejk@YP56>FP*M93
zZAzhR6!MWT**ZpasYJFvfXka9Xj!LR4eTtb<Iq)<#Zt)co*efXLkpk_^c(uNd97xk
zyP_2QB>r*wU1$Y=4UiEybXU`m%OssSQ$&<c3vrI`ZcgSW*=$ccNx?^^qnnE}o%<I|
zD_cWZ5LSqH!<~!u0!%B~PZmF1az5zp9bzX}0QJSXb)-0R@kHy({c9T&FZ_dFyY~St
zO)S~IsuhuHG)1JRe=qxv_-0v2cOAsneYqpaE#}MhzN>#gi%(KgLAR^T)O75c)GudC
z?0aeF18A)>{?5~+OgGiKvcc7H4SRvwnoOVul7?-(SxA-A26r!eCe8L~PZ&kFV=ji5
z%$RXLD^V}|eSha?Vegw68Af|hKXzqjR#wefh_@p~-+p<6nKKz!oUf)4Ed1_wCzckt
z2rlg<Pk4ru8{>nILp+(;*bF5}e!X%_DdN|yzLCw@c}C1(W!3Hk&~MzLihcW5@woIl
zSFIp_fO)@coB8UqV!7LhTff~O_V-18j?vc2bh%WiR*PuofF6T-YUE#3$eEEaFwbn=
zDr}i=`2#JT3)OZ-Y|{MhAF>{?L-(>t*@TIKAj_%YCg)liQP7mhZNDvrvVLt^mThIf
z#9-pPsMh9~faK-J$FH~X8CE-|7D$=rpI^Y4O<gu+4HQg}FMs4of7)%cvm^J<8;*w(
zFSXau#9ur?e3pceLpSA6qjb&UzE4k!4W0X(AGSv?&;A9UEfJiJbLLOE`87rpIv=IE
zOgPtMlsJ}@FfKr*x6Bq^+m7yv)-vf6YSv>Y#34bf>?Brr&mujBF=$j2juOniCOLF5
z;mKx`I{}PZ$qVWZn6M~Ag`DfKGW$^*&_D5aMzAauD$a>3Xw^C2zi09*Ct*hJHt~f?
zVMp}b`^Rc?_rMw@8EQ3LCYJ9bGPR5P%S;Z_vIL|Vc+(k*ensjPx5gX1zWO$_@WIPA
z=!f598qEwCuqx|d3Y)39OT@47;q;1Nqsjb<{RPIAf{s4`{(g|CF5{4dij(v2Wj5g*
zTiiDdOJt!0@%0<glAg^E3M^eM^8d7EA?l^C{qr&>-F(6rzB-1Hs#QVyx(cw2iR^W{
z1*wh2Xbap0rI={g6-Z1!L_<b{Pw=~SN7X&Pda^Tj-I18#nyTR?-IZcNJ5new$s^Q3
zEzLJh_0fCmaN24?CvWBY+3gA8hq)aW418E?1KbwxLT$Q-AugjH-)><hy=m4m@jz3U
z#j^^XF>?8gT2?yF*-YHNmJNxBbeYIP23&PmN>{AbkMk7FbQ$x@9L_;%Iw!am*k$i~
zHcx&2hNmWT(9@{p$hX|Bk~!R6n#D<;_f#ms=S@1;9t)gEh!&I%)=8*J@sh9ES0T6M
zZ3vb0ujVf}j|{wh9ul+7Cn6dfH#}Hm{xJ?8GNzs^E16PQXf66gJpTgBB=x2X`o?zC
z>A1|ve*QJH8!xOGU#Q<&el8W6!OhukVkvixP`e_%!w=)L)HhB$>}nS7FH{&v+O5Z*
zu-s`HG;Q2Hl{qMV=L#jcUx(Izu4g%W_e1fh8tU3}-rQ`@$sE$CRcHG&s4azkmwENK
zKn($<!g6CyOM5W)LdD~A(&b)iGgxwKxPs-_H#e4QnmkS6+dzBQZq<%9UF>Ha%S0eZ
zC<Z9;!e}&96lCXY!Rg0&RYsz54Vkl+c|u1q@)I2r;o}9VS@n`o)C<hwuIQUn0eL*-
z2=w?jI82(l0Cvn!82Z9``+oC%0&1Y*$Tavi(PgHg^GBYQIwS%vZ>a}q9u$xre%n>3
z6V)dl5~srHB*Qg`y*fsM8H?4q-8>XZoT|J~%h~wr#A7SfCp&dbnBZ^e6>{Ah_xjO^
zP7k&1L@U3mumvn`@=~+HY$|2``4;cphtbwc#?D>!qCWjLI&r+pI4OxuwIjcVsAWmz
z!dA?(Z}53~%NM^ze2>)lKObZmI)53Byp>UKY|kpj7kKP<>t0s*)Mjtxotg3d_|g!g
zi2`f(3HsGSrK3~9xguk-FR{+~=rh;Hpa+Q7cO{c}&p{K@QP<LIn1F~Svji*eKY*7e
zjA*+Txir~JJ9Lq{*?8-*S+XHCn2hpyX*8DfoBWINh=p;<gOE~M8OA(8EZ}kG9e90Z
z{UQ>AVM`2CIRG72uZR51MCRcR^yt$)RR)hbfoA*|qtF#?12jY<xibbj1;LOPuUl7t
zF}~=zW7zY>6Lsx#W+$Ka;#h_$&ZEoHAd5{Dp%-_jKg_4!Nw8|I8RC9%;O#}i=4X~~
zT?WbA)_gP8+%_q*%z|R|P2JcH5yPZZDORufA?KYu`^Gt9;gh%VK}F8Dkd5#^0LAGL
zhQb%mKZMZvlCE{ty5U{lrGxTZz<b$n);Rrol(&>1zs5&PKLz$0ByExBVK%f{Lr>M;
z2^uSwMOe|ekjz%Txa^g`58yQsNq?NYG_a)ZCy02Z8wtd|(tbI(iugCQ7E`mTt%U&S
z@J?{dku!Q=0Uj573#N0g7!M<yX*3_tAS!%bBW72TY;7B%NY?lL&#L${nM<D>%0J`)
z1KU4lkUNL*LSVW*U+IhkBbu5^Ut0bp6_#gj1^CnQFYCWeD~gU0p`9UsOyS)JQ%X4}
z7+9(;J5|(+?7Vu=#Pd8bTm7C7vm43w@YCUV&<~AzVH3ie3}aidoH&}sD<4fE^T9Z9
z(MHE^&cP91!1urqm}TLf4Z<$Q*dHGz?bkh<F)mOq9Vqw_L4U>HRyE@CyF#@nb~$I~
z=-;pIr%7_lT@?UI5Hz1arD@fn@$6WD*TIJoNe-#g4xsX}{fq6D<8&>kor3Zon-O*3
zfT6*Pg&}t^Be7yIBRKg)rh3Jr)3cUW>D_95(Dr#eaQH-$Fr>19sX#b9PunRPHLJim
z^+)&H>tcj`I~nBD-ewu0S&M(f+#1x91kS0TVzy^(QR&_tti3}ZS<=o};A+XOmlN(w
zuY-rh<fJPC`?C{<KxE|Do)%F+sxA?vYoc7y`F35K*S6iOgE>|{qa*4gxWXN%ZJ0c)
zhqd96TbOXys^eZ-;52P|flqk4z-BatnU14H3a|EkzQ>#6-P<s2G5q;vQ`y)U5^{fi
zRSlf*X&~lhEmpR!?V-t{G-mcER5Mwm4axjLD>+p!MSBgq3of5pui+k%pre_sV45vF
zHyW!Z5Nlefg6Xy^H@?^R;3HEbEpTi1Y->%7;@0&#c305afhi6SYP*%poSBFR751AG
zd5MEO#-G50H_g(oMY{cVmZq@h{+6Uqep33RsDDO6!$D82w8cL9YxZE*0jKk2n(uT_
z-5VVMaF(fwVz6n3-OI_b_ZeZtd+s+z48K1raFw1*N9li>s8;^<qi#ay=Uwe$*$<||
z_piF?N~-eTdg8^<k8SKuPkm=bg}ugAAR-^zL;+dc<mJAur5J}Kuob&JgHG&~^%n~a
z8NFz;JV;C1WErf`6QBhk23Y-;*#P5Lu#T~AB74BIjoZ0X_j|04&!)U@aTN=uTE7-d
zw=7jg4G3b~2&E>o8hICtHS8P)KX{XZ<<@gQJ|pX>>@%cGViPtM8@_FK!p(O`T1Ylq
zMEG+_{qp>tbo_fUiE6%X=XJl8tfYy5Ge|`+EvJWNi7}l2mXY1ImxzIivjcinW1P-4
z(gV)?(j;fvd7B?=^!6^lmbOs}fAQZ5!^_+E*3`83^K0_(X{M8<y9sygI@p`!UrE&}
zuUE<MU*0g8&)829WhzyFk4((iX*Nx9$89Cfc-^NRDnIQT0syf6qwl(6KQ4TD#d3f4
zXR~DtI+l04)(LS}eNs|#`xWYYM%nSt;Id4kOT97MT*Efjcrf)`vi(c=&6hO+wlfQ_
z?MJ>Av1mVGkNm2R<w~wc7R?JLGYaWDODaqv!UBx@)J?57@TY^^3m|@rZM$zdT=FBv
z|N4v7d5^lQSr`^cm*gc1%5{JZzU=B4yVtDVAjKm!E%gNL1iSipq~hLukklI(?-Fd4
z^-cjJAG6;^{nMZ0H^FC{62VJM$Vzq{>VmTQSN2HjpqaczhqWV?#qQQsJmQvwKDeDG
zQ1bjQZqd7geMX{9ZOmH%Sr((WDVotVK4ebQEv(Hg0wO9F<}opzD^v;bhopuoABY~*
zKv7K@4#l@6mh+8Jd%H$;9yeq^Tj;XXh-0)bTT-@eMVGCmvl;WNS54&%iF+tyYX;kY
zeI3a8s#njl^V|6Sn$>tz5I!CE(sIx;!ZCl_-Hj2<0$w`4XvB!e2~L@grjVM8jY!`f
zqb&}3OjeQ#9?EK2-q1+r-;*#h0hj!Bx$yBx&@RO)X0~w-_<AOYu!bQxHstsbWAm=z
zs(xpnz6T9eerEVaepiOY`whArdO7ucB#*7SHv?C-^MV|I&B)efl7Cs>NoX1ziGy!!
zCT#N8Vw`)wpE>rGnGGTBjk5DC%!cUu3#yO0MM;GN%f7=kKq>dD67tVGrfuBxJa~S1
z{0bfgZtDkChCH$fsxY0xf$C4z>p`$oC|U-Ydr)N0X2@*puQO4r=;WP&)jW!ZN?9Rj
zRI)Z7g!#SZ2w%KJElg#yT|Kharmg;2pqUT<_u@~~=Le+au8_Ov43g;lZY6dCP)0!5
z*hx6!jmm;eVw5B16XTUG$|hCDKm39Ntc4Ys`xatVv%gQKaZ1-p(8A%S;mQoKuEb2Q
zUB4g-K!wv4S9(pLmox`3is;*0rvlgz7Cr++H1`*VTeVv3CBa^log-zo9iVAfb1EvF
zK8+T0Q==8)srZ7CpwADn3_c78Zj{ye@B0M>1}QV--Ih~)cAn?^tXeRILItp-PAFWd
zA$~se251yHSZ=K=71q)D#Ox?AVRM^aoEJ`ntosfO?S@|n&RprLnb_UkSl9#HsReEb
zpwI32EKLH+4x*_p)(`-_#6U<!!$}LHZj6<28-avIJg3*OW+=(>lo)uI>oZ4ViKfit
zI~O$CbPNJg=V4MC0LwwjKo{iK&=4AMKAAI%Mo?HuJJ1|I4sJQ0Q3UhlQ;<8!3cn}G
z3+gG@3u<9*Y{OH4QUhpo-ZMbas<lDf7dkC|YHpU#9C0ti34ner+>qs58uk?+V&OH@
zt1qxRNO!ZG4D<yu#rVO({sp{I@LN4RtygAa&dl5%c(S%=^VFk=PGws`6PKGCRs<lH
z#XXr1J$tqAs}GPm$vc2jT@`l^4SP%<PU~B<1UNd4j&Zv3A`zWL0hpyqE)9{1^>uVM
z0W?nUjQwqLKJ&h7JB}&`U~bI!Tm#8^ZPIwqQx<8I)<5p)GxW{^F5yu6f-t#<>jO4L
zbX;C!qP4;bS#jL}1<?fjoe!1jJ3=dfD=|-7bDNp!@YJoDaR5ryfMu;OuDOg3`n0WX
zn|E%3ynp9h$h)i{1M8-#PiM+>@>IZuiok%Q@EhSYT6Xb>CE{<|SZiT7-=E^aCSBwB
z1K{{oIs5;I?i!Z>jrs>*;!ttrJ^t`G=$+wBUlYiv-z=lX-v(glS1?djgiJvfS8r)&
z=#)d3P88+7xgF+RtG*SJ;)nzE9UgGM<Iu$|hNo>AG9R3%u6NS~(th}V=H)eXH2Vke
z^=yweG$IsyR}bHD$GC1DzTw#ac~2>lPP6a7$L8;a$i=BJ1L3Zo%AZmc9#f7Q{s8X8
zA_xKKsIFUF2r9jP)*1F!B1Ma;28o0F?^weavyB@Sh2re?v|(91*iT-gus|QZ$;<mP
zkQCmMg=WZckIQw`@AdUz0iwWU-nFWw<n$_wj;|!+ia!8`CrLUpzYM)0nck3GCQ9x5
z)ylzc*>Tl0N5X3ueB!mQUMBjc+s0Ct<<yVqnBD2z;>bMn(s?!<ym)J|{tqDRgPOUx
zg&&UT6;rMw$U^8>X!xRw1#8``_hLgv#};UV!y>nki;QFWsWIcnxynUuWNE=TOH7BU
z#M|u=1Pqj)8#;wfs1{7flv+1`EvZ2K)^Xg@dF`SGX_}mPdue_ItG);|H)R<uAxwp>
zX>k{q{f+_>GdA$!-ZV~b4eN<kIa9+R(NN=Gi*}25XpDEZHvIm39q=%d!*dzwE|7)w
zves<-`NrpLSHb~Rg@ecVC;<w~)R>L_mJzRlhu>)H^5wx}gMLsoM?ie)!UxeQMa40O
z5A|oL#Y<A02SE&Ku0D;Ya6plLPZL-XMhA5{O)TQdIMr_K_~m6NKCJ#V9bAr-CAZbo
zdJU8{zDkacM%awXPBLEPsu=Pg4f;6soiqyiDBtWye@hz+4!?u|b1-)uht^UrAP$Y+
zl%KLrm*ky5rkcB&B4+JLI}FX*;tV!mC^;x1C1TOq)Bh*>o_IKjR(VX5oz!XircS_F
z%Z$vixX4kj3VL1gByA7>riHhpIynH{ttV>pa6G;_*-w-hbz6tHckIHK?RFqa_3P)n
z_b-6%f}J`tHo6?6Y+<+>e8k{y<`H{v_KVM`9n~D*e^ycVI`usO;Eb}L*WyfGeG*L6
zkpX=SpbXnrn?D4LGmux!n<tj|4qtP{eGM8lp4<JJ?o!fUVEAwBknZWT$U$0U<>JN>
zJyECNtccW`<wHGcnhG=8G0i6rK048|@96lII8_ne+#7QZm?=y@NnA>zr<Vft@5R|q
zHz~V7idUG#8FtP`|D6P4WzmVM`wzPEkdkWijOx{aLMZ@&<OUxTWd^m{^+r<_7>*+;
zigDeV@;~MOcfFt=>%s-ZW5tNU7Oc@?Vs*@v)hkv*kDjv$se-tI>Vto%L)<oV_O~Vw
zU|3s%GDMroOhyOr6nWvYO@Zz~p|9uI^xpWG8g4%>er6oXQgT?WC=E?!W_1+uDIKR_
zu~v&x{mOI6rk&k-j{X6|8hi*N>Jlmxyp%%v`P1<I)BWh5{;^s3qY5JSs;g@-qFZR#
zN&(;QwOX*|A278?nsxe8<UDCT{u48g1JTF@er26Y`1{4=SX?LpaV-pPak$8IA!2n_
zV4Ws#N2(vZaTOqt`X-2-`~%iWI`*Zi_=|g=TnakxqjYC3U<)%LuKfxB@JHvkA_K;J
z$(HKHJCHQj&5<R-6s>Swf9k!r0w@|;Kh0G5U!(qSBLV2N&jx>Ho@g@hWswfTY1@tE
zesn$iI33LM|1JSw5>?PW<+k3g?Z7m-+MCQYum_~rxLBPwMysoz^&d^Z<z&Aa%FBHd
zuW96-5*B^r)9@Rfch`PlG0&)CZ~qvk>X2DL1vbSko2cs%xA3e;e?(DaJ?0D*7->3M
z9^E*amer2)QZmOT=~J4~HXu4f|4OZe`H#UP&n<g<;Ar=JBa*qToBqAwp^bTF5H9HC
zYzk+m4&Kr0Xb~+CNc|C*G3-dN<{m#Y_ymp4tUQy$&mBmB*_hjL75keMa*X#7=6K4n
z_*CEMN_#ui|47m6Ep}B8SBx8lOK;z-n1Zi0MSUKp$yo$+jzHF5Sr9{#Q?%Zr3P$tC
zFMQXe{Ejru*D)4A^MpGCH)J{E{Pg1NOnX+oLMNKvc;YSC+vJC?O<~rZ?Da8D4@2O&
z>0GH7Vss+3-=#a3?AW&0N}dn)6maLEAAp>bI_r0vbDf&Yupq6<e>ZLH&1>h&c6X{j
z+C3UufHVVX`ce5Q9L&S;$;pXA{SK~==%t;FZH8}2|8<6?#VXvu!{d{;pq?q@%#}FK
zek1I%z@*K;opwXk_dJ-~{}!a{Z0kJ!#vdF+P$||q1hz;L^*7E7RKRwv<;}zIc*(Zo
zfQRV&=K*vAZSD2C-k5gMW^W${A&RaIxA;F@evY^ZHe2f)XG(xDCW=lMf0=iIy6b_M
zD8aA*ysBTc)T39Gz8a8~TPv0c$53=AxI<yVwa8ShBj6-;?K)pu@00a`VJ_UZj`cAH
z!qly3zB5Jig4w-Ic5R80DCgzf=xk}SdVGzo_x}G5vGY9ik{fK2Ci*YtP|4IQ(6D%U
zx)=7NGQnht%8ku;Hs{6ettS#ra`S7?u8SwwySutSI`rETkr=GG<%9uej;3Vj)+k63
z8@y!*sBh+7Wr$|}W)OblxjBiu@ZQN~cbQ%5qycF(x6r?bv{QfPzr_8l`U(WHWHCX#
z+Z}s&%LjOcKvH295_Pw3z~tAED;;kUo=E&_QdLfWUgkM2K=|B?|E($Qh0aI^qEkmp
zUEy>_@+fjEGYsY;K^xqYFBp^=Z&+sFo#2fC_V)KwKfnbItCsV*?=%2qae)+?f{6|2
z)4A8o$*#z)0d!Okz&(>2?>&GBAdvRLg;IJ|=Y~e7(C8$36MI_H=rYV#X&#{Q2f%G5
zTd&N(!!m>9U|pMQ2LNaL>bw@veE>d)->PD8xTpjXay@2ceWDED5S+;#S(24`5Pplv
zxbLWTb9H7XTx798v5=^`!_fmMYfsb*4!@Diz0G6JEkIZXT!*Y{CU*4dS1MY|0t7Uy
z9?semW)+lKk;^oFi>(>r5-n*0Wej>cu1+S-Q^GIv@eN?}{6E&x{uRV?so}?<wM`P8
z*=pkJO#g%JeM&fgobPeXtmFF0TR>T_Dj$uZohqlyj^pH{<Q`-{JD-J1_Er>HS_P=;
z8223m{;_Y{o7)M3KVz^*vKKwgN1(C#!7v!$M&_!ekEB9ikPK{^vEykyA%;75;<V)w
zP?uuJLr={3*w`2i(8^r3s^v2bIt=j(=TkLjS)Usv#LxhPW;cJZ*WBv*PKXO2Z8aDb
z1!S6|As}Yq#F{hr+tZiyH=6VX&<cT`M@mU_hN~%xRSGbC9=3?i2>{UIjw+)sy7y!N
zI|tyXY=BMb84LrETwVi^Kv(iS;4_^^!0Tiln;(OrC-6HAUuYu%J`&ExmD-z7qG6qf
zSORDp81eQUI{!jNVec?lNfFTo&*(>2H5{xoDSS%e=4nkY=(s2x=MM2Ltmy9wK~C_5
zsd*~k+^m{f!9a5!_5<C`G0(obAOmCZG&=nTzs*4aO^wLC>C1Uq=1Tjq(e8-@pp5QC
zukqkD25?2EK{ECFEFk+BG5!DO&v3>+`v2yv+yT@%XlN@0t+b-U=hRHhiZM0{qi^Da
z0<P&m7wK|K%Xlrs=v;gCCo6*Ag58?0?oVD+VCp85uOh4tkK7qF6|NAl2*<0TxGRmI
z=mzQSo(bqVy#?+D**)?utb}981*|}vnA=f)1|N$rHDl2HKRuV_@zQ^X_QP=2(rc9$
zOR=7aUbr(H_oisaaC;sW+?-}lvdo0)V$WJWxzW~&!`A7`RH|Q;`S@i+W+(e$EBr|z
z&=g#-0jaNlt=Y%@Yy&{|9qBqNfuo(_542cTTq!fFCSy4(+w%7=Zs}OpAj-TPu|JP@
z2pdg~GTo|)J&>2Z-;4uvmdk8O;IAAHt%K}`XQ^WN!%LWn1>$CV4=UBMfRcHksVB66
z?O#;ZDIs{J%ZeC_WE<|lV7V27laa0^4Vq$@#wsJebm^t--5vfqFD}=rs<pY6^60ez
z_dRm4I7>TqOX*qcBxkAENtY!_yB73Ytl<Xxu$_Ir*R1p)#x(^TW)&Na`cl>SaBh(q
zYYvf_YLgSF@;+qO$gSvpbeQELe!@n|n~G+R&3-g4gQ-ou=UZ;2vMVkFqwQ*jlZ-Fs
z^C6}A2%y@OZ#Sxz^)tR{+WWFSvE{U3P8erbhWs`v{yQ`z&chnNcW))EoDxch*eRv1
zP}r@R=8FXv8y&(pupjiz)i!!TdyOYQ>Kg`TYCa9aRrkrvKpd#yEQZ4A)$DJL+7q5U
zLk$DRkyR&3`aNr(<$(kxq=FDdedTV*=z{OwT|=>>!=tyS@&iMP+Lxyk3#jlHr*f2r
zWJMR78ymeVwmB646BXa_Npd;5blkp|uoRZ;5v1^`Y@yqDIwvrL?8*y;5UJ&9$WZka
z>T9vMTv)1-!le146fIV-nRi5{b3@a_3~N6G!&A=Ph4@yLCXUdBU&tzp^K0fU%6<1R
zHZ}`|V3?SrtoY6kpXzntwfSZt)}Oj@qIOM93IVsm=#R{-pHM=ALSKBWRv*j$1Ase+
zK-)3f8>=g`5TVE3#D@2}xKd!V>N%fdp2tXc#!V&|h)LAK&r8We6WVQ#W1VT5<f}}Y
z143vuWN1vc!ilcw-$eI2nCOweDg+5By|zm<y;rY#_|x20@MH)E^-z-@!nmGRoKE4M
zksz>~#bzZWTNw_ks71{8M2uO#W<%mSW=q@}I69tcjnR_&hTNBe_99HjZ{$Bkp1dhc
z#tFJM`rFJxAP5tkw9I6mf4d1?q6_8O%&;TU@r^73Dkcu)F$;#l?bJ+1B5BE1AE0<-
zqXJ0E{0)^wM`GYhjW~^`GYi`*PmFpOhVz58E^UEWqK{7HFPN5RBU&Z-jMnec2IM4u
zO5x*Wwp&Q-!$oxy$nXYMXW2?%|0uFb^ei1KK!plm*YzP*`1BRR6LAk^U~{dnwMaGa
z4*<K))M4{-WgI@nl)F#%F~E+nZKAM)s*sNy=zolxPtDN=(Q=tfe$@XENfPAbq^adE
zB)MzYJB^zDRF(#7r%;j*dI?ZO%5#p(_Vb)r1_L=5-2D7zX$`PjzVF87KF4(rk2II=
z$u+7Q8ATtVU|?7jVhGH$zVkZWZ?o0!Ahd;3-DGN}j;<!`79wF9GDS-)Ht%_0H=En6
ztbYW{o%rVN)FiU|MJTXdr=Q)XY4iT0l+ZAtpIoT#r=ur;LE5(}G^nI~p?J(XscOsS
z&ZD@R#=G_#o9$ru?4tqFs?rC2Ovc2dlLpnF3BkjEGy={NLVsTnT+E4FP?|qKm4f!S
z4iesIJZ@U3_9^SN5yIqtHIB7V2$eJ%4FqQ`%=lFWTYVyqwl*)H9~wi!&1@3a>@d=a
z9uK#U=x1+OB}RqyPV%KtOW;jbIQr?b`4OC`-!x;2oh2Fl!JN0^GqUZUVAmE3o48wl
zO9JeQmrR&ZxnI)bMVVVjRNmF>&3%T5MYu8ZR10EZ$z%MM5A6euA-$L_P$)rN&D1JR
zXCTT0ixrQyNAyNPpoi(h_z{eBc}%TlH0mqk?{9XY?<COb1o<7Zq-9+Dq5#Ccx0-|^
zedu1x7@Hs*j5i>r_4V=M+#A*i)on5Ssny%XIQKt*ii#z7TSEOqvI~^VJ8NFb$44w<
zC!c7AZ3>YD$P)J7C&T0ih;?;3peF9g3FSCw+;k&&5d5a#@Fat<3=xuRKWR8qww6Ur
z#v#7J`9)Xc8pD(`y-c=_PhQI{jJ&j)+ukFsMqi#X@F=4Vh2M(O&-8afR0+oxq5oXf
zscGc@yM%qp@tYT@s9{F>Ct<7^|8`-@U%311fBJA5Ih7<Lg-8Q?nHb!#2^W!CwRA51
zXtr7~c`QToXWn@xRr&D-ot(b)C#K5m-*6>mh(N~2#O`+64!`Qugn;H&C!tV<rb6ml
zY(Q41^52S_y{OvRH7L8evQFXYVz;(-rkwk)#2W&kTk(@eDRz*3G1uwR6sVQS5lc||
z1QhwM=!o_TfxR-vv<)rR!*5DVgW<=r*5*c?A3>JO)?Zo0_AB-)d+`S&mM7-*`x6-p
zsvYR+sGwum3>G@xlxGvgTkcimt=AY(o^fQe+Grz~6?JlB7aT;kobABfsz2(`5(**o
zwl1<Z$|DZg=jW}-r;K=`jm^3Bov1fEtD{Hk;}g9C9WskX?ykqY;&+e#wRsb?w7<u$
zc1A#apJGa*)q0LeT>Rltq&Y})NNvXU5uq%RMEc_VFRD>qlHB#9qNawHLT1CqH~*N8
zO*);O;L8tcfbAY;*R!#T11Hs;%|0F?+N5adP>A2~6Zz;0lWs|aWZ%TkiG7054x<yq
z?7NU7^6Ma)^~C1a`Ur2y-sl!N+60^M10(=$QgM%sj!P-|akNQ($#_6$CiMv5O3{6O
z*o7JoL>Ov`N$3XUh(2Io(t2VnD*oimDLj5>m2ZE4jY*JVAaDfxruF%#jxno&&73e@
z{;8%o*>4p-{=?L?N{(NTbi!K15MlPtlGDY7Wf^r;86A`2_bUb(nk^0SCgYfjnr`_K
z>)2SRqTrmf+(Od;M|Z|peQLlvG=utltXKm_J1VeVlhuN~BV=n#$75U*zhQufx45Wx
z?J-*|g{Bst_uz28Oj|u<g3I`51K|_)rcf-bvbbu?8kgJzIh*wI3&M6;1lrU~6d3xS
zglBc6IOIA{q@TpNLhlLWxqm9pBY5z9A%8xaHEjxgeH2uhLK%Kii49KoC~eU3(Z@me
zzTE6sKm0y1=}Z9(rjYl%m;wUJ@6ZOdA$QZHp&@H9m3N6?OW&WS-bGP*9>y{`Y@l}N
z!H`N@CTq;bNEpWq($d7%BvVm=lK6}7wZ;m_A9P+qcw1DmA+vkN&dzEJ8diU5^SB#V
zT7L2mW_4NrC`6TZx7|CF22MH9)iZ`F>Q?Wx{@<C!-dWFVVqxqe`rq`sLt88FHl&cL
z@Vn5k3D#LLWG&mJ|0e6iQds1v=Cfil;kDnfKO-&Ncw7n2zRQp<T+>&#Dk>P@O@fD*
z<7DT-;!^>cPWA-9<|J!ehMApXYrAw_F-^HKL@3GzXNj<08!YmonI-FazwIiC#N)g*
zHg#;;D>^qOJmd+C$FG>fB9W8Ls@1_qC;AVzB^9@e$J%CGkJ$3YAf34247UAW>w`Qb
zg3{RJyB^s@V<lxm#%hytksXQ614`Hfc!&SLV3a|NPw{C<o-iPc0fP6Zd3NqlBv;7;
zWpti&S{Zn+8RFCk;O!PD)N!I6$5Z%>;o?wRVteqBGW|eI9xqjy#j2gcOSFbq$sH@7
zUv(i8e(6s$1Sc~yXKoV{nuY08C1^UU%t7x@a`nHQPxd0J^5+>MSg3X37s4(jzRPL+
zcs=ug0ubiUOC=C@SGro*I1vmy;v)zfcoW=8na#z{j+SJPPCCgs#fJmSdIP&^LFlLF
z04AxDgL}k`gRpB{9KTlA02<rmg=ssx;D6`?;>zOcqd*g|MoR#rgv7j^J;BC02M;)}
zkt(tHlu&K`2jG5({(T+7VugmXFyWVp#zXv&#TS=}G0e2QjmjKL@BRDd?*;a%PD7eC
z=_C5mf?Z3!XzjuhX5UrMluS(=Sl9%Z**)E*Ao5T(bSkAsl^4f?PcHx>hJ^BSi|k9Y
z5k+){9L}!88lh3`4NdyWmnI|>G*g1N2FmJN*qC>DLGrb{eYKkal_Ek6PZ-E?bbezN
z5GhPhogoYZ-j>Bdf_8Tf03i0PEhQVY{^S-zEoW|#Ra{>7Jiti3jD9w7Yv46U(V539
zz?C?Cg+_?_Na&B=Iy!mZ--4DQA^zdQ{N&0k&*|WON~R~xQPxhVt~Oey0YalR4K;PN
zAuS**4D*TU1%Pe=U=}Ym*Nf@`6kAcrOM`Ua&=-tHREjrM`3Cp4H+e8R4xPs%9!TTf
z*c+gogB0ZJ!ubq7(D7@)!r5qjyy;J+y0(&{0Z&UBbe;^h*Uqgk0aW}>vD}v>Hy75F
zFDYp55%UzToW5YhzNfi_UwcY!ZtU*@?!-ZKyY|HObF3#mU7}e80wAzc<;!L{>b)}o
z+xvUrA_fvWN%RH=Zy_N+gK3^5WdOtcDUycMMFEyJ5(61|k8c@_WsG_t-KacCXLls>
z1P(<O);D)_(R*np4iEE2{{TebPjBuAo}}&PcX<KMQ*p<h*Kg-ITmk^tJUscg2hjSj
zWN|3$MpKK?L&lU~?#3JlEabd51o#WSRu>MV@F^;sz-Sv~OjTzLLaagYMUfL5-{0uw
zf9bdW-?{UD*t@yBzcQ+`4CnJC2O88y5c8LLCHa)JTzJcQ3xa^J4Suw{&ei;VX1RzD
zLdFC7p*h9#EFB?hbKgQ*i0^VNSJ``x*Iy`qg1y|j$XFEC#ZpiD0~phafqra?&66k@
z+%$-pwlx2yg)oHK?<Gieu-cBTOnpWeJo&g#46$EdU=A$ZDqlO=ovDdM7-afpKwye@
zCFW+DmmMi3W_YhA*VU``Lo5GaOi6b7H;<cbJIe6AiHK+{1VMQ17E>3$w||tlc!db>
zx#tm^?3U?*ej{XU@9guX{Wy_i=)bPB2@b8={+jL@gl%M>Y&L32t!ZquweMDdzaik3
zQN6{tg#ysa?8h4$$DstrF#mBNq9wwo-P-s}xx?pFf9tSfu(M#zT%b<qH}j*6Zu_oY
z<BhpJrU1)PDfy-|6|uM@Lr8DYVH(#lQ@_y4WO?J1H_B?gF~}-@cY|Y3{08+zt9*WR
zx%sQ17@j~PbpqGF66dkxc`5S{m{G_6A;;y(9VPLvm!Wax&0FTgur`6}>U0@VSU^y|
zXU+~H=?@@DXLsAIbV8@R-fd{ZROciXZ@o+5Xyb*B<*XtMqT26ycOcKaPn-B}mJXpy
z4L(XtEywJ0mB)NVPQe;BA}=gS&MYY<jzp5|PIZRodxp2S1bf$c!%||lb=y}$2G*CD
z#CPh^biQYflvx99NN}((^*aG~D<|L%flBOnstuXU`(%pu2#Xbkb<9rfhEa<4a*D}E
z@^2hr=~GIHK><+rNgfYG%x#eX`0`uunyC#>zM*)Qs6gGL#UkIvmBTZ>D|O;oq5ojO
ze*mlv>54xT#3|P3PvRX4@S>sKSLrh4eyfoR;=zp#v+I*n;)UwXxIX}4->luq{oR<V
zeVcn3gp-Qq)@JBQ`~RRFAd%V#*<-hvcR8?NjQI}IuwOhgVk%!BvN6~*pSL$Yj?E#2
zm9w!nE-cOL{tG@vO~y9+3Yw2C9+dWd{<yLCZ#1UKC9aj61-lA9?lxKKM<XrdCgE_U
z$Kr1e1S*#EBwWxtRPo^j$b*33E3E#s0i71JL+C1xJKb1smOqKog-k}URw|Z>p>TxB
zvfQ|k8#nR$d1Z|a-=}bLsSmfUPE9OH$uc;P{pL69X-j7K01hP&je=c@wcOBSDljrN
z8<C#Otf+q3hIMsqlPdT4qTldw(8=Kvq7ArylnNYc9z8vQGxVe$P0BacGI{%heU9xM
zx=9rtSNeRLhmYV%*aHYL3N3!`58x^+jn>HfyOGWKjpph17}ttA3(ZF%gf7dI8T-kY
zSq$}4b%3k5m&)gQ-P8J3Md2L6S$wy>TjpZS?wE0ZKyVu&mKtix(Fd;&vzOgg=?}PI
zb;g}J)y?GSrDm;#ZC;Bg+Mt}$b5Oht#OwZpEf{^cjW@h`<0v7up18*Zzh<&^B*J$4
z8*0(CpM#>lzQBaOVRIb?R}^%sAIW7pFe(T*LL2miJr3mk14zM91qS+JW4?7$48R<h
zdzvkHus?4~;7bK`N^Q2D#>zD<HwD3<G$Td*DSy}<AHuic!wGeqIDG1jP+KmdQE}3<
znQF5X8rj@Q%M97L+P`slc2{~GkEUwoF?;MZcONJGSE%!tqj$;eM+)iolv|V1gZ_h3
zQV*x6&BG)1&TOfHv+LuzY4y#%9M++}1+<;R^&Pai8(y?G@dvOOX~45<s9A;>%hk?}
ze!Dj=L)l0CqR|^q+5Q1!<z3C?1Bv_r+*2}$jM5JM1Hdy;%BGMKvoU7@-IrGC{{W^g
z#dN?bBMH>hlPPRh!iU7%JUk7|*48IW9G=9@X8ZvNDMfYp7Ec`KhnZ1=1I33fXDkWD
z!3j$>sA9!G08PZwgny|GbI&@(;LY9tnHM(o9g1%Wu>o*&amPwxbDt^~RywHczjq4+
z;zIL44Die#ZwPLF`@zA{C97syHH0&0#MLqv^%nO2u|J)ns*Xd?eBg5>v>T{gu%SI>
z`fj^@SmW6xBM*uznfLHDv;SyVcXJyqEa1%I_M@=l8Oh0vEnun(@Hmr*)M~Ig4E8&G
zD{6B*dDS&kbETfp_W~S-pt<UDvl~tsP6{YPKg+|!vk~)=3L*z93+%_6d4i=wisy%H
zHVOi1WX;d?12}`OQyGTOcEkld`satNFVWn6s2vb%<2>A&dop~3<`i6=8c;G7s2^xl
z4D&#GegwgEItVL!Onz1wgfk;i@!uS688dkS10J-caZ_CB`Ukz^+qBpN509w^OOpFX
z38CUmgVL4j7tvvB%`C9sr5%w+a;D2i=Qq^!DfrSiry!dDd2@m&(f@Im!>Qh!iU1Dl
zV;$`{J#5Y&tsQ3EXzJBEhQ0hwuK334SlUjDxq5?GI!0R&Xl7A`v+-|X4Vbf4KSYcn
zje*7Dapf6KD9EYza~BwA%s-?rrx0|&rNQfS7<7Y+fJ!WL+}riu9M46<u{ZBz1S1T+
z7S^|!{7gESnPs38O$XNSmqt_(*vB27q_$ms>E5MXtFBR~T~+BpD*CmB+=948NYGo;
zrNvBiLKq7z*~yP1K4UWT?`CnBNb0(|A0)o%<yM*C9k8(lV#WI-K1aBp1(g=>+Pq=5
zZ0;P|z=zVd37Tc7?`D8dPLn7Q+N(B?75#6yXWyTj-mE&&M20PG>zs7rX~vC%mgss@
zjwa$^GPK%4>Hi|{O~abXwzbij0E!?Yf~2qk3Pfy*V1bm15Kuv+P$HsI0wI*8B2dyP
z20{YC4iFX5k_ZAp1p}hQQZTeh6cjLMP=ZnVE;NaBfdHZN&cv#{_u1ch?sw1mx$pV`
zS;<;+&2G&(-Z9=W##`p&f`hKKnCj8^iem{QdZCVA!S<eb>|XNxxur)?GExgotHB^9
z>(bNt;n}B0A|vViGu7?=hTIvp<kjODdd)x1E|$>K%wLMEe;XcON_AZ%mZBM(NAb^n
zj1|*x;W<;<TgS&0`Ca)V>$en-eFbEC?!nLO*(&`0=kvptp3d#Rc85JT;#Ya<D|nho
zgPGj6eql~{2+YyO4~D-2l71w*$6@UGs|{=GCSVtNyy=I#ReS08F&?mzL(few$J~)T
z_0G){JeqyOcjOLL?L4xT&Umj@8f>0*Y2OjMlHN{vJ1D7e&i^8Rnd~iKi~dx6_uOF~
zzr~z$?GLGVs>1ss3x5oc%jPD=2z9=#5qP*pur+cVSmPrBWnHN{ci=0fH(h!3_!$^*
zhNf{((8BJ$M@->QLa`o2r)9eit<+i`y?rL}80#!d2w2F(uQ@2~^4t=oi8VFAo5%Z%
z^DZ-9{50#BOvt}6qxA!IO84zsVtL=#!XE@TerFEX<R+GAvGN6}zqbyXZ!t)BJv)Za
zq4orDUy(8AbB4|BBS_KrS_BRKSQo|-BDecT73$?_@<Nw3TJTnN^X22y%UmNhYhvD@
zy&D8Q7oM^2ykDeP%a6Yzu`Xvno1)X}GbF?6r?GLbv{c$?pCxnr{76%Qg!CK!@m9u-
zIp?#(r%Hvvv1`c3qQf+ff69v<7kB*d>eKVl>0S>O!<>2c9Ikx%cB*({MY>KtskCO^
z?zW4m!Di;mufQmvq|6p0bUt$0xE|(a=|b7ksuC)<M?B!IDp|}fynMNAzI`Q%=sNsM
z>S@=N^B0~55z@@x2Oat_GDp1$*9KqqQ0`?GbLvVlv(~*M-o1hF6|h!};K!DxxULtE
zD^{ypO>7rp&}K*ogU_3PUUvqL$M6MRI8A<3-7+(GB4%lB22Q)nvV-0GJ$7H4&-vw~
znU=(nIwSO8@S5O)XvYNuso~R!U5iU1j2?akZ)9fG<Z@h$gX|vELtXGDa$`fqgM&-P
z;hK6FX58|MSoakD*3YkO4c3v#srWEC&5E_r7Kp53O4)s;^_h3rZ|_3L^~fw`Xr^ye
zwyeFCK$|Z-QyLDTWRE%ls!bXl;f7tlP%cX7wgwiK3=1TOWLX1=W%a)bMhlO`u!WPW
zx7@5h>h<`TxIU#vIKq4ov&>rYPMejOCPLvF5#%~WJQ8i3Gc0&JwiE`2n@2Uigb(sj
z9)H!^|D%@mM+dXV&u})F6w@5Rv#zOEwHEK%g+{YWzZGsCG)$K{u+`~FpBKcm;bd^V
zTfSkSOml8zZTnO6stu=MILP3GJMTr6&pC-V`8;Lb)6&{n6g50CX8K^*ZpVVdu!lFy
zr>eLPt6<YwR#x=&c8)E)Y#R6O4eK5^Wl!dMT3U1ceJDFz@@n*`?9G>~D!N&=on$64
zCe2jDY>m!T@1^Z{NUNhQsc<%#pk|q$y3E!*m*_RKZ)*C;&$X@F$;uo0pq8)Std`vu
zBaV|VKYDubRdvJeJwb$fmfa2w&svDr-a2Bf+^D*AS-V96wR~=u?eFrd_79Zt>g~Nu
zCrK?%t~hqv#fv3YQAK<@9Zm@{9m-kH8nuUTKI{9X(@W&54D8vbrknt{3SG}1Em<&n
zmP*IRYhOA$%&bbm4@dl{-H~IvC6c(<taU!j{AHCxmtaGCS<n6;@yB2P{E40t9Y9mD
z%Ae!A1~47O8E-Ur-$!C)D$TR+#8oL>JtNa5i2<kCb>1xBw_CRa1u{DxX716_4%=7)
z87u4(mlMw<DCRGwYYz36X%_?r?JhcWI!r1U3guTspwO%l>Dj>&vPVOCDYJ*)TFou!
zdUK18>>DLRaY}QG`Bl~&p}%m`smH9qXP2bL`_Ax-31idSIXy*#XSkW|ObgC;0e(dm
z%|THN5$I9C*H7O-?D8DXKu%rB2sq%Emr6k!Q_(DPmKrz7Khj=jW3qHI9A#^h!NLLE
zam$|FzN!tbzH0F`NmM0K-za5GP5e}HKdwbOh*?Eho(d@hBS{~Q3&8=Bm4M*00_Apg
zL<BVoBFDR^mx4l;(cs`1*oavX3|%%ydVS7={m^Gdw^YB}mfWFMC(}y}@Dw?=a}cf<
zOPvbne4j9L^DM-Hq+Np+J$QY#Au^+NOBB|JR17va`&aQ2M*6DRpD0U7SQukFO<coS
zM%0=s$PrY=-Fe3T#98V%!b0i$wYCrTfS-#xjvc25mD|j>a@5oh=}EdJ{xyh@#7h}u
zqJe$|=Bm98?S+M<S;QzLHw^UA_anpWi}iAORJ{2w$V|V2VJfdx3=TSECrfO%QVD+s
zfQ&(xOKB6|A#WZo%l4L<ugyZFK-P6qthoApFM7po<N>IXnve*X$z*^GZ=f9!^J34U
zB7ozA(a^!pPo(l$NMU%Gc<q8C<7QHI8(34s=Ou&(j*fzZ$98*$p@}JJa&p4CJ<JHv
zJ!43wJ~Yvq;p1a7R|b$~TgTA59l#&i<Gsg7XV8poV2!jsF;!e}!Fx7mbdqAS*3&>^
zJ2BCR9h9X!f_}+I?5ryZ!RTKl&TO5->|fv7*;rK)E-T~cPyj$?j~P#N=OQ&-AD<Zp
z-p>9#76PAnj*73^^7iIR0*%o#1oq!+z<y6+vFeqz2HJ&#gFrntC4FWbcvKW*6T|GV
zgc@*vS7&?m+a_?sc;;Sx(xa)sRz6@v9ZXF<ZM(T4b)a&;%35N`d)v~`j_AyVP4){e
z`Hk!yZ}@CykK8l<-jp8G&^{gj))h6lI(W`RmVkA}^);O3xrzH+9!ZD0?4iaADZ%g|
zdsPoAf{u7J8<=llZrX~Qb$M3#IA>waY}_b9JNVqJj$~q3w${UMC+*67Y<5%IigxUu
zujexDOl)&cud@F(-GOVIppwgRdGKZCnKIN+B|ht*+PXCNX#Ip+9x<Qu`M9`OZFlSe
z%qRFXEsFVHH~r`RQ(rjURh~B$dv@Id&%jT+j~ST_zH-w+j!8rlJ?fpz7%%%N=@?O|
zGuU90B497lC05)ok*b#Dl+Q4yD1Hoscr7z()zWoq5`CVUukSZ{%-k?(@`95WbNzTs
z&Q~zF=72))Bd#-eVJ>!Q<b*spnO78s|AD1;Iz3ze5>aAzedX`(&zj%M+SragxnXS|
z{>6Ts_0dNMCxnAl$8>dAHd`leBcTm_n(`c)y<BCJ#QBgBbbc`j3N0?pAI%1zP7WIt
z5sExsc4vzk;;8-V1<j9Z^XA&IPZUxO{;Ht*OHSj#PmNdvgt#vUD&Iz@tjK>sif{J_
zjUpsIuGOv-hl$E!SBj=>i!YV76R0jCtJRf_nCxL1fzhWCI0V5ipP+eg8tQSRvt9Tb
zrZ)5U54HR>sWB!pnixkuo|<}Ga}`$`yn{t<W2f&Fm25Gu?%2XB6O6?|OG@o-*~C)0
zZSDXRYw-1GZ`51NIqnOH)ic`ISr#2E|8coN?|$|d^WgpXi|?c6Ypf21F=;K|9*1&k
zv2_>io-QLhsPKJuk#VehGD?8CIX<h+O-g$;iU>y$XXh^jkm&=PR@3aT#cdw`voYU=
zKm7`p1)znl<+be|!MUtogUZWq_FJK-jQR{I!GIZpzd^$2uDxG2-K(f+oLmCm4ZC$-
zd0bvE^HWspvVez!0gPCP;-;fx7X_|;<uLSQ0L%V><d*21p|{W6M}CHZ#ujwaKAecz
zq;Y8e5HHAUPdjE9|IYlJIjz>^p|FLm8Q*S~wLG};#ShCL-brZ3yu-BO_gsJ$6eGoB
zDUAe`DbjAmsG{!|6Y2b5d8eaZpm;s^|Iu>CF_oF}<Y*JVj}LoC`F*inVwXB&CE@tU
zio^<chLM$J?}?*<c=FT<A0Ca~LWj!kJt()Wx%{#{Xjk4jY5$=-s<EPTO16szrKmcy
z8Q2&59N}!5Sc=;kiSLVvwUSc=HkjJv^h|sol<rYxYFcZ%c%wTJN>w=y!^HC0=uq=k
z;ke7qL*=#yGpW-Ik9xfjtPJm#Ki3y|O>vHr!<rTBHp1<1e3LUBnd}2i?M;Mxv-{7q
z_C_C-ozX;pnU?DIR!riX>#sqyto;3{c0u*)?gx@-RlH^rKY~H8KhuWaLELsxX3F?>
zD#ueJQ*%yo8D0+I&sSGJN*E}fTZM_SM-q=h9hRnk@wMA!6${K7f7+1s6wNP|i;hR}
zBD3+;a;+6moVK3D{q7Pu`V`+`Ex!jB!MOc21p9phijegZ(t%$=TJVxP<rNDSy*gwQ
z&ceF|ofT5+>2mQ7Ica)v{<l{i+@A9au%>!IqhWY_KxC-=k4bRj+I3#40lgm~=My+N
zw~CkRwvZM)EcnPscH2((kJOB-L2+qe<^>HK_?@R_>K}ar?b@9$PSKKTHlT%g9Y=B9
zql_<oEl1l=HIi3Ph^h|Lm8VE(kA_IHvR-hbRCa9N!H&F^ykPHIMfW0cTVW3cm^?3u
z#MvFyXd4^TLe(7=*4hZV&#WkS<F@5VyrfUETISvDg0EP5op_G>Y-Q0|R-c=$rJdPM
zuPdZ)ubU}l1>5^Hzu&ecIBhnFwd>H9zINKD=Yg04Xt>`I^}EaICN_3g@{0W8=Q;j$
z(y7<M<Q)F<_BC&R*6^!#wYTyJJN`;AMJ{`zI&O=N*Kj#eB5~=PK_xhlqcZ{S@aVze
z);{S5V~ZZ&?U?8b!};97mIdLfEBU$U7NYJJA_5wf@48u&zXF5T%BG6G1smjp<?AAc
zX-ni|ij<?ec*T%a2-qv7qpK&F>fs}%575I$NJsB6mVX7-?OG4-B-^l$TKh<(v)&M8
z31c!OMYZ|!SQ$H1KEPF_qj)PbIKqaCso*o?68QUq$v?Gb%T+|5`-D;xIY3k+LM(j2
z{4#@DyHo+&mA=);^g$o5-lu@E=i6jM$P0W`I_f=fqqg<s$C@(`U~C%etMj6TeMZyF
zcl<q}UOyyE-XkkOjV6Zl>RH9iQ*6fkE?Dm!@%ZJ#+=1Yoa($8V25I1IZBPKq&gJ_B
z1C!Hdrq0JT^s%c-nJxLf?3I7qiTgaz4PXDk7ar!xZiz)J`{NG;x=ltRXOnF(Osm|X
zp%U-x#-`}ZqZ$YI6^w}c-C3FKZnGR5gl18-(dQ#=+25fK!Bx0Q1nD=+26(m67x}>i
z?^=P+jx*gsv#d3I*9}$;*O|{>o+Zs;gkGXST``N>L0ipw5Wc!yU_szLN^Jgsa@GHt
zz5S}kTH(~J94-Gk9Rc;&i*G7S-N7&3$w<z}Ix-z`-;>`@SH5EReg$IZ+&2Au=>|F7
zgH>g^_~wgfaq#!~FTB$&xi$IOV%yE*#HmYOjJ}B-elc6<EpjbjR_Ul)uq5W~%=I8D
zp?Y*-#?188Ev?816V~yylQFjw`(%|HgQV<wue8bI<Lo?zhHnQ{+E`6ryFD|fjP@RE
zos>}zojuYPWSsM*4I<Qa%uc9hZ7=5gEEyy4ul+Ie6`c9a*nK9yiz+=}E#Hr<VkGQ$
zGa4>2e)ZaMd0~BnA5G5@`lNq`A94%x3U2XeAkeZdSUyaj6^N{Z#>SVbo7(;LIAksF
zZS?(blY|+QHs`djE&AM&(qg9781rR~Leul#U)<;#9Vk^)UZreSz(6RpEPoZWCUe3<
zM$wkfA`YSVu<Hcz53}JSk1iKeWeJm?pEK;iOyB7>lgE7A26Y|q1d{BS?|(3x<BuPd
zzO{*SX=4O$HIE*;d7`aA%dYfJM0>Sg)_`3Bv>${94>1|Yn`cVr0(3=pmhXS=x~S(s
zFpBVK{(BW_(haZfFzaB`EKbBj8R`19J6ztAiV7RD?*T<azSVJwyqA?8Kc7;30E)L{
z^(iLjOhuo_Jl^_qOoEl$*><QrK0|M;7-;Czojao4p_Nc#-kW19w=Jo+pMt{3gQ7z}
zS-ErI_HM{l#vYn_T%rDO$N<@ta-5eGg1_3h^>^3DoQ*J6TVbH(kVGBE&N<#HZPI#J
zc6s+c<1LKj1^3W?y<(;7&vV3_;-xUC-`UUGPOU>WHUAg!wD{`w2%5)r#P0MnXbhy7
z`_nZdrIZaY?0z!^It%7B7SQu(N)0&m<{EH;`od%Nworj9{tfs+iYe$e52{xIYUA~P
zI5pClS-{xBvi^#U%(UAI3P|B?q6R8&Tm(!n#oeU^Ymm}NS&JH-F*Zrhdcc<A(e9ou
z@8TdCafY+h*|M+FmE}~;X2c`!*B9Elbci`+vJe3+_%&D-A?A~R?i=ju0b0X)dRqOS
zW5b;sgj*=uI<h2^qN4}H9da{s1bsbzxybM}nV0<fx_p9zvQ*ZdxL<#zyeWl^{Qz!>
zaHQ?F#5qGmW}5#*D{0$w=?rD*Zl~(bfS53FLF8AEHX#uYfrF5d5MmZSMr!Rq?!6+l
zB~&;Kj#G@?jm;vc26jC9Yj72tknGPevyu`qo0M<t<z35Z=YE>usDxT{JXVjR*0zsQ
zR5(|r0vX1#G9q+JUDKF1Q2mH2Y$!w=2_VwYD_Jr3KLC1?*<zl1H!se78<C=&x8lru
zT-@F8>&V4#Jxy#XN+63M^~RpWR4Hk`g`+cyS?j44idi*NESm}&1UuHnC&g*ox;#OI
z6j^i&KDSFK>nv>da&t?PX&xR_3UbkI(&i4*_%Lu~|7m1t)X8->H?185>nytOhR>WN
z&t)yLb#e8iE?Mn)ma-ns1r=vo&Bk7d2yEOJp<fCy<uzNs%KUtL6^_bUr{>Cpv7VS8
zfth}VAH(B*GvhWA2GF0crsC~>903P+``t|rb4`!b&#VF&5hwi{68tN73fRISu$7$G
zkBeDIl~9ah^-@#UCB2Jo?NF_UOa_d#)@VZ^X9@dao|sO1_6FoR@o{OH8Mtb?nKeg;
zz6-O)uSXU|QL^OEs7|3gvJiU!C{HRfw|niYM$ug?wX{10Iu4aSlpej=6TNGS^0>X_
zxn$Z~GBF9^-1N`UmMgGdp#6<KeKTRkCi>^7{k>Z$NNjVzvUs>|DFAB<1xerQPH;~e
z?su&ycVI4#JzSa}dy+V2XX+=p()9LiP`Wld>5JOlLb0(#!>B!s9${keyH8B~E0vD_
z^6LBtogk_`<Tm}8dVp}#(mUPFFgwF%h2A>k_?f!W=Zz{%uCh&y`V@b1O*|8uQUUEz
zk)v9N!f&1`A<5R-XP;bloPNaMx`Wry!C_@J*e#i!BSYPTb^D(4$P#6>B^rB+VmDr+
z<rx@kQyolfQP9q9!;^}zQuey;6YZLJ--RItGNve-)LlIGBPrX``1|rwcRWg@F5N5_
z5~r1Ux5`+MWucJ&L+usN0Vn_fPyilOYx&Xu%(DhQxcU`{Dts4w1yY|U%EQv(uwgzz
zF3ST`vJXCJ5NIec4T43!@Yns{(1kI1Cr&#?fAz5iGOA@my-8D<pwg>HZ68lmv2bjG
z_wuOsG14Q3#@XhjkV5w2E2y15r}1*oa$;EbbBr3njOC$}jh?re&-}jgnIErMLo#T0
z+p@DaZ&LciG<!aG?$>8C8#3`GGv}Zd`k{1S^y19yt!R*WeZIe9=Q#qeFv*ov;R`w8
z5D}S_K+#@90U%=)FcJLK+Wg{*G7j<x6=P&BJ1mjG+j3d>+b*NT2(KYk#Wg5wFCLp^
zOvbU7;>tbV1`DIkW!{`Fhz{;Soe+xI!53ew6xQ&gpR|+lt5^Y-Ar+0<T#jabrY^_d
z@YZ5wj_~5r=FBBZ3c8!$3mZZI0vTZ@vJ$hXZ$kM7w~+f`<ymG&o+xib#i1f{-JI#@
zcT|VwGvZq>e5T8inv1<29Fa7CXwm*~mdbpzUm-N;NvV8oh3sS5mgy36V_EGBo9h|l
zGTXBAFFd%t_vEKnB~U+n`G<7vTKKqWR;Z;AN-B^OxYhKAhGbYu>ZB=;&8Wj|eW#+x
z^@<sx)Ge^Pt>9s3W7hYu<xl&?kNdcyE#L?F$7~2$j0oH(RlIYNa~I_T?Kc{L-Y}E)
zS>>PS>)>^3RtnoivMOG9`TQFz_j8f|e5l63j?7*E?2T_#Jo_E&n)#RVnk#JCYVxla
z%4aEf?z-})Fh!++yb{Q7b8BT^t60X-hU`U}^Q4m8Fb)>Aqem?}d)Qvuhw{sWBc@RJ
zG=06^Y9%x@-osRme}^Yry)r>UF?NPy8pdxgk%QG>gCm)lZb4<}=W_5~0*SsZ-BTCP
z_f8);kN=fu-<bT9kxm-#(YO@ha^mT*UC@Zi%Dzwbh;@@imYpKuC;co&r0qSUf^lZ<
zhBlG0MEBmgdj}F3cyk!xq15#Q^r-q)Z&z7(@(dBAK#>RX9J$S9`$rB*WDn&%;Sfw`
z_e~6IYBIeO9=1VN?>s)}`_9gfj*Y<x`4ETb!kf;+?key2m%;4F3>=sLa6zHPaD&3{
zE=3o3Es+xe7SVJhGqF(JQi)kF{MI=L_!T}E)`r?FQ^MdPoB4{$$frgK98E_3KPE<3
zUw8W$v*w@Y@`0&5wDqsCUf1ZVx`oSb9836n0_|YoF6l|e*tHm=e+=IOk~??kz!u(<
zi^{>d2-mM50H$baLlYr4PtIWm#)|Pr5<Y;8E~qY`_OQIHd(;wkdnXCMM<>=s{7!h_
zn@^g#*!*YPVf>9;)9gX>;9vZz-(va|b3$cU+_+=Aux%~LHoA6H*RhEzAMEe48v}-9
zqy>^F``7s~6+LWhf;H8jxJNxVeGYRT5^5LRJ1yp19iOBzqhmYrgpxz({0P##iLEK&
zk>wEH?(XmD+BLy6Hxhf(ou3|(FafXuP>Xc2L3Z;k7|<+hezz9?7TRLQl}IzO%kA#G
ztCcvy<{doP9b<l7`I%ss-4+c;6`W}ZEuVU0?)13a*Vj;?1@n7lLlt>>`O2Zo-(xdR
z;a!d-bN>6UV(d$XrtrNK#qjQ(qTiUa|GOYohvG}iu1tK0;?>m~g>>X^qwO*<hL45Q
zq$ZziD})|}ewDS+&lU4z80AlNWXIgN=~muQKrr=LHASTG{9X{n$<{Rg>Aj0`%PXSK
zyARqqQxsaBJeuVub3b0gkb5~U8#OFBAli~9t)Y|3>h?}WGCuScxu4U5^J_lV5xyb+
zm~;5NZ>z|+StnPhnG!dIVN2nXn@%oeC7&J{lC}515v*)(zMd*!6WV*f4X5J$MZ#WU
zbzB|$V#r?ZLJ7olO$@8kb%si#JD?gN|NVIA;#-92dt?NBgQ9;Vsaz&N<CQ&9;{-HC
ztowJF;c!MozhfT^sB*x2dbWjwa(j?&)%OuP-~D{0uwma`DSA~TiKC`n>nZ6RosE7C
zE}kS6nLFho_8<#WQOos9WirT`Rx7gYHdFspWJcns8d<pM(Sr5gt?oeN@7q(1(r7Tp
z8{9^0FW+q0=)`p&773EjzN)PaJ~lDGRqq-GTN0e^N~B@p?tZW%CEHVLpgKtdo8;$1
zQA_Y%%*%Jhbu6-M?W>Xbi}{Ehxph-vUj_&Hr5q$bDRv|Fe$wI0R(oX90dy<2xeIta
zNT{D?_0jG$^Fh*QJ-tAN8L^m@wNNFj<G4Tq7T(o=Kq}E?{njZ4yU5YOe*h!?6lAOx
zbvgA&AZgR6Az^IL&IvQ33@M|i!`a9C(Dlk<URnMG-r#;A!n6jqRFxM|WGws=gZ%{}
zg~a4g3=w51wzXfn+u3%Tn66)e?GcYrHW(Z86LWbo`=tf<g;j|@-arQinn~I@I5>@c
z_QrSN3+FM3xH|3*qX#UWOp0?0WZ;+Zf_xWQxX}|NkZzi#PEL~m9Q4&q7`PigA~A>w
z0Vj4lwdM^}CQ*y0Il3Q+`t*Xh$^#<HDa!IN%)&hXJgF=)L}0J7y}CM1$<j7T6B1mL
zL!fJ{-8#9zoJ`Y1ml7z-q;P03@kL0(gpHC(%0?%pzPh3zw4Ff}V7zY}pDqFG4?l2v
zkn^Ba8h|$sL1t_zIq9wLQ(zMr0(M{8O-w1*tQ7*MB{mp}a&ictcd5Kc+>oEYU&X5_
z%(!8fNIYn}HJs$DMz~U!v@c3M33I=}C&mzY|Gg&)uVfd`W#S>z+Dx;M<x?qx0Ahh(
z0ll)ZvKVnjrh5^B8j&EGvd#lzGgIQ<E@0Gv?2;79A`9CsW!OG&0048CYsOHS@<aaM
zl_jW{W#3YnA=oe{gbCI>IOfLjm~49YkJ<A)nafv@x0Uv+x`$r0<hOXRX1tSbj;Hku
zX_;qCOh=Gp=TzW(aL~cKZhS}(|K{NrheLh8$0~c$el<99IsaH!UJ1f2H8{X?1Y>9a
z#z^$}RNC=Lm9@u6iP*%t&GC6f!K0(fiWtJ*W&I~7GZSBy(7RMVW!>^GxY$@(Z_(T&
zX4`GrJY_x7C*b+K98#ap_v?Jx;&FTZ3V^VsnQH#F#LnggScGP+n7#Nh+Gi*F%OEXn
z2pey>U-Ny-qP{xlQvNazHMTgzCC(xG{`XY0%c&CLt)J*b5*FvnQj2+P?sNW!y|b8~
zwsE_$MbActLFm&rB_q&#ia5r1Y+^Qt?9cJ%rpBLs*qcz@a|HRv3huSc-zL5w&g{!S
zdKKDU;(^t=>A6)vLwI)9UX?Es5idS|qOn(&Do2g*aHTgzUsslUz}YX5TPjwc619O~
zcxqBMJ0C^jYvoPQvgO+XOl|kiV?k*caDCu?g`=^zM31tvy2*D@o5xAymRs<HiC(x^
z*kx3NI=S-ApbYc9l~-losB*yZnmR9|K?emG`KzEsd~WjX6vw7``ON`zLfAtRkefWt
zway1hm7xuXnQr!$jG+Y6epOo_n~Z`{oLB|zZeS&Z4@<+Dj7{))u$lsqY8#Zw5|1el
zZ!B~n$a3Lxr_RY$Tv_sl;5T_eR^UHB5=Z{E=v+M1*FyC1(kpG63?=&xrj__*+fVFG
zyQj-DTRRd?i7G@{(C!iyezAmsJJnpLsDhfjuk2Oj1{*wdT=)a$H#+O9_)gJ@iVg%B
zV?&y-aOAxs{?)V)D3M(P8RHWaD8x*0G{$A*GT0oPLv&j!MNQ)bnWGjixcyhh<df|e
zweQz5XSJ;k=Pctfl9<f>i>FL!pI}cQoCmfS&qSRD-%Kx~OmEJ=!IaLRz=3gPdDC@!
zSpo72?Fb6&%?jRFsHOE2Os9rY4mR*l6nNXqgWbpl4TOA*Up)<JF4Vvl&V7FjmlxtH
zDZ@5*k0l%v69`MrHpvTtrktC!w7DJI`&!K7SBfN21d%S=zzRz-ChzE~2$vQ^K!**Q
zb6{HGSCY&=a(4gy;1x+snjaH5xOm0sKgv39i8&G7`r_EkAl^1Bz^|_E)zh_APYnNR
zxqMR9{E%3LUI6FiO$9j7mXG<OhY2&erckP5T8(<aY?!dkFLjm&Q-7r>utH_Q({CUN
zsX@Ltuy}d^@vVYzb_|X%bJxT9p>C>aGjI4`AE*ptR)5)~64l4Yo7Y+9XEbu7k3frM
zB)Kg1m&9tGIAr%I>=`eJbd=Q=!7?<m<i@%m<*&oHp6Z~;=QXVFQh5GK2?{}bpln>F
z(LK27X$bQy9>9Lw31{=jr$r{TBY`3qOsnk0;cpVXwp`s!*~p*1G835qrEvRTz0+>f
zj!#2WtdpVQ2@<)aFa`pF_k^OwVwf@`QhM2%7h~bO7M;9fg}Bb_6a|<@%<Kj7Q9KGo
z?&~Y$(%9d1TPqjcR0Tdq|4q3VpCYR5xI=;YFc)H!L53#InxkY4#SiFSaX&L7OZ9MC
zW_npUMeRM8Cs~SoqGHM^LKh6XX3W0^be!s2>rjg+ijF7~`dy@R;fCtZI6^g8hul+G
zFQe)oAEh>o0h2Ym{i+kf>Sbd+;MT?iX4bNa6bd3caaY$dZ>~=p%<#im<|)ogiw<5}
zUIg`;w(5uU>5}jenZ2$HgLa1r_s&WS<DQs>PZC`XRgLQl>&M3etC>klhU%4kY$67#
zrnxf)e;||{PIw0EmQI%et%p}+1B}RIpzLsrI2~of@+lxv)?694<WKom<|hFaho=qh
z5PGTdYor}ft&W3rh{G52LikhAIJ(b%DLzn;7np}q>IFI0!zj!O8KkG!_Iw2gGv`S}
z7;VCAPtv}{CDb=gkcfiZljM%^Oi2r;K2L|SD)cXTW3RD-c;*B{<xoNFs1bjrEU^~`
zK|5Yo)XC6a<1~i)ld?{XUJ)j&d@li}v07u!vMH(`4B(cUl0J6=+_3yPGE{(DTw2%%
z<AC4@*qt*$YjHk!^G(X@#t&x`Wph5R9Z=x9R(c<2)tM&+>)X1VKH&|K5h!g-!C>Mu
zJpHW@sE#*!j@r6Vm$aEmP)otBf=QzeQn3~-nEDZrjEvCtZKKc2?3K`~tope5Z*XwE
zadE96weOMd02PG*YuWjc9Ao|l+|s?|*a~&7KW{ZyW*0Q{Zaxc+1iIbH<0PUWo3wZy
zZ(0MM|J;x}-r<wvfzDEXR9E;#Ru**y5quljXJlmCW>Y#vJ2H7THIlN_Q*^xX?npOx
znsyx+*)Fabujg?qbLWHX7fdQ)oCI`BTU*Cw!6;=FAwrrO3awxEi+(0f>4k?U15?L9
z#uAkW7xTb+yxo>};G8AeIw~^CUPX@<8jj*ZA1;ODQ^$1&dypUIo4K<z!&TUM^jFZ_
z<ArhRv5Q~)U|u+-?*qAQCO@0cAI-G&uJhk=YTlfl_it<xgzbe@2a%RArR+tzrW5nz
z<%B?dMKF!#4J`S*0}aJmKO`JyL;ggbLfgneQT|${{^R~FT#YX<pD`GQ{#st;Rsz~;
z+milW)}Nc2E!bY@tn8EW20Nb(aAX9(Ga7yRxR?wq8XhIps4`pf+EjIxITn;kwQgU2
zUc`K8Xt+N_v^&~(UFp^PoqeM3!j9xK#c8LyzNr`^8AW5JU0V5p4<*=ycQ`F>rg$w+
zM;_~DS$MJTA-VVCl6jMnDditZ_!g%(%xAm02=1PQzG9sBzKgeH45btBaZ7wPDF1N-
zK;p=ow8$)y>hT6>4KS)ddY%P_u+vZaRmTDoo9W`SGEGQ|Bgy<*YY?HUtk6UQ5A+(4
zpR{^gX1WYO4@>t3-1ek616ns97#o>sVDL9%XTmA#-I8g@dW7}J7t9dbyawd0b>h@a
zNu#Ud^r3$sIN{)LySXp#q(gKV*byR#vsEpUiX5x^Adl^SXXRVQBZS@4hRkyAvz?t^
zeATe-@7EBi`V~=A5cn0=m!M)aT;d+hhbS+L2(@hVBo;#%6Vn=wmY^|7&!O@+t4!L!
zuvdB8V79s8QfVMB3Dcpn;HFtrMFkEz7M?LXNK!2=on^JyBU>>l6>TwgH84<7@vJ_M
zX4=rdN#&lAoq6v3Z@{Bqs8qLHy))Fpa-x%$i&*)Zv^~m7&<%s`xO$2Tio!YvfO|!r
zmVO0sv=7#-rbdc}8PAl4C#E=B51GZG#*%&z9IjuQJYUU#Q2vT^2B$hbJ_YlGFRtOa
znl{tTVML(c$LFrGe?<A=>DK|-sfz3B*DU;Eo{l(t!CBxgaJ?`70)7t1m^M~kD$UHO
zV7Wd6&U*UUxOp;ky>zxd4IM)<4PYm%hr}!rMel{+Jx$#B5m!D3L;jQ!Q!e-7DsU|k
z*O7aRLj{=^90i~J+M9e;UDL0Wjt_T%$DaPp40Cfd`#NIh!Q*)bZu)5x6J(&q+{?X`
z9+6ot#qh4n0~LYgYm$5noM<osx23?nZxn3YP0!1$xR{IVl-N<J>4??dn>+i5fv07|
z0pmX2GoOm>&~5{hZF>?VaN*1?P7P$Zp_snNr_kVqgPR7!{MV%rOT!2XuJ^k+Gc_Ef
zvwkV{OA^+!3iedurJT=oFkGOg_#@eSN=kbNvWoihXsgweEJcE1;HDP}#ko`4x7+I^
z2M|O1<P+z<AQokgeGRez@WC$QOZymhL90E1ocL$a<;MtyR^raH{I`c^&(iC&6mFnK
z!9)1ie{lnCR~U;n64Qe}cT3!x;(n|G+H@t%Tic`f&8}O<6&T`%t)|;@bc)KcW1jhO
z`VKgs*66BDx%nshQW|k{;Z?e`_iq-<jUZ4^iBZ@=n5)<!KYybIURLM8j{J{2l&+Dy
zkHmP5B89gXR#MhRo_oBv;X!)R?YGIRqXzd;6f61A65l-2yCUNo1HL=PhWOfk+d*_Y
zHo_DiWbe;F-{VWwn}eiJe_6V2Yxt5)*)S|eDzV1NbO3Lc>K;I#S4k&W0=#K1EBV#j
zov$F$Rx6smE#w#irXf|ynPdLJip3S7E*qVDBZFaFV`?l;Z#M}TW2PQ=%+o{M<7fpJ
znZ28*ZBiL-`v27bYo#OYk{0<~`rjXmNbk3{*DLbZ|L+U=-`u&TsJ^CjTtYF;q~F!j
z&1~%j*${DOz<gM4lavj+)1$~G1j>5Bwes0=dz5?K)JbpaLg30eWi3N3Rau*9EkQr+
z!L{PQLtZ?Mv8F)Z>!`&2$^Z@hw`b9-w&|x4dAE`0j4itv`YgNekSrU`g!P^yqRu97
zS=cc6IU+DG6Y8yKOKb{(4l&e-&%<q&McM0e-TB;0X(8s<S({WU`BPdjtXXg{P}ZA!
zSZJ$`ztcSkG|zAuJHIsS62KVA_Y+Va`qp6*%0}D0h5Q6u^%oLl*<&nK6Pt^0PE8K&
z^6Q_d2PUT1(oob`pyr{}ac{W6US*Br!w4}ChKREvTuG$HpgpJOd>7fe=Vjld{|U?>
znV<xq^#?=_4t_jte;)_otfi%&zPGT;Y1H86ZRCA7-SmiXt<z(ZBvy!ujfTT<MsxUt
z2;}{2(E)RI#6=)8wgu~^4OFq*g>AlSyq#t)aVIj!7u$A&b5l;$DQS^2F6A&RGSo-X
zppkz<Rzfj7w}<O9Pl}+da~pVYiR6OA()1rkZKr62*C$2oQ>Px8peR3XQ+A@q*pkb$
z5HLH~X+G4kwt$-!7G~<#O2fJwTYE%ccUov?H8WL9Dh-twY*IF8MVS$2B1gN&fT?>&
zfM8$5Suu~1{|=nDfUy_Z&Je-<;dfWhV2QUTgIx#i#~rSiheHUU=ehxv*-Z0AY9`3A
zjB*ZY*>0!16Csj7s4E@!r^9$~Y8^Klc1}lAREol5JndDD1aSsixUEg#rrF}c?k4Xk
zDh6^s@=Tg5uVs5D;q6viQ<fF1a(OgvArjO!fph3*ZBQXhKv8>u&V9@JlGA*a&&xqR
z!gW9v0hOld^O=T-;V?|ADz6WC+)Wuc3FGc{^BDbnU}1t%5(|><zp+=-&fCblJ8i%0
zzQD)eX&*iRjbc%EYNFD*H>udGZ=4A>mBd2YTG!Cr8%=pgI^^1RmH)$|f8(6?)qk4(
z3!E|DX9P*h8p_yRdQCmew+2&OqDd^nZ$jo_zdRJdRz1!j$ae1jx|BoBncRZNa4H&P
zmvMbMb9$clE=W&eUdIQ7G%a#!M$x)9{U=iZ>;$N}q$&V5nC&e@d(V|ubZ<Q7RnVG2
zKqhAa<ZzEV_vD(Hts!DG+x%ogwap}YXb7NaR1c;hUZZTfNmWPtv!G|9cNKwf?o54=
z$aYT|iFIm0aQO4_r;H`TE6*iog@_D#)*B4?jb4Yz=)yJ}%Pr^O195GG>R2bO%DGG_
z_vNH)mGA|C;TxJGFJ+XKSj*16nHRjXI9<Tg!Q(1|%n&XMJj144xO_nQ->i}U_j?w^
zu)Xa6^)iH6TG;2YL*g%r*y{iK%m3)JMShwQn&9X&w0f4Z;(v6f{Nl$dc{Bcf&;~$0
zMs6D2gs@qhM8l|5K~pkjdlO2jp>eQEzT;$IQ|NSGWVU|+y-_j(6($v}FzzD<rJj&G
zHIS#JCELE8qupGP7EU%))7kAOxGO~w&_gyHozwtg$_3AX>Nv=@QOVOYcn!1yKz^!j
z4lmi1;o5#3OnJK`Oy~uWNbK(*Gb#}mj(H}a!dNWu-S6@HOb?0^4mQZ_mk$y0<Ipw~
zovbV6voLmwhP_IdvADacgcv#V9dffJ484MLtV+ZO|AJKI#@VW;h8vu?DBh&<!2;uR
zU1dQ*h3b_OG4HKF2zG{>Xa(jG43^ldB_*VW3gR%1f?mN#kR8v>96oUw<J0O1O#B)L
zfYY)m2gkdUVGW<=_k*hro&$Y$v`2^y4TV!ulvxy)4J{YWmfbaFSPF3BduuYgBxk=v
z?!$TdPdYd_0MlKzn~jXRn_{p~M#XlF2@7)Q!EOcwf7DXtP`6~MZW_&F$y$&*uY1We
zGmd^Xy^gb#n2+10HI0P<pj4d=RukeApE03Ur0&YNEldNTiQh2^)j?OJ76OM+7@g&N
zr@Z9;W8QL(y(8CkbvD_DqD*U|BKh1@>4-?&2E5S6g=TO#t}w9&X=l?$#Z#<JN>Mmf
zY3Jv<xl)-31lC<CG}GMYa;-dmwxVOv%+WPe3nRPhpv}l^a~JN_gB-cL43tY@*tkZa
z1mY1m4-lSviUjqz==OEU#ZzKBo-1erC3yxfdx>GjGMf%Hip~#LW;Ua3urwz15#skm
z_YSzW-%TYL7z-iY4Y$0~hiGsF+NW$alSO_E%LZ3WflZ=oy^6`ln+=bon6_3j83ttW
za1%2$tZYf_J;2$69*3vbwt*Xbn0B_apnSaZBY2dA3vY;>#l`GFViQ6z?@R_if#)&G
zQ-OVUcTx#pJq*JJu*Mb`8!oR&F01%BnZbMyaHZeSC#YsNcXu}}RDV9qOTw?3`etiV
z-p5iOapyL-_oU(n)=$;>+dZ2<Fnevwmr>dJ!Ckkz<ZlXuwP<p5&{6c;w(SVdGyECN
z+;Kki(EC;JG<4rUl<AP#;`l;oYS>uhrT2$|`xBYfRZ0y>O6Zre7v?D^%RTzPE%ZO`
z`2R;f0DJ%`GBcfA5&vsUf`#`D7_t4_&4~%w?uG)EJw?%-%3}@9A6A+(3qT0e1`_?7
zn+IFgMs5goX^&e{a!zI8lDqReljl=n^pF>9lwwZO3_J0i!AZCfyR2fzJSw-w!Q~-o
z0-SKFDeO?oEnD^(`9>yVxvR9cjva&&e4gxr-F<;y-Z6yllpZzWo0&-Q-2b;fi+*p6
zX#O9`*qM)`WZ@LUe=8K>>5`fnfXkrwh@l0Z#JiNifx7Ac>oM_nIO+fMN+Tl^*&s}&
z@jpD&J{-f9PKLoS9V*sZqT>Y@2Y_p7;Db^FV{zwTbumDjR7M>T)V8-NWcqZeNIIL@
zM%n0Es#_YA$*OMTT}Q%f8hyQ>LLjIr666A59t~XpLqe^293k=yI9^RVLM501osm{s
zs}AS~DDD6kwNfse@ZP>n1_Mlo32-iP)lEPv)xVYd4itN47Y~#*SNj;Mwi}x*E(>pH
ze866Zgo#Xc3Cn5<r;4UHOQV|nKa#E#bqax%QJaNeXG3i-)cP73JBNq%4fX@~s^YXH
zS7AWRj{_fp!C}$f!l8O#Ht3RcTiPl7BFIG?fwpW{o3bd5PT+$J4!cPa1B@&g+(AT6
z3>B>r!Or?QmOYpZ@A97JQR(?_TfkkMV;wgRU67^X?GllAZ73~r>NZk`n5soy$S?>R
zQP#a&Uue;gnO<Rs%0jI7K%!TK3Vhh6Z@^6~Tn0+0?+06}5I<`sEfcVx!FXgj`lTUg
zM+Tm`R4FPpyq`!IpxdnjfnOT-VeEpSCBSSmpL>no*lPxsYYjX|KM=id4=;C$vgTG*
z&D2T1K_S>r47D&Zc-1VdM;y^ISiPBF*x3dQkV3|<{<+9=MuG%=`X_JIE8~K)UT|o{
zWdi1VGYc2Cfb(Vtah?OM)d|z~T71&hv_igYI?5i2CeBwsDiIG#!1s=Vf>H^8svq3w
zIL=GZ8ky(;H@kKp%VfQ?J}rPkF|?-pK{E)X6;LfC6cf{GNP~bz8$T%CcX#&;LGa`h
zJYi#NeH+Q%ofZYv<0p1|>X%3Q3o3o|o4|S_tlI#mSX@;i?h)~U2hwcu)J&id)ArV0
zm$N)2O*egS&1`iOsdRcC0~QGgI~xue;He*d7a5yb$8+PLE~dH;y!Y0`MMw=afKDDv
zu7jS#K0>VTIp$R=BF$iYRUC*T-Ie>!ba-1+bj)eD5ze~wEI_1rEbAMCDjf<yL>dJ6
z#9ZMl2WI9t7oSx&6<MK}Nf#kCW?54S)V{tsHG)5v>C*;NXTpdr&e+$#InU7~)Sghb
zUs5eL(1<_2>fv*jbO*IXn6_2K#9W8YOfNx%A$;qUsT!JH7$|6AFCSA&aB$)j4L5W~
zk<M>uJoxq$EQ`*#y=qHIh#;97Oke_@N6^`(#7f!`%F}voZJms!^$XK+_|qM2!=2w$
zH&_4jqj-UXkF7du;uD)cUGj?ZD>~#+@CVXXwIt+lX*guuZoqj2ohq7)?zk01TZSf+
zA}7uU4(49{xc3&x;&b)Y(kJUy!j;ZhKO;`4bzs)LSnlA7iz<%K4?OLKQ~`*}+?eax
zF*&r?JGZa#T3+(ddqiXI9_f5S%te-+cKV}ZP-phe`^q5~vinh}!bzTtBKekSzp@P;
zQz;_n$uMZ6RqdZ$_aq8wHBS~i@4a|SH*OFb)b{)jSJ(f=<@!GM>cYFa&?$ob3-W#D
z=tQKw)`RpKDGa_Ok#8edOOd&oTxI^hIaI2$YoyTV20b0HrbnO?!{3D<_Y{eZB$O4W
zhr}XG*D$dE`-lJb|AGvh@4}nc<-e&FT7{a~T)cY{T#XRljkK-+2m>XvZIiPZzKG0h
z$IrRI>8?ltnGSkGqvDFNF8r29l%+)J=x`q+9TC1_pxTG*i!9MEFJ;lTA~V<CurLyL
zI1ml)lr=v={G3=@om*O-!v}x#6O+d<yQ2$Wq>2#aO=GjD5}6DbB0R%4=N5et4|k7K
z*8Fj^<Ctf7U}JL~5V$T9>=d|nkYu5-*Qj?xy9cYGmd4q@O)}IQ?NhJv(!zpYSM<fQ
z0ro@SM{urj4XMaAEt0c*jg&9#{)ltzkx;((46TP*4rDCjsX0TriMj~gp`4rymD<7{
zQn3WOXx|Ne?{k$@F0C)mQrkO~Qho+)Lifph4KjvylV{b1GO!eS$jEO7DD~Kr$H+0N
z72d?MGE#?Xqji{Bgfvhf5Y!-!r0VI%(akhdSi7cXfR5G>6r6)CpyP%zF*UF{A<}-C
zCkefRrj3593M?(rDjcqReVG+hykBI*z=7O?uLqebWxQP8Gd7fGuT>D4tW5hwSj;g8
zznYmDbv{^;l#nbPxJ!caa5&6jW(rdi*vxX2Y|mH-4%8QCrcKYW4A=8NoH*f)#lsCP
zP1@j7fC-b?`Yy7pFXMGaTA>SUU@`-5m-5DHSEBzYk)om(;7CVDyAEGREI3}1*icuo
z8V5rqS_<#dJTPtWTlw6`bnv#83S7x;_DVk1lOk{v?HNAyg+I-rs8pq-jS7-Hn46SS
z^`^-tpQ4+9@^R>eQHTL*XN*iXzHz|e(n$}z?RD#<Log%Spil}2dks;)oLm-FX0Kfy
zT@jsR8Up;Pu5ewQgu`PHGgzTCG+mu|uB+3nXP|pdN@;S)7dsdWk`$>IKD(?d#va_a
z=*)w#;(AhkKa`XnoANruZH`rxrN=qEYh|y!Ng2>b1)><0@stY=;JV5!eL8pFR>VtT
z>6fCX>l7we)jK3q2c61f*IKYYxR2ic5aTz?7<b)zN4o-}H=W7fcFj=UasY0aMG$%U
z?~EQl<CScgLU!t2%u5dG=3Yymv!0Gy@%Up+ex_F#i?@+@?z6by!EVpgYYs(^5~C3O
z6Mj{N4b8<{?in3pvb|p%h*(!To4Lasv%XC|OHuNF;2))HF><qPJZNzJHR_9sgQFmi
z<Vhf0c~~Gee?2QhE}kw;6toM6mup8$$AztOlGdV5I>uTu)Nnhs{lwwscQXj6B*}>I
zZyt~Ym^@=^qGv*txStzE{O!fV(kF6TtwR43vwm^e)WN)>?G0TK3+CTrcD!c4Y&S@F
z=j|MDeD<CdI&bD&k!c)8rQo6%V;mIZ73T8U)??EVH+fdAtd}9;oc0No4C^bqPK7_k
zWe|0BJPoQ;92>pNM}~Y#VX_uuG$xa?O&vIR<#2SCtoHLN3vQ3gLOdF>14>6)$lCp&
zR2=OMw2!n9#T2CY@uX?Y>y9N3UL&j-GtNKOVxa8x`4lrm>8iJQ5UW91+<NHj<LOUo
z`y6mY+1ATf!$w^~OCynxVcy&*M!6A1^OTG(ej#R6xJ>EL?@UM5VcEHGwCBvy_-jFD
zi&*3c<nE%61LSn(c$V7bhj$K$pY@TS_kchzledxKBj8t`d{(=G9X;Oo0EWE<fB&}G
zM?psdxh*o5u?*cx-nI9TSG(<<7siU!{=b=fN>{B4N?NrsTZB<0o27`tgE)%Z5oZBG
zC=U6T%QG~eL9;IWVv#FUHFWk2gG2Z19`?~aB#@e{plJMcG0w=!lxAz|B$@S9Lvf}E
zQ7E`b=!r;Jys7ynBv^)<9%xZ|OtZc4FNOafACVCueGIrP8t4xj0`I8utc7<z%@qR`
zh6n8#kD<8-nGGfDh%w<^|Elsd_#l|HZ`B^fn)Po*m1nN;o{EaBLF^poml6MEjNo;P
zIt2{EAAjG4BtthDc?19LP`e@(jY%veGmuCd_ldNPVhF%Cu!;hd2N-waVT0t=lVd*>
zzXt>UoZ_)w5}63rcUNX+;O&v4x^N{rUteTt8{XqI8fi5S%yb_dbBJbdMIQ9VrVh{z
zwj7zYN6M45<eAnL)9^7G+oDJuPp~ay6M(y*JwVDyFdWq$kjwR1LloDQ!lXE=b>$a%
zuB&GDR@kdr!u*o<U}{1PdD6<nP(x+9&vE(^8~dgA`B3EqfGHJP&;6Qv_~XEJSR>rb
zhQ(Ni%<KS&wVuEj#^(*_EQK`k;pshcHQ)_1>`uAta6<q?VfhTzt9T|ySCx!n9=kky
z2E96OT4H=*Q<p)Jz+T>DSjcWqgOhSVA@!%PCdUK8KI=l!F!~A1IK=YtyBjjUU&U!~
zu=x`x^lBPlz-&8{AroU16Yuec0~Vc?6HaY_`5>JmFUmQylmKH3!(GC(l3fEJgY6Tz
z8V=Jv$=%y=v!yIwC74a9D?rYwLFeoyL38uWUc@iRH}Q}rFsgL1Cbwj`?m45#F6fDo
z0tFoX2MWlQt_b=$JvewVtJmmvtmf=ZXsUe)w1s?Ev0T(NsV;k?PF+5vO!ae<k7;7{
z_ME})>l-XBExT?Q4R;Wy_q^89(y|D9D0yqvldI<JdAbWXdq$$MG5DE@@Yc+Wg`cZ;
z3Ul+SHx@J_-6mf3O<mdVBX81w<gx1XTxiUvS$nlVhy&bYJy{ecHYJC_47f)KojGSy
zD(RG*7NJU;Ps_fvEnG71!@g}PWHyLYb%1heXnpa7YT_*c+A!8$-ZzLU+Pp)|H1Cf~
zxEuJgu8aHf+GyIG>|)&v+Mdd1Ma-PR_Ox{2xbDTaqhA5?_qWlwaNmQLkHm0b?GI!p
z%c6Gnm$|?CX-+3G>UiOk?BubIUr)sk-l0BjVTP&ZmZ_^bxyi0*Z9&gn&iGwylpd}%
zie~LR5`1d%&+M>!UT33o&Kn!ozi@N=J=^i%r~Ls&onLmYN*kaaHcPw0^YKakAyvq{
zv4AB$J~ES>C*hBO(XD;{dGse(bp$%0X#Wj2>RUeK4~xrRvm)2}<TyNo)UGQB0vXC1
zO|+z=UzASv;AA5kklSif*S5#ui<Od&oC=h-7b3v&<?bi!HH-z(B>PoqwXc8<SZ4)+
z_2g_zwf?<_4<FLw_WbHYWfly(q|`|z;ZaK=eRy*GL5*UEIry`M2)(L!alRX&AnLS*
z@Ssah>F7$4a=_ZkBwS@Aj@8zB?lUUD@FPLLv_oSznHUK}8JZ?PC^>X^-dWP#E^=q!
ze|g?o+~y*A`VgO_nQEO`Dy~v_#)^(yZbA%X<R<p(9rU`F^OV(gLTjQF9XYs9`2PNN
z`FQ=yakXnuT)*g1I^k1*lYra&iGxE$zbkbs&#YhZw##Y=LkyG3R2&Yq9QVlxf0n;&
z!=JUh;<WxUH}?{l{?kD#S=?H847cU?d9)tH0|bwn>l+CVO$r9T0&WfB$-g`YuF}R0
zUE`Eh+hxRgsQd@Ze4=f(Muke?I&pZYyr{!Fd{ArzoRgEGe|Jb{^+#OAZb0Al#cl^>
z!A&FnzbxoTNi?6Lv#!WP8wQR-I46hJPoW+Gk$_G81>mOpaT%V_$y+~d#8^;>SnIu|
z*uOFs7><j~Y^EtI4E`u{fqq>8AR|Ki7#<%0MR^pn3j2ft*H^`1x~uQ*K$8L{efwe=
zF+7kS0dPV_2%_3`I0>dH(^y=Z1cON<Ba{1`7)MY|GbAXhF2TJ9U`1dJ!rI{131Il`
z=D+@;+!-@%tHvi})|{pHZMP=-7i7!_bpx-zFH%nRALJH8FZjQ|fd4bwhJ%1&X@)E6
zKx08h$6qIqnW8}~s{d=ZTV%wS4ZyU<4WI8OpAE&>W~l+~yHpDc;CdNt4ec$M%Fa3j
zP)ppmEr2NL=3;~FRqJK_#Iux*YX<I~w)@%Tjknz&h*k58P)izmaikiA^F&}lMmh!M
z#>MTG4uYNWyk|SdDeHuHsTga>nQXH#Yv_Zp&h`T|JxeeJt1p4`ZW3c#+vpJGlYTW<
zI7-e`gQGBe@(ptO47nJrq*up|4B72RswWZt0QJ0NJ&Ng>aZ6j5ebyx+aDGTA9-CQ;
z{JYKrBg2E8bm|ktiV(qpap;HshsaZO3=-d4ht0MV%U^>Pd)ZX%?oycT4p%Xk8Jz{j
zZ8sYW82v_YGr3_b64Zek&M+8&t1lv1M6zmtROY)A22d9WGYbIYT^NU$?~p*lx|FjV
zrPSfX-8DGe@XQ|8yC4Hsx)eAvYTH|I4?u>FcLa5{Z9K3&(?hD08N7vlPb&<5r5GZt
z&n8d8`D$%&lUvyhoU?0+I-*^z!Fu3KNXZ7L2ZzLCKtGdIQ98>}V@8e%8F!w{aCF9c
zyc6@w7%*9~bfT|~+@W@zbOt8IMv>Avx?3kkg>~~7j*6$qQ*Yow9OEXV9L6OmFs*d2
zFPVk&3Q8xH3?Fh+%sK!;Bxr@sp#WB9q^|{>j1H<-+Vy3VGT3PO#aLKY-FzEK3x==}
zW^l_XNPtn}wrZi677a-rhAO`58z86bZI4LlqM)?BvMdyOdL2_aLD>5|LR!rk(}jNX
z*r-KQb>Eim%cdiAyHfpzy3=2_+9S!<(zwbHLpUjy9lzUeWcHy$H-jiERAGrDS|9*W
zOIGZZ|H;{tdD2&3!Ed9`i%BFL<dr+H=9nT*><;Tmlx=j9VM=eA+2pV?Z8b)aKt+lq
z()d*T)KgX<yDS7|REWKNIWY}NCyE16mGSpRIlJ+rL=o_;(-W;suG(eh6`fQ2Lwveq
zW|gpwr0TcDOz?_e6#nOy<U?BO+2_R4FqAxe{6tIl+t2NO{||ZZ8P(L<zWWC3ASfWP
zD56rOD^0p6y#?t~0*cZJNbg0ZSE-?Q2uXlQ38ABcQbLmuLPvTDz4x`}TJIkFjCYLl
zWsmdi{PQ7WkPsk~nP=Y5eP7q_x*sxV?(v)li_NC@a(vd}#dkR^PbQvhHIrm}W|?Sh
zw+}0)6M|ZTFYVWagNHNtpgNfW{nGV`z*>$OZ(_7M*q)z;-uGzqWwqwf+gf{q`K=oq
z^=uay)Rtj2)9&T?Jh<}kQh2CAO8VG$p+s^`ovk<kS5AvNHFkI(Gc_`B*F;1_?)F>Y
zlAHfgdd|SD^e6d*jlE~fR%%5D^1uS}4FDD$5PLImos|Ir3ZWn%A7Fuxz49A`Gm(*Y
zp19HFR-{WOHV|w^X2Q^X)3p*vd0T91PZy%S*|77p){5IP_p<)r3qKv?8Bfqw8MB+*
zod}MD+f!Z>c;Qn2O$u*Xb~8Vp(wThT6@U#pB7_v`WJimA<rT{a1XUOeIJj!{0f}?G
zP(ZVPm&2vX{g0s^(sX2=xM?aPvcz6kiobf(o?*<^dKTNh1R8oSZoW4P8$0_5#fstb
z+M_)2pFZ9icRK+66MJN!IU}a&%HCqk&lM65Yz~y`zLC-Ezab~dM2Cm~*YK_5{q-wq
zi{N0E4??uKt{d!fGU2=aK~I?b=686-4;hPZqmFqp)GJh{7OwCe2@B2TDD-tt%ZUwt
zYs5K1G^3{R`llWnmzKgXwVwuUMsxOTK*ZPqKaJiPujeIOt3T+oWwbTX^2wFj#kCh1
z2Mn!o@=q&_KL|BCD>uAI8mMiK*bn~~hxVhv*bnZ8c##Mbc?ul;!i=k?xuHPj^Fw|O
z8P8=LZ1*&Ly#LumjqTP7<>xPV9(t4aH!@{-l1Qu+4V7mz#9$vTevv-`n%HthXkaxY
zk9>tS|37Cfa1SKj?}BdEE?F-SBgqQ^K+Em(X1Y#5tt*{a21I|v=RG$GcyoO<t(i73
zpv7WR=G(nh#IF1UCv!Q=hoJYtZUG0xrmy=Xkjm;RIw;dKM2Jh8kxpK|0tn>fBeq2Z
zvpuLjRiG=Z9Zoou^%Ok_2)XfNu%2djO;p|-P=h9m0M<sJ-_{k=R?`fclsA`a{&rIg
zWuX%E^@;~B&!$Z^<v$%^$^}zk$@2jf4%d^GfLWNS$Y6rDY{cQujU5mNwhh?%6tasv
zuiIh=w6!{%ZgM(#)a|4b5V<;&yZkQ4?8@OB2=1NeF1iR8x_n3x#RI1k73xSQFVyf|
zT|EI>S$2Mzo<xLnB4ZiZjjk9)j$QlqML?1f5ZMcwM_c?8p2;^QGXL#7+jG8)?7C`a
z1gvSIjLemnT}YZKFK-ZpdO5%C>kYhGy-0tyy_SsAlt)p}KN3Sad~&U5Dd2WfERfBD
z_a1;hP$Q?bkY;FwdVs`4mQ+A?evZR{dxNMaBo}Hf1EMv}4Qn))?oH5fNdXQI5hmZS
zK*rGcLB2%?y8tq<ldpvQ4eA1Y))6^Yg$q>!AegnLKpRjzFK~AH2hRkAP^iVJ0#hPR
zrU$x54#7{zpJ@%GWgfWBF3u8%0dZX?WDZ}02DG)N_64+&O}^fHAo_xBzz}li3zR8#
zbIP6InnPezM?=HN@yB-YIbhT%Q6~v{iY}5suhd25B>jEGK~&u4XCD8%fla=v31(gk
zkaWa#fx>o$ACnIPCl!yjrrri#8wkwtVM6snf^nlDyWF4dBzAm12<06QXbIN<RkiL3
zPypy@^(J(w!d+axkb!5981G9S0CgCCerje7P&<AY_}UaZumt=<Wx@a|Y8Rvqe~Hi7
za^D1;8)^U;DTK!zi$k;Had1F&%^fFs<afaua?if~cRM=attT)ER36YXPH%m3H8Bu-
zWx>zEB0-X@sq2#QEzB~jo{TeT@c6-7$DQ4WS2r_&OesK{4#;p;6LT|NIfQsR!P5$4
ziM#kjMmAXwA>};~FK5Q7#~cXdff2IBLBvFeUoRR!kJRT(X?FJg0Ol=ceTMf3Vap*9
z4G&6jYYCLhPGX;*0JuV*1jWG-a5<eQPtapI4*<E*1dQk=L2-A!0vaDqroGt>;>;Ln
z5Tqw2sP^sdzwe84DEKNM?nVs^l!K-Kt#dmXsC+mRAmYLeL^O6Z<Qh#PL#gdRmw5=S
za%X1!sZSuW3-n23nnWV{cejszySFVToe<x+3o^=W=ipeA6O+f!0s2EhaSrzR_Lm2+
z@{oI(C#&tiiG5ABR({H_-pdy#Qky>wHa9EKO?ASa!)wXx&F_E*)JdI@<WhOUPl4Km
z98i6AT|T^KlZ5{WW2=$x9{>Vo*J9FIp$fF>x-dD)uLnS+X=CAh3*3I~%`Nr4vv+v#
z9!Qek@^j~DA%5=g;We3PA%GrzMH#S<-{B2;7)cm4^>ll9HPZtR@jJkyfCd&7fj_?S
zL<<nQ2in*`-mBOIeFo_kz;ILlPL@3q^zqs<33)<=hkroDLq&rabuXF^ARRhg8o=Pb
zBeEf^Kvs&FMR>d7u?cqMpE-ougd%Xj=|{tqo1O6W;323B-)Z?GgeO4+&}5Q2@q33}
zAP!$u9HES?Q#*(e5apleUte8OZI91%2WmdPDP6bcCLtWIkZ-^m2=AyBnr_P?=97yp
z*vnT+KX?EjZ3=;e;-hBx02W3)cr}9`T*>O=SVkTnATEO6OhDIz1a<tQwmIZ)3@BXP
zTvOhz-yV!TZO+N3@R0M^!dp}*&^x2Bmd_so%UyZBD_$R{bc+v2^Sj=FF&+5X#?;rQ
za25cp_lkRj2$2BK2XT_xeQK8rfCRKxSF|M)nGS&cj(xzN8^C@c2{Ye&Y4L_atM{0+
zF}48Gp%R}*XPqd}Y-~(}Dq3p}foXF+s5;R4TdnBx{MI+u9^T^kr(OWn`#qq-%TeZ0
zc?ea@w{8wzb*6z`*#lOP6Us8s)Ye6U3-&vRHkm_oFOcgSnQ7!e;%W!k3))D)syF2!
zdh(5?vEq*6ZsG%Y1aKB|h~O=xcjdR7zg=CihTr$uVV_?EAU(cv;gDa-r7ev>IF&pP
zu2%ur_Ao})vM&Xi%M2`lz`0Yo3fD>;N6hOThl8d<fT3}plyyxo%vFshuOs*Q72L$%
zpoOR#e}j7UEVmNohM|ZSw?4r>uZcgdN~6OIZS<$PLsH;}!0L1BpWR5<Z;6VJJWc5D
z2fK>|;(lx?rp>r?s_7{3Bdc$2v)`}k(JU;`6$9dz|1%E$|1sc~TDr~L^6>d|2(UZW
z8JI%9;n{M2_<=-h>S5ZP&+{IpS=^2A{s0_{e{NTzWX>Z2#!}vylE~Jq<(;Tfz)|I$
z%XMvAPf4U<*TE!UZ<CaP@k|ytJoTLKXiZt24w?+T6iDWV!^M}0^iT47*N=JVPd&R`
zO752+4sUpAPjAXOE1e`7U5t%M2FUgd#}5q-phL=HE>cIdS8IN&xn}F25>h_WJPW2Q
zkIwq-cYI7w?-F<hX2vNJ%YO7AoHhwN2Cc)SuoS!@jju4Af1~H{K(a?2-aRg!T%!de
zb~IG^V9~-EP-TSo0r?WFk6vC@2G%w8+sno^El1O$`taaNTZRN&BZhd7_pLR;8&>ss
zXthuJl?kglM&>9)EE?7br|NdX2mb`oj7&BJfS5pJfm{W^w5!}5BM_Ae^WnT-<^C0b
zBnVx8buf|IdS#FVY|l4_pP<U?T9SJ%CW923)!Enf4&qEIv~#0AsAZDQQP7E?&_-9C
zrv(p0hPACP@qzMXg{wbhaSo~`epfH}G4!ZyQhiKd85aD26=#d^bX{NI<3<9lGx$6<
z?C`4R*>NE5MkO5m!~nbIu-Jx<_DHO3)vT?pSpzJwYo~!t%9g_!(24gJFDxSK02OQH
z+~~YP_2>|=k9!#cOdrlFc0u%B@1vYeghN2l8B?-!oIAasiwD#DAVU?N`(r$lAVT9p
zhMU25tAnixT|!bwHP(&K_kvuCvH2z{uq&s_?sf>M{7Ybf+&jCUGq5tgN#VuAM-@sJ
zrwY#~Yuf?5E{A)9JX@8x=me~c!vhGJ^VfNjiHyz`ulPf8q*;Tubl#k8F_>+=a9f}V
zb+1T*0vqlQpkK^a`t90CAESJ**ydHH71OvPw9{{)CW*!GKHgjG6-kRlx`59c`>Jy6
zu%%&lYw|5CvMG7!()+|a5djl+$?JdW&u&WWcANd~J$<JCwn^}C(FF+2_{YCY<>gyx
z=5y@IkqAvCctNqFapDVN%R(LFtr>eK>5eX$>)JC|oV{+@{@`VvBrY4}F`C<al#Qqf
zjvq=40Wp5@gLi+Uwh^3|N__RNls8E*u;GZ&hg@up1cInLn5)I0U=M_`ugSeNwZ--h
zLO4Fe#KYVF!5-c`3vB*&Yiez{w-z52!cy`Lrw5c;xLp8Lu4m67=W`N5HO=hc=h%(~
z+APk8OeZorwP&6HOdhz9srao0eYxV@VgcNpmiQ@9O-rw#0bj%`^4l1a$nIa{RI*YT
ze}f<$ZLnrCmDz)n+KrQ@pf^{%u@L{8IQ#;!oQwr|D%+A{E)H<@IceB{G<xg@av)v)
zy6FK!%i=nKg%K1x*d`8v;%_Zz$aySwkQrr?D&dvcvN<6)*avGj*MZuT<}MkWO3GIf
z#PRQ6K{skinP7>I3gH!FXdd=Ea|N)7Y5=BznUQ=Rafw6s2jyRNW(-CxfmA}x<sVQ3
z<*0$F+u-ruaFgT?P{s{yi0nQPen_r`5d=|5+9OUp=T|{uiyg;~0OMljpMuXT5558v
zbrv<dj^D_hI~D-ys}T?rkoOC~5A(^BUwk-wcOnK{>XQT|y*yV3EZR3oAj7eN2<+W!
z%&vID6VHF15<7Ely+K%fcuj#qt05P;O9QmCgwS=f=;tsXEz!E!bghtB23%rpVup|5
z7KO8m!(_S)IS$W(0_1N{&FmNv-I-6;bfMSs@D^vH8?hS=x+W7E5X6`_`i<Vcz`k=4
zq+bt#51>$A$z&SSGV=g@R44!!AmfBI`yA}-f|waag_KJ(y}TGZK>7GC92!+)nLGyq
zFL$o(y957wkzKFzKNQ_uGoyi~7a)qr$cDg&xA4Gq`Wa<gUKNPLeeU2?3dF@VY>j@q
zCnzI+0-|e#s8}z^-6WI#uz^o&Y01BdhloFa>b5-k=GK&u>~U_hWWYbYGvi`1t_ArU
z#3|x_2(vdJ^xpmhikm+0Q@8?dbO4IR3G_VZSuhg#HcnH|j1w)82x)#nsO_0W{>{OG
zs0`R!Mq6L6!w{aAyE~+{UC{O1Jz<_Z&r?7^ib5U?Pzd668s&W4iQ0h;AuMeK{vnVv
z;ATgwku^<M_<Qc1%6=o`X+E(ZfQYaE0ih-$$#bUn<~u(E;x@sXIQ<g^a&;Pez?1_L
z6BJuq_ydA}_$T|)ek8~pM8yeElad9cM}ibVhLfh-B>6U%1)%oMZOoIC@&@tL)SxZ6
zL2NCiKv`<t46&ZV0gx^+Ay!;k%lS8I>rS265GpBGV``wCPsUOVNFdqd0>=R&Mp%2^
z9UgQD@Nu~lm9dR4#{n%eGNqCSu5Qm=0Ik}?fu{L=5SMbHW;xD6fx`Qa#pZPWwIM-D
zg{vVCyAlA%{E)n?CILu&@R(jQWi+<;I`abYyd<#$ft8#f;9P1MR~|9PfNo+v0T`+~
z3HV^<1A{#nAba7geO@Q%?|TN|esX|Q{7h$&v<u>$nwHb*jXZ5U4wUr+dXv=mvjN>S
zNNm0m4mf~-kM?t5gQi(~umSA7E2jg1wkpT}cH!PYtXL!qC-(qrgOch`)d@JT;Q05R
z^&Z2Lzk)=j0iP|fDYu8y?(Gq=Cm@<e)y^ot1`tcw%o@lTUIqfa@H!WgQ{4kyH{G6J
z0_iu+z>TR{WzNnN=<TNiYEJ|5$s7@+(%e9WyY?KLyTA*_EBgZ>*&*cM99tYXpgDw$
z2aKO5?wt4oO1_6Uh)hc@(9p-DuWS%M+OIB`k*h_v27LwLEO|N?0jtOh&d7s@S2GUm
z0{n1*XQ>#E@LW^vvwZ%BJn03{=gq!@vUHII*$rZd31wuQpZ?fRzM%j_!wR=RDq)=w
zK%*Kw%g{V9focYEGKv!O$TzoU2WrVI@}7`OpVl6%fpX-bz-V~2Q`-=iXAErnzz2ST
z^&t<F;U;;ogsKouT#vaYK#m38{{~<%ct(k_$jc|w_+pEz>9^er;bjN-sN0L}=p_(%
z^`a`R*;8cmaBcq}gnvp<?Zn%DqX1Z=K|PH#+DE5%fz*w^Xy{qf%xiEJiy@MD9-iRQ
zJu&hN3mtdsi=DhKslVzp&Cb-<@?7+30%;88oc$#iBo%<liAP^Z<^4YxxT^<%6eYLu
zAj4PYx_1RZmv&?_W~PIlEnBmDK{%_e_Lts?aIoKn?+1lf|IQ#5G`A+Ys4qOJjlQx<
zx&uw`+k3N3oBsn$N&kmS%4vEYpk#=>U)Oa7Fg~a*$a!2fo8bcQW4;|8ZZQtn@q}lt
z1=QlvmVaWG{Y@mM&nhmz0AQqlZ)88SE|vXHhc1spAu-T8J}7u!){U;f-%U8%YJAK|
zX{XN}8{W^I)}6WRX~wkyqu&}&!~Jo4mApO|AUi$z-KoKHAL}B1cHUd*;HH-&$w<dQ
ztZw3*%(KK6kR>wNy+NR|yrE!7;EP|n4(2{~!787E?VVuCT0`f)2VM8yx5thx{tf!;
zsUz*zXVk9LIc`=#v<wFqPAXx>Z^!v4w0xW^=t{mNXqT1BHn`6BNIP{~GcDi3VzmHY
zH0sT{x?=GVf2!$wbqxqq9O16};M(+VG*OPRCW3zF(NwH^)_65@0}Z_h=5G*nYE+qL
zW-2&YC-;5F1($Bk?#6qgLm@+edB8F{^4zfB(eN$mx3KlJ7rnaN?Uo=#HsLIKNnz^d
zR@Fq|HD&GWBk0s(9(i{@uGtOR$5=^86nz};umiRPbEU%3y;gh^aubY6n})|i7ZHzs
z>Uqo)c5Ve)*f&ljU9j8MGxS{z1mb;|nhf<FY_K~^6n^qM0#R*1&O6VCCM~n4UQ+sP
z#n)sa#$N&fb-t}{>ebm7Br>Bifr9b{hiUguGCfWGN2zYo-Hk_P+d619pre=jN$M<l
zV3n9QXjD@DyAF0Aijl2`7P{JN<b3EFPI;K1XgaLqUj|0%(yXW!M8+$13q)gq0}*8i
zpiOg<NL5Zzo?Svlt;S|t$FRuyoj;HkbIjiPOB)hjrMVlHZ4SxzzGRB>Ywa5AFDC%C
z16+VQncJYrP3X~{s%L!X8rnBH6?gspeLq7Oxxkl=Rbk4`K_w+TmcqJWsi;%}3L|ka
zMCWV|gTn^rD|nWL@-?~5OgCayu%|F_vqHh$Hs7~EbhdP27@<kjdGVsTMhsB_#kDXB
z3&hxbURxy5u8`Oln!N*^ltvuZE!Iz`YX2R4Mq>MD^rluKTbXz!7g}*s!i~6q#$6sy
zNZ#GDIBit0Y;#?@v5Xs^4|M-}Dga5=LZ`w$@p%TFO=)w$tO}1N*q`YDy`wT@9RX7}
zkx@_p(*-ismA^^}sG7mQa|gt{oFDl3VM~518<WJop}9U5&`ODsKVfong0wQ89P`MZ
zH|m6*y{yiOJ3;>nNSUo<$JM)E3vp$GMU7OMddu=bZq93K%bVLcBWiyMp4N6HGo-z*
z(#qk|eta;zFUJ6B7?U_(kNL5db+lx&TEoAy+~BsiH|lM##e5qFz`plS0g?f>+t8`5
zr>kyfxruYOo{#65^F$_}rFW%c^E~f$^OpPXDLs{1HL?!bkAxp6Or}CJ1#7f6ru7T*
z*q@4OrA<qi@gQmq%#gNL-iBj4<jal@TVt@CAA(1G-j@q@*5isY6cBK!DNxl*`ki))
zIwL$GZ2~Rcp9=CpWT}2U@_(*Cngc;h#@l;WR{|qeSOYE;fV8>6HxyCVt;Z%MVhzEp
zA7ye*PwNBb*4$^N(k`Vw0o|fEOD;P1CWr+&j)JKH(Dx7D8BQ|(Q8jTN8Jj1XQ#z<;
zYztCiE!aNxOeP97R4$IcGc>kt+JX%i0rPBy9t^$lfei}RGwD}VVXX;i;I@#v@ctrF
zQAgOBI@sTli1BE%bYFgrj~Su1_z+C9L!@^=zxpskApl^={Edq6ONN%E6KU&=4)jt^
z`iPC^((V<i);7bhV;lEyQ;PQDe<2%tn!LD8gv=#`a~HdvY>##6oC6H$jT3stW}Iz{
ztc^wgGS2R8dplzD^<|(1DlEO^I!dp0nP@juu~xHFM!aS;o;6=<U+R`$pP9pKWv2Mc
zX8u0i&q8zd$7NWvWgz?yF$t}6`(Sv{W-Wg%0wSNggGzY*OgNufMQC_Oh^!0f52Tat
z>cigxU6l=hPxPN9_<zhjlUrqnGBDe^N7jK=9<U>47%@-iy*{16QJ-rX_WeXUxS(jH
z=u~B_ZgOiO=0LI#={$-G)@fL8PniV<WyAY4M%^D2@_`N&fQ(v2r)zSXqh0Ra#K@5W
zkBhCzvrPrPY26>FnB2U1?|(qlIO^Zw>&41{pFY%L1G}QFN`Hfb^4@JSxsB2Io~JNO
zOg$e3G)(s?0rMjE&5tP>hNHQx`a-jM))ZQgGd*HjBV&px6lYb)$G!E?95CzKaS^<b
zWgIWpg%u$Beud8CchnoVNX&hH_Z+LYFdI?g?3NAdMAC4yXpJ_8M))U787{VgCA>JP
zFQuW;M#;~?^s+=c*$~O8gR`$Wp#FtpoGjtyZsK$kLUyb_>L93^7>ZX^pX7a)F)Y>h
zeEzhoTkQ~uxHH3@kMoMo(#@$ED{mpn(}$%~rm?WlZIzq<@aOwcVinHy(+t}}fEq+w
z>$TCGX!Urm(wHH8BbX5SEY_>SKceVGEER3mQYwSt=4h*vaZiaTi(Hg^vWi=E#-ry#
zU*GgdnI5iHZNkEs)<*a<M^s`MYaEc`Ebp<)zwQ4rvhWK+D}K~DI?d!*URo$+5b5P^
zx?jHFWcMY*8w2x@cxqmj+0#V*Q{K)VYvE~UUz&tvm>X7u)(s}>jbI*pe|a}aTu?)V
zvpBZ5D<XEr>%dH$E_zd`|H$mUuk;v=VNYaSKdDNx;)tD#VC=dIwxa(;Xhl^`j1M`F
zs6QsSRn8Z&f65_bIdB<)UG<aR&obYRH`;j|C+))U`Rrkr!MncV&|6IP`2I^KNxiBZ
ziG9Mzg~&$4gz)Nxx{Xugdhg{k1N-u1&lmXJxFqO7Y0lGITobkqm9@~IK7j_CjRuA_
z@52W}P@f}=(Q)l+T|jkg8CL5oDS>)2DI<ovkkxM?qeB0I7W__8Z0VvvuH`oupJOV`
zp|xcF9ecO^wNQsxLC=^+6nfTsv1<0-B#mP6bZw(Ch)-~dsCf;iX^F_~Y^md#-7m${
zkxSYg>7T}Ii>e*!q+1^G@f8*svX&vEA6ZCPTM!fd^)}C}ifmOJ3?equ4NtY2Jk#D}
zq!4o|I@31<>k643#7Q=&YcdCw+lkCZ?ytqxuIZQB?^W+@8i`lymT*7nYxvDDI6h|E
zt{x%Hh1W$;nV%Jk(}Q1&OFQ{;mHo&`jpK~<P|l8Eqg=!=ox8kVcl)7IjVoUa%#6yT
zrSmY2!CoBQ|E<7p))8N_1)D;;zLFc@%#8PLk{92!Zy${9SJznaVMG1plrj&0E_@_a
zoM&9etS-=O)9;18{3~dEC9tlkt>#V*(Q83J!<Mt<cMVBYcoQ66mGx}A-NZT+)~u9J
zF@^$k0T_&LE_Y9dlXyoRcTaol%Nm!_+Rm&5-DI07G0lu<|IMVn($Xg7JSiG;k1wrp
zcbKXAemXlg7T^qvM;^Ntd@CQ~^GHuLL*Gi3y47&wm(9@+r^&zgX%OeV$}Y0Ie@aIl
z6VhGT^^k7j3<|;4^t+-@xW_2C&|@uc<W@QhO*h;M*X1VXDmWcXkMvA)k#dwC^=me^
z-)eWg68c`3J3x9OqjLRaN5))7)1IpuM%9gA2u>N+0cY6rSI7IYf!}<HbalzcgfOkC
z+t26vCo1Q-3a6k5=1msLQYjOyj#dp3j`yQH-m%FIh!xp*Hu|5g@7tb!41apt<c8u#
z;4VEag4Xb@*@Y`N3*QPq^$ZHaZaP*tdj2>aPJ!z^%MMUh`Ybgra4#iFKPf__$`j*B
zOIhSda<oQP^ttt(?rGDlXvRu|3+%ZH>`ZtmHDU}9eBk0@RHjyfGEHJ)hAW#C8p9e*
zP~|s|PBTIZmX|dYmY!F}VD3y2!WWthD;<`i(+n^h{-qmj2%(s-2Wurb=imHdW`0$!
zy7s9PUrO({spJ`bq^^J*Nl`Vu;**osN?S7Mrv%q6w9(kS_@x-B-|@-R^RiwfIr#~<
zP07O2JWhR0V4>rd-A>6OEjv=tW7y}_3<ZRpk5BRSfR&tXS0B{yc^W4{{&sI|wtcsq
zzJ3f17gwQbc-NlhEgGxhW*#|q<zdxlW;Ko&XP8o>W&4y*R`>`-0rznSrAR+Sncw{S
zW=v^>OM^HS;<4+8-^O_~Zm-5iLM7{HYaeNFmUXM<Up7K%u_o_*QvAeC%ex%~y!CnU
zsaHRH7=kCJ9;&RT>e8EyJCNA<N=@?mbSqx1^oNhPm}Y5tXqG<@%1%omRu=`o?2N=M
zm1S@fs%`wmYh2BUo_unCg!;!v_o?-!lpI>FqpsIzCY5cn=oDmziJv^ZZzhyjxAtlW
zfxS66%VmCJN2{k_tuu=EIoGpC;<t3!d-`#|QZpjTGy8}i?|+%AF=;5Rp<T7*`t2O7
zRlJ_6V!=8bI(%#Fb>mb@N9}9ZKT5-HnDNB7QtNKwT@{|$b#4xoHq2h-#z>FzZl@Ob
zOXP4tZNGg1<OcP<6IYfji9FZ07@7!ccF{7WwDd4YStqBOVi`BQHD`#&NMpmL#*3}g
zaNDw0fS{pbA*(LP(p#<PAoiE*1MZcS)BW}|%3Oj6W6N&!rg2rD;cKnBcI4&n<0O>Q
zFX<~u#eE&0FmaC*KT8QyZQ3@!{QNwkpbD)rX8zM@*oQCbTs!tqdczIl9=dxsS=T{b
zU9vWjDUl}0$hy!!2CNhy^<hCe?$*&eI&p}~Oni08$VDKSxIjWB`&D`?W4ACgPN8+5
zmJ9v~Ex)rmo=xN2v>ut8<u#G`ZM|kP3+-5O8Nnq;bJ-Dlo=rK&JZnwR9)lRSSK2a6
z677ATMw!xZ$pt_6YI|rwY^+$%g8m|6(xzNqK6qYc6@G8Xl{`6&-dwo~F`5S{&9RD!
zsBvz*Z8lP&&Ym8$st>hYqaxZW4CP6lJB!_TJ;a*4D)ttyt}DrfZV)^&ueWaPfmn}u
znr%+ho>&Mj-{Bd(<@U11wN2FLMyR)^c4_WMsFjMq=%KaWJH2Px5$VN5(kq?{tatKn
zsX;pM7Q-9?aeJ-OtEKgDzCAEmNW-2{Y^p&Z<FpW2a%5S5HhFCbHN0XNP4$w>B2_QC
zW6e_~iz{l7M=@o{B0-3=t6zp$%XVCk)kaaM0%m4v6vlg9#R)H$bw==ON%g6yNJN$X
ztlk`}`R+c@lxF2%@8RP-P(J)f|Lv1h$^C|*^ceaS&e%Q|wWN8i+9dWlXF=ZO4jLCv
zTyBSULt$m*{_<fo>3vj{_r)IAjJm}@^kZ_5S!E}}#$WjBgEy0!<-NZIq+{o8Mp}f<
zT+9)AR4sP`(vrkh{KxIAkTTw%3b8|7(VEDGMO=`XV5jh>Jd4!8*>XxMNw~;aP2p@c
zf`)&!t#~rU?$D~ty*rNvRl4kDWf!3Zn}2$4>I>d6?mmoy8xC=pShv*dTbg5Mgk;q{
zaGVF2W27Z%SJqI6VcRO|?XREGBYbZc`Kd~4?jL(!sO;9wnk2nWbI>v_nxb`p>OOFO
zQ34?!&n|AX*L~Lf{yi6LHWeRO?|STQK7vW+#B60+2a8$WNu2%Ae&6kHQ2ftVvb4W?
z`F`fP4W=6CP(Qss+!COWHXmEObT)w;2J`ZcWILstNF=*rieraWUf8m`oVkoO4RV~g
z%r(hRx7|TvH9bb`O;gl41_%66SPQ_DQgth^VPmO9@Q`^#(|diYmADYles_-&mVH-(
zCMJD0wEj)Qb){}AYIaiw*GB)<7j$;@eEUZXHZ`1ZPoC`raFs5-m)rJ46X!Oh=Pe}o
zT3ayndOa`PtFx(bq@|gw0{KiX(c_h{jK*@Q#ewPlJ}0A9EY<a@Yq<U^v6`huyuv;3
z{@6mx?_OdNy6uX(=?h-Hl{u3%dZhiWr$bpSS_$;Ef+y4kX<t-I1PLW|gGoyz2o-fs
zcm_g4Lr)oz!B7+>YaBds!dq<Sglm&MUf-Vf^qTJ%0t+&u_GM6FzJ7J2r@3x2`^Zg^
zSpr`nZoPbv1b&(0Ci)0#X2NU^g_hVGUlCNn7&{wnDS<VcfB?cxzLpW4xPj_Xtk;9N
z5-jy@JO2owZG-P~0`~e+jSz+5z>D++W9<yVU*)=7eIFk=NoznCFpnJUe~1YyJ1Fu+
z$7JaCAgW|H{`ie>&Z}JiGZ4+e;2M1UE5UujbyU3#jj|Y&f?^7!7v4|Oo4VyeYyMzr
z!H;;pt4Ks5qf+s1$o1`+NyNOB;{;4_t>xUU<$_-4hPBNN;n4!Yd*@=mCb`;>)WyTY
zETS~)6S?mY?8AL!wWgEANp>(&ZiG;%HUOicuTwr=h}zj5brWAfWu6;$A!<H(!>N{s
zeFg9{gbDavt7T_ym^03e4{tG}4(pI&m@xJh*ZFvJYWM`zjR4Z|=8QiW`;;9Bd8jra
zGKv#mE^HfttY;A{xg#4EO^2)-4no5YP)LsX3BiS7J9Dfohi$HdL>^(RPJ>J}ZV~LU
zR_81n#^!4yMFph~9J(Ez8>xgv@_jI}%6d^g%=O!lK{i#y&zT<Ya*(vmE^6bi)0NZl
z+8IXwW*C)%3@{)7F{2BMkkht^$G=}UY_=pgV4P9leJg`3l@;u!-Jpe*Z}nq(BovyS
z#1Nl&Tf;4jPvb9N=((80YQayASsEtS2i;N}x}_cO`BLTm@)$S_w`X-UwA5SHm6eRE
zlAjc`Y&4FU4;xKE{qu-i`lZ9O(sB|(;2_HTV~qmYy-ev)KduAA?FTfCBmG3>1??%{
z(c@sSw&_{3V2ngMRz1|fIrD;V<`Xs+!>!L(A;dxHfs1-X%^X|53a7zol^T$qPVuY5
zbqtnFr#xYB^<MBVmBKO0F-^r78Ww+>3MJ)kgZVN8``UWmr+T@m@bAVBhc)hZtqW$L
zn0cw#MxJY1b3Xp=blW34Q%42CCdTRl-H&D5NckguDgkOq2IHYcx*4wQ?4pCgS{LMP
z8k)dzoar8%>j#s2Jg`^vEj+hLXZA}+y$y1+C(EK};AcydlypjVhvxQh7}KWVcGppH
zacW0`-sbGiY3CA#UjI04LL%=g!n(?0&R@>v*?Vnd+p_R*NG2Cs7@DmT9rLB%WX#YP
zl}4N$DDOoULJc>oh0J%i<@nm7d`#>hMnB`c<^<!MEMUmAER~_ksT5|pZVz6#+1rkZ
zed)ND;^1m%^Bz=CY;;IlNHUZy##GN*dHDws*D1TzWHkYeC9ylz1u^VuPwbg(%H3ev
zGfcf*-(RBdPL{q-``sbVDayb{vLI^2IzhTDXI|y}CYIjnDnf9Jrg+jmwN^z%W!65w
z-*G@*$c)=t8FP4Y9*N2+N1d5{IzDx38x?HCR0Wz1XEk;<Jce1LEkpY>qCQNdzBbI{
zvdH0L`CW?VOCc`2$f4N};C`>-`zy^{ukt$k^pQc``B}4s?-6gYo0$#HX#nLSTk~b^
zyLGoGla^v^4^?Mg$}W?2%0rCalrP~?-GsWU-f(Sx3}lgp(*JlO_mU5*wcY};YLF-m
zJUaC*x#Yg>HEnaoPv`a4wv!6F-t))lT(xZR=&s#|at9-U*EPbt1g0mfp&s>h>(V5G
z_mcc7@~NT&U!GYerlHx7N*oyznBluK^XDxt2ne;0C-80CbFyAUXUKD*f;;l(%2ke1
z9H$h0TKwwO1%n7&Lm9L_iH-;w#~(659sLuuXS9=-(>IMOQW^6;ZESC9n$`EUBYp$0
z#0<8aUr!K8!hXU6N86iHFz1OqlD&%uuB_cq-|Y*6^`br!n&sEqc;hk9l;Dv)bRzlc
zKJIEhT?+{Krq{AN`|S*vU?zqDv%vqin>aq5T%3d0luoYj-%pDRneMFr_e1};Ur}k+
z{@=djzrSCceaY{?ANl|A6)V_bhSQLgFkD~@CnsIuOPjz1%1Y0)Sp$H;0f-03(lzA)
z+4Y#DrK~%`E^O8BNu4VCI~j4}PA!MW@SbH>omq!HgMGzuRlw!<|J?rP^}Kv#zt7B3
zL?(12Hi1h~G-Brc$VJ1a+qJXpy*IG~&+q=zEc`zWv4BJv*Kc3j6*?5FO=^3DF&d&m
z_X)k@TUe4RM=$>ihWq5_G!KKj_BkD&XzDT^Lt9WT&l{P9Vnc+26xRCRSKGp5iC6_A
zskf76t=8pwx-XCz1Joeu-YDK+y!3|yTq%n}|BE%EbY%Zo+Whr-J5OHCNDUwaRX(|@
zKfVa8@||$J@2yi6n-%(ni+iz@fdLEb*YShg4>^5I`&E=xIMx@T>&S?3tBVJ3Z+lW$
zustsNfe<Kcu(lHLMc55`Zd4A(@G?^p;;SL?hR{7sFUxv*dzKkRU8JpEckQaPUIoc0
z#uzKu<)Y!{_;qEw@{}@G$zbgILIY=#3A52zrKEL%G^Pw2;doJobjNJmSaGBo&P`nN
zv`-f`lqzk1&zYVb2hLVie*G%(OFDYE@5}h89hTszt4odcm)7`Ew2Ia^_R_Yb&d^Nt
zV7le46vbE-MqR*N;asaKHLyN@Vc?#<jI=?bJRwn4oP*htZ3e}O@M-IT=46tvs||oi
zIvn(Hi_Sc7FoihR?3yBSXMDmDi^MC97zym|$$6VjLmzSxBXctw@3k;fkH=a&53uM+
zjy{iuSqMxWy_e;xs*L{p;lt2$Qke%XZBd&W{%p#AK$bK(o64DdpI0*FXMZ9-Gr&tk
zt<O@+<2M{x7vu1kO<ri(zU8W?Q3l^hw)%5kYS}6I7d1~tzi^qxI;q8HSP+oMHTP4v
z;itXI2G~BT8!M)4Nu^Jf{qN3EOtvzW49~qtDbR-(DTK+%f0>uFv0+C|QL|Vxjn(L}
zUW=k<H>(RB84JYOgc}H*U(8Yw?3NVoRhpLS@8)?!y!sjbw9tw%(}|N3T_*K@8`hJy
zO_aK;pof&j;XO$WH~mW2*t^0<cJltLW@)_g(6vbHNWVg*XK$)El#(iqh^q4F70-*|
z;_OHcITl-Gf-A#5GvqI=b-2N;;y!!7j6Bfv=TJ!}m-u4MFJ`V9$)7CAL9Wx+{TuON
zIgPvztaOwfLo3kx7c}w<Fl5y)d}gx4R(0^G^){C~<Z!aEAe#GE_T{MBGkT?3KA+|(
z-mt{SzsccB_QiGf;5b@N`lr;=+I8V+r1(ZVamoCl+8V}j{ENTh25FMTPo3#<k?fXw
zoa?m{O==a&h7cE0In1n3JZ^MYKS4$ORs-6$Pi$yPj*V0L-jYB{X=fbGZw`s8Qdi$|
z^An)f9*qrU6XL8#{JKJECzbu>8Sg3w&neqU#%xtW2%Go9*O2-MhFpfx6=m9nSfZ%7
zD#`HLB<4NhDDXSB#Go0{kWgDJv4&@h(zeOR5z>P@H^|~Y^oM&STA=<ou!i=hRK34R
zPf^K<WFLkoqy{u`{%QXLi*5;Jib{9uy6i6NbRa3EM={(}!70rTGTiAn)%})4>(h_b
z4f|N;UNwt<yc>UHOyD+p)}rFk5okf!{c6d!wk*vqwdrX6JHbiq>q$fFH?=Gd(PY@|
zH$84^Bdttx21{i}*FIf1z>VPjd~zBv+C|J-H6JY`)wsyq!?XY#5}k#AP-IxpByl^O
zey-d0P|Pu=?N^NAxWW%+aD=el&(|+i#KAF96TJL_;@4U2eba~BsovK<{_sXiO<U9J
zsgVWiV|O=qsxW32LgIb87`2qgxdWM6-|Mm0w7{CP<*H&mP}7i_WIEclW8H^!vNr@e
zY+E^byLA24<O*9kZ$%3Gv)7y!N6b`U8{&Ub$`ilKBL4IRL5n<>!e^7!DyIw@;=iNx
z*CX+EwGk_W+`r7+dAH2PHN$VK<2j@wTDlVS67DZ$CxMFwFS|0I6<TSO@(l?Ie1Rc#
zJ>1naVk;zEbIP*`4xGFhQ9gT5%ws5k{{++Hy}KpzPu~jQOA3`DXkuSGA3dlImnk1y
zc+yJUj;G<`8{I6c-Ou<)W5Udu!lT(9N5PU=cC3>gC4Q@=kZ+oWy_y(^Co!b`faKj{
zWr+#LQZs5cZe1y65zl_-w}941uQGBZc8idPiz4J2=>^lcx0<wy?z2PW>h9viuNpv}
zEyx`cTpZ=Zd*^&o$DR2+`CcSbtg=d7*Lcarb~Wj_DQmHhOPWa(5Yii)aDPuQewlxS
zGXI%w?5_G9+ZL7|?Hfh?dRE{*c}g=;lPEd)@lgB90mHpl6;HlA#6rwhUeQwKudN;S
z#qJxU!F85aOTvq`a<8$~BnC8d>7&v8$C&ZTTgIenMs9?LGIJ`!792k(!?k?hA;#@>
zjYZ}K4R$6<xxQ?xG-ZN@yY=IGo1H9I*#752*+(0UVfM5hd=+{E6kbwazG$1MN3HD4
zb;(0}9n7Y=){^?esHLkCZ8(=^&+tH7``$K$+L-j3U+s1B%Y(J-pI~q{A?rsKoE<ga
zo}}*8uLui8B2d?cj}EIIRc(b`x4M#bElxT5V18q`VtS<2+U<(Pj}@rq&%fwgX^iNo
zpXf<D=1gQ!z0)n)=ib+I72#>cX*ex*LVxJ)_ZL))-&s3UfN~4nz&nLM#8D6EBc6~d
zjiur=-f=o6Wcf8jE?1ddt71)fXK`2vDMLoa%uW-9D-_TxpWaa`MW&+rk-=$obmOB5
zWiQhUIuafDvt!MA`Jt7K4VM+2<AQ<lRa>zV9jx#A1An^BU%I*Y8!aGIJW-{Z5<#p+
zi5j_!CAKEDD;g2nxdgVizhdXuI47<>vf$M!<qgk48Z?{U!El+~uws^f@LQHwfUYpw
z&{Vlb>_o$$+JJw5F(Wk9q$g$A$DQtjmG<|I?4O?wb&3ROWq-7ok>5z|JoE^Qe$r=y
zDz-ma*DdyIVRx^|G|ledr4ZX&EDW;b=b)hc)1I!bb{a!(LqEiKB+tqe^s}Ybve9V}
zXTvvsCddDmw#V5yE{($iYr(IRpzEB>oZ<U=b5bAiJD<4f5!J2fnm3e9+KC#9pGTM_
z6-jqDv8nB#?aYJ2_#3@#gbFUT<F@xhOIg*Db~rjRN{m=|C^*%lm@>+|jf<KL!kEm)
zIauS~jb0d61>3%O&BCa+&p~&~sFd$mrA?fzfx-g%Xl-~^AddcPiOrjrv550qR8|Vw
zHXH?f(v;)z@0ET^3<;FajnG*2>z2XloLu=C7iYW;-KbFRQ%`MGkZ<GEv$CjGQK*U1
znDBx}tHN!jpLInY-ZN2k6?BNnsucOQ4xGuY%23TwP^ZCprFq*WE%ai-^ZUR<h??6m
zSLlDJ^R4%q2I^Mve$7{asp+|dB&?Uxywy=QZolBA^25R*Omj|;BE=x&4JGF4F>_V|
zifnnBxkyc=&!XYnLvFDHfu%}NJsbYQ4wHJ_$v6+x7_A0f5+4a|o4Ur<*F0cWyqgbI
z>Mb|EH#h7o`l8fGYg%I@MP3)JnY0wD$uXsuW$yj1(YmP1m8TOyUm^9zp_ftOPEBP3
zqKf^clG@OL$-auCxX@B@iA3fRw_{o6hEYk4*5HQWjTfV9iBG9z`}156Amz~s9ze@l
zIE<#k^Gw_%x_BClJ&2C$^A$3Amya(US{aUps#I|QDf?EUpFdF9mq5+gGr96*G(n_F
zh)XVgo4p&+Wj{8=@glxM;{KPq!dm>zmvqFgcqg?)D+w0s@#92x&cB@fGp@7bR=#JD
zW^dS3vqMl-`ez@g6gHSI8e06qBtH5PM$Kug>i($x-F!>=ySbKFl}3&%D&}7eYthRb
z%p--!wIaQ)+#eGGa`kz=FD>P|XD(8v=9*>6dNtI!W;VWs2S+(&No_4Pe?KAkDEoZQ
zaFW^nyqo-Ck(|Z7=1F<;ETrI5)nw7Pir@q-rTVAwospyu$hJ9Kb^4ez^SYAZfyqYG
zu_L|Wh(|xZgt;NDt=l#_*vFi+f*S7s4f6i}sHRF)$OjYIP~KZ9I3F~0Tf1(PrOyYw
zRZ!EOX~EZ^?Ny|b-~I7}Ua7fp7}Z*q)+_#JRHhZ0u04KuzR!vURSdG?wq`Q!b^dhA
z!C?R456#rs6YuC3!%dLg^1Mb!!Z?GGqZibc%GfMvf0ZGQ;xNPO_a;9D$7Weyj9ZjN
zZ4T;pLB(QWZf3`^>#efeD)HT86Sget1_J`eKEi!MTd;vj`9PI&Ok5D%G{MR9yYFj4
z&|9xQfjn>!-98I>u`8q3_XLY9=T{D4cTDc_#hpbRXx#T|p>1%-lb)=jOtTG618yuZ
zQ`?kBHML&2SX3)<ZAP|Pw02f){MimX^jP4qX*o*QsvnaKQ~Bx0zSrL9ZY)TMO7$<H
z=TmOiOIAg8=BZ?`3DT8(aaK)=c%xs~+TO=hUHr6ltZuW~dn{=BO;@-xXYu*n3E#fE
zfkLX)WizxK>SdJy0wRBu)a4%!$7oCc8^uTJLN?P0L#7-25TV-9Ecf>wV7B;24P~Up
zK6l_0ZK-b(gein9zlOJ$>Dm*{=2oTm7p782ce#f64920%ajy-=WyEoxvMegqBrHo_
zWfb8#ju*<xvYfKevoopnyYm6a216QpH|Gvq$&UPNWQJrrB=YhuxNlvqJIv+o%3rPt
zMimS>!N0dCYd3irIn63DK4XL)sF9Z4JrB*&#tZI5^Mq?F;T@*UD#brF)h^V<otgnv
z!7{Gt+dzc5_9t16KKSe8nu*7CsR$Tx`f~;=9zyq_lZyY#m=wK?Mh1f{ZNsK19p`Z9
z{lStEUNKF37x7~IRGu|~49(B??SF_1Va+ZikvrY~F+t_9y>Sm2PBseX*DkmD>Rj{e
z#m8bZ5_%6G<KI}MNA||*SNz^+@`6L&Hb$?6>4$))pNZ?yzAu&2!x}Wa>TM7#pT$3{
zR#(8PY#3^1uv^zS?@t8gYJe*}=HF6VZ2t}7lc%}<+o}1jp|mw4)z|8w4oaKOqCP}V
z1^BbGVl%>GslyK;-f5%b8h$VB+)6*@%Tlp3LQ_-}6;MHPI&{XIR8;KVARMCRM;z<0
z8*d`2V)#bCn}5%4osH$v>g`*#SI?n^yd5{Nt46!m&y;9JNW;~!#?pc{%C+<&;u0ux
zS5H24?33SxvMQT9W~Jt-bhcr5!EcA<mp;NxVs1Zc&`*wwB0PHEBj;^sX0>1L6<9%t
zQO;UHB^E_e*W6%6m;&8D<4E`b)D08baed~^$n?Si0wd_ja5MpnJKWX@*_PS)lis>p
zC-wF_XVgz@jO36ucKcQ8<H}WjMH?G40AH$<vCYo0Hu!ezZ%~foVo-pT>7`B5WejwJ
zXbo!#QbGE9m9L-n1`&Ffd0Z-j@0LAl?o~(YKB;I{XxcL4r?dm>Z%T(URAg0GIJ0;<
z%IDEWPTmmw*=AT^vY~2vYMV3{N$7SW_#InT&U4R7S&y<*^$h(CVCx6J+TaV5FR>z6
zO-p*Z&&IzIawJicXtScvZLI4TRjRx?%U=d&@UNQ0n7xJH@%H6Z7t3}Vldx=Cb}OBJ
zz-E`-OX{{Q^!0<$>1O1y)tm&Hhtoh$VKTd818%lB8<QpX>BEII(x#b)_;UL^mD$oF
zwu3dzgrH>VzH#E;a{nctb&^?|!?Cl{%pVC)qJ_#G351O>GwK-g$xLqdr)zI@mGFZH
zHrhwQdJAe%Mi?i=oox3A?ac2O_RNLRm1Ggmmn<8aTHHNi-t!gs4^1zC9G#FSL%&_U
z(w`p4)0<KKN!O%+sb)IJ8<m(0Z!jNk6Fj+@$UK>-Asg;8wvHVUG|X|*%cafskvT@~
z(qi9h5)p)ElLr?jdN8;47!I(iLkA&pSoOz<bIlY@f&)rl-+~#5*kp+RAscHl>S7S%
zXU6CyIH#4Vbz&&{s!2E|>AwBxX@2wE+P&j@8gcTe_djJSoe3Hg5f%h$cUF@KZ7rRm
z8<5hQm<F|dzTQ&$BF(;U>JHbK6SdSCG&pSvHGc+7W_R=&3ka`G9x8=QwWLTIK^sb1
zvT9wZ6Q2qjeaY>hSC8^Rv7z;l!zOH;_}K;s)0g5>>&5XX+AYU-B5%7L<pz|^&a=K>
z7OG&k_r&)ful{ailX|B7veqT=7lOuNcq!Xe`_a|;Uy-ZNSUD2~O*6QpHxtAT$J3*9
ze@aUum8>KLHLw22igTQ(O%G`HsZ%zClcJI=R&t24{A29;yW<RUB@!Id<2J%k+{2Lv
ze$}k^GD2~DCZdkcbxO+Scn-u^PsT^%O~N=Kr}h)vTetem*)*kVR8TX>+A1Mppwvve
zzVW9+f#ypdrPARg)^H>y+AJ~KS;?-~fcsZmW$X9fPPZY7hpu5JgPasvza+B}T0#=p
zVH99*S_L&7qZhE0lnf^|u*XePl&~<vJ<U3fyzBSWBGa}jIQj>d_jlK9VDrNZUk#c=
zC1%53*XX7A>=cSu8ux2ypsc=sroSk%q@|m<cxPHw_`?b!;%?`ZHRcjE2yIsSj(YLA
zv^|Hq;$_p<7*-dlSLUf?OADR)`hTYAeTwTQq4(KYB=U&9)J#kn)O90(BjTF2#|0&W
zm0wZX+k}rYnjQ5Hv7>Pj^sL01sim?(_Ln2dW>U=2QMOOJGICFfSBzh#Uhg)3(#NK(
zcTYt%1*PjqP*3YH;x~Lw;MLFiOoLTYR;n(XUW<;l&L|VaXIOD?&t=56(`4!Q3odKd
zZ#+utBW8<B?v;kp)RUt1CM{4VnJG3w9~~O|%}7#)hHgrY4L!)R$h2-_D}V!@C=eOx
zKfzt3FT>8EYiQ^crvyB2Xv{f|L>j4!p8TqCnmUEK`yHGtyfsJFaL(5p9_OmCnARr`
z*1OIVXS&X($NK>SvU0VnW?|un5fRi3)?sVvpw8wK8-w=Wpy+zr+Rrv6Tc4h0WyE-e
zVO@>B_KVhrlkEKt0S8WP^rr;3X`y&uWYOH>T#ugFhu$ds)G$;OyuT_BncNeJ22dYU
zqROWMqB05LmlX3-1?`UW4-N~K6q%~b<m&MX$rg$Fb(^N|9W`&FQYrkC%F=ML1YUX`
z=Mpo$X@{$gp7P4%ze}s5?ls=N!E8_etj0qVPCxJOZia4eFb-?kvxKjEPJ9>cEcg9B
zIk>E9J;G}YCsy3P=djtZ(M8MICUGspZ}fE-ZAVkl_^PhtD?!rtjBZpDmC$ipykZ1b
zxZZh`hDSx~z<ng5E#6zJIqY=GrFtoCNoBBEw>M}Cg2q6SeVAeyRj3!1g)UdnPheC<
zd>Xo*&!JVfaEDEnVH8R?gt%7?DN<ypaTvaS`nIi-ZgJDX2kepid~x%cJLW?qY{~JI
zTA+@D<!6@~MV<*^QNfsFsfgo*XDm<LsmxpKB#V6J!u-&tsnVbJRUIf1gOBLByWHen
z@@np+S8deAReovKT^b1$z&KylkkSUN(ceFN3ay*<%&FTHS_M53-LBtsp3CW<<!p*f
zQ-Lh!w<n#fYoc*3venhLYX;`9_M!8Nny4ION@Y}!75my0V^ENN^#rUu-tA+B4HCM2
z;(m=+@b>w5uyFSRt0r`UTFw9Xpt(%(%YXXeJo?M%hu6%OY#_?$SLA-+Er7M?ku%+v
zPj<b<FBV{ffLH)hP|juR{Y=MB+`{278O!ppGC*g4Hc&G$0&>YSVZ;`b$?<rXgH~mL
zjCvrMIOLz4d(f@G5+Y%2FN7nwzNxu&S^<<6JPfn60fHVJ<^NoD;g4WL=j8<tg5Zll
zD4yalnaSe;J)+{8Jo+K|@)Ho*3zRls@zM4mkrr1#s_K&<W*m?M_jdQ-FY;%nrvtgo
z&69z0e<;)j^p<Q_LB?&Jusg4T-<>Vjco}llB@eKOW3g|@+D4V(Kq^PhJb-bx0vWaj
zej^jWwf#dF2EyT*{syH0?`Bkith~2LHssdKhLHU9^f>5Yn)S>gaTx>#tPmET@ynoF
zex6JxI{&`q2rO`XT=Hy_Cjj5!C2RdDkaZgv3QD%5R_<(r$QFY>nH;3GR)EM&61@%`
zcMnfPsP;Aw&i~vJw;uzNXE{!Md=>6$%2xxBjAYBVegJ6__%YXT09gqzH=Z2cC2I|S
zW}AoZ&L>w7ECEv0SuR4A0Rc4dr?8pJy}5p}v$1yS6LP)v$1Fe+X!UcOUjs>lhYkUM
zXu#P`(7(CLpwA|GR{$qj$n{E7$8Rx1kbhpb1(+K@>*iS+<yt-lsaUTP>&Z;eUmIKA
zP~Z}MLA0P^LTQWj6k4A3_Q_rH+~lT-iF%MYEU2}@|8)PKuuDrzl#UJnzW|ro5(o5z
zF6ThTt%13g!v7rekef4dIuThw=5sb^W}O{?Tl0!D<+g-SV1X!LauZna-K&{rltZAn
zz4eEn5gXJ0Mc!LRwbi_Tzkvd6q0mB+B8B3`ixw%;;$9?ByilaLyH;@5;10pvo#F&5
z2?PkP#ogsTo8G_w?>y_AXRY)8IrCz#Wbd7kYpzKqd*=Fn9QI#f#9Y{)0pj&-US3Uy
zfSAL$3#7gT4Wm^s>;`%Q0<g&-y!$!;p?wDC_IrOVgYJ4^ndRbmx;YJqXI!3M3D*KH
z7&++G7P0ui>$=sKAKkSj`3%zSb#wFA!a#Im`;@MCYm*=D9oYd=kLUQFK>gRk1Ea_S
zbl1@mfSOfa(Xz37^w)Z5LN}hRzEoe)+F^l^Re%t72FV00T6HO62NduweW4#{!F^w1
zU7nHTY=YirptHZ@1tMJuhMEB}hr@$qFQHL@E`#Ri3wg>h28a?D;sYymgh_wA=e&ks
zs~mtP_P>T*pxxzQ907&jx*xOarTd8ZVKU^1rWBy*UnZpTpKj^<J@C2;UAqKa%VQgw
zVqhBtTG(UrxDx|z6Z!nts(|4<ZvpHtAP)rj2)Mf;FiNxhFj@r;&$a=xsKKLwXiKjv
z7y#~hCJ)zt*jIB2MB4&q?z06<I|(AV_6BHNfG0n-Y0m`FAFS#Mogn+g9xVV74S?3p
zy?1%fx2K{BZRA6tq9Uy=&Vcm*iuc*rzk{9r7^wtAnF9C!0ty`71|VVk8D<8>8>g-y
ztg0;z0YKF5??u3MISgRXXAgXv%O{TgFxuY92fDriEc!oP3=ANkAi&bU0H3AR$I`78
zc5{F$iJ`rQ{mw*d@ox1vy?W<|?F`agSk&des=0iM0GObr+8JNbGw-9iw*h`Gn$)E^
zz>#lH1bXeV1`hcQvRun=IlTfsljquJ^SlMX(Z?%Y11ErICY`(UUUm?~lRB=fiQ2A#
zWI1XYfGSsSamb+KdO+CUJ=H$|c;k?Wwz&@MVGr|F#Kn~#p{*rC_moY@)3f85so4;k
zHW=aMM*@gB)t=gMaD1Y$or`hZXJd)JeDZ|hPNp4)QhQ!>z0hFu^=VR6=^fqvg{cwa
zLAFWVrK94VbZvy=PvFZJP_8Ysgthn6mYdY>!&;u1^nIgBZTQZvRpiDl285Hd?fzYf
zRgbLW5bgf_e=-2F{2m&1p7|KQ0#~Z!IA}db>yHeilrFU6NIb<YN?NOBZ7iQ)1%yLE
z02LOlBzU|{Fg?F*L+-b%=Y8AadEG@n2Mf?`1eukqxO;RGppeaWlzQ%Vm+v~~tQ&7`
z-lO}0RutKcT4!TK&j;{f#uu*fd{7SxEah+J$nw!cuP-5!k_Qf6LU31zJ#@VlX3GOx
zY%>$6HDAA!oM~U9F9$Y-Upn-U%8uWTeC!0F2l7+vE61Ugwn?$nT)SFK+M287_k3@T
zM-3!~%Xt^f3((Mi@aInqvnj>czAh;o2^NC!&`Zg3(Q@sZrsU-_w+(62MY6F_S{_Xy
z`lOs5snEK-S>X8ivyz5IyAY4EDu*}CbWWvSKI^w4uEwgl4}3;5E-gJi?)+@z7sP7F
z&Cc^{#WH-QgkPR$NV*$gnNg6mMN?#Kwn2c)zBISrWSq-9+UPW^!o(1pqrfK9KNJ=g
zsVC!P5m9Ds5$~j4D<`$9uCp+2xLRnev)iP-Jxbep%2<nN3Y{tVV*?#760TEM*ZLaj
z<Y2t2FK;sY8D@x<Eh1(A8*xJ!`TP*0K2&jclFdb-cB4C3I=}T+ysiyVqG>MvsBU39
z(Ip09QBL+#UA^@hX+QDQL|ys7IfSC_mOrARX)r~F{#6lM>~{m!o!p(5YUTr`MivqV
zMY6rA$}xGAKXd5D#<mE(19S4r7`)9o3@or3oud6-2{F{;sJzNi5&c$LAY7D&dKbBS
z=8$M*MHHSxW68X%4Uaz?K93DJ0k@q@>NUT^%US!7wQ|i;7hkNXTDONzg9vL+ww{Rz
zmoh1|rEJG(f4^_d#!KTV>nU1EJb))9NA9qI=R&D>$)IE9p(!!4Q-cY8ROsv^8?qy9
zuQaaog?Hsf6Fi-PDbA8YVWUGfJ>B=iAzzpHmROi+d$vq~m$`>8wVOG5E}n7z@O*8+
zq+rA?G)hnOg*KrAQ{HWMjM+rFg*!H4B+Xu;+1QMBNRcb5>Z74pI2n(-FD#QY^4sGu
zCz*<g-5|tH^gtPs(^pmCdBxo@ubjFtj$vb<(%=VU%)&Yk+Ld1cQ~0Qli7vTa>Fc@t
zX+JUrM>YNdY5cLQRK2(vn%kC|ajq%n;HUlu$M^X>ufr(hxIFw1NL*I2CaPwVX@WfX
z&9=e`VsnR^G8AhDs%X;FTY$9Ke;)uHWc8<=6BLb3IrADL7VN$#!eM&RE;Al1*A?g9
z^9jrY31;EsV7NF+kU60I6KlV*+~KK8$E0N{5z^~Y(q6(j>u$}rnS*IKB>b{gf%98V
zQ)y6<Lr8QkF>TKKU@~+C`bDW4{%3i4%UzCB)z6gzCcs9;-al+kv+Sl{WW>%Qy7OR#
zGj8@yvRS$|Jn&GXG&RLeTfKx82YqWC8(L?FY%tSM>#K28Y>N_$h1zJH|Dh+D2Gc1v
zd#sZorRkGWc|sLp)0k+gg3bD_nDnT$KwKiVnqbUWJ5x_Sn)XU#w*<t`BFIeJha8#0
zKMOLd<CW{Vc0IPL`BT=O(`b>2amkR7TdC&ifPL5S$H30?g43Uf=2XZiePkj9UH22Z
z{4d_EUMQudG}E6cu?xmU!y@!F@%UDRNxzG^jb$V}bgw=jR<%o*BDiS?^C?Z%qiPab
z>B<aK6{Wi2pZHwDKPo6PiCh_t8x9Uhq$Va=X?_UYi`fZ8DTpUY<!WjE`jSHS9EBgU
zUOF=W`%r|u)UeSCtXw=Gk)<G;d!$Yy{SEKQqetCNlF1(38uB7~ws@u#nzWgc#xnLQ
z6n$hT#txR#-6*E;q`K-Afw$Oi>LB`1Rd%O0b>_8L2DQ%v_&izL@&}aK?W&x&lUbB(
zTAcC)UEe-S81vp4KjGJ=Khx+;W-g+6-{Y3=pp>HmwdDQ1X7nq~*(Uc-P(&4<9Br=o
zVSG`kY-nx0ghRQa(-&Uu+<!pk6gs(r-pfl?+-#Id%4q=^TjnYKP2ucOX_nnu2(6|#
z5@E7JG}b!Z$XaVd^CY1UF6wz_!@W8^jEdc)H<-_-a(=E#m}_<lt54e!x2WniK09bP
zi>vb&&Q6~*ZM9o@5n+R(fBwv_3B@b9<~+C&zqv=Nljm!;F3NA+W9yNf_uCen*)?7D
zr)d5i2}{bOeuG~sBy$`F?1$0YsFU>8ANf&`t2#5^@uZEekPEwoFLgmZqdg+>ogN2F
z=3U6(42^bts>GV5j@Ad-g`M1}7#hdDcDBnyzP0sc{tHHUb0e<T*;Yfd_GbScqk<f&
zo{d7OajJNjr)0`-rD$ma0mpkC>dfquWC>pQpm3^f*kVpQs{(PkjmWNZ!Ag8HOD<6`
zao|WJD<;irByDL@%riBBn=ylkRExE&GkaO5GxcYoPFo^2GRsTqOxXN{)s~yshef;0
zip_V>>{U^1L78Lfpx0pbTV5H9;H329meRMSy*86&+UUDlQEJ-Z4xZIYijTJ>0&eV+
zL!C_q6oo&ghAmSml#~8Zte#&Ob1G%3rSj1BU4sv2{M>g9epMBHdqnuvCE!AZUX`rR
zSW37E$vO6W*nyeZ80xsBQHTfaD$LVJ-?B30dq;Yb3OVPm_(CH>_{mWo|5a-=*{Y*1
zvaAZY*<DcY1ExM9b1AcxU_4hI-n%PjxeWT?3dYXFK{{Y>kFiNJnXVQqKkD~tpK~vj
zNw*+86p#V_c98LS{|8ZIATb5lpZNLAx<nuWdFoWuGUuctMITL0azTV;w01#AlFxSu
zV^i_a$2-B;Eabq(1A%g58CwhC$S=u7NZws_N8JgH#-fl3y*uZsvB)ul*DEgFirZI^
zKC>(wztLqebF`WdS(fh)8qO?bTa4jLA6Gu7W!=CvVkud%ij_WAzexEy>+=C+^0cUB
zO}K$>IR?teEM&DWSQq0}+*kI%`)a#;s@dFU;z^~|0gR7kIc+l}fo@060$wcji}iK>
zEJ7}CH;Clq$|~rr`vvhV>F9)tSg5KF5l-@w4>Oax1&=v<+f|+r9$l)%hF7UitkJoS
z+Z*s=*{Ux|1(H@Pf4FBtA>PU;!h;UXxMXCJ+Yn1tX+4NJHXP}~pY+<ipR+7weH+*o
z4%H2lO|h+)E~;AjZCa64JZ)Wd$vBDPyCLE15>?J9m;3UglX}>cV6Fa*z$T3H^Ioy#
zV@~4hLLFD$x;yGgbBz#7-Z1{cxf+(E370&(!Tbv6L0Lo=cx4>#!&-lsbFd<3vvOQE
zQtQEt7n5SuC(WdHOexV6856WCY$;#o6LwKasrHpor+on@*zr7D_9dwTIddnId8kwA
z{0ZHRk0p~vLxP4Qd0jcGX2VHj6$r0-U+?{@JMwP}L%%1c3oodMo$}M<`>#f-8XsGH
z2$ql!cV8Nk{#ZJKVcjqolbb=9tPrOiHO6JIIPfd-snMN^>MvD^6N(Rs>szX629Th0
zp;5AW_hPT1aOs1vQ||RQr@D=^FDy`Xn<PIaD{HYY=t49olZq4a*EtrLXb_<hO$6@e
z51zCA17d}^4z&nwm(-UJ-@Gw*r6gb38-8bB@R8YBL7<WyyHs`}TQEHz*1$%7H``pK
z4ra2HFdj}5w;-j4Xbkk5OT?9o`|nMb9xt?;CFnFfn>v5tg4#KPO)8Y8A}5d00yRzW
z)+!2}hu(D+)9hqUR#M)T_2<@{czy`~+L^{&sgy!y@8<7ftxX~4T_fW>SUk*z`{oRc
zujh^0Uu!lCZ+Tzj`a?1DZG^?y!iFaoWTK{6<WJ9#P-}TnYgOXWO0CWHO6z9T(s(#`
zP;Q;dQmi7q?G0mGd_$z4Nq*J9TV`f=^t}<U(N6*t2eFQq7N(Qyc}wRLy=UDiBB4R&
zC6iw@uHTZM8`+TSCDnQqWp_d0Wm5!%L)fW>tZ*dFI}Jfc^?ZV&G!Dwf9}M#ZHZ!c)
zl@7&0?sPDd4cjj^B4282iWGNRcuK@tHeb9C2#zN6b|`lQmm;_9j5L{trA>)MqHOq2
z+}H;)R9`DJ;@cXQR+XMvjl%mNJPT1G?Tcv(b!*L)g&`At>PC^`T`kIe*uxXx#5_W6
zNwPF>yk3F23HetYe@c4ZdK00B8RYLQ)1R;J9wQlx4%+s}ukZX7n>ki&)!G_=4oLNR
z*AY9qmO35CzqC{um%}B0X~&IUz;mNfuSKJ#Ano&~)~r!b1s-RbP`Z5=`RN;ecZz)8
ztNt;D>ezSj{7_e&@5x@K&)MBN7x#9_21_%dhIaKD>Md91F2#(NVBXqJUy3~a)tfTM
zbhC~0lB~gb+M49xPXu!={;v76wc9FrmGDa3IvxAIx1M=O_M66Mx3mO}T$_kZM<WW-
zacT)4V|o^^r2+SVSs%`lp!x`*A~gmN`#IzXQr-G&v-k;jGsvj$xxlC5WV`x+@fn@t
z3<E=~SwLLEcNMDlX?m45H&}9QOmcs@WkSse%Knns#8KXQ1O_`hQPXCf(yy^<x9;k^
z(e7cQT$y@H6m6-pS&u0x(Zi0zl{u6pUr;2DD$Qo&%RR<bv}e8*&u66IQYP3B2%}32
zYmK#UBkDTPuzaSvtp+AddTfkLj+k?E{HQ|RHR(C|4okFpQeqOJkvE}2lItN|`WDMm
z>GO`bo_bu@1T!nPm0+CnL_8#<)aykw<%V(ZuaI@xSx8mWc06KG#J%pjp{f3xwQcRe
z!j7KQ4kJZ&$)}Is1azhj|8~+eo22F+^0?_&mmGCTv4&4MZQ&tIvg+2TYib{JQE2U*
z)=M?9mKzmyGy7}kGf@0e*Gg3T6o{`$Es~-XlaLUGl!VB@t$$9z?ZK_FuXt@Doaaj<
zgh(Q4VgvsW3N)y1PFt+M%7_Wa3zf)|wmLc4>2_n6;UvY@GHR_L_n9y%cGe9xrA;jg
zjvRehR{GHwuUeM44j!%%8yXi3b`p&6A1_gWPLQaZ_iE(mFx&Vl`SUFiL>U9RvvA6z
zGlWx(++76CK;uh>9VRl0qU>!Nn0PRI%%0`=u7e11_8jqG(!TK19eYX};fGY$$GnWO
zgztm-A}lI5DOs5Ur7UL(lFKYllS_4fBNeK-+I?r-8Ac_UD+ABFy<Yc1tu4}&qWQvZ
zE{65bOpM5jf<L}~DjBWR$`&l+j+H^?x2lmF^Ju&g*IuMSZQZFb`_oZkOiV+@!_p5J
zkB(0HKjys8pP)C&Z^4_`v$6n7;Y6zpU>nlX+FENgg7Lc54gVlz?Mw@J89#C|4@ib%
z8D$W#G9u_`y1;=XS9NVfQnKgrg0#d~P4s)~%w#uaYK%{I3d*J%0|&|sZHt+myIgyr
z<ybb_C}3$7Ke~AT)j)PoN)^hQe%X;)&&eVzLB}o4w8263yU4*;O_{@6sBtA8=J;J?
zLG2%>K~~vdR#&C`p6CpyLM^Xh-8&aL8P#ePHq`rI<-|DYVxmgA1zhj|eD{r^jQ)tq
zpHIzFpYjWpw9_mIb9<jAX2=%0|KZRnHd$o1&&w=LTvzN!+BuEh<SsWY`^|4{2QH`J
z?n1S-q7~`cY*n%8M3gHSBdj_KHNHj_J<qOgGu;pyg3QXxYS)j?Z>gJ_Bko+mt4nH@
zydhRKCQ09vibnf8A{qtXG08TJQMWoMBi1>IrWv=IvgWT?ceLZl;y)(+*gM=PtWt~4
zwB9iqf_;c2pq$X*+@Bb>O)zV!i_r6Bx}}^AJg{$*Z;!E&m~%!QiMCo-@_A`SQlZjD
z+3?AWkj*RGJM5)cnnrum#)@gPVAX9;Ql5~5A8V**XG=S}vF@!{yv1ZYKTlm0Ro1_#
zX?;%xR}EXHP@iTce94lCweO2ljJDVC?{)w*^c}p=YHVE2g4Z-@8pu6Y7`T^j!=)s#
z0>Mev+N`<=CS=;Ffi)x~-g<_(ng|oT?JO9!=N~>5t<f!oXyn$9B!%cI5!LI=5%HCm
z?|L_PmU4X0N+?lkw<+1P{?tS6y{(^>qBFRiWvI<iS2`0;Q0*z-#u3tu(UgEY$R^S-
zhlR>RXdBvZG%FX^!BAuKV&i4z^$rX>$Sz1)o%Q<&@4(B5N-p_&)Ykxt^}(I`gI|r|
zni|4d>Y+^+1**H+=zHh(!KqT2m4daJv3M!6;jXC4v8JbC^45*x>#EC{!+kT<9`lX2
zXi*%GGpy%z=+?}hRguIuU%~rI)O#OsZhoqDFkI6sZW{QdAQURdx&;T5r~9oo;Y#h7
zyH$N1Rxwo1eTmK*%qw`5&Gd$P)ONdXcdq`833;c}rds6>wk^A`m^I%Q2}oQ~wRbY5
z5ZaBaEPnkMq#v_my(o??mdBh_k;)nHg!XPuJkmjrZ(p(1THjSPv+C<o7pkAEpwMNZ
z(Wj&Cjk*r=>j$a-6G4D3-%<<W-P?jk_KH15D)Q4^kZAS+;8u1C{SRXRCmK-f^MLl7
z0X%hOt)okUQ%?twaBFi1y0Bo-O(t4;>-}w@ojpJ+oUT7I*h1_vEsY3$p@$e@B6_LX
zq`t$Y*z*I?qEk-`Q0~bV@}71IB$C*(T#$bT9@$r)g;(g)=KxfM{(N`!5eP(t55bj(
z)KkvyiDB2r7Jk(SA`uhOu#7!6o1u86PXVFd+ATJ(89zqPtGFxB17VM#mY*P7U)2mT
zbVW%-%yAU@+ZuMsH7?NmLqMX9?t(Z2;~EB_WuJL@w)i~)Ai`AW$5d2_Xbc{w`0kHD
z>&CtqwOd<DGkV%r0Lk2bcMBec?%H0coC%s|bIYyS+W-{Y8e9RA@5<aNh<7R9e{K8Z
z7KCQE(s3OI;zkyP_KTq%03j~04G_2x-#m--?2nXwIv^)}x<zvYVrIv@etvYk0jzpR
zC8zs$RYdA}dj1276~i#^Q<R4RxOSx`yRinBD0JSHGXGV0U?%7ue5G{Hb97`AdtUfe
zN<x<x5!t@73+Tmd!UCW}z!HbZM~vv)22#tcWA=8AujpxR8R#o>uuTw-Yq_4s$rzgG
z!$_3-$t8&JudPAsbSF#Q7pL}Y0v;eDU}mxjHM{}n^5qPzcOOT6DVp0d{Xl4RCX;#|
z=l_5h?a*6S4}J@K`)}=o-X~E4(@@LqhHXF(j(r)x&8O{esR3(8&9VPEC>?qaSOOx1
z=k|^}r(Z!o?)$^<@v1?ku-Cdaq&^jma(-^{5l9&?`{eNQ5t`84zzK*pJvFp=Q#7uF
zz{3K7Ed}fYYN3AQ_gBb!x)ve6^-tgX(Uu3Yc>pNL`|Pk<c*a_w-8UT%eSHDFylsT!
z>MIO~9btJGbm%j7P=9DJe$pYk^!gjjznu!`5<+ON4MauXfV>ofLKS_uPY!kfDrF#)
zdv6((QM*M8c;VlQ`e%a7%2FQzx_y*^F#SbL?cCEv2{WjuK`a1qW`CGviI6Z@0Q4!p
zkA;e_lm$rX!W!;AgNAAmK{Y$@70`QgWLfJBh;h*uymgQ7x=*EySxNRH!~Br1=t~(7
z;EFP6X2pOM{>b8}2zlG$y`8;hZ$-WUd@^NKxASNaR<<w?&B~}-+egd_Oq~S;6j+eE
zcV-X}cSRNF_q;{hyU-GlfehdC!uj0+==}e6Ic*SK193aXb!6}r2!DDG7@Gj~ZK53d
z=M~2O-=QD?(R?bqeR$*dSeV~3KLc^ox`;-ZCgGzBdfQ&eVQVvn0U&`^L8SM#WCQ&j
z%Yg9HBM`WC78po4fs_v?vTi&;bT!Ln&{!m_0`PQw41Jqd=<cV!CNN+$z?CG+nk*!{
zlrx6=K|GsK0%3Q2$CX)-!%7xqYpXs`z~8)IJ<uons{avSvP>`{B)b$nYJnvke)B%1
zhh0jhCa4){0C5U0R}xNzt%IJx0{me$jhCKc_+1fA=U+h?-)=PlHo&>VzoylQofGlc
z(hZG#zPk_7D0Ousfv5udn`|!5S2rDxfK^uD=?gz)6TTfh2;EsWAr;WcM}$UO-nwLv
z0Lyjv6j+|c|M;)$>!xviBao_3sR9}kJfbylM{n_A=dNaKK%0IC=BxYi*mL`Xf*Q6r
z;08f{M1HKK83rne6xhY>4ux`!R5|bt?77>)g!x9BceU+zGVynDk^2Ft+ULnv(5dC%
zN+I~7=chG(KLA`rA+U`+!E!&CI|l(@e@X}UvW2+H>Cp18T6{X}APbZJq3`|8s*Co=
z_HccD+f0QPht|4_+m0ouew($3-N<$6ici@?0M3l&W<6Tc`TE-E_(lrUa$m*&BNPAD
zJ%{ZdkOwGcUEkmfz6*ywDBeyJPdY2f+2!)uYxynTMlE|#^M#w;0fyy7mc6KGlz-iI
znZDi@c7u6Vv?Wx<BeK25OlXeC+IxE|wE;awSh!V+uh6GeE%6naa0~Ax#!|*C-b?LU
z;m-6sk<^Z{>kGRhuTH1q(e|xOfZ%2iCA}#5!G?1xqGw<gL+ap@V%lu&v!?nNPqREN
z{HW!pFMltLCVwd2o!M<?XjNlZ{vQw#vWFO;;`H<RR6UP8*Vya}OdnN?bl`Z>W_~PS
z4nP1}LlLxf3l1Bnr_>+)DyruZkr1;eYb%7r-8d9aDB;%liGh*LfLGpf0`**Cn!YU&
z)4AfBmPhL+kA?0gf5E3s1kX^mWiJJ0yn4S>ZF0pkrJZKR7t%_k|2gj^JBuD~Bg~dD
z+>%rZ*^=Z!mxm1^-t3CFYmf;azUZ%9nC=UznAi$7t7g$7`nY|^aAx#dgsoJ)5>q4N
zQy{o_C>Jq16z<`naac>%<MsM&Fcv}Mixg55y@bYDFQ$%A@kehGC~;5|LRFP!zrKCS
ztGGq#f=ydOyuU6S2KlSM{(b%_LRC+Jjz&=JC1bX%bYVu}A>Zqg<yl4EQ0CI8AD_|^
zAxR-y#y^;jyj!X`E0+0G$<upF6X%>K^_+**&nbz1@q|(N=N*JI(@U{2Yhbv$;i#lJ
z?Vw%_u}mCHnv5-#rpySb;0I0SkN9#7M#xmyGG3^-E}oGnw{G#p@7m9j(CfhlTgl2b
zY`|UkfON=bMSuC+dP;KqX9>$OIRdZa(&INd8jB*&I1{2OrQVgwAP+HQF$>_iYsMxo
zsTr!GlCV^1dIHXW1#@zhhx`dC^Q@ienVPavh#dL#;vwFm_PaIgpRs@<HKnPA;P8-m
zD5=pf&<+|2e{}sJDTWdrjbb>jlQfef(<OQ5P@19j+Rqnif){BP1|;+z8^uLO4PPdU
zD%W+NMJvaXW$@9nCniu=lh!c2m3?`svo<~ub#?cw0V&e}rJO9B`hhPA^)Zx>Nr;No
zb_&gVRrjPsQ!3z3#R8>SP1v6ZP2v7CPBjK`BjYcR7M}=wD^x>hlQVjcWN~xya_>7n
zi6k{rsHAQsC!lyiZ22j{sFCWq=d!S)pmZWDWC>4>&(=(2ywiW)(fm9-*g-cQ!r>YI
zmUT_ZB4I}(mgq!nv%rX(RWO^OiHLtWC}E|HBltQNee*4$IfoJ%Td`3?)ze7nqW*rE
zj~?CT3v&(>?OfSJ<Vyph5etD4X&SSuPbp778415MEPeSWMJ8@Y0MBvGLcJ&s@k?1D
z7v8K16AlV<_~38KOHn6HJKM;TzWwrPKYQ;iblsQ5jXTn<P_t*#d3z=t9IsRN8fi@s
zX?fAZzFMO4QRPYqe6{kVoh%g=k;VH%Q?CB4hRkoKbCKlR;+njlmFqN}ar`e-G13TF
zTOO0u$i3?xHjgqj&Rh3YcbmYBam?xtJ29<^$mS(4L9)$y7>LggT?ut3^7)5saVKWM
zc^MRXQu99)$--1Mcv^P4i!9!o@WO*8jEP&pjLz{}YZmn<*;b^G)Y3&fb*s~9{>K3)
zLg=CNy_7|*7<o!%nrqS@qPm+yRtnqs>UG?IHf(*lxY^JOg6P_ibEarzIzfn6boC;c
zeN$G*Zf-rC$<{laR;H%7N*w0Ka&t2Q5v?2&Hd9!Kkw|~#OMDf=JqD7_i!Ye{K05w?
z#+`!Am<8W9*X~Dh<JZ?{H8$*rzJ2@rR6u((XmsavHYz$?hf1<LRk)(kr@x_M;W?Xy
z@P@Ho+_||cX1#vM?4Y$2Z4-+$xkd!nF*`j0jeJ;L{)ad?SGLGdK#IL+sC?nX>v&S;
z9^x!s`C_#%RTdd4(P4j{JH$SlaHi=DLkw&>0}2a$mAd`mjuh+7XBIp&-uhA&NgA==
zM3qwmF1S@=-;C3}7+bKc1^yF;>Y{`rq}s@%hb0xb5$~wN;$TIxE-5~yN^>>2h-h^U
z3N=o_Xu&awpH!WHZs)7FYU9&4wwU7Cq3qdk30WJPT)CGCkh5QK3Q{*&t|1b&BC!Og
zTvmDEfM}Kh%>H_5g+b*(n1IUQ28#|89?pcXF1`FhdG#I4gDcf@8jGC?s+K*(XGBl-
zp%N_)Dcs206w9+&7%yp(T;>H|MqL>3I&#(fO?&)$Qu*o48yF_*pm&PJ2WQm<3+_mT
zB8{T58kDLf<fVntTmh1WD+>PA#cLx~`-OEur3x82o3ymPy*^8K!%Q{RGoz-VHJjp>
zyc+M|@M_5S32-1ABO7kTT5g`<+)NaGmUv)ES<reDZTjhX0^u(=S4@@6k9Iu_8nKEH
zu=)?yiOEpXeuP(dw_C`?G3T;Hz5YAbHrnKHBNoFda?)c~kAVO@Q*L5ATGB&I*g$a?
z2k(R~pCB77m@wqs08}#@scp0Jv&gHy2tToK%>0O4LRNi)_`8SDNE56#=604Tw?*&A
zn=_7ICy#?)sFzs39uQnHT_xq^nqvLN%jW)QKd5)G+pA^utBvQ<5RDInGtALuf1tZb
zsOv>lYXB#`DaNdJ!n0(j-It$-o1U~=U5x*pU|7u2tPhx~s>tRgyO0YB3EAMaN~=~4
zmooJ>HJj@m8dVmE#Q5T@Qrf0y!TUUhKITqip*na2yCB$Uq1snh6YfBi#aKv{vrN_=
zV{HKF!`l_j*?3h`RNy>zWD6v+V(KmC%kZ2AyMB>#N#mThTYF#lYZDLCBJfjwJ3;A(
zGBw0D6uVRoLZ<WD{S4PKoJO4eZl1h*D)8ZCZY3UZiPG4%s^H!3#~4on7Q>zinK7DQ
zVg2LBT)$_6rpDoKj5>qDE#q5r`EAXXjNb8z^&pZ$i0SiS;7xArVlq_|8X8Nkt4ivg
z9|VMM_&Ux~z2C?PHF8oXTglFAr>gUIa0f7$tLW5O+`l*M*5&7`@q26cI2DaCyuFe(
zFFqHX`=;xH)7Pd9Y1)N3v$Mu?xJy;;YmHjszESK9DU0;a+fvCY;hzf7#FekACvubs
zC7^Q;4knS-xhn{zw%!fs89H&IGD#!qOb#T~3Ciy83ZCi8>MogyeLVALSA*{HYhDZs
zwo@?UwM5fA&93f{j>{U}@63<8I;Qbhfi^mTKw`V?Z>BH?UcpF-9C6bOxqPv(?{u7a
z7_WCE!v=Aj>yqVUoPWL33NAQYH(ilTOiqp@Gqthco;b{>9g=kVdYIj<yFkFI*#>+u
zH;#bSDvp5UaC8(CrNzO*g#M(a=8@urKRU)EYk#aK40_JgT>+Do;Z0k$_)m>!#)1OT
zODELV=`^h8GN__>uu$J(EuV^t5cnu@#RtCyqrPZ;8i+5#U|xfN$&V<`PS+?=@NO}T
zr`xV5;;Ea1FwaI}Zmf02k8x5XlTB>urm5a=%O{j$(i#UC><NOG9Hu^HC0XoJPv(EH
zd~<%yd#%Ma!^4!PP{rzj8*Gitrgi!U)D6pKb4lL2j7zlL%W;g<Mk+RLREr2ToL?H}
z7|Vf~lf?7Ia#YL@3VA}*`4LwpTq#6BZ<xlYtAY-xI|GSurGtrCaO;NL4pG+mWFN@Q
zAAL}b`Bi!~-A{J5(bX9_U74X^UcPcWK|*q%MzAjzQ_TNE6y2uT^v8#`ZCe60zQxj+
zbAO1t%Hk{$?tF5ddVO;4OAUvGLNKjDflM_0ZY^@~a--gNq0cPy+W$f4?R1a%s{U<1
zeQ_Yfq&<K?LoIXGb)KfPX~va;UH%s`E*RjysJ20W=Zkqb(>hNft^l{FUhIPKy<_nZ
z4S%;hv+14A)V|wQ{Im(-qG*y`T5TZ$OD^pUJ*8;rseY{?{+1%N?&qhnIlkFO>N%c9
z>sf1UYg9oBB@``Spk_u=`L)n*{Kt=HwI5xV)_Yu>N0c83;YZIV8)}F-j0PrwJ1z2d
zwYpMyCCv%RBB&>s;!%}A^jEQCF}!uW;CDg7F7LR+*3X~4658)1I4M#rC6g5`U<q4o
zNlvZ$9oxN@b@C<#{L_`NUWY>q4Hn5>iYf-bF#hu-&#wAi>6ChQ_j7CE=7f;jM)i0#
zxuRMorUbjKn%aS%UxR9&8|P>wuV>rtq1GCxJVx#$xfW`;(YG!=hvpL3kYCg-@LALc
z5QSUWdI|nLpi~K#n9&9@Bk{P9ZWnzY&QkleUl3yP-_T!igTa{iTN%nt3nk3~FOFIl
zvv!~`5ti{v2SqB<%DX}v1bLB;OTNgwOKIO{)h*6B#w_I|OKq2z$vq=Cj0EwO#GBC?
zIVr*7qpI`nb+eM1Yr_uuFK^tf53uAao*Lh2I-t^8+(#2$^mn_xs7>_HY}70*<y*{b
z_dh|ftEm|e)Ry)q)R48ZU2;@S7Y4f$$ziDD*JH67Q~sP()UDfDI6Y9<!)x7To_ZYR
zT>#l-B(7}A2u}14Ro_zACcKgA)JYGtnk*|~MdyJ7aquE*wPjbB*6Nd^SwGay*?CP8
z3vI}X@FYwUo=G2`Y`gmU2_*4P;Z$ePC5nbM$ObfcJLzrp!Ptufr?P@S<=AyJ9P(9u
z1uqK_T@f+V%V&-;Y2*n*D42#gZ9XnK)nZLSRMgK&-xF%;xi~jl|0aphsoSa8>3v^k
zv{r1Ugg#Gd)A_Ixd;Qn-`clIl_AWPZnT5$5=Ysbeuwt=Fa;#jvb%I>b4Q3fvjxheE
z619bTa3p`;>#e4(B#tN5s4ka1*HyiAfm7e;U{<{YSw8SG8&b8!N=pC&n~(#fW>kIf
z@Ky*lUk63=uE{ua8ZEQsc&TUhiNjPX$zGb&eR?0On8lX&SaDK*U+Ls}cU^G1{Q1u1
zivx2PXJc{lg6XjO)i0)kMsp=c8(I@XUM;(*gp{w7#QrtlD`Bju>Qa7H%i*ee@=%p>
z_6j?IOvCV{rP_|SKCje!)8>x3^^DxDT+ch~wyoN(AEr0Hnm%f4z)h;ySPF^Q-l>JK
zrA;64WaPu@Hc$5J3;XDjj@aAR5azxu+xaMC$d$j3hD4>e=mSN;NCTVB&^gCxORcvI
z`>Ht(C}-WXwexvrdA>##*u3-31#y6ocFb0uWYkHq%|`v!I9qe?Of>aN3jcAIim|Q1
zcX5A`c%~1bYV@Q883B3GuibkejEnI8j@Ozfd9D*Go17RB#rT=^k(&Bd7=}yrXF$BF
zb|_@B{gjLD?fXbv+v#2l=5*??3M`^N#D&w2ioEg3>!+WpDpmIUd6`Icif3TG0gZwS
z)1F4+4$}s>gF((4uT$t9zQw@GvV0B0X%3q>8Y>p^`<glV0DTS%9vK2>CIy|t%@e_w
zf{8o%kv*%*64K1<eG@S;aj*MtQe<YF#+BlL?^XW%Vn)2U{-MNR#T;26X8J89!S;tx
zfl@M4BT>_RM89q16m7}s&$6MzRQxR6hkivY8ae&Liij8rnKx!d@MjyegC9cg8u7pt
zd21$1mih9}E1XQ^p<jcY4C+lH=^Di2Qp(9~QePM8ug?qW4e8foR@+a$n)ph*tCbT2
z)v0|p!TeH=m5D>%dxNY9!CaP!{Qc&NRPI>@X|J|)5C4%euB;J%75>ta)<ZS(YJww6
zwzmXem2jz%Zkf@*mqTA(^p@~{Bp4`J2Le~x6o1zE2HL3)d)WLOD4-f=0x)W3Ys=DB
zT24uNt^%LmRvK&2rs0q1kMmeEPZv#2KMnlSHt-~8Wp_obTegz(AudM<;+%Ox{s;E{
z+t=7hU{7(E+#*HQ4DCEg;9se{^Olnv!%9X`snB{JRrMR`n!SVrHZr`I@ddv*anUiJ
zI)4KgsInfpFH{SPi_v1=(8!F4Jm-z4uH$k0gK&=7G*;IA^3)9X#qw22j|*$Oe&DC%
z(EaB`6v7(e^u+Ie%^YT+o5WUydzOTpzyF4*TM$yOQ@H8&rP7n>ldp&BeskGkZ<u3=
z3LQ6d5)Ns30}>ZEaD|P%Pk?93gdk!~F+fjFPlD`cBjK^Os*R*wsFP2r?@X6GQ3%B(
zRUYu?gz$38-cV5#ON9XPmYBk{*r<{rM`2h{M=xZZRo3)<1fW+(3SIx`op@p`TME<M
zQtsw2%c858W0+M_H@+!pUJ{_fqcvF^7|hk={SzUjAi_#-a0{$#)uou^PnpzB8zzjw
zSgU5V%@cG!=ve2cyeA;HuAz(iNf-S3Jv|eRoJ^oL!EXJQ<7`oJ$+^HD7~Ir40(NK|
z(E;LFfhcM)SSKyCTBRTeVF?^2f%$4e6%CLWG$bVIR$^x41AWo=w$833=bJS(dDoj;
z@YtKSn`sBCB=mn=oLmHe0aeAf6GWQGj)0)M=cfABUK6|K(?SrTkw*R6Qe7L<+Cbjx
zm>HY&Q<q4^hVpBv2gilD6Qt3~y6evlOJh=AUaXelS{DTMUp_Z-eZ>_H8O2C39>JL<
zY)(JV{^Dc8%mbZ?8GcJ%$^5pJejuWBm07YnU~=tya;hZ1M7?=!QMeSwPlz}(Pk?&e
zguerWE}$PB8z|a_z#6P`PBCBFEw-IHqpugQ-q}`-#rC(>KAT{FN$~A`j#`p}8&8uY
zLptnP;Q<ZT&${lfLv?U-_Q!~!QS5r1AK2HjUX*hg%}Vxoe4NX>9QX%aFG2_fO&kgA
zM3ysAyrGk6O48!W;U42<mfCyuELE?hMTvv871>{U<C7cV)JiY*Mv|81oEem--3mMi
z{+^DPKGcFWxW{AS^SjHkD(@lHF5?cMXy=)*;jh}~kw@e}I}%D3?#BB*RJ_(mtw@DN
zedYJ=wFWMM(?ztgCex{NpQT-dr0(U+C2By}@iJaSz`8yuswK*}@9~+}T0+-g+2aY@
z<aNrWZ3qq`B%kJf8d9Z$WOK0g$!!YT;w7HXBa?I5P<vTG=%&ass4DkiUM-@Gd5(Mf
znwc4zR(+x8gc!Nu9s$((x_kpxYH(ydoN<a6boc~{tIKtE8wV0(`^*A0`3Kpaz8Cek
zN#VP&+g?_SY(5AJRpt*#0N0I{nMkxheU=UXutzG{ddZwM6l^TPM40*C2e81R#O@Zb
zG;y>?!!F$MXlVmEoO#bl{C^}gx7s;2f!kL&mK#tNhl>tEKjv9s-bH<V&?7D`WOGI;
zfU!q+1JIP3;~>lFLRSk=_<wT1l;weNai5sqWIo~sLUXUK-*o_MW&cgGl&Kz9&BDzf
z{6oLLCz%m7a3x`2)Bhu6M2q)3?^h31j*$}^AhG+a4|{>bW48HLuNlxos78~k->du6
zK+=_JaXN85+y{NQ3WeY*4}Qi5XtWA2eSL#4GypPx!@D#Etly!%6FdblvSep=OQ<0<
zin}-eei%*FiQOY32SAZUpMLe$wqp|lXzzaiUV9jOAE8Cp+Lpffpx)rizxBWlf{D{|
zP~V7Qi#<p78N>mt2NLB+;2;a^Wls^ie`5Zh1w&xBytcrlA3>=To4mM)hpPif2c#~`
zeI@olHX+U92p;(gcx;a@uhi<a2jX3WKG;Qomc8?RqA>pE2`pqFBQhTYrh`{2;=lpM
z)W4@ePhSBqC_iExi-=_Ydut3A?XtGtLx3MD_Rvh!rsF~RDDvp9fu3^Q-+W)QM3Asx
z?eQPL6o1n_4bzW=q-_5w5L09c5q7sQU`#|~Vp|ags8uxFh_L?J6PMdpk2va<P#qv)
z{!8WGi2r;R<6s|Yp!q*-#9WQGTkCI)MEya}#PlZ)Fn;(TX0YI6nSNxLU#Si|Y_%N#
zdM>Oc8EnejU6{9b2NEg<&u2WC>ErAMfiMnMM)k{cxnTkSfGS?0sVamfS?DWzqYDtZ
zFaz|HA;J(4x)XAJ8wa)B@C;N)Wu@;M)Isu}Cj8rw|C5fGKmJ`u@H`Am9}^aty){I_
z#~#4-DO!!_cheWdZ~|Wn)$Sa*e&2|{L`C|(dOEsL{*^J1#SQ}-gzo8t{Iicovgr<z
zMrhlMc>6>Xy)8XbOlkS1sg+Q)C|&M;B>EOI!re$bJghn!YvFe8hW@QW)^V-em6OC2
z<v4{$vG*)D;~x;$KcE?zi0O@gKrwlWh4UHDyQ@6t8N5^bUdFrbPLC{niJ#VT<H%lh
zOi|ffU0da}P4^x-MjhzZXZb6l(vodQ14~<M9LCG4&i0<0%0c88_Cit>&KEY3PeQ^?
zP5Knkh@Km<UEc)}&)O~-4Nf5ReQm?|^fWIZaN!#g|F&=Z<$<DjkJ_jn1XUx|*vs~^
zIa19<<!VVRZot+)mR2F5#%<TAPJFHl?RlYRwfBK*1rB7p{7`-Wag|eteU8*xg7c|1
zouCZ^qdB3r0tbR>cp$-!s14P#)H<P+(PWRk#vfz$mQQwWthy{=*c=xG%r4fKVg)*`
zgs)@QRP)*u3c2#>^4X-Yf3p*Y_gEkjx8%b)By&%xuWFcw<#8uAPiG{Y;tN&RGe-ub
zG{hsO>=%W!LjI^E{;?<bT{f1`Xyj5C<*Js6Dj$dbnc!Cnr&dh}W#WAtP4#&62Wb&9
z$4FeGHA16kddNYbIGJfX@%eW?Je`AlL}wT)6v5;|pG4{8K_QrD<m}_ghG6C@s=6^M
zR?%J?e9JHtwMh;=ko63{pp($ONWq^t^!M<^Uw)kE=xh<Y)MmlM^pzlLz9NeYhxfC5
z6$OJuWl(~p%X(vBSD@W33$}`J0pX{o-xQ-Pr#bNGzE(e#2qryTC?()dJuhmbP1w^*
z((ap77`jZnk=Nqtzl|B}KSc&_?`Fa@222LjioZt)8xXfM;j*36-^hhp_qX0UJ7ffw
zb8#IxTzZ?$bC)Hp)-Rni3#j!8=>&hIpM`*7-);BVo7>j9p(VHIyE+WR+eiBw%`Ke+
zADouGu3Em;TBVTJvbBo_%pW+YYU8^eo<zK5oA7FhBf@HKxI&jX;o@xGrmz2gNlHcC
zTToboh89^OKb;a^qB*MZ%$lI(H6->js$=^-pGlww)PTrB4@wp@YuQUiqF!T^^`_|f
zo$6(I-`l)?K6TXZaFyZy(vnO3X@<3F>$w{jM=Y_&rV<ho+9c>3{Q+Dso4NXHll9D!
zIBA%{tD*^ctE&ArvD`_IgVXRFg6S*tx1-Id5whUXa*(P8_o69R5#o+;G{VO(LRSVK
ztjQ~(xY_*c))v!TYrFdbmqm4&;d5i5yH~38OGtD3E2$&kH*&XIPtz=6qzaSSK*hs7
zoK_X7b8dBci9SA=4G-Q%P0?8C>?qE<nbjjGoLT#85vSa{$KUl}o%keXBG>EfAGLX3
zU){F7agL=sTsUk0R*KX`I+$(Db#>>lzdmJiBVlikL<O;pPj9qM>%j6VIR;o;uN&BI
zj<Hq#cw|I1wNlPy)6W*9xE&soEG8U-`B!W#jGOWK&TS6Y&5Li->Zekg7ET@ntxnkD
zwdn(qeNsli`k^@Us*OyVw>*vku2;9--NKuhX`5!>D#V;m9lI8pVz4)qVq(T;{(Y%N
z{z7=Dr&&heR4$~P*Zvq%(yJZRe7st`RR4Wro}=leE%@ZaCV%$!*)cA=M(b|xR>FQA
zs_ckuH^Hmn3Zhy|>Ueemo9c9^Ju~oe(Sv0mMl4Wy(eBPMkm$2EwJQ;^y9Ed5L+;A7
zZDhwf?FTLnPEU^jZjDCYR_Bx3p@a1dUh#yPb7j<s!CN=u!uT^9X4}C@b;TQ@1{~Y%
z?!G%yKGU@?d|ENE1WV5Ef$pO{sU--o#;>F}ob^)c-!voV`boOxGpEGq=@Cr{%sd4V
zJn-=sU&oYqj$VofNUAtUC`@?%er_AddpPxz3tcG**b<@!AMrRmTjX>GPeJtcJiHW{
z8C>Q@ZCzw5<p!kmyn3Tmd_%ij1NE_F7c^}P+CIA!S;?1spq6puwT_UZ)5ni1IeHwF
zb&l5@pn7&IeTc)o^tRX|^Ig~o`-Vf6W9#v;&i*3st2w`2?vY*o3x$W{U)I`dN;Q-*
zCwU~ngBWOu_xWnC+24}wYbbQIcRQRApk(?F2;1#O-p2XGAbmWL#oo57E3IT@=Ksep
z<@2GY1`jts_RI!H5s_ICVG@3ENodJc?a`(8=SOm`wcJS945-0W&)Zw~Rty_ga?0Zr
zf=+vMyoLZ}jvB~b*8ZI!F{<sp`FB$Czb%eYiR}D-4uVGC*E|TK!utpG>i;$8#GBt=
z1?k$BSEDvTgxX1t`y%CQ@cLpg3<78FjVwFbBM`eC%?d0E{fDaO(?FMGh~YWaXY?P!
zo}tzMnZY9dw`Ko(QYrPxHP7ka$q|7GREsL=^xuiy|MC&EldA?l7HO1w!?n8@5o;~{
z1GmWD2(SZc1O!zC=jWSqXySUMc98G`iU4S`h3{&(cDUGtLD)}D*E*VoKVysC=f#Um
z{9m&C?@P3D&HK6HUr5xy22&#2(pPl*?{xORce1|&5i#zDqF`GfJ#HFy(LqpcR(hJp
znFIUYu8^P~QH&F?&n4=@y#xPja2u}7kXPOf(A2R7FYNyB1~KZBIRlLPzj^A(nrCfy
zjO)*fVd|c4Oi%v-X^D8a*zF+OM;@Z@)<0=5$Os0ui9~fUEMQ(Rt?lj~DU_e&dVNt;
zfVVH1@Gn&NtwL+@z662jPLP4<_#9TpAH_1$vsX}E)$0hwZOD_0TPuKOg%2Pxsa`t4
zBid}1;=xVRYApJfI(><+`fqwBb=#s<)OuXm7b;$T1pWU=0~p<T3F?!rBeVfTJ)2J#
zW^i9BNb}uT1y_p5JQ8+0+&;FPVb)=-cSHG6iKgl|0{w-Np^3}y;n}TtxgZjSM-MA$
zYQFn8;IgN8(<VmQ>T%g4Z*%e>!we7IY9;+M{${todYc>`uJp|}mRpv8K-Le~2Vo2U
znZE@Fm}ut$+drspene`y>^E`bK}KoQR5!mQ`=6d3iQ3vwLmW7=Z+3m@m+iOKid(sD
zdGuPnBHpv;)A_#AQ)7*P2QM!qQt=HYA_=iQP|Z=v;9goMxZ&DspOAIQ?*<#}Tb-~m
z>Gn6egvOZNmJTH7-JDNp{{t#S32Qb~S<oQsEAWY6&TXFk#Xh>v>VRNsTbtkg|A3bC
zn-)JDgiW<BmOku;)vrFxmuW$BGAAP9aGLWGdm?IN)q3;Z+60N*-M&N%ZFU71u{uP5
z654?6ovZ~xk%6B|!Rn)~`k!;b6HNeqye&^chb^`|(z`{hUUI6QSbZkM|7cIx`%8GS
zW%inLdlJrS_E^(5{9&;AFiHAC12g?8L96dG=6wz4QJS^3JI%{1{i~L?xtS+-5!h+d
zH&VJ@cSt(OjDBRw=97A6C=PU3<ooPzZxKFbMx69~{kfS?=h}Ap*yLNQHj&)z$d2MX
z_?eWCZR*c`smq&aAIYk+Zhdc?W2Uz_`qgL0SiVES14KTWr!z`T^Y2+((&svh8;OYA
zIgqw~EVKYquwczbF3IjArsjdND0KFjspL?v&4zcK^E|eqcRCO^`L>yEe-jO(ZQ*0l
z2+qb4rm?YCn0#aF?3w#*C(T)Et51g|^bHa`l_8qR#2f7zRIoM-Gu}8+T&yuQ>%RMr
z4;J+=guL8FYE0o^YaE_C;~IC@Avv;C-j<C}cuYGCPIP?`X`|s~y8_hYkqd6ank{uS
z0xDpKq16fNUFQKlgR*}>&zObP9NdhFf_v23Iuy;ck*Iai4{S#VSJbivq2&gzEXr<T
zLvHyBi3WF4eA+PcSo9e?LQ&()RueX^Gp$}X=zI%>sOsCRp^93|SqWw>Fu}Iz#)@~l
zd}`XQ_ss}bq}oM*Zc*`E_4cyF%D8jOH*Q-NXr{X4jPCnm4*oUpq{X?%!S=dIy?m`F
zysSPojtW<>*Ecv7zIJxXlOPX`{C<1O>hR1XAuWAj5W6v_E<3Lo@E1qo_DyhA{s>O{
zuuJOCf*KlRuqSc)Oi(jeCz*Yr$$D8a37|zl0;$@6t?eG~k=_dJFI6MASW~oD4|QGx
zvIxYR?K;DP2SeJE)v=?#dOZ?VB1>$Yw(FtWN8Bd2@g$3Sa$(K=V0FDl4yq#OBWi}~
zEjD4J>EK$hyY`&>_rZ9C?$vFxcjNc6K8_p0%W_#yJ7$K{=}Fn4CH^+Ea(gd865K%#
z=E^>M1#7#(_0p^4e_Ikew#<5Ajygi$X**|&7BD!wNDWGm5D%$FoUoVMevFt!%+<H=
zCdpwbIaSdxKFvVXlqzbIycCX{q_N#@-d@&;!^TRVFAqZ#r20&-+5Oc!5Ow7BnLYL6
z*UviPPZgbng*^Q9_zXlIi0{0ihlf1F1U<gzW@0qiD|2R@y}*GI(u%PS0H@N^?dai=
zqG(xXR5_!EpntEMuPvf)_TEm1KdtDfKbheXxf7`%Q4)3vT@^!j66r$M<Ifzseu8<Q
zFXKHjlVG#wT|atk*3sDMk-Hy6={1kYyQhI76ip467ic!AvU9?`cOU8DL9(RjfrT!Y
z&0e>eI#7a5;-3=t72+Z^z_?ksuhTJ=MY^!|(b4%CP!WYQ;!(GY!!u!icUW-=P?&Q@
z2PtK~2W5b_$;KNJrPr#MWS~YtD#I%wfp6@oYD0cZ_isQDC8l*LKb_bt2p89tJE&po
z@{b1^dkQ;*jLPtvO{!Gj?mYS?k{TeCa)Wq8y3m}K|9jj4S_Z`j>_fcS9WG)=a`pV)
zECa`oT%b`*$Q^(WdS5UP_p$_L17Ly2PRl>;+YeNInIEFb?g8n=Y_dmHd<BF>e3I?%
z!~oN~M+m2mRwKl8?lY#!{QHrWE(nm1kc%UQ6I|(G-|y+8Z#H6RnK<$INj}{T_6XQ$
zS^G1B%|7oI8DaN~`ub;Vy_;t*?ycM#ICbB_6&-(#+;wu_VO*dbF_QoGLuWM5?xQH2
z>>YJ7D&eO}3%1mMTRr97a^Hx1)A(UZrSXU?BO5w&zq7yeI@yC~P(}8muOSlrHvn_U
z9*I)>HQ7_sXW1WJwcWdq^EH#zd3?Dr&g$+K0>^3jX~BtzWcRfba5J#W7$_QO4ekH#
z2-q{~-bF+fr0=bCZvjd%_EZ;Zpf9dKdi>!}A{Cz+FxCS;T=xg)h7Blt+vw|ia{_P9
zR_^^#sp=x?-g{w5FMu|w{B>S@p!0yf)Bcak0+}6b%<UGUv4^&oMa#fZ6Nleyk>wq+
zD}Bn+uR#5$YfzVDw0jr%>Ky=d5E-en=e;%2EVK{Sr6;D5>!<g2TrdQwI{kHsM<>W{
zRB^RqXpAjhPFq{-WsCP68QtU^fc|eE`1p|kru^4$eu%Rhf7bc8YCxt~0lNJiz>n0w
zzCtU$mbE>FzFBu^Py4r%?#+n{^q(Go78#lixVa-TQosGx-vLB(v^@dM?5J0w?<P(@
zKXU+S0Hfo>PyB?Y)0<?VLzKB<xi!&|$~NBmVr;+pN`pisJ1#7wr@JE3jn>gOyP?Xq
zjBiM`+YNzX0NhTCT17(+rTr}n@R-{MV32rBNu|&0i|L%s)c_Z6LGr+u`*&CXRzvlF
z=0NKG;X!k_w}MaW|Ha+AxFwl&f8%83RE{;~m`^%sS>|MkGnHAXSWQ-rBW7k!BGB}t
zrY3j@@c?KxrDkPjNo7je!Baw`DRQ>7)W~S^K!%7XG(|)~P!2l3P0#oJ{@&~T6CSUt
z>x%o{_r3RC>$48`+H0>(TKtuzUeK3-ZOVA>{eVa|Yn2;e+{64A0LLZ&{vUCGfzv!!
zT{TPi&y!!ZgP^Jk{S!z(-EYJ^0PFuxtcjmoI<h|qs=B1!zF$b!!4T}2f#Ndg4-M0D
zErhNEL)GsKJDVEwa7->Q2ZY&eS(^LH2R^@%!^{9w`yt?ZG0~@~)yT*HBksP@=)Z(-
zL$!daWp}&+st!P3NxO}#YW*N62L$yr0HiVUGnnVN1k<b@@#utd7T_UeE`HL;#DRBa
z=ZEO0{v*(To?U8GD*E&dx>e7?EWv#f@$MUt|E1S@5Q4^(-20L$8#gMoD=9a8M6{%0
z4yqXpsfV5)zq}(4uhh00DRibh8x)_)ItBE}p9#V|^JHr@x~zPWZA(km%J*Tr4ZQ?_
zs21vN&$qb#&=?vP{|BBBwlB=cuCKf?JrwpIER8-K8#XGZJ6}&f22gG{1)9L9?;~5U
zHqrqlrbc_gt1#~YQ(*5=*sK4;ps&hF<#c{#xEePxV0d7Bbj5j%ns5rCmQDSy=z5<P
z4h{_iro$XRcD8vYpG6_lJWHNi!<nCt8&UX~pJR`0{l->Ks8>w6+_lSUs;V7Y!+%LC
zHhlZkub;QnlfyoA?9BL~(3)be&CcOf3n4lu%PuR=psw#Vk3I;pOn#_1|C=i4DoAGr
ze~b3cmOSB0H5**owy{XaA2)G3Tj^Kl9R;HI4*x(_JCB#9K;N?$ebd5f-FY`2HT8Yy
z+8fspr9V1;4HCaz+HZWxf8*bx7H?y<o@PjaSIU}8{4v<Hg-acVMQUrc+LTKR@P^rM
zP+w();0(JWRsM|^p-yG4N6Q(G8eL}o1Mp#yuiN=yW>lv6;1l2>cCGEGy3pD>GvST>
zG1kX#df&J2<|sETEVjd^z=0&Y3mTg8=Ob5bznN~ajOD2q=52o(o%3!={;2R$S0f{p
z6mUOs({J+If5+aQyk8)_t+nkwj&RMwfs<b*P`*Dl<fr6aE{%RFWv29`Ag=qK3COfW
z6Q8kI4O`CpheU?cuhCA+O637}_fn!L_x(d$7;07o61|=poxUQu^8AxpH{2SEy_Xl0
z^(Mc(2va2{lh6p|PnA_K9^sHBy+rQM=>yIY_ibl;S}v(t=?53yCXaR#E)N(;Gz@_?
zcY`q<9AIuxhLOn^poKjY+lrG|>k}NGmVg^RzbU(;t0(JGp5&GbD_V7PWO^K_`bu_%
z5V5CH0S(hhXSaefT%B;U-YX?!ZJxcAJ`7vcCYO{)mtuky#{<6Q4$G=9nJiRiXP0=L
zlNIx++)JCILCq}+{mu4k+^#Ts1Z5rffUFyB((~j{Bn?KA{M3D-4Tq1LQ;%CRhcT28
zdkic{Nf=t=H!^8>#>0xPn1}Om^@yr#T}f#%4^#9<v`Lz0oniOcjBp$(<s7}d8|AJL
z4EnBS7l-QcQ%aQ$BC>e4C${NTFQ&wgbADiuEUSJJTFw~jsmPz6<D?Qm`N|2-KM-aG
zxl-qBxHKDyv33ehvVNe~>t@t9PFDQ&Hn(;z-$OSHWn(3j=4kFK^|VIu0kIif(PyR%
zxY3P5`OY{66W>#MmRIuLRrOa=;5F~)!zCx?ey62`&FTC`KllfS)hOxJ1%pruf4FMa
ziHPKun)~#SLA~u&&zb42mDi+6lq%7%d9XNktR?cEgVwA}siPiLRPp8F2hlY1`_{5_
za^u~^07spbW$_zSz-jB?-ZaHM!ouPfnJ~x8HI>62s8md9ojxQx(K;_CHVEFPzqO(E
zv7g=d{M#f2JKA5*fP;E)>2&4g(nM*kplZwVuXPYNh9p#`M2@5Fe^?;TtOMl{-H@-v
zaqZu1EhE1d3dlJg((c34hF3j$a=QhLvwnJRei?Y~^vT|kSKV`Ejmc#<4L2^HRL5yF
z<L1cmD+>D>13vj%hKjG40oC=!sAXStChz^Tb6BamQH<eIM;66bmT<8TcF;$v=@^>1
zOidA=Xj6%B@E<PC&6`EsPKudGcdviEqUg5BJ-n7-?ZTSH)L4g#EG)YSZ~dJ+=f;}I
zE5&=YY6`z~TW<j-Em5$Q`WlYxk=FgDQf=8u5vRhg!LEnBt_{S;N}AqiL~r%j(>);`
z&Z>8y&U?AxAIRGs=Ld7d=hcTs(@S30SXp2Tf~ab}wC~Vpi+}|exTfDhe`HyNK-1Fn
zj1-&sv>HKpO#r-p;$8jY@z7}0F)!l&BoSGuKp70gZ)AvbaJFAW4^F92g9>uwjEZ~b
zx_;9r_>QB3$P`>yRZ$H`t8Tgw6Yoev(?j0d$o+F_UHXvCo%kQ<BM)v;oQhC_qEF32
zpSq7?+>kiFoSlw^#SD-J$99zv+~m+c`A)TJa_|7PzO$yxQA|0R+#BtDF3fP{V`k@z
z(H7CLD$^O-u16FS#TYS;De&<(%qH>x6Px53jYjT3KmDYkbOKFoinscKA>v09iGF_`
z7L9%2p0)@r3hMAwPRriM?AeZXgx@O9L89B_H0U2>6*Nf{9^h)u+H>tpU*}XS0qPl$
zWk=L4Ex3|?uUNyWvBoNO6@ggts+0Pdma0yha)x~HTHS(kNm%Dy3^TH2B5Y$I@tdaM
zg?+WGk6@*&o$yhkU?!U>;jC&|g8#2_?_HNv%PU#ua>#Boh>`6BBRsz~x7oxn4^D&j
zQXQ`@e2^8(Y-6?T9Cm%}+LVyu)Wm-Nc$~;h(B`4I&q2CJVo$M<mvwWZ?S#}nJQ$5Y
z^GjrFj%NyaGz#9;$$rZiNX@kLh>X3z8;{;r5D_sUeXFUTkqBEpe$E#h*dZVI0-d(%
zL6FoUK}-6kPf`C;3^J)<E_8AS`w?7$Np~&ENxCY@vsDjC8$P>*;E}le{SU6{%?j_C
zR13pd*OCaWU-a__RthK&dgb2ON8L+Kh@|bLKF;uD5(ULn>j_-=u74o5k`&zbQ7md5
zp|<>{&zBC7RBp8{wymj&kF}CCzOS70k$F~syk?fS^0drJ`7|xH*Uxn&Q|7SWKDLj?
zz=dQX<2N4AXcqPfr7{w;j_QVN8epGAUZBl)(pNI;e+lL{<m0pnQ^PFw<*EjYhe=7P
zjOEWLZ|`eGk@)g1{Gx}}fhiY_^7D*q@U3K9Q|abr&o82LgD59#s+y?mS>sQn_wBt?
zYfxr)KCry%grENGR*R;O`}F2EMpIMHeTHAEh;vCx#?7kDW1f!8y85c1RV;lycJ7=_
zT3X67f6~cmJ{gKMw1v+lOTrrpnCXdvy!FC`aU2LMU{{aRmda=)?^?wJ{0nE1kI_z)
zk2Yg9LO6jDi4ClrTyi!v!Np^zYwkOnHbX@hH+@5;ydKHHul9wOy1Jf<NyoW%>E^D5
z#K|I2|2_b<`7Y)k2+l>1lQ!+CoHP?6j(lL8wfP4khT8`?jI&}dU+R%Javfav8Jod+
zy_Sdfb4dC<Dkr#SgRTipETF*H2kL?nE<!(QhEGws!|=2eXx|>|;AF64Y3O`jc^(}J
z3v_EJb<sL5*l(zuT(J}S5=TGj0)l4@rA^h#Wc9n^4%bggVHNr8J^szjCpZnGg0Vj;
z*}F$s8<+$-F${L98-W;7in|YX;g`xvUNVWr3t|K5r^5)`0xGv384@>^61(xhgiI(9
zNv+lu%jwwmAa)^ssd6K{-Gl3zo53_#&h|`=AcOtII5o)RWSM=~%)~_F2EL+0)Rf}E
zNwvl>lj6sfm+7;|42LSG#)Y!Z0$=6!QQ1Pk(HJG1pqqwPHQ=T;;`5*|c&K4Nt<ANC
zX1_g}w$vu7>mfPLNFQ*eybHB1)|TC1Z5PIDNHiz;^1CIQRFlxksn6~yCHLp<jty&b
zKiC)o_vnbGmD-e)3H9fl!g3A@^*lrwVAqgu0c2zNEb`J9u#k<oiN}nRiER3{-~l1&
zm5&g$Jid}K%rN(8NFxl!xQg&(uLU1~$})Up$fGaT=!g*1R((p`Tzg1dU>Ti!Cq@|{
z!LDNf>-PP%k1(3+s;gG56^1Rd+JfL@*ADjCuo=yO%7_5G2MC_lY``I3lDy5XnLTIi
zF{PV6+&0|%I3M?2%+x^{m0lN;dCg5n0^)(|V&?+rx9u+k1$RXON3W>GWE#J^dvoO!
zQbbPf$N~vaB#LWtw#@NaZp$4b*Jy=*KV_eH@tD!>536E$GG)z7YX44IhhR)LW7z+U
zl{)5bVN=DTatVvdUT`)+>rWi#A)CM;o6p)9ZU7=H16i~X7?<Z8KeMcgFstM%Wcj#|
zOr5k#vpFrnmWi`S^<uvKp@QLgxpGSO1i-;H7Iu2Z`V>?4a_^XF6{t5GIeL3C7%0;K
z%zj$!Cf*TF1dx8J{4r+M+KwBlOS#NUcTOB{u2Ij*A}z`sIw&R+*66~pz0<vJ$d{?I
zC=ce^xDRvrxU{sho4I(pYZerwEsV3W8n5+LOE#=cJ-=Gp*B0B#vg`-%rDb%$t35xP
z)_@}{oREn&)B#~e#|sXFa9T#4IRH`_PJVaLhcDVEtZ!B;0Xoo*GU^~cIi_qSu{SyO
z7xc|Ug$=d;F#v+FOC;#%wiz>2=;cGB>_U{8@cqz%fu`Dk8{7<Gk1i=uIS=OSlnD1=
z;{SuLHb!|dkx)MZL<7Z3&-?jdk|}RCNG@%@SJf2GNituCJU>7dT^86a4o*JL<shPL
ztp7-N+TaHKwjz$FXM;Hq6hd1Et17SHi}2kFHL`mwK}K0U&dO8^PuQgFjvKf5p+d0L
zX+tG@@3gEL;DuACeTO{1aC}_P4q@us!u|U^*cvO0tOegSL3o4V5E_!rgZmFDl1ecD
zKq%!%jPFMYt+KpkYFsln<%DOIa8Q)o?2TmZta`p+LnSBK;8NV$arZ1TRjJbBa%udg
z$D*!Wpgj&1--~%#avk*tZf<=SekF6fdy_xu$6^Exx;Pl`pceAaiYI3XeFf&&$+EX%
zv3|ED08#$ti!vzbmkDjQo7M@&{*FoYmn>ouUT8EjAfbYLICeBE3yHgD!74M~IZ#&g
zoW(X;gn@ykQ%3OK>0rG_m4pkM-L!Z$tMgvfOTsM-<qyg;rh3vEO{RwFbYOCi@Nk<b
ze%3S2>8~1<{51WCH%2Jh;E6VbP@}je3E=fkodEcUE`36l%7DVyC7}{Pqd7pqKSR(7
zn3=g`Vm=`1oI3XLn7C_P`bE9sdVMGFY1poFpWTp+Kp29A)CS|6_<~ao699pa!IUI<
zsPHZXN`hW*BWs$SqLYhNj^xY>I8C@jGx?4Qu!p!@@?0n_R02toRX4$_%_w-7<gtJH
zWSPC4$5L|mEzFd(=>zJtmi=mLO<ncW6rAG+#xWWAl!N>S67X@E4C7CqXR4NUA;sq}
z`r`&@*p=@?4J8jkoG0Nau_}O7POOY1yXMw8qr6-r*G+-KA(Rv#IENG~!#qpGDB6!3
z8A#VQ*s>DOAct|<N~S*lVNzNefm13IC7ds6z^W7urB4U(IB1v~2<qJO!Oc2sCRN1r
zGJ#pGpXxK;8~1^O5?`nW86aabllv=KE<)7Z3pimWpoYmIcJ7FZ{N6?)4j^9%PX|!V
zLZm8f@4J2AAjOA8W%pa%3ba8|7Cx)w-@)BI13dowCb(Sy8RZ?2%mmg>tv7NI{!z-$
za*bUQa^|jMN8x+LfoR$)qvk3`wah4gw&M4)>9gvV>@Ji&jP<hYMMKK&w6#~LJvXu3
zlQ<q5G%hP?JTmao(<Ao<lk)@d^hSmgIde9QNflu$Y^>>Yx?P7gS`Qr9kF#PxUD;^k
z&47~@aT6WS7J|%4<_Bes0EY-m+U=~jP#$z)vR)k(8C#=nyk=I5cC~Qp&|+(`NX@%w
zR_4v{rkprf-{j<wIcW=t0f;p1U(w5p17$AIuDsPot7J_t8rk3`XVH2rv*LhN!l^$U
z0>)fAr%!D<xf?sJjbIZdfrfB`zZ#V($#isbjJyrX^^**ApE07I`s`5vMZ~rkFG^Gf
zkP&{Blk0FAn>6AE+Up&IchC3(lN6R@`H*5Kz~1snpg|{@4WJh%<nPzSxB(H3FB>Ym
z??%2kTZ%U%u0gFMro`HEp2#I5^6|-^-N-bG{gaO5nkj(B^SPJ*L|qpYT`u{f`pg2V
z`kYcJ4hP28^`!cp(G)F4>nm!+;k7rqkoB@AL4!roScR&CjDByUPw=A^{sW<1PXYD}
zOGhLBoXl6pltJyVSjs*iPvJFMwLZ@HsxZy!sO=+%1z*NB(q14W9+K3QX(MWZ_D@o$
zlJbKMmV%OA#%w`wuQd%E3BklgQH%RNr(Oj7N-=sfWFgG9^@BfYS&mRK;^(?lHydw*
zkCB1zG}1|29X{7}C98e`gjtJwS9L>qKr?z`3T-)|sVpMr2UA@AnY*rUyWjr<sVmrO
z=DJjABAY%aduNnJ>cGRK1TA0M@qTlSQaLRlF$&-JbHY_~Dw`zhf~?5)@^ZDmPzjd-
ziuZ$@6JXgiP*g^@e|vx^awX|f7s^)Nt)AMH`<!DUo5G?Fo~Xpg0Afn+r=>O@jt%ys
z+0hY-7XX|T<j?_g5)zz+u+Zujl3DODEiG|rx~nAbqzd?9seH^!p#IEi;p@WK;$mc`
ztn6F%!i7Llur;;5fK?ImFzRzEo@jS{;5Kd+S155Q?%e@Y+5Sz;Y|g&x^~1_>eH(BJ
zvc(fnWFvJrXJBX=y!FsMb&L#@S~p>YWZhnH2ROjZd|vMPF86I}lJ&c=3+rO4X@qK~
z7*ltt7}!{W%j2^?ols7_13K>{jnd=+qy$W@w%l#bDQs`Rr8kvoAtGqo8%2Cda#W9?
zm2@rj&u-Lm=>Ut|45UIWGy;ZU>@b(gwv5$ahv-q;*w&hN#bQuU!$HeIUZ{R;NiE<r
z+k*$57r^iHd1K1Uyzc|J(iK;Vj$U&c7t%gBbFB5S^ygiu74FH_E7yUj&k~^&gNCi^
zNgS>MUlaRTv~`|30_VV1YEEGkFIxKkwb2skCTuxr9QJwn=$Op>Tzy3D=o^DD-2yn(
zbXs1<CRuBpTBFm>mT(51u`OTk9b_d}0QGnTIHOtk4Y&?%5X}UrMefqj7*&ii@n@Vi
zRV}>Hg<o|6e(4h`4^!j1@1D^uujZ4RGP)8l=#`TU%l{a53K~eBSqt^wcpW7K-#%^*
z)VeQ*@f9%fHrA2Q1G1+l0&cUIQ~jXbFt9$GfjiIw<B=SVdI2n?L0`>$7vrRS!enft
z*DowSF!~Tj{AT7dGqNx}z}4$X7v8)WH|qqa<^A1)nly=VQ3=Qh`R#7)89ET5Lw^zv
z$fl!X#|;VdIpO`>gEw)|rTi+s3M??*<1<@}Trlipa&Yn%k9!L(EHb`qY}lcaHbL26
z$^Jfbf4==n)5tDS*Mo8ydei#_Zpqn)p^VrtHsklT@q<3V3pU3rAEh7)B8L0jLRwg(
zHS=LnU7b*_{P^<|m1uIv`~;ZKe(;HwBc@kBu9=JP1+$qjl3*3B0(|f_(N^Qr8rCM^
z(ts@AOfj0y8y^^7!jjc7wDjZ{r9Zt#H3V#AzdLT|DVvyz)3WxXiEVV6Z9D@&M+mK-
zyBRgUeK=H`OKa~cTWys06q2Yxlt2bvb=iRltuBuOD^-5kPvc*nuctfqQvxe_9x16K
z@Ugi5u!aPwyga5%S%z8y0CngRA8rz}x&3&ZLYQSVQYf!v%d)B)OhdsY<)jV_4ktYM
z(Ddd&6R7kuV|j9W+jcdd6h^xL_FC|9zUJ+C$r%heCB?(Tv-lPIm8NHEQYvn&Y{8GV
z4EkDFr+cI6c7F$U_vWgGM)kc{y85|Zm?5(5@8vg%KB}b4HOihCMC3^=_>Ns_TQ&=$
zD(?7)tA>k3ZGdY&woh8Je#VhgE0t}YN?CsK@llGKEXDW13oow*nW5y(R?7K0=6K&j
zTRP@7N9fA!t1p3PquW=n`m59xyKnQGe<0+UpA6MP#Bq(q+xVNI(V<EUY<T?#-!I`6
zMh82E+3Z)^6XGJm4mEslOxNG^LMDXwBPK=fEK9Rzg#HAk{qOCcE7DM6p)rJ<YVifh
z`>~t_t-uW^m#kQkuX#N>cXd(ne`9$4>zsc78e*U2{VPSj1maYM;+!8>=5R&T#QDax
z+PZP*zMP?0*^BRfG%lHc3%$Q@`-d5eHDhsZ!hoOCe*|d}bs?I?RTT3S!b4s}k_eCt
zvN+e>ko^xtcQSMI#&5NrK~5Fc=HK!M$2Ai_MCon^9uxNT6}vW$f?Nl*<u1JI$}r7C
zt8!L#?}<YtFLYiUj!6tuingcqd>ifhz3JoWzdd!}3wwhb`TDCzyJ=*}m(XJ~0|Hi6
z)~akjwQSnGvTwN%vEu7tGOk2n%f#bOxV4NN_Kk1ZcJ=%s@=}%F3h~;rV}zkI<czH*
z)Ts$WzHMlX%pm{bf4r`G=ysdL)v0Kt6B#ZY%oYCy^*=Z&yx`?<9u}6CHzMubOy4;$
zUqx^w;caqWAJV09Ol|mvHWr4ya##Vb9Fe#VL%r8;C0~hr^$*1J@}m@5SoWA1uS<0~
zEf_7u-mN|{b>QN87H6+v|4qauMKG@ve1XU5aLV8V7x7g=+ei6Y*dMO33;u99<Hs)S
z5?S@1=P!LC_L_-4>0fuFN6@Mu*_|%<vfz*L`hTm(RNsnIQk*_8uMOT+;n)d(c{!<l
z#$2&KA&pl#Mai`S+J=wm$59V>_?tx2(#WFJDgDYD3tu<~R<`e^3#yV=!|kK&>2a2|
zJ~bz;8EtCkF%Lgs*?Sk^6>a{>`7WjQbN_wU$9d>G$AkN-re}gTS0&>HgsV$W;>krL
z=HWvll!LPDPVA~<rE#nyh9@JF!eQtu^BDOIf=Jv&J#)gzYOa8Ly;(=(upd5ShftY+
zxq+a=8{M9>Xy*EEoM!%;NI}P>dDKKe8TzoxQl<hW+vAl~b_ZxQ;u~#x(8!!t1%2S7
z5OLaOKC$F4k_hv#{=yPZ%NE|4m**=$Xlc83|6Sd60i~=LTlN?1rbeYwn1^$_RJ^~f
zS*PRHQM?b}|NcbrO-qYv^^c54F|~4VEr4M=Hoi%h_F__{QRhBmDLzUv=i+)jU?fUB
zmoo0fUg$f$M)o>gfbF1JeQodijOF(Z$=!F@*?FYM(oGPF)AP&w(ou}>6=m#s*2@zb
z;D9DzRRI@ph$d>o_gm6GX+g%;uJnlBUfvP};Z>bAb*<B-++WcyJH0U`E<zGLN%1hy
zV%PVIss;`V#2E-&$1Qz3WZv1}Ba6F5HnS1cp;_3!jiqR!ne?#4Qnf_Yd+bS6YaR-Q
znL2)4+|mJRVpj#7cBWa$KE9?peI_i#wI@@DmAZeekE-{t7#hq`L4pq@26*+^CTEsz
z+&Cb3kNP(V(*0Rex<EmcWlF9$FKxcZ<inL3^%ds(_dQQ%j?3Oi{jiASFb10{ArGxZ
zo@pZ{%Blz3UVD))A)TVqQr{dxC`KqgU|j)5AYH()L)w4q*}XaK*Br@~B&9}6bP$JN
zh+in%7XBT)8>ikn#xA*N3AOXXUVBrCIXLa7?hXD(#y>{TtP_AI3a48JZ*TJolBHXK
ztCJV365G??(_5l&L76GvZ}|vYXUNWzc06;%AjoO;iEW<S9(Tdl5%;XQ7Re{Sv952f
zQ3SV>B$hrVCqqs<7hx)61rZaqI$Xk~AGH?vnDPEKC^qvi;8wq*KcoZ#SKwswkMcbP
zF3;j2IfTBx=ILbp$__H^6G##IV)Te@47~V<mu|&O!|5OoVH^8<rf65B;B?~X(2L@Z
zA^ykEc@t|SoA^D(o?c&)ci)tW>oP1{j7Mg;iJFUFr=_LENEcE2nB(GR&;m5^{{^{M
zRyEtxKy9><54v0R$iQvbpEr+(PHj{g?(d7C=auF&se)|N+GN9_F*W&-DBySVh-ul-
z7xQnrnoU{BU{MUxja`jjCFm)G_Q5`@!7n?eQgXV~5l1!`@3wzWXasM02p@kj@I|PF
z#L%;MRy}9n(h@;W-1Rq!1|~QhiS^>^tL$N=8Ysj~^bP}x8(!x=0{Mlxl<#cDgedz(
z4OE2_9kQ8~GONDR8}gSv-b6;H<=9sKWFwc+>=_<eW$7#JcblJ=RcUV?kqcP?SnQ8W
z5<}AOCW=I&&c#R$kvV8A;+VMTS)b$PLlKc>&`_nfYgtyZd6HU!<;BAXGh*meuHo{)
zEJTRiZY0Kf8~M<eCkKCZ`0xU{S2d|j`<Hpr^vtH&tGQX0!uHYe&Wsp$PYsT|dgsZt
z6f#ZpZZVSmYj#8Y#?Cprdog2RXy9>=vWK`a+TtCc7G#GK2N6bX*$Dk+L`&p=i8ep}
zSdDWbI)V>F@T=wTL*p+x0uH0e!xMlyn{Mu2YoC-Q7%~1IW4~^XwF{xIm)Y>e*vr+r
z9sSk26SfbP-I|BE6}1eS@3pINP$^vxWAB@?j%-TOLGH;%%x#ndz3GP@Ad^%Z<J3dV
zEG~2}n4BGRym|5qIN4GaXFP_>$mMyHpq89Cg%|1KR|Gt}I`e_8`_HXgPn>9WNW$Dd
zkPR8asL;CMgP~_h;uh(mXOvZ_T1jiBqs1CFYp<ZAi8ls_>xJveycrQGd3rw}OPL|W
z{``<AIPs*D^ozImo5e^kUx)1<UTo|@c@c<qae#OIM|}fjG`!^?3405G;5KGz@TRwu
zw=d-_;2AFdq1t-!$RWau38mVeQkdaZ)O?nZI#+maTB%88?HmMc2+KB6P){CC4M|#z
zDAKupDDK$Ng*+4*#qS1`dUM4kz&`-Bu9v`eg9>(=C*62_ca}oFkm6>XXfEJp1HsNy
zP&=mSIBL^A#BIiuu$SdF2+-I^AAw9PyXEurlCi>N33|Pe`AZ6fK`uKd7LMA;6|%b^
z+DTc5+u^b&Wfe^XKLz*zg8N&OX9l`j$le$vLEjkTo`=LeFN;U@^<f`n1E>k}BpVv2
zg5QlVv%R9x0g0e7HeDR*QUX4Tgy>Znn_>i6fKu%J4*>mNRBt*HMqHQS_BmMIXJq_u
zS7%02t-btmK-N@Kwt2)TMbc0o6%q-1Pwls92QvH#v^q4%0`JWUbo8mPMx!_-eHlh5
z`YNJYud20sv^e9`uLPB<H-1om;H)zcH!VKNPL-vcv`}pys!(<un*n^Y_t_!_<sJq3
zl#utv+uFIXTlx7VY%`+@0fgHCN<#Y$D`56VzRs%mnpF=-Q9#@EaXj|LxaWW=(N5am
zSxArFXT;aFrs2>rQ4@u3WU_9H2iV2p(?sh4g887M`32OgjGu+Mj|1@W^h$mA7~rP@
z2WrEr4Hp2Ql5t;-pA~|G!&;0CC4OHPH<ywMPzx6MFiQM+FF&+X<So5f?OhER_IJKs
zxb4bmEf>1pC{e%O3N8E-y&2H&=n?N+lo#?0c*R@6vHqLT`4qy4@qH_N!D6Ocr*<0p
z-WytMfW*8g*tkR51pndu2ZJ~R@C1pqW$!K<-61UDlmJ2wF1yOiEn%xW#uS~VtQ2+E
zg4TlhXvf4*^W2oMiuc^8D}6VOz02maD70atM)ksVzkHZ)Slw}9{4n+Ysv&if{qN@I
z)nloG8)8-A{i`~?*)UKRgF#Hv0g#XNh~M%FOfT=@=EIG7Irv1WkhYS*jRWWuwosK|
zz9nOEA7+LG*GsN1^czL!+gsr+?|hVMAlmjmOF7K&vV!X+pHmwF1Z=im$hWYE+5fCm
z$Yp@M-WH|V0)I0UVYhVzrA-A0X2Fbn`q*a}4D-xAoL90s>gjz?rUA^m<%8XHew=Km
z=L-yW(<s;kuV3ARKLIQp8fSr`z2}YBaIc2q_ql!c-K3sSEbIewSpE&oZV70!|ABgd
zp9N;+ynn5={XU?WsjMqtL_BKv=CEZAcjG(NE~#L_G6R72=-I&Of#LN=O}^UQN-h@~
zm_{?gT^AnTK*M6OUlE4#Ie6Rp79URFa6pH_Q!loAt*U)K&u#?v;*rnnlmeW(n`Sgr
zpl*pnjWFkKO!EsZAIPb!Wo!y8(%TVI>!2{??i9e_O}7qzIxGlyY{cT~7k8E8!eOZp
zn6qE;!7X$;3U&;PTh{;UvDADDv=Pv~o0wwWG$7!SU(?Ef5Kqq>!w2+}Bbayb8i){1
z#C0cK^gSDDA^VgLEurlL2GX=($B^)A2fjP49r{UBSV^OeTuqlWl*J=kjFJrO(X{hU
z>l0k~4}^JSzkiYn&iMZB+LEUfS3CQ|Wt0bIfLW^t$mYwq0hsk}pAq7a0)hM;7_nLy
z0*!h|SglcfWkC0<%wKbZT+GneqvlB}S=GGKd3$H>0oh=fS%z__=!Z*+_hv6(UP%E)
zK*x{pzwB^u+1c?6j5traMufPDrqnWz@6J5x$ckK9O}MB3?mtg?JPkYVt1-OpGC~6&
z_(OQN@AmCOf{&pVhJC<vVJHQ8t2cVBYA@R%)9nMsX8&Wg^dUYeM%W_iTxvw3)A`mH
z(9mg+$QhWR&HxN$n`TWNsryjZ$x0@#1oV(q6+A3?Ofaj0M=H+(QCuiaV1R<p+dGCw
zkZjI36~Nr8yIQw<?ODTlBA}z8oKv9Rv(>nP3UcHsZxnNu0GQuwbDYr1ej*Uj#`=vW
zV?eX1gMT;Q!g{4qA5+3!Dx#hi6MOwS1ax3hfVcz8K5dkg{r|7mv#M=$4rpAA9Xl;W
zV;~S781*sa?9jsmxe*Uk_A&K_&0%wR@UR_oFt`P!K=FPuD!;5xnZjQ>nc764&E0eM
zKNeKw?N9@s6*RPHAdmvEWetFD`Ts}l1MXRyj0@358?e6hoo+_>%Lg4ZFCdR{<Ex)F
zfTuhXiiZckD&@S!YhBhZX7JJ-ALbdxVf&c=s$5I4llYQafZJ4?e0y!2gI56DE3B2J
z-v>MxI;q(rzO$n&(q*UCXnvY6IiY;Y9{n<d(}LT14{gi3kwBX#jEZ^zIed3k9dOwb
zBNVlW?wFsE4aL@P=?yt$Wo6xrsd{c&vw4B3f|4CHa~b#KL&7j$00ri>i_LF>V=f*X
z6nwh+Fs|#mBZBICX3Nf!`^hmCsTYG2xFLuo2=qN?yF*utg^$ypl>GyV`MYU<I@$1K
zx_!s)5>|?HYUzBQ=N|Y^kP;t0Wur1_d0i3}5yZ;|yN@7Cn=xvYLc8n_ov1!RTfRK-
zF4;sm)_T~@!rE<;`{4lh6Cdf;L4h8i&jrS&p(Df(9<{Y%d1sviMAno+2<pE!_OZrF
zGp-QNz0g9wgOq2#=oGv=+(J3^wz99a#QE;I!4-!xAduCEAC9MLna#(UCtZdb77s=c
z`|moeH<A^ysD=Dy$CLShw+Em{s#-S;epl@0pJG8k!!GXXK8U>ULC2Atb|6W5NByk&
z9rM;*oE}e@CpKQI7dhi<WxV{-2c2+|cZ^BFM(>`eVHa57*N@RJh@E4?mRb=z91*LX
zlHJ*|q)S2?)~V@dH)fU5MeLZ9Y|m}QSae{5n*-Dob#p_Ye57Q#G`|>QjwhoRafavr
zP85!Sy=pgN=pzgqMHuBq)`rl&-iQZQ1Fzs~VqBFp+K07Od7q~`gpvw^kGMSM>3lK5
zm6z`(X6%~y{1|JmhEXv6LU=(3_EU_9miT?ld}JAa25e}Jn3{ppY31E`rEZI@c$nnx
z2bcTh!G?si{catZy6`p^Ov63aoD*(gJg745Tm3t*_G5~<Z-tLwaZ#q1D<-@*^A6Kr
zL;K7|HA`Z8(e`g5tcNM~SITW>t!bkm9S{9~{cqD%mIp(hvWoQ_QD)vP>K5mYoq|sT
zle$|?L9eV)?9hBbo~f%YOq+W`pDo`CmVlBTr&1_+_CK$<4o~X4>ta}VgJXEKMwL|+
z7Kg?SjVlwEB&0qa4)``d!C%xyz5ji=TahtiMc-gf`t@+ppk^Tu+GkfK<*~5alkOj-
z9KBNAD+z`&u|Bfar6L2jN(BWLW6A_1`jiP;wi1Zu5LR+3&2IT<%v{C}GSaCP)X4x#
zSR($gbqkfH&rp%cdF|!NWu$b!OfN2phZ)$bQ^;=ESf3IFTShw=lFs!EXcmh`ZG?VG
zdP_8}agDGkz`J!j^|v}q5izL@WqaY=sYO`n%fzLdUy#!aMEh}MTmfTOCOFeK%^nW|
zV|mENN5M&VQ!{0+2`?8tV;{eg%?hGUL2IVxz#4NYMvbCdt-e~PkelP<X96A^JlMl7
z|IcnOb39In3U_vNYLkCinm{v)qa)}0i;rm6E*Gx4bYTODDeGK~LV9^`esHBIcDRxf
zw3XPSO&F2f^<u8I53vtfth6p@bxWE$j=Y~7Kj|b35RH+%WLT3@Wm#*ebwSlw2OD|5
zH+QUsR{wEhF~^Ev<?`iQetAcQnqt>i<FZ?=PQi8hpAk6nejsD|>`CflG7oL4?8%lL
zN_#ui%vv=L<UZgs8M~clsYH>y|9CgDK42B&NW0dFCKvocAVJ)2??PV|i<ooYub`qP
zIfQyAVxyb5h1!r|!|$q|(m`xwE!AS|rw3+K4z7D@Nr<yt)N=b7<J>!tLJW3z3?D>F
zzi3l0`EXBg3WyU$iu8KE<IBKJS7*RF>0DB3>PriH&1VA{x<){n3T~iaV~Gvd#@(s3
z?o-#A$>=gT?4g@Yk?@UVpL-A=HCau9{!u9=OTp@{`2&p8r=~#8k4`w)by6?UWSq{&
zp42XCK#BL47h@ndQ0O0{>vrix6U%y)i3;hoY;<ksz&YV7iS1mb`|c^=q=LZEjv8k~
zDpbsAEPe$(!GCAwkWQ$VOghzepB|@Yg6)l+=C3pw@4S7Ta3=a%hBzP+O{cfoR1ff|
z+<h2wBz;v;w@pRUGj7QP<FA53%KEqCPdgE_iAa$PSSGyyW`r$c#{_Z+dj-wr&H6N=
z*Fbq6Q?FW#^SV&k2QfZHufKN|L}74U)u(#Fjz+?K_KSNJwWmdegoQ2o^|m=6MX`fl
z70m3enU)Gmnnz3r(_VLCe{%N|iSj}o<?Z;_yn9_ew2zh+E(u3!j`psfUyG0rsplpZ
z$FW`?qC3{mTy61$U18?U(+YKpX&r(hj0|f8bD=kHuENGd(s@mhO|8a>jw@w-4s{ju
zmwW3$ENNW^h76XsLs0bK3Py-V)S{@3?uC!1?Nw{EjHN{{cb6@l94_iO9k&Eyot%A*
zd8Xo(exhih@9v@oP&h6-_|Vf%^h&1mhHcJ-MwJus`Ae7Ja+-CEAOl>OTZjiyy|?2P
z!nlobtfpX=E5&cBN3~?qT6wuDh=*Ia8)ngf>O-F%qLup}KR#KXGFl=-Y1(%|CzOCc
zY(JxM<q5s*1F`{uI)e{Ap%w*2ll1z6FSeH?<JuS6MS<mH0pncJ970+P`rvTt)P_^n
zexse|D7EK!v}Q}=AhC~wW0KCvt|;L4-?<0%fEo!KL%#6rVzBG?KXGjN1vaQrsF#&z
zv#jt^$v9*BQ%*_0ucKoZ?bh)_rTAFRf-N@#<HQq*x}-wqtMIfW7d;nNq*{Fbg;YM$
zA!>CUtC61fiyRsdCIDRa^@&eID>278nmLth>xG72j&zSu>_`;A$y$*JB%&aQ!U4^s
z?m|P!OPp<^`74D+%exiPUibH!9`#LEM8YR9Rh2#^J2)oD@vVNxfTR|b2k12&OTz2%
z_=Gj>f(X$li)w5A6|@WR>Lp;mWw_S$8ZSKT0-%aqE^h{J(@y(M^%raOnZ8S<3f-Pm
zoRvj+J(3amz7KT@FfzBiP&5$R6Db*eVdQ{hH>sF0+E3=EN+MH7`Dcp(Ok|qL^3HyS
z4v|Mf0V!l#*$x&ZiIPHU)J^LT4VYFsJCs6Guqy472?_*0K>Vhe!)eH^%&wrjN&edV
z>8^I-Xs;`}pk3bVbFjOP>^3>o&khN1sQfzb&_aHyQ=SCCaYCumHU;m-s`$KZjlBVp
z&j9FG1p~RE*uMu7IIYWcHv^;}9Pi&;nkb`Pp19zVqw{Lefo7s-^ZR-O==5fXItE@u
zySxtr@-$iuPhx_Sx(*I?BlR_(O$R`lN+{tpspuT~GXD%+LbCOs8ZV}${yG1HHSft-
ziix`!GEHl}0T+m=C`qf2AK@Z2CU~=9jY8+sg?%cZIO@TW8@povcA@id8NJYS-_)0(
zoq|Xgx=NP3F^<ip+~A8|_#dxvxt)YT=$vsa>*2;8QrtG2lfdNcJ+9C&=oQ;bGR<gt
zVeCYC#XU@03Q9(A8U_Qw5x883s=(FrZDOV#%1tq9w!U7%LhO>-bCubw1PNm=f|m5b
z-BK<D`R~kX{(F%;b}W`vGj{<r6sNa6{c}xptRyC0@>Z)DeboxI2JGkH>Pc~TTIxH1
zt<i}C(e)ZZUg0FnHfd1<I>0qIt32RiMH(V$Sb)Vp0sT>cGG2Nw`$`3!D<b_^l62I4
z^_tk*iBbZa@uTe}xl)^cW1GmKC#lq_D{TREK%!vm9tKzYp`pZHY$34`V_nblrk~Be
z7EJHvv!L}8!3P)P99sg&oi6tZ=RK3DlcWqxSxGUat?bRRYwf`<=O);zMO=?G!0zP2
z(~<Ps^Y|Z)`cIWLh7jGGOCA-ZZyi}?RFb^m)$MP;$nb<@3HDWEar3q4x>zw$`@l9w
zM{EE-eMKQ^Ye}E&-seW`61XQ5Zzd8^=%?;g#FCFQosmJ2uT%M_O~>K~#)iu1Xp3{5
zf`re)_Y#j}El{`<nz`4Y<XXIX&~%W<d0`pMR)vw1(hBe*Be~`BPRF|x=+;p(BH_7h
z#l+mJcK%lMm3bYHcro4!RL!H49!8uDDXyQG92z$riv!}~dex79`3M_^hbQ?=|CZ4k
zaH{Jb<57q#eVwSW*VU<PR3qPDWpgEmS4aY>qf3BDZR=u|x`z0jL$7;K4D^6ZdATxF
zgDi4aOM3w$mmZJC1gntTVhJMbtMZVJtmuByl@zX&T^9q198zn&4vagQgVbt#GUTl9
zY|5hFyQIC#dO>O#HuhjAhycXF2h--kQ6+&mTLG2h0Vobs{n2Q5DibsdpndwYD>`=Q
zqwA67y+^i&W><yF*1KEw?TB!)GouTeydmzEJ@V;GJug#rhGa-oKt##@DmUyIq-f8)
z`_wz5or|`qM#RDquzw(@zKW?RV{7RBL_~^OHkj$Y%q=vVuIgV7*qq&6iQ;!7*h;+-
z-=tKToWJiNRxwmRTc;doH96qNd%GqksV>Be#XVvaPq1Fa?23*XQ1Z$|UZw7Utu$KK
zgTm<2&$1XC3)408+$oO3SaA*G7n6WO$d$}tKrgR9B0kH=R}9uLAmaGTj}pZ~pzEAO
z<!*?y^Aw&pfQ?4e>F4fA2Bp+puS-WNSf{|Hxy*NUsyYKp&(;I_L_T|IQB{1~vgEaT
zaa=<RdqMm%0*w7O8;{jO7uEwy9$5}FbJENh9szW^ab((^gN{XMsgr{Y?YWsg@J40S
zjMk^?ptvr!FzuK~YhYbia&rRab@<&2;gE$*F5{S<>^KSE($YkF>2b+$AtH2EsUyHL
zFl+8}z3(^s2fuw=$%e^PiY1uupwv#E`gd%1+6^#a4v$O=ta69q>7e}CD&#D%{6G^0
z+^goIfPpf4GwxYN{W=XM97*I_%kJ3bSlINf-7KQc>n#%!=`Q^kC*wYv@@sW-)9Bym
zpsnEd=(-(sA2n^kvE2(0K&`e3QlOP>+BtuEFnw#}DE5>2#|e#+@cEmuI<4pm`ntjo
zA7}nbttW1Xi0|kAuzOa88eSX^>-)T1XPq!T-jnbN5MQ5`lJ@#w^n_1+H1{s4$Jo!w
z%tyz1ussBwR<Zazhr`(`D|c`UIX(AuMog}R_Ga`RDepY~i4orxlK#o)r3i&x(r#zB
z)f%)*@qDi<iw*EbwCno|cT9L(Oweehva#Z(c(0OLD;v={QAyY0^DGZ96p}Fiu=4)y
zRJ^1m8&sA|zh!X^PVC2Qa<efVFf9eT9*xD2z$x!1p7yO3QAxJXmIFP5#kzjig^B6n
zq>kPJnq(gR)7wW$3(r(_4Rhy~7~MqsQjv9L<h{9bfZ{-$8dhX?aq?5eGt5z50<cd)
z+TTt7Chc{zF)6d5J;d#c<6dZq4JI2;jLCSkRbe-PIUw4YM?0ZoAZLxeUEgY%fa`7-
zxdUfS11v@2>h^vfX`Q}$_Ne(LBJd!NB2jDFgEANTl6p?LTgo)+=XY_)Sem&E>&1v$
z3?}`&)+?5~fVTe)IC!V(qa9psr?FMQL8H^(Z_myxjEbq$;Y!}=UwY)&bYT$5l4tDN
zua9*_N5*CY1wjqF$~^}AdmFiJ`w2<Yl~}>;3kR~Z1(Y_!y>A&7R!Ih4j7h04zT^p&
zgXZ_{ycSERhHw?)-6qw|dR!VlN4FcALR>K8v9gVQ(vOl#(n7NM+cXa6BsYFMc78C7
z<u<rebk5IEAZ1}I-i$g!Q5Ahhw-&bAR7&e*E{5vvCW@bfQ%T4Cx7@%tl%9VeThi4V
z+9JwQoo;R!IsEe*$m&$w|D2ch%^z_Cw7(3Q@nI^uuWCukQUzqui7e{9e0#F!MZ{Q4
zPp}8Gf4MKkp+g{e)|{}pt9z=4vFMxEsB{eJVbJC|Ty5uf1f9X~rXaH8)PHt+{x_Ed
zyiNQ|%Q*-s8(Nj&JisI}WwZx>3Qq1mN57Yuzwdv8P>Wh_FC50}B)2YNezn{tu*s`)
z^?vhj^Uv&Ul1!_FU<Fw3w=#Dq?Gs%RUeQhJAi*`xx2=K$twIFN7ro2^UsYDs!F_^1
zbqZ2NkzRM_>4m?ky$<foIel>>>~I+!^C*?l2WKZ|7B|Vw;VTN4fOTEUmyYiJ<(%aC
z>hTPWeUxeG^gj1hBCylXM@GA0n<bWyXn%9FCZp`#mdp?bHH1d6r-;KYIXa<L3-QUe
z#k$ma4kN)Fo}`vl7eRUz12cw~6#`e@hn0D}&SmdOCo|nwHTe7RcM9$-xhV6-VoxuL
zv<g`H1IhlB`>Gcg1Ct+s%L8)xjLFY<sSph=ihyMRn}8rkU|z`L{;OfLaU;unVUM&h
zsC;QBl0P~m;ZK;3b7kG9K@N>$L1v02lg5NmYgO|m<B0>adNZdb^Y!LdO#`W%I&+^v
z1Wxh-Q#F~_y^gZ|<nHyGswTNR=9)WETc&^f8~Lkz)KDVMcs5cf#woQXM{dueXv5~m
z7tUx?CV1-u?1t2Y{%0fiX*V|c2ltk9J(CjDlMRd4f92N`-#(eNyaQ29c)5SQZVrFt
zDpH%c_z_Hjk;BkOz}RF|*r|)y1LHFZ!V-$>`V-EMnrWG4>E?rkH)e!tA$s4=dn^T0
zJJ00h;D8h5#o@q$cC4A`=cW-;`~7IuI+?@yag7G{mv?}-KZn47#)C3cqPt*c!E*9W
zCHf(hOBV~nE?@d|z>MZNwf|v~_1W*vBL9Ruy9rK<Lmdp`uwTn;4l-l?rew75-~$>x
zVc~ObIXhLSj&om?XD9af2Xf2J+H+1snKT{HsC69H{DyKiUWXIBGj9UlqnZf-f@*Qx
z;yAsZNHhj$2~rl9Od72SVQ@TAMg^5*+&$(w%=!Hl0xDg%=<iwG#h^Ru?9<Er;VGDU
zysX*$ebbO!3b1yCgTKI#$jby}caU%*LWYI7d(5T6(y>SG<)jb#r9ctM6QZuhUIpk^
zeYrGVH!+|Dv4FUaAZQ<Oo3f+ZYD0SyMJ>>p#duy-z{SRQZYh)XC!J}+>c*<PTO3wi
zkZO8j1OD?DRyAP(kMi)iVQz*b@@{Rr!W7OS)*JINmlk+_l%+vW#<lNJ482?JfZSEN
z`2_!p0%U-gW=d<ck}JsAmTVIjWzCcN5w?4H!y+tE9>yV+JXorU5DO<i)z7Vm6k*RF
z?~y%UguOW-ZMxlP*0WkON(URH!ZScE3l#)WhD-qa`PV&It_$(3k%$E?r$9}CvO6Zd
z5GM4sai>3IyR=E*jePJe6)QSNkDb)cLfpvfSUct*yF(n0_dob%+I8h{|9uE_U77TU
zYlzKht}$hIHo}pgqq>?o3^QE<^XZA678q$?J_a0WPHJ3&?CzSLp_iMW^xbbOVq*P!
zro<mJb}$kV^h_x@jzR^Asw`wuqgwEIvy7GXBJr`F*x!^r-Tg~q^S3eEh^&0~IV8lE
z?-VB?C!PA%>zIDBsKeFmGo+kl)td7%wI}`PlIYTK9Xl2Sd$4R2bC8}IQQQkLa~kuL
zVBQ&OKvK>phS)N1+Cf5dDQYyUER8I|2a74|ggx&Txz*H2C{}{4#MyI@;Df=UdlNNN
zzbTXuEB;?Dv1D%AE;$$F0r~bB9aPLl_128*@eu2nnuB#3mwNns!vhmkYj#dyS9gFT
zBzml>E?tdCoq~j0ZKH=cD0*6w3>j8Q;W3K(z7tcwO+uap*T<w^!Ou=Y*7TH4ylpK6
z<;^1?xv>@ZCVhGYu5PHakhhV=AN6YB<w2rOOnxrUDb!OZ3P`6`NTWA5Gbh?07J>B~
z1!e^qF)cf_c<UMe$|`=O@RUjE$z$VJ?kbqLIMe%#QU_72ihgIYv%I{%ejdecwCQG+
zp7W9XMWWdMWLn5S*TLF0N6h$F2r~A3y_gl0l{(eB=*fxd+Wv9~8F{ANg$~2}q2EhR
znbt6?m5QEI#_AP~06enUeW0vEHEZ7w@L=Yo;E6V4k9Q5Z_wIBLq2b5HIk8NGq9u7z
zy@*cKH_oVRqsL{NMUjPlwlL2{HZhG_;$D5ytU1!<-V<_DdI%^{=Iji8&wOs;EQzR`
z6j1w@@LqNM(+6{!s?I>#^PyzKNCl-u-?*a7Una)!dAvJ|;WeYgoHVrS_mDp-bJE0C
z$!Bti^B4%GtuSpH{jD}p4GDi+;&H6+?i2_-e*RK8^fI=N#K1x<LUZd9Q707ojFkzY
ztwqTq=2)zuac#}X9esBl3=J7>EJx<aBzb?~oD*!l2|A>BcG53ait)&>uy*u8xg#c}
zf(%GKB9dXT!%sqit~X0_a5#DH?&#FJMS-h(s=MFbSBXpT^Uok;3?G5(TK$C`QZ)=A
zA9_;YsbZdCV5DkDl$Bcu<{!wnwe(HBkhjhemDEr5J+&?AaVC}0w0eDm2ybxRx9AE>
zxJ^S%<L)zexoNPn4zE{CLn3EBkFk<uWp**OC5@&@3PFi4#vTJPXU-n<>8e47oF-&g
z;R4-VZws?1y^FSW{U*$m_lRG*Lw=5sGe>K~4e5l&)%<hxNq_DZq22nGhn*cxuIY8@
zgS@Yyr-hVsO{-cL;hs^-PRC0~{iZc_b!;r<j*0W6Pqb6h7YMTcV7mp;h2nh&0`?&M
z%8;B!G`Zd}5xb^Me&T(vQ|tqUMvVV};`r_xTcO*)v@9}Xag<??Tv9FJb$nR#rm~3k
zUOxwMT8WJK1`7&4ns&@oj-IbuxL?cr?obKuPV`u8MYyW2dQ@-X<KusSR|;kd;&Z&d
zdv0R%3nZIc6zDDqXM*7Vnc&a$D8lN0Ab_-P7)IqB73Hpi#Gb3TCv8j6VIY63;Xf;9
zZ|~`zGVQXd2uE8T=TS0D_8jXO71O!;f<Z{Ov_9%Qr*Ph&5oJIsvt<j_BA4)1NMeD9
zqvB0Bx_t4ElqV+^tmh0MczxP6)uSb%eSy_zv8I)~#z#^z3^^X`_hhmv#+e$Xno&Ri
z?IUlZx7F3kS{Gp}vvb0O!&H#t3*ksIN+#74CL!eLaE!hbO?z%~Mx;RUS=HPVK+)Zu
zoblJgPJLTi%8FJ7)35gyv@Uv#Q1}cG=_3>~3?#f#9+T>#Nf>BdRCl~BFen-wA?RJ?
z{j0=FM(nd5Q13I{&Y_?4JI9rx?`7<)iYRrwDXZHR&NU=wxScsDaQjRbJcfk3otUlJ
zoDx2Y@oZe(?l)5ZTlH<l4CJ7HWp2yX(Nm_A*?neMThSn-y=(QX&ouK%qt&{bBJ8w_
zCh?#avhzxzh+gs<+13lWQeSs_hlB@?eq7pdqIaUIJ*gbRvILv%66#ZEE2ow&glWRt
z@CgGqfC&L*h}<Uyk&_U*jCX2rj;uHzABoM(*eMwhyjoZ;jWb1J@EjVmjj<<e(Vydu
z=1&wi!RMRMv7IoRnz6O;OHQpv+~o0vFDCV^^?uNh;L$kK$@}}(2K&FgON6a!{MJRm
z*de}&w8r*AYFujHYr;BUTW97VTVruOXL{6%`YuS6TjD1cwl=+W(X|-V+bTN4vM`NZ
z-2m~|9B<#o<Wt=CnSN}$=}gtjr}U5&!@t$5pDzjwa^z>8Q|e7_JE2|A@}E`IL5uLT
zmGk|ZUbuL?T-{o6{{26YmDCkk-$5WB(Fr1Nv&;ftyGu{+6DaG46VO`~%u$LUyJmCq
zo?D$?RJ^fpW#%qmlmdxCnn~sE62`3E?`In!@XMjHPRGAvWBZLYXDo%?>Gp3o%iE^3
zCN(G6*>n~$8!R_gGTD2N?!Swb3G!(Eu%8Yh{m$u7|3DxI>V!FKIO5aw-EAj&XCa`j
zp?CAY_x<-Ca=NWbp4UG89pn*66Il18FElNBdo~?0`ZC2u_k80ou6treMwMpg(M##q
z1$T@q!ArgnFI&UcXDmg4)I?d!0WCMM_zZKhJ_Jo`{krL3TkCwdK(^0rXMCkIqnp2J
zZu8+Cw^s`Vuv;rYcxgWIr%{K`;sg_hRgkM447d_dytJwL+wDS-ugXiWXeaZH+f7gx
z+RpTs?0fNO=%Iym(cAsT#P`YgJ+i-c#c%olxO?x1CbOo07*JRh(M1GSEFe`;!Y+y+
zN>M2yZIQMRLM#+1NhpR;16JB1O^9@*w-A>WNFYH$0RskU5kd$ZA@ttw_k`WOpXdAj
z2@gL7a&et=W<E2g&3vZ)elwAzwBgxu1x=N%{1M^Ani7DJA=$7HTrTiO5?_bw(C^&|
zM_eoT8LFfu;%}w3m%y>S&$vL`De|I2AeN%p0Q2K*i4SIg1De7_A6!V%ts6)qg6I+t
zq}GPe|7f4BU!lhQ#%66soQi$qD@*o7vO?kH{Nv8MquBndo9LsR<x0%$nwfPR?Dj}+
zetvPL_8(s=#Y2u(t}l|J(Sy%jmk!`9+fI|V{5D+>)LXZg*^}=9jnwB5YbGIhj$Iyy
zcLu9vT;mhaZ|Yw<?#h2Ro%5Hd`IS)8f<xK5i0#FyuPmQQ9`Kz!xM5RdqK*~_f-Xu-
zW*9Y5V{m>#PQx{+dV1gaI8Bg~wNLE8`FPzKuROWcHE%fKXL~@5P)E7T`Hd^b)964H
zICv3kmjf3?NM=K=kYL7h2-JIujS4@R``sb5+w|b50FZiUnWtGl3|wOT{H53rUZccm
zND<uOe(@x=w98H5FvTCxavkj7&F@}5JYK6ZYnK1rT`lssdV;i>kYEhG8IeSf>C9wb
zI^x@A<$#sLYu#^tXxndjSW$_3hut(-hnj@|{%?i$XNG88ffvgU2%!cfsfxe71uu5S
z=H`lp^*_i`oW2!zIv(azM<O=0=9|H#%I4wHEcIxRN_~#xwL}{>4Vzh96)c+FXRAI<
zZwAb6a=4fDqv0e3YhqqeR`SwvV4&OCM2sNevKfIOJ<c7M`t0llVg|dQ9h^StKxS%o
zcLKa+p#)KJjObE}Rw$$nO~wd)uFOaavS=Cv>R(uakn*}VP#Z#;fK9YRy40-(MMpjL
za(L}NTBEea`{f}QB7iK+YyJ5iQ(d&vaLyrBTx0^EsS|QyqGc<UzjIjc0{l-z00rE3
zJyS^3B1e(L@B3~B^0m~Dfuu1a@$-GhAqHGjocoJpnpr0eRq7w)y{9Rq)DPD~0a1kA
zPm}g#vl~10EbfjawW{R@&<?nlE8;3j9V8(>ab(`R456ctZ_7M6_(qb~4HE3RMBxWP
zFOdA}e--`4RvpbM4~U=qbm{wE7MiWPLL3r&THXD<%+r+W6;c3REo3<u`UsZM)Oa!T
zbln=Pj<FA<_v?U1V@X~BAvYWc(=-<6LC+4TT0G^_7t~b;J#icYx6WK1wCIj$DUh~d
zPJ%rtO5<rKAGnzIGl&f@sRn7aW>U85yd+pghW0OR`A4tsoy`(sRV~6`^+=p1cpV}Z
zXF@!g3-0^YmV_*a7>o?SF7p0q%Lnix!FDl#Ot4F$J>vcXx{;wr&HMJwYl!~FR++?^
z0$=?Xz~-b3Fb)$IdK{9b|6d^qfxhyB{{tIe(jQ2`wE%(yZuTQ__uTS%v7|Y$P3BJs
z1px}6ms}u%cCW|UKZ2!al-m2hNC-P>k~|}>aC3Pp=xOiu%?&=lt#Oe%#J_t68hTMm
z@)DGr^Qh+<T%G+@{)8}CfYW=B=J~B*SWw~UMG9b1GyO+$$C0@E0sE|gSi!D6fDg3w
z;xot?UIY!ne|y{ucv4>c0Ou9r+%d6#cFW(-8HA)^m`!1R?PT!gmmfmj{{NckfMkI&
zH$$G?oEB8z1v(dhdofL<8FqzF8lo#^rd0(LfBiGa-*X&%0p2Q6SP#d7d{{8%>VF3T
z02KuErgd{<?;VeOgVw=eGFPFHzqiRx{RM%Bm32UW76C3+H$JvB5DMy6!Bfg(06Ce}
zoKf&3x8nfG^vQ8CQwmw9oqU}IuKjJ*{g88W;@snizlF=?uY`OZEb|<Nu1qfcR}Q<`
zH>a0YHgC2xG=;nYaG1vS13^8mb5vgD0<b9UXPl|bKcI|5su{h1W>qV)2IQ()n{f}^
z*kk;FYx;Mj)hx(Y_8)HfN3^m=xPI>+6fJ?)z|fz10s>UBW=v)dRN)8?^+pE3+T?Eo
z7+`Tt5?aranG&*ueOv=3Lr21Y1~ybasJ!xTXG8yiIIzFJ!NnEbp1HuQN$8=K8}B4l
zRu`89rDsgZtN$oq&$mi*|6xD3H(YCD1Jgh+a%my|7B`nD^dJ5035k0|@zftM2!dz(
zI12zIHq=IdXMj=u-Dli!0Xx4Q@z`YNF%aHf`*FyKN^TDN%!2rRE)s)8o(gUv{t+Nd
z3kz;+U(C$30`r4A1nL089|o_uBrYMm5-7R&A`bJIS`B1Z-)#z5B_A0ynDRQ><ADp<
z^HxU{cz}y3A(sK6bKSuI*bC_Go<h{ma_#Od_|9K&2YlornEUX>H2%&Aa61QXtHJ%9
z2FH)&rGVCa4k`AN`d6ecI>T9J_XWUTNM4*A;7PbXjQ`se;IaQ!-__ZF9}VTYLXcNx
zCe3w#<;<9`Vy35ER+IqB_k=8US*i?8U!%1PbwDHpz?lx{f@Tf)2hM#+)L#LU`jg@|
zL}I%{K`Mze4Fwwc-@ZoRx&7nS4Zh}NCLxvz+XA-19R#4j^;jR}-tN)3vV(2ubQUw2
z;ZO|BnfrvTy1(O|Cjx?>spmd`j?@3AH)p-Mu?#ak+}yIBc~h>n3j&BW13mt~j^rP4
z-JF&x^SlYu_Sik$@VMsDjiS_xFG#_}4e<P)AG`S<9<HMU_poRnEd`h#a^L^B>c2%)
z)v^U_K^u|H4Qoj|bXjI~=|x<?!hfIRI@_5p5Gqqj1g^Y%baiHN0obf0q<U?l0yr*a
zSnCbPp&qy{;C9{X&JC}B^wAjuQ{-BGI8bpAKVb9xUw-hrm?_^5df*vw5UW#y3P(u&
zfZ!8^n*CRAk^l83`7rcj(A5y&9291a@&QZIzvu-bt_Jc0ngx#TUmt7H`j1~+Jh%5p
zlcXD5Kgj+1fHKvyoeU<_zYPYsAV8Q4=0`3(zQYM{V5xt<MVXP`2%InYyUt}ueVHd<
zOTE`l)>v8E3i{wUMA_902E+LmSW*DZb=22jkm#)^{|tR|W&OV@7~zg_3=H@*x|~Zu
zyvoh*25q1F2f|$U9m-u(0Z;&+qj}(Bs>cgpS9BB#R{_E(&<uKH6ktCE8Nqv=jf-QD
zX!OQbgM-fkte|ngcEpK1U{k4(4?71OwPs!T#Wamc{?1Yo{l7zkJbN6z(#f0<vec_1
zZ9aJ?siieA@L5ccnF)KGHF`oUtiDF_ciSty2a4mF=<Tn%M;U-<#7KGPUD9}R5#ZVk
z!cTc!TgNpFhQ*&$;W-veawRBTdu}RjRmEK5RvrQstb#r&tw9?_>J$4MG6-qL-L>Mh
zkap2ye#toVc?K?|U8S@|7?bDAq|F{M&saO+#A;kQYZp<R@o6I7_J(2H(gc%!cQINz
zuPyD0ZQzr0TMw4D_0K8)_j=?1+6TcqTOn72LdnOEXJ*38TDG$kUX-MR5?Bha0I|SB
zHgox4;>>?|1aU{J&r5)@Ye}t#@)Rn(+wz`elo-f$HZjtPd<m;vmQ(dSlD`D;Iva2x
zA$B}Na&qrZY`HJ<Hg_S)nQea4%L-na&lE>ZuVR7&?4z7lu(Pv-_ux*LIV=Ld=C(4L
zc16g#SIq-f=k>Kw!su3e^;Y1|-J2ZL6Va3Dc&sGyx>7$BBsfFU99?H-CWOsJFH%nX
z^NHrANWycrpG82T;+><AhU9B}^<yA$`Gub4_naRtUDDZ@ERSW=S%>tGQ>kAfequEA
z$dja<GVPl!Z$rvgR~HWuU@Z*TU_tG3d$ZV&6NqBm9P0PmWA5gc26);f+`9?HMKEYv
zbwr;)`T+Y2xT*<2zvt;TJ!2chBVv9v-#heIythZ&)nH!3oR|WmBG~oCMau3l%bV_8
z?+W?2UUy3CzDWAwSjaW7*v>7N8F$K$^_Kq4n>n!xDX-o>bj$oI4(5;U-`{0=t(?O^
z^>GM78`y1f&Dx7D(+hO`Kn`Oxjl}+8{hHIbfyVq<TI}mwyF{dwd>W|QKyS@L;#}*W
zS<zC)82h&b$krSF!8)Wgo6cDn+CJ{m(uQ~hf5Y98d|{B}ykg->+VbG~_3;gUlEQw@
z$nvPFx$gu6DliquPXcp#YB>aZ!iAwK8*9*_nVx-!gIp?mOrD8PSVm*#Ac4mVQf8H}
zZZ~LrA#V#SFKnDov_=|jKBn+H_KLUL2^1$@QBLugmD^Bw6*x!itOijhD?uOKib9-s
zyzj!vrCWurU5|8~3Kz7Iw@jcA*JAAQIza{9vvTo~OBc3Ze>THEC=_VetEO-#^Z2He
z6s_R%Mf*<Y*%RA`g!sD~hDEPTTaCKIuse=W6_xQ@0R?37QVhl6jG`)|^i1+foCGEi
z9T{Fg{S=%U0IxxJhDy+r*k;Ube`m(n)=G~*uer?wW&?&nVjH;9x(*jxg;Zj3Vg2rl
zoQd>x-n}Gf>RK?i9t3$mgQ=sp5|5!}v~4aeAHeHc>RGKTq_FWLFpL8Ihv|zv(gkW<
z(98QpxT-QgG_f&-^Wjj$`Suty!>vH{;UZf^V1jMG@0w@?QC3CZTbgx0ab+Cfryra7
zN1v)lCKec!nGS!MJdNkvo;%t1qQ!ll+C1)F8qqW>5Ws2tCTqu;T3k7lx+4-fOAF~2
zjrVg4+Kh_x9wLOqdRK~yXaQ=Wh(tF?;Km%RWlIz(?jm`9E6_JuNOxh*0Z*<*iq%&8
zuXG7y!JQq{szJPvkp4V3MrCz^9xkL0=u&*g0Y^(SxjK_jJbHL-fkvKEg*J4DJiu6d
z2TNFVhX@HoH$Gf#4<+G;7)IT((T%ITsweAT$X0ZP1M)n=dncy2ON{TVQ<;3ArpD~K
z&bXvBGt&Xh&8bTm+lz%YDjQ3~CYDSD$XS&xGejqam8Q{S5^%p&P<9G&A;03mNq6L?
z^TG3TOWRKQHhvNA_(aSo_4wEaqoSm;S>n5E+?Chr=w%b@c}{5+2xbVNPw*Cj!0#2X
zCIcTNg}EJ6*8W4tl2g1n#~s^JXcr3R2Ra%U2$Fm(N-6yS*_w5vda$OZ6e9}Xa3fZU
zL@BU{2z%Rac4Q9zTA;{@Mlc64coNZrY9e6dwO7`nyH<gLpGM=n1fHr4)q}(WC!+T}
zq$rr^@1fvXIo42|Nnn0~_|LP}<g@*RI=CT&pa~E%s6QDPgoPusc%pSVjOeAQW})-b
zoz45KuGPHVYue>2SF{XXZqHH3{h&3l-;71(gWyf34g{?A5M)C76FJS44V{!icANgN
zj+=t^Z{X=a!Kx{s0K{VmS<i#Gx%_6mDKJ=5{G#si!rl^fqlOo4u9X2dOZor?BVUAI
z2hQw$F}=joZH5E10zzG@LHB@)hxfw0h{J>Xth9E5c29u6^K_#?m+HA8k(2sdiBYik
zV3tEYd#@_?3)0^UIRrZ?n7`-T-){v9IE}7H=D!CW`}++?V_^2&k&JW+6a_$gc1Aqw
zDwY=~#(0a;gZv48Ao__&?AT|e!I@V38)2>BzXib#POBCy90G&xnON>D&C{=Wg9g@t
zIFDL!TjKWqGVy2_P&Vk_;NGefq=W%cm6n_Di7^$&13J!tn<RtCq0nyQ8l}&}fjn+%
zq5oojnzvv81gM0nT4)7*o^)@n*tj*Q3av*fycghVuhT5y`_{&0sLb7-rHLgbz{ov*
zH^^-Dl7rh`YfDSRT&0x{t%42%MU{XNqk9&r?Sm_Li_$inYyK5%xBetr{PfWqqL9oZ
zyK*3ge@W(XwVNJ|3d7zr8e@%k6{RV$aP{e3TzK*&aCHwv!Cid673ZR&^B#zx`@Fx~
z)ODDG&E^K;NEFf$s|EjzgA01Fn$qVEF8>Z#C5HlKf&`^l?i@O@9tr@6fq69lt&$+1
zMjkgJH{2XZou7;5NgOW;E=fckc`IKO6flxB#05RjI0#nc3g~eiMO%5%o3|(<X=4R!
zUb(Y`mPMDUMG0XHWtPI?#yYD5AHr$mMuUZLKd!+4r*A-G#IQ!BEJRge&x|yzX4f6}
zMz;*sl&8^4|GFUVFp`%z)qAl$+1#jj&X-Q0$Oo%SD^Jj16M`72&G*tA1SB65@6#K1
zK=m&-wYi>KUQ*!AM&$&Dx4?RX*Zvjy@E(|Ixe<CUOQGE$ymvDi2-tsx3sAp)u9CR`
zo_gGAmI>rw1Z`>x=LX^(EEqVG4I;}1SURdF889T5TZcTkxN$3&3noVH9`?Zdj?~@&
z(b|*4{Q6mFZd?2RH5K43dO6x@073zTu07myTja_e9H;<f0Q2a6ITfh&E7-BsSw1y`
z*so|OObSDi*i~CW7p(xo5-=EPj7aZ7q={t}JqFHo3P?${&s?WKo8(&cKQaa4-e+E`
zx%XzW4iGcU3w#N{a2>oKG*-#|h*>7@axqg5T0T<*;w8VYY+erqQ;ksw(G#G+;VnQ*
z@aJJPqq}RLIgy$-^p95w+5?mt*q#EHgaFVwN`QF)Osr&n<+f@w^{;?++%5y8!BHs4
zyP}L!kUA9S8c~p0Aduy=iSg&kJO~Q_+t2{$t^9yD<}J#UggcY~8MA)@^9^@Apq~Ph
z{+g1Ix1t4g!QeCmtP4TbtworF<okbI+u!Y4?X|n-jkRR<LhlF=?sG@YOt*mIfx80E
z<_~Vz{5Z%;kDz10MhMB($FD$nQ`rD27WkcQ;0AznYC^kja3ch_fz3B`WB2UFB8*K1
zDEKx5va8Lh$a3zC0!W<qlm_7aUZBW-3C>;xXOi4jI~&4SxBEev8wHnkyA{BhXS_PW
z=xhO@r$jgr(+J!sY{z@{JGXscNB~t_SChNIUIdPMW@fY<SSBpEyox6E&lq#|SsmzC
z^7mZg5K?cpU!p2Qqk&m{Y_<X}4pM~pi~DZ`qW{!VpNS=mg_+?{W*Whm&5Y0t&ApeO
ztVhZMvCnc%O5Ey-a}C>VrIrgpSTMPU7YwtQT?z#|{6OGU^Y&S325^uA7kvL8r^r>t
z0610&1#X_QF~QK10iX2%C`nB2n?}I5lE<EzoffnhDj4yGRN2%KSQS;eMo2l?1c*II
zq65nuAyx0HPbM|sz)D+n#2ddCZTLvl+by4qH!h5|<g%3xHy@ERQhc{hc~^9P{rQIS
z2UAOB#_8P(ip&Y*?X<ax$GVG$YNS1>g<nrS=Q((33Zh4I2;TW9je-?dF|YdU)spq|
zKl)zXWgzVak6@wJe%J&7W&27&-mPYqDAatQix^)0!yvB~&DhW{Ar+As*PPt3^*Jje
zy`^D!B)40J9@R5OeXG9<@6km4{(m@C(lX@0)}IvVU-!HBftq^kd`X+%ls@=ET`r7q
z-XGI<v=er3dRa$Rj|?wHTN#%Vd2VnDIdt>8Jw_ft%G{rW!?2ta%D?LH@KmgcCF%7p
z#_o%Fivh8knbsNRqUKj8JgVpUVUi)mlP4JPv0h@a;~^7ynB9hYv~<?f*fIpI_IS$e
z6)go8JO3Vis%H%Sm1%)=-;(=)@}w7csIBlY8Sh)`NR8~jP^-opv!+@+NJ;acjDfzF
zMvsXpPB^%ZnpN-2#PMUdHT^(>_LBc+1Zv25o7DRrEN}g6*{l`XNdKYbsm^L$L>0bg
z8?$o*aG><fF9qGql2+=NL~i*Lc=S`yuH&n-Vw0bE!0GeSaiPnjJO`5;aQn>0HiHJ<
zwrE8x?*}K7@#4x_JRr{OQt=0{B=I|cS9R=Nps#EFu^-734AIIDro(l>jV}tuZ(m}s
z@L$XCWzBTIbsAfo)VVBa>%ACYL>L!76-6Cc7#Tj$9#zZ)e7LvlqtKk@be7SKpa1?I
z({ncs1%d}yp*7CYBzCSi<k;=ny6h`9A!3Ffw1WNlvcPemkmJ+QA1>YOAL7dFR-(g&
zam(9Ulrw*&&goKSC6B?UtZz$$P_mh$UF2^fNdq-+6*~(+UP8)5+YResvL@x%9+QYn
zS+h{ZA6sV3E(uyq@klI)TPvOvByR-&eq8C4po12k7e#8WTccrT2cfd!5a&ZtpM=%z
zCWnhru>!rqlA3`>%?#3vP~}!vrP*~#^<vSQh0C#S3Zg@k%gg&EZg19(wVK#SkEH~-
zG#{5ZTFhuw8R*s<PSseV{d(6`t1Y;$F1%3}X^qJ=Tv&fA^LXoeZEebqX8On88f<Pa
zY0Gyx&HqVtkU9qEl8qz2);3%61_4g}RLY>*;Tm!6vjM>rb8XW#h%IV?7=p`66zdNX
zYw$58SZAI|r%L)&x&4R7HE1p*=BVzMspkq_dfh+TjZ{M_hz`(=-Ca3a7~y;RCyY4J
zvn}<LO^sHOyR9}(GZiCOhyA4clyhn6P+I)z$TGDnatxM?3BD5{3%fa0G-pioWY-){
z03?aoBv`sPkE*Ee&gQ~CGaKOwH?`THh(zvOIVzAc-wFT27BEL+2a%HvrdnoT$MOf!
zqz1Dx^n-_54GtNVb@&X#ip${{zP%_ZyUS<t^OceN;#Z9buXf>{YR06fm||*M-z8t=
zqZdh=5`ejOA_+swe6bC=W2o0%D31CJ!#4Fuv`t>yTu{6azedV})2X9o4U9NfcGK_(
zA)y^o=G9jv@s;UVHC>0bQS726gE9c!t)D<~rSBcrE)*%QNpG*QLoca}>SJRNx;<u4
zk3+jgSV3HuTX5LSSUX??An?SD*7;+5pI*}Rb8RAkEY<@caT1a4^HW_upHgt`Q^hP4
z`yPR&R0F%zB2J6#TKAmOPNgmIe^bkU)6?5@Ef<G5@*f_P(ICJRuK;fq@1=~|5FpcY
zv>b`zFrlMVTvI_xo6!QL(YPD?5QZq%+`a;z>px64tDMNc1aV<)XWU08xU^vSvC!+Q
z_W2Rm#*L}q0N)>gf0u4Lq?-CkV{Ud{ZpK}DfigN$lzI*K!BC}ZJudl?Q+Fw8NqTdi
zZ80@>c@9$MzzMV)Q0q&jj4rC|&{K&xOqy!K2el2qIg`(&$+oD!F2PT(;md|iDR7i+
zCh?JUrE=ysHTWmr1!lPlM*$O{cub<DiEw&&@-u7DeW)U(C6HHC)K_D`i1DV!RO&8A
z;dC7)RHCJk7l>`7mZ!&*G;gXXAmXN|*oz9>CQ<AK%H%r3V3@r5D}aOHh@1h;m#Mi`
z`x%g*=ty4`6*Uhq`45j5mo;RYXN*%E?Fd%qJV|v+_A_64I9;TcPZ~R+Ie-&_Tm=|F
zMi^N7@9N#f!Nn-mO~7jj-tweaX_j0iZP&lcK3keWE??Mp#hYa#$4XgXGLjZba6|_w
z_v_P9<q%uP@Qf&G(kF=)4w54E(2ADv)e!LAmInGqJ1?w1QMD&yyx0uM{h*bFeb!KU
zx5y$Qg~mc418pd@s!u(p+DOv(Dj{h_dCtN}#?C00ry^j151XsoUmUn#JO>)f3?8`W
zs&3c9P|j>5wKRnhVD(PsXkxFQV~y9s9PK_}B@(jhVYm6fU#x*fId$HAS49y>3^fey
zRg#81wVTN!mneJ%CW>g#y0(R<a2x$qrC<y?KQ#0{rEOSXsmuvMB<2}SxWgL(Wdr=^
zN#>iwJq)u&iuDNRK9^=?wR(TuOnL%hjqQYs?Lqyl!oinBQMR;<*yCIUG1~d0=O}^J
zlZjQ(!E9SVl5;59(qzqq(SpT*0yU_g4+gEoZ(VWDIF0%4t{|CMyE({@tSq4ft`Y)*
zUy}Vop}mMNJuUwIpml4?Z0Y{8zOBv&$f5a(rf|TiBN!Zx8?Tu&Cea|aV6X?B+N>vA
zC~BL4e;1IWz~M1GY8mcHtyS!%Y1M=n+a0Zwp$ey#s+ef%wP@c3KA0csYyWysx5qQv
zTK|>S0|jGrr`4=!nv3L_hqkp?DKkd-%U!J?jk=fRfZe2DXu=)BHq)fu0Av%@!gv3}
z(~4u^1kstTTs8|nxHAo9k|9po6mlF{og3t@?C70gCN5G9sSg0imN~io`^)lL7yF7+
z;0^b{$bTyuwPd>IJ*s1y-6$E5@|Ihoj4?sRwT$hxtxzB;9U;eFEP0g|)bMEmbQ++#
z+To(B_E9zJp487weR2S&VOjuJ2RyN@`U-<_?-bIFbuo%e+3^zsn#R`)pH5;*PgAc!
zOG;D|N;ywjDz7LahOlHVgs+r#*WnRSZP>{;bUcHzHdY}dZ*NlzG|mH*fOFuv-v%8m
zMr=$|`m4PG4-OwKaaF$2NSd87t5(L<6Qu6S6S1kP%D5|bKhkQPXbf3M8G;I9>U86t
zPM3|R`sJ=IL7l2oW`+f<N?}z)YzCKRg*4A6V_%C;!dim?YoZ$kiWZ>kbo}SZPuLM#
z^=*n2qM#tEiI9v=h$#x1-oIo09oMr3)6&Z6G1au&@!4nixD4v+9y|hMLyNOQre0C1
zTABmh<FCZW;~7}D%*t4IweX+hRuL4|jx*P-{w$Z&i(g_<FSyWRx&X-t4DC@L`D(s5
zab}Fz36yZA7JCogU6hP6NkueKwot<WM1q+|MR(zZ3&APC{2Lni6R&ttUoM;0>XZF8
zc~7H_<B(D8kPGYmGI0UDj0;bMi#<B&-pWS=Q`S)ZpiNLkgx8~HoY<E<Gq(Zw+gkwD
z%LalM1HVjr03BJA2jaiGMt{Jvm}ESSc3hm=aZYVNL@Z-IU1JZe0iGfORqNi-!YOjn
z8Z)HwWtrE=58OVZKo3&s*!L)utnP|L%NQdj^w?LcuRoIdslbtG|9m;M8j5AEjKds9
zx(N)S!7B^2mJlaXhdU;~xeW|nK^k`}jZs?=748ZSlE14I6n|k}UDyIJ;lozYP+DTF
zH$AA+>4UDKn9ZOye*j-bjDwUa>XXukAQYM&unib6!Vh;|oT<UAvmv)d+e~(Kl;XiH
z&v=^p>`*tbjcpxU3+IqA<%Y1|0C^tEk2l&zu6)E}SDein0CvZuhMqLaV)~A;hgB1L
zSoT7oX)ZThDXIGCZzX8~9N+d|N4<-n&YYm#TU@(y^+nD1arH;IEg&k^;ywdq*{D)+
zU0fF=d4cJ70#wZ?9gy<zgVI2QTA?lh&XlVl!p?N=O^+F@mzlEiM)Vl7I}l)fg>f|W
zIH@HNJ~=LMDymX6fTo<xwTIC1Dmdi)w$Vfb^&SHSLyz;K9=9OAtSwa8E*HQY9q<dh
z?$UCElmX^-zeU!k-%|k34$=Ogi}GP&X*b1oY><pg`2$!&*xcCg{%+qk)($2ZNCJ2X
zK2=1WI0nB(*#^x&0R?4e$wt7wwXOZ-7Kuj!16u%>*ycFDQI<4fm^d-DCx^cM-jJCA
ztfbj{u9>jk^T{R1!g?WddEH<#EUo2%v{udbS5Y&DO40}Hs}&dcY|X`{ilRPFkprs1
ztY{!pl3{Ijqz1@!eTS#8B)k^m<#1v7z^DUh(}!;&!v+>!jMzeDHjp3npd`b^Xz9ca
z1-NkOKrlK45OaDOvE7t0_%*+1#}(bBeFvg~zWI3%eNrKYor)?+vnVCk_Y|?+n30em
zpM^PjFvxjc3{2&F4KswPRkGLrHXdrxhCFoDi%DGL&S!ipnMWL+*{UxMVL5BO&-JWe
z1(8I`?wlO>k^yrLbrm+hME8~-Goh`FRRfz9Sy3u010fA3Ro2%5`OrSwJj14tBg>p{
zFum_gSMg57=i(bO?e|Egcm%NrQ=m<mXG_wN`z~Od$}g<%T<ZW%A%FwS!eouAx-_@s
zun$~#GF*PWoYF$n?BCj|)932uJ74Z9`(g*X%7`|s3J|PCnlCF*DlWe^$6l^-c*u36
zJW5sLHF;Z3YqQN<v^mVn7v<tJ6`Ui+YTf07NHpL{6Ah5CBC@FeZiB7H{N}b!2TZBb
z0<!F2_(9!n#8!jYbWZ{}B_3cj!hRF|X)|Eqg8Eq|j#fU$05e6nAU1t6v4@FT0qaH`
zm4Ygo-P%`9I)0SGU3HK~=J{+f?mm5~yN4*(tc*(1f-Oe*=(Gv4<{GmRIB0(KY!am#
zeky_Xf>x5|8?<{5A&ddBZbl+t!1U0cq>G%0XSu5zo$x4BcZNHi@RqA`QFCz|hQPIR
z1}9TWdZ0(wYMhL{_B(gxI+!Ajple|!$Tqd4mPS5z2M!{MSw51nyPsi-g<&vHQZ=@=
zZP05V9xq(%tC2@busX(DVrE~Q#$V{v#sh_q`wC1PIU$&zrQP9LVr&lm4dYR}Rplzo
zLW?w8111=X&HMe9bVlj$g*eS%jrYHJU$sI6D$I>!#(=(^p*TpH*dR70Y5^<U&1QO>
z|K?6zO|+`$6*d_IL#g){6yv0$J69$b14E?x@PDQiyvXVYM-y^EBd4ww5w@!mt#@x{
z82suH%BzrXP4POmH8XhNIA9f$YpwC$@Na}g5gWJ#v)2vYiKSEe)iZPMTP){xqZ&#U
z&7{;(I09VZsCN5_Ddk-95~udK7v#*lTAcRpceK9O{)b06Qm~%8>Hfd?!}s;2yY-9G
zm+%2x(_21_F%5PUZk_(ze*3w!9Jvnl>4t0qt9^wL_%g%vr_sd&1tf;<mmaug`flh8
z*%gNA?inibFAG_sJtgxVpmpTm2iRN%hC_w8Y_0*f=qQ*!a{p{b%7P3djEKrMeZS6k
z*~a$7hS}t?=2Vqg9mAZBA9^(0(TDquKq*c@+@}SPV)mRnhn6=fpqG!WE{BuFYi=|d
zKGKDSTzzSXUYh3*49laG{-qjc-ID3s_BOp?R$_iX@|+$NvDAOaM4%LAFb%tPMHia4
z+?!5b)SIVOA78%E9%G7)x!t{6#U4G>{-rv3vwABiXnzkq411eO#J$zXe5#%p=ul1m
z)4IFC9BFchk=*|Ked*(m2IOB(<Xx$|W^#67jk)#Ezv=L6m;5sdLDswOjIb=}eU(M`
zT~|Q~OgIWsKD5*qnw@Q<jKmJ~&;4p|bZ4Yc93sKl&O(d~(I2pbt6ln!zPO+Kuo$<p
z_(v>Dlzg`Gwn>U*qwb`&!w*n}Na>Di1x-cq{t3hL?&j+sqkJmt$vZbzI2j{vZC~3m
zL}}(RcXBLJ)N*L?Wp^G!4j?Lm$#f0;&orT&elIuKTzQwb;hBL>?rZRCKK-e@?I%tQ
z7Bf&^l&H?O^R`46#$*=}T)xVZEpcMTJ=ry9-~8q9;JO^XvViH@5L<C(GNsEdPKnKE
z(++$lqiCA;xHB2eX{C*xO*9(l#c!)r8GZ8G7gPQ2y`jvlu1)eaz?2}&uxu9BGJow<
zZdF(N1g@VwYPS~Xxf11^XKlJdY(w%^;MH+CF=NV7UyI;!Re_1%Vm@ts^pFsBDp*pd
z5Uc<nQALfmc~2i)ai7}c*Rx`x<jlfl-q}=TsuLZ~)NDwflX(Zn<fta3Z{dFMmQg--
zGK<5LA!F2&P^~)!$!kwDB6RI3X0hQWmR4=n+uab}E7^m3Gvk8N&NT#WoYongjELZ8
zHYOV&(bIQ8Cmu!oMq_r#yO%Ym`4Vump(P0fKFFE3EmsLC>k8j^9fr070ZYjI*3&H1
z7wpZcW7{5?4m%m4gsk0s#&URMRGm%3aMfbv6YHVL`r3g^lsc&pWY&eB%<pC7Fa9`p
zGIDw)md90Pb!>(;HY!Mar1f(#WqPwj`nzSG#SMiXhl{^|XG0Z^r|U02yHQuD93f$u
z?tda9M>9nHXUo(m*zF#d8aI4uZ#hg^$o|!ga9wU`W^39N$<LEvRj#oYQ%VQOAuq(`
z=3uF`imki}pW_)s`!|W-+G|L<3y&JJT&q$o3_5d9pN~K8(iHNFbnegfn33#ZiLM|U
zwL7WNM`RODpEQpU5V8N<pkHhQ!yA->0ydDyd<m0ATZU2Zwab83hiYx2SJ#PTZC&xc
zo}}Dg$$XHm7Z7&-w==lAKhE5J<FvE7W<IaxaI?eSc$aZYs{<mp`Q~$_c~0)tt{-@}
zR)aw4?003-S0%bEQ5?6a{3{d0%c2MKzkGa{>`swZwX%CfYjT<+w{E6st_^)c2ibdH
z67*~i5AUgp%{}wtomi2kpV*7<*FWz|Kek_?QMV^ZUUM5VE|X}1D7T`K@9)e&D~mrB
z)LzL_oPngeG&hs4XNniNeM!!$QNu4$6FcdxxdJVq^5&e%Z#mwJRtp=dH_UDlP!G@I
z3sW$;xFL^i3ePiFJx~f*E~Um+OK^oJ#e8uCs|jw3RM+#L@|om{6Wa=?`A<FDE>g%h
z%=Ci?sBgf@{+A53Ig?^WB%fifBysUz5o>Ms%dTN&?1dj@1h{!no|sA7k^zA>bx=Mi
z1FI3hE6j{<Azmz{&J+Ehv05;XhDLn?^vTqAMp7@D82P%x{*_nUX3-JlsLQ<~Z5m18
zowEId<##IwG!NH0B=5*<k7Jg^h>MlCqaX$2Q%t+|_>*>|<{7gQk~{-psRzTRLb~no
zv>1B8rpwB{8}5yDwQR}u>)9)|7s-Au^pGUB=7wwvr?WN0G*28|#nx=59xn?}Znrkc
zG#5u51^LE(V{lPo8&)bFVZTX^g1-?J7IF5UfxqmUVt$wpCtCH5UsUY#erT9osbvT`
zq4!F^+Az21mZ93FR%%d^D<&kUwUv0(a_Ux3+4`%pZTOkEV=~qTRZ&8*<Z}q06G!+y
zJ&czx(zQZYCR*7f8BD0ST`GbkmIun3b#C5WyB+YQXu2Co9CM^;rls)HeT+Xz)KTSK
zuj;=<>&c(}d|&=$pR-1m1kA{Bwa_9Z8SiW)m$4>v7!+WK6&pXOnX+%RrreQenjTv9
z{MP>MPsRgIw)t#dR~sT_fRh|<;gxO?t;q3Ja$;_pi;7BoqEvO0iUu~kjpTMaT*nPY
zm)#q~-d`l@A#K`%u-b65E+ymzCbO8_R`C5~<Ov%<qHqHg$8LmoB}0}eM$4DGb)kj(
zG@=-=Am11i#y-?d8Rj<&7&pl|%>vnAwLMLI%m0wPjm@^LV^$%DVucJmHl{hT3?v&I
z1c!D(eFjxiic^qGlqqrJNsfUR73*tXGefaaDH;%)rA2xUExZ@8qV*S}tLCoLO1kEB
z%i!3ZUc|sY8!hXvTNdOL`L+gIr3~im_qw(I;tD|*d6Vx^!PF{UvW*%7+()v1mCnY!
z0C^{iRLJWBydE2oJf2!k)K03GuB5g3aFDyziKn_olhNfEdmhmFpwA>v4o35>>S<aj
zepO;-VsxLmzmV@d{79@oc2w!WJrj_MuN=!TCGT#`&G8k(GW_K^rP%!jY8T64tH!lD
z9WwE~tSiV~#wA5Obj*oKMcIe;hJwYG`wR~Vk^sS}Z+L8~u>)+iPo=7zr4;JP1DBJ9
zGmr4%gA&}PK{Y!qJtMg0^FS|&@2Ncl?rUeS<>?h}KNPYgjTZ-cv9J`z<<h#cO2X9L
ztA|7RbrZDal+&E-Bj_M|j}&4+pBd$d+iHP*9(j!qV+Omt@w~?fZ2Nu()@P(+Bkf4s
zFYHeHA!^OU4jwZG)h(DU9DZ=F(#r^WEA{);m-U88>0zUqx_Z>(b4Gmp`zPrUXVTNL
zrYwy{N^7EW6DCA(mM5HEAZo7m6k}%MzoEY6+9O})z`O?PBR!FjJkv47S^Al1Tn*V7
zP!s?cSDu2oGePw%P1PwO$2#JZe7-qGmw1~$v`gq{j3-QGoL81^qOs%PP;VpcuvHgW
z{S;$Nd4+vdx4%bkYkG<0sh@$wmVQHxlrutQJDf^EwRib4B|J5kC*IjE_ClQ47lFx@
z`8ll{C8_13sW&9t^U~0au!IU|-x|rTcWr)uJX%v#*^JnJh$A;mfd$vIv`*Wtc?;Vo
z&k;4>N*KqO4(PnnGn(%^Ae+1>Y9CPn%4jjj6P=Tb2VUtBwR3k_Kuqg#he~x@ERRR>
z@Zmc<9kfg-4>ei<m(6$3`8ZrYZHcc&#7!B8=@R-Bd<HPHQnBUxXU}9xx##ZQ%W0;W
z1boudt0Tlpc(s!s3yz_OJV6<7o<l@4lYojY7h7eIxj?>*hfrQ~Svn30A`4CG29e3W
z-i+iRl0ccFQ$J3RIpld)(TSc3JyaarqsY<<hLN@cmQQ@)pzRmSvc|CZL>fL`4)0XA
zMRAgG*5TfC+4x0{nruz;%5JCmqc?)mhc6^hGiGN2&ohuOTt_AjyM-N$2Qj8Gw>CAy
zMSJ>day#cbo~9p#T^?5IU{G|@!G2<bGcyRwa#zURRs+=^_zqVLKFDHU-hb$zN@qz-
zb3k}h@R#(-dpd8^L2T02b1DouE)e!TXjw4-c(12bb#N0MN7Ncp?M(4tG78Rjwg!hW
zM(3u*cF%uO!z>O1;Ss|y#5037(w_88Tc7>$#_RK)wR+z2w(MA$Tb&L&*N0ikNSpb6
zw-!AaF+3#9ghx5R9uAoSY63+NUu=#aWN9O*3b^78a7{(Ttf6*3NCY(zY-qG<==z)y
zY6AhD&{Q2_jzPYu#;?-Dyyy1|pIB7Vv(oLW+Mi-vgKIo63X59U@A};_=<(b#|9V!Q
z&DglWBqyIwTe<j)M2q`)A&b)wnsLZ9i#2gvrZQ{y0Sf`u9%SNBw9y0YD=O2B?wNXM
zZk!*JQS!lhXzgHW+6P17%(GS?m&_@DpQzZzTPdz;nVAX9&@VM2`1BoTfmuLkPx+~S
zdyFIG?&8Esan2C@AuvVW2afIfwE^+tjnXtwPJF+Fam`BPE4|+8pU=tCZnf(|b3gEf
z@)Q8*|5+R-#zNN~j|PbBS{})JrqSX^v8h(@+^U@3sH;Z5B>4~}3AVHRl9I$f3WK+{
zw&RqN9OR(LUe2(k;iWNfq2C(r<nfAB1SrgulIR)1UvepTk`@*yL3i{!ADlSVa8EYg
zWohL=sqQ9f|7@o6!$seSlGFgxT^<SJDENcLr{I*lV~qz8$zQwrv_@DtBam0(3<<sJ
zvpNXi!)kJ|eL@QI^WL~+t3pRL?~{Sti;Q~z^Em}mi75vYzlfk2r@1PfCpR>_$C{`k
z$F_XOXoCPd_a6_3rZ<e$Zx6fY^>@|_zgu-38(N5ZJc6Cx<q<xSk$$fz9M=z!)7C25
zoN<k(GZRC93wP1Zt#4n)EOtSX7n%1C{JGKz{F74fU^<|?V9yk!OePJKErG(pK>-yr
z%Yu(-!R!9O&Cu?u%uY0Q+;j(Xns`QQ?Q7w8nS2wB?!~Cb#tdf=*eGR~Ln{z&CV!B?
z;FSEgyk-R#VO707_T08!bZV{w1d-R>q_2u6`SNr)+HHXY4D0^w4=3{*s4=M(Fh_cl
z!*$+kks$bw({2K&rlF+=q%Ub6u(H%80c8BDNBJuC<ObL1i5Ofpfc=1`GUivcpe9<(
zU-NUL^2A64Er*`eu6DGW0MEw#$}oc>mzeLt?(qcJ=DQPf9H^c`@_L1nOF$PMxJ^FI
zI=JNT42QaN0(PzgcHvQhi%>p(9xnDGfErtFNO*NpZ=e#lbviBDYqlN0zU-+(3%Yk<
zmO~mW_S~UFW05<$3{%SL=Hh!PFrZA(ukL{ZqgG#bP#i6tr!rbMi?R?rT!;2SRBW~X
zz;O{0mgPdbIR_3_`csi_T^ng2oncVCMLdRHQs2H)r{N!1^P-b*Va?~ez{_BR24-i!
zxGtMm`dXwqz<q`W&;BA$a$q3<wYpFzVBFk73zVU$Wc*<C;SJx(eA?)o*pE~{Fnd9E
zbwE8*aMG&-e2uz0JHOam)He%JR^&0Oh%lO_I3C_595+I&(4!MZKKJct9U}Hl;sgl|
zFB)zqjRa5CeYQe<WX4_l5+UzIApQ|Ksz-!|n}Q8Z@afU}rGQ3dv;fS({=BE61`g?Y
z!!BBGC_6Z<w`i^%t<;v!JQp4rt;=59{CRA)xO{B=WoW+*h?nu>GnB_7Dc8}%_FLCK
zx}?|nq5-#5@%!Jo#YcDm6aF&K$)af<Yu%i<q!-=!*nDRHulOEICS@bf;nW8KeK&?F
zT_N+^OyQvjXC~F<RsD^e*4AZ?Hu6;P%`97C9_ep-^`K5b^Ml;AnNdM7U;N+v;fvpL
z@~24g>-;~%tD1D{tIa%rfGo!J5Iks_BI%I%iy|CP4)L@;5T43we%63Bw%Vl^XpxDr
z1~V-ccw=9GW$X7^(8y|n&c{eaMb#^;^xspmof8fJ%*ivJXs*<>&bs{9={k#!T1UL9
z?7IuLae}OPz)7r>O$f0G3Ph$9*Gx-B&FO~c#Kn)x=MO8qTOLm|x_VHlfvt8V!>J}~
zP3`E5_t(T<yeUQv;}1-Ay^5D@E;drVW1J{E+R3&(k!>X#p=^Lb!vudw%OO)N4Q(I!
zRhE#*2<n6N4K~&HL+9bE6#Vps#%<r~;;(o;NZ!(~Qu^)VmU)UrNV)Ktk`_sRS!WK-
z{A|U%wE4XChndR5-4EoXJtW%P1z#mQ(u3ko_d01tCd^2fe6+Z9=E`tkr$=;`>~JpR
ztYW{6Y2-0hfX~>{Pmh0yZlb^G<o8=T`86u$$hJlh>doh5o4-Gfx_b+MJ@Ngd*eWO7
zMg-zxqggZ4CK8hwGkt{8weawy)Scj$o>z=tGJ|_kgNvqJpy&g(3CMh}#&hizkIF9H
zF}pSVI+GW%`cvYETKD2%t54p{l|XZ`ezNy(ig&IfS|G1;+KdfG)bALcRYd4#cgy2V
z<n5CAD#eu;RD1H1xi{(0dwMS3%IQSn6y5SVCem*;-fQzdsWR~6^_M&5mD!O-j=AT{
zZ^l-fYwZX!R%1HbtV}HbbmuR|kr}AMv3Wb8LfvQj&%1h}KM$Ab5_=+Ner@Z8BrnIG
z`F7_UgQS;vobxXEw17dH^0O_1d2)CiQ&iKMHtFg0f`=C^mvy{Ol_g`oI3=SgNQe+$
zJQy@HHi{d2yuL5L;?IG*r{39&TPv-U+rO|vrr&sHk)yHI9UqcoWUBP3qAW5g%_DZP
zvqK{Kb4SCSO9OKFB|2wynE2DHs1&1tge!?E+aEWs%oJJ_T-smcrl)_oU~4oR-PY14
zf2Y?Y(%@G6dv{le+@S%(f~-pIpzdo#f;UUK`gD8Z?@ta1t{8g%5n~p8!bjHFzcK5Y
zITV#=|DxK3l+;qjR`qqw%X8ExVxgr48G%8NLb9hus@kO$yBSj_Z9&hza({zRz&Be&
zJQXDxSj1B5fx55oVf2(@etC^M<v~p93^G(b#&Q`Zf3S*s6Zo|zqQYK-<Z*kowFnoz
zJUSiWqM~7&cc-AqO9}7hWI{0~Vk%WGm;4!XR|Q-?&_KcCEj|`IMNR3#m&M9cl_I}Q
ziR2#nGta5zs;WQD;=nC^ng{8%!ivpl+bgb~?a!<FtSgiAn6f5?<s+*<g=IUlJ`b@>
zSgY|rh^J-Wsp`Ic?^ElBS1~8`jodbV*)wXTclV@MgZZ(ZzO|DEw{u=3Tr8{G9HY5N
z1?~zy3~mTt)BNPl5-SOGd#Crbu}h^dFYm^&4Ec07sCQTPod`&iS6#=A*}}j{<)_kl
zMyLzYD<Lrji*Jf!e!Zws>#jfWuBBG_s_V_j82&r`&vCb3)1%MlR4-<DNn9PCk&Ui0
zPWNw=UEZ%}LN)R(a~eLHZqORN;@;yL=k-uX+~Zr(?Cft=&qrpDJ9jx{6XFNrY2)9{
zC7#W}CktiUs2$aPr-zzrAmbPpU!Sgb>XB=5ws(5>V^IPzZhs@bqpIIfN$^h8-OGiF
za>8*7GF@uDDD3&fRRl3{!)}fLgSoJl`wIU@1Jqk1kP3F>>xD%fn|W?dt=f3)ZYoa~
z*{I`|j+;SqGp`g6k6?z*Mn$UJniro^e}9sQsBpxQPglA-Q-8Q;$;mlp=?~?Ew?=Dl
z2*M-g@0T9f4E2zq7w*eQ9<{nV6>VK<j%skj{4)LS0@*-%HM&d{gV`>hQ+(oN7>%Wq
z`h#ES#hw1#d;h%ob=>JYce5Qx`Hv&6AFf`S?R;q@lv65|QnHG-HhB0gHm)f0RdAe@
z^_K9-#OlJ&@zXOs4?ztdMMaPPLMPHfi?$=J>#q;-0sJI`7EvoDdt=I>yM3?qooJRg
zcuv^p!zAKv&V`a_$5sPHf}mV?$l&(TuO-m(Cyz<KEo;j)^tgzZZf=Qg=3_QFf1NH(
zU#=RKxkT!_O&g52EFW})etGoTDNC&;X7)|vtbW5Yy+pL>MRk)I`#&RkRV8```bY$;
zz6dnZUqNKEZlOy5!^`!eXg&Lqh<F6Jh@QPZlOCexW$2sud${Fh`T4sfKGpO2@z1`s
zeN7m8_nAJEVqUPUqM{|U`9nQAF-#X1!|UX7Uy*YY=6+QKnfo6e3#%GfDYVxL^L_M*
z%J(lNmR;d2JZ$ol+Pb3sD4Xcs^W5%no|LqgQEL137qwLZS0b))?1awSG<41w>eZ~l
zbVGWfCA&t44%>YEqcuwSk1FT|^OI-?_l{EMJm-^2LT5&M$d!o4rf%dpM;1ziAv^5T
zF`x^*%T9m?UldtNpVauJ9c{X%JG{JO{;|EeBQb~ma?0#(=c2fnma|Bfg=>O^MzqZ}
zThZ|F@S>`2`*r70(lpXw5wCZuC|0U5vLK>lv`n?A@GqN~i#7I3)_Nma6LFR?8cD6k
zLX9P0_ZgOng`JnHJHs&+x1vwD1za%AZf_3pJws6PJ%tV`qv?h&KV$|w_Ru$~l;WAk
zdRgir#uZJ-S9%ZrYPF;aDhD`hh+rO^hA`s0RSV*aMKLY%EDeE|6>~f9u9wTrDg|Lq
z-Z`EPQFyzXlI|^Zbq43Co?v07YyBY%aBF;gWvORrRn{w9M0q6LcJuk@1J*b3hbL3k
z=cyS3KNYk!rdDNqbn5a*$k+36c+sMgra3a~dUI1|$He7(#anu@#h^A{;Bl+MX~)~k
zj~&$`Q(pNDkTHsExxuL7V1|9d#)#>sr(Hg8Ap}vV6CII=LDbBm1<Q<$POIIycX>4X
zWn=bJWJ$7|MqTHPYk3=S;>R>qwKS%#zml`L8{1cOlXOw+(;x06j+4>ADw7&^X*0{f
zUCMXW%0$;&Zl+S%3F<Vqq6wX<h&X)nL84B&MX8Zta_1ZCE#qUi+f0=fWs|GyT;JuE
zD`_1gHO1(BzMwJiRX_Rk(j8J!lcMkKf}+S#AK_DliA&a5&XF~eUlM;o>dUI5B1C6M
ze<e@#ZP|o9b@Y!E0ynWHUPVG5Zdr+eT3(;Ojd)#~GAtA#-RY|E(K~h;;S3$Cc4}=}
zvGBV1Xho8ycvoxeGRN1NlvOm|Q*)}Qumsobt-+unX6T1pV0|$U8Up^qa|p6S+<<yb
z<CN@b?p!&Q`%T2^&x(tqO3LRUWjT{1-45f|IVr}iZL--i5PP&F%f6tyu<hk$@YVA9
zr9T)xl|xw&yAro-+=YqFq)CLs^#%Kq2Q_$`q7Wx#pAUA#hBYDcsfWvr4Hm!I2lXqw
zE?1C1S9Nuy+TG2`h`OXXG9Y6p7FL*y#h&_PhHw4+I!a;o+x=TU-R4uHR_<{{-R^Ox
zx{c-rGT?0&)Dsj}i$z~KyWuEMu}Q}=7AEsndnJRwWLEeGNROwUtPu$*-P!i}vXS{c
z=-CuOLe!aRQry`sp?QA7E=*n}X(>T$U~bACq5HZjUoFN<Hw$qs?@rf--}BGE;$EjG
zC&Va;Z{jZevTn<{VDliW)P{DsH_+CM9<0A#{jd9Vrv|>X;J59*4n{au^_N>j{%KZo
z?28+rH|F!ZlZDc+M`lHeT-~H^MRcE6ex82#b>6{p^LTW7Gkqg#nDknp3aNS)YUcCp
zm}1SOurSTJ#wx^YinV#kzD(Lo(_;k}c;AsktF(yOzc633YWH}3<@RrG!j|qaN-x>2
z`gUhVUX(^ca%ETUL=_rpy&J!~Fl1_U^Uc$O_anba|K^J@{^qBsG2`3+%}4u2se$UL
z<m@=nYq(48osP}=(vW$D>@+ngeC4IA$QwqMA)*6YFVwyrGlSi3?#xSn*rc7^hLpDJ
z(?z87H^0Ykl5KSUkm&0j#ap6#zN9ME4&6R%nCZXUr)G6^thhf}an@mROr9Q)8Dx(!
zDs6&_-k$zT=(!y1f<cY2%JIp}59<z!y6K&oh6V)4^~KA_6Y*br#H6nc+;>wsgHOcW
zel$jvY_D=S(j1-i&{^_QNZ0C=9aH-y2dkYc@<35KW3le~#KKV-H@3@se$h5*e)sA3
z_I)qdjddQ)VuxGl<EIM3BUV6+H}6|}5kT|;Q-8|Cvrnh`e{CHOwtxI_0Q?T-R@Xqx
z7CO`SS)4@bT3_!PdW3afHpYR4Rzi;v@?)e9bU_#DI@WqauCMYLpo-zQ_q&|*q(xW{
z1-dUSt+oRqVw=sGsRl0%&eW9f-AQ+5vg*lIs=b!nEs#fOX0n5jG0PA8tvD~FF{u~9
zdYsdRBHY#=^1q%jz!sNQ%yL0n4y7-xM}q{#r$H_4py0fn>vX`<2^h`|)n?-cA6s--
zC}2BifqK^pwcbIPXzT6O=#ML$ndN|}(5QIDESP?=-`oi5D1mlg1CyDvo35p`plvea
zND5#N!*n?W>aTkFnQy*Kl`?aQj;a1ax}4Gk#OE~hs<M<kO;OQ@_{y;xL0;Ptm_)nT
zt$Ko5GZsEpkitrvv7jzeG>4jVGDq*7=NzC-u4JV%!i3WO60E3aYBdHf550|YFUq=r
zBh5?-WCyEl2h5j^gN(3+IW26go)x?^9zx4AL{zaVNFdWtkolwi)P(fnEE5MkSdb2F
zQ$k+v^yDTcA=DuQ*F<X6aprZ9kC@vv=t+9a*~H43&C=8k=Cq0R%18$W9^PJkCe5fk
zG!>5~Q&Z6taS1#VD@a?td*ftKfAP`+)f!S9aQ8rklB)J=aW}<WhNim<SqTU`V=ImX
z%O@afZ$Z7~ZH(~~RK%)~eqQ$mW~IY0iL|yA0!eW@5FTxy=t0`tu8P)QrerG7K+2$r
zMs1$7BEDxv4{ijGRw3Dj(w$EZ6qJ%lL%7zI=ncQPd=V{ntW_mGr3miqnsNV^2%={1
z3cJI(z1l|_(x9CN!C6^a=~X70t09Jd!1PxDjSbmUV`1a!qss+dMqcbGoI{O>c^#TQ
zs6XjWAS&^CBjS8?^;-P!JuD9}oJG^BqB(_|GK`p<4``d^&D@gI)mf0WY3MnleU26<
zs+u4>G#RU(<ug|n>u5j+8I>iveWV%%QPxQLZ=hfbr!~5R;XvsanyL4eD@Wdm=m`Yl
zTbb<H$%zy8Rz{usiqx}$!XLmdR<Rqj&5QasndKp~x0CdY6N3<rb3?(6U8at%jWO?!
zty(kdY%V|}&`{#USkJEG(@IpGuXDCl9lE=jB{NaDVKzE8wZh-6oN$-rMd#{_;XNLC
zP}VNa&2B4rpQ%ZdJHgjM#=H)*{G6KxN1tTC5XN}g6SPPvg-jfn=~_MbxXou4HgM`Q
zNOU9|Rh5l;$u*GQBboTc5gLvip?Y$j)3hoTuUDFq1hQNcg%QCG=C#|aqT)w77hkDK
zD~?8e@TQ$DVr85pt|R(0lB3l*<Hu<3(#oeguT6tGAga(Yx(n22{byGhP2-yA@(0c}
z%C}Xat5j&n12w@g40nofM|3c&@H4C}A^oxL4KyA5@Nr$hQ^)YMsDci)19#ns&RYw;
zwC40ikaqk3k@x1|Y_0wOZz#6XDHOZf(ne6)uxlu#hPJ4osHxgS=xz-)k1<4o+O4r_
zDk+MZXK76#M6|SMC>5l{JQFdGG3>KE&-q@zXa8}2=eo{6=Q__>f5gg?WZm~#_q6W&
z^M1e78Hrhr1rH_>jQstzC5HM@i5w@$4#o6L7Mv0e^ouFB9!fe}++q{H+>roNJ~J7-
z>W>PlIlAGk^G)gMf6Bc73FdbYxfNYJ(!iQBbrf6J4TspK^>3SxCK1+mfm-IgwmxH&
zfmBs?Y_(njrWwD_mKLTk5&0bi50m1NvP{e2wwRVZ%n<4B^Ez>w;^rRCq*2WueFQ<-
z+6<k}<S=zo%GB#TPW{+bX&4Zodct9&EZ#(>FuGY?6hHzE`*{mLNAF{>m;*HX!c0_D
za+jBlua%o3e>1byU9CnKSjKI$lWR1v8+BPESrN;|dw>FE6CK{7kn&g~is3^(eQ`ey
z(4A6^-?9$WM=~xIu@c2=5I3y%wp}EOU061<VZG!b$3Eut+IB=C?pKN1xom`AA8+v}
zE_k#o*9~V~&}w}PGh#F@LuiN9d41obl>ue$8A|OhBTMlHC0FZ<_Q;2`3EtBi=Of3N
z071s$``-E#?pM0Kp$==~Wr8k|9Cr`#l;KL0M!27<6pq;)+zUH0bQ!K5IDC80djzS&
z0({)opmyi?DG7GdU-oL^g@*w*#8&(ZVaHjTjZ{=oB{u1X1IsLy_N3Rn70<3@K&B4h
zGk1>x&4vE^8~BV)0BeEt-jPl$z$Ir4AGUBu6qdL%r4JfgdI4UC-(mly1TDDd?4Wqx
z!(GTsB)0#@-eMsjLqxU!l??2*xFa!40WNmh_Ui60;69QRPQ=$B0Y5v~u*;!cue&wY
zOIRqt3B98J$t^`XyBJRAcL!>U5VN(6e#Vi4^mm*ZQ?kzvPMX0DzAu}9>O|OD+bt+~
zEv$QBjn6dU-ypaJ;G={rO6F-XJ1dfRXh2s{-X-}6S))_lM{oB<2C!ltGIyt)eR3np
z4D9M%x2J=z7x(RK(@U98PgKLW%*3?~Yex4L_hPhmShHQKsE>c?iTb{;x)Awgj?fuM
z>**iZ07T!j8+2Stf}W6NL6024!o0B<0!y^p-A23JSs>)i1`|9XPI)zD2daB@hX7zo
z*3Vk4<(zYY=Yp7Dw)nD7<5IB9+te6F+25e*%7!`VYOjjoeg0Y}N#6kU?g2kZS5Az%
zk@>ZaZ?2}&4cE0C6bF<x8uAJ>9teO6+=-E0aP1RncBAuKGzO|xEbQ;hm_8R{s0Hez
zRoc~|W)%5TQ~^Qb<oZ4`1UFiJ3zvyF>qL;P2|Q}EPe?mEBzWvNtN}hoN9NLXj$N6h
zZZrcK<Vb+6A|jK1t!#9b<OZ~=>u;167PTr0;`-bv<-rCnVn$GGD(f@Y)mhrcb;~aG
zQEJ(I&{$in;q97=ygr#C^7y)NBn_cbccdPDV=xnGw_Gtjzb9ltzIfnr@NM3c7W)ac
zwZdZKYoD#O?-Aoy17@hpfJ%E*LAy-~M;Z<}mIR3`;tTJIbf7i+8x&Lljfjq$#4bfA
z5hUqlK+#JE1PkQaP(AR1^+-HH{O1EFg5Mh~`cVQz=I-gibXhAv*UNCNIB+utTGev~
zwkhj1hv&R_zHOscH>S*pfKl1j&Nu3{b#nxKx7gcSyU*Q{wcieb04z~sLZO_+Kzc?0
zeIk8I5z(_rv1<U_ot~6gRa&e1vR*z5+z<N8qn<?q64s_*sI`Nd+~d=OOyg~A-2G&Z
z329eTckuIF&wJes`n!gJvV9A16|dh3rSG^pFIPZ3h0E3eBZuZU%(|Hq)wB3S!gsl$
z<hefcp_&dSrRLU!6A?RW>OC_GDU$1u!M4VFuiM2b2NeLvuk{Zt?N)r%g*1LuhXf%L
z^~Rw0%Ih5PKu^Hl4gEf{3)Wd<eZE}LY<g{p9DZ_7Cqcv2XU(~=lXg`+=|)~@%Hq6E
z37V1F1o$GHl&%9SmHQz$p9O{i3>a+$(r@Kg@?bH^OE`vsj~Ir?$u0aQEn$G!L0m-v
z$)c3D1D;}+`LF#AvbORYg*XwIL6t!Bfh|e!rl_B<VanuS8u!rT>uXU(Ni~0iWG~ij
z$2=VDXT4pDU+>+|4+`nm`ju?B8{jUHgSIhd_gPkN!~|uCJkZj%M8YbQYGWSKWdfGw
zH9WICmuH!?oSmj^$%cTX_AP%|Unk%u?4UZpNA1RRjiu2=H-_ATBoghy1)sx)q(DXh
zo8PdZ<dKs2qq|kF@v*yMwjUxx88(qAPG-=yi7~(}+7JdOzsHKjD+Wh9P?|FlIwijT
zu3g()T9_6mF*kyN^faTvNhyzUJqt5DGOFPiDh+3(mMvH8ZLNI-)ZaKt`5KngXrc(O
zlNctuk6tfKN6}iYkEhCX7WI8A@!eS)=-JozGcZ)R5K%X5AyWt6Yb<OC49iB*tUf~<
z8m%=%WeeJPtA$hCk>4BiihDO2e^^@w?nQ;6%v8&0ZipTKwZ;vZk|0&qz!L~kK#oic
z7r$LRDnh_UK&U@I6!$IH<Di`?hEa-V&N<WPw+=NcfZ<82s@<N~D3Eac>4M^TqBelX
zB5qJ^ZpP$4C(Kp``?aN(?MU|Q@0EEOLW4#CqG7q_a2hNFkz{s_mI*E%LCRPKR?;i&
zwf^O3Fn4mx0!h&R^*)gtk2+6ijfW=4!1plOvI|6yb#ixV%tT{GAonl~HHpXJ2x~l%
zNy&62j{$<+w;X_=trCD0lj+l{ZZI99Dz0=m5%L&OX#dyM)Y{E+_X#k-tom`#Gn10m
z&$r`6Qa67uYqt;xuv?pLa#BhK*Jh|4;xA2ER406|S6-a_D2Av6!ga@#bPsCxhZLs^
z9G*bn`~AD9Dkp-2sK#mJ=n>&()Ho;`2Mo4o=YbN<CJ?i<%1Y4EE>C|JGgote%~;X8
z5$H&e{s$32iq|*2XMZVGljc`%7pAN)ZBQV3P#uSUH0zsaqKRrFN3VNqcvJjKtu9p|
z6K*-N8HFKSm>$-po;js>zPT6dNC|guyINX@v~VAF9#Ap<q=-nEid-EfmmGtrW3VLl
zZ2D7IY=e*j(8@Is>9~icN(K{?v?ixL9CLWI&8cQoH3@KKMDOlla^BFry_Sgf$+PRg
z{f$MsP|_#w0HO*9!gV-qq&TuQ)|c|gNhyURK@3<+-3|O|)!!DCmgH;q`6lYAlakxA
z)-RGD(CcP296mAcyde#^i+kIg!f@rFcTkzvv;5`<%89$EotJ~;JYWEe!>K>BsP}Uq
zFW084yQ+|mU5U<MI0SkOdjb8^5~C#)@L9YOS(-A+NFHi0repS(`x$;Qs$D%#v&Cy9
z#V67lbaGtR$*1oeAp+-?1n5dTCm(LEGGy+0)bBDD@7&nZMmE;WC?Z<F0XBYs#{V8$
z_JR^QlNmQ8TM*v4y5FP-j|8x5CHqkkRb$E-DciHO*p}3o`9Rjn@-fancYLPRt?%Y=
zoQBc^>O)FpD;hQ9dGlIv`Y3zb%w<ZH&2cPIsI;Zg@e5`CiMRHDJgTz&*3m$|90L81
zL;qd7rZz5kr4~@(KXP^U=2p}i2*#bpU#AW#<1AwQ5+gudt%=hfNC7YyboQd?*EXqI
zdGOO7r=s~QOo2-Y-*;*nGw%Mlm9DW{Om}&UiN2#lEV&H^tEsUKqdtC5F_U)XYX2~J
zBL?c426ywZk5N!qTA?PtSKwomM%D$5%?lpCd(}@wwpi^lqAEcAk$&oJp*x}WckHV4
zhUN8q<jzj&{ghHD7(18p{I2HeFNg5n=z<IQ)EFl*@2A~PDCy|y8T!DDwR~W3o%pch
zbu!&R@z&dnT(`^Tzvi$0elfar(DjD>+2NDMuGt1>=s|)-MrA)7r!J^RBb6+vZh5K=
zZJjFT&c(L?;8K@UGFA}1SFg^V)s^H+BAfckSslEz`!1*d0I4_6H;;27B=uiWpFsev
z>tqosQ)V=$7640+%7T3U2C?P%JR-f8_4qmXatUgMl3s|71ni%?>jYCBTkl-IJtRgm
zU7RfZQeIrVj6l?1-5fggc1B&&IP%@2RUDU%Ui=(-!6Fr%L;;ZU;FNRo1tM8xBj+0}
zcmE<9z|3Y%(DxsB0@Ehczd>T8!?&~BC1Brz;LLfeL(REz3m0kV!q+|@^RY@9tnmHc
zdlRv$^HGb<g*T;5;g5xHIb=J&y|D^TJ(IU|&OJroH>$TgC*Q}3_~PUG_RAqa^z;>K
z8KmOXIWMZ|jNP329muFxL9o^5&#kQ9Hf1qa>NAyGnF+&sb}s_Kd`j@jv`IpY!2zF^
zf00jZLh7i|dw=U1#3gJq50eL(@)Xe<<qgZxx7G8`ykijLdIxbOQO2wUBChO=>f;am
z78oPWZ4AWU{Q1KiAz`)FRsMwGYS*?_>8=^164~WFsc4MpwspK8TH_$qQlwp8?2=)i
zFSd7&&(Tt}n${Y%IuM9uvq*9JTw?FLUu^U`oT{^}lXE8hj(ONDY2pmrKT^xCAq*K`
zi*_=2xMut!wNQ|me!tv4Y^~57lC@RnT4(wCRhINUq8jH71ypS!OwjN`c!tYc9j!Hl
zN5tEKE84pS%j1NMra<webJr_XBPcGf6D7h79;d9Bg)i~-WjVGqilh%Z_Vc94?YWab
zs4G6vZ;U>RXOxdtXCagD7vA@dieO<f?;|aKGyLtG&bmaR=XLZ+a8s%yut=-6a60Sh
z{S(e#&wDUgZ_f7=b!0!2|0%9`R@(H!pQmo<#rE9y<W@4BNT?hy){VwWz(Z8kfT&gy
z<ZE>is-GxarV`C{;aleyV5LM}%2^vGk!>1z#fT_nh3;^0f>qw%XYD=`S=56}aYxnk
z57*C-FrURVcRRa`@RTCqc|vm)mx>^DDbUlge5u4Qy1qwu1TZ_Z`6Q!V^jsrULAP*9
z=bJNMd%}PvSyeRa+DATN--PIM78KUcpZqd{ZU(liGwAW|@=5Apz<Pa0wkdiSeKQ`6
zHV8R%bxisjw3555L#W?v)DLzd>hb>#;4j|q=@QlmZEG9Sur852f!%j6WQxG2oTh$*
zP!mj}QQAM#VsiU!xZ7=tGPXZ=ceHtHZol~{2>n*?yCCKoBsyq)xQ~CfD<OLA<4|0l
zs#35@9i{kvJiwVeM4T=<-J_uZMCl3I!yW0&)rg{?`pONCJov5_F?%*udO#>R_Lr#W
z*Z#8`DmS6k@~PuONrvL$coV^5h`hS?v<Nun@>ci!3RThPc|;1`*+g!b5L@mr*^gsI
zNRgi6KZ9XE0IclZX>4`jgF+wwpE+_bm1_`QW`jby>y+D4PVkN14f$?jMD&a2w-Rv1
ze;FnG#ls8lTl_Wmrs8V0nu)*<@0xPDcv;)Ql{55H{KAp$J$)7<#YHIMa;N!j)I?1w
z+|fff5|%rP`-x+>ykmQ~^(xD$tdoh}pceW!i~kyef4YR&gs4cA+mx8D0$3YP#IlBl
ztMP`LLH#~mWosgtzH#e~Y?Zrdk(9AqW>$b<Hnrc@+VkM|&h*lT*U)yGKQ^?HDHQ`V
z`d{~ce?DBo@3cfI|E&?;d*!VDtw%q9ziVFu_r0^R<E~!QT_8BoIB>54Eor94r=@i(
zEaftnzCH&xOsBw}rsi8RGLsFoTjONL!CH3;566w_LfkWpp*~XvmOyu*J;{IYC%tje
zAgaI|bzH_=L{h<0E4Rksr-B|dML~bW1lrMMAu94Aso*=6&r5;p@w2+06XNlFOsBGh
z_K~__w|XO6=SOcI=HXem)ENx*lvG)-+xa(>?-GI@w?XfDj=s!KSW9-OYjCK~(Sr)6
zo`Y`4P*>k8o~)#L_x(OvF8}oE@8t$1;B9mBu|m4yO`BI;ZTSVx!uEdxkNa~64@dW|
z)D6g1^b<T6ck*3}X;$9UT(s3wR~6^-iEUi-i-V$b&XmI4(~DoDdh>|qlN2X(UKOTD
zm*W>u_zc7QyLO+9=ZbHZC3(rvZkD-UtfMzighv-wlU@}%hc(XV7|T@ETB)#hymr1l
z%lnuJub+mGD&d<Q#UQZjV5Mf`1LNkco4*j+uXsZGG;?mGtmDKAFg-f+RrLC@>IQVm
zWhi_pe$F193GRB8Yd2<=B68lW`pveASReRdTaAt+$M?gVqmIQcTz7EJKl!c2hssBa
z0wFd@jgk@dsr?5vBe$RDO3Rcm<fJ5<tyNa!<!3OcMe6ER!y2i45*WPr7*TgEiV(pV
zwfo86`sZi4P5sxEW)5@Y=(_EdolHBK<Ll0Av42Y2_y*om=?eDBEXR0t+eU5FUW8E1
z-q~e=9ZPe;yVbNi>Z=TlLd$(ab5uOo)E#O73AO=1qh+m5qJJ4$z~62zyVD5vZU&E4
z*LRCH4~Drr_NvxnZEaLN=xZAOI|5uU`0x=Zoa7(4N0HK26OJwMS*>wv=NvGx%!H&X
z?}s%)D~vDRD|dXmBXHuLhOnv$6BjKEQ8e4AYd>#7>>R`0bL})!u<H7Zd)d}^{LRn8
zsjz8ZjEPB^@yW&uvBje&F4Ijvj?9DoQzN0;S%UMq)GVQ)IXGCzH@q#t=yKSboUZF8
zL{dyor-4LCLR<QP%${hO;@=>I&h*LtX<OxYsA5;?rxy5ypc@$y4q0aj=GrY?$*b|6
zOXt4K6=}V7U-K<9tk1q_B2T2t^U3rly(CXdCgQf#+IsB)&L`>E_F`nsFypK5?utg~
zUuw!JuZ#0rx_l951W#<LkLMal{-ODSyC|pzua56>1uUyy4ayfCOgB<CH#z2nBft8&
zB1fwixv6hR=ykbG0OV`h|6}-2EvU(H{Y+|_8-a`jTxc|@AESTm?D<9)zPIcG<nfNb
ztZxfM)YDr3&~@^#FvcOTuoZoFoyDCzu<JPOcDwNX@>i8T05+%~tF*?I?fF30Nm9Lm
z4VYgfxPwsvcqF3!In)raVwij<%vGQ`otc@b%iUc?E$I<X8DRuHMy(&aa}q$4tuxfs
zn&Nr^(r7cD)I_!Zv&faP!kBFX9!-GVC=#|j12E$U9I(O)6!+YW0u!3J3N+|tV_bRU
zDv~Q>0#V*s<ZKUk!%3&O&1-LY?bz8awQ(kIleX>C@RT_V)Yj;m;jJC7A<h~<;<lSa
zMe$?I_U5~@NbgSXK@Px-*!q$H*lS=2Y#~n{VGCrgGCtpd)$QLvj8`grJR1T!nMVdh
z@im}JM<ZEY)WipFkPY?8>3FvWO+Fx6gEvI<5{S+1!!?7D3cy)WkkP`~7|QtSA0lLZ
zmA06n33Gi(+6p0jn<j2fuYfeg<CO##>YHlBn?Qm_=Y5kIUp3EgHohW%8v}g0X;C2+
z0nzw4Hk4uX6Aq9J$Yucq`UISax13TIl5AKbgv7DOfymu#ZVl#api_-ccu^n#NwjM7
z2M0T%8AWw8vADTrMFsTrWto}5Nb&ND5dZ{^qb_k0mdT|NB_13g1pW*Mo2_+i@<O`D
z5pkn51i-x@3fWK-jZZjf<pj4e5Dv&F``B;RPp-W`%I{IKF0pxV5<|oRALfvw%K(D)
z-O>cUC%XYs*#*I9Q7?|51OA1;<zgT9Kd7(u2Layzfyntyz=sidr0Y>K_gRcfN4c^B
zvBTR5U%g>VWGJPLUA%4`nf35*kZX!9rzA2UL<j_e@99SsePlaI29ScrC&vR_K(aJs
z8_0??-Zzi(@3cnkh6vfn0vBrnu%HG`4(xn5?s1+AaHejc02W4UAWRMBw=VY<5SuNG
zR$dcw2}ld*UwI-qtbz<eK-W^%2NC5~H6S5t>0|EztKDJ{5<PR2|3`q0QGf6mj#^%<
z_F)K+26AIJ2M8p{_=;)bf5$%3#0`|3JH!5kypqAe&9zsP{5I0|nh;@Dt~l%TaRGmP
zS)+#e6en;-f;Djv6at*s3Usto30$kgJ<T&GuF4QdjcSsxZ$AMSwhRpIxv)--Zn?$g
zCVL9oM^_Li3TME+&bg9|&9%+~!)yUZ(ci8-2fqJ58;7Xe+I|=)?tk9@xBX?)!&Wqm
zo?-v+kBpf%O37`_fT|LD+g^GN2sSln7hLB19#MZ7aA3}-06K=R?F`!t47HG6u8xHt
zocLMlWpcS4rZB7DlLD{x=_v6}c!d$jvrk;)px|Tw@GrO?x`~t>>iv^xTdpWSF$+zJ
z^U&AB>3FHAhg9vr>wc!~y|lH}m+KA@ap(NlgDqLg>+%%ZnZOs>PwFcawUw$Y8!hYS
z?u%sH9hB0_;J#xTV_bMT2~o|P-PFR{Q<$bw`7ruXnqqF=yS4MZZ~R`%?HT8VBRsFl
zbShs(qIyF_Iz6bXqQ7Un(oeaI%yS}$6K15%!n{O>Oj3j+I+EAudQKGEur%>Z-O4_x
z?o*VX^VhGKzdMT#42+lh36J>V4^BOA6jtrn)L-qcTnX7RFpVwMuGz0NjyY5{7tw|~
zipj6d161^lupt&B<8Khh15_}l7~gfFvrwF5F<<Npu+W#%273&znWB`mJk1Qoi-UaF
zi<$Sba4o!^XIeIZJKxndXDM?2<>0$}I@KE#wZ3x&<_XF|$4t(>L#_N~{)ta}8g0oO
z94u97;UgTLtH>Wtq^20eUfDJeYdp>6rvOc@R%7b6wT0~U9TPn0&UmLtHzK$y<mcWh
ze!Z1$CD@+azxAdx=Mr69oD0!f#J3?Pa|^V!wY+zELd<PCH$~JXeZ;QF$>6qQk?2Vu
zV1c#k6%*&HIVCXdP?GdjM_u@DP~yAkE_|ViRpVa3%xN(ml2%E+nL_b<>AzlobPlLZ
zIK%2r0_RC&-%g3Y(h;eakKtBXy1#AbN`8RhhxPKLpW+;}5XQ`LO6M4(LenmKb`F2D
zzufG46{5&@+%O?V<2N&i7|rDN?D|QqBFn{MjEb$4S*F`$xVTG;qq}l4;)U7m+UNH!
z_SQ0N9VixH(PI<G_*|rDZv=lnxodkckgb(#{w|{8XY2bI{VNG$`(m)xhxFp^*Az>*
z{t1eIN67AQUBvIjU)f<<d|mIRt#A3xKe*}*`R;h4qUU&+NCHmpgw(G&=-g_?L_0^d
zxc-Evi9nFGRB*xLSYEB>t8adOr5m}7UMRbO!%_?T;eXP8*DCrnL%4pr%|s-O9%h?Y
zTQ<uFh(7d0B0sQV;_npQ{br?IOg2t#x%so}>x>E|GFI%dNiUNB8}fWuZjLE5C57*P
ztl!;BoOJbY)5sX*zbxX}%^o=1cFDQ#nI|DX<91ihK-Q6%#|M`m7L)IO{W`JW%r&~X
z^lMCD!&M{dj3T?mMMHVTHpZ3m!tV*X5>w~+olJJ>GdW`PcO;7>{mhPir7@jzPRSO(
z+O($%AabqFFO>bH)35}jU!H&W%YBHkLBqa-uzpM)>|1i8EBZH)M)GM;K+d=G2~K)Z
z4}~q0)V@IO*r#>2>c<)r;5l+|aLPK0v*}Rq{Y+=1yAgiqbQNN$Uo`hNnk=HAUs;;x
zEv99f+<dCAqcc}Z$G)tm`}L-p1^y!EZz&*<NVRzt?0ZL^VUg)%``cXcYW1=ScwFeP
z{{5cp-URKeQ$m!c2&O}`b2`b=X^9Xv$@=i4)1dkskNmq{M2=}cv23|Aax(T*>Ex`=
zvyv?%loxH>bN!7%VzTi;)ktI78;hbYRq=P|U7z$vQ@204I=R$H-pNzwD5NM>EIb@f
zJ)4_q2p8^@uuqihrIdQzu;#t+#zo<t#8<<CM`M_w7M&6s@DFV|_V6Cz)U(|t5U6I>
zGII*_;9gdIM7`YL>ANjFn4(Wot$gwoF)8|R7Ie9Y!!}XB6N*mZi?-${zt^@Qay8O+
zM)I<L<FN~~O%by<g;Q$w1o(rV<1|!q*nW;T>txrPN>pT{WizdmrHJL?U$vBd_hT<W
zFXzNfT#(Sy(n?>u(=Lub>0j7mrKqlhF)Z$Xs_DY*vn?lg{r68+G>C0-zZNR>y<2f%
znv3hR`wNPtAxtddYI*`B8v&=xl8LR!Rb_sTZr@AHlZ`DL4vA%cL}=mm+XPGa+hF(n
zNC-sFqc?ivN-wUxR4=8#H_?+E<Brv`APxto1EEn@AqY=9@xYVM<AvY~#uI)gmTi|u
z8qc4SfHWnl2&i4pAzju@DUM9;Y}OMtMSO=$XgAd!NGM;2<_09U`~ES^U-^1g)k3B)
ziVi7utr^-D^g5PPa~@HR?_c^pL|iuxh-rvegzgAGvXd5DkaLCF6+kAe3`*kGt{9rX
zI$nOCelnsVNjPn5+Raj<tjNAv86uf*;T4~?tLxPlFrM4HXW{3P8?DL8)U?vp>%KS3
z-O)D%Zpp3eh+D0x-vk#AXPUk<nO`!9x$HDNe5aG}KwsX`wq{cSCca{9eL+!;yQB5R
z?N?^a>)sb%S9DOmS@h~u|8UB8`g(A)y8V+sK4AC3pE_?bNS&(#U|PF?nbUyRufpgp
z!+wSWJ$3p|)N#*~unHZMKLv~6x`zXL&$}$H8rhef#YBX2?7#BGI~I1ME`PE#Ww$Mi
zPAz#<Z+UHKS{s>Mae05;G%@kbJrcOoedTVWBeu9OI$=6#ee&6WbpZwa#Z4&smRFSA
z`BiM^MR|h-H6>-o72Q}P;rQgg+OqXKw3_3l0~H^&9~b_(r82kKNO#i$jDuK4FQ*e-
z1Ptnak3lY!tbHGD&8U1z=6P}IR+!ZrhW)Bk0WV5RXFTOxazTDxOYLUtVwPq!%=i_T
zyg;lIS-jKL$9#!IlhJFMN)G$@;Zt-qzlH!Q_HnHg)z(@e64s1uE4^Tbn7b{-4JGEu
z`fv6|5O-JP0qKT$Tw08EQi`pl3a=o$b6#mG-U6{wA?;JJrjv}(`*j^2UEFEmH5HjK
z>a^pjo>uUgT2I^0>4^7TvoN;FV%=+dlJqbhQDmeoVK9#!xB`*rMKFPez)%`y@Mip+
z6Wnt>vs`?<&r#rUN}-(Wo3@`~)ol?>i)RzgiYD!tKBYy<tN38ZiQjvuCKI_zHz6f|
zgQivX%8MIZ!l{mC)`eP<Q=fg=S=!Kd{_Xn1MrkEtFy*UeW41avI=7Y{K83b$eT(Gm
zS;qeSle`meI<SOrAm6H$Vs#?2rSouTRQ-#&ny)HW&K33IzhdPm!x~y^bUD|~c~~u?
zWU?sLz;r=QCa2e*DPm;*5U(K&6?ai4kuvHO2Y;F{`;gle4k`P570coB*98W|mfifk
zVALSr4RUY|^1hnHXi-J8a{x`OFph55P@R$YS-c;8da(3P<n_z)Ts+RCmC}f0Vs@QO
z2BNV>qg>TuCHwk~y5bB6t{+kRs%Tz5VZ^EU#%hIew=y-cr7MKQUVFGmdYJqjT;x^5
z7hG2@Vx349C3OM|Mf1<(x*u`~-u3Mzmirxw69XjfRmS!`%$_CBe?8@usCUkhNn2-p
zsQIzY@ax&bUK8`E$JUcx3$&$S2%$;PL<V!Kmvpfi=2cG(e`urr<b%sU28aKCb93d=
zKx*i?Sk?Zh?xf|4vFJxhk9uQccehW9v{0LJ0$VLIUzqV3G$cdaqRwxG5my)*gg7Zv
zp9!nR%lX7znbC+V=Z}7h575@p7p*@5;=1u37W>S&@17sJ6|iVyWPUjRY80S6Z9k`3
z6B#moXH)&Rgn(s53ZBx{)z{>B5hp-*wj=oYcAeY}eeqF&vB0}D0%FgN7a?Z6Z;Z!D
z=au5<fPN705y|uub-(vs3$^@-KLZ)C!)_=A#Ja`Cb~5ur#oiMtFmfzlIGe6~ZhbOK
z$VQ!{CSp=D6Ipq+cxixmh;4eVWq)sedAB}bPRK&bd~JGjykhJ*9GsDdH}Mk<*dIp`
zvmIO+oi`=B`SfV@M(FJ|D!wNuUZ8&#n><~!*O1tD%qR@|iaER0`}H1@KWcNlFYfHV
z2nn@By&e3Jfk<Cs6U@UqP$-%f74%q&>t_>atl{ZSyr}y$qdU9KbX$A51{p*z|KLc)
z6AXhBGczN^@O6IHqh!h@C4a$}Xh1`FU9l2T{8(l7Be=dR(u3OGO5Ww{A4r%D+6Pu=
zVooQg?kex!8NM+xyo<U0aQ4)#y{>R|8x4`q>6o>FcKHFqJnoFn@<qkz4py`^GL&H#
z=n4qq>HaV3%a_v8z*1;fbv2$yzq!j^9^2qhHN~SF{ifj-3lsEbp2hcp9h>oq<IwmL
zL;nE(%u`tc-(yIXOYQO_Y|yF+^p@MIJd@LF6iT@060gjz6d?4NQRIsmQjRXG_7g8{
z2&x0Hj7_no>Oc_aZYU!)-!Z9-u?cEdaW2{xI*A#JVhpx-dJ{q}MccTk|7qdcZ+$7y
z7n8l&I&~Z(2D`ui708Lma6_1dg<xC8)k?LL&F1IgQerX~Y~7wv*M({A0UScb=0h&!
zE;ZwQ*RO-uj)L=8APh=Um#1Ayjuz?6{Pjn<7pww-YB~EH85Aq&=&{3UnDj%AVE2VI
z0D30KBI`M_U~UZSFf~4XjyUWunc-Nc3P@d-3XqCwn2{{Sbmsxs<N}X5T-47pE%S1I
z85MX4xR|C1F+Qad=uP`5fXuLD*a510gTl5wUTgf@%>~)FFZ+>mMPFUXQ^_M_aZU+%
z0E<ZZt`Hr3at~L_DH#4i(GB-6r{yB?^1M0q!!EZ~9IUj_krHe5f<wol;8Ag5B(@3f
zUG+_I=aQrcFmbcJ${3R%Zqo7*m7Egks=Ub!{;P7b_kvHUY(BrM%UDI$P_D&Wb|nWO
zu)Sw9kEwsk?d10m7X35YXM!7KuBC%MHwP)sM)HlNj|i_%_O;r`iirG~?!v5u?IP;0
z?cY|-)zY!IqfAV$MV1y9M?};I-Nd{wdcQGcow!VmwNW3hUjz)nIHoa`H9wvJ=$7(^
zXyncKNR&AO;WW4>0P3!imsdx7O!m!eZORa|w1zQjqm#5kK9rg=&W3M+xyGvhjI?bW
z9*{Y_=+MUJ*u7UJPphPNt!tsE=|5qu_|t{CRScRnlPh_z%*_uJjSQ^aZ4A;ZarUOP
z&Q}TW>9RJ|3^LUZcq!`vPwUK3J(Nm^t9U<J?MW3>a{A#Q<_=J8+$kmdn!|l6@gfbR
zdJ^$3_>YQWgu4}Mgd4i)+-evYO<CLGDIc|~W_AWPS{=!GGZP~S-}hA4IJGh^_<LwP
zTimTzm`w5wa7gueAJpI;7c)}ch}m)5_2(@k`*x-<{Os9siLqPoK)-+&7wV~dN{q}1
zX{c}J2Rob3w?6Ahz6nDIpbszS`C4j`eU~F7eFzZ|I{^mt2mMG3i~D<T`RiIElha4a
zgLQv2K8B2DmUn%W>aJy!6X&!G_!rgG>Tkv%y{gNci$|doIL4bMg#|e|wSY~|?d<Hl
z8h1tme|8W9m%IP|&e9>WQExO(ifT5_*qVVYrm#>Sk5S3<Tbr9_#1F748zUbQU;!zD
z8HsCHkxv4yp9I{e0M&i4roQg$uD&Vo_rgfV2)7B?V<uQIQ{(HrN}om7Q;hq@&hDiI
zy(_in7&InxU<6VJuggae8L^jc$AdE(qTBJJZjkKaY=qM`U8VvQQ6)c_C7MY`sy!(t
z0KUu_qSopmB>?IIV_b<#?83K6RLt78D>Ii@2pwl~=P6yo%qf{HID@$|OG7VnVE7r-
z`nwu$HB8;es@a$q5^py}MWq4*1GMg`qG9;YXh!WBM>$x4pD5tB0{~<desRlDjn-&%
z`j00^+H4dp6p&|wx@C?-dUw_s;D_zW!%ff8^Zh~wHXc`jyCO3FDwu=v4(PqGM>sSZ
z01YhVF;qcqt&h(wGuNWtG)@nu0ca#Y_C$!-7&@*;y4XJXfr9*$&eLIjr(b742{J08
zr!Buh$}R671c6HYUd*Al99uhAeP1)|aR#11P|um;;`K!<ThzF#ICy)e<sf(W>m4uZ
z4~)typyQ}HJf5+llM41@Y={za0e{J)#(7#CG@>@s#F4SJ?g!M#skHvXbpN$lVQ}V2
z18X?z%R*#&GTn8KC+32Ip>YP8(mJsK7EKt1B_CAmEOTA}h;LYxlMeMXrial>hb+_@
z3n(3Ze2jW`y8Q}w1ZzY6!jA8LDtB=aaXzDC2eZTzBP=AhmPu&u>XwToE?Q49W-B|+
zCg^TXU{sZXI4uG=&nvHj-YQEk4S}vg)P-3=A(pnqDMw5+IcjAG;8zhF#hzPWjm9~A
zmzzb{;bPm)I8w>ha6j)}bJQ*ZaB%$y?a?j|!zgTS1_Z}xd6ku?xzCf!SM<cH4uI>f
z9n8VN|2muM^YDFi72pA#i=tyoKCA9)%E<_y6Cv*M_CJZL8y0}MZEk4l<$P?@3O;My
z_{1%_pv%ZM__A|X_@#b?1ZB~Dx2(pY3Y`DGZ9{G>SZM8}^`3xCbSfs;Ll2CliTt+V
zJp%!>fYO5tnb?qmu!fs{(dm`+wo?~bjMjGMepcpHgv=rN33oLE!kA^Y)(QYi&AvY>
zW_ThZ`tlpp)d9ct+esEWBLn+#8(j^Y&}e+himV$pgeJW-P>mjFIg=ocnhT1zDjud8
z1I7js8%wlG8%cm!XL{2@_Q#NGr>`B&BNU<*$X&g>#k%RlQ=B|e;)cX-{o>PuDay~F
zJ*c#!8s%<l7>QY)Z;j(h=Y7GT<>yfdg<zUh4;^nuZ}yjhsdP{}0p6klVF-6*u?tiu
zj11ZXU=q?w4eh!%$}PLf33LxJzZaP&RarYI-)*j%=P9Gfi`y$R+|(DM?y^eH37uwP
zqaq-H9<D%(v!>q`Z~5${d7?_~CWA5s271V3`Z}lW8|uc1t2>hK&Zi$(lhgWz_X#X4
zmhF6oA&MF+tHdwAv>8V{*s5NmPbNUTHy%8=!P9u!B!l!f=pMm8vji=UuV1EJ(n)3c
zbl2_r@X{wy1H&GA@nBRuqaW_U<gl<9DsrV@cGykRQIqQ|RSxemm&Oc?roucupXfD-
z|8CPh0#VtQb!L7awHb{Q2FIUs?z+3H4d|_4HAr>Mzl{0D&>pm~)IcHi5BF^A^44qo
z$Oe`E&>k(T)7H0q78zUD!QV?#5!djoe!AR1SrR?QlpCL36cF$x+P?>TFHP|1gdB}R
z**xOQ!+}+ug1L=TirDhx0Wx!YonuKTm_YDIWzcLM+|B`_gRMtK2F<bk{d5}A3UMK%
zu~jjEcM{~KdGrha=l1{831f1#hSE1pUOlxFukQhIwnE<?)gi)d9&j~*K;SIr@>h7b
zSA~Er3pwOcuXewX#jm0IrcvYIM0$8Ba_^QGYv_!=m>PH80#A*>QriXT(xN|#sN&6G
zMQhJKPL?*`Y3Gfb;ImIo$Pt5>c2vOtw%J$u#OYX#2*0YgH$_f{E7tLfxt>xlFdq~w
z=OCTV(ThK~057e@eF0{a5j-wM{Vnin{f}?+KUu#uh2GH7zwy8TH7uCjCox&xKoHcE
zht4I4sVTE<egv`Ijk&R49*(#hVIpjvl>0UboE!!`JL>3IonTww^Vn)n&V%jQCsY@1
znl_kJ6?PHK<2`qkLCWH<lG(g!r1DyA`VEXmtm;xS-#3YT)~+sNpp0dr^}>y;US^$^
zk=MI3bl2YL&cky1w}qWrI+!l&C#RnPCFj_kMd0;l`xhOT?b6Zizc*etRKTGo=SR*j
z*6&Y{j!CC~$B}Tc!2>~|e$-z<p}U_=EWN!3Qt98Da<Itkfm{CN-M3RR<#|ykn+Skq
z&tJ5*uspdM0)1=OYmHQQ<>>o9KrdG!M@VA!)JXL$I^~^q1YXG>WA2$Az!RYjF78sV
zJ(^LF7}tpP!@Pi2GA03zs(f)|sx5%|%1ErIZE)dtHD(Szb5$;kGftY5`YbFdV>~D|
z6fL4g+raIZ8B;^|2&xU7oK1dmb72G#7~)x1m{L58O~+u_00)lEamEG;i01zAM$5oq
zeztz=BP=7`D)3PKSiCG*#^%J4{mwr&c7w=xzlb^r{*vX-Em1<@v#N=7bZCX1(2wJO
z%y#}s8urEIqAuzZh#2`*QwZP?X#%Xp%`uP|Sh*ZKEE{)a=yD!W!T_~0J>0dx1w3J5
zMO<|h`}c_BqZ3@?egFtBvYdokJLXY;-yB^$IAVhW#b+@JM|vxEehIu@kQ5!Ar?~%d
zdsXJQ0{>J%MN%^=gG&ww4fOZk+`WC?$3S|`rsm-~eTYYi7wFCESmD5cy#79=(wZGu
zs$ym^*p+y#g>jj&1WU{J)_`|S`WO!(u&X0z#QilZ55OO@D;XOq%FB@5)MFn#mJPIL
zlQg=Q#WI-M+DrobemI|=AFN^DIE?)=pzQLCCnDNVfo23>J@=4_D7FcIMV#TvaI-SN
z8@t{N13Y(OlvI1}VWzy?+8)M6xiIqcE&j5mm>*Uc($i2q)dphA!|cr=Hc3NOB4BT;
zapkHzhNd=r=~9nF2kU@ycwv#rp=B6S7;VJK&~TUSfK&@<goyc6L4Yw_0fpIyurFyX
zeZi3k^HB$AB|+bKB5E4PED8XdF&uu9-{M9dwH8d&K?94>_8D#?<SR85kL5Vx!H&ag
z>tF<S0av}xCZ?9Dp=nY#Xdg5(oD*0<Xgm3ZdHuF(18?NYqWb6`ot4OB1X4g5)!(f^
ztF*p=?lM#py(H#Ja?P-$wDA{YhNWhH7cTm0rkX7=W>#zMZc|WS==^4g)yA%q?l3b3
zO7B_R4Z5Bt1Gu0k@4%WA;<p1-RVY8C#R;RTvvXWaud-0<>SQ<kBWa|GrSF)|xMNdC
zqfNvEivpQ!nXc)r$>DsIi5p<zynW12-Avs9V9#2swO6*g@%X8Bh3C)-Q^FjnKyb52
z&lm&(K{$DHxCy!26h2V8xw<tL)VlaY?@~1R4RQaDv`vr|8&}Nzj;{780KxmxKs;m!
z&@PNK&H_2!+FDv$0J#Y^jtq}+%?8Zl`FPz&jzd?d)+B@4LmT%i$=ZlD0-F7}iW7QO
z4PB2p_)r*aJCwY^`Aj&?g*0pra)LOZ))-6Nkd68L>>QK!gfe`^-=M3u_4P0GKD^!p
zWV^dMKd&VNh1%Gtf(bV3im74u<3;-866q5h_e)DW%<4v43G~Uce0u6s$YSqdY;+I6
zD-8e)?UrKknV;s5u;(G6I;qctY%EX5&I8WP&!JTOV+M1I?}&}gkpQ7)Ui9DPcsT5i
zx?g&nIzz*qp4tp6_O@D1WiE5L{S7k5>(&5-VD2HA91VAO4UZ2AM&U(hwT5PCdR`JH
zhz$5Jr2;;Gxg?UMW$Gf~u`7s*m)|7`hbr<rWUS~3z9(^Ec-xz_0mncW057w-(7U<5
zKOFo=P3KK<;tV%*4$bHVlo+3qz^31MB4SH7*Kma3lFHW)B^K*naHZdmEhJVgY(z!c
z4wCkppz&xFW4A-wb-SD*(?MO}2@6yY3dse|$l?t@9UEKt?I-vKP*Z1i4>gqhKkz9(
zcc9@<u|Jj#dbq-|w9>sHPrak2@z&64&Eo#DYg76ymY4htPw!CidshIs&1N)?-ejpp
zPfrd>ORDwnh3GD8D3!D`0QV+j>aKsJTL3EgA%&JPb`pw(2Pkt#6gAFwsaH^FaU2OU
zYFPLtPZ0*6yrMH3z!(I?K^0@ip>uF_K>U7?vCZ?SUJK$Z#Y57l(V>8lXiavtFNP0f
zd+mlM8WnV9YFJWd1pqT+X2Z#0-RR27@RHbXnvu@<GfJu*d&V{&4zFth<!&k{VMFve
zPy)U~vX>D4R>zDAic)rO4cq(<d92a_u0>*xMs4j)4}i7`qpBWvRBq@vLJP>vFC#r;
zRH>lsiIM_i40HJAZ%~A`0X<IA?+tNvOQvgTneiCrI(!8hZFIN}hG#UMSC*tNKpo7=
zGQ|n_@;lsh&Mg%DyZ~6h(#g{>eN>Tgohw0IEoeAXhV-5*{#r5zpK4cU(}R;7*pZPD
z6EKWq*#J}!4IFiYhifLYLc?u2*8sxQ0!*CAE|~p>TZeBNN5=Y$xVfQznd-v6m<hlm
zF(rl3H_V*rHGtSXISxg#=JD?YWL^4T*pPHp1(<4M0>4>Qd0>#0ulEC;<xPDWg}EAS
zJC^Rxt^~+#AfbM`v8F6lwR!;FiBL}-bBj-5wAy?yM8LlnIkQ6S@Fy`u)E)}+Ax5hU
znCP0P*ho^Q1>n(D;QE%X(gW>{zd@4EQ|lmZ+P~^q!}Sqj-bPT20inv#bpDAsd^0o6
zNL0Ns_`QP($qG<VE>V+1vk|GQJ0Lv>?c?t53$v%A3Z#bH4}N5$nNz^CXD0vk{Jg{&
zpxUfzLLKn|0Ks^k_R^eGCoLN%4H?f`l_|rLNZJB*!-7Mu<oN|QAxpXC%|Ji)n%za~
z$2ODeM|DMg)D@GR4R1$LO8kIQUG>0dK3Jw;W=7twOVuXSeUi2a?YY-CH^v<my>Br2
z?oyl<dVP99z;xdAeNb@iB`W@+KTBbUwf7kehS4IRiTOltb@H>_+wf?&<MlvMMA{=8
zZDJb?&AXhCei+p{_LC~By_B%f91}jeH#y>bgTew0AnODuOb*ug{TPe+!jK0%_tel)
zT(f^}?qM=Qj(Jp1wr18ea1c(skVr>&-I3woP9_T=Kg4`tZVIK*C(eTFMyEkW<G%0M
z>I%d(GJo6#%st}^Y}{+o?dUULQ(&G7iq1T#u+EvNFd4WLs$1{BO55eKh^&W2G*B7i
zO?<-Xo-A}I&{<8+F~?gc*Zlr+O@+8u6fY7u5dMQr1>WT8^S~hAt#QQ0{WbT{WYb}=
zt43lWn@Isuc})h0THR~WB6yV@Ho0wQkVwHAan-Y3501i**p&kBcaQ+&7cay`3zS&!
zcyn`}IlT!)&1AFt`C}_c>LbHhwqqHZKwh@Cs5JVQ7q!TOQRe;!E<o-__=ExhGY>W;
zl|Mj<dZPA?hKek;;A>hno7O8pqrAYvlAby{xYT63c4b(-P<^$r57dlQ+>b+0se{SQ
z5aGzy@(NZ6;56E+Wo}&;;KE4)qFmh{^pqux)pl`SA=O2tWM8@YpCc&t(<uQ#cR{-U
zdF;PxKeutf+C75p!9Ox3WiyzWHhGCNj&mv_Z{b4l7ly>w8tcpl=113t?c3JhAm<K2
zbNIRKRbi)L{$$PE=X_R_?>4_je+tBn@lL|L9)m6a;QF7D`9FU8A8-D>RE+q?vHyH?
z4#1Yu`$7O1FQ-JPJT3N}BVYRyK-51CI)YkuG3P+XIe>1A@PX}o<T6N!w+~o>>1LBi
z*rLr&4`~7e)J@j@$Y2OSSzW_ej|5%fIs%0r17tLsK*gpTr_phjIF_IwfbYmg1SZay
zl%FF{rIFWoUIIfsGI{zVn+fbAo94e$#PKkjxdbxQ0ZLI9bsqPsq3MXy&c@gD?|=Wl
z^C~#n`kz5p{|^$8tOfe#y$69f%(Yp|^B{o00OZef_lmSj#9N0!n=2BJ*;oGa41dj7
z+r%~T{pX?os(mTh8q2f^koi9{#bdEmz@$X5hw3p}>EQ&i$pZU)!y;OT5O>wCj%(oP
z8>>Y7>Cq-2*bCS$wZ*=Rh)7e~e{sHOdLt7b_Z{%+S{)rvudg7VXd0F+0ObGgB&yec
z6^-{}AG19u>>Upxwy~K2!>QwX@(S_sDc%sd-iEx6R~#WnzxRK=U3u(W;BNUHl$!vb
z3mPE>VlM<ll=vJ-<96816JQG6)l3cd=|2n#R;Ag~o3~g#oWRUtXol4ZWZj4V^`nw^
z)wS$}I0U2QAqopt6w{+KHypZU$)w~h6;<;q9>X0Ia{Iq2Z#;z4hQ{GQP*)bZ=_GZw
z*=X4RxX)x))5#Ief{Q*Zz<~Jm1EsSP6%zFPQBBh;|Lf%cg?l_dJN85VVUGD<-<^cc
z1lh!e{IgBqR)^XR@g_fs=YQ&V>3?P@HV`huD$GEvv#DPnQQst0H<HQgQ!O`UrPy7s
z9HKS{0-a!En5Xhq4Za@uVd$l&FD?$l2bwRd_BfRp`*dnNPw$ZdokH5BR)wY6C$D?P
zIH{*^Ue(UxusYuO+t+&?tcdz1kt$}9#oXHk=L)O0Z&TZ)O%n!j0ISx@z)S|?L&LFy
zyvl^%4@f1kf2byIn2FsjnILs8#kkgD?UTT6nUBwHf<bD%|C@4&9?!Fb!^|u7uKFi7
zlntNFted7!ZN#>tNs?u#zHv;A3xU}&Nn{k9u2~kTxa!C&O{`LMOGvhdXDp=(omAPN
zwbQ0rIMBv71+0ovXG&ZFvjZCz1sQC+xBs0%+%VZT)WS-w$;)IfF70##YaRRzD(<t;
z?jI`Y!5_PxWC8@DyJyX$9k`Mi{0H-=mz|I?+Fej`d`BAJo?p^DtshdEN)PWoo65b{
z5SlIRPDW!t1Cap>yRW*U0%Usu$Bs?RXBGj`ZZ`A5O(?izhJVw&Z<BQtJpg>>08utY
zwg7b_^Q~>3@^5-*FCj3uFN!+up}!s&)bW$tR|qY@(qRKB`>I87HI5hiSp`u0H8ope
z`z)hoPM;Hg?>SA9%r}kfG2uf~0<_8{DP!|1KBDbB7Aun~ke2>?&Ou4!t;!$UL<sBB
zt(16CIZ<ly)UO8O((ZQXznJFzigt&<ygmNIzd_kBToen-b|<lbZ|XR9cYT21XLTF!
zDu)7w?@a<5*uyOT;NSxF6{B%KR^t;Oi`b@rJwIz^mIUGPKY_FZ%vQlAU`da|P1FxR
z!@E>5N+Lfn8sE&MtaD**(dL8y&2<%08$FMCdn%UPw-Jc~KEBKLbDQ--Hoa$PIoSBu
zzmA7t8b!sA<tzh8qbm;0WR04j5I51s`{X~6{sG09a_%AZ9$q}>x^XcEbhKMOgO5p^
zIX<^{V55ukTVox(;QDJVI_OFv$^HHd=$p~oT+Woq>C}K-_uu|1O&~tEO_GulcLE}N
z#daLk%Q3aP_hyZ+CP2?>`#S#^{Yoj34EQYY{moOzG)Z|T;BlnlUH^!V*qq+iwT}7W
zYo{09X+QQGa3K;JgRlLt`$hWh|0vS3&WyV^#{{rQOJVHsQ<1v6LZv~@is{h3I=44g
z%XSqAEG@RR_Pcb6g?sB*fb<>nce*2r=TiHFxi{6Z1x2g_bnJ3`|F$m!`H+!XK155M
zZH;`4H7C;ordl`D0460CEew;AEK2d3?XVdK7HF+ALUJiaTnf>c?lb?JO7#Erf3Z(u
zgWTV;dHiQSZm-MAwmmuipNIad_Rm;#QP$3|YyKlsU|(6UkbiaPD~iP>ai;X^3Hk}!
zgDJ_YmtJp($uZ>BkHZ2&Zuv!6>A6hp<&@CtM_~E6XwTtZ0L^&QLq_5r{v>G42cA8f
z<*KU(G1G9<ZhD*UGXm+|kW=Tm^jqn@<Z1ORbF&j^70`y70pp9qhMfRc1v-b!M+<>&
z%Ke!5tN!MIYFjTM#VG8hb4x<i^Wfh~UaZJX;!kP<>xBEh=S{gIrQyZ_w|w*|q0zAy
z6(^@a{^5TPaaH@E1PMw>M5IVK(P0I(3>m|0QcV~CxRPy<vC;l%=9bZ0uVEN}mU~`j
z-#L-g6j4hx-cq*3vC{$%iy#~*w|%Ko;Jjo?BxBp&#w!Dj%6JJHLsWhJ;y5qMn3NZ2
zESvbLr&Da=j#`H4D~hs@6LqxnU;b7B0&%NzI1GU{bd!p6B7@`|Z}N@!d{^nU64r?t
zk%#e`!g&3xIg3&b+Sxp3OL`OS@7GCTtBHDM@Px!H!PSEJ=T3T8wZX|zuz=^9r`^$r
z``YH!DI0pOG9G)+Mj=CbLh!dv(0Jc>$NC21ZA#qL&KZ7*A+b^lt4)%Lj(C2Pc<iSY
zh&p_}g*y^b`qeH%Ufic$ZSpZR8l`vRVb!M7z4B!TF($}FqQKNp-$i-BEXq@-L+d&G
z20qljaqZ4=B>kHX1|%q4)GwwgkumY+{@0?KUnZ%BF#P^cEwm@BSzHpy9V0CdkpZ2S
z)J}XLwKmojm~ZBq(6*akt^-f7C{I0irPj}HPSZj?M@6f-z?7i(`~bTtnEMw%w2$3_
zrt7}mO<B-xes|&CP<hYUp-4V6xtoOv6DO*=yUY~g*Nsp58He*tHf>x|tHYVD8xmuy
zJ&(gNQ4l#m&H_5)q~~oS^tLT@zUAyj%-^7Fv2DHGCO&<=HB)12__^P$1b_m3?6oJ{
z{AIh^QoMMU=QzdaZse0sPoapQToz(`0!H?~|B!mwls6EoW!Lq>xho_l<Xg+5NUR;%
zHwao`z11;hL2U?AO{YzWb}Hi@8GJ@>)U^ujsV<U@a7k?jRB^#YpF~4wr0Q%Z+(S-G
zZYMgfzv_8m<mnAauyai9;?iF7l5a=t%F5{WATsjugwJG~nnh89ImXz0T}+*Xu64S`
z-PQld)2)2c4L?5hu+H6X(84re=E2VjQnQk}5WAXgG*Wr*_o0q^cb%^KuWqog->1#*
z*Cz9A3}lvy-WlZ8zSdR^XSLs|#q56*-dQzeR+@RyomN-}zIFmX25gPgY|5s0c;oVJ
z8DHyvn*w&Sip59tR!*oQA+FklQ;8Ytj@Sbfp(R_Zod+{{SKrX?vw<Zv^CY^eyBcQv
zbNH!tVEVw9``h*5)UPK>lDD+#?A8kZ2=#2|uqy&5@ETIptu|?~?>)Y5M;P`qWpEYQ
z5)?YFdFevA767YSu50(Ck$xy0w>R2AFR##$#%AMFta}?*hkuDJvr41Lr?V`A4Yvo>
zjO4xS+x6kk#tI{dmL0W1iQxjGYP_9I9&%9s!idZ~GezIwErz?mFS`q~w#hRjEbr@j
zsmLpTZYCCaCscK}b<{q`rf!P*J&tONe=JlPL2RoR@IwYB{vYjqX;@QN*S6N7T1zcb
zrGkQjNYxVhAgF+V15ywHL6AZSu@+E3!WhN`aAI(xpc0t{!9b8ip%6$&q6GyE7?dGG
zNSHy0VM+oa0ep7`pX>Yc{rUcV?-l>(@$9qLTK5|EIXU~RJAav@*X;^5M^`x#MLT*_
zn!P8UPw1YoaPggQ>S`)KPT!uM=bNnNrg{+29KOk%r~AK7;&ahKq$+1U{e}~gi$?q<
z&wCsHcmY#wO(Axja>#ifd#3e2@7!ErY!*s=cp&EZ^7!rnRHVI&&z?5N;%<L`4RqN|
zM;q4a?}Rl^USB!&m&+nfC#bsKJ-+|sUGGv{PI#q03~|7n<nZLpj)>b%H=d^4woacJ
zg7waJIC4<jOg|0XuI94LDtBXh{wz^z1}CtuAd}{r=UQu>jFXKe5wnnn5#I@Bt5pS_
zdC}ajGRCmSp@^+&Yh0k#aYBRS&18H&YiIIJoyK=Q=0mgHgZk&}w%uHI&Qc8=x759-
zD>-B;iQ>D#2OXdNZjkPNb=dRNDS?1$^q>S~w@;`2c2zP2zeDmnzlx(n^w81S8(AZ{
z!o4`xpTw;*jHsEaV4A~XX3y<2AKh)|fzrCu+{_@vdBb``__I4^T#N!EaksMra40eV
zP)cc51^qYg(0~k9xk(OUdod$R_na<-vp!!7W|p+eI_@DaVcuPj?#sNY3N}1o2v?hu
z_jSGXI8t>kICx)ie)eIKpZ}hTrc=7GxHHv1U*8Ge%7bCT4qMc~ax&(Pm@qf>hD-FZ
zu<gj#ojzDpTW0SGmwRSv*_pHRfvxpPrhhD$9bhGVU~}wdX$gA{BfCdZ8xmS81$Bm&
zPj~BCt#N^ZIx-Nu9dmxFe#2tcEM(2*TMHI{{>#oF=dW->Ti@2X^;S6KFbP&?=tih`
zbD!!K4yQAvJ@2qnYC7m(_%=4+ND=AcHXN6$o5`jVkJhTXO;8R^Npf56)u(I>t)3$J
zyus;kor@nfJQsG<ru~y(&l$On&$&|ISi7L%R)xr=IPSK~+4Fa6xQ>g94S&*Li;Zt%
z3OSlv7#n@rD*q?Dl(GyD2(5gyn@w5#{XnHd9V{9ZDafJUvC>P`(>hH$;qmEodu@Pc
ziM6jpZNvw4=lrQVA1S%^1`F7&A5-6Q$nF7Kw>H;Af+vd;m+X+Kp~Un4m)oC&?yyt`
ztK&!0zZxSfU@%GS<g0g;SiJkx;OsA)(zhIBiQ4y{&>9gUgg@b5%eELHMrG5pP+U_S
z@)4=V;Z_~~%PN*1=fqSP%*b(5HLBxN*y9#x<5RZLVbwxzbLagLKUlGkPb^A|@IFwa
z^W?)jMAz`xdD2ANCYMpm;Gh2Ru^G#*@bk=w4m14Eq+hH*cJ)Q;c-p1S`ks@c?1{R=
z>2AR|c6~Im+ZSW;J|!b2xMW0~g^hCfU}1h{DVsY!bR_php-1m00c{e{nAlfkdcb$#
z#>Q&A=0jLCdhqlp3kfmC*z;WTruBcYyr(Pf9SL`4sk*GJpMJRH^i!Chz32TOGgx-#
zlO|JImrDBkU?&$d_xyo(O%{9J_VuDf@@57?b-Uh@$Z83&sE<ST0jISt6h!0?-M35k
z$Np;&Gi%9<#++UGqn(rfC$|osZK%Z~*NzVT*-YYfL+1ub5=)ODi<@?OHJ9D>bR(9M
zJU4fof0Bjc-O7aj%J_7I*AP^*?Ulye1I%cl$Sdyq&_mvuyPp|Tas4zS9UVQQ`Xk?8
zjWg5nld+7fg@WkOzcf`!yE!dQCi@~Tug2(gG%bil-}Gkq#bSuW=oC^}ffgKo^Oaq3
zbM@98w`$huf*ZH*QuoHQ+4bZ)K-FV5-Vkh&nD;trBOnmF3+O-5JBPSLJay@+c`}>F
z>jD%tU|ZBjl?IiFvGs!zfBtQJ9{ke=Kn3(jpc28I#H^jtjv^A@?IneC36FI{mN;y0
zPX0GsY#LNN*Q)Y*J^8qKGLJ`~_i0Qp&o73r`a2%%cYn2YqMzw?Tr+ugZfbFH6TUzN
zG?zcV_qu4A@{j1o@7M-BVRlW3<T}js+F>UatJ}*wLLD&Vz7HP^0putNhVD%b^(Gwz
zRzid2vAcjuhd_j12z&~IR`daB;FXpd%v1<ADa=j@7uPY*)QO|6)&j8wtkN-3lu*Ep
zT|c-G7qbQo;%w+UOX2|=hy|>P0ia?s{K_lMWT0%TTJ6O^>~A)Ml9>7>RQ?NO2SNLL
z^73k_HG+lpBxHBL;{}8^kw`onh=n2RRcp}&>hClrnuY~{_k$D*n%MOp2`nK@&<NU(
zHKh=g-e5YCh&Z51zkbK#KrGy<`X#_I_~LZ?b(k5qWE8n3WVw39MX(8<Ur+PpuL%WI
z^*{vSV<XcMjgaGUwHEyfpZl`Ua)7B2FFlFuo&_T49dqI8Ff-OD|Kr9!t0ycPCKyoj
z&^#7qKCU0S4yy-O?OY-J68IEFiA$79MpRe>mAL4{J(2nOLH5HU!McRCw7ZzWnOe<c
z!dnOlI<kH+6`zREzRx1N2xn9xrnz~Tj^8$Z%VV%c1-A%}HORTSwTJ>Otv)n4?mEnf
z@r*Rss?q?6h2}y*c;Lxkf5<Y0nm@G%U<*Z?jyy&bG)*pwuX|W((HsB`()>KL3r{%?
zsBVIx&}iLI@80!*6&vu=h^~78E^+CA@kz@kY5<^1xZq=lH?tNkC}t)d1>P1hUf_sW
zm6ia{1#ERY!bZt2rIIHXUjw811yaFvVzLaL{$^Exx;k++@M0REe@f6M?lfKUI_Y7#
z(-&Q^<EnXb!zZvWCxS5zzMn|iOIlrvidd`$o#2>f2tpzppuroR{|zwfk&f96QGjQU
zk$7B!HXyb7vI+y-<HRcd>YhLGiFaQ|#=d3-YacuOlZGs|hy~9aBDe&VmmppmA*5Kr
zV}v$_MLWiH^amXDwecKwF8TzBqELFLYHcC17kDo?7+VXhghhn|YTkbXiI1cq#K%Mj
zVl7D{^}u^0;s}+9<rOAKsAKL}6JTw9zBgwzJ}+t^1iYD!@BfiB!E98kO|S(STr2$#
zCpeIeoM+F<;&YioiFJsmUN+-?7z7#w2mc1f1Lk;XNLvL-<D#c4vCnck_%|T3{<!8X
zb=c!Pkj^n60{;bx0kCH+#+gnaJMc7wIYcwLiREn40eJ?BL`*zlH_!teCRK_Q7h639
z<g()_L4SjUj|-pVquM|#@v4NhPMe7Zu|@r|Z_ndEeS2oo=&&@s<6>k!yRZ=u7lRO6
z8IXz^=cfYVF)k)i8t0z`u?-;37x*$K*AId-lTvJnApe+f5G2V>yxAo1IL93T>VVfj
zkC{&17{XcLJ8BxBf8}wfnTh&}^m;82uVp&Mtw>XiNW^xPN+G6qzeH%|E$4F*nf7zT
z%tuo5LH_*5q+OMS&hP)3)Y3#q)Vwv{Zo=w%+2>x9&*5Z}V`;D$K$T`u58h(}67Oi?
zODUi7Wk5Muek}^vz&Z<1l`1L|>6FF-nKJdYt}Yb=kP@&h$NHqabimz$xVWhGB(#|O
z2vpU4c5gu`Oqzc80hM12n%n@Ru*23)X%pRvk%jo&6-xMXT@dh;`ua5>Rq&ssl?7A9
zqS=s;K>q`&UuWfw&GXlj!Y7xbLEN#u*{|?<!CW8>)cPebk688t8PQN04c#0-`Q}c1
z&l^*T6fST<EXDUq;<jZ^YD{#ICV~T3*P@H}l0wb`_mrA+R1AP0)`*X!nYrt3!y?5J
z{~p=F9p%0Rm-Rbd1LZCdL=NC(1tYMXAW^}$HL;+K^67&iHWQLKv_+pS%!f-L9t9L!
z@tpo&qzg|i*yJ@N6xU-3hQ~-PvjX9ieGq?5$cp8cRkvAhQEOn9JE#^>doScQFj-_8
zaj4TFg6OpvA_D6bC_&RI5uoGlb7|&G@FxgR01W-#`9A3}ub~YDE{V~2crzgSidrqr
z0POp5P_mkM;{&UAE#*7AO|;3Xq~VK4*%Qnt+eCzRh&1khf=F`!g@%1u^;`A?hG1AK
z3l6YqfEr)k4I=QoP#~?{fF7-CXc5#9M=NbCSWaS>sD0@Ap}-O<%$J#{8+wBKk4(+s
zaou!x2#J{F4b<p&%;%tuK>-MgzW_+~Zi82x@Uw3LquppXZ=j4yJN(voQzCeD{rcyB
z{aXdBi?n<8(v4Ml_Ww}!tO|gYOzc|Y!ZWm6rGfE}wUsf_99v(~>z%R61oA=f_v*4p
zOmU*h|9$ZB?w7KGa9YbTDV@#NV`SC%bFHe~grA>1pPI-CQj4ARM3?1*R|d9eG+N@C
z9tY3>le7rB<viDV;S~4;cRFUD@jrfJ_21uV7Nt)xo8W2B64m`*WFWVPWft@p7MXy4
zP35Vdd_1*s)9BbAzfCXKa9cMfD(%h#+t%Y%Lm{4@c3vSi{`9{0(5{+Zfo|97&jS_4
zgPkbQm!7{W=Aer$*pKFDrmn{SaSe0AK||ju&f&+&honnofw+a*^mJsHCH1K2TUhLx
zHru#D-$2J(&Z<)wHC@Y!DCB<(yBT+{4*I`W+uh}uRh~8-|J1%c=Xkj8i8Dfj@9VZ-
z{=IP4(p9TLiq$nq)g1hu924YoIdFwdTW%QGptjsU*pWInxr-uN+>biW6>Mm1v5mRB
zry}5l%M(e^2i=1m-qwD;pCFIdP0pD`w;5kEF?{o5Q9JdAoN_0%$sym7Ri3yKd+eK(
z3=96Jji<1$j86LHqjSVsfPFQv^TOuC+c)@`8|+El!?JapDrvAz*De-X&U}0AoZl$Y
zu=pXVH(oW%f4*X$_FB<oY!>KE?d1J<UC1Nda=`u(VlYzUm$x|v`t(v0A>tq1X(R4t
zR{83!H*f@q%l#Sq7=!HFCppHC?=+|7m+2Y+GQ1<JRPU4Ba}J#RBsS5986Py-%ebnQ
zota{3Hru{>sPLaJtM1sM_oJV9<d<LWsDI*e*Y7mPB4W5~fzQ16I9&UlAhQJve>AlD
zq>)GU(~@QZ#7rngB_HI1Cmx4H{)}^bqwf=|?si*)Gu7{pZpzHdyC$q6><Sb9H!5$?
z5;BBCKj^l9MEg}|w2HQsdEb0Vz7tBpjrNd_E*+Vf9Al}r+3QvGc`g_k_L(kMrCqW>
zZvV`@NzUwFXTq|=WT<Q2+*wl6SmINhUmO5eiRefTY)by&#Zx1C_I{{elv5gf2T3<K
z6Y9Cp@`Il=DRy9{=g!IUWs@TA|BnRGlPo>ZI#gIB-lML@Imm+)jkCr@0;{n{h)SZ~
zjH7@%k(FG8)2k77(IUnlXWY=!E8OC8s&G3}UHg}nq#~2jl>80;Tbfg%mp7hE^_RGe
zp=vQ!w;3-Pshh`aa7|3>mRPRk-YcX9_WTQzEqeEDA^H?TWbOu$BwXXsvTbK29%bI~
ztOj=0Vym<2mc?Abma{hs&5P}H42~fE36K}>QgJDF?UBUbw%`{>p3eK+h=2Ot0Gu5*
z(VE`rijG~EO~>iz;5mV0!rw>EZm~O@YZ(Y%W7%*Je5N}0^FU%3xsQjdLz40C4XWX8
z{o}25M_L@VY)Z|}@CA<)lIXi<_Ewpk3wn>Tq>@hyNWM+@zYE*cHa!_Gdz|5^>R*;M
zKMUbZ!HP_+O1n5^J`QB{k*(@3tj~Eq-boLLJ2wve-q_VuHi<%O{9tQ`YkW(;vcT!r
zdUn1f{l{Udb={q;z@p8mi()cr$<&{bwy76?uW2vWue639HY3V%bqj-hru;N#Qv@-?
zJW6msc<IiqQyru#=1Cs-ss((V7A*{4OrCv}RJMH@c8`XM<b12?S-L<Dr=E_n7y<=w
zGnd8)o8mTwHb>XCmAG)C=>wB~cZWJo9yBl&PT;RE4py7H*EvU4U`%@J5JKw6Zz-g=
z<SJqT{6*`7V!;pF>Rs~y;oRJg^k>ieGTq!xRfZdr)$44Npy}=R+wa`<G7T}=jw#($
zRGo6WS{+~C;L@BDc82cgJ8sk-z;X*l@vXh5b{LQ(Q)g{Xb$bVGlJ4$n%I{mDt}1LS
zY~1(V&}oD3Q<~FSV>6SynqM{F3n(gi9~C?OoRyLPbBlAIk+1z}<Uxx#)XkTR!p(zK
zgQWH@r28e6dHc|*BhSVva^et!f};;EWe&L_5HlU8wv|vAFk5<3uCb_gGdD>X2C2M6
zOn$=-G2JrH@PBESvGY;I_o;8XAIumyoOYh#c8#t$B;=)xJ#$;}K~}=J8Db&%m)lW@
zQ-iZ&zPI`M51L_HGVY(w+1t{gGZ4F{+Nk5)LCM9@W_qhlK>1cRbGCbam-lvd+?$Gr
zHI70Nrsw^4nO|0Ecl{KXjMWMlgY`b4U0v9;aVw?#p}Nq!r72)Z?DF*VXia|xej2i(
z7ct*blljDm<H<<nx2OCw-jr1;0#7HVdMxlF*Qs-KQ9m<)BC&0(G??!vWNlU1_TjA8
z8D?2XC$-BOg&OUQPv5KNp^=&}L>eI<bYgwdMp|3jSBM!Q7;I_@HNf~EFH==)#3JTT
z4CF;^0oSwi7SX!J{Jksgc6rcZn5pX7%tt}%3biIQNOwAYrJ>B#NXq=nrJIX^)^08H
zvF4AFI|o}J8GmWgp0k-%E*4Yq4i>fd)Juk|uhbm~&iFjc)H2;3w!h94Iz&66x7+Z{
zz2D0+jP)Y@%L)q0+y8P_^Z0DPpGChF!|y@gbR`k5ft}U5Zv!$QzJ*BjOQGt`rgLX1
zM)HttXWtL+#{SL;5$$PiuJvz9cuuYs1XyKnLV7KDfIiSKAmtN?6|XQBn+^EP?pZ(i
zo1fy+9^i84L56|d-Y1K<hj-W$dh0%)>LrI!{WH-zg%3`3yS?j~qverKa8pAYrwqm)
zT$fyAp9UXT?KBrje(WWgr#m~>v0}kT*fo)d$|9<n?NpyL4TmxUB<D^ehhn#L8|knk
zD6cpCE9>kb)fGO%8cs}j&&B;7xYU;vm5o6+^sjE8wLID?2DA08tx`Jx`LZgl)#TUI
z#^Eyt3q5;gjoR;s>!#JUV84&=aL%bvt)ZT$A*ZP3hqA(8ZS*~hJVx!DkAHhDrZ~Y(
zw4`3q|HCg~vbEI^D!XMl9{**Po~$l^3pJBW$^Nwtlw;W#qmffB!A*PMMOUQ@^bH%V
zMQIu6=l`yfv+un{$>gH(JfBNm;MKh5HeJv~T19wg@R$U+d%|T<?9bY5u@!bwcGZ>C
z&eBTv%Lw~^X<QUyn&<(t(tq8}RG7)~?QWOd1?jph&Cjt1iDX1uqU_$v2D=Wvj+lXE
z#Lmf-T91pGdM6RZ^rI7HcT2dOx{*@q4t2w=0{Qu(D~u=x=BPw|x)jqtQo|P;R&>fN
z(I{&eCP{b#%Kb)}xBwy|atgWxkuwP%$K4K$f!3+#GjPFY<>!mSXK4M@xA~fKlyIq%
z{e-7e9#JwRt~;C2qA+RZJua`Lz+BbHU6}HQ!6D{&`Dr)xdz|l5=?C-=EcVI7Z|EdJ
zP?f?SBBEbKZr*upTBJ8z+l0PLE>GW5bN(}xp|$gq>>g6p6Z3ZBO&i>)=Dbig0|Xa7
zHyntVfqPAd$nEFE)rv<2Grh9$YM?RS2ze2j_E>f|3UeLLm|-Psg4Qamf5$m$`m}BN
zQ_eB@#d+{3!cNJ;i0op(jATXgU@I)WMceUE&99R^9`aSQ`a?orImYE-^ajiO+@PEW
z>7DN;Tpl!&mw!vHf%9Qu)lHII9TnHH)VGV%AIlQ%@<0<pu>Px9i>5f`7)<8ME}45f
z9r1aR;LMff`Hymqtr^)+0!yy%-tyoV(N~~ovsRh<+Z^L)mE|wc!5==#j`spAye;Tt
z@}=KoDs~qk=a)qlc!eJI;u)^|LXC>HgLAMxqkf4Z;trwCVKhSr%09`R(B-7H{}!dM
zlGb%ZuB2;!UXAZy9i$%9B1fY%cg38~=*Z;AweGSOkE0}1|6sXbQfk*U23k*)dHsfs
zKHRHPq<|EzWsa!nrJ7=K2J=Z$0M-SdjYP9Oc}p+l9>`rlcvJl;Q8LYLRtB}Xg@3FU
z>0FwJJ(dEr+e8gtDd@{Z$j~u?He|jQ35Ty-m*T!27vy;K=!c#lQH)IT%`QjHiqV1m
zIC<=!2q5OK9G1+9#}r0+)m(ZN4OyyzEPqN>I2C(zK8TJ-zm)l+%b`cdFAH!{MUGaz
zI(I?gkDh`AB#{NFmwA38juR8+z1$4ie8{16Hk+i6$WiQcjQ8FiT``Yum6>+zt`l25
z7Db+JRKy;2#mk%+&SJ<h3wXLD@}aQE)J675xiR6K+YeXvO{V5gt&*$b7}w#qj3Ivh
zIz1XnkUL(ZjA*4tjn}M&wkziT;M*NlxqmN@DbsR&c{Jo*AT1*zNA8mAZbtN^(P>7*
zucFyw@@Rg_ofRVlquIa6aXn|~bT4LaTopZYKoRX$dV(IlVsOdA0-C3=QW`6_X&DQn
z#5P5u<Pe4<*aI|@%m<*FEyC#~!w5M%Zl*IW>Lke#Xk_14RT(WoO_Z2%Vv?=gE!J#q
zJej(1Od1Y=N+!8!P&nHxkME*z#=p$buF6`4^T9gmN@GO6`|2;p9fu7!YilhNzZ%uL
zgc->cU!uq+JC4ijCk7NP6S@+mb02%=c;C<-ia#a?7eO}=6&bQEcgYTI^rr`zX}jdM
zV~_IcVuGvr!=QPb{A@`L>KrZjsYk5B#%(KE5Zm(nslecS9*W=+xb|QYEb1;zViYcj
z5fv>{!sei|s9zfZUvN6GglYRpW*G$MJ4{CKC~)}_aO1qbMb7=<i3Y{|&Q@`=gP8ob
z`LEtIR9O_hHkVwN2Ff=tD4Yv&8-lI$*iFT(zvb{a9~p0n*M=^HC}P65T0AqJ{dD?A
zg;nN3)tH2+n{!mTQJs!x{WK3R+2HRw$BEZgB!qcaxkQ{tB+uKNY97Z7QN+jmDUrcv
zN>H^-h4naJ0jv)d%{y~HMd2ETrF{^~h|W+zSRK7&xo=)k16<h&uuR75U{FxxH4>D?
z%n_O~X=&-1>~g?T7sYI%{Q6jG)^63Zw!@+Y9AYqVwoeA`+5l+mXq(xTSdl5^F=FgP
zNiDuz5z1dyS$kDNh2%Z|E#sF8w6!8-rfdWaYUvCfUv{vmw4i!r87g1;4rnsEV%;Vb
zL3~@XAHCESBp0aSIzB?RL{*Sp%5iq66Jg)7h&sCzU|7(O-j6!Zz95%hq^~*G!*}XA
z$Wfdw$S4wyKb$JBRh$<C_6!R^Co?JsxBOg~7|)sRfArts!UTwImLg%)@A2q@ws=&e
z;@IreFf-|OuMC=9=do6L;wz&WA@Z4U@XzGAsqq&E|5clga34f3#mhb1<(Qz2Nz3p2
zI`~pHxNBO=jvrMNMC<YHKCuJ)m2~aC?sibTpM_vR@QE@fAfb$-3xb>&gL-*l2IwtJ
z)2-5OMJXy!X?@-JAjUgRHur{#Q(kz?2+c=!q~dWT#^hW9bm@xBd)saFE$Y<?I}6sy
zL%%%cbIoOH9MvPbQ~^LRy-tnvHusQEox)~{4xfu?JR_2&j>>udoHq(Vx$;pq-MWk6
zPAszJ*TQymF$xYo%!Y=^M1NyqG~0VSp{7j|UaeRL-RhSxpD6M&=Nb^iBKs0&SXI-C
zoH=Jfg{mE3LGbl2<;ziyB{k>%MG@iq2Xg0CN5TwS;uAJMSLoUi8_P005aK<f7>=s>
zJeHp+uU8sn48IY0a-;_yJs<<o#6>cXA6+ae5(l@*!2n!Fxm1}icY~efI-(-^GI*k&
z))FrVKKA1Viw*-5L%Z*bbgpp8&t4jQ06Pf8HdoF5-+o>@w!YZUbGueDtIy|uZwJ_~
zzrHH}RwAH8K#7170VM)T1e6FU5l|waL_mpv5&<OwN(7V$C=pO1phQ54fD!>E0!jpw
z2q+OyBA`S-iGUITB?3wWln5vhP$Hm2K#7170VM)T1e6FU5l|waL_mpv5&<OwN(7V$
M{Qr!=>MuS21A-L_#sB~S

literal 0
HcmV?d00001

diff --git a/docs/tutorials/sobo-vs-mobo-hg-selection.jpeg b/docs/tutorials/sobo-vs-mobo-hg-selection.jpeg
new file mode 100644
index 0000000000000000000000000000000000000000..4d13ff2060ef055e70607df2d8654a2696625a96
GIT binary patch
literal 496939
zcmeFY1yEeivoC&d3AT$Zo5kHdEbhK6vWo_H2qCaIUqWzqhv06(lRzN2LvR8lkl+@<
z6Ci{<zQ0>ns@}U*@7-Vh>;GQWovl6RbGmzGXZp-JJKfX$XZg=MK%u3ksRm%(6+SFJ
z0Qj?sZLg`UY^7(QtEP!iy<-3XHrmbZu@|lo06cd0#u%t8F`1Z}F%j$oxBwwQ4hRGI
zHg;Yfih6no;6E)t-UD|n0l+-}-);S;V*hK#J$nZ)I{?5UykpDRd0@Qnu*Mxm_<4K$
zjWg~rm7TMV{T;5p!+e;#4&LE~zv;IBg7^Q%PXB^${>B)zf%4thDDE(m(|^FW{{h=M
zV;<kh+>IPyd+g?O*N1<Uzehsi;BJVzJG0z<VSpE40H^~>0281Ir~=BsgFCHA-W?wR
z{D24`aCZg#C-whs8wMbOCwF!2?v5Tn0Du8p04Km1@CNwqs6KbCJKUWe@8}M9A2%T2
zZ~foZyDR^$S1(6Vk$;e|NR$8o@6(?@XY2q#kPZO9Q~vz<Rr2T0?@|E3Sp|Sz_kXYd
zq!<9;7kBiOf2Xk(0RVXf0K6LbcbaWB0KAR{0LY4m4aVjlzk_pk#CCK9fYT}fAT<R5
z${7G4e)umsyW{=M2g;TJ0R7k9t$GIlxp@G<aVJl1{^#SJj{cr=|4jM2k$<NCKkEMJ
z1we#2KL|K*us{Gd2nz><^=A-Zyqm1pSa^35e~;lG1q&Mo7oUKT2#@&Qoj~H<w86p0
z!NtSJBP5|FAtoWm!ovsdL<m5H%)($reH$Xr(9~l}s*>iuS!x<umJMQvh=GxD)(IUe
zJBN~?ZCYmb!h1GOQ8Doc5~^zID6}KS%R8gAWqwgv#oos^EGjy^ZEnfL&LJYQ^=1Fy
zFq8{!>f{%mQ&wK}dSGZ|Q&P&m@>RP=#c9CyU3cjJF%|&lKXmwC`in<MbjN-S0`4yH
z@Nsc*@Nn^QhyYeB0ECUhjEg6%sE-e}@k~7qEoq)5pzPZqWI3Ut7BR4;VO2^K#e|i%
z%+a#36G6np4UIBe`!_jMRMpVl3lEf0_6|M~kriz(7da)o{K7NJPvQP$uU-!h{rLcp
z;M{2sgaZQPf!{a6pqK;|TUHPxW>Aggzc2o8(t+wgj9)W&V^F#;CYy};6NyT@C)u$r
z`yDkLBmh(Hg2Z}k(ytF{M^hZ(vzWu9{e}Mv!j<b%q{~V8?5T0^1y<_+n}+}Y;*8|F
z|9xrr1;nMuX`(4k&}N#6p5XJ*d=KkRrA(Jh6jzlqJo+!j(0zTKqcZfr>6!{oGo<io
zg|y?;cKE{e8u;YndTDLM?5!+(a^CEwLPr1dhq$3__5bvIQiYg5KH73;c*Uhn@MHGk
zTqpI1@9{}TY?oX&^k<%lzdt@};$J+z3=<>4^{8Oz&v{$|mA}qztaM+-*y2>kTfLsO
zCwScX@N<u#GR-3p<Ru(SYso^`$aXn`gmGd8Qv2U;gMYh4|Ap5he|g|zS>^{{xt;O}
zpD6g8eDGX~cD8&vJL&bCkX)d|c~P7%__vI#;n?Nz?;zuTYRR&a=QEaT`Eq94$KnS0
zRs$a|7GQz#F$tdKLd}N9ySXnoG|PPi2I<0fu(%D7RlC{_;LJe9!Prc34!w7iikEM+
zk}R^-9;T?E8S-f&5*2*&IoKWcZly-@I7oh$Qv3lxm|p!H=`9N!anE}8{B623=FiNi
zOfjzzc8n#X=ToH<V_k5jHu*~jIs*e<mSQ+(48nnEpj|INUYK8!%{QkY&`_-q&Qq9P
zC;(@Z<q;I;7EaAi6-v*}&59@eU?a5ca_toT<#}-oc;#T7oK=a?O%Ye`&0dL=n76V^
zFX@JV;LqgJcQ6=Z2}8~L+Hvq<b61(LH-FncA^xR}JIv?H7Y33_)$94jS*)O#IzyL8
z1BbtGHOw0c#eVxGdO7+3{LECFO|@HAlo9?1U}0E#Tcjxc(^O6r`n!wqiejTj?!8~~
zQ!2YgEywFi?#t1~Hdix#&#F5YN4jsL{MG{eB9B@xj4p>CKWhoW>O4)Mzy6)w^5JtT
zwW^w%{9(#pOwmvF!*imZF14yJBuPFR>7g>D1~uI#k?Zx71zE1S_dJw{di!%>R&AQB
zNDoaMGsEv!9Qg<MhWxcmP7#3F+f5w2_|`=yCk%b`^4_iL=ciZK${g?o&m%T>&c>d=
z>bqago(j(@%D8Twm>YpcxO2467>~(`+$Y#y4|(P}fAa+Pby1m|p_K8-Qdn|A)b+Ps
zvS4-n`Q$peiDn7^D{|{{wBrxPB7$3W0lJ&-u;_4@>+`OFVEs>TIEO{`f5RBB0NCKO
zS6_Dr?HW@ikoz8b#G*P!ifhlt!=u}tk{GrZ{@SbI2^P*h3M6~>qj;lJ9@>1r$e@--
zRN-x5g#1kkIRoqAVqEp*sJrBq26M04y{;EqF{b&VvkvjwD^XD18@mTRCnnEM1x{W!
z;q;o^j=vTJeWsV{oE~O<cG&a>*tU385R4Q2BT1ZIKPtQ`z%MOS|9SLvz1*{m&J>4f
zpR)^bHS&-R#pK|)yJ7zU3?ozi0Q#hodt32T?iRdD1s+2Wx0qkvPEq%CICT~!<UDV4
zI*oiI($4kn$@AmCY%Y%PfB&|Y<z`$N@Mz7)bWK0^qfm(ak>HoVMjnFNvws~RwFQSM
z`MTX+_H=YfOnnGnCoy`P+#~Q#?tnXD^$(DrFr7-yxgC_F<hd6-J^n4e|6Kh?W6QUC
zvJj^ir!2qf|9c_yPb>!G?W;Frp}zMUfUch@9svJ!dr{oVho8=yF9*g8u5IgP+t<5#
z4ww!!`sGz$N(h>6N}AkU@0ffY<grl!3u{;L=qTogwtbm5I2BD#OVi8<${3XdzWG+>
z2_UoN!LW!fFV=&!P0f8Gel5R)B<hKT8sDLU-$JCQ=qBDU?kGZGi8(s4hY#QA({O{w
zfnv72{K~ca3)H53Ri#3s=!jNuMI4{XdnMzIkw(``5srR)eJpNiu23d9ZX|!I*U)?_
zLUV07rm5Ekq1?GRpyDwo<Vt|e4uOla(c<Qlg^&=$vZ*Lm&=zJIldRGDLvbuq@Jd<9
zD^yjeKZ4Mh7FOoE3GOv34Hl4pW&J`8mi}>(SOJRt{n&M?x}Ft2E0uRc5t4R#Ax=z@
z7muEwxU^(ci`E<KUjIGik|{_XjiDEaMA82luPCUBBdQ|7G;Qllmghc&Gqd<jQ=ZqH
z=EJCP=u>H(YyR3IO;Q_Uw5DnL8w*1hwCjY43nEmC80?C;)?qhjw$tYppq$i>sbm?g
z%0R`v$#TUU7AUzH*g<ye=vL@-2B_GuKlT!aQ#=gpEKb4l6X;7&0ujY1lZsOKmJ7YI
z^M&e95D#u94MzJb`bvwuG`!6pw~Z_Lk{67y(sE&2)F@s+R&X`12rIZ=1V|Y#O)Agl
z%T*aR%-yhXfQXdX4qV1ByH1(e3J~)0V%&+0X9{mzwe$yM)H_IohYW1Um?HYk$&X_~
zgz}!nMP%O;Z?)Utrf3Vrv_E|xMOezd$reXo_tpT5Es9~m`npRpwqD$zc0kBEQpgNi
z#TM&&<p38?a?MqHU%uJ}&Z~{h#02y1MI3QtVRl=DVivw@)x@)fY0+)h)9sZ4>nenw
zfY3x`P;Y`&^-RiOKOvvzE)PRwA9GT1%hy%rcK7*_Bq{OrDl388H!mKRq>v{*nfO>_
zm=n+B<L7fhzGW-$YVpQlgONNg7cK5Hn8;jSC)v*}&R1&T(agglouiwBglK5q?-P6;
zw>>aNiK<XDnBl<c^JR14&IX>4?!FTt5LN>`$#W!=YL_^VgU_uPvh7~uQTJhV5nSK7
zx!!41*AkFyg31ggShiTtoQUG8bV)4*1tb0dnR!ImVJlm7jcnvsbdnmgpW7tx*2f~l
z+jQzYb_qnyg%wE<+9;SvRX3dP!V2auSaL7m+QrOMRgF7Rktc(%NJO2FCer?ph>kT@
z!)wP2#oj7RpG-`0li`dqXzZei$7fB&UO%=B(I8=|ga}JMLN(H0%3x;s$%0l6uIh<j
zp-UI}Qa;Mf(rSk-Ke)ZG2HC>+9gRGmQB=QJeger+=Fg536&%;-q=8Hh`}AQ5RXi*i
z3rP`p1@Za?ErIyf%hrarQ$w9UEy|ahNPkP3eQf+PFVIySS);W?dcQ$N)Xjxjg@4OR
zP1$Y4h$fJXBNxCi%At1#8GVwvc{Kf8X`>h8j88?s7c$i65hpG2`E^m^U(mOneRYU4
zU)yN}!<&WhPJaIj@;Uk!l9Eugw7ARj4$kyudz4$-=kK5Q51X7vn9w##&PUU^FvW#V
zDiiTn_0-332P0P-gn}Pu92c!}9rMZb%evI~jiVSiy;)dji>i7E_V3`XKLB;7>CmX)
zCv=113!DrElE%l6dt$!7Uz!o1^V~9D;=2ljZkai5R<{s@tkZq;o9(L<gM87Cjx;$D
z-;XmRk&lmMj|R>K?l}!$yJ}<hwnpv=4O)EW6N6%_Cx;Bj6x>QbeXO`x@RGDp`~-f<
z;`M2~yV&-+^RfV-3sD7G%wR2z_!fQ|F3y5HAcH%hEW=k{lHvUfjF#T}yjb|!bs0T-
zDgR(R<J-i<ZJK-LA@gjcq#D~7Jh9pfeE!GjR!9nY9qY191ujgHs(P=e_r4)d#Mbw>
zQoMAa)H7O6dLaGzSxNRy)%fhaOM@;X1KafpQc(Q@bHDUl`W|(KbS#uy%M&-1OIo+$
zi>j3R0tQ>0w}p(0I4;vcC~><y3C2?(EZ`eVLR9UJz)iSWZCEDkO9oBIi2C^!>;@Hm
zM<GuwqFw*nqwrra8~)W3L!0_k#3Xiv<9^foJegj8KIMilvbyzD4UyWsw8MkP;|Hmc
zk1RP|lqmL_#rz$$3`*rCtF_Uybm;MjIjF2C*pd14jmC1R5wqK{{Lq^PR@(QObZHu{
z6uq7L7le49gdjAyD%*f9p8}H*-XV*CgktTD=~jMWP3g6SXN_R2cbZwOosOTVlI{r|
zEp5han=!&}X|BCJTDuQ%levZM*8O}VtPr(=!WG-hlvduy=maBwKhT0v#U{jt`TEZh
zIIEz`6rPEx2MUfulDDx%SRL=Y%I&t0SwgZC_!^TCmE@CPSMb#5{7zbU;}dqFxK}9a
z<jNPS1XPQsg$l(NtrWFdObG*T(R@`Vex5v`{i-w#T&?A1#0lw4n(2cVWHx3V-fp?t
zg>Xea*OJo-+FGgAj1$M2g?0uwGh9tYP!E>HK6SxRQm3fL)xtuN<5yCyI2I%Gu#%Nf
zV2T9?u88?gSo(PlR%Z@hPbc!1ue*G||Eo;Q;;WXyxLy&{2sW`={?ZP+tx$EFyJK0W
zbc}UM)fhsd(x`S@6SHE0CautaDUeT{oyLu(23Kp1c0z^)NQwBE*XAOiB&v5J`61Y$
z6zsi{qF5<5H*nEy$(FCGpoVsjO6tb)dCKKygB1*zr=BrTUu{hnl}seR(7-MerN*aZ
z7&$lijHvvbe41enWdVW<&9g9&nU+?S<ix&_ubgdOQCnZj&6djBW$PQGSAaYd=8q)+
zn}_-Ef@mCZvqoi9OF4L+#e^(C;5a^OkGnk1+h?F7VIMq7(#uV{6m#f}k`h1sl;+eI
z>NHNpB^penk!adN?e5y4D@2@2pL4NuOc-$DX>OIrm0uPh?M2Ab-c4pt_vSlK+EPhZ
zauG4+QTS<x+4!3~G7i<gA+)_Sfp~9C!aKu`Q;T+|2}4*3BiKLIFS`wb=ZP%1E~^UB
zD5-|z{-^#mL(UiH=hsQG)u{(k2M>j7lG-XByioTT%fPsz0};EJG}t_=bMw3RQ3j44
z`IH?2816$A4RzL?P6fMLPETB1NkQ`CZ0S-8n>CDW6#u#%4mOGXk+ok4Rt%}q5z&lj
z|1|WsdVI2C+_^;KnI)|PAJc;w@eB7)D+Jw=zYB_Z_<m<q^o_5|zR-822Pnw8Z*B<a
zms&x=kL9YT3D2#M$ZZn2b8+^m-!VNx)KH~nbgb&e&yUtzfa9F)6VhkbN3w!3M)qI(
z`I-|lH0g>anEUer9jUlH`jx&bldLZNep0j4A773;M_ZvmHC!lZ^WB*6w+@n&tN$fi
z%kb=h=~d*DikG>MwJ6dB=!dcql$?__1_I0(>nRFhTdjL2k-F9ug33Kk4u|!z$$?X_
zUZst+$a{FaXp`r7g2&*-HGu)wrWe@-v^3GuJwO7w{qi$<I{aFou&P?dEi1iH)7as{
zQ9<9w{tqzVxy4DE2KMf*pc|994s?W>28f!frw+~?Pu|-c?*to>sB<`M4JMuYxwZ64
z>un?fJp{#--nGh9DlwWI6hzo%H%diEDvyxy{!euT$o9(?95fSOk0Tu)EjQL$fCR)6
zd>A)RwCMWty#0j9RON|A4SMdqdp8nKW<BT5w_t1^;ZTvE<#V}3EhVCWmD^X&(l+@z
zTZ&hxXk&uW{MWf<yxqgf*EsDgmdH7WfR^tKi)sE%Has(JCK)NWQY>PFoLd7~#?|Or
zbSb)h-e!g_<0f{tHlv+YYfy|MHB>^7hK(YpRyk_Y7^Q>nMoy_3O2oVpR8N1txh_v<
zh+CMN0$m{HM=hKk$EC&wuBxag^R!DOJ2{Ml4<p``SZ2q0ic^_q^Rh`lV3n3Cj3rY^
zXdi$4)8_6^Y-fELXFz>^v;-F&_vq+T{E^+0O67ZS!WT%rxc;z4;>?h;QKMvC6nYa_
zlqY$q3LD2wDA%mu*VN3BvbncD+FWTJYSN1#32-Vw4AGWK-@|TL(eH4@ArK?g@+j@B
z2pv7!`BGK8E~;D0z5gq<Tvc4oB&~g#v4CeP9lU^nJ_eoUvh%6(pu~k5WTT0TM>N6Z
zd)<&wE^Tp<8voTzU-T3xfNv1Hj63!r3ErsfLLFSjlXW7-RoXjcUG7CTaj2!rey3r`
zrz`~fd1N(O1FFS4$d1;bLU8SJSD@LsxMkspF@(n0RCPKB_rQgc_kOpQ_c+qbSARa%
z?n*y?PdL_Gu<c7f1WXV_hLykhA@6Q`kxX5dEc(?}*f$2=6Q}5}lqlNTDV8OnG@_N*
z8Ezz%=GX|_=mmLU#JL}#0rwbFp!aI8jr2}sS&TMS#<*qB@nqB6(R3uhX3$=QuxM@w
zHy%>mniD<DdLvI27l17;MGp2}hzoXR|7^~ks!<juVj}pgF?m9t8e~u%&9_s338JWW
zEC7+O+xWTsbl)##DSt(p5+oF8P~IJDA@gNj@^lIBs2SW$R-;61D7HBy=7~s}YoghC
zbbu2J8IxQH>hpjp87RvTuAAfj0q~y@3>F33jQNX9y`>R7UiJKC-^65721|Z^ssGN4
z#N%9W)?e$@b=hq)TY*}k3T2w{ctSqE2zEzYb%JU|H``QvvrL8;3%4Q5flQ+~cjhxU
z%lrUQj`MA<CzDj#xf6q#@xpIEK#3Va75-kHDW&z^va~$l^RsnHhgdD2s6ao~hpu!l
z(dGti#B!A!sbl`=ez(K=AhQVc`RZJ(x5{gQY$Sx+I)iszr^F_Y>NyQevzNap5PAQM
zqIZB;|Kmb-FG-hKiN*YQvhfyzFffom`Jr6(Bb*GH216}@+Ddo@95u%i(OOxH!ky#E
z(E;fcmY_K_(*x36!nJx)@*>92paz@MtJsz?y6U+yVzvNDr%%C@iECdU#a}sP!RRGe
z%78&RekDdFuNSnBPSodWqPp1hNW?bf%Gn+%&M{JJrioj?xmBw`!#pjUTj!eK9|q7C
z#6e&Yw_skFUP($;o=4O6?!8LnNI02{rW5RWe2$_p#0zg_N*!9%C}c>XViDeK`w4!|
z>uy6?iE|G!tF@_Ph0r%#6nPj~>%N1o_TT~AkeSLj=lQ=dcfVmv5c{Tzn#jrR%~H!5
zNEC1p(dGl~#us4RSyQ(LvljGH(<Qf(n3tpWE4QsO#BG_d%yl3$PP%vtZH<du-cM^5
z#xlN~pWV$<A81TUoJfsUVugXG_X1@&H|BC254KR?C@hr0%pNVo!3o+P-`ic+&_<|a
zrsc(bNwHPtOIf#30=(j9<gCzZOwA-7b`ww+<l0kc{Dt~k#_Gt`O-;FsL27;7Y+Y$q
zQT>_yg$Ug8b=)Vc3?4~__B~oWB-<4#3{S6OBd$Djl)Y`UanB{zduYC^OnS}W&&b10
zsL}n(!6?hMm2Cb^!K$LNPhk|Ovhvds(-_Uej9h+`QO^C`08mMozT6`+(YZ1`4r(o`
z`Y?^(qo?y=%Icbqm<ii}(FZcR)Dx=Nxrw--=GkTaLlJUg2E$4`3FmK=*cJzK1wk(@
zDtgH^gK4GI;xbv45!5b={a~u{&GC$dmb6pq6=B7QNf${xs(P02OzfcFvf|=c`{cQt
zx|4&oGMFm|l4ib@8vC~8cJlP4Z;7OmvZkvJF@qGBj+dA^iP@zrllFtI{mi_2-Z%Vl
zBs2@P=k)`T8a3x-8<DT*h1Src<H)(g>M`;Bahl*_L(X^Qn&gvnm~=S<I(@N~`tn!=
zO6ihPq^V-3I=DS<D>=!W3mTWh;ugaMhP6^`44{{kpHr&;#<ih-rp?n0-H$|;>~B<M
z1o&vDYDqbVQypNd=<m0JP5lqfmn-I4(+}D%)3SO5nU5u5U&n0x70ygG8vCNwU+$$&
zl?Ugb*z-Rc3(l|k%Kj7HuKzY}Kmm($;Z%qCu$l$?ZSYF^p78W+CEMwq@UDZF9i8;t
z{V0j@4LTjo2_#ZmQ#FUsqjHzses`B?q6f@J=X?V5rF;)_7W2x%X0jR^A|rT!Ch+6`
zy*l%)t9LAi$W;=Uf!7%Z``y-9S~2YBOP<GaIqb6w>8%gHo_8ZJrxGv4v&qV`oOd}y
zS!yLQYxh-(s?0_4m>L=?Sxj&+`kL{<9U@UcX)s2#+uN9TS)FW~2_6T6)OgI&&SiLf
zCkw!Sjdj1t_zXuVi;9H{0O-wVt819>ncr_&*f6{{$`dvrop%{$o=I1&%0hYHwl&V!
zFV|>K7@?(x(<`)$9U0B8W2(rPAIBkH=a?YoCC76OX4S_*GHQAB6m%MI1=Ri$w~qJa
zBIKv2w!>O;p3=vsBEzDpU?j<r$iX=yB$F447_@1S+;!G?tiYSXEsK93Sof6SHzPi}
zrD={h$yBT9i^8K}*sgQUEE%J4tO}_i1f{2cV87~b=$zEH(FoGq{w-6_8C!E=sx*N3
zkh`MM`n_&6gIo(4&GLxm7@x)VlcXd^@jwWq{ZoUA-e?312c|y)dIX9cU3@7v)zD?P
zk2V+{sHrvG>P}hJ)@yOSuyA(EVgsSt*2s0|(#}s%vxQG$B7U2B-&Pys<s*!|WGx+C
zcDRUN70Ppp){Cv4kiA8{5nUu@vPP8>>^iRp;t<A>{t$Ra$igx>q`&|Pg;pz)UuoJt
z6HVg<c^q{rx$?e&C|}vwOj)q5-EVU&+;9xFQ#5%RclZ_I7Y2@`n)~pkktgx7*6d+_
z{fkRHsHIAJD2L^N0}G!%OJ?SVm_ku*R+^?P*hrzl61ODRlr_9Pf(1U16C20GQu3J$
zDTFPAw+WEQi>hF+;<JL*D2g^ekxUbY@q}pBeNWB79SW)PX~uo2hw@^`ttD(i*7Axq
zut<r1@$u4_U^ECAmo=z<%9&lgd{JHQDqhPTe8cjlHe1n%L^;ST7Ny6TMh!_l(0V3)
z$|_It*1m^x1IynUKQ)aBchL6pP1698e2OaJ{j$FT!8l>I%%fO@LLdc>#BaL-BzlC2
z__`qG)#ME?<?9}RwnR&Xu_cxt?>GPR<28TG+KU&RI`5|N*CL{s21t=`{3<5;c7Gd-
zcfSkYf8a+8VZ92B@#rDf75ZGT96xwe%XrV*u8t4Kv)(%bM|e`<)cJAHxi_x&ZeL-3
zyF^3%s`E>NlYYtHw7Uak0j%>3Wo3_Ir;@xXY#bP~YlbOkD2S-f;bDf?8ASHw%3K5N
zs7VJhtDL7I*GBixFhUW{m+p$;WW4hpj%!i)G)&=%gk%g6vqP>ofl~ZS@E#eHrI>X|
zed_~8wevAf=TeP<a`cX?lN!sSNUX{rJNAbJu89)#WvpR>fpf~lK}XwF8`pYnwy`v^
zyQ9V2sK@6-d_E@*N4B;26y-L-*^prG{lk6*KKge!dwQ;Y%{V;r1Pq_g84XPjnP&>(
z4yOzSiJdo7cusT;Vt4F9NQlDib!gOU;@P)CLc(Y4rfGYlsrZgbBtEs1g$dY(pf2lr
z0r*eVhk?%Fq+7tVi+K646Yb*X_zSHJIWyH!kdl<bTJ8vb34Oy0BGBw?S@Hr;OqFpu
zwbpcfLU|QCQm4FMa^o;(Yy@E~=lYb-Y@_w8!gF$B?^-7wLl6}jV~wGTkb}Rn%WEay
zh2rA~0epn8tljHkH=@}ficOxF+NK39u`x!|RHwb~rDF&rORqYb*fj|K^)OLT&el#^
zPM^z@qZ~P@MWcm~#?4>NA9m{;)t**NPn)cs)Kl(>rS^kw9wpJGCvx#l5e^(~A3;Sa
z04VHB>G?<{iwmpeHLXWLf_NM=y=08)LIdrsh+%S^<`w(bk9!NVcZN2Hk2aSDur^qe
zmbyjh&%MNv&xYYW<|xsSgpzRP_YBZUd5(bg-<5gc$K(vs_f~RVML~>LPao12-!F^C
zhcxwDSk$#AmWhZPI{to!GB{nCkA$|jm`8Fb>ee*hw6U0~Y;5pJib|%d=GMEkrhF6$
zgmDvs&%jMPmc9?{62k5kVK0!TxE8WS+Ql*HIOD9t*T3I>nFvk-uLQ}SZ=l4WYM4M_
zkvlKXo5)cFL=7W$5@%vj-PS_epWGLTO)~j$r@2IDB<5=uL>HojafCsXwvsy9V}w+`
z@3vJkH`h6?<QqO4-3mQ}S=GUiycxmO?^_rbZqwMs`Gg7zqS6Nnq!d_D?79e<6E{ln
z59q9#S~R%jNsLmq2prFbSvX-nbjOKp&VHt}`&^RS&t?-)h;V#nu-Q(p6d?vApDb*p
zGPhf=Cwx|Vvlu@<JL~&)Kvt~MM0bGNq<pat*Rg=Iwzo}`^?6Qp`P{bGQ0?U-Cwrsl
zrmBSGcblU%>~`6ySrg!CO&K#y!tyXj>zdzCz3iYq>J5G4^+gzspBG=ReiDS#b!cPY
zXw6kNHnP<C@JZbpVhe=F)^N^n-hh%BAJOZ+O|8IYWC<_JOF@1Lti2L2mw9k0ZLVfW
zgk-Fte#R@kO5cZYRTAWH)6tS$a^vmSc(xd<$}{8N`iaiob5Zdv&rn;l@sSAq8J-m@
zVQ!rXo>h<9ll@I@8H|0oK#JBV2bd^QD&o{)@MJ?d9p6<la>FfyJN7lpJ8(ZDLaF++
z5jsVJm`q}t-0pTvTs;3OGIinybM@SqcuSk%9A}bV@>CLin!wC7lBI$oOXmqrRU4Z^
zj1=VxXlHB!X+(60fUEu4J*Q1-+N)-d=_$2%aoOE&A8{Hs|5?VzIZ0m)Z|&57Ed+ME
z#`N$y+7OuKsKLhAeg6Q5c{7A!4`jG+#<?Nc8t(*!KIyfZsUaw+QsSCzZ43{$OX4~-
z*wb2iNNY(dwzqdCbtl8@KLq0D=W}<oPigh~j5)=hF8%?;)S(GT21|k0BnUui<}8Ya
zAd1O<Cwd>+jdJzv-pAI~70`>N4vm!CpRD@g%^jC2gZ8$LH%hnZlNPfq*PPDL8p+cw
zn2;A(hdZL61|HVm6au+=V{RlWf~}c*sgJkzr)ziKzOgKRLcqWuO<<|=R2w(dbVzZ8
zj=L`?w8L~!upG?cpoU31+ZlFfRlC5~>+R}SyYs`itkA6lysoHX=de);6~_9xJ`@eU
zfAnUJHU88S?{upu9>&OC6Md#I1yKp|3-?b96yyuEJE)MZ`ppZEu2?vLWL2Ku67Arp
zr>GZ?5`#5MozQK_jNF25gn~LZi-8dbzGEeRdZ801z=5c8tSXZR+d7J=e`PO;25ZP@
zi0$adZWZ&EgJfW40`^BLz%-~RyyfB#Q1u5W>m-0E_UlmAKs?IWGu}R=rKrg>b9OW;
zG0szw)1;s0DJ;#&lJZn5hDX`SZM29Lx)Kfc$~f=iu7dIdNmfnlVm(u|(-am8`IAw)
z#Q3``!q9Xzrl`pf-a?r`em{kuYL)jqox<{pP@V*8Y_3x0!|e?&+i6ixwOV*{nGk6u
zHCL(j9VEEd5tWEHfkdSVAA*^?=$=Kg4klyc22`hE(&mo0N6S7(#sYcGWf+3gU%W8Y
z(Pur0&lHhz)=!ahO&5}LnQHNI33lUS3(%)#sd1x1>SF^^Yywr$bsYiRuW>ln`%$Ho
z!%51&amTy~8qaVE?2i;F#97JYbspB~FD8(_hY+qlDj?d>FEW>Iy)Zx*rq7Q*kJvGk
zuXJYPDXrDm<ojM;t$zTfS{YK5O_it+O+mho!aW6hX@$S$$lwGGE`R4ci9gZdETT&d
zwB_o_5%R;VC$u_l9OVtNkKyJsVOIeg;ZXbv4_CZvxBaNxv+jhi)T&tRk_xk!(1Jfe
z0(+Up$0$VsveTEZAC!*BG@DMIP0Z_kc&cfc{6PQz#fZ3n-Ou>{J<RlMtGrr)%Yy5}
zeW~v<fuf9oTh>=^Sxl^6{nc<B5-5P>>6U(@Y~tglZT!}g?<K&NMSjJxbr;j(2-<%m
zet}ecIg^}_6L4#87iGfg4iffKT%-8b<DS@j&ZDSAG)mccsoOw4wA6({hPR*uLQJrY
zM2oR9Lzra%tjBy%(~t!QrmOvy$@@(AC%Tq$=|N-Vp;c<h>ilZCnw6yu_D%PrBg#wP
zowBsYb*QXLnP&#u$n%xb=|(YuNAfeGC1ABwTP^yI8y&*nd$VhV2?L@b(;+^#f>a~b
zSFQIzza6@gcq{(^s4Ajmf7NO-<tU1hs~-ebD?T4B#gNhN2^&e5%h+3B*3w@CL%*++
z%0_4fGzBg=5B67p)TUhVa4eWTR^^R@NA{x9dsSN}+`4fFQ(;bl)ane14==bF&)}Y<
z8`aK58WO7Gs_am;K-Nw0oJaYi@ph*Qd^6tGba}jCNJ=+3q35~>0~Cx9P!toX9ICS;
zfrniZXVxfgMzQLBwK9o748WLUlcgG3&a-TC4&)PVTo@3WeqaKxv>w!af+5>VO~*?{
zjK|kdew1USoq?B$1C@2H;<cLtWanUjOT<9r5m&k=L!hZ0HqgU&`#MKa@ePN{0$vTS
z=^udhc->sMK|EzdgV|a06WL|rw+{s(*e`SxIWJ7l`f;-}?#Y2My}ys17KiN|{gt%%
zGx2QcFOLvHg$M@{*5{s>Tc64Lfg9G8v6guOt%k(Tv%SkM_s4ZZPCh5EOfSElEDu`c
zKB*NOek{XYehQ(EDKJ8yBa_EoxCNjoVmDZb=S8#(J%hmet>pKK&ilh;K46@!vOJPN
zZrt3t&g+k8q-Py$AgvuEOt%i<KL;-|i9Ug5!v|0(N)mhs)N?i)=P8my;vf&AbIjoU
z+*GYbJgdIZwZ+2R3OA02L!&+`g>!2vlVLE)N8l-!?Kv^V^`>=>t<qq&6gjFK4>g^Z
zj3)r$bbk3<COMV{XSrn>rocIr;l0N2?msn%5<7TPT)-c)0>+Dw%0{|IN;haxyRCvv
zoO2>j=d(ek99j<<=mzap%If&FD|aoDv^lJZVT2`QzP3y_;nf7&pSE_k=Hj%osae?E
zb2QkOyN}1Ba=HtlmLYuF`cf5&UkFr{`ovyR%eo0Rj<>Sgmy2+wHWt74{>Bw8%M<Ky
zR;@l)t6|WftDfo8uUSJIV%>cb!TEiwOlP@`weKeAqEv+?ivQRZ66J&s)~Fsk8zf!~
zNR7OH5hE97{ufVq9W__J-o2Xi4<1dr39Gi)wN(g`X?q0ff+L5ALp!0|v4Zh|P8L0t
z&Rb3xV>ICf*zY9h+m=)8ON~K=Y$zi^RdUX5Gre}gG){VIOx@N<ShsjHt(BB#a{-bv
z!yx(os4|zXlES={ACK*I9I-~m;L)_<;)7D|it%mjn4K!tcN%K?x$@Xpk|1VgZhNj_
zd<t7#=z6dbotM{QQEqX=BU=h)|2KIAGolM;Y{yiV&(Cj`p!(6~TpQV?{y7~c9<5bk
ze*kjr1J`gGjg$_0k*KN&YU1>9Y4H6jQ39*`Nr@<sj^j3SCJ~}7<6|@M!}Za)Rww=8
zS1d<2PkLDRy0`{==OAEvm8OCu+*4A2f&I}6^>a>Iz3EF7-@{plFr7fR01XG8h6g&N
zs1(X{cD1(S0(q5#?NgH`1e)E;y3KzRm%GE65;M20e{V~cMsmB&Us({mDGbkxs@}X!
zSl~||E#~03wPvP<FGfN0brHrfRN)Xpp2u_A+=&DB;|W6q_o#S398kloV8w<FseIIj
zhEz3Z&RvcYDiydugde~<5Ew|6?>slltX$|hvX%f9JQtjQ<59YA%|gEhP;XS)qZp|V
z_jJ|hs4l3rH3LyDEZiz~3MaD4&YM}KIx?2qv+qREI^~**_ht3iwH*-^#9HWT(*=yL
z`J9=pN0O;Yp%8b0ShSdbE>3<_v9IlCt=4Wvy^d^ttYLB`P5v6(?vmbjPcs;iYLuhS
zAKVq4W^JN&D%^Q^)A{7}X)C#*vc#2znOV{JQcY~u1~{!sre*Mk9%*=PMyN0=w)9a4
zn-Bh-+sNRJki?~pTMYSv#8R63P^urb;7%in3~iNmV6e|z_Do>4c=@G+)=c2QgPNi`
z?mR@aLU*_pUuMP_iUudWnv;Q#$H$mHKogtCEm1>gQ&X3(RIVU6n4}qDyd!ikOAgoI
zMaT%xuPAh|M*N=A9zoxMR<zha-0{!SvP#IWGFdv~{mrQaD0(oi#7gL;fbN^}s1Uu9
zL$?~0JX1{$*Nr^;LopZ874#`aW!%gbR(|N}DSPKpBf1<GO<_6+g*KC1z0Q_r4<qC=
zf22BpBWoC5`*R*kfj2Fp)+FY(={(9HFV0?(oyJzCgv7_kj~RYRV<xiz_Y$rcw?RrW
zliAwBxMoWxa7|K5?hTNL@V-?$zW6a_aK8+dPRwYTBZ%p)H71X%dHDc(f<zXHimB4F
zFSH>`tL;ky0&p|VqSx!I-1BiIqGa-G(|AH3bMx8J(Dr?w&YYl8&kdJzHWZZ3RSc!;
z*eiPKZP@2dy%j?3+9VXcj^V8lJYRod{QLf~lzckeIjO2TZP`V<B>E*uiITmWF~5dh
z#t-}YSx-C?qoN$R_d-^7%DY@{PEINc)b?yV!f;+cqF^)vv&Irx>ZR`30Cz}64JPPm
zvxj$ZB#4ep`QQlLi@&Wpf>Cu6Ek8FaF?PA|{A!M%Dds*z&+-zMI+}#+=d2%ku0&Eb
zhvvGC^m(~MrOBs=`vWm%Y*^&sn;o>S40ddaG><FAoHpwAC-jvxK5a`+PfW#6-=@OM
z=Wi_RUNzW`5vo<|veI%aRQJ9Rdx7mt9P5w1ea@7cKBe2wAr9vWyE(r$z_=pV-*NP^
zJ34$yxPga*sbt^hcO?d`J@SY#i7J`ONxKpj)IWS%a4>eDThcz?PeZB4g{z9x&Re9W
zGvXb0qQQ(w+HNk`s{8>cA)W=gbTWu)$>-E!;~fudJ-N2)>oL*kjbkHmj|ivO#>|^@
z-zfw`+z)TCmck~%n=CA58D-sP{T|G@N#ZXcn{vrY##$FdeTBp)@bGFbt@0|-vN+9E
z4H%M@ZH`iPcuXp)X<(RPJXYJbppI}$sw~}v$DTBSX{tAy4|kPlC0Cg{U7dZL1F`KP
z^C9Pz3_6Z5;jfY&p|Y%MGV@TSPj1*J5@qVNVn1;%r^}YJt5sYb%p~!@sbs-f)ow_s
zb!ot#3V3JdQsdY2b15@Ng0(!A`Nc)??Z-73GeQ|h)gawSn0oNaixDha{PJkuk@P$m
z&UK^uZFSN8g(_-Df`a}zGqkDrBVm<KcM(5XMoKJu9J-1J#YTdakm=F2-tQ~ymTzyr
z)i`;7B4bkZECz>6+uq8b`+nQj!6z79bt`UbXSd%oQBm3|@J6%U?)v<>#D&b|LKQr%
zFO?rpv&wKgo`k+aPeUtICIB}F6$hE1%N#twDwPXr|AYy2HF227XL1s7N9#b9L{FP4
zZYd%`2QhX=as<8N_ldWcT{|cg1r!@t$xOttTq!rgelC{&Z!L<|p&QA#|7V$5tYx6H
zF-X$CCUki=K21r`rLtwE3ck(%H4ABcx_!s$OAsg3k0YG2x%@SYW+3S}`?46<h4h^G
zZKYoJWq;f^JVsbO=B-c=9=X}PE)V;F_U<4^wGglN>4;djuWjc@bWs#sUz?_=N^N(4
zNdJGag>uuZ#@bEzZ|N#Y_-tJ5mP_wqY_Z!DRh6<hj;m<S&eeGSwU_@)lEwd)BL9h~
zZp**U3z&Hw6#8@XvtS1|1%@vO$EPE;CZnp<h+A;Q(uRFQH!Bm~&x7eAT?>5`QX=QE
zRj~325S(T4DW!5-P3t3f&$r<E9{}P@@p_m2*l(29*H+Aey_pZWFEnzm^$!38DO}I4
zwYu&HsUHM4_|Fd)sW~|HySXRSYcmUDp`7-oGcu4!_D9shocUE3bI;M??K3=JY*$~R
zRC6WopgSK{HM_fNprZ^YL)f`QP#bb4*FJ@H;TaJyyRQvhAc&xaxZm1`5#KtmN0xA&
zK|%#cn1W60$Y9_TtP)|;#SacuNFUp(F_fW$;(ovZfi{OdEH6?d;DHq)OhF9>Ne=15
zJVskOKv>$hrj;-tuXzxFx2(sa*2X-WO^7X+%NK4S=axzVb-!8L+fZ7hs@*nMVHzCt
zvx$Q^LWrE6{sFRVBA%h$@NwVLYI~5+2WWUq&o)NKX`&CHrv(c<;!)Sex~Rr_5=|ud
zi>UVpTSUaL^{!>QTFpz5=L+e*W=5`(n5n!`h6I<9_!1gfSl1j|Mueh><9aTkn6-*Z
z>KmceC?V}GKKhu3#~dS_R8GDucd2q~5y2RfBXugFtg=)pb!>@3`tL2?O5T-|vw6x!
zH4XwuwnuCW1?r{M3s-ZN&+VALrm7062+@X5N~u1Gq$Z83NDW1sE^+01p$wF5uauRx
z{D-?AW4j1lK||hA;i_yBZr*a*+Q!`4E)S4jYDm4WpL(U;9P=GgDB2SutO^p3_IHUw
z3{c21>iRgmxcG&0@5ANW#3*llS4_<GtO?a*+Yu$ZxOcDX4^a<ZLk5WyEyEiL6jqdL
zp>Tg6I%7q+e9W9^p+95)7TwQgYp{Y7$9&T>09g|TYr8jR!}vq)PgB^kz*TUL4778x
z5=)XGu?_n|X*1#JVZADvGS##)FAlSK-l%(08<M(2iWNHw@sg{_aSE~3h#>_%@6jOB
zxazRqlK7g+qV1`3%(e1L@0Or~kUH-CmMBZLB6O*wvT@v@cgENHw(@i7&Zd}>IQo?+
z)*Gul!Vnyv;uuOL7#I>$5E}TgIF%xyUy(W_CxI=D@H5$o(kzw<%dIeB4CQryK^|f1
zXIi;I0*BdECffOI;GX>|Gr<F2;EdriZayx*$xJUn#-BXoweW>QC(>Kaa_g<WD=fC+
zXmRStx#jO4V-d%ai|z73&hbfEZCX(a^BM%14`SzdQvTv+#jQ2j3D8L$KXb-wnXPnm
zm?+MvCPHvBYxL$6JWs~gU!I!iPZSrW9(C`5@=iMuuj7;5TVI54?Sov>&^Y8VI4j~x
z_@Vk}o;<~sY&LYv2cswiqo#6URwAd=c1Z^_Rcb_QE5>t6>zy>d&4}H^P%Ji=JTdDO
znI<kMnyEkGsl{7*dEKhg#d@yC1l7^xT($HhWtmtiH-zBEBL14Yut3atTitVT+H@X6
z+Vs?0Y}3BwP(r%^xZa?J=aSO;;()F5GQ7Adr_SC>wO(~LkQGTfcw3+!sH2<AkderA
zGRKf#HW$^XG5MC5Z89eZL^-N+RqnUm7cA301QE81B(aq`n^}G3tZWQL@g!9}7|wlD
zocg&)j@Lu5<}F`cg~6bBaQ1v&gF3s&`PW!+EftPzb_1_QC5RwWi?$|Shn-1>gtIW>
zW)LEv3Q-cc_Y23S0l=<EvVLSVdxPHA*w#2V3-r-CV27m^@sza{=FS<EInOvIYP;lt
z1R_&JaznlxSL~9o&FwfCoTaW0#DZaj^Awx%#P^Tdx2D4TvDrlBuaq&c=B52iKej@)
z^h4dGI)&F|-da_3+6Cwj!kV>`3*`|a9iovta1Gvf?D0&K2puJXY$3gL$h*FjNKr{i
zVPXTAUdZ;~!l08mVQQ(&?@!nfHr24pg}pFi14;0jkQFSM+ytW?y38STrc4=9$y}&a
z17j|)W{E}<d8zMg=2nsQwuhOMY9!cdY4ynI%I{3<e#*=aw6$cNVRl>FS4JsoDEvAZ
zQ8hmd%zQmSl%w|H0|~cHJ>$z){ZS|at9$sJ)>#DM0s;7m62lqWynE4W>QaSo4~^wC
z#;9o1v?|Nvh|ALrRP_k$=UXaTFx!|2%$z&z1f^DnV%T;>2bzcgmKI?_W-+E=kkrXf
zMxp3uQ_ND?V5P+SIZo1?!89VD84T0ria9QbN>e4tn3OzF-pr%co=9?VvQas0+H*)`
zn>%*0))|PsdX!F4)<HKK`ilIU;|TEVE=QoXzwIc+aH^`tA}kLc%g!rhecrXYpFO67
z>||vox{6(yevlmtW4GTa3F%>QwYIt9b`}77Xhf%!z4i#jhMD}|J`CYcB9m%?JId=y
zOwiwt&%E@JJs8&0=B5_CEZ^_wJB$WZ?558IOV)Fv$7Su++{~GIP^2bzLF0Ar)1%52
z1^Id56pL(>prCil;Nv)47@@2NH6IzD%idShCmtOXgC<a;at}U5E1Z%uyvz9EU^xxX
z72~<+h=oLYqv>Xvx*AS)luO#1WTUde=QXIB>&N)+hCj(mw9%aFFfJ>eXsTFvKKpZy
zf(L?9BFHh_t$;rMV!Cdz&0B>YX$ru77iJd;u;n>~iu?S^m(1jr7kI!X$IU0tER&wc
zFUcaEk|D+^!Ji*-L)5<|GPbY3T~+S5E_Xwht62?VLMT1m*JmE|+`z&I<ZL_kV(S9n
z=GkDD<FjaDw}Tt0C(*9dx+QhQjxmOMvsAg*8J-ALxqWFFtzz$xFva<_+$91>x3)nJ
zwn~fE=Heb}eE4mF`i^U>v$lV?RP7#W;lgPoiFc1;ql`pT+2Cxu654c~Bg>t%-Df%;
zn22hzGg!$6EHJhTy)PyBCAjIAEdA*d1!-UkSOH5`0O-+MLI@5@js8&);~U9^3*Mb1
z@NvRL;a|jXXWEvlX}wxY)<RpR<BBH*o$A87pQ7<A;dW)d3kGd9mDK#7ZJvh1#BCBz
z5ee;Qr2fyO1hq4}wQ^G)lzD$0){)VHjQbx7Yr^U`F=F7f(JK5($5}Ren~CvmdpmiV
z8;Q!wt)W;IWq!@-DyiA!)_gBl*1Y$8=FRX+YJ*kLjfgCj^ZMtc@L+B}H0?x2O}sYO
znp8@PKDAcvXn|H%p_UdWvx0a^Hk^<DK{%&8nWGsf*E``XCCt?b*F<L{8yez0o++<r
zPD$Kx=>MxR{i1XO$C;6Y!zP~ri=ZHgOO@}Vw}OJ=`d^#HZ#`cAX!N-Ah6|A3J&Sr7
z<T4;pl~GCNkKU;Wyx{%9bvzMyk-8F>YZVnFSGy2(hV#HJxV%hk`~2&av$;=Su8_Ex
z&b#R;bz0^Q@(IL9ETzqe#fKFS;QmmPoKUA<pIYBo6}Z<W*{wq^Wll>`bAv&le4#MU
zV3{CQd#g~PiV)qx0h)B7<=7Zi6_3j0;q@41i%&-(#|HC#Mdzm8A#iC+BAi3c>yh7i
zmAH`b-RqvP_w<(nE?8JI54CY~{KIY;cud*|#sBjaFMXFbD-FWhb2|YQ-2PO0Z$;QC
zC><Fm08@-Pan^;lzweNPqVgW&2OMk(@aeuAZmavbV)&8FO=mg8)kktSfENfk*!-$L
z8dRjCcrtHQn#E6eOi(NQ$luyIfZ<D=u*-k8aR9Z~WwT9^G_3m^%nYe3`$xJ0=d)U2
z=_4xJ2)V}-J@&RCLQbAzkUCG<s{mFip$_Vw6Xcdb`?gC@XIN=8{5C(Sp6(|YQS^15
z6AnSzrDP=&suOf(jw<OzkMPyFso1g!zU^#&PcU;8+FFx~8A>HJUlOb$Nsap+CFwab
za<0Zt#(r4_Qf>UcIy4Tk{ZeSh$QWpU*hzOn@^YzXb{$feBE(8#gDiaJfW7{Gz9?y!
zw`)1>^w-HoX=3Qk=W}wWa>(;lGODUV^S8t)-cRl#AfgkMd!(C*DzM;7P^U)ab7thI
z!;2%&>EYW?Zl<5ummR|I0;Gdk!#Zg;g2;3%&*%>m%>Km$(eILErLwyj84a!;dXzdL
z)4s6?3GJPT^zRN$jT81e^<yZwqNGWUGv*kXSc6o4NkN)D`V1|j^jRBg$QVG|x;K>w
zrWaK*Xbcg?Uv(EMLc+-88;P&O63o`>I*$1g(D+NkwfMi@O8gLZ6r_3AZY=*gZtyrJ
z$&XBF$Yy|c;js43oNB@OagX%+!ghk<>*S>|8I+MshDMDkc7Nz5HD2T2Y^wjS?VkUl
zO-7UDh|BfkG;E&`ul2qOd=(d0bfQb}xmQfHOtepx;)Uu7OV(IS@{|XT2d@5J-*@Ss
zI|E#b1X<S7>jTgo<#lm>I2A2kcd%c3biTaJMs^0-3xcHL4>$b{E+yFQ`P>NCX@x?Z
z?i1?G20ZrHv|Z)BRtixr@^~}%KJ|X(eoYU|Gu?aC^PrEtE>SfI5B}w+LzQgkK=TAK
zDv8WloJc=^z<{ep6)vx@6!3GSnj9R#vDcvTbk_4S2?iOP%)vR5_G5T9g}0Y~ReIc+
zuD3<#-49B<mkqg$WHh@tNF-qId%0IOfa9O1t<B1t1ao<_BgiWa^b#|s=jtlt=4R$(
zKhGk%tkK+*EaNs2#iY&aZ9DC9uvz4W62rA4)am^nuu7%u7K4QZuR^R%V{sVhiaB89
zR+B690UzV-MjK`oG%L#;*3x)0LiAa-O>@3$W=ZmM^L-5%%b9OY?WcnO#bAcPI67in
z`Mb*R7DX;Gjq*_5qa<Z#W+>e5g~&vxBiF3BUq@j$M<tf}5azEx0E15F@3sK9U+WJN
z%!E&Ne)fGF(^d6Aq!V0U%N<=zHU$=k>wkWny+Wg>ONF1_jQ_D~6F=aiF6Z~fd|oRa
zkE6a+bIKi?qC%Bvzw8j%yR^g~uMj5NJVT>m5HGLm($~H0zOM__7=LXNRNFkP;}>!1
zqMz{DzQlQc0WS^0H(~rT`nGwUobi`cyx7Q%U;$&6GQDZ2yGA{3MbDC_{7&(OyNCE=
z5-%q@RResa_(^3<1MX1IQt}^QqJJ;TV?>Vr;fK5(S{|~O-!4>BG_@o6_jG@#htVC~
z?R7B*>tq8_pSFcgB^jQWTg)1-vrK<!&;J7mhA#cql{vF!9<#I1VAL-Vot6Gxb4=&Z
zlgzU7<ieTtF#2$|+9n-7wDFW1g*X|K^4mH`9n(1)J&*m!Ku!{VS-7!D=dl<Pcd`{t
zdDXmQ^o}Z2(%ETGH`?6a=zM^TJ_|qbw)qd>OziW+@RD3Uc<teO>FoW0nIxlIeq6t2
zUci?V&&skv^50cENo=8JGO{kat;#&SSnZQNZ8#}c{!EP94xblB@M@pEJNy=nM^h^7
z!ZzQ8?M{_&l-IMvUYH-vEp*d`_yh2ToI5`uf#fDnMYl+Pc`nZnOB*^IYMGxeqW-Jz
zE!9Y>Kn~-C<EJerM;oalnb5UO$<IbIQiDMrUyK>#T>WhZddS&Jz_uyB=t^GyoD@C+
zVfFZXzP_K-)dzV{C~kcv#RazQ`I6nUvq50B@8`cc9IbqNcJ9vFr6q^Yw<hw_-rwrn
z|JT(66Q`}KNS;+$F-VoJeDJ~0-nzl;oXhR#%(H#*vS;K5k5X)LjKeTTEk)Qge*p8r
zJEQ+DFG#UYX_?}NXeZri&`VVwD?9mmS5R=GBv8)LA_1SCAVhAJTxGimYt%g_{@M4Y
zpDtx3mv=j)$om5HR{`Abo@cMA_g1`bnjF0U>|;LJTby<r--pZVdjk{}Z+3nn3;qXl
zZygk8v;7Ya!QI_mf=jT$WpH;1PO#wa!QI^@IKc@d!5Q2=A-KB}V6dIM_ul>8?{00?
zZf$MV{xdaCPd}$mpY!Q{o;lX{hc`1BU9z$oQloK5Q=PjV1?y=s+dZVDe!}lOew#Qr
zWv#QLBZ~1?aU@Y7WUIf2usIKfz?#KuV4@dwnJsm|x0D|J;`bjQ_+TNqG$z3RtmYRO
zQ40z*x;JmoYj*4#8Y)K|C)e(1p**S!Y=>QBwJ1Jl*+5D1@q-uDA-vTKs`pwO6rzak
z#ORVyO^<w5pIZ8HZj;+@LLfS_YlhHBhIx1P_%rw7cqbsu$HrvA?4@(vPv_q^09!bq
zRKKnJ5gqvAdlfJIy@pm$cz@m*=PrVhJAHgqVDPUomi)L9B8!y|JkB$icZ-9xBO!MI
zCs0<j)+on)1G4`op~!2DglF6_$(I_;Tj=0JlI6Ezu?A-aho@Ut)7{nm&wgEhxTbGd
z&e8pA)<b;s{>_!_`Pc&g=F;#ZV@CM14(0LcT`P11C>J|g4{|{ZWY2DYZ!5o8f~J)J
zGg0Al3vx`Ekjckid&ug$W<gs2=1SkJ^7@Y|8!Dt$3@Qu@aum*apGPL53s~w{OtFO@
zClnuv)p(;cvu;sBagbL0Tf~-xM1RNMpD)xv$u;6~y39XvZgBSkKVvPj0>AzET@WO0
z36=5TQNly;D_|a!E_z>)kx6-L<Ms0Ex9gag2^$Hw)(2Dzz1YhyB3xS)ll1-129)pT
zz9n{@=zn+x;E6cvTwTcaVVmt&OhMIkJ~d<2o+xG|`fuT`fz6)!i~nfuQojPU+-@fy
zf3`}5`Mv_e9?)<{rCuuckMr*u?Y)EeGeqA<@0$J72`bmRC!q&QD~kURSxAU1Nq7Zp
zJE8bIR%C=%q(dn|&aM9uG{Or<VBFzuF(Ra<u+BX|5;~i3IH`y4a|i#WCMl?FUz{-d
zcUS*s<BizF?thCDQ!;n@sC2CNFJqtn0!6@Si)_qSUG+a|YlL!=`0O-FDg8gyEmfcS
z3t0QVAjbR`v~toBLZTbDK-g28x{K}UN9Nydr}@v$em~1^pwqL2UakJUYTABy1+1Qr
zr>n;!@kK!gjz5Y2v`)q(Lav$~A$FAT0a6igs(0Yaf`qqr^+@r1Fi}a|nf=L|JtEVW
zHX*WX<1bXc9LQR9%Y-Q@u8tf&a8khX0Ysb{qrmw5k*1HV<_<a}*@TfsJMF$(yt>{l
z2Lb7U#Ej_5ic&8x>8}8xfxv1}5{Bn@tWX!~-)r!>*1b*c&Clj1QQfCwJeQ&3sowwD
z0owFOH6prOIaaKHH)yhv3`#R~v&3o6eCp!C<hm9_=1Xb$Uo^0XC?bFTpGjz5jEEgp
zIo-S0KP)m_2>iXrvhZfU?MVnaF80HX=*WL{piL0LZ1U$;h1eOskiUKLh}uq+Iw?d;
zeaXpq1w<;m0`koTpi*Y+r>870>)xF0-q)*R2_#~A=4F2cU_ko_4Uj?o1Fp%_(?6cx
z89p83dkz?(qasQXW3Q*Jw7&xGWoZ(_p>&i!i}lmL0w%dYL?YhY{!IE13+fkQju)4x
z71e-0cux$55Hs2r!oLmvZX-kSLIIka-0O45lo#KEuH*Fi&@SCYS92=*!gKs-X#**|
zpdF(l?@JC0qpi;gnJ==P4#SdemBlGFPqeTNW)5XH76!Uv0&&@2x>H{PDA3wF$}#Py
z>*dTB$zHclnfF?l*KwM@<o+@BePu!A8kCzUuYi4$#_4}hWJQUZj!<s(3fOE~-`5+0
z;`i@%=+CJ}l0z62sTh?1FX%D-w_JxPu|U;{5z)|r{;fX#$p52y{w}$HX#JN~zmT#n
zI8SBLuKVNux2lPj{wEgrBlte*LDe9=UOrH`c?DI&zwrKFaKQh+iJ!v%U=a5&Xw#m!
z)&HaXn5|lb0jc6_?^~5IOz3`Mo<JbZ#X-)vGgNH{fXb&)_nf_*u@8rGX*T;O>^%d9
z7Z#vxYI2JT1}r<S&l>dh=6uBXFj{@l94Y9k&gA>cAZv`N_=f8k;m}ZTI<TYbFp72o
zdp)3&)#}S;VKU<RPU%?B%NNua^(jAmY}ild4WB5G4oRR}`K@ux57tT95SS+&infsr
z0@mb*wuU$9muI#0|NF+Rh2^aO|NDAh-*IQvT&<<Td`qmAY|K?G;IOeaRMS<U$;lF)
zBC-k+aVNyz1RAWV;q)-U;e;b#FGt+-z92g?&?<$d3?{X)7^mG#bItP|%u~-h1`Z|~
zF(rb<g9lz7VDAL=J3(G<A@1CGHQ=}8j;EQlr~al}J_`=`pL3~=rQ-vha73I%5+3)7
z!YFV|&c&e(D;`4*MS(^b=JZkQitRDfQz#ZmlU2mtTc+@`lt-b9ty{<KKQJzCzy7ol
zVr9{vMtAGdS)FQnIY$5WotJ|8@lJ+R|Kf6MB^Y4f_^HDEJcb!Dvc4Ade)8C9iFLUR
zjq|({TT2?`eHW`E6TbmN?l!xY*WrXwoEQ312GJ%EhHq^E_y%*2dq4zs0x`E|4po;z
z(`ye5jc#B74Tg%5!AB0b_%;H03C4$?WZrD&o(Nv82s$2h8P*Xa*v(zV#Au!?p>y3z
zUpd-mNc13vKW8q9Y|VLNUWD$dXiHzIM4|!=l^=tpL`6C$jQh_J$PdRlP<oNqVCtTG
z$n&3Cwiw}NvCH-LH`i?Hu<}-5j<7f0HC&>z3jVI!*&+(?&sE~}EoqnZdJ8Z#7T8bn
zJ{l=a+{L)py!9&O2cj@vNsM_scfQa4(|EmtzK%&@6vLS675=>k=-YWYboFp=`)ezD
zCx3&yh`gXsCC|VqNbYd^@V7c18s`RS8zL3<&E0O_op%t^)t?LK6n2wLDtW*9HBSX!
z5I6-XiNfAm>LQY(>Q5h!_SXkFyN)ru*}s=@t+41+PY(w0HrS9@Pi#IFKW-jF(!+o*
zHy7(gZtJCl7{E!T6`s1Nn&u=+a&z*vhOWR4DA(k`)%LjNBlU*!<>16XefXxk1Pjlx
zrF&2$!877Jj%-k>yBDIf!S8mf3Y}m!Be4ri+0mAHh)oPIwF9<Y;Oy4x4;kgf<=Aq!
z-3)#RYH5C`%6q|!VX0K+r!F6tv7rLfG*zmz4Ak53yNgRdH><Y~Dn54=bul5soaEyl
z%kB*Wb1;le`A5=T^dgK@*CWC%)IrH!inikdE}N%h*A{4AiawBULlI;PfyRg!WJ2oJ
z;V?5kzavVsFi{PK4V9%1KlVq!Is1B=jHSf(0aYcOxiGc=y!RFZ;`{=1oS7=<9rp>p
zLzgmIBefIcoL2vhV!oS|a4BVEKh;SH|IotKa4HtCF7nho4B#m3aKCAVACNbzuM@Bx
zY3OH#ZK19BZQ`X53@tkSw10=7PJ_rM{;~A&F2Z2NLv=5@qkHQdTnW1AbFZMC`K?4r
zZyG*Gr!t)EI;HGwwUqJ21X6PT{FyUxhwRsKR??dGB`fX-66U2`+sh$t+l^6=P<;o{
zD`0^z=#;4^=&BnV>4<p^x#8|suJCH6Q%o@NS>0=^81_L0kw{7x6D9Lv<{iQ-V1Q=*
zvE}2h=dWK+XX@^5LF);fdrf_XN$${e#Iq@_pOi<*LAw@CQ>X1bevr}=w-e16&VFb2
z`tm}lny(<pXOjoLSHMkn=d=3l<L$gUAo76{Ei%CX?QOR5r<W?i7ae}mfHndr%KgOf
zpgTI7pwwZDqVigvxp&IRFIBiN2K=0lj!J7{Q<Oocw36#Rq=e{W1WOZg0vCrk98WE#
z$Ipauw>HpOQJQoHDdV~epJ}usE@iS%TK=SLi+u&Kk!FgKhp>4?^$9$_0&Mw%DxQ1m
z*)J+J8>j@O<6wuz29LIgEotrD#&pDPWuGs#B!})n4=&`CY!uar0@U^c@eZdC{)Q1g
z=)vCSXy2O&rkF~$jUb!b*#b>-qMT@MVPxO7Ar$xk@=KaTY@|jHLXeFV7iIpn7kWSS
zG^tGFed$Ddu>R#nkdo_GVH%$cU^fERU@P!T@Y<8aD`2<8nrcHB?6Ji^C)`3uc`37=
zc$9DtP4x2d+5F8C$0*I?a||ESk3$vpwAw>M`{hg4A0%G?wrF3iw#D#O2S88{LK<w+
z?Hn0PLnj-odTS4VRDvn=jc@3*c@gpz2F=bE+&-q%*iWXD6~uJ-M~hhBK5@cx<z+~%
z_2yx5o-GN<NIw>lud-?hg+eb+1u#QoBlKaU1HuxPSM>5}Qt<m5K}_NODd8x9RT4Lu
zi;%97iig0(5T04lfk?L!9$ea&&^^IQo8uCpgs)q~QO{pbYWW5~Of)>UL()khpv<-!
z7b{wNW~VLZM*zUf*&SrK!vD@I+@*}`0-Db~rLAWpJX`Ry>RQkcbKEVU%f@k>2WxcY
z>T7(h#}xOny(+B=QoqDpodMIepy@MgZgcb3#kC}C7_`2g<8Oed#=(vj0r$&`0U{gW
z3ms$u0?aJfLO@q(!Rg9JuV8i<Q~RvSj}nX`h~!LsRW|)p0_s2N>67S9BP_-w$a=(`
z%+RGBVcI>r+|*t=7a}kVF_@-b0ZfP62A<k|#y!~0=e9x3L57Cnr1s_JGyvLy<58N)
zHQtdM2(#=9cA}5B3y?rO4CXIPb*P+CgV(8B@9Yb*p|*@__QR*uYqeG}Dm$%U8Q*|~
zr>scGXVnM2-qY0}@oNIHTbcDo<INCW_TM4hbGa_C33|FA@yRbWe*yo@uH^DGb9?dP
zIN_qH6OpxXe@!+e3iA#Gsa1WjS-;6?uh>iMd@GZqdnNUvu4KsmI~Tgp{`DJlow?jT
zUa5{fm5xPXAl6j<g04@oT?>!rHUy_2BcJ{Hg!ZF&bTO*KE!XW$u)6Hj^H;{FS|^Co
zI&_JdJdcPmUSV>`o}&?~4DUWzLX9S;Gu?+g{;vRFK+qk15IZG1$pwYh++@|?eWUxT
zvEAgfoooF`y<4oSOG|m?sqiep13ihAb1{n<$>Pst&~7PWt$s&Pftc7sJ3N2T=b$7%
zh8`$owQ;rw_m5sC!oQn6>SivzOY86-yGvQ90f}R3Kl}G3GRoE!SUNT$8ZDgKb}ya2
z^aPGMN9vq?q$s_XSw}o4(`mDR!rE3H6$xDSR$r11S}2WL7aJ=%VvkA~K7EPV9E8}X
z+GU-~K-4cQy^2KlIU;w-m{51i1-&;46z(vY@jf~X0@b_3-_RD^)sh?#?)+ke?w&*4
z@ZyM1v>6Ab$%oN+qSdAFQ%CiG24cZHk8J&QfA-ABd^xUhH;lgew|_b7PdMQi_VcOk
zM1!S@C1k6=wFw4Ff$9q`0G$zw<OZUw=2K)M#>=G%q38fEQ=sS=?rg*QBY&YcoH!gs
z%;SS0XTy<T;#(`QVV<~5bnM)eAAKV!8KMM){lZk9E<;@lKP)C%NUkU~BsS0vu4+>&
zx}^m<3CIw30!CuMxyITYy27TBylTkzc#bS|e*eqc_@iQ|`ylw+Y8BlsIE8GB`BrXg
z%_2g`y???t`btazVX<8636|r4Pp}(WxFP(@5ov@e`U13&Wj`jOzh4!KghD99fJzKD
zwf_~cb=FOS<UhPCt@A)#7UZXexMAn_Ofgt0xSzu~g8ue#xkwCFbN%2C)<Yr2W1y1y
zm5ie3pR`xNg`mhwg6)q%_G&TM-LpbEt)^GN_zcX_rJ(Rjg7x{W%mU@7^8^apfBHwE
zzXGORy_z4^yQe4SYEAZzGyU-|M76JPHLs);Uj9^K_RQZwL~^?q2(`%${5$7%BGR8v
z)hItzo!|KJ2UW9IK4E5l_fMMb?yQEr;1hXCv|CZtsQ&K;3%8?y8BsY~leGnydy;}z
z0H|eko6yAlUktbWo8f;L!<S9gn4a2MfOVAHIdA!TY8ZW_wMKF`1+V@&t@+&H(-Vfu
zBkHHP<y*%2>Y6JMbiTusJv&hT5_|4;_hJxIG<se#hOu7(pGKfhJnx3Mnhifl$slhA
zlxHl;Pb25HZpv$1uI^eSG72Dw{@mpb)BOI&ZXR!!u~_4oaNh!mH_hBripG!d!)mCh
z|GJ=H>1~j(2X^;c+f>r7XBYi7M0B9v-J6K$D|v+;p7JP|&4&pUNDxo+W_@|zf-d<@
zpoEq!8}q+(({WD_;dA)?*O|E(;+5!7?$?A$-jsCPE4$#;EZWrWgYyq=_^K%e+IOwu
zm`@n0&aVLb?@y?B@Q}7#uQ~eUWspK77jy`;9;q!?zs%>!<MrOCTgT_XSSAe=HL=;z
zLX+o#ilCWq`#&yu&5G1AM79%yVR4$EvX7siEcEbKQVA;c-Ww!O3DEg*bAF?X0#c}d
znqYnfK!L>GJ-)6|u-psU@UI!IVI0B#UwDcf%)#xK+&}S$K8N~OG;{u2mkvpk_vYu>
z*Vk4?7Y8EgeaW*8n0t1?ZM?2t&*Uz9^N9Vvl$51~GWgJ6dK<!dE1;$pvyt-~w|}5g
zu@jRK@GnK&cqOo|zEplM@lV#6n4f{UCxd1Sf$nd3ydWh<(N{81=tV*QSAH5<H~R->
z)nbVZkfe(LpbtvYkKcP>=_COfWOMjW_x~WZ!~FOUNQjanAYk=Bpd<SWy5_%>6MG)~
zDj#Ud3jGv<?*D22XTZcB2n0us8K3oWnF>~{edkYM<M?0hiWwz+{i^<Z`7Oh=URGl$
z&*(9ev!Hb;O~bwdV4<qq3lp_;WvTiLD+4RlNA;@po4-i&kcqRorS~&1RvCO!mv8L$
zu~(kSX3V}ls<}E4$(X51o6O57*3#Tfd&V&`d&`W#`uQeAD@MgrN4UA^H+ETh*$(y4
zC;S=8TpT<QrMtT~sU5po%qcj4>`Pe-#CzYGQy@}Fr*P!@a(!DTj+gumLyr?B1;?2M
zI-;}=BKgI;8;F-jV?<~PYnKxDxZmmNdWO#}`c<uH0a6tX5^4x+7LuO0JL&E?fB4-V
zttX2i`{Sk280LMR*eUrdfQ|@H7aFb^#HHPL<7@A0(ikZr1QJqEb!L=g+67hJOh(Q$
zH}|E5=&f|3>Yz<qwG7JKz6lpo7(Jqs{3ao&&7z&8b!6ppe3p(Yr;0B>GMbF-TNA1h
zlaLh{&7O+F!4a|?)-k9mi}Ey>KU6z7N>ZgbTltH`xtS-Sz+iG>wl%;cCB{shS&+8j
z=#8?K;>fmW75f`k`40?K*_9efGOI7?^9k6>5LY3Nsp>P4qgB{?8BGDJ2j;#bg(wbP
zB!!)uI_3;bTN}IT6!9t+#Znl@+CUE^8^*8*QzZ>!xFdH4O9txEc9|1dYTx(H$l*#_
z49?K{Wxo-!bka3oAjNcB*}R{V#WSgk@oOh7C@ZTiC8wn-%8Ce!v}VlP9X#F=FYRoI
zBUw-cDv-Ide=%D>O7f+-l!&7HL(1$(J8o>kHqXt>;%ePmu8sUAw$d(DZgMoOA~i<a
zz`qGC`x~^HQ!LWwbl1bmp}lKv>78OxTQOQ@ZkYpFeqZs1<e4Rf-D3>)>Prze1(NJ1
zgskL1xdrJ4g96y3tpR3Pnp9WW7;Q3e9#KO%x-}iD2^&pnEQa?V@2~QA)^M@oDU{Vd
z2M(g8vbR+$in-}&jOiou`Vgk&^m#cx@rDs?s$wpB?udvq>U?gXA~+<QJJ`HNNz}WL
znkz;_E6`NW(p&YdkjysYLd@brI6K;R0hS3eLl*hjbbaR<BgXg<G)xV*3T8el76oR_
z#&LRT84k{1u2`kra#`^~+~l3Dfj~SP&eHFXMKX}yI`5zT6mXnPiS<OyHHQ#IRjb*9
zt#1}QMO?f<Ev5?1ra7FEd0)K*wFRvv4P}YUoP3<LOZl{uH;3}qKKt}BvHSaFQQw_H
z606}4T8BEdQ;J*Ob&qE(Y`5)`$rZDdFsmw!>#E8Fg*AqoW$@zYW*)}D0ODVCwN)E$
z_jb(2C&a0I@jeBQV*AO0P2w@DH4l{rV%~Re9+!}oPDuO5fhicMN((C&(8<g<CHo9C
zYO5Ki)3!3`wXw+kCt%-w<e>%Dy=RT3d@IPHIG`A<zmn|_Ngj%7XZJc1eG4oD#%m(^
zYF2OG|AxCKdedjws7@1{uOpT2`Eh=1%+{*Iw~<NaqG2E<qTT-k8d+vQf1@~DkEjBj
zaXTbCFaxNvC9fS#5nt<<Q^>+)90pR^k0luOqAOvYNVdUG_MopE{+{xI0qcdrK$HII
z5-m06u(41`SFVbILaZ>ChdRo^U1l3Jf9qbk`2!I9k`a@d;2=#}p}Ch%6O%M2JySc{
zRLr`-ka7%EmJvu2FHtlW6v)bCPxmg{MQ$P1r@N#ROwq-Rw?y!-8VTX}O$a(yd<O?n
zK}pBeldEmOAC=}M)-U_QuZc}YeWYO7<#NwimSCM?C#Wx2Dt&ydzA+4}P#zFg5$B0T
zt2151jy+T8klEIi)c_%t1&BV=*^%fc`nDQ2VpwN_Q`3c!V`8W}L7GP#%zRNtJlqw9
zN(->9@+r#3_ypy_vT>{-3ev_{D~*amPINad$^PRV<{Y5L-nrA6pqV?Z5_Ob^_j}ey
zmu^kiS~oA5rFiMAo0xFJ)B_pt9b29%%x`SQEw}o30=P|`w^~q2N3vmHXRYZQt%rY1
zgvPm=8B(_7pgZ(wo2e|!6ZT=t$CqYEax<xwVHUMxn<uehk22Gw8;I`@i(t4K$cw8|
zJw!X$YPvS&Gv}ohB*^Sf@&s18mDIdhFxrY(B!Swui)(Cs+UN+PzDq}IXpMyN-VX-C
zjcByoCurfwu)G=C1jDA4jv`ezxj<7BQWOT`IZ&7hSp?^lXKt!{WQ-5d%lUS`SxDI2
zaEEw|RJqDq@YYJQEmVwvUFCV3%T@6d*hlyF7wq9yzRvAp$f@8@$O^M@=hA(`8kyyo
zLY!Z8KVxb%G`9apHWWdfK9;Aw6<CEc)SrWcBXEQ-L<1_eZCcQ%P^Gg<mBA>g>XrA8
zmfq=~t|*nyWwtIeQ_fW_$RG0c6eKhkV#doN?Jp4WH9MZgrNke@1<gYwziBy0%}CV>
zY2!VG!%@jEwrUjHY@rQ)6fbwJkWrc(Auj)Hnf#fPF+njlui$NJ)D9S#n&uhDRP{3>
zN<N)?)}Pi*iejaCIubKiD-|tVSz8Yp&Xr;V9VO|aERH<Y8dMY&>fRW&&rJELoXDWh
zj7C|p7zXBaw({-HNKoGZ<}*t>>i_{;jWsgUM=3RS`!xztA$X&=sTDQ0(!K`tVmcYc
z_?#4#JUdOp+r-)QF13wh$mVlC?R&Ut6@^Kz=6c_Kfp6H;wh)gpAru$u`YJZAb2t$w
zRt8#oX24d#vX)ghjieFSY~de36VV36)JeR8@k|in=eSK88kTi7?r?5>(UWgsc_%I3
z2%GR2I@N1cG|~c`i}1<WbeM&1+R<@p+<jsr=n>S&qn<kEThx9<zSDEm6+{uV-9&C_
ziH3gE{*s^Z|CIK5yq1RD=>J*b!~gq}Yp&bMpgtAGP=m_5=hio8d3KUo4y0%Awn*&I
zqgE5*>l}qW_fu9VhY$n?UIB3L?Kj*%CafXZ)IeSg)$h-$?fu)FPDpJ#p}voq-eIT>
z3Mcs$@HsIHYWjD*xqSQGg%YbBAwXii(4>0BAoESa!-qeYV@3K#CVEGsl3}wku?E$2
z<2NxF8FZMsEQd{f_j`lfBR3~u7CcO|GOk5DJlE_BV_;vj&j6yCwH5(W4j|wcb(MzV
zeE}?@g+g<@_J_M+Hxui8w&Ci0p^961Y8+QHT1{+byWsS!!qhkuVRC^cnk2Op*}SAI
zKq4(aR>Pq7VYWaIwpA(Wg*iYB&CsmLQA(*Z5Yn%}&Y28;Cg1|ktYKQF^IW<B^2@Ml
z%)X0leJ>|=ZxxL5*T{`<KI)<tjwZS`cktGQlQ$&?3wYv9D|OOfQ4a<ZyeB^hUBp5b
zRk;_@&we6kQ#t#ivPOp^)OesPdMZ{I;EV90vsvhAx?JRM=gx>?yQauO;qQ%fh5FM6
z>3HqeQw8jt5(~VZS*lw#TSv841=e9W$*N7DQi|+nWVHF}GKHj3wRB*fD$a;3w*XK?
zfHRhZMn_A>go8`SOiYjqK}SOeP499`5E_XLXEv7gL2zrFRMpbxZB(AN+6s$WXrAI?
zYnp}-tL&Bz>vAVhyx3Tn3*OFEOEN3_EvpQEAt{LhX;yxQAhgP@g1P#LZE$`(30`jf
zpqCQTz|S{l%t{`0f~m5uH6o<FZkx7Y->Pg~IzlhaGfCy-vxjBlIe3@}Sks=6+_>w(
z>3QLA91ZB?u#J2UecG|CO*=076H&-OlrBHb$Xtl#ulm|;Pf8X~msjr2F5bZWSef(4
zOMS1ZONQZN?pxB9)*9xHb#l5z+pWHo)anQ{@yIRYs(C0GJ`pFi6z-Hbrl5F4^18mo
z8*CS-v|b=6Te)5h*dTsk<KFxsTxCE(6_Z(XONBVbQquX&8+{$r*{thHwHtdip2s^8
zk3>~r`EiWRYn|O(RA7%PFP`a%;#1>x<zS19@Ef7s!cs9x*=8?|g=e27RjQcZgqJ~}
z(o1X8U)?LFKpVkBjY7u07hwUzuBWhmw!NzQ16R9CL^#czNM{Tq*Zo4YrZbA~J+m!~
ziA5V5{L&g&Vf?;Q>&vA5u?q1*2?h_l)o*>cARMEsdE8$EPW*ub-dA=6rmiwUv=wY3
z5*Yz^M`&lid#g{FRbhg+>xf0vL8;gm!gszM0A8<g*UBLdl?w6b9Z>jfXTQ==z;{R%
zh&|%d2$5C7d$|IIKkXPDyi074S96i52rdJM{Qy_|#n6-C&(r{yZZ(sPGxYk=Q)5n`
z-qF34R5-Tb$nEbzU)!68#utmB!pONd;yrL`oLB2h%fL2|{V|4krF)HUThrUU$JH!x
zA|JaWIs<Zl-09hCSNkT=k)UK&WDjR>d3Es21+?AS5X`08eIpz3^7Za^^ro4YaN8Gl
zekR0J^*$eqLWN!UjXRV<X;5-uhzqq@e5I&G`~yN_s|=!1?7s}Jp9)I^Knm+A>v;oB
zDm;uRT~_Ogi*EO*NMIKMoo$9XW3K>qSTrUk6`*$_R{q^j?d2)O4`WeAu<V|T+xEL%
z3q7!~iQ;`75$w+8NMBk5HX$1~f1>ssPUorRT*<%TMiBqln_Lix!&X^n!P+m+1$!SR
zxvUIYTf6ipFw_s*XP{{lwoU~dIdvecN)og=`WP(H=s)E15K}S=aj7iazXGDS2YvRI
z$Uy-w?YQ>yl(t1{%Ljr;o-Pvwj1|ATJY+}>>jXbQNFnV;UIslRQEPW=OcnN7>s9Mt
z&*p_ABN7Sg&2XKvb_+~Thl0>IMC1GICoj%BxRBSIK3?BQU2aDx)m4nCJh@sb|A;89
zG}-$xRFe_lcjJ%>a*JaBsa`Ytr7kPob8yAWhJZr2zCF<9IYZ~`$zto=z@e_IZ7_Hy
z0!H|pgHIyx4@_&Y;I3(x8Q9w}J{YEDVm1sGJdb+=j$?|#bRicP7bnNU9v0_bYG?hE
zZOzuv40)-cLfqoM1Y<CYDp2@~yOM94q8Gdapw&^Q*VnV14rfJp%tR)%WBvR-TLeC*
zC|WHy2zH)!8#hF1j8pd5BG*|3vpHNf>`zo$(_MZv`5jZgWa4;T4yeT(0$?_6f4rOa
zN$8MVLjBrO$H!{td_DQ?2lM-~$Ek!w9lk)O+SacH(<F{%H6D6@XsiH9?p4><l~y27
zBVE3*2)hcwNjl?abmcu+n~jYwm04G=K_t+1ba3Qb6rhd4s&Ay5+U79Xaq4D2+Rwr{
z>XUCI;#+!yDjI<*oS?LVj8DSMc;o>?7K+Lep1ep!$7k)2T^eK{3Pf{(9dXQ@$GZ66
zTqq6x&JMTJTTgFbBh2guJ*MZS0!ssW?TtX~{CoT(#W%jG+u#qf6YJshkLSnWS|^or
z7_=|_!3Nv*>lxzwFw<^vd+Bq1Sz4F787S-muFW*>YgMO%!pdJXHs|C#K3)aC<-Vqj
zn!%e<KO=hL-&==6+bHOZW~CSPL}2g76ef1!eQM$Vu9x~mclP5I5XE!Ay(g>iRVc!x
z^x=VB)72D?cOFsr{A=;A!CuqF2X3tUd)R`+&|s2$9-I5)J=uVblP5m*fzg%g<JLC#
z^Mh$#f*Cf2-3%b&N*B?0kDJPi8^<(3&%Vu|>0^5Uz$$CqaJM{X`BYXlhTJI8Yk#CU
zFlGZ9#nXVOW7NKaILmtavI=P-_jSG<5by3=oM*-Ot+v}c{bXQHV*~GoH*Pzw3v@9*
zs{BnMTD0o2IyAGkkWe4N4rBcXe_X`oAo%_g`aY4!ITabphiR~};v7gp(%x4ag?Wj+
z`!RoWTxck>tkQo<b*>ZJ&?)Oq0EieF*B{GUb#+}zQ^>q+HRWc4yL*zqS?_VAIqKb9
z!2kOY#&&7!{S#+3!+NY>|H1hQf^LEjY^J4|fep6o&xFwa8}<MSSm999J&8GnIqsDQ
zl%1G&S4VAtVcG30fC(oyA&6^<A>m1xxe!+&Csc(!Cgk}Fd;huasXy^8#v`m<NjfiE
zykH#ja9%Flxhm)=_&)NWc>};t_NBa22NlUu2-S&kEHi-Ek`w3;!hjjGEPw8KH~glz
zzn%&eK3JNs*)4i(db**^pk1dOWm_|4wzA_LV2X>fb-Qb`?&}r-sS*3ij2YqGHCYq-
z($w4F>+>Jj+Dq3XCqH;luR2@D)F)a<+P8WuF}eLkN0*?-k{P+Y|K#oyC8nthQ#c{n
zK`^@Mc4c-vCgSD!b#ssdrCrc{^vf4MwH-_xFd6ekr-#RHgcpna`fb_e37#SBU@DI>
z!oy%q)7jMw^IFW(IN!<!=aUA%ta$L<$Pb|f3qm;Fsm7Q)$U)`bz-qi;q=PpDWM~)4
zneCW~6strDeCOQ`S|^?TvuK*$o+u7~zbJ3k0a0oQlA}T^61{zYHYt8c$bwvX!qvh;
zk9*Q<A!YGlGWWhqsaYJo+YSPuewbnhj?r189^7s#Bud)KL^=6Cd2bEKQd(b)h&Pf>
ztati-<b89o^0BpE<mqOvV2>6>jS}OV=BlRc<_iD5pUe8qhf?fZamv#rj>W7&s@?Bf
zKw;RS#uyilxp04AA-BF|#PMZ;J6Tueu$Uz2!<`=MbVijHnk+{`N9(SJu7S-9+X0_L
zIlgRj>0%whoDh3!b)`^Fo#C)@3Gyr>o{qKM`_d^ARc7j81xe)!qm<nIEX^=8I`8jr
zwnHEM<CvzTM8wVaN=nzujk1bi`L6@G&(R5&z>dr|fw)XGxwsw^*38vRDZ!aJ0(!V9
zg{*nV+BsqItT|a6EFc03vRrXqL3KQdNV0}5J3MCFV$Q5UWPjx7ECa{KsSv;=C+Q${
z&*uBI%L@FT+}}nlG#+sE3D@H3KG3t8(G%ahn2Hc+c%r6qUm~D+Q4fZ3^hIs=qr6R!
ziJ)N@-L4m+nvrCoG`Q|@y{d)?!=B!Hk3!tc>wC6Vulw`yrHnsr4=PDTU~bksJR4h~
zwUR#${TL!`w7Qx7f=u0?ngw;qekg1{B=An1J9snK-8dB4-PDeeV_%OcX9=d|R3U=>
z!>6BBL?%JZ-}UhZ5zl<~?}2zS!n8epS>+e=9_sdX1I8EsM9MdPhdM7edt>RG&{_Pu
zUIAaUP9oW-y8`(w(UAgt+X8FkoAd7-0X8t^z}2TL`|Ln>nW0<5hk~kCK*vVVRGFb%
z*N<~+;q$)D41L4GfKIyeqpwG;YU$8S6#Ip%GGMr5>B#;(%Rox|PNz_oGFJLn_d82>
ze}5+~COr)u6E<iK=aGlQN}SmXBcJ<5a(6hr(z_EbQ8L|}fllTTg%HjO=a~@Eq&)q(
zs%*@#m>NE&qQfwJX3dc>y$eYfv=Z=wS`p9K)qNuLFNWGCk%F(Z1(Mki8*gqX0qMCV
z7=a_oX@pFR5AiB0fw*C7OtKEF2`Q!^>dstni=g0olyJfNLBjbUW7^+EyQ0!jPsVFr
zomr?(GH99}$s@*0EEmr%LflfOPL&p?FN6Sc&o}zc!t9Bgn5!c9U_+r1F_vrUQ3=Lj
z8Hm-+Faeuxv_JAuMmy+wic`33dTc!OChZm&HH`<5i!TW8E{1k2tW#!D7li))<@h4R
zr9C#?9WC{;xBox|)+dT3(+fqx(<{cHQ;$Ty|Ac}CJ7XT35WA*=TMyhW#eQ^X#{?dr
z_5nBm)^H}#y9Ag*D1lYw?dHsGk=LV-ksU@FhrG3?y1MC%3brS{u@}vIRVBL$5h<(#
z672Q&t=u{8@ur^gY^r#(*L%#-9uo66QaIB%9JFZ%Mos~W>+-T#Rq!H%pCV|TFHNR2
zQ-|DD6yBt@bjm9yRtwrO8#0ylpUhV2DWn@K#euJHzLwBUTE{6(jgD_fMxDfcPiHrm
zIY?CHcm>p6PlrE0h7`U63hZLQMxZ$hnd$aONVQ|45RYaQYI@qKYfN8D%r6$l?Mnr`
z!xTuFZo(vPd4pNU_8fd<icGS?IDPn?QFdBR{2QtVfkZCcqYfq+sv@G;gN&*)jciSb
zMwwckHu&=MgfIxJ^o^%%LT$r_U2r5nvLKD&8?*U&?aLdQfObQxt80$2v929JI0iAn
zKBR<Lde;I{vKE~uYA$le!1)81)9;r$k`8yh5i96gzU5n-tVO6~oQMM#?GOK%E{2P!
zF6Kl}y?}~|Es=){1Ggy@DK;QLzmCK?SBbMe3?(kO9IrZSNi0)v*+Fs(nITJtgvq8B
z#jc!Rb9hp&!g365CmO4oy@FE#xMyIe&w;JW#cK56S4~{02{sObRGPd3q#2<>3DD|K
zHc(+@u$Y1ky^e|!?BZn)i4KDLi+%ClzG1MARqLOZ!2#;0Q_n2CuUHV;eiyAljh1#q
zoXxweyqhgV4M^tFr5Nwypk){G^t9te7U1EKrs<zm41URqGoE*#<6vf#{6ppoJT9li
z+(WAhsSDjYhCRH#t>pzTbN(VR*u*Q7lZno~N!kSJizY-fAz+jj6%q~irG>KCq5)zd
ztawVQ6KNUBBy!05c_aR5z0npkMM@pUVk8F8s^l<P(30_Q$*dSCqS?91N(VpTnVt5=
z;*d^uT@qgZI!qB`?)6$w-zX(WZ#~#`7c@jgKe|8W$rk9#Qc1?^@{(CZ;N1Vz6Z{Go
zIRU3N9sGF@d$}L88)x|e>c1FyKx9DbE&^{=alX&WHdL(~woFd3RM*C}`_=6F;67e$
z2NycnIM0JPV@>up1O)eGDb74sI>%FD1X1?mVQ?l)6xd{H{m6a7bgS(sMo7&iUBN6{
zM55xei(J)P&EZVtQ9N?F#c_H1eOt><`<Wr<S=ns3aCY=7+z*+>%HTr9EYThQJ_Sp&
zWS6`79JLdE?XIM?<ZVc=>P$LD<y<!5hbp*~RUx&iWYuJ*>hC}6DykGv6L$A4cM>-S
zKB*3GkdKYR#1+e5jyNEX_((<H<Axcvs$f;tvIiJYdvNqc3vxqi%15V3XdX5pkGR6w
zy44sFkS)l}uxhwQni0xnJh?MbS0^OFA#vjzWh&;SWu$xJr7_VNMNii~$pvZK&hg`*
zac~fW33%4S0hFGU8Bu4P7dz~dSBXzQ1JySW->p5=t=qW5m~&7FAx$>cYhS`9Jb`8>
z<4|~JlMS@?KR2@c>Y}b;!;;LJxmIZM(ZbNi9z#BMctTG7cJzgx!BNN8Q4{7eiA`Y~
zTV=pi>kqKDm#|#~UV!b8wO5qOSE(lD2lAmmpV?ruy25gWl6s6+RIT75oG}orAX%iq
z!c5PSL0iU0tWwODA)QR2ZRvwM$gFVyS5_~eWOv;x#lEA#P6J7Zfr$iys9YWm-|hk{
zJkC0E7NmWB6uk6s<slgXEU^YjO!MWbAxC{`dvtB1=1trMG`0Q1{ZTauIT>MATtIl0
z{$QCi6?9In=>WOn?X!%HBuLCc`tp34EpzmDSbl@pW9gWWN9K*9oODb={fFOv<+zB-
zD;)9u-j!QGTc)7f@(mdAL5Am*&zhmgrn%HSVq{i+tEo|II=?{vuHpzKTeXHyAX9Q4
zx5d@@mwQy2a79csj}0=t<%4@K#L;L?2e?udN8|Ir<aTvgPNnpNfJ`o-LmHBkO-w7A
zvzC`KM=Wo9bq(*~lw6pc{|UuT#Z$qFK;b8@Myl&fejpmSWgIE7H}TEe#J*VMyG^ji
zJ_QyI`)4yRJQvaOrO%Y*nn<Hr^@<v0=!E)<9W<Hw($Qq!q}1dTp>H1VsY3~DH^Vu%
z!YS3!l8*cJ#h3}Kq(fXaSZlpT`_PIB9ljG@oN8T^-iqckiO}^DK`<CE-K1RTHLz<c
z$}M{IlQl<;qAF$N%jO&7?6c$x22HZGM$0b37>nC|sF`B<uo6y?<+fz@_9c{g2M!0{
z99xoLPsc|jB%`?L9|>)etGmGHy1+-wlPoIeb+RO6#wsX|9Jze1?&MZUBx)K<i9S%<
zT@8>e$ES<y8BAx7XOe+PLef^qEq=gxdxLl14)kH@V=HT=^Py@qJRNG=lxXLRe&7wv
z7on?8Mjk&HkTe;^`vE^cAQUf%G)t{ioHl+ji90q9f^FNx4*c?el^pv2RgPiD&Y>20
z9^nFi*P?YH(yjkz`p~W|g1w*FPuQ;8l_JasZ5+>ZWLvL*5c+EfkC~4HraIP=9h)q4
zlMm3=x$6pq^CJ;|Ofrj%72HgqO>$)Bd8YFpdg846qO}aIpgFE?vH0E83C@00HPTWC
zbK3p$BQ8;V<@~65ON}*)hznJ|zNVd%Qq{bNh6L}WKkp}h3wrs>%Q&zu?LgvN)=dZV
zUrct!127M!nOxF)B*E0y#l^zQ>bW5OfJKHMQ{TxFH=T_d?JkZ(j>xqKGN^R*WDXm9
zRiKqMKA+KC7>hG}kge2^EVb+-_^QFOq@bnn-Mu{>M4m%FMX#-qVmZf+TB7e}+vfvY
z^~+YnyOF-AlY%o#T20h}e65^hF(l%1vmHE6c=pdEz2+Cann*0Ynf09IW{;^dF@ykO
z0ox(q@-}L$5WQSwIWZbRmMnFd!U2l>xNtt>;gx~f>FB+Ci?63+lLB^LdM>4$JT>mH
zB=gclVG@2=F<wVZjua9PQf2Xm@!FSMa*zRf{45$Daj1f0f0jQDU{lvltL_Qkd<!oy
zV>buITps?4Kz;^CKUU9P&qia2#qDCx{>l8Xl8aCkw&I)?o|ztQ0&IUSV1%e5#T(F@
z>rCMV0E$5qG~y!4sowN+>n+V#;|xM8wi2dV8FP7uRcn!3#PbdUzYA3MM*Gumb*0`o
za9AQ?WFF(;C{6eFS$gQzxEvFsEN~bug3B};<>+UIhf<uCKFs!+(d>My`uvSwlc`S}
zyAf0ByV)A8K)cz4oQ0%R<)tvORrCvgU#pqg!g$;mXMHJZZGbr;UaKm+o=p!Oiktjm
zdvR}x7iD@(y-o}GR+a=Nk}pn@1}`t>eIWrQU2H}x@1dg)RvcAcfno|_Kcg>QuRA#T
zvs>67Y7CmF)aWo>&hx5rV{KN_{$yUZ>Dw{o@L5a6SwdxBHkWipotipgJBLOtykg7Q
zLq18JC=X+W61x$}oty;$-z1W9*VKx10;8w%Db(~G27BJ#Q9Cr(sxTVwQL{N4A%K80
zc`TMPEJ;#d(V;9aMrq6zlOh5iKFn~>J=w^B3j^EHzPvT6FFTF7)>gY|bq!6`E=NB&
zw#EAz;kYxkPkXjJYc8dVCtW>*DNfP40+~!lK37>8h4`1IU{hNkvaNly+Mud=qco8K
zT>G>rb=IMDX)vWwUKZ{YV0o*Po*CR~EOpcJ#o5IGeQccRD^&xe;7b-iZ$y3)Ll(WZ
zCddLyHwVP)X#zV|t75&`=UY8C#21KXx$H75GbxkhJg7xJd9dnjBO&Cgnf|_gqXvET
z6d5rC6#7gkRZXZk0IJaXP=-j{Acw8s0911T(KQ%{;LW5VYC@wH`O1JdA@Az6azg6;
z2luoZm!nF9ZPbO7iE1e%Y0@;Hu9Vzl@!k=%X1rml(yuq_!hqY^V3dP}(lR&!?G^$1
zNa}Ix6Yo>PSwAgXWyd~Sp}Ef=O3x|bytC`I)`A!8*doAJW@O)rVn{kJ7*iB|W!+Wv
zmk<{yvdXLRez~wr(o4w8SD7>H?WkIUt9dPmU8~Jj*J>JjX7e!k2DZv2ECmN~VG-e1
zk6@f9QL)8c{GH#gs0sY+E2x4xguyw!ktrG}-wP28kq3FJ12lOX)Mxcl+_<)~w0~(n
zRFypN86nlDkBMC1hBqank&Aso`?fVAX%U(&{xh?V2xtFYZ;i_58l}S6h`K;;XX6@I
z^k8lP*3|lTi`zsn_t+dy>gJe|x3|rgo$Jo04~KVACh<pH&jIPj>ji<vj#%G-qC&ZF
z<j9vWHT_B*7bpVr9y&<CA497r&9W~U6}X$KXe?8j+B&8oBF4d_&^+KX2a6#E0MKTi
zCwMECF}dsPxY_-H4}9QZQQ<h`pxqJ4aT)M}xa8#zq`qTha{<oCNlZzcwUy;|hOFUH
zc;8s+T!{p1EIS6sL&9{fs5d}A{B%f|D8AkNA^M|6<@x6^v@w4xadV<8Ts8S?5k*FB
zX#z*|P&FF!dz1PFP%?(`*2K(E)qY(oXmeB5gBmv2q(lEZA{#d})8YQ1B8h^c+f7>w
zje-hhjbe~_dkT{;{xIOr<u7ZN#yC`rjBY*0qtzdCy;04*p&x>M?)OB$=-20P04{-I
zkLGqulyQ^ckeD6VOpn{v)n?1b!y%gaIqbS>*!2J(^LBn4DGs$Lt9IDEwLcB}&SUD+
z;_tLKGl;w`vbU<gT+b<8qdy$_KUig*k+zx{*{|~U_}EG&`L2!kxIvA9-+Tq*VbbW^
zz2+#<;K|=bM%^{Kxw_f4@^|C{F~6d`3*%1=+&G;ks@Y#?>u~F_9e{S(pE}&m)N2H>
zqt2<0Bd&Ek@Zmf?3+<kbHHYH*K_aCp%qHB1nU^VO$Hd4T)Tgd}ujaaJUA9DC1QN@l
zpoxzADBmigw)`*M+b=2p(B^q{2Bg|}$KoSG3TwRQ&Lg=9DY^-dmxK{Iq;%NMEqX6D
zr-t72g>n{wegr@UduX5sMwZ=K5)k!w`}^K=&ki{cgWh6Xw(qCkU(N;G1p_9QihRlJ
zPHwA4Iw)l*ZzcdAfBae`AP2EI^(VbcbM5^kux~&zvr<^&DiOHpvPR|=b#>{eiybD)
z60%`8*>c|9F8JZ-s}J-j$FW}n)jxmu++U7gYnx${A|V3rkJE1eYy4BJ?FnJLHOI2N
zc5Zbl`FE>?(k_ir=>aH-l1Or@GimK14eax_+8I=K@@S*+F!J>B8VHQwCTsX4En7iG
zt=Lilm0|h{Ow}T6MIj|u293tNfRSv9hJJK;`s}Lh>FKM<_JXkHBqs90G8+~@#5fmN
zUAs>9p7f73*B%J*OK#@euC&Yktu*2=zSwFx@7{q3B<{f(@UzotS5}9==GxClv6*&}
z$LUJbz`?52T#V$0!?0OquD4ZaCzvn=rgj2Eq@kug?URWKKAXqijsncU4rjb$oZYhV
zw%j&RLuj-H2@4*?oS(HOD@sNd?vZ$qo8GuzCG(-nixEhA${NT8|J)QcL}B&muL>RB
zrbdfe2EcL28s%T3E1DvrDIQFOnzu*+RvH`%Is@=#6z=ptFA)uu-X>s^t7@6)HRZjX
zi&=Kf%92&pmOfRf7^N?E^(7&(c~FOIRM9g{=cEngXDH4|6^JsUJD$biqAM9ERw^3v
zUby%$8UG%S&jT|+SAQM+o(JWSg76qTVA9_k*2PRXPmYg~B_=cPyfZ!hwt3o2ZlVms
z>KG3n<y1EM`auexMGve<&Y7B|;q$GRmphYoVh22<Wd4|xY9z)+v>{Opp3bysDiYRw
z=k>+#c1xhnoJUttvnvT+Xse8D*e74?VCsHG4bKW5mp*(-p6(cXX0-C)ix|=Ti~o1^
zdv2cQCg~ZB$?`JOIA+4}ahiVOtYtU6N&w{$SFmeOl8g;x9yT<r{7KmW<HHa?&G$CO
zW8hcO8axGg;pmZCb<>r&L1Na?r&zee@zmT(m82wd0WYQ)&gv#b-SjK`@fikO_}wdw
zC?aqj+Y1`&I|cl>UuoG4xx0hZAHo3pOnK->{p#Q4ocBWNLq^;lylBwuv5WBSVC5pL
zi+fAs0(vD`qghDkN!2ttRHT(TaH4m!VfVO`>Ch<XG}9C)@H@q5@!m(%zvJeaeQ{&P
z-i+oTwZrY>5a2);*jT6>NaRw$8ilPVIU4M|ELq+7qcaZRJ;qw1Lc2%EHc@XUkf*-{
zSgf@!5%j+ez3aAWKkoNv+V^v6;Ibj?4@G!fNQe)$`3&cAtH+Nh0(tyshErd8b9Zhu
zX5+QqA?S8=!vzgJ?cST;-pKVIy6v4RjBQlj5Jsc`6{`Jhn0y7ynvX<L+}n4461dp<
zr_Z^0?vT!LKfmwAY!>v4)(nd=V5h$b+P}7&hVpH&-_M%Dz3bRt>_AXw6!ME>eP*P@
zZpuenmoeoPCx<61Ywn*hp{nTcZwtPmv1m&>`5lnCkqw+WO(_N=JWu4$3`Kk^s8e0A
zy{&-lr|@4C4iEM8LOTNvC3>_aeC=HIM-Gts;ljvzGa0oVdT<L_y3u$8Ai)q9$sG!s
z#M(NOD%XEs>-4UIpRdf{wNB+8f#tW?%DZjnB|Aub7;k5%?j#n*Tf?DHyPHckrN%cc
z4ukv!=qp2N4e<4az@bV<0IV4SSS^vH5GB)*$eRFFQRzNGcRjOIWHkh~$Z7Vc#X6cv
zhE4I(;C4L#U5ZUg*xqa$al!g`puujH*JLCInwzq}@X$dg%E58-Et{uM6m`FHALD@F
z`<0gV>yK!v5%m#HYt@vLbL&G#o&N5u<UU_hx&0CS1Coqk7GKWEe#W^j0Wno>wSJQK
z5tvuMqcHy)7b_Ht33Fy1_#S`P!~;M$E&M5Fu8U$`vk8d!Ehf<D_gKfAr8lM<PJF75
zAH#6M=9#%>VUC_RD)vwS9+&nTfKVvj>U0vsOBEiJw?@9UM=#4qvcG-4n=?q=E-rZ@
zWrPl@Tq@~fk8-@+PQ^d(2>eyOyxb^@Z_(4{U5V9=asHVojyYc7I2KatBy68>(w^vt
z4<aKBgN^z<OYw*1(zk01k(?zk&2|OWETU+7R4bgTms<+w2&F>1%4}6fjf$bnZb72F
zqG_1-6V7a*TvcKQZiM@ZS&w0rS%s#Ryyi8rtsC+62UG^ir&_Q$Iu<^Ce-|FRl&KUN
zleYCL{aC6});sSV{|=#W=}WOWm>4hv+t{o&)GDk0ab`~2?ipZIWmV~#QqngF|Gsoa
zIgLX_hE7-?TYg%|vsh}Pn2wP`RkTBMNLQ74GFC!B?oiTE{VCLAf(-Uxn{nu>iwG-@
zjOWFgdO0z)c^+YJ8>L&M<hznRygvPbm~KjHY06b66^lWPYLq0A^c?P0Q)(rPbP+di
zvgD*{wh|Fm=2HnSD2FZ3)pELj!$xEGIyIFGxM%*Zjch0iE1_s(e+)<eLM;OE9hEnZ
zEXii1%qmM;G)qC~Ay@AUA?{^V{^#4AkS*>a_T9Zv>LJmn;_{i~jg%NQXI!jx_RKJZ
z@`S80v<|S~zIkt3N+6ObS?*&qp2;`>nqPFe#8)ZR#*O)I4YNMv0>ORl3ywSPO@84V
zyq$=ECUcAfv{vIAD;BZ^C7U+Mnj<B;Mu-bT@mnMMvT2ZdSsE7&O4y8OS4u3yj>gd3
zbjR@~vEyK~8EMhX^(2-Qnmo4Fg#&rUGWk8&a0QMLM8BdvX5&rvqb<%ieO4-rJ+Aab
zDPL!WKnE{gxNKh12f|d@>w7zb8?tLqom(0ci>5Uv(7>6R8&eolmdfwl$HLW9kjn@y
ztT|?q9ZsMSGpNT!92I*hLW$Ng;PR9Jz}6yIGZ(L72Vm~TF4jZ1Qv&X<;2YPWR{e3p
z5m3M|Mr@K%AZsmC;tv#xx!+i;gkfZEwx^lmu(q2M-vIE40e2~%_X0>G5l$Cf){pI)
zu}z<3^C7k3r&lmziJm7fx%Q=48C65SuFVotH)BQ|+WdIG9($jePaXz!dTg7p3l;$`
zPcX~32Bi5s?Y5Lv72GyMe$+)JqTfvomBoTi*F?ju%}cwx@lD-9nTFU#WZQ-84r<-h
zb-b5&noSfy?`G_TE~%r3&C(i&s~bo2c#OcV;s@r23yQtBKQKaX&wsrG1>bQY5r3LG
zD_LB1oFe;Kn0DanRT2w-?(od2>29ZgfwtPE;2ESDIQGBsc24cNG+m>OS8Ut1ZQHhO
z+qRt*+cxglwrwXXes<pPN9;bS>*_w}lj`bOv*sM5QCnc|VroLi>=sr{e&F#}6`&07
zM1cFfn;!O9?nQi*UHdhsy0_~naJE>hxQy60m$K;z$`cvrbi73UZx^zsq-I=r+VeSd
z5I`piDbXP}5XN50_R+A|Ij*n7rsu??4Y%4zscl?}E`f2PmU<dhOtlcFgc`fe8A@@r
z63oq=Ex9ViK}SlfwRdr}nC)T)M>B>oV7YC{SzEhgXs#+D)1bv`5&0!=RHu7Yf`OoV
z-?0XBCll1Qec}eCN2nfC!B|vpwGYmn?JrJx(6d!WO+ji-W=7l-q0IYCVfvEf_=e&A
z{y<WS8uG-$h*K7}85f9{y~Yeo|G-%9|Iiz1s?S_Wx=8Hr&5T}4Yys>OL47F#Hx9E2
zqaB&+?u4a;OnwXScGuew(bB@CZjyA;P1Ym_8{xD)){>g{Mh6a<6eD23g|C?j@1n*=
z>EU%H?u*xKKE9T@!395Ld_rI6E>wHaSD_YCOru&d-BBk=bb2!zQS;BFLAbEfN~Q}!
zg`fqgP{Tqm1Bqk`C)G3yA-{Skjyn=vVuSOHMIgXWP)&GDW-`WZHO4jEe2cEzf|q~5
zv9S04Tn_<`mv!NN|5%bZIX4(C{13>*<tx#+`ZxC#^QacLS7|1{f~!iyl(+B_$DOnA
zq`4M;R+VNYjG24M?TQwyksDvgQXXyX+gp02ZAMzghy@74qG^snt_%^uDhM)zgwD;N
z3>K>zlvr*<QnuAr>9IX($9)MQaI4D1r%0cmVr!JTP|{`F<ab6&rrj{E*Iv#p>e5dy
zhWk6fOGNP|L~ek1C@pm4rGjihF##LdJ&~Xv+OfWMAu#YCkmOTldW0O&{w5zv>^H7=
z0&k;#(N}_o7@j4DUInnK@@z>TV^_d~EtrzDL2$UrffFcU!o4Xn)i0b!Sl3p<b>JYS
z)NETuL_!2mh{r|l9SVo<CHMh2glf8<Hy17oQzCT@IWUMH1R{(R;C><q;jK^0eVU`7
z7Q<u9gL3LJSkB~892#!n<@>feOEWyVkWBTutbO=*!tEb78`|8sQ^k68*gsG}S`29d
zi-s^HfDrN{EwSGlg3Q47Q)O>bG-GHI?k$$k-pC<pJs5;ss1k9lyMV2qR}lqvAOScr
zQ-jE>n>f=j{UhB;r8%EM9n~cKC=j}I#4m4|qcRCKnV7mA^SHP8Daa@zcfTPFeW4&-
zbXK~D3!~kYOhl(PVnC_Ug!;=rvNS8{k@ig!7uH>oP;S1giY_caf((5yH7xQc!X<t_
zfTG2Ob_j#)x7i`t0jJ=j<it7u1bEdEOTd+D$_bK~$-|OgAg*MNojN}U;fLf?MH|4E
zsYpl_W5^|rCZJnVz~N>3Jp^920J%)|h;7`a*@u20ES8E|A>%oBDl1FxL~KYpR@4zC
znJLD?Bj3%)B+694Bq%e*LmN>oQ8=lXi6=&HD<k73=3z`^B372*M!QkFi=E8ye`H42
zsym@cR5M6xNTq3BfQueHn(JDXk|~$OM^4R{9#fdY8je>hEU!?xVe7`eI_4HJUuUK^
zAC%T`L0PGU^=K;Mc|9_$2?G7!1s(`W2m}WN>IjPDRlVr5pb}P>cKM6TeZW?B)pd=}
zVdE)rGsEJ7?DUaAEqhcXou-H?ni0yT;UeY^BR2q7Oy*C{Bj*j)H(<OfJu4-Z2qQJf
zD(@>1K~<6JaK)V~t7?-w5Z6ZiV^#^eyjs|o<E{^D=>VXJkTYju!_Ju$4m`%h7hJ`S
zwKT#;<%~}h`Ge~S*{|e>(~GrkaNhbie%_Y_5eJaQ$r@Z<u>z%UcrYVaMMD8&CY`TT
z6}vw*At@u0k(?SqEuX89o>C_HmyZRpqY5i1YcMDY<4ix@Ps%z!JYSuAOsq@s3JrHe
zy43?oL7FBGzf3}PW0+X<(M*ZGuMEa@!pyB8K^bL=^VBM`Zf>fexClkIG?;ZkJ6>{@
zNVO5e!JGF3=eM<NzRE08Kl$3;`2G{A_0o>Ikg3u$tx{?!`H<T=RS6SqmUgCO>Qup6
z`9!fpHG424kxPFrSDcH9A|C3{^+1+|<Saq5$L%A}Ix=aveg)gL^)NmeYeZS8b%%Gj
zvWrcb%sk9-4&{XeUDfg((Vm$CIqWnn00EJVBF=0|?$}h$NC=})ZemJSMs_e<?9!iv
zzih->CkDDPmOth;F}rkOy`ydczTA2eoIaCPxmW>iWlU)fZAt2hne2nvi&xOl4p-Hg
za*5bX%L^pwyqSswwbPb4dB~am@tOOXV9`GslFhhqPAyEHoJm3dFk*6+5Iu}ogoRu6
zy+#c6(2HFJ_eI@=gxb9<_AyT1*}YPLPa|JCeGXGjMu;vcF@yL?zUN!Udyp&g@>kIk
z3!ij3v2Z}8M-E4<8m$>Yqt+n?f*rf*xPISATv7JIpM@=1WM)A@kS9y10F6LYRvzjP
zriwN`NS91U{DM;WR7f25EWK1!K&;%YxORD>Pb>*%U3-O`X>}0>Wo43bCq<@wVM%`Z
zXYnG!VkRD5R@h0pNU;Jwwjh^BQEoDWVk%0q=l=ZYBS(5{p)wB3drkC1##$AYR;Hd@
z-J(M$U4#nC^m5RYnatabKBd|moGn-myF5NRSzUlty)IuU3wEO<l?2MQ5p_UjIF0zW
zUtB|2*FL_+;V1$NiHQ!rzB@K$q;RU`Vv_d}{L<~AuMT@ffZa>`MM*2TegB^24Q2|&
zisj4%^gw`&?u2&;3j_6e$4`Nnq>Q+H0ejpaQo?aX$;z4bHMbf;xA;~qsA9ZgEHTyc
zu%jWkbV*~}T*-V?#&dBgVU*W(6uI|}VQQjF3(`EYcfGs?K3zo`Bo2V>N3Dk*Q)I<N
zhD0llH6@rwn_Et5Mdd0>jN;{SzymY{_Tg<XJ>%O#gX!g!anyA`=Krh0cKHve(yT7k
zbhDL|o9jc%CLGBT5Gi%hVROwdSBhFS;e||s1(}QzHlKFPW=DXe)kuWT?Je&(ADX0s
zM#?CKDnmxt$;QDec84HFAZ7<~o(^!;h7p&8UmdGT!xbCRs;2I=l47Hi&=O5brwL+y
zu#I^rS&B*~{&n<Kx|XCaA9u?BDYq~`n@NaAfR7#T<@x_!)t^xMTO!~jr-S$n^Zb2#
znVAIh;n8ODgK}`Si}vx7^pkDC+5h5HOCZkk$%#{#5MI}>xI33HnZ3nd{p!vNz7J%n
z6Bz)fqloPBq4}O5*ww1u@Gp*+(xj}<q#L9qI@>sjC6+rtatSjLpnTXUGcOieDqlqA
zV#V#De(^2AFGHh@B>DI683kk-6^fKH0CJRRn~gr!@@r}q^%W@u;#w$jAlw{&;K3Xo
zS1E?2PV+WmXJ;2@rUx_cln_yGC#J1L?Ty-`cn$q(5@vDh*HY)ll0Et#kQ5$HYWamV
zk9-=Fxzz~1q{^hBZ+J=SD4w|L6p|3H@+jy+j1t<9rFK?6Uvdbq3Os7~(%e%trL6Nm
zA;iHX3y(@tMU20gK!cO>gp7-Xx=+r7Q=qy#@e*)J$_<r^gOS&`OY||)X)+Ci5|&BY
z=wVCHYq6<vzRC=7>d$3KeaTOc+BPC*7us6Fc@@!|Qp_2$<}1=%p+55m`BGEAW)Sw&
zO#0PloCh&j(q#Oa`D|Pwdclbf!xeu;U1TQ^WSly}@S}v_<J2lOM7I~AR_)A;$qBi?
zFdy*9ExFHsyR0^%e%*<<YFSdkvNNO3q#h-wdosg+3Wc&f$@SCxaYq5S@LSt~JHd!T
z1i8by?3}p2JQAi|w!*^6P@^5pH*3efpbPf_G#0si=hTA)*!8^P)ayZQ1<?n`#u3~b
z++ZQO1l2Q%O6AE7akEP16<gbC8Q@$kmx#&|%VBfyKy_^`7G!s2&qJFUrO3?2U8CL*
zL+ZB&R_}u#t{r$|f5A>(k%M^n_2ZS3Yq^6Q?20e>gSC<zSgTDookrHMq32z+MumuG
zw4gLJ(KS<~#^p5O5bSt{)J2#~)*n9{O?)P0H0gRmg{Uf_b8wa_M5ZKfq~Q$A9mZj%
z=ukF-UkFCI@ReAk1xOt+iC4wQ8{#_h%4$wgt<smvQ8Os#K=E6Ad50e$c(Y?EAf|;r
zjg(qByJnhbj?}C;hX=cL70R{9kJtL)uuU*%Ld)7YC_7ymaWE%0lTAGflUr9Q?=>DB
z3_3QZms$)I11@tx5NrF6X;^8{D4gp|p<cmBCBi-#*8G`Y1D2e@aYnjQ&(&Q5)-u{K
zTkeoC`7fX%tR=g0u~hiL+&6$fNu+eX!F*z&*`-932|<<RVGnA8t24^;ekyb^D-pm<
zjR?!s=<!2T#>#*Fq6#Slz75}j6jV8OYrWHT;XlMr_&_2{Ibu4f`yfQl&X}b^YR!kp
z#L7%hn87Pwb)qM!=IzH_#GIL$N=L3#ew2ZqBb=EX26hU~j|Do%{m`6LD5U~SDvHYv
zk+7rydztAVYy*X5^Mj)3&F;+*S#zF6QE(n5<m&KOgITgGM0$x5EgQK8tYvyblvcJ#
zUHQd!&_e)(wnl_ej`lKE4Pz(}*@XgJDvTDe==@}kj66@gZl1*9*9xc0ItD<WjUj6o
zKX)d9r@N%?B#&OH%Vaz3q)x2O6dqCokuF`X%tRA!9T@&dubW&xp0M5S0Z$5pFOKrI
z6m(5vQj;0TiA{Oesb~s7Av-H2$seaFn<_@d)GWrShZML$ssqluWV7{5%lkdWMp-$>
z$}~TQe)I#mo&m}mMM@JHZs{Xid=?UriLNThG7=n0%b}V^hxQglUSf#q)lcQi%%)3J
zi?xW+bz8)UISIFW>g1T%PD-TW5ttx^)S8GczfnFLP^wN`fka~n0%O+<{`$(yu3HYp
z0Rr!^@bTu4d2LJAL4G??P)S&VMSX|&FDQ)qnfu~bThSzR|Ci7io$q-piH|8D+|&)4
z(Y#XNgVfjYY`xn1FntZL&w1u_YwYfV+Y-PJL-D_}p5gkdSCwPEhX5bsF8J~dg=BcW
zY3YARYZ;z~`I1~7Oa}UXA2jd$3ZB=-@hl`2D#!I5B4g~7`{rLdmsEQgO2?giM~#sm
zqPzi$B}C5ndqBfZAuEsAFaA#|$DQtb+q4p+jm}2$Mfqe3Nym7Hzk3=Ep%=8p_~Hq(
za=iD5xXnhngGww_2>(1s%~3>zMXXQEwnjeaogO|pPV;sAtI>)woS@boU#oH)=B_nr
zrH7-vV1~JH$}tNhH>Vu|-i8Pn(+4G=4Kus1mw6nfWX21F!ulKzG*EoKg#y{i{SF#S
zhs5w;-qBSgV7Niw`oxR6LU1VD_)N8RSgMt9`&vGps~zPH0~+7k>cCUrX}57N@Mffo
z3DckmVZCr!Ii0J46}eE3hXY5rI$r&qL<2qGm@HrY`a>qnc+SO+%d9JV0c-BNWwyP!
z^wQLZ`{#i79~NV>h=};wZhHt~O;69Yo3}<}KxxyI2g1K=(T4knj~~DH%VF6!n{k1Q
z56mc|gMER$*O#+uBuN9>mJb27zSWnLN{vrPp+2EZE8C}KiG<B<ub1OtmmtUO=cOL&
zDnddaVv@wL(bpX#+z6&Mx6!RuYy#@{DubAB9=Av!;nf4fc^9DGP+Jhtx-JM~taan~
zSt@<d#wk+76Gwb+O}wi%_oKY}%ws(w$Dtax;q)a$s!V`bnJ<I?&igTu?_GxoQ#(<!
zVRWtmHP^72vP}fE;gk*jUqbA-J0qO<Hau3?mo*(oo4=G!E-}q(li6Nicl~QOYq^4h
z)r~5T%FCw-OWN1@Orh2~id4NqTAXpQE+j0d_4{){M>(Ww7bISmjys$v$!4_3<0?F2
z3DTHAfb76;us>{|>_AVn*4}ZhO1Tk^xY&M<#BpJIgSHyElejP|8`p$KPN<n;M3u>F
zsM00gwV)BE-U}0(TGC`o3-k6_(v!I2#0tq$)M&;99BgWRf_hN2NB1ZvVELkb!l;}|
z{{eBb5o5|=lzRR_9k<lVO2ahDtR<+A?=-*as8^y6dif7XaF}IrBuFlIa3T#oh84nb
zc<-oZ#g<ZUFx{~P4|7PD6xr#J-{eFXrCFYf7V))U)81!qC(`E2-pfAre%S0h^9^|g
zZQRmjiK4hR);(QaF}gE>R~bKjMy#BIl{uv<np`ENl!e-AViPsInn}%>+02{-SzmJ!
zwdGwgRTgK!GvA){s1WaLe@b4&rGyPJB#j9>*LAF{6A;1enQx1clRBHRtP87ztg8x6
zbN;G=UP3h}G(4Hb{~|$AP3^=+7N|o-9$BSyx^~`CP4<>8L>q^Mn@mikj+GOeo7thl
z1^l`!Dux)rOg+$Odrl2d{=(%ByyF=fz(z@6w^q;U)Gx_5nt$<$8F4-_%?RQXu7ZFp
zuRu=PNYT<#!FmcqlVWG%my=DKl;?DzgQphU<w#`?gGcSbY~&5aDkncUner+O1Xe<S
zI%H(CurQf+9I=WSHh=)tHxZv1jY+uB;*bP~5h=Z26_YqmQ<7hLo{DtECtXI%w?fWC
z_4b!%Qidizj7?K{8cEDcp(q)jf}cdGqUvO^+VI>J(}PB}j38=8WFw4l7OItccF8fG
zwrrR~s<lb;>NkzAUM#xsiqgWuvg*uaaZ!GBUVQrXqA!q^nv)v|IgKyukB{UOL{Wgy
zY8XU8sFOY)OL;nQL6a^6W??lR<$&F%Kt?{$XWkU)41vn^J2fuf844>9U<~q&GIe+Z
zJpvtoj<+<4@*_xOxh&K<V;M_@61TD9yTY~a#y!n^EQv{)osa{MY2sD6&gH`@b!Civ
z;EU@D^Xf2qpWt!D59)!8S-8lHr7g7E>_x6BhJ@C-Io&M)1(r`p#(!+B${EMx5NI7e
zlL3ISt3ZxF)q)UIy*vI9fm7o^W&gYPW#K0lLuItj&+(n#FM>qN%j@5t`LX`HdYoT=
z8xT%4MrQqQ?fJS}%R!R&<+XX;`sP`$_ooB!(yTzdL-^F@dqpN}Sv5QD^SMLN=wjYa
z2NeX}pT*2Ow+?$4@T(&=!23cB=Cx}d`SRG$6{Q`%%mFS`uB+f9?Hjx!{GRbZ4xBdH
zbi;`}SdSujS?nisKfUay+}eS9h(2iIu6eZ4gHGYt#t#lM>A#?pV8UAFc<y#q3}FJ}
zYS72E5qYI9S4{SrE?e%n36Ag9WY#?$J0Qc0h>1Zk!y0F^&dSu5=<P&2%KbNL<gfbO
z-t`Z$4fsSIoKF1MH4~7OB9Au`t**;@A5S*)!=8V4ORTN0okR!bymK*Lk9dP+T9BGt
zKA)}u;J$7jT>67u=a2?t3D|5Z{Gjk}Ojm;jrksCVkUC%F68B-n554=VE7m<CN+ZWq
zPM(3q2b|E!oE-!-_@f$d#iIS~@%{P@+y`8qeKI3_g6`|<SJ*)v6XH*4-aEX#))pjE
z&dxtH)8^eiDN@cd60jR_e=+8n4gqAKuwDoTLB{dVj2ZWjWG#>T74&DB-k7xrZcaBl
zXzqBI4>OTtB>NK0v<{r?hxi3}vJ$JgR(pz#u#Bg;y|M}zx4p+zjAMz7`hM@=VP|Y@
zA^?vBiQ32pzQ6x2o*|h~g>d*mJ0LzzDm;NL9qoInUy6yJG`rzioe5;2t1(f)PV5zr
z;1ir-dC>ELFD!FME1lUMKB8E4AD!0gu{m`Q{b?5ZFruI~ILh##B=9kd)0h&{y1$e#
zgUt7JIE>}NT+{|bVF$^JV#Osi$lknceIbn4Tjn);shF&lkZ3cF3vk!N>QaM|yhB^o
z4x&92kuH=?r!VHR`<_BomYPpkX2)m?yf!K&@uf+JIsH}4Dq|f4Z2}pDhFjef4?xj(
z!VxX;Muj^-bLeAAJ~)_40?%L1JJN;9FI;u4U0v5ivfDZ2(A9us5?|n;OI6jR%xYPZ
zD!{<0sgOfnn1xg`ky9?RG*7@DN-Yj%Lp|$;LXf}CJ&HVx!AJw~gPNBrpR^0Xr$lhF
zjUl53xneKkoM4*9fA?Z&8eI~M?Qy!2D%?m?h^(tI8qr2v(v^OyY*iHJlGRWkRKX!(
z!uvPOfr%PfHSxjj`bff5Xke|lszOAlGok_P9%w_c61Wgrp&+08agw&9s3gQzr3NC8
z;T}A5y+TkpHJh=(wbc7IO;ipNaPEBP?PIra*0`*j$mNIp^~MVZa^|N9rtwRO`>JBB
zY+^J_ikK9;tg)J!rl$HjzyDD(@>wZU<C$q-(c@`IWu<3wgdAAG-{^BBa~>t*@1P3e
z-Y!^UWvTUj`Z0XrT&i&a!Z%Sz|9NWtx`MvZYrKl}9&hSjN@{KSgkhRAutqnKJXLRN
z-x}45xjCy9PAZW>-%1iTURCAe&2)PUYh!KJVW>!D$7zP?Kw0nRWH}aZ6Zvv{v{Nu*
zLIZEaU;J;aI8RA-N<5I-9@_1LM_Qn!1|8nKJ8z?HY>$P|zSn#LggPqcUxct7U*LZd
zXWZ<FFyUJ5j0WU*-#E90Xvg0?Fj_6d+z!tOcs@5P=8F@!UhrJ4LdSTft_=OUzvcf4
zmW_<?n}#%dzUX>^rz18M7ys&o_UnFMt+ER%t+@R5%43wMB~0Kz2tnrM^&NEC)Qvzl
zcy;qI5&h<I+5{RjATdTxRqBUXV8G@WPSgRiqhXvXFk|qNrHPC8?t#!6`7TTl4woPI
zqH_HH!Jkf}_lM&X?9q2U?O^WapUUqy9;J`>mT`&d#qS?Z=K5!(FJgtafOkD;mJoR6
zXB^%M|CG!VLBNa$cqY1X!y}?LG6(+txe)_n#o^Su?B^-yKOp`zV}`!->tT2ipZM;G
zF&&F9F2ikBeqfA@ils_f{^#B`(xy$GMX1{(UA_gkS~w*IFD~O^>XFb&@JaSMWh|=l
z-ImERHbzy#OD7>u(E3><G+J{)Fd>TUKkk1{FwdFmVmD$L-Dw-d2pEm0NP=F_3aj_e
zYWh`4Xz)IMJ0oC$_x)xwW#RM(NE|556wUNHE*xmsU_t%^D#Gv#QN$H1{_{X22?JVi
zBA`$Txde^<siJpf#s$zLty~YTk+?^Go=hp8RktqoyaPP`)m5Q<wt$&DIUEipml$*L
zx0TE#%kZ>FP?(01ia8K3$WISJ0s44Z=&Wr&F)H@*F_Ey)<2&yxc>#&r#}6jIoFFxw
zQPSvCx=It4p5Zz^Q<m-#S4-RRVU@rfZ14gl_0>+iG@OJCgynjS%Fu`Vixck6W#6wW
z0UD^xER#`J90(G63PgB%tKoBKdbQd+^wU%8b9=bL)h0KS+s$~95Ph;v;>j2h%kk$+
z(RgwEw*&Bgt!S(ixxs|J-54TzZCQ;~EQi!oc6FKEIWLVGT31KKHBU)>(QNtQDcY^5
zhg?qOGI)8qJ(UKzLg9;w2bsy1l86kM;^WU--xsN^#6WNGn?rj6Ji|>@zpckg(UhaF
zU{cHLRM8gNU@OM#_;$Bh^QuqB4hP)0ndTi$*^2kfP8*D=$*UCNpmn@J0jvPto5KV4
z%jVSXXO}}a*`8^RS+H5tS^OaN5~U6;y3Oc{yADa-r&!%eFS}ToUOq;~W6*n23N75C
z2ZGd@RG7-(jNQ-+Y7yk6ZzEUbS*QhCk(0N2DQjq08FY4PO6dd%vW804ai$D@m1<d)
zm+|_k0?rgCRaV3gWOnfNvQ)tI7r8&`tH@FKps&u>On*On`ccDm+McJTJF@t!_H8w(
zk4S5~_~agsp6^EO=~JgeL5{}`hFbhZj&+79I#r|pYpQ@<3RVj2?tegtKZLtFG`_Yc
z3oaYB?49kg7c35*H%&Jv;L*1&yXZPyQ_KqEty=9C+IA>*h*K>5wj{64{;DFLfGatX
zYSyMCv;p3#7y5!F5UGm#fppFa8`4ACI%8f6JbzkD2xt-F4ao(gj;H#5cwCDDr?EV*
zeIP{iAC`^0xXtgDHrei19y9k88@flsaMLaL>1~hFZ`(9_W}{%Pbep;<>pU<dF{p6N
zqCgQOFc3+ai?*xTo34|nwY&qP?vi0y#+i2QV=b~_F-fFNa+%OXNK%jklQ4}+R#S?}
zRO~|e!zIE$`0wJOt#YEmw!;PC8nFHx+~!oDqAW3xor7<}b8{mR<Ts}T?wGfH(OXnL
zV&92bO*#q?tGptgOn3lqK#Kx+K8>l!KdL5}iGwB~yegtkHrjhJ-v5Bypdy95^}<I!
z$MxAkJyd@B=+a$`Ym9#ZwY5+M#LbZD0zEaRB4485uc$D~)Uuhq;#JzDOHxN(MHk4$
zD<HqodFHI^&S0bpxxS?d7abW7tL0HWza9e}KCRUnkbb%|B~XkfhyAP_+>?FFj)&NY
z?{0fqv|QDKZFk*P-Hv^`7VRrBQQKV025l4hCaF}Ds3?17v)LecZX)>zuwMC=t-F=-
z*`}Sl4R2JVOt9+3(}{E%GgoUGSZ4}JIOJg_4Lzn5(=P=DZQi?CqLQAWtWiM5ThQtU
zmiiT&$~X_%M*0}e+z5xBYC80i(*jN6C3QmB@rzrx&UfZ5zBFE0X_4tPDe0)2hlvH=
zm~;38e*QXsc6uWqwA|QUA!>D|b#n^*^SeCBcYCpT<MZ&PdC#C}`|$_!=dMA>p^r9k
zU7)KE?7i)M@y;TJoXRo4y&rqbOUJ1F2C<h;1GT&EHplckn1`*3$hKMQsOf7;s^8!B
zh<t`|_G{H{-4uXU@FYqskmYEP^Fhyz<anSjRnGVxF-?(T%*Ov2T{7tEmJ6MxqQaz~
z@g}Y&pC((c9s`<s?Byi^;n{E>Stm^8+qOO7y4^$M_wWKEK6o<{zt5v)vp!SZbluS+
zX38UFxv{k?HbAF1i|$!7`j%r7ev2My2PPRU)#5wI86}SQ)YUHyvwM+a`@Kj1)P1Tm
z4Pw|ZXxA)Z`)-c+!18Kop0yQ@FiX6>v-35(qF+`*!$wd+4Gtm&u2$K!qneE5rkW*5
zpM3dSS+XB}4Ga#<#8h&T#2D(47ffhEM@xiuX7rsI$nH=ZzGau)ZNGJ9*~VjS50n3n
z;IiAkI=8!~w&eo~oC&w}I}=E?>;NYrOxMlkKZ3?Th!<wPb=|J-uQq5W9$R$m24CcJ
zW^GzGDAzALKR>FKNz-ZC0ba0dWi+7|a!!iqNa|_$<d^E0D5Oy}lFQLva&uH4AGk@l
zVVu-agpWvE>aeTOl#ga^uc+-qa||eE3sc8t1QC?XV2^cPcJ=;S2oeA6>K@i*>JRuQ
zzqEU6|J4ueq;J|)cE6@r5EpoAvlnf^{k228Xl#>7zRuuZyFcaXZ;Z?9rBl9PkD4Fh
z)V6Cb+IAT4cF0;h-d&9=>|uAGBcBV{MThBg3*RP80S{IRd8NY0ol<@Ko8cGfZ`HPO
zGC~h28R4u-v__VAr~n^boP2z#-?yMf9C&CpL7YsCT^7#C#hM!vKtZXisqe$Y!^46M
zXavy))2v^2*Tvze(iec?r_g=HQ9JYT<=mR8yJ?h;_G`QEaTl?qWyc|n)bVz!TC?gd
z)w+Hxbn0cc9Vdjxt)EY7!GJ5X?L@b^J#SISK@3*S!bpp^-AKwM%AnUhFODZkWek>&
zxvZ(KMJ$&lP7M>PNv(vg6>vv6mU2rImMlY+kRS6Wr9?nY#>*Mc%IwA_#cVi$&22G^
z{GyEKBz{T&^7J48?DTE=G;R1@?*eYZ=eO6KV)Ze4$C^1-Zrn9tJB*?Y+jOqcX~fbe
z4Z}H!V-<3M!!Zw-7$XIA^qySSDJGk?>UG*itaC@@ve)dJHqK^tNktw@W^=$Ma)uE(
zjOc|xK|0lX`TUcg0C?*?l`TO3;9FP_BuR}wk>8rNtB{_v^(*kQDK3nS#X!U;%k^kA
zTuc>r1q$d%Df3ex!AO<N<YJ5DL}N&>jg6(Tlro;@4$k)LCcY%Qe<*x~-ZEM=ZfAX|
z-`E^?Fm^YlH)z0<A2&ZY7g{_*ugWdz`0`UcN8q;1TXk;QyiU!IJVm%i%)(_EwFxt*
zebu`eif;0k^*Knd!p}Nt)Mf1F?bw=@OuNC4xNT$F=D?$6*<;MAworqfjxxwq9qlKQ
zjdB==Wg#g4BA+^vwvlklI8LchF2}P_&0<}?Ak?yvu}B^*+mcPFHJ!K-Xx9;ofsmvv
z%K0Gf0=F_J=cH#PGAgDE5HV^I(&OOD4(3GAD!M#d^}`B4_oKb4*xM0+vKJrLjoAGR
z+16gQCBGURv3|$2Y<Jr^UVky-3;~+gW@iD$StgsEkLHstT6I&Z>{0IBg3)6}X}FnX
zTebo*aNiH07WFV?a5W9K-xn^s|HAUG*uOJ_pMzK2v~Lm~&9`V*msn-H|3+i}jrVN^
zQ{Ml1wG}3Ut?W(B7|Y|$99t-Z)GrX87$q-6Mq{R)v@T2veJ>T~u9#m?Ms@9E=9fD|
zV<t;}vW9y96HBgOZlsXG1(vL=4<=Q)mp1*yW5?PLh)X%TZ?=PdquWEg;h_Q_qcC>>
zHE*)*4KX|HujO&F?(JucK5x-Jo#?iGO)^jOn2evZjyCIaok)~Q`aH6uMY3y|N!V)~
zGSh9)hgd`FlK-0W-STW4bdzbGkC8<_i1u~#9WNIpNe9a`3LPOJ(ctpMF)k~}Iq>hG
zt~gN;)xi8s=WHqoC`awfFH<N>b4i6D4@EJfQ8_1$T5(p8rMZp_VIHun+Yp_mgh0?F
z5hje6`3h`JaQGeh$gStqrMc-o)~LtqK4mp~Ok?=15kEEqY`V+!+1G1?XN=Mun#RD(
zMwpkFRne<?9cPnJoez-+Ov2y0i1Uc&;I7<9EL%p{VOS&WR#mtca3iNX=Ot6g?`EQ9
z2MlvyJB*o+ATS3a5jzrKo)L;giaQ1X2puM*0C!xqtBXUakZ5o94u#1Twe9B;%tRGq
z*TY&_Gi9_!Gs0X6*Jz~FiAWjX9>B;fC%IBMe~kfx`^V$Pr}*m57D5|F_~Zsr9$vLo
zjSh>ScANhUc<waHVT%DXtqT1G<2Z9!)Piiw5LuQA$23n8v)z-+c$<)iLa+S>M<@U?
z!zjnTX{t?@sck=a3U`5@1pk?*2owJG-omka+PhV}^h^dub#i&qb@*2E(KTD}`K_YJ
zbytpA@4hS9wn=r!UZ3}?$+HA(`jNMWvmsh0WX@tQBmJwjAN{VwVQTj`G<DNmYdte>
zkn29ysm1Wyv27b~ROqc0Xw*}-bE#Q0pK6}zxhZ&&>C-*`8^2-ob2v>d_Z2glXVaur
zoI$@}-8`PfjkPct?0Vr`oCBV3Fw-JNk9inQd5}c&Ik6xNna1!mY7}05pmNzQb-7)c
zdqZ9M{8%=TJQL(Ky~D3vw(4BUa>$eiMZ$RjAfZ%7F@ElGgt96>5Jw8a1fBu}SUj_5
z7`Kg?kZ+LMg-_VAPb#!eu;;+$m3BDDMBBB^f`QF7$@J2;Y2GjkX3=Ifi8=x^gZ923
z__O0IUiTbi;ca=YJ3x|-b+Yro)y>(2tGa!q5I0X}5+*zrOd&ypr_w;Rnin6_C~sQ(
zs8u%7iE+8DP0T|~m2^5UudEG@<$1Q!a9V-AdadQfz@3S!3sXXhItq#sxPLZYvv28h
zz*Sq@f5&g;_)lv}g>SXE^X2!}YQV*YAQG?mj5kSmY>YFCLgrZphQR(tz&Sp8%5=`R
z?`wH`<a_K*H7a)LR_>j3GmqB`(<sHiOFr<hbyrjjexkKncqc49COW}V*q}ez1D-79
za^X=c7pJ~thER3+Y+-dZ<jmACW3X$M<e<WIn(T~7Rw8*CFH559^sUx966tB*%Duz=
zS{nfx9RUXIYInCayZ_RC^@0ED&GZ%U2G8P>L+@+!NwX$#r#s2<bodPOppSvlu5xX1
z$r$yLeU8V59r=V|wz&wuLiG*%wqCaY#H?%v|DbI~3&!`3D}d);Ts{=L4O$h~(2Don
zkO@l7s}eDlk|=SoDodHr6UA@)U(BJ@QPKpj>k2n)y|U8rE*|2WR9R*?Rn=$&kOIEp
zhG$QXEa!Xl@r?E*IyJX#?F3b_MrqUuxyg(i7bg4}*^y8N!hFG-e{{gXohd)@cDw%8
zZa+#}$dUMpFH4)PTk^2o;>miOhXtwrf{6LXC)uQXFwCP|9gS$UXO4XywQiP@pe-`_
zRJ`#C#CjzA<YT9udQ8=v<NB-dUB8*W3;9<u>Xuy_4<`#<3Z3w%ZITS47=^nklNhSt
z*ZBc4+McWP4u%pWX<F*Heui#oY|sO}a&f|B>M+Uy+)ZW(sk00Rw;n4z-;=>33KP5j
z#1%{c*IowrXER6YXO0!U&YnfR5B=Q@huXXQ7Jbg-Tn|1$fl>QJ^Jd8%G~kI$_^{BY
zHhIIeeLQ%@s8#DG-74Yn23_j07Ma8vZ=%UKURTy>C$e^aFR>@||94Hs|4~%VIHdFX
znxe;?T64>TL3rv_li_c7iSL1QJx*-5au?^tU4g$O@!2?i>Q)_|ADe4_Sr2QY=QEMO
zLD87>aRb7|=D5Pub$KUgJ9r*V2>T(w_>4V>iSSlsPpc(}(10NXPd^P`-i`;n?@gQT
z0LLf9rC+)ga4sK}EYC9Y$u8fs=3dmasB_02pU)rvpZ%5MKA^W~Lous3?0nnt#*o{<
z8@SQcaeI!d7ju$g3v_OaK;NA426!n>C7!!}wgFcW7*A_}>f!Ym!ZUuF(vo(r4`XRX
z1SP5#vg3uV3lp!QISRE(78Iwr`Mnv%%#1m7iG>AmMhaBwXQ^_s;%RZmH_-qn12U)0
zK7{IovJ`2u6#v~+uLggt5uQGNb$r(!nk~M~lZpmLSc%B$#_Bk5i6B-bjmfZo;olw>
zb2UByzs(NYNC2pcJa7|eooo!wTwvIk!3w;$CxLn3z^u-=Nr>aY@`IS7CoAwlkML*h
zuy^k@p*f_#SakxdpN!IiE<Pg8+m!q%BkNGS4%3>x39OR!L4TP^fZ*1^RRMb@m(j2}
z)EGwLL<}>XA}<5@utZq9(<y)c=n5RX(HHA&2d<#;^ifU(&la7}n0~(vFqR_h`@f;u
zXgZlXi(yZDYXb3?yDHjzSx*RmkQ>#q+?-l?e78F4fit9AV^`2VxzpbSTw2%5!}bU+
zb^sFI<HN^MDy}QnOQSle@n-wZN@m&l=c(7#772<ozCIadPU6L>-?HW(MF{fS&;)E(
z<q@J#mVC>RHcU2FaT2&+GLPu*L>3H$FLbBeREco?c+kXQissLBgRb~JZ26g$3qn;b
zUi)=Pd^?K0Uye9}@k*)v?f#2^KR(<L7&b)gTSkF9bZ5>E4D)1zRcDVQepXAMlt7AU
zk+A$x2et_Z<~;jl_xH>Rcya)^&K+?Yht_59DBhyLlej?uU@8jc5rMe+5qDG1h}#Ue
z%!UxM@Se`k$%3LG?Xg?6|E8ljC+~q#1KK|idTtG6i@?qsMr}Ro_@oqZArf;~+qMr;
z_xX5<TZ4TIKUzPBhv3i8ixiIfZ4(jE6y&#cd(Wq**9LFvQeH1hfWU1W%lX3ha80@C
z5z<|o-<1I0r@bxWC$~^uZzMl&@XXtMyEDK;mxH<!QI4_;n<JDj|L+Cf%W*C~sG4U+
zZXsa}ck1KKG)Jg!tGS=Y5gzz_CkMsV!T?!5LKeY@JtQNrfAb07`*_&W0G|FhC10N|
z%XAayQ@{K1@)%d71qi}pZ({{9s9qUA)6c4abPMK|K)wqi=d%U_oy(WY<l6faP1E93
zVXY_CE?nE9TM#$Y`<BZEcXY6is}#X-NsqZMN<Kd?L7)2f5r+<8$B84(ug>ajng#Zj
zuKX;V(;C6oRdE>?UK7O4`l5W962_ng%<Wq;E)Yo6&c>q7SKH|*1Y0Gt!u@DJ5V$Pw
z!Ix<Y`L8G~iHAMML7iUlG|P!&M=1S{_4<+m^(l+<vdU=h)r_%EJ>k)SeqT39qU}kU
zAqbaNgH)mO{*h81xXT||oG^v}{SR6OO_upz8_ktD7vw{OE>{8}_~o0W^}9z=X{Dme
zNjlpjn;If4pxVC;5#|<V%Q0dD@~pRd7@V8M6MiS01f1t?>}5t|7HI3t!ZyP1mRG_i
z|Lc7nJJ5QA1>Lq<KHi#}BEs+CywQs%zPr{olojV;9t7YE5X5P*Y@EaWV}cSUXQixG
z?s{?SO&wc9nRqx52ajuswC#`T$&~9%iaBr0`P4`_9DdJxHDBO#mglt=sOvX~A%5Tp
zp@!s0f^O~iQ#zUV%1z%z#;6yXO1(|Z%mcUV#NQq!KyRu4mtF)(B$I?SJ^pyOG1*@2
zLNn|l1e57NBnsM#hf<3T>I`>18VHu0!tcC#sDup)6{55kkE+Z;gY6*33jtj(9YcQ;
zsdvB=Ugs|Hmniu(M?2sl9SaP%+PAI}U%BV^_7})NH~7eo6$Wzi=#`Kw@Bv1Ly+3P8
zAR|nxZ`sW%X^s44P1ae<8RlZ>$eySI(#E&Bt!o_{+4C*d)2J?&(QrW*{Fc4r0>Gmy
z!rOG>vn{^S;GnvO$1;6zkdaIizK;aFMAQ%%@AKUIrfG|T8wgx~b`*}U8SfCYG=f&g
zGyI3{ATJb@Hqg_6p98U@`3c&q@zKzaCO@*=p;{kT)GH#WS%VulAobGAbiCY>`4eyJ
z-OVbm2d+AgPhzPbWXaW!iSN}-$#He|GhpIjd9^eA>8_`$FdUE1huR(&_wIv}O>W)w
ziElpx&h{?gKcMX4od{_cD#vey0Qgyx>pcrks&4rFPW+$j%1wmjv$<y8o<Azm|30XP
zI@Y32cB$ES5bV3+Zo2)w&k`fo>0Dli&j2sSlYxx*<$Z4zTX37+49x)wo>!v}k0u4Q
zjDuS-&aUO;b<sIOKHD4+Sr<Nm;DKNsI!wSFy7qMCb)F{rL|q)`g&;jv>00oaj^mwz
zh)C0Ja~$Cx2<w3shOJ}#h2|o<BfdEVsz!ie*rv7TdaO;D-H$8#cISDA0YO*baS<jc
z4U@l|?7RpG={2ZMzp8VrOL;tMOH`NvF1;;N5DXZ|b=Pr-A&zl+_V*!dP{TVs?RcE)
zo3#ZiV~pT~Ofx%MtaaDb*EJ%11Fjl+IC0ryBPD4I*$?Z-!$0--q-MJ5e6?Kbv?O*w
zEI6A-Mu@F<1uCqD@-6jIyzrXVAPmCZoIh@o6~tH@Xh0x?d7-a@!mn%B9^20Nc7;**
zSe{n#*}0fq42FSXJAf^Je;(}0nV`9Nd{)4SDR6HSVt)cb+^^%~?daAQunG7Dv&FOF
zDgTaPkw;3)WD$Y?V30<O>+Md{@j-_DAo&HVG=F>VP{CFk-UtkQJ1Uyd{B**c9{-6G
z`37FSeiWm6w&3}8zI#(g#_#vXcQ*>2A&Y~;`{E?H{*3{2*bVJT#J7<M(?qSjxy&SR
zSHS@sUw|xXn;@vq6p*bBMNaCo|Mu|Y;m`kLqIROCocTwf&`0FykO}$f&E@8_<`+p7
zSkCKMU+svi27a)8{tuxsdTp=EXu{a*U|CSqyVgHIt`}CL@1vIHkn5NOU(2(<GP2<m
zYb*Zh_P}0bXNOZGkL&VBvcZneY22m5!cTb9VvdewXTXo`V`=bV1oeh5U%{fkZfo?*
z`x_%HhR?}Jw#O4enoG>k>#hZV{+|2a!!Bb`#oBbPXT(h9d=>Hi-f3h6?@i(zjzxa)
zSPM>3P%r2*oa-3$`DN#FMfW!TL)z%~hQ+1zvN>+6b@b=MX;BhJD8w4@%}z#-SET(+
zER}SEqkPtVSr#(TZ`&gtU~2DT@gVSg`Ak(S_~--a7M~awG(=T90?Zi)UccT1mD}P*
zU!rOj0q$qy>UQ#n*M25yw)fm`)zIs@&oV*40fI{VZ9c+2FDa#yOAn(q5iTO#A2ecd
zc(g_ck7;UQ3p9HY)Ng3m8BQWHfVbx&iu#}h8|a^N>z9vM)VspU!%EESHrjxGn|<U5
z_fHHc8bjbS>q6gY9!P`!#uZ@X%Y(_{$q=uV+q2cCfHpdvWJDTPhoHYJ2Laf)9^*Mk
zR^F?<D4hB%#O0F&)iXa<g8v(u$zbO1rC(CyY29Yvw{GSCI>bogBvk)zjE&?cGQkLC
zb#>n5c(^dGj-_sY;!_Fm`fblCkZy#t>^s^s<+FnT?bm)08|+90un~`p&}5dZMAqz5
z*>-Ml&mcER;%AannUPgmMJACarRgWvOBoOF39Gu=K1-aAtBNu0AcbMOrXH$paN4^{
zn*kFI{MurNaXZaXA6B~%2V%l}x%Rc12b{~mGYVF4VErf?p_9%f-872J@g!NP=i3Si
zmo*;&!6y_NN-(5ZJ;yiP>a#ctBBkJr+Vo1yE{H~wbt<W4@pz-TVfg+8ney_}vi%+a
zjIeFQTcbKUj1%H~(tK%Pm<}e9$znvuvP}Lo-8PA!>{iW7Zs-Qr$V1c=WOCGj=~4q3
z$2Cc-;S==1Z|;hd*-2TMWffDd&DBy>><dK{md9c==#pxhjnpn~(&2+H{p>`_GNK$A
zqiG3<drzT<9|k<#kZ2Y|c{hP!@z7&7P~NqYW+br#jQmdq^(*&_3-t3QC)jPqI=CP0
z0$J&IC4=r_yU(|?Oc&#R1qb_=Q%r(k!vBDrBYp&#)Y5(_jql(?%^|`SukcKJq@a`<
zlHQ_DJQ0^ioF|3SVSh%zl?sH%_!Y-s_)zN#R5h(Y7vG&cd9#8_G#iFU=J+5OJ`Em%
z89s!!65<Uaem%5%+OKQ(Q};QC3V`=NPqP9@{K&RpIhJHQYqmXvtvBS;5`H_cS%G+0
ze94f+J3tZ$kGnGn+iUU@H@O7qdTZ@EB2Ifq2MD)zMLi~sBm7>NQw>{T>8>5l@kT`V
zJk{v4U#)+u`8g~Ef@7%9q#@8O@sM)87f!^356p|ORo+dKu_EfekXLTB6I4>49E~<M
zA__CaBRRsL!Rn8`Fk(LcwD>^XLreT)?~VB)^EG^nYhV2H&!ztY<M4G_i#^>g{2^6<
z*1a}weZq&rbt0Z3S}}Ts9$O4|qECG(VDpt`%jW27@lGLfi_LiliC_ldRxhVrwgJE{
zs~P3KHA^$PfJ&T%1SSDQbVp8FtZbdLNT4iTnGUJIEy6@1alXSXr7D~~)Ky9)%XrYp
z+l4jaEe%Z5O2=8^k0+udB^vAL#x9ttgeyvxfUH_&Nz_ro!0_Y$rUkKaU2u?NR-2>o
z@H+Tr`KHtM>D2OXc6!U5#G-BJ?XC4e{%PZY__)tpCw_9*Lo>$Q9_oJ3W1VzYqj$|H
ze2*S$b%LC-u3N5no8ie_AIHs8y1R|veaGOR-il$f&WV28R@+Q-J6MeQOVt!?!7jRw
zh|Y~83b(5=?bB%%DS(<B<z#SF5AC^?(?QYOuwg=z%8p!Exg!ldLqtmv+POGaae<Yl
z6+69>pD(3!`b=_Rc0w3Zj2SWZbSe^(FexVmtdnH^IC8Qa?{#kEXH7qyF6RsROY=9G
zxW!kv`Q~n7W76?MYA3cve7&_($ivY3&*0U2$486yU)Y<)s7aePNz+8DzDDfS(1`te
z`UC3c&<5d80KlhBbjxGpH`QZ*)imOs<F`%cK^DG%`&l&d>9=F~-KLgr>TFvCwD=c=
zB!gHmg<VF97>c<_Pvv~IfT+%Mu=ts|Q_W3I4q%RaY~6V#R(+*fGe~WNtcb`4Czu+i
za;8vrx=U+NWeK9$ry87(A%ig(MJz`8*w7^!3I)55{T=b3WYObK;ES~75GjE0%6+z>
z5FLP3e(kpR){O!8K)>hxMVHm$!{WY&{3<QBLoV5Ry|eH@{MItt)0Vx&zQm06FYd4I
zYL3{^r(d#J!fG1`7sl-JjtAS5*CqpA!A2wl)gC-t-L&b(-Gg!aTvN^{*Mq3pRxJlv
zzAiH@;x82=1-NAptRRI^?i73NE99f}?lu0`um05;;w?WIaD?!gM(}0OPVgU4%H+4s
z(Ko^&%zr@RI|AQ>F#9G9VCz2RczE9gm(SR+m2jGIH0lUss4oXtP{$`WYIEvE=VEH2
zGh(V!`hi^>nQ9V>@>!!$OXu+uax&>6^iW*F8&Z}A#rG^nYs0-2UvfXB_Hsp=H1_k?
zVhM&n9J5{iczca8ONMnm{DYgtZMJdA*BdaZEbQJR^k*6F<4uC5dh<A<S#2G)2K;VS
zuIq1h`s2FlaG~olYSAOqW8Sj8uj)8rFQx>|yRVsdUdDPy%zh^f?i<WBvMNWJ9_4xC
z2kB(D0D4Rb$7*^!N_C7bd2f;;qt%qZ6(wvr9b@G-R@T;8G*W^(WRT~)o$s*GBa2>M
z8mSKx@q4j$<c98F50ROpSQA;5k@tTEU=DR)9(WEhM*4?5j>HL~w#-^|ywHaI|Jd;j
za#+{@p5-tf^f4errr2+dE;!@3ZVX*y*moaf7#El~!0=^i8NM@@Ol9B>613{C!Ob-H
zBk^P6gkLx6fvmFE0c^tWnx~ooHqBC<Mg`<BifOO{En?J8(Y7oU4$~~~UrM5)q`taH
zes+tjB&B+KF{s}R0!6C-N(*qR!x+_;i~P(}s3R4)V}mFyAfrSz9q3s7N%#<JqQOE#
zT{_>NIG~}4WaYC&(i6^n{r&?grI8p`7<`$!w}0H+>UMC?xVO58vA8X6)EGH&GGHIK
zTlIM~y(0ApTeE4DOYs1<2wwA?ZqX)Nr(N|SRo}98KGrhMVMreju-NbPC*5qz7Te)J
zZ}^hi{4^{6SDGzWH$^)cKS`8Hx=A(@{_A|pok}JQrdk+dlujj%Hw?YHhYCUEbE!k-
zBZc8*U6SP{E-5iLV-$;MiDMF50)Q=|4iG$g2G<}*2|JHlpeg^A5``L$Qhj<V9VcC$
zu9@&zDr=mQC;f65_861<`{sB5v+$<d(MN5kG5q=ASETwMP(lInU0zMUd-{xz9b9VM
zUHBFX_j}uBQNy+W%q9`DL$z5x7J6P$ubEt$6*8^!jXKrGdR|JS0mm?opUuW+-$5e)
zOz}Ew%&fa69rOB?bEkf45r($X*zUwWF!H+Th-uKcc?4Dw^AyX)@M#J-^|8fRI6`f5
zyYqY&P{*>W8DI61&Zkhb(vBPjHX}-(m99ElW~@@=`o4m!m9yAXLfgrj%etKGBt!Yt
z9iLtS9oht$GKv0cqxUCv&oTVL@D*I|4eAg;D?sT3*A0Ma<1s3|P9M6)80x-!-N=!x
zY!}Oajj+0DPYZFdr$JxWVGp<K^7W`Kdg>|{VJ5Newkv3zM&DSi7hms&y9wC-S9)jr
z=Hp~%R`<(DzSFG7+5EQ&LzlltPie=rF%zsBwI)_)w?#byn+cu|g&wKt+xhsT`K~DQ
zPoqe)s*1dUP70~j%Ra86ij{Y=mr%A|c=`Cmh%{d$xk`?(Xo|p~R3F(T>$wti(F&@b
z^cwSVUj!O;8B%G-?>BVg>&Hr;N$3qg%(3#fRU7oT9a(EZu9LakalV0=*v&z5XubjO
zJ8zN*&_39h+V|fvYTm<0Ebx;;$~>9=KEic2m!Jh8eKG*3JkknG?7QhMY@2k>)(=~?
z%(QDf;__co@Z#mlOxdLZ+BOm<^YF68PtwUP3*QMv!Q{~s@F8XKjg|k-;}ROv`#Gmn
zub@+}78{lQ{cAC;Wc@sppNlauLt~DZMh*s5bU4C`Y}G0G1a(BkLZ!l`l$`6?M*Ezk
zI@}%r=$rqtf756-K>A_#WoowlernWV0tfuMEz#U<EU#a6UXOe0@+w+p(_CaW@zS%H
z<;vDv$Lv$Mb*pEm_@}rhI2DrV6t&JaPvk~7HQO{S>DtMs43_B9bryH++`kLjHH))M
ze>2PIkVMfdXwkbw(YoD&DX8RBD1=$C;~vKE=A2|=2$EMgnRn-9)FSGb>mwYa>1aEZ
z<Is++#p{7mXudt@D9dQ5$*LV2;2PtjRuYOUO(-ghPavuubRgQKEj`m7(&XD%nU4X!
z>z@4x1hG8&^R}Y>L9Yx!1Q@ducI3;tJP!A}A8iB??%nn?dMS4?|Btt~3X7`=x<-ct
zOK=G;!QI^<xVyW%yAud*!5snwcXu6Nf&~aVxVsJRIg|H0=Q$Vu&42aX%`-jSyZ4@2
zySuAutx`<1xud_ttewp2sKIo&wnVbr;I>$zo=jt0u^6I~&pXB~UlECDtoANkuISQl
z3hs8$LNc5W(zSeh%Aj8pM&j&fW%#3XTeM*p%DWZ2rxbywz&ebchPxkwsz_Npmuz2_
z@FR0lZP}gGoNneJqOwu_8u8nNF4!*jjyztTv;wn96^SGf&Ic0E{qBANy}(nH-Nj_=
zL*n;{UDjKHpp~=Sg6F&yF3Xgb6=nKK5eQ8)A*t{bbVJ``k|baO0wR5?CC<p9C)CN{
zu$C;f7?<gw__kIQ;QAZKISX5rbUMw`&^5rObt2}hUXq9q&Aj_Z0t0a_2xaH_$)ay%
zD{9=wv5y%8;Y>6dTD$!k574<dlM(q*k}NVn-|S6@L4J-%<_8-Sw3@k8llrGYVU;2Z
z^_@=&7L}G5c)zU3XvTLckg;jh)k<7SOtXmStZ9(nSqaif@4Zz$)IAipwiZNh!1sQT
zUqRZi>cu%D9qjv>;k9g_t=P!WdD^$REYCCr2k|$))RkzzU{L6)o%GV7QfigvXJCpr
zP9mSYDgp81_kF#*+w9zIzbm#=T(N31zGEs}9$hr7ouO4OFz3ckPqAtOBTsITzqd&U
zSSn2Yj77an5hHZ?jnd;U;G<7a#cgcyn5B&C%s@oUHD45`rI;}@TZ1|Tn!sNzCFeeC
zsc<?elgleucp&X@^Cugxj|92Gq476z1E<8P_k=ZRBBvu8+uc;0F{YQinFtnUZ5P^h
z>~<TlLj=WnEeAYV@PO244DS^RS&F5a^<FxU6|6~5GVQv18!%b6r=AC!W)Pidti<rO
zKCtk)fPW%VF71uo8fLd&5H6ZR*&!xgK~D)UjrujMleOJm`5^M!ESo&;St9Lk>edF=
z);d;(*T`S+$XcQ-@nJ#uYzi^VLv?hr^FA7Z%JUS%y4z;8%0$1+>a)voC>PZjh5}U>
z%bHU|ezlAa7wP$K&v(#w*O1G@nUL}E&=<RT?6zM|h1v`fIcJ`pFD(fZuZi5taxZSa
zZumPG<^ydWgq*+(#Mey=@p-*>Qj5)9wmx7-Rhf2Yf8G`Yoaf}lK>dA#kY6eMr!kTd
zvnH>(qfL?s7y3@%#rd+$@Wtk)>ieJWmS>z*BaYcCaIND$<MnIF)bbe1r!y~I7F8Nu
zj7fGc5OG0i?hh-^Ghxf%N?FDFzF9(96VtJbpSIDKD*cL9Q4M7`6W?l3Q*?7?Z!0LT
zl&l=N^V7bQdz_N<(d!*`m7M>Ex;j~lMxqV{(E_}bmK~Lc>-yioAc(lVb4_)NL*h1z
zscHLNirodF)wD6sr&wDj!ojJ4u@pZe_Zr-<^>_*^stuC1+^*cbJE4fOu>OqNh8s<N
zpPU9<))w?T4YwM!8(Nl}9x+e7><v~>8Wpr`ijv6jC|?fH#~+JOUYA5EBMr1eav8VG
zF<x=t=kL)!z{*`UEUM|9X)@H3e=4!W;xZLbjY&HxXH%A2oc%5-@in{TQpQ4<MK%R5
zRTkr`%I_}PgF&ndMgPccmhFq-i|~!u6<lCJ#QpqVK+_7^<=w{Ry@fEz3wuXRI#{33
z&!MMr7ks$cRpZY`F0z&*O=EA~?MT{vuyWc8zW>@9kXa`A%#|zqvK9lL%;|e0ob0{m
zVoF%?+hDKD8GGz=R4bA61)p8jX!Tpmw+@NeG)Q`XKo5wbQb>bMVhH#mY~jDgqd05l
zQGd92h>)-SS=BBJXb_~5?PJedERT^lk@Y#%Uhk1RD3Udk<9l{2?|0#x@5x!b3KM<k
znSadWv^Y;weUhh*+|RX3=gx#iAIkp%It8pdyb|6zDWQpns3I8L*&y<ho(m+{D(0Cr
zZ(Lv&kJGeni7pf7kEFo0KaN;?I?lQ(K&M9e0{#y;(Jqs5lE>-(y#sbjXrZpl(2N@e
zb+6_9dc}kps}z2^bOy=V=a|3*-cm7)gmGco&oPeq<d%vthzc$erYIW~6HF?0z*)a$
z@(Pa*^d_;H^TaWE;qX%HqekFIH#NtPKXa)OX?iK8GkjTiKG5bz+3006<<n&|QTYsv
zZi97oueOh}j1TZhaf8Smq&G3QyScUP>z)&BS~nmxtzLnJ%Uzu}E#TH7uz8|oHBZY9
z#r;c1(On8L#CO1=#i+GPs8O?WopDC@OuZC*rd6Tdvn-U;;cOgx*TECMV&1Y;TNBwu
z(O`IedbO`ZrHK1&MG&gsXwLdJiQXi`K?U9e(bKsFnIssnf|w0duG*vt@EIgJMg0s}
z{{rNrEih6&v8xAKM5Si`;QL5MYh=Be7RjmaO%4B{<1tA~k;TVL#qHkB^7z?|huP@n
z=G4i4rB?gp23lbbb$Wv^@WJT?=S6n;B(Ugp#s6vlVW&xZ|Ia#ydCLaw<XQt^pr_%A
z$n)>Zd@H@P+0#~i$Dfk?v-}S3)Kdk51v$bFs&DB7I;V?LD=#V?&~y=9<|?WxOQr>R
z<S2<Qu58Z}_hp(DVTnuvgK}zxwD;o7c%l)ApGngZMej=b!ZH?S{wM18zpEgPi>Y}W
z5L;>M&Pn`5d5m85;L;+GhoT_Jim3Wxpj2(->dT8P@6{|ba>KW{+&P{?=?G!9_dm#)
zk$(3i!zv-dqE8s33NwQ+hE>eIMHN!xMZlV|-!Qhe2$w}wDkRX9%^gO(GbJc^5Z`p;
zv(0+!@I?z(q)8|nBpS$_JKY&zwJ{f9zm&`hTymE7{Y8d~RKL~1WB>g<j2>X19}lOt
zF5WJJ0>dXdDpz*sTIf%hWH;QakYR6JSb@4-doDCQT^e9_%xI48oEiV(a<P-iSr#%L
zRYOW$8vjEwaSS!+jqGGf_?&z(!w_CErENO*D@`iN@7}x7A264G^@~bqH$&f4TLnS{
zeKx_<0u88wzIHB8e=uRiP~YlE6Pq+x^sG)LU5A1533%6T?|1TEERS!VophRT+(3c`
zGl#x5H<B(sy!Q`~xjA7+P#=phVk18{$(K28V{E;4^TSgp&npno&#kW`{_NyD)?mMQ
zZW@4Rdob05VP~*0Txq?LIAgK+*^KdbC(PoG6zx<3T4Z0KW@aU{e2#%<n?43CBexmt
zzm$R)QTRb_q(I?T+O7Vn(|kT3$bxn+T}h8Xzx&$5bLNk$aa@SU`mhEx9Uvaj<yU>r
zo%rJ74!BsRFS3+)YXeq?)DP%F%r*98ypfml`8;OdK;Uv~J(JJ#t7G^_%wEl0rcZSE
zKscCP8@9LuvWj_3B?ZZ^4DaNyw&=)Rndkrp>FHmQtYGn+L>gMJi!tuk{q=uj{7Ntk
z=qloK>Ja8_b-JB*6dNER6#eo`_i`<H*oxoUVM=<PT*zd;uabqu^1Ca30`WBjM#dx&
zWy2L-7i}%HUljgjiZ{WiG^g9F4BJm6;*+1n>+Hxtr~*&q4XWjXQT_3&Ahe*n$82{x
zl$1gdS24>RsPSsC<cxMm1ib7pp6|ZS@30%;FJQ6!A!)OY66WJ_HZR;5$MXuDI=R=W
z?Rzu0cRpjAd-P|#fLuL42Vw_W;yFJ@6udBK2bcM<wL?4dbliwsmo2D9T+X0~Jm^5r
zM@qZ@Mzq*C`E~pkFu9ILv!EVH+HQL>IPCr8Jd(|pP&N+cd9Vh}zh$a_{C^w&0tV;3
zpIqZSg^fmgR)`6S50tCRP)S{yHo{}hHXa=RYv}b4^u3SxaGSaXJjd}5zR|1O_zQ>>
zbne1C4(LhouZs;HPS+?^DA1K6912d=-R!w|Jk!1yzqy}1F(+2eXe<&2zt#E=%a~-K
zzAqvVzU(k9)$b+RIz4&sDr_{?wJ8RIF+i2L)7MlTz#)>llgNzdn{n$4f-DPQ^uF=I
zJ2=sdgSBHLi#g0E+?}eQ##pySrp2P1-t^N6EdK($S=w3V^Y?l1IxBjHU(iSd1z9Up
zc&UNu!MS0ddy63*#(}{ry<{7c@j3Z@66<`S^z+DkJKKJ)yha1UL|^X2vz(D$oU~nc
zz0=<&7e?cUZtpK94G*mR+D8ivjDc*S{k-(ROlt_yxFGO)y=FYfI7h56?|Q<!q2u<d
zEJ!Y@-p?JTYLS?|j!SjGiU=sd79XZO<<_OYa&sWO0>8QM6ueKSZThGBVOCk$GS0;?
zQn@|sb03^t6BAGLlw<Q~u{P9YExgo5TA@O0j8l%6f&Gjnib@|Ay2^hpiGT2#URhUu
z^jKqx@|18Ao_!0UQ7UTQiZQ<<@YW-MXd~lnEy4c<U{3r6AZZ#nKRKRsNgceImoeG#
zwo&;H^mL6>u`HNJnv(7@PB4@CGiYjU_^+;tCurTRpI;x_PhqRxW|&U9W9w$FGG%vY
znygio_VbKA?#@)Yo(!1`x$fNh-W+Yj{5nTs-RAD3ob?JO`c!Ct@Z{e4BAL^z{O$nM
zxrtsNjDMew$6=keioFF=mB==b;L@&gTQr{h{^zbu;N0BqFJPa)H*8~68nlv5)!|0M
zz|c+$4DU+~Bb(Lq>k_aG-d*rp5m9OKjz0F^nF>u*FnB;_D>eS&p(_r@>pXRxU~al_
zyccqmakSvSHs!+QL{k^!R2Zwhh@8u2b&%j`7gBaY2lQU$i?}#!F@DS0n(|U`qNykg
z(A*$Sc-08B<aK_IWFrYC)Z94gzvKK@rM{w9_)J-)3To9Lk>7%E7l#dc!c7f(Yok>>
z4@FnPg^F+BzPlJ`Pq*htDDbp`OEaOA&ePJzY41dhr?L9Ix*=aEd|IrZZ2t!(z7^1K
zckWSyrmh!irs}T$1JBsS<wfT~QJ|%c>vKf@3!`>$>C(Spa$X3TZ0EcWNOhHngtx5w
zkUsqd9Q|JJ(fRKh_p`<#lTZ8q5Oo8~3FXb1wTry?F9y34Mqk}&2e<wOtfoeMaM=`@
zhLS@>rrdx&A*jZvS5wS|>+o;o-92gE{afFmWfn8!DiaXvp>UCnf)9pbDww#qa7xMd
zJ>Fjc+tr<@q6;YHqc$;V>_f|2V*n3ox$cR+0K<Zz|CF?gA5%Nt>hlGNpP@(Z@WI7<
zf=BPiW)3R;o2&KnrxT_HLW{s5#~|^!k!`>70|n9-7K3BNV<0CorfIv#4|SNe>>ev3
zao5-8K6Aq8<MlW!AIM%8uppsN=4rGZY|{D8^|Vi-UjH5?Rm^p~Imo7=J-_#S9XWym
zRP2|sB#0gH(PioEOqrS~y$Y`X`nhD`B?1aDcpj{DKi|8DO7d;x5Rjm48OjF+&*IBr
z8FY?WE80*?sEFC1;i&FWsO0gv9OGP55^+4L5yLqhJjHY<L+#KKLI~WYCj=>cMW8_-
zDnWy3`-Y3Nd}rb`+4Y80>97N9cRJJk!LQi)Et?BXxNeGP1{iB2?B{zAI!h@Fe1hl_
zl}Z3#{sPLo{{kZ7C1x9x-XG$Wi=T{sy(KBJ{Ai|FPO1@9U+GD)RAKjmP|z92!PI~L
z#K92GPK|c3Cq>d(q4pY6Z+=M<bj-=0cpsShl9l|Hr(wLJR{Hjgw<C6NULvr1r^5Lw
zHMqc&u(5J(`Y&K%fp^;XxdtF2hU6EB9gvy)mah@KqV@lE$SH0>Yy3aHEsPBN3&;Qq
zD@6Ax9gCb*sfz^0x&FAmSB>!4dK8M}eB`3`e!NwcuC3Il^0cbHGj2%WBndj~Z^a8%
zj)$e1htEzv=P%E>t<<EH{tMV{I34f)Z+gw5e$Z^m;#r=LVTvEx^5&qM*LZ@YGmS*}
zf0O(%`0a|YPjsVlkaI4_IMR+bcqCk+wR>(`_<gK0>Wwm1ArRcLsO1@bY@;GF11II}
zrTL{W_<)XNqpKkXXQXorl_`N(CYauKy_Aaly(O0%iUXOtc0hYUqOm53o1w(44zQCz
z+@nQSG8|UaSQgO=sIWJxpF8ySLui&+e$MxZq;FHwd}LrGJ!~yU(3I2NU|r<+iX!T>
zHyqP(&r%?k=qn_ZoG$W;KhNV_Fi{~lfIfa}r5tkfu<Qd}I4^}cFAO+!jEddR_K)9c
zfm1i1h@i<{e*u-!(P@qAd<oBpgb74FSv-{!hP|u&*(}+dMk+?jkFwRK`A*^DoGo=v
zWhDOc@XoIFVgGO{y*2&ZhP5@UdCbJfuQ4`pc#zYA3}o`|fje0;YPWns?>YYrfiC41
z7<Pg;5M(!srKAv>YL?!3JEVg&v0Os?@4@d9484h3w*M`FhMSy_!b>-yJkc2)kwE$d
zVoi8wmyq}iT7gDlF5<hUp+|q(zW^zrzyP=H>a$NF@AK4Zl!F~rtO4vqZzHNRQ({xh
z;x!O)^!I%M{&O|rsaC41>lIwdRBb{klV6^!uZbdfj_VDi6+%8b3j<ysj!p*ba(`l?
z9`{r^pFhsyi{h^34*_#(N_7FY6^h5LM<u>)*Un5hCCf^BIGtmaA`^CUO;tDW%n_Zo
z0avSV#?)20pbhc|(ts`bkNqO)orPUbmW=lsE=1mJ@XP~`aowcS0XfM9$&C#&`Wo+E
z==H`EIb2Z<&AsR!Vjk?=uxH^Rq9t${ZwB<4Izf*ymoP1V0p}&xFcbbYIXJS|Pyz^F
z3(}U>-WM)i;u1+y=P6HBfA8NFRjdooYdqfI3mVc1HqeUPBl0+td~<i%d3GtP|FPRe
zy1Dt~ju(Z`%}B@jr6YXkSF4X^N(AM00&AXMp#SxL#d{^Abo#%56OLAc#&U>Lv-U?T
z<)>iP`(Q=-`0t+YehBK<R6!AZ2%M~s#NE4x{A+jK-IFBgEK>^<t%O$>TVV09V*GH4
z-Wnj_(g?>%^D0>0*IFX?djFotM(}mkgWw5D^QCVaO=3OvRqYI|)POph7iJ(m|1fxh
zy|sX5@yJF6?F20hT%-7U$2Jl+wPx=f&MuMxRvwLJIa-q@eORdI9k%~(IPcEjF=FM>
zES95r$=4Qo=ZpWae0-iTBs})zzxRt0{=8hq@h2N&kox~jzzd=FxNS5Vx(Wy%GxR~q
z)I<W!Km=$a9<^ly-RtNLJ@T^9d}lSH{WG=lFz3YzBh*VAC~Mhc|K;}?*Yk8;7Ke;H
zxdD;BX0$5N-4Bgn`)$%KDRt&V8zf~+{y;GqA3jIrv=3MMQcejn(&&-=J(c@vN)zM5
znT%|PS&BGQ!1T;W(n%T<373H8YS#!<=?JvW=ijSD(Z~+hCQ*Hv$CGGn^s^kr9<DFN
z6`;yJ;q{p(LQ(F(x7+|6ts=q|P*s65iGPP-p6zJYrDV0|B>+GwZUI!YF=_#&G!!v9
z4=u{aHy%$3-_5!YKi%)gihX*%fPlPJxFn0B1HYc3xNY=7*l;=I{{jg6-*%8*BJ~h%
zZ|@|tDk(5LRC0>j&|n{wDs|ZL<SF#{v=tZ2j)TLeET}cv%edfV=pU~!#eFD~|G05b
z+VM~+rYM8xa^D?Q;PaGz@gWn$5Gi!ri0+GSy@>0gfJ{rf>fG&^o$3pC8$#(0cUjtk
zC#aKa#vP1z`hhV5gAUIURf8uS4&K%5XUQUVDK`03gnV&$f!tby25PNV5HNSkL}#6B
zD|jI>haa)#9|cv!F2N$JNP%?#J8K(1-OUnowH=CbcE?hs;t^>2C>G&I^X*`Z@neM|
ztg1Sk9_<IH8kmN=^<4Ro(ucA+sie`;lDHibB8hA+O(8(1*pBQYxc$$C46KUhZyx7u
z<!wbUMBh1xU7UE-VTClR&1zxw{{Cz+u*h+d`g{e6<&^t`YtcNzDTRBK`Jx_34{`^x
z&)eV#2<1z3gwD1r=oQr86TKF6a$Ce|RJHCEc<VU?(|e###RRT!?2jb`0O@J1@Mjg&
z9Vz9*_CZ7vP6AAII{du0nyPglk!4Ek-l$nmG#6$6NVKT%1H*OQ^5y2uzD<Oc$krR;
zPL*NTTTRG(G#y~1;ibkDxBVc49`NlXX{(D!MBgkJ8(O?x1oeC>ylrIYh3+A*VMWN#
z2)B1sN0|21L0?h8{H(3z#|n$`>22whpP}k#laba^am$JW_m*>m9*siQH6^}lmi<VN
z*T=JN=ZgE+?3-o6h^y9LDOTOq2Dq8^Kwr70fvZy}D#eKNr^&O(RWSDy)2^%r5K%}j
z;T*P6qr2J3liG8{X46od%WO{#G039qjfa=hAHQy$avESoVeKa$mqgQ7s3PALe^I41
ztk`clfajo}ic~ZTuvpIt(7$IP9Z?>ReR_FNZq`@!a)6&T3WlpA{?YA)cAq~M%eCPx
z83Z!TvT9W5I2Q<<@sRXB?e?}@@=^j>HRxIzY;>KPeArfC7*4sIvaU{7ENsxB*}(Zg
zs7ehYzkkI2n#J$Z_@N%ePN-O`S^9@Yf9Me!RHFuQTWA&JvE6&e>{k|sbPH>A21YLX
z(`y7OP`OkW%hFvwm6k{)N0-F2=Sh!>F{LC^@v<@qhH=Wd7_+d@SJCi&=Sb3k?d~3!
zpsN`5;b475v*^Px#%xu$4Wc~bJh5DBY3gM)^q!oBP#>)zQJyU9E{_Y{+a$~j2lfK3
zTGk+=ZA*?E)Jh#^t%Q>4GamRUR7>V5<Q8s4ul^vjbAdsi>yl|2|9PZU4UWDd$fnCm
z+`7eZ2K*UwGgM(|?>B~Q4lFCsHjVqkcRg>@7?&7c(<mG4OqfEu3IiKC0wtC&I07Z=
zjcM(c8gTUmY2PxXLqE5xY(rqBWsCzdx%VRiXX|4H9{m#rBOk?md9=61w?9n@Y-&7H
z#^-Ik2r70TI3VVDEf+^Pc(>d?zs+%nAdO)83^IYdRGJ83LZ$Cr@wVR}W-|>uT&2X-
z5f)ktE4&Vu8-z`JVTet-!R8syr1vzt1!+2MLk9_k;W)Qdyh1<cvlIU4k<?{ri%Y3`
z8h+OKNuH7`D@#1gQ=~j59PhF3$0xDdeVVKN^T!JpJZx&tbwA4^;vapnrUwB+2$VWU
z=)^s>4L44|7xTK(Rp3}FaE{~{5%3I)!_*rQz_V2k-dnMlND_t-7FnXm5)R-F9>4ya
zKm5`)W$84&aNNc0uc1|F(do!zSUF~?82VwvaoU7ijGKd=jflo2VES_iNQ;$eMxAw0
z>A1%ZSmLW8r3%#F3K=WOp0$yba%=pix?kDIH^iQc{Rbnqv=oI*oGRjE&q+=h2l?mi
z&Pb3Q=^n}>Ass~B_D;a$jpl{F^K<7m_%9$<%P_6sOu=iM61@1WL!UTEvKd#lvxZgg
z=Xj&^r(Oz_>fcSv-%dpCT|Te{`*g<wtxs|DE?hRWQfzQ0^Rj#tx`sNw11X1n!IBBH
z&l2zkstWI9-cH4ddX}IknssbL;fhyGAIr+Q!6G$|VM%9N*?A&^8rSk_*v^~kQa`79
z=8};Qsl^TCR5X-D<gq)jSeWb|j_|=n*L=3Rp!ia9QP<v;M&r1Pm}laaE{@g=sBLhs
z99Z-s8Hm7~oBN;c)q<^48VE&Lr7d$!BOa2337=+Jg2sJ}xcw#b6-!6CJ-(V{AP@HF
zK_1MT7W3AeIuG<%qP6rvwrBqOwmAL{9je2AD`d9c!=5=+awr>TDMu3!w?i#hLDy!V
znSO^J?l6ln{Xz24Sw6zo>P}S0s4)GRq0^*BJ6*{}!Jf|M@l`J~M*5uwgxPj~;IoH3
zCM6y`Tv93Jx-W=4DsuM>cRO^pSsU)&Me-8sa=;zw{J5LBye_Zcfy;BvU)(cxaGZP0
znK1h<^fHV8qNlom=Pnkm7tcJ)s%1P)7z0}c5D<$qAm5<R*gDX%;%obwK0`gtxG-7P
zyv*qSrtr%uAXYPX4E=YL!rE#cWfHR6go?_iwqsIBCS)B(dcQ&x9X9PqwaQIz{*oHb
z9n|?=xBOZDA3u3L5A8T^KWbS-z5bcC)IbQ;Rpw60Qxx=OMjOmS{K~=0AyU#L5+@SH
zLp9o{Ej8_+Jn$C)W%Y;D%XGG4*$Gph!p4^@_at|oN=SPdE?Pq9FS`^gLGBZz+(#*a
zzj`D?O(PhDFUG!8xwO4-b<}88jm(($aHnsj1t=xV`0FqeZVRAn_$*WZ33$A-`8oCI
z?#TM4FQKIaeVOy4&-C8tubmcS+hv=cXOJHblgP=C+qQ_%xFj*=F#DM?;?pU!khs|i
z4z2FmM(<E?h~rlP)l=N@vx+I)ZI5O-@GG*w?T~^uzm=Yj<+^Bp!(hjSl?|NOXFIRt
zl~29*Jxt^V7|S#P9d4B48PpP()pz<-$>Y?L2|%;Pu@uV&o$4#bbY1qJvKh>z?ww9K
zy*pji8H)2(p`#}CW0(#LzWzgENSQ@3v_)*v>=vU*((F(cxFoy7(oPmTZE;k}@KX7}
z8vKYHYaXEi<}9@{6&h6<B3HJ<x-Ww&-JfhN#%3GqUz&x2GBh6^4-@6&zU}g27$>oQ
zuoWvv$qJHFd<=Gc5QFq>)AX<0tpz#2UC{qxiS2cqY1Sq_GHco(<Y>NEIHVr<wARe?
zsTDt+aBK>Y$N4M90)3fr#Nh&bFcuN;U=`%`$DwAt#Y<L8etQxdXqAUjI&9xGnK0>8
zuSdq3Bn+}ol>lSpEqRAfBi@^Cgi^r5lhmsxL_Z-Wl}BY~s2#5}G-x=>|K6&wwWVgL
z0`Hkh&6Xfn<tx<RvL|OV(^6x(Ic#HP=E>(O{!)bf@WA#ZNKVR5|3qK)0rU_X=rH*g
z(AY)SP=FAGw$$1>@M*qf;V&Sp0M+R=s?RI`G!N&t_M*qRg1gGG>yE+mb1T>Z%YogI
zUBSy|^EDrud9z_yAHdn+(E3cg<ztki*xclFG&timx~zZhd16to{Ef3?!JkuYPwuhN
zHI3ba!z`3Kj`+bOhQy~*hrINYI96&K?&s#}=HrIhcke7TWRG=KsgplEr6|ebur0dH
z{i*y8`}JF83le*APF^+#wd)mVVr$|XKDd2=Qiokc5KNlbrQc}$;jU}5@0=-O8M*?m
zjI9NWTQ9$6rUXJIQ=0F?K`hp#XIh|j>RGEiw=?9bT7_tCc@LH4twu?PAc;)mXO-rl
z$-PEg&z)v%ueqQ_e;pS6PKheAgtN9`%OoEi%H^U~p!d2O4@Jbhc}#%oFuMx-!~m8w
z#%e~#uS_SPBLCnRnN&!rYpOe}Tuf?jWvMbfa*3k#S4Gv|+oa{D<I<^2jFE^Hqlb0;
zhO{q;4J9RKsEl*KjdZl?0JVpzW~ls9-E)&IcP&z_=8nf{>u-nfR6`uy1p&U!R-*xw
z6)#C{gA>hK-kK!cP54=U{O@Ai=EJaA;#YOTo8V{o-#TYF4Bb;PilvvW>*^x5S;)|L
zx=q_}j)N!#_Pwtx0dF$ivJnJ$MZ+G^veS@=Yi}*OZ^K%{fpmo}=nperx(8qB>ZmIi
zh3KhMG2XHN=;i0x^u10tgHw<$jz0EeEGMJR4rjfo#;3EDwME1X4(x>P{`;^Uq;`7h
z91|b-T1C*gt<D<_a|9sv4nf2^7X{JY3RSJd<3+Q~FR!grrfo*rg5)|~bpq>$-Hw0E
zQ>j714o9t7DX(8<E^9Y)o(=P-e9ySR?kfiq1&-jA#-}XH_A?A+7Ta`(0(!p2pL(*f
z2Qp217Nb#;UC8J_Wb}a*aY6j)qm~8C3`xonS=>w2&mFXdvKqOog?hZ@g#jU{r6VQQ
zRBZBP)Cxa$vDnqlvA%Ug2iW=^80F-|3$pYe2XR<D96;VFO$9*7sg|ANf#Q(;*`~XV
zDDB~EYR}r2R(&sbn*&G91h8qVm~9gMxbUpMa@MEji6%YEiKZlK1sqqzD1Xk$l#cDz
z)8ST)igfixR(&di1$#!7@kY5crv`EF$qa7Q3?g;1&?#7Qt<zt~Y~8AOYw5=#P=ir3
z(<<1m9%N5i&Qr9l>`bFEk<S_f$_r=eWsPR%O0d)Wv{J5)7B{J}evxy+fW|J-!|%!E
zMtnB|H-2>m(K`w)Ue?(Lp@B=*tS#=U>}n0`yZ03vl@jhF?%EwvPQCCPS_kYRpf;Rh
zlXbldvdAI&b#E5o0rg#4pT33$P3w3ibJ)D@c(2`6wTebqkWV-)bbMFLwvuC?P*uPX
z<sru+1Msq@J@y-w((!$(30ADSWcwqc*nrxuI!8Azg<|tz(o1B3%CG#IZGR7|Maklb
ziBeKt5&fg@sKwJ)Yd)S!RmSfVO7eK$E4E~iW{r=jf3J6$TjgD@-*g0{2Zyg!JXkl~
z8%@e|3iAcX)V~OJ1qr?IgMC$G@_bLfRlz}Qy16??n==%mGtyREp#le%y*_E=j_Ht2
z^JYYg7t9swE^6_>Uo8$!8;191jWbROGmYMqEBiseHnbk%RDN|A<yeI(uCNc=#NcC{
zxej0LiS-b1P%cZL;a1%=6-!cA9PSLJqcgl557X8bUSVGqj#VDL+o!j*RkSbiQ9|9O
zoj#Gp_XT^SVC^3dV+2HJ$|}nR8BoJ*o=%WI9+H9tj326cof3!9Mp)ttP<9P6hR56P
z@=C#3ZNe)KFSs?sEyiedY^GgXop+(z%?e4}3{-n&9kt_%MZc`U2XA6l&6y?AwoMaH
zhAWp(<F-po*<kx~0nO}r_W{#3pl#<_#FTy4MHVTyYqQRoi#QRvLv(A_)n--)5!Xi8
zH6lf1>l_df`#62U+WscHw6vN&&OuoHg(s&1_qf86ayf}usVrooL8dtCfk4h@`9~gJ
zobK(Hbk0q2-ki+LOiZv3RL`)Nt)-8UyTHB``-Qi_!=Yrfku^YWSvf|Fq0Z{ux6L4@
z_GQ^$<TSmj_9X4u<5?)t>J&#6F8`>r<QJO`G_RH6uA*M_-JO4zWi_tjDNDN}PkbeW
zhd5$0$EHosv{}-oQHSC&`3xtjR{>+PO)(R)%mh!JgRLrgHNf1aDFB5*Dc1=`W}M=P
znurRUIBMdT3KfNkJ*$*y;-U{&oN64~qYmU?+vJ8AB}Y!k^kgRf<IU!mL;z*3>mkdV
z)a;Aqa$xm?=Vjkk_!09-q}3ayWGm#f+Do4Dh$k1=c0Zr|))(xrSKOkQW{<94?Iszv
z3$89``n8e4aWIUYDVPI4wZwdK{MJ{!Kkc|8#a!s~A%!eP^mWQI%e*mYicvm-&Y9jq
z%In6RXU80M6S)g^gUh)6Kc4LWZp0oVs?#-Ld8i)vMo3R=EIR&uk^$4Hc$F@+G%1Ij
z(syU?%9Mxq<}@`rT463QdB+#@1C|2c)y?Abu#cH|#0_3nR6SLHapZizljO6fDH=M$
z`<^{y={cspu*&xOx^~M_u0Vf4sH$FrrxEou<`q`SPbC<FTa}#e*K`zE;uXKhA*O=g
zCm+>mw+K9}D84`|ljykBTQEzo5e<}h@(~@yZly<sV4xF+nGMY$HXg`{#GLy71#rJK
zx4T13Z@D2VjPExli@o-+*?Ojk0&bZ`%ZJ}_uTsqKfCA$Lee?AomqE1UEbd!f`L9nd
zM7KzmJ^-}Yw(E+gcjaht?I^DUOZZCwdt`x=x@?B`YtR^(bLX<<$?6C(JFy!`?>J_Y
z2gP|O3>wfedQU>?7Ap|K-NtsX)8C@b?*Hm`HELf7cnS-wyYb#1x*IGgn+e1drC4!z
zSiU!_GDh>qu*-c$Kf7(<(p^P)d_`8r>emLbXm3e+(L?`fI+Pc`Hs|9To%dSg#=~LW
zaUd=72y!;;h{BOKcC*O$#on6V4cj+VHk>W%cKmVKu;BZ5{{pn_?#4oKR=HP&MF??4
z#SY4Xi@tmHlT!dcFSeelWQV}H(I8)%73~1YdJDe?+YP&3PCr#PXqZtT`jO8zbHRNg
zaUnS)y6P*gTUd3SA3@nDo1gJ{juKQ~<YclhM;z)*ZIuWK8R|7?)T4LF_u}MLr<8P`
z;odJTA(x5~cI6_z$iIxO?M5{an|7)dx9!U3gpgLE^X`Nb&b(Yr&qhk)KYS;~p~?YL
zX8T9Mi|^xKz$|`Vt_SS8Ynd08wO45~zPNW)#NO2E<Udp(+n9-q>7)ylud3Ouo45A`
zPZH|w_`73e1H|^l7m?9QrDD$GWg)`H8%ACf5%)Sc&NH<6tK6?Fv#hi6XF1wC`9)!*
zjA+{G-f@QNClf=?Ye_+Vy0m$B>9L}o6a;LSMAvr#gvDxpq?%$QN)$n77QnZUrLQH9
zwa<0TmpNgH8Q}N$ym6$5%jymE*?n*FqcQ(^XoI=~Bu^mATC2&URl&QH9%p^%)`7D<
zSRVGcWMZWN9;X5$ZVIUOVu7ubah}Fno0Ujb;QiVJT;BcGY3b$;cz5)6XPYs&*wzGE
zDZQe8n=i=n#96V8y_+<T|Gu)T+l;6e*w9b=1ZlwUab$R4%mMrbjQUykf)UWsaRVRx
z={mV$`%quhPez8U*P7cl8}0Lre-+NXkLV7M{;x^6TO-hCDS`BGTX*AuK8u8#XiT~-
z-+T=B{%p`GK?q)B_I6mqH8$$RqB{fCEKP2e8Nj!i`{j2CqVt5*{n(&hhlx*6)?w`-
z1Vlr8Qsb%<iI<tQxO0QxSa+aGeYb`2=8kxY@^~_w<^`E{3#h6P=97xZHqEoGGabdR
z3iN4D_!T5P*joHJFQ*5#B8MHQiCDHH1ksDZT-`3zMY0H@yfnC9KG}|B(>+04NgXm8
zpZf|W%EPaF`q-{}wdQ!Nz=1kGAwfl%kQp8XlEB$fGk@p|NDto(*!!+ZlZgTF8@w4l
zQ7+?5;=E6!Y!{$E#SERimlb97)o42inC2x?5$ERil&W518x95j`cZbcgh(BL^>^;P
zT^weogoRt-s$r@xD+YAC<A~eOm|?H%?^cs(O^C(TjvCeic|sDG!m~T(w&!3T1UCKV
zTJTfXh&)Fa(W<JA4cDm<$SM-80|qD~>cfZFJ55A6m**Jq5p~28q$vNT+g(j1cNLd~
z9X|+vn%g{nkL#2Cyw^8Jht2$j2d3}9m}xmA!mB<08(`<L9)!#_Shp$liGKFvA~6ys
zq?kb<_EZ+W<um+-!*_8AM%#p9$V%m-<(()w+pnnvO_#?uMkyU3qUVdlMu9j+nY|a*
z%^v4gnPOP?`i=dvwXI*w`1%Zn$J@}fr5qvEc7mi0kr^v*yro1WK%He|4P_F>`8(Dr
z&(NpvNXFmbIJTGCS1AJYcu?>2q@_1LUVN*3H10QMTwm%`qO+Qi!qr<;uX-K5j(%UB
zm$fG;hp9f&(rg4>1bp7_NML8OZfA%hQ$UUVQYm=#y_YWl02k2_fU00KeEF`u4;jR1
zTH$a4t<c&4;wSiQPJ@X{$V}N5-b+u_9{Db$9o2;VeFhjV1OV7CaB8NB7aXBoRgUy|
zWgW1F&Lp87v1T_?`B0H9xN(;JN6Q)vlIv6;*a&atdX{5dZ)9?N_{qo5>{D-du9Q7w
zS=}pIVj8~S%vfdlMO50K6F77*aeos5^D14GSVw~q>~&T67XYVD0ga2+!8m!hoF>kl
z{EqGv<a@slGI+%f^u>#LZp(PVQGg1jzPMTX3*Bq1kxocQ5zikM7zva2o>M`)r-&Su
zrfn@AZWw>^4z6Ub9Efv}d|es8D@oDa3;t89b#`WNMC`w_i{<@PbaJO$M~Y$W^=EMl
zqRZ9J61;2O=FR}DTT$kx>typOXejg9{5SIT%d@=YYb^DH(Z;VyZ_nGtZvKO_VIHyH
z6dTMU2I#KwQRBhPfBN~2w;P>}jI+!G9&H0R$L{>0vBkyTELeA%Q2Y>fxH1ykTKv|=
zsE;3ubkSFo5FiU8ez?MWrt5^WS0*hg-@(P$DLOY?y?v-mYc~RBe&AgqZvCylPPpo5
zo~zTwd^ur;TKj2@L>O%f$e7<=Z62o}a{_sK-)ZrIGczKT!g<2}B>g#26Rd4#|6}*j
zC*?p4J$>e=aBrKvxP4f8F)<pgtHO1WB_>)Q-WTo-u{{bPRCny`>RW+8z{6HWbOpJO
z^r9E?u@xUO?>DK%8yyPK2@;e=bm~-O7I~waRiZGRlVAqB(drN@A`tL$o@c5emY9|5
z1R^A^DKBP0M^}SPaA0Sfzyn>8LU7>P04oA}LHKOM0TBf!BmqYQD!lnl{#)sqABDS5
zWy1Q#3$47+_GEb}0+S|0IL=ZA@uRE*LI+H9X3TbytjiPvS`?Rjw)e}b)CbIotjuK3
zaMe%^zO2wuww#j-tkV(daIEv`ID0`4B!2<Oq9iIw>2)br6E$BMgTKX##xSqP?+kcl
zd}uLi(B`uH@zqv6;pgsXO8w++5e?QM=3`lDFIH-@FF5ibX}?y4`3667V<{39>)GV+
zQSm#j)EEcq{RBBVz$yYyhRelUl)$3e45O<P#UQ2I44W)#z5T7WYv;-<0LG;1mVR{)
z$_D!j*j$;<U~&eV_dfa3jDjC*b?A*>!a4o|sv>AbnGyp)K2`-Zk$(X}AG4sTYl?lx
zM=|9DUcvX96$Mus8(^h@==DlrrNp-vE{-?3+~D~gxkLWfB<0cNu!~{!;N2t8s}XxA
zYM&*Z4L{B3atLQN7tfp0zuy@_#Jh4taFkP804F+j&M$eq&})wT(>kZUj3WekuSt)B
zFYR^gOoE};d8@-WL@;jr5w;}!?XmJDzx>?>Fth~iFMvidviIh!l9!JZJunqTSOx87
zvm($)D*G>Be){1CdkyW$ZMb04%&&knA}zE8dLI;uK7Roa&(k$Wd?V!7y^YXnG*ZkZ
zo}&IMz>7#3iu}&Ymn^SUsa!;8;OG{qfyfW0Q~_-LJ9WU^_g(P)c13ouz8Yl85I#Hp
z1unA$PM_!+Xr$YhJxEK1u#r#_`FbRe5Rp+D*15h4UVjhngqEF~i~F5x+wY?J@U3u7
zm)`h+i2-_}Y<t_=EUn_RuzuEC-2sm?(nDP6!Af!&U9Yxse*wq85WG&o>qg+qOAA*8
z`D-Bq$X)^VLrf8JuP(h#@Oo-n#z)&e#mDuP%&IU@8G*Yq#iR~;3>(`1q&i>lJGK7X
zSz$5oViy0aV6Nw|IW|G$#i$za?Y3n3l61?4|D@#X$p1cBov6jn9n>^BA4FDO_p)?s
z*_N37XJs${*!FB%51Ikt^jNhnbX-Zxg|-n>iuQUUUn}@6Na1$Nbx~m;ADV(7uotfp
zAP#-x+ZBG1Ps<rj0OY>#(wCk%h_Dg(EqY^K2i*>%?Q*-q?6b(iT)VU=)Y#@&Hu)>#
z_-A`7{{k?u^<Ek@`Y!bhk4RJW5q8dK`|Z=M!%SQUq=z&S?R6@lciP<bR^!t&tVQ(D
zaeLlvSB=Pp2+eHd&j`NIGn|5Y^px~(tpo3ObL|&ZIE8kG^|vvhx2g^`UvssAHqE1g
z1V+e3XW;b1<15zxG>Ht@>R9kOl`5#E7P3ri89dWLP$Ua3cPk1uo9h;@EXPhSU`yb9
zK3-h~_%SB@nEOIHnO>?d`UYD9-TZHlGx!N1e|f)UU<>TSikUcu6XlDys3@MNlV|!1
zSj$j3_+go&?L1AY>F)emOCZwQtUFjh3%>!3DNTI6kO#(t=<Q(o{jO06Xmz+ao{@<b
znH5}hg_W^Mg9tK3ps>4B53zgcgv%hnhz)Y?;~AQ>!wwf#frpxINaAZzb|RGATBp*i
z2TN4%pwmN=2-Lb&PMQDrL<lnenReK`GE7YUrzf-@o+f26Xnk<<7`uBsP;(nF>)EWk
z@+qtjMkUC38gqs2=Cd9qV>tOyzbcPz|Nl^B+XgaQ&!pNYp=`nl_GrB1lfFLKd7sdM
zUJV<x2vEPi%PRov>Ci*b`ft=$dQsl$0-6|@g$9lKzkuC}x>rN}zK&QX)O@3xAiE%C
zYy#-h$*p;pg#exSSD154H)qo6AK$o%MS~uALaXMW{&r7`io)QQw30BVCo5r28wffr
zJb?%!TtMKwQ=sgv{wKR^HEJyQu@7=ZUAftek??jz72iq(g5)eO1VVPEOY>4iWhaH;
zDe9n=kKftBio!f584xX3cE}h`h;lx$YzEelm0zUuZB<Bm8-fqOJ0QzS#?3q3T|W1-
zaS@n&2g&gJmYoc1H0{m_$>LyRB58jN6LfSBccR<A;(jzFl$MCF{)dx{8C-Vx_#g5S
zZQ*Et4t?eu*(gmF?ZviVKuaws%~It5#LNB(P5o~lp#N_!dYx0^;`lXt^tV}-z=qZ|
z@bpQa%@m^;z)-eN(tOy000F)&zF^d)%;Xa4U=~tznf-)3pV##>nQW6v?Tf##ULC0`
zy{ehMU9l2kj3fGLlU3l=o(`k<wcTEx&NMe|1bdd4eeQ8?eDdeNsI$jVcMfr%V(iUZ
zbL0qy)x#ehwZ)eOegH?QHL_ziVKvRV<|CNMjmZrL2@VZe{{@Jn!s0m(N8Kf~A4Wmo
z--q<Wtjyz#mwx(^w~+eLu2N01(W!XWGd0&8wi5i=mfkB@BqcpRn<+DkfipRXv3Zr2
z<4WC4)#%&vx)QY70PfettzH1iT%QQ)L(#_NR^O*hFTx^np`a?nr@FM-!NI<YO@Vt!
zP;*SpEdPZv*1TPOt-2?DcI6um?DjmlB_deR1%k1aY;#PM`T~QeYW9it!=!B4)+%Wl
z+xMKbn3E)v-oy9i!zb|pBWX4GTw9W3-+@ao=eJzyeI$V{S4B_gxjh}0hm+B&8OMqY
zKw%VxjOmE8npKjwl?Kgf&ozLcyBU)xd<e2SjAl8YUN=g}euGTHtYT+)DF%A}U(JjW
z^ns-Oj~#V=EznN<$)DMPUncwp%Fg)qWs`&pqtzFxp{AgQj)b)i5MucHJg*|*9sle!
zTn{hSjc@$fLt1~u;I_2`A#SNb9UYHjOM&!b{$WFd$A3mzk6OIIgbLRzi86+6l()#w
z<2QqQu=8mOuq{`OT98Afv=J5RkTr4Lcj(c^N_e#Jc@x<{gBL$VdMfNebH)W7UX;QG
z*A2NAW<v7xAz%}^n!?_K4-Og%Dvrc<m_kMk2}`Hq*(wca3lXR-b~vTja529GelL|!
zXMNxfz%Y75FuL63o^}xEK+W^|A;Ol=xO~j9Xdt_c)AH*7_$Dl>l%HP+*`!mlW97U!
z&nw(vAz?8%S`n#>#FZja=07=KZVVyShS>^M3Gdb`CnIjtOCTyQ-4u!P?CA*`dLQAS
zI1T4=5q2`Sn}KBP;zsgLpeImt=(G1a%9k69c2DbmUj#J#Sd0kDY`!4SJN=yP3>3H4
zPSkXawXQ>?21HV;*`o~7F;34H{qpyHN9*KMXpJ=+0Xkbz@*xpttInCmiy1~?RdQF)
zPdpg6nxQ49?_JhmG65k8T7Axw#EZN5BxE_R@bOV@c-kECrhZ#^ekY?5U-P=#we1x^
zp6A>EB+DNtP)%V6a+rYHX-vRyUca>!fz`z?oEbfmx5^(6`)JNvP6YczerG1A=CW3U
z?HVd4hWOXxnn{RTWh7k4ucVW^Mq|t2_m;JK`tpbf>OD2Tl2=&|c3A<aV}E}1Quyop
ztly#d)#Mu>Fd(kN6|!t3I-bWO_CQ?#&J&*PudOeGlVU7e2CNS4&kXq_pYA>54?}V5
zjdn8OpHTzYk-s^RnULUG=}ym<Gl?s}pbU#EyBaaRf&bLv5@^Vc%i1Cr2HfgaYMf`v
z=19lVYjUaJ*zN{k@kODBGTOjR(IQ(A=f%Tg{qDiRu1k&n@K13c!3ZHo0&APTKVx!S
z9hY;wsHj6kTj+twjG$aSyGZFOCSwj;Q&t(66aD#d8$(U1hAvVB24oyhg+3JFGonVc
ziswocL|F{!?q-cfHjV&Xa;`Bxq9Wwko*WkK$iC=pdHgu`<aUgv$5bn1WXq;ur{Jiv
zp?j}(?aP<3TRL`tv_G{uDTg;UIT*{pMMZ#=!b)uz1S_L7<a+Rh4O<Yj7j-pA!aOy_
z+6=-uksP(BFHoi%^!wOZ$dVMLk?I@G$NMoY^hXkBZak03M@nCO8)<2dE6W**-vePd
zjMX&Ml1EW48$pP|JBd-Sl*gy1!$&m&K<qV|k6V!>z{4%qQEn!e#zIcI&!2~CV1)#Q
zIg%*JscEX_UM8k@cmLR7f7O-M^+=B3TzyQ6JGH@dV`u+ftf-^_ax>B+l4&sTFhPS3
zxph2-u`?bO+y{0$;@=|)DAxFiI9~xw)5Au56QKS$g+w!lxT*e2l0z}{N4}OFT}*%m
zTq7jzgCSRX>VuXXvTO$$3k;|TzD<eh=9|@aVy1~}`L%jq;RShH2&!0ZupPeZjVG#U
z;FmKo%1!a<E10aOS+e{LV7E}f{6&6ozAmb+HAp{dugf6EX-d1F8}IP3NeIeeu{lc@
zj(ft@75|>7Z36FW1rzk%WXk6~Pi~?VU-lubE1zi|4T^Dw^_t#el<_o7??A>6>GeVV
zlN-}`<>SiuAJ8?Xo7eNw(m&YQ8yNHLUFp`4!1Jrgyf`xNtN%PRQGoS1jt?nI&`QeW
zFQ5C30xF){%VSjHwbM(Dui|=|*EfsQAQa54+B5Il^PTNUhO!tZ;7Ip~S$+r{pPy&_
zK_-H+@LkXSlh0?ErY*TX>A55I<oKFLCqjLfRUd!njuMaSXL{gKWSD6L$$K32fO{Xb
zpf3cd#VI7!>q2opwgzcBch&Eq`(wbWR6p)kro(H^^>Kov4Q~v*n|SW(&em52|0y)L
zSGsVo<_T}Re+c)tP0u0!Okuy<FJr3Z{)ISGb%cgQ99dK??PB0+uqIyy4P_l>nWzcS
z)K+g@3t1#1w7?nLsm6<0WR%u=ss<uH5&P^*6{8VNe)G2XkA!LUuS*+ZB)Fd77?Q}I
zo~ih<<Tk`=qP0Lq7#`{67ttNYc-;aegNSnjzrzE<c5-Hwg79WG%Ai@-EY_JP3-4X}
zFeaQ$h__tO2@m~<=+|QGK#Uxf%`?wWm0^!7LE5MUFKZkCj0sql&TRlW@M}K@{1oQt
z9Ej*@{=Lm>P`DrN%=GQC-m}1)8`%X+-#)~$&hZY-ghwp;vNY_+c4JUvxwF@hdGhUT
za~PTe>o{sY;;?E|oG#_z?Ln!bAN|Y8c@wsuDb^)20!u(3y5#0T{wA>y|MSw1&)La@
zItlzf+A-zb<OGPWq}Gg&4K`54)h+}-qGOLw8to7qCe|6KKq1hiFOm3FLFH&8kguAY
zoIe&Fz^apv`_19((evr{U?8S8nc+!>Utn4;jR4;}QGtj8=hx1xMPtk7%Apnl<%;BE
zH~fda$mqQfWQ}7g6Pc%(zqKs7r+<#=>G?q&|D4rxh-`(zJ6fDOL;m+ON?iDxI&s<Y
zdKyBk^UiJt7N(yx;uPjQ^66R6%gd_dvb80X#(%zAmq{bu8kz}4V8~W{@s$5uoJq}J
zzF$8?j!Q0YO|6jv{Unbw6nUNPeRM0-$%Zgk7*%SmZp0bzTW~X|h+6JDDg8VGT=dYG
zPUfV<PgN4-u{4bD-!Vs}*m?PR+4ExG$uyT&cj{UB3<kLg;_LFs`pj}*m<~idZ%U05
z!MSpkv3pExj>plk;fwyn@c3Q)dyznvM!8BYoIJI7JUTn(dxp%srKD&W3I8WPrBv1U
zRa)Kr>di4lDS7NNYh>%GLiq^^hKwvXSTQd<m(t)WwV!Rsa4PWpPZ3NZ*f1;7Owo1h
z9j&(^qLP~nk4Jr~<ILWx@5q__7e^azq%QGhdmov(zbx=8YoI6D<#AUW(uXO2jtO6y
zhrzBU*Q6-yP0Gz@)&TlSNec<9@2N+?MNl2Mbf1jjDE*oo`z-*^*{w~KU|2O8`}9i6
zDC0Me&HR+Kaok(NUXi-rOfGAuNnB3bS)(+UA3(Ra$FH7T{Rbf!9&a)w3q~zofZs$_
zm}>tI*0B5UvELGY3L|oEQWj*B@>1XSfWykApL;nS?Dw=iTjO2fM58o52~qnS(9yBy
zM;HH?qn{|?D5uk<h>;7;X2$Zh%#&qGu-n}xV_{V}8EHHi)1eAaip|4+a2Kb4k4Gg7
z;0u7`F#SsM@})z_wHP5V;7}Wl((lWYNaefCH0*C6DfD!CF3OjNFOHNsL!RWkYP9h*
zq+ZxrrXMI{dik=3URJf&A{i(R8Q3^TD1LW*#JWIy&to=$Zo{B$*YKl;oohclB1j0r
z&`u&n+gA@`iag|r1XoMuC~mmPv%a|43sGX(xt(!M<ET!JX_2oiEkDyN^VHM)Sj!>G
zWM%tjRIo85*&tbf1%2!D&$xUxHkT|pjorDX5!i_LjHdLbLu84}n-X4+cB;!+eY~!P
zrGBZLBdPGsh5h%Z)lt=)=_TdIABA)0>HYp6=H4o*t+4&}4ek<LLU2m)6b<fP+>5&u
zcP$hN?(VL|3Z=LuxH~QGRve0y_B(w0oUzBb+*f=2uitmAu`bq1-Z7s!f3rt#US?D&
zw>m?SgZPx0MnF>`z<(6gsZ=`WsfDlP3w$U_ijygjogVUz`L+xhB$<)aTt5#9X3q*T
zS*%KCWV6`<v+GsXdJT{?SGAtlX=%g=hwD`)G1WLy%GNQy(bw^BMMUu0aCFiLS%`7D
z)y5XGCjbyPl%Vpnp?8|aQxyJ~5R2YB2~&dwP`AW~+B5njB|8NmBqatqqKJ=+J&paQ
z3I{KC(w$d+Ulj*;qLMb9hpQ+(*I*z!y)ZR5lFG<M%aS^c=z6CQn>SI`eH#xG&wy&V
zTV8?9A*F&aT+(C<K;XRJu?PcF1nbOGu1i(2FHuVr+sxaspp!P`Rx;w&h(>&N@i6a@
z5)skZpSCJLuu(kPKip5Kw8bkK<jTuT;1+mcZQAnU)wgB({DXXg3L{A2y|#L95LAE4
zp0$$oYAQy7zzkl=$m;F?*m(Gf(xhxjmq3f*O)ufiruiGUR16ppyP)XS@1OO{qe_?H
zKBIp-;>KOT%OV=Xawhv~dsEb1GmmZi*G7v?xK3<Fce1W!j;^_Zgvu!!xJ*Pn3!`-?
z;F(58vD%wdi|ArJHUqhiNa<uN?u_qBUFk(2hyPY4wNW$^>qLkV3|^PP2gJ8f(ugMF
z4QN?5X~igiF_Sv1jg-i&Ka4*9M%b3DU_dBUV~fek!Yc=S?fijF(r8Up?BtHAk@p4y
z+UT8HQqw%(yazUA!rQ@Du4XHuz-G>D6y!<T?90#3?6rAnHP$Iu6>4T>Z|59qOpB^c
z!qGa@3pg2yP1spX(^ME^4ik@<%qiY3zK#&m85uL;NJi%!=CV>Uv*YZHAS=cF_l;aw
zs*p<WoxPDF($PdH=^P<pz|NZ6IvoJzTZsOKxugB#!e^e!_ojDOBcghfmAc{uda@RB
z^5t>+-s)B3^boH|w@@wOx>N;eRVPtd7c^xetd5(SYYq{PcW;Vr4Xb<Y?dYM5kp07o
zGHDo6$jEFuee}@Ca>ghA(Le$>$3)14>EHLn4%?gJVyf~aoMSb|UkIZ2g%6xJUi(=g
z%CQ}cf{m)rxqck+Mx=n#%3UhVaG(J=knY~;6x_y<%;vUx&?7xQ1dF!+=}D0DNz^{x
z1nCWz!mk60*X*;In5f_0(j<iV)K3tcWa<Z-%_3WJ>k-w4QL@ti2e4A{!Tvj1KSmNB
zcNe!>*Iy6NyXEMwUHFXKw*z0#5<|OR6c9)&k^os*Sq}gJCpuB@)Pl2S8=`P!YF{U8
z_6#Me-AfnEq(7`3OyK*asVxn61~;r$vc8uMBjambEKPV1!?w~{khLXBIoKP>=j0R0
z;7N}`XM*M!s0-2kfP}6MQ#k!&EdiN~xqdW_4OMAkZbg0AtousgIIKqMo-FSe97Xjj
zvd+BF$K=4$t1!U2J|u>6bs_bJC#!WyiB&~IBaOIu;;3XyrEY<#h1lxNyLVTmUFdXO
zRY5|6g!e)gBPgMDZ$JMsUw1)930!sE_IeijSz0%6tI)R)c(Atno8gcG>4kuEWK8O;
zNr*Q71JABMp8LvO08%NZio^e4P32YZ>E1c9pf<N`a`}f>3B7qTkWKRYz-{2Tn)3pD
zRmMg%XjT7|d8Pf!B>4Rwz`c7tx4e$n7tC4n`3oJ31(Vv&Ug`%tc*C#HsXL<Ly}=Q=
zR3C$eUH<{MJ_>i}HgCma<p=ADqF6ZLwEC1+&+=L3F5B~mH0!neG*h43uYJVOs?>^O
zu-=fURz6Tl|M5b^fGF`0>Qx_?DiAEtP$wBD=@1m@yK>c8#Grr%tD`9FOo)6zVu7*L
z<9J;1FQJ;L$V;5vqcwc{8MZAD4<y_6+pMI!jxMY>_{^UsqPrZ6hq*o9I-d^_QBNLs
z&{py~HPU_O^Fh-wX#Qh4;)nUEMvZX5^T+<10HA@KRR6A%kw)y{o?jw!IzU9E%Ox3y
z{FVImF32IyW{s+BKKCVnbMDb8GM51Seb=eHN06-?{+CdT)4Y_}{wqU-l=N=Hm)hNb
zNR$f|H38go?%2k^DWM!=rC)uFL`yxW8=w&A+xnHwI!qOzkxYtf8tH6+3O`NVDAHE$
zdaDAZU!KAqdZD>`IXGsn)>+qSO4xWYh+j{YPlrGDU-Yt2`d5;Y0-eeh)O$Qoay6^m
z4HloRL1heRS!ar`j4e*+9gAGc*MoTl(3THI9FZg{$X=+xJs|6)SQkYW0V<@E0S+hf
zQ|u`Wd6!p}W6)&xSNbJ1v~v%l5$KzjU{SR!luH*j>3bU?f46MYE9{4=u&peTRHtUM
zC1XXblR_*`jBIHXHB>sQ67>jKrcx`BT})+a;UK!DAwMz_UtC^3aUu36gAUP1|6WQ^
zM{?&$D9Ab6!mi|@)FoT<S@D7nq7<ipY9j_eXUsQHUO+5b?!T8cd+QjfUhNc9TDBT9
z?yIZ?w82Z)#2&?;j%S&~inZP^R+*lmNgu#}aSsc%yqGi)$Psfue4fHXZ{L3BF;yd$
z%dOV572hpOvRfR7Zg5cH#Shf2eb1`BjE$#zwV<UNSIe(w{q|{2(^;bm5?d=QMbWFU
z7=h%N!BQE5>!7`VS|VU;ZNcQI$@l)sA?mw|&gY?Z5#E$Ma;nERn*^BzDGJ)(O1S=W
zc_7it<D`1sudq87FZW!HZXVWaWw#y#o<^lp`i--_L0n#C>xph`zM`y_p3=#!hn3c?
zG{S4Z!8(6VcHQfx93eBx%R8>+09A^K>gvm0%6X!CTuHuiWbZNvi<U-5{p?i%?P{Ln
zo(~d|K3fyS2<L$N<ov0JcQJfx2!7bpO;fhob+;-A0^JM*W4f%0Azgxd537X^3i%yF
zg?x?bgLr1DDRKjM)Wt-XOJ=_QvIKSfC784G{S{sx?IyM$ezhFOOG>J)@?T9(3wpHo
zN<!BzN&+Lrx?T2;N3q4rrA+-3&*XAH*!p20y;txB8Kua(OIGZcAzCEOC_{l+Kdp^I
zOTd8hhH=m2hp6O^pr`8xEx{(V7BHyZ)#r-_Z|Xr*x5?it%|Jlt;R9`zX(ijV6bVed
z4YWkjmT!>D2(<Y03)ZqYYq2GPZ5>O46OE!{R)nkF^DTy|ov|QYE71ivpOMXDAkgO4
z3rOvO`;)xV3RwvJZ?_ir>0o<ixH}8g807;Xa&WJogF5tGJNoydeh&)Uks>R$)1^ta
zV)R{<E^h=z@YioQNB|nI@h6iy43d5!F85K~ufduh95#*I!PKzk*>*SJf<XP+*$63?
zxmv+*|5uv1=|lSm*zSs6tt<B4?DZ_|e0%%*D!*sb2X8;z@k@L$g0FFr_2+?IYPg0F
zp-C`-9w~)yB1Kzxf6vAJz0tCzt0!W<KAMt1bD);w93u?;A3*bp-rJtZqUN1fTJ(WH
zIDMbZ@Y<7mQ<Uj2eY959NoApsM4CW&mR}OS{>%!(1+fL6S*jNL=DA$c3JSOfi*`Gd
z5eI2iQ5|oHdQ94vqhhG|Cj1*-nS{Eoe6Z%Eu-sT!<#fgG&P_ZrYeON8H{EfW!wu!4
z3{5m7z`+k(MHi<v^039*W!J3=;?Zfh32FhZh@4Vi2JTafwD02yx?L}lEs$HNQ3yE9
z^xob|Zc?5Ufx<;FQMx1<#~-MV+>pwJT+z^j-7Iz%PVaT8Ur7)vy=ym3(V-4bhD{dx
zGYNdVKMtZO*qX&3MO;I^6RwM1W}mI&eWsq66z1feaR@wf`qVpP{Ms_zC}vpMD5Y>c
zlPpnVYPivUffgYtP#DPLvwhLA*?;lpK8frM^Yqun{hRTM-Z1G8z0)1m`thq+WtgHm
z+Q+Ns2qO!q!RazxI!-nMt?nySPFgQnyS4P{FH1pH^gCvhN*k@OLvv{hw)jOd+?hg2
zm~&5l%|&_f0haz&k)=Exqh0%xeCQQP9J{rXgCkx-C{#Uau1S!=xNnO@3Md3iX*TFo
zsyr9x>WMM=EZU?R_eR1b4&@)wuvFK_Vs3!wZYO-}N~EaLHG;a8Ou+Z$v_)=(z)4DN
zgH%uBT~0g|a!WS;l3k?fR&g47Cd+*X;wFT)|LDHZH~pejGm-vlB^NfvU0ZRAb#0lU
zjez_B<>rCPmhm5mg)~%8q9pRn+7i`j)d`08Le@Lh9YJ3vq2c6jc3zpaA_p7b7Fl%;
zNz4i_$J}o`Vv*V3cVnq!*NTPQ<C=q_-3xeU-4J(qF@;I8&ADIFPw~O<^7LkXSSo*x
zMOa6?DC1Po6a~}4VYHn&WzA9Cq=l^h^e8!aLOs7QQ(Z+>MTUn1!WzSGj`I7nS19rw
zbwA7^Qbt_uZSL7qC<90pg01z1w*s_)EJO&O&lS|X!tJd-x!a(tIEnEdDUcVW%lE~0
zVxk*WP?w5!V*%7kR4}2*rg?<eL&_Q7Q!A-Ti2rJ(BzP-EW<>om%KcOgPPCWlrD_ZC
zHjcVQR(fM{By<}L4a+uj#o>b9|Me<HU`R4R@6&{@LT!I<5bN|ZTy<3+0?rX~v!lps
z`sJT|vFeb>@bsCtK(LnN#Gx2Od^>^fLhPO1Z=8$qwQ#ppLDzMGo|aK6bZ&K8Vi+Ad
zS|^)|rQ0sl4SW0Xn6V+u$*%XO(TnHaN}sdQx0%v0p{p4}s&?_<0FM59_=l5ZYk_uK
z&__XF&+s#TDX`<=K(%1OPk`TD*=d(&xQFz2BfQ(W7}Mq_<t1QyiMHEc1M0#DCwuF*
zm-XS_MI~dw+i`A-0XcYopDoL+)lyu52bg<6ZPw5KKLAgXEs$`^-NT$f2P|;&5?q@1
zB057OvR7s=D*Dh7(<^k8b9)V6Bb#+P4=iH$TSsjDeYZdcI{j<-`O3$;rym7o)<|#J
zweu{oRP%f#^z(tn_sspM(86g=@{ER}S`0$z3rpdnu==^8a+bZ>PmZ*_OO8ZsMn97#
z3HgMpvX``R>Ht@vh_(p|MOA>D@3}Rdw4B?cBR3AF4>bFbz^SMftAZb<3@|w0k{eE;
z1Uz1XTW*`X<Uq3diTn?l^}{)Clx+>bJ{)Udhrq6O!DCSw(ZMOqkeiu$lRQpR_wC@d
zr4BEHO=?Wn_E3ze8T#=(1k!9zqZ{agoNV8Man`-x+-YvGPW86`*v6f3{0AVR^9vMh
zaL@f)zhEm(SW>V*Lp5I1kAvTWpkK(s_Q**69|iG!oWBcGIns~L>Sxl$Q<M999Bs4q
z;2rN1Qe|n#?|q8l`8vtzxqCi{G`O<pT&U+9irC9?-Nf3e`c(THtCsy^!q@Y&gQdPX
zAC61?pRIc{+yd`m?GTflk$&Q>4Y>$@BugRq#IKU*{(~R#jJRi0lAS2kerJ}`Clqq$
z3*2I!I^_I>7uRdYmdJ7h#=YkX*Q0Ehn2tlmQUki@7{l?p=9V1u+bE=Kx8Mo8<KMi!
zfBH$xVpQ2I^GZ@9EZ<Z#L*}`1OQV*uXDaF_98o`RC=C*Q?JF3p^u$J%zBzuUf{e|s
zd7zBFu9LyeAg!h#l9t2?C3BnE9#x})3W{{ok3!K);Dnx*t73kZ<ywbwofbe~2PPdE
zK^~0a($cTF#O2{GoyH=}A|fOj?KgfJYx*W_t@?Ww)C|^DMJfxjq~zRzJkeU_!#CU|
zc0{9a>5d;AyB6X6gg-<W#RPt}Qe29VnVk4F+LFCYQl*z+f0HzyzAqN8Is{x=BB6i7
zUG#VsNc^E0=$=%3V?aJa{fv?=he-4f1)xzGx^4_VOwc<4{6d8v=6GR$?SoK~>H@i^
z5>#)v@5KMU&+9_khk*--6GcHj+A6+g;FWbC0JkyQqdw`mNK(Wr@_QM`%BmkS0YX=~
z|9!S|1PKLjW`D?No%!JY+6ciim|`&$b2*e1`Py#|qoBLf*TyU+Bd@c4JG29wGz$b~
zFkc6%M(w%30%P(S7=4=Cu6Ka8hh9g+akBp`kJjG(V_b+mIqf23rqy1G#?@d{A59=|
zlZdlUkVUj%X%02WWv5A59;a<rdlTbEUMY?;1gXm`V!>~8f%MI=+)B-{uk)!l(Qj%E
z#!srBISG9y#re>4ANO_tvHUZi9SR+f_^srr%%w{=VEKUi7pI3Mql`#%Kp|1p$zrp2
zrCYDY&LR`_1afllPDfU8ZvL?%nO;SS#Q_Q|AN$x$5|7xS7fxR?<jBn-nLvE~I$+TM
zdORDLAr9r*iDv+@qMPyu^p;_R$kV-XfIT?2V)$MciBlFS_*KKmN9cToiG=kGJNO|<
z*+jdlKhr7~<d#5Z<hm?H3&J59vPm^kTn&ag7Ih2Ec1pXqN~0-@<%ilH3!O}uNmn6u
z01ZouRQn)zA0Kvf%=N6e$2x6NBp>rjWr%GVHuooYn&{`Sxs}=I43pYZrB{0M7z>&G
z60quQKj+2UP1jg~Yjz)#X>f?d%Zi!LO7k<;`u>4xqy#S=I=ctzkC;{jrQ#tlC<^d`
zaM+UycjOyEFuJ;gQV{LqhSu3m+p8&vu1G&auQpw0;Yp%=JKi$<!h8$mNf=o6<266h
zh>1U2{!0!LO_dd|ZNiGjRh4H-!7kw|<x9fm`=Rq?=jq3@d8NYC0cOx3F@<F%M&Qu#
zdQ>X__w*b2tvMoI5_g(*Xb5^_I^_jR&GvG$^2oXCrDMi5DZZ!QccBhYogd;+pR7PH
zqI`A2)TpsA0y8HlWnx`b?21#Er&zX?W6B3UqR#6|u<px!Z5?KYd9_cXga>HOY&<$v
z662d+$_!n@&R$_1kqyT~Rd2G-hNjF;3cu`E5;>$jSxriKdC-$YYFHm8XWOe~Q)Ct_
zn?aIzMR*Jvz6__#nE0@Yaf{ID*|w#p6$&WSzKRUl@r+*BUC<S(lYWgJgG`&WCyQMK
z!YPi*gG!&1JVlJ0bfQ$aKr>)HbJzx5L5FRDF<w+r{3c=x@a+web|W)mo%t|bdYl%K
z4t`liHxH_v9OBO@eI9$ni%<#^bBUUC0Jl;+KF=&UbM;JJzKE|;9_|Ok{z2{5uFxWM
zPrEJ?XAzv<iEG+3Xr^#3IoJyXH&QTF7{pc?WPVv?Dr;?HhHB_iDL2vcLDfO8mLOt>
z7OR+uozua0IVOtuL^lhk#8!<nb4(@vM)9+mlGAD|85AyU;lk33gXu!sIuuHli_!T}
z{2sI{{Z5|`LlREmvtzNb2SnA|gK1i)>~WpJs%pe_$~O4PaCm`}VrhCMLV_x0=~x;|
zTE|)D5%sP>T*sq~DM^!QPP=&md;6_2eRV}Km1Qwtr{MmhgY*V%Y{<uf!wal!FN&7<
zW5b|^@E`N@cMmG`W8J?}3^nakZ3W|}R3}+1l(I(06iKzlI^b9*KdO_rRg{${I>;<-
zZ7tc8iwB}!3})K*O6g+D-6%ckgVcLV4QBv=J=mh3hbfq1Oh<Q2`S!FRhMD9R-Rn;|
zNvCaIMU-gV=TCF#Q8BbQY*3<IKmbYi=N{$jj`EzzPm%I!x`W?@Ka0%0oSz*1mWb{j
z1XtFNis56Z;YgD4q&!6jDf(4=m#xTqB4{oCRR?px5QnHuJ_#J!Pec5O?~3R+^QkIK
zi%NzwP?=>Z9rE$lh{rGttY&Y<C>G_SN%}w=PWC2usI7maquIVrGyE(QPjU$WJ8i+T
zj}gLj8-T~NtFE?aP_-m_9k&ccnky}$&B*zP9+gG$xpfz$@Kt&9ja%}~=|S8^+O9pg
zzE!B%6$Mz|YpHW`VkVYOs10R{1q~b9O|#hPRxFVz@mBgm?*pD;*FpTvP6St3c9kcT
zWOx%C9yaPV%fZKg@m9s+fO9@{V$qnIG=460ceonM?ZSuLXeE&_6&F0}H^lO`V3Q)y
zq9q8*&rfB~UZ^k^Xvr0g_BscX#p7GDBA0ZX8A%#OSf$`4H!g5OTQpeUy+r^7GRL%`
z55?oVZ{JbKnue?~gdWeXa1A}&NEZK!oyE2bBYkYOgiP)ki9z+bOvepc|Cfn~3JAmh
zvqVm|<a!uT`S11>NF$!%80kW80u21(aqn>iZ+rL{D-ABh?5-SgN7@v2xBoD=g(vG#
zFgWz4Jf`+fl78wNd24K{31%Gv(wJD{h|$w`N-@L{<1}8F%m<S77o=(vdF+HXZxN!c
zPllFFr>|$gAj-8uz|B<%hgLgmD4(D80$~MX18Vh3D3x^HVz<hmE5+083M6eKz2%QL
zqL)Q9FatlA%YM;(wTn79u(2j8mpgpE>!IWclwjFkPoE?rNfBun9gY<3QjrQiiWj&{
zTGa;d5Xfa)d@&ctHw?GXXC7-La#F?o)gauY&Te+V<n;QZ#$A?uUn)n8?&pi3)agoP
zN%OREMes+;v#mP_9?+W33d8t-qhyL1bU7KM6Erh5(M$&f2O9~rbRD*q>esfrg?!Fl
z14{sWx~%8`1Rpfh?vS46229vYr5!iPH3w2oPaRT~(A7*mjAcr<FM&`}biFAkP+gQ&
z^)Ck8R*x?2FeE_8?2RMt5Yky?V#p{n@#J)<nt3U>kJ^RovJi$=&Wd+8w*x->+jx1d
zTfGnM?qJG8n)8aVqU3>YVUJtU9soBo2CK&RgS4={wVpCgj&sfbi8i$VZIO$c7SDLO
zf*8CWQ6TS4Y1b?Wk@|jw^iUIjno|`<`VRoa{w3+|=GB-?{;$E#Dt2HlB@@k_2i?b`
z^yoF9>$I~g>hTf&FA*0fmVgQ@WHF?)byA7PKzkLra;l)nmZSe3pV!H~7D?Geap_SJ
zUdv3YSsBMLs8T~RwwFe^g!GCBf$ty2E1Tp*-Q8iJ+$588`<#<VDiQX`N=14A^%WY!
z(Qqi}7BRypLzToOl-*2x`_*Xcq;W6{c(LS$-WpYs+;@RDSoqP?s_>&L%M;kB7%ExD
zaZYX$*qKTD6K$2h=42!wiHV@iFxYb8+zh@576hM>Xj~5dSeX%{Y5J?D@Q7)POml(i
zl&L$xn0uznJ@}iG)deg}<o+cZK7U%a_<}MQTcUE!7y$#NsOYuiYJVV_u38-7JFYPi
zRW!BGew*&$tFBVBf43fSFNK0-`R?QiE8A^%C`t$`2aB`Yx9JUeN!j*NC1q`Y|Dwpy
z&Kqo@I=5IDFvp~$sv<k`gyMvHpjRNn&$zyw2Tq1sv&lqHAJZ&rl%ZgW{C#O%g^}sR
zD7lxJQKc?pp_(luLWN|aK)EIZgXp%F*FVQ!>$72v!9v=8{vn00pyMYVzJ7oVSPgXX
zPm9-^p1KMcIxs0Dmxp}s66t+}3Ur?d4T6C0G|9}VzAlpv@jx3KUW9fFRsZEy@r2Zu
z@pF~G)vdvfd>w|Sj*><Xu?64n?0eI2t7?;$-dKMVH{sII78G`QQF=cFBIp+U9!JJF
zN;C-Tzp|CVO7i<mu~c-6JS`bni$epLi#9Q+G69h;+z5A4)KN8}N3KTJY=Au}Vi_AA
z@f+S+yAEauC#*y|#wU2blT53<DNg3E#bmX{QqE8rBOzw?h<!8gqqK|?F3G?tm^{KE
z_m1ejB41F6`L_MO1|g209s?B_#N13KY!;Xp6%`q<6gnRO>Zq$&Vo11S80bg#3gc;@
z%sn$|x{d7S*E7KOOG-;qE<O1RtpF~2^Yu4tNP3jV78`)6N{HtxD$KQ6uyew5a<qN`
zBg2V|+ek_08#RptOT+yvrRdDDqIHlGWhEZtI<8Dyk7wQugE0dbg(XfFdodlIb-u{+
zZv?)tJ$U<0Lup>ugrdru04VfXUwzr?Sh-Z`$$z&LxjH7NMoX!^oM^;DdA@}pz9Q8u
zYiJ2q6NOmdPR-bMkHHf@&^m|A8nXr9v3ilr)ylwy(8Htb*zX{(g2*oN{oP)l*uyDa
z;O7A(UuS*B26I}vS5quiWF2nn`jQb4zH&e1m=C%iH7FSvW~8m2S&=@B6T*x7r56$G
zZrK9Z9Hfml$nLtPdE~h0Mp=Qc4@tm4{55B4!UvyQNKrw+*s@m`urG$0n~lqqL79HH
z_uR|<m<*{r!{}kzt4(*8%Y7SK_~);4&#&+3#@=U3aZnlKDKB^?4n=6_T9@gx)j(V^
z9qbI2gjCxe8Vv}Ignugb#21S<Kt+~9OEk5Fsgr6p8S{f$wh#twpyqzP9NY1DqgNLs
zDJ^be-|DPJ%HCE8OLgsGPl9GhfV;-;X(qNSH_ZvWzBXn~La<No_X*K(jWLAP_-iSx
z!Gf}=%Sia{<6i2vMIh(18ONi~I&YIx##02U2-IR=D+63qfl_BCZuplZG|46C-3-Sa
zHNM{X!%X3_5}p1DypO{m&zYbxq~mCKxy^SQr8qK<g>mGhC@;aIq^{!reqHcvV@7~O
z<iXU1dSps|){i{+A9otb`{FD5%PZ0BSfB8ua;V=Gt{geOcvkSs`H!DW+94sy-{T_!
z8vCS~?^ZG`24DOy&K4jH&KO{$sKEA1%eOtSBzjPU>xxToSSuZ1NXq5<@bXSGq@E8v
z+Y>Ib<rbZ|(;_MwSUMYreZ`1vp}N_aXq@3W{uFHoFZ4g?(sv$-(~HRU7QWlBiB1#(
zCav}J@`AlL26qaj;5D0cl!;|<gK$xq%-zSNkB)~W$+8~c;D`1E82zlTF-nLz-H`F^
zL0-pH0+JiWqh;&mp&%~8x#B7M_a!0WG`z5MeV4RCX)HlK()b{Qfb?7B#nux3XK3`Q
zAimMhfr^_hlmqRYy}d$)Vy93qy>Q|&c@O3W{Hvqk6+aXRaF>qJP7**Zxk#ZB-_ccQ
z6Mkfh4HgQ}Y##V`HqK@)V<s$aVlc0jOVxwO1XiSDoc<(nmmLyy)vHm>9sTEbH<(&;
z@?4;*>{!Rj&R--@m~2FvOQ{+vcMBJ4wyfT%()g1|Pn@;k8$dLYJ>9XAQ#z?z%?ZRc
zS^b6uCt^<?0jd9l{}s{aYEBZ2dE($z--${YwP;_?R3xq+OJ;p8)2PL+tVrQVAW=}f
zqb!ZQ*-ibt_>dhx-ZZSbYw6;lWh-suWBRcl`^CRl?iKz3gka`p(z$#m^KhGq0|kWk
z_gY~DbX0Sv+4C9w0`Bx@C!Iy<>edowXPQh~p?|HOJ<aF@L^<mMfg(VDo;(3Ha30xY
zo(x?A6nEabk}rN8s!(Z~mzx0y;WKC!R%Kv7NgZr;HsD7|B(1YX?jNrV>qdzhn!?d3
zs@*7HRFXR#gq6wxB6Sx%BP!_}h<*+6#dAh|EX}4<rBYMNm*-B7<7L_RyV}7Oa_J&-
zF{Mq92OEh}fVJ3j(+EHF0tebNU-}p0@x&`K3kuHUMO*OpCl$~-=N>CR)TJYsJ{d-F
zr`{;QY~-|+<zkhyQ<JM~aI1=y6huX%&p+frqIBM4iwv`b8q(QX@29s~q;VH!W6<GS
zML?oQu*?lmjP!AXn)koHRSPJD5}30aak9`PPgpPWJ<I4F)o6e9a#jX0$C4^qFcKM8
zXt9pV`Y6^Q8@7iv3!_H0@KoTU(`;mA$L^CCt|5Kfq~)TJ=ZBtBW5a@eOtB78mG{9U
z@TqQxd>fuzPbWmsWtQAG$&NC9;2QI8OL0Mo;~WqrB-e|7U$L#i3<Qn>f8U>*ucNrP
z$uyKwH>Nq?+6QQtebK5o*~K}N8j8bZGjT-PX)A^P2QWAN2j&gqnlOFalb~2P8<Obl
zAs>!jQ1bn|n53lO#Zfl?70#*(_wV8Sx((8{4!dqH5ck@pCCKgae5{)_Pik9c>&G2L
z_9p@IWlwaKy(nhj!yki~Rxm2>-D3+-AUxU^*5|uxNhj9tB4#n^+{jHCN9wtyOUHoH
zSNmk=64K)v4S;BTNV&S~K~89(v==yVpNT%pv=sSj2TSrY+JxWHP@H8bZjW~y1f9Fj
z$~3~*6Q|k}fAdt;{K2Ch`AoDaiy{6ddah<*$%O%_i++V}%VLyf^mWQY)lFuSuMbH=
zA8Z$pQ=9$MAV_WdAlib@*v<<iWFt`JuN2Q`ThGM^?lAqC(285%v%QT>iEh{`VcfiM
zxxLC2`sfx=&(js6*JKma+S0i%F!RMndd~uc>5rR0(_u-|vUoaP+vQH3s89H<u{&fU
z#Hw8knNHy3a!czXI6WAg<9HWvCuCtvUJ?vKdU<zBbAfWrf-Y0gCm46r^E@ID_I2l=
zDA7){uC6M0o4lC|y;QX4;Q>cCO$%K&L{w~9o((|3NGi`NDHKVDkF<_!$TO5K*O5Vu
z-z;ev{L}CLW4HZj>Gs7!(XM6Jxirq<R_!p<dZ^i`IC3?Dg8;o4%!Mo2>UvBQGH@m4
z%zFjz^D<=KMQNrXkLssN^ldsD$qI=tZqTV(b^Pm3Ur!$21Jntro#*FPUZ2vga!d#H
z7@Rl^1%If7ELktexvh{48R8M{x_jgXIFE=EsTRs|6(TPf2G<zQX2cK?Ukc$zmoeVN
zF5*HOL0?JhY;wsMfYP&}euPMjNyOn%2-1ljP9e#f_K88AuRfna>A+pRuw}>hvH1e3
ztcT{~6}EMxgksT-ipr${qUGyN<KU{v^zw=lfx8@IQj%+qAB!(F9L@2PO*|Z?UqJZ=
zpZT7ne7AQ(CDPj@57%*MBz;V<5NbTnD$OY{s*-Rjt2s_%R&horVTyk-dTcGGD<p^N
zn0LVt4V6kwoTX><hGg2QTz&gU>o>d$OlXRTY<0;55CL=Cvf7(XRo43_RJHv!ss1Y_
z;O+&W!ZFlb36~xH@><Ra-F5uu>5Sa%vqDpXR`nZ&KG_4R629KL-BdfI)%kv-0ef}D
zxd!US40~K5`V}6%@tb}4u*QB|X~Fb_N<@D7(=Ub}uZ5ytPvFH7k2Cn@F_UWct1H6%
zs0z4mGn^6t<Y-}Zv0cdSqC;x<Jru?5FJ}5wrFl{mUM_pjktA=$scwIkeMoNghN=@y
zlM+%D3<N8I65DD>c|s_&xjq-Nc{Ov{|2c|^ys9+GNC|KW(JL7EC;mC9>|hsz8I!+y
z`g7wxM);92r9bf;zp1-aSP&iIj$`GW-dzMoCYEy52OV`b4f|SNJQ7R2zB=ro6zYRR
zoX+(aNRS07jS5ugW$a_?8y1tm(V;kwySK1=vXkh7t6_%4p(#AW+p8<1Xq^;7`Ypj@
z-kV9`iz8)D$QO$G@(l`QK5^PwS;q1iB|TiH0`;nqij$h9dfG~6Efv2EbeCe`JdL-+
zU$0W}^A*Qx&FD0X$V6ynJy3t$aWhl?8nh7E%K7m#{^HkO6^EWAfz~vRmDWxaK}~gq
zliZd|icW54nK=eED=M<rD{$P%!lj*s2uq7nXc!}wA2<ZnP$W)G8<oT;o6gCe84hD-
ze1o*IJ2{DnL?><>3Ic)Pgqv<Jr|y*?2z<Df9E|$c@Dv62`ni$nZQ+~ic)Pj8$o_94
z$3er6l`Ul3HT01SInTFmW$dG9t@5+xQpdGwBzUzO*bv%jI_VhOIiOKOImOo&to;(@
zlws6Ycu7^caEQsaF`*@iC5ojx3a6U-i~*rBA=`Fuwt#rYa#B<IGvC>ig2tWv-ABE4
zZ=DsKc3`pX(kmL@dyb-}8gnss3pg#6`+i<nUztsbiO>HoWr<LTKk;L-M&@uT6sGp3
zY@+=Hh$!FhLPDm9gT3(zF%EC`kQfTL^<zOB#$e1826HCZd$`YtF`EEqncNO;&coC4
zb#+B;Ei6m5zew$@CN}JL_O5*0lVihkDHPBmB4?@-Q`da6LrKM4q57sm968jW9Lmqg
zLYG<?LK$I*)q(n9kqfEJoYf132xPegt^S3{fgtBFK`=M%QxLkFezs~b^o&W2{UtkI
zcEWIWz|gj%x(lTt6;qeZsIG{;_mDRAF(Gox%k5UOQ9}B%GH|HHB^&&R{tnnZ&S#e&
zM){Jkap>zp>2IaFXv=sjU=;fqgPiVKj9hRVi>bWQ05lLOYEU&v6h}fv%JY(}S21Od
z+Fa|eQBZ#JE(7Upeml+%?r7_#mdC5b9A`#E5SwdZ25<9Ys^BOEve(SzjNYX7N)N;o
z?aumj2=<6kaeo8T35ionF-r6aj#rgl@{lv(s36(2z_d&@uEyD*qa^m8`XD<qEf<2t
z=kUEuhAlwuNvT8V>%s#55FzZRP)T%0G}1`S*qTw++?C#&xLaupOBJ`5p=|7Jd7~TC
zj{4@bcvi2rMEzW`xgmqhw`plhKi>NnlVW+Qo`kTU%)2Fepw?=7d{z{%)zM~4;j+PP
z`e$iFiC;lf6K|edX2y*18L}4nh)%OB*fd?-<+m_gT4LuJkk^kf+w4AZlkZ;1rLTT`
zF4DMqc%oTs!hpR_zi3Q22V&RgtWj5F(1*>TQluch4%cN)v&bemsS}=}MR;v+l;3z(
zDU78Uu!8auOL@~F>SNQAP@$-t6vSXes+tSm=;BVlgKp)<OL=(GRWl$(IerUt4<IC0
zVmiJM2KLZyaCix%&nj1|7&R}<??fYLpu9yYIb>OHs#)kNYX*qw2+<cH`-}=<$_&~!
zktL$cH6O`8>-0~{_O}TQXS)AhW%Pqq!~FATR+TKm7j4pxM+XMnJMmOvL18E_1|U{>
zrzPu4sbw6c>i9z1Lvo5~NY$&eeyn@kX5C}Ax|tdjWnHkFrlhDd+gT@7*`$mo`UNq$
z`~EI{T>eeITym@eMa^R>_qM8EK@`7Ty&EZ?MRjF+n9!`{ZZ=SJ7vJj!Z-2lWbp$I(
z!4&VxSDV#FS7Y%a>0dSnKjSvrfmT!+(>){hVrEr#E|XKO;?NkDgAA2<u1xg-F*l`@
zmsMJNxy_;a&V=f4EMDnW-mP9b#V$UbfbKQEay)1oK}Hb9MD|h2u99+r^xRBY4MnXv
zAWCLgRg{gU-0$~b%DPcbMC}wi<rulK#zg17)wrAnZ2<Mv`x+aa)GU(IM9D$WWKIym
z{}tG%Eqi;K-dQRwwHfh9gC}!40>+4LDj264bgwWYl)P>zOLnfAwNPTKM?L-yH=%9N
zxtCI|?WCLURs52y?9Y?V$52p&$S@O5`w;?isF5LRAu5BSXfb|085?ZbeVrnu){NC)
zo!DQ~OJu)hqGhu1ahEjqWK>&98hPBSO!ON7S7OOgA-x>NqjcK%`GxIw2jK!-j58v)
zeV@P|o{&)GB5`8*9jVtn5$?QAmu88oI<EO^1AhYt6ana=L-p$uX(PuxdH~=<GZxAZ
zW6-YHqyZ!o760x$qm5?ggf@_5bJr}{Rki!^EdVeVo2{zrdq~|)Y%SzrxqLFtcxF81
zUC*!MKMMd7gqZSf4#{+U!eu*JX`9@$tc!E}|5uii?MGb+|0xT`=u+1g7kiIc=C8fB
zHb9=8Zas*`a}qSXi}V3Zk`l$?wh|d|En8VCe8;LRor{HCs@aLi%*czyAIK4wqXZX|
zEHD#YGfcr*OyPVKTob5%iL(;y2cf%*9De?b`_05H*D=@hV{Sq64K%~`OP^{9Js9s4
z%8HS;@-J1n&>P5le#f(skORE?4w{YkiqMONgn4f1=_~I)^&8})qYZQ6XwL<9wi#A9
zbeh;`>5mMZ$Wozk1+^H|B<x0X7zJ}=zs(hq5P>Api7U%iSIWBtIMSgE{-Pl0#_y&B
zK+XSg!)P@#+#6JMT{sUn*{A+RM|7r$%`HhLkVf!2SmoB*7(sCGcqO=rwA@0&iTHP`
z^dtOr*3%_@n9}!4#Y=%Q`KdujPYyHOXh$aGXe*MvR7H)h*{3Os1xF3X^l%NN!|4yE
z_&t;%{pTGAt~Aa%3*_c6b%_RlTaCN8UtCfC#B<-!Sk6x({#RP_|0Lb}zh%|v|BKh$
zW>vU~;_12JNu7SY@_wnO56==mr_Nt~at|+Fc`BJnYxu44JBAT^<h3gydkI(W!<m8+
zIQ~0=hg~{C;jda5dN<qrolrsq2FLfvFXRr~e?fT3C~xQj+sckB%KnOO-S_=gK)OHd
z8h)|(Nt-X0_4|Y7y&LG6hjS}+w;9#0y{G=3R_*&H_;C6-Aor<#VS=3XYX{F&NVLp{
zLy-~VU*P30NI}Q6rjaN5wq<jqh1ujS!Ph`)r?r_c+#lQRY|%MZbsl%BYGx$kC?+L}
z#s8FydJs_mF>=v|6)n^RrKsnNBK=nPX!^OC=WhO)&?G6u>ymFs^YI#tgR2nruqldC
zSQ6|Pl}Y9jn|AtZT0LyIv$R^)PZnSbYV1o4HVSof<{1G6y)=~;erk-JJo~-;eCB>#
zXUs~}epBqKeE`dy{`DStTfEC}FOOI+MDAx*H!GhJj23@J--_5sUtAn|)hJ*D1AphC
zRfs!A?EN@zp0o8n-xh~9>t=(I&bEhWn$NP<nrLh=5d1UCLumjZjVIqVj!-&b9Fomg
zGIxa$@w=U+*YuI(w&s6@wrGSQXzRzh<q_CaLe3q}vn^+HZ72?~qfk>OE@g@Iv@-L_
zv=9pMX1r;-sQnvm(Z1>p+%mtM2!4cE4VQ$+qbCLxu>|Re<$WLDhF$}D<MpUym#m^9
z+SsYd|C@W)ydGZ<E_$Cw7h>rrxMF%WMoL4C$?s-=FZXmb<U5I|H%21(D$ezZJtRh4
z>;4{TR)HROco3OB@bKNFsCpY-*o<;lH9^h&p*}JWd?eZhFeC0C8R4}~MmX2>bm$aw
zmVk%GJ3^}4?}4+qLU;*!v>D!|&&K<*h8h*nQEC0OFhZ!nPvLLBtql;ZkxNG9UHP#7
zuD6s7?`ai$@ro;^`OCjkOCfdH=Pma4r_IUP^j@F5gt(~4$%yd<KG;hX<tgwT<~>X!
z7<uG8DTDplvGRz1>|ZaHZSG}Y*XEhd`raBr-C<d)$1Y8g)1H8YI4mpJ?ftaa@6A0@
zv9gUjpQ{B=V$aT#izCtYDv=~#uDZwXEU#Zh-1U5?U#Cw}BBjr>m;GMvAJT#MaQIAC
zeK2c5*~Ua&|G`*oSJ6j5-lxY?)~p#bGNdznv4{1!Van^Ot%)eM3;eCEK4wj@;OY1i
z%FlyBnQq^s>h(u$P)mQlu(!UKNXMl2-=O#20q1ywf9eb1%VNvD=dO7gm*JT{J0=7u
zyHHli8dqWT8D#bBp#?vHoihhgM!)|lbRF>eH$>eU#OHHu2M2%2>kjrD5xu?c-^P0q
zb9SyY0ywz0qcsNp+$P6UL+84y6uJpHNgyRm61*P{eRQY$;)?g`ct`$;z6~H8<wT&+
zx0`fi$GF29pf)fQy&f`p!2I&}I-XJVU6J9PjPHB_kbd1J<^fCAS5gH<x~u2D_;?<4
zU*PVdEpSBi&(BMsF@w*WFe*WUHZD4#8MgZ$4~_=%b%tY~qiW3PsF&~mB6z>GxQgc|
zEv7<*9PR!3(F1M~E5h1n(I&v&+32ly>3+~hN*G6fAl((k%GV=~7B2n;^aU!b^Sxs|
z9`7TV22K(^j#=%xjL2ETsCCW6$OKWp`63EB$$O!jiJ`4<zd9l03zz7qfGu^&P<&(g
zmBZzVKA!C-0et?w?Ss?HH9hy4NS;w{*`|#L*ry-(ZFqws-;LnG@Sxklf~?9aPNNt#
zMyV71Gj$t{USH}ci|Ts=nO}|l7zGh>b;*>WBtn3lb&I=dX}(}^{L4&QJY>!orSKZ`
z_5K1l{cPtv_|vnVy@byCZbZKds4H?QF!Hlb<h5GYDsy`q#e7|`4_t!h#mK2C6BDo?
z-hL+4|0=Iv{;<aVMMl_XPe8ap*QWOhx<e^OK~?)9P6d8)8?y79&cI(h6tQs@K2MH@
zA>oxwM4$AB{QTV*w@~ml{QlXoVf)(l*V}eG=Z5V)7qSg!6gprFZP-bP%hTI-k&)hO
zOY+`_uHLZgpCh0ETI+SHhm}ru=ARE|pI(+<KY$<VStDa23hffd-?ZIMMSfK|ZEXE>
zP2jhGc?SRasJniNxSH!WQMZ2x#tZ!KOF#9xS8P*wwP9!H==Wb5?%pSuorufK`@ccb
z5+-;MIBEUB+C%k7UTySs@Ao#z*BJUWp64VN^}RR=|AZ)wpqx7SK~k?MdhB;<7`we#
zTl>MN-_MnHle(+1)f3IEL`MqA#s(_)y3;*!4HWk#8)pCtU02zX1vTdDL|_f!X)C$!
zwpRp;3t&0+<x;tF6>ZlaG7yY*s0Lh7qg^fqnr@3pGQeR)D<DxR#z{I@asPF_0!LV|
zbQtyhRzbff5DAFq+;WQdaXnr>w`*Ypyn4g#M7%s<HhU7-O-5}x&|*P|_waN4q30fp
z#uJerCaj;x;YhPyl*A@Z=_cl+3Plh42!Q@ge#Me;0RZG+L%>lTG<gJhW|z${Y0<pu
zlNtE2QMywI{hqBP8g@s{MLuzNqY%~Uxv?xfbMdld3h%s+`djc1sWuV+qdeGZop$z1
z=%Y6Tv=7op79W4WbIl@q+e?$($7GrJjNfOcL*6fn@%5MfEe)ENxX{Ch!A*XTgbWEG
zFv8T#8~O37h1iQK^|BB<{)0TLiogGyg`f1UNcfMw;lnpnH_yJutq@eP9;xMp>8RX?
zdtC>!Pp0I1A6pM54a@enY*NgBQm-p$B5fe}+rl@x+#pfV$F#x<GCND7A5q=Td*VHV
zOJo-DSDpBAmpGlL(9QLSlk-ueuY=#t^eQ<UyuFM*v2i6$0WT+Bo!1G^&Fv|nd!gp%
zs#tWMti=pH#fVQI4#lAfuw?YmAB2OR0lJC7J7c7uc@FF%ILH7S$@tyA1}LH((cd;G
zyKaA7$t65o&{{JFyLM#;@f`#!uymD0U(!n8rf9CVQHL+jM0@VlkA&V#use;BRZSQ9
ziF+x8V)!f@BQNiDCH{uX#0d8owH|Eo>D&yGJ<S-PBxcPdg#@)seN%m_9CmN58^Hfs
zr6hO;IE1?Q?d=8KKL3kuj@O(xvKI!dbB0hxNv*l~*i#l6t0eH#s=g~_L)j*qUS1Gy
z{*K=?%ZDytB_a0OFc`-pA2MLKiH{U?-^T%OG}T9kOTSkOLA$Hrhu)y*7$Wn8hOem`
z3x-C2SeX>Qbta^laIdrWmmnY0FS9PU>Snwx^1ld_nfTO~eThWTzwL;}URR*tx2>aj
z@LHQ9lWi9wegz;rcN~~DYQe_kfrCc*=15<@^~s98eA^1@{Nn*A#`SKjsWLr{2AWdi
z&nAK3;$5!0Vn0yyZmWVd&BifdKQJ+)Jtl*f9~@@^JXOIra=Fy-;^-}J1ou+b2ffS2
z&LZDS_3xy0?&lbOPJ1JPgxiRRHP36q0zV96*veiw30%9<d7wk{sf)eYWc%0gko^||
zOtdfHqChfLx-8FwU-bu2h~2@p`rC<-l9={SIcJ;S-uKcCwL!@%Lw3^7VMy1X%I1e8
z1d4C}xIa7Aa<9LeNZPaf;<yjo6z%w8d4>L)?e8D!_RhLbyi0-)`2<gQb->^e*bxE!
zeSoj~h21NtNAKDpqI{D*wR&}a!bXgwQOEE-Rmn~Z5#MjDk4BNQj!e>{2fW(z9&ZR)
zl*sJS!bA~X-}KAJYy8W-RpqDN6Xgx0L$Vkp7!oaTUpv!g_vKjEAMtPsq;TbX5tf+n
z6uPH_=*2BXXC-y5PWPCL1ZAY8u>b`_BL2g|B~<|{<v?7t{{Y_Jl}OqMs*tH*ix|XV
z#?PY@G96l|u@WJ^#kMmss!h&FQJ|-~jE})*;|Zlug@nvMutX=m0U()Uj>}+D$Sz7C
zcq^>}rLw|_*5@nWT-?p8_Jv{dXk^K=FKlU(gm3MxxM4%3u?_0W74#p+H0K<A*sifF
zWEY?^r*7F$E{d7iXG67eVjQf;WGgPmccam5CSs6DLc=J2OQZHI;ZXR^?QE?v6=>#g
z=P%=%#YKzo@Y8fn_Z4FgkRZKuB;Kl<g^c$K{b0m4O4B+FZx+Ki>a5Ls#RPz1;)OH|
zz`aB9-SQf+h8+~<^f5q!-Ob*yuXxF1<NXODD5+h}Cx=M>qsRD@E^&htMyB^a8jl1a
zH-cEmT3^%1Et41z$nm9R7|N)1-+Wf=rTjAHC41AFp^a&hoWReFZ84S|k*G9;QCa+J
zFgt{?QS^?-{lYzrFqX)02miHKN(e4!!aQuFfZLBMPfUu_LZSKwAE=omY}LlIBv-OC
z7*$f8kH#k#gsMKQ5mgqibpp}`Xw)Sw`2>A~6WiGgtEvYL?}$;bARDkaql`acY7tIh
zdSaT3-G4NA#ONPO$w?Wa1W?2!^9*DOi2n!hX<-p5tFdiEkBsxnU!Rr!HSY$5-r54p
z&d*qk%B=;6cD+$~NgKC1x7Pb|x2-8Zdr-u+eZ41>7!K;3^)Gc!qNU04{qwUbIFkB9
z167jA-<?ew)xBKy@SWa=Uk8hb%RUJkw>am$EMEffpJw^W5YYR;<Az^qTQqCgNzb6$
zDl-nUn$cXlwO_O}AJru^e~%nL;%PTS*A~U02U9U?@xgb4jmbtZOV6ffid?%9xTtuY
z#@WNZIG-ia>Q12(kHkn%EQG>~|JcMB6@$mp6&KUx<A$DOknho%WfbmC$&H~waT4Ro
z21Vy9SSyY_EaUMHszHam<IiA8eape>PGpp1h`p>7DH{q;?^Rp(w~dg1Gq?bo{Rf6e
zcep@M6WoYxRY8nP<g!qIrwW}@h-9eMQTN-3t|ks%b}J4{<CRj;EDnVwk$;TEOyb?L
zA)^l%L~>c~uZDok-niNSI4I>7bONJIItFX%>j;sGtEM6sdZ;<fXU9-f5{&sb8^^e{
z^EOoP7v$W8vp@!}GD5lhU0Jxgxtx0xV;Q{4M2cw8DXqN~%%^5y!XjZsL<jEGcW(K?
zPfX}f#v09o+XA3pU#PF!E@YDD@sl66%cAzm9{s@`x{6-z`Ju0*X@)<a+WkX&Ed~A=
z$?}dre~2nxN=&{!PlYXjSDzxykN1SO3{?IAA@@z!x^^StQrY&Ft1=HK@a}2;1)lN%
z==*!BL9_b4^z81(+V-s?GQm6e*5^d@R?g_>B8HkL(D2H?kMC&1fX<D~>Y-MjRK*_O
zcBA)5O%&Ov?|V}JeYBQ9+Px*{EDYiO@8YKi5}-^CXzQ}^Bk>J1;or-DFKrJ!dODer
zhs=ZftT&^+F}6f;HmbdSz4cZICEnAEV!ZIypSMTHGvU<F7OxiI_PQSY-rlh%(BbPk
zL4S02-`7I=J@)@^v8D&VYNq!wD?|bMeJe^P`kx1*d*8<UM_))^CAjZTY}f2*&|k`;
z0fpD63x2fb`xBk8_0swm^!5LJ_8!d1{Fkf09_piq0)MVXLr~>^{fqStYz&@Dl*WG{
z-GZ6vyd-J|FMXDA`>-qOQarP#_}ivVPhx@~1gjfTr+F$46Nky^cS7-RznO7?TUA45
zj*F#N))@p}eU9H6%*7jjKbQ3?(6;q7Pg#Q#J&o%Ptc5VK!Tha&zMGZovqms9EHUil
zU}sJc2HYK)OgyG#eccu;-K2=$5u}*(%&EYyk#L?2IHp4TN$4asur)M7E}`RY<`;aw
zq8kw)oN)hS%GLKLE|JnE>t<FM;0gNH`_Xc{Rg8VpOf}8_<XE9E#IY!4**WO++V2Ad
zAgAiyYgXuXMuYtOaH$<v2~Obgn9H{F(kE4@JJ`Nk_e3M#^}6${2+{s~;nPvNtn52#
zYs%;=^a!#+S^+qf-Q5(@C)J>CF}8AHk@t-+D=HqnpNMU_pj#L5AF1m{Q6wivV7NhN
zMZCKPVWL=NCu`b?8M=XGO>38(9~LoB@(~`BND5s}Tu8}|0AL(H_1}70^5RP>w)bwg
zD9Ob~dRMQTzwAMtOgQjq^Pd_-M)vTff`k2ysoP_%N*XI$hnmJjYA#DL6qJUaJN3Js
z@pEYE9NZoU!QD6OxEtG|Qde~+m1uDc?OvYwm^S(o-hUEuHo7H54S$~%Us*)#PhS3`
zH-7i#JeD0p0n+|+DYO;v?y0&X1#C9QYF9;#pJfHIzbD1O_#4FQSpC92sjuFsN+x-H
z!2jo3(XR0bEaa4FR#-T3e~-3ta47x_B$fkCM}15RdF1FznRI&CNv3@Sf&{lb8o8z`
zA=?3mmkqtvn$mt`OGEEg*zn%LDPIyjLuU>`d-t+8fE51^b8i(ER};N^HbH|E9Gb=n
z?(R;4yF0<HvEUY@akt>E!6A6#!QDe}cMndczyFyzb3HfnoZH^Lt7`A6wQB9ETE7?I
znDY;imgNlB+@3KQe|gL7{a2F%>-d;SZXoF&plBCnk+HXrrU_flksR#}FLx$!+1{MZ
z#XRRtwF^2mPOdJa!3%-Z*{fn(y_91&iG(mc5)#jF$zN|7zLy$)HojLEgJRU#8y7t$
z5BQ{E(4*HE?*C<%_|o89;Ik!p;<R(%em5oTs?f<Ro_*TSQ+p$1BN*!uwFVfqQ$ACX
zD`>r4b}caR_5ur0f?;TDlUSf{Cd@4?{R6nDCUEe#1c=&@0rz3F*Z1h%bs7Dz1hhkp
z6`q=q+PCej=|w2icnj8q)r)@02OQE2|M|-?U)i?|pNBx!C}*jcp(*GFN%>2hg>yGL
zWK-JMWJQHd0cP75y?h1pkI&va%Fbem<Bq;Bb;GVNJA(%fP#%Lo5qNEmVM(D~Tpy5V
zo~p*RfYQD8f0cW!i4#7|Qg^ye?$pFs=u6(6&UR|K3{lwUa3WWs$Ky*J?h38xjCH8t
z!gajP#XdWO7ylvw+{5n3%NEFU#k7~Ak8+tPZe26af2dP-Jj%w9R+s&nnKFKMLwoEl
zo9e3`!W{TI021%;M`q@p(1Fz}_vG@)-fwya;y<KiO)<vu?$1WXHl@=87PJCNE__o~
z4epP#6r+yWxEQ4N>(1BRDd|6NV#_${^o45}qC^fry#Na|%NSG%6kRWfQHKtA{Fsf~
z%N<3ch}fk}h#Fj50zf5P?BIx{Hg^uQ)-fNC;tYrHCg6+0y*Paudq7Te_d$G?FKtI1
z@#r_WWRxK1UBSwyEqSPkyr;~>EjjAw6HIJWRCzvh`y0oVCRN6z3U;_E<(POFAEw)w
z<h#3$TIrbQ%N(%8L(oo*A+OBVZ}d);kgFP?;Lb@}xoqq}x(ERHl5X3Xs!x6xm*e)b
zM}Pefa0VX$(^6SQeD(R!;vZ*(zZR?`dRVc%5yc<}LyFqK9Yb~Om!OYd4o%XU?~xB1
zycxUhWt4r@7}f4H3H=>yD-J~#^)9jNrbABNYoqYE`NZWQs}swMy{Sp0%-=p2AoE^P
zzsWdLYeSE04>qZi-|APC{{ajeZuD3T#f(+muNjdL^`Y|mo|A_>wEqCl!<PKF&#8u}
z8Us1}{fBFK-l)y@pOps!|B%Fpy%eJ{!0Ma%RXsc)+@dGmQ$;$TawxXRm>~92)R37Q
zvIf8V8T4_=y)0|Ne^lRh;fh$U>@0@4v@HK#-xa2t&9`FVqx}Qpw5ZHtF+vR~J1`vk
z4%3G-GdIs2Ay)8Y6ent~i&i_7<ti!%232hx#6~t&xVmrB?j+A!J#30kUnIZdI<~re
ztEMY;gCPDeuGL`L@PDK}fl$HXF%}Me{;-86v4|}7^r5f6{q?Y_PBpJOT?pdb$pY4H
zbf7N;iARD@hw;hTMYZ99R)!7Xf_s1I>RAG_{D#!fPYHD-=AS?MKUSjGQ1ZzY@j{_?
z&2tAZOO2Nka=@sq>d%{i^EX1)I1TD^yRMa#1Lbg_mr_E+*ktbi0jkp$q1{U%s~aOl
zWMb>0?zK_;Ft%AK*_8_S0Nqg(2blasrp~41>0Y>ZW|I)-6$R}VPpR0vN_inaGUyj~
zO_=zoIT3fX41^tP8_jh4o!Ag{sU#^x_!{l^p<tdR(0-nCuA3!d6Kiz65DV)#$FR+z
zhL6xBI>XYm?ho5AqT$i->$eX)@GzzzX5}rqB=YiL;r-+$=ZUo!Mnk60h7faz;@t?0
z?kqZbr0;rLnGh0T__uHd{I>XRWoxT^qN9B$=ad)Iv?iqDuLinb@bm4}g0AT{)>AVv
z^kP%1&PUSS(HHT-&{ylfBKRjMv)+uveqVglB+?B5_VEynv%EO`+yb{;>R!OXVe}5j
z_Q0Ihihdw3P&8|vNQVW#a|2u8%|y1=rR9=pj=fGdFkZ@Q*%6E$zD;fV-;1Y5ft}fE
z_qv+?dc5vv()jki$oK38p>e1CMvAh<*xSw0olK>nd1(Y%qDxy32X2}m+4JVfw_Omk
z|NG}__|yJ>4c8j<^08mzsK%YRLUSlmb2&*fKz>W#^hbrg!Y@Uy9d{c04-oP@eKx=A
z8;`mRj5m*Q=d^sFtB%B*#c3QD_fq4X+Wv9`8P$yA#)(JwdBIOTn2chA(Z0^K#X&fO
zlE;|bP8`f%e9SWEXKA>@A>;9IC5Ej<wltzw5BW0XANwc8YNX`_&L|`;C*V9zsR7X5
z!Uo4yyIVXJtH(qn`~GXBq#5yD*@8cjdPWebw{5!sI^UxJVzwIEhEfqf-<UO1knul&
z-Mi~{%(4|Zvipq)6M4K;M^bN%C><<?NABMZ?1tiUI;miO13szcr7qRHIViK>eskRC
zMT%!0L>_nU#_~I^7q<!UL6w3XvwEA|lkKZm0Jq*Z^_pWwMz`fCwRo)47i7cRb8FF;
zJ}=C&^y~S@7DQ*>sz7)5i%Z&1VsEO!)jT4iheM2B@sFC1&HB&9Wr0DXDo=x02LEIJ
z!@r=S(@fr|jj1V`7asor{FNO40N3%y-S+=oUVP8}vTmKLy0<+liD<C2Lo0~&fK&{F
zvRqKY91`fbW9H$O#EFO4V0fXnUEccvrTFwM<JI9mAEC75C>VGeyqhao^*S>*cJ&ZZ
z^bhdEN<(AXw{oaaIScxEf8k5O;AiFbF!Vwge4O6-KiA=5pJ%i!8r(_1S7MI{F_ThS
zC+oRpk>7M}ZSV4vYW3AqhCR%S+#_`1%Hw+9R7Xc#R(Ce}moHcmo5@l<aD%@j3;UcP
z=U&FPCT_#Q+v=ylqa|21T@i<5Vz!)C7RS<*Xq<4AN&FXjawFJ$d?3ly>MKp-oN|A-
zsKoF#cAlRRN_@Zj*ASWk^6Mxb!+(I$`EBhW2&Y7i=9Z4nHOrvuA?81TGiCtuyxG&M
z@1>eGEF>t)@Qe=x{p4!=>z8P66zRB4FuM_-@b%heYWuMpot|MKycM5HdS1@h97fN5
zt2WT}DSW&#x6hk%sO_htnvup{0a1=6T&-;!nX_8Lps5sruHgj#0A8TeELNk2SDu3@
zR6D)ZPyfemVg9#oDzg-$q}?=0W-FQ5z$~x%(RMz%Slm=MEy#e=m-pkbx%)s-1AFZF
zNObPC288nl7vyTqj5*pDcS{*fMA;LW^oeG>cXhCHgB*PUZkVJy;o~$*o5TsD8}@A^
zrQXaQBNff>xh)u@JZd2Ubx&mv%2qL%G^e7{OUo<F5{`>ooQ$g;;AGEvGt04;m+{ru
z@^H2eJ4u{b_HVmwT>j-%A@%?SF$<j$Zlc++{K0zF_wPpd`Fh=2iO_i!AR;_RDk0qa
z8e{Qf*fC9758!n6V)ARB-nTb?;l*8d@M>=Ep#cQloQXQvKvcN7dg#I+g+i9%4I=|V
z{FdwJ7gOm0W_Jp2q8q+Kfi5a+(i$#{+Y3aV>x^F^5f9z*1L(4PM!+DyFM}KiAnN|L
zIKRb`A&BZ&+CJ0Hb~9#okFT)La#^i+!+h(TVV9x~^dCS4vC`xDBu`8>;4j1N#VwlV
z$<naLA7(`X^-N1;vi*j3Uf_$@Oq4WGqkE1V@q1LCp$neyLXOWOG*{T4S)JZ!74-bf
zJZ`uZS&{C)8DkU-K7?&sLjB-XKq88UJ9c(17u{O^uUi+ld$-F(^=aPZ&Wx|w73NWF
z39BB@k59{0nqrs^?zg%<vO77ymd2D%yGWPYv5C||nq5vzgj1!R_^?~(SO7ifnjD&D
zJ%xRLi*|-!(nsM5*+)<J$I<ZTXnn%3!Dp8&BbRS^f`vE7RXY*w8S`3NBGF3YaYb|=
zOLn)v_o>N`NMsQw9dQwwKv=MA%xNe*R(irIZslrj`NEcR;p*IROdsPgB74Vje1A~n
za7;{sh_)Ff`gZD%NLklD#Uii@yf`K|hl+E*6X9}qTqI~k;aLY}>3gnM@_>oEapK5~
zF1HtiX1R^>XK~p#oMRizsbEvnAJSuj)7G%?Z9lGpo|vX39zF#go+)y|3p%hX@~08-
zKY+Xx>%x~j(trf7x=Y(mf$Q3`yu(<0F$iO`?_s`=Of=+4oe1SdM;=lkD*S?GtJI$>
zjOGgaI4IkV8a^*}>#BMWzWXQ><wPcD(hxs_m$LSm&eN_g`1qk?C*W^<&+Wn4R902n
zSjEL0!b>7UU>D8QmLFzq+qr0(82ntw{ZKETmlRCu{(oP?TECCx3Gl~aecH@|eDWDQ
zFS?ete7kKlG9ldzxYz&hrq{O8qgnF%l($bQ-RntT0=N9I)^Esz4}H%o+T9_qyZ^f^
z^8fYoX@}LCpw!!S?QUex^US!(>$1e_`9|DXfBEiEUwGHwqqech5JN4wVr>-vSjsqR
ze^EgR*bb#+-}^*3oQV?yaWvx{tIIN=(2XWc>~wc!&d^m0ln4<Ia3qhR%n$IVXcsg&
zH<V^3-i5ahKmGk~FYC72r4%XSascUQDGPuVY)L)Kgu?kq8AZJ3QH%}ML%wd{g<j4B
z5V)}I6uxrIR&QjvyYTqD#T&kj7n~>Bg5Q#F;8O<z1bZ;v{heUv9o@>kLW?jYamF0J
zf3=yfL@|)GRcxpg_SvoLnq>j7HOlTp09GHpz5yrzaIn*pFh8Ixmtmg}BOXw6F^0vM
zA95C7N)m%LV+Kl;1GpA(AakI-_=Xezfp?Ym==j-uJ&oXM+YMVi|2S4EYV@E(U73Fy
zwM&I(cG?VQ!vBtVSuB5Ra+CK2eE^y69@XyAerSo2gl>=1NjFo#v-Ah&5~A7>q0b^J
z>)%&9_-;p29IQwwQyjQ)?#;WwntRIe*LGgf_Z`DW@k?oX>czA1?$OdqyE?G!sCkOn
zt?EShJk?G?#A{N%Ru1bMyV=ZZ{dXx&4v6-fZzS8tMo0Kwa)Qi%r5q|z{F_?}`7g*G
zSJ?@i!6R2d{|@ufOGjd8tEX0qpo|pLOhiH1a&0!bKF>O3u0a#}Zzw2$X!BJh3JY1#
zQ!8;j(Z=BLPmc#{fyiVJMH2d1hmkwfRN}5CvuuBB>UbhPYWU()XU&s{Kr5&ISlr2l
zqR_=3USQ_)tH8;VRf(5H#YKp_l^%Stv4%<F(Yn}Jb_?3U;|c(Pa{(JJu0*d7g1iob
zGo66FvB7J|r_Kb+YY9Q1)J}E_HXQcLY1s7?SbwScTFiY6ADu0Mq;d12e{RrFP^Q@Z
zkutZNC#Ja#Mz;64{@F6SlI>|C_heIgWL2swMtHS<nwP{vB{m!7F1X0~cjGk<Dc}!;
z`S$}u+gPn(^0f@jz53_;6J3Q9=X*gATmJ6M<E^lN7YA(rwBk9SQjA94;6j0u{1S@T
z{RZj|vbYoxeQo7u;TVqbh(6Q|_|J+mtuLV_lpC*@=Utg`*Rrfi<lpVACho>#Q1eDU
zq(9G&zY&hz`A=Sr6y2PPIW0(%Ke`XDrlCts`+VGuRM>b-8o%>%4t`$mE~$iBZQ%O7
z#v_&bhR^4`z6bKI{rIVZo725S_dhTzyl?-X3X@<JB42oC{h>PA=tlS-^DMuO?18IO
z7l36Ehhs~HUqlxjLPNUg#NG)!?WEbZjZKv@b|*2Ra2IogI^Wx(0Nmivum;nEx3D`F
zcL}-%jGR$j`cLLa_I^cBa6bVBJC?c_MW`{uCFFJ@z7SWN1|3rIy_`gZeOp0VsI6|a
zq4JAQ1KMwGx>KmVC%)tcwH~c<!#WLZ*%o0i)Z^iZZJrp_o<%M$#C}<t>3)?OUe!L?
zoftPqVXnyJT9zirtr#~an!#*Lq%Uie5O`PTX4@&`9Js_tPeHACCn~>@#6y{WS-`%y
z!n8Yd*7S0LI7Hd}uo12F2C@+5s-3ZsDu)yZ?o?T<mSZLso&<8@iU0!BjzrEQ4jx0P
zbAwaw#@LaF9&k5K`ui%{BETG_@o7|r8C%gvjPdTVpWnX4Pj$jOH|WU+*-;R`bJ>~)
z2-pz<rjTrjB)a4<uxpc+(;){#x}OHV3`vx*8xBd=tC!RaStBs$o6D>5e^TEX7?;5p
z&kklG(K;IbAcYD)XoIU8!hWg;0j)%$0y*4lFX-9ymVUIAiW~&>P_F%=kXWB?;^)h(
z?`_}6kUk$L39ff_@e?9T|LRl=u!=2{6slIAhxOfV9tZUKo=*<h7_ws4`8;8?^1GdU
z+EHNXBCVvw2>SVMt(-?vXlwfDUi;>$<DOJG&>@xWRg1&;`OAZ3t^Y5Lyj5Dce*i7x
zpRi%0)^^8m8D{((QzrIPGXAYHv&ao5(XJB8c>l*z;I#oS@J<sp*cmzIULU$rUMO{-
zJGMhQqyPQWVd!A`ncB1KL;U~qXU)Jnt*`t$)6a~5%#Yt{aszcjo-oQ_3S7jDDvpd6
zJF)!aENZQJY?h!vSUC(!XDP5azOwYcJE&bmjYoyw{%c+J#vF%St>Z^B-IphFOQtM~
z8u(IqsK3sDtL$9Dn7_PQoYaE6)cnG)uxCF=m`|0G+Y-m!Gg$2)HRPN(J76Ws$>at1
z`VMR=7_uoNbXsJl)Axh_sT&d%FY?16F=h4dga#n+USTO+RI$|bj@(aZ+kxxuV4;Jt
zNJx*5nxPz-JY`Sh8GbjeHy_#BLefhtVk;3oF3|hYgLt2RXZ|r+3tm9??4@*06^3g`
z*%O0J@u<q{Yn<Z;k1A^CwToQ=Ia^^!KJ~A^G)z`SlwgPLqLe2IcB+hQM?0^|Xf?J_
z2k-HfATZWODTa3#70#GvZN%k`m?e?x1++(%4nHO^KhwS@uwO=5*twblc;R$5{%pO(
z8u2F!U1$lh7#FpV_(ic4_U4F3&44%0Ggp^{rhDt&u9bXB_|koK{)2P&wl8xjf4U!N
z{u(c0wi$`McFrYYUMvSpXd(QneI1;040s%M>h7Yr9*0+5Ykgr__oDu~Hf?+Fg-Zg3
zGnLUW`v-`*<EftKjyHT<^<OdK;?HY{{Z%X3`{+{CJ(CigWAYqT`ZBa9peO`U+4{V;
zTKdDSEbFAlOzz4N-$}Ri;O_dcY>TA(o=~avAAtLg4^xaLxNq@nOcn|Xbo6UydIIYK
z`s(bjShJkpzK=RIlE0x01#H3ugZ`$RjomDMPfeUd?eBu!y{9q1rnjEtQ=KJDI`|Lp
zS5nmqN#ATkkwfP}^1ml_UKLPSD4!Kmt!GHQdA_>Ue*ZMsMyMp_aK~59fPHCm@M!1;
z-_6Yud0c~pANN>TQG1qK1{-s9P!Q`sfW&$mL&#E&R?n|`<Xb|e5YHP`^^I0r@|Xd=
zE#;?2YAn7hSUc;nZi$sT7+q5WUy^+F#ctD1%{CUaME!N4I~86301sp8Fi|T$0x0s(
zahWXfS9iv@KwVBaIgMZdDeU1>*{pBNF9CodB913MzeoDMl9i#2i*1E@?#M9XrEN<A
zQgDTxiq42gS2cO?_0G@?Nk;PEOCI@rDmlu+>tVaw^Oy3WnQL`&1x-go*N9i&%Dhm!
z|JODLbzP4A=e9_??ulV$9!BJIq=DxswoN^w(uePR&YAtTf-3e@VR3V)RTJ_uv?7XK
zwc*mV^LJQ1a6v&1X1{d&kGOyT$J!nFVb-2^r@lbiX#Bij04vfVvFYGv34{A-3B@4M
z@t+`2%{L2O?uT?`fSlKSGf($(`Tc4L(CYo3N+H0F=h9h5E`|yF_i(%2<xM53ju*Jt
z;>^Z;njAYI<ggrLXR4pjKt?>j!gjl2u3z&T@%AGCLIOrRe*ZFlkq{tEuM*SrKFXT*
z9k-5OdFLs8qwA!YJ-UwQNbEq|^&AvKhjQ83IdI366;8%eJRlg_=?iI@#V4?qmRibr
zu^x)ehz3dw%Ra>iZJgY55JXXI<YsqpO*`Aiv-C|q&OtM&dM*QW7Rt!p5`d3R<GRt`
zX)ksy{{h;h{sCy%h7_(HVm?jHhO8p<n@4N@{y$YS=zpCmW)twy{<Tc{qjP5t*%zsn
z()kdxh9x89+g=E^%jQsfwZsd|B{ly8ASfKTTLy^*qhjA5+nKsPHh;@+2atrHSew;h
zN)&wpLx69s`??2J3}n>tdDGyzFn^(DJOJBLdeGj`*E3Q(7P@Khovvf4F*g!bc~>jE
z+E0mE>srHx9am;&VG4n;+PuJ`x!%YPugA)>>nA^d-aiAXSP`9=K~AkmmvD7RzR=uK
zh8<%NKkCixKoy|#9pDnhB(Q4r79L;6oT<a1aopSEqFcwYB?QOPoGE0L_jsCRNNp@U
zPgz<I=p`qFX>gW$dw0~4g<}Vd?CwMe#5Sqi9MJm{1-$<Qa3XPq^>YoLs&AkUThZZ6
z<a(9v&eFVdZ<MOF@)eW%;UB>U^I;5}>HBTDs78W?HJ{%b>9JhfgAoOO!oI&UnSba7
zEi@1HK6w`rY9)c*rtAp?zI(mm#pzv=Mi_hzxB~gdsA^}1Bi@wN0fjXj_))@NTI=~R
z^MO@`h6rr-7cFFe5RlVhAUh7koCbfcd7Poy0##|}9w(vK&Zfon*=4w9bj~N2A>=1n
z*s{*=!5`2K8FXoLW-L1qM@*8FbWsy!`hqs8hzsb=7n3|g?Ff-b`38YNy3zhNUd0P+
z9Kg@EAYCWHmQUqGR;1||>;e1$Oy!fM1u$6?sT($S^BvhH6{W8oAJp5ZVb^?jPeVxo
zluD#YW%;+ItM4x+#o-aStDkvD_1Nk!Q52@|gNC-JsT;x%z&k<!+`KDk{3@pQu&A@g
zkWf%mZLZ+JEl*K$V%cIundKUQ8+#hQO1SX}IYftK?)yzGj$tL{r_$cP(4d`7pSGVB
zwRvnUmM|%dKv1A&oquh<u-lXFDfhAWtCttGi-<4iv0~3@6`6E3+W2{b=+bp6gWF<*
z@h2KG;Ob$pC=P}n)%rztH~w|9j1}$=NbsUYf2GzRCmw!kv@>r?UkJWc*glVA95(yo
z6J4J2LQZBQmq}ZYE(lL1a<(@7`gmE_?)AwS-^G0Eb@Fi<cOw3JXVSvTpDCmh>t;R1
zj%L+V1)lo)687f@TmrJ{h7AJeNN|OmXTjhlGxB~97cyR<=b8)2YhRGB5?7sVb4L|N
zbND8uwR4rK4bN~ChyqT0{_Auso`bB>$sHN~*C0;-W)PJyD;0Ix_c(Jt(m2U?gvE7O
z5W*^RID9-m_L)yo*1=ZjE8aJ?i?!08hW(;&?UTz~yO^tUrf`^66dpv_h$KN!eB!;n
zR1$!ySfB~5M8Jw=Zt%FYfF0F4zF{Yz8_^EIR%A^x56P83{sr_(lj!^THvEv#vL_by
z8R>>zHA(Bz^#VhM>7Cu}6dNI&syV<iO-~nIV%#xch7=Iih`l*P8wSR^@&w)q+vzcE
z9DjRW!DyK5xBmifA9iMs*uyz`|B8*K9k~ojv`}=p99z33hKtj;7<jGkA|k`-3UVME
z=8uy~yBg-)$&D4`dB46`yL`<sHlxsp`RCnM0?LMH)%qg<@s`vU`d46n38^+Y7Ec?c
zCeFg)qTJ9GZgOefa~bKkfx_<)b=@8Po$X~8pnIFfz)wBT%^DmsQpB|_U#DUIzbKfd
zfF5NpRiQMH*XG^V07lPn|E&c}cO&3g!1yglFrIYP!1yt)zhmhmKJ&#^8SDFv0MM-t
za?VC$&XD2{kr0O<W?t-{N`|JdErKUw_@t577wLzKtp?&<Ln2QHrCA08JUgiw54TrT
zvGe{mRrb>{D09e*)tfNltPmn3Ec`Jd_L|IbDIvl-q2eG`S*GXaJjx<3&7Do>B0|I3
z3t8*i3he6;<H>5pr54?UWU<Z4WJo@UvR}jjZY;LF3qxA!<JYehV^B*X&y5ddEDk`H
zwDjg%XzqS~NU$1&1g1;9hzUa0?wo16pAxlo2IU#xnXkAG-wbT27=7sN5egaF0kM$(
z6xoe2p%VIw;GHRG*$LmuzX4&hiiIt}I~l3e;GrR7X?OZ|IRae{{oCz>w4m8Mr}SfC
zPv8E&63(SL0!5of9^Nt`TAI>aYx9(;A?%pr=LBP{O143O9eHS2W-AgpNf^&l9jHvg
zP&b<q|7!!F4x@(?yX$%v-Y*F0QE!IKp$CkK=|u0A506x8gm-^oe59Cf?C#oF(8}h>
zx97+O`~?%mZ8e~r;l@%M5|K-F*Tc@9Fo6UgD)V>AHn1T{Kp#<pP+Xw;m^vGIz_7MK
zY`iyIW2*K>@0d&eYPk4>(OeOa84JROP2Mpk{T{esc&jG}sdJ;{4aK31-HVdV_#ZTL
ztqZ7_;<fJJ*>j6!y4F2nh%zDCAq)Ad*-kjH{%Nz8(?6i#9c^`oQm#a<DFH^-J_S7Z
z=XdtDYJcUA!pLya04syux2oy5xPzfRmahQb1`6)EM_~=W@W%t+Kdu{XEhs+>^>cxc
z{JrlV43TN`WXC={97Wxo&9w3^G9L8T8}T^sw>A<aM#QK+!f()+$5&k9iGGBk42tkJ
z$kO+hU&-+aCl^6o<En!hlui8mLN(&$zsFrS63!QYud*rNS76?;1hdjV2$8rhR`S+5
zW}#dxu?Ru}eyF!)gr~B)l+uWZ2ur?(FW$bQYrfOHAUM0Zh^kq>HixNw9~ar4Qw{36
z_cLqQo{N@JV~VJ>2YY!v!^95Y)0|8e)L@HxEx*1$bVZO-j$xkMYOaC+12Fl{QK)Lb
zvIx4Cra8u4w^obK9kA<ry)39L`|2XMa#hMvuKQ>qQqj+LXwC1^heo0?h&n(Ua5mN`
z6qot)v%}N%uUs|=d#(T80urzX@fT5${^M%5F_nnXnZjPd$w-IaO%R{|?z>qroO@|;
z*Pn(W`Gi*dV(Cjid$pp+mU)(#I{=*BjiMNSD(MYs4=4B9DEIT8aWDO=J6|2KxOQJ3
zAIdjWSjirlCkN(vPC&-+EQs-PL+PcpTbq@WSP_TRrenTx6VVv9$Lsh@yr`a~*}nZb
z!B;zhL$@gGAnR5A7B*B7c@0DUr+dSNK8lUiK$7^c)m-EQf@B<n4KuCOuJh_4iZ5@`
z6f*l{ZN8aZT-;~nO#O<@o_VEkUp%h5F;`s<rWSi#ijApf1bu(xKU{Tt#=l#506m{l
z+qgYvx*RvO-#vO@yi)FHSHx+~KXK?Fhw{ZcIY`1iYjpm<ec^~jP8K73u(YNO>#r<W
zf0a2$c)ciUxbNYdtlPGX&E0aS!giu@FWD|`wSQ0cyc02AM^|rc1$~@ZzQwcyn*T;y
z6$4*hRpCsY?6boXx0UP%U$in3k~Jc(gy|pNY7;bEFc5T`dgkK_ScT;}bz_hKjKY_D
zoz6KIJD#nG4ZBdKtPr)0LoGuiZLYE?*s(J+iGP6H(VvO_m!tP&J2k!`83zt6mnXp)
zyks~Eg_mJDaNZXLf9<O@^(4{NCbnEb&(dDN?tt^vWqr}qn_K?JFLyFg!bF!Lwg&+=
zzo&vCI=UsMU1xXJhx+3Kjp>`HsO~Az3^;C&1)RJ3RYzAewY-qOAbZ7qzncB+dIJ5M
zx^@5R?IttM-W|KwA46pDzz=$Lnl`sJ@*s=<I6+V@Y}EAsXRh1-_XIm*uzS5oV{m34
z(Aw6^+VRfzgV6n6Y9llaxQa9VK`gx==l0`a3u`0A0pSadU7fo*Kiej6h=EX;0(J{w
z6{5t?a}jSF>SSF5cQ8ye<ZAHKKftWw%v<nVuDw(gG+|tDOR|uTJ}OhYZqZPNdA5;h
z;2g+4B)jDCw|QLS5uQ-KEZxc$kJ#`QuU*~P-J&;*Tl;$=zxzvYW8@OX4sNDYaxD%*
z|9c8BE4aT#RDmWq68<|t;47xCfSuh>hHm+A^!Jy)C88s$WK!(j^XP4=3P#kD*fk_@
z^e-qrXj=-}C`8PwwdvU%CrGI<QWBQNTWOAzD}N$p_<#$*iA)WpoBLaa?m{hUqB{8~
z$k7|$YVA@RQZveO^;zoZFC~2q{<!8Gr1`oFr-A@Q-zSBZLrr;1V>jl<=Ms!hdfz`u
zc*UZIrTL(nUl$GP6=+oMw0So@v7$zt%YKzKVU=pNVwr!itg>4!g0b4w*1}XEf@<D_
zA>(pp>F<|$2OO;{9u@}>{Y!?0t+bjb&r}9DlqB^uzZ{=47BGe+7HHl^sV&$wo60H%
zaJp9Lxx0Kyn;ha57D}O`XAHIcpdl@XC*{u7pC#0Ig;8V*Z=Pbe4K5*#Zo3`h`Hk=o
z@R^kiU5)4Smm(>181+Ey*8SV7kqToCh6{geg$mWArU^c(mhHTuyb&+BK^f{J`ss3n
z4CQV>%_1531fn6$*+>Tt!Y!qie<zMPa37(Hn!r?IsOr)a$G@Y`#?2lw>jkMbBB13W
zs;}f9q^@TAO{R~^1fsFgM&;1`kaEs_l`2jLQ9F9oELe8uC?2G9KVrHxYtye_wB@15
z<EMNdVw=yG!2R>yK4PG5APH@A87lW-1zY=GAqpMr=Z+CM8s5C%iAWgdhrelox{SwM
z<yBS1!+{tfuUMiz3shKaxN9h|*Os3vUm>Z&Td=?$Z{B+zZMK#CQJ$MkWr0n?qDhx5
zP5+?{o`>0%*!IoZYY^WI9lmWYR^cC@LyHj{+EMR-q&$$UpPaDSfgFyRh|aeSt*xnh
ziVV{|n(!<evGG9IjyHc7p_vKn&M9?6k}00mqDhF$$HtT9`ZPsCO8B`q!<B|hm&2OY
zIF}(Mgp2J1i}}<*jLfG}&01qr63bXaP0;rm_;L{*ew<QBbb%m8?GSfb3LAQbk>213
zF7)O$v#?QXjZqbvWVdo!8Ea}e<&T9RDG?r8W>+)eqa)zw<UUVBKJ463W>ypg0&5Nn
zl>JF?HCzc;9OzQE=Tm?6%4K<^#a{6~vooBcy!;UL@ylrH$YF`9PRf<Pe5x6G=b~4k
z(64gF{O@u#5q`xheqId#uWZ%_#rj9Cr0Fbr4z`@{K_RM!K$O68vUU=S^!TD$h;daj
zS>G#}(b9`}?Z{Cqd1;Aq)(-;fU?CD+s4@4SHiyvc?=G4Z!81o&Ge35N89jI<^{oPm
zgxT3>sae>$b5c`cz2EsRHTDu7Y!DjJMHq^k!xxGm%1II8;fpsNp_M>1Ml!plPkxgA
zslFv(4t+#FP8GE&rlL~NY&W;FDN_zzfgj>(oazwP;r*JHk{XG`zbqbstAn|$4w1Cc
z-3L|Y*Nf{N$~s6)NqX7Jog8CQ`{>uXgx{?xiGzOCXz{OywrU7P?C2vxGCTAV`5kEm
zSRy=_{SDHG2YwtgkbVT@BFq#&diKv9OCAY)E>8|+&kd^+z$%EUmMtSmzoCl)!1R=C
z%t`vopTJBZ?b1&wIP<Itn}d$h_^Bk|ZoK7)-uWk=5~|C4jblaz4G3dQH>+kX8j9wY
zf@Vb9aa33Ffc4<#-?9e6X94VKBH=cXa)P#~@?SB-X@_m3(gz@R9@AV}<ypmj2G|$c
zcERHSW%n`c4hvNu1#qpBy8Dr|Co=X`s#e62Q*O~boyz;x?;(+bg^tQ-W?3x8pUrOf
zQ&O~ahH+=rN=&#6fwZHloG#YUd!?!DRv-7wWI{<2X1MakqA|@{g@X01#%&?E5({r{
zDpf^YbVTCS{Qfl=sVZ!E6`>$P^B=`7tl8!6Yd#5;^EAhKxWWN-6y=(Q-l9*(y1Us8
zK+l|XS~{Undm)-HnxBEg%jJ`jO4?udF^C%Ld9cC}V&7|s%tcaWER-)CKOxtkwWNIe
zk#FD3R@PjYQVyr1#8@aI43(%bOyFUFZ>lV=(cMfK*)Cf7(Ad|Std0w&l4TZPX5(Rt
zxw*FD<ZMi^5D)JMz2Y>Zvr~jGG&&0G6z(BWu#Kz63SG%EHnvkxo|x~waA;b~Yk+**
z)CZ@r174SU^_G4*`=6@Qdwy@wUPZqi_QQAu>Nd98>c{naAC}q9kRb~<VJYCPMWBOg
zK8}2fZ9s>3*3J$Bd|-()DzvJk7`Xt9wGew`%-7Xx^A$|c2ju|DKgTbPqm>~U@%7P?
zLH@8Ucyq`Q4#dmuKTYE4O>c4;O?cN(fnBd>pTd{>?VPjDPYX+pjg<^KoY=F(lm<oR
z$iI(Vj;9TZrGu44A#;f`y%J_vO4VuFz-v#5eF|dh9LT%kP4X~?9G|B{`IhF!t{)^v
zt7fA;0*BPTCpR%*d^UZzg`5-p$N3;p9*lQ8{&I0<?T^RP{58%+-!60}GP3Vv5&%+x
z`Ev2I0(R7n2&Cq?<%h1Y+}HhOD#UAy-luekXMB<D4;%X4dSv;d4eZFgAm3VrtE*O-
z7XfKEKl1b=X=RhgT5t;p-JxRoaATs(Davi`AAsjvBtS58Np1RO0aqMT-*2haV<WIW
z#shEDe}-%4jsE%NYY0>Mq5q8Vmy^|wKq6NQGPxjFiboAfQQ+6DDax^R>e3LfLt-9v
z&ZhwGeH%nhnb<8f6qKNSvd4qct7$J?wzdYNoG~by>Lijx<IHyg4q%A?HK{X(9CMKC
zur$kpUzSP67xsQ_v$3<F0m48Pa^OOc_xV&H98?)XhQQ(rK#5(Ixg>)vR_wlT?47T1
zF&6n=3_B+P_FO=9aKoGjBW($Om+;hiam1TLh}y`ZQWB%*5qDOk4}W#xjRj6((Wc$X
zexfv8h~>6C_K_J3vG`NLrCr%e-d$Szm@Am&qR!h+lk@0+2nz*CYg@&g!9I`S1!>-f
z4!{9-_?XW%2v#Vj%c60*w$bE4PxnLk+o^B`c7C6Ux0S5w2g5m&U6P5KYVdAjqH?#%
z^wZ^2OL2b<HimupK;s;oLiL~yo}`lRNlyweRz`>+7O?%Si@k-?-Bzr?9*&RkLIKn-
z20RdoqNz0#!O25>+ThyIhy6eGddg~7&Hi9aD?TKy?8oNG%pd!xL~t3suDI~yvo`TS
zaAjWre5HqNr3hnTW&%|zXLgLF6uUq6PrT0`@<LDzb0dt1f9)?1=$AGFpy;5WtsxO@
zqe=1GY1gSV-5KycH`A8nb5WbG!6Am6aHz6BC|m9`+e|6ym(MRB&6eEFb>EDQ?JY&i
zsj?shhI^uJ)Et<SE=z#n1|O{C0Pc@10Mbp8oJF1~#*Pq(F>?Vk{{<XG-z@2jM|Mz)
zwREi9ebz_2^hY{aUSkxuc!I8IL9&{Nc1~Vz$9y*RUBR(enyNgmywU_t#v}3j$UMk?
zz7&LyN`;S!gN1+<K;w+9o3jKaedj`XSCoESVV32_VmRmn8H**HF);Uu*BG^ltFrT;
z%%;ha$y7|8JNj_@t#iStNsZam+kI9-L|I{Yr)EEmPAp9#DkCN*U4ZU;umGF{o%&16
zgMtle^Cx((M)sX#JE9o7V+~cIzm-!`R{}~^1;J0YBFtT~^E8QS633P43zc7MOPbMR
zBHq6AsivIWK6U=+*mBh%DoJK-b;`*UWJ!&9X=-|zFW#Mh?|F*|1R95z3;|Jw=M#2N
zMLk0<M=1<l`qPA$EP7w_m)Kia>+7r?LRtDNcM@~pw!#G)=Yj{rQ)i)zCt)3T!w8an
z#9zT6>Qa$ck7HrYK;G@w+0N7;FXyvOco*!hQ{H_b@+Y&wIKiu-??-873+jY(GcIIo
zr3M`=0f3b?ZXyiw*=!|&ohXL0yK5{{ENcQ@xG^`3-lhQ{Z05xIWa#-1kh+@hHdUKN
zLY>FBqYT0_(p{J_WF?P;VU0yZ+5s`Y^q;Tixc_e_LCn$SF34M6X7Ud(cuom@ap-$L
z?k;u%mYy<sS!%nv`m+#MRB$id`rxr?=|P6oPCqW#@JC?pF9m5+YrJX5mv2ml<?Oo)
zPF_^MsLL^`!K8cDL!QPVeLX9zQ-&@dBT0}3E@fimC^r1USMLu6u+YN7U#=Y?6S+;p
zLF<^{-QU<o_xIXMGm}pk;l1gUlPf3}(CAz6g58d?ut8La_KjQb(}_e#{yDIRmAaOa
zY%=`WX8=LAcQ_m?%w}P5Qs9-Nf>tYb>!+bKTW(Q+_L9|3zt`{k32-fqE{N2MI?q_*
zXiRO49cRibL^zCGjJ$!~UH0|sM%puSRdSb){=7g%DGR>G=Ht?Y(8a~Aap<kb#vYGn
zM*IP2^`1P%WuMxq%b<MV!85*`j5xIv7_i}Sik6AuHd8sK(|-~kJJlby{J-es!2^OE
zN;CO~F?#Hxc$ZJFcm}ISR|{jo2sA+V>m|&qiN&on&lmrf7O;6o%}dr?7$SR(9pdU`
zH>tm{Xf%gxgbVPN$gGCQM&k>ML+#K|+|Q5uVEZZ&<ZcqwY%RoIz|Kx}WVbf=RiLu4
zS0#rdQPj7yGiR0UljF7aK^P^}l_S6RB_R7)n~v*bK3=9oE+h!)d2elJ#TQ2t2r_DU
z35@QHrv0{hGtUZh>%Rt!<OkCTD7F{Sb0h`_Kn(2b410bJ%?oL`Rlrw5!|c#DjuD}e
zRJfN!z5db)hP-exo=n<UnnPi*x{45{Igz{=rl|5|IqgX}L@mp5WV9}UtjEEQTfpYw
zY9herawixSqVpu-Bh6k}bR=~2CAmyjE+m{d)Afr?Y>9jOz`<Tk+BnPn^x`i%1S?CV
zxDgV`YZ~VdALojgDA>evarys#$Bf6uPWrr#csU=#$UWMwWk*FMO&Qw8jKXcs%~s~r
z^|2o!4@ID(rRODNa5Sq~c&g<0WRkF^jh~BpYSw>`uEr$LF2bee!d=hA>{mc{9eNwi
znCYQflS`MQnn>KE{|V>iz2NRAaGzOG)aI_xRezY|2vQAEi~%N#`gVB1VMm^HeX?Xd
zTDV51@tj5DQ6*JgT9t;Zwd_%mWU|+XdDSyT@y@s{gmJM@^Zt~a&p<JXQ$CPipV2Z7
zMZd>kFUrT7IHAo(@!qvzOgDjm5{Fc|Pd{<7GxQ2GdUa~!#u!AYX;lL$!#E?tT$3ua
zwD6Ua!UQ3blojf#*;h_Q26;kwl&!MbDm?o_I0uURU!wL((J3ke!&_a3(lL!chp`VH
z_I)dc0^It)J!R+)O`r}&RhTHMTWX~xEiw9K)9P2VKSv7kb`;=e$6^+<=Q63VY|#+P
zsI$^1AvIcFQM?PL`I67d9M7po<4aWPH~&-JiNo|0dvYSPyPL703|V=_U<gi%cq<K8
zl-puq5$^?Q5bw#!Ei%Zn)p)g$E9$=Tdv>|dT@jNmH}+m`*HiOFi~Khe^!mXn`P(6&
zi~3C4-Kpn<yg6w%a{n??Wx_mXHyfyHf}MGWFV4H;*~&#<@3-`=wLjB=HpUErCk&tA
z`7>lLJ(GD7@!N~PiGmgfwfL)->r+C_pysmNP@^{erso|jEM&{rsEWqp08-Whzy0iS
zyV~(zE<5;$M0Pb6!URCpy}Ab<V-S^-`-@(DXZ4rtE%`RC6%vx6KFH?}I5e&i@S?{1
zW4<%RI)xphZ&540p&p|}6cv)SIF1Z{l3OY3Bmo}2uwQ&1w+(g)(|)tAbiUvxzd6z_
zhh8Gh_4}@R8XF5-5qt}xx|t*PIBu7#)U0~9%$-NAWbV{Ti*>Ai-cm8AenL}&!~Q2P
z&xPN`z*;SNhJv8hPO{p7^O)dUK1QLRuBe_oSsu%P+VD;!yw>N-rD#?A$&ffVwIM@g
zmU0K%I84TMM~md54U7jD#p#8aJh>dZqok@k<%S~qFvao=NG(2V)qJtMg!=I}pRZNQ
z3P6G%YM<Xp>a}9;iXxXbT_;n#H&YsYbmN|X?@g~!guuaLj)0>EG`$h3m&7FEv5iEb
zru-AXRn(6qetE%OZS+*LJU>pZ$r`Jft$#JsX{|A@8IYF5Z=u&NNQlqirhU{bQHsE2
zB`4qL$EKns!&#sw#Z-VJ5ImCP&6%XuU$Q{Qp~tB?g3RiKxL*$!3!fb_Z(|?Hrv91r
z;k=bHexE;+1Y>E8LiT$cD>^85jwCIlI%PCxF-qT?wWM^xj3BEV#p1}+d)O*6RZX*g
zM0)Ph9KODU4wRe~db*o3m$1KEnjQO*4+(u7WjZ?{*Ta<@v{6cgtQqc-5g?j+l(c~*
zbR)++Wb_gc`}i|WB2ckvqz%YWv^O>#7`yl2&Cj<^u#7pm$igD^ABX7wT(DL2#Ce*+
zs|EhIjir=fIsYKWcE)W1Oub`XOe&Iax<CX?AAiZCkFU(+XBsfL@6d6>tx<H%rhL#D
zk~ltSQ>a>JX{S@JVeN@M5KuK!Kc9H#$Nv*msdy#iQ+OsOF%l^D+m(^|lLAH;9~mVh
zP<`iT{%t_g_GGe|hkw}9#5+3Nx%tsziVA$eFy-SCZjOr9rv5t%g^~rE)Kdm{Z#*i+
zS0*xXW3TJfPlnE%*XGRK3Gdg;+GRi=Fi?Dwc1W_rYsT9Wh;ZLWKQtF*q5}Il?7T5Q
z3zhqm>bmz%eP;2X888w*`lz&BQ-U{B8JaG_RT3U~Dra8)mV`@#GanE(Zq^q*%Lp6N
zMkTRoan%d+38`j+pN_384S0kwbmP<jH!r^^283xc>RhHs)N9kf6<lJU*CX&M1rL_H
z3>6q+O}D7nrOOepTC-=(3Ve8f&Qh-iL1pz0^ApdcSJdW&;K6hlXrl-os8uR3kSNvO
zMFlI%kCCgC;ka?pMi(woGGiW;e$#^oE+(KbbJ4UqF;9B17$PH@5BG;7@FH``;*VzR
zGD5Y96CX6njeyIspSS0*>G-uMJWEw@%9^rpq>mpNy)5PGksv0rx^tH9g}Z;{JxJ2y
zSx|zG6+h}oQOI%b-s~an@1fq{4ZL7zN~65VR=^FqmDd;{$_MvPXB)?DQ=Cr>;pBjg
zx_ej|u2{6HO4-=8YX)gakC;n7p>rkAH~kWz*QX?bq$if+0|#Z(+%S14a+7i^>0f@2
zZH~MzQ8)ddafz{~F^J2BsxuLzg3v`7Q3&Y6Ed18CeG#$wau@_3j(+EA5Im$|oX?7J
zxH;g6rtYsYTNw4xo{U4VhI{TE(Kv5;%9jfF4E9QTf%ovJ{VX5+ZB<H?sH#3(Qxl!W
zP)>kSK1swN)3WtPJeP9T{wgNV;07Zqxf3k}RYm#enytmKb4L(5n&Ve<_Ffy1WIq<W
znN*O3aG>IO8Wm_^S^@bJ)&k}}WB_C*<i}f2HhH62cz2uT`O`ce)CweHHakAx77{=X
z%+L4R;j`?+S@f%nhBwF_$baY_FW*#N>^X#Qsi>TzD3`2%MR?ptsK*vJeGaMr;80#|
zjC7sUNHEFV1W%k^TF^%t(#R%6JpCK)C&iCSAk~4+R)6}8)C;iiBVCVmHN0q8fctbt
zd&zV~DdQ9K38rY>MW_Sb#I_q=N=6rbc^{P0Zk-|Vws;kXF^u5}Nxie^AO`$Z^`i<A
zdD^fh7nfj~X$tt*bZ^*MkuY%t|M1OuoM`hsf(@Vq!`tG!YI?PzOdK)=d|gdSX@-DF
z+ixq=AimP3)6vAd4@?Y(v$Pxp*$i`xM5wpd!MK;)S&<>;mWG{I*@%=5k6aCIA0P?i
z)<LuAa@Kz5rsj`2ZQ%&>U;1lnXraekNVTv$uoPnv_s7v+Onq-3>}C@(9sUNH%@r3R
zr_K@VUZPom<R{G9=t|V^nRY-QO?$rMQ=)&1l}^m&qK;?bVB$Wd*T7YMNy(ZLY88+6
zDGJYsvW^(nU|xX7=2)ZP6(IR?iWO4R?4D*}jq;E@C?{n{j>&VJVGh{hn{|=KkMdw4
z)J)c*(Pz$ePjpu{maZzN|DHVWW2;`B7skh(D?<vBXP8n@FB{^N!HF6!JuvM9;Gk;U
zsCKz>59st(j*p<wB#mDN!4)lx@z*AA0~xg^liEY{Cw-(CW~)dus17$v`N37Gd>AsC
zzw2fXAA*0eIHi9ov&t4QN!EZL-BLg+#nDP>nw6x*Rw`tW(bHu4Ww=`-&wYF3mb9-4
zJro*;LziWCb+uW>?%Af7U<^j5Eest{AV3T+hzPx=Yct<+m*KYBvYvmv#a@sJZmO`u
zgxtn&z7=c9c<;V2aB|afD{s>^(Ujq#;(8BH?~cqJ{R2Qg)6iL$74;-01icVa$%MUM
zatM91O_&*=0Q`~YL7fXxZ>t>oDm;S2%lf<%^1wsb$X|G*LYZ0P<zBd~JLrUDTg_PA
zG+C?VcIvKU8R@T*Un@V`-`AuTMp>9t-rl01){N9PO3o0aW%nKt--7bvXG#V1gHlvH
zt{X5-$zkY1j3CLTjVv<Xm4g&BvZ+GCnM96#@KzbKLB2n}R$pSeIpJ~0i4_G0gNloR
z<a_w{cA3=?p|AJ|uk;l*?l6(TU?p)*ieg)v1uE2>4_eI9R#8crN=mGIcENTo3j&(g
znIF>7{~R~fq>{w2kE;?)+|f(8w4$&{bmg+1q5;MI;giO0cO;MTnX5FvrfL2r`l9J9
zOK3w=n46Z7h&Ns~vXeg%`K5IBJiY9o7}a`i_VRsRYMO_zsk)Y^G`lfFE;CzJa_mQB
z8ts}#s3^U(u34O-)a4Qvq}*`e@Q*+^^cOu5;V7F)$JSKexV{e9?dOF(0X3fATQm~^
zg3_j;DZiHBa)+zNMT<v9*~!8@Yy}Sn=73G<luvQQ_fv>L9xnlZegjb{?8r=)L%C&}
z?s~giZ+Y0Jl<C8lW2<24jO&uTObKHWaL~<hIRhlmnK@$<LlYzzLF3ktfQmYQ*=@~<
zk5t#`9rgC2Ly%OIfaf-o>I$T{skA0w?Mn3>6-h+c0AJAh;~!=b@4X==fGQANzBBkD
z)k1O@dOz3cSC;^i+89O_D_hNNk=1m1jSo_LyuF(nOsNmg7D7?5pMJUzo?u<D_Wdc+
zg%t^RD9>T--C2K9YVvl1N{x@h7i-+i@4fGG#Z31m>@^bQ?6wo;7tg{@gv|7CPHtTv
zC9svNyA<npO&o@<;WcwSwGl9GUP!$W5NbKN{|3)w($XpW;zq#wjcxAkRhfcENm|Zh
zh#d_Kz?u8n`P4(g>*v=t{XhuBciHfJW+Dpz?mVZkkr8{-g0>G=M1*9=hi$Z+d+4fm
z<-@mxPJS$G^U-M?O0WT*qlAeN^NsfKNMK)t^J~@fkUVG_lkt1p?7&ctldlrVlH|~U
zBPi4DRBxw1Les@wsYE((fb(bO?-l*64vwksp&D>$HC}MWff{j(-=Ft7lm&Dh1hCNC
zo7Zvxs!K?ezYDAHhRARzR!94P1Ao>+`@#DHao$J^K`dZ2T^qPVw=1~TD`9v3lpg!l
zOEQ~I-oBscmopI=tp!F4QQq9u8lx5;+Dc0lw`v$kJchp{xrHLRy5=F#YK4YzfFQii
zVniR!M+R~t58#ow$1Y99A@Icyne@^$En^6Wn75j4I;K1&WYH{=f%G(@EQ(UuUTGkV
zR4h~cg@ipct@z8ju!;x2J`DE$SNziYsC0fvoXro_>XCWctM_w_n&qidnzo4!zh!fJ
z6o|?yNse@zw21s1*{Ku!;;84iaO=GzE%w@8rf_KZ3j(kBP5sCplq4tkqNBc?smT)%
zZz2k3Kh|Z5<%u2k9|5@8H3xMRly-U^dJ2@VV@8+|E~}+<%1JWePAjF??qhJ&M|FoZ
zqRP}8oe@@W57>q$qhS#;Ia!o4fbWT3p|zrX2x!xf0jJt1DQ<c?dG-kQTe-R_yd}i`
zaC<~;TxqFNk}*@}Wm<FofpS|1f$Vd5xJJ>j_d7TNcAJQZLb*!xxszFb4tcGTVzspS
z5vh0BX<gW*sx^)Vm{!~Za6)z8-oq`!n1nDn9Oh+MRPX1~gDW>AR{}QFQe`A5@IWqH
zCDi)qb&r97^j>~Iy}e?jjA2&~Rl22`zS!>@^Kb9p#i<p0(kMRFbUFMfsd;Ki%A-_K
zDrf~2sbRRHnxmFzaGcmP!Un~Qb$wt@1nehh3gWdC^b$e!K-ROIJ-M&I6E#$-|EcPO
z^L&aj%fyw0w!F=^=J#Yr8p47MpVe;<7kpLban+BdIz`k{nBVQ_j+(ocM{I{x#xd{i
zW^4Vp{*q>g-kupC$2=hp=rRA*><Nzvcqpb|2ROy_^e3*<Y{ggJ5|H+<^EE#U9HgCj
z8XEZQMA*%t)(xLjhOG@s>_2AP4`JwjX#_-=h-|g@m)qyQN?pMlUm=e(cL<Se1@|ZR
zPl@+_L5Ad9pR6syCxAD){WPUlER5ULYb6ah6`14VwtyEyzb?rHP)@$GL_p#p-B5Wa
zpg-lWaXFI5SUpytR9D2OY8hd_lzPRwF{x?wHViJx@)%uLvesRH4E_h3UZofNo&Sru
zw~UITYqv$45F|(=!KLvK9D=(A2<{RfXpmsRU4v^w2<~pdCAhn58VTCCcH^GY@B8hu
z$2j+%Kljfa`wwGKt7@$pz2>v3)>HGDQ`MZv@YYMcb2SMh;!&$B2uYoHdIB|SLOsC5
zX6O+QX5Cb(`NsP=?x(e&+U8nUJi#Gm9LV8RG?Jzy-sm3*VSw!QbvU*@?xXP8KWoC~
zx<jdOptRplQZ)az>Y@u^6DNsg0?IzsT#($5`ns1euQfI5O5Kh<&*U|BPoK!(#<HCa
zQ{w$+9m(xKi+m(hrh5kH>j)vdb)Tzs!*lXo(&*2DK4Ny04T$=fSk8wTH3_$0*gROj
zRy@}w*KPH<9W03+fYN1ihH#r2ai5hyi)47?{oom@zT=m+SRPU5kqKGQaQSYYgHd}w
z_Zs;f(8Go_J<fStClxB4<w{$&&b@d}L7exA(s+W)t?eS_sxPN+d#zeFwpqWj8rT3y
zBV*DfbDyX}@j<}}uxe~bt=$agyZ%U0-Yh`ZFLd|N8IxMyq0i+n3cabQYwC7_Ln|7i
z(DTJN$8UPtjoApi#QejU;Q{R4@aw5dRmQ2A%UTe}&c*rOcD8EZ?PaYlZK12!ae;}K
z=dl^0`e;7`d2H^~?5m@gp|k(1nj7OdJ7suAoT=>CACipe!~DK(y$~r`<rGuh&w_hZ
z<~c=C2}8ganWd%djzkQeAOWA78Hr}$u;(z=v(Kh<YzY5%V$bln8nfxukx(z)@43w{
z1hLyl%1fJt>#u4FNup(tZwuP_e!)8ZmJX5{*RZ)jen(4K(fyg2av`GJPy7!Yh9)DI
ze$CgC7jj3w{9(lF8MANzqTo_<o0>fO*2JG$JdD!N=5cghvZyciJ%S^?EZ|?U`=r{!
zg_VtCOWGo@ZWt7mLEc=QqpmtUJUXVFMfyQYpwPL#IFZw|g`(F+EAwYVm64g#?TpFr
zu)?U6C&PwR*hxAdO`|*ALhiwNw2TLj!e4Jw=j&_PCy|xzDsEU&=xJ!X%BD~&B;7@|
zRz_jUTjFg#7#KeKN>6)wvfyX=C-uN&T15>#kdvDBBxs`Y!{?36;@4sQ&gw>d5tin3
zNJ4oLx*qzBY^`P{hTOt9McB;D1M{nIXxh|1u+Og+YP0i>Z=4u<j`}v>3wYe8WU5rN
z(4GNP?F(HcZ50pQ>2QlKhx@h}yo@<-Llre-q+$;{l$oCd8I{VXhR5;BQKAlzgwmxy
zD9Q52cqhM$ruJZCNSA%P%Pi0>9f6}iJ(Hv3O6~8|dMs%lz{SW~CU?(aUeo$9NIwNk
zkY!yg`qt*r&&-J`93wZ1m6cw{WWqC1MO}KJQkYW<WB&*()uLE~;p-S{9)nrIXJ)eB
zMrN#+@a^{_N)OdJ;%MBj0K&pG1g09d0``h+tkOY`Z7q9!chT3rqLr>MAW;ADA=z{W
z<3K@A{J#J@qjH&PyiI7gKMNMk!&&RrFn=IYPV{}3*E{5(VzWU?o|613g!0Ge#>PI9
z!x*ooj|L9XeNO&3G_#70#5`ThGC0Wx)??+i;(~Xwx3*wtV=XXq2isI(c`G1j)D03Q
z$z?yeuKs#B`oIn|t1M$2o6-M+%tjIvDq=I~;%1-H?0M~8tGw@acC=Aq;u%b^PDUuE
znt}WN77dOhk|QpSD?=2BFk?>K+c*3Ao~2C%ZA!@fphiS5W4Tw$4hPq)Z7PJz)pS8M
zad!YA8o-fyhQL2fv7Lm9X@@XO>;EDFPt$a2oR3vKW!Tct?}^n$u|%A9>lo{wY?L@+
zfmz*Lj6pbaNBZIo9|?T_A;dG--E@1SABPFXicvx6yg66*?OnGE(Z_?Di}2jW7irT#
z<qAZtHe);v9(=9j4ZL8+P53axve;rjK+b%&wxEZkgX0zyL}DW2|9XZx=H$JvX(zkL
ze*4i3MYsp3EgvRCd$NjzG5N@{E3R!4(MRBB75Ht+kV?fqD7UQq=|%8VmpI|j*Lu_C
z$=^sonqpIb9T0XP^A%MFrETgszoCB7LHGF?%67Q)O985c#`pQjqLkY)qkjPeiSu)l
z?6f=dnFD@b(~2bDa-atfNUP>E(&Az9hXL@~^|w36+iJ3~1>aO>eYz+~@?;^$H{iyc
zpFa(B+=40%D$d<>yZcnPW&Z`}Rz@QQ>mGWkZEi?+p^R}FJ)@;SxATC$Svb~Q_BML|
zlk;H_{YFl%)vX*RX7ze6-D&6K1|j5tg$C|GI(u$%z+UPyqy~sciy5}gvws(#8{0$J
zb&83*46%vJlJ}k>mvZVk0JWk6!%S#&bvO8Hj5esL1r7OkhYg&8dSZohL?jzdGouJ4
zAVuychpuF@szq%wO?lfWLp;K!ADT9U(XZ~LzeZktL`^OI;>a719z^^3CB|0$x9Sl3
z$A+&wSr^8F$>+5qF}i$f&A}R60gaf=^S4+jocj6U7*6`c2ZZ$KuIHbYHB~5+3pG_#
zX1z?Li@zDAOiHmuC}Y@Ub>DKD<`<hL4n}DMg}(F-#}N%g(jINHK1_!JKJ{!HjDJQS
zb{WGfrQS7!i>NRT_@RV~mNa76gG@S0hLd_iMKwm)a5WDv^T@ut;nuQ0b7~h_#(icq
zFP^VMe*EmMjIT7Sa}&gmrQaaIcg3}n2bIhQy~>CpqWzKAugOlrk*z8J<((hvPw9l&
zR)-H!Mrmbnx*xU*g{(VR8wB*Dr&wc?Lhvo(6@FOS4CVOku)+~GG&4t2B5x8bn?mG-
z3kPC9HWYN2LhpY^wLti{9^oIq3~(2_eqo^xEtua#k&hop*xijz(3$~dOVKOtV$pT*
zigVFsD5el5D#&G}`rSn~*PW|vDam>5*BW`<gzgrSy-*S4V^{<`7xVHmia-W`6lghi
z+T>0V&tf3_lXZ4>cWfHA(^1%1cr_W6Xp|*6Q%YasVr&lRjZH~rXNL=-NZrK<mTD%d
zL4_U?@q*xJ#1Jf}9yq(AsgyENkZBIa#v)(L5@9k5P2qR2CjSnT4DS#B=n*-#g6c6E
zd>E(*WsU(iaqb<#yeXj|9Hb)TZ-hkA-})J2V}LQ}WfL7(zSkKu=vAeG4>3;KeW11f
zh3BDx9IMywisN+6!WRju)J?*&#g-|dujcyG{qw{mG|sPhvNf7FKuK?s@#yoZ@IIt<
zU|@Y2F<UzD@=85FhZj@#`(tsTp6=Vp07C}>WH~rObx9TYRFMv;93bm6xtkBCq&3g#
zLE~cPjOdX+ncpNtU)u1MR>0g_G`3<vI!)qQREn*4T>f;4i~b+WJ`5;>jG<n$p!YwY
zowfTK*~IuVGPqK*4=qjl<m`;Xay43R7nY>aqbXFB(9Qd6NzzT2qP9zfgt=s7nXa(%
zQt=On1bgE^``YA+v**jsCA|Yw{T~nW6cns;->EGi@0`ys2-vcFHolJJpGX|Qf5#-%
za(;WkG~1iTjT20N*Qe|9s1HFKjUx+A*lzMr6ySN*Y*PuzCc%?)`=j(u{m%!@cZ4Hi
zOk579oKU7MB1yVfaS5g(d=4do-bAh>apowc(DZl7n%(Lhg3On5L+WNP3typGSZqco
zYMSGEWb(Q^6OAEjP#qs4M6LU~QUUkrTP^e{I3(^$jI(6~6G-_?^WO=-GHR;7GGo>Y
zV^X0jR|?O>ij`8t`+%pI#0bIn_u$<vpG}kJec6>c?!|N?(x6;?Fx(^vAb~DD6l5B<
zjDLSbPXC2PJe`wIATV2=3^6BT`!LQD#AVeTN|^4QMYn6yBhQgBKKe?~LN-Pf&y-qs
zx4!;+7uslM?~e<wmQo`*Gc)e$bef>v0yNZM{ukhI#vcB-!dQx5l7kgTeZ@SjDB_ug
z(_bn@LlyS6l@+bZGI`9LM*V~gy+nDe-Dvn?U7KGgYJ{U|&K0A6pNTe;;>^lS&&Wii
z&HDAgJ+GWq7{Y<<-2SAQb%OJA-~~IGH^eoajU|x^?`!%jsTLZt&_AQSjWQ`N_yN1m
zE{N?pfOIdhM)9gBm%JV{#T_2l$YEkjOh@9(!j1MSh2r;OtGwJ_RGan~<5?5RlD`qC
zk^!q6%>Q8!c~$NimKO3$If;%a-n<CCpij3DeFqye62IKT?1KneS>3mbz83sCI%6SL
zjvoTt2ufe@%U=88h3Id$j#G&kp1109MXwytR(RWnqZD2jp~zp}shhrfMe3U-&qTu$
zQ)VhDsP5V>D7DpFcR->>KQ^|_K;4s+n3xoVYNTUCAizl@9RiI*LCXAHbhJ(=fR!=T
z?)E#+0nHdN#oDJBURfR@U_}!r(&|R3W8(Re!|@Cb4cs0`iJ+JYkJA$jXtv-7(yR^4
zJPOWyMmygUV;#tv=>g*?Q(b=4`^Ju@ue)4AZ&uxXdDh7uqv>;RwjZN>b`YJ<r40SS
zIJHDl1+U%_;i~@5#otxT9#tJ1=yvUKyE{O$MlYrUMT5aHF{Ra(?=72hL3hhQ>C&)|
zm=XXUF&xO*CW0WDLeR^;C032xP2<MOg&DXHm-m?jUsnSH!#0tS0F%ejC*8vH5Alv8
z#HfL7N&P%XfX-7<GvQxNlGN$VdrCDtE^$d^s#p(G4Bt=TQ>lSCZWkw*;ed%&{Kc!1
z>hQWL+(|zXpxs%vr`ih&>1?0#(|O^$2B*?PA`n+qGmb2)&jex%+0M^3SWxcoFGeBg
zv7G{>>1&DOrPQ)+h@N0S!#y*PmD46;!;wNXtPPQcnCI;y%J@m3YB8>&7={AjJ`^@W
z1(g43#&Rk>?TnbQ_Gajd@TeWkBiZ*RSVmliIDYug>7S-YH*I%{s+-+T8wVd!q3Ze5
zc)CPeLN+qCa-~(}-)Q3h2*h4M{mC3UeYra>1*doMtJN!kmaTxDdWgu2(naLstZ6q-
zc4Ye(-J<5n@ZWbJit<YemFk=~6Y;%{Ln(2tC(R7s?%M_t?D68$6_g81RC@}?u`UGt
zPi7c82I*P*c?b?J@If~g6X|~Mq5p7&{a&VSV_GY-5d=y-B!x(j^CO%$ioC|wji=W|
zsv%V-WMR@K>+aw_G{hH4x~0-HiSj{qbUkUcDUohg3h!R1zfNafo{orHw`s&naWB?{
z_+0Z8NHQVi+ol728P1e%=IklT@g)os%6mUhYtSRdA~({gNbO!l&V;<%$DdRoOT~IC
zEoA;~`^!jP*v3%IZ7VjdxyI#G@~4f#Jaq|x<=$R@y7+=`BQJMJwDf!U)?iLP;a=pc
zE^&Fva{TwquRk~J%LZ{CpS$)ib~-qIw|m1-WF9hN_r-cm-{_@%N3&Y!@W%)evoAx$
zJPb7h=en0SN4xp{X>)xVZWyk!Ir%Y(htQJw(K^BHGKx*1Bsn4*S$~?lJG=1^dAh`u
z5=8oZjh8EkTkLYgH^wxM%Ys@IhZ-Yg+Y=+aJ9#i}e!KEZb{g#Ju+hf#1?5*Bd7*mg
zf`yC7NEo-$1k>Eh@Czm=r$3gQQba)p!WWw<$DO_%t=F~V_BG%UT{T34L<L=<r(8C#
z$Nx5}>s>Kpmoe}UTfd3Ek*lGJn(HTOXEl2UK7AVvdq!3>JxgOzRMwvdUNJF$-qzLE
zxqPoR8&g)8kc{NuM!JosllCS*6B2%@(I_}uRLc>L0abc!gp%oKel^{t4WC61`6NOY
zl3Fs{pZ<y$)r<$hcV*&JN}>zp<=yVJbRd-{$PDsH<mx4$j-^}*V8W&8TlSS4l3cn|
zbQuW)I!?eU90!D}M1;#%RfNBO_|h!~@jNKOI$ePG!00I!O}I*slN;27iiOmtC_Qru
zSWl1tu+qc_T8hZt8=pwsT~MA>?z>U?i0UXYT=h}hn~;aK)Qw%pdE$`csNw9lKHMH6
zVx&-FCR~NK@t?d=l<!QFY1S|B6MN^=!z3@IS|6Sb1f#3eV+iD{Z+C_}mAZFuv@3ce
zF#mS2(-ZRa$M?u^b}|ehZDj>&bDSJU^j08~cQJ+{&L3e=&DOKf*IkPU@>rtqJjL*T
zgBlUw{|0bOhaDnpvrC7hr;Gj#+k$eg=}V>t#JWHNRbTighMS#$F!NX`9(_<K``L=H
zDJ8Va@xj;9sdUB~9UNA066<`eh39I=kJbFg^#s^{wab<_5mqYc`)+(|VYFmjB<bB8
z%U^`%Y(EdwfyR{F(4+COk*-fVFMzM){VFfPK8q1@b(lgo`yZK2OvA6KwmxBL&eH-0
zA~3}G>aS2+Z$ZJ#cUTx%C1Oi!WM>e=bI{6Ol|7cUn(LFZ+l;AqpKrH)s)1n$BJ~sU
zV&wv)K_s}#tP&1FDEm%H^^>g&8w$k18$&pJE3<BalJ0YHSQ{JYWmcpuUx9g)b#Jnt
zMPFQ!7gX5pa5+=I5Ps_f6nqp>wd-p$Smc4)=5%Ge6!IA8j9z?eM$>Wv8TRR$9S{;j
z-WMgxo2d!hJ2AuEO#bxC^ById8h5@2_Pa82Zpa85aRBYlfX!8Sk{k$}VV&VnE6x|x
zoPMX-MX=%l&eh9ZGnMPl6pXprJk4-`U*fbxrI>Yhg}0o&|F^kh_v<E_rm9f57*o>`
z0YXtfl<RB3=8bw2LYQp3e>HNF@TBnoT5cb<L8aRf*{nhjRPoc6c?b+`+#Pt2CFv``
zjG%r+^!r;6B%?8l#X^&T+91FfdpEX^2Nx=b{<gCvRO5^BG!(aK^A~45;qFv^7;`6i
zb#yg$=Zf2@&`#^!>1tI|Ez$hj_cU{d+Gf*1M2Ylz)i3j|Mt|lBsOcAdY|lybZI0Yg
zKhFH|S_ZR!7N?6^<IROj;HNjYWAb+-9GC?Wd0gDTc6z~Su(=6Nf6-~v*$R9n^w8p+
z#FttJxCs&o6B%i5-;5-DsdKH>nmHH`5abg~<Dg57PfF$GO?yV?;zDe#TK3nI#|NWd
z@1a;x55yy<os;lGUY7MI_J3E^INC*_FQj@~@8<i3mz%84M1Pa@N&Pk|)jfYQQSi;C
zPmyM{u0oF@r{RH5{B~A+jOAl}U7R2*ZhLg@hBRUoC_qEH6ND~lk8vl@jJclF8QUo2
z%zeK^{*kSaLsP_L9=W)G!<1RB^bJM)W#;%#dD4)!(nSqm$tP90glK#PE|#KP10o$(
zd5@hi%(Lc8Tna^J)`^3Iak@;GBA30*Dwdr&GCBFT>2~U_I*6U3P#o#vF75Xs>(2vX
zN|7;hHF@KyzJTv6Z<$FC!l@;kc@{^#`mLyY-RZWD`2jpM^Bu-l<5z0wEl2Bq#I$XW
z^JqrMT>-$hn;py9OrNtcQ5_ScW<%6fC8=Wr!c2gDJa%hn6hJU@q*9}=+b1+yk-KUq
zvEYwb8_HrG7e|~309(g-t%9xvXd>!Pk(%)0W4mmpsfAg=(9mg&wjro7*aPg~XZMk_
zExjv1!C7Yo_+lWd7qhuU3=Tamv0oDlL(uBj3^ULm#1F-D?RvqUsH?uUdl7IUrezO=
zp>x<71qlVOz?FK_8I{CCYQ?}~rzzhmbf3xv_Y5xxL<iYlV`@6njZ;!X4<*5i>^ofh
zI-xS{gvuuV;-1@UBi9(n4~LdT3Hp*aZRYstu_8+O*9|Y4uGxUg|K;)b7vR<X7m%Y+
zOT3Qj3u+d+mUg`Bl8bArq1{L2;ylu~w6WkAVh7T^8RxQ)oDIK(OUmmEeM^!wg90(Q
z(J=^RK|d-OEf~Xc*f4)MJ0GRL)+qiQ^SX>V8fa4Ms3{Ps`<5|sn~QTba*S%MpCWRi
zMSxY7ycqWsh&FuGk0<rk8Y;bu_Tls$Csz~?kd*0^`7q0EA5wbBP14{(=rVspARt}I
zHSq(^d0na@pO!bvLQt5wKyEbr(<n3Twza)*LAHc&kRy^olv&LJ|Lge9HEjs8o}`J5
zULa<Nv23?6CB``CKx9ZiI^B5_537qaXF(S+X3K&xZsV`QAaT*~pE$X;_N0BQN-9D2
zP6sHJJRSBDfvw3E^z@sV44eUqhxcwQI)|SO^oz%VQH3^3`Plp)4wM*WKIoKwEYD9M
zP;}`@?KFYCiLIC!p+20%`0WOi?19l~kY-GV2VD_m&;5|0EO|>Nr=<BkQo&7eW@H|N
zlF#s1m4{xDL80i?E6m&u!u^B-uQJyk=wHW`-beHD;!>rh#}bKHCxCLfFT|SnU-j1|
zpv}~B^kj@?wi7|4Z~z7t8{}bC@G+W4W$m)bG?#<H)Wa4B<J%BG%?^;{>_Mp!-Y%0%
z$f$FERx?^Dt@S^bcJ)sR>_GPeA&cSfKdkeIrpi$}lgUxKYEy|s$|Na2y3=1MeE+VN
zCIpZC*<TpaHa)6f1r^+TjoZum!9mi+2Ju%k@*CYtDXJ}#RbS*92=A<xTF*O@e0TE2
z7f#=(-?BuGVL7_NYuJ=ic+6j2=BH}aS-n1T9U3Z888uBK$mmPHH7fk6I%eeB{oQl8
zPC)4g*|74mk@MT0^XIP=)Z<doC?hB@IWwiTO!k!ANH<hW=oj;%1oFd?)uqT%xjt2S
zQ=j1hC6`uKeq+}m^^5xnZ?ydM*G%@a_wuF%Gc5#=H`<+cGOkyPQQ(=8r0H*o35!@y
z%p~z%+XMYWXnQZ9?XHc_8OC!RF$*McM(K6-w-x`1^}*B*FAS?n*JArJm1%(OVJGbJ
znUBkNMadSzT1&7%(1zbvjU(&jhMFCZWV^GI<*D@)#kyv|0yI6vT~GGD9e=6)(0wuQ
zo!@_{Hnmh8XV4}Y+<CDZ_3hU7>>3UkNYAfY=b)&A{OQ+hBK7ult0g!=MFjpPq%xw4
zY(4Nkd}wPTqLDo|8@NlmAOJ&^>eo*(?g(^)HtS_)K~EcjKpi-ANL}fb(64$Q*BMJ^
z$X?rWzd)>5i=WRB3!#2IW`xVyz$+8a&(*cr9(yIgwxf|~;lF?)x9vqINk@L{peXc7
zFxVeS>l673LI}WxC`2-+y<`3u2lworBm*kFb6QSK%}2zHPSYF491BDtVtgJ6#`gA)
zDE$-U8OQ7W;*ZmHIR~*5VVL)6ZIw2mR!T{|+$(6oN5SuXj5zqSHM^SvN(O9ru#{|(
z<n+*bZd4vFXAyLDw2>2`)3M1;A>oyQAR8@Bq$<znDD`r^4wV<}el@<I9EUfF@$+&<
z>`wX}_%sS$vK5<Pa5aZsuBZNaXO3za7rH_>r;p){CPxNVNp(Bi5o_?3-{-0Gba}{^
zT?x3BUg%QWv;4^=yhi;%g)c%DC4L3`lyG_7^*cI}>p7`D%dAi~?AeB8uXL}^)5=4J
z>tsSv{SUZRoW!_&eP}ESi08!8nmqCRnoiu;-otdJOJ!x1Dn{TaMLo>>e&}~>q`{t|
z<^A)!0<-`IV&J;L+joP_VjQ{989AZ8E&sm&B_T${NU9xn>zlOvDc90%Vy7lc)~=#B
zKnjt`vXom_G^A}~_T}nIa{P<~yFJ87IoC2<PxfE(rry@WU^_Cl$tS4AX*&qBkeo_l
z{2cG4uHN?^XOXeM@WPL`?o8jb4vY_l7BWiga`f5hpReWuLL$!AUyHMc7aCrkVG;Oq
z{sp8StslI<pQ&ZKsmgq$m%Vr%_6{~0)C5bP=uTXcA(YxG20H_kfZDTS4TcI|&J6^+
z?&)P6ck{4s$eo+6HNg>rE@Y5MEpcS^Y$G=)7y}LbpF3IaeQ%73b18i~#%`ar7*LV7
z^H?q-eP;tInHCilD5ap?(mXQ|^rs?K&#Z5L+vfEJl8?MF75Z*k7iq>?lv%)w%yXV1
zoY%OL3g3iub_u~l;U(j*W_kNwPK-1#u9Z{suB?xWO;yyo<ei9@?yue-0~Yr)T#UYD
zzLQS`!?pAFDSOc(H#^y<sDwj8+VugTAC?8>mjTUtOFfE<()~r@x4O#yuW180JVM0t
zIp^4=L@TlAVqK4v&@eh&<jP7*l0<)V^}eC6?_f_q)}*b+8WMin`xlTm6~9Q@Dll{T
z#Ip3Nowtt&t@D;uI>b&M<xJ_#d8b*CEDqMW`cdhM7#Yf9Z}^3!ivDfOXyGqy{yIj0
zTi!&66ZBiIDHYzsLCDEpz$%%zF!~{|w7zDa2q(igmF6$NAc>na(0g;}UB$!R{;KPa
z78bo34=b{g<yF^tfBdlh^S-qIaceVR^YxK}0P<bd-DbW3Bv==JLfPq}v{d(g)`{5n
z;K!*!CuhMe{I)M!7k5JlD1pr{cje`Ab*}5fOu=<tB?j0VKzo%M^{_a3I{id1gdVW^
z*ym|1fz9FW5KFr;p(I@Wx=i8LdT8oWqF0A@u<zVlG#BUGi*e=gdGdZm_=_KD+jElu
zQB{e``Zw@8NIz506P4zJDGOd2y5=kRE?OB^rvwaD?)uW`;TQ{d_od!>3IAo4jP7+`
zUuU_0!%w+=g4O`zxyw|5RrCAjr(e@<xQ!Vtv72+^cwugSn(e$LN4q6Q^8I0^=lbS-
zcOu^@d;S7E?*4cL!Y%hn&R**`H9)7ngX2UhQo<>ue*FeVQqDK`{nUCze)PQ0?@$w#
zFUY$}Sg40njHW)(PT(5}xOfuvF2kZTlRGm&805T<-?P^u50P5^7_B^0Xhi!uhXZ8z
z+S5i@sLOMs<VdzZXRXNv7K<>th+(Uo7Cydt@6UM!J=yCKYp|1lI<gW+pZN>uq;ioB
zm!9RYur(BUYPi_X(dT0jyQ<F#vEIPuxRG952;MM+iNH@QQ!#YIHLqdEk@quBj5k#o
zkMwf>xB92CqFx>l9_wbSw5d<3ls5lIj9TPPUq&w}G$DiGK*1<0sMf!iRw6EyeZJLy
z{)BU<`pYbp#rVK#e%H%tvEn}qi7R3u5x(gkwf2@-_kE!izyfE)0rBoAWl12`NyPFn
z)%)I}0muyM(&dX0K$Ji|nTmcd$OKX8-{!dVNSG4i2lC&sWj)f%z?uTGDb5bZGDre+
z?+|`Mo?Ru0(9uppfl%a-f0qFPx@2Qgb?=?59+{PWn?8PCU(s-{$`>wltXxm3#Z?to
z1zSqULIxtFapgSL5oe>L<X}q^x+<ly$4|_@Va?Nt87LyAcDDVX>itdQqxbz;-7iv<
z4g7u#*EX46i)T|Kjq?Kl5pWA}1h2jH<Uk>FW=t)-`i-v}$b6rIW=k)i$ZoCe$8Zg9
z^ZF@R0x-?L*Ty430S>iY4@26|=qZ=J8ta6WnquPjJ+ZvGdMJ4xiU{7*cmSG2Gs=mR
z-Dxbs2f~fNs~nWc<LmpV?cac<=y&wyZdD}$*rC;BK6)2utWK1VS7>Cc@I)?D4n^zV
z_QI;~c7*A=pp*O<z3esp)6LHOL$RRBqb^bS$Tqeqtj|3LsIHv@0FrF3#D~!+)phU0
ziC%uH>Sf?9Yhbb~6DZTVQZX`ePI)oA*zY`B4v!YZT9*29>y)W5k<s<$EIx!o=<=wB
zGJ+Z=b^Z-!5mR^$Kw<W$Ol4$1Up@L-*ncWuurw*GOyP$}Z#Hw+Ol6H^_l>G!W;}DP
zn$3^0uEx=JKN%@31jGAUEfxu8IuveU$lQX^q>$1U3D{aSL^@zG|3_*oCnFq5!&foX
zvAjr%YBWh1>t-&=I^ecywbIaU@j)mjN2#V*RC7nE`c$Ln8NxhuX{)m(@|5J+l`_sa
znxKd1FNx*4RXhA=V~Ho;;R_Mtv0Z}4(qz9;t?K0y?dLdzuPD`L&#t6)zi&B7_){(V
zi4R810W%Y7VzHBq_XKxptPelR#2XJwo+zQwcw5hEzCLI4|NH1x$rDAtXIFCObJ=<B
z#~UV%0NVl8d%*~J07t57(u@D@ZCoz;&bQaMln60w&hnM!bDWvKl{-S#{&HA6m~k}7
z>Q9mUQ#GOZ{y##v*SZ(hLOEam9VEFFp@er{umJPX`q#nYJYV;QMME$E^Gf_){LcUh
z4585Y>An@5gx}R-1AZ_rLzEC4tA+PFSKKG8XUUc%fiKxK1a?%DkoRvz`~dwx&hSa9
z-1|ya{c=q#v@eWW;&(#U_0b54Pe!kTi+#W3a%VB+;ra8vsM{%R{9N)!jdH=gq<0@X
zN}mfeU9Pz9F2zP~J+AdoD_)0{F_%@ZH98HhM0)rL|8w7GAeZhtpPzI(k-;2Mye2O5
zOd6NoRn+eVZ4m?q9dYsVfs?hUJ9nX*@8G0fwO@O(Ki3fhyWNKQ_J_OcZkAd&{yXCy
z>%I%C6`#M6(lbw;sB~hxJkgAabYL?*8J{r*o!`J8W^5=cJ}>#Z+6>&rGJOjF8Tb4S
zdqUcYqTgtMsA#!PO?*05`=ls3o=Tpf1!f^$@wB;J<UAC$AgCSn9{Hj`>*33{rS*>-
z?lJ#>PnGrIEB_^?aHCo&@|{LG5=2>t8uUFLa60DmkAPJ|E$lSLcXX6csWc!@C}Tw>
zDd>m=^&g*F?S$n*UAym%j7{0)eTYiSp!EZUOr>A)2Vd|*mvOb6E=J);#9WEn?jS&Y
z+fwZ*o*#m<eetzCCy5p@SqTqD<pDFt+FQ@YH#rMpS_;oQHgYH4hzXQKB%g%4&fJBV
z?$lWyk`UngFPQy<=>LL>HUcVi|Lobtv@peXm{m_a|GAJ0!DIRh@Y&B{M#*t*WV~+I
zLij_aCL!b8RP2iVKM-my^pv<GQGcpa_za24xpYB5+xX_2Z)x&Z1w>@P5s^WDVdeSn
z7;>YL#(ye5JXIq`y=Dmb@fVP$IvV*8W_N7<9T$5sZ*3a>xYqCO2N&1ISjCLeDE<rR
zD8bkZZ|VIN;*$nFI@yoa@AVEYZg12lN%BmlJdH2!``S=sTbgpXNKf<+dh~~`zVbU<
z8L&C26%mw~YrN>sl7EeC#qfDY<TSIQut<tke#r@~nDI?io}xmJBPHSs^DeD)FbYIw
zzq)xJy<sgJ?Obe3N%e3QcHZN{z8HJsIOca!4gP?s=2#-MsRA#L77=Ko2wVM#;{eBz
z4Mc7UtG?fvXvbM4>ZJH%#me3(BFsZ^wnmC@`!W%PLy*JXd+qLo@mS?d+*EfXQ;(^T
zxvZwV0ie2(+#ao^);W$?n+^P4k8aB@SX`FLjm6*WireUl6TY$X84_bh?ef1YXa5xR
zD1~$LHeefJ!}>p*lwrtgaF~t}nK#MorXuRc7ojZ8hFp+pfo4_9AVwk-Ps8rkTX^79
zJH9f9I#~Da_jis|5*qSJx)3I7t^(;1HvL<RDcx-q)>5DZNy_=l9BQ4_qAQtK`tk1x
zrBbu{HE6V{Fdk7oaC0xi=r_J+kut_qX%(?D3MIdC`IdGkC+paXT+iP9s>L*xfy~H^
z=OzADZX1yurl>@;5!P}m?j%m!OCPpN4jdrz*pTDVJo(3fIkD%l*E9T<eDERTf#j|7
zU|cUWF!JJD=c3095r008Mg1Q#?6yZ^w}+eWtY)`h_G${h36Y%6(!krR+RJ-eDD>>T
zdV}c4hp$kjI`Z1POGua)$xeb<ystBS!}lZD2<Gl+0GTbVZX&u0_lO(JavAOd8D17r
zTWS&ayN(Tk$>gG?!JUC~SnJd~qs%ARXNV1z2LSB0+%i`wU?wydKb{{Lkkci3Z*Ny!
zVAOHjCFDcfxp;ecbhR2g)p>XOJxcPK209eSD}08M1UZ>XS}an<xLH@$DT171Zyvxv
zD3L>BmsM2xlh7x~`lWGOY`C(#F(s-$LzgN_g`<bX*x>aW2Tp;D+}%5VQKH(O&SChk
zm`7xSzkpZ$P%!XQQkknAHLKR9l8!qyJEd*b1MbOHt#kA(I6S|eNIjyQ7OARp?8dFu
z`h{cP;i?yNGz8P>u*oa*LG-6SZmpkS$6JG;{rsMKA4~F6(!HZ!4l&x37otF=FRF5|
z0rhC>e!Z6t{9lM+*)l|oJ)7hx$0F~};PZdNVKZ5Wx4&?tSOz0nZPnm;@&~O;QX%5(
z2C_mbO5HzlBJ`Auxv4Mp&hhJEgGM6eW!4Fl=da^~_qxMiV#jj!R+4+I$iWc&V^7#z
zMVW3%sd=BcQmR>#+GViXW7j(yc7yJ~s=nOMw0YPL6o~R%8bjCJCoVONbWRc<_&tcd
z@{YPe{IWQIpYCg*LiN}^eZ3Zp5o;)p;le?{<R>NdyoBr9(!Q*&r1Q9dC8{?bLDv@_
z&ErnPiT1LazvP2QUgig1(_az8d{dFzwF)w*Ml^Ur5IRIhMvsT0)LI6Y=>_h%-@~8l
zf{FA+$rEwjKgN@>^VPD+Q9Yt6KJ0?lYOFk&XW&OS>UY#U=9k|K{CmD%Jdb$!@S4R^
zfx6tjRzJu1f1fvAlrmmt!7JDjUw$Zm{(w#L#8A3k^SP_ji1g+R3Gs*tB@+|%KMbIM
z5%UGcCi4{h=IZe?d|t4vuk+{|kFWRBEz@H>2G_l!dY8#amv4+L7;N0E4T>jYB@JSe
z#Y1?kMS7tIWE)jW3!gQ?wDY%o4gLab6QDhs=MtsO!1ug#SrL)%=KT-u(4Cf%uwNVp
zKt@_xc?=DNo`zj=sD!{D*obkv1`faT%}Yc?hn_t)z;Hd^3-v-1<v?i^4@d8<>pd)o
zkgEiDC&Kw_T!Yko&~~xqFjqJNhtE<0VG<<}VcSyd^YO2qR9}A)+8BdK9HC(DO&!UD
z7x}-Napx1pobb>+?3txS&Q>zyPQX+a{3;G!7rV|bIod&|@yu)iJKn|EH;<^2@vGBo
zKQS%&?p8M&odnL}uq!HBjg3A}Hz^ynpl&&WKVTD4?Yo!Qe%&Su5E-IwZCYBC?Xzbc
zqEHe1Qrwv(#pN&5QjO6ec(DUMmw4M0kad{00K@mb+w0J$kb3NY>BS!o3a{Tpn3`}2
zEd=#HNgNaYb7J#hM%k}PM9$&MYM)(~7`0GX=k-^hD31Rf%Uzl3Yg`A8p}shhV5)fo
z>e~)Y2?yd9G9ORuY75U5>dtQ>90e5S&~V0soF$fGw6{4B&<bt!)w8<dUjTKvo<SrY
z`XNoSLWgUYIa5gjv@gd({~xO4P<0k6)7A5g#IvIoq@JplmU{1c6yk-7kY94f`ms4R
zMJ3Kt`F7Fosb+ZNlnybEH9-0Dys8|e;)`MNdg{iDF%|=1Lw2X?uQwE&f(2P34`)gj
zZL?@H9wjy|T`TyEa-L}WIxmPRz56}tOH{IrT7(3r?|dp1v#5DKxY3odU&SY2*W-2k
z(H3nJtAsQor{p~_1|5q2RzpmhWFD{FF*U;<B*cqD<u8;|TE2<$r0E@9!o4-CbOZoq
zl)AXe<#1b&*u9P(>kn2_xB$d{t-2pt<0w4+oR|>7pz9<mcS7?9=gE@DBggNftT^P^
zB<DM_Oyy%(DdH6>FtYM50RHyW%1_I>G4c+4_rw1KHZj#6n=IvD!0@}Th^%7k@;|R$
zWTN70;unnDf$C1rPV;uR3HU320b54?$sL>Pt_YX!oHHsn@$X2oJh%A&1mBA4jpEX8
zgxY_}?lfXLA8yLQBY^z7XEQB_fYi^cAD_Jhu<uhJ?+L6g5wGNb0plfz*&jM!lSRbb
z)4AuVRlxhl`+)X<UDXjl0aalm`88I?&F~~lY;`k=$6NeC?f@SLfkC8asd=BZ^<i2g
z$<#;7Ii-tX$%44`1u}L~FQ}2qlBuE7^9M?z=4Vr_ge`F-yq&C5e$92q@arB2JEwI&
z=r<lcEhm(AWfg4^U-{C@Z!<oYIc0pnm+y6l+3C`Ad6AItfE~`}U{$f_)N=lMw5s4z
z#&@FgZL_ar>K=Q_kgn-W>FAmb*@+)74oSm)Eo1VkP7H}63sNmiU25J@|1jVf7fKxM
z|7<9}Dy!4{DT?d4;CUK!C|!sj1-|%Rh(<%huv!)61ke6PoYC7nM@IG8x(s~RCW7f{
zo$D=WooXYaVf6bq4Z{>yQ@O1%4FIXBX`vW4oxUv~W}o88=J5R;<bo27olRD-GgYn-
z6!|nc4{7!{Ejgws*3}NQf)txAa!dRLU|9(v73|9Svq+(j@M67lsm6WsQfqUj<R18l
zZO)hTrZ|GIw&6hjYj3gIWC4Tbq|^^@o0mVi0v}}vp!GO+AH!eJb9&=86XCA7zBkQH
z!y2-tBX1CfML4;S2dgKD0e`Bc0^BQ_4%T`aD^hZWD2e%B<mUkb-f5cbz#BpQV+keo
zi`cWNcTbNx-r0|${x@z~T~CBeJg#4F^b)U0Ftw@(?AEH+sUV!px&hu$>c4={FW}F&
zwH9$~CFc-h*iHGb!6XEX6fOXWCi$GtM+D}%gWq!0{Co*~tqh-z_7HXoiQ;W~)w|PI
z;Y$Wkp$)XVRMl(dME@>fvvA)Xi0c!9uce{iF=H2orukn$F_-aF-Vype<DkXRTCWeR
zjQ62>h=_3y`kTt(QZ;4_+!1oKsqpp2{Po4Yci(Xd^d0ofRQy&e{pt6BxVTTb7&p7v
z_3QOermoc(-G5KXAx)w3d3DzNWBqiHaqpkIP#v`BTI&t?r|-(B0I#(}@IU>={HI^S
z$7N8>!@t81`oN0#A8I4-8bnR6)!}K*RI|Ol0h<a|Hzqt>NI46d9VJpszlpU8rS^G>
zijn8b6~n812Cp-wj}CJ(>O}UrPLjE_Ii&Bk$#66(8BaJ-A%0i77nW;&U|zYTijCRt
znWFXp!U)9A_jcj8ClkUjb;5TqISGJv2hKJC5BAG7!1z5;<Cds6Tsecw;G(gddLVs{
z8}+%(G3Q#Gc-Z*O&>OK_#IyJ>;2qbdG$hg#9Ll9<q(2Ynf@P}VJhDmI5X~Oqf&!FN
zA8{VKF6b!q#E}y_vmhu`l;t>UPSUrA!B24a`&+JCDh@Z>+@ot9lmW?A*&qB624~Dv
zzZq|@u$Edxo2%Va8QsKZHz&2u3kBRwH)o9HfL3<*<HHV^n9%|FcYCuBm3r1fhcv~r
z?9|m-wG_aZ9RsQJY5B9@K8&$sK4dc&DJ;nbda~8XUTFhZM2)9x8`d6<s~r03a@2-b
z5qJ+>fAjH^9*5k$Mn<@PErs}Yx=24qq9;NcmPI$Xl9|V&QYY~*V1ko~ad$I`OI&i@
zFt^+g*wEfDw7?YbdB0X+2iO9(^M-_K2dVv<y8Wp7Oz@^7(f!O!MBh`OI%S>(0Y(JI
z5MX?r)4D7$@1}WS{V#lk{)LZ52F?=)oL3bAJ)FA#hGRe*NluRRZz>D%e<Ac2&|3&O
zZDQOz^XOur`hTJ7U*vKB2YIR!|JRN_-;TFDEKd0oAohdqT;hD;KgCAiSuQuVIaIeA
z8BdG-<jedq;^(EelPwRBsb|7B0q(Wx9mS*q_tM8XWW>p%$!X}$TE*fX$OL9KrZ{0@
zp8N|aTAkQT>KUsF4LQ$F@dizQ5RAC8R!2mg3+(xmNUgqJ)P305;>^Wx&Yy+S_|CH{
z&0hlN>+T~Fz@aYtkt0f^-GCdydX;k8#qY9$d5tt`V}zq!FRiR7v}lUMG>mx0-{L5F
zHN1wW#^9o|)KWQPCeTL5ffL<se(twpdfFhK4-1n?`sYU{$1Q3CZuc(tUTB^(BW)*T
z`7D6tzRTNV&~K_rt_ztgNEifZ)Bmhwd4Y(3eV*ett=PK6YKfijXZ^R}uXA1120IIv
z(V`Fcpe3T%n@*Ff^F8}5P?p5U<c`~H!Teoo<S(8Sn^K6>Lf|bba>QKp(MguNmCgTf
zZoi9brNI%!&s=KcMzS;^;yqp5dk^D!_sWRSSDVHc(onDpNu~6o<3;o&ddm3z1b_m4
zItcKDEsxbu-{|MR58R;0Lh7;~(y0NDp$fb<w8h(B<Q$=U78t_CbPU7Wi#k2JGmiJR
zzit0a_EF!IJkGn85M|@?;+}t{+6B*?+!KBb%r$#!#`Fse62+3i`YcP!v%M7K<=CGs
zV*8ER-eJ!k&F_~XT3k-7(G*DEwVOhL`C<~%jOuqYd*`x7>QI`0f9-(XEt)M%x}ML3
zcGYa#RWG?Fnm_q)mK{-7OQb8?jWD=}uas0B%VaIa9V{iM%)6f5JDa7N8a@bK&83eX
zJ#qs<qFsszCHyP>W_7_UpYWyI)V#0Nbk&`T`_kT<DSFtC-lBG@)!6T5+JHxTq`Ung
za}jy4k6GlW6u-c+dbveRx}ssZefXWhK_=@tF{v$shUcf{;~&-damj54q+d{>#K^Ir
zTlM=rhp`80_Y?=OOz7tDvB4C#k$R$d<rkz0WogdV`v^;{9Yqqzf3RNRtmjw|??DMq
zTdx_IyO;k0G8m}oazCfN0m8^*f2}<nuhNnr1)FA3affmVYHDpa11MdQ`6a#%Uf1l>
zQspDnpog8hkJOQ6qUhsFnsuAqqRsE?y?pTRd=g+mjth*i3$2@lUhAJ>Z*Kms<s2Jw
z#p_Ezynm<W|4MXp`Gt+_xOcr2J5(JPOS+nY_k#!fsNz5w&)AHzb8z)hz}^Mv5L9ck
znd52^Kh^(GyBozv@u2qPcm;S$(L38qw*O*0DkMGC7d|7qGnjg_RbolC9PSzSaEAMb
zJowK-Sa9yu*$nwu=RvHqW-Vr*apj56Q%B{OLZ0V@E09fbKl`UMx2In?sxC~0;yd9>
z3P(FP>NR_SCzs`I3eP`M4~<PfO%W}i5u?YW>uM~mW9i>LgAh=gt?denKI9lFhKCZZ
zP8Y2wr^xmm5Bzd5yjyq;;_d<UA9HSQ3Vtkz3nC|b!10Hb+clFKu({!m9XB7@B9ts~
z%>yM&<ANvC5>yNc)!6&UL%%^QPx;nUCMjRKS57W(FR^84*OAl;H18Uw6On|Rbpgj%
z&FlE?n#Q?1f+g2|0v-OL;F>lgiy~1r5xP$)Ieq^6iTi9oI`{M&5Y`m8$YHA|^?MTP
z`)ikWMYuAY-}3ZZEc9D9aL98<E-%+<iqdhs0<$X6n*3sl1s&y>u34IMm&WdRF(QV<
z$rPJp?#&>$2*_e$n&GjoEm^Z?7h)&!>~_1t_ELJD7>lS%poaBb&r!7XNB!EN)Pu1r
zM4@BfSb#-!DLSgh<ZsMRp|z+D1!|h>I8|@&951V^#tut4!a)=#_bh=CfnPB#h9v`c
zuSNpTQtGlge4kVOwCr0cIQ{~%vf+rVIeKWNJGeSmOaW}<z_cdDi~{SZ!vvrvHyI9z
z+g~(W^59!<mfNn0*$1JcQ>jp%KTNI^WHojg_IvF#%TfP98fe%?-sYU5<$~pIzCUT3
zgCrT42p*E|_{IOyIr79qvOYbP%(Ef^k|jg8oO84ty;g24vf-O{^=l9*%zBIBR!ubD
z2|^hB9^w19nt)S_70&+iy!j^ZXre-R3rJ|WlWJv#EEZBy0C#v?2&5d{cyU$5cZt$%
zU-^@6AJ3k%3~8NqC<SSAqu|czOoq>S-Q3pQso8B&j@>&N5Uk(@8$u2T6No6=R?I<?
znY)jYr;%A;6yJU~ok+-TwizrH?@wE$6x{U5MT|}1IDDzw1JYD;;xz3YQ|fCfYz6{0
zGt^fP66_W;x=_`v_=S*)A1kj7vNuz<h`N#{P`GlZ+MO?e5BD&gC6^P#Lt+Kp`N6@o
zucZPTO>%$4XlRvnV}IY_2^UVuy@n_gZ2TADqq;iPNp3LZ6tRVM2oA7=om$kOpWC|c
zgEx)^^YhYX!d^siGt-ibqHynnw7q>m-L{CRvfdY*JGnYt!<BG8-q}F7ciE!zJxc4)
zPdj2DH7Iu;&?h<V!{Eyd3~+Y7f76#d4e%a{N87J-qPSU)yN)48nikg}tgc>3G)>P+
zVY{ZWw|AoG99=$d^4{@d9aZDp7lh>VixtU^WpX1#3z|``o8x|)+ucO)6lh`m1%x1N
zs!eFH#Ir6hNG_3oT(L4C-{$XpSP%SS>(CU1zd=cbMx`MXXG%!vwi00%ZncPT$lfLT
zeO>dJe4tT*vx*l0X&SkmX087Vh%CQ@pcOe*DJPT<vp1jSlDeJaVj4%c9;6!eE=Dv-
zi=pf<;wWItVt&%$LX&ofH?^Z~k^VuW!6W~Q)#px?=!iU;dMOjTmo-HVK@dB7N{yB*
zvI4nl41_Zg<vu?^4p*=T5kjPmWxU33H;r`=$34>;!xz7?er*q7p{g3ba-i^n*)Z=B
zmtS%eRKo8Rj8IMQhTZRM?>e>P;~w<aP#OV%{y$tcxVKGS=MJy%c=G6lV}@||I+cj4
zb-RE<Xj`D)`}J{+FvTkmWITlJozu=A&@P?Jp=x;;U96l|Gcs<%U?y*0xgnaTyTRfB
ze_#bDl%tJ(<EoZFjG^!F3RYE$<1of9+4Ykoqc;_TBMTW(f#^E=q-IQS?M)=wRy2!_
zD{msci70~EH*UB<lqxw}Czt2{`A;rf^EZ9hKWA6LjzJ3fFw_>~6d<X0YQ|I`V@X?F
z6yNF*d?HfgEeZT<SAJ<%xA0;h#e^q9-+&_mc6zfKZRESemkr&=ad{1VR5Qk1KfH#U
z{SftqjCpf7m?pK@=B}sukYss3Q6?iriJ&#So4g+D%8;9FFR%8yW5tmLoj3z<RAcWN
zWsKQOHtnup6dty}SvAG(;SUwUL|XumQ2jlbc;sx3MyAi$l9lrtZ>z5GcG2UP6ofRt
zO#;Ai@Tu}~r)pDKF-j3+Yhg@kF}~5?$2)!Sz4?W=yp+#L?pQBIj}@Ike*u)^DsI(Q
zzaYto%RITX<3XLtkU$^FD1Nb$RYNMudyA#jkqBeD1EY#J(=Mk7#w3|`R^!=#@<o&G
zF4BovMo)I-S}LeF<H`fACF%^@XFCP-sRvO(j_nK1&i>D=F?C>?$*J%Er$K-W^5pPI
z13HHI{WZNQRm0AW|DB3`xeKp{(#|fqK(n^#L*GZCvMwFN;6ME@DPCPbx|DE1l)sMw
zH(YA{>i0SG)E({i%SfkYl~ZMUgN%SngM~{jV&Vnxu4n1<dGKF=Xp$M*;I!%L7qE%;
zj6?*$xciD%#>&W7dRG7z21k2uzuNB8u#4s2Y8j1u7Gd*(6ybM2FD*v4w~&JBMtK{e
zZ<XDJsLQ_?6GapALjCxqXWcu$_ZHAntXuP0#PeS6ClAH4E{)FShTt9Tu7Knt4!ihu
zs^l|WrO%a2IV$b{ce{iCAE|}JnW8LHb{o#s;Va5>)Lk4&t_Ho7#{B;JtM!C55eS+-
zYQHX<+Z;nnzybAcz(%CC_dAZO@aY7~fH1w&Uy~JA$kdJ9E+EplCh}mKXA)+`;6vLr
z%W@^EH6;W^<}cv0XWZz?VUk+1f|xqCR=IM~*4vo8M8>?gV1*G{P^Rn>I>Y-@Qe{OE
zOAF#J>x^{tc)K}BNY1z?$Qlw>)S}3HBUVO)f;Es4#@MK6rBQjtiP7yiaDU$hZCU~8
z2h?yWUDnrO0WXJTVzy_%J+sN_*%u$w+b2;9#*yQnU37u6=Ac(YY|fhyHg7ggP@+ig
z%UsIW-%2*r)gSd=3vz^aa=bh$0)rlLxt?)vNwBhA+m#;I$5*rbmej#JLkKBN*Z@(o
zykR`&ZTUMzg3JwelX0bdDed>F6h!#vbp@ill<b#Gzq+5X8d6}NJwv7WwwcsOBf!Eh
zQDBYNTcl49<!Nf0(B8(6^x9wOcxje7YxPsO^@@uM_XxhcyErc!V0xJs94Ai0<1E9$
zi#uJV?=~F2{sSz^@`7M2fx;W>L*PR`7reAc6=|^LXuBNYxLXIx`@=tHU_67^TLsSg
zOT<_C!j6ZV<KxTqc@ENg+K<b=e00r3lexWcnY&1!9vvN%bVb>($Ae0TV*gq51hDt&
z3DNM}ImcH?T=7}!<<j#dhl}XBDwr}`YnU?<aG*L!49c>43AR}0BG-b9cs{7J3Ugrp
z#Onb>1qWM=wQA3$J$UT@nd+m~qe}|Ic=4lrg1(erezH}+oLQCW)nrVfEJuu*v&8*}
zR%c7=Y&Uo?xuw=1L4(t|y4}Vd|FuM0lOAlZB<jl$%8wbPG`a*aUJ<P~BN(%_)RLO1
z<k>{T@{F_-it>|+0zUk_+!#pG7x-Ze7n~HFk?fZry$(K{0{YUvUF`L5b9b9bmK6zm
zC@Et3-gDG5Uvj^amp{KfpBuP1pJx0VLl5mu|DKtc9&+sUrA59JM;z5s-wg9qHVA}y
z)*Jn5&p7O2vVA{Rv%urtY4h8iJPN+L4<qgjQH^57XwhDA)GpTS-hjf3I~RExx24(r
z3kz-+M;cZ6=+lt-!VhsuiqVsIdJgu;DAG~kHafYreJY*I**uf2wka4pyR9Wbk4A5}
zazf~VoEGwhW%+{cPx_hP^c*UZ1X7q3XYXs(nCi>T83>#Uw|~Ck{x<)Xh4E7?9jf6s
z(jQX8^}9e~5nF!r2qNl(prK+1L9W@_A>g4O#BPH!L5?S$CPghlR)m{As)(KOn>`Kg
zbdTT*GX)J(Nk-~pr}T_BL42P-v5-u@eEVIkn6Yp}oN}>VjUv)EZzr2=J%fRF2n-zk
zo<yq$+ujjUC{xZXeXVZNtAR%dA|jzE*^*l(!HtV%ur!-Y>W{6tNpSXl`M;QZ>!!Be
zFYGtCODIlocXxMpEpA1E6n87`F2Sw1JH;(H1qu|};%>z$&iCZ^oHOS&Jbz_oXOhXD
z+-u)^uXSD52c;|`<s9J;%|d|*o2UKoP@ShMc4I-{Gr>v#qZYT^+sAP>OiiVoUnEi|
z@eSM`u+hixvOP#!0}AG{nP#&XRxbD8Cx1(eczr}S6FqDiqER3`7xO0s(SBsosl?vT
zlUboFpu}|ggS=aZ?50WNd<V`OjBqz=3YIAhE6-L`)Ty$pt5V&~SEes3$dT2a%aZng
zbWf~Mrhs@&TMl4c?ZP-tm~}S`<7C^fTE7DR1at=2jS5UOlGJzMt$~@m`9q2DP4S#!
zf5{};>z-4HK+rM>8cAzIXzdM}C=%Pg@-UI+sGbb#xQuP5k0aP;eZbf&Cz{*PQVK#v
zQ}Re5AbAx!U;G45L+Q0Unt&cJDM*5Zh0TFs$E=CWkAFDRk&N0c03&+K42k^yW3ZJq
zveWGn;-)7GoaF(QlkQt!Vq2!=2BaXb%96#nJv?rYA(umMs$D;?E`Gx7U&RGp9IQjt
zgV%6Hu1^Ot%7l^69!@3Ze7WtYSitWP9N1a(LB6?bSKEpA*djDJaGPSGMDx?)z?X!$
zVfEQT6%VMlnp^BhyaU&I_RK@3iECjSj++q$2I^K*e{2cwR9kNhrpRDdBOCPT7OMNl
zD80_4KA0N&DI_wiV)k~R)cNnEqG;I=whflD{e*!p2)^j;Wd3|6Tw)S|diuAC=}mK=
zKyy+%7V`FjV-?N!J>3{%BB(w<;QRs(Hcsq6ry{W>oPKZ-3=r2S?BJGnsqe<lO9A2z
z8c4hST!)M{Y2c4UEXtUiRaLgX;Unxp*E?352Cw-6n1sG@K48z|S=P~E!)Z_K3)6|%
zn1=Yz#cC<ZHX^=;ne)D1&Q{IWA^Fg&`@5i!dOjT>ig(+a7V>`p!qZ+{xzHR6_Ot5I
zG!~zU(RHy;PB#OVP~YStyDaAzP;s?(*C6}oVAvZ)-#I*zV05m#^VN@fVHgbxf*t(P
z7Gd7)V521BgP)Z`z{26m{V6Ub+)j?f7Cg~F?+{{jtosDx*h+5@7R$8s)w<~0KwtQ9
zx5u(GR8BJ|*(@aS6j9bBY9e^#DfYk}cRtp_d!yM%!2ttk>fQ*kWtlO1*eV;R5jK)}
z{?yAZXyj6Kg@L;N3%pIz9qu@>cnk)|bC8(Cu-%c5h{<YUZy<!qytB;!3NX^2u=OQA
zGW*HNtWwOzuzv0~Y17_vBeRT{SFrS$$nd+e{A27}gxei^v*Gk7s-7e4Y~%rRd~|le
zy3u0bT7P=zP6J`cnUHK<&||}e10-J^xx|~`w(^>(@4XXz<+E4V)U)d4QW@Y_?G((j
z#6?$tG4MG-&;G2o-Fd5sLi)YbpD(=WjV7z_qV@X^VGz*=>XQbanq{v>as`HlEi#5u
z>LknI?Be2Bh{0UOoI)-;ck0b8-a^7girD$y1>uzH6N;jCtR)rs?)C>Q)5J;3;c?r;
zS%X;IPFG57ZiHlF2al^e<h%mCo8PXglCd_%X6#y<@t?31Mb%9u8LN^JWvABA9_1y;
z6B)5LQgCy3t1yv$oMot&mAmLlBdcc+*6gu~v1xV4ZoSFjXX3dI@2^btlUR1!M#o9P
z?3U6h3`Z+d%3t~5YxKS1($WKH<Bj7pl_7DFY0}ev<Dv`G-;b<SFu(Pcj1>Fm@M%M`
zJ^0#PWHC}{DnPyAp49>IN0sf|$GaT}QGRbk_OqJk)9^~_SQs-P?!s*LRKIZ_1NEsT
z=9yj=3kF_pEC1!YneSt0{z{_mdl*fKL=?Z8ZMkQW*;BFo`u#A65cIw6t%@rQYT53X
z{j?A5Y-F;tG9yUAc6QcmLR|C*73(2vV`F0u;rl{dR5ISjX=QT$6IMcd^JD(2rv8T|
zwn``h<@On(NlYZyx*$pOo7J+&o5VCNeWs9HQ6}wlO@B=uA!?p<L6&?}ZOShMCm&(i
z1$Y@08f-$T#l+m3D3l(wDOel0So%eYo`6*r!;GKtS`>WzS71t{hm9a_KQFAkKnH<I
zfX6Hc=+bhQ%J&BZ>^_5_b$9inh{Y$OuxWa)G2&h8#6`*#YjA0G<I~Xmht(9Swe3yx
zq5JlOV%9BRDl*itE3z%7b^oF@Ny5`=XB14Kf2Te%Y>%$Eoid=$LANvJhk^mNDR<*}
zXXW_a?^QW-WHEjhuOGPpT&SwH>c)jv+6O2Jz5U;uMmc6WjJA%JQp6?#G`3D7hLDQ}
zNm8K`E8mk;{0$k#gDQQ1Sh7v#hNqq1TGY=I_*uyRBy*|&Yh5k|0>B{tL8GwEXSYr5
zQD0;0(BJ3wF3r2-uWtxW_sPclP~V7c;=R?CB(TMlz>w#FtkwM%<P;QR_KUD9CW^nZ
z&kLj(U)X3B)UX;NBdPFm46S;$f<FG+u3T_(+i2_p0y}M*$Ui$b46++CijCLLjS2Dx
z4qF33ktm|DKtT54?*CKc2?~w)qcRRp&l3DXS`~}mcU*}k{ZNm5BQ)TCB4qWrE#-h-
zHT7pdm<OO^ZXwfVc)C@&J`4n$U+q#m{So!P2IbR9Xjb`q-kG5EJSLuv5=1@3aQ9!(
z5H5weZNtM0Y^|-AVVToeZZ+{6pp@ZI5Dk6?tP`GY5&_$ZACELmh4<r29xrhhcz*uc
zhYLnwlmhTBcmd%SOQ^KhfNIjq-;S55@)#vKd4kEHtEGjYU0R*hzBCTxvo%ds6j9`^
z&cL@|WJXvM#@8Gx-Eg(hLjN2=9~mM2f4P)|m~b`8w%-`V`cw3^-9!Ul-Omm(aPa`1
zL&)GkPE4{DxCQKBOQ=y^x*}ZFASJy3HhJ&MT4@-J)gn5B^GRT<Kki%h^(9n3{9y|k
z*Q*dpHFpVB{ot9B5rA(^On$ayT#o+~6f|%Q#l*_*=Q_W;E-UFFix7Kw7hpl7gsSn(
zj6}frS-2a2ih|Eg5J9iZ8u9<z>*BTom~gwfr2xa;D1>{-M7VK_Lndpw0oAW=+1?*<
zi16-RS%CW(V@DD`&jIwH{2VPA!=6%aiSxEfx4jHGhFG^d!%|I|G#?HcDsF11ss48e
zZ)ap;5aAAd;}a+ud2q~Y(r?kPPW3usY0vOlIG&A}A>$3m!;gW6{R6dw>HMmk_=m$@
zH-8Nofm=2{_m_a7<@2d0Jjmx6ik^@{b;O$FS)LufEMT5*$p%+UR3Su6j%#nnfaLT5
zHBx_XY@1r>li9+$X=?wkin=9~6LKcO2G(bC(kf=vNu3!47OnCW><o&ZTG1yeE&M8n
z1eMy$$UIytH_-=lpBOwpp!*hP-s+MV*|{B3$reKZ+i$$1f0D@UZ!%O8W!s)~AVQ|m
z#$RbfS%O|uA2YY3iAJ!;L|OS$i<zRI)W*hg%TsbyCK4FVpcy(DI&Ir`mLx(qi`%o2
zLaEk5z=@1U-wm0pTS|@JeRs>sg_+-hqk7W?nTl!(tdAae$!M7cI!s@`upW)LV{U@1
zQiGaiBfgIf|B9&)TYaRRj^r#@LBZo@KWv}u=;0}?I?&I&x6QPmx8GHh&d@1amPb!V
z>uyck%F2&oag`SoGoxUupA|efG>zg0J_6T6@U=W>%l4$T`!Gh#%T3Q#6DplwO`-`_
z2b==B;i*V)y4hVrHU7>bL{J>3oQ}pVQ4jeawmRb^_*|Z!ZIg5;ihy382g5vCTM#-z
ziZ%E@n)ZKl{ntWyD~SGb`@(kdd39}!KDH^QY!a$JJWOHU$p*=1^g&R0n6j|iq)2;c
zQAl{|gQ1_D{GN#`k3il8x3%1YLQlaj$BCLb6H~;+oGCdQ#Q6n51E*}{&zpN^*2L<p
z(Gh=I^x1U13dJTC0+?W+rVF!K<La@!xDtDh-z(j2x|LGU%0vHu(})~|*N=b8aRFSI
z&6iyQEZslqLU~st9|~}u?&^TSZ*sClHo;Qt+1FcrpI0D!F>q5vIxSxk!cCbfPo6%~
zi)tIzE39Vmy-AL8kft$wimW{y|HjD_@8|mZBk^E$%JG0xI$iiWyHOvCk9sogng!s>
z%M=tUrqU=J@xVzoCthg2YT(i7XoSRL8yDd$aB{m#YH`QJl*L==y6^nB)pTP-Z?BK5
z{wq--Bs0uRZX&C)!h58C6px0<;_a(s4Hwexrm`QV_Os1F1wX~Z+U%j6pz8IX3Zjx#
zd@<VyDAAHzuLEu?x>#kqo?FJZh`7KJmqOu#Id(v|S%j9MspFk*cP`&V*+9YW5~^nl
zs&4J9<BuLg(x{aJ2}*KG_3RlIcUHNt`gOK6*(qv$cKK=Cb$ao^<Ulk+qaWTMRo;<O
z0ME!_Lg|m9)<>kgsOI4<xGdH}xHZ(V{LZKeY2~Dr-k95RR`-5l2^TUc>=c;mUuT(g
zc`|gaK`ETPoFofz$YQ7q&|`NPitILJBzA9J<Tgd*wh5!`49@qm>r(D=Lap`=%p4@~
zne}>7a(Xosu<B{15^*d#<eO?bjL6B!IXWc`(=>4`pzr$o9j~9A)-LnQ{eJ<z|95tM
zI0XZY!W0oY0#>J8^TeNjS0|C3ICpG}A+r0-x_tYvy|VLEdcCvTql_b9v<E|JRw)pR
z-p4IC9m8o~5N#5;2WI>JJ8*JC!?=)AeC)P8TFMq?7JTCyE~UJR+8-CWN;7bb7H*+V
z2FERKD2RM)h8i$`qGfjHXwla7>!?Cx$|FqS^Qii!auV#OQAk%yE%I*Xc2P=*w?u3E
ze7EBwE}_f!q(@xh9Jm&WN_>Mf15%z*n==#)kWr%Q49gf34!CHuHk?oFEaexaEY@Y#
zzFO0lC>m{T!s(~x$+V2T0G)qJH)Cate0FClmW6aX*!gk;R#?n+fx^5DC{;S&)jlXj
z8-pDz(2ad0K8~q9CKu=4vTZfOIUw{nM)>!KblcDQlzpB=S)rDMJ1WIk=)s{y?S?<@
zJIb8!Y>b%47!(q#D-pnm8X@j?L28x89nOFrlpiv!jD?pB<)!s0=uycw;+Rn;a01-^
zMv%g>-t$C37U-m>fHr6>ay4}hrdmpxdEzEDWo1b=Ky9ihkIZGXsj0T#2F1*@>eG|x
z=7n>W85tR#RbY}I-KP<wy19*!p#k7+dE7Y&-ZFg>v({_&;FFWBGMnr++T1<Tl<+EU
z?`3QO{Q@S%cP~`KSutI`zg&&+=iXM~iObJJ`4NN~<pzA&v?m@Nr!&}1b$vUxc_rZP
zf~*uFORw&Halx#BpdWg*J~DFOLkuK*eM3H3cx7+WwFwb00<{&~v*9d>0HDzu>bfod
z=PC#adGbq`09GS)pXa}WrY_uppe^&?1tg}o^Cv0t+XHNlU>p*1AwS0v1!mL=Lm~6>
zM?@p*8;Ev=vnIPL%R3{3Q$}X>xRrB8A<<F?54%vsrrvT5gC~Q{l^U&Q+~We0kY8D<
z;9BIx00+jk%R;a}KR*WjkQ>-~1DfC(p;0-2;d0k)b(mmzhKSt0hY@SN&itd#*@hP_
zQ<U(-orIlFAx-MCbxF<FMSXS$nHai)tSY`kkqR0uZB0YFDZU?GhKNVFwm;Kw#4;hn
z%@GR3!Z^8e#nKfBw&6@=n{?lz$+sPpX&J4psW*w}l78zuUt|^#Skab`K(=!}7}1vM
zU1AGmltX4lv6#PJF6t=Tf}3!aYYS@FWjU+WH415cNaH(Inwal$>wndses@&)miQ!G
zaQE#c+^6Qz8s~c9RKvRT6_w3{A`dZj?3?pUi|-~u$pkV13N9Ht&hP;2Hh;-?$<8s3
zR4TpE!>iJ_y!1|mr&?9fXo)7mdmOuRMh}eEW8ls!sDaOtGlPcl6>WLi_S^7{hW;(F
zqw<ZGh%m{k%uDJ@!iSW5Ib;+m5eaIYo!l#a(o47$*>YP$<Evj0Dmk$l$sF11a62su
zg{lk=FmtN{mE+2XN^g1-En8-rc6vZdnLgJ9)?qw;`>*f=esuzkT;}cXJ#u#+YxWTK
zh}3LGABSAsF-j;Qq`c@LH0pU^LO<B8uD;t%^O{(izAc^tveN7f;rP9MO&JkTVO}SJ
z#f`DBr*wrAk0e1vVOGyH?Wz(3I~5bIL3X1L&;xWwetqmTbt(cD8Ib`UX%RedkB?Ee
z;+$%>B8`@s5OONUBA2r?qX9!Xrb&d*?rtV=ly9PXZsTR7kp!E=5qkDXsNyF;X{4ks
zP@kJlUx_)*oGN<*v_Ur&)#02fS4DT)+Aogzf~!T1Tf6F}I;>I7nVut0_w^H9vRakp
zlyarAExl<3lL<5Y2jX9oWKgKmShsye(3VcSGWWYZhZH#bbuC#cdx*Q7V|#}YQ=F2R
z>lh+sV}tvtcB(wSA?#U{KaNeYu#x3sBNYKN_)L@ODX4=I2Bxc$o7o@HG4LP|`ik%J
zL337DOJ@wxQ6^4#&GR}F{e#kld+g$zx`LOFbHU?6>o6lS@w*vUjdS%uSm}@!N9HoV
zu_{jY%rUr**yXs=_|BaUvqdPegpzss<W;cTkiZJX)Ul?9yPvBrGcQFyuR=$C1s64e
zVZPYVkhv=xZOJ-=7zIeL&zQFJjiw|q4<K06vk$mi=s@OFq)Dm7C;J1PIIuQu=a~Kd
zC0QoQ(_1$R=?7(PXI=olU~z6`+_5&&wwD5HD}+~X#N)%LQ{S=ZlaYudA2|^4LjA3H
zA7;-oUPUjGh#z5*{!X=Fw2y#8ZfMcc5y8f{KoM@to>Z%?qJ(EOBAdF+iwA|2NAP%k
zvN)>aSE^o?p;<HzHAu#sB~oFs_hhX=9HU$Ob$+81%{CkgZKa7pRVw5N`vB6|HzcaT
zG8m~+SgBS?mPbY`mI=ufB-Qrqsj)mmtJ|staH_Ch%}Zw~t3)kk_SBuxjfAIDV||!0
z9ynyCS7xY<FDUbh=k^G%5L56JFd=ty;S{P0k+J+_O^PX4;9bx4AHWMObT4+-ouA!w
z<3wTDVS;eGK`?et$?7wlf)H%PcdL1&81rnOAd-?e<_=tx|5URaMJ<QVMZe7JT1t<r
z<nVPqO@(pc>K|I{A(fOSP#NQfLzorbsO=G6L&I|RMz}g%ikRq~Rwt0M?2v{Y%H-io
zAe3QIcA-Z-)6~mJDPKPZ9*8cYPN+ZE4>n|X+1z*+<n~C~lB_g4Y=lM`-eP&qbiRB$
zhIX5Rm`~LFtZTO~X11X5<Gf9s@d~4E3x3aSpfQ)NUbeLoxACuzXVTK<xjz$#qoRKw
zLl=l}qFG>%NCcEvC~Xr-e<AKdyAEDK{fpZ3t*5W*UoDFIbo!{{Nc#}JDRa7f&Sn0L
zk<>8AK|~HuZUXSW2Yb3Dgv(#!0{UAwX#@edm~FU3-GEO{2jzA|R5;zVMdUaU+V~=3
zeAv=XrtEH8CHn}s^<vNsm~YScc(ec7ZWXxKhY)@7_7k_G74Vx44?_Y(>=^4?E&xL|
z87%Z?jq!q-8UqeVj%jDQXhze?Hv3M?C1{HU{qDU3*nODF_9p5CgEvvKo#lnEDX(2l
ztr>z8UoAX=8bPpN@zLJufRPPb?P4JIs-HIwneV?eCDiEu07~wFrFwCb8+hj5xRb7;
z-Iq(MC#e8?NOTS%3${gLkt`<nk67CIY)b*$zk5AT$Oh|}Q++v74p{D12^1X&zb1Xq
zbGnftWf#c*t(xtWaE0@E@HU8gtI3|SCUJ>(Mppk`HQ$tW;owABpb?lg?sO%D3-%~s
z=pG;zrcKG!?l_?s?vDMV1(-y+qt0Q7k4>WFO|eUvsToDM#P*MkYHPMnO}cG%5CD>$
zSpob10MYT#0(jwaIgKgJRF2!5P_GU}>;xxaLFQOe0Ga2lf(;S@f>}vP3QI#m<|5*U
zgNbOC1q3i3*D?p(_>L8*QDBpXHiWjVgdZCt0NpAH?6rixU-)2tN8)E61=fwL`14pQ
zQoh!jR_18Poo05x@B?@(B2XnM6m))qzkC`6B@KJ+^e_hP7c3{HW{hTs7@O@b7Q~Pr
zXDpVEZj^h;1Gp}UOMuq0eA;knv!wkNDl9CZ{Yxzt96z{ZIL{6Z&hq%_RzuAdh`W}}
z?T6)%7<whJa9iv?VgskrK+t~La^!PtUdu8bI=u!Vf$8v)J~t>_L0@h#UjtuFSss!s
zCCjo0fu}{j4Z9Yi_2UWv{IUkrwSf21$7v1rHRzHM(h-DNOXO=2P8#EbY?k(Pn8!-Z
zFvhw6g5+t`l1eFaskFzkHq)8*Q$1^qbjKT+ixu<g)F-LPW%+o&2r^i(ooC(+2caoN
zETNTPh8V?mfDaSNguI~(S~9Z0zTKoqp?!sKf?U0AaV!*FnPaUZgP@&Ygk%-dT6x{b
z_8w|QAO>OS&T9ocdK<kLR6>3^{p0?eT)M$HAHpM26bl}ITwZPPE{VS)1_*(<2JQV^
zW|xSrpqP<yW;_gJH6EF$OqCXek){=tkH-zN*IQOflv6nFuhNU7DloFDtz|fUUat*b
z_Wm*_C|_r@NXPT>j+EKiK_<DjvT<7bBsXKKYB4#dS_@tIMp1bVuokV<<FbtV!|t92
z3zeq9xJc5}f{uPCE>WwEd&9%~ofaFCQB>E}?LA8KJ$2VbQ%~u{3}tkK-dNkH8tTnh
z+Cg=%RiwKdjuc0dM^gr`elE7C_XL;5W9;Hqz*mqCDNeN9SqxJAI%4(fWb!(hY&^Om
z=I{_9nAcflRtTV`ErHCriy$@yy<Oh#$MfDHJ1ctQIViHE&HB3<Op5G_Il(pxXC|Z@
zL%4_-jemGFpRjUgrD*VbQ%vF4Nilb~kxL+01Pj{!?OM3kb1y$4r2|_k;K|!aw$AP2
znfbFTmT(p#9`|^0>1NHm0#r$SHke|F(&$*^pj|JAy-<;zc@}eK*&yEMTgv0WWqIo^
zG-AvDR=Z=?yR#AI{=EY6e3KVE6l;PhUgUy_z7IE?%fSJ+{`mTZI{gZf8PhG=p?HuD
z?)IAriIf6TF5(21pW+~dw9s@pw0v-o@=d-7xWTrnv<Z4hAV5lZhzZhiDfVrOwGjIo
z)ZT^69uXro9{JoE-R*`RE(mCj^)Ui)kiGdwe}P5<XQLZrect@piTg)EiWGlj)bU+x
z#a<dJ!)T~iV2vPd^xxUfjYdzHZlD&Cof;Q1gH6E)Q*W$%pXiTBO&Ri4BVUU+Q?!Wn
zdb>|+uQpdfV^RopDH7N{q}Zk=+$i&SeCGub9>c5pRUv!XHV{+cW5IodibA;lKkQF!
z{FMaYLwUd~P8o}cAQ>2;`T_`?-Yfn-&pzzJLFm$zySIw&^Q%21D*ty!h~kj_v9BU=
z>~c>e$r@0L9g<uXwJs$H@CCE2&SV7k)UAnoAlHQLH?%(BP&AzD`oWCA+;0lCJ}5cK
z<IDqmV$3P@WH@fudr(OJXbx9CryHeVL;R&|)9Rwa<Q2ss!_~64dQLY^3I(h)vT!PT
zc@kfk(;Wb*yc2n@{C}|lu+LntmVXE!4?Ku>?jKzq+l;Z$`xO|IqWiyN_v$$jzBO_E
zX!9%j@y$XlT2u&^*56MkN4(uU7j4!k)Qm$=W{y>=!=g)|W-$I;!of)M7&gs)>tD?K
zt<kmuA#pOX<DIuxpkMx%!nBj3lfLe-Fy*_TUMsSwQS3^9Aa)nez!mN?{2v%F;Cjft
zsm^0qXBSoZ2kp@&Yh^GB1bLKSz3Q~yXvNmC;$!_}fxWd9(^H3xVQnVbG$lh$eMZa;
z4}KB~{?3>V_es_eNz<IVtx~ZBuR&3+=3E{$heTOemCJRa2W>p+j(vUzAz7n-^y$!l
z{{oHf)`N1AeRWY{uXRutYE+y@?6IW7^q$_#O9WlDNXw7Xc3K|`<=Bp1COX*Z+czu>
z0uhlnCzQJN*(b*DD#}>Q;=b1bLq7<=Q;CeB?y3SKuFXhzU@1=^p<p#^+#Tu#eY+Tg
z-*_F>`V2ApwYAmdnE~A#S8p-<=)W^6SmIZRRGolpRk{pqmD4A5%25(}Uu$7;ngRYw
zI_G`F3<pZYm+69#SgAu6)m>yP4R9(THeDL20KFylW?^NWbT6d^B6s{Bpkd8#A}yH)
z&o6Me2-EHH+vv9j_WETUbbY8pmx^9|a<Rb&ipA$TPIdF)XcF8SZyC8F#{h_kL{@$w
zw;-$D=$~-z9ZhI)nie|lFH9`uNr47+a0$$85<8J{Z9S2YZW=#`GRf&|>L%!zW9iyu
zBaa3?(#8(HJ{H<vJDQL<3k<RyLuTizB<-~sU0_8h0Zoh-ag15Jv+<F-i(AmAt;+ct
z5ho)dDPLbRs<E_&0UV%B`dO>K$#0uK*>Z%mB*YzNXbGEj0JJ6CGE!qTzXBuhH)HhH
z{No1EoyA9gQ2b+f6?vkxg)K?gI>a8RWfO*38=$5J-PYL@2t@^JbxCd$;jm~`Tx%gA
zBCY@N7YPL5A2x<#yGQ=l!V~=eT6nHPu05Plqtt$!IfkFP|DJ!B?#5Jv(-db-=j1u=
z>q}BU@tR=S&_*^L+CzVI9dSYVszoCLgkzF6S+D7+tfy3#j>qZcaOf9gmj*&eRi233
zkn~QaT3P2ua{R|jv?u@`m&dKvV0T6~ILmjE<>b1^9?w&&4W9Bx9Bx|rs+ctW4Q4-y
zs&ThkcCM-{9R<B4O`bllpU#n;<&{}@Xj!t&w(HS)9h$1yb+Kj%{8!YAflKb2A^YA4
ze~bSZVx~egZsT1+lMSr07I^ZeF}444Eps(Emri>^+UPlSX|rpV+RGXQK69M^Qhs)q
zGvU+nQ~AX7RsiATV=Kr@E652w9M6Wtp`G=-f3&_6qWGLobOhbXFfF(}xW8kaEsCYB
zK$}n$M`Y3|m}O=-`8elZf`rce4U*yhebLW*j6k<E_5*60dvO9Gm~}EQs5pwk>am3)
zg#y!P<l*}-pXf=hrLiZ64$5oL{DB0WjQjP3*l|CdpM<M_^6T?Mvj0ZyP!;XIiOAKp
zL{LBH>t}?QM9v#>>mRfK`@?{-iT}5D>TY8ye8q5na&`Q(*7bdn&`+}pc`+%<J&7wm
z-rJYlr8F2%?EUy3U}>5ZG68!@9)NqS7bljXG%7^vMz&xUga28Vuzr~%c1xH#IFqBp
ztfspBBkL0Lr#*5Z?;8w~?)wYF##*-QCV5w+j2Vo0NbNqOun*>Q(EM3zD0S_q%lwz9
z5pP>cF10Y#e95nk@oNQ7mBapt$46PboYz7KC4v9hcw3vD?rk)%qo3H$JsG7!$<6;c
zD{P2z8f}_3_F?<cwu^yEgul7X!l41Jj5C|_xYU4<7C14=a*N%8C9izij~pu_i^hzw
zfA3|;EnvfT5)S?lg~CxNE8?|49_GS_!({svo1Q9mNSSh%u9HAGXSM}4d;g;4O%n0t
zXe<0t$oF_`<dFDm`6pwlr+$sPc}n$g175YhMx%juzqzr#^#^%UY87;q__AsS#_=Nz
z__e#&u&?QEon8f*r6T(>!ZgG>xERzqqabHE&!@QRx&klz1pps$Lpm{L*eG=pg2NFJ
z$6CEIMG~e+S%4Vj@eGqtt#<PzkgGwd&qbt?CQdi9;*)mrseALQZH~(H2T`8guxB50
z71kO{A2i+cn?>pgi2KuIFHmv7de_g8H^(k75ypNlzp;ERPVQbMm|AH=ybI_D?vP8g
z@L4D;ZIAS7&&ue;WK_wedonl$(5N^_rKg#UbF1g|I2Yun3+4O$Q>rBm8u#E$wN}n2
zj7SlB#KZh-cAbFb?|x0XT~g}yD|}}iq?lY6LAX|PLPT%MG6megM{yd<H7<B;;Oc$8
z=URhg88BkZRnbimND|H2(G=ssht(q1%zRzCNr4bNT`6gx&Z>?FQEk8|JQGSE$Ms`@
zF>=&b8z{;D{8DzrTB=DKwv4V^Q%;a8^sF4?G6{4?Tp5_x`I#HUEvzp3U*Tfa1WB1+
zcf`J?$`}TCBa=vH=FKs3va(e&<uQq?YN*Udwhx=y&`j|%;VR<N$?PpVMgQ}1`<vcv
zy@()#>ZeQ+W?Yg7^5A1VD;42&E6Od^&-<c>H(eN422#XhjCDuuh4Nce^LIA<MAHi*
ziAAl6BA3!B`jQvmj(ImUPfeo>aey#oOuFe!86lzDBW_DU`)I0Z^A8V>S)-%QRA^h=
z)yX*EHVZILqKJqVv19A`w0oM9>Q$yo6AJX19ZAl;ktyQbBjV`${bj9c1H|$Vy2>U;
zhdHTXBuEao9<A!XhD^<{$Vn3XYw7d28fTJ!a@GlcD0T2?<sL!U5S=h0#o;ha>GF4*
znwq2CHEo~Pj2)ev#d=liRRU$<`^-6|aa$-(ULDer9W=v{=eQld_`9x`Qd#^#oUIyq
z(6yu}AxpxJY{sP|w+FsIfHlYaL@8Hv4wHJQ#HVI9LVBWx=`t2JPaE4S7875^ta=u3
z+m%Zj`z>L5*fE+3WnjsAoYkeoUCF`{HPfv<Z)MXYQm52jd1HQzZdFhscs0(*V@t~u
z6%XM0PL@rf4ysDDLyNIvD#D}4x*o}lT!1^st;R(L9OQFxjNzI0<<tMW$_gXpjntRv
ztR&!I-K1vi+2GL&*WTqMxOX`0`q|w+C*dL%BCR@&KmOyO!ciqe&(B`On?NF6Gfk*W
zG>*foF|ka5ovkoEYS*>VX%a!B$I|KJB!C1Wiv9J)fe5SYnVxH2Lz_6I0yQrx-dcn)
z_F3Q9;T5Kb!L8%{{Xy{q|Fi9nkaFUMc!dhT9oz_A&3uZ&ylir{49YQ<d=b$oR#ZCd
zq<wp2rnoP006e2@zhdK^R6X1aE5zNW?xBY{AHNyoWuRvG4>bgirNb-~9Jys_1ZfI(
zeNsS8;t#$#iDGKS0!_-)8Q@dTWFw1dk|Ik9T?P{gzgT2Tjct9xf~9m6)mF6l)yL}L
zYP1En3NOw$UrOUu<Hq8$0MON@TxB@{dgae`hux4ytM{g=GbF}(;?lA7ye1CdtsKzy
zP=Y3HhB#-E8dWRiHVV3Q#)NjKEz@EpVyw#*(oh@($=|-ymM46Cx9S~3A%_3EL~11$
z<3lJ~4gz&ad{>((!U+&N#tYY^@UTO^=c4wCBamWjEQga*+0Il^)y?g(!TRmqo)f(M
zDy{i_0un$yQ)~68X9j7Oa-%1~et7U^NBEKKN#IJdR2|s&eaL#}Z0g>XQNTi{$C4w`
z8V9je2+3r&iLUx1SRk>~wo(^NLPEbxBZBW0Cob>8#5jdzTc#nGFO?=El7gGIU6d0i
zaZc=!W;2q2pha973pEs77Sg|FHw^-tZS3#|NHh%hET6K5TZh8&nGkpb5e%2(PoI=o
z?uFHu79#Z(u?dF#WpZ>K2(;527k}WtzM(r<@c3mXduh|SAa?qhqrnokXr*GeC*sd0
ztz(gECIzI6MI04S)IS&_j|mN(g7VU2E;D%9|7g1D%Q+6P)h6>UmGveVvmCH1{y;L7
zttibMG2D%(A7Mn>Ox64@@*f~WI2LWBrGUSnpb(NkjA>UQ$}Zp+2pShtqjpc|MX}1r
zE7$_bX4?jE0Wcxh!pj|QJ^_t;vY;?OtUjl16Irz&iJs~XB5Zc;U_(#L=+Qy7%b9G0
zE?^sxP&F9zDTyqFwD&|mqalH~=zSWI^B4xDyrlTY-m6`wg~~q(KbJ&i%=Xz<5VNPZ
ziyI6ge!6Z979;Rnjec+0n^;<dI!1r<AoUT``#;oAn;<XjpGk(>qo?tWSTGm_=ez>T
znAupNIp#SF2~qgN;x1n4j2q5(%zAk0Vj!ksmeEc|XU`IZpq2xv+id}b``&~K;~Kck
zSpjWDMH;U3JBa{0#YvE)IM<W>q1vj<jYWDq$d+~P`ePo}>wf-Fl)ylAR4A#tjf1&i
z{^#RXNY~F1k<Eu@&slFQxJFh6guT6N_-`+^CFFbO<ii82uZEjYL7dN37M$hd^VI(V
zh(eR+W_5ko(VcI|{X{RXhvQufMbB$~Em3!$WprK-))hs%-VpBwQKmcJP#gu51O&>m
zCIH*C+a^PyKOQbOR<jcoo}YYotL#%@{*=85WTt%a;P1~K*C*BuIg-da*cZ)bRde&9
zfS<$1Sj3Gdmjh;cj@f}t&fUodKlZXNNWgG)xH4`H0TKk)H?^h`u7<$oQxfwb;j^p;
z7hf#smt?P1N_i!rTh57Pt9$|CJ!uRh-bprEt9jO_&ZgqC){Z7i0CR683btgByDOSy
z_r*pv(z^qS`M0P@kv5=p8Hn$;dP0l)sKh(&YG$3f+LauI8jE+6)W`$~;v2dWlbXor
zLR##dJ-T=ey1N7<7uq;ocYmq|fJy)O<Tyea$YhqmZHcdc#Q2{v#YMqxMk5zuI7k;{
zFm<RiZ3hf(flC(E&>t2vfj7H$ZF!+Vj8hteFtgxwcoxajfux}2(eX=%$5C1ozEmH{
z%ve0Ov?#qe2}`ZT>)X;tf0VB?s^nq1sOSel)5d(vxz(X!Hs5E(?nfq`CPha9VsK)t
zIvNGx=m2CR+)>g3q1Cb7PAugSQi@W&wGOw=(p3?edQlcCnraK`4xy1er4^6JuVZ6!
zv#l@NHB6PBy!$+D<MK!>dMA0_My3jmVulQ%@Mb5zpW10?@Zc~nu`IlgjTW2|4!#gX
zO6cES&JdZ<7?yA0$m8Qx@3inDIb>H77<^S-zWqoOg^xP^)t<-mxaLV_`B(0a@(e5;
z6A4+74exLvS&bs>^<@!rS6xk*n=>XMFsSnio^%Nw^(fzOe>$&=$3jPzlHtMRqHYI>
zrk{n((simp&AdMnx5kw{*5294hR4jmUys+C&dDdKD$ZlAWO_GD!WYN)t5!mtqqKlC
z`jzR3b1eR4H<}g|6PoZvMv)RzcH12@k#+9*dV=P=w?(jf6e&ahSEhJczCu1#+8T=?
zPrbZg)Cw_`dPl3@rUTqyN1Zss$?rjeRA1>LNKRrE7<d^hKZ#w<Q49+N(3$n4%c+K=
z(Va9VMGFbaaVQuY;&5!(ly;lRYdt+I<L0b<Z*HVy<WV27SUv5jtwq<_pZfMi`dhqP
zyiSN%y0(D6ke!USm?iVwu<c2^#}iSKeU)iinuiGt0u)xKJN1PR=B3nvizRNhlnq9S
zyD`i5Wosz;m(++b!@2lrup1S``lUi_{Z3e-edO*(yUk^zs)TIen6G0)1p79*0U4M!
z_yy5vD}Fe#xEBPsYOIUGNI(S#K@Cf>7rXV7<t(5gx#}O7gy9R+M<%MlPD>!Mk!H~l
z+TSF}@1jpvQ?<Q?fuSvXQTXRagNqY|T}TgSMY-8(C@CSdr<-;H!dU+SHm@ihPDYEO
z1g)0_-mcV>;oda?*y2IX<n_8<CtJb1x~-x^frU$h)c2m&acz@<1qbsJ^@V$|uo8`V
z>I=__<VqUY738Jh;q$c}O{%Z>(}8o_1>o*a*gA1XPf*ATPEIn55OQF9vy=Ob`i_e%
z>pH-wju6~3)mABFpS$=g(-dH~Nn3R0gF_bM8cv&mB2T>Q%{ebn9m;lR^tV@hTjc)t
zXwU@L;(0q(eRz>t;-l~BIwdIck1rhVd{Uto3Wtqlb9A{A)}}*#Rm8D7_#xqmx!JxX
z>|#b5h}q;aoWI&z>Ky$yT>)p_pjtBM*Oi%+D~vc<>F?oeP3qoTni@649<uP%IZz^e
zC_>$&SV@2rfsi6DZO_udS{!_I+9E=@1v6fcKLEP{M!aN3o&d72BfCbiVoigOmRO$D
z2}O`6`+aeY<855hEUmwv4_Hj309}XsDR6@!fe&rGhI3mWx1F=O>D8RYKy(C`)$6^V
zE3(?Mdpidy8ctlO7=+T6aOA;GGXmQf%+^aVH=OlUJURJz#QotTkEu|Lb<`wK@Xv%K
zJ^U45O+J>Ip(G%<AKyh38h`O0K>1WW^DolhQ1Sq7^`!*DOS=CdG_3!}z;O{B6Z5-v
zv+k{u3EY|QWNyF2y`EQKIB~`aBAGTtk+zN$Q2e{;*beftb}huxcij$)Hn_NWCT?F0
z$FgX$=ouGS7d*e-3nY?yioU+=%Y<!mJk>o2Gq9Z)0~qA+*?BN{TdX_9Anz>Lr5q>t
zdYH(l#^=L{3-y=}<F!&&zM(>mU&j+YdMRLgSH*>!=`V@0MmZjK1$qg-@*KX%liayO
z41$m?`0o5DBLVRfwa_`?n7e`;mCy#=fuXhRqm?PSAswnJE>}sw^0M{8^_AX;0={E1
z#!2ZM^Vhc1Tg2f5wQ2$#51!?<?&F`N$zO})wV5nc=dv^Kq@(pw$N4lP4$7v-b92HU
zG7^M&?ai^|@wMeBV(=yy4~(Br!ZLsi42@r66j>|mgVENQwb)O0+IN03(lPz@GAdpM
zmqsEU(w~-TqO#N`*KnnaXt5_Twd$ScDR>RdNg12^<EFc9ej<T)=<8I_=H^f^FxF*O
zn~I9{K_-qMXRZ84x13XR)tue**qZWc?7=2vO%YvJuq;meXkcV{6SLi-&Zv!vE9As3
zkAzgigOR&j==XI79c+FbfL>i%>%pThEFi6|Gu$Fz?m}gPT5D(hA$3k^)>eV9z+4N;
zg=fX4TqnR#F|IjN3}<cjCCKdbVCJou3;kJ0|M<4T&p><F0!ESbjyr8(D%7S`0dI1*
zaO<$1eiKU8&t;gK*Hz~6EQkN?mXX2Dt(-RdJ&$4G6Id68t}ugJET6NzyhjwKlFag*
zJB!(Zu=4!}Km}NiVw<B3Z-rm(MYt$hShRGu&iDAGKee3W$V2!sPJGxw=(7Mg2ueo+
z+kb#Di%Ea$iUhUBvl5Y`nN12&NUMes<_slmOhA3TG3N5H*`gQ!YDj;c({E^Oq#`EI
z=mbcN;9nt3e>Z+^8~}qn#QGi7;cq0!Zh>$SQ37$pd6?aH$<Gx;DB=YQ_x|2<Yojfl
zoaSp>!`z|PON&THE=>sm{MM+}iyfn8yrYXc7n}cPf^W%==7ZlbO&oHIzVRdIMynrI
zUL;>*{r>%Pr<-#wo==v_b9m!^whPIrhiqgt_+gN849cQ`oQSMmiKQ0&Q?Sz)U|)E9
z?`plynIFIN0Dg$CBg`RO`Ft2q`F{K5KY#?xl{f@MT5TE%^z}G<HJ*)uaZ^akZrTVw
zums`lepmhKR>)HJai_Ly@$MdOVh0a3Y)YfO(y`Vxph{0$&1|wf_7kHX>(@C(e}^vl
zpCkL{HQ`y}s(L(5BVV6h(?O?*{gL5h$ZmdN4pRK(0tD62@F^VW*lLkei4wbgaexU<
z1$H{Kh)nZk*nt8@<zjb&61|8LSAHylcd$sc*XtDXjl72DO1jgEzel#CI-P<)8rMno
z*po8oDeYy6%g**drIO*`kV<9&UnmFPau3&ODSvl{YB}6E(h#7E9<YzB+Aj2a{G3!=
zlnm4F6MVqp4{C!C84FAYm~;@geaHF_jC2WB)(AjtjBVnP3O&MNHS(P5#Cbw&Lb+;g
zdaJSpE*zXh8d2&Ps?N_++#q$wgTJ+)-=KAyH!6_wn^O6$-3P9H=}1%VMKOMc?+nWx
zdTH3^QVw;+(YW?MS*EI&7@e}(Ur{V#@ovxLOUu6KigF&<_@kBkHAnMVd09(P>TU(i
znX&>$G2Q;Y4{FbID-8c8+E^(;4a}=igzppM4f32WO31N9IlvW`iFudu_FIQm<o@5=
zIQ(BzQ+^>N{i{byXe4QSx@=`~HhKh;1LKTt&F)FLO#~pdrsf05B`;@&2i@4@bdJiL
z#)J>PAau&#p@jk~h64A9`+){G+}d(3S+dFRUmFzjLj@FwJA$d9K@d(Qk`Dj4L0(fK
zLx#-f$keG$NxMQX$s`3X(ER`bk`pEV)iotnPTcwU7473Mw|tJg%%ZFSHX<`fX6_eW
zq~lZ@+1S^pF)kSgDVyq96zA`JJHyI{#FZ2Hbi~$F;k+djCGUy71$XA60}1!{&kuKY
z!5V&pK|D(wsv55*$tL}KyHDnG>8QlFGjmrvYU*zC75-=Yq#nz`roJzvFoi{O;<o!u
zfMqqf_sU3qdp<jx%<bV8u1&koxGG8biquaCw2ScT4_o-3wvB9q(|fmtfb<}~`;(<O
zFg!zU|E2`z8sQQ{BJb(nFO@;wq4*!+IYqZrUxWg0&v?;2D*VY+9paG%jQ{vu{ky_=
zzC$(?86Z8H?kGVy)z+QC_Bo#y7Z<%-Ni=RU90hDY7}m1M5(PiJbUAH}_N^_x|GJuo
zE0)Qzb|Z^DYP34JVzZgaDWXY{)8XbE7(H)3$u78{9<KBI_uiML>%=`TA#N^Cw2?M2
z5WEK^7{2Zt*my@6Y~CsfCNntZyJlFscN_~2VQ<`V>`fLL$hrchaQ&^(Qw-_v?|DQX
zc+%~QDFn*Dx?=^pOe(h2vNewd3I@Ch4vKmZ()=6^7=af#3F;O0yQF!>mbC2tyqavf
zmcWIIZ8>J)%zrx1t~g<5%q<%uVhFdD<o;34W1s{&O}R^AxM)KZCiDAfzHovS7Uy`n
zi4sQ6tGRIIYVdvZ`-Th!n?l{qO{ljCs?R0n^IC26|9b2#l1(YH(p#Zw&GHIkbYmJt
zMe3$vPdO396;)ZNOF(_8$+F^xxdcXa^gR_4g0J&cGlv`ABHCCF2UVZwXoeus=f2j`
z!H?P&-jmI{vcNyy+m1jF_bNY6@bdx{TuWV^9YXfOs8%<XW1$Y_I7H8l%!E%S8#caw
zA{0h0gX`vWU$=T)myDTcDpLBQ&JtbBppVPjABA5rV3TcujQu5WU$hJBcQ)T9&Irh}
zQ87ktew3176)BUsP@D4PlQxQ5%2*bMKMvmZE?mzl*-<N7)ve9E7)&x}HB{C$m8Cs+
z%Uxak1#EKmz6x-FJI45wwMqI!S?=>2@p5ocaVksaWo_AFsDtxa5*m?@DEm}AbgH@1
zhu2@+8kDQg*`Fl;ilGx)-qSmgVRr0&BQaUEKV1`i&ceBEobg7v8Y;ybKpqtDyA^VB
z^GR>uA3xC*yYfWnp^v{)uZA;hXlhP@wiQ|i`zo?fUebg|!ivjx#jB0%|MS+1t!4&G
z`9n+f;1{swogdN*zule46UEDwW*01P4~eH%_Kl5&)0$m{w)g$$r1!dglX7l9@FGc}
zpl|;4iyAoFw~~P9idoeS<Lotgh86W<kfC?stWi$e{N+-glEGUwn7jImlS7=2iS*W$
zk8v1s?&h8)ax%_iZZ|MPr~7Bp;WQ=Ahq7_LpH)bg>MX=q4`rD-lf09vo~NBY-$xt2
z;KWkr_AJBD&3qbLG1x(m^}}i`v8bp^r18SRV>Nyba94Kafmja!q$W$7#AMqy?gfE>
zWVEA{U?~bPGGqnN9h9pl?){iu(OEjUz40PQhJt|_v<*tfc?>%k6XakhZ~)5=yatK=
zWUNWj1avQnL=)js9xyAAqQ;sZ^nC8WrQm3{s)5cQ2)`JDMC#1vC0-hVmHM)mj#`nN
zB{a)lO6*_%;tsO^+T!|z{dqV^qfZ@dGAtq~E~#)bY`8yO`Sur#ZHN)iuh{DsVi$Hu
zTa;@%G(<?#2#Pq^WGmc#<$i+?tGE~R7PPpW-}Q(@^#2b)JDUBz2xwEWCsgEsNXkZW
zP5+ZrTX|Pwu`6bnc`>=*mtPIt_vr-QT`rG4HS@VvCi<o7T@H-%Yw=BgHT9{PzC(Wb
zf%35PdEd=-0a51pf-sSP-|4Su>uNaF5=%0|>vzN4C>SjO3F*O3GIxj*3|DyD8Nk1R
zqk|H+atlBh&9!}IB6Za9UF!Hd5iTyBy6#FsglqtJG%)PcK?NQ=#{RgM+!SSAy(iZF
zr3&E>)004rWMI|<S&*jInoE&I?r+x_t^+qq1PiRClO_rj*xHP5)-}%R$zR5}#EFCX
zsX}ji*(o}0N|JYw32WTVq)TDoxMhEuS^VkBeOh(Hht(%FZJ0>wUmqVZ(C<%wTW^is
zw)Z|o|M6vfw=PTWLq|2d<ORR-4BO0YXRr_XL%&04{!BkQ(-RLY<U3F#qW2k7`{qBu
zx7Uf@#zWYO4PG?aAFp~lKPCPTFjPVLEE-4(DmxW^u&50#ObHS#L*lL3d1C`J!Yb+k
z!ND~V^&;=RPhDGcszU4Km!9TO2jd0u&#PscpxhOl{wz6IJ%()C!6XbCuv<U5*Ia$_
z<?*t2q?*vhSCW40{)ZKf!Yfz_{r{C^v5qBJy1Lpj|BH2W9^3dK%YuPl$*Yop3|2Ys
zDc3&#%e@G`R=K(y9Y7CJ14BYp(UY8_L-F*=1hziUF7AsqitJ+|hVriVx$Q*fm7kqy
zyb>tnu8`l2mm!;^v1sA6z~VoVV~sBXC||uC)l9hp;6`&RbM@N{2VP*c^`i>8{d;K)
z>{pu)k%g(cl-Lr&GN7nerDDcNtyYsXL5@>uFaJjJc3XoW`b3^joYlXgEef)@&Vs@u
zFfm{r-E}HBpWb1JwQ;?dP);Msau?#&UFGL5buTObyLw>`sDU}G1u5fLtfqIY-W6bR
zSf8@zIa}do!m$+b*2yrrw<lC85T4$|!&5R?zKZCgCuJ8GWC^XY38Ngk`mB?#%lWiB
zr=q>nkUguTG{x7#XRr9SS9sH(g&C2{GM|%}+_l+$4l%$u3i=&QfC5*=JAAv+$Bagj
zw(PUh?>U5@*U(3zlb4$sbB9_DhdarO6F$NIZ-&)YgsW3@ZqlaEJ-`nUD~zUN%<p8g
z!dDi7=eB*Il)bXn%CSN;hgWGhhYezqtw~>!t0p6y9!F|AK)u|Tk&$LOiW|b3J_<Z`
z?Zc6P3#z8o`eiT>qAG_xIgqH>dUwaN!96RXrSma#WA}T*E$OE}Ga=c;?v3pIV+X|&
zD#M1%p<8gYt5;9<tu_-jz!SqpKH;S*U9Hw$LuR*7HM*4Z>(5WHXA7mUfVQrb0_kVQ
z=?!+!fL6Nf%+Afl;OV)R8{Dhp(V~Qbqwf|H{fmg%dvH@MdB_*R7iT{k4e$MyenUr_
zo@GRskP__9@bJV<VdG9wRNv7*h9lXO!%`p0=2gOP!4LxryOyEUB^lHT22Ks#MZT%B
zBz>x&^sSk3LwgV(S+oyiM7%~>@TyW<I>^g-+hmj+9bF&ut9v!=3Y1_gW?L@4xEEeI
zx_D_E^Scv!a611#hpvU+DPYk$(0Yplic^V?HIf^rzOUk2&+H~$sMZPVxa<6q@}IMf
zwW}tUk1R%T-+nrj#Ad`&6gS19%e3gEEe7TN<qfAuYv$G_+1YxLwngt%Ra6PfI%s!i
zxD4KnsfoWNM|leGnx5m_(83e=5Aa!aC@@-xTX5X()_qwdsD24&0ABT<SJ#6gB^jiV
zA7}YXrB!p2xIRT4U!Tkt6EomUg8l5hd*jz!oOCYTC*9Gm5{?<QH}i^Q3vd~2Vdu3L
zA+l*`a=_n^zB}uV@xGd^yg!#K&pJ3bS5xxHfy<wyn@rch6oS1upIOpTLhCGHZRDS%
zS|qPu%Rak=@Zh@+6b}nK{izb5$XjK1j7EXq<dfITW>F(uZkGxj(Fq%b#hT21uG;<f
zm|Pb1Em#}A^@log74<iiaqGaV_a5?J_x8{R9{xkh%Oy;pololzsUKC;4euvYgWmTs
zb^#a~9Iunk$x@=+CmBDyj{`4bKbqBhK|G&SWs@G@Atiwg{1tyUN|N#(Z$=x5pJ)fy
zDQj0`>y~T&vWReRZ+t>CiYc@8zqBW+itv>;=$P(M9h1x>6_fQne6qhN6dzg9=)ann
z`CItAL-p_LLOKHM=^FzEkO{H~By=BP*p{d_>0C5907G!@h7{j%2}6#p8<9X6;{#tz
zS)iO-S~_Bbf-{m{82H@i))QZDI3#NS1F&Xx^H-yBjD92f781@BL4aS4*zbPB|DZ=Y
zs!~XX{>CHuJ9o%Wbb~0&DA`fr$fEi);X~65KO@fZweSno@$;7i(XC}oIGmoKHl5Jk
zv7^VpfkT#~=RT`{t|W=C7HoIL7uUjn#3;7!xD^wgyBGYmZf>{B5ArgDH>)H^3m<$s
zT+I}FKFgQ#FV<)vcc*wg{cn{|&0jr;3kP=nljBay(SA7R>6)QudKplEpB%w4;#R*M
z9VBq8yQ}$2bQ1Qy_WR@3O_6u}Wf4WO?uLO>onE(BHM7LhH<V!a;rPod%0miBT~Pgx
zzbpSMM^njn5OK->L)}}2#SwM=o(%+t#@#)*yF0<%-92b<3(&Z`ySoOr#yx0AaJS%Y
znSS3h&zbW)b2WG0^ksF`?yB9Z*4nbx|CeC$Rpal3&VhXfs0|8o)Zn!&=BsRlzUSBC
z%H>J@3ky|s0P$Q*kjK+~(rci8i2h@GGL2Q~)ooxY6-C5ps<d)Mxvel0*Km9hB=`=J
z$y^yw7B6X1bTQpERv8wsv*(4dLMzM2EvTE@+Y2r(bh8zTg<~8iQ3{*2)C*7w;aBgU
zZ3#sz&n1X#qGl?@37&iN$h9?JOV2{XI0PTUr2Ae@<!2qGAPnTXQ7yDC4j&))>{q{|
zZ4fNi8JqU&?a*V&Gz(;M<Z4t_{!TA3`!xdDniV7--wp7=Lz@yXKfM}Q@wK+LoD8r$
z*<ZiYz0S{$!8Y+dVr!8S9}3U41?!vAXBGM)5|Wxq2}mIrRTuIZfVHubZ*6-x=CchX
z7D23dqe5vF2)mj~yX7wRs+N{Y>F^vJjk7DVi8RtrqQL4{rCAft$5F03@yMqhz@ay{
zKf{H;Nj0c!R^+K&a?oAV>-nHZHlX+M&8dv2FW-aA>egKN*6rX=<(cEKtY}emy>-S1
z1^jEO=naVt_5XB=gn{W58U9tf%vRc&9{?c49~m!(_v0(>gmaw!u3b7^1V%M;8<r#d
z(dm5{ywCt-CY5;BZ_o%Jtw=8`Sd<yWr7M^+e6niLsZcTYb-q1~0rc>%2T<%83&#=f
zlZ~_-oSYxT9o>+)ZB1@+><oD5+%+tYoGer}Aw0e`l;_?|wa8$_^B){@{ymmVg;_m4
z-&@UKJRN&DyTA4=&Z}`<+00`N6gVUaO%b~r*!JF@n|?YO!3o(}+1uoYB-0v@<R>)5
z8+u^>{H2AByk|D;l%CK3(6ohwm4GJ8_ia||Pd`)Jl%NZ6I?S!Z)#7xly|BtFNEuM#
z$UJw_-VX_z^M@uq@6SlZPEU=uF9OfbIZJ^Zeph?t(%64*_tO1^iN|U=m(#B1=ilu>
z;dj_le}EPZB_aby8FdY3e||LXkJqEce<R%@(LdQpllo0&i`r>WqhcXAgJu7oZ8spC
zz8o2R1<Er{I$9alGxB|d@$&ZbvN@g3>skT(9bZlrh-}`>t$@n<N<dnNZS_4o9abmi
zC5mA0YyCtbU4L)L>R`=flbE4EPJ9ND)!tZH<XW4<+1A>k3RL%)1C9ddYAU1AeQrz*
z-I>5-V=u@7gEhcEK<HQYyL}O4T_xBZ<LvBuqEweuu2F<4fNOEfFQQ>#eb9O&hTn~_
z?L}vu@)sy(NlkNqIHY=l$v~9pNsygJBv6Bn8}{0Bb>kaX&+t0m{HD*?df;3KG)o_v
z{t)yA98A~YKY%eK^?|ZXPwzWJpT@FC?$=7WM8Mp$9Tl+Q{{DVrbgfOJ%uehcHfFC9
zKbjaCR<(SuOhPI_y)?{iFG%!yBr5USn3yeGEd9hu4x@j?Vo)C@<5gkIW8h!>M(}e_
z0@_y`@oLQ_?#jsEGc9!Ze}Gv0I)<P5iVkzxqxV|n;&svz57pz4AQ5!X-xk%I1vn9}
z_Y7JTsP?bXAc8=aJZL)n9aKIp%yZV^X>o2C`WcqMm&svoOe!Cj5SM*0l=$jLH`&Ji
zx?d!UG%o7ZM0|CE1i2P9P+c`wG`;odAK>KSj-4#{1aT_Ee<@t2oDfw`A`qwI(fA+W
z?CiC6#H$Uaa1Y$Bl4ft_;9p7sX@wPRBHGti1Gxx7lZ#x`zu7j@IWUH}VjB1--2x~G
zT=%K>-aKKz1M%jE{04Hb`aC(Yv|&D=1HnWCu&@o2uLGDo*N43!Rsi;}UsHS!saY!I
z@U0*RczhTv7MYzsDKOpKxta5o^Zg9=D(}l?8!y!A(iG4f#`#+@y8rPvD4%yK2xVv$
zxuw6MkRg$6wvCCOTbktFS+U~4JOl05QYo&^M~s|6%iGUgl4ols-Pye7T=xJyvnXwP
z7hSmWLWvh6!+g}BD~L2qUEd444z<+)Pha-_Q*1g#^<CG)vvS1?+MRKJKdK1k>8jB4
zTHfOl!&AF5Dlju?Cnr1GvYOjC<2P7Cm7<^fyI-L8hun!Dz@cA+?sD#@(XOzP5OTNm
zP^P8rWRX@?<US5J-2Jc{tgp8e^uz2ZRgY(HX8j{E!801MIoHlWZ2<$<&Ngaxd<4-c
zc^x?h${jMkL#15<55pe3+ef~@&YBkFcF1uF6C*sFdiBL{kwHG0@N6A_ym(xl=K`xQ
zQ6&=6<gG8)G>7rO&Y0%U`$R?B^;2vPyST&mFGJTZB1igHl7I3bn@jd$hPfM4ygb=n
z4EMb3x|g@d|0uAkcB@*{scTYKowIvj*oErYTnn^~JSELv>D}A%pA^;uKi6I?Bf0jK
zP(o~w*%D%Mm;V6H-$xM3sX%1va2CyVH)TqFPsc$+Lu7ZMxQJ2L<yyq9H34j1>Nx6I
zAQO57hg&UlAY@iK*68z#k6mZEvsLsKRVWozyWFo@0FUjpihQwQ2kPQQw<soFDcJfD
z@g$<Sa9Q4}uarXa0|y=cBY%YPUGxHC@Q~4`Jl<7h#G@a9tdjtA0mB`{)WhExDBd?m
z)kL&TBDQFdpR^@kd!z4eULJRG`n0MNwrCKaw8gtWAoqRO1YudIyjI=RDF!}&RvBvj
zxO{z&GC6@D>{KRu*#NVp6#QtTt^WY+exd|TBuy}x8xNvV&$?x=rxM5oM|C2Qhws1H
z9^TwRQ-1#ZY*1!OCQw!=#tDU?T@lRc2P`M5$J?faa>7p-a7T~zw;R#4zyFMC?*?Ff
z&!~_T;lX0G$Gx*D!6^row!v}v&jk|~;CxN1Hs~rV40Gz#FW9ID34#Wvl%_BHB8oeR
zCFAvQ(Gu}V-1MMpG@zyqXb|(j&yFhbS7?hwr?^u=8rXD5*XBar$8bWYr~d#9sZ)%a
zVQfdH&9$N3&lG0s#JVA~{*VD;i&^l=1YD(t+S~yKkmqAEQj!tjZxGoZF{njlay;`w
zEPcYz5K^bYTOnNd(>hic=XZJW&O-kKH-G*kpZ67vgSzHPA%Xb1dwpoy0Z=`mCaBD?
zX#lPeS)MyO^;y<TyfTX5xPgxyb_l$%r;`gh?l|tpS@^Xm3H#HRD!%K^DXs_|UMkSD
za6JZ^a%LwIw`zW&;SaNp<SxSf(P>-!a<=pB4_H^Re}J2Wv;KG4nr@jw-T<t{zT2xg
z#x(PsfR&Vk!N)Pk62jKPvP6&~S{5>!mc5JUPct3UQL=vk=wZ6ZYo>|=?c9JBoPU73
zt;5Xn$Cv)YSGvhMzSmkth=c2&Vb>vyua>CR2smBn|0sAHXp+RuFM&Wl7*e;~T5J9R
zD*i!~9<H6AXdn6-Awue!Z`B%pLCoO+K^cGk0nk^6N&8jfU7_<yyC-fclts>vUP~ar
z4%?b55tG}%sQ=@TdCo=;ga9&`_n%*$GK8F!W&YLWHey3|5&|%&V}RTYr1>_BBBh;m
z$q&nCt#XL<l?A=?KBq@+A8!5iWw#<@Xx+PLY=8h5AntvH{%sjjv6pASA}^v<)xTk<
z<8N}SsPzQ?_}UEd;z3;t^-ukf2fn`junp)t715?|zjswREB*(_-X1ZDWDQNaBfQlj
zY*PYo>S9G0TmuB(GMb|M<(K%u+SsQ}!g}=nY~5$(cmUQ14d~0VW3x>rwb3M>HXl+E
z<bD6P1qTTIrD|$RE!WEh0_k~b?mHWUbmac>)$2)ve6_FvJ)6z-GY^5u)BOMu>_dZ>
z+f+kv(`ZSh{3>Z`J?a$iQ7Dt7Aw9W+s)p#!{NL_-5kcKOwIy1C1F#6W-Q0vGV>mUc
zJ*G{2AjyE+R?FA|r1h%&z<{Ga>Vrn~=vFtN%>Gw1<ZiZux>i(pF$gUH0-|c3{p(ih
zS50V9OScpB6Yf6;&Bh8A$rk5S51dQq0;1n}9yQOtQ~d(QJzmVQptx3^=6-BkH4vv)
z{hm<RLhfp(^aD-1dc)@p;lDst+dK*%H5BSWH!QOMv{YKd*!c&z5}$}fL;r=ex3+g;
z7@~brdEEG!v@#(m+njH;PFU-EjfUCWo~ukywSRyrY&75Dn`}y2P@2BOqprq3zyhT2
z&4aJ;Ar=)xvCzk`qj}P9l0Q^0db)K8FXAXmK=xSh58&tl@sBz0D`^Qj8IUuXZe)M1
zU&BxN{S?tW8QU;3@d&zQ_LmiS6<GWa*a2c;fe@oB7oFx(fm!ma_+R#Qk?>&_`G%GO
z!7R$VKI&_2kiAV|Z<1s}fG}wgFo(VtBm*5{fc9f>EG6-xhhz}D{6CDaRhFz)a{Yen
z#3aq$nXpE_{{>=M`jE~lU|8C^XszjEeJ+O-3^5fja;^O_GS*mb!&-y+J(71pw<siY
zr*!aBY<UtAcbE{tbr2KOZ}}+ng}ismJ7~=d(skJNxUP7JPhv#7is!z@E>K}S)Iphk
z)xdfOd1BE?A^+<3R{cL;cBK%+Rj~|QEx@6utFdOCZMZ@@X5ucS-Xxx@yHWvT#K^(O
zmoSA?CGLWC=~V1H2|NS&u+E*LxIv;_`zp9(c>T=3c|xuBfA?ps>Q^>{iGZ)+5h5Qw
zW51ShD@nI9-4A9r^9Dy>6kR9#X72CIW3a9{NFmxW-lz64(vbXa4>Ce)p3|-MVSU|n
zJb0mQ_l?!nj1qRrLp!q%_>lKWvS+9Yp%%wb?d!RG0>QoQwc^>||5Gay{#VrmE1$2<
z_E20%_58k#QNY+{dE%z)Rd)hld5LiVO@N%791S3BHpO`1D61-|;dM$jRdS;r`UEj~
zhXhEdh|hZwczP#*jMq#>LOWRzf4L0)PA1=r4~58UXx+cg^#HqYp#sVKp5h<#ea<4|
z)&S7>987i%hcQ6S29%gmt3609_LG^>PLRRVTae_}?8$|=VVMuXwFzu?98$ovzw@fH
z&>hr1=f(@E(n{awrgsq`_zL-GLR7)K2WgjaxW|4Ip#KkRiYbL<;k(P>V_#kIGpoe=
z8#Lw13ZAS-<(E}ioqld{s5zzAy0f;|Pk#56C2M=X7gEdl-0Snt9g;trQqkR>C9Y2K
zNb_8u*aX}>PE&f{!Q1@kY&iLQwxg-xhuSjv`WSrqBi?ja+v986{eubf*P*Rg-Mq)G
zpO!xiu=<1pBO!B`_MMBqE1=Mm+o}01Z9^Q#Tm%a7N~?aRMIsv1*~4pQeed-_j#-+K
z!_KSZNIzLm{_nlr%X}ew7bjELVVD?<<EO{rcb!cSXFPLoCb2okB{!1+EVu;c|1NJa
zlm0IPfeD%d0x&0W8i%&iUetd$V&b$kDInI99|*!+wlsDf?3|D>KNh00Y{CaxxDrKL
z5&(2L`ji$JEI@y`nd2Dl=fds8+|SDWgD+wl`m7ju-s<M@ew+s*0BrpKKKz)6x4%TN
zvHo5nS!6?f1d1Rg7{I5C2uPqYP0siz-!^ezP6TY#UpGc>z686p5DD<_GS+#y)>9D*
zcx}Jm=pgbz1!W$`2Nj|69}?fTpWUFs1@gVPU<bDCVyI+G^eK12bRS0a7cl*>`o^|M
ztW0u{sb1~BLdLoh5x@CY7p&;&_P#BM@P)wVy?bKmxmC2y>%Lz>EAFx!g4062zEk0e
z7!cJb{g|z1Qu@HYj0ia-5c*BhinbDgezZF<;swza4t6_eve5QttD(R3_J_GIBpT~>
zIVilg0KQi3=4>g(`|R!h+WkCRGE?+>1*ivKQ|nA(QKN6``<}WfB>3+Aooz-XM{*si
z|4yju37&+|YwLUW7>j4oYNTV2L$yyUYa4QqJ+#0U(;J$B%<bids;7-j$&1C#MK#Bj
zq>d4h{+2nB!QDJS_Q;TnDA>TK(GAvHM;v6`A-LUevNzVt?W&fZ(dfS_<%ru|NRd*E
zd%7eg$q4cbuj9V@W;(`Wwn`4I6|<Ko%pREevqk3yy}^$^tzYao-tV%51JxUNl#g51
z3FPy9WrCI)_(5q`o=bc<%g!LQ2{dH^i6@5E8_&k)eqw^Y$Ut?QXqFT<x+!Dx8p*qI
zQaEfUE8!<oA^zyFxIqUP_3*l*w{3hI5EQ%m6M0mz2KNmLIoLGTeq>-#NSZieAh#o2
z_b7o?lypK9A02v&eFOXezy<j3Il+5FROX*SP8{ReDE9VusAIz`CAj&<t@oNYm=7Pc
zhv^fW=2Q-mHihxU1iu{l#oVi|3H!2F=RTd?-QPo3_elUv*f)LyU|WbT$G?`{{6LPe
zyM?N%qyXuvVu`EmLwD@@cU>X*hoQBi{^_uCX)kdWaGvAInUzKhIR>KzDqqXyD8!3x
z!&4yZgIzYM9BTmtzyDTDqtKTdwopGEpEoSti=~HDq<C->{eIY(eQ%%sny*Bh_&1V!
zGE77!6Q)<O69Pa{(i+kRe1r+sRR%|p_nqmwTcl$zM7$YFvj=+aMh`LqB>n&|dcna~
z>D}iP9ZY!kN@Cl{Fd;v4(0`eMig1S1jR&6%`s*ei07P+ZjV1wiBv6%bZO7ub=m+Uk
zOc=T}h4#SiZbo~ebAhXARtJ+nMR%aWJcxOu2aUo|Bf5YmG%=XJ41Cx<!2ipc$}c`g
zdtrb(tL6ec)Z;LL%I>=Jpu!UdQ_?GT?A+*j;ItXjcBfu9Cb4k^c?Dkb>oVN02p%qe
z#&V4j3wN2P)uY{CrdMw|p#=xF3ck4=&h|ci6R5nl{rG*fNPHGtzzgAg^rd?jusZ+A
z&KVQmei6wRirm%wk2!nY7Y=VyYbe{WCvN{O6hY&Oc&Ptqi~gF&aeGpbz%Am0A}WE~
zZq|&u`Xm1RKO*p?vHSf@Epsf&kXPuvroZk*w=gojBGXqhSGZ++@o*lr3V645$GF@Z
zWB!W}7;rdZOUzls=eA8I)LI;p5$bRr{QFNHz>j>R&`@TVoB$}EcxWFTlBfHE#E%x*
zAEG15N-9FoYbW1TIHmAMR>~0?&jVgoiU}7LhMmYnsHG7Z-)$#Ubrt04Z!lsw;vZ23
z4xV<UNDr^e9p?efcDruE_#p&aRbF~-cl<?YkAyG4b@VZHFXO9?jIvjoA|;bNh)04`
zM{RN8=s-Y#JA2s21N2F-+`t{7Tf~6Mv4<nxN(FtsFBQ+buS!Nlz1-G>$6A?!EKuN9
zdh<M32!bO3-*%}o$ce)Ym<@w#%pymoJm%o4mtE`mt`D1HYuioI1O#lFaTjg=INp*Z
z0@||Dp);GpuO<fh%zs%(RSmv-`>M$)C>%T+6!QWw0LYuWJgAP8{0IoTFf6E%MQ~xz
z_#*JZZUBf#MG>|nJhOO0AODkj%=|^X7lmXM<2Goo{x2m3;5=9W8sM*I+7_x13H#{Q
zVsF@C%6`%4iky_syl0qe+QPxRF&a-q5pI%r4;D3HB8KT8Y943Z@=qQd5DbYf%iAVR
zaL3>=dR4tZbH|VY7ua8(Nv_XcOY;Ds)={JyU4Yx>G@-!|Ji~u3#g8)GCK(u3fCpT_
zQR?!+yRc*ar<us~|L0^u^J+oDNCDwsW}*U(i)KpNoVw}vH0Q%3m=2Nj|L_L%4>tk_
zp^gHVV8VliXL#)|d<#}*v;>vk8w!OS_T268rhOj8xPU|rdu@@8C=k43@!nowH&|Ok
z<m_U+Nf@8cppnmQ<sTr1C-ZU3<8w88Xj)fjS}XEaqmvns$@#6OAv*W<Qr~T55F0+i
zRB(?)1SHb>PBu39uBEpTHPt>K5@KIbkZuTXGbPx>F)t{jf3DuqIhzFcKsEBr6oSsN
z*b1{ku{S^Bd;2QV<^9tVKJQhP)W<Ivzbq(cFlYI7(2z$~4ALz{x2U_PVkEDs=Z${&
z4FSuR@Ma0eV(_83#^i2JE{~U!ufTws-hmMZaKoC?f4Wd`MDzbke@dIKq0IyC|3LnY
zuhNVS8Cz=xcJn_6?)<WOv6iX}jkk~CO;O)bfycx7IKpbK(s~}G74jKU05MC`gY>=t
z26Fz(K2srgd$3}VltHk>F3;XMnsX1IgC!S4mGd}y_7{w>F9c+>*9{R;UB>At)F1R8
z=G@p530$EZzTGMOB^ZR4AdZ{$*x-M2@k8sJ<XwC^25SKNi2Qu(YpMGj=rC*{bP7p>
zsytPC+68W5;mto+&{EW%xWjo>uobPx4czNGup{Gytp5Llk?)udjAumdGP_s(y%i?K
zhl)s%<M6bJqDQ4r^?=KWv&oAb|HwbUCxxRs%cI~IOwl)MG*6RCXp)^jqr2llAW{=`
z<fnLk%_`1!>Yy_;KC?%o-*N4f<Yx^CkN{=X@GY-bsVuDo-5YM`Y?bJsS+Hw>b)#NU
z3eVi?*<dhGLdavP{l(QHPR-K*?^mPb_E26CgO;3P?NVtPbo;^!t`2Qv24xmks`|1<
zWtgRPOJgeC#F|R*)FH7id-u(4Q!K@l|1&?qc<^lqOtKhuP0IY9qTDuPvX?#@&hN-X
zN=a`+JSiuzu=}c4Auk&SC?7iW2dk8>i09vC-rhR1qhWxTOWRGi3Cvc&Pn-|;_U>Y0
z7Th_b_pB?FY`(zHwN&Yy@wqZ@PR{j)(z)HM#<$Y(KjH7t3~DwqN<E0PAN(#SFn_|m
zvIq$rIrb8K``UM&8Qssvv4(-MVFsLrq{sZ>9K?>4f_G-yLWyAa2$0<7wkWRrB>R1P
z4?MeOvM%(vrA@g<6K7t2j1(nH&(W*$DZh9ZFypKwhlQ4&JFCP5QITXDEn+sEn!(FZ
z#!VZdki${Vo>nGP#`z*c#>1SFnaQ1>l$7N<)!#35SJ5ZrSA2l(Mh+YY``L<&Gx0(J
z$%#G9nPvqscG|&fZe)Z{yXt+UfxT(Viq_^k1m79hZ}wE_H>|3nerSKu)U>=umC0aU
zaa)XDt6ZQt!jh()lqwBt_EmQ#=A#9v%@X^5((9!6@$u=*>fpExSFqW#5!Vv)mm{76
z;{SYJ<o-ifI0AYK=A)VhI2E8bq6xh^&ZB=J^m-}rTIpAwT*y?J!p0*YCK&l9*H|p!
zI6dk7S#<w^nwnfHJ2yM3@oet-Gv<}~NRZW7{+i1-BX36-lNtF{o_UQ`G7{XaGEaH}
zJ-afITCSyi&D7nh5-e_5xZ)?SU%6=N$k0BVcPEWFWO6w9xNnr|X7s#=a0ktB>Y;e!
zvlCO(otRklXfL#U-=>ms1D|c`x;oKx;_&m{;k7zQh~Td0SbA=*FU+S^`$CP4!qZ|h
znM|VDy&%&M4E+C6`p|qHH<1D1pv@^?{AeOdbjF;MJ8-Xk%<Wn9YI_r%_K#oZO+TkO
z@HE0M5DYm{nO(=BcUG0D>b-S@xUR|_v%IznqPN9J2KXUppncrlk?nCO*M0tW3b*^2
z-=W8Ei1!}ivulTV9mc=oZaZ$#Q9#evxTR(iACz%f*F|%~#W``U7^{8jb;FVgp26$L
zR*>COR#h+$da9G#X~$NO?iG+CgJ;n!Au4cP1EiZqY%dT_WY|l!r$(m30|`E>I7Z5w
zXRv)SzV)P>)uP^BAy-P>7eKvRw36<oXwg+LwD{F%$DKx72Ji!>LHCd169hEW`NQVZ
z@2PLE4g&zVkLwQDA1!@rDc){?cDQV-3Po%uheWTla-%gvm>R~Gz8ej1t*iU9ZKC4}
zhe}Ci)~>z6qd=-n{(Vg;w#%6~&I-&mxdf6Imx-T5;SoyPyek7d<NCZ<o0;spt(B|@
zCV^)4VgTMP$1o>Xbg1X`VAkw?dU>pPZg_x?uw*x7`Ft?o$gz3hVO7Br#&KiGy{#D+
zKU;7o|7&;7llz2;C^Zz(<{LcYS7Z~$i$w+4(6}FoOEz_AO!SL!9EjhzL@RU}zDG1<
zp7O&}v`hA_oeIH_x0nts`>#igjpbyR&fZy-#TxGu-4g#W5riGxOIfoeyUfYyyj%rU
zHu@-7Qk3`jn6!ff)_|{3LVm}m-R8HmUAU-q-0OtDx5@F4*I7-`mTijA1Ta$I%B^Bn
z6<a+0WhD(=n0L}~mVM^1O0#fyijC+N@GVEJ6v9<9q9JMT<rUbDY3ixbm2&tZhF_T}
zof!H-w^t#TRJ>3V6M6(&>^UiM+;j|gL9`dKD6>PsyRMGJpwW1e-<dVof*pI2m6gDA
zc5VWJ3t@(;xn)ZNO2`7fbh)Ybbxl2$eRH4@Qx}s*?e;*SIWdMjA@NW$$owuc3$oq|
zFXii4Z7uI%7>30H_9ChvSFi?paG{_Y$5wcpGk2SuP1*pt;&k-Vt>h(JcV#%75ku6K
zb?ZH)xn8og<&u(oC3MfX!<wm*59t<CH1gfPGPY1a+f!p?W=0_C^Ay2Rf(80~8wGAp
zYOn2^59j#M8v(%yGw9QBtAMEnHq^ny->Y~A%o1Y!M>Dy$x|Ty7%^k#?yBq9hyJc&I
zc(pKj=ll--t>~*v&3pL|g}BFN)2;$2K^j`59Vl;4o^#T*5x5UKepY>JaiDd$ZAw9~
z;op*!Sk7JI^m6>9n$vwU291>avtN)wj|)`egYhmd%Fv4;6M>ihIlL?vmm?omKDKDF
zZ(wg%PTUwfQ38?fCo%Yl;@2PT3C_wr4D|AL!&PV6LQiHp;T-sKLV{sZ99w29U?-<U
z{go-U-eKvuh19Vu_bj0&&FZ>VfnQQ$oQ-jHW+dxUFpTT`Ffz0C&tw61Q`!ipJkJP$
z9RcK5FZ%KKr$^~Mnl3qp5kGAC4OI6s9JDV_q+EiO^=~!WcK0fRbYe5tUsC<KZ!N&2
zrxgwvrQQIWg7`+FoPvPY&o4Yw@j~>8q{OLz6VOx=NF0@R9=T?E=OX0(?BiaN4F2Zn
z55}eMdxIx_c6}VvgKy~If02?|_YJLeh7cYqjGe{jP=z$Nf=K7)<Fk&Td66&2ovbHj
zE2buD3VLySwAnGxWIA=ZCkGP`TpKO#o7+3LVd8k3oY*XCw8jWik-?_Kj|k3CkmNlJ
zhhWqeFSXgVUv@(k@!0!b9wNkRf){zGg`b=$(GcK%Kp{KN-HH1FbfN!{3pK9|HnPV1
zSMOL7=HurFrb2Cv!Sa&0k+?wtq45AHwYbQ-FrK3`p=jn2DvF)UO^dZ(5pI{nIHw?)
zg0rtraIZFAE=KWud-{99V08}_=`iEoFc_u=r{5s6{4V{$LgY?EeY1m|#eQrp_&bt8
zQz+sH_Qt;se&NFs4^Zvjkb#x3Fxy5~=C-mJ@=tXAKYT568YBe-$a)ow&mhyB*<HGk
zZ|FquI*rgX^EPoR0q+Eo>0y{IT<3QaegHza3nf*vU)R@TBmZTP{J#`duqmFsSLZSy
z967J2zh@he*)EeRZc#<Nq%w5oO1&*E4>J`8XWOrs%FNagY9ZR9l&Owi9wR(76rZ&g
zPVka-mCMi>k@lmWlS4xz?YWU$?-}gredW=yMD&O+u!k3ukKjJtU|6ktk7Dg4>xCv&
z*<j-n{yfRK)r9xwn2GaP1>fH?WmL2$)2}!PIA<N7p4g~r$~!;9v=Jc4hM9AQhl}E<
zdFRlQ8kEN{y!~8IS|(6M%tFW?ABP(BYf#R39)>p#XdQ^ybkXZ)T%2PYwvdn!*FYo~
z+(Rd=6=L`0$sb>ou9eSInopy~aQN+UdQB}+?}?}>M<T!E;a72xJnCG5_a)N3+&Ph9
zk|Ya}YUrzIj(_VU%5ctgK_J%d8u3}av7Y<1H&eFjzBqi+Yu!4#F%&xFGpDq6wCp%$
z;ColHT6Mg6dLiX+41kw;SorhoP+%R0Z*#V{1w^W|Na(|b50=>LU2ol*=3OhXQY(V(
z>v*=wW_I{4;BwF%oDeb>{yP><gzgXN6Xo~-V0EtN<yyS~3|@PC*{<I{auyk_v?Wui
z0=B>vt&C`YqwKEV61=|7Hx`U0>^LwHiWI)_uN5PV9MSycjf^o`{l7x}|JKr(kGSl+
zlOAa^5b&))dm#Rgj=82ukkGkMm>=|Or|T5xA#q#q@;ghfc?33B3IcW)iE54O(x@m(
zrMznD?;MS#yUB%m?dsyv+Xzg}{Di_%N{IqJO;}4G+pEbbiqRC7q_m8DVYQsZa(QLo
z)L2M^51k4jK=4#2Zy5$Nk8Xs+PWt!Xj~JGCOA>KzW|_rd5sn(nk7g|^8gV}VQ3?jO
zU)O-jLhRHgYflPl&gA1R|JvFV)poj~NfP32DndqU3j7ZY6s>_VY+6E1b^OrY&$cFT
za&j!M1K@X-H&+)1e2jWN7H#s;a*D=>#){Ul(v^rsd09G**a?hr6f?6nkIG726;(~}
zt={f#w2yks>V#C9JUBb^O)B1qimMUSR>6=W$^X5|zKFq{=1S9v9mQ=&^{}o61&SQ9
zk@6W-4gEU3|J`<(;u?lM9=P<W6Y%1Miz0{Ci$rcs=z=~Mc(oA#-*>MScNAzF;br6W
z3T{8Y++a5xgxmbeME2037t{kcj#}XCPX?T1R<Qfuf6VazM8%_?^%Ki~fOTG-i1#eT
zUK~AwAOd({19r&dNm4hkP$SOU-G~N02cEBgJ4!|76Cf&t5aG~-|966)i*J2HM*`cX
z@Sv$!5H)e<p-Kg-G;L5>*0kzBKn+@@vqJF{4K3Y^#LJg!r%+$BXy2qzM<!DoJPB%s
zrWpXE_>|7+iM}gJOkS1U*7s_OXX?)HA!3DCW%c9A6PR>`bEMfKW`H5~o~%xfAxRlR
z&4L+2C75meSQ#fJ6BN4;ukm^4F<I(vyEwfJ{kSHtSYKG#7njB)O~(BLt-1{|d~e3O
z4%-ojiBG=9#Z~mZx}^-^#XLA-<m7bUqw-ADH9X#`2%@twuxny64nOeo8A8y>@SkIz
z%D}0HgZ{TQya(JiZ=>;b?3Hk(sx7!>u!z{6Q4S_ek70fB;b`)04kPH=zdLmr|9+XM
zyue1Zn-nGhMa#DhXJ^NbFVZMNP-nb&F=+uJX*>e_dc>7;>x?=|q}k;NQz@S~q;6Sl
zQJ=CZp#tc|mTl^}g^>VzUOY69)dQHWBPXG}!G=Z)31NnMs?%IMYG3sQB+is+Q7W#x
z!+ff5RjLU|BXAc=7=q2~(^HaF=f@GE4WslHlDV$mz8ujcyQW5SNDf&*#adC&S{kvE
zEYb8iL^~(L+QJ%O$y>;Rk(i-TdF`iB?VgK9Yi*V6LfH*BA`MM6QRF<rcK;x^WTuQ-
zW3&AU<!6vdx$(vr^3<-s&u?EojZRLjl&KDH)05bV!5mqNLJ3Vnm&ee!rlKvVVz<cp
zOdET+kF>>{IL#(u3P&fnjN+xM7TR)aL))QRo3||<CJFYrx+3pf{zFUNxGUeAVCbhl
z8$GSDRKod^jkAiXJ6o!G3fd)Y*okumWvwgCTS=Rgv`koW3}tAZS}6Mu4F9Uqttq)@
zUs^zjZ7~zL?$q_!0fQDH`XcB9LvWgt@mww&D%fVB*%3wG$|U#VJKVtQD0HeQO6my9
z7)H_lCZy&eewtQEElC>9t$!I6t8j?fDYEnxbI^a5E(BtYZf6ywYhYrIau^Vh)&h$t
z6@YWfR4Q2z*rDapymt5vOfYr>2wTnFnVjL8g{1M`_!_;KIg_8Fu+_YLh11+lt_aR$
zy)T<`HeJI?y(#}2RbzQ<VTvdXE6yXC8?o)xmOt#gPg|Ve-cEJe_I}7x<KAWRO$BT!
zRjDLKW<U|&Vzn?$8wH6z<IopGqv+4cv@!D*hGdi{PV)ZgC3MGYZQMK(U$i_Oy%9Kd
z?KG_3+I&=4_pOge1I6{tIOI=u_5lcC`fANiL}%JRRWf0Q%v_$n|3uYY@<unV`Ktao
zv*|=`n3O+Nif^TsPh@BolU@GDU><RWx4q&5JyLK>jR&Dz%7J>FR@;V3`8|vG35$K1
z29&<jC;1&HrfZb2l2OcLTIt0IhC9_3bL8*#9wiajXllk3X`ALS(D!pNq1<dUm5k9R
z8=5ENqX)vrJ)>*Ipz_jb_U@l23mD35rLOwX^CO?=<=1o48o2ZcknUBKktMgO&+F7A
z>ZY+A!hD^r)rOwj4!@?TU{B8|NGDgklcU_R!&e-@h<J-e0qQuT1<{`(hz3V=1yKP;
zsl&vxd|g6kM>2n<b@q14%<H&gh;0|9@pc7LH0@=jlDN8@&08yQf^#dmY}50oqBNH>
z(aXPNWgS~`BGU}NWHTN((`(VXEtKDk6=m(IX+xL5jP$eF%!bRlK%wZZlTr+8LMjS8
z7(ouoEXxPFm%~mpqh4|=lrL-+=~{Z0n}C$CrEiHTJAoC8XpyCtZ*`1}ahgSG^mtk&
zS@g6<S^8AzQ>PxXN{ZzaGz#%_KorGNwci_LslBcZOyJe`iU`h;G`M>P$@$eRQc^BA
zFOa~Aw5#Ar+_5c%7k-*))Js_8m`l0x&!t;5^`D<l<6y%}g}Sa4l&I5Mv)pvQGV5=}
zw3ji|axA;=`uD7bA(>jn`<g3T_E5ex1));mw})?u5;yX(m3Vo4=y1J_QjKK7VVj!3
zymi!F8e)@>SJ$WO2mE}OxX$5LIvF;OsrJ%4%QMtkwD+&8a+jGf{8D3gT3x2zTH2*p
zoq?tOyZVi5NN(40jrYbnNhYUSlfIO9H1(84x)7ilX@AZSrAh+s7<}H9D(AK#k6_xC
zkg*IVQw2I?Ho3l0?bEyQV)XKAcuR}X*vfl~GE&gffQEt84!ueWqU72>2AWfl!m2)A
zt+cs*AaXhOP2xgpE;%PhZoEeYmxtBCHdxmI{3IThq6iRN76R#a;n4(dj`3bi_R)6n
z!ORr{Q-oAB<3rQFkx$mNQuT<d`&%TNm3v{z-9=jo!`0GEv13QjrdGlpua*9)SnMaI
zmmesXah4v!XE&7&v>YYo=S27iNCkL<ce>6mCC-DG8OGqhNg%ikLjY>AKx)eu+=2YO
z^Xh6Q<5gg~{#NMkuQ6%yB%cXzv5&)@v}Y+W5TvbCy%(&+HRi7uu?MBqCzhj-?3kLe
zgc#IWQm@gG4ydQ@PMFj)V$)%;_UNYp3sQbX8;$&Z;#mo#u(k;^mDJa_CD(PF0{6>;
z8tg8GU@s+YEPD}V2Kl&vU_BbW1O8#AWZG~A7JDsn6SE@AJ~?&!a%G<{OJC@Jaj>S2
zrt2^#AJ9jKdEt7leWy`ZCvHikyde2|B3)EvA8ZH|GFtB}5yX6e0^W;L`4aL&>PVjg
zRV7u^7n?drigm&4$C<@R$1*%=rCRo-+hb_#wv27c8LC(>dJ&5lU&>L<<k81zD?25A
zYHHHEd~z&ZR##hkp1=o+iR5Mo^0^5v>b#j`!o=xC)^s1$$;3A9%pH)M&!^+#j^T2%
z=oKVCIA^9af1zI%p{cU>q3M;QlcGU7#T;)ehk~+ib0wFgV5rf^NKz2kZ(@o%xx<3d
zlgRQS4mM)t1uCQPQVhDz&sv9h_Xl7zr?Cm?CO-M(lDJi(w!l>*eMi;(0WxOfEOA*C
zJ~!M&aVeLdHWGJ1`mdA|WLONzQHAR3n|Uz*>~$Uz*)ObH=1)2tV1z+-oDNY12pMua
z;z0f=Wll7SyqGrZNth3i$;1)7nOiL0PVybB*9&j(LyaI`73(|yK8>#-F-C5Z=F@Jt
zu3-NC{ucTE>%5r6pfm2mW@DZ4-Pa1A{M5^e#Z&CRb{Rs!?puE07nDLTKa6WL8%GNY
zzm)-{qHnY?rj(CBug`86JA-+TT97h2NEVUJ**2Tz?}HD%gyz>bV7$?D8hz2Ja9{Oc
zLCM|Muh)M%qsdV}GRA(6c`@Ca(_g&%Z3NC4U<mdbKPHT<3}5zrRP))rJa)ZAs4iY7
zDdz<9zUORm-#xIa`gr|$m>hWUH4r*y1!jM3*Nf{kbn0o{3*Lot#`bp|zgpHb#z1}Z
z+!`Z)Zl9xgE^);Gv<L=><Y_;4yNsP3DE<<tj9Gy~eH~MRWG`eoWLiV@yKvdR!x1b;
zjdrahRKuj?TXne}8M+Qij|pBt9r6PJTkIVsO!WjOX+yOlK0tl%*w%V}b=aMLA#nAs
z$aynzglUt5zo_ACmsvzYqAr=YTf3phya!;O6QehR=>`L;<2I5VHPIbh$Ib^8k1W=;
z&_j`K)-AYyY)uf582YEqaad3$SdyPV+e<^Oo}b&{)$6nYhOb`ttE&gRC37ZRNhLzO
z9-KLDhw+Rjes`}!8d1u(iK##R^?a--S1sC2lreG99Mx)JE%LkBS)<ZE+*x{GR_q;A
zOectEGm}3}y57)u=%CV?Pbi2dvth}x#Kum6#F-KNcjZmt=3spBVm!f-ddDx@CP=0R
z+!dSVcLS>PA~vB<J}kdirIVr@kvz(3b^Fo>kvuLk#&ZZ42w(`&PhYrY@NeL%bx5kT
zu$x>z*D!vO%ui7qrmCns;d4*9RLWNRtXLXr8DrHMvGLZ}x#>8_7FtUKzA9I#m!nR^
zzu$==v=sM6;WolFGADn<5XQ-j;m~F%Jw-~+SoJ|-v(t#4h|wt1($=CdvP<f+FT$fk
z%ynr6K}S(SbmQXn=zYBC%#_B0S;R8K;m8bWqj~o~aetQKM-iXr2t%g``!33vvHy1(
zYPM~-esou+r%J7(m{V4}R&1GxMl~VW&LEY>`MW$7?E;RPuML43TQ|;zqV9rzrbJ4X
z(pV^^4i0uXn+lc^1aMoUXh=Yin3bIf25t`zoAY53QHV*NWezj*Pej!r(B{G-9<rJ8
zq5g@Syj4}dy7H01XiK~yoBAnG`Ro%lY4c@wdAarmE}L3ICJA*OdjdMnIBArxHYHm9
z(NG@)ouK^Wt>l)0vUSP)cmaLNlI;><m$nfUh6yvQz*Ymd>rj)HtUxl4h_amsYr>bQ
zv5=L%p{u38jM*a&saI=oMKw77w1F4<Lq#)IN;NuE5qF#+GhaU?Ig7jk7fGH-MqY<z
zN*%}GRNZ?KeI`CSgi6E!ZQ=_b7H0)H0z~OGlkc08A5J;W_-0X5nCxdc@@Utntm?6m
z$-7FjKI10T!PUg`xR;?BF|V@Ezco35_w&_C^qckU<;>-=qksG^ZJY@Y`j>k*3Bm`f
z2Wg&g-(~y_c|)w6`ljGXIJSnxKCuYcjwouD@q4R95}5iDYd9o&#*J-F4eX1_Sg0Pv
zxzBc4glbo^qOm*%3kQyVHh-7~=4`2a`n^lGhek>gvLqa<nsJeJHUA@cyA+-oMoFmf
z=3;AoB+dD9HRiwmBHnyhV*}zU-W8=fu2A~5A1?YV0`XSeas2tb9*>5i2Oh1z5g{ay
zZ#lLQc^a>+w^9!O1EA$`R3ueUMAy&%5!_OnV_bKKBu~tlWb<M)>J=!zdiMTxJtPvL
z{y|n!9T^ua8F;-Ii~tHS%XD2wB`ld~oUMIY)lc{MHr-g0bfC(ol~C-2hCyT!FzgAK
zeB5wtyL81}WJ)$%)%W3PGn1XU#tHHM>UX(Ij@PmIf&hGHqDs`|Kq$XAx*el+`fP$y
z?EwW*SP(BnxW>R}gcpqW`vEvcMOds)<%5V+vxJU~*lj)vP6PPEhu%CI^}#i!lE(TP
za4hGXGlPyKn(%#|pfyXLdxT$Jo7?<{lrU{69_WqVO4>b-8shym17U9lZ^n-jP)OqN
zih7Axk<xqm*A+U$&nd2Uw0*9xyXHj#u7``hR;a_D6+fT=WMwKp!z^xC!ZZ6FmGyW(
zu-n{9!Dx`)UJlgg!ngcML6=y|b=Wr4g6zb1Ms>f-lD?@!cHc8WSV<yubB*PLgHfID
zvi&%?K)~?E;Do@Q5$EM-%zw7(nR!KT3Ef^(|GWE~Q=zf+;NPa<^d~c?^Qfl4cQuW#
z67tO<?O#_qu`gpf8rDn~1x)dx;^~G<ju8NeCj77Z6Gw$>)}08eCaqr(MTxgP$-B$;
zK~+QaLljM^XY`kyZ7td0COb_-s6|%N_&EiP&RRdT01?hRc5G_`+v&i)4ZPVfGLe6P
z>b3|aKhM)8nVHD7YJ=6Gk`Xpznr&NXI>l9`mIPgB>~sLLX=6kxc<J@h4Fdx??n3y%
zxis*L<KJhHknbVajb_0V;y-}y?(?aRVHZo$UF4o(&OvpZMaK@Klx;kV1A`T>+<fW`
zCs%Ww-7u;_J`aJwh7Q{9G=VsID625P;zZIh2Zx<3`o0yd%xJm%=WBEMa@1cxbnQNv
z^>We2aJ{H&NY82c#H@FN7{A-)AaULpj{NM8%D&Zk!C@e!Jyntw)Y0B;-4vLK)){RM
zo6Y$$qaponl_U;O_>wl6+JiVj3GMUcvzkId*j-O7zB~G!0Ghzv#@i?<^P^WC17kW3
zWOW1!C9*`dN2!3>_N<Z-nE3VugwX{ltMs&}La?t}z>+RMWuah~R<=d_pkcJPLzJTz
z5V!7(Q2soGnXRci8uL<XUjAL;bW=O8+KN6b6^kZz*Trfl<R=G$lmw+5G%msSFEudQ
zHQ8;FBx;cXGPS?jzU&19pIFwn8EST+7~}Y3`6o2b-FAP#ouP=3)KApla$P0qANML0
z-Ou1Q(Wm7}WG#oU;IVB(#HO^gSgLaQmE~0?@(SqaezGG`TU$=;A`l-0hao5vewgp~
z(d69?sO9i$r~~*&))Dp(a^hU!BYI{fk!}3zx`wD6OjLCCs`P&lJM1X84e*AybFMPK
z=@w+gf7;*DdDQltN9Dx*yp{sea}QdavG$zV+qk?AET`&)BglSsy*X?SMu0$Dowt5k
zdoTe=S)YCt4$g_LMk4bJxRN^RZc#T9EdJ15gQTB~cquU`vAHqM_1l?uht6s29GvPg
z2k7S11>cIQz~cTVK}D-0zheD`CR717v;41duS)mx(|=N1AQEYj+=p_*4q#FS9+I6C
zGCv&K%+6!@HVp_aGX1kxvoK(%>N2^qbHMM{cdECFeC*3;<XrgP=<oU^IeuR9$JH#6
z0VrB8k1y_8<~1?)KW>FJf_w%b#Z2-|xcjd2K370WyccgCmxzi=e9_f344<%A8m5K~
zmgP15whS1K*#ssbH7QtR$~<q{sgyQZ+-8M?Ui<21j)eH>@700XW%PAiEbOjhB?HPL
z)hQu_#GF+}l&W_^>@&wj%t{JRP$M~{6BW>R*D8ycU6LV)zjGw)dcUQMobV5o3DGfE
z-o|tyYI-aj+_$Fg_JGU_nAEH}>8x(tSS2z(t16RFq2$Gj;-|o~Ov&>@Z$rdR0<*2+
z`LM$xV|9ZmXk!t%M#FYpP?U7T&z)RLvCl9x>r3bDdvcO>7WK;$1!Nc_>57J;fR3ey
z%G9lL%}Hbt*p^BP#li^ViJN+>Hc|1G`=~J`3Ud<*vhF}}5_TN4%%RvP%k@j*5m_`c
zDlncA$#GK^<4aAjJd%?Z&^mRQ0ui%P)hfB(mu)(e3COmkYYz+KV{XOwIGf9<J*5%l
ziBY@vsz!?vTV5=3r<)**@eY6kf@AKNKvou}=2I}@D$`I0`I`4eAKpxtTD^TABwqHN
z;Ir9&MR+EHJhyZ7o}T{4&ypYHRw=kE-~LC7#<Sz)yfZ<yg-i~UIgn=eJ6G3711a`g
z_SX`gC~GvxFt(fp<o*GaSRwjv(r1^^4KgsV_?U%5xTqQSY!QrFWyf&a+d_|<e<?(h
zNo85Ajsy5U<$Yw^ybS&WObSkT9-(*8j=T$x{FNPxkcm?V52@GgViqjExJN}3#9;rG
zEGTNzHqRK0)lsfKFV0SyQ)tjgqeoAT!6BljggumkmvU!^v!*wkYT9DkeB3ujB-fbD
zCQP<deV%86CjJUNc!r2(7)b?R6aB=(lzLoSOWR>_!sA=J%4MWH7W8F?CqAXBO+b&s
zKDrqD!w^(JLVlvfQZaG-l|C&wbtMnCAQJ~Va@b~hGKwvYSCG|z6VQ!|zxf%Ni%r#h
zxh8q2Jk68~hVvMouVm?3$lXi{&$b3lc88S8mksen$i8?0VVcI}UWUFTjhTv-TM(92
z^K%o#_tW-Mv*yH0i`c4a_ekU9NYiBIk!OpLBR1RKmR@G|b23xJMoyTEdku9Q!i7ri
zeK!@K5_2Q}j=1l8X9PE_BRi(R!gWHzDY=`1`y>;=FJ5CrTK#$W9;3_?gIdV|#V&h-
zk)6OV5~sO5K-pBi$cG5li%(6d3P)I}=tHysH<KV=5feL2E=3aG8+05dMfpiHomIj$
zkjV$S5`F)R%T`tazKcC^wH)7Wx1?dCU2K;|^kjeRN9<_{XleHmn|{(GXZQeZ<kYR4
zNUbdCFB5ZsNJQ=U0W&3MieYIrBBhjxoC@tmYHWhm7ks6>Tv1qYMjD<|g8ozqv%$5w
zP+Z`;EYAcb2vaiK1(41cy_g<|>}-JyllgNLPxf9G+-5S21JvDxE?$Y2*4QSZJDH46
zR-JT75o4+P-uErNlzKaJSw6NXDI-2wjbNN<Wz;$v3*07nMY~NGNrNmuS(xc{1#oU0
zfUh~zZ^>gD4@PfO{#s)??Xq~<y%)`7Vg^P){Eo9i!{8XsqCh`4!Y1QlDpw^GBIJGI
zvDUHx98$o@pHb+6UEp*V$F={2==u!-jIgP7MC^k|oo#-u4Bz}>T|&4gnX&dd==qVp
z*IzPM3M4#3UfOKLS;^E=&SV+KwXO0nQlL#1sF0QViN*NE`v{5JjGZ%)Dvc~QRyPPC
zZ1KbUSS%DhaLk+?5f&UOQ{{vvus?1V2~Q5PrY4rbzdr3+i#Vse>oOZ83>`uh9R;HA
znqH(};!)9;1!$=urPho=F>_I|BTS@XFUZ@u7)2m1ek#@`F5{sxoc)qHDNw+iB$!*#
zG-D=Aeb3&9KJQ%skZk8WGTL350AU^$ro4&njG-Imi}iB)*3Fg-#tUEd8oT0Y39wfm
zV(p(m61ig;HuR3aabVorj)*Ebh(}3b+J|e~HmAqtr)Oo&Ez?mWDrt?E4f-XEd4EB&
zZPKBq;)88QeLuEmVCSw2u$>*rfwphe-RqU}lH?W}z~i!^1Qo(lxXh>@JL+fN+3E;U
zQadyWOAoixx+<rQEC&s9YtqKYQ&zUwor0Aakj{s+7-Ew&M`Y2XU14aY;^lGuDGiO-
zSrJzcsjZF<i>GV?4)G&5h`I|cVc=nK1s2LWxW?5(DKtVkleveN3z9F%<nqx^;aC-~
z&_q?RZHwVusuma{{EZtZYAevFhhOs5^<dYN6Qk&^C=efWYb#)N=+NV(wzkc`B!+Te
zjg1uV)0N0sxZ+-!DQ!~gWEiq<Yr@sr4uE#l_w7tbVO@Y+Y()z?QM&?jAOic7?~;l{
z=G2{T*_w=}!YEaXeAy+HWwq|;87kz96ZfyF&Shl9Q{^;Ps`rRe_rm<2^U@MUIT(dS
z8C+(AB&-M`F#SzDbn5v`;}E{lyv4`cq2D@Fq18miC+OG11rT#9eA0r;C_kf<N!Bh@
zpVZWe<_Iqy&8<HPP%r*+ieDPmom*}_A?-D~@Etn~<Agm*^NSt9-YRV$Mn*Ud!7;0x
zIg_E_NUuJ>`TVw=zwMp<HfovMx5|372HqFucNMl!<I)#pEZw_mWO*?c!+!u7hOh87
zLt$La!_nWII~8`JQLQx?FizZ$?Frnvlt@2mAac;5F<?v%DYH#Xs7$Hwox6KyaUzCk
z+`?c14B5d9*>J?(04D%|l*M&0Guy7rZpM{FWl)A|UYI;Bf<TvDcl*#;2geSHE2;+M
z{X=qDD^I2n*26z*e}EsfYS(#Vu6y#ihdGL=MqeYGp%Y2lMpa|D0s{?#YB9BZp#2a-
zcpG4+_!aA5=)r7w%gBU)4`K}_Us^guphLf7>Oi8Cd=E%Iej5g1Y}GIQ4botA7O7zW
z_E%3cJ&VS2c4F4vlebbHzj_z@yS`8e=C9uzst;znTuFwOVWsDHVW}ZuS>NWrSjxVi
zL~=TF3i6{3Kt<ge2o`c1UBRGa;+fMM8E746D<GaF1Cu9L<2YeCWXuaRb6M11u;tQI
zBVBGr)g&SfNEuLOQ4OPYJ|z=yEj5*X#w2xg;1sy?GDq>zWr1lBPQt(}3LaVxu!kph
zrCOHk`eKn~zD5k!YIfd^!#D8d#K`7EN1Ig}abnVEf!s(bB@A|A3}sWZL=LN!U0N=U
zl`JGStW)FmRxK67CTfE@D?j+H70>_5RN8NXSCYgkQ!>+awv|BRjrGC21(iQC*HoI4
zIzZ5R*Ox4OHg%HuJJuT`y16Vu)Ir_nQm(EfGhlsj6^XdZuWB_|c1l(A;=1rO3Q6?^
z>rs%cTb23*fyV;<%?S(`21fU$<%*_^dIa0T+pZ?rUy{v!5zMPq^iey3T_xl?CITJI
zXWYC<Inu-}P;r36Ev4U3MI)D5^?xz<Rt<4PTeoQA)<_2rjVA>6;O-FI0t9zy+$DtI
zuEAY`1P$))jk~*h;{-x(@BN)~?!EuuJaj$OL(N*%V^xi{=A2{LOK9U0H7hEY>W_fr
zpiBhs-(79@hHC1&d;@lmN;QJ%mF1v{y(unau{~%Ki1OT2!ju$$yj!KHzd^r$U>34|
zS!58rr&`f46fK#ybQL!}B*EeI`}=%SHSjyY+_dz6QqiSXhHJ?7yF(4(`QzS?ur{%c
zexn?)dIv;SYYPP36>P$(t$nsjd3GS*c0%o!n$Cts0#wS6#zVbPP>~Y1@`&rOmH;dx
zs{s~2|1<@-SYzOza}Pj4V_bh4P)X*Uu=Z1eiqnn0z?EkWyAhI`QC5v-K?60q4#x@H
z*s4n|#f<jV!Y;eD#8d7yam>Mb6={0ORWVhh2-QahwLF`C^`P0KF`p;ww9G;YufOd>
zL?hjhm^%vC;jq`AK4G2w&}NDef}v1;ccmP&Nf84>gh+?0sFJUXf!R5Zb~$DxagKC!
zm!j9X$jb*0-`Kz%iZcHIy0tG|Zg>ax=YO+M!sp;iyUy%~UjUpu+QC&_JAcYJk(p>W
zehcM}=Yg>B*S3FvaacJaUhcj~PGP%-)u1*uzw;jNoR~~^zXKYxEL^1)zlTRM<$O1F
zSw{-l6Qy>0uy>%CrqkPL2z3Vwt^C*OC!FB`JL6`3G}VRP%cD9#m+|LHyWW6B0c>|a
zUF}Ml{L?}^1uMh%eZRs1D8a4`!AOGWfXy!<p--0}V1Seuq7O5P^OhgK;1k#OuK+ih
zKNoN_PsM%!7n!?Db8;w!o2{{bPC41ds)t#QNlv~A0QwEk93DnA5(UaZxcW%-=PO6S
zsh)Qw?Zv9o1ynF(W7fkrm{qV}&vM>__3^tRRy;g4^mZyXyalva*KFf^fn_(ig1)hQ
zoT+0lk>e;{c5{+HUVp>4ZJ2RFtag<vHa`artxaWCM#nxl;h|wPFl6u+=Cq&yCLm7t
ze!z+vGvZ`s3O{bg4Kx!L#L*_0%c73LUc#RRAhODr__MnOOEAQ9fO8sXd#vA3J2-|`
zr+CexmJkkMfdv+W2kAePMC8@|0Bl%R(@vq^nJ-XbEF5C$>9T&ucp?r=7^B&d8&iBY
zKZ8Ftu1$|oc>Ck`PIGkD@*Q(q-c4=m23}9Ai3gh^%>Il)JFlGM7nFbKJj8ofzn<y)
zwwvz)1NhxeRgi4ik%hfPb#r=d)r?&o1Jb)5`CwWt0~U3#V%ezN@R&(_PNrf(+y_bq
zC3#YioHRRYd}20+ERA$ZX7||U`iE<DPrE9e>0c;J-x0L2(fx3K^t3#&u4>zUjgoY2
zAPAw;-fmyWtl0C%%vM>DqFFKDC;h;upsxsdDRMi0Iq@hORvHY2xa`;{<R}W<j-(||
z(ie{!O9qgJQURRAe6nKmAGfm!Ag<HT&IiyaGh@|pY<c!#B?(*ZbaF^fk53>9uK`=c
z_$27mi^)Yc%2r*?DJ@ADqoBU_k14o1K9fF&LsjD#L-+HWIu5$`f!qdxh97!4O(BeY
zUP<b%dBaXL6}n6O7e76S!sAWHXL;xNQVgbs)<YF6HK<kS8O5!DA;(Y4*WN8!x__<<
zb0+Ac$37K74@aZI<&)$Sz3_0XO0a2jCipaA%AUW*oDXh?kdk<Wbm*xe?7P(^^Bx9c
z0RS((KNr63Qf|%I4-Eukcw2~AbEk#RbE))_SI&KL`hGZAEP)R*iyL6cG<_+=9l-$Z
zIkZn4i2Ct1L0T^y)rpx7Wcn<np?N$T0(-)Jl{X@~rPRd~xn{Xhs&h<`8#h7UXRmta
z(rG&<=iPDTZWyjg%db#ujN?trxak@p$&?D+bpf}_+R9qq#b(*K*o}MoA<1bTS+*>m
z(>(#NnKP|s>Dh}CtALKpTMn7xuMG3jZO!MnZB1W{v}wt)Tt4Aah!*TSHk_O+YEuR0
zD!*0+LT;eRwPXCg$ZXEUQV_sr&(D`^4$(^Z<}BH5{VegZCSS!^zM=_TP|*zJW8CpS
zY7m37n?t_Xqy?P%qI9FFEqdX!1Q`7;4=eSv$yG)5|8;oxk;E*9GM0P%vtqS#L-0oc
zXkpv=sj=fvr878?ndhA`Mm13IxfIRIWeq7gjJ=^bz;M?eK*=XZ?Wb>&(sDiUMM-9%
zq1vBJcn9tZ3~06_!Zx8Y6}#-xn)94CCHAQ_9{sDgzRrnt!0*YfU1gU5e+N!$i2=J^
zYwYY@K+jaz0Io@)Dxxvz^1q^nj}?tS1|GK!QHElikwcC?%d*`l8m_8D;NWOut9)XM
zckGG|NzZ8U6assBsp*r+4E0-fgt|faf1AHr+vp3HJ?K^qc!q%{i$zKY^2B@Kxi=>c
z=4cHsTTo=P)d)#pT&IDP4CDXQA{PEf96ywb>sZtDkU(YYBn}dX1?x17vm?sic-t6b
zBwElvFijGLwG!KU3EN`s*>sn_&KJSyUja26iI%ANdn1`ohk@?<J5+eeMu>CxoO?ft
z7kWJ&;|&DHsK;?jn?ACS?I7@}5FFo0*pj~~mO*!gX}1Yz&}#sdtcrggV1hqzaK9o2
zKi>o~kscZgnx@N9BMiU8HxN6NDHQo`oj^eI27Mh_Fc2%zx@Rcrk^X47S4q4_gmPXp
z@n;!mwD4w4Y!S1X-ZVT-Ae|vNw+u>}%)svl9V+T~bs&>d81w%9gOBs~-g@l*ehHnv
zm-H&9@X>Z8W?inxi!&($>1HGi0wOblt=orr@7J)Sv18&w?q6M7dciwQSaKC8S`8mv
z_+=^x+;siw2O2nKJ|`7^#LP)bgbFoL6>=rAn^NS0KFx70)9ya9;{xY6F|L<~(B+fv
z8EP&Osmp^v<#S2I`#U(*f%y?2Vh%pRs;Lyqm!%zw{O_LbqqNgjUodLorn5dPF)0a@
zEL|S;OOFuE=%ort<msz66lo~<zE@!!VE=+I6`G<Q!(HBQ@%|^>iK#aOO0Tva{-At|
zr=l46TNXaAq9*gFo4v0L-6utgieghz6Jy&ERFQMP=Oczt)|pv82z)vqxTtVm)?p7r
zua)NWlQWWb@>XPoAou)UjtKNO|LIa$-q2(0P?M2H!^S4W6fgT8gn`(cfSb$7CnV^v
zprD{j@MR2Od|RpG%(f?|ix+N?)olA{&T!VYx7%Zfe&-0=%(Pd;s_PVD8p=(AYtrjx
z0|3pAxD+BOtmY)SfFYdgyDJ3Us5~I4r5gb0%X$<b2{cl=ynj~PVt2UIAG4Cf#6)Dm
z9toZhNiK1hVf|mMC|Y*%ooOt)(^0>3m8hm^=G5t)T4W#&3Q2JEuug8nWi=kc_U6gj
zz-L0O|DTF{FYZ}8Qe`lja;2|lSY=+f3;BaY)JW@EO?<bk9D%-pCMZ}k=*3>+twz|H
zgcg6vhP$AcGpQUdVeRAB(mSz1s;5?&A6n)eQanYdQ12lVx|yg%TR$cw7tRJ`(nWVH
zk}s$OSIEx&p#ygw`MR|IVSI?YDu_cVq{+VpT#rA>lylnwT)2CH)upi6uS_HihPgcw
zbg^16dDQ_~EDYD52^HM82|@cKre3Hsv#6sO-2tiwY;VE3Bh9Lt&WELW5i8L;Tei5&
zNU#x0aQCY^lVU&>&Sc@v?NHa_+Q5ksn&uQv7VDE_KZ;V6P(<$n`G{FhbVqe^qJtio
z;96ctd9R;gQNyB`YoOSj#qvrLp5N>eRm++9)aRlMJmiJAm%KghX5VR?6dO+;MqP~4
zF72_t;t1VUq}@ggHi60bGMDNV%+!S@vnag-3H69;+hRsv6lo>ZUNWC11lP<IG0k)R
z^w$IPZK4gPT*q&VdO1w%onD^Ds;CBMm50*RBQ0k#B8jz0<3TR1(V|v;HDNt*RYKf*
zEj~t%!IT&9vczHc$cl>p+jSjUweKa7=JO26kn)veUU+`(kavRzSK7%_LBnKOGc!8^
z2q%=7+t8yn-Jr=7KDb;?dE7~-m5h+5EFw49fO1MRev22gy~|UvGPBhaJs^mR4G5OO
zPVstb+0R$O7BGIzhpSOuM8PeRCB3dY>Sou#o;=SFhnX57#Dp=A-P9%&;1}mP>0tH1
zhJC}^GIJg2SA;G#iU~zh+%~T#w-EyS15%W}p{A4z(rymtOfMe(ufLXlXpv+uO!PQS
zZl0XAdlU*=@3$f?kA~L>UEv07ZbE0U^Y~0AV=ErBemxdIF}WttbtIh*o4t;}0ym4U
zGE;S(-(&|mI0uM~_`eCZU+SP#AvAl&ZZ8VjB%%AsTlK{igW1dI3!x#eS+&EwUclwh
zdS@k)x1WP-|Ih&KC7Z&H{8J`{o_Vb+$q<F486rd`qJVO3;+269+j+2y?F-v4A0P&>
z?e4lRS!RD#DfiBnJM4bG%gH!$9TH(2S<Gz}1z=nt)05#9_-5j-Z(C<n?&{=XB~yf^
zsnNVY5t$DsuAB}xkB`UeT~&C%{Y&WcbF)QTM<J^5E>do&<9)Gs)Q2AkiyrA#yA9FG
zlE4#=D-k{IaJVdADsxAM)0FGvjP3UK51EIT!#v^TP1(=4w>aMBZ*Av1;}!~2<O2Ry
zfAjFY#eU^$xK!4AFhs9wy=&NAggl_d^kvumpduRVI4xlL5@*D8*meXfBnyyk4qzs-
z&0B0dUywx;^z?#3wHfd{1hiW{A4V)p^O{lqcj-T~W9;{peocg$fXA8f1ibfQ-!zSK
zfLiUR%?ah*Yyh-DK3weHzI1-$XD`5#EYFxU6hXgLuv4G?1_4ll=I|nNr~d|IvdX(Y
zE7Xm;h2T8PK88fS$O^O$+S2ExM4J@p{aO2sVztwD8#H+mK)WF1d6;MWc8%@p)*yE`
zu?SyVE}hk2S~6iTQ(=YJIo>8*>PQtO*g@#vOqLiX>aR=x@`q4FkGv~Va5?2`=lWmB
zn$RFO%SrY3%W+LhFv2VSp%czC?5v~3!R~2L47N~Rx+tY~h<E;?wsGiKd>Y&0d5P7S
zK{u3*h9JwL?~`GieZSVmaEN2GPd@AhjrBH4Ljz%Rk3ThBKR-LMUig2-^}Zhv+J<L)
zG+kr2_lIKr?xu4y!<4~;e&u-@eZqWZw`V0;e4;0@cNU97@eemxqP$+_efkEEULP+z
zmz$}^`0p?N@S?jmoAc}!x71vkyANfvZ!a-=&NltU^zKW2rN7AUcFcKs_)F^OZ|?@p
z5y1}i`l#?rXnXP`cC{|er}t&WA97U;<UduLY9y3Gzxbx<#^T(JjuTWuW__e%pB$9@
z0$<by<-Fa_mqsETpKLS6YxJmDZRTBL>pH{*5aOIikZ#Kib4<<ZgY-@LXzz`$j_ul<
z$pGlKNbSrBPR((4H#mIOoQtT}9)urpz(QoxVgs&%LPs8(;-H1+o0?VM!m31Sknwn>
z&HhGAx`_`{kkhK0>l89STH%@G!IFLbTO=zNwx+Nu$^y^Llutx!!j}S6a92|gQ1hlx
zzsv#_fXjo9;t%H6FDirvu*KSy3S(AuBM3M7Lrti@m@IWe{Q-{|9#$Az_wmjhY>f8k
zyaXxMA~)0>K!BYEZ&w|Ir&%F&Q)0+cQpf;4>+z*4xKoivJFL23r1k*jDTp2rZn&G{
zy?HLtm$w?zKdoJ&3^~YYu|!n0NLAvW4Wq$~36qq~#?o=GxxjXvSwNT+rvOZj*_?g+
z>+AkzV)e6^oda<Vth41Xsd4M&#lI!`)0ErEQKhG(mr*q}b~i#li+}lk8rO~MS7p{S
z4nMxSt)~^|J|V9<-f1~VfH`&RGHlm*VR4FgNtTJdgFN>;Ao|Hzut)ii$L_igg~yhl
z!*Pm<{n~lyk`yy_4vZ|Vg^m)iH2Z;j{#WEh=lSaxKN?33jBHJ&&-ZfOBaq}bMQcIl
zYz9&fr0H)%msa%z*2X=8uUd1=dbHi*MWkQhcgLKI+_L>@FMTv8(LSCtgIr{_eFNO-
zwxfk7AJ_BFmvKsp!gYKo9DJL9FDd^R-sedP;#LVg?HCAAmbk0%xw7RPc{*b6`U^OW
z$tq?VWu6JSf8epb!F0CW-u$L{6DG30!2=C>yry{d7Uhi#0&+aN6WEqa=s$aM?NLNu
zZ_j5MjrWN*H8Ga<)?e0YOn^J?*PEVK9Vi>2s(0>i{o&#cr;BLKS!jpRt=j(DlqqWd
zeY+IrWL%*Li|9wDY-`HZ#m~c<*Kjy{iamchetf}_iA}zBUU%*N(IeQSlIrpgFhGul
zeHkD7rD2T<fJkbN0O=hJzardmDJgdvAe7<C%y9#tv1R84xOSSQ#rnQFg`S|;KDP>c
zad`9_zbMNE5s`%BEa9Z1N;Y%WO~edt*OCTc_oK08RflCG?B?h$G$T8HzU<`78+3Ku
zPJRdjO}Wh>i}3-3<8=_p&%z{xRhnv}Q89(uGM>c1x1-B(h2o&U9#@Tu9N@3At-pdO
zD75{M{I_wZt#X<BS-Pyk!zJ7xTsK3MUC1ZDgRM{(y?6oZhYkqw=;((z#&5r<Hq_cZ
z)!QD-qZ$3gVFvjSP7mKy&;9ytRYesM)Q<Kc4tmp!i#K@?CcH(4_O@j_YrbI;c$!!t
zN&dM;?53xkWnxY6!K+OxKsE-<Ki*kV{Fb?@V+`PBG!%P4Z*YZdEMIIGWEe#+^l;+S
zFS&VQI)Z2Gjrmtt!{&1C;V-kwNQJ>8{fcimn&+$~<CN`R|A~OBL+4worTN&f0uVN7
z*SHFypc%B;^)`s)$cGllTULBkQf1R9RK2=8H_dDEYU@h=rwpwrrZw<kacw@Ln=p~O
zoAQmM$wi^#YMcN0AJ5;hAqLZ>P>xV<vuWiGF~LVPqn2nlMs^o4Pf?8U9M+34^#xI6
zxQw|M=nj)7XJ&hGcWpoq=$GaDSJPg&U#EV`>7-ysaEOikW$7m>>f4=vfa=6h+H7M7
z>j8`?9Y6Fx6^>Yo!;<a7yrxs7<+&7HeUXE-T?-<c;ARf>F9?^j#2;ZpAFm8o;Vp7%
zjs7gD<*lNue=OHs3d+&Xji~TCTmHB;>e-StY~qr8?OBtqJ;jn11Q6fgA&|=~LfN~>
zU%#(sGVBJ>GG}6Hy`07~0onyZE_2$R?y#Tk18Q4%lAX7N$U?{yEYUWbg#UiV3f*e-
zuD2}2A$(uoPb30V36MhYRmIv=Yl=D&{@hUCSe|!!ul)R6F(LGT>u}Z*(!pclMy^tu
zqx0vue>*N?PT*#>X_flq-ovIRdCL8nj?gUUt%!Sd`Ve9R;3^DnGBu7|NwdF`Dq5N$
zmi5(0D`By72|oO2_4@Tb$yB|j_sP@by+jB6zDpr>uWZLd$~Pkxz=wc4e-6_)JV@69
z?Va`t-UhRF$U~DKX5;K81Y!9C!88UnZudeW`}gFTx82SLN+dUT*{<5^JzPQ9FO1FT
zTih2yM4~d>F^v-cNb%>5djJmXWW1alR~kK-i}*vh$+%pH$!IT<!vw#P;Q<FpARfA0
znJJ}^TlGmECB^4kukGAfygr=z1IiAYd#t|Pr9VT9r%f8#0u<rM@3V5AM^@2#KHFc2
zROc7WikF_kA(cbEMeYIH<*9HJCJ+pYqkXu1&+B`hO)&+XKQ{w~Q>%>h^NJR3Z~bBu
zh72`JE<a|lAmG-ueU|;{Q(=0aQUe<kayD)7nN2nt%4L3vW&5=`cBk_95b5<R<G|1R
zeh1$5uWx~Z6uKhx^<hOA^MOux6g5=uwcjPbJ9vj_-uk0T3MN!Q&6gBK1yS%)v_dqm
zpDUV?F%?RWfZW!5^DpbbEDospgFp|cTyyFT#mwT+WIh5zE2Uf6>1%YCeI1tp5z#Bd
zkW{|~rXQ)m$0EVRqBpqYZ<Vr0B;+-GMsq64d@SDG?%mE_TW}XhqpND~iG^xV)9l9g
ztaji54^b_VT?sz&Nw!~%Nw(&9rQi?d-37-s!wh2j4v=a#LmSewbrljTPqh3*l!l8m
zy^{Kcl?)dDyf|efxFF?ae^)<AKl!mumz)vDg6QE*5P%%XXzGvLzB<xm3jHLEC7PD%
zWIcqa>;v>e=vVu}>@wa>g0175xXijX6b=_9GAn&tqxikEs`l%Z;`Bh;=OVTd3NO9#
zsh|ALb&0!}+Qsn+-Zb<5?;u6jia^i(v9z}3|J_v;MBRoTW={&1j_>3#IMad$fy6-c
zx%Vy~-ra1o@M3!CF4xYI;B&fCv+*J#&rvcV`~$q_LIrvvEX8MiUHfP{hDSfyXGqRL
z&4^AFFYY8x6^GYY;LV*WlO$g}gVFmg(XDMlszW4$KA!r00_k!~5XN*8BV$T(LJDae
zMoZapeh}-{P-c2sdQ{U}ZrP0jBB8&iNJb@z{CtVtsdq;DMmVcSybpFm19yf)>u}K+
zBMO7I4n=9i$HVLnijV6f;iYWOATsxupyjcTri@tIZFNJ~!fn-pmnu@4t56PD-k<ZF
zthC)&-76oh&X7pP^ZlyN4_AkmQ=|`<mxr1fjl9EOwPD2`8wevmCMjAmJ>jPWHs{X{
zkF9qvM{fG^$2!0TSkIT09o}EcH@dCK()~!=Afls@1rH#oM`yJW&Rh1R_#u(BEt?#E
zpcDP-o$GxV0yK{0?s~u251d=efL1s9uz5k@|IzH&Q0XGl@u<t9L-6u(2^W#Spa#4C
z?k`2#YBf*wW^Mt~c}R6v5J-8+tBK6<#m580(>GKXyLC7>@md;ig_GUX)HPd3a45N1
zS8;)(y+Lu$*nrcHWDGk3z>eSh-al8)y+u>yumleV|0ze6czfCsC2eIo#>W#J>!%ha
z_djW=nhxKX73r4`?bcEFlR7RB&X*@^)B`P%f=$f&PG%BI;7x^st=Sa@NiEqhTl_gp
zs@r>8W13vji4l3FiRiv~uoERm`17PN4sGpKew$bKQ82x+sOkCYFput4gbY8v$=`xn
zp!xv<AAYNd8p{n*JO!d;O_@dZ<2No@+&$6k0?JgLC?(qgKT;p0FpOqUv4u_c$Bg8Z
zRnmwEmzm}%Dx_F%*+FhEiCexec(1;_&l3UsH(>_{>vTHyoR;Er?upZkkQovEvP4U0
zI+qVi_xm0K8v3S!mploq0dYO%h<Is?NQ`)Cpy&Vrha>^}OqFCn@6VNFC&K+B95CI_
zf(huXo+E{_kt1}xn~gb&1z^&ACinfV5N}Qxx=Ek+Qrvm_MPHo2&4QblE@z)jG7knm
zBqhT2;Nqo)>)m)DWc;zoWG$6|hVP*OmBdrr>wB<PZy*FDVHRo}Nn)^riapf>9eSY-
z0$Vd>j9><A_cB+=KXNZ*N+$%eA=@lj{B(X#jTM0SD3#_W1LSa{^t7|NT1E@JM?jS}
zY50_log?_Z7)z>~_nkBygsTwr26e9xJ@x~8Jg-R^G7vKmf=m)T3`F|Rf%ZF#%+AkI
zj4-sHl<Y1Uv$A{CYn0iD1!4POVnZ}3%Jd1WMlbQ{wr#6mBsG9Z%}RbYb_&z`JUtr9
zlu3g00%v5B98wv|G<`&U$$$@l>aj4VF??!vV>5iJ^|$RxZ3jj4JYc^u#lVbNYBq$L
z$(J|EoMIxHR7;kXdjP+hV79uhA9!K+l9v8`Hu(Pl7CfJ1&)@OYk#hNaljU;sXK$vz
zAbsr@?Rn}7{pC08DPlDiv=8@$d?5>L>*^nm;TTMtmYag`wb30%%`K!07+x(eZ`0FZ
z@CF-^h~bQ5b&AQ30{TY&0qPK3;XNWmWb!$Xf)Q(;1P!YvF2yTe&$ZF@r=V*P<rUl{
z24p;7l#X*a_7v~%4IUw_3FgqqJ@l47LfmN=N39S2$3nVfLPRazp%KFe7j?R+aKYc~
zRCtCfczD>UgFA=_X;{=XD}~vF8D#H^nWR`v>-y+yve6MnBIUyTiBO=uB*c9A>FB{S
zmSR+VGG<|J*HP~6Gb|{5(HJ@!)5n!uLGMZQ(YX>7<hMr$h}XRj5efRX@CvXlT5l#G
zB7(}TS0UXf@jCWgA7rgk&FjbNKYe||lg}hL;*|?!c~b3#Uojr5mn0rrPLdM@h#m_N
za|#~<1b1TNJEYobN%>%fVc1x6C2=pLe;QRh^KPGWN%T^meklxkLzxhVM;<~vDv1z&
z(AQVLg>|t$1u}2LE`}`aDb(s&P9%5Syn`n3@Yg<=u1ERnv!`vXM~9nkAQUj$sx!p*
z0Xd{O(M=+nOgVa~Bl<$a1F#~e*PP@d;Su@RE$NOcIg?Y<0Hj_wDDSSQhS0KB-HZn!
zaasr{oW!q|yic12#Zw^3CiRoC{OXXD^jQnObXLvuI@**WxA3h%d;)B|b-`|qVn%6B
z5<ujC!Ws1{4pXD_O~oS=XWcSdma>A)O?smQm29%I7^1hNG8jIjhLHdePJ8|7`OZ`_
zZmRuyd_M%Z;n~pj+0;veiG1G|_`z2@Gm><Q%fCOBQRa|Phxe(<9w6pLC(>^PBY}sF
zLEkVUA?)F036>(@8HYfVLCN8#JMLl^L+6G)p|hf#FVHiH{SzfKj<-@S{lUWE!o+H7
z<CNh@&5FG!*S+2Qg6ZqbgO>;{%a@MxgUu(Wf#&DmJU1d|O5e2^;-U6W@#5|0U7REa
zzwM7C>!R76ifzAdkIFx)me^!-Cq;n2(T8!XtFgYAxF3r7CH?+BRG@p9jj)6;ekiF1
zAf+3ysh0$6`?3`HajqUn(`hc(bZ;ee8i!J2h0BwTZQ&=#6-y$aABZEc_R?=7Un3*h
zDbQ#1y@G$z^rmJNyDg{-bq9zAv1@7|ayxsU2cWa>K)MK|9~zNHTCbw|YDvR`-s@Cy
zZKd-_r%VgzH$5f2SIFeVaN7`s^<aeJj)E^iPqx;f5Ij7ZbYwbOaRR46yLb}=rr2L*
zQ+@)YisC;qiZv(nY|V_{?-%}f(~}m<LE}yDFN1X<_=c(r50xxOX?n$vhrBrys_Vg|
zc&zb3zIKIf<Oj`D-OrIsaEmFr=k8`YkdB1cSa8}=>MiPr>-=;e2f8WzAT4JS-UpOE
z;x&0EWypv+1{eN(8)UM8JjOtOJCWX<6Ia9jo|Yg<G9zlA;_I*(ron?j6u$k{*CFX4
zH2W>@DG(kR)$bKc`PgYzsE9?*mx!dT1WpEahwlUu{58GC0lnzq?c(2&r{l*$N%}#G
z!BS2EX;NzD=(z1hbc9rqoP9(=CL08+i6EO~&M;8|@@p40j1hy3ZxGg_7UrApgy(r~
z9v=E+XW=0a`M+1pk_d~jm>(;yL~b9=6(beNDlWMtj6))DV5oG>`I&$y&jSe>J~7i^
zX*^LVk|S~H&eCz<dyKx>qyWP^j$+oyPGS%F9SPMeFJ=zbv*y0*UL;&wDm{|*$?SVx
zJS_5J2Gkkmy73IKJ)%?}6}JTD`1+IC4Y%mPixLj^=YwuqJilQni+~_M^l|bB(J0Jo
z+{D=q38#m{Tr?D};x2U5TAZXG$cHdCwinye@05=rh2QXB-%-DLCXKom{5<@({2ty|
zQ@d~T#3dpNOSivi^8e!a58$!7IrQXQ(8Ys534MwWpI~?F>k56o=)6O5>TT9C^9VJ$
zwe0U0wNcv4b$M17)6~3O4LEHxYhW>n2H@zzh_jwSXkko9lBV?$h%=DJaL6>?>{xn%
zl+-|*d|QETDSK$tW0vN$h<V2@ARXt?Ge8<64tKvSuj=hP0rIe;I<<Gy8VdGAK6t|W
z=4<o}5g&o%jT>uic^n349_wf!hvW_2hGex&gvA)_E3~zffs<qoDv^rA=0okzZFPz?
z=s6YB#I`14=?Kx?v;l)0!7{u-rdNU6xym+}VJ3epEt!8{+9(xHh+PN0RmTw5xWV&F
znrMdm1*V1~5s^rKJv8`c-@O!mQ-z&V!)XyY%cw~dk)nIOr^x=&GCh)xT$Z^tfU`H8
zk0;xK&LM%ohW?HZolcxUhQw0B433IWdPRx6Hzs9-#_|L{WvTf~{uya8Oe~7W^<!hr
z?V$aiqnKujV%LsOA9}gzjc?^lcc0leR^8WAJFS;=IR|W#81;`Abgzrxf3>Tnd?s>B
z=G3L+^??jBhocxwvh`EnQwUNK*ZTobC~3Sj0bd_bH7!WBnGk4gBoa9@$Az0TZQA{#
zSaZXHyE!g3gjxObTV@2bkZJTPakPzKsz<C`;VH6!Rvsdy-+Jca$Rqc<*B0ad0J5Br
z^u+D7UEgb2FqO<z>Pyk9`Xd8ym=-=8hfO0XcS1?>!Xpz@8QIGl^nrvw=|#xLC?of>
zxk|X=x@%-zRGV7`uRIz9gvx6uyiB5cPbWf`1O&zI>d=-225jokY3g;YmbcUX0azJk
zEEPZjNEamA!7|-}o(92U2m3%L8M#zWlHq<yFo(mw_J$SYR!Hk7Jsmkvg6o=lCR<q$
z$=s}iftH&6jJzY?&GZR94p}(Q9p>q1Wj7KYa?e?J#*@h(pvRmN#h>vHz^t6a^LcE2
z{c8`H{gUUFt@XT}5-)x%m4pr>UbdB~4a~t4J25dtA)vwi6S1%8G@(#lL(i2L(})jd
zVUAVU7f43)53pa|<3n5XYFBWDwQ~dz{5GYJ5790cauOFAu}%Wd%xq=Tm^T*_Bjknh
ziAT2}T?eV7zL&EUC-I^aC!y>&i6jXi=nw1-m^+axfU<~P9r7YFk5%%N;H%Gcix>*E
zc<X!N@qp3efYw#^E$HN~hvA(!ZLFwMFH>!FOO@)D;*9^Lu?ug0DkkyhyHc<$`Rjuu
zfoe&hb+Wv~PngrlOEjqTw5bH_owuq(?I;w*Z&3-$`ZDS<(ZU$gt!ESG7BUs(q32Ki
zb8mLiW~PLaV*4ZZ3tJj6iIaj^`@h*gc!2S~o(fPgr<>M~WTNEQr%W@tt#78%nWxt%
zr+0t?aZ{01!*@2KNleyJc2Gf|F(rBQRnvl(6#o-3-UJA4v4mS64k$8Wh4IZKe8`*C
zvm=s0ZwFaXr{MSh2xdhpxPlW{iomUATmd%wPBN-1BWhGxQ#gS|$q{6M6Ih6G0yw>9
zJen9U&4NVAS>zZYpS52rvSvKRgd?JWUKS>W&K_#zf`s$2u$*7q>F|x(4Z7PxdKWcN
zc2v1pQ}H%4@U;81`(>Dq-2l&f%T4^Sx0ebjEdYJVo_#WOoG@`3yO@rBc_)cmIUG-p
zeU$yOA02t-%@%%u#^ERSwEiVUzz&c_;T!(NU6oE5CUzJ{5<XsW{qU!CGGP%E<U%^6
zeMIa~tC1%FB_lN!U+w@h_?bEVnXpKONW(mb<pY-JaQe;U{$^43Wcmwb?&G(x8xt+5
zW8YjZfAYx#4l!y%#~qKM>)jX*b#7uDYC2R}C7XESj(8Oeadvt1Cl%Ts3&Y}|>pb?_
zh!^#XuPKUuBWJ8K+0IhpR%CdbX{;tGe<)TY;Euhb#yx%uy7`9T_7s2Nc-c+$4{(sn
zNj@g`O2@vd>^3m{2P=%kC6+|0xPE*DJrDuM{W{Upf|5Aczzv-4g^1P5252SnbD*Lk
z4X7nU`NaU!&9Bunt?xb%6ymfZy$V2}W00Us$3n#Ll0?L@wv?cH3N<GSLx0%9%TXio
z^!qP|CDn1AF014Up(Qc+KODAOx;u>h3YmD|%6;3@0=9xCC2>zAZc;y%O`c<rG7EkL
zk0l_d|5gri5B_qUA3TIb62OLT1=tBSrvZ_s0|K^5;dHqZo9L7nXI|}X;P#-N9&m5`
zW3kJ%D;b3&FSsa&p#B^E)<D>vStm|!)IN3Vtt<}(c@x0M<02sH{gOSG<yZpah)ueP
z_QQA*tK!lDbS+$*cq|o2$H!|%L(&vr9UlCfih*F5I#?Ed2xZ$29}AYm4<q}HH_+{D
zDDWdKui5EG5UQm^#w>%bcYvG4?ifK*cMzDubyR@ev>#1B5|{sG9nN7}Zi1Ls6lzwi
z6k|0r-xolZmFb|#bdZt66_(j9WW2OI>n;X3U(WwU`oHON)T5$WH$q|PI}T^n{^lx7
z%ThY9d4A7-X>8H|p|N0QTtj@A_>d_obfV)1S<+n0ji?sFQ!b@JRkzH-RhTSaGm^w(
zu6|---H+FATuNPF8_0n<z=;mv7z@O6;*dp0=9)?{Lc#xmgu_XVas0<d$viu*VUHq%
zks$^Bnd)LOfn*hd*qnxs_IHEv*J*GLR?$+Wb`nY5vOP}_L#J^P9b}fxyL%H)lIY_8
zOw5t%A|XMYp+;==#-ev8bGirXw@ZJ(Ep5`Z{4cC9dE#x!960rpe~Jo5vmBM*&yA*H
zZGA}eV5$<v6Ekw3UP7p8KeBK_mH)kB&prsb|04Bh5qAkp=b{ToaN!-p8)$tlirHa*
zXRH-#&Q1jXIZ}?&u=gq>W{%y_B83Ui(G!8q3wpi_Djru9d4I1*=Xx*A2{Q}t#6e38
zBOv#okJ`Zr?W@L_Iy>R9VH~i|MRt9sYmfhP^dCSTT*hT7&hkC{asZi*M2}cBEfTH!
zodz<pB!N?}sU>I#z@e#*r;~X5TwCvh5k@K>&RM59n*M2A5rW)aKf=$=AHMVas!W*>
zMA{ue>_89U!x6ZjeQ;jwNYd4YlUP?ciJi=(Gkh4!WR)68+`s7SB6@M^H3{;d9P<pf
zw$;%B9vMfu(8Is~rPLg7vv}+7-)sT%nFKKs?8i)8m<k`+eok4@BIQV^^J5?-sr5W;
zgyD{S8VuSu4VGQ=UgQD$+8jzxHMg=*{noP_{V#ttkNDDSe(Yj7z=%C8rRF#miVq)}
zTqMh`oE%*+q2dyCxFo@_)NooKvf6hLKl>0dVyv8n&JRIccFU}Uh)>6w3`coKeM#Zj
z%KW@|6yuv8SZ|wBzCp;n00vGT!4wTQL^Vt293zp_C@ItAIcS@?!1M(9FgLV~3~F|X
zInQveumKSWqHrH4i5c@0Nq6{l@Nd>h@p!)0dE#kX0UZnS+iS5HPRLn}WcFh^KVN)K
z6vLZ$e6f6{Cwgwh++$roLU*qay_c#az1Wb+K|U`rIS|gR91F!zMQ3Rsz9I-iZ2%W0
ziEQp5Wl|n4Pkl1h)P)h=(+L!h2=m?vS&ltz_AwuD02MQE42y7Y!0kIx9)8P_J{~OV
z&gWx~0b^-p=N_rl_lk9Yg}t&9xAp#cv4E}h9{`3FLtI2lvO7Pid(3P3i_$Y)Zc_cm
z>O{%JKA35~J2^p3`RR#Sa^GzjSNX}Ka(nVPs??N<m-67<cBbzm{S%u4Ow)e}t80^%
z#BPexINud+FKv|BR{F%Q0MmF~Vw&8f)utTJ$*~hGk%ar3`)IloirR}Ay1ov-cm%zr
zJh*03l#!hbqbymzHvI7l-<^P<Y+lC1;NEx>it@hQd(B@@+UsEWoAkPb)v<XM2fHwn
zEEA#FV;p+=>Iyq8wkBwy7x1%s5jA*yT!K5-MMjI2-rZ2aF3PU;;g{?B@71C?*|ws;
zCAa=nai59jZu!touo-sx`yXI|hg)=f|3Qpqm{OYZ{fG#rW8eNEJUC8C$+~u2oKdS^
z>buT)rANQ^zc6l-vLn$0l?l!dz{#nDlD!T(vLF8dVW*RuWay;CSHduvs3*S(qjLuM
zz*n_o&jZ@;{LHNUyEM&=M)~IcwECs}^!g==yV%vq6F&4Ln;Z97dsvK9LC`@*BMOzf
z#ti5Lw)lHX4*7yINxZx*k)AjUZ_B72mi_E;e#NL7fFB{h0Frge#f%uksN3SoAAYMY
z@s9W;(aJoaz-`lt=9J5EMoW?To(Hmb?*{C=_9#oK;RX`e3jN#>B%&VF^BT5t4r8Jz
zo8S!g-MQy{zL~(V;}cxqcPRP{vNlZ59Yw7y^9iwSP|;+l>_N=SZHp-!xb!((C6Pe6
zE7t&S*`rY6ehll{TpK)`D>ir7S7h(0pKoUVE~louMC4bk$!H7%c57|KtX<n}(9`#n
zcWY4MU-m><xg-@DJmcwj%w<I0e^p@|f|(D<(%l~ycYabmXO*`GR33=ZCnIhoi`{dG
z)bASb`)J4L_k5ZujK^%c?=`yKDn7k97zli6NozT5gPv6pbg($qYNHv58@@iJg)*1A
z_wR|cv^I_S7@a2@H6mVpwbkwv`Ni?7<8$@Zs7WR9wk9NOSL}xH`b>7_rx~_vZc}TO
zG=pN+*H3?ZheljJ2j*Vgxj*mOPNv-)7M7YuVdBW1oK+CCFm%lGZl7E}3^T6_|0U0s
z<@SoBvb(bo{eMwXX#Gmv*`f6^9#Jfs=4JJ#3_caU!TYcO3kecSxvSG@rKS2CisS!<
zYuv@$PM*sDK#SLOwzKU2dp7!`u*h+%+ay;UP~(5VPFXGLXRimJ9nXh<@<Uk!v{XiX
zG@d{Fts(x_`9Gl5DnLToQptlm!Lh=_F8%}TB6vRUnMuBENK>%jK5v}BAL0UZSGYwJ
zWs~hJjt#8S_tY}=aFj0)TOG!G(aA~B!qT-1$7#nsgLwX1;_Ey0Fg8vhXq<FBgNvfe
z%grlgwXkUs?icA|(b3m;e+ZNIe!rv69gKbm5i>%)Q21n278&ZYrJWY(r5YP|H+V-c
zk8@{aAz6yEpNb+_xVW=lc|br~Qw^lQ+YTdySu)frQvvy|t+?>~%Hx0-1GcR$dTdFN
zrx${6^~~?N+r4}pS_MHSQqHHSs|AQu)3GIa_j+H(H@!mrUaWG-4yJ(ICI(;V`aR2}
z(}&LZ<38h3v^6H!k$(%g|1!Iqyy9Bu{3i#!^hL>)!mO`$GjeA_>EiM8-By$&==pAc
z@|)$$!=H?T8L9E<8o2GBwNYIz1+z%9^jPBVlT@>F2<}TjQ~45qqhy=KUH@rSdx9;Q
z-2W5+hyP!xgcvP7LlM7{)=`+IQ8rJPPH(7CIoll;s^GV+%K*Km1kSmolSWMt(?=;|
z6r&m!4f-hGgw%;h(V+_AzURULdg;U=B~#KliB&ZpDi~K#?*Qh=Ph0|Ngu0E8kUoMu
zKVN0a+iEGQw*S&E)Bhsm;}6p%Wt#2PP)ptm>3$#4+Z(8BoW#HqpWtQ=&GWXhkg&8i
zn~;g)AQHpCOVm&G;eCw2VMwhstTr<N0obvR$fGg@&bb~uo}4Kr$(nhkquaj->zw$y
z2N+Tath{*sdA9%gz}5MOY4dAv^QxEQW<OsX`HKU&l-A^sQP<2&#x0+lG*aFx8wK}#
zwBThQlv_vUmGAFcC+sof7hm$f2_K=J(JxlbBrD#FCr`HF!W%4S<I%Q)(5TZtG>&tQ
z$J<++W}XWrx3g3v6*L04hMtEx-}4*}i?=i=1-2)Nc6Z?=9nUY08;hOK1@tzGvh#mm
zKJJ7sc1{-1g7lFMpaI8QyuSr6StIYs4vYW3@b2-R)p+P&80q~{Q~q8pLUz%w-Wxe_
zAkuUDEbKu(fLqBBXIMW8Tsq0Z{_AneAweFhdtJ$}K~gDMbtEF>@*?(mwsRBuu{HY{
z&lG4n{jwzLH@}XDfIvju-6(vmyHtnNgb+4Q(pZEhygj(=1uFze!bQ?WP$DV7!rBN}
z=wY#)3u8vT&X=8$zp(On)|b|LwhoD1lkIj7GYeq<0043d0-ng3F%?j8Ip=k?Sii7z
z>Ha+Mm2psdBm%7{ny_?uYJY%>@K`R)@(n7!Y*Nx;Rr*j`xCzxmJgJVaM(&bPxKaN@
zKV{^z)>Yz5*FEC?5|&D@T@~EZAyp~mvg7QU#ihbCIdR#tUax2a-&~rro6dmei%_O;
zESXS*)2I4ARvWIjD&zNp#us06+U_Gq%sW{=_mfGBKs-YV%ul(Up|S2c=4J8X;O``%
z9_Pj8&O_Woq#WWMxce__EnzrOejm3T0<_O~H6-=wI8^M;eibV@lgl;y$X1mItd>SZ
zU)+)}D0-sSFTY4cT`Nt)f9uUZO_C63YV=6hg|*YQv$~Tys1mV>>(6}do@)NnUbauZ
zN$5N$+0}sOnS7R0FwPohX6YZmTbG5nqw#Bzv={mEKLFjL1;=t%TL3ifdEfx@571X0
zWLZUaXytI68m*0WPo_w;RLPUB!GafM5O+`7MPp9nt@=rnuIBPn+VtEd?r60{D;xV}
zSQevp1$o~1uX<Z0L$j&_qMHv;#Nz>d`ebZONfq2qWR7^H<R7B~+n8=0Pq@M~oFrVA
zvN7yjs+3_&7%(!JJxpUKZ18Kq<7s`0ePVKBF82P}US&>_hXK(u%Md4d>WU!L=3;Y0
zn{;vD*Q^c4#^*|Szjm@?9Y&)Xqb8!JlQhpq=+((qvxnv5vLMhi`m@0UVIz~d#yMxP
z-ZsxOK_>D)K!O07P&1Xy8>+R9WQsFO`q(SM{7<cHPTO)d-!#}|l0W`(rr-aZuBlXZ
z5R*Jm_le$FtJb`hE*B+Zw0KY9{kjVVFn-i~)ZfXZlOvRV$o+vlEc^G%7eCF_Z?V3#
zkHMXgHd@pTh<;{{^fXPM+v*<V)=59<5=$xWezn#9Qe<9$zSxkff0q@O(1o5ty~dvO
z=Iqhk$>0lv*O_~UVE;OAk#<@o$9Vl{<P!ET40fd_jr_VER!|!a_WCi$Zw|p`Yk4<=
z?GEC=1_vYsY{sZtB`a<#9!Eiea;C@@7lzvS1~u+CvZ_g{7+cM#(&`hv=w-t@XXVo-
z?Kdz=53a=L7kxjiu2<2u+z2c<Lbe}+oaAfeYnCe>q&p4A2T&qU8rH7AS+|*W(l(6I
zjo4>ck~s34GW4Q%hvFGWSUkGd-7Cd4K%`19kzDn5(!oo+^-!KE%a!*lrfXTuSEEP=
z3L!aCN?4JJJGJY?I*vcO`{7%M{j_hhxakLN;F){0d1*;iM7VmLa&>^DFmgA(z~NvS
zJv)S<By8)u|7W`iLySf9%YmT*{(pdQDDRHHAP<}$&qycACon>>ryX9QX*h$EOQxo+
zaJHXmuI9gAub8%Ti*&yYynhlKruQe;$2-&XMWr(oZK0FyAb=-dpM76%Lhp+H0hT*e
zDI3JlUU6R|wq9vFL*X~*z~BAE1E3qTwX5fgv;?jt1QW_@L_HLAmuk#&bjSK!n^H?&
z_?fcDp^NwLalGjcV@3q6la{2g4NML&q3uv?K4v6XE_1<%U`d=x1>>AYP11EV7@2HV
z%OsszOKfTO@dM9Q<E&Il$HadfeVkEs-$-R%9c*R)PV^w+Pffkh1)=A~k}2jp(2@Ch
zhHvCPbNu*ricdrN$oG1l-|t9sT5yUzkA)iUhn172xuj{DJC-;OnoMO>uvk>1iBEr<
zF@o@Y&Afj^voB6DO9{DKTm=f0Q|@u^3B8FE`MW#`zfIxP$A@~OaA@Ort@HWotbF^s
z4IW>uNr%k3cyVuCW^E?%I)f=A&Ptc?B$IfHv`dBplCj|e5PC3sWfx8rmjCclBcmU*
zU;iGxXo!a2Bb|L!K7kXmBx1g68L2Pn<K{a~LrhlD&#!e6mBp*wPCJuD%rwRb!IYM@
zROhnc-;tJA%$VWUR0HjJ<>BDaCE@QUIY`B;st60b=~tJ0YqIPWABq&e83w9>TM)%X
ze#7nLJuex0jvTLb<=J)E+e~C@A$l%YwO)Rt>bhiVylwLDF;4bdK3As+o%PU_&KhgC
zr({*jL^d@EmqxsMbagT!WAcE6R9Wn>mKQU<)7IHru@JYT8=`T~`8=LUn8EMdj=#IP
zo`8D$rN|*b{PRXs1+S?nkU$jU(+y@O%cmpccVGgc9eZ_k{mL|#>^vYt3+^D2?p#m4
z+w(O3fd)uXMlr2z0vJfpkCzg7I#b<A`pW9h@WSxe1;gKU7}NDoc365*O0Q4iiB&pv
zU9I$FE^REasc8G`{d!h2+zCm^WDmWq9SwxuR4!B}TX2j<k4L{OeLn2V46^2NI9Rsx
zgAS54?ht`YV<&BrS?BTE&bU~gC`BM1Hi^65_%vL+@1PrL$dYc@GyR*=3uM2|B#(+I
zPn<)Pk|b4iP5Gr1dzIAHwe!+~bLGUP<sx350*w!py&odB_+Gk4iHVWrr+v7-<KP7Y
z{HWot1C2)p814>k|9V{QS=M>MW3m7&yPos1JZIg$?fsDVW)e@}3@h|w##7BiodnA)
zO{i+NZG*#dcPc2B;|AYGgj}CY!&U!*aGw(vI^f(#M+VmZqFAP}^M<)#2yHUfA_D@-
z{eckWq2CmnAGWV(;HJOb_V&l0m|4Ac6Ag2XIUYSzHBv%?G6p)%v=B}?IY~SfOC@}?
z8Y?3V8qyEhq`K-?7-5C|c*HdPIN(Rh$Ck&E$8RfN+gHXz>mb<$w~sLbnx}uTUJsiL
ziR7I}+u5*2Dc5HpP2=mFlep)+$$20F_o8%DeX8*MNcv1I_$Oi&3uW5VaF&haoqo*#
zcJ$IYXVF-=G@WA5%}%;y7Oh|+m1qiyC=aZ=W)dzF0_HcX)r(PBNWx5~#tCUIb|!@p
zs%tCUdTPbe=4bfv(T)5>SB|PGz-V6_2Xu{K-Spu@y{*OBCbuYeW_RMY2OTyhmauq6
z^=KcX9m1<&^`6p=*P*&Crd@@<UoVIRcg}gN1f?CH{sH(_6ivxQE)Jnl2X=O^<ZFU<
z*wffBPFR3`SA7)anRdFmqEQ1DRpw2neGc5K@)hNSr|W**;Ki<|ZZt&<P{_>LUY2-1
zeDJ4Sj_Pp|MiGQtp{PDXf?I>u6e&GGFu*oTG=}d;sRXMRPeCy?s>U$g`7p<yHi6pd
zk@&IVG4s)9`MhnpGPn-XP;l`WK?kLS&q!;BO#5w{RMrvG_NphDi*)mjizDW};SPHl
zE*`BCALC>^6NFHfbF04RC`e_hQtLzPRjB(o!Ei<3kaaShE*IT!Ht|HLMUHdajqZNq
znX11A9;+OUv+Q(!ShRt*EsFtTeztP5?vzbpOOX^oVy#$33b%TQW<l@g?G{OH*xK1_
zf!zm`;Z9Riy_=c97hZ}m5(&ZTpsvWth%RIvEXP;ynSG{(l|JPgi!blafza!drZ0}9
zPG72Q772DeRv?QFp;{SiOC>`DsP-9C3o%Q>k)Ug+jX&j@pqB4C@%$nAHg@e^q^359
zMg6B_s&wr%d#9IU=jDR|lBXI8oU>sHVv<z+aEBPMkg3_5IYGi?yoC~|=9;Kt{y}{0
z@jN6r(v(}zzcZtB&j=?qtz<m<ko>DSKN0v*E-M9SdgX$Qker5);rk6JN+^oo`Nr@4
zW5ku2Ma46dpt=7Cg%wgkS%HVv%Eq<>O){lAbr0?9zsr!ijYdJ?40?wVJZTHA3he~O
z#SR`09>jMM{%0sX^<QLN&OdH*Zm=<hFBSY2x2#u4whUjgpWb*Q;@4bpW0rQaE6U{o
zA)8qNwSMFLX>~N$H|{=ANi*m&w^GtuUazi8%ovyaUVB90ooTNGBW`kkE|5sMBs!)z
z*Kb~0<m@3mH#JyS^*bi|Cux^=UUkM@|G9JU3;t?cfUl_KvP~ukhHC#0An>=|kXNa3
z_8T<LhSj%mw_Vh@X|&^P;R|2rfctVxnqu2IC8KjDSO&r&iTt+Xm2!=EV{ZM65hU^|
z^sZ(5cb|EnQn5af1T#<iLYb(5qPEw8jN%9^{J~1>>LQ^3jq{vgjH1xonpvLnaN);a
zc&EaADOAZS3Pj~a7SnNkO5HrEY&X}<$4_;olS12LTVuf5NKHyFY9Ox(Qd}FT_8%ZJ
zH)56ciqfVFr%5Z<!fc7WBHMeSgN3N=*SetfR^(l}Nv7@JL|Ce?LajXBkmtyoU}&qu
z7DsrrC3PDMTw3vhY{;&H%tl18+wj4UlR5#pO{x?krMsO8>`H&Pypsu{fq(u)h+itk
zj1a>LWpPgQA$Vp>-3w(Qj>?}D6owT(Rs1xe$qAv0$Pw$g-jBoU7xmtz<w*0V>ec8$
zEBQX0hXW0onXK9zQ0YG{xq9&Nf6Kh^HCYtr(x$FNuac{(+|5^kw%#5))20}MH?3pq
zcQ?wec`A^Gsg|cU?PsD{)PY3q)$4Y>{g0TIbn-2>Om22bR(+xfiS8+#-I|zV)Nz6t
zf5qV#7NqCFDBRKdNjEE#+0(5UxW#Y?V>L9(+c?=@ZN8ruB#ygRIV$S;jykdRr#t+?
z#luZ*v{$#qNmfNHo^sItgMfk;{+>sp$jRExTSw$wK0WDq$JaTMuEqrWbw?HW=4&nQ
zW%C%2wdwEoimjK)%PBw6W|@`jcF^iSz*iDaU{@~FgRlQ7vC99$+gEnQ(FNTSEI@E4
zSa6p?26uONcXxM!5AH6(E!f~57~BbN!5xAp43^9DzIWZ9aKH4IUfruspQ_X6?AlfP
z4c+m`;cbGfElk+w*lByL1@so*!$<3uT|O%2)~e^<!dQM>9)DQMCW*}oDUv5BEsbya
z-lSNQGw}GO4Q|HI$`I1$$=37SF!>2wZ?JKral`-gBD^IMRclpb6%CopbjtNgT%bK~
zA84syVu3oWBb)Hw#s0~$j1S&&$fm^HYQ(X@l!uo@{n`9(;`eR~yA6;B%Mr9yf&}MF
zY+#xhtW~N}BFh?iGbr?>S~dA@_Sm=a3!0Gd$rvPoYq(W(B+PTKzF((Cu`5bXf{jmg
zKTywYm}!`eFg}8EZ}3nc>?BRzufMRSvH*s#&uH!Itpsvw(Yq_=SzXgv^<xn97Fm5;
zWl7HSQt7J0XR6<#vMr50#^1Gk1>Yq8EwCy;e%%#0M)vV6_><M$+2xbv19^_USl@s+
zUd(SyU-?PcM$N;o-C8t>vxqPq1y;jOqmGHp4Rqa=xJ<|WO!p|ILg(Va)p8-Ahxqqn
zS1-qWVz3H?(+81gH15`C3@K<dgp<%S)2Z{b=F^ptX23}SY76ZFTlwDZrlv|_fnqOO
zfeMdV*E*(eu$O2>R^=Bmkj@Pu{+%Lkqs(*eR}P`vjs~HE?tb52I4%xKxj!izjoYf~
zD7C_7tk)&0lZjLZa62xi8wv16TM7DBbW7XwmY<^55WvWH2UAU7om<D+W}zHr&0-|*
z%FUCQE=oz>XeY`y0t3-+@$tO7?U-6c*y)eI8iuR1mdy+}S|YsEtjVdPv25GIdU;?g
zW#f=A3cMCNSOr+FJE4#Ph7f*#XCG@%a@>C~?@lVr9?1CUcO77V@yx<xB}^pOVR><J
zUU<>2dzF*y_-96OEuiO@_uAi9uv86w{aT8QEys6nGvMNa8TJl2eLY%eN1KZ+TS0mE
ze*3VmNt6D58^o@N(J2?lsVy$>Duwa={CN0Pkc)W1Vk`4H4eMo+mavP1eLv!?PS*jC
zm8W)oDWl8W-O370LXDP^KmsIKR-ifh$g+59?>M~@j8hbUUWAfMEOhkD+IKIP&>6i=
z)Q@L^Bh2q8{3B3g!$5dauA`55qIrDn<Tq3OwpkCYgmoA53U#U9ItxK-msSPa8Z&sx
zgmKmyOJzG7CgP>Tdg?`_qfP#yk1Y>u)yHJ?7uU@Ip*}F@rx_;;V0{mYDQ2C4+jK_-
zYA;Rr24hZYJxCxeOya6rQ>%a?>ZD_h<{&trIng;RC7yt<Pa+<Uhn!zI5Sq;-Kh;6z
zUJgJiLQ%}^ZNwBzn=6U1ZQ3Ric{f)Ci-&e_lQ_U;%|R%7_n_9XHj1U7>vF}b#1j9R
zX33)HdBm#bPmfs(d78%>YBS>};R|a|NlD5G(KYhuzxc!J4GvcG9M_3Xceqt`<{QF_
zj~~)F=7AWgG;3vqBHR_$f9};fws6hw2S%lSeZCgd(}}g6Z#BxvtV>S2#)qvl5#Yur
z23310Hu(&g;E$`%701mjBlel>^3ryz2(TQ8!_`l9a~=i;zUaNEy-06(qi^59PJ=W3
z{=uN{_tF}|*|63IZLnJE^lYSh)&7kKcK^BsCR24diqIWOS#L%$iN=gri(1{SVMP0&
z4^8wQY~HaU?es+!5}RDMfjIYIr_pMNLg)_H%AD2Lcme{BT(}6Gv!)|{#Z3WCm-Ck8
z4dl>j76<J+72r!|DC=C9ge_1oG4m=Vc6`h1?(g%v-%P!DAEu~Lw{CZlFN>(pugx9y
zZT>@VMyJ28vNUhRv5t@RzLCs8SKtZI7MvS(k7S-_ApZYg);XMbdUrV%FfO%`Ka?vs
zhj?0xUSTE@fw$3}(UQ9x0>EcEzN6NCyPIvqRwat3jY%oM+1DZUi=Sqp@8qO^Ft~s>
zYhFY?*M<3@s7=N>|Fvj+ODC-gOzvAfDf*81a=mdCa<Pai>7FQ#M{LJxL0Kcl=dy-R
zKlg>zt<ja&sZ|r;9*LT7Ty7k0H1<V?ilX~*|8z}*-M#a6%g$@OXC^wG<AP$(a!6-G
zAT8P|zq}2kZk!A35P!jQM&xPb1y8?j(jxWuDiyrL9*)DKffuh^95=C$U<uo>4ZNCp
zMyGF|l#*8~0zt`o<zmbrn%!Sn#4^yhq=`DQr+1tt+K^>=it>vRW4^tP@vt|OUpZ_?
zo1f0dLo7r!<)k-QjufQ~yzx&F@L-oQzf+2=<RZfP()h)Xn;B7$X^(A>)$dl*LADbO
z6s*^X0&K459Cy}E3A7emz=j?o;0-zF^}P1K1yNpC4lV2L4FZFKz=jML@XH2$7UKs-
zUd7eA(d%=YKOxpG1?f#PzQf)oMeV_BB+rUfm#%NyW*I)SP@Heg3R!TnX&<usX{|4&
zPu?M>nwHX%E3`!qq*e(8J5I*p&&gWj8Jg=AuDjZLz~)FiQ%zZAhikc(8fVEGtrZ$s
zYU4EIRHFGvqLZs(!ry)S$r}#urq{L?k+ccizy+;&%cFIz`tdO5<w}K}wvVly>~z$%
zM|KTAdu;>%w&+(5<DSpV-%WK7ismrON0GmiWu%<@qYZ>S9m|iKmUXY_>+_Zc6uq;L
z^~**bBi0Y)GI$P^qbA`yM582$^L|p9o6`)y$g`E)p^YgLes83gx6mk2^O_p(S(|!!
z`c}bzMxs$#(tO&OA`p?A#SpsRT+UHHfK(0r>JB%b?`tmn+(csXesXY$?UuXwv-=|R
z3*RXvdJl0%zHz+!;*UecHk%D)*j4vsmSfVo^U0N8+jzv<Wpg|P;h0&9T?`Wwf*eJY
zw7Wd3+a|o$#9`5sFzXcS=ushKpYHTyrhqXpAZ<94NHqbf8hT6lQvaztKJq^#hxNy9
zk<0}gN-h`IFRRneQo@?l;wskR;PUY$H6*&33L`qV*?2HJ1&m-`U@!_{>V~oFwX#5-
z&fflBLG2BW#f!3Xvt#mOl36;nT(g*`LDUo>#TvsiA<u6}AhBKCU#SX9%ebGIP9;Rs
zw*#pjA0f42#v_WWvew2XowP&dv7C0PPQP6U**F%4v*lgZO<~(SUKb!QCIzVctgf3@
zUkCe=*1}P8FVX08y`42jK9`lRhu!PgYhG$$Fb6V3k8gYIbG`T)uvTo6hlP>U4Vum8
z`>*vAkmFv2#N^rRxohQFCJi}wrqO!kGL>xLPsLEKF$s^5PBxlQu2a6wTW7sb9mTU<
ztj`k%HT-Ne5Bzj|Mq`R%Prv&_j1;Q*0=+@LG2oY1(KF8k)t~IKe=q~1f0Zz9g1wlO
z-+>#}Yt#52IrEW9(Nc5GG&q(O*eI#lpE^?Hl&3MuVSR5I%}px=l9cnN1qdnV4u_$x
z;b`+w&q<O&uEDIr3>Vv@8_zeSK5O3UY%#<#v>UY7&HBy(p40hJp9Zg*w$k2J6uQ6l
z(dZwh*ffrFO!?#wpe~TS{}C_;+XDGISh^`JS92EFusl{vGH|nZS|*~_6>}Eo2Ksr|
zO436FSKD(X**?LHlxsb)x^C&QgA^Jyzq402+M!Y9$}5#uf--109Bk*+hGi@dzc_nv
zWEV2!xj`kMyZOlO3=|1OYd*WZA68vDv6@2O5vG8TH=~Xrm^?IDmPby@n{s_++rWUJ
zS*IdbF@#sA?z-jeRcENqbm0mmz0Fpe*_WvvI*=<-@(?ZIZY<vEJATmnNBv^;Tyt{B
zh(}Xf_hNpcWjt(EDOK}Sz0dLF`}|})q0MN2H|<A(x-K|8zmqPAH)UhHWMfP(pc=dg
z{u+y^h67rXlk@MgIJks&QH}{F2Noz2WvM{L;i|`?$ZX|}ju9yk%`x)C+)iziqPF9T
zVnaR&175@)_nU!6X?blS*#BT!iq?q;%4eNh4d0pUt;BNX>v@xZt)isvEL+yvT?G0h
z@l6VxTLWtuau`MONW>kP?sk_3T8KG49Lu|`_R-ed-XPDo-T{i$-W~m>jnmx+fzYP4
z3Nhlb0-~I5TI2ia)#9xP=S-R|*2b;aE;Px;a}88`&k~fclXO;UI1E~9YUZ)#pUpn+
z5q78?PvLyT)|l;6eCOU1aSZN&gnZ`)aoKjf1F8Q9U;^CG?GX$MB{w_fUsgP3@!uD<
za5U7GY&dq6j|ObdJK<FsWf}<GnH5;WUdKSFm0Q;pf!6Q%lE{^YZ6C=dv5LFl3c^0e
zZ4|@}ViPrtm%G4dR6=154jyJZp31S#gSQ;Vz`y8pAwNPDVa(URH`?tdT3k<1@9MD9
zI9SkR)wRNMQ>y`k!)31{LW_AR{WxrMgtGEOrjRXv_2lRH`a$MP49xyg@ChUT=glv0
z>s!4WLec-hOMf2|xRlqC*q#fwT$(21jkPyc-e%IQ`dFcvz9Du!bm4cE>nao2J3%Lk
zZbsGbKQ$+>El;DY?>_NBhg8x@BCx)dTplJA`#rC+Jxp5Y2`I{<#n_F)m@?<ZL{zl#
zG}#>`jr4RaY0cFl;*%#umO>!9A=Nij!RL(iv&^r~f~b&G;wmc%UT~So<e{c>Tt*k>
z6MiM697GO$fqTz9sK4+T0Do^^zfFU+LEBhVT@@z#cC2wh`|ut0pq0&Yy5A>`KE0cc
zJHsIc!9LTm3;jsVp{5-BoUAJ^0v*70tp&!%#G5eUTw+O}$Z8!&St<+rnpw6DQJdfM
z;dl3TA4yb2=ie|-gkvZ3kP$w-ML!3<Y5Vx%Vayz0u5?Gd&N$xqaICv^0Wj~;8fzXl
zmTR(@E1LyLjE3Eib<p2x<yXw}^`6-NHQkNYjllJ#+vU<gqqb8<i*7K*WRqNf$vRWb
zmajg|fGOBGlYzXuLXXX$>yx0g(v3YQ>Y8@K$!;1Zj2JG!#X*>S)4;~Q%PFhJ`;m6D
zNhx-OSsSEoO6xc&xZ3B*$$Zy7@{nuboK5Hg`<p1XFo8tK0<nJHeno4#CIUPXJ~cK?
zX-0$hlV<3qgrNFBD;egu=4$94cc*zSw$KaFRCX4Ea;ZmC)?osR?xOIJ*apxx>>qNM
zH*!a#A}gCs6KooNYA5*rftRk3RrQ_Yl*T##dX-H2uhQt1ye2LI=@rkpE1z#rEeXvy
zrq^Af|DrD&w@cmlOZkd9bfR{AHy^jlpq1meWY>?L39~BnR7=W({}N*ne#<-5Va`@6
zgF#p{<b*MdC&OLC@j!=hN559a1;5@WFU9wnlE{awIQ4~((!7X=GtcYpG<q}t4lZmU
zj>485QQs8N)@E>a*ktG`{2vTTAH@sVnt|}d%$f)Y+;A!&1AOeN`!l(xNi5Zz`p5YC
z9bHQCUoF#9<Z<jUZ_Zg3P?>BMT#Z;2>3UyO9z$F&Y3B?#6ntfT-zCo7^eCGotXn3t
ztZ9dxfKU-(O>e)Z5!A<+cnP5KeOm;f%8@Qph_p9p?H{IUM1YQTt1_B?iF#FzWMC8a
zjJ_NLJ~o1Ja2qv_AJHHjRmWAIdufror@|sapSsd2=HZ~W*;pFD>-z71uCd!Npw#_{
z*4!49mD5CyMdyTi{uFfVW#Gi~((xyEuN&xp^Sb4d7O3agW`-7a1|GJ;al}u`vW;~d
zi(j>fGQ@|z8^7JH=6UJW#xzL$^Dt@@EfE<%M^z80as7D1GVoMw?MN{`XBiY`7UfAi
z=NwA$C9SMlh|t%yCc8`2LP^m{dw1!{g}E?8WfH$3%82oWdM*#yr$}KcU711Z=F!??
zGCr(Cl&ppA%SA^w&LmLf9eEJ22!?l3+1)#BzGw+RXuWgMTd6Bq^D?M(!o}K#9b_?S
z`o(eO2b5z^+PW21P5hnLamN!Murt{6?#()d%d5bddD0%^zm7m($7_WmMq(^{mx9|j
z$8eiiOUaj$2!rNgR|?b5y%ZzXWIxj_2qSf~E475QCA1jb^<@*pjWGXc`~H@VWG={k
zN5!z{>5Svs7^&2F-*@`y5mY?iREWw3j?D#)FZ>u>##;hyH<%zhq-)p4L8Bzoacdic
z(4Gr|y4;l(d;>wA^3xU_?su@c_j_i5$!~$>>+aD@9hxYwzG==sOz883mJ=C{K)*Wy
zNHc+BK|@h1rhWuZd7DzM<c6TD?tapr0HN3g>nNo{OZVyeS=x|;R;4ISi=&w)K{Y96
z8F&GrvjzRZXYSx~t3S0;h#@>)Q*dIKFq3MSWF{Y~C#PF>&#O^;JH8geAD+tZkq=hV
zsMUG3dzn^y@88Vb!)}NYeJr~E@M(RsnfSCLiRUm}qv#j!O$aK2A2#p#SGC&}gW%@}
zMreri@#cG;@1>iS;OORIfA{H5lh9p6!$h#Q!ai94-1HxeGD_B?b4b@FDz8bV{#u>P
zV;64-8~T>G85Q&WU;lkY@xWH?-iNRXw>P6%h->0LS-nmxxVN_G=#CNpP3_Hk^t3|B
z5cj2Z`TIYZs0W1?9@UUCu<bFlH>l(3v7fde_gtlP1z40hu}Hq)Uw-(SSM;ptb#8wQ
z?U`tRGZmEa^F9>TKR4M9F8}?OkaRm0@;eFjsb%KkAB@8rn-x#-o7!PR29vS8umO{9
zh|*KmdYi}?GgZhzv<|6+YPXEYlz)ct=Q%&0Gj7u|CL@REE6u)!#L<@yLWh7SKkksT
zhu^JgyJINaw)|?7UOPoBUJX~h>iuZxBy3L>{jArUqaxGC1mFhr@X5aKQ2xNy<^H}0
z`4`~(C<65oL-O9(EPX@L{al~y*UkU$!@sGr{`;s2{fdEJEgmnU)iPgiPaiaHSM=YW
zm)Ne?#NJX4m^~UA{E9>f4j0BW`i3cOp$>H>L1$Tyj^B1e7D(UbHsATTK(>Y=k)@Q~
zIZx8(gVXmJV|iJrvAUag45|A8#}@H1=z$bQ)T6QGhAikU_RR=U&tY4iZc@u(^3i3n
z*XgC3{I~!D-G#9Fd}tF#6Z9H?D<B^Yb<T%55vV*l>wh08)6!EmjKG@!y^k^ULnQc<
zP{680i@pwlu%3nX@p-6@H-Yz`(`_x#*_!@^2qdcf)D<JK2CIJ>Yi0M7B@jIT+)WZy
z-4L#k))&qZ&qaz@_xGTNyQCT8exq7%L#rl9?FdoOsS4&mP%QC>R*|6Rb=A;Q&0?Lu
zdk87OqNGYYcG9utP@!?-iAP<38w{Knq^P?^2hLo-IdSljxu_-SsC9@0-B^*c;N^K4
zRhtzSk*1r%Wwt*mFd?l5I1ShT`7F$)EcF9Is%5(7d7zTxR?C1guh;k7%2vA%P73NE
z@;Ra!uL4ym$4ePOQ9YmbV&4l{mhys(s{IrCh0m0EUIAlHO9jpNLZs6SU*Q0wLPjl~
zR;J${ioQUNs*OxY(*wCMrm?omFQ#`JJj+RMuBe!c_0GBba}VXoLP#~mi92%a{^%!M
zIviUIF+ix!jfwmtuKHeaBW=Gw*KEnZ;`IwpkP#V==*L{6Nw7B?3GbgikeP|EwduP(
zzv8{=I{=^B)gQ=Dwxoub@gL=%e8;N-*TAE*7o`%AG!!?r+a<m5C#S{wh9nfSCe*18
z7!w0S2_vn;`)3BIbnfAw;huX&#5V1az9sn=9+BO*)kL4klb`Ka#}|2X!Qq!h3tHqX
z8AVs^@BTdd_z$K{7%q?aDY^)pH)qT5dJ&X%Sl<$r;a!es;(h!N=9D;Tk2nr6^z=*x
z*?HvVJKsO}s(-HY9(aH17UABhgIG2(hbGpdrmqr@=fB=tH0s~I5ic=aztdx0DQExm
z27$G9<~(t)!UEg08~R@VH%9b3-&{V@U-!W=NA(tUy+_agyF+*kO?v({hRToHvfK-(
zXLt{fzUK!fpL@q||6pW&|8LwSVcaIsEuaq&RefI4AANFJZ0!Hz9Fl1$jCIrDxDdB8
zyyON6mxN3vT7DdbIYVsK?(ch5+$tLNLex8cxmhtbte+)$d{9p?R_Y0=RwX3!qlenn
z^#`M5bwBZ_dXW5s;qrB93sS;^vDfe46aH-LtHA}euj{TwQ`{=q5+%NS^G!5fX8dvH
zF;rA9kJ!4r^Omp0V|4C?&}s50=s{99%I=uDcK5}5h8%Y*TKTC8>Qql2%$QAnz4+pj
z@DFC?C0xkk(ymu!@Agog|Ld%lk94^5>%w7QhXJ%nv#B2;=P&uy4Voa*0eg0T{?`mR
zGLb=Fk71)#Pco(D{9b|$*=7C#j_(RTa%+~}r3k#RxEMfwk+ET1$T4bkhXo)MF8T@#
z2@ncX$z79~;7KPoVcdHWeW}?o6hxkXgX^O5854iz1TmxD*_FiLvu_cGZ-gkNFo`kA
zXqK~)QNpHn&#NE$C|;y>RkM@8=xB3mOwHX73R80yCMZUH!%Ahlf%TBCF-G1-c9h;}
z=I;;7#6$YnKT3}H0>3FFmL3#GUJd_v2trGFk!~5TdeRXrsvx9XY<ur?dZRW+lMrvH
zGm|%y;ZLsLM5JS%Kwgfqr^`?WG+<FynLlw=+}rs|Sp+{%eLtl~+3+>>4r@%|Wtq6p
zQ3nqy>m703HDuE2P5H8DhBuK)DzMt-yIo!u)j=U8pc#yOvYlD|5PGl|Gj#Xla#Ybw
z54Qv5pz|WR%Ab^skrdWe)pFfRg*J0KN}W@$$dk+DxQF%8B)El0fK*&NaYsxr-RUNY
z#Mx3o;BreY>Vw!C+ubz=J_{JVk+7`H<gDzR{35|5m@aYnsVQI8ZOW%rrBmu375p}f
zLyw9>53F_3C#NC8>a;kt&Fb=l;HG@+Yk`<_)o~p9(#ny5@d^HK>CB1p_kJK=v-cHh
z0^XdhIZilP5PwIrC1w_XHMv5Q7mrnGL=m@E*<4bynr8Bs+c<RSi7QigN#Bc#3ZI#f
zv-z;BnZvzl|HoG{LEiaYUMT)w2#}3^IKE79Np^7Bqmp}`(AXUQXcQMJMelxt+&xE<
z=UObQdjjE6${k7vp*v&QU3?9^(34F*GW+2V_j?GN6>eBZ{;!NMGCs%R6|Qe*SxZT=
zqTG0PUsxdrhdN7KT)bf<Zp)7c+Bhb4@=8n89cgz<u;*h>m+purrP8=m5(^@iuBG8a
zRG`Uj0FmrI-WxtsFIW9Ir3Ss_XhqAZ%9p@f-#xDHd?yI<9FG7}E)^n~)Hrl!c+g^c
z{FI+fPY=Bc<by9#SG~sq&%AAs>|*PD8&8#Ijd{}dwEN$sD*OQ=+#Cf)jK(@&VHM>L
zCTY<VR21_?S*$4tu(F7zB+?WLR22CTm8H3SWV<ir?OS>4H4n*k7*M9diWj({%v~VD
zHicOUjxTDd({aClurn7#+gOPDJ-Sf{!lDmFekPFx7`ij_Y<kAs(L6}@0#-qT9P6Q;
zKAyqylgbtz_{!;UvFx;2f&!HwtMv+Qfl75-Ju0KJokjjwaJpbhGPiIhSE?*_vSN-T
zqqpyU#L2g~$$Zg!$slzgewE)5XG$;YQcb7!gnK4bvMT)pi;lxB2wHM*LH`jE^yDjB
zULYX0bQ&fjjGYfJHZ7=tUpOH<BbS_2@R?`=OC=?^#2@d*Bd*6{2gIh@G#Op~Ms|ar
zr<Plyg*_Q-U0!y(r2DN;vTB}B=Vvtnz|C=7Pd&S5V+SIq$1AFH^DXp`Z>C|du(=Wi
zt5lzZiekT1B3TWty@rn4N_1JTcq&hxN<sM{exsJhsl+%^%9q)6mOUy{dg{3fbwIJ?
zS3cUN9I7NH(rqfzF*jqcdCIFW<?jm!58N!@W8_1V4b(l#HnqNTOUqV$!mV2H>>;vU
zNJ)}kgm1<vuRv%lKiu!fHdi+1tdA#b9GR!++rjdb=TS_O8tNq|9Ab8KQ<bs3h+X#1
zN14KN5rH4@Hx$*cW|vJ)<D<O=$%jxD@G|n|v#Q_{O)1C{%FD#>Cyuxu7!^Vg*<A+`
zs8RikBJFoo&QmF6$K(=}F)LM%_mcDdkUT_H%jpsis`o6{pt6mX6!=A%nKQ#_aVYk<
z9}oBVu=ZFhbY^nvW2XR`Y5rmcZrxHpwHWx#q5L^il%~F|obvetL608P69Zz{42SBY
z*lHBLs6VR3*!4{f-ymr}e_lyOVPJXd>yQ!(vO~lQ0TWH+a*2&y3iJo3ORf~!S%u#~
z`M%4dvD8gX)dF`7_(bCcpJj1#aApQV9W^aD4sz1d%YKNU{bZZwS$qvrr7SAr=r(s_
zOVA&Xvje|<`(BOW^O4Y1nII5>9)Vr%tUC9klM29zG%VY|=Z~A~?4;~nEt~_w7eo#T
zqaIff1SxhB^wbn)y~USl);Z;A?n{Hx1fppKxs>cObA<B($5FxYfxg*cX?gTntnv2t
z<?+5bYO(`VD&u)X!`Pg9G42iWvWlZoQsXGbsmvd#T$htaP-^j7%tWGU!m-Wu5VWW9
zZrwpko$+Ni#g3<x;>PqJer8Bf2}}Di3Fpx;6$q!q$~)s&&5442O|cve**)7lo%R8C
z3`&g)@%#cBH4O1M3}I#P)jdyk&R+0^@OsU%Wpdob_O35C<H+<_$4AEWYz0VCb5z+v
zF9@*AK1e-1r-bfhNlJ7w3*)~sQ%RVpU}`anFm_P<E0^Kt>*-8T#WkDQ3sNP@Lhbfq
z82v)>)E_ZwBFF)WsWFBQ&<rNpA>;Xax2^CAn;2|G#zkgH4Qgu&O}U(mVj(8A!Y<9y
zJUKb6s`P~#o+xKrj#G6%t|vKpHy#l_AvQKfV>uB{(nr@Fq$Yw-s&83^@jY_dgq|o-
z``VG3{@vVw&x$!D($xf1HI+2)FkJ#|G!T%U^2KPnlL>`uZjwq##Uj21Ip#NnQ2nOJ
zHEz(j=y&162vQSEv?9r_d(F?5`CafSCo$nUGI%!L3p=*7QmMKcYVK{U`HRGJ!fJS$
zONX$E4r4!!tY9(~+B4{F8ahF!(nN4(9JcZVi2bacTld~Yxw@-KPMfU&qR^nOp+<*<
zzDGMdjXhdFHZ>gn;W13Nn1-7^w};M~j@qJwSoB1Y+VIE!&q@S>0#YV%^c`!kpumw`
zTK{W3g;Wy`mmvk-`{}F?ZwLP{ItON-;024VU_Sfk7krh(^3r;W)05WWbFuS>je_Uz
zbhm@+?$=Xkb$}2e_F&(Er962wnM8+l^S#XIPGd%6;?kO9tFvj!GKAJOUpYLEGVI%2
zV+S5GX*~&K3HS;ItOy%xt5QC~RB3;{^ThC`H#C2E8Op>OiP(5S_j+Ryaj;prQBgv`
zldJv<zd2)^`Fa91;A6B`m8QI456)k82LhalmgcA{n#EE2{SB<Tm^6DkHh?<a^rXx#
zrJe{ckyuXp?X6pR1(gPJR0LK*z)t)~J(>`Sf4zNUC{m2;!}ZTKFKaX`3xZChE3DS8
zhxQ~fJgGMrVg8oGBI|gmXqT@2i<PLh!->5i9Zyk@KNvP2ejhHw>tIgWuPj7<CmTTH
zK)pjp6JpvHac0=~TQb^O*RaXcVY6N2bnCDk^g00yk4e(HYcrC_cvT|WAPv#fTHjB8
z{Jr1L%Ay9meV6v@H%SQCxw^Mh>bW$#m>mrf;FfA%j9nb$>Qh}auTU<NpTbB3so&uN
z=Iv19Z;MTRjk9Gd4vwa+%dIL@G8i?~hi*MfP-Tv15~nQX<E)Nsjt+MuU9TP<WDJr&
z>?|TaCDfBO+63K7rP=x-CAk{-U|#+jx-za!V!YI;yeG?QI9JkB<9*-L^fJ4*8dYyI
z^YVFu*v06u0~e>{HHMFlD=e6p5dLnZ&@-`f<oeH19~@Ej9q`1O>veukY^en8);d&|
zOG#^^xTw@T@CZCkE8CaQFsUUo@+eHp7l=r+#FS}PdsHf5SSp}Tiq0j^$a_+wO@5lD
z5$w++l#^GINPr~<?qIOHez91}c*5q#7mf?+z(5M+CrUE*GpSWEAqFTaSPVwDS1Ka}
zQl~DFbYuVvq?k0Ah$!(WzkUt?W(qN>%BL}7Tf2>G$!XXga(cMNX@W{YdZ}r*eC<0q
zX*mibDS*t(oQV4`OvOJi9}LFlKh||C+|Jg%rz{yDr>BM=H}E<Oh(q1YCx55@UX6ct
zQ8Hu#IK8LvT|z`LY<(X9@4HNxhduuGm8hC2ed+({O(zAqo_WzX-;eY2oZZ4h03)8K
zj8>ux_tEzZoHPz-CL-DP>PX0H%5^0nd&Qi!wZ_;0bHcI$+ZgzUJ{w&-q&BKx80ESP
z^8273SQlo8vhRKFr^F=pR-`=&sqDw?2NA36IXD6}99Rsfs4Pm?f}*;;89$*FKIH^U
zIA!@5`wSnk5!T_bPF9U2+V#e#Sz%j4V3MhD_MlOOLO7;CTC~V+mcO0^@jl7zznZT!
zfCBX39O-!*@E8;p$qD}}VT4|U38z1{&)6n$eGt&)Nb1V=ed*yc{k<gV_#2%*!G~fV
zs}>EGibp!U<o%M<MTPq1Fgm(~N@>;ePWU82*KV6MP>rY?JtlrD8viO2si_aH{z^j&
zhKgLLGHD}qWLTT-;o`mJV*Kn#Rg)kPgZMXT+g~pBH=CQ2sW?pD7n!f4Y(5P|+7|bT
zmrtz&J+N$~_yxiRnD<c2`pQW`lZmwNs`BCY`W@Ny6;!D31*%!V94aBkV9)yGSpp@I
zFLv*1C#Te%hr5Qb%A~yFXy^-X!yqL*UxbiE>PvNjD@2~rjN%^*gO5e1v)7?vR<W<o
zkQLEzIk|$)l;HQmDo>CTkpyd&PMDK!y6ZBdnxb=(h}WQ0CyJd6>#2Wy#Ae;`u{=%j
zlw>OCv)0LNgLo3=r{kP&bfEc74>NHN)djUkq&kVi*~IKRS{gShnQM^LQg(R5Sx%W2
zjih1)jTVZY<X(XRNgKz2t}N&aIEX|!u&l`2O}d0WC5EOPC!7J@?sV;2Bn=xU0I@k4
zhI&3HA83)aoJ(iX@6YN*U=np58hTh5Mr2QLGa2{MZ{aHTizIP0na;kGMAufy2pO+{
zj`G7noaTWd#OKqKgh*wBbf`Jg!skBzR)AEkrwm2A>0fp@iJtwH(Y67+fg7}6BBamj
z?Ks<+tgrM~y5t{L*vJI+Z%QWEDE>-ithB)VqBw0gsSYKX_WEoW9m_mjrtP)y@G)s*
z=J~3IdEF8%9ByfX%E8mFGAo$=ebSYmPgAJ~Ic;M`NOu?DrP&vHO$hd5|ML$?dkQ$V
z8Y)fcD9oUf*pFh4*@(te_mH-Ed_+Q{#E(1;%4<5zYp#`=mhA0RS;4DrYQ9oGrK0_8
zJ8H^#6Ywm3tSL*aG)!<qa|S}stHx0>_0U~*8zKh3^q5|nClEb~Acnvp!}B^1XFp2)
zJqX_Q4zwxoe!=Y*Mn@~i;#azH>bZ>i2g8%^s+8=)^bclt7>e8E{R!1|3cKW87=4Ng
zxGHSgYpu-w!_jrX{z2~<0DoBxzGS;hk-*E<1euTCeuxrDvF!UEMkR6-j(*)&lF+AS
zVG~6l!o71YyzIRRgz0Uob0tRj^$F4RgX3RaQO?JB)^)o-jx7zHe~RXAEq*sab(KZX
z27=~-X;;!*jXlX?Si9JS8VS||tX8}WKe~V~hk~7Z;X6|#NL&X<R25*R;w7-6&`u)Q
zv-ds}V@di*-?>PN;+4!E!heJrQa*9LQJeW~6=@eb61#VeAz<@uTLIy2g@UA^o5djS
z^u+V;?yk6Y_UFG9#vzvL?^3luQXhlN-&^iF;kPV>b=$B->2~+_uM1A|oA^^)V;&fb
zrw&P4k5<~s=Sw}EKFg8P8^zu8Q*2LYE}1W@m4|zq;c6Y$b}Jt>+O9C9&y&uCixlt#
z^6ds%h#u()3U+6eDLTY9FOy^3Cw-e#CP%(ck19DN2NSinh5Gn&O|iJ-EmH!2QDolM
zA>$srg;YC|F4+Bo5h@6&hB@3Cxsd*PyATzO=ez|B5A`JuYG8W6a8sw4(`uPj;T%m=
z`gS;(>GN#zIdV9by&P~@5C|U3=2PhY!Q;G;wVG7f1nS11+XGCY3q^FeVvt`61i}Ec
zqraX)9-cW8DuY+rHl3ad-+mEcMdey@4yGSQFG>F@;^J=eZVR7yI&^50qpC$`yDUO;
z`&?mkR>OwgiO_D9X-e>nNxJ^9!2~pB4m3a=5q=WJLu!hHuJ50=<?8ynU_zHs+u_9U
zRF1wl-0z?>ufb%;gBQricKCnXy^MN|?3+z8nb-)Ju=s3cAK6QUC-a?VkC7wz``MkX
zEq7%=yh<^-(l8iG9wIHZV*y;YJ|6_A?QdZ7U&o?3)j5wiZr!YY4n?QA@8T<hy_uRh
z)#og)VBRST(jau4GYy=)RzjfG4`HP`03aq6W)g+9`J+55>=pR;mogJjP>RPMo|e~+
z{~8Qbk}C9kEp+|w%<gb?XrqvdjE^}C>bBxTCmE`1i2Qv6%@%d+xNx@u9sVij2W~|s
zxlg!u3o2X%DUXurd-AwY$2+jKI@lV!ep+58l4w-}oVtMC^;|z3w!e2*hxA4Dh|l#$
zwlba#TY|Q%x!92Z;V&k)h{<9Hw03m=)6vNAP<_a^wdT0YA*+FYjWy-xD|{VqyXjh_
z?qC88llryD2YUnq;nHmTsL?ZA{dUUmfq(~-5Y=tp#7ttVin<D1k6jb|D*OEkM=a})
zBaric0XM8guStUq_EvX8v#592#ZU8GMZBQbV^M}ZK(z1NrL8o0!+MdtK{%VtG)wvR
zswO#*<v!6rQS6V-g6a<6C!Nxj*@>j0Q%$ut!AF<F!t<M%=Z9vA-txeXSMCEnd5eXM
zJipN6&Fd?4VgAUX+-#ul1MU^7TC}g{fb=TjH}bj_io>O~)7IUUi^@YH^jPn$1-0!v
zM}dBIK7arA_2<Jw6RpprOyBTknOjag8NIi5-{>ELDqZ3LB5b$$V7`TZFP?W!>pz(6
zAzSVhv2a(?)I6~t{5=9|Ys<{i(PH(+k3xLcUINOmRle<KeoqJ>s3&4R$K?yIlfV!D
zx{c0l*bk!|!D`7<ZG65rBtLh52!{0D1oPwBdXj0Pt=$CS*lh&G_`Hi?6=}aIN=GZc
z`dY+TJ4tX>EcC#2E1>l)1b5%g$H4RPoEB{hYDrzPI6_@!{=pQ(jB#xWL^k0Z1agm#
zxsCGZR|W}Blsq&~KwW1bg**EJHXGLC08|q}WxjrZBA=^N!g2dwB*Tvh(MG%7!TFzI
zjsAAzzS6&;+}A4|M+)MR26P1RjB#%b@ogerx({ViG+k3+lG7o*Hr}JQe`dHW&y|TD
z2~!BQdAZ3V0q;B36&06Diru(vhiR}02XAXV(Gc|c9v6)Ey$Kp`AxN95=W`uSX)T&?
zy}Sx1D-|(oGjdS-^D(NLe$%K~qCGslUv~^G(PoA=?cSWW3>BO4@fpS=EmY82bGrXj
z=SVi6nwE)#CBBNJ>3&$~MX}jNI7+WnK%+B|LVn@dH6iEi7ZahpzbH5T=4-{BT6*jZ
zNdmAC?3w@Rs5<G?lF4#s{%%6)qAUzdQGpPu#VOf*8IHPJ1#n^`&BaiHpqS{~fwy~A
z=ogDNFLQ~!*jLfM9y26OSJClGOxD$Mxbk_@O-Q<VQ4PEhAZ@k&&#yWlNyuFV2RY&R
zeBABe>ZgN6qSWy!kJwyu$2LWR9*)K-^zzba1@XAXM;<r`kg`nuyFP6-46L+Rh)gj?
zIVYr+bw36PDg4KmFF?#4ilzNHL)E`Y^O{D`@cQoO=Vt+Zp7yBNdeiV17Z%94_@vJ&
z|C}3{pmD!S8&WGvnKb1K+^bT<KbY?n3v90o_#yvb1nep(Uta(OJLjb;?_f9F>3jW)
z8)4W#m`Plgn67~N;V^yVM3WldQ#HvwBSaI;Q!Toh?3Fs_g-Bf68#VjL0crY}>N7)p
zc=kO$)^bMr)M`c?ka_iWu-Kg(AHlKN)C9aA5tvoKq6It0G~jIFG>9C+p$JY@d+__P
z7Cl}fc3N_f9})H!Ucm}GIaRf8byhr!GPaNT@xWLDWf*~0HExqBJuQtjIU|$uBsZ#D
z$vpfl)?d^fNOk?mh+hr)gj=jWeHSlE-d>O7axvCEcFBC5M@dt@WvB-kyjJ6#PnDdD
z151<&O_dm4pHVzHT|n+UU5%nPj=L9{%Qu)yEX;mLjlQ?td;aiBo}D4bg5^b-yTl(@
z9}`{4URDi%*7406w;5~#KQ61Zl5S`CQ`%Zk7LaCD>jCfPFehJOE}Bibv^>lv>DDec
z!fv`tTuOku#htyV!I0_7WXCwR*d!Lo?+-4}7|PTmPN`a~&M=ELCW1{_QBB7CXeXm;
z*J(aZ`PsT8Gc^uHZ~!0jls?Az<gvi{GeCO!R4|k5+vLLMO8zRyu6l)@t%eL9lA<Ez
z_$LOCf4#V~x0D=>lGU=kW&~(5O(W^O`Xa(CFZM-uo-)Nt<SdAvo4xrb!%u;qQg8&C
zXF>)Qi_Sqd{IX#=8nx=DkCXb5x>^yp@)CUMYJ0A~ZGR*PN|t3%htbx{O-_+a5%scQ
zj);P-72}@Ne){W&7vJ|4sO1_KWy&qstHvm3l%ZIwGcOKxq6iTxP!}qvLvjm~IE8qQ
zXrykzS2HS8Q?tsFZl^hGIiRX03t_28hErXxtTJ+~=}>==PUT@)_1@PP+8G+u8Dk<!
zckxor@&sT?PP`eexIS*RkxCX9U+()z80;wG0XF2}uq3imQqL<j4`LXSc9D|bGuZza
z0hFuNY1C9#7#SLaPTe(gc{9C4gDqcpP49If^m-JWMB~2bA^=XT6;sh@Y%UoR-?)Ld
z<isp~3LSq;x25>EB>siZWw3Ie802a2<tT^kxpkD5m-H0FDmgve%J3q_%PF1IL6kMv
zlk`|G6R1;h;9KePo1*pfL?}T<zXoK^Z@vNa8{~-=FqD9zr1MUlIei*pY>kcImE1iP
zWRlau>t&@<S~&d~?mkupVPK#Rj|DZ~j;T+`L7JhVIluy%G@8sz%5#&&2{wYojDQ8L
z`D?>Q0i)f$=v1vk8e#TA5Ta>gxpIa^y&c2XwSepVu!FK`4aE%1JVN=2LDxQWM((w7
zKR?;oY1B1>PeHP?(uJ6SB&8caU4V(CIv(_%V~JStbi%rPQm{n2EMK);Q8`Pdyjm{$
zM;YG7#sWFHj<TGuBvbXeQM$+$Su#X1H!+=~XND>D^JH1VIf0BC`<z^dKuH8m>10`j
z*=d!p0WA(VGK;&g*6+3Cb_<+bHz=X&sOiuurD>t#G-G3G(fnMbGGhyx>*9pINGT;R
z)=_ft0=!d(;!oUWO|juvv5UwBdO+Nl9yyS0Y=n?EMP@aI8pnerIb&&#f3e1()BzRf
zTLK>s1t%KRCZ=DmU*nu@E-QYDYOnIss#<YtD2To(ztZNh1)@#65-T-@V>?eZE10B4
zYg4A>VJ889l3h{AmM5e@P`WsRkxWOy3NjPVQiK~aRtb_?#=CydnQd|~uF+yrVrIf8
zB}hki2eBy$z9^qgO|%N=9pb9R1g`tTe-m_JHK9YXwXw4PVU+1H9v+5)l;ZkP@~=!i
ziN8c&)u(bHyx2u9{{mSCA(as+rKr%DZ)%k>Nm=tv`cD)9QYrRKAT=;Am5xrUR?+gn
zG)I9CcNsVCj*3hkcQbvBpt+Zq3-ayxlSUBfsz<}33odu&H#O>UN{f%YSc6Xy;}Qt!
zpj`d&F@>x}@kqm$#kD3YLp{2~dD>%jI%_Tel9|jIc|2YRUK0@^Mk(mxa6FCr$MVUN
z!e&fQ4T5{=aq^X}55i&PsWqv5rS?VC8AFtSu{f(RqzDi>eo=1EZZl}XJe;e^a(pTd
z!Uo`Sy^o)oO9`M_i2et2u73Ys{?)mOAM$dx5=FT~mWL`rb))g1(E)%gNrsv6s4$j%
zLObmW)zx+Y$^rc&AFZnhnKeRCvWCJ%=;+IpWpk1VB~z2wfH!{Wiwt%k-`P&0L6*t~
z1wuxMh)f}{i`Kpnwd#x$JxYGqEnh;8QICd*(SYMyF}H9`ZmAX=1qJE2Es8B{S57p4
zX_h<De8+^`2rxfCJ2fl*6#wz*yC855T%Tr~aF`MVRGDCClCRCm7$TYy5kjshXU;H~
zJF_i{C?lvu#O%<Fa9dn#&z4BVWc3vM)=z$WMEQEM`xgUO2c;#0P2BZP1*uw#Qff#j
zHdCB7%oSPAoFSdI269{uGwAH?d}PlFv9L>G;>3Am^aqX%2}P4q9(n)&R3qrwJ2eZi
zfo+WsS@WDUc>^Xx`m3ARTd;0mM4l!3$t*q7-ZD|&>E{1nbUrsovqZycVv1a95r!D)
z5Hz@=?ccXVzmf<d`bD#)>~W9NM&)t7+T4=Jo!6XlpfmIMdjaN2AF=N!=;l|XccI7~
zm`5}g{EXZD!KhO`s{XjXSTaW=;aKz0*qA;0Gcic*rXt;0YA<5lR0*6n>Wm6xba_HL
zQYXi9WK3zK1VNbG9N`+MtXW!{R(ozX(He|CH)Q7T%^36^%=@tz(yJFK?Y+=l2e9?^
z!OFBcak3zHDD?0Xc)}Nxx>$vtB#C|tezb=)PyJw!tTHiz3|BRpF0GB<`r%6c?@Ym-
z4jC6tnGc#T*N8p@oxAl@g1>=LCf9(OB!m6WltT5~6ftDEOvV;BkbmK~MPw@_;amkG
z5t!t4Z|Bco*I$A?Qy1cE>L*}2lD#)<{d&==msS{_L)^vO6L(C&BJfL-tKobz>hNjq
zNt2I4g4NLCtAEX+1$i%?E=PqYxMEQgNO}Tf#8{wvP{CIBe1L}O%1<l@3B=TxyGbTL
zAV?YmQs&jF!ZWb!P+u13BRE$PhH34Eix?una_I1qs`2a;5S&EG#qQ7Iiqzro`Z%i6
z&J-@Q<HHjC_hD(~mH4ZF2Kum?$iwUex#v6iFA6krSEMq6{d}}2f8tQ*&*zfdhovm!
z`T-N^YMPxF!Q7I_Qs!NC6v&^_Kx;UJr(Psp%yQW1j+6jP(fzLwxM)G+GV6iD4%boU
zM!atjgk^iPpDZBolfZ(E%Gvl5Y&CD>Nk@rChH~KwzjzUVwgNA-+AaEXv@S#b&yswl
zI;|UI+;J-e5dghO1|D;9<`k;0<Shr-1z$>7xD4AAGwYY>Zgyww!7dnXk%gUFb?_2u
zK8nxm>d2_)9a36R?HOn8#U|&b0s;y0syhlUT?kcX$nxVUmuWHTcLKe?(^k@fV!<6%
zX-8Ky@^RV-_4T+BExr6&_)eHvx>RXr!`La7%0WET)Ec)b<#JQ_T!EyIs55*J4u!oR
zuuRceg^D`0?(W6hq+|Nw5H~bd7$J#z|0sznZsq5>hf40<8S*BgpK?DbZaZU`>Dbsa
zq5~Nz&(WkSbY19cTeM5H4-09_PW3v&0ODO$v7TF2<+q;0h$VVB(_#u@QYi(DljWjE
zhQElSe+)OD@!`@8%L@ELjLK=;12``J!c3(^SnmWJ*19===P3{yDim^XW-WI`i+mT*
znCE^VfH(oa`|b{@4^Vg1pX&Rm*-sH=so|!EfB7mWON-k~>lTB)C_=<2IAGZG6UAw~
zJyu?`96h}-1`P}S7?t=lXRl!<-n<U(P$ybwg$+)o9?8TiNA}ZjB!i^q_c=%T3^l!c
zlN}snCR9x|`(nj~Gmcob7<(N~c}Cc%qWY=R+^>p(zA3^=tg16~_PK`m0Qu6~&z<zC
zIlhH@km^Ovt+$FhzB%n2n9ZdRvw}5y-{18Kd4VjU=aD+24GMTfC$HrN%nH$sG5|?(
z@!gRcO=jw{<W^@ZTOS!F&tV?cY}E{%aIc}BP8J=1{-+NuHS(mq1yYo8+e8e*cq;jn
z!Hh+UM>(**W(A}pl30e>;IDD9-6_b-!}*Kugy9+$Ol^caqf)w+kJq9*DHt(=Y&%Yb
zr4XEs=!~J?jB0L~!%J#*pQ~e0)Wg+(hEMcLBws6+As_D5XwFw;$Xi(+q@<_AYK}`X
zAdQTuqMaD6cc`8Cb{to5G42e7NgOKEeph%pd+(3(bHXX{mXUbA?n{f2nXr>#`zgy{
zERih|zkyCLUAE>i6Jx8H7y*q6$+C>%?gBE$eb|@5T6MW`za<86jFOdFYgu;R2e<I2
z5AkB*t!QqpQ_~uwXJrCq1i!3zQ(!t-8DGRK6m<ml>F|xH2#u?@$TYR+^`i8lMonhw
zP2|cdlM@R}@>H^vVctce{1vg|9kB?*mIjqb3~>v4ebvKdqjZ&%gYe^BW?xSE#+x*C
zU11buJQ4XMWvghw!oFEnvZY~qKOjcG{Sj}1Neg%=PK`4vabMYJwApFs)X;FXkbbqU
zq}iM3_AzU;yvy!dd|6m5jD(o{@N@m<Ib!aroS6aJT84rohw?=tK3O+=cC_&8Crn$Z
zPfJh|2661fCAQGgLZd>5*#i;T3~wDXfeLVe&d0_jJ%=pmOD1E&^u=}C+!zN@Ca{`O
zba|vog+^?byLXwU{D<#^<E51TRjHY(Q<a0()S5l@i{cPC<Ui-xJaYbA64-vbaf^rt
z4>qFqq84%8B&WtwwYkqW*~cT2@v{=77!XPW)X0e*zn#!EM?csyT0D5J*ev?F^u#bN
zq%*Fc#(7E;g+t18SR>rXyy@ju)@2CR(samjeC6i8ExBBGYcS+FxZ)+0fPun##>-iw
z&B!n8&bSpJCW7OJThW0^`yCC$x)CQwLn!@daj|V05nyJ6-3835b;%yZ(q9z3Rh%tV
zOWeBj&kttMec7^e-BiHD;rs)O2$>bo1o-k42FRJh2(?t|_+*XX<$s!N|B=M-i;e3J
zpu+RR>WZ3)cn5j0ybV#MCS^T~wKk!Bd@Ro}BKvH*=6gemYIOyRni<FMd;s=brs@<2
z&voCgA4SxTwhE)PEfjGlkhw~m;w<PZm5vMrGtM#5huFuW;Xm#<6k5<wZoz-;`O%zA
zo(H!>%!6w@w#$BwV=pMI5k6&E%Frt`)baw3Jj+iklg6o7Jjt|%?G<CriOMem2y4s?
zNAt#o<^edo0no~1-h4n-G&|-uNcbn3ucmU}YT54^3^hU7OAw1Fs-)CQjU=0u{n5I(
z9C(PK*+D?15xzBxxM9%tlHPpd)AMRII&VDR_a9&Q0kU#G_5`1o2wVQADV{}$sx9a7
z<#wiQQ{#>wuCeC5+)mFd{5?klB^o1Yfn<Eg(6Eu-03cOGk<Kg=Ld}_8?n2Pd#_sBV
zKY0f@U9_mQCQbOTsybaf7EiF99&B&%)O_RgnAG@23tX@Sf(O9b2}xPt=MP>NhLdO&
zf71eGobr5rwJ#mw?W00@yf#p?X(0g;)OzvNO`?{BQzsGTgu(&WSLPKLEK9yd3-6_8
zR4%=j1XTsB(|FnPVL_SWbO(f#MZCkfh%@GZm{3aE|595j1d>9}c~G?=cu9;l`~OuT
zl@g%(5s+OPA9I#RDG(X6W3!vvH)8k_B+Ox+H_DIB3f--j&T)uM_}co4s<d;t1Ld!a
z4*mNHKm8(64?QsY=P*xHA0L^C7Csr==8#*}S>AuIOXS|)skHR<xY=j9kRg42=3{oT
zyZ2o%Xfl4jx{8cx>`hJzN&4J`pK@H6-GpDSC`#2u6)fgL*64_RfWjHwK!z(Es3-CD
zF!m+l-t;~i%Z#_1JpCvHVfA&0WS~-RnI$u<eXxx{nMq2JEy4*8#TXglCFXZy$ng*#
znu-cd4LY?ck;loSJC5R<A?i2~zOjALwZ4og_xPh<fRxBv>gy0GYD2wSk)Y_y=3diP
zOWqn^lUsK^(+U~EV1KIn#(+{%qPEwSsL5SV^HU4Uhy(++pt|b5Rk8y2HV>aJbs-Xb
zPDMkxt6r9(r$|Kit-^|g<<Xdjg9Hg=pEMnZVx#MRTr$P=4SnV_Hc_~#Zl9AVnuFBt
zGT^4Gc~oTaejnLPd-8NM!>xp{LVh>6Tc&j0rlJbEBwCi;Qi^3(ZD$BA9)Nx=;ggM1
zr9!8kqK)70<P;IM=13H0xzlwn0F43aT&sR8iq!5*e9h!8o!eQ0BnopIhwaIbJzQ^;
zWtQ*QA%g<xcSur`7AEl9rK)a>6sB}yH?0}Y#tKG@XiHK@T5CWRbPQm9dwnI?Qa=uJ
z2epx6d{PB05daOV1G0!DocHGz$rZ!{6@Bxm)h^Rv(19(l8K|BMw$|I?>i_U|R#9<v
z!L~+%1@{ITm*DR1?ye0qZoyrHyF0-hf=h61f;$9vcMr~gId`1L`+VQl*kkRdT5Igu
zRW;{V!rrj+mE^P7W#86+Ks5wnF|96K=#tP5X)0m_&VUI8<7ot{$1BAj=oj(oIXJb5
zfZ|kxbL84Jw2Z>)=ZTEz^_v+kY<&oxPUKd6B1wWWC|K><5qX?`;pK`MqsQ+ja1~sX
z1ZVar3%}G%UV$#^%eKB)nw_`+nQR%1fdo_`waCf`*;1~&>{13Ub7rNQv7;q9ScqB@
zQj!hZDLu~H{sQmOjN(Cb@%I7PD`F*%K~dQcj7xgtroy39U_)pIMS>#n9aGFc>}%fH
zPf?z*?$#Ioa{wIMwWFIKQ&}F*6UjK+CqMX9V0(Q$O%<AW0ODM?XbDwsAYy>wTqcyS
zRSB+U$aeha;TsT6(TS<_KX(G>^Z1vWI|NGqqq#nUn1=itZ4^Wpu4{<gjdebMrWx7%
zOKRndIVfT(JrchI<~{1uyjN$A{C+tZxt9)@C-F-$ckuNU@X3tlyTJukX4hm^7h{SZ
z3yDK?0g(_8;(fB|t_uw_oifD-mqKW&xni*6Zf2U)U4*eTzLxrOQy}v5lQ_O2=oi*(
zVL#S2L0}NYpbukUD0w&CV@};3SNi1>cAy_HdOMeU*3ZDvz?|o)i!5B}C&)nJ`p+AN
zl;_7V&S*XjdPLLtA9A~p0EWMYedKtu+s!^<#g=CIEq|3t@kUI8C+Vjdmp_Z^w6QU`
zujb#5{Q40$c@{8#k25kJ;?tP02KZI;$itMSFIqhCr4?BmC6Ba-Rrx?eq?}W84pn1{
zyP;QpmzsW~G3^6vUoH>02wE9+|9A{x`wvPy05x#EZ^bb>@M|mVE?3>2d(Uq3McMd<
zfrRi;Mx%~<t<f6yP1B&T02N+kr-Egcpd8!xsD_pOArcxxUY58Ep5dH;=?l0U7ljpX
zyf)@e|DqwdGKN2H#li$^nUaIdL^%D>QTc8%7=&rF6UuHuOwLd_vMFQad+LK!K7Wir
zn&m3haYfmKxf!%P3@#|YyeuSV_lxYaA@SJicOb3Vu>Jmi_Z#o6OOZj+ZCG$1`l~|A
z(oo%Zb3e^ak3lmBB2pjwq5m;MQLL`Ah@SN`;`!xTqfP_*JrQHy{aR}nV$rN8S<&M;
z!FHt(4(QSN9SC4iYG^QO_4d@%`SRL&7k=BccCHgw#30FG+N4i~Xz_>fpLMC^r<K&9
z7>Dm|lh!b|sa;84yC1ibicUpS_63>;6`dX(B+K)9G6ZQO1D<kj&dG>DLn0#-?O7}@
zRE_sxR|!x-VaaqOGQ{&fned}*zo&=%9B=hg)d6FvyPaj*y8N&4(?Ijug5Ta?R)U^L
zsnNNc+tQ-UCQ_7<!_gOD<6+Qgb&w46FE03Lt0k~Suhwq!yv4zsLy`SJsq@oWZ>?S&
z0O&Dn3y^9PLlfa9;+(rt%(sv7oipZtdkamym!lh4PH9ZOW61yrz>PlR7x#P6j;Zi5
zuyHZ&FSK>BmH~foupy1q_F(KuLLo%FVRtbm>N5RXL+NygzRfUeS!JvT@%Xv!)f*Hh
zar22Uf#Vk`Ms4Cptug%81YOVndN;_&!?G*3MFj=wkLBQ;u$K)n+h>-I2b-qwihh%0
z8=zn>i=@ng%7VT8LzV@EA(Ldo)T+X~M7|`|x}A>D{qBg`?oM)61a&#JAyJl7GZ^1G
zEx5~IT>BQ}c<jZq?)wajDT}sEylUTlJoVy;>gcx|?Ku6W)3`O<$A<Ox3C{IR0GS0E
zbJHJbY@Pa)&DaIrfNty~I$mHgr@9|!7FnkAm%EpS^GhQ?{2ymOhoAzQJ^wWTW0BOF
zf$r}L8-6YmG7*)LM;a@ETKMQt!U?%`!d;g`IIz#*1<<3Hb*(KO-HQ654exf-_eur8
zVFs6t%(z>kHL@)TOAqU;z2sg3Dz643(IVJOViR`S{hVO+d5i5+DT4>lk26B_0S|^}
z3E_oc?5$M|d}OQ>fCeiPG(L@VBRyFWcAb{^6Ut{L6z>S?5qeC|s^;U(a^kx#`s}Y^
z`|#!&&>0_ky6WvjNlD6n9pMinjDM|h-W{Lv`MS^3A;vG|VJhKq&?W;t0!4yZ626l$
zS&s-6^4n9eX&rXX1Y<NAcFAk5PIWqSSXdAv0Bn8k#_7*iXoXs%=NHTN*Uu76!`T?I
z{Q?zU5=97UlB`rZD*YRg22)U;j0@u-Zn2Pr67@6KA9-H`O+xuc;d5h=t+xs}f1y*h
zWHw`=xi=1LU*Evx!WSWHV+huZs~bTaaU;}^AAhU8O!Ki6UfP52y2PqY%E_5ZEUNF<
zo7?Hk*OOppPO@9vRCAnnPg3+&8W8re8==5g1HKw;;R4rNL6|`Ii6JD+Na^M!6|sAI
zA>jO^(_945u5m(mJLnJQ5TTT|y$Qwm0!JEr4Zu&gh;ybs4kKS{K}xKIAdF6Pf8q58
zhdK<A3gS-SSJ-g#6M%iBx-<5im6m~&aEP%jN?dX21gqc8K?x^kqI|B#`lX(#oez2t
zf2^zh8_IH-1`zy}h6rzQ9^ES|+)+M`-j=yuRTcu5;*v6IfvLydfT^w)ii=k?f~(Ta
zSo)1i?x*Y&&Uj`<#!s=#q|DqfLaM?|42tl&JW8MhTi-koPOQw-RRihAr?D%p4;DH9
zK^QwI4fYY)i3maKBGIWH>hYMc5Gom}e?7uu_pKp7@icj`AxazJZY#Zr=niv9m^?tZ
z=AbXH`(2y(cA%Mg>dDfd7ivg(>9^Ce%aYK)4HM$;lsTUlDoxr7Wvf<kYNau^V5j1}
zBCb<H=yk9%^884nVsd3=s;w6V*CQn2sqXe&2jsEDP1ikLxig$=Kv)Tl=J~OB=D8KF
z_uvX1<rg>ty`JizndPtE{wAsY)hW*a6`qY2lCrUa=JWET7kRpbM*gzYjzCf;f$PFb
z%>wc1vFp$K)UN|ieM+AX-m5qK{66npc?WeLds{%u#l|lg{=CY7z}@bvabq1C%3R;U
zS?(a7+u5`0y*+KpTrH2!9-ru%!TWEFURTU%70(CJ&-B#~VfC>(&=v0oe-a2G&ucHJ
z@)+h5Ys;g*4*y6DzcCq^RX;396mw?;9XS76SJEM`!=TNTR9Vkua~XOXeKDU!SW8Q5
zcIw5s$AkiZFCR^u*r5W{7E(N;YNoT3<0LWpo}VGu{dZ3?;b@jLumh~{F`{sZ*;~T(
z8(CC@TlWqO{hf1ns)k7V#x-kaHXzEuD;PR=i^wCsXG4I<9p)FnQ)9Mk0i?TxLl+?A
ze8vzu^fbsnz>Q2-$_(KOES*R5mb?gra2X5hg_BlJDl^4B%NJ$LNiK2;R1YJLCT1!#
z;6)Sqd~Mv*`|?KjEx0BQs=RQXrRJ+}X7WG>SkN7q8Kk{+lF#E$r@(5=ZXCTXW%?EA
zukiJwQmHcQf*zLNK#?L-LDjf|P4ua{e4&QUfz3+s9Ob7$4RA<N<r|i`&h*J`xvkm?
zYL8{hmshXH;0e~wp#DNm7XZvepXOc;?+4jg*YfTFi~=CFbN4lzjqbp`Y(Al`emD{f
zpt{XMt7JhNicOam?6<PaX1PU~lb@EtY+SO1Kc<5OC_}ol@roZx%S;=I?8KBQyU_0z
za~AI(*=`2tP%sqYLhLjQZKc12T(dGo&MO-1z=wn3WR)0y8veqTrb-hNqC9{R4X@Qm
zV>{riraG9C#>{l8?<g+8xh*v%KH5XF=TnQq;}|?J{%ZNYyY9NT`%ApUB7*|@<!QGQ
zbnQ2D_PG9ux=w=-rbW75?>@sG?8Trwdq9Ce#;>UiXJviH$JWca0T-vuEvRaMnCmFE
z2c8WMw-7?z`k{1W=bH#YK>%2|T3d>CDuwQJKf2+?Tiv!jOrA-;qJNCcOZR-@mnn6$
z1!9k_t<gNitAdK6LdUsmsL{WHZ_){iK=n7}nU07kP&%Qrj`dgmlCp%;12#q-P47Xt
zF`ORnk>SdAJ`H{xQDZjXm@3H+G@PATa|DP3mJ6caA;r--wHx~`T9Dfbc)XNYL;ZNA
z8?UbNG*2}{EK#MVXb~pz%=^@p(*WZ|aR?8T8V0WgaSmx309CR2FO`|Ug(USK4}OW7
z9h@1D6eMyRybpDYTKJkFz0jPGPacUOPN_D3X8nLFaQizl)X0S5X^q|^hFYobi8qXg
zSKr~9J$BDzQ?MyHz>|v2iX2l(;@A>G*n1jk<Eij7OrhB2o*X4N=^J~wYs#6G?fL?w
zl~^0c`ze{7Rn0kRTFmQRiq`yif?-?aT96=Os4o}M#Tl{97)bmD&`o)Qpzc=aCcN7K
z8_p=X@C<8Z$0S@mgw267t83{@qr_O6!;;A$z97dpP!+<eKvk?woi-9;t(tLLTQiy}
z^#{}3Br(a=Qwm*DTfpf(x}!P)l<enD2qwjFrxEuM*H2e__)M_T51PWIk|i;<<go~f
z_`m0>5#mtCl%&W)8vI#QO4ZaaCke&{J?y`@mV5zu7y3YV7+n*@zA0$+M0dHF{>e~1
zSI;Gyrib}WJ(=o`O0Nno64J!N?@Guk!KDUP36*P9N$6>2*J9=J-rUn=)x}CCV>LCg
zJ0o^V%u>&AMQ_HDq+~zDJe1L#{Gq0^@GKcm1^25L;TEF$Yt~g^wPr7BaY>0Nz8P{#
zi|4Rb%E95#CSNX?S?}hHIQ`*V80~B*6^+T_N8=}OMwktZwD;3Gom8)wJC48|!<p>>
zc4SW1l%vky?pOO1m1l)kIcvr!$o0kGDY@vuAmS(!@p7oA^kY7f3$Wnv;hzvkb{U@J
z9vc>Qv)qT#%a&@X%+Wu}k_hJ)X_aKw>TZ9f$QMt+Hjmy;dP;Vcmn2_UF*QHe(}M+v
z6nz>3xCU!rL|`2Me2TdnX?w|ctRjNY1L}}6lTPf*#{7uKq{)&wGyL84(rt@Nb~wGQ
zVh$FN^iH+oSt}Bfy2)wGXmUV*>4HVVsK$;83BTM|rzA@nm&bD7OYJbA83Co=2`u($
z<WI>uB^fl_gU&Yv8s%TID=Vn4547D0Jv=nVcxB5HxL~9$=x;pMB3Hmv6F<ylYm*dY
zBBh?k#r!|#Ka|tqMr`3pI=FmvDvM;07P%z-Ou$O&^6_&D`y~zQZSpc)ioKMwXGAAj
ze&A7WIZ7!hMm-I^Jy$tBiJ%}Y++`I!OkcTKk(oGUA12lmeH(UXhp$y&7A&MY1Ushh
zJak_1qL2mk!~IdWXr~UM!|<@?2U0;l4^tMb0P|5%FIz7UAG3X6CY=f<raepL@o{wR
zJI1{P2IrqHO?f#rW!nucG7zSUI^RL;jj6M#yfjRJU}4i4m5;E^=u?VnUL4X2pWdi*
z1_u5iFv`nB>JKqtE@I0t!aHRyO=jz_H2M?<8+d+7op$p^^&HX#Ir`{anqOS2wq*+9
zYRXeob913(>^4ztVNDz-T704FKR)gLD}V2Zhnt+>D`E;E6AL}TRJi}FD<&#AjmuZn
zpD6Q}BGdq=m)qJV=d*tf2#ujx2@@7+&BXgFi%K0@&_p*sXMdM-V#mU^a(e+%+$P@^
zu_$(+cHk(!FzhpI=+(HVws!4~<CIhXS=2Mkq{*Zfr`b10nOeY1Il*B4L1pjD>Q$*m
zSb-Dmed`9Hgbo+i1tn33%8@u7$4th@lKdb^jgPT+rNc=TA>c%sfM%&0aM#emkf8B<
z*ROWlA>wRs2V`+k@)Va-SK_ka1Vd&=+ZX)x8JM`^u#oR&k1T!$gAAoMKBCdW4aK%8
zONvV*2yul=)k+<NhF-#}atcyt=^mQLCi$4V>Ri)w8?ky|!tnB`e&DGlTMRs3)9~{;
z&+C9(jGm)Jks2tLiMjj{GE4r1vtBRFu>=W9%&#&;zX@U%&o^4M_zz~vwilEmYf(I-
z)KYNMN^4pH1(P|ofY=pS@|tRmV)}|VZgk5BwF*Sz6^I>lQCpP<CqY9Sc7tdfH|PoL
z+<$=@vthGzkQWMn?uK$4T$sr6)BBeLnAB2^3S01V`{<9NXsj}4rvkMC4QW@^qe!wZ
zvz?G0S3t~MUgAf%p|}}rnHdx@6{38abc$OBQmJlWl2MUX8M2oat{7no7^Xz4X(?G9
zGE!{50O5e)%JOzr82-xcSwuKC#mBq<GgCAtR*JXi*aAZAa<V>fCGTQeI@?%9C!w)$
zUML$Zm~S9aCEYN4oAwnHKn_d|X7>Mvg)%<&44p#oU>I{Hz{GR#Gnly_5u!0aJ2M$6
zoaR#;bTElo>E-}s5T=Ez#A_hGDEH8udh^z?A|VwwURatdN3ocEQ!GC|ui@od-~cQg
zHdDb1Mc8Y5o4<wVi&UW>Yt<Riv$0lcIo0V@PBPbK>htOUVN9!PNVSl&Kms@VV+4S#
zC`uKm(+{amC4ao&f%x2hu4?;LoFc7H=K*lKbdcs@Sd3qUPc}I!L3oKHU4hyq(quPM
zWY$f&AT4<>1H?i`Us|^sUOpw+5m^Otn_ndu%H>BT*nRAgMYNb?>0*&s(~>?^x}iq%
zV9HN&#s4SeQp^-{I>u^WizOOnW%b&}E3L^oj8H%htZU9kyfBpeW<-aUd(uZ9=)2i}
zP>aLqJ<}=l_uzvcwy6Fywt%Y{+p(LAN6Zob6}_PUpxhBKhKGI6QwEi@|AU$k*@@_t
zXU0%54Q|K1FWWuCZb*^(A#GcWcCtYsP8-0EC1_&ap{&xE&B2hR(#dx{>efttJI!mw
z$M9906=5Qw`5M-J69zz1?gz}YTu#w(R-`>|-0mZH+5Ve}ATpqYRDImFROnBrJ0!mC
zS@17{&gKTEMYP{ifT}v3eEeYPxV;i5Od!2taaNUP=&)YtZB&cv8oa?zp_=VCGJ1U1
zK<?drK$9O)mxXpeVc-UJ5r5`9P_ED3Ace$+p&*4Eh^$q0eY~@t)IWCr8)w<3Mhm}G
zr6>a1+U;k1H^CskFR_$Xx9Uq2rvWh~VbiSnu6E8B&gZPIqA=-`+r&om?i(fCoHy)x
z><DN|iq|F<=S@<^Yw|EDcoTtx36nF7rUu9i#zl_P-L#;Nx^*1z1nYP`*J!60{fVrY
zcVl26yje3kO_W~qFAXNZOG7nONkdbGY0}2uP^qSJb>=_~j=n}jFjBg#TvfASB1-Cc
zw6XDIWyNhrL;{EjrMsX51%yU5PPbpgoy2f?op43)#dy+zEwruP+d~FS{gMHoS0SI%
zsZF?-*su}nghh5C6)}Xi&=j#%jV&V^mp#a~qB;hvNB`P)>nj}G@ABdzs}+S<#T{y%
zlt^|_6Ew(?n6Q5MZ$F^w1EtBV=HuwW%gth$iB8o61~3=>KPro_YgmsLAiLD)^NBYe
z0M-Lh;P5}>lNVStK;R&MAu*GZx?K@JLX^B7kJ6#y5R2Sb)5={U`k65}mqQ1fWJyBO
zn%F!#PSzwUc249j66hdc1?1=y-Ve?$r4#epZ93KB;D9yoCoCEiD$$A|i=4#32P|uL
zA_Qpwl$@>R$Feh*wJ@3%+|5FkD_?_+Gm8_!L;X_+y6HqOW2dUL2up4%<L6`s;gk^6
z*Ya<s6XTLGN!aO7KPxx_C5FuFJqF4WrX!;FA@TH1eZ{Ic<@11)0$VA}^S1_5I%rgr
z^0Fftc#ekLA2c9sZBS8W9&289{A;2`sRfx-0tv(AVEEp27g~JhwVu;Oo&-FhNXT;?
zrPGl-zd{GQDa&<k_&TY&poyuom?3?f!eFUE{tz2HY4OfGy0&HtGT{DiLm^J;yKSXr
zzZ8QqI*o-U=Kx`{skjpEbNP7GL^(MY-gnw<u6r_p_MNB`{d70hD?x(^u*0?(vo7Bk
zRTycwYWdxEIDhdD(d+&g4zxr;0W}}K7_C;y>v3rt8m{_3xz!l;ANL2}i|6*3v#J*;
zinP?{=LjI0Kb+^1m~bh2v&dBy636f|#aM;8xf7Vhk^roU(wKOZ&xZAbjFYYfMf0f3
z5)EO?*>m)X2CTMxiEe_fTt5U&we`K{mZD?DsBK<6@iR~Q3D>?Z`wa3!doxz6%)GdM
zlhuKf`|4L#9e!(r(^%uzRxuU{kBs}J>t`clyo=1#Xo3+HcV;5hAsC*{^d}4w?-Ju`
zCj8&B*k6qNLHE@d2aPyH?fXs!Irwc15Pz+geDzr5LJalrr~JhJO$-}DdAFBtLL)|o
zF*&TrKL!|WO`O8X65ZF<6bU<h9Jq+sfT7e-b{fo`N%>y9L2QUvF$_-}oY+4<qeLj=
z&H=;2C(IBmq=EucsTCAItX(Ms#4LP<*^F@IWc{y}ia!(Jh%^|eh03e>z10h#&l^Th
zssfrdr2K`s<i!%B#jeFmF)EXdQ~yL3g`hGzKp`KRSu2Lca|w14J-P7)^+$;z;cj8H
z0@AP1<m|;zUTTw7+zfW5bMDnlK6PJN`=~mQzeqQFat)c1k<l$83L0GG18AL$^pK4&
zQVY0I&d)pi7J|v3OpOKcLS19rli#w)<&)HM7T-2gnS$9V;+x64k%oS7^1C5Pb3#G3
zXIBgB1!1pCosEJkabgI((mq)#k|1h&Zs5G8HVr9#I-}-th&a{~RhXtrdiheS=53ya
zccGRa+K;C@!3s_nX~wr;I!=3}-Y{So9|s{w+k9x5M*NFg3_ef-P0*>LJh-zrEq9Za
zApaYr1=VO$nUE!BE-CEvLhA-PylN_UK;Nvy6Fqf@Kvg~ZCpOcWx_ScD82wA;NFKig
z77aBn1yEi`+lWlAUrXcoq;#ROc}#5x1&&ZO-X*|f&c2=5oCdawgfir(8xBrD_Dv9m
zAqMhZ9AEJEv71vzjS@f0jbun^E?*;iDg#Ptv}?&XRN2%&#MZ)qgs+Gv;?&Yo5ihgy
z@Y(G+lMIRqRLUmPBvccNF`1*1$x7P4H1cU<N|8(cxM<7uB9EI)ye_jcU!nptsP;d;
z90x-na$vxa4s1yJMt^2D^Phrxk;q#^(R<vwxMVo2N<XyiKh<+6l2aXZN;1)@AbQqW
z1)D5#I?dGfSiLlZ6#9~#u3r+I=XAG~x>yndgw88AQ8LB9cPLC)bblfg3%r_m7n<xG
z>&CkA9@`y-m@c=YaA@lVWDj<kLzG9}9p`faL%I%W+Au+}ilFMb<-!>Cz48x+(Bq2S
zG-twVK-AtrpW>TLZlZ_hv`1+l7~b&Po59p_EBVdNg4Zwj8rS~#kcpjAE<>K#XBoDQ
zP=e^A2-gWg_2N}dQ$sc;y^~rBSWH%64qr+IP0E=3_k^5ctcAbkm+BtYGYTa*hpyw`
zc6B|uV!8HjsJj_*qI?U^;czkAmqev#&4qqirkLM+V=xrdA_K*_`s(Caf4N3+5kr@M
z%j}ib=<ts`d8Yh*3KmYuR?gM9)oZhL9YK1Mgr|WYZgLK`po9|<wB9?Cu0K#Ni!G%t
zKcp_`k6hU#>O&CJ3x_olB*ha)V0E_|WLw+Cw*t;r7%0HveI!k@C6A%68NPX4UVy_%
z516+6Rj3n5puwNU_X`3fcRC26oDXd3&~-U4j5*(*Zg$`8_1*>}_Y&s1{V$w|;m>sf
zRb~A5WA#0yDIibl4uiw!MA;mCe@k%*XWK}tSm-;bof<Ujb0w;7rK>e7`G4BhL0Sy7
zaI)o8Xv=>x&@<*ImCsYxLDV|-CwLAu5mTCu4AY_0eK03@8@_x4PX+~Q1uzBh+lvf$
zx2&L+EDOan+e>XU{ZA{tbjS0PAt9A6;C}}0uB-NYodPgh7bW3fmw>cwDEN;4CEM@I
zJb!7qyNbx6_BoGR@>+JZn=(4HZrQ@VBEa6xvdN^@JJS1=T+oza1H)<(FX_xln&{Y8
z8Rb(1>pU$wy>d2z&=*Kw=TGUTHp=;Ov#+f}a>y9BzLg*JI$O_IjKj&yHOOYT$W44G
z%*SRIzzo|%{}*ZoI)+)9suHZxBR2i*9W$ncsV`s1PnLE_`4X;T!41sGyp!LR6#0ds
zxTiat662Cvk@uAnv+z6c%{j1&O`iL>HnX)YL3nwnhXh#}qMPrr5}x<>8_JEo=fgRe
z2}KO;rq@i7;Fh$;cdxIG0f!udkk>Qy)^6<I?&%CaiG)z-r}vIVYzr-QfG!ClJPXG^
zIb@AGtCEIc-K@P21q*i-mK4y=ovMV{go&n<&!}59>`^OCQ7f2bN|JLp--gi6m9v{N
zpUux<eSk1T%%UEp8DS%wo?E%@`(o4gC@A)le01=E^fDob;}vaP{eQ?(b1B?ux|mqe
zm3*{ljpWNy<LDwZbHCSVJ|HQwii7i$V4~-l9c52Pzb^8Yr&bRqf0oES%8G1u4~wz-
zDSpa~VE;*B^8lIo&U{B8&C2X#i<*=Oo|HLI%z>pqJNeQ>mC>IQaZvsBBtmh&gZ^eY
z``G=2Wk7W%TZ+!u8GVl1G%{O7(OSP7CA}E4GO6f4DCDY$&C)HRgp@|jcbL4C@@k3&
z7itIATumH|Z^+~MbUvK@CMr9B{)0m8BlTh>OJ3v?0(@nfjW~1xb5%xa04j4dqlG}R
z-AX3_Q84&-PIV>HqvRo8huCq{eHWw@P%-u<K|#0L7^g@(voA?6p&ulnu8KyV{0B?^
z>9iQ@pbV&!aypu7*#4ypW=>19OK&76+{6Ts!Bx#u7RwOh<+@=d@TM2#*9VAJTSt+t
zI$NKX7G|ozg+I=yY_61z7*b|=VNr9(C8;gv5;EpZmDt70)^0W!t@<ICl*T$Me7^!+
z%FLKSCvDm<*-pzKFPDas0j!f<_b}Q;Oa=@emLCio3Ku?wsO3~gf@M7xe|RuMSC~bj
zp?^(KV9iy#nbuh3s1lUhu8lP*%T*|zqM}wCdJ3a~;5w->=L;xGQ?Vw9C?z6A>!y{v
zj6T_xqfo<X*J%N1DZcGVq>(YR{Cv`2OO${bx+~-j2~iD)Uj2l00utq-!uiaO{^1(k
zyKlFIx}M%_av!d0jc@PiYXo=Vlo(b{x1OGgGIPVzEBHV)@^8bNWxfE?1bMCY&Ko=M
zNeq@lx4HB(t0vA0h=QNJrd1^HDRFK0b-3r*61K2@4fT#)?&T~3s?m8~I@kd4J6Vsk
zL$rI)1I2>jfMOi$Pf77slc8F4xTPoD;8_)$J~08HZ8o2$*kf{Mzk(U+DAV2Rrq*5`
z6SMGWyjNKl!nDBeH9w5Z<_7m1>pR?Sq-q5H0;qy5fE})zeOiv`>q*Y0Y7WG^FKdGt
z&|8o+*Zct+xy8qQmti$mx~MHW=Q;UfE0`Olcwv9Sip2ej@#z~y>`;ENabTBGAWkc2
z@S5i1$ymhlNO{$FYRy;QeodOn=#NsiR<CscVDpwbYZk%+Wxm2q^U|2a_{koMa_S2(
z9M>bsw_bVm+o)~WnlUI0@2AkpLrQ0Q-ld($!F{)ky-c`Nz90WP&T)q?(nNeUL0#L5
zcucrCXcs<}KEnZpIl*_3yS5Ex*+H9(r+$Mv9ts+nVR5p5?-uSsSMS)166B(3;Lnq#
z#yu+kQi4<(i)(mJLF*C(cEUiKls9B_32+#Vq>h)w7Gc7(nkc7;<ps%OCxWGsGDYTk
z!}B4n^B;h^_xjb1gtl7W(@)2HqlElMZMux}aYG%4Y2fSbNw-<op{~92$`MF|I_8&L
ziUs!>PR;Xx-%3j97`M%4%x?iy)3r|OHP21$PO0=!)2_lUnn4RJ$g3jjN)4VfnQ}H|
zxah}g0AD`6y%JJ5CQ_sf{Cg><c)A*vDtQ@Vk$Lkk8S+j6K87!rXi9mZ_#SNi=V$Zq
zhYe7Yq)y+gC?$ThG%@ZY`ZDtr`RP3E-01B2|8u;GdbRU&P^v;almR!amP=T*f6ufG
z#Hr5SZ5XubV6L>p3l~}b<zlI_fw^j_oOzQ!zw(~m#;rg8O8v>oseLvVr<uq=il4Wp
zM6BU%FB4EII6|My!uYX@Xc`uDV3bkwkIxwc&1kLj&M~!aOB=y>$y~(XQ@%TXW7iQI
zb%e0hOQj%8S_n0tAnf0h`+a(XFfJVGhQQsycIvqJDB;*;DbSdB#oeZ?xXT$q+ic7K
ztNkb6Mq&N=_9nN<NXWN{ZpBymZS!$U&B6tUh<>ywTOf88_yLN2cA0L8bq22%#RwAk
zf)4V!q~9+v86V4I)_v~(SwhtAn(H4n&g<6NBl875D+qB{U|QnBb>X~#S2i47FLd{J
zZ#EqJ1Z|J`ytj@)hR6LO8SYnK_gPgV%n7%ar)2~oA<7j?%*g9fB?d)*VtaulMkF5~
zs)APpCA@pA(UyHEBjtoBA6m!t4A%&mX+#|=k#u!0%GcS7Yo?)7-hc24%4B@{FF|Vp
z@g2T*<G;ijDVBcJno&G^bvwr>F=x!tOQ|MjcKk8Y6`{#Ukysqz4G8|h5o&9|!TO!{
ztDlKzsgFsc{&&V_gGNiW?ePI@IvZo5F4@L@EaX}Uu_6Vng~VIuwDhoNX0fo~0?y{p
zXf1&WoSP$w0{7Pw`j>e1QW#KSnkXHA-meyagKrG<)*w4H3V-c(ZUM3!g5c4p3XJ?5
zrtA<rsuG=zoK*j#LdCu?%#HLWimhn8|DaIF2(6VwzsZKdIZbj1>fU%%oh}L&@s5<>
z)Ryg_qRP|h4e<(a{VE88h=<ooC8g0yFS=vi+%-m|R<u*HS8?m7$kU-%6pE+weN%#P
zxM<kloFG}Jz#|q#_(>Zb%i0Q~r^DWZ%XpVyBsg|NmDmV@IxEZr-EIo1eu!0A=5nq;
zF_Q~eI9>;ZR9ij*&KgnWO4ZOx{`ko7`qUJzL`F5?&m0)y5^TBx*-z>!5Vrfu`O`EG
z0Ymx^ukPg#P|XEh#w?uq;B4F8UQ)b(7`e$R&pT|b16T@a>^s@hZ;4o-AxUoBKTQ4)
z3btKqoW4|CP$-65o!Z$52W(#o_^G5IJIF)O5M8VeBvg@Pb|1L)!kXPv+Bogh5hU(X
z_%VIjgu)1C2(2t>vusr2s^m)SGOl>_7rJBB#SY1@!&C{q;7@EN(4dUqU<AC<(Gva=
zdB}gzWe;nAjyKjntR%kwA5@*x)He?o@9-;91Z^nb!!>lrRQ&eXHf-iT9+>TTu-<^l
z09LbUY2=H9(cVfqW7D2<TUa!A+SyU!Sydv!ef__nJJdcPOFr4=5J7{6)sqjx(#3Bx
zOr^YSfv1zfrEQnCZ6u+6*3)QFl5OGqVkd+G{t0LBriiqZ6G9jAhMv{f`>${G+1;QX
z@4Xr|#_Q`V-?{`50@mAym-}leZr{G{tFsr5W+}Y?S>2pc=v^9J`7TFU^<s2icoFPu
zxxZ4xWz?0v%EX2EKmSY6z~q);)+YYwyso!v<FTYtN?H&VkqF>J9oW!mJ^Mb+7LoF}
zRdohcrfU)BUaM<b7muI2OK{dFLyiVJr~*4Nj{)ss{E2$)LO|92D}ZLueAaeyZ<8m^
z$6d&|>$2F-5@K<<Xnr-ELGxe`g(H&B8MH5w2UhNUBZp;ho?&T|k^b?!0sM^O-;oao
zN+T9-cIlP8j4Oh=4h#f!=hZja2SLu%ZEbdpi()3JkRYW9Cth_U4<*x?scs=Vf1?)U
zF?{*A<<o%+=QN|ti4S=x8Iwa89^P6ED!81t<+5r0p9ZOXa9TpBONA*@-RPo%oc)zM
z5{mhlIVd-}6hEVY5*hGm+wktA?lGq!1u{{R9Rk1=Ut3Z|Ja{)~qAwCs_tn2Q-n1==
zeiMVdXQs9nzDnxH3ht5?iSEIE0Gh2zg<m*W?CmVT<(ZzP?#$hb4>(ownd+g=yBj9H
zyl<>XU@#pZC8=PvKMD(DIw!MdpJIR_%Gt;5_|QePOdM#lvf_owihLFyK*DuJ{VQm(
zaH(5IfuFH5_$754Ef`x%oG!FQ!O*WvC#*x=g)}i=j&xVJj-~p1Ky*ykHH}klpb`Vy
zGI^T+>>5RZrcla#B);od6;urRSSO&Yq*`1V`ACE}ggw6MV}6yyK=73?+T;YQ2dP!>
zYzhC;VIKbv3ggH7AJS*E2Cb~fSgC!RC0PsannlZB>hTHo>=;XPJ{YC3VO5K_m5Xxv
zdNSsxoGEo(Rw^^Ka>=xsTwfN-HWl|5+E-u(A}<F3mwy<!$+$T`X6pk8g^ADM|D1nc
z{-Ew*1c{(A^<rFBa7UPZ0MEmRCs*Iw@*D3aC`OxeM6EXQEmoSe1<bB=?LCTZTGpsD
zL*bFN;%^#OgiQs$J67-BbzM$|e}YxktJ7!fb<!TaL8=?SHSwNEKR?`HB#xuCvatdC
z72o6o3{oG%6>OT@a17`3f*SYmVp9doX5K**uh!Da<i(R*?$~<z3OTq%Zm!JKl(Y(*
z6)y#*Z&V&SFiw3PipQ-~ZA#e6(}OGo?a#Y+&o0L5PfJg3ogT<wZsR@tLG(yG#~4z0
z?v8&g_6I4p7VGJg**&ekw)QSZD!EO*T1~CbO-CF5Mw1!;X8l9f<WQ`3ixJe$PqAuR
z{eF^uHM`>9lJC3O`kvie>zyOe{OF!%)M~MXTcy=8ZeOp&I*8hF)jArE_)Q_4_h67)
zFTzLDIg+P19_f1YKFHXK)C16SEa|_L@80hP{g~@ygs|zpc7a@rd<i=MMRweCj2J#+
z{OsyMi4+Ojfe23{Pyji^P=<n#M+Vt?Dwyc>@_P9EgturoJ_UI%5qm#wA{e^C9p!n6
zEg8AbbWRIiFGE2=NlHrh5yU{TVWN;F%R)izr-(w~Qe{K2NnNLoQQQ}qir26s-pHp&
z(u~nvC%XW>GsE_W?%-X;=un^PO{L@Cnwv#~_HGI#qA(2Ik8AlUo&v(=J@g%8RxjFr
zl^NoV|GjD0NH-IvEgcJ<|K2>_ILc8Jj<^?W$YD8nM-gu5EP+TRhDlT&SLBChC-|y3
zxyL+BWB>FB#Ri}j!x{N4Mh1vw&7G0`(jv_eaorLL!>DQck^#nL6}Qh)UNpZOa}lRJ
zZgh^18K(eD%gGILdG%0v<A2-jOK|~Pekp-0bT#WvD~SL6>4Kp+^ynDVi+Y6~N{Z#t
zxl?-KwR~kaZC1zTW^jevfpo3Q#NEc&>V3D{5TnR`fvLBh#ZTu_rkqA&zf6X0mPA22
zQ6a)$#nLVy=e)6Wk*ByO7&9ev!EQdqO3-+axfc%2+NVggFHR9DfulO0d>l>2lYp)H
z(?EsHc__S#bU;ErYh>xvQAekVVI<<Flx7~&=7(=ey#m>{T)NC)`<;>Q{aqzVo<tu>
zSes>RWANrtV=p}8tISj2Cv&J#s36z8*^hTMH+p5oKmc{^%Lz|2k%3M0G;yuPzC-d}
z!;mS)l^6{XZizl|>?m*DL>G6S0N(OC5hJ71MDT0pm4dzYyx<`2&y5+!D5sHUHKd~N
z@RZ!B6NOQpfL)UT)q*+uf-clJi<t5<<gc4qVKgU6p#;_qR56=B6Rg>P@UfrsIcLQR
z%R@=tHVB&)ZK2nXH7U0)@2@$~44dWHHjTBcLMo0v6BYAi@_ZXC>MqyVC~3YTP0~Ee
z%Y80#U(My>aLtiYu{)YUX5P|qHbh}_FqTP(-yHE%Hu0XU(Im`XG@!D|PG2!n^Y5xL
zOw6K2az%=Qyb)s^4X95#TR?;dhQOTK$N%WFKpY^AQ*xD?(Zy;+P?&*vn`rE4{S@M6
zA2!(P`951OV8HSHpYWi4xYkh1{C5M7sfk|3@M#)z?Y2vW5sB}eu^fB3l>4iqnUeWg
zZ)^iy$l3GW2F&BcWsjX<^WIA7;FgV<)lvhgWXm@b1-wdhde=Vkgn+LBmJr;rsh+bw
zinFOMsJp4i4l*)y-s_InA~HU~vp(i7!j5Z1)KH@NNY<HPB&dil>SQQ1?HFPRFo+Z}
z_kRbfzMZ)puz~0*!Srw{73LKO&;#)Kl4^LR78$nfSe04Lw^|QsO?~S5p#>Zf^!hg|
zAv<J-=hJzL-TS7(Oy|Co9_#pdf9)^kTORKxlUH)wpFNg65^PL6nLHsIdF>aU7&AwK
zO`<#qYU^$MMdJ|n0wbQaQqi1+UzS7wYZt>G|Hf(0H3xk6L=g80x7?k*Hny-)HjzUa
zlA097gab!lN8+~$jPd7r5vm5fH%&PA^*g+RAC&;o!Uz7G+;RpQcKM|S%EWYFG6~Iz
z+{L%zS}np@dAH)UEDTi$3r7NdGG!KdN&=b7>wE-@irm<kLLb|U#n!d?1<ANy;M0#r
zUk#IPZqjY{IxF%Fuhnzz*Pi<|<X-#M?eqBYwR3#K6BZ8%o5t~6zfG(<g4+u`fMg6}
z({wJs{W!~x@@n969cn`R715V7p|cj3i;~d{ZpAa9HU9ICusz2V{#g{`tt?lH@NpVC
zR;e#@;GA9*)&a^(0Y4#MbvDe%|0}@x|JG6ce|B!Aj#%WkjbU-iBmHu^2<>iLLw8~G
zp?B90|AQ)yyEYwvia_h`^=0|M{Pr{OB`Zd3HUn=Nr;yTj6IW{P@7S+87CR`Ylk`h4
zPgiigSmt-bZd>O*I%`y62co-AfDhsoGx?G?X*A!_pDpRu9jDRiuNpZfg37#|Iz_wn
z&uZg_?GI>vzMjsV+74_7btIa8EmaKOVwfYYm*}F1&TR>Ev!6d-=u2uch2y8JQjQG%
z@Ebip4oqfzz2JYCtd-?aUV&CWZRg(ayPviMyG_?NQV=7gcj3J`tad(ctMc70pVgA@
z)%vFtkq{c)YbTS?MA~~1Z_Ic;+EqK*GLo=CWtK|}(d{i%B2`OLKB)#vdiHCq1=`mw
zc4ay$#i&?PX6~e9&P1Q@^0Z?Bxv{D7ZZt3mMTP-7q8Ai`;@u!)Puu1|6fB4{Uy8)u
zk+*$?j3}k=Hek3RTrR!yl0t(+W!aeEFqV}HOr|ndR6wE5hw=e=$b-L3;0Ist@{<iB
zzthqR84np9w+P8!kPUA*AAP6H^Zww55B1uM_i5NEzUGxu;cfj`TX#D+A>NtHX8peY
zbc|@K-$*Y)_<G5`=e*AUnQ+^5u9<$?IIi_2ODW)|O)&XIG((BHJX#7(>Y_B*Bjhs*
zjY?%wT2$OFDVq?rTckjtSTFl``E5nh`g{}a!^E5)`t89ro*zJb7~N@Fb19^spc7<&
z89Uqp&wEq+;fTVJtHipUr@o4KHRBHje6#TQg=pzGe=;mgPkS3o0C%_5%roNF_Tt#)
zP%I6YApuG&+8Lq;En<KzVV#K#Ks{%xk9!&LP<Ep1gh9(U<3smSBpUTBd48U2$H2z%
zfE;vmmKH9lq@}Cg$RHm+i^YN)GnjkZ>wxJZCW!<nD2v1w-a8_z6Ekk&chY$4p20c;
z#mhxYGBoNg|JSc;X5+(lxXy2zI8|-Ju;Vq6`(QySzXlRJGC}eQUVldE2`kX&8n!#~
zh;>~`rdP{^$PxxpICK4K`wJb>u@_{o;pTE6C15rh-6F<pfZ$kwj6H2oQ&_`3z1GD~
z-w?Cc$TLkoeQYCo?gTK4(2zHo80o4#&Kdki<#Z-#yX%PcAP=XbQRz&sg){1E<G#A#
zbMhJDQ5S8unmxE#F~MiwI|a_U8nHtx3`*HQZWr^`vm0FO>={4a^`2Uunyx)}|E-?K
zJzg7|mfI5#y4s^IH7y-%2w42HvY_bVP9OC(*6y@8W%On70&xqNbDV3heMwlmaJ5(~
z`<T2f8x7aYVw1Z24=Um_n_J^^-{dx`=q~KONYB~zZygsjxOj|AGtIGVR7kVpaJV&B
zN;k=&cQdBZjaC$Ro%tlZ(L8Y2vs^eHXkA^p-PL81DmUvuZ{qK4bmmvaupkvX4;r4R
zC5YfU6t<zGlcdx4N1}rqj43m|=a`Lmf?8=!!<x>HUp}J6Zgv7mc-YJb0P4&M`}aGB
z*!G?X0Re~fWPXNjq#Ea++dreU4)ncNYB)G~Y~Vp4WiekPQB>-O+@E9wAhil+<^%_~
zFKr!v?wr2@e@7DrjkXFxN40U$wF(Q7aC|;_do6NEfC7g@S3cY$1LB&%omW9;D(-Pd
zp)M_%HLyeRV(m{yQTj4lq)u!)M<LhcDgSh)qyP2RJZwZF-y>Ab#_YnTUy=!RzC7A(
z&XYVGvVF>+t2zOq4$hd<aqmWEWu5od%Zv3(=s@Iya2%<Wam39>=-))f0p4EIUWh1U
zs@z^des-5{@w7j1Bz!kKNkQEmw`M4w-ZMSjE$n?WwV1t{<XcY2mm8n9r?B~7c@O<$
z(9)<ADX)PqSG`MhT1}(XC<Y_R??4~xm6S1OOrrR=;kY2jypO4}=99jsQ@grs&kZ~o
z@G(gO2~Ps$-LZda|Lu+Q^w9OYwvKUk3-nH1+(){IaKs9ikMm{Szp8~#MKtt;r@`@k
z!}GoH7{-545Yi+^bs_nQvn3+*_}0oWTPZeP>n;5yt;8=9(%){(r6bVN^m2*9mJ1^X
z3`m)~w1Ol#!HzHtC^Ccd8wU2?KK|#bqUd*M4*wgn4M2ds)A;n||Aqxs#kkQSfkmbJ
zGcZC=G|OAdkag6YN2SkjM>K6h)(NoBK40tu<P>O4v;D*Qr_^u?Ml>y~3`Dyx=Th^M
zF;)%WrY0ZR%P`WbR`#hv?gRQ}R#0uzRSHj4uptEbyibT57+JI5lrO4MpgsVfQM=rI
zmuvas^TmCyLGW(r^IzNkG1&-dyW63zO5hS2svx1?WXWP0@sY7qHCD-8#XwUal^b#Y
zG@4(&oa8s*(AdHQ#R#F{sEg}Ze%yF`0=@@xzV+oinq1_tuzNxT{5u|UJso)4F6A1}
z|N4Sv{aCwxxx_tkFR)aSQk<<g)HDL9Rw#qjdJ3Dy*krq|0Q)|%XL+wVEv*KtMqLuF
zvS}0qe_`!fBydI!bn8~I!FaM3IgGw{7mxOn$7Qe>jhVYJd|2c$HB+C|3y%euK7Pda
z^mbjC9e-j>#~@)eqDYj0A5Xj8-InfezliSl{!tK#;!mF0Rp>9FXX6)92(bp%^12aH
zOu)dVg|N*CJwjp3xVAi=SuRIrbL{=_ILgTNp*2NdJ}ca9bPx#eJj(M#*D2iIu<RU6
zKzV&AB+%;Myrbt1a%pMY1#(Vsjn6C&n>?#m0mKO14)U4AYYhYcpo!HvI)z`@9o{eH
z`{SW7K3!ie!Q*c0K)sMmFGAQyc&<?5gMrJx*#-q)%%phG{C&k%PLZ9*Q*bx7ed15j
z%khw!VSC>kwjyPk;_a_`sXY<$GkV=^hbK>`BS9O6{%`%cw?*I4MQ9%0z2G*+KN3-&
zZ1xyzKidOJ{Tl%vuqm;c7WutO95*IgJrg>v3I4u<v>E6yz-O0xbW<Mq@}AETf$_el
zhX$35Ki!Yfapiu9R%rv9LIe$q>x8}jXWIpm0<N1t_BY?Ajm+W8_MMD8Su)s!Ge|xO
z%AzU@0%Iu8!|EIHg4+VGUG6)y(Bme*Y5Dy9wU*Du8{`Z7d|e<gB5VN5pbU-viDPN-
zMB!kwIUojh@AQ>yhR2d`pWvX)XacUF5Pf0sW`qC5ggD6*Un0D4Y8BBLIjpgl+fmZ}
zn)p`vbyQ<?o#Vjn5(;|5FQS@tgzw8{WVR3NF(1Q@V)fawI!5CrzfUNF;BeJWnG%T7
zn45&pyOg)v-)6&QcaF8sWje0;$O($10ol1bh#A$h$VISXFd<y<g<633wp7qv9SzYO
z9}UO|>w98<<u@4Mz*Dq@EpN8F{N#_LjllG8CH^)3)v9M@pXe8Xl?R_sx&u6d@`j)w
z+gSu=(BlXcd1n_w69d}oDIO8GXMD4#eTV5(n!Imy_!88`+%7uW%?$fSn@g*hq@=UI
zT6hf2&z1$zFHH7zPU}~n=FmYwMo*nr^HtOkE@Ff^*j~n`gIY7+pytgOP78+`_VL=k
z8=rugF3~TrU<zlz&W3Ao#b8%zPz3WevNfrlUnf_qHvA0P%PyWG>{7V!qM%cO!chfi
zU7{}h=M&sxjFzx5DN9GoU|)oYC-DG^#}E(iNB7NxGY>jC!Y@7#OW3lK^^3cctt1Ed
z_rTWH*Sm4%*M9<6lScj1<08P$8Y-y`x1GD@jpN_ts-wA2wmBg^uL1JEQJSxohM8+z
zt=~HCHCz+668!SH_`JM$PEH=pnT&f`G}qDN1j@txX{Rzg`@-(#6XqTM;)%>pci84+
z`Y!k*dtXm6Uy;fhMEd$W0PR*ZTytixi?0K^T|ST1tP|$%-D&3WhG)BY{hdu@gTiYn
zFYjK_$p@`9bo^a>v?<^I_90A)@T|T07}weuQRv`5*Vh<lt;Np2+5WB`F6$G&JPv^|
zC#0W!p4~yuYiV~)ZkVR(mq%oN9Z%&v-+y+_e*dDYXP)#QlpCaV!HqRzY$ZJXtbU{5
zPg)&4$JfPnFEc6O+^4!(+0JLP)gY*+X^#Nt)@NJUlR>uKm{2x`uotD6D#HubYOGTN
z(jN=d$W_j2iSkGg2kl}mDb9QA(XTaf_!foNMQj0SSqX%n^Pj)dlXFwDSjkn>b7|ty
zkQ7NX1%SG_(jD$_$3##PJlBVu{QsnMM61$)9U7=|v$EJBA8~?y#dN2~*`;LKx<h5S
z>XD0X0|!87ZVMGW8TAG8#_G8i{T%`rixE{^>w97Xf|W=$%Wdp5>73Zrt@7(``dPMb
z8j<!JQ@r3?fGON#5n!7%asp@|f;&##Mp^T4TR?7`Aba60{de@MwXJ+zIj>?tBy5#V
zXi1ASO@O>xPeEOYR@K9S97KGG-G&n08<R~nm8$uff%o1uIa`A&5D%%OmiTxwM=Js)
z_Yd0o=~z-CbqjA&IT}+G)SlLf`GlJPj{k*>idE~*J@=5nka=tsas9mDbmGL<x|C0w
z{A@v%I>%1tRwl1Jh0JX$7{t#8Ug3XGi71GT4S3luzzIFRHg)HX^qeRr65y}#>$j#)
zQRsm8GUirT^HG`yj+zJM%)gB_1gVX3N2fCiI#rbv;IpNDq~&<DTLpMM+P=aJ-n`As
ztj#P4AgSmn=fOD678#3)h;}Xw2xY_A1w=B+7B_&Wwf4S=isIwnyHLy}fjkeO(RJ)j
z^&2FjxX)K&`BUh$HSbc4k$WhQ#Drf!-i#C+eb{OiZ=s&8g^=*;&l4250pzY$*$sa0
zF+q6!EY;v&?>;U^7AE1uFd8rkm@#Kz1%e~qk(wlY^{$*mtUeBrdf9bhcqw5&+YMhn
z47-CH^%6da&AM4SxlU0G?%;Odj;!NZvIAoR|5<;sa-U-b8cf8m+-)KLIHh)6$<ipN
zPUAS#yx)!grVfxVyxp(1UFH;$kt*@NX^e_blhfzNRwU&lW-$H~N(!+4!#O_3KK&>1
zP5Z-pStRV$w-q0E*k8t~o3R&(DIcfR(UisRZByGDpDD!FBidqvY7ljcfAQ`cNc}QL
zGDmV;Gp1vPYNa5R3EMOnU;;Yw-CB}bvujzAbfIwoRtInR>iq&sI_sAk{UaA9yDQA-
zL<zA8rX|k+9`lp|ev?xrcSR3)$_RZKs)fhWt*S%#`Vh}z_kyw83;@@hMz%6k;0)Qi
z;w-E%+<LOB2hwQAl9l!-Hn+92Y8oRZEg|iVDSnP{0oaqbRs#t{u(vP+Py)!V@<Tq^
z?B_hxqt_ZcJ+rzVSd-UY_CteqE8V}^>{)fsb%y`tDYaha+x0p!)o!-v*L-66=83z8
zzGv`jK3PQ`DLn^FoD+{WzY<%lDjYggEw$hu#5-J&%;oqu8g*Q@Dsfoq^_oO$K3S|#
z8PzZwb_%wANl!J6hJ_VFNBN1j+e@O0c02%)?@1uogM-hctK#4-ADH-BfD<MLDvL2w
z)_AefDpnQ|2I)8CD!!@b8%<Ih=Y-zeRAmz-OJSD8mgjy7>`|+?uXfgMFrS*9%v>D0
zQ?}D*H~(%Vr(H+ve%5?GRyTUlN>&S!nsam-W@N)BZ9mz)EXG^h4tXhl5;Sa{ddwBE
zulDW#^x#*XIV>)iSM7}lJO3@ZE3W3?=4tKW&7QCGBwV&XE~rm<5cS-#Wghg~#ZSL7
zdMy9?!)m42j5?|&#x9aJnyIb4on0_rn+O!5PzB@^)&l3$7k5AF6kFJE|A)D^{%h+G
zzJ#H)6bcl#;_hC8Ln-d=?iwiWQlNODK+#~u-2#E4#fn3LA|b`S5S$RSo6qj|dH#jn
zpKo$sbKjXWH*;p@%>LomYqpdO&iq`I0B546{pjyG!e`YM%-YsI{KnwG-f+nhPoT7^
z+8J|3Z}oZ9CGrMrrh)W8=0?ET1|s4Kjs2$=z+EE9Or)MmSEJjEd-*!q7*Rcy00dla
z;_k0}%2l%4YUaL3Y1kES>OQw``8mk6{vTTVj7%w57ofOm0}DwUvsvNF*FaV1fiS#Q
z!F`kC8ym1%3Omn@IiY$kzjEtJai*ZVE<7YHhx?S#yeI?hr13o|V<pLYX(hgDnX09B
z<WoIE;-Ue;xTW(?Ndxgb+_7A9CYKH$Fj)6uHvda@G&ey~@z?}T3dSIsEmp70%ZD&{
z!*8J0R=}3SY{c4SBUQT{AQtN7K9%JT;_l!DRfrjynH!1;nCz(J0x+C&saNmyefGNL
zL#GxoVBrK7{y2R$yj&V7Q@9B?1&Ia?>w**&@~Eu=BxAPj??ty_i94S2aLc|9X*6hS
zrbO+YYdF~bhlcPUPJNP%S#LZW{D-z7bT0c3ZC_+#2^C%(`QAMJLp$$kig;vLv3gXt
zY%tIvcL`Gtnk68R<$nGgbKn{FHX%AX?vHmq=y+7fQ@bA$W-UMBUOJz=ldYm(^EZ0V
zSH5P2>|94BF@}wg^24{P^eU7F>I7G4^COEm&%4Xz>&M@=)SWYy32JgW5@GHdywb2W
zwk+jM+{`tf-7;kOP8{p)^=)Rf)AnjId8qrhsKDkww5FTPO9#{fywvp@Wq*`o@AT}u
z|NOC<-7^!gZe)>u)yZOI$mKCP{hh}j)G+_tP|Y0FdlNexNX3j~0<YkGRc%!Tv5M{v
zKP<3L30|OX6lI9v+GbMRzJ^E=uJOy76;!C;;)ePk@aS$7q2H!&6*vmnCwxhxrV|Rr
z`WBBB=YSt9!%oCiE=(6w<A)uh^`Xd|Ki<>iD2pyVQ?ZPk@o&pYp1#Y@N;ptQ{GvuY
z$A0@kT;JN)+&IvT+Yj*Wbl{uB^Bn|w+ypP`V2Xy^?$6@F9voK!5dkjYf}t|V1>0q-
zxPNFwo}bg8ZU*<wOPCAC-AQv78NSNol3i6=+4r4L$EFovKv)ehN@FX+?v@a>3p~ry
zx}fQXC1Xn}+mZVj%mxbDcGz268Xj@z23{OzXSSTSk2s}~mdLEwwt6UB?v99?f1led
zAW1zH89E%0YUGY%87z_cn}iz3TNl>UHWi7h@$WACeS0}rsyiMEH2YAgOKR}!nTJ?^
z%)?c3LhiZb*FPLatfBef^fLxIoB{In$Dl4*Ph?Se#2?g<FGPF!R4cyA=bX<45#Wya
zYJtF^X>@T9!8ng|Xb1!422V6W>lQ;4O2z`rDo0b?bptALGni1q1IwX-?%p0$d?R6?
zE?k(@gzq1J93a%@cD_QbX{s0Cde#d2ls}bA{XrJsTg6BapNFa{Ssy8Ae;9nv#S?H>
zWb7wiNM2h|*I`wKJr3I6b2_6lOrLw8EF(P-qdfSuH|Iq!UZ_hs^Jd~FU$CaFtK`Jl
zB$K9E^cM>_*oq7#IUzFGcVBkyUN~>+q1Uud264TDYC(^)-TQsa5bfLL*WN&s!arWc
zury`tW#=&@gnE>RF|=B67pv}<a*l2C)+)H22ObuRPyRX9qRO&M%oDsub;*{xQnm_X
zb7(UzPW5)jq)S7=cUhe(P2ScemfaZtHz*>_?N0drE_p2f{|y9fs4LD6rO>+sr+?wy
zE>L4JI2-FFN-4<C<P(cIf0bOV#`uu`XD8unD+5wKPn#*mgni?o>1XVL=>5rXTzq6w
zFcJ210ib~0&#w^ye>%1@sf8GT2NvzoiBWje_L^>Pzy6A65AL$7>ikweKTI^CIolxB
zsIJ&yOW8udNZ!-}5N9Zq$IN-XCpn+|P3g&OfoA^I+mc!*m2x?4OWB451NqKb;X}pS
zh6ZHZD<x_@#zaw2Fyn4CKH7JU{WyUK3rkT5G+H0ISkY!;3>g_|OGe)NKaJKys0K$;
zBUsT^Z|-`lifvyyRYp{*nLfh9?)K&NgPxeSvy*O`$rwU3d)68->Mv-`mlM0}MurCu
zaD4vSvVEQBUkzJY>bysq`DhkAqB2;$U2fVgL~!swG(zNdY}oluVz0K(^jgE?T8iYb
z>`^V9<wdaqt?n%x|0Ck$?#<1W0zKl4_9*brkRgmJL}@kqb6_(G{Nm(zef_vgW3w^*
z4)?AE)l0txwK{Z7fM9rpfga4j?|vn;QinZsm}!ROZXP582JZa9y>}!UO%D+Y%ZVp_
zgDId3I+_RLB#CAbm#f_@qYg74!Z1&E>W7dP+d}GF|HsS^G#lLN&Mq^D6K{g+Pz*dR
zf+gtgJ~*(h_$)R$`&**+<)+-$_{KjpIyus+f$sCaYdr@28-KW!DCdiSu)0X(>CEB~
z5c}xl@4%CZC;AckKC;Vwh&p7q-<EKVd!lc_;L|}?Z)k{zq~PEABq#qET&*U&BIrTV
zcs}{2C$*^GJKHcOLjgk*6OF&G%GJLWs~S94(9QBg7xcFbX%=^XrSdMYbb4lQ9A7=-
z0>75{ta5j<vfNrePe_PZ>nI{+ovaVPj}VcuNqTyAvoZ5Xy=o@rzw91Ce4+*!@v}(x
zFA+1!Ok8jy!)h=KP4NGoW+D@8;$R8+E~0TLjiXU$QDQGx2hp*=ogO<rS^fi|DFa(c
z=-mIE^(X*147wHpW}Q=hFs2Eb_vQ2G+6iv10|K|e<B7A3#LsnPu@mhhguFIOGJMmS
z_|MJfM86Eb5Udf6(i^Y&Q)0mUiH(a{`D1RO+yg~cB9Mbci_jajBCl;lItRNfDxWUn
zp@u2rSKP*rvtAzf$R?_Yf-FDVdq8TOWxF6GHb2dZ?<Q^0j;$?~*DM9Wxj}$gDvQe(
z>#hZDOTS@a?e3QrE#Ve$e-G1_lbG^W*O@(qosf&M4_1a6-&Q1E>_(W<x8Y=+flW5D
zotlPelGVVJXL121*2$`ne`pc3*Lfj7j89{;Dx&7X)8!$=Zr-mtZKHDasSP;n8m&qe
zLJPRK*z@1Ndn?N}{;O%KjD=&xxYuFlVo(2CPHbP9g&oz79E~Zo*OdREM(3Q(Lk1^%
z<mXHrNabvMlXd>MuRo;1atc453Ma@it5)N(HG?}uB?tyc=jw%uxSgsjdIh`h96^r2
z*a6qD3pl_egMrMr1W;X@as-3%PFMlBPMzJ!;FYbz9*;92D&X;`3ENm;ta>e00WVeO
z%<~8^thh+p@j5;HikRW$9{R;*8)w-gbBwV__0>t$R(LK~b9lhp>hL_nmYuTv=PkOX
z!OH1xEJ<#Cdn6cnlF!wTn|}wTrE0zY$&_&x#dP`OsA(7E)K2$lVq|-I1g`n>fq5k<
zb;DqtbuZ4Xm(Gk&P5<cPuE43)lF*6}$}<c1>kLf;rUCkSm7|^A&C*ONC~Usi3aDp0
zK-*jVRWF(>LlkUK3PT6=tngkIla1IhhgebOB>r4Zs$O-%&??m1)ohVtn(4LO0EMz{
zc)H**A=hC^7xC5P(VJLe#WW&=H3JR~JZ3c>8dW24lUMc)q~d&B@q!5*m<nE!<-+_c
zIUui_qLlk@Vwme7FQi3VID-dIea90KTq^qP(ILYw_5tZBebqV~0ozyW1ZIWUd-GvI
zdc#W4#SyvokasXOhH{k@%qZduCKHd<-t1;bZRbL(AvVlLkRx^?Y>?5oMI~m%VT4<O
z+2YqbbwHq_DxXbqPU(vYtSG~;zUg9-+po-o6aG4pQ=dB)BrwFY>3T<|WM|zPWk^wI
zWxf!L)#nS%k^?FKaq-nbDoTLS{-Ncg(GF_CQU`Xd5~u*nV@Q?OU*Xin={Bto0wVQQ
zaBArK^7k9(6r7QJ8`3g#h|Xvq1yEk*TFgrYICzh%>@n5zG_2%5-0DvZiRBv$e~6bs
z4P&f*n6aa`=;JP0jgnw{LW}5O<W`!N3)Y)c68NaAYN?m-*M%*G(xr+oe+gNw!j?8a
zH~rh>GX7Wb*B(t)9GalJ5Hf|g&)|Yt1clk*{L;rd#DTT1|AdpXgfn^KHMWN`uLPkA
zE<r7hasdxQ*}*G{zo7dxnP6jw5x_Z>74IxToGX3n+73-+?msjG5N;{1HoAT6R2Klj
zlb8{py26D^<%0Wex(_!VFGVyWQ0eVQrW=>9w5R+gagXomP8rq$WcbGsUwzZwY?Nbb
z{K_e+sD5zFoJ<=sH{$qR_SHT;%G7vrMeGj;<7n%jUl8K?UGNLtCPf-??&SnF!hNf(
zp#4%M_cfHk+!673r7JyUsAIjU{J~`p`bkrO^mFV1>Yg4!e2-WXE!JsD)-ji$Kw<{4
z>8_cR`r3+`Hmcj_8EYhcHqVmd+9VW*blrA^3$Ms+SBV+sS)ERg+ATxjs38<Xq_HY!
zE^)DyiOXa2G5JNihPolA-kjpM{%e-X-*LpPP|~Dy>@uHg%hosL_Kvy}*tQqMhYG%o
z1JJ||#b2@m5AK|k2lTD!mAk(ZA*ba+dzK0_^FY}D0S>%f0fS5bK>;qT7)cYP(r2?=
zAKA2)RZYG+EWQwA0d5e5L!coR0W(E->;7Zib>JIB)Ow%b0x$cg;sMt2_BLoc7@>=z
z$|RJmo8kKLb+K2ytfR3KJPj1gGqH;-Dv9ud=X3yHT~@g++jNxd2L?tCqb<?btF+n=
z&F{6ZFo=II=5K`wyg-A0sAY38p&Z(J5-^Yw;INd|U=Yz*&=o8$u^Q&nRe4^+B1mBG
zu>SGIT*v!!eWO9Vjt5@zB1&l{IR4Oj3`naj;$>q3NIjg!T>~0yTG#$_asmRE%7THl
z*8pQ?R=(4%RJD#QD-E?Ox#5;YfO&DsW{|nH>4p{R@J%7EyUl;hl&rfD>66^g?t0ip
zqa5<qNot>u8E!M`8#z=^Mzor3Ma^Quvu3(@F|dgGZygfFb><Y?zr7XXi<Q)xH`OuM
z6D;F39FHtjRnf}eb1c{SrIyN=F|Il9SJzrw@#@BoMi$$C+?_fz{+NwAE0*qMSNLnx
zyT{Y1m+3*iSDFv?BYbW{`h`q)mmovHA__;YEtXWmtbX@`3jwl$rp$1^5x_6^c52;|
z+gOLgpi`}wpSB?472FjyuqQB^XUY-UQ_{2=uuw?Jh-^2tmj!4**xkQO6p&PVtuA(0
z-)6NJyhcr<zBBnomnIOWDz5yORH9^<LVoB`xMbml2$#U?mTFQC6($bS$%!3%@0RmK
z>m0q{kFN}toxCgcwup8)VEsi=8y%Mo5<J{UV$_@L(^Sq%L0@~Q7s~DLBU*0i4t3l~
z=e;c<C^EyzK`wAjsG{E!#N>tu7Tc5P^;>tO6I=><JMnh3{bT%8-pT~Z?nRtJ!*L*K
z?0!3RGu{%Bq#w_w1{tzfjwR-#%=T+~H~+1Gay66wt+hVhvY#ll+=9uE(=<^m>Oo`P
zm6RXkr44sXY)~D`i`u<icK)H>t&6Guw=}4|pD(M<GUw8XxKM9%OgmyGG@qY$9~lPo
z9DF@Ch5Sf9avB5EI)9@9C!O}kq>8$1Gk!fS=ah)rbJo!gl3qMspLv8(84R(6jyCcP
z255PvT`~d+fS+LP0Wl(?%a7lh{xqBXLvwjN1E#Y~J+?<6WT3rTlU;CF5&J~jsYjB;
z!vS50Z)lId;Gw7TCq?twXl`0tUC=DEs<-{{-RS!oHPu{AUY1fSRLj*S$uNKA^=t0h
zxnZKX3(p>I#{%+_+;>6yTRnJV308C)fR*q!05d25yODF)rs3R!><ZPQMixy}cZ0<E
zf(`iic$r%yzgFG<)*I01WYdZbIb3omArzbl{P-f>osr_XqxmWv2x~XIcbFBiM??+#
z81E<u-E=3J$FnH`W2VwbnmO4|?maeR^@%qmz|7>i*tMQ==7zexX~GMxC93K?Tc&xs
z;&1bEIKx{Rz9y&yT^juEeqmpCo6lICutT)vYVJ7Kw3o<mTaw9tratU<W2-iVRuqdA
zfb_+N<s&1`U;Ranoi)1nuX{GS7%qp?0cmD}Ox^wQKp?@L0>|c&K$9J~#R%^&W-A?)
zc12c;_QI<i#>hcI;tR$u4S)C)1o^$5N~o2LtKk4V!gVv1IL~Shg)G~&$x`<-iXM{<
zq9;!Euv0jVI+wiIe0n<s<UYhV+VYR7=RZ#{_0-^>%g%k@XYOM1*VfQWR-0w~FD|rY
zi%{}v3*AhL{rv95t^Q8Vcs()3D}S8xZ_*|d^3a=sp4NxA=2lqYOn>M|!byR=iw|bK
z%y4R%*NEw(?s|Y*r6Zdez*Zy4$*!9_9nx@+;Y{v0FDriCirvn0c*#AR;Dw`4cdhL3
z4-NGIwazDIz`<5Bsy-+8S>w;w{#q!`YThH(qE`$u%W7}rSJr=dEFkCfv0hiiuIkqQ
zVNeyO)+zTpr@OT6S@M=Hh@mN`FjGOy$a%KY($q!0HIK)DZiag+Ge?77KUQ#>wtO#M
zmF1a9cp;V#%X!KB4{bN=g?XVeFs(tU`7;R*gp+44>KSGqJbKuG&&}}KAvn|XjJwcV
zKv&i=_z!_uoi3ySo+i24J_#CMm7Ij$|Cr2h&E{pVXGCXIMRxh`&9@mD^zN_Vkj6Qb
z$=Q;(&5O_Qav8LU`RAUmOy*Xc7E#McPmttl3%4j9dv7uA_CA=pW~c?rO$!uh{S}rP
zDhH@<9MqUdZr#Oev6PqA==%tM!CRG1Vs)Uzba%lXWIrFbiXz|?P!>P_s_i%SE^MEb
ziVIQ{8-o9XrV46NjWlX)?;ELgKbqsoMJ70O*9$gK)#%?si2Y?<n&jI2)g<TbGdWjr
z=gFo`ClzPW1vL2M%VA-#b|FY_zh@H=FLtU$@Y@%jrMDB39rgb4#j(z<3cM~|dm~u6
zB)TPZ*HCqaw{aL1I&%$QZYMuf{&m4*amjUjRnAiON2d6EF8!R;QHcN7-V3mfbxv<=
zF2Ha3$hd`VI=W3-buRsM;3dNVAy;x2VCig@?d&l4_|rbh;d!j$T`nt!{T|e^gG_Cp
z_sms@#0j^w=df;|eXXRU<&3OL7q*0uFr3&X=UiUK^^-IlmV=m^2~u9cA)%tgB}llT
z1Wy2<%{V09rTo*6>m(6*E7q;L>o{sL13x1CeP?t%(ileMgNsjCT|zpT!p&^)o`=!A
zVyvN_8Iu=eNW9vvT2{iBxF{neUA#4UQtBq~mxNt*6Y-#&n{Lpaidvc!PXC~U{geHP
z?dhT|!U?ws-W`w=Y0<Yb=(I*fV%<@6&x;RldU%BYL%X{zLB%(Cf49azG+PPcPPC42
z^*&gPRp+OI$<x()R0=f<k3a@J{r%kaI6hbDg0H+&@%HNi^*sqmVZ=+Hmdj~2c2L#3
z_-|Y}iRpY_(i3QqD&FAisAml%w_Pwl*T84b2P3aU&n%vGaIFJn@W@gJ+h$rslm>dr
zn!_2~FY9J4QS4i#rC}8!9L9=+Sa5CaXFW}FV>8)E1P;?=oyL65OP}8vq2M3VD3NeP
zIW1mm>ly&`G$>-n6zg={PA$li<lgZ)-JC#y<UBx*`aPAt5F2^B{MitO{yTyBaoIUV
zoury*ZSv*vGEQy|svkO{QKc7G*EiGivwLeCL@)OWGU+v`m%lG#_Ub#jj@`F|e8%7N
z)3NP+Za^VTp>KnB+#NC&%C({f+8&Pen<rZcm1|ow%v~Zq=X}@xgov83PIcpHVWF}C
zS>Gz+6*T9_IDIb3=nl6%=>qI&F0F}t+KpJQH)8@&z3V!0@n-3F0}s=UJ?2d2^B(Fb
zT--Ru7s3Q4ZHCm(s(R=TnKZTv8+hGbb5s*8;U;4Hm42<r%$^^4B3GKEsEx0jJ$Eus
z(<=QOiLIMnX@CB3QN<O>q&0|jHf<Lq<U&4VfpP-WiDDpLI`-0{jMm6Q!u&L$qmWBa
zKAiCd=kI_-gY=Ca?k<HFQG-Hmd}3riLF@A;6F;XN{*ZfRGfn`oH1a5>3W47}raHht
zgL98wNYtPQ)J+~R>uW(|KOVJYn+R-~5gc;Th&{<5u~kT?ykH8khac?85fW1<;ci}h
zxk@W*#xu=a*eCg&`0bZ4^LWplgysat_6%i-!z9^R^4ypGA8pc+g6M3D5)>&SMSP2P
z`-C#Q@j%)Uk6tc#<z~qvDreu?L>i#r29Voon_=CE+u~ieY|pG&@&mp-;+;}&UY}iE
zROvDaRR!aWxZrlpz03AlE#jTR&cp=-Y8{t@mmL2rBNFf1;4e^)-3<wzCvCjpfjMRH
zT9gdpPWTGB+xqqgm?RIf4j=;8=4FCy0+OHnLV4zR({`jx>zt6e!=$V=t@R25;@6DI
zMD(A=uI#%NxQ}=8#WuOs=M%bqXewlI(@E*q4l%t8ab*tdQY2h|gP1>?Z3SVUA&!uI
z|JhFN;gM_T98M2>afvE=ptz8N7c-9F6{fBo&6NoRjwmQp)&(-avU=mD0>Z_^dL}RA
z&@k26Y`HY_Sh<3O2lrjEP5y&W<IeT2SpEjc*?Y&Pgtu;=2mn*CkQc~YzE_FZ7O$g1
z#S2y6`Z`RCo}pEqdJYBLezo6U3jj%R9Q>a@qD?7KtWZV|JKs5=s<D#1`C8+H*kDqL
zALqU{-pt@ZB3r-ndlWFK47}V)vp4uqz(fXwTj<vRg70odSQ^%be*OjmjQCD8jaX;P
zxB(KWM8Pm-OK>UC@y&-s9Q$IM9Qs&no-Zh0KW$q(hll?%_N-+o;b1+?um|^ZN>KAp
zwJN;T;I~-hxuotmK8s5UQQ$^NZKsmGI;2yEi*VL`5lBn;Yh|wfTU)PT4!gBe@p<xF
zbrUfMyF~KIr&sUwM3~UO4&51Yqvd7mVWebCVx){=oHOJES;g~oJn+FCx~v@^(2*Lg
zsB)pZk^aPQ_)gN@Wt$}-9G7_CJk3IT8L!k0lCeWY4HCY2zMd@zUkBVW=fb`WJMA{#
zSgx(kH3s?*I@gi^hRFjWRtcfp7duJR+OlqB(QoAfj01Y@b)f1JzotDi2_713KF6r4
z%*kpU7Pznsr?6Fjva5cjq8+EWz)|+KH71JIY3sJT_~+XgGp)lb>^S*5>*5h4dz#v<
z`d1@=0Bx|YofS1?`|Nuw%<Amv49Rqs;ex=iZ^-@0)BC-#Yi_}#4NhIO=1&RB!k$N<
zy3<l*AEBJ94o@dA|9@y9&QU!3&)gy77G_~KU5yD62CarTYv-+lUB0GukRb2!<&wuj
zUp@c@?-51nJ+fTgxm>I4Ilm9~Sgg3loP1)j!7xXHy5{_wYD<!f8YT8}IR=&SG2-G1
zC&JXMN;@i5<@qc5{E6};kpjsSVi#di`AlCM>75;V<XbCp#D7-P5lhopSz)0l5$T_y
z!kugu0-<tHx0TB_(^H=E3?^rWwzVb;L@vxKOalyZ_<^^|wY%2txN~{=T-O%0Rxtsb
zZMbgS!1Z^k)l4a1Enu~R>fQH79N|K2Z*Bo4JARYN1NtuarjZM{wN_H_jLd!0kRtws
z2|FI^A%m)oX2&<KpKR3ya>I=IM(d1zA2NA3So3AkYtZt`vCq0=$|Y5^r$Q|))j(C1
zl83L#FXr@`ihmLt^L7|;BXj@Di;uRI275}8ZfDu90JDvSUjVnKFA&;`9v(x`wu?rR
zOFd$>>~Y;4)OJ&MBy*~f2eRp`Tjzc3k~XNA5p<Vn876E~&!&O4X9071>x}khjrs5@
zTEkcG^AKrH*)Q@2FWm(ljKh69JdWN9Jk5B-`jto>F+UzK5ftUfbiGnhxl_(SO4fWu
z)Y^vUaSH^7ysb-6t5vZz)wJME&n)6#ip-3%wRl!QNInvnKc|1<?QA!H*_`k?9H*JC
z$ZyG7&7vJe^}zSfVRnF0+i)VF!r+<UNrtx^#0nqYUQiQu6E+Pt^Q3J5?JwaQV&PfD
zY<8^Ia7ZKqh#kDx2CFAh*n3j!P1GF2pb~6B%X}6kO)H27U&B&=%Zs{;%@z}zTuym_
z4nVtWr=KzdKsh^v;FxB8x$YP{nukh?&7%D30j9Q5d4=1O<IZc0f}N@?Wj4(5yA}Aa
zDiRqqG6ySPls8+ZGe^2Gkl4;=5C*^fBT^V{vu~06(LWm1JBS168={?oJM^EWy7I0^
zlBobFJW)iD(5Zy^#M5~J=Lux2U5Rm>Y{?WzG+c6?YuEaNVMPsa{GIUq^CtD`BvNR5
zIzk*800YGKOSrQtyj?!6;zrr65;K!Yl1BY4KrQk_CD#7Mu@Ei$$H2aL<IYUqwD&*K
z>iPIl*>LU%b5LawSL4}S3)<(eS#JuMeUI7888qp$B+vAe^{RCUE4)TFu};*1$<`P-
zZicWYa=QnmWyrt;%3OzV8kR8vm^7tA&`Qd$nS#eJtrmNte8g?DEAJ%CXbNvDeyjaM
zt5HkTJcxW6Cc*)GDh{i6*`c^9GZzvq-bW(($3(1%AX%~Gj{(?DPg8_Yw}7Hk+qIGV
zU;O1#%wA2*25J-MG8zu&W6xZcrnv48P0$smCSGV3NCcBGrd1Lq(FD4bsW8{CR(yE5
ziG@E2jIklkNuNM@H8djr-ecLcmk=($K3F!!N<}xw?kQos!a>~ELKf1kM6(6x#gUBx
z`jRE>b!iDbRSQybThleaPdSEdH|sVm0}}gRugDQk^{L0>C9+2Zj)=viuFk5xyzz>f
zQ0tBe!cL+)Rj+r^QXwqFPSal?P5A4o@R<l&5}%}cB$}C)oF<hO8R+&F_)5EhMMRJ{
z=IW5y`#YAUW`4H2;@9}g#))nwAsR5u3u)dGPLGIlMMR_6>@^uQ9$nzCKXk8|EwieP
z9#_SItF{?7DYKqbR*tReWFZ3<E%W>etyu+!KE*ME{P>yAp#H$)sg&kbgj8vh3Jlvz
z70~sUgMf<qz@|WuqXb*PqIrVqLcN2*jRcZN<a|Hwy2^gYsBL|cskfq$aOavbyh$>r
za#w1qK9{b?Vc^^PQDx_aIpf^&)dsTHVmsQTL}kn-T5I$tH3!!8V7Tm3;YY#c&nPy^
zqK^hL;|f?pS-rlF!AB|bM}n$}jB9Rz@!>n0DX`slB1^)d6R)vre>%3`E+cTXps$te
zG*@t*dx!^<kh(T_6%-<4e=>oG;)1DI7se9DJtel>6TF~2+Wnj#+OWNCyBzz`K>!Yc
zcQ(!A&JM%I)bSO4b1CV<mLvzRWKmn;d3xQt71Hz%kcGk4vaegfkTYGhjq0E?vCst(
z{n7CUTgyoj_Ofp1>%0p=T^;KIBrov{sk$^Juy50U4>@~I?EJ)nB3;9(k-{OLD^N`z
zQ*m9MW(Ik;&_);_Zx9IWTO&hpX?|L*;YZMFLzLS4B_C{NZ^fw}z_+p$V_Mz_LvcE0
z0^Z9FQYbF!Uc>1z?q=#pTE62La+-<ih+n;M7!EL{k47Z<>edr@K)LT*HuzJsWV#Z)
zq)?1h*6(Mr<@_@!+E}%A&elNWb-TX8D?`oD%rIfr#1G$<C<-<GL3O(<7H>^3_R}}H
zGc|(5l%j}JR0ipyqaGMw5d$YPq3Xz_&<6{oQX7@i+2Tz>h{D;WLj!1T!M+>fs($~(
zWwG#L+a;szhI>M``41*lr%P_|nxn)!AzVF_GD}_7Eb3Qq?>I5zokr18$9>#q(BVtZ
z*3oQ1g-&*ETZ`hcD2d($TslGM<CO%&O+r>NGF;Z(3ROnnj8?D8<t6fx(O{oa)lf|q
zpRCqru}@A~5|SYw+kM8Nj`7O7FPi&+PK_!D^|za-5@S?zz{k#J$${k(m7~kxbP2Vg
z=5zV*KmS9+29$dGD<TSdn#hGGP;4fkry%2gdR4}I9iX`6{@@>d-vW#szvOApzR3#Q
zS_@2uX1IFv^t4+n$=7GDaJ`rfgl?jG?_I(D0E<q(G)_X2D!Z=wrvAqVA;-6ZWxD+Q
z&j>@^=RDG1r5h;HTaOT5jep8xqpzl^({mllul_3BZ34AVN?1m1xzXa3YuPa3x2#K|
zn{t>B@yMUR*c2t$?64{tJV7zv5KUP7ytGUx%SuH2BG;nBod0Wu!Jw5o!s9Elprsw7
zI>15Qy#?4Y2%^$v{zdgRgO6w4x==QwSN#$AX{o(ExwhCneAd{*1mfjP-fj4uXdlO3
zXV8Jnhi%USasF~V(V}X&lUcL?;}lIqI`*xLlkz7eW1Ui5DwN=&_En;{{p#J2YAuRR
zo8sBs!y`@`w;@#!c&D$NKBJPQQdQ1~+nnfY8SjL?U!pF#owoDf8ENK%bcU)WdAj$}
zqDT?GA&HCYOq)3CQ=G($)@5e*E2zK3q6I|4_dAsl_{+-kd-MAm-)9bm=c`d;02*k>
z9yV;{sNI7H(N!Xqh9X3!?9)i*L&n-hqkIiVZ9z_gF{Lz7f2{e{qGqvIQ)F4)S+;CQ
z1d?8v%+Yt(>8r?na#z!Hl~vZ$6XaX`YMREwCGFgm6;t)45VdVXKswT5a=eE>=aZ!F
z#(0&U*7Kpk)abJk@#8DDev7sTq2slI(D$8BLZO^17T;l@6^#nIxO-WjHGdFIoa4Td
z&x$+3`4pSTzhN`mIo9BmCJ#oC8?a_QmPvw1@r_m9$mL|J6SB4>`Iv*UeBRGwD{`Go
zLL*-T1dK7YP7z=gJvp-1_B4s{jy*o4%cFOW5N%*a9eLsikuQ%!oVuD`utOa4&^XWC
zy4p<o95`;`(01Y2#~_(82>P4or@z7W5Z9UBTq$R>I<DBL9U1ZNth@$oC_Bt*P-^%L
zv?V-vZnSAR2agzbTQfk7oDKFgbvANoWF0|sAw7B7sLss)O{VQ<I+Od)$q8N+vYOu%
zO@fD5hVe@1?y7)z(00Jox9`9J%wNVH@FK8+oyY=AdJ1<!p_9`8pu%)0H_c)(r&y>?
z!ztQ--pM8Mi?z%e=MtyE9AoA1Fs0S3r0Tpr8S#Ajs1p|{jfk4s6o(Npi<{6TX}2yI
zyNoF3r^WnLC={K!SdA~^<16P*Nn=eig|JJ?_Q4+1>95jZAnqN*8BocxML7~0Jhj3O
z!T01Q&#X|G0Jqc!yns4Dmw6UBLX7-X3O!(V8vB7OxD!$5Ld>SSYKdI>{5X=ep{pPt
z=at%$RUZ^5HQ2Hk=>Gnu(+NdkEVw6HG|URF*iR;&v?`z#*(J*7hyc@{3rxg+<SsyF
z@@8Bm(AWv)>ze`-TwOE!59M2c;|cP~L|^P`NWi+bA6|*JuoC`?Q`9aH!j6iJB;XX|
z7u(ZO81)X(x}AjnNBI~_qr)x7;ZCT;4b`h>R%tow-Z&(|v+J7aWq#q5^zx*N#}Et9
zGR)}JoNIU~65#{{{X=V2BUF&^4gDUt+T|tTc5**(vO<LivM-l&dffC;Pvhlss9=9S
z@0@79zB+FgqmF5KWaa?a#KT8+{B5+3Ah@2Pz`T|Uu~hwL(_q10lqILv5?B+b?~;k`
z?d@B3@tox7<ed__Zv63__?KUQhL#^L9GzQJG?u;R%0H!$2CssF;Y3PjZ+4d_$cA9{
zp%&n^?6E$;02gF4*Kf_pU^9F*yF50Oyr3ei(Yfv>b17hz$U{vytDh6z^(p<Vr!;vw
z7uW`JDq{V+BI@6{5Uk$Aq6E%hHi>ZPl0PyR+A(Eb-iPiDxSz@L4pS$|AE_I5>}e#1
zp60Snqs3Bs57nRFQpY#7)#w)U<b>O2<F#8MdrhvL3QPTwTx8r@fY(8N3)$`2n*&9v
z0s8^LVmgBnn|YNuu-i$qNxlc4fz3*SzG<+bcwscGmy<|QYa4O;#l+%ZTiNWkqB`bf
z0>_6BtNW5si?84Pz#8g{X1|{3AEV|!rT$^`kn%QlT}*A{_O;d0qt}p9Pv+<?=L7zG
zDYKactsSq@r`sCO<*3t3c*yDSdrj^A9E6O6Oxn-S_j{Z&wn^V#UVEUSFJvX_g(%~A
zd$MXvQ9J)b>vb(cF!!NeFb!fSHyEma#lwCPKh#^`n@dzA$}lC+-AwAG6ir8+x{t8-
zfnHW5ht2*&D{q;4_;#f=x~=qhNpjy;@S>~nX%u2aiwIcDRC>xsRcI&j!j;}goaR(K
z#_{Z;u5HXFo>Gg1nLW{!pf1Rq=2Sk$^K3^PrPwm`7u=WrL(}?d-j}U_F*}q>-K$4}
z*weNcfK>$AkMEcbo)97<hX0{);w!z8``=&WeVoG>`Sk{at1+8bYF>5ZsfY=68FgRj
z^(g_Jdm@X8$Mo@|HR(13WROB>1fOl9T<K@(H^1B7p&BRC*WJCgxEuUVbCpmnio_9v
z8{*F&Qsh0?Idtv`OvOL-)w;OX53;h}@b+$x#8(;6rs#i&P;ZlQ5etyH4?HK!+?+qH
z3}U*FOuDW*&2*Kb9(`QkQKx>w;+<faL3X^pKgELw1?-V}ZE8D*bxYPX_sou`-tSi5
z{6iC)sP7BNx2HS|@sC#Awv-#J6sE&O+y554<7QAvn{pKxq5g^Yd$?V;Csl;e(G@Ig
z??hO5cQflv@<K$P(rI4fP9W3Zojr#<#?y8-zRXeDrcKX{w;g%F1<juKk8(4%it1Zh
z<&DMC>X!D){Eh(Ae`q5Z1&xKr%YS@EF|S6B43(~@i_`<zX;YcQtp*v?x^5@U#WBCR
z6!)GZ_6wXIUuLh)XQdX|b?O+4OU#MCP}M>8oqFMn2sC-*k9ttc3pp!sUj*m%goa>S
z(u9xq&n{=saN5RWXg>__(FQz8Z;tpygGAgABs*6Zmo+kB+on3d<CIj@BPv`|W$vRH
zDvFl=y!6il2PDj+&*O@+_|m?3`snyNC8nZ}qei-ki4V8_@wV{ksVnEGF#2?)QE#F5
z37*B1yB{&Sr*cv~t~v1jY(`*Jc+z;}wmUAEsjtY-Q;Wu(hPqnnT0m6>NzrG1S~q%5
z`)K;{Q)t|43cXW#)c&-hwDGk0T^^19V6zahb!^vYcqiypDUGb$@h%I$b&Xg?HTzsX
zbO<*%=BIY*5&bz_vot4(Q(}~is4y*-Ia3LYI7A!9rWDw&eCKz)RgXgdW4InET{FvS
z#FF?ED?4g6D%Eh`rq<|=vdcwIC7hjy?vWwkY4jjs_kS?*53QL6y8eV*`XShP8+N7i
z|9k9zNZ5Rc=iQ<^PI)JG2elkoHF_!(+kP?%=eYep@B|^~yHN0~JtwqHeWdWYc<3Gb
zKUZu)sTcHJX6tv`sg3pJel+-<4Zf(w)lUW}$fvYls_C7KzQTFGIR9wnzt?kPKzA3h
zHA8oH{hvu;7nvyYi}r7ujZh#Uf2iq8`-uXc?r)T#FM3RGN6^}hSOZq|pP!utg!;<V
zo}*%g3eK(`3F^}yCzQfa(IOBz`bJ=BqEB6P!ASgn@cBwcW;3ViF`gUeMrl5*b=N=H
zJ86`uKq}O`@rCTcrlM!HYSD7ZN&S14K8ka)IzOqGuW33Pi?5_fSnGABTvVF~vROsB
z-g=kNq0~uTn5~arj@g@ItvLw2lF6X>D2n(obP+-of*)NhVTK=;WRrSs7mv$7!f-v9
z_v=Em;Q<YZb+fy;NT8s}{5#C~`Y<oa2+Lji?RDUKm$^h`gxpYDL!b3FY~mMnAF-mp
zD7nw|qt8xaYF{Jm(>iqvN{Oq$Z5`j&6ss}*^zZi4$m*eMk^+I8z0bAJ$vN;?VRG*F
z$H$_(WXbQxm`7oh-tKWeH`ASIjagM-NE$`bcyQ8evK@n24H^BJPvT#2)G*v$Fec1+
zR41WKL#?b0TMvsBk>zvMAHqUVqG2U|9&=?UXH4ulp)oZ%!XIP7?jY(UZu1xirc0vY
zDYf=@+iV=ckM5#6Uv~n~PSC72-F&_eUWai(LWF&lE=;$;K#%DaddSRodlU(V!;)g3
z<2aVVYLO?Qe#8x8P7;}_NBy00lnsG3PGzo`&Dhts*Y%5Eqj=TjQ=FDn7pi|~*Wa`K
zAQ+rf`VoRcg3j@SBsvsB9gYfuqPNknj4|YGhRyW2h!#s4y}07<Ce65Ib&pUhV@sSb
zf|!`^Q^)CbUFNDOZJAZkM~YQ`ret%)(OP=jD!yYa9ZKh7OXImpO}2_h)qu7>O|^gs
zKOFy@4&_8zG(o<%XN^BdEqeho0-&eWXT!Mg(q2yqhZ5j+1f2Q3J(6V&(NR8?#yjhL
ztGn513SUgJ`s=<pw0uHkuo=7pN30h2ow|G&?dAjh9#8c3H^v#XZGrCg<AP}7-#&+y
zN8>7DiIniGWj^oidUsHvTTX=Y>V=ZmaQrK&=**$5LWUITd6yb<59=Hj)2T_Vnpyq1
zma2J7wIN34HP_I|@D~}_2N%DRSqoxp1u8X4b05W>wqO74<*Uuu0>GZ=o~YpMBREkC
z^LVH!@55lYU*c4V*$Jg!-Hq^Ln;D+p+M-3l3f0pv>y$l8Ds1Ml(0uuL&NgM@yHoQ-
z%M%vT%^7S0G-6qRknb+EWR2EiIt((-1Zbqpur874thi@~!KG&eVRp?^iDE^6`g?5%
zhn|n5xa3YOWS3eNaE_fV2*q{Av<z6N4`36PD1Ov;1u`pF5s~QJN>;!5^rlg@!Qpcv
zCpvm}ewkR+3EfX({7-ASg7G$#+r-AKw3$uos1eTZ@O4s_P&H_6>Nf9qs9U#t_R>&T
z_DU||zVhbtqd&~alHB3m+lopCFam^jjSoxvdY2Ce$li+_zc<LDdi!k3F4ytV47T|g
zHP7X=)d64Z$?XQ1s!#BYd3J4NINt(PF1}?*kGX=)7*OdwqdP%-332M=ZPz^R;3aGP
z4J}<04W9=2f^ECu-Ezm!6FnD;^w%=i#H=Q*GVM=8ttG15E=vSh_HRf!<v$9ly#)%9
zERSu*JBUb0y*9N$o+$ZybKmSfs65E4uk?VHS0k74K0^PY#q{Ap=v>%<<?}7sEW2Sq
zyqa~05USZS$psMY42fNE?)G#G1(8&78)`GOjDewBAYMu@H9>mtO$UOPsuvctPCk6<
zzQT2aXzKS{;lt5S;z_|4vByM3ciZOWI^B&lrwo$iWN2nCiC4TWGtDLK{*RvoTKrmS
z6;-vFa8Fa0YU)2&Bwq@=K`JDax>t3gvBU!DXPWbZx&jN@6;73FAYzPo{%y4&5eIu~
z)iI28)MOaj`OEW$Q{;8q9OUdltOG#o+K4U03M^=MI#t_)Ijq38*xl!fd$yY_=ix*r
zfx9Nh1Il1yM+IJw^8)7-VDT_OhRqIQ=|8$UYw_$3V!0HW6C+|@>rgns>9FdEDf&4_
z6WssYFpb3P)N=}LV1VK||68V%=CDYnW3?qtfp~jN)*4nSP9U*z{cQF8?^gzl<*9E?
zd%}%h|CC5v7SjDvn85aoIc1{7%k!1Z&1yKckwu_X(MHjh^a)ZAWxvPM8z9@j=UilE
z3n(PaoP2jb07!ElW}&;Y0`aE&`GW<NCni)%Fq4b-ZSvra*oOgm?R4^wZB)b(&U~PX
z$7Ud2??$&r5>Lw*BML|ktrX=j<y+Xm0`7;GkrY%Pex*}H8rGy(FA+K`Q2xF)IDAo`
z6QxYcn5n|?_1Rp}+}1NE$FH9Tta!g+?=-6?l+F(|^pU+SYZ}2eJ{=B%#>u0@ykMy*
z5xD+|M@=ROnX{!PD?HEI-ly-Uxn?RK;H*oQfNj>J)+BXU6ZAC1p=6XBafwTy3A&g(
zv_*+uSlsxzm*3kJ0VF1Xry&;;;$tpg1>3e|4=7Q4J<$zAQJ-VR^Ql{2b^cS2^aw%N
zL1FThDQm$7S)rQ#wft4#Gjc_ouhn0@-{<HFKXV>XkYjv^iozA$#;SNOqAyS)LDF_G
z`{K~qn!#zU{m|+kTH*xaX~lEMV_pNwRWT!TP6<W;i+Z_R;E2&(_dg)vlH*n!1sJvN
zkpEa3RST9n*|+EP8p%;-pSGdWw9Cg?G>J`vfh%^=w)Gi2SuU~a-$*;zQIj!IfiKDj
zo?lp+6^#p9B#RG>5iSa*iH^qC0wPP~e;FuXW+{AUU_IwP*DbJNZAyut9y|Cbg*9Mk
zs{s|&{j!PMs?66zc+-D|=9yzG&1XqQ_p$vS+K&T|5NcFNLnir#%7em(oeOyoAr7<N
zro1Wv?p^G0h`{6da#LL7z5)dZcT}E#z=pYTtpJ?v2)9UeovLsyvi}Yg;ReV)r(+c}
z1nHAiRSpZ>42{}GzGJ}fjoPb?%Uq>XVetl)nP>>;km!1K${#w2mDk)Fv%E-S&`WsZ
zr(*CnPaO)lKmT}{0A{>a_X*0u{D_H5Ii{?8Rv3({w1yw>J%T8EYQis^qM<rOz+WpH
zX}|JVwQ`j1?hn%!fGd=`0;cz290-Xy3WpqbL3aqK#3E6Df^9OzK4Eq;8T-os{_YBd
zcEYt4xLd;IS3=w#;1_k$_MSzJvHwNXQS0g#b#S_+yAOTqFn~cIGKCQ98yC>fGR4`C
z@I^FZO;Egew7hHYOTAxuZ&O-W=9RB0zj}|YIO?d8deAa22<i>_XmMl=wG`_7NZ7mZ
z9LkH~XJb(Kzz|+B@WfUE{ak|FcVF?0?GkB+QJp_srd-1`VtY=_PL5k9+D2EpJeK~<
zTHH*{NY4bSkZbIF6m*fDP8H!2*%31<p|78x0;9?9$;Q>1tg?;@FPCqh&G@ZB+@!KX
z2YDrW4PY+Bb&EMn8RT#xmiL=C)~xii&#2~dUJrZYk~1k_o#%hTTW*e`c#kz}@!=V>
zcFaLH!2r_+UkrcHhW9-^Bf9^SagKGOy_Z_#L^g>R1xrq{BNjF*1+F~?Wz1WNAq--P
zmky_!KhU!2Uvf{Ime)+XPMbKmbN)Xdyd-bS+hB+JeQ3}E$7Hd?h0TbsR95h)+G*w!
z|0ZZ$IPICyJ}o+DjD7d)e0$}<&}HF=t14BjcUarP7=$!S^tf~&eh-TCojQgsjF-U|
zF~gteFwG)x4QhCq*~Ky26;z@qcYkmcl7K4*Dt+sSCs<w?^!wgy=${thUdDLj#A8bq
ztM-Tr8s%Po0Hw-nqBG=g;Or?{J6~;oc~9}>Wa{83-xe)0_^$05+FCN`=|iY(Nzi+I
zvA?xhyq^7xJ4Ej4@UJa9?Q9=D+K}-YA#Pj_!U-riQ*aCJy|p_?e6dL|-}%@piVsMZ
zlI~^gEU6!fU;q3NlG`b<^QSYU#5@Rz6P&XUaD{jbLL-zj7iD}q&s!iPZ-_pT^)wOh
zH5n8e`E$#~%91uW=xbzr;Hb}P*7rDVf0Lx%-fz`$DUw$Kha-pA{=d$4`x^T+f9(6C
zZ~0Ze(n&MNe28w3H~ZjlOp7*sww)p^zTdBUg9yK?MP)nzcV0C;A(g9&p(5vt+c|e5
zhjW#2c741mg{N9Zef9B`BejRCS3j5Z)-`{kwTOk!&=fZBR<&uKpIRQY2#_z^a-b7X
zMH9rCSOnkV^R5pc+Gd_Y8l%08&<TDS$6$Omaq#uru9Z<Ev-0~mkm)We#lO%^+Cc5Q
zltGpsP{b4Bkhav0A3}`FfAu?s&VSb4ww7wULn1WrCe5V~JcH(kc5^u~LRPR9f;dyy
z6Sn<Xu2gBD;Ss~BTI!^x$9lG7U&*#fbGs8LfU~oQ*i$rK!$l8eRF~e}d>8ZVK+K6R
zWQ6{;R-tHH5IOU<F2NbI?`iiEb-1#W;aZ~XYY3XxJB%#<$sJ&$<c)Bsvx_-?d(8x)
z(MaTu<Wr6m@Ros8bA9|huDn>4@H_*C4%#ft&Fz%znGdbv_dv(f-wRZV{qe3&BZ(cp
zF$+8gG4!cyUule<)VywdTq{lq*8hIOm?PB>-y$e#(2gkYK#o;mewDh`SpSRvR5i0c
zCBw{Fb#%^vJ#ze&>J+t-seL*4(#{z_{7uE-V4xAtZ1!j44?4FD!b&GJFV~9yp{ao}
zzeSjTrdm}@=t~wskF{+q_ueZ?C)StV{1x^W{|fAK_U&hm-QdMm#j(}HFJGBUj)yOK
z-)?YYPPKj7FPFYtFPj>S+0q?pNT)O$7O%e?R9BR&DoQ6r-QY_B>IUsbA6?HQjI<!4
z{`!CB1m^#td85=S!j(02?&D-{szhN|uoR0Rfr*XKBC`4ZwgXUkUvflN0p0M`-^z>k
zuCa3Gl-Qo|uawu}uGi=8=XO))7dNKPP3zR{O1%X;Md_sa(em%{ug7t$YBwGA<hGob
z?bkZ4^nCXT8n0ds8lj%m&+wL!!2GQ?zUPHi3j7ZcL>GB^Ba4|vxUp`aHHKUJjNFeP
zvdMIdmtBbgo!}`n)|PD4T{17w|4EFiQ&WK2ZLK$nZ6|sG=jA>J&Zvwe)iwT3;pWSR
zGy5}4LmD@_UAD$BWP-U!bitcmh8oE;$7+0KVV(O#*~n!j0T~)JTIwIa`RZHHf75Q^
zHsqVuuW4OtEKT~j9yRi+Q`&@-5mh_c>*DC3jmcU!MH)v4->eN*PQQM(yjI$7?sfLF
zP0XNm@o3AcQ--YqBWg&B-ar$9%=!qirW1buM)LLNK7iE694x$3e@`3L0zA|}!uD3O
zn|Yv7;{jB#aVN?^rlX7$X(HP#FHBsp9ih3=Wh7ksIn;ZQ*G6Oh;Sc}($PB5@Ax2G+
zmoh!h=MfvZ4y8aEOtaJa9!BqH-AJ41{EphL>QJvY2RlpRSH^{h&uVVtnK=HT@r{3#
zD>T^&r8Ix%i#Gbs7e|SgUD&CSfPH2QR8oH+akyBP^ZJWw1KlqJnI$deZ#Yg*YxsDu
zIx*q>^-z<m#dcy?0OCVSoHC96q%>2pQWXh$#pVN)dXT>DX65)LC-Sq<{CuYNYZ-Mf
z_oJi_vsybJ4a{~D-o3bSXIGNqWjg!i*=uJ_Fnt_;1Yg7Y6Qr%zJMAi0hFR-yJ6X9|
zVX2@{VduGXa%9H82z6YLxT&OdT4(B^7uLtP)p&<?5mUVC!Ipo8PBTRZkzQ|PcAAg=
z9O`R=QTDpVQ;}X|cjv1_)OPjB_c&T$D1O*pDF9fDCec8>@DGjob&c7r=DXO9hiX~(
z26$hFUExA>ZGFK${NOK^S8Y*)7@eZl*z=kQuLD|<m-tspL4D{x5eJQ(lMQhbkxQw*
zM!p|Hg(e;wFe+6ZzL=I^cqWVr{Pjh?)WMS`BV1_Q8;<{Vq~1fa`c!H~*wB;7Fo&lf
zzD*F;hwt~47rW6{nsc}{sxWgIu*M}3DYG{iOMSE+g$h6axm>CX(&HZ*FzmahwE%93
zRAAU$#VM`+Vd+J1GK~x~V-;pq3<aX-PDVH+M1+5soDaWUAwV>R=kDHnPN|HwDr^uH
z63mO6JZCNq@0=EjSUnX$9S!=Zq4VnRIq?FUsrXwo#24>5<Nk=_7rNpF{6mv!ca~h<
z*(F@kO#Z$5GojOY7A;}{e<PsqyAjGJ4eg1`<d{Ch$W!W+VIiU4(3Pp3)g*u8E-k(}
zNvg7i4j#dQ=F(RTRQpoJcl&2j=6Qj%_($2a1E9X&;;%L2hQdfAn0ads&lxcOp`D^#
zM@WHby#LxT-7328-1T7|$^Ut(Im0uWy&plVl48iTet%LzK!#&^@;^xa6yrVWvkTIE
zydLb)^D)ksfth<7z5dc!enS*P6^Z3ojORDADLXgg^h2;I(z5wUPctGsDCpNqLaW36
zszD*$KR%O`4_{WwLqZi~Kf}x99ufHSv?A?B&NKKj?Mt5oG{(0J%$*iwn8mBY6EHs!
z>{2}NQl{J<g?)}|eGhH*d8pt197H7Uum#Ew?R`X|BIjKWdPF>Xew^dBU`XM|5BO4$
z(j0SPv<_p16mR@xQB>y>(4~SeH4@Er&5znZhWUNXYY+M_-*O^>3etq=gamlF7Jbz1
zXrBkB;!9E)UcVM_6%d~5oiS));KpWpbugkDCe9l)IL*Yl#*oQYeI7&<7yDlJY$kt<
z?ekX*83J|tve4BVHj+7X-edk;ij<+)2iXvm_=!0!_H(Z^!W05Zkl){v6Qr=@RSmmM
zoL>U(Yn}`tuZCQM3;1PqO&NNDyav{;=cXS&bs*2}DQ<KDth7KZ__6gvAoqOHf;n-f
zJkJCFyc~*T1k>czciWGMr95D7A0~-?g!hm>@h$N6zVbx$e)d%TWl97k3ZjJ#Q`p4u
zHRCTxqt%$zya^PqMhWco5xVq7q}@GI>fy49cHb?sJd@RbLYC2dy**tI+?x;0j}?^&
zE)p&n^^@7Hb$QU(Db~AIj;P?EgNH?!)|sB3(e1rD8Q^D||8=2{?D}8Ky=71wQ5UW|
zFt`lvgS)%C6I=#&O_1OoAh-<f?(QBW!3PWOgy1eAxI@0ncTUy4b*s*)`}3Z^-Mx2L
zb@$%A_Pf`5UPw#3B0akOD`(hR4Yp(78|~}R1`0Bpo3y$v^>P{Rzm<WPQ~MvD?`?vS
z&?Go?IeVo!&jToXD_qEtpF$ajf0jRt(Z4l@JuFV`y^jtqxyLR@?-N2_`(@B`_J7-<
z6cVvyZc%04_hN172quKR4&dz`Xus_L&#vvRX1`|91enyA2ZWlhJxT_u`6OQd2k-+&
zV|cBE^4(VMTcQ7-S}bd|y(FxEO-#XsHIxC`aA%b>UkJ9BcBd(#?i4QW^xoZ411}dg
z+MWfQZ7_WMGAEgan)saGsw%fxN`w)m)e;Kl?=3t}jUu`qnqZ2L#@VRd6s3KQ&=Kb6
zipv8nJkZ@jb!LY#=gjcx)u&C`bywX;AV}@l%Un#y%iK#*IqKDD%0I*(gMpo*m!j7f
zRA>1*=NVN3V_s|QE-jaIG;9FA75wd+r}hr|dUA8|BW)vm_f0(dD>;1W_-Pa60yP~t
zGDq!VAN|kuM0;Uwl<Zz%@Culv;g87hgHM3{b`9!t^TU|Zzh#}Ucb)9*&wqbB6O=7*
zut&K2qkEl8M1z-ox;{tdndkoVy|tOCq2dRvS{Z*V>(iF*f?D%{v?S+_@wCmrKFY@{
zM}1;}kFCEinSOO$IkgVW9JhAn*TUWF(zE_jV+b>U^cSGGZ~qgxWR$6Mr=wLh?)I*d
z`z8Gcu#(}kiMkc|#MTE`62HUY|E0kIT<#zoXQVzvULwW-GU1YzKMbiS>PkR?Cx$3|
zo3T$Cg)PoQgmj5$ev~DZ{RBD}P$|l*poey|g@Ao=jh28e|0nssrMq`JFMG*Rujb#+
z_4xm1tN(eX0d=iJa)ByMa`$chC0%t>a$>QgN&!Cj`3>^=|9lXtn<7{@2pX~JDi=vA
zFXqS>)w!e6|14sz+^YkxQNKDKeGUBox@7z9j@S2CauVnN_5Fek1fK`+|Hqx&wca1X
z_B~6;|L7VjEeZ*Hw#LCS6{iRw2$qcbA4iL^st$g8X``N&KES9bx&t1+p84`^i69+P
zt{m5*L2d~y!dOFtsv`O#aeeAf1M&U-Mc7cPnC`t+4op1$tCAX?VIx0DeItmFg!n+{
z#=I6f_ue-S`lpWj(w<&?woBcFcA`&B>NF~AI^QA`RFF*tvnjQY#p5w-Y}U8_)M9L{
zEY=q_q*bAYKb1u7B{N42G{|HJeb~32pPmxAT_kp+_8Wp4G)fa~L|-CEMLu>5kZ0E0
zZ9*mxOcLOnLhldzAM-Rc8hUnBXoCfNEl3~tQXz)Ha-8p4egf4&gU`j)Un_R7G+FfJ
zUgjR=tD*u;Kb0sD=1I|~7c6*W@yWG@!zLsR)otNA%V*2+`#C1h{ThbSNa?KmP;*#D
zb=&_E?E8PS8UFv~44b)oHY7|VIdZK!@!}7x_dRaO^!&MG&&mqLx-KZJ)pDTRi^?;9
zU+}maf#QnR`U8>3fAe!n)C_1!iAyJ$Cq)|=VqGSfHx~X<de0^1;-T)G6*M9Dab^le
zxk2~mJ6c}TvTvI-h$Wi=1V{D?zhuP{3`P4i<L-QMnV5%Z(wyI{aGDN{e!&WSSgxw|
z@m4ft6ZV*37Z*GVSlw*g3gD85=ZL6xyu=0tw4V}Y{p%g8qTT9?e0{LZxER)gz*BW|
z0>1%f9KHRK4GyP!P$vTo3E+-ep1R}foo*<4{U0u@(yv~80}0*SE-`BZ=E(hfNP{0h
zZ?v8l;$GA3<Klh~$JvPOogUk?8%xeW*QuHs&IyTW!qbkAjyQI3a0}l@1@a#~uU|?s
zs<Pb<-|cl<8<9a0#c+#(BwM1+NHk+ufGEr_@b@!wf2^2^g83;a#3$u%U&|kD8)F|1
zVb9gxOjsR><_Y%<XfMuJx7-Bl0dn4^Gku&C{8ssIGpf3g9Am4WVeElVQIb01ih(aR
z9#>y$8Z*6LpKtEGn>8>!AI?V!a=?yfOZ9Km=`X%OypaMm^%He;y|X8`yQoA(QA21=
zbo9E<l@%LDV3Kz;QSf7lZ%`++i8}KLiehZMTGT$*=j!se$(WI9ZU0=>Tz{m%c7V)^
zWZh+=&C|!EAHfqFuu1gk(Mrnw(GJs#_WcE^Up*YdCkOZ@QfhkJ!KDYg<6tJu^|J2#
zj&iUZ9Tt7rfhK)v6zc@}91<>yG{6nyt`b%NQ#;t)E=f8CbJ{Nof<!%jZt`FPRhhO_
zzb^>Fa|tE9Fgv^W`*(Ld3!_tZzqtLaaZY%`J`Y>cNTd+Q)N%FMsy!?X`VT;@^XR)z
zp9h|&=+4;VNn@z7>S(|5sP1_!o~KR}C-NocFDZ;WiVgmI|1LwQGp0J`P^HL?3Np?G
zMZoHRp@L#em2HY`zQ|qq`whoO3H%4M7^2)?f$4AH_c2W>DNzu3&8F*oqBrnztx4->
zv<p`dD(tn+8Mr(tC)Bn%k_LcyRaP?A)CoN_7J5O5!r8M<X)A&S$DFCNj1{{>2d8#_
z`3hhd<elg*WnXHw`WzmW+n+&@N1kE_32|^{jE#osyAb~f2OBtHc6jfoKRzEeQ6tt)
zw~s1c`<%%UU6Yx;{rH@|(VM-wect->!THN8rK)1DZXk~_0Q55Z@s5$Wm(~quObUMQ
zhhRi~h8{KmJ(0|YWHX?}i5ub2e!q&mM~plJxx_$D&STbndQgn_eDU)tI|$GOd%Srw
zUlO_dT9IvA4Y>|0hxr)rks*0+!a1&cV1(}E_mAQNDXI3vkisIB7i3q?<sJJh$N^HH
z1A3jit~!x|X_F#DM|LrhNVUuCrxN_jd9VN00<XRyABmBgt-xW*Ib*r(Qh%pzj~pN~
z?^AWtiEtYlhcJ-u0^s_*tu4;iLHjiKrgj5`9DZ-WrEXo(i_u?pDE%aF3EtQfSIF20
zj%Md-=(Z+gMvrp|#5}zPfmd&Yiq^piUEYpC_4j?~pf1y*;kCrV9^_x@FAs+KLC+{I
zaG+_!W8pJ5^eAy&ilUQ9{6-NBG`6C*#sdG7&wR7*PL+Mjq{zrt3MBjn7Q1bB6uR;f
zn*Nb{PHodGIB@41Biooon@+2(I-HFqK8e34rVJ=Sg4pXsaOfz}=3L}J=>>H{$(Y6}
zH@#<Dd9r~f^0RVZZna)`k%y0OO)gEY6Tx9>+d-W=GYc559|d{!t1oim8=4jZBbPga
zVom=8P~s1I+MY_Z5Xmi#ej{r&<<bLfJdj1{Mh>0c$5M8GsW`rnO@LfF+dErVU66hJ
zy|hN;P_ccp*ecmkz`{Q1xhp2>KM>O?^f0w}pXhssRojoQxcy~Z5K$aN?CSJeZYS%&
z(N0V*AP^%6xi@gmbe+Y%F+yAsYMk}*2Y9ivz9gtX@R<_-5MJbF<4+$bz#}o&b+16;
z?<Rq6RPJTaOSOM={NM^X@>g+Z>MvUyw2PSJn+e3+pVi!47WkYQRTJlX=^<uKj=j(J
z@3SDrkT2##uz`o&9k^9OEllf%{t8GqOcoyy5aKFiKP)9i9q6<9+k`#q=+~sN?JBs5
z;;Yvr;46nUp`_LMlt4IV)#h{b&3r%D^V|7eB~!_GcU)pZkV+9|w%C#xPUj)5vrq5#
zH=~0FKP1ca-yC~rowd)1Uq(v?IirJOGyLrYI}>eTay&kwc5|nH5KSW~b>A^GgH@6%
zMTi~2fx;#p-;8)^r>;x$_?;s8#+h>;DLXhN&}VxSMqv(Cr+XHU350J2%=1Go28dmz
z-Eaogv647pi@l`Jv$7fKfeLxA{ARsK7UX99pQ|J)dvCOdwcauRPJeeYgYPF&Xa$Yh
z=p@2sZO3scE>7K62p=vd<9%e|vfla5a`s)Lch3(H1!rrWO{^MP!B4P^9rY&~{*6-M
zXV06%=mn3&-&(nmE2mogV+Pc;uO?hWQ^^q1PUnO7WP-ggqezF-qe3)a5HW&5ID(+>
z*=L$a8a}Fh<AFI+BDjs;qyjxpm-NM6tI3MC(^i7m8-W`^P1czfsQcgpZElk1{ru(Z
zYJWY}B2M(_KUZrL(wM)mJGjV>XC|lTkp30O`C0EW8S*+kH5TF=<>4SVko?^ld3iZ^
za}fJsUQ8Zt5YQp1<a&N|yp^j<Y2Mh<+8mbX++!KkphkACb6B%96b*L2Ghdk*UIOaO
zi1=jYu&Xp4oSt6nRCXx9f7&Z_{%0cXr2JwzvMJQ|zK)pr1AVa)L05YZ;u?P?n^2NJ
zJ?;Zf9neq7XfI;qliQ$D2%c7KMPMoZgTI)sr1}~}v~w?)@^;fFlg*2c!*_>B_%;g8
z7~?*pSe^<c*!^4*<p0v->^u2s-(Lb5yu22vv)i?t<M)Zz(%#z=T0u`b_p6J!WMb-l
z-}i_dqEJHEe|Ub%e?CVlM>Doid{M#@|CexYD06Rn6;nSzqW3(uWcRc=Ukrl$f`;{a
zult?mWf0D-SMUBeyP$c)dRtN^)+U|XFhL$l;wsq3VZHl}+(eYLWcxHf<`2&tUbg4C
z6D%E}w1#Bdz+>J<SD|rX9~HKq=d#PIeE=;-RsCcC=^G^@DL(R!q~$MPW6?F_MNvvg
z32Ql%f=mc~rhUn94k9GSly-XCZ=9zZ?5|g$PC5a+kR=ISC^4*_KRK4r8ZC}9G|Xxm
zc%bxf{6yL*@)poK)eOc7l*akeSV(;^Dd^QM8+CG5Qaf<tv553={^u$^ho3kv_it<#
zHSz)Op3{#w&THUGt#k}YUxwL6H@mU@{}GlI|J%s+$JU9p|0+ox+=zGPI}XoLDVb8!
zZAD>Hz8G33s8O{@A;SJ{Go^ye`%W`;)R>C}Jz^CFlf+Ih7GWX_)?!Li2yBhU5B`aG
zkbNjgl}xZ$e8Lv{4?x`*8TXB*`cybNWu0P9OmxM>)~bRBi2wqbe<nvn(v(-OzFpsI
zu|-~TuUx!KVaAS$s|<XH1QNtpX=LS+#0hJNrm=?l=OwtQEM-3)MFw}MW(4w&7LX*I
zw((1r4W`eD@iRJ}TNQQIRXRqbTci!Gk)e+A6_^Ro*Ra1kug0<&#8lX~9Pq6X7H8%W
zO-wS4t;C1~+l2~Hhv0%+V)s}REAkdZ-4K@J*16qT8xbl>-z`%Q>$S#KBJ?7oZ{t+h
z-3XzBdUUR}(PDTjIo~q$*gyHPe~eXKQu3cyip=mFH-Us#^Cf$BsVkV%qYSddYljO2
zG%Y5D)5lT9^89(t6PlIyd$oAF4)<CdtB)4PCBsWVOfK6?ef)<ttaB`VrL{Nbrw_b2
z{s@y8@t#5hBbS;nYdqaB+6hKn#bpxYxI2W2VM0#@w_W60MTG@t=?HZyv^7ndh90y+
z|Dl7N%OJJ9EDnWXdiC$(KWAds02VS;w>f>TZFdmtP5#sA;f5eSSH~{DPE?eb)|7=q
zG2~iAB5!oIbpGI0#8$6F2TiV7{n=#0l-j`N_~w)=T8>|7xiEupI$sKErflTQZ9}Ue
z9RoV=|Gtfcpij9wkiFy}3YG9}XEG<6kY7c6n1@7^5H|@92vd+>u|{LaVN%HJd4}>t
zl<9n&ui+bPm;0tlJWWxfiv|mo$ibzV!Yv0e6A-{jDX8G>6(c0$TZd(Z>f&}BY|@F&
zhtIV*CrmD}=;JB5`b?*}P%3^jR|IX)-?FrpG~4+>wC{*eC*n|GK@9BG@;*rn;nn4I
z-1Dgu?(}Uxu8DWd<XYrv7v$|0j&ITyU)N7#x2rCQ@JG_P)Thap`F7W9i`k6(h3B9Y
z#UTh5=4Z*?)~cpn8CYsG7TFk4FkO@WY2ITqWEQP%rE11F<><ol{{Zln%T^dJQBxe(
z^3`;_*q&>39MsEmvr_{?i^(eN$}r0lba-*E)rYh);*?;+*lte>Qh8U@5lA+8Vc!yT
z`77aKuwsb6v7wH2*f#?cB~vD)0v+*1F$flj?$%4Plb4~vxn&-U^>mpPi^FnXgbHbR
zArQ0y%_P*SpXK2F@LSJra3uHmuP&#!leDj4Hbugu%rOXq<gs>SW6wNDlTAxrxwxlb
z?v+=mTM@|GhJxEijhgWd&)OSeU-`m`qy@a6pL*2GO(gX{?2?d+xm#T&gD25GGh?S|
zjf|$)T@SE*Qg?95(v*c-8z!jFXkr?3>s8LJZi7aV+>9(tgGZD5<7$XN^Svo(6O?6*
zWdx|)VR>H`+cx;5Ey)>VWmB{tzw!hnW=-Pj4u3e866Fwu4akdgdbdmI;7&74%|X3z
z)*^)BlSA8*)u*u2Gx@RO$G-CRm_`#P*=G<0c+u1<07%)7O};lX3Xr%Kv3b;{PY_c!
z2+`2h2Xd^&rdMR;3=S*HC=}lTDR8sOvZO~@)cUE*@8I+jd1Z(eSaS7E!dTg&;@q7)
z4N+irG?O{F1WB2F1Ou!o$=_wDCudu6N5U89i`-9J01n)M4CnZfQcx$W^ht#hKGkez
zT&)mzq#iUtpd$Dot^De%miA&N&H=7t*e8T>`E%t;`<#N}8qe!Ariuz%8tg?pxgP_5
zu;y+DzOvl@QdX2%_$xH>IXw?O&>|ot+hfjjAsJzJ9pJ7|hiK4AV29|9g-cHViA`*G
z@;jn@r$6i)mE&nC7V{suqErY0X}&H|Rch1{y`kzMT|(mY(s-!I7ls0E)c~%5HnS}3
z;sZ2Qn{sL<21u)|{fz-I)KO+Ica=8>BG5k4$y{8DyTo2$aI``AI$$QMO+3ca$V7BP
zZ#p#fW$?L+E1{8(qMZ<1g*CZ(g-$E&n4ZLbQ0YlELx(mKAtjQxy0vBDBhB<6drtnN
z@StA{o^@@xgK71iDk=#CsDV%2MiqrX<aTRCd{uJc#Kfr(xI(_P(;;D<5hT(ee~=Pe
zcCU2WqWdfx5eK%)CDnd((I1wvBCOI<g1AC+ikFA4Lj+1IKLDTGb?qO1)@v&Q{1Dwq
zsNw0{g->EoQA;_+@*@y<e8PBf3WSgY9ahnBT?uS3vmGQ5p5cdL9*hUS-=^xU$r6E3
z@IM1#f$p$-R?S2R)G!?8z*Nj?sPpN>s@enC&h8bWgRivzhDEEUG1;9dnQxdU%nhf@
zW(P5*a@e6>AOFm$XBXbOi#qDbUF~G}Ycs@zdx&Uad`ojixt^ZsgsU?4_8%+iN}FDr
z<ryw9e>o$6O4E*WlQJE>1I9C;hsC*q&3OJ!#89)m@{Su&lAt1%n_|ca(fCQov>rxN
zmP+qBmcd<L4r~@fCv6AAJEj7SBv|;3;$MIH9+D}n$S=ig=0o4jf#*8geGsy?;=^d&
z%fP35cvsA)Y~?gzIVSy78HwC%37VYBEJi29P)H#>yP>vrHA$J=_{oRbV_C0(PlW2(
z%T~a-HX+{QwL`8VrWw?McVWRh$Yk0=m~HxdNA#;-(wzU2#^sdKvAo6bCQS`FBTu7t
z+H#D>{JbAsJ3hPihqB_u^uhbOMUjAbT3St}3Wn%;j-SMN>J+&dQy=2mQk!iU7}apR
ztxGdVB;>C8pMrydI%;5H%F^XdPDy`#5t*7$QFzFlsmA3){7=A-I-%eSriY^n)MTde
zdM>U?-76NWpkM>y%F|oeY>$}^Ge0V#5TC3noBf@XY91^_vnCpS`gKzxoGJxXNRpIy
z0eqrd1q{<9IZN$5XCfaD;J*HAYjvVhDdZ`uB@z{FB5Y+Bk-nU>Tv-gOCK%KLm)BsY
z(L<W0`jn>9OPEsrHcM1^S5zz8FdMN|HT@duiYSI5PsRQm+YTuk@iEDWZi0Qw#&j+A
zuAb9=;=bA)Wh}=^Xfvd~O{@i%A0M<mDC$XULyU^G2%>B3rHZr{)20-%G%E0z@30|1
z%Zi~rX=M<8GC^`hrb4f^RPCQ>ZZo!+w$VyyT*|6@$_n)~n5qd9Y>q$E@Q3w2iw5v}
zJk;zlzj|uc*;@k3PIGmkqUOgbjp&O4)pXoZQM%D|A6TB<2)l}*cW#zBm=^h(tS47>
zXpvzln(|WBMfno@w76H?Jj|i8TuOF+4j3XS%o&~k(lLrCv%`p}qpSS=5)2(KMvgg>
z0=O(>G4k;|?crp(4K?S9hX(&VH)5iMb7SOxth+={@_Y|(f8iT7)pf3R-N-W!(5q*p
z2mGm_YoX@H&l`h1d$C2OKwWinhVkVe<U~8sbf-PNkrS7jH~a;0J?M~1ZM6~-b$|Jz
z&O7MpogRD|N?Ae4Dl1;iFT<Cy#yRlB3@}x;gQRC93prieX&$Dq#yG?+aq|l181&J^
z#nMz6)5Vmh*|oHUmdA4#)a0nea#g|f%4=Aw%U#9D1x+HYRF*EcJ&?tO;`FtisodI)
zYlXjY{;WyI9-dFj+A9+!E+k|LUGq-Mr{cN{qe^FyRN7fmLvVCU`6b{Qk&+yq6)LTf
z%(9PAldVRLP_M<$%;V`o-5KmEJ;u+Cfu5?sn9fDLNE=7j#0KK~jB@H2diMp%x4m+8
zYv?B{Zd*IIHcnOf<XVkfyQp3&ZZ+a!xMyh70fi0HzB?ltDv1wmR`Qgpl6<OgVHBwX
z8}hCsyyL`Al|kaxm7!YUn=2WV;?dk^ijpYZwP3P{$8Q=R<+P6jw$WWS;U<Evz|Rji
zwNxwQ_WhR6J?<5hg>~xjSD1QHV(C~&9WU@tj6$b`>)^&wZdd*ZjFYKp-C7+)1-KO?
z^91Vx<NB}dJIB3erp-kS0d8W50RwC`CUPD-=?{m%f{=_5*ZHU86|ox1m_e51^cAW#
zeEn}%8#CTtqS(D<rFZHFs(_@=r+-;vCL$YRTENIXn?!fMW1RJKza&m2Ypn;~`lNM(
z1+gAWw~~IVzjmKtivB66db<bPPv;%3M>T=B!L3`RI4|Hx(!=aaa`zYT8yNOwR3#R>
z(8#T7XVn9y)?whX__r`x%x0L9c80-WZ>ON%)No7V+V92hFZiNZk@MCq%u2i<J_d0X
z;EQD;Jaj@aLVN1CEjf6H4E$9}U*GBTR{7ocJx*`%J-j2;nxRCyv^j&jwa?h5m_qi^
zAqaFEbA9=k2rO4v`xp?`XuMB2!;PsJ`M!-4ZfUlq0L=BPX^wB{{5z=_WaRw`AlTbs
zfq3dZ%Be7;UlE}k^rI&&VpQxQdfk1NvnqO>x7TuT89+Ec#`%x)(S0-#Tvg=qd&b*(
zCBMgq0xpGbM*(cIoM=wpKyU#az;@plc><hLk+nA&uz@eSx?1rfeUHx{Fn3C7m2d9J
zOg_7IUcq6LBl{i;N1Qvp*U7+0<pRh%|23A@?iJdCar|j}ET{ki>w;vM;X;vmlbl(G
zRX_}yM(&FA{<B-jfL}phKM~*zh--)rqov~9oCwv92UzI4XT$00dEt@<;Lc)BFsi`A
zar>VL!!v&^oXLGcJ+dhzzWVW_FHg=}YgN-<ja$YNJM<G0Xb4T`_)H7iunK{{f_~ab
ziER-9&CTPuF6D$gvh0))YPkSvht^gGF;6y*6_)oQkm_9TU%$_jR(N>2&UK442v5}m
z`?qu1)mFaMN=3q+&A<FI_tBLO(GDpO6gFeuOs4#CpDYV#*+g>=PWsV0mb^wub*l-Y
zr{hk`wX}H@Tj&)W5V&z#3~lr=lC6>;^p2-|=~8J%j2k&zxbisM#ZDg?CN{I<C^V+*
z4t7UU^D+~OTmiIp{<^<nLo(6{M)7n}2NEoE#jYlXDCvK9hsC0xt~L26FmHfs2<F^l
zXq1>8RNF=KD(K}u;YO5>G*%vfFj>T<We2h$IhlA0!tClwT_h(hRkL<m1?g+P<Vl07
zq6b+?z?6f|U@uaL)r4G#wwkYKp0WNAn=fa%4!3q%`8cVOr%iNq<q}f+NCsy`8h&(e
z`}3Hu&!rr1>-EXvQ^<OZG#wOP{xYko#sm1E9mD+5Bs;;2AIYgnl`674_>&a<mf~q(
zHFP2voI!ed{igRVYB$KJiZd<E{<F$T8b=3x@^Ct#dtT$=T;y(02)|&0ET&nk4mJMI
z6nbV7`-znB(K#gHfF0rS@Mcd~>dX%uGF42|f;2mZw#}6=3#M}=n$0@?$o8shHbq~K
z<?c&kt+CG|d((SK34RCZzqz29B>WC6GX+8`zcLbog#0O@<iD(Yovl#!{*Y5-MLt$6
z$S3p(dlc<c|Cg?Z3IEI1sq@Z6%Kv#tY`9b_bm3v5ceNH^?i$|_H4khs4YVS3PPn}t
zEQnW-4199P*p~p-ygrB38CBhW{mQj?qvP~%_%E>U*U(lfzDI?h%h#Xv&Q#*l_uju3
zDy*|jT4&6$2zL@i#sHTTQ;8n<hlU$X6{I02Q|x13VOEmlyio=D{M#%U`{L8wBL4w=
zPnv1eDW&f?a<1e%+xc^wd{YbA7V>xa^^HX_%jLkuoNDE(1$W4>{zc2Ha#6H>{x#re
zl&XVkCG&mg^IEO?sNYQK#z_hNUHnr2#r3Ic^SKQWis(9uy=H+N+vSvxZdZo%>SHJR
z2fT^X^B0)qx{ldCNqR%mwZfsHyiG>RO#UjR*f1zIDr$f%b~oON?Y1(r+Ps|8kbm4P
zMtDI6SapJbpxURM$g&aJC97;yXKK6Xu~^`mh~BbWZaS-neO_@x`pFV}2=)(l>Ahe4
zoIXlD(f)0EgHpOqYDYP-<(<l-#tYUjHYGP)T-s>urtKabCd71Kkp_3)9W&hpy-3z1
z9&2c<X!JFyzeW3Ar9wz<c80o_456hs{|o*+ff;&KE$fXH8{2<{8ZTNxV)h|tVI0&o
z`<S&rZq4Lz-GJ`>@NTtFJu<3e%m*K)m8fPyro?p0rb`DNnI7lOn4RY0+&eEu!uUy*
z>DKe&`~u`av_uu{7=E)B>n1Gd?yhg9LC&t_voCWm+Ry6#+D+HZrOJ3Y@{!|C6r-gd
z7##!snf)knYv-!l9rh7kGb#wH(>U@92HgYTG~L}1LxF7CJUlKKop)FmTubX<Um<@>
zWt%pum}08sE*aOsaO_tp^f@*n&+<96XhS8g@vv8t-C_0g^4O!xk!~SI<>|b6Ri0#>
zh%dL>H4dfA!y&hyls%Y^yKs%t5akrm>4oUx4=U1b<|~^i(zC|HWn|^}T);C)-iPGN
zO4JWo$?uE>LP*glI9}hgX(>^}FJ=<A+HE>Zo;}s`ip$&_qhkI&TMeeKs%-hD(5_x(
zmhSsw*(Uxt%BAEZ1C%`49z@olHqj+aiatFYeUgCBIMQo)J?m!3#yotI9pF{r7{5r)
zoTeriVM1-CfGJXk?v){ehO0yTX$Vn_L8JQJsZIjIXp6rPp*=j8W^1oK$j7a^kEaa%
zE~G<LmV$^SlV8Z~Z|B(}<I5WHg<Mizjk_}N%J>02>p8PQw_Ya#Ivxe>2xrqNXKstI
zi&A147#7r5;yB?B4W+T7`J1(9q}gOhM~FCnvobNRhbzmY?VKkhn357pEwHVorZPe$
zCgq~rUsXKh_oIs!8vS0$$1Xh;k98+X0r^=ke_WdRbM5-}+5N8l=Ay5}#ynADPU-S$
z@9NXA0vd6wXo&#a)?OGs;`prPol%s+HoOF{RV{7K#no_hOY{Fip7ekcQTHxhOz@c)
z>-J@z7QRd)zdq}B)4VLN1<1zN=XZYL%;=GsCzTO>3+gPn(IIC2xngwoCUJZ=By#&4
zn_W}g=5QXnu0*?)9-+xp0A`qpOSZ?4AkELu$mgVM@Om>-t)hB2Z>K-Bv-g)KZ81$F
zn5o*Xq#aWo>O^e*NI;BbjuuzZIEK;0p{|NtP&Zor*f$s<-Y?EOboPkeKMj|O=1>>|
z^-*TEswJAkX+@7v#&JD))fuWvcdqBqQGr(LVqzZb^e_YNoO-l1{3g$al?bH#3~bwm
z5DbRUF`|*OtO-8MQFCJ+LOFNdm7#I?U=ZJ4${ztL5pi3t#hrVWZnor@Zscfu)u<8r
z1~U7_v~F6=)d^+BtsFkV@^-W|#sg}aXcjkEoQI#Xt&t8LJu5DvbXr|W`0OusL0)z`
z#A7DcKqL(5e(hi`;=N$Be^yHLTS96M+!aM#lQ^1RcBE?kvG5=qDOL<wBRSBpx-q%y
zJ8^_TY?U0si2|je>?~U0bmT&B)!R^CyBG~R>4(>`O&NFN&+i$FSkV#|2?6|K>zIbx
zEDDyRW&>3wZPY&g*L|g<X7!MO))$|o{Vsp(ufYC)2Pl#-&=zc1ul)Z6<K+LX3LPhZ
zxTV7RZ>Ofu<?N@Gh`mD-^#1q4@2sIgPlr2<4%Rm|@XRYtT@TR=uoI_<Ml`RgL<C@~
zuRnDDs!BPCx)7BZ!F#3s@!4Q#2Zy3C&?t5iulqLmxctQV1f@}E=5aA{NKnZN548p=
z7!xB0f8rXZ#?f@Au4b>D5n6jR`(aos8@}rJBQ4wf{DS;-L~D3T@Ug0VVZvU5Cp$h>
zoD@}<GBRD%6=Uo+$+aanpg(^OZ*b?-t?^NbU3u27)f(CgYhcVL4P^FANwh+JuinAm
zZK&zdRl+zPoQ-j>C|PpOQr0>4ZIGR(EjjT|DXa864B`S4%t)jp>nd=>*;J%!^s6bU
z7IcOW`s{3uVY^HG7;sB>DC5ds4OcySbkw#S0W&GhXf7Ff-P4_nszi#3@Y}gEUH`1m
zd<8Gt{-t$8hYF?b7Ezs)DA_V!h062F=^VAA@+If!ap+edTcRrE(}eIKz;NiM>|UP{
ziuZAy6lMpoy<_^ZgL!>8rI_{|i;aLc$QbU((nkA8k-ea{H?S=Z1J}U9IlG8CA1L3m
z;<i!IU7N2=Gss=p51gDHd`rwZ^kY$Y%oRoz$7t>?BmwH=G#8b!#2>_DeNrT>A;OW7
z`U#xsxxQhSpARs^RnH9kz|X>goBd??WF#A^xJxvWk{u#S&-a&+X?=3iU_P0uiyNLw
z?DkVL5}lx#C#B!fD7UYtkLn=xM=k&hHvr8><wlrPSey6**HCr>htLIO)wO9yy>DxY
zukdpbTAxlcnsc?X_6-R7(?ZjtkFMeltptZQ&KwqVyPu|8IX-A<;E+bo@7gUEUJ|_C
zZ76v;F^|ZrpgqE1>&kHRq}s8G9jww-OZc^yQKX*OstE2hz%1>GoXBvQ*r)c51x7C7
zn?F<{@19srTw6r!TihvLU{Hy{+_lWi*mQ)Z>set)kiur7k&{WMZx=q_c8*Y{t(-Xq
z#;v6l`beB(6CtH%+>B2(0lMYPB7Ls_IWy`G2?M^v5z1g6W*3G0ZMr@Z^)M`&g9E`T
zpu{}k`L)iP6NE;-2rL=(CKo{&hi<m_qF`uPN2*P%>{^3+jch9QEDEXKzVHySvR`XI
zr{5|acOELwV+G#+lLcx6^@+AIpnor_x}MM=&Xsc4$K6DSh<+mU7qosQ;nqcQiqKpK
zMR$d0b)r9kpE-lv{xdo4uqlAK0kz~+r3}Y@+;AM?@i&cX<u;>pe1gsUD8YVVdVJA0
zN}UjII6GPwP-HY{iM<c*x_66r<zPLMitanugs+<(Zf1Wfz5=@r*{0QVTk4bqsv}B`
z<k+ky;+`sBM9Uf#vREXh$!AZD50|Ioen4|>R!WuiEJvw_EHo<?R?UYXG;Vt_$8;(7
z@896Mjqmf&zVGIrD_}KnxSwQ?XLP8avw&@l@Xq7?&Q|5h-6{>ZPa029Fk}P#mHJ0T
zSWB-#MDeKh1V&Z#x?F4d#hHBVc7>e6700^vmRiX)4ZaN|sZi#@ENyq4S&Ry6X%4Ta
zDAwq*A-(bsGepj8icmSOj}@0968B}t3Ve$vi``*Sq^YrpQA_HUr)i?1po+yPSIuXT
z0){VQb2=lc1ODms2NDizNm9*jtkKK`EiD(->1maW0bp~$pq-h{guI9GY;-x1tDMx~
z^bnQ?hHkiXn;{Ws6?f-%i1mr?SXRRcPZ{yX@s5FK;Lc!@9)Y?fYXqsntn44~^enUv
zDm2uR^bCjQXYo~cW+^(<h6WR;V%a5N%TSRchRW0{FOCmRN`~5sDRV`C@D0Y$IN;a=
zObXdbzk4o*4<%b@9J+P8QKt2n6@h$@R-wWgYSon*@%++c#8b=+%pHqQIu0%pjWL`&
z``y7;>~Z!yrFb}r#8xP(BoZbJTJ*c4v0TThz&ZIHoB%&rY%XBfC}H6|nQH!xS7-5S
zan$9o%(P@UmK6^<5w*oP^h<wxf45)@xyb!f$x>X~)kGnVE`eLH<YJ6;EQn!D6rUPF
zuzXkIxTon=l7Y4j{;WIJQyQ5%W-n{e%9vqEfgP^#v3;6njo0^khS?I~`Cv{I&Ji0x
zB#{j3+XLJL^`tQe7U{`Op~Pvc83!e0E$0*xB6qw@H-lQUM_7Qb2E<4j0x{Tc)NP2Y
z8jRIFjZ}j=SR1Lvq<tXJ70{f&6`5K}8APA%R%oDJbzMwb<JJPssOBo?DdBoW5B7Hg
zRfRj|e$hmXL$Ogr%l-rSV6`1ZM?`9IF4HB@qz`YxbX%k?k?tf_%yJ+$kU{s%JLY=q
z*&|*Kz2Vm*tnn;2ckjkavSJ3Mv1v0`R~xlu9qMx7tf%rl3Y1G{(@BJzGt+W?=FBpp
zmYKrA@DNUp;jFAcYu+Fk{_zLv(SRG|0JB)cFz1JjvSc>*#Zn}zr=Wq8OumOBe3j|Z
z83>f9f!?Au1OCxD7(&Z+P1PW6$pX#vjJhn?%(%sm**IDfO8V?#Rf*gT`L3rQju%p%
z=<t~1<#cHkSE^?cl!s)bfkd<#J9UN*>yPz6JDF}Ss`c480=lild`*dNyAPW7_vG?C
zC%XX2iA}djGX53}wv`9>d&ntk5v2G^VJ)D~0(k-J?kn)0<WU_KBHJ|KQQfqyT(D(N
z%vHxn<>NBD#kb3vY&E&fjorQkuglOkc`4G_dkbA|s~O=slb5E`%}kTe@<I#3D(`Sm
zYJ=n+L++S|>8oscqFSy`TlVlhT?sxcZ|3Rs?egF6-u_`5q^0gK0iX-QDZt!>EK7EB
zE`8!^#%GPB4X(uU%@z@-=teMwYxYMM_=3j?^c~a*>?>&K=u*5=is$&ATnb7S6+G|V
zZ6Ur|RB8aaWNloQR;~a$jHyxt*}!<Axp=oxH13f+<+|NC6KvM8Yr5pvy@cA<$3+-F
z#R$t#aYgeike}v0WAe7RVxkNZc{^WDRX)j(ihpOx&F5gCuY)rVtiE<{Wt71nV9U1A
zW~vxI<VjITuKW;Z`O`uU>2i~#NBZWfD1g=Eu8TkgCxRE&!`gV<HB(+nOu&lXu7X_s
z;rWkOSR9jX*UxVUKEtO*&<}d)G9UEPsOfXcZ<Er8X#~oPl2W|{EXMw_;oXVz>t$t$
z6%gBZYIFU>#~N*Uunx{Y9KY&cVH6UDBGu#5F#f|etj-#qR_7u<{PDfGy7kGAvIb%z
zB{|Hmrl?y7m`kT;sqL(nK|`xu4!?~gYz~I{nX;s2l;3))%x{tED2AiND4w)f1&P@F
z$Qs=~Ziy_vS?HJj3A%&r$r5s~zDj2gfzYj*r7Quu(NV~o;h}Ezj-&Jw_!&~mR^Q*0
z_CG3PDi}(Ju&MNmB+xGe$>p%kD5{|*rH9+&Q$6vl?5WEHX`{C$O8L|EV|>rmLbfc8
z{}ngWgPxpDgEcgmqX&Ht0KR4s4V>Zmk-D3nh@|p_-N}CdGTDw$BF=aUSAC$(?P3P2
zkmp3j*?TC-H)Dtu<~(h8VKA7aUT?bXOIbg0MIQIg6+Kb?2Vlwq8{C7<9rM@x7co%v
zZefmcMXn(4CMZ|`MOw?DHqnWE@yRGI1xC|X)=HY5Oa(10_4=}1x~A+xm#@vJ-*n1G
z$7f_tBUQ$zmOIICh^aS^Rfs|;#S$gV6(o-^p@k!}+Xm2A=JGDa3f3Gn&>DjO3sRR?
z+656?$VQw6VP08kxP6aW)T67m3Mwv7M$e95tJWgdR+Uv^JFsHF9~t*snM5t0Oeot3
z5fsv(aQ$W{w-#jfsaLE`D;CMddO~G88ABo6L9dL9pF~Ih*82pr=xQ+QW@X&FcqN#8
z=KE;i)2}w@<z{D;Irq0N+n>w3!uCmoh6#-6Ul@r_9d(}`V21|bwdQ{t=o`N~s6(U_
z>ktt9oUwA9hNF5t2VO4uA@4@-3&))tuOov*8;)1&guUl-GJ{0+A$11&iM|UU0$WJ%
z4bCFPw?8-NcB1WD>HoYfrFLDdR)v70<;)O$xNoM<u9)uM^iMI@z6IN+Xej4g;#W2B
zyVNqn?UQfCj=TgrWq?BiRic9OjFFZ@)SkbGK#3E2Zp%enIWn&-ta3+9wx-^+-%JPg
zMyZ;?k7rxFzv-X+f$Ov<xACiq8+@Np7bHF5_JOMu?<%xUCvay}{EHL*xjSgv%Egr&
z-aN?%`FsEMc=Ri>^!XMG{Y8G;3|q$q#b&Q%Rx>#BOlGSQ<I-spSyJ)a?+!b-FSZq`
z5c$!f!EC@1G340K8fOe4@aYcpZ&S3NvxudCZ_>*sNGy-voLXl{OWWPISK93;$S=V!
zak}&9#uwzmSG8PAHp*=-^Y=V5+ln0V{1wJG|8(-jTwK_W)QCO0RmdmR>MzqwELR25
z;4|Rk=Z8%a#5TAY(KDw%0#4Y#<U-`_!H9zPrHrlx(DOoKEIVtzutx``LchS)vcUZP
z@>Fte^P!taZ%Diw$W7H_ra94Eo=c7#x;JBUj3ORHg5|pIK?Yb?6kYp$$M2w*<)3Hx
zvrAujXC|>%W%LJ@CMk)ovYSL@b(%tEBLrm4(UqZ0!EthJtL%c2f;S#BuKlEgH(&j$
zQiW=2T4Pc=BMP6sSaou0IC5JqmgMNqwW{mi6r^o$hGvu!%<<p}FfQc|Ldw1Xt$t$&
zGTd^lx#8(61Voik#j^a3WV<`8)IMJzrKr3uB_)L!|L!Nrk2tH=6a1Kgbn)u)4^hmJ
zD`6|Kid<UF;5uQ~Ym-B5pY11K?V?Bno`Hyx*R3r5>#pkP@2y@9F}bam6C9kTA4YSI
zKky}Y`6zr<vH&YL?x?C+FW&w{PiLzkQIn6Qv)1Bz6NR^RHDqL@r)Q~81kv8GR3WDD
zM9gCCkw~Xzguvr9<;`)u9+n0)7jii!6v~6JXR#(fAu@y14+@d#eZCWEaUDt<q@nF8
zkD6XqM#s2Yk$li?=Tqi@K|A~DnI7eFH7#-zlEf=?FSeO)j>}w?EaW4b+#S^I*~KMU
zJ}`4)PWB&SlMI`O%~vYxm4u}bNszw9#u<<1tnzEtG?36f3v)|DwMi2S(kiQ!OY~!6
zx}~hOwE0Oh6>F-lQBb>M`OH<r)6H^NyY0x9ph_itRScGSsyHF|e~6_F4uXRjWKsVy
z<#fW}s;Ho0<}A2l)3=DL+yJOJaVLktccCHDhQ%GZ_tgezgX$s_z66G8F3DP<99}kM
zTm_v}VXuA;6mzJRBh8{x#!v5@yZvKJ8a|ts_cRYn12V_hBTWAB@GF@kVPYVu_*|+T
zl2EVESWa#u4u_d-0hr_ktO|1Vka;k=46&NeihOqKjO@Cw;X}OgC@9lW3<UVh))MKL
zdQ@M7+e>ThD0-GxTz`vo{r2IWAFog>MLsbK_7u@FK=8$ExdS8=2MR{C!I6DJq;?Ay
zww=j72wpOv*swV5u**E!-LqQAuoAKzVqAP&(8neDO|{^Q953uK#)csU*1epez%W;3
zhV2<Q+bjSZJ|6$;%*zgm<?s?$&V6k1-P-Mkd2A(isa%6DDB=aVDKq>WH_cFkq&NOv
zeOFmeT-rKI{b*Le8)Vk#)i$C2C3M*2o0-A&5QrkA<hARO(yq&~s+pp3LO}TAg3cZU
z{6_KH8S|looyV$#t?L@!y`OvdrpN&14oB{_m;Y!Zcs?P1?|}Hpi_@fbhC1w<0lPSh
zIZMdctUmBBJRlJODkh7LxCW0NfuytS-z(;O(P<cJxd(oj>_r0TZ+7zA=I=XET*$p9
zqRkng(M@hKqXnjy^t$ISZMUC6VSQYkXD|){4D9^{7O#3{{i)mCZg$?CKJ~zah+toj
z<-MOfkxup_T_3ZZD1kcVXuv=1JFXLIPgpq1zgsXlHc-l$Va+(cmIVeWYM)}?Gj1CX
zbGk7Rw8*0VU8LF(VuY#I30S~7FoWv+Lqel?i^RT~5H&gP!!52sd@?;tIz?zF>SGzG
zgs!*iKyKhCZg3t+=Sl%$+8?AHO7clr9O5Was6DJ>NdQ=wNU1{TGlNCBbn^2Di2(e=
zMxgk=3Wh5uF~@^Ze)wbab&9LASy7=RIjP3%(TV8k?vu~KJ#XX%h)@?fFIpZ%uS<3f
zFZg57&YzCBS68bN+H<7@qpqRMHE)^arbMyEM!TR9dN+3==Zv8O@;L~2d7E~U{o?Ax
z>ie9nnl$q7(QjJKgAYer89iO6FdV<_B?;jx3@){}Hf(T5MAywBY}2~RBRR-`ACo{1
zfggz-9)w0*!%cwLPoXL=?J2vpH(DQE12g9UZe<|f4}AIU`Yd?%mv9_^ufHQI#`!oF
zwvHi6L8;oOwtxG<uv1%yXM=Lq=7?ru#9hzrgiRuELQRyB=cO3Q!nRbJh&CK~i<naq
zgT}zj{UtgCqk{_gLui+0{9j&1^y_;vZ2MwrqJ(PgEDqZ(2JvEXA|s)K)kji;_E_?<
z>kDOtI8I$!WGYp;D4a2u#PoD;<do>kAVauSnd8`8__Z8`M`64S7s5`<NPqkslmt6}
z`$&_ktHU2X1JfKQ-*CG5n?>gwwUhZTM=6+S+1wYbRK;nyUJM9-qQxRrv(-q?(+8z#
zC($vSyxdNX)w{8H@U(~;;El;>q?3Xt!q>yyg`#A;X(_WHv%06BltyI&nJWI?N<n2*
z8GUEbF031<&>$4(_W9aA`&{;S!dbIb@<PG!k$H?e2YL~(on<ioog1aWr#0vAnq^i7
zDb1PFT6o%|Rgieu!gA~W+M8<O;}MyJc<c&p?#Z&(aVJ<N(N)xj`!u&@>eYq35<51=
z2U_oh{{V7|K6zTICLs8C1k3)5Tf2)HtXT;gj^uX}=#kTA3C|3PhR%K;{j6B(dB;&A
zDKo646GKqI0<tVAWnK1|FH^*=PID@LY8t2i7$5f!Ez+2sl*5`zYrPYpm)u}qP1d9K
z1^|^uiQ?jYzB4=QrXsyqiCVDyyA)c4&*8vx#lcOUvq10WHHWh|k(kin-z>EoT1o+q
zw2l<=&_kA$!gvwn{uGie;T4cOdfWmG>*Khf_(tyvg!NT0cGpdR4+!%hFw93su8*$6
zkZAzf0l%~TFdts+y7(T9zVaiXSBb;X;?DTjMkTjZ3CC%`P}mI77}F{x#p|1@ES`<8
z)f@9!PLlUpaN;N9m9IM3DXx_vk_vZPO%j6FSGUckI}yI~rHNl{h>%^eN;a|+E03*!
zyZnF`Eb0|1;eb{+(4UJbC~;erki+eErFCPc`12znb!_Bk{YtKl{ZeaD^SMfM^8}mC
zfG&M<f1g2b;b(~q78;J<eI@}+MJCvXFr+$JO4_}xTCG@>Sci@qah<Pz{r>?JDmN4E
zx?N0=!;w>98TaP<Mt}N+*}ZLBEb>J!eV8bF)h7$#LgnKJWy256bmi5=0sZli3iZih
zxGXJg&*aZ;m1s)C9O#2L^US3Amvwj_se@(H^Yk<2qRAAnH@`}0x}P$g*Bd_F4zLLc
zPloi!9AW%${a9?O@?30MS@l&>KELyyDg`pRuTbzbL|=-1>LK6z9vqdcrKyNjbs-nt
z3~hWb_#e(%=@+Db2IS?ZC;JD$Mw|_PH1%rRc@747i!mEAGcy}-j^^a$=4APUQT+@*
zacw1XQ}F(k2}2#-g}vvr`5jK^Ui<9b&EkaVq|EUJk+=;Z9$?9`75!6~LKh982RDWI
zw!rcs(ixLNx7=wJ1jhLMIi=cHA()rQdCB>er=B~>UvfD*_v8M2Ply~UTIu#l_@SHd
zt{6nXXfYn7gcpbPc1-juu>in~SO2Y19egL@1J&rGp$H=O{%Ip5{Wc~;FCn=8FVfKZ
z_bK-O<Wj`gsec*Bh_9G>UjPX(Wt)G}U%p~4vKEQ&!hMPC{ze<PnyXQDKa_I9_{P`k
z<|j>J6ci_o-S36?)c6^JCDFBMon2r50q*DDXe_V#<%l(k)fHK@pQ&HxC;tO5{tv*$
z=S+7?GY1O#6LGJw0_VLmVf>a*JPD*~{S-`5cOElo;aFJ9#zc}8Qm`<vNH~-5<<-pz
zFejS8D7Y7f{8cuVW<f3yTh6Q}93Y1{`>mrl#M%0XPg|Mc+{u1F@{Lno1&C{<_u7a$
z9v|T03;#fdRQki`kk%6XAdrQpeFRrPX(AgchW4c|;m0DvqzbUZ*AyLTEI1m)6`L#~
ztdpnJbauZPVOKGzhg%^)Uki;ZI*(;^ZHQ8xJ<+;W_t8jg$KR>|#=(iOUo7HYt^tWw
z9#aUx+YyRv<<;p!?4x<5N{wRMyMe=bY$kWdmrT<ka%XV|nsmlJA(|w?kPt^CtT9}L
zIS#+`JieaAUj)>_4P}DPr_>3kha{I;8!-YP`a$*&{ct<zZrA{D>mtDShggG{p^RHO
z??~74mv>B<DcPU4i+AL|&_f4DZ;xOULpWRrAarLYWVA!~)5!bk0wto;hDvak<ft#v
z?6aU-?FA$xtD?7jGaaPszI_r&eptp(iWp1*Zktf^fSctf`&}E%zhOZ=Awc|8po-MR
zoO?9VTT7d&1-z93jsf6beeKf+f4PiT>?Nrrku)Q#zRc>y`|geAjNLBk4+%E@yX8#3
zUho<j`0Qe_`Ci=7lR>O+<i*HlLbD)A%JqYWOooOK=hq@o3a2TG4DovoilyCzbiMKQ
zH*1Y<yQj09J+uM!NrwYr?S)@x#gI>nyYcgtmWw$8cKLKwC$nrK96|v)BT!h)Y^3g<
zko*-Y#%7@!iJy!tEirB}voImTZKJY4DI*69)3mVw*bs=;)F2MBfJV8Q=xZdIQYzaR
z9Z<CApWvNPRAJ+?gwpfL$(+{BWv7XJBcV>~${jEFvvR>g^`NC2ADeik@HJ<0HRcyG
zq_8u>f31R{G{~qEl!M{0unum)@H!eoJs9%-ayWZVzh?bcrvGB;-j2y3d>Q3}Bb@Sp
z)>loZi8#W9+@TF4(7(Bckp-%soh~w_>Zlv^u$O89=mNw-FpndMfmILJ9ZdT5hD=D{
zo2;V&kb}h7+YjNB8#fU$5$I9k^H3Cm!M8(MDwND7B=x?}+ai&fV?eI;MH#qF|FZ%R
z<zSZui4d2wv;#cfQd>-s!*TTA#rGsXuO<}p$6D3lp!TChE$;o=b}%1GFb&DCxo_TM
z&g4E5!5=McS$t}Y8g<M9+J@wG*AfVy8iN6UU1gh)ez_cFeG^o?L8AH5=5&cao-6NQ
z7)dPHm=og9^}Xd%?52oHlvB9vI2tsI5**QR){li`hypmpzzzukTn?P6He;Gu>_=>j
zQ7>W`=Asy$fsQPrMTAeOovk0+>(I$h5BGuFj)Mix35G2fW4l$HlbF4ZPl)+XrAg*e
z6I2I7Elb5!+!FUf`pccxnQ59>3O&QCImUmG?%V5Voi+}}NZaU~Lzy)xLc2kU6H<8|
zOG@Gt@1)0jOUKwyT1SvR_PVeG#o2cSy4HU=Z^9z%Dg~NOoFaD}0l(e%;Z}~DfVScE
z<F8XO?|U?tgVh5zIM{FK8{TB*qqB>+9st*m&blr|{58yxm*<H6E+bp93JdW;y_7cQ
zh3lFJ(a@H^7jgaf*#W!1aN3_?U)-0^qt^->jdmgwThIifY+cn}gq(3Y?F(yJH-CMz
zhq;!EsdE~A3}p?Qj&=aIVahy^o%FE_+5y&ZKxR3q_s7RAA{>EDNP#&Zz24$XJrg$4
zEz*ar-vZ%YE3Gp%?K@u~KHjB|?ROui8g3aSD{IC>ZY22cb=uL-*C(Q*eh5CZf7gju
z)QrR+X};){68K#8)dlN^<Vo{LX_X-t1qGUSap3pte^o+QIOH3K|2?X<iQJe_Kmu_?
zeeYxEU#IwM+s2vDq8EyVu>Dp*ssr}F_?Z>ap#neur~h6qjqJQ8V6L7W@o4UJcrk}`
z^?xz<mSJr@(Yt7{LW@&ei<jb1q_~tqix)5M?yjMNTU*=<6nB>ZDFi9*n&K`Y1Puha
z`Tfti=XuVD`~7~+%$}L7J!|%?cgcIy3aaGzpZ+pAMr*XbsIujVs}q0FoU8jU&sq1O
zL=gss!e>#483{{>?s6nkk@r+ykP`-k@2<fbz0jJSm!7jW7c&ILCJsf86RkcjQi>jE
z<~rNKJ)EurbPrt;XOpmmjM}tUUr_mOCr5^gk)}DyVH@MH(#U*8odu>Rip1d}yt_C|
z<)>Ph`_MfcSReY|EeLiO(jKBMNhgsZdyi8%-=I**=_T&hGHf9LUE;l*P|MMxvX+Ai
zO~nhR?vQdggwPx(0pYiG-s+rYKsx&5#u$kIPiUyYiiSbZuh;_zVYy^VnYkw?)aKo`
zlV*KHaDg$uKG`1`6fa+mCV6aRk5_<vzcyxWb=b40i?pJmEPVB5K|ay~*FR_>Q^!Y`
znAaS4<kKtsFU*#EU;#R}1;N<>sPx!V?Cl$>kWMV;Pqtlo;U%vJr9)H~BsOaRr~-R4
z68wb5>1$%GXd+`+r;8vaO<QA5AYvkBR~I(<?#1D0lEC!x4jQVmp<+ay?yDZv_}0i6
zS;a!N<;-rb+6>>~^1xDI-!J#_8Q7_xU=2pS^0l~pQU#ms;BPyX+K-l^gJ&4Nb?u%f
zZN3E4LyZ@w5EC`NB^H2Je+^^zZtr`CFI4IQPgpmv{<=h>0MHlfSYUzfKwCTe52qoB
zAT4_D%^{s1rw0o?4K?=fdc{cVp7^zN<uow3mRJz_*Hh1}90q9DyAq{hzl3!8?ycI|
z_?S-L*n{&jqX{qY0fui)8HAo7VRhG+yCwVL_w?sHhwB#t6ArI8=>bA(gObjpGR`6X
zQQ)6$Q9(C>ju+J2Sd$O`u-X(C##c<fav-;M&>2CNt62Nh&BV7wUFOlD)&2zw-S_qH
zHZ>3L?^9NpE|n!80e5h^EI6UEUTzN*wFSkI)aqUte*`=NBnNiU+x8jtYZ>K}Y^^Wo
zdPhBvfDoETKtlCx>wjBa$77)tPT%A*3sXS(%bL&!Aunye+YHhs7KjBfw%Hc@ZeGm2
zo8@oyzZ^$g!H1U}Xmgw-zOAeWP2GAgWb26*pmnutwZ<&Zy2b)sy!^R0LFXgjDtIH-
zRq>$=1`;F6#50%q<n_O_Amnr;S8!%Rgih<adw<buR161xWe#tkza3;vP}X}(iwv~g
zFxOLW<-O9_#mI$(W!etkUpuY0&X7rYtN&~0YpfSR6boa8>tT5F=TWtQ+!D^{dpD*0
z5!1&W<6F<w+T+vwi6sKl<~w#iP)+~25Ck+S2Si$?6QbJeU8+6*{?9DpBEl0a8$A|M
zM`^Ec_5<%u8T_M&Ksy<>zgCwM8m#*QWDm@;bfT0Z5I+7<h3j+%b!Pwq_;*}h>-Mtp
zvXNA8@6{KQ@XpwtV8<E*bWXgJDC6DPGQ)NAQdffK*rkNzMgqOSQ08Lkn%RCA{{8`Y
z*lKs9kclTscI+(fDB~c{$lEo3Aa{KSv|9jS;`JD4k2t25!lc<*0CQA4``9UL`F){U
ze9F%E5zxX|^3qWB1fItEA>Kn|>Pri>GZeG`eVpQKFnw6A0@Al$b3DLDCKsUx9Z|ar
zm^3We@gyACr`s=eyOm&tBaj&-Th=6uD-RW|LDx$c#KfiElv%V1@MpyG3FI*8tD*%<
z#8z9$F@N;-*1_65-y=W<UJz%VuC?s0Q&haRE<LLKtLNaMbwxCPZ{<Qz%wl8VBwN#C
zuLPX5*tZg8M!j(xs*4_{`?FPc9h!PNLyw`LEgH|y-|ZMg8M$wPCnN2Ru*C|!MyD;t
z<EBG#2o&TN3bezzH<)yu2RKQS<86y#F!o#v(mP2+L@N%`2^yH=I_{a#NB%iIA(@bB
zf~`vzPDU}=pN+?8)~*<j$XZTBxrWQ}k3-4~gN0mvbY&JR5egRzuD{tl={XDWI;<cy
zBOSlNi+=KXe5Iug&s??H{>8qa$e+Cl8F5cxi$?5O_m-~e<UjlHU55L@8K0cLH6c++
zF*OA%l>YWCfN)siObhNiG&0}Pg1<9MC}YN!m18U&(kWQ@sxOI>w<YoleOq=rc4D>*
zIxT0Uag0_Xa^0q5VIF{_7cC%LN^H6fy(Yel@4%f2qrM}r{p<0cU9adDOp*LGWdsqI
z$BOuW4$JY7J~!f#V!U@iWn@FNkM?U|f7~rktZMKDsZS5aI|)N5^;ysNgZJ|W*opV$
zZz)_Z0gQDcvkWdDEL66g4+FVfrQPf45Ouc=Rs+X;sIkBII*#z>3*sb7D*^F-oxlD|
z+Gv7EnO$xWE>VSFKFVCKUVTv(?a-N73R#o8mBN4SM=#bnsJQt{>1bc-%c?M<hys&x
zP?>)a%ronXceN87#MfU!yq6esuG1n~2f#IGkCXiT?`I&Y92XfGl963JLhoQ!+*lhX
z_X`|ST_)(%o^K}Q$v;SxvH0zC%>NZa7hBuGC+m4r>m*yi$omBACttEP?lbPJUjGjW
zn#ksM9M*v}dJM%gl@|RpYFZtq#AtoR6r2-A?kPbjLtI>+u%f4qCXn5?2u6J!r9dNg
z`L}Tc#ooor30s7C9v2V!%}&BDW%7AZSc$doSfkzE29WMviM$$}=OFc=&Dydrg8nW^
ztlsD2JdoiWvGe$$G9>upgFg2s#vo4yDSbs9N6Ct0fwxsJsKK<t&6v3#^?XvEW_X={
zjPP7n4&x|@CByRz!`}Er&qRx<p~`|uC}u^_vxy~Uwe886$v3RKC-`~V#(*s0U7;^4
zh9QjsRc}g7#qeLhG~NLsW^E3woo?D9-Q8@Rh>rsrpIIoTnVxc1C@GYE*n6eB=`$>d
zX_xn1IXhqKYZS3c$EHR?`iHbZEW@ck?0-GOxB4%_EGW6>dUX-cl%(=h*kah8m%ol7
z{a`7O$C(nvqV+bP`Fr{+b`I_z>7I*1w*9sZ4Gls<lR19molSUJ1sz1<fLSd)#ZLxI
z&-P3{HPL6LZn*+Nzi#AX5N(hgU5ONVOdQsT4Cleemb|vxh_gqc?rWGxnCjNH(nT}#
z&KM{@MV9^lqW&%G{|8*4H0$Cv<;%x>dZbmhDK#0N`mQ!IDWq-aJ!RNa9^$x>?coyF
zp~d2#PrSzi`(oHeJKO&JND9oTGmY6UK(nNu$19NwDn7N6vyL#=f8)5`?`EQaw-NnC
zE(D(zLW0XP7GSA|5C~SG5D$WwiWaFVl40oyz19cVSfytP=CEb>+=rCws@&xClpDQ9
z!4>!<QzBlI#{I%duG?kq39`cfzH9Z%1DliYqPK{0Mk|?xbdX$*<c3i_xn?Nd<dShr
z8mNw<b0IPfhNMbSQZUBS-LvL((s59YMSdk(i+uLw9ph)KAolC4CwY>`5FD<WH_hgq
zGXbEzzdXcXIN-wn{*9XxfZF=&^{PhUAO2`3(Y39Cto`}Kn`_unKSgc~bW~(w-#Qco
zXEW-o71#g|y2ZP3Lep37P4bK@Wj+PKPArKtPhqz#zxu(X*Fl4RUB?#E=b^!cJi%6y
zkj6jtQH-VCNNg?shcZdnNYZC(XS2{2hw2uos6k`8n3;5r+FWm?AB&$$a$AV1U8D;Z
zdl|b>krJe1l<J_<l~KuBXlj-l;x(LfMuKJdsHZI?fN8J@Fw%eLE34t-5hwfu69u|w
zn5RRM;taa@UJ~vVIGF}u3o&>dwS-0Pk;ELCqpQofQHS>h#Ow8J5FIzpEUcX1JIHNm
zWQ1PcT0iwxlMaT|6q_&+Qz2>_a|^<E*23<X#7oEM?7<q6Ul*3?zrO6V9r<biCUiER
z=)T>onZ)Idikf_x(D)>(HtfiFIXoEN>TU1EOU|P~*kc{?1~<ukYrz9|?+7+QTph_F
z;<xu@V6s;^^lCX`CV;6+(tXMhfJO^|l?$+q+!emrqn^7oHR~65==gF!mu;+&TSycS
ziZ=wMEGIa67V07X1mP6GLw18PWV+y5(~dmFob5E!vO(SlwUsC-`3;4~g07F_o};aR
zIplqHR{U7t8JYnHZ*TG5IL(QkDTkilt!qqizFj8dqW^1>Nt!`&^B|r`Zm@wtCHZlo
z<Cz}DUrj2oc*u4YbP`iTA8=#jnF<_|;_nF?kAvF)E+2fqWYxQmq{`LCNQIDkD$jF9
z3;tyzqg`F2*$B7IyU~!I-^eyyJ2+k4n$QUJ^uFFo%X`)ney6Z=hn+|$TWBvW&&Hgw
z`^KXs-hF@O`C4lscO;f4{=Hoye&Qux{=LF-Qn<ANQg&7QziZ@Wq!rWc91T(x?Kf|J
z1X%b!?d_6!?Ivr>WN60Re6H)hA+Yvf{e97OP_QNF-9yrKa;;_y*5_T_P|JXbhvs*u
z4@SzBw}btk6YiPdFE_w(14#!L+LsZVbhmR1$PaITmqc$a8~tx(SM>(Z17ojyH0P<J
z&5v~bH@H{8N0ee0z)o~k*o}h=!~s@#r@|xPQxWhfZsjecvPzC`4qS8(!yzH9L_#i(
zMq-ny_(@tHMBlC5PbZzAhizIxuP60M{-;o>GyU6M1Yv1b5%{UueeVGJ_Y}{&RqKv0
zG~Z8^qB(MoZn>Dw7KDUEKq|Xx;1_}S=z60|2DFLg=%td2q$!=ueYgtz2h45R0LiU%
zE<Tv!62=|Xen0&{-{a0SCAZE^GC!<GQIk)Tzmj(LFUs!|;HW*}p0kF&Qp3q<XCMW_
zXqT(dI!_ZVFpVh?lv3+a=VU_Hr_66!D+c_HxPu%DUZ3jST)?~eN{<Nps4g1>eCJdb
zT*ABO0W;EDp-N@n#7^hiPY-&S@)eKE#EJTd^4#JH6-9(0WuF!>H)bROL`4}Vjvcb!
zFHuE%=`8n19_#Si2dYjsvm*Bi9(8`uxtJCqZ>M{n#n>vv)U9ka-k3p2sK{Pa($A!-
zk6qyr>qT#tRIW+E=U`LeQ4t4A*^ynYsI|jHaU^EYffAEpa{$ITnb>K(|C*|bc`6Ki
zx!}(`DGnxTJHM?!>TVS5%{}B%65uZc?-A0?Arer2VfI7?)r0<|ZDQgC=Q7k{GxRS@
z=Svc{#XWpw4g38t@D<-B=1)EU=f*<C-yN1nurk!?--g`!YcCgb8NUELR9Q986YN4K
ztldQMKICL}7z1u*_a=J&@~dgzT;G<?3z}6?h9@#1-@Eo$w5hDl@4nd`czxGIgQcBl
zPrFb`!ElMdJ<K=k3SnTM+DhBu{zV6SXdo0Csg3*i3RQFGINADX+h}T#F>1f`h`bh(
zp3REfE+%>eY|*U=GIj|!?;bpunP(MUz<On6zWGDBZu>kjf^Q1N10Dhur7j+X)+skS
z=0Re{cl=vNQ3M@Di9H)9Rn)|N)KQ2mao~mkfauqoz(f<Fm03c6<+HrPoXW$>EAIW@
z0(8;x;8R-hpp>3Y-VWvO{=WYfA56SFRvD$Y(StCSi4*p{1XfM0V}Q5B7-j!R+<r4$
zo7$Cr1gN#qJI~B1&oe};hm0{e-_O|2-B}GV4ed(b_`RZ^csIIrke-!PB}~#0SSXvn
zmtJxGfN{vUn+=|NIFqyrTMG*%zyBY#R1PCcd<38dgT;VF_dE1J;WL5jK}{uDLp$cL
z_fCm-?7`Rg`e-F7-%XUw9%n?(9Ot8(fYsU{aG~~n&%jFX$$;!o)TQHEqQllr^Gx?3
zRj6eKTHnzFC!Nr@-nP{zodLU6W7{6wH68(G>i?Y<=*K&s9BS7Yo8k{jLq>Q$hZFC(
zlfvV!e~K^6fM>G)=fr4HIsxyc9wVd9B#ooZEB(H>ADmN6(0g8#M3e;4`hWTS$BbD{
zHbFZCm}cL0&NmX^wbGMt&kml4q=xM4^baIJ2<WGqWQJwxIu)XBSIEWIiy5)nRKIG|
ztyNr;IIK;31wl&0t&?n12QOIXjE|o62_*FD(YPG5l>CzbO160%(r2~QA(LSc4a1kj
zCMe>mx$e3g#iZLWY~nDkwHZ^n<`a`+hI9X0?uXVj@Q1A=#Vh^W9#1^cW#4D(=0Rx=
zuIYzQ2POCB8LIuA@1(6{)KrPuO3mV{PB4zn+oefyJxHdMB}D(diQ&5sI%_d-u$(4+
zu)fmhlKy*-_C_Q@)B!c*q*?{q8sAxUEj*M6(g>;=hAqH<;|9nYr#Q@$C?v(28+Q{G
z@Z|ACJ6qOL{H-mw*&xv(JsIopH9bRmBBj+Vi-}eIcK%mK_~7Yk0p`9}DyN%VfV;qR
zw>;cj>9ry2ujwW97x#{_W!y&;eKH?(OjBQCq01iOioSeDbMECebiT9Q5aLl~9`tRE
zX*U`8>D{pr@KAgkaKXD-jp(gAd|fV_hbSiwJ0Dn$<W_rf$rzr*Gm(WkP=^DKu(H1z
z0$%5^R_afzL3J&>*}Cg)WilUP!^G!_yS&hiaT6#)CIcchCkyjI=qIf*xt$6BC{oD8
zNA%FOhp(H3Fgh+-_(LDqq3^^#(}E`{sGZT#;&(&nR^p2qHR3xPrL{)@W^0pP+g2ov
z<INeu2DxT%{c-z`ae_)h`Xiv-(%o50oc17#a=xRjR(3Qjz8cY8dw(>8vGCD4>f!fN
zVZ6wN%#Rn9PRK{VhYPT)O1vZl^>pOViPaOD{3kSwBFA<`Tejw?1$MAY^jc@FZAN9%
zZk!BaAvRopo&d7l%#Ti9fv~&`nsU)SnS?Xa+zMXkBUcP%F3Xdk9Gr08Jl-{&EdAdI
zcdz=`jylJ)PZ^9AKj^MA=wZwAIhrJnyOfbOZ~tAtzq}ezGEcY!!arr)H&{KK8!lrV
zUEuBAfY%=ZnaYFDz}Ho3Nq07VCu8&ze89pt5AI0WU0sR@OG2)X==z2!7!M!i=VK{1
zHm(vJS*o3dKl?i%$j!7WZ0hjPMfP>odI>I$HnJ%~@#r^C|9o8^RQ5e*d3A(G`0yTY
zm2^~oIhjB8ybi{Ziy!&brM4I%8gU_Gw&6O^wcZ5huB2KGC~7EH@z~8@3HEyugfF+n
z0vwiUuT+;e?6))mcoiun;W-e^Hq?Qw``jW?d#*G+g%bk6tNcP8t>uak?`o#N_l)3_
zUN>2o3@p%hd-!1Z>w_DTofsLFpZey)z}NN@zx$7mr$^-xG`7IW<V#%BR=bHS?-1te
z!B{Q5e(&1W*!i})DsbR*lz+%!I*aUfX&$)rT{mOiq0G1EsX)CdUsJ<}i*H1ivu$*@
zKjFT{iGEIvI&q}gGKm+AS>Z>Lv4_1hcgdV5w2uISGV7(E23nJ$1iFp7b<@8R7n>=~
z6Di{0YI*k!CijPe>ibHsmN)kQ?F$>pywPXsQV5XU;##wMYuIqQFFX8D&;6ylHnFMx
zHG?lZ{BG8x-t_QePWZ*@Eq91}<yz48AK-wRyOl___$7*~6Nt|7O9`$-KM{KL6OH%0
zHI>DT8rUTtywmojKDskKL~r$)`=e=e_8W~LIFQG>;N9k(<0C)_6Fo;tyCBM@|9(d>
zo7P)9eDnf$nhht^oaw+MW9x3(+Wrx6qVw+R9r`ap&=2G%%!P;)AA}A6V4ArZqGTm7
zLAKuY$cX9Pfg^$p#m#=d+j!XgDy!vYB~Wc8QAfL{b9*MQ|DPtB0Jk`J?_ZX%i%82H
z;+Ps@%Vv6+qqP$?a@AbQJa{Y<O>GxM?8`(~`2^V8EEp<OP8u!*CN~9rqfzVbhR}D;
z6+O!ZLh^CLYAO*8b;K9dYJRgBtKgWzneIn`l7D&30dZk|Lh4tzN&0<*b=W!0W;VnS
zyuw(H?j3sc5iJ%q_f=<eCUaYpoTrZfk1FEIBf5ffM+rYpPAB8f51-x-L?xf08bCFQ
zCy5p}^|e`x9~5u@l)uVSSD#imi%|@i_7YYmnq?`!RxMu5O%IYf{o7{1&*(Vh;3iGF
zPqOOwTG@2z+tEZ^O`;f-+G_9J<#c7qzU*dXZEdcw$ssAWH^%XZCf%j;Q_s(gVQ6%w
zkY0nvlkh&vH<-?JKkn0ZpRybeYJMdz69*J@BkJIR@RNw*FZLc0d}~;!L=_3|<R_L)
zN`T2>KrgO_C}Vlt0{w%~=3ce71M6E*QSsSc_dJ;Ui5X^CE(*o6bKu#D$HFvQ@Cf)$
zsBMc*j{X0$yWAH}c3I9F<)m9&6{pb8JnZ;lgRX?Voa<xz%uRt7SL|~#xAc*m>>reU
z)H;J7&!3xlxpFshKC?v1m`T>z7Vfgl{nedsi+>U}{_-$<8zcMdmtYWP%l)&egLs8$
zn^aq+w3kEEM*`AFo_6DxMl3*HMCb>v^z7$Pt-QZ(?#nX<{Is*u85=8!xS?*CV#R)%
zzB&BoJDNH=klq;Jpt2Q-Q})g6r=Tm-(p@@t9AscMkx{yku$dAADAkEhQ6{yfiUJDm
zfQ)@{y-BU82K~&W0%l{=N3>hoo5FDUWo<rUxj<^Glm1)}p?a1mS_c1!z`JjpN`-e<
z|4qb}EFO7w8##v2A{RmzD-wQfGOYLM1bZUp{5;^D+Dd~m4>VEn)uqnl6#ZRcpD|$K
z31<s38$@d*<H+lyBfZZXL=s5&T7Pt7$LS$`b1U&6BtTFezO=RKELm&2ax~(f_(@RQ
zf3R@)2>LB}u%YMZ5*M@b<9hT?o8)2~cj4t9fb#yOAf+8koy6%0d?>o+sAU}083425
z0TFv#)IYa9!R*te^2H!#<lVQni(~wQzz;xFl8WL+3sUr~TD#GKsS?{D9}eSQh@YMv
zIpqfLt;OOe;mVY-^qv>8dbrsT;K!zo3f>+aI*{tplMb&+*~#2xi`$ni5FOMI{Ozpo
z84En1ua4$XEw9X@<63#C#X%nX(L|Edjx!<JitBBOlKc~FoEMrebvNC2V`)BmWuCX%
zeYZ{iu_hG3^+bz(mj(S1HVN@X0+H5>p~XmLzYs2DaVQKn6l{7$r`cb*KGwLbJ|ZeP
z0O444GvvvBV;wocbZp+`O6GQu59ZYeCfnXLa^4+SO`Ufocl(`O*1_oSf3@+nO4m$C
zxX?-3tt1jJ7u-CzPDN>Fx@K5ox*wk^NresZNgrg_{jrSm_OxvFoK6XhT3+>ysUVdn
zU5w#M)st5E>9C{AA6sG6?jxuE``^cJx$@>8jy3t*j3oECzlDd;-F;+s5y0|9$nYM+
zH-uowqSIYCQnEEWv`O|U?7T)B@#<ljoBoFCevHL$DkScNJ56mYuAMcTBHQi}aFnJ^
zI2Hsq2EK@#1Rhq_j6=g%G~4*gq3?Z6uA~c62`jG|e)bDdSNUaZ)VnX?R4@m9>SM75
zp~{mS8=ERRu~7YKUuz|_$4-lk$@Q+jP5N~64qMl(6qM<9l;9PRe`lNgCLB$blIO0K
zCKPNJw?B;8D4$WzC5YtH*BMsih$PAxH;<8&M5@(=xFCal?gXJ@-QC?!jEh~SkV#F~
zD~V`Gw|#e6H0`_*CM714Cmr8mHx}97NOl5pjq|pzRY$qpoGMb(ym0wxz|gu$4@>WM
zrMjH*Uh=Ea)rbFU!RJ}Ds{3U51!AArgHp3PZ-Nkyw`RXBpKxhxDtp1p`{1Hq?0k~q
zO~9bLBW+Ni%a%lQ_WVPXqpLKFuKbOZL~8m^w$N8MlRae>Tnz08`=#IaI-50LFmdwo
zwi2|7T8)X$*aM%YX(mZBc^|)9b-mm}hBd!$A76sm?pf#aJG&jsxKe!+n-z1-;*GCU
z&m_#6z*%9)_P48sPL1ba>wh62OSf&^WE=b0SuIvQ<^31>)bkV2&-9F=*X`gGf}LzD
zyhW9>FP}-1$06OJif8$oy5Hsts+iUud-(FxK*VI>o!5vuO-Eu$|6u^Dx*EG>IsOZZ
zZj19&>R~yi#LQ^NH%xAolu3E=Ujw}0%L-Cpzu?nMP6WQ^zjUHi*Za!TEeZAbHK?U?
z{U>H9H_v`ewhh!p)R=1#^m?%%Jw)ERahTN>#&4e~#Jm4Tc-@E>Ov<tdHLY0kF*&=M
zr!I+Ohw`~_9A^TWK16C~3Yd8_Kj9Y4rq^`;dJe(61NVNcJV`OAW3i~sE=b`T4Q+o-
zls<h;J{hKJH5^j;K3(e0ptq=xA9}+c<Z~_2g?sKtFK}i}JG591l|`yD#c=aNINkjo
z_#tc`-#5Gd*5{UkI&cE5swwLUC^e5wgyUtQ(Jt542|we*juW}Ep}$s1R9y2(&|JL2
zygy8?1)%lRH~xm1xF?Ooa257s@`+VDs{2{%Cg2w7JraxQ1aBp9I))evJ9aaOMxjLT
zfYJ%SG-be26>GcqldcobUA+BHeo&Dr6%3@l@?9ML+Qx``5iF4qgxmY4mP_zn8({R&
zz-B9Kb#3_WwQ4dA$ZuRPcG7R!XQQFTS&d}8fiNypT*0q}TQ9tO_`@Thas7+*2EMVi
zdduD0D68Mg+bdLWjlIdO_f|aB!c9FCQ6V1B2_{xOz9{`;A7i?TH(w*8{&Ds>pz}a6
zxyOn^H@v{-<qX|<^4r8($M%XefR0zYVTs-3ggJK@`#a7bR6DU7<jSdMzH0wMI?>Fe
z9DR)G>DeNUYu|<*0nrRa?V|baq7VH@-ya+ZN#W)zZi+(Q+LrMzgI0;KNT^+u70kOJ
z(zij<ryL69x4|?if7{R1H&p41-l23%R?UWd*@bS)%H;Xbnx)1UR5oR5OND`{!q+)<
zYGmJvaPPycGYWsd`)p{C`_eW;?)R{hu)_YwC>jnOm20O<8Le^)92w@6EN1EW#*oj*
zIT~R${`9gJ-#?@W(xw%WretMN(F^-KxqUq5m<S_-<umxL?m_Q6k)M6;uo~WjHF{xs
zGgeJ%%b@9y>|t2ChhmV?`XXr9rOnmfWMn+FHrVb2wsgjZMUnA?!|Mty;M9h6y^_5S
zx43r;qlrVPWq6a?7rKrK?2nw;mW>1MZ!lT;CfVe9t7r143wZ^kwCb_ZVrK81!`Iqa
zj?^l)<4#!r^{$I~uELP0V1~5R5fKkl@a4PffA{W#)PbIoN7L`hu|v$39nvBKu|&&%
zhgch76JgK!iQ~bfC$+I4e+1?xlFK5q39^wlZKBWp7FOI{ymVDKV!z}as`AyEKeM5*
z)ul=w6bcGk@9^h6+xerE$im}Ud*t5tUQ)zsSv2gRgEv@P+z=RM=$iZ@5@H)GiTxM0
zIYQHUT*&<Dlx%z0*|uQ&ClKxC{y<$|=2K9*uSYE2TTM{l>~!V&>g}bUY7KeAN%4%1
z_K#D$q<?B%BHEK=@Hej-sdvu0<xMIv_<W~)Z(3zvujb?acQ6EnC?jJmku$B?z(yd!
zv2ZQ(>FVtgh@vp`GUR<aP%YU{DP%ak-v7d@FyG^;UpeYpeRjUVQeas2(qh@zy-!to
z$ZLElu4KQ}JlQfCbu!|XiA`zzTHA;@79G;*E=e6N@3NYX^@*35ua*i?Hh_6vOe(1P
zvh#%SrR}paJ^Fcl%fo-#S}Aq(FE!L9-YL@>^{tQVI@#lygj@+8S3bxcd}MUy=Dx8`
zjBE6E|12quRO=Qk-Neb?x>|uf47=pU`V^d=tTp&?E0wGT{}oHv!d|QmV0ST_R(kc)
zZOPwd|E-^7BkEu#^7t*y$7~~4WQ<*%3F0_n%EyfA+oF}&Z1#W)$@@m-GaG89jO=kp
zCmIDLo8f#gv@H9rZ9mWe|Bc|{Rm%GVBPvK`v6<_ao4<bbAq_u~V=X;0p1h1x>b79I
zE`ZZYomV0FuS^9#6_smD=tfMbEM#}$f!w_vOrpHW@798{-d-$he+29w@+Q?U^J<%&
zj>SF#?k`bm+sAgZA+<}tk{bucy7{`coSW;K-1<?XEdxu4XCs6Y_i-I=*~46gGpM8M
zl)vn8g5M0$sQrC<(0QG^dj<^j2tX=^l)JY*0#=;`6Wb32Rjc<$OYj{MNXi^^-Y4Ta
z8|A*$k@p#s<k{#5qwzlYlWVSn+mHN>ND<r$#Ej59efzt3qlY2Wag_<fjj6XE(Fh8a
zMc}s5`8Nc=w@S5@C3XpBx6XH{H#;w^TlM@@(6yCX5B<Qdh|_I4_lodHZ|Q!nFmN>?
zen@ip$Zl<^I{hsjgRiMrLZpih&bbgGlc2rD;q4NpQ6pY0HxVLMO+PwLwJFSj`PXkj
z0}A}S<ky!Dg%O4th2(F%->Ta^Td8>_GF(Xga^~k0g;&GwL`kn_pe6?f@+s%;-mkz8
zy-zZsgku@r#EUIFNPdq83b;GEj4}xfOy{@d*5bU$9;sJ%C*C1yHz+gZWB$CL-Cn&8
z5;B@vkCMkOvs)`vO0&B^Ku0&sS;w-H*U36WG|i8J8`r2_SEqH<Wc!XtaG7_*L~p+L
zI5sM{YAL>6DiBMJ?^8++vT&9zLj3`d0*iE_QwQJ2FmjYxKB3!BB{A6WS+#uz3;M)8
zJ8AlfgS}+o{G;l^^vP(gwy@QWt(2eEY~$Y-sp#<g+HR7K?EZEh=@4#ohGg&9)@jtH
z^Z~-zP4sXBIb@OAWWSL&etiEk1eViXs<IR#nvfZJoBw2RO=uh9x?EPe=nA@q&AzBp
zB=A(SiF#-!2g9;~td4x|=_GwtD7!LqDZZJvKyCh7wk%o3uF+F&*@my`>yW1UM{^82
zvls0Cq_uq5S8_7($)vE_{LWDpL0WLA#+aqbl2=u^!;*!^KQp7_RCM6m_Zn5dQ2xHh
zqnQ&&R?2jnt0Lz~kdymRp9s*Q<3or{8xWsz?K8+f_)ZBL3hI!NTHJHRTO>?tKlY2s
z<_;SA;cc4i=f%A|^<;4MXK}wv_pVu*v)$E{DX7JWj3t)tCCik=n&=ofwb?*DYX>cM
zjUn~Oe$?h^(`Y1J?*zGD`6*N>3oFoQlf^WR<~h4aZ15nx1x?hMd~8v5huUw?lyFs@
zAXdeMQyod3{Gfta9pBe4%8GTgi^uY`6<_(ix=0H!Ny+%8v0eMKILche%G!$U#C=3m
zf8&nlQy2iIblwc|Y}F_3lg{(q7d5bPLopOee{X%CF51giR(<uOett<yNk-T2T>2_&
zU_iGtQnjgn!v%^I?e4J!)l`=F4X8y<pn4IxqSWqOf^Rt<CfWYDb7_-eE9h@f(xk+4
z2#W2@yiDuj5b*5Ce_vV_L74rg$<a|PMdX#M=F{a=E}@r!H|2xu7(3gUk$oaF4LEgI
zj%?SvE`}gM7RB~XiUp4cO1m%ye$}%}w19sgQ_>{@I{ZYH5(5|8&)!Wf*{H-$Inyyn
zs0k`HYJaM(@1A?_?9%RW(yMv=qnfe{9OY_b#B;4T!`P=pz|>^ssTOtpo0N{+bKu3-
z>l8r}-d+>5eQ)1?KgC9dJ4Opk?Bb9$m(q=!m5GSNLX&C~=&mL3;iw1MZx_-;x1E03
z+2Reaw7Q`>XYfDA14xyAXNs?7X_LjrI$gp+5&d(Wy3@!UW9+A6FG7*9p-I>_8}K4(
zQS@wZpe42#dZrb0c+PV6^tJ!s$vWj2_~9QNO-Id-y6l2l^Gss6DSXAsIt09GJce32
zMxSQr6Aibr_4)|Y*L~WWiFVkgFe`Ghorm~OdVV0k_6~(cgi>y%CyTj|+4j*m@2Hv|
zOBItKe_X4|2xJ{8P;-ap_d2ngkGOZV7ScQwW&%aO>HNNwp?S+t7uS*z35JBmAu(8m
zdWED&*|Be)uC}~~qD}K3sE$Lh$<TKU8m2~hm-(^oI9Fa#RlPZ%o_Z1p({XFgb7ryM
zp7?BS#W858EO3zI$)cUA>?NKoD3WXoIWFPVb|9ELfTO6U+d-r7!NZxIr!Xj|c9<fu
zqgyQF3)^~*&k~Yb>1jAf^jIfJ@@)2Z+?zlk7de<;!Zox9Y1D9swIq3D?YHJu%hbnS
z;yW!0vZ<wGy9s9XAIji;1az~2V1h28vAUi{qg23S6iC%|^`^1aGH-3A25Psawv6f>
z30(14OGx-ar+ES$R`ci&7>{L2MhWEen0Z@aisgbI0op0hx_Ksqer%0|+E~4j;q&H>
zNUd)tWFzeJ6WYbYz9vL`<j;lTFllTgqY&)+&NSD#g74mGX=ViW3w|&Q(>ECI=LqZ9
znTSru-LQk4%3AOPBcblC1`~t9F0dRXlB-X*4VtM93vV;8arR^#tvYx~kA3I(T=aPh
z(tQn&gR>O_;YOUxtR&0dIA++rprW^KbLN!JXNMgwAof>u?kC2*BlLo1lo`KTR3dwM
zh5C1-dN~#o3T+nZRUGT&o~EIG2%NSZped8pUD^6<@qZmiiIb1=Fcb{;6@rH<X1%5>
zGlZfC(Gjsc1~W<B7T>4`bhd;wrwTdvIs}<h`T=aA2&?rwUQ_VXZhA9wJhZg^zHZ)k
z==}-faCw+Cvj4kul;s1<%9M4mZ08(ZM5UQ5x@2pxFIEM&t{j0a1>dnWCEYH%TsFFw
z!YHm;9Zo7s1{=MsTEaG43`-OlTIL8SH*lhte^KR@_1YPVm5c`S=9Wyx0QcwW8rlv-
zwB|a7jbl5g<bU{t*VqLvzY2;s3}l@nXsmGDC(|b9Pp^9AG6L^Yd&4S@R4XQ9>iJ=X
z2MNC)`6VOrk0~O%)KZ2KD)R)^z2Yjf2DMpA>I*J)JZ`MIR-cgI&O_U!gkz#^gDw}(
zEx+8x&e?_c1e&?qeZk9n@fNu|Vl4VF6TC*&QiB#Yl~h^L{fGl-S5aunG3qK~BYe4l
zDz&cnN*fz;-=B}_RU}@gsq<?)c%xHsc+nzMW{M;CJC5U-X;Jb%dvs5pY+AHUVWiv5
zX<F5*S%bB3V+=7`yip3aK9;euxVn;6s-55bgRuNNo`0>BJsNjXuz`Ot-+Py2H(8P@
zkhx91ZG)-sulf~1L*@>a?0S?C2SgMqbe%5dHxu|8B%K3SU470eDznr&UulXUT3Rts
zU$IFgFYs0a*(nOymOK}l<fK!fN(K4*>55w8_pqKTQ#)k-nz5Mjp8j6WT_q@3uJDpV
zT7>6I9<4BDe&tFO_JdTQuVLAWEau#piSM;72m7?!hrqsN6^x~&2@XWKG#RsbD)P(H
zT5srx+ajJUQMP|8$lRi3v`e|~4agF`;A|h&bcro0`3~%r_@%F}m)XX_g-ve5o!3?k
zOD6AENtWLnqb(+i3OzXG@(Ps_R8&{g+gs6Z^ozJFsS;t;veNIk=`~P0xu`r=CV8eX
za`D9csp@!2GlN*w46?I;oD^5*!;YLoxdf>Z!83AUARv2y+KI}>;8Fsw9r1-!s`m6l
z1RJ9>qwLO?-Pz-s?l&&*4blDOM?loPc{nJr6{=I;veXx%)ah3<C^HtYLIw<6w2iI_
z;BYY;?-%7?@r3R@s1ojmg_JA${!JEgDQ?|lFM)M)xcW|cFY<F~Nsn-xWMg~r3f{#`
ziJ;fI%*@pERAAmW|4fMj*iTeSP`R37eTZ*7?kTU}AjD6iBvc_*_R)yC{42spRSbjr
zZrX`9eawmW!Pv^vg@{Unpm&Uo2^SC6_dV*S;`!DSw2{OIjsX+ZJOaKN2ezV=PFq$a
z4w%T?$BODFlEX}<?Od^i+yiPH1&>V!=@+X|l3mV?C#F<2m-W9$CS)&d{oJ4kve5oe
zjb-P;y?XPrf~K&ohW~W#=N!GLFhj<tLaO2_1=g0Pw$&N?hm%^|Tcxzm{6X3pb+yK`
zhF=O&n7QLv@{_>KD(J5X)?ZtJ_b=Y>p<NjRmg!}cTyjYL0;M-xBGi`@oL1$5p3&Ho
zR~mMueJlBQ9B6JMS{VjYbNrTb_ge55O8D`@JsV5}W!3b4IXGikG;7U%Iq6?Zz!l$q
zrq~rj+Ls=;x)YcC1F&1R1P03T$sm`faTqh&L^_-Tu24y*I~9kmpb{ZA*#!Fd8at7?
ze#FZ~RK+FKVXmFysUom6zf=OYJMhI1b#^nK!c2?X^76xbf`Yf=qkLKVlJ5@=Fw;yN
z54L~TYtz4t+pFy@8-DX5EM%4I0*jbi=lLet*)&vxGIF%<A2jakX}S#(Uhgl{>w2<g
zmJi_da#L(U83ZlvjVzBe_&{B^a8QIxc5zlG2aU{)hPEq>E+q`vonj^b!Bmk#JA9RB
zt2CQEaM>L#^G4XWHkf%w&5_h!D^O_<K{>}?`vLwplV61#^9cr9-#b`_H&OS$EZ+M-
zzxqr+wz;36z_x8OHA~EwR6UNoVM|@j){>Dzo+D)F`qi7N@3I0)B+E|LZWR<sxpLu{
z+l^46WWs{~F2yEEeH;`2vfs3VLS63OOe9-MDm7n?FcdP>*N-Qjv?#kr4%$jm*=CcY
z3-R-Ov?Xn%EKL@5Bvr%O9WjLc2UQ)}RzUwZRQ1sBH=O+x#G%0XflO^FzjIfC;3L7}
zCYnQY<ZiN;mT5!C3~m1dF(WI^X=xNTtT*Xr##5$|AB=!M7tacGY<@F@Ny(KVe)|oi
zbxHY1>Uo)j(M1tRWuWVz7(4yXVr66SDInHp2s0CDm#D_E((+>w#sqcr1~*W~>1qu(
z0Ik+@Ea_|s^G-^J)zDy&b)p^OTiD7o)7DMq#mtuWbft08{kLIy3j~Wm-+p5FOZKn-
z3$a?X&VK~NTRz-jBwce5py8E73$1c9Q^l9MCLwRLd}o}@DbSEr2X@))(gALMIC9oO
zh|W?q=Z6Gg7DzyrurHcX&FWtyd4mblw>B^tB{XsY4G`mQWYMrK`c0cAPWsKeJOYj*
z^5zJHdA?e0<uE@iDLH~vk8|o#EqEO6Za>@-grEo#7qv<gP~pC$kM+aVQLGD7$Hv#|
zc(0*Id$=I}!!d~_T7v|EVJhwqAK6{dqD<*g;L?voBUU5y-a1M|HC**sXvFH{l3{6N
zorXDY=xw4N3;)OuMXs#$#fr$UdxBfX?Lt?xzq^*up9}VhI|Wyq!KUEA)+<@a?-<do
zMv%F6{U;VEoy|Vq5%~zV=m?J_l~XQ(u=`8cv>n`~EA{t>bkbNu7MCFxH2U+^y2`JT
zV#Y@us<N4ia$OR&!8}fXn)_Nqs3hz>?P;W3D!J^{w}~-w-RK9cO6Tzup?{3i&C*N-
zDKF6;&JyAGoK9mhT3i-5QY@*NTvE}sAu-4~1~E4NzjQq;pP9EStiHW7vvanDkSu4U
zfJo688_0Yxw5k0Hj)vHdkXSa!p}8^3pqhS%%x3?T1AS-*ek|Si1X2I0ur0)?ZZJ@-
znYzUdPBNCvMqPEnK4#Xk7X9-|<G5)>Onu4$`t*kpFlc2y=DVW5N+g%gSk>S7RWEs=
z-58P)XP>>umvPGRg68O*V6&)KFw7%8IZR&Jp9${cPxpCMt;%;uD!dns6}bErVS5uB
zpO6x~UJ-$p^SgBHkhWz<_&MetMBv1mSB3`t+Bbwwky8lD17vY)Fwpak1bsVq1n)06
z)<{wm`pvA#Ho`={Of;Y`oi3*?rf^XM7?hWxd&O;S303K}@=6W&N+s#on-_(KQaJ7S
zjFo!Oo`dR&Y!oqkgs{d%V37xUMKc9OV-)eLewI|jtuaEyA`O#gh0#e_xMLA9&3hl1
zaWD%8#f{PUuyeFxpzhY(#;fL!e|%qNW+%pZIjRy0p|d{ZRi(DG7R`B+&wXR-u$us;
zF;aWGWoutiXauNBh9g7?^dRiIx^}$Rdqi+F8zKT0_3C%X4>K2Hcg6n57hRhvbj5d{
zsH>llyVCMW>?E!X;=+MNko*YH*Kt6f6Kp01Jh03Tw_tRPgebl&e1yHjhQx@czTcYk
zlTsRM;1ae1zVpcEYSE~4^HJGOHB){G?dPw@V<*-RBPLVtwM0^>Dg#C2bmV+e-@L}g
zM&x?F*=wfE>Y>6>B;VBy`FvI;HEm|T_Z1|ZEWGYE@$j5Kv<CUr4ta?N+8RJ~Rynwk
z%0XKf{Wztq;Gz3Hw=}En7wPrEO5|l=-yVy+sYN$1e0PjMMaZ_LW(h$y=8wj#5@yht
zRmvhUyj<+tU-rLf%!)0o@p<5)N!X;l{QLQ&Ap-VCKtBbdtm{DGL#lTAsl39zy0XaY
zFGaKqn7_H!zhNH-kPUGrYm|mO6nsqgv*)l&Bhm=$94-@OV-sEdPRwzW1|-gwMQU`z
zKtN(NY;}dQ;i2S8S3w4fSs&}bbld|t#=$rz?w*|1qD4DO_Q@O@u@hjv<JH&&Crlbt
z37s^Q&c2VEr?QYYwu3Kt*k{pIw%f9E7lJM7@J&>632e98c}&+Uo>28}Ao!`*BjEF@
zw42ZZ1Bat+D1B;=I!VDCDXpV`3(P^FEH>bUMPAxByWB{vTmOj#pZfduFKve^Q+F{|
zm1hTUq|3Y}Tt+lKkT{L#H1N2?P3Y2Ev}+r*`X~6<tBP~uGTDYF878W_?BgdnFBu}%
z`LRXT@mrg%vF!@k4t2Gd^+?kv{87MYa#wAO>wVgSk79kR`b@y$ZvSiZ=looCJ)#rn
zqwx{Id1XnlW8qC+RxS8;QuA(7Ze8ZsJGOt~eU?|9w!$d8_A~m9BD+Q*z6Bi218w{*
zOgFvDxAC2*1F)3oOf<^lXYIr|<m_pL{i(}KeJ|lF{)GH5|JhGhi<;5;6N!@Ut;T|_
zS2Uu@r@MzsacoCQGa(w0yG>(_iXrOh|AAI*tlb3-OGSca2F6B3nWO?3=Nlkx7br>^
z=KWzwwc?=l-?>$|xqe*8UyCNAg8LFYYf%e#Zkp&gu0(-pUZxxZbB}uZ_5)$oz6c@~
z2wKF$cf#4DtPHjI6kqbFd31zkRR5ki8_dY~fE3kT(_dg>6!>TVn%g=pns8ARiShF1
z??qoF8UsVvkqLnzrqF~A^p`L0ipJmu>DY*pwaJ&Yug#qHb22TmFX%2ESwx*Y)mpyF
z8LNYe!9kUN{RDy;Re$14)C*c%?-B~yZCrNde4wIkKKin2X(KLZtjfqT_Mb&dHV?;H
zLq<FZOFeDgSaFOEgXvZ<v8_fzJ=W7!S;vPXTB~1%WQ_egT;F#&Q-`pBImpS?3it{X
znNNN6JLgo9SLu&tMp}HAu-=1%`q&NUIEyRAkY<zt@rK>XsjPb~Fa|AZ1F^wia#$e~
zvEQfN2F>h-h0aVd!X-&Nt1cMHap=J-UHz4U%%DXW2&&|FV#KDB$ZSK3ueNjP0=+TT
z^D(NCEIDX8PFZv@?{>W^{|*T-+Du%Mq4Q{zbLW1;q8D$p`@xy~I+Gj?RiWBEbc+Yy
zZ>I^AADU%qbv*O@`u@OXeby=5qXxsAd~`5iZHqNRo0a$ZE?@i0V;0Mlv)4JdtDXWR
zn-g~%DcP(ZVU(`sfR>tDC_@{x5`=hammN&_V~MU~L_+#UC)+S3@jnFGRQ8it;lr!u
zko@xr+gR1wJ(+#uyUI-(qQ&*1s}^+JYCGpMzW8^eF$~dJI{MA*qgCAQ!_1j&_PDM#
z9<9dS{X1v4k|7u}pa-O0Y%8E{6?#|e5X<6CU>8$pF!_pwdXRKZIhOihDVgHwx9LqT
zNCcS@8*T$Ji5jWK42_NhA~B}Kf+jvinPd0fYp&582Mutu>@IUZnK3(;c|90t{#-Q5
z9!Axis5EHRQRRQX=C_h2i8IDH8W-Aq)>#7@H3zniEL&PO8l8Z~b^QX<GQ!^Gi{ksu
zpo#HZ^|@N;1eQEmMtS`H@|?3at2e1NXynVd9BHB}kLkux#xOP9C0hzxCr+_2(+e;l
zTT8bS-h--N<Zc2EhXIKjsC1&rs4z?SukHi_yDNhI!N#;3X1-VDZ>FQsg;zK3UlL%V
zdaQZ-#7q7tYi*QyINcGu!*<C^@>7mXjvyr&+_>*CJOIFFSft=9Ti$)*IO~Iz<+%a+
zPSjz8CFe^yIN9njknLL&uoxrSm&X<y?BZE)?6tQo*vHIz6Jn~oos<^3uLdTtRYb>G
zDdsBvTV5qJl$U6oBqU$iA`H0o(HvsnK}}LW(0mBHrbheb-)|{i-Ijv}7|U;1i@`xy
zOb{;3)h*48cO;}6*Ms%_|GNcLj*+yHQVMtZbT_fjKawP4W}oaIFRdG<a^Sxj0+9yF
zc!9fG@&z6N%q0XgWdV6-Wx5K!-;mcOCOT)zs@><iOp*Wf0Kv2|_S7^3EMtade2$1r
zk%y&c%5e`DPh|h6IeW=XyS}tCD?XTLMZA-<+vHi83IezqK6pBAaNRP?@Ipo;Ojg4_
zi~9XI)3X4Jho?8J=-RqLSd6uaUTa+~?m#&p2D*R+(B6$f`JlfHJ+^A~-Cs4{hvS`U
zFp73{41N$*TCWZ%gZIc?BvPIHetuuc1vKc-BlNHPRg!X2-k~__xI!ML@-JyMEckTT
zL|T?IB#7FMzCdc_;Su1@aWKyyh_cH6DaPKBqyeX&Z`gonj<-rbTuqF|sdS13UN52t
zKf2Om{LwB4KF!M-egsT45gcR6daan=EaeMWSVq^6wL6T4?dkrLh>$;ENU2*+Qg@%&
zs$y8WA5L=T8yG(YZp4*s#RX|qwhwlKMq+1@RC3yO&J^k8yupJGA^ctJX-5D2w=oCl
zgZv|dgyxT^aPrsk(|NoHZ&(CGZkUsh?!+d|cL(0BhF7n{a7P%4CRWSaYYAvTwTl*&
z=sq$R!kA1ao7A>tKn2$C;pTsufn*OE$atp}9slpA=YtHd_dic_V)i)6(tG`D_fEZ3
z@nG00Vxk%u1XK2lsd|Kpcr9MMwlg65;eY|$kfjH1jC}20^P>0KW}#nvAgrxr6Zo<u
znxgOj+svGQL!Mr2yUoLwLE(VpmN%4Vp|hi6PAcrerC%e{ZpW7BN1%Hb_8`;FD&z6P
z_E$63T8DajpKiru)Kr7@_tR0hN>fcfo3PjndlO&PBS7MFoBN5C;4jH8hRQT@SI;pU
z^Z=$|Dt1ki`jfIj1*M<oiq7{R5(~b*JHM8`!TN2d8p@n=C;S<;_@s711A15xeuRgR
zl0;|@(S9d%B-jT+xZG;nTimgtlcY2klYIV3B!X?3eu|r47vwV`t$AVU&7QxN`$84m
z7#YKcW!ZedaCupzEjJ2khoDjVBYs{RRI;(LaQxp+`#+o9J%8u$31{4vx(9vA$`-Ni
z-H&N_=S|c`PA%NEc2xZ2_6^P68aaBZPA2tN`9CF0uk!`&gj4Tfb)zAc;v4VIaZM82
zxI2KQ<vT~a|CfG5N<0FJAmO)+)K1w^+GZD;LyY|Yqs_w)SBosQr}P3vI?vab@H>pa
z{8Fge!vCW^#@8(i0ZgZ)X*^OOcN{T|y4%xH&h5nX`+WkD)yA~H-VXy&gU(|8M0*k9
zTgoGoPT3LJXY;iNr||t>63VLA+BYaP;m3RT)d_1caA|VA8?|T84TV4P#l~R8_XEbC
zgfEXzBY{)D!Eoz-D8~!cll~ILfdX~Ny-etRK<ReUaOmzHkkBIRa@k$+uHMV@_H<88
zXD3%Vzaf9<#ZN+qi&I~oooixL22r=ED}nV(tIM5*3HE}g5BbbVDCs>lrk~2p?z_^m
zse?COX}c14pDqz_>OVEd%2IU*%U_{^)zL(({sG-VmJ7WZ8-^T%7xha^@8zar%9Yi}
zZ14x{w^GDcU5=0cad{>RC<$V-#J`AX-R^AsOk#>FRT+M@XuYee*k+jKPq^9K;pEVB
zJ{3eO=D9(Hb%^*U>Xs5MegR*a*INM!Yz^kmOq5sf>eT)dX?5r6lcfGz-+k56Y7Pd0
zL^HhLb$^?*nMmt`&n6a5>_g*mzwolj-+X!FL^-j)<GK;*)wP)PDnU6aNln}Pb=$HT
z&e>fB(+x(pu!PRh9Hn1bywbY+2j#iv>)wTz!I)96*bPJ;0c2Qs-4=)O+HGuzQj(8@
zN)5Z^$W#Wc5X%=Dxw%?H8Y*KcrK-$jbG7N3D)btqzhg#+OZWbE=zTS>*zX}?=6=tP
z*Qdl-vB;yxHEye8Q<3%E=3`aKw39H+QXtt+y~1xD@>0JPW$bC)j21MKNKcBN`4x`7
zoMnrsd)u?gKQE4l2MGU%D?l=ejVq6V4$kyZq|aqre7g^4m8Z6HDi+Vxo-4+RaZ*)}
z#@{Nd5K)yv%H=e8J+qYLx!`0@BUfWe!PrcB2>8LL#z985i>t+{rdn+G7qr^<jz#eQ
zC8mh$UyV=9lF*y>G)1_|Pm){xnh}m?%Taw^uBngv9-`3iVw4)&PkU@7j$PX&!bQ0B
z+h=zD8xjk`|Aucp%y9c1%T>&O<Oh7+RwE?e+SDNoA$^-aWE`qq{L;<z)lbx6OD6jn
zh1*}Idv_M;@w>q(MB#&3P!M+9+1XUdH}-$As9Gfw51i<Z!70mWxrhU;-p=k<jw^5U
zKh7TaEBrsay=71wQTy$SySo$IWpH<Q7~I_g1h)Xe-C<yG9~j&%xVuXr0m1;m9TFgf
z_we3(>tDC-`F!rzuIdk6dw19FXRY;H#=uRzS$TRNV6~I(6-ZLRn&1ADB*U+^5{%LN
zKx-+3#diTJdwqlcz#}&dG`6VVKn7Xg`6BDe<SNVK%Ky+e;$K2tok(AypXIVD$nLYB
zi+ge5B8Zh3s)}K1i+I(;C8T%Q_i>CDghGGaNDhFD9e1QYDQ04;S6M9k_^yYn9sWS?
z`YB~?M&HDcB^;RigQ+9|1s@jX=qO5e*e<+_F7N%krAu8yKW3@N5bSfxq<QKA^i^BM
zAk{ORjRus5w|JQ!;%4Eo<(unfpHXam0!Ufj-|Ew03xwja6$V&Fpr2OFF4>p4sVT<@
zFj^k}cUcr+S7>6+r-s5WGA*Fxr_2gsWW4s_vVJR8{iZc#)#=0D8Lkfvdf7S)0QmO?
zY<I@}p8D_AdN|--Tv;Bk6GD<aJe2<~%7Dp!B_E%BRQFOz8d_of&<ZD-?Rg7>JWqCt
z&@1aF$$N`1hGxY{#p`@rG_=DpOL<?(hsl*Ius)Zo1rS4QppwAXz!;>OKh=Z-oaAQ#
zvX}cXBm}iSB^>X?ARyJ#e@nL$<#$dSOZ++dTzZfHnPSY6$2ZJ(d1WA7!`|*ayWR`X
zae1jmjl-1llN;CZEd+cpR#aRUeK8Rxi`A(5lWf1g0G^U^;YPvfz0cI8jv>6e!L#LY
zHva#M*p;+k32Ny?Ip;a60KQwwF8#0=PT+qZZpG`|ec}yt?m)dAExG+NOz_-KULuYl
zK;=zapcmzzy8ig|5h6o!R28L^o+uMYdQKwY*Jx0q@*lWVDJ1MKBO=^a3bP+4dY&FS
zHw=qkpZ{?xrW|)*2gP}qVm}}66yV#!!|#sijVT4}sw6AwlFD=TA%Fc7-2Dzy{P~VT
zxfC($4mbw;hi905;v5R_*wp^iQuO!Fuf$=uhS49Q_j6>3EbEFpFe<@=gO>PeXVL2=
zB;VV$G1ItD)-MtPzgz72v&P#lr&otvEdpfrLtcXiDTyi0CPBQWxpZl(Zm)qT9J^sw
z4p4vUsJN>54xFBEG#M|!-;#7Lq!x{zZm<-y={_~b`^7Dd8KD{eq<J^aQ`alzPQO)#
zN=dwX*}R%U(`tjH<i<m^b6cCT&aZ3(_T=ggqjEg~M9;jNTPK5d_$u-^^pcjy@s(6K
z2$F9Lzv&}<V$EYJmS1(PgJ$nY;<83|L_;O=h{TR!54Ns7%)jnGz&iSs;5)9XX+F;G
zCQ{5NzHBP8FJEy%$BqPy@ol*w)yv~q+5cp<Y5A4iK}DM6qg=X5X?(Va$ipLJ83*J}
zML97<<#g%`&-?bPi(ic2xWZ^du%7%nztZ`=B7Mh{+PXM8Y9_U~Z72CFv~FQ$hMnYk
zyt}zx{M~=tqnBwW!#-6ra@K!AIOL~0K}@M;GMHtEbTN$DO57yWA|R#W<JGsTS>GgB
zX#OO#!=Cfk<o9u(V5TVlSe<I3Xe3)3%kX`fcm|oUOd!2w`R+^CUF{3P|9oTrXG7bB
z^es~|e2FUwi?8K@bKTalS*`Qb|NgmdQoP+GRJ)PA{*$!3f--2EG+0>T@6RN-KC$?;
z<Q_yU6ReNsQNbm3^h5W5C%=aK2jYJW_dYh4mfgHE4d^R9IT&qdr3^gyu{4Gpn8K!o
z`!JH%<FXfML0jZW1Hbj~$rbZ?Rp~)8MHlEX{Jr`U<{hq*S-!y&t`q|<O$6~yBE8sN
z<J-@xOLzLI^A{_+K|8qv;OxNACc)af(+GZX;5T-tROiAYRiub-BlQc3=NHb$qXW8+
zeNPDg{`vhz+Lip-a3q@c#Y$?g()d_@FAi6N7p)N=)j#x2_^Azj9}h-}OZ!PKiKA&x
zwu|E>1Q8w;&iATmHjy*@>3TkvhSBBm;8)TI><9QRT3W4!oHQ%`2agHos+!&4L?0;w
zX<w!7|0kNjpTn9`vN$^UrCLuYgDie4;qLarsGQUGKX5pbukxpEZ#Jg+_WSLeQhkY~
zShIoR&o!u%fv8qVPd_%OtjvoR%=yj@NFzfTlg_rUxLx#6D3ol@XwH~rCMX7BJ#{Ov
zU^fc`2(g|F^!pR5B{#1ckQJXvA~z?U7>f*uz*9VhS@OBJQ+?mP;-}`nj?T;SEEZU3
z=1%a;SE0bYKOuP=1%f5cVPwV^gvgUb(p`&$jY)sq-R11@BZMf+9m?VFuD<Fdt#Og2
z{09zYV~hLLkhQ!UqS`ht_5Cl^Nm-=GMZm?@(DpEa^-lX1caNL$kr3EgIzI&az4>gC
zOOSRW_GUj<;;}7+NN)FwNAdaha!Jg!qy2(<TPkY!nbCBdU^Df15J|i<72o%aCX7x=
zY0D&HPhQ6gafeChy>-C{VSKpFKse-QhY}haNa@m;70Qs*t&fJ-VI8%0_zR)Ze)>A{
zW}xVxdhi8;D9Rhv$m0TziBi&Y3M~!|x3-kvh7e+|LF7L8d{hRf{&DFTxbBcfagbKB
zH<aHV)p*r~H~rnpJg%g!BlR68Y;b$a98<eKnQI^}cIrOtB<$6Fd)<9rC!AXSs3h`F
z$x<xHQ?uA+(@Fg0p6V(5(^uitDq&k1a73!Pg-}pzMbE-zARMYrk_GHrp&!rsTZtW&
zM4`{{tKM**+{;_!@W<5N>rKz!0=r_xwSdtFT1}#hm@Y@LtshY=?&~%S?<z?1p%3&7
zbg*YPl&Weya23=uvA<sGwO`aGy7tIs{<le3v%r~d8^>dn<C^LZ_C3k_W%iaeTP2D2
z`G#gsoH1gd*u&z|q*yVuR|_i$|D`nCE1+k$zESEjc=-JZ4R#V6CMOHdwGs0jn&t<E
zEmq@B+V^Yrx{rZt4Ib%_t^@F~qSC%MIP(qGl;KBCZo*X8_wFI#f#&?HB8MI7OR%@F
zV$Q}(i^Xu9>r~@_gb-~h5*i~+l(NtON6KjTQRd-<J*pd~2^_OV20V85I%)2vgvnL9
z)0E>sM}2Zb2-Kb3*e|0}kXZI#fCW(L@D21>1EkINHFWfw)G+=b7gf=hgx%^W-M$|B
z8YzaJOMXi4A3_5%l7Du&Fk#2$6UAhOVO8c|uUpESvKO{?Up_5;8eIQuCT<Z~-eITq
zPXHo<s;i<f@YkWl-<tBS%f0JteW;EA;0D*EG08I4TQM+C9QncW56&#AQ?;gYOX|=4
zniCfdXUeJEa^W1Q;@Yb|ij?jf9CGX@KW}%^ad0XNXOPWQzv~QVxNx(>#6SO?ViT8h
zR}n0&ruA3ekO!y=+95s8(Ha7SyFAzDcFOQirQN@E4n`3nmhMN80ACBdImB*M^JtXe
z4f1f0S)snyofnE!3?tLP+}kiBhKcT>_)qU7Dx!DCU$BJRBm@6(iJt*<xsNhpH4$PK
zvA6N{>KwPK@pkQzgR<p|uwKv$#Au1{pL*VV4?|2;-a<80C%)M`_)in~SF<_7{<@R<
zV2w!X#V9xD`#!25g+eza3Qo(5(($Q<uRM9I;f_vLUXfD+A0KqzaZGOmMfw+UXF>MI
zo`2w6s!<S>`afT`8>IGmQ9U*x<lxZ#IBtAC-z#|E{<Q$|MEe8TCPkrN)Zxed|9x4K
z1N~@DtzoncK1~}!5BGLO+In>gzdFb^<LovPh2%YBWrr7Tjt3<XUa!2bAEmC6{k~8I
z;o>q0;n1x+iS=B~c7>%-!-;jJ8lS^_<`ezne2RZJ`bh$PS1*gn?>!`D>}g%4uPt^Z
zKmVP0^-bGL|H~atHG6;B6to^6K?6tGbT9LsH!_?P>u@9MIfO-(LW85^V;PC7l9jo*
zJ}Jg=*b6Vlrlh>fsj$PNzc-3Hct=f`4yFG@Prc?Saz)3G0V2{(t)*?AA<ADIx--%&
z9Mds>v3pAngF~1&g&#hGEjNykZK-gAgcba~PQx#BU%$UI3I^|Y_r2zV9|yujgk@ze
zwmOT<9)}gSvD!8Vm$nt5Uf-6bMkq%?544}<%&xV|mWk;vd$O)1t~r83`*++>yS$R<
zS0o-EZkvo6YB@i}&0ntiHC-L|eag|yyqJ$?p!z=OSZm#rNP63jd!;ith0;^J$iK*~
z(m}UqSks<$pPp3dgvE@&vBc=W%)%g!KzNV88^eM8bc=w6Ci7sttT;)7=2_Nze@okz
z3fFgoi2p?@4N9ZP(o&a(lPQ&c_LqT0_hYaAJc4Jcl#wW7kVK9yOw+(<jr~ZL5xpDV
zf=QBx9a1$-bUSR;%99Df*3PPY^DvU+XFnx9m##D_t47#it<KKVOw6W;4OLWO!J&Ca
z5YF(*(Xz#BI~>tzRgg@MS1qTet}H>44wGk&9FAu-60qnr1~fi0aJmJsJ1UM+b2(aY
zWG7*-DJA&|EcY<!4R}jr;yW4VaEQ5uc*&ox7?&nU<bX5S8*3+zjb<E(OmZtIIk7Qn
z8J9&(op%EdYlaTrRJLW3?*7JyWew!L#d7M|lu0I9W)`yA?H1{ehdBFXrqel`W^895
z80y=DxsBC0=`vSMzZkI18Q-xd^t@R6aoX;~_mrEj;Q(peOav5c700Cw7UCz8t5p^V
z(+vEAOY)Y9Puo>PFd6KV4$Eq4dY%DcG}C7fm&$qGc~i=KAOnN$Dz!`&ClIm0mba+$
z&4(El>pr&ANFJwP)nK5QVlYWC;N~WJtY4+l7<8&R=6cq|jOIF`$ioFto7QEEI@YM*
zQeTPjT}UP^uGO6$ALmiUuRdmbc2SMY-mpGAA=YEe^2Xrfj@?Z52-OvqyEUX~2~AB_
zUvhJ>Icwxm&Gd_nu8!AfPFrF*qqp-T9dFiQbiw8?P%9PzeGm?k@&xNp0Z$E}zRSAX
zzK6}~PH9v9(x=y_qjiB=6~=<qMbtM*ivZje0zIv!O7(;2VegoSB7QwtH92MS>i3>9
zTb5&^b+kt}CG&#VMs^8pqG^(vv7ea|54PZyLraH;IGFHRUN9}M3qrL~vh~v0nXVBC
zZ7ih2NCB81=%U*^sP_x<%MgF6e7fc&f}Sfl&Wa^roegcmDIsg0Ni<UbE;xwsAf|OD
zl1Ck2wVGkr%)cZZF=x+O$WRZSAMa<bv_DOF)Kyg#vz3$suEy61?id;{;iv1xEURR-
zu4ubKGFuP5eJ~6#T$QJ%GRI01;F1<hJ9MS$gfT@-b57-<;M~%)mtd8A4Fj>FsHbdL
ziIOfoRR8at5&~?P+ek5UVOoYcy~N+Pv^%EWV~z2acMP2TMUbri2fE7F(-fRSiXQ#9
z>KysXYZO;wa7g27(u{-l`rS~W63>t7*=}MASz3L9I|4y!Zm4Eq2DyBQMi%qxm={(I
zpRcwc5hhOGm8E`CQZSwC@jCuPWP_(k%1nh7G!nJq;<w4#yo!t(Gn+X^oY6E}rrVZS
z1N8oMF;kd<WW2#3fq_PMtFpX@jhxE#6M#2SBd3a~Y>^uOlNIzoaHpyhBwQ^{E3N6{
zkG`?}mEER%Vctb53Xq2)a&)vJZm5A<^R@s`%rio*xP~e`UyVdI^pWxTa7g?EKi=;B
zbY9IWk_Aht@%icf=|%0g+?sx>!1zis;t>N3eRW7W(?^9KEIYr1?DGkT^D{ADz9f6S
zS{rimS(vH9c4@a2Spq^)t}shFr}jG+pT^NX>3dk5oBqQ@qcby;?mxzEIfH&HXNiQd
zU`m}|v${mF^28{Lne1#b^7At2kpe;F-_N||Y7`zh+Y2L$S+mqm<&}?S-*_}BSXKJe
z8ZotNlT}v4j#DJm4$d+9slC{Ktla!A-lq$3h~3gLC1i+K_;>hEj4v^{HUZgwCgwJX
zOS{VOkx3;C0+bKN?}G*Q+ShNmC3QXdpR(8Wvz7o^RwPY7o<nQai}RK_O?5i<VaDYY
z!LoKj;4=Fp`k3UAVB;_>W<e2pD^gjJ9EKQHdAx18VCT~Iay{`F0!H)GM_Q}5xbnq!
z|3Tv3o1AETztw5la2EK_$+vM!Dn?9desdaN!V!om`hqvujs1~71^Khq5#xcet!iJn
zPP#gy2J?rgWuy!_7jE7IxikVSIi3ZhwZ4a9CLKwfz-}#r11k=~3TA30=uabl0X!B+
zP3>9=P=8gS)sDXdK(*ou!Cxo(5#7$h5CmVMK|B<V&oDDPpt!jQ^{D8M7O&c2mE?8s
z9YGElkFf>R=--7irsAS!=&iF?fwiz5Yf~3$K}uDMis%jv+L90l2+MNHkwtoJx`Em-
zxPUjB;AUzdjQ*@-U2=a9M-Z19TzX17@*8j6C3ju<(3sgImZHiK^)&jq(1mPl?J1Q>
z%)C5@MJfFdrX-IS^4YLXKV}s9I@vQDcy|X>uuL6XWlNlye*_rFSC&yhwe|XJSG!bo
zCe9TiW%)c6Ih*Ll<|QB#uBwcB4^`&banuf4`Z_H0YR4E(@3*BL!|>E|MnLTT3Q2XL
z1W!>fl{i`PVA=6@bH!DR&ymp$S}fJX5TqHYw5c@>xhnkHpuGHLU2Eq3<>|-(1}e-f
z-<o22=N(LBE(u*8i4u*4d=)kW<!FoAli*X1i2;@uFEL}!6!UW#Zr;b(BPF(^ntthL
zkxFQWlIrT51qo2Vmo_E-fg;qi{D~SLK$JDZK~&wMA@oKd5Ei(z>9j)+o0C&`fo$)Y
z-4g4jwMfS9#`SRm8P0XlF=iS^pYsho6e$9aVTa2*%8c3@dDB@|5s8c%XN)7wC&}R6
zz%xTtzPS7x-xcHTq@00j<*G%Z9fq~f)N7x~X0n-->qN}6=|Xbq=@V9>J@f@SWYTLD
zfxAHn#WT6UQ5CkV#4qTu7(SIkh|B{>%L4y|FvG!;zmWcQ{hwy5cGXIk-i6uHK=RqN
zvS(wa7f@@{uL<9eOW>?Zg@usq(F6epREl}L^09TxXyNBfP1@R|Cm}Z_p&E!pF<Xyt
zA`iq4#1Bgq!+tSs`3ae9WC*4g)1Avonp0X%P;{v!X<>;cgg%3i6d0F*lL=XIdW_Qw
zck$8xsC<od*m@fY1f{<4DssD?^`G7o(tmTPv><sYv>T~h)XdddN&oFl9z-9c93M5B
z@HF%^lEKb4US1wD;>{}W_h*8ty|T}FF)Fq1srTnaFM~A!r+|2+c=hLMsdii2iU)67
z+jwgrCRvECTGndxrK(n?C7-1>d5BG+qAC<un@!SMRLYowAV)24jm^l8aCNPaFp|v#
z>G{X%GK-T+f8I;6M{fT2l@FO?hBSa=H3jYDdP<JUOc2?=kS4s8iB3FovUV0OE6Y@p
zhk{1Cc0GaeSQ4AU13~-RWKlX<I$^F{K~Z`J{Xd5PeJt{Sf;(X8>DxSj^V78%?k&+D
z{2m(mh*tK1_Q_2NMwa$)=12b7Ai*s6EfdG<%P>j1sZ_Ej{!l0y?%G;|f_p&i%d7m-
zV{cbSZ1Yecm9LL1M3wi;dwsX`KE<Dt9xHrs>&S0{!~~oc;dn7+xytNi<}zlRwW?tZ
z>uI&usFek>5Hlv;qFB#T-?l~jU4eLQI&rK{sx`E>MfK75vMHm9*n9*lZ1K5zJDKKr
z+a^psUpX^^dvJD{9yqW#z*VVh&Okwf!c3BM0m3LTqv=I-x^}T#GNPB|@C8|xH)Kw;
z$`Rvrbsr7ip{VNNszmNhWdK=VVgjr7k<-A4{d3|B9Hc^pt%gtirGUc#e<gkm>&IN6
z-b&_q+bDwz8<CsIw^f2+25fbrhSNWe4z{XaX{hE<%N?fSBO1$kBWZ_A#zD3h<F})y
zqx`uM7S1Por*st}msN*8*Be;U>YM8b17G{OWHSV*#R0nZF2NYBO$2MF`VLyZ)|D}x
z5lX9K(8lwiUsrSzlBIY>0)ZF`$#b2iH06ON6XhS!I4xp}epTRIVdI)&JkII*1X-!j
zAf?uoY^-P|bBa?KRi~6_ja1zcThIwqgAL>h0teW|^eoW}Zdgc`oC?GA#JEju*&#>)
z^m55M2RAUWgDCvPtb<5WCtGElQ&<;-9`HFvf?@bWQo;|V<uh`8y$2bMTz!XS3g1&j
z93-O;^O~7>TAIDSu2u{roUBUBe0+oi$cy9AgsXgShBRv-=Sez*ma>1x0ao6{wx2N{
zLnWGw^xQObD8Sj(V+8EA;gz!MFJbiza?BdDb1CYJ8xVUn^i%yjiG3Tk>FA!wNVbrA
z6TU3fIC?CC=dRKOTNIusqgn17dtzKG{Il-u1iL-N?rd{1poFyT0j^(SSx%2I1w9wM
z{)!qC)dMtyHRE^=E%6Wwrpbm;`9UsN;<hztRBlT|9N@@Y*8Ml~vB6p16)j9_VNyZB
zX>ky7sA*8Ir65|2Q;?PYjrk;0Fi9Dg06o8iu?>pudm9N**CA*zV%6W8i^xmvyCBAx
zrsWZ+QqPo+AX<9L(ee*z)N`~{b7f@3?7{gag*g1J(D>l2kYg1;uBiZPlgtKWU<fkq
zoLq6SOUxHFbs70^jH}B=jV_x_kgFfU3#y0ipMkfP8s?{1X6LcW4aoSB)O_8xW-{~u
zGng!p_w&fiT!MTQAsoL1?@Mk~h|ER@t+MyScprJ&{J7zn*1<b9i36ke5=qsL1?$Xo
zmS_&`$Yq+zQ8o{nOvC{AmBx%wd4vOt=&hyNSnKm0{&=r^cUz_$N7YGrcRNB!0U!$D
zJC1-Ys@;K{Y!|-Kyn32$LsKvCbb4dhRO^&@n)lUR?SBM`2Bg>Jk3?%iXUoAuIeA=x
z50y(-rtYF}3zi<HqHZgZ<n}GmA~(BPDP%M#k}{hT&{}&fdEz0Frbdu$kuS(saVcAG
z!8$+3R?Uuj*kq+yU4_MdldnArC;B`%41X?J%}Fb9fQRkZu261bp&)CR-*IuaAR#(w
zr11x0#<NkX;e=bKQtpYSOna?@rs3)1sZ;~ZrR;vC>W@HBVkkks634>n{_%LnDZ7P^
zQkYw;PO@+XJi<_4ZLo*CfG@8_u<)=luYL7vI+nIDD#E35iP-b{bM%KNYGn^erN$l<
z!oq-;Kd2lQwZ*g$uoDEd{L$Qd<KXO77B;6r^+5egdsPsMDkE@hTgRVzO&tIrsoIng
zoC~!SX(2JP=(Z}p)_nChsSf=!8(~fb_iHKgsd>Nd&oXhryFx*K5`{01|9^pDDz1JY
z1;nKRH7MnR?N;gqmPt^foGB;>@kRj$X>N(7m*EP5IF=jY`52C2AlOu2<4N9FOz2oY
zUR`3c?8>|2s$3nWRnJUV#ivS~FF9)eaB&g2;<>??|7~qM*q^br-?T<M`Y_p<Pp2$X
zo;wNoK$Va@{yn`}l4KRNhcSS837ImNMi+uDcja>ZX(qK?vHX()`48OId|B%CHm1d8
zsPUzdzEDwq^Eb{MXGW8EHzRD?HJ9%@=eEXlmPSuzw6xTdU;9`24pNIFw{7uSOvt)J
z|M9o_l@Bf7-Q@fdE%ImiEJ}wxip1<ksixbkpEKw9;&bR?sZVTK);8V@tYrUR%IW`1
z^=;Xut6{HU{I*kq!~NQ%sJX>Sgd{nru`v2EW18UCa0x^WKw~n;EJ5Pk%}x^erh+$R
z_#tAc%SXR!oJ)o^9OVI?8boH_gUr|IpWn`_j%l@F3#ApoI)oyjO3QujwwTz{0$3W@
zOWR$<V>|psT3~$exFH!0y_1~+L5Dq5ePwn0el_Yw9m3DN>XlqB3)($o(3rrDfcP@^
zxEMGJ;7?+G04sg|-m_##gYzIkT7T1mt=Y_=<nm%w=~IalTfAY?l-$<lCswKkG#Ad(
zPeAW*>6_-%zocT*)kb1_|B2b+GavDd*QC>&BbpEcB(af=aH5gBZy!y!aFU@)2terU
zsW0G<^oD>Ww+*hn*DI4rN7RBRm33r&!v@|OW5V~_Y&1EBwSMHtz*WMd#P=+;*3yi~
zVTNW>4yzwh`#vBuI?~y83LKO@x*DcDuUTm`mfN6n#SgKj_90dtNi)Y~sWa*)4e-KG
z(V(^wmB~}LRVHj#i-Q{Klxi{qJLt<GtEiBeHbNi8o5ey_-jYZg&OG#IwrR4NglQMw
zi@qQAWA(#@tT!c|eAFd0GRH&M?5guO7@Hr|_LAIWbX?t4-)n|q%%F12vGdtPte9sn
zIi*aAY%NgsL`y7~=U9=8*hh3JFYc;ba86=DNc)JJKD;11(y-XW+SeArwTHj--@o~d
z=a753=JOfZr1tA=pOi2ZMA4TORAbzMsAfEZ?H!f0G}VllcyDXHJ;I0z=cYq?ek6q0
zp{p>6G9HdqRbUZQ4u3mT3M-`%wUGf+Ed37a2Yox{u^vp_Gc+~HGtBSoUZq+?r49T3
zXy}YChyr@CD;NyQPv++JP=1{4C!BZm+NWvWJF`w0-K2~4SP=i4Q#Oo`_2-j2{oK!A
zblYPl|6jI~4udtmEhl2JJ1^Bhe9GtRK><s2-%Nv<(qG(c26TR;oC1v$%rd(a-&FJ&
z_-<R{JL|McR1%)FUM)z~j@iyA5Hqi)p2rGmotKVn6%$=`9=dk9o8rV#-JXcQgvfWe
zOFUBs1blBYlZ<mzy!4*Ua?P;Yri~SL$h7H&m}uS$AMboOYyaIwRj1FZk$99VrW30e
z`|f3_G3BY-wR~4cT>R1OdL3ulCdGFC3f8h#O$tqhEf(Nj;?0p52e4TDMz>bv1S{5z
z5>AS#Mm474Tr^UO;_8@5W^|hK&$-Y3JpZo3qI|<G5E9Zh@As7;Fh|2CIfAC%IHRx-
zk1Yp}Agg7{3KTL{%8*S-DdL19=fT)VNT##PNv2z6QJkHKqZllJS|N?S<vh4T8#_2^
zxE97D6!q~p(5`7V|9Bfa@7wct6WgqJw?X9909F$0w723bL)X@acFnJjbEKt3^sB_N
zTg+^k3JFV7xdPh`!AjpR=c8;mbKUoYOxz6%m<H}hLr&^rO-dZ`O)d$?ZrSa1<i*G|
zqLB2)w!Gl+QxWkI(7KM;8y;FaGa~*0OtWgM=z6eEw_H>XxV`mXW#GN>r>PE_aWKmJ
zZrr*rUJFU>jBRDY1OTYJ57LQ)NCWI41OxNKx6-!PMW9Y)swp?a0&}p_k3a1Mi@%C|
zta3yyNfj^I@g_v-Nv3aA8D*Vg^_PhdA3d<^y8X*%xB{?G9lK`X$kWx+{8ya-HrS9^
z0a@p^@pQr}8TvO?EZK4i2))GLUFv`GJxfQBQqkAuz`()fu-~^gReTknLy7&P5yP&;
zqiO=sEMIVf1#jM+#4>(2w|?2E9h=QaK2vaEld_AQ`yaoNf32{4{N?qCoysI`E)*At
z9W8Qt|M}~~nleu6)=QV#9oshTI;!>U;)7KG#$Kh}_jM!AMMd}e_Rrg*@izwpFUM8+
z41e~*G${>X5`2g7zdLt||2j;^U%mt!G`jjvyx{clZKm`%xwXH`?~7r(7(#bTpVowy
ze4qRNUWtE8OQZRFo&x;X)E)YLWcb83FaTB{zu0p{vcI)9_GP63(#uL3`|o%EXz2BC
z;rH^mgI!CJ=f`oOuTdKRe!?#rvNKF%PhoxiN0jtKS!S;)+)dOx5V_}?geyKmispLs
zvKUPBxk<Mf`EuR!a#k~_Q#h=uHP~84RIh^mTLUL<8rHS-BA0xfY%^Ny?ZwA#Nx+dB
zu0-LwvfCHUP-VjNQ9lri;fCr6Y+Klb>7+d0_zdX~mg!N1r$6mO;ACH2szIQ@ptF#x
zuGUZDA(4F_n|aSYH*$|!g+Qg^<3$^jAO5&esjq__+_uz!L$`<OPqCkPAA?|%rr7!6
zmQ>F6Kvcr~2tKzG$X$zw)qdx5-~-ZS5Uct2L_>ZNo*^OkP&0Lz0395OAYAZQ@03IQ
z9uLorHpgk_lkmZZuFWPY|Ad$sPX(bg{R=;!jBJ!N%cu4_Qm$_$bL}5MtM(RO9W+~|
zuWGOA)s{$<=s8l&lbAA26oQ@|lGko?e==*LI^taU-lVs$J3g(%vC11xI~~iA;e>?a
zM;7Oext2RAV>2HC5$&KQ)WA$@J+wd4lc^^FY@Q01yQsv0o&sj;5>o9ofXxvAQSlGu
zq$x)g3YmT*s<2gbAj@m#T&0Q1f+DFB;a|}=0_q=M&$GNHEqHaEk2xecIjX(aLv}05
z)u0;tqOOs8<>XMcO8iyENxYw9tR*=*n)Ab4X-HdJc~0sfg`?Mt6W}NWgKsSra$YZh
z4~aK8!>#*uUkAt_Q5E|;5+K%ppDP4U$^-9`*O3e~xY^TC%whBjZwkBiITDMx_r@%W
zHw@?z3MP{dj`XU%P8{qHhphMv>Osa@Ntd1CkPlWxlwgPkRYWc^FQ+Wh0MBmql@XuV
zn`{Y=O)reJ+i8%)Kc;}5z-vEB81I{1s#b0D09D%(5BQhx*}>D@_s^^cRDGfCP2YME
zcb<>;Hmz4v-w&d1WWV?}X@WX^{ltR<%%lzC_KI_tAb(i=$NN$~&mvot9WQG8o5u7%
zx5(3C^@s9_?!yY~7MI^)iHT&3C5ZcRjHq$or(dL-T60|h8!P9))X!~uCEM3s746(^
zZ=wu!?tbTsC1-xebAG$`=X)o0fA;*?+9lqTVt-thseb=Euykpc_#2mIXr*#Hmp1g-
z=44GK(0fzeS?WsC#GKDo^qPIp%8%0G%YpaN0W-s&d7-!xi@5l-?Z=NH|CmI6ReuZ%
zIx77OuKVkB{oO~mV=nmO_tpHIDZcQ@f8dNDI8TCj3JR!zL5cRg5|lZe->+vDH(_@4
zy$JuhzIZlE{5BUnUZXCE>Z(g`fcqnPV_>k^DE8^R%cARMKz@)T$v37BJa_BJ{ft$J
zC#cw+p&b@VK=~gyiNmi?Wa{w_A3fj)<U>bo++w}VQ5bS*$c;u)wDdm>*8U?%+RKrk
zY&u=-()RD9319xZskg??ook0ltoFhnsm7{BlaBP5MQ)s3pgXlJ7Ffww!craouq|h(
z=U8ggdlsgPV81S@cp5?|7a^3(5`EVaQYlPdeFC=j_MuwK5zl#VGjx6Mch#pwsqQ_L
zC>hlu1(U7Q7Y&aXCa!aa`9}K|uKUG=HK1cj6Xg7Kc_5)VrC@UFiu95)pz16vXPnYz
z(qr;Wf-vrsn6LFV_Y0#Dj3UTfIW1b`l37MFV0-U49uYC;6^DGJi(doF3$=KU;fb7P
zOmT0leQH9iz$@Es4K-u8l=9>RM{qxSveq5C5K>v&4AbGb3djd9C0hXn5{1<ISI3gG
zXD2}91nnx)y3<*T`1-0<d28tW^hWMzW!|i0`ci={EP!Vvo$!T-Y3`+CY?n7^MdTw{
zR8%Bf(z>lrrXYZhqKezYvP`{JHHpZj3SMK499A)c2o+&gL>bRoqD?^X?3<8twDeGO
zMZ6vc=cM2Wu<UyMQ4Al<jdu$P0g1xUJbwgB2NFLhplsjbH2%evu8PAB41*>fJxNd8
z?fuyQl8#0FD4`t+?@Cb)Aia--%wjA*vV^icrSjGhF;}jqrn;FdNSP<q#H-TBF>9!*
zrD|-nZ}5f8&4{`hx>m11ndp0v>LF+RYXb7mK=h+B{UC1^!sTR@o+Q$L(#9*N5o3nr
zH%&{XSz}csB@E^wO%?{oSotB*KQ<g)^~Ayq1vNP|^4ILeyd&t<=d|?Ml{X#ZG80F;
z;)}ibS-tZaLE`ks(zT1VWbKuVMN|(}4*DPCHjK-{k)FD01rmued9RwA5?J(*69d_&
z*n;SzTTF&CX%FQ3k)raSL{U7-gd~%Yr3I16nQsB4zq+c<zLbY~A;#L_#G4DPx0-Tk
zfiU<I9}@+)gB#@8pn|gcYB^GhD<vrgGDE|EK+sI2)*Tr)6${oPR)73vO%}Ckj8es7
z-=qiCM|&TyQnnWcFBvr}sp?AABWg?Dd_z20@4#|~2_@stR#o<8WYNan@FLBc>8x0U
z=?C+&u?}$QGn{0*YI@@D;M41~By5Zq+h*b@OHd)v2UiQNQm}UoS+tB{cEq-aJ=9Gr
zg<nwadHdo=gphZTfJOo{0OClXCV$G-HHWFX;c{@#s_0M^3uHR9^lC5~1jhCx#Suv%
z2D5u<CjF4xh=R3(9FPmgV(DF6R_H|&J}-@n=~QP3L-{@`%2zK}{fMJVFXz)7LFXp~
z3Zk2h;vf<5SV%!R5d20W7+=@`Td}5m9*zkor*eAvuAPvWNhgpL6E$<iqzaQeE#bQO
zl|;F51?JqeR{B|ff`A1Nm7O}vyjJ;Fd<8^i#g(@rnqtO1bsnYHd_v4XqcJhH+{)T2
zP9s=wt3jqfSui?A<-Of0k|v8%2{XtQ4P&)~9pY`hHOe-nx~ikaWUDE!Dvc5X6EjaL
z3I^UFo8LM;ic;gOH^x-xgYf>M8wXaOA&($)nn`$6=2iJAWzkoyVd_Pqqj}BYhOtmE
z5UeWc#oPh;^%Cr&#>t-f%7Lu)iv2AakNN_*WYlEm_3V1R*cRA;^+Z!yO6(a^8*D;4
zu@tS^T7Z1KX|y9)m{pul+EBhIhgU34Gah3l98x4!#GEb0$#loh(?UvXTY+GpTycG-
zgmtRu!I4-{_ehAh*7>z*ISW^je2pze9;%9ND-hJyLet(FT|`l&DF)?<Jrr?JbDlUd
z%1dA(@OMsnG+QDw<1|>>TR!Jty&7Zf4x#4@;ajjj{CNTMyBOH@4i<k6^ADhGAwAkW
zvrX6QFe0Otd>9CJd1hnJ;0%c@EQ;$oVq!8Rmc~;(O{fXeeDV@H+9w+%-ebX^{9TBh
zyA!+!Pip=`9<23_X#v($OmrfMsh^C0QmL7M%&ukYc1S3x)=&ZSva==?@}vyVwE~l3
z-I6VHm{*m3hkZ@>1qwuS1D6j(SHN^U)OS2CH<~MpE5|E>;S1TE8kNlpF;&oPVwO7i
za6t+HRa=z)8O9l7V&zmLx_N$sVT-1IMXy$Fq}KUFB^yHf*+%6&2}=?q3;i?{gK=t*
z4a`upLjsb=Sbx|g^Q|BexjL<cuC7}UD&Q$M8s?V>xk07X7F2=qcg6}%TG{!DCh`Q%
zx2~7+UTU=&d7z@ie2s+}*+`s>;Fu-1D(3tbaB(4$FZ#688Y4p@F<aT^mi>NFc8&TG
zY7q{)!Zc3ABO7(?Mfx>QCcA+Xd2phIC=OZD%ChVVk5^d5`G4S+9!%ant^l+Dz)dLw
zNY6~-2h;y(#xT@dP*^a9(t#~10Ldz>8xo0G>^(?5`q=@cwWo?&wWnf-dW9O(Y8T_-
zY#2+i(h#Okv!VbR7crf%lC31pFG2@nR=qEvxDXXzt5D6C#02XVnO`XEtg$5M$t>0B
zt-}&fP|Zgs6ZR&;I)$_Xb#2XgnriiRK?AuoaCRK;N46UPN0qL5K~`oMIqClIhz2pH
zGhC$(i};R;z~Tuf&3>T5MmaRFI&(d-LcI*Kn_3nRG)bt&Or=%QfVix4duw`x`POCX
zH&hax-WSy{o=%oh%FsUyVw~9m2Wy5V6GYuE+J`g6@px5<2%0NK699D5((xf&(8K+C
zS>KRKcrc3;H#ia>PNJS=_z#lYZ~A}Aq3Wk@EuUnIQAaGkOG+Z*L4evRo#k`$Fi5w=
z{h~1(>hy>|(bt04Y%*(ldIWmKeO#GJAS<o@m(_W)d_zlL7)CDDt(tu`Lq7+bkOMAt
zmG;!kyr#&Za#htY=13ye+nqT46W)^GcNA_fP6BCCH*eDuZVg$^`1bW#a>A`kyo43=
zCe!#sPHf&#ZPwAfI%$gb;fDTH4_=g{CWh~u`MU>a*;JFj)4HHaZ%l`dR7G9+6{4fk
zFY}V=gN7cW`7i-AKp^l>;exo_n)HdWmSz<&0)bE@)4PT&HhiBkl-3pIJKZG|4#qR<
z4Wz^5_F>;|GqYMl5j6bi-;ye%f$}RuktbD3FfZN2L_d>?eYL*2C6}=smul9gRJKZk
zAc*P_{@4q;D*80+_{eFq#$ebb&urI+#Njv-@mI_$A6%HdlH%!}tc#!gsrc#jds0jz
zgNesrb&e-{iVIDLC2`~_lc$6YXIV8^C|*r{UiI0V(7+zV&A@B8mcS=(mnRSejSvl)
zJ6xj4BGb%p9|%{#!f05zeUx46b@iun>re%ao1$1x;7L8`Jz-L5aS<+8+lb?ElBFqt
zVr?G+5U8BBla_|4ld#VvkStlPH{u_RtUan>Sr|LgftF2U9Semu4)ZQz4QQVX?bWgq
zdV)+=o?Gef+WC4I*qwE1^D@$bt7_4FZPQhBUOn}Mis(oL<;N^jgC-?<+p(kcYgu^R
zb)(u!9dbh>T*_$0IgnbpRl;?-P7n+$Q~xUD&F`UAH$(==t`3h&A(L-@q-7uxXeQ+A
zCnBTuNtmcbp3_|M<@EoA?DfCTC;NV!+T;GK@cWXz!%uTon{s<T-9~#;Y<yyF&CgMp
ze(#V9_%cjuV^;o3^%)@>jb_QbCU^C^yPv8dmxg3vzmBew_vtPmK44t6>E028_XEm1
z#6R%o>|(u0Rr<%lTI5r4@R%rl$Xg@u3^^)}SCzFVGm}4AHICmGe)MtZ1*AqHpL~Ia
z&mkJkcN8=o!ouQ!m6AMTqtQd}?2aNWj}4EI4YOFZ>2^FX)JMdKRGlRt8(;!=c$#-4
zyQaj58R#_k#|2YIhDh6v;yqU5ca&+XvL$Dm%4)HJq#JyLkW`?J>mHCMN4>;TB0*I(
zI)mt6KDt%uwZWzS7e9p6)%QwbSpWV!Iysd4Q3RE7wd{dn=4UE!yMENZ2|v~e2{F=o
zkv?=)b<%Q_Nbg91k7b1KR~y<vsxb)8+J^=P8L%x*%2A}df>pHjp&)mms^y|O9D3}7
z34wGXjlQ~RyY{U$(Vod{K)K`s!z+)Tu`?Ozp{q}0aLh{7lZFQT1h^PiL+V>2`^lbG
zEdec18c5WxD2~=dVBgMDsnH@3oWq(M1+*us@=G01jfTQkM%Ptev#B4d+ZSuisw*W3
zp?EW<SMFmO^QEGyl%fbvDY*JLk^KxJC8LwxRiGuydm@Y#!_rcYd|FFFFt8%r!OG^#
zW8(Ou6vHu+9~li)iv;RreY3376GX}6o?=u>OJffSZv72N%5osyq8Q*2i5;}mW>Meu
zb6~B`5~s_`PLh(vRGfYt5+e4;EgK!*`JVU-i+4<&`V8h>V;5`S3dcL`jly}JZO-Jo
zAksvnMBPHJ>sq3W4fr7S0^?{99wDL#xHWC&NO06(bC;>VV>uP1LpA^x1B#4R?KSeY
z3n9)Cqd@dQ0I7OP&j>v)KJd8&%at0M4jqZ32=A0nmYJd`xrxw#<CzrPa8ualBOwx_
zI>ZG1TJv^Iw*^)qq#IzlOpMB2FqmqpG~4NW5fYt+CDNU#nwa1x))oTggVwf?3LoTF
zpyZ<Ca<MYa1LS6JhGC<e<OPSh1XE<`nAZdV?}_dpF*!@JST_}<wK2HyZUk4)x}T)w
zJ3{SmF{|u6%NJg8YX$;pwq87`$*g$R9)@`tENHm4qhMo@ItTu#GkQ0{C#6b9ffIVQ
zx@n77m_j&PxM(E46f#LT{($qGE#s<i6+q(0W@377DL1UAr)**{KP5pB)>%x(yC}oq
zq&29yBUASPLN_34Z#1gd;@5^Xt=?hWRTlOujkxio@iQ@^E;+?p7tyFy0LL@3aY+UO
z7`2h6z`bD_4RWO*{-`G#E+Bg7mqbo%Ks5!%scl&t)lHT8K?DRG#NDl*_Y$eBFI5-p
znyvugu^`dqBzuJ?7d_^yMsDSMqLucN6_R`QAN%p1e>cYdg>f+DrZF==JCwwFpHU{i
zRf<T|Y>{_Y$M4A#8^T&II0iGhn{S^Bjx=idh6f>|V?DVTTD!C1h}(3bq6ms(V#m8V
zDK8o4N-%e{P*-5okrM=f?!V+bJ?67^A3lgQa$D4nk{Dtxt@BPG=6sN4jD?DMiaBVh
z7p!;Z5?874q);Re88IFDv$-l1fAI8KY4wWLfV06zj9oba52+bmC~vxgRW))!dbPk6
zkZ7nDbb7`_{lrgRRKj?~awyC;A9`kBs7PBvJM^IlO#z>oPL!o#_1Q&k!K+LZNEPE{
z&C8(^ZcDkWAh5F(J`_iehl?21`l3Iz9k|i+Ph~iwea&Q5^pyza?K|S{$f@#PFOMHA
z%sMoYpO=KE#HgA8UXJAqOp6~R;kfR5_xV}0lA(2v<MQnD9A6rI{7TD9mfRK3m}(*#
zlG~R}lWW5SuLq#77%dZI<AFXhSI*YlDvD|^s^bvdKMH<h1~3&v2{M&p`PmM3vJkV-
z?AyG@;rJcnl!ia>Lt=e=M7d37v00CrA4auN3YPKM)p2L&XCekbb!SE&>BLCRVATC@
zkU9pE>Eby2aIfh;!?K};e?b8#p>G>>g77s5mdu3(CvD>3m~T){YFFxxkyqn549*J!
zM_+lFAMop2gZ+u(;;RGS`Rd2@W6L(M6brr|SZ>f`|H1AL&1)eXAD!%v&%t&tz?b>z
zg?YN!R52(?mP9#hGkXglHpSb|aGB%;%O{`350UX7&z>dCq$aqTHhF!PKf9YK1qO*Z
z+oq27o2!z7rX{@ZW4%!(_zY;Q-|5t`KZjFsXQ{a7(TdP;`cbC|M@e_l%fMcHXM^KS
zLtLO&N%zJnw}okUP=&jjnew~1w`s=pJHtQzLO7=P_dw%wmz(rgX3cEtEcS~Gk&qP@
ziTi%EINy?z5V7sQqi&5~Cw2mN(k*=4jQaj|>Vu_xG_#oJ9g48-@AqR_QrT|hpOuae
z%K{MBe*fPxUQxz-<t#AJRjkIgX3TI4AM@$<K;s+a2KnoCpL4Z!5Wp$KOAp73<`>&w
z=_3E0{$0qo#<$YytmBgx<9EpsSK$3Z34dgeR9IS7!o5h)Kak>Zf5-8)?NYpuU2I<m
zuvtPj9`n$1p_m$!@Uz<_$^7#lV+F4C7Jz9$l+GW2|K{@L0r9V<c*kM^+8P<I_R4q1
zKCgz<sFw4>ohsava34+WAbno7JkzL(oru|F1ss#xPq{QA=c5v30jd+6dUDaD#HQ*W
zmXv^sBN2d5b62%cZQ?e5_5(OE&-$?ust0uEBSw3@1y#8O77B??V;Eij*ZJ?s!-u^}
zL+{ly%u%G=W3Vxi+u8@iurkd#*t%7W8|TlL!UI~OC;m}G8Ym7!OGENYPA()i#^nfR
z>)@iM2bs-O<1u0%H<7GqP35dLh3S$-Ym_qTzoU~ww#moFp!z+jCpPiBKCg@j@`<FQ
zrfby7bSYmr8OM9+jt%SNXd=#sZ(m=tsq+UP#i=<@(V;V8Fw1n`_Ij5cu@9B(=;Mk+
zO6B`7WlVKA&1J;rExyw%Q>C*q=W0>ed_^v<cvmeSq+NmJ%Ak|<qR~DDQB<Fx?|f!g
zm^d-+akmcr%2=9+B1FhFA2o6Z0t-hLxS=Ql0MQlHNYPg|TivrFO=u=5L7>aH#49Y+
zIEtg(8QgY<ej&9XSAC2IO{Pf9664Ti*ICzT@1f@PbwVjCB+s8-%u>(rMZ6oKM4eST
z-%l6)d0r5u7R!~;NRDtlt7ZMXWm)&ZF1RLCi%YK@?L|6n8_~!Cpckrjs>I0VRojJq
zQmDwtwr}cP$P6OdVr{Nq8^^yTf3{6-L$5wknzn;JVm%P6iDW#n@S}I1DBtpW|3npt
zDcVkca3g=>0{V83YSI!G8E+ECiu^^5yCB=h4_M**k~DK*!P_9L<BEfy>fG~4H(Eql
zv*ztll?`#thgU18SJ$hWU|=6jYzsES5xA7hFdWhxD&}IS(PM+Q4P_vhkOIg#(aYY&
z?gf;}r<}St+g3DAx~V}^{z5?-G0?e+>FC$6P^06b3u+TYV(teDqsgr`fVn}$GBsJ7
zjAs)LO<q_MN8|SI?w0q@4ecN<ag8r0%|PderqvvEj@0Xeo`jMx%QutyQc~N>)@BiU
z(%IE3+u>w{^}9cs5ilAFFWXNkxS=Q2b|n?3wq)B48g;+FFi|3oCkzO%HB0nM{rYbk
z`BjC@`S;bYlc`unf&9--J3yPOL8@<&bdw}Q|EN`Z>8d7M;^|Q-0&Bnd?v7WJ9afmf
z#5X6EQ^tP(Z`m7>7%62y3u&Yz<Kvb#Dk)KIRBuKOQ9Gk<&=V)>K0G+s#ni$;gXJ(<
zqGWjF)XmgT8GdnAqH!jc2ZU;c&imzB*l6T)?A{Igzo!Oa=i})?k2=3|hm5u%<GBI_
z672FPJD+K=CBQ9cBeb{gQrR1U<|MaQgSMS4RBE|hs7KtwZtK=ImV>F>pxEB<onI#7
zWZZPFAtNBoaX0BZmHTdM7_t2_-;I0R+4^8t&ki!MS~C&bwp3<s)2FC$n4yPY{VL&#
zwxd3|ni&Kj%k$vyFUk0$Fp36h^$E45?he02zrOb^^#-);VTKWhNI&_wKQRPab@nH3
zeXe__u#7cS_g=JKvvv6ulLO4?+O)RL-?7|Y(PK=?QocGiL{lk7lR_mw6_37LBHxCp
zZ|4cA8WjyJyG6c7dO}8-$(O;=ZB#;2bZLIyGLSwL(pgZC{esWcM;6EsiR?+yPc}Gb
zMMtO{M$q41Rmr+XMww_-0)?OE^|^d2-#1smnN^>8cIs=E!rhuIqP@}{B|W&dk&?dT
z-n1a}Cs7`m{c0~HwA}Hhan0(lfzP;~db7#(XUJI|^QLp+MK7aqRYGnhA*_9y%ZeGF
z#dEZuERi<*a&H9lyWNG6JA`tZ<IV3VNY;PrLP&gIg$u>a(B*doQ8&_G1t6vhFS@7t
zuK2>khkHkoU}27$qAY#EnYCY0!`Wd}YlXEslTVJ5;g2O@gv(V7=%%hjb02(ZeJ9x3
z-&vrnwnd)E=OMt`chsG4)=#4=_4EB-qizQCw=nlc^c*wyCfCN>|G+8fvB=TKk}pkn
zs*)D3%(rRm^1;4VSs2RQm^NOPr_!NY&NoXV{cieQ$8{gPn^Tk(YeLAR{`L6F_lzN9
zu!cA6nS{(yUM2dZfe?@@HHO&qS-o;EGd(jiK~ltlL`%=okS-#8wqIJ}0gYQO*u(Es
zHAvJ5^2^feqzzYnW>z+s<GLRQi3N!Y_eSleL-teG_lA{vL-9_F#!abJ<eh*~XA(>#
zl7yCd5_KY>S<TjU?94g(XG1TSj%_D&ogwDMJemimmtBoEb`7YmW`V?PE>fGASNKA{
zb+nwLCg0Co&bgGX3SRX{LVWS&5mv#{pI$U`eJ+=uh$2y_%z0@pBqUgSIo#r;8gCZc
znmACID{Y?5?I*HJS%#8tD8vH)d}k#^4E=lEyDPaRgvP={y5?Zmgo7B#JmqsuQfOK%
zWl$$fwL>txXm%#`W*G}u?(5g~TOL*k_t|f9N*!>s+k)-(x3V@{>0VR+8%(LQN&8M$
zWO2F(W-8iv%lOGK8J<Lu_J|6G1RW_a6PfG=Sp`0R=2|R18%=Mbz|bG?b64+_C3d@(
zw#*&Ka|jL1KkQU^XrkGpAyy;_DC+a`q%`Z&fU$N4`FX$)KN`Ou3&YBm&`Om?$ermz
z|K2TznWVf|DG?*Lq7J-1>Qi%l{jemA+-)o$*nfiAnjdg<IO|#b)~g5NHe~5`7?pgv
zS>uEUO>%3z1O``XrxG&=AID;$)M@;-JxO_-ZC5}v+iMOw!%n0E*wlpFj-NdLGg3Sj
zgeSE;MLH!7{`uzy2j84B9t~xjK?wVK7e+UzTrbq!d8}uiE|8el&D#%Hh{IS-q`m#*
zx^!M}JcJV%XGU9LnYrSm^U*Z_QA9~wKj?>|VGG!Tta@e+#QTO&^J}F_@8~$|?jD87
zI}@J!(8Yy*I>E$<r6iB{PU?#wKY(N*CK(RJ6SL>d_C13ECzQO!kQ*-w9Xi<!Iw*)~
z<#n~A&pjD{dX$i>WR0tgPQaU1YB>I!N}yYPM5vNFOek2zrxRz9CO;sD%gIT-&o~$>
z&hjQq8{|i#RY095;@nRw()J6shQPJ0B=e2p8VvDbEM^N-l*z;rZBv*q2y0g#V2QQn
zNf@t~%yh6Q>6WQg89-M+e6m{ye`FuQt6xjh=STGhZf0iZdjSr1zknX|<O9DrZO8}&
zxP6ekr>W@@=HxL(@sPVo;@5af{}*p()z(%QwQDF|+@ZL;)8J6t-QC?Ot}PORySqzq
zcZ#IALy_R_u0`IRy-)Y?_x*z8ni+Gg%rTyM-@jW4OYv@jS{}lB>AKM|>c@cy*ky#;
zn))rqbmimwXcrQ6X&Q9#6RL~U)itOMf-$OxMQGta!;@Hv=DDkVlW}qDCPtFO5P59H
zf6{j@E_^vuI6^!QM>cMxT>QBWW0~Wi%zc$8awqkj^o|T^Dcxt0ty`yr+p)s0aS`sY
zmDudJ{m#zLnA|#2&?Y$nM(xg$JIPVEL4zIm=9t@shK<-@Z>T@mc)4H42uUJgu#d>Y
zl^2!_Dc|R;<KL%$Hmu5fLxqnT6f(u4SEQFq@sY*Rzu|s6j}nbYrW+J@@RQ0eoguD-
zV;7+1=QdnLF(FBUo{sA0GIs6E%eap2&7LW`m{t!(_E!6SSgb^9vm+~36U^~e_!G_|
z#S{%}m4wIYmHDal1P&&HE{HUGukUwF6_PiC_O2|0c>X=jSEggoY>GbeY-8ePlwr$^
ze&4G+>N|Hr41B1RB7`dl>AUc;&*$IeqG3HrIS6HrDe92&f4OVN#^|}okq~I7)ovGY
zc6t3i3TFamwp3I}Q;@xm*t*RFKDj7|6WsHK7rb5cJaQ-TAlnNv%)hMoWF)rC2d~ve
z<)!!A32?Oga`4-0(HBln7Soo<(FA^Wn-^*lN9dR{ohbh1%M|6`pNz#3`0vSapo#d}
z%uSSB@b$Wnqc$*s6ao&8+CZ-?gzJ5Ky@XWa5piPX8!!E<?Aafh{jKBK<QdPnh#{Gy
zss1LsZIh%qQce}tjPlQ<Ewv6`TY=6p@uhmOSSIt?!hAMh`Vb3@b7LWl;6^oAOaLcy
zUFBgK<6n~!X098AfZ@F4wygZ_SbZ;(xUICCq=hdn%wyzPYW6SlS4Hw4_ZWKB+0U!s
z<v&_czpne^K?$EEl+dg!U<R{@eJS4uAb_x?mNPYy_WeYxgWn%SF^T9#TIvNL3@-n)
z3;Wp8tPY}-tg-Pdf-myQgjs~D6N5QP1xdM~Z$0a|`=vhA-?GxOf8x**gRpah@$6K4
zF=!G5@22Ca!jbb%pFdmJpL4rcS0GIlfV;tdinFT(=9h76M2gq3mKMA?LpXR<@U~=6
zyZt<h_UsIfbYIKrB!YIZgEb@y$}}UIFn`1((OMns^50$dVD|dn4P`Af9DS{vQWW?g
z$M7)^{#2HuKIQLp9k^Naba}D9WQNG@>;CJZ1jAV2U^Vby+xnvn*}u<>%iads%=B{G
zBJgDvP0Ld$*tJBB<ktNC78QlvYh8NC-E7|4h@;}!<R$+F?YDDx(_4=(=G*qF-$X>t
zX~;dvO~V9k<iJ(5jr<QK+>ql}$6)<*nd?~D*Ama)=9Hj35IZV|>BnEEw#Tc-4BMvy
zAD^eGjh^jhz6<s%8SpaO{CzEza=En)W0OlxVu>o11ZQ%B;@AR$eF})CeV?JsqKs}E
zpGX2Lm?1eU{N5JH`91RE`|=C8@2Y)6XF__zU(Xqu%CgMC%P|h!vdOb-)3=p;)2Jmd
zl)^b0&B1&wn+hfRV;;PpHSdg5SYVsO9+=etDw|H>CT=R1aEnGsBr1%@O$twJUSe9)
zr34Bl7#yD5r+p2Xq{CCMPW9aHBEAyiPEBG-7X;M=4Z_|n_(h=zAoC20dRUWdm>4e~
z77w^z(I0fw_m=ip8NCX9H)(&rvIhowJJua^U-;DS*@aHn<~X*>QV*G@h%Nv~zNSl^
zDLt2sgXEjXas<tRMjcQ$noc(xP)aAqxqQa5`I{N`8k=U>>i#zyS+q_Ire=#!BYJ_P
zA$AxGM8;hpkqI^IU&RH3?~J+ZvI>RR8e=%1_&-dFie++z6`%J6FjyI=Gq8seJ)0D0
z5=%zH&u-@%^1Vsz>R`eeDPZ5q_SrllaAI{CsA-n$@doqgzQCi{jaj3>s?U3{wp0q9
zxAaIzGRY^8Y%2=x#s3b!uN(@+=)&6YBtyk^1j#Q9OWj(XcobV$pb)t#%Hd;;j%K8c
z4FF%K#W)zKO+TZh_V7`|U5b20V<7Xo_M7vBhbK`VAnl0d8p6YdK>_kBalv9?OUO|Y
zMB`9@2}Lea!;%a|UL?e0b(>wthA@q*E2<wIR=1>m{YIC@Q21KmptUO?C>QK;E?Xr0
zT3&$=F<(v(^EzwR^Q=7i2Pr=SS!NG?!diRpN*PGtukP3@TAGGh{0pDKlet-g20!K3
zX9MPEd)`_P#cJ!68^xb5L!o=AAtO@q#qQ7cV_r^C_SioSdgYJCH9DGbaLQ>-^Noym
zXijtfX&-lCA`?v%K64MUlg>Hf;2+4<MoeV;F`?ZGa@IUe>zuyuD!bNvpHY;^0C1^I
z;n88+Npv3YOJs88(9(Ky5<E5Q*8(8Q`6HIiGp%bE-w*U_+6;NCps3RvrI*C<B-`fE
zG_>!ws(Ao@>+v*pqH?7yJWg>G(2wt=ZVI6S&6Yr;PB$>=Skf?MC^97?<6b;J2uNfe
zPE;IC4Nr+92OGTHP@!4{rXC_EKf$Nakz+rxrYXLkPF;9NvM-WfFV167P68t*Yql$c
zivQ3T#G2Def3|S3YNMG?nChn(Lq-(e-)vVtb2kBW7r%CMpyoezA4%J`Z}`;4FZgT%
zSfnV4N)66+Iz5~L_OqdA8WjqmoXGTFBBp2Tn%DLxT2>jxT2_xF`|0OwbKp;S%y_XC
zbJ(F#PZZ0SB2ctNCp7|~LZkYm`X}>GX+;i(LID=Rwjb#<p>?_GK+JFfzcN|@Bo!M<
zer#&E#Nz3G^x2?muURDEe=zyo>>^e8LumF(q_vE{mi)u9S>j@b|0(2XMZAl#s%NE*
z4Q1xyW?6aUuEFN+hu1I>ch-lS6G~--VW`-MrF*UV)^}d<F$`fw87jNatU2dA2egxj
zH+vfz5uobK8s#vAn6%U7Z*U7}6gJNEIuW2shbmVKH18_{XSf4-in9c@KEdIQ!keZz
z^w$X?)S)7Dy>!)hpCnJ7cnfLoSv9Yfq#;Ms08oWbW|gD$fXKUCAT&yGZ13>4Yr0CU
zc=#_Ka(PCYAnpgoCV&5=RGcbJ&h7fBsPUo#V&#P_6yjNTEj=AsF8D)#bDqz^I#SiV
zJj9VZ3R2!YTM0r|-c2(cc2KXPrJa?+%ByttSnC3VQG04t0Gnf*Ax(FUw{c0l5HII?
zdW`lCyESl%{tlp@a*-%&lZK{Q#ED5yGJYA)g2!oit`7DaAf0F<a@L@ZzRqA)%L7=0
zP{P-?Giak4!8C`G@Wu2}QndJ<q+oiU_$tnF?5fr9F^kkW>ef8*VP)kJc0H^=UVANr
z<<`RYa>>ye_E94yymWwI;v!G`3K<sz)SDs4p<Wl3%U!-ezKt1wDCTt8IR3S3r`P4x
zc+-_7{RSKCv-#xPI+pC5VcbRhow{+8ASBDG1w_Mb^M@*`YxTn0P%qahJx8Pz8lrV-
z3$x8*cS~|e{+4Hxce@*NU_eLoBZppOHF=C=HWD?;(c={nr$U2CNsi5Jh`op<jM-B!
zyLMbq6ZyEy#DZ2l36Ml@JS`-`geYw9Tf!p7!G=DXL@7nYW#W$u5uIq>Lj9n-@Y;HX
zSc{Tf{9V{E6?M*d%Aaw93ZH_Y@_?2xEaJZ*5En?_g3Zfq_?V4TebyZvXq!udFYFlr
z`Pl8Y4W9Dd(p3(f>qMJLa_({R*56`hz#6%&n9d;Im%*Wm@rohfWiiSH1F?<{LCNx@
z3Ojw?{@zxNf>}xxEOcC1**Q7cF)%HeAW&+f%$t2w7HE#H6crWqpdf~T<(kggO!#$n
zC*cGV{Q1qQ+sV88&&{*q4Pd}$g>~Vg?aZs)?6G{pvQh86sieRG`bg2dtny&}>&P4B
zK3yQ{>86mvjC`mRPeeuQmewl7BkI(N1Bl<v6Z9OVlTTrO>ONG?0a!_(cpOZG&RWJ>
zxhJ7vNua4qNpXhCp{9RkiUUw;k<^abpOPlo6%vNOn5d;tv1yUTT=iq$@QI})hB{es
z3`b!<_1wAu7OwiP*G;e~vN$#jUf{VItCePZ8p+ey6NQ0TDy3Gq2izXESmG<w@HF!F
zesXR~RADLD>&V|-WIV|dh#>|v+c$)Z^i(t|@`0rJ^vu*8$c;4NKe;@!j1pyInhz}f
zYGR47glm*D0!x+S+8>SAup77tmn_)YoZ{ptX2!Gg%QS6elG!C9X|0@O^en{~FyL(u
z&4$u)bJOCu*!PJ+Ni@(gzbsiK@zZVx%<GIWtzi<VjS7LSTwaoAuccK$f<9hbRmxAr
z<y(?!3UTFToGa_i{fm2G%j(da6S<la^(5*e`ma?`Uw9WfCcUPxBx9_CLQ>XnB++QP
zv7kCk=^!s*kr*QDjVBYI)#0v0Fg#&VKNlj+v<M>ZGzg6k9=UH-UtXOwyi=XY+)u!&
zPzpt3w^P941;s3efY!H&gMkSHtp_WHoK<h`6DgnfE_7kf`d$ZV4L<D`>lIE1o|o}{
z)A4Y)_lZNFh>9tVoy^b9MCr_^I65dRQdSY+;)UfQwV@s+T7bbWc9udfr+!}G{gb0o
zV_i*qAR;-nh{BICp}=mVVa4^AyJ=boeT42Hb<T=H@F(mtm4y3>T7@z2tYN1M;|KpS
zEa6J3Je?J~?TnQs(+f?rocj)O?l?~R#Fxy$>=+E*Nr5|L?rG(zq0Bf$G`&kD8PK_1
zL(wn@Q*^Hw-jtaRP@$f3SjAt7r=?YOjam_v8D+t@aKs){*TJ5hB0@>`hmxItFBCU<
z?uW7avMj?lm>4(gB<ym9QaMd4u>39eC>5)Ll8iB<hL$P&F{t~KRF?!=Pi`rpl#~P>
zGUqxhHkJ>U1sk*n%qlM??4XoQ$;iv=PxHq{IvhNYd&nUj%fn@opPQB*!3DV2|4DR%
z2z?X;bH?h{J;zK+-~3IT&RGQEc(mdaP!l$R7ci7FCTelQnKN;4)i%AJ=&g3h<y0lp
zGTonygGOh`JZS0#&wgJc!!#0HrKYBGq^2b%LPfJl$r)ix40KG0(2KgIH7$H*6B7y(
z174$?blc+YqiSY@?xSu+9cS@^XLUbd35xzNRPle_$9)%8<_l;L(ucb+EwBF%=CL6f
zqxN-}gzG;T_`trBuDT|%1%)HRk2=v;uZm-^FWe_BQqPw(k~y8_qHe&X*zxwv?e@8}
zqQ~$2APt{?_627l=Y#D!tX*PmmhrmDDZZseDE~73BfVMik_Yu3tbUqejVg)Zu#WM@
zeJhWa@>pHGU)P%}QKsK5kfzScjKQ(t`hJJ$rvn*=L*vD(FVW=5<G+KC6&aC~-HTos
zpN<JmSc*fh&~xJ+ViwL%p8v`_|2G?t3PY!5N-0*4XyT=6yfnK^vsKWfTvDB}Y+%%(
zBy;@v@N+{L>utQ*(uuxpk@a49GCGO{mm3fJ4m4E8cupMQ?uPpO#rG?>b+6m|k6aRS
zWt*?K7>lrE3IuailY(kiWy&L>o=Xxjz^dpLs4?8ach_exJKkQI_)g$qw{@ZH1Qe5;
zGH!Rvxij9Gq7xpC+;}&23Jad%*VY{Yxq6o&Ix#+&(#LsvnT24cJvc`F`x<E#irA@z
z9GzM1)hHfa&XYvLcC?-SYY(SvQuQcljB#B58S4<ExO8DSiD8qdZc_M-|0Xxe5$oor
z+z+Xas%4wz8fj>?0KAMbb<g3co3&(r`o}!<)KIXFS1d@)OjCfnA-9ajfxl6><$Gld
zxw$GqBueaXz~rP-gb;F41-vbF%$=$Tk=>?$kC9QLz&j*V)U@6xa9QI<G4%5eA?epw
z+(>?u!FGjK=GtQgjMy@e!CPkqa#|nGL6tdowzOHQjQn3_MfD^Z#?PQCb(O^u&G3P7
zFN0gqyl=x~_^DE9w(|w32rE>4lbl>Rtj$1|j`*Z!eZa}R%#CdbOPAO4`@QSiPuxAb
z!LE5>5lXD*JzY8_+iM0HYpAMqNP0F3T4#8J-dKzpQ8-`y>&nNJQEK(G87rc~V_$NU
z0;>d!^qFL-+myV}!#*D``s^gIuBHqxd~yiH2xFKP+U-V(REge9K%-eqNP83A`AHsK
zl}M?gh|4lLqpNtw>r@p^oTdLF`29AqT6aQIfN6|rtm!37FoZojJyxF!ukl&J<a)f7
z1NUzx=|_*DkuRMhZn_fj%hy0-@2{AVYdp#<Y<|O^$Z+9-brudHUo-0zk!>C1pMO;@
z{$wG;Rw;;8v-vJuK=a&|ksYtIR~4H4jR&~|;DLx+*U5V+2l~MY!$*4ai~J;6w8r=b
z_RQc5g$a{t!=2PgoJsxC1_m%xQQKXbJWom2T5_aXQ&u;rid)6}^ACS6?18aRHxTlC
zv-)^GFoS^>=ydks>3I@+Ex>(2O-%`t)cw!T{>lHIrVN4kpz>jAS?gp=%2cL|zII+O
zszOh?f+@RHqZq;8-^Ik;P&i7CrD(*%!B``rw7gpWR4H72RC&<Dsq%spM9WQ|>uED?
zV^wY%g?~bDg9c6AV=&;>x{Vp{ZfCb^PVOB|{jHfP&ME2HY-4qYcaJC)SduqoQqrm3
z0xJDQV#Kwc>wwAO_>=I3$(9K%`)%Az_`}Q11em5R!yhcLk`u_q!y`i6OL6c$@;bxc
z`qvLRs$<S6No@Y(6&OQV{8#nw`oS2SlNur6OH9se>B&OB*XZO`uA6yK)LQ+?m}FV1
zcD^#IC6A9wX-K7Lb&9E`44-W07=P1J9hXe@4pwk5T4VKlZb8nb{0<_VA#HJLE_TI2
z*CREKInnq!SKSU~9co1@9{<<ZU&cGY3N+ANaza?L*1(qUhnX*8q5C~DHNA+7O%1+_
zs#GQU^2dEeWgr73MUts7?1<;i-O_YYigGxF<#A(SoG6ZpHc97;Mqk*3^mJVh#=&0o
zl#a9kRf*r(kZLV~8W}uMYr@ahn#<n6p&Kq0q=$nCV!zz!-oQvxLtHes<>R1#H`8a3
z$o@^R+be4!^%qn;Gv6vmKtd4+hKO}iPBFp60iti%or3Ss9~zT@eMLasAWF4VKyG<v
z!EylusoRX7`rt^xhIb+)b4|}O&FkPIpqc?t1E~ieRWVKL0cO!*g%7B5it!>xdwXfv
zeui^GaeB||#b2G>xm3&+bQ4!ne}O6#&8oL_x@?f>JzABNfX?@)PF8=6|3%+WWW%@2
zN_!S9U}ph&hG>}HRi7K*0ON1m&iSL~skMz+ZHX5u1I7VP&UfrY-hnF?J+6j_=X`ch
zr-do?JnhE`ZbbdjDA&iV^~V&__*T}j7^e3oTX)wL$v8_JN&O@0X3)~Dd25dC)O-Gv
zRl(iURQ6P7|EtQ?_<Dkm`_t~GU64WtMkxAQQ}4rIhcH!HW>L6wEw6IXO-M)_uZ2sz
zsL94&bj&Tr*}qG$A)lbdkWzO2;ST`GtMiEk)=po-ehp><`-M<ulUexb+xy~+ijwBR
zjdD-R%<S-b9q*k6Ch+x@sOZO1oY1v$!_o0mBOgY|%<8xP0`^8sT^-X{qRP-*Ke)5$
z{vzduHt|hScCZxPlwl%S*qF-7cZ*OWpn3O1pM4uC(n!LWMs+kvz@5l*jgo<T<!M8&
zV_c7y?>!eHemTQXLf~o3969<oAH#MrQljdY8*=mjxcWc;!I(VD7Bo|XF)YV07K(+$
z`(CcLMk1-O=SsBrRw!m~Zzw~zvtc7Y(*QqNp9V)z6;XY52-4QoSz525c~A&fJ@voD
z*1|U^<{hlL{^y{=R6{M%PR_~}N3F<+Q}+_(SqYBivig=j`k7}w5}Nn|=VfWg4|iVw
zgXvJkHYUj=2S=!~)wEuQ5!6--|AY_KaF?cTu0qbxuCl0lnvBtSAkBhlRK(JT-%nka
z{R2xT&7wsh#D+uVJN`yN!29cA&w05VUHse9j0mRMIL0xmN8r~IfjSmNuDP5a9UUPx
zM7k+M$<|_#T{qfBwP`Zxw0JD322QtSP)S0jxEKpCNgGFndVgD4)6?+wB%F2Eo{j@K
zY0MYi<y#(iq)>(sWHD$Kw$fi6XeAnbMc~4rhFMvUW0V25Z$zI9n0t@zN)pmz$^WwO
zxjNaMUeBaj2dAnUUzn>BVfa=<x=h+jh76!bOM6<S?<a5W&RgXrlcps<K%108=Yfo`
zIF#n2c7-PYwM;PqTgG+-j542jtiSbHBXU`LdMN<K*R0-?wi_+Y1E^UR1W(JNno4&P
zHq42F3yR_`G8~Q)uZS<-WXP*X8Pm|+WKzneX3#v0+8s@@iaMe#+Z_(S$JmcVih~g#
z7xVd5$pm>=1}~GE>y|(e_-V35zYac9Bzul2RIxwfOLfJYRT?jr6pIO_<UjX|X*!7C
z_J$Di936n~wsY_)vBK|DOOFY|*L#TSbQs7?6B_nmoZ{^i^93;=^<9^CQvI;x{g(DJ
zwsrLJocCZOa@c=*3H6ZURtymG_pQAV;g!NNbQZ}BNLv}AEj%36#0dGck_2j4hR*!Y
zzK;DSE?p$JIa^rRxfsd@yQ*^61!*&EbfWqLfp{Xp22laG`{%Q(O)#H1`>+_A(7^&7
zH59nIgIN*M7>)e(6cfTx?Htj(oPz7ti_ZUSy@iK#OD<AUA2L78sb#;-dGe^-XOhm+
z3GNX|y@<pl7k0JCsh|nNdiI>?Dqrz_dhM|T%YlwmBF}b4zu=Sj9L%(*`};yajtBpX
zrh?l@Pnz@&_9;nZ0t~na8K%#Z-F_?k7d~`<K_R-XH-Ev};@5rY7{XndWqv!VI<NI`
z{}j0h2{2fzWXMFBB#Ah&o0v_Z#8nJoY;yd@AayQ(Il14Z#z=}X@sgpq6sQS#p%Kqv
zh^E^OZbO&twgXW<8NVl<_NdnQ_eZ`%Tn&j<cf2IIQDrYTik@xVQGoY+v54xqqSUhX
z8}C~@2RY@$LUX9@IaUu$uPRm-)+sh27TA*ZX|`Ff_W9<~m<5*oY|gJ@qe77W&hYk^
z*(+4}sK{}w$urGjxV%WOrsYK76!57{f|3|ygMO{%X7Q|^qe~2|>-!SmF}A-Wd{^JI
zS1%RLfyM{^O<2^!?$o<KENMUI2kE-*cCMiCn<F!9!7T<NZv(AHt^ZKrnqw20XHtQr
zk>klORwlV5s3QhGQKBQGhr!c^nj=#Eq~sq>)E^3+185l}&>x>To>Z8*qYXZ`_3Aiv
zEc!_E@xMT4pgOnadnTk_4FR%lU-dWm|5<K!(f>xmsD~Styop=uhaK;G>q(`I^CvUx
zv$ys0x$NzTztTDUd(9v3CHmQ$>1a!9)%!mfRb8QzfPk90k=yvjfXDI1fdo~25=ucB
zp9J*PL*htV5t=xpf-bJvs^*@^_TO&gH5VzQcEay-(J_dANWPbsTZ;s(t(Mp1!N((g
zujNJGx=1?_hMl{Gf~W$#3%*#+Y%R^j`C3QaUa@MO13Nt?CDkWp$@_+`AD%A8xvu)M
zD+#~(AVh`@EuGAy{pJJOZVSJhLEtTy`xi_cijO_^>ij^M?2_dl-O@r=)o}$fI&1T*
zt`uLcq0+;PpLbW4;y%B7gFIeceOjY0!4<_nJz<0-(>sw}HoBG9VM~5Sxr8VO5`-f~
zYF<20Xh$7Mtir6bxu0VnLAo2HOJW^Ic@|P0c!x$LaQLx&lfuk)J6Wk=e|v1LCK|c!
z4)k5{4QaVSqj9RZ>Cx@|!O92H6@Kes(3``0PC^~7lTuEEeb>|ceFlritI?v;j!8%H
z=nIjFVD~C?x{4F81F5P~ln&{M>R&f{gr`VK=!evm;sn}wX`JW!@+HUz$r!2WvN0^k
zpRKGunNg$J_w&coVArS<<=}SzaOm^1R=T}WZ~U^ny#6?Tf16gsBu+Z-l3!6sA61dc
za9TBn;x{C)OjA&eRzgxh6ssw}`T}xCLo27565~K*Nlr8Yu`7Q2i3a-{=dd&p&Ta@!
zm5gf^{;Dq~675<%yd*o;Q77)08gI_m*P7GY&0dT}gxA0#2ta*59&OJ;9&{h8P3t=&
zCO|AfYl_}ycDHb)igq_MiOz*Hg^fP*+kL6waJs-f&TpZgly4PC%4k1D#$1S`wHUsZ
zkd%^^giAFk6YQ^4{4E}}X}F$%SV=`nIFXAb6yvA+Vo?oE72kg_Z0Vd4c+!ufX7Za~
ze?`t1848*|lufxUlrdfg5roE=>tgczWpQ(tEOunkdw*xRwy(~&34GUSC}6#vEe}S?
z)UBCXRQSn&l8JN2o+{{a;+p#{jO#F82aj*afRObH$o2!@A*|55Qs3#$#%o145D9x7
zF8zae3*gGraeeJ_rbE|6?<V?@``s%M2L79Ke%99Se|j`zPxYcP^hig0Bk}yU(W7TY
z<(J}<GH-Ayep}-6SMeVw9YP_@4?f%D5ccQZhqE)x0OMmX<$;}}gh03IyhtCzy7!!s
z1oA+l?We84=&>LBn;~A5dz~1Dz^8xT1X?w{*@Om^Yctdwg9UsK@@kf5tX?Ze>OQZ!
zVcabIJ}Unvh!MGS81&3o+e8ZYFtk+QqHhY|pY;+w9>s{bYrEjcH{pTZfFk?ryJ|c~
z`aZ@Qmz>`9-ET%@d<s0r_*5cPM%~aMBN+>O05RC3%UDklTr+-u*+2OO^3VRSPpY_N
zm=Yyc9C=;!dm?W_w_+u>SC2A+_{RA)osXtm80M>)wwcM_MHIL$^Q*l*7fkx7pgwdT
zRA34IhFDp^LOyFXpzT7a<?pWWdk-A86av@%RYvj&^7W4TLw%f|SwkPmzpkgVexDcs
zX~)ht)n)e6GewMVfYv&S1jC5l?wU1(A8Zoxbx0JX*<q_luF_9QArskBAx#>}e8F}T
z5p^W>BMFAKzMYlzl)``H@2}>9qgg4RZ;;x(AmHAM#q0T83?uGa=_7KM*#}^l1s(1@
zM`Y>e_~xG??4MC$jfAMF6ehL4)JE<up(IaXYuMcw4f`=kv)LCagn{Lg447SPNG}w{
ze=XVF!l5A0;_aiHp}=4nak;pf(z_SJZj52)=Ngd<^b5yuvdc|*;Ziuh@p~4k=h7?g
zIO4B3=HwNLQ2IIy2dOG+jTaj#EHe&^5UEvBElPM93+gG#X~BDJm&2yonWgEM<EGS)
z{KQ_lp%mKRifUjq=pMSwwle*+cr$jbU&)4>FfO^L0GDIsn2*{*6h3m^>K!B-p%P1H
zk3q@B+RPD~rMbN3^A*-MQus@9zIY_HwYEd8K226Uh^vxmceQiD!Ch)W)>d&zp|nkX
zXi!3H66KEYMGFVP;dvWnIRREh;54h-x{5}2KK?@GS8qpP5wa4;tb=5vf<1=wqR1C<
zhD59H^y(4<Rn6haoZG65809cz9##)>)<=;?%_?1c%?yZ|%^3W9jj*!~+y9pP-c*>S
zRJ68Pexv1g9)j<MEAX7IT=EgBE9%RCWgOi8gy)xt{E)2CwuZLEf~EY~Q-jbCIs|Ak
z{?PUzTj(-+eQp-5^lvJws;X7QX|J}(y5A;JCQnsiEFSoTo%G2^xcN$e>Tp^?@SGEP
z9MUQllT&{Kc6i?NReGl4+q)OT+*_@UApExCxaCk2z{nP6&0guMx>{!ZxJqb+E9M&`
zd(r43fnIS4=Fg0k4&iv?ZwX}1Xi7F;tMnG$`m_;1qI%`{bpvqyVsqU>BKoFzAZtgn
ze;HVKP`H6Lm%+2$)PZO2o$fiX8-`*}yw`(NxHt;_i!*;prB#s_EnOuc<89D8{B4Lo
z;=0j;+Flzk8a{WjCakGmZah_PT<9~C(Rpcq@te;Isw;S^{tYLP_IWBut|6Ej5Pj0`
z9kBQH$~*X|=Mi|;;oQCm6)zX9@z-#Dj5ze~Z(y0=GAQKO^GFwawaxL5pw~&A6#Qz^
z9j?vo4raO_nAEq-YxHK3PKGoSILcmdv+BfH{bF_7k{P5=$YCGBFCD{@MkDD?91q1p
z>)4LW=faMSm0-Rm)NjO%CdHidJLjoX<UDRJ!Fw>&L{nKj^T-Y}VA}EU|Ei&yDmR>=
z{3UHPI~58PvWFuI)iJ-8@b+XlJczS7`0%=(*!CGREV#jM*h1^W2@HXBi-8@RmTpA%
zAiZUsZ5k71T`s6DQ|I*aYxM#FRDOoeSW~^wafWm}r0H)IR3J#vy4vsDvri(N)h56b
z;_$xKeY?=;)B4&HK;b#ke0Q)8=q?|!@1a*|F{CrBcSP$%yiH<;L(jMBpdas{1foP9
zOulH)Y(*0%4M7Ai;*I2z_MDdCGF^t|=_}dy%wOn?Pzfg1!bBq_N<U97BSq11)rx0v
zFbMO<tLVH{`)j3^F>06vdfEiwnGg$buO$3=ULKkmdK!cbZIdP3z}_hCLgztEh9XGA
zIY&G1bvvt5In5V{vs>O1bOS!n2~dJfnSt|i5YV`5742-?t8X>XkagKBbj;`5mS|8l
zve~2Kk(29-kSU*K9cuKLFGP9Fx{q${qV4$HTZpwPeeAEH6vR_aA&+-F)jSEUO2=N4
zVLFQ++PB3S6ISD27``wYdXgu@(a>PFRccHLyVnYjp6ux>Wu~tpzGbd1R9*t(qhr)3
zD#_|(d=XPs#5+P6WSZ}2##;1<%F-N?U{F@dqk5wFhy1MED25Y)8W7}lQMYtq`g<!b
z<Dq=c-N0Fh%~Kwt*!8~hUeajPwvoMtn!XUA*y*ez-qHqLdVdofHn)lBvpoOl2E8jd
z%ch)>yO7pyY5(<{pzmw!(DVOVZ+eqr=rTL%-~#@#=$M~g`C6G_jdvyMoI%;-JJ7h;
zDT-^H3*~xsCy!foD(lua24!WU>RI8qY>+#>SQ9=^vQk?Cfuxv7xEycF#)s%%JqbIA
z&RDiv13h)j<As*0LurCcrMft^qmq`LBQ*s@Q+^1JQR6b8;*aXj9J-sFo2Z-e)77}z
zK5tu_HVK>c3n;eYy=JXD`YOZ6Njt%|p6So45SvApYk<;wQDb0L8;QsaLc<2Zt9H5P
z9$<`m$~xZ4G^1Ha-}iiE$T*|L@bc_y*M^^@=lUy)kMU#YAFmVcRceY=G1nZs<^hCg
z$02}_k_ME2o5GH4wvmPR9iFD`#K>m1DM9BHTb^q}ig67q22VESxJyy95?@ZMc2x7L
z46Ko6akZ|Nf)0w-+kNAFx(c2>H668y-3ridQ~M-@;n3a2Hu{4rFc9@wr0o^%2Gdy#
zQz)ceNYXcuw)_0)a`n7m*g3t69re%SDpUHHH>%0<d0Nwc9|wyt1Yt>(pJ9{F_=0}K
zcdStv*XJV;$LVp#w%OTcenS^L<#mxa<E2-6L3(_iVZGyFcmWs#1pFbkaq4cn$#7_v
zizogI5)V~+<%LdR0AznRxMVpjG}i<7zI0o-bgF+Qj2L(Fj2sfB-<x;n*ZkWm>cz$U
zAe@~|m&(i~n=Zkgmk%%<9q(w=WKxvTzO|>eXgLp$xdWc+xhMo+%zu!6Aiaq^t6sml
zeozG(t>qxj2I@Skb2zmzSk;jwCk3dGDAQ~Ac{NQ0t?jkypFgtYw5afrEqNJ)A3HBx
z{GFIyumeg*pGGrnl0`au*S)Nf>>hQtd0gdLu3i-JS)ci?T@jsM=b)*}1C1tV8`=2}
zA!d<}zod&zPSW73JScyR0A}Gu<918_gZYj@)Bg<o^&iap7Ng-y$PVUzFmVMD|G|8?
z16clpxioAHIGhi;6$<$e=05!%dd-^n46J}3hp&U`&+N~ll60TVW-C~y&}AOEN~;w$
zi9^%5fIp@zSwICW$j@yw;?cy%h83qtLN9RZ($Jf?2^LEumZjSZ@e7oE#`O&Jc{BUq
zWF_2O9cXByh<p<L`zBuyV!+C^Nq{PnVSACoW7A{NQ*$eJ=B{_XY<=L|4n~hAs?{uC
zInXb=4BX?!Oj_?Yv|r@_ZlRa2x(o%V^f~hdjUXhW6e}Gc8qz^Zl4)fLpDdilZms1J
z?Qx*zNcD^jto+Tw#)_C>dKN~PFI;Jnr-Gr8(Z!!LEhTkQHL4ghlHtdrb<)F|G&KLE
zv6m<$N<S+P^4E!>OG)g7r{Ff0S%2Qi?sy3w(**-!wq*y5i=IVpkap{TZ$$~Yx3~vV
z^muDb-F(`GOeJ;$yyslH0DVu<dtHwy*5$?-QwOk3B!kFB(dRuDfEN$SJi%rBct)rB
z+WZBJwTsF|XSS(kh_3sF4pmDN2TF!W({WD)#4N?W+i7sgu3-~vrddvmQ=T&F7A6JH
z2H85Qa57pqqlLDvv0yi*Y#ECvbtoOPI;o76wjibI?hn%Gl0g*ZE(5BpCX<1mBpM!G
zmxTSW#P2S>p6WtUrOYCgKWO%#*@@|r(XoEqe_Oj`_2!pb;RT`n(3r{R55_>3RTBE;
zO#*ueoo%!6GTW55sk`ZaFoZQY^>O7;3v{=4*8)9TVJ91pGhiyGVZ7-A;fE}{7WSGG
z&IRsH<dH6K_ut^TH?_X2*GOwurVOXb?H<ecpy~zDiC6Mz6@Vl96n78=USh-{l!y*e
z4(0|6M1jQl2Vg}a7<7{A8)+W+VywoA;<FgEQ&kC1KPh-;#c+`cf>;0p#R}f??DAin
z$g-n#<T;23Ul>zhu=6v1qFKNQQ*ZIM8?^(c1KFUr2$I}PAmYBx)<@r+uLgwLwpyT!
zu(C~`_qvgr>6awYr}l|trAjvMmqcm%7D?7j_^~8pMeg@8ohraa;WL6$^UD6lJR6_r
z;!KKJb7papWd@%$Ytn#Cvc=mUqfEI_ID93}$m$DKDT-T)QXlA;aUrhYts&YqA^2<m
z_0{0he=uVOzg0ix`gNP$*rDv9_WqCW-rfJf7+M+6)e}h!=GlP?O|{<uvt7lVKaz9U
zel%3#)#B7^igU1}6vP|SAYVVH2ab)&BQId@j!krPW0e)lM86S|`qK=Gp*;KC)bC*o
zN_w9k4!>zWM>GkEasNwk$ntR7TrquOTOuo87M)>VvWvdzu6+z()vV=I$bf$-*Lhnp
z3niMAz}kcaIrUAMY>7hrjR?Q)IT7fU#b|e(H?GxwLUEI^{>`A8LOOtkNNpA_HJdDF
zYr^L#X|55L?Ww7zB#~q;iTRZW&nhZ=woUw(PU;*LnL$w%NMTB=SkvzN#Bum{W{;OF
zzgzedar1MOj^<&U`6e!_PvF~Y$=8INUpJAvY$-R9HzYe{dp8NYHsgUb?VAvEH)r%I
zcW3Pf_QP&(O*?NAEAzF?npXkz9p;mk@f@2TS0U%Rt1Nm{kv)J{F{=1~Fg25LffYO6
zQ|ry5U)J0A^RhbX(MwWnduzQ}l<O{=NVHxD!2u>RziHZuxeTnqXdbi*8G=NU9#lX$
zq<{PpSDvCNypt|5qsCN;$qm$bvK8zih-pK4Iey2jyVS8vDXC4QZfUrYg(5mq$b)96
zQp?*EQ*MdaFL7@-6gvtJ8PDwr*E3`{+$8Y56(HF*Z2f(Z*|QJ&@_S8;<8Hcb@E}d3
zrPu`=>sr6%JCq2k@m}$!xE?voBJk`1IM)A}?F19kB8%^uV)wqzzC&;|#hEknns=3P
z7cBeKoyC^eJfgxmgkKelJAcln0Y<PF7;NRMuU@j~E2=_fQfA>3<macV<izJG9eel`
zdm*e4K&e+CD_;>~LC!d-lAE0uuVrA$N7fs<`zG>+{U1yb1hV&cIZ$KmtlgYdj0jM%
z1D^VBl*8k-X=vH{$btX_`<^dO<j=VD)IsV7{0uK&rR#3{88*E;Twd!Y%(+Vy+YD;C
zt?O}?+YGhe%bK0yAa2IZSN4#vW_{4$o_5z9^?A3=R31ROi$uwzRJBDViQW@uU*9~5
zb-G0<mpXPlgPlKvettS}w2W@0UC9#mleUz@Bg*{zSDarNMT;`1_n!{a$jU3U<fGFR
zFqKA=4-Yk>k;datf|;`k#Dy3HBPn&vIq7-k_9xy|9if-~t__{V5UyN94P~7m@6pc>
zfd&9u%UZtoRh!k`lGXQy?kjyl>x-stOCNm>l1;F6o_CCO^49C_h-3VKOM|q1n|@gg
zTfWh&>alURhqdBbjOgM@+4@uYsD5Glrkmc;h;ySM`|4lcqtz?aCeJZ|eKS?TZMnjG
zHZ-=RwZpFEPMnl!c9!dsB<xvcXC5rT48W-J1E4b~juI@KXejgr+ef!7ebJIh3aqZq
zq(8)NN<9XTd1zW3U>CE0`qA@GMRwYJB=hHIG`|<uiitO_5Av%75TIb_apKK&4dOE*
zBvO9U!MY6Yw{SUQ7<h%a)PdNrAOO)Rf5m<VXOVQKwpX(9^Tw@ac8KzZ=-h2d2D_E|
zM6>c6QQMVM&9X3wrR%C;m(JJ-e2kU1Jzwcdj-qXsGd*U8pHg!UEdY6#+h(EwFv>6>
zh|CwRE&&@);gAxIW9{G@p_5w>>X5IQf?p}8O!qiwG@5xT1Dj;@1{1x{{}a(e#s+?v
zgK;n6T*}9J#H0i1+UvyIu+DJ$rt?aXcY^H56H@DAy=R%zedpA6hEQyszHF1#iKq59
zMi8Y*C)<sub|qmzO;lf}e>C1Imtx=ihhDBlAwyJY)YRLmYF3il$b6;6O^O-=)}n91
zm!<lD2d)19GbXy^=TQ|2zc!<$31(?(fEa#R<IyVagLx3G!>`Lf5G$)J_poqrqdKWy
zd<s+3Rb{_V)0_M0?g$|gtlK5-^iP}7BNv4STQHSVCxqP<pHcH$PlfVNNTNg?t8hvq
z7uL~og#Gk)=`JBsDTPJ}aYDV^G7v_|@dT~Z+|Db8@TtrK4erX}H4m-7lY6K+Qt{dQ
zjg}xy#3Ek~W<`8K<E0p+qPde`IcG8PHS^T)R09L}cRq+f)NzSKzTS4-)(1WpzNfnz
zB>C)})o6wXOXKgstaBX?(h<iQ4WLNd+PHYvYDVNAUodSGNQu4)KTHwGT0(pc+cPLi
zYU4(_j@J&4o8f1H-yYh}Ae~oVTB!bD{z*7ev1{xl<wM@$>WKH50A8AWYD*oucQIA|
z#3nQs8E+)Zp<|hTB`NqX+V^@yOAp~1gYV{QrZwW$X-mDr8ZZ3YI`x-??^^nSg%DNm
zE3zr(|6m**LW&s)hyDE5p7Z76gEE4Hg}U`JYc4@B>w&(HjRZx@w-cNfH+xmles1Oc
z)#GO`2Z{eya!-cIAu}p|8zG`_XEy^;Gw3q<<VLr@c8o6j%O3Xj0&sOJ{{^sazc-O2
z0GOM8#P3{^X6D^IP3ckW65gCx<Sj4sKylGem>~&d=X#upCggcfCa5$xM7JW2Pcm$w
z82D2})m(TW(CM(+uneOE(|JR0<x$A+?9Q%;Z1pb{4h-{Pp?-?8aLpV{G!aoyPuOo&
zIS-o-jW<m)vRVHOIfjM}^=kLfS^<E;%l^%NXI0aDp>(LpeWN%{vcjh)vl;BEbTWyg
zG;da%v3B&$2q8Xp_4n^LXV=Ysfs|{m;kx)ayCzd_-_mY^FB;bBgbu+7bz@fD7VRcF
zEby7mb0|+eTpJZndC7D18&**6(x`MBu35${t*qnA^jP4xq?R_ak~Ud)NS$w+L*SVA
zFW!~1sVrf+ByKAk-GgT7yOGfxcn_N)G|DW~E$M!ujL{Cq6U8FmqM;>^;ZWwhWvyL_
zP=)I9C6tF^)_83r?_FsO)UhW`oUtW~M4-@}v;~fa7&;^JhVLXI>B4-I1#a!je=r39
z!Jv=?UAyjP3>Ab!q+4F${I|An8H}1-3}<Xz4Nq6yHf~IFq`V!*U30X%Z7(6#*~Eak
zSLw=Cmsk5x@cVuv!1W2WZu(883y{9u*A@*n@7>qCWOP<HC^d9?wAUHiYFg=ciPg6r
zZ!@5iu}FKNi%L0gRemcS!aUR|++b?qz~Wb}&FocJ=sK18S){(4d}>u)hNlSY@9`ii
zfu#j!o1zi!Ftp#|!w{XSk&Ix9_6r4(s-|aM;K+*BIMkf;IptXvas@pEk%*E9qVxj_
z%v{MTpHnz&%GZ~t#{9Q*YBL%nTS?^o&U?d62_PmjdafkGI*wjTqH`_;hSh6ydNy)7
zj59fS9(0HM&r#pEUO56UoYRK_MjMQP8|dXz4$b;z?b85Mw3xuH)iT94<rGd?<~ewr
zKI>xjXeTaEJLM-D2i^HRiRw*^YVWUd_6J@p6<T%H>dd~0BGNb#l|a?f{V3atP1+Pz
zoouKChzIRfe~cnLoz7JZ2W=)#h<RH)Sjg%B{SXcR!H|1PUO)L=i2jBE+MQ|F7wRCl
z=e@2~5IQU1Yl{g%0B29_V_ZSgriMid8AOmpF3HnC3QH;3I;)&99JP6uY|8&U?x2mE
zm2mi1*ID!u+8OJsNH)nI^QHlM{>dyggF`ux6vIGgzkS%9BrZtWB0NDM=(E0Ao^gEB
zsgJQWQ!>DAT5EvFESlD%Y^qPIlvPPd)Rcy?CRavyUw%Q4hg0~%nf`8GtwaQWMH3do
zfcou?*7;%JIRdKaL~rY@SRsa3hm18&H*Gx{{{5HLI8W8KK`?Qs)@s572-ZG>SJOY&
zq*KDV)Gvnq>(~wlvw{J{*3fk~1OS%XZqNCK37wOf5|tWdsmD1OV}D6?^U4?{IYo;j
z3N(N*9?Ee$sCg+R%5}>GBb2ulLP`<IgzKi~n@$Q-qa%axWbhSe_Oz`|$k&;Wv}dRh
zi6~Ttc(H?g(64^vMv>U=aJp69Uk%D^$-T+zhoE6Qi~i1QYnA6ITqANdJU&<AjEgB~
z7i#9<X&!vU4nA)lv9xpHkG$e9O#Ba~ImICYb67E3Br2=jVjv)<&d9KQ*h;_OANO@0
zwLDqO5fE~J6V=+9a0#3640WZ`1gD<jDgR713R2SSIwMHWpAt_?mzpRZ1y+g&f0ZgI
zuM^`oB{oJNz<?bH)I{ou-WB?-ii4--b@Cq!qba)i8_mT-{|$+6FtIiCYYw!vJ7peC
zuc5TpK45?+fhH6BZEcKJn`-CxJ~isW1NNcg@IJIKn^+@s!-;GNe7kg03>NzrlRBuI
zEKQygL{1jcIZv5fhP&NAqg5{~rje#Ync~@GUz4ew_d4kOpw)Zwh^n@-|Fy#FuYWd2
zc@&+(99XW2sFhm!Zh^<5N;i&{LHm}8$VEI}oeRg#>aXUE5F#sV=(%r#bOcJXX?R@U
zTEZ~K+y`lmkZJ=Us9`mVFZ|5;-33~@h7@+di^i4O=7{L)l;*oD`9`G-*EIMfz+obF
z7W^D>D3N(Iiwm>_^Z=sZwq}>tHKDc*JdpLufqb*H-#6RI9A=6fdKM!BUnwpH5V+GJ
z?q@`9GQ18(>M6c-LzsMHZgGXz;%44ol+wWBAo_gi#F0um)mmg#Erc10u`j>IaT<(L
zQ&p$Fap8d7f6X55jdwDiuM?KY{`AJZPH5l43Kn$k|BYH<|A+`~%|W4m&+|9cZM_K6
zf7DpTnA!AkUUnoj?we}<+dUqrd9>+4ua8iC()2clb0p0>7Efe7!r=<a5SeMpFj-z>
ze_eCl3Xrw}&R92(HOu-c0LE}HaHfityjMNUu*9jUcs%AfL9x_i+wu`0^3+qpNMUxh
zNqv^84KGm%-4#a0F{^QT0!i);TJNHTy)UsEpQCt|{64uDpWuY9%p-2WG0l9!?vdjx
zl?nc4K_nT-2`!T+a7b(c_6Ld)bmjS6=HIxpTAHqYZNB%q{$igp!FFm}SoPLecQJHM
zu&sT3m1}`qcw7J|)-ETobIgENtAkS8tT!B6pmxBF)6zsMpu6v}z@i@^eAVf%Jb``s
zD)n`PQV(7560>uYRwh~}_hjc5d*H_Euq4Jk2a!b>bR+J_M9h7D>#ERRRY<GHDb=+1
zkdj~xbyic@`6v2$UVDkRE6yy7J|h^`)5n_rNACH_vXZgiD2HrX%gRhxe|S*e^YRVF
zF2fL8_FB-7F4uZsb3GH^)J{uA?Sn<9^Nn%F`BXu!i=%RwZM~3jF>!jQO<=)V4S)te
zF)AtQyNO)4;tkH>YC!s7KeUs5Vc%@(HX1!eKSVvly+o_pgla{-0Doo_<wj(k1TSH<
zTBziP5_w0!@lUhT1oh;Cs0yDTPuJR9#4Rl|TUkzXQ6KH5Qz&^jJaVS1B)nR;)XGfs
zzUsb(5T^9^&Ro_{?z0&ogcYt5QG{UZ$JZO!R-uq4Lk8%J)5GwCOeKx}x#hz6EJ%-)
z;JAE~FK!j3i+-b9Bw)Hlg2A}CT*?^`K#XG}4}v3A7l(8Bpk99^GwZx!3EB!+D>LN;
z<7cDxNoPBqP!n_8CfNgn>So}vT~vYCw|hA+;o>WwzVGRlkps6a3}TeEx&n;;s^~J{
zU6mV`R}Kn1jSVoNli+yb{ixR|A^XAet&I(ZB9vSJLd-)UaAzE)Dtz=D^hWP|v@7U3
zj3PwQ0zq&CUmvpMlWbo^oCO)rLi&TGLxybLr(Jpr`a<kp7;I~87GBjFFHs&<p=VsH
zt(4xz=O|Anwa(q1%h$EP97Xo;IZsY8ta~hv>Xt7BEx~h+O(36$HodTetinsB@5or0
ztt+D^{KfqAhI6H4vfp$b{K{OiIMbA;7fSVoPxOTj%1>_XW|*j1v>8uUFr+fo$7Mq4
zaAY~qdU_ypq>=b5ST@2TKl&q-yWS+uFSnf8$DHX0*S$F+J)D~B14RFBW`IK>F2H5;
zo-1U8Ip*^YS3ez4QJR-(s9CFO<4=GSNIaToFW%&_eHVYJu?<o@<h$Z(XaqTAN?hIR
zE?uoqNVT3yqWze|J2TE_BxjML4eYSVTd|-6ePZ4%DLn=4P`9WnkoQzPfYr+Cpi!-{
zlvHNgYG;7LOH-XT73=}6HFpu~ca%}hI~QfyXS~##>qPgbA^`$N3-#A)%j>FkQ&Lfi
zH_7L?W+C~8c9EIuHRU$1JF_*%>E}jMf@dc_+Sf<3&YGv7A%A^mE?Mg&yex?&=SDaO
z&H-$oK9AH|2#8Q+!jz9~!nI-ZZ;bL2UZvAQ4G!xa>wGLk3m_L_5AD#+aKq5cP=hAY
zIHO!7!7tie`18(EZc9<s#1Uz6VzsCEy}?<Migc-!_vTXLj7bmwK3wwBJ{Q})tK_Ty
z_K;drf4NiriOWzw#OybktfH!VLD=!E`>X}2j|o8zrg^jN0Hm0^OyxdXve+G&>MBE}
zwx$HiB|TKDd6_A#-g6!+=ep=SW-Va5h8f_L1wr@tAiZI@MtK&_w-%C+H}NiSAA5G?
ztk#u!`^Sk#7Bm9=M@5lrYv^m}?{nRY%o4!4N$Hr6e&CKiuoKtEDhZF8KmN4L;e@Xk
zPlymN4DnP8-Ic%cYq35)Z=MeQoHf@h0@kuZ%6^h#HSth{0COFhJk$OC`dr2U6K*7F
zQUc?B&Rw&rY+6PEX8-3LuN#VO?~S*>&c_sUClZ$H5AWHr-a5nG`##oJZI>8ZcEhH(
zHj?ro>kj7(uVNDE?E578KhAuuv6HQe1@Lq9Mp(*O90edk<%w1wnb9>TzLvAWC*TNK
z{(?=CZS+;zt8;Zwx>}1;mSct}F4`EjG@r*5k2f4j_#b&7*;iR89-bFMY=?%OOZm$^
zLas7-?4>v&N|3dhO&}3i^23>n@uX${UuLJpzPt}gIa&6nd-JfuH+{DZe>3uEPM7fy
ze88DwAIyd9F8$S-azxWPa9pLA7IMiFbP}kr_Nt_-1VO7h9Pk<Ia3-``v5|I7wIW`g
z$_1n0Dp}|QvB`QwDL44MSM<EOy?~GLw>{>~VCH3d>1>uRy%uo*zo^qnt=mlaAf0p{
zRnwp0LCCbBOvwp%G1IzB#WfXgXdN#8S7|mT9{Sc4DFQKNGXZ{3z6t;N;<^`W*9wJa
zLl<w%hf8FpCt9(D`$$1DQOfnfDztKbEnlDcf6(nickB4dM7ie+Ag-Xm-<0zrcctc4
z+IlrJF8?BQJIhNJO$VaYErlp_sSn_{QKD4&B&8EFOj176mL3F*9}1B6H=qh-WJ?@v
zCTYSZix%?5o&hx4!7HSj3AO1(PEiwb_DD)S$;aXRd$Ba5*hTPsQ4eFqRb~5$O8NeO
zc>Bw!IGQhd7#%{e;1Jv;Sa5eja1ZY8?(Xg$JV21(K1lGuKyY_wa37pO^Um|TYu*1{
z>)!kEeZO{f_o}Yir>bk8bN1mTG?y=0*vPUXRTJ(aECdX^E5eNV0s;NJ<?!FmBWaGm
ztZw`EMnd`W4dw~sB|KAfPe=t8-}D5xu5GRP+@tSwin#<GV+GwttQkLm#X81$qh3qi
zDk`9%9{In85T4P`pXVid9DAuJo-2DjQ8&~70UTb7{sFATV6K?=Fj*Zjnu}nc=HtzD
zKk|~lVOj7^9&g=WDp72GA7l3;28FYN3KiQa(^7N>QK(YIei(H<?;o&Rn);4nX!aFE
z6O4-PWu@mQbi=s)PEWTtvWsAtl#Z<zzSm!4^_p3SH8HDhw(R>!MXl=l1>hx#(i$f?
z%TpYUx080{2902T4n0l49S~>AS)|)TX~<jK+<eB3PqIidIfmvs<d7uw-Ov@qOU?__
zF04;oAeY7dg?tS2A&kM`>w9U&CAo8qO8(9Y9Ul7fmZz^j=VCq4Mmj1c^MtUmMRy!`
zQhs>U_Pc{T57z(uP<X7!rHTWGd`;p~8T|)<fAnpoh*WA{-U{(&gvNBWrkmFK)3n^w
z<?sa#dK+bpZ`3vo2B7)^^ZFr$E-Q5-?4tw$U>i<0fAR1GxpiE=;&sDbEY`d!dk9&4
zs#&+?rnUob%%R@la8sYRBEEAP%_CP)tBfcO!DO_%-voTKvILT9Q2HvF0f(x#%lEwU
z3}}(J^smKqCOpPh9?w138g9YgvBVa~q}*3Kp$?}vQ6Es`+A*V3Bb0?`-dUg-r4$nB
z66Iat!oVPgE{b%9G*o?R%S(K@VjPq1<)$>_ZqWAY#%?taO<&2yC{M~P?g4hRN5$K)
zj+!#>C?9SewCZ+@GOV*KFenl=aGlI-527^WJ6PBvp|Wf=YH~)+B!N{s(X5inMD4L@
zSfkY`i|0P}8UdDYj5mvu*5_Dfbh8qU>!-rOwUR%n9C^{WqUtrQr@}rX(pZ8{^LkYM
zPK4Ks+5~)Sw)p|=kNU+Y<qW1F7zvzkwwL^uw#oOB&Esp}2s}#~I6&ux$#JS5Ghz!v
zl>HoI)`903xP@zJfLo_fZC$T$u+YpGAS4A;J(o%`_7ZXgdFwdL$Tp1e`kOQv@y@WW
zOIAvI6kC@~1$^o7O1G)+&TU+V)y!!E$E~1O%{Zi5W|{wR&;py}JTs4Ijm%jgALKs<
zb1Ig6tRI);2$D)i&8Pn^fzM;tq&m+dQZJYOM9(3_T*9rZ{+m+P=_M@FlLK!DGxl(F
z_mAh}3NjD_?vWAo&E|zYW{+`-krDl>9=2JEy}QQ+Lk7qyR+XZ9oC;%c9OmlmF~&Bt
z7CP=5By8Jjme$`r0yhcJ7p#C%=(!85ie_W>t)aG2{BUsqx`CD87(V4iqos){6WC^H
zFE#SfWz}kJPxqJZA(2=1S^2$9=^6_z4WfV*)y=2Tmk6AJi<)JDWA0*|MXiO~K!GpX
zlChD5*^HU0O7kCSzjuWK4-w_5+r49?2ut*oy3QDm+kXQAo^<m60IA80dj^Rgt%?*x
zMc+Q&pI)B3o?%Qo2Ej)C*%BH}w;_0x+Qe6T<D+-RzHu*9w#ILC4N-_X%%Eo2hpAJ1
zLgijFiL<Dd<RCN+LoQQFhAk@Rzv><uH(Echj&SbBY4FC>e-_DAY%lW|ldK4~;$-p4
zE-|;tHhT<H#TtZvw<i&f@XGK}h-BY0VKq<v<SZ#va=YA|kxD~@!=T39Tcc@oq*6wG
z3ugh<hJ!rImE7cPnRM`z=h8vxI~^-SKgVO0&sPXg`L!)jVyB4%iD~mQQm_&<mh&#`
zDA^ohe{q>%$|rf6ru1|3$|z6FUCJ8Qthc6Y>X{jT0&1a`V5S?UlQ(7FLt{u{&-9Sz
zwan&E%g4>@Q;W5!<<Fg69yEtu$AFNnd!AWDe$+U=^%$j3xDRaKXvOYOwXx!`{bgOn
z{Z_D~QN)6gf%1#W?G8(Y$<1v$g`~nuy^eOlZs(}}v50#B4nCbzJK%)7voq+6xuW>e
z5L-X*<_rAlKS1;vQH#DxgZ9Gi6|pOf0=d_^?>UFXtx)+cJC*m)8wrVdOr>H@b<vLd
z^W-iEN#dG0LQS(sS5#t_M3)Sg1`oZLo#UDFGRaOQJNu_BGX!lOaUIbuqokT<-dnpV
z!uAAzPUFKk>`+>35SzRd+8_&jEi_Vc?&Ks-{+q8Q>FkeKt2nw0&HUdv(xPoYJh@o&
zG#O*edpY&CM@C+_f6{pFVlQX=>8U_@3JLxJke~f8&3BwZ^^tip&Cdw_#)UYChMZ3J
zV^$FRlO<Oe$42lXBw!u)*u2&nbk?#0&TxoPX(nL_TQzM=&H?1`J%3;6jaibOTlZf0
z2QY2eyk^cK<&8P#^>-{7V_o9ct%+KrIBSq=RmmWVGu{h7nD~Cfpvc~3&-3nBw#X-{
zzr}g5L#<%xD=ir!-Vx*MNwX<!(-~y%?w!iFi5})r>}t$8F7hBXPZmne%h8GyGvdDN
zXpsXPD&&41Xbw~sMz9qUi;@MP|Bqi|DTikA+N>`P4*f^3<-k0oEka5#*99gd&l6~U
zMi5ze@9()KIOp`5dD{9?jpkN1@>=Yns=J*<FrW`ptz3yewL8m1U6qR3eeJxj7O9)E
z3fmC6M@7r)_daVXT|F314$?R>$+%&`Nl=&kDl!a6SWUYvn;r#)Kb0rG``~T(p>p9<
zuI4CtY4S`2yB&7P@w;L!cchp-`GvogvUN%dg`@o`P&oswo2zXlA?~wMdXvGVwO|<W
z<3KSKnD8QI)CnUgHfq*l`Djm@o;x7=CqX02;NIG{#Hca!L+56PyR>WAzj&L)(fQXY
z550(Xb=pDZi8uyum|-y1h5ecZM5GD0%ogkIxj<Hha<4BL&rJ{kvdEzJu#&9w8cCzZ
zmJr2;%dyS8LT<|m6;KsnY-V%~Ybg^LV_fh|FSGG2@l~gF&E<|9&ES9JSkXlr8~k-`
z_c$MZA(hJ;N^D@YsOBv4`Ge4mVfaU?zPMG)|GIv-gJ}&*AHbi0acs|~&2A1{HCcB_
zcRghe0Wppbo~<|ES+XKF3RC?Q;!t>r(hg#^xxr{h`yiAg_>V3%h1gDcL4VhH7Y(DJ
zVP}m}yhN@1XN#&n0~V7RbTq7ryvwDMXzHeMWT%Rl$h09kBm}qR(<P)Ia(C|$gzD<l
z`eU<xnHSU0Ad^*LQ+SmP&o=tOdGJI?nzUc8ztB)#$NQ)RAn(a`Qkr4l_q4#y#!n;A
zl*S;!0~nP#*xGtb?NM8l<+r8kpn;X2((ITWWx3}SM6ycQ>#e)nq_!IqbQOofc8$`U
z(07G#l+&ord#2+Y>Nyqz!b?~=>+MD%_)cJIE7lq`9^8|lR9iujf~M3nO~ihzQe^oX
zk7?-4&fBo?m(`8(s6w$x@vwBo*O)uS{9hrrZqRsSn?EiXduBPvApJrAZEf%%lH`Cf
zi>u{-#AW>zmE#MC!fIe>w4(#3Sw5eQl~LQg5FRP}S!AKZaE7l0@XBaL>bYYMAqBaP
zUj{;G4lD9NiK?h&4I=}M1{KySY6#jp4hB2XV`PCrx0t>%Ug*aZE%EFWg7BP~z}FG%
zV4`)b7Fa)J4e+S1CaslEoA<d4CO*pJ)5NP<U^6NHkVprrF?uJI(&OXE5juU%WlSbF
zvBY;z7W#}2Kb@aQXeJ)%<qu=3*5sgJtC*n3WEfK=y?nNp=kaX$>HuTd{3%39dM+A`
z)QcK7Ia~c;@m>sjYrhn})2+`+s{hPGlSY>07<4Sey70#s(z#!C47*P{YaOH`(h^W(
z{(BQLS;#9%=a5~SG7OMnqQO5eql@VOLYqQY>rp=4AryY1ornj6z~APR+6mfyWHk%q
ze&jpk+PhFWaG(EB6J7kf(i(zKmcv#>gaO2L+<v43zSfq8c@eiZ4f@zHa{jyvdfplw
zeOe9c_E1Y6Cf#$s@0Rzk(?Ho3AqzhA){BPK$Q5Vc*<ICj{>S6CAK7QcfiqSxslxn^
zASMRWD4x?R{DTf&<+C-Ehrn}QNH=wp1ofD8s#VfurlVWAa@Nh>kV6(V_IplhTs{^q
zVuqgpNfh?#LwU~+pucz(L9BSq7oFrktA~GvvS!Ou4_j$V_A<!g`o*YU<tm<=e9g_e
z!AaANg*OgeKr&KXW$qH75KAQlUK*qQw{QnusqBgk7>6$5&z+%fG{Z=4vF6^MdjIRY
zo|;ZB=eIFOo@h(Qt@hehrGmCi{K8eorHE=JSO8bu;Ee;jQl8Mg?=;9dG3Z6Ay*1d1
z)v>C@@S*t`{qY&|Y|)-Mo_j2ya1$S;Euwzaa-wA<197jCL!574kM3fEJ;u982pro3
zqd6n}5F{Eg+R8H)WYg=_Do2{zKQQ7;@z&6~D#p9I(w-FPO4lo4kM*+ZLFFdlIIh;}
zhyb80KQnbGM|uB$i${PpH0O!{Y&8M1#trR5kla#p-of>%(Bf0$x_5-O=(H3tGu&OT
zMo7e0pM`y*RpRFCV8Nu(d3iX8FblT(tg6JciH;k1vrg+foo>!p_EG@{Ez`|_D`k46
z@k$B$Ca+ZnFpljTp=JK?Ks!*+yq~Y=WpugYSh2{5r_4jIQiI9cP%|Q9RKsdcs`>Gw
zjG%`S8ZyefK9j8cFE_~!_%|C*Ha0VNnyi3}kD%hqFUVZjh48^J-_+HOeGjFB)hMjv
z>up(}oPMP%yVdH&rv}4CXyTz$N5Nsf*V*g2=nQh9_eA!6HBZ7zY+8vUv^Wc4wLa!S
zdf0GDWZ${L=)RI*UTD$QJ0PC<cs3-J$zRX3&>jeaThB*hdG9X9A7cW?`?p+P6YbB4
zCNXcN$8|@|Ppu;++2h|D8#qcrr1|p$P107L3xv#?F~7eVWxJXgofZDMQVu=br+?YD
z-0)I8?s}sLM$z|&Cbgd5MVsH}nbZ;WVjLK9LVWJ~?|PgbwRQuPdudlDX7HAP593;)
zOrN9X9VuEQ%-5qlzxUmsjf82gvBhLb56ZV6Y<i(?E}>tu(2cXM;QP+V<J&ki$~=E*
z);lYP$7)r;3IK9ESiLjF0(IYV<jlHDx_qrKu^CMXsYAsjUg30P{*hOXT%0p|?L9I+
z7pacPP^evn4RkSxCsU%wH{Oor%KH1hYkSZQP1_Qz2;Bnw$CN;0f4APlc4xBEW&OC3
z%WKUjPY8{Dy}}NpWv!@f8KUn{<e|{XO0uI_%<yRgKZY(|@stpkZUr1MgyuTPrf=Bx
zp!`?w!k=v4wA<00ExWAGD()f`CR3)G&O61BCs$|(ltPYKHDQ1GZS|&O?uv@6Xx3SJ
z5Cg`kpOI&Uk!*cx#dYbNoR#So)C^U&qVkVn#ZrtY`^=VK^O>CChUm+|ii1$XALrX?
z(RhcXK^CMLG0@L2x}h*B59JDV2XWky4rH*T0`ld}S(=Epg$V5o`s;(6*6;H!INWtF
zo}+Godem_fmN|5AcC9+0&4cgXgP3y~T1%WS64Sdc{EN3tyKA-LFZii|N)56=WuQD4
z3+GwMefHsM%~k(<_xCdUusjY7iGk@FouvPGO{;-ObP&!6oWaM>f{&aD843)u$W1vB
zwzN`B*CL`UzAEg+>U8tpNYi*iZm(j}=%$QZY7dRmy}bPG#FzS$`^z>0jG;<bB)b?>
z=XCuW{vwy7FnUe!siQRv#!ux*_(%XddD!e&%=Vjf>PY^AY~upsOn^Frbyp3sncJ2V
zoEFVDwMEcKkF$drgFGJlMx)kv%V)eK67FZRf+?~jxi<ac1_j<6prf1BAF5nYi>WMX
zdwHR0h}7dpZH)9d8u3PZ=}WKjFN?b!q5767H*zg{miU@v@054$EP`S`d&=;7l+$YV
z1xXGH3GiFJKc>ktU`=LtNuQf@9!k#&5P}^!mj>!XVT<I+0qoG$d9M#%KQB)AUvHZ@
z4$y5a+rS@R=d+q#GwWdH)sI^9)}w0HWz)rBmY(xl9v?I4lx#1Stf0*2G&aU!txmUd
zN=$9hakg!XyhD%s%u_W+H$fKmtxfN6`B%^!uVtO$vLY_?l@#Cp$^jfJgCZS60k&QS
zjO#ph;<Q~vNveEx<*c}4)#c+>uKy&3H(2h5!5Fbko6n!esg74eHIRv;TPO?_AU4}A
zA~v4b%2`dovKoHF^E!0sfjc+ep}v59L6bw$uyS|lCIQ*5?-r+5#{%9YGvc(<bB|P?
zZ7011-$O*<%J}onr1^S&&!-DeuQOS1htoz~XX}6bT8^*Lx0T&gyL9SNr(?5)n+OVC
z*H&>+^sRLKxAR<cVP>;xk0qkJ9jmroVu=sjC*PUT7r86hC`<W7V`+E*7$5NEmgYGW
z<;Fx&V%ZTGmzzolaMPo1202Q8RG9atr}}Bh8(HiwITpgCdNKbf<n4YxF9r^}Zaw7R
zcr0J&Gsl$VP+xb{Ki3_G_4%&S>pPT5HC+M6yPZ3k@f8rjv^uOhqMpSlixt}qk4;{6
z6GHaq`gF{$@tSE;xig##r1}kmbTj7pwp^tOQ5qAEZsssAR<@OLI>ls*R3}^*!0OD~
zW#3qWzLIIo^GuyW@pkk{hvldthVPHJt1BLV(iYyu$0bTFZI@#s(32VqlFFg*U=!+V
zo$i{_YGiXUmc21Pg!wOm%#P3%1Qscev2edNo@(FuIL6hn8a0H44L2pi--4B>lBqY_
zol+kNWiEvMuWiWTZ#B#byZ4TRO-{jH#KC+~0Z)Zxa3v=)<Wg%rHvC@oC9jP5y$v@2
zlHPAZkc?aV0))aSilF8{jE|ntN#rNJ`rQF!woG%uH^t^}F+sw8q5lBYuUi!KCsAa8
z{qTp;96JiwEzh&HJg*GNe;*MCn{JcL=HZKk{f-h=Lf#0i%QG<AhZZ5mLAPTH`-Oej
z55ZkqC)kKB4r{l66Q}ypveQJ^>61Cu7@syyFO2(hY%_ZEv2OtX0KU5-Agla<?#?@P
zhG+~u;-zl+<l5Dq*W=8>MqDDx&V&8@2z~^RKW=#FemVK|4(xfzW{&MMHpf`dAoL;X
z(v1Jjbm`)CujJITzR*vz@Jx8<mGr?r<EGW@@4d<xsJ=Gf`>WH!4MZOJ8l4;%+zfV^
zeZn}P-(Mtn7!rQ0c-pAG2=33Ze#WF=j=mFiCl`o%iMj0jeR_O1U%q&1EP6X8IvcYK
zGR`URJ|l7J6jq^it+>8CLS=Z`us_`V--C&7OTtaJ7r?I+TuW2AL58N+nxul-4db?)
zQ$fb?Lt6no{g06DhyCnUMH)T`Xbdi`?lI~4LrJvW-9wq~0Y&Kz+=(0b0)X-EZ**)B
z(h;e@i={s#Zh+st-oKJ_THpnnBS;%PwY$x3A)44~p&b7x&<-}cWq6Qi1Pc9M-p>on
z!Dd!mOIhDA&V80SSCK1-hDjTkUu~r{5J{`+p46INW3tb&AKTCEAuY>P6$4~8^N;~;
zjY~B`kGi6)5&1E%;P;Aa1xa_F05@!_3*(%^zDigzD47vbQJO>_^W?a~+R2WKU!SRZ
zjE7@G$pi#Rq0hHpyxFFGPk#<!FjU`lXd8EbU0fGM$R$JEnm_N17VYBgoCj|zB4%$V
z@u<Vvox25>{bK2c{sH=uvCbf-(PHCI4PwTm^8AD;!WvOz)eXpc=`8(Q4<Ecm_#UYi
zL78Wfbw_VS;0)mte?0yL19phvbvK=UQr62r9tMmtP~Q^?U$CJfmYPdSdTY<!)0M!u
z+xbH!r?u4pG^no;y1+VVLv%_$ISRNNqU*yw^{N;1)2zM49JkH)cve<%v!j})pwb@g
zX_nD~&E(Q%hW#@p$5{N2=q}=m=s!tBfq_AB{PUiWrQ$%R76?SXdF$mU`|v5*D)8%)
zhpUV0WOV$FT}h&h<4|;1Aq=vC$67GR5*xwD@}+{>x*omj(?SAyQt){aW_=Q|=3U29
zZ#UD_J+=8;vUYhqW6_!Su&Q+Qlk43T<P%l2ACzJ?!@BTwIQbtyV*2lq|Eu3YEx<ae
z98~fE>%iUr@Eeo8BUW*^m51d{xc_wGv%^rR-yQfo$NUg1uzT|=)3!KhV7A-C3cCuk
z)fyNLvilGimHW`}&gDInVOPgv44p&uu2XyF?R&M@vd>#>o?(LS$!hhZC$V`w1|zIx
zs+(!-o?Pq|(UZdUPPOQ*AXMgmCQvO(G3=ZCBzR*@e0OtDnDrD;SPiqB&oKP|&lK-O
z0-tmFa8GZYS>+ys%g45iZZu(oy0w|(Sbr=$?EE&(=jSAJwqoMCDT=hBO`P<i5$Ke0
z8~lrD+Go9h-?+f(lC){8X?%lX;W(4Wc_|Dw@H#<lOiwf~BU-*EVhDq%Pn_o)15#mF
z$c~4Xrc*RyqVc*Gf^Qjj+dUm>VOlv-dS+Hm*)^ZG6&T`Z6s6--c=s6Ci>4I1@tJjs
zw<V%6aO=9BVM2Biu1iL~has#~Kaj<;kp><L0$$X6DU5qRT}0MB_Y<u<kr}o-U3o22
zUS9JI9q$xBsm&l#EXRawE*;)R0$&_xU=cH{ey^Mizsak?Bq%($JDL)Pm-SIki$7Dn
zNYbnqtYXv!V?3Gmgu(61cJ+#+Dqo@whEW^9ih|pEfA9e|;^cnaZr<*LlP2hR?-NkS
zKe-bdGNp;rkIl-`R#~E@(?#Iyqj>mY>)8rxvmU|MDP-ZK%a7oXh`TNsBAF^Ee2qGS
z|A`S<Gyj!wKOF6PDuZ(dg|6&ZAmI#ruaQ`S$I=%^h8_=9Y!8{j-fX|eMgq<;Smf8A
zw_VhiXq|X2Y{K|~w`erF^4V+($x-cf9-MN!X5w@``h-s>=VU3LE3iuGs)~ypg#-l9
z&GJKvqZ!#%aO5p=g-wuE6$&%sX)nh_P{40@FRlTWUG>Sy53nqnXbL=*hk_x+6~_qJ
zap%lvZPn4AkE~dE*pSm%0EfDrO(4rc{g=LJA=3uyJM*vwv}O8GZfSzyo{wka@w`LJ
z+dcO|it)on1kJ7>Kf`YdZOGJJwb-(8RQ4(KF=pRQ;xweAbR+(F>w1lqf&nEuw_ZC<
z`Ewt2M~h_LrBzUHWGy)TJob(j9S`^VHHw7usJ>_C=Vud;Vr8YUrpIC#x_l-5ER^-P
z7v)*Tb3Hh=H+)T{CF?xusH29&(|eO(a@rB$kQ!vtIP%Un-?0m#@Ykaf=Tl5EqGj=d
zD4{{((R#f#G}^4te%bI>!^XW_Yo6q3nMWQ^(h#)(ux#+r+fo)d62-6^8@(T9@cxKE
zTTsqcO3C0>c;Husywh*(c^7Fgjxw2$T&apmW+)k(2E+JMl;%K@c)-lElnTMeoJmq~
zFA19}#a2Hv&t)u2nqJJ)!pOwN6{Ug~CyACdsTLtIx3l)1H}0u;+xol{NM{`l>-`E#
zJJ2w_xQC@RhZi1U9~Syet7AnXp-;bBbG=8DI+;m*H__YajB4Ev^jfaLAJgo;SIP>}
zVP4dot97>Sa>_pJXP*5T=RTapWZ^EoI~r!-DouqoAj9;la`wr%Os{G|E`qZaft%y#
z`3to|6|==k7q{efzj8bA4;9+djO3BfkE2N%l~B5LNf-Jp&Tz_Gf#IgQvG_)kIY?mn
z%1B{oe)968zPQ`?YA4^0x-HwY*0{KzPI2$q11$_P_d@6ekA?0M!kajC?XAP+P4ZoI
zy5OcETWr}d+w9BwS)dL}!;5xTJwuIGk@xtrKVF()*UCXWsaFMCY`S?<pDlYpb}iPJ
zZK@o0PS`sd_KOtlkH2gZH3j}UE0rz{;MtlYcop#*%FJ5Y_6Lj?QTGa`f;^?=$>eFj
zAGlpd&_l{q5-mQ=Q|p<N^~<efa~>p*O2!=V{Wc~EY`muk4Gzf>TNTP(@(mD%MqfES
zuYJx0`RcN?ElzCQC&vkK1nAHk^C@Looc5nb-N1BO^9<LXHFI^_hBHXlXt|0aG9Fv)
zi?8!q3JHo|;}%$zZrsoIGtXL-QaLp<Hw00`E`iEXV>UTxoYpzG2Ysx;V>wvqq+f}m
zaJ2KIzP~FsQ!i0@KLSFd)&}wn$ymyjrQHG1GgQe-OXWYDgj35;(4S~@<MHKi(*<gB
zU0Tdn=9&!G`Dr;h371}3L=YaN?!_mHl`{w9@T!2VcdB<c4y~XImj@#*LG--(b_Zve
zgCZ$C7v0UBLFDIn-~hR<RxPVdonJcV)9cs4_UVEn{<@A{zOxnz`PMCZG+`sGGj>gW
zv}<WstC-QxRMb{uTkb4*D9)YW?)#wYCOy{nhlx0)Tt<ATQ*5K+cSnwZ-#_!Lb6EM!
z5<sNC;c`u~xH6Br5&K`i4Dyopv>3%REvSrEo*Z?QEmWPy%_c~zUUx{4k>Kd;b5JRx
zXlZ`n{sdF68e%@Hu|-lyoQq8AWPXgGG2?<Kn(b9stA|ziYIO!<4q`r;ycju<&+7AL
zS=S0NHuhqQVyp!boVV0FT5hFGz<gYZ6>aPOT<`q_n#@jft;KEP=FGZTEGD8R$l|uF
zfU+%b`Ezb(=bbhe6lY5y|4k4R$RX^q3zGjUBiK;w0%*ZL0qiqLw{OFC>I$}rkofox
zrRh_%7`swB>G1v&$;_Ctc7<O1tc6|6*J|u*9(c0T3_sEl<y=0h484*AH55i$JWf4_
zS$Qp-SOIc7v^^B{u!?lcMk^{sP$c`U40T8_*BkwPp|w}E8O)8XB@}JLYJV{U4317L
zZ08Ld^Vn?t{MQo{INHlr*&1K=kZMsVDAaVDMQ0J4UZXW^vHCNMw4&rYRUA(DsmU4s
z8rks`qGjEXT~nXanviAH5GPEmnVQ5f^Yf=nLKb|R199X+YK90G5U*;SZj9u6<qr@#
z+BhZ5hErqY;W6t;j3k>Dbw+xI&)>7zxp=raB*;gyGP7jmLGe67Vj_IRXf>s!%?#<d
zT-c_4;pouh4p?b-u>h7<T~-JzncW$+f5Ez8nOVrY34hG7bxF`Lcf`C&w>1n>XxpM!
zi(4F~7%x(wh&_=n=-{G|!HQgx#9$hcdO`3~nO;Y7p_t<e%d|P_`XvvO$1cg?v=}2j
zqRe7N7LViKjfpXh!ZM4a2E5MzWAh-@He3o!YosK~*oDVci6xA4GgXb6F~MmZzp^XQ
z?=gj!+X;p#UVS5^kwc-*CI~I+h%@MrY=2x46U+|d!h*@&Oh0=MD22vf!NID>Fg=Gm
z@ydTLbbLb$&}zTuy^prJ2=_X7JoVM%mECwSn$V8~-rHYawt=jg^FGC$<=HmtnWl>E
zjU~t@d+*}Z>2Es(_f<1?)MMLr+ykx8`SO1(JBo>30EfdaALq_zd$eBD6w^s5I31vt
zeau4`w-P>LL04JIOD8{qJXBe589d!$DU_KI8O!Z(DiZl!9DaP2tEe!b4nt`PU61(T
zHb378Za^`WbkE6Qkd_8;x;aZ{W72I+QowuVS+gdCl6UcUy6c}a&wB1ltlqGOSE>sx
zyP36w=4`At3bBgYjMSVI5Q+mjHBWgVjlqlOO=q0tLzX~_LSb<R2%Gon7(&Nf>lQkT
zmp>USPN&almH&J(hu0VAaL+)$emR<oSLJ~v#X5y{+`BOv=kN_B^-UHfF224ZS3+~2
zzQdOyw@mY$?&uKVtu4tM@;(}YU5qpcF4qrhZm5_*5cNx3Aa=j3*$-T-fRzW-{I0dh
zK8jz)Yxa#m*zf6skZF2L9v?4Dbl3%n&8Mv_3u_BS3jCU^ALh0ZylGA}2pecx&|bfw
zu}<@-&vLeue|s^8JU3lru6_=B>lYE!sP}qi*Q5_Y2v%0okj~iH-db;B(nMv(eM=SI
z1X%@2J(Qa=+1|>PJ^rR^RH6uyMUTtiXyvrebLML4rIHqa;i$7|87muiGX+qPEnG@7
zb5t^htH?j+6bfb~Xfwi!@zHM!oju#Pt)bhm^gfrXF?U)vwdVW-V2OSTU~4-g%^?x@
zRLW~@CF$L@yz|#T&!cBGtYf&38Z+xC8||v~iZ9iJ3BlujT0N9ODB!oEcz4>tX6pqB
z#UAfuL#AxM`23~>`RFAi$-d!U?$DAsMW?xAmWdE&u_nnOn}txmKS+9{_**=IP}|Lq
zg>mOCKf`>7_!X;8Qd0Le-|YS;EvYs=25xdyik-Vgg5>Ig;PJqZydgQB_X4`v5mh+T
zIgH$HMqIO8t{V~1&daUA%{hmIIejq<SBH&%0Jp8nb>oPwusm4YiQ8u8=UI{xo4P!*
zJA`AGO{+msp9W(-2X#vc=S~)5rZtN=i$<Gt<{{K)V^BPLE9kSMRO^skW1jJVNM)O@
zRMiykq$5m2evJ{P6_3&&ep53yz(K^d50THNqC<|yJ9aBT(~Ve-dXKD1Ba<ZUh=@k0
zEP}(V(@0B48vWrTAHX<>R%R24)ODJmz(;)Pi{uxQN3G)|6zyMYoE9!)S*(|hL!A6O
zSbguY^ga!Kbd{9Ff@OJR7t;dZXq%qUe-hT*fFp1ebAAFy(@bIjfJ)kFxVRgIBC`Uv
zV-;yF!W2Alth#d4=9hdek4OAYWm4J<i+Un$!CV)k<0hXc<uA+a$2;jL@B-|4-@~{*
z-87}Ac<mp+WFw!dueO@k1X{)7An`(SP1>(xEhHkah2iVxj@Omb*};JId#FP%=DpgW
zEw9k%MWJous#L*>RkJRA4p+&eRt|2d;zk}5zv<&czXE#eIn*BrThNE>8uTs1eNh$_
z919#X=jacN`4W~z9bzxVd5vBwwsR0~Y5b7E3Bnxj0;U*1qr2B?<qasvzo|`y!?9UC
zeoMN21l7xrigQ!?gyiN9rMS*yWGHvWT3b!hrZpC#$fbRS9=OCukorQ67c#eE^Tc4`
zh@T3tNXRPsu9B{y<^<b4gvFVr`J$F-*HPDPjvAff)E=kfFHD@wT#y)~$qeV_u_hT?
zl77OsK&GA6;8|gR97-fB3*(X<-*;?d$T>87@}1)pAoRj#@oruc4xv=JkZ~-0*C{-2
zEbDBq<2ob8jS@|*`XY2e(aK7<7n>$Qv!c3bx0F-0egW#v9RWp|@1wNCgy-oDRCJ|^
z<zu2BQv|cW`*gKBZa*FjzVLLu(Y?X*dL0Y7H$LMwYl#gMEa<g5IBq<Cw(|0}v6gAb
zBJJ$of1B1fB~V)NCt*GBI8VB3)^%8{Uq0tWagQuppru&rb{<(G7#y=Ke^A=@TP|}D
z*<#&zE*rz^SThSTx(CuNJd>6@OFoN@zn%C1`#1%HIZ<XLuJ`{om-(pIWWVMdGuD-B
z9GgZroyma*X`)CuY3;EcRt1ZR3Dgf2@o`a~GLt-^K-qpxONs2yEHo)RK}>0eTuj!S
zHmVGJa=zvmtttsZm{6AtSfHrC>8gGDj%GF3dJy$=XKB#?)QrW=c_O<>ul4&%ot(Zv
zQ`_D*5Z!Cr`WhGOl3f=IFTY90ff2B#sU~w}MuOKf50}^OZq26k?J>`!#lY=84BT{f
zKc(xe>$<{nL)idXEpFY!m*LR;q*wdbpq*Wj&IbtMWK1hUB((C-QVuiSU!_bh9o#_U
z&r~(f+m2X8kxn;|)7~-Qjq8_sd?#foE>MrK@Rdw3D1%Ud)yYpqRMrxEWO!;6mSYR2
zFo3HqRz&|I1<m$)dR!Ay*{ud^u0R7{TW?oC*FFBGf8lR?Q752<I^sX)S)GYtoa=wW
zp98hx+mz|lJ$n{xSw@V65iLbO+)&~3ck`lV?z~P}c2Yu>nsM)k_F<_mXRX6*?R-e*
z-w0_}Yv-IRy<-|kT6yyXHQcPQvDrsnQVO$swVOviv4)h*KcdiSX~;3eNtDj2m6od0
z0V6DNj{Aw;`5;)#MFVkV+%46xN;kf<=J4m{@FX*SKvPGFnRAwod4TS>c#m(Ctmgk(
zC0gvX?D2dJ-r^fy^W0?`qP}dM=w>5P>aeWapo3)5_0g@^Ci&*uW|*ck4Y7K-_E}}U
zPkm4Gqz7pNnRTA=GJG0MM^78FY63ASs#B*kkqBphLUAk3|EOM^0SJ*`Wo(pU{{xpq
zgPYMqa!HW=|A;#H|Brl9R^G}F^SM~Lwpzb6%2jRMD%^TeDm9}mD`($L=ND$Nq}|+v
z+vadSst9rX_@0_^6Ps=1D!|3Yf^3vXjGJtcLYnn1V_1}j{{xbM3uB_t(u`yCXf5+^
z>J-gljpO|^gSE+sO2rHTmTxIUj|pcZ4v=ipkBSpn%4j(WKh`6*^2P^ld{^XhbULxO
zrK0xV<C{HcUV57w8^uWAbJJkC8rvS$rr0gu(-)3$oMDo;{>bZ6{buGyzV&YO@+i>w
zD6U7nIx~n3R9oc2aeIGqKK@=4xeK(=`S(Kc9ZS#~Qc|J{BI9+|w8UWZ`CYgH4^7*&
zZvsOc!WG#LX}#mF6WJG`RRL>qQk&Zfn1>IKH$1s*u$}9MblGNIa5viY<2rwIaQ)m6
zrJZr+?^mw9K#d^sp7|?FO3_feBqqV$z-uZUa&l4redlujFGBS}L>Bb{3!$&?mhAGk
zn*b0*MhWGYiWgxIrfq0cJM^le5p{nnRriZ?dE(|hQ<LCp4SbM9mNmw6qmsvLxltgy
z?nZ!w0jRe<cEDroRJG6A^YtmRxp&tryYpMPo;i5PK+D?9g}l$K$Dm<%cQC&T2oD}d
z`g{L*vzlFOKq$N16LPm;Pgs?LjgAn9g+XQMxlP_n%;8SiZ;XZMe7jRKIE|%`G&2)d
zf=w7Rh@rD1Vrb&8dTNX1;&-cQTWsJAZMwg_Z${#Hp>*daXOR{x-sa(@VJMb`&Ty7g
z(C*6&%)v2R>9d2Qv%V+25UR6+;@J0g_^EG^FN>o<5I{TDd-7pN17ESQ{B8TOEjnY&
z(<;z%<eNqeTR~m++8j;5H5|B?t{T%|HD@sTgHP^$UkZn6bWLwmnm<c$n%Hjvj4mWz
zD!+dKiVuNLK(}fBO3>7noY>PGo2k$Jr9Jrc0yr@gel>peRPl7ee)lj#dBFwapPM1Q
zDjlx%kBG^28RCPY<Zn;3VzoSE_v^-k)jc6D6zay@rtwnS0Ag3or{jjZ`q5Z~mNz<k
zlf#$f`xZoU;p1rP-5$J_vEaeC)gRsRg>_k&!GZ5E7m0(mH;GWi*zSB8W^4&0rWU*A
zPJe+m3z<-!gKcb|Lw}J7+lrqiN1uGwM*iFJuqk&pXL=@978fA4347F%BK%<4XAa#a
z0SZaa;8MSv14AIi;!kt>55AGBVjc`2*{wd&4hPY|TUjwZ`{&T=M;EffkEW%8xTocy
zX$v07=*Btyo3U*e)l0$l^KMXjT51<FaIcg@J33`*UoyT4yW%MxveFwaS@)$4slMVH
ze;w#ILXU==5G6@6*&hQ@_}Xmmp>9{rk>MSTAR_la3Ex~*qPKcn30`utL-qq_NiN8I
z^>)w$uOKU&VgrUNe*<1f{>=P$(}t)9@;|`d#CxGPOsw<6abkY(g)4}lqP!z=3aTt<
zni(D~75btMeh>rylm-Y`A}=jm1{-TO0~z%~rV0uS@T-U-R1#&<Zz{`pT(rdM(lPtR
zCL}|U#5>ExkBI=24@*LO)Pfv;i*7wCMp+#vbE2Hl0fVwTQ6Yi1=_|(C<spLq014rg
zq-_t}Fkx8i>Ij4EPH_1Z@g?b4h*?w!dT&&*v+)Tq|AB}d?1FJNSe+<i^y2YlCqzp1
z$EWIabC6*Au9(3q96D2#HQRH*w5{_!tM0ns`Qi|9QjzgQ@`0IH7Lx0?<=b)19GNAF
zFWw6llM)@m+ez*J0JSppD}T1P4ebYRh57xUJeo2evxB`ks$#)!OJMq_$90FZ_@!;3
zo4tXcKp~N9ooGCt0Jyf%REEPRBK6cB-N0CrbszWcum>}~EmFVCz0kNW(WHzoKP8^W
zA&Z}MpPb+`^-_N?4R(kHtQc@;MsG-$e@FiZShm&zKWUM4`s@*Xck<w$zYVZsub7hh
z4zrL6QT0(F+R2A&D4P!ypR!Db^lxGrKfkCQI4(d0QW>!ud`~p#z3bK;^rHzICvp;l
zEGbxtYYNQ<@e9lxhU9aV-q4&#`7>`Y?wL6z2|F*DL^~2h18?Md`zC^C&JAo2{BI@<
zH-mz&U-7~}E(bgx;vgB|vi}2I1t(40?LG#K=9L$QK1IN+=iy-*CKjCbpV>8n-;6m+
zU>cFNce5~P|NKe;LF-O?w{+cugzkPNW{e<9-ExU7R*cBla8EHO8-5oRAV6?=-bOX<
z%@4)$IkDT6*H|pqMk)yO(1N}Bv7;gdNj<9YH*t^=%l2!V-y%IWw#cjcX&LBnOEUbf
ztoJvuh~S=x`;j0MNcJgE^cZGQXYhi>5>Io+w?2)>S@I8SB{tMo5qMN_HNr@@Gq8@P
zbCaG2M4;Yn6USV1qPma;b$`SxGXE1zP71yJpS=Vt;V44C@R9RP|MtA9bMF;>I;r)D
z$Fe)b#AF`@!Di06>OC*n&T$Gt)YgOOB1z=$i|)wkB+_cgka4GnT>usT(ztiu-Gb>M
zq^w|e*M$QEzVH%>{SsXWQlR(CBS!lvQSeoCmaqaeynkKsqsO?o0B~57;zEA53iDo|
zJQgD7c0jF%Dg3IV7#kk!U#iN&9%i}s|1Pd`W*x5ex}<P0Ej{HWatg08fpZWtf;z^9
z+G{nH_uFUKeLP3kIKL7{wy*XDtP+9=2k-~zDdmCyxL<DF!Fss~AD%3-lS)5S4*xp3
zqY4{!)6$=^`jsXp<(J0&qdX%TVI_9hVvoz51pzu^5Vd<1I%FK+Wk68SDQliZBvgt+
z1hu&ZD=SD7hC3s=imQ<OBz5!YXzY*}Z`&o{uX7id9}u6QV8em4?-AAlAIkk|<}WYW
zcQ-WqE3gp1s;btHrZ13q1$nT*lwE$6#9cl=Qz)udgV^YV-3K!A_T_|$Kh7JCs;a+M
zQra~Q?7$rag<wyu=<h2DJt?V1zcE$W1fOK{l+|zhoV?2`-2q%8d|A-=2RJ5^C9l#}
zd#yAvHO;=c*ZAt=wn3qIeXB`aRrpF%De}Fp)L(6FjTO<9PdUXy^}0s9{14CP`n{$S
zq_-NRq=Pz6%7?-48{DNLD7$JynZoJ&2I041zk(JrC}2k*FA+t&ao<<CZ@t&a$V!3{
zk6M%zhv@Q{(3*<Pd5~T(`^~+0_e;qDUsH018O1Dd*0Yv|^CpPFC5F6b#CVSkcV8zW
z_y-T${r~P)%zwVUfknAf{sS-tRHY)_#QpM8>qvZF!S5068B+{S@kF_<k$94TX~@})
zCO$9QN_$_&J~>s)1k;Kd9mBf*f0}G$i5@8`$c-2~RG_A0L6M@GvhT%GG(Wdy%)9;r
zEO-eECRN4A4*L*>1ys@xsmUi%TxqHW#(D{hjSE)Ic;nu>_+Q#Qs;cZiCaQ-2H@FWg
zw4&Trw}O#`yXA1VG)c@5ZIz7o7uRtK#J2KGKK}qGiH{FJA3OXuvtb<jig|nR6!H8w
zyEqTAA9tkA1sVkvm_7mD?kp#gc*y2}(P=gANcdBvw(4`oQa9vPD1YAE=Tk_MZ_6sd
z{{U4qYezu7NDop}N~B~o`}$>pTMfU8axu9R=>SHaZ)EVV!_WS?&JtCGW`B^|3$9@a
z4w}S2^C%9bWMRb`t4PJxqe)-UHlvL-Y<_%e1Q%ii(RzNVV(gubOMO;bK;OEQ3~*nB
zNYtlJfy8(SuZG0PkVz|5K3IGx&3k366l?jU`g#QYy(Yx*S7rK{Do7Ol?l_4C@ss6;
zvL6jW78*bP0id6%sE(i|AlACSKRb|!FF*ab@*ZvXW1d{#emEA9Yx@WI*dh_)Dtl?>
zy=B=ByG(o03$Ky9np^R9D>Jh>dFiY%#ur@wPG|>+pr*X`h^abM_gxbUHZy=bZkMTO
zK^h#J7o87}6dJ^(1(=5{(iG6D%Dz*%aK55>ZxEi&_WmQLrE!MjTZr6$*gw(nGm?)^
zT6F#zJRm$&VaZ3c$aej(Mf<2p^-5U@Kebm^@~n2YNgPSk=;mHYuKX^gDuU((TuISl
zD1Bcho;Q7Zd#Bbr8a(x&4SQs6n>Yi8Q~lr<j&nQ01L3!j{~=6fV($N7|DkCi3RWdS
zW#XmerDs%BGe%6-&iGtcwb!4(9P@xY2nKG7eU%I|%>Os$8DBtCE2RGeH{<_l2;Kw{
z+6lj=^$v@Tmj8dger{+u`5R983uQ2HKR2KnD=20D=l->>DvWm6J+&0e_6BD<R7;f#
zuXR8<1FGnU9*5NYZ4G5#Pnu!-Kb}pstPDMQws)VbCM`7i8mA=Mkf~Dp_LI!~!q5%;
zPPOE+PnXixf(}!dzb}u}5=2Ek8hB6|HGu+G#YD!$j`t6sQ%EDlrup&wA0Xw9<i4zm
z1_n#CP2!}g5ZQYd(4TysXRu#oIsD_48e4(5oLv?~;>bMF719)jo3eeLg}$NWHsKgM
znpXHd65r0F-Tf2O?!}{S`-$^6+P(dac%osaUJf44ylA#}SfbBAZ*68sBepT-Ac9B8
z5hlx~@xO2Q@2)hR4-H$};5Nx5wuP2Ihx|gOfn@*v(OAqpi5}S;uZDTG1l%O&nkggo
zC~s?1i%ag|r2q(=4mIXqL)aGZ<0voA?;v*Z@|V%+SQh~BRJEut;6$MFsrbOw^$jWO
z@B?9R{03-tnPMOSLl5x@+=pM(8Z+?!t?u*xI%>urlFI(&V&rDY<j7;lW@pV$4RJ7H
z4-2T{UT@aHO0bCHN;QdCPro9|$s}wj{(3$A<2p%ne}jSvWRqply27IFFzeV<vsf)s
zY?(CI<8a^ZX4S_j#Pd@j4*5+HYRXx5A%zyq%HL*&5Oc^2&#5uuG9IU)I+lcN^=IvR
z{9Cr<Fe!F7oY3@dsy{AYk2i4|%=Duq?8?7s+_{yQk#}fS&aW-tH*3m_fGezrRi$na
zDh6y(ID_QBB(i^35ij7%i_p_&YZJ7x6%Y{NASUHJhPg)C0<56NcF;z%74+N=FE%St
zq8E{60fLH^#E506<BVaZ|KGRf(H7&j{avT7TKv5J`yMv_vwJy?)mn9LhkFIuSbSf#
zowx<wxsy(gCbxS($9<<}qs-^~$L?KVUf{Ve(boVs=CqwGize9&YGm4#v)T9V)HwV7
zSK~RPt^;4P&`6ensJ3QZrd1rND6+xW7L`9G`Ga!h2BHkFtaL@cwB;I%6(|gN>7Yv<
z2;Z(<yZ(sfy-kr9JZgtqUJ3|CT)OglXnDcz5Ji&l+-O+$78~#%+EL6qAgnd+F5A@3
z=%cqj*tjv_QcBRVT-gJOA{V8xQ1GXI1i^pp8u1sf%4*R{tj3Q$Sh8uSU3b-u$$A#%
zq+O$Kl@P@3yb-lcqJcxCK|JNAT<{9)k?)Qr{QnP+|NC3I_Kx*EXa4M4t)gyz;_|LR
ztfQ(UPk3~4ki47jUkPSsxu3A3*ol5ET13toI+}3(^CtuO+9?EaP;Qt%($E;N_GPl4
zCSyP*v9AW;)IV<`BUtz`8Ctx(8QxVn|KR()1!VJ1VyVH``YydgJ)pDcw}RMR<ovA>
zpJR7d6=G$yC8-(&?<~XSm`fqp=Hor_F(kRheXhQ?F7TnyH*E(#P$sDrasd;l)}4l1
z`uml<kG$)8Q?1>%`YbB1<Jo-|lvPT(zg3?_JC;nQ`@5~UEg9dgK3Mq$I|e9-514!t
z_G%Py*ERG&;@B<0wg9vyNHI%!8&CF&{sWXN|D|F6a~Tz0|E&?P>sL!#WlYn1hx4H7
z!=M12_|J{)5B`xc`hvb}O{d#}4;N+=(s=O9r+3FvUQEdRjBpJbaVg+T6Ou@1qZ=lQ
zodtP&rP&etAhZNenmTmaFT3o=Tcn9u#rmI0LZn@TJjDPHsmH59Nj}l`53RftyN^^P
z`o<yhU!QLP`%Brp@#?}lTFrMm0l2Ue?B`G~s%3y7sud+QK$tBb*kAx3cpilQ#j4g`
z{xvvW`1q&i=dHjV1mPbdUXFzrG}*`ieV53lu`fLiuKg|66Gt}gjA|U?;JGNoF8%n=
z@A>`nw|QIIPaVNos5T3qtCt!AuWq^aHlB@q*Rr?HJbOCBsSv-3Ac+X=7LQ3Cz*2wx
zj#HTd$#f75A&?8e**nQk?<Mm4w3zwnaxvLIfT7E20rBA=!cfpW&km_1zZ-ZcNXho4
zxa)P^r|am9o!`<RPq6aZ>2jR}RCwYqqQ9cDPagDGeps$Oq@FcLNBYr0D7X>ay;(H6
z9ynfRLf-vUp{V0>_q0Em-Bd^i0i0oNimhJEiL6@Z3Bzhx2wSs0&35n+bB07gR&VA7
z5}UWmHEX+ccEt}blCj5tp+iyhe-p|_I`4}rTu`5cL^J;r(ytyW%YMB%)wp(gt*?&O
z#=b?=!nO{%dC<NoAiH5JFz~QWci{m*;NURMZpK*p-_ck_#*BrU1-O&CZC7ko!xHRh
zaz-xk1eZo60Id@clb){vmH}-sn>%+=kBGjUIlLO6CU%Djy!yr=;X%&h_=DOTE&yD2
zt>e$09Kaqv$yx|POQ+iW&tM9sSEF{zal%9?9q<sr;08>PNcJpBSO4<?k<Ql#037DD
zehRYwXH*NI#s_P)K$p)hF-gALJBf-aS@_;=nn>nR=B|_LPWYn_Wboh6S!%Ck;aSBm
zgK1LBgfpu{PS;HE*>diSf^*;TqC5LLp>E!jIAfzPj<pk7#y_Sh@JS?v;&ULqN3xxs
zifkTtcJxLF_W>LX4dzDP)Je4;!y=+rqJtaJol}s$U62k)-=ja*v7pcj5FStfd{3*o
z+rFNch<&+LwK09d>TbrGYJKTwIzvQ`HPKlTz1nfF@?Hk@eGx)EC)>&Z{!Vs!wjk^C
zT71w3&E$UPIaZS>FK0jbvnST#wVsdh4z8}d{Q=N8g659#q!~M;*hM7e;zsW6!OBYl
z%ejooF?R9uT}Lq;`zoC<r53zWXF)xRA<+;S?s>h$N@iaB1Vn3=T`Z{6H<Y68uV{jA
z7?J_J=R>GZK|J4ey2M`iUPsrC^U7Gyxls3sYHz^J1K>nagzh9@o7s8(o728dL}f8M
zxrjK~$9pL5x5Gs@tU`g+1JV~ytaG@7zvMa;<|D#FkU5@r7|CR>rYpdJCNSt$S5u4+
z@#-G(H`N4t+I=dinC3p`X^7o!7m+;ZgzO@F_gv3*uxbJy)kZMta<*TX%@wX;na^_Y
z7gq#n)!QlUklOfEVREL%!Epn8=PZ1uapNZSJK-BZ3&cjOA6AuRx7<Mb<50a|2dr$&
zIy2ehvHfeZv(6>fe+61jg1jg5w_R_B(hBL$Z@2OzILF1*K?;ZV5KHc%9;p3SwgU&Q
zT05xSD@k!;w8Smd<PL;w{-U=e+W@=~T^I!S-KUDEiN}jRmb=7l`y?Q**NzzD@*kk&
zVxDzqTljV1W&irj)KIkjC&)dzluuuHW1Rr|U^-P1!B+3MPB?7opx)1QSTu;S;Xv0D
zvf(n#L&kZEjT+7%4vGroAS75q`{ry59-<TIA-~hScKO51x_OVfN0TW&#4!o8k|!7V
zgzSQZH|I#=vbd~$J>nlXKMDQ!j)7=SBqHK2Xf?b&T1yJHP53j-`jo;i9U`ut6FVjb
z*4aee_R9x94WdmSi!x6;Nbec|7Dvuob(wuMyp!Jt+)HYsry|Mj$SE9LH3+C)D-heB
zNzZ#J0o?<VNZu?^T62j^A9akDVX(oim|irbYJdv!ob&hQTETBuRDG+5J*wIZ`pDG?
zi!W476Y)N2n^G8^$Vg{oJrX~crMIc!GX4NOcTHMu9MX`*zWete4QSOvewzIeGeAoA
zu{oWdObYd2QoAGrpcZZX-VaC~txW*{^u5~Mhk4K_oq0_}+n;k&P6+c0v0gDC`2_X7
z$*XIokj)?5Wgon^;!6l^u$kd249kv2las{0VoQaGc=M9z$o>OpPSt;K*PJY`(&0yP
z;nml+#_*UojY}%A+$8Vyw*@s`V-@K7n8Y?gFpaw-lGp~mF6?3R8iJ3X^m5I^RYOGm
zYoDkcIze-{&8|-|x?S+S@0Fr`{G`I#DUeOYR9O3X$-u`H{xb7?&os-{e{(<n*DA9)
z>G2O>p+4-j`F_Pay8PH)vBN3(8-{S|myRQZdzoI9t8S$-QVJDVSmNiTn8swy&ez&6
zq?jR3@4qd1*a%W{M1>dAnErvOy<NdCkP*ie4SC#{#2*S@A`Aq0KFh4mx4>a+AX3<w
zZ>ySL=Y=huS{==ilSW5%|6k0#RZty4+b+6rcZY=scXxLUZo%CGEF1y^U$}d4f-Qmv
zcU`zkaCd^c<lFpv|L0<#n{#pM)IN7J-BUGH)7?|u&-;iWAw%(&Pn}rX;R1k@`~gRP
z^UQVcWPg!RBS0pLOCbOgv3tuaZ@FdlAHeVAn5>@p3-fZj_EM}M<`p4JqH}|vx@iVY
zUfm%EGI8S*7Q~-v)zb&;I`@a_JJTV{{pAq;$|lS@1}Sh*os0b7a#UY9zmydS*VfX3
znO||Fy=VZtEoAKbbbTI4EF&f;5d>jrKX%13OPKWiEz$-W@cDoODyr))u@qm|*bmbV
zsUP4b@hmnek{|1Ee)xVc;jaobqW@?2pFIvPtU28OzPTHH#*!E>3orf{h&k~GF?Sak
zteZsH|MjW+J#Ql^3lAL^kbhua-6eLF<Wn9{jFj7K!t5=tPS+xye5M*H>m^0ViJZwZ
zp!71%RbuS781ee<MId+PLa9yi#GmlF@?96WmTfcCWX++zNIfi%?K3sjVneineQ~}M
z&^X;A-W<~DaVj4s9WxbMG~#*pO%B2m)-PdsHjRW0LurL$8F<-@X4mUluseOtMVS3q
z3u=0HzpEz{=!T|0UiGv;pxN^u&a~kN2zhII0Lj8pEI!oGG6z|$B)&)9BbJH>k&@cI
zU0;}K7Mf@OkM3LlpDJ~+Qxxfs&29n)zHu8|AH&%cu95gG+TPcQV{hrU=Xnc{t+41~
z_M#~C!JLOzIlQwz3Vy+UC~?f(xarYgGe12$Yotc<M$N^*;uaZS7%hPEpyr$Hz+`$C
z55ztGX{8|pt32(Ter2sjSFmjRF(pdaYL{8H(IcAmaiC=-3hJ{{5V-)pbY~=Rd1_(P
zTiRk5v_=|&=62eu!a}5tszQo)abVC0+=&HO#G(z#>#iK4*JP5@@Fz-nm@8!$Ah(g9
ziQ6R5pGLLZ9{U=wW2}_rIhgpevN$v15sma*#OZQTXFENhl$cBxo8Z&7_Z5dcSiIFW
zsi`;K?)T>xgY5?#$^Mw*^VmeQD%D=Ntt!gDP%zvi++(Ka{A7k?MizEY%C|-luS=L|
zDMUZNzzQ5hprZGhn<SD{)6=8to9fkoa=w9&tR9lw833^jPRAHuJL(h65@Y%w#gxz@
z|4JIhcJB84DRDE#Q>6v}4buvF{1Fwh02-cIorXgurR)LRqcIERUS@8_>`*H%Qd`mA
zHB$72nw=XFDuH;K6a2q1{d5a-hYak7oG+-)hWRT7?k-^E&5#Kk7*tBrW)^3$K5`n<
zqW&nig+&dzmPMN=b$%hF%IKg*-%!S@={S-RV*1XH;|3#?;mqxJC%kWn;5H&o;<mHK
zejHBRdnJe|N<(XohWXFf|Mz57<5g`aYac2nMkv_;H0ic(ttZ&9s#yHoPJFZvZB251
z__cn>*D_|NbWaYuBdpO4hWu;b+AnclM4ICZAjsdsj0==_&Lvmg5bdXmQ^kmCt7M0u
ziN9)F640bo(L`@+>l{jrV$ItzI#L&0jL(RK3gC-;ED*!){K|ayDRQ_I11W5&GZB62
zeJ_Ecz-Cj^o25s<W`~@W)Z62rmayX2e&;|xk(;-yBu53}Ur8UYgt*<<pJ@_=^Y>sf
z!IyqkSzr-*KC1Oq2_Jn&Xi0(%pN83BDW{f>QWlTH^5`e?vTi1E8RQ1I8O+2WQZPSH
zvN&20R!&zY>xJ1<c0Kw^p28aZmu0f2g#5l?9F|Qcx?~OJPyUSHgt}00md2oRzm{yM
z1vSly(<!kmL@`2K$F4cl9(lss)Qy}3<h2tx2~YxP*ZNtR`D~5g-I-)dF<Ji77#WIl
zAVQ<5!i`>J)f&;Zl0a{!K9Madnd3YvNKX#(8kNc+@v+@ZPVu+-V?lCa2>L!MNF5E-
zDaG`((zwiY`BQzmy)I_*%g{g0ako>@@x8r+a7mxT`EKKc(us(1<s>0U=D234)eUYw
zTb}JBv^!RPP(IxM`lcG|B!^=<fhpW{(CbUx>KK!qYe3BjwK?Y1G<9>^k-5b<G*T|?
zdG=~FNh>x~(uM3V)T@Tn8bG@>w&&&?x*d~a^@lG2!lZ^H2bG$Z?DpPEio)U|i*MXX
zip@k<I)>SKWXl?%Nx}haELJy6t*j%NNOE+_$QR=5C4}i=B?jQ3zm!kyba0*n_FSr5
zj=X>ILVJoCTh~3dhS#9?Qs)oRZTDg~vv%K@eRH`+Tb&McTAy%s-djS(Q*{Me6NW7!
z{P-MtCvT3H{{uL>-Wo{%Q*(9dRf^^Wu5cd({r!t6Q8m^}p4UH=cMkb`cO;|G>ky%B
zAArU2_qD@kBc>zw3pA6Qk#nMsTJ8Bo=HCy3@2BK@1t+_v{9v_>0C3nQQeM`NzH1EA
zS7;#*h@x)XcXgL&Mgh#)Gy;1Co*xu0?1^MXe8%)4a|#G|fphxEF4=PKWy=s9H3yab
zaz8j9h9D<gO|2NaIyo$Z>fX)=dAc9kP8Vxv3wMT2GT2}DpnW><k}_3|#{ZTcF}`;C
zxNadnUe3BB7UwpQ`Xc!(UKJwKh$oYKjP(ngwYqit1v^bqJ)!eVU+$WnEn48TusvLM
zg)5jfEK5zeP_2*m6c`l%J5&3r%EO+6Vv)P41jh41A3si|RLfruszYi{W?qPHPe-@F
zVnO*VgfpE_!mxzEhNiy_qAsMt)`U|_&3|3;IHR}d%b(;qwpVA<uSFR*oMep8wmY{z
z?|HZB(B`DFC|;R<(@6DPtz<h<-K{)5QJYgI&Zn_kPt8l%;m%Q4?Kc7r{+@`tO=#!g
zw18i0_;A)7&P=Q?LH{Vg;S5@|FH*4cj6#^*aoQyru%Ccvac-6HI$?dWHK*Eq?i<F@
z3>pi}-tSC$S)#Ux_KIW0y1%6D=rHW8t<3^u^O2N}!snz-4OXCZk^xh|vlI_lJnria
zA9M^r)iwUU(%HDK$b7-sZtD?#(~CG6qt@#0UZwVynB;7wUJ(2BQ<*Urg7mC-GI6Lc
zJcN>)1R&-+0=J!=luaVRL&;B)C>UlA_KgZq$?+H5>#rHHr;H;1F861eT>u*`K`P1Y
zPLizu1^Lg{e3|d{tM=1j{E1oK&^uAQ?x^$e{`X%3FqlF|nZUN-t8q@mLjBlQAd~7<
zFQ2hnNl%GSt(sG^=_Vmet7}v6SI}H9tT3`&|Mo(M5I&nxmQyr&%%M(z%)`bGWI^~U
z%JuJfM6d_&z1_Z71u>BM?}r4k-Xg3_a2pO&b#gP|)YUoIFKTsNCJKqg6}`5i`StHy
z`lnsvF3$3Rrz*ICAZhCKfm)IHa7nm38elN&B;27{_j{;2xyWUo8ni&;a%*_ya-Q=q
z9y&DJ*=$jlhyo_P;koT>LgZUhQ=G`F-(6p5c+08C1RWHrTV1m~{>x`ws>)}}_f&_A
z>u>#i(X}&(aFr6Shhq_RbUah`$lhu(V7!$;tOo5(lvs3o0CnKI$$1ufkKYeo^|mGX
z`d;ig|8;FX8c2{7u&jl$#tN1F9*AT&zz(KeE}-x?a|*^ZG1@$rbOU+~VDkurh*3K#
zTOS_Jl5CpRdQ4}TpRMwE*|(Ih=PE(>pyqdvxpl8r<dqT(@%+4<LfvC5%S2sM@|-Qr
z%clE{@gqb-b58h_MUD!=gUd?&?zhAh4OyQ^@3Fma2Hew9Q(HE9ZDGc%rB#DK_u*VM
z#cQ7}A`Q48>JRI^?w8<Trw;16OBl4~C(!p2_vuLwq+8wSiB{X;_zhu|<+`)YhZYKc
zw5xF+7;I-bw7BWSM3@06_IUB|)t|2&oXZ+rEm&m8M`PO`CqNlT-pLp3{HQw72cPA@
z(0*aGRaXFo@Vx;UBLtR4<eadD3TETmNxt8@kDcwJ!A!DpJ*&`8#ppbA;H>zLrzFc8
z@__@VQ1{NrJiqBftFGhMe)7@Bhl~qUi=R3qAe|g`1)~SUr8*$CFLysNSyucvKCXWK
zUFQSp`(GPNJmZL{`fjohiD(a$su7<;oa6pT&E4S7haZ1LPV0X`Y!PiKC>Ygd67#2h
zL|ULP1|TfCS*AroAZM0M9D-pE1c&PexvWupl|}f+XE`R`+(1)?bQ-+D;hQVf{qfP_
zlu=o?<uS6}g7Ezh_!?S-mK6&t3Evup&S^K8a=y3^V?<VnhvRsHQ$oahWvWTlZVnpI
zKEtptR7?QVjZPX3j1-&va$o?JN}^fOifcwwYSGHwdh|>G-q|L{6}&r3q)$j7Pp%vP
z;8=~%>|kla(1gd>;qyVc`%WkJw-HgvM;xt6=5OT9(1U|KF_aGj2g<-I1YZTEs`N)t
zVsu)xKWQI9R&9h!81&VtbZjHE8q1MS(xt!OV#sjk6~qI}%-8b|B5_L1a+3X??US~E
z;NF-}1@(NLJ1uInF=J~-m}#W4oa(J-b1@5##uLZJy<N9*3H=}jg38E!h2NfEGpv#j
z?v}S)mRPn|pWoG0wCi3&Zu7`M>lFE0-|)oAm$A^7?BsWjEg=)Kmcd~mmnlhPnE(13
z2@Hc>h%8ycimBc9Y-z?|%mMZTlJV1(T8lVQ|J%Mi9vR+TkR}9ejqUGf97nw80Hxh-
zY%HYN^Io0uT#)kiMDINI0&4{siS3)H+2#uPN$OD3>obu-lj;s1b7TYpc4@WlZZpJk
zenP-v2!Y##xucp77ROWDa7D=j&n}NqzA*U!ds9o7w77%__wFwX%B{A97-tUMCqZ1d
z+DQLHGIiT5rlFT@k*^*slg6^T6OJ0V%uZU@eRE&IqHyDSkZ>g>hc3p#Rtdq{!Iap0
zR-u#rPosbJE9o6Q?w$9Um=qHze^F~YDuPqvS>NoFQ4h-YH={TQ4d5`4hsk$>3q=Ov
zHDrOE3`j-y=uUlC&Kn-Pu{geiy}U0$c0XbLGj@?sppidE!_ZwcnFGEmq%AN$J<ZNF
zSh0SdS-JOrOEJwRou1Al82XhqtIMvDoco8M7{9Q=ALlldSPZC5snG{Ut+U*pMNopd
zH8<)^=tRS19uzXIh!UjQSJbqChLAvMdb2{GWWy-XCD*`&_zLjG(HM^uUPqK}vJ{?@
zF4VS#J%LYN`puVT`$gN&C8zgxCzyuEi;|xBmsmAf?B>Cgy*D4(Fiv6rWV$!??E&wg
zv&26LdPR``YZd<AM)ZFh-v4tqer<Cf5x=eOwcmrZsV^WSeLcYxWfQI)*dsCb`0s=V
zdd|S<Hyq|W^M0f&Iw^9TSE`D>n#`)IwLf0n-v-pFAw=upKGEX}v-AR!SzMNUGuCc>
znR#}Mr3y;lm8XO`k95zxe9MXTkI!T+;`!yI?d850%_2F;sSU=9@IY+SnJbwq`x?BQ
zJ@t}P*u!&G$}RGHnZR6Z^d3xyT*Z{V8ukOq!LpKgSw4FS#LesoGI6+ToxWl)qQh*Z
zm{k-!m2)k~XP8l>!m)$mbb3s5luFSgLLuoozmSwmlXHkqkX!lEe~vZ%{7oZ=ATLEo
z+)sp}a_ao?+pv~4_%(7or`p+!UJb`zh>y+`@4nm@`eTX{5a~`Ho5T~OQ;xpOwYy?j
zPjgam{8gHuN>>iqrb+08B;TkcW&sId)GK_>GEuv7;apcMNXs2XIFcsF#n#d}hPpF%
z?Eo*8t1NaiGm+ppEQu7`gV!S#W6sHx@t*qb?74F}gU69O=W`UKmTRuJ<wSab&st<i
ztC<gNxKseLs9FvS|KT*|z``PZGR7PTYg=42;?5)8VuPw%X*8sCbYwZio5Nk<$Rrp!
zfAe!g5d+0BBvG98zK|A^%E~z8*`UAgmYq0L^p<pnw8lfne3nl;sG_*CaI&hI6K}{d
zKNg9lkhBP-rauF@tDpVHd5Bj#zi-sWS<I?YUJ)KFlO@VjqIHO)f#B$n1?JwbnUGBO
z{tU}8Mtnr1Yy&>qX8EW2ENNely-)`Z**k7)B2b$0*KC(;m&R?HYu{y}1+!wZRVGj{
zK2RP9x>Wq;tUFM*jsTgq{7TnY%_C1@$~ZMl{Zl(iRWkLXRmWHclb5mdw!Zn$)>_7!
z*d8T1KT(oAD@=a-zBpl+eAXa1Du)m;Smy&ajAf5rc3g{U7nyyX9}OE|m0lc6C^<ed
zwYXk|i`JzX99_M>u`;1|=+agBNzjC!eX}=2T9fz86TUyT<l->kvBJZZk`uvfNJnpm
zn0S|Djn?dou5|cU!BL-}0`a<#m<X{j!_jutp15PteTw_)-Pmnj!(R-m(naLDioF`z
zOojbNWe8F&XOOS0&Wy7dqo?WU4f#2m8fpk}1*i9$(&%ib<dM&YbBirdljeBqBp9L1
z<i~Q74NlAuRbX-zT4<_nJpT&ahCv?i|9OS}v<S^USOqdu;|q$Nq?nr5qZqM_y<*(U
zl*DKgQsOvu8S5{#rNui^4|WQx2)e{->!935P~(73wSrjtgBwVGD%vsNP)%p`aGqGO
z!j3>en;jEI5xt~{L#J|b7|fNnNMd_Puf+^Q-{vuTZ$iE$POXe<Obuk&d@qUOx9sZ@
z>o`8;m=qJ~Tj8EBM<YFH1WN8<|EiP#w^M4N(Ue}R>EkKGMAd4@a%ahYW=2iJ#<%7D
zo#M+tsX|2`w~Ngn?Qe@7-jOs+mY|6eE<<?tx$N2z*Ab;wJ?%@30BjOqXqB`hdEv^r
zLN67|mzQIxToFrEm2a2|B&RmRw5~#<T%;W*zT|~~#UP<Yp*44Me1?#{OE7ou4hcvw
zGVWGlmlF*-m8>)6DJb|%_E}v;A*r5lqqhT5XvT7TuJevMgVWI(4M#ax<{m$&$;3iv
zu$x86pM=m&<?Q8YtaFE>)K&#ek@Trlh*Q|4O8X9qbJAw*`L2_jmE}vsxjRBRmc$&R
z4l86?&1;BW@nGZ+TsXTW=;35#LXPD)ECI?PWxH3x#9A164%D7ViU=j1f(+Uv9O41R
zX&g9D55M>retst8UCPvljFQ%luyK*VV5iXNL-go~O$q8xDKa%yBfBFcu7rb_Sfz18
zLuvdhN2O3wQnFX9d!{}_eLfb5alo9L$wqH?<BIQ*al|(}Q9k&$C6;X>C<{l}cUI(C
z-@xDi5zkbuT-us>f)Kvg{RcrOX1)4C>omG!zeQcjs)b#*ErP`Z@0^OrvpHXTn?V{S
zL*BM)*tmAa4CTqR=XqNi0W__eNtsHjL&4?Ir21cStHB|99Hoq3m=`hy?>R7H+Jr@n
zXRY&?W2aJc1+6|@%y1iP2zJ%kOsh16#m}7ks`U%_OMh#n4yhMF+;(Uc7f*;#l+aHg
zFS$vRnef_vOid`%Kc%1KHb`+al@Uj-FsJ0y{o)Wbu71(CqLN;_nQT!aOz_Ge$+oth
z*qziTebIMT)&7lY^ikI)$rB!Ux6;tq9eQs2N3@^MA#7GTL#3~QSyXDXZzY(kJ_Z@b
z-4V2q5cgO0`(VY6+%kZ}tm2uUy)2B4nWoK1-EI!=jEUoh)mcuU@|e{2(vIEis!p&T
zWz}D2k>t1(v{Y%TPmN&He=K^A8k#7YvXrm8TfZA*E96#0qovOmAv!l$a(nZ}&d}l$
z!To^q-3e!h!bM#ZMCN$upyHB-eUQc&2ar6X_a<tV_M9+yy0Gn0k>Y;|S%wiSr+TW>
zGSZ0-Hi_37Bvg1U3dJ%z=`EGwpEzFM=oONYlt(eYW#z#Z(bUSMB`GBjhkNfuAJ^@T
z$+C@<J>n`BlKWb&#1IWB&|~k6Yt!@2V@o=lwhJ1noJAj!TzA}S2}Y_BDw2PMTawk`
z=-|o8*9->~`zgoushNmVXYhmB2+`;FFLRWXMD6!G?eVZio(Jk^rK8woeOR%^u%c?z
zXKgmPj2xv|v;tT%H*?qvEPY^7ZN(5(etp$h|FQ=Y7$aO2;Ys)j(mryDeUHrKe@d>D
zqReqv@#`H4F0<mCR7Z`u7{r^(#GsHiojygF_9J~NLnaJ4AF=3Xznnw6Ve~XC+|DwF
zH1bMFB;nhNxF%7Mm#{u;kJRtgP#rlowSIZpSP+L!s%JyKMQ{^tV8mQG9c-q~B5pPx
z19n*vONj(8S1~#LLD^0lH7tp<VJ5wD49y5ILkTs!UQuvxxcv+&fkm!I9fJcPD<PHB
z2LghCpk7UXJd|y6UhIqOzm`Cz+Qr00M*zI5c~2q*FU#&YOOg4s8HEHk7bvDI%jx6t
z^w-cF_CSxt?)`tkUZfPFVw#q|008&a9<-~fSdfUKjOOibH|lHDX}c)?)nx}qsLGq8
zs51tpXzt?{uDng3h%xtGhN02p2*-H-yAk>niMg*n>0B84arAGo<?fQggn*`L7L0+w
zkQ%{5W|8xXz3#U_ZtbQ!gr3cabK!{LUh+&F!YX>|TIKF^BpNdlY=wyG>Ic`gb&5$=
zbVDwOrSt)CW+Y6y>wE2)pba$YA(5Z8?XsI;xV7|Vack)Tv<-K1IOsmPY4Tty>)-JH
z+5VgG@bXmeAl+*K=lH!!f?>q{E3k2VYo!3k(v=hlr4x~P$!&Vh>#km{H`)6G49F^*
zxp}v<OZ33=y*lRI{~<HZu>ej4h}rl|fBie%`9)^@z9-K0Y#m+gs7$vkq-XKER1h9#
z+Pf(I3HfT9^oyG#?GKSl$kM%AjhDaXn*wXRJ%sj~f!E*h3Ryv0LNdEmi%oLQO9G#^
z$esSy<L$FPt5(F^2T%0BU)^EQ*1Rr+u1FUc<VHxlyslln_eLtZE^FGrZ0jBDG#G?~
z(&@b8MxMZ`z6`-I(4DdWpy%A)tq-!xjP#!<OF!TJRnT-V_f}xDXh!mNcdp0S!-RaF
z`Scg$L%5p4dCz&8-k2N=P}PnkG}1eN%-q;Ykn|Aq5uw+8y1TCx>9DC<L0s7K9MeS7
zoLuDAwTEuk6!!?^c9xa*jHR~ownFj1d6w(fHda<`np}~mIh;cs<gcfv#+)vi?3+DG
z%@)nZN9r&NE!&;$k7Yk6*1D!Soo%Oyn9vQ!o(zt<UJB>Q4W01~`)BDC)n8lH<>3=R
zBw&fYnWGLlIrZ*Y8e7%Z5m!bjkLR2ZKj5=0NcB^|706}Or$m^=Z%j8I>1O&pi2G?4
zauHC?aVR^pH=5eZpT(0P_;v^@qelZ1B|Z6J9Ut+0q?;-&NsA65C=|K_EU>4N>mI0S
zAJ+|`2>(8?ugEX<&4`vk)Agdke?!eGRsG}?&n~R`StnatNcxt9ug#CSuuHrsgDWYW
z$K5SUb`wA*CVIQU`NQm=9f}LQ0By)LR|486EgV%T&`6Hk_r*+GE<Nj?YP(#rkU?Mq
z{oz&fDH9588Z+lyP2~YojBOzc0{n)G_R@`!Yz2*xN?u=H7TXV1h$(}Zy0lt<xCt^%
zZifH^N0uZfpl_6J1StZDE%bb%(h6vaH06bE<?ixBfQ;MbS_xXZLoQ}>%-i{9kaWiv
z<}$|A+IuAu<=JS%-=k6{l42Uv)XIXz`I-ULdbUNA$<^g(u~w3q1xfjFUs;4KFzICt
zp$AyWbkV#oc)>7X)O69|@yB0?oK<r4S@{O{+k8-z#>&8hxLHvlaM9uFgi{rFR#>r5
z=|(v0N4UBb5vFpC#(#xGqm51*8VtGX6v!FN7e1x_Q|&f>3RMr1w*>Ma{VZf&`@KsN
zA?01p++9PSfe0AFj-k!v-5n&v6tX8ObLk1>n^ygsVJ_sfSaFc5C{G_>ZE{Pjks~->
zV&6enqR%k-4yEBKZ4US9s^LHDL35pu-{zm%F-M#`RIdr}EW=Pf<ursj3#1YPxJ`4S
zzn6RqCTz)Soi_H{GrlS#Nf$9zc+))Fl!#4Q2h?0u<zl8^MEovu<PQrDYB{l-x|3Lp
zeK-3tpCJg}>V~!S{dvNW@3Q+|I}`A3HxFeWX8eid`{h3Dl6;qX*_#=KnBtXA5u<zg
zKKonP#@8iFSgG%Y585CMwwI}Z`%Q`HsC{pt6Np(=(<Wsc&)!royobVgs;9GWW%pat
zt102#mWnrGF<td(kMR}$i0;+AsxgPLd)=!S#)^T!7f!%};a1QeHdmyn936oW*Lh$%
zyl36nTE-7(hM!G=12{FP{UvvC^XPupmk*0JjByoxfZdq4&hZ%0dsASeaHYz7$Kw;;
zF0=8;vT;T6%l(?lQC6baB$`v<1^O@>asq=fi%EbaMj<Fvl(xpAOseLZT+k-)quyXT
znHAPUeNf)W?;A1eP(VJAACeyCH9U|J_ATP^_z(6nlolF^k2DUigVnGuP<N1ZklKvh
z1olO^JzXyGbX&&~y`EIV`kFZ3iB;4El=08haG%9PioBgWI|c@O;80YbzacH^p3~|w
zA_X~aZ{K^>3;c$^xxADq&P(v)Qi!g@n(op0Lq4!KH|UWi8WFVV*nt%&&SUACAesa5
zrZ~HvnoxW4O?VQH&usP2`d~7j25%>_H`8BvLo>Tky%ARrhXh)+YHQW+d-4BJm=<^t
zY?3Em{D2#cvIK}wqCR){mGxdD(iR+iNJuCz`{W1K#i#~9sN*z+k}?{Wg<&DO*sX>!
zzod^N4{0BC4Px>7srXx0*FMzxNbv?SEjSBur3m(-BFC|nzJ-w1yi%t9*=Az5Z0cBD
zoaBjxlxPm2D12MWU3%rBF?2tfN{{pg3q|7~9_@tZ*n_%&e$fiF#t-Qtmmk!Il5pG2
z!#Kp!QHV$$VLdq#D-Y$WtaX3<Haa}+^IcrhUAAWhOR@<OZWCg4Rk4NG*r@!!t}*5e
z@8~xSr3vd9)2~_>xS)A2MrPaqO3?uAIl9YF`58}~2P2L3>_=LX*fSU4fcsJeGf6?b
ztTJLB3*QRz9wtoFpWP=P3E8^{`s!|Q0&{DI=6FllaRQdh1dJ~jF{J6DAqx0+`Lh9{
zS`7qqK^4<%OZM$a3A_rqkgD){<zM@wT!2OO(i{a;k|W#`SeoKmp{%@mS@O6~*nbr&
zN3-vV8=I8tWsqZ-qvVv(?Lp|h{02y4A8ry{37F{>UAXrYerzEJ2>uGFX2c%JTv64~
z7dqZ<HFO(3tsmlo>544TYDE9EVx|W3SZG5~zR>c5P?%VF^L{XgiO?Rlc#|Ij_T>K&
zx=rSPm3-qiq%O7tw4x6cu%hXP1?gE1XWc{C#r*Y_iLH<G`etp3SUf-d4Jb`zK0ysp
zxbDoO8gvnMXrh-SEXc_u)IFMo#yAW0qbOzSrV}@NbFDs0X2pKWVAM?N_gpm}9%z+G
z+SZEOZ^E#kPy2*p)Lhg~oJ@@FxSd2EFLMYlNy=4o()$9M&HoQTS&U{FL@Lb>1LBgc
zBh%K7#3i7kje7HD0O?Z8=-D95%fl~nCoyH>mM&`iu9gyFhety-$8&s!o7ZJzQL#<k
zmc5HJ1@TfBA%<mSArd&}|DM3Lpvv3~0GRvkMIiTq52(67VJ(yd!+gVhb=^22_vJ?<
zf1Cp|GXx%NeTg!Jz+x3gQUO@tifIeOaE9fkO<yN%o3|~iSc0iYE-hK=@&EW@#h+^q
zcb!yKIb(5rZi^M!lR5ZU7UT05A4}^)O;RV?C4NZ%wD(_)ozfF(FWL*=%6c!buQvy`
zNQriIVUHuv?z!w`F~3ekka4&X(tk)pM6lfMz+KUuB1kLzrrWFvq&>{fs1<crgv>=z
zt<6>6c@_m8bFrT9s<7v(K!V3#gF(I=fnHeTlgXM1YFU)3d>GrY6YXM5)VlsApyF0t
z$xw5faAST8$B5ZcgXMS>CTg>dN`jO+h_y}0v!a%rQ(LIAdv+dwsyF<grSf-V2<>IB
zza`lnIu?v+SIRxhrzWmSUfv=LnKE9??Q3I^`B<HDrmr-;l|NJHQ`yZ7JqvbdvZuy$
z$4uiP5fOA~jAs?}U?um1Y{Op-Hg-&iO3NsEpDvUs@nqup_rv#zZ1><+NAiI@RCf6?
znDCzvCe)eQUxYwf6VEfVA*D>`kOP{GN(Y;Gev{-8W=*t((Ni(g9R9E)I`By;iS?+|
z4p(?<6uco1&B|ycQ`9#}j=s1nK%_J=Ak?qBwoKT6Pt_@jl9Mv5md~FCA5|AgK^Klq
zk6g7<&B3ACY2Yp=!`tGytrXP^lkcK@Ce5&OVPByjR)~*3u{`{<y4qmiQ>u0dg~%mx
z_hK+FI!jIW*W{`$!9V{2#IQmSc(LNB=Jn{-6I)6y5%Nx3I_5L<w->FQjGh3`{GCSa
zi|%)=rlv@bj57ade;FBO{eMsB{(l3?n!4WAb~J^o4gzWB?RsAI<iO1<*Uf5K<5EQI
zY*7z;qh87Xa?AvK?m@pyh+WZI@d^hz$t!kAwv>N&!C-IH+Qx<dGXOcar%oFOcxw5)
zIOIa#s6pPJUPOt`RMjW;R1lw<CHnL}rQcm;q%4K`IN*wR&ZEwStE!=oY0>m&{lb{n
zpsJZmh~vINSO|GdowIhoQjAo4W=yr<swctDQuH1$O6^TVoW?%zDEcYY>rIs;C3uP~
zOwi0?C(m5gxl@*}Z4?n7^QUm|oo3L;l`w69*iMP}BkqU3we(<*DD4eN9%Ucz_ltVU
zX!_*T*tnK2*A7vfLsaS^Qlxf$x#g4xPL}1T4(vgqoz(fk>j?x4D#Q%}A4kGZI3~U=
z0Ftjp-)FxY>{#KUKCKU`OF-D^m^aB{So{rdWE=%PSLSi@DJ|Z?%X;E2prXK9Ffpj9
zp`mYqWFb1F(>kd`PfBUH>JurMIJnQ@Tp2O#E_GjlzSr-b0+O~Qsrt0gHw8ozHa?f*
zp<3T`CZQtVuoHp`-7oMlUhOs-HHGvGsHwQR>F!<>zga5ac|SOJxF?1wGQ=Zbo$yaC
zIqIYHOa#xB)NTY5Y#XrxycR@e+%&{7oa4BguqDj7Bo;*@E>Eq%iaor6&;wTOyYCDB
zT`_mIyqxo~i}YA$MS7G+SHC+%SxfMDTL`O<(-sv%bgiTJkxxYxP;nho(u37LD>3IK
zi@z0%3K5rbj0y1}dZBO*w7YLo5A6Q2Rzm`lLR9vucK<Z)JE(D%>c!OE`REdXNHn+%
zLq+YGuB`N(ib15ggW<Uc#4?tsSxfc{S53P*MyD<XtQfdAC(8MXichJ+kWak0qs2#}
z;bRWR^UzfJK7O_(Ds5PWI&rVsVDeQ&5Zc8;3~A%>#R_=88MmYuw}8x?fgnpHt@`Zb
zB`=<il{6&E_{##WmRG5I1p(QfmW2}W)1<fwcRX#ndgds-IbHIxmQb-fd<goW8P{zX
z@uA0Xl6;Ug*TsiBzAXtS+r3YvP$BCC=QvC`CEhkM`}FR2M7q{gfj-UOi}JY&Jz=>D
z+P`F!rP5V;1--rte4dAX;!K(UtRTvPzus2Rzv=z_i!!o!%nys|JgryuCV@KcpWAdN
zDXBIr-{4Y{?e>nOTM@eBqMI4)JvnC89=qF!aVpfITq$jCRLd0wA^0GO#LnJPI4kE@
z+9?W(Q6p-HtZwPy2qptQDgtXSfvj@Me2TSDbn;2~LgT@3t-8Eh1ADk^FZF)_gx8gY
z<kvw4X&51tyKX)?KamaQF*(%H><3P`r6h=fnf;VXssq#F+rqaSlz-Ba!)^y*W^WEF
zqLJA<K#kL)PlDS4$DwDzx4j$Xhf&&+pUuWh+B`nee_v<5PyB8wFZovt;Yt;Vixaa#
zu!`gh9nPk4Ah%K{(0tL=Lxtd~A(!Wivql;6Y5imv!uKK7^4Ru0<mRSVhbK`Is&G#j
zLTUBzI1DnGmHAgelRhvy%@1Ji4`nnhqcVyABBinOd~3WI`SwZ4;<ROt|F^W+L{%B+
zo4$6sx{U+@A94mP((W#|8Z~mH@#sLKL5yUUx9%-+q~_02xxSe5I>k=T&Ft6|CG22e
z=poo>CqagOTF#OxAHcagkmR_R7;~a@B(oyaO<2WGa&V36M8{@O|KOAmogVJ!Kqml<
z^qeaxJ4IWVgJFp^RJ6sdD^MR`K`_=<GPvyE<_VSY4RU<vO2s>9U0DLFQTD|I|E9t*
zW^_n*BsmG=myhihoa>u|Bm13Hmdk>FhU#1+rGx*GE~1;CtFA({Ak1Tkom@uy69TJj
z>2K;T%{uWTweg_L#^dIiPBnVssI+J(iR%{5@|j(*&?6Z-)H`@oup*79KFBo&rX637
z6x9W*GFUMYXhuU2S=yB$gpB*RTF@js{8=LKi%cY`s~Eq(vbhxg>Ceofe|(TroMh*p
zufy=TSggpn`{g$`;H1=fyg4lpmBAjD^h}+1A;yHb*NnL6j11sYupHAr+B>umoi5_V
zhq~#alO-l$8GXToemb$r3dd3==>^}QQWW8I4b@7}TTEF#{h3;jikfoh^(T#D78Y)a
z42IH*#)H%wja(54e~z^5nCI|*3tFQ7f}(rgS&iOEH<*bo#IOVc-l6NDRf{botu#D>
ztkc0e;b;<VGBVV5it-$J@$^VF#N2qIGA^b;CPt(SMhqI}At_9T{%L5GM&1*v2?~+<
z^t$cZrwr`i1BfpLUpaN*LMclW$IgR^wep?h9P%GWXg(2SQ9(-wUR~h2<qKC8iewHw
zSbfP;&P?Ys4(iX`6vl)3CSzJY40fHOef601`ILwoW)9Y8%~f8dU1G)Tb9vcb3le~5
z89V27)fH<66^wEuX;j5vj^^;^Jha7t!ArRV=ZbXFs7lb^IryI8A)x^=NgBv+`2;c9
z${`e$!Ab{YO>-Lk`_3UeQO~c23zdKvyiysSk35v;fkvMB{abh2qX>C^#&KDI9mew)
z^s^{$jy#ej8Eaw?`)(T68%Km`t|UuAiV4L?Hc-VgTkj>j7OXWRMmJ&vqn>$Z-pBkb
z8mh!1&)8lQLz&)PovE#^#!RI{Bd0oeE*UPHk)_sy$%>3eKRqFno8A|fnAW!o7G+9?
ztFigR>Fm>`Ch#FS-H+iT%Q?mkD?+Sc<NRILyZc&-gzfl7Lr`3flg6GpT&mNX^cUlk
zF08qObG$JG-c_mQtX({v1coUD77?eELoKT{hnk)%l<6j=q4@oBaQh<4g9hi~>5%-k
zI1!Uos%mk_<d3z%^&aHztrl&OxVU5Irh;HMr8Gn9g+<o3lDhy??&ksb!c=X6s5Y>!
z=4WLj@A_<r!w%2%*rbxOQux|c0TWARCSQ7r?~TDDJ&kEPACTEJJY)VIjq@T{{`k|f
z-}oKAt49drVvq_jBKYk`;As&>$w>R6Ac<K)f;6QHp;h7a3ne9Aax8X70%n{^@&_To
z-i@)oTNud+k$m5f!1$GeBBP>5^r@0OS8;KB`4-Cja5;;cA+}2NEx&o0de9GbDL&n8
z9>{#>>GHY_yY1Eje2j9)cvlE)sI#l_`|ZCMJre0WerAHe$0uiw9;Ug}H}!yiGs58v
zUS+Efesv$uWFu5rj3k&Q<qtTi{$kidO(BG|hKy)vlVMKS^$W0Q$>=nDLyCB^{DXev
zdZ$%JbwCPm5VX@BA)p@rB3v`gwbAN#;7amPFE94EQpVIc@!jZ=-IV*BF{6h<SZu;~
z_4y#btAz$&Hb*K}BXv$8)yWH>`O;W_A=U}5os+G5@RAr<E7!}IjjGNMI(ilO$C?$r
zDrP49Z^PppW93g3CFxx4it2M@Fm2YsgP~nsuixskF~CLN^SK4g>IMa#$eF?fdXU{C
zT4g;P5-xC0d{u(hZpe#%F>3TFu88bT8j)V=x~X%Wk>#1@#OaiIT&55=UV!TeeQ{ep
zY$rEFzU=x<79WfZGkP|SWup-6EYfmRBMsxSk^3^>i_CAyYiFqlhgiez4zMLM|7&{u
zyB5sa*me6(FtFCaIpWJZ&3Y}kKuF(3Ia0q2uJM%s7lUPu1DAt6&~I8-H{)N!c{kDF
zudv&0j*_?kMyfEtN&i(7wspI9V{D%X14Y}wyN<aUr|mSoj&cqrdc>(*$Q!gzG&D;j
zq5`QGGH&bZ47nUnR7g8q@$@`aA8NLVXE#s&^W_9;qMLqdf6d%x;i0mKXgyq6Dqw+8
zn|~OTBG!t9vq!^xX*I;C!pq+~ZhA@j6AGItQ$gt`K8c-^b~{EFjbzPIyU){O2Pt;$
zOt;w$nT1pMl_T_<H;6p@+Kt$`ZeHCpFzx4uq&v~@p%nD2jt*fC$nN*^coh=r0s7c|
z$Hz&@1GZG0mkdTa^E)PIzx=lZG}2C@Wty0r@8%*4D(F?FYz`8YO4Ds^ab`-eXXVeW
zpVX%FRHxdN*f4cR1)n#gCbrC9LkZD`AQzn8Rz{iC#Ex-rnMg?$uJcn>bp*nZ`Uu>G
zce`1;KW{S27jC#4Y?)+GwZLY5pXI{!nYDTCK4Tu1&`)Vc^3>Uc1|I2npt)CijNIC4
zm1-10Zxwh@HeAQf?WM|u>Nt$ty@%_A-8gF%0yw`&Uc%Wk`G1x?d8gN9bU+<ItC^51
zT6AuZOqVUA$`xRbF=zm(8d<S0Ye7C$)a}V-nPLl`0F&2y2i$dUMg`eNyNYn!?PrN-
z5)O_r<wGO0b*nHvp$v|`x5icSC*+p$6jKlu3cBgDke||q#kj>Zr|t<q7w%F_{XN@*
zoM-Tq9;Y#<7B0ngVfA1}Pltn090K~%iY$x^`n8iSTN(;?pAvC}Q27P-vFrO=1uT}W
z*2nttbB;g&m=R%UUa-gd93hseQ5^?Z|0}1D1yZJ25&2n$x}V#4f^oQvNYwQ%;d6)a
z2ydB5PkexyXOd6mZ0)IQ#)W!BYskWeAiwK4Z7qfx<3$Z-qWI2>1m!n^b#X)evlU)+
za(;yZx2wm4li3)TJS>^NI}Z@&k7pje--WnrI&SaOl8@HKF9}cE@T~nF#ilc3@9Xib
z1K>6fX?}e35gXsgw0Xu+1Uw-<-~0ABd|(SNloY`CFC7LoCqo^Vof<w^IGD7Z3RK;%
zU!<D#a4?kX%fheis)_AX(|r-{&*~zYfUkVx8cjKS1*RmrAM2OHbS=6c(%fSDz76wL
zGRPi)0j2g^kKBdJ+Xa43tfXLsDRhEzPE%occD?&-9}@t2qUL#0_jZndQITBT+a2Q|
zDfk%tX3II{yi{kwju=E7$ZMQ8%cY_obmlM?NZ7V#(K|5p!}ME%{GuY9dNGidUu(oc
zdqn(ELkMVc;}VMtSb!s90Q#bf-;vZxzX*QMv60@nrZHWK`mm-|c}(cCSZpOR3On9-
z-0m-p6uP-BejuL(tbdr@MF|pJ&fnQ&M|biR?Zz3&GpYo>Ucz?}0&y|vLxX3QPnJm;
zTS1r)w5ZsEA6leGSAIyIK@pWXGB00{7sSl_fejWAAc{r_-0W;EA^5-{Y0u--ju9ye
zth<^lfAr3Ei3d|k@V0_1J9?dFo(nMPh#OSdRBtSyjmkLtW5v|n>n&%Omg)4tG0W7S
zW%8ZSh;%GjKSzLTmj(?eH%_njR^`h4c3LAtvbcz*!=D77#2!jcFvSIVssG)RlXq%*
z=7(+1A>a(N3xYtXr4yC(*G0IY65QcY>7m+8>Lx|@w6da&DKQLKM|({9T7tIpcv!=9
zv`@U|Lacj@Yo8}DxH5tPuRl|z?_Hg_w!=GV@KQtaxUYf#R^Q+)FbR<=ULdJb6|)fV
zaLk6j6R#Y;=P%xHCB<|@Z4%{|Z`*&E5y(<8v7a@PB5m;+b}caRlEd2-WkwsUN6O^T
zB>_V7M9z07@<Zo#L9{sNmgbI6h{Lgk2qqZmTp#7XV_ohOl3xo|Fa!ryIvI#hpJiuP
zRWP*D!3Jfvs>{f{3A8%yr+mo>JiI))IS~Yc%9$+uxFiF(P@O4~Twp=A{S-T=y`zlI
z;z}>cUWNhuTMZ3or(~GccB8<-QgBC&wa{@PO@2QjY6Xoeb}X|`6cCU~mpkCifx`sr
zpv!HeT%bNqr*4bsP`OiojT(=lfnsCxlNmvfM<U9!EHynn<%fW`=ARj>ZPh5;zJWZ@
zt!ULWe)*x>E`!_OY+9pGkSoHUT&|rSf`v&t{_Y_TMc7M1`KhwF)K84XIh2Z=e-)xl
z`0^PFa<tS+W*U%A54eX7Q1X(P`80I0`$LQOcq+IZjy-t(`5@qH%1I(ysRmbb5m>|{
zG5?B^<V7bF&hKi!bQa^C`L<eS+lGby?cT7Le|YcsdTR+?1F`U^w$bgnxK$XYkV|i=
zb8SV1z_!oYP2hn`_1R~QHBQ%L%^30dD6p5rKzPug9<LrlncsSGnmvD<uGbUWGB`@k
zLUe6}$ct6qAKSF%+u2$P_An>a4b`{*n#Roeu%&VXQ9ku3BvG;Y=wh5IVC+jv<Vb1y
zfC!||T>-c|)z6MYVF{RAwlx?VpwZ(WQc(3Cv)Fo4X3%KxJnNermkM}q+_afZa6SYb
z(>G&&^0+jnQ;2XZ?M8e(O%(7!#Q#Ci*C?dd^{THM5>vwl*_yaldq0Bva2p*IYj|Cg
zK6~7z5<cnvFh`ST1q=p?3?+Ut)!hvkxrA%&O*_F1%@@|R@UESMZ>GFngQ1W$5BmAD
z?j-uANJhHegfpK$MA*GdwwS6W<L;rpU6X;+K3$d!o2_&_UM1y7<KX-gvB54-4l2zj
z<T#xT7-m2RQ7lqWkN5*7nz2*}7<8=?v3wCUEoe~Wspdv6QgSJl!wAH<waf4w3V$;0
zH5t>n->CjG-bO`_R<wyP8p@CQA|k+f40q9=8Fl+1Pmtt{C4!9)3w^r5=N+;HD)PSl
z)V;Mhc$2i*?{J7ecT&~>0}Im-QB16%e=X>8*$NGU=(8i8%~iy$F!94l#|?Kz+W7wp
z7UAFUz7cS@ng4oXmb_7IXLjJipOli|V8J5R2)zhuZL`?*H|L8Q&d&2i9L!}3JqV!Q
z=`BIdGA<@|Vh~QKK-G#$Wf0e!l)2$yp=--sSWPHmv!ybYHx(ZB_AJpJ%Jh4xRT<<b
z<`{~*4a2!wHd8+t$}>nvoBh|C5XAZ?;d_h-`&Jg8>Nl!HR}LcM9=NA)$_tJE07eke
zk<N~Cs#yM;QT_i<E+=v9lxeyhWo)SM)IOu8)gTg6M;|6yqHGiufdk&}E5U8itFVNP
z$A;&c;7)!(hewXlhbP|R0JQ6i-0Boe0n-}s^?v|2phxA}#MxwuC%hueDCh2DokBlo
z4VLhg?o05+crO(!EFoI((ZU?uGu6<e+YIb8&~x!DCbbu@j8Fcm_LcT-)Td<(^9NYO
z+itJE-6m19^6Fs7U6<%)LQ=TXtoHghbro{0)E+p$6PA@jVrA-KXRmjCQVV#D=${~V
z!%Fky_3HcRyE{Gk;~PH`bbORRv~bG|=BYO*5+{4m?&wTjRGqQ*ARI0u=fTY$!QD+>
zciB|fSCdd+ck4UFI`oVG*Y3v!(u1&_QesQ-gJ+t7W^s{kzHakBoVs5b8wLM`!Z|s)
z)R5Xuw|rCmt{244{11lkC#^)=HQGg255?D143EkQ!CLtNBbw^V?)MEf$q&ik?%==H
z21Xc+%|x>Izbp$+n0)Mt`^esX$p(m<eLF)fM_Yl5ghk7Pyz|GG_hyxK@Cta5ND>@c
z|0<d=hyL!1bbR>)Jf7D0qqDQ{j>_2e{NgpeI(fr=kZRrZ>?IoT_P+2f!1w;dx#OVn
z?6dP-DNDMYtjzaB;EykZO=QxP!rIfqEeJBYTVa8E8UZ|0%Xf!j9i6T{1?aZ}^wXm9
z5GM57=ITtE#IR)w-^cwbcdZ3^V=?!=0rj_?qxtSzBdA5N%XBaxP~c#jtWcC`J{3Js
zRMXMKdnMLP{q;Cj<1oi{d1G&VC~6wWyt-%-Bb6uWUqU4NUjhP0<V8ZFf8$Ur3WP$5
z^s~so0toZwDnGnzZq-4A@7@c%S-!E6Tm%~KqRnyvlB{|x1(B;o$4P_!OYLp2J6eJz
z4auggqa3%ZGevoOi3aF9vz_)d--ldFB;;Lp;~Dv#AJgjG_LA8DEYbwWfPUL0>Y}CV
zZ$K~H;y4xB#aI4#=q=D}tO(pLbT{I@+)Zc9K!yAr-$&dcqy5Ww5pRTuD)Im_n1PI+
zdQop%omaOnOt-B<t8^cL-nx*P=6Ll--qK)BAopa{G@mGz`p9v<AJR)X*5}uawU@rC
z_&~mMAESI9)U6@7Uf1#2sSQ`04Z*{IG2*-fea}9no_^~bry;vrmDd)866{@9XG`u#
z3?-DVfC;md|Jl_6)?a<N!7%CxGKD-@yO(pY%YM9HVo}WOj=1DM0Cl!gH9!A*%_Fz(
z5TuZQ%N;%P5<x*xZ=5d+&E4(bp(D2*2vSLW_i{^+iTUiBde>ns*ze#or7ndZJgbPs
zs<5?4m#9ab!;dn*#n+SV+Q5AszkOg2D|#5~?F5pKZab8!`3cC6E_m{JaTJ)WV_WEm
zLk=^xa*}>&PPUfsw`KgdDBkVemjab=X4_AHbo|$*2hvL+Xym(=(B^Vi^|$X|OaJWV
zo;>Ug$&a&Eg5v`=&eHMG2Z8^5kuT@JM2<Wj9=p_C!`ORN?DQ8%i+So`Vr23g<Y$Ft
z{S7|K@R{-Pn0nkPjJO#MZ>!i>np{gI#c&F1s3OV?g?n(hCbP>Jxd5KS3K@7T?Ul%&
zEO<==Npl)}w!Z5oj!TnMVgfzU%z`ff8$I8U?G_HV{SSBEG4N`w!u$?Y8H3-bCMYnr
z1<>v>Gsh$S*I0t<6I6*wFLBoe_3=zS)nCt9e`x$B3S$Aah{1h9M!r5n<K5WhT*`We
z*)wHM>!hY^#L{g5>@1gKu(DERu>WJ1XV5D=xYs_(hb%`tOkFJSZ3ViIi?#=I*mB(k
zaDF=6Q}eJM&3#|^+W+d~=SbA72fQ73J9;5A(<Y74N8MA83|<H^3e5w(+|G4wW&0UO
zQriB#+4<&_-Z=BMdzXefJ?oG7MZ{DtXnS#gbkp?m;rJ)5^!QPvzxzCg;<~`a!52JL
zOdPv?o&iM01`$%i#{oBM5h&+InfFHNV~0J=R2uZ#2jwCGfUkd3_cgHNYCgjciR<nd
z?$+hv+tfGWHie@t-|a%${+4SMv#f;XZ5U{Ud$S(EB{_@u=(8G*=po@##(w|-5_jK3
zIJ)}^G22f65s=@zC*_2_>(8HKU)RzXIx1C8cVEeg`+Q6TAMiUNL;`L;)_reOAy4XG
z1w}$_(@rWrs0jM}xjF6iV<wv|!gWd<Kh@nJGQax~o^A%Qj)`6nK7}0r2T&|3Mnzgr
zc_}UStxxm@>t0=49D03WDFL_THJduMV=TeCAQL1>c4@wGf*uz8Iq2=3j5omEVSqwe
z^ak^XOK>p23G|AyO^Wa;-%GY)WHH>dm%;Jpa|<SUVngAi<iCnzOucC~I5)egH&>@W
zUll$=CATBnBt(Ul#f?sM-Eu+#b|cM&y_~)oD@4>*Bob=oB<}Uc?>@a#_;U7DLV03S
zH?*g=x1LJ6cJeK<_Gw`Hm+B%K$<Wl?oa?V4w&a^lwj__r8Ss#}ee_zKHJeT(Npk-t
z3M(fUXKbZG-&{4LEd31!E0WQG9#&7@EzpdNubF8_RKk~>eq?|~0{qd#qg>*Yc_-0L
z!bNL46?NmS`(AZp>ps!pLU*CHaAXlFgX#6<ooTSA*AIdb9mg=-fMIF5J_Y4gtSPm=
z46mb2kyKwGQi-e^MSzLdQL_teV_vgSVwJVvjyx2D7$P;n^G;M#2vJ4&4?r|r7<{`a
zz4CpFT$0w}?}i?xs+ZsySoj(a!ACbbeoTvE!0*FKVU9TbKL8&7$Z}R8p(=|-YXmK;
zi>z>otG@5e7BL!N?8r;WfcNxnUV4QWdK&Ei0KR^BQFzvKekeP5&rqwz!h0Rds9H`F
z&<6ekP_$eGc*Av5>1yaIX)LPeVKx?T|JMu3uJ-mfOwv%BK2gqUdE`xvOL%cn5;=l`
z+7g^g_WUIzrj4B8d=7Wa%AXjC%ig&Ih63s-9IpGw9}RxI=drvm48end1*sKJ+ljBg
z)Xd(yKAM18HP`>N))Tj40^Ae=m*{K%ZKi|&rLn3V_oTJZ1f~VJDSotmxBD=j^x_g2
zj(84Scym#FUo7jfQ<3nj7Z?VYZ+sqJ$Ujha^Ml#3yU1Ya<`s1S^GT55uBDA$p9-^4
z>WtLuBXJAvK)K3OLUifj#U0bY59U3OGK7w3S-HM1sNQG95Im8E@DCTgi2uRdTSmp%
zL|ej*6Wq1&;M&36-Q7KSaCdii3+@ma3+}-sIKkZ|xZ6AZ-kCLXXRY~u-Sw??``;s7
zc23n(wX4qF`v~&~uLL!5QPYtIPuo8)>F0H$=tf!VZcMr$^aq}1;N^Q;U$|0#rQGS;
zLCNV`BXL2;g75l$4Vt&Xa(U>-QHlEh0I2u<g)^IaUa!VT^!NBYFT!FyZ&#%t$8uIk
z`B%B!P9kyc)R4YjZhS&N{}~PmUWOXUc&O$fOeoUd`{LxTiWxAGdGHV555n-aO@!v}
z%J*B%;p&U3`Ro9%QJ=8&-q|-u901hzBpX2<=p-_c)Vc{eo%>NreCety@*+aJ;U7S|
z$7mBzg_8cWIpxONW6~XjoWZINm8V(hg=xD|5v`h%Y{dmaPc?w!oE%{IxZBuKrRb=J
z5v6~(-LnJ9m>~Ry+2D4%3%7@(T<8Qb&=MKqD<BaxtNxHtB)91edW^swZ}e|{AQTZx
zzE|i*KU?`rH4FsZF(;ox(HLlW5Py^TW7Te<X0*tn+IO_3#;pw|Z7*w=c>%i9Ug9~Z
z>IM$7C%3va&V?cFVcoM#4%`r7mj=$Mqq4vbh(EZBNhzQwFbMH{7COTgZu<9T?n2)U
z`PtrY<q@2q(^|!VX*~3jsjQL~jDnli9Hf4-cAcD7_PO2-1;vL!3~%2;#yP6uS`I<!
zvVsns+`Yuij@IFYtm7&r^bca@_c_QR5z;r!9*(k}lhmyy$*m3o@Y5Vifx9pfHkn%$
z#7qz|_YD-q3ruvhYvp*^#11t`hZ(>k9gAR8J_yLOd%03!DD5dOBY)3geVBy7aB(OQ
z)1uHUXH7{9eK6>K6lrK0Q`|YKM4iMFs)?@-i5L75s@j(!NRyR_r)4;*LQW(z6pwqI
zpWc+TGmvn5*~6db1}p6ZvT+ohlpaYWgpa4NJjlp<F8(9t2h|^O#rK*_f=}vvg|P0w
z6l!rq>L&IJ8XpSsYYU1v<N}^b3N+0R;pd&9MWZDw7H934-*>|C39Bh95Yivg!D-hn
zTPJA{ObrsuA0nAC32MZci4?Pe;$E*&6wu_3tZ0iRqOoiRbP{{6`V{yMC!Nc5c&+~6
zQAM~<1>J^0r~p0pJE<clsZCQqg~JcFITRxzNyAAK!H)k};ZR1scNurkUHYDpr(Zj{
zRiImIwaxb?5-LzF#PdC}6~izA2J;#i^z=PTV=0@H=kShm4^_T(O^X0wy3ZP4P&yjy
z3A~`H6ZWi;IxaC^8Zo9=VLTl2J&a0ZEg3c+ivfQPeiK7~qzl}K5%JyE7cFM0RsHWx
znDUJzb_aw$wz-E0-JjQI<g;V{!4BvY&;FJ5Vz=2&c&%?q-oxtmy3Xx#(78MVKt2@&
zy;xQI<AHouw?-1C4(Aw<-(b4G&zk#x2=?zaMAgUs5;P!)$wcGxLN97O5>2RBGx!If
zp;)*(68r3X6ahF(p~D}@dYT?u)12;=P(e2k;kXYv+VaTK9*1~Po1jS>YTZR|O(>kt
zV}Vz3n~<<_>qyTLfv~&Z@#;<rBh2&F(SWdL#f(2=Mz-h6EKX>j5JIGO!k_MksaQJU
zwMBcL)h%(HOQqi8@v|Mj&Om}t+?&G*Iw9f<QUM1uMwp#J@!3HbVeLH|MM2K<C%X6N
zgDn*>I->LAeL@fS@$VUz%dTt9ukL3JSt_0f6Y;<kGN#wVO}<kwQKY`Q9zzpg%RLtw
zzz**4DbAPyn8cpJ6p}tvZ3Py$0TP0SKEYr$0RV8F&kZKnm^_JGM-h?L87x6;U;k%>
zvkt5SC39A9SaCn*`Q0SPF4ekVe2@ZVc7s2|s(IZC-fuR!GONSAY2|YDrluxzKG5OO
zrwL0-b7U|~!n4kG*;Xu$VZ>en{MscV-x)YUDTu2NzAX9$2qa@#NfTVy+cCzy9m0tr
zI~HY}zJ;x*6HOz&-Kn7Y!9jxKLW~Ns0*AdmcR*8MK9Z)sk<wj`s}~f-agabK=(GGF
zhAGHH<4re_q!gEkQQ%8GFd9=*FfBZ;IQ+2oc?}oFnl*NjRU#XO?p=?A+^%)4IqANG
z^_n%uF=ImboT9=v8guy%fa~wIGuw=tE{IFQ73BLEp=7>m%YC8BxOhcq+@e`S-;LL6
ztWAH-N&HBc{vOKkK0Zw3xr}rvF`dcs55Oo!h=xk&0IQ&0(zsm@l8oYU_GXqvOhI(z
zR=Y;sIidE2R^@fb=E7x+pQ6<JICY@Oc&|%?cAQtnm=H}#z=2wCy@heRP9h=90=M}@
zwV#5Nqa5U=m|mBkriHp&tOH#Ttzg)vj2IF%8>KnQ<QDI>C+x@?Ujl1B#;zxoYmSkV
z3Bb_&t{5<|=yTrfhZ9ah*W1)9nD|v^Ky+4nFMBT<RWu#3R$m$wI!+3$N}_@>85dp9
zZz&4?!do|v1JPiv@RlZhxVFU2pMfq-dp$)t`~!;qeQfgI|66$_F1%jfESh>_!~uKg
zN=KOYsi`_&=LYhP*ddkGz>e+ZslUCs(FsS1hggudSK?W@d@hsts54LiVkK-}6yU9?
z0jL~kJ%9p=!6keoO?xAwUwT9Fyu81FB%p^0Gaw+lmrh^)JzIO1f!jrp_S?v_5nU_b
zkAm-$t2ruy2>7aw>l&;R$Inu-%g5s;U4~$M>{aO29r9sSN9;gyZdzCYsM%KzEce5K
z#w7NQ0Z~8kC<ZEI2KQp(vF!|pbT?LuBF0Th5+pCNPOS*NO9TdW<8N59&33`@UrU;Z
zQARvUmh!8IeH5h<zqn(}gV^RLx_zse3PUB(DoX?Y0UTTB`u8sT-b@HQ9;1}o@A6&k
zs(154Iw*P4>#Q|IdAKov&w}>%nS0dbzvV3v;&Byxz%z~-4O#>yqdy6#06o1cg!qRb
z?Sb+1MAo)=d#8q&CS#o}i=YZ6eW7HY3b7^JWNIW|<YIe%c_~WTo1CQo7(~y~L@v+;
zLR?m;?)9MDQ4nq2+s;vOjtTK7NXv<0F3ppXu4t%M_7VGO$UY>ef9zj_6?L!mT<7Vc
z;8r8zBgr($uvNWR7}EYyGQE%Ux9JN@KzczDIlt@ZyKG;w6z@(D=4khc1SCz97kG6V
zNUBJkPX35gQz2PiW|I}WE{qqd^sms`ugU-RxUj+H1HFNllf7-o6DIbj>a_O|#)H87
zI(10-sfH>cleez7g*PRLmzTND-=fq#GHSfN@dToOsn|8+aF?T&h6w-Q1{|2fM7hy0
zZ3(kqMmGA%OQKalD)mVlLJj>Vw37?skd&OYN<>JHi45UnSQvZnh^2#`0VX(6TbWk`
zUfaV&XfaU{9W7p&{oOJGm2-4IbyiO#3fH1BRUXlEEDH7Oeg19dwvh<f_dBaHrEjOb
zShHrlr78ZU)VhK9O6GUXj_yfjJlG32{$B5d%7bZ&jk2Hr_CiJ;e~6|+5D|i7HKZL7
z51LMmg;Bb_eui)YiKx~37o@acAheC|&&O1&riGXH3?c5<HGAO<`*YcLVnK;ALE)*g
zh4atjI(U;j^0tI)fAGy=QZ^qK5{%sTB7pip4F(#EBjar8GD|KI(WMd$oZ%KxC#V3;
zze#i3c*qgip;k_w(-o{hm-!c)yOGnPRafOrl$pw-ke3*PY{>UJ`{-#zk>&Vc0+o$5
zGqIZgam_^=$wp~&5jyU#T&i*Bq+_lY48^SUwqMoJiDY2Pi(^lIv2-n+^hsCLi`#4*
z$4-#al^}#T8K)x`5Xe3#;__53{Tea}afmV{xZagV*8SQWW5tih<AQ!M<vU-Ey8FP6
ziv3^T;iF)+x2srR1mieFo_k+ATu%-w_J{M}qi|FlKQyC-o$OA`>ZXe>hm^<S`CmgM
zKaf)*KSKAkXjMnOO@bBZH}|n1!{miYTHqZX-J97-WQt$be0KnOanr}pn6MK_dgPYu
zX2IJZ8<OaIKjrPm@TWkh2JPXB$ODcW^&2Z?@fR2=51LQ@DNJf@0#;&^0>-B`H&#j?
z+q2zOA<nJ;094Ha?3HEw7XuWk{`RIxSO@mI_(C?z!w8xqDk{!SOZLa}Q7=NIj^1Oj
zK6cG`_LLA-G*2y*VP2r6F3=BjGFK=Ch#)4Hb{&~!_O6a+uLypJq7sH13v^m5Pzh?5
zz8WCqv3z=^d*-MJbfViN8*gX5AVrZT3YV@*=k^lv7nBeH&V`x(11P*{EhOmMCE<JA
zDsO&9S6j8Dx~&?=J^u$Fkxc2+pB1R%!&$%P;R@j+q<VT5DyWk*PU$csr@HwTdgK-i
zxM({ytUM-k{&x^bKlgv#{()1SiP{@pjob_@m{48Xet*DxfdBXVhG#@;>%_M+t@V;q
z<Ciw~51_jm*~tKNAAckL1HJd(evGWS6xH8T9?`>F6V6o5{b6}5aw>xn`4POQ2>b82
z&iL})_VUDZF*k1U*e>>gn3V8#m$U`;?N}ZVec4R}^O>H_4xCM>9I%xqyvGZON4{ZG
zqH6>>U7SHmCkcOTWp5p5Oh(!GmZKi@kvW<h*sdo3_?%c)<F1|TTCna<U+kAX@7M+_
zz1;w5tQKlxnEmzpfYVa?aBi|=C>((v2COKLKtN<{eO%OOVqHG#j<SID_!jlT-lbpk
z8_wtV!C44lBmcI{m9R?Pjp4t3S=VpoxQ?n7$W>7)EOxphUp2K4&Dy{s{~8{3J|*)H
zz@jH|w8k%UzV9Olvmj)@<Sf}-C~a4uXRJ#&3$i?0SKt4Y@83uH4*)JY(23K0pz=}M
zhwM<0?dIO?nXRJ039+}|%y=E;-v^4vfmyeXP2RL);R3Hg5SPaK*fu96k?zPHcurJ)
z9}-faKU#-oz0%2igduK4x*VuWeo(zJhxnwbamYSIk%aCtIQFe>FVaFt!Dee6iSf!H
z^$~^u1%~cOr+1ux2^U7eX6g=_@k$5ME8xG(=OMeGpWXeNLclhraY$+J@dL9LQ?P=~
z5SZf~+wqZ3itGiwA(mBZT8buTTToZWCY@(XdDgJXOn9BiYHVTiqv%8ZqF$oxZuTv|
zg3j~yok%7c0w`oz{@aQ({!+TygTlSDe)uOhXeip|ViKQChJ#5)XDELurSamS=zvI~
z;hDJC)7+0J#r&6yBVLQqd9L_wi;shlzYd@}qfU#tJ|6!(s({+stZ{*BwO)ma^xu)h
z(}sm;e5^hk?fJ+5$~y>Q)~`!Ai>yg3sft7IK~QLxjG=l)`=F|8y*{4gbs~A^EuI<l
z`QW=jD?%Ht{Mi71fqG8dwhU2kBDh(MYR}OT9~D`8iW3$O-s?;?F|*y<1TVZS{>HSj
zFaCcBk^T3f{lG8Spc!`;V@Ke$kc(5VOs>|2s;G}mlNE2`*tzyn*8Kb}Zd>Mi>Tg*>
z+2{4N!NIhpT!n$#Xx`{(nKhT|-GM+%Bni6+ozBEUfdu=A!XGN3?*_dgRHs9J0g2Q^
z+2$cHOqIIyY4X&d+-8IwPD)6~jacEATfXwPq_o@H)adLc!fB3_^bE*VbW|uSH6t@E
zf*EWVodX3~28^0cJ56DG1sM@N{<WkCVMM_WN&33<$3#O*@qE{V2P$M!qRRJlP$~gA
z;pfZ!Ib~&pxC4<<AC&3*iKYK3|8NJ6&O6b;)SqkTsey&1o_v{-5u<GOo-9heJ^RP=
z>#d9j?ncJYlFFz^m5XaT5N;p*%Y#Q^g^*z_0t;;(|Mq137ALbR;g&z9rTZF4Zl2|J
zI(NIV<2b>`YReRz#AHw~kIrfDqN(FJu{-)`OV~Y$A`*%iB;fn#h=xqZb9Hk)7V~3D
zoP-^%*A>&2B#h$j9Yr#Y!adQV%ZDIlQ~qA4pe(AVWui;)n`-mtvrsZPO9lYJ|8b6=
ze<?q($8f%d4*XdUEtpUSw)=OFv%)zmRcOoXGE`=cGHPOBVI3YQjgOAeDIR<-TleRX
z$D^vJwm97B3p<;Yd<wz%4hW#S9Q3&qx*cdCH?t;6uR|MRR}CH|gF4$v*#>DdM{E_i
zdu=IC^Db|RPCs#o0GPo!5!sw*qGmRG247-h+2={=W{wY=|2{vfZpbidwTykMYYCj?
zJwoYS`hMGTHAAqj{5p_;F6SZz!)n{x>D4%P-ERNwPxa+5^&gN0N`OV3usQ<!u5`4M
zSAxd3WXOzAIetR$3|Hr%TgFAGz5ju8-w#w5Z8{z`AUc0uvS&8vu3a67s0cyynlClc
zjz`^I$r%=R-tnqBHQ@I~-ZK}sS>`&P{Xg>5Z{~ud9hQMl6~oP*8+vJhSJ$l<Jq>T(
zjCGk?IgxH(`LySwPTf?mYYiVYO5Y2(-v12qK@h^t&}Y98KRa#zMS5YY-@Wb)2|dOE
z>sOX{!V~-0dwQK0RkE)F_OG>ZE`riqv(3MgM{cwkpf<kL`r-K5X{~YUC2^x4UbiC0
z*Ky{xIzyS0__e*N|0X5M+W^62&euh&FqB#EGnCP66O-i{J0-rHf$EL(vq1?xC1>e8
z*P~03QWNM5!fOKQa7xuAnsOlIV}$Y+k_Y79ju5KAMiQoBiqK;_<y!dkZ#H^*fvlo?
zFaD^waB@{7nhX>eEfwY28|(d-+qF`UlknRBzMy&(;N9?dGH5tIIkpe2#nr5%;678Z
zCNeqwBHx>p%0bqL0}<e&CzWf%v==|r2}WzCy+6TENv!Qzov;g)@a>DouQq=^A@=H?
z)3jid)*G+oFU?IK$Vc|>Z4oBSsS2;PdX$9g+#x+g2ZM3CO^2)1T@@P35}FLEB*Lk~
zOWmQy+l4JARNopu*60o8*-sR1Z^#RXss}{s1xnT8FDg{gF`t~Kb|9?KPjufX7W2E-
zSgdV_YScb_W)_@Ex=lCR5AuPlKS<25?_pLerw1Jizh`yl=N_^qS@^D583Yl0D(T@|
z+>8UN4;({g&_<|Rjl;B7bzf<Zv>-Tj#UpAn0c6XNsTo~z^-|q{2mVZSI-^ji-R^Qz
zPUQRw=2^(I_J*qf!nysy3>KjzzRVkB;mG&@v=^IL^i2}8=lwB~S{R*dnM*^u5GaIS
zfYN(#k1t|hb^+5y*u!(Q$-IT59EXJOrdA34+IKw2ZS3Y$zXk4t=I(?QjUG=o&tgJ#
zE~h2qxT>3H_4C8i#u@i?aaSw#G{_U;@K^%=wW)EQk0kyH(iY0Hs!QkT`0;F-01Sv#
z+ifr+Gi%V~xBphm4b^->q!?j7z|seRazy)#;=EHe`H}sje#TQXb<bs9RF@G!A7Pk1
z0V?Sq06e0B1JJQ<<ttIGKtnAV>d^!TxlIJ_&e6esA#N<nz@AT~Tla!AOr?(NZC&H|
zE{TtUy)^e(Iw$A-l{*3Y)$PEkX0Q%;GJvb9*Xs9Qe`#U<EfcgKks{`#Tj>luXZu)L
z7xTdSrJ@S#pdlGX+@0bEIe`y0;xiM-G{t!-S8G?gAX{+mM-BP`2{cl-akpOC8#y=G
zJ8RPA1PwR#jOWSU9PJx#R3724t{PA4wXZ)^FIo?;doOwj*P3!+X%qPTJFaO_N)iud
zb}9p2mf*;UakB$_(u|wR{{eKgU2mNHaAO|b{kU_S1@b#2vI^3#(GCw;ZK%aQxUFTr
zG8+0r4-2>r2!Gy;eD)XAKd#s_YjUZQ1dCstv|0EECbS`#-2XIBea@#61{R<|JBq)R
zM35Ye)Z)C^YrQX430o=N^;~ZMxVgJA2X>>S`P?!%q92eNtyVrQ*J6C=EkJO5N5v6F
z1NWS{ZIV1~b7v+mAMIWoNN&<BzY^nZTN|GiO5XFhx|=c3x50)Tw%pbu#x%!6i(X0?
zq4S9ehTA5ZNJWbpCB7hJsGYJfY3^F@q6HDo?mhqlWa_5W{K8uyz@wQ_x$IKNI+0a(
ztO@Dbn^me&;gRg~3Ln<*NpXvP5^g<WP{Zfu(g7%MUQ5YCF{;Om6G&FcC05b|j3H&(
zNg=gz#3Q$UYl6>_c}KohZ4MuzsfIJU(VjSF!VkGbnr~VXBAvk#C7D;Q3J_Gj%)L>Y
z0mdL$#EA&eDR<P-pF2W~7uh5<m~(qiYw#Lj;~Qwy8?%;vBtp=9&$dsv1Wv^&FO8O>
zcF_Kf@LOXm^9z?wra5-|H(TCL?`<}hN*KIHhF;=7d`M%ug@6qi_q<ThhrqN^AqM6?
z8?PtGuFsuyD@I-6^)(%0-o*Wzb~Wz$_yw9vc=IVIE-;jXzc(5sx^!%A($hBScM$XT
zDkhi6<5mUBsfpW<ZdVoDp))UK#?pi25^h?Cl6c@D?2m~G3ze>2!?CC2yhs9*;Q)+5
zClW6MKLbn*%$70{^34ybH{n6qZ%#gu(|U|J%AF#h>ks^Zj_V<x035Sa0hl*dEOX~~
z>eNjl3kZUQ$gQ>Ob$uCfROf$pT0pswdFhezmpYZqfQuVD5#8ZdQ%JIkb@MOvD>Che
zeVn_H(b#s$zE;Zn#hh{dT*2Y$(|hM`;tSm~PHBN}Y2rCk!E*~8F$SNfF@kckob+Rd
z%*QVXcUrFXy3EOWa6D1kpl3)yvYzZ?yG+0@*}xpD=$i*ABD7#3NgLM96N=YQ)E_q0
zRX2>XLXWcO%d)Ju3KGw_kfhTh&p&5Am(%s_VncHGzfdEqPcc5s`WK!0PuDrz*azov
zUtoI1{bJ^nyT`>Ju;kz71!PX{5%m(*k1O`|{w+Agg6zM!M3D2|zl6tGcYmJW|4X~9
z_*wV^ZDUfa-FyD>B&KlIN4_Jkw01H>{%sBSu=1XuG`N;u);I`j?3Lmt#l$<p0*+38
z2~O&zTXO3cT@cxuPD67Y8~c@Hlyx&XiA{&#MI${VO6p2vm<X1`n8~`CZ0c@>xb6fe
zvS#$GVrDoJNxip~!6b1j08(3N=@-2c){$+j*M9)Jt&p{&K+nc*Oo^oR<K`m%swBmS
zRb~~Ws+cmWdV+z6wsGf_Nf#efxe&yLvURkX4E@9*0j-!G{np^M$s855y(y<X+|gRx
z^wTKu{qO90V3Lc_sNZ9#RHW!!H&(}!#pQ7`i>i-^@Z1iW!&ev4rH`_@K)fFuH<#*%
zd8cGcGtBZojgl`aWB~_PpA$=Nz(U_wORlJI3Aj&|s4J1!WtsE%WP)x!r?!63#{_(<
z=A2$YZ)_dy{!Dh_kv#peJ7)Y30N=x_dhedh0Ou`M&A5A2{sB+HpH%5LF~r=OI&`)F
zzMAqL2H8L25-LY>zD!{%piKL>E#Jc-d>&kjBF2=&n!=PnQ!ibT{^K={X^3U3|Fcyw
zl}Br-wv<wY>s^cw`=kDKO8<)#vVX6ZqyNK!=>H!MMEQRu0r5K8&4d5xbx{TXIlDnE
zBu1Im`Lw9SSBSIN$7qMTSJ1rF_{Gk#cF)Nh3c)6CbUqctWmT8lDJ2Qp&1<;!cac&i
z5uZS){K@3M9-ED5-F-C(R6U6oq?IVhPNz~II#g4;xy5pwvq-t3DJuu&r{hzJ#-ffG
zyHJJZ*71>(^C8>yJX0Rj#wJn_KcGXMJwmoA(KI-0Um*ZLd_Kn`r=*h~0G74D3RSws
za8fe*@uvB9Lb#%hUnf=;Yw6CyH`1g1>d1rg7X$%ID8_FsADCn(_Hg=L^ebzzz10uX
z6^K$e5#PH|1UHh~+2v%7s*Ep4;Ce&2C!v=y_kSka|CNZ||1aIK|7skv_sCqw!xRD|
z#^zFXIRuD$XpHfVJe6$J>pjhwXXUIEP9S^e-H!H~imS^DM#6XqwauI+uepZ-uQ*LK
zRDKY6LvYE;getP@k#tFpw|ek}1VWh9lmOQ@G;gBNVGW}U`IRj+aF8H|>FOi={27EJ
za4v#;E&!EpcD&jr#N(?Ii(BCL4M54e4^=_3+2VU6`A0#0HyJam7>T8@^3!5Uh+4au
zQsPc92#%0Jnv+APwU%Z+tSYo=1U)hJs;Dfs;`658gGmP{x$9?_wApAXLD#;F1565F
z`wiH1lBmQ2?mk3V^l*Q|G)b7a5Ws8cLQvKh@rxz3rAp{+9|_|=*l_SXI<<X&{Ea8#
zemy(#cX;#c=_nOb!+|(VIz00F_Wr^u(x<p*;GwPU$~d-|(7}w};}KAQmf4%v{a&gU
zDdK?d<}0pjjw48LPfC(6ueO(~7pY=1b*@MhWcI6@xO$Nvy>)S~HtNenWJQ3_kq7R;
zh6&&SWb}G+r)wCVpK%w9X$uUWs7XvvkgP=X3la?A*NNFd;Y~~&9@JV!_K$k;|0Y>w
zM$1L{zt{P(fKx%UL6!AE5cJ)vD*n+H?zAzMyA}V(w{&;|LnZ8R2*Yt}_(+6)bi&#G
zN1faYI)YhjQ0LE{OcIWDA~X>~qyFypgK~#;!q`3W5-dqOHFsC2O)ku>43D{A+`Z~*
zl3&d9zdXf&%Ap!2C}a940=x~xf4XiYS*(=x7o<P$>1phgArxA2N;+&Bu5J}r;mw<?
z7S>v`3l*aAEMk2KxQ^Aj&0ntV^%M6BYy7?U8viADUgtm8^<IJsZEUz<zPM4FQ={~{
z%||^@_mkV}u0aaO|L(V7kHB*E7>eWN<!$xAy6ddDdMK#v8L4gDajv+N7(E~$*!(Mo
z1$YDexlOXPb&A=E<E-d{a|I+#OjiU^a`m=s5+Oj50N9rV24Q~a{6vSDPt3PW%qbh`
zB!7`sc!bvnQ+VY0Ns6920U0i{wR>KFOD%>f<V~=gUpgb7<lSS;HpRC0H<_yD`v?4p
zf+l}tw97ikvgLwmX}{P@e=Ki*-Z?m|M~6l*P!@sHH>fT;x}W3=f-z(#HgE?Hy3Xsj
z@-#s=<9s6%H2ip~{Q;p)Zgbu6XLaCp-70JW?tVJUBT@5CtwZI%(jW*Tj5_4MK|bX*
z5D!9FvFH*On{u2OIIE7uhOmjtJI;fbnuB-~{lhEgfma@P4~MZ4HO&r+rVr_iwl9NK
z8Xh@PzaV@_8qqXP?}}e`+ON$502w&FycB45qf^8a3e(89NroBBq1;?oiOI{pCfE>C
z`rrUs$eBPr8vn<X1FSu3DK$0qE5k%2oVy=g`#kC@#u?d>NVdqyMFUxxQKV|Ho+N>0
z9CX146cTO*KGS?$gF4$Q7I%mDi-LHyC!H3)i0iJz4yPM?0;`~zy?AT4!@635rlWhE
z;crq8q(nw52VGuQ6DmtI#aU6|Um!#R3kI^R2I?TY{XZlhs48#HX{AYaOyaF&u}}>3
zE8%v>M{?`APdZAZ!Z7%Lf1jYGudMMACEXS*>*$C!OuM8HsDB<>@0-{J_svwG3=6UN
z5`b+Y$?O<6Uqkg>MqvcN{6DuWgCUTt+6PR~-AD+teBEE2FC?p9p?dn4CRxAb**>Aj
z;|VxTClUomz~MfC`I}Lf#-FVMCc`ASS6Ud@fdiH!GiNCYgWA7;4%}V$E8byVwCFbh
z`vCS^cBtq0U7fc`NLCMo{sHHwuevr4y1tR3nCC|jezM+{G<&p5V~@XoJm&`GBhj^4
z$OQ<BfMgX*6%HXnLXPjG-H!RbSw(1fb&(|OM}51eEckB*j*bm2`SC7Icrt*SX}wM&
zyz=z&FzTl*A@J5r&}EAsc+g&&lxMN>Q<2k$y9z5Dn96zDx9i!N*p?3@HhI`q?=O2@
z^m|%h0eLmT+xSkjPgouN&aE56gJT};vsZ!66q1{WNqk6T$%WaHr@NI=OgU6(-Cu2s
zR@RfS61V6e^yBBK;|a+(t{vZ4vAe;upe1ggz{Zjip-rZc+eH-mG7K}bo!;wB+e3uu
zggj!OEUlg;v6gefH)3*UrH`1Ys4az#x;D@FMwEekF}{|cx35gl1E5S=FoM*vJGG9q
zpw!Yvp#A;fO4RfGya2n^E*V_6P!^M;pvkDWC7~_x&7S{GA8Xwg7tfcyME8<KJWMZ7
z<D{*_NM^Sq-*pZkxN-r{Su22ZCvrpk8~;IZBpGCZ*z*^`P@Jh5qs7~*cJ*R*qDN6R
zw<iq;>ZuW0gwIm{Y=FeIKeMo#bo$#OT$Hf})3-3rhG*Pv)4ZE`R|lXn{B$1OukSpS
zZr`Timrj&{nRQdWNo#>#KKgP{5gg?TPJT&7n43(RI;8U+*VC<DM}E(TKz62vT5k`w
zg9yI4$6fw*9e+xpVWuvLwgbp0ABt`0)W--iJoE|On-JhUnJmGpDRH4)I~yX1j}R~@
z@Q|V{JUab29#V~pQ;CGN+wv*F^W?Qio_H@civYs8{wkTAebL6%&cmE<H6A)PgGRvo
zn25JY%m-%Ru|y5(z%o(gJvZPXG?TVameg;>T=C(}NK+u|K<|<ZG5;=NBJ0I56#E6w
zgt0tb7TN5QTmDD380TH5IxWPl+n!01#fDH^Hbcs~X^)UrePo=Tpwtqc+PPe|+-#~Z
zFY)tjGR?&P)?4i%%3syw-yhwbMJT#gJRwi&!gZXhApA=cHy2{P_11S(RdMjJ|F-**
zQIBsVbAOT}@lOuIcV(kBSOaY-Mhnm&vv5kan}y-$u8LvW8ji`3Qn1drL5JIE{*T0}
zmmW18)X)2t=4~o<3i+vmyhzA%>~)#l9ER+?R785VY#xW*F`jB1dt)w>bgIQ&0JLsX
z68qDh9_YWS?)oUpjS^<bnlWz06kPRG`M)fwa0zS;e>0F{{z?_rDML(+ydC)@JvKE(
zIifU_sWvW};kWsTw^BNoWF@5yfEGJ@Z)y;Gx&rU9cV5*0n}4!bVCmaV)Ug0N8Z=3-
zQoR3rpr!!(d98UuLQWzYsemi3m$U8$6kvL!RD1?Sw4im_-$e^euR(nT?mGwa=@2~0
zJ5GaU`jQ0aMUUaSPW>Sh-Aaym(d|h}C$`6T*OO3a%vkIH#e%Qaaa3>`r}-bJg}BVZ
zUk{Eq=wf4Sdmyaoz^{<)hv_nXnN~zwieD=a@UMr=0A>T<tJj{&w#|k=M4FKuCbhMk
zV}AMN91Tc%G}Y-T#ZDGT69>agyf!voO5%}3Zj;M%J{%3YgqRAY*E`RH>&d(n&cUVU
zrKPsMZiX<QM*<kCyE-A1O>TN-gvJV%JI`vkf5;hXaews}|I(4%$>dN{Tg(ZU*NBZl
zF+m}0Y`gzN7nGlqo)wM-`dS#<ccFBTOa_Bk{zLvi`b-NT`q>dhxaOERg74NKTE_V$
zM{`l*fy9|wy!sIFny4`|H(qH5caD;72U*uFqwURHmx0WnDyPU?o`K>-p&~gCBpH2z
zDxR5|fZZ4G*MVa*>d|&z<_#?Z7f*MY>{i}}$J;Q|LMklQd4xqgv*|tZ3-ZaqAS)$6
zj(c+!t7duI1W*#`*D-Z>Y&9t5ZDYP(cH^`4eNeDwzN9hUd$Zje2DuKNRsEf{vi+~l
ziU0pS-eu(I)__{it^E(6=OpQBu7=8U9hUF%5@lGApq&h*pbGd0a`Lh?^t<tuMAd9K
zB3WNN9MtR)15cvL;?QxFf6~5itkW0lXVY@D<&-JA5z^={sWj-POF=JRZnTNo#w$lR
zQ;3=(&!zV^%kPYK?vjzEX`PJP<Yr@RU6rnBYSd_GP;JMH3;*Oa$TO&^G_Tq=M1LY9
z$|oQyiatWJFy#K8A4+mmf#QQG3dnRZX5#>#@ot|l^00hu5u+U;(#pa|;3l-t^RV$X
zZa@$3a=CNzD>K7`^X@<7umNV^r!zdmnxXO~zg$yKP^YIgw$f65E;~Am1rEk_&SDSQ
zK)b8{!jmHSPCP0W&iGm7ts@MWVE!v!iE{#luF|q31ynNwIYjQ3;J)4Qst?MJUXA((
zpm;%E8Lq`CzS9e6voqM43mTxMk|V%l0EI-!DKF^#))m6#q{wH>;whz3qP1z41=<X7
z*d)8X9?l~9y>NK$9ytl{4XMG?<n$!?J8PhdcP0&0MF?4m=c+heP@Un+W)M12%kb%@
z6WsU|4#r|5rIVJL!d9T&B~yr?+E|j~;j9SQDMl+BrwG)zNQb2?q8X8EcrQ+#{CrzT
z8<eGN<du!KA(V1vvm9${FIi|9swxADhy#Y(@GFZbjz`2Zp*y#<Pth1a-gB~AY@Ctj
zQhBCC5j!-|YU7bne94MQcPGYr)q?WiAnR|z6(_EI<+(cLDkoTO!8&vb3)EHF5Nb{y
zR|(aY(zt_iWOI^{DO$>k2$YfGbU!)#u<eU&6goO3-hAeC_A>-|cmBb9ivC076uBR=
zGG;7|<8Ey`@SIdN>3B)Bs-UmHY<87!KE~=QTQh5^0z)Y!22)`Yu|jzn_3Ut26dP~!
z5P2mFOZ2#(Y_W!O9987}nYKXxMW<OdTYR*?&k;p^14dtQCn9N=FZ(S8>roPFbhZSC
z7IB7MI%?JUvSRd9s@dgtinLC0toT|b7_0o=IVqTV85!d2Xr7%z>?Pbv5x>@(hpj}f
zblo*HD4<B#i*X<btU%O=LD#6(k}l-tvE7WJs-6C^;R8L%>)*`mNRLw{zrn!{TwuS5
z*b))4+-Swk4>fBUVvcs@La6m=kZ{A&+-}seie(QhERm05Q!QiDWmEP~74F4RvV|$A
z1gpj1pbB{xIN=3ue~STlcuaOi2Vvxn^0T30Uc8Lnk{r_0EcAyZg-@WhO{Nk^G;@AN
z^yiV4-sweyWp+N(V|Cr+%WA}2SgEVo!invV*_JC2OahC(CSXg3e`!>u#Lnsua)dV$
zlSbQj8t8yO;a{GKlOk71feB$S+z`{rl5rV2`#2uOcxsrZUK+ZTwd6bvfb|i?(^zo|
zE?(|4aO<S#;Jzk>(2&!0ltxt36Rz$kr4O@EE1JpXC{v}ef3@~W#ON0CND;`$^Pq<k
zivp`*&ttm$1NeT@aaZva<Z&HHt>Z8_CF1(2FKH*jTjh2a>3o6ehs*X-X(V);f6dde
z*6v)q`Cm{>STR`APO@^5gS*t7CQ>b0u27<>93E#;Ko?6(QWVdU0tKfrc<)zZHUdq_
zv`~N*ux+qHQOu1T&yOe)Yh{L;31=jf9Vog9iw??Q%|Jd-VM+1Hj#R+Dix(R%2II;N
zSxE_s>lY?WDU6M(sV^=R7fU7Bhn6XfiE$J*@TqX-fX37WS`4zvf$Lbs?dm1E))DHd
zu9)X4ktlv6OTi^%)p$<BVbE;l#qGaDaf_Hpe_r4LZ5(CCT+~bIr>f1U(Y!y)E8^P<
zCAmnxV5I1kYaj)yMw`-OM1Lk8gZ&=j&CMvKVcBH;T-6RE@ss^a2)*nc*1_jNug0WC
z0J>ZTRfT}x77mk5@O{luIpKG<;(IwD`8%S8bE~$Hq3^v_^+*EKK8l;|t`x0Q@j@rs
z>LYxd-p&23UwghnQ^yf7G{|c#dIHsVq<~5@dx-}kzWcNY*^>>do;>>w{kB}knFU0B
ze#t8H_IdU82M_6ZZqh*}0`)=gc{wkR>w}vATM$P)*x?-yzh~6X>l*w+6ZlGQ#yagl
zzlq;KrY-gSW}e}BA6AyI<`Hy>c?v(N$B%s^SG*lxgTVNBb-7D5z_tb8FF7|BbRCFi
zlO)!??fK~5lFB`2*1hHn*1AQF<t2@a>3j0<fE!-pLj7|0GF<<1GB6ia*8+SUs9Cz4
ztsY~p964n#XFjzKc{KLz;`j3LS|*5~NWLW!&c`K_B=+~YM)SEj?PX>e%G;Tlndt<c
z1edo#GvL_YKSnxS0<&^IzJHl`^4$rkOCt^3eOmJ9Y;UmgIy?DWUD>--d5Kb&9?5pP
zQ`=}3mYO|yR2OIg{r+^>V#3|rIN=r1`DCCI*6Y{ZpjdMR@nXDQ<ZI%~jV#N-=~GDX
zki1p3y;YWeYl5i1`=5U29cMgdn1QL_tD#{l7Q&=zO8()un`!qTX>IW=)J~oCEu*cD
z-43B{1OKLS5*(vH^~M6Bf`2xaOi3Iw%w`*L(B9z*EEiu(=8dj)_CkV4^&_73o&oVU
zXg_|=y{>}IO3k&wCD@g;C=pRt>)ei$uIm8qjV$8;T?dpMGQE?!KxspS0{<&L&&|sg
z(@ieo2;GRTPU3{FTy#J5Ma}RdFwDsUjDOo#e2ADiyt*fMAk@n{U#G><hkzZ4WNCPs
z0|+`#R_=r{-P)hRh;dvJpI#*mr$LZ1Cnr}<G%PP)nc+ZQUaOim==+d8ho~@9Dxr<l
z&`cI6-9^}%BBY#FI1_>3$j}-6cRiHe{xAa@kGi2=a06@14&QHmJj8%RpKD!uJi7ea
z^3pE{5pFMF0Y$~~1of+O9K^=alak6$rEfVh%YQ^@CdsnBHYoMnSy6Vz)OEGX%g7Z4
z0FL5_CEd;(aMgxyvi+i}TCPd3Xb~k{8P8Q`OrngSNyEq(8@=_?O5Js;NzJB{O1FEO
z7Va<?ecdwEYb&f|4HamE!VjmeaXI;EWpV~}mm!zizFjf;g@xpP699P(4SFvII0$+@
zwtZg!Hn?Na_0HwKvfMn`iou#wrS4Nqdb}L4U0cx%FoGVo&|o8r1HT{rGv&7hd?h(<
zjI=63UkU@Ak`NZ{i8NBoZ&R%P{J<UAw%Z1(J`XX84vee5z(`;0q6Gno`DlM+Jn<P%
zt-8#RlYU9xI?KCDq-I>=gwa@}lBlwKa%E<^9X*y%h(fE(DO(@zI;^ck5eZuj6<hI_
zRQ6P>5r)*UxvQ`okiW)VoK0aG5Np_6Olq43OZ10`T9~Dj2zJy==<ZC*EmCp@&o;-m
zAy>m{LH)70IA;yGo0-YJyOyjwhe?J-%PF~0QEpyVNsR&alJ=`)mP}U02x|+(7EANr
zU#Uw=4YK3+EeiXBFm?udbLgD_1VikeK_DSLq?^;409Uzs=<Gy~6ZO=4l~$be{nI2g
z@=|AZwPI};F>VG*e$<oi^x(uwl<TivXa(u2?IlarZB+?J=yF59KvgN65*@I~5KJ_M
zHnp_TDgLkn69pRp%T{jjn>%l{a9(V{J>yec3b*M-;O=IaWCGho!43>1nO>)FOSD?d
z5NV2bG{v07fu2dY+>Ns=HAcHcf!7^aXHphDb`PfEtCC!V-DKSUE*~a!g^5%UIl4H$
zIP3xs7!4*=LBDVz5!ZA8|6wAVr-w*2<GiQ+!9D0PLA*0Vm+KFu%g!+b1U*{dPX1L1
zV>(giY4IxIvW?$U3<uNA;5_L4{Zu(!uZCQ(x8?aO-!e}WylKzNnzQ|ccQ<eSLU8Z`
z{p!I9vN8C@CKty?8`Z_-IWz*dPLo(`T_^!e@`gY|=lyaDnAwM5&}_CQ=GK9l6Y_~D
zwc9L1k-Hl9XlFNbYa!1;HY*tl>c_d_Ws#Z73GM!M@dPn&J3Vg~#H|ov_h%1}WUh$E
z=^B<EyzX;uGe(zbfo4GOalirs%f^Zmb0iFD7(x{gJ>KgI@v^6-4Zfrj0RO5VWsx4H
z(}IqvzBfQI-Zt0HU}Mre610v`6J_UD*IvTS<jy))mC&nQcu?RDHgp~`fM0vhH4L71
z2UF$U)EfzEbMHEvEcmnUg6eSX63&iJ+k#Gkf&=+wh4SbvI&+aFOGPzrby?xVdl@Ae
zZ#Ql58NJA^QJ!UAfa0d1ws?Q?{(c$%^RSk6J-0LUmjdb;LyASZ1@{}K=InF-UM0uW
zN3cCMs{A2YthMz>lmNRKBo5H8OPGP->Z9)nftdvf_8<(wbH(}3V~e7x**S4f_pSJr
zRf|zC-_sh!PC(C}rLCzN4h|tgY`OO#JdwaC(9Lgx1I+8Xg0nkafQXyfTyrC^&%V61
zl@(@rh`qnvMA@e`CWsiF`F&exm;#OGU?#D89B)#@F(3e;w;{IRu+UVR&v}r`2k5M2
zc0ByU`J3I@CSM7Lki{DzL3J+T1sPd<3=iSsvKm^{1?eA(D3llS4vOij7{uV6-y%T~
zg+rnU&_yv0ar%kgOpbh%+$c|>IAL>zPy=U)L_7C|K@pwHE%iNkq`mr>)?4G-&_rvX
zR!=Tiy(*MFJ|e$a<YoK2r`bkD5%{|O!TjkB@!);!{hH!PG=j&vWjXMnh9jx^lc+S-
z?lj(<h9*eyIG)W&vGR}FqOPQ{(9haxy;6m0Hg@GfEhTC4RFNWTEH!Y2>?Q{fmjH;{
zMWfF74~3kdor(ylMoG7;epE1y1H+foH>gRXYXX+dnK%<OVxo6E0)xSu8lp7gl47sX
zYrmKwduL(E`Ba-!^&FiJO||8uD%C|JRU9=wXKs}^EcF<3!jij`9(wLyNwITZQk1cK
zNxh|6r92>?Li9alJRJ_E!9NAB8Ewy2G0X2H1i9d>teQM39#WCJG?76)ha?(nq%)M|
zYbgpkv2MTI^lG&u%*^B470REJ8w8b<+5Cz<N?N8uxmo$)G!FKlqnRzI2w6`0XB{5s
zWXwG0X$E}Q@1w)X9R|NlnGT5-U}AG3!A7v)PB)=C+bIk_oZp~3@Ss^^q3XD~taDm}
zqUq<Uyg3&)1)>DWD84+D!g0Nd6sm*MbKUTq^UB#hLiKRbY(h5+$U)PrOc7Q}nzYPy
z(cKLIz;-1#+GnwY?N?FgEfy>dc(?-P4rd~w?ZqZcfcGw)IqAuNh~(=DtCm=$CYB_W
zX0-#!zSFfhZT-gQV69(Fn$xo!cfrmeRC#7@dZjFAGov39i-b*e5oG}+xk&;kKD;&w
zhS6p0aMMi3&R|}b1j0upcb_=BA|P=FpibzDQ>4My$Qg5<QdhE}!!?Y-L3hsEP|S4I
zPKy5YrJSLfX-*RQ4x4JoLL;Q`Grb5_nkY4)TZiEnjbfoMIow`j7Bu?sa#G|804<<5
z|6F;O)9$DlM(cqOk6C{Z!ldN~u2G#E->ygf?iqBHZEh4cr3IXg0&5K%Nu}Hzv@FN0
z+45%cLK!8crrD?6BLhYnHkH)03<46Fw7wKNw^LA<S;%hj@?2Sp2l+@>)dpiqLx4Jv
z`X2zj1C#3|RLVW`vVUfa`8m3>orQ=`Tq95@>uX&-_oU19-sTW+8!CVYjkO?xQ)|UD
z6xY0#&yBLDdK;P##P))2kwyYI+a0v>oMw6HZ5&05dS?<7F;(x=sPPt}+M5~TPEaWz
zqF+V}NOZ+;#LUe3h9E569{t=~@tX^#pk#fJ{|(*PZ5fSfZ)9$Wl;#8YtUl87sU9d9
zbsfl>6Cb=gWB@7Tsn7~J@1X1c(5L8Sw3qoF85Sqf2Zm2mr12LeCSDDCCkwB9!A{}@
z>bO^$W-R{~C5?Xe2;5N+PkigvNJAqXoN@hMhzg*pw?Vl~`W|)}gEyfHMf4*2@pRcj
zL=}d52?9K&yLOy_l!UPo4TD%v`*n<w-}|RhnJVj_&jX)KqY892sc;E>XO1Smmb3vf
zXqEY#<}2{k(JK%7oJZ~qxaS)S3K4J$U@+~T{hbQvvQ9n^@M$9bMXCIn_B1C}^n*lq
zGznNkRf%nD!ycBc-q@r{yN_%UqcJrS%?yl}QoB+No4Ox7iq)n(cO4Q{7ZVO!Y((h;
zi1>V$gx9T8clM5ubB?nG{w%^JMH&3wuz(1EoL&VXOJuT<lvnPdkwI!wPO#XM%fet$
zVIXjc9F%gwcA4(QG6H?wn;lnB_<ZOcB3dg21bX$<ji6-I^RGl+!<XWh_^4((NlP30
z>g_th-ci;rBD!2GX2#;C{i&Bs3r5q@3ivS#_X3VxNR8%|5jp4OrMl07J-(4s!+@pN
z6(_);2nDyHbruqL`Tf=>QM69}P!k!;vfgHI_C|5q-I&FAZs|qayneW9(o7a1u*lF!
zuGxzk!JTOQ2QWzf55OhzLma|{t@Ga<ZM&-n2nLT2liA_r3fI;`>f-(43%|-mgBK!v
zA5c<R!E|Hz6(NuHyHqVCC}MAskM-9RR^jtL+JeK6>ea_e316>6rijLEp1e)MLF0|e
zQzZx>#r|#zVHtkM$1wkraF?ZdoL-zdS2dMz8XciGo*k3J#5`Ed(kQRNpa?i&W++e{
zPEg8-Xsx)f8gs?uQl~M^m7gcLado{ZgI8Cdv)JJ!E6jo;TseafuXv+0^t)gLYk$6X
zLJqn>^!vCw!I$;GKZ(Hdfs3slAk=8lX0}H138rG}YM^pHk*ls4O4hp<q>!bhe@pum
ziVf|Gi!?bNAR{|kQ9e1ZvFi8MRXKkPn&<uu!cxAu_G#i?&2iR13R%Bn1rs*778=&c
zx!IEy@eyg|k&NL&2jJRN^H^5VYyOqc&LWshqr1k|n?D@pB($woNl(kER75go)quyk
ziKxT-p4Ts%nL75&`;>z5-IssB3njGu5OJnS(AU#{=k({(+VNctPug(+>+!Q#%}GC}
zF-uvSqJ+2>gVn-Z+Ac*iw4Uk#R=(8^bZPsq>|cAb3|@{X5#(iWhXM<;-`EM&o>VPM
z#wp@=qZ&494H^bWc-7U?VEL225bR7R)PZ9!c9ju$niTdZJxTLO?}<V&7v&OVW$5xL
zj)!THwasO0i=f%&ctfBI@#$^ic+*2{i%TpjXrfH*GqTf&l28bZCr1;MKY<H#(Jsza
zW&%;2YM>Ilzl$tEDV?AwXjHN_Eg*!2=r(!cj@C9sG2!g-<8W}J406sj&!h`lm6K3-
zc@N!e+=WQ0Dp$#=Wz4}uTE$h7Fl5G}+BRwu_C;W`m!8VS@>BEp*}8o_`F`qEtWesa
z{hghNXI+44=6X{5EM*2m99gTrMPgxQPs(yIc@R8gwuOWoC$fuN@=^PFp$b}ziDs!;
zQPXIY&g^@eOMr~G5*p@Q_z8hU84VU_GVcp0l*oQQ`&YL1SFz7J?5gH%=RXzY<BQ^w
z)lf;I2KfQcJv`l_)%{SbKX!{6S;60dM7ZUBN?6*<T1MuxTJ5SOiDk~}@jlVfI2<_N
zXBQ^>kw^^kWN^|`88&!6(?l{=W#mwS0s;(*>HF2|DaQp_><F*p_6dQ%cMt(QP}fDk
z5rn}SY<vNV@PuywC%c5Oh%xt1ppg7jAYt|&2KCtuUABM=oHIcc<Hgcs59)bbDM6KA
zW)aodLT)Xe%UlovaWPeuW7M!KIq5hOKc8BtrPz}u5{gO2gR_Gxbjm}9*%UPPMMk_j
zLu;V#Eg3nS<lCW^;rR;@z;In0qs(7oi|)qoBc|fJgl74Y6nq8498zpSiA2w8)lG-t
z)q2Zcb2uwc`P%#`Sn}yr2i$Edg$?@L8Dt!J0bq%!7MaRfLp)wDg6SJf0ar>6c(QST
zsZ2PJ#jY|{1<KHJXR|X!69E^;?~uPA3~W(7V6Ua)wZTSRnM4NZ?OBUxYE(%Yfo#5z
zN;TO(wrn^Df-JGY<T7TO?4%M0BbL*K{G3#gh?P>G!*ovyKdCH$Qel-*4cO3Mi0}v1
zbE~AN)R2!wS)+G%(*!^_BX~Tb6>W7VcLO_e*8Ng94p}=ZBvElbrSi07ndK8Qca23k
z{Gp#u&pS{``x(+5TpFXSzuFdPwsYfIrcjho;)t`l+B#V2kGKg*rMnO_xpVXv`>8H-
z@wBj_5gvPK5bfi)rE9XsQW;3?bW6%ID6%M-#tTDN@WvnoR4XZEFHrBBe|{@P(!|OG
zPT)%hDSbxbwxUJf%;Bewc+0(dDAOP(H-;AcWcHL;^O03(X?6t9;Rt*JPm4!Dm49d5
z8;*za%l;ZS%;N*?M|`grglvtf8vd#3+nZ2NM@`4yTM~krZUh7Y9%m}0;#fuu0Gn=}
z$A-CGgNQ+DInL<QNxyEa3<N2I!PFtaoMhgri1*bg3655w@|QZ-6#rk8XLvlhT|PQ7
zVE-hKm#o;^>6qNkXrd0NM@c6BUI{s!M`jlO#}mIbOp5~dnv=fPD9OGxF~KhPmaEh0
zhJ=d?UId=)O9Q8J4l10>o?|df#`QGl&kJ0q*xKdanaf_VU)qjQyS9)I{y<7S;?~K8
zzP`fyP8$MWqv5OVT=vro^y#L6#kpA+nq?YfMXSyd5xk(|^&;Sk>UNZir<?P0F;@*_
z72wCABAB_$b@>-CU1?uvP(p%K7<GTt2fx?lZ3`egv2&rT=1ur$xb9Sh$xzU4+4(Ky
zpN(#Kx#obk^J`L-_rYTKah=M6!q+gWco&b<bS3Mldn8FW=ES(we^IHs{dVpHCf_$W
zYRl@~nFq??PK}ZR+}YHUd5?G>5obsg{E=41qyA;Z#90#)W1ZK=?p@H_-oE3(7C@cX
z{?V&3fdb<d0}--;(E`^rXTg75cy@;hi2)l?Ee$(@;8}KJ;i<j|th=0u9CI%T+>v;q
zb~@la5BP=Y;Z1`{?NoX!>()HhIp@3m?xtQ2(r;hb0n=<vF4JU_ejTb)q|F0<ua~Tx
zQDTqC`gJWE9c!YO_JPO?i_cd;7giO5xKrlTCopc!kmestA|Ic~Y_$=~uY1|z>J<#%
zi@~tLZ)s~Ofe1qk8l`*xgSoeAi?fNgMH>hf2ol^Kg1Zykf<th365JgccMa|qTpEYq
z?i$?P-K~M$`POsx**{_3a6vuQCGV`7HEoRXahhUT1Ao3<_wU~%DDx^P2>Uw7yaKsv
za8b_LNZDh*iV21ylTfNR5V7yg^sgOF-RVP@eJn=7?Gf+rkx6KX#E!rmiEx#mpe&>;
zKnPV){#(1(U-+w+O2Fx>pb@38p?RqWUp7<b_mT;LZPRpp%#3)Fj9zak^&l4453N7b
z`nKs@93*iyhEN`rQ_OvZv21rlzkd#J%<-UL44&b`!Hr%R^JVQ*R+v+<^hcN*gcC*r
z#P|GtymlaV?;pO*^u@@dzJ>DSPemt5X&#f)S=<$4nOpSX$so4<N)Y0?6O(~%pnx@N
z;}MOlA~QNl7onEOiYO=^P0GUYOH(0dnlec~k_S>g9zkuG0_NEo7N{W$w-)>Mh(X4U
zP39CUJIc@d7o&|TVD@W=x8tqDWTzeHei`A=4U+RGe&r)<s>?i8b(O>2Es;-ODSMxB
zHC%pNJED8%NeD2i&S2qT>1fc7%bCdH@cr<WG4^2>+&PFuq;;Pl5chA6xfR9r?~alT
z2nc24fVwTw3iYf4pbC=@?ZmiGv5b56vWK#}z<t5w3|HFL9Y`y;Z@fbkkeC?A&fw3?
zVbz@|QOHe|g>w2mosb~@y#|S*Kg1rq>Jb*7_Y_&Xf`NL8Ue6LuzD@|MtHJin9WmwE
zE|?#^BQ-PZ;KG-;xK!|yEuGY_l*d!Bkd{{i;|pJga+t==oe@WHV%r}m<(^Ocm`TVW
zfp3^krL0;R5q&jAnWef>(`o!Q`SRV-Qv(F<(Bz&G)y$e%H=O4Ei`$D9VSl5;JuMU}
z=p`r0r&XRTenSeFjOD~L!s+Rm89&=H@UU!;^#vmilB9;CXk+5JkxO|@hrPs+iTUkJ
zww>bI@|@)2dLt!l!+7x%v6b}OSoJwp+qJib^kAdi-t00~8z@4*KKZDUbrW?6nn`+r
z2DXZ>jtMqGS^&|0biL?RxO=5o1rEmuboa#PgnrQ7Ej!JgXU}BD2nM~ObWMbGv*l{i
z8=>wGu)PdDt86o*#v#~u95z2=EKZY&7&mPU$C!aoN$2vbk}g*q^DP|#&G-Y}&^slt
zl~yOYR?WV%+<Z8nM_q6qW~!qkzh3rBfE_zpP+CD&$1v7?`Ef+64Lfs(V*nThYq=ko
zy4JiduoqqKW<Jbbuyv!S5lx<MbJDu(v*I>Ym$Phybz^RqtNq$(f9%JLV;P!nELxS%
zW<(k)umMvp&FAeSe%gr(pTT&M_!d=(%9}h0%TErS7A9DP3iAHebq>W)qV}bLL+rI<
zK<Dya>d(}DD5F-D9l?+!619}PolsOkUW|cY?fYX?MKe&ZThaEk=C@106DzMtNOq|Y
z2V{r$p#eBUZa?F6)Pq+&9lF6Jp`S-fx{Z9FE;{_ZHvBepUoLWEY*J`+rAY8CS3~}g
z7Zju##3Efm-8xP@A*pzNS2r16Z6Z_d#sRhn!ax3?uYXfIxUjo_;v6iO)nz|1<!y#R
zhH<r;i9yRz5_mms_i}iF+uB*hY1%%St}jq7BS67ooJ`JA7xsqNZI{PYsvhnL4=Dyk
z^Vn;xxr<v<%+4WN?&ZvM&{JTc!=$qk8g9+{E&5oC5_&S^UZVc(AL=mzrK%gBAUbUP
zc_~Hi!wvDP$UTHNAi-xfB-+vCP&Zx-LWYm|(eE2xCEAbR8A0Up$Y6Z%$(_|~=#AP&
z>|ujb1nQG!Jp#`c<8qku6~>-EZQQ)pq*jrzO1g(%z#mffevasW*Q3%cb|eMp_>Q#g
zo#}7dLry<Ak8VLn&dHDo)89ksDbC}M>eG4W7y2eu4j@Gdh?S3i5B$Qedd$lZRi;vo
zKb$x=Lcb@26CPBm?c&2{D$8QtA~ibPi6Qw#6WbBdaW94v4GwoN$wfP?7oC+u_}8w|
zh5I&&#F%D!rAljFxQEow6p1n?M7Ue%O1xXgagDUzXJFJ%PUkE}(8o-v6p>8bxj@9<
z;joGva!eSUusbfl5L(!Lq$tHM#}H}R|1d{hPvY+*bg?FAFx?IPVmB((gi@!2z^k4`
z+BsM@pjZ%>H#M_&>(i8?a5!*B>3*9nL#gKOYrxJ$aFA5Steqy!EVRHh?Dio?#9VFZ
zaDVeqio+b<c68~lN&AC1y_`=M-CaCv{LS$Psu)3Iy*(Z44W|W&NY#>)y~^y-l!w=r
zS)s+Svw#E((+-)_qCf{NvW<<Y8jxGuoG6m7dV5)0B4ZpiJaVI?dVGB&Ar31oUP!Bz
zm`U|F*;Qv#n(pwNc&G?Y-`YreQhwraY>mH5HOsPpm6MGWXLfU|9DW^-cZ~-_c$UW`
zp>|J~T}d$2k|8Hw(_YOO118knAwMQRK8FeR_PU;6I~n3Vb3<!U&0{eS<`;<j{Vg~@
z{L6QSIDWmh5^Zk+W>z8zK-`uD>V(NYZ(>*vOuUqno>s=xo%5Ps+yE&Fv=}O(PH(R6
zou4k5svI5Dv#^@%-DiepS!~?(lEpM(k#7oF;}huQYDcC&$$pae&jft_!B&>`^Y`xQ
zw<Q(ngJ33E$s|vxKKEFOU^u&X01_`%4wsR(MF@1}XEJ0tXM#WKsgH()pLArLwKw1L
z4egLEI0TDJ<45|@gFUN~R8;3av3Rh_A_SCxqUQPvrwx=o$1Cz#k<kc#-nU;(i2QQ8
zQUj;%Xo(givZ^c~49OB)dX6Ei!;({0{i@k0H&+$?eWF1>zJ(iber}9BNT`DYDjPyU
zxW20)y8LieKz!+{0WZuO3pqcMG#HzL^^isgI%f=i4R;=N*5m!i<|e$5h7%eF4%ETi
zTg0iFU)KlYGA}Ro2c5rtmjXxUZ{JoR=p;X2FuG!LL{(kyA<^5doi#`v-harV|J)XR
zH+lAejVf^M8i+)e-;B_0!+Dm8zN~ZXN&fW+<_-9zU;P9%{SqN00QzfmoDd_bVh2K7
zg^vhDFiNsS{I9!*;Un&El|%(UBiuXK!%;rkgRoPL1KOy3V<vEsx|!6S)DO3-Q(vJF
z{xZNP(KsI|2utoRmB&$;oTFXNBp4}oA0lvU&y|e{*9B@?LGULDBnN_1blyrtWz4Ct
z4v&IPVv#1Cdkl3YyYI<x*SshXV_rPI4FWjFHyhIZrJFege3xjUN&zHW6p`3O&X;(w
zrzo_n1Pb<{;&j+#mTkK~e%62EpGHEE&GIyoYu8APtVL+C4W|p))hg(3d6i|;O_oWJ
z5soIdUmO7lKrGUIir#*9-s^Jim>m0t-VoeuyPgK}a!@;gGnH4Uz^5r~6Lmw8r1Aub
z$d9G!k+~AC0&LoOddGAvO`wFBF<y5kK#J-th3ZNIn!fJ`w~twoaFm*M{OFfVU~sK3
zrAA=mH$02VA5gmh{AoYborv@Ddw6GJLgo~vGrn5OTB4;v6{qn_%vsa(KgI@M(8S*6
zP{E;^jkPuC|Iau3zd7*#Yy3`a8%Mfm&qG$i2!7T*#moRyoU#$xu7%XSiSF({<D{}t
z>2O%Rf>;;n7v}GP9xA8EcYp?dOyt4ZbMGOG--hKUNp~I~XTycdz-yB!NxJT*Tu|gA
zVe9XDXqs{Or;tzJg2^Vm8+TjAijX@UzH1V<o&4>_o+RRU&5vF?mdaeE9@M`XwW#!!
z<CC%?e(KAejs_jqLa`^MKN88P?C$OahX$`(Tacd4cFBy3B3waj^`xt(?~hN??f=Ny
zp450|ZhI6EG8+{H^%pq7mPd9@-T8XY`>necVG8<Mp|$XbJfCcDD42|jvZmU8y}45g
zVErh%Wr)`7zEEBsl2I$0kYbmDUjR4I;^np_vrT|-RFv83ZDx*J>Q_bc{Z{aa5m%Ex
zV6lQ3;J4B2ZMXB7VT%Bp9o6L~7YaYf;1)5oFATG?<9oLH1dxvQj^P|+pdBQZu4}~z
z`*(aT7}_i26hxyI-CdZs>w76SJ%msRYgJxegz=72ZTEcN1lTOe$)ihcd2R~aP1HpY
zZ=EJ*uQjW<JMItZSgbQbf%4E3NShe>)4P_OGF{P>X;kV}*;GEZrF3NU%*qV<@;GfO
zSstn`w6qB0qLi?T0-Dm)H{XG<IBjf>#G+bTAya$m#!DK9k2Bi4!waRSjH@WCwoo~K
z_YJQ7p`J6-5a4BJz?szsgl9vm3&3GK97L#E^8+qdOV7<c#J7my^EabV<YVH7E)o}4
z#)AK`W;2_vZtaY}7}GXt7=P6gLs!u9)GV~x&5I~NHvziEfUF>KylTH1oPz|cow{(#
zR-1!snBo<h5qJ;~$*RAGAmneRr2#YcNMTKvVDX$i>DT1b=9q%f7^57$^O0X?$tEb;
zna;yA8?io{qx2vhKNRnW)60e6GoRJ&qj(z>&#T<NC}*rhTiol5qC_o#$4X2SlI<kx
z#+*H1zg#Jkg*3ep>WOg9ty^KH%?Eml=gnkmU5aIb^-J*&_$!pRfthT9Fp9~K$%7x~
zWQ@v6+6aA&n%ylr{Z>FS;AfIn_X3ock)&baHiD2XhYn+yR<h9;J)KkN!Goz6(jb(p
zgR+t^)SVdtKej0t9XPK?j&)waAXaM+emENVD#`E`-lv&tj$xh{k-5s+_Ii}BDGTb*
zmeRc61$df5DZ05O16$m`ppSihdX0;|9rA*SiK5*7#AfZ&Q0Y#^yZdy9IJhVf8hrs0
zgh>r!-JTIsl*64#`tgya_hRm@HzAn9Xzh3t%6`L_|89ROAx>b?{~|^oudG)B24tXQ
zz&Z(H+T8pL$~aM$>Iz{De4I~IR4nPqQPcgS{k)H7PWlyW)5)-+lwz&vF%!mE;u#-9
zCVC)qGDZl`T!d0aXiXyR_o>e#3@&-CQ+mJIf9S3^qf@)NL(^?*Z@XyWQ}@KyT-NK?
zKVf>cZV-Z8_J*IEcbn_dWlZak(A%C3B<>0hbm^}ZnUVK&l8Ci%Q?i}dB*h1~{!A%4
zPj*tk+v<Bg2>|6@z3{spb1|Pyy0#bu54!ZR&uZ7JI$8}bjOK7S6#KYLih6&!MMIkc
z@O62e&Esh=h3yYerl$}Yz?b$!y=iNZjP85&gYiu8ydPuIH)6ES;(o^2Z8Lzoc9zCs
zv0Vqk<=J-(ECxRh2BessQdn8|fP*B#q7Ufjt@sQBpK^CuKy;-F+Kj2+WlHmXL|7^m
zXM&(US2-KEbCpm_7Dr-!SCoV4W?`pMp_)S8iE=X*N9e`=c@^1s`PHY`PrM%k;>aCB
z<r)_TbcgyhrLS*6F!Z>Z*THp2M~jEB4MaR6)p<wLG1S@e1;v6-N%p~Q!u)JIz#3=X
zgRB=P0qt}T*P{(jM-XQsQIp|;msgMrib$2~m<dLT$SKA!<3g3Uoz+J0g%R?LGoT^Y
zWS`@<|E{0^wA>Hj1SoM~k75wzNT65Vx3PV+P~aB8;0Kc}(#Nqei?=?1x)s;SJCtqG
zaTJWM$RE<!p>>qq=G+j(_UTE**c*Yc?dX<c4$HYQquCV3z7vWtP*CfJcWMXR1|E$`
zwZgUU9Uh!5Aq>@D=iaHhf`aPU+$eOfFQ$m4OCCSEr6!u`xK;LnA)9eGYgs~N%0t^S
z!brj0OVReZ{h>hjutB``!R;i6$S!M8q<U|li%1x2B9sUPf9pJozJK_;6?_vDcH=vs
zp&Bmja`s!7ouEXjrw?y(Srl#WQri!KL<7c!jx^hCD4!3K5A!d}>i&~l76)<}gGp*e
zDa9DikGZI>+V-d!He&LOC9AeQig%W{o*{UlmP+~Ad?@WLD{iFCC@qV2<^q}v3B(kk
zABkzO6~6EpFiUFbG8cB)+^(OAP?{56_u^lTu1ONz70Y4T|Gi%)%kty@N}T-J5k9Yq
zh$IGU{2yGK)gG{zk-bBd!b-cl@YcxF__-#xr+|HH6RU_=AErYtf8}~3cSZLshachD
z!|;O0o4D=8`p0=B!BU9Ui&y15e#n`1cl%%;;SKJ>D2z(h!|jPg5DZHblb$X-=?u$0
z|8mxj`lbtcCKL$P$Hir1A~pkDTrTwjlS2S7=Ms{;ThX~$pq}dcGyVKM@NT4>rp-5c
zO`|HNjcc7SxW3?Ib23?cA(n|sTAYae4C?O1!`{N@;XF{!{g*csA<K3C*r^pje`jg3
z<JDRLYEU<noR?c*NoN8+XL|;SO<hlH97<SYgBjEnvUO$a-3IsDq7NQK>{8Ue7e^73
zyq9%-3tyaq{ROcosXO+Qu~9IfNM;NFCA9Sjo1*`o65j4!bm-J*la!bz5WQ?ytry>a
zS24q^+Z-XYJqx^8LUnUZ(G_w7KWFAjdfxsgM3>Vz_z%eu=T+|vr%%l~)J}V(nw{q}
z389+*CDHu3aj5$bZ=@VGexp_)DbS#epNxQ_AF*#}hk^vxV9Ix_HTcg42E{aL=;(8i
z!cYMr<UNnkw=)X91LV_v&F4gRo7^?ngcrG&XBV9Xvwf=e;%wC?pif_0n>e;tgGmIh
z1R5AtOh)SlEv(G^IpIr&oT~7i`54@YVGa+Lmd4AK)2jFO3<%SvHJOM!Uwo5?Je%{T
zLuin-n~lNt;sh~U=aZpz9cP+`MuQ)FULR*DRnZ7fVX_+=liOc>@M85)6&(WWzQ_2P
zqU#OnyUs&(vNXlq&k5$3_FEBlBojFrt8`eYoSo&+vzq^bFLiS`*$Jq_PFib@QU3&e
zPk6SIfK6_Vl-cDFelzH=C6<9Lc@vLdeGF4w3@l@{h6wXdg|PnL|B^2=TQ6#(6!o(Y
z`z~52Fqu=*S?M!8o?qj`^@)6k`jsP@v6Y_QBGDoj+Cv%Hu3TyM#%{{Z!M#}hF5jPW
zN`dW=!27{<l(tTn9u&dKU@V~n0^dxr++!y`z@CYv3X?d_*gToGq|!|^VOHC^GLY%0
z)_s%OuoKQv#+dL(c?++`xi9|jPdQN|IH-IVrEC^@rQ&h3!l>}5-^ID2G3T|s(6hsJ
zy=ey_2rW>&eqigu-H5*&vA<jRV$!H5kS5~8S1q^LvbXmfL$m3kZo4_A8TA<rrjk64
zB3$Vw5kCgxH8*L>!c>Ew6pZi|ib(R=SOGT!l^H~*1Nu;j&^V3=j8Rl=>4Aaie7g}=
z(^rfr6~7%&BU|19Dy;2#N3-rTi6sNSe^H{T%7hb)iQaew6A|>z2{AH@7bnLTb}!kN
zlp~h;tC;R>{Q=aNP#~D_<Dh(-l(In)qdO?+2`)#@{6w_zY9eJF=!28%go@Nd{;-51
z=5RVI?mX#bV;Z0cPpqK$w|+1K+?+3Xw(`n#wockFb&_*vbvy02in@&k{3G@bAjK^k
zftY6h`=0s#zuCBDIU_$c=fPqne=OtsIS;cSk@4|IjKEf65N#JdUkV>HPSFSr(imw}
zO`dh)OotselFcC$3Ugw>t+e7cDY?5WS{JT7(Y3upA$DHG7jkrAIaXu@<0W&1-hgS9
z5zCN*Fk?C!dFd=UatvYPBQk;_>VYb0|C+X^EkAw%Hvw}Cf{9OZ2SsPB{J3PuK#C7!
zvSPt-__&(J1PonMQQ(_FRb>f0R&jlba6^(@m|;hD8#4lQ(Wkr76yG1w(L=3d*if_H
zo*6oRp_l>&cX!Wa#+0mTMEuY7M0Pth8KE<&a~Yj#*yv|x<4r9Hpd@rvVR)q<WY)-f
zzC#6aejmiH4h4W%H*gtAjs{)iO|jTkEgT1_Mf!ZlP~s?W^1ro++}wjJcpc7;*S4=>
zikW&j6)cLO=Xf?ZQlFv9je1eVwTtyu0n>GpGk$W|7J5sTSq~d6&9DPwO35g-e4B8M
z=vO<EwBJjJfp#-0RsEn3{|;|hp9+8?j#X5+R2^RO+0XT09$AYOt}ExGBV9f;e$5Xw
zd--@Gn6Af%8g9S7d;a_HwWK(sT}u&m(Aw>xYQ}#d0#ze)ub1Y2V$5L7H7U?+xf<AE
zrv&~pb-jJq64Tb5g;P?_ez*J!I_qYRaE!JN{g+TUEj6p*x6xs->(3LivA>*2WqP38
zY&LPa52^@+p(-r*Zo8s+rca_>%+(H;d|dV$bLNpkr{4kE^u49ZBrA-AS6ms-`PYJZ
z5>S4+LZQ!VCXb334+Q_98YJv0i2Z;%6+pzC*S@J{+=Q8xj*?t^3}F~8m)I?KNagS&
z1k(SC<hs}go2*M7isY`P_cHDB9WVuU)2^9)@**F&hdxBmpTgEXFABH4VDF5qHz~aq
z5=rqKIPZ9UtUxss^!L*xWpI2O&UqaiteYQ!fb)}}8uc>6s?$`{is+4zG5`?`AmEEy
z=K!LYWnW8;a_&MWfds5Ddu9<(CTvKWuZO|fJHW#%^{OM&nt6<!*9>(wz*0iB*nrTh
zTgjePqypQodJA0JdGrR2Lb1^^1|GAT^LyS}|5!ZFE&>5dNUY%X^&mcetGhc#;i{SF
zFcAf&1kn4KIFmfD8BN}dJpvE2ITyGZ3LUluIvJIPH!TBEAuXvoPiDXenu63L2Q!x&
zBB4<AdkN*Pad?tD*ielDjYnHl4icEfk&}I)J&E6h=t@(6wX}?8At4z8owvKawy*Lj
zyyCQ_Sd|>3_?>`=PAw{w(MK$e(VJm3$Bs=zJst<kcfc$kSJpucLXH*{@&-fs4!Bl)
zUub^$Xkur1P-u)V*7nLeUe#VoOsFy`1p>SKT!EcPLYxfj?_wp69)+jfth|e#s%g69
zYuq@GKI44{KwEh~?Ujp&fQ28QIvG-<rWNO<#Ms~9)nOaJS6>Ion$oP6;vzY~t0zi|
z=yO@6>2$BAiQLYxww=y;UTs4>SL?j2on?P~Z5y6FibFkV=4Hitr9VOl5?Y9L&Zu&f
zUL+E!YyPGkgj)0+f4fUiMm?sR>_Izut6B4Xb!#B#9P?Hrsr^FudFpEZV<R|GnB*Cr
z@b0!YXI!2a!eX+@CjLD91$YN*w}k@%P6>^?t}njH;=cki)X>a-gIu~L^lm_Llh!h6
z-i87oUiphTf9V{?31!~K{PAQDPtG5~EX`+$x$+numXS;WnXLdz65x)R0rMoKLf0E7
zN|i9hc#Kf$PL+ZSZc~Q6!pohO*oS(yje)XOM-_=NH(;gn#K!XV50<yD>@YbgSq>73
zEK;JOua@N`kKZts=$AIE0+l5BCqrmJmGDauTLB+G-|PRVJXPthMitsL(Dwsi&lOeT
z{{8K_rzlubg-ei!{l=kjU7JipZ?W&W3e!|s1?*QaiT&{XE^5kuj4ElX$N-P{?>7JL
z!y)m$<&`tV%9ykG^FWET#r_EjC@SzJ%xOpx%zZdLd1F!1o28Qyf(0wmwTIZKatt{S
zwNg!Y-!=}<t#YoOVrNEIlUcjY5Ru!~bcrmbZcx+{P+XzTs@Do~_KBpbuRtNvWjw@r
z5D|c(cQdDx&N7(y@ewK4=)Li;6agxLhwoyQKqqE)TMUZU#jRYRR~Lv+V+qs8B$3h%
zj8X-r3Ln(n<GefYWq1czjj3doXNN&#ODRiWPRr*(rGL!s^{m&xr@{feZqh;Gtb$9L
zM^Kp5;`6tUNLXmG1;raL3n>&YmJotm9$VoX6#WkHpKFhh>J7=;PaaveBmPA49I9N%
zR_S3J0S_}v|2(J7U#Bv#OOluD*3{`lpJ*}~3$v0xco#PnNAMI-P@-B#G5Q?#Gl%H3
z$l;dMYhQo`IntfNzR$0CfOR}>P&TL@yqYNbRQhWUnR_EOArfDk|N2VcvBMskBsno9
z{K+8UY06LX6e??btPyj84u2tsx(WjPN*wvCL(w8syHJ&ALF{ji!e4Kv;~B3XH?}!S
zs)`B27adv33%9p95A7>+EvHwv_sW0n1dpygzCO&9X&pVaDCWEaT9;p>8p>F9W?#y{
z*^{rg$DyZJLZ}Y{1n&S0tcynB;)ET$tzuQ)H`_!{A^ugo?E<G`pLxZEKae`hLX{2h
z+ZUA1L$9kE01PU~gCS-HLbA|N`MvYSTX6#nLS?qqPU^RC%~A|IP@_vZeUiHH8qAgX
zH(I4w;<`Jqs8Y2fhf`RA5Lh#P()1Tuk|<&2l{R3uRn9`{*=o+WGwMU&D1Fi3N{$nj
zcIZ#%eS=>83WL{9N~l}vz$;eXiac41Affbj^(`YIs@JpqiH%M(h?zt)EO)V?UXJ=k
zLVZuY@?pF0F7QrGaai}z?7CRZt`Cz6AQ*IsN}ZFF;xFhI6Txz}QiOP|#mA?{S0!{A
zRs0H=+a9|OZhJud?n=bJ7tl0p`mmu=_J)@D_K<qOqLV4u^-SLa0gCrdQt-K>D{&!7
zfS?F|izM~zeNa>VDX<7hWwWdzlIrWIQ9u&n3N0y4X<;yt-XDCnp36NJBqzB}AfFsx
zBDpS5dn;^L&)NOu=Au(B@(u`<hIeTyf(L<;<uy-^2G5eEb0`x!eOek`cyeA(9y`Aq
zKrn@}7S8|s??}@NPxd6}riwTusn9@w<yb&jQ2N04rNQam=IhIXnbqEkb3azkukj7O
zCiTIJ%fKe_@dr1jYX9*5c1fOH17<$*p)=~V(MPK}{}Vw96t4uMf%VGm>3qBn4`_Fz
zv&|t0CZ6!`8zCR<DxpX3jK!?Y`L;jm?eMk*C|?y@nuf&}Hq<LZnWHE-+6USn(peC%
zmAM}l%7yB7f$c@vG^0zZPhYY+)XXbdXMLrP(B&aA&i7ioAN5!~tojI)FRObIka$_P
z(u4ng8lC*@T1}{<a;^##It&9&Tx7uw3C#cw)wQBzTde8qmbv+8mit7I`$3rPHmr!5
zNX{YbeHn2_`#w{&kWX!1`SuwT{taZcJ+9A4cKr@W;x?KYdnq0k|DU!5$alN?!MX)W
zp0)e18mh>CUc6i6bOBQDM3ez~omO?8cfSMZY{iPx9u$A2O?I?vgQ;j=FY2pEpF}q<
z<9GiXY3Bd!Vp0l3eew<e^KW{BG8Tf*cUG5)#S2%y<M@o;gW%Ht8l;9*4>t7w&F$?W
zL&4=bwBY<{JEQvOxbj+!AF@*hK@esi2#(Ie2*$pV^MCWIIHSv6Q>W{?D+!0dmTSQG
z|EX(hJ%sex2xwL*UAxmRk7bVf0O>V##{sfHrCj^Bm|E{-dC1AFhPc5$=NNkw^n*K$
zxXEV5jeV08boS@8D=xsK)OT%}Xt$gw4({F}&vd=9-_ILo??`+-COKJ3=fp>ZO-3cs
z88i7QeN{X$>t|`>6VhGr9Wj-YQ*oJkW6<(@qFD63N&AD*^uo@8TopVs=tG8Jx(0dr
z$x*7&i9gl@PpE2Vo&Bw;Q04^|Lq~T$nuhrko2dM7S)Q&9Q>Y{}<PNqTbJ&Fo6)<sO
za2~d%x>tox*PNq+zZlT<N5HmU*9GhdJvGwom~JGl3lI*nFCU(7Y~0qr-b&o$@1FD)
zfgd-LF9!N_Tj3w0n9ps2zdpgi#7V@L{r80ZUu)M`2VVWDpmG1!+XSw#;XEPYdpr=d
z=}lJaNt(MxsP0A%teeS+RgeCyy@|97=+ju-7K|+n%he@Xs=;Orb(j+NfRmFD2@&w>
zS-&UG?K!?6u`6EhQ^lp2oDwV|kEIlNNzE=3MYy?zuMDd~t88ua8Iwp`6zhZ#Hd~=v
z>V`f(xVoPkGm2HM2opEbjx8t}#g+|M^AE0^8}k-^a(NNTp!d@XTS|ZYKK=qFi}n9)
zBSQ>a!Ono3?hX&i*XMEs?@bc?PiKc6dKFsm*YWFQIpgJ<6}-Fs@iJx03us6&`jJ1^
zfPnsD*ALa8jGehwx?HHF3y1cnyBGe&`fugzaJGaU?N|~UnWV_t=)Oua{*Z8d3F&5D
zL|NNeZ}yfk135tufA3s-tr1U;=0Sh!ZX`krDK8+5gCqlbkZX|S?;rPCI;33w)>_MH
zAo~s8!x}o5-sV6=XR3WMP*n{F;5?)q;6zN(bY2mO7P5sxI)@E~gp#EvPr#}me;SFj
zHA*&IYg8Uh$GPxHGH?z%f?8vu7g*H2{bb%l&chvQpr0kfh%8k>E^xMXF)2AJz9Awi
znfSalOFlI&KO+520XJ2C1P2>OW-y19H76~QhK!7g7V+`MDIviLwb!4Xksg{*dIo)`
zFZ3fa4skfvgPy}dpy(esOOpZro3WfZ9G#sV6IeD7GH<5QRtVC0*jWqbe3~FBDFlas
zy3m{1J5=XqRh;N8v|zTolGmL^$8o)#NeWCZ7onL^kkHOZCcF}mfp=A^WfjCND|o8M
zw<ibo!2AUAc`iYAH1xBxHr)yW?u=yIOg<SmfP9#(w4b<|jPaz@-Eeug=?vn#v3dzL
zDc%8f6i6!#RrOOmW=6_>tKv3R2C|k5_WR_;lD_z{`k1uHcAN1h022}<kgiRZYmrEm
zh*haD&!Y4Elded!Bs5LoPd&Pc+1KwNu?_E!v!ea$pNmL^GZXunhz_Wzv%<g?tChyW
z^0H(6aK;=J8o2{0g1!V>DCQsB(P0O(7h~ZVXez+mS#A#u{PguHnPE0=wvz@;!O7Oj
zVoWo}EtT`-0*kqkuBI%C@#MV9sT3h!sG)Fn!*Gp%_!}fybyo3zIIOsTIX<e<CPnKf
z%3g3^Je}DL;Ssl&AWWOdO+#kw5m&;^gZ?5NhD}=59vMXw(wx=%r~{$?`pxol;K1$g
zYP}u*3w&eoJHYR?2%h=n<F%*JJD_baNzXlJc`flBFyWK^4w%ufD?ZcZw=Xzp?r}He
z$?m<>-<a`0gglitU?eB+QEkeEY6%o$-$dK$A<%~!Lh^`+ua!%X$9}iuq~q(2TPK8m
z%Pj`+Ze>eYia`-FKy4Kn+-oO_!#zAT(C~id9l*hrqX?GpIEj5jx={EA=?uZF*v+u$
zyd7WD1wI}0M156*$Kg>zeFu~-!?Xd@{}T9x{cy7moa<%nTx4tz0wdT68ocbZg$h`&
z^L*m?{-RTx^E6opg3LXm6*5oc1HURu!Pd+#>ky_$sEt=HWo(GO>3Y}s6`loZIA`%|
zgWa3hJ!B429>A+ceXK&UBw)h|ZRmZD1*CT|w`(W(d$}Py%E*RZ2PW_|GuQha5Yj@C
zYWCDH2ohP4{Pzo<ry(=*{^zno7zx154f9zOU_&dl&?C*#vI%7Ukm0~i!x_3*QaHHE
zYvE;xDer)fB^9A5`{PZ1a3`L2kZlCsi2Zzo0RN>V6Jy)3j^A-Ao!LTgHxC3S9t||y
zKG_I<b*GDAvM=IBt+EWbK}|AMw12aHTwi#BZOnNps9y$d^ABG^zJRrFD(!34k`VY{
z>4M<+90F$;bg)8;tZO9gNGPwwa#D_9-zb)Cq3d+I*<0;p*%uwR419(4P&))ZWF%z5
zazbXlLIbIoUg3WS6jhz^XnP4@Kt6p;j_w}J7I<zNyXkb&qK-c76nps{upcrDlkytw
z@m=6_L(fU^%yl@EtH8n|@)+_S*~$Dv3Q0BRui+Q2`#WcZnyoQ~VprT+MijVWyzx?E
z!v%cVZ7bb*Vdvsm`0zG=TKwj8dQ|+T-P-whKtctqTKTB`>Xe9aKQC?vh4$uSz^EM^
zu<}nd#C|q&4trz^f$4K%OfdQNYOUV^%Qo4_!0WlSF^zkWrDX+fLr+-j+ff-}8++MM
zwyuJk)iS~9BxK65Zy#JU)L)0)(_`O?2HO%|^dPb_UbpdVc@8-T^)D6vkx^e-Qx>>I
zyy)^7e53>zNWXekTcfPR%f~J|LAX))hcfGH<GUQJhkm$rn)+AGS|7dxPUhC`>N}Sv
zB=p8~w*;TI9Ror<KdDLZVK?E#*yJ8lJTsN|R`c7EKoCdHN{mvQFWC29=(7OcVjo)y
ztnS4jwz68C%jm-zWJg~cx)x+Ie~yI(Z-4xq(62NTJR+Ej)|KHa$~!_UtyIF`FZ2=f
z%QR)d#4m-3k)Dj;j{>+gRL0`2Mnev{eV(}d=-nh;ZA=+|53?DXD%WgV8L6Rg>(;Tg
z?s*5GrP0G(!yR|lDAAF1T7L)h_cAZktW@Ac<t#Iuxcs`TyGS$liJ7LJp_D?h6S<+3
zVuQgXCp8Yj0R%{h&Q_#^$@~z((ojkH^6Z@C&(2JPC-+AMgL24ld>m<FWTcN<Ff%^R
z(Hz3B;;OG}4$a|~<xa+P=@;luPCRJ0>t2VoTZQ#}e<`!qL(Nh>a(Iw3>cZD3<}f~o
z5F&E2J=^6RrtujiGuR;FJ3R3R5((vjb3{h_0b)Ey0)xHAEC;0Hywj7B6)kpWhwE*)
zv?dhZ{GNMmdpSN6c;wC|8(yRnqvjGZWmMtl+GZ<zhf#$az8DTg*P^nxtI~j{k{FSs
z{(ZcCiheXlOkA~=%&(s*dQ*83yOfb8+pWU*4iUl};Et$A`DXa^vPY?1ilzlk*?uqW
z%0@Up%~0@H%O=oiMwkxmga@&RAS)+`AZn(u>83eM<Uszi<;IvU%@ln8J(9wNf$2m&
z9@YEdk9`lavJ|0G4_PUxnpJjx6(+o#;@17r!F5|8pi`50(anzVC1cTO(Wf?yXNj?x
z7rXg_QRp#6a0L9lfu8}1sA%cZeFF#h4j}wOsJ$h{{95<uv*2PNaxr=P5QfTs53qYd
zY77EilQm#MxRHnsLK16_rYGoymfMqhU*zlF0nf%4&&$xURXZPr?v;#Q?$<<OngBzn
zyk0I)7WWu7RuHF;?=0|;e#}#ggX((N;LLmkc)&IB4miqykA1s^Fn2WdXaZkVwl>Q5
ze&ryXBM0ul6SkZc{5YjvX?+Kn+?VscZMWmMo%(lVBatkhB|sELbQa{&b+Z$C!?T0;
z{Rp{pse%Jfe)mpYAgN=*teiE!0~iL~kdC{ny&#ETX)E7j+^`M?QJ;%v)&3OUcbdKf
zZgMO<-d<WIJHfXcye0~WcdJ|M^v02>f!W7x0fbMQw_bN=kSm)J5}q7;Ev;(8I)Gf!
zDpOm7ab-ZOd)apY@4Xwu+Fa{^@FC$VLnc7I0zuzLA*sN72XM|(>z>1>8L*<1pLhvX
z>)1#xz82F2_j~Z*w!RII3Sn!!7}NqGw?Tp8i{<&CBA6cMLf0g+f*Qh;C=v<*(u(e5
zes+s{gex#?TvUZz__1#_`o2>4!t2O&kQ-MKe7Lmn%IW$JxLQD8Qhh$V%mFQk{nu}n
z6ylB%g6{U6SWO2SHZJ_rc?Aqi^rHeo4=IA_yP*4CQ^-Z#DmZcT`rry_UT))YQGY(S
zhis=AG0DFb?c$Vgtq#uRa6wdFV<|sLJ*33-5bMBi&veWwFCVgY#eZveeh1v0pFx0~
zdS!fDqiBFj5}}hT9<XoOxAqrSuYoJdF+(;IF3ljZRETEL{RF9}5NcSVwqN$i`kS%_
z*{MMY$@o7llhcaH<3ntu5eL!ZJJV~~8t3E15Sanq2w9@?hul<hcRp$i{JG9ObA`x^
z;6&_O`&uX1r4O|wcYEg}$-NT9vaGGjV}^{<lqt2&2&~Sq${w`gIj~<h0NeDpIscEQ
z$gu$4&aFUn{7K^+434Ja!bxuLB`RM_&_V0gq=5mt)oaj-CGb42{&w^M3$mzzmY*QX
zw*~7Qh^Dv@+HkWgX>i^qP^#sRte~;Fo7#GQ=?vM%V1{^HAf(xQ17mMJXN_KD^@|Xl
zK|iStykOGTI{x>v;js~%1Rj+)y#r#rKMvux_1KEtO&dM3znqgv&0HQd)Ir9zC>lHy
z$t!6FkrgE?_Y!`|fe84dC6~QZ$3NTnMX#}5Yjptba~0IxU!zNS=GyWh8nH0*tD83N
zy*5o(RGX`}fin$Od;J*2;9sB|!WzXh-=VG7vZYE|U%gCt$fKmm%!v&@H`h{2Tn(06
zpMnk@6R~*L97XBaQ9?@>anXKDGMozEJ;K#j9djkFKqLZOFI_e0FjT$aF{tO$P14)6
zmXb9spaqg`*>Ke+50VR!*fPyANjGGE>T4DMyt5V-<`*W`vgxm+aiV#4PcWc(hU3#X
zY#T?;;{AeEBwwcR2uSs{4%?Sq4?>fpra?hO@(~M9HHT@vdf{yfA`yH!tJc?^ZN(HD
z&}nn1Q{5@HZcq?MN2wHBWl&t`1aX$YH>1WB1;K<#@QV`k2f>JIL2VH&9869x;<+@+
z;X)p}4HpdIdKF}$9Fh=lV&RQ2FR?CzK=hwCJ8Ra@=P-vx-Zk3=CWFLR=TQMfUY*2&
z*}j}+UZ-yRh7Lqm3x@@!HmR|(2Kr)!mF8|6CvyhAPi|$A2qA7wApMnV!ggVh&%?3N
zOlN-F@~m@5*8G}%BlrnLt44C`Y`9~cZcI&U%FE?7N))Q!41hLkw=)>Em|Jj^;&YOo
zO!r@mjiFB+;AM{h;V<#^O}pB7Tr7xI^*#v^^I1Rz`Wcsy-si(C%*_>GI+PsePudgj
zQ$$NC<<HMSlBeWDLxD_U<&YK?XtFF&wp(Xp4WZ%Vl?J-IIQ0C^<C&cfCODw}<mg^~
zVioM-nHoTr2l~Y8tm;mTwool#Fi6elw&A@n<k5{2`t=<Ud&P>rz_Tv^b7&Fi){kvy
zFQuCbz$G6xY)I#F&rW~2k|>Ng7bRqZL{FM#hH=Is4iCn>f;drK&-ZB;&r0*B<otqY
zYQY|RherEr+hu!UBz=be)y)dCfBf*jy>M0s_Z{%WjQTP`^2Ws-3%w-NKV1I?5;4f9
z1T@377|ee9eQ|PUAhF;E|Knvbl7)wGaFCij%kOWmFJ29nD@;zz0QrAK<6-5yT`p+n
z&zj-8fhi{<VFg2>%+kKp<Z(fStyO*M2u8UEb*p_^&?KZNozM&lFBjv<!uBw_uW~9a
zP3}S~`K|xg+W*gkqD8o|bY^@&tS=;Cw7}?Q-}v%yw-Lu`YNSit(jh;PnGvqq-10C%
zaK~yBr1Yk=I<Md>M1Uhm^cKMtuyCut-Cz0lY<~<(SXdu@&E^%tK0%Jg?%gJn>vu9J
z2wrSN=#_EK521^a&<MCcF;2pd-{&*g#AbCCvyW9mlEkM&;9(HmM5S(@!}JjJg}6Wp
z3_Zc_+kKkrj?cRS-j;ou*&*b32Xi~?eVQbp$!~cJ6#@=F$J@V7dp}Mrzv{<LgBX1q
zB(9A&ELZ!y-G_^O9v}A7nXy{D)nz^;LsKFljMIGFvm|m8B~Mb?-EIfS&j_dKm**$S
z^*J3xC3cU<{W3R40>RXw$ao;EtLQUmDqHhpI${dQ9=j0BI<B~PD2ZDVgAhi!sA}&>
zBEE(F>NZ5M1=u#+R+}}83&U54uC9!vF#A$tI!>(1E3_DErG&GY?_bft50Yi@$8W`?
zJN=?Znj6WBZ!VObp`fkRoa*8DeHC>?i-0TaG3_zb|Lb5tqVD@8dFxG{ysSt#Jti?!
zDAA=yXBOAPmBXXm<Ij3l>+CtOY_)9XZ^ejl_a&ytmX~E(H-STP7Wo9v<tl>4tJ=jg
zHfn^zWd_y>t`h<KR12-bI#AUqs4F&7$HWEcd~b<f>U^m-=!EiQV9j+6ef{$`Dic3u
z`QHW83_jH%=@_*FM*@yTX|@A#EgLNvs_#+Q7}9w7*5&D1Lw`^hws5-9bESTzM3L_z
z9xbKC`%_SBuw7b@;>*j6FvQw>BF~%)9JXI@7Y`Z)q)_+hZD249g|ARpELyG8i?<em
z18$Rk;#B4jx4xK#oK<Uyzx>_sSd20zuvsvxo@}#L_LWsjR#oa-Z!&wdxk6d1Z>)@~
zwf1=;p4Jp^u3T_dm;N0|Sh~*~|FeF%IF&(#Avt;|I(><48a8>RMJ{{>%B3Ym)2On3
zF`eAsNM4RIoN6mdtM~KMH77Fkp)fRaAKQnkwBO$l-8E2Yxw%m=GXh92GS5C`Om`5z
zF#=IGur8LoSZhhpPgytaOM4_%Tn#jAZt4>3JZi+rlBkbk-dd|pzP8K>Qin_$#`Dn+
zI0*e3a?qb?Fo1reJ2hKid+umdn6mxzIqJlw_N-Cwp`P2j*=RkDs6%bVvO%%#TOw41
zO_gblNis~o024f}<H}|)8d_^XEll#n8Rw+WKo|EXNECX8>RbYiER3q36_;4k1(%)q
zml%t$3Sod_6%DGr2RyE8S<Yba6(aeG9{^=I^NbJ15XwWhz<0iRk>T-c`GU8d&G=gY
z7#8gvFzwPcYo&Gx9>1^_<Oz@r!~#oQ<r@ioW=^O~KQ3fzIgO6_jblF8@5qXd{<p|u
zd-kvUnY1Z{=$<5wl$M1j%T=!%5KeTYa^#VE-feFH@cr_q`ml{a5w^PnWZTA|;fyxk
zv{K=Ur`-PsdDJXlB3&yM9U+a6$B5Oq3nLJerA4V@b3>72Si5YKKo;st>uM!1D&CD0
z8u5in3{w>Cu?|`S0@*rScU`A!y1!@gZG2SH_rvV(Rhqs1!hJG<n;OGGf4l`_kw_F$
zfn8|_LYGH7*lcIwnb8k&Ngtk+yV)R6e`6_Jx=LdyYzUnktlfl^y_o2t;!3H;<X@yo
z<22X_9k(M8^S_w*4I&Rai2cDt@6zTW4<RKF;UWL6l*|$*E{xB}(u#B@HkVWn$>MF}
zF=reKyT-dIFopR=3>u|m4;J7c)QK^BV&5!-LY2}(=>`C1#oTY6TI@EGfr94Evl_LW
zCZm|_AuxyhuKg1s5t}HXhhicPwe3Mtb;^xet2)$Qx-B@Xh@P(c)HF>z_j=L}9R_E&
zYnz#L9dGXd(OqtUFiqwz8`p^^p5&3rQGv3FdU8h~?PaCQQO4z*<$#J%8lN$n6B*fO
zTgt$MZfNGT&aWQ{dBbXL$E`ik(?g*dzFc&{gA3xed0r-9KrNU1NaD>0{pJ{Zc$YFE
z-i{xJ&7Ikd!@7}Y7$n-L`f*vr2y9dR0l)BmbaU?_ZtG6QV-6LH33Q_+nu~r=mPJ1g
zLpH`hHYYmUM4&WO_kT@sY$7m+7MvdX;P!Kn4CQyf$|opm9{)l1U3namR7s|vcRNLP
zDO5>aAt+}k!qE6(0Ju%$rn@8W+g45Euxi|EnnBX>*tKN2<<GlNx^1wm0tojye(iL>
zxoEQl>10Zx5k@5rEG8m{kq<taN^~uIO(GW-=6@SGZ2{b@>vOQe+cLw3GUMBvxDM5K
z8|8TDUE3t!%GN40EEO(OxUQi|#~5Z{Ob4-~6j_ZVnYI-#?KDc)XzwrX@uvMuG0$L|
z1Oz^Im>!xze3Z+7SmDNPE?se>*POY;nN$5%KX6-QomGKBecp!;d4|HJ-C-8vE6Mej
zx@74;49I%XZjU>z$h9PkVFqWu1?~}Y3}YS_%A!SOb}i%kuxW#7<Wy<mDXI2#=_%vG
zSs8Y>D&7Hc@kyQnI3|NmKf6evXKZUV^J_({8y8RGSI(3)E0nY?hU2lSmRR4^evMHN
zLgoBS?;DM0_Ou@OK}NMQRwv!CLRv3jTB}`2K1|N9SFyA^mdP-$C4a-n6-bBBA)l?q
zl9o_}YbLt=CBF+4c2XUx0mE;HQr+mb(aTDBjstfCAv+etO)m0gvwXz~zj6D<Kf9@y
zxf8;qK(;zxe&MaD6kv4MOs(ikC8{L$txv;WB(&e%vZQ}nev5iT2mxAk3oKfKj-2_$
z(_(UiIT+hvk=T0Yil9Ga&GfZRFQ-XMV%NGxCwSwg5Xq>fmkSSFJFgD6!&&`@g%om0
zaTs#5xdbvC<D;Lxs)~)W=6IV{rBF4Jbj{n!wF#Hv3V90Vl^Ivo<S&{u773njtl(ZI
z7~G&9><!#FKi~e<i`3u?tC&C=*sY0&<MCGc3|zRv4LU+X_G$1JD_!spa`?xLEY9xI
zEDw-5AM)gQp@YWvUw;RjsuPM;0k@Y^$%K6xvNQ?^IQ7p;ar5TV#MkJ>BZa7wzjw_I
z91d%12dR~e1WxMW8#^<BpObYs&LUeb5Fq}JCG%xGY%T1pMPJj;j9diH!`I@5LZfL!
z>EaGOgt&Xcz@Co1%g+lxLajjj$m~T(lqUf`MK+n@zN)T~U$@+3;<>&MMX3~O+pz5=
zXo?@j`AwrWzZ{3nOWn9f<<{?;&fDA?aiRt^Gm~EXZFy6PX1O*{aPxX1<sJ@ZrqW5r
z5A}w#fQ6I%8L^g>0uP)zoWN>~P24pVV!6Cl!tB#2ccnqLL~c>c-65aGI>X$c+7$C}
zfn_|YrN0wH!pUi{PXe3B$No^z?-u_4_|Z&1=qSKn)P@TTy-oRKmnrOOJvL(gp|uW!
z&5;F;HxV3_Ir63VP17pLX|GqH#&Ii|6gIO}&?E*v7IoIwAH(3gYHPOQ=vbwLW~_WU
zxoperFrD_V3<iw9GlFukhmd<BGEq6`6vv>Andw?z{?+f=A0Y0XY7BbA*}EI4aAYn!
z>R`h(zr4|x*hwjHL&HVpVel+UPAWU9{pE0dTc{(&%gvI&#)iOZo~4*6mYI^0BS$AG
zsm-y=W+E-dWCX44OXodwt^0dIAf2u^Flw>RLLTwOSbdAM)nr3AMe_3&R6id3)~7C#
zAX4xq;FCUD^Vej*%a|re?(6#6<u&74+u~uh(6VfHZeVnaB)-NV9Y}KpI9}7Lg-EdR
z0@B|NV72=2=)a?66~Egmy!(R!^@lJ6RPyDlO`8;Hqbn3T6!tqnsqo@Ym7#ym<;|TB
z&eJxv7&66^2x%OCg8*tQ{%!*q)8)t?fzRKGO+M345oX=E`+otu2}>NJj3N2M9_B*b
z<OTLz0zUPc!aaq|(D7br<z-RRUPd4OM)VsGfbMl4^|?f<ri4-sH(M1GkO={fhDoR4
z1t1oEro!!FMu>`@K?D4Pa)g@x*+(V-&5yDoh;IiV@<8z~0W3A4he-sJfaY$C1H9*r
zL}d>h0f!30PP4)CsW{Ee=c4#4dX6&;cSw8VTh_SwLr<na<+*Sr`A>dFlReGu6@I(O
zArnqx>_gBib)#}2^KSG9S&vKImus8g>!UfrA2d+IHO0SqH+z!Q>(M67*1#Xwt6i>}
zQ1UvEdwg`c3jA+uD@vnO61%4RB~?*nSf~Tq9Q9$UR`uLW;dZ<LaE26zpV-hFJzz{F
z@Focnt(#TnLc=t?;aERU5NC?H$y9se6JK+CbUT?H5aZ~(oWM-9p=#>6a|>WUbusy!
z>3(4L(xkTQB1i8fM2Es;6`dAC<3vZOCf(jzCBl;)4vNR5#`Px+^9#BV3nTRtlXz0F
z6Rpuki*6K#=hyCS6Z4Bjn-sX7YwD1s4t>^S$Ja=fyVYcu7hnse>8>9(CE*SMFa!8t
z3jNoZlWGZ~1IdVC+5kE_{n?3qIi&(t&nn*2a+Y5%Lmm23%<sFcLL?sTw<IE|SU+OA
z`~EWv(8+qDHUS8oun(*7#?3g$t2kNJA^UsOb9CMg-PGP`HKRi7vPo<R+kTtVesZGq
zAUdph!yGz1hp`mGW*U<7od~S}ywrJwXo^CGqufboRu5V?i3N!+XMHHNX8xNj`2J}e
zTL4T3h~LW+8d5z`d6@5r5S|YSp^$|)H5nVrQRTzNh$|`YgGHhgk)jU?xo(+T+Xzud
z+qO=W#Ey)D;d=OS;ECYZ(Ev|XkzJbtH}k{O9tXNm=;H@jNM}ffeY!B1q`$haBIuG)
z&Nan<`oZ4V*kP{^_@x^Xl}n*WLq*QjL79Z@;t1>GI0BHO1G-SPrka<#>&XE)8+|Sy
z4F5JLCSeA|H7rv8bd6A`NbkAK8WeFR00)Y8jI3ID2wk=eaqgjQqlm4LmG?!n<4LRa
zucgP2?N)<<uUZzq3?u`Y-UezQ=WL(r=7Sh4TtO)&BPL}Q$J4~#ewI(gR7FL5vL?7e
zWC<ZgoU~ND?97y43v-jv5xIe2#7BGbwhQ#1zi7J6BBsd+R2>U)U{l9ZUc^<X2AOn+
zr-s*`YGhIfhwY^}#4wx5Gd`fBW7-F=I6A8s*N3yqrz6HmM@(NAu#si4Aru6tesA*C
z%pm-{|4(Nf<_4IGVh4~5>RgCGHI5c6<g~`LLGr%$#aRR{i*kJN`SuVUC0Kaj_wiy;
zJeW<Mktd@~YwSDahSMgx6iKFB&`Y~1X;b}l1T95t#d@gN(h`;-`X^$OFjHnoUT||D
zDTRwI5b?oiV(f$B0kALW*QAg!(~0vi+j#9LaH-<n>zkkVk|QG*zxDUsUe-(zE?zd1
z7_^=*g~IYi!9Ji(&Z=p+C7n}yg;a**$bcC^mrO}tx#4*)n*ahx*wmbLIKU+zQ8_sR
zmu5Ujg{@Pq>$P>@wOxPzjnrU4``O_o<h-07WQQ_^O)Y@X6~c1_ffM<(bw1PhgoI``
z(Dr4~UNCcHfwvq(bi*taASFpx+j2hfQbWISSPKI}IdOSx#q=S(e&`+78NRt?rO!RN
zSFRRFEwlbV%)M1mTkju-NhuU4P~3{UI}Nm0DGtG<xI4k!ixfhz;)LMEodCh16b(U&
zmOz2xP+W=>`7Qs8-I<-;>z%z!=3JaPbIwWL&-;9yuQaM=7^nzljuZYD@B}09NPS7p
zKJFq9H`ha6(Ejya-~VG17<>2Uk!~(X=K1ExrcKq>?LV0;mPUE}69x;Z5S0eF6ix@Y
z!upsYi|hWMcn$X=gGHea$ql2R?;-=!pwE3p<+b4=;l_}xaPYJ8u#0m&_z4k%M3dH(
zTI8B01Vs0|Bfc{t=X#?ozMYorQi&;wAv8k4#{CsDHSEjT?g|Oy_!a9Ab@}hPze*fC
zG0FEN@g0MGVH=4beI0N<c~mKJ9KK_vOa-^dzuFLQt(QqXTN*U<r}{MdbdqDl(b?`1
z%08q$SNe(4XxxC1#WLT$HA54X#)mcK_6S-EOBp`?kUqvn`Co&U0B`zYX{%pS*FKjB
zn*qwo)BTiJtmI~}=QoLr$J+m4AYaVH3E1}DHyImM|Nj}o7p5GRCj1%~r1qf=q|&?a
zl>U<0J->aBZuB)S%O90qS?!Z|qng3MxN5>-xZ#_8@o{xXc>Kq8cHD|Nq_WZ3s+mNc
z+!|u^1`%phLm;AK^_zZN0sZRPuOZq(y|rX+INdWyn@pL;Co%m0m#e_r&h$LzsRR;U
zYm`XJ#n}7MsC#<lf$z1F-?oc(dvT~svY8qt`gL22YEY!e;Vt;@BY;2mB@xlI9SP0{
zH2Ck;?%MKKDj@>JJsiQ8zNbU_3`En|BPo3&mDDZPvg>?eS0{7TJ(b(;Cc-Xs*p|sW
z3i1IMgm-oIy}bnQUj?UUdBig0RyXPY4~>UgWA6{=T(9x~<kg>3sxY;Kzx><zOQ^i9
zL817KUc?5s1he=a9(?A}V_@Nnffedn!AL)34Nk)ewEfxR#5^du*!H1P02@BGT1aP<
zfPZ{Oa2+_CP7=C!?D^C<I+)L4xtTV{2x$nYnRee{Id%O5u_2(t(^HNXd}zhhrmG8?
zS^B{|j-`MhAGV|}J5uwq*N-}KIA7G<I*xL1EJgND^en>iOSOo$MO9UmjgN{uP*s+}
zKU}I?BRek_GY#FW;yu-V*C0tr;5754c>FtEiF%MzKzl-{`k5f&WA8vH@v`grddozU
z#RZ%BPwBK}wh^pl;@%TeidmOdm3#p%*5)O<#Up?4VHTJrmK)H-Vuv+^5rGJQ4G`E_
zo=OT~^m4<_#kZfrIvBe_Ebhd6sUaS>y7&XivXC;sQP!ww`$7;LU5r3)eCmHU=tis{
zZo?M9@w7qL!Avp5Mp~}$85^%dH@jF7%A6@wAoW?qlvm=qV0@~FX2d>nZ%VcQMua@W
z6PGB_IJ*R#P&h0b$N25%w+4vp%j5jMW!9E}o`=?}gnzUO37rq(ehH2o=GT9BeaB*v
zEF)~=P<JTDkic1<htW|Q%R}`uff0gK0=<Y5SrSX|8kqL^nc>~yN>0<SYqyDJ_%Sr?
zYwOCiNyVCw=`oCOTSj6N2B7h-P1{{6U$h^vj@9p{$g<N}N_5;t9y7Zs>gm<8(%Y~&
z(CFq?^5f2q%|7Y+8c^9oc?PtU#2nik_GPwMj(DkT<=#T!FSBy9@W3-KNEmW53&jfx
z88`1feqYJCskxczlqFc8)i{A$;sY3+{bB2WC+w3~uo0K3euo;d_q|sID=hOaU-+gi
z?*hVC#9KGqX8C;V(2karZ$)a;kC&?ltm5`W0K+#=J3f&=vxk<=3kE{nVyA!o)Ld8w
zQU;b9DLfQC6`)oq1=Bg43JGVMvked<Orue=1p6}}K>RG{00?2?##z}CLHReim=Rh+
z*=+AHG+a8H{sqXU%A^7u>*1v&%2Sz$Y3^sawcBb5Z1kWsaxhOOX1*Q#$lALXp%u;*
zfxU(k{_0SA`iyA(aHXirgO_q{<(j?vPp?Ni|AMPoja%$_i%Pe^-j1=IX;BZKVNI{u
z;<lVKltoeuc`;bUF-Bk+d1HQajffz<IzL-{sB1DeJRm`b_+8s~^HGJOqz=UH=z9<0
zBlZY+g8eL`T0<yR%mgwO6$PS<*R@xqTXe$-_?ijqRTS6G9Q^#7c&jvNV=yCXvwpza
z#3Y=Y6<7Dm%E~JpWKovqo5h%;i+1sUG^gS&C;V+UlBajhC*U1mPu(>^m-MDnKAj%t
za$SIQ`LvH#x8rznlV<nM%CrTUil=Q${@U2Yz2jM%1BjOP%)=ewevpk-10M*(yP}n<
zw2m^!nh0L3Z5ym12Q$$&sy51dP;$#vVS&v@*}Zd#k$Wm)7r&VU*fr|h<S)K^KeX;H
zhhT{RYd(F&6i)Tavn$oD7vn#S8AEVQy%g0g?5vSGoMrV4BjS{{o;g{zQ4Od5SO)bg
zcS@gUl1OMa3SuAo+4q6#_0$KXFy}{?;1Aje9X*{)gO_5)Pto$9D888S`wl-bUa#{n
zsi}F{(UMr-<!)~SEI(5}*%HVc{7p7)iV=*&4RtN4B3G%+ZI=_79qz_iH40tQQAitM
zREx=YmlRsKr&9v!){FWRS%i~Q35#*&A=S>TMd2%ddF$hzEGXf!Q{S<wq<XN$8r*ni
zqIIdRQ7vC!jIi8npYuwBgGNW5oP><?D)>E14#yQ&Y!->_1OD`#?imqfIEA|90#8+3
za1laHj!m71=eskn*22M7aHj(wrqhsSPFXLfoR;R)2afpNY0mK#$R(8mf;_HnvR9BO
z2Jjz70<AE7I=+p4*4xZf&lBAsaYU6^lRO`2@soeqy@@FQ9Zf2}op?90@nl@s9nIG?
za=}U&Eq--~z!w~%r5cbfTH^K`>k*XvbB{jQ@mmwnh*`Y~!JZ9U^;b9gl}*Ql)|lqe
zH~lgo7(VM7WhPSeP=IHy7M@^L-=kZrE*U%bH>@I4l5lSEu>}b;WPfLMT2QKYDVIpi
z8)Qp**&i~4yr|MeSE&q=o|9O2y_FW_f9WEn8mKoVf9*`GdM0S^(USmcBlGt@MX$8>
z@CD$Eb=1z5nz*Y^US_a#m8`WW3%pI+$kRF!R&o=PXVEB<ClF1HW~^Bjp#Rk8u2Gxa
zv~dStgK_$qoQyhJ_-s!!uP)#txRQFG5r8?=ql9_rmnC9p*>tq>9fhg4<ef<GmNNDd
z%lFY+?5tH8V-56l)g+jXy29_ZjP`lLo~u)(SKV&D^~PG^D-@!Dbi6Zkd-AL1y%plS
zh`+6BR}%SAls(?NuZ22cOc6gS@q^!YG?*q&Si@I^$GQMYv)-!+nBR&2SfPM%tEr%o
z9W8%WSh<;(7OCCRHlDURL`N7HMO`PW&?STa3l2h<R$n)+E+l-PwvM&$bLP9wkJU&%
zZX3@ek7zbJBmt5+BGc&@xikvx^Dk_`!22)C9gh2AkBp~2aeUIVJW%lEieI$>y{SHL
zCZK?`)V0M~Q7S=LIqO2TDxFy94SQ@DMz6ozc~#UXoGg8cW;oo%MRbO=;r%C>|1cz7
z!<mlU7c4(&4lcRl!w=wxk5iUcLDv(pvlP8}RNM1*-Rwvq6cnJ-5{d$P0%PB@m^BFo
zTqxnmr0_ZbSRMQw4to514bI%V9o9ZRpq$fvmbUN-mJKF?mj!N9zd6Wpcs_|uqqx1~
zC6tARZJ0hN2>J2Sw&<|O5oikTn_vd+4L#>(O3-K7$<QOc;M5|nOO1J%&0gVPUsk%k
z*=;W2`p%S}ar7;zSZN+^Gt2$JSz*Y^Bkj?fHV^=C)O*6krrx~)12wJ^8*dNsXdMCY
z&6WX|Lex9dHZUGD=q#47J6<{~>3Y;N0@hjoJIC5(?$+%537cdP3QGv>W=LH}hK259
z`Dvdm>pRRowi*lE+f~U>5r&HVhS>RVP;W}tl;}qpR#;6DdcR~pQot%x-e&>oii(jJ
zcZiXD1-az2bMn~KvZaFyk{x+kXq=H%xGzK<X3K=^ZHCdT)YD&;;{S-39f|DSq_<Ia
zQLOkSgw3D&&xWdYv!=e2G}Ge?$f(>#?<(*MjX}n2enIStS!hy+1XRh^V+j{~>i0U!
zR+q8i08<0R+8zD&nc9diunF&nyu+=34f#49QYbX1T;6#_)H>E0MotXg>26PapR`pm
zP=I3)7r0n_nJEm;Ske0)CRm>-VDfbi8zQtXQ7Tk~124lt#_NxTA{4@L^@ii=XQEX{
zwl(bSv5~i<#16%H3K4^iM-_u8^8Cmp$*bvR|C(sfiD~YhxL?$WbE<3eK)4*8VTc`a
zmcW4~%{p%9X~F5Gmdt#vz)uXr*<VIzS@8?I)eU*8_t~w5aQm)aj~pSo^y3$t%eP@X
z_=?FiLU<lndpiS9S2&9Lcii34xwXhn(AU_ODnf%n8OHtG{G0l+YJQvY(IMjZ&|ozO
zZk|*VU0r*APQaejW%uP=;+(TH*YCZ=x9yFVI2J)ap4|!jv;ZVz+XHg1n{|Ks{R+s0
zj|K#+Q$hj8siVmfbW+4XVnxOsB8&v=a}8LNN)G_$DA1)x%f1<Az96X)Zk()epx@3Z
z_muc|iM$Q@p)PTf^CD?&Tw`;JD&r^FUuPN>mu{E0z`acMk28Pl)xO{j_8UkstLl}}
zK#a=CeOso*4kPD$jMy3RX@}m$$rA2g&{TQ1vzPuzNMP%AhBM+^z5chjoP5>izN~m1
zdPFgA+uI<yf%a&F`J(I(9I`>s^PRC4)XFWR?6&+ZQ0Kmw?gR?p+DO<p)jRv>QlC`b
zQ{R<^cI>D}#h$gU7~;cGR*f(=N>@avzlwO7M)R<VXXXV^Xox`M(P1lp)H*uA0}>va
zKBF)!EbI_zF_)<%$Wvx6@}Zb(CZhyas#mHqFIJl=|B*%mm`yi@${jS$pUf)NNjnZD
z#l}7v*UF8+evc<8;8x+rpM7`|S4bpCTYu1eNI&4M)-7mWk`6d+HF2YavB`GU5@IzI
z*zgimaxA%djtWC#nXyE!#W`;LgF@IHJIan$f1&voZtomJS7+;&>o=eRNERUMO*L!K
zGDfnm@Bld_G8LKbm^`CEI>;eW4vl?lURxoCCmM%+C646&mouRU<VjI{x6?Dz%zERl
zUj044!8Rijm<G4Ns?Ez2MZTS}jh9GKQ~w_QmSd~*;QU;8$CgU%-*&U}5v`;3jXJ33
zct79?!SW3a?z~cW;+w&H_+k-Yhmu0>3<J^A_wi|UGvaMKWY{G%GwY)416YfF!6)sY
zJJFy})#eq$y`$?lo2M2ROCedIzh#}p!*)4l3G+;rfKUftG7x1>94d<$G9{v*gXc9w
z*Dhm0;i;<TnkJKktPx6zCx?Vrd_i?u#~b(se63WGC7L3a6~$B4sO)Gt<;@&rB?~B9
zQPWh?)SQA=jK7>@jvOzIkWARTm%D+sgb0VPotbxe!LMVMvrY@KBw20IU8^e`_-qMU
zD9A+c+7A^#i1a3p3fQzvtxYAHXIj7)Y?~@Vg7H9@d=t=oeT4KL3Ot;6u#3HHG~Q8e
z5j{cmqQsUQ(TO%;Yktb2JIdBjv1sdL`{Ym`HjQ)<J2GZrg>VtiwD=#7DZyjl!9SW`
z=Sq0oW<0AZ8u6TN_g}k%7K%#t?n&cz2$S_SrAbw3sk>X1$cQMuLjpx-*^b|P6Fke$
z`TUP+e4=cL5fZ{41d^^k>NtWapZXv7F%^cZ<~wKFc6x+>mO&WEW^jz*D2E2T|J5Z4
z5(_AdyKdFT>v(vLJ$1U~Tz>?oL=8`P-W&C}$A2*=V@{{9zUTz`tFvkM({19#PLOx_
z$H)1<^#Q)1R?r?)m8Kt|cv*hw9sEf@m+w5+FlECbk!0Qu?&W05;ZmhYH%#r$t<$_i
zs=fX}Bno(WrXnX4$QaBvK7LsQoE%7vlNCQ&dv#TP+qZp0Khe^Nwprr;(!2ZX>P`a!
z)UD0h7X;8sBg{xrJIJ5F`$7}}>AFZ!x|kf6o=p^^_VP+@S3P$d_JLRUG~r0gwOGQ!
zkOM*^)~|X<te=NhG0jdpx`!Z#6Aez5C?C1_TF4vD%tKvrWGR<8yde5f_TBY>$=n;d
zBkg@<;g$Sv@9HsEGjCEV22d1SfRcg#Fn--FHtQy@x7$LlratLBZEJT7L>xD{2Rq))
z>@-Arjqm+Fi<l=H50bLcF~$9qOs)6Cob&j(4CI)g_b5qzDm^kp!M}~>G<y9DsD2y(
zcgr3X-Y6He4*3)+$jNDl?{)4o4!!n8h`BTnbk@!y$)JJq455`!B1M&XHd~gvPnUx+
z859)qxsuw6HK%6z*(`i15`uVAdB%a(Bk|8)gH@k>7Oj`c@>(HU&ysor((sxOG%Flt
z;G1PvR6v5XZLmeuG`0)4IkXHZZkbgBCp1Tda%gmC=CQgngmUtrDWVKN<i`{schgvk
zZNxE{@7YHktmO6!3B-I%ly{iEzhD)5zMCRN`>nz6pA^LtFf}bdlcL3o1)Vt#<FVAM
zi9TqJY=T}+Fa27`(5wa%6bK9DBuG*yq?yb(v)dhho0Q)5O13M=jX70@1^cAN`3oXH
zat1T75eskP?@r`AcRx;-SBoBQD8PCtv@`TnVdr)}$Rizm0<tZ~a2Epzlraa;<AS~G
z4G0k4%5r|nFigXn=n-kW5`2Y@LlagP)laTTE3P=Zi)~NERfjQNmhpq7Y=69d-yZfe
z?WMil86;NqER>m^f@M~@rmAJmaeX*D=_O|QYv--Qac8F=<wfsB6Bu-^9hkp+Tyk!>
z|1kA;wDQlr?NW_eb8JF3qsyajYcD+T&2x`q=bwY)#+&hVTGX+q1Gc{t(nk>7OLf$?
zJuV+^PX)^N*tRUyk?o=ewXy>ikHq7fhX=*Fln<kT3rz7|gVs*D>{DVW`kN}<pDlCP
zCiSGfUuv^I<2~;dd{@94uVAa^Gf?H7T?Kb3lrg2QEH`S&b^boMUW+TNuf5A|S23PA
z!4d1WKc(Lrj(2E2(RBTJGeADj7=YO6;1{1uYL30`-d>%yq@-mt96h><KV9lI^EBZn
zZL3fh`4(k1I+TY5%aO~>2HaaLCH!=$r3x#dCH(5pV7iL()j}H`*Yx_W-Br)K#-i@n
zatQM@ZqsTn#O=w%z3tGRv<e+RsJDkYvn71KXR-lp;H7gNYL{PmOBaugty3`|W%%SY
zVz^apc=|jsn}2*VOQe;oUkRI~`OdYBuPQi}J94ZDX8RUU_q?-zo}U7GWIm!>iS9hc
z^>>(X>wBbHSvC*0pjt^=BQw>gH&_D|vKYk9vREA6Hrq{RvThAvVS>EkV{xrAse)j_
zO&=BVn|cHn8x{|<#Rt^~?CTFGrkvq8&2o0)pR<6-tOKdN>r$jJ3MMfLk=RIM&r$GF
z77JzD!FT>^R;LfpOwKdlniFZ#{#?N1h#KG0<IGIuGFBU(-ye(AgcN^+6lhjwG#{{K
zu#NFukE`K-RnX6zn97<!WoRi09f@ujocVwzXOl{sew`x*9ImpC&_)W-tSTzPb*J9#
zu@>`>wTqj#7l&Cow#j~oWwr_7&@Sd!JUE9*Q9A2zoYgVvrxU{SmmzA>VQlRu_GU{u
zD~w~}-M>ba4hKcY9`{l9{`1nMIz|Kyr6k)P+ZP9u*vlGV7n{%SpBe{ZlB#M!egie`
zGQT26HUo1Guntqr>*Krz)|xLU?x6oLqJKKg4n|@+pI^XzNc(y9%N<n`x9!5IxM82H
zUa60Nhn@MWZ%7|GFX1~(T2R%a+{7qrAXnY>s9DFufTWZw-kH5m8ijtVh9Ua<jUx{E
ztifSfvsn9jEVoMf5hC1mju(J7Yxgd~5nd;OgC|PJ>v1G_C5v|EsE?nLX832oYC2Xt
z)#o*tDIT*A^8(Z49h6lh>M%Yz@tuM5;^v`?zv{Wz(Vo@q&;d9m0|%;FgF^k|*O=B1
zrrcHWCBh2jMKL_=UQ0I43<b&*v=X}Ez%=PEygHa!SeI|t(}bW}@1yQ@5%+g5mO_=H
z4i+Ekd+ZKNC-Rqgzi)IM1%yEqH+jyQk$rV(wWJ=n{A6`{rT0j^UG?0jnu>r#xbdx$
zHTZr{3icG=5SU(NJJciZ)C8Av`@>8RuWIxUdhH>g<t!9nS>!wDycW($&7qPmVx~9V
zPYorSh;eTL-*+WQw(<~lUQBkMc3Jy^1f*DKK7BeKRq&{b(|K{%;`t((x;U1JDd$vS
zOF>*cT0R-Olj5u|7bA*E1*|N<WlLVkG}9yfvB5z6)n(AqO3JoIxS0)03F&&r7IP_?
z?|E{eq~1@x0}3@oJS_6qF*cCG#W8Av%Pz(NH8ytttzMJ$)DHOdLo_Sx?>2#1H|%tq
zUh+kR+5sAd2%itXf<q}2`1`KilJN<;#ugn^vN={R))TyubjYZ;&EcVY@j)M0iSOJF
znzQMfU1eBJ{Ikybju2sa4^t9asjf}or+IFTi+mgUw}=D+Z<`qC%=VZNDtb#T5)31M
zv%kRNoXMi0M<BiMkL&0?r}9EJPg>vvjh+^mLd|z(IGszC4fu6HyIfC|rDMhq?{$-e
zExQ0)y~gN7WJ0O~%l9Nq@`f&u#k3oJ@8wcb-cqH&=!r%iWWPnd&0NHd(v=;4(z~^^
znffqM9(wy_G957B6lG_6IzTS0j)`dnu?vT`{-hYPY6Pv^HaFL89QSRn&agJErt@7K
z0Zh%OoSMgARFlofQ~_>7Fu!g|6yajYQ*lqR7i-iWwe>U_14%j?YO!Hx*27a|VLhC$
z_pNNJ1~bTj<gMCR{hTqD7*F~s-lriR36To+SGiNfss*F8q#tx?$X;uXjzG-0HUpQ#
z=nfWc7tO&>Fu7G1p!07xOK+Vat%AR+!a?U&*Vk^270)GhTnA*-MauSAz6cQM6RuCj
z-mN{JC~!yc6l}Yr-D3PD-*hg++rpLvYWS;xqNS*AL0G_{ZUY&Qcs78yliC*gw14O|
zaKRcSi=u01F)M&+`=%Sd=GR_mawUH~tM--YqniCc_6WVor5~dmq*>nH#@f~CWsqic
zYnp()z$MN&I9X41<GC}}4|UG;RF)6<0{%W{b>hnXN14mCX@8nm0<xwo(@qkqhn^mD
z-aaSL6|cdjo5<IumF5Ox5u3(ktG1H?mY-Pr&31}dKHE#SG3t5t;oV&q8?n=et!F_a
zpe~QDG(U8(sUU498r05D*~K3_La@hgcpDJ9xZbw7W3&;)eWjXKlqsf+1<eZbk+G|E
zxOk+T#cT5lnO^uSCvGv%GE_jO1yn`OmG8$#m70x<yWZMiAwkiV*Op>(s$*YX<LThV
z{LUMtr$bS{EO=|I7o!*6)r#mE=8wAR5;;uu`_){(7GKzG;%LNXvaQvIj&MZh3{{7Y
z9?ddQWuOsIhiRv5tPV0eB*gOY2M~G@*95A|9eL=uF&8+6dh}Sg-AhN!J7&5UviR1p
zT~IvQq+<1<i#U$cc@*|QGvXWh0-#0JLUNHdmV+yrEhn|1U#3Wo^7$5-0@>(`Q5S4}
zd+mX0AFSq~znzJ%72bsXGgY7E&`MYG*G8z(7`v$a8Y9@v)07q!7JN<Do!#V?`PD@L
z_UgN-ptB);QBUm}37@cW?N971$;PGH3wOy4Tu>D4<eE$NosO{P<LOj34_COEWa9{d
z&d3d|pdA`De6~zWI00DkT}em0T){4x4>w_7R!A`GTHViad%=sHHbn!+wax%y0`=O8
zb1k?XaIIH;rx{h!14Tye=GI75)K;yfk_Q%PRM<!|ods!R*<W@MC(<C>|MI^`$<B+G
zBxUN4+E*~X`C$=Bvfsw^mh|~_Xo&rP7{QjZWHOXv-3X;ibi2K&e+jpag(0jn6i~SX
znr~OcDz<Jlv5DpWKn+xK-)4>ZKxo>NwbNA}5w}KKZsebP)MpEGM5hWFu`QRv#E75V
zDYKGST~d)$lWU3Z+3$x%CePwz`o^tW>G|XN1aui==1$moqdZ=6su@q`^WM1YN_%oB
zYw}rt0d{JWdV8|%Y5{*~y|c1#x=)nQU6&<Zv>pr>Z%1FjFIzwsYi_vd*0F2wg@-oy
zB0HX5Qh{Ah{zC~bUh8|e7jK(}*-@0yJfIv0#h&x^FS1;Jp?AUo#)t}=dP}G)1lf|s
zpVn$x0s*DJ!#$bhpf#g(iisqsWEKWK6wowkLWq~oG^TQA>@Ur6qVnXj7DQYctCcvK
z2GcY-A<mKKsxsvw+N85%O)SQzXMak;hwa@Ns>L))Vo-jjTw$;i_LnkytFb<jdq(}9
z&837X4+P+6GnoKB{aJ7*f1yXqWUt+_qY>YNy5FYPR3Sydlzl$depaOOt+q;(-Gu<2
zj)w4%2><-yM~y<qLt4gGfLoFH;6e5AP0}4<fjJs5f{mDVZZv8zCM}yRbWSC&w5NGt
zhdvui#TwltUC&}4AlI0!)t-IrrpdweeI|vrCXqWSuoUjcFJ;sdT5WBUB$lqnO`^7C
zJq=|mHn#qq#x?si<|2qg+u)_YrVTlB##60Y6+H!7%>UZk01bHf*^=c-_|#HHyZ*7C
zaEJNhoi6g9Z#E-a7f<k|x)k&rkd*Oc>l~^up=jL1(7s$Y^}qb}9Rf8aeiI(;b(A%V
zs1|Wx*UX5(+PMaNL`*B+zID-Y#4#4nCo%rq1`{Mwwt3#yEs|8?5fn_Yd9>-QdAuMW
zXaD1s?X|0h0*(gea@#9tLNL*W=zkbP+u5((-mHrJhp|qJB5Mc|47eedTduKCI9%zB
zev0UiD+tZy+r%j>>l_}bEEjDW#&7ldU6{LyaD5mVVmXu9<6_?}#N1L`TE7hoPe{&>
z+~nXi6Pkce{Mz2tZasc`nF8Ppjk9gYD*Gu|{(dOT(8~9@Mi`$?Go)MOBzNr9tFyMU
z1lmnzP4FvMuRk`_lXBppJG`yi-Ob@zeFd9uFJ2e6O-mKZ%991|{Uf&QW2b$N8nGn{
z_0ovkV2}zU_!y%7es2Hd8~1tWV)@yKZ?kOJEob`9;9A6K#?x1o;4epO+37|98?`sc
z!S`8W^G|_RVKMj&fN?JUuOfYS&nAk+Y2aOtBx5M~Ar6U6EDOF+gy2_>Z$37$5{(r?
z71pcPoB4}TTumbb=#0ZSjANU^!hhW}-(F(~8zvsqOFyLb*F_y_EVj3DJ<Vv&XcB*A
zb^7gMoW0H62G(RVQd8(_<30qwa<ArlR`i@=1^gp@{_ezp@1s{E``>?_;r9GpKIALl
zpIIC|$<=(6t?;L#@3ms<k2d$d+OIMocyE_4{Un*cQP(KQ4vqCVdHsPgf0N9qj58_;
z{gXv-{?m}V(@4Mldkcf|apfy7hN-fI_q~|V@WiJRx7|#qFl_XE_1i{Vrc4R-wFBHJ
zv&+!Gx$t9_h-5XqcC<x@`D7M-Y58jc<Bz8lqs5!@%3iU%te7j)7F1xkpt?#mF3Rv*
z@`S0<hFKgAm-fJ4=07ex&wh6dy2vhse{!BG{K;xCO`UMwHZ%WMUG<*YhCoO>N)1~&
zaokS+%@+;smuxl#Q_R!qqCtUr*uIbN<MFv15Y_WcQr%r~m0NTATUeLIIrhyws+nw6
z-+e0+ChWE5#e9EQO$JeR>@)18vsYZcu?QllH`~i3GTZaYc$&`qRYRWRi;+IN^heds
zrtGvx5obIBUn5OB4?astyhTNA>v5<`YWSp3SOU+RWT3cit-dxHGJVPXBiTt_>Y#AO
zjnrX5X!o#hJTOxyJ4oetl$(^trr8&L*n$T|GKuo#Kul$9Oer<P?GoU%|CbH_f0Tpb
zW}jLFa2|6HAy({r=ALLW|8hB=9ShS$KN8r1mGK#gqhE^m)lUuWYyZMQB_^ZZS&rwB
zn<c812R!LvgC*jud%g3DZL|mcKQ1AmQsQyYCIvT;bBKY96blQIsraHKp|jvap02Dk
zLCoPahq{m>mzf7K+(y42iC5Y9fo@>NM~DNH8ax0#SfFItPmya-<yYO;>n8{5%&UdR
zl|!<5U;l$)x)+Yf<S>!W#Rr>sYY&WZjT0n>{&T+AI|g>`%DBJ!7QI#6&T`YOc!2Kc
z&<!p9adSELJ}2=@r|++=-!tx5k=8mopcdm<GFF=Xigk5W6_qi$q&MQ?!jyD)L4i7~
zN@1h!Ql}*0JT4>iYBngcH`9Jf?<j+%ro5b#K7BeVocA01P|j0CBTs{d;ur|;Z%YWN
zD=HL^m)qaxT1|NN@kq#e{SF2tMg=nRcqZg^N`zNU*c%83YkU$&ujQQlJwS3mVf=xR
z(u<R&L9?8}lFRD5pAf`?iAVj+&?)D@cyaq&#hwqk)R-?P6RyfDj8O`=9wI<e9nAkv
zgfEcpn_}}>C#svk#)rbGbp2(IfjoCm^<lh7CD;~E#z)3dlpT{UwdDg8D$mKz=<mE+
zNb@#Fxc;C^@RK>XbNWZ-471KV?NhG&AwO-olKt<sJEi*i`gUV4g|~b!Y1phwTVgV{
zxF9qatjxKvIGWv^$lKHyqEA7bbokLE!N31Mw{!A;@q>Qz&Wxgmc`2Au=Zrkt%Q&f+
zcPVSZSod&sd-@l`yZ$~5eV^5upsxMev4%=QrCy$SHC#qzX!=PT`G3k{o^Hth$A?0O
zcmLX+KaB)4yWn;U3PdGPgbFwKP0Rn^4y80ly$$Hk&HU#$mh+XI&rFgb(|yDmOJFEU
zgXc)idF|zyk;SGb?J4cQ#m?w&v2!9d6#Ahee{2L{!hN-2R*MO1GqMKnLzn+rz7$6X
z{wB1Di)0-?cPQrks}solS7{3Ln*}CC(=Ac|{v9QL(rQ_2MSxK6ozF(L#-qHsX{0%m
zY%a&ZcIqx4RsZ=eFneQNs!T$)kgN*H(nT<r>`!-DsklT<MvHNgE}|ze5KgCDy1QMX
zoR;VJ)JY9Dw2@36^t8ko_zBpwJ=Ox#GPg*Qwdig@{oDwnB<c3HsTJ<nZfv5$l2747
z)KDxySLZUvlskL~fcu*=L9AIsA8FN3)pV%2Xjw>v=-pV>$j>K$Ev&c6yQCBJlcYMO
z)3qNi9AH;d+eFzJst*sy%v!bL`LIs_LKOOWKviAM%zs0xTYtZ(O{>uQQzbTLY>^j<
zx488ObGvkI%RG{LnmwFc@qMyIEeB0SIYU~eC=RQ{mhy%Io8Z$6f#E+4X4@O#rwsy}
z-)`G2wCtx@H*J=ZtS$cW)2JSR?gbhe0P>p5Cd@JY-FO5<e~d+1Iv@om1ki4cp*d}(
zs!`?_?y*S6LZa)V-LsFUQ)i*(cRFV;4ja+_QHQAlxV72y$rlfBpvcZs54{PdF-4#&
zxQs9_vAfN3<uB*{Q3wOrEq6|c%)Zk%Gr1CRt`eJB`a+ui8fbm^d+vLX;2SaJS9H^C
z)lqDEiaajNOiJ-@XsAf!0Co-qi3eX&RIRk+**)uK&~J&NAJc;$U&#P0y#eBz=7`%a
z&{OfYej*^hBGs;Mlk6m9)^QNVFzMzCyHU%u$wWL^01gAfvYlf5p=|~yZg5~<Nzd2r
zqpLlBhMuMfa71(f`&moZv(R4)%K_ijN3Ng&g!R0Y%8s3fZKE*2eS6(U=9r~j&|=Gy
zo6fHbc2&R~c5cLR!%EZQ;esmM-}iM_+?bPzqqGM52CRTuj*1B-pRUt>wY&Zv_C{AO
zjuP0hSl>k|MkAL*apu(@{Cui?b27ZO@At?CYBTe~>nX*iX^=RcDCpP~^FQ!wbC3ao
z@q|p={CU&uIYMi`|Ale}>n9k%%_>d!N9*8Q)cw~mqG8MLDc|l2{<YmRbxs^bte(H7
zrB6e(u!I#w_Bg;z1(V7?jd#_fecEjE9pK8Mz9}&5*w7nmb2@2RA=$TL(xM65;~Kj=
zKu()?n6C&1$CWo_wy16HR@JcR8#pm>al0__>pyq`tUI4Qri1>D_<`!De!=*?Q~<68
zzq>%$8f$E+bo~xZ$S8&OopDSuz_hsUWXlpujIvCq52w=E2)Xk8@)GaAi>02TUjLGR
zxtyy%kG5TU_?wwY2KY2vCm|r+t$uQ^A)W~~1vX>5lR-A)?Xh&cl~34Hm;ZRDO7!Y+
zdmM{i6Ly43fsRhv{qt#^#3z6AKPKJd;5;VO+Y<eUL1giFx`Fon9tH*KgLwq@z0>o@
z)CME<h~FQL#6TcY@aCfc19q^V>e=@%sU=+Zk{c03#7*R%%3>Y$F1<8Wb%Fa~N)@_1
zsy_*|W}-Y;6hio_)pE2Xx3^0Z^D>fke}Bk~d+o2%AuXV|y&8hwVklkO?O$Q-UqBCQ
zwS<@Uhp$}KbQ)~>uZr@WQnQ)#0=?wU-H35Zz#~{3qIAe6d68!_(*!4^{jJyR&5rI5
zuw@Mxv;|<AYqwW#8kThi3s?o<>217lOXFH+TvJK#{V;45MdPUwy+!7j3b1tX{F`)M
zU6U+D{tc3;Y~ApRoO(q!zXBI0(}RYby;p{WwqLi(&;aA{nOXn8RDF;LaXG;_)2ON%
zhJ4&hqQabl+`Qm|AD}m#yNg6%pF00#hdDRHIV9Yq!FAMFfvk?@u#m$(8EMy~Dk`Iq
ziz2`p;L#f*j>Y({A5U6(7dv(IIre<nJi2+Y7T}hL?#W9&YhFP-cZS$iy*O_&v0LH?
zz<|>@<AqMiu6XHY19a$A>J6w&C^8T4qwrW{RN^4E=u>|gQVIMUkPiUr|BDwk)Z~e;
zJ*T)p#&0UPyueaeGHT+0>Wp`60U4rCE2mONXmAjPpLE_C+_g~-t|9K+Q5MUVEmyEd
zWOEwd0;*tNT%XFU!~X<rmCb?<N-30at47;>E{O}L@#=T_G-!iqck9!~aI<uf)o;5L
zQsV&d)|hbV<6No@sN3`4Y9JFq(wA7bpau4TIc(_Kthi*N<3w#m$7K3N)u-MV=d&vb
z&pTr}bzU#~DDKxMX0xgv#3CoO+3I<6O}X{U$V9Xgz?wWFPVr(kq7Ww$(jQE;bSktT
z$l{pa3LG|h0jr#iLz_Vi%b*|K^(&v3+x<fvJ!tLj>}s&fH{ccvumWCvEvHn!B{!0A
zburCwGm`sews5C_Iufmx$*iOfwCnX1ulpj%clqj}$=Kewd>M2-QGE=z_&#-O_tcf0
zQ93zdVmx7urTH-#-%mUmrI@9RIS14q*H<G_+p}%?UCd^w_VJCXlL<&S=*=5Ue?p*;
zU;);B2l=n)A{{1}XD?kwo|U+8ytKd}XUX@hgjcL*54}6+pq~zNYmyrD`+%SCZaa=c
zs(qn|AkSfzPSEM-L;c(HO%^-<udnV$nbUcwSs)z3gwsT|mXgXGSLlQaN9HbeqoU3R
z*QLvBfWpm2W>@esKV`FBwqr6*rogz{3p^i|LV(2YYz**+fW!z+a-+<=!6So6eMeH#
z(VUq*vDf5yWv@yqG1=He{EZ+qrSjB6pS60ZrzVD-_dUcgGH~*;&M)MNCK(T!Xn6qd
zw)1`E-2v0|yMe1Vj)!Z_U|Ek`j{6+@`1H131Ey8sZadN+FJf%n06A{<7uW}PREHe2
zy+~`3-iRsTA$yXPyeD|`(`}USEf&TH3?FjMkF2v&o^-QnZ<O{9%OnH=pOu=-<RQ>E
ze11~S5*0ga*r#k(C0{cJSaUo(HO14;Q;CXmCY54|K<~6KCWDm!!w|Vx1lQduw**|D
zu=3u?H&Nz8AX9gMvvi)PU%8vQ2=QIHK4Ot?w=Zy?QQSrqJtob!p4Z>SIUajFz4YJl
zrhpDZED1}vH=ez-Ps3MevLzi%KW{6BJ-b~#YbItVNg6#vSjihT`52j(pQn2@Icf=G
zeJa#|REy!_I{EEQ&pJoHwe|SKB2^M5NHang$4TCjIa{H)y9(tEUf6bC<T?rwZ@iPe
z8%I43TV!KzW7jSYS|T1dqPUnd&%;Ayn;(DRLtz(>dVs@k2mWr&nf6PB@ILHC2l>`E
zjlX^)c9eB%XzVkmbeXs1Ug;!-=Qin`R-{{@6C2ZNn8fc~J`_o*!HBtacV<h~e2b@b
zlK{b%H2XfJw&&bv9}(Li++&O$OyBllLkn+Rq0l|y^|vI;-N)hoVYtJMPiWce`#~9;
zXhD3b1NLeKzbLnYX@R$TWyAAMzmvICi*44MEI!+}Xg=k3h-JRT(xqAe6i~x=^L{nc
zxp_R+7uwhkq2Ko!Ah7O##d9$0N2DV2+9m#E%rAKFk<;ua&8UbDg$tWofZ^;|fHqMX
z`|ocQ+<#tbNZ4VMrE7ige_L7_SrzyF%sifa*q`@8$0A_!N+bH_v*%*cKyw7iM&seH
z_MHSQm>bf*Z8h7zz}HKeq5uU-w&RenUInW|M7dN~WjXMGnE#R*K>n{E5Rd<s=sy8a
zj}q)KyJh+d=cbjmm|JZ*EYS*MYn<&9T)AX=^suei&2CQK5u1k!CT>4RQ8_i_dk4K%
zS)<ERZ`1VU%3BL+U8fdZXMbz*2~+i%O+KYC!hf1>;v~pLEckc5ciL8>vkgF61t=%h
zuB*quQ3>2a#G@y38SmQuL}sJ$7u6B`0**ho=D9LcrV3w0<djcht^PozW4#Eqv4#*G
zuS4`nrYIAQ5zCH80Ni8qQnc~M&+{DjZI%IVa*tc_J8AP)$ttD_{m|>_Jh;}KiO35U
z(lV&${pgcfB>AnsQ|aPnX@^r$<xQdmtuFaZ)=xSiI%{r%1WtpA^z<J~#-f}r4WzNs
zq!iXm{am;7UQN<YB9&wFK6(1PI101VNLgG~*|ov!Zyb3Bd=C9qSpMeIS2+4_Otod#
zZ8|qYn@>=ycwa`Gl6_2=MOX^4NZHg=g1xeNrufFmcA)93Llqm(T35RaJfME4Y{y;%
z&wP`(Thx>cp00Zc!ngf5eBjklenmfv-Tp;MmUS9|A$UET94gc=@Jdb7w_y6Gqvx4!
zbv-US4}zs}=9B!a7=_Bq<FCqP8|kx^7Z=<VSWBelOo6%^;9k2i&<tmGto5_a(Go{)
z>Rv3&sPBcw-7;iu3TsCOXrH6nng;u;`l~jJg@azn67%Yn3lCPJBmY2tFJPwVRp5Fo
zo)*Ermm<qE0cLU?tz?hmrzf5a5pLF#-)hmtD|5a?nAcUC?Usfn+yX6d{cZ%WH=Edg
zVS*9>a#SJi`_1D~G)*V{#J^Gr@{{FfM75GgGKZlb^SB19M>~H)D6;sNATF)Y*-g6-
z>$ZeQqph@5I{Ey2fH9Z(F8k^34u|?$I9}rE5Fhm>r!eo`$BKc6ubF|Mq$Bv9p1?sH
zXn4Tfl#m={fF`UrM-2F>K4qD=93iR@XX2?_)9GViCYQX>cb3)rgl0ED;L7qxzoQqE
zT5XG0DQ+iWEm7vVt9HPqr3T})az!T1ThE-qN_(iRm^{%_JKtGkJs}{VPNI5<?~0oe
zoc>k}&z`+}t~{T3rdnMcLNsAOy3gkI{DP$To~;BB$4~#hL?{Oqvd;Q5LH(=dm}(|U
zV<0Dm>HhL(8%SW}KMcG2gJD{OUi-7qwSdyQ)!Ux((QF5(D?-ajL1YUhj0bSmAW3Hf
zCd!F`_R8)nIq<ZD7na<VuL_m}hFFFOfBLOZffDNfbg?gSEHO13+qJJk0ah^+El-f?
zIIM#}M?IZ|Cf3eV5#U%vk4;h#pJb38S+w5c6DA;VA66Dc;}?w>$9eyo@WOI1QTFi_
zC{ClYu{YwbKZNL2`=5iA@Mn4+{NM8YEmpRJ%}RGs=I_fS2L2@3_TbZ8nlLV{+Y1s0
z*pt>8QhBd34qnh6)CYL1w?|ATS;)w0akz2yH6>ZIeY_|TBZjPB(Fg$cvPZ13NtJDA
zap&KvK!8kMcx>xq)Q^>8A5v7dc&PeMaVaX*8t=KeX3Ko9I$26l+7Ncw^u`A{IsPbD
z+>SFLSm75a{l@%oBKMEjuf0K{wM|UaUO#MpcXLFKzi~~b*BLPvuh9QV26v9(r<5+?
zinbwt7xnJhpYJ&}X<lV>UYir+w9M6VSWBhV@63MH-C8jvN!W#nbCY5PRqc6334S{d
z0t>}px|S<r$hqMnFC{1{<?aVNyaP#ji>dQn=6El0QaVqfRudD+=ko}y4Vi`W&FZn#
zd|gsS`HGVfGMe8kIqlOD<{v)3CBsi;?2Qk)#e#%dd`aBzMsi06n-7jd6V_FJ@p&__
z{nVlp3g^`Zm(y=hKP^(86+f#P@0D=qV^q!gO0btd(a>0!@rtU&Q9$~k1TYm6V)(7w
z)C>Nkiu;|lBc=85)sYov58oMT(S|5SD7!q#$G#n%9)0Gc{A3BJ&#^hOSSwoiGX{by
z>-#c~f1s$w1_J9<^W?qBBZQ=xgtA2hQeGQMrm}NxGvs7@)9r^IQqL%fq$s83Wv6fp
zGvGf-<P1*yj0I8fH6IJoUN^`Ed>n3KzfOh9=TFpD80-g_DetJ~yyo^{j+Ib>P^&%Z
z(cDqPKYcBz!j7zR?UvmS^eJIdQyVeL0^&j=xM19E;OFhhtO!1|mok>T@0T7d!4Y#(
zW;1D1T?sZm=0P^@pL0WGO*L=hvl!D_>1=Kfd7=tSKgFi%M5o0?!0d62$Kr{>$*vRI
zWdt^;Y&G%mibIRFWr09{wbSK2ep&(cfq<!GgIR*Q62z}>cwXiUHB(pM(#H^?hg`o(
z<IO^u=OhIViXv7Cd;aFtiwXP}l(R+%7EoRB?Zz4pO<NPx6D_Fs@Vm_z#PU85i!%CR
z+jt!9T|0gJI7N^uU~w{6(YlnW{=2+I5={CvcoA;C$^NPrruJolG}l$nK{J_qLbE`7
zJ}})u<xeEtRtb0d1^4FlSN4K5n!oy78_sODSiR)KCtcM~F7Gx;F@RcV4MgK-qyJrr
zdHJmGM0fonC@W<6Pwa%B7$uA;RY>mq2w=m{;&AjOC7Z5UxXU~s2`O&VaDynA0$js9
zmaC|04PE)Sc~073=H<r-#dl|d95arp$;?k9OJHxaG%9m!V?m^&>&;TG)tDV6f4;jw
zzmD#Yy~xqjwtHmQ(4y90b#&t1SG|4ry^u#=$V^a57s@weR-wa18LPeaCf7JuB3L9+
zz{tYj`6bhhA5S>SqD{K<3SrWKTD^{Tb~DwqSC2)%_<`OTh1#^CK|9=OhcloL6OPTJ
zO~_;#Pott_1(JU9(M=2Ti7trJ^h$Oe<e=HOJ^{X;;<$CVmrv)f+D3lkAn)3^<N&rG
za)P-)d{GL^`e?5q`i5;Sn1II^vEqojNm48cTO4NWdYG|>!E}Z$&I^3SAwOv41^ZT-
zhjLCb1eiuqL~JiIG&|d%qMRcm850=1McPgZKeJzf6~sTaIN8x{x^+5=Ez1U0-G{_0
zh#I;s#8HQJlRC_~MLiZsd1p`!x&v+kfGPO`v{EVQ4j~F&3ar1faJJSyW{sNETC%L8
zQDISNdIZWj`Lw%1YiUQd8T~cctI-&AZa12JUOsiJGbH1yS8<GtrB>jO7+bjrMC2!d
zI|smDD$WOljhH^}>ERaZE6~_o%x_b+DNPK<2D33}{~|rED}jEGJ$q|KFJrOw^(3=P
z<Z|#82}s`y<UR#T?Uth=e<*+vwxf-vsIb@V+T@=E>bL)4tk5bS_VG!fqqN4?SezqE
zjsevluX_J2MO#BiLeM824ukxX-#?zfu&W2<^3UTYVzrKB#g0)K4#(8YA_0+cRYW6H
z_T8|yKywV@XY~2S_=GX?hwt57KOcD!KcfXdys*<DVK_L+VU3aaI)sOjT^?gCXTgv_
zfH8!d*cdBe!}bz!D#J$LBA>Ss5E+pi4S{$h%*Kyda;Qo4W-IjzXU4Q7+MoNJ@dsy`
zj$!3}D@*7%-(909rWgw5gS$@T=LLJxCksg}h))V6fs>EA9vO_1)(NH#`Icv#A+*An
zzdW9hZo6cXhkX(8h``Vv5xo)F9m=bg$~}OwC&pm81}Gki^mNPXD!^6HTNgI!^Bae2
z#BzPgu+G1bwg2U|@7Bx0hFJf--RhHLGHoKP$f!2JAE}f{;iW@pgwvp@tgl#o!W=7>
z_d%FD!!wNrOH83#!fQqR4ViewhGH;D#1~wzhp3}=mc}E%j2q4oi&fuY0R0!^nDl4O
z3AwaGB@mDCgk735(~Y8R!R=$B%>ahJpr@YQ)0)^_&M_JHSW@vqh8L)(;5kLr@`P@U
z+T-Gk-Ljx?Im2ahZP<$@o!CQlt^c~)RB>x@W^;e`)?X^r|L0QgZM!kASW$7tR!&GZ
zJL^r<FUD3FDPv;fTcDe+{K@HksJ^*0{$!bXzCgidYnSkgIrFO!sV@7HgYIJ#2sX3T
zb6BM4)FsF(0anW=XwNk-`JYJ|rjh0RtB^yYn1j04Y*e&aQeM=N|6&(L2OH}TT--Hp
z8=80Kpm~C=q~tHyXd0hW6!>ybJD`trpOpNm0BRfBCt`185k77^9qcb>XtjPq@KXQf
z5?psaj6s{*<&E#_JYcqhptTyR+HQt>X3QxvC(%)qlW(iZc_!v`nAv}zi7_=l=%#+~
zZ+sUs!c;y8^y{d3sjG7(;BewL;D{m%<@3w3(S$!c6HNoPfq_~yT9>S1jS>rC98)Z~
z9Jq-rTsFi?(#cA*hxt?fN90La=$h{%ZcDVeYhEU9RCOnJW;JhO&$5_W`t370rr5?)
z0{po}9VZ}fy*}VQ-K@!GruO1Vt-1om)=CxS+GV;pIUnCs%yu|Gf4;c0{TB;vY|^Vu
zH#yb0E(g;8U647)s!JrA9k_n}Ba&c@?tWbPyCK#7GJw)x)5q-r1!_~K%SZG&tZ7eX
zGRcYbODxtO@Z}hdLui$eFh690Ix^NuSlvlB@g_?e4h>i{B??9JY4^+8xvNs~d6!R-
zu`h2809NR320v&6Y~LBIzcXVRkQxKaiw3(fG2$PTzULfnqMUt68JwbF=)z;KlJqj;
zZ`8{jP*muOcdqexO(`Lh9ywSO`5`8QpDjxrNY$UXo005Fqr<@4&$2A}CVctuP8$0z
z!kGW6E84}*4T}=SKLZk)Oto(`;qwORpdLDZV4Xmnq0W#wmS^BJzQ(o*5%UM1r?~Ij
zZ^CD(*57p3!HtT*-&JFWKNZz#QU`YvUM?=YV~euK0|fG&q6sir$tEB_0uu3UW`4;j
zmie|3`&R4FVjRC>#^pW_w^zBn8`2@E*ZVt{wvTLQJ0{JEir<9B<0@*QWSo=^W4Ico
zf}BV4@WlIcuh9G1Su=;RKf>qO2bw8s7G^05xRqF49=828x(0>D*W$KAo8ECJTcv~!
z;C_|3N5Sv2t<e*;?Iq31BGRSleriv}&rvH%9e>l}HtMfgRtdYIvsgkHET6vNfphvk
z`v)=3P-%k4D_K@TVgjlaZ3kw;Fw=m1r3d|R3OZi{wM0v3g~JGL&`a+?tMe~l1c|ZT
zoRyIxyLh4wX1FhO-xxf&zIPor@KBj-xM2QKW?{6;KfvYaI<t4)Hs8D7y$uBo9gD)B
z1lE}OBT0+kODwmka=57%WdlNBjseE&nqRe&Ud>jAL+inZ^ChQ!{to>Ci+l_5A4?5+
z$w%Gen??e(pe!_G<s;4+1dKvMqTA!i&wBXA9{0VLBx$&+f`bgWrLBwA)MAsX{uXt1
zean|O#Cz#oq|R%HpD`qud8?OeuDz8!=7>K}CS;?0VFPU^>c>s)BDRs-dD7#DfRY|P
zq+z}O{&!903#b(PK6rCPpm{p~48K1UC^bGxALf?Et%tSD*6|516$g{-HXz0_-Tz1t
zdDrIT4!hb#EZ4RT&uK0NarAp++T}a8XpmWpBq9;&3ZhKE^2wX2P$bkgY+@m$o{vm5
zhbI4O)TXkEA^Tx6<$E6>uIg1Cs-?Dz&ie8C`^AIWV!Gp_SrgUzaS+yl*}s?&P_NqC
zjN63Yq%-F;r%YXU%Z$w=prhE+hx#?ZRZjzK%<U-D!l&#$-KBLRn@3*l3u_MFf-_~1
z2<R$^cpR*rjdMmuVj0&OB8UAn5~*>>p%K;Pz6?K3_g~2ltsGx>gpVcLjlkY<*+Z}5
zn=61YnV@VeJyuBZK!eghe3tG<i42T)VIHp2A=2M!D?U(ddST#iTfZ{&$a!Rq`#-pQ
z>!`SzZC|tr1W3>hZjBRMg1a^loZuc@gS&>{ZoxwE;K3!h1PB`39U6j$rg8t?%D2zn
z@1Aq_edFEx$9ZEs#$a`?UbRYQRn0kT)~xvhd|)%dwhi9`WpM6aoQCHf!(tG|cPGTT
zzjwZ9xAA+6CER(;s%qpROsdhOMFrN`A6J&IwSvqk$!m^3<W%Tjax76{Dv6*4#vDju
zR?n{oIdk!ln{=b@#A$a-OdwB>$Kbt0h~7E_P|#1h`ylq*@ylZ4s*^AuGw$>Rf4?11
z)?m3lt&84xNNL}y0IEpG>1`olKF`E2T@c`dS=v99;ax~10OHt)HToy$d9l#BlUl6o
z)aSN*xGb0Q!fRjc&(~)IHL=PI?zgwCT(%#$NdBjtO-AWB9wAp*Dl~3G0V{V|=2$gN
zvC@9c$m_<40Ve~nG+1lbX}+ZxEq5B{#iM+~_<Q}!!wc`(IAb*M0MYJ`Oskl!48a?X
zmG|#fgLF!!F~B{CRl!*?{3r&$h!FhZji^&Q;A&ixK|j`5BdA>iYKn{E2&VOhf-GkO
zEwvy&s+94Xi!h<SF`bQTB44LdT4q3Z2!=h|x7AK1=smo%P*3Q!No~gzL5y$1d2(wP
zLD7pJMRwAW<~)v@oz(M<PodhgVQU(8m}U82PUXM0dQXv6Wk@HgWe|49Z%4I!AxY-2
z6fzg`yJcmZ)vxLxUVwk+qvj%1`}o{Iq?1LJzRV?CNHjk0HFbrQltbd^uSV+Gsrqe4
z%48IuLFo1wCzeG?QBmPf!mgd-!Ojlc{gm!i#=TelRYF37CtqzNZZ*Q~K6uDZ%&5Oj
zktr-1a{<+*6(*Nm$RHGu94b=<mBc&@z{TY)3kOUd>a}+7l+lgSeZl_st{h`=gscPX
zTqo}<_nC|z-k_wS_Gi|oO8O|BgosD$DL8Du(R^L-L95U9gXX7i1yN;6J5U}Kndk%=
ze03OITOS@%OQ9~e^Gt0vEzK1-x+r_<F0civ5tCqdU#SP7Mfil1@|)OB6;w`_HGkZ@
zo>j>yP9!XS-jtRpJ4z@k$HX9?FtUSLj8~n~^33eBumD}kOST*ijue{H-2x7r4z2(`
z0URx|9QU^QQIV-=CIED665Et9zr13p!|)&!A?2F~QPPese$ZE9+tP6V{Go$9L*e6h
zLA8&H64$o-C{aGahBK!bG3`NbeojS14M!=pN|s^sm*1{6q;6yj$Y~Dh@PbY~tks96
zUmR;*Yl*U3G^oqsUFexQxC@$o(A$x07)9vhkfcdvhbldtqL(o?*o1+JTlW{rbu;^*
z9{b-}_X>(<Zqz}!*dJNT+j@$oBzXn)Glv6FP(Fv*6o8et!A0uNi!?nP`gf!%<^64W
znW#@Jq6}Z{zv6HXERHE6dO>hnzI9gUMU<LYg2&7K3{Q?>rtJlt-II$3(p;4nyr}eC
zI&xX1t}dPjOu{9_=(7ak#wt-sTX6#A+1KuKoN1{yZ#c#{pd!1<I!e!b@vfChL|>*p
zW@_|lx(>IgF&D{KHa6N+C1tVjy!ldg!68rHZ75Prtyp0-rV@|8{RNe7e;VDE1euqP
zyX6!mU&M@KIzqnl`K=|5X(6(%%a>I;^kFQ?G7oMkT=RNTjIUDc>WvrXZ;DIuyG!j`
zQZw(%kcKL`m5GFf91>Jobf*rWUwb~vY)$S<oxUlwqPgEtGM8<3*K|+a1Y?A=XI+fH
zKBi2YB|+CR9DJ#5*Z&~Qz@k1w^QC;B{?!`~4Z9t29zFqO5~bWC9UB7jvPsj5P8BOm
zqL{ZoAz;U@^44B9dxi5arU-|CXx^np?r1l&p=iQz&J3+~ozMlQ#b`xUZ2Gdh`WV}r
zXI}WxyO7eeE_@8!_^6Ss`;nAzFM$^vCPMmL5pme>fz44dQ&$aQjCM7c9GC!7NBA0s
zUDPWqXi%~r^(^*Nrj~#f>X)wv!EM6@j2PQA1TTfG+~3j2>+x7Q8=;quwnyW?;-Ig6
z@hIrav8Tm9@Dum8pkeJ5OTPLrYGBu<f)k$WMP9O<hY_J+Q=n2Km5WiiQl*tjaY-zF
zirh3_DSp(VDdW#dyp;X<kxfZfn#=KGr;_60smJhpQZt?B>NHcs^HX%P+D{eujW}Mt
zDx|a67eMvMcr#8cm6M*SK~T1<I5E__HzjrB9~FD>NquV;B~zf6=FAUYlrVsk_|0Bw
zWpkT>d%1S&v4{EzoOpnUSiIsFi*J1Bb5pJ1=OfyQ8Ws8xZ)Ou2nBQ@-@uvxV(j+WQ
zu1~3YK}MKNAp@Yl>Q>QQVjKvIKj0>HJ(O4IJj#@?Lf1ITUze8KnM`EI_dijf{@jw^
z%A7sDSXR(HX}L90!klei{BhM7Pd!h@m6K?!h~OpEZ6bR)g7=Bf{_Av+DUsf>B#)N|
zP+L0+@WfHD5yME#+wwzE%SyY0NLvo<Uv9tl!x|oVsZw5L1|meF?9@qB@jGS7dgxhE
zxiy^gS_CXi2?dE7c-sntkOh*fZ46a)K66joRU0wogN5F`Xww_aAnT`NRAp2mu)sB<
ziZy$t@vH#)EI&A3GP8^vk?-JB8b37F6H<>g#qo1|7TZ15+?%T?EHEN4UhD2Wz6P-I
z5n8I3Z}~5tYWZmDf36l%`~u!N*w@nwiMCVAVN@oWu{_A<!lvo86c>|333rsmtZi`(
zjTviq-%Kw~Kk&nK@1g*L=NFlk!<e%L)dt^q)Md=>M<<$vCkbmS{OnJ;HbPTp?oega
z{qj-cX^L&sY>(0uc0poh2Cg!mRBE8BU?vGsL5CoPXd`K+yUIpd^TwL@k7O-KvdmG>
zzC!(QYG&Eq0R?j!_2-JPXN%2=w2DHliCSOf!A}nsZ8;;=nKw5V<$4rwX>vGd&f3V0
z2#h$0nfE?d>3%FVVwz0{tIxZK7GS@sNz&O(Ys4a;ODtn}UC@vJo#LI^!J9s-`SZao
z{~Y(8H-@RuOp4Txp9N{G=%>0rX@Xq`WO3w~=4G8`zE-&K#jtP)lsnsB=#=0GDs5p3
zCcpMgQPXyPiZU?p#)YQ$TBec4c)#_#p!B!vcU#03lO}0txpBg*(A@*zi0~{1GHL*0
zvAAZ!Gq%`>DGZ?_lMs?o%^|$e(J#|n@tT}Wg;9dO3;I%OjhcfWgQEju6eq@`q(#g$
zqvNQt_M7Y9SeP_Gf{l~<onHS1lG8`VB@oh<eQG0Lk*cZP@G~WSI2D{zJ#q3(*|b~%
zgKq7a*`D1;{W3@%H+x!oauS;w@n?C0*WVe_5{U*oKK%o&_y5W5>fgju`gh)0|F4o6
zux`$WZa5tF`sfOQ67~whk9`8zP+#`42lzdwR5-t_KifW3XFt3|TJ+|jK)sE-@@mu*
z;8T)Ad%JM`tFZ}!yuzv}gHC``Fue0Smpr2$S9px#`l%I5d(g+LjaX&aeaUzyLY4vX
zsIhw#tGw6s@#c;e&l>>fJG=bXuYVFP5*MGFjBKk%!zp*f0tfrn#E1O>Xz>oB(_C>2
zGbiXjlMMSp?vwsYBA_fFSAqaZNMPij7k$0<)T25cxZGu_qV9TgCy5(k)vqC+g`9uI
zYWRg3f&cN0I;ti#7sbr!e<-hQD^}wlb#vnR^FS0G-p={YYO0S3hy6h99PG!)Jt3N}
ziPc*bBKXkT?H3jUAdA+VKsL>KVyG}WO|#9tDao6Rzo`L^!wc}vd!6z4ZAqjC{ysM{
zAjI0kD{N&9UPS=uK-1*82>KSS4OjS$B^F|Rwp#?5$&gpTYnDjS<tF6CikE4Ws>CP`
zM2<3Qk!CHoqA=|Tcd{xwo}?gp=i9+nd|TamsFfu{#A_IJ2)-3PUA#H5f^-7^OH^Yz
ze9ywosI8muiXCGW8{seHP)X8|VwDZ^g8Lq&2jw3^)cpYqyuz5%oY?^zt}M~RDLBId
zPe6yT=WHy_hG5M8<&Jt6tdd6C4hm-xB>$&O()!W4tez(?;FR(~zr>3K*aRMwABCj#
zqr&GSlB)_L?OQgBLBb&hJ4N-Kp$pxn?PM>JBMU#v3T(^{bs%%ygttnk-$|}F=?)w?
z*Ur!&pN55ZlkGNCmF5o^ebN`Hyp$d#CElXHQdPCGcs%Mt*uUQs=l4-J4?1vmSuIQ@
z!t`%17%LXey&IMXkT)c{7|mjyJoR|+VLUwz7YQoSc?`VJl{;{9IXDU8iDeZM0(E{X
zm!RC5{5D|R{tfOhuPbfCjP?YFg3D9LXv4M*y#Kwp5iwoO{!$2pD3{5iWsGm{jkR4w
z=N=mqWv?xTv3+~xY$=RR1&w*V(I1V(!ixf&r$(0oOR;h~8+ww#gH-Q%ui<rfc(rL*
zpWBbKZfEU*jWLo*ebyIW!Hcl)dHYP;H0qnn8$4rTs|;G7x9B0?mFNe~=KF9$<}c^p
zSpuN(ZE&)JpO%e!Uti><IAl3|@Qxp`OsBC>_s_z=k-$Ulo}43Gv#DKScgHnXUavDS
zo%cGoR68PzxYdl@3GjMvo5$Y~v@TgQ83d29M-!%>J-k~_>^3dZDW>fGHTV6e@6o=~
zqi?(S&RtD(fVNI68X_MTu$*%qpS7oHp-Zd76jp1{7dSHFs{N3quRky7?l~Hj0cQ@!
z(#<CuI?Be=s4AsT(CjzaIh2882p1It$N)&x0EAeO6gfi(4e=QSI#|ZObO(XN1T;CD
zv?-X3KCCpZXLmb@M|FgSa<(+(A@~%l#k{_i#5~RD1ikb5ez7<i@px{mJGp;X+wSJb
z%Gog?9!HH-H;oMgU_<P%+u2UFNgO+F7NnJP`__61#ak|OG6P3Eg>3<Z8vllZ+;gPH
z>%g<c<Df=kiXen_Yu}Z`8m}Cu{^mN>U0tc~aw>BTDBf!|`_x7p&kqyl@#Ru%(6#nq
z34k?oFc=8>9r38~_sc?Yjvc5GtlC_AL_Y>rt0-C@1vQ0FkRB~{$m?;t?Cl?Keue{K
z{K(#LejoAxKx^t?$4@mL-}ChfPg^VIK9aVV%^eC_9!8@{bn<@hMJ!`?iysdcU_VG-
zYaS3{_fQ-GKrL>jL8oFfY2ATC9@goP0?6e+?VBiXRIR;(=Bk;4F4<dmGIoOoNq*Ek
z?G)Nv@$q*lO&gDlV;=X8zq<$<KnT`UZS5|765~^>Je#@-!cG9dwDPcO&;Yl=qf5$r
zUhnI_K=9h~2idTh2#H_3-@ewIjhw;M<Wx-P1FHmW*u>$V7py5SWoK$Y{730;hY%0H
z!*#dDB2NASHC_OKU`?LIT^Cus3g#ZCqFP;4ZlvakUr(+w(9!RmONmy4_Vrp9O$@@l
z244l?8DQG62bA;z_<HQ40OabAl~)rd`ao^is<5we#-DGo=|s%^CJ{H}#gX{M(|L4q
zW{NaxuExKUep2bj*#Woruo%DlbxDcr_Kp|*<zt!W=Vs#;@)}?BSRfCIDmnlZ$WO!-
zfHbWJ*;g2K1s3U?qH&c3?M&l{hx4Mbb_ZT$?u?6a<DyamgMtX~GQBJe(PQsMM|3~~
z1m0IafPVWCwqN?C|JI3f?KUbQgbU$F9P?BM#S8dS`~@0G{tG11CxLdY?MU|YVnX&7
zHDb%lG$Uw{OpT0V;dp)o>26-k|0lk9#6z`Po3Fzaw_Urpg=24@Mt(t!I~Gjdji4AJ
z{Yb0{#;w7szhUic3km;(-UU$C<UXO1c;a!u6YxqrXdW4d=H9JecANV;z{AQBtT#bh
z!%bf%%f9zaaL2JWX<vI_xaeuhxs${1nPp+lNV@}6j=5Lk4pfxaxwhBA=52Nxm~ct$
zLQx7F0AtAEReO_fOP2BmX^CPJD=MbPPv45kn929mJ+4DeI^|QVLZzj{njiZ`X{jas
zS$I-t700uZa<fQOKjq@R_(YPcn(;Dv$6{w^cw}3F&T^D8?v)4m0p1rA;_D7X@;Kso
z&)eo>gVuSoR@S*@Co-93pJcYCEW;Pp@+qKHt1L0i%omnR4E<p4;sGInohkX0hpU-K
zV&EKA&E<&4Z{Ovwb0cHZmVD1%eHG_fWWChgmCxm)<@0ahE*{I`8>a)&4To#u%IOum
z4Nbe(oWG%Tf3Kurm!v?*6e?<25}FcORzmdF=P5(Aj1q|wMfz(cQgC_>F*vh09uxal
z%vLuQ=X+bsM@N*S*Ct%avFqqKt=*dBD^H%`Bvl&<G~e-^e1FN=7H;iI_N06I1q{tW
zCmQL3|N7WFFhKtMUAv6-EYk-0_e-2)cA}a{;S4KM424ge6;@B5^15O}Sfm({`K}$6
zlwAXj!u9eWRQQD7f?9VJzwfo4#Pr^rm%Rtk@O#T$t4Bue`(H<jdWQI+Vsqj|XoP2g
zp_SL-6%=R**Kyj(95XQUj?Do0$5A$5TZKd9bgcwq>ufQy{bUO4?FI%n$Mre#J7g$|
zmGGG>bh!9G<HeA2pyHPGaeNP;3!+sL3_X@`Q@El|X-ucGU=(t2@S^gVP<<<_{y{_@
z4IB62hkU_Z@%f!dk?2duNfZh)Mt!Kv?4UFBlXEuVD`ilwH{$=f?5$5Xh<m@E%<4JB
z7gh6mJ)RhIuxKQj*b5f^LLMd3Fkx_sh1_dwe0^RyqlbOk*YGNJ-8xIOFVpR}C({(&
z0Kc0r7duH(2)8S7!755Grp&9I0U?yyLUf0ND{LnrqMCb`()h@@V?qDac8aZ&Ws##N
z>kBUt&{`?x6=AifHgu~l8u#>nfqbNk<YE0mY&gAZGw(^SnrZR~vYaBK#IG@-4b^da
zc1C-oKVtHO61?3yo1Ui1V8GWE-UUcbPzbC8z-9s1y~*g!O~qB@i&gbXH66J+6)zv)
zS-Qx*ehvVWJBaD_!DyMmk?7=J_p{Lva7F7z;p+q#_EJx27;`fa);in(h-R@aT+_&~
zv-H&{EcvL47SGSdB>t(zBV|+j_w^1>FSX08fq2?0^RsQnx&(l0qj4nYI?cOI74FaQ
zByv+cEC-j}t2yvLN-y;ckGOvVcKKsPHN_)VtH&-ErvWHxqV&H&eqtSf#31}p6Bg_d
zfssH^DTlk0I3+ov;sstI8Nb_YBJH|lm;uB^HhY!zU!XvE<aoG^I8K`hu(R#y9dx0*
z^A~8XCtXS4tN8?dTt4}HhSS)MziS!t2>f-<QNyb9Y(o&<9j5rzNTY7hN#$%p7<{DX
zrUEGVa(>6pVXrV~lRvFd{JQ!D`C&>qjzl=$k4*#|3pNhRmc!aC+UHP<yEz6#rT=BG
z`ms^@0NAqC5Qs*5Zqkk$2+$!2K5wlIkb91U3JzS`Yw^eFu!Aw*p;mkqa2c{yhr0mh
zmfx-3pq`)x0mKA^rhnTT!9{l9%W>OvUEv1meB0ys;u@B1d)M6@%W~{7#S`xZq~<El
zZ9$MleQDDQh-At>tkS;@JkL`9c(WQl{A(%cFVIZ78lY7aU3?Gj_og!exGbx<uft&w
zLU}vDaa5FXcyod+&JSzu+70Kowg<6<*Mi)3mNzIo8we`U>1TS{(o_i0-?38~V<9Z%
zI$#NYUs6Bt7pUBcUHk!)<lqs>;PfftD=FK_?)cfRx^yHu;Gg+UE-@jG{cj;0%;xq^
zT1?xZ9@Ssm&szh#R2*m72<Q+#e{P^|dhJ1X%eS#7&&8<>$dLW2ICrkC)xX7^-!L>8
zZIIPrbis2LoPWB#ljQ|UTnz|9R8f`}o&ZRea~F$8xsUZeK0%Aw+jdevM4x!iV(AdO
z+;f-qRs%)^?_bn|-s14=1qbLzqgF)>#XO*$kyB>575<UG$6FVRKmG!3W&J1=f4DSH
zawJa`r_vyWqX!0@Y?Ji`3TaCI1v*vj7WM^3BG4BbL0j)CHlD^G0+M0UA&%)%<MHaD
z#_XzUjH~yt(Z|ZJ_sYn%ucn(360Br*h*-y#E?tHdL9u!hjZ=22jw_6}fZ^#$`xl7j
zw>EI~VQ(K1f8^U;R12i^Y=3UtCIX<is{1@|27vc0ZA(vduz_rx3{3|G@7+p696s~}
zCA+Nqe*-z&<81kvCOqCWm6-grXhW0jh}t^(S;!cK(`Nh^=z7}PpN&6eT4*&S6ZX^K
z6vd_Pcl)P9tOBBk7gI&T4rfzIB1Ee$od|G8R^Gkp5erc(U!F6eV$JPZUI0lr_5?6l
zdlroSM7;ZL+tG>^!GLxbptSd6uf^bPzkB^x4<E-2O-=5Zk5fnjlB&jn*$T-en99%M
z<9;f|jSmiD#wJ)Z490xfE~1bAvi<yvscL^5j}xc;)9)s6_r}*FR@cE^3so%1$4$D9
znS_H*WSWNq+cA~}v&@={ujqS#7$K9)N>OTKS@`NCXACMCmJP3DN;`vQ0!jRt8yu8d
z7yXUr?s%(Ca{;mU!T}I_7J9`>sX`8e#p9_A07xcn5;L*s3;JGWbaUY|rOnfh98b({
zgVV^l&Q2AEOvi6UozH~_4m!uA(<?cx0!yZ!Rzy4zut#OIBJB-Y!qOZ_f`sn$&rgS8
zonUTul3K7Q@G%5}b^7>|=UNZvF(ZtT+xVy-$+3xfeOc%Jthu#*bl4u$gH{B%T&QDE
zg{=2;9g1S!_KANtkQHOQayJvC+!W>7y-|-&mZx>^Vh-%Ei`A__Y72KTNh!p~NkLec
z?ROAH#EFV3gQMDm(iq?ov&2~gH>!c?eQUfns%R8`?q*q&Hjiiru{wXRh#C96;?;6-
z0}>|xPM2H9#h>u^(mZR@R%etDB6^q6aDPiV-M^g7MVTvo6f0#w<AZ!l3%-Vg5r?#b
z4`StZr|W>rL?QkfNANodp3^F$N#csLy?~_XF0nA;I<u)4a<qVr0Cr#NF3SY0>6_eN
zE)s{cQFHXs;>xJdhM|&A#yE}V%Q7lrQhwl)CQR(}$46gy)p47V*y2@;f4B|}-XO#I
z=_kDRM5y+SUin&Y6T&}z^m7RK3ev#8HsV%|b&S+4EOgZ8l0+^-coI(NGBMTNKSuSM
z_1QMTLMtY!CJALZm)1S3f){(><xJmsp4O=C=F<Dz`N#N!-B+b}Lv%-F&2&Rd1(i^S
zx{Tr&Hw%Q%^2a!**E3W{aj%TZhOF_1Wk!isKava%VBp^{5w<}nHLKG0sq;LXFujHM
z8gyChWoei;vtMae&xD3GGC6+a&`y?BR8=aNp6Q@-C|56|$JlCNOb~vh4(?|8l>RY+
zM9lh$;q!HxV=>HIY?tDznOPV=ik4N*&~&@Xyd6R?x|~Hh4xdirW&HgcRnx&9##kM>
z4`%z%p(bjy>P13kBtEs%iBnbOQN^<^!wfq=M=<2uDSAB_TN!GSLc+wzT{b5{H3sf^
zQ$At0lEPre42>5I<@#UxW&AipGb(TkW@hwz7*18}eAQ`+n3M{YixU(y>hMAt#Ghr4
z2`E8t-3!l)eYh_8187qj#|zhT&pnvztH@dEiiSB<f7+3spH_0}Dbl3U+r6mdRxiQC
zw=r)-hJTh~RZvz^8v9_S8T-psY_X5&^(%Jc+xQm7Q&I4bnqQVQ!?IB^LO%(vXuAVZ
zbdJ-CYE~o9YAIZ<s<!zJJrGDYH9Zq#LLd}lkpRV6nlN1#yy;vye3h}37uux$Zb~ml
z&E#c*e#YojeA-IO$qIhLJvFA>!N%zHLb&FM2xm;)VF)3ycoT}T(I$a-aC~HIc9a9_
zIpW-=`r!L_NhqtIOnHH3IMWe#)H4#2mS@R*E87op#Zh7OVtPF(Y~rt(hvH+m(#d3n
zUcPxvI4b?gJul{h?Bdn-zQUlNVXbfU-ls3bNF{jNc<xJabMC<px-NS9lXjMRW>r4C
z%%4k9f1RS0S)3OrlkpXl_H7wUcvnZTP<>4E$C#;;kueSkfHINm+0{<vS&LHhvhY4V
z;*}rj&rFX&w$t}{9VHkwKpm??v5ZzgB;S5H7>U}OrQh0TXn&z`_WXQ65%pR6^-{@2
za<RL1i9A!`*|X=K77n_ZYAO=CrMoPjJY{{sEleU|#zG9?Dn~~g5y{VK&oK+!8?rT}
ze~stjgMH|yl5}elG+6kM4lhAMVrj$*-S0T{rKwoH%4xol`ix4Q9@V%csM*GIEfLQg
zJL^OftvH1HI*k#3xNsJ8)aUGme+Pp`UM?*fdQ3T+lE{MBXkUmy2z_QArc5U_#=bqK
z73vBf1H!=U5aQQ*e`3fLdB1J8wOEudlHYF8J6)UKD;Tcu?uGP7u)|spfv@}PnGW#)
zqHBwN*6tnrinCK<^RkNvf$}fU-EX)r9PQGXsc5On=6CkD^G|cMBw0W67nk|KgEcgU
zkzb|^o70w%dm6?r4xqP>Yr?AY>x!uuEnoGRNPHzNOQ25<7JzD$wyq?9QKlYYcuT1C
zT{m5$tlUc{mia~AoR92LR#|iL-uZo*f)5Ojo~Pm2A|up|&Uhg}4Nd04w4wuBfH^0R
z*pj|P3j?KYx`JQ~kFh6sM=?xw@*-HA!B!V<W@C$c+Bzn^Y<QzN&G`rU$M%I!FC^3`
z@cVq_3*P9v#}v<297N!+QaymR<WTU@(7dBH%!XB;{j%^bGgcX7U|<a6{^ITf#$0RH
zx-MVzpkmzm$(Z~MZ_!n=E}Q*h4C5E|?+i(NIut`MlE0#N#N8&m*2`3hjGQRmferWH
z9<V5`DVX8QbMYC|m1z5!(N=ZT@sVMD<s-A{Grl&jc($+Pwuy$TtDsM+l$bQn?Cy~0
z0)6>(C^K~Eu9kl6#q&tlq^I*QJK0#G2ggk~0_k2VQKxs%o!cOhn>9nPsrIh+U-$3M
zOh}bABkco+-W0W`baZ}PJW2gJ89r54W3h|R`DI)A7t>VUn$|YGd~poXkZGv})nIO7
zP7!{sX7Ye*0-0(?a>-yz7v^>8zT*DfzDuEZO79A*^xJgMIkS*+s+ns-DTc)@<62q~
z0adv^UF!3K7BiW5>R)-o?d(W!`UPoHKAh`eY^P|P)>{rqKNTDz5)DwhByh0}9s9V#
z<i2d09LspPg#B5%SnZHhk5S*j`EAylH!yAhMZm6<>2#Xb6`xI|SoX9`kl@RUUR%2F
zkza>k14V1xO7f&h9%x1U#VvZ%H*!n^vh#ANm|VnfTT6NAsVcvM=PjeG&G3X`NQR>D
z&qMYJm=iOKsF}ZHriyE54u5cnPNL&bYGLD)#>8THsSw#_GCe(!q&J<7B}%b3jQ?6t
zUq_Vm89mZdt(TIx8RMUsQV2A|IBj*ynI)2uYsWC3584SawjWp&kbms?Oqbabd_UL|
zDq(i?hW%-O#4L81uW~Dc-7j3uvyf4l&s$nYVv3`%Ck0vZ#k7+`>Pq_}!3yPdGq${;
z$;s54ABU^7I0TerBr22&8CG6tAi4N^y!v4HIaL-az}3yGy^nO8!uu4zFw{k*LZ2I-
z0JTv=)1rGMee~;CqQVt58NG#-(l0v8BIU6p>+%ey&CKZS_WiS>u-XE}o}55O!<dm)
z-dN*E1-uWhmT*j3$yDdFSw7wxvvD~W1*YkxQZqA8_FnIN87>b))h?yCQ)ikE`+oE(
z6<huTp0t+5N2VtsBuunXDJTdc_;}v0<ewUtPie>EoCQMn&eXMe^>z%zdZY0XHV#iG
zl}GUQlcBDil3v^bzA1X;Q$w#v>QVehSA1p=GEZpECQ{x=39@j|Tq6BB>HmMcDe}L)
z`SLCV2`4O+yX%wWxj5xoe@Z&nFS$VP_l=y*fnP*?b@QJj-1m9VTba%6k)W8Bckf|8
zehu~a^drt^b^c|Dt(!`S_2dfC#z*^0ujRpZ{WF&q-vJzSQ445GxiX7nCSJOJ3t2r&
z)q{78X64JY#g9Zde)wy|vyJ9Qc}BSNbC{C20>1bl@cmm%DT(=fCR6w{Vz19-nsOSy
z<`j<O2OfY#EV5-cnu(H6%C}PQonI_Qwe!x85D*@~QE34cEP183UBv5e#VsEI0{-I<
z!w4eAzSnv$-gJcXZXxs(AufAcMug@yeZ5deOY)E^!EwIb--ro)>n^A0b$uf0pw@@V
z5%X44*B{+)Azt8IY?$~D`blx`)>;!I?Ay5_4*SVfEx8L&R`33|c#5xbl!o_>YW?be
zIx3dyAz+RuglkZG88(0#D&Uw4)y}?>6GQHbNFXOm%S@6Iv{@5r>@iW_o%=Y=`(a$z
z1xsM(_uAlV6wnTW;vx0EwJA5KNja-{;RHuD!Oh+q&D(B!)6lrpr&OBN><ZlF#BzP~
z4m~@A^6l?+nt~r^H_kXF6~^xj23FgX-ugU}x=a=VhM7tqY$0+J9(}J|@E{u1I?T=5
zV<DL(fz&)*q+<kQwdG`<9g>z+7=o!Rfi}REz?~<hDrT2=MRwTkcWu1FRxQkDfO<Ms
zxtYL0%3|0CpZ?HsG3xsbj)lA$;5Zo`GL%JW=qj~4O|IYERZxQ-{>fuo^m|``{Xre_
z=J=Fod*;saO_8rFiO)tAjx+cWTktmEeqlvi)uNh`f^yZs#^++i+d%R1jF1^bi+jGA
z2`e<6JSef@H{^Z5y$$fOcK&7rkI>4wgD2Oxfw7!&?%Q=;k$mSnl?`%?14MYOyC>48
zP3EzU0t(-u&0yuBsbIqAQT=<WucesVg+JWG1K&aruyno>ckj8lom;JnDA$sBW90bb
z%kiB)oa?}w=K=$7Y&3~Nx-SolJA4iN1Q2t0!6=A4a-Sx8qx|28lq!hzE<f@hR)Oic
zj^H$<=r*5aPBNFHPNSQ<B@M~BPnIJl1x#qokr(R~D?(O6eDcL|5eJKp6;d?9g4bWG
zA^}`6zr%KS$kDZ40|a8oIj-QUmJX-SpfQL)Ob|Xn%q)k~xCqsh1i++1y%s38hS%QQ
zUAtgc><k3@)!V_Ob22T)dvRDbb=QJePqBJkreJw)>jzfuRXn>&-fxjH*WzQ8&K47#
zbhg$zd$TM$vEtl#MMM<qKQE{g0+7|-?`=nN-@*NkZCe*Ue}*Ycc2<X!AJsOY_t9o6
z)QC{Gdq0XhO`3B4EbG}fUXB)om!#=4EyE6RM_AzJ>nlsR+gSct&mr#TFv6nw1{~dW
zamxYGn}$ffjJFS0<yGF(lMi!>XX^ujCUWnxc&0}>0@Q~_uNJkJD4&Q3-TB^dg@+Vq
zsH`t#GEAJoDIsX7<zBaQ(Vne$1NDKw@8vow#2`bk^&rc)L~`-JVGcLz^BrZO*Za9a
z{vWkR412%Itsnj@VZK^JFQ++9haf1O!~?YnC&KzAH&t#ox1KOew;ZOn1&DPi+Pnw(
zQ>utG^mQ0SGe@?iNzb|*k7w6!(02aJ9p=Ib_l*zqckUsOacY~Qg^ApeiinRqA^`ji
zz<R-vyNHi5M-HEdrTv-!%|=dh?XS9`m)OX8FMsQ9Tp`W5+seLU)1(45pW;Bm^FX7#
zm*#8oogta%AP>HfMi;eB;|D@3<~pI@-(ibzvmW{-$XJHQEo-c%6sX(x0URU$9}!2O
zeB>>A{ftZ{$%*cA&esS4ZHfyi1K#DSQ6tjs$)$`K=pr!iCeLVkh%Smq_x3yP|HE7|
zRR))Cydt*DvU)LZC@uf+H|G0W0#XZB9nw%UrY1PWk;=)fS*%JqakPU-TkqQ6ZvpJD
zeDgXZVLbm*5hoZ2U{2(|ko9mTQ=A{j2WA4|`H(Wx#w%oSZ<8g6qu0Q$kG^qF)281=
z!?EfFvC>bOE0qR0HierBU=2Jcz|<1f{SzsrMmS&t(!`qazL3J(>p}<Z$>J~3uex06
zMQ9(g)8)M4{cjhtCanxg(r|w2m?sDuWH@a@n%Lm+gQqi>2O;)5g@hluED-b0k+yDJ
z_o2}&-Pqzk)<$Zra9y5c+Lxq-Rix#5#F!%u(*8mCCK^v<VrG7pt}uczBIRm%Gt`}!
zl9uwUY%uSzmf@F_F*#=fH+UWZ*F|_1bYi#Sh*0h6#_VwsM!&6Tb+l9Qwdqu_a&I<^
z^55?Jf<FC*mSWy0{*@b;x`R~LT<EZ>pO~X%hr(kba0Aq*KSu*F_6Cieq8~L+G^eEG
zhxHaFtSHiRQ-cLT4D?$ldfuC;X5{jwq$oN@QK*c<-=5Rv-6aT}fNBXFo-CqDkbqEs
z(Qpx_dCJO+Y4b#V{9G3LaU@5@OvTd}bLFcSJ8~_lv2YSqzv~8a-Ce@dZq!9CU_KeS
z%o}-3c+5*s0RKqyMxYv@y{GR%AXX3w$O02u@;T)rXKDhyI18I7JulWC_p6jdYNg>(
zNq^RR##2OgAij+gs;VG(AlsW4$&B?_V4I&J0&CvTWx&W~fZUzJWq<<Y@uvjfDgmMZ
zHa3Aj);@iCXjsk5Vx^7rA<0+S5vZ7(p-l3)yTzM|??cfwxE-a@5u4pexF5E46WTCQ
zB)dJ2pp|B9JkJDQ>-;iP7#>j+7v|)4OXv=Mj&bHDjZF^D%=9IziuDN&LB*tw!#9cK
z+?0-Jtb^sD=%)cRN%rrQn0q7W^Eo#EW9Ud5-YbpM32S-=@@COmh`#AyqFGzRo@hOI
z;KK&8{d+FL0&)Ro)0XG>{M1sRu$s-@@WU$U2&`^2pjAL?1}NMCI=O=={xk)ZrVVUs
z`NiVPTBfhc8*Q!qq4eFCkhJtLr2UrW_~xV6Wl5p1LWmV@mB1ll>o?SAo*XR)*PRtw
z4m79z^Ow9)Q5F_9R&F<(G3D5HbTwH!N7^dE7j|z=TlaosL~aW>ibEJ?bh5)?)gn~?
zXwD>*l`TP()tb0df$QUE#tFJnnqi43QI;ofjfhVyi`T=sHl*U&1-SCRM;<0U9e|$r
zS+D}s`|s1*{&xFa3Kt|35kHa_-}zxS#40ft&xN6d?tb_!2xgXQo@7cCpL<-XNB|bx
z7$5AUxMu3quJM4IRqD8n&|y<i5h7A>_ar%1HCNdDFGB`AzIk%Vvd^rFUijZ@4R|jt
zqT!N1kS@okK`?bm(0sGeOLB0E*?Bd)wC~m~$JS%S+Tdo(ZbO+l3AfWJ76q3S$Q52d
zu+dmmk^9uDo!{@gkIhD8$F6NZ1vky<n|KwE>f^K-dfy0r`?=Ws^!I>~*K&1#M*YpV
zjNM{l7kD|kuxX7I<Ux9R*50NcDXRCA&a}w+0RhN?iq(L+=lORsRjdYzn`K$nkT!IG
zSxtX62nbYTJ-!IrXEvov=h?h~A2sVri(JR2jWaOz_Z0CE0|?Gf*9dVtgJE8Tj57Mo
zY*BDg0@Uv`)QOe{$1D9c8z4LT!+9PGf3ybO5bW^@Mzd0Y^Ci}xJ55g?n}rPun(rLf
z8?v>o#v1B&I&GDBbwuTU^Zu#__k=AkiG4lYELJjN`luu#f>y~=#j80P<6U92Y(cU#
ze|ZS2U|?-f)sXYNHC-3Vi*&>$*<V5GkACkaG!6N(S8$`>Gv5(8>m^dB*00rV1i}a#
z2nwq1UjPg!X{@jI*Xv@!-}3Lg6J?d<p>FOJQ;^)ilBOo_^?K`3EB#D>G*~(}YCA#X
z2??sdHy3qih=ww+7WvtrP(}ljLXO?vs%!a)p<oE1Q4A&OZpOa$k6=^W6{?Ad4cVu*
zrasHorM_8oqUdHbw8*MWXmq-io{Ql*#l3jaQX&I0w_Pau*6$tp*AML5C&CX0xGrvO
zbd9?#BUGgQ@12H9qccCndTgvfQOVWcdZW01E37vmy$n?f;uC4CZZgr-?L!OE$CDy9
z4QE|j3Q>_y5a1)*?{Dz$Z3<GbEO`<G8S~oxjkGymdBvw&?|5tDeVE}e#`xvQ@_T-l
z;YT#X>I{pC?<&pnRKF86@_G6RJ{`LSp`MGeGo}2}Mw7k8J~q6%?s6lx9^l}^GUOcL
zj%IyNYF(oJo8Pt_&{oW~ks^zgt|t+Lr&FO{e~tyneN!Q~iCn@>;M#$2pU<TB%r_h4
zg!Hm^z;f=a6SE{YW-8szu~H7AjgjsQKvseg@7LEorO^9Y;8jK>1Sn|lL*Cq&_fJZM
z!y$rvN_3wqq?KA>d%el5G<)HmzeM=G4-N=tIY~^{`9!g8$`@%xREfHGav6S9hCL&7
zgdj)uCQ?i`7x3a9_^5eJ#PdQ^=ilXWc!~Yw=ZzG`%pQMt0ZjzSNzCtC?SV8WPWrBO
zK2U`<+U&j*B5zj~sCR}&ZFtK?I^#f+Ziz8zgSW_U+z?Qs^k#2>%h5#Qc3)85B{Zs=
z`%I2^_Fgw&A*bmyDKiyu4p74s5*0tM#t5;R?S;XcU+wN4&d)sgflqzlt?|7z&gh0F
z1_x<p{D(n*W1CVtQw8QzyG5-s8ATkd%ijnXU*2=AWlUz$Of;;E{v>-oWY&QJ%$7^O
zI6wF~ux3KeHGVhx{q-O={y;=EmNS$U*Us1LT*kgxc&9`NmMGJBZr_H{St4p*JxS1w
z!j`zZ=n!E-gw0P#$*!V;wtq~l`9;FDK9BrepLAN_m>>_il5B&stBAkAzNRKF#fBI!
z;d)R3eZ>1(e4bUzB|rYPIQ;@XFJ_ZNxGX}(Ap>I?OI^*VdEE<(A1nO9|65b<`!xr_
zx3@kn1+4LHe$FsQNWAhF{XIutH;4p8ESzSpnQS5ZYJm6S6POjFBAVVpEsEgz?^ar(
zKZ|U6`<z(mFs?Tz@An9|`ZFYn(<ik6HBK98ly`>z_yCbYD4*YpoaTR5KHA@}BZ2Wt
z9A@3ib-6aG-<Ov67s&I|L;%UGn6niXvVZ-!IK`2{eewIm*&?QOsDZA7Ai`|A*+m&w
z&e?cH#LWyT(|W%1S?ag-AXIu2h7m2(#wsNHc)i`Nht;oo*hh*rl}yVXClW;OGORar
zLuJl+ju^?X`dE>AgWL$PiHTkswwBl0Dt}+{T+uBg8tD1ZG&X1Y76<TK>}ZqPzUqIq
z*!bbz%f8vIqXDw2o4V%3R_txXI39V@BP$sIn@_LNtyHuS)H%S%_WRlo<#xv~5f%RC
zVn$?VJ}-*D1lJW7Mv4}e+8yXCC9fO#N?clLKUjekw!8rH7``=Lm4x}>V`I}dHwm1H
zZXXzt*}l8lc{_aHa&cldDv9vViuv>nLf)teM(Wg_-Zxu7vATXpeiZt4QL7ElQ2>OM
zp3%?*gy<rMh_M1p`Ylh#?0ZaDH5~g9a;#)T9T6{R$2oY_OAx<>TzNqre&gFg@rlbe
zm;fuBy7Qq9&?eCrP;_ev<Rc<~+Q2$&)KLOw;c7924al_-v-gBMW`&^tg~?=sG*bbO
zkodD|=egjEvS+<a0P4`W&dSH6^eA4=yP6fGK-#gM(7UnbJ&f^4<f`|PAz+{J$(^Z@
z4`Uy7m=CYT4Vlv%Cw3E6X1FvocHZD^OqNO1<L72R$&r%VR24>fkuv5#GJKQ%srxKD
zK{&Vxx-a~n0to(pbQ5Ef85g&P{c1vdB=V<sU^(Eb+*zT89_XYMF?r6#ct4PDez@ja
zGaa}&o+}>cpo<^F+&qnWUsr5c6jQ*3D`?lJ5y%f&SG96bZ+PqD6^+Bj&xJ<r_8%q>
zS}?EK+rJ4eSo_bTO3`DHsChJk`fCDrts`7doT;9-U6pjP=3feiAAO%Y_iiKN^Ui=k
z*RiQJ<&=|-K+V1a)+q#_-Dey>Du}mfGqy#%YTXHtx=TltK7=#>yWZ?4tzOI3h=>jb
zQRv1Ex}KbP76(Bijc|8o`n8EZcV>^m>tc{6TF^bhdI$Vu@9cAIYB3>3z3`zplW7kA
zyMiB_g1(~p6=b<@qC2J2Z<L*()_Eqv-oL2&G|yb-#VO&s(VdQv^%7Z{T>i@d%KT$9
z79XU|M-8yS;3?%?E#4sDKt>Q*GK3>OSBrncuIhbc_A>c#xor>lpjbAVzv2?S`lZbk
z7KpqCDW)my*`Vl!-x5cMpm^;rApf>mxy@A)p2YWom~7n%82nK7S2v_`oev9{^#m|{
zftnTc={>;-ru0;TLgc>sNe{qm7Sc|GNOjJklEH(19ZL<pO0|&*G{4s(%Q@N3S4gz`
z57YLi)r4mWP;p|yI@mBdbWVTq3jB&8?gp>!;a>9MHU&yw%!{lN_^<ff2K>}LJWSAr
ze_(p%vKiPGQA1iVcLQDxlAGvLr4hO7iOiAlxe5X6{At|-tIVm9f8F$eIaoD$`^pPQ
zLO%qjh?O}t%t0*txjBaC1GQ+rH8NwqVXxMyjVv6DU`LoSp^LOt0UUR1n-a2tN;4~n
zy>#Vw?+1XsR7IZRka7{q8P&L%#r|rx9WX|mX#cHm0n4$!J6qn4;%z@RT9VE(*)Pbr
zZmqSmg=%to6;hS?yVXlv#9BY7w21sta@6#e)eAz?KN%kVw+8&r*Odn%Yl&_0|BdC;
zsAuT#bi}Z}(-=91kOuqbk7n(-H5RH{R(@<T?lnxk;BpnM01{KOvjfCXOm$T`kge`I
z#frcED(uaBt)*t2^wl;NDDOS=FVJtu1(#MUvcv(723Yam&+H~kxku}m@hjrezd(a-
zqtI*7+i(+@V+!Djat*k3^l!C9b_IU@?=<iJCo=sHH*7gp;#oMZX(5OH{#KZqYk|YK
z4+o9TZWyi}QMotmgZkYVAY^bzXHa!TnOV8NH=;z_LK}pOOJ=;$r|62?bkvO&7V6e~
z>%&C=<oZVN^uUbq&-y3Y4eON|_g0RGUSv9icCF76Q^t%QUIA8aiSYI#j<R>ARIL~w
zqzR|t!jE9P;Ucs*WEB&<(=F7x7`V7Qgi5+gxORNIb6xA#=KHf&|96}KP}76m(bhX6
zOWB>z9~o%!TG}Pha$^@|G>MQiynpqn*N{WAbJkZ0?z8^_NgrMD&^`aCvF(HAU3(4L
zAOF?^Ol53@jfyg%M~P5?=%ifo&i<`lJLj4JOw|esQ4*U8V0F14|DG!mwOw}yzPt*+
zbjEP<)>uA>+CIGlpRc#=+2<m>H7>TJw!7Yl2`v_>rq?l@nv|VCu=xe<MHgan<b}gY
zYuo~_X?!@RdR^eatWn?Y&5i#-Q=N=$QVB2&;EGD)L$z}|Va9^Oxf(d$Hx_WIZb<$3
z#V;_6&nthv3#XjvG^h+mT6xby%qB;X4S5;XnALk5L7EG&1c>5=SZ-0gc*b=?%@tHK
zVt#ppOl!vduAn*vX|8u+I2J9-EpV8Tn0jOiQY<)e0onU|hMH&Nq_?u5Q0=S+$u|Ig
ztGzRirS+l0$Dl);{D!FQ7GYRz;n`kb_CYi%5fvo=UfI9RU5nVH3^KZeeYX|0{TJws
zSX$@Hkwv_r$1ee_2&mZK_4v#Ug{WT8?o+DOuA@kX4N6z6X4HRRJPld~cqHQjcAzkE
zw14_b!1C4ISI2_FF_U1kyoALc-f>OSEd331aDv5O-c1BMG(`Kr5|T%$sCF~e;8g8@
z4z()%uieiy>bBWc7FqBJSYiH`gX3RMqNQLZ9=`(u(#=?p^l;ysbSZe~MzDl?QFx<!
z9L_KG7Xj8M0?G-|wRNH+hwp}#o9oeRnDYq0Q@0^#X6F0>N7;yQ4-2_eR9XH12lL(f
zJ|pOE38(VG;N^95fW%8>91rxyM6aNZcjV96%q0wI>=SvR>DJU+d`CDoh^gP?m<lW)
z{l%w81LQ4}_#hDaav+@I>(ky)^*UN-Fw$;bMB&mbE`{fcT%d!@0>*xtHjV+g?~Peh
z9i{zpQ1KD;U1DP;tXf-$aS@x;3ozX)^M`Y+dlzJ(ri~2PlXG`~`J~_TtU(I<EPqAS
zk&9Tv?b}5~Zd|K1)!4&&IlNflltkv?UM0_OX&l}^Ml`$M`*)@Z47A=B1beZ5B_fCK
zAqM=A0{kVRM!|!Q<W6n-cF~?38R&j!@xkdOwUM>8;K-^CYyUdfxFQ#!cDyjkEe$8G
zGgRFj%^3{|2WhCB`3M^h38#q*UK0rnrzs5!U%bAu&OYjsK%yXyGTm8&T+-<Jctw<7
znK2BmE#LNb*W!E6Mi#Fc0W8qFnG6441MQE^l!v;PMz2N^TT-NUGXbq+anjspVl_RW
z?J;}l6>+@B-UIi34r3j^j4VwS4w4z8ISp4hmNJ->vnK@j7E(Y<%!L(Ai$aCEJ60N}
zS0iKY2Jf;^L_>}1OQ7_Usppr2lyssO^mGe<azw<GGu7BTTsN%F<a9i!lRBE}Ydx-O
z%7#N6?8EWrN=P+*X?Vj)KWW^C+|lqI7H&YYU0@MBvY|nWUEpGo8=I&qsYTGvjh%-l
zEZu}hYtbtrfo#F+hohhto+S{JKWh?}&VJW}f7I4fe6D}BW>d}%Iz_aU#>R{OivB(O
zXWp4({i*o}TIE5%y8@mSa{rDbGJ`xzSgxjQR$;DCmq-#xv%tj>4W|TZM?fYxD{=o<
zz$8f<2DjgZj2*J4SY({~KV?x4@4w7ugv#mPus*Db3SZ-xgVkKgoJR&?5N3>sHGsTa
zct8G@o611U1W^Z>Z!r-XV7a(}Pe3RC%_M-|d)NBZVcLH)2B?_O*?-8RV^qJdl*(WG
z^mL;W)|M-mXZmzABKx$w*^$>)^6$UIMmdUb7s3YU^hv6>IQ|0JU6r+x%SD}vUcFh_
z%=17iD`)yw{r(fjAshu{{3ycb5k^3HgY6U9h1R7ua{q2B=wbco`gOYw7ZLpLnC<_~
zHo^ZiCjZ<(>*HJny6#Acs;Dj;H`tQy<=XSF;574+lWX4sVg;Zh>s^zCm8I-x%X`rh
z@6PSl+q!w70qrfjK~^W17d)n$dB|gtUnp_NmKpF;cu|oCi+KcGyRD-6&anLIfB@`2
z9_FTzHlm>?8)6A>V<a%tPVc|2;h#gvRKnj29fNQNy7mSQq~jUWM~*_@wXqxy1eJO(
z7S3jz9YhC?oa^tHHeLsa#r>wJ%OgQPTC=Hip4y;TkC|BX@(OK|K#oU~QEwo3-y-)z
zijyD~)wNl178H<K!Z=LFz^SZwK^%Q;KEH3mqnroZqkJ;ifR;pEPy7U)K(rP8bOUQk
zRQp~pkZ(?W(=-xJJ2H#Ef+p9vnx;%HEz3$;sp5YEeR5($TLMBt8u>O*D9tQRj_r?H
za(o=k_uiw6-QW~=+{7w&=MqQ-;CSE!20r$*p%Dh6SNM6KUvJFf|B=&&dyxPFIq2{x
zdMw`8P#^$nqF@-=x!E??M7#BRsJ4{HvlO3hYd3R-CfIi=q9U%L<EWcS>hWnrmhfET
zW<Fnt`XDO_R`X%Izfr@j=6p4UzzvFeW56Zmwc)!_xttG>@UgaTJcUO4&2V0K^)8(d
zc76NA_4JNL6;jB}XMuu*PqZ2FZGCJd<kxoOpE@92#*tqn(46;g$$Qg|st9py(8WY;
z6J#_3*s?3TSp|18P>+$CNycllt`4GGK#!4vhC*j#x4d@I8~nJdbAs(pNEh?j7M)&1
ze-IWFUQbLY?9bdU@i>7}J4D!vnUUIdbyU5B-})3J5xHS4+B%%A{=ReyJ{#z7-kC(q
z639p`qG3r0ShxK+`czPP*3riEk4gB4KLTSh?in&9GOjNh*;&SMuTI#ZiX`6Dc{4}4
z_s8WfUR#Kyw4q>pYe5aF6h(kha5ZcwpM5}z(HFLpAk@nGmNrBG#<>}={<#aa5X^pH
zJVL;}Wtrfu`#00X|L`tIIy%3*6X1b}8LtTyxEzdht$Mb)RZCx3D6h*ke)4<ZhGPET
zYByf>z!#E{E}FmV)ze+@_oQ)v;$m}~N6dm-X!sLj7#-pbet(}oY%;vzf3=Z(KD*jH
zgV)RMQ&4&oVS+hOci#~wi`c0|+P~b2?wngkQTs;|bDI7q#iaN~e-U9u|9}Y<mL6#q
z9fdv-xdJ&KQ9kJd-r2>MM!J-7jI2jqet>;n0<hw!K4W<M(ilxp^emcUx9yKCjiQ!x
zfX%>Tn}N%L45j&z#rNOV_sr<8_Gc)JR&=8RoH7tyF}1iT5>7`r6s@(mvM^7YL~vvE
z0(DS+WR^Dhd*P69UoKqPzd(_~)kBfxeHtzp@3D;q!WN<eALc;VD<N(DgITIBln`Ds
zh6w~@<-q|eEJYbf9F5}S&;ZN6U5JKmLMYbmYVmB`SBHBJP-)O3*I<hQeszeMJ}om^
zw`h?F6N@)QtFSc@{%<mR-@LC9DvL$rgsjsjE|^95F}po5uqp?<G&FS=0E+JrA-oEA
z*U_K_nK;8+XfL~lz*vu>Sg)OVsbcGC(<0V+b<>5|21(J>_@Vf<|B2`R^CrehGw7Fh
zI{-5nop2u^(q|H~UTXQsw=ez({{X)?X32=TI`~I_{8t6?znL2^{#YLZlrF+!7qOg5
z@1|$wuqx$eqPyYI?^P%|2&+Ilf%)eN69a>jD-ic;qs;;?cqfbD%17BM4^bUmQwQD6
zb^oM-C9bodWUR@X{gDNmeZoBzNx{^_<kCgGkzOI>2fnFk5vnv|L(W?i8v{Z)<;T5w
zZ-CDte%X)$uCbCb<;lD~a20~2tPMNz=UyWaIYu^X!1&|2jwd*MewKLmO;(}g1TBzH
z@!o{HVg#Y!%mq^`4+Z`^G+elOw(6T$9*QTxcbgJik&rDDu6Wt@kMG*%VOjWQJNXb+
zw+Yi*?;i_}NPvu92|#1>cis34WJ7d~<|if5SG=*MExX_8P9b*4`ThUm?k%I@Xq&ZB
z2n0xiO_1Ok+})i)gOlJ8+}#G(;7+h$0fGnjpuq+h+}#HE!TmXtXYY6KbKdoQ>wJIq
zS?fFhdiCn=`>yV~@2<M4>Z;p~D-GxNn|hA7M1d}P%i)oyArujD6^PU?GOPCrz%`Yy
zxhA)ja1~99OWmR)L2ACuxdGZOL9eg2o$s*xZy8#!$Q^DA_{Olr-2OH<GIu3*G@L9H
z3;0HZ_)rQ7JKTz+uath5`akp$cKGzH4KtkIPIa8gIA*_S4N;F-6BdGLY{RzK_Wm61
z<+8>5Jx7dO{5JTVA#x?-m_z2j-^@b~KcSM=A2HsmoPD8^)<@WmUv`3_z~#;2pA+TV
zA^h;9Rs@?d=!zx%^Q_aWQ*ILn=ktdWl0aZV440OZv%@WeLnl{w?t+=PoG<RO%k~;H
zL<wdDiQ{=-I={X1n~v^D3H37!gtfZL_6JwRs%T={Zr?N3`LjOyoV|`f@w<V~1z45e
zTD=#u8f|#@IkS3yO;{~+!gkFI{!%<wwW*b$rCM%QfK0G@(Gh=>Z(=4`-=~kyc&by{
zwjIB;S?IxY6E5K+jtj1pVj80ZCjB2}FzDhr7tJHs`Yb1bK;qR@%Q4uSlv!|`>c0mT
zOwqOyq?R@=_PfUb#KoW>Py^uBD31#}4WIG2=PLaaQ40L&9al1f4pK`unkEiC(cT4;
z;(dIwx*+SzZE3i?UtG1Z`91GTo5O;c*H9nX)@b~V$k~3OT>v%$`dRocLopw*B}%Ja
zV(`VY-gYL<ZV9XS62jkYai<FMO1SdF43Q%cuBXL*w+FGF`EwoTa3wu$n}AEb2bce#
zn03~%@8%@R0d=EzMT#$f#$dqhf_?(RRocvdvpqpXyFPvAL#_2k_c8f93<tpcE$(oQ
zXgx(XO^#&r?GTj%Q*(x4VEJwD#CmSd82m1kY28iAWsBloi_Vq4^HpoO;HTXsd)Qr%
zSrep#K-YX*3Y0FsYdP7d<;CQX<gUO~qWva{*8rxHR(d)@^cKOTo&xCvycfIXk2fp{
zR}a89G!p@LB|RZ1g|Es%!}*oS9kr+>@|~}5VlJ23+Bf(djNB}T*WqJw?KoXLzz0!=
ztLX*5*HiK~?EC?)!%~MwaGgxzUK9NLw#vK7?A>dnQ63rJQ8Md4sxy(ft8kOH!pB}E
zR{Q6wW$2pM9W`9}MgP2Op1UmJ=fVHHj`HeSB@GNYiT!)Ez-Q61HwPc~?{hKm3&$l&
z_hSEbCztOYj_d1-2WlxoVF4!i8W>M<R{GsXWUOAfC33U+A$DFG0dK&YRj*jGHgJn5
zF5CBNcDX$^!8YPWJ%JM|yFUyA&yaZA_T|m(&JyXaw5AQ*AVM4i=DiYbklA`8f_U{t
z1_LYHtAcslU-R}k@9@R}))xX8p?uj)B1qyoHRwQl^rz=mf#%;P!oTm=EE#HlA!KY`
zJoPh*os(Pu{IUG=(r%Z2{Dtt)@um|7sx)C+EB|LK?%yBvKfCoun`O0|#y>2+!(@BH
z=x5s$B@1#7^*3}pYT!KGBF$clTKP%G47%xqmE*3Al*gmu**TfKWB^tmu8WMMH2QvQ
z%-ICnsqu#E-EKfv=Y_tTJoC_LZER2AzF7)&-xR%$Bkkn}c?&YyxAv9yeJ2^)i{kAG
ztQoC#mtMkrF>~Af1iOJB`N#W^Y#ARs4;0(4iVB13$k?_o9y~{Ik5QU1uN`gCRR$;I
zun<qkM6bgr(?uz9mqD?=7CN{^Zu9j)@jiTDT*Y|6sm1O1&1O^+CgtkLF(q9mu3G&m
ze$~74yZj(uK7{zZIM)>t$OBbL?*V2fVwX15Fyv|))c7+>xQfr$qdNDh4YgVNyln-x
zxIp(5pdudS0Z!91{grn8F)loR6e|K{!{KDUq}cwXisWQrHQY#3t?`CjiZ5?Y`RQ>&
zn|w?}fSX^P5k`Hq0t4C?gBlG*QNLgns_9hT4a31YVo3@RuFT~B`dcz~7RQ1-L6eU`
zpxh75h-vX`pP2amS-1Y}Aw+L~Pe?6ddcck+T3*0Vt1vJ7aGT8jQ7ixLPJ*cTh%YTG
z)tIpN>jZ&2IY_u#_Wxm1{<G`&FEb{hR}*AM+$9@s>Pnv8QbNclT&{ls^w@(NN%>S<
zj1D0QGP940LJ2ULejX-}DmB1*NAfK9d1n2R1ApBCia6lSMoRS01KAD5KaRc2J>h^a
zpM>4?=rf-y^v?bjIyl$t8N7-xy3ZBh|D^q&cKQFeE9Kt?)kD8JB*D6=XFY5Wz<1#I
z3*ntN)~{pKvv~54-*g^Bz(!T$3!C*FAysI4LLsl2?9Whko8?{HfD+O68w<Ie^CK&W
z!t%yjW9!jxCPH8l@#QiOFuG6*M8m-RUXJ6=X>vt_RgAtt5eCBrJmF?%GT055+kw~+
zKsF8I?&cvK9N~cd#C{L=a7ehD!>yhW=G@NsO;Ag&feF;^H>fp5dv=02>H@F}pd&L$
zJ4HW5K!h8=*DV(Jr}I2W=bm{$`d<i%c#S<Etg7X1Wwo9ykqP1NuCS#6BCWeTQW%Mf
zdX#Wsllh!OQG6;)INX2%QV7sA31@w{{}!n1Y@RF@$s+xcVEG|rEtwDWHlQTFP|cKJ
zqt^#Kk5REgz<1fnm^EQSuulABxI7Pzk5Ssj5G^U+CFyAs8A(0$5JR6($wDR9Y9k_0
zF<~>Q5F+=#)I#eL77CwdiF<GdC7_CI|3nKKs}pneWcv;iGpYRvU}=*|xW%?hkblPZ
zwVjQq1fWJ>#eK3PV^1^`EqqMWC&#$+)@dy?s){&>I{vERjDf7%vOD?*tQE`g7Y(cD
z=NEQ*Qech&03gybdsHtQMBCRR7XRD1`fu%)q2e|9UB``9BYcMO6Jt084NRo=0Kc_j
z;s1Wk-_F-REtFkCi137ouWKW`14pljeHH>Wxs1O+^LjLs$3KtuZ%_K6O(s_(R^kpX
zKYZI4$+Er1tY+@;A866Pz^DI(=5t3Q4dU}>w)(|Jut(f{V~(Bp*My^0xE|$Cj}5PR
zO}apL!O>29N$e!y`LKzyf8P`SR`dU3KqBxE-A^a-)w73Z&w%pwX4ywSq5c@z_%A{|
zk4PE!@5}E{3+mP~g|05Hl^-_d@Gd23LMSnp9B;fA8T0nq9%?Zs9_`6Hjbl01qZJ;V
zpbOAhUYlrzOR$KDK`NFa;4{J9;r_-nQ4Jo7$Q8MG4GzQ=b7KUqdpB!Shj6H+HKG>T
zzvc+*a|pl)azS>J9<@RC@bFW#pQ>8sjPk_FD{<;@1)!3;L)rs{)gKn$g0%W`2ELnu
z5N>W7Jo?VT9@&E4RNuk7`%T}8@_R4+{e}Dl;E~5S6Q|(0!)gwE%J{KKgkH7nmcI~U
zdes484|lr(;x%~Pd_h>QGMlpTt@-zq<aM2E<(i++9;NMP`R;w+s9?gzpG#_63HaB*
z*w^)R%q!=j(w+E7Ni0$|Zes>LN90<iwc-NCSz_8s);YoxmL3$&O_Zx-LV7I4JKJgD
zOUepy50G!eSYeOtEMN0BMms{<KL(1s;Zx&%21zOJ(AOfZ3vR&!rmeRAENaA(BC^sl
zt<W~Vwh6)xZu*$D@OaUKh&g|<KZjA&e7_$i0F{@5aSy)%MUU`vSh>8#D{nWz;2;!7
z^dM1`_PIRZJNQp~>QB+DZ+Mm69|-IzRzK@o`nKWax(*&MIFOq?MxG`d(|@mId6U%u
zzu?o696>(!T6nIn8nGBx_%~{FZ_JWd&;Bof;^A@6HC!3Tv(pjyeBh>PKF=B)=Kas9
z5xqNY1N@m~{svgdN2gOlau*97bVmjE6fG`wBz=9zN-`yS3E>o2S4ZCYu}z?R)a9v*
z{SB2Fyg<ivZjRnJs&K1BpZ~PZ)|+_)sysx~6Zfsocj|onrZ|bgy0g3Oc_6JLh9jFe
z^C?kg9Oca7(h_kS4$Q6Whzcyk5^g|`uWo>90ZyK`Q~j}g(v+iu9?t_cjq>kXfF%yS
zL6Oo{rbzLB3l&Ul0Lyceag2R)le+*HYhx#3T869G$V?{lQTTb_Z()m`*sId~E~w#V
zzqIWiS$RW0pY24<{4USu`G0OT*7&GN%#KS<{cTI>bZ+~n3RD9t{{?0LFLsl^w<SC9
zqsTch_Ni+=94($>>q%J6%%S98H)zuaM08Gz=Go<LhAT|PwXa8?HvRu(Utv_MX%6s_
zyo`n?{na@D;3PnssB?!hV4g*!JJ}aLrz^&(0Xtv7P9Cn1dEyhIh5X6cJ453+=<;|L
z$TN;?pdJSQ4c|7Ya_`^tWRz?uxv(I??ryJ8MS%6u%9lQ;*Guy=5xv-Pm@j99t2j<z
z4x&F7sk|Iam(hV_TInZqRZaU75kCQ5H-Vs2$UM*~Lbz3o7@(3uj<f)6iTHV>hT*qy
zirG&!xE+#K+<$g?t#IItHHC_nqZb%nTpl0mX2le}9YL~$<xm;9?bW(=+2BA2S>pak
z(Do1lu3Dam`zWj=P`w8S9lI~}TrMMS&sctFhtZ+m-TT=Ajp*U|{x#U-Wbpny(wN3G
zYtT-XUM;jNvSA+(47O)=y#8!(cx5u}WiO@d@#tTrsQJl*r|E(3bY;F*XC-#5%IPNL
zbMPHh@}VaKwJ2A8Mu50KSNz<LKn{#qu<Q&hbK$u;+g-3<$v)ZxZEW6OOsnx%DAxGe
zU=-2C3rY)@UzIVR^kP<i78R)oCfJ2amh8y-ibM@##cP9BC)D9Y4kZKe{$n%=WQaG8
zCb8)~W5rELnM*R2zT7;oIUARj{San0o{cyo>xF$mSO!05)(bHOo%wGHI?&*NA$}F=
zdCsF6Z&(*uE*ZDgS!+$DW-;p(P{?$1b^VHC9qvW=sqywN-_}1?J=0F+>}8rlU5y4@
zsdECvqgtil8<))ojF9sq>k%9)P&8lOi-aq3-TaCkP`Hm8;8|nip2}X3_^G3)b%OjX
zr{Vx0v()B?bYcZFM?1ipri!iJ89%i_F}UB{%QPMX+nU9ip!VkGGufl!56$k4Qm#!!
zB@pk4_QQ9$n|ssypXK&gWRFkw!N$)T`+I7<tI$2GKQ$rWX#bXSef;X-AwRojYwA2e
zD@9EPeiFkF1IwQV_c*o%c2DH5HB$f30wi{HwmGx%^HZQ`qwd^4H0@@G7g#YX&^N>I
z9rw}D%Hb&7v1dNN)PY~(f6Uef#$!yf%;!&^=>WULYAN9eU^=HhiSh`E;9iZsQ}NZo
zUFNS}qXwv8NBTW{bCkB;LN$l)i{U<qF*YD<1)81+p4F(Y=RDHQ&Y18`FaS<7$Vwss
zCob6Xn_wl8Y9nkX%K_rnbM!&V=>J?|3Vt#G@P@?4zgVEy-LZ+RC-DW{)a$Mg_6m|c
z54(P94ud&tp`Yuhj)T#<+jk-s)`&g=GVeo`D&~$l9Ve%yIaY+b&cKjC9NR0yoc_$U
ziB&-=!@JdIVUPM%su)!NH5)3f03yt*xk12xcw|bwC=ZK{!NeX}&jW!}gw?$meAXk?
zJy_D}SZJ<?u#wEJEm`WwBJTC5uD=klPhiJ(sROxy<D;l0_#0tI3gn0NwHD;pk*x3J
z=uq+Ydc#?JswI8{{@b3aAtTs{#@Wceedq`q_$ky9b$365|LNf0+=zen(=5a`MA51{
zuchWE$=Ayp)GX}Lkv<RnEBJ{2DtD9Z3abPj5YK}FZ+h4fH&^AlT*#t>@~9<a|JiB$
zKL?PAh4Vc5ms(Np2dqFnUwz3ZkrAZ}`{5aH1t$zE*vATo_pNKUlw7>UvMjBZBTNL7
z2}I!!rlyv#-fFY?UP4z6E}Q5#DF;|lFojN0Ue{IfN|b$MT)-ol%*ml=7#`$Y&`)DY
zR2@wskWo)aAo(EomQ`Uhk$~WLuv{gZ%c~Z%9}km7lTHQbEulXMs;~3?V{3sqBX~Kl
z@DD}Wlxr630bt88H|zm8XBWpvtrE`mkeDEZFzUk;LX@VO<at5di`Ovb<Xv6LY}`YV
z2f`Ga2A%a_H+2)PLb+^`7LOA2$YkTU_M97Gz9pvY%3)}|T;{(CC>>F=eWmD77I*NW
zE1fNEN*g5y@4kZBeSO<!Mbb{~6Av4+!k#<)D$t$6F<*^qU1}o5>0Xj>L9c_{&#>Ty
zD)GgsKF>jMfXLI11q3>TkHGns?bHWLyL!NzrN$L+?z@D#PSMvVWTe;2GvefcrqQ)3
z7WeID)DXT63Qa8pDA`5NC$)%yOMO>un<3f|jgZ9(4Rnjh3HiYvml=668S;6{!we}3
zU$ZIXUjvM}vtXfJBe}Ww3ZjpMmK#GoagWPWXXX?#u~^rQ9vkf*W%sqqe5C1mY(i$`
zW()YPydPrkJl^2tpEdMdSb|x#?>;Fq-^JpJ_ca>_0s+{${hg<WRTd7JN;iJG0j90Y
z$;?T`iyvC*q;Xhg2l3x9)4pQ=knV3^IYXrpffm|N71M+E-o^Gr4Bdkq5fMvv>e3nI
z(HT2(@uyr-0bsq;82)Dkg^4oZj3Kz<m;5|U^*JH|fzS|^as$b(!zj2;>y~h}*J2*U
zYEgq1SD`1@7jps~Jlp;8OZn-}CSupDZC97o{$hk&AndD)ZD{-gez66+4RqNbJW)J3
za_bK7Ad%BC%m=QWB_l&Q7Byp)py2b{?)xO@8QgW->#=zej|#vhT`|2h6AG$)6uef8
zuS#XY?(?CnTj?T;eX5WaU(auxo%O!Gy4{+=4Q!)M0gHM-D`S(fU&ib5#YNThMGYfd
z-^e`YT0*yauyMhuNWFfzs@GF#@3M)_?fGr}Zpew#F*tw20$%L>Fj))moXT~TGwFKg
zCy&~xpWNo3Ug~5bdNb_2TLl4=_Vh&CclF_a_GT)mtG>F`_S!<=8=EVtNXAAtdfS6l
z&Ip#QjfRmRyvy<$C%q=NZF4(nY6+%Set8bPRI@RbnGz)vux#VRH|@z0j_?lq?u$Z@
zKPT|6D!Z2SDdfmzW&aC#Pal5XeA|PR;Z7(&ByleDrVQ?Dc&(LFo~$kp2=aZpzb|`Z
zELqoz%NJieLVUQDa~uD?h!o$TntTGwevwAm5qk2sKkr#nPWn2^w^++Z7|SnIV_Se5
zNNn0gG$l266xk?@_!ZjS&MAGKtu<$`%C}OEK{e652k%^`w(S)j5NF8Eht}e0#G0O8
z(zhcL1nrO2j*+I06xD`1!lro37h~z+#k`d+6fOXF#6<BvOoN*%cBG$v#MAjzaR1B$
zKh$P7D8@SXAWFCz<$1{&mJ$>rme%#w4^@k8C~~tH?)kN1ZQ9jMBcQ~I{MH()T95<o
zkmbSVdzQ3TxO*!6tS3@}DKL*(Fcw)5#8TN_JO{p{xOGK@6$d&}|GsBt@)CbAzHDTA
z4AGZ$xO*<i(YTV@zgj#_jiwjhzhc>nTGhCc*t2{<9l>k+eop4kX)W6owa#>y>kCl%
zx;+PXAB*uJmk*&5kwfrico{ps@iTb7ReV11c88zDhGY#(%D2TnLR-b&nUZgl1!yRv
zl{cP}>uJT9i0{L5Z7EPO*W4H4FWz3-31$g0uHL)Nm=esc?O{21agj2VMnH#B)MV4J
z;61!)3dpz5qknnq<>>RTr5gUWZSBIkz|!@i8UTd!^{lPV#MOYX{R;|wsB#%eIt?y4
z;;^v)La-#qR?N?3JcU<sRuCLmcU_RK^GKgp-eSYb3I0Ohmg=V4Ubq30X$tr^6M(h<
zLKyP9MOL9l`i>efJ{0r8G+lJBc*y1Q3BC9tN1EIPpgix0c=eEOvzDcy>tB+iDsV4y
z4DSTT)fk!u$`6V_yo^ANC*(G9LN4oxcCoy0B&k>1mP$Je%FnC!cgCt60NCHN$&RlK
zQ*6~14?UMpT=Mb-6g+KxB*->}%6z$C@6ku0IFzl5V!RWdT7gEBK-=8Dv+RjHCc<CT
zBrg=&8G<O-m!neb%MkKah5iMh<%+{!2zjAdn<8&6zh7MZ5yZ0ro<{>9YNI=W>noe0
zYyj*ITXYTXh6g!K3wzX7#L$7siM;chQl7L5_CBN1OTYrER&e1y0u&r*1-|nZTkD|z
z31I8;Zf9Zd#g1AswtrH>E3$-KlG&{DnP9Kklp<oUiX6-O9pn#Stx;mmvh$nf-IG_t
zfF3XAsdzWyspmS2(EbbIA?8@W$JI7^poBFOQzeL^s>RXQ{Yl#TH*D!2U?gHVcKPZI
z82KbyHII>d2J(XgT{VHD>Hp<IYMXx95GDzqW{H3hsBjy(abtdWq*q1!?Zor`K(pbW
zy~zJ_5{g;*G`H-?SI5JZoskTQZ_WGKtk-%4oToQ>uX<><hRBmM1T;{}H&FyZRmm?8
z5xjX;(#nlI(j!vAB<WUIAwtwnY)e(ajdDV&>P4lfDbdnm-5d1l4Ey_MMJe0<)?!Q|
z^{Q@f>vAkE=f2&2-96(rVm9L9VY@(~mv))2;;$?>%k<@Cdb#l+HXMDeI5}$X!LT$5
z4Juyf9h4CnIld+O%sw1O5adMurSbG#Y^r-SoDekpItZC+&Qk^7)+~D=;e!{~Fb*;b
z%gR9928b$(NGPt*%{RMzi~-Leuuh*@>muS&b~aQ^@Qd5`EOq)0)*rjkit<}K=(~ml
ziKOtBi?Y|ottk4YGVyo=*!@hSP1F9|+Lh%m?`p&LejV|BC|~>RA$s4S!@5rk{JK=7
z_ueB@f1bt>i2K6KIV_AG%HB;G4ywz|^rdQA_53nYUv$}K-}D1g<#64&Tc>M36<}5W
z25-R*--kJUhrk%{*i6ZTEeTIyk!J7MyTA<WjQ$m)6Lh<y-B*^-(W=Q~%cS4!@(SHE
zL31{)PcZ}U!*>(q)DL~$HH@nX>5Qr$-A#w21URostG~We{E&zvL*9Hm_CyL#lxHEP
z>k*3RJKMQkh;Z=RJ3jmEz4jo>dR41S*|DBt9SE%ZgtnvDIA2NJJa0an?Y59tzet^I
z+n`8JGH?Ch<s)}%@X)i#bBAk$U1?Cjm%!_TE_2kvN|<I|`Qb3%@@V~cJ#VQK(4=9_
zGq4Gao;f+i38B(|;7PSv1^H@kCUq)pprpOsBsEM*)!oRiPyd*5IW_*3iQ(%I4)%ol
zB;)%dt_Y3LIy|ec@1?xa!L4e8-}-3i1T^XOQ#ggZ4ROI36|MfnM<Q!5DJy#)fT!VX
zbfzP(0;EohTTP{IrFxZbw#)xK+oWx0Fym$Q(Msx)=!_50U59RAvr+e_#~b9qb2|%b
zq6NtACd$0-Gs!Z(NOhAz65B}lVyiiv(^3jKT4{&%Fw=p?K;6E{x?Uk_h&O7^EX};t
zxSb`51$%*cw$-muKXP_0nP571lc;e+0%_9m!z-Ji$|{j?JQPBOz1WZCMV!$ulMDlJ
zYR}EVlL|dBmLMvB+yj3z<%o5aR53O-viA%~1X!v4p+-o2GCg)xy{*u*s{p-glaqxK
zkDa5XOpE!W9f(}6$2^;W5!UXa#lo&ZDt!fx<$Py7z8BSqO@?K?BoSfqQ9oh9PG7mh
z(lK9;PdW4pnA(bE3HTJ=bo4RP+;OF_UczfNmHzsCgo^OoQOHi+IK}>{$0m#AlNH@S
z%wtRK4_$PZ{9w%v|I(77l%Urkp*+^2v`p?!43*6ZMYoPWt;!ZF!m6o6{eMD;x>Aia
z^7m@vWQuY2rD_MloCcDo)D(7K6-cq<5>%TR#q;g!iTS|kG5a3MHyf73Hcr5HS;q<U
zdWVi4GB*pv9Wx%fN4$QSxn_4aEDyX)X8U|c<wJc|oPCW8j0R;N8wrZ#f83f4-Z1r_
zbUqUrwjzQQmSW2mz4X)?`3(Dyd6pWqLNmJut+V}y8ZzEhK<KW$Tv-Lu{A#5#Nu#W5
zWldt>ee^6wErub8#D*MzXgm8UeuCtZ`r3#o>P6MdAW~6|uurX-s*I%^(wC(IFIK%Z
zCj}9)n1rYWJ@S(k<rH$sv&Oo-_19|+dbBVEfL=#&{l09s^7U`8-)<~olr0jYOaiUy
z*Cy$79pL@hj)9fZgV(K%BUUxf7j5}<eHck#!5OhFZ8ODyL{a0fW;=Zri|w_y@i5)Z
z+{UyOa2>Ij{X&FYqivE|G|vvpr{jz|-40+9ESqWyZ^iR@0!}pA!s|tAU+dc~>Xu<7
z_(m2VdY!b1$Tra}Py4Hvd3ht7S7{RW$$&DEie49Qrci5WMJR#c<JdG!o^XXkQbZ1?
zhul4hwf2`f)H;_MO8T~I^kMrdvs;#cqc4ZB`w3D=4d>A;)BRine)~5Ly<;uk<LqjT
zubp(tt?ySF-$*|o@3C-p*hln<3cz%>f#`=E!FGuLSpCFlww0YhoS!>+n-`C7ctR}f
zd_KJYvk*j{ykzTd@N(YFx^LLm$v4UI>MW&hpr<o$y8Xs27tf@xH_V<aGte%D9B3Vk
zzQ?qMvuS@<eRZQ`ogz}qO4A0};NT%32mGKh$$GVx{mx5`Cl+t=jL~yA+jERIG$5Dl
zVa~FOA=UBddQj(DU9&(~muBI#eD(&v4%g*bp2IMiI!<d4dl*51Zs%1Mp-=)_hd!VA
zlgDrCu&2dn<*5(f%!q|OGRqErZPr8}k7XlM9VXKf!h7A*Zzoy4Fn|0`VZV3N;((1T
zH@-p=wA+PR9&c2zbPJG1*lZ8-cefs4=w}t+h@CY3fXjc?OG|#1E#)jmn<Z1^wD%)b
zP7CLw^9t4v)Br?<Pe4T7e22`5USR={EKP|5NLzzYzubrg9k<_U`@6h&){?)xndA-?
z!Was{3l8sGPi+oFz<@)rSIzZ6{L5Yfy#69O+m?6JJ0%AqI-0~Pn~d+|x?Qn33X74e
zE2~Gw8TB+YzAid@QBu4;r`t*5n0$NSl%5eP>QolQ?zh$LEBRg+jC$|V10pc?M}DEG
z_4b*N|BXp-<ya+mWO=1KvU$|}6Pxz0ULLlpCz?S8>WJ6DaX~WoEY*vRN^}!U1#bs+
z1GHzFXH|-aQ(RokUC*A`STGs7vuY*1ok-HybLCaidcL7;C#%QxPDc>ww{JwgZzq|}
zD(&+6*()CzhKFGDyC3vqs}Zj6ET=cABunG{(Jba(p@zWt&CB$F?o?K~;wj5*hL&wZ
zIz|{r#5bA=*w$2nD5dfHD3eSUXsZMo#aD&g$YA?ziM(Z5+)b}(1ikYZs>;ur7;dhK
zCa-%EMwKLBl(6y`6cYU^iAl69Gich6(?W-!&bX+ja9^qkjbSoyn1lo#_LO_h#}A#|
z2li5|nT0$6asATd0=H|kXVndRRpUam@1^)<@UL%TO>e;a-QV`BsyH$mLKJ0}g<FwQ
zJidhl+2=%c_mcO#ANZqYoZ5MN!vu!}*|F1??+i+2T68A9#J;|b{8^U}M@th0TM2%x
zql(UuqWXzZF)v+8kpZt1SL&2;vH^YP)m_m0Csx&*o*26_VcuMh??{~V+WuCCM4I?r
zSt}WtIOQy_v92Qe+uXg=CeAGKuZS{CcpZtKXto_T4$6s?*CD6``tq9F8%d-RoQ7$Y
zpx}NN<Ed}_samO>bO@UJjMm}JTn(Jc;<UM;%I$`<Man;Dl;wsV)A$Qdx}8^Pa}TBv
zd>L(CO1)$30lZ%y9+B)DwU3T<veT;QkRQj!WJ5q{L5MS0JzS}L(INJ+aJT+aAv&7b
z&;$wV$5n-=6#XckdVSpM!9^f;R&eq3P_m9jKa<J)7+dU+$_@^_Thix_Paik8_PvZ-
z2DcuUE-X3BC?r#Yu6=mEZO)U25i>M;#HO`0*4g?~82U&(lpJ%H8#Hb-JBVa7Kw9Z~
zFviRqwQ2ND^+s3tNA9i%8#C~2nKPW^>mb>jW&_VW_Kj=A?l2^qM&6P7ca6hGC*AYB
z0XILikOz5*5AHj7qXu~|>Oy=z)Gb=fpFl9eG&whuO~Xd4>vfBUqs;Re*cF2@vGa?Y
zbIa0^dSMRajFM4BpEM1r%bvIj${T{)^8E~pi>!tcfNa}}_`7F0R8d6&#fey?a`Z~R
z86UWGHV2zs->N%pn=1MMZ!G8ER6)u5bQ|X^nT14Z8&~2MS}&|lm?}>j2U?l=W;W_v
zEgaxGs4VB%(;Dj&zI9Vdl&15d(ziF#u?sagtYTdYEf4Mvw^|Le_L&{WbJvsF9<A1u
zK4I5w$BjC3DP=n~_H>XW^HttRm!HmhBmEYTYTaY($QM%&!t?pTSYuf4*CSOF^-H4M
z$cGw*UzgJt6cA=@9FpwfB!iVxyLb~_7@u%(hvO{7SCQ+jRNkSli0CWP`D5|Ml1vD0
zSwd)k1iag9@2Rg|{=`xgV%dQ+Ha}lw?6J@Y*^qu4VEnW}nre2-s<){b-7WMnYTOKK
zkuurWfLBppO@ZoTJ_q`q$<ta`Q^VlH;(`m<-&a1ZKI}AVLCpO5{ctv3B{2zp##?B+
zLre0{8l4Ik^U?TgP?3RWcG(y@&Z1i)Ad!UZ>(Roxn`gw!n+r%o2NoK#$<cNk36p|)
zi%rL98%D1kv$57_hSU+6y1tR7KDTk-bG(g~Hsje;u^AmBuIKukcYS;&TmHp?wN!6A
zgI>i(HlRymM#pS0saEa_ks6gE5waEg%Ml3qN*^Gsb-ab<z;mJeiRO$pv<@YiJuvDM
zg<iDnV1Ia@+x&AU)FXQR@a=R|mI;fOhAk_px8vR@uG5pI?PmS@l|p`@&6Njviq)?G
z!_@sIxfypv`p-5UoGQ}{caaHvs#G5=HQ<d{KB9<sXi1{RAfcHeP-gc64=~gH)o*K@
zbf~NlBO^hSQ9t6(y5&9j1uP)%%ojPohOtx&O}0+dD5UUT_R6pha$H9D{dS;Jo%v=?
z&7zEko9<Vjjz=S=MvuX%NEbD*XX0V5V9ct;n@yj|5y?UMYH|TuH&}oe&+vi0hm}7M
z0obWNA3k8#`>ow|!ZnPG7`3Z=`1G01TN9<=g4-y+#)%~279kE&g@$m_aD>>S=zU>_
zj!HA}JAe2{(LEdyc6n-3Hgx3RxzX$n8;Ph3b?|k^BsCPl$a=KtgRQ$2X*=aJSN3=O
zg`ji`T64ILSwPzO<y;ga8xm6zt$O6(f8u>{iREj@&aWDbg#Zm}iL=u}TYl(So^!L3
z34KR$bq)+*BnkCiP%}7iCOQoFea(cdEJ-SL`wjZ|!|g^IEh|*8okXDF4E5KRns{%M
zC7_|1Bi0G)_ftnn8HZGa_mklc$8b)dpJ+S4l;C;|9O&t2>UtP20?(a!JR#w_Uh*NZ
z5~S+o!e2YMLU^O8Nf@+Dy$`R(?$DQZ5EbFu=-j7@XPJY?2q}rX!kcxVH>bhOc?6*N
z&uil>$!G&uD$+VH_<|6x`8b*2dGqYWpBOu_jNAiQ#vUz>^1U1apxjX9{cE68^sv+n
zw5IM(CI_Wm-#fn<xbuhi&*D1VXc(nAUVpy}j^&WF-rBZ;c`L0ojP}XyETct%I78v^
zV&J^uomsZ?kUd40`}}xDl6HTQjO$PQIQU+Cxu7s0nO?f*8$NdqGEAW=I6&3Jae~rX
z{iljb)6|vZHul$IHJpLN{~jK7w0FHm_5W#{?RC96P=6{*>L-X=wAsJAt~ZgW0l9z&
zT+68v;>X$?4^gO^Wt)tO$0?6Yn1jf3fB5BH*5Hf^+&<X?RI*$#&u+Nd)Z}CVQws|%
z1+E8v6LK<SeqH^W-ZyK=pU!_nnsxKNK5s8G!>88QPIP@+r-HM6y!7{b#6TmQi)5(`
zO4NeHW1nKVTDrj<C3M*XiH{GT>%9C;S4DUB^5a>BeX34PMd{18Pg~(7`jc6i3wSaj
zO{d#X72<h#eS9H6_U~zMgp(Doz$XST7>3u3wp_t6Ey%@{6XI4siFdBsC$v>$kzohA
zRMCbDpc6bsx~o&jg%rNcZ@qh+2mB1r^FYeARWM`9g-HsGdtbrNj6MY&t<Th6#&k5}
z{w;_$Si8+rMSp1oU(3|>a#>_!eu08ikU%A5Jbzl??RE9MpFjVb&tdl=5jQ)+Dce(H
zbFA7GQdJ07w-foz8*!#fOS5IeIy$7@3usp^w(%nF5BC?q#C&R1xgKtr<2KTq_WP}f
zzw6}SQGdFg8GO29{{Y!ImV}3Jm3RFe7=!qq{^a=mzhg@O6~0<HKnb}O#wueT$qU?G
z&H1nLIea)M*t-hrt0ANQY8cURW`(hhj!MzyszCkE(JNNW-ff_YIy#@pi}G?>%Aq2L
zLcB7?K4eg0UY7@<pgcCn+xX7SXlU7QUxy%0>RlBmUT_boo8krVw(LIqy~NfR{(Yj7
z3v6Bug@#7L#Ft#WKIhOOf1n=LS9ab|Ueu8-MiUd7Yye0YP6CQFik9dPOaI@J^nW;o
zSKqGeSy3qWei7?VO|bcPTv%rfys>Je-qs@ads7SSB#po<$~MxRW?HVXS3kTcM6#!%
z#Hm|?W!HAtO_{J?7&4I#BR*L6L{?2#xy>S(w^{r(K}!?^1S(TE^^FA|rVw`DyM&dd
z7Dp0QXcrHpDSRyJoy;tWRdp>9FjqgCeO21B;PztCL%HM&5A)+YGl=nrvZAf^LE($!
zq^T*Yqp6OI`Xx(;#d4Y%s`Fnb+V<~&OYST9wxbY>21pxKt3xEpm|2~22I()ih>bA(
zDE=?s(DGx1IV8~7$>kR+qN|T8Dlh9-no_Rta&Fojg;u9LbSrqR=iK<R53TSs4Jto0
z+o@N)-2oD;g!)gBU_)ruqj+O{zd48e()vJQJoLg(O0~#%BQ)+u@i>*8T~0cV&g%jc
zbfnW*C~Zmt62YYMP!}j;(@CW#EfUMdpD@CVyuj}f*m*kDIHC96KsCFqNk8q&#X0kA
zy6DIJjW|0GuFq?Y!*|%jTszU4T=25WyI{7Hm~VW$`hOu{vNFX#;60F*UB^GX`rP4g
zn5E0YYxV5+?J@X6t7@AkvrC7sx^$!r1UG)cUnl%wg8wa+hW>BqAt`_xQTF7}mk4kB
z52_&hpXX9KfKh5%mG9YQhZ2(gRA{gdeo86;RpX520={e}R(GNI?mO~#3>ntD(03yr
zl>FXm^*)HSnt1fD)oUpa9K*YN1lW&`8lr5pueXVrJmnlLw9FZuD)K}P`Yf>Q2pG|h
zQ3)jKnI^v*tM4HxKprp-*`0Xg$TeQ_i>uM+>zG~k#2>R`9bXwY^3;U1>6+FX>d9KP
zb$TqZq(X9*JWqxJi*!<~Z_rDRE$d{F_e@f7HG~?oe?OCHObY~r-wr0f)aCu@OUL&y
zZr0Lu2ya_KElUX5a#rCg2693)#`MDGlQO7zRBfNYOvA-R4b42-co5S4<w8-VH@iN&
zk&kReu5G>A$cF>(tKD#U2ZRdZw#+Mpt0rsl>tv~nz@%pA*G4e?mQu}dupMtH4lSfH
zQ#_L^U~kqa?19>H!FI`XXXV2F=F`L*Gd(H6HJe<Uw$oOF4r~_B6sQ-NpkbuV0*KD_
zx#2e_+R;TJBo1;es^7lKTMQxvzE7QoIjWWe-Flh5*`XUijhXLv?`?95>N-p|G=IhE
zsxikZ$b9^%2r-C!uUV`!Og4tsf@`rA4ERZCN-x$D@TlAFn3@0li#0Z7{x~e9A+d*S
zc9B=|v}wLVPXZlhr1)CCj^(Csll8FH>}jFSVd;nV&pMO9-<$+LWJyBq(&)OE8V}nX
zg!(TQMaq}Zu66YXeYmJ?9N&=Ej=2FjKkx=DZiMxD8o-uL8PpLIY0uhCnsf^lBfzZu
zmO&z}Cah6E7>-pcQa*={C~@akS(HqDN%Z=;Z|3rLp`O#~Z6Rw>q513gmRsM4({vf4
zi^42bWU{cBnBJ<R-L-nV@7K?zz+<1(b)PbL2|X}k>I`S~D(xaQzBi1f_!_DirsghL
z$XaJGTBljXQBrRW)Hy(+JvcZiJKBr(7xaGZbY9n49o1FTEM50cx#?C3XzUEefvL<#
zJbK(0I*rp<$5`$1c*oE=bfxdEv+t@S3JKYT>nh7H^RNsUqDB)nbp$dlUA780lOs4a
zq-exg*>@;Z-eusA{5n$M;iBjMn#98r{}va-zWZZcgoBL(fl54b%>4S>?R<Md%XO7y
zv0~lYJJM=BW>?xFoUQ&3&0Ax2i)Ovu_NfaNnSz5hY3=B03%ehJUnOlu$YS}GER+}6
zd3fI}_}oi-MCU)$0!<qg(>NKp^1}UJ)mkp>O2QQC7wkEIEdHD#qGa{Wa89SjKFLl|
zpf7D;&Ee$`?0xB3q4wd8`Ouqf#*HB&vj!m=76Q{d^%i6rF$u*{qLD>(!k&U%O!V9!
z1WNIYR1e`jB`;#dhT7|-H{316q_xCM=ouFkIHSY!Rf@tMgHxdtkGH3KJ2z4Kk>cbF
zM~zYzbr6nzlVIMo^GCnjg~%pkjUAQ6it<J6UHN1xqwBaOk457;#Y9|7Z%sb?Dfh_K
z$fYh_mQS6F_BhCI`yY{Z2(!l5k9ytDp3s&>a}4|w9kor?GQ||u=d-%6&AZ$*&@T@{
zD{4qp0XyXk7sIL_yDIQ7UndNF1Jzf2I-sNvyuh`40q60aFh=bf*ld2ba1;y>MgdqB
z1f?tXe_?vjhwC%h%Z)t)NqW4)nvJHymRK$Aa;apC77aQ0B}F|x6Xt@>ZY+y#rO+N_
zV~q6@ay29S-F-0^o{z)huaq@iryFS1$egV)JlTV69px;(Ei1^t!9Wziwre&CRM)@(
z_-4dN31>_pt5u^A#2O0|zlJJ&%Eq4uM6U90&VOrgBE0NvB5xU|38rMLoV-vRRwE+c
zG>N96?{hLNVJ!30QnjI=u|>4EzSGnw|M>Osa(cCb9yhdiD?B`m!y=hI@e7T7{eg%W
zvmQ0Rk)#`_5%88NBy93qyf1dpD`cXePbl%xEhCE=7Oergf&%O}hS~k~cuDzH?^@y+
zb>yv{jBkH^mzQr7PcBd|FV*}xB8Y>_(9{KXjnl*d$J0&+5StaP1sh%_TH7XAkbFxR
zjssx4L?V?hWRQsLLU-*l>v=Uip2Scoux^lv-&@572jlRl8r%{uyBI#D6qC%G<ux#$
zqP&kJJRmB3|0B{)zF6bdfL58ww5^@LaQA@PS&npIlqNQYL9&9J<d;B83&+o%32N~W
zx=Ds$G@2xd_xPUP-&)WKoWo_fb+DVc5%YUV7)O4?3za4obE0Mn-D&o2dh*|(Yoa2}
z+&57gYb0O3WcfbtUJ!+Mj`=>+_T5NvS%Z>5#5oHFV-1JeNJiPrmHY=-QsiKtZjskE
zWqA?%&Umj$L|+I8AyOiy>){nAz1)gT(6fV=sJ547tswC)(eG}SpBb}^FINF_GYt%R
zeTG$ay~!NK3i%&Dw$>_-6b}_cheyRkl>-&O7M5-}CnvK>2)i@k$ziX9h10_bBToCB
zKNFLTh}>@7vP5WGDf!;cebFjW?jTl<y(HcpS`}ELEvw9#$$F(e)c$e00mK+a!N4&B
zRG?G)7#bWsO^7UMQ(LS)CKIed+eRKuYvXBQoD7c01kKPEDt`Nw?akp<f7F?+)E|Ed
z3Ey0uuzMb!Z&cOKqzlll)ayQL=e1a>_vF?pgx29NO}+|f7;f3gG|Obh23n_KbP)0`
z(KL!T*54NPehwz^_w;$>Y(Dk-(ReK0&YQa1WO)7N<wC$hCGpygJ-mHotwD|K1AV*s
zZ_Bs|lD+pIqWT_ysWWD|4}_e^zme5W{iZF@O3k?oZFMCS%P_182MZ1KIDSg+;4YZy
z#)=lx7GX!h4c1rO>RDlWFLLC?>0dE$2Pb4J($Y*hZ&~J#c8AdHbl}x(7L6MUNw^=A
zUbq}}jO$e%XC&8Nmy0~PkF@>NTKV<?KjP)Kk7A}>*aYbn`52B#o86~!IF)}2^1h@_
z)aQ&z$l2&t>j^Q$^2BG!nw6NP)2jHGG-txa&}p8f30O2ml96a|W9M?TBh4pGtEnpC
zxp;Klg7RTM$s#i0l?jHuaJTVW#-D>)G%MB>>EWe=Tctw`$gVo*sxOR`l85BZVP7i?
zlBYLwG`h7>qqJSrIZ?|@b*MG=y9eoMs*4NFOq9AZ`tD$4Yrb|{KGd#9v%^O+qk3$_
zEXTeE4;+lQYn@fZ=%oqTMXQa6C$}B;b#}8jqP`Hk6-M7oyd1Z^-N*FZ2wJJ_6=Dpt
zR>x93px0X3$&68|5JvuCE6dHJr`6M(af4C{=NxOF*)NBsvOd-cTh8}Vwv9l`xGB+B
z><Vkaa(A|MSO{o`_e`hI9z)61U=*)vG7=2m2m;aa6)l4|)h$Yj&fZ1+^*F#Rh1$}d
zb6bqMt(us@mq@IkAJ{O;NFqZMrAp^*Q1Q4>WjVHsEum!=3b`yHOfL}Mw?Psusi`;~
zYmn(4l@-y$+6Ma@^4z~A7>ji)ZfGIq;@i{Tr)-bOHOaOXgnx^X4<DF$JhqOiOtzyl
zQ7B3;5JiH=aWsb;Bz`dP<UeQpPLz(0+?0OJ@`}DHoVy}LTM_4*oGcf2dO9N@;Y$sB
zdKiMlo3}cpx@o?(JvoA4d=|7&%`*fJnq{^6E>+eC<Eet{<c-;C1{jgMCesPMGkb}g
zRB;W{DqWhnt$Wm`9Is^@`7%Exw6~>hJc31~uP)NX&<WG16U~z?EC5tsE{^Q;iMC=&
z?&0?VU;1p0=tt#+dm5ZPUOPB6azhaAtNX4-%iE})o7%xf#VT67`w2u8M2BO&R`&wt
z6t@llESVj=kKHs4Q>|eSmY0p%TXF)MvKb?*Euj2rAcPWHuzJ$@&Qy!#O;ol|cN@SQ
zt?hL0@X`qax`;t3_zlKzH6g1v(+TDangTbq7c^!CQ4i(CQNQM7&Lehz2%pB_`&JMp
z#ZfXx0??;+CKvbNvU#&%C>?gjeL;!$SmJUDgGSql7Wew_#ZTb1RecBT7Z*MEn@0$y
z_PW~vR83%H)mZS{4n<AvoY<<-WzudEq=10Zi0T+W*DK3n`<ol*foSYMh>_zD?Wi2B
zh<M8Y=0~euRSvcs)00CJ`tz(dr+Xj!mWUh;!|Js|GJ)-6fHsvwbGUs=yp|8Z3>D0L
zLe>dq>#3H;zDMl@7uHG$y6s;b1~Z)ZY@Cuvbv3+4U+RLv=hnnA7ch|<JCYzzv@M0R
zHGh}enkpXKV=xO@`yJ!@9(^S~8zp8Y09OmGV65XfiN7Z<&>6Cy%^0~Ip546N|D@~l
ztDB4>%|fJuAD&3=bLZUDq+AI}9yfkm6m5GcO9~J!SqliLlZhjW@x??*PZDrDW4wby
zLOiSf)K9wPH*gTd6-&`hOLf|Ys-%B+nM~N#240BbKDrFrJ%B?mQaE9Lz5zG@4Le_y
z95EajKkvl7$ZcevCP(OJj11ukm6S0i^GdJn19Ls1*@{b`BctHUheyZNp`8`cZQM}9
z>SJ;u9w6nr^ee1rZ@*Y!VV^Pvw6=$ysQ^%2Pw`xBGzIX+Oz765LHc|d-t*RK`Iqa>
z6%dI4w73jbnmyW>|C8^;4x3A;!KS*v9LK)ll@}yu>&&M=Uw$IGdBR9~#e>NY?>N4@
z83iTIy`6K2DBo_FVwKt($zw^R&(TW9vTR)h!+V^!m21K?#_okdJK%NS_J=Ka!WZlk
z6?)0zrm@{}FKoF7=QcC^EEaWl?QT7bLEZ>v1l_2rX;YnDia_J+eNNgBqE@xN?osEU
zvkN8x>cZjv5%BCVvUO9>Hea7H_=jD-+AfZpZMF^fZTsSlFCHNl4Lh5vylebm0E)Lq
zt00SwVjcxf#TW0xzE^+KLJD)*vQ6JM!j9I#q`7$#Q;8*}lYC>IK_W5iS&&)-K0)*o
zQA(a;cR!tt+FZGMxMxwU!sy<lg16e0tJi#&H}=IgX!`0}M>`K)$K3gFOXNt+AR<=6
zCZF%K{)<0wPHe;(?6h64x4QjWa_r1nrD|>v-(v9}>%RU~F7}_;2mDSrJ0;FWdOSN8
z5MKBox2`gQ|7pI<McU>+uc!t;(_WY`%hI<4{Km%xgWB=}w?y6{Meh77G3cP5lOt1{
z9ZynoKDM{)`wHEMw=i?FU39d`*D0l03oY>}PTzA>OX<~ZIbNI1T99WzRES<#<JN-`
zR(beXQ{NSlP(AyLyf^y96!~rG?22#iTJp(ObJcRxOjLXJ>=oW^uxLVL1J<bg&|e5V
zrN@jHrJ%&m{a<L0#npy@9{K}`R<iu3$zi}fRxHt<8Yq{huR-oV8dj5_E@hv}oMlQ6
zILoIyxe6KM;~QRyX0Wn+D*qx^QE|@bI-yDm8O?~S4OZ6N^tF&N>qcfsPb7K4IuLL)
zFUT);WC@2VE71Nzs7j^!pF7jTab+h$a?%xZSxEq^ATpm<Lor0}cKHGkk0v1ne8J@I
z6@>D=MW$O@@OW=YV__7w7k$zdMS&>{gnr`vWpOVhT2j*ePBu-}1NZCnbA0*Es~DS~
zp|wGML=Ebyny#<?rTeQjt2mV@_IaoixKo+4v*_a<t%@zbOscElS&I5Gro;dr;1Atf
zk#f`doSJPM$;CtKU+>NA9i%^(<RIhn`nkRoGE_KIoTNM~6-4&dw@md%Hd#Utk*=<F
zZ<JXQWftA2DdF%iz}x132RQs6hSOxIpdXFSHO1QD9LOm+{W~yAtdU`?N<0{MA&;Sc
z?mEl}Z=zc7r%%Pw;f)g=EAn&8r2*Eor~XWr%o;bcY@o|7G*zb&SrTwb2$~pU^v+6`
z@128mcKqb6N!rJibJNWf)S;1|mgQJi#Kk5od2NOnOG|4afT^qBR%_#cdvxTL9^d*J
znZXJ`9t*h958G>_K)P?VZM<sGvC#Xm(s*~xKt*4)^nU&r$F9~mP}o_iUV?cjl@c$C
z9{Pen>f1Y`Y^%0P2us<=Fh8hqvVD&6Tjt~n7aHAS0*LtgfjBcRT7p=VEe^-GU4ko>
zz1U6IbMbYaldDW&{aL>5Fhi5-VG&!jxNk;I+>z-{`{Cl3r7u$mp&M6QC0uNdfNLWP
zXi5lc`^%sZ1X<=Kqblm)*qe_RiJQU~==EnJ9yTfg1@~0t6QLj&@e@&_w8UeWumImu
zpf^~uT7&=zO!lx;{?pwaaCH(pMqZ%&a60kXn;i|!an9eD=jxgED`AwmIJp?n-EBU&
zDVMEy&hFP@j=R0jl#-f-`(y2}h?loepsfDQ#)w4k{xGK5Id^xg;_jMdd{nb2@wqjD
zK-!tZZ5aw|YU}{_PJB#+S){eyAF=LqRrYZ36PP3tBS$jp)@4$XUC8b@=Pv|sD~nL3
z@~aBv4|gC79$$RD?g}HLfS1M@AeVQcQu<)wbBG}%HUt0Y<3!mq#5~@sQ0lJaq9iUb
zQ2Lz|$S4g{q~*T8E0ClH*Y1t!+=tAcx|W;KbIr@32d~>kCF?f`98U=Fs^DKY(A9E>
zp)!rB;@wwXc!<mg8A9~RpYxR!hz5pp5(emf|3X-Q1x<?JSYv1YD*QIYF=)rq{JR9#
zEqm|#+!#&`QH5sKz9r@7A<s3*CT4+{hPBRe{hQM<!wO@Tcy-@~u`a4%9KHe6E_bY&
zJa(v5QAc*#>%gZ6>WLG9INy%MGvT9Y>}lf_-`3ZktJ`ke&k@Ha#?n+B+j_H<I0f(U
zT^5=}F-0yvo(CdT)OS4g<h0+}j!vic?eFe`3-+p%A5O|CWIzbaywLAP;SRnx?!2Bs
z?*&Z9PNmk%x$w)U1T~gTxIYo3t@6!<dFk_=9qe==(=jU>-;#ksa+AjV?n)M$!xv%D
z)Eu^!UsUkq+)gGx-raO>o^{~4vZi$8<bFuXs|f|rH<iKAg_ebw3^lXj)Tk{2LuWBD
zBEwGz{jvBY@J6oG9qLRcJuasmdstlf6hRv0{7)#&oX(R~#dtrInKa(KSQnIINIvMc
zEa?C0w!-bC@2!0~-5gT09qe0X;h<?jfp*j%4DkKIfsGdTiwZ$-wNQp%6Wdok{DnYO
zHLN~HVBe)Ac`{*y_Q%dJjiOOT;L+oBXs9)z_1%J(Z0@Uw^CS&gU^N}BagRT4z3D4@
z=UnHzX3XBqZo2*y0WK{qQL?hPy%+{Hu8_dts_$D4)ai%0b^i8pgmq@@dtT;UpUZ{-
z%{SV9Ib0_Ysa;8^zU^_Edb2Dh^+JHw<kI(Xas=`5xua;Nnoue6#<nBGiV%1!3hd1(
zkjvcF2;4|OYHbA@6;uS%^~_yu%mcM@Nf-2TS}!0hBC!TFu=VI|aUQnpcrg|lF*+Xh
zk35*tlZLt$=%DSon)lo2`d@RyU)`)FXG9o@(THIp_f&6o(g`qdALhuI;L2Va$4H@D
z5QK6eM(L|bN^=B$5E@THxul{u`O*^(K$k*~*odOf3!}tU1-23(;~?whv&@H4vCQX1
z#(J8dz98osi}ys_uz$K9lZjTylq3e1-JP$2Svz##%vzwsqQ4Nx#*a-H7RiJ!k2*z7
z+js`!VsjLlMpwt7j%3$yLR-%%Kng*9KI>)yZ<kXmD%1}DRbzZDR3_J^0yfp#*`$@b
z{PxBC^06a0$N8!U3E#_cZ)U9(=dO1-s0N5$8>kK!-G`Nd{}*p>*%VhBg=+%AEx3~)
zjWya>u;2|eE{#i|aSiV7H16*1?gV#-h7dG3X(T{^1c<!DIW;p?Gc{9nK1|IY*dKO1
zyLPQ--S>4RP8^`15YLn+Dng-@6D#cp76j7n^jIH)jf73o46@dXRdXNS*fkI<N)N&)
zP!^9G`~7_Itg=K&P4br?9cYc}v$Nis>fsz|kCvvpe)&>U#gU;EYKlESs;<Cw@8b>r
z#_ys!5qUqaiB(7(gYhLfRN`BFsq4TV1Dr%el-klH?io1JT5sbqPk289f5h@4K+6?O
zVbp5Yg4`kOdvVap<SJzt4O<ItJLb|@w819`#0~2i>--sPG@UmqJAy>;a!two$9FLN
zcM~pE90-%mJ6T3^1>}Ewgv7Qv?ABBun)X^eSa(Fay!pG}6UsJ^NHmD3w+x*Dm!ybt
z3q_FF)l9zFd&rQfPWEv)wSLn?36OakIKkZEL{=Q}6q|CQRkaF!rI|1DbdIv>jOtjF
z02&vY@OM<|7@CWztK;}@FKbZb(b4DG{j25U*AZ`VTV}}c%?S?nc|+(M$a!^^F;7mf
zG6lgJylyQv%mFAU0lY`?+Qvip+k@Ce28wmhUkkhaBZB3Erl_QXe#z@nz5oz;Q=Vzc
zYKBU%udnC1vw5l_I|Ck|SX?C-kIAoEb}5VK=?VO-2Y)Rk>3<qFxSH8q4Ei07pL{TP
z(d=>|gBGmp;_ut0I^E?u-(V4z>(ub4V*JT<yk`stDWdfu8F(*^`UQLW;3|}G;c{6X
zFBFF_MsUtdN#R1$^L8YAkyD$va7rQzAi*Y<qe?HYR%=d{J3}?8!G1uIMF(b(h+=yq
zD4Sjb#*&{*#XSo*^BO&ZZJw3*b^K{i>LHE|70c7dEDfSMf7G;z1bwXbYwjSHcXn17
zrheQ*6MHr;zpD~Nw8lh{ePGDdF4UY53oq6rDVkKHpRdYrg(O-f9I~ysySE7-?C+`p
zr_2(1UfckfsVKj$1QA<_^3sedIP{Vu=k+++KFwp<?>(r<a}QRxL~jICCINLUo2$7v
z*}41yF((N9@w_$WtQZsG9b~fg(L&BNpvk~-goqjYYm(u|rIz=-ogux`z88B*A4t@U
z^7!AsSEXc+ohPfc1C!!<UdFWsZXUX$h2yV{;9DN%kP@@aGSbgYFHW(?NuimJvGaRq
zS?whfnsF|Mxjr}>uTbv%*yBEGtYc6hIHYzA9Pj~=Md}y((nq7+3<#qT{9D-<6?AK&
zDx1OilKOKX!O(X}4DI*wfnp2^bJKUqgC3?ElpO(D2O5IBk5EjZG4)3{bE0DKdHC<e
zr0K*MiOBasW;_W0&bh?*z<^3Ei5B0V7@z(&c4%;}&q-1KF>RkLJdHcy9uRk*%lfzd
zlaKP^erZT3;CcxBksY<kw85=84?YMLwQiOa{u$sS=KAcj2}~HqN__G-JJKV;WUsKq
zb|LC<+?aa*=PH;Q?*+#zejU!6RI1CCdTgU?ZT|!`6xP6IRa*lJ7?Q~H_qxR%6Xoy;
z@#kcwzF12QQVt>wb7Nib6?Z%w?SG-zJV<YFJ-b%x&rt5fymdS?d4u{tSmB|e-x%I?
zw$qIsG-Jjnjy5eSuq;WvnkO@gfngD$-Y_0@*UQBImGJd&VQ+isG$omdeY)~x+K_$e
z!u}Rm-5K2--kHP2SD-F1h!`G12~znc7wCNP>-n!U3Pzn{%&VMpnUnnLAmoh0PJr}|
zSO6`aY{ToYXZQObM%8a-$C|F`44Rld+u>N(eOyV2kr?rK|HIC(%P=n$6SPj;_ktiN
z3>LV&a+-+v7T_9>y`%WvkQ7_R;upBE<9}&s4U>@2f(8G6+z~iA>bhdI5!s*28xs3V
zy<+X0fCW(a-|PMH_#uZMeGk7R^Jw`q{Vpx>dgUkgg6`qLBAu^w_@DG1a5;#FSA@5S
zLpUW}5Gat5%gfIN$_x`TvtSg(c=GGzK#FEF&8f|*sf@>=*%(QPq*#swJ!+ZwO%-`6
z4RP-RK4=GpO2GcS*L2;$yI^k8`RDQeq_&stl^9W_KwKxOHCSQbTN?y)_FC+?KfPmM
zot~Nb#4=Z-3Ys30DbI9Iixq^EgIoSZXRMK<xrc1nS+8;JG*M?w*PynwWlkycR{sH2
zIF|>g%f$rLz2DP?mIleXaSGbq_Q8!%*B<AeMl#qvK2CUI_cK0`6p|19-s*W7=uV?z
z{zG{lg0USdd+Pm%a_z9s{2xjl#oPZ-&i}?1r*f<q|A+FaJ~a8Cok7TdDD^*?g5|5h
zAOHPqLX&zbwfGO^U-Or^|JBDeoyiY@BiVN1UYCDB<ryOQ^}%g_l;Z98i|^PH7bAao
zA6||ljOPULY@x;hA`|T7td4cC%MItjC%ck@5HTy2;w4>*@XeA&om76uOA4G4wjQ(o
znn!`36Cc?6f1Bn14`##vLVV-@!zm0rdEW9mivH7oD2b(BK^f{VBiBot`t9m_<sUCt
zcC5G$1k0#hhS5Gt$?;wUkKOUKQg7+~Egq=WJui}j{(jyz<2=$?vp=6jqxaHgCLYd8
zIOoYDrzmVp8ifUGgCDI}Y^Kmj1X3$aa{YrH=sXOyyk)gr0B?Eer`>P+O+Sn9ijU6t
z%nvkOl$q99X@dRC5LMIt4dXIot`qi!wy6&JJlOM0*sy3-b45Dw?UvZ`i17jL)ZNdt
zYJH~RyZ(wcS<KNR-XpBmUCXMn^xF~?aX;-rU~DPA*fW_Yi`Z}_qT)Bt!nXDVy4mSQ
zO*isGdb)j?M&~C={1evT3NoZ@kp^8VXGXNWs2?yRxLm-9L{V`}ExjCWMamNFpwJga
zy!w?9g_U6UU*#i`|5q>mJJXF@-1D{_$~Uu46Ejv{^0WR8Tq3Fst4+N*^fudh>V>kF
ze2gmhNS%e_M_0r)_&GkuqfayEF|*>RJT>nRec#?A_TB5p@!e|PK63CbdwdD#S1x#M
zHT(=v%lp|4w#>x1vhP&;#MSzR#h_ust4K9oB-OTi4vr@pO@JKhoa1tPtK3{)$W>|J
zY^J*vAve!j_EsgM20~iXE0}>k!+4*Usog7;8NpOo_&&c_9H_xO3m`xDu(S7&gZL3=
zx(l>@mFoXTw5hP@<h_1H@zo?>$n7d4H>3sZ&}!WHt6+Q5`ogKd`b$H$?jD;)hf@E3
z`=+ygCel7>%%UA{a)B+_Y1KmV!R4{l0sd#}DsOW)cCahy`OAg8<{pt!r-2itRX^7E
zdV({iZTnBo+i6aBSGPvadd6Ls>A-ELPOy(rBXJxF4ZByh!8(7jnngudissR5x9P2b
zA}(hb26L(rNfnNkonbf=sYm<z0^~i<9a@?pHKJPyqDrqmk*kxhJjB6n${|35k9m73
zg|q*+{t%<z&U<jIJ)iRCmmQnaLbn6!Zq~)3*KGivZWTCb!fhvul)P-yp##zmb}r=F
zvL3N$Q^ZH!C7@&;8%uJ4fA?Ed?$C0(Y!NhUFBG?3vBcPz4{-T-QkrhO9FwlCM*ZD8
z$H?hc!^>79^EQ<vTEWe%ZQEbM0g`uNooCN7$|N;nvCi6dt23{4*$biN_gkGS!|zg|
z1uVxGFPc9<+a{HFP6{Zq7Tu=e_ADXYy%OW3>I`wB14ylaoHD7*kKdI%ayoYJ^}h!m
zZT06pZSJ;i7FKR?XKckfz4Sb$e8Fw2&Y1Uo&EniA_;J7J2Pf=P0LBYV99fQl-nK@j
zK%@fq9w_ib>zD8Kr-&e@E2P)gd{Bq@d)F(;#WArZKcl9*j`W6`mmp6#!s4l|Ii9|0
zPGXCTZ?h!ZsZ%wbf7WT0HB^0v14$&2qm8Df80Aoq@J?j|N>1O<s<v*ITjsCEsBZ}<
zU5&z@S&=`!^<R;%aC4QT`-_>FU@aIE!&*Ulw_!0;by)9jB#A%5Gr@Y>a`k;-uge%=
z!@FC)nZwl;uxi?D&$wx?-?31hZPU8tlvZuPmGycjYtuVm_MXC!vchc>9BzW3uQ7Of
z9UqQorAKQs`Sm9rINqTv#Kt?C1JU&DvR_=lk^7xmw#}JQqq{RHz_^j(vAKWHy6Su1
zQRiYmAYXpYY6gbK-u3K8K{Q6}A<@0vf%$uCi}GVk41hGLSUqDUC?kbk+=&*e$@<2B
z#j~3#X8~-1i}6N%N*bzsW_nB4lUMEQH`d9R$(&o0DI>3Eh8)eAXU#a{I9}C7i)4%E
z@zUHlm-FKX_({y;9ASO+7D$rXcpT#5j<Peb;|L3KHAzs@R?qp|>$~pN7GwER|Mk)?
zgKb4LiAh>&T`C^rB_iOn`4u))HPv&u{?qz8Wz80>u__xg=<>4`R9>R=C@@p5+w#*@
zl#ilp=%1Z(8a{~n5X(^UVa43hWQ&)Oy!|8SdTH2Jt49^){wi5cbA6&mUor$U@qScu
zK(m%_MfqkrTpzk4{n<mVr?V|4XGs~b-opV#czomN+p3;;)q1>O$&j$Zwj{5{wCXYO
z1M%RdPnXduAPz9{LitHp7A|749)&&sZ3>hl=31IR=GQ6m4KznO8^4jv#MiW+^m+P=
z8Gf_*Mz@c!w234k<XWSNQyyf!VwS??O1`g<x0<Pv(XrP7eF@VNQ#qD-o<eLgt0I+g
z8#VDVSmTghl^RA9=vPuIQam|)5yGW;X(EZ3?h=ty@^^feYN|8esnD&GPw44~?-y*?
ze;KT;xs43G{N~5r!^R7xA#3Cme0-TK5$rYg_BwukJyh{k4^KqZYLQ_hakukjrOXW3
zq8e~*XcM~SrkAov8RQW&Ep>W9-84%GZO_BvDWLqZpvC;Smq5YiFNdzY@uI`9?pNSN
zzSoVJ4Y}u+LJf!xzFhvkazpvEK;4yr+N!*bm@fHyYy;Ozud-2YNkv_21sbNAzyQVN
zN|GLy%HxeXa>3O0$Wx5yR&Bz0T5__bk3W4wK_nqjzYG(l8DgBz0zvbVh61Hh8ThZm
zoWL}LPr4qb?0s707H72DkJvI6+O>xYpEzCjGnzU|qVtO~DMt0FYwMVv2qdIi#C0ej
z%p6Pvw5&0Y(~b3b<gPQyDh0D-nJh^Jrcr1f2{riBK9Q=o6m(d8$KOymE402(7zE9A
zy=fYwhZtif-w$YTVYif2H0BtAN+zh{Cavv;m5)cZx_LD#?1Gt*PjR>O0%}WY$~qNu
z_3psway8m&0NKfRqU=g_MwNbH<cQNW)#xQ4MMeG+^<Rm~Ax-E?G|#>{#k<`X#*Ku5
zpN@~->q7e9E1I{)tQnKe>?8bdiZs@NIjv=xip1nRR<)X{a-0fCF#x?<sk-_;tt_S5
z`~`t=erh!i$#4Nj7#pFkdc6>TxLF~NfFvtbPSzM9h=g{GatSOG`Ph1h^+pKmHIS!u
zKe$h6{m+M^8UFs^`6ePa)lj7aR_}iH(RxPUN~0mOK{c>um0baj8^nyAdbWZye%p~z
zV0%{0b~s%yoSN&cN37a_<*$lKWsapB470l`aX83b^3>AOcA#Zix8;?cj3&MJF+L*8
zxzSmrxlbSYN##1)I_bi@PBv-Nor@q*4oxYCKIg1#dcB@7^jgqsaLi%Fz^R6^-6Y_q
z0Wfe*I9Aq>R=xu6CxlDHlkc!qM|Aq6@@al}|Ej0UQgt15cvr_UFLLTNj@$PeZ{EG&
zWo~h7Fg!{pP?;Be%Lv6lQu~wSo9YZm*8^d-Y9fU!)yMY<|MF=6P*o-j>eiI9;#6k*
zu=8;5Z7WZY5*4L^sS{t0!wvVsu|q(yAr#!{Ei?yBxUmPjAW~^pl+R5Q!<AU5^Bak-
zUsvDi=~I<5j`D!`w5x~z8rOBSY8djJ-88yXKTNzcAYr%v@#Hn?Ai71@k%uPRP@$3-
zzh<k1;wFDw=ZAQu6mwG|$4Qz`mJTF=#qBYP+Q`dsf&Nh{*^E(_%*2X>rBb^u5$M+j
z@M0(ZhqC<a>|**9oqIZMhcel$ZfFyj8_N=2Rq07d=6ezx`?-UMjNR<5!yDe)2+i7+
zg%7-CH5g1BXwp#?k<n$cf0Y|LwM+6StB|8gCyIpj{5iSQ%y&Y#aNQq<Xw}tKDi#{=
zRHukiIFmYIJ5;YJ4i!{kq!Gu)Y{<h+v5O!fq<o`e1IOV8F=h^iP9{`>=+Mhjx)nA0
zKG9co&H5OyD*(|sNrcpp%_y5hvMR4JACCZk9YDl<a7EcUlo3_Uqzxa$@;I1^!sU&(
z^ViM&iciQ&PxlX%$c;!*XRAtDm(V=l2v)x0`B*~Fl3_`mxhbU5VrM4wGk02>pJvST
zmP9=Sy{>Qt;`Ky#5XDsqI@#mo=hxPclD}*L>maC?;9AI82k6a&k`oYSP*d~Eo&8<e
ze<%bfWIq1aXkRGFctaOH`$UB4*Qbx6hzOb0uS`ixbB~#FZ}um4(mi#pn^I3k3j=Oy
zyyWbGB#GR&LhuI70?wJ1UooE4qkb&ZTDZl;BU(J4IsM_y$*A0F$XUKV*9ItVI<^gB
znTb|e(x^x<BU3CNh@1ARW&FOnzFz>=26NddS`;ZFuR1JCU-uz{J~_lpRiTh{s6#d@
zF2*z{Iu+os*>tF1h(w-1c1}uy4l_4DCi@+~b}sTsw^LYSZnzl4TA-ZA&NPz;W;-Gi
zZ-97wl4S>a&Jj0<Zx7w1*}aJ*Rx@xLV_8#=$~X*SiD}TNTs@A$saSapQJ2J?CLL#K
zlOL~;k0O7NHG-;YLQOHvq+4p5WVOvydj{3faUj%G{6E}V7*&SD_g(x5U*QY>V}dYb
zO+*Qd#dD2b5g3-5KS9qDA;>X89oBL5=a>pNPB~82mv6G}TVL!Uy?)G&O3qrI%^<&?
zc1NR9g}t?BEhErwP^=aM*c}_3psC0%DVSQE6#?KUKDL43U);!2MbZwVT<fD}p+2^L
zr9wpEQlJY;7vG{VS3OAAprnkG&|&&mn*L1;bQ>PESjVd%`}cBhEFMp%P2CvJ_Y~6+
z9vOy%0pV1OW#p*Co$=FI--h@*Ypk4jL=EA+S(trOsH#RiJI1U+rvfQbCxy6z`NlC&
z$m0eJgbrm^l*);)82=`UrO$XB;H(Z~=6E8}YsftaaK*Ueni8!#4vUF%^a^U_@eIst
z9Muwll^`MT=ZvjD9S8JAbrlNrNzQj(LR_v8J`-#Fb$%i`V4jF-guAFK)dlfCrhO6+
z>I7cFW?U%)NdpLbTnh!8dJW~IemZ)dklVCG2Ej~r5E6x`Z6N)8yod)%G}ZETDlr%^
z0H(&J3td5LHOX2zYD-;c7clAdc*-4@;srcvw+b3!vvUg=lUe=G-!0f$Su8lkD%|XW
zz&xZB-w4@-EI#d<uDKh8qOcS^!Ug(i-Z(Z2;Z;Oem-RVUkuJB`uBx_H<KUy}0_br0
zVya`0d>uRO^*C3Fb|!E|DFsYav@tW9L~&H&(E74$=}O%Qfe9sP)&iUi7X0@mDtl>6
z7-ewoZw2~~u9FG-X=<+IW8eeY@o4u~lw&^8LM+9YBGFPM&)S2zb%GBZnyj@~@|B=W
z>oS`M@p{0CJUKO|Uv4alSpy5;y$(4=ZhQ-x<wC`hs)@T93}qaCX%UC}R-liaKmj0f
zP#7S>ju&Ix2^MiTD3%aVuP)=%h@Vxpg=pU-0&tkv6Ira8CV2tTuC`BkdQTOIl;gtU
zVqd`VUIKBInKx-No_Blj5`Ebc&!!`VxZvgw>n-kR`8nyfX*dXf^FNgK*?&J>Gf_dg
z?vsB4s;JDE-ey*b{<yq5HFP@Nw9jUBa_0(n>cvEeVdHBb;~+x%Gi)Z&(X2w!2(G>j
zv2!*U7}sb&(96Xt*+8kfBG&}}s<Wlkw(-N=!GX`6D#Vz1Q4A#_VvKBDym<xL$}?;;
zrBZ}eYiO!nW~A-Xm-N}65H3wP-5BewT{E-_!tk@fCS%jK<9v)Hgn9Ukb!NbJ#Zo)o
z`-a?Q@Nd_Sr37mC5x9NqN68=eyYWK@Lmvqm+ckQmT+KGXcb07*ND|wYTyh;c7J8eI
zp95a<y<9}Iv3RI$oBMW0Otg~7UB(#0*Z%6UI(3FWkO|_tV;OPup0!kt62>fF*~P<z
zOh~Kh4QcGvh>&N|VA($Hb!up-{P?&b2h)8WnxAVPw<eI4juf(bU1k`JwntGvxFF$N
zHMjW@&t}5pEm6g;_fz^Mf~%j=SEc80i1+vDW<URgmw|9F=ic&al}4fe2NIrupt9?J
z*jpx(ugbZ%ZP(c@#kS&X9SAOOwHLA*NzjQ+Dk}o$y=I!u+v>6NOE=HtLzh|Br;<$1
z&7){1ey0s*gyhfG&1<|-`fd3c<LK8)qpnTG<>YgDJbet>1sl)V(G$Aw)N+)e<c1hJ
zGSwPtGwJp|q)|t_ngo<=X-KG_Alk-ke5Lxj9EH`vk5RijL=j8&zBAI_?B{O`L?<P8
z*>!r$w$Lr^^$>T<%X{3`TX4oWuxMKvxTd-$1v<~z^g8Lqn_l?gsrG&@wd${$%d~k3
z(3zxeeVnYX8uK!A{*|@#RWHCQz`27K)uoRsW$x2QJmFQ7EWFWm*6%Gc7GJXiwm!z|
zW=(mUj>@}R1#kF{(EXOL7b@~Rn1&Wpr!G2JYgQP<m!0*P8X@{X0&zI@Udrq2Jsc9H
z7**})w)KW%YbKDgjDs|%&n7GyBzGPj?P<%0e)AnW$5f<pST{do4ia*W-@7{Vj5<3f
zACj*JU0E4pk?DIa`R{zL=lMcS@^Uj!hev+PqgJ-dbknvN?FIh|#I{yN<M1#~fA=Ct
z-CT;(;r3a7Xy3MP7cPwW)KpkVYl)ukb>;JwB0ym_7$H0*_E6~3m*8zoc0t~xmFnEG
zkV~h%N8M&?#6ndb0cYuxTaQGNB?i?|seM4u(ekK~@>lpqh?rI+K+!AkJVa$ZVShsu
z0Su_;YG?r>7L3GWPhL2G?5rm-6g2UwtrM*<EjOLtzWRO_xVQb<{g2A#@85?(6XpZl
z-hUI?On%MBb62f<Y@QkOI_4d1<+g?Rm~E_wgY*27rhagxObM%%Nf)phR>w+FC{EJa
zUV774fAS`L><xZn)>ThlG|}0Wzj~SPkZn6~HruxA(|FPD9^19s4enYIGvU9IU2)9E
z@9pyY+79MH045f6=P1XNT1i==tsY2gguJUbYB^X`WGNUeGb5_gZR*y=Y<w%hVxBA&
z?DEgSh9Xb5qbEuM=v8zLZroWJ3EsC5zooj%>@h0uc#AXVzeA2LLHRvgv>Hx5yvAT8
z>pDLtS3ma#Y@oE$pu*sw*x^%SR?SuTlue(Lu!m!k6_Y)jmc3Yb>36^4tuSr2v-Kr?
zg!Oe$$&hn^knI*;m+3Zr7LIW2*<MzE8>diPP{mC9uAZ@ko}DvA{OYFoPXp_D5{qz7
ztyCt&#_ULfmw)psp-$5uSVknBgfM-QH8t52=677iDw-3O!s*(TXIbZUF<cfJXK@G?
zogYoqVl}Fn8dOSBXoA|lqzAm5;zCye0ym$LEq^S&s1#fU&wYJIP`CL4-?VQ)e2k;0
zuOq1LYyNHh)gk(<b(f-EsEuPI@=1WXRjtJsmWr<Pt5PkO>*M$5MZ_QZfyN;m@6%uD
z?fEhuAF_;Jj}sK+6YScx8<%hP@9np~kW88b)Q~pC)+CoNY(R#t&8yGsZ0h{@xjz8&
zv7`M2kz6u`>ub8=>IIq&3Z+$6?5onx4pr+`sai04dvTOn5aSzQkQ#xRr@Oa#S*$g=
ztRdBKnKompM8i|Uj=g=gtlqn-KafovaZvErbbvGKv{{>tGyRoys@LVif$+Lx|8cx&
zlT{73o`*%da_UQu1KX(`y|;A){qLsxuabYa0m);#jl~_=+zS}hwq0eHg@j9{-n|={
zzbsQ(Cwv#-3pU|addBh;zoAHkGx0opo%alqNv8d-JjXoe$A+h{;$(kgw)i8()9}c@
zj*AbZhl9~W0>#h<9sOI$&{13PB3+dY!}<=CoimA%aDP-F258`@apR_tk;&DGQW1vd
z;(+o})t5{Sj{@c6z<gCaO(GD8l!8!RxwR6eSgMvi6v@c;!IWzW)&Ed3Tv=cuIWRex
z9P$HKQFHkB2|6U30ec@U-^r?pzM?KwS|VB+AlXf%2$Gvweolr>c5XLW2=lMfIbMV1
zj$%ZfgM3qT>nfJ-Y_ScQBjsCjBg+c<yk(Vg=FLG078p6N`b322QlE;)^Z5K3W6kXq
zUe_bG>C_#IZWdN`<lRbR^*Ou=<7Ed{j)Rz$*Bv(DIIlKhYYW8%Ygsh%R-d_K_4&XO
zCB}MphFJ1HOPJ60nc_J(&<F0}weo)<X~}AGk(+L@FuuH;QlfLZ7&ari`eA4y_5n+*
zATh}wjx6$DBYuz-$8e{2_w_5drK^R>a2rV6$=<v3M7uDBbRjF7<k^<>sI35~p5SbO
zR$rhl8doYzX+O)0t3t5Qu<~HKSfkI%4E7Wchh#!VBRXW+X|S1?eMln^pudLS<WAd|
z(3q6Sxp1i7Nr(xlH0|D3SVRzftK*FVKRd)AsA<AfZmU;P)l(lN-Ds2o^x&E}PEf65
zH=yx})lbdY3XQDU>M{v~$2?%Q4V(3mYUCfk*z|M|(<zBOX<W#DLKls~W*0#eDzj$S
zcj=>-g^NF(f=#pT1&a|BD(HZtT3tz21nX&@EzN}cXF(d%$-Y}?|JfYzjEhhC%*mdT
zZuJ!-Q+4T$#8dkI<waC7rE=(^+Fai+Y-7Jr|L3rLI(4BL=fdu?U4l4_2oVP~$W3`u
zqo>!JXOw-3LVZT?TRhiEnB#|Apq4fV$fD0){?^MJqOW_dTw9o1OQMZ2UPlUMT8`a0
z6s>4rzk$i-WruN!zGLL&APZ29oMWQFbbV)-!&k1|$%6?faOe$x>cNb;gTka@(Lmnj
zU-)cHSsOor;sLdv+1GL$h7%PhgsEDXI?NbAsj*KEQ|j!0`arr$xW^gf4O#IB>w$*V
zylyIOCXqV??hv=A2A(ZG?q0eobx|g5Y7M>@eoh&hNEzQCYC+a=3~by|+KubPYCe*F
z`M3bjMg_|<jG<l9FKT*ysq$YA0=F{i+*rjfNN;>(Qs|}^i#P|GR>^-Yt)(0B>Zn3$
zyhG<0o>m4MCnh_QfU?*9c&>(ei9y42N|qJVv8>xCTTk%z=J`aExPL~<%L&P$nVf$g
z?Q0Lm-ei!&iw1N{zCtag<9F>=dySOPY9x8A<E%BHYCaaJ8wuY_tK@VcvV*1N9KQEI
zc<-Z{*=R4eNz$I7rQ-=4Pe8K07jZ334~Ln!XQDC38P?)V9n5?Z&l%_Cy#td8;M8KF
z(5)u$VSLfcs3W*u2xYJ_YBW*cTSKba=@<WpqO+ZhTd#A&5i$E-=jLm&LrS$&Zga{K
zsH;z{Ix_?R>B(Wih+JC-v(pcve|D}_K;r&L0-e~58<!vZ5zkajvO0Wg(okFSkjpj3
z&mM<H8Hp_qz?H=W*PCJ7)hK?PbtGc*JxlVpNI!S8epbhdkZJiox5jur7aY1wS#v!|
zT~@N=-*bC(J}ru=rsda-poL>(;fry%7REk%(YVW&ve@@3i{ks5@?VywK@1J`<A6Zv
zZmy(zk)%GGD8U-{x*umua&fp^NIGm#K``yL2tPKun;hOUw8dc=YaF_Efp=a3ju8to
zp2%PD^GT3Q%p+aFka2ZRn&*|_Q_y!}Jj5f2{PL-hgtd-iq6_q`(6<(X3yyPlsO2dU
zlj4vRt*vM2m3?e-JvUf7{_d{M1t1+KC8=m<DP&QwB$gu{={7F>P?Sl%w(ef?5?R!(
zk)eyN<ywD5-s)<bRM)Clx1<_GJi3Qd_gT~p8@$>MuUqR>sK~VUm(r*`VatcC+{=0g
zX&6wZwOe_pRL=YrHC574)=Xs9YtkK#g)Y7jbhD-*Ik_-VEvizKSUKDtUgFrED*W%T
zN+He)P&D9B`lf=Z#NDPgFt2{I&M*nFqhTH%b!3cg;G%P=XVx|N=A&|g16ZmIj_wgf
z$_d!|b#THhpO>w^s2~URblSr8#-TmsrJtYgE76cNFN(asc---erhE0tzSO5)a}YD3
zV;`O(a)^o6KzI0$PM^-x^-w*I``cF~a(kgM_Yl-ub~C|NdbR}j-iIG<PFg~G$VO1*
z`SXU+nj_gjMxH+dm?b|w+)vDa)IGHjGEet(iMU(&K>g02YClpGfGTIAG#<L3qbvlO
zLB%(Dk?FdFUQfl<^NkS_N#9R$ufYQad9(k%eQ-}S*P+%8LzQzy#FIzq$SgLIa{@;%
zk7@XV`(IEs35SdM2w=SwBMpspAgoy^C0bW9M=NJ`HiHwFS!#I}j*agoe~p(~$?*Z$
zT7{RBkFV%W945rDhBlFmwxmol*Y2D50-P;^Nw_*kasg{@d*QM`CsTv-GNrzTGIlTC
zGS9V}HJo$-FaG96z><}U-UUbV{#uYQyBZyk0Fm<2P%+P{9yJQK&1fCs4D5+a?{h=P
zdOh<spYF?>$XM7e<HOq$AL?bj9I^8$NRv=2sY)Nn(c*a<?XkCFJ*`etLaCayy2dCk
zQ%_cvyMP!_Hp6zFsMJ9ATvdIVZkJBQ#<z|RJnrTCkx*t3`s*O$3AHycvNT(XA)5UF
zlRbp}E6SldkA0@dCt9-F$Z>^EmKKcMFKLHi9qZYSEgnfMF3c-wEbcv&Q&BuZYMQy=
z5u-6uvVR2#K%$-|IYo6PK-+#o=OIYX?zz@k#DNv8bIv5Y>c)}w?wzPenm;-=X;0EC
zyfnACNby!1powl9jDC(tlSrD7kIG`S#J5m^954BX`-668HqCpcT3te%0}BDNwdHaw
z%h)J@e@1Mrt*&RMRaC9Vnl+-`DDc?xD4iWDYGQAwe?XhJAfKkClC3(Pjb_Iu!$)Nv
zN{LoQJa7;b$!$WDa6SG)&?Icaa8TmSj}|FnCf@?Gz*qZP;q!*(kAErOYXGaLUqEOO
z6~xqQYSPdCEcztc)%Hd$adBeG1%<(=Voiox>iQ@A{I(Miazp%r>d7O6sg#q7PBK<1
zMuv|cv7+J~_`RFV!o%OYRjb=nHLTs?Pk2R#R2IdN9!$2wvb^)xJZ7B=C!QAZ_?q;h
z9;snnQ(!1ikV|f6;ov4u!5vj*2DhqC1q!WGcPB}hIf}djY1Qw!Yl{!(Wt}qSlT9=D
z=S?m>$+av?V!aV7h10L9(HiNK{Asya=Dvz(2wVrY#@#j0A6~pTjJU?_v>3({#Z`Na
zRzzaup+-Weww&rq^VaA5*r}50(gVs3_;@Nynxtvz9-*dE?4`T{MQq8q0=q{UBV{~i
zG`M!gG~Z|{VEg6@YSNW5=r&9G2Sx$pPncu>Nc`UsOX=+AP7$Y;X#GovAp4G96FxX(
zd{&Mv9@ow$#+g@L6YL~YD5>4Da3U;PdPAPhnvSJiG3{5|HZC~QhM3J<$Qr0gyCg(1
zKi2k*9}tnVMy@<e<!ZA%MI8c2Ag(v9n4L3^N16H;ltjBMF*%aZeZ=d`?AP_@LvSZ8
zi=7o!?eJ1UX<<R_Ll8YhGHDHm1&)0J4s?ad%!dCYxL0!>Bkw>kt=di*<}3lbowwv*
z=Nk~QzE-5(D~XNvZ-))T?5OZ>>W~A=F%CDiHe>m4o%P37a>QF6(`Y({4K5W6Q8xvL
zb-YO^J^{;3yG9*|z`N#Qu)hY|zO@dAS)K5NG!~82qZ7Iyn{Y3C*sab0VN0u4Jeg!$
zieySf_V|o5F@^BlrE22v^>|O@7y6d6Qkf)dMmM}J#geSgW)D$H{E!&8r5~y4v!5v!
zwlxu(k#=ht?N>|lEqpvvuzo}tZnY)^dRg>D0aOQ*YiH2c?&?G{Zd1vSWKV#Cwehl-
z2<$Xo(y?n`4~|bQT4eXDzWczNe>Nf`Q~>SVn3b0GiwRF+ya(BM+}<erc}p#`1XUSb
zTxz9a=#nL>ovF_9%wZi#u_~9k3fn*Rx4I;xd4~`^h(<i#dEI>}k+O7pHOC1xS6OWc
z2MGU%sK7BX(Vvrg1#+%S0=MB;Np!kTq0zfFRL7!sC-I$z{#tQ7sjIjFc=$?}Y+6ym
zA|Qf<I!~eE*Y{y0%^C<<_IpWGYMN-5&UJWETUH)t#oFeNbsUpi;@%OQ)bL6*Ap#yH
z9lC{?C4uo?8;b-qc|JslXWp6Y7m*O7RqB+{CH*<c7aB1oeyg#1x@Z1eb4z_v-`c0K
z)v|;`tn!AUR{8NFvW)m~B(l{ZI$JyUxlbNPEKSy!>XV|nL!ONpRmr&l;Mi*6Pu#$*
z8+EEhU<0x;iT=yU9J)SLoSljdHb$9<{wrB3?J*)aRTBQfQN;6+dgBB%rt(C_#9@zQ
zzmyNGtKu3KKk*|vTgS-^nz=$QNlr5j&t9X+=P{2dLBt)2#fGS5q}~7Xt>xCF(D7Ut
zdu-=b%Su%h<;67%sL(_mSjmq32r5C*tIP|vJzPAOg*d>ZCp$x4X&FWP42zwYSyn2S
z$swhi7=>1Tt#(;uCr7{umDQwA?9OS-uf)6|hMf0uM$0c6jb>LZ`8eyH(h+wHiOz7>
zr^7UB3Oju%7H)uo$)B;45HCU`dfF+w+|Bfw2t4Kv<{>-E1~8wbB&1=En@(BdPF?W+
zU8zX!zP5zs0M9k{7u6Gq7)`56Fa4Z(OQ2mB#8EYsan2NHkby=M{0S^Crbn;ArKBr>
zoQ?(&rIJNeVx@h2DxIIoi9Gs}jRPt~7?`}bC<h6lx9+daIJMM|3e>o&m@mkVw<p9p
zM4e;UG<1v+W{8y?k~GqffLDzaLW@7ttD1Q-ji)kmaN~5Vr~pb6M51jKhzGvpN~p2<
zjkVW~{z!Y&SX7@I4awl8Mx#l8n0UfBRGh!3{>vG9Z)TyDlnHMsZ|rfDISeR2uQUId
zY-#dkS3I)sb5DH-(#EP<R|RmYALr@iUWJ2%2q>ep*Kin$T$it&ItcmAKc{}F%}HxH
z%BEk;^kdZurz!%zD2;sHP~9;LQ#Z+)$3H4fHhurVH{8=Cb}v3!=b5x|)*4{y8WvH2
z6XfO4z&)dlERcR%=VqvMbAAh$Fq6+*0|eswXXs|suQ}FIA=3?9PuCRMzE#{~(8Yz3
z31A;!jA5rye@KysnV(@bFx;FvX6hxF`yQMn(Rkc_)YCduRD^iWY^4IM!ZZs~Nb}w>
zE~>^7s%c~1o!qb{=&-UP<>WeMDv*F3gNb%lbSzA(yVT0!9G;@w%s5=mFG;LiZ#9Fy
zc^KaYb1A0Q?GY6n{cWyoos(LaDOP{7GTIO%I=lFzh7+heH~4$MQ#~o-h|#ViWQ^3~
z#we6F^mS<o4?R<bM*W+A_2Rvu*cbPmnrmq5^+k&6WOXgRw~x<X+7f%H{}R)_STkse
z;gT`y6qL<w{T#h&Qf5EExn2KjX-1Gbn1bzr{$sXYBG!!+hF{v>q}aBu$yX`=Z!NWm
zwLQ&Ycf7FiJD}13=`#BNHy8JRwKM%cS>*mFWeGfmTOY`N^-Fo|Vd%2oxBbP19a8sy
zWaq9QJ`gOt>SnyW-QL;*nA{1T-6Icyea*ETlrRPfN)#-!dGU|UAN1QNk{}5#gMfEN
zQfWJ&??+K4cLYZ>Uc9LHqa`Z|`~dp}Rc0^2CZPU$JI88`LM0ZDT&m7`W?ey&9X@vT
ze0=fU4`z4dSBB&5l@p1fyoIqa0Zc)v5a?rl$2;M*tIWQiy|%RAuaeh24XagPCa;{_
z?4T(h)1R@yYGo$9VT}~_WhCZy5sG-Rg@M}#QDgp2i<71s@9%)r?-Gl?b_mBU%Zcfh
z1on^FT|3Us20P`Ps=(PBx_Vg|@ol}?LR%xle%>HeKYs)E#%C&5z=^fEwd&*6fpjwn
z{@k-Q93)Mkv=)s!pw{z#XVXAAe;AbRu6c-`a5P4MjCDWaq`|j+x{CViW%ir*X6vGL
zi&dl3X>ZZ^`|n9rZCjP_Ifvq4W9MZ#bPtATldC7v;|Vf%rX(Ucs_s(jB}*5))w8XT
zxa`WV1@w5~)hnT0-`%V*G5If7dg79cyV-#@qV7jYR$;((>kc`5(t~U=826(EUy456
zM&3t5`=30TR7>Tadp6y3f1>FRR_hTYQ9uj9_H2k`n|pbc9=311oGkiG$8lQRW-sYz
z+&Ze&@py(G5l@WZ?>l0v_xY9|cT)x>shufD3!AoHUq*U#4UTo^^LFSoZI9mdvF^2)
zdhpPWejBphCE@PQ%u~G}bLXa@HO;&3n9USZEB~2OG0Y+(ej_}3v{m1J$h!1$3Fru?
zugk=oH>0KO=tl3Ko$xl({=V6rk^g-pV5gl|w9B|}#jbtpM7@36sjEIHW1)Q^jqCTF
zq?P;gIw<NA0<G*+9j{w0kcMzxu(I+AyXb?NIJHi($bQCHgqO;^NKm<7A1_I=Qv;U(
zs_J9{_8m{X_ajaj8M%f~WNyQ;`(}t!@<(B&<9=ada~JVF=Og#a=&?Y(R->*a&aT_`
zxgp{9L+;N`r)t|4g|>f>toj<HfJ?A;jn2)>R3jz<XT3NQ^WA*P=KoMEy9R%Ceho17
z+xg*gv1Ir2=peICZxJsjI`Bm%@)P%0wtckM;4W*R9<ADeJt?>M(+{rpMkaYy-K-_C
zmz(<PlsZiM3z?B%rUUZQ-JCqLqp8;qnOwhT1q-l}IPz#PY3Ox0FaaYwY*y&yhiEI)
zR+Gt)H;Z8gsBTV(XqIx(i*BCH%XRq2vFWqzvS@;$N3CvS4&T?C+$-_#uYwDE;nykX
z!X9^w`Q*RNK})H30BPp<Q?<~!+}>xOt|@1hHuaskOp9pt>!Ad3j4p~-zpc|qB~N-Z
z#+LRd!eK>{CfW>}bMrDb`9IF);n?!EI+9N90;@SCuO%PP+}`ST5EJU5t_;56+|Y1~
z$>O{^GxZ&+F1!6&VH9_G(@$OahU#h58Y_Lm%7i=XP&s#9Lh41J?4a6ItO`S=l2$P2
z1J?neNC%O(s(HkIV7f@ZUj)@B<3;KUT2D9;5inj>6HONd92l&`=>yP};Lz|3j~iC7
zOgx>*YUWYZfG_iF=vbvu9_)ht%eTfy3rI7CZ}g}ai8oxc?t?xOiU#!4tnFJZC#`7d
zbJHS+d+w}Ut^l2(##-kKu6YbR;Q4%d&N+^?wpg5Wj;BA9*5fF9C_|K}_f2S-<0)6r
zXwpP!!cAzxx21TPVP`Xq=3XM5pE+iKslPO`bf<{RjR4+0=1b4n6U^~m8qom{$9;fE
zG_x987|eK^r!jICjTP+tn#g5hscS=bsE|W*fEn4d2$vLvF_#NNUB}=3hY}-}P3Pr=
z`Z~O2G=!ar$PU%kJY?M?!u&PS#qwh6fQn^5Vf+p)vH1q(7(C;XRnuEi5LwAxyONv|
zk@^D5{)|SXb}v#CI$Xm-u+GY;otc!wsQs2CvRB$rh`)|4j`v6<7v1=HqL0zL5+X*3
zPlfh@>+>No8vA!glUGMaO@@b*praYO@*1+(w)}Z>QN~p~r7Rm85_11EdZ@o$lrGyb
z17(8e)&z>Fz6D7c)OJ<jl{nY2Yi-6lG&?uR#5-FwufW}6#%$GA7H@o?!Z$Yxw6{TF
zG12lmwlb$B7CRcHmo6k~GxrO8eWU0-O%Fp;T<cWS3;n$h8D*A~_e%-k*%fMGG<3+7
zO{A-kM^#)&0rBq$(y&dtIL~KOw69GSOH*oSDyD&2TimaVNVFxeLad8CdmFy(@1>Ig
z<X6y8SWBan%KfmW^J-Dx^yjks@O>d)g+Sx@BlVbryRyivXl9P@Z_m4>PejHSyeixI
zkvMc8L}QiB)5om!G>GZNOk|8QZ=mbM8v2@8lYo>H!Yn1(bD}2kz;m1eE2PtWSmZ!M
zwNlj_QBK&fxb71&nP`VX9^)x1a58yLOL2KUK%Y5!cJc#4^3%kQRYQ*^^I)R9@SFH-
zo*EtIH!Q}i<FkWh)Ri)<$gw#pO$&e;-oiinBsqe2)Zdz>M}=Z-(E#;DbUg>*@?85l
zJJC}#D>WKU{qi_TT<oCQe-!uxa_fnXs+9cq5eLFL^dh{?A~s1D=$X)V2u>rLo~oL<
zQaG@^%(jxkGtu-WEZ{2+L^+K&_9cxy#P}3ALX|u1q!96r0$+Xgl;D@t7?EW@=G2%8
zO)r-`;e`!w$7Aj-s_wdE9KI-QhEBz_HD(nEEdowu@vP%;=svsgS!p{ISFX9ovC+w{
z`uj_xE7o-!NR74=)}1>f>9SXYI6!<x+46$!rm{vB+`PT%C;;4B%%Z&CD;KFauPAtY
z50{YAb@A0b?d9)*k82I!9xWZ{jwI|`@+<cH2U~UBK3oNZ(5y0h@S%Hoaed19R+5H}
ztyNq(^sRVCx>^N+s=7t$qAXWdj=zgCK;0vc-8|lxU>YAJXL@x+%z!qIi}GE@Lv{3m
z_euS3pHO-Le0wNlS;U=BB=~{+izkxQr$P?jQ*_>l5%`%J`(`ON554F^DM#Z6<prII
z($OmnLypu&O;PnyinNcAyc}j|#y>CZERh)UC(^dKh>a>%Yq(QwiXuC~J~qo&v_*{~
zs@`cW5BRH3Rn)Fz<xhy%r31?|e1mYL;kQR^W_@4lw6R_L>dehoCN*9-jb6HRJ?wRF
z0jVm9=B$pW6I7bHRcV!8CK=Bqh;J29b2{A!x5G3b{t}=p0ivCc|BWHfJYEC^Kso%0
zR)I_i<v<x7%ZG-<ogtCHH!QV$gLzVfcoc)XIwwtTtwLYMvkMD4Gin`(G0Ai;%xF)t
zcED<5l_0>JAg6$1zb@CH_&`FE=d^PCXDjxcfQ>avX`vt=eeIw_1V@fmQbe_lk%^7F
zkm___Qc>m*H${#po(4rFhTMg_)CaZZz&pGJrccMdn9c<{s4B3)Ir)wYwC2>QtJedD
z-Nne}@!*BI+{3+&aQrq%0ZG|?(n-F`WGr_I7o}#AfQs=1Xlr$WTjc1Aql1*jCAx{J
z_YmY1QAnu}JH8x4z0qhVQQzbk=0>XiEPZpmn$|e^f(g#iNW^vf;A$Q>HZtnM5?GZX
zvYG<ja@5k$wVG1KnN9S)%h=y9_#t6f0;a54EEY<#`7`3wFdiq^7ctOt@~2gkz{Igi
z-K2odPQfxmGfhQOa6Fppd41CFT{8ZGS=fEA1v9~4hM5^7Q#GK1fZNC9PHT(mhrm+T
zaHg?YF%_b<woHmfN2g8PQaww}1060&_B0?#8g3vgRkZfPc21j*fmu~ab#XDC=MV7&
zx-36Cz9D6d=@tR~kIA@RB6Lv6uEQIGloDGPUvU?F3{Uq%%F@9uwp!LVQ4qLf@{T-<
z@?tBYxo@c)+qugdkWsE`7E!T&#;U>oPMTO<hF&g7+l<b+e^#C`5Ahv1O^1ZIdJd6h
zeg2zlnmIDX0RzPwo*{~Fr-C)!sW%RQ4KG_{(ox`J|HcyWcDJokdFAfG%qvLMnGNT~
z7RQuC12u9}DMqZ`L7C8f(CJ^d29q{p-nhW1BhhZ~={}BJpK=u8DMUuk?~6PZ-5S3o
zj#Hj6sp~@N({p7CEp8(1;7aPz=?tcI@MPu^Zg#jLZZtQ^77<ib+d!)rM`!w1Sf_Q9
zX?5I*@(E}#s4*`|lyS29JyRsWJPBJ~Q{KW=#?_33i7+H4icZ`dlfirxHd)}dp(up0
z&8%*gpou$Ve1TVIsXP>U@{{}+w+IlP$yw{g6yKDo{Oe45<?bn`GDFQn8(YVEJOZeO
ztIk;0pr&p~y||OpWzvK{Q%rqIO4eh+vzY+G!dLJbP3S}ip!tRs#SHf!pXr(T-@76j
z3=a;#v0DQAt_=-L(rZpmhyMlvj+d2sC5kmM%eqG8^o?4C3jpl%$iMm-f@0x+uu)lv
z2~O~sa#j9-PEmO!`0Eu;?*-7ezK8wza+IHi)(aYOAUMK~YVpt0XJDTq)ltLd>Q3U;
z%qh{Z{E9<WS|;M4b;ymO&UY6lrlN(X^&lL_pfR&h7v7Zcwz5o=*PxCDm$)JRD-k|y
zew-s&?7pVwsFlv`v85oslngXYl=F**`_M709OcU=KesWi6&g{S&m}kYD5p8dX8ikZ
zPahgsQS&~6Y~s<8VUp2>_F(Op+iX;HSDL4C)+kCHxRVxW&S~E%Eu_b?(&qRqU&{FC
zLZYP*tg51kt2z(ag=VNJE8#!1t7kF5sml2N*%DJ$Sy2SOXtffcxjf7fz6Yu;Jda1R
zJ29lh`(hG-*q%As3B33E+53FINkp8)t78GHY2B+2EYN1h-eYS8PPWMIES07WJD`hn
zYLQMqu%KA;^Q%{{{Gts1c*sUJ#90lje70H0Ci^YHgH3x&)rS5ZRO4+MY|7OGBbsN4
zxfx)5LF~L!wd1HUt)H?Xbqunp+9l`cgGOcw7LQRXWnprzn{mZfDUpoHYa6jEG^W*H
z@wa2fq-n#kYtIs`Sj^K{EDPvqBDF|hPme_9l4nXbBa~MJ1zhyQ+@xKGC@9=ojT8o9
z*rMS_asyIvMG<e6veze^g~r!;kho+2B={a;bX2vgD_xrnTXKy#RwXM|XT3Dj!OBL7
z>gOloCM@Bt^_sB{^2L*pwQWJNdgl!t%U#kFV}1}1DZqWrcjz=Mv!Beka!$PPEsA02
zlqA+2yG1-!`5@1|lqT?KA-l{oS3^th{+7S4ZmkX$+nmHmt@ysuqh4@Ax>%_n`*;#s
zra57fkSLfTC{V;4YT#l`ASGdAU?^e1rv-HRNGNw8I`l@Q6qI9u8ACx86DTC52s>~g
zCW<62*?6~}j8xNL6M5Bf(sOZnvHBlMT<;{h<~XQY*cc;LMBt6{@!5uQQ?6FVgl+X>
z$+(k_gNT=qO1+Q_gKqWtc-d&)fT}K&j4~X4DoLd>ZXhy5IS<7^0RxWH^lSheaj`mx
zBx!bUj)-Elt^7IL=8%}7*A03B-?2XyruSZR_U(Gix9q_@pR%)-w|TK)?b}k;@BE^{
zMXRi8#9QX*1T)5W3WrZw^9I83Z`*%|I<}fbj(UZw9{WFX4nC)U>_5zAHT~M2?X}DQ
zLi=@=-No0r5>`6MB<q*3IQ*L5%M}+NisL$Fcw%MmiBlK2FDbX`q2J0->U9DXy~(OP
zTe(oF6&h<63BFOC$c?U2RD2xb0cC0zc@b~68N8lWiyZ3H*e=W~{9F1TN=NzI0V%@%
zcQ-%h!SEX<?(%+yO^`!>@?Ya0!{Bw7a9WGqmWe0qPU5$x4>;n~9ji>6X*@NSlp%+&
z9E1;dvAe!7I_G~1>4IH<Qv7RQMaA^#dvB}3&h|-1$b({*am0M>U+a=}+zfWY8f4v!
zay?RR0&E~%f~^aEB-EQ0YUDAdCMwFx$*Z#>Ij-VUF4dr&%LEcmCo&9bowjTLNY~;2
zE_Joxb=AIY;db?Y*LI8g9ZUGcqu--X+po%N=a#*yg_pP$;fhJhc0+rGA(xKx&syPl
zNVctjldvPSrnKXF*k&#p^htKx^NMw5FJIKI<MRIZ`L1s;V=$xh3k!LZ`<Dqtyw^&w
z)7}a#L7QpWrgN%!yK4KFL-;aakD?#oI$`tVhNHu(9QhNjgVPy7n)#MkcZ$ebvxf7*
z7DpwJS_f!i6sXXrQ@)HBVUL)=0Dqt-npdr~<T_I~OHI%&R4Jq;qezzx92o!X9rDQi
ziH3e~^ExJ*V%#ZdeE#3(xy_+CN}mb(vMoI8v`x=1&GA1j+aYaVnvBaP>`1QhgjSg5
zS?4;8xmQ_?x%Fxip0urIn3R7FU8?=<+F0DUV*NbxZ|+6CXMXh(C2w*}nsI2|^WN?|
z9^e*m_TJul=-M3giVnK4?+nas<NC~T*C1@nRRd5IUJ(xt3`A=wH&|*G5mE5`*ueKp
zHi;ic&tI|5OuEEt{|wLQRYr33q8s4h1{b7ACTeM90e{FFLOa5tw@g{#CN>@v9Id_R
z9mL7Z*WRY3kC3(yvX3E2*{oYz!f;M-x!uorTVYJ`v|s&QN$M-kLTTIR`Zmkkr<);e
zo8R+`R{R&AvVITV50rJzwv7?wf$IdwNdGtL-ZHAqFxvMF8d@Mgixmmh7K#RUx0d2w
z+)8nGFBF0lLQ8RXw*rOY5GYWb;10n(xOZ~y+<Ruttaa}?pJwiSjJ(OSlIPv_?EU`*
zYF$zEwEP3C9gd$QUCvSS58m;$#I3gf`b)ij*vk2)5qD(tXD`ma+w7(R&XvSjo0ai*
zhGdJzr5Ut*jC>JG@jUwpo<&$|^@Lvx*Mx|rNtej#M9MWAR{=3usMA|+=|J2@74|JW
zY}gu!#tk=p6rfWDyLeSE%YC?=_{B`X(#qHVjw&AsJn7o;1x8UB3D)^vNo-9YDoxoJ
z>n=5iy=Cnmsy810Fp}T6TLq0Urz`9$-$=3?XBQG?wg>N4dCdm(0gvz=Dc;TPIFs1a
z;q_kNPK~r$)Z){p-dXR~C=%uv1WEmjxG@#o#=zYAdl|xCKFRn?Gt~Enr!G2Gi4~I|
zqaXhF=Cf~JC992sp{NMY$$H-t;SY1OBt^IcEZ<~Is2^3IA0ISJRM*b5C0Gj#S_x!L
zF*=qF)U~M9T`g4&TQ8EQ%YWZ(b4a?Ai^8;0e}9Ac_%_<8BbA0{NQ3Y?H_VUWaKX}j
z-YebmSGDy4Vgo(g`8d4N$e)C;CF%WCXW(r3bIRhIL;z2ZMyjUuJVUDBW@J7@mVDzi
z$5<2HQ`WyXtdT_?Zsq2=rtb@PGpXMfso7~PNf)qUcSq^Qvb-;1c>7X`IL1@6NUZdn
z#ky}qEt|CX)!)MRi}5LEQ8)dNr7Nze?_QjNH}#^5KNlR$=AY;tO`7Q5HSg3dw!AZI
zoUI!CgK%ubJAN_KJ7rc27Z_;!R4dOqVACLP6K9^u9>udB<<|jET`*)^o!f`ndMPZp
z8aSz>wXjo!Zwg#CaZFQx-1Pq7Ai%QLvbm&}`D=zWT|Z*nA>w1zu+`18%_v@sOz_U(
zq9Es8j))@w%`xRWJ{KJPDi`5B9lMrytAyga%U>X&{JyK#OAfKaF{q-Hkwwg7$kddF
z^*(}s%zE%kmh=7eV~ws@$8~f|FC*P*X-|z}`oKI8ji_Bw%Y`uy#K>=l1~jvEOUQ9|
zVG}!BYAMU9cD6~T6TAm0riQPGBrh)yS=r`vbyjhy0=kz;n%~`uK+a+-Kx^FisL9`s
zF-?k;w7te^C8Tz&3P~Eua<A~W_QGY2wVq&M+73`&n~STYDvNwondpC(s98Gs<oS<G
zc1ZEhy~dB|=v#|{kp~PEdrRhC`m%;oo>owGw<y4PAO6(B*2om3xwE%}tX=|L+-DPA
z`v)LM9;hipRNs~ENl%oZsMK^$hNFA_LcR$FYd$_nO?#R4VZuS`UD1FQbHYF*E4-fj
z;I}0bLn2jqlyZ{P{Z$IK1%s@<y*jT`{@%5X%sT+?OMcqaTEFQOFuNVKy?H?bwJ~wH
z{o?c76H!f6`Iou0K8@)VijsPc;P+38Y(v2nZ@$iGxcmc{M6+tfa*IVKNVBP`Z7o$G
zB7nP7q|KkNN4hv;*|Bj>da>6TBgALCBbw4`IKo!130)N+-`@sGp<_Nc&rQ{-N9=@y
z?2Hi&LxHU^!}kw*qz!&uQ|^5Ey*gy{IyhG)7d}M|Isqw6B{BW|1@-qb6X$b1pFecv
zkTkh(m@CV0ex|af32{lJ{5l*K>_&+`W8W$J!`>gh`1QT|m_@w_xvN&o=yk4817x5E
z0p1xeG?q1b^?ve#Cyr%@r^wpL4|_>OyKHxFa<v(|VsNaGIWJp_jHm4tH{M@>C_gSu
z>ClG~P}@e)YIii|&PhR252N_)I+i3m9-s}M$_2kb8h#Y%D~?MxysIn~Og0u5`y&WF
z<9NxyviS+lnhFXR=vC!rE0m}**DCSLq%2aJI*=`y*|GRw^<619LwmyAQv39m%`$|J
zNof-w(9Re2!&d7_Bv05fx{lh-P2Cct)YOweVz=5|Z4k{aCCP^8;?11nP$OKoKG<K!
z8Tq3Pjb!*8uItCRcAijRzx>ipm9fOwqj+j0T`P4GWo*TtqN6reHXBy%`6ok8?LO%Z
zcDHt3N%6;--g51=p069b<|IR3wD)~wJW4`DH|(D<Gjri%UTYImCY2|znneu&)Ojbj
z3;|PB&ck81<%*C}T}>ZUw~8$lc4Gby)kF_1R+l-k_;=bb*9Z9JSaTapbN*^SP`)2h
z4nQjlzMCk~Huw@D9boeHyIRV-=Ryb0q4}vw$vr6S8~+4G<}-O72+lkDG>T7_MC|$*
z6gGvuHw={B7%xS6T{&W<R^1HKf%+dS>$3~43eQOP>kEFrqV3goz=(aS4cB~4fUGde
zZSv8eFEd0uS<jD}dirS<N8?MW1Bp)Ja|xFNHL+J-0d0Ln>F!K`4?Ft2-ME-3zgiXO
zHwGD>$(iq~i+RWLe<Gtb;%xFT%W-W8F#sxv2#?k%lo~MXOzBl%Lu$j{nAl(#Gl=I6
zrlB<UKBy>viSl&#WVo5=^1NIK=O5rT_ilay@9R@1xm2yhq~SlAT>E^T${LYbRJ7yw
z(#s8m#Ml5Azei)UFYPe04=4Op?9$_X{G&}>ee#;{Bq2W$m#ojZ8y9BDRIIlZ3eD=W
z#^gWPcbmO3iU-s_C2mB{WCwI^go|jQscrhy!%EFMfI_jCn(nGBsxig{O7vYc8tS-9
zd{h=fX-|Qk7VPE#WDIa9P@-^_OTC-8B@xch_{kJJa!7@jxr@9+{7KE2PFBB_b_vpg
zWq>RU@o<Y=p1ii_kcfHtT`WB{Baw}%)K)gM@ZycActE?RUl5N=R$<|HYrA)E-UyvV
zyyPk3c{bu}n+*1LLoTLl<%H6nU=y#U<+Qju)AT&C?(cx~(6Slmk!_J_yxwKmWl=Ii
z3sI!}=<+?HB00Zo7>jP0;5VG|d>UI)a$(1ugB$+7pRW%T746ri&GDmPYBNYqJttL&
zKO*^tG?_JlQRI4s{~1fqq^mRmNy^)o5d^BtxcTiM{z4XA>{oC-L5s<rgyg3>OK2m~
z<oEWxV71G1A_SWEKxcIUxguL)B)cM|^zLyxCiy|jFZ1M?XV%lV;S%opn&I|!R1d^O
zS<T52<*9N-#I8T=N`22r4+{Q@eB4p2O+Rt!R+b5+UCU&k9B&{vCo8L!*`Ra>KyW&~
zthx|IFsm|8WyKLtB%BAC<%62U)!ta^67T$d`xi6Tm~TfqBG0%oG5Tq&24VE30;$mV
zC%aR7UGK@*b@+IBd*9{0;mFS7&HNE(&Ku&5jN{tkn#j@y0^Sb3)W$&3l~zyf>zaHx
zx6;&X)ij^k3$MAg2^L9W?4f(ltR#+`u1C?|$&>z)vqU>7Sxb16#OiC0udS?5&k>bk
z=8wb}UR)D9zWkr4A_kGWW6{_01Oidk^lveCY<j-QaPuK!g$hH>P9Kac<R#KB^q#g+
zhSw|S6KQwnoj=Lj12fwg9SXcxLlV16tGqnvE_oW{r}+-}9K!sSqUa*YXmwpb^KYiO
z@)8`v1~?VdguWjvWc)48H7^|Rq#&nCIi@~I%rackGf|Mt#z^_O(CCzNZ6(K=BVMRl
zrA&poSC`;c!G25dqyze-!b~H>U2~o#t(T1Z-Q-`N&?=9Ks<nd12Qp<l4&HFx$w-la
zvSQ%m>fTdF_3%I~_Oghwkke-6X%dH)jn^3;d$BcNby7VCECC|0pI!8{>;M6TdT-ab
z<y62;GlD(VM<5AG&e5p?V~;<_iqoub7$0N0;AHweg!nzLUdm}ThiN`5pB?z!;K!l#
zt@x=(WN{&gO|w^4;**emkdG>i*Es5l;iJ=9YHfbqGssJBIKe7oNhboMl5JqVspg}g
z1w=1Q(&U8@0xNW$@@Z+xUw*j>CGIFA5>A&paH2YPnyatAwr2v-AfMl#a>l>$yOH7A
zOZ-Y@DpL9>%sMXFI;?9YQ$L^J$$jsc`0{O&fRIqm>iL$GBsU|+P6|oF8K4~ixSrap
zAeK8+dYU%bbP`|~p~zXM>8BENOE$4N!DUF>{XHpfZIBA(0#=Brd{_2c>aP`X@A>Q4
zkzCdnx4*GVOXa3`%vvjU^(3E>=2EG5Zv4UYb6pQbQZNtF!Cq5ZCUUYAIi<QJe~x;!
zUihIjRyANjJ^HBO{paW3xtW=JXi%QNZVg-#;>ccofN$~mahoxJtTi!IwI_eNckW(y
zs-5xVtB78~&DXz5MX9Bvio^_}T;YCwxSCb*%+EG}Wra^e_DIJP>3$Fo7b`a~nnfGR
zKR!E|2uJ1XEIgC?<e<y=<+nqxeuJT8`HsR%vg$rTxfiU*=!x2=LY(kuNBTxsNLeoQ
zcI!*Q{%)%Fv-hNb`1Bi;v^2)}U#Mp#gkgYvBY8I0@2?Nd-@L-qwaCwcr-){~WEB+9
zNGupmK4Xdu6$}BacZ%AqPQzAR1X%C|QhpItS7j+TJkU`p7xDEyBewg<zhedqiJHS^
ztjV=B+56~O$c9sZs-K()qC_-1#1~$u{!%V|-iI<u;pH}F8Bu@bHu2`h=G8~2d&H;=
zhgv%BfoOqz8!@TNBm)`;2A*;)8zNTNmI84NV#WL-vHp|86*U$$9uU`6^mGT^?$Pyt
zEsrc@&8x7sXU)15iL0{P_`;husIfg(N6aRo)60ik=j$KmW@E<7G`QO=be4XV=B8_*
zX|05vD&~A7OOoSVZe?Ry-KIJu+S6nIgbxQ5$RZSNu4LBIs8M44PRk?9Jm#5ZX$tR;
zxMZ`t=jXEz#>AVFAYuM@r8;ax3Z*<g@2<C8CK18-MI_}y>rMt50vhyd0q=N)Z-f(`
zUGwG=hhmWsY<#s-ZKtPTqOk6j*$AQS3V1>Q53}qeCLh{0=ngdP{RJ+$NumoE`V`3y
zUY`%OjEJV6R>;X#m@5237d!GgqtH$s^85Mv(9brBeDNgo0(mx1be?UvZClHI(Nchh
zx4lvU^tCC5c|BL+_W)gNtkz=ufF#j>isdbQJBqvEchU7%BjzZc^IHJndT7&{yjmy~
z+4!D(BpFREzji6<_k8WWV}tT}a)s0(<^*|dW4roS?-5F_^xErrRM*$Hq^Hx?V$Uqj
z?Z3P8;%>}k(nzwypSZondUgTCWYfs1;?s|y%$(J%vY~u7#jx14N(6b7^f35~9DV+P
zGueTm?+M5LiJ7kV!9i=h@KW5aSGDllZm30^_M*)6fKrpwm%KL(?<$>)wejSDN_$q>
zuSHeq@d~HZ3nsrMRuyM{7WoIz|4QGgsEH%pX~VrqqJl_tU8d9~C-b|Wi6Gi(mu#kw
zN9}VyVLKG-CJ7SRnSN1I`sQ=mw=ZYWU*1L<cjwteI;CsO7t0fo?lrsp$Sx_7<w8r?
z_e3YQ#+04^jK**aQo?Ad{kr-4Zz>>yIhR}h`L~!oHiIBt&;>JYtt<(GBB$>&pU}0P
zi{VkqY6`x~i;0g)HJjrO$vY2dEGgxlCr{x{)NP-Nj!s-lB`K*)`RfF=DZ1$Gy|Q4H
zcTtmUQ9M@m#hzVi0=&$7XJp)N&lEJV2kuizUpvIV6K(0~VPSZQ@uW(Xay;y5S~&r;
zm*CLnEI_Cs6$~<*chQ00-|&h9&Pl!dJp)Yx$;wnWkU*Q-_rv|Foy=nIv^Lf-y0qV{
z|Bm`O(EsXev+h}(a<a-$zp^ME<6p&C&U>M1i#PGT?N~cg)`%Hs7*6dIs3!_zgmb&l
z)lK%o7^B*z!cYVJ{fawN9?~5c{a0A)L@W(1v&5ic=LnXW?2*?j=5Mpq_`a9_<mqa2
z(UUMM?o*3gweJLfeU)5%3~USqd8|}k(QjnPI*ln3{f)qs!*&tEd}}?lr!PP`^ZJ<u
zx9Aw6^V19!1(t}tpB{?@Gk4uMp@~g{U_NBSRotL-iYhf+)|yoWm#MT@xlp@s|6`f}
zzp{{^3If4Ws@ZLMV*{-j-uMezPi#f;eSj8QG4yS|c90Z#T_BnD>^?=*eo~}}G6ZCF
z{neZy^)r7i`^8pcJJShD^H!x<h?_3Fd0b)YTFOA@OH9<uTBS{IrR3U=i7ZW}uMWlZ
zQ!+K=KfY%oRkr&oV`PxSde(238p>Fg8)E$O9;H30{66}L);?wI8ht!BiBJ(XiGmAH
zA*Ft<k2KYTP=hLyD*kW)<LWEc_%eOBY=Iq2+3|TasgcZM+OJ?e;Ewn_qdS~_dxB(B
znUJ-OAVN)x^+8{cmcU{q!gij~E9vD{0g?6JC%Wa@^{rp#zY~4q@xa83<dRij*l{h^
zGIDtx6CX-1<?S8`r^WXX6&(-9{Rh}(g_9}vL@yVDRnKK6dn1^}HCcP6iEFpN5ZG$i
z3AKc73}_>M?8>T>0~?-xp5jw~CZ7_<8IYirTdpb3$*m1}hHLl668MS0lhvbbRyu~^
zhLHqVlmoG_p%HSTCm_+y*^+-pXdp!2Lc<t9DQq&l_Tr84Sd$JsjhB>lm+Xk?*`zC`
zqQNvy9hM|(PD%K{?u(Bv(I{z4T+_AG;^MVls6@a)eK|B?L^-|tN1~hOoR5i^hO;7>
ztvV~banJ<A;|m|-7aIH6prr(z*+!tKtBpSIGbBjuYLrWy!ye{*KnE`B9y@&%S*08-
zU+&`VG!-KiOe)L^EA;m=DCXOvE35+rFiKPvn8Jy;ay~5m8u=X#Etja2<5}sd|J+Du
zziaX#+SjbQBeB(Iu+ej{ul<3yJ82TLwdD*HxuiiE4ODTMjfI;jd=X)1B&P;dzQcVs
zg5~EuPVP1FUsRY!$%iN)gTbBWkcclZ1jV}jFuB)k=YK*5&0+V-h9sN)w{*cTO+ENP
zITwOIAT-dtRaddmvLtllAmx9lOYr}jx4AZ#->~jDpXr-!$~EeoQ#&0vlLw<<7mB4O
z+OSp0eLAUjw$F)Iu<}2S>%Ab~$!kz0{2Z(9rt&=u57!jdQmwF+OQaNMd}`2V5ht5j
z*(a%gWiuW3-ZBZ}Za`&Tu-sog({50-_R!&cpKw)z!=b@3C9FMe&inQR$oTv-mD!@}
zp34|F0-`F0VHO~BJAZ%gM9b^YV;*xI!hae78H5A<a=wHbCRwt(&Sd`*+%lmXZ!2mf
zp}8RR-cGrQO|Mx3KnO!DRd0B@aL|f;CkBe^ZDvioML|4{LjCSur-*E+$=BjY1X%$I
z_0dYU*{4C{hd>u%{F%-=mw-UBCCMt~+1<FHL3W|)7JD&zibbZCS@DNTeBvHVv~)hK
zjF)>*qqhrOuC=*Bg0Vm{8X0+E-wD*sPWlR5p(7A$C5|kQ1;!t+5qvtWg9p`i1er_^
zK&)|m2rh*D%9{;oKxBPfbU*(vauN2pBQq~9j&JoDL(HY~$q#OA<sdz-$L9|;6(c<q
z9zpOKXK%U}4X>Ze1T3_lYk4swM;=by*RA9Qn5=nMpU(;b0RWq~m|xxOe9H-5riE4c
zJ@`*)TeeE(`z-<#%VBKc==MS%ULdP~;M5ZR#Annf=MIp<;BB>|$uZ$^tD>5!pWQr(
z^gvDit>j+OZV9a&9r5CwUxnbWFelo^waN#MVycHg{sBQ#F;B7})%$DiP5SGVgE|B^
zb$AtZ#9gBsn4B_~=GykM^iSHGX8S$cEb(@};RXz{Z8u<#(EZ_BF=v4DWFIB@rYcGW
zvC<w}>BFz<%_MeYb`lbohsiJ)ceJ9<$37%4R=bujytba+YF`v5w5<!y{dflb!W*FI
z^@|!s(l0wL=5E<-NQmi_DOwzugy@OBn{PjXau3ezV(ek8<TeB2^t~P+>S1xx<r|{#
z=&>EaCgAw>KLG8i*2P+8pw1Hgr|lF{i44?-%;>W`Orl8JTL!?oSY<W7$OM!|Xw^tu
z{8VKEGSef_>{<Er@4=_jn{_59Cf6p{w%fWUh=5VEZPlA*wP58|!*MGw_V*hnAxCxH
zFOF#at5?eLfTV!1q*RMPiCzd;+Q7D}tdFhzK}(72^yOeMSeZT>Ma=WU`)W#=$M04E
zy#E}P<<IR0ALPIi7j&7)a?Bi~9c*txsNo)zYn6u$WZY03#VB({t_VUa+(EhLhUKL>
z(InrF`&wNMAZZ|~B-gd5D^!Wt#TSh2N^^X?$VF`570JTfMSjkkPY4+i7uI^2AD<3q
z&0IHSw8M_EIyqZUKnHz>*DIqOl*;VGisd`kPam2~o8?V05_v|kF2u-PA|G4SK4twz
z%@{v&&eC&ZVQpNWoa%XFx&oCwyben|bt#sKhqs{-mPxszO^Om604xx?I(hOmUX4)g
zHA4*&`Mq#?UHovGG8dy}v?ITQ62BwMvGn$i<z^wv2U{=%Cf;Rs!jRzD6~MiggubzC
zm3_}t+)0|wX1rm5u^!+;2O=%+S<oZp!?d31XjaQlRJGZHLj1Q7*L00$WKSg}r52q@
zuqsq1bM_cub(ywZlwa)3@sNC)hTt4s`WQ(#Dq&JTUAp<j4Ok(3eD1pB^dM*-@Kn7`
zXboOPGh*kI#gqkJ3a|fRBWKron_W6tu<|5pVv69l64&fC?ssSANnBg_k1wdO@0({6
z-|LW&3`rsOAnj@F`*>lobNJ7ZVVPtY@SHul7hObmW_uMqMViik{3+S0wIK49B4O1q
z=Q!sSr?$0K6Y`6!O#-V8&w41w4sdJ0hrGNB?o_`t-J+ABT5qhiHD7fTVb!maHZCwM
zG9cg}S_>!8khjhj(YOia{>}3m5U+dxWV?^sM@|7*J@Epk(eQoj!dV|$3;V*v-ORT!
z@U}utrJd<g%{wwe+6)<-FA7Tk0w??sHQwSV{7YVjHgmq6>rlOZoT*i`2Ww}PgSRoJ
zzF6$6JUhp)X0y_X^whxhtH`G;BVa4vU^emA_@68?XH(p9U1m;{{(Vj>`8-L}@uk~R
zGO|B{Bm^t|_+0Od)<?LnEQA^)i{|_onO%$%yd)qBeEj{<i*4jFQZcs2zYUxzGq>B-
zqf_#xZ(uMJ-!jQCA}DBYjcoNFKpTwub+QuB0S+Y`^`q5oBnp7G;KVbwW4ZYFW4k+~
z<O)mj=VDCN!Z}0{{6xKhhbG`IVuBFQQRTa{Ix~onc?fCw)k2y1mY;+Xjty3Felw!=
zBr-Cjt(t@oASxC3K~QDKrPIw>K7l|VxxL>zg@Yow>AterO*~ym=2)Mj4q2VshKZRb
zXX%}RJ*V^kWOS7(TC`#~)&klN>zT6JbSD;|mAOlt?)gI7(mC+F918f(i@MH)7zBBj
zl^5??KHN!0V5X71(%IfB#2_uMZS&|{W|ofAqsKnG4jkZK*DW~dBvJJ!vL9~{JDwb&
zw$gz^@G|=3xF}X5S9_k8tNj*6R3J0&;*`gtw$cvHASZa~y(}gXZb6Yfx{VidG~eU{
zs%ZLETFrU4m5&xthIqwl7Tc2sNC!wMOBymEQmGX9CEi$34pNd1waHmkhDLo8{7Tfv
zN&mG_AYGXITm0E4!tLB-dHxqGpQFY8PBj{b9aB_OnI{63|7t9i^+<TqCQZ5Ngw4&J
zKZ@2xD6zr)&755cX9yZk8$Y%QI(+YL5*Z7>6}oitU7O@^QKSfwWcOXg*XI-<xV>3)
zKd)c@VqSjXbi0#YxO-ULSX9DoLfrTxXi*?*5_hdf(9a<3NjX*|g0t0^Wq3c6bt-Qk
z{0ia@r7=-r+T;d&)nYZ4mRFZo8Y<Yk_*j7Zs<R}SJz4#IpFlo8ZgyJwTPkU04lk?v
zEtj6=)A)7SA5gOG6@Lq%TtbkQt+FgCSH1h3-H$j>UR=Tcn+*prpVFjYmz4tA?2?X7
z4v9E;IaB2d3oWELzrUC(`Gc#4(0Z;>^fe9?J7jIG90!Xl{;E3iwdgegp)xbGWpCV-
z@VDpW=c?~cy221V^$f03Q{Gdfh(aP5#20?+*pAVSbQF%<T|dNDM}B4(z0IOYToiuj
zFnZH#PMQCE4Ik7Zy9BS33|xNEbPN>((M&E0bFZb_-}(sYXGfmwopm!MSun*LiJo3U
z$s_0q#cY=e$F@qoWcisJ2!30w^;N83*aSCpUr|FVc=ra`4@L*NGlDM8@Ww6SvJoj-
zA;n?r7I&RYXA$tsI#BM!^h(S0yi35(LKiKM+lW!Y(~Du9nJ?HtyJ^~)<H4Df$c4(H
zbEfd1lyPHtRr7{n`+N0+Pa!-$?lgbaDZpLpV%Gx7>P`7yA*ACqj#^$&jr^O4pnH`F
zKIy47unePUn+zkP@b>;BBTd=dLmT<~)tl8vnhyEmwjOMg`F-NSddHv~0%*F7IX2Bo
zNT1OKDz|KTA1%j0G!{YGppUUnFFcP-(U*ZcjU7(95rb8{bNM999{kH9$SXy*xShq8
znkr7#@zYF=DnxF54R?o&uE&1;zFJRD$>2V@@Cz!}1KDmnowT3L;o;%KJ2;#$v8EZ4
zq^-vop#Tsm5xh@60i>uD*P*(DMpDfB=ToZXmx@WQZIW=Q*gj<x$+G2)O%T$^exIoL
zp@JycPQPZD49dKXkbF{3x?OZED&okqQHMBIdMC=kfN7u$vf!5I`=de?zkv(lE=5R^
zN!?ebBUL?DdseAnJ!g79d^mnG9^Hj3^qH2{uXdnMf|-W%&C{<|c2tF6ttSE7ef91K
z>mrbNAyx*@LRvXF&%wCkKf;@>9+usFG!k>KcaL|gBl5euh-D!@GILf;4!_+lBksdq
zcH6miw=b3^Q*<Atx$e>Jj*<%j*_adkB{M_~j@$K!S9${11?mKIDSf>Wy$K!}s<aS?
z0IaPkHWtNTw)T^SNEmW(E?dX~N?y%X^r=w=F%)=GMVTaAF`^jY8F|rYEiiI*u^_%w
zDIM0k(mo3=bINE{2;^O>B-t5_{7AC8fn|SjLhD#fdw>gE_g!~gLtAihh(5MlV4m==
zN#X=F`n|-v8gdnB+UK*^V~t`Stcx*gsJXiI*Dtq_EWy}5$2TTY-y3Z~q<`Ee0=yfC
zEjizehK%I6_D)xh*6!2g#;>XtWP`egOxIp~HYF>i2G|d>L)m}sIdaH#({VZkWs*ib
zc3`Y{aVi;BP=ak@>lGLUn7XcS^g6&SHEktJ*{g+;26;mZ|D^$=I=Yi=B5nO_jHbAQ
zHkJ5bChO7F>%6}Ulpxu0lK+uf!RfBkha*onDdIi;|CfJt7jW9`EU$Fdv=PZ&oj5-m
z*I)Yw(7Anntg|t|fq^cC-GYGQ0Y1`Z2q=eQb#IuolfzA*%K=AZI!UH)hM&1!r<tx9
zzU{dyBDw8%3a<9Cnzg!uDG^ZzyYa~&`{#O$_D&H(xR1{$hwm@!mzz_bd{X{dznZpN
zs)LRQD@RVm0{p&C3x=&5el854gJI^|IF_uS9;oVE8Yzz2M95W8eW2ZCeht@AN#|IM
zqP0oiyVuuaH$GcsXVV0Iv5k#V)KSOnth5$6b_6CVV-FGKc(G@rGfPN;R&wJ#H3Jxz
zh}A{chpx+Gw6VRg?lots%ft~vhP!=`&3tGx>2gRhO`kkL6iOtrsSZb>fe&qj{6W%&
zcGH92dLZ!9?THP81*8}n<z$dyw3Ch;G<FMqp3l=HL2JT;+INJEp_RLRRA0u)A(DNw
zpDcJ59f-U>33zX_qIGMYP_jS-9%}QlZKhr5LLGOZw!K3}GTJ-rM6LdUgV}s$V33P2
zmBkQ8Rhi%ReNzNddQ=%V*EAwJmK0&tDa&_Pxc<K*PL}71KgYUO2VR%96@i4aRQhNF
zT=cVifGVO=*q*2&M{k=UU%m7PZEoFI#XgA_k{q7BQsU0khgo;lVp2Xwb%KgiZALft
z)qKmV5bV9TyB9_J;Vw~w!Ry;I_E#Ge`%6K4LOVuwVm|9qPC(_5e*o@fj8Vv_k8_=B
zd%*2u8Y?5s2#sMb*+{!pLSXwyRaL;vgRsTsSNf9|NF#qdOGLiMDY&vb^x{r&5iHug
zatQ3hDAGX)Xbg--NcRS-63x8Z6k5bqfM-{D4rW_l?wS9|hm#Ky3?kG%w0d5mp5A#=
zZ#yxxa3e%ffSKRFQ{Hdfo36yZk#9q4CDe4MhFz9VO?92Loo`We<U{GCj0|f3(uuQ8
z)u09`EdpwW??ul)NIzl?xiy^a*$pSqoba#rzhg3JA*O3}+s<>0GpRUhbZY(Ce(B|Z
zTM;_k)v|igok8koB<Oi*JJQt$ap=WA5jry@adU=1P-kzlzuptpFVZYfLywQdlNwXZ
zSFhW@{sSnMzR1D3I&LiPbhZDzVyT*;>-}3?<RYJH=Ew=O40mZe9ZkF@qPc6_SQNkX
zpR!$=Z{^SNa^vBr^X><T-rZpZwZqg+oWLhn$Moq#N4yeNcXYVD?^XH}0=!q6-B6?D
zZB=Ws<OHC?GQ%TejJqcJue+mL@2z=a{&yRLujysAwUo>k!1A*89=0p&Vo4@UMk2VB
zmXUOUPe1zr99MY(mo+lL=z{OECt~$N-%|P#$r%D1;n|!khHi`;EPThKRru1Gi`SQv
z0Yax5zr-9@v5aRHU)ax%$3ZW;Z0`<h&h@;Q(2R-QW8}_7oyYFgD2>>pYsgHYM~@yM
zxm18-&@weNuH}s8i<DxFgN)}E$nUIYhIa}?r06_|{9G<LhMBwbJFYFOx<L3sP(IY|
z{rL&OMRx;cJ3IGRX_PA+#i#<yO#}Ydk?{T4_m}&WZGra>RC_dSR5;<~bOMJh!mRUp
zV(FY2w&8@VAtzUEM>Q}GK4g*#Ddl#O$Yzngxxsjb!9#|eNTt0Mxefe$W4Omf;)j8I
z78NqxN|6~$ci76s|7=v8r`_+E25PF0yPqAP=A?t2EhffqA#PHI8Seugb4L{=Z?7uI
zV<mSsW|PUO<+p?9Z$Cu{V_bDmyF=uJPIWGdB?H6|j$2OhdCs7_{UYBAHN5S?kPCN0
zWaV#XawqV;`2Dv(P3&H4)p6j5OGhWy<NZnYV;cy=&ZAk~Zg0ZWEY{)W?T6n(C#cK8
zVQfn={yRN272}Wi$HX+}gC?zY$OJ*5tC6hegnR){xPcpLwF2R?d44z|#;-WlA`x0W
zFpEQ7heDQ6Me^^i(6PG~EHeW!V(ry9M}YjXYZwS`J<isqqxQ=2w~?J62Qo`Xxfj?i
z1i<abeFq4!w2;-4!Tr^rQpoe}l!Pq)?p&_x#_KD5IpP^L<2T!b&*fDr9>CuYC*bg8
zEPu8(BOmpeAs-ExNB{mk@Q>^fh6AOaWNA5Mb1e?Tfd<9D<^JGPE^=w$wLLO({?XoP
zt6v&3KhBrtViH5<3o<TKYLvaqry_@oh~R9UsSxUNe2aN#-(#s|qU|S<_7$KWZ3c2W
zxSeYA^!5xahH9V|=O8^E;-|E6N=;;2_%3VETK06=6g(7aQhaSH{tHfRUk43tR<+>g
zf%5|xKjy{Ne&A!Aubo(B4nG8B61428O;O1P9i#C*sU+tf4c2X>siW2h1|6g}3Hx^B
zRrk@bD8#{^m-Qn1V6xJ{-DN5oUGJ=rlaW;0QUaUDiZ<cvb+kRL4!qJEh+K8?`vp}!
z>;L%;jZsFOZ%&SN@)QS-L3XPC0VIr3Lo*SqM^MM+h^6MZeSK_xd^?1@k)-MKZ=|b(
zW??Ih<2WmUVGx=Y%o``<{EU12CKOTqJk(Rn_jj=}+xI_2fVf9dcre6~63cAp!wS|+
ztFB;~ovH_6HdFG%Z=Ca$l>^%*p{oJGyiX2(qUfR3N;j-7b7bOgnAq9YrLkdnGY_IO
zi{9meZ~FLYJLkMdX;(Il_95Yo3*pV?%v{dii%(%s`~XXhbnSd)xG%o@A`QO9oK}Ta
zRbdJo5ZnRFVSt&}PT9D?J{5Onkd3O9HJsJr$Tog!lKm&vG{MbBGlYcqF`}V?;H7Az
zvO<7g(K&}==^(KRU7JsZ#fkF<SUl)5k46ahc;N`OYoP5)<7OFm^}G|<bJi(7-9daj
zSR(s7f;doga6m9*?}h>y3v%2!&sP~?&JKFqZS^?mssqbP0bkKeh-dLOIoX&YiWjqP
zSQ)3zgM`O_jxvl}c{r~e5_SvQH$Qga@y`ra1*>71Ds0%ScyI8m{|s5kCJcP!Tb_e_
zt5!rDkE0`m;X#Rj2%Pei5AG{_V?tIENweWCCEamE#k3#rUX_nV5p`~l4yY@11896v
zBr`Mq^GL_TR`PeGmN72*$YLy+Su0&V%*!QbnudeD#h%{!FOsS$o8de!_RX8c^Y<E&
z4)R5<Ze*`;Ti1m?NeQn9=(bb0HuO`A#7joUpOQ+djT@tLNOtbNmF~EP_@yy^z)YW^
z2S%9p3eLBiJZ9BD<t&jVNPnElPDSp)?3jkPgK1R0ZvZtPVxAh-Glb04UC242YxcEZ
ztTcA6!1#;0doZ2$hjg3&3U)@o^6~lFUNmfX;lK&qsu%O18GSfuCXD@mBM}KO_PE82
zQ!Eo={3l~JR{MX;q5c01y^*~_Mq`p~xt<RRw}!#NT0V6hU;kr#|CpnZ@gKl2E^HKD
zWy<<I)n!?r@l%Jt`Ve(;N0w4W=ME4_-lLw;9yanrXEM^%Sd#p4c=}v`ki0RQjv<aR
zn;`TLC!TFKjK&}Y^r}^&eV-$DFgXZRGTG^SwVwC6?dMcJH|WiZ>P`vAs7Y6^IWh54
zG3Oz5B?K9^#f|mL2}dGBT$?-GhQBO&fy)Jj46`~q*;rPdgo`V&h2>HAr8sRnQ6Q`9
zuw#J)KR1!10=-@OT&>PU$5izk@ec0C*qY;LxR5-S|3_qB9D}U5|8W~(u)==iaa0e2
zGUO40DC_Fy8}hMUz_#5I+ec;{hdxskKONVWS-e0Yr_5<BUZi-d1(Ej7vo0*`HaZ>Z
z_eA8t84_HrbZg?)9F9VciTlI?Vzzdw_VZoZkAl!uNIO8_72m#rmkK^aG;pyxxNl{q
zeVQsbsxHS_^-LSn1xIx>=>Eoy;L5MjxGqr*Xb$C(-VYg{$VE}jwCTo}$;Pb?i9h@T
z39NZ75#k=z3-3#bl!UB$9n<t_G!k5_?S6D80LQlU#p3;ec9AO@w>Q?@wYs%X(=&LL
zN8d=6=k<d^+)>EhV*+OBetXk^2bO3rS*TMMCl-A6Ps}J<uQG!uRyePuC0LJsDgM*U
z`4(B~@AVPa@-+J*vJ#hQ=>@~iqs^p8aiq(TOi|i?lF!Z}Pk*Q`$jICEs(wjQtGgYG
zL&Jy$Toj~Hx~ckj&t?eE_O<>QLR1Wm6JMXxSz#u_{M`CRSs_U1&76NoEnB|rX7$+9
zq#^8@6Y$W3vfi01hq2w!b`QT0%B%l`jxH*O-QyVaxWkEeMCbhVAz(gKF>t4x#;^kv
zSo<_f**Q3K3E4-Uxw;$)#A_fBJz))9g~qk(9h~6Ei>~*qBGvRPNFBrnJF-9D>L*UR
zGZWvm`HVM@8CrS!J*1vspuQ1OwL7a%7SqdKpTvGXYy=zFPE$PyWInd-=myx$cZ36H
zxN}vY>9XFibYBWYk40>c_wWw~ydzY$Gjh<U-UY-d^i)qB87{_yhwFi2YG3(5z}=oz
z?iXO|(*I)2dTMZ_`K+Xb9|qi_rsYMgDU%)$fmU9e-5P-JT@@VP^ueB@rkdZ(g6YTc
zicthErIcReiblUuVKHtNV`eNHSs8S!{!<n}IiI_Gzw}(<XxB+Zri!ccl;vM&U;Dm%
zJixK}9{?73wt3L5Uktt%BQo(@hEJ7oBa;li`V2?Rfr4QR{h~a$(d~p}cAEp6F)(s4
zphAX-@M5JU>(FN<qS6rNIICYJ(fV*ZBa{E}2x|d3HZJV7v{u9_-Pb_}(<?1^-JLD-
ziDgR){MjTVxE)X#<m;EJdUuQSp$@z4#YjEsd<%SeE4c6{fM+1DI?**ih4sO86iah$
zN)Fo_7M!nq=UnBdZ^Hz6IHQdi`L<LJwXTFhT!qRUh`97%-eRGTCgPWKGo>k8%SK|Z
zTwPjLvM$#c$Af`=s?!zM#2!`ci37DliAE9pqkUT0QPm0>4$#a830~xsnK_BYiiT29
z23x7h9<hR$A)UL{L3s_0;&{;I0FxT;EwWEw9LqIG77V0Q+8X#N*aqj)HQ{K9J+oa2
zyv8vVB;_lTB~m`2rzkH6*j9H;_q-%0lQd%Sqjj?^?34paNZxb>+``LEez9y&by!*Z
zy6MjQc(gyv><xN(-#VJ*paVXi-587V9+!aY)2_PF^XF%cc(Xp6W}}VsVzRm&KSbAs
zT5beQiCy&iI&0{Gg=*x*>Sdg*qK%igJgYRnae99Q$r8X~i3~Q4PNlgbM4+faDCk9J
z$b$6J-acv|-<8(?fQj61(^%}(8T6tT8gk`xh+g}Y!BHKJ(fV@R87Ks+3D{(Cz9;Vf
z!NQHX>#Yb=lfRqYWGSsFq6gVUc{);mp$Al$CP6u?5j;p|>6T2z%%}h{r#3?@84uyV
zZebV=4XDMQmn5mf87KnUJ=*IVLLpH{p{w2VvyplGMd)!GKUGUeO*#egV4^FTrqN|}
zwu0Xr#^YH{!&d33M^nCRP;Rg+=Cd(pp+;fnxB6xdY24P}-PeiEkURxZ$Sg<742+%p
zs}-tP5QY&~{H$lqB~JZdtDFPFioUho=(6|=8*xHc%U!sH*!2Q;W6>N@adiz=Y(-8m
zb4lOf`mKnCE2T~hIVEyu_Nlo6ezo%=IOmpQW+vrmb0tD0SpuJbl=E6+Hx|F_0EV#)
z<~X*Qa&9cmN3CWUsw1n5J>L8j1GZ$x&Yah^XP(}g4nJs5KoPy|n^J%iI+JGq$B4*x
z!ss7!*8Sj9Vb}N`Y8foZ4VY}KYO;Wwzo^K;ipOnjwd#4<BWpFR@W8f8{>hkHU*<s%
z{12dq%YoTdD@JnxPs9&&!{(+rM{v(XWJg8oJsUeK<xnO9sqL1y(Y9P$O9%xCeEGbj
z_R1j5x=(`J23J-<xY9vT%n5)(`n!umfKq(x{#{eF?o>UVtGHO_2D47KJ4DXT6O&e|
zLNXG?<q7Lwp5G`6!SaQV5_|1J0GNw`!tN~|A6^2#_++8hNoU0cm#V?P*xxvIvfc)2
z{D%|~W#rsfjq}^O8xeM{JqbR518miaPH%<;<`Bc{@t@wE8|?%CXAjVYwW~6AE|YZ%
zl#>dU)3Z5R{}cJtdI-%os=WX6I(D>XU{vdWhbs<m5&#)?T0QLkH*RG>w^<kk6yT>p
zuj;u@vzOXheJk7WS9|X+H(K65K7Xt^H(*h<q9F9x2a6k6j|qJ6_vheHOn98>v$SS<
ztXYR#T(VSdT$yrK7yjl5l}u?QM&>6*?daa}-BNz#W=w-$0Esx?6E}(2P1@nRgGZ4U
zA6ALdZ7U_^rNjGWIcxS$I4Y6Kg#I)<lKD)ubdl95u9thI8#lL<NFgCfNkYh4s6dsN
z=y5D5e%SORCBz!J<Zs0h^Bbj6b4B#le?9@@f#_%EdODYjWfwf}HGIg2j}{Z{K1@WZ
zUv?XR2JfU}4I++g1l@yi<>|Lcjy`SnKOedaOf25Lx_pEF8i}MF*hVpr=iMc#8uRGb
zqyjEx#@$_U;FMy*9;)g{;?mh-iwhvGq0o5(YCopURQYcHm_Gz-eggmu`EDLJsy_(V
z&Q4^CE4^v?+yv}{*a^a}JC-kBk$pY^Gt5|jo$C&`JKQNboJxD<<3XpO8aBG;alj$o
zNO<pj*{D!X2wyWv)Pi|vIFlp#S{hm^7;e6u>x>kaB?+E?>yf{Gvl2)eVSx8(aK+jQ
zw#lri5|BwQ^$Ue@;JthAJymD)87o6YHS^cg-E325Res$X7tmIEZ(5s5Q07JS8c{ru
zh`O=fy>zd>=P5!;-V>32&d0ijdjky&;aZp;*DViKBT?hcYP#@@XK)1^4h0VB{LI<<
z!*~z5EE}ky?g=!x961*$>(h_(D5kl&oUD<>7{m6TYXt$a=wI)Q`Sh3$^vO%SzmUi-
zEI*Mj+38#zP2LiF(7pld9@2fBYvnDDywdWpA!4E!BLlxrF5b1V*a)7d5j|0$@~QW`
zbL|`6=+N3-TR>J8U12<$U+P3;X%Gb{h?GzYF6|Q8+g^if_U4Px+TT6*E|C_!51(1Z
zplsf*9qZd!jseu|Pqj4Gek-**1jdza2nT=f8T<!W{L?Dzxpzu~H7T*h@M2P+<p(^{
z0TV_qhSel~pm3OTHLe)HS<F=|z(gy#L?^o<cS_UIYr&16+kq};Nj%x`4UX;&0N#kh
zd|6I)CpyJ-9lqn5-y!AzmH5Fwm)jqo24oA2TyTQ{4>t9SszU01dZTGM(uZKIb1ctS
zr&#euAsas0Yma4(@iafrL0RMb$2Rq|>cL6u62;j%5{i+*G1n=vot8g<hsbCcBbDoh
z#HG*_rT3Go!+L&S;`5W+KtdWjk)j<jw)^6x_C&3X5MeJ~ANzA17}urr8PV-#XgqXU
zTzMG?lmO1@iX;qrcl^wLHE?vC8z*!w@t6Cts^;`u?LVH#|5M@GmwTxLy9jQczZkfv
zdYTW6e{Aenp1M0GODdGEJ`a^eelC0;%O)$Y@!4_d3YVqRiYn!p$aSlHL-;l*#6m#r
z@oE+%6fPYqkwr1u7F~<%4nYODBDeDUVMHh&+np<$9?L}_J0WhW554p<#^Q01gDO1?
zPoi>dKP^6dCP;cJ_)~7Aa5kPAuw2-6=a%_9IN%gDcWCj&3YBSl<fm0uBLl{ih+t)S
z5eXBZ(_1a>KRv%|^`orXc{Fg1YNPAGJ{4<|0HSIYs0jW6`jo~ms;A4uV5UCY3Pk?^
zzhwfoZ|EqlfN@()SW)|wmy$-;Z}7y8^VoHA9}^zO1Kq&U%W7p0CM&u5DB7Toj{z<d
zL|`8<&2uwvJfCBSlyoej?D4w+8;Rl^I@Js9Q}XS<{{T{gGNueFk~#ws`QKi;^-CD|
z=5%c?iv7OscpZH%!l&;Rv}lvxe+oXa$O*21`~#3OC#7kK!5EHlnauC?gCVD@S#rKt
zVWj5isMidC+A?f)7=Xz(u$I?MvN>B8m#*mhCEhQ<r@cZ4dTzxn)`HZu$x<wDbw|yP
z=qS)39xX%=Nr}}sV8vv|dp~jI69v*ox-+_u5^}=cF(gosVb9H`2_-`0tqjPd!%&eb
zt^|m&Mxfz)b?9vl<=!J`BUq-n!3-qZTcjcx@A{Alb(|)8dwp!%R*9RvzMJgBWSf1u
zmc{L*5W`z@jq@8;`_poDQEkkNUkcZR1I^2`S?kx6FL9&8-){9hu-T3bbQKW@OU}fs
z@5*XG#AJ+Skq+Cf!O<-xHrH4D#H(*N#78UnHu8VdTc)Dy+=7kv8dO`He(luU)sxLY
zTsNzI=VDkUfycYiOkQ6g7ssJY|1Gz#17oARu*fMM`vaVC!>(CIq%S1p{;O35CJ@a>
z7?+%W_DuC6RXy<3np%uo`1w{0+yCXm(Pvu;!;!T#dFS$1>3{#@*vn^U>VM(a|K}e6
z|Fu_|F5cZIyWdSy+&Bw9X+T%o*gfzo=-}&tz(8_5EOhzIu+fzIXbuC)drkx|B)_92
zukCZL3Ym*zd{HMK2=x)<F+G&eQLyZjxEt0sTHC!>A?A&p=sml`owB41E<GZ0f7XoE
zPgLk%meWLC1s3z8EacmTrr)0HPypi_iVA!JO?R@ycWT>+A`Q!ib8aKE(X|ZS)2%e~
z&8?}u5>6L%co#f2_6wI>j0k~4<^^8cwUoW@D^BLqsu8@r9{$SDm9IDT+{~9dqPTeF
zm%L-hGLzR=4{?&WLKws8#j?A5(GD2D6R95C*Yw06JW)D+U0I_oFuG5#^Joul)*r4*
zSAuA_YWX-48cFq+eWctF<sE3=8|^-nH(?`&&QKFhxU3Vid$=hBk&v@le|{i(BL9Pi
z@TQ?$>rZHFnE#1T88zgQjjhVL2H=c7%MHEhh(=iJ6LjuDZZm~Act;itiQ>%98$H7f
zT}VJ5{H2R_{=iC!Hswu|r2;kQ3?=UGE*y?B6>nssJ26H0AirJ)|6uIj&*;JSAsVj^
zaftUQYRNIHq!xs~zVb|=q24b3<fW3R%1>#<ds*zPQ|h{OyCQ{Zi*7T0=JJB5O~kqZ
zgD5K?tapJql_o6tIGNv-|GqDW@OH<SJr_UTGtMy-$@Q$~baWz9X^b(%G9^4c`%{+{
zo`slV`SSQWf4NI>9(i{L-5K~xhrCPJs^U_*DzgOFR9ULVJyvk9JlY=k>BBX)56Ux%
zqUh9q)<|zYax6$akLA0MdsG6G&7AT2mivNC*RL<R<DmEP509|2ET;SI!{v7y!hTqN
zbW}^rzyjKRRTJiQ$q|3}$bY!dXmA_l?1LG!e{G56COCV!?RmeU_jQg$uHA4X-A3@W
zlBikmGVqZeZU6oO$g;5Rp-VLTW8FcY#h_d$1LVbhtxijxTM%i;t#8RG6;4C@+PUi`
zuLnU0|4{o2>wxr3ZTANwg`L}k3v~!U(0ft*!09Rvyn}=+u3kziqxBItm)V^gw_<9y
z;v|dj3&mjc$K;ix_kFdidqR(KP=v%y*$ue;<Z^3{e(fgx!Tf?~+1~7{k%X^%?uFCS
zEVO%#p0!L7r`4m1E2oz}VLHM+K!Y5<jl4bwo1#7agQTpi^5|WMdc*c#7Ff=4G>mCx
znGni+NQ@eUF(kb`%u1C%zAZf!4itvw4F|j*UuKQ9{ak4OU7Tm_#^R#CBh8nN;+BfL
zchJ5=P0r;|@gG3FqYTD0EuMRbtEU;fK&Q@ACQ0(B|J%OI7rjo~_23clgwEjjyV6q$
zu5<o>0Q-l3fS;C{BAtPV{_FO(&aTmidq3==u&lM>v}~rSWr-~WxMNaH5l4|KBzTIc
zvsQ;v4BE=E`_W!zaZRG(N*l~np?=$fZsUZwjRj)u*iGINzgc}j+dhu9ucPsKaQqNg
zkP8SlA9W~n2mT?Ja*v77rlifxM=#H4hmP*qlbID+7{W&|g7S2ue8caqt1!6yVamu(
za+KnyFIaZQ<zBf8n%@1`k2Jl?Q)3jn>&pKY1EYMFaU)OS6jJUBmZHr+u)KEv3ZvhR
zwF#b!yxA(C-_G$<9P84GoOw<rZ1)e4r==P{Gs;DP(mfUH=>7C;WrMjYdWL9tg7iOP
zElU@=4T-&H3hl2kj<OV!uPe-p@1Pws`94>q5=)JG@tcoqvV#TSZFr+1QuSpZFpdEi
zbm!gT{9aZG`Pd2msIH1F7?WuFYG8#nHoP#q|H3=Q7e8K1Q6E|<mWT_0A<VD%INkOz
zZei8!8T*gtlBaj9H(=BC`djlGMkQMN$3m_Dwt3_HBe*=AfCc|8<Qqj?Gb*gQK7K|2
zEpF(YT^u5LKv$}`uE+~G2PQq)@Bv;|IN}G_LSai!if4R)D8iirv*X6=M)C|h`_G|m
zmn5gf;;Vs1H$1)|>D#fZ+U(c=09Ju&K{z2OAb|y))FS$G&jEh3J)4gQOq0#%JrZp8
zSB9`U*#+Q#^ha1SOm^q^_Lp5KKiw9$ANpQK?$UWVvSr8gAalyx0?HUlQPG)OKShDb
z`STnAVM)vt&85bWZ|UZ#{{VBH6(ik4Jm-sn!*a|bzDeh!G__j3C+-_`{nr$pC0?|@
z-B*ywAjdHPf8cwOfmgHfaE6`9iDo;NpCJHSWMAyCi+XZ**zU5abrO5N6QwDImBQYC
zw&`0-u1SDP&|EXwxjo#FdfGq@Ysded>He=f^ygsLzK{Ro!mw)*<j_st&^#qb`v1{?
zQcDBN4f=7HclZC7T-mFrf2r#qzygkzfWAa+1s6D)S7$5fctfv&yyphCuKgnL=5=rp
zYr|1w$bvUaty*D7x{oB3y%8kwXQNMcFRGquDZLn%r7({PEwr##+rEQqo=m6M@kgb}
zUWP8l9X&wxXvgko5heq~^cW|n2<?{mGgd;hz2qjZ+mgHVjS?;uIYqASy5tTXyYycm
zB5{>kywBb*k#~$=C4lt)FbP_^u+m>f^FiHF&NP?{vouXH?fiW!G~X<IIB89U1r#?U
zBbeE*Y1#VJWALEr*Nt(?=+P7wN*cCg)jurx905L8K-|HsyTg%^S}}~4nS+@=^2JxN
z`htWCAW>lrQk*bJ`JaeCM~;=N?tM`}bXcI2H?WD74O1+JS}g&&uA+7~XF(q@?moGk
zP#mo|YsoNl;{MHe*tyLZw_E@PiTpYIH3{;00G7RJY5v&f*+EUfl_C30o}sSj!x8fT
z;O;G>;_9|F;UW+q!QG4C65QP-!3hL+cXtZ}sR|wl?hq`vyGucEcXut^;r%M--uK+@
z9^HM<pC0#)@84dvYsp+|?YY;Ub3V^p*plfXesNVYovybEvY$P+`G70f2SSrcm-$kx
zKT&ymcYcM+B>kqiGaj1d^L3c$W@oG4IL}OFrmj4t0f^vgAzkd!Nzvn3vwsaW6*(79
ze|ZsCh3?lyTTzWc%Blk^3+RB_J_*I{<3nCCk*RE++`=1$6?oNJ$i^cjW37H++FhNi
z%&6~@7~9wFl!UJ|kR-(br?$;pWcWZzXYK)}z?~k@dtC^IGDUj|6=Lzb^8g>u7<hY-
zcNLH{=t%hsw;=kNNi+xA^zBk}Mv@m&#cXc?eBMJx0Lub8+%oJs`$irKaEeHkNrqHy
z*U|-kf`b-{l~h7SDu}A!Ey#l#R%GMXnP3Ko3#duC3_{U@FQ^#2r#4+w8J7L7cU2(s
zS3Rtue&Y9S`0bJlGCi#%g1Jp1A-|})dP?APnpSmqCEpOcI6YrzyUvp)1a$%R?SH^(
z75}l9u#+D8{Fn#AZLRjp3S{Q<AnCro+wLc<bN^FPj)z%I2&3JF)r`D?v0yfcjG)av
zy}tll`gcY1Kn2y&!eZ6ROW2KKRhR*>`50E`7msH|dl>Ecy5Fnvo>pNhs0rV=yT~CA
zJh)2#l>JM}OORL(+*%S_3*M^V7%h^4WP1$l$+P$C$Ct23uD4K4!e;cR_`yq?g4=6g
z@Q}2K0{?^Xu$u3)<E4MR5&zw>2_MYZ^8DjiuxC=_hJNc_(pBUyfYgF#upp#89Qs8;
zSmdA@ieXutv6**9CP=1cz6z}870K6!9)Jp0c*r&m$bDr<eX`tEUy^<3=g}#rJRfq;
zS+~otM28M;Gy%(S%PXkVCe_<Ua&`fQ6IVQ{W9Y#PV$CEg4d%%)oo<dY7pj#|j1Q0V
z(OkfQgFly5q*g-DQ^v4<qQ;7EHGM;baf5SumX2&jDn9wE{V5&??w>po!p8H;?xk?N
zH$We@%zqkZCq3I2jyJs)iOI=BLz4E)og46d9~?X&<YJRUy{r8B-5$9lt|kC5)!~ch
z&v3NqdiALf3mjEs{H5Gb0=!b0{ZFGL%&!W3><3qN*g|*5;hPuNS2KK(Yj`I91;mu}
zD;|`36Lh|v>kH({R$h`?i@;!mwGJ-rB}upeA4={&eGn6v%;$1Thl!|Ukx#e1d}yFA
zA#o;`GW2fk3fvYZh>lYI7draqVE=FC)c+?RR}s<E_aywI8sz!6U9Hwx2ePa=zfO?O
z-tPLbJ!?e6-=f1b79?P|^`VT}3#}Ul=3uXv-CvAURXiy%>sLm9oWS$Q*j#|A<`QS%
z7qRkV#ZH*@NIAcNn_*9(O}D#ttIzPa69zDhhU?Hj#l-t;TISgRz+2x1MkN%Xxx1WF
zB#+d}{5INka6Ubb6+sy77@T=24K_qucp%A3ERi$_$f(JUYB$KrscCzbEMVPP`5<kd
zO>zM}y^fAt6NvO`_;#6Fdg=y!ZA7-Q)s(ia5~F%_92s!$-lav3IrN^(OVI{H=i!0A
zl=Tdwe-TLwlrIDp@ONR7I77^HzN()v_IYQUsJWLF;czrqJ@!Jiksr_wu0oZ!c4&Qc
zIX@lU7v21Vl@$PsD#g)*IvgEr%`OtLIq;>=U+iTN;ZuP5cPZ?>oaPp0+D~l9`dUGm
zWLPRDo_js+rNR;<4|frAF|W8sGik2#xMkjDYWZn-J_xjlb$h$pvt4y^0zZ^uwIbVY
zEl!*bpHX#aylVXM3CkN~Ymf7L&%rp`b!(*hqh%39k!vF#OF5w)Dt2}r>x-cE<hrrG
z+{ft0819#SH`H&yx80YwlkY2Fs-mht-zw#M3gq`fCa1XKE*CLMHDQcU-&X!9yeha@
zQ}>aG6av30KshTJ@cghgQBnWIF_KdVAB1`}+jgjzJ16P0HKcI&_~c?oIorsO#A=HF
z?FDGScQk;glSuE(Kl1?q-YL(S@58sdxVjsf?)}vJC-$I(>@Ai4e65sKA6D)^&N=E>
zm?S!n7abaH`g8Vy$ewn;me2NPzMGPHdRTAum*pW-9Nx_Vbut&hIZc!Ud*zSUhv*dT
zkD<*F9rnTEigv|Rrl^skdqGOB|K-78T<pdN>_MtZT3XuSWwuh^H4{a(da$4Mq+yhU
zOKJv$d7c8sWHQOoRXPeYn=Leu;Jtig_w6#t?tiFOl0JE`-(iTlDF$we2og8GsBXSI
zhp7EgQ+HRj0TqR&gSasKo2%3xp^6r7D#1td1m}*nR!HdUs6xJDEFY@SeM@KQ#nAVl
zYlq<^$kk20E-P2h`1bu{5YdK}3H{3djpL#SVI4d&C1;vTWhLecZB-?43Iht?s>}Qn
z7mAC#6cZhQ6v$)J)sh1ZtCbGjXpDyD(T0W)wr6rnXZ%a)ALxC>&oiO&W?RTx+kMZf
z#-#V|Y%f~Q0%&@<qp%nf45@z$BC(+C%;gGfIdgcN!Oyh3OIurPI{=slcn<yrxKY&Y
zSs++LgD(<R+^Haw|F1!c{;R4oB5KC;zH}#qc{kQ^=wcwx__*?}@@~qS<g74jE~&vw
zG_Z0sztvCB_3M2sIRJo#xsuW252_J*)Xlt%bibG*aF!{Pm?G62M?M97!F6J7H1J#$
zjjCCHK0COTJBMCftQTEPlL^&Q0D-DwUnZw|+7D7~w$XuF7;`h$VGe{s6g#8?GB;o%
zU5_R&#Imtab|`CZK-+#Pj#GEn4jI<UWd_O3N<r;);RJFIamUoR02{vxeDNKMs4IZ~
z(q8kDc;<1%Nb@Thsf`jIMM;QwwyOR;NWXq+6mi`AGFW+z$Aq<H7Q_<i^O6r5r8xB;
zn-lD*h{*}WmJEi}L#A_K-$F=tc`5s^F5Az}XP1_7uTLhF)4N}aF|JP^ulJJF6o&&6
z9=tVE87`AWXGouYGj9e$GGwAjs>}Fpw?^b^Fr{3=*S~al#^_pujl{qr#GXq65$_Zv
zF|M^tXCK>dTC!)rJ5z7(ZBtDOhZ+T>K=YUFH-jrgIC(ApLaL@dRC%^ip6n=3>ur!s
z|7V}N*HwOF0FHe20T&BtqhP%CMGq1vVE5Hg3|glP92@`>$lLC-tZNj82W_oZepM;s
zLH!;<)*y`4pz}^>r<2d#?gTLjFfIk}#Wix!qbvatAZ72&odSmiQuK1}#YP(^HlRc`
zRTA)RbK>#cyHtEFwUWFN5<FAPI`K4Peck)&J_^J#zDlsZ#1=~y!P$tRXm!MbMh+C7
z8!-tO1q;^hBL`wEG*e|oRv+zTe7wr%+A0TiUG@Kfpg%Aow<Mn+O4<gG9+lOt6nNYA
z!Gh-^U8saG+LgP|Y4+2Uc04nn=+9c6G1=5}=%_aQ#$r)>Q{~pRdrN{3>$fko6C#AS
zs%wlzKY<>90W6?L!Ay@xs(>@fDKBpSXNE96uTWhOC=tjgzP@^9L((&qagu4;AyN>=
z1k$SvSV`lucg{oPzJekkM}xN?(?E0?l{-e`p0_rDk?{rBanmR#*AeNIxj-=y*ZLvh
z?2V3?$i{IoAl)3xK}yc&^a&e1@8x#ibe7B0u;VL|yVb?eQTTQRh6nmO^j;)`$9G#r
zK9_KtuShELOx}rEg?=pHpSgJ4QrEd-&uCZ>Oa|=-vyeRA9xRPSG>pF#wg2E7MVEN&
zC`Sc#_)0<BuVJWbJ_`E=$LC7R19L>vqAQI;s+`?&48cZy;JDS6>NFJ@hIB0X$yTz5
z<=Xz9e-Jv!SqJ-8mU{tmuL-io61xAIJg-;JjI^M`-TA_m#k2(HySu5C$8B*x_$)O$
zhqqxzu-T*9l6|<RU20uyoOXYEN6zYoltK1<&)%O`B)fOCe+WBepnAEJsRtkx`k$GT
zIGkUhC=Wf0e0ai0YL2SxsLM)v@JpD@a)W&`32MQz6D!Bj&O70A5l&T1A*d7dx(z|_
zqma;R_fJm4%L<93b8z+GRS|bc1fg%In}CN^<5wDRm4OYPjQOK;q?7De9hG*ol2qpm
z$)~6f)0D#g0>EOD3=b}p+A~YP-XVz;L?Px?tcV#>Z7mfcb=tSgtX;8ZPrHBCwO*yz
z%8dF#cXdzygrQs@^0Agwhl^AoB*6b|RQ4v>{jH;^cu&O&dC)=E%kq@{sz}g3%0e(H
z?8A;BEt7K>he9asriQoyNJA@s{vgPL`jj=btqPlx;@3PGn-kg7K+8e$po?48T7tSB
zMen0bAD3hNv`X7u!Bg-P#`FbwZbtT2N0MQo`0mI&_a*WZh7$AbcAB^?$SG#)^ns`a
z;}V0lT39=Vr_lc}&kkv#2;?{csY<Jaks&8TT3g2x=Q?R;f}L(a*_<a9$D+eRZQUZc
z+#5$DGHY6vEun2sDgS7~UHv2!GPea$Q=E9}8Ct=yV;w4i7@XcX^gMMv30^A0??5+h
zer06WJ(Dc7pgUu*HhTdd_qxPHA@5eMpjX^0!G>%Zw{ehKs#aaY*e|Sc^fs`~#0Zf{
z0!-U2=dKW9O?&^@THoi&VjqRPpbMDT+tD5Fc%kdVteG1tTm4bIMsr<|!@M)$+axCS
zlh<|VUDuo*Xnv3Y@Y4hEx#sf{+G^~B4s1L<gX&1#-Bbw;zWo0#zuOe?tvYOL8t`#r
z_N0s5`>QHVwfcE&P9q~4T~f%d`?Wpe+JmB=4HZ)zx+;=2yLfs;Ur`;yrekCm2UdX8
zod4;nn|5B0&v|41Pqn+}sV=2?*F)HfmiXYF2~FFB;B4>RNwb8;yv)AUlOjQ4`p15s
zZYfA3A*j{)BQBp3c^H0){tVmV+O7qDUv{9g9>HKw(57jWtV+T8i`p8<4|}6muq&FI
zOu~_TN5p?U37ujU)w$>p#|vgp*@ny=|4%;zD4zaZ##<Hr?c=#7)#!|V_NTeDWy<}(
z@8S3c0g*b?|9NX|(g*nL&EzQT7u=vU!U6%Q&e${qL%bPbk4TLqE>l0fZ84mN05F10
z=oJyz&=+Pe&M_XU0V2_EL1q+tz@EV{Cv2!;(81ao_1}$#)Jt8wW+9bqC4?<0L)$aE
z+hhK}|1$*kdD%?!cQxLai^OfWTX_&B6p?S*{Y|TD$kkZG-$i7~2tFk$z~3Xs`yZ?R
zX{UPVmmVL;>yMUDJfVol{bvDs@IMr`o~qpbZt<-nae!KXX2#t=N8s5}!1B#$lWZLV
zlZ*HNSdhfkmhcmY;{GjR!Ik0A_NTv>rv-Aaoced~n!XPv|D0`HcIh7?(ma0_w(ZYC
zBIk~h8=6$LUtj!@R~K*KNcFS)l151P3`JvX32eKk1L4pl+lt<)LR2bIRDu;2)7k#H
z?sx(I7;}ezPY1q&?7edG15KwJ=0CQ~mEmt6|5mn7kdX23(=37Q?OZBU_T5Y*O}xSM
zA3Nd4WPQ%f3!`dTL}LSw=ef<vv7*fo^E~G@nKjp(9F7`L(1Kcn#E85ONB*0CvPRBi
z4XV7{2E!<fg#vianQ_59)L%x9TT=`>y#(E$J<I{ryruiL$%`x6NK=m4e*U|b(=Be-
zM6P?&Z#8+Q#0yE!yPL}2SGae_8AnH|!%~swciHpjUDZZyUFWvW;%*OP-}CO~4{hgD
z4-J_YDk>>OD;Am|Pi-Sya6)AF!f3YpI5@*Nt0yOTMA<RBN#m7K<qyhzSj=kTwBA<r
z7t%qZY%`gf$-~MM8PN<&e1KoyTN$vOC8*fK@gD|baNPA~U*g)OVN(}-BPrq5a5){q
zC<v+TjASw+DD&8q=(7}x^{F_&X9TLTu_|;YEW{IrKzjky0Av0nBPSCl;^|j%W5Xe{
z1k9hH1EjT{bAj$VN@x!z^bQ18(KN^<sHmSdKjo+m+iy9meMZ^;%&n<8_{1}A#TR1G
zsmyw06KyrA$S-gG{iCzlQJ;DIZzQFhI5{B`xgROC3+X;7SdLNsn7eGQ5#8kEy#4tI
zrfyJ?^LLT^%NZkLLyDTM+!%yc2WTIyo?WHY6r@c|GUli^<oYv}H#|Vh@41dQKT2fL
ztJD1e(MZKr(&3oP_JTgSN;{7k4+yj4($WB3NVk#WPPZi#EQssfrZ8>jH-=J;^-3BM
z2(5H}!j0xP)$`Z;yLIJ`#}yhG6qh~-b~#elsi-eY_elIU-Cr$K?32^9N_h98t*5JM
zhA?Q<PA5%`kCEJ+D1+1ggUv^;QGU-5mOFu87#N0ws?y!dIT89Am<tLjvvR|1GYISG
zhUjcMQ|Zxj=qvC0qiWDN2TiU>RGCKLt+L=<eY6d3picZ{g{Ag}vgsj1BAQVkQzmtC
z5Ql%-pHs2f%o9*I`7`~y0=$=09AdwV!<h61$AtsiTYz{$R`Ta3n<F+H6&Zv%AYpyb
zHNsx_ZIR5%-1*Q{_ep86iuquh>jN{(w=JBXmT&LSv{Vr?=@JS*Xu03z){;=D$MvdW
ze9(`Qdh<yZw2d>FuZF|e)F+oS^PNFjRGDV#tt7Wj_`C1&U4f<?x(GdN;4SAj<Uino
zikaxGLqGDd@fUp{JR3IF6hRx!2^tJpjn!d@0AF{eij(O(Z8!F$xVS57GL*A0xLqUZ
z7cg9OHr~z9s8c0oCMy&8iHXL_Gte5PN%4dVs3PZb{x<1m<!8BEzzoNF<GEUdIh$R_
zUNSzjM$kt}0O=5_B_+_Ld2nQlHml4vFAi6Y+_9(V<+hPDeu+0(?m#{Kaf63f$Cy)M
zN~yM!k0<VD6rI7t>McoY&44={#b!@O)NSh3=ZtDRAcMJ215r^@8b(i^HVtkdRiudI
zCOvvEy%H#U6<k=D*9L6o5Py$bryhcX(!3>qNIYwit~jQ4X!>bCXU3d9uHeWWMpy8&
zwM?#@tw`ePG*FF_knS2g2m47kIWmbVO&$BA;slGS8AFdo&Z_Ci8R_&IQE-Is)$n<j
zt2L{<62=}|Bw^R54$XrL|A+O09@H{attogdtxT@oW_uO0cwda(D`P>fTwcCk{ROWp
zMg|My=`5rge=&XBk(%`GPC)rNZxyK<_*I^m^!58);{2Jl&@3j~f?7WRIXa`UpKoKg
zv}Q{wr?@WK$BE}8P4oLpHkNRwf78&ESzUad@Ec4e(&b7Garw!RDPQuFdXM(4f*dX_
zmNEY;^6ts}U=hu|g%kuQ30^Ql`~IsC(F|!<#IRpIr0m2RWQH2&2&Srf9PiuFU)R)o
z!jS)$N52<ttYgV}3iTdrhrl{Tb~27rubXK^!Q%I$0zk!~m0QX8NLVOZ{D8f|W-!XD
znZAK@{y9<LdqgvqQF2+E3%D3Y^Us$`25<oeV*+P7^!gKUC}AVrr#nvdDZcj`Z7O<G
z@P`kdm+!Hpbh}gT&v$k<k9Ng9t#2i|hHMsNERs+gYF|g3bCH*q)M5d|zkxw2dx=?&
z_?J;Mpkb*&U?F9yTvnq#O|*2BLONy<jk9r+v<be@0am&e8}2!xS$cPXj5^r|zHK#c
zuQ1iTF-&pKF+<n;x7&Bd=Q^j~t0$@I4bJ3SI<R%c>u+`!KRUzQd1&@BdL9h;Ymc%o
zbru;SAEE`5Q)-OD+(sTrMLBG)1!)$%x_Ej?T#!BD)*RAB{Mfmda5a)k*Q2)DI<Nwt
zvd_iad9y>18$6a3WDBg?x!H3i%2NzWBT*yHre38?{uEbTTu^c3f0e|5J~NsZ$590g
zV%Rp{98Ad?TdtQrDQ0KOBue_3BmBVG&Z<w2{=)2MT4|3~XkF2wxW#K!U>6P1BXDni
zYCY+-+orvFo-kWY@YKrwVX1w>vt?DUCABXetyQZGa$&_a#P=<+;wjLtH~%$->#P@c
z)WYOahQFi-Y1!+ukCx6<XaDA69R5aiEY*{y+XUppe2FZa8WiEfDj4?d`17C=xsn=X
z($D-D>YX8TpvHGnMFzRacm%|s-DbzF@L#uFz1h-24$`$S2Bn?)B`iGTOq94FGsbsw
zRX8HV(&q+m=6?9HQ`*Okc`a;Ii$5M7*6Fay(@cp#V2Ltnk<AxN@gwYt6PrF2j^HrC
z(7LD^gKDMy6A$AjXiIqGj%D{>AqY<5B=WR&jd@utqsA%BA!0+(X6y=IHUhEX%+F&R
z`+g>{6y#f{ik0UoHaC4OiVQE}v|7k*wkf|$q2uH!aA(QqPK%Gu4O7GZ#P2tKPhoej
z&Pk`2NOX2Fp7m|R4hk9hx>#1V@Ep^4yre0@TvlUHUzWTms4=2jUX}`Dw<m*TIlSVv
zFTZ|{K2LdMDiO|AcOGuem%G1EcxAD@Bs92U*RT@4WSeAKopRnM@MelyDSAjmKhqE2
zi7E~E+a)aT<$%%a!LO8z$T{Wn2By`KIeDd0ZNeJAJ~9jQd-JvFsD<DTPNKcb0Ra(8
zKS?I!lpYYf5QXrk6nh(km_<4EG9aOhSYgKNA#1_)<lnMDEp#FeVI+ADJO$8&G(8@)
zpNq^t7RFXd_?>n0Y9r{5JswvV99l~I-FU#+KW<E|zLO4YBr*EAnuKYe+mk4I41`a=
z3tO^IFsUV>DPORxT(nrZV8#S{ZbR&_U&N!e!wG4TDI>c$C=etFLv&lZb2ne2)M&JQ
z2)8w4j$nRsclCIM{7)8dIo;D=OKr4%|0rkjq1FVFA@mJ1)q?utM-UB7dJs1Ji`Q8j
z-n?jAT?g<5cx?Y{vY_iGDDQJZb6IoiF=4kdQj^B-wdPUwv&JC)#DytQuw0M{ag6aS
z{5?eqZp*Cyf+a`6AV(Re#hb~W_xLS0T_#zKL!vjyWt_(sWgOJn$MGj>bZ>UqS@fzb
zzZ!>&o(?r)b|X{Qij_^r_L{J}%Ax)GrB9)bo*hd+{QcJt)KnG+ZUu$nZv{p}>DU?6
z>%$$y2pf9*#UJyK-YE`y|12GfktKBAP!vzPuoERZeZOQ}Cg|vT+|Y}1ti{@*G-~ij
z<9a--u~Oeukw)5u)(r7lj@1>>D=PhzmtbAn$G9Xw<iZRSd)%~UXt`un>XCZLECqj|
zTd#-KK2tHy4Ysw7tt{kIEZKTE2zTO8jX)*pB^xdQSw;JcR=bI^Ito;ZBN{lNFfTBt
z!`QFv5ovAmR%=#auI&S`R1oUOK_ye`cJG(3PXiWP(rH8>P*SJ7L7#3rfu05wG;xA)
zqcdZ_D5#jIF*Cf{!Bk%&Ghq;4A<tY*!Q)3sM3Amg#I&tkD~p%>J4IkZWD56S;<~9&
zyG+wB*d%O{`c3?L7>S-5RV_H=AdJs~s0zZ~Yr#B0YUzPCAdcDH6TImbVGLa_9DN&M
z^DDWwfLb=G=9^D|kvD?^3aLN_mD6bs;h=(uy2NL0(q<Qj4|;6rS*fvM$6lW(udLYC
zt{O%0*501r6Y+-5m@WBUNM~#<YA0z7E?LUWm+;kzo8h<R*N&!94p1vZvkPc9SsHSp
zf+<&%5H1;|#`<0s^sxr-FJ8A#A+SLgM>4ytQ_xme_hgx59V5jrj+Z7nKYXQcw+aDA
zlKjLOz4j-qEVATt{G|2~JasUcnyOc*NrcB>g!`el@X|V-Vcz1qal1J+Bjv%t9YUN#
z(x3!;Y7}5$e{`_p9Zr?;BWv$U!lH1^BJ0;j@oIvydi{x-N5j{T0gDbThwt&?&K7*_
zS5TcLd8F`LxLOvaYh0v?KXumjQtMa}nn7sYZl5Fi6^=eO)6PfV4zwDjSx3LwMW(8h
zwalaqQ1P_v!Ts_M{2`oaeFfqBuO4EoFUkjgpX3rb{{j%3R{T=UmSh(R`I+Jt?(gPL
z#7VNj;EMBsMVwvvn{(|OBwC7I7LX7&r;Cf)IBkB;%UG9x;$qbI`5f1i&<)2{$7JY9
zZo|`3|7G-D?Lf<kXWK9HxpzwiN|~l1`rRYEPBOf-O3_5IlDxaGCr{$7Y8LXI%Y_dv
znG+Wvmr$^|v`xk%7h9j<0Mq7&ufz_)&0=7w76P>K_S1Eg>(B<k-AH@6I4RrNTz`tN
z5J7s~xMYk$W``CJ9#9A{^Vr=gG1s_P=wFLrXo<gvImGdiPW)tZl)hQeC4{YjoxCDL
zZl>|nqh4$tZbVetL@<x|A^maabxI2W5zbkrC`xPJhmxlO13}Gxb^LmNq$j2$Pk;uL
z<b_dJaBz%uC5`5Zfgx%p;{XH(3`8yq2|a%&YV`7iIYD}QIuK;izPm!hsbr`yRf!oX
zsnAoX+}Thgyo(BYY)=tKcXWM8>AzLq!kup|<^3@##Vi^0(<frily9wzAKelih7&cq
z!<sGV8EO?i;Z-(aHIw&xaI<aH1$sS-#?dsl^sG*~-B8|0HE4SIE;RdM`BH|pfM8*N
z-(V0;g=OZLihvLQ9FzAWWf}&{84P$R9DMxEq5SN&<{x$ITbs^p8(~R2s1j>WQ^#}C
zK%lH@D&9X*sW}3Lg?MY64uzhZZR6<AD{3_U$XWWgpiS6RiUUozeONW2Z>jJQBH!|=
z0R|dU!|ixr;Sr~w8>+g&Fz5x$D&5SrLO#s*Tz72p%@|GRR0ILo;hb3n>k2m3a66&@
zkRzs0rX{p$FQ(?6Qn%!vLc7Ck32B8@1l9Dm|Co?W7<_7^d;Z~e`_D9wEnxj)%nX$m
zPz+zisd<rs#6#HCzfG9!P!@hmSVG?gzpc_i*<0A+z$Svz>Fn}Gn&*6Ls>hcXJ#TEO
zpoubu_>tRUV`J}d`0qx)Q*tdwWcjI*@Tugot9<Cp`~TJA{o6#iu=t;w$qXWxfG0Es
zGXDa6Dn$oB!;~}X=`krrbNCYJ<+X2Lwp|WgJ0S*_1`>ep@!`q6D(2#0s-6BP0A7aB
zWD1Ss;8VCReUxOIU@ntt$G5g=(d`K3VxakV|Jh4eLUsdze*_NvtG=%PjZ^XeNhH93
z^uxlb@p=Sa^tzRqd*%^?`2kebDk!!J)`+?Wvogw_WnH%i`F+p^AnCVdp7Jj=S#7lB
zJ3U*cE@YSc_Epbi@cFW``B<MpG@XqnW>jOji1=Are}Nf6vbNO$iHeX=R%T7}q8Ujl
zUEq{unDN}sE*-8E&8~!uK3UgSyWAi2Oy2t<J~)`I5ke}^SXWduqJ8aiz{EfWcQ<<|
zsgd|5i7=W#(xCg>cJ;m(dBpcVEK}3(QRfSP7Nb&c`g2Ae%g;P17{BCh?AIqeWfY(e
z>GwFB>BLINW9O*nfBPLyB#|nYA@hl5=;zOE^o=Z%?~N8F@{1e@2-xrwL5RtORfGGR
z`3K*(*|&7~gtgW`*lYw@Vod7^m6YJ2bX9!NK(~fb#RRZ?=7aR1Gds_UiF28T;XP1>
z6{Ec0Owp27DoK#<{W>4Y`N81l_8<-s8)0sJ7C-lQx?svr`UuI3Z>}iVqNZBm0Hg+#
z57V*w2Zi9z;O>eXTnn1yv(a{`m+wxRpyEh+V>(nmvq<;Xa>n9Tifs9N;ryBfGGgZv
zu^$p}iWGmZWPs2e-oS{!t3nnOYKQrp&4%*i<wdqC#0Vr~De1p$@}YNVY<)ICK=iye
zB0pXD)X2&uK&in|`xoHV=h*@xY~}<h>Be=)WSm|>8m0O*k&&0qDdQnmEML4-9tyhp
zuatp!l{{~LY|4fEq?J&zJ!7y~<Yi+D!EH7_ombepSVggAj#FRaI{6Q?4;V7Pn>F7x
z-v$kfjZpc>Hme?Hh;{D@6<_o93#)fUwqfWovC)N3nstS=3FcH<%XMy8xMTll*H%;u
z`IVF-ouN`;5x>I`;!K&TG%`f+aZDzmCOe#pg`N`RL5O}?Gv&5!f|yOpnUAdvAb342
z)uU=aIETr%?QHM^Bu6J>dme^)6$*1iu6XJB&o?IgR9SPsl@i|rA|(V>!?;=Z1Lwuu
zQ3EAThu>i#7yx!QMiYjJL-XW_NeRnyHraW^BOKz<gIlF1;#{6&H*!;cUqu$h*hcX&
zfj$?dN-Gj4<it(i_3q{rn#W)v$g)iT$Z7p5VAVOR<2+MsU_wN7nuV>S(`o!wZ{XX?
zyCBSA(qQ5Zf}Bk@@o#Gv<CrL~oJ;VB6r8(~P)#G=HYSy7eLDcZ&#RqrBszSYAdoNi
z=lc8&E69Z<Z`vd@!|=`{5vz=DW01yTR)!&-;G+qyG%tv-Szkt<%K|%pDv0eRo8#6T
zd%THa(1N<9ooV<%4VUGKvj70ExB8v5nu$QomfNDQfSn#mnwut9pK!h}Prc!_SNr~n
z{-)Dp<%C+ekJSe*ZKYI65m?sGdYmxZ=E#tbk?wEnJU`J4AgdKo$0;OC(Y@kd_)&%U
zV%BK7vWn14su3a~&hn+FBwfCg(Bx^4LXO9bps&_W66f;(IGtF1{&l15SMxBN*zM!O
z8G_At96afRW{F}qV@e9L&2@$mZ}m1J5#UF7PAbIj;c`ET>}YqhhpB<@`TQ5YcKKpC
z#Iqsy8DZRK1kVJrmhfPNszxW1@lD|~DZYnD&%I3wWRfi^j?8f9%AX%$q3AJGGW_7~
zpgTn>H1wXKoOWg!Mx8<v9_JO)<3}*~0@Np>G>oHH&Fi9J;60zA^eS_T{f%3gK`4K!
zr_w|Ec?I0_=Jkas3N~RR%51H(N)Jl+FelzSb;0iKDFTFa_fDUliF|3c^3UHE^Yn{q
z8{}5>i{3Kgzyb362tN4bJ2K1n%E?W?nyUWI1O5iCW^`#+AsZs?W>@XGtNtn|oBxwq
zDO;KGBigh}XgH83IcFur5$VfW6)GGaTk87!fzEptY*a~4x3^OQTVcuubp{&<?2CXQ
z{bgQvbTNX-dnX5#SUxMi%JVO;rzb9<V~4H1C4_Hu)a9a^bv}H0q8|JG>jx(YyWb{n
zS>hvu0tMv4_^rJxV~UF0cLBS{=$K3)Rdm#Jt3WK631Na9|Niy$*@gLu|0?f~r5KB{
zFSU!5DUAN*_lwpzTR}OPY0`;<q&)o2<KSM<SBm>8YS!B0+g|zhgQnA=cINCQU99Sb
zoFoD16<0g|f-y#;X!_i+A?t>oN3%ik{*79Aj#nn2JF)k*$-3E)<S&It*6<gMTH*qB
zEj;RyaX($(AN<_9sQLLOuE<sGoROip35NFaOXPY3GA~GHezjmy8T@)7J3WGA*+bUy
zFF*~i*-$$P^P0(5^J@EwZT)!Wn$k)Lk6tpWgrM7C{Ou-oAlT6XZQK~$gl@8a&+7P5
z*-smaB3%8{N)hidP}yg_NXRm%aTi&lSnW9!hEjmcRJo9YA8Q^*nQ>{FZW7_T(}Q8Q
zjOEEj#_!xvaEl>x#9O!V4O&f5U>XP|5b?)v*I~QY6meZM15o&eR%AdE0y)VR5q-GX
z*?Joa{2vpFY;l^<4EXNbd?6$J(z<tk{Wl!Ha#O!*%=E5Eeo(n5Rw*c)4$`Oz^k&Ks
zACfP{?t3dyJ+N~@{x)4)$&X^E(D*|eM=Fwni9|AMd??Aigv0dco#%Ph=IXcsIa)o`
z+TsK+*p&J>VXwxQRp9))Z5Xuir#4i+oMh7Olidx!tNrc>hhG<mS;eD>bu#LsaHiCn
zDnRQd?6H2N$^bv!>bnuXcA`AO(Z}#N8m_1ZY-3+>SPY&v;GgDqhMeAjIm=m!=RHdL
zGDsRT7`Ua#Mm~o>Y8Kf3c&3*eP%Md_%&q=4eK$#IBJ)x3tp|VUIXfIucfnU}R;)p~
z%#5(JUKa_bslNcBUQeP+AIU*vEmf;?)C-Ts4K2>VwIvUS!vuk|G>^13>r)iBrE)Av
z<XxrbxGs-A0p~D+_zqI6)2AZvnWaAdKF_Di8VDu*$!*;U$8|X5o!{NhmldypG_urJ
zfcx0N_&4Fq;}olFm@g`Ovqb}3ZdPv!<y8kcrp#Vh%?K)>2Cycya4VZBe0zs2#UiiJ
z!o>MzbAJ-sd=jTx8sm3&%J<lW#K`Rr9A|p251SKi`nw~R1omql3#=#wCv{r&7E5?x
zW27UQJZT6T$X`6k6E2xxFb2!}PjqrMeJn)vRw7c4tQQqNVG2!X2-w-d+v2#z>=G2K
z32;^l9Gm@wCSQ?_DY4-&yVnB=zDezISAt%J&2z{bYt(g@aHtYUbMwpd1*iXzmY@YU
zaDGZ<z&W&(psIU!t)dH5<V%WEg!?&>uu1ZJSVTcb?;Lc0WZTN~a{MH;ng#FB5t8O8
zXm7`6a3lC~ncyPHv0gi$+fhdFQ{$-mEQyDIZe_z;>*aIY5G{%jyHcQHmPv*!k1bb(
za?V0u^^ntWlijWG)}zc$`3R5Yahb_-G(#4T9nV)!d#sTpt`wEuSkhRfGH4TRD{_*`
zqXDhz)mIOh4Ak2{E49KI7*#WWD(nn_KReG<z*T#p7yi^wV9=7shNK3GDDGjR8zm$w
ziO0Y#d~6);efc`yl6#qOK4eiJzEFLekqXO;+Sb0_yls#aI%S*0vbXr2=GG*|uE|1&
zZlM>SWbFpUIm|wjK3}I@FHJoUaX$_2CIXgq#?gwIec;P?S8Zy!+CJuuuZ|<>w7<U!
zQA8sH*x~Tc!YGMyD`+>|zJK~u+NXDLXe^cyuWF)i<0eLD7WNtw<2N~_LedcV_YOg3
z7Q{YUawNM>&N#x08uoWm-UH`|C&yo2GihGW`98U{I;K@$HJ8q<_1W%TG{JJtqUeK{
zPWM@s+DFh%Op>e;m=^73@3!MvCTMX`*Wyl1P`N$Ct-db!>rjQ6<%paiDm{_3T<m;E
zpd_j!--v#*#?(t$P*9NW8R$Lp*5?P_6pcOquMd54F5X_gyhKKUYIMl7`HJ{8Qy1j~
zz*K_YNeW8RA)c&w-FuWF+geVRTVkR38gg8k!n|iCzN?x6-LjRw;#In?H3U0jc8e&R
z_9a#@Yj?AQeE*Q|TK$k6m<5)ulTi=H-X*(E9Q6@(pGHSYMKObgD9$*XUWhCK|C<Lq
z<(zpd)6I90DXR=Tm_<%1x!vxvk$M!2Oxth7)~^M{81C#QJ%1+4x1k=B>C=;HQT)+6
zgQXtdE5G~E`bSD>T$fBfH<wmItU?dPH!9@iNxR-KZGL)*D}JjgR1~9xkmSA)HbdVF
z4_>p1h`h7BI$c&)b5>p8j6s88F9Uu=`U-v%@64WEEA;*1p!-PlfLVn4eSG=A>4Hlh
z_9pEZtvW}qN*BO@E3Mf)%iK|(6yIuR(LVA0vMimehue|VoxO_OD&lup4()gZL0GO0
zMc*1p`N)C&C_)<U?^*A;>3&}A7xh+7YH)q~?sniKm8)bU)~#eb;EL^9?TVR|l$wwn
zwr%Y-o)wG-A$Q-?e~G*^HLG9w3xFCxsz3KL<wkxMI!_i=WuPIi-nhM3+7K0Q$+CdI
zBwV0Ao5JE3q0m%a&_#$P$w8O{cKq7m(?iREPl8(16~@P;@xe5Vk<dcY;s>pS?41H9
zf>_TBDFMT>tO5VLWLRq8PWk!KOxvEX&>NzTVDDXmgtZCFu#;XP_ZX(2N)tNlB6Jzh
zxaovHp?H4d16=ik&6?VF9lG`x@CBi>!1LR=Mj}J6yX*QlQ-l%d5}5%*Vn%xyBlw^B
z{548nbpn>)0#Q<^)6a9cRX+ft#a6^HL3~GJN^7{{DChypPaZr$&axMU0rb`I<wuTr
zjWV|@vkN2;f-|G1bVQMXvgQrH2HjcWogr5Q;=6WX@mnd(MqNl%Q3&E(J0vYP2oAGe
z@8mA@D<(?x+16ME!Y2UST>rvdy~2v$M+lWN?`S&==aREaD!kRB`#RMj6^V(jA>J0E
zuEL_{=xe>@o+gH+;9TLbU(i)J(tZ@?<M^b=t9%K(k`{-Xd+*XP@kL(4-tRXfcYi2W
zV;?#+oT<E#2>IyBN?36L@0gU>A$o8ZJmrEb<A9p(j98Qn;L|N7YUZX(q0di?J(ngj
zOuo3eux{piMoOFGxsVnS@<Mb#e}smaS*&mEoT#gl$<o!-8$%nSb?R0+&B$SpS%vVf
zMD6oNXh&Va(N+v9VpJDSMjHV+**&K#9rNwM?I5mUM4r=*|6@5EjMc?CsLW3{?>6^c
zi*s8qKcL2~9`o#RvM6!Ic}#q*MHPB%aiR~JDrDiy&46cmxI7;$Uw@_-po}bsO9s0`
zr=|RCS9qOE_@@+*IvmszQ=CcLM1&((y7M8x=SH(|1Qs5$5*N(+bV0#IurA`RwC{l;
zY!!8u`OTPl^_r-G<Z_gO0JN?l=EXff=#@J+TRS^>hqrewO3ncPyl{KHLsIYe-lU5w
z>F@<4N5jpb*+ps8cciK$kt(Mb(g}0$6l4KWNX0FbZNb5c6AMBU#10bRk*@gYcQd#E
zBoy~xUYVblG`KY`K>ED*mq3G6MP6C*>`R~B)enV`7{Ex^n*Z5w<T9yW+pqB8qU{@x
zVd0zeA&r(s)Zi_{xxw#R977_!WVbi9D}K%!ffJ3Ct_`_cTOG6xW;^ZABKMRL=WY(9
z6&>!K0xyXUY}^$QZUI+P1P;87pm+EViQPbQpU>D!<^21Cw>HAhkt-#Y4{kdUT(xZT
zEQ0`k?q%|PBq`N|PEIWJwg+%blyAW)x21q};O)t8mnjyiq(WF&TBCa7)>;GVcd^j^
zo;n8!{3(#7uPTbs@<4?@!!HH^iY|yLwBRU-I=JURN(BJd$Zy`F2hDhiE+`=Ar@Do*
za*4a~BMBlk%#KXn#qq_x7=WvE-<=#iY5OdF$O(ZDr*aM_Tm4YoylzHSF#*e~W2RZ@
zWZf)CmkP7xsS-7EIQJ6g1_Cx5DlPV>GtF`n9VGj3%>zt_mIe#<x2}MWdP{@2cjxOw
zje^j2qVH+JF-di*D(<A2qYI$+ReqZ}Pq<h=+lPCO_lVk-{afuXvuPF4LTtpm2udDT
zdVZeUBg5$pYlUmnssSufJqfp`jYx)*BmIbV=MFKe;IAkBx0u_nMkePad!TZ;y6zBR
zq<|CIj;gde4|1jS#?5dy)S9C)2yIL7RrY>3M<b#M#axFf)_H~7N`L8|A{+3zNlC}N
z*X8Q&hST$H5~=UaMjSn6<Ge@N04#;TG1P{_jwA&Sn$Io$4$jtTC$@IQWjof(pB@Zn
z-!;Va=)aU~?5q_tRk(utZyU1AQ-1Tqru@ARWuHC&>aw3d^G}s-xW0cp+0;5=hm7zu
zol-WPMnr%3100+qpM+IO*g)J9f8x#KVe#h2)<Na6<GIUddlh8=eJx=BKr!f9m6DH$
zVcZB&`~_ewfuQO3WjskHy-e$$!_pRQh#vXBx<j`j^X*|Rgs6B;gyYXVJlvPJdE2gm
z@NUd@4ZezUNdcExqCklOQMW_!W>WJY8)tYWAp*`f-1*KB2sw-#*k3g>1{dCe9t_ap
z`h5>Z%xm6if8srsBDsqYtm5-xLMBWk+fX7~uZo=3MG3<6aBpDW@^zTzysE_fDA=3H
z7%LlN*-D^NB_2w}B;PJYbJx<(Lb#6mogQb`#Y?-W2mzPIuJ}otuwsmSgvnY_n)`<k
zoyW&JYNP`y+^Q}GqDu@Kp{Ea(CSnJuJ@A3O8G3<;d^}N6PTFix(*%R2FkC0(b>}`8
zL}Iq1k7wT`idV_#wW$w%Qq(vG7hdtJaawO}HQjIJ#nW)4$m9rdM&P?`$drC2&0v_O
z#R?;dmqnbKpxMc%F$iLSPoeHN4G&304@9~G&diD-F){j$N^;mVjazKZOFG<RSeZKy
zWu9!=6E|Gr3rA-Ll=(uHfoYgGX;bbBUhy{JUZI~9rOAq>kPalMk|$(F1BI~}enbvI
zqcYrB?xK|ftpEeLQN=iL@t#-bIw7cCrdyd@xO-PUr?8aT+umHc?vz<na&!@4_L~>D
z8|pz6ToF@KHME>#a`^Z|zX#=o`Aqq_g}K;_n)Yzw)o|FQ?<R7kz9fUl_KR%B2G|Fy
zb4eRy0b*AN(TaWk4YVdLzEYU=8gMHE?)93#N2-k}(q9bd1}uALWFqh{A6q(7rxCWD
zFg7dZ5g(3vbYt{GbkoGvlc+_?rdUyDJJ@>!Kz6e4#uV->@&}s&$u}!(cpA?!+h5=z
zZO!oL?LdIbL6o+o2TSna*jo7L_OYiJ)8pb%cTbiT*y#kZ@v^I{9L89!fNPr5*YaMT
zogJU|T(<B0_guZk)PAG$OeBAO9fO9w%Uq_{#%3*V&D@I9ld1i^p2TWm_Or`LOko`6
zbdOF);fig}a74KPK51}FVI{{IGs<=pbbZuis8S=`Ml=C1B!6~`R2bmRV*+j?v`$NB
z=fZ<1QP^OC+_wogR3I~cd!R3t?jV)9R{N#IH!OZ4DNR$KD@h^<rm*MDW%GC<gb0o)
zTE!&UQNkWQR#(%^7FP$d@ViQ87WSvlRRfEzJGo^QP<Bs;jwG|9(Id+HyvNBpZ}d(Y
z9gQXz0(p9s4C&qbQwb=lwAf+=XNuS-#73nS+n46rdL4b2#5>4lQ9vGU!{}Ny<Mlv+
zIZ<g(lip1sWj7Q;1@!Tby5uEMl|wQK3VYT87s?O<Qa<86)y;c~FiKlJ+oZ+XXZ*qq
zq6jU@R^DYwVI6b-jl!nvcP&-rMxt*my!~Ur)y`$|I3@7~thrMiyhIx$y;xZFCcKS^
zVg$=Pr&|`YcOE%oVM0Q8-iq@zeM>b5Q}~FPPIqWGVH+8plC4Nyr@#r!nOYMp0y;z@
zHN)Zy8OoqhB+yDxDURyuGvU(vWyeJ-<D)VEM^4FzPvt~UPDapGjNvb-t@knc?%RhO
zz<0Z>Xpnuf#j}-cL)y~5em;@6*h;K1M-NQO$gRAj5YJM0Pv(uhH@$b~$J?nQ<b@MI
z0G9slQxq$6Es4IhCWvF`_hst5l3|%0U-C+5v4_<8&;YwOiVWhD$*g&ARyc+^P@Gqn
zH$|ue49U;gAfPDxBEGleaG_qQk)uAW*;{R0M@Ge8&U(aL0$##NMSoFlTP%7R&)C2%
zA&zeYv&44%NphoFKItmxUG*sa>V&i3gqzDu3lX@t5dL9OS>Xs26if{?$=iOumd5W0
z96RTQomYihC=>t2KIsPBP_pA{vHG3%iUTm~s!d?Y^RAYPGj7*raCxXY(S>j4nI-LC
z{b!g+`i9NrUd3hREPx4<#6CzfMqakW*^y3KWB5zYW`xR;*#D8B?%(|=;mmgp>^Pn%
zmYswVn9US3Tay!q%esRTpJ4=cce&F%<NtT?{vQk+Zh#2c%PeK@W&L5pgd@JywFB4w
zUjPTIUz4I6=h8(se{%KxJDCspSuT&!aPE|1RF3d-$zOo&wfXUwQSTTI;Ry%)OKF={
zHyEn)zw@inIko1ViTE!7i3FK_+so5ZU)Gc9v+AK{0@WSH17w^vWH%^@RSBvY(ZD?C
z<Q8z*d8x~8q$~Q)<zV%_)}DS#?bym8EMG05ep?H3*E5b6lHC)cA0*msw~hj9eP||t
z=-PT3*lvN1;`@lCgP8G{DJ=9k=%sQkG2IQr<7Bo{F6>ZPpMCI=YE4vAgP_9qkboS<
ze-j$KyaGQx8#0wI3_kcAhlo1uS2aA<Nl@S@tG;;dAi<A}d<GO&B#hhXaqDS-0FgTp
zbCAk?M3nQzXVhixmmQmz!n*t)Ff_ACj4uVu*9vU-j7B#Ws&~m_^rex3TT9Cn*&dco
z$2Fhf%3SfB3&^aLYCkuwMdv|>rs7@L`o@4yUd+fjLKsV7BJ=>>DMh!T4wkY0DGz4+
zAU~nUF`8t)TYvl^d1S|tv?>?}=Vn?D__3iQ%v-1qlk_ei%Q+Ff+JtaUpiVdj4#ogd
zcTM`}(*E?^T8lW33e^*;6na^hkt7f;GlW-JaW0T28PiRU_J3Sbn6Y~#d!VPZa~@K1
zD`>QGeXnr2lGp3@;OXIlKOQoh?72qUXMUGd=+_S?RG{r~?O5*<GvN_}72__l%5eLo
zAwV8w4Eg%j-1u;#<K65c7u7Wp-;9b&Arspi;Hp?)?<=GxfMPBQ#W48{UB7XeNF+e$
z&PCU=*S+90ED!r3F*nwuCxDdI?-4#a=dR#Lefm={83{|%!RL<yY~0+}0at?wD_B>E
zC!U6(AVENwzhc<uXPtzjED@VW)6ES?$wW{2Ie$!r|Cw5?NO@(0@V%k_iB8?I(*vyt
zP(<f`yIRll;TDQc(CB<*j9?<TTT!0QE$+n|)4@u%us_*sx#s^hs$_MgwPto*xTSg(
z-{i;5b@tgO_=O>gl?qLCGH-Ta*P$5GvbV!KC)501M{6EA7k++YGZsu%V)=%HeePX#
zDDTVHN+2;-ysCjadabhc;+11WP5D-LE+R0T?{0Mq`X>HL$9vyrexAy8?1FZTXg!;c
zL(1G`APj}(K!4M_<_2EA7<3BA<7zer*{gXXpuWxWC_5QF186*CH}Uv3CdSkfr}x_a
z<JSHiq9(c|&<8GiBa1+1juefv0qF0iD<-;Af~r*5Ar<lkyy*u*e<|UqU&UYRkmr>i
zr9|P;KFp7xUok=j)=B-ZbRu{LD1o-+>+KrkuNG*cwMgX!G1ttrKon<al%=dgP<8Dk
ztN%><{qg&jbY)wntwuga#c5rHq+X)!Mj+RI8y0>eUG%?p_`jtE3$H*7%5?PMyy5w1
z`WxGJm^YP)a*~eiiUvRDUxKePwh|+Hx9yM3FVME;&<*n@OM_jo&|gh5c0YwPo#X^=
zWBnUpC`At?R|v2E7o|(-{InP9+nb8N0CCHkKUKZ_@qjN8&b-Ng^z8ZfzmxX-XK#-G
zB0cL|T|9xK@^T%*^B2JH&TCj8v{iepUvvhT@6FLM(f3kyMfSMfPR+})DkN;seljBF
z;HUv@cgP4>ei>3hk5Re7AQR*Vk#){TkZ%O!9Sz6L^OOgtklhw?f3dWGZf}2PTRC{J
zu+9vRdx0y-<lZwl(}2S~pXtyaw%IVy7G=557INRn@$?&T&@st2SX)X6W_xfs0@`~j
zWmq5PWq3^{Ko5VhTE+%i+OL*YU;hP2@MJJ}5Iq<L4Xd0%@m^64!;1_p`GcZQI4c*5
z;Hv%kMU;e|Bw-xAapm)CdNorAE3<P>+<c}t&>8ZM2_8tlK}TMiVi#oax$zkmXRqY#
zRIYTmev|IL)#?K+{M^LF)^N~WQHW-k3VZi2(T<;=Z}pqZxwI4SDUp|Rb?I;2j<sJh
z)Wb|ahqjHJk6TlSm#Sk0I-XbE;nEvho6jw+=$^S)h4veo$Ewl3`$fV}I<W%AN=R}K
zpQXh;zfAQk^Y0AXdkDfQ=c%P!_)TaU!Dw52&rA@weIL&!>zI^jg<36pB3#wg*!L<}
z-b#>@d3&87m~<rZ{jvouJZ`7K&5{ZV5|#v>U=>Yv$nB;P-OW{$D74gKi*R#q^LeAa
zmRAE58gY}8%bW-&Msh(h<QTWGeb`2>(fR5NWhw>8Op|g~$In}}ywxUyG{*^}F>a|r
zd3Wcfg1$aImCXK}4HU-{b(lb4a|8NOcxObL6TBYLLv>w6^ONAM#}3r0GcW#z>?>Th
zzq?|D`+<Y@!7?!hU>z=F-VNG;_mO`~hSa^BPqkQILsoC=2ptacrG=Jv$4<cG$?tP#
z+#Pn;{qvOr3%SZss$ok@lKx`lNo-g#M0V_n>!SDUy~<Xqo<Q*0R#`MsG`!)AjLS9>
zngXD}R(&hI<_h5TK!Qj~Zb)%E$AsVc?0PnCoo3V(J}c>(gg#++pkz4hE<$%utb{QP
z3y2)I9t(!C#|W-6R}yMeku%zMS`S*@MZASGfOIU_yz0b9Un}&&K)ZLb2Xw`8u#Hv9
ziWqDE+Kp<2bMl+d<yHT~Ch$K6x4<$fw7)b_cHiX9iU>2&U3g(wK*7=}|B_ar_qiF;
zz4bwwIJD~>B`;_<&B@svs08ckd)9(qKFzDme7F49v>ULkdiHrW=*lOJofC9Wk-Tr^
zT@5>`&gq{}ivFm_y*B$7wJfN@uCBdR$a<-(2Im!d6rpCTOzQ$Z77DsgEzJhCv9boN
zq5mKjfgJ&LtSdH^zt}vtr2P<slA*X23iW3eHK2G0!`Mey$^Ir${x8yE|Jf0h0LYNO
z$mRLCNcbi4Hj1P_ed+9i*v8u6iNM?P0`sodT8~g8#WOhcVxwz_f@H77{mBdQ)mCA)
z=w}T>YTkwzkCx;QW3;!(e+WYTnX`UN@(o`f1jClJ%IdRUmZ<FHbd4~qjlG_4*UsU;
zi+bh+p>-gURtTtiVb=+RqUczd&t*7o;80M~xkISsl;6P!jCm4lnxG9~ypp<*pmN5l
zdtMB^_X3PMc0<S&l=NJsfwHBd<PN8PuQqut?)DyBh|XpMIT(<sAeFA|Mj~lW9hIh}
z-jS|X&~I8404ou}CBYe&H+kMIq=J4<uLVh;+jo~~uFvg|p=!Im7k<y=^456W$T8(x
zNu;$MtV-U1?Or6(;-f;A)L$ztnsRs1gzKf4e67M!8PD(sJ7c1`;SvD%Xg{|VUoQ5`
zt)L64Loz)l^3NiYN!gA^JDWFlQ<BrRL&_6`2WiHcsXI%>{~vX49T(-c{tXX3$j}W#
zBV7XsNJ*nKf(S@Af^?Ti4qeg>5`xm*>41QMpn!CPbhrDt(Y^OM`|R_)zw>_H&+~lF
z`_HVo*Q|BL_geRwJJ$MMN8Y=q7F2B+SGFG=zl$Z71uC<NN-u9@Nl0B_xLZaTMqgEB
z_{7wg2h{#h|J-g^)>wQ?Dn~<cG}f5+vU2rwTO+;SZlGjV-;YChL%xH|hf?ckJxYk^
z&{uzUPLQ7wqR`y42p00Np5@b6IUeScw5J}Tu$O*&boylZhuR!K4Wo)C5jbeyZRqU6
zMCdP`A+&<RWeOG9`&P;v=WTHPeqv&0sCwS_>b;*u$0=@D@5MQP5dM7m=tTBN595~&
zCkq-M?IY+{i6D`{m#~8kb6lz?FIR)UmPs;CY?YPr0px4Lbi*IATAh)sX3u6d=b(nj
zzJX&KwK&rAj-RJ%2b|@W^wW>W+Mc%vL13ablDb*v?=@n(zU>Wj`Z%4`%yhM!UzE16
z8T#J{Wfw+hZZ+X(9elUgTl3=Sa1SYd5cm}J<n)pwhn_sp1(X5P>tsMn9BnXcM?6W9
zBr=C-nNk7F88aA6oN@M=avT3xAou$jC%}M%E_9~sw^7l0D)oU0eW(>1|34#q3&(Ib
zjJ^?hhQBz`F1?^`?5)0K!J{F$3S8~f{A~{W%Y^vrKw@|6V7|!40$5f>eOb9L8@wa`
zK9Pr~Hx$RJx-Ra2Ix~SORu-kX9O0Mw9=_jg@P~l+N)VV4e>aiNw@3_*(LZJvPSHLC
zvfn_@&r0UL_M<M!TkgNC_NJe~UWxqw16lhse5no0^gNN(+*6<RZbXE~(M!1cG^drS
zU)pIv>~e;Izvn^?-r~2fi#|W1oKIbAf*Lv_COBB<6?hx4J)L#jFcQER%$iTthcxx^
zu%{c(7t7xJ)A6I|=@}<`vSE_BPjeWkX6J9dXwk)KHz({rI`-uRGd1U)XIC*NtZ&%<
z2CAGBs`ufM`8gdZz3hy6V6UVun*D$}a}WV`xw1|zZDkJx$kd14{sz)0YS`8Yxb0`|
zI71aB>aqxz6+!CpapNI748ldf%xkXbz@Wq0z2qIJ7YOwTAAxv#ozpTilbxMWJvqfl
z(3Xhc&iR;f3hO7%ba0(0ePGe5ti<5IwL3KTJp8HP;pT~sfV`rByYUaZ>l-14>^gm-
z?bdEYj9;YHy}csKJYeZH`g@e>>^nF30@MrNW<ue65qVK@3!H*<OCn39nF^V4iwH#2
zGgOS+MSU+<-TC{K0Jj<c>G{mm(=~_3K%SZg?6wdVddBgJ%`-I|FuH^|@(+<lK`A<B
z>KPFK&b}n28ff-V`BtLw52HEXGg*k){z{SgF19YVsvk~O6=-*=-nGSkpR^kSKi+M-
ztXC%+ZQg6KMe!4B<Y!<$z?d?tW5~S`Z%IvZmqFznfZHDoMGmb5Aw$(GQQEFaa;HIP
z^?m{@vQSMgYIZg9bsIxN#CMZKsCfQ&5;fP+IWZmcjptaTC+aF4jwD|%M(8Wh0;W#u
za-%kU@U=nma-*~+xwIyLMG#ad(Ojq|kY%R!sTY-|ZEJ^i`p9<xub(BbirWM@1PT)N
zR-CuMl64b%>5|_Y-PST43j<Vi{*{8c<4glHz|McWNaYPzs<Y!>mD|ID=*hROIITyW
zrlef-41Vt^2s(RTJ}F2r-$nicv{O!QfB^*nE?W2=wiF=O$Ky*_D^7mfJ^-HBpz>nz
z_gfn^ep%+@^d)>xx_B@8j+?;($FKe#TchFj(&avtfGjNHxzHz+41nd`YQJ~W05tWw
z=YEkJm_Y*s=K@xVdd&3-=1#O7s^7ZP{L|TPMASB*#CwksaOMX$RdY(&vGA$){*Zvd
z0?_yiP;db70Mc%40SfBjXM|o<er>ly>30my50<(H0SnaUeH%1A2ag(X3`CmK2&H77
z^qK)|eKFKG6z}$=%YCm*fb=o9Mvec>?en|zf1q&-%CE|vU+#FPl<kkZY5<y8f5rfp
zTBIHTe4_1g<u*P)19IyCAY!lOwQUUG)KCABf&M^X1AmWirT$OAabu{`C(}o^=6#(2
zO&$O@F#w<M${#lUjAHcAtuU&y<@n6xK!?j4CFKp3m#Vje1kig6#(j^SZ9s9E*ttjH
zFJw!}8c08#T~Yx4zSJ!ZZ|mD9DX|Aa3RnXang65-u$hKh#2p5hv)7gibSEpk9YoXs
zBq6y1@_-7U_!ZyVIR72~pMw;LPq4^sIxBC$J%g+M+W_AV4ge-x`?t3~jr|PzA9d8C
zW*k2w+!S={#xE>x2jA~-%K-+nx7@@0HN=bm9DeD)hEwrT>bm#p_9;J8P<Uy0KhSD<
zIh*sDaijB>n<l6hL{G9zuk|vL(0pM=yj?>JNIe>{`MtW$<qqJ9xajjv$b+Hd?ZS&b
zoXgvwSAbupi|ZdL9~FJiH}k+UcA2BNwd(v}x-qQf-d>B#kV)jm1#^(p6mGF(Zr^iX
z_G5YUfY!5sE8`qj!FW<Mj(_0``!BmR=8X0u<sIa<wtBfL%K5?aKpHU_{1tam^hfrS
z{E*!wYOPC|XAF0iOK-@Z!ui7QsrpTF<RMdpM%S@{g@`3x<kh!9m%2RUj!!f!GcN--
zCpsp%IKB50LUo*rvRXtIzH`vL*w`2MKQZvLUp-p-GB(WUT*P*k2qZ<U@U73^ymZ%=
zXSzQ&yaUj(l(`)3C4Vd7Ae}%&w>4XQ+%&kl6eDN_ZsE0s&JU7{lzR1g4KWlpv)W!h
z5kR^jjgc4YEKCiAZu}6vI*=EB5F>uAxg9#~ebwB|CCu-fa<sao>3rR0+VpmEedE9_
zcdCz}dS_M_`DJ2@RE3D0vES~y%i0CX&T_k&_sdLy2_BAnUx+5X8t0og_D0ZR*5|j*
z?mXa|9}`cBeMkAUwwr7F`1IK(Ns!d1$rG=!EU_TRC>wz6=IKykhrt~6d+%$Ur*$tV
zD7tl54SsY$@nmqG59OfU7bJOdd)<;71eNy#M<sy!JTpXgHsCXT+#$X}YS5dk4_l2m
zSylxV5;)`srHk?(GdCZ*^-93^+wGr@%?|lv^Nl2Zc-@;m901Jj$<V{&t&NkqdSe7v
zBrvN<`IFW+Z{oa^7Bu!5zOy<|pI3Iy5#9R2<;0cG;hs364PJ{QN)@=Vb&#Okg@$d|
zRDubO)**iDA}=wVWbm~C3IC1m@;J7XxLke<jv@<qE9ZRS8$^~YQkbO-H~L1nW*0~n
zdD2Uwl^z(*ii9^>+e8Kl02U|=A|sQcuQ!v<PH;|Em~&RPrjWT2x^ar^X}(}u(PjAg
zwd`TfY=n%7U2h&t&}1Z^(1{a2teFH3RH_2CvPCIp0I_bb$X)=@qU&r53v~U?+x076
z2U==Pqg`(i(2IVQ40tZR5vkl)BL|4}jB)v0+HJj6r1E6V@z45vK;JLfGk+=^YrCnG
zW7^jB`eh6wpz+p@vERG))f`c8^vp!4!ghbj8I(vey1a1z0yLp1Dw?<(yf`SG)qIG>
zbLXiP`KHF4%^N!Z0SfbocS8lkzqDoj;Yx4r-tkTs3@}IREF1=~4*se44~K72DDr!t
zaBGX$ZDnTl*#|hFUghTY=%=JVkov<yK-#6pAD;bep>lP_U_q<M<LlG>J<<iSxp~rW
z?BR^JUUnTYh@3bEDKRWld_y0I-_2oIGWbpJ?009?!x~AljBgQB{~iB7RE><a)1dwi
z=})zPV7(25@E_rl`W@+QX@;$z!Pxr~|GkztK>95ZGPXUK7P#ffz8Y?H$)@elB+^hA
zj@MovM+fUuBD$<Z<{_6L9&%HqO4rol4kxTEV3cV78XmWizBTn7PpmR8X9!ELpvkCl
zG-=Sj<3MX}4;9j-FV%BAW5zUeLrs319Q(V@zZf3>%`M1(aFg+WV=|ty^hY-JT|}n3
zSUamez;cQ?5B!>SbFnh;j;2eVaDyb8vL3e<+Bz|7p$1y~?qMdfufy1AvT(jOls((k
z$AweBU2(&(vBt^b5On1JU8ocjZuTsBvq%*!sz7QZM*^dM;|#rBT1}<h=Nh&(Zt9Yj
zFt!!fRZ~XBr>PJ5FjooCtx=Z0KQ#TpM26&lrB)w!hXL$d8X0(&Eig<mErxuGKjOJt
zmyEsG!kJ|<yCjr6*#0tC9`XQo)ZsMTLdH^0chFod+JumMy%#NZ@5y<LQiJAw`#YwQ
zP6x{`aX5;ov;B@WGj3=MlrDTVLJTTqC~pkj4~ZHSpgF&%5kMT@uYV%4_rf0W?9_Uo
z)bB}1{_vFf2Ak6leLm-ua^*X1(wU*+Z57aFm-#-s4)pxtPnlKsX_mYF2YP?JInly;
zPF=3%K_#eDYO&m8JyG^N)GO|!^XAg1ph)HVMvNpb?`iv38=L#yIwBqKEPv%4`G(*Y
z66I?9GY9B4-GDB1fvR$6J`YGGq^ht6Jj^bcBj=l;6>XL-ml#l~%feEyAKn-qHMV>y
zv$qi`W6rnR`aQo*{`e3vL>8ljE4`CsyP0k8-Wpt^NverIyzMW-{pIWUp&b@&v)bA5
zQi&iX&d79BkkI->CXmgOn(f>?O6*5}!4HS^9*hc)-LKu<WsC6d-(Oa5UTS3d5Wn}m
zoK`$p?RIx3Y2VHkQ<d-Tl>baHdD?N-Rq|q(yJ$4qtqBf)FjeGAwA$zV<Tp?u*Nfp}
zEv!B2?W%(f+xHx0@ICD16zk!GJqvIxPDE~tSLjq@fipwVEQzHd)kSDg`TWuv1^u{4
z)%|DU^C0p2u!v=4%C^hAk4Uz!fR_&X6Yt9+Ws~(f|8~A`pm;}}SCY8>>+{XOHT-$S
zqMcoPyt6yjAnlEB{-3=JmJPcyRGHfm2@<R%(fa?OghZ&_NL&yL-<M}Z_*F~r^4tGR
z{r`<K{a>C8|3dBsVUfMqK)-am$JPDIcQg5pc>Fi)29`J<x3G;5Xb*G+r8=e-eteAl
zqyw6I5Smt7zKj?d{hGua@D1&~9rb(pa^K)^<FG=H<)O`fxxT~s_E7&lbS7~+AWN$V
z-wWg^<O@M=TS_IX(R7f|;h^=}%<FBqWl{9iYJx&W1r>%x!&`Zpq*QAy1UGMtNRQz8
zUcEDteL0n|>y|?21|_8GOdPF|!(_vGIkD>4tJ$M=Zb8`drh>>QpQ3VD&$F862M6C5
zfdBh^efjfPyTP?qRcFkX6O6Jp1<8?pq=CeGFZuUn2aEF%Vcx!=Kz|}c-N4f&&FSi5
zlo&ft3<oUbUE~ZQK_7XvLLXngh08;sw9;n?^X_(Mm8-ope{2jFq0Q_@$vu;I{Dncr
z15M3MA?&Ne+nUp*4&SLoZgPE*Qr-5N%q}?yKB3>kHXybW3M}V+x6|!_b0gi`os}Lc
zx&@-nIlvtm%X(nj$P}t?vuLOy6zrGNVC_&}x;Y^?%xpe-BCb3^v7DCOyL)n}%XaKH
zg@G7Nv3yBdt)|<!QcWtb&wRWcgtdp|HXDPWpg!J-70+4T<{!EluhBi;YZ=!TLeX>N
z>?M9-+!`>`9qWi!dxS0gt%iYWeRF_Hvvg(lY*Z5@<7WkiI}D<z7*yg7XMPZe+yJ}g
zK&a?nvF0T~?QibeZJCSQ@Nb1D-e2Nh9Bv^|#K1DF9v>!1JlNY>{ZOL4vi79R_iRW^
z8DD=C?}pT9#>f0imXd7zwfpH6Sp<8&4aO%NU$@m*%gwbO-69zozP-A4vTxl^u_oo)
zH%vG_O)ZRwK~$gE095oQtbaQt!a1NQjqrlv9&ubvArb_mnf8@K;j4!Z!-yzXmQl^W
zwg20L#E74ApTlrNd|m?rxUBjr2Ne8XeTMRVZGUTKRq}6#{$Kup<@#(a=#%pC&Mo_;
zU)egXMMgktiZ-a|y2WfRljOql|HlO-MA`uLF}|WLNQt56cyB0>IN%{Z4bntAGZ%h+
z6S$a)a<{TYGFQxSJIdX7#C?v<UiR%?3x0aVM%l6MYddDbAvZEqtMEAFUEKWf9{!R`
zN#KSomNdHH-%oJ=$urKs`1gP4K(=-$vt{+nlUdAkc{IT}hI(Fnb2)ws9V5GHv}q?$
zMEiwp%QI)^XDd}$xm8DdSkXcYSi?O&k`sEaBi}sNv!Iy(xiCJ|A7nQL9|x_`YlC(T
zL7<+5Bw<*I5vI_cgxEDFeG~DrhbP#N3arxH)v%&1nY5ho5<knR_}pa>l0UFrF_!Y!
zXSBGJCmrzoz(+fF0MfF}AmxhaouK>RAhy(9<sKEi_Kp4XW#f*Pj+0%Z<0R9!`Zw}y
zlP^xS6{nV>G533UKUsE(yS$(M{OpKK;keXoY;xU@Qp{an@O8RmjX|~99FNYKrDCdi
z$acMQsN2hdXiX_B83B)_3qtp^T)c$lHeBXX8*FxI7W2p73p!cvuS}2lGn)29;cqfq
zAmeqw9jSF8o+d1?*$5m)vUT-{1=)~&nNa)<BW{eP*>t`d!QOZg*J<bJ7jD{u?QMyy
zi>b}6Z5&Gs5VzumS-3n}8u@64v<y<-O|##qp1XdDe)znX4#(HG&4uI|Oj=+QT50IH
zKaWm9&q6ovC{ZyzaR_}I+RI{T<<0)_P1Z|~N;`981!ZQ`ug@hfLbG9l9IP2439PB%
zE7IDJHl<02mK(NNK1ZWg2Z~ZZhk<-<q8iQTpPs}I+g9nW&sIfVeEkg+hq?%n8ZxiZ
zEts-UO~mBIt}=UNfsX>2NRi3j6MWYBZR@*8{A$fN+oSEn_Xp{s=`5Mh3dtJX?Ngn)
z6vn{^^M3tYK3aj8=;+Be>nnGE^a`=DhvIelR5ej);U8HMe5kO9Vx0<mWE^+}lX|YD
zPmszQtwCTt=@<ZcJsLS!pB~Y*%~EMB!m}6Qr<8g#__q1JdBX$#zLl`Y+>iLyKGGyR
z***=f8e!fgpf<ykBspaS*FW`6(o4=3kQ@1CU*S%hdc;ZkB2Otb^ausWH*ZGH#8J=-
zRZ9WqGWa^BS9Gpt;#AB}ZG8PY)avPXt%6jR?f1RuYf`CMSeS111I;9-Djr65-d>@U
zKtV7~h6B7r+uwXeaX_Jfh9@g2HcSNnA}sq}S{R5)-_Oa6t{p6Wba#&T%|M+}-EcA&
zrq-5Bm<6q1n_e7?9oQJ1C6b9H%ox9zah$u-0AS%Ne|9-lnS>oUHK1HIYIAO1oqTdZ
zm1P|AiJSf7s1IgRx_RBaRS3c?{YA2A06776o)TeR94{qD)P{xB{U%SP)Yy2{RAuU9
zDNap9a*}Gwg>e|YHHmGV?cCF$wZ))E4#7P7cVYL?MYUrWWE~NDPP9D{Xb2#;1%d<9
za!s;X@CNGL(M{M%)FC6;O-2VpzItP%7<{BC*6sUkeeP6IY{E9xron6|+eb%SC*+G!
zqGjFG?w0&;HKI{eIl->>QMGacq@hSLm3%IVE}5tM;PXBFJI%bpZ1Qi-RExmSWVi)^
zic&i}4s`6mxHF1V{JJ@KjVJgVpUR8XWqe%71PAqL#$m<j#iLrC7lcd-apIVjAMwb}
zpk1hDySpEzE1OAYm|a4`GX1hfZRI9x6GOIJ+~QYiW=<97SSz%n%2@Gq>7!+M^V8XU
zqsd2~acCF^*rF=d^;IKC*Jtp_odx|sTHh27<0YAq95jSI%uGRf3P{cgA&=vtnDRxq
zx#^jLKUd4D71E@5PC15BxX~m;iz_ezm*xETDL-^@5^T5Eehl~NtEUUVNWGYen#}W;
zg~05QDTHzj-}w<=%Z4o0?r_W|wQ$tShIb1b)~eK{i-nJJb<*o(3QSzJO^7QmLg8kC
zB5R6Oy_oUhwu}SU>su^rQTG)(d82c*5|f%VZHVHXAog6x%Z0IKii7zd6(33k61e$p
z5TN&t!Ls9$Q-jg^lST7)#p<?N*L68gJspQK#q}C!qggFj23~ri@*s%tB26NtiJ0B+
z6!COR4|iV#PK)K%Xuqo$7&ai;OxkMF-?5C=<Xx;zZ+cNSZIob>ZIjB2buUTp%6!_c
zy6>jMC8bEIdN4!}#=|UAZb1KLbR;srtN_eNyz+sDWc;&As~G+;E6Y}NI%$^^s=H|q
zr=Oj{gM<hy<H>0}(<WSS6SU|XBbNAiZ6_0@OmUmcz7eCn)W}c(S&!NWBbYQEzLfEt
zF`j|Nr)`UI-P3UMXzYaAd5u%M<ErkLdK#wTee@Sv#ck%ZHfc=#=?tfx7TFAbGB5c%
z7*d&-L&%c8Go{7H$r3`9Sy1=e^towmcuP`}-B8vJ)jsBYk`s?&Wk*8LMjqpNbZ&YO
z>|Q)z;aThg5?Ox(2l3{@l*5ez^q5#jD`2}_qe>Aj(qNcfO7u2QcMej(HG5aDb<K;r
z%kF(mbj_Rzdt2t=%ANE=g;-5vy!)w^iOve?!e-7)KM(>ewmeFi!a{34mC*h;c_MYH
zx_fr1!gg77=PODFRe6}UogC?vChY_!<byTjRsJcYbXdWo%6Hf9<~@|)^Li{gpB%6a
zb~FrMF1KJej|Z>xhmWz5Cq>OCa?Y`wg&d=TR_xJv+IEk+&>mu)(4BkM^KlKVU4~lN
zOS<V1zCYwv)N{%W(9aU7@|ot`JnT`2zjx3dTlth|pJ^K6lTMz%|BXo@MAL#`X7dr=
zN&#|M3_ESUNMZ^Mrlz8R6x5^^f;B7&f4&m9CxY(sL9};)ctGhl5K>Q?3p8UY7=I$L
zAL+YOj_ss4XHm2cw~BuN)#P)Ef{y3SW{>(SbQ7LPsU!~+D3Y>!CNmL#WY&<8|K8wW
z)Zp-j2YlqjSUIThXh6eU!&%L!*I$*Dn%uC9`Z>?t04E;?(I3)D)H@o>*(0245SSOP
zsH&PUcimw>)Az5O_d}dHg&PAhqTffSBpk-A7ijET-o={91}HiM8K2-ddAUIvUHBfo
zefU9yPu+}8Ng_5gjDvQ=Fk+$~i5w5Z@5ZX>iF`>2t18T0JmCH6CL^2B1g_i{*87jf
z2=9hJdfnskg#RtqGyc+NU2$ZgY?bjSEPNFSYwbja@#(QK9QvUC{-`Kf5YkM*<?@2i
zg^82)9M}1Ge!H6CV29y$TBI*KJIfPuQzo+P^B!LBQ~X6gF$9{f_gFu95w~=k+@(1J
zmucrm%f5z(*}5L9YCVU-D=bjVC#q!GvsVZj_>X~H$Ei>Vbr}8@x{A>9aEkzjgp6d~
zx206!Mfph9H#^2j#bJoIowg4@ud#W=Ee~|YW&tNV%Z1E=PY+n`uqZ|uk(A^w32p68
zO-*fNA^41Ms`9a!q>`LCb793oWM|4JJXDpP_jn_v9}UhR%oE|C<zdm+Z>=k{qh{wR
zaBa08h5hL}M`dXNF9E@vDO%4;m1affWQPTl+;C46JE&T%3(N(iVl=(vAd3iLbhf)}
zTAm@-REW{AmV$o-lHR=^BuKQ#9XtM_%1IxmO8UelK<N`|s_qXtiGoFjBU9Gm5ZQ{Q
zIDb+w5rmr+cK?o(1wR~C2*E9nHn364{;2FYcy^*+HAzjkj9~K9>NvU*@jBAMCvf@P
zLP^5q2_$=>oznM78Z0Kp#8gO7|4K}ytrV@I*mFn&&KN?)70Dp?I8q@={3wLatb(1d
zs4yg##LAwO0ZVf-;QRc!(kK0jAA^JFZspBW%Gf1*DoyKet;V53hs2Ev8HoYgY1G<Z
z37}tLHOCKQ74HrsIf9a`c?|hhn3<_yd~puUj1dr-EX<5;(!2EfW0~k}w7yzOQuzg}
zYBxuEIzg_5w6AS^<QXe*@)a_RSufvWay+Gnv-s=As|=K=@4mI@?<+LzQ!K_dzbj|=
z7(Dcr88V2@0S;K1?1Up9vglNH>s{Y}0~2X1_x<jyH}^h6>+{q`gmo}1uP$RKU7_nq
zEF}rCOtvH`*(Wis%}dQ*aWy@2AsG#7JN=`jTB3W<Yx)N6;pFs3T?wg4sp$N)&ZSox
zg-vf{M!Lte5EQNAuPQ|QY)uR#o{2j-KVZ{-tiwv-(bMI|f8X;F>wPPG*_71OlyENT
z;5!gz*|Q~aTF?gev27Z`I`Sv~c-9yXk+o$CPP3qy%Ahwj$B{kc{p6E)OmEg!^BO4(
zp6gmrR^})vehA@~jzH&Uo(aQYh{1|!xGHbj_YdxalJej4)qXzy(xS{zc0z`%uB_)V
zl4CXgGa#K8iIS08;OZPXQwNY?Ii>J?d384(WqEo*JAJvm$XEaa?y-APDfl6@tu3rV
zR5GlMpY_ioC4D*E7MEMKF|1AK!n^RNfb`{<<R)8OJQMq<JH^2vgSu$lrMHJ4&Ta1i
z<qHB_q=CUj1HCFW88q#LS#)uqBjs1Cn`Z)A^#~enR}mD%huZ`MOy?tKp?)O-7zu<v
zj>@l=Aj${1|8ELFB1{kv=mMlCQ$FYwz(_8d>6r(1evSby6Z%-)CNle3cmmCc{%G_U
zWe3Z>lYlqA7fc-o=zu+i9S7I8fEkE>O$li&PGp=r(<nWnzSM%1Mj!aaqHjC^1P}K9
z;h-cQRHeySX-ND6=!yQZd08qQrH3s2-xQ)ARD*!vy~?z|plM<s{SU1Fiu}JlY878&
z`W3!@ARvHJ9D@gxfRi3)m$#-(0{(j+{0ZVeA;j?$+<&T|?U|<_R!73tt^o_N-$w7#
zAJMx;mo{LQ9!OH8r)BHO+JGIc12*H?z6YXx+wNC!P#&WJ==JVy*90aEK-QlCBVx!8
z8wC;6SI0|3e*|@Pxe_q<Y$+XO0AUJ5dk=^f947c7_|M{RvGZ}gY6TFR=Y9#q7syNp
zgvsxJfC|W*1n9li0J0GO4EXOZqH`kOsRP0@!w8AiUxp8K-zig<xw&!KlK}*uemyVB
zMtgm$g8bG(-~^CH#N?(gOV7V_0f59F98c(V!S@H5N&v=>?OK)H!e4R~fD`S9^1zhP
z+p}ZrUb)>eu!526;eUCO>AKDI9KNT&fodDSB>o6lr>4)GsA!VR_Fc~o_$<5;?Q<5g
zD|(|mWbfadg`slEMb_ifn9+;>tb-#|5Rbl_#bL1@mxWNZFlf6$Gj=hOcKQ&8>2uP*
z14TYt3!x1XtHiaLd>MCB(xh}=ATc_8lg2uor(N$iBy!5ROeo!Y_4MNfZ8F~vZJdV~
zU)}nKDFr?KJ~7E&nE4#s9UPu|VEvjik`8tE$Iy|V&CA2`lt+1!@;0w|-#SY?eB0A<
zKPX<uWTn0PuK59@hU^$V0h1lk-SBxoR4^Z58EN3bdO-n|7r7EM<7%PoRAWvc#;Li!
zU{t`)w=Lx((t(ye%8K4Aj;Rsly5S^l5mAku(^rtbNZvLpYuKPiQIEx7ke!4{%>H|l
zE3~?UDV?X<a}|^2EI|Tc0LiLZB6-t$wf-uFa28s8r3;w7JStJ*%?$o;AWxEQ5oC#o
zg@ilful2t^7PcYQs?=1?B6>@(2|q}Dn+TS5$QZ1Z$&X&qpkEK4gB!9uW5my<i{PXH
zNxULq$ombn^Ns1p7t1NdPa6&Ili!h=&$@rC(&f5(k%#H-x@Zy(-g(SCx&7Hxz<Rz0
zNmxY2QIVw4nJ$eX5+n5J-RGdx=|}E!?n7BSsrL2HEL0P0EBm6^UxbV^1PqAFmW)A>
zN7+|Iiy0-DC^;T4B@^-N+4c4l8?&T@Nf5;PeAAIZt$IC@6DH>&kAt-ET24giVQ<9B
zpg9!A?laj0<>8C>?6<0<@;oQ3D8p@_Zm&)bE~BcdSH)MV_%ZGl;9EFgUL$`c%GYjM
zKNx~^xMrtBz?{9y;E@}`Vic}J#C~!b(rX!EW}D>v8%Qr?_$DikX-xd##nD|X*I6!>
z<q$K<eOY0(A3wmkXnB!JM09TamExJI-BC!Q5)UQ0d9Er=I~mzx*<MjIL`Ou+MBI@x
zK#8Mq9ZuUiKw8gqU-S1D1oGj!HS3v}mwmCWl?U^56BqhS9Y|wx$uQZf<6$~=RL4JU
z>Z{VRtw5k94bu-n%*Yn2`;%#9@j&3b3x_WHuoPrTT6N!tb|KS^P6tB`6Ws$^`t)ce
z-wygfp7+aqNFU>4@PDB;)yIyZO;q6=dTuh9rJXs+MaHIzS%|FMwe+|v$Bkp`5M2(x
zaIx15f^nCLA)?EeYU8m(cHp^_rS=<n`)J$8?N+IFHIk6(X6ShiJw60a5tn*JT}DY6
zXA_>ee4+SKdgsY#J*B+)`4e}2$LZ}Wn7gg`paGvfOMY~3^@RQ+hpuwfNeeg4?sndk
z`(!~s?+>%hq<&sS?Nw>2G8Rjd%p&gBkM-`;YVnq-1hvYjbO>`rGmy~4WryUac@!GC
z?ViO3is4A{Rm^)%clq|&w}O!t<KM`m#qU%P<2Lf@!2$sCVLmsq5*`mp6k0i(kCi*N
zf}uUQQ6l=)Zt(5Av8tzoS5*^C_*`^UX=6?w1va@Jux3&vnb&EmccYrFbwf=sqcH@=
zdxA33cyqHE!wr+4e$wa<s;P0BOo&}-dc=TZ$KUeG(*@jSjm)2(;9TA>#>}#^HHr=;
zH4)#6HEv#g#>gL`x5>kQ{;r*k5TgDPRb-G3v>P?MOxTTcimgUEbV_lPYW0mOrtvgn
z7Ii1HY9}cD(l*_tncrz>{dx2NmutzCO*;4bP@+}%fO)_Kc8QPM3xHTkP0BPKsz>Hm
zDow#u>?MiBy81Ce@uQxLGmM#m1pl!v<ZTcc&uZi#bs&x=u9DK)lBX3;Z<HYUxm<#V
z$?{$^fnb>Um_;-eQ3TCefNUOu7X}iEH^vXa&wz9j8K#l)Vf8$`?#a&6r+iRzN;$t&
z@EfR(Ft2(}@JIiPoiE0THZ{`3i!}o2PeNQY*0pWZ4|EwCx7|cUnV!7q<mv}S@5PzI
zIxz+U6It_IBH7HXAGgx6;!7U9u4x<tR~WYoay<TS`+iR_Ir1FqaNnX3S>%XIdJ;44
z6~+LnL?DLcV3jmd&9g#N!F9Vp`4AK~bi7b2$w(PRzAcHlh<d@qotv=pXHHm}=XO?h
zlrJhK7gC8xY-E>-&pc3sU5XzD_-u&=GLhpC&I`Ou7g@Qk7I*tGS~TJlLcqM5tA|WM
zT`1uqPzARCR3e3yLyvPk5_09g0XN%e&~s!7bJsL(4;Z$>3sjJ@fcu1apo~q`H2^DX
z@ky*JQEqm>Y|CuEs*Kdx#Y`~aX>_uXhN}Bn-rZ;SC{hMTSv?Exl`~(1^=-<eI=-oS
zJwu)nPspSGL&m&tQQCKXb;<LQ$9`vDSrh#=nLp{I2Z|GuLaI4Wr-0svYH3sxaF7ME
zIZ`tv0fLFcso!gM&)uMoJDh34uHu=knt`{kb)7cTx9w_<{pq{1H7|u!tH~>*As61&
z<~<gQb3yP0u2HFLGlwKPx~@*tMslAT_$LL8z^Ta*x~6T<>dE!D{ZmsHDNEE^{1eG;
zjDqyy`;lP0MMcR&r+qpp=h#Fp$~H|}yO_&&lS^*lLOo7vvQTw`l?FHF!_RmfAhEVy
z&$)U8qWRQy>v^^ffyK_3Ve>=*UbCsP6fXUo$|+l?yN@vi9NC0S!u1+t_=*Noty0^?
zUHU7Jq(3!1)v;Lm1n+j0i4SB^CO<O-t1~<tl<+f=BT{oWV!()v;Jt=<P~&o99IMZ`
z?oHwn(N*;r`FQW=m#-J2uh8N-6Y;M|u0DlUV8v8E6HktiA7;H+vRVEZP$GN@Lk&(M
zGcxlk%W$0xmUBAD_MA{)jCX#;@dl>;!4J~N$V#RjBBUZ0pPV`_w5Ir!ghISc!4+|G
zpvNABeDRUZuRZTcJO@T;-1sr^xL2I0gN4V9g`zZvldgh=tXiHEer}`;Ke_eNM85^H
z;ox1WES~td3YtNw@h7-LZ{>;X>_cCfnK+<XFMrh)68aKQb0;anQ(%q6;<Hq)qm=H2
z5Ka4&RTaIvD9omj`5)A=^t@cB1Wj5g`-F}qLXc9{#Fil(4|*P9m!o%+?BCU9Q9L*9
z-fe#R`1?1U*n+!X2NK}HoBh@FU5z@M<gDGsDXI6smF8K2pgPoXs|u;d1{Pk75A8(v
z(gtM647B)0jARGt`k~_U2G4t@o;wozxbz=w%Ijr<r8tBfjazV~;ma}I9{j$DS&U4N
z`=T5E!8`BMpo(5?ez<GJJR#_12&63B=$7lm4;<w3^g~1PfvoK8cNhl;2Ya5<4h#%*
zhQJu)W2hNEbp@hBm<{|mp3b@U*$H#E&~E4#rPa_@Jg25kt7j-V1kZ#(hitt`duqPB
znck@)7MGYj_GeZR!2MG2nMp>&BPX_K#5|8ADE0oENtsX~f`eg*K~kzn+UxiXm_Q2q
zAxiz)hu3=#xXXFJ9QumbO}W_++<QUvPWugub*&P&<mXylCT!Wa?BvpPi7EQ@kww&T
zv_;koQ%KP~MjXgQ<A+IpnnvwB388CMe8b&YnF*<VW%)->YOjt9JW@7XI@WmBC-rj+
zzANbFJbLs{eT?z=J15IF>Ud<D8C8h3()Q~&H4o8Za{^<plM`OYa1#pJ6MozQ*3_ZX
z)`0;E%ui7vxL<yh_&X61zxVK15B5o9{qjWei`5<0Ai4#rE`{PEf^~b;*VoBu>5ov#
zWt?cT2NXSL-aVypW2`wYN=`^|w(0-idF}{x>KP}t7`M6~t$p9s*eiCHZ@+u&N#DXc
z{y}S1XNBknI){jlRu%{P#@N?Dy?uSj9wOwPFCN{pdIpa?ds(G}C0GPRQeQ`iDg@}h
zeWPH+_=H)FwmX&FiqH9Rz!OGK7#o2Z-y*Ub|8bD~>QZ|#d9=``7C(k4zjj!Efgbv1
zZ!_ZFr04UoM>rAV^0N2HsV(FZf}T;y`O<qC3P`Cf)>x-H?PrL*qNdXns$5o7k_ICQ
zz;fDGy7*$RmoQ!z6hm7TQFNcuDvsN{S5`8O&@i~ZCe>iMr(1x(HOEAqY)}w<>EBw|
z5li}sanZ**iYt&!MGOo#l?h=gZ*b%~4lL`xF!mGK&j`Uq)s`lN1(;wLSgSOhpF?Xg
z?TE4E@AY_ceboYw7;6wm1}+@y)Z%k|*96#!sUx60CyK~)Fww{sQm$+5%Jm}J8@MhJ
zv&mpDDQ#y2em@RFlhHdBoPv2>$O$nb5TwmHK>T?n9s<=IY!rT+sn7`d;bfx4f0f`Q
zE3484i=j}S)F32$1w%*Gju{X^=hwb6#hg?(^q?YBe#D7D<|)6k^iCLq(FhtjmzBxm
z+Gxh0sNDZi`QZZ`#@9VHm|;S(9s0opTKvrP5`L6*E|Q2`(dADe19dJEs3>wN&R+R-
z{E!Zad<3@WTiIi>I$*mm$L`!z)Wz8qNAD-LHh#AUt5<DVceES-@)DSX`zH2Opk7gU
z6CxGZ^3~$w+JCZ77)C<+qtC|<{p+S>K38lH3t78g1fRL{fAl&x0_!)BnYbW|ja&51
zS->Vlg6DaLs&8;=)n5(!-!0jd1rjb)(NMmsZN4yUn~Qh@yMtcVG_gVQPnK&h-Nd{s
z)~-iqj{hp?2xQ+T!T&Gc3GBlMLP{CrcbvB|VRn#S>_+G{Sogd5hj5qq*87koPGE3$
z+SczMq&st+^xCsMEiQUl@<oa&yRyN4g7!PETPi;-RP%0Wxj`<%yx}_)oNHq%*Qi1p
z#F=@9f!8dDpjaMl9u7xojVsHiV(;_fMG#M_X9Znr^W?Q7?KE)rhAi(%!s(3t@LDz^
z1$P{#9rV4X45sdddQt66E+NC3$?)+QLa8PL$HlZ8T;Rw(UGzN3!5xegp+Ol~P&pN1
z0}pAQ23$NG;q|7oedKZ6u?H$QJSUwYs3?WDWFs1N+Ndc0hAH}a<El~8jv$(HmpfDd
zSpbk!-Fa)cqb@98Ao@dSH?A#-1ml;>Cn2fFm+$5@-Re4^)3tSVRNv~7zSSNtjJc%M
z31xq7-Kf#V9wSng`qYuBJ%;GqS7rMRG>zMK{lac7$bp=6LRz;3w5sQy|A6QEx<U<~
z8taC7`$}iN{mDaLr(so9^vBzh?%Vo~Grfybjw21ghAja|?;^Cvb#64eW&~GR#6rn2
zoTV6Jjr*YkN&@<5aUk*H*KhF67WZknwuZCmF=9as3a3jn{rax2_ydxQJD1ws$ukqP
zWRF9;LFc1KpThGgXoIsT*L&_8ibREUPKx;xLC}e>E>=q-kqU%HEIg4vur+?yfBLc2
z0S|vbeLmS<#Ih{4_QI27bHw=4qykKC1;P$Vaexd?@m^q<nU}E2n@8sMd<(>5Vl29w
z*UBkwD6n(=dL?rFK{mQb;L$q~hdhw2d@u#L6?<d4Jbyx|6^cX78GwzGLh!>9rq|0p
z@L1z6&oh?on$E`5841q=%w_jGvzesSQ=~?%`;x{=r6Y^m*PZ059Q!jSJ#c}k`(zsN
zuMrgOh9{Z*i-D%{hD!A9J+>%0Gaz2Pfy}D!b07VOrSqwTSzDka#l-8SmMic0@MSn_
z&Zz8-aVjTzr<VAK6BhG}O>ijiI!^TTHCR(!HDmA6sil1CQtU5g#8vIPOV{Ccr-@ui
zRfPMlKw9|r!;x@tLMJUZFGzVpjPQ1K1+r-Cb!pB+w@+~G>gF*yJDj0eXT5kXJsFo_
zMXze}`&nMmI__P2_dB34%MZ27s|)Y8nuQInb^TX+QPw9)tv{s{KeFJ@$u~_NtEe@`
zzYbYOWc$?*f98ZXw|Z!dKBp42jssOkx2{Y+l<^#?u5$r_KvdYVz$XZVb^GzmZD)v>
zt|I8r$j5<MNsDaSzF{BbiaVklB@S)2!f_D^<e~QsI|qq!dxnc#@1BiD&xwv7_Q#+$
z;sK=mgHFdxuEM*oJH-T^vWoD#Irw>5z1L>7oC!vERCp)q2bGrgh<93EGye>{zcNs3
zsZ&Bm4fA>Pu6cNHP=wmaOfgxstq*KMRkBOnZqWI~NxZjRiaOf+(%%OQ`rSGdO_$<_
ziQ!Jp_-U_Y2)7Wqx$XPd)uo%v9isd-X>p9smRCn*qz+P)W%scs=o%pcdQOx%%hAC-
zwp(TXD7jg$Y<I`w9rZyUBG+-|w|}rRCwWVbDSDmki}WVi`kbI3=r*dGs9)2zp;hkT
z@Q+z5yfMqX?<!#_ZkxlO7$tllvNl@aGZKJ$F23>*4b41Ww$~v~`W1m_BwuRk{aVV^
zgEdAb71NLM5or61Dx!u1)H+KaWutT|#VN)Kp2hmG)W4_;==O8R0+>op`TTM&IraoV
zo|hZy6B7If+cT7DT(Jox4Ab2wS|*=At_JZ#ny-C1^p+hQ?`_UcJu?HACy@w77nLM&
zrWBJ;2F4I>#b}yzaycJt=M(+Y{|~Sh3kle%#%cwh*qb+){XHT5KQ;K<;wu<Ir}y6T
z)pSNVc$D8p)c?3r3Ll==A?$z8IWTFIr}*(Ge))f-gh(61nF?tC4fTPN2j7Xsfs9>_
zlyt=gK3TWq2tA*Rl`y>^njC$M9)37=29a3K35=wm`F4vX{J;DmX`AMP0gt)kq`N!J
zRz|3z<Dy<%KMEPtmD6yzS?)R(7?}7Qh%`#<ejTfplT)G4`J&q86a6C~iSx>2X<!Zw
zoNuf3iqj&C{CID3=mE@hF$Dv{gUD83=YK}cJ!l?by8cRN{PlB$RC&bl0d{#28w+0h
z60Qx;jg^N6sUF-kdm@wz1+Ow9uo)oXTbVdaFnkXiLgd?|Dlbn9Bo<W_`-kk2drwP9
zzZZU7*ediqeFn5ow%^&J&mjuDVxW}tM-WD=>t0wYSUwHf->ZKw%JLZp9u>+>E}#g)
zRd%^IJR)wK$jsH`0cmy4b<LuD<qZZ+VyIy+k)P~ZKX#b);kBPg5@(I$a>VU7n{<AV
zw(Ik%Z0k-Ir@y$ENJ+Hgpz)mVs%+up_VHCZt^cw65@Mi!Uzdq1CQ<atp13eQ=8l|}
zg`4?eC?_jAM~`I(+5uQCgLEFxb%Zp_+kAA&1!Um4g;EBOO6Rjsh+{I(vz5bqY>wbH
zv(+V|4N8dsDanT5HNr>3IMn5v=)>8-k}y2Mlsg$T(r^3F>hDB<QqJ^xPVv|e$xMf(
z+Wbg^@H5^TvkVnR!GdS{a1+6@z(!1b1cebD#-Q(dOTv}i53%Tzx@r1onA5krIwsWs
z!*sU57h2ipaJ%a;^+&&f<~)kJ49Me@qrnnrOQA<NctfdB#xU5cCet7p^Dn`4Z{%Gv
z9`FR&L^&r=_1;Ur7njPO!j{hB^=K*ue_0XHo-GwYjfeWkGEcffI%KQ6>gp3*d?$+%
zEa+WznChWi*OzIMYE(06k^q@fA4(P{ypq93aIu)a9*G|z<&bcE>=3_p-B&1O6b>#(
zk&#GIJwTQwGlAnxwHez{IzVnFs{IPLbv3$yn~!@o19e-WsMgIS5vH!CdZ~dIg5?+*
zRiYS;Wh;j6MeXq@xZ>D)*EG1Gn#a|L)s@Ufz>RJk)|$cVboc&V=YA#Ot@4)DC84cK
z5S0gzI?#7W`uPIw+`X(Z`Y*(Jn1q7MRi6aBWklk-57)XQMMk^9Q6UcY(z$wz;o!VT
zDZ|wsqwtV8xTMK~IJW|oqxRt7s>0otHCq#$>BT#!V7w0;gX8XPU8pU1c%)0>_d(#M
z4L!7Ve?I%W{!4}oD_Bpv`<5k~=co>JQ&}ntb5N}gHSX?-?r#q$w{jeR=nb@BSY)R*
zArfnb2f@N&P%PA^dVafTK^dU325cTX=z$hpzJQY9))yPe5nQCLuQytILqW1R`}`Ze
zF$$1U0kRD*;z^$2w@ZH4>SdeczP^MWG}K4y0sOPV?xFVEMXLv2{l6dQA5d>h02Efb
z*xnoSGNv@@pIZ+oN>D|ejtUK^Wz?0M#geR>oXTI?`i3`mSh(6cq6i;+e?vu0olW7f
zForr$ap4~G;`Me6(l(WE{TRFB3byyb`pJ!kk4?l9&!+UH<5P?mi1x8>)Y~Z+zLaCC
zU!32`XM2|N>xU479v6=zJ4i|Nnz@c5A<a%~Uh8>%<B-Lp3w;s?uwznlr?<bDdw9F8
zi2nun5$=>)NRbDF9iDW6sIS<#+%z^mSzn}?9M34Q+ucl#Uce-{7tQcQ#?Cu5RI`_|
znjY5r`#$+N3ZCA;ynAMH0y^}tkjzsInP2}F#(4~s96#Ie{&7zdM+rlg#<_3=q`Z0B
zcGXqlzZ%z6F~Sn{Z1xvzi5vy0Tydm~DWDwqpWzv65o3XW02arSS;0qN^`Oe}jJRLE
z{`q<PogPri%OFPqOgQSdlCiz_WaQSHB>b+M_m$L}dB{+KpKsl`U0zSYGvXriyj$j%
zM}Mfe{T--_<OfHo-+lp>{#WBUbl@BVCKhCzjjBPb1l7okp=iJEgobghm`fNW`<wT=
z^H``bihY$o(OonUh=(CsZjuU`fUR@?)$4U@_;BF8t5xJ@b@Y~~$9vfcmI}vRjeaSf
zUL3URH1aXt(Co#K*S#SW->3L^ql=70-5l!0e<)pl^Jpj&o`~6!vbFR+)k!qH(EeI(
zAe@%Gb-SrOhIx1dA++l;){g4@Ryo@9%=_BGc2BFK)}^ckRx}I!?tP<ai_0W>H-n0|
zNXhkpU&^rbY)+EOeI0$oyd4q!mFjDu(9?7=Zzz$TGE;uKoOlzn+K?QtPZGqOP0PfM
zrJct@!IHFJfw+}w(T_vBU>yc!brW@;iy~s^nslJ;a;?R4U|;jSw^6zt%EoyZ&n*zE
z{JP^oW|X+cb;D4)5(01fXmexKl3@Utksv}2W?3<7kM|c*DsnrLLso?Bgg^@te8jdi
zv2-nu77_f6nZmeVe626wRlJUCMv7hOQ4r|VD+)Vkkl<-vx8v53oa?#6slJB+c24yk
zNC+8uA8cW(KH!glAoF4ISK<Z9JhEScBjIrC=~S{iEfO+>kmd0N?2&9nh6Dld7NHlC
zt<-gtLaN*s2+m};K;V*a&SHbFE(RVFLL!cfr5<v#qNZm4$l9%C(xD52VF>hr2yTAu
z71G~8?*>;mocj-)K<{DC28=w;kI(BZGn5_-obJCYZC;OA-=lK8(fX?0uc8#E9Q5ui
ztveL+ApO<aNIvQ@bDchzTKxb6*lX<fn7DVLIR-7OxjpKoLZ<jmjZd>=A{V)%&y*n-
z_=gWSiexM4S|EzHVzN~h@h1;>e`=Ew2i6E_PymDxe`9iCKrbtfKR!O`Xzb+!e_p4u
z0mWB7EjbEJ+sp#`WEEMS=T}_1WDp4?52|UJ4Eou!0kJZfFvFT>?a!0Ubb8*(4YA4h
zA%j?<RSc<8bk`@8<zZTCaNJ?)qizY>5R&|*_0MVV@inp7NE|t@wdla}S5k;fU@G8D
zZzE~hOKHzyNAj9*lj)qZ$)GuxEfowm3{)&{x+qUhVO5DP$jeA0B25ruN7llWL8Jmt
zMEM1?BVhNSz<b4t(~PN5jPvwQV61%XGea+KLN4F+7j+UZl1iZwL~0)sx%z4PQ+b7Q
zneJDcSVcOtFiK*9jJsX-QuarO(#iOkBvD&OA-sb;K=L4XiV5R=x?&bA<dsc23aSt6
zNo-Ef35(QK8RA4=tmyf6zGC4Cq4QU4p?+u@$ug3A!CZ#hUU!vWLuW@nZyP>fS&dl{
zVLR<JfCO-IO9VW^SIVNL?J(u(MWuYhFOw3o$>2htVuWY!=giA$A<%(rKuezW>QxT2
zL~T`YLP3egYK7|1TJ9jT1%uGV>+b5ZkHZh{<3NRW6}P35Xc)Hwj$&I<aQfOIhKG)5
z^_Fo~K=M)Bsy!^da|``SrMX%qfmie+mSCe)NR{y`|KfmXOmrrmx{h%2cT94ULD_@G
zB1P&=bZ&hlM8n)!atb*|%j?FhQ~}p~>V8%-d5>9Cdi)6MQYJ|kFDlxxC-lkG<~%z{
zWf$;Ur+b>;xrf!vG3kH5i*M(-V-m-sYhpPsGK#{|i-iIqLZwLKa$3x86}&poNgt(2
z#9e0sOQ%KpLq7=bTE*`Uvfce4(rovTxF;pibpgq>A6~iG<;R2mQl!7Nism_&FfHv-
z<COtZKt9TeMc>5<wZ3Di@#(YI`bZPg%L4u~XskmmLhaMx;R4eB9w(L)o6kpWk{Vnf
z-lIf;ITR?;5KF=$KdOcB7ca<+;A_{jz_tzV`_*ooevS9^3C-uIhczf**)9_;K98!P
zqLzCW+q%);vJ5tg<j63G=i5ht*RJEtzF#Qd_ryfj!X;4Mv6M&9O!n$@Q6bp_G2Hdx
z6+$A1r&nn$bZJEf2V0%vFv0<);xZh6^kkBO9%UMT3it5vAsNQBv9R6ZDEE8pg;(e(
zIP)a}XU3zJUYrW)*kwq|h{)h*G@iY-KAl_A0S$ml;6!H9J5X@(3n4%dQ3y~FlmKiB
z<gGZx0Of(jv2dPkYbUVX`j#>R;B5H)ryQv%EGyFDTdw;jsFB2V28-Wq_kf;%ZS|*s
zZbW_6Hb8jrBZ~uQfvW+~K-_AB089@6kpZ6<l{&zna1NwexwqE|)FTgVxXSz_pWs9P
z$@wsnXDinInflbeyFBg`_$3hFb09R{rH(zT_M`HNsRFDL^*D&FzU9Jb3i?S9(Jlvg
z;=3CDr!TsvXJ>p~q?V_*uKZ~NwWuB7<Jp|mpSJ$)EH1U+PjQTDv~;CvlAT2NWM3pV
zc%_SsZ2c7Q$IpN=EP$5b)+hkVUsiJf_V2aL{=-AB)K;9fTRZ{cho2-KfB5hx1^^+*
zt*0F70Q15x?|=IEPZ1IIII~H(?R>NUM-<X=9W1VI1JpeZ^f~#PP;W8YxlqIMygxDc
z8P8Wp*t!8Y;M@{+pd-Q{Dibg^iRE8q{W&Cl1}U_G&kqRFAGU?Q1&sZtAtCB<YxG_#
z&|-1<!|5&ZD}NgbCqSHi{~_ZK|3rNO4F*?N{}N*V+Fyf%|DVRlU)6`BqTcH-rd`xj
z`$Al2urT$%ZTOdX3m+ff#&tC&o(*nCLJ|SVjG;aPh&2DRU77!nPb&V8UY%98Lue@B
zaC6%hXH&~GHc!wJhpf1a&?S&t6g1k9b+sKIF6!@FQ7dzZblN=w`L~7fyrJeZhXd-z
zIa!Q6JdW$jamt?KwXbI3^WsEmfxKHCbOd18fE5XMI;3wPs^lf%V|$FT{JfbMDU6V%
z{WBi)y+C>dfeNaXy<`jy)I)W9rz;A(uVQY>gWBC1(V<T%3fkA>#Y4gZgKEOTrs!mg
z_)=EJa3Q+RV*Yzb&&l!v=!o!TA=2PnCbD3PAQ(~6-NJLC{Nv*4XTO0MR%?NsKs_hT
z&LlQMT}rt>dc<YlE;RX#S_utjyu9IfdBkbdfi!I8onE%FsS*)BHq<Abkr=~@e>NLR
zH5A%A{Q{(iu{H-}Q}M57mw|H<SUe0{3^AHzB0<%p5Te`6lfB3dM+uUQ;lko)wKVDq
z#+(S!4_b~EW+)C4bd!+JI;gE)nGQih#p}SdqQ!$y`J*U31h-icG%N;nMzZ!7s}4Qj
zy7u7tfF~KTHqu2I-G%}&AeHUFqse@kA%ORBj;sRy0;X$JqnczP%f%YMz~j@H%>kBy
z1!6MdEmUG0kvtdVLz)kVkyG}N<$_>FLuxxqc}CGhIt|rTM;g3@gm_Y@V}}9DZBAQt
z7}VkTRB&o7!rW>cKK~U<sy><uw0dC#q(4qMS>Q#MS)U+NFS84IY|P@^TnJq&*b!m`
z?qCF0Xt6*W;9ZKyLs}EjdKKnj$K(Ee1CON1Vnx65D0QX=ofPV)G+-UH6wkQL;~`Of
zLaFE?6WV4mT4!+6X}3(Y^q1vc>0we#N3yOMB<4fak;Q+Tj~?oRJt=58fQ$rYv`fYi
zoXC7qK=?V{IXdM&g$2iYLIe&l`5J^@55JzmFo>w5$BATyI`$cZ=^AjFNL^&Oz9qPS
zOc17&h_-bO!taFkEyHAmWF$Rf!h-!n-CaN^&FGj$G7L@&_>XrlTdfnH<l{^GnOK;h
zuFNEiy~6sWpXfV*%&`Az!Hrh*sQdxu0mi%iE-x}M8~6*XuKTI((nM4LYnxM}bj6ab
z%b#@$I&*X!ZV7EuQaUjs^6)-Givv$43bt2nZ^+te%&t`?9WJs>#&c)ofo{PQ7i8-s
zx_8@8J6z-7Cga_3_rkerTU#FrY(BswzaWuq@AvaJcj?AY;NK>;VxEKDjfk$A?mac-
zTk_OT-xNNv`_aE|=E6_k00sA*wVPY{>e#;(8*BFN+2AVYdY|#j#vSwZQ~Z_Q%e!Cx
z@%?V^k?#+GrJR@tmkIk+cK^k%XzPY&HG3xlPaAW;boSs;@abeaciQ_al`8G(`j>8D
zkYMpwKY#HHFiV3DD-;APJOUQlDB5_DVFTYkx!!0s0q=P~KD%0e<a_vi3+SSliN=p)
z55EW9zVoJYzQ8;AYIAmZ{S==c?FGMY?f@=&R!To&-G1-v!Q()G*eTwz`<BON7k6SN
zNdM-7!cX5^&Ucja?*(0yvPmnaQ2JY~?!FMH3nCbt6$IY>idJLDiWPYdGUhUx4z(KR
z9kp+$?2Lgnbj1$kU;46d7cjWWfew?)>N>#0nB87}>83$b+=}JkFa!tJK)aLph$YLV
zqOJ$!FIK$IWnj_(C&-X627wbS*88TNY>=8xn*>OfI|-#~N)25F?8SB;ug+i)bO)XV
z_$zABz8dG<FLUyMXXJ5UWTOcmA^B<D6PrMA#snTB^!-6+qU=LZ&SP8;%w*k7R(pY^
zK>}lw_D1V>dtVSW(@)d3#;6NMgLkmEhey-sXc`?&qrl4{2Kl7GXg(dyr=w|fG@sJw
Wh@b&G+HBN6uxa-ZSH9K%zX<>&bdlo#

literal 0
HcmV?d00001


From b39b8e76d963301cd160f98185f479aaf4797605 Mon Sep 17 00:00:00 2001
From: "Sterling G. Baird" <sterling.baird@utoronto.ca>
Date: Wed, 1 May 2024 16:52:59 -0400
Subject: [PATCH 07/12] increase default steps for saasbo

---
 scripts/generate_scripts.py | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/scripts/generate_scripts.py b/scripts/generate_scripts.py
index 42b18f43..3dbd3810 100644
--- a/scripts/generate_scripts.py
+++ b/scripts/generate_scripts.py
@@ -132,7 +132,7 @@
     # ⭐ {"name": USE_CONTEXTUAL_NAME, "options": [False, True], "hidden": False}, # noqa E501 # NOTE: AC Microcourses
     # ⭐ {"name": FIDELITY_OPT_NAME, "options": ["single", "multi"], "hidden": False}, # noqa E501 # NOTE: AC Microcourses
     # {"name": TASK_OPT_NAME, "options": [False, True], "hidden": False}, # noqa E501 # NOTE: AC Microcourses
-    # ⭐ {"name": SHOW_METRICS, "options": [False, True], "hidden": False}, # i.e., visualizations and metrics, e.g., optimization trace, Pareto front, HVI vs. cost # noqa E501 # NOTE: AC Microcourses
+    # ⭐⭐ {"name": SHOW_METRICS, "options": [False, True], "hidden": False}, # i.e., visualizations and metrics, e.g., optimization trace, Pareto front, HVI vs. cost # noqa E501 # NOTE: AC Microcourses
     {
         "name": SYNCHRONY_OPT_KEY,
         "options": ["single", "batch"],  # TODO: add "asynchronous"
@@ -173,7 +173,7 @@
     opt.setdefault(CUSTOM_GEN_KEY, opt[MODEL_OPT_KEY] == "FULLYBAYESIAN")
 
     opt["model_kwargs"] = (
-        {"num_samples": 256, "warmup_steps": 512}
+        {"num_samples": 1024, "warmup_steps": 1024}
         if opt[MODEL_OPT_KEY] == "FULLYBAYESIAN"
         else {}
     )  # override later to 16 and 32 later on, but only for test script

From 0d4539c76c9c03e591fcde58cce231fb63969d7d Mon Sep 17 00:00:00 2001
From: "Sterling G. Baird" <sterling.baird@utoronto.ca>
Date: Wed, 1 May 2024 16:59:05 -0400
Subject: [PATCH 08/12] update gen notebooks scripts, etc.

---
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...ical-False+custom_threshold-True+synchrony-batch.ipynb | 8 ++++----
 ...cal-False+custom_threshold-True+synchrony-single.ipynb | 8 ++++----
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...rical-True+custom_threshold-True+synchrony-batch.ipynb | 8 ++++----
 ...ical-True+custom_threshold-True+synchrony-single.ipynb | 8 ++++----
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...ical-False+custom_threshold-True+synchrony-batch.ipynb | 8 ++++----
 ...cal-False+custom_threshold-True+synchrony-single.ipynb | 8 ++++----
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...rical-True+custom_threshold-True+synchrony-batch.ipynb | 8 ++++----
 ...ical-True+custom_threshold-True+synchrony-single.ipynb | 8 ++++----
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...ical-False+custom_threshold-True+synchrony-batch.ipynb | 8 ++++----
 ...cal-False+custom_threshold-True+synchrony-single.ipynb | 8 ++++----
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...rical-True+custom_threshold-True+synchrony-batch.ipynb | 8 ++++----
 ...ical-True+custom_threshold-True+synchrony-single.ipynb | 8 ++++----
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...ical-False+custom_threshold-True+synchrony-batch.ipynb | 8 ++++----
 ...cal-False+custom_threshold-True+synchrony-single.ipynb | 8 ++++----
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...rical-True+custom_threshold-True+synchrony-batch.ipynb | 8 ++++----
 ...ical-True+custom_threshold-True+synchrony-single.ipynb | 8 ++++----
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...ical-False+custom_threshold-True+synchrony-batch.ipynb | 8 ++++----
 ...cal-False+custom_threshold-True+synchrony-single.ipynb | 8 ++++----
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...rical-True+custom_threshold-True+synchrony-batch.ipynb | 8 ++++----
 ...ical-True+custom_threshold-True+synchrony-single.ipynb | 8 ++++----
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...ical-False+custom_threshold-True+synchrony-batch.ipynb | 8 ++++----
 ...cal-False+custom_threshold-True+synchrony-single.ipynb | 8 ++++----
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...rical-True+custom_threshold-True+synchrony-batch.ipynb | 8 ++++----
 ...ical-True+custom_threshold-True+synchrony-single.ipynb | 8 ++++----
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...ical-False+custom_threshold-True+synchrony-batch.ipynb | 8 ++++----
 ...cal-False+custom_threshold-True+synchrony-single.ipynb | 8 ++++----
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...rical-True+custom_threshold-True+synchrony-batch.ipynb | 8 ++++----
 ...ical-True+custom_threshold-True+synchrony-single.ipynb | 8 ++++----
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...ical-False+custom_threshold-True+synchrony-batch.ipynb | 8 ++++----
 ...cal-False+custom_threshold-True+synchrony-single.ipynb | 8 ++++----
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...rical-True+custom_threshold-True+synchrony-batch.ipynb | 8 ++++----
 ...ical-True+custom_threshold-True+synchrony-single.ipynb | 8 ++++----
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...ical-False+custom_threshold-True+synchrony-batch.ipynb | 8 ++++----
 ...cal-False+custom_threshold-True+synchrony-single.ipynb | 8 ++++----
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...rical-True+custom_threshold-True+synchrony-batch.ipynb | 8 ++++----
 ...ical-True+custom_threshold-True+synchrony-single.ipynb | 8 ++++----
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...ical-False+custom_threshold-True+synchrony-batch.ipynb | 8 ++++----
 ...cal-False+custom_threshold-True+synchrony-single.ipynb | 8 ++++----
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...rical-True+custom_threshold-True+synchrony-batch.ipynb | 8 ++++----
 ...ical-True+custom_threshold-True+synchrony-single.ipynb | 8 ++++----
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...ical-False+custom_threshold-True+synchrony-batch.ipynb | 8 ++++----
 ...cal-False+custom_threshold-True+synchrony-single.ipynb | 8 ++++----
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...rical-True+custom_threshold-True+synchrony-batch.ipynb | 8 ++++----
 ...ical-True+custom_threshold-True+synchrony-single.ipynb | 8 ++++----
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...ical-False+custom_threshold-True+synchrony-batch.ipynb | 8 ++++----
 ...cal-False+custom_threshold-True+synchrony-single.ipynb | 8 ++++----
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...rical-True+custom_threshold-True+synchrony-batch.ipynb | 8 ++++----
 ...ical-True+custom_threshold-True+synchrony-single.ipynb | 8 ++++----
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...ical-False+custom_threshold-True+synchrony-batch.ipynb | 8 ++++----
 ...cal-False+custom_threshold-True+synchrony-single.ipynb | 8 ++++----
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...rical-True+custom_threshold-True+synchrony-batch.ipynb | 8 ++++----
 ...ical-True+custom_threshold-True+synchrony-single.ipynb | 8 ++++----
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...ical-False+custom_threshold-True+synchrony-batch.ipynb | 8 ++++----
 ...cal-False+custom_threshold-True+synchrony-single.ipynb | 8 ++++----
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...rical-True+custom_threshold-True+synchrony-batch.ipynb | 8 ++++----
 ...ical-True+custom_threshold-True+synchrony-single.ipynb | 8 ++++----
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...ical-False+custom_threshold-True+synchrony-batch.ipynb | 8 ++++----
 ...cal-False+custom_threshold-True+synchrony-single.ipynb | 8 ++++----
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...rical-True+custom_threshold-True+synchrony-batch.ipynb | 8 ++++----
 ...ical-True+custom_threshold-True+synchrony-single.ipynb | 8 ++++----
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...ical-False+custom_threshold-True+synchrony-batch.ipynb | 8 ++++----
 ...cal-False+custom_threshold-True+synchrony-single.ipynb | 8 ++++----
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...rical-True+custom_threshold-True+synchrony-batch.ipynb | 8 ++++----
 ...ical-True+custom_threshold-True+synchrony-single.ipynb | 8 ++++----
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...ical-False+custom_threshold-True+synchrony-batch.ipynb | 8 ++++----
 ...cal-False+custom_threshold-True+synchrony-single.ipynb | 8 ++++----
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...rical-True+custom_threshold-True+synchrony-batch.ipynb | 8 ++++----
 ...ical-True+custom_threshold-True+synchrony-single.ipynb | 8 ++++----
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...ical-False+custom_threshold-True+synchrony-batch.ipynb | 8 ++++----
 ...cal-False+custom_threshold-True+synchrony-single.ipynb | 8 ++++----
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...rical-True+custom_threshold-True+synchrony-batch.ipynb | 8 ++++----
 ...ical-True+custom_threshold-True+synchrony-single.ipynb | 8 ++++----
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...ical-False+custom_threshold-True+synchrony-batch.ipynb | 8 ++++----
 ...cal-False+custom_threshold-True+synchrony-single.ipynb | 8 ++++----
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...rical-True+custom_threshold-True+synchrony-batch.ipynb | 8 ++++----
 ...ical-True+custom_threshold-True+synchrony-single.ipynb | 8 ++++----
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...ical-False+custom_threshold-True+synchrony-batch.ipynb | 8 ++++----
 ...cal-False+custom_threshold-True+synchrony-single.ipynb | 8 ++++----
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...rical-True+custom_threshold-True+synchrony-batch.ipynb | 8 ++++----
 ...ical-True+custom_threshold-True+synchrony-single.ipynb | 8 ++++----
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...ical-False+custom_threshold-True+synchrony-batch.ipynb | 8 ++++----
 ...cal-False+custom_threshold-True+synchrony-single.ipynb | 8 ++++----
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...rical-True+custom_threshold-True+synchrony-batch.ipynb | 8 ++++----
 ...ical-True+custom_threshold-True+synchrony-single.ipynb | 8 ++++----
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...ical-False+custom_threshold-True+synchrony-batch.ipynb | 8 ++++----
 ...cal-False+custom_threshold-True+synchrony-single.ipynb | 8 ++++----
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...rical-True+custom_threshold-True+synchrony-batch.ipynb | 8 ++++----
 ...ical-True+custom_threshold-True+synchrony-single.ipynb | 8 ++++----
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...ical-False+custom_threshold-True+synchrony-batch.ipynb | 8 ++++----
 ...cal-False+custom_threshold-True+synchrony-single.ipynb | 8 ++++----
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...rical-True+custom_threshold-True+synchrony-batch.ipynb | 8 ++++----
 ...ical-True+custom_threshold-True+synchrony-single.ipynb | 8 ++++----
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...ical-False+custom_threshold-True+synchrony-batch.ipynb | 8 ++++----
 ...cal-False+custom_threshold-True+synchrony-single.ipynb | 8 ++++----
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...rical-True+custom_threshold-True+synchrony-batch.ipynb | 8 ++++----
 ...ical-True+custom_threshold-True+synchrony-single.ipynb | 8 ++++----
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...ical-False+custom_threshold-True+synchrony-batch.ipynb | 8 ++++----
 ...cal-False+custom_threshold-True+synchrony-single.ipynb | 8 ++++----
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...rical-True+custom_threshold-True+synchrony-batch.ipynb | 8 ++++----
 ...ical-True+custom_threshold-True+synchrony-single.ipynb | 8 ++++----
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...ical-False+custom_threshold-True+synchrony-batch.ipynb | 8 ++++----
 ...cal-False+custom_threshold-True+synchrony-single.ipynb | 8 ++++----
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...rical-True+custom_threshold-True+synchrony-batch.ipynb | 8 ++++----
 ...ical-True+custom_threshold-True+synchrony-single.ipynb | 8 ++++----
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...ical-False+custom_threshold-True+synchrony-batch.ipynb | 8 ++++----
 ...cal-False+custom_threshold-True+synchrony-single.ipynb | 8 ++++----
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...rical-True+custom_threshold-True+synchrony-batch.ipynb | 8 ++++----
 ...ical-True+custom_threshold-True+synchrony-single.ipynb | 8 ++++----
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...ical-False+custom_threshold-True+synchrony-batch.ipynb | 8 ++++----
 ...cal-False+custom_threshold-True+synchrony-single.ipynb | 8 ++++----
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...rical-True+custom_threshold-True+synchrony-batch.ipynb | 8 ++++----
 ...ical-True+custom_threshold-True+synchrony-single.ipynb | 8 ++++----
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...ical-False+custom_threshold-True+synchrony-batch.ipynb | 8 ++++----
 ...cal-False+custom_threshold-True+synchrony-single.ipynb | 8 ++++----
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...rical-True+custom_threshold-True+synchrony-batch.ipynb | 8 ++++----
 ...ical-True+custom_threshold-True+synchrony-single.ipynb | 8 ++++----
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...ical-False+custom_threshold-True+synchrony-batch.ipynb | 8 ++++----
 ...cal-False+custom_threshold-True+synchrony-single.ipynb | 8 ++++----
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...rical-True+custom_threshold-True+synchrony-batch.ipynb | 8 ++++----
 ...ical-True+custom_threshold-True+synchrony-single.ipynb | 8 ++++----
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...ical-False+custom_threshold-True+synchrony-batch.ipynb | 8 ++++----
 ...cal-False+custom_threshold-True+synchrony-single.ipynb | 8 ++++----
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...rical-True+custom_threshold-True+synchrony-batch.ipynb | 8 ++++----
 ...ical-True+custom_threshold-True+synchrony-single.ipynb | 8 ++++----
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...ical-False+custom_threshold-True+synchrony-batch.ipynb | 8 ++++----
 ...cal-False+custom_threshold-True+synchrony-single.ipynb | 8 ++++----
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...rical-True+custom_threshold-True+synchrony-batch.ipynb | 8 ++++----
 ...ical-True+custom_threshold-True+synchrony-single.ipynb | 8 ++++----
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...ical-False+custom_threshold-True+synchrony-batch.ipynb | 6 +++---
 ...cal-False+custom_threshold-True+synchrony-single.ipynb | 6 +++---
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...rical-True+custom_threshold-True+synchrony-batch.ipynb | 6 +++---
 ...ical-True+custom_threshold-True+synchrony-single.ipynb | 6 +++---
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...ical-False+custom_threshold-True+synchrony-batch.ipynb | 6 +++---
 ...cal-False+custom_threshold-True+synchrony-single.ipynb | 6 +++---
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...rical-True+custom_threshold-True+synchrony-batch.ipynb | 6 +++---
 ...ical-True+custom_threshold-True+synchrony-single.ipynb | 6 +++---
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...ical-False+custom_threshold-True+synchrony-batch.ipynb | 6 +++---
 ...cal-False+custom_threshold-True+synchrony-single.ipynb | 6 +++---
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...rical-True+custom_threshold-True+synchrony-batch.ipynb | 6 +++---
 ...ical-True+custom_threshold-True+synchrony-single.ipynb | 6 +++---
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...ical-False+custom_threshold-True+synchrony-batch.ipynb | 6 +++---
 ...cal-False+custom_threshold-True+synchrony-single.ipynb | 6 +++---
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...rical-True+custom_threshold-True+synchrony-batch.ipynb | 6 +++---
 ...ical-True+custom_threshold-True+synchrony-single.ipynb | 6 +++---
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...ical-False+custom_threshold-True+synchrony-batch.ipynb | 6 +++---
 ...cal-False+custom_threshold-True+synchrony-single.ipynb | 6 +++---
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...rical-True+custom_threshold-True+synchrony-batch.ipynb | 6 +++---
 ...ical-True+custom_threshold-True+synchrony-single.ipynb | 6 +++---
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...ical-False+custom_threshold-True+synchrony-batch.ipynb | 6 +++---
 ...cal-False+custom_threshold-True+synchrony-single.ipynb | 6 +++---
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...rical-True+custom_threshold-True+synchrony-batch.ipynb | 6 +++---
 ...ical-True+custom_threshold-True+synchrony-single.ipynb | 6 +++---
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...ical-False+custom_threshold-True+synchrony-batch.ipynb | 6 +++---
 ...cal-False+custom_threshold-True+synchrony-single.ipynb | 6 +++---
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...rical-True+custom_threshold-True+synchrony-batch.ipynb | 6 +++---
 ...ical-True+custom_threshold-True+synchrony-single.ipynb | 6 +++---
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...ical-False+custom_threshold-True+synchrony-batch.ipynb | 6 +++---
 ...cal-False+custom_threshold-True+synchrony-single.ipynb | 6 +++---
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...rical-True+custom_threshold-True+synchrony-batch.ipynb | 6 +++---
 ...ical-True+custom_threshold-True+synchrony-single.ipynb | 6 +++---
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...ical-False+custom_threshold-True+synchrony-batch.ipynb | 6 +++---
 ...cal-False+custom_threshold-True+synchrony-single.ipynb | 6 +++---
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...rical-True+custom_threshold-True+synchrony-batch.ipynb | 6 +++---
 ...ical-True+custom_threshold-True+synchrony-single.ipynb | 6 +++---
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...ical-False+custom_threshold-True+synchrony-batch.ipynb | 6 +++---
 ...cal-False+custom_threshold-True+synchrony-single.ipynb | 6 +++---
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...rical-True+custom_threshold-True+synchrony-batch.ipynb | 6 +++---
 ...ical-True+custom_threshold-True+synchrony-single.ipynb | 6 +++---
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...ical-False+custom_threshold-True+synchrony-batch.ipynb | 6 +++---
 ...cal-False+custom_threshold-True+synchrony-single.ipynb | 6 +++---
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...rical-True+custom_threshold-True+synchrony-batch.ipynb | 6 +++---
 ...ical-True+custom_threshold-True+synchrony-single.ipynb | 6 +++---
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...ical-False+custom_threshold-True+synchrony-batch.ipynb | 6 +++---
 ...cal-False+custom_threshold-True+synchrony-single.ipynb | 6 +++---
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...rical-True+custom_threshold-True+synchrony-batch.ipynb | 6 +++---
 ...ical-True+custom_threshold-True+synchrony-single.ipynb | 6 +++---
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...ical-False+custom_threshold-True+synchrony-batch.ipynb | 6 +++---
 ...cal-False+custom_threshold-True+synchrony-single.ipynb | 6 +++---
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...rical-True+custom_threshold-True+synchrony-batch.ipynb | 6 +++---
 ...ical-True+custom_threshold-True+synchrony-single.ipynb | 6 +++---
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...ical-False+custom_threshold-True+synchrony-batch.ipynb | 6 +++---
 ...cal-False+custom_threshold-True+synchrony-single.ipynb | 6 +++---
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...rical-True+custom_threshold-True+synchrony-batch.ipynb | 6 +++---
 ...ical-True+custom_threshold-True+synchrony-single.ipynb | 6 +++---
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...ical-False+custom_threshold-True+synchrony-batch.ipynb | 6 +++---
 ...cal-False+custom_threshold-True+synchrony-single.ipynb | 6 +++---
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...rical-True+custom_threshold-True+synchrony-batch.ipynb | 6 +++---
 ...ical-True+custom_threshold-True+synchrony-single.ipynb | 6 +++---
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...ical-False+custom_threshold-True+synchrony-batch.ipynb | 6 +++---
 ...cal-False+custom_threshold-True+synchrony-single.ipynb | 6 +++---
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...rical-True+custom_threshold-True+synchrony-batch.ipynb | 6 +++---
 ...ical-True+custom_threshold-True+synchrony-single.ipynb | 6 +++---
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...ical-False+custom_threshold-True+synchrony-batch.ipynb | 6 +++---
 ...cal-False+custom_threshold-True+synchrony-single.ipynb | 6 +++---
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...rical-True+custom_threshold-True+synchrony-batch.ipynb | 6 +++---
 ...ical-True+custom_threshold-True+synchrony-single.ipynb | 6 +++---
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...ical-False+custom_threshold-True+synchrony-batch.ipynb | 6 +++---
 ...cal-False+custom_threshold-True+synchrony-single.ipynb | 6 +++---
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...rical-True+custom_threshold-True+synchrony-batch.ipynb | 6 +++---
 ...ical-True+custom_threshold-True+synchrony-single.ipynb | 6 +++---
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...ical-False+custom_threshold-True+synchrony-batch.ipynb | 6 +++---
 ...cal-False+custom_threshold-True+synchrony-single.ipynb | 6 +++---
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...rical-True+custom_threshold-True+synchrony-batch.ipynb | 6 +++---
 ...ical-True+custom_threshold-True+synchrony-single.ipynb | 6 +++---
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...ical-False+custom_threshold-True+synchrony-batch.ipynb | 6 +++---
 ...cal-False+custom_threshold-True+synchrony-single.ipynb | 6 +++---
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...rical-True+custom_threshold-True+synchrony-batch.ipynb | 6 +++---
 ...ical-True+custom_threshold-True+synchrony-single.ipynb | 6 +++---
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...ical-False+custom_threshold-True+synchrony-batch.ipynb | 6 +++---
 ...cal-False+custom_threshold-True+synchrony-single.ipynb | 6 +++---
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...rical-True+custom_threshold-True+synchrony-batch.ipynb | 6 +++---
 ...ical-True+custom_threshold-True+synchrony-single.ipynb | 6 +++---
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...ical-False+custom_threshold-True+synchrony-batch.ipynb | 6 +++---
 ...cal-False+custom_threshold-True+synchrony-single.ipynb | 6 +++---
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...rical-True+custom_threshold-True+synchrony-batch.ipynb | 6 +++---
 ...ical-True+custom_threshold-True+synchrony-single.ipynb | 6 +++---
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...ical-False+custom_threshold-True+synchrony-batch.ipynb | 6 +++---
 ...cal-False+custom_threshold-True+synchrony-single.ipynb | 6 +++---
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...rical-True+custom_threshold-True+synchrony-batch.ipynb | 6 +++---
 ...ical-True+custom_threshold-True+synchrony-single.ipynb | 6 +++---
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...ical-False+custom_threshold-True+synchrony-batch.ipynb | 6 +++---
 ...cal-False+custom_threshold-True+synchrony-single.ipynb | 6 +++---
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...rical-True+custom_threshold-True+synchrony-batch.ipynb | 6 +++---
 ...ical-True+custom_threshold-True+synchrony-single.ipynb | 6 +++---
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...ical-False+custom_threshold-True+synchrony-batch.ipynb | 6 +++---
 ...cal-False+custom_threshold-True+synchrony-single.ipynb | 6 +++---
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...rical-True+custom_threshold-True+synchrony-batch.ipynb | 6 +++---
 ...ical-True+custom_threshold-True+synchrony-single.ipynb | 6 +++---
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...ical-False+custom_threshold-True+synchrony-batch.ipynb | 6 +++---
 ...cal-False+custom_threshold-True+synchrony-single.ipynb | 6 +++---
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...rical-True+custom_threshold-True+synchrony-batch.ipynb | 6 +++---
 ...ical-True+custom_threshold-True+synchrony-single.ipynb | 6 +++---
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...ical-False+custom_threshold-True+synchrony-batch.ipynb | 6 +++---
 ...cal-False+custom_threshold-True+synchrony-single.ipynb | 6 +++---
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...rical-True+custom_threshold-True+synchrony-batch.ipynb | 6 +++---
 ...ical-True+custom_threshold-True+synchrony-single.ipynb | 6 +++---
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...ical-False+custom_threshold-True+synchrony-batch.ipynb | 6 +++---
 ...cal-False+custom_threshold-True+synchrony-single.ipynb | 6 +++---
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...rical-True+custom_threshold-True+synchrony-batch.ipynb | 6 +++---
 ...ical-True+custom_threshold-True+synchrony-single.ipynb | 6 +++---
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...ical-False+custom_threshold-True+synchrony-batch.ipynb | 6 +++---
 ...cal-False+custom_threshold-True+synchrony-single.ipynb | 6 +++---
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...rical-True+custom_threshold-True+synchrony-batch.ipynb | 6 +++---
 ...ical-True+custom_threshold-True+synchrony-single.ipynb | 6 +++---
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...ical-False+custom_threshold-True+synchrony-batch.ipynb | 6 +++---
 ...cal-False+custom_threshold-True+synchrony-single.ipynb | 6 +++---
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...rical-True+custom_threshold-True+synchrony-batch.ipynb | 6 +++---
 ...ical-True+custom_threshold-True+synchrony-single.ipynb | 6 +++---
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...ical-False+custom_threshold-True+synchrony-batch.ipynb | 6 +++---
 ...cal-False+custom_threshold-True+synchrony-single.ipynb | 6 +++---
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...rical-True+custom_threshold-True+synchrony-batch.ipynb | 6 +++---
 ...ical-True+custom_threshold-True+synchrony-single.ipynb | 6 +++---
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...ical-False+custom_threshold-True+synchrony-batch.ipynb | 6 +++---
 ...cal-False+custom_threshold-True+synchrony-single.ipynb | 6 +++---
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...rical-True+custom_threshold-True+synchrony-batch.ipynb | 6 +++---
 ...ical-True+custom_threshold-True+synchrony-single.ipynb | 6 +++---
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 8 ++++----
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 8 ++++----
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...cal-False+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...al-False+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...ical-True+custom_threshold-False+synchrony-batch.ipynb | 6 +++---
 ...cal-True+custom_threshold-False+synchrony-single.ipynb | 6 +++---
 ...orical-False+custom_threshold-False+synchrony-batch.py | 2 +-
 ...rical-False+custom_threshold-False+synchrony-single.py | 2 +-
 ...gorical-False+custom_threshold-True+synchrony-batch.py | 2 +-
 ...orical-False+custom_threshold-True+synchrony-single.py | 2 +-
 ...gorical-True+custom_threshold-False+synchrony-batch.py | 2 +-
 ...orical-True+custom_threshold-False+synchrony-single.py | 2 +-
 ...egorical-True+custom_threshold-True+synchrony-batch.py | 2 +-
 ...gorical-True+custom_threshold-True+synchrony-single.py | 2 +-
 ...orical-False+custom_threshold-False+synchrony-batch.py | 2 +-
 ...rical-False+custom_threshold-False+synchrony-single.py | 2 +-
 ...gorical-False+custom_threshold-True+synchrony-batch.py | 2 +-
 ...orical-False+custom_threshold-True+synchrony-single.py | 2 +-
 ...gorical-True+custom_threshold-False+synchrony-batch.py | 2 +-
 ...orical-True+custom_threshold-False+synchrony-single.py | 2 +-
 ...egorical-True+custom_threshold-True+synchrony-batch.py | 2 +-
 ...gorical-True+custom_threshold-True+synchrony-single.py | 2 +-
 ...orical-False+custom_threshold-False+synchrony-batch.py | 2 +-
 ...rical-False+custom_threshold-False+synchrony-single.py | 2 +-
 ...gorical-False+custom_threshold-True+synchrony-batch.py | 2 +-
 ...orical-False+custom_threshold-True+synchrony-single.py | 2 +-
 ...gorical-True+custom_threshold-False+synchrony-batch.py | 2 +-
 ...orical-True+custom_threshold-False+synchrony-single.py | 2 +-
 ...egorical-True+custom_threshold-True+synchrony-batch.py | 2 +-
 ...gorical-True+custom_threshold-True+synchrony-single.py | 2 +-
 ...orical-False+custom_threshold-False+synchrony-batch.py | 2 +-
 ...rical-False+custom_threshold-False+synchrony-single.py | 2 +-
 ...gorical-False+custom_threshold-True+synchrony-batch.py | 2 +-
 ...orical-False+custom_threshold-True+synchrony-single.py | 2 +-
 ...gorical-True+custom_threshold-False+synchrony-batch.py | 2 +-
 ...orical-True+custom_threshold-False+synchrony-single.py | 2 +-
 ...egorical-True+custom_threshold-True+synchrony-batch.py | 2 +-
 ...gorical-True+custom_threshold-True+synchrony-single.py | 2 +-
 ...orical-False+custom_threshold-False+synchrony-batch.py | 2 +-
 ...rical-False+custom_threshold-False+synchrony-single.py | 2 +-
 ...gorical-False+custom_threshold-True+synchrony-batch.py | 2 +-
 ...orical-False+custom_threshold-True+synchrony-single.py | 2 +-
 ...gorical-True+custom_threshold-False+synchrony-batch.py | 2 +-
 ...orical-True+custom_threshold-False+synchrony-single.py | 2 +-
 ...egorical-True+custom_threshold-True+synchrony-batch.py | 2 +-
 ...gorical-True+custom_threshold-True+synchrony-single.py | 2 +-
 ...orical-False+custom_threshold-False+synchrony-batch.py | 2 +-
 ...rical-False+custom_threshold-False+synchrony-single.py | 2 +-
 ...gorical-False+custom_threshold-True+synchrony-batch.py | 2 +-
 ...orical-False+custom_threshold-True+synchrony-single.py | 2 +-
 ...gorical-True+custom_threshold-False+synchrony-batch.py | 2 +-
 ...orical-True+custom_threshold-False+synchrony-single.py | 2 +-
 ...egorical-True+custom_threshold-True+synchrony-batch.py | 2 +-
 ...gorical-True+custom_threshold-True+synchrony-single.py | 2 +-
 ...orical-False+custom_threshold-False+synchrony-batch.py | 2 +-
 ...rical-False+custom_threshold-False+synchrony-single.py | 2 +-
 ...gorical-False+custom_threshold-True+synchrony-batch.py | 2 +-
 ...orical-False+custom_threshold-True+synchrony-single.py | 2 +-
 ...gorical-True+custom_threshold-False+synchrony-batch.py | 2 +-
 ...orical-True+custom_threshold-False+synchrony-single.py | 2 +-
 ...egorical-True+custom_threshold-True+synchrony-batch.py | 2 +-
 ...gorical-True+custom_threshold-True+synchrony-single.py | 2 +-
 ...orical-False+custom_threshold-False+synchrony-batch.py | 2 +-
 ...rical-False+custom_threshold-False+synchrony-single.py | 2 +-
 ...gorical-False+custom_threshold-True+synchrony-batch.py | 2 +-
 ...orical-False+custom_threshold-True+synchrony-single.py | 2 +-
 ...gorical-True+custom_threshold-False+synchrony-batch.py | 2 +-
 ...orical-True+custom_threshold-False+synchrony-single.py | 2 +-
 ...egorical-True+custom_threshold-True+synchrony-batch.py | 2 +-
 ...gorical-True+custom_threshold-True+synchrony-single.py | 2 +-
 ...orical-False+custom_threshold-False+synchrony-batch.py | 2 +-
 ...rical-False+custom_threshold-False+synchrony-single.py | 2 +-
 ...gorical-False+custom_threshold-True+synchrony-batch.py | 2 +-
 ...orical-False+custom_threshold-True+synchrony-single.py | 2 +-
 ...gorical-True+custom_threshold-False+synchrony-batch.py | 2 +-
 ...orical-True+custom_threshold-False+synchrony-single.py | 2 +-
 ...egorical-True+custom_threshold-True+synchrony-batch.py | 2 +-
 ...gorical-True+custom_threshold-True+synchrony-single.py | 2 +-
 ...orical-False+custom_threshold-False+synchrony-batch.py | 2 +-
 ...rical-False+custom_threshold-False+synchrony-single.py | 2 +-
 ...gorical-False+custom_threshold-True+synchrony-batch.py | 2 +-
 ...orical-False+custom_threshold-True+synchrony-single.py | 2 +-
 ...gorical-True+custom_threshold-False+synchrony-batch.py | 2 +-
 ...orical-True+custom_threshold-False+synchrony-single.py | 2 +-
 ...egorical-True+custom_threshold-True+synchrony-batch.py | 2 +-
 ...gorical-True+custom_threshold-True+synchrony-single.py | 2 +-
 ...orical-False+custom_threshold-False+synchrony-batch.py | 2 +-
 ...rical-False+custom_threshold-False+synchrony-single.py | 2 +-
 ...gorical-False+custom_threshold-True+synchrony-batch.py | 2 +-
 ...orical-False+custom_threshold-True+synchrony-single.py | 2 +-
 ...gorical-True+custom_threshold-False+synchrony-batch.py | 2 +-
 ...orical-True+custom_threshold-False+synchrony-single.py | 2 +-
 ...egorical-True+custom_threshold-True+synchrony-batch.py | 2 +-
 ...gorical-True+custom_threshold-True+synchrony-single.py | 2 +-
 ...orical-False+custom_threshold-False+synchrony-batch.py | 2 +-
 ...rical-False+custom_threshold-False+synchrony-single.py | 2 +-
 ...gorical-False+custom_threshold-True+synchrony-batch.py | 2 +-
 ...orical-False+custom_threshold-True+synchrony-single.py | 2 +-
 ...gorical-True+custom_threshold-False+synchrony-batch.py | 2 +-
 ...orical-True+custom_threshold-False+synchrony-single.py | 2 +-
 ...egorical-True+custom_threshold-True+synchrony-batch.py | 2 +-
 ...gorical-True+custom_threshold-True+synchrony-single.py | 2 +-
 ...orical-False+custom_threshold-False+synchrony-batch.py | 2 +-
 ...rical-False+custom_threshold-False+synchrony-single.py | 2 +-
 ...gorical-False+custom_threshold-True+synchrony-batch.py | 2 +-
 ...orical-False+custom_threshold-True+synchrony-single.py | 2 +-
 ...gorical-True+custom_threshold-False+synchrony-batch.py | 2 +-
 ...orical-True+custom_threshold-False+synchrony-single.py | 2 +-
 ...egorical-True+custom_threshold-True+synchrony-batch.py | 2 +-
 ...gorical-True+custom_threshold-True+synchrony-single.py | 2 +-
 ...orical-False+custom_threshold-False+synchrony-batch.py | 2 +-
 ...rical-False+custom_threshold-False+synchrony-single.py | 2 +-
 ...gorical-False+custom_threshold-True+synchrony-batch.py | 2 +-
 ...orical-False+custom_threshold-True+synchrony-single.py | 2 +-
 ...gorical-True+custom_threshold-False+synchrony-batch.py | 2 +-
 ...orical-True+custom_threshold-False+synchrony-single.py | 2 +-
 ...egorical-True+custom_threshold-True+synchrony-batch.py | 2 +-
 ...gorical-True+custom_threshold-True+synchrony-single.py | 2 +-
 ...orical-False+custom_threshold-False+synchrony-batch.py | 2 +-
 ...rical-False+custom_threshold-False+synchrony-single.py | 2 +-
 ...gorical-False+custom_threshold-True+synchrony-batch.py | 2 +-
 ...orical-False+custom_threshold-True+synchrony-single.py | 2 +-
 ...gorical-True+custom_threshold-False+synchrony-batch.py | 2 +-
 ...orical-True+custom_threshold-False+synchrony-single.py | 2 +-
 ...egorical-True+custom_threshold-True+synchrony-batch.py | 2 +-
 ...gorical-True+custom_threshold-True+synchrony-single.py | 2 +-
 ...orical-False+custom_threshold-False+synchrony-batch.py | 2 +-
 ...rical-False+custom_threshold-False+synchrony-single.py | 2 +-
 ...gorical-False+custom_threshold-True+synchrony-batch.py | 2 +-
 ...orical-False+custom_threshold-True+synchrony-single.py | 2 +-
 ...gorical-True+custom_threshold-False+synchrony-batch.py | 2 +-
 ...orical-True+custom_threshold-False+synchrony-single.py | 2 +-
 ...egorical-True+custom_threshold-True+synchrony-batch.py | 2 +-
 ...gorical-True+custom_threshold-True+synchrony-single.py | 2 +-
 ...orical-False+custom_threshold-False+synchrony-batch.py | 2 +-
 ...rical-False+custom_threshold-False+synchrony-single.py | 2 +-
 ...gorical-False+custom_threshold-True+synchrony-batch.py | 2 +-
 ...orical-False+custom_threshold-True+synchrony-single.py | 2 +-
 ...gorical-True+custom_threshold-False+synchrony-batch.py | 2 +-
 ...orical-True+custom_threshold-False+synchrony-single.py | 2 +-
 ...egorical-True+custom_threshold-True+synchrony-batch.py | 2 +-
 ...gorical-True+custom_threshold-True+synchrony-single.py | 2 +-
 ...orical-False+custom_threshold-False+synchrony-batch.py | 2 +-
 ...rical-False+custom_threshold-False+synchrony-single.py | 2 +-
 ...gorical-False+custom_threshold-True+synchrony-batch.py | 2 +-
 ...orical-False+custom_threshold-True+synchrony-single.py | 2 +-
 ...gorical-True+custom_threshold-False+synchrony-batch.py | 2 +-
 ...orical-True+custom_threshold-False+synchrony-single.py | 2 +-
 ...egorical-True+custom_threshold-True+synchrony-batch.py | 2 +-
 ...gorical-True+custom_threshold-True+synchrony-single.py | 2 +-
 ...orical-False+custom_threshold-False+synchrony-batch.py | 2 +-
 ...rical-False+custom_threshold-False+synchrony-single.py | 2 +-
 ...gorical-False+custom_threshold-True+synchrony-batch.py | 2 +-
 ...orical-False+custom_threshold-True+synchrony-single.py | 2 +-
 ...gorical-True+custom_threshold-False+synchrony-batch.py | 2 +-
 ...orical-True+custom_threshold-False+synchrony-single.py | 2 +-
 ...egorical-True+custom_threshold-True+synchrony-batch.py | 2 +-
 ...gorical-True+custom_threshold-True+synchrony-single.py | 2 +-
 ...orical-False+custom_threshold-False+synchrony-batch.py | 2 +-
 ...rical-False+custom_threshold-False+synchrony-single.py | 2 +-
 ...gorical-False+custom_threshold-True+synchrony-batch.py | 2 +-
 ...orical-False+custom_threshold-True+synchrony-single.py | 2 +-
 ...gorical-True+custom_threshold-False+synchrony-batch.py | 2 +-
 ...orical-True+custom_threshold-False+synchrony-single.py | 2 +-
 ...egorical-True+custom_threshold-True+synchrony-batch.py | 2 +-
 ...gorical-True+custom_threshold-True+synchrony-single.py | 2 +-
 ...orical-False+custom_threshold-False+synchrony-batch.py | 2 +-
 ...rical-False+custom_threshold-False+synchrony-single.py | 2 +-
 ...gorical-False+custom_threshold-True+synchrony-batch.py | 2 +-
 ...orical-False+custom_threshold-True+synchrony-single.py | 2 +-
 ...gorical-True+custom_threshold-False+synchrony-batch.py | 2 +-
 ...orical-True+custom_threshold-False+synchrony-single.py | 2 +-
 ...egorical-True+custom_threshold-True+synchrony-batch.py | 2 +-
 ...gorical-True+custom_threshold-True+synchrony-single.py | 2 +-
 ...orical-False+custom_threshold-False+synchrony-batch.py | 2 +-
 ...rical-False+custom_threshold-False+synchrony-single.py | 2 +-
 ...gorical-False+custom_threshold-True+synchrony-batch.py | 2 +-
 ...orical-False+custom_threshold-True+synchrony-single.py | 2 +-
 ...gorical-True+custom_threshold-False+synchrony-batch.py | 2 +-
 ...orical-True+custom_threshold-False+synchrony-single.py | 2 +-
 ...egorical-True+custom_threshold-True+synchrony-batch.py | 2 +-
 ...gorical-True+custom_threshold-True+synchrony-single.py | 2 +-
 ...orical-False+custom_threshold-False+synchrony-batch.py | 2 +-
 ...rical-False+custom_threshold-False+synchrony-single.py | 2 +-
 ...gorical-False+custom_threshold-True+synchrony-batch.py | 2 +-
 ...orical-False+custom_threshold-True+synchrony-single.py | 2 +-
 ...gorical-True+custom_threshold-False+synchrony-batch.py | 2 +-
 ...orical-True+custom_threshold-False+synchrony-single.py | 2 +-
 ...egorical-True+custom_threshold-True+synchrony-batch.py | 2 +-
 ...gorical-True+custom_threshold-True+synchrony-single.py | 2 +-
 ...orical-False+custom_threshold-False+synchrony-batch.py | 2 +-
 ...rical-False+custom_threshold-False+synchrony-single.py | 2 +-
 ...gorical-False+custom_threshold-True+synchrony-batch.py | 2 +-
 ...orical-False+custom_threshold-True+synchrony-single.py | 2 +-
 ...gorical-True+custom_threshold-False+synchrony-batch.py | 2 +-
 ...orical-True+custom_threshold-False+synchrony-single.py | 2 +-
 ...egorical-True+custom_threshold-True+synchrony-batch.py | 2 +-
 ...gorical-True+custom_threshold-True+synchrony-single.py | 2 +-
 ...orical-False+custom_threshold-False+synchrony-batch.py | 2 +-
 ...rical-False+custom_threshold-False+synchrony-single.py | 2 +-
 ...gorical-False+custom_threshold-True+synchrony-batch.py | 2 +-
 ...orical-False+custom_threshold-True+synchrony-single.py | 2 +-
 ...gorical-True+custom_threshold-False+synchrony-batch.py | 2 +-
 ...orical-True+custom_threshold-False+synchrony-single.py | 2 +-
 ...egorical-True+custom_threshold-True+synchrony-batch.py | 2 +-
 ...gorical-True+custom_threshold-True+synchrony-single.py | 2 +-
 ...orical-False+custom_threshold-False+synchrony-batch.py | 2 +-
 ...rical-False+custom_threshold-False+synchrony-single.py | 2 +-
 ...gorical-False+custom_threshold-True+synchrony-batch.py | 2 +-
 ...orical-False+custom_threshold-True+synchrony-single.py | 2 +-
 ...gorical-True+custom_threshold-False+synchrony-batch.py | 2 +-
 ...orical-True+custom_threshold-False+synchrony-single.py | 2 +-
 ...egorical-True+custom_threshold-True+synchrony-batch.py | 2 +-
 ...gorical-True+custom_threshold-True+synchrony-single.py | 2 +-
 ...orical-False+custom_threshold-False+synchrony-batch.py | 2 +-
 ...rical-False+custom_threshold-False+synchrony-single.py | 2 +-
 ...gorical-False+custom_threshold-True+synchrony-batch.py | 2 +-
 ...orical-False+custom_threshold-True+synchrony-single.py | 2 +-
 ...gorical-True+custom_threshold-False+synchrony-batch.py | 2 +-
 ...orical-True+custom_threshold-False+synchrony-single.py | 2 +-
 ...egorical-True+custom_threshold-True+synchrony-batch.py | 2 +-
 ...gorical-True+custom_threshold-True+synchrony-single.py | 2 +-
 ...orical-False+custom_threshold-False+synchrony-batch.py | 2 +-
 ...rical-False+custom_threshold-False+synchrony-single.py | 2 +-
 ...gorical-False+custom_threshold-True+synchrony-batch.py | 2 +-
 ...orical-False+custom_threshold-True+synchrony-single.py | 2 +-
 ...gorical-True+custom_threshold-False+synchrony-batch.py | 2 +-
 ...orical-True+custom_threshold-False+synchrony-single.py | 2 +-
 ...egorical-True+custom_threshold-True+synchrony-batch.py | 2 +-
 ...gorical-True+custom_threshold-True+synchrony-single.py | 2 +-
 ...orical-False+custom_threshold-False+synchrony-batch.py | 2 +-
 ...rical-False+custom_threshold-False+synchrony-single.py | 2 +-
 ...gorical-False+custom_threshold-True+synchrony-batch.py | 2 +-
 ...orical-False+custom_threshold-True+synchrony-single.py | 2 +-
 ...gorical-True+custom_threshold-False+synchrony-batch.py | 2 +-
 ...orical-True+custom_threshold-False+synchrony-single.py | 2 +-
 ...egorical-True+custom_threshold-True+synchrony-batch.py | 2 +-
 ...gorical-True+custom_threshold-True+synchrony-single.py | 2 +-
 ...orical-False+custom_threshold-False+synchrony-batch.py | 2 +-
 ...rical-False+custom_threshold-False+synchrony-single.py | 2 +-
 ...gorical-False+custom_threshold-True+synchrony-batch.py | 2 +-
 ...orical-False+custom_threshold-True+synchrony-single.py | 2 +-
 ...gorical-True+custom_threshold-False+synchrony-batch.py | 2 +-
 ...orical-True+custom_threshold-False+synchrony-single.py | 2 +-
 ...egorical-True+custom_threshold-True+synchrony-batch.py | 2 +-
 ...gorical-True+custom_threshold-True+synchrony-single.py | 2 +-
 ...orical-False+custom_threshold-False+synchrony-batch.py | 2 +-
 ...rical-False+custom_threshold-False+synchrony-single.py | 2 +-
 ...gorical-False+custom_threshold-True+synchrony-batch.py | 2 +-
 ...orical-False+custom_threshold-True+synchrony-single.py | 2 +-
 ...gorical-True+custom_threshold-False+synchrony-batch.py | 2 +-
 ...orical-True+custom_threshold-False+synchrony-single.py | 2 +-
 ...egorical-True+custom_threshold-True+synchrony-batch.py | 2 +-
 ...gorical-True+custom_threshold-True+synchrony-single.py | 2 +-
 ...orical-False+custom_threshold-False+synchrony-batch.py | 2 +-
 ...rical-False+custom_threshold-False+synchrony-single.py | 2 +-
 ...gorical-False+custom_threshold-True+synchrony-batch.py | 2 +-
 ...orical-False+custom_threshold-True+synchrony-single.py | 2 +-
 ...gorical-True+custom_threshold-False+synchrony-batch.py | 2 +-
 ...orical-True+custom_threshold-False+synchrony-single.py | 2 +-
 ...egorical-True+custom_threshold-True+synchrony-batch.py | 2 +-
 ...gorical-True+custom_threshold-True+synchrony-single.py | 2 +-
 ...orical-False+custom_threshold-False+synchrony-batch.py | 2 +-
 ...rical-False+custom_threshold-False+synchrony-single.py | 2 +-
 ...gorical-True+custom_threshold-False+synchrony-batch.py | 2 +-
 ...orical-True+custom_threshold-False+synchrony-single.py | 2 +-
 ...orical-False+custom_threshold-False+synchrony-batch.py | 2 +-
 ...rical-False+custom_threshold-False+synchrony-single.py | 2 +-
 ...gorical-True+custom_threshold-False+synchrony-batch.py | 2 +-
 ...orical-True+custom_threshold-False+synchrony-single.py | 2 +-
 ...orical-False+custom_threshold-False+synchrony-batch.py | 2 +-
 ...rical-False+custom_threshold-False+synchrony-single.py | 2 +-
 ...gorical-True+custom_threshold-False+synchrony-batch.py | 2 +-
 ...orical-True+custom_threshold-False+synchrony-single.py | 2 +-
 ...orical-False+custom_threshold-False+synchrony-batch.py | 2 +-
 ...rical-False+custom_threshold-False+synchrony-single.py | 2 +-
 ...gorical-True+custom_threshold-False+synchrony-batch.py | 2 +-
 ...orical-True+custom_threshold-False+synchrony-single.py | 2 +-
 ...orical-False+custom_threshold-False+synchrony-batch.py | 2 +-
 ...rical-False+custom_threshold-False+synchrony-single.py | 2 +-
 ...gorical-True+custom_threshold-False+synchrony-batch.py | 2 +-
 ...orical-True+custom_threshold-False+synchrony-single.py | 2 +-
 ...orical-False+custom_threshold-False+synchrony-batch.py | 2 +-
 ...rical-False+custom_threshold-False+synchrony-single.py | 2 +-
 ...gorical-True+custom_threshold-False+synchrony-batch.py | 2 +-
 ...orical-True+custom_threshold-False+synchrony-single.py | 2 +-
 ...orical-False+custom_threshold-False+synchrony-batch.py | 2 +-
 ...rical-False+custom_threshold-False+synchrony-single.py | 2 +-
 ...gorical-True+custom_threshold-False+synchrony-batch.py | 2 +-
 ...orical-True+custom_threshold-False+synchrony-single.py | 2 +-
 ...orical-False+custom_threshold-False+synchrony-batch.py | 2 +-
 ...rical-False+custom_threshold-False+synchrony-single.py | 2 +-
 ...gorical-True+custom_threshold-False+synchrony-batch.py | 2 +-
 ...orical-True+custom_threshold-False+synchrony-single.py | 2 +-
 ...orical-False+custom_threshold-False+synchrony-batch.py | 2 +-
 ...rical-False+custom_threshold-False+synchrony-single.py | 2 +-
 ...gorical-True+custom_threshold-False+synchrony-batch.py | 2 +-
 ...orical-True+custom_threshold-False+synchrony-single.py | 2 +-
 ...orical-False+custom_threshold-False+synchrony-batch.py | 2 +-
 ...rical-False+custom_threshold-False+synchrony-single.py | 2 +-
 ...gorical-True+custom_threshold-False+synchrony-batch.py | 2 +-
 ...orical-True+custom_threshold-False+synchrony-single.py | 2 +-
 ...orical-False+custom_threshold-False+synchrony-batch.py | 2 +-
 ...rical-False+custom_threshold-False+synchrony-single.py | 2 +-
 ...gorical-True+custom_threshold-False+synchrony-batch.py | 2 +-
 ...orical-True+custom_threshold-False+synchrony-single.py | 2 +-
 ...orical-False+custom_threshold-False+synchrony-batch.py | 2 +-
 ...rical-False+custom_threshold-False+synchrony-single.py | 2 +-
 ...gorical-True+custom_threshold-False+synchrony-batch.py | 2 +-
 ...orical-True+custom_threshold-False+synchrony-single.py | 2 +-
 ...orical-False+custom_threshold-False+synchrony-batch.py | 2 +-
 ...rical-False+custom_threshold-False+synchrony-single.py | 2 +-
 ...gorical-True+custom_threshold-False+synchrony-batch.py | 2 +-
 ...orical-True+custom_threshold-False+synchrony-single.py | 2 +-
 ...orical-False+custom_threshold-False+synchrony-batch.py | 2 +-
 ...rical-False+custom_threshold-False+synchrony-single.py | 2 +-
 ...gorical-True+custom_threshold-False+synchrony-batch.py | 2 +-
 ...orical-True+custom_threshold-False+synchrony-single.py | 2 +-
 ...orical-False+custom_threshold-False+synchrony-batch.py | 2 +-
 ...rical-False+custom_threshold-False+synchrony-single.py | 2 +-
 ...gorical-True+custom_threshold-False+synchrony-batch.py | 2 +-
 ...orical-True+custom_threshold-False+synchrony-single.py | 2 +-
 ...orical-False+custom_threshold-False+synchrony-batch.py | 2 +-
 ...rical-False+custom_threshold-False+synchrony-single.py | 2 +-
 ...gorical-True+custom_threshold-False+synchrony-batch.py | 2 +-
 ...orical-True+custom_threshold-False+synchrony-single.py | 2 +-
 ...orical-False+custom_threshold-False+synchrony-batch.py | 2 +-
 ...rical-False+custom_threshold-False+synchrony-single.py | 2 +-
 ...gorical-True+custom_threshold-False+synchrony-batch.py | 2 +-
 ...orical-True+custom_threshold-False+synchrony-single.py | 2 +-
 ...orical-False+custom_threshold-False+synchrony-batch.py | 2 +-
 ...rical-False+custom_threshold-False+synchrony-single.py | 2 +-
 ...gorical-True+custom_threshold-False+synchrony-batch.py | 2 +-
 ...orical-True+custom_threshold-False+synchrony-single.py | 2 +-
 ...orical-False+custom_threshold-False+synchrony-batch.py | 2 +-
 ...rical-False+custom_threshold-False+synchrony-single.py | 2 +-
 ...gorical-True+custom_threshold-False+synchrony-batch.py | 2 +-
 ...orical-True+custom_threshold-False+synchrony-single.py | 2 +-
 ...orical-False+custom_threshold-False+synchrony-batch.py | 2 +-
 ...rical-False+custom_threshold-False+synchrony-single.py | 2 +-
 ...gorical-True+custom_threshold-False+synchrony-batch.py | 2 +-
 ...orical-True+custom_threshold-False+synchrony-single.py | 2 +-
 ...orical-False+custom_threshold-False+synchrony-batch.py | 2 +-
 ...rical-False+custom_threshold-False+synchrony-single.py | 2 +-
 ...gorical-True+custom_threshold-False+synchrony-batch.py | 2 +-
 ...orical-True+custom_threshold-False+synchrony-single.py | 2 +-
 ...orical-False+custom_threshold-False+synchrony-batch.py | 2 +-
 ...rical-False+custom_threshold-False+synchrony-single.py | 2 +-
 ...gorical-True+custom_threshold-False+synchrony-batch.py | 2 +-
 ...orical-True+custom_threshold-False+synchrony-single.py | 2 +-
 ...orical-False+custom_threshold-False+synchrony-batch.py | 2 +-
 ...rical-False+custom_threshold-False+synchrony-single.py | 2 +-
 ...gorical-True+custom_threshold-False+synchrony-batch.py | 2 +-
 ...orical-True+custom_threshold-False+synchrony-single.py | 2 +-
 ...orical-False+custom_threshold-False+synchrony-batch.py | 2 +-
 ...rical-False+custom_threshold-False+synchrony-single.py | 2 +-
 ...gorical-True+custom_threshold-False+synchrony-batch.py | 2 +-
 ...orical-True+custom_threshold-False+synchrony-single.py | 2 +-
 ...orical-False+custom_threshold-False+synchrony-batch.py | 2 +-
 ...rical-False+custom_threshold-False+synchrony-single.py | 2 +-
 ...gorical-True+custom_threshold-False+synchrony-batch.py | 2 +-
 ...orical-True+custom_threshold-False+synchrony-single.py | 2 +-
 ...orical-False+custom_threshold-False+synchrony-batch.py | 2 +-
 ...rical-False+custom_threshold-False+synchrony-single.py | 2 +-
 ...gorical-True+custom_threshold-False+synchrony-batch.py | 2 +-
 ...orical-True+custom_threshold-False+synchrony-single.py | 2 +-
 ...orical-False+custom_threshold-False+synchrony-batch.py | 2 +-
 ...rical-False+custom_threshold-False+synchrony-single.py | 2 +-
 ...gorical-True+custom_threshold-False+synchrony-batch.py | 2 +-
 ...orical-True+custom_threshold-False+synchrony-single.py | 2 +-
 ...orical-False+custom_threshold-False+synchrony-batch.py | 2 +-
 ...rical-False+custom_threshold-False+synchrony-single.py | 2 +-
 ...gorical-True+custom_threshold-False+synchrony-batch.py | 2 +-
 ...orical-True+custom_threshold-False+synchrony-single.py | 2 +-
 ...orical-False+custom_threshold-False+synchrony-batch.py | 2 +-
 ...rical-False+custom_threshold-False+synchrony-single.py | 2 +-
 ...gorical-True+custom_threshold-False+synchrony-batch.py | 2 +-
 ...orical-True+custom_threshold-False+synchrony-single.py | 2 +-
 ...orical-False+custom_threshold-False+synchrony-batch.py | 2 +-
 ...rical-False+custom_threshold-False+synchrony-single.py | 2 +-
 ...gorical-True+custom_threshold-False+synchrony-batch.py | 2 +-
 ...orical-True+custom_threshold-False+synchrony-single.py | 2 +-
 ...orical-False+custom_threshold-False+synchrony-batch.py | 2 +-
 ...rical-False+custom_threshold-False+synchrony-single.py | 2 +-
 ...gorical-True+custom_threshold-False+synchrony-batch.py | 2 +-
 ...orical-True+custom_threshold-False+synchrony-single.py | 2 +-
 ...orical-False+custom_threshold-False+synchrony-batch.py | 2 +-
 ...rical-False+custom_threshold-False+synchrony-single.py | 2 +-
 ...gorical-True+custom_threshold-False+synchrony-batch.py | 2 +-
 ...orical-True+custom_threshold-False+synchrony-single.py | 2 +-
 docs/honegumi.html                                        | 8 ++++----
 scripts/generate_scripts.py                               | 4 ++--
 1154 files changed, 3078 insertions(+), 3078 deletions(-)

diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index a9c76a6d..5b6c46de 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "e1ef5215",
+   "id": "3c72aff2",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "c788e0b6",
+   "id": "6631e8e4",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "748e6dd2",
+   "id": "4514e203",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -67,7 +67,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
index 6966d41f..1da44771 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "8a726861",
+   "id": "91a5cb5c",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "9e8904fa",
+   "id": "df46df73",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "c73576d7",
+   "id": "9cfe45c8",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -67,7 +67,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
index aa0983e8..ad7579f8 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "01163a12",
+   "id": "26753923",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "32409e12",
+   "id": "6897e593",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "0109f7ee",
+   "id": "adad3bb9",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -67,7 +67,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
index fada6660..334749a5 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "f78a8144",
+   "id": "688e3d20",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "a1af47b2",
+   "id": "10bd679d",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "55b41f3f",
+   "id": "9bc48b22",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -67,7 +67,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 939c8c0d..53d93065 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "f4679ed8",
+   "id": "4c194001",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "ee099c5a",
+   "id": "c9dfc950",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "f5206987",
+   "id": "8dccee59",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -75,7 +75,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 2f1cb726..9a67f7b9 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "fef05336",
+   "id": "d29b3ff6",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "39ddad51",
+   "id": "bcad2de8",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "05964b52",
+   "id": "5ffd7775",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -75,7 +75,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
index cdf529e1..f43ec0e4 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "e87a6015",
+   "id": "ea53ec54",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "dbb128be",
+   "id": "bb0026e6",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "286284a1",
+   "id": "199a22e9",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -75,7 +75,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
index 47ae15a2..aac7751d 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "9fbef3a0",
+   "id": "28b4b609",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "61ad52b6",
+   "id": "d03988eb",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "0a2a6ee3",
+   "id": "2bad020e",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -75,7 +75,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index 484a5f4e..531a8ee0 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "de9aa73f",
+   "id": "420602e1",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "988c2396",
+   "id": "7d43ee5a",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "bde45f7c",
+   "id": "eb13b1f7",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -76,7 +76,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
index 857db128..15616632 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "6b89aff8",
+   "id": "7cdc6995",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "99c9525c",
+   "id": "1c5b2968",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "0bd6f4af",
+   "id": "e30fd0c3",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -76,7 +76,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
index f592b4bf..064e5141 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "5edfbcc0",
+   "id": "ff9c2b6e",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "5e830d8d",
+   "id": "40607b37",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "6c75e623",
+   "id": "3f3ac62a",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -76,7 +76,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
index f4f35ad4..d8befe20 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "5bf876ac",
+   "id": "d57bab03",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "52ef0547",
+   "id": "5fb918b9",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "1c5fd795",
+   "id": "d83f889e",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -76,7 +76,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index ae2a6d02..e4a30665 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "925bf767",
+   "id": "3ae3fbe9",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "3a475baf",
+   "id": "751014c8",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "b38e0e76",
+   "id": "ded75669",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -84,7 +84,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 2564ea8b..db07159a 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "832fa1c2",
+   "id": "e59eca32",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "8d6267c3",
+   "id": "4856b9ad",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "a041efdf",
+   "id": "5195bb27",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -84,7 +84,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
index fe535c71..63f75c56 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "11686cec",
+   "id": "104e4b2a",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "3164f51c",
+   "id": "4dc998d4",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "6f243124",
+   "id": "fdfdb3b9",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -84,7 +84,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
index c12bda6f..0ae5b625 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "6c3320cc",
+   "id": "fbb86e3e",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "339a4268",
+   "id": "033ac7b7",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "2a46a826",
+   "id": "a4a09332",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -84,7 +84,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index f1dfc7fd..67f70600 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "bace0043",
+   "id": "d94cac35",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "b5738132",
+   "id": "4572bd24",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "b4afb559",
+   "id": "418f207d",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -67,7 +67,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
index d5970564..e007190f 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "98a35974",
+   "id": "9552f5a8",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "05b56a4d",
+   "id": "818db26c",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "69460652",
+   "id": "09c2a39b",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -67,7 +67,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
index 4b6fe55c..a33fa58e 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "0309771f",
+   "id": "f879acbb",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "76b9076b",
+   "id": "821db2e6",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "fd258642",
+   "id": "43f7d98c",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -67,7 +67,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
index 2bc4a140..00d5782b 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "7cd15507",
+   "id": "13226cbb",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "a470c694",
+   "id": "1b342fd6",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "9c913f72",
+   "id": "0e54d4c8",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -67,7 +67,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 770a6d03..2846805d 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "1f34eefc",
+   "id": "0af1b160",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "f509459f",
+   "id": "df53e43f",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "ced9b6f7",
+   "id": "c9e5f774",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -75,7 +75,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
index ac66448b..500fd98c 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "063c5634",
+   "id": "119ad12f",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "97f9fb20",
+   "id": "a5a87550",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "ee9005c4",
+   "id": "8ced822b",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -75,7 +75,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
index 43c94ca4..eef17ee5 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "8d12913c",
+   "id": "bdef28f8",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "317bdc59",
+   "id": "935d6981",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "972a3aa1",
+   "id": "886efbaf",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -75,7 +75,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
index f8b4c1b3..b964c664 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "e01eaf59",
+   "id": "79441368",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "9033a02f",
+   "id": "7a09e948",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "5c47ba29",
+   "id": "de8d9f00",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -75,7 +75,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index ade86885..ddd7d9d8 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "b3eb1701",
+   "id": "fcbb9b34",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "b2f37796",
+   "id": "2d7ab89b",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "a441b16a",
+   "id": "0aa1684a",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -76,7 +76,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
index 40f9da51..db7867d6 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "6094161b",
+   "id": "87862220",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "7cc88fe2",
+   "id": "ffec31d7",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "c72d0716",
+   "id": "8794c5a3",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -76,7 +76,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
index b9a7d1c0..d7eb12a4 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "a4305038",
+   "id": "d419d107",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "006b8f18",
+   "id": "5561097e",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "8381dd76",
+   "id": "6db0d9a7",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -76,7 +76,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
index 62cf4736..f0fec98e 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "fdcc3936",
+   "id": "73aec1d7",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "5a9b7831",
+   "id": "f61ead08",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "778140be",
+   "id": "5653c311",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -76,7 +76,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 2885b761..61632eac 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "463d3e0a",
+   "id": "7008a25c",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "64abfb52",
+   "id": "fe69ec87",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "43813a0a",
+   "id": "db13ecca",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -84,7 +84,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 7829a962..9d4f1251 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "4f87956c",
+   "id": "68c4ec9c",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "fa2e8add",
+   "id": "c578c2b1",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "46ff2d9c",
+   "id": "c8933bf1",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -84,7 +84,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
index e96d16c8..37e6b073 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "588581cf",
+   "id": "f4660294",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "731e9063",
+   "id": "020544b8",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "f1ba2abb",
+   "id": "3da0923a",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -84,7 +84,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
index 545f5bdf..edfe0b23 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "eadc7424",
+   "id": "766ffcac",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "4f1b1145",
+   "id": "e49de922",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "be43f1aa",
+   "id": "206b3972",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -84,7 +84,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index 9434cbd6..30d2b716 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "0bc1f36a",
+   "id": "dd89deea",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "b9f11ef1",
+   "id": "2dcfecd0",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "3f4849e8",
+   "id": "e078093d",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -67,7 +67,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
index 411e393b..b274e99f 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "1a550dfa",
+   "id": "5e76aa36",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "8a8b3fa9",
+   "id": "d4b742ee",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "85fe4721",
+   "id": "b00a1625",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -67,7 +67,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
index faa5fef3..fe8def89 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "f771c587",
+   "id": "5ba92640",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "c5cbcd3f",
+   "id": "7997362a",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "2b567ce3",
+   "id": "effa8349",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -67,7 +67,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
index a89ef446..a063fe77 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "4d76b9d6",
+   "id": "5f2757a3",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "8a07fa68",
+   "id": "aacb2469",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "26fa93f3",
+   "id": "3d54a089",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -67,7 +67,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 8c0c0e03..98516173 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "1805ce17",
+   "id": "9a2995f5",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "0702bca4",
+   "id": "daff01b7",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "8990cfa7",
+   "id": "439e8345",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -75,7 +75,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 5e5ca75a..92b4a426 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "38f1e534",
+   "id": "e7ba1a24",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "2b092d59",
+   "id": "33d32699",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "cbec62de",
+   "id": "4bff8516",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -75,7 +75,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
index a613b0c2..9eb31b30 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "f2987fb0",
+   "id": "f567dceb",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "8ba6dd37",
+   "id": "83226df0",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "a0ad653c",
+   "id": "f21e1577",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -75,7 +75,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
index 1e34ccc8..e3a7706b 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "6d753e0d",
+   "id": "eadd61c9",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "7e3163a8",
+   "id": "b74e2b3d",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "4512c94d",
+   "id": "9a85382e",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -75,7 +75,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index 2c007f56..21e80df3 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "fc766589",
+   "id": "877e3bb8",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "e605044f",
+   "id": "c56c7a56",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "e2ae0aa1",
+   "id": "27034f7d",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -76,7 +76,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
index 29454cb0..2cea3b4b 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "fb638df1",
+   "id": "08af8078",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "b41384fc",
+   "id": "2b15f0fc",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "99237d66",
+   "id": "ed7b80dd",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -76,7 +76,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
index c2a91bf9..48977091 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "5df05036",
+   "id": "c8bc2483",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "acec7fa3",
+   "id": "45cb4883",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "4b92c9b5",
+   "id": "2918c13f",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -76,7 +76,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
index 89d2543b..37af84be 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "8efc3ea6",
+   "id": "961f4139",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "d5d16b68",
+   "id": "30318a0b",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "a2155563",
+   "id": "bf5a9818",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -76,7 +76,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index cf5af693..2491f82f 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "29228377",
+   "id": "dcd3f477",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "4a9f500e",
+   "id": "e27bc48a",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "816df962",
+   "id": "578fd066",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -84,7 +84,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
index b7ced610..3402c464 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "a29da79b",
+   "id": "c7d3323d",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "ed8aa8dd",
+   "id": "4e234466",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "3800bcd0",
+   "id": "8f835985",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -84,7 +84,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
index ef67759b..d2ff5c7e 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "3b7e8596",
+   "id": "b536ae5a",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "ca1f12b3",
+   "id": "dab1c268",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "e964030f",
+   "id": "e354b853",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -84,7 +84,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
index 416dca94..b6de1717 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "cad5614d",
+   "id": "e022924a",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "19f04ac9",
+   "id": "f4056d43",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "cc352f67",
+   "id": "4385e7c0",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -84,7 +84,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index 06f3ee5a..deda4bd0 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "308d8782",
+   "id": "ba9edf5e",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "d38ef8e9",
+   "id": "e75b3ff4",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "51653e27",
+   "id": "89c79a70",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -67,7 +67,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
index 257c0c0a..981da239 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "6d49d74a",
+   "id": "e9c73043",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "69475389",
+   "id": "cb7bc0f4",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "85bbf3f6",
+   "id": "1992dd0b",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -67,7 +67,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
index e96a39df..1433bdf7 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "ae8d8eca",
+   "id": "8f76b98a",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "4757ce68",
+   "id": "4a0b65f1",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "77621883",
+   "id": "99be348e",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -67,7 +67,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
index 5f789b27..d30e9638 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "2000a5d6",
+   "id": "6a726770",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "737d4538",
+   "id": "da8fb3b7",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "31cc1d55",
+   "id": "66ac82a2",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -67,7 +67,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 86817439..844d7c75 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "aa0173d0",
+   "id": "3ac546fd",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "43aa540a",
+   "id": "c90c7c72",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "6559507c",
+   "id": "1abf6003",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -75,7 +75,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
index e0b7d440..399e7bb2 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "debee3a7",
+   "id": "ee5b7f81",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "23285382",
+   "id": "9671e3f1",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "8c92aec2",
+   "id": "476e7c68",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -75,7 +75,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
index 0415a20a..7f40932f 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "3f10734d",
+   "id": "82387b97",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "90939167",
+   "id": "89d38c47",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "9e0c7f4e",
+   "id": "4abd37ff",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -75,7 +75,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
index 76c1ed6a..97266212 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "e8ea7725",
+   "id": "415a3148",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "943de2a3",
+   "id": "386cd3d9",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "8c7d0e79",
+   "id": "cddb5572",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -75,7 +75,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index 4a2207c0..62c0bcc9 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "4ffb6971",
+   "id": "24e194dc",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "4d92de7c",
+   "id": "2cfd9f89",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "313cd601",
+   "id": "0aaba0d5",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -76,7 +76,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
index c0b3f670..bf2101f0 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "cda7d22a",
+   "id": "aa0d7fc9",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "af1d0822",
+   "id": "870ac89f",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "322a4b99",
+   "id": "7c870f0e",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -76,7 +76,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
index 860ceeed..c2ec9867 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "7782e018",
+   "id": "54765b86",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "c6ea8674",
+   "id": "402bcaec",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "21142747",
+   "id": "801aef52",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -76,7 +76,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
index 79fc67f7..3c04d416 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "3e4de994",
+   "id": "58738197",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "9abe72cd",
+   "id": "979a5a7b",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "fb3d8da1",
+   "id": "54d62090",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -76,7 +76,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 49016e04..cfe93d49 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "5aa71ed9",
+   "id": "b6717029",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "4bdd5219",
+   "id": "bea2f48c",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "33bf4cb3",
+   "id": "0a063aaf",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -84,7 +84,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 55441d40..4aef487a 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "0dc56395",
+   "id": "991ca15d",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "5e8afeb5",
+   "id": "95b46d60",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "91639db2",
+   "id": "1b39a104",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -84,7 +84,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
index ef72f678..f838fde4 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "73e97fd0",
+   "id": "b7775a42",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "72255021",
+   "id": "453c39df",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "84988a41",
+   "id": "2f71e974",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -84,7 +84,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
index 67f21069..39a79fef 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "bddf3aac",
+   "id": "b1568b7b",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "aeee228d",
+   "id": "03e70dd2",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "f5762fd5",
+   "id": "e227222f",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -84,7 +84,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index 4d80f1c7..a922a083 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "e9b3d4e6",
+   "id": "715b3cb2",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "c42d587d",
+   "id": "d499f14d",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "a0b904e9",
+   "id": "6b1f3e7f",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -67,7 +67,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
index 3ed44a75..34735d43 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "1f288820",
+   "id": "ec8308e2",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "c8cf5066",
+   "id": "7b364422",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "f4134c3f",
+   "id": "d4369c9a",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -67,7 +67,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
index 3d439fa4..ad8f8e4a 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "05d0a022",
+   "id": "93e0a20e",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "422b0b58",
+   "id": "95bbac69",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "60b5d0e5",
+   "id": "b81fd4af",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -67,7 +67,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
index 5fb9c5d7..ea346126 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "22896244",
+   "id": "67d8f5b7",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "827c3522",
+   "id": "e4a85faf",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "0794821b",
+   "id": "25484fc7",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -67,7 +67,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 29c6c5da..f3bb688f 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "31ca5ad4",
+   "id": "f2c1d2ff",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "a7ca4d55",
+   "id": "8dc862a8",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "b1937bba",
+   "id": "72f63dbf",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -75,7 +75,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
index f3874cd1..e2a4a6c4 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "1b2daa4a",
+   "id": "b9641930",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "841a0892",
+   "id": "8e654185",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "8e0e7eda",
+   "id": "086cc35c",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -75,7 +75,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
index 4807a703..a8fa4268 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "db663080",
+   "id": "8ad6526b",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "d472a331",
+   "id": "15b26fa2",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "50581785",
+   "id": "40c504f9",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -75,7 +75,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
index 475c91cf..8b8a27c8 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "2f494fd3",
+   "id": "c4516172",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "78c0c3b7",
+   "id": "560eafc8",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "ac6ad51a",
+   "id": "10d34b84",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -75,7 +75,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index f53e225d..72de8521 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "729d40fd",
+   "id": "9ceaae93",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "b72f7989",
+   "id": "6483c1d6",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "6142bfbb",
+   "id": "9692457e",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -76,7 +76,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
index 5da346fc..31a17778 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "06e9247c",
+   "id": "d45f3cff",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "67845c73",
+   "id": "480994b7",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "3cc85a19",
+   "id": "c4668e93",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -76,7 +76,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
index a8ffed98..1f3168f6 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "dec6026d",
+   "id": "1eb2f7da",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "5b2fd0f7",
+   "id": "bae78e13",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "ce91a666",
+   "id": "dc26a640",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -76,7 +76,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
index dd098f80..6f6ebb0d 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "598994d6",
+   "id": "54feaec0",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "41ce5c17",
+   "id": "2afb42b7",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "f60ab7b9",
+   "id": "f6031713",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -76,7 +76,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index c544aea6..be6c4742 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "af142300",
+   "id": "90385307",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "75c0acfb",
+   "id": "9fb9033b",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "f00ef955",
+   "id": "46113ec6",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -84,7 +84,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 7239b7fb..cc39d462 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "9bd6691c",
+   "id": "bdb65d26",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "af6d5b51",
+   "id": "454f44f0",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "844a8b74",
+   "id": "5cdfbb9d",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -84,7 +84,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
index e6f8f421..ae7e1573 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "fd5e3895",
+   "id": "8c046c46",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "0af348b5",
+   "id": "61686e75",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "68bf68b8",
+   "id": "cbfbecf4",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -84,7 +84,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
index 115461dc..15c3d1be 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "77482089",
+   "id": "2d3706e4",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "85c103dd",
+   "id": "df8b2d6f",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "fc693430",
+   "id": "fddaef32",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -84,7 +84,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index 3d8a8f27..8af37170 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "bdc1d380",
+   "id": "ac69b461",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "610c693f",
+   "id": "d9c25c8c",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "c2edf01b",
+   "id": "65c4e17a",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -67,7 +67,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
index 5af9144c..17cc7dee 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "540094d5",
+   "id": "cfab83c5",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "3eed7cb6",
+   "id": "4b677afe",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "5d6f4c0d",
+   "id": "33c18ca7",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -67,7 +67,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
index 96200920..8f83fdc2 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "f27b5bf6",
+   "id": "38242bf5",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "d9d2005a",
+   "id": "1e134e53",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "1b9bee0d",
+   "id": "33b1ae03",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -67,7 +67,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
index 941c1a99..b9cd4472 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "f88523a3",
+   "id": "b8ebd7d4",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "8b580ece",
+   "id": "48e4b4db",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "1658e4d7",
+   "id": "a5ac52eb",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -67,7 +67,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 2fe2790e..7e6203d5 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "de3ee0e6",
+   "id": "8a931100",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "a04e8213",
+   "id": "4fc2b358",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "042b87fe",
+   "id": "7e1b4ec1",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -75,7 +75,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 747cf7f7..b7970437 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "783059c3",
+   "id": "26996dcc",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "05f9792b",
+   "id": "110e17b3",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "2637eb8b",
+   "id": "69b37ef5",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -75,7 +75,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
index 39f1de05..9a2d3606 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "5de5120b",
+   "id": "452ef899",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "7dc31759",
+   "id": "02ec23c4",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "8b9924aa",
+   "id": "6a979fb4",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -75,7 +75,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
index 06632088..52cc06eb 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "4161df7f",
+   "id": "c6080140",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "bccdbb6e",
+   "id": "40b60e22",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "3630b8da",
+   "id": "fc98e417",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -75,7 +75,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index edfde44a..13648ba6 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "3a988036",
+   "id": "49a15777",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "e1dafc54",
+   "id": "e6623b2c",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "becf0f14",
+   "id": "20e31e95",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -76,7 +76,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
index d0369592..ec0f40b5 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "8c482389",
+   "id": "70de055e",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "09166ce1",
+   "id": "61351022",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "e7e7f5b0",
+   "id": "edd09f22",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -76,7 +76,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
index e1030bbb..42229584 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "60e3235d",
+   "id": "3eb490ff",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "1ea3d1d8",
+   "id": "2aa88de0",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "d297e7db",
+   "id": "1fd2fe85",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -76,7 +76,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
index ce330d44..b2da127c 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "aaae2869",
+   "id": "ba3eb2f6",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "21d0d940",
+   "id": "e6e54b48",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "15d2a570",
+   "id": "1db482f7",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -76,7 +76,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 33a39900..a9a7b16d 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "41444715",
+   "id": "63b53a7c",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "e4c8d7db",
+   "id": "51e5438a",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "8e59d17d",
+   "id": "8224678c",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -84,7 +84,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 4c482bbb..deaaf678 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "ae1b2a13",
+   "id": "9a9ce710",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "9ea11c45",
+   "id": "b507a1a1",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "00562e60",
+   "id": "1abb6a24",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -84,7 +84,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
index ae977fd1..7baf5115 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "74b46ccf",
+   "id": "84788d6e",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "b77ecf8d",
+   "id": "beedaa07",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "458cb1d1",
+   "id": "6bafdd92",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -84,7 +84,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
index cf202e06..455a5ce1 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "147dca33",
+   "id": "a8e94eb0",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "66fbde11",
+   "id": "e121d3f3",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "6394f38a",
+   "id": "3e4b14de",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -84,7 +84,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index 6693132c..37663ad5 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "5942f570",
+   "id": "46f78f06",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "b4968481",
+   "id": "70939b2a",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "d9117af9",
+   "id": "b27c43c7",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -67,7 +67,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
index 587ffc01..5b1c08f4 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "028fcb28",
+   "id": "4d2f67c8",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "3d630929",
+   "id": "21ef1d91",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "02bd9708",
+   "id": "679f2809",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -67,7 +67,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
index 5fe1c55c..e1a9cd13 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "a323a55b",
+   "id": "be9b6be1",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "b2b14639",
+   "id": "7221e02c",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "53a330e4",
+   "id": "243084bf",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -67,7 +67,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
index 5a22f085..ea1d758d 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "c5f3f805",
+   "id": "0ed5d25e",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "575d35b5",
+   "id": "da6ff6eb",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "698c35e4",
+   "id": "56476404",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -67,7 +67,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index ada1697b..548fd835 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "5ee06a05",
+   "id": "ce112e72",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "aa9bb216",
+   "id": "c2ad0a46",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "5502cf5d",
+   "id": "1fe2ebde",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -75,7 +75,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 930d0fcd..c179f012 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "5c712240",
+   "id": "de943999",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "7253b8a4",
+   "id": "18e33465",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "a0115de2",
+   "id": "6eaee22a",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -75,7 +75,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
index e46da44f..d6a4c7ec 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "566fbfdd",
+   "id": "e451810c",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "3f74df8c",
+   "id": "01b77c72",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "d2ad9062",
+   "id": "c254e624",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -75,7 +75,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
index a90e8ad2..47078522 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "cdd2f157",
+   "id": "46b9ae50",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "701e31b1",
+   "id": "0e248b2e",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "e58534bc",
+   "id": "8210f025",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -75,7 +75,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index d086f518..224fbf1f 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "a28cb441",
+   "id": "7e00d4c3",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "9b007330",
+   "id": "2bd04ad3",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "785eba9b",
+   "id": "0e136ee0",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -76,7 +76,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
index 4cf86889..c30b80f7 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "4d475a01",
+   "id": "d1cc31cf",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "b5c552a1",
+   "id": "c7a7493f",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "1e17484c",
+   "id": "45450df2",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -76,7 +76,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
index 7d051b85..55bb9dd5 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "d8e13205",
+   "id": "1520433f",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "46407832",
+   "id": "f94cd200",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "b60202bb",
+   "id": "64a2293a",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -76,7 +76,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
index a1977d3f..384ba456 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "2b7296b3",
+   "id": "f4db6770",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "e20067a8",
+   "id": "08d19d2e",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "9292203b",
+   "id": "c9c56f62",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -76,7 +76,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 93d7a13b..21210962 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "650a30c4",
+   "id": "7b6bc560",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "628312e1",
+   "id": "a32cd349",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "374325b5",
+   "id": "37b356c1",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -84,7 +84,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 34f198f5..a87cb3a0 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "3c7679a5",
+   "id": "6224e479",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "1d7a0418",
+   "id": "e8c2ee21",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "fd494132",
+   "id": "0487f8d7",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -84,7 +84,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
index a89f6a77..c6cecf3f 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "fd98a974",
+   "id": "0b077a97",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "3003d1e8",
+   "id": "8e65a801",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "c02bb1c7",
+   "id": "2cd69b68",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -84,7 +84,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
index d12e39ce..74a38860 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "0c696f9d",
+   "id": "39774651",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "3bc870a6",
+   "id": "89e8dd7e",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "aa514d6a",
+   "id": "7e1dc774",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -84,7 +84,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index 81b5b2d8..77aa67b6 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "fe3e4de0",
+   "id": "2f57c642",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "ec9df69d",
+   "id": "1592f87e",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "10ef5685",
+   "id": "fbf64bd1",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -67,7 +67,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
index 0f8112a2..66865a2b 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "81efd795",
+   "id": "d054213d",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "53b3b5bc",
+   "id": "ddd79e5d",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "346240da",
+   "id": "b6511f24",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -67,7 +67,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
index f0acb3bd..65023887 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "e360002f",
+   "id": "e7020e68",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "045f221c",
+   "id": "4082ab21",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "a6a824ee",
+   "id": "e6bd6df0",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -67,7 +67,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
index ee1fcea1..58abac47 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "624366f6",
+   "id": "104705f2",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "e9854ec7",
+   "id": "f7a1530a",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "3a8042c3",
+   "id": "a856efdc",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -67,7 +67,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 016addfb..43d9240b 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "db38c5e5",
+   "id": "27b0da3c",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "c0d99a93",
+   "id": "0ce1109f",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "d33b97a9",
+   "id": "7758bbd3",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -75,7 +75,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
index cfadf09d..04ce8035 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "cc2ec09a",
+   "id": "2eb2457d",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "4e22ffc3",
+   "id": "ef3a4efc",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "a216ee84",
+   "id": "5380a873",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -75,7 +75,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
index 05935123..f5c96ac1 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "f0f176e8",
+   "id": "afa81bca",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "cdca1856",
+   "id": "757a1b59",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "bcddfc8f",
+   "id": "3e6ca5e1",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -75,7 +75,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
index 50c3abbd..2448e87c 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "ad75e3be",
+   "id": "e4fd0acc",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "e29c1e0c",
+   "id": "bf2349f2",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "59e0dc7b",
+   "id": "432d27d6",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -75,7 +75,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index f269cbb0..d434b665 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "8cb38d92",
+   "id": "70dbf46b",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "6ed29bbb",
+   "id": "fe7c7124",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "f4551045",
+   "id": "e562eddd",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -76,7 +76,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
index d78d9ddc..bf627190 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "db34e737",
+   "id": "ba2311c9",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "bdcb3bac",
+   "id": "d95e9af6",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "866b92a2",
+   "id": "340787ad",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -76,7 +76,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
index f61b7067..56572737 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "7cbc54d7",
+   "id": "0bcbe1b2",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "52c1f7e2",
+   "id": "d96774db",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "cb3a3539",
+   "id": "dae34bf9",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -76,7 +76,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
index 2165ef67..934f4d6e 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "b061a608",
+   "id": "34f64061",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "a0676364",
+   "id": "116af055",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "7010c877",
+   "id": "4d70b152",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -76,7 +76,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 765d8e6b..d65d01d0 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "7f3ced62",
+   "id": "a2e5816f",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "bbc5b5cd",
+   "id": "93fda3a4",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "cfe2bfbc",
+   "id": "49c4a289",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -84,7 +84,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
index dddb5ec8..b4e3dd33 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "dfb83c58",
+   "id": "4acc30da",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "276b8e6c",
+   "id": "818bd239",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "680be054",
+   "id": "432b8d48",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -84,7 +84,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
index 2a0198fa..37511ef8 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "b2a319e2",
+   "id": "b1038d08",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "dbcace6d",
+   "id": "9e9ec2bc",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "7f5e0ee1",
+   "id": "e198a03a",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -84,7 +84,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
index 5d837127..d7c4006e 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "381d57b9",
+   "id": "4bd4c72f",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "1b744fba",
+   "id": "9abc3142",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "3767c3d4",
+   "id": "756deb0f",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -84,7 +84,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index fde59bfa..84adbe15 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "18d37f42",
+   "id": "957dcabb",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "d1dd0ecc",
+   "id": "e4f21627",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "49e50e60",
+   "id": "1b7c56c5",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -88,7 +88,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
index 93a902a8..b54d3c7e 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "e16313ee",
+   "id": "ac0ea6b7",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "acb6ddc5",
+   "id": "9f269f2d",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "c5cec17b",
+   "id": "43e45c5b",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -88,7 +88,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
index 49d8242f..0ce4c9ce 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "b8c9ffce",
+   "id": "2fb0c94e",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "fb0248b4",
+   "id": "0423297c",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "ef2c14f3",
+   "id": "cab488a4",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -88,7 +88,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
index 9d08bce8..43dfb2f3 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "29b741d8",
+   "id": "3c6bbc0f",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "acbf8b1f",
+   "id": "ed20205a",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "924ec251",
+   "id": "b4eb3561",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -88,7 +88,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 9be1d665..6c447472 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "227d7b5e",
+   "id": "11bd1c10",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "c28e20a3",
+   "id": "28e5383f",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "bcda4a00",
+   "id": "a02b38c7",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -96,7 +96,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 4442398d..d4b9062b 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "aa292d4d",
+   "id": "d7f36fe9",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "7f4d223d",
+   "id": "139b43b1",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "cfd761f9",
+   "id": "eff092cb",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -96,7 +96,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
index 8f3c98d6..2abcf67d 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "4741bf23",
+   "id": "0ffe05a7",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "dc3e6de9",
+   "id": "80cc0e2b",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "aab3f4c6",
+   "id": "3047261d",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -96,7 +96,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
index 7e5baf02..288e4918 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "29519868",
+   "id": "31f91bb7",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "86f606a0",
+   "id": "642fd671",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "02fefd3c",
+   "id": "9644fd05",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -96,7 +96,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index 4f9316e4..76d37304 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "c5983eac",
+   "id": "18b001a9",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "ad622ca6",
+   "id": "7cdc8664",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "4f0e4f0d",
+   "id": "3e11c9ab",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -101,7 +101,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
index 22507dee..60b2d6bb 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "5cb5a717",
+   "id": "75afa164",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "824be3d2",
+   "id": "2043d304",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "e4b86ce1",
+   "id": "3b06305b",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -101,7 +101,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
index 0dda9039..ea3b8811 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "fafb0308",
+   "id": "c51045d8",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "81432033",
+   "id": "e59207d7",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "311d87dc",
+   "id": "da4b90cf",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -101,7 +101,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
index ca54e551..9aca2aed 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "aebbc36d",
+   "id": "cbe3b5a5",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "b16f7c84",
+   "id": "5e8e8f30",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "dfb5bfd6",
+   "id": "8f199b91",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -101,7 +101,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 731eeded..4d0e4912 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "44cbcb92",
+   "id": "f8e5ebe4",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "53e1a7cf",
+   "id": "11c0f580",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "2ed873c9",
+   "id": "6f28a804",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -110,7 +110,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 4bf6b3c8..c68719c0 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "b9cb1de5",
+   "id": "5966ed11",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "61e5db5a",
+   "id": "a3a729f3",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "df616db0",
+   "id": "64218131",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -110,7 +110,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
index 842c00d0..8c23fd2e 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "212c8442",
+   "id": "b842330d",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "13a6c5a8",
+   "id": "ace08468",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "0f2ea3e1",
+   "id": "9c249e82",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -110,7 +110,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
index 1ddfcd05..4d9d4507 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "4756df63",
+   "id": "7127302a",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "0d9b9dbf",
+   "id": "ef44ef56",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "497d55e1",
+   "id": "6389a38e",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -110,7 +110,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index 2ce5bb49..c8019b02 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "698e9314",
+   "id": "1ddf62ca",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "f6be75dd",
+   "id": "40bbec6a",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "82dfc976",
+   "id": "acbb3687",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -88,7 +88,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
index 486b45f7..8d069be7 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "986e366a",
+   "id": "1f6aa826",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "e866dfe1",
+   "id": "118c58a0",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "417f83b1",
+   "id": "ec774d13",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -88,7 +88,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
index 79dace9b..60993a38 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "e7ca6531",
+   "id": "dba67a85",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "5b982c9a",
+   "id": "94092bf2",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "51068280",
+   "id": "23b58907",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -88,7 +88,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
index 4d4b2c7c..34fe50a8 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "64348864",
+   "id": "0b35b944",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "9727c5ce",
+   "id": "29680359",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "971d065c",
+   "id": "fc256ca0",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -88,7 +88,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 0ee8a685..b0cbaba8 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "6eddf2b4",
+   "id": "55760923",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "f9d51a40",
+   "id": "4a32aa9e",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "7f1fbc04",
+   "id": "f97d50d0",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -96,7 +96,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 66d1b50b..11808e2f 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "483d10f9",
+   "id": "4df8f45e",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "9676ab65",
+   "id": "67b99f35",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "63e93f97",
+   "id": "b5430410",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -96,7 +96,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
index 12b758ad..4bb345a1 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "d065cad0",
+   "id": "7a2a1370",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "e37f4bd4",
+   "id": "2fd172c4",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "9c82fd19",
+   "id": "cda0eaa1",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -96,7 +96,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
index 59e708ab..788640c6 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "ce405495",
+   "id": "e812c86f",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "13144a3f",
+   "id": "d0f6af9b",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "49a61563",
+   "id": "0adf71f5",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -96,7 +96,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index dbc6f923..e68012f6 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "ecfd021a",
+   "id": "e7cb26ed",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "dc72df71",
+   "id": "e793914e",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "58093441",
+   "id": "c195bd7a",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -101,7 +101,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
index fd9b1c4c..6af4ae15 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "15957377",
+   "id": "e8b7d6ed",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "885fc3d2",
+   "id": "2bf488d1",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "896f63df",
+   "id": "365efccf",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -101,7 +101,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
index b7aed2d5..4c3ce0a0 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "d6d11478",
+   "id": "79cd14f3",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "f2f1431e",
+   "id": "d1a6d921",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "0cf2af2a",
+   "id": "bb5dcb0c",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -101,7 +101,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
index efaf3523..4955320b 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "de6ff3ec",
+   "id": "4f521809",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "417eeeef",
+   "id": "2708d3f3",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "ca0cae5b",
+   "id": "ea44d52f",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -101,7 +101,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index dd706951..d837eae9 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "15cb65e5",
+   "id": "0424b988",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "53cd1f32",
+   "id": "c3f57eec",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "e86a08a1",
+   "id": "55a15498",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -110,7 +110,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 82111fa6..fc1059cf 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "6e76c1eb",
+   "id": "ac02d024",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "bf52c7d8",
+   "id": "2d623925",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "f437a8ab",
+   "id": "1f8f50e9",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -110,7 +110,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
index 7e463089..c130e848 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "7d6ec5b2",
+   "id": "9152da76",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "ee02c25f",
+   "id": "25cf18d0",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "88405d24",
+   "id": "d020bc66",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -110,7 +110,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
index df21f797..5dcce68b 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "2485a132",
+   "id": "2ef939bd",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "8baa634b",
+   "id": "692679f6",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "36ccfa74",
+   "id": "f987f61b",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -110,7 +110,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index 46d8ae99..f5a686ef 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "0abe6a13",
+   "id": "4f34a0bb",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "e2a3275b",
+   "id": "93016563",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "b561368e",
+   "id": "a1a3e64e",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -91,7 +91,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
index 0ebb75dd..f155ab6b 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "794a1a1d",
+   "id": "ed0b2b0c",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "a007d517",
+   "id": "d2699a2c",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "aec3ff5d",
+   "id": "6b27e7cb",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -91,7 +91,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
index 0cab0d89..322d2083 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "14216bda",
+   "id": "4f8d8339",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "45cc7642",
+   "id": "2ce1f589",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "4fc7cfb7",
+   "id": "8278d9d0",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -91,7 +91,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
index f4ed61b8..173f876d 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "db654722",
+   "id": "01b028e2",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "fd6660cc",
+   "id": "b9b06718",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "992fa347",
+   "id": "e4882174",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -91,7 +91,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 8a2739c1..9083e937 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "1df418e1",
+   "id": "01b82a6e",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "c8a8035d",
+   "id": "bc9bbe08",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "258409d2",
+   "id": "f1293294",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -99,7 +99,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 9d78702f..9b620840 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "c847d8f1",
+   "id": "032bcad3",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "c76a64b9",
+   "id": "35ad585b",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "2df80b01",
+   "id": "79acd4c4",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -99,7 +99,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
index b2bd06db..cad66098 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "7166d08f",
+   "id": "c12c8608",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "53250c44",
+   "id": "8a36342c",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "f2e3b767",
+   "id": "7b8049cf",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -99,7 +99,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
index 000434b7..9e58628b 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "98c89ce6",
+   "id": "beecd470",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "77a07dfd",
+   "id": "ac64fc0e",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "6fe97fd9",
+   "id": "baac1275",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -99,7 +99,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index 2d79d473..da38d31f 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "e68481ac",
+   "id": "564b267f",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "c519fee9",
+   "id": "be18905a",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "2376b635",
+   "id": "584cf86d",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -104,7 +104,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
index e2e67849..5442587d 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "8a183a27",
+   "id": "35f7ee76",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "ccb70fdf",
+   "id": "54ab58ab",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "f988f793",
+   "id": "420ee452",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -104,7 +104,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
index e970f452..f05eda67 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "0b8c1763",
+   "id": "db58acee",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "0e10754a",
+   "id": "8486dd91",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "e98ea612",
+   "id": "a842accd",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -104,7 +104,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
index dc774b18..3bb71f8a 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "ee0a7507",
+   "id": "a30ee9b5",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "a09165a6",
+   "id": "c273d9c7",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "835943f6",
+   "id": "faa5e4fa",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -104,7 +104,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 3fec501f..22ecdd1e 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "76168760",
+   "id": "d4538f9d",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "1e8d0f56",
+   "id": "bff8ed5a",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "28fd0be1",
+   "id": "d1c477bd",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -113,7 +113,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 7c0aad4a..2bd0a522 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "556a7340",
+   "id": "ab0b5260",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "54813189",
+   "id": "6b90e116",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "74f57a44",
+   "id": "ca036472",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -113,7 +113,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
index 23d125c4..60fa8ce4 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "a1ecbacb",
+   "id": "5271c30c",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "889176b8",
+   "id": "3a44fb1e",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "8253e018",
+   "id": "f69b3d45",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -113,7 +113,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
index 32efead1..85eb4601 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "1ca5ee1c",
+   "id": "7a821fd5",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "b159ce8a",
+   "id": "42d25e36",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "c8e4aea2",
+   "id": "93ef697a",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -113,7 +113,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index 11cbc3a1..d4de5dc8 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "8c301ea0",
+   "id": "e8689c2d",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "1f05c93c",
+   "id": "a1c127f2",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "08a92331",
+   "id": "46fec5ce",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -91,7 +91,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
index f58982e6..27c314f8 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "483bb8a9",
+   "id": "99d1d453",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "6465e8c2",
+   "id": "c63d0a48",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "06270192",
+   "id": "b9729894",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -91,7 +91,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
index 78b039aa..ea9c9c8c 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "c97fa7c3",
+   "id": "d1c55db5",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "ca6e65f2",
+   "id": "a814545c",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "db57cf90",
+   "id": "021784db",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -91,7 +91,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
index c3686735..eacbe79c 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "5f138d2b",
+   "id": "cb73273a",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "a529128c",
+   "id": "91a3cb0b",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "344a0ddb",
+   "id": "4cf79235",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -91,7 +91,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index adc3aa5d..5b25712f 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "81bb467b",
+   "id": "aac57b06",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "9f2106e1",
+   "id": "caef3cdc",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "53017985",
+   "id": "8233b23e",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -99,7 +99,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
index bd1fc431..7d9bf50d 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "cbc4e83d",
+   "id": "8b2384fa",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "904e76cb",
+   "id": "d00ce284",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "46514c85",
+   "id": "2ee339bc",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -99,7 +99,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
index cf6f1962..28cbfc8d 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "7f3f7af5",
+   "id": "66d8da36",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "5740ccb2",
+   "id": "ae5bde0f",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "0e7ea107",
+   "id": "c88fc81a",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -99,7 +99,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
index d1a392cf..4b51d94c 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "d804f5d5",
+   "id": "2bbd01b5",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "017c9ef5",
+   "id": "83e1e689",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "0b10cf49",
+   "id": "a899d188",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -99,7 +99,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index 9ef514db..007976f4 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "5505a029",
+   "id": "48848d6f",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "c07fa98a",
+   "id": "af47faad",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "1e3af43b",
+   "id": "740f2e12",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -104,7 +104,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
index ee40d166..d60ba86a 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "585befbf",
+   "id": "d08c36db",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "c3dfe70b",
+   "id": "0bc0d2d0",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "fd8d9dcb",
+   "id": "c6f30c01",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -104,7 +104,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
index df966623..f4706bc9 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "7bee08a2",
+   "id": "1bd9df8d",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "85d4b0e1",
+   "id": "3922d3fa",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "62181162",
+   "id": "492f61b3",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -104,7 +104,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
index bf7acc4d..d84e5cc0 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "fd0e557b",
+   "id": "1c19e2c4",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "4b70a4a7",
+   "id": "5bf6c074",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "28c954bc",
+   "id": "bd053e6e",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -104,7 +104,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index e5eea935..365d71ee 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "66974322",
+   "id": "27d0f5e7",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "f63a06aa",
+   "id": "5549a197",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "205a13b4",
+   "id": "042a046d",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -113,7 +113,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 2acaafdd..444cb3a9 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "bba75920",
+   "id": "4165abb2",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "e381ab48",
+   "id": "760f611e",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "c5b8ad41",
+   "id": "682a4ff9",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -113,7 +113,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
index 5564353f..db978f9f 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "9be1efc5",
+   "id": "894c43f8",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "f7144595",
+   "id": "3a32ceff",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "4fbae740",
+   "id": "bcabdb1c",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -113,7 +113,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
index c08aa075..9c680f6f 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "0bba8718",
+   "id": "7a4cae12",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "c7169d2d",
+   "id": "d5ec03d0",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "7fe08926",
+   "id": "a33480af",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -113,7 +113,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index e4fb7356..e055b45f 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "06196560",
+   "id": "fb0d681a",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "8c15668a",
+   "id": "bd757100",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "96c51440",
+   "id": "0b5e872d",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -88,7 +88,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
index 93221b51..f86b5ead 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "b5b04d37",
+   "id": "1b659b77",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "c9cfd99b",
+   "id": "c05447bd",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "f8af6b42",
+   "id": "33c6fa1e",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -88,7 +88,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
index 787bc13d..bf845afd 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "c927f78d",
+   "id": "e9d3c2c4",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "866fcdfc",
+   "id": "55744f81",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "7ba1feef",
+   "id": "9f60497a",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -88,7 +88,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
index 87d936f1..2fa1b787 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "7503d662",
+   "id": "ab0b9dd3",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "1ee6657e",
+   "id": "a34875c3",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "dca25e1e",
+   "id": "290dd02d",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -88,7 +88,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 10e553af..f4d7c38f 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "7eb39dc3",
+   "id": "03170d29",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "2394904c",
+   "id": "ecc584a9",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "d2bff280",
+   "id": "950ec2f1",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -96,7 +96,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 0b36c075..21d0350a 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "b3a07643",
+   "id": "e9ed8726",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "0c852715",
+   "id": "cf22acde",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "18654843",
+   "id": "f80a6507",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -96,7 +96,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
index 49588c33..d82f198b 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "a93a512d",
+   "id": "e17c0a64",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "fe6ef939",
+   "id": "6a345215",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "db4f433c",
+   "id": "4f30fbb3",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -96,7 +96,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
index 097a1432..b06523a8 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "570d7a4f",
+   "id": "ad6e3b66",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "42077840",
+   "id": "e7e2c60b",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "60d4bd17",
+   "id": "c06f9b4e",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -96,7 +96,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index 41b6e645..f1ff1f3c 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "61e04649",
+   "id": "537a3837",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "5f695467",
+   "id": "3a27c662",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "1f6f04d7",
+   "id": "56ada987",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -101,7 +101,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
index 5dbc9fe4..50db0ffa 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "0b4efb21",
+   "id": "8a87b590",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "d3d630de",
+   "id": "cca8da46",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "7983ca24",
+   "id": "3ad993e5",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -101,7 +101,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
index c1ca64a2..c8418c49 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "035bccc9",
+   "id": "76437655",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "d9174a91",
+   "id": "92c5abcb",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "47f0db62",
+   "id": "ac7a7896",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -101,7 +101,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
index 8a7f37b4..c643070e 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "fee87b84",
+   "id": "ef80a896",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "de7fffbf",
+   "id": "34c9784b",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "dcda93a1",
+   "id": "d5dffbbe",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -101,7 +101,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 3b8c6665..a8438efe 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "a4dc8401",
+   "id": "98e35885",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "8a909e4a",
+   "id": "556c9c14",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "70b20535",
+   "id": "b26f7c81",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -110,7 +110,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 869c09c9..84f133d1 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "aedf7498",
+   "id": "a7e8c15b",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "28f3b7bb",
+   "id": "8d5d292f",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "899386a0",
+   "id": "494faee3",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -110,7 +110,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
index 67f66f10..57b6969c 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "08163dae",
+   "id": "4b956abc",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "10b4aea1",
+   "id": "7e1996c5",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "beed653f",
+   "id": "88af34b1",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -110,7 +110,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
index b95d432b..cc8c4f80 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "f9717064",
+   "id": "8266a27f",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "19a9e262",
+   "id": "1ea12a97",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "3b55a083",
+   "id": "857e66ba",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -110,7 +110,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index 798be757..b2ce7bfa 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "5f883e28",
+   "id": "c3e90a79",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "95b22617",
+   "id": "350660d6",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "8e10215f",
+   "id": "f9baf005",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -88,7 +88,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
index 8385c229..91bc900d 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "de7d60f6",
+   "id": "88c09317",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "66f592fe",
+   "id": "a00ffc0a",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "c44aca04",
+   "id": "8b6ade1d",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -88,7 +88,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
index 4acec4a1..5cbda727 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "c8786657",
+   "id": "10eabab4",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "a9f933d7",
+   "id": "c46bc6b0",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "e27d6a86",
+   "id": "bf99e4d4",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -88,7 +88,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
index 60fcd278..05d48b6c 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "4f8bc3bb",
+   "id": "88394ebb",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "62bcd964",
+   "id": "dc402aab",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "defcb937",
+   "id": "fd1b31a6",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -88,7 +88,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 6dbed201..936eb490 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "7bc8af47",
+   "id": "363d79bd",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "18648655",
+   "id": "99749769",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "6c26f6f9",
+   "id": "09cc18d4",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -96,7 +96,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
index f243e0f4..d5ce9a40 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "7bb6483c",
+   "id": "750f9408",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "2cb9d5f8",
+   "id": "ead79667",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "f2269f8a",
+   "id": "4fdbf010",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -96,7 +96,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
index 6105c206..87307ce3 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "4a8e783c",
+   "id": "2e6ac0e1",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "d69cd32b",
+   "id": "421e1084",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "b7926b93",
+   "id": "dee3074b",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -96,7 +96,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
index 911cf7ed..f152b25c 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "88572407",
+   "id": "47320cfd",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "90697771",
+   "id": "53477271",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "2fa2ee7f",
+   "id": "9fe7927a",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -96,7 +96,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index d2f73ba1..8854e1f6 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "9416080e",
+   "id": "a544a1eb",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "352cd827",
+   "id": "8f559884",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "9625df6e",
+   "id": "3c96a1d2",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -101,7 +101,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
index 23749c91..e60eecd8 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "78bc0ab9",
+   "id": "bca976bc",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "385de27e",
+   "id": "8ba5053f",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "d67c6793",
+   "id": "0047fad1",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -101,7 +101,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
index 180231ad..b412da89 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "7df1ff62",
+   "id": "974e2332",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "09dcceb3",
+   "id": "5b087d2f",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "1d7ecf63",
+   "id": "d44b3452",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -101,7 +101,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
index ede36255..97363ddf 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "1f865567",
+   "id": "e663dcc2",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "d0d39ad8",
+   "id": "6e5f6721",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "249f12ae",
+   "id": "7362000e",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -101,7 +101,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 109abdd2..dd53cfe0 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "ec225493",
+   "id": "e86b40f7",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "e067fd25",
+   "id": "b5021e27",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "2d053aa3",
+   "id": "81f00e5e",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -110,7 +110,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
index d3ca5906..564db607 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "33dde906",
+   "id": "a1a38410",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "2b37f03e",
+   "id": "22bffe5a",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "3f192e28",
+   "id": "3aa7ad75",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -110,7 +110,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
index 3e5d2f80..8d7b7f2d 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "fddc2e00",
+   "id": "a6ba7f8e",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "5b0beb81",
+   "id": "aabdc063",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "44b46b90",
+   "id": "a48c8444",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -110,7 +110,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
index ac426231..a4bb10db 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "dce6dcb3",
+   "id": "18d37d61",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "a9076ea5",
+   "id": "829e0708",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "5d656bfa",
+   "id": "1fa35911",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -110,7 +110,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index 71534884..9ea63708 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "9ce0963c",
+   "id": "f531276c",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "828a000e",
+   "id": "2e7bdb4e",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "1b4b6d5d",
+   "id": "1ecf4bea",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -91,7 +91,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
index e0cc7ec8..82fddc0d 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "48b219db",
+   "id": "753940d0",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "48931b67",
+   "id": "73cfe2ef",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "a56eb0e1",
+   "id": "a4f487e4",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -91,7 +91,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
index 4c8da6e4..f96e360e 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "89c8cbe7",
+   "id": "241afc3e",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "a24467c5",
+   "id": "0eaa5d9e",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "c9a4d744",
+   "id": "d44737e9",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -91,7 +91,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
index cb37fbb9..1ada392c 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "ccc2abe5",
+   "id": "9b8e4cbd",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "2061a729",
+   "id": "9d7af2cc",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "5f3cde77",
+   "id": "04216b6c",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -91,7 +91,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index d9877115..62e7a108 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "f3830a5a",
+   "id": "90cfa28e",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "f5176021",
+   "id": "086366cd",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "ec20e32a",
+   "id": "bf451783",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -99,7 +99,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
index e0107aec..53f05eef 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "0b115d08",
+   "id": "32ed1f45",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "63d84422",
+   "id": "95aa5e0b",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "76155efd",
+   "id": "aafa134a",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -99,7 +99,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
index d1de842e..ac47be34 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "9d688797",
+   "id": "9b951497",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "ceb4761e",
+   "id": "ce416adf",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "f70301ed",
+   "id": "e33db031",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -99,7 +99,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
index d585f273..d436cb97 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "23aeb714",
+   "id": "86c2a6e1",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "f9c97bc9",
+   "id": "9775fae8",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "f508066c",
+   "id": "442413c8",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -99,7 +99,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index 60896657..7b9e4851 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "78e24824",
+   "id": "e991ee52",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "880a96d2",
+   "id": "a7f98a32",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "5146b3c5",
+   "id": "5c3cc8a6",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -104,7 +104,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
index d93e0da7..0bacde49 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "919d57b1",
+   "id": "846ed55c",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "2180b3b4",
+   "id": "ed992b2e",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "c61ff920",
+   "id": "eef293f7",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -104,7 +104,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
index 38c9b607..6c0df6e5 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "5555def8",
+   "id": "aec816e3",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "1f7c3b74",
+   "id": "3a4af067",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "791753fd",
+   "id": "99cdad21",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -104,7 +104,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
index 892f13f6..b33c8950 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "9a47da00",
+   "id": "fca24db9",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "feda9bc4",
+   "id": "a64ee449",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "94f93116",
+   "id": "96c8f530",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -104,7 +104,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index f5f12740..bf5e7a93 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "524c1fe6",
+   "id": "dabd2ce9",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "a087dbb3",
+   "id": "7928dee0",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "01e417b2",
+   "id": "96e42d46",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -113,7 +113,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 9740b9a9..bf55c1fd 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "b3088710",
+   "id": "44be0063",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "367a91d1",
+   "id": "8d6db7ef",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "ec23c31c",
+   "id": "aaf5c609",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -113,7 +113,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
index 7b3ea78e..d06dd63f 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "6a12327d",
+   "id": "b03c4257",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "15969a8c",
+   "id": "fea2a9fb",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "e7a7659f",
+   "id": "bcc55158",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -113,7 +113,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
index a76d3e0f..dc396548 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "eb0831b2",
+   "id": "6cbb4127",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "ef8f0313",
+   "id": "f97017db",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "e6c88081",
+   "id": "fba1d079",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -113,7 +113,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index e6fe14af..4718125b 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "1724cfdf",
+   "id": "07536673",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "70262bbd",
+   "id": "cefff47e",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "3be7d42a",
+   "id": "4fab7d0a",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -91,7 +91,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
index 904dd085..6629c82f 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "e578a766",
+   "id": "9d61a965",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "6954032d",
+   "id": "229e566a",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "be25634b",
+   "id": "fdf3b5e1",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -91,7 +91,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
index b70e71fb..aba4fcf8 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "c358f2dd",
+   "id": "13ef6a2f",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "e7043e7d",
+   "id": "e3bcc131",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "a1135be7",
+   "id": "a13fe2be",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -91,7 +91,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
index 7c62fe90..77cb1e49 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "861d7349",
+   "id": "62b52722",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "941488fd",
+   "id": "4096b351",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "f1f76a26",
+   "id": "4dad6461",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -91,7 +91,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 56536a88..83acddef 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "d16621c4",
+   "id": "cd6adab0",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "8e7003a4",
+   "id": "48996fab",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "1b1c94aa",
+   "id": "844dd28f",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -99,7 +99,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
index ee19d6a5..2c900195 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "f6c362ca",
+   "id": "08899086",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "1121ef83",
+   "id": "017e9fd5",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "9f5619ee",
+   "id": "3f40ef77",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -99,7 +99,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
index 8e2e1b37..bd427809 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "96332805",
+   "id": "8104245a",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "ac76eb43",
+   "id": "92d0c3d0",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "901a7322",
+   "id": "636bffeb",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -99,7 +99,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
index 23135a4e..3304b1b6 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "1b858a0b",
+   "id": "d9ce4e81",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "206c59f2",
+   "id": "8d9bcaa2",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "126c2ab8",
+   "id": "90b166b6",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -99,7 +99,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index 86446fa2..8f445020 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "15916a33",
+   "id": "19252157",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "3de3983f",
+   "id": "6fbe2549",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "1c50a890",
+   "id": "ea152c1c",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -104,7 +104,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
index b1c2eb10..438a7691 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "d6e7c15a",
+   "id": "9d04b7d1",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "75995219",
+   "id": "622b3ba6",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "e0a65966",
+   "id": "85b96711",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -104,7 +104,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
index 999e568c..2b5984b2 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "00c1326c",
+   "id": "21382c66",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "dde707bc",
+   "id": "f0835413",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "9d0b4d10",
+   "id": "d3da23ab",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -104,7 +104,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
index 9a7f5095..308a1c5d 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "a37a944a",
+   "id": "cf401c5d",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "7fc27888",
+   "id": "c2335b7b",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "83b58039",
+   "id": "3fcbb417",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -104,7 +104,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 1e703d99..c34bb350 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "1e987e88",
+   "id": "d22e867c",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "6dbdbfb8",
+   "id": "232ac431",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "1298bef9",
+   "id": "80e91bcd",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -113,7 +113,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 5f9fed11..24eb861f 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "5d65cf9c",
+   "id": "6aa66782",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "5d77d86b",
+   "id": "99c28ea7",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "7b7e49ca",
+   "id": "5b151317",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -113,7 +113,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
index 2c1a65cf..5e434750 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "7d358cb3",
+   "id": "493d2310",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "ca614499",
+   "id": "9c9ca7bd",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "a1942565",
+   "id": "499b38d9",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -113,7 +113,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
index c93712e7..97d51afb 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "27543e5f",
+   "id": "2f074c20",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "52c311e6",
+   "id": "1cd635a5",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "d18cab76",
+   "id": "7f8fce1b",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -113,7 +113,7 @@
     "            model=Models.FULLYBAYESIANMOO,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index 9bb763f3..1615286c 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "3ee7a8cf",
+   "id": "c5b1ed8d",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "752fd538",
+   "id": "459b19ca",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "62901c55",
+   "id": "841f6dfc",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
index b1c091a3..402e06a2 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "6d55bb2d",
+   "id": "f2f942e8",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "fe94bb5e",
+   "id": "46f55dcd",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "a803af8d",
+   "id": "b2d5bc1f",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
index c96a4207..7ba91b86 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "a3176a44",
+   "id": "dcb23ebf",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "dd20e82b",
+   "id": "a0bd3207",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "8261aa97",
+   "id": "f4a8232d",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
index b176c6f6..a5f19559 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "d362efeb",
+   "id": "29d498bf",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "5b958208",
+   "id": "fedf28c8",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "ab7b1716",
+   "id": "2d76305d",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 92c2dfbb..56231ed7 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "75907ce6",
+   "id": "c988f615",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "aa21a502",
+   "id": "afa1bb3c",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "30354644",
+   "id": "a444d24f",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
index b717bf74..3f82e109 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "a1c7d9a2",
+   "id": "83ffb5e5",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "4bf976c5",
+   "id": "20e70001",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "496db9a9",
+   "id": "04427e20",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
index 898c7c8f..f1b16ca1 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "a078dba7",
+   "id": "8a46b0de",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "0af0efeb",
+   "id": "f408e336",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "caa88e7b",
+   "id": "cf264b34",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
index 14976ac5..4ecd314d 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "4c129188",
+   "id": "449439b2",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "7db4f059",
+   "id": "b461cafd",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "fabf5264",
+   "id": "1dd39c8e",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index 1b2fa728..7c79c9b7 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "264bab42",
+   "id": "cd0d31e2",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "50cac0cf",
+   "id": "54b16940",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "5a6b2bc8",
+   "id": "e2229984",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
index 6633f28c..e2e7ded9 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "50358605",
+   "id": "8b37c198",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "b594fb48",
+   "id": "29c316f2",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "e40633c6",
+   "id": "919be014",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
index cada4d84..255a051f 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "5eaea685",
+   "id": "d6341d22",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "935c307b",
+   "id": "0c07e942",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "5d880f35",
+   "id": "29ee2428",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
index 0e2dfa79..310c10fa 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "57f2b5eb",
+   "id": "1ddb7052",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "99233d68",
+   "id": "0d078cb2",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "9ad28fb4",
+   "id": "b78abc0d",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 8cc9fd69..577b9ea2 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "26ec0a51",
+   "id": "6d89102f",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "f20989fc",
+   "id": "104589f8",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "8a23c565",
+   "id": "4ed693bf",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
index b4d51e41..e4dc95bf 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "7dbaaa04",
+   "id": "1bbecf4b",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "b522c2c3",
+   "id": "8faaa84c",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "ba37894b",
+   "id": "763ae647",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
index 121d083c..8d96eb0e 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "ab92e00a",
+   "id": "3e1af570",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "3d62f0ab",
+   "id": "670d8796",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "1928b330",
+   "id": "39356839",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
index 4e144423..f11512d9 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "b8921d32",
+   "id": "87d5fb8b",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "e17e6058",
+   "id": "0fde0965",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "85e32732",
+   "id": "a30e92ec",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index 28f146f2..09029358 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "66751f13",
+   "id": "71c4db1d",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "161f5e79",
+   "id": "4b28c811",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "331233b9",
+   "id": "e19ffb41",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
index a87db33d..ed07c808 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "28169ff2",
+   "id": "63760e9d",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "d8909fae",
+   "id": "8fe29240",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "6e125641",
+   "id": "d6b9152a",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
index 8cac7b5d..73f6846d 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "e2ca695c",
+   "id": "37061535",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "6015bbf6",
+   "id": "8260a6ef",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "54701916",
+   "id": "8a343afd",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
index f8eeb647..7234f53c 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "0ce3adc4",
+   "id": "6f78be0e",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "2db1e607",
+   "id": "52154cb1",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "a0c3ed2a",
+   "id": "44a9960c",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index e8cd7132..a75e2b81 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "e71f149a",
+   "id": "dacd14fb",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "280eafb1",
+   "id": "65a5539b",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "8c1195c8",
+   "id": "e4e8542f",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 57c91d66..bb1b951d 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "571af0a9",
+   "id": "79314c81",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "1cf171a2",
+   "id": "122f9811",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "d80ddc4a",
+   "id": "c6c8df4b",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
index 73645c77..dc65385c 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "904387af",
+   "id": "9bd5de9a",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "d1e7fbd3",
+   "id": "47bcae54",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "c49a8580",
+   "id": "c18ec0db",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
index 62f45887..1c9da47e 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "919209f0",
+   "id": "fdbff250",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "ea7f51ca",
+   "id": "5a496931",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "e507c389",
+   "id": "638f09ab",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index c6e13408..fc3833cf 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "08fcb7d8",
+   "id": "8834714b",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "d789cfa1",
+   "id": "63f9b022",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "14012258",
+   "id": "3cf7d666",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
index 30f8cde4..a57e69c1 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "9eb03999",
+   "id": "4e53b9c7",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "7cf448b9",
+   "id": "4660ffd6",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "e991ccd7",
+   "id": "ddffcacd",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
index 5c5751e2..ea3043a8 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "a68e130f",
+   "id": "28e0b2c0",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "e711e88a",
+   "id": "83d9d661",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "af715f14",
+   "id": "18b96646",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
index d4d5c916..54dadbc0 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "fafb2891",
+   "id": "63d04cd1",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "4d40ccfb",
+   "id": "89d4c979",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "1b9d131b",
+   "id": "304ce4db",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 51b36769..7572d61b 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "8e4479de",
+   "id": "f6ffead2",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "7adad874",
+   "id": "eabe3188",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "bbd1439e",
+   "id": "d55ae8cb",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 1a80ea13..50f7e25d 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "44419614",
+   "id": "7cc1e279",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "13087ae6",
+   "id": "218a19c6",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "975d81b3",
+   "id": "a1263222",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
index 2a4004e6..0f24d8df 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "c1535d24",
+   "id": "f82fc11f",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "4aa17d9e",
+   "id": "62ef0d01",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "c38f4ac6",
+   "id": "5e472a46",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
index 55a0a772..aa114ae7 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "71567e9a",
+   "id": "d3fd735d",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "0eeb9eaa",
+   "id": "fc4b727f",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "69c1c02b",
+   "id": "5ca02398",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index 3b7cb6c2..5c919ad3 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "64beb32c",
+   "id": "49fe6459",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "e2ef10f2",
+   "id": "3a3b0be1",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "c1b7747b",
+   "id": "c3940a16",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
index fe834299..417093b4 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "a1d866e1",
+   "id": "c610f451",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "8ef872c9",
+   "id": "db9b1883",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "95e76289",
+   "id": "90160442",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
index 2aba82cd..5299fa70 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "3c9139be",
+   "id": "c735fcef",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "87c4e17f",
+   "id": "5c7ab843",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "9af8a538",
+   "id": "d6d0281a",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
index 6a156cdf..f1227207 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "c0ea50c8",
+   "id": "b47e6e5e",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "b01643ea",
+   "id": "bfaf8cbb",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "db4b3d92",
+   "id": "7e15d6e0",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index fcd3c84a..e5249898 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "133fa0dd",
+   "id": "965d125d",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "4581e37d",
+   "id": "1469810b",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "2b835179",
+   "id": "4cc289a3",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
index f19e8bf6..dec8c055 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "0bbef435",
+   "id": "b86a2204",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "b5c220c5",
+   "id": "7980aa80",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "2ef6d44c",
+   "id": "82f4a3a0",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
index 4aa4c529..41c4e06e 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "dcdd5f8d",
+   "id": "6f979da2",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "de53056d",
+   "id": "aab96e3c",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "5fd4f7de",
+   "id": "1a01ddea",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
index 23dee69d..25bc331e 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "00f1f88c",
+   "id": "cdf7e66b",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "a39b66f6",
+   "id": "b774c696",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "94f11ba6",
+   "id": "e83fda2d",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index 9127cbba..3c21d395 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "70a983c1",
+   "id": "300bae1a",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "e81f440d",
+   "id": "8fd5f17d",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "51230ee1",
+   "id": "d2f19b75",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
index 7155dbac..bc008958 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "ea7d2c0b",
+   "id": "f81256b7",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "bb73ec7b",
+   "id": "a62764ad",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "26741f66",
+   "id": "d25f404e",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
index dd9dfa52..33603834 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "c22b4a0e",
+   "id": "2f8b76ce",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "4e866bf3",
+   "id": "52eafec6",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "f94019aa",
+   "id": "eb061673",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
index bd476bee..bbbd46e2 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "ac7f19b5",
+   "id": "604f1ad4",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "c6071d59",
+   "id": "efee1bff",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "b8d8d003",
+   "id": "9691b684",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 8bba81f7..1f109aae 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "81af0ba8",
+   "id": "2cdf8557",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "35d9f698",
+   "id": "1544a398",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "c09feb45",
+   "id": "9f75110d",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 33c1953e..de0ae905 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "852068ac",
+   "id": "25ff7300",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "f25df77a",
+   "id": "cf73b178",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "dc34a5bb",
+   "id": "5cbf3161",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
index 430167de..c1a1a214 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "e89d8d58",
+   "id": "0e3fcfa8",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "349182e2",
+   "id": "1b5f790a",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "4af7212b",
+   "id": "ed325408",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
index 275536a6..acc25352 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "824e44fa",
+   "id": "431eeccc",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "adeb7fc1",
+   "id": "c7b4ba3d",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "32e973ab",
+   "id": "b81e4bcd",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index b5d0a992..d6d2d932 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "3b68d351",
+   "id": "9a09d89d",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "a4464ea0",
+   "id": "5da0b60f",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "835d38c1",
+   "id": "a71b2684",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
index 6022bdba..38aaf048 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "7d62afa0",
+   "id": "5fde6966",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "f9ed25a9",
+   "id": "a487c444",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "b6f1179b",
+   "id": "f9bb0f48",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
index 663de522..6b399783 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "e2ba087b",
+   "id": "9d34f7c2",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "66202533",
+   "id": "8c2d6daa",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "775c5735",
+   "id": "8f84c913",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
index 1abeeffa..422c4538 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "59b4a5ed",
+   "id": "386559c0",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "def4e168",
+   "id": "48c076b1",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "16826939",
+   "id": "a077044e",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 49a8f2c5..0cb7903c 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "c04d93a3",
+   "id": "b17ebc6b",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "3f66e025",
+   "id": "c2d4b128",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "d41b3810",
+   "id": "f3d7be71",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 22e3ad86..6db365bd 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "99bda5f7",
+   "id": "43121511",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "e5f6004d",
+   "id": "fc2c48da",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "a431d07e",
+   "id": "c70f696a",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
index 3204cb92..82db3ec1 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "540a45bc",
+   "id": "da20a9fa",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "1a19d924",
+   "id": "313fddcd",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "10ff5927",
+   "id": "574f5f54",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
index 39d4c6b7..30080072 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "92bf8a8c",
+   "id": "752ab345",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "f2797ec5",
+   "id": "d4244128",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "ea121aa8",
+   "id": "edb8e61a",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index aea268ef..6f92a834 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "42ad7019",
+   "id": "38624638",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "b4e5f00f",
+   "id": "a3ed3a87",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "6d15becb",
+   "id": "e7cf70d7",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
index 97da2de8..c84cf2d3 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "cdb123e2",
+   "id": "1a9aa422",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "8a2ff184",
+   "id": "7793a135",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "bb5fa5db",
+   "id": "ed5f5fbb",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
index ba062e69..e4651e38 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "ffadefa0",
+   "id": "3c71e9eb",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "ebf3fb26",
+   "id": "f503c2f5",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "56927c30",
+   "id": "a1efab2c",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
index 98251fca..feaaa71f 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "be08d625",
+   "id": "9f6eb4c0",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "9650c6c4",
+   "id": "a8d407f3",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "0b77215d",
+   "id": "3df9c153",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 700d5170..bff43bac 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "d54a5b37",
+   "id": "e8b2dffb",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "57496a4f",
+   "id": "90c74dc6",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "4445148a",
+   "id": "120d333f",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 58184637..b80e7199 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "b016552f",
+   "id": "3ce01242",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "51739063",
+   "id": "338c0e7c",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "731fa727",
+   "id": "4fe92330",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
index 4253ecc9..0044555b 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "61abb195",
+   "id": "c7a83517",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "b6373dee",
+   "id": "0429d116",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "bafdd340",
+   "id": "1117756d",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
index 637bc29d..b3bbd381 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "14d3e976",
+   "id": "fe093a46",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "f0336963",
+   "id": "a1eeb3d6",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "2aff7c23",
+   "id": "7ec7c0b2",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index 76c3f0b1..f2b73004 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "bfe5489a",
+   "id": "0f584241",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "daebb749",
+   "id": "760413e9",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "64c9596e",
+   "id": "4a69e6f8",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
index 70e3d2a7..439e548d 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "dfe22e9b",
+   "id": "21cff0f7",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "9dc4a9e7",
+   "id": "648e786e",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "2509dcc9",
+   "id": "e3c5a4ae",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
index 376672e2..8d6fb63c 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "88a9d2d7",
+   "id": "d208ce00",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "fdfa0de4",
+   "id": "c2d0742a",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "b4c65e16",
+   "id": "727166d8",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
index 7c88c8bf..142068b8 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "8bfbc34e",
+   "id": "bce2d788",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "0d44bd9c",
+   "id": "0bb12615",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "5b77b380",
+   "id": "6abbe49d",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 2e3d7b75..874cdf73 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "11b93f5a",
+   "id": "b95077cc",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "5378f13f",
+   "id": "9fbbd115",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "47c1997b",
+   "id": "54f2f232",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 650e88a0..a39fb3d0 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "1e29f7bc",
+   "id": "b7afb75b",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "4d2ee863",
+   "id": "72fd43c9",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "6ff34f6b",
+   "id": "c4deefaf",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
index dd1276b9..33b84a11 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "4f3ccc25",
+   "id": "0dc08f37",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "7fd71cf2",
+   "id": "0fe96c9d",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "e9831f16",
+   "id": "79eabf63",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
index 19e8c646..47979648 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "7da4a46c",
+   "id": "c8832c46",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "52b56c5b",
+   "id": "39703dcd",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "253dd37f",
+   "id": "25d5731d",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index c2f19165..96e204f8 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "fa88b970",
+   "id": "552e65fa",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "ca7b048c",
+   "id": "aa6773f9",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "6d368c67",
+   "id": "9883d9d8",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
index ec332980..6458cdf1 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "021f24db",
+   "id": "9f573510",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "6f91e757",
+   "id": "c293963b",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "4a3b4e12",
+   "id": "92b8034d",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
index 0435f71c..05f20c34 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "ddeedd65",
+   "id": "abe89e70",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "78eb1cc9",
+   "id": "cec24bf8",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "998d98cd",
+   "id": "57c5d3b2",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
index 7ff0a3ba..fdeecc68 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "59bff723",
+   "id": "63df3aa6",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "24f40165",
+   "id": "363caf30",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "48eb4f3d",
+   "id": "86ebcbaf",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index d30384c8..20bfc227 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "00425679",
+   "id": "964f2713",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "ca592a94",
+   "id": "a85d5204",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "68f12631",
+   "id": "a8fc4629",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 873a9f2f..d7543846 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "c8f0a79d",
+   "id": "ea1a7470",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "0bb12a24",
+   "id": "d67dd17e",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "f48304f0",
+   "id": "babd5667",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
index f8fcca97..67f8e2ed 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "ba58e3d4",
+   "id": "85e32830",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "cefd0733",
+   "id": "7827fbf8",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "2543023c",
+   "id": "d23bf560",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
index b12b732f..92e8d6be 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "d6e20075",
+   "id": "cbe85017",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "43410acd",
+   "id": "32463ae8",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "bf5582c9",
+   "id": "a7b552ac",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index 8206e281..96f47721 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "e7b42cec",
+   "id": "bf17cab9",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "5868144d",
+   "id": "d51de4bc",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "94dcd498",
+   "id": "73e859e4",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
index 39df7080..e797fb1c 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "7a531a99",
+   "id": "7cf55707",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "e4536aec",
+   "id": "19b2c2b0",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "86c83738",
+   "id": "e9a260de",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
index b6e52de1..cad2121b 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "5dbe85b4",
+   "id": "c690ec0f",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "c5388c62",
+   "id": "8954eb35",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "b2d2621d",
+   "id": "9e636e0a",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
index 6e97a5fb..83cdf2f0 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "6dd010c8",
+   "id": "6f5e91a9",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "8b6bf742",
+   "id": "075363b7",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "8b277fa8",
+   "id": "338b5858",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index a6e52161..a53f7d21 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "0acd4eab",
+   "id": "ed4c5351",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "a2c6bca0",
+   "id": "69748dd8",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "4b19946c",
+   "id": "bc0a9786",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
index d3a47053..fe834f30 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "85ee869f",
+   "id": "a2a432d1",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "42d6cffc",
+   "id": "8c8d7c56",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "6e0e635b",
+   "id": "70a2132a",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
index 7214b1b5..2b91b0ae 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "433457ec",
+   "id": "84b832d4",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "a5df9e9c",
+   "id": "aba54bc2",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "b17e0979",
+   "id": "e1b65ad6",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
index a29ed5ac..22cd5842 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "f618b174",
+   "id": "5b5dd15d",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "14e8ae19",
+   "id": "d0efb603",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "fb220607",
+   "id": "b6599517",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index 76706c5c..c33c8254 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "a4959671",
+   "id": "08571f56",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "622350de",
+   "id": "9af9d1e5",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "0642f1c9",
+   "id": "46a56f37",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
index 7d86a841..c6f6e7b6 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "aa75677d",
+   "id": "59ab9814",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "c7c09732",
+   "id": "5080f084",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "3566e634",
+   "id": "8b6a9f1e",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
index 61125d55..61470842 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "b3b6754e",
+   "id": "20a31562",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "b33733db",
+   "id": "f1fa6b41",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "6c01a737",
+   "id": "91bee408",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
index 165dbd24..4bba4eb8 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "63ebfcf1",
+   "id": "18dc89af",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "120dcd06",
+   "id": "021fbbb3",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "bf7c83cd",
+   "id": "05d65e0d",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 32e08df0..9799f8a4 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "6050fd67",
+   "id": "fe6025e5",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "112e123b",
+   "id": "4e4f30fb",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "16934070",
+   "id": "6f4e653b",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 25399501..89dbda4c 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "dd900b54",
+   "id": "0749bb4f",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "eb24803f",
+   "id": "5080e1f3",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "97e465b5",
+   "id": "bf82a27c",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
index f375247a..63cfe7f5 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "2a05a300",
+   "id": "e09695df",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "1857530d",
+   "id": "5d03f4ce",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "cf8a3ea3",
+   "id": "3e5d3176",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
index 662beef6..dbc69084 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "a7f0ec88",
+   "id": "d4b9348e",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "2085f43c",
+   "id": "1d957bf5",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "029a3ad6",
+   "id": "4ce15530",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index c130d50c..da0e14af 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "f78be98d",
+   "id": "c2a51b63",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "fc4fcc6e",
+   "id": "10037754",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "be11535a",
+   "id": "33065d29",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
index 1030b5ae..a7db772a 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "83d2326d",
+   "id": "68b4b0d3",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "ca36a960",
+   "id": "1ab7f599",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "7a3e3531",
+   "id": "d97c177f",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
index a8df7d78..0a372fd0 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "7206387b",
+   "id": "c8219559",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "87decad3",
+   "id": "b87f3a50",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "bc84cb3d",
+   "id": "83aaf1f3",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
index 87cb5080..1ec130d9 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "1b163501",
+   "id": "b854b794",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "6f45d153",
+   "id": "47062066",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "0b06ed7c",
+   "id": "49507100",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 295ea841..db84bb71 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "66c4c6d0",
+   "id": "dd84dd07",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "c423984d",
+   "id": "6ba2c7f7",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "10d05b5a",
+   "id": "aca27f1d",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
index f3d23447..f4b03ef3 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "f9c3308d",
+   "id": "e2f8bc2a",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "d3d9a707",
+   "id": "680988ff",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "f920afe2",
+   "id": "3f3a766e",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
index d902a4b5..922811f5 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "46c40632",
+   "id": "a357a8dd",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "d874a2c2",
+   "id": "a4e2f1f7",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "a64439a4",
+   "id": "d056aeaa",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
index ef75a84f..4db34894 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "2e4089ad",
+   "id": "add1bc29",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "a21e776b",
+   "id": "e0a99a10",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "b725ee35",
+   "id": "f81b309a",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index de2a917c..f00a7e39 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "9fe35560",
+   "id": "6ce6a6dd",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "960b6ec2",
+   "id": "5975d1c9",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "be780080",
+   "id": "187fb763",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
index 40f62f07..474962e9 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "d6df4824",
+   "id": "3b36e058",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "985cb07d",
+   "id": "eee7e850",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "de8cb9d3",
+   "id": "8689d2ee",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
index 679b474c..d8e5dc89 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "9825a3be",
+   "id": "bc09f566",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "c544a004",
+   "id": "77c048a6",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "36f2faad",
+   "id": "44402b6b",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
index daa93af4..3f869aa0 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "b96a1a33",
+   "id": "dd4448a7",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "5dfbe9b9",
+   "id": "dfc49826",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "40fe2143",
+   "id": "6363602c",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index b9ed08cb..f99af8bf 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "616ac130",
+   "id": "86326fb8",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "f27b6c0c",
+   "id": "4f28ff74",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "f6bd06a0",
+   "id": "490fb670",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 59a20efd..d2091520 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "de4094d1",
+   "id": "d5d654c3",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "4582a2d8",
+   "id": "67a8f6ba",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "33ae325a",
+   "id": "97f4d55c",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
index fc62f867..77ba174a 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "da0284d0",
+   "id": "1295b2ab",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "08c6b8a6",
+   "id": "db808730",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "b5b97e4e",
+   "id": "16279c3a",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
index 12fdb7b1..3183faf4 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "04a77ad3",
+   "id": "6f194f5a",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "bac179de",
+   "id": "ee3bc8b4",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "7bfc6f55",
+   "id": "0e8e415b",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index cd049777..26b9113f 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "775ef956",
+   "id": "f9a48480",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "66d6b588",
+   "id": "c0730108",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "6b7e27cc",
+   "id": "50a3d1fc",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
index 3fd60e07..19d405e6 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "0bd95acd",
+   "id": "0ca822e5",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "2badb940",
+   "id": "92381846",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "e7200659",
+   "id": "a4f61fcd",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
index 1106e907..37d03b51 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "1aa87cae",
+   "id": "ac6180d8",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "6ed54413",
+   "id": "d0de4baa",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "692da4dc",
+   "id": "0c8d2c08",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
index 302d9938..d4868354 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "6970b75d",
+   "id": "3f70467a",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "753b5df4",
+   "id": "2c362ab3",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "4b056c14",
+   "id": "8edf8411",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index e9b16f36..3b9762fe 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "d6f2c8dd",
+   "id": "cf1b907d",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "f7c83920",
+   "id": "7e5fcd39",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "0c12e947",
+   "id": "1753a9dc",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
index d519fb98..3d8c7e11 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "c1f6db27",
+   "id": "3e6176e9",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "e0e3aaf5",
+   "id": "17aabf96",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "3a0f0388",
+   "id": "88bf21af",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
index 10d6a895..c8102a9d 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "7bb1fedd",
+   "id": "7de094aa",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "daa8ecf0",
+   "id": "a5f2bbb8",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "ac65e235",
+   "id": "c9384c20",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
index 18077cb9..82b27087 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "2462d081",
+   "id": "3c5684d4",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "0619fda6",
+   "id": "5e73602a",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "41d2d96b",
+   "id": "5b92304f",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index 947f1838..6c01c4c0 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "62dbdce9",
+   "id": "d509cc9d",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "b84a0030",
+   "id": "3e64d57c",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "30399e7e",
+   "id": "dc92e5fe",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
index a375bb8b..dd605ec3 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "c5b68548",
+   "id": "6fc1edaf",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "9ffbde4e",
+   "id": "aad987e1",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "45c03856",
+   "id": "3ce74dbf",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
index 0f31a35e..bffffe62 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "d1d137be",
+   "id": "47a4ba22",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "a16156b8",
+   "id": "2570bdf3",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "dd9818a0",
+   "id": "9d4e5634",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
index 60f60ddc..a48c6559 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "94e4d614",
+   "id": "8c59ad23",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "109ccf0c",
+   "id": "5d6cd215",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "77a31b6c",
+   "id": "498a30c2",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index c3e9abd6..a5225c98 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "f60d8a42",
+   "id": "fe7f845f",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "d19c01e2",
+   "id": "d0bf969a",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "45cc4c27",
+   "id": "233fa577",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 70fc466d..d4a5c786 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "d872e38a",
+   "id": "771279de",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "aca0647a",
+   "id": "a0d8c2fe",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "c61cce11",
+   "id": "e1a2d021",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
index 986144fb..44a2b8b3 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "8ddb2288",
+   "id": "42df4fac",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "9b88d60f",
+   "id": "76f8235b",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "1543d9ed",
+   "id": "1923e982",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
index d03b834b..22afc9dd 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "d68e876e",
+   "id": "747a8279",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "1b395503",
+   "id": "075ebc0c",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "3725be64",
+   "id": "79f8cf19",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index aa2afb27..45d4be11 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "26db6c86",
+   "id": "18fb9bec",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "71211020",
+   "id": "1d686f50",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "7415fa07",
+   "id": "5d65abdf",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
index 82a40617..be7e7cb7 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "1aaa3343",
+   "id": "98a41876",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "de1300a4",
+   "id": "1b7cc7e1",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "c1121652",
+   "id": "19ae1ec4",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
index a411bb90..70c0848d 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "a6b86eae",
+   "id": "83a72d2a",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "40c7b797",
+   "id": "9668381f",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "74ad10f4",
+   "id": "da7e37b2",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
index db040b1f..620b50ec 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "a5fc6b5a",
+   "id": "d76fc3b6",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "d8f7ffeb",
+   "id": "cbe95119",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "3eca7f22",
+   "id": "52911753",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 9b167250..0f4c285f 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "bdfa6f3a",
+   "id": "7421f6a1",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "facf4465",
+   "id": "c1bb7239",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "de73c9d0",
+   "id": "925c2ade",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 011d1ea4..99e1dd7d 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "a5c60b66",
+   "id": "c2186091",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "13dcdd25",
+   "id": "269a08d6",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "07ab0eaf",
+   "id": "a1498d6a",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
index 718380eb..32b999cc 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "4d47c485",
+   "id": "50dd0ec3",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "3e928323",
+   "id": "35a63d1e",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "756e8248",
+   "id": "947bda64",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
index 7b36f604..e7365cbb 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "3daced55",
+   "id": "8792ea28",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "d54027a7",
+   "id": "19b94a1d",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "35009923",
+   "id": "410e65fd",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index bdde1dcf..792ead1f 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "dfcdb9a0",
+   "id": "24c8ad59",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "b30761fa",
+   "id": "cccabf9a",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "119ea94b",
+   "id": "4f3d6c2c",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
index d2103ee8..c8534d66 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "35ce43db",
+   "id": "1a7cc631",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "4e818555",
+   "id": "a976d632",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "fc33f814",
+   "id": "f3797da2",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
index be611147..4951be96 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "326c85b2",
+   "id": "c541a56a",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "c9f2fea6",
+   "id": "2001a0f6",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "0fc58666",
+   "id": "786fe452",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
index d794044a..757090e4 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "12ae6295",
+   "id": "74b20cc3",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "5c0d8cb1",
+   "id": "eddf8777",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "101b0645",
+   "id": "6f2ddf57",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 9923476e..e58eb150 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "b32d51de",
+   "id": "ae86041a",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "175f8d6b",
+   "id": "78fbecf1",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "0ff2132a",
+   "id": "e7de7cbb",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 36b9e06e..d7899482 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "64cb2883",
+   "id": "fc294472",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "b84c7155",
+   "id": "3a0c411d",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "8cf636ac",
+   "id": "8564bf23",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
index 1965022a..71b08154 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "feec70ae",
+   "id": "cf73e692",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "4d76dab8",
+   "id": "47ee2cb6",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "f6625085",
+   "id": "d9529c63",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
index d641a236..4be72cf1 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "8916bfa2",
+   "id": "7fe15bb4",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "a2496b2a",
+   "id": "22318cf2",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "0beac96e",
+   "id": "46c89216",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index 46269d57..995d4b57 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "676203e8",
+   "id": "0ef1c565",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "3d6ea0ca",
+   "id": "4febbd21",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "a7a73d2a",
+   "id": "b5558f72",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
index 5d8f204f..1a35b550 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "83b06def",
+   "id": "f4daa3c1",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "6c5d2cd8",
+   "id": "6794241e",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "4275245d",
+   "id": "628bf944",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
index bdae0afe..52a90054 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "61720706",
+   "id": "c5972f15",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "6803517d",
+   "id": "0664b4c8",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "83114ded",
+   "id": "f0f2f127",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
index 9e13ecaf..69011aef 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "63942177",
+   "id": "f5b5d19f",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "9653b66e",
+   "id": "c78e43d2",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "3f3d33dd",
+   "id": "028bb017",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index b87f55d4..495aa41c 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "60b1e988",
+   "id": "74efae09",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "4093c0b1",
+   "id": "772acc17",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "c896e0b6",
+   "id": "c36a4ae1",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
index c0af701e..e7b92f78 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "d728df12",
+   "id": "2fecd76e",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "9ebc0948",
+   "id": "59c1b2c4",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "e95155c3",
+   "id": "fb40fc24",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
index c4d9dbe7..ee061550 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "8cbf229f",
+   "id": "5da27f0b",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "d2b594b9",
+   "id": "4f4dd14c",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "bb5fbb78",
+   "id": "36aa3a77",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
index 482334e4..9de9c3ea 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "f91fe19c",
+   "id": "1e74525b",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "3c9c9c08",
+   "id": "d874f093",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "5bba4de0",
+   "id": "3b910b7e",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index 2a8b284c..5e2dc428 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "482e7aa5",
+   "id": "f4cbf91b",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "408d0cb1",
+   "id": "9481828d",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "c865ce10",
+   "id": "64931a6b",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
index cf502215..6f5e69cd 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "6c7419b8",
+   "id": "3bb11100",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "48c66d09",
+   "id": "8a5d9901",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "ad8755c3",
+   "id": "8b81e4b4",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
index a65176ed..f5c61bcf 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "5f22440f",
+   "id": "ff144aab",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "bd777ac7",
+   "id": "12d2719c",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "9ce4feee",
+   "id": "d35d4dd2",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
index 10686a75..9d637e85 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "840a627d",
+   "id": "afda3be9",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "441a8237",
+   "id": "4599251a",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "82b7a128",
+   "id": "beacfc5e",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 892de12b..9a571528 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "18fc244c",
+   "id": "1d61b049",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "9532bc2e",
+   "id": "26d9abc9",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "d86c28f3",
+   "id": "297ef13f",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
index d9b85aa3..d7c6b597 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "5b5d34f8",
+   "id": "700de161",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "2341b195",
+   "id": "83094c1a",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "dda9663b",
+   "id": "317e66b4",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
index 6d8c3fca..163b8a80 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "b162e0b4",
+   "id": "74cd6aea",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "550ce0aa",
+   "id": "54869e8c",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "228f0666",
+   "id": "844a295d",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
index 8547b7a7..ed76205a 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "296bd728",
+   "id": "9836e9dc",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "4ae92af7",
+   "id": "062be324",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "eda3369b",
+   "id": "3a69d3d6",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index d1a30fda..8cf1dccd 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "854323df",
+   "id": "352382d0",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "db8441fb",
+   "id": "92e7693d",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "33f68ce9",
+   "id": "2f7b9d06",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
index 053c7253..0fdf43c8 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "65ef1bf9",
+   "id": "ef504c03",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "69c7d3ce",
+   "id": "c54728eb",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "99545f10",
+   "id": "538f84dc",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
index e0574017..3321de56 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "70d96228",
+   "id": "74cb1779",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "fa11eeaa",
+   "id": "051445a4",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "fb155ff0",
+   "id": "73a06ba8",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
index fca8ce84..99233848 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "068f5814",
+   "id": "46ec6788",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "e7f07fa1",
+   "id": "ee4cdc55",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "7456e537",
+   "id": "b87ed750",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 2aea8ede..7d34dadc 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "b4da2f4e",
+   "id": "f2d3c374",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "744a688a",
+   "id": "d49c1ea2",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "32f8e1aa",
+   "id": "5a180502",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 170a8739..1fb05c72 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "44e65665",
+   "id": "0dfae19d",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "b4f05a59",
+   "id": "5fd434b3",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "2bba1b04",
+   "id": "a1a483dd",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
index a7047f70..24cc4c7d 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "c7f05485",
+   "id": "0b01a27d",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "b8450fa0",
+   "id": "5f8601ab",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "01a7d9eb",
+   "id": "fd911fb8",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
index ab636ed8..df6577e3 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "abc9acaf",
+   "id": "662e7622",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "43499e3f",
+   "id": "792e771a",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "54a6d8f9",
+   "id": "345428ba",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index 62bbf45e..4b713653 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "12e3782f",
+   "id": "203616c0",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "afc2d2c3",
+   "id": "7898a77e",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "d07fc398",
+   "id": "c38772c6",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
index 0cfb5a1f..6d9f6ee4 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "01b7fddf",
+   "id": "21c13bce",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "52be21d7",
+   "id": "f1850774",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "d1421df8",
+   "id": "45881484",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
index fa2d6c71..fe1d0c29 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "391f9d72",
+   "id": "dffefd95",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "75dacc0b",
+   "id": "23ce5dc1",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "748dba04",
+   "id": "49c00b50",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
index 44ca623e..024ef2a6 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "d926e2be",
+   "id": "f71e97b6",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "1bf9fec6",
+   "id": "2b17f3de",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "ff94ad91",
+   "id": "cf98ffa5",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index dd625bc1..2642e321 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "cb1bc3e2",
+   "id": "51fc8073",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "bd56014e",
+   "id": "42848b1a",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "79235531",
+   "id": "bd0aeea3",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
index fbc03e60..b89f3131 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "a9d98bb9",
+   "id": "dc9c8cdd",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "68ee89bb",
+   "id": "0504640c",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "8b2c897a",
+   "id": "56a6d796",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
index 0b910d8b..a6b1b910 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "3010585d",
+   "id": "37a3a9cc",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "a4f38e83",
+   "id": "79ffcac6",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "4194b848",
+   "id": "796690ce",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
index 1dcdcfcb..b09f227e 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "35d3a5cd",
+   "id": "9de9c8f5",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "bec84a92",
+   "id": "45d6368c",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "3ca5968e",
+   "id": "04ddedc6",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index c4511286..cb56212e 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "327e18ad",
+   "id": "7a0f9441",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "f7b862ea",
+   "id": "cd0a2a27",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "4393af8a",
+   "id": "036896fb",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
index 4f537833..b95176c9 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "0233ba10",
+   "id": "3414e1d9",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "1178a929",
+   "id": "6bdfe683",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "6483cf0f",
+   "id": "5ba07cd8",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
index 8a963a3a..dd02942a 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "45874118",
+   "id": "f68d76ae",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "33f2d3f7",
+   "id": "812a233c",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "f682f2a7",
+   "id": "b6e872c5",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
index a62464f7..e54c2439 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "26116226",
+   "id": "2e67bc84",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "ef3c1f13",
+   "id": "5dc1d8fe",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "5329b3a6",
+   "id": "82a5deff",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 0aeefb1d..564adac9 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "db51da80",
+   "id": "380364da",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "f2060aef",
+   "id": "0531c05a",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "1675c490",
+   "id": "0c7215a3",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 9da05db2..be92ed01 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "788e846c",
+   "id": "ea8ce55b",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "80da7c78",
+   "id": "4225adc1",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "ff0ad674",
+   "id": "5a16161b",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
index 88a116ca..34853a8a 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "5394aa82",
+   "id": "c2739475",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "71276d12",
+   "id": "2b003cf9",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "cce13a9c",
+   "id": "f232b088",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
index 10d6657d..e3caf519 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "d8a00db5",
+   "id": "e0f370fe",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "7523c457",
+   "id": "9e77e76d",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "cc1d3a00",
+   "id": "071525d3",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index f275c125..a90b3892 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "3dc0af97",
+   "id": "15c72ddb",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "fa94c9d8",
+   "id": "a1df3d28",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "edff4b69",
+   "id": "04eda63a",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
index adf5b5b1..7eb03865 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "600c9977",
+   "id": "3a02b72c",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "6a69f8db",
+   "id": "f16c0a3c",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "059edb3c",
+   "id": "fd0a9510",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
index afa4db32..45ea568b 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "2bd00ab9",
+   "id": "cc3e5ffb",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "30a738ba",
+   "id": "bd09291e",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "0634062b",
+   "id": "9d4cebe4",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
index b7b02de1..d7c60e22 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "989f4236",
+   "id": "e0af9bba",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "adb81b04",
+   "id": "9f2d3780",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "6c663b82",
+   "id": "af490c4e",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index ebf0871a..ccba4874 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "c3e81a60",
+   "id": "c47bfddb",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "771fcd3c",
+   "id": "a2a850a7",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "39e5cb4f",
+   "id": "4a2390a9",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 981f27dc..a2490e18 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "d604a0bb",
+   "id": "77d610f2",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "174844a4",
+   "id": "54c226cc",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "1b440ed4",
+   "id": "4eaff11b",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
index 09b85b56..bf0122e4 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "13957913",
+   "id": "ad444e5e",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "98fcad81",
+   "id": "8eac66fd",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "ed69a633",
+   "id": "6b8a6e61",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
index 8cd184ba..910eeafd 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "87acf1ee",
+   "id": "6f6189e4",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "bad47474",
+   "id": "44f424e8",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "137462a5",
+   "id": "431ec0f4",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index 96046d5a..e8ef0903 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "0300951f",
+   "id": "11051f56",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "4ae06948",
+   "id": "c8b69a7f",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "be11003e",
+   "id": "fed107f5",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
index b23e1d9c..2b0d719f 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "a9b08448",
+   "id": "ab365402",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "73123106",
+   "id": "868dcbb2",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "b7a69eb6",
+   "id": "753aa8c6",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
index 42434245..cee2a581 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "47585247",
+   "id": "23620803",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "05c8aee3",
+   "id": "d4831160",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "e5e40e77",
+   "id": "20cb79e8",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
index bf932404..566fc696 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "0d816f06",
+   "id": "65434e49",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "9875e267",
+   "id": "eec7e17b",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "9f87e946",
+   "id": "ac43cc21",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 58f7ce6b..048466ec 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "c42bca27",
+   "id": "f4d3c325",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "95b9e4a6",
+   "id": "26f84cd2",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "9eb78e0c",
+   "id": "db6b0f7a",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
index e64b819c..e280360c 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "e315b4ef",
+   "id": "3b19dd26",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "b2455d0d",
+   "id": "ea0dcf70",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "e5b36cbd",
+   "id": "046fe5ff",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
index b470700a..11ec4022 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "ae64b97e",
+   "id": "095ba0ea",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "abcbab2b",
+   "id": "ff46d89e",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "a94021db",
+   "id": "796e9b7b",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
index de69ef03..b1656db8 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "f6f274e0",
+   "id": "75d27d06",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "347d4cbc",
+   "id": "fd76353d",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "67114b0e",
+   "id": "74864d7f",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index 84d8fa6d..64742a1f 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "5cb1e921",
+   "id": "bc2897c7",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "4d6a1a1c",
+   "id": "8a7d753b",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "ce42ebba",
+   "id": "9a6d73cf",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
index 98eb2a43..aea99724 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "57e670a7",
+   "id": "600cb5ac",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "b4059d03",
+   "id": "3c5c3c33",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "66a17073",
+   "id": "34067a7b",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
index 04d376c9..6c904a5b 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "570353ce",
+   "id": "7c71058c",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "98f81f49",
+   "id": "dae2b4cc",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "47cea410",
+   "id": "e01c7d82",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
index 2c521fad..070b9274 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "e8110c4e",
+   "id": "d14c4524",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "c8cb0b8e",
+   "id": "9a71dcf9",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "dfe0d040",
+   "id": "3fd14d69",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 723bca34..ecc7b9a7 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "c435e522",
+   "id": "97dd9a6c",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "7068eb2f",
+   "id": "29cd3657",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "9e238cfc",
+   "id": "f6843ed1",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 6240cc42..b1689ef6 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "08696c87",
+   "id": "99ec323f",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "e13a5012",
+   "id": "6439d79a",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "920eaff2",
+   "id": "56d94b84",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
index d1dad59c..154b668b 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "7c85d96d",
+   "id": "e8cabb70",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "76023bcd",
+   "id": "1505d288",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "9d79021d",
+   "id": "c006db2c",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
index 296a9528..6c01313a 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "c310e3e1",
+   "id": "49266d05",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "8eaf0100",
+   "id": "d401567e",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "1099b681",
+   "id": "2cefa76a",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index 36812e02..f6794578 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "b70d34a7",
+   "id": "ae80c305",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "0c184ccc",
+   "id": "c464a198",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "60e40cab",
+   "id": "41dc3898",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
index a1aeec20..d58c728a 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "b628a978",
+   "id": "73217c8d",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "aa881488",
+   "id": "854fa799",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "7bed19d3",
+   "id": "8ba4e760",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
index 90bc079b..297d4dc5 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "67bdd875",
+   "id": "f6438653",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "0dce612c",
+   "id": "3d16b343",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "fd6ee9a5",
+   "id": "66c0fe28",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
index 7a563eda..6ee1c5c8 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "9d2d682e",
+   "id": "e8d8fda9",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "0156e89a",
+   "id": "133ab699",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "2077ea9c",
+   "id": "ca2442aa",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 29c5e716..77d1922b 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "b76ead88",
+   "id": "30fdf7d3",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "e6d4cc28",
+   "id": "f3c8434d",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "25ce7208",
+   "id": "fd32bf15",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
index c551555e..53b55ef0 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "4da40591",
+   "id": "3340c350",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "6569c6cd",
+   "id": "403cca7a",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "996651b9",
+   "id": "336319b4",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
index bb2bfca9..29846441 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "068602c9",
+   "id": "48b0b376",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "893bb340",
+   "id": "b8fe2a33",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "c7189a11",
+   "id": "903f677c",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
index 0a4f8be3..dbfbe5ea 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "0f471d16",
+   "id": "761c9264",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "ff460ea3",
+   "id": "2abc4ea3",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "3c678a4e",
+   "id": "5388709f",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index 924731b9..90034df6 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "0349652b",
+   "id": "b6b82cc4",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "d1556ca4",
+   "id": "5cdb7bd2",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "d4ac9555",
+   "id": "6322757a",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
index 131ed1ba..3315b8e4 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "891e5c8c",
+   "id": "6fefad31",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "65f7b10b",
+   "id": "f80439b3",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "12b540df",
+   "id": "155589bf",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
index 1ddd0c26..a105be76 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "5cb43bc8",
+   "id": "b146e78d",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "7eb8e644",
+   "id": "b7bb4395",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "c0dedb34",
+   "id": "69462f93",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
index 05f22f5b..f3e04939 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "998d4909",
+   "id": "264567e8",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "844e8713",
+   "id": "8381dd61",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "8581f6df",
+   "id": "a8a814cd",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 6d2ef876..8d7f8056 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "dea31dcd",
+   "id": "ce623fa5",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "6bd6d08d",
+   "id": "2c6e64d0",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "0d693a16",
+   "id": "3ef5e19e",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 7b8018f2..fa512a72 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "cb928d28",
+   "id": "8ec9ad67",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "bbf0cef7",
+   "id": "56c38d02",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "27a50bba",
+   "id": "b303a8e2",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
index 7562a59b..769ba159 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "f9ae7293",
+   "id": "21a6985c",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "22ba0474",
+   "id": "98398b48",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "2fbcb479",
+   "id": "a6d7198c",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
index 11ac600b..ede98959 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "533c8e7e",
+   "id": "550724b4",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "44d9ddfa",
+   "id": "b20fbd3a",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "b80663b2",
+   "id": "d8e8ab81",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index ffaf5d20..c103d092 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "944c55fe",
+   "id": "84330730",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "ffa9e6e6",
+   "id": "9bd75094",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "890dd7d6",
+   "id": "54435c19",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
index 9a3bc1c7..214370e9 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "dac69dbf",
+   "id": "29cfc3bf",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "dfda3f77",
+   "id": "72db7553",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "42712c33",
+   "id": "3be2fbeb",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
index 028b67e4..381b8ba3 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "3fb04101",
+   "id": "752982a5",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "e7c53750",
+   "id": "aa47e08f",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "2bd340c9",
+   "id": "c652875a",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
index 9531bcf3..6b2272d7 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "598ca63a",
+   "id": "752fb088",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "65568ffe",
+   "id": "55b5bf0e",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "82a2c301",
+   "id": "73a7001b",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 3eb65408..438cd03f 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "dcc19954",
+   "id": "a6c774e2",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "b69169a8",
+   "id": "2fc4e9cc",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "ad1ab17d",
+   "id": "24cc15ee",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
index b20b9c50..82a41d82 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "9877541d",
+   "id": "87ae55ae",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "5235e8b8",
+   "id": "5c8c7a9c",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "724eb6bf",
+   "id": "a2797559",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
index 712254bb..ca918e7a 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "158cbb3b",
+   "id": "6c56dffb",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "551af08a",
+   "id": "e5d5562e",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "7bbac95a",
+   "id": "ce35cee3",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
index b07a3c26..b0903fa7 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "af6c8ef9",
+   "id": "e57cd185",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "c3b3e947",
+   "id": "c5eb8700",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "6a7d7ea2",
+   "id": "65dc0e05",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index 933d8e10..b13f5e4f 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "d9f1ce9a",
+   "id": "ec1944d0",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "d96cfc92",
+   "id": "c1807e44",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "d1f383ca",
+   "id": "7bc7fca2",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
index 95c00a09..38170d1d 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "dbcdbeab",
+   "id": "34733e7b",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "44cbe7c6",
+   "id": "3f711870",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "e3590c90",
+   "id": "49ccf533",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
index 1ab50daa..1ef532f5 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "7ce3c066",
+   "id": "45c50fea",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "255bf5bd",
+   "id": "79491bdd",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "d1e4bda8",
+   "id": "2dd8f575",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
index 1f00e318..b529443b 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "41e1f864",
+   "id": "fa8ee54b",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "dfac55aa",
+   "id": "8df2317a",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "49dc9939",
+   "id": "f0829a49",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 6f9cedb7..7232bce3 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "61c77541",
+   "id": "9b710c07",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "ec768b64",
+   "id": "d033cb55",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "e7030b35",
+   "id": "213e732b",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 4ed6c824..178f818a 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "5dbbd24d",
+   "id": "cd32c5de",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "357ccbf9",
+   "id": "721e9c23",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "40f699c0",
+   "id": "89fa475c",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
index 49c53bcd..5336312b 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "d6a39288",
+   "id": "8a8835f7",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "3ef6c32e",
+   "id": "d07eab83",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "c0b7018d",
+   "id": "ac720356",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
index a7f693e9..6d8d3045 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "64e16158",
+   "id": "8d609d61",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "31617b02",
+   "id": "cad4b3b3",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "1daa293d",
+   "id": "a466877f",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index 559ec2ba..d9877170 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "60a879cd",
+   "id": "ca4c4697",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "e702e7a3",
+   "id": "ec08165d",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "1f254bd6",
+   "id": "8e50df72",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
index 45040a17..ea27d55b 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "0373c5b4",
+   "id": "4ed8c76d",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "f12657f6",
+   "id": "e1c77286",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "10f5f2be",
+   "id": "8f5ccf3e",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
index 975f0e29..5e497172 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "6cc54f45",
+   "id": "4cce95b6",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "af32c137",
+   "id": "d19480e5",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "5206dfcd",
+   "id": "84e42275",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
index b075c5a1..a6a3ea83 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "14479ed9",
+   "id": "56fe39d2",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "eb51e1da",
+   "id": "5d30683c",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "62f54e30",
+   "id": "134dd4f2",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 13e2640a..87e96807 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "24cd6027",
+   "id": "1ced10b4",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "49b290cb",
+   "id": "25b7e6a9",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "bd8ff7bc",
+   "id": "4cc96038",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 07232bb9..86c77edb 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "11587641",
+   "id": "eb5d3da1",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "a1838fca",
+   "id": "9ecbb1f2",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "fecf259f",
+   "id": "5301000e",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
index 6ad5235d..81c3b11c 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "8e437d11",
+   "id": "5e48e31a",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "d023f388",
+   "id": "03024203",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "72773e10",
+   "id": "fe82da3c",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
index 90f43a27..37f2b6ec 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "a03f5c32",
+   "id": "95170fd0",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "d2748d1e",
+   "id": "11970b14",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "238fcb7d",
+   "id": "ac07a214",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index f120f009..adf0fbde 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "94b4fbe5",
+   "id": "b0bce99f",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "59030df6",
+   "id": "8d39d64f",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "37830dc6",
+   "id": "a4ee6ece",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
index 4d701b5d..38a03aff 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "2e2646fb",
+   "id": "22e317f0",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "f571015e",
+   "id": "f0e37b14",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "1b96d4f9",
+   "id": "960a9a46",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
index a7c915da..50019fa6 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "29909a17",
+   "id": "f4347611",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "5f3bd047",
+   "id": "ab1e3836",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "e97d6dc7",
+   "id": "b37bb143",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
index 386f6169..b30b06f9 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "5cb823d2",
+   "id": "e30700c4",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "a545691f",
+   "id": "8f2326bf",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "febd2a01",
+   "id": "11747a00",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index bc5d0be7..4204f8e5 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "7c88b8d0",
+   "id": "47ad8f2f",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "0b08a77b",
+   "id": "28a1a029",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "87cade18",
+   "id": "3e8936d3",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 9133e0f2..452ecc7f 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "e7fa2fd7",
+   "id": "cd7f935c",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "5cc2c872",
+   "id": "464b8db9",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "e65a670e",
+   "id": "4e81b619",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
index 3fcfd8b5..f7f699d1 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "c169adc8",
+   "id": "6ebc6dc0",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "72942122",
+   "id": "033d0101",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "99c5929c",
+   "id": "8c2c10e7",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
index 1dab4651..df492c8d 100644
--- a/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-multi+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "fd6446f6",
+   "id": "d741c0e4",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-multi%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-True%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "0dd3f006",
+   "id": "04d6a712",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "baeb8819",
+   "id": "9ac07c08",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index 076e6ac6..eedb25aa 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "4b3f6e8f",
+   "id": "656f1072",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "4dc3ba60",
+   "id": "fcb00776",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "42929bb2",
+   "id": "b33daf3f",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -59,7 +59,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
index f2238612..5faccb59 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "718dbc1e",
+   "id": "7f52e1a2",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "6051dfe1",
+   "id": "60f669c5",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "b7146cb1",
+   "id": "e4e73973",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -59,7 +59,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 495b97b1..a4e4b6fa 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "7d630ec8",
+   "id": "91e83649",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "fb5d96ca",
+   "id": "e72b8e96",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "eddc41bf",
+   "id": "73a1e840",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -63,7 +63,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 405c6b48..0bf550f7 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "e995d559",
+   "id": "f61b1ba8",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "c12bdd27",
+   "id": "1bf3d685",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "e2b6890b",
+   "id": "4fd3f37e",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -63,7 +63,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index b8ecd780..a1d4b89e 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "8c55de57",
+   "id": "a373d09f",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "be8ead50",
+   "id": "39ebded9",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "955ed08e",
+   "id": "5a212849",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -66,7 +66,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
index c7496d00..dd5d1df0 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "f7080c1e",
+   "id": "73be7b83",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "34efb07d",
+   "id": "128cc246",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "592b7a90",
+   "id": "7f3ff1dd",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -66,7 +66,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 756db031..c947d03c 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "3eb78db9",
+   "id": "d59456e1",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "8bea0f8f",
+   "id": "1ec5bc40",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "339a503a",
+   "id": "c2810636",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -70,7 +70,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 28b5caf3..8c386fdc 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "7ae4dc6a",
+   "id": "43d18839",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "ea2e8cd8",
+   "id": "0bb0621d",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "fa4c090d",
+   "id": "221af001",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -70,7 +70,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index 8fde373f..14a15c70 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "3aba0134",
+   "id": "e1ea2e33",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "d02dbb74",
+   "id": "02ecfcff",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "fb658367",
+   "id": "99e6186c",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -59,7 +59,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
index b57f4eac..9f8e5689 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "e25bb99c",
+   "id": "fcb6213c",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "e80ac178",
+   "id": "96a77cd7",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "9aa2c54c",
+   "id": "787b9ea6",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -59,7 +59,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 23e4ef6a..ad546f04 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "380aeb1e",
+   "id": "0f866f41",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "b97fa28b",
+   "id": "58a68a12",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "c06653d4",
+   "id": "a9720e53",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -63,7 +63,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 4325f0e5..08471e51 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "179f2797",
+   "id": "841dac3e",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "31488f76",
+   "id": "4cc274c8",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "1fcdd863",
+   "id": "7a5942ac",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -63,7 +63,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index a0811ae4..4406f7c2 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "76c18664",
+   "id": "8d96b7f0",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "1811c815",
+   "id": "b469f44c",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "9d4b279c",
+   "id": "9eb86e4b",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -66,7 +66,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
index bd549416..6fa478ca 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "06562c6c",
+   "id": "371ef085",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "f09ae211",
+   "id": "9dd41fb6",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "f25fb953",
+   "id": "0a12efea",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -66,7 +66,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 7f02dd95..40c0b1e0 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "11186098",
+   "id": "fe234a83",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "3c8ec81a",
+   "id": "30f1de8f",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "98413d4c",
+   "id": "4c9c7d7c",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -70,7 +70,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 129a58ea..22175ddb 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "651f427a",
+   "id": "30cbbe04",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "9d106285",
+   "id": "a5d162ea",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "3046df13",
+   "id": "e15c71aa",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -70,7 +70,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index 11463973..109c35be 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "ab1a5c14",
+   "id": "4c8bfc39",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "bff83dfc",
+   "id": "9903d8df",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "7223d74b",
+   "id": "c1adc079",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -59,7 +59,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
index df6e7442..5d35551d 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "c09d9395",
+   "id": "a45cde32",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "68aef8d2",
+   "id": "f30adc40",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "6aa21911",
+   "id": "060b60c3",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -59,7 +59,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index e3011946..acd85600 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "d3b90e2c",
+   "id": "e90e138e",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "dbdcfe81",
+   "id": "9e12876e",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "233df030",
+   "id": "f8c0c28b",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -63,7 +63,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
index e0dbbfdb..c096bfcd 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "e1fb805c",
+   "id": "9386a2a1",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "edce46e3",
+   "id": "30cebc6a",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "9102206e",
+   "id": "bfd1a6dd",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -63,7 +63,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index 6be54ff8..ecc44bdf 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "f0b839a6",
+   "id": "1c4c63bf",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "46069d3b",
+   "id": "39c7790a",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "c5769794",
+   "id": "94a55ebd",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -66,7 +66,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
index 433fd03a..11ad06f3 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "b78b6410",
+   "id": "8cfdf7fb",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "8c827b03",
+   "id": "c41974b0",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "6be33ff6",
+   "id": "9a2ef2b5",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -66,7 +66,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index b0cd5dee..15e6c35c 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "bbd837e7",
+   "id": "a88436bb",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "42bcd51e",
+   "id": "575223e2",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "eccf0c07",
+   "id": "60d8b4d0",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -70,7 +70,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 79ab3e5e..0a47dd73 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "00b396a5",
+   "id": "0a0920f4",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "79dfe306",
+   "id": "9fb9b05f",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "5967823b",
+   "id": "4bbea43b",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -70,7 +70,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index b69e9462..62dbac97 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "8244110a",
+   "id": "c729d664",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "7a5d58ca",
+   "id": "facf8bb0",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "2fa3c8f7",
+   "id": "85c9c7ad",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -59,7 +59,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
index 36b9b079..bfd497bb 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "7892c8e5",
+   "id": "992af99d",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "479852dc",
+   "id": "e0fa3404",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "fdcf0a3d",
+   "id": "a115d940",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -59,7 +59,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 283a2a0f..9e8fcbde 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "0d20bfc5",
+   "id": "b94c2b48",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "a5263b67",
+   "id": "8e1c9d6e",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "13906842",
+   "id": "502e1807",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -63,7 +63,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 28ea6d9a..37646182 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "892f1ee3",
+   "id": "b08669ad",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "53eac230",
+   "id": "abf21d9e",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "3b433b1c",
+   "id": "ab0d9096",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -63,7 +63,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index 8fe0d8f6..530a5b02 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "651e7525",
+   "id": "c5528f3e",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "301492b7",
+   "id": "4df37e68",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "38e12024",
+   "id": "21343f31",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -66,7 +66,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
index 497c2c6b..46f72dfe 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "f5849d64",
+   "id": "cf2b4081",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "5832db2f",
+   "id": "acd25267",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "6c323c23",
+   "id": "bc3a29d4",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -66,7 +66,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index a237af1e..d94d5ae3 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "afe711f4",
+   "id": "c5fd3837",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "41434255",
+   "id": "41e1b758",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "473865a9",
+   "id": "3abeb6c6",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -70,7 +70,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 7d339763..6d5c6af7 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "1e07bf4e",
+   "id": "9c33fa4f",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "26478f48",
+   "id": "b520d1c7",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "b6fa1544",
+   "id": "59bf02b6",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -70,7 +70,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index b8662dfc..70ffd470 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "227b100b",
+   "id": "adaf6e31",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "0fe0ab9d",
+   "id": "43d28377",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "f0b27f6a",
+   "id": "aa7b9d95",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -59,7 +59,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
index 32c9723e..fa07e818 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "40f729d8",
+   "id": "3e5cd6d1",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "dee44321",
+   "id": "3659f006",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "831a11b7",
+   "id": "b0c4df94",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -59,7 +59,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index bc6f008f..5d2464fb 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "6e80e485",
+   "id": "60a6beac",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "0ec7ad94",
+   "id": "8cf69070",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "120b96eb",
+   "id": "226dcbba",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -63,7 +63,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 75fc34b9..b5cd22fc 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "ee98db6d",
+   "id": "e0ad2048",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "1c842e6e",
+   "id": "729e945a",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "7fea5160",
+   "id": "42f0955c",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -63,7 +63,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index 372dbe13..9c8e2842 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "f4766b3d",
+   "id": "a18d887d",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "6022d421",
+   "id": "5613bd4e",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "ed03ea3c",
+   "id": "fcdc26f6",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -66,7 +66,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
index ef2daa0e..1b536381 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "a031fafc",
+   "id": "d7942745",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "65c77b9c",
+   "id": "d73ae82a",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "cf3aa621",
+   "id": "4ec4bb07",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -66,7 +66,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index f6a483ee..bef044a3 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "d9198514",
+   "id": "9218eb72",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "4f2d23bd",
+   "id": "02555c19",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "94c75138",
+   "id": "f9e06089",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -70,7 +70,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
index c9617518..1c825fac 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "2c717957",
+   "id": "38a80ed7",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "220cc77d",
+   "id": "d7f224be",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "58d07577",
+   "id": "a5cd6f18",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -70,7 +70,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index 78a540d2..bc2c69c0 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "069e38f1",
+   "id": "95830ddf",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "f7825fae",
+   "id": "be76e318",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "478b1d9e",
+   "id": "4b29e1a2",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -59,7 +59,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
index 9e933d54..80f18658 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "f5fcb507",
+   "id": "f854d1dc",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "3b324aaa",
+   "id": "e506b69a",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "1c4cdf9d",
+   "id": "955b56bf",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -59,7 +59,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index d54fcf0c..7aa10ca4 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "2a1dba63",
+   "id": "4e09722a",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "e88b0e0a",
+   "id": "fa7fead5",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "75ae7c49",
+   "id": "33af6317",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -63,7 +63,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 5f87ac23..132b184d 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "83373871",
+   "id": "924ba593",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "ccdf8cef",
+   "id": "346b8822",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "e5c6cf87",
+   "id": "51dc6ce6",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -63,7 +63,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index d15fa043..c88de77f 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "64c1681f",
+   "id": "aeacaf5d",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "bbeaeb34",
+   "id": "3d5d324b",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "50cc4648",
+   "id": "88e648d9",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -66,7 +66,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
index 09129693..7d1859b0 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "5e3ddc90",
+   "id": "5d460835",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "ac349f56",
+   "id": "7d874eba",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "407151df",
+   "id": "e036adbb",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -66,7 +66,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 987ea02c..e14e9069 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "97f81925",
+   "id": "537515cd",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "1eb536ef",
+   "id": "ca031c3e",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "19708105",
+   "id": "87da6064",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -70,7 +70,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
index d6810e57..45542e66 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "ce8fe3aa",
+   "id": "72ec38fa",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "7716636c",
+   "id": "5bcd9f0e",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "c0a336df",
+   "id": "d3b55671",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -70,7 +70,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index 4421fbb0..23f4eae0 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "d19bdff0",
+   "id": "6a5cd94f",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "edc48ccc",
+   "id": "13b3a6d7",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "50c2602f",
+   "id": "2b1a1304",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -59,7 +59,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
index 88333e4b..c30416ee 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "52f6f22c",
+   "id": "afc2a6fe",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "5e7dcab9",
+   "id": "d787b5d3",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "2a4bfe55",
+   "id": "9fac59d6",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -59,7 +59,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index f0904cb3..ad36e438 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "6b085a0d",
+   "id": "2f4c43a6",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "4ded30b8",
+   "id": "a5f87d60",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "1459f716",
+   "id": "a5e5c9ed",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -63,7 +63,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 1ef51cb3..4d0eba76 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "eb1f8129",
+   "id": "e4e1874f",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "3064ae2e",
+   "id": "96869ad0",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "4c2587e5",
+   "id": "b52f423e",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -63,7 +63,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index 56f8fdf7..fe19e279 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "4795d778",
+   "id": "a2c1322c",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "d846138e",
+   "id": "e5b37933",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "f8031ff0",
+   "id": "5669e7ca",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -66,7 +66,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
index a5535af3..2f1772c5 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "f8b35298",
+   "id": "ef0a65e3",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "f70bd852",
+   "id": "a042a8bc",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "20252744",
+   "id": "9ea5cf2b",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -66,7 +66,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 43a9b1a5..234e1906 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "d2700d1b",
+   "id": "9ddd2f55",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "d2b8537f",
+   "id": "2576002d",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "a9b1aeac",
+   "id": "b3a4eab7",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -70,7 +70,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 4018752b..672155fd 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "5d4cc9d3",
+   "id": "608fc8f9",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "f1b7ee4a",
+   "id": "b44b63da",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "cae45a98",
+   "id": "9ed48bec",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -70,7 +70,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index b7748acd..6d72fc74 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "594cc9f4",
+   "id": "f5827601",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "5db16dcd",
+   "id": "68cbabec",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "c6929bfb",
+   "id": "144bdf32",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -59,7 +59,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
index 5fdd5afe..30fa4d45 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "7975ca3d",
+   "id": "73ad6488",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "f869f602",
+   "id": "625f7bc5",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "bb6858d2",
+   "id": "e20e2fca",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -59,7 +59,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 48c5d585..0e91cc7e 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "743f52ed",
+   "id": "18a4d7c5",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "feaeb608",
+   "id": "243503e4",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "b15b543e",
+   "id": "188b3989",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -63,7 +63,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 24617b61..941be24c 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "11972be2",
+   "id": "088b33df",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "518b9e7f",
+   "id": "85362455",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "c3a9d78b",
+   "id": "bc2626ff",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -63,7 +63,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index bd13da42..88cdf998 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "3a8c932a",
+   "id": "b8ed34c4",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "7f50cb10",
+   "id": "685b86ae",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "f23bd124",
+   "id": "9150244e",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -66,7 +66,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
index 621144c7..98b38f43 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "52ec298e",
+   "id": "e4fcd980",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "4cc1aa9d",
+   "id": "5e889c43",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "7939b82c",
+   "id": "b8a960c1",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -66,7 +66,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 012f8445..6bae7374 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "50f75454",
+   "id": "e63e6e9a",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "b19e2f06",
+   "id": "26fcfad6",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "842cf44b",
+   "id": "c7022692",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -70,7 +70,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 09e14fdf..01d6714f 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "d0117393",
+   "id": "14dc4f19",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "c35b199d",
+   "id": "200aa4ec",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "ae9f1968",
+   "id": "26f4793a",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -70,7 +70,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index d34bc37c..f2e9809c 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "90e396d5",
+   "id": "8dc11ba9",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "700d126c",
+   "id": "99302155",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "586d068e",
+   "id": "f4fdb795",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -80,7 +80,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
index 5ebd33df..d96995d0 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "438df5fb",
+   "id": "f1bedbc1",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "36fe8460",
+   "id": "7b7f476a",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "e48ea97e",
+   "id": "1655cb1a",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -80,7 +80,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index b7fbac2f..f7ca192f 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "0f342705",
+   "id": "60c95504",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "d940e065",
+   "id": "4889d846",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "91aa5e41",
+   "id": "e45c937d",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -84,7 +84,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 6b4ee3a0..99910338 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "a52b6dcd",
+   "id": "578f5b8b",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "bf3f94e8",
+   "id": "e34dc16e",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "0c122d8e",
+   "id": "4a3420ba",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -84,7 +84,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index ecefacd7..02bb1e1c 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "22aad172",
+   "id": "4c4f02f9",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "1ae77563",
+   "id": "e0bee551",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "95ec04d6",
+   "id": "6b0a5dec",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -89,7 +89,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
index 1a3940a1..cef90daa 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "3781f0c8",
+   "id": "850f1d89",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "661563e6",
+   "id": "fb7890e1",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "05f61ff0",
+   "id": "25677af5",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -89,7 +89,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index c00d61cf..b571446b 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "1acaa045",
+   "id": "7ea0b9fb",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "2f78c2c0",
+   "id": "d23523f5",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "d3d91c20",
+   "id": "ee964fde",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -95,7 +95,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 478b70d4..38e14930 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "9f64c7ec",
+   "id": "ed51206c",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "ce74e3b5",
+   "id": "6fa6a81f",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "71111edb",
+   "id": "51f5fcba",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -95,7 +95,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index d0fbc264..43e103c7 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "9ba2ce5a",
+   "id": "e8bc56b9",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "fa13a13d",
+   "id": "2d67dfb0",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "ac8b3140",
+   "id": "6e124989",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -80,7 +80,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
index 32611750..7e471897 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "629ae703",
+   "id": "059c6c10",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "a60f2ec3",
+   "id": "8edc5ade",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "a86a1f35",
+   "id": "4da773d6",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -80,7 +80,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 20a55154..bc047c64 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "068a020a",
+   "id": "9a25ddba",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "913e225b",
+   "id": "98a7cb41",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "75681f2f",
+   "id": "1b1679f0",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -84,7 +84,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
index a7ccd89f..7ce12cda 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "151b97b4",
+   "id": "796c254f",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "d6181625",
+   "id": "4683b346",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "83c63a92",
+   "id": "1ab0f31a",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -84,7 +84,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index d887dc69..b5a210f3 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "a2967fa9",
+   "id": "f2902d33",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "8605ff92",
+   "id": "72bbecce",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "1ef1997a",
+   "id": "48419d93",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -89,7 +89,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
index cad095c8..2d638382 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "5e7e148b",
+   "id": "f3b43772",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "5a239bb0",
+   "id": "0e653b17",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "c39c60c6",
+   "id": "843db72e",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -89,7 +89,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 6afddd70..b5d0cea8 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "6495b6ba",
+   "id": "59022291",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "0f5445a8",
+   "id": "342116a9",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "683cf8fd",
+   "id": "98069853",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -95,7 +95,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
index e95d49a0..b3b6c8dc 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "0cd2d8a3",
+   "id": "2254dcad",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "06c4a8e8",
+   "id": "9129b805",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "59c8a5d6",
+   "id": "0722e6a2",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -95,7 +95,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index ab44500c..c731dd70 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "0928221a",
+   "id": "fe461893",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "8e046eab",
+   "id": "dfee675d",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "5052d994",
+   "id": "824838eb",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -83,7 +83,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
index 4eb25c06..81a52ecf 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "15dcea86",
+   "id": "95523d99",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "e266ac0b",
+   "id": "3d0d641d",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "311bac1e",
+   "id": "52d95482",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -83,7 +83,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 559dcb60..c45c904c 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "ccc98508",
+   "id": "4d495bb2",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "7c7274ac",
+   "id": "df77cae3",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "515d05d1",
+   "id": "4634c4a9",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -87,7 +87,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
index dfb1d635..dfe8eadc 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "8b0bf999",
+   "id": "e6e79330",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "095c27c7",
+   "id": "de1993bc",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "a310828e",
+   "id": "d0dc306f",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -87,7 +87,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index 33d51903..92590015 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "59ae51bf",
+   "id": "09e2e6d7",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "6b673cad",
+   "id": "6fc52b17",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "f0e81d09",
+   "id": "f43738c2",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -92,7 +92,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
index 326df445..a08566df 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "38accbc2",
+   "id": "c25e24df",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "c6b6dbfe",
+   "id": "c21c44cf",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "3bfd3eb7",
+   "id": "3d51e79f",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -92,7 +92,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 6fd463e2..803311a1 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "fffec6a9",
+   "id": "3b3e23b5",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "0d4b6ca1",
+   "id": "3f037490",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "27eacc70",
+   "id": "39fa0926",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -98,7 +98,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
index dd6c9814..6555ed47 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "2ea5bd08",
+   "id": "b09b8500",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "9b79b36b",
+   "id": "17b086ae",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "5620fdb5",
+   "id": "8a9d5dc0",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -98,7 +98,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index ad747a1c..6624b1ac 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "14fd5c6e",
+   "id": "912ac3f2",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "d0564389",
+   "id": "6a977cd6",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "e6f6cc37",
+   "id": "b331cf66",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -83,7 +83,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
index a0264bcb..eca37478 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "f48ba1aa",
+   "id": "43c967f1",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "aa81b0c0",
+   "id": "4e62f3dc",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "21fa71e0",
+   "id": "055037d0",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -83,7 +83,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 8a68e762..5219caab 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "a1694322",
+   "id": "de72bc7e",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "8f812a9a",
+   "id": "e5e70fbd",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "a3e48136",
+   "id": "d51184a9",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -87,7 +87,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 1718b3d2..cca5991d 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "53722b86",
+   "id": "1acdad0f",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "a4aafcfe",
+   "id": "f70eba1f",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "03491ebb",
+   "id": "4d0c6192",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -87,7 +87,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index 11c38b91..8ef99156 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "ee23e34d",
+   "id": "becb778b",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "4295c913",
+   "id": "52dd096d",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "a89a9188",
+   "id": "d03e81ba",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -92,7 +92,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
index a81b077d..de37ec4e 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "e711d33e",
+   "id": "513ef722",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "c7ea2543",
+   "id": "1a866459",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "cb20e0f1",
+   "id": "f8c4a68c",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -92,7 +92,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index a5ec13a7..53f3aee0 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "3c3a0c88",
+   "id": "916bfd12",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "5f196445",
+   "id": "3aa445b5",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "d9a7904b",
+   "id": "d520174c",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -98,7 +98,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
index f48292ea..ce52b279 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "fa970acb",
+   "id": "e93db9ab",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "1a697d2b",
+   "id": "4f59e4d4",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "bb780d64",
+   "id": "5314ff4d",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -98,7 +98,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index 4e259d1f..b0c7b87a 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "b617d838",
+   "id": "efbab614",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "68c69f97",
+   "id": "6f14eeb1",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "ed145b0f",
+   "id": "26e21954",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -80,7 +80,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
index e98a3877..32d71abe 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "843becdf",
+   "id": "f99053a6",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "ed3d25fd",
+   "id": "46a03eb1",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "cdd6b121",
+   "id": "92c83b23",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -80,7 +80,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index a8a47682..2d80ab6c 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "4b5a0014",
+   "id": "ef48925c",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "64c95396",
+   "id": "64001c93",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "112b92d2",
+   "id": "ffbf32c8",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -84,7 +84,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 3c98a55b..59a01c01 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "2c6ef65f",
+   "id": "bd1f95be",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "4ae7346c",
+   "id": "af55be60",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "a16b2a5e",
+   "id": "ad2e9e83",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -84,7 +84,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index abad2c8c..8c3f5bd1 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "eea36a94",
+   "id": "16cc946f",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "c0976f97",
+   "id": "e0b05bb0",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "301a34d8",
+   "id": "e1cebe63",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -89,7 +89,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
index 46729728..befcd010 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "dd0e25b8",
+   "id": "b811e971",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "2b8ad56b",
+   "id": "ab291ebc",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "cc8b5ba8",
+   "id": "af1e682a",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -89,7 +89,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index a945be5f..4b66efef 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "420c5ab6",
+   "id": "de8c2b1a",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "554c60f5",
+   "id": "525868d0",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "8e7cf8d1",
+   "id": "f31d65c2",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -95,7 +95,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
index ad1347ae..a7db886a 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "ee3c21a9",
+   "id": "5c9d6791",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "46200e8d",
+   "id": "8961c803",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "13aeaa60",
+   "id": "b09b3935",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -95,7 +95,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index d6f1f927..90dd527f 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "69e2abd7",
+   "id": "88e918a5",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "64a32902",
+   "id": "95b8a7a7",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "2bee8ea2",
+   "id": "ac615109",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -80,7 +80,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
index f1defeeb..1aa0f4b4 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "6b3b1739",
+   "id": "a034995a",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "2a0c9fb5",
+   "id": "bc52a089",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "6a17e8d1",
+   "id": "3ede4be8",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -80,7 +80,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 0388feff..b3139b93 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "8f8942b7",
+   "id": "2a78019d",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "c05e03e3",
+   "id": "db87e021",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "e1b8b61c",
+   "id": "fcebf4d3",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -84,7 +84,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 465c6c62..5345c3f6 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "e7c9b646",
+   "id": "ff951004",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "4661a0f3",
+   "id": "dc33dd91",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "1f6ed3d3",
+   "id": "34d47d5f",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -84,7 +84,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index 1bf9f9bc..3d611b74 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "87677053",
+   "id": "3342c8c0",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "5f1ad9a1",
+   "id": "aa64f556",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "0de40a4c",
+   "id": "5deee04e",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -89,7 +89,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
index f41cda69..372f182d 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "800b39f7",
+   "id": "c8812735",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "d2f19d61",
+   "id": "4757e8e5",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "9eb47ad2",
+   "id": "2bd3cce1",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -89,7 +89,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 247a19c8..42727ce2 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "65a8e0d1",
+   "id": "c72888fe",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "c1694fcf",
+   "id": "d7865e84",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "0b89f4be",
+   "id": "d3a6ba15",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -95,7 +95,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 50d7633f..729c88c4 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "560e0049",
+   "id": "e380ef8f",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "2877a052",
+   "id": "56597004",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "ce60f673",
+   "id": "2df49343",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -95,7 +95,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index 2e285e06..b3d131eb 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "e976e49f",
+   "id": "5092a172",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "0271f5ab",
+   "id": "b09f11fb",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "7e8b8a2c",
+   "id": "228f578a",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -83,7 +83,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
index 0d7e9b7c..f9b38639 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "3019153e",
+   "id": "f2fe81ef",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "e1824dec",
+   "id": "f9235937",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "dab8fbba",
+   "id": "04dff635",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -83,7 +83,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 2253fbad..df459639 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "2628b759",
+   "id": "2be65e3b",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "72464c09",
+   "id": "3999e068",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "0ea44623",
+   "id": "6094804c",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -87,7 +87,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
index e60342f2..7f30a8ee 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "d29fdf04",
+   "id": "28e220be",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "586b1f94",
+   "id": "a29c9fa3",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "6117d861",
+   "id": "9ad922df",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -87,7 +87,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index c5ab4606..b00a1e7b 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "2b928021",
+   "id": "72b446ef",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "02832f01",
+   "id": "cb0cad4d",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "18feb154",
+   "id": "af76c616",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -92,7 +92,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
index 2d939653..d26d56bf 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "c52c76d2",
+   "id": "399b98ed",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "b36d08b0",
+   "id": "485c5883",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "7fa258dc",
+   "id": "870e7891",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -92,7 +92,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 7f6bbdf4..02eaae19 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "68fb1df6",
+   "id": "adb79304",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "6b29f706",
+   "id": "7bc6f8dd",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "4d5c948f",
+   "id": "26afdb27",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -98,7 +98,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
index c0200048..ae37f752 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "eedf3c44",
+   "id": "d3170c27",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "0cfbf674",
+   "id": "a371d162",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "8eee9d53",
+   "id": "e72ca60b",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -98,7 +98,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index c642695a..63dc8a1b 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "e12ca18e",
+   "id": "3dfcd3de",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "d13c6da8",
+   "id": "80dc0c0f",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "6f75831d",
+   "id": "7b989412",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -83,7 +83,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
index b4d8ce75..84313d1a 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "ef9dd33b",
+   "id": "7720659c",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "de57ed96",
+   "id": "79d365e6",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "2e8de985",
+   "id": "7d2b45a1",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -83,7 +83,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 89f2e4a7..e6a46bbc 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "ab931b99",
+   "id": "2146300f",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "e400100c",
+   "id": "62dcb373",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "948402b3",
+   "id": "89b4f474",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -87,7 +87,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
index f3e22a8e..6dd53050 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "22d8343b",
+   "id": "d4559ff8",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "5be7832b",
+   "id": "7122b6df",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "3b482a77",
+   "id": "89e1f38e",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -87,7 +87,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index 8eed46f4..dcee0490 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "48004815",
+   "id": "ac3167f6",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "03403c8a",
+   "id": "3ed4864a",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "3d4c0982",
+   "id": "cdeacbc7",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -92,7 +92,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
index a9fd8b1b..6dfefbce 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "5ace816d",
+   "id": "b50ce1b0",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "eff6c87e",
+   "id": "31fde227",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "6764dd26",
+   "id": "f63ad307",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -92,7 +92,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 5481dbc2..dd9dd162 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "4a30748b",
+   "id": "e32e3f77",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "b1996f30",
+   "id": "251b6e84",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "6e521970",
+   "id": "06ffe4b1",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -98,7 +98,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 44d56e95..cd033323 100644
--- a/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "3da61997",
+   "id": "e6404dbd",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-FULLYBAYESIAN%2Bcustom_gen-True%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "eb01f69d",
+   "id": "904ec745",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "4edf3e96",
+   "id": "b174ad9f",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -98,7 +98,7 @@
     "            model=Models.FULLYBAYESIAN,\n",
     "            num_trials=-1,\n",
     "            max_parallelism=3,\n",
-    "            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n",
+    "            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n",
     "        ),\n",
     "    ]\n",
     ")\n",
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index 00ee7164..92573111 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "3d3ab515",
+   "id": "443f08ae",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "4f3e829b",
+   "id": "d10f183b",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "bfa3697e",
+   "id": "098bf317",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
index d4348d82..33194346 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "af1b785e",
+   "id": "659b4cc1",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "f6211b6a",
+   "id": "43933cc9",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "ab0d4a28",
+   "id": "cd3be30c",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index d4abb9a7..d6d32d5f 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "cf9dcc0c",
+   "id": "2dab0261",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "46b5994c",
+   "id": "10ccf6e5",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "b5677ecc",
+   "id": "c40c7c10",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
index f0fa07e4..f70e8492 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "43726ea4",
+   "id": "48684d87",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "80ca5eb5",
+   "id": "3a0ffd78",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "6891611f",
+   "id": "1ae4fee6",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index 859e9278..a604f424 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "bbd704c5",
+   "id": "081f9e04",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "ccec9f33",
+   "id": "384da025",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "31037ff7",
+   "id": "2757a13a",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
index db0b2b55..0b3a704c 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "36e75100",
+   "id": "3a7f5079",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "21eb8fa8",
+   "id": "78392ff4",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "d74e213d",
+   "id": "30361c83",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index bb109426..b9e307a4 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "2c57560c",
+   "id": "3aad3497",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "581484d7",
+   "id": "5920ac9a",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "596fbb7a",
+   "id": "2f7e30bb",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
index e36bc61d..24d2ba0c 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "e66405d0",
+   "id": "a6bd7754",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "cbfbe9b3",
+   "id": "814c1fa1",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "323a8a08",
+   "id": "d9221240",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index 99fd3fd8..e2793000 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "c9a717a7",
+   "id": "77211e4e",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "eb0d8391",
+   "id": "42cea336",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "aed2aee7",
+   "id": "5e3f7920",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
index 649e28cd..1b77e6d5 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "ad9dc1ea",
+   "id": "9940e9d8",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "9e08adf1",
+   "id": "9d49b32f",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "4190f817",
+   "id": "952cd52c",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 0247a28c..b1eb4a25 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "95b219ec",
+   "id": "b245f6e8",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "0d9b91a6",
+   "id": "7791ae60",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "3f373d51",
+   "id": "538e01af",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 22e0722a..300f8e2a 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "8ef66213",
+   "id": "9de0ed68",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "f0f08a5e",
+   "id": "f1ffb36c",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "b37e1bc6",
+   "id": "c50d0d57",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index a97e8675..c68ff516 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "5e6c6153",
+   "id": "de6bbd2e",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "4478ca45",
+   "id": "7aeff10e",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "9c47045f",
+   "id": "f40d17d1",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
index 0bdc0a0d..5344322a 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "cae96c95",
+   "id": "fafda701",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "921d8021",
+   "id": "96066a82",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "4437f19e",
+   "id": "d0b05b56",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index fb47ee60..6aeee33f 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "e9c5f5b9",
+   "id": "37763105",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "c45f363c",
+   "id": "97e4c79b",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "becd589c",
+   "id": "192aae5a",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 6ba6b5f8..e7a18a0c 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "f88ab633",
+   "id": "55c4a93b",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "cc80fdaa",
+   "id": "907d2086",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "0f5d825a",
+   "id": "90f1a571",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index 3b71c902..cfe54a88 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "61e184b1",
+   "id": "b0cf9e1b",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "e995f45b",
+   "id": "f0239818",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "8e86e0c6",
+   "id": "81b0244b",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
index d88ee686..2fc956fd 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "f3bd36c7",
+   "id": "a36812d2",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "d4c5b0ef",
+   "id": "ddf3f5fd",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "5e1ed948",
+   "id": "1dd05cc1",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index b7edc14a..c3a47896 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "f271a78c",
+   "id": "a591cf1c",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "fe4f28c4",
+   "id": "9038e507",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "03028b6f",
+   "id": "f845a908",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 7f7a7ed6..430d5644 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "f04b7d62",
+   "id": "f8a046d4",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "09733f9a",
+   "id": "de7e84c2",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "9f478e12",
+   "id": "809c79c6",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index 022f9c1a..234f4417 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "64e5432c",
+   "id": "a6041e58",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "4c4df3a3",
+   "id": "42d8fb1a",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "0238db5c",
+   "id": "2dc65568",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
index 8cda59ee..2a92a1fd 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "76ba7604",
+   "id": "ce63fdde",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "6c50d6fc",
+   "id": "d85d1a03",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "0f136d04",
+   "id": "a37d733c",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index b81942e4..c8e3872b 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "1ada45bf",
+   "id": "972a7603",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "02f02cc7",
+   "id": "bcf8bcc1",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "94eeab29",
+   "id": "3292530e",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 555aad93..ade41370 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "ed303e01",
+   "id": "0e996da1",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "953eb4cd",
+   "id": "158b1edf",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "1bf6959b",
+   "id": "4770af10",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index fed50ad1..504cde86 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "eceb4460",
+   "id": "aacb1db0",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "0d5f1cce",
+   "id": "d8f2534d",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "ffbad0aa",
+   "id": "d67980e9",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
index 789d2972..55363193 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "8b9bc456",
+   "id": "ee9aeedd",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "5b5313d1",
+   "id": "84aac07d",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "148bfc12",
+   "id": "6512935d",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index c4330cfc..a56a0e5f 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "398171be",
+   "id": "b22d58ac",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "2aaa2bf0",
+   "id": "c586dd95",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "ece5785c",
+   "id": "c816d378",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 0be198dc..98de299f 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "9e8fbc84",
+   "id": "07f4a6fd",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "ebf3eda7",
+   "id": "6825bf46",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "9e8d13d6",
+   "id": "91e9b3f1",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index c52dc569..d934fdec 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "c5ed585e",
+   "id": "555575b8",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "6757099a",
+   "id": "f4e29924",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "086a1629",
+   "id": "cf34bb1f",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
index fccf917b..b1bea06c 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "8965436b",
+   "id": "f8544ff3",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "7ce5d735",
+   "id": "943d4827",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "dab54ba2",
+   "id": "4ab35ad4",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 08962260..777b29b3 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "e13929a5",
+   "id": "f86b41ac",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "c3e9383c",
+   "id": "e2d99221",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "5671fc9d",
+   "id": "566c66cd",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
index bd216166..ec53a020 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "276309ee",
+   "id": "2af62fce",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "417b5d52",
+   "id": "6714e06e",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "bc063ce7",
+   "id": "eca0db77",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index 727d691d..cd41425b 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "1b6282a7",
+   "id": "0f9473a8",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "1f5bf807",
+   "id": "4b72ec69",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "799e5f36",
+   "id": "bc42b266",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
index e7dfadb0..eaeb26bc 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "fc5cf2b2",
+   "id": "0e0c2b04",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "4f1682f3",
+   "id": "bbc75744",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "24512784",
+   "id": "662e87bd",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 8ac99b23..321f3c8e 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "29cc4dd5",
+   "id": "0d9da312",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "8308e317",
+   "id": "6c8477c0",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "cc62d86a",
+   "id": "ecc62290",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
index b4e5573c..86d302b1 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "4a6482ae",
+   "id": "f4f1dbbd",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "b12805cc",
+   "id": "d9994f5b",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "200d7eb7",
+   "id": "69b641ce",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index 7a021e34..4f2150e8 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "b8ec7493",
+   "id": "5aa84af9",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "d1315b99",
+   "id": "aa99d9d9",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "faddc0cd",
+   "id": "49a8d787",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
index 4ca2921b..aa659d9f 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "ac8c581f",
+   "id": "bcc42b42",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "a0099c8d",
+   "id": "0e296c4a",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "296f27ac",
+   "id": "e8d8020f",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 2699557f..1003cb4a 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "719fce71",
+   "id": "3901b8e6",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "fcccd107",
+   "id": "09cfed4d",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "3e29612e",
+   "id": "ce0d7134",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 9180bb00..fed9f8e4 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "cb6bd6bd",
+   "id": "893451d3",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "099ad36b",
+   "id": "67f8df8d",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "f0be45a5",
+   "id": "0ac6295b",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index 27b27c8a..4efeb1e6 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "43b3aad9",
+   "id": "a23e0b2b",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "08903688",
+   "id": "7e0d80ec",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "f3bce120",
+   "id": "59bec292",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
index a8bb4b32..7ad15b7b 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "9f293737",
+   "id": "f06d211d",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "802f91a1",
+   "id": "4e416676",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "0b89cb29",
+   "id": "4dc97e45",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 2e19ca48..2eff9680 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "2b06ed31",
+   "id": "dbacb770",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "3a12d5da",
+   "id": "b1afd744",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "4f5f8bc5",
+   "id": "1f40ae15",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
index b7a95cd5..86cebcd1 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "3d662188",
+   "id": "d7066416",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "e4d2c5be",
+   "id": "57aec7b3",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "d8d368d0",
+   "id": "df6df64c",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index 991e79ae..6a70f361 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "4fd1fde2",
+   "id": "05e19a2f",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "c4dcb131",
+   "id": "94bc0382",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "61a40a2a",
+   "id": "b4de8302",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
index 22358bce..503790f9 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "778efbf1",
+   "id": "3f7be91b",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "3a09b581",
+   "id": "4e2357ac",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "ab50e236",
+   "id": "077bdb6f",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 06694798..68452a27 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "515b6f5b",
+   "id": "1a8e8cd6",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "ebdd4e76",
+   "id": "a4f48b5d",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "e83544b4",
+   "id": "aeddee9f",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 259fe18f..1fdd4fbb 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "033fc592",
+   "id": "06a86b76",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "64a405de",
+   "id": "150563d8",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "fed38aaf",
+   "id": "74599c12",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index 271a8e7b..690717e9 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "1f67c902",
+   "id": "1a44371e",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "d95d02e5",
+   "id": "7417c2f7",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "52385ca4",
+   "id": "4d30f572",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
index 9e2475a9..e2b95a4d 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "8b3ec886",
+   "id": "0d15459a",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "21aa469f",
+   "id": "9ca6cd8a",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "959eb12e",
+   "id": "496f6f5e",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 4efaccfe..770b9822 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "b1c2cfb0",
+   "id": "5c97c415",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "02348362",
+   "id": "90eae12d",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "8cfbf734",
+   "id": "dbb7f36a",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 38e224ea..18debacc 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "05b5b2c1",
+   "id": "f43c9338",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "ff5d9ca1",
+   "id": "5802fc2a",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "ba19f5f5",
+   "id": "f85b4049",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index ebc203aa..28fb16dd 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "8c301a68",
+   "id": "13a6332e",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "5db973d7",
+   "id": "b1d39080",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "da2875c8",
+   "id": "abf5db44",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
index 509e51a0..88eb14bc 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "a98bf202",
+   "id": "d6dc0574",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "1de9d8e2",
+   "id": "b5d59e47",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "61b058fa",
+   "id": "3325b536",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index f95a5fd5..5f019883 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "0db35aca",
+   "id": "e5fec0c8",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "5eb454d0",
+   "id": "dea4a758",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "c12327c6",
+   "id": "9ec85924",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 95ce1dc7..497ba59d 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "3628abd3",
+   "id": "92a204a3",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "8888757d",
+   "id": "338ca75d",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "b687754b",
+   "id": "e8070056",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index 713491be..9c640584 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "3778a60f",
+   "id": "4bd0fd83",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "aadf515a",
+   "id": "c9b7af7d",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "f381c012",
+   "id": "d601b17c",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
index b3e56d0a..211bb73b 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "156e6fc1",
+   "id": "ca15409b",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "f5e966e9",
+   "id": "e353e794",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "5ca456b7",
+   "id": "7a6bf0b5",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index ce3e3161..31a42b65 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "abfe5e90",
+   "id": "221cd960",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "3342dad1",
+   "id": "92046d50",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "117452cb",
+   "id": "db707920",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
index ca9ab4b4..7e1d2861 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "f3543655",
+   "id": "1f1e13cb",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "242c5030",
+   "id": "9ba869e9",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "dbd74611",
+   "id": "794fed68",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index fdcb3735..f1e17cf0 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "a3318953",
+   "id": "579b8848",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "c303418a",
+   "id": "2eb23276",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "3af30fe1",
+   "id": "a69f46eb",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
index 705ac642..0a3f22be 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "a2ed0198",
+   "id": "6a91d710",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "e6bc5549",
+   "id": "cb515045",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "8f693d30",
+   "id": "3b01d859",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index bd53e0d1..f0252d75 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "3f140c8d",
+   "id": "b70dfbfc",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "39c8d96d",
+   "id": "1adc08ab",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "caa66845",
+   "id": "0b33cb9e",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 56f135c3..c147af07 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "de5b4d5b",
+   "id": "b9428959",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-False%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "15103e49",
+   "id": "5f96c5a3",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "6a77cc30",
+   "id": "1c1a1d79",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index dbbcd3cd..2a1b8cd3 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "7eee7fc7",
+   "id": "1638fb38",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "d6cd2db2",
+   "id": "919a9a5c",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "91e7ef9d",
+   "id": "7be379e4",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
index 9dcd3662..9ed6d97c 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "1de3f678",
+   "id": "752d216c",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "682a1d2d",
+   "id": "ab2e219d",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "3d10e609",
+   "id": "702df922",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index b2757a59..58f47b4d 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "f133a157",
+   "id": "5be5e75b",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "c02adc64",
+   "id": "2de2c552",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "2541063a",
+   "id": "feac3a3e",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 1b67f91d..feed25d5 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "ce542f04",
+   "id": "1e31e0ed",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "3b0ca93b",
+   "id": "2b1332d2",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "53181f9d",
+   "id": "856ea79f",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index 5297e8a3..a06ff2ae 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "a4783cb8",
+   "id": "20357361",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "147782f0",
+   "id": "50469701",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "d6413936",
+   "id": "58901f11",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
index 657d377d..227c3395 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "41b86f08",
+   "id": "0a742498",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "3db24574",
+   "id": "fe9f64e9",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "5797cadc",
+   "id": "cb96ef01",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index edc9c4dd..9e09e428 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "67112956",
+   "id": "dafbbf3d",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "7f6e10a3",
+   "id": "0f8e3863",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "8e952375",
+   "id": "7ef01699",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 9d030671..fa43cd95 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "0e0ccdf3",
+   "id": "a004b90e",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "2b495251",
+   "id": "36a0e399",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "b731eab8",
+   "id": "df13b4ca",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index 1fee8586..b73be9a3 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "117203f0",
+   "id": "6404ca50",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "e8a6d7db",
+   "id": "e2e92475",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "4b36c157",
+   "id": "161fcd13",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
index 921d82ce..39c25740 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "26676cf2",
+   "id": "d4555659",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "a2a2ac5d",
+   "id": "997750c3",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "04af8cd2",
+   "id": "6475a51b",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index f671a0fd..130a0105 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "78ea5669",
+   "id": "399dc283",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "b2371338",
+   "id": "888005e8",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "4a11ad44",
+   "id": "da8ce11d",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
index f46c3801..0f311279 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "0d499967",
+   "id": "3e213e33",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "042426f1",
+   "id": "2ac22e39",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "34aa6335",
+   "id": "efff9968",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index 1ffea8d8..add19061 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "e829f5a6",
+   "id": "6010f937",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "037251c3",
+   "id": "2a15cef4",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "2e58c9fa",
+   "id": "b0f65a89",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
index bf5a87dc..503288c2 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "8ba5daea",
+   "id": "bc0afac5",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "d0459ee1",
+   "id": "a12424ea",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "5059839e",
+   "id": "c145cf3f",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 5849dcab..70529c08 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "68424b0b",
+   "id": "a7702e6d",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "169139db",
+   "id": "327bc0ff",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "e51071e1",
+   "id": "539a397e",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 880e1fe8..31e0a119 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "8f1e15fd",
+   "id": "39d61114",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "7e2313b0",
+   "id": "912a0fb4",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "2c52d6a8",
+   "id": "7126b36d",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index d76e1a0e..02e13a54 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "0fea4d0c",
+   "id": "0bd45f24",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "1891d6f1",
+   "id": "8d81c3cc",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "5cdc2caa",
+   "id": "d8bdbc65",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
index a0bc9e96..6b1b0d63 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "3c8df398",
+   "id": "e157e893",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "59929309",
+   "id": "04b9557f",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "9054fa26",
+   "id": "e12f5263",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 1b72ade9..6229d08d 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "6e40a441",
+   "id": "46dc6150",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "80f06c7e",
+   "id": "e51bc163",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "63ea958b",
+   "id": "1982c43d",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 810913a4..ff29ae3b 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "87cbc170",
+   "id": "45067928",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "e26a98e7",
+   "id": "bfa71a7b",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "f486be62",
+   "id": "8b5d9f6f",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index 0f124c41..f1fba757 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "a69e7833",
+   "id": "51f3b754",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "481e7274",
+   "id": "d3dd13ce",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "0471b2bf",
+   "id": "56bf5f43",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
index f5f1751a..d5024e61 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "5522cff3",
+   "id": "55c11ec8",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "7af3cf24",
+   "id": "aa4f4577",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "8d12cd97",
+   "id": "e906e071",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index f3041cd8..d0b3eb89 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "ad242685",
+   "id": "92a95c67",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "9b670ead",
+   "id": "f29b76dc",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "f00addcc",
+   "id": "d860f231",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 777bd21b..292ec6fd 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "306fd83f",
+   "id": "435ae999",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "42984953",
+   "id": "fd659a7f",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "43b3e794",
+   "id": "74ec31a5",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index dcfeef9d..8afd4474 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "71d4a62f",
+   "id": "0fc2b97a",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "42b2fdc1",
+   "id": "03b8a2c0",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "771c7828",
+   "id": "a4bf5d64",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
index 29aec9b9..fccd1153 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "8f42d35d",
+   "id": "2642de47",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "a9198a76",
+   "id": "53a8b3e3",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "72a208fb",
+   "id": "609d9783",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 13062f55..25797e1f 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "2280d19a",
+   "id": "bfea2047",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "11ccf1dc",
+   "id": "ae0a1b02",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "ff3e337b",
+   "id": "18010ae1",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
index cc47a79f..53e7df8c 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "73143f03",
+   "id": "65c84a9f",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "9babcb74",
+   "id": "76629a86",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "f1f79f9d",
+   "id": "95097950",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index 8cfab344..f3f03a34 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "1bde9bc1",
+   "id": "0bfa6058",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "4cd00087",
+   "id": "c589bce5",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "24725845",
+   "id": "786ad805",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
index 46102c59..65d8eb65 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "caf55ada",
+   "id": "1b8c2b4e",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "bd4a3827",
+   "id": "4f17eb75",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "b857c2a6",
+   "id": "e654a384",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 17f405f1..088619c7 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "27e74276",
+   "id": "49658131",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "d0637693",
+   "id": "f040f4b8",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "418de8ef",
+   "id": "75e2d16f",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 3a4b5e3e..3f4ab371 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "9e8460de",
+   "id": "ca778359",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-False%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "6e1240f4",
+   "id": "33341b90",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "e6094c77",
+   "id": "aaabca88",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index 77b1ab38..778d6339 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "996e03e2",
+   "id": "24ba2117",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "4f58e044",
+   "id": "e0026858",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "7712c726",
+   "id": "4e5fb61d",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
index 52c18e44..d5da28b6 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "4eb05bad",
+   "id": "e7fa2c19",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "5ab3a4eb",
+   "id": "c2ff23a9",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "4e3d118f",
+   "id": "7a3e1356",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 65d0dd7d..b280e59b 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "7758bc49",
+   "id": "c8459af9",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "30131a67",
+   "id": "884a8f83",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "763c9d24",
+   "id": "0e0e01c9",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
index c74aa547..314de215 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "3127e8be",
+   "id": "aebc00d5",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "b2f74ed3",
+   "id": "4874f8e3",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "96bfcd40",
+   "id": "b1250829",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index fba114d5..d6aecf0a 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "8b8d993d",
+   "id": "5e70e23c",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "8f8531f2",
+   "id": "99d0bd0a",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "eecf0771",
+   "id": "e0046486",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
index e7f42d46..c4ee0b73 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "815faff8",
+   "id": "c90ceaa2",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "8f35fffd",
+   "id": "c8525574",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "8e981c23",
+   "id": "45221d3d",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index fba08cc8..d18bef92 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "c72bbbcd",
+   "id": "75b584d8",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "fb847e48",
+   "id": "1fc276ec",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "3d153855",
+   "id": "58ff1d7e",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 7c3dd9f1..b1f1034f 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "597cb798",
+   "id": "d6515069",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "e36bdbd9",
+   "id": "04a7dd0d",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "3b523a95",
+   "id": "ebd046ec",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index 166270f1..33af8ef9 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "8f6b48fb",
+   "id": "7052e759",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "7493cc88",
+   "id": "75ba6731",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "81c0ff35",
+   "id": "8f3fa0b2",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
index dac16b8b..5caeed8e 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "51a1e576",
+   "id": "63dfc33e",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "d64e9479",
+   "id": "cad85973",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "3042ce15",
+   "id": "ae28fd0c",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 970c023d..8d27bc58 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "b848724e",
+   "id": "68afb1dd",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "8788e7c4",
+   "id": "2f146d9c",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "7068c50e",
+   "id": "4e37cc79",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
index f6316642..eff00c25 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "9bcbeaa3",
+   "id": "1dcebcc1",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "b22e7ea1",
+   "id": "db7b6fe9",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "8d021181",
+   "id": "3686c71d",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index 788c907c..00d0eac7 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "95531537",
+   "id": "943c1c8b",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "0fef3538",
+   "id": "8f976548",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "46bd0de8",
+   "id": "6e713663",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
index be67e3ea..98b0c203 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "58818d2e",
+   "id": "24738681",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "c813b621",
+   "id": "6f846386",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "43a8fe21",
+   "id": "49bb099c",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index ea2f14ee..6e705b4e 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "9490426c",
+   "id": "ebf86f9f",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "b9508024",
+   "id": "de031e8c",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "b8734044",
+   "id": "8be553db",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 13f75fd7..8d3c79fb 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "b9653bbc",
+   "id": "ad9e4227",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-False%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "bc4f4bea",
+   "id": "2a1f77f1",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "21f395d8",
+   "id": "b1780024",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index 104187c6..edc0d191 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "79122d9d",
+   "id": "162ef72d",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "fc882ad8",
+   "id": "441b6b2d",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "8fea05d6",
+   "id": "9a4de6e0",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
index b159bf46..19558583 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "65586bd2",
+   "id": "70625ee9",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "63a0e1e3",
+   "id": "2a5eb02b",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "d4e74f3b",
+   "id": "dc0fcac3",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 8f519823..d7780694 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "3771bbcd",
+   "id": "e8e30001",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "5df3fc22",
+   "id": "c5788223",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "1ecc2758",
+   "id": "53cdd4c2",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 25bf8d79..ed084f75 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "28ad47b2",
+   "id": "7142fd74",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "13a9a0aa",
+   "id": "a35a5ba5",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "23c57fba",
+   "id": "588b7144",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index 52f17a35..34c54df9 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "0231bf13",
+   "id": "0432470b",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "da7a2909",
+   "id": "166e1e9b",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "0226e5aa",
+   "id": "06588954",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
index 76fcb5ea..139943ee 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "9f8fc47f",
+   "id": "732de8f8",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "c8a2c192",
+   "id": "2b8b5585",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "1658d479",
+   "id": "26ae7d92",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 0039a12d..677c3996 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "a9134146",
+   "id": "2246a2f7",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "52be5d3d",
+   "id": "5aae703c",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "51e958dc",
+   "id": "c62b2a25",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 4bf4d55d..906138ed 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "1635bef4",
+   "id": "6f4a7609",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-False%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "c1a1b4f4",
+   "id": "161d1d16",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "8950d714",
+   "id": "2d461718",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index 289784fc..012e0992 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "c9463c84",
+   "id": "d8349645",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "c51e01db",
+   "id": "62bf9a3a",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "5d83f592",
+   "id": "c71714ad",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
index e3399cc2..eab0eaa1 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "a64a8bd5",
+   "id": "03936a43",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "08bbcff3",
+   "id": "ac77d4bf",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "39c9a4b1",
+   "id": "14310ece",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index 5f084652..e3cee8a9 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "bb093bfa",
+   "id": "d849ff22",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "7266afbf",
+   "id": "9ee4be2e",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "b994bfa9",
+   "id": "6a75e4d0",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
index 3afb584b..3af98859 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "d9d6be80",
+   "id": "bbe868bb",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-False%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "eb37796d",
+   "id": "e4aee1fb",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "07854f79",
+   "id": "95de2fac",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
index 47ea78f4..76e42f34 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "58c132a4",
+   "id": "d2a7430a",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "545c439e",
+   "id": "e2922d7e",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "5218c5d8",
+   "id": "58b623bb",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
index 9d81e07e..fa86770a 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "a81ffa24",
+   "id": "8bff8795",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-False%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "16104d87",
+   "id": "3ee75221",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "43df655e",
+   "id": "8985406d",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
index ad0cd0b6..399b0347 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "c33a39b5",
+   "id": "7c24aad5",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-batch.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "02cb5ac5",
+   "id": "0d6f6907",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "433a8206",
+   "id": "2bb31b45",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
index d804ecf6..51cee5fc 100644
--- a/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
+++ b/docs/generated_notebooks/ax/objective-single+model-GPEI+custom_gen-False+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.ipynb
@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "c50add83",
+   "id": "99b501a4",
    "metadata": {},
    "source": [
     "<a href=\"https://colab.research.google.com/github/sgbaird/honegumi/blob/main/docs/generated_notebooks/ax/objective-single%2Bmodel-GPEI%2Bcustom_gen-False%2Bexisting_data-True%2Bsum_constraint-True%2Border_constraint-True%2Blinear_constraint-True%2Bcomposition_constraint-True%2Bcategorical-True%2Bcustom_threshold-False%2Bsynchrony-single.ipynb\"><img alt=\"Open In Colab\" src=\"https://colab.research.google.com/assets/colab-badge.svg\"></a>"
@@ -11,7 +11,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "cb38d394",
+   "id": "e276e860",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -21,7 +21,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "07fd83d5",
+   "id": "5f00f3de",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
index 9534c41a..5b0521f8 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
@@ -38,7 +38,7 @@ def branin_moo(x1, x2):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
index 47866dd7..0afb4d5f 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
@@ -38,7 +38,7 @@ def branin_moo(x1, x2):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.py
index d3a14c26..28357060 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.py
@@ -38,7 +38,7 @@ def branin_moo(x1, x2):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.py
index c119aa06..04ab2c50 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.py
@@ -38,7 +38,7 @@ def branin_moo(x1, x2):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
index 79249d9d..4b3a4507 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
@@ -46,7 +46,7 @@ def branin_moo(x1, x2, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
index db79d4fe..b079687f 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
@@ -46,7 +46,7 @@ def branin_moo(x1, x2, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.py
index 53ee85c3..79b6399f 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.py
@@ -46,7 +46,7 @@ def branin_moo(x1, x2, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.py
index d6fa68eb..80e791a3 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.py
@@ -46,7 +46,7 @@ def branin_moo(x1, x2, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
index 722bcae8..1bf6fc08 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
@@ -47,7 +47,7 @@ def branin3_moo(x1, x2, x3):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
index 31b5637f..9b59bb32 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
@@ -47,7 +47,7 @@ def branin3_moo(x1, x2, x3):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.py
index e76536bd..23f2af87 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.py
@@ -47,7 +47,7 @@ def branin3_moo(x1, x2, x3):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.py
index fbeb4f5a..1f01978e 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.py
@@ -47,7 +47,7 @@ def branin3_moo(x1, x2, x3):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
index 0d193897..041c7c05 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
@@ -55,7 +55,7 @@ def branin3_moo(x1, x2, x3, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
index acff4d21..8152b96a 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
@@ -55,7 +55,7 @@ def branin3_moo(x1, x2, x3, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.py
index fe731e97..78453fb9 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.py
@@ -55,7 +55,7 @@ def branin3_moo(x1, x2, x3, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.py
index b38c93cd..846461ef 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.py
@@ -55,7 +55,7 @@ def branin3_moo(x1, x2, x3, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
index 4d3ddd85..e8ae7a77 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
@@ -38,7 +38,7 @@ def branin_moo(x1, x2):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
index 2abfc3a8..67b29d74 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
@@ -38,7 +38,7 @@ def branin_moo(x1, x2):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.py
index 58b6f12b..f72f4431 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.py
@@ -38,7 +38,7 @@ def branin_moo(x1, x2):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.py
index 1336abc2..5d2d94e9 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.py
@@ -38,7 +38,7 @@ def branin_moo(x1, x2):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
index da5dbfd6..2d0f70b3 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
@@ -46,7 +46,7 @@ def branin_moo(x1, x2, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
index 132da9b2..541d8467 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
@@ -46,7 +46,7 @@ def branin_moo(x1, x2, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.py
index 0dfd741e..44638633 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.py
@@ -46,7 +46,7 @@ def branin_moo(x1, x2, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.py
index 37dc7e96..40d00625 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.py
@@ -46,7 +46,7 @@ def branin_moo(x1, x2, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
index b2cf8111..8d7ef5a8 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
@@ -47,7 +47,7 @@ def branin3_moo(x1, x2, x3):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
index 111a0180..8cb367dd 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
@@ -47,7 +47,7 @@ def branin3_moo(x1, x2, x3):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.py
index 49c105fa..a7379830 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.py
@@ -47,7 +47,7 @@ def branin3_moo(x1, x2, x3):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.py
index 2ca63a3a..5c0a3783 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.py
@@ -47,7 +47,7 @@ def branin3_moo(x1, x2, x3):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
index c16b089c..32077735 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
@@ -55,7 +55,7 @@ def branin3_moo(x1, x2, x3, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
index c8dd87ad..5ee1a8b3 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
@@ -55,7 +55,7 @@ def branin3_moo(x1, x2, x3, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.py
index 94080122..dc6cba3e 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.py
@@ -55,7 +55,7 @@ def branin3_moo(x1, x2, x3, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.py
index 57924ce6..e64e9dda 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.py
@@ -55,7 +55,7 @@ def branin3_moo(x1, x2, x3, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
index 7d46e862..e1b1da3c 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
@@ -38,7 +38,7 @@ def branin_moo(x1, x2):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
index 8a39fa5c..9de27aef 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
@@ -38,7 +38,7 @@ def branin_moo(x1, x2):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.py
index 9c1e526f..9d57ea35 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.py
@@ -38,7 +38,7 @@ def branin_moo(x1, x2):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.py
index a1d2acc0..251cc6c2 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.py
@@ -38,7 +38,7 @@ def branin_moo(x1, x2):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
index 6b6b0dcc..b9d06f59 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
@@ -46,7 +46,7 @@ def branin_moo(x1, x2, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
index c3f6cff5..a7b7f88d 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
@@ -46,7 +46,7 @@ def branin_moo(x1, x2, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.py
index 2cfeb893..bb442d5b 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.py
@@ -46,7 +46,7 @@ def branin_moo(x1, x2, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.py
index 8d1858af..35ed4ab0 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.py
@@ -46,7 +46,7 @@ def branin_moo(x1, x2, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
index 8ed19004..7d8725fc 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
@@ -47,7 +47,7 @@ def branin3_moo(x1, x2, x3):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
index c40ab565..c89c35e1 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
@@ -47,7 +47,7 @@ def branin3_moo(x1, x2, x3):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.py
index 737950b1..9a2da963 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.py
@@ -47,7 +47,7 @@ def branin3_moo(x1, x2, x3):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.py
index 3b2cc2ff..e5ee4f47 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.py
@@ -47,7 +47,7 @@ def branin3_moo(x1, x2, x3):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
index f4b36b39..6f26aa89 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
@@ -55,7 +55,7 @@ def branin3_moo(x1, x2, x3, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
index 8292d0db..d318d950 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
@@ -55,7 +55,7 @@ def branin3_moo(x1, x2, x3, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.py
index db6456f9..4aa39ccc 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.py
@@ -55,7 +55,7 @@ def branin3_moo(x1, x2, x3, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.py
index d8a4d7e5..d52bb57f 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.py
@@ -55,7 +55,7 @@ def branin3_moo(x1, x2, x3, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
index 0997c2ef..84e9eee5 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
@@ -38,7 +38,7 @@ def branin_moo(x1, x2):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
index cac44af5..32a54866 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
@@ -38,7 +38,7 @@ def branin_moo(x1, x2):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.py
index b389aea4..e8812a3a 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.py
@@ -38,7 +38,7 @@ def branin_moo(x1, x2):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.py
index 78f52970..aaf06438 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.py
@@ -38,7 +38,7 @@ def branin_moo(x1, x2):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
index b247719a..2bc31962 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
@@ -46,7 +46,7 @@ def branin_moo(x1, x2, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
index 9d33924f..77fe4b72 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
@@ -46,7 +46,7 @@ def branin_moo(x1, x2, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.py
index 46b3035c..61a6ab2e 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.py
@@ -46,7 +46,7 @@ def branin_moo(x1, x2, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.py
index a4a0c4ec..4d35c9ef 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.py
@@ -46,7 +46,7 @@ def branin_moo(x1, x2, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
index d99cf420..0fb40a0d 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
@@ -47,7 +47,7 @@ def branin3_moo(x1, x2, x3):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
index 9b75a67e..7479c3ff 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
@@ -47,7 +47,7 @@ def branin3_moo(x1, x2, x3):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.py
index f4655658..07a28ff1 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.py
@@ -47,7 +47,7 @@ def branin3_moo(x1, x2, x3):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.py
index 70ab8ec1..e689cc57 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.py
@@ -47,7 +47,7 @@ def branin3_moo(x1, x2, x3):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
index 3241595b..f6c560d9 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
@@ -55,7 +55,7 @@ def branin3_moo(x1, x2, x3, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
index f713690e..45b5aa53 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
@@ -55,7 +55,7 @@ def branin3_moo(x1, x2, x3, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.py
index 95f470e8..7b18bd6e 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.py
@@ -55,7 +55,7 @@ def branin3_moo(x1, x2, x3, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.py
index 079f8eb1..5ca1f168 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.py
@@ -55,7 +55,7 @@ def branin3_moo(x1, x2, x3, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
index 84cda4a8..70b30396 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
@@ -38,7 +38,7 @@ def branin_moo(x1, x2):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
index 5a94344d..9eee57ba 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
@@ -38,7 +38,7 @@ def branin_moo(x1, x2):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.py
index 12357830..5a59a7f9 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.py
@@ -38,7 +38,7 @@ def branin_moo(x1, x2):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.py
index 63663a76..03300bfd 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.py
@@ -38,7 +38,7 @@ def branin_moo(x1, x2):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
index 7b952ad9..e978f5b3 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
@@ -46,7 +46,7 @@ def branin_moo(x1, x2, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
index 776f4517..34ba65dd 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
@@ -46,7 +46,7 @@ def branin_moo(x1, x2, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.py
index 02a104d5..042de344 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.py
@@ -46,7 +46,7 @@ def branin_moo(x1, x2, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.py
index 3d76fc51..dc3221db 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.py
@@ -46,7 +46,7 @@ def branin_moo(x1, x2, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
index 2a5e0605..56860cc3 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
@@ -47,7 +47,7 @@ def branin3_moo(x1, x2, x3):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
index fafe9779..4af5cd6e 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
@@ -47,7 +47,7 @@ def branin3_moo(x1, x2, x3):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.py
index f7616907..0a0b7a7f 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.py
@@ -47,7 +47,7 @@ def branin3_moo(x1, x2, x3):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.py
index f7e4a159..f3900d32 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.py
@@ -47,7 +47,7 @@ def branin3_moo(x1, x2, x3):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
index 6dea4140..c00902f6 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
@@ -55,7 +55,7 @@ def branin3_moo(x1, x2, x3, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
index a9e4f267..c555f473 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
@@ -55,7 +55,7 @@ def branin3_moo(x1, x2, x3, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.py
index bf99a640..1637eec9 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.py
@@ -55,7 +55,7 @@ def branin3_moo(x1, x2, x3, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.py
index a990c48c..cbec29f0 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.py
@@ -55,7 +55,7 @@ def branin3_moo(x1, x2, x3, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
index 32410211..c98c19ac 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
@@ -38,7 +38,7 @@ def branin_moo(x1, x2):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
index 1db5b507..a170e953 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
@@ -38,7 +38,7 @@ def branin_moo(x1, x2):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.py
index d968ae1f..2c5cd172 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.py
@@ -38,7 +38,7 @@ def branin_moo(x1, x2):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.py
index 07682bf9..6f77fa0b 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.py
@@ -38,7 +38,7 @@ def branin_moo(x1, x2):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
index 2ad94e8b..9eeafe4a 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
@@ -46,7 +46,7 @@ def branin_moo(x1, x2, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
index d059146a..88969149 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
@@ -46,7 +46,7 @@ def branin_moo(x1, x2, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.py
index 30e1bbe0..76e01dbc 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.py
@@ -46,7 +46,7 @@ def branin_moo(x1, x2, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.py
index 96a7a42b..d75d4468 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.py
@@ -46,7 +46,7 @@ def branin_moo(x1, x2, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
index 6fcebf00..259db8f0 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
@@ -47,7 +47,7 @@ def branin3_moo(x1, x2, x3):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
index fbd5ff5a..fb015584 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
@@ -47,7 +47,7 @@ def branin3_moo(x1, x2, x3):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.py
index 66fb00be..35764615 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.py
@@ -47,7 +47,7 @@ def branin3_moo(x1, x2, x3):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.py
index fded4fb9..336b339b 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.py
@@ -47,7 +47,7 @@ def branin3_moo(x1, x2, x3):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
index 184b9670..e0eefd81 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
@@ -55,7 +55,7 @@ def branin3_moo(x1, x2, x3, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
index 486d24f1..b3f8f064 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
@@ -55,7 +55,7 @@ def branin3_moo(x1, x2, x3, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.py
index 92bab993..4965aba2 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.py
@@ -55,7 +55,7 @@ def branin3_moo(x1, x2, x3, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.py
index ac5f8a4b..b1436117 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.py
@@ -55,7 +55,7 @@ def branin3_moo(x1, x2, x3, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
index d8d6bc3b..d9383655 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
@@ -38,7 +38,7 @@ def branin_moo(x1, x2):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
index 226fb19f..2f181da7 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
@@ -38,7 +38,7 @@ def branin_moo(x1, x2):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.py
index 2b8ef921..1b7745c2 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.py
@@ -38,7 +38,7 @@ def branin_moo(x1, x2):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.py
index cd631ec9..61ffbcf6 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.py
@@ -38,7 +38,7 @@ def branin_moo(x1, x2):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
index a43a14b1..1f5cfc96 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
@@ -46,7 +46,7 @@ def branin_moo(x1, x2, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
index bbd5cb33..60202b35 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
@@ -46,7 +46,7 @@ def branin_moo(x1, x2, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.py
index b6765f39..9b7f8efb 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.py
@@ -46,7 +46,7 @@ def branin_moo(x1, x2, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.py
index 44eda978..6c9a4bc6 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.py
@@ -46,7 +46,7 @@ def branin_moo(x1, x2, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
index cb7eef17..7a6424d0 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
@@ -47,7 +47,7 @@ def branin3_moo(x1, x2, x3):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
index 2adf275c..b09f0058 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
@@ -47,7 +47,7 @@ def branin3_moo(x1, x2, x3):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.py
index dc1e6d13..c1558c73 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.py
@@ -47,7 +47,7 @@ def branin3_moo(x1, x2, x3):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.py
index 81a128f6..bca53e17 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.py
@@ -47,7 +47,7 @@ def branin3_moo(x1, x2, x3):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
index 54d3e256..063f86d9 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
@@ -55,7 +55,7 @@ def branin3_moo(x1, x2, x3, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
index d798216b..9990282b 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
@@ -55,7 +55,7 @@ def branin3_moo(x1, x2, x3, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.py
index 3e4ae8f6..d4d1d2bb 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.py
@@ -55,7 +55,7 @@ def branin3_moo(x1, x2, x3, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.py
index 6abb671e..aae656f6 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.py
@@ -55,7 +55,7 @@ def branin3_moo(x1, x2, x3, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
index 455835c7..aab7564b 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
@@ -38,7 +38,7 @@ def branin_moo(x1, x2):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
index 7e5d7df0..f4d47ec3 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
@@ -38,7 +38,7 @@ def branin_moo(x1, x2):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.py
index 96b5a301..02a62449 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.py
@@ -38,7 +38,7 @@ def branin_moo(x1, x2):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.py
index a7b05dfa..4ee96da6 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.py
@@ -38,7 +38,7 @@ def branin_moo(x1, x2):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
index dfed9250..b2074780 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
@@ -46,7 +46,7 @@ def branin_moo(x1, x2, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
index a83551f6..22f7dfa9 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
@@ -46,7 +46,7 @@ def branin_moo(x1, x2, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.py
index 5a98c766..62e477ed 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.py
@@ -46,7 +46,7 @@ def branin_moo(x1, x2, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.py
index b49d8787..18c7d0b9 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.py
@@ -46,7 +46,7 @@ def branin_moo(x1, x2, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
index 2f739641..bb6017a4 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
@@ -47,7 +47,7 @@ def branin3_moo(x1, x2, x3):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
index 0321be0b..f6309865 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
@@ -47,7 +47,7 @@ def branin3_moo(x1, x2, x3):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.py
index baaf1e6e..04f4e4b4 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.py
@@ -47,7 +47,7 @@ def branin3_moo(x1, x2, x3):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.py
index 7cba8fa2..089db70f 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.py
@@ -47,7 +47,7 @@ def branin3_moo(x1, x2, x3):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
index dd243c70..13d3b8b2 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
@@ -55,7 +55,7 @@ def branin3_moo(x1, x2, x3, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
index 341c044c..9381b041 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
@@ -55,7 +55,7 @@ def branin3_moo(x1, x2, x3, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.py
index 5f6a8aaf..f0fffb84 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.py
@@ -55,7 +55,7 @@ def branin3_moo(x1, x2, x3, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.py
index 862f96eb..1a3d67f0 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.py
@@ -55,7 +55,7 @@ def branin3_moo(x1, x2, x3, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
index cf8849ae..1e31498c 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
@@ -59,7 +59,7 @@ def branin_moo(x1, x2):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
index 67631f52..f5393f5c 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
@@ -59,7 +59,7 @@ def branin_moo(x1, x2):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.py
index a94c6565..bb3caa60 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.py
@@ -59,7 +59,7 @@ def branin_moo(x1, x2):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.py
index 8fccb8d8..1b4d23b5 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.py
@@ -59,7 +59,7 @@ def branin_moo(x1, x2):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
index 23db33fd..06bfdf88 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
@@ -67,7 +67,7 @@ def branin_moo(x1, x2, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
index 0d076d19..d70be0c5 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
@@ -67,7 +67,7 @@ def branin_moo(x1, x2, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.py
index 590f7954..4091cf23 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.py
@@ -67,7 +67,7 @@ def branin_moo(x1, x2, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.py
index f3f2bd32..b158faff 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.py
@@ -67,7 +67,7 @@ def branin_moo(x1, x2, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
index 0c80b85e..177a3ce2 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
@@ -72,7 +72,7 @@ def branin3_moo(x1, x2, x3):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
index a232d724..974de135 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
@@ -72,7 +72,7 @@ def branin3_moo(x1, x2, x3):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.py
index 0f9b7d4e..dc9a7142 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.py
@@ -72,7 +72,7 @@ def branin3_moo(x1, x2, x3):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.py
index fc73717e..34ab735c 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.py
@@ -72,7 +72,7 @@ def branin3_moo(x1, x2, x3):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
index 451b54b7..3da5b6c6 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
@@ -81,7 +81,7 @@ def branin3_moo(x1, x2, x3, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
index 68450b6a..5c19ba68 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
@@ -81,7 +81,7 @@ def branin3_moo(x1, x2, x3, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.py
index c5c17f2e..59f0cb92 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.py
@@ -81,7 +81,7 @@ def branin3_moo(x1, x2, x3, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.py
index c41dfbbc..05f39e1c 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.py
@@ -81,7 +81,7 @@ def branin3_moo(x1, x2, x3, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
index 97b2e629..f7d4fa92 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
@@ -59,7 +59,7 @@ def branin_moo(x1, x2):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
index 48192409..d82615db 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
@@ -59,7 +59,7 @@ def branin_moo(x1, x2):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.py
index 57f7bc50..e8c28537 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.py
@@ -59,7 +59,7 @@ def branin_moo(x1, x2):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.py
index d5a9b659..da8ee791 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.py
@@ -59,7 +59,7 @@ def branin_moo(x1, x2):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
index 811c2b86..af4b2af7 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
@@ -67,7 +67,7 @@ def branin_moo(x1, x2, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
index 3db47269..e72a7358 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
@@ -67,7 +67,7 @@ def branin_moo(x1, x2, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.py
index 0e293a9d..a4f16de7 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.py
@@ -67,7 +67,7 @@ def branin_moo(x1, x2, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.py
index 8f2a923b..d18981c0 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.py
@@ -67,7 +67,7 @@ def branin_moo(x1, x2, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
index ed4a1db5..dca70bba 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
@@ -72,7 +72,7 @@ def branin3_moo(x1, x2, x3):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
index f6b20f19..03509d17 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
@@ -72,7 +72,7 @@ def branin3_moo(x1, x2, x3):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.py
index d3fc9450..8011627e 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.py
@@ -72,7 +72,7 @@ def branin3_moo(x1, x2, x3):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.py
index cdea82b0..81f307f2 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.py
@@ -72,7 +72,7 @@ def branin3_moo(x1, x2, x3):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
index 76712b76..a18b714d 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
@@ -81,7 +81,7 @@ def branin3_moo(x1, x2, x3, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
index f3417690..784e0309 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
@@ -81,7 +81,7 @@ def branin3_moo(x1, x2, x3, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.py
index 7adedfb8..07c85799 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.py
@@ -81,7 +81,7 @@ def branin3_moo(x1, x2, x3, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.py
index b4a0b711..17393dbf 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.py
@@ -81,7 +81,7 @@ def branin3_moo(x1, x2, x3, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
index 0fd377ae..fc21ddee 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
@@ -62,7 +62,7 @@ def branin_moo(x1, x2):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
index a1962e2b..077de390 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
@@ -62,7 +62,7 @@ def branin_moo(x1, x2):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.py
index e0f62738..4c095b7e 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.py
@@ -62,7 +62,7 @@ def branin_moo(x1, x2):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.py
index 1806f45d..d279e7b7 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.py
@@ -62,7 +62,7 @@ def branin_moo(x1, x2):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
index ae196207..065c0a32 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
@@ -70,7 +70,7 @@ def branin_moo(x1, x2, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
index c753bdab..bae9cb0d 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
@@ -70,7 +70,7 @@ def branin_moo(x1, x2, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.py
index caac85b6..004ba92a 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.py
@@ -70,7 +70,7 @@ def branin_moo(x1, x2, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.py
index c3b2428d..3ccc602a 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.py
@@ -70,7 +70,7 @@ def branin_moo(x1, x2, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
index 4c8ccacd..71138c29 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
@@ -75,7 +75,7 @@ def branin3_moo(x1, x2, x3):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
index 4d702119..076f5f17 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
@@ -75,7 +75,7 @@ def branin3_moo(x1, x2, x3):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.py
index d032770a..942ddaeb 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.py
@@ -75,7 +75,7 @@ def branin3_moo(x1, x2, x3):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.py
index 802ea2f6..15df69ef 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.py
@@ -75,7 +75,7 @@ def branin3_moo(x1, x2, x3):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
index f3873e38..f576f906 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
@@ -84,7 +84,7 @@ def branin3_moo(x1, x2, x3, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
index 475184c3..db6d99d7 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
@@ -84,7 +84,7 @@ def branin3_moo(x1, x2, x3, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.py
index 49ff0c23..e5d028d4 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.py
@@ -84,7 +84,7 @@ def branin3_moo(x1, x2, x3, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.py
index 3fa3f0b6..d22904db 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.py
@@ -84,7 +84,7 @@ def branin3_moo(x1, x2, x3, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
index 2ba188ae..8e0816bb 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
@@ -62,7 +62,7 @@ def branin_moo(x1, x2):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
index 09477df8..1a2d1801 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
@@ -62,7 +62,7 @@ def branin_moo(x1, x2):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.py
index 85426d61..2fd2addd 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.py
@@ -62,7 +62,7 @@ def branin_moo(x1, x2):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.py
index 0d1f3512..3e8c9c11 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.py
@@ -62,7 +62,7 @@ def branin_moo(x1, x2):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
index a4700167..c1f298c7 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
@@ -70,7 +70,7 @@ def branin_moo(x1, x2, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
index 19104638..5422735d 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
@@ -70,7 +70,7 @@ def branin_moo(x1, x2, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.py
index 7ee0c1e3..fadfd60a 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.py
@@ -70,7 +70,7 @@ def branin_moo(x1, x2, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.py
index 51d6fe2a..ce407a72 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.py
@@ -70,7 +70,7 @@ def branin_moo(x1, x2, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
index 7157d4af..f4d54ea4 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
@@ -75,7 +75,7 @@ def branin3_moo(x1, x2, x3):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
index 480f10dc..2e0fdb78 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
@@ -75,7 +75,7 @@ def branin3_moo(x1, x2, x3):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.py
index 82c40019..2603b718 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.py
@@ -75,7 +75,7 @@ def branin3_moo(x1, x2, x3):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.py
index b1ce0468..2b1aa86a 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.py
@@ -75,7 +75,7 @@ def branin3_moo(x1, x2, x3):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
index 541612d8..e264c9bd 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
@@ -84,7 +84,7 @@ def branin3_moo(x1, x2, x3, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
index 99bf7b42..86a327c1 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
@@ -84,7 +84,7 @@ def branin3_moo(x1, x2, x3, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.py
index f9e889e2..03ac412b 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.py
@@ -84,7 +84,7 @@ def branin3_moo(x1, x2, x3, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.py
index 22faaab6..65e6f41f 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.py
@@ -84,7 +84,7 @@ def branin3_moo(x1, x2, x3, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
index b7f18219..08910def 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
@@ -59,7 +59,7 @@ def branin_moo(x1, x2):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
index 5a19befc..0e5ad0dd 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
@@ -59,7 +59,7 @@ def branin_moo(x1, x2):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.py
index 0cd3912b..60c53485 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.py
@@ -59,7 +59,7 @@ def branin_moo(x1, x2):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.py
index ff521986..f2fe1981 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.py
@@ -59,7 +59,7 @@ def branin_moo(x1, x2):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
index 75caa9b5..41c8add7 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
@@ -67,7 +67,7 @@ def branin_moo(x1, x2, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
index 0f108789..7b4ed306 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
@@ -67,7 +67,7 @@ def branin_moo(x1, x2, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.py
index 1798d21b..3d6e7334 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.py
@@ -67,7 +67,7 @@ def branin_moo(x1, x2, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.py
index 2391ca9d..9f28ef55 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.py
@@ -67,7 +67,7 @@ def branin_moo(x1, x2, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
index b55ba70f..f5a5b960 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
@@ -72,7 +72,7 @@ def branin3_moo(x1, x2, x3):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
index 45219ad3..0d7fac38 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
@@ -72,7 +72,7 @@ def branin3_moo(x1, x2, x3):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.py
index 8750065e..b56959a7 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.py
@@ -72,7 +72,7 @@ def branin3_moo(x1, x2, x3):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.py
index cece7f71..d0d28f0b 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.py
@@ -72,7 +72,7 @@ def branin3_moo(x1, x2, x3):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
index f95333e8..ead420d0 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
@@ -81,7 +81,7 @@ def branin3_moo(x1, x2, x3, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
index efe37658..fa96e862 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
@@ -81,7 +81,7 @@ def branin3_moo(x1, x2, x3, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.py
index bc5ed6cb..b4ebcf16 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.py
@@ -81,7 +81,7 @@ def branin3_moo(x1, x2, x3, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.py
index 5b54f4d8..5e2aec68 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.py
@@ -81,7 +81,7 @@ def branin3_moo(x1, x2, x3, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
index f61372f6..fefd45a2 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
@@ -59,7 +59,7 @@ def branin_moo(x1, x2):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
index 732f85ec..3a667049 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
@@ -59,7 +59,7 @@ def branin_moo(x1, x2):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.py
index c1747af5..01b1d598 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.py
@@ -59,7 +59,7 @@ def branin_moo(x1, x2):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.py
index 2a3a5464..d9cbaa6d 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.py
@@ -59,7 +59,7 @@ def branin_moo(x1, x2):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
index 1075ae71..af9f1c3d 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
@@ -67,7 +67,7 @@ def branin_moo(x1, x2, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
index 429140cd..97bca5ad 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
@@ -67,7 +67,7 @@ def branin_moo(x1, x2, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.py
index 0f3a2ba8..46bc9159 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.py
@@ -67,7 +67,7 @@ def branin_moo(x1, x2, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.py
index af3837e1..2748c214 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.py
@@ -67,7 +67,7 @@ def branin_moo(x1, x2, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
index 5eb4675a..47bbdb40 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
@@ -72,7 +72,7 @@ def branin3_moo(x1, x2, x3):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
index 9a5421fb..8bd4a1e1 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
@@ -72,7 +72,7 @@ def branin3_moo(x1, x2, x3):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.py
index 485e2843..b9b48d05 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.py
@@ -72,7 +72,7 @@ def branin3_moo(x1, x2, x3):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.py
index 03294970..f859031f 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.py
@@ -72,7 +72,7 @@ def branin3_moo(x1, x2, x3):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
index 940ee960..8265ba96 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
@@ -81,7 +81,7 @@ def branin3_moo(x1, x2, x3, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
index 2ba58cd2..f5517fb2 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
@@ -81,7 +81,7 @@ def branin3_moo(x1, x2, x3, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.py
index dcd529d3..da05e98b 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.py
@@ -81,7 +81,7 @@ def branin3_moo(x1, x2, x3, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.py
index 6d608ab7..cb73d996 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.py
@@ -81,7 +81,7 @@ def branin3_moo(x1, x2, x3, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
index cb224da9..d0ec189d 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
@@ -62,7 +62,7 @@ def branin_moo(x1, x2):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
index 835fd646..6cc83613 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
@@ -62,7 +62,7 @@ def branin_moo(x1, x2):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.py
index c2c46ee3..653df808 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.py
@@ -62,7 +62,7 @@ def branin_moo(x1, x2):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.py
index 7080d5b7..cf3be3b8 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.py
@@ -62,7 +62,7 @@ def branin_moo(x1, x2):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
index 49ac7783..edd04b6a 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
@@ -70,7 +70,7 @@ def branin_moo(x1, x2, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
index f076652f..4d0a318b 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
@@ -70,7 +70,7 @@ def branin_moo(x1, x2, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.py
index 00415fc1..5eb937b2 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.py
@@ -70,7 +70,7 @@ def branin_moo(x1, x2, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.py
index b2e10695..43bc5db8 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.py
@@ -70,7 +70,7 @@ def branin_moo(x1, x2, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
index a1fdcfad..e31393ae 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
@@ -75,7 +75,7 @@ def branin3_moo(x1, x2, x3):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
index f8db30b5..de1ca373 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
@@ -75,7 +75,7 @@ def branin3_moo(x1, x2, x3):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.py
index 1308061c..9dc10d07 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.py
@@ -75,7 +75,7 @@ def branin3_moo(x1, x2, x3):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.py
index c8f0e628..f8fbbc2d 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.py
@@ -75,7 +75,7 @@ def branin3_moo(x1, x2, x3):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
index 595559e2..d0354721 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
@@ -84,7 +84,7 @@ def branin3_moo(x1, x2, x3, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
index 639d6d26..8bbe369e 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
@@ -84,7 +84,7 @@ def branin3_moo(x1, x2, x3, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.py
index 5141f8da..4861ce74 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.py
@@ -84,7 +84,7 @@ def branin3_moo(x1, x2, x3, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.py
index 52e37a39..4d0fd697 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.py
@@ -84,7 +84,7 @@ def branin3_moo(x1, x2, x3, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
index 044dfc53..8492a992 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
@@ -62,7 +62,7 @@ def branin_moo(x1, x2):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
index 44e5e58f..3729eddd 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
@@ -62,7 +62,7 @@ def branin_moo(x1, x2):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.py
index 80e88008..289b39e2 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-batch.py
@@ -62,7 +62,7 @@ def branin_moo(x1, x2):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.py
index d457eb96..662b97cf 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-True+synchrony-single.py
@@ -62,7 +62,7 @@ def branin_moo(x1, x2):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
index 7dd6e8bf..02a01ac7 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
@@ -70,7 +70,7 @@ def branin_moo(x1, x2, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
index 23a3e558..fb1fb201 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
@@ -70,7 +70,7 @@ def branin_moo(x1, x2, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.py
index b660d495..4220c96a 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-batch.py
@@ -70,7 +70,7 @@ def branin_moo(x1, x2, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.py
index 9106abcf..07d17697 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-True+synchrony-single.py
@@ -70,7 +70,7 @@ def branin_moo(x1, x2, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
index c5d88b63..5e3439a1 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
@@ -75,7 +75,7 @@ def branin3_moo(x1, x2, x3):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
index 64330c43..d80e800c 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
@@ -75,7 +75,7 @@ def branin3_moo(x1, x2, x3):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.py
index 59d0bb22..611a5af9 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-batch.py
@@ -75,7 +75,7 @@ def branin3_moo(x1, x2, x3):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.py
index 2780cfa9..c8e5f8d5 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-True+synchrony-single.py
@@ -75,7 +75,7 @@ def branin3_moo(x1, x2, x3):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
index 53f6d588..93041eff 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
@@ -84,7 +84,7 @@ def branin3_moo(x1, x2, x3, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
index 25e90401..0ecb3c87 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
@@ -84,7 +84,7 @@ def branin3_moo(x1, x2, x3, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.py
index 585b0bd5..90715ca1 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-batch.py
@@ -84,7 +84,7 @@ def branin3_moo(x1, x2, x3, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.py b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.py
index 89410847..84734445 100644
--- a/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-multi+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-True+synchrony-single.py
@@ -84,7 +84,7 @@ def branin3_moo(x1, x2, x3, c1):
             model=Models.FULLYBAYESIANMOO,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
index 1f51ccb9..d334f8f6 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
@@ -30,7 +30,7 @@ def branin(x1, x2):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
index 065e18c6..6f8ecdf2 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
@@ -30,7 +30,7 @@ def branin(x1, x2):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
index 03a676ec..c06793df 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
@@ -34,7 +34,7 @@ def branin(x1, x2, c1):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
index afff4ac6..65f62e3f 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
@@ -34,7 +34,7 @@ def branin(x1, x2, c1):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
index 0fcd505b..ee9e436b 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
@@ -37,7 +37,7 @@ def branin3(x1, x2, x3):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
index 1e731e2e..a5ee3d15 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
@@ -37,7 +37,7 @@ def branin3(x1, x2, x3):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
index 215640f4..4f9ad836 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
@@ -41,7 +41,7 @@ def branin3(x1, x2, x3, c1):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
index 7f184495..2556fbd9 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
@@ -41,7 +41,7 @@ def branin3(x1, x2, x3, c1):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
index 40240c45..0860e924 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
@@ -30,7 +30,7 @@ def branin(x1, x2):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
index 4c5766af..2bd702d4 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
@@ -30,7 +30,7 @@ def branin(x1, x2):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
index 3e4a6d50..d076a638 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
@@ -34,7 +34,7 @@ def branin(x1, x2, c1):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
index e0fdc426..029e63e2 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
@@ -34,7 +34,7 @@ def branin(x1, x2, c1):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
index d96b7a15..c953c2a6 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
@@ -37,7 +37,7 @@ def branin3(x1, x2, x3):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
index cfacd203..7515891d 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
@@ -37,7 +37,7 @@ def branin3(x1, x2, x3):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
index bd597270..f3891501 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
@@ -41,7 +41,7 @@ def branin3(x1, x2, x3, c1):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
index 3fef4a39..7f2fe986 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
@@ -41,7 +41,7 @@ def branin3(x1, x2, x3, c1):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
index af60e3f4..c662ff65 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
@@ -30,7 +30,7 @@ def branin(x1, x2):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
index 8947b696..708c2ab8 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
@@ -30,7 +30,7 @@ def branin(x1, x2):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
index 8bab6e35..b7332c5e 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
@@ -34,7 +34,7 @@ def branin(x1, x2, c1):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
index a0233dc2..751ed64f 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
@@ -34,7 +34,7 @@ def branin(x1, x2, c1):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
index d93cb7bb..55d808a2 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
@@ -37,7 +37,7 @@ def branin3(x1, x2, x3):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
index 820b2562..f34d68fc 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
@@ -37,7 +37,7 @@ def branin3(x1, x2, x3):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
index 6e7c3272..2de7603a 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
@@ -41,7 +41,7 @@ def branin3(x1, x2, x3, c1):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
index f1eac028..a97a8e78 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
@@ -41,7 +41,7 @@ def branin3(x1, x2, x3, c1):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
index a9dbcbee..7666e39c 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
@@ -30,7 +30,7 @@ def branin(x1, x2):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
index 7cc2fcb4..ac111288 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
@@ -30,7 +30,7 @@ def branin(x1, x2):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
index c9af2329..62be1172 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
@@ -34,7 +34,7 @@ def branin(x1, x2, c1):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
index e9cb39c6..eefbc28e 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
@@ -34,7 +34,7 @@ def branin(x1, x2, c1):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
index c383f1db..7cd39ea8 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
@@ -37,7 +37,7 @@ def branin3(x1, x2, x3):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
index 1046d271..f035e728 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
@@ -37,7 +37,7 @@ def branin3(x1, x2, x3):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
index 939a4037..93876677 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
@@ -41,7 +41,7 @@ def branin3(x1, x2, x3, c1):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
index 884ea7a3..1319ec01 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
@@ -41,7 +41,7 @@ def branin3(x1, x2, x3, c1):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
index e106552a..20266faa 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
@@ -30,7 +30,7 @@ def branin(x1, x2):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
index e359c773..19f76c7b 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
@@ -30,7 +30,7 @@ def branin(x1, x2):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
index 4c2a129d..6304699b 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
@@ -34,7 +34,7 @@ def branin(x1, x2, c1):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
index 7de71b02..5736d1b6 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
@@ -34,7 +34,7 @@ def branin(x1, x2, c1):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
index d5839053..243e2b02 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
@@ -37,7 +37,7 @@ def branin3(x1, x2, x3):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
index aad8acdf..8305c240 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
@@ -37,7 +37,7 @@ def branin3(x1, x2, x3):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
index 7c2e8744..72014fd2 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
@@ -41,7 +41,7 @@ def branin3(x1, x2, x3, c1):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
index a6005129..c109f122 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
@@ -41,7 +41,7 @@ def branin3(x1, x2, x3, c1):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
index 1a0675f9..189d0141 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
@@ -30,7 +30,7 @@ def branin(x1, x2):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
index 3aa66bb2..018d9f7f 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
@@ -30,7 +30,7 @@ def branin(x1, x2):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
index a29d50a6..9190890f 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
@@ -34,7 +34,7 @@ def branin(x1, x2, c1):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
index a38970aa..45e1449d 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
@@ -34,7 +34,7 @@ def branin(x1, x2, c1):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
index 940a1cf7..9d3bf67c 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
@@ -37,7 +37,7 @@ def branin3(x1, x2, x3):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
index b185baff..148224db 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
@@ -37,7 +37,7 @@ def branin3(x1, x2, x3):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
index a0f954ac..290dd053 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
@@ -41,7 +41,7 @@ def branin3(x1, x2, x3, c1):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
index dea8829d..f65ea6fd 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
@@ -41,7 +41,7 @@ def branin3(x1, x2, x3, c1):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
index 745cfd0a..070d6cf0 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
@@ -30,7 +30,7 @@ def branin(x1, x2):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
index b8491d35..38ed2dc7 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
@@ -30,7 +30,7 @@ def branin(x1, x2):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
index 264dec99..c6ae6f8f 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
@@ -34,7 +34,7 @@ def branin(x1, x2, c1):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
index d8426948..663de6cd 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
@@ -34,7 +34,7 @@ def branin(x1, x2, c1):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
index 47aaf9f6..f05c6c13 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
@@ -37,7 +37,7 @@ def branin3(x1, x2, x3):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
index 3bdae9df..bb075fd6 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
@@ -37,7 +37,7 @@ def branin3(x1, x2, x3):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
index 286e9edb..68e685ba 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
@@ -41,7 +41,7 @@ def branin3(x1, x2, x3, c1):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
index ba300779..c7994461 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
@@ -41,7 +41,7 @@ def branin3(x1, x2, x3, c1):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
index e63cb09e..b923d07e 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
@@ -30,7 +30,7 @@ def branin(x1, x2):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
index 2cc5104a..7c273643 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
@@ -30,7 +30,7 @@ def branin(x1, x2):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
index cfd3294a..9c8ce875 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
@@ -34,7 +34,7 @@ def branin(x1, x2, c1):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
index 830aa55d..bc7e8b32 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
@@ -34,7 +34,7 @@ def branin(x1, x2, c1):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
index 69d6aa9f..516b15b7 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
@@ -37,7 +37,7 @@ def branin3(x1, x2, x3):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
index c8848ba8..637e093d 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
@@ -37,7 +37,7 @@ def branin3(x1, x2, x3):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
index 2503c39b..feab1fdf 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
@@ -41,7 +41,7 @@ def branin3(x1, x2, x3, c1):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
index 4d0457ee..a605911b 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-False+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
@@ -41,7 +41,7 @@ def branin3(x1, x2, x3, c1):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
index 4303521b..104abf3d 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
@@ -51,7 +51,7 @@ def branin(x1, x2):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
index 213fe7f9..f4a0f355 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
@@ -51,7 +51,7 @@ def branin(x1, x2):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
index 6bc71bbe..50aa23bd 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
@@ -55,7 +55,7 @@ def branin(x1, x2, c1):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
index 46530892..45cc4f0f 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
@@ -55,7 +55,7 @@ def branin(x1, x2, c1):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
index 717a5c01..dd51676a 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
@@ -60,7 +60,7 @@ def branin3(x1, x2, x3):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
index e6018dc8..12912602 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
@@ -60,7 +60,7 @@ def branin3(x1, x2, x3):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
index ebe91e66..160f04c0 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
@@ -66,7 +66,7 @@ def branin3(x1, x2, x3, c1):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
index 40b39bda..656cef29 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
@@ -66,7 +66,7 @@ def branin3(x1, x2, x3, c1):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
index 46e2aaa0..06e7bee8 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
@@ -51,7 +51,7 @@ def branin(x1, x2):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
index 99f07ce6..c30db602 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
@@ -51,7 +51,7 @@ def branin(x1, x2):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
index ba9910bf..26e4a15b 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
@@ -55,7 +55,7 @@ def branin(x1, x2, c1):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
index 12d05b2d..8ec44073 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
@@ -55,7 +55,7 @@ def branin(x1, x2, c1):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
index bc503f33..afd42084 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
@@ -60,7 +60,7 @@ def branin3(x1, x2, x3):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
index a6c83563..f8d5b390 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
@@ -60,7 +60,7 @@ def branin3(x1, x2, x3):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
index 7ac9609b..53143123 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
@@ -66,7 +66,7 @@ def branin3(x1, x2, x3, c1):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
index cbb36100..4f82d96b 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
@@ -66,7 +66,7 @@ def branin3(x1, x2, x3, c1):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
index e74f8969..0fb6a037 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
@@ -54,7 +54,7 @@ def branin(x1, x2):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
index 8f4ff0cc..1c12c9da 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
@@ -54,7 +54,7 @@ def branin(x1, x2):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
index 5f12d704..21e1d89a 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
@@ -58,7 +58,7 @@ def branin(x1, x2, c1):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
index a594904e..79484a7a 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
@@ -58,7 +58,7 @@ def branin(x1, x2, c1):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
index 9c57cbed..8ecf7b39 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
@@ -63,7 +63,7 @@ def branin3(x1, x2, x3):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
index 6902d11a..8648319c 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
@@ -63,7 +63,7 @@ def branin3(x1, x2, x3):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
index 59433a79..7d9793b2 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
@@ -69,7 +69,7 @@ def branin3(x1, x2, x3, c1):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
index 811aa68a..375e2380 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
@@ -69,7 +69,7 @@ def branin3(x1, x2, x3, c1):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
index e50f2e62..57362724 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
@@ -54,7 +54,7 @@ def branin(x1, x2):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
index 13c777ae..d289229c 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
@@ -54,7 +54,7 @@ def branin(x1, x2):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
index c10968e1..8111b2c3 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
@@ -58,7 +58,7 @@ def branin(x1, x2, c1):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
index ef7ce092..cda52b38 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
@@ -58,7 +58,7 @@ def branin(x1, x2, c1):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
index da08c0bf..c3dece72 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
@@ -63,7 +63,7 @@ def branin3(x1, x2, x3):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
index 8048a06f..2b54a41d 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
@@ -63,7 +63,7 @@ def branin3(x1, x2, x3):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
index de834d67..164a4f83 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
@@ -69,7 +69,7 @@ def branin3(x1, x2, x3, c1):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
index 89265544..f4f12067 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-False+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
@@ -69,7 +69,7 @@ def branin3(x1, x2, x3, c1):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
index e9766969..b8022833 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
@@ -51,7 +51,7 @@ def branin(x1, x2):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
index 03baee3f..2936efea 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
@@ -51,7 +51,7 @@ def branin(x1, x2):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
index 14c3e163..d4b97249 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
@@ -55,7 +55,7 @@ def branin(x1, x2, c1):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
index 1603cbcc..b4849f61 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
@@ -55,7 +55,7 @@ def branin(x1, x2, c1):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
index a04ad759..0fa73c51 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
@@ -60,7 +60,7 @@ def branin3(x1, x2, x3):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
index 05f8ec09..ce4c42eb 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
@@ -60,7 +60,7 @@ def branin3(x1, x2, x3):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
index 1fe451df..e068d538 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
@@ -66,7 +66,7 @@ def branin3(x1, x2, x3, c1):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
index a6a6b817..bdb5f410 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
@@ -66,7 +66,7 @@ def branin3(x1, x2, x3, c1):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
index 08bd3bbf..11fcbfee 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
@@ -51,7 +51,7 @@ def branin(x1, x2):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
index c530913a..c7c91863 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
@@ -51,7 +51,7 @@ def branin(x1, x2):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
index 21a293fb..618720c2 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
@@ -55,7 +55,7 @@ def branin(x1, x2, c1):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
index 9b6a1c01..0632c7c6 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
@@ -55,7 +55,7 @@ def branin(x1, x2, c1):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
index d27a652c..c575d88a 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
@@ -60,7 +60,7 @@ def branin3(x1, x2, x3):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
index e18e511e..56644c1f 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
@@ -60,7 +60,7 @@ def branin3(x1, x2, x3):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
index fca0f4b6..56ff3cd1 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
@@ -66,7 +66,7 @@ def branin3(x1, x2, x3, c1):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
index 1228b601..78e49837 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-False+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
@@ -66,7 +66,7 @@ def branin3(x1, x2, x3, c1):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
index 042e443b..05a0c4e7 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
@@ -54,7 +54,7 @@ def branin(x1, x2):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
index 3aaca7eb..c2d0820a 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
@@ -54,7 +54,7 @@ def branin(x1, x2):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
index d9c0a0aa..292171ca 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
@@ -58,7 +58,7 @@ def branin(x1, x2, c1):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
index 27bd005d..616a7f64 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
@@ -58,7 +58,7 @@ def branin(x1, x2, c1):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
index 9a9f328c..071e0164 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
@@ -63,7 +63,7 @@ def branin3(x1, x2, x3):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
index 6b246c85..2b041624 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
@@ -63,7 +63,7 @@ def branin3(x1, x2, x3):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
index 9765e0b1..be9ddc3a 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
@@ -69,7 +69,7 @@ def branin3(x1, x2, x3, c1):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
index 7a4dd575..1f2a6318 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-False+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
@@ -69,7 +69,7 @@ def branin3(x1, x2, x3, c1):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
index ba53594c..0acd4b36 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-batch.py
@@ -54,7 +54,7 @@ def branin(x1, x2):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
index 8669cc7c..56a6e616 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-False+custom_threshold-False+synchrony-single.py
@@ -54,7 +54,7 @@ def branin(x1, x2):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
index 0ad484a1..8f2b9d4e 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-batch.py
@@ -58,7 +58,7 @@ def branin(x1, x2, c1):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
index f18f9a05..ed435e0d 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-False+categorical-True+custom_threshold-False+synchrony-single.py
@@ -58,7 +58,7 @@ def branin(x1, x2, c1):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
index 0f32130a..bd1660a2 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-batch.py
@@ -63,7 +63,7 @@ def branin3(x1, x2, x3):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
index d1f65af7..a76e1051 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-False+custom_threshold-False+synchrony-single.py
@@ -63,7 +63,7 @@ def branin3(x1, x2, x3):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
index 90590018..413fac6d 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-batch.py
@@ -69,7 +69,7 @@ def branin3(x1, x2, x3, c1):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
index eb79d14d..6a66325c 100644
--- a/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
+++ b/docs/generated_scripts/ax/objective-single+model-FULLYBAYESIAN+custom_gen-True+existing_data-True+sum_constraint-True+order_constraint-True+linear_constraint-True+composition_constraint-True+categorical-True+custom_threshold-False+synchrony-single.py
@@ -69,7 +69,7 @@ def branin3(x1, x2, x3, c1):
             model=Models.FULLYBAYESIAN,
             num_trials=-1,
             max_parallelism=3,
-            model_kwargs={"num_samples": 256, "warmup_steps": 512},
+            model_kwargs={"num_samples": 1024, "warmup_steps": 1024},
         ),
     ]
 )
diff --git a/docs/honegumi.html b/docs/honegumi.html
index 6ad4b6a7..fb9d0b3e 100644
--- a/docs/honegumi.html
+++ b/docs/honegumi.html
@@ -10,7 +10,7 @@
 
 <body onload="updateText()">
     <script>
-        var optionRows = [{"hidden": false, "name": "objective", "options": ["single", "multi"], "tooltip": "Choose between \u003ca href=\u0027tutorials/sobo-vs-mobo/sobo-vs-mobo.html\u0027\u003esingle and multi-objective optimization\u003c/a\u003e based on your project needs. Single objective optimization targets one primary goal (e.g. maximize the strength of a material), while multi-objective optimization considers several objectives simultaneously (e.g. maximize the strength of a material while minimizing synthesis cost). Select the option that best aligns with your optimization goals and problem complexity."}, {"hidden": false, "name": "model", "options": ["GPEI", "FULLYBAYESIAN"], "tooltip": "Choose between frequentist and fully bayesian implementations of the gaussian process (GP) surrogate model. The frequentist GP model offers efficiency and speed, while the fully Bayesian GP models GP parameters as random variables through MCMC estimation, providing a deeper exploration of uncertainty. The fully bayesian treatment has historically provided better closed loop Bayesian optimization performance, but comes at the cost of higher computational demand. Consider your computational resources and the complexity of your optimization task when making your selection."}, {"hidden": false, "name": "existing_data", "options": ["False", "True"], "tooltip": "Choose whether to fit the surrogate model to previous data before starting the optimization process. Including historical data may give your model a better starting place and potentially speed up convergence. Conversely, excluding existing data means starting the optimization from scratch, which might be preferred in scenarios where historical data could introduce bias or noise into the optimization process. Consider the relevance and reliability of your existing data when making your selection."}, {"hidden": false, "name": "sum_constraint", "options": ["False", "True"], "tooltip": "Choose whether to apply a sum constraint over two or more optimization variables (e.g. ensuring total allocation remains within available budget). This constraint focusses generated optimization trials on feasible candidates at the cost of flexibility. Consider whether such a constraint reflects the reality of variable interactions when selecting this option."}, {"hidden": false, "name": "order_constraint", "options": ["False", "True"], "tooltip": "Choose whether to implement an order constraint over two or more optimization variables (e.g. ensuring certain tasks precede others). This constraint focusses generated optimization trials on variable combinations that follow a specific order. Excluding the constraint offers flexibility in variable arrangements but may neglect important task sequencing or value inequality considerations. Consider whether such a constraint reflects the reality of variable interactions when selecting this option."}, {"hidden": false, "name": "linear_constraint", "options": ["False", "True"], "tooltip": "Choose whether to implement a linear constraint over two or more optimization variables such that the linear combination of parameter values adheres to an inequality (e.g. 0.2*x_1 + x_2 \u003c 0.1). This constraint focusses generated optimization trials on variable combinations that follow an enforced rule at the cost of flexibility. Consider whether such a constraint reflects the reality of variable interactions when selecting this option."}, {"hidden": false, "name": "composition_constraint", "options": ["False", "True"], "tooltip": "Choose whether to include a composition constraint over two or more optimization variables such that their sum does not exceed a specified total (e.g. ensuring the mole fractions of elements in a composition sum to one). This constraint is particularly relevant to fabrication-related tasks where the quantities of components must sum to a total. Consider whether such a constraint reflects the reality of variable interactions when selecting this option."}, {"hidden": false, "name": "categorical", "options": ["False", "True"], "tooltip": "Choose whether to include a categorical variable in the optimization process (e.g. dark or milk chocolate chips in a cookie recipe). Including categorical variables allows choice parameters and their interaction with continuous variables to be optimized. Note that adding categorical variables can create discontinuities in the search space that are difficult to optimize over. Consider the value of adding categorical variables to the optimization task when selecting this option."}, {"hidden": false, "name": "custom_threshold", "options": ["False", "True"], "tooltip": "Choose whether to apply custom thresholds to objectives in a multi-objective optimization problem (e.g. a minimum acceptable strength requirement for a material). Setting a threshold on an objective guides the optimization algorithm to prioritize solutions that meet or exceed these criteria. Excluding thresholds enables greater exploration of the design space, but may produce sub-optimal solutions. Consider whether threshold values reflect the reality or expectations of your optimization task when selection this option."}, {"hidden": false, "name": "synchrony", "options": ["single", "batch"], "tooltip": "Choose whether to perform single or batch evaluations for your Bayesian optimization campaign. Single evaluations analyze one candidate solution at a time, offering precise control and adaptability after each trial at the expense of more compute time. Batch evaluations, however, process several solutions in parallel, significantly reducing the number of optimization cycles but potentially diluting the specificity of adjustments. Batch evaluation is helpful in scenarios where it is advantageous to test several solutions simultaneously. Consider the nature of your evaluation tool when selecting between the two options."}];
+        var optionRows = [{"hidden": false, "name": "objective", "options": ["single", "multi"], "tooltip": "Choose between \u003ca href=\u0027tutorials/sobo-vs-mobo/sobo-vs-mobo.html\u0027\u003esingle and multi-objective optimization\u003c/a\u003e based on your project needs. Single objective optimization targets one primary goal (e.g. maximize the strength of a material), while multi-objective optimization considers several objectives simultaneously (e.g. maximize the strength of a material while minimizing synthesis cost). Select the option that best aligns with your optimization goals and problem complexity."}, {"hidden": false, "name": "model", "options": ["GPEI", "FULLYBAYESIAN"], "tooltip": "Choose between \u003ca href=\u0027tutorials/freq-vs-bayes/freq-vs-bayes.html\u0027\u003efrequentist and fully bayesian\u003c/a\u003e implementations of the gaussian process (GP) surrogate model. The frequentist GP model, which is often the default in BO packages, offers efficiency and speed. The fully Bayesian GP models GP parameters as random variables through MCMC estimation, providing a deeper exploration of uncertainty. The fully bayesian treatment has historically provided better closed loop Bayesian optimization performance, but comes at the cost of higher computational demand. Consider your computational resources and the complexity of your optimization task when making your selection."}, {"hidden": false, "name": "existing_data", "options": ["False", "True"], "tooltip": "Choose whether to fit the surrogate model to previous data before starting the optimization process. Including historical data may give your model a better starting place and potentially speed up convergence. Conversely, excluding existing data means starting the optimization from scratch, which might be preferred in scenarios where historical data could introduce bias or noise into the optimization process. Consider the relevance and reliability of your existing data when making your selection."}, {"hidden": false, "name": "sum_constraint", "options": ["False", "True"], "tooltip": "Choose whether to apply a sum constraint over two or more optimization variables (e.g. ensuring total allocation remains within available budget). This constraint focusses generated optimization trials on feasible candidates at the cost of flexibility. Consider whether such a constraint reflects the reality of variable interactions when selecting this option."}, {"hidden": false, "name": "order_constraint", "options": ["False", "True"], "tooltip": "Choose whether to implement an order constraint over two or more optimization variables (e.g. ensuring certain tasks precede others). This constraint focusses generated optimization trials on variable combinations that follow a specific order. Excluding the constraint offers flexibility in variable arrangements but may neglect important task sequencing or value inequality considerations. Consider whether such a constraint reflects the reality of variable interactions when selecting this option."}, {"hidden": false, "name": "linear_constraint", "options": ["False", "True"], "tooltip": "Choose whether to implement a linear constraint over two or more optimization variables such that the linear combination of parameter values adheres to an inequality (e.g. 0.2*x_1 + x_2 \u003c 0.1). This constraint focusses generated optimization trials on variable combinations that follow an enforced rule at the cost of flexibility. Consider whether such a constraint reflects the reality of variable interactions when selecting this option."}, {"hidden": false, "name": "composition_constraint", "options": ["False", "True"], "tooltip": "Choose whether to include a composition constraint over two or more optimization variables such that their sum does not exceed a specified total (e.g. ensuring the mole fractions of elements in a composition sum to one). This constraint is particularly relevant to fabrication-related tasks where the quantities of components must sum to a total. Consider whether such a constraint reflects the reality of variable interactions when selecting this option."}, {"hidden": false, "name": "categorical", "options": ["False", "True"], "tooltip": "Choose whether to include a categorical variable in the optimization process (e.g. dark or milk chocolate chips in a cookie recipe). Including categorical variables allows choice parameters and their interaction with continuous variables to be optimized. Note that adding categorical variables can create discontinuities in the search space that are difficult to optimize over. Consider the value of adding categorical variables to the optimization task when selecting this option."}, {"hidden": false, "name": "custom_threshold", "options": ["False", "True"], "tooltip": "Choose whether to apply custom thresholds to objectives in a multi-objective optimization problem (e.g. a minimum acceptable strength requirement for a material). Setting a threshold on an objective guides the optimization algorithm to prioritize solutions that meet or exceed these criteria. Excluding thresholds enables greater exploration of the design space, but may produce sub-optimal solutions. Consider whether threshold values reflect the reality or expectations of your optimization task when selection this option."}, {"hidden": false, "name": "synchrony", "options": ["single", "batch"], "tooltip": "Choose whether to perform single or batch evaluations for your Bayesian optimization campaign. Single evaluations analyze one candidate solution at a time, offering precise control and adaptability after each trial at the expense of more compute time. Batch evaluations, however, process several solutions in parallel, significantly reducing the number of optimization cycles but potentially diluting the specificity of adjustments. Batch evaluation is helpful in scenarios where it is advantageous to test several solutions simultaneously. Consider the nature of your evaluation tool when selecting between the two options."}];
         // E.g.,
         // var optionRows = [
         //     { "name": "row1", "options": ["A", "B"] },
@@ -18,7 +18,7 @@
         //     { "name": "row3", "options": ["F", "G"] }
         // ];
 
-        var scriptLookup = {"multi,FULLYBAYESIAN,False,False,False,False,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,False,False,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,False,False,False,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,False,False,False,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,False,False,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,False,False,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,False,False,False,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,False,False,False,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,False,False,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,False,False,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,False,False,True,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,False,False,True,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,False,False,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,False,False,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,False,False,True,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,False,False,True,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,False,True,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,False,True,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,False,True,False,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,False,True,False,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,False,True,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,False,True,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,False,True,False,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,False,True,False,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,False,True,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,False,True,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,False,True,True,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,False,True,True,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,False,True,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,False,True,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,False,True,True,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,False,True,True,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,True,False,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,True,False,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,True,False,False,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,True,False,False,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,True,False,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,True,False,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,True,False,False,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,True,False,False,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,True,False,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,True,False,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,True,False,True,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,True,False,True,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,True,False,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,True,False,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,True,False,True,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,True,False,True,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,True,True,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,True,True,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,True,True,False,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,True,True,False,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,True,True,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,True,True,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,True,True,False,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,True,True,False,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,True,True,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,True,True,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,True,True,True,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,True,True,True,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,True,True,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,True,True,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,True,True,True,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,True,True,True,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,False,False,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,False,False,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,False,False,False,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,False,False,False,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,False,False,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,False,False,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,False,False,False,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,False,False,False,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,False,False,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,False,False,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,False,False,True,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,False,False,True,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,False,False,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,False,False,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,False,False,True,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,False,False,True,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,False,True,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,False,True,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,False,True,False,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,False,True,False,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,False,True,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,False,True,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,False,True,False,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,False,True,False,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,False,True,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,False,True,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,False,True,True,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,False,True,True,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,False,True,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,False,True,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,False,True,True,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,False,True,True,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,True,False,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,True,False,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,True,False,False,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,True,False,False,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,True,False,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,True,False,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,True,False,False,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,True,False,False,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,True,False,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,True,False,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,True,False,True,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,True,False,True,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,True,False,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,True,False,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,True,False,True,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,True,False,True,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,True,True,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,True,True,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,True,True,False,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,True,True,False,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,True,True,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,True,True,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,True,True,False,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,True,True,False,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,True,True,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,True,True,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,True,True,True,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,True,True,True,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,True,True,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,True,True,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,True,True,True,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,True,True,True,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,False,False,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,False,False,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,False,False,False,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,False,False,False,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,False,False,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,False,False,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,False,False,False,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,False,False,False,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,False,False,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,False,False,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,False,False,True,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,False,False,True,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,False,False,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,False,False,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,False,False,True,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,False,False,True,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,False,True,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,False,True,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,False,True,False,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,False,True,False,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,False,True,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,False,True,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,False,True,False,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,False,True,False,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,False,True,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,False,True,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,False,True,True,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,False,True,True,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,False,True,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,False,True,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,False,True,True,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,False,True,True,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,True,False,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,True,False,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,True,False,False,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,True,False,False,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,True,False,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,True,False,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,True,False,False,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,True,False,False,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,True,False,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,True,False,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,True,False,True,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,True,False,True,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,True,False,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,True,False,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,True,False,True,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,True,False,True,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,True,True,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,True,True,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,True,True,False,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,True,True,False,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,True,True,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,True,True,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,True,True,False,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,True,True,False,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,True,True,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,True,True,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,True,True,True,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,True,True,True,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,True,True,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,True,True,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,True,True,True,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,True,True,True,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,False,False,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,False,False,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,False,False,False,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,False,False,False,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,False,False,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,False,False,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,False,False,False,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,False,False,False,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,False,False,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,False,False,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,False,False,True,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,False,False,True,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,False,False,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,False,False,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,False,False,True,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,False,False,True,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,False,True,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,False,True,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,False,True,False,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,False,True,False,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,False,True,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,False,True,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,False,True,False,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,False,True,False,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,False,True,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,False,True,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,False,True,True,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,False,True,True,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,False,True,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,False,True,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,False,True,True,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,False,True,True,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,True,False,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,True,False,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,True,False,False,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,True,False,False,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,True,False,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,True,False,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,True,False,False,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,True,False,False,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,True,False,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,True,False,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,True,False,True,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,True,False,True,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,True,False,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,True,False,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,True,False,True,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,True,False,True,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,True,True,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,True,True,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,True,True,False,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,True,True,False,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,True,True,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,True,True,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,True,True,False,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,True,True,False,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,True,True,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,True,True,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,True,True,True,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,True,True,True,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,True,True,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,True,True,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,True,True,True,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,True,True,True,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,False,False,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,False,False,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,False,False,False,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,False,False,False,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,False,False,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,False,False,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,False,False,False,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,False,False,False,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,False,False,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,False,False,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,False,False,True,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,False,False,True,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,False,False,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,False,False,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,False,False,True,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,False,False,True,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,False,True,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,False,True,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,False,True,False,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,False,True,False,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,False,True,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,False,True,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,False,True,False,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,False,True,False,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,False,True,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,False,True,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,False,True,True,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,False,True,True,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,False,True,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,False,True,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,False,True,True,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,False,True,True,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,True,False,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,True,False,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,True,False,False,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,True,False,False,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,True,False,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,True,False,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,True,False,False,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,True,False,False,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,True,False,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,True,False,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,True,False,True,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,True,False,True,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,True,False,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,True,False,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,True,False,True,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,True,False,True,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,True,True,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,True,True,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,True,True,False,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,True,True,False,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,True,True,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,True,True,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,True,True,False,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,True,True,False,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,True,True,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,True,True,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,True,True,True,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,True,True,True,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,True,True,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,True,True,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,True,True,True,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,True,True,True,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,False,False,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,False,False,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,False,False,False,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,False,False,False,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,False,False,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,False,False,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,False,False,False,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,False,False,False,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,False,False,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,False,False,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,False,False,True,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,False,False,True,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,False,False,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,False,False,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,False,False,True,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,False,False,True,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,False,True,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,False,True,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,False,True,False,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,False,True,False,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,False,True,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,False,True,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,False,True,False,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,False,True,False,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,False,True,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,False,True,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,False,True,True,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,False,True,True,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,False,True,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,False,True,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,False,True,True,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,False,True,True,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,True,False,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,True,False,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,True,False,False,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,True,False,False,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,True,False,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,True,False,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,True,False,False,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,True,False,False,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,True,False,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,True,False,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,True,False,True,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,True,False,True,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,True,False,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,True,False,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,True,False,True,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,True,False,True,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,True,True,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,True,True,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,True,True,False,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,True,True,False,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,True,True,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,True,True,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,True,True,False,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,True,True,False,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,True,True,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,True,True,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,True,True,True,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,True,True,True,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,True,True,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,True,True,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,True,True,True,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,True,True,True,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,False,False,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,False,False,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,False,False,False,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,False,False,False,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,False,False,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,False,False,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,False,False,False,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,False,False,False,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,False,False,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,False,False,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,False,False,True,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,False,False,True,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,False,False,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,False,False,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,False,False,True,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,False,False,True,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,False,True,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,False,True,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,False,True,False,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,False,True,False,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,False,True,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,False,True,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,False,True,False,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,False,True,False,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,False,True,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,False,True,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,False,True,True,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,False,True,True,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,False,True,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,False,True,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,False,True,True,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,False,True,True,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,True,False,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,True,False,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,True,False,False,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,True,False,False,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,True,False,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,True,False,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,True,False,False,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,True,False,False,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,True,False,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,True,False,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,True,False,True,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,True,False,True,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,True,False,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,True,False,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,True,False,True,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,True,False,True,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,True,True,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,True,True,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,True,True,False,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,True,True,False,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,True,True,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,True,True,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,True,True,False,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,True,True,False,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,True,True,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,True,True,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,True,True,True,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,True,True,True,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,True,True,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,True,True,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,True,True,True,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,True,True,True,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,False,False,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,False,False,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,False,False,False,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,False,False,False,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,False,False,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,False,False,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,False,False,False,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,False,False,False,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,False,False,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,False,False,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,False,False,True,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,False,False,True,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,False,False,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,False,False,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,False,False,True,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,False,False,True,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,False,True,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,False,True,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,False,True,False,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,False,True,False,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,False,True,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,False,True,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,False,True,False,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,False,True,False,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,False,True,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,False,True,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,False,True,True,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,False,True,True,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,False,True,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,False,True,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,False,True,True,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,False,True,True,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,True,False,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,True,False,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,True,False,False,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,True,False,False,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,True,False,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,True,False,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,True,False,False,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,True,False,False,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,True,False,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,True,False,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,True,False,True,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,True,False,True,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,True,False,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,True,False,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,True,False,True,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,True,False,True,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,True,True,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,True,True,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,True,True,False,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,True,True,False,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,True,True,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,True,True,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,True,True,False,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,True,True,False,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,True,True,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,True,True,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,True,True,True,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,True,True,True,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,True,True,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,True,True,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,True,True,True,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,True,True,True,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "single,FULLYBAYESIAN,False,False,False,False,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,False,False,False,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,False,False,False,False,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,False,False,False,False,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,False,False,False,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,False,False,False,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,False,False,False,False,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,False,False,False,False,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,False,False,False,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,False,False,False,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,False,False,False,True,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,False,False,False,True,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,False,False,False,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,False,False,False,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,False,False,False,True,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,False,False,False,True,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,False,False,True,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,False,False,True,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,False,False,True,False,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,False,False,True,False,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,False,False,True,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,False,False,True,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,False,False,True,False,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,False,False,True,False,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,False,False,True,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,False,False,True,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,False,False,True,True,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,False,False,True,True,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,False,False,True,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,False,False,True,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,False,False,True,True,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,False,False,True,True,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,False,True,False,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,False,True,False,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,False,True,False,False,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,False,True,False,False,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,False,True,False,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,False,True,False,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,False,True,False,False,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,False,True,False,False,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,False,True,False,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,False,True,False,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,False,True,False,True,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,False,True,False,True,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,False,True,False,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,False,True,False,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,False,True,False,True,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,False,True,False,True,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,False,True,True,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,False,True,True,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,False,True,True,False,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,False,True,True,False,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,False,True,True,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,False,True,True,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,False,True,True,False,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,False,True,True,False,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,False,True,True,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,False,True,True,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,False,True,True,True,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,False,True,True,True,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,False,True,True,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,False,True,True,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,False,True,True,True,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,False,True,True,True,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,True,False,False,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,True,False,False,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,True,False,False,False,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,True,False,False,False,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,True,False,False,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,True,False,False,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,True,False,False,False,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,True,False,False,False,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,True,False,False,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,True,False,False,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,True,False,False,True,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,True,False,False,True,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,True,False,False,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,True,False,False,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,True,False,False,True,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,True,False,False,True,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,True,False,True,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,True,False,True,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,True,False,True,False,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,True,False,True,False,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,True,False,True,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,True,False,True,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,True,False,True,False,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,True,False,True,False,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,True,False,True,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,True,False,True,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,True,False,True,True,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,True,False,True,True,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,True,False,True,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,True,False,True,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,True,False,True,True,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,True,False,True,True,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,True,True,False,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,True,True,False,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,True,True,False,False,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,True,True,False,False,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,True,True,False,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,True,True,False,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,True,True,False,False,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,True,True,False,False,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,True,True,False,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,True,True,False,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,True,True,False,True,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,True,True,False,True,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,True,True,False,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,True,True,False,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,True,True,False,True,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,True,True,False,True,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,True,True,True,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,True,True,True,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,True,True,True,False,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,True,True,True,False,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,True,True,True,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,True,True,True,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,True,True,True,False,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,True,True,True,False,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,True,True,True,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,True,True,True,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,True,True,True,True,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,True,True,True,True,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,True,True,True,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,True,True,True,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,True,True,True,True,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,True,True,True,True,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,False,False,False,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,False,False,False,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,False,False,False,False,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,False,False,False,False,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,False,False,False,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,False,False,False,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,False,False,False,False,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,False,False,False,False,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,False,False,False,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,False,False,False,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,False,False,False,True,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,False,False,False,True,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,False,False,False,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,False,False,False,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,False,False,False,True,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,False,False,False,True,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,False,False,True,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,False,False,True,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,False,False,True,False,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,False,False,True,False,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,False,False,True,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,False,False,True,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,False,False,True,False,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,False,False,True,False,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,False,False,True,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,False,False,True,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,False,False,True,True,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,False,False,True,True,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,False,False,True,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,False,False,True,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,False,False,True,True,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,False,False,True,True,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,False,True,False,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,False,True,False,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,False,True,False,False,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,False,True,False,False,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,False,True,False,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,False,True,False,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,False,True,False,False,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,False,True,False,False,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,False,True,False,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,False,True,False,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,False,True,False,True,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,False,True,False,True,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,False,True,False,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,False,True,False,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,False,True,False,True,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,False,True,False,True,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,False,True,True,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,False,True,True,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,False,True,True,False,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,False,True,True,False,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,False,True,True,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,False,True,True,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,False,True,True,False,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,False,True,True,False,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,False,True,True,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,False,True,True,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,False,True,True,True,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,False,True,True,True,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,False,True,True,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,False,True,True,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,False,True,True,True,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,False,True,True,True,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,True,False,False,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,True,False,False,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,True,False,False,False,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,True,False,False,False,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,True,False,False,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,True,False,False,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,True,False,False,False,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,True,False,False,False,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,True,False,False,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,True,False,False,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,True,False,False,True,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,True,False,False,True,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,True,False,False,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,True,False,False,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,True,False,False,True,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,True,False,False,True,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,True,False,True,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,True,False,True,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,True,False,True,False,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,True,False,True,False,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,True,False,True,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,True,False,True,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,True,False,True,False,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,True,False,True,False,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,True,False,True,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,True,False,True,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,True,False,True,True,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,True,False,True,True,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,True,False,True,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,True,False,True,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,True,False,True,True,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,True,False,True,True,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,True,True,False,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,True,True,False,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,True,True,False,False,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,True,True,False,False,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,True,True,False,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,True,True,False,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,True,True,False,False,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,True,True,False,False,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,True,True,False,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,True,True,False,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,True,True,False,True,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,True,True,False,True,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,True,True,False,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,True,True,False,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,True,True,False,True,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,True,True,False,True,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,True,True,True,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,True,True,True,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,True,True,True,False,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,True,True,True,False,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,True,True,True,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,True,True,True,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,True,True,True,False,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,True,True,True,False,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,True,True,True,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,True,True,True,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,True,True,True,True,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,True,True,True,True,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,True,True,True,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,True,True,True,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 256, \"warmup_steps\": 512},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,True,True,True,True,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,True,True,True,True,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,False,False,False,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,False,False,False,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,False,False,False,False,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,False,False,False,False,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,False,False,False,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,False,False,False,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,False,False,False,False,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,False,False,False,False,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,False,False,False,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,False,False,False,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,False,False,False,True,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,False,False,False,True,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,False,False,False,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,False,False,False,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,False,False,False,True,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,False,False,False,True,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,False,False,True,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,False,False,True,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,False,False,True,False,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,False,False,True,False,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,False,False,True,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,False,False,True,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,False,False,True,False,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,False,False,True,False,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,False,False,True,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,False,False,True,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,False,False,True,True,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,False,False,True,True,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,False,False,True,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,False,False,True,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,False,False,True,True,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,False,False,True,True,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,False,True,False,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,False,True,False,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,False,True,False,False,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,False,True,False,False,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,False,True,False,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,False,True,False,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,False,True,False,False,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,False,True,False,False,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,False,True,False,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,False,True,False,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,False,True,False,True,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,False,True,False,True,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,False,True,False,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,False,True,False,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,False,True,False,True,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,False,True,False,True,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,False,True,True,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,False,True,True,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,False,True,True,False,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,False,True,True,False,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,False,True,True,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,False,True,True,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,False,True,True,False,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,False,True,True,False,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,False,True,True,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,False,True,True,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,False,True,True,True,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,False,True,True,True,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,False,True,True,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,False,True,True,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,False,True,True,True,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,False,True,True,True,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,True,False,False,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,True,False,False,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,True,False,False,False,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,True,False,False,False,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,True,False,False,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,True,False,False,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,True,False,False,False,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,True,False,False,False,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,True,False,False,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,True,False,False,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,True,False,False,True,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,True,False,False,True,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,True,False,False,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,True,False,False,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,True,False,False,True,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,True,False,False,True,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,True,False,True,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,True,False,True,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,True,False,True,False,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,True,False,True,False,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,True,False,True,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,True,False,True,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,True,False,True,False,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,True,False,True,False,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,True,False,True,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,True,False,True,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,True,False,True,True,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,True,False,True,True,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,True,False,True,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,True,False,True,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,True,False,True,True,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,True,False,True,True,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,True,True,False,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,True,True,False,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,True,True,False,False,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,True,True,False,False,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,True,True,False,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,True,True,False,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,True,True,False,False,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,True,True,False,False,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,True,True,False,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,True,True,False,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,True,True,False,True,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,True,True,False,True,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,True,True,False,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,True,True,False,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,True,True,False,True,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,True,True,False,True,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,True,True,True,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,True,True,True,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,True,True,True,False,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,True,True,True,False,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,True,True,True,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,True,True,True,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,True,True,True,False,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,True,True,True,False,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,True,True,True,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,True,True,True,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,True,True,True,True,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,True,True,True,True,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,True,True,True,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,True,True,True,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,True,True,True,True,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,True,True,True,True,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,False,False,False,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,False,False,False,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,False,False,False,False,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,False,False,False,False,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,False,False,False,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,False,False,False,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,False,False,False,False,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,False,False,False,False,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,False,False,False,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,False,False,False,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,False,False,False,True,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,False,False,False,True,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,False,False,False,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,False,False,False,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,False,False,False,True,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,False,False,False,True,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,False,False,True,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,False,False,True,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,False,False,True,False,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,False,False,True,False,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,False,False,True,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,False,False,True,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,False,False,True,False,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,False,False,True,False,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,False,False,True,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,False,False,True,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,False,False,True,True,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,False,False,True,True,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,False,False,True,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,False,False,True,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,False,False,True,True,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,False,False,True,True,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,False,True,False,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,False,True,False,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,False,True,False,False,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,False,True,False,False,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,False,True,False,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,False,True,False,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,False,True,False,False,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,False,True,False,False,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,False,True,False,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,False,True,False,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,False,True,False,True,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,False,True,False,True,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,False,True,False,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,False,True,False,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,False,True,False,True,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,False,True,False,True,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,False,True,True,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,False,True,True,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,False,True,True,False,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,False,True,True,False,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,False,True,True,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,False,True,True,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,False,True,True,False,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,False,True,True,False,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,False,True,True,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,False,True,True,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,False,True,True,True,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,False,True,True,True,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,False,True,True,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,False,True,True,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,False,True,True,True,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,False,True,True,True,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,True,False,False,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,True,False,False,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,True,False,False,False,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,True,False,False,False,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,True,False,False,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,True,False,False,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,True,False,False,False,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,True,False,False,False,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,True,False,False,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,True,False,False,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,True,False,False,True,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,True,False,False,True,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,True,False,False,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,True,False,False,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,True,False,False,True,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,True,False,False,True,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,True,False,True,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,True,False,True,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,True,False,True,False,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,True,False,True,False,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,True,False,True,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,True,False,True,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,True,False,True,False,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,True,False,True,False,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,True,False,True,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,True,False,True,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,True,False,True,True,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,True,False,True,True,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,True,False,True,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,True,False,True,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,True,False,True,True,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,True,False,True,True,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,True,True,False,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,True,True,False,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,True,True,False,False,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,True,True,False,False,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,True,True,False,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,True,True,False,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,True,True,False,False,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,True,True,False,False,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,True,True,False,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,True,True,False,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,True,True,False,True,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,True,True,False,True,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,True,True,False,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,True,True,False,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,True,True,False,True,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,True,True,False,True,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,True,True,True,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,True,True,True,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,True,True,True,False,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,True,True,True,False,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,True,True,True,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,True,True,True,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,True,True,True,False,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,True,True,True,False,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,True,True,True,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,True,True,True,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,True,True,True,True,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,True,True,True,True,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,True,True,True,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,True,True,True,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,True,True,True,True,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,True,True,True,True,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent."};
+        var scriptLookup = {"multi,FULLYBAYESIAN,False,False,False,False,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,False,False,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,False,False,False,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,False,False,False,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,False,False,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,False,False,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,False,False,False,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,False,False,False,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,False,False,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,False,False,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,False,False,True,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,False,False,True,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,False,False,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,False,False,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,False,False,True,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,False,False,True,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,False,True,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,False,True,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,False,True,False,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,False,True,False,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,False,True,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,False,True,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,False,True,False,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,False,True,False,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,False,True,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,False,True,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,False,True,True,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,False,True,True,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,False,True,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,False,True,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,False,True,True,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,False,True,True,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,True,False,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,True,False,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,True,False,False,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,True,False,False,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,True,False,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,True,False,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,True,False,False,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,True,False,False,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,True,False,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,True,False,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,True,False,True,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,True,False,True,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,True,False,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,True,False,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,True,False,True,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,True,False,True,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,True,True,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,True,True,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,True,True,False,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,True,True,False,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,True,True,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,True,True,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,True,True,False,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,True,True,False,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,True,True,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,True,True,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,True,True,True,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,True,True,True,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,True,True,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,True,True,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,True,True,True,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,False,True,True,True,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,False,False,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,False,False,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,False,False,False,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,False,False,False,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,False,False,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,False,False,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,False,False,False,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,False,False,False,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,False,False,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,False,False,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,False,False,True,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,False,False,True,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,False,False,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,False,False,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,False,False,True,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,False,False,True,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,False,True,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,False,True,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,False,True,False,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,False,True,False,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,False,True,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,False,True,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,False,True,False,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,False,True,False,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,False,True,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,False,True,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,False,True,True,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,False,True,True,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,False,True,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,False,True,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,False,True,True,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,False,True,True,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,True,False,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,True,False,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,True,False,False,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,True,False,False,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,True,False,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,True,False,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,True,False,False,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,True,False,False,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,True,False,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,True,False,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,True,False,True,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,True,False,True,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,True,False,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,True,False,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,True,False,True,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,True,False,True,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,True,True,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,True,True,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,True,True,False,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,True,True,False,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,True,True,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,True,True,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,True,True,False,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,True,True,False,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,True,True,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,True,True,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,True,True,True,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,True,True,True,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,True,True,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,True,True,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,True,True,True,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,False,True,True,True,True,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,False,False,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,False,False,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,False,False,False,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,False,False,False,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,False,False,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,False,False,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,False,False,False,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,False,False,False,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,False,False,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,False,False,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,False,False,True,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,False,False,True,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,False,False,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,False,False,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,False,False,True,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,False,False,True,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,False,True,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,False,True,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,False,True,False,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,False,True,False,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,False,True,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,False,True,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,False,True,False,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,False,True,False,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,False,True,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,False,True,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,False,True,True,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,False,True,True,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,False,True,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,False,True,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,False,True,True,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,False,True,True,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,True,False,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,True,False,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,True,False,False,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,True,False,False,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,True,False,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,True,False,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,True,False,False,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,True,False,False,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,True,False,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,True,False,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,True,False,True,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,True,False,True,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,True,False,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,True,False,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,True,False,True,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,True,False,True,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,True,True,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,True,True,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,True,True,False,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,True,True,False,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,True,True,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,True,True,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,True,True,False,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,True,True,False,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,True,True,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,True,True,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,True,True,True,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,True,True,True,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,True,True,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,True,True,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,True,True,True,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,False,True,True,True,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,False,False,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,False,False,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,False,False,False,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,False,False,False,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,False,False,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,False,False,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,False,False,False,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,False,False,False,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,False,False,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,False,False,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,False,False,True,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,False,False,True,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,False,False,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,False,False,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,False,False,True,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,False,False,True,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,False,True,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,False,True,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,False,True,False,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,False,True,False,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,False,True,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,False,True,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,False,True,False,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,False,True,False,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,False,True,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,False,True,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,False,True,True,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,False,True,True,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,False,True,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,False,True,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,False,True,True,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,False,True,True,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,True,False,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,True,False,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,True,False,False,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,True,False,False,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,True,False,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,True,False,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,True,False,False,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,True,False,False,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,True,False,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,True,False,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,True,False,True,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,True,False,True,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,True,False,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,True,False,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,True,False,True,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,True,False,True,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,True,True,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,True,True,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,True,True,False,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,True,True,False,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,True,True,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,True,True,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,True,True,False,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,True,True,False,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,True,True,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,True,True,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,True,True,True,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,True,True,True,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,True,True,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,True,True,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,True,True,True,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,FULLYBAYESIAN,True,True,True,True,True,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIANMOO,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,False,False,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,False,False,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,False,False,False,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,False,False,False,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,False,False,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,False,False,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,False,False,False,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,False,False,False,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,False,False,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,False,False,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,False,False,True,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,False,False,True,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,False,False,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,False,False,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,False,False,True,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,False,False,True,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,False,True,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,False,True,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,False,True,False,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,False,True,False,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,False,True,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,False,True,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,False,True,False,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,False,True,False,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,False,True,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,False,True,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,False,True,True,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,False,True,True,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,False,True,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,False,True,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,False,True,True,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,False,True,True,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,True,False,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,True,False,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,True,False,False,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,True,False,False,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,True,False,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,True,False,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,True,False,False,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,True,False,False,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,True,False,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,True,False,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,True,False,True,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,True,False,True,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,True,False,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,True,False,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,True,False,True,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,True,False,True,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,True,True,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,True,True,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,True,True,False,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,True,True,False,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,True,True,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,True,True,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,True,True,False,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,True,True,False,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,True,True,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,True,True,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,True,True,True,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,True,True,True,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,True,True,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,True,True,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,True,True,True,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,False,True,True,True,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,False,False,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,False,False,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,False,False,False,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,False,False,False,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,False,False,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,False,False,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,False,False,False,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,False,False,False,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,False,False,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,False,False,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,False,False,True,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,False,False,True,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,False,False,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,False,False,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,False,False,True,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,False,False,True,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,False,True,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,False,True,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,False,True,False,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,False,True,False,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,False,True,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,False,True,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,False,True,False,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,False,True,False,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,False,True,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,False,True,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,False,True,True,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,False,True,True,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,False,True,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,False,True,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,False,True,True,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,False,True,True,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,True,False,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,True,False,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,True,False,False,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,True,False,False,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,True,False,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,True,False,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,True,False,False,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,True,False,False,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,True,False,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,True,False,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,True,False,True,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,True,False,True,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,True,False,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,True,False,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,True,False,True,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,True,False,True,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,True,True,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,True,True,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,True,True,False,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,True,True,False,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,True,True,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,True,True,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,True,True,False,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,True,True,False,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,True,True,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,True,True,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,True,True,True,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,True,True,True,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,True,True,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,True,True,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,True,True,True,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,False,True,True,True,True,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,False,False,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,False,False,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,False,False,False,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,False,False,False,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,False,False,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,False,False,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,False,False,False,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,False,False,False,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,False,False,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,False,False,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,False,False,True,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,False,False,True,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,False,False,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,False,False,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,False,False,True,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,False,False,True,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,False,True,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,False,True,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,False,True,False,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,False,True,False,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,False,True,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,False,True,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,False,True,False,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,False,True,False,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,False,True,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,False,True,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,False,True,True,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,False,True,True,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,False,True,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,False,True,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,False,True,True,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,False,True,True,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,True,False,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,True,False,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,True,False,False,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,True,False,False,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,True,False,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,True,False,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,True,False,False,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,True,False,False,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,True,False,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,True,False,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,True,False,True,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,True,False,True,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,True,False,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,True,False,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,True,False,True,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,True,False,True,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,True,True,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,True,True,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,True,True,False,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,True,True,False,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,True,True,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,True,True,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,True,True,False,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,True,True,False,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,True,True,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,True,True,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,True,True,True,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,True,True,True,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,True,True,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,True,True,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,True,True,True,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,False,True,True,True,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,False,False,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,False,False,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,False,False,False,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,False,False,False,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,False,False,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,False,False,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,False,False,False,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,False,False,False,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,False,False,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,False,False,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,False,False,True,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,False,False,True,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,False,False,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,False,False,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,False,False,True,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,False,False,True,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,False,True,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,False,True,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,False,True,False,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,False,True,False,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,False,True,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,False,True,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,False,True,False,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,False,True,False,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,False,True,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,False,True,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,False,True,True,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,False,True,True,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,False,True,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,False,True,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,False,True,True,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,False,True,True,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,True,False,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,True,False,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,True,False,False,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,True,False,False,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,True,False,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,True,False,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,True,False,False,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,True,False,False,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,True,False,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,True,False,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,True,False,True,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,True,False,True,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,True,False,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,True,False,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,True,False,True,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,True,False,True,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,True,True,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,True,True,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,True,True,False,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin_moo(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,True,True,False,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin_moo(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,True,True,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,True,True,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,True,True,False,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin_moo(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,True,True,False,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin_moo(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin_moo(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin_moo(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,True,True,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,True,True,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,True,True,True,False,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3_moo(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,True,True,True,False,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3_moo(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,True,True,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,True,True,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n        obj2_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,True,True,True,True,True,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3_moo(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "multi,GPEI,True,True,True,True,True,True,True,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\nobj2_name = \"branin_swapped\"\n\n\ndef branin3_moo(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    # second objective has x1 and x2 swapped\n    y2 = float(\n        (x1 - 5.1 / (4 * np.pi**2) * x2**2 + 5.0 / np.pi * x2 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x2)\n        + 10\n    )\n    # Contrived way to incorporate x3 into the second objective\n    y2 = y2 * (1 - 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 0.0, \"B\": 2.0, \"C\": 1.0}\n    y2 += penalty_lookup[c1]\n\n    return {obj1_name: y, obj2_name: y2}\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3_moo(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"])\n    for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True, threshold=25.0),\n        obj2_name: ObjectiveProperties(minimize=True, threshold=15.0),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3_moo(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\npareto_results = ax_client.get_pareto_optimal_parameters()\n", "single,FULLYBAYESIAN,False,False,False,False,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,False,False,False,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,False,False,False,False,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,False,False,False,False,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,False,False,False,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,False,False,False,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,False,False,False,False,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,False,False,False,False,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,False,False,False,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,False,False,False,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,False,False,False,True,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,False,False,False,True,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,False,False,False,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,False,False,False,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,False,False,False,True,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,False,False,False,True,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,False,False,True,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,False,False,True,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,False,False,True,False,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,False,False,True,False,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,False,False,True,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,False,False,True,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,False,False,True,False,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,False,False,True,False,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,False,False,True,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,False,False,True,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,False,False,True,True,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,False,False,True,True,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,False,False,True,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,False,False,True,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,False,False,True,True,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,False,False,True,True,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,False,True,False,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,False,True,False,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,False,True,False,False,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,False,True,False,False,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,False,True,False,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,False,True,False,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,False,True,False,False,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,False,True,False,False,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,False,True,False,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,False,True,False,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,False,True,False,True,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,False,True,False,True,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,False,True,False,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,False,True,False,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,False,True,False,True,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,False,True,False,True,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,False,True,True,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,False,True,True,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,False,True,True,False,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,False,True,True,False,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,False,True,True,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,False,True,True,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,False,True,True,False,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,False,True,True,False,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,False,True,True,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,False,True,True,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,False,True,True,True,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,False,True,True,True,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,False,True,True,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,False,True,True,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,False,True,True,True,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,False,True,True,True,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,True,False,False,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,True,False,False,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,True,False,False,False,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,True,False,False,False,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,True,False,False,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,True,False,False,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,True,False,False,False,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,True,False,False,False,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,True,False,False,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,True,False,False,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,True,False,False,True,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,True,False,False,True,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,True,False,False,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,True,False,False,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,True,False,False,True,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,True,False,False,True,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,True,False,True,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,True,False,True,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,True,False,True,False,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,True,False,True,False,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,True,False,True,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,True,False,True,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,True,False,True,False,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,True,False,True,False,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,True,False,True,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,True,False,True,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,True,False,True,True,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,True,False,True,True,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,True,False,True,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,True,False,True,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,True,False,True,True,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,True,False,True,True,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,True,True,False,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,True,True,False,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,True,True,False,False,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,True,True,False,False,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,True,True,False,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,True,True,False,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,True,True,False,False,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,True,True,False,False,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,True,True,False,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,True,True,False,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,True,True,False,True,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,True,True,False,True,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,True,True,False,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,True,True,False,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,True,True,False,True,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,True,True,False,True,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,True,True,True,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,True,True,True,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,True,True,True,False,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,True,True,True,False,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,True,True,True,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,True,True,True,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,True,True,True,False,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,True,True,True,False,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,True,True,True,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,True,True,True,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,True,True,True,True,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,True,True,True,True,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,True,True,True,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,True,True,True,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,False,True,True,True,True,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,False,True,True,True,True,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,False,False,False,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,False,False,False,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,False,False,False,False,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,False,False,False,False,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,False,False,False,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,False,False,False,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,False,False,False,False,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,False,False,False,False,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,False,False,False,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,False,False,False,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,False,False,False,True,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,False,False,False,True,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,False,False,False,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,False,False,False,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,False,False,False,True,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,False,False,False,True,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,False,False,True,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,False,False,True,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,False,False,True,False,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,False,False,True,False,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,False,False,True,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,False,False,True,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,False,False,True,False,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,False,False,True,False,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,False,False,True,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,False,False,True,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,False,False,True,True,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,False,False,True,True,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,False,False,True,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,False,False,True,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,False,False,True,True,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,False,False,True,True,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,False,True,False,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,False,True,False,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,False,True,False,False,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,False,True,False,False,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,False,True,False,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,False,True,False,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,False,True,False,False,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,False,True,False,False,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,False,True,False,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,False,True,False,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,False,True,False,True,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,False,True,False,True,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,False,True,False,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,False,True,False,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,False,True,False,True,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,False,True,False,True,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,False,True,True,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,False,True,True,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,False,True,True,False,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,False,True,True,False,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,False,True,True,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,False,True,True,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,False,True,True,False,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,False,True,True,False,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,False,True,True,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,False,True,True,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,False,True,True,True,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,False,True,True,True,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,False,True,True,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,False,True,True,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,False,True,True,True,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,False,True,True,True,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,True,False,False,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,True,False,False,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,True,False,False,False,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,True,False,False,False,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,True,False,False,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,True,False,False,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,True,False,False,False,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,True,False,False,False,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,True,False,False,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,True,False,False,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,True,False,False,True,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,True,False,False,True,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,True,False,False,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,True,False,False,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,True,False,False,True,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,True,False,False,True,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,True,False,True,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,True,False,True,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,True,False,True,False,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,True,False,True,False,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,True,False,True,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,True,False,True,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,True,False,True,False,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,True,False,True,False,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,True,False,True,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,True,False,True,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,True,False,True,True,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,True,False,True,True,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,True,False,True,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,True,False,True,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,True,False,True,True,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,True,False,True,True,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,True,True,False,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,True,True,False,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,True,True,False,False,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,True,True,False,False,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,True,True,False,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,True,True,False,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,True,True,False,False,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,True,True,False,False,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,True,True,False,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,True,True,False,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,True,True,False,True,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,True,True,False,True,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,True,True,False,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,True,True,False,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,True,True,False,True,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,True,True,False,True,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,True,True,True,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,True,True,True,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=4,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,True,True,True,False,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,True,True,True,False,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,True,True,True,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,True,True,True,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,True,True,True,False,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,True,True,True,False,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,True,True,True,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,True,True,True,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=6,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,True,True,True,True,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,True,True,True,True,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,True,True,True,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,True,True,True,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\nfrom ax.modelbridge.factory import Models\nfrom ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\ngs = GenerationStrategy(\n    steps=[\n        GenerationStep(\n            model=Models.SOBOL,\n            num_trials=8,  # https://github.com/facebook/Ax/issues/922\n            min_trials_observed=3,\n            max_parallelism=5,\n            model_kwargs={\"seed\": 999},\n            model_gen_kwargs={},\n        ),\n        GenerationStep(\n            model=Models.FULLYBAYESIAN,\n            num_trials=-1,\n            max_parallelism=3,\n            model_kwargs={\"num_samples\": 1024, \"warmup_steps\": 1024},\n        ),\n    ]\n)\n\nax_client = AxClient(generation_strategy=gs)\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,FULLYBAYESIAN,True,True,True,True,True,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,FULLYBAYESIAN,True,True,True,True,True,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,False,False,False,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,False,False,False,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,False,False,False,False,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,False,False,False,False,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,False,False,False,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,False,False,False,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,False,False,False,False,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,False,False,False,False,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,False,False,False,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,False,False,False,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,False,False,False,True,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,False,False,False,True,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,False,False,False,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,False,False,False,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,False,False,False,True,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,False,False,False,True,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,False,False,True,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,False,False,True,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,False,False,True,False,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,False,False,True,False,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,False,False,True,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,False,False,True,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,False,False,True,False,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,False,False,True,False,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,False,False,True,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,False,False,True,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,False,False,True,True,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,False,False,True,True,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,False,False,True,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,False,False,True,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,False,False,True,True,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,False,False,True,True,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,False,True,False,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,False,True,False,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,False,True,False,False,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,False,True,False,False,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,False,True,False,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,False,True,False,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,False,True,False,False,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,False,True,False,False,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,False,True,False,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,False,True,False,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,False,True,False,True,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,False,True,False,True,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,False,True,False,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,False,True,False,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,False,True,False,True,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,False,True,False,True,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,False,True,True,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,False,True,True,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,False,True,True,False,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,False,True,True,False,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,False,True,True,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,False,True,True,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,False,True,True,False,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,False,True,True,False,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,False,True,True,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,False,True,True,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,False,True,True,True,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,False,True,True,True,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,False,True,True,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,False,True,True,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,False,True,True,True,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,False,True,True,True,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,True,False,False,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,True,False,False,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,True,False,False,False,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,True,False,False,False,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,True,False,False,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,True,False,False,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,True,False,False,False,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,True,False,False,False,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,True,False,False,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,True,False,False,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,True,False,False,True,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,True,False,False,True,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,True,False,False,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,True,False,False,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,True,False,False,True,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,True,False,False,True,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,True,False,True,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,True,False,True,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,True,False,True,False,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,True,False,True,False,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,True,False,True,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,True,False,True,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,True,False,True,False,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,True,False,True,False,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,True,False,True,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,True,False,True,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,True,False,True,True,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,True,False,True,True,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,True,False,True,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,True,False,True,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,True,False,True,True,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,True,False,True,True,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,True,True,False,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,True,True,False,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,True,True,False,False,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,True,True,False,False,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,True,True,False,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,True,True,False,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,True,True,False,False,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,True,True,False,False,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,True,True,False,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,True,True,False,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,True,True,False,True,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,True,True,False,True,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,True,True,False,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,True,True,False,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,True,True,False,True,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,True,True,False,True,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,True,True,True,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,True,True,True,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,True,True,True,False,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,True,True,True,False,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,True,True,True,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,True,True,True,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,True,True,True,False,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,True,True,True,False,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,True,True,True,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,True,True,True,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,True,True,True,True,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,True,True,True,True,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,True,True,True,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,True,True,True,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,False,True,True,True,True,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,False,True,True,True,True,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,False,False,False,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,False,False,False,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,False,False,False,False,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,False,False,False,False,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,False,False,False,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,False,False,False,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,False,False,False,False,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,False,False,False,False,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,False,False,False,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,False,False,False,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,False,False,False,True,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,False,False,False,True,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,False,False,False,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,False,False,False,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,False,False,False,True,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,False,False,False,True,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,False,False,True,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,False,False,True,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,False,False,True,False,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,False,False,True,False,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,False,False,True,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,False,False,True,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,False,False,True,False,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,False,False,True,False,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,False,False,True,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,False,False,True,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,False,False,True,True,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,False,False,True,True,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,False,False,True,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,False,False,True,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,False,False,True,True,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,False,False,True,True,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,False,True,False,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,False,True,False,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,False,True,False,False,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,False,True,False,False,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,False,True,False,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,False,True,False,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,False,True,False,False,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,False,True,False,False,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,False,True,False,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,False,True,False,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,False,True,False,True,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,False,True,False,True,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,False,True,False,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,False,True,False,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,False,True,False,True,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,False,True,False,True,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,False,True,True,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,False,True,True,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,False,True,True,False,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,False,True,True,False,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,False,True,True,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,False,True,True,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,False,True,True,False,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,False,True,True,False,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,False,True,True,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,False,True,True,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,False,True,True,True,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,False,True,True,True,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,False,True,True,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,False,True,True,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,False,True,True,True,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,False,True,True,True,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,True,False,False,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,True,False,False,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,True,False,False,False,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,True,False,False,False,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,True,False,False,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,True,False,False,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,True,False,False,False,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,True,False,False,False,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,True,False,False,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,True,False,False,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,True,False,False,True,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,True,False,False,True,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,True,False,False,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,True,False,False,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,True,False,False,True,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,True,False,False,True,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,True,False,True,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,True,False,True,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 5.9, \"x2\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,True,False,True,False,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,True,False,True,False,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,True,False,True,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,True,False,True,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define the training data\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,True,False,True,False,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,True,False,True,False,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,True,False,True,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,True,False,True,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,True,False,True,True,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,True,False,True,True,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,True,False,True,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,True,False,True,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 5.9, \"x2\": 2.0, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,True,False,True,True,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,True,False,True,True,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,True,True,False,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,True,True,False,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,True,True,False,False,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,True,True,False,False,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,True,True,False,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,True,True,False,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,True,True,False,False,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,True,True,False,False,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,True,True,False,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,True,True,False,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,True,True,False,True,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,True,True,False,True,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,True,True,False,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,True,True,False,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,True,True,False,True,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,True,True,False,True,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,True,True,True,False,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(19):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        results = branin(x1, x2)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,True,True,True,False,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    return y\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0},\n        {\"x1\": 0.0, \"x2\": 6.2},\n        {\"x1\": 2.1, \"x2\": 5.9},\n        {\"x1\": 1.5, \"x2\": 2.0},\n        {\"x1\": 1.0, \"x2\": 9.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(19):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    results = branin(x1, x2)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,True,True,True,False,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,True,True,True,False,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,True,True,True,False,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n\n        c1 = parameterization[\"c1\"]\n\n        results = branin(x1, x2, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,True,True,True,False,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin(x1, x2, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define the training data\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": -3.0, \"x2\": 5.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin(row[\"x1\"], row[\"x2\"], row[\"c1\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [-5.0, 10.0]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, 10.0]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n\n    c1 = parameterization[\"c1\"]\n\n    results = branin(x1, x2, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,True,True,True,False,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,True,True,True,False,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,True,True,True,True,False,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(21):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n        results = branin3(x1, x2, x3)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,True,True,True,True,False,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"]) for _, row in X_train.iterrows()]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(21):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n\n    results = branin3(x1, x2, x3)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,True,True,True,True,False,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,True,True,True,True,False,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,True,True,True,True,True,False,batch": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nbatch_size = 2\n\n\nfor _ in range(23):\n\n    parameterizations, optimization_complete = ax_client.get_next_trials(batch_size)\n    for trial_index, parameterization in list(parameterizations.items()):\n        # extract parameters\n        x1 = parameterization[\"x1\"]\n        x2 = parameterization[\"x2\"]\n        x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n        c1 = parameterization[\"c1\"]\n\n        results = branin3(x1, x2, x3, c1)\n        ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,True,True,True,True,True,False,single": "import numpy as np\nfrom ax.service.ax_client import AxClient, ObjectiveProperties\n\n\nimport pandas as pd\n\nobj1_name = \"branin\"\n\n\ndef branin3(x1, x2, x3, c1):\n    y = float(\n        (x2 - 5.1 / (4 * np.pi**2) * x1**2 + 5.0 / np.pi * x1 - 6.0) ** 2\n        + 10 * (1 - 1.0 / (8 * np.pi)) * np.cos(x1)\n        + 10\n    )\n\n    # Contrived way to incorporate x3 into the objective\n    y = y * (1 + 0.1 * x1 * x2 * x3)\n\n    # add a made-up penalty based on category\n    penalty_lookup = {\"A\": 1.0, \"B\": 0.0, \"C\": 2.0}\n    y += penalty_lookup[c1]\n\n    return y\n\n\n# Define total for compositional constraint, where x1 + x2 + x3 == total\ntotal = 10.0\n\n\n# Define the training data\n\n# note that for this training data, the compositional constraint is satisfied\n\n\n# note that for this training data, the order constraint is satisfied\n\n\nX_train = pd.DataFrame(\n    [\n        {\"x1\": 4.0, \"x2\": 5.0, \"x3\": 1.0, \"c1\": \"A\"},\n        {\"x1\": 0.0, \"x2\": 6.2, \"x3\": 3.8, \"c1\": \"B\"},\n        {\"x1\": 2.1, \"x2\": 5.9, \"x3\": 2.0, \"c1\": \"C\"},\n        {\"x1\": 1.5, \"x2\": 2.0, \"x3\": 6.5, \"c1\": \"A\"},\n        {\"x1\": 1.0, \"x2\": 9.0, \"x3\": 0.0, \"c1\": \"B\"},\n    ]\n)\n\n# Define y_train (normally the values would be supplied directly instead of calculating here)\ny_train = [\n    branin3(row[\"x1\"], row[\"x2\"], row[\"x3\"], row[\"c1\"]) for _, row in X_train.iterrows()\n]\n\n# Define the number of training examples\nn_train = len(X_train)\n\n\nax_client = AxClient()\n# note how lower bound of x1 is now 0.0 instead of -5.0, which is for the sake of illustrating a composition, where negative values wouldn\u0027t make sense\nax_client.create_experiment(\n    parameters=[\n        {\"name\": \"x1\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\"name\": \"x2\", \"type\": \"range\", \"bounds\": [0.0, total]},\n        {\n            \"name\": \"c1\",\n            \"type\": \"choice\",\n            \"is_ordered\": False,\n            \"values\": [\"A\", \"B\", \"C\"],\n        },\n    ],\n    objectives={\n        obj1_name: ObjectiveProperties(minimize=True),\n    },\n    parameter_constraints=[\n        \"x1 + x2 \u003c= 15.0\",  # example of a sum constraint, which may be redundant/unintended if composition_constraint is also selected\n        f\"x1 + x2 \u003c= {total}\",  # reparameterized compositional constraint, which is a type of sum constraint\n        \"x1 \u003c= x2\",  # example of an order constraint\n        \"1.0*x1 + 0.5*x2 \u003c= 15.0\",  # example of a linear constraint. Note the lack of space around the asterisks\n    ],\n)\n\n# Add existing data to the AxClient\nfor i in range(n_train):\n    parameterization = X_train.iloc[i].to_dict()\n\n    # remove x3, since it\u0027s hidden from search space due to composition constraint\n    parameterization.pop(\"x3\")\n\n    ax_client.attach_trial(parameterization)\n    ax_client.complete_trial(trial_index=i, raw_data=y_train[i])\n\n\nfor _ in range(23):\n\n    parameterization, trial_index = ax_client.get_next_trial()\n\n    # extract parameters\n    x1 = parameterization[\"x1\"]\n    x2 = parameterization[\"x2\"]\n    x3 = total - (x1 + x2)  # composition constraint: x1 + x2 + x3 == total\n    c1 = parameterization[\"c1\"]\n\n    results = branin3(x1, x2, x3, c1)\n    ax_client.complete_trial(trial_index=trial_index, raw_data=results)\n\n\nbest_parameters, metrics = ax_client.get_best_parameters()\n", "single,GPEI,True,True,True,True,True,True,True,batch": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent.", "single,GPEI,True,True,True,True,True,True,True,single": "INVALID: The parameters you have selected are incompatible, either from not being implemented or being logically inconsistent."};
         // E.g.,
         // var lookup = {
         //     "A,C,F": "ACF",
@@ -83,11 +83,11 @@ <h2 style="margin-bottom: 0;">Honegumi</h2>
             <label class="row-option" for="model-FULLYBAYESIAN">FULLYBAYESIAN</label>
 
             <div class="tooltip-icon">&#9432;
-                <span class="tooltip-text arrow">Choose between frequentist and fully bayesian implementations of the gaussian process (GP) surrogate model. The frequentist GP model offers efficiency and speed, while the fully Bayesian GP models GP parameters as random variables through MCMC estimation, providing a deeper exploration of uncertainty. The fully bayesian treatment has historically provided better closed loop Bayesian optimization performance, but comes at the cost of higher computational demand. Consider your computational resources and the complexity of your optimization task when making your selection.</span>
+                <span class="tooltip-text arrow">Choose between <a href='tutorials/freq-vs-bayes/freq-vs-bayes.html'>frequentist and fully bayesian</a> implementations of the gaussian process (GP) surrogate model. The frequentist GP model, which is often the default in BO packages, offers efficiency and speed. The fully Bayesian GP models GP parameters as random variables through MCMC estimation, providing a deeper exploration of uncertainty. The fully bayesian treatment has historically provided better closed loop Bayesian optimization performance, but comes at the cost of higher computational demand. Consider your computational resources and the complexity of your optimization task when making your selection.</span>
             </div>
             <!-- <span
             class="tooltip"
-            data-tooltip="Choose between frequentist and fully bayesian implementations of the gaussian process (GP) surrogate model. The frequentist GP model offers efficiency and speed, while the fully Bayesian GP models GP parameters as random variables through MCMC estimation, providing a deeper exploration of uncertainty. The fully bayesian treatment has historically provided better closed loop Bayesian optimization performance, but comes at the cost of higher computational demand. Consider your computational resources and the complexity of your optimization task when making your selection."
+            data-tooltip="Choose between <a href='tutorials/freq-vs-bayes/freq-vs-bayes.html'>frequentist and fully bayesian</a> implementations of the gaussian process (GP) surrogate model. The frequentist GP model, which is often the default in BO packages, offers efficiency and speed. The fully Bayesian GP models GP parameters as random variables through MCMC estimation, providing a deeper exploration of uncertainty. The fully bayesian treatment has historically provided better closed loop Bayesian optimization performance, but comes at the cost of higher computational demand. Consider your computational resources and the complexity of your optimization task when making your selection."
             >&#9432;
             </span> -->
         </div>
diff --git a/scripts/generate_scripts.py b/scripts/generate_scripts.py
index 3dbd3810..ab8e7dc4 100644
--- a/scripts/generate_scripts.py
+++ b/scripts/generate_scripts.py
@@ -73,8 +73,8 @@
     {
         "name": MODEL_OPT_KEY,
         "options": [
-            "GPEI",
-            "FULLYBAYESIAN",
+            "Default",  # e.g., GPEI
+            "Fully Bayesian",  # e.g., FULLYBAYESIAN
         ],  # Change to "Default" and "Fully Bayesian" # noqa E501
         "tooltip": tooltips[
             "model"

From 2b9dbefc8b915501794aa5948b61a21534c21ed4 Mon Sep 17 00:00:00 2001
From: "Sterling G. Baird" <sterling.baird@utoronto.ca>
Date: Wed, 1 May 2024 17:01:11 -0400
Subject: [PATCH 09/12] Update docs.md

---
 docs/docs.md | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/docs/docs.md b/docs/docs.md
index 71283278..b490c673 100644
--- a/docs/docs.md
+++ b/docs/docs.md
@@ -1,6 +1,8 @@
 # Docs
 
-```{toctree}
+```{nbgallery}
+:maxdepth: 1
+
 tutorials/sobo-vs-mobo/sobo-vs-mobo.md
 tutorials/freq-vs-bayes/freq-vs-bayes.md
 ```

From 0eeb6e1ed6cac1dd1cca6030a354e21d902f7aac Mon Sep 17 00:00:00 2001
From: "Sterling G. Baird" <sterling.baird@utoronto.ca>
Date: Wed, 1 May 2024 17:19:58 -0400
Subject: [PATCH 10/12] docs concepts

---
 docs/_static/freq-vs-bayes-thumbnail.jpg | Bin 0 -> 411116 bytes
 docs/_static/sobo-vs-mobo-thumbnail.png  | Bin 0 -> 50244 bytes
 docs/{docs.md => concepts.md}            |   0
 docs/conf.py                             |   2 ++
 docs/index.md                            |   2 +-
 5 files changed, 3 insertions(+), 1 deletion(-)
 create mode 100644 docs/_static/freq-vs-bayes-thumbnail.jpg
 create mode 100644 docs/_static/sobo-vs-mobo-thumbnail.png
 rename docs/{docs.md => concepts.md} (100%)

diff --git a/docs/_static/freq-vs-bayes-thumbnail.jpg b/docs/_static/freq-vs-bayes-thumbnail.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..248a2f3e5e5cfc2f4ce70a398fab60b1fdb24c21
GIT binary patch
literal 411116
zcmeFYc{r4B-#<Q*5Sr{Erb1+GAr)q{AW4>DveZ<_7Ao6d%!uq;2t^q!k}Q*bH`z({
zErc<Y>@${eG28FzexC31+|T`cj^A_t?&JH%_j??lIh+pWnrp7}eV(uP`~7+^C;K~_
z1lfPZz{mi?0lri?R3H%cIOHOPbJwmvKj3RO_`}7+#kG4k*IsV!Jv{t-`T6<w^6?4m
z6WTAZPjDX}-+r<Eg2E!AqN4mz@dIKa2ZTgKMgDvV2Pb&XZZ2LfE?yA<J^_*c-!JwT
zh#(JlKQxPz;|OGzAP1))2fG;pgFraA{>+wx1M=@Lj$Pm!_i*#@?&SkNP_`eki-VJM
z*KW=~vj#sM1YU>i7UU8-eB%5bVe`A(M;?ise3JNvNA^NdgQ!IxMefwS$059X#l#Ot
zNXj2oP&}rjp?O;C%-M4nFX<Z?UN*XN{f4C#@}{+oy@TU@CubK|FK-`TKeT^9=+m(9
zh{&kuq~zx*scA1>roVmnKKnz?$J|fFC8cHM6_r)hjZMuht!?ccU;76JhlWSKkB&`~
zXJ&u?n)^M!KwVqk*raU%+dF^c;(&1en^^xL*+0l72+Fl<_ioPJ+<)Za*yRTfPQl$=
zhfnMgI&aQ>_mS|ClTUa=E+oDwYT%VUWkC_W_qcDbn4HG6JoS%gf0OK=6D;IklI%YO
z`%k&XApD#h;O21(LJ$z<^|#TAkiW)XYw#Ba{=&dt82AeVe_`M+4E%+GzcBC@2L8gp
zUl{lc1Ak%QFAV&Jfxj^D7Y6>qz`u?Gx8rffC-!DJB=?$eek{b90Z*o1D<j-THpcCI
z)(aDHFLPWnbtu2e5|)^Lrgc50M}ES#B_We<pB-E~D^dzE*BAL~{3~l<y=BA7a6EPK
z5zdo8Z%FR>F2vp#mQaBdD<l=awT){H)^yG%09Q&qclO%grhc*^l89-gKH~3VE^G)c
zjp58`75$YBd75XyhGb>uFj*+-yMMU-qsnSQ8BguF^-N{LnF*;quxwL3E)ny;8<xrw
zu1zM-41EH*Y7(Cnw~Hh$c0q2vEhhf`liaSI6>}cy3w_5Qe*KwY0pwhCyvs~^*w2*V
zGrf+l;~_Hdhy57xHd;(BRmAEV8xjX`?Xy%Vb5k{X>8>_i<tTAiIK1tr)$jj4mz`Bx
zUokMf?vWEesWXF#6>%Ck`jHK3w0y^g^jZwC*bsG~cTa(63LC;}6~~60H#Vg+h%<)&
z_`n<Pn_ciFGT*p=0$qp`>CEB$HX5S%FOQN!or~FH-+H?~+UhI4I;r!CXD7CIPn_sW
zSXAjthZ`nWhF-_-ih7<DbZ+=HM8sVGpBy)<x8iCM)4qK9ZdT{)1uOY#!eVDelJ%r+
zAYa>I`|y0KM+Hr&!c8ZxT-ycD<3Xe@jQ6vY(Wh+NXt?Ooz#kiH>h>6xLGZax$DVyl
z6U<LM*CEB9cdY#g{_i&R?>G3L{9Dv&>qg0dFSBq#%6&p-gb&~S2?uTg?(cU%wC)?y
zocC^`V_fyggP|6^9Qq<=|HJ5zyKL{27;<FIIv$e6-Di!XJpavxY>1SyA;eCW2A-1Q
zJoO*{apdVc5~4MH4+&iB#qTT;XON`o=y|`3N50m7^3oIP`&V@A|FC<I$Ef#yxgo+b
zPgYi>kSb#gYLm{|yXVIqdVD~pS;j4|@H%-=-HSqKeb}mHbTlR)TG!~wmn`U<AkH}<
z+BFLrIjPjf4EI0))jNE&{lF-dubEgn_cY%Jbr6U#FG}?hvc2W%Wh@of*<f`~0E6>!
zSAE_5>}vQ~`+J22gMx#|{wd0}Tc?AJX@!*ZFB#uVe-Vqr?b0DkI2#i32`I6n;Ff0O
zSDxfg{Z7NX#$o$%erTMu5193^5Ws5Un)ER@*pRmgWCG)$9ggZl7iUAbJo;DEh>o`O
z$^mb&k%tnt1K!cxi>Po7j-3fMWVaU(MJ7typ#dPdxw^deu(M9~AVIE@H|4?d<Tb=z
zrpk!)yFMRo1Eo&S!~Ew$OV?3pILf74a2HjU+-)vqvzM(w7pu?e3sZR$dYcV-CIec9
z+x@uTM}rZs*pMiEDR(nnJraH>1Qv-^j8|Nhv%EBvWA;uLBJjknhoOONLF_29l-Q7l
zmNqs-{3aVB%7)CoM<=j|4lM7fFg66L4yxDO6-xO_9kG$>*iL?nA4P(>3KDYA0LT~{
z5*`iYup#=e78G!%k_{n)=>5mH7MHGrf+B%RsEjftcWd2=4Vm%AQ+_7I;5RhUNEWMy
z4Qc(&IYR`T3DLL>bA}}w5*m}hQcJ-(<D(yZMNHcRd7#C6@wX|W?y@0Gsc<!1@6a@g
zPDijIQrRfrNKZ@49o)1Mg@D>#{PSKGbTDBQb-i{N&H=#J&1%wj)@+aW+u5D<^Gfbe
zC<thIh~w1Z4x+S8BO_{0<(dj~t-SU)l4^_De&99el)n1EpDz`jBI8=2JQyj;a%%;7
z*<<{2QLu--tYc3fw*p1<`_-1akIx)7Vk&RvG2hm!0ySh}^Rn+^i_Oo;BKd+Pik=!w
zS;Tdg%Y0*3Fv1cp0dAQIY-VKHei|Av$=5$T(5UtJhy29mcLCMjn=kvG-4T(vy}84U
z>KS1OtFuh6Pp>k)tJsi!f8`yntUbgZp4m7ZRH<KELX-`nttaxd?`;HCv*7)?iGF?l
zbEON3AD;z(dh;0~9frI3P3xG<5YR)-WAS=XBic~nwT@&dt^sUA?m~sQ)z9}2uV|-A
zYzM6C$r#av8J-}bVPjrjd?oSm3`3m#>1-imcdy(oxyNeLP0A;BBUiJr7g?P|iivs*
zYA;qBZP;SdLwh)&G5&oH-|4GbdR*b)UM4zgSG3}Pj+YvJ<IVk@M_MFfOR69oD<*K_
zIUK17C01)oS#5yx^)s{zCS1e<1}jdWvP<086D^PDj3?{KyrElw3uVlbX6&-9gBhe?
zY{`{t;oV(S5rq<s7fSw10s(VNo>V_uV41FsMz*q!>{lavQt3H4ezaNYzK<DqpFZx)
z)h=1Fd)5hZ$w_WIdq6t?0=Od;qs@jCYQc8Q<sm^AQO|G`WI{Cd6a}%&IlXAQ52&!|
zU6$+Z8=JaDOy%j6K&e<B*o8hwM=t)v9tUUSlK&xw7D6fLJiO-Edqa69Hp)&=4<>~l
zQy^#T`WxWc5N>f8;}&)S+c0@(2%}F|LI;`+l_}R<H$xw&o0$DzbaUxos?Nyz0ID}_
zi7WfZO>75TB;G-3ea!zO)W_j~1_!W6*T<lNl=qB>)Rb&|0>i}JU4X9gu<28GWGH9w
zo6lTc_<cBMSkwr9KSOml8?v8au5!|~{NwMj+Mm`}zw*fX$#Q3#GxNOXH}GT1ZRKf~
z?K4%4LUWhbatn>8b!*d5BDFC7E4K}YRlabtYUClfYcv&`?p@c6;*1pLuMb)k9dLge
zyJmUeRp8##;Ju1(gZH|fY|VZVW_Hr(%&ym{Og4ma12*o=kf;2>3ou}~pFqj=80P6;
zV%tk3-P3rh1`&=C>~}onI+O?8Kq3@v{+5b}4YzlUDiL^RA8sG3)ho#;0jlU|j1$oL
z&|#?7_R$A@{HEk7&A0ZS58I}1vs^OChDnAla`sg(<Bb&R0t_1vqf>3TQn5<!sXflf
zb1&k8E2}TTEh}~TDqJBy3H3+NIoEr}7>1uFoCvS|WpSsj4b;n_OAOn-V~q2xXL1q)
zkKklc7y!ZIsi5v!xRl&{TdB|e(A{L|jLZGd!t0`*ty2_z;sNW?E<|`OGUv9A>&^4%
zo2N+L2X2>gtcf=f$CYQGO}aMP%u{WW1v5F9HH~xC?y({D)_EJ_G*u8_*)Bv0%^*Yr
zQSeZNK2^V4Hq;{aX?gWCHe}BfN%xvj3-$`b8c?5sg}~*LFH%z@eBPms$#XT{b-ev8
zr&6=k_J@(1hS!B<k)qFzK4%WalY<eT^zVF?xO^x!M7w{a2~RP`k2T!@HYvmOChW0l
z+l2jqcN@Ro$<vNCPeU%kw)A$@&-|zI{k<ao6MsWxnR%PW6H+8-thxrhV@dr;v2Q$i
z;_KxIms!DnBHohu7JKaD7a;C`xk=`JR&OTHAXXN0h^i3b{Te?eK3pJj`@ZjN$XP<$
zBhY!o1LN^XV{TS8kqr^nlc9odJpt0lY{)_MK+6eqOp9b|jmX2FvV#PN*EdLyIg_KA
zmV};F;3yli<TNuiif@r*ZpG-o2wBrj40`?I70IVvbcgewVuY|T6vZqd4z^!Moo<Fs
zYPw7%Mrw2(=sRs|y>MM$a&1EN=DwkI-2cm!-mU?CzZ@HqTtqdbC$b@@2`}!(b}1R0
zQE4y;`pwykn|@n=j1B2Cq#{}S)Ih>Q{J<K-i))ncQID*b#krInv%?mkES*6WF(5|o
z3@6<E*g(efsc5BAe;g?-ndLV70-M<Um9);{>7>RF3Z>757a|s4fzic{EGTC`8{z@R
z8R^$_h?0|?3EN9cY{<8fItDL@t{RLX1)eQX5%0dtuzceIiF_w$chpn?ceX7XjciN!
zZc1)jqTfB5LvS|rf7`|?xn5*JG2E2fyS)&M_|Ovylf<Y~r@qkOij*>jUn-J^uNB(E
zXD@%#_ejlGh7|Y!2Ufp0f(x#D(>bkzU*{~@Kn%$z%Uao^aKtWvQOR)IivANW^{yCj
zJdtk5;{LxpKIQ<7z@0u0JRt}1(Y{jaKD;KYTC@%c4LlIrN-~IH#Ycp3b0*==ZTqpr
z+Tw7(-G{*~g`-$X^`YQey~A}Z0CUVvQ|8h5dSpq3R@yVYef-5rP6G-e9N;cCey0+c
z0@YJ2?oKwO#heY{vu`DCoL<BMFm;fup@PsT7SA{Vo!nII`>FnTQJlxLbqJ<SLZWS=
zz!Z3#7K!-nRm8{usbChg2L{;6XtE)1Ds0)1FWWfa6Jq)SdW01|;mL+@g7GssZ$q1P
zFq;hBIk<!ab%zY1FX9+Flog>3>mUN@kR4MNuWX^~?L*8L*nv;3;x=(@wCu0j71uT5
zf`9NA9j*UCz_TIuh@*w;G+V%s4dDr+WsBu%js2`zpYDcxy?V?Q#PMI%4y%J}u-9=R
zLj_p9w-1CGjc7`r-qS0Zu56qWW$fSo_(Eb2=NjRA00RZC)&Q=f8Dxwtv!LSin{hWQ
zklW%+VU$#GIgY2XA<=~2TeQH+^mUSH)BK9O|6KC<V!zajslLbG$YuT_zKmwQU_+b`
z+lLVQh)bv<@Jo304)!eP26=tJc7Us?wzMK&S{-q|5<M3dU48sEkL(-5ToG0o7`nkh
zp77*M*yb??cLBK}MSq0rc#E-SLyE9-V8~6FI}(hrK#*cs$46>SsN7^F*PS6&?m-(n
z0m<8TI&VxfI;5f+2*Z0Bk)RZZLH=+FNM=Lsd_-*cU~B%{K#{w7ih@-2GN*th>{FV}
zE#~lRFJ_f<z0dnD;TdS_jnHTYxQk&$GpHRYA0h+#H}OYtfQl(2-2(UBcVJr=E7Qx3
z{71wXGE@^J!OX#U0UJWX5E+(1(sVIw!#Z|&^5F31&CQ?v;wDw}vjz)OiVrs6Wg8eg
zMPF)ys1PUHL%#^hn+5HT!Nu0=mi>ZBy7j?mn#UqoUAB^*%g3llqsnMETqIV0&uM<q
zO{15lgJLZ~KM@Q;<6Cnb)NFh?Wr|^f>7$$B8g${ZtZxBV76YQnMrz?brBg07-!G2H
z+&N^Q4GHvIssT01m5K}9%utmgJ15+xEFMo-{+wmB+_13pBaO2FOGG1Db_^3Rr*hDt
z7wQts`h9q;0s|v=&K+-uPhDZzDQ~9zv5Ny{!|po9lpeEmqvZ|2Alpf<;B{8<o4B9|
z8wm5_znT#w%&pTXp!=$mrPTs&TpY=_Z2E^)ty9n3txFW#Gem+s<ihFsY~P|^1DT!@
zR-WMsY|WTT&CAGAmJ2fqP@04LCyb7#h<rS7XMZg9doOnbiW9B?w64z}gWup=Y`zs4
z6^<&uo4RcI<Tb}`;r1iM_pXdkzy<VrK|mg<C$e6AWTn=0I!e{M?%EjbM!Zo>v7u58
z*Pas{oWn%aKNq?b+LE=JR|PQgl)9-%jOC|>_w&9VZ~c6}rni(gh`(!VL#wi%el4?Z
zIk;k9^0_p&Q)BIwlH!U9mJ2&MRles9-3)UrQ(MBJg5vRfdEaZ-YdYU#93J()nv)si
zSSFS!FhmJ0^{0kbN^xP+-q3x=oX2S1&_|DcFQ65ncE~2UJV2OHKXBW~rvUxVYVg;y
z*|M(L1OQ^XtLo;Ai^C0lm8FNG21W{HL5Y82yaqL_GUh?w-4S)GsYaWtSvrQ6`-;0G
z-W`HaH?yGeaxZ^9tBvoT%RA9ga7(LIawQ?*^P<);dVbdKoc)^S4q<u<ID(6a&zu^^
zY7M4t9psZwxQ`<lr>`Z%;VFiw188HTjO&!sCE}6qDy;m9D%&d`A}V*0(jnV9q_*Wg
zDvtCaTV%&G+GEq$@@{pHRmzIJ@()N`G&7DFi*si~1hB)mr4ofS`)ZL6S+`vtU*`27
zo;%1N8iJlj1xwOGKcek#*7TRTo%LoK^j{RA4HS;VMVRS(Vw8Y41co$qIE_pcm@>Yb
z`~B+y9h12E#42dXJ^={M)EoI4+d~vnWu0cgC~dKaookn`bf5LO7fBUqIq$&WCFuX_
zJnb6_P|1n*Aqb2D8uRDsXfNw85|{i0#8zO1{mN_0^`PB*!rf9m31wcA|Kl8JO{~3w
zba8-4D-Lbym<~c34d6=fC)p4ejhLG7`7+Z2%uO6A1|c4Z&T9l?&B78kgjx+@N->{o
zDtFn^P?Wf?&B2ihbskmCrZnGeWUl6hraxk-ijr{ykfk8?DB&eO?*1Qd8VDKXZ}~rX
z$SonUe7oqs`iuXPo&ERZoqyZ1gpVe^0Vdw%d8*;$%_3`A_xzmr^8`b}DP}feHVo)w
z)xx%;AuR8edS&#mcJ3g$!uCrd@z-JXTsFk@?LQjQLf0)8LIIQ@j<GmQbVQA=EYi|}
z3P)0miGiw`ZKPwyvapjgPj-U+yxu`GplrtRtc==l0~_-BDQHohZ5aMQvSUICKyQ9F
z!}4D!Nnobe?_{uea>ynTU#Xe+nsU;2&<4A4T7d3A8Dn(Gw}|Q&F?-0->bhc2xvv>_
zd%EGW6BVbQA{RjxBObeOK(&RDv`_c((3biVnCRv57;3G{0ldfy4X0n9QW0FV>f64*
zui^c{RKimhoYe(C$IM_{sz$|n=c-y*=+~_sTT?qRVlPwn)0=~NZQiEfL)C|I)OR~W
zKc|Nwmp?>RDCbJ-maIU5DI*H#wZ|lJ5*Y14th8%?T6J&6x1z*uo6E0Cjw>)Tl{fDF
zTOju*nEStR6y#bT&CZ#yF=7>xk4DwU@*ukN`}AO4>dV^GIN$;s;;bGidE`TL@vz#F
z^RWVNx_Jnn5Ttmlp!;=NQvL93tZrF&vW~s%yP~VGc1fljkV5BR90Cev_*RA8b~M%&
zDq?TD9C@yIUqNoh-G{h~4RJ(7OqG;wt)sO}lDciDHVc%)6=EL+D{^Yl#wZd%BK;&V
zqA$1|R)Aq<;d)NO5Wq6-tR>UbH60__n&Ef%am{k!TeLjN^3IKf_K~l6%HmWoJ&M&(
zEg?s`RtI`<4URb+#n6)rBox@5>dcCt%dD|FXbmi|Q!Qo73jD+&T|-vgdk}CYUQ@g*
zQ8bQ3c;XF`GvM}Ufw!CV<W-OHj9>TK6gaC2BmN<Y_!1V)hOAlWfT%zf<Ib%T?Qz>@
zp<u3HWCORezmS_uWIRt01pi&acUQwkYg*Htnx^7p4vcI@VP2p6IxnAoRXA+7upX@c
zr|23VuHkz9uJA+X#K_Gxkk3VdbBf5pfH{>nal}=S;8Ha>HkqPp+zef9k`|L6G_2S^
zA0ylV_~^GW+RS&G?uOSWw?BHW<R2F94wLZBy;C*2?^m#HVf{m9F&py04LS-nqI7N2
zqIVh>kwRZSCAqKB6z>JVY<A$p6F5>OZ0#rpB+j41=2$UTAg_s}vP|hVnZ2*@J1aOK
zeW!%nvz$jFgd+B6BuJyfP1S;KPdFC!-_!n(@`5wg=SzD>n<!%y=|jCpo{EB>Tp07-
z`L=qds4Tv_yI*Dd;oFa>#6aOv;RR>T)x1?eH782y=E*UJPLVCw0~K}b_=|w#l2DWl
zwiQP)SLLP^j=ZOQOBZN;LPjr7E{(U^IO~)h68?4P&f1w4tk?wF(5|S=I3rGG=fFWn
zhd?L3Z88z$e55!W7x6P(W|5^k=<Yk|njD#bw2e`-+=QFy)PudZT<2Z}873DGaEq>O
za`dGSm^$@9Ki$O@j;Hs=BEk*qmf)?mpec8uyLC~d8gof_-dnq=@ddEeoQvx>8};TE
zzP!3e6fTp#vUrnNSyx|`y0SL?Eve^Js{esCBZb^7X#eVdT!XvLMX!g8*%;m8vM07b
z{ncN-jHn!KaRy9Td@eXz`{5CmjBP>QK_f4BOv_-v@@N<4EG3k79w?&@n4XIWWx)J%
zT|Q0AP@~%pfVq#zo@q12J!UdoiNReKzuJlwtm$8pf1Eh+(;=>C<^2pb3pxQE9j6Te
zj&viNeC)wtt0|<NO1eAk@QaGKd(Pq->lTQqJ$yi(6~(H~sleH9eLg+=Rmf$vf**PB
zpcj<4a#*}ubUTIR^OTPQRyW&XHST;~Fwgq(@UdWPz^poagJaWD1o`DpM#^;j<T@$h
zDT$)+v~X;^#pZ|nuiB`3H(RXbXG5jgnLWX}P+BTzxDy#CnRt(!zG=|cJ<VuN1mmQ8
ze$DTFkcp2)=*Tvk9EMnDi<bD^{Ce0;oAGdZ9Om2QEp6p>n}r3p#oR&<NB@<hUUEt8
zEib>nN-}+H;)~~FaRYNR25;mOLKs^91UrM1ogP^n-SGAO2#~NE0U&a@UQkA&%Rc7o
z1WkuiN>83_O-r>X`#MWa<ud3)S+ROyWjPTnKE{CvvTlez+5O1lkfWt9MBSVD9h@=)
zkYw&=_3!u@WaEU)E!*Zy&OsIL-y>c>l;yHGI_UW(UN*QkqVRWy^blOq&aP}E!o$J(
zZpb~;(N|~gD`&n0D=qy5Ajl~K_xlPSw)q28Sj{c1rD<y6I4umg_!hm=-%$!^s-JSN
zIZduPNI*3b)2}dD;8KA}`;IiUIBOEO5!lP!ixRGVB;JJMScJ=`b9#xr=6Ly*Tk3NF
zat>q+4X(B4@<Cg`b7R}WaA$pywH#DLx59PQ0pqM%sqJW%yBcnWjN0m1&w@^=k8;xh
zFiduX3b~Hy$3HomL=(&po4hdognN9fzL?O41CC5FQlR3vk$haoBJD`9(n4C3$v~mU
zs2`8L3(w}4F03~?nZ<W(dUYQVW7&^yD7ieWWnn9^d+yiw7O$<Kyp2;WxHXjp{N}O(
zE5vR4IgrhU$e(J#(?qMm0J1<2RthWeWQfcdFyF!zZ=z2RGzrZQ+}e1Zt-^g~KAIjv
z<OEeV4Y%gQlHOrM%u)3s5M(#1G-7#*(nNEII_53ByFL<m*n0vYRIK$r!?Xh!;Q+T;
znpuLI&cMP7!=^iiYTXNt#S4!tJ=#i)sBAy5iVBzfRrL{_W>r0;Y1>gIU@-qJdBX0j
zznuM#<l$h1ajgmX4W%PTtCDoZr<dQ0ElkM`picP8ynjIcKp`^32CU*24wUsoq!J!n
z7P32!86OM?0~J)Af}>di*;KNGhmPp9a@XyCF+TZr<%4c`Us*($Gv+ok9;=2C7{HY>
zZjtCU0xI<_bq3b53anj5us8^I4BL&X@M~Nh)3q2d7;1f8$gey<7Jm4?J>$d*0x^x!
zA6e^oyNF~-4Hfw4s;#&QTo^vDB7Xb?GA`l*&u_P_i!9zcx&vZap0Z*BZT90~Y4#>s
z{j3gpE8K|CwcY#JnRBUno0C=!;wj7pQA~`>$7lym&udz%y*cF{b@*16{rF@(k<dZ_
zRMHq}?tH9T{5BX)vJ$|>NT;k2@^9W$550fUN9W<5^#rLGYS`7fpb}~xErz1L#GJlf
z2z1_zZ#|MRQyLplm6kwCI15i<jQrk7{=HL{!nmW6TJnAYdciW3pLL|p9t@t;qhRHR
zEvdvzsRBQn8fEE1`~%+jiHDF&yQh%}!@=-%;0gHP&vbasBll@_j<&!%`(*3uYMfuu
zm&P95tQQ5v7<Dt7)R(EE?DV@?-H}V@8{QW>sMhWFv(Avk+dw}*pRP`OPK2&RE3Kd{
z(6N!amnTqX7bSI9BQ}i-w9Lzsx_5>ziu7JN-u{IYst4;+UmBr1GR`q`G}d%9bGOem
zE5F>RDf;;>r^?L#?uzon;&v=>ifVWR{H1=LicQw({ozt^?9_WTWOwXwRjGhH`)6)D
zR{^IPH&i5xqZecB{I$>3%G~Z*o#;yrM+WbaV7NZWr%4}n)R#7?+GZ5Txp4Wx?DFd`
zhb^))^=3qY6mpQY&#I7Xb$U)}MViHr-K^h!stLpO1E8?U7zJuX6iBxPe7B4#o12^O
z3ay;)spRmxO^?0_^v^sd@;<V~+SKeknDt+T72{h2QKMTk#PC{we}}QS^5<)56;@-m
zn5(Z~fG9Qt5NGjmXQ4r-dHRWX=&vW6QqL=dUM76KIfL*e!Zf70&8YC)JsH_HMnOR*
zj+{AcS8~UJ6r?;>zndXQoqSDAYp4*H5g)u4u6yOEzogKvV1KY=`ZmTk(Z@;P%K;$+
zxwgUkL|v>buHp6uETSGx6%W5p!HfvGi1&|5JZ<Jy95<R@MrEbjJnwTC4L@L;CCl2S
zyj{Zlh?phU%d#QgeOK8KgE9z~zz_z7FLXvpz!jX~`$8K_tW7@skgvWrAS6=M`Y4*d
zKz!}_U7K~S1T4`#od#K27*1b<18Fh1wBg3PN?3VQT=K-P^U<S2muFB`_FqJ)>TNxM
zGItXq>|E>^GJGbv^<2I{xjFAK55&ZC2TimVLv)<>X=vFe5-#{#3+j?n%IB5a!}sid
z$K6@{-C(zMZv8&bTEl$n`2hElNmuctd@&j|wluzN@4E#P2sXtHPqD@ApEzcsHQkc(
zqDRQ+<EbQg#tDmg1jV-CvB(V~;NtjbQ7$0)7^55+jYhC|b*Y5r^an_ru$1~xxlp)%
zAOBZfTG=BUm}errCLVzE{?@WtrbUi@h;j}$j_Ewf6J_|gs#}qB1H7)z5S5)aY~_nl
zRlZS~*2p?kbthzU9-5fksx$Be<#18M@A&)A?Zque!j1Na<U=p;EGTc=SvkMwfB!DG
zT{eV=@Er%pwJp1SVi@*qLc{Bo>fLIVWkfSft1IvI$T=*pm9W}(_i~TnN|CLdo}+BY
zbojd&shXq2@Sc37?+PISEK#739*F&M&S}~I17I+2+WWq5+#tzc&P7g@h&F?eyvI?w
z2#UEor#V$&iAYwKR_iw&Z=-*1KH2uul_5kmY+^$M(MPpiG$&*hjcs3T8<0=dP)9^K
zdPaVO$ve=caT)fqAPing&l9y;RaItdQqsF0D(|@`WY5gNmXUK@zLKy;m;^e_IKpB;
zM>{Cc{xqg)Ix^+K!Go{@2$soeCzkmnEZa`=P55>i1<9*^IB{U2(}`Q+xALqfpz8Fb
zk0D5!5_`JmQXQ-3e%s+IH3?0cU~@sk5gk|ofSBG_wx{gF50?_%nUie~YmB7w7N=sT
zc4L&Oi;a2HtK!yWE9u2iL!u$iNP^EE203$X<NLk0hJ1{*6@kE)NuoO?5X5VoF|lMa
z6!gl(FH^f*w0$#lC5MG{oA1DeYp)xD_`c5&=qoc;y)b^UT3pOajw<&0@{b%@gldq9
zi<@*@z7?gg{jFK#yU)^!oD_T3-b9iCC1Lq4Syk_i{>bHe83tF+yq(sPIEet55?xLA
zxP&84J?QePfUSS??3qVZzz(t@_C4WKrG}B-Ot;o^b`^fK)4#@!9e$?JZtJIWaNQcu
zgSd+iLvS)4xs!YNT!d}itf)pJUdD=dJ}s6JpZaD&v;J4ZQv-E1MwP)!D)4GXKFIUT
z9yRtIkLUOvRs@iV=?Ss;{e@U1h8a~oy3jG{`X`j}BUrl8F{!gkwHpwb6#5^dZa}@k
z5@bGOocHonZrsxHr8&H5IC9}jG=o12ss{*Aa2CKa^|~JAenz#{V4=k^4^clJSq1*t
zfW53J7B@qX()I-bnXeb8bmIlyi}fj{JEAiEPpoTeT~AJqA0>SC1Z1IMTw84jLARvl
zVnet0wLI2dWuv{bfn?(SkAK?4IOp)=PBi;|go}D;bwO;ZIbQH)<A|jD0%28f&_hYr
zk0Um4guEJpxC54fu(}zYpDB(*c^>1h^`stce<u6}`@R@0Oi-3Ps91bSLJQqf9EiSk
z?rTHA`-;k<@>=VPyqV;9GV$8Sx54|LWhH+hBoUoaz%gWF!V^F7^s8FSWQ1VTGgTFF
zjCO=sf8mAn83ZX8RhIgOqCVY~^R!kSTYg)Sy6g4&G3M-P;>nW3+&|sFc%dTG#Dym9
zjCJ}oO0Kj_8lJs>@39g5eJ}TXH<*s+sviD;G|`^Ueg&IIy=C-*Wvi@x^^coG7mt8x
z4VcRop2p~4B-LL5OzHa>4nW`AmVCgiVdVPnwV=+?Te1c_dk*2h<}-|#q4ioA<=&OD
z0>1_=^WU`2YTI@PBT;U<X9w{8!3aY*6s$VHVL;>B<Unh+X4})8*UfiH6JA0^XLgtG
z;jRm!P7{Ua;m7-yH*blXRSa${h6zZw<cD0kySay$TD<p_U??gZK{3RQ@mUV&_<zEj
z9`X@OyHH!PzCrirJ*pk|^>{6FB8#dABW$ikEC<m8TyVv>NbfEs`uSw5z`;ogIptD&
zc*<rF77i+^3j*x9v1NgwdNw2=5ZB<QqaCj!dKU3-6njUQLTod~T%jCIB;UMqy3cjL
zN~hr2@x)iibxu^j(-NKyIZ7uHcPI(ldm)=&x7iR{Ju^Qa=K!kf{MgTxq@Cj1cB++e
zFppE20fER~(d{m9t-+*MlGZeV&a1E%0BJ!`Y<l+DCvaWEErZbv*=?CHnl|&3unePo
z9d8x$709`$5w=@pJn{?oz$7taFLM^$7E6%S=g#9cDN{Rpd%3@-{QRS(uFwcvNqw^X
zT&&M;_xqMV%J&~nnlI_yzO&M>+42BvN66)Fz&&NjQZeoF?&tcA4o6iRDOy<Kgp{{C
zfCeh+HZVlxw4!=O7S2`^>r-BF-)ofXP?p?L<Ocg*jQ^BexXMb4>(dcXhOT7F=X<i5
z3!86#=w4aUY&j2JE2cj1>r%l(H?~zf$k2-8Z8OF7x&R0LFyjjIeJXjeVdr4lIg`?Q
z{nocjc_N;%%d;WZbd@JJp@5uetG0<|;h{jOs>USM<Zy)nIdU*O7VrnZHNSV${hp%E
zT~b7^#`@MuzK6|x1cTd|P|M=S_R4SrOwesF<G2>hU9A?krXQ&Eu_3iD+X-{(<g`YV
z7)G3WbyAud9|w>5)XI6_egl8bo81o$MTf9m2+*LhA@{)?hmBv9hVkgTQtKNi>pbZ+
zK?e^D?N?^p8+iRYXzo@qESz#bjw~f^t-2|7d1cSkv`SVAgSW*WzZ^teWl4;Xur=fw
zbZU^rnONTwMq4)52N37=z`8oB+^`8T+Sq!X?Ed(y=5d}|C)=ZPCgM5YzWNB?q^w2~
zNfoM|k5mHGo6v>i4qnYJx?$q?WgfS6;nvKPL-mKy@64%tssc~H`GmUSnC}0oL@0ni
z^UEUuS<K*MMqrh&eUj%Eu~NOGI`DoC<J|7AvQD?N_-DbDEBO3xnmai^IX6wid-3WR
z(tg=N3%_ofh7l<?h`nw*#SJqFyj{~Vr)QSMEXt<lOCO=s+@>f-(Ab@8savOAoId?f
z@(}y{I=^i|z6f;)So{*O)r1nhjSg(Cb{M->bbtJQ*(*%v_CrHZr?V5nh~hfOs2NQs
z+$M#FNJHM+=U#mGy!gBGRqoD+J*s<a2}!^1#)g!jTHH_mnp|_DOl&OfdzZ5!FK0FL
zP&!i1YZ=vBpZ2jK&t-U|ykDh!OE5oju{Y$D-odFBT%;gIjk2eR?7!SKXydJV>e4>G
zJ5Hk88=xpX`)2T5*Pu*41j8igH666L9aX>xs}|J1UK5<wa}GM44r5J>IW0RDxK;<n
zhaF8CK2}=-Ixg6Gl;y~*7DLdUWr!C8vuLp2%A3D}Phs@Qs;3$QR3{7aZ#U(*lz((c
z|25rsL~+H45Xw3Y8jL&jGH4;{QmuIQ4|MB_*fC|J473-kle8kb4iYMRS~VClQYF`?
zJ)Y!h62nW8CtOc_X_kt4^33vu%=O)m7E$;BEZ!>*b>7K|@5ilc<TK=0$CpuGAZFb2
z@IQvQl&U`i{mjt$Tx7n8?Kb_D;`ztP|5)rTb;D7{TD^O^sS9fP+@D8LC((W^-V!Rk
zIpAu*Q!G#Bn1q>;P>+Udch0j%H3_Y?Z8OlNHtH#oQiDP~M%m|0gSLCL|K94eG$qrV
z00=XUnMovlnA%%^#x=ECdSU>uN4>&4n-X{?%<NOX<1|)ZAFrRP>3!#S)8Xf{f*|*^
z4R@+Yk&hZXD;TPr+hB2aw@fS14?zO+F*0R{*r}2+8+6Fla{6%C!v`0Ojv5`ZycOS8
z)p1sXgJy~0rRK2q*1cg!1BrRSm6^um+r~NV_AaUSI1eTc$P8^w;LhPIP>fewv24g9
zXyH!j!E%Tts|-=~O8}K>dy!&N5tjWKe<SUy^_ay4O5nY^Hd-RHtX>NHZMYs4S!V8&
z-wsXtaFc9((RY&LE3^Q+hamx!eF7tnpAj(|8d00>I;-WU)Mf=@9THm0Yqd$xXjmC4
zK~t(r)_4$uB1~Q$ursOAz4q`{<N=wN^Lj9E;2oGwE9ml4u~BFmzpOgc9`8?hQB#PP
z_kCoGg5pSJjWgH7$6F@3;=eZIGihJvvwhf*V0Q^Ztd1WY!te(C-oXp<(2`3o)DPSD
zLoUw>Qri!mcAHDJ$hUm5`{4dS(8RP+OlRgoz|1#n?}s|lgpKZn?+KF^zFWV6$G`l_
z()^vmpkcotO66C-z{dD4R%!`L3tOdl(>paE%RXzrAp3dcM!7!j!in}0gTh#e0}jVm
z6gdX}z2O@NdN+!N>9#gZ1Xl%Hcim2HV<Et_?jq=YPM}P14MvFBC5$<17T1;c%N7PO
zE5M>l45I~tGlT8J&1M4jtOj4=&M~*Qg>KUm&BkCgyon=Q4_CY#0GgLfW825Bz7Lu7
zN@3=_2+jsVAgk}2pH%5n*NC&}*Iquib~@Vms9oN%1DcFcR=*5K#5BION9?!QxPEpW
z;{(_2D2d0GIF1&a-joc&#d9~dhs9k--ABwHTa7<`df5D{r14A0wz*#S$-qqI?NK0m
zrjrmgwZCjtcp{|O6`Axy2}UxwXNiO8W$_bv-GtD%=AGN`phC8Ga-kGWCIW4RD?_6U
zGgtxEjciCdC+Zuak-$B{&;t^tr=T~R?^h41%SOqXcX}lu4!P}kQkZ!+f1h(N5hxHS
zeU$k7!V^81I(B5LqCAbR03=-P_biprbgE1Ewl7I^>;(P_MH>dFibr^=->Hz{e~lU8
zt?AKmn4y==PEl;sw~u9tvmpde=5yqn4Bq>J=bcK<M^Q~rHnO0<Ptjg6t7@g_8i2Y9
zagay`dHA(aHFu4h8FPo9Yjqp`IJhNBSpdz9JLX#TyQ^5~`Q`fow#>RGFp6pl1=>uE
z%9IC2XE+NMi#tBLdh%QSF3NS!mE3LiF7u1kzBkh*Zdc-D){W}Q49ps`D|`;XXU(`*
zE0(<IUT7Sxw4^#@<Z^XiS*OO7*8|H;k%JHAzwBP`uYXMdy-_j>EMrXDvbrAu(P*$;
zm0_t!<X>F>q^QXKdE4z1XoNyoqx)H4esG`uhd2!j^a#=-^>{_sPj!yjJLm4mC~^+4
zv{}Pb{9vPpHtJNX<l04V1O2cU!gh4*(!nPvKJ);nP@<W+kGR9(u%p@c_8qCIt#h`n
zqdh=$6OZe-%*6ST6(S4ErUV!KB+GLjkH4?1dRF73z^Pj`Z-LeV9epJ0u<M1UrfS{c
z0&o9OPl+=-nIXZ5iyC0QKP^Ueh3^&yF|hJhVI`<4jdMY$J_>+=2Gt92qYCW|m{4!M
zV(lOFxO}Po<d*|9m*Wtb&24k?>L`r7^~Ag3qZ@Y5V17Q$bNzH4qq6(q`}8#hgShDG
z6vs14O}ovwH&4BFO9zQUyQcKfohggYul@AB+>|<`67F-01`xRscktsix0$aPckfOw
z2heUAPBS<1_{;mN5cgCb%EczHeeAx>na&*Q)67#B{7tj@knu1rAZXTDN4Rav!9ao^
z{x(?M$~0QX@;zEzzjA1E!eIQ$fcuDdnwCYrvu?5J^?9qZAe)}^{VQeC7re`g%i^-r
zUG|BJ@F%%ZP&68|%x87Xn7r@JdzTTf?||^Nms#ZleeJg4*F$AV)5Qjc2k!^N7l0mc
zIie`gkkvw!TG4lQ=yLL&6Q#+i<&mGyXxnnjPS&c;u=wth=RL(P*K*`EL3EmE!)F=3
z$X{J`wH3*7$1km4nd3X?Y8|x!K@tH!8V5ip7pQ6qXiX2l(R?3MC+Bj+<V!3?q9F@f
zsRQO-xQdGgHI302QxbG3L+niILN_%>N%z9C1fi~;7u%15f|k$ZpsNhu$X%kgU)`(s
zhF+38r4F;u0G1d05atvW(O4H?BHj|9*&UN>xb>zIHEYOC(JTP1CNhe}0k9#aRHrb{
z^!}Zp26Qo|=4sD~c8=}mJD_D9)%V9Q%~4CPF|+FT|AwC)Kz7EnRE!06-hc9~Z1Xr@
zpdL3RuHpHF$iGRHa4X*SJ?E`mH1Xq=(vvB<Xoyn_8xnt;4-lWO@*y5rG;>M)n4Nm$
zOP3m=@?|GhbbFI3%i^=naT$AF&f=q((poYv+ji2!gJ+6HOGoBBmqqFNXaemS@Fb9o
z5JDNLxF4p=$FKjiH5@e$M9NaOrU_9jB_LXWuQ6S-yX;ff>r~oeM_aBAw7Hl2w^|0c
zaz#|fvIaLo#7@wUOU#~%lcb;v$bLKrk(56r09SMst42nV(i+AJI`(F|sCz)^7Oyx^
z>D6-#;WXM43M+CfyC+yJou{|sq@m7L>PLwVjw*)oFM7?{B{heNsxVUKX(eiVw)N$u
z^I6aZw`+q1nlT4FRXpz1mcG2$rn664+VlbsDY^s|hoW3XjVGH^)MJ+FC4hk&{)`hJ
zVQP&F3m2yBleu}5u!{#F+Q_8>$VqF91CvJ5_Hi8deb4OWwlK}cyrk=cmh1{#48xlg
zbb5?|e%E|#%Uxsr*9GXqel<N!u;kNVg>&sfk`jWwr?>#0^bu?7rITWH*IAz?u8l(G
za(^6mqawB{(xh!%eq!JMd>O?w0#9!Q72+B^%3O9Q&f^5!@4=ch#$rIQVU|8gnli-3
zn^W`3>C-J?s|bIeJr}_PXvkoODrmjz8O}h;oAHI!cZFe&kKd5o`m^6(IP+jnJHvIP
znE=T5w1TUx+M45@tq^&q{Gv+dJv~RgleSz{UqHiMTh9xoBe9a`9q0U7l%p4kdBu-c
z=+n#T4%WsC4!4!p!gimjN0Z%i=lz~d&2K5nh5wppdPn#u8;a^Q=_)Pv6FRTs&fiyr
zjBDV3@3R~EyNw})qg4zsv}W4tMS+;>{tT^;hgn_1NxvpB!9_V+mc?kWz2^JZ5Zca^
zNm=T`3VONV?X5MFQ%^6M|49BKs*82RHQXY)cG)+W-()CIHj1rJ=Q^uxDz{H@y}9@G
z$bqxm%IOfA`j|LtZyJS2nwTBC79opnvJLx89fNH=L@a)6KU(iFf*@5$E;p72I5h=o
zn`g?Wncw&9ysx~jWje6p<9=cCYrj}!O=aS}XgLX|dk-NDw?#ID1J|Go*91B*R}gCl
z?)8fK>`R<Het*p1{9)VJUhc|2$;#K*F->FeJW9EiQ*G48VpykRkOD=PwYQ2=LViU{
zcZ$TU^17v|3Eq8aY#Y1t^&uD5lL3<VkOvq?z%XmV?Xfd)Jizj;gql-P>%H?s(aA~h
zl(jr%gO;13j%N<V?sw+gc~epwYF-w;FzmeI{=tmeCA`P*di$Yxp4_~(8L(B7ZtFUE
zsHz6IPOdX5oT+$f{LbWI%d8Hm5$d@2Bp3GGf7|)@KO7nRpN^FMe_rSQlW&z`nIhW|
zD>k@!4WBg{n2|r62i*Y@*pLgNo6vu_+@0~TygDp@<z&&<LToTtWEj9VqPCA_N`c%$
zkqt@lDchbP;Q(d7M&dpIad|+Ye^Xeqp!gEq&i+u4vKI>`NXcM52}onoh|_wofIbu#
zu89m6#`UU$tuOT>4CN|788kF1fFHGqERgWj+a10mMenIxKR!MPH4T95tkpKro&#;_
zUl5^iS?afQHj6?or++=VzqWnhnqI?$gILS?&2=P0#GVCaipUB~wAQ)^2Nk3aA=9qk
z=cY|EaZ74!2*b6$f*H<+xOWoAcft1Sj)WbdjL>dg(8((~ll-X7=K$f04HImA2sz4z
zwBZvF6EWawfJ&Znv;H)d4{g|x&Z_<#lwDc}CvRM^^?YF;hxZ6{@=$iIT#cPtGh~3I
z`;PJ|Qy~Dt=tKc(Vg>P7kx{2{B#xzt(!;jkC>BzyW1ZxJk<b;4Aa{g?G~9TqtOe_&
z;HEfUnzdN7L1_9)A7JqeuptT;KjFUb!7&9CIcfbwki(Hmr`8_Hh=RSB{)i3l6#97-
zI5~=q=k}R|y<nZ87Qb`u3WTvmm`|X2fjJ#c$|9?U4rA_5VRR!m+URG&aFhc8CwPyU
zz>2&A0pl7E^=%E)7z9lntl<9EWXvKrJpa3f)~h&DByP<dE(bnE3Cfg4J;aXV%ff5u
za)6;RRX!o~4Hcf|^HBAEs%;Z@^5G87u(R!=AVX&eLRnW7n(=flFeCFWZ^MvA1Pr-b
z38Pyr_0n#*Qm~9@93v`wrqWUy`pcP<1O>2S1#HNrParecvkD#t%UhoMo<l1SrBwlb
zII>v3EQQHTnX&22DQJ8B#qw~&7t&lmG4nOj1aB||9x=!s^>RDaptN(c$O3T!<qO*L
zL#z|(Az1B6v=wc-zJsiM*VCO>$A3E1A&_Loy&NZk)S^nYq-Jtc)T#<^cb8MFZH&}k
zv>mvU<GC0O0c>fRAQC>~8lM79`2g$+c$y<JJc(t}@+1Ap(<nr>$l>MNo>{mCSCAyS
zjrpdyQ8N2hJPYm1qYMg1RhC3>od^r%bp%-}P91v$ShS<6Wwj7|v~wz-CnqQ28tcJW
z74!o_FmZ&7iPEo~!LGNzBS%V%W5Zq={Ei^&L$E7=$IjM`UJZJkK@T0yLJJu!U-Vde
z6`Rt7k?jBX6T5&Dix2cSN1x+q(xog@R74p)><_U5PaIq^rSBWVDG-{*Aj|;pS&F!h
zGgKn;X&Wp03R8{HIQ9J$4FLCPDPsB-&S+TYSS~u`)|Dl@_vKoGt@|HXHVBMl2bqIM
z|7t0QOkvagIFdIERNidM8GN}&fH%Gb+Wh9zHqM^aT&eJlGxYO}3@e2ShRhV(DIRyY
zX_M}Q(`TlFcwt$DM4X!^pg0$YbBU7!9*=QRR>Q0-I^4u}<`ZPyDs>9A!A>{Nia_`7
zAJoiBLN6YW-52VWi}9tJta%y->dN05hbiAT+cifegfLX6by2P9Lm!gg-ZcF<HF3E}
zpI`B2ssblGg+2v##xj}DGiY}Jz0{rJR+iGxi(4r8RQkbBtU2ArL_L`KfJ5TFxYys0
zD0ZGY2;ug+MhL}`@?d30-%y;U3lg4UPGQ|Orv%00{j-9K$1m)9HCMirs3$p9iercx
zPZJwrmM>nSY<SK<g{-UMz|u@v`^F0gSY!P~DmO#Si!75(q-!;u!=~VCqH2F&-B6h)
zFRx$jUb^*tw&Y-vX~2#f3C9p_BX70ApbZ3<ofG{V#!oT=JAZH-@8;!uHM5||IlM6i
zs32N0thy%R!bm+PpYR=S<Ml!ym|eW-^hH~Q3~-UIG;-K+@WD}+Zf(4}GvPV*D6Zi+
zOB18r=b?GI)N#aJbtu#R{#OYRp9f;U>W#deClUwqjx|^uKM#_J1<vW7X2L$qO-fl#
zBt|{A>643}l7w8?6Xm5ZGSD(F@GxMDi7g4=u(}@@G3-S)r%7Id2f8euhzVGMkb+Hd
zR~T<VFkN7H4A!brJ8HRfXYt={e%2QAI6SsEYj4ryymT|ujpo#BBTtCZP-d#pL@Uc`
z!rmrnN$dvCwx!@H3~B4fQqv|c;P|n>Y56?#dopXUI&9f-M%BJ49Un%b6#OjZ)x@Wu
zo#>{MlgBL|1)rKKz-qC2r}$P_{EYil<Wu!ib1owfs@g2?NIU|L7rvuu0+k@{an1xD
zAo<an)!b4s;a)Lo74Mm$!NHU<zZ&k4DVKB_Di!zj>ti0>`ELy(CHp_^G1BFNU~r(-
zYtTE)O!T=0z4DndE2A)^9%}T+Q*$)5Jprc7uk56-jqJox7^+BClqH@e00J!z*@2^&
znZ_hgBWPLFYgtl0^%;lz?>1~-Y5Z0Nvss4*BlzzFE!2p{%>8lF$v5ibC#}wJ?a#bP
z2jc-vx)#QA7^{gn+=sba2%yoR%P|K-)zLh}E5TE1<xBjtcQy!Dll0>xCPLoErrHN{
z7u*MI&qp+}B>PJPeJsOUhm{@2jvtQ{M`UAtaSfITcaw|L6Ztqk9f=_M)RMsc8P|yw
z7!9*>y-Rn$mK(xbgLSoOFTj1v#^?YroYz|*00CTO=zEMigio53cY-d_;JhX^0n9>^
zz{H=3rgoLwfJThKQLzX4dQrX!S5pYV(vvO{Khk{9s`$%gZf+zrzyO8SmeqJ0m2;D6
zS#{wNa7^CfH^pS>hgDsWASj6lsB*`9XOzMZQL<AjJkWQNw^ejo#p-EHa2}onu!vXO
zlJgyfvj@^VSG_w|Cw|t+-fsnux)3E<y4Yr%D8eZNwu}(`eG(pNduOnK(zmj4NXT&J
zwtUv&SVGul%q^r*ZF-Y=*)wB-=6#lZPudTGoxLBN3CRf3Yc`~;m^A>~t6x}f#`A!)
z>*uKtkNXXSbC1CbLeqB|t%VMZ867yH_~Tmuhu%R%%WB-D02p6Y3^VNWo3h6g<z?^h
zv3@ji&q)<P&UrS2A-&z{fS*G-Av!|osn2t^&c!jCF3RG@b#q{^F_v^sY=8RdvIO;=
zRrR1ex-?FSY^(}z^R<vS`2LMMPqgbHVi&@R4LN}AyM01Nnu5(7NmG1l9EBd;?JyJQ
z)gKHm8YGg6V4^F8-R0FmVXq{99aNs9D&InYRE1PJsX6c=fsSlQm3;5NP4%`vvtl4z
zpl=TvJ2N26$LUwsOx@mLJEsme7o(mJ=UXsCbYoT%QIL3F#CQVR<+&jtb#Gxfbj!Aq
z6g%s)3YMY}C8i9K>(dF#;yCv=;}ackfDigTw4jA0L&QW}w-)R!>c+@g+q;F4v3<F2
zA6?R}Tf2(&^B$_KcwB7I^v2?xBr6y}!Xrv8+m@jevBq_)lx(V6l78lgZ($d;v=2BH
z-q;J-c}&y-18Pq|`TjeK_Kax^hF82P!`(`EA7!jyDLupB{ugfFK8ZGz4f7CT&W*K~
z6Qn$e;32riWiU!og#-~n&&1->`h*^D!%9=uF~>{GO>E68id;r4@1Ka0yYYh~)=8rK
z6zby;SPBGhPWnlqvzlm#9bM&}2U)x4I|y+^7gONaT$AY&<kpT@Y7dzxVnZ1&AElr1
zTzP3=;5BH81CD_OoAZGb7-_+|UYcl&5CNXlK6gBmq<rsyPNbp1ZhlFIYISuw?U!v#
zl%&fquubps<L>6!q`)L+&JDw66vYM_DVcnBK<8M_{Gp9qUTO%9LL7MBBczdGQ-+za
z{(h@4;DIH}__iiRfyN1xP}lVnZ2LB>ih3PkRfB32v5Tzs?f{vx5ymC(CR>IZMJ>iF
zcYOXB;o^f2MpoZ~8z3DI#O6deHl`YT0975Ncq$EhhT=ME>%p<fCoyKw8n7LxkBw(2
z0aeqexZ6VIRM=CWn?AQr-hn&FVw~uORok}TN7kkMT{Oc6herfUPc{BJD|<I2e;<z~
zjld8}n~4K+sl+kKwnooD1sWpiyG>Bsi*2mGEHq*D?xS=Fvzq`C6%WGrxmDV^-dfd<
zHo$F|u(azhTVJEXZ@UgfOC1V(FsHdGk9EK`coJvz0TAhTmGMjwMb1ReKODH?Cvg3~
zX?@|}=r?N!=W5D}Rv=|2P=#C59{v{F2}`;qpO7wt1WQ?QX!Tf{C}Rrw37zw&M{&%p
zSq9c8?V*_wqm5qjo>1zAsU80$u2l6!f%9lEd=bb4ZT{OKHBE^EjKH+wozDjFTg=(E
zZ_j+cIX+bG@m2>^oLY3@+m3@8H3HXuoW9B}I2+)g`Y0XsH_5jm_PsluESdtUkV(p5
zyXn?rssD$wH;;$<ecQ%IL?IJpU#3EoEea(Jl`V;qJtiT`gjALoGf~JMJ}Al%Qc06N
z!q^qbzVG|ajAgvd*6-^3+|TQ|pYQ!Vzu*1({h5ZDd0*FcUdMT!$8ns(gjY2(bQv<L
zj&Rg;YC>qSXECixv$SgPf<$Jhu2@|7Q=|vZ1#Fy!r(k=F-yCV6&a&1qWxq-3d7euz
zm~C;CyW>pZrzGR#s9ssxNgwPShAX1xhb_|7#V!5`h0n`Tr5G=?)sZdnLkkVIYooD>
z1a2<R4;RJOUmHkaL2<9<KZKAhqyajA{)`XgLj%?GdL{QWH1DWfwVS8a#7~$Z{J%>3
zpdyl~9N+<@laCIdO0*-hBZWpr+`e`cYjU{0e7sXj=xBv8xWK+LWByb_@@v}pl8-d0
z;oLj7?{PVLpYV#C$6j`2j32_b>+$J|81Tvad>H<a4TI~*&Ba3}QO`jI7S2eZpY<?z
zx^1d8G?P>n-X3fLbQm#p`bM?8@ArN(cIb<&TM<moU>bZG!LL+eDndtgB?e<{^xiT`
zSiN8IOiSPQ7vtPBufvfHC*mZIx#ji&xspVy?%Op?>@m=zuv#~LgP%bCj=zAcA-;uS
zHfV>L?HUWK<)j)Mr^onl=4hPp`^s20$1D4v?$`bKW4b2#{pYALAw*X0J<gm$7!N4I
zv|_G9xl^Dg#`m2p6ICi((tdi-CFh~KdQsfMpw$AZ2?X$ysd*lWwiqqa>;C&e9wITZ
zUK6_;?u2CGWcEa<WgdBCMlmnTdYKQ-v1qRPS-`DP@><0GxUu+ORaR}<aFSo~%m$zJ
zq6DI6zE3r1iOrJoHKw?5Z4oH~qdl_%&$tm|RrN+I1@Gu6ft?3g+Rf5@P$Y)r2O90w
z=4E=T)Ez%v$y<nT-z>)-M@M(zCKK@@zFN|X>zSoJUcwJOrH{V}au&0Cj=?aJuu|Ce
zEjwh)<5QmJXzx#XW1pY9c}7?=^Z~Ynh1y3`1^jt!35-`?L5UXj$mCey<h1NM!nZ;@
z!@ZWhW++ZCP4%(MW5)`c>$)6S?8M}6vjNUl`m)OpKj*O_Y#D%V73u~uVW@$o>Mmnb
z9kll7s(;D1l*@@`%DlLlOf!zY0lqzf&7U!=HwXVRWAD)Q!LPJ=VE2=oi!)^!MD22W
z)<sy(E`QD?PwYDz|C@Nd6b_LpK+nd69*aANDp9AWbO_-FXn$Z=VT`Mcx46mAX{jx!
z=wU?Y9OUbntM~D&oP)u~bFmVF$j6x-3#e>g`WBD_{_O&RRM#G)nw^Jmm(Rhh79!R_
z=DOdq7w1Zh5y=b9f`U9h>p6d_-!w6Lr}PF;8!pFXlu{9>js^PBV&?Up#)+L)B}Up)
z?Fteoz&*1d`$w0=^4ACs>y!_iVoCP3yqg}U!V+_jX!dxd7?)=|#y@y3(s|^~%rP-+
zDZTYCOE0l>%9M6MkG?{^OWL4EQ(YN>pah%A39In+J(OIPG<x-Aok+?PsXP2S%SMC<
zq}v8`dbgqDx%$~92fk;_Bc6U7LsC8!af89quxF5!GWh&nj6Bu&@EVRDI_p$hl8gVB
zg|*!SjfP*HjvW^ZHLRl$nqR$3oOX#*b(eY+W+5yE;x3#lLl^L$7Wnv`CRb0z#%H`r
zZ*Q3|eGAZcofA{mnwOE4J?P<3^<ryJ!3hH{vQRzxEhu)XO#yYI@7xAw?x^v(2NRNp
zBSQOfR<$E??I3`P*f3DRq&;WwW&OUA4IyK`6q@vN&m&(Y(g9>ckgS+o%2Oy_dxgA4
zQ@HWeo0s|icWmdZGiL92@>4~p5Q2+LP10IQMHZ^zqq~1X<cV(~4f_uK=F#BIFz{eH
zs48eI7!tn~Rfnv7SsQJUSK(Iqdj41xZH0MIk7l~eWP@@!8VYPTr@gsW9K$MGZL&*}
zb?<*z`}Q{*S2gcfjYhzi%0<F+!BbAI6qe5LKky>cInm@cp%pM4!A*VP^U{DNlJnff
zl;XlPWQhkslvgSu^u+<qQ``?Pc5>ulJHdD8V8v+kB1}jp|G}Vxxn3<KTDd9o#w|~Z
zH!Pvf{^fWpkY#R-F;b@(NW+e6M)NaUwh*gP_m`dat0Sg`Az!=|ovq(RvKuR(-&eCm
zLeP1M)2UG<uM~HJaaT!=o3i&xzemnJ?zn5#xxWAo7~^C)(-B5i<`OkSV~$?MG)5WG
z|Hj;g0Kw*g!-E04M(Ood-Cv_Fc$VM3^edoEjlBzX4C4k~VM4cp%%?`)I+;`!lm8L;
z(cLbQE~>v6rDDhPq*j;^=^Q(ikewi+ZXRwabc_mer{4agd>izzXIWOBy<U*h!%#!7
zV-FUWgF-d>TIpbMLoYLXH7W+a)Eo^6U7z1DO%$l?p8F=>#y-l*mr<M2jx99f`1uD`
zF1S|hPdnC-r9}ukPFA8Md1`x3JvA8Ux*llt(}2kpmVamFe-+Hikdvho<e5nVbrF{|
zFbB*%jPb?z=7ZRa80jvItJ<qd&9vu1?P+cEY?ESMwH>zGt$Pj3gJe*X$e-MxP9nGY
z3ARudS*2}q7j!@Q0_yjYestfTtr%PoZH);#%F=}G!j%p*5QpMT!o7t&V#>>5djbNH
z^S@feTy$HXWt8r1&BQ|r4&Qw~+}4{Su-T3bos#2kYU)y)Ddrz(>8@b`RH2|ogqZem
zkD62DL-c_GD|6-OUw5zESeVnf2vfF20C?z%6K8ggA4@VKSw^URAjt;3PUR9ia9v$l
z3VZ3P_6H!CMl|(TsVJ0&*8PEL`YCOV*b_l7e6mzRuIEc1=JMost(!w%g{Lc(i-MjC
zrft~+mO6r*Hek9nWq0u-Wnp#k!O-^4^MoXAS<Gp&X&q5u6hb#quMDddda<ek_+Z`@
z!oem2V^II5Kc4Xs_>!IY)GQPu+CI_%pF56S*O^##O_i9m>a?KO(iNfD&vaSxc>HuM
z=d!Vhf$3Wp{n!_yEVE+1{0v3RMe;2PHIjBj!N||~+_jvC0HL{k7^BWVRc5MVjBwA9
zT&2Lq$>5HvdiUg3SiaF$qaQ+JaT){jAHgn}TyAZ|)qgBr0g82hV8UdvOI*ta%hfhR
zpl*RAGz@pX8~5282z(Y<DXNSzb4mtDDH3B+uxMFxsHjU^z(!>ES4=+p1^GIkU33Ey
zfP8tuF#4g-_KQULV{hncURav2(!$FI&hyLW&#Ho==wG_9ssRLD2>metezh8{b>vpE
zR8<^z#G8-Oy2rK`a)S_cL`X)x85v_MT~mtAQcK9Wo%h<?H&ACwck$AKBvAcWfb2fB
z0b`S7w)L*)Yt#rcz|)+%MtA~&>tcUucWHs`fH9OX)v;>qf1IblNzx=m23F)LB)t3x
zg~v+}orqD0;@odFKwddl)|K<mWr18#xhty%tf7W<bmJl`HHp*`b&e?PW*PW9Hi#$d
z{aCYf_%sLi=e$wN0<1j<xx!2>`{gq2<B<IaBdH}P1SR>Oq8h*Br^BVUS-;5b@cj=N
z<Q8(+oN0)PA+CUJ^onrp--V|h-nf}f9CdNxcg+v*g|334?4%oUX9ldpyR;hh>OZh}
za6(tIaT+vH=E1WkF4ItM?NX5<jPQ-rJJ}5Lz)p_!IsR)a0$e|9%7E9764ZC-a`QN%
z+eyWKB@n}jBtCpb?d0eH!ZJcQ)v6Ib<lSTv<jI9{Cl@rm?Fp$A;HjwF6|>q}4Un>0
zNu?hhp{$UScN>C0e?)?_H@n|g;o9?F#kEK-kEpt`XK_Z&=LtWCVKg9Gr4dYAX_k7q
zikS5F)9d=wcg{EsMs;QSDy6$EX3xz|;+rs`xdW0!Hhiyknz1aOVT)G_)JnJgc~awC
zD<~UcTb5CqJU}e-7bL*2PME(LPunt<{V%9c0-z%2U!cO{e}Iap?TsN0Ca0%A?K<5O
zbzmHl7-{gEa7kO*+24CDOW>URLr$LaE2sowi3~l3@E=&Ck)+fZ8#L9_wALIF&@ZU|
zK(!>v;3d;;-{;J@Mm|~Aor{!Q@TTAGAf=|%aUsYxx8XCFy0S@u>B%F%*g_?5A8upz
zqg(#fEq(=sWKZaiA1BSFRG_~vSS0<tqoprcfH^fu40kvMMLr-wk1kjBy=9-@l$3(E
z<oY7vHTwgSZbL0o`2EYn(rYo)xU54}`SRD)vd?o5Ej<M+qAnoukL<$Sg}k!^%0z`r
z7E6Vm=X;+y_EblU8W%X=_`yT1(j;aebwJr>_)_lN-}^7S0#N;vrx5UA2woR>J@$ya
zU!X~hUtJ}zU=XH<qsaqIP79i^2mVNvXAnr4>uf*wW8pnEnN?~1mx$i~xN61~PR7A~
zbP%?5u*jk1nDY?&k^lREOWoc=Xtm!?2@md;jcQ4Ou7C0-?r-_02!BLX-zj&|7W?wx
z)aAWz^nRkYj=T=*$?BFV!m24bd%8cF=i?NC73JWb7*o(2lOO|HE=O2Fqq|oE6e>g8
z>6f~_vV+kLY8s=g8~P_Vjp&yKMi9lW^rH<^@hu<MO44cZke;TI61~!)?6}l6m)xBr
zYp7V90MdmRg<@U1xzd~guj%r%r^s)VU6B!92D(3jLa4jK`#3lIcPVN-iN_n{SYPU{
zg3enGAe#IWZYj(r7J3T3CCNM(LVIEl{hl6~GL84;sO&2&H|X^eD;~WAA5+#$fgMR>
zsAul0lB2{iifY8Lzo<|CYW6Jomb*`FK5kQ<|7BM;$8lP#-HcI2^x#4c1x@aZmwdlB
zuu<&!-Zm+8Q*7S^>2U49*3JxGe1*uTX%a+$;&0$@^?8i~TzJ9h=Wq5vP8frmyh|5I
zZ_-HaX_|^>A71kN(I~%l`?ZAbV6t4uZNCH%v!b8^*T>G)J)f1PeqXUZNDU*r#PZR(
z$Q>;TJ|%dzfWQ^|Y74~`9)1ghn>H^H1wmyXIIz`2TUm=TH&9%vEAPA$`-RN|pQ_-Z
z(_l3CWCvX!=+c1dl_Dr&%8#`3u~v|ZiG7r`{#{>g8wpYo0O>F{NKJV!G|z&f?~;|=
zKWT!8AhE~Vee?z(nwx+2@wPiPtRyA1Y~3kE*or&mA}`7cY$#12t7D)kK~Q#{n|i7@
z@|kM>^RkNKmx+ajYo+!V>gnwrK%KCA(^V{EznXCNw=n6*e#22OEAyAS>vEHL=(Nn*
zTYr*4{eu8_y_Ki;bb;4NcfwsS&pUJ0+z-WOLHo1<sAVQ|fd`rENcvk*Q=jRMqnG<z
zyZ*o)=5^%?!)R6w<EO#CYQ-4$eVI!#?EHMr+IsH9_{m`{m9};)cFf)+*jKgv7eTA>
zw{xP6Q1QhhiAGvRsvE?8NJm`MA6N-{T4h;C=1=zU_OBl5=Wodj@*M<0vRi`pU%OSe
zTX81X`|oZQmrzzpH2roQz3x2}&*VG{dV^6ZhoskWQ=;U{F>kfE*J7*p84hv>=Wm+N
z<gLQV-jj;)CrdvssyMjaI4sqxaHjS7c5A2!W+!i1-PUAVu+01?+MhM@MB2&W9}QXT
zz<-NVZdwWC)@zYUYx_w|Q%)#l<2O0(nUJ|ugo=GFjIpF$am;EYrtVg;n5?iAC6mW=
z+O-MAg+Y*h4ev&)1_jHQ#)zH!&1z_H@rKb}{8z&7iD*!T@;qxaAX7|&KLhaYT<w|4
zxU92wh1y+f$lo>KjtPwi1<Hx$jkw|dF!?n^gyz{N3bAkBDQ^xXyuLQLbD|19xll(u
z3K}k@<zZFQc{N!df4$eYXqJpQEZp(+M<%>bH6D_;YBC<lq-c1CJ-;xojo6X18<zlo
zx(nJcNj&=zx@_8a6!9uJvkCw6RHXr(*L3=1mUFRvz|7^IIk_aZl<q=~o@3KhtdwN%
zOE|hQBkl3TTJX(rE!T!+zn`D1zes}CQAy4$5T1ZA?FnA=8lw#De6(IvfB16>f7BkH
z(!I*ds=jM!{`7i~)u3>PWCE&n=j~Wi*sQu*(a-(A6s4l9U*%_mz#x7=l~xW;OB*AX
zP;(I@AcA`F#kqoP89IJ#^GCt!W0U%G@s{T=fbX@^QKNzhAeq*pc+^T~v^>OBM%Y?-
zE}JN7tX`WdT(cjWo7D*ZJfO~VA{3<RV=!?Ozzr?40;C~k7hbqD*xd4UlIhnB<BMF<
zcb&q|dFFgrwQ~cNsR;;t&HbNg+NJ*PlSI~*;M|Xm8euzadXlugpzBDDP4UZ2E?)B8
zX~@{q(Gl&P;W%7?Z!EwRkXR|c<kU6E>4i>B#2v5qA#CO5Z|;}JykYkWyjRm+0-lQu
zUH4n@mJ4}4jO|-TVt-N6_K)+K2kB?Q8-5_C;C%!s#L!jW^N%7%{-ahL+)e}kS*Obs
zVs=|~E|v)vxl716a}WD2?|CfH3M*)!NHT+8Q=lECUty%qM$I((UD~+vx=b(MHh#3&
zr%w3Cvc|xRMUDCFsp@+jx4OL+zV;S4^VL1N4^-I^zLHs>t8_1FuY&?-WSwMWF<F*0
z%HGda>D|?C`*<i!pLD(Z?7(RjLiT^pk>>xe9nb&U?|l;4r9a(4P$@|O^QvM$peHt2
zdvPnnrVO@Djx9+tih1=f0e^*Jx^O$!7YE8W|I_@`w}<)5;Yc^#hulzKsaWjgRWkCu
zB+l)b)58=%W78~)Vm>#eAHh%Rf|5o3@pxK*hbiQbtc|B0MQ%ENVBOmwbZ<Z>+!3Tg
zy8j9zg-BIJuPgnpSLAmx;n-g|IF@x;u)K!53s<zRB)R$C`%Kx~kMj$oW*nae=9TDz
ztjsGNpsh<eTm`ho?x;MAy=bu7fU=Zub_&OMu7upSHRG6##ez*&cLG>Vj0v^`rWAta
zwn!kV>Ahis`-5NV`oI0X9%|hRiK6!s#z19CYCT#2PeTM?cl1wIegE!CmH)LL7VMfI
zV5&b*Qn9^u*|@Z_@U3RWvBcx0somhm^$rNgW$L1*pzhc5u5;pAqOD~SC(frz$D<ya
zcn^q^qt;fn`Jpd!XKNJH{6zz2j{`&AY)UzlizDMG&&YArAB^WKCyZQ<xH4_7tVU8f
zpHXf>2JjYmd<L8pJmuU|%GGE5@yMrU7SL{c8}P}=XFAxy^7oR`$wzqqlchBxnCoPC
zJ(AxpAi6<4w8gUgLe$-mGA4+s@9|ualL^Ty#3|cdt#D}2rm@iA=+&clxLsE)=-!iK
za?>1Pr3$Pz#Zi=csz8+CynRDL9ral|mIDu+;%G@U(pC}%L>3p|APfGPdn{fsCZ6hN
z3)>_NpPhP07l0;a5c>@zJJ{o~6!O)X{R6OP)PYJ|9q_~*#i%jzN}^rg#41@v8QV<C
zaQ<#MfXz(X%bS*IG<h_g7p;Ubn!cD7ls~S$w{So(C~(rUrB>^zr_7B1(=W$fHqP8G
zx;e4>x|0`p#WP=jxhFu`Yz@>8i=RA;`t7=wM?UQxg)hmAEq(vZy+CBU#p88ri^!+g
z?2%&w+vWtdF^(IKe_-0(;3mVQF&!|bKGGABsy4YDuXb-1cF^XZo~sdoBzC6_;z~>p
z4eEEBey}&-XzRQ6m3!jpPr~BXq)hDu51j?_$$_kN3qx7Ze9_S_dJg$0!rX<ZX>?F{
z4n;2}K+h;OB+t(|`}aKo#khMaxLlm>QI`F4Z$!dM%68A>Hu9Y^h|}9|Es+^T2P~V=
zOzbZ|91ByW<@2I|k<(QC;g2Z?tdI(|d$}!T)>i9HSP$SEac<BHD@ND+?tf6VChH+@
zP}vO?S=LQXB91uu3b*Mpz8_p=c!tCd#EK$Lkgg&3R5tegfsJ%9;{#zkh~Z+U$wu?U
z&g8@|t1b6Oe_#jzJ!^1paYndOS9<zHEXcr!uL$$q3ruZk@NoXXG++!d@YxhV%4J6O
zX9Tv;Gl+NP$ZgC6)ejQrmgH+!=$p3hCu+5@&7JI4QN&rBKIFe5#K6`mvN7|jYcz6I
z&KEFQY>DrcB<!OR`&UwBax5QK(2ovmDUd}b?ddvX`I(%w<c9cW{c|kaL^|IF^E$DC
zh+3B0t$uYn<t|JNFl0gD0AzbIsHq9U4nK`BV;o*W3NP1Kc<g6>zUB6FAV0>VMh2!!
zCz=SuLCt~#DzESMc%iPAUzlkN8`78ijb*{8qoOB-s}Pez;?of}y?09zuZX=XoMl=h
zk#8;$u+jnVrnYMA_5ClbB+mAEvp`!m5PCWt!Rsb?qQ|1>r7>I6WlzV>$xe>#c91dD
zRvI*TA6UYQcPX4G{`xiKxJBfp^<yw=T2i|k6x~1te(HABa<@vr*;zg-_Vw5I<Pd)7
zzy#?gJpNiEUN9yJ4^eDQp@7{MJXJJ;$*KT*0J^UWb%WY$z|T{W6r65!67Tv@WY~yZ
zUGf92#|IJShaj2aqI&4Ej694|;7K{BfHn6=Jgf2{y`L5qP5EbVN3n7NNi#B0_sD`#
z1U+WKnTFqo6>%)vpO8FW86DFc;jXd2ye0DP*M&4IpYG(?fujHjvQgSo92R6eL2g}T
zyy@|(8Oi%Qn8C@4-i#*4Gr7PnkRG`Nj=2oeT?c_rrey(yP|U!egGT^OYvhb`S=<bY
zd<}t4T-R%BlDN9(5d=?Fg<QQ6Z+(v9#vc%06P7qAS^$=Q#oa?j*kZ?O^#6&DhmIgt
zbR=3p%1^Q;hEb7WSb4V^=^pv)qIAn=o&*1Uvyl>we6)cow%7x?#{RPIek3<tyB8T}
zyyQq{yVA)q+lRZ1BbipCX+ab#Qdslah@=~C;|(YCiQMiM^I$xv%!Ezhs2fP;b6M79
z<LH323t~;8YCUbTTKRh+5NO{?b2Ls+j#T`$p+NQWw@@RIW5`xw>coqQO|!oi5RMn|
z43S=zZ4Ua8Gwh5~ZHZ78373u=_&xtbVkV2LR}L@R9qKAS#PSz10pd{aU$L9LB5ph5
zDl^25g}^MRNv8K=dl8#HC=h5?FM|CGPC@PyWM6qVkjf4}V_uh=MlvG2fM-40jKh(}
zf&1By(E?BE4U&48u=xJ}C~c?zfr0DFxWwMnfbqG?ni^f4JNkZlz~9YSJTvG4x*p3z
zH>!j#ljRziip8N9KR*8cZCx|cpzW9?+ehj&@h-?2kYB5aCBQLLgOn-LoMuw~Pk+Up
z_pzIXy1!ZXHO><S=mt{?=z76!o0UL|`Q(Q&W-b1#`VTc`B=Pl6k#5+d;QbwHjzVp#
z{aDFTBB?(II1L#|ntKpQDns<9V3)~1uoxUqjk0xhc?8`w8%pGQK1`wPaw+PHg_hFW
zEHk2~V_h2_K1sV?5S;t-&3bV~fa!b=sAkWnf@k>}x`>{>xMpRhURh~>TdS_*TWzem
z^4)Cl3PLb&<vB?|37!3D@BzA5=cl3g?im98yy7J}&d3QBgWHb-36T)<@R&`VVv*+M
zclr&#i+T@VY|Z|fP^7y{=qC&bQbtL*HJ6-gX&!0`)5hl;Qv<KHCq5-7-#;|EkP8bI
zB7cr9?LD*Ym_`RRm(Vf&1UDpC{fSJP-ols}GE8nqf(SzOWckGAJ_uYPC_y}x6JUrj
zoDPm;axuV#5sej3jZ)J#X}N{iB1_|t$qZ0Ntph-EwoO|cXw*dG8M28_wKL+5K6m6b
z3zTir{;yM*S*H+?{FPKEihZmTV$s#IYzsL)>H{QFv=ML?u9HeTqLI$xeq7dRHlBon
zFTcAI`MLP8uLyOg!Bl7Psb$(+=b&j7WDmX11-aefv9Lqvv!#QN8>$RtTGIo`?6LCs
zVpg_S5;N~LFIee@Y(z0KG7l)_xUDYuo_C3Tp*?9eNoYhtvQ}#>*#Rr|g(XlVZht}b
zXhMrflLY_e!yIFT@$Wx|(HFY(ZL(buH|0nREfXZ;iDAJ;iF5s>q9`tFt<t{H#O8_N
zmoXpj8nA&f#~WNj3-F|;p3;b5PyrB&%w|elq=9l)mr*ZrqZdUtdAtUeKacUQ&n7-h
z{sB8~Ru1%O`B4f`OTBICQ~HgMP>^+)=Z=Y*UwK8O^%*O1eFOHV+8_T~rS4~g7}H8o
z89dz^%oC_l>v0cC*;aj5Q>5e*3men6(Icp;mxXR-8AdR0F98~}vFlUf>O~iDBD;z2
z^Km=ifL8vMg%8V7grO(2t1CBFBu+*J>{w52N>)Cnx4!;;=DNnfufHGAp5&E29eno2
zSMvM(M=tw&AGJIAbKBYThn#wOHk4g%eIptrggFILcR_7wGWMgy>)hH4PHh#<exG)n
zR^_%H0n;ym*Tcr&5&Hrzd*t)-d<;_a=V@t`YnjnOeAJMLbegI@(P7-9kPVC@BOOBz
zg!A+2!}we1m!S&MLcM^MrGpAtJNmud)$j|My<Rt1Pl<&dB-F#BfYUNO7B6rm@)J~M
zd|vdpQ+KEB8tU46xn|)i)XZLm#zdsF4Ds#bA%@#NT=b#$<xOybn%HxXmXtn?Y300p
zlYPbIZo29F0|ampg?caRH&ALxY<ChU=^b(_n@Om1W47&1K2m~*f3lGat)eVWw;Gt~
zE}|FXA(>DwMv=BKetlU7)Ww0PPv_Xb>$0(RRx=xPfh3SHgb`OJz;2jt#~I;^aP;&I
z&@Wi2N6rHK55f|HqW{g9FtkwT6(L8FLcBc6va0%D^wtbvLOW<}sC_7;tCDgf$VLIu
z^m0xZ9ZMKDg``pRv<whNP=`S8u>gB$U_*w^M_`(m8iP9w5Q91`pQ<7!^U@Uaz5vVj
z<MGCRan@dr-(C>{{<+|@XEmvXyWyZoa!4l&)>XEI`@X0eQ@>T3*zYSQ{w0SG?$nND
z#}1$h;W4;-6XFiKp6|!e+WcTO#r69IUdlTs@{->iKBEHa&0n$gDAJ8Jb&5O5IRj3<
z9-&G}px`ekt{xTsho9B!AZ(k^Z5oCLYZi**FPKJ~3J$vss5rS~5Kh#ck||R`@6EZK
zZn8eCTgRSqK}5z%-`V#9cOGbkFe89vq|%T~EwW54b&B@Za_^nJ_PsQ@HFYQ9-Y;rr
z{cY#zf?%7rLkFj<U)n<&6ArH*Tq*yx*YSlv3@ZbC;BuJK7=>QG4IQ%I$Ai~A!-G;&
zj>l7krfdEHMJWhwU)2kOU9&I_B}Uz@oETlVP^7#o?A+BV#v>iX2c{tDOVVUD--Kb8
ze0ERCh)<bZKyfKY=1&6nT$6=qC-Xire^=@?Ju4cq@xb~HI)o$T<A6SW{jH!)F_|aH
z_2bf3ZeKF!vYyp{C~UnCK8sF*21xPx&>X49j&!fGfO6>i5#9tIVYpc>WtJ1hbe7vr
zrJMnKGRF<c#uQx|=Olj#d!+C-`b-?{TFlwx^UZToJjy63W;d$HG!EygBs#Gql9B8m
zl&jxVP-+(-Je8=OnIg>jWF6fo%Su&*oM=86Vd$ebmwU~*p3h(Ann%Ph`9@)JL+t<%
zm*7^<l6ENsvcR+H;JfQ+oBQv4Ros`uy!r}b7iyGI9R<~mWySJeuaa8xPLm5Q5trop
zwS}1d2!1T@xI#w+-SNZ6o19Hg&feBnUi@O12_N17ki1Tvxi3J^(VwK1QQ*(ZT`m|i
z7a~pqw3(cew<mip^>}X64hRhY9#hlfT;QyErn*O<$$;OwT~K8coEP9UU7F*k@%)#h
zec<5!C<9GcSIFkpgx1P-{lS^au#=#0m}Azu4EUJ`GT{=8uejM!IvNZzQ2~QYi0Q4&
zm%!Q&v8Rp%hB5;&c<MpI{3b910IL;^*wlo|nV?6ci~7=YrTp;EBf$eV^A`2xHf5Lx
z3je?)JECzb@!R~=Eggga-5RWc1x6Oi{k$wvKJ0twh*?d-`?MfUWyL^zx@sy-9eiR1
zO&OXt-QijDA%cCtJ1)bwo&~-_X8Rw-T;S9QJx}mp<Pq=xfx)nCX?H>9e>u>rw4%rN
zOY(g#qvN1pDm-|88NF3Joktfyw>TV&GpvG7t)RGgqE)->%BYhcnqb&w<I{GH-G*&q
z9xA_DV4M&QaNx*xQ|d2}t_U#*cawZC0{Mck-stfP>*Mf857-4gb~;WHC9smremp<(
zw7n*EA*NKmLOQ$X?ONvo<DX+z`1Z_z?Azt|Tj|&Pb`BiBEHyx2D>^N^{A>0wN5wgo
zXQ)qroPC9Ht%BSh9$=VwJJ?<id&_DW6zhQT{M&%NKd_~4Wl&QPc#5=^+v{Zj+%UUP
z<qATO%w8HZYUSU51S6$$srM=sti2OXbmu=BloZ8GK)OtX)QCZFQiN{jw&<1EjNuFG
zUViZ^?&AJ@ErY<SqK9N<#2kkTr|gkl8zrdbS9CPhxzmxR9g(Nk<5Gg&lRe7fH9dau
zgLAmE&54Aoz{>fK{sCkdP747-O!zRW0WtiniRu;YRwi=0oZU?!8zFxu<fT?|8Z&%_
zG{NLt%Kd_;UCYVuUv6th+BceUv^a27(hy*rl+T+d8evthsfO=~?+Fna5tQ<$Rmh;X
z<H7&jbX|T2vPGUhupf1~U~~1Ns6vFL=*d)iB7y_R9uxuNSLZ0p$p55f(E+pNKWmof
zcJ6*YFd3&8tP3u1Pzm}^ZvRd(?I%t!=8qf9=tvZL)2!(ZC1v5`2#Mvc22S}KnZETO
zgkYHJr=PXFXBHZfyz9(vvR|I>_b*F@^!()l8RkJj@+0eRdDR2=dg$qG0qaN955D}8
z^%SX|RZwu2d0>L9KUetVlRM>EuLg0`0|trg-m30>Q#v-fD}UW6?b5`rJ`cv(Zg55-
zYa}aMMzF_hZK%^fYLGI7JfbTRlJ>3#V|F&}JRFX}XOs@>jHE129({@%5qs&zIuEKu
z(n_oLAC0EvaM9yRJ-J1tC41ZFT`v)&-e4hC`b7XjR<tg=x_al?Hlf|w{QA3O4X48P
zoseCGiLscQam)<R1)@Rf90qPifcRYa_bO01xc^<2JU<7+e#6yC;#`6cR>5VOf`hI{
z+~`$uaVvTM*8d4-OzSNCA;%ml+d-Cuh`_jKx96?|s%B*8{Zf`Xa^=jE4li^Q@eTqG
z+HOMzbpp!V$HHdmHf@Ln)a6e0dl6YnSa_#rNxRChg8J2>`1@uC^;fFiz`aW8Ti8*Q
z$Vq69rVr)azPsY-d7)=A2A^!n=k!B@hTh$srevulk%&#|DU+`!d*&ar0&^SYy()_b
zyI9$NnU{Bslur6lQ>8lD@ePOwYF-lx;fCewy1TVwVms$KX?pVsGE<dtg!vm+4Eph|
zDA<sjgG6^$_*}MSb(J%_U8GN&GTm^RD2kg*sS$f-$H=E6eU;6qeqNFtyI&0uZ?P0J
zHgnqsA5jB&<sUI>$-Zp7ODGTlNfsM}3x-gzbZX-U`;xn3-nj9sho9dkzsT$JTQ61*
zz^qo*FhgJ;9_@7OCpsfuB)00H#V@RzYYpB3+IVJQ{w^2upc1)gVH#O`k~E+1=T&M7
z|J~X0ZL@6fh7kP>nnWS~)%^H_T~prxrMlU~Acriw_pgfnKGw-GfFs>PMGYXikuKGi
z7cX2hU+Fm{wT4vt$s5fJP{&&=xP0q8RAo>R+PsAeZRomF+dp{ZHaVXnuM;QZQN`C&
zr_&3ZrZc{~cz<YeJXP4^^!m~4sA9i6Y-6)bZl}<~$?Q{>&LXS{a|7J{u|1PLVBefO
zbH@9kMVeL1J=xbx3raf-1G3%yY>d<yMmCdkbi`f5SzSW8XV&I%#o(PUHkew&)mtL;
zWfG3gJ2N9uF=zS-<rZj?A3TRnISYq)S{dKoJ=Jz&fQ#Nm+?;ZQgeAN~K`EYoqDzTx
zS@)m#HL`w_5kGi=kO<CA2}&R@yrmwyuc*@P)^`Vo##m;luzqbrCvO_Vn9u-R232R;
zX}5#lB+WHs*(8e>`7i95aWo|VO>x-z85xi9&xo+c)whQA$4?9hu37oVWIsA3)tuc4
zLzAx7#B|~&Gf|?Lv+stMZ2hAwxkS&W2ssI?Y<`Mi0X9C&@IiJq-j{*TP%lA!!Fy=I
zm_&H7je9{=eFv7<10i{X_XC&I(4O5&2uckQd0T2U0Wus#l7!LWP&LgKa}8RU8Z0pt
zT8aL+)OAh9ke+-~Y#+-H(Ej?zUE`lh?E)7ZOJjd=&JuI;t0OgvnS{KDc+rEgfmkPT
z0`}*F+D)%TJs|~N5PODd*Ta&*+RCYZ^z|MxH;OaTtaDW^*Jbo^RsmijIRYGYgWmf!
z$^n-s%a)1F>-FA61n1!&LRjCu6=0c|vz+4J^|agWTGHu^=lT1P$v7Q+xz#GpiDrpW
z-}IbVt&#6e$mvmpAOHR;{{1wp!-FmXMUe1KM9!VIcDY6>|9q+s`c&k&we^)>Jla;x
zLcyS;A2A7%5V_jG_o;bH%=2A@$M&WkjXl<l5~GKe6OH%n4^AR&QPLrM^3C%G5t{Bf
z9_oA{%<$CHF1IOi8Kv}Ii|Vu1#4ayQhBkG+9=ErImeZd|14^Oq5*h#`IM8LG5b|XU
z^81_ttXJ+?Q*FG+hvz@88~b#qE)n54M`9?B+gDE9{1QnztcGxJMcy{Vo%_>FHar~=
z9^~jJ7&Y2w>7s7tSIM^qD4+BfeRU}XZKE!m=p8GHDtLpJ0d|zvy~jy41pwz9n24z=
zSn`%!k6I97%f16MlAHrmCWBK^VD^k<EaGN2&cl7eEFmXuy-B95gVbpMl$Fd*^hM7g
zAX&odC8RsyK;8?w#=d|R{ig+$%6neZ!rIpZm89Nua!e~VObD5hc3W&YA`9Lbt)ELD
zB-}iuz4Ff;z<FzyqsUFE(*=%9rAEz^6qE1aaZ_tIby(ZP91Cvk?nOIu(jSstLom9N
z_E*v3ID_i$FDD)e+kFr@W4JAKhshrX<e*AVnOgMoP$8$m6hUB-x^nHC8fEvuS_i2J
za6tFUP>?jdj@BXK5y;pWQS7GreM57=uEaDZ#XS0{3O&gNo~tDKxL4}OardH~3D*Z?
z_4ypDZUj3xf7wW3CcNF9TjIDzf2yd`TlB|cuXAtodf$Hu=I2nNr{Y)rNw`U7Avzy2
zuM-e)Ab9@r?<#)%DCS6;(>CEAAr<LF92u!cls-P<UU=y-Lcq3BI^Fp7rg-2z^<!QO
znmoxn;{VO5l>go7mH*$b!|nqo=>~Cj2lI$Im&Z&pK*<jDO^|MR?`fexmAb<H4=kr^
zpe;=ktyth3r&6G?PfOQPhhLVk7%_!k%_HTRLAcD@9h2>9Hn}AR{%%JT4OhA=y&-4R
zbNsf{5)R_|tFvuy3cd`2(8wKPX+&7avcD6PYRh;Js6Kjsht*uV4sxWYD@=aE6l$nt
zsw33*L#d}=eVGV0OK>zp$if<X%#0(~svTs1xIIw!d4c~|wY|Iaug(_rpLMo$iq-Ni
zomZWyNq@pzMNxSPts;!SAX|DYJ039-Uz2dDW$|_=$8wktQJ6Rz53Mk3)hDRecZRH3
zKuEc1FEbe*w~ol8C``liFExPRw55Ow#MO!5W*2=;X|N9944#V-{4YNOUb9xyKvp1-
z&b%&JPoyZxFynwfog5j*<N|e18D20NP%!Ba%)m$ok+H*22AA=s4ptmDNOS^xD?TEy
zwR!blC&A#?nn2AplI#lfvv>wO@l6yn_P>f{fkjgMzbRaI#)&78|E87wvduzgg0(ha
za>R#%r0ji#c=C6H?B&7UW2r)xiKZDk+Rq1`VWi1B6ZI334&r3{kP{nMtBjwwlyhvK
zBsvf$zoPadr_><1JZQXMVIV^|oktS&dL}vHdaw>CY=c25xEnPG$<pP~bjcjFT9QJe
zU`pXHDf5>`W~M7?;7N1#k)L1@%-=E%3C42{&^uglbL)dTCucu;3qSB|G9ZU(1G>2%
zRjgRJ6wYl-F&c}rj*ws3;{+DooKxhd&<f?bDfiGQBipdk(~0X@{qNdHcCD|ntWzG0
zVz9=S=z<b#5Zr%;>qU(JxGdgph=nJi@mKyf2bkK03ai-%hNFN<t;{1|0lBT+VNJoQ
zwnH9GwS#;nX(*Ne!ONZUlXd~0Jx@G;$D_WVC4Tymj=1!HjCVi)QC!_!+-!Fev!M+C
za|Ixi2S6!>+;j)~*Kx#jb_WX5;W5lGt&EvikLjuDI5XP_{ee9mSOM;e`Sltck=o)=
z*G@2(NuRTc;c25_cd=jDZ4NAV9>{H@nfwN!$i?d<pflnGh2Cjd*1eE-#P@H_C*cjj
zE}Wx~sLq8AiMO1HE_4$UaPBW)57TW(rE{@05fUN!N1fb1*toE%IEB01_5#CBpdp$9
zvoGT~BbjcaU7F?IZ~NgJi%VHGhdXO$obEy?utokcQRe*|=P|+cg!;H}X<WMW24zh`
zU$m2)s{7`jpS7wGoNO>PpG(SX86F5Ly*1*WcU$u!u_o(<6<r@bV{f|JPR_pxagdW5
zY7~0LFN+t{N2mMVmnUa%mVa14hQV-GOJ`6bNN3Y996}XH=Y$~JiJN`>%6zEeZwa>^
zvmT1?u0|Wxh`djB4X9FRJawjpb``$!MRF#h#rpuZqvqf@FjY@gy09qz#mtRL8+_5b
zI8&*|OJMrR*_&XYH8Cfv$kHjfp83mVYGqpoQKqqno6c~oD{Iayp&l(&2IoaQ4nm*j
zS6E#I^_n}UKYg0n<D^GJf1%nNvftMM0VWG+ju??_WKB>K_A-tkHF3rS%gqoY@udq*
z_-HFnAK<-gwm<Hg-&xfn1jGp_&(+Ao0Ooi<&tNUAia?RaBR$nqT0KS9xCzY%xYC#y
zZB36fiBfU|p!11bERl<zPwbrJ&SLGEP<@>jpx%tt{BBUvwNvoW-Smh;Rk7=68GP%r
zeUNSq>35-T?y}#8JI;f4GF&%Yzxe@91GWoTsOve8iNkqyJ=~$r&NHKTiCN1`U!vpD
z2{85T9IodfntCb|{-XnuZ>inNg|Ztac(+&VQ~VRsP4w?*Z_m=-fbwhU-dnVf^xI$N
zjUz6R4cUQ*A0EQ#;mbBaOH2*>TU)Nn0inIwKen;@k~>M$t%+j=^}As0&9Lt*rToC%
zZ=w28SG+#ePR9QNN>N?f0lIDc499B6)FG2Rntm?YEhTmB4U=Jj%;N81_Ti6s*6`6;
z$>BsF^@J0H#@_e$S~kXnWPr09ilj$~VHY=U!tp3d9psK=6tHjY;$_1OrrJRl6A0I0
zIE+YE;aKsyW8Q1-JXsna=P>VCbgAo^@So!5H^23(%H`84xXZ0n&J|TKPdb6wW&diR
zp2G-g?8aB=lKIy6r?nWr7}Tu@unQlYiulIA_uZRov_niew%f4oXkg+4IjSJzDU&;|
z4mXrvPdRJLNVV=gCYTmA!}0q9qU|Wl29@$DO(NpM@=d3<;<Ci4;>9R%Yn*1Hk!}Qp
zDOr8SHAd4wtKoLRdhWRdfeT#1k4g=`@-qbKZm%<;i!-ke^Q~k}@WxXocj!&U0(Xsm
zYJypxdo*skL@6{vGk}~f5icv^x$}b_@u*0}soA#%$DyewdP-=XC+o*sf3ua9l*itU
z5^4t4MKB~M6GcBV)9yd@yF?WS_c-F7_Vlt)hQR5O>nbYRq1B>HQI6>i2G4LR1ISka
zBYBa~G|u0n=`&SeaH-!@vZlv7%k5sMZnDb>V!G+D=oFwtk&~Jvj!H)j+84c_3gxeH
zK_u`K!H}R3vOhDz(->R}<<P(@D8I=zFV=UZz}4n^BJF0y@!W4O=U#=x;aGGKR)6s}
zR>mPfkn*9c<i-5qrJCagrQd(nUgRA^l9w4j$Yt?9)9?e!RL`7duJnA@s0$nxch9TR
zr<rY)#fx|ey0Bm3o?`hAR#))=W+~&8-Y>&o5GWMKHHuTOlgnQyxQ;Dc{%~8LR+%Vu
z$0QIm_d;m5(ZnM4*wRDZhCQC8=-})7%c@E%rP%d6YeO&mMz6sk*>d3QbF*T=t40!|
z(lfj%_s+R@`$NaC_?F+#@|_n#P_F_9B6g5G4ys@B^X%QrPd{)30&p*_SAZFr4_Sb4
z5CECZsFf9y4^;y&vLi^_G!|jJ!qv7pTXkuaztHAm_B-saCY4A?c59mFdT?s~P*4=m
zXA_HKnFMLboQGS+>Hey-XHpIOHcBeazU}u{;_aZbMFDa>$(4R2X;S97<csnU`BLL=
zzxCQ@F5IV1s>zLmiMxN5j_}V89Sv!EWeFa$0#QQ`cr@-2;Ye2kbZVjr`Ao3CpV~A;
z#zmMjTw<pr&_R)gQdX-Gt24AK0cc7x<m*07;Q7VAkuu+&u3~4@(K?X~3(Un`1RV>+
zo@RU$ks|2LU_n!*|4)+EC&2A@PZ0$tt|9^a_}i{&WjZri03YEY6b!C2VAdg#%uqAq
zNW`ES$yi}*r<G-ShB~RKrtrh?r8{L+38R{Z*A0b^IcaVaUSs|uy8!4$K3br>Or$wK
zTaSB5&UWOTieXlWtpB5D-8Dq0njBp}!CEEgKOHaVVVvjgH?!fSxE2#;v1}^7AA1LL
z2SoLbfQXxk>pfGu;aOcX`YsV>HIB*&j~awp$-pe2AASf5x=!+rxPLBKg{`>f93&ug
z*MXncfWHo!p=3Z7q%{}RVQ59KXI|%#xAgFnh(Sd($#^<qEtwoNJ=b9J1!|>gwMd^^
z66u4N9gdUhkjH>|r3m|G%CJKzq2m;1(PzHVBY4VJH`b4g_oW>?Q2I=Kv@ebenVvm?
zGV&3Pxs!6&NNBCyli8%6XTr#53f_(@L_a9sBAf(P>^lffkHn?Qg3l>dUBQrxkP9_x
z2QfJ3&+J7V5*)CJ07!ECML`811h!+H!CCs>^r>>&{o0bcpphlV<O0wCgi-^NcJMZ^
zLdnuWB=k{t!SnTm<g#rb4l5|y=YF|u_3Hfau+1$*@h1Vx!C>0JtTc8L{88D*#MO<#
zfrgl$Wi}fGDVYj(^LU`?ZXo<dLh>68Yl<6F5+O4Jg)z$%|DShL?>Ij9{+7&;+kRLh
z4W56sb-}`Fq_WH*o>KQ?NL^_dR%l?&iq4yQ3uJ^t4ULDD=LhTqB6?acA>U$853F_o
zY9-SUv~5=JAgLz+sRjNY#5YzFaJB3c@5Pn+whYJlyT1g`c61aXM5DnpM=s=)=yz0H
z4T609d9-N3$&1R0=T#&xf}v=8pjp?ou<D9)|9}OAsKlR_u2f<SYjFX`4p9fntXjZT
zadkpaCW`1B%WLZrlz*`;`ibONirMFViCPB;00zzxp;I#PRhLO4l-qr!M6Tsxx8V~1
z{vWB9(CE>W9}E3N5#&^v_W8f<(SGDc&R<($$_8Z)k|ba|amCEs$&EvMsNqz>2iIyI
zJ-7dG{Rz^R4#H%F4&8md(wn&-e1s-54gj7T?qhraZcq4(n*i{yu^abqKN!gQmv55?
z#vD=tXZ;r0?^_CLC=GMZs4D&WJs9^kdyP>hyl(=MiU*^vQG(j1q2ab>iPJR$YYLBF
z%!iA<%aMGoGkS)@gP~Vd9AQ?PHax}l^;)ht57tFX532>%nvXbJgaOULuTm)Ld8<_m
z7zR&fWgbi-oo(K_ecKLle=PSs!d)t@eK2wLp&fTiMc!YJLelk#eayBRt==k&9FNrD
zxw&46E6U5fVD{E4@gD|QWeX%a9+GVdMcxxF?_A3Sl=Qr&PyFfELhU0(GQx6TPlGGT
zsJAG7<m6}IB0ztzHL_?v9N=pF>hPT?wnQ`J%6pbf_z3=RK(-+%_w{#hb9<@``lc-F
zMDL|RGE7b*nh6F$ZdnEMQaG0!<hyr(4$JVan$O_1qF19=bBhd2{}G`lli^L?5*VFc
zDNp0Z%KrRw3n$^Ft;@lotnuCX>h%Pnc=*Ih59EETy!(EA$!@AZk%7?Y1MPb-2{+b%
z2!jadL`IGmRur1=O^Ocsq5KE7j|WYr677kTk?!TS#KR-B0lx_<7x4JD%mOa0g2<hp
z#00{8GzIiB6NhHt)ITuX5~_Rs|Nfsk60gCT1oG9SLK2Gc7$kN&stoRdEwEYG0PM0u
z4ioAgBFDCFV?Fe6SHOIgtvztp$b;9HkkuIvqXBh|u*gHqAV9L%&#Ljm#Z5)e@CuEZ
zxFTyaj+gW!|99ml(-nIV#A#bn9^?55rY}PoxZUb8&N8d-f@<4-7*-2c=L)3pc9{EM
z-@OnYT`IS|?^l>bd#5LFBmtcuYM%?XUD54RP!&C;nf~2Iz&fY)qP5twcHCurnd@qL
zHnwkqd&Hm;SBUNx&FkI>_c@w#CmhjuK#0<gJ&d`{cuU8TjBi);U@WHE+Rn0Es4#pz
zc}$|_jwmSTGLAx$;A_6>QF#0oc^#1dZ4_$S_SvNu2nlnonSK9pPwb6z2+EDF%=iYp
zQ!GngERFwd7UtI4iR=4DJ;GBSp?)W8b47aIvvEwAO1^F}bN4g6@F)voKa>P43r>;M
zrv-9ns$F%S*`!;#q!~8wavjce_LTf0*Nx$#DbeL1cln!;Yn{fu1Syoyl%+f|@+6T3
zE5PhW@oU4EdB-4J6E)oFPTQgFb{|3UvI<FlY9Wme943=<9mA})#UxbB=b0`|8>xWM
zwc7{0#kk9bSm0nlz=j6>BnXg2(7it}BgW^_;G1(p1*#X_L<ZG$_@pga$%L`AJtPHU
zMBT@E5Lk29)AaIUO9Dohwvsk~<)uVMAOfDUXn{ckNwqa9y;uq8yD?edpXrMXZI`qY
z4U&?D)Ht~v?<pm^KtzK^%$Wu<NBypdNgWG!n#Nw?vmTisP`2KfHmJjkV5IvMA5R>G
z0%^u+V`Wu|mzqxm=QQ^~BGW)QrUgBsIbsk{u{+<A_PK7`=}Av!N98xYbdJJn1BWL_
z*Kr|f12L-JN0u`LpV=l^RG^!~Egkt%F4V6{Mlo@#8(+@R!ZP}wc;=zLODW!n=!~Ke
zpCijsn>s=V$fj9P(dImWW;Ho$kDvPsJ$xX`y4z8UE1qb;hx4__gjGrAEf(pYA$S#h
z!%C%XUxgMZe{YqDMP#Yi+Elw``(wCJo#3luD(rDf`sylgxW(|_QBGzYbE$NJ?@Ub$
zKVW*nHv{Q8gg#){;!h8N<d~I+*_}O5GU@C9fS74wRX4Fyi(8a)=prAPB7|e*13YkW
zsMsjNwrgSD*rsj2H*NNrgAzD|AoWhNAo(*MAgG)JU0{Z>JEdrJv|YTq>Z*g0%bibA
z8i}U5paBa3UYioht|?l!d$N+$yt<~wrZ(|@8a)c?@f}0B22>H@sWYi1;>*ci0d#mp
z8g=cciR58+o{QX8KanM_bOF|e4AubxVD1eo(WN$PZBMBhIV<alMLoFm(VX~Z>y1_Y
z1V3>y2QON4T8c8DEom=jde`sr`}(gc@w)?T_&^wAkeu3tW2bAz;v2_L_MB6_k?8%!
z5?Edi%p&Og(KDtE-Z2g*NXacBrP20Y%uW;WoxW^2NyVclg?twQsVmm0iK+cM5k2PJ
zcr&TBs`&OBhKd_my#aRwQZnw@Iplw(x?7=iU;W^?PM5IS%z18>by?Oos4DGJhA6Yi
zHnI<Q55+o~QQ#i`^93`RKC3j2uw_Q$93KNf;ey{sU}%&x!+Tl$DP7=^x{wDLPXIKp
zC%$n<_N(^}Y(Js_0rAfwP~E8ltZ+xv?{oMSQ7~<884sPr*Nw@{2(D?p8$$O_uU!2a
zZZ{vZt)l-uyQ54G%ec(#N0z^#U#OqW`2q$DWI@FtNcsx!HJp&9-p93$Aq6$po`+wV
z$uwyl2~*lFNZo`yunNd9vFheUQ9H<@g%tjTM@2V;hZ{ykn1Hhto4nIhv2rvPG3$fs
z*=-EcBa+}R6!7b9E~gPqCODNEjpw$|NaJC`SiIRcUC6NhsdgjSpZY7vAzVq%<u}1a
zKN@!SLckj)p`PIZ)Y#K_E6M0sK~+75{q43|y=HdfH!sm$01W_6Zwhn)R`;j=z@(I#
zQ9Gb$19Ido^PoR41h@<a^qIa+dIPF(o(bvo*FfPi94g>uhvwdy&mRYNezSjIJ6*)L
zfqAP$n)U>?2g(1Lek5<1cgbJyaGUzDl*!bkQ%=nWxv+<4>DJ`fUK6nzsa|{0P|dB|
z+rv$ahSU_ww7%Ep?n0RzLKj#THVd+!KF0GN$6s?bif=_8h-im*PW>ZFjO?__Sdn=3
z7fROYAW~!2Bn9qTQ<Kt`w8b$O{@$_zHUW<sH0}?-&-a_UDP&g0?(=OXlb@P9g@jI|
zhL?<89k=&yQP*fsQ*$Yg>aD#U$PRA`ZX$V8(2ze#@M>I39aF5VaPW*`?hS=HYODT6
z|5E_328$_<g^W}zd}z_ysC1?3wuisR8RwU?VaWARy`ZvOFaUCeK%JfV(U(^65%)u5
zg;}A#d5(qNh3&*|+EzJ&KK}+oE|HWG%9JHr#%Jpr24C6~HM{jt{o8G3X+QHnzv?|B
z?y*pn9?c*?vD@(<!H-JQIlS}_AT&{eA<--hJ7o7Rz})z2dF0xi8@`%u>ZWQiSEprk
z8Icqh{mDb1e8ASFWFRH)%(brA#Xu6;Aquufqbx&BN_EH~LX+)Iwv$=narb?@Codgl
z>c-+O$~WuZyKD1mHu0eewi{RH?927tJ48dg>R5VEbpH{o=m7?r?D13^?1)f1W@{Nz
zLR;1$AidG=#sq|FNWYfSy})<-%$Q*Bz6!#qsHkYLTE86H+Nr!e@{*L#2MaR|nA~JM
zB$*QJ=}3ZGR!9ER5fs3k=__g+g0HV(W^&^M0+4#-{%23#tx!E{@^i=a?`8~7J^YgQ
zVFv-238=D`)pamYz`MoXwc+KyVa;|*zOeG<i;4LI$gv;%41s}JQ@TKdVOc3T@2U69
z9WSqfbpea@eXC%Wg$4#bU2RayRP6R$J2|(yBgzHT(=R7Ka?sDBx;N@g#q2PKkmu9}
zpG_I0w8p11^xICzyH|wim`@b|-}cd%@e`UbT{Y!?s@g;D3@AefzNCYAgCtp-MyWdh
zHI4%2M7e}tYL<<jya$hkOEMPM5(S-<9d_qE>hda#bp>Sr_U&b*Tq5IB)$PqtVH-2o
zPG?RsB0_+&>!8``Sb5B@dergfb~_Ar30X*}HBqAAfK}EQU-$W(UYT;c{I)(+hAZxk
zGmX7v%3;X{D_~*D{r@oco>5J9ZQCG-h|;3cK?n+n8k$N`T2Q2khz*b;Q4uK-5T!~;
zDAJKCpr8a5L<Ed<q$G4iL^`1pdWVD(0x9=wpSP@M-dW$Q`7!I8`C(azA-S%7?R}od
zd6Y$Nte15ISMU8oU`aW{tc8;@(8rlsEF~A<H2k7U_jQGOlk2)a%+N8zzcGltZ6unc
z&H49{s&ooR2>LC#={`DM;d1JmMTGRLI;*9ALH6G?NW-Ve&PFD*5d_nUh=1=<NS#<E
z`fWZwSUEa4EiVM-`9n~F9NG`Mp<ibc-_ytCN#xdj^DyyKJi)Y3L^meV(b<dds)58?
z#q9fBsN=ZVktPKMqHpMj;`7t+ouMp<5i^w~l-q)d8y6&J$F}p0b@{8anVvzd5H25p
z?Og(uT?D8-xfjY`7^Wa4Q{OusZRD)c=Z@yuM|cB0NNcnrE7f{{#K~fpy%anuKQd12
znaWdi$|8n52o=FB`1BFh>6lk+7{k3k@lI2GCpWM4DfTS2tep(;dn9IMC@ReYvxd$Q
z$mB5uC}*AFLVm78ygsH+iR#^nuM)M2ETrnVofzs|%@cc)Wy*Bc#I(KJG+Y9SIPBdz
zphn;IRy636vAfn(#F5nr$R>aL*Hs$UpWI3PN_MG8`&B;Ro@`qEgyerqrE>{HlczW5
zFWD$=mns*Vf|kyN>)|gwcp>H-FyHhR?g0e+TNK1EXYYgWbUowK37_G7*W9%)=3Ko#
zEth@G{M!w{-*bgK^NG(l5+rIoKig}Wcq8*i4P+`T@6VsCG9N-v-wc4W3=<VI#XQ<F
za@)58X|k#mb&yiDf8!!oH1h~pynx$YueFZy&9h@%rFlBdFZCqteS4F8z&`ue5A7oP
z9P|yph=!`pm+RuaHa$4i{IG|gwSJ1rBraC+c5$s#G+=28@{Leo9;CV8B9SH8_h>TP
zxY1U*Kjx8DZxMJ&Q@WX-#tm{ue6$zl$l_k9ltJ_R+F#Wq(kd*I52L>QA=~fY=8p5a
zj_=k5K<~hYX8esas~|%fD*aIqm;6zfyERU%%|4KW-F}mV4c8T>dAO7Pob=1OKPGTG
zmft$Zb!`(7MqtB$nnM_UXkxhe`By^JmY~d^qCWmzS&kmdHX*tM{S2#LN|gCkdtVjE
ze~Jh^G!&5Yb2n+McJGnXhgMbg1+1Vd3@3onq41|T$QbzgI?3MD+~zeBV*K(eA25Xq
zA=P`59K)&3qnRtR_c1A$Ti&}+Q2GPfQpi)+a@`X(v@mk38FE#VTsoYZCZf#yevHd<
z^l!~L(>XX#Z6abGA%vG+WWLQAURJlshIx1+y{=B`yN_l+PiJb;ZiCl)Ml+dyTv0;&
z1laH&N&ZQ8h93u^J3EgF0%~rMn`GvQ&61CZjrXtZ@Fn3WFeYQ`Kf}gKKp6y*GpIFW
z&iOD0K#iPwe}ax|6H00LS(o=yTk4?NPd^RTp<c2;J*BC!&POw41{ITE{%j)Qi4{pW
zVruMp<IyP9K+sEFefqrW&+ODZYo9@&CtVv;e-qegNa8EXi1VN+ZfuNc92!kHxff+D
zpTmY8ry12%aRqr%i^$XzkL=;IoF6Of_8IM~$nStVO~hx4Ge7Ab8>Ce<-VP6WP#*9Y
z{?|5M=)uT`Jz~!PcE{!ac5}X?w!X{01pilzs4MdZkubyc1=KF3fX<i&nKSzLAm!@y
z!3oxN{d)K|6UurC3#Q?K)<dQVvrU)Z{6?p0fYLC5Y|g}h9sC2Jip2j-^Ew57!8SoC
zVlY6ooErXNqaU=#ER^DUzcQM^G9h7pz9kb_4S?;h2-H@F1T=<}mXLuNY^`@SJQVPm
zG3S5?>SqF1?ep-NKPaK?%D<`-J?L#qtLRdIH1jRJcpxU?v(68p>kH)@uwj%jM^`Y#
zEE&JA?GotDuVec0JC7eQVFQE*6Z>f9Q@*QwlM~#2_=#0bpF;VBNEKa_z{H|`K3Hi$
zqyr4bt}(olo5bC|gf`+&b~MRiI_tzTwq18Q8dxX$YQ&}#;dVVrnVq+bUYqTwY#T~0
zaO`-2O+y(CPK&Ms=5#RWPB*7{uh$Xw*a0RK2)YmfJK_q!TsSY$+$dTNKI-Et`1l5U
zx!|L)l&bsXYo4!ut@!^rhc5d;zly0h#tWm)kac#+)M~dW!eP71-9sab#fKdD4bQFQ
z&T?F2oZ6>_p`9eG?D#4aHRiS|Bm$z-<Jc74mk-Of;Ixw!^^yWMz1yaWVgqX}lS&Wn
zFFjqTf4#LWM0=2jx=pdnoeCHW+#cEL`<Cl6ld2VASEL(6oWKHsfjDr0Cs3RVKx+yK
z@$!pt<^ENC;Nf}kXIiAqbT+1UUm>I8<Pzt+7k?<_7Hl_-);(#>5TMvLYs&pv$?VxP
zaBMl7T)ju!T0}%*$(Hd$tocb|s}ZF(r!+p>FIyt&n-JVraGK|PW>Xl*SZu&NSU)z5
z(qVqC5z8MRRV;uy)8TK%&ddMgnli%}vW2t78DL=L6}JY*^q+IQ$qgGY9($3`kzh0h
z+peYptU2U6%?8e~2z>WHSE!43yfpO?&hZdSQL@mJn2~)|$P#4VLn4wu<Va_fjeFrY
zKbd;<9q!@OBuD&@mr@pv1_#ZV8^{F@;Nfqn9457&%$%ax*kX_RMK+lg!1xv09bT=8
zPGemet{@rO{2r}idey2Kf1*-DJcW_ddGW|MlRHw47qa+u|DWjt5yPn>l{EJ0yIt-M
z_XnYoxVeU9magM}2mx7Gyqq7(FSkyU_4>MQuhSXGw;rQq#Bblot?Ksz_I(OUQ_k*g
z^L{giK#-6H0rN{z+;G|uIFs}9wLwPJxnr2BSR}Op%z3~u7ervDv(7n>pD&47JzUo!
z@dZKwj~#4!^z;SXh%zUp-T;4qHlnZjS{D|7&BCl@;x^BIf?f_F<StK0DPu~|JbqKl
zTB!F@OV^ku91y?PioB_-HXG9yHy`_U_9Y`e6Dlegg3HDf;!9-~6s3ZUQpCM(%yZp5
z+0mVFDei2N*j*{}GBg5E^Ppr?!Y<$~@iQ@BaO1BGJG>=z=lWSQlv1}X8KJ+i(fg~x
zrhy}`6ZfK-WU${jP{VpK-_e?W^y?#~Hp{-A>??9ltS-@IXjYAAEoN4Y<Rnh$vQv*5
za$#l0UbCq6#UAkL9HVPO0ct0_TH~KH;;#vfY#G`YJGJ1L?Xdk5_Y&;nn|l3<-io@D
ztI9wp-l*-w8_N#_qx4^Yl{LXnnqD2y)6w|*sDek9zvi!nkDeHOJ?Aen9K=i}kiLxX
zqpXK|t;)1kASdGmtQ(IV?ErwH(L451-HjLT9x;)<x>}feL){}Nd2-;`+S@5DbHSx`
zF<Vm|C5tzzZ>IgZW~H{+qeK`24%C3mr2ywG+VGS>x$(>HHT~F)vo@N*hFO{9_<p-=
z3?I$%uE|$@@~!4vf|18fM5Mh6t$qq8>bGn`F>5Ic6#j&H>f{qWNjyfrpu^T2_!eBm
zQObG?A!b5C!7IaD-1VlPl^Ftl=V89|o%+F=_$e7lXAWw}pMC|mhZT_zFT6g%<V350
zo`o<|?Tpw4ZD<FKno1BRkctr-_ez!bHZXIq*~#jlgFS8v-&?@~#oVF?HDH>A4aKZL
zq1(8*WN@7la<EJvbLK&-tqM9oqT2)j&JsG!DrH~r-c1ABT@^qt1Ig2@hGfhFSTrgB
z*Dr#phBti|y!{ri1}>8cOkeC@uYVdBf`BDo^Vn5Y<^czhKhW?U*Mr`%!+c3*_A>H!
z(X@md-Q#|5RO(s1KmJ2wC^)h72(T>vj%pq8+Aga(jQbSFlGB%~UbgOkQKZVgLLgrw
z9$^h*A&1r#>lmvuLQD?grZruWVP6hN4kHct3pEwU{Ci41{tvI`SaZJ%n9Sc^gVSK}
z`d7A=-CMPSNUOA}*u2A=1vRDdo~^%qrGR1sY4X1kBJyaJ5qdxTF#n-ydV+bd6I?L@
z6`<(J(SOqn9*hwFsvif@^%EQ3SO{iC9s&;2z0<gPTPmoWk_%`6KX`BfvvnNNeIUn~
zxaZwp{PTg!N57_+{zLm81)kS~ByEhVn0AV@6Q2j$!zuEUo%Oi*^iMQQQ@J=aQtflp
zTcLjYDK$xJ@EFRnh9DG)NkYrIfS*;9)QNtl@`@^-65qF<1oZB%K4o8ciaE3S4_h!?
zvJ9kh3YD6RS&x}}%e`)&*Ox6oEktp4&Fp}#@D5rQQ!k+_u^gk*8RlJUZdjX~DAW{9
zUEX4U2GU+*uw>(Z*kTBfC$rUEk4h*74svc=WsYecm8j>4&;PJ-VLWi57~vW<<d=96
z7nW4@BZOxA-AYp@le3%bKYoa<+v2sDWoRB?*6FS$F(I|EvwX1K*2ULz^m<?j>>0>@
z9UzuU8a0ZS{Ym>eZ2I7f?Zr5xf4G>a_kmgj6;8_tAb+BU0@*UsU#{RX#i8=fL*YC3
zt&sfFvKG+cjH&JTR`WP}IZ9^6BUL)zWs8jmBNek2+@kTfV_9kybV1bFAG9L6iJyjH
z+#lNv^RlRK*;nL_8{G1}xN?pSjMf{55t`$t;rL*}S;QL<;n6RDiT1GWsgm+}u!HBI
zvGveoUeCD5518YK4B&oJc<~$H`J7leeqwy-zOAIj6G8S}c0`vlLqO@0{E97^2APbT
zG-rZ`=pQy47CL`hh6({ES^`RyCQlY`uh&((ceQ87<echJYn$EnbLWM%*F-M@<aj$I
zn!G*~LoFr~R&RcGt%?s@T`OEwaNPJ=6Yr4u^pKd^n(Dt$s{hml{QWz72g8^gtE&rO
zFzt()u-3f64wub1cCRxKk{L!7D_8^ZB#(e-shYE1aY>6-J;N+Au&tGbwhA%f#JMzv
zkO%cFw#CMkVMX>yCu%mQAKXUT-fi=L0QK&-T33-xZUqSH8vr(dfCQfPV7PFN*0k0U
zS?o}SiNe5w=X?vlHKYi)&^pQ091mf*lKE4n2>b~B0Ea;CYJW4V(lx%Hf%QKbu-#b4
zQ_Xh|gT(OR+6?M*W}&V$`ghgHJwK~En3cu1W&>$wf1eNTIu=?1T<P~NeIzLSwKe4Y
z9=&0J5T=PMUVI+1=e4dBNcOa00p@W?pC9^Hiy;~JQm*eCSl?;~oNrTPKy=2<0)*{i
z)xG}aQOZdDv1^^Ov21J{$sHJVc<Df%OCbtFncWU+xMw};TB3d1Xy=~E60Dl}3FD3f
z-9>3w(|6as{6!mfx9$b1awNEYwLu*;Vkd9&n^f84zI3)V^nOg+6S4?1;t65vh1FI+
z*B<ck7!zU)uzsT<0}Mw>Z@v27$u#f0c8~mFyKCj$A{*1zuS?1`+1Td@IYKe&EL)7P
zODhPD<oO$13*9!4hL)-qPM-BV<7_YddZ^hQN6&l2ankO^WKsIAavcjopnxZ@G@gGk
zrfTQbkGSD;M%>mRYr2gWaR_F0jfO@T0jV4%oD^}&Tiq|)bEog?ro9U}qIH9zOiP(u
z7ih}7--|nud8Kt+ukO1P?LZLxKK>9WX-}ZV{7$x;*HTbB_VPqwHTpIFcIr7MHP3E(
z3Ax<K!Qi6>lG`7UD_(lN-`~1p^JnNrJc@XPoKO8jGo#$Dr`oJm#FoBSoQu^i*9@C^
z+Igt;`0IBpwMj&|2178v4$i}X<}KJC>Yt>*W0a+xv#{ogj^!0UW&9IOB!Xq4*%&Wq
zyI&-4Ph;ZUArxd<mP6O|={>7_OueQRyj;E!iv*&&*Tg@=<++}Pu4(CEBfDu<&$TxQ
zJMZ6=K!2z2AO?juSHO2R{NQoZ&;MG{^$GBoZX*)lWShJ9(9*P9w+J$NdT;|Aaftym
z0wIHbXQmJkKS)!8Vb7Jc^Ov6L(L*1u$}gw0G2Y-7;;CGHgffgHhSSc;ckSNut%tDP
zQCzPFO&$)s%NwEGXei*`N0yj7xj89$RWCdoueOfS{P|3sln8hr&v3usG{w(FvrC!m
zy&9N;vLv@3TM5y7J5yW?!RMripc>6REJJp&fRI+Db?^L>o}(}Jad>v1&!PtirE6vM
zJE$8OO~daV_Ox)SxD1?mra1Y>6Z^b_&F?*W15<wn=%V_z2zxXSCMS~YPVk45QZO5?
zFAz+W1}mhLohtA%kx2zS7r&|R+P2^P9uti+l@E{KReFCWB~`29R;<p!b%YeS=N2_*
zr#)!5%kj!&pQ1n~W}<vrfQwTL%~md3{=1#kR2W!k-W|8BlGd&8mA<pFX-Y$k4mR{M
zLl0cP=ZyWdRVS&E&x^>9jW6)7_I`fZrt);Vybe?m1t^2^zZ4{ykvzm~c6pM%yMkv^
z>a9hKE0!Su?3yLBP+H^yD8KHBv34YqXSG*NeK$#Mdght=FU@zvg%x<2e8ZtabQj?$
zLwVNWqGq-X^k!8MoCB?e{t3!9xRZZurXqQ@x%~AXhlo2jc#kG7Jjed2aV6Y^lRhIV
zL>l9dp_Hp=-jlFOhEd+wPhR5pFV50>mpEnj&m^5iByX|PpMydql9m%VZ4Nq@OMS|6
zwUu3K)|Mu3N)+5%UFW;ln*#?oWif&MnDGm^NH{48!Q+k+plf{lfpJoNMXm6bc{6{x
z;GvbS?R~2Bj1?$l>g`qGQhvVrT?~)O_s>`(hR_8Ks?W=?*!$m8A{UeNvLLnEm;%4k
zr56hr_q=oPX(~+#h;zj%J2ty%QyL5~kt89S%fvG?95f<?7q4`id^l$$aX^eOBv*Ql
z!0VxHmcPW?vV5nuQgNqdB;pC{G2t$Oge8z~LAhH=I?8p>qiO=#OX)8n71gvt=@Bn^
z&r~OGByh5UiV8$}o#>2Q6-QzZ_~BdThS*_9G;Utw`*F{7n$xt?$_Ryr8rb)#jDm}B
zFMX3MYkSX_EA0Z;{yeT-;ud&g;C38b#dl5oy&aeU`Yq^e72!Ob^kt&lH;8q9K@$_q
z5C`5UM{P#^i8h(_v!4$hcesC@<PR2$yG-emg6#a$KV`lEj1XM(e%c3L3XFOWXhXwB
z(};i(9*2y2Q&Q9>BTqO$VCxn)w76mYhQvNf+!6raK1+L`0iuJQ8lnPR6ytL9^r_YP
z{t5Y?gr)aEQ0%*1T+iwEK{@*GNw+D?<LE)ehC5mfD3!(FTLOrVRW>vzQtO4`GyW70
zrY9?eU|AD56<H%fs`apY8&dcxyCinU@v<I+f_P0J8^rHQp=|$Qv%_TJdNIRXjpzIb
zq|TVqpYK{_)4l#21h^uP1E43(!>2WxUm>RvO>R!jRkKaWV@m?1(McB{92IC#&JtRH
z7p<>U<e~c!f*;7UwOzelKW}sfyr_Hsr6xs`b>e(9IZwtp`;qsVrKfjd4+p}QCTQ~J
ztCw=n9Rx8DP0fp!?3YpB5itDW{EqyJwa^Bx=0}tqc=R4Jj?nCd)>LJt+FIqdyOR0F
zvp7Y+MF5kW)(Bm3Rx1e5a4S`mYBvgZ<M$KRx*Ni8B|23rt~~5JsXBbs)nQ6Le4ry4
z^Je(9{xPTh?#ru)bSU}eI_*?=1jV*7iJ?LBm8}p*nu147(IU;YJvRP~uJI%3(CXnJ
zxEgF5&N`U>X_ckz4hRjG0I9kVeb;I+mvIDoTX7oD4^sD&tL)CRZ~vv+dEw@7Q<p)N
zn@PbkQP?@;9_)X7#hD|Ond;B{sJ#!qbQ&}pi|VZ`Z8xCTO9%Z3q9bhu_m@adg3`H-
zn8-b|BJlGhFne=^pEvez|1GaMK_i$P--}rh12suHvl7>Yp($uE-k~!<7FH({Xipkh
zl0Z}%3O)_)3t{UlqF1C|c~@2!RAArl5m-U6-S@y{J60UQ3^V{hRy6Ak{t%|^I2pqX
zSp~fayPjm(f3d+k2@Jtx*7eqA=$a8@82|Y1n^^{Ks)XNGy>RmXDqj{DCv!DocjfV-
ze}5v56>YPS|DFc+AK9HoDD|&XZnx<@q5qbE{3UZZ6J|d9Ay~Bp>P<vO${qmNI6s2^
z>MihLp#oNx0oom$ZBcaT5oQ|In{4OTx+jkPQ=3$iM!f~D28Qg!^vm7A*<|q&1YJ0q
zq>lBK_OR0}s+&8n8^U&zXdkjWvKYpc>sW0w-qgLE-;Sd9NmmD%u)Sbh6VP13pewy!
z7&^3C`6fyARw*H(1&1p4)q2e!3IEj}@lvJilbG-3O&RcTXc;ubTj%=XldZQ(O&Yh)
z))}06YCav!ihS@^>3H1TLG5R!93OOroa4j5Jb!i@e|wpq_wzyD5yTg0u?9n^s<%N`
zsGLk}{;90g({fe_U*sT?p&G5pbuGAV8XTfdxL6DbM-nJN9ig9Hve71TS7b_KzBeC)
zpWm0A9Dn^xZVcEm3;$u`sDbCGmJ2M47j7krgdAK6x;04Ba^LlDMw_Giv1LxtaMw;L
zzJPlc#YGjjh2k=J1SD(x<)!8`KL=?W5-h=sM6vc&6-Bh7l|fH=1s|t%a0J>cu_m*P
zYoFkv8Cg)~qe@q~u9PIvZxi0$4;Ssryc?Vx7vz+T=!UOEkcCI6&uL;=wCfj|eCeD6
zA4+E=axFGh0)d~@ERg4|hd-{R2~DoL%8r*;l|F4Nee|{XVrQ=!tqsxRv3isISUDcb
zQDZ)7u4unE)j!;vK~viI{?#6FGkyxkyClU&5&rbiHFd)larY0|^jPGbSrf7P6;C~A
z5ty7KB<{LG*~=z)GjiN4TP;!-M&{=ZSG8~iYH%Pvcr!@oU+^+WB*xJ^ilb>**2LfK
zX~;OM!SGDQ;dw0+Nfm$G0D&7*Z-YO8+`F5(wrtg_G##tjW%fu#;#im#-W^QzJ)!|w
zOLhGvnr%Sa>TAf8(uHf^zUnD190*882>AhKpCY-U>8PwtZiM%H&OOyMh=e?GR&8T_
z6|_5?CrX)FjEg{7rV;d@s+62vAVpnTgDcs7Jb2t^X7ZBQGO*7CD>;p9MEy#0qhC@b
z)6O@kELoK=S0N4f@Mo9&n$h<$^^a@Lxan#-<E01nYI;AFkHuAMK$j7=nEI1w4l{B!
z=^9JEuVALMzvkwt_a7@S7aJXOO!3p<;7IQQo8@QppZRH#I$z?}vQNf6a-&+`*`^;Z
ziRJmqq;t+3<ZlEwJQt>37r&ozon}56f8-xF2kqkVyG^~C#rNAvJyw^~877O(Xiy8)
zYhg$qEus6oyl8>b6Blcg=R}I}VZOB`XJzxm=-F8NQygmun)f^agW1~t6AZqy4kx8!
zOV+TSf+Cu(BOS`4CXNP*TR(BvYM4KC4-L`0x1UdsoiR=^4NTX1>C!2=&#4epIb;8@
zl^e9KYEkTA%mb!{T8)}l+}u7-4xg60tS{ot75eTk-Wd&G4)dTMUZq%D-K68Jlr64$
zLI_J7tTi~z2nBmGQ00{FxG@;Id6?_LDxx#D0sd-2xPW1j1BAkq$Ax^evkkTlU*;b2
zsE9Q>1;1knG5JuZyTPH!&(I*V-cA)-lo5{D4$Jxuhn_m%;@N)RQ%}-#v>7^{BEp>8
zS2Hr@lSa;;(ny}1kTCeL?4fQ6(^@S)Mt>pB0F6=z)|z1ebdSRu7f}Xz5bk9Y>MJ!%
zO9M7GNk4#OrqenbS=vOD)|4dNy1Z|#WA^p3>XY80nZurK3jn`zm@;A<<(Y)W*aXz_
zuiqwbPeYPU$(onM)A&b;ZCF7933y@Doc8bzoP3YGPEmy=3%x$G{4@}M{{V6gF$Acn
zCgkdAk(W=*BfYB;2mO*nEj$X#E?SCe|9(DO4Si!(gal<xMg0W7%ci{@!+^Y!FYd8X
zxP34?F`_tp1}MOA6pl$k6v!h@a^q-D*ZC7saQPfHO^=I1)+uzwl@MHWFOUQkb&)#W
z`#5QwZuZ0UO?K4x`q%TJh|8FIeT?%uKdWa#@&I8nz~7<gNkLM*{CjonX6vXYFP=C1
z$v1j%jxnxK<{~EBPsvtR_JLT=Lr0S)&a<bpvC;s@9b97pJZkpZ+veIZpKa}`GjlTK
z9y3k81+S6fwdP)7#_z*8>R1wBDutOG56wz=p#H*6*h%tE!PN-L6oH%I@r5OeQUO*z
zaLBT*Oy5k8yF9Y+{Lb&-)u`jw59si*q5m+<$R70|NlYVVazez$*<QNu-oa>zpyI0l
zusniLL;-{+<RU|vB7*gvthKt+wT@kD*nru_*C`pGjZw<9S{$&6j{pR*U(~6oWwWX{
z%o&q6Q5JWPMr1~a;>OH_3FI4{p@1-2ql^snoC;9pSwh>y<SRaWR57A-fE1Q7XSb!w
zzS~N1U>)M3AnMAWkJEfMT3c?$zFAe>_I?+yU(KOLQI2NF75GWmS9N=@>bfagl)07I
zjPhzi^%32OSoo9pP<x&pcjYyf^nx#vdrh4xQq*}qfRnEc2SD%<(`lbXx<0xKe*!q2
zxKh*a4Mv~3WiY%VqgNfdeW8QR^RI)3j4SBol>Ho<(%t$}ONQ-F+MKLE)9S)Ca)6Mt
zsy@AH6Iig@`SvJ2@EJR9t^#mioO~(Ir=7x38ZJo1&S8(kZL@h=7U+lOSyO}@Q&HV^
z2M_+Uvn_eMY^V0UIQ8*AjM@J;{RwBsfU11s1s-g=&w2rG@e?ks)O?rkE{52I0l@B;
z|J#4gf-F`8K4P*t9{^%sV}4Z)&)`Az-=<2QVEP*|@(Z)}MW4#rhxY(0Bht3EbtQmu
zJmHa1?)i$|6Ziz&X+YtCJg3N0y=lRd8G0Sh^IUG}i{4Dmgn(LkuAE>-AVq3Kumo^X
z1KAX222vVI?GMIv@LlfMqAzDy>oSg(nr+{@#Ijf=6iuw^<gwb}!frK2?&8zNmHYUv
zw_#>$>_~x^fN$<XPmMBlcp&W(GhN%|=^q(B=k7=E0!fCj_EH4x46!cI_L7^)w$82J
zS-Y#BOZM4KzvCMv)<G+948iM|;m1vW=Nr^}UB<aVzi0EH967LGXtXv+CwN|!ojew}
z2)YgzP)2>|hp%wJS>Y+<M~S7^e=b?PCgmNArg#4TJs$Pj_AWTY_Jf&OQWQm=ReuAs
zu*CGu;&+;I3UvcWI+s0pd*+a*5cA#luNQvu-w1;TH)B)@g4$gEWCB12QSumUx${nv
zjcMYT8Rmu%4OFO{!8BQv2qpeHr7R4WlCw-+&9n<#{(iS&E=!zk22Of2QR;=@Kp*$}
z%x5`!zsp@UAZFg`0z~_ab`foc_zNq@c%a+_iDb!9juors-tLpHy48BY>RDYNn_7GJ
zE2*nRw+#7v<8)40xQJO?dJOUg0p5-VC$2P_8<;@2P>Oe2tf=_M8Iv>(+m5(2+t;|6
zY~rc~>Mi)7?KCAUXvrSy2+tYEYDPp-zTPvOp0kv3<|5nd{rT_5PZu-@A)vf%%aLbB
zxWSW&ldI@SO#S~!-5CT2g4by1#&vD5#2=gmWd9_BY5wjWGZEuqewg*MioYnwnI+b{
zJy-^dFVz~qPI;|$ndBzbzXf6FPP8Jnu5ZB^yr9#J0&&$>pCl=O@XaHFk4lcQ!6<k-
zi|00le+DXFuL&^did)dRvKg#6u+Tz?u$2AK^ZKTPjEkSg3?Kf8uA9x#?{k-t0V6UU
zW+mhKXNEo)oI2#YC?yRsKdnqo({n!ks#t<GDqKR2Jk@+F0)6x4{jiafkHYU~)^R4f
zWn1wp5V{tE)rEkn+B&$*`vqK0@ldnA`?QFkO)A&tLKb8k`;uy<3AnC&wqrk%CKZm4
zb`aM%82P|zlskjcc!P$eU$J*KeJ;<C+1sa}J`npXPyJvbs0Cp=mIGsoiXTGGzKq(Z
z+{MeyLcvKNrMNQcFlAQJD|42JvU3xs-KQ?wMzfI#5iA}&?bWoIM5E^8Zp>G^!cw_C
zcMlt8B~1w2x(b>;dE(mlDxy~l#T|etmExa=jvu}M=nUyT2O<gzL4izx>6Sg^sI6Jc
z+MYGbIXQ<Q>j0~DC<uQ%<y*fV#o{R@)KJWFBh>bX8FYLrJX6=8r!w-)UyKW=`&dx2
zMue`^;&~8FDco)tEOR05&k<2Jh${<}`L3&Xk=Rm1JT92`WZ(N@&A@(W58*Z(X-+#6
zAEGTX8EABg=KU0<KZT5pMf!Z@D>ySM%zvm2NS@<Zdu~^$0o8BCt>>Q0R3A%f0_6e$
zT}K02vTU-?bnslo25upakTEg1X~Ak`enJW9`mEvO;Pk~H+#+|7_(Jh%s~f)>IqyCa
zXT2bhvM~EcNb#$RQ};9lJyw{Y2`wchLyvDp`P|$(N%-)Bu>=-w3qwHn6wM;rkM&E)
zdh1ZMy=Hv6@ywZg?0xX0q!WsW&5vmsKqaQtbjwaub-_c>K9;1dJp*9<eoXx(-SbA2
zd};#G+g2i3H;TJiD0=?5Rr|e(x3~=UGtc<>VCS_QqZ|N%cNFumGc<QPVinYo{Otnj
zneN*3Jhopgkvt*j96j8^uW_M@kP8W%#(k$ave+IHR6z3-3+LpE3!OqlS$T5VQdxgW
zm1;h`D*mloU4-}|$R0dRhpw8l(~{aH@V%c$8{_YFtFX!#ZM5Ypx)PMsvK*lys@9L=
zBe4$1Bhjpn@!C=bXkidA26wWg!6{SIP3hkGxCq8Iv%w6}n_kW7SD@j1Y)=8LeS(j6
zz6D+y_x_{b{MC{HuNb<ublmVR$MLt(FQuhGVt_JB6iU)y3b+9ni5;*c@gfwZ52T4B
zi&b+eS~g2IsKw_V=eCV*Cm^Ztzme4;<$fH^nL^)-XyFx7P>SlQB2(F$yQ!P^+2=fD
z?`LKQba8bd$aUbwjN~WrQSp0Jp<fICuobVb6yjWu4lc69$T=OOW<`^CZrq6*C8n}^
zv1E&QaNd!x=j5VvsxxdT_kJ5F&K-$AaV*wd;ZPe$-O9%ax>nePbu9KNB?Uy;I6ZnB
zZXG30$OA{5lL0ddeZaXYr~K1r8qbvw>R3~v887U|U9S_hlW&jPmF&Mc`i`3s0^YbW
z%>iIOhjnE}Y0-2<U&Ko<05$a4X=z(MQ-OUK_#3eY-VKIViKUD>N-+=L8i{gK3&_wA
zKHqrU(4=VNT{PP=%7k*h0S}4Am*55$mkyX`s7#ZEW`=WvSZplfkqL7u1Va0>H2P+?
zHd?y(@|Z3~61lrWxle&aY_IAY_phy|Qvdjd++CAHt=%$wYW`v;e$a{9P|bXemNg6X
z=S`0YFSEF-CVEL>AuG9-%Lh<Gw&X#Le}piJjT8>R@*>4QM+F@J)vOvl@UJ1`G&CMc
zzKV^W;M&SBL|hH<5L+x?=r#6gzE3gyB=<&L6JmaqsZWk)2*OCdv8rEiA1Pt@cf}_G
zT6EJf+<<z<w$R`FFIAgUk9KoXdEc@)B0^b~qFaqYDwZA@$x>hbaW&P6E7s{Vs3#hi
z=s<-J%&ZR3O_(3lAcquxM}c|U6XLDQ5;HC`Jf(@tp@@BU*LQBWo!{6v>ku4ghJKBy
z7sK;UjG^?1x+glFWGyW%J0Ea1MPL5;2&&#VA>rT5hQ?cM{yjX|K{fG{mJf~;A3z%t
zCVaGE`GDshzTpoQ0=e)Ld;7Y$-k~_af&K3xqPRJ85|ojq95F$K76FiG5z*@is#S0*
z+Tt}`bv0>Z*1@=`yI%3UxFqV=*7hWa%|i#*>eOMApo4Qxre!pVKTdq2SJ|iZu`ISi
z@6Y5V-66GYWrh$}-Hxsxv#{p&mkx<+o9V(8g9zSzUythGZFW5xUyS0#D8gy^bsy$?
z*FILiGO1ID6NLbt`zitO@WGy)N3$l+8}usG+<S<+Hw8IuL)BWvA?KzOVq#bK()Tt&
zOToubCbLFuX&#rpcR%28Hob6}_k}0o0=g4RzS$WDMutZo_<!?@>uuQEyOJM>%rpL;
z^%|1VF3V_Nw-Z@40axrg89-*d6N_k$8sPnD(R3|!bA&W<DZ!9QdYAB24_sPX*6?r_
z5qQo>UK7<w`(ny(L|1CaXSG}ayKo-uL2@v&46==rrwCX3PQ41aIiPlkmq^?>YS($3
zCGVrqx~aAREjjRk33BXd2i7lac;C%{18(fq<sAbw;1>K(r6$2Vt+RJEo5i>~@kVI(
z19*pEgp?>UiDkb0hS|CSB>77K?*NY?_`hwLG$U4(9cG~s5hMiG6}V<?U;*fm<C$KS
z;68=}r&Zj)peyW#v+o2|xD4c`J6kmaou1{yuBz5LoyP{O;w(tR3nr*<LckhsW2eM3
zW4Qo0WrT?B|C^?;|29}H9Me#`D1;Rm7X;V;FY+C&k;R=yzHS~(D2d&x&F0hjhb?SY
z^QFtt*)ijTMg%T3+O=O6kuv+AUcxPfQ~yX_{)4oQ9B@8)IbI?*97J>r(cs|IaWiAR
z;X~6?7HbtB7L!d$W;cIRiARoml{v(=tPUn8ogWSYLs!xpLRoWe9VALaN`j)+721H8
zhDq36y*^@M=<{8&;wU?A)D|`pcy$1lZ=EzXZo!|_GGzI$NXV7MpCn=*ykvqQs6SP)
zLjT;HZR5&e-*sFx#*E;R3jahZW5eieS<Q+>U`XW^JwpU4v2po=(j64RvC791Scl95
zLt8kYupH60&h2oVL_h{<HzOM+j9{k9L?rJGu2Bs6wsNGsIC*EKbIMKP)_|Lvn%mW}
zPbDRfe<z>i3gSc`901L)cHk5wGJW<>(UgwOyk(yf;&+m5guMKq7c}ht4Xy>?Dg$Gg
z$~Qsi`@p&UDGdDy45f}hC(Y{-qO`99F@+jt=>eY0(zosamnpCZ@4rX8H}y)S8LdJI
z|B!8G?T5M1GhJR<op~-_>ySR`?Xg&2TrF^s#2u=LB|P?sv^)aSX;{<{xE$60{zd|D
z03lL*qst%?ei^rk-X`O7EY{?joWE#M``HV7yu=>tcj$@-7YHg}rg`VLmHxxVyXbl7
zTXx0$;@`gPMg05_={dY*s(gr5oWbc6XDvZ)U))E0HCRK>S;Q)j6$0hNuD0p1Vn3(5
zk(X2xzRgv*)$IoPE&{OiDGfpAV~EjiyKygO^7O7NKcrWMZNcPPncLA*1kz`STNbN{
zrBPZQr{z3j>*o-~#v%1WAb3k?WC>a!U#|#noVN0iTD;XfyNh)SpA<(qTo}s4_wPD^
z#dHX~fnX!LX)WO;1IjyK&sUPrCf79ms3N|Vw`TGZTQnQ(836$&MS%(`3^ToQ3|nqH
zgQ>G8NH1^4vknHou+^u1a_x9uA?*HXshEQaj*dNsdr>6^%fDhnjnRwl9#}_G+FNcK
zt1N%fH6p?G>=XAV7E8qflTi_dAV*!A?mlKd${Q*Bq&!U|B*0qK;rHd3&iQn<Im{zT
zQ@S!Wifp$#!b*?A=ItY7+pui3wQ%=n2>agcfJ55(?P5JsC$}8$8*Si*a6uDf%g*pj
zIdW#Q<MdnQ(s;p~s+cz6K?$d)7mshiQ`iFOh>qM<w*rhCs|R}6iLcVAz^;M+`Oc#U
zh);+zBZg38LpROnMi*MB*&9<ZW;n3^WvuC;>C?Jbl!WV3unp7O`0D=kZ)y@lQUMux
zHr9{P-)3f`cG|M|aU?7w1;>pUysSJAh8Pi;8ALC9c|e&7gr-_#bJkU27<}cp$R;q3
zY$R+bFJiZsMOZ<n*!RMJ<Swez5Yq;z-MQl~k9eXtI)xYyXo}OwYQzCrXp=a9k$UdA
z5&fJoG}%n%sUWYzNrD_;!A)>$o02o8XXi>dS=lTJ3S{rjd9infB}ES@!S9})MI3Xs
zSLjDwoZ^Gq6xF8d7&drUR=j<a6qhT#sRyk>2oOlA1P+#riEQfL6_!w9oDuc7@oVUZ
z+s&v9&}AH~z*#FcPXse;UA*_-FCJRrW_e`>9j@#M<fN;D9rzAjg0|cs?j7;!TQ+$j
zYW0KhHyaD6)9c_@lWj|$Nm`F<<C}pcr+F}N!FW`g*5Xs~D~j;zyJM>UilBxKFlC|@
zkt3QB`x(}rlbs@qw_P;f*+tH_NCv3Le3`M75B1K#Li|dM`wHcgh69k5Io2m07O0T6
zasEv3z_{{2hSR1G5xoLVmUgW?RHvGsTTQ=nLGGQ;MWN}5gW5b_8ODF<6p)31b09gz
zDlwQBR=L8<bL<6svL3X@sUcnndkm#Po~yZT#(e!;U^&rNGdIfXJ^xrPq0i0jgngkx
z)*X8R@>s=7hT`Xi)chEvmUl4Mbv9hx#K_2`M-=`5t4o}kq5kQs=_kZb_Z20__tdKe
z94hvhm*uB8sPRqk!}x<Jr+$z-U9gl1rI<0NRYFaEAXI<qwbjW;auxbjGy!)_NWD4K
zM3$;vffso<^Knrru?bR@qHQ%5--wH{nP`MdF0d4O+-J;gUwdaJ7~6!s3_p^HC`WMo
zg^3QK<@}y?m#to^QU3Pz%-F~{+xLiVBl$!<=!t(h*q)=c>W9pMJB5}j8jy}XnvPnU
znq;zF;6Y%=b;$<^<kn`5aY>4LhI&eY^^Z8DOT_Q&V^-@CaiCqqa-bkN9RbQQC;GL^
zenyy%XccX-siDuH!JaIYOnpTgDxl%Vf8HJZ)dIWwv++%^oHb$iVivyx2@><IhKGW!
zA0EqTO6^4EfjFFNmmX~%nAOdS>t`y^Fn7cU!g5`?Rg_kz-$8a>gF(R`tCbjX>up*g
zIP1?F_ZB?DN9!0$WOOQ;92qHyz84QHU7&Y|nscpfhxMu-gs8m|>DH?QyUfNsi{~DA
zKe_;e_!hCNgT7_8H1R)VtA7ji|B*Hd&g+m8#*;q07=t6`a`!j;0eX`N)UnbFlqN=J
zMk1#4aF*a(P?)x_G%Z#o1dur9p)^I2kl7lV>16!Trb6kmKH(m{toL7ktya<>2f-x|
zDVS)dGnJVj_dY{cj8;!(O@PDZuzXdjL=>jWYs85=Hi)C@T4=)a9)c=t+njX>M@Rd?
z>k*Jin&<S$#7fkv<H^<=$@>Jv*!KI#$36^#Uq(aG9T;A`aO**GXUNtorMq=ZhWanr
zhg!*_#CZ>f@U)fny6!QTtk{vj&zC<jv#^thM+`7t3<*aG&>TLrqsJA?iL)F#KEqmt
zjU^62PyIkvN+I2v)vMXMYDBTwX%->JNC@rYfeAjeyg&@{XPr2k&PH*hH!dT(u`BxV
zw8sz*w~Nq6aNyTAuK;pIc>vdETfYF^%3`aMm&;1QoVLSUWn09!L3h;=k{BYE^&7U^
zlFW1nG{YFd%i|dt%5j9{P5}J~F{VveMflG_a5D@k>L0cZbs+JrtOp4zcE}Aa)<M9|
zknxCzFUP{$mbY7iR2Tr>Ie{(#n$D!Z*3NI?k=PBuDr7bI!uQnj_nuH%8Bu$k%oirW
z`aB(&q#Osvynr3-K|cf}!V3dgkUR`2g|OkRX+*~k5d4<O&sm3TLCtnM8}oyS+==^t
zJ&aR^bqEGtKNw7biLWMD(XMRF;r|Mp`Z}D(2>nJ_PR7Vsy8fHY63j3R=$A2Vu9OHS
zhOmV*WCkg+`hlia4ZfI<ZA1RSIGgV(+JhDzx5G+V1m&kWbXN<)g{hTig#z_;Ef{-4
z#WD4kH6rjRg5y1V*06wfm0h=4k)bWpI4f~Wh-QH3Gmo!1?le7dU|F1V>$F-evot{{
z;pw4U><oU$B;pU(>LIeSJu_2R-W7kCnB^Vda;V~mW5Z~3rYz#--x(HYXZG9Deta&(
z#-KR-z4E@cpRm7MXRa)s@Be1+C~R9DXxqY}0>It_(~^#fMxDi+e4lS!e(O#vEHOp5
zRcH@K1`yL((n#Tf0eED(9P52vOR!v;X+)>BD5N&T@LT{e%F{thf^EtMujDdu815|F
zt@@*R3?J%VRJu_2Au+}G)Xx6<#0qpb5G}R`nl}2raB6sUZf;<YBzVv2M*O4HX{f(d
zg82YJNiEQ3-M(LG8@4;XL|wDYbjiNx<Re=IkIFizIDOvEKy;>>6B>nEl%SfBc|mc8
zaYAs=+-H3{o$hNucHR(J9f(%|xj+n=-e6{-&kiC*sGJP3tlt$PJ-@PzwD=QGuMJ60
z_YhvJPMSK)gsY7lb3J0!#U2Da1gu!xX_I0#r|GuH53k=SD=MxzB=Ac-OET)SYD?Ad
za0uZ1ZS*8jvkLq4IihHSY~?ZZb-k6fuNQxOL0HlA%YtLt4dWX%$(KhiCvfff!>=-2
z0oYF<Vd7LfBPR|qlyS~G`{r7v8l+{R;}+bFWNG4aL>RUR3g~Kf%(q+HWNf5jh8_8?
zLRSl9{Yjhb`ylWAnR9p|@ox>zwYqiaWRKUX&$`vn>+;HyL&7UwN)n#~zi{+o+&V{+
zI;lpL<ZT?)iCRm|rO#DZvcHs0FmBrxn$ZPFbx0@b{6OXfZx0^i3+*?%pA)B?!L>U^
zoSguXaE5;I)OWPE{!<f&IV&uG`@BFsvC%TFTDj$~z1eXe>_iW~f>P#vmY=P0Q929R
z{-vpnJIPYA@wY=<JwV2~cY@Hz7lGA^<cMF|mB)eSZ<U|Z`>nf?LIN1nIK!FTM#eOo
z3oP=6@s(|bx=5b$#n`Z3<E8wRIww1Wy_9(d*%V4Mj8{)gL(lM9P46&-da2Jm*cf+1
zr%tB0wqB)(js>3ib`+KvX0bJErOLh;v7(<tyKTqJ(UyBKeNUHohrHe1CP=M_7T0+A
zQt1+~W*4GA03Q!RmS9DdgRLs4fsm=FakH{VO~Q8Bn%7Fbr=Z)o*5Y*tWDd4Xr-Dup
z;8I4G<X3U8D8_b+=o94z5vx?nZlYZf+xSdkSv<&E0Ug`Fzu*s3#E>u55<igJbD!K9
zhPYMn{8;&j;5n5G(sPv_@e^bVk8mk2viBAmQiDZMRLR9d1)h+T>e7>MgD=##3ekQ{
z!6lLQCn@Ur-^N@;crbZ(ZN6HqFy87Shps&^pbP={>7Xx3?fs01B^I(JTNxJGD*N=C
zKL1OrTH8jxu!0*c!l#bwMb_^7BFeFWYxbbYVZs?0iq`ZU5r-ztY<q#Ov)(?kiP_P<
z`yU=PnFjnegzi~CQBYGkO+g{q#g{z&ole-c38XltxePwjq3DU*pP+?J;tn-9_HS3i
z_H$fJeZa_X+tv{w$*(%<tY&xl1+?r8I6T75dZcp1WhEkl5JyX~f;5hn;m+`XrM;|5
zs?V;@30(V!?d&bX$>g7Ei_jwCsu>tEVT21A#))OD<Y$D(eTvpmV**FgZjJ>V0~<$f
z6HsM$HsAy#VsH>w&Cj7IFjhNqBEu7afFp?rNz@Z?dyE;kX~^8sCA8#^GD=SN{mSy*
z?w3oSGWG9E>A<!CEKa`Z5j?Kl^Po1S*>mxvL0jmZ5sthy6&nBsyuy<2D2D>P26fSI
zSob6A)WCZ&H@AX5Z3WtNK=>pEgmK7LRwVUI-94#)tf<JyaH)X*AY*w2N0SAtKAv@M
zR{x9Z5u8nl*2@$V&L-DEt09T1nLq<nFwXVWH)ML;$E73dO0w_1ENGm4f-c)}0z-Au
z_CIXVx>{z_R>g^4WBWK}$25;f7*8gA&3qYTG_MCO^ElD+!{?yR-Kih(M;SNBX`x;C
zvMC8Z9lf^A$H&kXjjRDI_c~q$b&z5hn%${WUOIr2dzrc+=y3s(mWcS18~K@78u)PM
zTEF{x>c+t1aOe4y4KWk0JSr%bl)<$kb5Vbk<MC!%dX$&sAJO*=4~Civ#C`IVM9u@y
zpKRC}zBkqReDa>h?QYy7*~o9QRuzaWo&*>PGYF@4R02p*+Rh}eQR%~^j>Q!#@&^Sn
zY5w=CfO1d7H$0dsgoH5oDfVfSTlVM5zFV2G%;*Z|AP;k@y0~T|TGwfEU_FLZnGF;@
zbYnQ=*yfumRN1IFI~+^BKezy!zRE0Tf;8AKP#)BI5}$_wx$JkQJh2RB;`Z=vf`m7J
zj@aA!6$xnvKnjs9lHCRb^S>vUBQ47RGXptLBU*7(F)rtU+@>BdX*|q!ROe|wUHPvp
z%}_bCqS!CK32{WJAUkA0^i^14Q>5PO3%lh~B%%6dbm{+>icpfRejiXkJx&g0E#_}U
zWI|2^!QWz|?T-*Y(_*IN%^<fNkwXXJkCKUt_a;^_lWy*v)M)aZl7-=yH)9qy)?K7I
z=8h%D!S*CPATI9B8Rq4Wbxfi#4e$9wKi5n9ttd`=cBe`oWS-i`{7)(!CmH|)^-okY
zR;}v6=5_1WuPI6@YXulKWVf#Me&a;Hu?XnUCEnmxa+|8;e)jPSU;LjScp8SLuMPH4
zJw;P6xSs)Hc;-5|->$=>)}UzQ4WAoFTE8B3kM{ee!d&*|O5cOCj?}_f!2^VySxf`Y
zcBC13I8JonRI9<Ipi}#<f|<N_1vQ45iBhItMQPTOE27*|D)Nw~&7!Hp_P-1Jc!Lu(
zgAow~5M=_PXi<Ip<ux?zs#u{lk!tsJ&rMxgx^c9fM%+4W=4dULf9n2=kn?)bcFC6H
zz{kx_#g)MkH)XG7qaVfup4{7H8l*G<1CT!_J!W8gcPU&ei{#E`tD~LyG?+0V^>&)e
z{We!^I9(HUo}3SJ(dF(@;C>f>h8Gm-JQydBLD*E<>Y*>z?T(*JKhgYWc|7wGO7%x~
z?#JqDOd={LN^mRdz(g50xtbfUebzZIjL+Av@!V+p*xRbaq^A%)WbcO%E$7rzpYBfI
z46_hryYrgODxKAZ*@uGD4Nz)*sN++j79ZZ5+KxIXdZ-_?=IT{jyaTLF^64h%X2KDS
z6ZEK=Ti(KsczBF#tirab@$aZJ9QDVa-C-i@$A3*-RM)Bb3S<wDq>0;Vfw?2Dsl_&9
z^^zSmp~%(~AqBcAa~jy<+y)rEvuqTl+b#Fw&-&TI-A=IAbQ(3ngYg$=wNtQrA`L9b
z+CF6^{=>TlGRJnFslYY&1gI^{Xc99S&x=y4?sLZvO)I*uEx4MAG5cJ<#3p`eoAo_>
z*Bdtm;h||<AzNlDcquB4TrEAetNv5aYHs~s$)yuvSg_*-fMM=5uogYTWPv3Ej^MU5
zKM~eAW-EY5;sc{*Og$)sXWDCcfV(DqPiKK+DfPy$OJZCheCX?#dLuaLebwrGj2b*@
zg8vD;RMEo(d#7iTP4ybrIqA-8>~%tWZ9t8zFdPxFZFbTrdgpp|DM;RN3Qmasefx=*
zj_Iw$Tb<@r)jlMJ!{?J}?1)}StnL9CN9&IH^C{&+i|I9MDm`|W52<z@WN^^}NU$$!
zD4j`3Ws^+!54jpPA&~Pl6R~0ewwgAIRnu6&Rq(rnVONdJ{=BE<&)MfDPD!V<Jo{8}
z0`J1i!AsgZPaLa3<T~Re@u62s38_4fqC2RO{XgvvS8Dr$7{Q#I$(MQ`CuW-R`1)K_
z=sqnE2pWJdLpy06p%$nC{2Er$Y|OSxt>;YK-rsqqnWO^SLFd^nR0U>-7%OB&U+PX$
z;k%c`&oLcQ57GqDG|U&8X|p==xue%hqcbEE4yu=X)(ro91GE&Y7XjLeC``$C&T~(G
zta4P1is-Ip&dY!(sc;ojj%Zepi<my|k;HZI<j4)Y`f|6(Hbx@D$39;{0X#ZyJ#xm!
zD4g#$p?gjO8XH1PWxr%#|KT~!ljp96J+P=5yRosZ%9i?d7_u@#CN%O&TqWzqKdhS(
z&B}t*uLr@4pv}s|0yl!3t9S=A&C`!g-@+qKAHU`AR&dFqf$$U%xd}tQ4TR!}Mw+VR
zficHExf{db*^TGy+WKa;hxwlZlqq^(KtjSRWL!2}Q1tG!9h+{78k-QU8{nERyS&YA
zPiKZxZQN#udE_}s6~6X)Eqb`27kS><Lx>9jF~&|0y0i6Kw*mbHa*gbRPfz(I$#bjj
zEqwNm`8tG(+%sfir~S}8G#y#PS449?i~9ZEOMD=nGR_6D#b5Rh8v(`=LSGn=Js2%y
z#I8rW!I8gd%i^)ho^Zvp(|l4btwdt0vYW$duF;s?Ac7%B&3@CotWU_GqsEv0x?UoO
z!)6v_c7k=d^Dw512Q=?>7vT+r7Z}o8YK1VO7#@D&M@9O;1J&I)ux0K>+%QBgB4|dc
zxIHdVcTS{$4Yu0DO(^^F1&6zzB!v!evY8O08K623$wW8fVO{7`u7Dz5qn54vVbbl!
zr%}0d$3jPL5eY6?^+P{}=x48`%4Vv^U9pyV5N;{9{|47an`h4WQ29uVR-^>I_{uNL
zO<P?^&vp}qAM5>j(TWI6i8K|2*IK3N+^l$Q1G)x7h2q%tV?g`JrzL^VMA@!KbL9*f
za(sOzdqUgfI^W{nwU-XwrPAVTU1Tj2`CzXDZ>{Cr)JLnn*YnLPU1A`22brI>Pk+0I
z?!q2%j4<eX^r~#6xS%}gu+bZTE+OWL0aPZmS1$4WisJjhnUX8n$sjpnlS1dEIoK32
z^H@+*&reH-o2uT)WhV{j<!|zv$V9WkmSQ!9-+P>B6M9wi?DS8^$bDzF*--s-XLK|A
z7&E^{{8zj1+bO!Lym^1fGyV%EXWNzKV~JA}Eudy5PUY5VHDt<?``fKcQo=WW`?pE6
zcG}*d)tJZnG%OPkdgRe7$xYL|NvEWv*;qD8ssVJwHr47dX#0)dV%~h5Fvi#X!*<ir
zQ-k?DDr1+MmFnF&G(U-ps=&DVDVx-e5j%ozIYUb-8qIm!SZ!p>WQU5}6t1gmz^eKv
z`ZxS2T5UP%v!7GLPVY}^{SIwI(Qb1~1sa!BKi!GW$7)WuTKPoEMdn7_ri6dguwUYS
zGgx!vD3jlR9bslYIO0N9Q2%b^ag=;ijhoY4kJC?nwTQ0MMmIrm4WJYs%9tTfdpg2`
z6Bb6wLq~kh+V6lpQH7;T)1QWh>&oK0YK*e%7x%(4-ppEty*;NR%6nvu?^*ZY=<gAS
zid4hJu*-mNg#MeeQ-g%0d|RNVXZy8`ylM1Yw7Z@`S76BXC~4E-Y|iDIPCi=|^SZUp
z9Ha@_i9&oyF0Ki7Pv5M%xa7!4+y1n}#jvD~gmV{Hdk8nb*2ZDoa)w*o{^kQwUgaHg
z71ERHfB4_LAnyO7<~{1lX)+i5Qfx@dm*EY8+%V7uqO?tBF;G_xD>ozQ@q!HX01Q8}
zSU~R2`x4<DE@qhMLF6WIWnKYzHtE&Q5aafwR{PV|ne-vl)`P*>^YJZ&?62W#e|Tt8
z5^rt+mueHi!$8~WQrW%DW4BAy04EO7*c$WZ%Y&jHl~XW(M}Xg}F;$S*lJ&;_gT1#7
zi?VOmg+WxBQ96V{X%J9Sazv#irBekYMnFoC8ajp&5Kxfr?v9}m=`M)@>6(!mYMy=j
zyzjf#`>k)Uy;ttN*E;xv<JJQPW`6g1Uwxft6onmj9Ay{l(OCKgXRT5wb;cCQ6x3EE
z4c@3J{<<as!#3l<BvL>anH30=jmGqbH^aES<wxZExiZu9jc}rK(IuJ-a+YP$$db$^
z^dZ_!^tle``29WW5{gJ%xl_<2;*B$ebkm!@k~&d4usN`9z0da=d^8Qje3|l_Hzyca
zr%?3)3h@NiJd7QnMH`w&F>6oUFMgVXY>v)2(?EfcrOd@L+n!yjoLNj`o5)BMgC57H
z<JJJf>C;CXzKBOZ&vaKHVbb)N%+|Asg}CAk-AS675>2g`o2@LvE7v{X*tG3rw0sAG
z1T&(v-N2TT_t5`1#2<J9@ok~*qm;El@11_dMbO6Xc6rg9^Le)+ZbPhXtw}^Ete6-=
z+kb3LjZhKp<h?VDQo3vdCcQ0&0%m#FXLmgDGUkl-7T4GiVn^F1N%r<sSzIKyi!_+q
zGe~fxKjC6P=g!_e<k$fO#eL^e68$Y8uX8b@VC(txt5WsWhg{}6*Hj<xnsA}IVN@3a
z7~eEZPT%A4#BHSp<&deJ7)N>9^Ohh)_W6LwnEI@pZF(KgSpVbci>=xV7I)oCykgk?
zcc71jqlFR;$v|8BA9|UPck%6-ecFGL^-$xZ%z6nl<}$qn-)gZ+w7uD^E%AFG{1@cM
z58Q(eX68b?xsQS-Q6yaUw2?vtVIx{EiM()Nk4&p*cEZrW?L$91d()Wl<Hw90<jr{T
zYktbjnHHnnGkSDRdA+^}!-uE9T73G@s_A%5VU|81Hl%MuU2bCVywQa2lR(*Fsc_P<
zHb&OLlb;m>%aHmeM$X_3OU8jorIL7G!yK&OhwvW4h~4bY)f+z}^*YVMJrdsHo=*LM
z(ZOCa*v}Lmiy|gm2pBq<AT5Xh$@pO!>hcN^Gj##?$m+8dfM;cs>Gs%UkK9=C;he14
zt+_Pq0n>)6pkEL^w-~E~<O)t8tw=D%JoS{BIw#ibx%|+Jr8>_BVw#$Y`m)P-=Wlk!
zIn|$xM9fkIB2IM6sg5lFa-z0>IZ@-<>YzagA3+Q)K(-3UNh3NNj4)ef<l9ymFE(o~
zZ-~f9T(BUC%z6ogY)`P$0GXdXg_Q}%xTGQ~vEOF)Mw|7b;zA*BK5P7}BAw}De5!L0
zV4nCAr+fgJ-nrAABSw&VRh=(fy!S1D&gBZ}4yU!Z=aN7-sIhGQSLqgUrwJ}<Jq+^m
zVaCiB2W>z48Z^{?Mi)t8O5A}0TJxe$vk;c+z((3ZsO8e7hh{YSl<YVkV<e@l&2H*9
z-T6Oo%Iz;Gh`UlG(IQ`#qe71HpP`DYfcQ<&XVjFFdAAk939w@L4NKx8hkhIHk+(Y{
zD9O)0UG~E_j5(C6h7bFUrMs`!R91(+3Ia&uspN3LK1F}v(7B1q0|U2>yZ9z$!5X9c
zW)D^D1#W7BrkZU>9)kgK10d<|=z)-4$mDN>L2u_ey>GMjmlgU-0U^7ypW~uWX^1e0
zU^~(1*HhQbMMFgfDvOwzsq0>{Fiq`#Oc*jX&{MB}w%gY%PmqK{h#BZ__0v!PXiB19
z(1Gm*h+Pz#Dx$!l!hLzRMM8{nTh^cN-G5)=sMuqqduVHa&g02a-69(82^leKfczRa
zJh<3y9ByJrceA**X`oW<gNN4}0#rE)F9Uwh6NqnPQdQ7BU51q9W5st=py~Zovr8+h
zmTj$!C@-F)5?aj#R=6-BJx)Qv102R1fQ3d6&HkdZgIL1odPp=?SoFS8v1s;tzZeF(
zi3lVq8ut&J&(r|Md>mncFpY}#*lx{dx<D00fK(lrowrv43jDUEm#3z{q~&)t42yJF
z-ln%0`c?DT-cFl;P@V5bCVWSpqXH|<We)f=5JSEW5Xb<8Qa5rEEMIoc8Fb)EJ-4EP
z#2KW4dSDn%QGoru(;XCE4GRI231#NHy+>8g9|gAH89d~WplCQ;BD&aZ8g`MA`fO}H
zrPHNKka+XC_iIFcBC<&X#g7E)L3DFQsQ83@!C&20xB%sykchZkVg74k37)>DuOUl;
zsT><OyynjD0?9LvrwZDU0Y>imQ~+byp$RA|m|fo6RPzS)F6Na^?h|<4z<gL@$f`%3
z1PC>=yPQ`Tw&GGP0M=Cya8`>Up15d=PH8Z<W88p;VdiyUU}ejOe{wLhQ(Tu9&}z8e
zjP7nkMIC=0E3&Kdda|g7=7w1s4X_pYYRDN;^gQ=9t&6oXaOuxcd~5qNx}owJAtYNw
z3Qr~I@#ZVL$fh7YpL;(1{vQc8O^4)n+mK#1=f_{rpK8ve(}m=IB1b2mfH=~xA=2UL
zB7wO^3g+?M3-WeFBiGab^1wmZ4g#s-I0X<q+yS_}0~|{7>F!U);D{fLA~M6k7<4Vr
zj)PtV>O?&ryBlqMpEr8^r$A>~)M)AOwW;OW3#ygJ6T)h<f~p=;0>Abi=B#*%>}{V6
zz2iB~$)C5T79Azz<Gq`Vju@4r1?WpIx-hV|vkUj*#GWVPJ~R(wvB4ZCJ=HKp*bR+L
zj6CwXv$@L{^9{Y1Q6JyQI{SAts53{1kAu&CB9OWx)WSq*+>rnfam0TQ#sgoV)>I7V
zP>@T+)8itvWX-I~QK#$+a@n^#x$zs1OA~i47?y7|`6Ls+XpL68_|UtJ-G!+o(6gty
z+>L?qd@z^NU%5U8LEaffY%oFrF>u`rPCN*5w!#yF`d$(DD#yUD=z+|y1o(zE)ssA&
zmt;7URevLZz&!;4ImxFH7v2}kmO-TP$*yi0bMdvlKR};cuOWba03-<-{}QK~{{k=e
zg6Ax#VRUEzv+zKXVA+QG<Hf;3V9r`J8#XL5g}o@n03y@a@2Fn<RtP?8d6OnD{8|i`
z&K=jA&9A@QQ6zcAq8=j79geG(cC8)^PbK!0K*zR;2M2m;yskY{+1hoRD@f_jEvK3S
z8c0AKQ4I(O_w+0h9(gpYwNCB>>nV(*rSXqbCY;(IaPZ0JQK8>tBUPaCbukRP_s&N3
z2I=0^x-et!A@aFbCACqhZ407bvZtEFa<^h0w0VTcu=_whf8Hr!x6Qv^GXEe?vGh07
zo{*<JW~QVk+s(9truKyhVZ2rUL(ZdxX{t+{Y3z3xf%9CH^c3<q^!m(vv+e0nslnj@
zzhdRc+m))aYWV2v!vHuHQ5qoD3G8U@Ur>xax%{$w7fszV0`(S;mlW<^oHxXqP``7q
zziVnSB8x+X3f|oXo|4sy%QW-U?i9k?n+qaZos&Orzz=~9C%;+~ukF1vV4f!$6N-}P
zF(q=@P6l@_&>_b5Z3kH|`^3G;((*61!$e$(&PmAev3nhf*vtzVWHL{efya!^nO93D
zkY>5_1dxz^=ufSHwVjsU7`u=fp<lWaN{WC0wXm<n9BtCfQ+&vpvJn^J`ka>`<%SaG
zu|O<DbqNPF9lh_hpjE#t1Et<$hs`soLf%e#p19}|_uGi;OCI!>Iw*iRkpv@eJyIpX
z`zWvV!82T{k>Tqi@84M;f1I6rR%Y1l_h?1Nr{E?IIFKg(<Kk?etpokQ^=}$04j;eE
z5eN$d948FtzzT5N{i7sCCk=FJyN7o3==z(K^{1LtOMCqR9|``z0i@f0PF~ywh}#yF
zq4^uN$1mi%^xxFD{PdC+zLc91ULFcsphnMTVEms{O!&P+ScixJhvsL<6UsVWrZjXp
z9FVw!WxN2Phq_eUNf|n%=@H}nIp>8xKdE)i;dle%(P5@DYDm%uWTQWDbE$g&z=;pD
zA37*~0(j&=XBg2OgBOcR?O3{5O#B8VJr2h881M*wPII34dQFxq(|4QEh5UQGWK0sw
zGWg?yPE)fjCjf)pFZ)^LV_Q`2?nBhseup<pl^O)t^^wu~P&2x;d$(bCP@oxY8fsd_
zaY3=ZjA)ezn=6c<R<Z9WSI}Y9?8HByk$V0z9vlag8K0DmR;mbk(RSS&Agt8i6Jk`I
zlJ-t^B@;jWC;+q~)Z23A=SA*TC%)LcV}bZsx|wQ>VO#?L8T+Rs0I`1(&RRv4=}atA
zH~7tg^WrtDNk>bGOn(GWI~&~h^(MfqFRRv*#R9H_WJ|O*<}2p|0siHF9IEph06LF>
zkNTl>XRmk#a_2rxi*Q}kTk5u_>?y%{d&EES>YPjKoSUgTWPP5#vLwxVih1=oqhuJ@
z1n=A!`^|*rvQUzkJvpz3F*v07Q%!Oaw<ikma(Xp3_SahT8IYX1NK0Tl!Bj0c^Gaun
z!Iq7qP3>hi)XdK-O1zIJSW8z-7#%XUs1e&ce6Ij|{jmc0w5H>o$LRSdXchWChKQmP
z0fNI4kkMr-_6{b=^*GnRs&IbZa~dC@1RkE-)U8e3y1xgE^OiUW!tMfN7#*OQp=FM4
zQg<ni4ryvcWBqGf>sJru>i6`0fYAXfNF}TTL|i7W>nG$v-W0`6bVy7pGNxRscYimd
zCe)3P{01fNFVmgd=S0)qebNjGI#cwdYCZHZQPN0t+v+I4d(YQDI>iGu<6}4{&fi+=
z|A(%BN@p0O7Q0NCFDoQ%KjdIMG+U~od5P5QZT2Ftk^X)+#r;i%Ho^E8_SB^>2aqZ+
zIqRZ0b~Isb9=eB9Rql9-)vyK#H_j@Ss)MbzBbBU+TgZ{^>#MjRf=49<fBisN79`nv
z=C}PwdH;NLJphhH7mTNFKIR1J1>d4K-aef&xX;WpCn9ZWl=_(>*4Ol_dxD>*Oy<0@
zy+?)(_1Lp6M)PV1KQbwvv&bSwk5+F+^i#VDXQ;&SQ6jSn4oq-?24lFqNM}=KJqrnT
zVd%?Gvf-Y41P*1nhERl=0pmz<*kB6G+;S&DWrsa$d<&zI7^v!#>q`YGfC-;L7S$4s
ztZQoKSE(si{T%e5Iv-FFtnFG1V0w_WoB*F4Yd1)%!Hy*L{d$}FX1<5vow)8K0;K>8
z#<BtYNWA&;HnWV_<hRaZ!zwOz=#u44-x%?!o%jjt#zp>n-{IFj)Ce$SshkN94TeLk
zKqN-i^laE{yM}9-icq^>oc7|Xgm|U*5HB}H{KtWzo6oT*ryd{@5=a4tuYELu@KFv(
ziFb=8A=j|rxgXDlho}S1W8_vyiteh*FjX(wn(jq?PkzkAdv0xgVo<x2iB=u_zMx!K
zU&Rmrr~OnZLA8V<4oHU4EbT|qX$J=!g+KTP@Spwm#jaVih<R`vjyF)DYoq;bE{WXE
zmsuMIBRq2CMZR3?o>CWj6_W6X?#|TKDzmyciRLh_De%_X6s1Wz&>qKTQ$OTb$EzsC
z<6^#&e7iX+;Qmg<OU)GE{?mcjDDwyXfkViSB8$j>$`L9l^O>)B1-39ZXC`r3A;*b_
zmR6YMqN9EFX(`IVBook6;Ht+=pvT(1DQX6{8_QWz=2%YWq^C_Mo65>xD~fRxd#3?|
z@6rUlABx>#m*p}o8zw0YRRlbpKT(ro0|3eglhMODFEDTqt3qGa)JB*^3eghbZ0h9w
z<07OqCt#OqH|P^WfTUfkYP5(?m?3qyPyUzfYilsA_d9%5D4?K7xfluCcrgh<5vid{
zH>Z8v!$<!-V&Kk{VqSOy0ZqD^9#GNbg0#nZlVCPK<07TZAV;el7~R9oKOYB?(i8br
zAX}M7?3|*hx0y@>gkK!h0PLV9ea9~_g?S30Ab(+M%SUuk;6X&zZsxGJmMxQ?4}?UT
z7&u!_=?9)qC(?i{A#s%drd4sUXaZlXAlcHr8krN}v;Ff&*a7d`mtQW3@q?QL-(<YT
zGhVE_X&Btv%R5EK$TMjL>`6Ym1g6Aao!if5wojhL!-qGqLSs}B%^l1D)2s>P{d$zD
z;F$2yJSFbMfTzM`syFim2b!@vbXkpfi@J6gob){~J8Fwj=4Rc-g<#PD`?tfSb?+pd
znO>~+rM!urq0Iv!QFVC-ig%i_B&1GfFrb1@Xsw$HtwJ5A-Wc3nc>P2$6&QpRxdaNi
zOJ5PUfVFECrG=9hRY~nFl-GqC0)M)>GAb32GqAe$xNk!2ho;Rv@z4oDyV6V>4LODB
z^V4+fU$k>AK8sx^$M<}LaJm+_L_$RL=hKl0BJh+(?+1A5W<nj@UKr>LZGW@to)g18
zy(1zH7$qnww`&ZrEQp{BSr@zG1o60VT>+VUlvXndthd~qj8*@9EUj>%;h3%^22kiC
zNs}zHS?)$%ZF6Px6HyhU;!}(W>3jWB;>&V`RWlf1Sa?_r6?Fx&Ht3JXlV@Jhp2|WI
zzgFeXB`Sev^<ZQrm0-b*xSJjkZ2vuJKm=Kw&Fv`4RUpm;2$V6}bSPIEpXgR*Hz{xJ
zwLJ!yO{<GdFkw24<3|j^4J{2H`-DdCRF-e4cqsyIa};wAU}74b6{>yTe<l|~mU;i&
zG3>s%kd0R_7~uUK`7NnKT#N;I5~HGm(rc*?I?N~&Y;Fp^`_}RbkO;`8RL3=OLT@iY
z0uffQ;FASPv`Wkm!~WdfI-CzRzup7A5%gTOI}ku&?w$iv<!!>=T%I6+p>^ynLstre
zQI&gKjE}?R7?ZLk)vwi|0GBxti@FgP0y<TKn29Oy=h7{?GT~AjH~Y@#1Nm0*y5aNB
z%lrWI`*0Eqnks#G>_wrGDka&d$Z$$@Hg;?cJZokb%9w#X>xP7o`$_E!<k6Fgd?gA8
z7Eus^0!zmjiD=V01F&Bem?X`ZN1>`*V+_d1XVb_8ku>cPQcrS}bSl!G*cv^(IT=E;
zs!p%ddgI3l2P3o!2$V3eW1tP|_ihtXWtK|mCZ`N2bF>Rf+wc8><Kp(HmD!AR-!uk6
z12bQQn_fS%FT*6XyN<+j^u<pKzKd*5;`&|rW_zJe)kcyYTnDr&D(s23MRse9@13!1
z91lXW4J%Nf&}uF!UQek4)&6&qjbb2dpXnHTF>0v2;b*(y<+b*$$F4VHZhxKZPDtOJ
zy?>2?n3MM&xP$T)!0k|bZ2RY@SH%nz=$mX$svsf6(v15hWhOH7I+A<%jNeM<z=06r
zDzLxzjnNA(WTQ~!*!_D11!)#T?>Xu@gr|NRS|3QQYX^u13aMV~JnsXOZn?gXwJAam
zkrq}wnr-sK7e2pKta1l^2V!!NZz^GvwFSDc^~Dg<%Yf5t>TU8*53y?np~Eq%k7%hb
znP0pwO}oe+$NjaqJ9kbyjh$Gy5i;E*4huV6OdLENlNdJtta!*82}QF2YY}s3nx(1G
z+!0A-qz8sC^QYvks*#H0m%Y*7N;mD&8!G+IsNu`BYY{<2-_n4Iy*_tfus=p6ENfBl
zaJFVpZ54i|d~6Dwfait$Wkz)fdxBb0hY76?*lL15lUx3rJ0OTP{J^wP2SF^!V&-Yj
zjM_`3>Xir&BN&9{YZKw9<81mq28Y0E*LOgmoj-6KkX)X)*;@w1ffv#$9rvns*E9U~
z-bt$~F6n+3m+#;YaVN`%=FPa7>Id!3%EoZr_>{_n*9NRDY2X~u{nf?2GsX3^`n{`m
z8|bmbk3zZDDkIo}kVJLP>ct=BJ}^&pq@whJz7tfM3#HrPcId^o|HAEv!Ef7|SV4<6
zh_H1Tq2TVE26zLY&PD<Jplc5c)<Vh*m<PxL{&zVGd|dmDDWD&ifK=vH8in#y#dr-5
z2GUv#sY5rL3ol1bH4ir*6PtozZCX{N%AKhd5xXgIX42{JU6kM4n(H<$!~3~Jbbd1y
zc+WuFj@6S94I~v`{^n|aE8<2<(@!)#L~(RYp0y;o<P4snEh4wrBR+rN6u@N^AoV)=
z)BC97Y&?8SEDIQe-(Pyz`>4pMmExQGvQt&tUN4?EJc%*<(aoqTCxE;0VcU=rXz(un
zURS&RVLU-+++`?G{HxQ9l5$Vlyu-bRz067qw1>3IEqIvFb1`5{O24QjVVR#`ODDnl
zBtTsVFi2kP0EO__jfX3Wl!X~(Qcp?Nd?4|?O@c*+HEg5y8PZX2S5D#aApGqUy;~dZ
zpE?hJ?XtebK0)Mo86pA4xam3Yw4=dm{LA3xT}G(90_uHmS{}W3oV69c8aMMppi_U+
zzKG#i1bAl&Zw!7xirG4E5-hrWxcB51)486Xcib7xIWXVccDD(<Bwsy&y|;zmWl;6J
zQXXnYqO&pozTxXW)fr>BGi4R7vi!(fNr{}f%~^}D^j?7t0q}*{2hkaktz<X!Qwa9b
z-Etx<Lc1p;Wl4%auR#VeKo`dAvJvQ6GO><a7#6v0BjaL#k@v<o4fUgD%qlkP39n)g
zxHx?9sZLivEFo@L5B9pGZ|UshURSoh_h;7O^?G}OWVlI#X~0mi8eyuKaTAztF$H-F
zqjllcoIn!*w56LMC^D|Mn@V@`sI_!taRyK(W)ub#7c9}>uG|>9oa*<L!^Ny~_r#u-
z&$luElJAS0I7?4=Q<g*ZTL={kOtz_MTz>^M@qr|QXuXHLB~c0RYGZ0O5}O2&nZy$d
z_{fYQBLTBhmzOVai*qL=%}}{32UZO=zWJMHH|yUnFMlXD*VezSmXIQ);yq-`<o_$v
z>JSJCd4EgJ-MD@kxO|x}DP~`c`G71(x5QEq5c5&*Nh^jNo`!bQ9(H*}0i|pK#4Riy
z;B(*9i^MKNNNlKe9mvV14W2zZw75>c`Vv)A?L2pPlz;8*%G>vYJ%?16Pi-35Q%ej#
z3ll%Gzl;6(jaEl43N_ch0In0<oF$F;+PSxQhxyJKAm1*(QxClFK1?VYp;Iv~DQ|<N
zL`CW-+Lb&$c%#uP`+V<u<WC2~RFo$F%%CaWh0%yJ_?nyMZj}Th{p@YF<iba;#D4u+
zjy-o0FA?41B<fmBnLwI1@5zW!UQt$qr%=b+6Wff))uQ8FD?j20j(-Je$H*hDw9SGJ
zD6v%=e(g8z;|btA#C;@BRuWo=h_J&Z*IH1e@Ez){QD<qbC0yRdW9PRPYIwOC04+!r
z{{=ZPubtagU&?<7)keDp-|f4s86w(h_6?8e0cs>mS=)fUjGOY4lr`_?6Czc;mW$tC
z8SQXd{1iXXr7)*cT<W4{R6_Unw2t3IlW6w#$mb(c=>;^1Hu1qka>j)iK(%HeVBJ6!
zkrW#<(*C<vE51k!vjD+iv@Wa_sPzE;)sHDATY`24QjVC?Y13)ZW4tkhMYFgH%p{)F
ztovky^TQ-0e&E9?g;?89zt#g&vm3<2xER3Wm`I$fdiP6HT;@rGe*i4MBE5{i@}Z<E
zG$%Eq@0pM0zFK%JH^YI)0tfTwcHuIn8pWqv*p>~@nR`<Njvt~3?awV@+FR`E0Cq_0
zRLIlu^d3`zQCaM3MBasE-hlbG^z9MPXMw`Je3M6N_=6!pZQ|wb1zQP_X6KE#5F)md
ziy|%p-51+{o^{XiRQJG>M?zpC6ZxCSwI3#>AInQVm?7wZ#D#wN0w1|Z0!+vA18YIj
zWt}(d0`bzia6AXj$+2DH96Je1r`oGwdT6`!pymn5CbOs`hjljH9JvqO(9Hs%TG|gt
zXaxHo4N=Zugsjd~)^b;S4=&@WW9><&t|9mhfmDe~QJP}%0c7+0?=0X)o|K&RU&DH7
z?>2YiwXpLC7`ajGtf6uq5*PV}sN$k<oc_o@3`~5e7VeEZDap1>Y*9B>%}GzqCS%o0
zOOjg&(lB;aGQh!l!)(A9j{VN4nMJ*ZdKCfFTW<O}_og(<pZv_1<YDf+hKrMU+WtEr
z*{=Ws5>f2-+#|?$M82GQ$8~d$Dd7TORx798jj!~On7Gi#krfC+0G||xN<Fp(o|B}`
z2i8mssDZ-+g;u&ZAzk#>+~eO8omVac-I#b_!)K?wc$-+NC!~d~^X?bfl4lByZEo#O
zF_h6Mn9@mO$n&aETg}<;vl4o@UYUhH@znwP#-U(*&v&r3zvV>J{1{eP*u|5@n@9FH
z+`<%_Mk9S4Z+vX@jqO(ymAD-}9c}ty_jgkIiIedrCmt|=@RnD4UF-6zr!Ca-dV;8c
zKJs3KvLIte8r8)Kq;2tzAdu(=sOfFNR==y6X2Vpgl622Ae*uk=QVff|-_brXrnHB*
zaO0%mwWN9=G>`s459t_7O7&T(0wtM0VA#V}9?jYlZf(HzXD1(1k}1wYPZ8*ob})n5
z3W8h_9O&XIH(o+uQTfr2BwXf>?qRjcU(;S@SldJIJAj%sKz(o==aP?sX_VFIohB~N
zjL1u8>F$xT4~ye)S^Zg-DHXfSc)O0ka4esDSS4EjEPolD##+AO!!wcxxNDK^rISde
zEk{X?BC~$`Q^(lxbKNOZ6wu}gHsM6A?^<LoY$e`I&z<Fs8?BlW8H5vtESbYEYyNyA
ze86Af#I^dXiUoH|2C?<YW0F`3V6Osz>N?HJP`f?dqNh3f?Xa7vP9jIkdwuf3Ct2?P
z@Wa$;U>7({NNqN$T~rr)XMATjI*V6mvLp||?<nhivTG`Hs=14o_S>NUo`7@{-PJy4
z!tel-R^O_p($3lIZ36VRhS99NgLr1~s%m|SCy78E^aAr|jTzI7Qr$XRKc@l)f<WzR
z3Hj4;uebMU*?cc}*3P2{x8uK10F#)2c7?J&%_$t46@s5P@DiGWgIW(2yv)zgH(4)D
zA^^<ow@qMEiLBV+4z)W|rqlfZ?xOb<B8nG0JEk)USVBuQ#qwIZb-r5*1rkgguTK*D
zDyL4}+d><kTVaLOwRCnD)FuuJniDsPwthX`{VD_x4kQWuNE*OM5MO{$DhYv2Y`N^I
z*|T|fa9eb`i5p=Q;1|Zll<GHZ<<)8@iumrAAPvI|LuR>N|D}X{Dt2xTYvVYA+S;6S
znG(D*>>~_~x5fR;(>cRh@sQc<GV$_l29WN9$US+{t^qDwMZ)%EHaZCpl8fjkz73V*
z-K=<pYN>oT_j;5++x}UNBTUA0oV^@5&)<0~q?(7=W4cE+@xgRaG7Hed6k_lfNkyI$
zD=PL{A?W~fEd{#{P`#zXVT{!-Rl|}GU8x$%4)*q6d{an}tQ`smTGMRjAZ+^?|4tHe
zv!rD2?v$j3CuG#+A@vCPLC6X|n;fuj`>3>l>In)5ETjDRhC?ciH@TbrioLH3n8+=o
zU1C?{mjFT_V&vg<{HHTP?o*bsINp@lF)&IQAc%q7JTfc<R5wM#n)u_R`D}gk?;8DN
z;Szg|PsiwB%3^}y@1%b_L^T<lH}@?VdOyB`548Nd?xq55qsI^VoUEo$+02oj$gBvU
z{@!A{`|9FZ1cQ1Kwtn{8;y2x~)JlM97q*<+0}UfSPZn+xM23E*887uAt<AqC88j$W
zcw+e!n6vQ&R>_&0sRw(ob%5;#0%Ilz6wMNUAj~of2MFZbYk2Ihb4zaf;l?A;eT}XV
zgTvoFt|$bMS7A8T0?*a;BA~J{q<Gym+g(dxT#k^2B{Pl%#9ePeY&U`mdt>V+=oZFc
zcXuWvXsg%r@T2?1m!1{Vk$DzQI|5(=1P5!#4umB(4kA&o+kfEL%}kxbQ^s1cEC>rj
zTRg3=<*mPVe%k-6z7*Pqka_Yi(=6s&>4-*rT~m-L=H=)}H1Q`0QytsgS;6i+3<M_r
zN@rHEgpFw3m7shJskL}4Evi4_d{^@jN6Vey7j0@g!d4pZ;x@1n4lFyZKSw15+s5;F
zXx+5-^BL>&tsfePAKo`h>*594J5jqp)PH?y3}QuGi?;MLnR|?WscCojS1bI$Hxd`u
zaB-n!55Q&;C=6Ut+XeDFYa44)MoMg$bNq~%6cm53<M@SYT@XlxYHcevMMr8g|KdAD
zBFr`Ep5jf1|8^e_Q7P4r63;u1X$1W11lt9H%?iu6-e64Xi#V2Q)N#IqSnG`CU4ZWs
zx~(>oiY0`hc~+>4Env$Ya;VfmGFE6|xKJg@FSyUoJFfQ-jzaU)9r8X;%`8cao?>1<
zu6FbhOASJhIYR=l#AA~D$hk*5&W!{C4&DBAM`n8}&WC^@`rA16|La%p*RZyCke@0T
zAU8BUL7ef4UrLmIg}p7E-d4RVIV>fav7xU@)ZC~X5cqs>a0)Mw$BF|6bhUJGAHz;_
z@x|2_w!03jUp9Ll0LvD-sG&>JT5vRwFrG&pJ79RA=Te?`M?8rw-%m}R*?*~V-J1JR
zj&;V}tkRJE2Vz#rtAGEuI18$7&o}TbSOAtc^1H6zZdGq5Kv?sAj*Q@>WU$ai=k`{j
zSJ&5$hiY^YEdj56Ajyon$CVgh=*q@j05hh3OIGq@9PiNgPM7W3CJ$ru&Fp1cCBsXD
zvlaEGPcdUg-;AS{RlJM;q)2}#yR71&HC_d~Bm2^x&%BP%3Gb20j^Q0Io7zCw5}f*d
z#WT;RD}gPKy&Wtm<P?qpf@gDpSka=?(*#}f1!+&Fujr9)s#KmAK}<0Ik+?>jbBG<0
z37+X$nh0(+^<Q8@d41xdjE~=%vZebLvy^whfWB9Tuw0x0A!~Q0ENk28MTD;<fwh`C
z6P|-K5I9R`y8ntjljmQQpDnp?|2YOpbD0NJkE7r)LWtE6lN(GK?Q&FHUA-P(w)ah&
zv)DJa*t2UzEhL2%6P(8|bQ>y~SajJkA}CNgOyP^s-BH^W1*~alIp!2?*vTYTg;qOO
zIWE7~$12c02ydfue2CpBWgP}qcU^A{s+OX!O|IQX)l3csgChW<7wcJ<$x^QyaS>Iu
zu1<B4oOhq>?YQIMvrg^HUr@Ln8w2VQ817p7iM$%LNsG?bndII>&Y^gqaNi&&Uu=UN
zL;ze-fso48`I=&Au5{xvZHR+c{%uaU+Ws?wQ|J~D^ZhmAySn~GzwPQAuFl~K16LTh
z!oU><t}t+gfh!DLVc-e_R~Wd$z!e6rFmQ!|D-2v=;0gm*7`Vc~6$Y*_aD{;@3|wL0
z3IkUdxWd2{2Cgu0g@G#!Tw&k}16LTh!oU><t}t+gfh!DLVc-e_R~Wd$z!e6rFmQ!|
zD-2v=;0gm*7`Vc~6$Y*_aD{;@3|wL03IkUd_%DrtoKv`X*~OL^A2K*TT7E@!VCVX{
zjt4Xz`UeiL6NWypN!G3z*JeaB?~J0qF>zdn!R-qq=;p;14VpE0uN!}|O3ytit$x{A
zOZS1NtP8ehRmJ|6amx0o_N7j`^a!I0Rc}rG`^9bu_p6DLB&@;h!`vgq6OdJ83w#Ap
zRRUWt^mZPq{Q(=Yyx3~V=UYAw+2t(Z-<<*<%>03~Um!gX(8x)pV$U+SV)mHEz5(A9
zcHsedVdDo*#{h=@62y8oe}QtH$1|5hUG|1mfJTuPqM^pb+{iffEBi=4YL+EGqdp3w
zP=c7gD9}Q)`tQquMl-i2@ON*IXy`$g&9R&e2l)w|I=d4}<Kk2_G6y~N#j=)s&2Q$-
zdyqP>#(CGyWAZP#8O4f6jJ}o2iuSAR^RFurNRdGwhW(8s@#^{)vdF7*xH^ab3=C-7
zRdd?165eeSwy8&u!+J(xjfNwB`O-SiQn~AEsnhL6e|%9q9MCJFv~YJg%qL&dJ5yLp
zcIhL$C>qe(dw;0QMcwE^bE5<Fg!_j#yEonwf}??#sA}r2^JmVKI&Te*ihL~Kso6%6
zzvzN=WFZkw-^n948go=+*?1qf|5YI6w@AuA{{<B<U10W7(Mi&(3Eo!#oi|Nz*1rvO
z4s+Jw?FnUh8lv8Nw4C3&%oB9(!?A7H1JX1ma0(sMB43aJo%L#a=hTbYuYaldC5r5N
z+;6x~?p-}{hvPs%D%ada02-e^QVVO;=0J1*fz!tc&^x)8aTZTfCvwaR>Zsi|5|!0H
z)7BR@tQYTZbtlS~eVhM8;@*c;hh0I${hz7Df{C2P_Z2i`DO>-gZb~~kA@eq79KA*J
z;Ij9Ne70^zc9+z!(O6U`%|jousk4dCp3hyqkMBkuL&@O1!#QVE{ewAD-jk}?371>d
zR7XucgaYMfQ*=*an*9BD^^^&j!Vy-3d+_DVrWBUA=*m4O-B;)v{V*80H*Lv>S8r-u
z`Da=3GRkc)6d6R=)$Y4T+FXcdSN(zW{xM`>xqem2VV*iGrK^F9%AdrK;5fSo{twc3
zf0Mwwy8i!V0sl*%NfyVQlutNS+ImMkimbmzmqlW8hC77KKc>aDrrpR$gP$$D;kkou
zbZ##bmili&?(|=Qq>E|*6N{COy`0_TNkVg6cyXLpPXuA_CaVQ}%YgIDi240Ck%4Tg
zGHu5IGf_oiM8Em#Hy0|#!JRKk@CwR~?@X#teade)Og|8Eet?N#XO?`6*nM9(vNaiF
zM#1YysViAl=3!-8W$9SK5?**sx~jS(D~j%WIsUc|g;qmU)_`6U%SPXE^2qdvQK$61
zx~P5b-lIaEW&3^EQyqR%4QHO&XA2qS37uts;E+GK;FejIFja(bN144SJI(wtkx_v_
zvQ{6)Vow3a?SPo6^9tCaP;-RZDp<$^2y6v0F<_(|L_<R~Gz!%Tq0bJ^J0nhEKVowd
zbPHNGp}Y@MDrC=hVSoC(yIN1bx%W6^JL?Mz-5VTjMVBN6-^&fN4W3v_OEK2iqpaf0
zfG8ccDlgJg$>+`F>}RoBmQ!{)OSk+lQ>v3?ox1}c?S!8uvBhjzyM53PuFuQMs8Qq%
z1<cy#e;u=38b#L)g)J_ePcO{l?mqeT=ozDa<PBMSFIzV~8A&Iq?ea=E-}S@a-wf|%
z36S0Dccc7fmtIE)*Zu^naC%HW@|bQ@Nh`7n=hI&};vW2+W&KDY?;#0s)WI7QPR9?)
z4adD#7-d%-D{IO9E*?;_e(QYPaM{$cF0Jtgj!hLyqF4eaGSGo?<ELAa<HHw5742OY
zeBJQOEmt93UTCS&V{fOm(sS*6`(*<5)^{2RrJwOkS?)$7DuO!jNU_Y%7K}z7^fC1j
z|GbO+U%B6_@BP1VZcy>OUS^`uQs$_Ag^c8+X~K;Hl|qBki|U&hw(c{kqHl^{ByFe(
z>{)A=IIZwUnt)Ny+leR5#OMef8Z)J?a0>G<c+{y^ImuP<4~LIw<#L$&e_^C}V4{fQ
zF*@}fL0xUn)QcvBRq)wojXAO@yCt2xv`nWuB#jEq6%pxLzNW_O>VLUIG9efKb5_A<
zla@ZA$DV8x%V|u|SiPS3%8mdx&|N&amHzbH7{hN->V|BRW$SD03P>@M2a7l*9%Z9R
zziH#X@P#7^wP>nBmr>rsOzg4C?}oFh^mgKxI_F-KD9rmfnacIXQmmAW!I3&cYLTa<
zFO(_`Ys9PrU=JY%<A81P+rCVMY3DJGfTb~Ns_3l?9euuqW0i*->(1d6cKb?aF{ku~
zTqDxwa;;tpB_3FE;^+&mQ$R$gUDjEG4Gg?;l*8W*9|;RTM1*!|G>hV^QtX}PvYKt0
zDs7ebip06zG2J$xCO&HGdJVit(P5tatE8f++hoE-0e)5J5{v{(j~P?MBphEVmdK^N
zu{_OXd*@ZSaS0owXWRcsnDpH?$$gzFrG?!N?XAs?i390<);w!g?N|>zIBA<3Vg!fS
z+ZHqrkDb(CqFz5t5feVt4LN$6vEf=<An-E0j*8Noxps)FCf-)F{S4pFMgN{dkUZUW
z9lK9cG6#I?QJyT9X~N(TChl1IR8?)_x-Z(nIAEpYd1dur#q#()WX;oCKL&P<u)#=E
zq78BXMe9VD?7%U<VpguIGGV8#S#eBhM{T0HE)TI0-=d%t_Z{=vrwv>fvt`TRCewGe
zhkZ!Nm4cUkJO!n67(o=1T9EkB>QUDZ*A#OP7M>I`rxa)vrgE8hpA^G)e4wfe?V?pl
zmcN;0X(8*Du(OU9?OeP?qY-5Yze}2tr}*UK&P*G=@OHymvl()Zx@nw`dUoEu@i2{)
zhzBA+1;nAZ481Z{8zc9?`*KbOlz-DLyt>N%$DM_)=&~B-Ozi}kRbt<niat!;o6=#x
z1i464+=FwLny>wxU;2-Wbh+sK3I*R42tb2^SJ-`>mh*&YS*`7|W4|BZo@zIfKFi8%
zWK5DMkh9Ps(z_`|nW7^9fuRp7{_LF9{H~D;etBu?1wJT}9g|Wkx@gxly1N!luV~1L
z&#a}I%cC^srz^FuHJ&zdcpkQ$!)1TXKKhfn?_r5NKs!w%WX}MN%GsRz5~pncQqExq
zT8^lUQm;sy)cBlZTqIQ(gyL^Y=4hSmvzeR{6<B!iPJ8+-ZZl^BL|}E=W^#P%LfgwF
z$L;;u);Cb~v!qxS7KqZXD8JvD_KGX~lty#V`%E*h-aILL%iw!dY7;M%D56Ly{-8}~
z!=^Yg$}GF&{_edR@Zo^NvvdyDv0)0G@@O<UPjAy*?;4VA*IG_94p*kQH3Satg)&)r
zzG8`2UhDBnpD3Bad}EpP#BNnRd+3X`#JFSPO|v9SVAdiug`&A~%vUsqivVwTovFG5
zP#EG5^<)sqn!T{N{>@Kvko6}Mbc!>5OjXG%(CfYvn7k@k*$8Eq{wk|vd*)*S={dw;
z8G9Wg+UZS!9j8BQs0x%!8~fnzJK#P~DQ-QWz@HrnUg0=SxVTGWrgR{?&fpp2>cSKk
ziXyXh+?E|@ouW76RwWnyeCz$nLi2Na&Yvfzdo-gxV6<6Rf(bFoqC5Y!?yllJ?jWOZ
z3jZ$43os_w6-_Lkvb8t=`a!+gLxqmpyT&EiSu;@TwFT^?L>)+Pm5z!m6uB$f@qXVv
zN597cul{2X`I4Q~#%cmY_>ge|j^##sCoDBf@w-^}Q|~i}2;`g99WH6jf<lG%DfBN3
zRqD>4K^D4sk}>J1or`CqnN@~}N+H{NM{iB}VttI+QtT3#xF@@p)k&4r(%Nj0OpVe>
zEoE}pz_5S&VIt3WX$iME9aMEoDpS-NYgcECxW8XZFx9vx*OUu{9K$!HIYFhlZ2N<g
zCB2I%!|5lPF`vt!j$HZjF1X|5ouQ2=+CW9D6MA>s;Nt*eC(nIV)`%eO;3quFxoN>8
zvQ&@%KI4!U%L-U!OxS=bO?6VVJ8&M*{3<Ho5qlAiWff+x5pOiMnO(<w-+d_Tr6<ZQ
z!_-}8P!7#gvx2?<3EDd@;X=~PyeY_as>~?om6R*GHdtrB&#yEYw9FGY&k)2k(IM^_
z+!>)2I~@<_Ec*VR@+kk8PImuMVebFh*>G`@gf#1cx_uaBbz;?Pf$)vHD)Few>qH0s
zb+v*<Kq#+BQBui&Pfl$`+Z=so_k>D!OVFT5N`5%fEtyw1OZ&VZe|(Bfq}WAtAr`!c
z)=vCjxUbN&1+@rB01_byLjI+wqCyOP$=Q6x;oB*$*dHT?_gUR5yfVePBo)ao$Jq;e
zmo2fhJ{M4G_{6Ktl7NG6Be5op_6>6oqEhy3C;o*`i3R<X<ei+1ijSz2hK$O_`Lb(j
z8X~PT&$n_P6^q?vj23-6A|L^f_jhON>pb7}!YnPV)s=>gY?Cf+2ajZ1l~C+|`0zt!
z1-_2Oo5lF`DA%i4SUmhTH?n3spqx1NvuroJ4?!o*Vk3=P9=KB+uO`8))We`+d%=&^
z?g_i#y0N=lF%OFCBe8q5ZA_M%F}<x6>SfFL{q8q?$_Iti$7TB6EVX`*SQyELi8wWW
z8wsk3C<rD3<vq7doonkP+-FG8J;{yB-lzuJD95<U*N!bm=R{J?(lty*P$}J3>9wPI
zUzVixHBQsZL0OjSHGqcyc|ORu#z>2G{-k-$YM(v$Q*`i(j;?+kPmN|J|B1c~)f0H=
zi}2c~jY2yvR&T>-cb0=)@0+O$p6Y_pNssF>vB=Mda|}@x4ZUk3ZX^#!Ypse*n36Vh
zCBxNxya~Y|PwwNB*gZtb6N|o|FUe`Zfkv2GbVd)-ecDd8F%gLx$zRoFbSpgS$BZEt
zqE_<aTpoSNdj2+=jMn2#{c6D2yhTCs&{LFJ*h_S6u&s-S)dHv8O>Lu>lSl7b80;~!
zhe@^~{oJ#yjoJ#kCQs3VE4gq<2{h5d)bpAy_^xIt6O!Fca*9_;FsEdIay%-Bvq%ku
z33cIRpY(|SX{1PN8GbBWdIvr5F@`64oNK^*d8L7SGe$#}isk>>)AQfW`m9|j{(%El
z#O^xB{jKD))d9OTnA>g+ob%4tw*RmqKNVwFzeu!D<@ad0OtxenN}rhM&vMzxera1h
z+1jH=P$`=cySJA5OLpPaKfzTuzbq?nH$N28x0C$xlt<of^}`BX(YmzW0h>d%tJ$o!
zf1ZWJr(Xp7CUj_q6?O!&lp%th!Jd6o48K2B@Od((R1*Uu%e6*I5PO-hknV7@+)Z*|
zCaW)2E+$rOX*4WMV)(E}X-lq1`x<{TM@jo$y>U=6O}=!h({Ud$G#pkOYUIBs2tW8B
zjc5JBtzpFkI-{fiMvy@JF_BucyJMd0pQ7rN)sB7%ovgFdvc?4k;y5>m(w(Yq3>lGV
zj_5FFz`1;YQv5%gI`rBur-t5Bge<DVlJ7fYagqyZ$&6^dByyqj`Nba&A#mYmv_aYE
z&?tYNP;X<g#UaZA8yq4?HX+mA;IgsG#vOD(uUIiR&!e$$^Pde-UuJva+@e;xQiDC;
z&9WGHWy~n$rwAz9N%@>j<r(}SF0{(gHV2HCE-lLz&!@>2N({_sig{wS=HWs+vMqd9
zhj~4Ep<yQbS4PYe>8)7b!)!9=MUjs88`5wTprfgawkt;y8H?G1W;l}gDa)5SeU?6|
zc>IB*nv*OJ&uE~C;?z(mqP0+XwUv7>zUtQBHzBx2`cY-(ud`lQW!>z`4ath|`D|~p
zyHHbHO~4oYZACiR-^iA`{AnDUT3(4FGM%?dNSj`wfG44H%CZL6^dwKp9qUqSU1&JJ
z#ItL^m$;NK=X@gfT`|EbE4#$8{eRi3xhfX@TZ%<QwI><fysNGx6RhYvq!W5Cnq>aK
zVKUn`8gQqyYR}^9x8$HnpD3xfS?qy-ap8FbwCD>U^ecJ2#U|uZbkZZq74y&LSzlnr
z;cX3frbrln_n5|-P9+pN&_xTNN^sRL2{=!f;d$UDk5*r?35Oy~DL5cO4q_GA0AJNL
zY1~`FQr*MA9VTS*&WfKjXNcSId^IKiF5U@RQDfq;LxIk>fa=G4kTpeGMi~Uj4moIr
zEovc9vEWnW*(gxlXd9w%hXcu1z~?`3hQO1tSn~IO;H2RF9xXVMz?W1x<4LinP@Khq
z(3`o%n2q4%9@!-O&pzAaEP2%=<LLg<n~DOO5CKd@;DGnw*@1cdR*@C(l_aeO?~26G
zl5D#Q(RH1&n60i|8nX)P71a{_6^)v7JH)REnD|O~<h)*LcPz(b@hch~e(Rqu_QeLd
z=(V)iPMz%&?Zo;59__$jO|MK*)6T^oVFv}HZtyi#c>ZBg1CQ*J#0@n;Ud8-*<;gca
z*^kF0F4s8FIU0tXU2j7h$(pS);A#Kp_=@-`+@=&!afmN@Mn$%>4;*_?`L=BDZ|=TS
zSt+Jwv|;32=(cBNy-qqDWiU0izcg__y|LDTl|15mdy|tMhS(wBZL?Z|e|B>C8`}@*
zH>6*q^#?T-AMt!D9zU9QRmRSa?BDn74)`IbuF$OeLl>y@U`&@m#BM1b=lL}TtDg!f
z64w1q=IG-hw3D20{T8aW*i@91)Y3i`6|vyVTsPM|1eAC=i=P3>$KThQ|7AIS{)em?
zn;WDPMNXf@G&|A@oi(qCeaZ0cht|h=Yf~*TfyX7i@VoG3cqgixM(v#Kk1PO)5}lDH
zt#W(@zuv!N*CQOK<as=v%|rhl%UoYAvo7!rXEm|v?S+c1qi<y!_U>rYd5?yTAjafa
zip$kR0O{yp0!1;%qUBOQc+VF7VuO?W(a9;o)YWX(^Hs@|Bpc(ic0E#6JG^f6{D=lR
z&}v2`QyQ~5r*HIV9+3?A#?F?`L1`Y~)~NLD4??+K!Qws@xQd_gP%m~`O_S%0ZSg2E
zU^vbzK;89hO7G_Bxj4#S11Zu~fAilHQ2ieXUNnXr7>oGt50-cO-`^c3TDNaub@T2p
ztR2m2k!$)KcTIryngqr4<YaYrSMjE(4#iK|cYzGnZFy^|$DkLYGU&muW$1R8l*pOH
zmo&(Or&B+4z};#D^KKpO0*2G^q7G%8VHf%=VkP!S`!)&0_3fxaqs8T?^|B2(GMaAd
zn<mZrJ&9So!mpEKhY$Asz46AI@Y9v;^Nd)F1yB8iSOZpQ*P3s_c@(;2^$8odfTvB*
z@894Pi8@rO3|HKES7<EWpSF5V9~L3*7}>mEplPwax43KI&)ylu)I?!e<(v;yW|mZ2
zl0t;jXf_+$KtGKkZ(J7zFAmwaHy7w@-Xbo0ta@|bR7qRCzLhvd72n_h6h*HV>Jo92
zwn}!O04(6Zu>7PW=iz{vr5wGDZs(ACMxDL?nMq|1M(%d;56N0S$MU*Wni{p@_Yp4F
zU5W8j==(c(7#@&xOzV`#jlO*jR2B+<#}|JeS>a|{7OO!!zz?Mus_pc*j!$0V%z@8~
zjpFF$zAa7hK9}Su>8~5eY|N-PWVC^5;sHNMYB#f0W<tgQg}Ky~if0p!9g2pJ@d4G6
zLA`(S%m4llUH$w2d+`6e^T3T;24N1s25Zn9H2%ehsA4DKZ6qwnNecsjw4~!f2AuOx
z$;5xlHILsuZESbJ<BJZhy2Y0$Dg{i^bL^C{;I2c`<R<l=O=@HlFF0IM)FoGt^1bHk
zmkz-_H`Dv%ze6;8f}8YnFzIYw;>b68gOgnj5E^NU-j&Uw^{P)#1!`u9o#(a{Q81*+
zqR;aa8r=cb7}Tp5aVzPiYDMG53I<j%;^8)wT41pMA*~Sd-HEi`>euyO#ao=knyWg`
z0x31L{?t?c_he2AIE>}Wvdof6E0cg$lTG`CkXQGpV1srzG7LsoQi`-oEwFC6T^*&9
zyxw4+bP~)aC$4Kel{dg)<~#VPs5v=z$YD{L<oub5Yup-tQuxl0Fn76$R<x-ECQ}|=
zluDUV@0;z*`jKEYij9hAgF!Y%LNDHOQ6P_OA@|M*;XzFzgO1Q{by$+q$w`#lT6J-g
zEahi%uBY2Zt5kxSgnPb<nKeOMT-_QSTGbu*BY?*J&8lLT)1_s#g;IpUN(!|y(p6ZS
z=0$Es<!*JMNBFfl(7E;pV{@Z?5~;#SjZdy}wQ8@%HDv#VBUe{p{I$QC^uG<o8M8Tg
zv;Kh-1<W_RZzx;r6wF6`0c^hq$Af>D^-^RhvKri>qBJkx;E_;7GbNQOIwpGk(0%hG
z@MHH3PRe*m%{9nvw?bu9@t6AXl9+jkn_D((2?d7R)DISB71(M&*~v3@y@SwTDQi19
zZF8!=yle?rs;r?(7;Uh}!6;*IrIHzL!fax;`@kR0J&1b6>Z>OGM|)DXA(dw~DblvD
zdAe%G`igScu8X~MRQ0aj(db!~3n6}>5nkRaw8fpN`k*mN-L|<M&RKjN_`LtVfuIaP
zxSOZdo6}9XJ7MuwF~2Q`OHGUnRTGN*I?q>R;bPv(Xhpt?mX3v-QOT;U$bz!5^a59w
z;Pbrn`G)u&_MAwowfihCUb&=I{{9;_q#ArlDh-D0+KpDz&xTq_BS`^uN-zl1p@#9z
zBYmR3yX#M9SL@Z=3u#wtex9>DuK9z!d1ilHqxG0lrxP$6&;Q{r|GwM4A6p{{Hmovy
zgm$WXi@)hZ+)^%~@5eO#&~j59Fy|?-5^X8-(iGcsNy~zf1iC6$EnPv&QWZjC{(YLM
zW;=r|brU`f6s%z$I`}(d77hOtrlleK^2)pY>pa;1%)|ZXcr^jSCHP}m7gy47Ug6F{
z66J}nOPYmj<8*TKcfvZjcA&jjNw*g!Q?|C|qn3#Zt~X+4Ky*hX2xMez;YJEoZOz20
zK8MDyNXN|KNhRBX?k+>+6kCoqsW)mNQF(27Qq5_x1Nzf@EF3w4CR;v5HiSPO<D?oG
zaJ47qdbm);zLKTX$@s_Y@;|0gaXU`mmz$80kSc6NB5d1~)KoF2W8LKKU7APR3pE@0
zHe0Q3rLA7dU9|~QXq%#SweT3v)zjMbkA%DM<I-$_M`ppU@$jTb)sgY3j~DissA4zG
zz3w@sz)Th7!rOP9;Km*^y+3enG9A>LQVbu{^MKmi=NCgHC7>Q3Y>%yfsw<h%ofP*^
z_Rbr9ETldeo2@Z(l#04nIPz4n@PTNtryLozOB)tdY{mboj!yn&U4!Gdk*|ffpfPTG
zeaPVvEvQp{MtzvEyW4}(^RAO^L+XaE=0R!Ohz+36MG^eeLa6UhljkNK{=NA6zW?Z>
z`)~XE{?n2+^#Kq2>-m(GU3W@PG``-1$h;u&lCO57iSnkyAha-{r1KH|2L|@!@{YT4
z!C#|K9rVUy;VL|FR8BJ&+Q=ich`(ztT~@92$G{j7=KNA2j1($Z8>L$Akhi!%+jc)=
zM)jG0n%TZd?VDJ6Ya%0ormZvy{wFpzc3&;n*!kYX`2jxZceUf+=ac^F>p#=+MtWIW
zW#NDE{7^TQiOMr(q9ZY^RT1Jn$UDSwGJxDH^_0nTlcPIYJH_sB_g2?L24Jb?D}pqO
z1hQWGKW;C<S#s9Ucl5Y8VLU8r!DKWMccxM|y|6$(J(nJA>o;r)+4%u0raSMbfiBq2
z|E!JpsS+dL*SF#DErW`3PB~3zgpE?ekl>$=lyhmre-Yx(#=sw2&uov)?do1Gg9dAg
zg*`l4@NDhnN|fMhDNGa-R(&t5_I&Z9zh##^nbN7NhVbm0?$s(yRii02Up^NRJ!#av
zjyi~(P2$RVtDzy|LiK$wv?IzUzCIBuX9XN4^MBjt;NMgD_!qoZ;2(;%&I`;yp!o2p
z<V~|DKWfk>708~crr-SQk!m7_c3nQsN0PptQX4C-@oY=cHE!#94ssuK(wdkaZLE+o
zG>F;bW#8S=Hrah`cFWAS@s2M1f@w<fmBWg1=fu5uLq*z90~udBy3OqY500QC-_nhm
z+JzTSES`U|OPjzHFZ(-03P0(&uVp4I5GA4c+_fg)D^nAYl;_`-rF`*k&J;8LH(6#u
z?U`{-K5gXQ^<<fEKe#6-i>{Lk8Q8{(%$K!)A<8VHTJ@&Epvmf22%SINpLzX3MCRT#
zwUrH?$z&Z_?pA?lKT;=m(~f3h`Kg?V|H0mShc(r%d&0553W$X!2vJZ$5K*eqq9Pz5
zAiWdmy|>U38z5b}f)eSSNbiIW3W(H5htLTnp$3vHX1#mn?D^)L^X=K^oY`}I=bQP1
z3qn{AE9-gI{oMDj+-u#5`R+WMjChpa>#@@0FE|1ADy^*aZvkd=suIoYw)I0dAop83
zBrEiZxX`r#z3i;@ca4^g522H?2?0N-4e~aiCyhzxmBv6A-|^wf95?%EfmrOMScM&)
zEgP?~P_<$_$=oM-FR%8tQbtvj<tO(DHh=AewpK5}y5muLq#}RVMdkw=<0DA9Krq5U
zQGeAjpSK1p|JcQ!;m5*N_SEVz2L=28dt?1i@|%CaCKc7L@t&jW3LW9usba=Cu??)*
z>=o-@Na7L1v~`o;?=>NLb{#2cza8)l=t*qq%n7Bx#ZZGhFJUv#@blS-{wSwh#(ABP
z(<O^K=0gU$3lrP0*AhtXlvZm&`^i|QhrvpvHypRz1qSVypXBPWY5Y1rRN)$ZN3%Mm
zt{S}Gzk<8d-!L`($KE>^XS=7z`eE|B3x{$050xl+Yw~paeS}1-k6-2w<Boe_Qk4(R
zj8)lJrsa&=S7Mxe&KW<NJ9Xm<v!LzT88nrfT#n?q*(SkaJ+%S-_;E|?Ope%!Sm><F
zX**Mfs$@>#=^qBhb@vyd{A~tG>P<*S@<i(ro17`7T7%y%{HS4}JP`Z;DcAz5w~{dp
zv6Gjx@jajB#Mxf|M&I&^p6NU7;s4k0ajct1!M#v#*?zaN#>1uA<$fT8O}_@80{jxY
zlcT3Da3d0nyRN?PThMQxIOrMAxL;i?8=-gSPxFFd?#~tddtDEiD%!G!`e~!aj73Hv
z#<d|QT+M)ApxNnWe==xQ_0&Cyyv4KGicTwVMZ6hPZ+vbuo!p|$ub8H+lpL>PyG!Sb
zT+6rjiZRMKu_G22Zj@du8OE2#_8+R^|Kk<TzgG49fAM$!D=V>l$&nJLuXbf{Q&`eT
zxjXipMr~uxp`sV2Ocd$k#D{dvJ`YD~J+H)-e=oYLt>Pl-1e4R8c4}G(I7Kp2A+7iJ
zHsso7zo+KX^!>t?y`s;pAg?8lIZz&K+^BNuPvm5HiC~;HO_;fzlZq3bc-D3h*>T3q
zB<m^DXQ%sh@W&+`-@ZP5kzap`^Y)O2;roA{ZvCTP&jCCEVkP8|oI6@|k3FD=c}Z1O
z1I1*<X78ypztX?GT|}0`_EQ=>J-=tzUH8=e-IIBG2VXWHEg_*XIo7UUQE&w__5H4e
zL*^CRvaW)P8C~ev02OH4tdAgt77zQ3yGC0$vw?3Q>DlN0biJRP>>iKtdTA#Z1iFYc
zKm*KQAdgdewm-Gk#d2&gy{B2!r+5hlp1SVuY)IF$M><U$K;l1eWGs-u_@|#9eP$H7
zAhEWgy?`;r?+D_FqbFj!dM&hSa2f`ce)Fts=L4#C#J(063uoTjE57nXW8m4S=pLs$
zlQ#W7>H_{}9`~=_u7SUbvI`ifa+2t?G4f1$r5@k}J^K5<bWN~fqIE^)s&_fM<{KZ!
zv40a2EdFW{@Pr(l695~0hg0mWLG5}I_9Nj$^@zfHm$cU0(dur-DS4^}T22RfEdW+R
zew{%JTRsrEKFQotY{0mS#~{>OBz@kBR+J2;k3vt4OPA~x(f|~9dDSOTB<n!h2wG2t
zFF1Y94(-w#Hk2MVe53LPI3+Lt?b+?Wy>0`A(~Z*Pn>mPX<Eo(%ITxXi3_qe`g-{(0
z{9&PwRSo4-i^7tfzc{3=lo*aHd^2pSKlaNnSye*6WaN@!No~}Mg?W_4!?D&}F|GDm
z`Kg9kN*#wltA!i?hnwdZUFyh(xQ_wMWRoH?Ouw;ysMvt#yL=GbHSWx&wz6hKyyh=E
z0h6-P+zeuR#ozUjnQ!=9Y6XfF)Uy=j1$e7jEk^Fnb#AI0K^pWdNCKeY_!$m3p<k}6
znW1Ks9-k~}8zAQ*0BO$HcTVt=TE0PvEeK?A^E|GX#zO+PH}gGjL2Vl!dAASttSm>6
z)de0xuj$)1txlli>D#Tpn0<#TG`u7Y1^5{+jDLT1pm6ABc-xKX<2ZL^AcKeU=ZsoW
zO?7i+zVx#HyV?Eg8}`3<bIWR$IpvPN+<0>teOCGk3h#2kINV9HpYC!@(lr~~eaw#|
zNCb?iE5t<(H%f(`HZ&rmltO>B3NQUk2-Ls&T3@?}-9Le?#u3?|0gs0sN6i$b@(t|P
zsY>C`qfh5$_RLxh@$`-H^tno!V>PHe_;ry^S<cdt9sh*d!Hr)^SHnC$*bNr2r<@3n
zT5JcggG)@BLA({WpA^L2ld*%0J!;yyMq}_ueSpgvph^YaMY}l*X$ost1t3j#=isM+
z+a@Rn)9i`@!y|}Itjc%kk9-qEY^1DVY;45KNy@Fewpj%V`|U38B(b4X9V`DFW2t5y
zv;EKS-F`ZZ>@W!KIB}2lE;)pRihp98PhP+P4-g!xC;RYH%$+nb<~~#8Fr2>dHE&yG
z=|$@OA7_gXABs!5yXQP{K7vfzNSDO3#O5|`kK|&G55b)Jg8dLTOpT2#07MXMq^53~
z%m;5RaD8g%T_Lb1t2}prD^U;HuEn0dgmU|SXYdrOjSlohKm6({oTyQB(fy7nU8s!o
zPUn_bH~a|F{H&tFn!*|+iZh+^cmPYo=q4Rs+VyT7AI>`bJou9+y8nfCS#~q4c?Z)c
z&`0LOL~fKVAW+JcN04IY-PeH>-vEW$zW9+n>~0U5VEP4i-eWXCBkTxLm9*^9{r;ZD
zXw?=>MI0@sG?>>p2Tx3bRob*W1WN@diC)`#0c$DUK-o`+u^MbkYk8uCh?pm`SFADk
zf>Yb5c@I?KM~8q*R9FwqM3{Rw=FD)A#QKG(vKHA=Au?`05Vy`iI=#&ezchK364q5s
zuAAe??k4P89zn1s--uW?K>E4f_mAD7d6?i7l*k_QW>Ll!hg=AF;st=1sj81^^d@{f
znqrTggGZGHD3G4kOctA<$()a-9*|DZ)=ig>AWgkqiY;SU2PQ0$6|j`O6|D>prCjC>
z7u5&$avWjnUymT{7~W`UDd^Pjx+0Yx6^WF>hE;r$-OF-C%~+=e`fJ2(!qukte&L=5
z2u6>!qvo;1=UHt7D=H+I**rOE))v^SvWLx>r@MruUNo)U%{_v=4x{=wDQ~DZc-%j%
z$sOD9CWwXrryaazoz7S196?@TwuO%%1_h|Tiv6-Z0f43R?Ga=dW_JV$)P#noGH?bw
z{_|(OISg8ex*G>jL{lBn+av0fwP947U@PhuC`w}ntWbr+u%$6d-)}V8Xo~&_k`_D&
z9A+Isg6d{kk02rp7}_p8D%BFbJpdlrzX%F)Wk71Lu@^=Ts)iq)O9GC+rCEbDjx25p
z*@0>qBigOt>C`E!pJfHeGPMqS=>1=qP9bsi0|GB?&rZIrHFE$pZJk7YG9f`b1;+Pd
z@I@#=Of4Qfyc5+jKun^YDMD?QXmye}T1DKGhYL_m?MU&!(izWG@C1hP3QRlqofn~s
zVvH?gGWXU0GJrDPzG@_yGD^9VD)!5gQd?AV1gU&WwQob-9I&P&N7kXn<-r-E+QKi{
zm(V7k%%H~Qv^wWF-GyrMjWJYuoOHb)V5|p#r*m@BO@shWinYa1;UXm|as3G5q(`-f
zmxGEu*ah@oR!QW2>SNL!xQAgKoXlzX7LN%p`ptwZ?tqOK+aZkLYk2UY96@@pyUDO^
z0tuWcWan8P=U4?~)a_P=UvGEB4ldeJ9+81=F1^sZNnH4fb(B3ebPusb)$pu7L<PC_
zwbHi-FwX$Hha8=uwt=~@CYsrp|68SAq|qoa24@yDc%FJKah`ED?kbQls1Hn|FF2AG
z$44iRAUFkQ6#q=#Y^4`tPzxNk{cP&0{JX>zOdYjvfL~yS4obd#CfuxL0%7zl!N2lA
z#Kbx72y*Ibd7R50I)q*kJNg>dgNrkIMShb+t;S}?(bRI6K3<XX7@zMe&&4F}96`u@
zF0F(J!JjFGPk@4R>wp)|4poV6m3dCTDm-QbAjOMehNJ@5MlZ#=mc9+_ytkrAbML5*
zVx5MmMGi24J};)A^R{hIKH7q>IJ*<C7gTZcG=pFJ5oA3V!&MBA$e<buo0iyUyNTKd
zh^Q}mmg}gPn78ahZR&$x`Z)#+_r9Z_3!AlTA2Z}Uo9B3wN#hyF4*l(ZH>m=>vJV}`
z_M&Hp*+m&k<tXFtR#QJ>RI*ji87o*MoXGIfy~2DsTGa5l-RNhb<^~JH5ov*|$!%5!
zh$w~C?%&Cq_BZtli-=)7VPeY}uTJ?|e<E#l>~{><_Bpo<qr1bbt|qSZRoXvH5RL!1
zo7h>Xq3{?I^*H@6*&gJdl5qa^#x$9{9v3oKf@xS{e8NXgX7qc_=Xd5TQ2rGTio(xp
z6zBy)^2}-zCl)369l{X$4HjwnxrbjWm|xrAx$l(~EA9t5KK5m3c@xE0;i?wDvJn{{
z#|suK-2X`Y<oG`joB4-~{U1>w1hi6IkiSQ;rJz37ymIg>>(t#Lg*&q@yKd*Rra!-q
zn48d6RzA082bWvvjNxhWy?A!{$}7Ng6FE!;?~K8{K63k9q`m5s{f@B+l=$-3(D)m2
zzuc&VW%#H%aL#HDc8*-2{Hu|BS``mxoanmd@Esy?f#t!I`>)F{iL4vPw9D7AOFGj0
z(+gI{f*Ol*#^Hv_Wrh<SH^bx~(&wZVFppV+#lyXGB2{rs`|0(qa~)}T8Ca|R7<Y_w
zZ|HpZE%_aq!$Xxs!AOBRu20hDc-DwAKDR3Os~2~~W~ZQJ*8COh#T!FeLzYHt(>(_L
z?|nsnP8@oRO$Kn(<2I$C7jbi={G`jWKBb3M4L&+oOY-ACZp?Xt|I_);-aY@jQI=z~
zFphd1{3LHXdjPwoW~+4}`>p63&qrsjdLQCS$=>)q!qc*p7ZLIZ(p&CmW1W@jQBIAP
z_7t^+36Zx3Q@2Q`ut7UgMQWR}&jqM5eP;T6)e^UM`1h4=*lU>D5HWK%P#5kNJ1$jF
zWCrfr?lH!9j?3<@BxG7<7rz@c1O;IHZU3fPZ6PLOK|gOVyZ;#-(LwLU%QI83vBmNt
zu%Qs-=93&;g~yQJ$fdNuR4`If{>yiREk3Z981t8CHFIN(9Q4G=996k%!{4QOD0han
zH@xC?Wiap4KOEzacX3Y0V`%fC`#AAnPpp&wPah*w-0#@`>tp;Mq4K|zS9)fkSCD)7
zjR{h&>$UX0ya)!f8xoxD-2bd6RzS5Uss(3Zq~+|`IFXJxRq4CjWflC+rlDn5baNQP
zlmdQ^g^>h}_PWSMS$X4_cElWD(Eb@nuRnqq9YNlNA_UZJRAf$QW1R;Sxlhp?rdo*|
zUZDUB3rfu8XD8FPG$$-99}E^et79*+YUu%HqR)HSOom~73icZn=5ls@Z-{{!>1x=y
zK1<@&rDvtccEh&UIqt#?50$yNY$_@~TQC}+VwTil%?c~xfZ~eB3AS$P+u`c`($z%q
zgKd|hjL!W9ur@MGT7OL9)xL9+B#6%gnB!j?r)H59+mp+KioFOcX9<)+_+!NuswFn^
zeb|ob=gW9Qm6>c{dki&GguNsrrou=^mXNPzP4;XByC7NxPCvNm>IP390q91Yqb~hW
zCtqj}l*xq~ua@1Ziu>-x9dSB-gF6kQhJHs@Qb}&)SMR^!(a4vdkfs{A)R%Y!IZKsy
zdn2f!NYOk7PlJOA+PW32N&TJEEb8J)QGv11Gue)U+YsIx5j52RQn%+x+G+b|PP|`U
zMP0dicv5V0HbA7y6_XsWwVQyw5p}L!F)FvP2r3Ww$o?9J&15Et9Iuq6H?kNV+xA%r
z!+S-O`B>YevSFhHhnJ0!m%h^kmQqls9jZQVc@}(<jY*y!M4ly&>za%JTy<uvdd%yc
z!4E7>73Q`AQ(#lao$Q~ttqmfb9!__wcWrp7kBi`gi|Tqu%%6jnjR(xA1b0@@msane
zD<zuBU3;pI=L77EF>z^adTx{1RK+Q0*}eOY;(6f*G>883=-<5fTbWD~GR2?OH6HLn
z6W}A&TE`nkHKKC_sKx4KJq6u5;?xa9<Wd`kX%!N>+uhj*pOzqx2sp661<O9QnqxH&
z5^87;1QRLXEEXTkk4xnr(2>f&lY&j~_EV=`PEvS;9YbEL{;*FGu{a1@LSJzmVV%iG
zlfj-G-`P9)diYiJVJWf~Vz+!zBkDXykox(dlWFsFVuz0{z=ZdJ>}CC<FR)NeQR7Db
zW12wTAPPJ%*bgaT7wHjwu*QT-(Q1oGL=%UmX>~3`%h8N;H)sT*PY2HV92>8Y36lB~
z-}eQxQ2Q^EX=dfLB!GSNf(5ZXRng-@5k=??aCVS1PxfwiJAyn9@=cI_(S(_uou?Tp
zK1Vp$A$R%@7`a07RQE***@sW2XAQ-tj**yWP7AV=w!s68r~<nH<1w=H<fCZ0Qh4SQ
zjNB|uc$}t&c7alI1$y%ENmNP;Z41n?Jr#pZXsg)Q2Xrn(pAH}W`()%Uc0CRiWN!fU
zA!Qq0=FJkms_%b@UV#_8J^g?xkXPTAS6P%O^4%uMu><s?dc9wutGz-8WMFP9xo=OA
zoZ=rXBzFQf9e?h31uX=5$K-XHV&xfHK_>Gds(L{0B=vdkbh8@W{Pe1)V#vJ$Usx+O
zy4L~zn-w1sadI#5L)5%+TDvm{6f{dyZLV=4GhB}CwW17Z4C&Q5cyWK3&#alK-m+)J
zz4KC&U_h9Go&*C(^Fy)k_m>dKe8>+cXbx)Az*&qdS(v=SeW<^5h?|JlW2&aSzggjl
zUH=W_t{8Bg3XE<?Gbw`eI2?)une-F1JAWKOM)TLP)XV-V(KMg%G5EgIGoWHmhq>?W
z1N>Q+P!yRSb#MV_?cPM$kg?mcPs71Enr){5*TJQ~9s?aI>I8=)65q@ri=+U0SW%#v
zyho#OA6OLL*QCO_My=SGk05t(_<%wzQ5}9#$cPdQR;v=^p@Kh`6Y(0s%N;f21c>e4
zWJVuOreY7mlcwN%x^zGw&%!3zXZjUtsR2i$x{yG9Xh<NL4nR7iHs|z)v^BvPDl&aY
zhn|9tB7_chq7%n~v)eNw1FQ3>rg5tMX)wV0A#OKrWVm9(QIUaq@?!0yt4ENJHFOk#
zQ(z4mvt@}Us)50#YGTA%3m`xV+QqP$&$TDftMw{3;QD95>u`#>5YY{b`kjY_&dYr#
z3QwVz0Gsv!@Y0Jrj?2O$D;fQ{QDA={;A}yM1!*;oHvJ6>@HbAT*IINM%Rvu9!Mn=r
z2-5ZGPP7nwYm<ap47d(1?vvBQJbiKa<;BCXxx*3pCF<p>_L#SOf#zrX9KXT|@UKI9
z+R~5u<{GqV=2(@>cfwBymEwA+qYboIz^&HnEF*hWbgRx69k{pmILR2B*h3jGMtZL^
zAXEF3H$y@nP9SfdIf6K2foF?H5WRbW3;UQY*li&!2y=uQA_IE=sD#yAvtM#Xo<wZm
zq^J)OZnWROk08o4)I^}|He1By7}Frw<!v^q=QUs#u#v8|e~VmKi0-Jv1OfbIM-Y=V
z(;Qlf3$ZXfnqhUM@M*HuSPX#<6@>hynmuNG1SzTrQ{PV#L&M%sVIF7+g)(=)jrV#v
zD($yHaCSmyQiyy;y4nWrj+sFUD_MAP1$5iR(U!aN*UMd985@voaOxGJY33j3))B!}
zY%O!oH1Y?QewSgKD&fEZ#XoBjBlMHxP6*+D#qN|nv}*Ml+_#SYzC7K%z+twl6;{Ji
zf;!vdgS&qOsS{u%z_TPf*Mx{TT}>9L@jGBpt^<2gD&s3KWM0Lfc`aH(F}!WEarhd~
z4)dBOiIKY~OOfxeZ9Y*hDsnCE0DVb6K$)bF2&o*|v^bHonL>5Q??>6Gzc-|RU9@|E
z=FIMn9>UewA3=y}(90brVA0ED=fbueE3g5}a%JCNlOahcnhTci4l5RY>)aMw>6D%g
zXr1)3t2=@SS1rbL5e*_wAq|Ao$)_5idqnb=?9w~?iB+yb+TyEcso<UfBi}wXe_cz5
z7&eK<C)}+>e#Cs<MUPJ7S>CWsA?!EqH6YF$;&}XLCOy>p#ysmjJ|fhLC>u+)9E8=|
zX`yDlk01kv>^&k^rF=d&`T;h}ydhxx_zGj(Jq2HXR_s|Nqhn#`Mt*G?GkXLX3!m)_
zWmc`Mk9j-z(*)gvn<vO~Ut*Wg+Q`lr6G`2j#Y>C*;3p_}g!GD$F+&?UZTi(zJ6H@w
zCX2k?TKnUzP2uPdM)heS^2mh5MPm>8lHMbT@o$MOOnW%=qJ6UIsq8d2=QHYpUB#_W
z$p>?NjO#2&o`y#a$Kr9__<+0l6|Lr4yYP8GZE^yt`v=McX1bTKdViNG;UHrcGhT;>
zg$!$2Smoue;%t9*?qW=dSSpi`>Q=wB+-yO1Jkhe0FIR@B8W1ehO5LaPIcJ{^5a4ko
z1kJv0I-+rI;K+e#@(5Hg$`KQ6dgEEE16_6x?Rqs#Vp(8tC;Gi92TsGVQ@;}J^XRHx
zhq-$_l41Pf6OG;HyDRuyUEu2A!50RmaPDVcGYZ}gI-F23H5>WoDL)UN)wPH>_vGCZ
zQS_d4LKkH#IFo64wI}u;@3wozYggw~U#Z=H4>o9`AEo8IsMhE1I)OvmS6HB2@(i=u
z%fs)e313lj=tBxqMQTY45t{lFrN%9&p^^hOsQq%|RkD0N$87Z-t1mezSKK?VK_46B
zsh|^z&JExs!){BQ|8#O-Bs1ur3$Q=OO55<Yt5=AJJ#sK0ztMLsjlLh~KQdgIb?Vmv
z6fM{8wv!h2jo;NF5n-$1#K0pSwU}6Ar0VSNcv{xnkA1vs<XUZ~6bGfqG*J2S>(XSa
z!`mXshD&$;39R@(`TO6;Ji1DLkj-Jfio0jjF`(i`Z<}$WJPcOEvBs?|5IoZO5OJn;
zpxiJd%&};}G*=1sqq@jr?>pHj<H;uLd+s4i_<@znXLOGqMv=Ftgk|r6FBm(#0Plyq
z5v_yd{F~c~RT}VX1Q4n3qkp=|^<_3t?MeMvOT>>Vg0XH6h}f}@jQi~Qgsf)(n`4(@
zUATe%8AqUs+PZaEj@9~^>hn5!9Gh-4^;&6JMA^b>zTnuZJklQVc8^SsD^{?Gk`Yli
z?RB^;fb_*Y9D>=YhbG5I^vyA7-n0OUA9iWLABPb8#B(u*nb)^nwiHjP@wBv4Vq7+0
zhV!*=E(CnXN7I&$=l0^AE{)T?D{|z(?c;6?>Jz5Rp)Z0<0(;UJQJ@{ql~&9s);H~N
zm<qzfP@jEP*lz!KjD2)a1)Nl(<fm{1Nm0B?{$w<f1>)B%6g=oy`spmKlKs+aYa<35
zgX!2L^WDnI3(%*f?{ccBepofM6fZGq=9cI>p15)l7Se56og?()o<RQ*<h{=8*b3w4
z=yVlo5A?FG8ZkTi3+pWAJx5j&AkA+HdV^)JI5zRbP?{J58S19ulE-A3EW7aqTAOfc
zD3qWelxaaef_&OfTC1WMAf%eH#{&eul3VptIbBrCqsE?I5dC-#)&O>N0><HI^#F^c
zGr%Ml_Ad6sf#&uyihm<I*g#sRAxpghwRH<d;1W5o*oSSyG~)I~$jZ~(oYm01<zY!+
z6nzew-5>#+5JTz%OfRS+KP0vD_A$Uy*oZV-zjFh%h>Vk?xFhbmum;Bc^y)`0*Ydfm
z7f%gOV%C+2wFl2_;070K3u9bSlGN19F##w!_zGNk5KPH1yK&6a5-O^PhAbg(&Tz~W
zoOxe3<W0Z&6<L9GAkno{g@3G9$(*AVbTtEB<o1$4KAm%;SJ~&#LDhOh^Gxje&=Ex2
ziR)H=BflNBZ>*ZCn%)Z(7!HCtpZf~qZh_mhVV7{yiw8##>qibG`@S~zRCIKiJ?jOF
z>`v~=B7h&!(d)(-TEQ2R{!`w!LRt`<^1CG}OHeZJQ^Y5u=`8T_Ho+|Q829xhel31k
zfSE)d=*{gwxn!@8&B)Jj8n_Lzp04!7j)mjBllV|%6*0Q6uwcA-OZ@>E{!!kP?L90`
z!dOMINP-M4Bnu3g$701q(c^P?>)5YM1Di4%X4bHuwGL1+UuOgk>eTNE-Xq2LVyL}{
z3TeqcJFNHpq)hA#X9=GE@?^yp?6IvFaX6aF<UMPE9G&f*!^T2@iUIvUX|Mp1raEG3
zzPiWNnAcQ7PhJ!Cu{vP3xr{sba`ERkVT$*G$QrxAk__G6CfTi!KP9nxAawPv{#B&T
zZv$Wie$M3{Yx{`3x-|p9j31d@n|fRE986G%EJqN5FCTjDB{0wF1Rg;a{T7&)Mo{If
zt&xx4G~)@Z*2^`pSrg|Zd?eHm@_W@A2iY0F2JOKm=$#2oquOgm=R0gGJ{Qo;T|dja
z+@B|Zi%e0uc=#;maA4ls3Z}r^<;<jskNx}n>wkwp_g@H`-|YQhP#95h_?T#WgUu4;
z)%|WC>qW@T?nTeUd5(ECc2>JDq{JRuV=x-7x5`-dODGCj$IE-Uy0If=Xae^2a6GiB
z8_R@f5!4)`wOxK7a>k<e9;|Lf|0glJI!xC^YqXjE>zga;PZ$jii?a3aCgqG9-ihRI
zWqY)BHSYD$aH@>QPEpLiTWFipDbnU1p|@-ouep1|5Z5Lm_nQ0Q?N(8^iQyFC7~H{!
zq1|0nPn1YZ9Am#-?(=9VrdQCNrM9BIz?Fid6b)25QXkV#-@(?oY;q+uOM`76hlW)}
z48ttlZ{%NLyC>5wTci390ki*8f$st#WE@^qs|7~yHEE$}W%6VIDlAN>_BxT}X|~*b
z*^udaeX=s(cV#F!=gdRL3Q5&K6$7G3pIE1{eWEUkMDnGF?)Eofy0Og2->}Q|Y8zGT
z<3^EfPF1I*DD2ylM);g?v3eTDv=Nor11mxvye3v%O;6;mw^&XA0kv6zn-~YMk%G)2
zh!>EwS|qRC+HPaH@EYg2DiDf8IlzMbj+Z2v7&SD=e==a)#{3`@Y3}U<vO>IQptj5C
zVE7TM^s(f_9#{~W4{Rw?pS2td5}-YKZ3)xOC$X1z;B8Ri3k+m5gMqtSv;L!Z^;>gv
z8~YZ`12_l6Jn!R5nN?x4=<60xqhRmob&^IWR9<~hK7wqdi4|GQ1fnz{<94j>OWD_5
zh?tDDCaTn1y*%|4swOHcTGDq!wqW1N<*C0~1;w4JI&4xO`Ph@30vL3hU>wprQ3U*&
zlEtM+$X=@kk&YSdWX82=(&e@kWc%!Bl%MZA-ffH=DAir#Ac<&yK-n6!XLg_>GYw#m
z97?)_+_4ASr?IW5;g3f5+b~r(XzKJs(oPur8j5Lwmw2EhcPZDRok*h{l(uxMvA-WR
zladr!s;b^0K8p(bV_hHGC$?+|qh4NW^0v<g7Pqz@Pd)r}X2Dntu^Hl;Ka|r8uwh3!
zpSl~j6gX>C6~G==3AyvPH$=AM4&UEMqP6!Tb8XS3imibcwQoi4MPj#q2LtAHf<iOX
zZ*~F#lUn!eqV=`G;*ID)M!$+C+A5n?DtbO}uB(zBfZe<!O^4@HZ5g-4mB!J2t9r5f
ztnHKNw8!TWAVlSWy8NgKiNK6Fpw6#|Beq2A9=J_jf8p>1Hc|W=teqRfn=oh$KI-NX
z<OF?Ef#YTfee3!VCFbCD2G9m&s`#39r_ky35d@w01EsuK4+Cp>0?DKUR=MjhB?K>3
zz5$qHABD55DrI8OZHUNJUEQ}4GCUakPqHg;UfjLz01aJ%5!D#a%av@_8=<7m8(w!#
zK}qABvVayigQFg+_yC;J1+D-dxV#4Zp;36*gkon{(vR+~XFJyuUgp(xBvavwh<&hj
zvz!i2<**j7J!u(-{_+uAbtYNp7T6j6pe<1jZGm6w%p?7#Hy!&pO{R)`VFn_l4-791
z+=IKJ9cF2Fn^A5#<4@1fgyF=TRln!#i79S&3!RsGf_xNST{N{J5ll$8;&R5bV=fD(
zRBhTsC-#|j$~fec4RQ(WN$=JdM$GOTzdXfTa90u+OOT9nOBfwWPcM$*oy9gAP1BJ}
z30n0HF1N&Gl@&&}k8MmV@0quKUW^TUk)2m?z4AAx-p>7B&ORo&?TW^09xGy3+MiQA
zg7`izq#QvQRt(c)ymE|wFdDaI9O@=8yTyOqWHmR|F1MYjh<+`53ddB`7qTAZx6TIW
zEF~|k!H9kOdiQ>dqXkgi&puWys|FsE_h$>(6lnn$#Ki|kknx`RR5}K&xomV7xw8#L
zRO7rKGv6Pe<7ox+zA~$$LB_?9!}rdTlhO}j$BR<}ov}XMH9aB3<5=a~lTn}bM$AA!
z$8}6CJsH@7XGQZD%e~1X=0FtzAJ}y<?5k86;sL4>;Y{Y=^$H3D5XIu#Pli1A%sedI
zPD>ud#K0?jE@lm^e_^o#*4}d0tixwJgT;mgeA7msF&aMtB_;e6dZ92emXC$Ps9Dsa
zuh;`SRChkw4|XNtmGcAHVPvcEkY~-w#H6yOx!NvN2lj=u4uvZ~f~5Q{P;l-0rXv8~
zVxF4<vyV6K`X($WMnn=t`utP9`=Xf?8?ik$_=PQL!vzci6>5THrAFT>mW#|Pyp)sy
zF299odSpNPg7mHCLa&1k$uq4e$vqf|D>lnXTV%YS8xajxweu4#eWp{g1}RqTfR-?}
zVd=l~jIDI%kLJK^!{kB0sL3OFOU7ZgFSK5Eqv}v>f<x`>*2m}t)7EBUM<7S;U>>rU
zmu&FazS!c8cYOxo2y#Yl=S=f-RcGruLjHUX=iVnjDs&RLjA{zSSQGa<7*CV)bjqhM
z==T_r0~UIbQoVGoZ6)P4v<jYsp%K4?spRIOyV(^k`o@?oOi~BgCUIo{Tf(29*=wO~
z8D(kt(BK5~EHbo$*vmPDv^-zNyZ>Qci?H?QawwOGmcB*eknx)xkPQ_{I<qQbx-^3#
z^ZC3l|B{C{;gT`G-r@A4<q=RpWZPuec9dBNj<?4c!8+C3_W9gc*s~KZN8^#NG)&Ke
z(h{xN&IlI19IO-fq56UMrz@|(Gq7M@^Mgf*s$Womst*e~Yez(u7Nfo(eItI5MgwZx
zw5=Gy7$_5sjN%XS^3!_$J3VUjYP6xi56PA=zU*T<V!qC40~?tok=FIvxA-PzMEsMh
zzy=JAvp(;o9lYkcbBkx}1V^J&%dB){?jaL4Bg4W3^vz<X9wcxA+kDM1!|mz!(^s6N
z&*YAc7BYB=RBdWsb<&#p7Ma^QWfh?Q)K7+4QGUd!f+kEyP?$VfP@}|FFx;?N^562U
zD3y=bol5(dRB1R@M#8hRd}+_D=re^=BXYR^^3=7z3R;aThI1;U#b;@QMu~oF5^R@{
zpngVqVY%2Z{<>^XUbzo{bBFf4zg9tH<q&VhwT#ao5>ZMz>0xHgW#qSqk<Dgz8@6}M
z{)CYie#j#K<D1#Kda>N2^#6%N>i^I07aMi3CLa3V+1KiHIzPP|TRBW6EMdpDJ5BGB
z{bw2kbciFFV*y-dVe{GUJ{MR&(;T+nnvnD6pqIM(@(?N7H4G!-b?)dyhh2bWhW%iJ
z<>8w)g?ml~230vT0ohima;G$Y%{u{SR`;?<8BNK$$bNk9jf+EcS-rlUVJIhTFc)h~
zy2E&bf?7w%b#A~H({t9!?&Bx@8kygV83%g5^ayM&^zN(9y9704B78BRUF}BOcmEG6
zB-z~&EJ$nW)JM`{W*2MKXC<UQ00~Q6f?zlbM)Vn1*_MCd?tAR`QvGDmI3bvMYqvNq
zFfh7Ie_|q9-^?2NwKz$h44V_W@H|GJj3kZLx<uq`Gw+C!6@yEUAjmAfl@X9GkZ@_*
zUPxYK-qG^gQ&i7K4nvuCBU=dVxVwyI?2PjQK_m5Ec<4o?MSIEi$yO9j7ygTV55|aW
z#WL;2Mtfysr#MgEh_!V2#l9mb1f1=}Ogd3)!kh{l6i35RJcjk#L8YA=+F7i$Av6&C
z38vLk0%NkIp=Qw6q_<SMyLr%ipOYijH^XQbiAmJUea(`dpoZ)$18t;)Z*`<plxGX`
zbO2?s88jw7%`2a1(jeP^7cBk+I@=I2vHi}X@k`5vSXNZiT23qZd%=(HMWpVNZ2aS^
z%9b8YI`cTC9I6P(IXA6HS}foA`f~sHedrIYVLAF;<GFqfU0lo9GlBl8emcfZ<^dzS
zy+mb}u>lS*VC@-`Uaz^dCVZPA2n@qnz_Xp(#O@<VQiAL}Qh5bchHf)4)Q%FOv%j#h
zJoxAbHdWx4r^{~%d}47vpq6AGlAF3*fYiqf!tOUUnV<P8`iVEM7r_kDUG}KcpLw=i
zb`=92crZYtRB2u06X554<4I1{fsRS`7S(LdKVz>|fst>m2_M_TuuLfUAGoGrDswUC
zbGVjv#S1f5m=7n`qZD69k79cc&Vcg0H!cY4$>EIvc&>58e23Eq#y+vJ73?>3tI<sj
zH>aB<PJ{2pfI?$vdVT(cQ7Vp1az8LYDskTX@lAVjP6;WrEp0fyCq{nk{1x^t{S&r+
zunSgX61}HPR$`nadY%Br9|9GGc)gT6+hU1ikQBX&?O8cYWCa2hHx%O;N3~|3><|O7
zG#ZMOgIOyh7WpBbOalOnjLX;SJAz!SSDZ6a4;}gTyPf~hRXY8+_!xJQx+=KgQeLX{
z3q3^gUk0q6`POIeMv3w0m@e6GE%K006clLvrf*Nz;~vTZm5Vb-Jw^);IF_~IOFAEP
zJ|mF#X21v?^xHMuFBU=R#1~>e-ql!YaYwx_>7uUsReZMgS=9fpXNu07O^u32&Lbw!
zC7aC$BIq-6Xn462AQt0KxkweL?TbsNCBY_RqF^GX`|_Tx??n(&ovd15ev`D?g}5r9
zyQe{k*#Zi{)RscLdHAhz$hmyWPt+~AbI~EW2E6!=<iJqd0H@sOJ=J`R1KjU&y)5+z
zk3;4Dc9q5SBZ#Pn3SZOr$pLsYNSaKdnk8PoK=$FsLm6F!nV2Je!C@ChE}+s0q#vl2
zG!3|KzX@^f=e6M7O#(BZAjH#zEg_@7_=Qj}cjB+w(9Ufte?|=&eZdaYL>ldJkh3yv
z0>1=xz?P<FusfK++81q}H}tYp+9Ecf&>PtGX)>>9wg2f&2e7~Wi7mMov=k)I%9b!J
z2$MBVebLebORz8rU?ENB2qRQ<se<}{l3iij=<L1Dkqj%w{dkFy20tX9MN(%-_(53=
zarkZmbRxQgWAivUPPK1rhx5tr^Nmh*f@a@8>KDH4c2U4%9g53kgefq)!6IiEq3|*G
zi}_T5Akk_oM&jL$HE%&+sK^1n!Cv;Z8~~q%|9~$>*%5m%8#;r(sCE-M)uI`kk1-?b
z>U(x%j{<Qk;(9%5x1AaGq1c<UcbBY{YpIlJatpPT&@sJY+77#1T7$Y2ftvtvH+Pkg
z`zo`sK%?S9X&0e?6d56}%9Z8IC2ne*Ln*rlR2)hag1HwAU^nSNm0`;rGl89|=W|;a
znPKcq3equR%qPOUDq33cX4!q}wrK<71z~!{P6<YmJdpH|-x6_%qIX$7g8c04GK9B9
zUtWF{7;aSm?mW7l_W-AELs1G)0V#S!@ydPR2YSVH#sAM~B{9KBn|ObRXo?TdAx<nG
zXo6q7(;685edv#sVC}v|hXUHmS!0X+RV@@g`XECncTrof?g~hGoq?KBECY;7ym;H=
zRN16E`fo9=m<EB_Y_nXwLsowHI(xSsPl2kyE!1^W|9oV}5rkzYK@b*vR((^j$B7#0
zW|-5A;;`K+_c7(F*EOwsQMJXa*-h3G){Wrnrrs4ef+*R-^?Qgv*T4Hs@5*L^Ea{tA
zQzZ~#PCZ{14%9qfH#rZY0jJo0S-uq7x&762k@*hlOMr~m4EBP-XO;A646W`VkboBd
z<nMr%{ILjy3$wr%1yYR}TN4`3>F$7XSaI93&n+B%MEE>1Y6VMtAr#E2Z!kAJ^Wg|G
zm^%i$iu1?-hSBu96{ls80$`@wH^`4Nq@wSfIT`l+`dW5<{?wKEUDl192h>5@`CK=a
zyFyb+7`Z+6!MTuELT(G>I;ABFPQ^%ER+iY2JMKu8&o+_B)H~OKWgD)fP01!W59$wz
z?oB!bx;ajjC~E@=^V-BJMwdzR%m(wQT#^{PLN#!Wh6$nFk|CnSVD+EFe1kf5c!-#e
z6rnKL`-8vPq_qmKak8$LT5R=tNxH0yk)SFsxJgV~$G{}>BX(PCc9@9efX<qm)3o4=
z-k`=98<qkW*~Qp#<_T;O)M=Hn0(M{j*fmn@r0!NS&>V_?p=Q#c7o2WD$1nU|y}iOv
z+vd<!G~4K-Dskp4YW(LBBn%V^lUUB{E~*A#XVB?8dhL=fRHoS93I+Z%HD4%QN%m(J
z!j{2WWG20lC4~zWQv7n42DlySg=oF;IP{-Q*>OYVO<r9(CUoFShHc4KrZ;*Gpg{dg
z(0+dQ-l5Zj)S94a=!fENd0KuW3;ES+E4-IEYAN#$`e(i01S|z4I>UpKLA!xT<96`_
z>K%Z+U{l6CY+()HL#V%KehV*kniLU^CKz2n#3|N7+6#We1{Or9gw9m13Ro0Fz^9S?
zCq_!_Oq%lCAjKJa?+B9Z$Z?=WtpW&0ou9ViVJ0~Pe}z~7^={~Y!;f|~q+MmwxwCmc
z)>gx%W<1zzf7t(n6v;Se?Ff>Gh^5GV^3va5S)Hy<)_jV7Bo*zq8++pTA6?QH>`LUA
z%bz^R(9h6l`Pu{xGM||^DAw8eb|~;8hxS$UW4(RtgV3=)RJ`*y<iQL>gC8!R=j=rT
zUr^7i@!1NgAv%KZ|Es^+Cz0kYv7jog3HB=Z+x?&a6{+X3I}%z(;^rUnio(?|c^crx
z-A9<2BUYYdFK6AVl(KbBuf7u58pxnxEUhJd=L%ciUXcUrqBKgdX$n^Vpxsrw<%eXa
zmKEr^;^(Jx=yHx1b!OboBZOv`Tc^o0{dq?J{$6|kS&uCg;<QB(V-bqFgWkTu_{k4K
zd7US=i^c?$y>2SI<#7&TC!VM0u(FoHxL?x3WN+J#eHoN}eK*P3$|=?)@0@}fBlDy!
zh(MuOjD?l*31i0hQu9uTlMcdYe54Z>ZJIm9di2(XNxV{Rt@{DrcJ!zl0dHm}I3?^%
zsGxr3fET3g-gqO_a2BhQiy3a$R^Mxx(FA^#qkebKX2%v_l~|*&?ZR&yCiyU_ZEk8l
zVZbeZIIm0BX0YCyoC>jn&$7`(m1cJS{Bg2h--+yHLSD%)&t9S9KU?TrB^@8@Eo0CH
zVQqRNMT$bYqtoo7vqX8J4>f&e2j`UeinU(;*IE02u0wI9OKf9GY}|a`tsfkJ#r#pB
zu*dbQW%Bk-@_Lp;WN_#=T8flxKQM)$3iPkL2v+dXHS?=qoO8b#hR4lpypuMd)ht*Y
z9>`SrjfgCH&x5@!Dr5Ovu8niw=@&oGW(k$k<S(>mRv5ktxvldg7xMBvB)#H)Xa@ej
zgY@Rl&dT`haPh!r#C4K)-|~_DB-0!(E28ecXmXh}wjExk*E-EA5QuqH71^f{3E{3+
z>3|c}XeVA6SQGck+-zJY3iVDsmD{?4!PsIlo89iZBvq{4zob&;1NV?#9d*8h?Nmi7
zE=ojV&7tpsi}WKSB@U!0kDNEZxehQ_aMFwzva?ON)_}9)q%1K;Qg8kA+@Q3sE@svV
zY8sS|9TY6qsND$lIKX4&nh;+ivux*VI~;2F;RjF{Ja|RLP{aYJMY7n2W;re+?~hQ0
zXUiMZ@I7xoE6ksxiI9@ke2Ih(jxXhXCC?wXwzR!GoYCt;Ux3Yg8LLQEg1z&4H#DAG
zReStG<0$5rtFeEK^=-{IS%pJYyZzNh3KxT8G)Kt(*P5bnRGr69>B|(SfG5Oo?mjn`
zvg4$MsY?p5bB7{esHZYrZb>G{Ip75)my~K<u<mv8wLlFfLfV7~?UX%R9V@8*sJ1Q8
z>Ww%8e0uj+fp^taiNky9m!E<*x)@i7D(04YnJ=wln(v%}Ax^9yy`43J%CX>%j0Vn9
zI6&On^fD=v?rwDPm}!CM)>LirA6?g?Et0941G~L)@}F*%G^CCWEd<p3ae|03BMPm*
zCNVa18Q#s>BY5$6Y8~5Ve2!as7!v&8y{Z1WA1dYLD8FBa^sr0vE2b<vFfBezoGb^J
z4?V6IDLPq2az@_`cYUT4J9H%>I#1fXF`9^Li>603GAAjG6mGnwUpLV6!Hd8Xtsk;j
zC8#8^KJNU9gM$MG$VxGU!E2LI=n2n+E=21rGICC(z(2h0wy0`s|D52O17`JW)Z=Aa
z%yA7(T^nS(k(87Pd$U;(3!SZAge!z88ChgkpfF%CakJ(2n8A5p=9sxn#z{m<tAwt&
zo~ncE4{xy0X07zsV<bw=0%fzzb!hD`ue`@954z?WtmlV7lW&25@oqj}a~8Z_KvNx5
z4?4rgxrIiPT1pI-Ns7MshDXd*QEHkOCsVHONbAk-hoc-e!+?9krod)9j#MW;6QdNw
z)zV-CoD0memq3*V&a+wpz0CE<XU<<9wg<d5m9dZYuwQ!4UF5L=9kl??Da}ETlgm^G
z0^EFW)yx)zvRt_pzo-(}207qmyYe0Lo0CVk99Nzqqt;?oT~Z%v^G4_4`3Pag1$fUo
zk_@!e2B+mAdyq_Vns=?Az0*h{ECg{lWH4xI1{8+HsA&5yFlm^EzW-;N-!Ykn`H`ZJ
z6VdjgWu!}0-ID9DDNX@&ddU|)?MbH0{t3{zQy24=(LT^kci(+n+`_%0%nE<awdw?P
z{oHoGSr(^7Ub#JQO^v)F(|6$owtIVGH{BiokL+?It-e=FhUHQ1$x5=X(+^bgQa)xz
zX8-(^Ryrv#CnZF_wHYy29Zpd<8JK;))y*Nk=I;j2`RmhOX(z7F(c^Bl1}dJpk><Ut
z=L5>%cv?CZ9vEAxJLyQR*N~<|P%~eEb0v{ChIYSbZp;KO59K6L%S=PlfKDu3xv_65
z*ITVlDlHq8;IGO&_#W>_V5S;t18UO=?{gAoR;Z#$i!TvAV1Nk@;Q1q5{@2>h@6Kap
z(a|GSxIgaY=gY)v$OkMa#}(c6i**ki$1eI^jCfhZ{HG<yS|40C|Kics0x(mYBUK`+
zFBkrUyXw7V$%a2;UIFV^LU9WFcON&-QV}Nxfm0_BJ(L11oC@u>#(##(lz)E%hMTAg
zp|0~Bvwa<cuufc+uDN(+wZg5NE^}(rkv{9^M8np?Oc3;n+%-ZcsDl0VC&Y&K?r)}=
zqY#riQXqNBzQ%uh+nqv*96LP4^!GIEAjmVly6xi!;zf<}H<^xc7`5_F2a6wmEflnv
zT?(Q;Dijku{bNT0E-YK|tKV?2Kr?PJ=N3B7HuQ;zZfwm0?g#?Lik1Ru+{_nY7FhdL
z8`#epunPNKFdMmjsn)Eb&sO`~zMlXtC)K(fN&g&2tDjzUS2+=sxRW#y(4GrMLs=x}
zG&D|gcpXLvRbLf|#9?PX`#Ix)b`s~$%~^^awD?pYTHbF5(JA|Nr(NlA6bQMq6+u7T
zkZc;V)V@i*jPCfb)P>s#)vM|`0CYe{@^R|ZGTr-qW*$-fiJ$>sCmE!bFwW@rHEYP*
z@*w+@I~M?Z<A1BsX5nrbg@>v&p#PpK;ERMcIDTUOZVx|64a~@sn7TYv^k#RORhjmF
z8%pUt7}9*qG)*vzg75!=E}8a1nMD3r(S3Jw#q031$87C+YMd?AsD{{xu|dOBYr>&Z
z5hE5E9peU9)YG)!6$7@lj%d{LQkVTAu$ENr8<M7*oz%U|{4*q{maEPwMt$*=KND3e
znp)oyH~?y~)s$SjDPkx_GVk1(gB5ushJY@0X~A54X4ycjEl(|(kVL+zQI*@fOg8aA
zm)Q*(2J(6X)(vB(x2!J1>vJAp!Tk>YYc}b;`CNJ2!Ji5o5#~>7Iv)eshEq&5_kPO9
z(&eRlJ)|#2V1AjX183)E2u;ZEg%kR5gBLeJi0H!zY$nC6BddJ#O1CR5Rz>IW!D54b
zKw#_c&KKPVkb?c5k?hMN39|-~qJXMQsxLY8Ypw~bgFP!hD*S3aihp%ydW?ckL`MQQ
zYwA0x`vExrd{naxaLuS|+B~&m^?HA*mnQ1m;^q;=3>)U{JDxMDd}iM{gsi(MJS%No
z`h^v_dK*wcwu)wzCR2JDME178ejP`ULL-j8!wt40h#_GIPUc$*)74?G3MsD&si{zy
zJWM8HE+Bt`)rUl!$822<2^5Bxlw-T}cs*vIOa?aVTpYt+PXK-fwV<F~hxWRKygbE}
zSGTK9{n*&>vESPxwq*bGr$04z#QK<AEO9S0*5}^4aVp_2Si`FTsapP<WkW^ZH)ih%
zY^OTiI-HjEN@z2!o8!(G;s1Q;*pApzUu(~s;%^a2nV)`@t$F%5Y`}EO;bU{N*2Vdj
zB-QNC>cmOeKd3?8lJE}o<Ahh7uy#YQ-_EJV3FMU^=cb`wlj<%&zAt!z-MTQButDUr
zs52KIfJ^@UTF50m_|?mlzOiu7xZc<~ikV4)O}gprkv`{f^)MiBD`Hl^(RwuxV3Rd&
zG>0_eW(VUHgr>=%%X=XnmqWYjAiVKJ46jS9%vbGizeSxp%sJ$Y!P~C%seqU16|%h_
zo|J3@hEf-GZAH?hiKq3a9BexYVXLBGcLCPuA_LWl3-Dt@S3&d!4%2IO+(0+Y3Ms*c
zc^akV--0${5Cq?G4kKHDtnHgv_-XxJ=;?7L;CAdvD7C~iBohTDJ*T|bX^mrhKTQc&
z@62su=q@%J08o6Z1i+GYAX1y^@j9<U#cvX-sbR{PjU0!c+r6a?4AupWM^71jfJx)h
zZL#dqbeJbMDoMJDpO(uexh3XR0{|Z09@ko2>rdGd8!u^V_Tmq@=$>`HLW0YsW!y)K
zEXKlB%2%rZ#or&kPpE}W3FU7I{D`+=4Am#7s#r5=Cj2QW`)c;zjl}<=pWlgylv$yW
zY?8S{#}Hl6(szijRh`60?H6_J^8Ui9jYGVYktg9s2*eU!r1z2PPGrfOQ{_56=**Vl
z{k`JniKVED2&@mwOof0?-iq%2#mOgePY9qexH#txzw*2+3t5fQ3^2dxeo^m`*Vm&_
za}ia_=tm$K%2&o&e#1n=oEgVQaC=VrREgP-g1e|IY9icyJ-VUO-hW6PoYedMyxro%
zzP33G;&8A-E3VhbMYQg{C#B}(ZLn&ZI`L09o%?$1;M`)Hwrf=IhiH+Z4I%7II`*oN
zm_ofetMRykk<X}^InULd&aA);3KJNPpJkIVcq7;e>Z7V)oKd|S3p&?`7+3i~?>lAE
zVM`Z7M(Ni5@~Y!}(&AlNU-`*0O#1<U(=PMUlXCo~hYeFdP3YnqdFIzd0C3W3YUHY`
zzR@T6j(&b$)^OWscReE>EVcg18vb*=_1}6e`H-ELK0+tN&2Z23I%3>$&hH3<@G7(y
zxIGxD9?GCp6g4omB~a1h%rWGrz1ynD)7Bgl7PA6^ZQNGZolk=fYTw;^#UK=c7&1|N
z6BahLC>77ZaT;2rVyr5b0X{G<{R6mT{u>$#{pbJNE2h`}TMN!wmkX%QkB2xB(?(XZ
zjhOnMjlR}I{)HCspY($Z$Xm9OA@A#XeH^llIs6MPEy%A{VeB)F0s)po^|$dVCXyXL
zpGJu@@YKdEQiO6by2On_uU@XP-86sigk$%+?m2MT4mqDPC=lsGo+3*a5j#>mZegTe
zUvDui`8n;kE6G>zPMekgvX2jV4uy~Z``!7s?f(CmztcP}RiN+n$8(AMH`^2o-RUt4
zrcQH0qn%Br;!3ik>5oqq!bXg~K7lFkj0u!6M6|tJ=nKikQ@=#mYvdRhW@_!q-X_2P
zQ)3in6+74Bw>r+;9hgiM4Whl~6j)T}LjG)ok$rf;huAN((m9`JK@evfh$8S}(b2Rh
zyttBu(#7C&wQrMY9Ct{UC7KN%r07Jaf{jb^ZTid+C^{oUB!MbBN8$&2OMlMnmwql8
zq6zzA(jTCdkMK6T`w8V{KC`rcR}YqeEJDofITBf;HGs+VR{Gw)luI*9BaL?CNbuY?
z>*QmXn#bDQKPQUsuU$B{T+Mu#3df&U{U~G=-XIQ2q)lO{1GLEdt`Y>w4_#khC0#eW
zxn45YcS_W4qL&L}{|9~52@j{e8@rhNMmAs~^(3iDCD1IT)n`5EtgBBbW66myGDipY
zbbtjBz(1?_a&BXiX-0hkE!fapEcjaY+JZjvbRB7e=fhGtt%xnRYVRv<p^Os2XI;I<
zRAaeOr*c?B(aR*iRmEL1C=~iy8yWmKvKgod4AtSia?OUTf;>3`g3(;OM8Ou74EueS
zG&S$jU1B?ZtuQj5ctKXVc;nr{;57RO4j+ZBskM3wS8{?6qFA<@m2{$=vi2PV6<FWf
z>I{@Fp!{SghU8w%3djLAc5;2;`7hkPcU)83x;2cVpcL6sL~2w}1O%jaVgm$3q$@4b
zO9Z6%L`8a&u7E`8Erb?&k={jG=)L!Z8c2fQ+IPR_eCNHV+`I2N=ic+)zp|42k~M#8
ztvTm-o@YE`7;qgcO@Kth4=C!UFVg&Izy6R-B#_U?$wYq^6y87!<C$vAT}uim@J_$o
zTGxF>hNS3W!iUR>pCw6dRqRFGAq_vEmK<ofh4gQE$;0PaSy|uv{3OXOcO*3aR?@D5
z2(V#B!`iDW+SEDeSnvuXTS<qhm6gR&XI7%#e$Zg2Ng~D%*Rm9@dPLd!WlS3yU2o=U
z8tYqv3<w4lzQBj62Of6mCa^F{7RpD}S!_UPh`{_qm92Z9k6&sh?#Bot25Q$ox@qny
zCbv1rfC6VqB{%&Rz){oUMQCG@1r6*;2g)bM=(_In#&bD4F`|&_R4Qyl(3@JG47|^i
zdkF^i)rQL3jOIMksI?^ZgfP2=@+(E-hF_Z6>B-I;B6Br$qD2w`dk3Z;e@!O<NFZxX
z6y08evyK(xuKX4rH~t}tUMQYum2<gm@$%bHd`qLt%^E9-8|ANjAF*AOY9~AZMs5p;
z#oDumY&-Nn3H?o1R(m6S=)WUj_pfq}|6EtfU)qj3Gj>hAB*c;#%ZFHaID+atVbQoj
zojQE{txva7f0CI+_fWrLJPuaInU7#fK0j{EIElE`3yz#7!%>E16jcsb*v<uSL3O1X
zzk8fAt#Xq8*zuK}gYt7c*g^@W8<mBtaDkigu7&;sp<L>`(nlbP1Yo)TH`%v8YM=bQ
z7m?QkWT47Rc95jFx(Q~@_b`R#SQr~BzNUi+yJ+{T`xN4SXc)7+b;1&wM!fcbh`aZc
zRFc#(Tz`%zkGU?~J4bs1s6DYu<ROno+!Wm(vV~^sP?ea(z?Uug#4^)E3DJAYXaW^c
z0F&hvQJ-nE^?d^RCKybk&kqNdMI81~MzXx}v!KD=WRHDRz75Jdk;ytv_iGqp2HrE5
zpbv=UR4cc5_YoP~mirF%!|b<$Or01Sb_*FyfO+lfxuK>fArnzpsC=KsQuH{9#GF=F
z&3Vmw-y<)hcD_N}%1Ni=wdlBjx9INpPqITHp#Sr8J-b&iX^epg{V>vCJrfLLd2o^#
zDELKp@xwznAIUS_MqbyDv_w%iuE1mE1g}fKD8!E%K2>JKJ6;;8k*Ht5J+KEot8jmD
zNALs6=-W@SIo<JT9m0=}hMnt71bx4>wUeY*)x{N`{8#~gEY;Eh-Mi^?!=i@`)T4bh
zg-bW;-$jM$AaF%UZxwJa*E=IoaIvoM2vHip)9$O4bD=TLwB{~qIKt)dAzmB#s2da@
z@QT}^q_GB^ixLGI@;}3<tD{p#Aj1MB&RiFg75ykcEh$MQr1cbe-JfK)zEh>hhX;Nr
zs|T8AnXKoxJT#sOxP?u;l=0FQ!7P8qrMD*;<s6<0VkibbM7S@(FDIQ;=mA&KNr}rm
z%QHbA*0f>+`ht?j0QE!!{IcuRuHp2WWb~yl_8!P?tM?E-JJ6|GwE(=eI<cpGr)07%
z@&{Z9Pw{J;qys!uhKBz__23Q{`<3bn!FF7)d$m51QM<LM12Hb6{<#VMOW?JPQ}p6E
zViSRrRks3RBaDsD@0_!Id!m0G>C1s@4>GqE7ou8EaEe+;ID4M%_S_SnBU6=%u?)cO
z0c6m#inQdew;q$B1Ns=7kM2a<!I<hQ%#X5PjMxx`O(HT*ClJ(VCJ0!MKrN{oeRW}%
z>G%!M*J@TsRkVI#O{LO%1QCo~;X)#%x;SHglF5E%#QU>xjKhU?!6vn$g!{g7xcBem
zd0Gxf=s6{dc!?W;>rCmV`;jM@dtS%Z4@pp*@QcLlQXxWtqTQOp9gM5Qlxgnd1n7`k
zd`}8RZ7b<3$cy4t1h<MGb4DoKN;(w~BXKnJkTSJE+ae8w3|47W2MKf-i%<Ttb;U%u
zA%qz(k^vg1Gi-)qpMg1Tu-y8XD>B6#mX%gCaK7Vk{lHIkOx_Fu8p#?sJ)(wc;$>4y
zTx&ZHtdxTSJe7e+nmtq{4qDHC`QcQ-eJhf>cSRuFK_k%Q#xqWzUfvo~l`a_Ro5la$
zrH2!JcK}*FvMaCpqMWL5nEv3$2}S5^nNSpjPEPTLqUHRSz&2>R$(pTi&Uo{Y(Swqf
zrpKd}FEl9XP3l?enRZ9~!}io+)|pQW`0@v1U~d%#2kLD3Q-0kj?lHpRG@noz4DVN)
z!~#YpHOAdEmKY`}L1jB!Ljd>+dWXTYYrE;NGvl3--0k-0a4dAvE+R50d(EX!JTv39
zx^z#We}Z^;IuG|Cueg=Te8Ua+Y}QOkWZfJ*gn5R(OpT%ryIi#HaK6BTlwfq8>fXu+
z(Ik=@PO@c-x~UqS_1VuQEG%$hH#)dI-<xX9GIU0(Ot_ofLt;>nsya|lOMcx>Njx^9
zNuXW(I0gD5{!mPd;&e?r=>&4fLyAEI7#%@hHc>MrXj99UJ(<5Qsa5!6_T{G0fm|f?
z@a*2<bG%~Cx&Yi|mP636&3@SonR3u)ux|`f93;w{azuWzrj5ujzsPu-@r==rrm^WK
z{X>HfeE6kYC-|YVZWHHgp)}CdJTlOi+?=#J@?ZW7CE`C-2Rg$oNpDzWGYQmzYe1Lz
zeJjXqPqAyy8TG_@m4K2l4@cy@w4NnRIE%VMWw+s4!RXZq^6EqJI&+kw({48*o5bIn
z8x$vQknY(lsuz0CrDILaDDdq+2igBmL=*qvIVtpm!3#Zp&`*WQEg;;o0=j1e=+tNF
z)t(17*BY9pKW_}mI4gQitLH(^V(dVXD}rM+X*$;>+rWf>YJ#*cXPw$!2VK7c%^6dF
z)uU^v6%e=A>m}*0a9(9@hNUmB&)a0`c`e`jZRyhkYhD%cu0Z6_ey@4ciB+9=ZAE#m
z*ayHQSsE#|@chROvb=OMDmNx~!(YS0FMl;U{6m!Z-~7sNCu5vYO#KjUOS|~$?xX&o
zQ@45eFAe4hGW@Dsh?Q%1Rww;~$7DcVvbx_Sl3JwAa&V<d_ciRvGBPkdSTCQk4`^=>
z1s6qy6G6*UtCx~alin0dU%U}@uTI@}Bj3|Q8%t__n6q|t#hJH}L8+VA3L6gObCwz5
zT>u44GvhXeI7>4Ojb&+AlI}wGP#6&eJ<gRr`;)dV^PT{x_Z7g+TPik^eF{6|)Z6s|
zMJ9Muu==J!)j|X^*@FVQfR<oCC~l(l!Bv4BRO;!pAd{8!!a8(ABLsof_zXCB8sU|7
zSlt>sf0?pq-zTFwW|(LGHp4;`sq+-CyE1?EZIcn-61#V}{<@U!L{Mc(ViDrb1VKa(
z&VL$yob=JiU0=}6ZMj_GJ1GSa+$8JFFtkZ}A7hi0o^qL2fDL5a{uZLOR*jA<J=<&Q
zb$(I3!-o43Se2+@$=`gI?(5aUy>EabmnnAI=6h8B_YXqEnwl8I6a6-;X_mQl9=XD7
zo+Hb=WBF4;Y8!L#y7E*@+bPuL0MJj+Z$n&cnEj&1JkXep$Fy(08ij0Kc+CZI=TP^9
z%)!|`;cW@R(>C7d`gDf{K3usdo|Om@>u_OD@~!Mr7bLV2`-++tu=rMQt&s|C=g`|I
zg#3R*V>u&UH(CB`<<4Qa!hD{|_s)Xt>4D3To@gd(a#~!9yX?os7mSQAn@0!p!^`Uv
ze!L5JS{tvpBfh$QH|BJ!&WCl+&UA@ZFK1euVR9rb`uW{+?4bsWV*hGS{6A$p|0RzJ
zU2%1-5yj$!?L*6V)4~Z>`EEVLM_2>=Mzix;mU~6)j?y<`O~d$MZmg~B**!5i$@$Nc
ztqEZ0prD>-ssC8ekoH$>TcTrrXP&Yp>x@)mqD|k(hEZY@IFJ^*p;5^&d3@wSn>hA4
z<+ap_{+A)8U6o!t$$DT{_{W=r{f&$IYl85}+3<yIKzfqWD~fbTTxUyuE|h5BbJ+-W
zWOZ-hAoq4w*)Rnm1(S{KYcgnxufi^WbQx;53x(eWIuXH^pn%gGS{13BZSa-%ZQv!7
z3Po=*mpkq((-B_E1M0ABG3*eZ?`w_4s03e4ApAJOO5`Fo)g~es27SaviM?L`M(CZX
zig><uD!{<mg&FPJfVgNv)Poj`2$y0aZ=y+yi2324WDMVPV_(|IR=e+4gGJ?32p5UZ
z`ZJ;nHU-es018h?o6|wyX}!vevCQq&)Kb;hr3L$q#kM|T0IXgcV{H-v(G0>#ie({9
z;&_fwdvaYVl7Weh5)GhTFv9@{rFQ>Mvc>CJ(*W^yq=y!?$AqY2t4<=N;dtKGN-r=?
z6t23<6#XR2&hZ~f-GG!tHFMZVDX^hSrlv7>D^K)?=X{?vsq$|;_fk9-8(HlW%%;U?
z9J#3bwJ))Y$k)DmY@FuQzaYhhV;Zt)U?duUxeEpVvQ8ZtdUUCI_g4Kp4g%1M8%H*{
zldj~$EniQaA9~XUCgblG@5A$Ggd&+{G;kCMS4Iov;v?3RP$S#sPw(yZ(oBHr8iiL$
z=N=W^RMcdj9JXRB7pL4TKo5VV=w;L3&1*lD0NUY%j_!l-xA!}8l28Fx9ul=~eiz*d
z*Y8WcMM^m3@zhg+4bB->*?&keYA?(8iF7e>%k%Wm`LQev0rrKGOMNVbA&cyyDTJZ5
z!Ne3+faQJ@iW7~L6>hRwy15P@ZC-5DRx`G{C7;nl|A^Ic&n@urRe)~&Ey5cUfRk-9
z;ntOb2fH4>sNb}I9lC=34#gEjw&;}WYp+#kPL4l09J!p55uI{Zp{*XmG8YNa+Mzah
zc2JhM+V?b(-i#~rF%KS8aaO{ThB00)kUZ2LrsfGhOu@&`4T2KqlQ+tqn91gb6;j$A
z&YXicKB|3amd_k2=Hr7;d~?WK+&2^pIUj+Yk!t4QyClvWTT0Y$j{0TJPF7v&lxHSR
zzt7|hsS;2v`F89b(UQuV_=Vs!$xg6m+Jb>_g>UC`h`n5T<$#mO?O8yk;3pYd+`{gf
z)cmR{XTC;*W6P1vJQg3|t*vpg`8rhr&uqT{VlZ5fo-Cv4J&g%k?%0T4Gx?OYhFz2K
zmH6PWYigYCLk}0#gU7!{um~P(fGK-#y$i_b+X?ksdb+3|)XnOVr0h^qBHOT}=!N)}
zymNWR(|*Ee61#FluZ|F|&mO1~DSKa)_A#ln2uMzu{f%k;yL#t^p0Dq{C|c>!*iKq~
zg?{YFXj<OjCf=cy&-AkEwX*k8Zcv1}yJedLkDH>4yTY}YJ9mC9RQS|6Hac-MQAGGt
z?eS^F%YP5<d<-~77pVTDhV`%fKmSQ>#{b{xfax32YD64K*_36^NzNr|AM<D_{pyCE
zoGuRf2%B01&7AD4_j=0BKX7Ni*KHkh^7$jzt3yECfAMZXvi4KnDEk+UM5d90zF#~o
zGcl%PW%Im-fge0NP!F^2J8~58Gd?bEt7B3k%N|yg%9r5D2d-Vg**{0s|1@8s-+iup
zJ68KF<g3G@@_wVH1)7p;axihLP}|BHWXnN6Lc#Yb&||`|(2@YX8X4j{8FGz<g$P}k
z34;``M8a4)_9n(9s<Nz?5_(v;$%o&t>#qyYoD1$~cD~!py9(VYfl2n3B&4luT^~0f
z^|>^*T2vaTj~kV2H*^q{Q>WEl81SasWM~H}U%iG|3GnqmdEUY~xY|GU;brvSl~7)S
zWCGg_*rM+|@$oS02~@EtzH#Qzw1{D&)i)F8f5?mzdN+^I!Y<|(tT`=xy87N`WL-m?
zpr_w1t>6MO4JAz?R3_qyhlIXJ;LNo=4Rl@}hM&Pr2>flP!-W19_<>H8z0#SQ3bqXL
zwQs0mIcglV2|+K0FDw%<nO^8&A;><K8$Pt$f<z@&+fOH7SIf7xpEymr4q4GFqoZF>
zWv*4L<2&(TiCei_^$b?IVKmkrH0EIO`4#n{wjVm|*!7roc5hP&G=<y|l0V%${PONL
zs?YH~Kq$|unM2&_Yn1?Si#e6$1E!G~`3-=!9R%QD4xF}z;b%(yQ2q8knuZ!Tg^zMc
zL{X`^SA*OhE!wr<Pj(HnzI|m)Rv%wkeTdH-185HCwsX>tS@{<=jDX<jZ&pAJONW+A
zP0=NQzgmMi@}XOR`BFoJ>XOYOdnA8S=G7qq^_J~iXj};<3q2S0u|BYxT3(f{wMzon
zgZJz-<G~L-dqt)gHI0@6{x{b%|HNoo^^cg<&9oudT6<0exWtD~{}0jlPw3^*pUsur
zuj?wQaSk`*;R{#Z&TG(Xq)(`aJe!whdwWY}0hi;f*<aUg^_^?MI-F?u)ibqiMn&a%
zDVwSa>!NCVUJ?+|T$(bOwsF^9zZ)!M5iV97H7D*oKR}JF_r2U(wB*}LNjaXG6V7fi
z;p6Z992m#w|8Vv4&+-jqqfUWmg3@V*Y1NuftcsyG(c7P)PBi{BNUFtCsr-FiB`yD2
z<Bv&N{i#CpbfZUeR<#2ia?%`e1N2X#W!>a)Nw_Fhc%$^@t-V{=nc%D!hVkf1MIFei
z161iMDz_1*=l0fiprx~Lz_%M$wKUv?b&Jnb7g7gv?eVlBidI?>7son1nCc4i&(a*b
zJ2S_T7^!f^&~K+Vu*(a1s~#t>O|q6L?tx`UU-c{(=@sH1YW;9UTE1}O5cy7&)O4Zh
zM0~H`BcFGdNvsJ^XE{c>o#x>9u~RHiC4296*zI?0(Y=sCw<%@in$Y#GllgJaWM`4b
z&F6fq8UT(3fs-hPLk}ftuL+sIp4nUCiSs`_d)}9{0H}ELIDB>Fu25jaEK~)Px>?L+
zM@}i!VL@+k<lOD!OcPl%djbjv@HQY)zR~srL;cAM9=$Y8Ln^Po>8ba)eheV9IZP<G
zk8Qtsrkr&T^|cnziak$@t<R+q#v>jh)GOiftw}6gpb&Xr6$ox9t%#~;F5!UQ|77-I
zj=P+M1shkjdJP@!j+b}N$55Vb=INX&)$n5*E?Jii+dYeHuuE3k9mEP}f3mvD>)fR=
z0KsaEHqiL$VBukR=ocQ3>=Obk_T++>A;L<c@6x{(Xwdn-?kC=#Z|^tP3uh>2Nq)=a
zt-pwj%=O7EhVplAjvh$tGV<)A6!F#o1*io~dk_X4twr89mAlzQ>G&=6;5gNGE>aND
zkf5(+oS9C*1ofL%#-x<BJeY<OqcC0kOFKa`6)W#*9~ZUGvw8W-x-MSl5#@C>p*&k3
z-A`1^^n8b5i;&bTuM4%jl=>vE*MI#?2ddkJ264A5vpXVfZl(W3JY4}n7ydO=KV7Vs
z>h?91W+;D~EU-alo%u&({In*f)a(f9&RJ<`u_~+I*2G9&x5i_Aqc+CT@%}Xa(MAF$
zMZ;O8ujnnJ2wAl!l_o_J5l>eP6?IF8OzVx(O>1)VbyuV+YeZiINzp$|a{g~S{ym;P
z8TvGjX)bcS_SQAl-%?huykZI)E;{y!q<}V^GF3!?+DF#hmJ7~wR_)!V-U=t^iR5+-
zlrN~Ruh=ipy_2yH^}SD#TiZk^6}c3(X5ybc+L|@i3O&0SdUHH|yg*a`YMVe=1$EKC
zDX?n~T;wJcm-?MI2vV+2&hUXx&wuR_o{H>^%Rzo5N?s@Io<GseAxct_Rs^2cDvA#k
zs<1DykLYs-*&!N}KDka<to>4uD#&r!q&EdNxN5uM$mOa84seG$vhFJqb0?Sy>AMBf
zq9vfxJ*nZMDu)lnxy-7y;N9(mky&r8A(De}&}M%}VT3z;DH`BLZOS%m>+36OV6sPH
zz3iw%+L15m1d5Fx`nrlj$S3ki(%N69yvbEaK^~;8_0{wZZ91z+^)f>=5OI6ue2;Yp
zFZr`>-p)R#It0FuTJtbR)(p}HbR7f)xW|3ujE5K|n)--2DUIuHu)%bi+34oL@$LkD
zV2tVA8r!HP-GlZ~lfIt0L%>VNf6chm@X?eeoX_0$a0?@>dw?FMEcN>?61vHzigqZb
z3Mo_5JEhU}bKnTr9P83Fmi2^<9^OcIq-`GSAoA6tpO}1CmR59>3{<uaCi3xMC;hiO
z5l9Gz_lTK0+>T$jR@PaD2^{fgoOb>|7yM9vE4Myv$uInbZ*VP9_#h<ehmI1)ZG#4J
z)N2rD(vvL0$GgnJT<{7^<uJIasQXRzcG~ammHS=vE3fQ!c^=46t>kBxsc4NI)2JQE
zEdGI2sdl;0?@r5M51)Qt@yEjteFo(mXNSVtykBX!FC01Blh)}TLgb0Yq_=K<tkMT`
z6q9h<Z&i*(4I&83cBUg3I{*HcH4o;aGd)*Ujo(9K-)@;WVs{qan~2~TPh1nW$jqLc
z^QMESo}?f`OYa0~267@7h$X}DZ^xW^{D4cEBX3(@^&~NWq*gKyZ*rn00MS5b)P&E?
zeR_RAXW!4|8J4tbVp|b2X3R4cbET2#gs?F49<cF^w!%WajYMlM-O{CMp5a3J`rv#H
z&kl*<SBB)6oGYRQjmCN&e2|aQf*#c(T4*XnR3oK*coj}sfk3+lA8z$9xS6o;w}k%)
zx(IgI`Ni`|v-Nzi;q#(rOqnKD<Txo58pc9&zy`HONXUMs)(OeBoIKQ$+A}+J^d`oR
z)QSUx-BnXJ)TkU>i3p9a49|jnTIWsTY{<booXGZy>>Bb1f7hKi0}D^1zkAZ%=}p_1
zMU*6OaF~(wk6C8-3Z7ox)jdid+c`p@UG-Pm8v&Me_Y3SywE6t%(`>U!y-Ll!oP%#0
zkXOD6c=ViO$heOF_@#WB`TMaBn^!DPP@0f;)^ft1ES&3aeV!wB14%0KWj_p^Nb`dJ
z_7uXpYNY@EIk&56^*RR+{#K-xQrKUvGdpOLLD3o&%qeokv<hA~U4h&-ufEq`WNM_Y
z=IC;>bUUI58kxl^f{_t@+=~QM{o9IhXFK1g*gNj%H5IvaeNfIpNw_Ylk3U}3(i%1R
z-)aT-M{~uDLdO;aZXdBdqQ{00bW>os0VLTimjADm0$3H0cxxopPhYMyCbN0~_P&4I
zboA{+N7X2A7A>nX8@W3*AN?r$aV1yIjoO)J_}MZ?I((k7(UsJ@6JG*Ou5Ta;?Lcp!
z^WxYA_^WLKC%nI<0gmevU)N1q%8N_lf4+JSXjiSe{*T!8|H=PxR;D1jL?%D|weA~f
zloW1fb7vM5oG>7mIY)TDJ^Iq5>5Xc`Lz8Gl*E!ttaCTA_rvv@=Z(Km<p-^ZK-hW<_
z$WT_G1h^8s;7SP6a?mK*V%k$~5gI#chtn+hL|EUej?&PMD$lbLfgX&fSC<E#FwhPO
zPquisi%>#3M%G?f2UnF}>TfkD@ZC?GZqVsVZNZ*iK~D*}ol?##$p@48;Z5}lIOez8
z2@Cfzf@M`YE3Wir4eGF4uud@DDR+L)iQSravTJPxxA`gAUH%1B<V<m76}C`nXFseF
zxhCip{Z0AdTABWaG@O0Wt4u!fCPCulDTZ)J*jn{1=epTu3K19YNz^47ZNOLnYPa~*
zjd^Mavn_62kLoPCwTN_l#0{uqHPFIsqR4%9EP|9V*VXR_?i4-wrV7r4sm<6{7#L;A
z=6eZ@!uu3&q?gsEwwa2T?HYE;s8Gi{_UN%7&R855_?6QgY57QPFz*5ynoNOx!k+rw
z`qL&pq?N6TdAc9Vp~Ncu3xc`l(Omh2ts}U6KXc96`QFM1OcY{9yhSRjy^N#BOmjSS
znj-<g_2T#I+M6p#m}tstCC9kww4k!A>UKz-LxYg$=G{*iW+t18Zoxic-xr&lEeF&3
z3bn8oG0RhJde4)cqt|IwAZMU9eXc-2NAb5<%KzE#05-D(z-F#fbg(_W#(wi#I|07%
zTue`ww=JzKG#sc<#lKu<d3t~6X`w>9c2aA-c9m>yVNAPD;IR^BoHR5Nuk}86qgpb2
zC@fA>(S`MQZJqxQWd;7z4CUoouZ1x1nN_5p+n(^eR6CA!)uJ+#K23#*n*Fx<5XK{f
zNN&H-D>gPY@rfC4(LPnMyv1xKUX_R?%%6{n+DD)bmTo=085~s&Lk||%g!kkyeOwA=
zM<7MNHBi~hF=rq@*E3oRHs^!9H?Q^Ee?8@BpRR%xU(NJ;k<I1!Ffo$l=F7bIh)hVu
z4+qOO+)|YH+JJb(>pVt4^6H}5;wh6^y!cRCP;<1<jC1>VVQZ%0!-2c->E%LAqk(*B
zhhAwd*5VAf+XK>hnZMW7D<&b7%=}Tr@O>h07KQ<HlbC0~I~80o?pFiv2Hr>v2pAT)
zxrY{d34XP(mI!(^#TXh=E9)>tP1lCb$=A)Q4vTjeHf+{(cjlA8_8nOw*ZJy;wJWP`
zxFF|peN?ltImyix)Ia=<<(c3HIYQP9vG$j6-pYs;Z8nu!lOn|T>Jz^D3WcUaoB5<J
zw+)t@9tW9pnC_%z8D|>O9Hb*;z5JTMM#$dHnmy5Xz~Cx$wDCIj<_=S4+WH6RQ2?r2
zGL1?k+Rf~CDn1(L$cIz#B?xjnt4o$v<<C|CmA3<qZJrKX{iC`Cy23uv>|0E=(<~gB
z;tZa4khy3LO#TKVAuL8ooGELXv)X7kd+k|ylq+Jlr{afRXaPvmxxST;ui5s97efgO
zc=V&%Z_yO5o6A$mZON#q^LwLC@>_wjA5C~QEkfE@u3B6+8VNjQU#({z5)`_#6{jw0
z?Ly{G;?GITY?Xh`lr=0Mu$^9!FfW#>Q5_Jd{^DGuqNheAplOl(93OQ4;PWKH(t5ho
z;&df7e*4Pny1D|*j#G=!FQS|)OqXM0ZS!gK*-<TY?*WGXCoAkml7(nj^V!*Jdz3_~
zdP#gE;q3ajR$q=MpX1i;K*P;DF!;Ar{+&?%jS|?co?R88yrCm_{068dyJNT5%Yp4`
zgBOPnvaeO%{Na!s*iRre-CF!}i#j^y5f>$pY8RcKf;3jYGg%$kdv3)0-PHMOedPuO
z31U#Vh}WJ+Co3b@`sLg1R`7k18EJ*EfK4I4;MJ{%eEE8j&kGGV^LxJekntb&AR>2b
z=lSU(Oujp+sXcE%93!a1I-R7H`_!gG*DDB-VY541dmdTndc=Yuyv#6#w=Vc)cGWf%
zNV8Jl3C%yr0R3&W0J(PH>A?L)?Lf{ixyub~ZnQERDzyoi-Dx$$kC7vxy3M@%;x)7I
z;L~e(mt2AEfSg|IYtoi_$A;_tiE>y1`NwRCAgs$Vdkns`Ucd0<iVx({F+ZW7s7c^g
z^mW8qA08GDCC#Yg3P2X|MGX#7+t$M&TM%QLq{HNSV+no|-yz3?F3l>Qw-vlh_)E~}
zg<XTxJ{I-DzVv<TsX3=1;pjRRNO!I`0GlrC{9mx?-&0_!tn<{Z?&)=Tvt}zkNy{i`
zHZLe1{#x6#Vu6?vhgCjneSB>egAA<2&RHD~j?{S#-$g>nJ%_ek%XIXPG^doSC205z
zW^IuyKuwGB4X_#g!+5s;)Bp9KCpqgtofy7j3ql94O?J>-^By4Y13bE|2Yu-Z4AwRS
z=k-ii;upG14yJ5nGRteSntg0e?Q6R~I7Ky8IkNil^vl`%LA~5SiE>&L(XponIqTWu
z)f=+6T{oU1k?EK3Umee@3$WybS@#z0m%s1&s>AmtFa3=2eYM}@PX94^sedv;{o^X8
zBVI>8*dp@^BiVoP<%gI4Lw<ZD#-#f2OY|CVPpN8tj6A9RffRZsI8&xhU)FHckNO0=
z0c(VTFbA~ved*@Um({)XZ7Zt2rKQ3`bz#<eA61H`_7b{d9mqRAt@2A=55kKs2XzrE
z)YQPR2EXhmxKf`meF5TPOb&=Lk?p<R*fE(lY?01VxlK|^*W9Xwt$b|`?#7O`kjmV4
zhKR?wokj?&uh>Uauo!=s*!L@nY(oTMRqN}Xs&#25f_1KKf_G~;#skxDY|tZ7>wNoq
zfbR^>{E!I?kB~IRyCN<@-Z;^ad5M7Zcy2Yk?j));W9(9GHKKv0RJW*JhX#fPVP9D-
zL|tH_Wb6zvD)f}?w0~L_Z|ZZd+NxBbx6FuI4I|Q8@q<_VP@@Sn(cCmG{NXu1PjqhL
z2bp2eae90m8F^G&y&^uBAn3U>r@NT%kyW1iG95m#$yj^euMg>U1x}+svfv57KA)s8
z{W%%#a7AMUNcwW>vb#$PAh=PhKgj?_ZP5TajKlf6Y@Nb3Tela5?)>XXKT6dKNEG4{
zOw6-k&T^ph3zT8veyv!-L`jT!@gb@dMh}}w;t3HHaGkmO0&YIlx$9Q~!XHr%cS<wi
z>)qx~Efc1*n2&}=97DWOyHy8xIRag9nH?d9sIa$T#}^ocT1({_#1J0W`)igu#2VHw
z<sTfsPrt3wrrHa7k`z-t3QYLuzq>%hC#%#LHy8AP_jU%w_UHLa(UUMb!;l*tU*onZ
zD+^f*pRdP0FQW2LCJh>izwR>lh7p(j?8l_n{w(xZB#3rO!vJk!{#3AM6<9R?*LC0h
zhuD?xpfJkG&A-A7F)xm?pTnCiKdc>*+i2oY(|om(RuN-G72E@h&$XW$Go;YFAWtcd
z@^XTv0b<zKF5jnOHluvW-&PAEsJ-wEYu8@gGc41KToPC&kGa1N4fM<f2R+nPcaw%h
zK<J22rv-bN71ab4*Q01H2A1|X7vep*DCCi+17^MZQEA{oOQITU$W<DP`LgOgHKWL<
zPd5iJz%9uz{qI_+{=#|AYF+(tpa>3-2_*~IoQo}h9_!F%j8{Cz?dO;2=61enwXkdm
z+M38xyw4N(gVZUE*PCKWLu$J5B*;fPYtVKnY`$$!McoSU0AVRO+p4Si6+?4E08E-u
znhLICfazY=s=E&|*!ITSvD}%B-FcLaMJ5PU^Nby*7&Bp1(k4_8u^`o4p1=jLQ&-ZM
z*_R`pvC`BubQX4Vt*fL0t(mM*<)0Yx<^Z{%ML+gozTYJDt};-)xDSM4Pk;i(Ut|xE
zBbsV==8sxF)lz1CL04E;C2w~|#Y6k~g<4G5ig1b{Wkc8LTT}KI^<1=^>|`ShPCmkE
ze4PMc(UG8qYEMqK_iPrE+2>CDjRSNBUAxukG$Jt=?uzn|*JrT`=u7eQw6caVb9Wrl
zD}6kKI~Yj=yzNseD(jqq5x31`SXKGndwH}vjtv6Yxhn#QtrK7OkqjrdvzyWfQX3Fj
zUmH?{uOYVg=8)LvgwX&a39o+?3Z7!Q)u5guWogO9*L;riz5c{v3gpvb#DkP7ph50V
z`LTTeyt#9Vg6K3f!FXgH#-&c(%%yzCo;~z+$HF<#1O6Wh%tTk5byL`I?BgLziEn#Q
zt}9Qx$-MPfL2aNhW?(1`=}F6&F)=^Q65k0os)A|c2IL~~B`HbH9P4eBaN3J<EtqrM
z8j4dG-GDS$`fA2Hi3{anR8%2mett6G9=wsBVOY227FWr71Cp+^{^IHQDJwpJTgIfW
z;?eL&GEODZWDsfL4gcCztI&0(gp2exA1y5+DtI~D8tA3<C_miqI7B(%WeLxLmH5?<
z2{n&f$V^_Aib}Iq{eEXpRCr&lWcUu`c6>wS;W4;BwT&{^66C^&T5ixj5DDFd_#?{N
z{ZEAr+nGMkkFFD(51DDm2`T$1gP&w_`g6}~Il0d0{%E}qG$ksGN7#7=y9~tt00>EK
z4u30E@N1s3>}|RkKdjkArRd)(R#nyfRB`Fr&gnc1)p|4{mlOq@-DO(JEzWvme=Q8{
zdlq`&91*x(x6z%eq(iy{@QcujjQZ05lQZ~_Fyt@H?%(~L(M$+<VGTT081qM$=jV>|
zEJOGwp|{iOVn6<dUjb!PcF(qvZzt!f<GV^1mLiz*B{S9R_%>;4GUK<bp0g`UKj1PG
z4Q!{VrEF5Tglk2$%6@VX%!nZODxDObjC`N+(v>KypGM@z=Q`r(2Xs8^`{0+4W7+R+
zMK#r|x3-BRxw~p2et7?_e#d|L;-J>Sw~Vw8oRT=%1-HKUzG~2{u`!8`8Q1LvPP_ea
za9f?>!{PanvM@uu*5ifGlINpUi%$G?4jL%dIB(RVs~buqy|$8_MQ=+VDh#Ll`9af1
zJ6&A(3@08`SvioMEeVaYm`n|jHGE(@7rM^Ch~vg_7I%G(JEf=v&9`sb&N_x%Hl7?a
z{7Lq5Lz}*7o#rF32o##1t?Pa|gmi;G-Y!@+3<FTo5Z12Odz7Y|Ible*$jt~8J2U+1
z)cU5cIQtVXsG{g@-~7sE)IwCi%PGGIbi|z$fz~h6E0bS3ndY@0VLSR&BY-uizp{~X
zQh%$PtV{Qer2|!JV6Db{#b(s3Mkuv>->6bVp43f;JNNGOtrJ~(VIwj4MD(3;1gQap
z0WTeaIKt;<&;f~^K=bo)JN75p0%chJ)fs9O-3EgpyLs1##3;~qD*xup`-Y0s5TrmB
zVnI!5pl~#}ck^0DiSR5v<rE7lMS=gc{ZtljNK;`o<yI#mmrz@xjaIId8wplnQ)=Z_
zQsU5kOVRM4b%}-UO8;Sh8Dg(p1#p+N!P=S?R~a31jKofy?1Sz0?OmSYn9?TKJbpmF
zDR~%nfIvpB`NQ}F?8bsn80^Y)#t8*d>@s-*+WHpr3Cf$G21M37U+{q#OmkZ?EtfW-
z1ahsl2K1~p8WOSs11-O>U!Mc}X>`w<<P)Qc8&^hsm231T;X%`W)`~g<2R<CmZT!75
z+x*=xbYxT*&R&qGHAHnUF*A4HwDJ?NQ(Xn-+rz&Z>i+xC_b<HXV2_{<R2pv{sHT-P
zdEwp;?)phKeNgBg+Ln2HNY7H~sVv8j>w(HQ5~_@4U&Qpjs}fOMo77jacjK7!S$G9F
z3aG1!WF3sIhR22Eq=kM1YQ}%U_4$AMn4JH&-k;ufhMoycHIUtAAyRAWEcnTC!t?;Y
zWs41EgZEapv33A5H|}fC_RDd+_lvs&pZ$Nh0nSE((<5xpt{jLvhqk$8%a5t@UVKtn
z8Hr&|uzo&&Kfzs!OeY?&f5uK1&qtk~dSO$+soqT!&(6&GseoGcs)K+VR36*OF<ka0
zqT?a|rNGK9%Gs5mTG{!U&)x&WiY?6nT<_&O3Jniku;36muCGts0P!Av*n0njN8&qY
zmmQP#e{B(sTcSf+{r7%Rw>}4`$%~7tI?_#a9>E?O#OX!nN^N#7Jy`*&;rTsJ*+wzG
z<9@C69*;&3$U`^RF3{XnQgl^j60gsYv6=JMG!6hd7MzaK$$pZRwptr`+8cgEIAW?7
zLj7kRN~(Wk)b0L!YknjqIdMo=?()ztej9lrTEn8U;+4SOaqr)6{+6EU%CRl1v{T{Q
z=(3ov)He>8mcf)TM!RX+;lgcHVNr5uf!!G0Hgt4#1!>7=r={gb=>a{@GzaJgZ0YsQ
zvwK?~`ef#H+^{K5kOupRZ(%!E9>l!NaijGg57r6aPQ2#?b=ucdw7ZrgR7SAZ!l)T`
za8*@&u(w%!b{`a>p?I-uyW+};=n^CXRP;SyzOT%VlaW7TrrZ!n#qIVcuLFJ_!3e5a
zHEe~h59k-$lJO6tY7yC8u~UY1ur>^HxZLG5&m%P-S{IGL6N;;}H{NxhuF3Ns$OT0e
zADinqA?h+$a6-x>%f4;0H`0S^SA9x|bZwu(&#DKu)2mBU)u6$nA}VaZ+-EvaZiGyv
z8(?b;*QN#ZhF>vLA4ZnkKB1by4-<f=71@?O8v+{y-FaoStMGoLy}B(7K1W!4h2~c&
z*fukZVDHe6ZfDtmlq+Q^ws{)b%DRqglJpmSp^=ie^Zm?^MGryW!1T4!I8l|TXV}wB
z+I5koIem)9bssMUR(lhzvC1t1K6BGKiQ`wk3dDs@gz|Gl9he*(&NPB2YMzW3OsL_K
z{N8O9g?Xkg@?>9gMYXy7B%{Msd5FKF+_60*s4a~w`oSm_(oKG}_>}Ano}M|_VuEN4
z0ki6HVL>5Rbw#wtoexDXmFM*Mcdb&4KOLyd=m1Xx`Cna3U7f|Auv{A8Ez_E$-2y{a
z6G2DA_8DO}0<~B)oQ%W`ONO#a0JbQQKy22h$t<8X*P8eoz@*f!7kF@w^3~^|;q^UB
zW@amg&xgP%{zpT)-wpcy5^(%a0Kt3ubwl>9L;BT~kyFpNw7Fg~tZ717<q^SR$FCf}
zzSzn%5<z<jwMUWTqgp&9JPhocHLMC4CwRKd<t42f0Bx~4p+*_S!=SljegUP`si{YA
z#DY)@8!vdJRa%i$==g2X#`+6`a&*~1Ywy)FA4?xA_jt|1m>LwB(p336J^<9E|C5@H
zJ3B>TOBSaaEJg^6s$wB}=0|(F&z&A$Ji$2UUf6E$8cS4k5DL`50gWP&L-})+gwCXH
zpnMVE$(5Q`G11IbHcI5Xh3$^2xeE>e7ie#@#|YaXFN(=l?!Bi`KtENpHG3uYif#ZM
zz98^P!X`5GHuBuRUs?Xu-ZM4%zQii-L=Z*nHw9_5+a_APn(Ad)w^wuGbmBh_mS>Y=
zJ-#KLp2x2FKdTn$6P{Z1kBe%cs9Dd=pOHw;GMc)h<Q_tOVI=n(iW6^&%h$s}!m<ZZ
zn^9v+W;dul-rehuu<m8md$YOZ?gNx`4C%RpOTH@WM$_rOzCeud%x>))1AY5R#*}K<
zR({$xJn7<NS6jRlzh*L~=zaALj9)M>-;_^Et(xRR{O&!(YsS^7C#k#?)QTG8X$MnV
zq|T0UKveW(1@F9#e&;SoHeah5m(FC0j+^x#(%)+7&ukH*=_qMxrezDyr*ks7j0-!I
z#ag#M7)dA=d|P0D-9cdFvfmT)T+~E~27N4=efr8N5e_s`0cwa=qCqC5O$~rho9G^4
z4&Lr~)t9S&uS`Hor@~BbLWRxwG37m%K`g35ffg_!WyWPlU=(;Yzd&god=fp4I1u#K
zZwCT<uqDuXL%bE^g>qe~8yF6bJe+ixu9}T!xltRE7;~YKBY07NdSRD6Ds+`(cEz4l
z2?VMNr%~F?UnAR=VirBrO2PK8P)IaC0K``!*T?V2C)}DWkdIJ@#LQf5u%%=5y*_kB
zT~aG%yN8Ifb$=34(eDg1ASBL?Y<->uEXiL_jM>G&UqTDrNvY;>j)xUk3O{~q!QR_a
zxzqz&6ja46a$^`)1G0bjy$FwThi%p#9*eH3@)oPG<3esk-wdf%VWZH-hq}RB{Vx1F
zMb_UPS?QFIXc=^HN;Kx=WDE~~<aqkIK-u#v<f2fv-9RA&H@m<8IzRNu$P*v4$|Y~g
zeUwBw>-0YDf_AzLD{t;^Y0QDoCSJc*TYmSs|1rP0{<7}r?K3ucv3BK{&4*p0TfJ6B
zo^cih!lbK7<6_nV(Sj;$!{<gmOhlzBa*i!m`9v5jFA@jrIm23YtgM$p%y*16o4{o~
zQ%~EKo#t<wL#BA$kFGLSE9Bn#D))8hw&sWPF1n?22s}QS)VE=vlTPJ2@%UqB6$cX_
ziTmQm-@kgi*qEmkKd>6Sa>xKncdED~`F8&c9WTBYhf`qNxKeG5tV-PoN`5`dyB6iW
z-$#j05MuUp@74?C36zby#B?$&bd@OmgXnVTJ}hwM#WnlRTX*3ggH?Nh8^z_3(-$ze
z&5iT&?vj1E2t)w?PZh)XkZsA1PM`b3Q|FPI>MHFz6twMuCm$hL@pvb5dDKbEHMkGj
z-4&zj6i623y%EUgG^5R<?kZ)+pIiN7wz5;RxGtY_=V9v1P34dEQ_$+G!5ST%jp;I>
zdRoYLcl~H@y8Z&(yt{vZ6(UR6@1G=!s?*1nM!N3a<nQGPmea`gG#PdfXN!N!@~&`U
zXc-Odt`{Z>)Zj+OvP)75ic_wR6;&P-WP7$&3MiqyQ-Sd^P1U+IZQRkK$wxilQ(dXO
zIOmRq2ifnBR6%2}5_%`bt2yl!+<z5U{|)0Zt$?nap26!coKL1T=wx_C{k|^Z-Un|N
zlPG{6RjiBE)%BH|w5U$!EBk67$ixIV55Xj1?e)BP$ci1St9rp|Zi21pxiaYb@E(Ep
zT>jY$v+sl<74%B6O8p&S-1d$7XWkNNeV1@Ghr)n*Hbhpl|FWySaY=hOh&$+*iZmER
zRDL?@I|rZG_*J7iPk{^6yk(Mgnb^>QuK*%s-uapLl9yX;eQqwtT<8t&Q9ZNCzk_%;
zgI0bI6H?|#n=Jf<G^ZGFKI5`pKYZMadpO$=3Zq{yG1%~S4WWRkRStdp=zd%rPS9tz
z^ie>#aP$`RIm%f(cd{{EJ7qeoYM@$@Q&-;5VGUm@TPQs+hKsf}S$P&z!@e72R>Ltt
zTS%tk6)gC05#!5y=qK7bfmWA)owkA>hWR7?`e(tx^&otFhGNbtYq(@_NubT}+ZeEw
z=0L;ey?WLj5ta!C>gJ4uK4QJ~>eV~lbdiu6cfZsX6(I4~#hC-FKi_`Kiv9V-&@~HZ
zd3s?LYOOKSWy-S%_b#?6$bKr$lp!`gx_u+w>fWjcZ}td>b&i_Vn9?!4#+WG*$uvv$
zIcgQ!ogWl0ua~n5P>=RC6`hlSeD}Y0y!W@p_a8J7MgQtV;GP#A<9)`Yv4_Szm*)!$
ziVal84acU^a=%~^!^GW-xj`g^v}vHdF^K#ZcDR}`V@21{iRUeiZxrHzN+cB^yZV#t
zeK6FlNf;i^%t2Q#qo=JEU&W$2^)$fHIgsN*uK=T722hCuss_T}DQ}?+Y>t?fw~v8G
zo<1hD#$2T?RNQJtqq4fK<NF}dIR9%&|4a0<=Ro`UGLr18C;8b*>oCLnta*MtYL2uy
zjY&@sipa;C0`qO7apnM@-bgE7db?Lg3s?tAZvWwN;AMG8b2(Z~kxU1nR$i$^qe^%5
zX1VkhwvUL5JtZL5`E4*M?iQ6<PJ3$}62}58eBfZ$y!a`%4P6JZZl8jkaO)u_<JYY}
z$wWk43G9AM6;5;IMjtJvHYfFTFfN9%^%7(r?MxG_MX5c#t;i0Ww9p{{75{4{XSiNp
z3-@m<m=bFCHVJccaj~>ysyz~@esTHrb4)yA!JGCWU*im08FYEFn-9BWB4~ZVyz>wl
z89#3#f@knSd7j8?VMsAuX54Y5zWtZSVCYTE6@+QYyHQ!SrEBV2!^P7j?=vQ}6ve6;
z`W2aquLCj7Igcwi6T{APxH~^OBx)T*u6pZ#Z(sqNK`sy<3HW?zOJy5AJ1bC0eJ)~s
zp%A*d&Og+>_ryw9f8yo36j%U}H<nTvF2Jti#jmuoXblu}ev-9;DYr9E#|JI!<soKv
zzvS*(^gZ!FNon0bM}J|_U7TQ^YwMoWno?eI*l#@1k!c|P;cnx}J}MC=dfIqG8P2J$
z+z4<iO(C<fwnsqOyK@avV*%rShO(0`^SVjwo;ya4%-7Kcx4k}iY~pF{Hi05CETQ}x
z&`}XEMw1)jdm9EyQ-N&<VK?Y>I3oOpu|9|SSpC*+&+Cz!wyZ5F916HM&VI#h{Cke`
zJWXIGUxA^UXdFency${=VJPnjjUVEwc&Fi{8dRkB9Ev8?xkRb;F;US^!M;fve0lJs
zmm*!VYz_+6EhYAeQYV>}9Q#PGCn}R{U=tPHKglrCX$@khRHx(9{CJ0MbL3C5R>kyL
zzIIW2PrpX?3QqJ9xM?S<o)(WbKL-FGc8{1#uL`enSM5`$BoQfP=kD}zIp~`9Y8<EM
z9UnkMx3Mn@*;tXxw(f9Dc3Qt`m(B+zy1JXbFE)RhrGL4$_t$EEfBHFfH7r*k%xwzf
zR+#ktP<Hf4$!oi=)hYg0$s_=!jw?Fre#x^dvx-g*ndpC!<bdWFy~xukv<iJ%o2%G3
zuKd7u6WG)5HT{+H;-8-C??2{~7UT-rsYVSAPF}oSP!j>0W<Wb{?a!AA*4}rLy!`eT
z;)dtky4u%Ewj6h2?m(+P-ag8cWL+vF58R_m8xa<7iA||VAlk`nL|*pGvRGFL>^bF*
zF1c|}e!SGOzA{;M$xL4KnWnK*=7m?m&aZ%_>9;ikiJ<vO*4{F+4L4}}ljD^ZIBh&A
z<lkWFb<E!fOO1A-<n-!ku)Whcre5J{yLmKw<8LIF^*ldomKUF+*heK5oKoVe>B1>>
zV@{ON)_Vr~lHBA+8q$T$vySa-s1RlkJ;{V8rR|CIZ3@ro54IS|Mu!UQGPbY1L>Ra+
z{#nQ6KP2z*mvSNhgwFs;PYLR;d)Fw3oSs*}KrwP<Jf5dpu5o%;q>PUUyS9{dM7mi)
zrn*FpZ2nb;5&CO9m|{m?1>DkC*`Onzk*^vIEH?YX@e<|2IoWdd3`}l(&Ba+GFFyVO
zjVzyD^9<c-I%TVBC@qJ6($pZw`)!2auj82YaPK|t+T9gBe)YJLoIKUZ1SQ7nMBU0q
z8PUd~@qjNYDs6OCqQR=DJaU|)wfxKWQ+2Zi9&zq{j^P56v?768w{#QFcj~NcEis4S
z+C0y`Y)r+*qrqR>Z*~6=yS?eeDF=ulSZzgYtB1oS4<m(3ENLqyt-qqeU{bKg!EwKA
z9qgmPdAURaIpU)aImA^Hq7sXvFpVBsY0{Y5U@m}k>^Xh%qj5&-qF~wck*M_PRmt5k
zyDKFIMlG$Z(mMVhUc=DnSx|_=b-Z={&C|C4f|^}g&{Hr%><D?29^;z{3IH0u84I_2
z(S9W>KNyw%i~~9=6ags#Y?c^mD?WT1FAU8<lzpRS<U~UHUkF{b^nIwYUYYS#%Rs7I
zhXjK304+Z(ar4ecTi14&hT#UFXTG$gN-+ovn>~N)b=NwTqIa{TcE{#SKQ#caFJ6?d
z&6N`rDN(wA&QB4>4SP9K?R%M2;H!^KeO{1V9jp6egnYA<{n(t8bZl}RyUZjM+{J;|
zm2cP@8SGC@g|yH|<j$jwfi@$Oa$4=+9NpHLIdHh1s%y<+m=?Ln^R91?<-hHt)Yc&W
zW<_YMt})FMc8hA!?Ofde8{R>WC&0vB`c+-~FRN_mnq^#W1d5StGO*hIIiBN0!f+i;
zrwpaiWL9NpskNU#td+z<^<g1jrKNyrRe=>|#ZvKE@$h_YQ*4RGL;a1OI3wdv)gJ#C
zHq+liivO(d{w4zDXJ|{3p9z)|_cAwkIaK?kPt5~PTrf&(1BcQR!6g-u-zPhLq`PY9
zrUH%te-2++9D>Nh$q`5#?{vVAPqZ<8bZhnY#l^n9TdKN3jTM@!ub3tDD*0*Y2hywF
ztlz%6vU4iFsiUcT;kI)8#hMzRf>vRj)#nXU9B-I*h}Qz}=M@4I-$81W{EE~lVp_by
zGVMMz&~zD`?~de~%?({6w;9}iR@y&{Cfk}Ob-U=Vrqv#+bWSC6Cp0QD>V*FRaQ>xN
zUKDFbr6rC0g>21xEQtn8b>D7MCXk8;N(9SflVLhu`RnLmP2;;QPdOYOzO$z4dN%i#
zayq6)Hu<Vx`wkb4ACzyS*>{<e$Xa6>`o3O(##o~Ku1!8OtM6eu9h*7fo0IMaQbom>
z1=zHIH?RAjzG9nL;+osXE8dlTRQmMF<LM|tlSQ{(s@2<GXILy>-L7*>2syAk)*-+a
zdFhg(wN#a2#-@0Ewj{Q+)nocDHifx5pww_msH;cm9AfLtc=bs@60O2eF|VT|)naT@
z4paObf@MuX;Klc3=Tb2N^bn`AnDUgWz&fWDAPiw};SaZ?8^;r{kGN`!%HA;W1H^Be
zNybf-RDMd{tjBZ<8k>}kaPD!$o5K3<qj$#b+k1>_&dZ%gmEA_cdHYIfVzL!3rK{yl
z3k`F^_{oFXl!nuRH4!M1%P|z<L0SpTMuXOa+c(S#xV0;|NQHr=WuHxs?$?JNsn;ZR
zZHrdjdfQb5)`S$p<nS(YGa&}jgT8=3OOD}z`bHN>Gx7kZ-_|h7rTwG#d+x05gm#Ip
z;Hz6d$&3-(U&s+{4e$=)z3x4+qT_dm+TPFtP+xXBq3?a$qAqrt1o8$)P3qzK3Eofz
zwu@Q#ak?!Zx_I-Nu%#HVp0l$-{REb$rH|HEbzq?$dzw$X|5UVyqfov)6^dJV-tR_7
zH=|;iJJxIN!P49;vt^lo`U|eSYF6G%K>Pj>kZLxUrhhb(DYy4VgxD{qbS6)iv)Y$#
zX>i!D@m(MP9HbPrql|-$d?J7qeRZ+ghr~dOAT``Uv5Lnzb*0JIF2fnzIK2KSAewY5
ze!f>i{-JV%zh>D73pMg>COn1Xhh>(VPoG%I3dR_;o&Z?%<NDQ(={WS(&_G%!|4QGz
zKis?j%BzUZJDEddou~<K@b)ET>jKqZ%@Y)DX(c}QPv(a(ilq2YYUjrcqE+rZr>kRj
zD@?^~^LI42jd8R)nW!6xWDz4jc0NuDy<O=1&w-$P(mauu6JMuK{suXtelcwRQ01~N
zqi1wN>ZzItZYPNMv&*Zv76X`*JGT8ocO|5)v{3G$<CO>G*AM5bs<^#AmKIydt7MHB
zy)K3@RK@0;u<ElO^o&tG5UMcfi_!nMG?ns`tjt9{t%mDD`<#+}#joYqKUaqu17cxc
z_0yFDpz8XE!$MzW?3gA#x*zz(JcsGu0lp|;Xw-90*pSgWv+#J6p^fb&3OoJCL8HO0
zqR90D*F#Io8@1_&0eDD(@1sF`c*(ZmFIY^KlzA$8gm??<w(^$9VVeyx>Gpb^F0d@|
zz#6YS@J6hbYgV4^8_uXbV_cxJ@<Yc;Zs!(4b=Bst?7V+%NB$T7oPX+hvwLOq8HKuy
zOGnMS4-?++K%O29Xc}q**p%URE;~P$(&$T_6IRttA+b?M{beJ`qacHCb^TW8ZlUk~
zm8hqwH$?kazkA(p`3_8`oHse!*9`N$Or0M*&b+AY@2{o#n2xz0cD6GUzlIhCWV_7t
z_ZXICL*zqHHk!L{SA`WlE84&zoY%1x!FRUA2v&y9--5^$9MPuR&m#K0ds7D$tvx-S
zA$0of1q)w7Z5P9@L{$5S_9HQQi$I#}|6}jF!<yW)t+Ap4B0{8tO7GI8B_aYM0s=zl
zh;$+#y$6*hEeHq*QlxhR(!2C7ozP3@olruAfZyvmcg~qPbI#m5bLW=tx$_543UAt5
z_HXUI*IN5mT4-uTx&JQV8#HqwB^A(s@09NwdaqiRCEr<34>39+Z)LO?u3^0xlU>L?
za-t@8dcO{Y*>3W_h8G@A8$8$)hsmq^z0cmAS(ze%cR~d*=jK{pV<cp~M4NArP0RXx
zmV06#xRzN8&uyCY3hkn^F!uIP1iZHm!jP$8L&k8`W&=m5F*qqctSy1mU`SFltHmor
zzsuB#dh+vw7L|EKN?k9G`q(A<7#u$DyXvdEPw!!W;Uz$=tAEbp(4b_OjWdMjr-xMf
zI^4eU%^5wdtDQ(aZ;>uj8s2(I_*g>^EZ^#?>}2^ZpM?>%-G^HlYBdE}moM~=hY~i^
zxC9ARl+v^A(jJRleDJQ~nG`~t(W>4seC84yjLV%}JSvOwVve)39?*K!-0bzxQ0L?f
zj^!BY3PpF{8S8pkWl(bO;<$0>@<b_|sF{vDtV1tX&0rwRsp`P6pl$M*U=CH+eK<f`
zAPwK5Pz9n~QA)iS_=jQbA9^kkM2QhzEOR_@OPQbpN?Q9QT0@!+nKVHIG}bSice*1#
zIa&k530lWyx=dzMJ!?^=j?OeS&<|?P3M;VA|Mk^SHHP;DCD3qbOybcX+BcDAs3iMl
zd3dAlBg$>ZW#RRbKjlV$bql3L&s?({ai1#nMcK8v2;jp736z|pfI3kN=^OCPVKAFa
zRztQDOM27xu}l%`q`8EAxovA%DjRi1RC`q1&;jXnS3@O?Sv*!Js;Y;`Wu<AwYUyLf
z#OjSZ+T*Q7(pP*nR985#&P!hhsE9*3Lho-oL|Ap->5zQQbidn+YrNm1dvTXqU1zMz
zwc+{WV{Cpf-_?+d6*lghpi(-HE1R-cqzDQgeaLEiXa%FO-ZIs~!S5FC(QmGfYF1^b
zz9L{4DuI6F=(O22TV(+nc%VOKSo}fABYQ8Znv}0v)t_Gw`23^ckj4>bDjU5zC~|UX
zOHx9}<im%DQn<Vh>jl5Jj;3wppou}|Ko%+|VyPBq84wG`%51eDfK=9rlIB89x3&ZI
zK0O!jeA6KrKd=YgrE)*t4YFT1X1f?t#j$&TrwFQ+Sx?cqa`SJUsQ<aD$FKf+|Izm`
z#+5fcaV@@IF_;<g(D&7*>*$+I6r%Kt=PjHR<h5<)kZpmY9=|Dq0>V2jrz{@9HM~k6
zlnT<{xOg9NIvklv$--8}C|K&3Qgc=%n1RDE=uP|TKn;U=1UH3!C+T$WV^lrY4Gm^%
z=Cja2fTg5fTXinf4+zYvK(PF|r@;Qa#>YoN_NL?^I`c~yA^zFm$c)|~)82dsV(rXg
zRpexN7hHI!7XXSeErz6ERQ3wfv4>0uB`bg6JNNSdb^Q^0uhp?+Sg;^>XjVTkGJP>R
zcX49zkO3^n*<?&^uNmWXbO^c5%d=P-n_Lk%7o|)iMv+UJB|PRSglj|B`yO?>BVOky
zUN#`UejP)$T7of5{Z2pyF@;>f>y%sSbO?DpJy*8XbKll@>I62s9a8_fH>B?B$KdLZ
znnNWICqnO!DLlFS<-8Meg*Y=<0M49wptCAx-{0t5yzc>U*CjQf!Ci}+OgZ|{gX9Aw
zz%lyRR3`@0=V~DfCDID_h4G68$($_NoJ9?ml{+6|*QN3>Q}>#<xVfoS&|Dann5<{g
zTbEkgmm3+!U0Rba*=W4j{-`>rd`{co`bzHUm)?tYD&U>zC|y)13qY<6-`w>Ao9Dhp
zQ{V4v;VL*$89w4_Iwd|GnI^^+mPHHb^M)aNc7thsx<f+}GCwiu91_S7JoTp!iVwmw
z(Z`@akAL;&{pcIxJy^9TKhfW;)-03wzTgc22+w``NAT#6*Kq23Wb>Gi(moq{s%9x{
zMhMNc!pyL%?<!6ySwOGkX}nhV84a2o6y6HgV|)gXa1Bgfi1HSSh8zjH2i<G%mu^ol
zIQQ31Z~suU;qiPv0S`!SUaPFUov$)rC#b6eLzf?Z{XFBQQ~tc1!`QWj@IIkns;)b#
zTQKmvL0P~p1!sCc2m2{SuuFT-hLWs)cq3m)cLK1US76TXF36oX-pfNpEie}sn-2|I
z6toY9fGNx-Jb;w+*Cg!sb{v;^69j`1Jx&@aEHH7+&u|^=4uA6b)9G!(vlZJujFfB`
zwxCieQjNjst|xC3>Uk-tk27bcEBUWKA_1p2SpE0_1F@n1=ym>oX*;tSi^3_>_b(!c
zTngZODsdf%1K`2{I_*z8n6%&LW*KekKgC@44|{6P&iZm+6Y%>BXydW{PT(2W98LK;
zsA^MqO+A|?>9S8SVxBX9DaO*9Sxs*ut(R;vq$xS~sitpi)!|!iPr*zDH!p#vD_57P
zu^>i&PgNsy4?kbYH{T*r)kFZHw%}rCg)mFLBX!2I;=8^K5Hbxo2rpt7aHPxibx>ca
z+lS2p&&Ac7IMU~?1vp3kY;^q>mP&+#Z_Y6Inm~O#?N7q-JQySMp%<Hj3aC-j0wb#H
ziDUE^JbFMY&jg7zwY8t$iK1h!hO1PG6;ZisWqF7fRY7ipxjipqppCTOn1=J1Vchtb
zqHKP$gM)&zBWo~@0S`9=>QsrSdFE>{Y-BT-c2H6~y(@EVMAcB|p~(AK8RON&D2!;E
z%IG6&$fkJx>&f0TEMF6sqp;^u`+Qe%KK^a#{yo<-!QTl?kFfZ#*WCh~72eZECo5RM
z2e}bPeryk8xS{LC)Ux=a=N;(m7EalB!nGB4-t|xh7nqun{%NMg>kPGN=xY;jhy)G(
z<?udilkiD&n`&wF*EIo+)YwN!BUOC<0F_Hpey3O`K((**^0V<V3nFsz6KJ}l9pRa0
zdk<@r^@hd6^ug5LF*!_U5;Yr3mFsI6TNsN<pgrE_ysb$|J!KTVBn(Ukg#8;Ad)qe;
zR=8H#7Zg&pnM@-@sG;ib;QsXq#dJu@_alz{?<Y-h5P3kP3g$#`((Ri%iUX2w?)5*c
zZl1ZhKu5#=i6jAHV6hX4?>bqV@i$30$8=F=ZmlN)UuJVflzwQxfCeaT8m9_}Y^F+Z
za{3{VpxFiq#=haBJ(eH5%{UKWIaIro@(+cye|pdLtVHnJ&6^RH@$wW_+OY_NbJ+!o
zKu50zGOr}J)umhmO(f(h3Ez7?=^iwVVelagEP>^VET%m)zn`A^JfkLa4!W^N*fc|~
zw`Q*|bgazN7;l$&T5@K6chxy^THcAJ#@3Fi5#aD@X8n!4b=gyzR>6%XCHnc@vsIk7
z(dma#7i#Tv2D3L7yC2606!mK1ycWM@N&x<D`8E^dEmqaW_fj%a5*~fduD;YF4=fj4
z*MU{*cQtxv|BmM`E%>qMiC+ZWU%prNVNpKnbknjQO5F(|B3IaQ5M^LUl>E@{bYkTT
z^u+g=U!)EpxJolV=Y|OQGoY(dgCcOHf{rK||8-fXJxdW3qr0Scb&Gk=Tom(%g>k7%
zJx=l<!a9GOk9XMfFPBJU1vUm4KCz0A-fex^Ux??*5DY0Z5gJ)=RF&2GXZ)7G=Xd{h
z+bcc~Khp%(9}9WEm?-|*b~mq{0=F)=W7Tw$?N}t);}=3!eX`-MdEyq<66TYN1|1|U
zP|8iap2|;)lBDh4%ggMcnkG`sHB=wqu}P^b4MG!Mm$~h)N#Bs2%hZ@i7tk07DvLcI
z@#r1GK#p5s8&wJ$&i|ET^vm+qF&ngYBgl2Y;)79yEb*ly8ZQd#tgtw+Fua+866eD_
z9uuH~XF4iUd#w7MASd=jD>2E>lMf?08x~x5eM%fQu2(efKVIISeMdBdrrvPPzjp3K
zv_;%u)C&7#?MBEQ<BEH;9wGa}M`ki@u?Y1pdGh0N#Z+9xE-=h$T9l~?Z7UAns3)z)
zhqY*#xiV`%4^MRx)0BAwp*iB3MQ&!Jd;O<1lkz28edNv)6elHuz7v?!d(&Vcp-F{2
z8EGi@@>k2pLb#ce82<L~>bbBhT<94}S4O=ZhBPeAR)Vl!V>}$AJh*S;UvJ+uzst+t
z<iTe7oxoY=5jX)3lR>Ykwk`^bl^Vdz;01t!ZF7z^<VF!ak*Z<2z+xWjP361Gi~C6F
z20CC^VrnYPJs<!Xaj00E*tmN2!N@L!Z8sE{s&&5J7Itz~LW!yE#&-hS618hqsyqFB
z$B)E}4d0tj)1hCMBcmXejEmma_W^Oc;MyHYY320fAU$Ng3XX7kUz5<s(`)WK!CM+D
zLkpUC(VfDt!$vh;(5(C;ynVV@|Hn)<9tE4kkfW7OLA`z#pgE^$3V@Y=h(Sa{V}oRg
zykMKir}AhuACM7*NLq*MDJj9+c;a8DzVL7sZqvP*1EZ%-Gq<K5?|&+EqQf=ZAi;dP
zntQPLh!>EZD<A*kp*5Dd<5E|lVRl{@@|rX$nGXj(XI@zy_ar`IawO-FvQO8%J6!s7
z%dNm>S*NmsZE3XjRSs{fT-WusA=HaIVVi<%>P59qXj#H*tX>mA9Th{oyv23*p#!<9
z`h}^NUWCMd1eTRD=HEkyzkQA~F+W$=!rc*79`QDRPOZE7I{{~<9zZFVq_>_Ay=X16
zqyWqeMuf=G7BN*|v!XULW+fWoA_Jb5W+)odFYj(O$u-*=+Dok$>=eXU-pF@wjmG2;
zwt~{nMF<+`?j+4UHN3iM!=MmySX@RQKvd#5ziLqaC(5Yauvo&4rBb2L$`Wx-_LLf*
zwBLTce?9)0fX7+7=nKsaq)hCY&g)karF5i?&&P)HWN6FnAAl0{Q3f?5n#sv+A_jcn
zClWOKmd~pmH$FY48&)6T<>lV*3e11LS3q|8;fa#Pq^^Qhi-}t1NJ;^~xVCuiD&IKs
zHjx7bc$?UI5l2i2danAE(Y+)<x@7Te@Z_6CYdu4Vv4}Zq5X3DVBn8IZ2+VKDP;&N1
z%lMSMp*SMfmi4%M+?}~CMK;|35b>Gwb3$j+%^;o0-yv;roLMxa2I3avil<qTb6v5b
zk{)T~6Tf?rqvWfGN!~ONGYANOt0Ec=)2!=nqA!1!eUsGvL_3Qw&$?7Xo2OtFv2^+3
z2m%Ptc-Xe9cZdUVV2%s+eSOsaz88e4p>1KGzdYj5d<Ng!+~OSS!@=2Ue`z-8NMhD-
zGZwAlziFGk9sw>*;TV@3(9w$NS4sDPw0VzeHff5A_r9Uhp5*5p_h5W?G#iHtZZRMm
z#*^oCWWq!QwI;J~a%Vjikq9|Ju5gP;Fwb+deYRArF^CAu{FXWkL{3v-x3WE5!2`hb
zy``4~4;|1vl&MEpo?TM(YI662l^|P5FX2!+JG`Z~ITkUyWBOkvhEnMmvsPeR^{<o4
zMPkseZPRhhB#<X}4NFNY+J+L})YGDQ30fA*n@tb170bsBm%@-;dutk-Dr8$-_423k
zBk!Ny5Cz5;#JgxlwmDee>{KMTQIiT)F^nh~ECy#w<bFyCVyF5e9{E{$z$@r$5>iM@
z;nei>z*trLYrGZLLn}ZvStFSK)2xYzifY}l7c-95VXFG{{>7ltD+_X`CtLu9^T&X-
z6hRdSIUskJxd(g<!T;nRlKmtZuYx8e!HLNArG&eLC^=D12YSicy4KGV2H+|4$1P(H
zZz}OOSudL>3!~atGD@En_)AnP#M}%KDEve!jys1AyXKnOP^1Sd5?Q1*PH33Ac_6%c
zKw;Npe(Awee-^4Mp9nbpMrw)0F7K7<8FAp*mC`G-WuLFQr=iMC_#2J4Mux>22A!WJ
ziDUuss{23F-T8-_$4aK!l*9s3t3yvNhzWhvhAc)P`!qOeM9Cd!KPSPXie+Im#fa`0
zimz6NSCxYvRh5^jMJiQ=ZQuXmjxtryCu>o%&1UGKlW6TfS!CMDK)viwB?b}h<P2Df
zM9`JU2$$HwasUU5cPUktl2R|e6S#2fFXbxC#d~!G9tDj(x2#J$7NdHqnd|Dj;5CrB
zW>r%qH6&#`b-FE_lS^347KP<dRr{#Wrlsfo#WIu}`4lis+v8mb{Z;Gq*CgX#;(n^o
zRd)wqiN_F!_EViw%)ula{bjxroXM^_%gH&Zqu)K+EDKmW=kW#D^y>RyvYc+wkA5>W
z76=+sA?d;^UwB+ct(-66#jpLd#WNKJRE05M6S{qNSYV|L2G~%@FTq5EUY&dt<Fs$x
z^%`c|Dd?17zja)ch37Qnk(WKUlQ+%dDC+?c4ArIeKQ1iFEQoQQ0bHI(p8+n<UN-vI
zPU6J0_GvTMs==i<cNW_Y`=Cm)@*_E*jsBi65<r|KrGAiROESqPFQd-`f@o(?O2sRy
zDP3K4XWdS}cVs>3`GM-nMpmv?<-vWd#ayCDn7<n2inR;L^}7wSOsh|q`mIfSx3H0u
zn*A-vlDCdCGLb1|Hj<*ZBaFP}PYSqrVcMUbGl)=aBe)8LZ$bQGD@rQ?{e{ktwgtiO
z{Vjso*5i@_Ja4&qq%|y0v9~?mB+SNhy~h#F)4R0LSa+^r6u4Cr`llfn?_z^&W|+_m
zL~^+QmGF16vbMI?89Hy#4>vdG>0c3CX=&KoGnAfDB=6VD022E9Ba#XZDY{d+`Z9QJ
zT1G^W2MVgg!SUke?;UBb517Ly2#KfiAiM=F#Y-Yh@m!4go&m%OTsqjV)1RwODy;cF
z-lWWOEc~uqnx0ufST8bkrVw5<XR%dkdqIJ<^jW40)@^C_tXrG(YeSAT9sQ*+H<t8+
zlIU{&1SCc98xr_v&>mkfq`2M(en&Sl*RsDM>-+?0pQ|)-<{U8Q1p6YO*&XNcq$jnP
zamnyHz_5p?&toQ~uEZpQ#;_iY(RCMa!DvSRqs$4sRCaIXW0kMBJ-NbUHVv%-7F4ZJ
z?=`8!@c9qwiZ!;8$B)1_XJ6{pDS3V;z#y^W81KfYh|^Y9REZK!+ZyHADWy#)y=`#5
z(>Of50jln=KE(LwL=hBi;%VfM?YB}%n@@{DeiK}>e9m8~qzrXhz(QUKO=))N6a`@$
z*MMjxw-C+RZlsyJD>?_+JU373gL=L1C8xk0d#GqGXrmzY*&(d4ZWqUKY~rej{wQmd
zt-u&inqO&BVO@A#itd1Ip#D;C1ucKawq_8vBt`@1c;!@qXU-)UVVpxw#@se{&HE_S
zjl)?3Swcr1#@hhJi>4tOC%C;AAjkN}^h-G{T6gXm^#zR&I-#qLD;69{PlW?zO~vi9
z@^1lRX&3hO+y3g6D)%{t+eX(Mw8}eO*qnq1wAS8DBS}h+vyq%YUlrx(*lH!)$(~Xf
z{D*~-zv#eZvDLkqQk3*TeA!cC#h!J)e{W%j>*{i|Pxuw{sfxVZ&`0R?6BFMs=@us$
z5;5E4k6GaZ`mh@#pYX>M$GIb)ZE9q>W64N@*W5?sYe5slLQHiT%GMD|b}UUc;bXm(
zSt}`35oK??feY&odxF2@l&XJfo9H<vf$&nKfZCW5mZxugVv8^C&DhonRFHQj1Yy=&
z8CZ7kSXE2@9$U+It!rS!%oOtB4sUw)+9_pG^>BVm{(fLg2(f1O$c}+*nbv@7FVjna
zkf+#3aPh)}XG8G?8vBKsz5d?wXxl?bzVWYjm6>oam@rH!K+U3Ayz%cSPMrNexa}ny
zYvZbryx{Ql>B>!+2N5V9#;-}8RqkBwPU$MvZ{I&fP$nNDSunO=f+1wCRXl29aa7dh
zk7Jja;sK+#S3^Q~3+yqW5pe&?iC}FW6Rz9uykPwy*(gmjlDi=8O|{mqT&`VIX#Gz`
z9e5=r10&2}GtFsWviu1mFTCB38CZMsXZ*xli}PI7ruD5Ro9r*1yCY_|uFFwiqeyPe
z>-<zz)&hv69It+JLB4_{yi5XtiExH!Y%mL=+=%)^*wnP#r^f+hv;|z%DOE3VL8CfF
z`S+CtbcHLU9=*(`yb`pU^ue39-h`1H&lBOX38oz0*I*8OPHk#vH-vgEP-3I;8mivT
zbRN1{QkaG*vA^9xOAIo4C+pdOS;n&T)i^D$o!)rbUuql;b3${FdLH(`F->*1Tb-NC
zX~2<&$Mz?xJKq9GE2cBZH74+3QQ(WG!v>r<hy#7@q)2vhF?@GqcFUA;gM6U=-YFR#
zzLVLqPT?hRbk`QUxXJwD&XIXOTo(t-A4k3g(^&Nw7mO-eOuSW~eyGhqNz>f#)FabW
z^<MPIbg~Tkvd}S_wc$K;WZ9E3+71NCpVRSEuu7*-%2{8pTzppiLfMedT<UohI5A+;
z8iav-37e>RTvi1^kRQh1UQ?Wz3PgT(l=4u)@W$Rq&d`^4vt;k=AUA>TA#C5x?~2#;
zQWNopzBr*ix<O7aL(5>LVlbt+uPW<G`01~&^MAYE9am*3&g5sVN9fM6bY&vlXUeM(
ztLNm#KrK(p4%Kuz>O?tVK?MxIn#i{op;A8*@wHFa=4$^mFm#XQvNT{JuOtPuG&%n%
zrqO@+Wq%1$e+pLrcKh*lJ7HXEu2pd6ST_c(Fsz_;qU6~-Yslz&I*+#Ahbpp-nK&%T
zusX}7CNuh3Qa#e_OEhr-hPFb?r$k$)S<6O9CUohWHm@|tzwQOCXhNgoF2>qst`=>U
zmn*#`O*dDPj4U#@U!G1y%28?Mn`Q@R9o^uyU9Nd<;t;llh%yM`Dq;bdm{{SwPr+dW
zRaQ%$5WpHzepO%C?Jg2cFLZhh2y5^9eh=dOPdxYd+<07HsuSQk2M!{5o%hM6F!N-W
z^Tq&=VV>)e#yIEDLfP}rg5HHZkgA!(Vc{AHEAKi8%}`42#M4tF`+`;(SDP7}cECJP
zUThRfNXxI{tgTf>FN7HrosuOTSs$sZ`0JL;@8$yr$v-<P&+LRK0O%5R`YrX8ED*9>
z^_^fwxUXM+#Om<s!}voy({=A&r|$%H7J!88xyrY#`okVaA=QJp+NKEBvpCc0ucUJ-
z3^WG2Ba0M2-tpUQV?O`6tqpIQa8C&<@h$wiW>Au~dWC&k(Z}xumrgQMy=R^-_H9z)
zE+579rNopb+3gmzXbf7%q&<SCq}_OTHT>pxf>%LtK?qB5>SwB3vMWK5E6xNf{ED)-
zfHK{~e|JrZ>(BEcP7bK}!Re)}V?;dDAz!Rt)q-jYYL_wJmpVdODn$r~JAd|PI6$e8
zv0EI-*}pL-k+of8U?hntlTgvG&AqBMO?^c?%4igoshnoT=h)EFX7?V5n-BjaA^6jo
z!6^~!SX~}LebG4FxiFhOjWFC5HcOHbJTDQdv1K9>-m#=Bm*(HA9yw`N>m0P>9Rk0D
zK}4dbz~h~4<t5X^q?hh~I8@gH>yTB+!dN|8K=Svis%!c;S?hc@P>5Gu*H$u+n(O)s
z0BZS*#v4#G{86ubc(FZL2r#+>aJ#9EG=qN)(Y*PF!yIFf>ZHMHxcuCvQ_Mo#lt`n>
z@k8;*b%XB&vhG@dn^n|^#*WjEW?ZiM9?*@m(#o>e4DJ8Mh=E}}FaZq{a|}h99X#;?
zhobLz7wy(x$HRq}a(ZYls%Z=@+*t$g@S2WkVv+d|knG|-KrCs`D-GJUZQULfe1tjg
zqLcYLU7t_Yxc(TdUN(uqt=R(uLjszus1Q2TKTEmbxtR-|A`$`1$_IClvCe=l_WZZ?
z+44=lc~U=n-7$l&<4eBz$L+jVSMMKjCBA}AdtU}z{0Tjum$E4gF_qUzA5U;hz-}GI
zo9`gIY#BVoFqBRD+9li9EF)L<z7r5V^yLcEz~)Iws8tW}F;#!Ah!|NCtYi)?ltw0h
zbrs}9U-E)o8Disknz~c}v;>~3?nC$y{gQ8NXoM?u1wJ+aU7i8#877*bJ6#i&@Vg4a
z3H)+1<!qu{tEh8l>!x=|o5QZ+9}iHf?qD-X@Q~K<3Lfgybr00uv{REdLKcF}GN5NJ
z4gOBRC+D^GV}MA^)>Qh~mgA7BtY^{Mza6{3uingAq}dsfIRwxHN%A4h7bRKRxW@c!
zntmSWc}`DO;X-qCPn9V9zfBdPX-(Bls09U@qCqcszKqXI(?e)b{;pe;jy7~2(2wHw
znHVNhW-K!h`UKwm6-eXcD0p7JB>(hiaON3H%I5+}>GRKgFHDq)YwEYwI^1a_qdYGC
zg#^o8PS%*7(D-^=JBH!*8_Bot+r8YO63k&7T}8%91?_Xs>)0IFRQLHNVNg+m&<V<u
zL=sgqvmdSi3bIbE{9%nD6Sgmb<*E7aURqSO>IABXHatZ=M8eNb&g6K8mwA(`O+|ef
zCnl9ioTS-xt2<-4J)h+il~C8r*yG_W_I!51VKuG|M~WuBu@)DK7k^Vy#=hJWF)sng
zWUC-o4zILImZqfj>$HsCJNt^D5kWvvqN;!1?*-2M6bvxG9gaqXf4z?a1GqQ~|8ONM
zB?xo1Fy()^g7yEy=V$!{j?;p})m4>5O7GdnAp@LS8&Y-Ox^J5Hu4+~JofHHcv$Qnj
z-M|nhe(`{&I2FKJ87>!NG^6Y+ZxejZKJVO{?=slDAMTrtm0WM|Cf^UE)tj6cS17a4
z%hH@JJOGVC6Yx?vHtU^jwaV()Zq(z22OgZxcv#;d!6@^&QS-RXOpd$La%!dfHvq31
z-#<9(Zi*r4@nV4EUUTRTFFyUjl<GGct-GJZkNwwxJ{rQiNThNW`cZA?>!z+k2Usm^
z`sRL{x?$>5KN*hmCJdK^4=NeS&8Ptx;n<?96dLQja6-yPAKll+=cS`RvW!-0sVQ?I
zCB~Ep$^j;&#F>-ObUgh@vZVG7?5Vb6H7ZHP{ybKlzH@y=@X-6parJ!phToiR@U@OC
z%bkK?Bs+^q`5EK)-wDbH)eQ+e)o^U+FyE?3-E|QGBKu}}$K3bsr@@^S!R+X6i6#xj
z*zL>kgj1cQBy=HFO+jW!aKAM(Jjj#sB)f(UV>L}hvnbS~AG1^zTX-4EmhYkR%!5Jt
z0NI=?(G?0_<mEB*@Ob-G?coG*<`B(M?=0VDPd3g>!i%!?krUS$KgI2D7Ct8_?x-ca
z?A+8N%H~DVIH_lc0Mj>Pxo;7FFWt(umeG3;axW{7^dfHk0p>K4AvMKN*3D-AuI3^<
zrTjRN3-jpxvYbtVtR>Ij<S~1vV9<O0L14)%=Kj+<``cU2)1=%kvwLiw-c!Q&g92q0
zbYztB5skfS{usHDCU3{#syU5ukqZjyYg}e+%_LNpA>EK#1E<GaI$y>C|7Q{qg(%r1
z@5;o>Y*>Kxf0vVeY8?75rNCWP6|Ywj;UCyL?lK&NxuDtS+kI`pfRV<!TH;2+*ZWCx
z22sPrLW@A#Jw=zEWTf;$-93qkKIpA^zPImOZ6C#9w^VJ<A}A%x3o}cCK1_C2-le|V
zW8~V~PzJu+pFX(#nqW7A?}Cyux8$>u;aG<<=np?JNLw_khXTaC9xfsov@$|9aDSbK
zJ0SN0K~ihQcLIRHLShYOT-~O*lei=k6LSPL*T1&3s5kL$(+j{`KRY`84uW>Y{Vzn&
zzdcv}nU8<;&qT8WDU-Hca)s2DTBgo)tkFo8nH18ZPj_-aQv&{Km)eDcvn_6wE85(d
z8mQ&e`nW21S$z!w4{}JAYtaHRWo4k;-u@Mi<j8+k;W}=irWvwGlqI23swjJtzeXFq
zDbX;lJ7XnyX`YVrzzqCK%~hr4$Tn1{w0@9v@3th!$low4hmlP3oF(R*jEH)9lmPus
z&&kLkHVx;pb2>P;g?O=<kP$5tYXWJ8@p&gsT^X1**N~-V_>`0n%DyE_q5iu~m(+BM
zCl?7rY6w4h-WymfM3s!%&V@Tio6@%RR|Vlx=e;+`y;ItFvu73ROL#tO?|)Ms3G^N_
z1|}m7eF|?!*L`F~OSUm3iQLGn9^uofk5PH*_My9)G}N#LQG1jWMDw#@9(SpPSe{Ml
zU5f%H_L=Za9RpI#bnPs?8rKPtT3n;+#LG)<M&R{6)LL6h*em!oT#J#A`*dQq`>gd4
z&ll|mda&&yoIC({jQ~=EgVa(exvP(C>pb_lC0}<D_fycx<p@j}G_h%Iv?P1T{chJB
zJal!}yunz&Q~AR2BZGaYM*XCRc?07Fx(#Yk(#z0<rQQKKQOV6QId>v%)KvoI0Z4P4
zr^}K4f+#kEpo{q`mBZ+KGO|VNwnai}I5oZUm5K&NKOk2nf=jqo&oLz1iUenOU>{Df
zHk$FAGcSOS)aV^I#fxwr>}GB(;*oRkgnFjpbIl-29jd7AYJk`G241UbDDFuP%!%T_
zaHzHE8qIB*-_l@(FfR|=^?AWoJ^(Z@Qrf(?O}D~d|Ip4}_+a{>JEu4~Bx8<GDk*g=
z$tj?@zePSVq#yvGmoL!+WQM<A`mKVnyd@@YdPHMNQ9enc>rJLYdHoDKK39$2eMFY}
zdH&?{N-M57EqD9X=R@NX54S(XY<1QLwmi<GeJ9N@+&Gl$BxFp)b><+CfxMV++(q76
z*3R1oh$@M_-}<q~N#C}<6C}tFvrNMKoh4Yqp}A)cgehYB<VQ5{S7BkWK$GJEK`$lw
zpi<whmQ6`Sfs*~m9k{(pVa|2e*w6Q6+OmZMWUc9h?`~&<2V~bnt=49Z;lul9vui@x
zrk~HKE2UI1QMH-KPD}z9deJ`wu76IZDjc~5R}(GI^ti6Kjhnx6T&%2*<Xyge<z(rw
zE#1P^ym<u7J&;mg_+Ej+CwxlfC5#9y-y{LdUE@#NT+nYiEwCGP*F^eAGMdgNT&EZB
za8-?EtQK|9_P}o)gVJkyK3E%-%`7yB+e&<V{yK794RFHzDWCk6Xj_!~3+J{H!`Ngh
zkCvAOjC9^Q`Tc|Rl>fxu`1{h6|3%*qgnThy;}`Y*MjPRDojT@x{WTsqZ=N+US{lss
z;C#!Xc_M6GCon%HLYnD8=uIl(r{)Z?`All614q{f?pkP%QwR`0d}?zJRSPbv<8eKf
z!`Z*9St#vpdn<1@IW$d=xsVxH@<EtLbcWamBC~4HB3_hT)Wcq}W!e9?h6ifHZ|}4M
zT1x?I2=sH%(SXKy$8EsnaVRB6FiMQ0VXPpnlk-uJg(od$0!v;sU@)(>^B~ExcrmWw
zbk~EOUe?nucjd6^YBCQZ*_tJN(wu}M_YD`et|oO}bYo(UF(l6I*g9^^s>o--U?S9)
z{A3n<IB!_;^6+g`;bev9#7w{*U!eOTz8BGb7cR`FPL#SixBxfx6vO4CZT+f-0EI!e
zl2y|g^i)*1ca=1<4u0u$+<+bzKqo@;O^FuzMNJPa3jP#_o0(_aXagLpT<mMgkXfrf
zQPqA9Z58xf?P_xV_g1<QwScD+a!8Onj1|2h-mGrfv3YlUZ+oRTu4$g%O<!3)Wi9sD
znz(<mF#@pgQWoXPZf6=&U*F@Jg>D7|-^7|BAx99VbK~A_y6v?57S&#u)$#INSSykU
zuQ@m{FF(X2NVccj-k`^Lk5uQ>wYNg$PP^Hnn&i-tg7w};Z{3%r68_GPas8y#$!+%3
zbPy(><PQKo!H?$^ZJl^>{X;wBrh)OBN`4<+*cuFfRn>Z$O7ro3+I-PuZeDQIt$?gd
zD7|^MirEa8iMQj(L1HCID?pobK0R0(YU-rPSFJIN&6K@eX!OorDjW9xzowFywOD}w
zu}qS-e%`8)rEnwT-P4ha&)i8CW2r<A<mg#NW5*`HfrIrHNAm5QhOx*LjK*c2p`~gg
z-&1FkHD_VqRm?swxC*^JYE2a+Bl*qkTVrLNTjPy(QYl#vuW-=&T6IryhD2@}IKeR0
za+Mg^O8uRt7uo@0wvRO)sfHG`m|?|K5Cg7-Yh{dM!MDe`U$yp`Bz&8VI|w5z%Q&Sk
z7OyZY-=M`F&ezw%N-PBm_MvolCIy4+vO*;kI*$=Ye=&&uA~FEbQLLPBWfYhLBepR6
z@Tfs3FdgUCpvoKo{3}TS)cK#L#{U2E9#>qBOrV|ifUyXu+d!+!^cK`WLeyA4L0+m#
zPfksCk}D(1MMb_S{!^MAg=S)UN^PP^7$4;tmq4v+kCFkKgwl^W4u46&&9T~aVTf@Y
z;Y7MJ8DFpbGKJurqWHi-xZ|ybNuNRRMLtnrDdyE&=)R=&)jI%#^vsYy|1yTZIXqhI
zc`7wHqk@s#)2a_IjMfser0og+z&vp7^;Iv}rSTB2C1~@${HOlMZ(O&M5f<^1YV4*)
zlQdp7e^nd(3*5~~b?wU?DH-A*phoJ8w>o);XT<1Cd*muQ^&OxjPRDHO*GW^)eF5#b
zv}}*nw;B@Krt8Z!AcYnetZoDua5KcdcnLd5%w41LD#*9MmXGus=-zFT6(>~3(VURq
zn)hsDDdh|sA2@|p?R#gm^%{TH1M}{NRr9VN$fKVZQJFF`v?!B!BsCb0&$mJ@E%}dV
z*Ov<ke}gwUwp}89DKCc10E@yK#rc92*9?ebj%igF(@MQ#S;0P^bZ{PRGsKdLJH478
zRWKK0gyX}@Bqwi#92vENj=vI#wBwpcH<;gloW72W0eD<!E+e~G4P_^%rLxmnmBD&n
zNJ*v-oNu)8Md_stCpigSm8xeE(e%y;bgkZ-8U`*5XY1+m<0{2mu0S4}Yl~)norQqw
zl+unOb_3<fSgefkHJoraJDaxV+;Q!IEf=~Zb{GK4+JknQ_A)Mb;FAnv_dhV)49O=}
zE8adeUvUDp<0WuKbKXHZk3~-zhHqDC6COoNn4Wi{C6$C4xFTEDRdWNfRyBh|qE~YT
zH#mkuwXz$g_<@7f{9hBf{(wI67iY&W+W-gHmqsJxFh*r@qA`8ygdpw0>mg`won#dT
z?WyugJ1zI}OP-KtA3B$DVFg@$H6)mi3YgX+aKaTK@@%lCvS**fF%+(<?E|_Hlgt#Z
zrNG&LsBK6zOkBPP?j`$hhPS`o_VIjNL=cz#{UG+AC#mUz0Wx7&VEyE2Y3QHb?|pkp
z%L%#u6QNWl@Atp%?=xfmwQUf=Y#lrP_P{B-&P4u_QH9^<Vz%j4H&A&pj#hDJ+UIJF
z@!2L~#lu$(#)!=6D$YeXtL{3~;96anbWCSGyK}jBV_al#)s5q*%o6@Ln8IkfS@HNF
zq=EIZ(8@Ib=SWrY-VZ{tRHs|>62V6ySfuGt34LY6Re49Gax{yg!iPr*rE>HXdFfRC
zRN(!nNfRJG{^COVTLQ?<y27$y9)t1SOIy|pfD~RY8?-CmmNaUpj5((VGq`LzxP4WB
zABz@&Y@9JPRPFI_({w^~z&u{iD;un*9>`ziB9x(K(CM#FQpW@UwaO7sqFjYcn5Mv(
z-mO+<1Ja5d05r<w&PhGuI=OucBdQn`=peEd107*7L~{^-FuSF|C~VvW$~jNOPOGnC
z9Nyze0(gn?00}^)d3~SEx(-;8x8($lJWi=kv1btL+RHl8rHK*`Q-t^%#NcDUol3AE
zec%HzwTwKdgUN&X*M{X@b;<N);q&jcc;(2PF5xr}U+l#l0o@-Ix1?-ln@7e}n;Ss&
z?4UJ)qD!#5<K6vAw6dIZ0(TlN)-uhpZ4#ot6ZjmZIaoq>m5iznM+C~$H1<0d;GjkG
zQ~EiK=V(ZFmI~bP^N9ppUXu`#EjNS#?{1?`FCWdwy4p4*jSn_@gYxJ6-RAob%&1^)
zAFY;06HB)qHM|!9SiQFbWDI+88}*=QK+fbi$y|M!&|8!IHa##vibEgBSo8?*{f3(G
z=eqDS_PrX3Jr&lgCnNdmN;X6x7s0w7C@xutr^#314tNFZB+{QX-CR&ZF~YKY0m!!K
zT?Mo0HSmWsGw8yjmWP_Y0*I}|EH+nGKuk^g{j;Xg!S#OvX3HqPu=tiyC)+M}Vq%%E
z;#fnDSNE1aPglu^lXUc5&}%2O=!-m#@{?3;2Lq*-V{Sz7C=@aGBF4gOjW<BiE?;+I
z;o`K{4S+Z#Ei|ST5M5Wh&Kb1f&p@^(;PApTQZkcwF|1Y~h%56sfNC<%$lOtlcqy?I
z1Y9`UFq#10t+{^Y-2yH8>2$kxkbpDvykk)h$pUlhI=6MyT`|HTD5?35?foK*53J*c
z9=_{LYm2QORAPrx7Q+y0>*%CyqI-Yzp~-suE)2DPcRWQcg5$WL9sFKZ_Qj87{ZIS<
zOfDd+z#Tyzt>(r8rE$C$W4hz;Ji7j^;~)+Qr~eCByea>^J(x3>Chxd`g0oko8zLr+
zs-gg|)rpi6KLvG{2Nl^iFY_`)fQe<!X$8fq1JCQ!?8+X$9PXadQ82%yK*?~Di%QoK
zmQr?;ZT1KD-5cPz{)%{O_L^tT??cF46L}0AciW~DnO6k?XS;q?_H0XL$|9258XB*>
zm(97vWb7ISug(mlYZ52Ff0LOES~6s?mXvjw^V27bpIAO&y<0ybnjtX~))1`k4pfLS
zuvh+!YoM^N*@$x<Lk`FIXwGrHZHttk(RcseuFF%Qa%H=Z!G^KMB3Gl25>l_;aGWan
zPT=YhG?7w}pfC`ZtyIcVtRuie)pNH)UvN$(w7!sMIKJ2lex^cHWWq%xL9DxSi=6Ix
z9Rwu_$#yjd{FyX9$F=S<m7-%`N+ZYXJnZz?0Aq(ba$ITJM=6)bfeVC=F8l>3xzj^H
zqjVAgIP`&h7n9@oEJ=S5Ew{fVu_5M;oNOfp7rm;yzd%?eIp#^szZ{k5BnD*lW=yxb
zMi7oa+y@V1f0X&wI3jrX+OX?%l@ko<Hg{gRcSOdcK6u)m1WKyxY$=AyH2L9Wlabvb
zd`I_>LVXI6A8mNji(_XKS^Z7!S}agK^J;;Y>k@|BMbnORsT_AH$R8MleA96%s#G%2
zhsMak2XNlaanPXsb|(^sQ~LW~RV!KiK9g8N=VzK!LK3-fDmnREuXB`a$L5SH={-Ui
zlKi~70Gzk5vy6CNHt4CAo4o$m`CIHsll%4MWMQ1M*B5Pn4fj#Q>49BbOwI{()+>=v
zGC8n>^B5RhlDPtmi2wA{i1^gXa`=xC@#TF(QQv~AoXrP*)QcMb{CxUvynnIdcB9|a
zo11IhR<!H1UOKJ1FkkJ>4E?;!chX34+-3$YOl_xKiR{CK&uJuuw->5R;vk>VM+aV?
zi33xh%(%T5r$vt269Z<10n4`zqY1^Z9rH$d$_DY4uc=n~+`EiZ-#&n!Y*7wW$Fb)#
zfj}m{KdIS&HCX#I5ce-Y_s;~QMq{?rcgKSSzaXh&u+alC=(8(C)@)-kZnpN`wS!bg
zs1gQODJ~!HH`6qctw}o(w9?o-@9egY>f;$5f^}P=QDJtOtnp@3tt-nm8>YoC#4+po
z;N<MF6)ak}a$uZ$lX&^us@mrkw@VQ2@WOU*>F}eP?3Ew<obVP^og;+rPQ(tr<768B
z`HuYgXQ<rEtyUPZaTvfJ_-W?l@0To@a@s=>$^8|aQzSeTXJ#RQzm73bvd3$dMjtAL
z3MqLKr}gvR_I}a{Fk}qR;?;1JXrynIk96+n!;#~+iXTv%O(;av5k~uHIbeV0`ly0=
z4xm-CzS$(txd2_IgS1iM<Az{xY>YXOdZvOM7$csvOgo;8!5Qsv#nw(d@m<0-`NW;}
zNa_a0P%pt9HtYndj`5o~G|&g;4!QD~Jka=%+dXKX5%Ha%5dJN5#Pi}>mbA;D$%fap
zD>zhhB52r?DO<`PY048dKt~Z}XIS0eV&ZA)yWpG@IRFe6$^qYY#KbD;t}}u|PMmQb
z7@&wuUNThOcR-cNEnvhGz%VL23Po@G;f?Zef@(moPj-m4uqv>et9f*(ZURSp?5bnc
zVVj0S^TV7%bG;cSZ4U#t4H0&P0TnfM2fNFgKV}Nyd-K1T@2?+!2y;4I{Z2q|%n+2E
zygu$7D1_T~?5aTW#G5G4nrIy1BUInhhD61Ii-cJm0o?tAv-SH226&!a#)YT)9Bnva
z3|+b@DL+2vJ$t_}E6@B9NK}?XxM)l|OWH0JL84Vat%bZfJKSacd}%ZUd~muBrPvAE
zDArJ~M5$_FPj3f3zWg-#fS&6<Esm9u`%NW_vfUI72V2qeAmv&P*zGs}q439%w-#U2
zLLeTLE0S(YI{i+AHjJz}UM@XsRj2j+;%Al)rsp39(u}#kz=gY2kb2#B(&QjxIcj<e
zd^7gL!*t9^&D~S7eQRtZ`N_S#6~~&UZGp@$-l|(1*uZFlRuMF0GwPzMpWTq=@rw(#
z8CF-=3(_J+U;QN$S3OMYbIcPD=E~55_a?2RLa$~|**%{He<24H<Tf_E%s0!Q-Kd5T
z<NRk@Xe4R;3QZBRUq|S<Bw8V)j$yQIR71C$!wXHL2z&PP*NKR{cn2h4TG3R-R2l|v
zJ_jX9Sg%NND^u`Ehou9C?~T9BgoJ-#Mmm5Qfs{CPzaXy06N<SS=Y0w2F11?#hx4bh
z>Hno$5Xo?!i{1Nhdnu2hA$6rpM?#jYRROIs`5sC$kX&`N>g_0_C@I-zRysttEUmv8
zte%2a`<op+ip3U*tU`1++z8J<8-Uuq`}2;k_yILYU32(mzvD#kaR!I2&Bv)%%=4?7
zYZAe(1<9^v(#*;tpUdxaNbH~MsbloCDZ;c?J7v;3m%Lb-VRK%jR{^b-w&UdI-~qE!
za=X}0lEFz4jd<@|`o?3L)r=vY{H;%A(T<A2Zyu-?E$#@|X{#@(9*#~Slb`<$#>-zE
zS-M$7^RqD3RRexZB=#vVJZQypxkQOcLv>eQhagCI_{;I|#y-@Uh(54uofG#6aa~{4
z&x!(bZ6TL7b@$^u<wp6uimxao`J}c;dP?qQ*pqU-0ETM0he-Mc%>|=?tOe+1>2RKC
z*{<&7gGlMyQLd9Fccf%JQGn)!ai^`BFj94Ar1$FD2eyuo3vc8OYz~ggWYDtW(a9g%
z0FW&{u~g+*vJYmWZ5jw&_H)rs{7%3$517H19EO~d&X|p#Dk&=46>V%&wV(Y?P!zcv
zcicAHyJ<N)6CAEPAvFI0MGuD&MelFdzwUuM;gxZK$lceahxDyb$L;bx$%<`Nl}L{7
z1STjl3gr1fW}E)@0Q=AJ*W}%g@DqOaaMtvL91=NO%f|7e{G{b2x4U}1l+j&dT~DhZ
z>oVBcJEp{W)GBi7@4RWa_8_xuo4M$1|Ew~aYJia|lWuYboANlE@0jt#{Wy2dux<My
z-k|L5N>w~8OTgAz-0ZB$Q4s!|<i8oh_x_n7d>Jk6R?APVjeRTjhHcrO&EWqqw*Hs&
z{@g@^!BW~dmlo`i-pdkfj<a6frJISqUR;A4PP$48TQ^3~V&hZz%KLO1wlaFtwLA4P
z7|OT#!oj(A*tg<2VbSWN<pV|8-KdIA&ALn#W49Zr<|ISJI@IsPuk*k2%5i@n8ww)3
zpA&L{{sPa>?a@C6F#l?m{+AtY2I}L7o>Cam;ON$6*=m>auq#-_-MysxxP3#vI>up7
z!Pzb%`se%8)3+Cj2&@UqYn_ID=Do?oF1gxltMAgxC97!)Z`rjF*lOCOrpmjfM=nI7
zMrvYa4coY8-d6cxld{J?EJFq4g<!tp?i`1p%_zG_pPB~)9`PDI@zDjEvTe08`lT&-
zE~-)C_FVBpLfSkZN(?2vJ8K_l`AM;)iY;C&6J$@N>dgm@n*FT>hz>=9VVgzz)^V;z
zjKiw*?5o3~(U9=rYC43Hi;(6%Kd7)2q(g2OF_vj&t`^VP7Lif~5j+(>Z8c!S*$ScJ
z#LLQ47ouEIhi|2>1bsvQ@Rp3@NVPDPd4ss~<%A16{y^K<#OZ3t9gQlQ9dE!jozxf3
zsjK8<ehpxZWOgMI+z_$FexXBcX_E$Cl`Ii-ojG1VFkJA|ln5wKB*Q1I^uB^q;@vR;
zjn;A!p_ES}+uQeeGQJI?1{0$W>l=_0w2evRDQ{4!rhNy_Uo1>)-<22B;N*d!!H4K=
z-|`67%;$$Utl+xsQ1&JiXsM`uryWTvWY5uVI_qQT758#UspY9AYouPUW!u5_-q6m=
zF@HYpq+$<e^UV@LdAsXf6v5n&1quRgo(iqu8*n_wMy};^*3Ba2yJh7mj;^c{X{x7P
z<SkxvJGt1v9q*Q0t&flE6k#Kb!tp%;>VV}E9-wx-^fwLf_-ePDiSGx3F<gEcP{@G!
z*)zL%d-tWpgQWJ7faWV<+21-)Km9MK#hp$~_1=x-p@L_}5o8BTNNKbJG?0mOJ6=-?
z6q0>kH|2etxT~kMEvu-hl1`Xfd1RUora&B#z?G<_B^3+rPiX88?H>Yu3;Kp7`BpV)
zYlgnuqWI&P1e2YxRIYJ1`Y2FGM!9oRE9po&!|kdqCWxQ;fsPOM?dnQ_ou`7RqmDP8
zITV<bz6Q@%FFZ+8eiO6|D27;Uzbym0h|5gh33~Yz*A65IO9K1)^(Jr;-g6p^s;WuX
zrjoWu8v%##bK?)zI0O-;9Ub}l3V09isVv<m74Twyyw%8ceXZIQiHot6xmkECYGESR
zqjs)$mT5KxZI}!O_dOl%+iM-zM0s`AS{S1o+fGgd?qe2$`#4fdNVAJX$ws9lJXxgL
ziB0~J6u$;@RkX5>93&J_&(hc-+Jnn@9fp=7e&$X8PI0>chNcW=Y=F~N0E~H9?7ha@
zSa2!tBhC#`aWRcJQ4Q-5xpVH&*8r5ij8{v#9y++D#`C#j=_7KRc-aQ>Xrqo8|ELx-
z{2sG*<CyoB*xpm4hsej9TI0eI;W};7gal{py(}$KsxJQ)ME-dEFCh6hfU=pD4yS%7
z1|Q)qzO-EKVR_|YwNAA=(#;sO;bCInA!?`p-tH=mqA|z)V>(li{da=L(JctMg^k8+
zR4z3KS8CF;KbJwPPSm;*HZ%Iy@ePA->rCm)NHt%td!FRnKtLD%dh;8qURC>PMh+K$
zwtdt-&aAAcJP2)Dq8xOttDH8hfV1VLHzoX5&n!J*cWJr6w9;<tg6Wf^xTHB~SW??G
z<MHD@AxeHDlld!cL#mrhJsm^bn+~Nww*9L+=HE!}egF{aqeCERCSTpvS}S`pqtkvz
z&4eT@S&GCn0F#M5S3JaAJ?weG4l8=IVjx3I@31M-&<yCXj48W)Fe#PFxRM^DQDc~M
z;&PW?i-xLwyu@~S2>%@8^4e+U%o?95(z1AUj|)U{`x-)CKhIV8W7;}p6=>fj&TFX5
zg+D&;U8VYw6LJ(Fa^KxqZq66oFt|N<bBKvxBjUluhkH|ibuJniYFG%)Og;f-ZgBTx
zj}aFRripzZJw#mazGnizY5FZr%d^3QqgEIPn7j9GgUM;bp-rM{0kw>6NBdl!pp#nR
zFygiwvGZEWMW*j^JyWkgsNZ_M2c0lDhfA6RTuwY)66uF;*FvA1D$lY$@vez&N4Dh~
z%hvCl>_@vtrIop;CgjuQNE03n&s=sL)ZO9D5+)&3s~EKLD7gxC-GAW((w-qO{g&by
zat`pAU)_PksKiWi-*9`|Sodgp-<xrL=p=a?@gCajeGy{2NH$@aC|U9?`26W$Ijt(?
zZ~!xEbebCD96HY<B5^{l8t|m_ov8sUb=zP*{LIG%YJ`G%B(+ldfTC7?(zi9Q5pu?W
z*4%;U+N!uAK&gUR`CF;`KMh`*<K8r`tM?b=zbFWe8VW3Bok*znvooZwW7fw*28@nL
z>SLYk4o&UeOAjWwqi{y^3Z8azA8a$rU~VxB_EU6-G!IQljQs-VSCr)X9AOjWIsi{&
z%H#Y7WvpqVoTe(2-GO+HG+0%L;~31y&b=k-bstEB9{hG`0{t?9Qfv5^yLz~=xzDNp
zl0N@gYvMmtOZm<ADOIIGyVDO@rJ2>v=H=bJIaKoGBAY*+HWeo)7OD6dG4yD0GHW4H
zV;g*PMU&xHux#8ZbNxFL`3m>_Bh-Vg%7;>yT09TXq+?}=PjIOW$cxErSiP`i2|-WH
zJNSxN4}lj^{H_><)4D3qZ%)X2lFeO|^+anS{z}{{fl~Ub67LN7B2@ZNQg+rXxuEbz
zDC;MaI&F@ACQrPMQKsAvlG93o8lwlD1}HPjB4RuOgU6hv6vJz14w&xt7;rCJ<~RR)
z0wrK(#z!8jo=w|L<EnBye}wFLnr{;GnHx-*uEAf2YshX}mVe{M3eY{6kwFO`EP^%{
zwMQgab2nc)(Rmbv4b$Dt0}-3J!Q@W8R9ca=cu@>#Q`d4SN@y%^OxGg(ou6%To|Nd*
zngQO3rOqWq%b`UBk=ykqHTttW=h}2Uqur-Jl58laD++`*fdfiCRWapQq9LJJwqZ{a
zz6tJ1v%+_CsSMpJ!Y60KU0=WM&JXBTOBWr+^8sTU8|Ljfakg+ErnOLepRc#+F>myz
z7i{G(LuNQq!n!*)5g(9c_%j&YCPs2@<I28L_5$h{m+%_5;0;;${q<?OA-td~wkrMo
zh(mUR_7mh@c8w@{)3{NzXlj+l7ps1YSTge_oT=72;pGkek8E$hls>X8EAr;sWx9}n
zzJrV?uVs|#9T(@Cn|V`TJ}*NjZXWh*`Xb1QU0!<S3F$~(^G*_?gpW=ueIz)m$&HrS
zPm{Ku6Hh2Qx6#I&#LJtQ04VLg0f;nnYEF<C%nx7F!LHe=O5?MT@lrQa{!9CT0_^9A
znhgLGGai;2mqRX!$O>7#y<YEhD+T(3BID9Gdb*G6pL{_@<8^SN4)aPn1E;OCxwb??
zFI+XYnpO?M=W4G)uCA1gkhKrIeSdbtUZKU9l8rU4FB4_~k?rG_gxEYA+>98%ZLa~l
zAH=1r```~bR>QxvW4r&PbH-o{o5d~h0LLGC>(ie#asSlJ{Xca(#-#l;m!#9(NAEN<
zxkN1DOr-?UeI+a3awtmF@o`FeWuxPKEojh6heqEN{f*^>#G&{ec9#MW^&$-tw1#wS
z93Js9JF3e5#B%$w=guS{Ekq2M2U!d2{mjbWKl7*ORAN&o?sHtp+nCe!(2@f|%WL&=
zsxlJe$`b6%w)*qhpM0hx3m?gta(^wOfboyaUxVs+B}_W#>z*8F9=5zd1>qq$N|!k#
zwNkeJv;x*b?}JEXcm1{letkTLqf-qUQMhRO!$dyoKUU8EIRsxlhmQGD54e^WKRh`3
zf9$<ySX14)E{qkWh@f;KDk=g3(u<I&ymTRg^cImOM2OORY=G1V2nY%h0qI1#)Ci$h
zsiBw9dqNE)#4~-@*{kfm-nH9x&UdbFuRmVNWIAWYm}5NiDfe^VVse#{S~wpbf<|$F
z#8V^>88IWi3L&8B4s!vX1x4E1*Ktu2m9yfNFJS5Jo&s4)tSd0<m}mA5(K0+z^O@A5
z0skt06o@-+y0Eg-%o4vcM%An=O===&`@KmyJtOZu%Vh_CZ44!D5zwKv9Ltj<;0pux
zg^6GJ42ufrA2A0eQXd|^_va+NWZv_XSgK6QHo=G*KRmu?2dg3D>j6zMtSt+-zhIOf
z`JuDhZ))Ew%Pg>hlvmfV$Tn}>bpj94up#4gyZY&o?w-{{IWQ;tI|7ky_w9gK?!ga$
z-?DpcBaQsR*}a>`9BDUMe5!F3vJ9o@Y7%M-dS${{&YI+}C;BiSrcH(BPrA}P$Qd`F
zOpzC|dRj8yJ$niMT-6;6xu66iq7BSDSq$SI@S07y+P*>A2_CKWQzFhVht)jEvFd`g
z?SyapVH!5-NGHsUg0H%X9U`aDm=fc5G(Fa3l8M}6?no1Qv#6N%HsxV^vk1~pzJ9(o
zxu~|R|I<0sF`UooS<Pc1unyyC?s&bZHet#QXF*PY(zh5q!xEkrhLYj70OV~j7yp>j
z_ivbL{sBepADxG*CY4267pSH~19)H>l7;c0gu!x%o|O;*@!Sm#2J7hG!gH{-&Ua|@
zExPsSx|tfhi`g@5=2h;t7~JM%VRPKD%No5pP`s&!O@?MJ?H*Wul#MUEar15}>-nd2
z*`ibC=HAo#T1}~UY+}D>3$OJ9T`7Oozxx}r^B0PlzhawlBS3=NPz_OGz$~2oKQ{}1
zG&}!4JFbM>jUpjKl|-jU-9C+$wVLnYwG>cX->L9VWC4OGqM??)yaMl>#Q4cG?dwlE
zllNRs>jin!wXYk2OcD2Tfuyj)d|6M)UUXX>{OIg19Rppds!!%iu+n3GS<=;c)UguD
zCUOw^#q|yHe2v)ivk^pJBYS{C(w(t!Q}X~IrYMkl$49spJ4PrFwJQxMNcZcP8HGCY
zwXH>~Ff5cZGidu)LzDuO8-3jiSyN(qK5o7;=Qhxf!##{()wKL#{qHn1(KC`jQ|Q-$
zLMH8q0>|i*sGDMG_pd^q2EeB=ksx3|q5h20do<qh{Kc`NoGoda!-9bW$Q_&z(=nmt
zQ!SXUt@62&Ewgc_w^a7+a`rv<0a*qS+>GITzStu5UTxocNmYD#)k|%kchfPdb?h}$
z*ie1}a)5t=8AqmaY<5?JD|=03U@BGr@Iy5x$!hRUmgAD6FK)K_#nYC6r}8qy<|fDx
zzHyS-<yu9bUg<cWYpb-1<PlnLR2*d<+T^1fz>@n&|Mlc>JYdoQGTZ4wA2pGqhv_of
zIm+z-VK2!hNBvCV%r}m<l_1NztZz`1@hzAM5aF0X!*3XEXkvqRPjWN!8WVOJOUm4&
zA2bl#GT*&@h^nu|90N*sP~3;v${(eUYkazQ7hGv-)lXJ%ANeJBiiZ6Zz#p2aJ@{7y
zh7xC{yYrR+KCljA^B98y#26-dT)tiE^LKVWMElg|uR<gtRSqv;M&5pniok7q7c}}W
zVuOP31mTdvR6*&MEvBTj%!u@vg~~vb%^LCB?{Z<7x&D42RZxGc{=X72{KNkv@`2PH
z1wfCEEQ$K~x-Pt#;WIffXajOv%%e+JgWJ4bD7D6&R^5>&RD%G(T7;6L4~fUm?}J0k
zz*zX3A-+N}->{v@<K&U{W*)7H;g85bsR11qb%*0pN`(#C0*CX*VokNcy}kGH3zDt@
zLwE$li+_R&|DDD7&rXJa_^AGOAecL0(&)!+3h>U38HNdcA*?^2^<Wk@=Mdyg0}@-t
zu5DNJ^CX$Am(5rvVkTFLYJ}zAZ6|gM2j~T<iU_G)3huqdqqQLk<QD(Lg5nKrLhyTI
zej^zMWdv2n@8(|g)`}84rOZ<5S880@w}wf&_ATh@<M)aj+#{-X`eIJ4L(e?zfTI4m
zlcG$1<61>&W%LnmlucMq9{QYjQgglUJ^WtZ4ZTmd3_fxwKhHw}-6?GD-JoHqk;+C7
z&_TOB<R;Jr2)0cM_YJB{cK-LYGyb|}_3JYrcAhIj*M>XJ`c_r+{Bt5Y0#pcef7S{i
zs>;iFIC7(|6(3<!+d-O<;e-vKc8Yv`%`dl|qu<?R6tNqZ6AMy4vH{EW;A^Zlekv{1
z5L=Hssc>Er{T;t(3gWf*QO<6&8jPVxE-miC!ZwwdT(m<PC-f|e*0BbGS8V!OcepUx
zWMj%*@9t?n0ys8XQ#pp{PKg&iz1N%{&_@c{6(M>xWcrQY0I@h_UQI8zY!_;#d-a_v
z0_>hJSTQsTe7^J6vZDF;5-BEd*q$jm=b-Rd*dbnwU=bU-6kwKLkwQ$yo<YrJljI|+
zxZ}G=)a?d9_SDOVl}Z9Lk3SWW1gQ^3vM(_eO{oa*zqmacD6?H!+AkMR>PARHnZmgF
zWsjfr!xL^R*UfsdFJT!0wih0%U~QC+h(QwkBGS(epeCgMqN_m_nU%saKOU&D_~MV+
zw<~X5ul2GD$$do-s{bV>vw|cGu6VF-iPu9$)9S5cqi#Xj1gQce*)vyu(xn}0r^Ic?
zf5<k2%Suc*?nzjUXh;@+AAL)^j6Kug`qA^jgJBy-c}U%Xa^mhce?vgrJgaRk>=&K%
zFZBPzf1NM-uYP9_ol#cN@1-jExTTK<<m{h)hvHrv{7Sp*Z`20{?ceOCP+D*PCekmo
z4$u7IF)Gfw^mNjsqu9KyVrqfEsd!%BMQHh!hf?g6czCr=ffdYr+40J*)@rA8V#gEB
zwnM2#KTB1F<B6oWmsdkg;xFCimJvNqnC6vx>aq$9<HsL%2>%l`_vbE@|8s2}!Zn8x
zig6#y%1_sF8^gu3a?jtVPw?iPIukO5K4-Ne!I5GLl3g6X%ty<VS>p5!Dh<8!OzCNY
zJfLUg)agG%qI?I-3|3oaPB?n8SvWR6kCJaa#CZvhJO@W+<RMQD!5mZg3d{Nnz<quf
z9%^25i#7ipynQ4YnbzpZczswftH5))$z1jl9%Z(=6!TlGknw%oL}fPNDpWm3>fJec
zUlF(aKN#kwX1&ZBV`L-4D&u69@1{G6G%P%OHulTz{UIOh?{@BrYP@V@ztWSNdck)N
zdV6m3E07-S8WJImB0fBD)-TJ)`4Tv=7nzrDU%=xgIUzP1JpK|%nx8@_Zd&wnRpl1Q
znTpQ9&-#)#-n1-pb3M#uny=jec(VJhIx$%FqBq-N73)}mx+dG9C$Jh9Q!iK!gbe`c
zmPUV0xfZ4>4{#+S&6cark<1%GEC>O#>VQM}Dc8$lS^={2SWTKxU>o=(Ih6?`leue1
zaTUw!KiORiC!oS%g_y`d-cIGlkDyWC`&ZHjKNSn_PHw;|?fk7{9f$|Dmw&`kq)LDF
z+p~wYJ$ujX)@(S=KS%W1{Yh8cZ^Q>_Hu~H<6GZEI?H@v;6Io$l_QtD_f|4wjc;0bF
z%eRcoK;q$~nB_q@_XzKsq|*jdX%VdEwXEb<1eRl#SDV6fEKDjtzThQU$v&IzqUZJ<
zcCUGd)|vT9cWf?o_%(rH+nHe{F;H-|TL5$3xPI~!a_W;YN1!fx=EFHRj-%l#dWllY
z`;fNUYNBx?ioyR*sphuwqlpdguEuBI{EVmf%bDkWZX`A8`(rywQIj->5o@;#vJ!#o
zX1jjC%j{9u{;OYfD}Rq+>_3Ay_(jdOs;>3WC|18Dcu6!pnau~hY8#Q%Oe%k|4M;aG
zi8}bB<xFPz3=Z%pGH!Yy91(bRbj;kMg;>0<{V}<aMTKl!Gh9zJQ@v2cZEJYs<Ic6O
z5j7e8(dMRUymf($g+$$VWk%Y~B!%Us^2N`knyRL3fEm93k7^M-wf_QTt3b2-q-*Xu
zFxIgGi;zQfAg_TXJ1Y$g{O{|6{|`KVeVON;ar?(7vBeDYm>c#*Z$mlfRwqh_``zv)
zu)d2{5!xxpF`WC|RGg2x9ER_g<_vLv>%a6nB=^n|%52A4zTY^stIr%L9sNmOtcm0h
z^u*sftuu2{jqieg{<wKz9xMdH<PNA>#k<~+PuYZFpQ>N+4G<|#e-bCZ5+x#cYBFgg
zYJKoB#TMkmG2t5UDY9M}=LwFJ@!x^GpaMczjLu@rqnlpg-*vmi1w>$X8uAK1SGhdB
z*|{9x<k;D^XT)p72dH{9{HN<zsy>b<(;Tnnk4cQ1wa8XWiOpIs-dWM_xm)RV7wk%|
zp|Fx61P)5gIYfwz5ltSy2ig0M39S_l><pA6>-os{OSg(`Z`;gGnitGP`^omiKzI)A
zzvwV|-K>^UD{7d53((!+k2?_4ZZ`e$z{vwt`lV9aV@Z@oNn0Q#UgKr=QFU!^r=l7>
z?|WN!KZXCG@lyRsnELiC+a8b52-Xx)F_YFax1Og@8Bc(K7W7B3W8~e__C5K{!8`n>
zI-icpA$E&XL}ozYpdwThjeT4p0qSg{t5>x)$Y3VJKf8w+lcZg5A;+KxmWSnb1rfM3
zuvo2mWcyq<UxGmZl$fx4K->tOkHR$$=6@P<KDPF0BT8+1gr0qS{S`q+ypcm<d`XT7
zyFGKUVG7-FEE*QJnqiWA^wiX=Mt9l?GMa!2-h;}8$b3k?S#e@eupT>Di;yScPlwi6
zlm-6EoT2tH4ckvYs+ggx-dMC_AMEOek&)9V>Un7#0^u-Yc`3uLJ^%x~kTUw=877a)
zNwTSjoK20~X7=r(8Cy1;n%&0_;R!yNGxmwMe$v4|?;{i8pqjwU;a>(#oa}}#e~=GE
zZ1Yf!|E|l%e%xzt%c9jKRshN`r^e;Ny_;#-d}Xkx`P{yK(Ob-70K85~fVkD<2{~Wj
zvW~dIn&uL}CHN%AhXHa%guAmbJ$~-0TW{cHf1y6`#eKGJO8B=ohnqVhi}0);+qWQR
z9N!L^)PJqx`cCJuCxfhE*J6bxjfAZa0_LdXA97{?GiUaH$6ViY$m#?d2HO2KNq%I(
zdatC9M!l9gJk&Q~P%v%aJC%;yn(eP<UZtboIq0ovTVc4*<x!`6gnd;m{_Ui>CgK@S
zJJTz;0Mkq1kdSKeFWzPYYuoJxrIDj<$GKWRF`J154c1DM5h^6R@{o0xe#`c~_n<gs
zDa!`e8(9S*u3uX&KL?E1{Ue45=|#l(5DU--L#e8aAg8<No`UuoKV0?-^U_dQ_fn(R
zZzs6~E*B$80OIH#+S*S#t!+m_<-I0Ax21+%dEnh(gpn%MiD5%WS?>){KDhl40+)*n
z>W0w_in*!Fd6F#C+Nw8|(g+T#?Ng|*vn_~{{*o$0l^flN!NHUV24FG`#MO!tdE0$H
z$x*NHXe_u!7NMM9DPVfd?rSW;9Aim=@K_lh4~6Mpb%V;Z#wI)YCy%NkGsZ!o2VYPD
zpU0GI4$}u?fTH2s>_aDKmWHIv;B}(_*#o$O@mm{L`Z-vB3Yg40wee<f^ICqo$~@MT
zaJ$F>FS9A$h#?EE=cebCPn@m-unrd@!B+MFMQzlH!0~q^nqgvwIuGDLZNWJKHc+V<
z2~!Rc=8fLHm^JK}8<G);VBzpq!Z~6_ztm(FaNZc@qj&p0U)xB2M$z?VB-m9V9hP(9
z?%r^}@0TytoP^5kKO-QH!fDLae$Ecs--;5?a4E(>BuobJ4@^6X?}(@fZ+mwEQZi}M
zQ~n?1QRVftb!Z-#bGeL?r^c(;I5l|{>f`BmxUiNEpzpY1I8AMo`Oc^>=wjNSA;?m7
zR3Kj~&*-u<@(FgciI{j2qIfVljaXZmhO+M)1zvqo!Z=3fa$#N&@lertCb9g;{J!&o
z7I=^`&x=w{DF|@kRH`S4a@VRalura`s@g6Fiu6=#x)eJmMb?Zxkz|mk8RkbAFFbgi
zU2B>Pky<o(7e@e8sZ}-qFcADZNv<P*KE~cct|A16PmQ+2&0D=J?u~j1QEKPE=<69j
zkrVe`X?Wmmj`K1mn#=UJWIA0}4p@0Z79mOLlhPABE0e1>R4N1+H)je?V|1gmovj)q
z*;5AXs9Xk$;2@h%^(2$g^i$0@&e~ineij=O<IC;==Un_8)(^BM{u{RPzbPa9s>l7a
zHT6GP>dqTiQ4J$<+$n<TnHd;1>7F40ih<9Jjbesd05C4rDYyzA{&+o2=<+&yw-e8W
z_}OaM|4!30_-d=UwJ|@j(^JQXiG1z&^{VCDny8+fk!#3BA9Z||5y2)i)*QM^4Mhch
z#HSktR(u*F$3o$vciB@;UW~Ig?UdvAyeb1<eh5h4Ri`KtN35Xmt0jw`3Lcj%W^z&p
zl?VCOHqPlLjc}Ln6D1-+R?it0q#>aroNts=JkmxjFe9n*`cO9CnWek_*+R-H0Z~Q&
z!czDTtlQsoIQ;p%VFArFiFO{K))8~v4)OwEjpi2xf>te}VB?WzPj^MqmwQq1n?nzU
zc7|?>%%H+Sg`mh<{!wVwQkcx)HK|DAOc!Lg!xq03L>K*$hZ2E~D1MUiBU0BFzWRH(
zDlX8_I7_y$oSkVGTLFB#P8FT1UWc17vrapR_xjniKsEH2<-HM^dH@Zt{4vYxWboZ;
z+DWV>_@X~E(YT@AINYqqxuX*x6Mm}3fuJnc>XS17!sMOG^5hMrbJ;$FLxlPjQ6X^W
zcg#o;_H0pJov=ycYC2^Fgr)Y7hCYU{hxqr`%*|oPo<f+AE|?46l8>zicfC<M_s_M)
zvVP^SdfwU?@PLTl?I$BiN7s27!xMTuxFNxFp7oF^hcN0Isl2|M-Tzj93J%ue^2CMn
zFnJ1P5fL^gF%+e4V>v`)YAMe0^F4j1CgWuOiZgo3T1BpDYVyi>FQCqsi!M`i$+ymL
zoL?$0J9rJ7EPbm%KW9{FJd^M$@CLd3yPRI<=aq!`tPt?+cl$a!Ry*4W+$8zB!<&-q
zK?sNA{pB+6?-aH-v;4EMu()|EUVCrzc6u2j<+6mCOhNz{^ycf%ja)&_y|nG40KfKe
z+6au<A9m&bm=kO$LH{iq?^~1DrMD7R!!~(K2OmX5^LMHjal-nB3REdr(v0Mo3ry^u
zCHI|uOzUcQT3Tq@8{-V=_ILVg;CFHJe3$-+AMx`a?6m^=8BP>JiGF%JrD-@L<&l97
zD*lF-jPD+yJS5NGkpI=yk;D9`LF+v|lq$q*$wO4E&#chTyhWcdp_?^xaZulXDXT`U
z#LEFheAR4hB~Kjb_B+<>!8+8(Uf)jpQLvutc;HbM>%f=)Q3!$Hs&9HlXPA`r<#X)A
zy<g-l|A77O8~B|wLS<k7$n|u4LrPaJZR?1h%v=f5cYQptSe)E9uqqP!7thj75?*Nw
zLF$#j@|I{a%$beIQ{JAAAMTS*B!3%?vZcqCp2H31)AKxa-oI97@Y-_j3QG5L_T$Qd
zgg5RjGR%t8E_n-<Jsrth?wyvHly?@t8Uk-U(D}vg@c)d8?B6B6{G$i^|Jw2NjL*-N
z9yEfA?a>(T%}Gn`gs5?JKnM4-FOXYbecF5`(Kh;}z)SWf6CzzM`vQ#59eH-e>g2Z&
zT%V8It3(?s?!F7pA0PW+opS3<L$qQ`NSaVfwnz)sfsekSQas2Gd@;Et8taA-6y=vn
z30T2D{+HQ00^|(E@b~Kr5mfLQtFfuQ2j(S4U$cc>NVdNi54P?S3yt+N#~z53d@BxH
z(uh8?w_##Da{TSw^euefH;h)wHOXExsnd7zBf^AC&7^{lgM77Cr2YfRCk-aZd2i5N
zQt}6LIL{N#!mrI;X_ss5$Y{V8{-pciu{d8rjGOMoJk>JSxcwUK){4o`!gz^xOI`S4
zF{IcfQns=_G}SQmlw3*~f^}!_3zE+5B1n@s^H5Ix-YehARRB>4&oC<iGNHOn;}`pX
zk>ZZgRM=rjW$8=WIre^uu?BA;;b2fS21E{}2$C5IY%N=kg!-kT-vG%{F_s_3t3Jk|
zg`U9<T9Xr~ePkV2n;qwBwaWt&q&dtR2T_p(mXl*wM%1#O*|&LFn};1pO=H&f38*tA
zvJ+QJP^KjplxES=EY`iT1mr3`0mHmeNi6O6J^a)z?ffQ@K2tDhzsicJTQ%)7A^Je8
zve@p*xhwbR`N^0cqPS{9E!rHkUVHN~5i+{j>VcZ0sW0u}CP#oITZ4uT1)L*Y6+g2&
z^6I#2K#=cT@=jSQSInUIx`0w}WRVC-Qh|s2Skt@$XPQtQ`vh5PCNolr@rLu46=UAB
zv+7cuplr;Gj%tZak#4bkgb<vZHi08uiHpC?xuw1`$4*T6Ww3)&;dAER#p>q$T|Aqd
zLk7NU$C3`<-zQ5JaF{m|*?k@ZVJ?w;Bm45;S6*h+T*+yR8R06-9e^fOI$)0Myj8eO
z_U}6#)_bmabsLFKU3NF%NEpN~@{bTif?R?HnW;H$(fF`^Irr6WqrD-(*!>6POiXJp
z7gyE(*brU)vMH8)N>kz^&ppQYy|<Q9`)uhMXLudEM7n4fshmZmvQo#a8%@Il*Z>sa
zScV(uM702+Q@p_oZx*q^9IxT=c72QGk+R4bz=Bp1Fl$`DZ6^xSe5KO&u&9hH`07YS
zZi$}x>K`KXfM5JKGHU-s?ERTm=3o2r|LFUX#H`1K<92;}uD=A5J}5kye3GM{x(YAN
znlF!<W@t%$QBY|Ts`7g=V~B7)KT_q6w=2<q1;|hqwg^8}V@bv6vkU0`4(hG{<ul$1
za5$6UAd_Fe5o|ZK^3m2f-Wxh43h3I|J<HKnNa>YjhI9BF-}DM#(b~xoQMfDVSo*c^
zG0+<Ny=pGn2DTybJrf^EMfUhZv(+8MVz0!QsNvji66<QbmqmUA@((SKAYZRu8l2lg
z41k~E5oLb;*%L0ZjE-(;K%UC_A7!rpy}z-`h+b<pDTn*5Xt<<W2b-R~K$DpUorOFC
zg#^mRYp#BZ#U-%B+ZS?=O=NlmoKW@kzR<@T-;Jg>yB`RJR@u7_*7r1s7`}9v$2W)%
zsWb<&fBTl8eiQEN`mz4<qrR_u_a%~uP&Tq4jOg4`j+gf96ho|!H5lO67z(XkqcZHB
zXxzO@xASB~#k$DRFY#w{%jZ=CGZMTLqf#HwWDAE^?!lH(i49NZxL4F%z?JB*w(&Zu
zs`KMOX}rITqE5zgxshvXTakt?MDg$v`2~fIdW)3aJ8m}8nl{-wXp#<w9TMd+CrJy`
zbH!G@YaKLBya4{HQn;zjSs=^t;#_B7^sAZb5Q(r%D~rhOq`r-TfL*X}nPwEOY^ov#
zK-1`SD-%&`H6(GIM2?lyF)PERU0ei<?g>BG^PhB6X&h2>N;aRAF?P;E+6yb+@Es{4
zhlw-+Z&^8@{dl`07r&a>1tG8bbWS7$WL9WbiG&-2a7y<=HZNB$6DOG+?z?8_o;iHD
zw(m&hR=?Tt)TK9e8%c!rbu#Yuk*y0_!VlAA(N#6`{l`{1eH}t*!rr$U-}691OUCbX
z6Yt1B5+Pyc1j$YVDQR9`%2sYwtnm(>W&rip+OSUtAN^d*Q}T(rieK#3O<*%dkuIXA
z0*CLeJnL{NV%gPEc!$x0MRFk7s*P<>f(>RdGqjYC0H5rJzYN8~|2{D%#}qz2;Qf<M
z^R*VzZ2yUR5P328M}wMGHfBp%*k4-}JMjQc0=lMm?8B%ty})dILb8h1hxfp}{o;Zi
zKr-CpAw<1=&kF&$xMvnW>8j*mOIFNK7O7<<lYqFU_+YcU8})p-vp=f#Iwn*UIhDcq
z{ToqypmA?|)KWM0C*6s&#N=Mg=(c0t>1Fn$dx3HDhEpjQKNk=?59G0}$*f&Z**py)
z^nd`ZfpOO^q&%TM{B>tnhR>79aSoS(K*;E;X{JoZ0!!3h-bdpEG2X7vtHd}6k;4P|
zc<LwJ!fN2w#A6Iw_-+STM{XQH4>C7mw?;l;3DW6nU_(ybG7`Vk8BaC^#wAV-JJ<zs
z<?Ta<i@Rnb%?Z2T$R^`TcPICz`hYgSd@!If{-b_!Z?+%#$R54T456O*Nf!;X#38#I
z_tNlLn2P{Ce(A4EsW2iZV>KpA%j+9|(j^9AH|o+T=cIqqeRBN$Hm_E`A@<3&+kPk5
zXPhZkR1w(zdl-OG>QtL76R!_iA)H1MxpS~zi;c@TkYvW6be97BfPW!2dub!rF)XpW
zy<}~Wr7yC6AkrvycbcpMvUFWNyzy}`Jft|(`E@oRiUipdps1WI&jr=+Z>jnp%B>9(
z!z4E0oCOnT>u08N$npi;W4Br5egMYDhs;3mb9F70u>tpjy-%@wg`(ELe71{k8s5-g
zIH{?e{k4;?;gLqNZDa^j=0$uX@*Ebye6E3?FRA%%SqgW0lgYY1PFX5SB#expcv_H;
z_WC(el@6-UcwKR|jJqD-78Qp-!$jbTDR~+-;Uco-pj7<vxVs1P*CQQC_|7NxlddT{
z@AUYMq5T+h)JD{EKb{?VYi)0vFqu)F>xQt~A_Q(=*|3!#V2|b@&4Jyz3P?c#3x|Io
zy*qANg8<$GW?pM0bD9=dhRZ;1s8d_w{a(=gsEK=#?E4fs>iGu^tC{URmsCa<=nstS
z@YM4H%`pxwWo04DVUxX8Hq-L~qSL^WM+kt{sGMGVpwA`Wu3@Fcun1IGZ~3|mqdfhg
zuYi%vPr9{25<<Tv!K$1NX!gAI8*I+<D-o}l&?pS2O(20J#zEt&#f&}iEr`B<nequ{
zUU9uJDrx}Y2z^ue^)iz6UIoy802ADlc_G(`<+fmhH6%rMXSAbMtl}Vyg}t!qhDLut
z`${0mLR(W=sLB^z2vyQM+}8{Ik~0Q3^Z2};a&RhP;3EUo^#g%3OiT*fpqTiRuDtvw
z9Z#lU*v>F^5bg_z3DvkS9N?cQiDvJT;Bz4IHE3vp77Rr_wnL*35?;A~)maFWu6n4r
zAN#_@3t6I$6S?U&6UlQI5+rs<W}gq7>g7SJT=v1PqzC0xs4B&Q;(XnyqV?hgg6$!j
zf?Y!?X1&j7ADwF!vW6fwLt6p4ua~?L$ZpP)U)PZ?EWsSWC^f4SXcal=71(<&sA`(X
zsH4o%17f!js#zZ~pRo9l9L}+;y$;b5!%lsIF*$1<#hbnVN!KTT9<WUFrAYGi@}RWt
zU8_)}c;-mjv3(x$5gU6-R`a6p;}S~6DC-FH(zZrx!EzwKt#tAnH^h`n8`7+4wz3Yc
z>&H%|oBgDFUk^aFY*dFj)_K0JW2Zh_$$mSop6d7(WZ;EKx=f6!c}B~<+duo0&b3$N
zP{iRLS#?I)v=A{}qfQ$!uS7~%p$_tKTpm7pAze^v7jjLh$;=JAGV6DBZ*7glv7`}U
zZboT>?i9fE4~xJzCJS$p3zpkQ&SLwG4AH9<B;mt5;kYh(iVNrd;5cU0oMgXsQu!n+
zZKs#v{g3Gz;#wjDjWjRb6h~tylBA%ly&Fo7%UxJY5yzsHT@8?zENUjBs#W|OGY7@e
zZSayjdzJ#bC&;Njm%}OAmSFIznBGsi+wkgM$kd0CxZYP;-u1W69S)RCQX>_faWX!1
zk2Hf*C9t%Z_v+#ug=f|!_m}q3{5rXl7U$+C#vp@6z=A2TkF74^rvlH4hf>x}J6-h!
zhbkdA7ASj;#L-Fz$Z*=}=Bf7Z5YEc{#@_2){bPxT@UYA`O3HUdb0RZF#CJ3X6*$1=
zO_S1oh*isM*=&k5S=hmOlhMoQ%IvJhWFPz5K#RXH*(SsPGSMy}xCD07are`kh60mK
zO(k@`#||~}13~D~7L(fLkUE7nt-}w_pd4$*wkZB&1(z6@D-h{b$NrMH{u^ibpWpxM
z^I;n~*VMu3`9rs7q88qdeayl-LN%pzhwHo@+Smj)sdsz_Ro}9&O~~_6x|-gZ|D|A1
z@545iX<;Xw<EU1jd9%Q!$izF>$JI-<ExH@T`Az(liK`<STTJ5Z7kuVMzviHs!Pgwo
z=cEy3Lu@gEZ1v;!f6&|fxgK!-{t4}TS;o@+1bPZC|Ka3gH*>qUKA~$Pl}Ui$Y382<
zPtRvtovKSOsNB$cxp<*qkVeeo9p~`V*Fs4stD9|S6D}#s+Mi+O-qwF>Fcrb0wS4hE
zvK~Y@8)_MUP_12h;5>FUn_rvkZfoS{2ru;~3kmgk!gH?aMGT}Dpc1J8lv8Ap#v^mR
zR}$1KXsp~vl_(L$?UQWCN3b{GV-E36<4WrG)rn7z?5o!<A1b6})y?|y$L&=(Tv$+m
z>v<@r`)R|P-Rc@_-z(kiRl)%E`9@{!JN@pCNg^vk`RNZ{Ub%bIDMm{5awS_-b?^mN
zSW)eDG6Lx1x!aoico(n0HU3rTP^8Y^gL0fK)$meNIy}_fJXqndcb%K66PTSqPDsF)
zJS!QO8~1Cc0+641IC1@QN{MX-S&9HUMfD_MU#36EgyYRQ(-D_MwlTg$DD}KZgV_UB
zmN)KM9=vvxZkKQGXjkw8B*tAH@(aK}Qk;Q?G27P#F1)3Dv8GM2Paoi4`rnzZVDe3g
z^m#N>LX4w%?1-PnS!kYloMnJDv;{B~QtgIrOf@5$sM)|+@-Gh-G4uF`B@E4Cfjv4w
zo(tXHR}&d7!Jb><>36&uoQ80Z0E$cVSa#aV)2Fzt{Ct)s--X~gHpj}0&f<f-*Z8Ix
zszkkkiwRE`zy3e~h2~JG5-Osu^#j_cvKz8^=iovQECI}VSQL2K+qgYqS?z#*Ic5gk
zThm-ONz@DX;T9mth0waWD1F26(H|yHVj)drzvOALB~=61xbI<CbB#KrZlG7EasMaX
zcksaHYD68jv>6BC$=T=W-i7naE}L5`P*?i4CpjKhc0AJNGwNMqQ4795KzyH9#r-lt
z>{`SpYM*kYNs>(5QYQr9gJs+*nA=b2*Ry)o3ls<bEv15RSz4A##!$Q0)6q+zUQ)))
zQ;3L4SfodLzN`#IDNFNpIsYTnN>Ye?WBSl)7h=Y^q0_CtGC4moC=GY)?BrvCLkE8*
zW8kTpe<RU{IE%YDCk;BD<xr=2cp3EED2<CmIQJq&%f+UD4D_o;-pvWp`%Y(|CaLd=
zxKoxVdS1``7uM<T*D(Ih@qel{{nt&ZaO7?*^J15*=Hc|A(+;~HKt+Adcx(m(bmP9?
zu*kIM6VCU#@mW<EcX3cGb(+{vQ!#@4=4+e*J)aVLGY;Hz;{3j~xA~JW!<wbI@rt(&
z>`;^bGjSP~6Q`0qUrJq!^ewx^P@&#$d!Ig(DoSq8^Db`0^Ej^slPy13M?QUYj53UQ
zk~c1gOR`6KJo9jHxNvUltewY&QaAM3T46^tOJGL=1p6)VtF^AFF{QysqDG#wnz^BT
zW-zQ9Knz3J9G^}OwOuQJXUjTLB4WFda|N)c|84xTg00;mJP{A(AwwJbq$@EY?v(ux
zd9imHd6ur{3k_teS@fq9sL9yN0RO9*)<PJ$Y8OnlBMfOF9~DGFsU>}eHcp!XZnV>#
zpEQ~&AgiiSzkos<@)FEi52pQ^@NwmMh2tYTm(Iy8rR7AcGX+IoYRy1LS3pEv(7E68
zub~M6VyYM3%o>j&wYFH!1~x&?mzWM;{Bd>QGHl=3ITD+V=})q3?#K5{SF3dal_%CF
zc$X|S&!<!#Ne>Ao0cIj`pVNmTPt^vX_vX`#J5)|8{KG&XFF8|%2~i@_{;<aD8Zs`h
z9bkMjIJ})TGB^Rn!d-sSZTyybU|gl$c)<)=gHt__#dZ)9n0m`A^bRM9%>=mly$F>%
zH^HQG@`HOWip@aK@|@=|yNb9T>;Da{T?~XYR*R5~hs%H|=gkq~1cOT6@?)=q+G?Bv
zh4hdO^)gEUIQ=)*_F(&jDnm#DRe`)vimjZCPG9&zGdx(5Tb&~gPR&t;o<J!b!s+Ou
zz53zes(PaZ+K3z(@9uxU{a_m4H;4g-7jjVk2Iy)8!q<Iq&9FnO*j;TPI-Vd#fcAEP
zXzUYWh%$Z7gZLV<M}v~`6GDe!W-?%4rZ#^==6V;zpk`?5b80<mJ+Pd73ZMT)m)f&H
z)NvH}bvzDKiSoLibO+}EV({;wkfvPPWzPpc>A*PwJwNGYLumN~YzhRm4~nZ&<y8#?
zLRB0u6@Xo)-W8{wilsWLueuSRBRf(#X~#8b(#QtgHsrzpv=Tyn3)~p9@cWIePOu;3
z|D<E`BZkm|PGL8w=ox+S)evH|ulHRL1$Zyc05@d4CH#{P{ih#&_a~hZ3E2s1rlQ(l
zxa~ChU(<yHt)hva!6~dj@Px(CWB^3kb3k=ii8HM;%Zh9^A^+AtcV5<Hmk+Msof@wS
z8-o6%yZVz(GJ*LK@=qh6K=;fy)8I%C>_#js6VSy2Fn9u_op&+j-bRCr??;(kdS%yY
zzjb=9P?V?&QMPgZqEGm|>8}B;W^XbQTCL=Lut~rLCz%MGqcC7fqOB~@Tb<ohy54<B
zAiCs6ZVUq~(TCw>WbLUHB_CipK*@A6xrnLplYRU~(e^+>{EIL$!>zqh7-Vy5@~}E#
zu~PsiMKR<Y3zTTKag(Swzt>EwC?^}`qhC<(WYIZqhiBi}j=--LEg7BqB$jMXf9NMj
zelXZJ5V)36ba+YcP$VW$sD0I1bOk~i;3?q7n^pQE>mZ=~T1N8x)=3s7nC7+~O(u<H
zb%iACder+abzvCUgtZvtN=TPrNq>mc@*Ln}Ri_>|W~$a}I{YXN-gWurgffeuCWY*p
zk+UIVL=O^xzCkvXhytW}pl~^^?$<5^{`_((e8)UfSIA2)o4OKjC1{VDP6LuFFOZ3u
z2j#)#&EhyJs^d_6mz5lyC7ehIZb0v^vVh_|WFbUd2#-a<lD|&cD6(kj*RFiE4lV(5
zNF-imblvslUXh)OzB8aYRlinjA~6n4J?CW#BOj2AQkt>vFo&w}QDmKbDqTQYgHtmt
zF_y-@TgqMszC}0<<bf!@s>tYSS0$#vW<vMdJml0z8n+<`jz_%T+q6A)U^I@Qp2JRe
zm`FJ#%|<R>yN47$2n(A+b2;=)vR_@YJK${=`rc?hoK{zY97>yxia8$;>^EMsURH+G
zS_ZNh_Q+sar}N&Bdv8&M7&;wa5B533ZYiy<9SJGq?v|qUi~2sJNSVD0qS)`4kg~~T
zDqlK^;>DinK*ZJ^_?;{xc6~T@Cg>s^9;*7Zoi7{>H2MKah<TT<+eSZNt|G%rWl4eH
zb~ros(CxnPaALR77d;E;gqQCJQE8R(^89N$SI-Ck?!F8YyQ%&<w1u5Kr$2uz@K+}e
z>ye{-02~;B&-$;t_`e{0{tE(f|GdthMfd0xW-v#WY9T!^-Tt((kayMt=7l+vXKF7j
zVb4gr**#3OSMbr;|IjBhs4uy$FBK7C|NQ1WsVNjeIMH*Z|FVu|talLV$qeS+;FJRQ
zg|PjZhRNcT>>Ph+`JhdI_|k^>2N5ZdgmO^8iR0Q1PSNZ80=MPlzAisL>Le6>FZSBv
zxMhIBJo+!F{i^_kQEfx><&v&3+&AAE_PBV2&A=!6AZ&TDAk1R{+R1ud=z_q}8nx>M
zl`EPrS7NHn`;D&d2FcW#K_)WG%>XFjC1w6#AUb=eUsQDACJ@HO9Q_CFkbk|7j^?r}
z97ko?!mZBr_^a|>%g7j+aG5G8Jc;NSEM>p{F}pqKOm|BnAl|CA@S5DdD^2zxG$O<r
zxA%FIf0Ruik64|XfszG{QcuMr&%jAt^g@m|)SvB+=>!a`1_aWFZ&lZ}#a}8NvD&0U
zua)V&#8iterS6N&K!-mNAkT5&?CIgxDrxog7AOlbco#-Gd`-D+Y<TQ8t|a;<Esg>k
zdU=Gt|Cj~#4VaS*Z9Ec{k!m{DGHz=-k5#GjPBjy{;lc-zuL?W^yjrJ>rn5HGGADuX
zjEZXNpc}u5G#l=%wkue4$+yX!Fn3-o?5acFz%Q|cK_0<EQZBN|)Z{L0K}fgh8i6c)
z7-Z^WY}q_<<4qcUmgTRaXR~TR8852ng3}yEkcFfU?I-2fdM_(B!g^9pOizxa5d=d_
zW#Q1L;XU^4NGU7KG#bFfgspNL8v_(pm#rku8rYEWCuUfN$_WlZCi^_byUS7enMZLu
z7$H2EkBr7i`uG+E`8@hbHy*EKoN<W8aragW=z6X8B7+h;*S0S34z&7CG>`w%zIpk^
zTiYP|fpL$RwDp9FcoO8I$UwCm#n@kr^g!5z9OvIRa4jeeaCJ7LOs>2pXbiJZv{cpi
zYT&p=Z$zx|;Oh)%&v>95GIU(~&9is3TiOH2IMfKQIL{(aSa$i~lZ-J**{%wjM=sxZ
zbsLixt2|`<DV+2GI<QfaEjK*g{$!%CkbTe*!b5$QrdWy%6@P9OW=p6wW}S9crEP>k
z2W6=rl;xjEQrv6j1M8ogX9|qO()VITr2(o$f>C@<nz7?MPfsPha~jaXe7hR{kaJB|
z*6X#$mfQte=~ijAT$dy89&du(3Uo+4bGcq~L^EyMD2(i-4h$UyH~<o`GwP@V4;V=X
zzXZgd=o1=r)x_Pekt<v%P*=6$*f})d<-E5)<UwYSGOZkw3H0;*65wFUvls4;X7+)N
z$b}x$sAKny(QMWt>2U7OkR9`c6U7|*iboc+n&&Js8&GdROtqKc97&Pr20pRQ&Etg@
zT59w;Q^qesTZ$tT_r4Zm$>!9k5tG%cmWGB1*cLb~h1dB}UE0&)(R=9ocM8<&L8rD6
z^ve$>zo*zWM8?@KIDE~Ur1v)=``COQ88Rc|b!#@?0V#Jmh308XJ1nYN#`8#sdGTA&
zdaIQ?QT~0&a!*X)GvuthI;)oUSQ1hip=Tffh*=5G$uN5IcL^H~<b7BluSZTjD%y23
z|KixmKbIqXB;bj!g&IJ~{ZZ_b@--CgfL#gmbv^WP7OjVqx&x&w>K}d<-(N)U{m8IC
zwg?<O`;#snaQ*@qpryAV0Jv!H$L9_);C!BgVc#)ewgVV!oXWHZ*I9vW#BNdKcgal*
zxJ@iP$#oO)2}G1IC{S{qroeVa8q<**n#MdbSBZ_-SWYFQ)sfE?BIxki)5YCrLL~Tx
z$M{tn3mWiL7@xGB^WM-@1J&Cp&<ycRRlE<TQstJok<k%_(gNu@SGUyD^sMBv0`X|Z
z8*ar7gXO?(OcDP`i)&e)EN{j|-XO@!09BZS-Qh!cZyK(m5jj2WuxGwi*!@7Fu8*99
z>D~m#e6Cf+0&6A8`!`!tZ!aTo%2VhfJYfJbm4?3nyz$Ad5Q;a9D6`T#6+y#LaWc!j
zhq|T<CR_nk2iF1`JH`4Ifj;>ype!?$nj&6J-_rZU89T;IlWAFom2EWm4cf3Tj|#1E
z%dGSe7j9*fvYH0~AHErUv95W#Y9Exz)Q0sL2+U+ZJ`$8sIj{ZgEVA@6Y6bSrtb9$G
zbk4+x&HE)m3GglD3XCWOY=p;!a1`sLR%WrL)c}Z*R74ZIEYc(UtUc6GaJ|oAB?Bl|
zD?d7oHY-FSEc*(O$<Idi)`W<oI3Ef!H1(dx=8slt?xy{oepNp@V4$NIn_D4cPJ$%+
zZ7Ye@H=B_YA-^RoAX}!<Y=iKzF9`+M*}z2jWVsiPPsU_vug@tD5@hYYwI8;XY(rzX
zlSysO^j3SSu4y%U7Rh8>@1H^ki8Bq+sRC@Ev9riey2{|%;E<4tGkk%Yl4x@JPdaUZ
zmtO*+vJM9mJ_F$G<@ODrY!Dgmpy(W#pCd=7$pgSI0}?0~skZ$zDWHs8@OAMZtZSKO
z^1aEMfEC_`yiSoOfltOwC*LKP-(m8>^gzgh8@Nd>>Mdezu^l1*<lb*=iq(vsRbH-c
zQE<Yyt*b3h6_ipKZbll+6kBvYIT+j7yk{?4CJY`Zr$?%+_+I~vP3-P@`5$R>{~vzN
zDX}MnD$Am40IM1V`KcX|fZ7W^2wPgrUmA>4e)G;C>io-U)>j*vmUg?S6Z7W0AWFN%
z;e4EOyqDN4ZF3u=o9x}+XK#J|+q;zzawIWRm;8!HTOc5brgY*P|MP8u8n5$4n_)4<
zmu@|||HJ(j^vsgcE{9&;{kz||xVp1tP9_lYMNF%|l@?b=nZMzE61{L7gfOehHceK6
zek(htT#?9d+6E%I@D0OD4j`)Q0lF)h1LBCpsi5$N#j*A(h~GqU>^n0(GiG!1lu=cI
zmcnd@hdH-m!>mewaW4HK1;k(bJpaAV0zh@PK{%^bu=I-SmDUO_MKs}o2GfUb&Ls}T
z1|AE1VaDpDtI`5-H`6-}CmVR9$3=Q&XkaEf&`4mt+{Ow@5Pu<eP~M<A@Cmz_XB*oU
z-42pq`MlfdXCdKpwwZ%KIWIQ9{8TZiO-#QsMrinnE1HGIvxH<>Z8H=9;Bm2`oOz%&
z<z?mapLA7pr~inz=a1;b{(>O=C!f#%?e<XuO>F64J#)%$QAuf9{JmT3%Lnuq{qT;e
zblwT8-Ji+1+C3XESiGK#x$XUX;D_HgedcS*Vrrepb0o$lUl1zqVPF22vJ#`CeARhl
zdyRAfqN;Xed&0ri-Qo}Ws!^L7IE~z)XtQ_!8l82-Yy}#;=)WknEW)vc>s4dC>xEKj
zbayxUk+0|{X<-n>uaP2)%YLvf6gSchXRjyxMn6&HLRI7yWqkOP&K;l1WMG$GBH=!%
zax^T7&;>VYc%!Of>6n@GFLWRGfOz9C%2>ZX=PW1Xr-Wp72V_~QIyf*S#5jHwzRst4
zE#G*do5~(Lqn0`iIkm*k+~;zo)Y*XB;qfSM+VPi}L-R%1V2ui9kt#mpMg&<Hk3$}Z
zpL49`EBZ#dQMPUqZ=@D=1GWB@0(8V<o=+0>LBUw=oyYFx<*~S=v1V4>(^Ek;<taPl
zP<`_;ZHH$Uh#EgSc~5Gdoyz7rB{P5H-i8~}9<`KeX{+jfDLPo?*jx2{QGXzBcAQ8a
z_9G=sF-H&B!6R+d5yci=%^MNP=ueo;m4S$NF^?B~Zvk*Gqu;HE`i$!N=a#FA6312y
z#kO;KiK?BECHu>^K_;}%`Fc8XobG}J3#SYGnTT?4Oq2F-yX(#5cg{4!dD_LpH-UWc
zh^~0l<MHY;1PSwL5d&vQwg(VxkFJ6~$mcwKIuHnO0-_>!y2*mNvF_TADVO3qknZ)o
z`$=0HXOVTcsCIDls<RqiPq})}q!^YeZQN%~-U8!B_u8~6Jy0r|;4>hp#H>vz+jf|X
zs-f^w^@xjZU=-6da$yi5wa>6c0e}N3(tVckNJgjj@?1uBk$nQzmOaszP3YQBI*q$A
zG!n9MEbPDn61xAIb7S$YRRV0@XF%TNb@H?<_8dun1D8~m;yLb^;JMJNVTw%FhnvQ3
zs9ZiY!m85*MJfKjl^4h22EFd*y=sFMzr1q6IA5LURW~w-KAup^(YLnXequ)B{KGw9
z<b?dwWjWJk<@Qku#75OTC-HC87Z2$txDI;#YYJg`azE*c&^Tw+aLv>l3BWD1R^QHJ
zoE*xXZXEP9)#jw+(Qjj1(4mP2wTf=At277R=`sBg4QhQ9LE|UgK-&^+h4n!}u0PuI
zU@kEBCav^9Fe~tT;K}OwTj5!daR3CBZ-NKX*OLsg_Y`YJ(7Wk^2ga$Ovqe1U^Xb^N
zK9<!R?w;BkU<T+}r^3z33?LI*9s(p=E9fIjDTp24Mj+Wb>IoUF?XLiwPxqw|uv+yB
zZCA(60!VHx-0(-S)u$b14dL(3xbS9^3}N!vP04s2l-VY19u;dtP9|gpu2*<M)^d@N
zW~dpKh_q>sl~lm<mf5bJB<k|PAyp*9xBNsbdB6*}{!{N*vcGvID!=L{U8M0rPkrx1
z9)!sGlddK9Lb=rNUF~^n$9^A;6B>zJgm?hHJkL||@XRy7MLW@bxC2uG04YG6d&|2@
z2S^b+0K-5~Z@p@SY|A=k(#9vLj>x=^zzK7!5F$Nr%_O9QJ-Uqss0j*o>yUDry790|
zAcYutcalr#(YVJ7hO&z90kY4>r$~2J2;<Vo(O*AH154CpVMB6;y945b^hi?Lq208`
zni1$1nrc?l$J3FD*#L&(-y_#3&fbYzNpPTzfJiy#yhmYxyJ?zdC3jd3?2250U*lc*
z5Ryp9WxGWHeh*JCkS~7uY>Z`R;%X3fT<ts(XaR>#=YrnB_E`_(-uL$!Om<?nz;dw-
zz9ncs!y3pg5y(j2FUv;?PGBn$dzm(e@HH|J<?pKSd)AtGjYgZ50h5enX4yGtY)4kk
zBDK0Kyl`JfG7c|yrp4Gy4wrY3xiMm(4U2n|8ye7^EUsWmGsFtMtvg9Oy)l(#!hx;!
z_vpYLI!zWH1_y%7W;&vLkP}E+sT;K6Nfd})3Ro@LQFkI^2d-{Gq)vA4I<9adDk}~c
z4t0wUQD7ePBUPG?ER^4nLC!;g9g}BD#(d0}fc&f_D|8zJ&I}!XL>AIHY`#0)@xZqg
z6bpBFI_Y_uckxU92pM5qBVXRTaSgKF#UXRBXGaXFGrD++D)prdf_3>)ChE^SO*!|$
z`Aq%TFK`bXMl3VT06RCqP6fitV29ltV4J-vC0ow$cKCof2h!;)hif5m9+jX~EJ*;y
z;7o<wxKtUC&IQEOm_ZX(XX7Wrm-*jF=yO|E7D6JucdG9<wVFloiwu<0)NSRRV!Yzi
zXzKI9ag<}!2em_ZjuDR%(Sh2bjOPbHH>n`)bbDS`jtA&$M(X-yRpbiHSI>$W$tpE6
z$&>~H_UQ))cS)x6d?ScoF%@b3*P2@C5q)-;iqY{|q2yy*@XqP9dhD6)d7~I^7cP&n
z&h8bM8gLC4v2F<vzc5S`B2dPy%a8=rqkK~g(`g-$GbL|j7g^hIRT2e#+GMZkWL5*y
z#j<BJUl!xA?+b#%h;VzZ4s}wlmQ{Z84_r@X-o=C#?DPi_bHB9NkvbO8X4h5NtVrHp
zDW7*d`rsj>%=5d<Qj^nvrp)_4`rL@r;8QJoKr4(SqU6*<JC0Jmy$V0M+yqw}z0A5X
zQl(`Zot_(VANM0{CG*WYY2P%3JIYb~VUxAQ#fN5NQ#&JYHn@-ISYD1^Zf}rD|L}YH
zZ|gqvYbP=QiNc;wv{qv{m+rn12iO|?-4&isHihSnkinfEtq}=!uIwS#Zyh5WTjF0z
zDhm}!m<-2xedT7jMY*3n;^M`={gtY3+zkM-&MMgfSqdk(N&c0F&N)UQg*#jjZ_ln<
z_#fdhX#(@j($jjnyNk@O;Q(L7=l?tvNyf2fN1rQf@B>+mJ7l}*57@KbsQcl%eWJ#0
ztparB!*`YgatV{bR;#n1@h=`=hPRn{HcU&6a2EF>ISKT>`xF{OD=B5G0mOd8|D27N
z<l_V9oYrSYD;WA>zR!A1F_1Z`+lE4gTpLDi8dX7jk$`0%jh#+dJMCR*ADvZ(E^t0R
zG$|hUZSR!kK9=Et4V#2uBm>3BI@;yFon%RZw)u)s{Nvfsu&);vxcN}fCTyzsa3A3v
znw!Qze$cqbdV@15E=l_t&$jX6ZECJBd$JDyY?o3MKMLS2wtoLdfrg&W2VdN!2>aXV
zSpdynEt0s`z`|uoHH#UUAJj~XtnC)fb24O?N}fpE%gfZelYa0n&Fx36tYhVT#?*dr
z_|xU~j<0H1>F`PVd?PY%2028N_zh(yA5`nj-eu{wGEpP=Rld1A<9toh0uU}{0MXy?
zgz^6t6!PzR`F}3G{Id%4|3#I^5F<TmdAq^oVU*3ejk`0qOuX3b!|1y4rI1xij?L}g
zdf#7mIQ8nvlskrje2X+`MZ!d8_PV&w15A<SiLc&smThPqFvr>I=cT?Q-r+u3RDwRe
zWX8ACIT!iSbB`0W!hN1;D~S&Oxw?5Kq45jZ`t$eQABmWCt|!O+Q+?TVoTyjGVUui1
z(n+sIWO6ZW!;1}G=Z}o_{6$jv&(Tz?fT{7B)F21T$2WCi&{g1pY*h<~>;5-%1S?uC
z7xy@-*CuLn4Rr4;yt2ewl^V@xmsfPn(j7#JRB~Y!)O67{sLy3C)weF&){lnl>Dab1
zy)8Twk-22~HJh7yqp>l+>EAF%V-+D&1xN<*QY_0NShD1$CZVsc{KLbpMw^f5HmvH3
zNFzUIpOQnW<>EnSy4OfRFV}3C>hZH`5XH81<#Y?R<mB%8lsJQ8x|5pnN3mvsfO}VW
z#b1OVYGjUOT#0%7tZgoWDfo89sw6y%%nQ7CtBC42%SWTeq!XjwQM==~&l(vvEjt!(
z;q3>d+Cz)GZ&n~vK<)e#g2qnk_AFP-lQeTu9Q_a~n$Bf=zuAEwkT){!6P=10JT)D}
zA)j{tP1sGkV{{Kz24DMJj!_t``UVW~(Z8~xZ%?qB>sKoVyxtSxe%F*ycIHd9Umfzo
zy8oRSufFwsGZFQx0*2he!{0j{grmbXcX)J`Jt3S*c3IvcIIL6OTJ^kJX?k5lU~no~
zZU)}X<L!}wEsqwv?EFO^D}u6Fu`EmbGU_ylarR*AJmz6^fyWX8E$dy=pBQ8hX6Ir^
zJ6~bf-5GFnd64gFTs2-|OzC>xr;#GomXluqVuL$>>Ms`hwFtidM=+0nea=4{$o@+W
zWPjdp|N7Np;AY<Zw9582ZX*42^4ix6hkGGfb8DlJRrHbx(AEkbKA+}?e+CFgG`fF_
zK893W%?x|z9{)xEOG@*}C(a{_7VjUd(LL0Rf`23@5{CUbr*9?SJ!NJ}BRqVX$p<a<
z6<r81b^Rf$zQ=6s#N!G*SLfwY^c(KP6^mE<c-N0ks+&^0@-<L-vK$G(i4d40n@97i
z<6m(l-Sm_V==9ZbcpS*zN0oc*?7s30ar#lCdbR@Q7CD3TiOdKq>YFOt>SojBni2N?
zPIcV!lNXfbnD^z+Rdx7Z?0t7!Q*E|x_&`yr0--|`5KsgZY0`p)E+SoNQR&j8BPBr;
zkS-wd2?|J+8j&V7(nX|~NGCz1OG4d*B;4(sJLk-NGiT1sobR4{?z#NIkL*p@*?IT#
zzUx`fTJKr_>ZZXlr|Ps=&q|&9^<j$5(ASn?zFwFA5jj)0wCdp|*lG$w4hFrX@?haw
zLPCcgD`zuDT&|1RQmHa)>ME_b4DBX5+=X-rctt?B>_fO4Bs_65oQs}IX?7pTrLC{2
z>k6i7_@<*ofAaA@y(X-?lq@U1PJwfh{N2ju!ygA&_1kOFQe}eI&7&F|N*$257wR;6
zF~wlQgJ)%pxlzga?AWBw>abo+);?q`5*LRBg_VbD#|Wi+^uw;MiBh_oGuVXQhMa9M
z*cTQUK)a2si@XVX&e54|1C;LNa3NWerq_ivgozmtD3$SUGO9Zr?4%@7`Lle*Z<5i=
zRqXKwz|{u0cOm6PSwjhOH8Ab@uIU$dH*8YyJs__*Ku}xKXHUOad}_Of9(Ku$dW>w7
zjE9bFmG;ym%i4AE*r;E=&ylvtY}-mgJrRzQW3;cm`BM8~<Yfs?N{RGg^Oh}ffUC^_
zpTKnM9-C#io5UNY3aO(fLU1cCca;@wUwzj$wD)M3^>QX&ecDmp5&a~G#^g3bi0kv`
zcHta`|JLgIAGN6mMLz%HVgL5w|JNLM5aEw~(A+nx@(Y~>KHJ^qyrH`no!;6c_>Rz&
z3JrgJSHg+JFa&P$7w(*?3<IuoM&-6Txvva;O?TR@ajXh*NO{9-e7&arxU7&J%W+aH
zRu0b^t|mYft#K}z`trKICK-t%BRlG>FRx3&^xoHbR`WT%_$Uw-3+}0c)Mx}qw@Qx-
z&39y63}N0}Cv;03DNw&Y!gHdg0J>)KSC>c=uZd0>T))C-Zg$25O@c3OA@{rRTauan
zXtJkElCjr5<bg|B?(%5eohL8-hkKeYgm;3qDiPj!6H_5yo426JzBsJaua@^W<))af
zoob4>&dfb*l5E-S{w0wInVnOXyH@^_?f(rk_&zgwS}h$%jS-lX&&TDSy5jwKM~*Ei
zx}FaGdcyo^@KpdO{g`|$?W&nnJx}nM2+yZX-6bjqut8;CV+kqJ81rR(;FVyE6{_4?
zqgkMej7WN{m73jJ-KbHeR=q?UogT865L@G|yPm^YkvB1|@!7kB8!1%blS#WUF7;Y<
ztLmAH_XT$uC=X0`B7%DrcXRH>lN^TPe&c0J8|)ubZ2xf&;ZHn>gT8rxC6WF2P&lB_
z_7Zc6>Z?quOBSc|N8d2<^|5R@@lk&nX#whS9m#m-oDj2SivVKO<+v{b6^l0IhTRO;
z@~p4WS{t@0TijddYAsTPPdaZm;3WdgU$Q{e(Zc;ycxPe2CpUo=3roLe?R1Guf-FbX
zMmC}Ik-|^ekB+120?jj93F+GX!PiK`yk4DwUcv2MRvG#Ass)uUi7(dIe|QX7?L*=l
zOt~HfUwBa1)!l)qN31F2fph67@+_uJ<2g@>#WY^fLc96&YVxvWU82#qphQQPjKqIV
zEci{GQ%$!wJ+ZUw(d%n5ncR6mh&OXeV`{pmSWnZn@Np%4{S3o$A>M#>?B4mry@ioq
z1LZG5lhU5OG+XfVtNkuEM?eW)%Hca)UAdw1!oo0>IitZ3ZCzIrX;|_6=wU$ga&BND
zv=`lqU`=4_Ld!*~<p_Jo@Ke^T%nfyqr(u-&YOb)isxc3$2-9li-|D3CXgpz9+mv6u
zkX`xrly&`&Hpd@w1sthrpiAOS8kPIE*H|@S!1WDTd*MxXG?C?P_P<ld=it2f9}B+x
zF2D2j*bMlzgw7_uZIxgOQ)O0*hJ3GtM_RB(Y!$yUHYnE%D(v8X!wtJMUQU$_SVBPn
zQT+Dc=~E5Qc5ghpJaIGj>s*z874y_fuI;r$ZF;g&=RVGzh=#1AL*eXI2~WnY+tOaW
zP59Un722_3Q}tA7)Grb@rQGVbdN|!#Aor|4N9=8f2Jh`>M6noIMOWD6h$TMl^x&Es
zoAc#atiIXrF^?_s(WL=G9ecM<OkJyww3?f)vyAgMs2Ub7Px~SM<LkGR4z0Wp%)djo
zXhp(Bt&}1J@fA|RaHiVxNfA|VN`;MX05{VNyp!q%en>u;XbjOXYPLDjIa*&-=$UUh
zIqwCTnHF%V=59*<-`@L=2F31VM#nFs*S6nH$ctQ#oUXrP#17nwiLp%lbLYJvq~wdU
zmxjR9q}kfZ7O)ZS$csl^v`?O}$&ep!_T5VlrL-!3?8~zQDoknzeE8}ow1XF4Uv5@)
zX_FNQ>cAq{GcK}y$B#~%nQQDl^bU1dI4x}X<d%q44wFn;?}hECiOXr<R9t4)+O3V-
zZG@UChw&a#25F)u%Ekf}Hl=u0)>{GT1}@4#!<VC7P|=jvizI&6=kkdY68tsK-**)c
zUzXGy^k)xTFdURs7C1cMRw9{G6NP%8#lE#wJZxyO`WDpFewS>DgXSB*4UPZa$BEC*
zHA4~tEyQWLEC|oj%PhN?M=HH?66%&1sHUG5U`(!LMi&*OYkYSSK|x9feKor#+*xDB
zi{Dty2fz}Qc0QaRYt3)dpNy2b(4kwjJ)h?bXK66@pMx<?EcyFs_I!m&S*qQp?o<W3
z*W%p7+Ma(GxZc9|v5)Ynw&>_AAiJzsx=DE3e5^(53+p9o6XMa{bw5q7h%_D!^T)vm
zqIoRXE0JwbLQ%@K<koZL`>i|P$)Vl}f!kA(W$F4=F5gTV&Oeett4|TF?-v!xr(RLu
z66iKDwG1)pU0QN+a<@&wvquTjKU3SfSN%s0_jm5GGnu(r9%B^UT$W?Gc<J<EqyufK
z=*po_ii;URqaINrjggm{LKJC(%af0}zd#EB?08lV7o-R#-Goa)DfnT9N~WjPeMm{3
z@pE1INoQ-TP1b0tY!d+<VMFb}_64s>1!^s5uie5NqMoz{=Cs~Zhm2|l;D$V=r<Wue
zhbt$nZNP4mxAoHfI~Kn3&~wxw(_Qi=N;v(96PFv=&Og*TTf2+S`cY~t+%`==K(?04
zt$S6Iv0ZR}BYwAeyR_Kyrt9sSZa8frn<(B;dYsXKg;(wmuzJzotPlHlj*JH6pqc8Q
zXRQ3*)AFCW-V53DlP7g-Z-H65jG-E!33m0&6=#z(xikPu3%e6@{>^=y0rl<MB(Py&
z(-$SFkz`w;dAeBh0iPR({Ud#<2V|A`xDxYS0#cY_y>7;DeIGAaJ)Z4Ey@+C&Y(MFn
zc$e*}f_=iJ=&g(f^IvU{eN<R2F;?&-;+$2d%;!%%Pd~G0KmuK5qF(E894nz^ZlzBP
z(eUa4vyC2k7qR6G&qeR1erOUWu3|e3%BLyW%t|f^GTIJ@QfwDP0~sJBWOT&(;#t@E
z_sKsl#|ei(dLCfIs45F;@<du^&L|Q8OVM$iC+0A^kj=W&JYj_*vn;v8o>N!Ow|RBU
zQ=T6VJ@s2sm#@QcE%&F@jyk1n?p>F)WgD&Se)m0H`x#w&dRy9rPoWpe8h)B{>dL6b
zXpE;@a!nxE?_o3*Pv7RIfa^`Gse6u(Ybn~v1w4stop`ed(yPaA{CX#!6!SmzOk0c)
z^4z#^|5~<5Y}gUmls4)RZ$xeOBB#+p(9=)TRvEf%V`8*<F2N_uvA*zMEPG0>h$t+5
z^ZVp=OY_{w1k*M9oc^vMr&b`G=zSeP`9e!S%jkIuc^FpC@pM|O$XXjo+Ui#~g+|2N
zvz+su$n%~#rp7TAFuFZzeO1=OT|Ta{!#jGqztR%LvUq=SJN@zTkf~0wv}u*z>w9dg
zwjJaLgdi)d+6hLA($~`TyGDH{cP&*yB___=CNV17rqq0lrtz>j6L`8KEym54-JiEy
z#Pyb;X@Z$4B(vj{1~lvUGU0zrm^?7$|ID2JC;#V&)Y;NC`4_YCcMHzgsed6k6i+*C
z@954uj2%(DvuwUl6#LTm8W?$6FCdH0@JY983_GUiKTD3vaLV8N#CrKqP|8+AntjPT
z-B>n`vzGj~nXulbYbDvkCEu%FI-Gu*LxZKeJ-WGE=Pz`_NaX@3)bjcmaNX<Fi*ktH
zY#O$HTqszJRxg1cLH3g9!^o!c#kL|n?44D2Q+A-P^Cw2J;4X>fYlMsfwO4hP=u^+-
z5Dnb%N6Lx<Qan)J7mY*ckA;Y%Ie|lO0y-^5su)9VejXitQi8`_#~X0dVT*xOtC*H?
zxzRkI*t<1GD~(6neOiiMK6v8k{ADn^HsKIibD-e?Q(V45(@o;{tCv2_PuDs4#_1`a
z5vPhOLQCv&%c3}=H4+ied_9+Pc`bhKoc<G8C%>V*LuOsXmQqc-J4iEZBq8y}{^YZD
zp6lQ?jZPGYtViTh402`1=ybs9@a^}4n|Y_8etfS&xW%%HlmtsT8*|5JTI4(LN8@z5
zczD_a5<1ZA0Tueh@ak9jI^s1+=g*vF(0tHw6#8;WT<ce7>n}Ug1<?7x`A$~Wg*uY6
z^gQ1~u~Z3JyE7Rt1<6>qY-^ndxnn(N-8f3w*36b)GGU>3>99apw8{<Nm^pDA%9qit
zoAzaz@w~Gavt{9eM72}rMNNUmBWN1JS__I}jI0xKQ=9YC#TSp0boJb9`)OBgcIhNM
z7_Pn;e?+SaYU!mWg@xl3x;(dswIq}KtpW`}LQz5nKeEH2r+)ucfaXA;{%;~s|7BF<
zUxe|~O%^|<Djl&LOX9y&=iN=q$n<qQDIw6N&kkLV;D+}$41It>HZvoqML>#;*NL(^
z^pr()c|<&m3C@3O`kvbHky0>w51^z?4W8YH1jI|UVZfY{NHs}9X0%j3_G+<PjYN`e
zl2&5~y9UG?!6xU@$oPTOxb?*L0h4kz!^NpY8@)Z%;2T7-xT@|6eY<4-Gul%xu}xjP
z*TO|Kd{j*%F1?YgoX}H{_HbchnX*u9N#13E%u_`JvONw1(uN~hrp|7|Q0AeeLl1U7
zc*KsUwk=P*_qZK@o+0^JUuh7On(0ww(O5OCr^33E*m0?dSNid0CaaD}n(s5SQZXjm
z8yD1hU{Dov&sogjsio>SXju_R=};Q3iH3Y<xJgVuIF-=Jm6TQXnS@2(<8`Ypy*G-z
zUqVnYx%2d1MglHGy7gOpraiRKo%MN^L+C|x?DPq+gZgZG`l?8xU+npWYrJ#&5Jv<4
zfP{9!l0JI3#N7GG$8*Z_(3PP}&^G&1arRe6tS`Yz<)?cY4WFEdKfVvS)-S!D#X>)q
z8sUq82FU2?^JB($?C6$U7`J^0ESOnI*o{XP6VybZk2#(qw{07Pm^o-Zl68O}XCmKK
zUf+pICqBP#%UZ1o5hPTrwUniRnvO`k`34NBwk=VX9m+dU^6TX5rX1NuB~LvKh>W(8
z{rD1h(5N*w5}zyz1_As^10}SFpQ&+*aqL{=D;33KMe@H<*GHr}>a%I@YibzP4XpmC
zohWbA@N0z*I<*h2c$t`$Vk5sl=8RnvO!<{q=@tnV4L0|21alR-@@x1&eKYJgf&Bke
z(~kpT{r`}#{x`+>SAk#$IOdxpIw#e|$_{x`l?bK<2`X^46#ui&iMr?PBKacct)Uz=
z-)5MQbfBeSMpc?IF0OGd6q-L{?oP-v1isBSx|k4Lp9gYyS?ix}Epqgu7v@GIva3wR
zxYr9_s}$|1g#d?e(rgQ)!?|_eUQJ%$=}Zx5nGK&9FAH|X%At+|`ei;d@{EmhOgS$d
z({}Df;CrWuetmS_=p8WBzg24cBq{QwO7xBj*K9&e6`VW39lvbiwV`z}mVGh0mgtQ?
z)Ypn(ZAUqJFn?Y>liyWQL%(SAxlZ8mJu?aIEF&*7kt1;3ZmbG@Za~=Nl?SfxvM$op
zw!86P{`zf6FQP`J6Q>pbqv>RTD=2o9Tf9%nVLYlhw`@W^7p&a~c;t;+-?F&mGo&jl
zaZ!=q7vHF?2cVpBd&X}0;StR3)im+?Cg8O>tJw0kO8gC9PriOKJNl{U+1b28{!-;H
zU%`#bE%Ti=$E!!mUdK;6>(TPx8b9YbT|1K?cxJTWbk3r;Q-z9A&jpcidPQ2HMhOD~
zIq=lm>dMM<>W#<EqkY?RjYDgWiM)RBiwQmKm7Alu*Nl;$e|cxH(CC%`s(ExrDMx$f
zl|U|k+8oA7nrWGOB3WCXlP_w3+^p2X!j*r)rk&rgrT%W5E#gNO#JE{vDB3+!_97pP
zi}K6PR%`f;!C=0Ev3q*&@YCl7FZM!8Fi*hZ7iXy=)?gVWU&*)CMwmKU5%!F&sSOR|
zFwbv63_o@*1S@B?;$8_dD@{#n-{CexR=xgKw)%~u`uzJ)eWmE!Br~u*((j{E{5A9*
zC>Q_zl#74T%lYN!|7E?eh*<O#I=r)HJfJUMrGCdcz{l&3dkC|g)y<xJsDQ_;7OSUA
zVuYT4IKF0mzpYn!$zideyw%WJb@UP4*!cv>#$v%`jrTQQrNA0G4)w2X*q83&2QEh&
zF}ytfNXJ^_KMX{182(!|hJWYvL65>e34;HgcK@Hb7JI42sh1Z*fRv`}{OmM0*TX((
z(2_`HU9bo;xIbO4_f9J9KyhI7xLlEz!<A@9G5Hu@miy_QMI&NybBgvSC=#`xu%w+U
z6Q-6jz;D$wcyegv{-(L1+Cw+#NkQ|ysu!NP-MPEACx^Dm-r!Gnm&Q)=%DgG)W_nsJ
zqIm_}BsUcF^v@m6--K=dGedQdjr>Qs>3`a9unMQl^zzyX`jLQ=7&Obr+H=*oMQbRu
zQsO$F48O5rR94dA*4K!ZoU&;r+xOP*PI#`}GPe<kTfgI0(7#y*+|8y+GQCphpLZ%w
zwJ>cj(-ZwhB^3FN@tl^(w}1PeF}MeX2>)v`%Kklv{coMOdzy<0fi`7VNU12(J^r;T
zHnZc8k?WVnlV8g!gJP^+)%T;06^TG+V_|_mmhwhZJ$wqhAe}&JDd^0R7rKU4JU7kS
zv61!(l8hd(DOHy<PBw=pgi_SDCU_GwTRvM^OxpiPeWU-@J^r=7NjSE`iA06H%+d1V
ze|=d?gtgjh;k7itog+_t!8FPL(C+>ow)@~3|7XVeh`Orgy_C^J58otXm%!*o4HG;3
zk6*!jiu78-X8G3+|LE@LM^CsaK}pd+&Ro15Z{M0-Gi+{m-Uz{&e?4u%lfzu=XO3dm
zlV@@%o;j128#VX0pc4#)jL%)|*>N1l%)`!wER=SE4B;PYdwoFF{xkgJzv_RZT>^e^
zv7NYMG;n?U3zH>=CFi^opW~itvEcmmjFBfkBB%!spkqcd5k}m0fxJr*qMidF?a#ko
zX>RKKdSdbylLsDO;#1;$Ur$<k<@6v7tqE<HLtO994!qKkIBgupkm;#oO$W~WzbsAg
zkGd+=!Qe(-7(1$#;==nUuQQL6<xP+xJ|>150g*R<yH(-f`1jrV@n1NeG-k{CHFe+S
z_%Myr@5+J|HLpo$I+T(moXlRz+r7{COyv#UZd;3;8)iAv=&y0b7c#<E$~(R+=03Uw
z9b>W>*9PsamycObXQbV=UuW)@bd}`S;O0EaB;LuLRDF>gWBGJI1NyUC&41{{Gb32R
zGIRqZbA7A6NN_Xz0!1Pp4TF-e<Y%^k4u}`;fo1zYW7+=MafSXUIRREdxEY#h(OE~B
zV7VQfpjj3cJq~qbej#-lY~9M~ZF!zevwJQ64DXLyrcLXQwxF*(vnD?4iLZazfle@4
z?19pI2#ZbhP;I}|$*ybKIhep+r)hXY3C7Sm@uef7LOyNM4diM^{u<}{d(Iv3vj0Ch
zr(BBFolxQFR~it<$%Sex|LPai7{0<?MOi4x)~$80?u<mh=u1C4MN!Y>FMS@bh)-4T
z{b6<EDN5zluj<g#q(udoTX~{B$4;qrp44#f!C`j4iBYypL<fG{=^A?<WNzYe+46z{
zXSA%*yGlnUhRi_SFaKU!=z~_!zs+>|f93d1UUjs`)|1jLs3gaa(GRwh0e(C(ckjln
z4Ziy!tM^9c@@}+psqZuMYo6>0FVe#t89teZf#Db3Rfmd*If~uAk#tYoN4u`LSk%gP
z*?as?Nn;9>G{j38<5<4)<NOo0@5LsnNLF7J58*bqa^!LSJnrIIX{d4Y@3EJ^{L!ZV
z2~9hoZ2yJS>91}}`@i`4G#t;%uN`q3rb06nBHg}rUCRmawx~2qEEHd9!)MGZC`39K
zya#iibaSL$d9qH&&W+cf(Orx`;}UXhO7{NeyudKE?4)`f*V**7HW2cf9;}vKNqJBA
ze9E4H&^>imK%xJ;Q|XHa1)=r$BSl_|T<1|V9(g{Rm8U9%Q?^eV8YFQgtev*7cL_=Y
z`Pv^cln&zD|2N{?zw>88`JZYj%xHMLDHJbw%ah8>Tsd}zL7~h=ZLZa;;{tmWYk}^<
zLg}1MZs1cKr>FPLzQb4AW<q-|4LxcuTPoJepU_Y0FAMRsHU!%g%3Sc-<M)LQ^x@uH
ziYnNvMdlf%x7HjFGktDwXS7cENr6<Zq7ILV$>6O%PXD&<tv4>OWw(E}U;3NvLJnf2
z|DkvFuh!y^R^xyA@1*C^5Esq3kxS$=CFA-fGl3`5YBh|W<y1gpomCXry^>^t?iO|7
zy3Cci&y~eTPb)?zE*Xei*VX6!4xjBCEbelY4YBq3RBv(lbzP{7X-uk#R)v-VZp*&@
zA~(Ly;%-ay*pV#KLxZxwD%ITDIAKB4pS$<j{vpf#Km8f}Kl1(P&afJ$q2)6@Ja^{k
zH;ZMBja+Wad?##GW_q8`biWtd#GB~-t*f5$Ln1r7CN3~>NlWCq=A2et&DVS!wNK9U
zb)wp)E9kPepU=@zTaO@tV5;@K8YI_`TPSOBlN!JCh%I49^Gr#GK()QH-f7b4GL*hn
zt><MAv`Q?(j_0(v+LxlftSPj{6fpX!N?|zO{7Y|kL0U%;3d6@b`ua<2QO!4tLa_=F
zUGeNU|2mNmXy0F{TJ(RBe*Jf&g9rEXKM)_!M#Eui;_z^_=-VdAdR56K5}6hyR9Llp
zdg`LZGt;7EOPL4ewrmDWoe4>1eT}9l=1JDN1+6)2NoY}<iM7txsII0+A53%o;*F+s
zUzjL8?V@#xn8nE5`4PskoJglJaFhJc-X1_YrEl5tJBRz50zz&)M_iOCL*RY!!l?O2
zagPhfCX=GY%h*AC_xtY3KDge&cphNj00RdYIKaRG1`aTAfPn)H9AMx80|yv5z`y|p
z4lr<lfddR2VBi1)2N*cOzySshFmQl@0}LEs-~a;$7&ySd0S5jI1GJxlAM6<ip9O5u
zYB6{e_bq4MY^JKp<{j29eQyk53iiuBWa0H9X6JJzSs!^1)vT-b3t5LavwFI42`NCh
zUR&b)`B%MSWvwK6(s9HYRI4}{tcsTHg7+YEbx}EJ_DaF>NCO0EYP4YbJ|wtnAJR7P
z7iX|I1F}p)2Z|)x^s^UNa3;>i7w^IBqQ;1I-?a8?S=JdT<+E>pHO}!|^n;?ks-jRR
z{51w}Pk4yhys<g6_uYUhn(*g`7@7ul0-|#t!gkiKkx=t^D-Jn$q3ZkF>IbF<uZ8O1
zfISTCD6**^40e4D+l9A;!hJEgI<-Ulkbx^hsI?bxAZ-(@?gnQkpC@{>=GYM8(JVE&
zFDz($A<F*fX&d#$G^!s4_Y4W{f*g<|Do`1}f(3#oV3*D?vjA8doE2sN>n57r30udR
zE0<o2>_#1p){k`HbDpO^9IDp(_oA?j3-8~_N=xR87)0;{L`+<cU){cprD^i(xPOk>
z-*gvjfP1JyB*6c}O~rkPK8nIQOtbAsR?zx@A&CciP&UN&A?#YeIY#s$9+7$QuqV7N
zi)ws>Xx~QtrynI<ZDq7gd=&*ET$=VF5jTLoeaKa88(8p8I+-j^yJJb^%b7(`I8+s=
zyI_OhsGQ#%qgyh6JERG}J@Hq@_y>{)tmg{rkgr1=2I8^S@-xff?>&;2pGxe4eQF~N
z|MzXo{JaZ^O>gQ^c0zgwhS{URCfFPoV&1xZQPB7v=jt&eSH2VN=BK|mk1)AFG~#yA
z%}rXlUAaA15OT)B8Z1>M06VU<A&uW}rQgcMn7KwdlJ3M;+vFYy@|{FUOv>%nVg8|F
zutU_{xO*Us5furKw<L{z=&fE}qB*<6Mp@)8O41L?gFHxRLyTME3R>Lm_<+Se%lb5i
zdHj!vF(;jWNnmy%UjRmC(XnPVTxu-X)vYgXy<;|$@*LC?=3iOKoNNZb$8HTKe?~dQ
zzP;0kOQ>e(q{1T=n+r*zUy&f{uPyvHud?{}Y_>ai&wt;-8VAqj-=2-`X{RhY^gNo%
zaB&|ZRzOsyVzT!kg8|DVD0;s5U(V+J>)8d`O+M-oaM=_wu-u2BNB1H9amYyX`2_Ou
zOsaMU*ryv2w-0&EOJU8T%ge6CMDe!wMUv^!_sr}hM@nB-Xd0-0%8iCBilf{59>EDl
z9&u{tz#czt4Eqq%svwCeof9v9IroY4BG1&KP3ztRaM9BDO<-~9n}vth#L!N(<=nv9
zUbH=V!>{Gi9Q}#!p>UHo=y|<Xu-mqOryH%2UO7?1x@mkOmMhgw#BB?@C%MNI>`(0M
zoOy^MJO>6_276b(l$KO`KGT_l!~r53iRv)D0YpRw95iV!$!`4(5AfFVShFfi7uD<E
zr%$a<Jl46TzBe6)xs7p{;kk!6M*Ris3|H4$l9peXcg5GMtG}M<<qi7WfuGrxh&W7@
zBbh^Mjvib;$HCU+;XeA;rzM4Tho{HK<uy;;pc($LgYKXCB7yIFq9#xMjN~EuM~msj
zIjmbgOVe(q3A^^Gfqmf+B!p1|`VCVFTNRiETah<k0AkuxyN^L)T)01Olx5OAzO>{y
zOjH-l(?ALbyMuRLLpC68ZB3yc3rDu3ofKPH<;|?7p$<iKAX8@9kYfm#`>BnDwl%07
ze_TyGB#U(cA5G>2j6t({8JHl9u7e+BOJ%FK{-B%IyJ%JRVBG7R%a-G|HsyT-`<|K8
zcU9H+1iXtc%z5C)%LkcjWr}CC71z{ZU_XL?+39co4NK-6qH4_0ps>pzm8wP$Tadwy
z@nPerkM*<CB+4Ib_J$*wO^bScy_4FkfY>{QT*l(G{`}0!jb9AG5~im(DsRwe+|T51
zFbtW+JZ?HYx;bisDS!Js;8xvX6A{!IvOOV_%mMC-kI2L8pXm*;7{O<JSDbv{D#Hjt
zMS~XDsi@9s)i-bKGr8euPR;Ne(Nge&#-mb#Lplu^YJ!T^*d-b{z#b0+&Tw@KD75jG
z<_0#RQ<*0RveqZUS=3=cgQ0Lgr?LyPeNWIT2tsmm-S_O~wJibB(Ir2d@aa<y6n!s5
zRZW{D0*6@HnnQ3b7@f7KPQN3v{bTpTZN)6s+u0Zx+>R9xcWn2Fnh~fJpl#=Ef8GC0
zM{qIvU1*Vv<*5YSP1HEp;CYaGjCg1*T#duR5@FyXe95=eMhb}Hfz?4CJq=2OP*k!p
zuT=fFXMr6&<Cv5|8|t`)a5auOZz$Jg^OrhA#0<emIzE%73i>{^y+%z`Ls*_PSNTz^
zz)&65uL@l+STWoS2=L#2JNSJl|FP4wOnE{`Z$_J2O2FqtuwH(0EPEQM3@qP@PJ+8i
z1L7W!;11d|niUgxY3`bCN~U`&XXi`V+D^d_-$T`EvN5!3qNwKUY8QhwhegtcY}B})
z5S}iiXG8pcc!ey75AH+U&<S;_*l`rjF5o%NJ6WKo`yK2<KmDeq22vi`ha%i$<qpuK
zO0Y)p1_BA1BDB8u<fl$<ol2)OUq-YdqIg-`(T5hB#0gWI{v^}QJA>N2e$|I>>^*AE
zVtMX<=PB!t%x{Pu>D5!6oiJ_`mYms(xq_O_Ley}z!yje}<`b~t9(Nw;a_Wh8#{HUf
z>!Hf=FUYFh2Y57_4N%|4HQoU}Mh_90Paka0I?<BYDUX_A@7BWJou@`)NDSD1D49f7
z27wz+z603KPQaKK0w14ue(wAZ>h$F)q|<g4fzEu3=wLlgb_bXXfaG?zlJQ)0gq;!1
z+;pfa+6M5%89=RLN?lqlu9tcpro&bnewq!r$!^x5JUd0(a=uG2GJBDv;>BP0WkTd+
z(`$8f&l(_(SuG_>&(e43h0wS7_ggymxQj>#m5Zz*E~~m=;j=F=l{IlFCo~ffz1E>%
zXszcg^(F0`7LQ;qsh@sg*w9L>Lpm<w7)b1R$QpA3HUei_&OhE@?UC5l5a?2b#X<ai
z%;zFkX~6j&MvRP-d*Rx|w>aQr!OPE>-ah4h@~a)xl?EZRjH%)3>^p9qo&elzI$3G;
zRm;(9;Sqs@hQcv`=v!+=1E<A0JtF<PRj6m)6Hm~K3V5d33lF_P55c}{P#KJIuyHB;
zL~9_~1((VFmZ4nt!xcYjqL=5UEX{TthpDOqZe=rQaSpGYm55#9NL1$%N4Lf)QyJWd
z)$^SL&T&QK=?=q8hx`<#rZq3r4>pZZO5wBl76u<w>Gk_n6qwXb5?9+FbFp6#7#)rM
zSi0AHNn}1$oObPu#O@NYlgz<2hfR>v@i6Wn<v&T{r=)A8;GP80BuD4q;1e@*m`7XM
zKN8v`9gT1HBD)K%WXm((^gvKoeaz=ED~90ET&(~`O2-3~@zBzy7`$G>^KsJ5x9XRn
z_uq>=?Szt!|NmJ#;Gj;|!E^sdF%UencY|z6RA|K<b2b|n*cm&SG4t8?OKI}^czn}S
zbOOdLVVAcOP6D-uu^!0%Ou{hi7-j*~7@mUu?-<+x7Aq(YMB-nW^;u!|td3b0XRnwx
zi+wTwn6e39G2Mp<cYuXfYzlw!Ovs*06OvMdJRJ~&Yse7bWnLRQG?v=4OY1&<SScZp
zbhycoQVSmiw3gxP95mnIfA&!$aEF;Vo3q57HuJ%)Qms|9yS+M>T@5#WITrw3&kIUe
z0fmQCfd+f@-Q;#iNV^Tso?);ywn<=N+=>L99}`IniT8{6*t~`IM!!J0p?A4_3`iu{
z$A0?weMsw#;LH1vtUl^x0|HtTpe6Qo2+9Fbj`{b-Q^(JKWv*4J;@Q<%k@$WFEL=Ng
zN!9(Hx#>Zn>!+VinS)pMwF!9O*1}gkWcASRde_|?##wm;t5!xfYF1zELxeOED85v%
z`|JfO?;|`H<uGcU4HfQ>prfh+eeIZIC9*vR#v3c&_PV~k=~HWIaz|gMfV$MzQfJis
z`z+Rei$5*&MGeHv=47r1BG@s@Bx#CRJm$5<Uz`htU&73SQqua>Eewh5K{<k52tfv*
znt8yhMlGYZfEm(pbUz6q5eI<cmq<lBe48b6AEGWdX3j%Qrs^P~Q7Zxp)KDkd-4}01
z$~Fb=sa|gJt8S=C{Tyw&RKiMUK81MeNM>&`7`HV@$J57q6jTz1lNH8tV%>x0nXWt+
z<A#vTn-yTopa2!H4^g5f!tD_B(hpINs6&f7$HFzD4azm?dlekFvRLUrW5A|0i;0r8
zb<Sz^#T7TuY<K&~)X*nKxOyWDgCzmmf?ys%KNq;5qg+W58pw)z(tIb$D`efJu80&#
z4wyt?<*6n#<TmoHeMnN_ob(>E4ajPD&<nQPa{G{EbHI8Za$5pLIgH6b#lY81RqXd6
zcZz2yf?E_OWCvy)R8AN%D~CKl0QA#WfGnAYyg;2stp{Xs!dDc;s9{y4wS5R38oZVS
z!I9$M2s&Jm=#L2#-BbfsJcbehe#?F{4EQOtDd3>nBboLgWjh$sjZK;@=_t891I7gf
zQWd$&!`YEhUnEJM19D*$^9nIOv7Lm4JELRLGN1S!wbPe5%_|`P;bm#ViA|Vs-yEVG
zMlha%)r9Rs>_ElrR=e;O?~*g)T0;ECBh3`ymk`m2iu{zk<xGAuyGxlKhmvesF5E3?
zj*nBa)$`(NG=#!Jc?jOpuO!h%T*w7HeY>)b)cDGS>NL6!MXpOp=jj9k6ww1A3bO{G
zvae2Ub|V;^6UUTRoyg%TAqH}hwtJevX9iu}R@}&U@^U%LOU@U4J_l8YuzoPvhY*G^
zOz;KIA@~|0i!~kPA`OU&JybgzU^8z>kjmr=CtJG5iJ%<2;#c2lYjw@3Wn%<Yr11zS
zwHLu9+*pxeE!KvFH96%S&&cC0Myp<eBy?b2X9CVM5wK0g#mpVR9FHY)jDVe7_YCeS
z)jW1zm6na!6F=pA_Tw3c6EBW-27^?)l=2EGb1OIyfaCgj#*g4bbVuwm#`r{!D!<+V
zyX0)9k;g%9C<!dCA-RBxR`z=B?eT>1JCkE$JXa5C%^`9t|FF&Q?{kNUtSb`44bXM1
zKy{{j@C2ku6jpv*ajceb6x^BI^X$;#=cP!4>TFjqbRUw*3ibt?<px{N7q2VKB0hkU
zFSS2IC85QJ<DrzCW0+oOlA>Yx`Q)UK_7G4GWk~_%!Ra>vCo?2~J=U{WLB|}ojyQyx
zZ*7AEhv7gG{Lc>o6jeZi*hhU>jRK&seF#*GtYkiyNudRGL&6z2^-L&ST@5_nY_4{k
ztVw*Hf@_rTw$@wKM2L!}_0uuOuKRlqHVv9hzK*<n0rZ0*sM!}V9yM6-Z9?Z0IeoQJ
zRkfD+o)^!*%cwUCC<I|wV(*#B8V^InX+NW!*LF`JL1p)?btg6E9@Vn})Tf23!$Bz_
z5w&m`lao0fMAe4H!&kUGLD792^U{L`^>Otz3i}m7I+n${xN!xv9HQWIV?zXIsSw&!
zfR(DAhN1+4!gyA6qdE`uB=`oF9Pkb6s0C3>>OO=A^>L;Qg#`@|>3lIMl*mdB^@Wjk
zK+X4(92o<2Q#C2xR7?S=ddI252F+LIz(tK7hoyv`#cb+gY4%X4uAE>U@Rm(5pFmqQ
z)=AZ7ZAI@vZlM-Z%4oN#;KbGhz^mOZM5RDbs=U|BgI-VZr8B;ebQB+WA*`r;^L9d;
z`C{Y>$jRvj2av)6nyq;0HXy&Qve84P@5+6%i?|celwXfc&Tpc74I-89EyIpAoWH#w
zqE~7(dRUW(iPzMKbsjN<dbG*iAc?aU(+WEk!N=z`t3AhU%8+%N0WvF#8AIT%$K(sB
z^)*|#nmx94XqhRzh<=2c&CVoT$It_gxNkod-bgcvY%Axlsy|YXI>h@rK;CyL3MWTX
zbkNF_*A0qIX8?&gN!1T0E4#kVQ+^%8<Uvqy#5>3KVoVqYrXq~LDdOqIjF%ifb6+%;
zV6x{3OF?P#y3G#zEL&FbnFS=>P*s4SN>P#))~&ae%#zW}{6%ww6WA@r+0G3W*lUHI
zk`wS<W+&P%lxyGkVd=5XS%+G@iAK6fKu=>VlV_9WNpIkA9}dyilxmCrDK4><x<0=p
z=fSqT6DIdSM^Yr-Wiy7#cx!I)dvdcNd)(*ymJO+1hZLu8sq>KTHl$LIY!KEAh>oj_
zg)T}d*^^~41$;4@cc^7^pVhg8LJj8CVV=}(6ye&>KnQ)qc^6;&D<wEIAm8rRsoHuv
z$@=3-(KOApid~!N+$~J#K7^I50SaVxGl226v`&XU>5q&Z=ZC$mS2<_^@a_(YP#&Mf
zIO=tK@B&i8Twzs5ud<C_U~~QMx~kEuV?^^cZ30|EG^hg#RCb_phIxx`TS>5%ud&km
z-1=+^G(C1x5>bNTX<`Cw=dcfdG@QaO-;&x4ieOn}%CL-pgqqeLPwshuXu$yBG;hW{
zs%y+B#7!~a<QT8~TIODMtqjg4PCH0;CUAWaR2dN&X}E-oQ8n6!bf2F8%(B`*1nmSm
zVcfbbh8WXsW_GT$F;}51v8|EAF%NakdmmDmOc^2<@8(f*F<;*UtyIicst?WH={;_0
z4;au2Hr-PM^o&MSgl!*$sJ^SL&^_v)4E6#-;9!RqnrH{v_4<8)$_ct}w;$ip2<Te@
z7Yxz*Su|-2K{@2IzzZSkf#1Ca)zxl+!lr?9RpJlYKgc2Ui2f8mYIak|-~M|jT-R6N
z@`O(L`^JVBZmVwnU1!Vg7iHNcBl=QT$M$Z3cDFLJ!|l3eU1G3YZtr%QZ#>Gc{`%<C
zz#kGBF59--E5z0S_H8TT)bIoLDd+XEw%OTYrDMz>qFH>T0lg-n0B_wCULVYGDJ~`G
zU1bs9rf8_z8hBGd;9x6Cs$FoFM;)V{jG~XYwR3Axv2!97^24^7vCj}uh~%ccMDhY~
zeHUi2W|CV~v1m(#B-vKLQnYT34m=H@Z~|hjX;;-WU*ubsRy81><or^-Ks&cKZa_U+
zK<J)Rh$ypE9}*%IL=AIA-l(LD;(McM<oRfbh;GBc0I@dIFG$fPq~NgLB8eL5XDjoe
z?anj8r>S<-^{QRaLLABftP2*Al7!AUH_7nE_7#L(t<#5J%nto}e(R@L=YVQf-iO>f
zOv!CJ+Cs|h5{s7AuTBd?dwz^-$}>+R*|s6prI#|dmZhnoF2r=|5!~GJE8*>kCPMZF
zMtxg@H2kFu#>=%l<R0W_Acm?zF{U<xw5OjLf|4eU3|Fh29hN%b%|9@$`?2-S$v(NQ
zwH4MqmPzU;>HBNE2j`591%NO}xpM82LsRg}j*=jz+jDur5AcXx?kWUn4_uS4$We|U
zKQyt%G{fJ`KU%r_@b-p~%X5bJUEcL#8(A!o&6x6;=6wjw&Aj}C**LW$Bk~UFgXecn
zMSX98Z&!oQ$epYR<jtxgj0x3=un023g>lkya@^6a^-PkGuNZ`Vk_kILJZdPEawkQY
zju%3;cKeC}P;=U26)br<Gt4dWiy}r|G^A3L_Dn@VTOsmXFe||^w!k`hnb*f(MmmT8
z5jZoxQgUS-`jM&u)c#H0sR{I6>4oFu<sAv?(Y`l)Gh>W65^VR*?eb^{$>#5k(;d9;
zpkB}a&lqSx<99x&Ucmpzb9@l;#`;!crHxSu-R;=qnPr<o4M_0VK7>gR2U0cWB{*m`
z!u?T4z!q#h7jS~*58m8d;xhNL>JYYn84F<vXb)`XfS9KR-vwX=!FSyx(de?X)Mkw1
zQ>~tQ0q;$;UUsM&$Y>)@fpK@r;VoDXs-+$lj;L@?CZOAyOULe+`+IZqvP#{RpPkE~
zuL^>KSIY>t0Tk{K13P-hN$dC?x7-cdOR>poX`4#owBP%HyFUx{_IntTv1Extz-{l`
zB8<7g@NkuLk&RUkrzd}#+ePk8^gXam2u|%RS?nE9p3_<|J=N!wSUuZ+Jae?Z?k_Hr
zO<-L?=LFEVjgF607YBLDstP}gE6B_Xiu?1z%H|s*;JVKblFdQj6-bu{P>U+Q6b9^H
zIqgG3Vg{+EBk0eHmF`2ZbEpMP%$4Ph);|g6b^J>(ub;-sqr+hLRK{tL-}->VPE+>j
z{Y-4;x}J(L$Y>YAgc4eH*ypO3r#(wr*<*8hp^@eHs2NE94q+e&_t-GpT56bP^_wK>
zypkh1auu_TLv5a+f>D=!$j3)G^xk1h)IwesOIYy|j5r4}daH?D(`dVasMi5~G|x&Q
zsB@W|sINr<-PBfd3`upF#7&92NJ(kpI{uN`0=hB%F1ss0`_USpb}k^{kX-jSazP+9
z6m>`WTN5P_$$89QHc;ev%`gLGCtAE{{MF!>#TtL9`)}LU*;hC+7u8|CeXBU3uD#o2
zGy99ir8<v5i()|mC515aeMnadaDiGsy$@Nug2nf(O5(dw0>OdA4vZ6CFNkto$ME4t
z;fEB7?R4r<1W0YnXZuKNAT!8gRTp#vpG0z!FF=nYa*?Wst#mfOND!An)V8h$;r*jZ
zgW<UPqjIViU-1%7nzs_J!<8uQ;D_c1Fq5tKPP>dhR&Pi<Y1tSpqkXo>a~C^@SgDzN
zPvxcdBWa1uJ$++mORAhi>XH=&&4~44LmV^<Fv1VaI4w2md+|O52|lGwRBJuliIpUl
zhNkT8Lzc0d(7Yux(&!84Hx}Bbs_LS|qBSlu#nRBP%iRlwXJ7z-$Da{_s~e{K5DbbE
zXaM*lRw038l@Zk14CpF>GSF`7oC2BS%Iw+))RP)PCgSzHH&VqO99|!{<sAZ2k@xc+
z&c=Ev-=OcyB<RkJ_2CUZG@{z2=RBCF7R*v_+-1N2esUtUV3rvfkNq{6nUVtr6E&Q<
z0yD!r9EZ=A+O6j$-DJ9h;F;KJMjQN#3BX8UW~(#Tr2s}M1{`6HDr7eHx3NV~9)Qm#
z!Pott!?%xuLB=tOox7N$pm+S6bN||sP^Er51H?W3T*BU=Nfa2Ioag#r$I824_mZ_*
z0~8S%sEPX!r#RIr>4!}h!vpsrwo~wL?7Y-_vKQl;stP_e$XIE%67HzOhTS9gAt&_V
z)JD)B?ANd<>0|mVMwm_(dKp<6^^lZRss!mF3O^yk3or1C*CdNke{tCjB_^mzVZj9}
z^XJ5|HQy{CZnQtA9s(}RF(VgSEWbFf>#kfsr-ZVgI{6k?VMaZEn5U%Sb?E!c;iIAO
zJdeNWTl$0QK?3<HEE2H|c?Wu{5?1u96cCwlL@{bJ$okJf$o!N{;ah4K1CHKwhfVlJ
z!nQ|TNKWYQj@w{xJ!x?k)V3^83nS>ov`~g`)YGUg8IX!UT!U}wVGXS??b5rVHd5d^
zpgK>D3WQK^uVUu#(mS0S=p`{wgIGTW>ITi*Ao?U-cKtvOmhc2K1M9`S$)p7*dukTy
ziy8<ejZz2_q@K<8dG9?E#!bMPBDkF2W=vV($7Z3LIhwjC<#V9-)w$iAu%l+MtPf6$
zQQ-K>mIobBe#vPq_PMM2UcMZCW^V(QdSsROOD7mzujDMuxu@XbTGx?wr_v0*EVjYU
zp!)G}DBNY?XYLUC93>AaB@NzUSkq}@z&Q3fr<ZE3`+bpgXK_pvboUzpL1oyO>*E8u
zuMpZYSzYBUOEs4id2g=W;^BD9H0#>@p$kk1+q#R|l0ns>Qc$v(%2aadA0v)=g_vO-
z>LS&YJVYIgHl9n3y}|^m!O$<O!|G9QQRK9>FwF7@Y7>gia+zQIqWEY}3Xq_OFMv}u
zt26{$J)^)lsKKscu>oo053Z3kAcP@D{T%7CnPKxj%-{Iy4f^-sSejg$+rBXpoUpr>
zfZD4^$N{^1FiHZISGpPH(V4jYb(yjfhbcwksZ&k#YRZ)7<OeR?(yIBHl3%rRPQ5q=
zW}fW<o@i8yFp<4At+K5DR79SKY-xW{Ns%QyRF;NvY2aqYQDVd^5a>sp)r1E{H#U2&
zPf59b-O}WN)~FvsdZC(K1c#^`HIGalZm+GDBv@yF&YcB?50o*1vpem~a8qJrgbJ7A
z{TeRLbCjZQH@<3qeYdIxWU+$IN|=7D>^`JgGvcJO?qW^s%#X)B=qc9YsE?Q;6t)CI
zqCt1e0K$mYy&lWC?*;fT_$m9`n|f|%GpC%*zrpg5T>-!GK-5m{LlR@ar3_t*+Pr`m
z%v>SzwiDnVmZCXOc9#6Cr0-Sa(STH{rldKw=^HqEdr-QVig|J(3`FV|AuNbeNk||R
zjC8<yFstCy5EGc2*<&!+hve{LoKdqN?^t)Df>|nPQ>rOYvuA_q>I8PEEuhA;(19K|
z07Nn5H25E$dI}^&OM?T<H#qn?7=r^0{B9UP^qG@6yr~9ycz+5CwA;Bv?4|@FVWt|4
z8?E8z+h@jII&CETWxF3#x?TT%(Jj~Uw)SzMm$89`S*%{*l5+v%ckE!o+A<NdDU5i%
zMVukCBesKxsFfW;r}3Z)!cCy`>yD(5h+PIF`y|62TPVXEQU=v^mS{kE1WLjE%zmal
z!P)~qf-lXg2I*JOl$Diyi#`{~c}6{p_4@_Tf<2|wP=ckY;fSHm)|uU7!Ezv7?)IQ?
zQ~{_vAo;Eb=2;5q)^aV1N}8iK__EwFei*M{SV2!Npm(+WJ$J@<#}PZ)y>8w|q-Y2h
zOrECQT>$U<VKCQwKyptnAnjvu<MTY}?Pm&f!RYTvEA;(oq*Urr(5M`uBh0kI8Bs0o
zfw*9w32ubnt75(9aU7v^A`tDZo)GLF0@;CtQc4;m1gdja^8>;@vfCDY$f{#cNU0dP
z4w2hEMYN?JO(53e@;0OFaoa@u>bY;-*%H@uw?%H>?x$Zpn^}o%JyMBY6(v4TJG=d3
zL#21-dKDpZwH^F)4FU!Gc#HK=uxLYwE@}4~7w^|QDMD7b6dFj<10M7vAJmu3x#^04
z=f#AHw(3tkm0g}r3o%=wO*l#t$lDTYl&&{A;yq$g%{JQIyH%T_9KMZeE=NT{Vgnf<
z<eE`g0X+@*zBl9D_ppY$#NC`j&obVlr&)nhDDn%$BlvRT8ugiKaKWAr5;XYiYSO`;
z^`i~2TnjKL_u#Uk3EM+;)pLhJ8u#ePVno(v_*g=#8XNKIQMb}t-C!?a9ue^eFH)0K
zk7TjB5);Bpg57W-2A79t_0lC4czJ}M+VRJ;&fZ5cP`xqtq{rrMDCty*r6!({S#PIP
zj^&LLPk!7LJDIxp5**)C=84pSnF>}g!sgsZuk#ux>SI5X`EKdPn~haV`tMf&FcoHZ
ze(x7lv%U}dP=L2S@PK&v!JW$L&whdA$L~5aTlpR)Kl~t=Lb`=pi=T!N%`&icj?tUl
zG4|H&WI{s-3o{Z<b^<}6O*SG3Y)y5B3(mA#$Ruxjvcswpse^}pX8d`9AxM~;!#WCg
z6DVQ3FMq3P2<WF@?p)OYpVbM#m!Q=`3a^RWa#GQn<BUrh?U*>5c#)pvYb=zW-!YAR
zWwgqJE!Y)^^A3^eDFPy`8l=B<FtYs^8_#(chN3}S23V;_(}_c!Y2ghDYJwp&BEml7
zO$#loNM>XjvJ;Fg!0D-<Yls1<H{N@&1+;dzOirFyU*(QfNC{P%FU(@uhivmuGElge
zNScLl=+v7k>>7vl(x-E%!6VH)f}so~w*Os85#5ajBn!wfO}~J47c|iWZiqKkgaR_h
z+V&k{=c+&saGmw=xQ5R*m%d1grZB^eSORM)Bp9RubReToZYT$dt3H<6g{X(_g4BQx
zphO_W$@G-$(qKbS<Y#mfQdJ6IN~~vZte*9v&qEgV|8cQ8tQE6)9-nY@Iq<yOBg68H
z{+fape4-#ncSgztD>VShM9VfLd#(Ac%Amy)TAp!BrR#6E<tU2S4pj9#nH_zZQgEw5
z=k;X%qdBLXFQO-oxQF)AHLxjdanQ_XzD*!=JZeUs1ibTrD|o@OI#|@^Feh=!+O9LO
zVK$6r;^D)-yby1^^mq==Jlv;x*cZ<0T)iD#9}_9r+5d|RkrmHMX2;Ccv`*=l8aJDh
zNQLl4rL+1fD63gh*imFNNU+Jag;yMcX-gM9Kl)K}j345okh$^nICYo?s#zUAn{5~z
zJT!sBw%XTvu=I#d6pzsR0$=G&mup@(LwAYA&I*A7o6r6G<r+T+Oz~~;@Id$E_CVg&
zELFJ{24+geB2OB4lmSI!H5OwZ{KqHc74K`))rfev2j>u5sYk1bnH>mdgHFj}ooVqG
z8*ACO)!VSsiyh+-iVNi>>T4)j34An%QmIiEvv7~KU0IJP_%=i8u6bcF4`FrfF_JAn
z;p(u-De9}-@$uf`e(l6FZ^`Xo!ewXcDgvLmvN4BXTre!DtIoS4!nv^9d{+^4?ECsw
zHo$z1p$~{{-bxTG#fTv}q)@N`I9r^__dtC|`lyRJ$+X1GPbsDG^NK+HaXX`#)&+^}
zAAjlj+_X^{5*sD6;w6=a57*WYl}Y01U;9TrL+0L~Unhbd4D2|n3H}OYs74R+PHA=U
z+k{}KF&5<Wk<_Cu#1U)0J>KDCb5sijG@>3+*TnfQRiDtC=&aZ_x_!TG7WLo;Ex5)L
zQ4_Y?K?dW>vzj%yW#1^@_r|NdKI>A!wH#sKB?F`+2<%rzRP+cEc;Aeyr(JnKYU@t6
zX}o=N{8d#-G07jDun%$d0L086qnvZ38uLG@T%0>)gv2t$rRw+6H3sa^RzWUogSICu
zIanSTz`5KcUn*?Jd2n1U=HE7t)lP{#G!oizl!xrHU~oJ*7zn}XOYG?x6CKh$94kI=
z(BYb_9J9L6t>9J|7(pKuk6~WRkU%l%<4x)CZcMTt;ybg-46aa~P}Ks$)>dW+Wn`^;
zjZCS}d!N9SAUg?w_|EE<^WNnikJ(IG{SySYml<yDxKHV0ntmw@mv3h<yMu@lLD^Ff
z?Fe=ups>hdSCD^TRC`rr^V8I)PoMnuA@A=$D9s$Jp&r>oFQ>&?pxgB?`NT=dv!waP
zxTjg6n9`rQ(T`)y8yIJ;xp0_pR~1#<GI!<bnNR2WZ(Kd6$St<yMC*(7db`Dqo;TQD
z?uX;JDMbw-1#V;CY)*IX4HScOI;MD|dz%anlIt`%ka(4<Mu7G-Vxr181TGH}KWTYa
zsWau$Zrr9>Q|AX`8sHjA517s(9tL`ZtN5Z>nV0HZ>P}}UB%Zo{?vjS$A|ix*8feGC
z;+lA=GXw{;U{&tgI(2|oCT{xI+sw#g)?ge4scf+Xvlpbt{OUe2uf@(yt!+o<gJW^N
z=_oKitJS>fe~H+h`lZZaL~X!XP)A6ZH<th5BrgQeqB4R?hZxiu6lh%j;7lF@*ELYN
zN|b1y*_t3lmx8y)O^I$24MqZ89_WDUEsU=&x;?hfb)XX{USQ1&TqCnz*mfqOSC(;^
z$1MUn1|@EleVbzI(3uk+X4Qr4S>700-Lm9a`GK>Hxz*n(V#HK@n)?KH1Gndk{LI_A
zGDA!^j?fk~U|AaK@-IM~!OOGw%i8vrt(b30+k_A+W)Y8(a6oFg)kbQ_wy7e;AoV1y
z;o+VV>;`B;A!IZ-%%H;3tW`o~UX(|%P)}B765=FQD@H03@4lyXVbQTsVzr>?6FOHJ
z3G*FFE{H`nsUi1LQ-#!xFwfXcS6=cf08~=w!hlD!jhahqA5!~)oJ)u;!_Vy=YT?lP
zD5c%Hx$O(UZb9#DP%E~_C-?RbjMW+gVERfBB^RU{xid^n41mM(9=ma5yN%dwBWAVF
znTd6W9fY~*<_YLFk0(eaB3GwDR9eB4trI(>?HV6@rz|6+gCaxm7xSiIFjFDSk?aG+
z6>1&oixeRYDoCnT^PV!CDgTbqL)pOqQE+F2qt0vM;rGVvg3?BU$}x`*D{IbC5Z9<e
zp-NyV&v6AKy@Dl}a+#R^FZSL%9P0P)8y-m_nh=thic(3EEm@|LEs6522{FkQ6H;Lc
z^N}^{L`9LNQYbW)t!!golE%J_m>EOa%~-}TOV8Q&_q(6_cU|}OT=(%j_j4T2a~${e
zM@N{M&wQ5iJm2s4>-~DYPP8=UjizSZ(m8S($*!<HOm|dsResraZ8`l9d+Ox{-OS+1
z&iNOKo>o^<<fs0+cO21{8wZVD?S-c2(0qyhrF^b2$ZMsTNPdI{v~<3oA@>E`oeIu)
zU_SMS*=xu0Hlt)xrR#%Bgt`7%s16Pv!P?WI=;DxNu@JZ16aGht$EU&(J$9n5pk5VS
z^#8lE#s2ro7X7c4Ee?n40DO?Mnu*m$oc7US=yccYo9=nyo97!2)jal84KgWNbLDR3
zY4)4Iab%nVlXU;pr9h{B>rUUdf(dkl=6e2L-i!n6tOEPr{ohq=ep9(x#KDV;&OI->
zHJ;F{41E>eTMM>p)64jhYyNzJ?7}=xwxa%Sb6mB^Wa*k7Avo`B*_~^%TZ@DqPl@2b
z00O#30BxSEE&*7wayaspKj#`bhPtE=B(a=xpz*K391_Iv^#wrvDF}jqkku66rf860
zli-&GD$L+`BZP$X^j*)FZ<-C%Wt>%Z-aK|M117+e=-DQUPzMz(Zd4%l(DADGUsD8H
zR-XFmr_SPGbOWl*E5|f2>^3!M+mUaOkTa6E|2OPDw4^?b=4@UP;>$qw=AnAv%SZqg
zB-ND~sk1_ANA!9(XTIb_kk403Jq>-L`eVq#9{nBu-#@xSpPrfLs5XPZCCkCE3o%t*
zr+NxBpbzgm+5=?{xIe0Cov(cXA0SI}2-Zv-Uhi$eNMr6KlV%m5Rrmhv2HU<bVTz30
zaNdb`l*#hoxS2!q?+NaQE%rJLb%--<IR{cy*y^k_F~p>t^Je5YU422D4S#M01QgLi
zuR+v`>j}H`8y1)+ymsJ!qJdp!kef`|o8~RIug4>+1qQ!~sjCd&-VWW&xs`8|gBd}7
zD6Wp8)00#m39h|HHKCa(1f{}-**+dyH@|75J=bZtrpu?hA*do5mWY^p@T31HzFsrR
zL<dmWO`g&vHvoc<d`fs%!liG|&p(gGyEejbOA`>&4iB1Y4fd)R(NYIpJa=v!QFt)M
zKVKbFQXbFlQ@nrKN}@&8TuD6T0npmWGeNeQ%g`X=EA{z>G;TxJScU1WBiUC~zgzho
zB@Xk&o4NX?-7_qF(NMD-a@Y>}2IlfFmo1#c(r<!q;OL(hkozi4J#t;#f*<NUxqi2g
z7{v>CiT&AST_|cFX&<3oGFIkWBZH1ZGcQ`XAStP|f@IP}WHw^HBrF;%YHptXkuLjx
ze4|cw@_(uob%;*fI-F#!RhQSQ`>UT&EMH$`^3y2<FX!E9qd=MHT96D{BlGUflcu8x
zYhuKo7>02}KFhvQrTrno?;!^dT1{st7Yn`LUos1EJ9e4!RbTbvvl!)Gze*>zKAESD
zk65rmy17`jPN-9^!T#lf{{FcDA%~qm&eSMCZD+sre^bAImomLH!ptA_OOq#e?<r1(
zW0`;CH!R0`<r03*PLzUXb`m5>6aGMU&CLGSxyc*wmOJ*oD$mR<*X&lvo&9yDn;o8T
z#i#ytmem9U%~s~$pR(mx|2=;~tT=D!lD807S$i`L{32)fiSgPYy?&&4Xo#0oDzq6R
zWqz#aI1QzLN3JmYtK66~#S*UAw!a+)oxiwZH=M)So9r%Fy1otfSOYSI)<J%R1ccCX
zf<f`T)a9uRBO|M9s>#G2<tPTS5cdQ36<MMd3;FoR%B=}fCo%6%8C99uXe`$56jqh^
z<!K9O%ccKK3-*8K|G$Fer6{$kT^}s7&wp!vAzWp3xxaz9swUVyE4cSJU3i@>MTXuJ
zEGdjFvvIh<-nw{4^4ZmjA;k$@^`m};9u9>+ohNrcixK@y{HfBQe*~&)?IDHirOtXI
z@BfC8-O!!=GjqYAo0(RvNY=rGhf6Qur~dtu{9pO~J+cugU#UbZwDYF?x_@wKs4L=x
z#DcGH?#uUeBsh{rT&RRfLOq<*1*&Po+kV6DObD~n>r9Sy7~R3>e!ut5Y3jkrTGiCy
zGLfs=kxsp<08LqM8}yI{Iw=W;T7tpYd1+`oh1f}f{i|fvy?alk=D|d19mS{k<Lcx$
z{FgW-(;G}=!tLItJDOTN#tt8Cye$AFR*F7YqKQ_>B^mU$->}3(pfZkrT8P&>17X^E
zUV;!ng2}qgKS{mtF=C<X!gGoTe8>Hp3brQ?G~P*{X(w>JKy%bR75d~l_6!7pyT1Qn
zti5q8x0wjLR@Uu6#pc?Ueu)~j97`9mtqLrM-`AO!nf&Op-T6mw>CLd#R|a62ww?99
z!C*(R@*L0z-y1l$Rn#nsCzDhr_-Z0!__X?LffX_^?3(Z-<zGqa|Mm0#+06u7qW{;=
z|3AJ5JIg<oR^vU|D^X(nLA%3zpIn~GT?_uY0uoaRSy!C*LW0UMU&616I6$;xP&Dp3
zRO_o#VGJ1O^)eW5k&nI?XJjONUgP&1rJ*L<abOD8`U8gKvrtU{GL%pSai?LNd4X80
zzhwXHg_jmaT4V*&M~_Y6l~qYN7F4JI8tORE2Ig%yPn`+Ch(E0?=NF)mhL-jasp7Eq
z->@iuOHevqNEZm_a{lToH;zuM_AN;&seIDJd3oXJQj6@(c-SgJApYhL#esoX<9q`0
zgeYPQVL%_q6X+KFm~4QCJCdA9w0CGAPS65tOoAqV-)UYnKicnV5ua!xXKR!op*6jc
ztny4m$VnBXf{A&=b~L3(AU^3z3-ej5Cx{mmF1JlPNZg#3cME%~5+QYB{exR*>1h)4
zylwK@IKvq(YjZ`R>`stJs@7~GShy#o<9WJ&mLw4$PCcz{qUpo+P>b9mxUOHe+o2Jz
zQ}71TMWhErq1IAM&{62~OkPg^g<^5drKc^6g{|Mkn%D2xaW&ldANComA?#v)HBXGI
z(R6z#f&w`5Sx3rkGLu;Dp#@{C<hy~RB<(-6@mdEH9c<TRQ<g?-?5r?i|G}#s5B+`H
zD;^$4l<S}tXa7{~=G8{t`X}i;=ir<AehYIBPfeb*huLsVz)S8W$XQDrYJ9myW!g>I
zI9-iBMkMvpRbeeMfZ_^nUqnRkd|&cTG28L?jSSwJzMGWqjJr}^c0GT;0G;Ib&mN-e
zYAhEN`?Z$U`+<OV2a&gjUZJtU`M}1=Y!^yps`RB~OILt@?FMf-r~9@<b4b2Pk0^NG
z+!BY}b`HNttcGaG%9}oP1JaZeRl3=K`_!?RH^#+9;rd6+>xE68^^XNN%_knCHWTAA
znv;!8Webi~=e2ZOcqmxL^la3Nu$suO3PH`xgF7$HdCJx19kbnly+}SqIBosZuO3Y)
zPW68f@gsIe01zDSkzHK|F>IP&A=oA8MN3znc)_Xewuoq|Kf}Fjlj~IMPx^MDwoUo4
z4T5PmKSUZ;13VZ0CUa)e`y}10*`hXeV05A~Cvs!MHa9>Ui0DKE-9xPb@;X@w*PE_<
zz;@e^x80$q+l@qRqO(=gYKPNw*~=g-p%tK73nYfTUGUWf!7Q1NEL>_G@0p_jw50fj
z_i?-E&qA?pU}2eNxLe?O;L*@tAf%B#@I=&ed6q{`!sbl5tUX<&1i#MKs?#+0kBkid
zwij@~vR2pUWhpiD@51?#`7=n4iaalcuoV!fmY~*Y2KhgG8CWIAJAkH;@e2RO!@q?E
zZGWKq)WT81n&9DT|59fGAW0~51)r_Jbj%QV2y~#<i=f#AS@9->*TnfrVA}r(mHz|v
z@S;O5V2LweA=t~HkpO96Onr<nG{m0av~$&YE(xZc*L4&Ln_V3Dv=_GbW>?JZxpy*H
z=m88IXelC|1A05%Km;DieqGH{j%8e18VbAHQ^2;$X}EeZ9RBRX!J9WNbXSbQeY}is
zBq|T_v^bbi<krhZ*L2zx-rHGXga6utVySEu>OY50SkvdpPBRGcMi($S-j4b))Y)|F
zty?npE7TRigm$els@2*0^=kJ1@X6VNV$F9t2mEZxALH-rm=!I7d3OTG0Msuvo?L>!
z3K>P3UO@9xN#@gw4+WddfX0GQhhARQ6to^kh=9j*bau3y|GrXeN2h9MX7zb93ApG-
zBK6-cpTbbBK5}fnQWWstppO0}`Qr*Vu_&{rLDl#x(-VCy77GRW^U=6XJ)>AarhobU
z{YKxl+p3H6rR5p>gjsljcmj9p1VO~_U?oJ$xA%H6CA}KVzO$+^t}3EF)9~C7l>tsf
zL!jk@kR*BY<xj=ffQAfwKvP)%>kIW`rN%d3RRh9JL@hS^447aZaXMB%&3cw4&-Phg
z!;g3*AHR&R>HmqG5u&S;<OO{$u+p-4^BBc$jfRiOyH6<cnprhGphMp|-SQ^q)r=YX
zM0_A@E2@Q50=gWS8d8x>Pq}2JKhQbN?i?H}pDp#f;jUJuB!Ua#0LAqnE5Av`=<c0v
z9rY{CZBE-RTjAN6!F$7zxAqZ#Rgq{f3Cd}|$R^r7u-~xfAw&M-8B6F`0`ua0;t2e6
zoUV2bC!*!^qS@G)uG#sLvK!ZgR}RImHnT1GWub~()TH#GC&bs-_}&wryZoLg9NSUO
zh^Rz$rur=mc?vVe$!9u*^N6kf6eoeWD{F|R*+i7VYAnr5>Q-mo8Xx+`K7!CYA|#o`
zFosTZwb>&4cSMqpf0Sna^yqiQ*1;dGvNLk~&g<GVm>*Y7d9d%F37<OP_FSK*NTUad
zG?$N~(;}q=r^MGmdSwpk>P6vLrQ@QyOp6&nDiYD*%(MC#7SWa4GrVF1?QdNUqU_Dw
zc50ovTN?b#4XlTSx9wF%R8H&?Vd~3HK3^UOD}ns13XL>aR6GBkQ+CeMVWG5iAGNwX
z;Cl~(aa=jq45t}-6+Fv^&{M3lFN1w)S%=!{DJq0v<8K~wH7g$1;a@9lNVHdIWyk4R
zB=ZRsjM3+=Ltf(`yhUCkHWDKNZ^!?>K(@zMqtZKP5Sw6nk-zBFbbbc!>Dl~Kec6Hy
z5Xee-?rE%n?!1_fnEGD1sm6gjeea>@X3!a$dUWng#<f}c9go0dklGK>V1^Hm!(Wkf
z0c~HCqu!KB3&WfPjb5cg#u&Sz2Mv_Qk?oyNZk)^mdW7B8+AP4;8)7ua4}8MZpJ^rY
ztj~eiE%gFUJ+00(tfMmR$;kql%J|@Zm0mrlO~Xq0lz-k2^;!I&mq48Ul4sU;&`6)9
zZi}y{lnmWx^lAv4j*QX&im+a1wJscK?xWatiCE}SmbL!D*EdSGC*kMLMquY$t*!>9
zsVP408QT&P>3}=jo6`}hp6{bq>lotXus8AOYVd(n?c>TG`ZM>KsUgkfkmx0yd9@7k
zcw*Z3TitzC7yuv^+vttdeccn$HM~B+yh_&M1UkvFj3>Lu@JLUX*ZI5z*uQtA(7K=I
zK0P!%m3eo5I+;W4M@Eh8t7Mhcqc>wH5l!jk(T{><4ytV2mVCo_pK#%;c+M|2$XX&E
zVddGksy{Q?Suz|#<LTGBRoTS!>%*0Gclkmu;2-mdb<`bDtwXgQr|D|xP4Ds@u&jjK
zx+%_zVY1B)B$G+SpZoEYoVW}_es(SuIDjPNWIXQ}2gDGOn*ikB^yF%9zYS{xp;Do`
zTSlRDg);P6cP*N$hNCD;@bilm_I8#{-W~t&8)nsY=!t@TZ_H730D~?mZAG;an0EdY
zR3ijVv(8z&p;{`^_PKyzC-=TQ@%n^}s{u@kyj5sfNp`i03xZg!nb95Y@lea&Vyg1{
zdpXz;+wQ%jjH$j2(b-brR;nLq68$RNGf=IoXV-Fmp*;Oi7AAQ8!Oqf;d3!nIQ~KsF
zB3C;^r@~};Tpu*^+<fv{QpXo%%`c@2<aFU9x!nrHa~unBQCQAf+VDvVVyn|2Hr;$g
zf}6d2g2+AOgn04-J`}J174kw-tm}Y2`JO9FGy45TM9|Nst47maHG2??Q|i5W|Aa}t
zvt)T0=-6e$fs;=Q3q76+-~N$jNgmxNY#z_M#w7D(yBQZhGV&#reUbIMM=+g!-0Nea
zels_MmwD<la7wp8+?GxpWSlMO)X5uV$tUNOBzyb5yy9BqdA$DHuj;#iL5Jx?-y6{e
zkC+g8n-x9QFNH~&$ZOuMq2;K1ZgM5b=WiE}f>0mk#WxaV{e8gCg6<J(XMPi9M4O{^
zh|Z16(x6Z1S%{9L5Nj5uDE5%o^!B^>S>r=NinD9<*2!X83R+h%fM8xIRHy5e1k=$m
zrG0{({e_p_G-cn=r<|hEa)j1@FahpA6VbAueT0FS=ZVRbc20%z{?rEl;W5gbrQiwq
z^fD9)F;VkzCbInn4Az@Bx%Wen&n|CtQxC*nxH|9xPI;#u#ZhNv))SPCgmSS(_4>O*
z6pKZ&)igH+o3F?YgQoOfy@42GxGi9@+I>mx`aZm=O{cKroW*$?TrjQ&V1oW|tlVgr
z4BffdWN#1Sx{_l+d5c@bx-l(-7x2!n>P%E4mH8Jwc|^-|Fn`Ef<V#b!Oe3C$?QwyF
z8~8FbkQp|1%c{_{;QF*&Ueazt-Vfk~R~^rRvjO%SQHVv2vfsGU2}(n1(UoR5PI=#X
zprQ8Q+4^&-C)E0?*OE_V=%(tS%oQ%$Kd{JqA$rd2mf#p`B$hc;w=%MeCHHc3<{O75
z_q02i!uOuR?T;&o$gV1Jf$5vOZUz4&!-nIOZR&a@xl`}uJGCbZ;aF_lhZ>XR#S3KK
z)(}SH&zI~*md7_A<anZ46qE*H<pyFsfpbuvmvtLruP+6a$ZO&mKENkvA4Ya}2j&~A
zkEUh$=Uw}L3Qf`5c>dZ#K()=co1!iTL`85hZ?W|jtd&`*H`vJI)d}ai0^x8+TmJyD
z7LS8tcA|h|z&u4>RjCFE;)id@wSD~)eB2GFRt7ni%;5^^4+FMeHadF=cJ2~RhxI;=
zxmb?^wZr|AeYXpDf7L#?f|88B0Qj$(&^1lY!K|gbuPOd@h0;R4QrD>Gdmnz2{tYYH
zc5xes_{(?W1vp(aXWy^pB%kpdTRCUPq@hyBv<u1u<kK@a1eo!cOmsO1CUL_7Zljy7
z9#7HTmPPds%MAt`*T6ht8p$0$#H}HfsAAqUIMoXdo@(<(C0FFstY#cJ+Fa&>nR0~H
zvHE#&<^BA-|Hv;rd7u%z$IuB}S|xSOAb-vMKYbm5*E;;C&={2`3}{SGkW1T_D2tkd
z=>h=6*8t{qBuHtAHvd%asEal$8*~U1IW7QJc2VhZ?OICM_U+;2?*kVQJslgCWmk7`
zkAo*(k~uU9dh^lcsu%SvoAgP&_TRi=f_-%#n!=`kCIBa#mFA?=9A{#}p)2CI#*bgJ
zSoI7?zVJ~8eMJkNoP5*B_TIvSxZU2RYjd#=hyk9_ez9LXuS;=IBpe07JnxI4;}V#Q
zE*$HCR~SdNxa&VZr#VeHUNpDZR3;Ro)nWk3$WBM~Q8Y&lF`L35P$~q{-3&$F13$bm
zQf0@kYbXaxZ=U(ML@dn{Zds-dn$o@E0-ds~&odt<^^Vv3rJ)j~o2l_jwqhI$E2c@@
zhLl%>&2G8%4L8gW=c6A&N=RBhVVgjacaepN)oFuEmLx^DnQpT2oEx{e84QG~PHb~d
zMIkSMqxceA(V*p6|IX#ThqbQyHkz`K$&%1a59$|ub{^pWp8Q7w4X0}qv-H@XO!k}W
zeg&%K#dlOGo4dBf3LCWF1(19;dyEC2y%cJj!}+SDM|S8odaKbG<*EBrF$j0w79{KJ
z*+0e|vW+}<&O;9-p8lLPTCwt1?AcQ9tt8XEhr_vYjW~0lBS6{WM-KECEG21LG8<O<
zi#A@lf2qY<%Lu;wNQi4==FT}AW<Fhi?DAS!I{%GE@y*K{|9U4T9#JDaAnlWjuikkL
zArYz`Hj0$-xiNXnEY&|_G^j?ozUSd_B}v&;P1d4dlSaPPJKJ2IMn~c7%){F@c|NT_
z6_7)T>Jm8J1;HB-ceZmD^DirOH}(fHf@agceI=do0y^UpTT%32U^}F)*@PDvel>M|
z02}k|!{WuYz217ivI;9A(Yc6EUbA$XE#+0upo2jN@0_&XLyO9QpGOgQI+8b%8SStz
z?Z0`rp|0Gkz}D4BjjKD=%|Tff0cKK_+qaau>9N6&u)Oq?XOSJKU$F@Q(xC3(IUlSx
zn@xL=h`Z4J^7a9fiag;JFkc}Zh_hd`Icu}~W1f8gcI({GoiC5L%9ts5IEm>Hc4q3s
zrBM40-?8GyX<B*LCdXHgv#RQ~T9kW))|$<(Vl|sG^cyrKCIvg0e4b<4!!YS~kp{H`
zU(GGLN+<q@{=)wXD*a!d{WtEv|8;NwAMQ*{1Wf~u#%~z!AwC+A_T(x6t7<<wM|PC}
zrYs3S839a^pspWU+p|xY0p7aj7cVO)Qa}eWIgJ#m&P4u?@Z+C0#(#h<)*9X!FMYZv
z#v=ZL0V*l}gGmEw-is$|EjUALLN2pBfW}G;9Zx7CMqc4WvzlvhJIbB=){?V#9D41g
z@{kWqz-$K~z`VjR$eS;5=11Q?)zH+^cpur~?WQ{Z#z<CehucSDMLJuQv$5Qk?ax!s
z(ivnZ>bDWdL73s&A*$WCGu1774hAt*PrtYVdj}I-xEYRA1jaZB=2^ZXEja?WlV_BC
zyCZSbuxi%{kD~KpBimAgSR>T|{0PC3kx&KxYa`7N$!{qXAMN*9Mmu@<pC3-7%-G;4
z0?M(I7{VuF!?d2T<}2IclnzED@FuQn-A*3OjY4x?^<5ug$$-Vqws-N@Y%SNSnR|o>
z@2NcDXNn2-nsK~+XMMr<!V=bmmdkwhu;T2YE_!*J^cS8P&IGFUGTGf)80yF@i}8K0
z<f}6zs=vi++}nCg&-mU8c+66jej^%zQ)|R*{M=V*^?pR!_E2Nir}mj_Rjv^FBdG}D
zN$1iDpoXf|#~)N-<;JP5v)WGk)OejUZttxazO1{D&XXe7jVK|vRH8=tcUR7SBv?l5
zn|aR0GJL!IH=_cSU_1J!@w|f&C{a2(Ux6g74gtJEj4B!kALVN`<tAB!J~&u+ZiH5p
z`1fc?))k$OEF<$|p8sNp#LPD<OQsM*x43^}wyw-r?+{tJ{LDRr5122}csg=zyAGPS
z&Ga{{5v9WN|BI>F0FS9w=?GK`F|91A`1tsYCL|^jky`0Da~zOrL2kFcieIpdh(gL1
zREoYEuB_h5+EZ8a@hf#tNo|d^`_i^;&1W$yLJsbBUrp5|Vnlbw3!L_1$l{#1Z!&lE
zOSQ978-^8Z*Gr+hnronKjSO;>vP|Dq)wm_jg*byfkJo+V0S7|%ADwQc*QowXFcTW>
zk8@{<CK$<h6ih7~>MPH%5|(<#h*?*a5f2jtA^rVh1<D*W_X2jP9w*nW(z#@4uuG0@
zmwMi~BYVpct0}8#G~f!<f+EHp9|b9pJe?pZY+G&6-Jivj!XC+Bqtp%i-UPju4QJ@G
zfZZkV2C@ZaPl08ByR$Ck3&ex~90>XT!+J>Un(zcZiVCX9dicHas=l{0NQ}b438oOc
zy$gq1Z?_oYJdJCJ-w+%PQQqTuF{u9$2^a*PCMiO+V#ETfau2{6&;N$$yx6=b*a)5h
z!UPG_NK(*5;94R2F_L+T#93tl)$EM~Ak`=K`oPo|;et)z+jsld|AuilwgPXK43h1C
z^w&Q|RU1KWt)}r7h;6n*KTEi1bUcD-PFgkawH4F=;7pfDS%dM^L55xgwRQpnCeX6z
zpj3k+M`%H<3V%>h3*Pf;qyNbpfKOE*VO5w4^^5)8^dvRmR}YNr2Mvlzz}B?c{;zxh
zG{b`T*#B=Vm+&2AnZyCAz^1ttTK-9DLEVFD9jiBj(!7aJu!#5Fy>6whTd`F^{t!#5
z$iG9bJ!JmhfU(^?Z+eKPBG&doN69|9pI-&TFH`qU)FdM>15VX1u2`r`Carl^lWudt
zeW8$;>wIK3uI7?QpoHx>x>KpDAvlO-yD4s0ifJ~{{QQxP#Gm#I`EXriBuUQb1U<8>
z!)N;i#-Z`ADZJAbG&9^9uAa2*6=#U6$FaX`KCD|HY(qeXgdW)PaJdl{GK{+m2Gvy)
zUA!m{(N!si(KgM)Mg-h+)FwZaaec&9>RppQ+hGGCsm$~6BvFR^tmgZY!hXta8`T8E
zUC*NHh)-8P=m_<IgB-w#>o+1OCRC@-(Oz9VO~lY4m4L@L=_#VLJH+Ae!-tQd8bHN#
z+BO2o1X#A$lM*iX+~$%|v5Z@L4?7NtUln*(5B^iwX@H~^qZh1rV3?s(f5T=Tuj!+o
z5txpetEvJ}77<$^_GuVzfd4>nXo4z?Gr^*pTy|Q==@jZQzCHD|4}DU*Ln5(%Dh>>Q
z7Ty-JoQbN6pb`wzcw#dliXgsBO;R@MHrhjPIUnX^_pmKm^+xtlvMN)3N*JsR>6|$+
z#GoFUqW}skSz!3m4gy#z=fFp;1Y4Gj)qrlsJwP!+iH!ojIWk7W*UKNyIX%jz?O#Y+
z)XG;L+{r>8An~NV7<DhzUpo{9p-WGJS&K)hyoYLaNY`xPO<k8cDxucA8!o0hkq_-e
zby*dRvIg={<Y-vFtw+Gy%VRm2FTGnltUZnn7YjX{v%%#6+!}J7BzYkue_)Fm1Bkr6
zcPmdwlFoU<C9<F6`*k@BrPWA&W@VTN!wHRbYmn1?GJe}#^kt1-qDc(N7IH;%4jd#&
z>sI&+^Dd@s3iF#Ry@Zjn7;n}NN~)Q-VZt)%5S_HaqGxki{Q`w~b|_0N-c@Yx%MXr@
zO->T-k7gv0!`~tC72PXeHyVhhTt+sMBLrJHbUJkgDSt`H^z3lOS9lr05hl1%sYWMD
z^RS$kuGvGU*K*@3Puv+VNSFMUs%NDt*g(4eho4`oC6MB2GTR@pHpC+0E6v6ujh81~
zOYXN`xN!XnV%#~lj?7Ub))FHLQW{1WdZ98~@@CIZx#!H*9bP^AcWhrf>CGrRxzO_Q
zg+OhmIE*!H&o<!b7dj+3k%cCUj&IPT$4*J&!oU{+C%TMkC5}JM*?JUNe~D`Qshbp%
zcdg1b2DaJ#qH5`h8sNy+V{>cND~e~}tGbpRr&Dn%(IYB7C5Bg&v8uwZ<bO8VI0dce
z-2yxdj&*irZ;8zjr%M{Xv=m;mYt4@huoZxz?}OZsz(}nYPf@Si-uDED$!uAgeDqwx
z{#bSxR+qn{Y4jAz{;}%=%<4Uk*S^Tn;AnUJD<5_NDXJ}AkZGiO!>`gX=i9J*nK?2(
z9>xQ#%m{L*)Pg6I+|ah|;yx{Y=3Tq6Rfhz1M9tMoyg&>NJXf|L@hBIm&yP%VIT`De
zeuTfWeEd|j2beLP13gu#dR9Ig#|n<!PMwu+_fqXsQ5d@ME=Uh4vkDadv|$iS=(dfL
z`gYRBg+CMC60?dTxnGV4@`0{paw>jYAkUfVAN8*%!?Ew%5FWiVj0(JRqPFAneJojG
zRH**j>EQNKvTbSVl;`)u_ut?r_6aYb1e>&3c^DT%CI~AD4wg2`nq`+)eHSWDXT!Kj
zC4eos9zYm=C1@F*b;W(1!<ua8Pc2nDW^Yb&0?@*tH(BE=P2^;kU474Q-Cld1!*Z1Q
zzWYduwBQtK9-6^015Gvx$R<+ErL<8=J4e;BF$w=C$+qM1uqE<+7u}DXTeK|yhVn=8
zX&D}`)6Xb;BwbAtH~y}L{!X0L;+_EQtS6k0Z?TzwJ@?*~VJ}DUv~62&nA{-TFItEL
zkZY|=wF(mCt)Al@8aW{$?N`@is`Yk0>8uW1+tEKxozbTisy7pWo}qa5&a7MK#qpeS
zOslMlc-}Yd&kV!x7x2%>tD2lWxTJinmkq0|F;BGuGf~-Mq`!0DyWDKc6B3Ha&ts|?
z=z$8_r-{PXc$jy#t!pHl+&x`^EWB%k-cqR|VYMAgF#GHHR-ru88U*tI^gm=uNCU_h
ze2Gb{TI`&WJ@f82@8FB)<O#<m+IPG&I2e^WR8^I~*v<CjfF)A!4Uz}+tjWc;9b4MR
zeDtnl6`6h(+ls#|F1Sw}Rf~F^^ngP!J5Vnw*t_L7jCFy>6%4ybb%BP%InGG^Nerqp
zW8AY8qmN3ANxDRP+Y~GV8n(grj9fh!^BX1rc$VeTZC{)`srPJl^UG2~z%$T?VqSzN
zX5j1Q2YYq+dzT9}jm(1iS!$04pncieh;tpcY{0NS9eN_Yk@{)2^ywYLA87C3?k0!t
z0aC6Xa4;0028i6}xV^<z(ERYk$+U&bU}FjDYQO7-Ak&Zc{A*H<Bc72ykZ0-<vo@?(
zl5qRV)b3NIN$=<Po^@u^W`4u&-D<~mWA)j)IO;T~jlRX?bpx7d2B|0KzK1+ao5C>j
z+0Ojz64D?v`i79GgYBD@-M%S*KfF!HCaHdab_2An5p@vf095OLKd3Ljr~GxbeBU7$
z;%a%^u&HwPw@szthM=xs+7cpu!?qS1=37;&`@g2Pyf%C5kS2X{fLr5NySANYz<+JB
zVTx3QZ<6$?J@zdBVBeunwMJg^Ao}U}5-z*3h!fNGI3wnKqK50|&XNjAbA$Aon?j<3
zc=uWgh+B9HO#OrsMj__8OqXubnLOKKv_rG@6$zn5VM@5kZEy=+crHDH1A=emp^o)h
zRUNs<+<nzXu;);2(szt)O?sEe|6s*|K_Et25_UWWk1qD#<dhkvX!Wt{irYh_r1F!m
z`CG-ma~^{x>K{|LW%@tdHnUmsyhrWUf;!-Q$3S<^|LNkZB5RRsiG)?uVl-fz@6Sl*
zM~{KZu8EXQ9i$2@p$KS3L4M%{+?ZccZH$~*;Tld;mfwp0Rne4nP1rSMT;kPn<lr<q
zg0vY3WNh&uqUy}U2P&pacJ=#x_jvm0d-K5=+io=`aiW<aT0`9!8q{HL!HRiKNs7>0
zPBi&(cmmTV03);(Sq+ZBiH0W`slRp^ntl)?W!%WH-Vt(7)F-9s#EZ@LG(1o%(B|ua
zq2XY2!8I*xuFIw7thX*{>z~sOlxk%xYq+uXIl4s0O-}Pd-bJ;E*XxOhNYp2r0NGFS
zX{X|ll`}wbl-Xask7H4<_vx~a%bu>;^T<l&1!HR3t>cKtpb10e?465brq`Acq)m3L
zre)Y<&h^|5{;7_QgB{b_6ZhkIE%7nw3N^toy-t&Ysgo}cxvFt~%J|Uit=brr-H*F0
zbPlyNefA5}x&gh;$e)&1iBf)9;FIQZO7DSF3w`MBGdW>cz>oeh(V8LrC(!<IZ{=;T
zs?g-~fU&l2cAS*M*$Xjdp#X^&c?d*nw>ptChmlK8sNoEW9ACKYgXIwHfZ%{S9mzRJ
z;AK7s36LvDR$fuUk=t*>WiXBhE>pbaZw>Lkgg$jE<90<hjTjg#5qn4jZ2~cC`mAYP
zhHYZ?f%%^Co-0oGi-jI1sklBqR1+f^*+Ps!x*;|+GVm1p8n#TGT7>aQ_4K3|)AfNz
zRk}#oPhMH7STf9#b%ZYbbp>_|ST+7?g=hpS%(H9r6Wor;Y41c4hdnJT-)H)N%Ja6`
zvc*dI89Ak_;3$t&Xd~Kcyv=_fNRsx83H_x^TkIp<m>vTu3luxlfc`jt|K*9r+B=vV
za4sHDX#jf9JdI2WMoI$V<R+w~oM$)h(yPWE!ea%~qt}Cs+;q8b(BDO8+;F$alo0xI
z%TcW<;bn7KzEaK5%fz)jz&N)01M^)P(E+4KL+3y91h{QDU2J#5b_-6;1`8_HzzFvR
zwl%8dkH6zby@}Vy4?CZ@Ilm*J*V3&$>>gKAP@h4Xy*IjvEQhVI`%=t&QxI-_o2pZm
zdMoP8vdiJKSYlEMesvQ7xch@aZIT3?LVqBK(>S-dS{!XlAk6U6mlzmNjH(8Xf(jH(
zp3D$UF77oy0vX$<t+La>J<ZD6GR$h2mrp}*_M>Km>BI<}V&o{ie#qgVj^}45Orf&z
z_YZ#SSHn}!ju`gOTnL>6Lt=fGTQA8S2>R|27Uo@5r<cXA2{WDO#3oz$(A}BmAJ%0w
zP^^9QW-)=|P@CnT<xau+ek%F{U{oHfA0@`jnCuQc>pj**glBnWFNpM<TBkI$GjXwt
zWQ0?iY%2&o&TNl-k#eZS@S4NMmR)b`ET+-j=}|nSFC8iAouT{M3Lx0!Cw2rR9*Evn
zqx?vSRz_l;r4F(e%>)1&iVk*@5Po#+=c8j~d4_BKpjEI{k*$UNL`W`pUmPdsr17tp
z4Qw~%wv5?rYO;9IQrnJOL;V6kKK`VQHC;4vejv0@e(?zCCp-q-SS7Fy9^~2e?gUk4
zJWpg#r=TIDVJe`e8E<rKmx@3bP1NOpaI>CJmC?=4`E-fk(|+9z3j9%tn$wKeO`%gI
zeaiDC>zewUQsH6uo!0C`#{p9%n_b<bX><Yz7ZU2$;$p2Mbkk88fc)P7QEla5h_5=O
z+K!(Dx!G>P^^F`(8=>6F!TQxUvy*cZpDad}2yN4de1>|&^JEzT#Y@%bZ~$26919i`
zpD{wC?ydFDHS=W3=zE_siEK3UZE=~Dk7L6+?zsmNcT&e6AzH*?ADMuwH_g<i4@bXq
z;vDG%9lYejsZ;W2XH^?EUrHKKU>6_~kxUyz1W9H|<D@!bvl%_<CVgYV#gAT^ia~~(
ze36gCT+LvR9<n%h=GQ0$BAwBQ0t>c&QRDHL>dbSa#3Ro_61n}8RGxG!yOP8q&@%}!
zBL^zww>Kq7mlA@t1q-7xDdXIkP0pe`*<bX{nWOXJcG%BO!o+Lo0nM^aM@~t8E*D%u
z{e;gTS<<#3V=C^C)O~$3pB?Ep{d}9{o5RpNK*SQ&St@`N$o<PX^e;|<K%~=O$w-_2
z8#dfFj%*Fqf~8XDp2hQ%IBywn4hk&Tp{Aq8FTZDct_azlxrc7wTxQ&yzzYmhw<HJT
zO>H*SYHe!?0Q973=#P=*j7R|2W*AWFQ$0U1b)WrI8Rv7q$j@8M80?5d=Q|>1w3+(#
z<!3T{@%4NYLo0bqzD}Z3PpXq{hoB7szfFxn!*~j>CezxkN1lzjQ<v!flyQD0td)p*
z0dMyo(iDiDqla17p>TUR7GCnP%Qna4pZo5p*yeZ6{O7JGsoP*T41bb#>xxvwj2F(I
zYBMIyCEoJ7jTP)r2HTrW-RX2Ux3Fue7rEpvHHjv#@V2JYjnnJUl!xe&?(VN;y^nnA
z4j<0*@N5z8j)!p{3N}6kK{mLV%2tR4z9QTCN~vZ^3d0ev64TURa7W%K80dSURwdR7
zKgJa-OXk&m{A{B@L_Hw_cc|a}=)(0qNlu+vrz2N0V7y51W&*B%JHAOXKD}C`%Wp(e
z3>!p$p#^e}Jw3Ds?k!V?1CpOz@E<^cffmTM@oL2F`%JK6Iv3RJZCE8fk$DVquFWUb
zM>d9HFiO-o-HDWOZ5Nlb!dBYLo<^xQkR?)Nu2PgqrU%FH40EQ}l-`uQdF1vrc}-U6
zYechRmDLLggt(v{%)ClPH@35Ojbo@K3R9FG1l_a_ByQ)wLekQZn@sj#&oepdRtdG$
z`!2W1wpVCd7&4-6G=z)IiG&3JKWxQ1Xi4W9m2()}llW|boPHE&re5oWn64o7fttLX
z9ABvgZp|QZw5Qyc{?L-87kT@Wx_QIR*GE1q+lS*~NOz&DNQw!P;zsjJA2k?|^c>(P
zoI4x!Exv-F9ILy^_!B}wr4p;kqDF?IUb;-K^dEaIV-~XK8LMRwf5m3VYW8{cV}HTM
z64p9j<fv_;k>)K!MfHmH86!KGdlpUHIA9qss6O^@iSyJ%<XA2SV#{a#hFR!BXP*D3
zufrg7LXjp9yf+-0kr5R{x@5X|aKReaO5%tLY7avbx|9EOi5^ArWKG$HUpR%Vi~g{Q
zvQ*6&BZFP>?)JsjL1GTt$?pzp0n6@fMrT5mfO@&St90<-?TV+_2F>4hUnRX)U(N0<
zNg(XNb}`OZ8@?~Grs`eYeDnJ%@_V?+XMQ+|Mp%K0ho`Z*4MirIv;?ubPkH?DX$hrw
zaHjH!E12a!sNDxr&Af!SZSEzku&wnSd1-%dCBl!?VN5aqxj@02y3u4a)~Sx-zFRRw
zT}PDrt$k6t`OD8Lxp3q$62_mi2Oec)Pk*|Mk1Ex|Vq8iRTltwkwIU`GA1-R`vF#!u
zbngc$9h!1Jz|XNd`1)E>Wz?wN`Use@0dWdVl>M7;a-L(K%hTt-IYeA)3&8ji&Df+p
z8>Adg=5H|ut-C*<uGMcCiUsGlL42|N<&z|U5Ypn1Q-7$rU;F`Pql?P}2%9h032Lp!
zl7VU1YX7W1PX;frv8aRQpqYGNYF<*td#0ZVj6VP~H46up;M${rS=ke+Cio@zS}gPn
z!Muo$P#Z<g@J9B0D1pD@+&p(nb)Z%5g;u~)>jo02VXh#OglU9X4SLr3(TmQ<7EE1@
z9O_6Jdi#+ld-*WzP+gFV5!gGCK5P~|Fcn0`9AOw-+@{ML40U0ieZxdEl)t=c+vM<L
z((z&I^^HwiEFJgqKz%uxb?NZ>9g!`&(<)8bTAVNPP5y)Zb=xODbI-mg4H%-9-b#o*
z_Zj!9>dTV@sb^tRHZWWuIC(dL4nl&rzq3-r>sWC)Z+!l${Z|VG)Ja@UwINt6F8tXT
zDLR{L<$yt*)B3<Hm$VJpKhU*7T;>OFSRJH~5%)s@4K=!(YlBT*7`dpo-3jt&RK^|e
zNHcd`Rw=|~fbZCc6C=bBik>O&tcb=9kH&3yk$SkkV9&{Dwt^#!>tZFnad6BdSx3Rs
zEgUw7VP*=%&vBi6ASK*ED%R<2c`9NfCur3AT486k!_V^<#*R*H-k^>CfGml{%h1U)
zT!!!EkIAMv?fx=vYaNvJE&y!Sz)`YbWDm<DN*kG_J5CW(eOmMVmtukpaIT=BKY(=V
zJw!P&Y~;mI$nD}Ot*>vnXaB25t5q0T+vBd>yOvHT4B%<$QP(A(Kcy)h+vX5dJv#R%
zLKFr<63g&)q590tvGhXy21li;=UF}vUI$Q5zB}4lGdyK;o67vi12M?d93b0~8zIg+
z7-x~4^7pOO<|NS#pBf3jX{60>wJ5*z`rhEHFaC2Tqe1upH2!W9J`w3jOd$KJv}SYi
z>&xPsT~#n|B}`r<;sW_906lJI68MtaH3br2ypHCxxPnbmbSm6}b*oOWhsp1ap7!ZG
zqPTiJ*Gla1<pXmRTW0c-(3&?_NzhRW!f<X4U$N2c`rZ2*`qmEeMs8nG*E3vq6XVZ*
z&3UX}hX_~JV|lzD-88Fz_KxpR;?tur*xl!tMPS|gmbK0aEyH2dm1$Qvk=JWe>Isi8
z%TreqQB-7Ue|-fR41*#>@{*0-UV{BY+n|y-3u-_z@?eK3&A)zL+=o;w+GL||nXzC|
z-CrUrGkkjJ>4hJDeqlnB=r3q7h?YeX5y)TDODm*f!u&5-_ImZqZ}NW{8&>hbfLQCl
z5u1(TI5DaWoj!P$^e$m~pPB}x>VNN3YBJwQ9PkG_v7tbQa;UR#7bpB7*7ni`skQ@6
z-Dc{HbVbzodiE^XSPWalRtFAwHYQHSwl$Q{`_sibAt~uUw%qM!r1Q3JVjKLuduO!!
z?fl3<y3Gxvj1ViqC%w!5=4*0{9&vsHQ?rRI9BSE}T?OJnT=De8Yi>R}8l)xFR^&-X
zlOr4q9euye8<?9Nigtbf=B<V}_4ucEzZ7bAF4usR0pjeQZ_+r=3cbLbTOKaXztoiN
z`Q2)r`IUn3=g%ZYN%qsvne+9eP1uvBd!y-9YUU4;-pm}#j!Kz)e9l?&KU7Km@2*n*
zuj~K+R;m6UH0{-w>PVoCK`R5^c4tlGL2J%&9vFhYfrUbUl05)MyzhX1AlIb(0fAl=
zL<LRc*mN;s1ulpx;nM%!76}0sx!S)-V_$i0psyo`UbB0=D#aCj_s3s#|3$?|)91zz
zD?lj(zUALZh2r>;&=^y&6&mnC@<jmI>=0nB2)B|j@KrtPeJXnc;Say;mm8hRdkmKV
z=XP}nq+c4)DL|o%<z@Oe_;ZDAfo`II@lT0$IyMHH(eedueQDvy9I^!1b`U6Y`pbcQ
zSpzKOc|UCQXVLFphesv-rPl4!KbIbZ9CPl5riYP_fyGzt)k?T#4C$IA0)^CZ?2Wvv
zE>v@X9<V$bZeqr{!-S8jH6|#`8zJaQ)(v_els>h&$<bnu{>qq!=<(IZpw7cAQGqa?
zMJxiB84qnKyl!O8z>AHz?6wWtDl;%tB=OPtr%g6#o9^0RF>tgJ+yF>@yx_lm@HcFE
z^4}e3V*c8aEXuw|<967wpKEhwF?}O?X;zBi@3K_|C&Q7?sGO_BWN7KkpSm`j!NaU*
zvJWr?6?AB@qVAYioX%u%;nk774uyv{D4Dn=3S_?lme)DtJ(t~2Uqv=z{4T`HRV1F$
zeipd5I=(Y7!FG#x&{}j<zHwHaXuR3P*Sk{3<@5fz1FrxF(Q?oWccCq$=|Ym=VKe#-
znrR7N)#m#DV^bV!bOwoOueoZBypLY)qXMQggV;snsEYEQ0ynf6B9NVd-AB|}Wr9uM
zWcP#sfycP!;%R;=u^HlM*hCHiBP*L)ZqWc4U;DsL)B}t>Yk(RNL8GpJBWdXv|01y!
z!P&j{MLPo-r+12(5m=ecdW4@u7xutV0C$objOFbTz%DQHU+!w=fAZlQ6I*RL8eR#7
zIV`nEBOTS>FzXm(nGQ7bAy9M?*e^l<zU|Dj10zHiB<O8E5Eyo=g}D9>Q97C@n<+S(
z%n($G3SxTM2%c1ez>*vfO|K!xmRZ04lUG3m-&F8_sqC3vyVeJd!9gSIGB`ERhXtj<
zcH~$-HdjD;4fa~#i{P5TeDpKf1VAiZ!O5z3l_<78n2P~SH2)59L3Z{pA<j6cbdovB
zM!GC@bI!e7#}rIKoxL3avG8L{9&$bf6#7zlhQ)x9*{E&o-DvXM9K)$lZ^v)gwjW)B
zCqyvcf%<8~RZcHgo0DlfSiNW+SK?%s{nLB5`7R64d5q(#hn^|tsBuqX?HhKv?DzFI
zD}IMPplaf@&bSERD)U~m&a40vF`h6Bicj-MUh8TVG)2)`_tW;ROry!<TJR7@fP>iZ
z%ZYs9%X%N*sW^-Ly6(U4h!z``!QTyT_xDiOD6DPmh{&fe>#SQh6z)nQ_0?Q>3`s1y
z1cMKI|G@;rL>p{EMs<4}VoxQ^92S!w_j+vg!#JP}$ihmo`t@?v;>_px7Y-7R-f1b6
zQ(E}PTleuru6O=pnU!5_1^4I`;3(PBHP?9QfMY@IEa#1Z@9u<J)tKpg92jPw>_ox2
zGHuAKhVw#Tz)_cU`6B&`li(D7fLmV8MDW3bGDUYd{~gn@sM$#FQLI1vKF74sd#&V=
z@4G=tTR<*VMEUyp18L&$mpg^gWRc~pmB>(QjpE$a9-X1A0z1Ov&I&{xagsbLR|b#a
ziH;gy4Eig+?d<2QglF(Hy`y~f)hC?jQB@X{960Q!OUZ~HD?9MHNmvW`ux`OMBe!be
zv~ca!qO6NKQ<SXnv_nDcz##OP$lazY;w7w`d1Y^);ggvejosck8Rm8G?xkXW0@oIo
zSd$d)PR=3k70sdDcOE<CInN;9{EaE5aR%Q_C@V3^;kPrJOG|scJ)E*uls%MR-1J51
zi;hqeu8Y9DWOJTzzDEZc?Rfj}B^4dy5GwpBLUg{d<g;k0S*Y_UddQHGp7a`90>59r
z{Ox?H+@RO0c?e7dgKIz(ZR5YiY1Pw*GUGIECmd}Y=A|6^xe)0l{Cq5)|DA*39s>F@
z+uV9@`fu3ANoK*dtwt=dC*1MHb{I~40tvb-zhOxz>kv1(B-EgxhV*tn2s3WoTNT0m
z<E@20!NI)ev^jt2)-iP6+ezG0RH6Cn=(*GvLWz7E-1OiiyZr9;lG2z#WQByQ=p@**
zEr^pgRFC-N-|#f|iP=HQnU3RdmZb#@)Q$7Z@laBnvVu!Rppz@8`qM}`#0_BT=Dr3l
zw9pzLVwlo-O-KaxrhRXqI%u%Ayu++{GuH!@nrR-bxiT~A&%156f&ULO#l2}Jn*FF>
z`s?7U&Jy1;_5wqkymNRb!rTY!sImAPCfFp*v=_+uM+7PBmE5mdiOHd~Z6AMb;PBTk
zrQ=Q4*7ftG5sVY8=9&tk7<%|l1|jy>ehG(3O!TI;ht9D)Srvu@2gTL%+3x$eE~ju*
zOs>K7nOxb8I#XW$B0(T#&$=a$Et@rTeQh{X#yaJ8<|g&_j`{U#VU%ek?E_lQB2j?6
zguSS+5nr#@zCJHEV!=0*5Naoy0@WG;@5dx`ac*p(rbgo0p=UoASRWr`o^a<4nn5S(
z!-3okg!vuSVt8@h{yvkP&L=z3m_@}?4)PMa<KiTYlc)GHFW_xtjz(5P>nqMWS<%*8
z25xiWqa;n9D08S0u~ip8LlnQ3es?ZAiRxKoL&jf5O*ix8oN9`EsKUkgtvkOtUYI%#
zIEKA3@JTcQu<l6r?Lv+C1x5;iVZlmi>4Dl#KibrN?Ie85)c7;F+d^MKM_M)%^v+(n
z0yjKS1j7pGgix$*L0s&8BojmrBD#|@8z^!3gu>Peix)58zu<yxL`E)O2EGNNApzM5
z|L=n6njxh+qB@KfJ3GX@Ra1@Z;VBhh6J{vipG~C;f8DT0RXBXLcKWz7fUDMLavT`=
zXq*_!guu|0q++qHM=HfaGp)y-(LAeXSfa5@^Rm~<GRzCTWd)gc6k<jxu8&gSc4k*Y
z4LC39g!r|MeWB`a>+`+e#D~4wAW;!}?!n#opMrz9E(mTSHf-TNIxpqqot|p;Yft{s
zMF;Jq+FQ>66W$2;ltUbqq~-5r!TG#HeY0J4C%a11-#)Ce`gPp{wQa4xj<6N!j!Zyr
zbi89^K`T7l*K{NxboaNjqBk%r6*S~5l7^t12rOlqXX`RM$1UGiPJ~v<uaBr)I6Cd}
zG9;HWWw(k7>KHcTw`Ol9r2t=3L=CXi40T|B{|%dcF5gcqRxJ4St?1Fq79vp~21bQr
zwR)(@=i(BUy%|H};9+^`5kGyWklE|^2!*&8JQsd6Zgc20Z$DkK7+}HG%#?rh49~th
zds1i(HRd(T=gJUg(no#wGOT*sdw2$gcL<)|GVo1%f8^hUs(?ce1sYsoM=*0|cQFOs
zjX2paRDZQL<ASI=rCnEK6z~1;EQjLI3${En@H}Opi4CM(Ahsouwh)()N!#aRB^FFF
z-(GsJ(Yh*ke09W_sf8vYq)B50uzyYEUJ!jHVUP|E#G}-wZ=nGfTmBBrnlV2MfJ3ny
z;GdoX9RS=HaOgeS=l_T%2x~|CmqNGzWNC#nfhX*4ut(qkHoJl16&Z?z3a8}*1}4)E
z7cc%eeIxJP5!*@hLuzp>P%T{U)!oFGr~73p6LRv4RZtbJS<D;C8inET&+UN-rUNO7
zQkn^tLQ;Lzky(1T|JBq*yGsLaJ<aXn1BN6p3}^t!cEGdqIF+|UR>w*^id*`o?Xc+w
z?NZlq-4pFZwMI3E5Da{>>a{@l{S5x-=>ewzgCX)2Sav<NNV6H9*l!i$MjH-sK9Z7I
zmiQR9qME^<<pcvEw>HPYO({3FQYF{PVte$(3@Pofo2v#y%aV_!AMqX~l@}-Z0}u8A
zdMz`d0l~DU#)D0U(S;mpgZy336tCAq{8vuuF_3Z7@tSQ-cb19IaGTzf6P351Z7zan
znj(S&ot#?JN*jMVVQ|x^V%w4CfB^ShW#{h%A5lZ3VVL%GrRG7weypP#@7N>I`oja>
zWQGw)c6QH!fbH@TnpjsY4<a@Q<*i>P2<&Z$HI4i^Td~$(Sd=+5?@&k05a;*{Hi2Ye
z@2^+Tz!hNb{{nswqNU&1(>xpaLTu|{C#wOk9$zB$G;cFFkRv$I@HLQu0zmpx|9lcm
zb`gjs-2^tsMiQI6nz80t?FrtN2({*N$^@WGqz$avNuO+a&%Ev-XLgc+jnL%wet|K_
zD{g?n;Rm@BR*zEGK?5fsV6oDrOXP)^320zgX#7!L{M+86lWr2jc2%Z?y2d1cu0RRc
zHtB9NFPaEVR{$W~oRnjGuXNIv3DvelZsMo=K}Y*YPpjMnC;Lgu!=RU`g6_7cMX##j
z!1D9qLYW}m5jG11!hZ9spyhcVtej$C2%XTKFQK8!f_WKG1H@h5{zq31<89$=A~)gJ
zdkG@tNNc*7wd=&6wlqy%8eu&&aJBz8%)Ah?c?4A1faooU6j<sdfl{sz+#uA12o3^I
z5R`Ufk=LlpYgFzYSSVOJ*XM%d=xsN0Iu6XCXk0JPHyBT&CTAd*6JCxyUR#C7^Zo+2
z(H`9Go=L)T1Gv7Eq<~QW?6Uiniti^akF4}*55e=BP+<G&Md2&$dR5oE`9HrnuL93Y
zPf`vtqLf`q660yerflfF?k2_PMljIrq=uvowj#VUEV%~^>B)>o9a3|tQhfC_7+7L<
zIfR<?(|PbP-`nZ(haEN^Vo0XNA8%hMd)c&4nEI!dS?w2y3rRksej+AC&ksAH00KIl
z(>H{^2j*SssBOH9v>bJsRO0(_X%U5d^T*{5s6uyN;#jcV309al&h{6p;*S<aO3o^8
zGj!_+KtCjW#7E3khH<U1nwCWg-=Dtc%~rSQ%dCpS_+9*B)#r#o`%%o0#P{-pZ%ist
zi@nUo^Zs$w@=niHd@l}I5lj#KB)ujkDNAKGlf*;q=@T$*AGssP@0Cki02h5yMgrNe
z4yQ-Bd&DN!NAmH+2`f0qPIPIK#-Yd7m7jEaR#W;hs(B?S@nYj0(hP&4uBi-6jxH9t
z-Aq5%{$?mnLDA4%>ixJ0t^c*15IO<LbZFbB!q3)Ge-prt(>a>*uv)k^a;<+>7bpGa
zN{s^C7`Bv8lh^&E1G|bRiYHI*m`PD;+rI+#d<UcR-*u~^!CxL+w!v@_oKZ`mBiEw>
zCOj_plhOZ9I@eDq`E@S`6cdIVvU#|t#V2nNO5Q^Te|A<Ko!oBov+Avr_gYxR@W_v&
z$=g&Eyxcm(mzVhoID?k>HfD2!$fXVIosv&X-0gpa^}76-Nxok^dwyW{b5WLuZJzG`
z!QPvPL;e5z!$f3H_I1i$lI&$0DkMoI5i#|VY%xi8W<)4!D3zs2))JGfA(K5xc4KYK
zsI0?`WxSc?bHDnY?^(Xz^F8Nxu5+L3KG%I;_aCltS!Q`Juh;YWd~8g8h89&XAsJAt
zBd!cpxPPo^sC<~^^}s{*8o$o+9XM#S!*PGxWzU15p*pkxMjUPJFh~#%2rs<_{TKd0
zOAdN@5Ivp^ufla8;&9a_%|#?TPM-;t#}v_w#TAWG!`vY3U|I^H#<LAQzItg;cU9Vl
z>>l!l+e<MKuEQRXF~#`kYeTdH+t${s?T^vsy?9`%S?|8Dhw}!~LY3{2(5`4BO*!=E
zoN_e2@GtIU$WGXreGQJz2X5MeY_J;qC%=t`7ufy15D%-({5!GUUo4}5#jLbZHyW&J
zvU~@gqpN;uuvXd{H?VVP;?hLdY&kpD(l2h==ufuQO2pt+ho4xJ1qIhJ;cV%Vv2NXP
zZ!dpFuLv7A$pWntUi>6rrcctY<a>cdVDa~`J7xUZ@8SE?+|)RBVh015Pd)2OGM<46
ztAz{pIP(Qc#rsb-2@qrG*kO39)i+{N2~Jkm{BlyT>nGOo<u2TQ_)HxPysfxA3gOEn
zEp53BYk*`1oxO=elu^o9ydS7mcF;Zg8N1?zA!f10F4%kHH+SNmsqZamR^ymQLJD>}
zxCiJJ7qGV2BHoB?Bfzp`{Wn-qG&~16G08e7(}~{P`IGG%&KEtCi(d0F75S4b5Lh)h
zGcGdAaX*Mno?yMhGL~p{ES@f~lYqGREjZ2LrBI1Fb&kb%ka9K@P0~-G?)xV0wOQZX
z8~oYSX+qiIBI*Nn%x8O+wg#!09!K?lR~aCtpsO_$=uu7e+fTo2e}<i}W~U+<!XdNp
zrv`~A>FF;D39yN*#y&!NVu9M$r{1Wa^lOn$b&w9-1Zr&G*GG%Us>&GctxDTSuoPy!
z!tsEvRVdL8bQQ@I<O6?Y-w5@HpEjSz$O2NFe#rJjdev`!=Ig>ihRh7#@h4>LE9vsV
zgY@LWuxl&<u<)-znlr90vn8Emp;OUOWfw+j*mcLHrVNmeSmtFrHWY09+ejJWN260i
z3VX+#THl?NTz={Jql6)hoq2#a4#P&C8rFD_VEypgP_g2HLcgB<3@&xru4vN>U;%jX
zdjU{@((^g1u9}s|-_OO*ym**TfHE)D4c+azqKRCQxpA4wUsd-KJD*1TPB8{#=I4>V
zC2c*E;~!T{PM!G*a-1lFID<Hb_bZ>;wXloM(;X_SW|E#8y=>t~h8CjhEC*n1w6#kr
zDl%o!DKJV0&{xhs3ylTlMj&x!lPMgk+3t)^HdZZMa`pfmdWqVF!}gZumpOysS4T$&
z0D*P8?Qd>*uyx`KSdwM`+ywv_P@0vdpJwi+2dK5RkS84uCD;rYd%V?4%`a^qIw>2l
z%5L5=6}Z;VJr}~*w?#uk1(dScA5XulQ7)I|duva;HyQciVs8Y<6tBg0mcwS&)O%T?
ztp4xp(2#Pw+g`li=9Thc<AGCccdu<6W+&A)D+9SsNvoW^hJ8mvM{|mY?u$!zt{!t;
zO(shj=LWwNdWiN-gOue{sIhL1RI9d0I$FT@sb4ec%MW3%xex4@9@*(_Pg4;r9y+y*
zXgpMdzptt1XDFLBeD{Z@klfmB^ciM85S{o(szDu8DPv(`l)no(XeNVrsjXJ~_13%X
zrM@#;-ubB{{nSfW6#aSYIbp5Q@TVG@PNNLI@~Bs>zm}{<G$y@6yRG-8j%*NG;m|=`
zCn0XIv(A|wT03xbcXM^`e!lrV3OG;HAyf~HuUY2`6VqWTT-kWlJ(A}K9DnE9FPt_U
z35a#y9#T|48Nl{N@<nJ3`W2sEss3=!T{4*cic`jExp|IN%ou?-1FCzEmpx+d(KkZh
zbg(xK7GKXeCVwL`H{ZkvKR5z{v)kNEaS(`Y4wYwp3sX#V84&n-CNxXCd=1kj#uRH)
z&EvmwcPbyJ1-vr;n?kq$+5da}W^puJJ}&@WdbPvT!$02I^LflGwMFZfe0Lq^@MKjm
zn7TvpY>P-rQ6x?mx@nJ)ju~Va?C{7P3%ZpFIkq77!Dotba!p~Sx-?UDOlxg#?u7R#
z&K-|?y6p6M{B3nV9ZQ-1cfDsMkA6Nb#+bN@8YM{jukIcu><vn&w@*Ki9>^%GrXU%j
zMzi?X%{`#gx&J5I#bL6$@wl6qRvn%X7?#AVVSD<bL1EzH>T*887AkmYQPla<-PYzz
zkD6-AAeTS)Anvp=2Z#s4h#DJlH4^>Y@Xmw9M1Mae!N$rjxpx(H*#~~R`TRS-M>iL0
z8!Zb$(@o}ABWC4L6P-~x#CP9K2LhH8!NrRQNfUgwuX_Vb>Y|Mm{j>a@*nD()_uQ(~
z;TGs{uash2%H3JK1=OC4nRlfJo%rW&aio<fG}wW8C}4DJg`zztHL9n5k%(IVIY3gK
zl2%`Dn=xRBi?qn=()BS^?z2eV#f)6N5nS($7t3GAVNar{M1;84_hbF!nmmq+btBb{
zPo@P%$`Y*PJEQgH{}LCK_$M0@Gg$^I)%T|5FDK?rWe0=mH+L&fUI=4%<C(oK{{+w*
znI9PrbeGVwD}$S>0^ZVx#gA^kR>3G0J)^b=+Lg0KwfseC3}}e|BadQpuxiitB|^-f
zY#a=a*(m*Q0w?Gvhki80kN!xPO|B<Cn1h{R+Xz=*b0qdL#7~l7DaL3bR;X{rPM}Tj
zb?M}|_2D~H{uUV)C)zZIIkK5Pf7R*(1zRGcm05_JDf?~+sZtYMDJaYgPKYqRuy!kz
zw0MQlC8en9%RbT6#e(@!VXTK>5rwak!QLog2CKa5I9tLnog~4TU?1g26DHGGN6Nuo
zPjNxYg?(~km{Op$QJ#{%#bI_jhXu+7Ko8DK$5oQ(o-LMY?~U7Tw&#X){=$~C*jWS|
z58O-r!L7R~zH~*WoCh1dm)_ccj)-H+?dKlU{YTc$j;d63Qb`<YFXSH;($O$d@v!97
z+sMh+yH@vm`s`#fy>YzFs@{Mz!&d}VnXjMAd~Gr3QvHQP%!DGq7Bu{ip7Q_y+5c_4
z`X6J9%>cEFsfy&I;<_|B81O2}x!54=RVN=u?Wr!GxMZ&ZHBQ$4e(p~QicR0PA(@*A
zHtT#16a-@D86sOWbEt|^M@t^qB#4B<&kt6Y+RuG>p&%OBh9e>>AOBu<*w+Ao>Te~&
z>F{?*#tT?$8ePB&G`}^dUUbVTQ@PLtQps`&2?OX%XLB4!D^zMi%Qf(U^;gn!IOYXV
zz9?{QS0Xy?*jPgZL1XiZP{oVy1~!-ustpZU`}QgEPW=`LBE7&;v4)&zW7*b`44C&y
zK=bRnJqCq<0ny1eI872zHU@eS1ASyE;`GT5!mg%(0`JS@kk<(BKyxIg`{5NUHeSYQ
zirumVwZ|5^_tI@Dl>1WZguq3f8;?HSE`47nzH<$LD~|t)O-2@gy+Vumd=A>CobWq&
zs`g5x?MqnA$xgS^m$X8kPUSj|y20uqTm?I&E)_oHEjzeuR>lm~WoL9T4ngTy1`m{1
z9GYY8eYhq{WmZV)OXc|hCC?#61zZUHX4>Qs5l0q#`HAXtdtQFO1I&96S;RYlIp4)=
z8MV}oyM=|6i<78NFEmme+q(T2$4OTnW<(q}b#aMbbb7?YAI4Z(ckF__+v*ZZ2v2@=
zHB`c>A@1DnpZEDGbpvs1=o59On*DFdFGzkjTaP)|NccYP<VMwO7lBVC4K9pCFN5#G
zI0#9>UvjlvTtB)PUfz7^#LMJY`?!s-Wiy}}h>u#<ZlnttA0Q>Is>jtiDfxY%VZ9+a
zQ0O;;+Q#vcJ$6AmRQ;?aSD0Rq*Dnu1jK6&cRn0yN7IDNU0XteEH1HDg8SV5<QJXrP
zo!82LSbQestKH8Iw51ztz^}<5-+RX)`*85+c`Ky!n;Aa`-eU~)Zk)&JxMM5)eb+tw
z3CglxP?(lb*qf{SMXok$Wu{`x-G=@GeDD696xRnVYW@$ZT;JK%$swR*`#?3Ky`f?s
zFbrm<PhJ|p%5%l6?0IO_u2#Z57K;c*w`jA(p@JE}gzZKW!9we;Gv81@;y<Tf$5+#;
zBTz!^V2g4`JO|6Hv)z6C&T6c&^6XMTj)uUC&L8ZV^f}j(b$g)9|H80?eDfJsDW<Q9
z2w~^)@6XX9mS5VGC1W_!0&;;2Z7=ga+DDjvB+?_v>7!$2+fIf?adA2(r-$>2;ylnt
zM|?uaGPtQa=CMwc)kNJTI?rRn54n~BE9J3>R9v+uP>hohZE}}a)>|%Tg!Z*9|CsFZ
z;+uX0!u+D$OFl-90%8%aqf92>>-!!mzsQDg)bkbwu+NB??ITxG>SpAV0!^_b8JB5?
z3n{hFZ?bM=zJuM>j^UpD-Dih%_6ehgU?MeGIiK5SYnMATNVncke3E>e7IIE!2mgj4
z1}+?Aq!-jtW|kcv7esW`30>FJEA(1>zMZ&B@SLIJ1g;vt%=?{<nYn@8#n9|uF+N!i
z3WsX9>9;N=xsTa0Q)Bm^=l=HZF<@{r$GEY25F#l4$WYv9({-W?S;RKeCTsVLUkY!o
zKKax~M3jlL|7JGt23>1rK|Mei%)x?H6O{li$UndgS?9N}W7@V!C>|IW(QKjD)Ii$(
z92W}qK5{ckqdX=lDKzOwh`bJjAHq;<5HTz*NP|+>E*+9w?UJ%@jkI!HF`meqbzVjl
zc)`(vt3cFf#Oe-Fl@;T={S+DitBgwre3PlVj$J6blu|t8KWqNsyM%*$;cIM*T_WZk
zx=X;5r1}84+{E$bs&ilPnuWrN+h6R!by|t8#4&QxKqax%i++Ul!K#B{ndWQI9U|lI
zM*T6eTqEP+VEyvZ^~W=}Ey9b~tOAf?Kbs!eN%!z4It;|0<QSF)uK#{%_Z?wu0QySX
zw{Y;W^$~Q9UykhVC&`JRQhfD>(E*NNf8@ccMoM2?WT5}`7}j^k+J@=2Z}-*ChLfV<
zdd*sqUZc$hZcy-?^UbXFGehn}4(oYe-kw%u?O>;|W12ISkRs4DyMwud`$)wry6cbV
zYai}h3Ea3W27RClMoMSSBFS<@TWK3)wYUD3dSEFv?;S@sgNhi?%X=HFf7|7heh2Si
zFB4IIp`T_V|6G{<1G`Bg0Xh`Y0+(X=Kmo)?A+GtShSSqEU(*g>{_Iq06u?MfLpieA
zG5aFnl&7>ul(dHoy9j39tFfuo`uZ`%A2aLGI*3X*WgR<?hp(BWF?NC3>qG2kkg1FB
z!tAIv)G;AlbX)Upx%z;~2<~^*0NxG?!!IuZG#ShmKOMwD+3k2VFq>hoZoE?T9{+Ph
z*+ThxD$m{z8=U1R`@i|;iUP1RQqICtvuV@c<z0<J7%cI#sK@M+sBz%xm*-1+K-s30
z+`mi5@<9HUa9{1S(h<4xTt7QS*V9)KTC8t4j!nV@2dOxTXuD5bg%q^Daz;c}IMqgr
z{HY_VoRY@c(>1GM@04oy!1TcT!D<wbtIEm4bHjVTuz@Mp_&UvnBC(9muSvUqx0KlM
z{KV9vwh&>;o!$i!Qky9fv2Rf@$lx?(<@w(Du5)!xist?9AgqVc2`&hHC|6L<r)Krq
zx(p{+UtPl!M|EL#{TsACy`9RdcAa+j!@0rs(SVI*Sl_BM^kak`m(r^Lc{L&yK*2Cj
z@ayBeH6&=aq2(=<f~jRv#iW4#FA8(g<Gbu}jtp(K*->2ZhMkO;ywZ^0rhtLd`^M^e
zbEP|0hL66Wr5Puf`GA^^6KfV(#Hz;KU0?*g`aJj9PhUs*3)WpBnhk!2`VmgF>%#7u
zib@@7d!at);cxup9v+xP&BHq};>HIk>lT>HRBlkA(j_h3te#slMDO?lDhsbFJzSN1
zu(V|c$N4a`rB(##d+*W2{n-6S<$|v<hZfwtX9!~@v^Z$I%(Cu<?5Ll@qgnFKSSV3y
zsSo}IcN%`94^qa)qWHEA=TeO^5}`Rr;nN>#YLTl}pOiL1ifrO{u}70o$?3qgXR8XT
zChtgKAb4MdSw{4{|BKsb<FgF>k{A>V0AHS2>Qh@Nk96Q=sKJHdyHycNw7SbwxuPpQ
zY<o6(QLU`5UQ#`ZEfjH%TAUtGu+-q%FlH?<C>A38CK}peJ!U71eP!5C{IKmXwx#d#
zjrUa3&R1pac9i8-JAwH`YVZL=l;`-nOm|ucG>!I<dA5n+_W@#=gz$Yj+tO8Um!xsz
zgVb-|y?f*k?REVwo~v@bwMWrscECXb7j1D*3r_wpuxWnwN7~iueOFHiee;}}#K_<O
zqZFK7B^Ed40Y=_~>abaYc?E@Y`Hj)*RJq_kyf|nEU@wd&vm_Z>6nwk}@w8>mb^d*~
zVBr>KQq6_2t|=g_<^{+&vT+BQg_;kV6r=X_(3U*urmTm@?0?qHjA7?gNd|3X^bD?D
z<78XIS0T@`1p|u^u|tvtx0jX!vRN;Df7wm)(ZnbrWQQW=`ygp}Qu*v8vdpX?@ZQ$N
zvzv(4Xk!<e6u5|=KrSSSxVsqcNOK&klj@gi!-i|P)q%1)an#g4S<udhUQ6hhkehDk
z6Raxf5#dj8d91jNpDF3ULMmIGt8s?+UvtM-7~A>E1cc=moN0`*;}H9G6hwUV`1clf
z5aA-09!vi3J{!1CM8iFBx8RVPW(?dDNG5!ogCD+eUV3pnLpy8qsN90EG4jqa#W88d
zu|z^ULNXHlvqg5N>p@Lp(;@z<ZpAv_<(&2s>-XE9l5-1MqH11^>1Ifj&TE~q+CkNc
zwDG209Mjt=po@cJZxpFR*ZdM`@p?*VFZ9ieCi8r+B$=JwBf$}SulrZD37TdN!Tji^
z3|`7yoWgzc;CJOCYcn^lo__h6XEin*0c<nfS-Y-=X0|#@w5J?uSy|Psq8a^q7?wK8
zKiOBx@4Hb^>N_EbAw|irpBeE}Y#zcYYS|%=e)(Kds>=?s)45oqcThcWv4X~7;Sx$k
z$pbyf%vF_8;EH}+Rg7LlSq1qkEcH;7oNt?{>d+=1PxI3DD}9E;JKr^ccD(@S<s_6~
zHC!WrCI&6~Ot|Zuov`ihXw3+$JzBsin41kP(rNb498qOm6^bCGz<ZAG<8wT3(gIYM
zLQ90b3>{+Lk+}=Qo71L)gVTK7ic_j=5gcIrLIT@2oz_GFc(>H7pUeqLWN#R8Rp=p{
zbnU~XhV<4sUbABt>#-Pska^-#nzic@>MbwSMwF3{Q6$U08esf=9NnVW6fm3OJahp{
zGGHj|?0U2|aRfESq>&Zbf{pvR!MWvhhcruF0eFlN*3|g2AZ&KnPsyWmp>5X~ZNhg`
zo#e>K{F6->jlWOWVJ~JFK!N#@Y83w*Vw#{+*YzT&Q<YY-?oyvR?OyBoc3sx-LGh20
zH6HZ@=;c;+CXGCM{hQok@Sli#vF#>-G(Ko9V77-o#$bPAQniRY^m+d=HIHwrUvPI&
zJo9xg<RBZOt<L$la;)Di#YC^%Yh~=8fzwVBsDR2S0=>9oV<A$Guebac-zzU0a)q8K
zeQ#tposp*WK`L;vRzBEvq%{kuhhO!bqRI-7+eJ>5f3UoH(QJ<Mkvc7Nwxk6i6p4nA
zZ$gdQ-|^f+iX2pv@=|6ar$@CM$AVfS4F~Y%Bs5PAD6j4#t?Q}Z(iAZYu7`A#7W0x5
z*;)V%ZN)NkRzT?LyBJ#A<*J6-JkIoP*+p}EXjC(mr4_0}EKUHWCS2HQVf1LqSMpJJ
zNUd9sLHEY-p-do^nei*3AsyjY%ex-Et7od$QsB!*^b?W(qlyyE@*~wy;r=L`tB|bm
z=5b*A1HeNGdJvp!j&*@tTCI)c`G-_)ro@L3tYi2d9Uo8LTNsQ@kK~4rM|!_HbB=Ii
zP2Y6vI-&;l9+-yETM(SLTpX!?bdU(*#)sbw9q}Vyt>8(Dw!3&7`wjg7<YfZ7lOa5<
zre4=>_hNW_p4gCH`IMGiMd6QYX=m7&lMq@pZw4;VjStSbMaZ>5>;p!MzpcSH?CAVL
zO!1LT*m7MMDG770St6m+@a&F$qsOgUdzY*Ej_p3ePHDH*+i)|b2yTW86*Gg%EP&uv
z^`kYCHCj3?{&CM#T6KUk?5{e;K0NE((g{%2s;4o5dzUY@kHtf!?`Qnx9NBn-mhB-!
zuGHJ`lzyh-w4E@oRPBr>1v~bIP5l=VQtJYUa}t`2S=Za>4b>#K%-AAHRzR+3%MGT^
zK8EF;eTTk_P4&hEg8U%~R5u)$a&iJ38&st8D!4*l1Vl?>wlK;ljWmB|0!uH{YBnKu
zGsW>!Lta*%cc}7<umcA!+_r+}p%z?TyMzu?PDG;*v1R-=W2$TdvnPdMfKs}FdDMJu
zd3F!g;ceY_$q_Q~;<~M;u2;e9Kkl=G&$vc@i~j7q#cIT{?)&a|LFz!Hm(~K-65ij9
zyA2|CqJ#mSVzxayW$Gb?emThdS<HUpn67N*GUQk^HIETs=v<EwF%R)8F&C|T*K*nE
z(|#5RbbCY5oNNRX0x%Ja0j)v5Z3H=#K#rPWx^&^HFsXOtc)!^&FVoLu$wY+4Um3%)
z0cbLcwF`jb-V~6f08-n0@EhZZ7#J7mVBsk{5+JD3Cxpc@qH&8Hq#pJw>-0I{%ND>I
z9TzdrhV28EC5O?BM}9Ia900#=VEcKY9v1ICN=OGm3MIG@j)>jxP4QvL6sHsGPJHF=
zVR1#njo>qn0jV^|(32Se-`R`(&dZ1f3urK_8wRQ))_}P5Tw>&bQg7)?w}}=03Wq*S
z9RvLic7q)%0e8C_Gi%jUmT?LFAY1RAycGVMzr5bJo@QsuTX(q?%L9OWg5Bg}RFA1b
zV?V=!zYeTLDvr6G1iiP<_<qgimdcrJVkZd?U_u*0iSi=D#s(xmeK|*~=Cq#d??InU
z6BdQ^fFJA`bc;fyD^cUj>Sf6DcJqP$@uVY|gK*yyVXdWCg@FoaarGWc7g8jdrg4>k
z4td|g)@bJbAk*-IYst((^S8((ST&}tFIFDbFw$gAa#2%^<9l%Zv9;ohFT3V4d?~zi
znm;r!LvO8<=cXRXG4_<A8&#M-E8F2-P6rbzg8}8b=tzq@M_|7`WRKIp=t@M3Y$ijf
z?IZg}(XT|FbG_zbUE(Dj4&2Q>6jyE}=Vpw}WLTo5=|(IbZn_3x?5Zo{D!$yd>EeOf
zL2-+=xhLt{%jTj0AfcSLuAcV6BcR40-w<D^EA9h_$KH4MlOI3Qf*coBl3tBLf!5VS
zx;;+IZ#CSVuGPieE{(p%X8M>t7zb*onG(W#CcXucJOp8E>r`96E4Ul0=qZ`+{}klD
z=#-(pb+Mk1gy1pOp;omRd>5y><%ky)yB?HGzcXMeh_@e$2{ePHgHp^EJGp)i5*A{p
z6J(pxA?KBu9>3R~3g0X{Lj<kuqk)&x5-Wi!(FSU1cid17*>rK#VoJ&7@xlrZ2?xp`
zz?cAvzjkzd9lqh|y?mtPJl9EDgztH`O-Fe9UwZ#>)nz&rbK*mUIK}hf+v3N!yik1#
zS@OU-dK@=Fr-ApNy1Y$PUotT|pXNpl8rBpmdXoHd`>dsLr~k8y=AFKj*bbE9aFYi~
z!^&->$T!96VqS&kv)J`U*-V7}ZyQ<=<mw7_SGdSUoNl1RKWvg(*}UQP;~=4jQx;Pm
zp-KJ#HdctL|C}Vm;{^yLa$#@g!-LcLB@7-}WW6BTaHUf_Nm|)#0LD}BWTX>d=&_qH
zrl{}6A3NVP&U_&a1`#-^)fYA2yxX7l8Q+eeoI%8x>Oe3`SEnI5Mb{=zaBbZ_)l1>S
ziK2c>b44ol1E`Kyna%K<Q737^5V5RFE<LGDy1F-{s7IQSB&@o}&bE!~<N9B4<=uiq
zEWU+G#&PJLQbAja8mn)(<;(buye~c7`!gqt)i-r!`#!<;um;??0oNhv?#73E??<=x
zqiCaNuboc7FCiYGVc<4~i<-*#NEo>7O<;*=h*s)TCwAel34VR)4nJuItNJbjJtx{p
zc)ueo$nPIo^}Ff%dExfUM-el*X`S`RR5&0kV4_WBavZ3t4>TgD?@x}qd{)XLcAb(5
zjES;Eao2s)O#Uv`upcsNpSU$cR$N{?u+ib?1+T<*A*82U6+k6*T=99dfunxk&x(jF
zUG{Xk9jiYa5J{+k_Ei5u{tI7QWsG#cjlDuVF_SN3$Jis@ASl-8@!~lmqID7{yENhJ
z<jYdnzP>LlXBn-2fY#k7){gp2pje+a7O7ZxaTqFz5G;xE>M@qqcuhbyzIo2Wwql>k
zErlJNzitznp_cBR!ZqR`$rjFTCi#C*>e{*;sm=l-yq)suJvXvlyoE)BlwN<rwxC5&
zEoj+Q<ES#Xwi$s7<f7r9Pn7oNifU)btO11W@720BEs?wXkxiNu&z6aEMI6omC#gJX
zeB|hb=pWLn{}c>|?XZK?B|xfY8#3`EN&MlpZX#i?>{PXH-qXnH<o4goBx(eIoXSn4
z-J{yH8m7fJ*_NDq<Fz%g9q95^d(U#z15gte(ZO`3#jFB-)3<<VC-Q0MX$-Q2|Asxi
zN5$A5wk&9UUVYQ04OeIN2?(aWGY|_blgf<mdef@UQj?q8#g-6kj7V^g7(i)uBmSGP
zc<0`atS4VF*N1ZpdstumE(*;e4D7%i@&j7;4b^zsvX%Na;8$mP)jHUeqd`~;@cQ8l
zHE7+EMz;2hF&s$hPrS0$dpu$WJ~|zh!SOQuMmN>TffwtCH@P!>bGoHHv(M%&pR?wq
z<LA+zG$BQx?>8@VUWDxnI<4y}67OrHs&TiA{#NQX_pBRu-QCA1$b@E@$Qy%^%ETx`
z4Na@TTaoqUL|J~a66Q`AFvVW5`&(Ma>&<OY1#>zK8DsW9-{}6v3&J`Z?ug=HcE%*Y
zJlAgjHh{<CSwVEL6LNyNZ{`IGhK`{h0*M=BW%^!*H#MXaE9SFfcdy21+Vs~#VZ0Yn
z`6}!T(~+4F$z2g?z|4z0{t`-Tbv1CPyJ~~fCVOl<))CF*=@(p`JEDB`sz1w1r56O5
z@TSMW4`l&*VhGM37hA;;Vip_QmOMBmF7Z^ObYY&0?cNj_mXFniDnT$fT$XW5FYgbl
zhURp|x_IU8PYw=jLXo^ou!6?CMs>^YhloXxZ~ON>14X08)}9Az8WP2Na{LOD-I)al
zIaUvvVh2kMjJ!*!=}TfLM-|MAO9XtMvI`R4Lo(p*VFZ}LK=b?#+wu&x2gm}%2cDQU
zG~M?J;Y@pG{H5X&#%G7Cpsp@;bPq93{8+*H2Z={T*`s$`n~CZ9X%n0WS2pOM8Iv}p
zZaf9>_T)X+w?z<r2IP-)1K@9j?IioOx}21cEj_mte&aK)aV_uRrM@?btcU77z~*lJ
zx3Z%e6QfL7O*(w}y!L3_8(ecuAj1pfwXK7Fum_xE(Lt7&S1>`KS)o_SS^Knv?MRY{
z+$(gSzMGv=W;&v5N2ndi2Vv|Iw`}km^hWPXhK5&5_HVTu7Xt%R0x)Wzx)8JS0|SVl
zC_K9nnN=sw_HmS8b(I}c6v;sCVt7z`2kfAe4BL`#s1vVvM0fs(hKDtqpztt3xM!b0
zZ|Tww^KD(cw%=WPm~r<J*!z?ZU9+ILs&?$MMbc8C$~RD>!<14}re96=jkj9KyB%QB
z4BukP>aq(J--+*6LJKvW{P}f^?mAud;HgQEG*g(F26o}hQM5o1*#LAVBFO6_n@Rd2
zE(_uWllP|Pj(?abKjlb$%>EtSatM7h2NR3tzDd2@SW=Vl(D?Mp>O1VfYvWm3?MY9u
z1t7!HyoaRr(ERWkZ^;5HH3elphLOD5eQGkdO(?ce2D0A#J`eJREyXTfW1l~GntzX?
zLNxq2HU=jagdGPO;?7J`&YlS)@{t*hjz)e75QmU9_Gk~5kQNQA0gU*x36ZwotsQxs
zFIWDLvgrl9=Fe{A0gi*~?!iJukwtU>lC&Ip36+yfPIe*$R{b_o{#e?2D(%|UeQTLp
zc@Jki{2R_N(gH;I22=btziLMsQ#m^Lp3q&3s;cc$7ux`9^<-6OnMI6_QufDWFRl>1
zFYEvZGGOth(2<~%Lks$wii1g%t>=ZNddp#?#zP(-#cRQyFfIS^aL2hm#z9z$^Q;Dj
zq0p?>q!sx}gg>QjQjClM&A`l4jAZj6AN@h!Y!}GrtS47kkcZc=@fT+*z+nGo^l?yr
zsDO8$XG7{D8xVV*?br4;6n#ZMpNZUMEXHB_K>_!Y^MdrD)WoCepW!zV<FT#i_pA5Z
zK_4GK>pd6Pyn5leSvHZP?IKsd;MVppAkJ?f=YQ)<s#$?5o<}qh+;SSJD0n|~i{xEW
z|Co5etXmHXq6eW{%)zNR;k)%+iZVUiV0F*WL#y!iamU6(0Uw`TBOhB>MVvslAmBdU
zioWpsD0!U6Wti8f+w_-5VY|3CIrpB%etGjRxsmGAIFOCD)ro#jNFp^4`2T3^FZOUD
zD|o6Lv^OSy3$BYyt;h1b0_sW3J%V)8aq2qp`*YH&fa_seNO}y_bM79!AoV5F7l&;{
zLMobMb9f6ZMZZ~kcvVSWjH;C6r=(1kb>ug@@_@#c)`<&xm9CPaloDGI{iUZrw!fd_
zFFcFCiVD?#VJBm0?Sp`X2TQp5OBg_UNkMGxLw7;<$DgwNft$NReLlt4sft3&40*jM
zQeKnt*6;8T{M31EYwVm6tr=P^W%VOun@*C{KNRYF8@XI*!5x%69UN0oJM$6*taLC=
zKxwnEG<ide6CQhHlN@;d*^@X9_*c8>6mIinbuwbSi^U^C0?g@bzh)3Xb~e<;?%CmY
z<DD_yMfT(ImVKQcOISR%bSdu!M0`F}L+aVHvI$^nPM4|u*Eu7FwSNR1SHRh-agcyK
zkVlas*x8{tTn&N(K5z&a&>mODMBNq$ls5$nz9B;uf7oA<(y@|HY|MfIlVG<W2e$a9
z@PxZyMnj-qM1kF07=Ba5?*wbi*HqGTQoQV@0qRJB5nLKEwnYQ;-Q^D?n^48PP4u3t
zr^~_$r{G^Cj!jD)AG2~ZB;nA;HpD#5PHDOCY2{9Br3+8J+_NW$9Ez_c_a`#An9Vp3
zm$dPxvtm=?`eeX$e(|<*?p&Uua52gHKDu7Z4E7Nw2ps$gqD>NoP=Q&4%SZK#_dMgx
z@!XYiDR-Z2gTkLPj9IA1CnV;B;N48>-LV+9Ts@~#sUiFF&9?RY9wjh?zdFycX3yfg
z@+X_3L@az7zQcFflOgQMI-{BdUm6COX3g(55@;gRlUgt<80PLW;0Ig<04<RY%;qAR
z=FG-Ky=tl=m5|LTjWdNYno^;yZC3S)>YQ<TeKRldPZog06gIxIL(PBMB~E)F$T`?a
zx5M$7!R!aFVxxhAz>b8<YDkL<S~BtPZLq$zy|jGW_YL5=9OU8EKDV1qL+lsOeR$*I
zaNH~6LXx5Qv;A||>#cOEVBO|eWdP;HWTW~hi!e`wbjxAHI;FCiz~qWHef_E_kVqik
zqePI32Av_#uJD@$!ClDo59w74>+Ei=sqfW1u6+p?J1)w$_&bsEka}6^EuwdY@@a(&
zSz(fg(%SL&L+5B@NQOe&aglTB+PQ=1dNew~9n2XK*9D|R+ZK_Pg<w%T3-@H6K_5ZN
zx;XLASQ{hjmRj*RNP>ipGga}9Jx^(F^en$pXRU#O=l_Z}O#J;?>WR=jYC#K}cSU1A
z)$_sf#&Q4QCce9;wbhO(U6ZbyfhWoDg)qnfcwEhrt&a?P6MO5N$P&oeDu30)O+u4L
zfQ#6X5Bf<Jp@Y3ap)R*-AFTwZEt+*Lxe!3}>x;RMmT20b)TL~YEh9pctIV^kQp~!Q
zj!FT{XG{%yAcs~D!I8Hq*0bb;DwboPzm5wmZ21oBy83jrhuhYc6uVkQT)*?;hJmnb
z8MgE#-(RRVz)O-5^fJ<T0F1*_ULy;=-;i_yZr^T9GvAS(jWnX$MINT!tV9-h%j*pV
z30TgM)I+Z%!GWjJCa+rIO<o>Vj^X}Y18%ldby@{vONkGT?Ra4@bhiw&w}-DhpQ2wk
z2g55CZvbWYYGmfD+C}*GvU?P6M}RHL#UiY(4K{I-Skh_H7u$p=S@KK0D1N9-F7%rE
zpc?ycW`|m7#gY7~xS|(MGY`Ww6?bz`^=3?UnawCpNPKOv>jD&8qF>-2{IT%B_`QwK
zR{_yrLoU>SO!)1=O?6>?3Phjjq1XWjv>B^Z<jlJ&>d$0}_1ZD=ir8@F2DkqqK%k01
zmZAVECc2_=GnK@ndoZzEhU?-{Lt!&m44NJOqYfAZwbdeaOw|8LqM^zF-3w?5#)3g%
zWmWWBr14iwlbx~3A`sT@e%Atw`Z*BX?z8z9|9j$fx(mZ#gkcv%+OFog6(n*)W43x<
zJUfTK(z{LUVI~guW1-3YPd3*+*8Pzk4RCLRdp0wi3d-Reph;0?3xH96*ebFTUGXfw
zr(jing9ogBJ_#_6MJMWXq9+I2gqTu@*ShDqXRSJfAVtTtNX?R|s}Rt|Hy`qPRG)ZC
z{Oj1Vc2_oZvRz(=YS7vzYVH1d0uhRKe*Xy6r`iX(rR=-pNw`v<ZEhAXj4r!?u6ArQ
z>W#2VREJB=KQj3$5fL!gS(cST?TLI7uwW*y3f?7vOk<+0XFJ|QF4SEc-PrxY3x<yq
za-K-G+b(|E6(Z7$dg8c7AYz1lP`jzN@s%aE%BinQ&M7Bad}+|e{K@8oUEly7svo(#
z5Ha%N)Lh|bxN40`rCVW+CoWV$6=><dUSWlq(T163d=_C(+d-}!kXl^6OGPGcOiNqH
z7FIYhkG{k|@_|w8)1vhQIg>iiPgkHA0rjvq{oEsu9lx1^!xR(8*hO{t8Vn|id#j-%
zg~yyqu83l~{M88?t`X;`@)$AN(+Gjh!m<y__u6);?P7{LVnTap(NG9w%Zc@cs+y$B
zR6O5_>bw0_1_msGTI?o4y_1AG5E)6$tNmDh_4|aAPvFF)^Qt<3K;iu(BFDb5{la<T
zW?mrHJP880Wl78^ju-Bg;@0N|{hZCSI#=(~D3j;*%ju?F!{UuU_Au0GIy5O}QL_YO
zK)XWWN_z_pog3OTay3zWF?#dbK%KEHFB|=4VhaV4x&hQ=hr@C4o2T`n^iQ0)I+<~4
zW@c0DBgmBL3vDBDCm3^-)NDNn%CS`$3GWWr4=Zt|_(iaI&FQA|8P@w4W|hrZ+9x#{
zPFq^tdh$}JxD@+j0|Om=y8I_w>9H$KNThpa%8(@l%fNqN7=zgyI!XNIQmdQtJNyAy
z2T~QD&eqTE`#-{^{GZ}v{(ty+Y%}zp1UHmn3-}X*MA_ZsOh?Ci+Wh#>g}=9CXCCVK
z4?ejA#<n`cPOJR7H@tZu25sz+t;Wvw>64^r_1{<3j^7A%A?;W(#(s#s46182;6auy
z6rS$guN<7gy&o3NwxfcR1cB)U7?C>HVGCU5Tz`Z(BNObuWx0fLDB3Edd<ht?n2K=B
zhWc2%dDOm6XPJeRH{;$13bb-kw8azMB37otu%eI%1=#_w?&|1~<2#q?cVdk-_pP}p
zQ`KkgZfftMr5DFW4{OgXMTRAeVPAsp)04TJaA8ztIqO#2sx#5$wka7H6jZnnk!{mP
zM60%G?pDJyvvz|T7e{3`4_Mi|+Wx0xNijD+(?OCQFcZI#N$-<XUGH#u6)pXyvebe7
z(p|P|_N8Fs{=Io8E-uQ*T=Dl;cDfb^pvT-prBX_mGZ2==>p-^%1(1Y)zp)NUi0kTP
zhv?7)Rh!30lF!Zd3y(SedfL$ksr|{;nU-uQ@GZU6_xysmf%^L!gBM<Y|2WL(iZSfx
z1i`wvR9osKkfQthmAIs+S_c-5S&Qt-uuQ%^D;t>21Qw<-w<5V=py~3|IAYUi?~C;v
z`-r<N8=JKUzM7%FmPX@aKCaQ6uqV+lv?MD4z**Tm@z*nozbN(yR$r2TyaR2Pr5~_B
z3#PqGsWg{vj~%#l4HXlu`d1!3yCEe^%!RJq_jO-XhR{(oEb9Y<db#vp`Nz)>Ykg}}
zO1XhpdB<r#|8kGUvR^rR*rh{=8EA(4Qu5a}LG7=P<tlmJx-V{Iz3W!pnmB4Vm-z0D
z@QG4hZnNDQ|A2n}k0|MX`!n|49Vs92JM_?EuzNm`*&ZUs8{Hij^_510EC1rZxo{ny
z+SO#6D`oX}@1G}wvsos3ewW}eSzhPg6zYHa;$8Qq)oC>NI(e94CE;H$bxU{VUpP@Z
zFA7ffzx>fZ|M4EpQ@x(Y>JIoD*RFNX%g6#cq!y_UQnzYH#KA?I?1<&aHk$KS=W```
zFF5wN_?Uk0!6!2Pv7KPm+*-E90&rTsh-%cDeRl$rpm)<iI<d8}u(n?5X)tX1xMY_u
z`%W1!tJzHJLf&tdqkfE0680Ho1-cx*_f?}lri73-`jHElM;OP_rl2Qu4Qu5W_wGIo
zH?H8HaC`NZDs-_x(HmRghnuO%ezow^lj5wmax8R=lTim~`A#b<#CJ7Wq+gd}_B?Lv
zQQYmo@mgs^jIn_tU?kp}>Y#??)kg0=<eWHlQKP0h7K$mo41cCK!RiNvJ2DY1SM8D}
z)TSga(_g{WeSIiasY-LpwN5t8=NH@^GgbhGD%D#lhf92U-Zn`g4QmVDoGo&AEc*=q
z;igmzNOCO(BV>5OzEC*LJ>hcG!+hKb_GTu4{a%>GBv84Ur{5yDLQkHh08{Kfp?TVp
zU-81|0+Qc{d)k3bWHGe-Uz^~6zM_6V9;%-3D*};jqf#yvbIBq*iB%SCdJgXA-{k+z
zt!fG4V%W?V1vEt2olW!yrTDshd5^P{lQ|e~djEq(@&D;9>r&A4850=`9PY{JG4jhW
z8?8?=moN*CWj$~o4eus6g|GAJQf|k!X_L>e(P;P!WExDk!c>ECjfzOBNRq0t8GW|0
zbT1C}vO=)1!4I!vP?otfeD|MA&A$+>mCJ`~LxlS{Q$z>e-AxzdP47=+rNeJv5|2h+
zV`3s<BsIaAQEwk|?oZPBWV9Bd_akZnkRUcOwW!ZKt`B7=5Yy|+gP9<;^Pfq1;E{bD
z&JeXHDKJc!j~O>YwgVSOvBj<1E&Pk&C<jOm3~o+>m8Y!hh1{SNf5G`-Kl1bNAI6C#
z8!9+=+)R~YC!Xdxu-`s50NX0mZ*c%Vh&zIFPf_xPRik6B9g@`8xN{9Kfj{+gZ@?*A
zt@<o>ihdli=hCIR!R6T+3%BS)_gn@3dIk}{Rj|&cqHVLd@fYa5%R(W<u()9}p`3l%
zitZF_ZCMBIR=K1v;I<-C59dfm99~}rUHcg}#vW#FWK;m93&7QpHvT_l?1uk%^7I)0
z%Qc}I^qFv#d?xnQ7{VPnQ9g^|sQoFx^xgKp@?vz}T55vY{8O=#3s-`^kMDUMsYyxs
zGPjj&`}(*XqX=-{uAy6e;XnSMc+x@VZ0-Zq(W028O(*HcyjUvo;-`|FNB8?)`siQ#
zr6*<4=-qyIJ=#--cooZ1lMCH|-VnzoqutTb2C!;mmmgm%Nl_(7Ch6npRIOXd2J2&w
zya4vxYBwGMqF?+<-Fyu#o_s6JJmYMOxRjp3qolObHj^BSm(y{}nMFcMTV06V^Q)y%
zX=g@-MFP!>r*+@|2x1QoMAW+pP;RDxv(y%BD0r`=rFbDL;BA08Qquerhbl)#K=-Q4
zU3Nxqs1PY64d;oG*H}p^B|s5ZhklG}G~KT_YI&jmw}0~TN%|Eh)(D)>eYnVpY;03P
zlC$h}QQFJ4FG})PGiXm@KCr&fq!VJvrU#)A+eE^J%117XS`+P}-%<mA)fzsT9Bn=h
z{>m>h{3E1JKjod(c~RFSG|31u?_3dVgS151>FbDGIEX)<Y&#EK9*jIf(Hj~M&A8z>
zyeUR3FSxen)N;tdA2jU1h~F0>tR7>IHRa4IA|WP1I_nzVy#FqfE+v5Xi7wk=H<S94
zjT`O>OGMY`*Xe|Y*fortJRM2CxSi`#_TsenQ8QzFKWNoBk<br6o18xWhja@Xt!wUm
zGZVbQ>rhooL)~)EE2q!^vfn1(&4Z=RWol1~c^h-F7lq2BqJRZ1zDo%BJ0q8S-W*09
z8mu^cU1`szkxFoF@Wz#+$FY@QNSckV>w2L3C!35nnvW6q+O(w5=bXmI<Yd-%IYRaT
za%D!|7M0EFCva`xQVBep1fIpq1K<41$0l_Xd90(q?6o@=fOrN29sUd$YK!VJk>^JI
zo330ZF7;Sv-AKQ_Ui&C2XD}LmeI!<G{||X{v4q()=>?b|<KXMiB-f_IHk)^G#o6A^
z)Ffs3U;Ln$^!j*bVLgf`TzXV<l(dxmRJ7{j*f5>Aqu>SJ6XcH^zRTn$W<9^t&wQEl
z@#0a!lKX+P?y&Q=Y>{Pj(1*61F!rEZEp27{r-TK(3>DYDC}1i+foGFfdbG)JbuHkk
zXN%^T%d8NY$CFu<?{#ycrSd1iXQkibNJN>~@m-B}kP2ZLhoqlm_$JOdn|+t=zvMR6
zx%9!`>F6%auH)G6CaTu&`>U5C1Pn)I)Yb3uh&AmI7Ph*!n++us3j9lysrc5NuqFUk
zIfxnP1{)-=c}jh9&BcFXNm0AnG2@ve;Ru<JHm%KyLxfwkL%RwC2h!`ih={?Tk=9_M
zrO$l4fk=XJ>7_|6)QZzy9HKC6Bew!NIi^k|+9<lO%0E*bj5e|xv!jg;p?&%|s|~uv
zb7nQP95mmKG+(his&sZOFvd>yW^)Pl7byDRP+uv77h)38d&hNdMHIfTPYM<*coL&^
zoxuJOJOAt3n%s#&nVc}rcW3hl@cG>NFNL~+huUx&EY(zs9E^aTUfoba_gp~eL<%v>
zu_$&{A4-d|ny{i2+aa&y5wG6yW1_QsJ^e+_<8*`#IQmH3k5I@H*xN2)n9}OA7?5OA
zj_q2VP;6xh5A&(y=%$(UsIm48Hw)o^XV#`s@GY>17?E;>Q$ItFBqTif*+%bYd2VJI
zY%BUW)`OXhzKIZl+GvH8uJ}lcg2imk;P+^+X^RliN8+jCm(>DO3z8jI5pj{q1(2#m
z?ciMv!%3SV^03lv=XPOB?w4{tt@DOHJ_#3HO^1Dy&oJMfX!aMnfRpNALz}|tu&wH8
zG3e^9#OfkVxvYF6pR7cliOMs}4pu{AI}zZvdkPJs6j|Svh>g$DgADbf?@k4-SKmMi
ztZ0wS;V4&N_W%RoQKV@<>wx!Y0onO%xp;x;@uPb=-6?<g{vbfg2Ay$00({zezOL&1
z0!t~0ZgN1Cj~4GSMtB5Z;W7)c-J63><9&gOoT-u%mN$>Fu>+0udvI&_Mc(p|mw^1-
zZ_D$*N$%mLQ-u%4%+OrDVL(g0I`j2~F4S8JEmCjL4dU}V8m;r*TdR5O4<|J{GRvEf
zz`KMdI69gSUJkDP(6C0HJ>IIo_Kaik19k#fquaG8QXFYs6z0d_LwY8c<j!~p#2-`>
zOhiTZa}Rx6lCIL4c%yB-tiEQ&;^hICr1}W5rq_zN+B$8s>>Si6iQH_wx%-!iwC<wT
zum~A(1Kn~OW!OW3)BHw$ynjO8^t+p<{e`68W}$e{uZ_Zm6E@{w92_<^iFE5{9Q6M>
z=;L_C|0YX0>GnDdVB@bT3A8BaDc#05O(MR!Z2DSNrha}vQ;Q62wL3O(M`4b&2L#Vd
z5Y8BcQ8qH})lVwZCAyeY`~>bs^InD`Wvd0TYXK!mA;*0zs>y3OvDUHID({Keksm>-
zSiK2g24YJUBBI;V#?v~Fy7&fXC&X{6<y!j%$#0~dIC>`~T1IRKY{H!L;Fu@5MxJY~
zfdZ)VWEq49Pl?aqUV#CQ6}Z_RHvBlKClfxw1i(5dpobLl&Aw&eXp1~h*`Q`TrZRCx
z?A8r|vCpfBZ=)XzmtebDY6S0MB6$-l<fFl7%dU1Ry_QEl54akHy#>;<)V5Yr*sxyS
zxuDv2m2q=fR{O7C+84e>s@DXo!+VB5Bte8p-4|{y@{&H)H|_C9g6Z~Emfji&IgpvP
zox8^pn*S~^v~pG(INQ_BAHG|XwQwo@!I2x$&)tTotpi@R9f<0LhyK+c1y`bkK4q<3
zocJbwXluW`a%R0!Ft=C>LjVxsG*|<{ikB$Al721>>MIWGS$SQ5I$SpMgZBfS3)RO>
zOM;jM3kK)Ib_8>KqTXQ7eiA$?^cxWC<u)5?(kz%*2)E5OL}7elj`@*i_aqkdvjZuR
z3KR#9Wev1MA=qTv8wSN!$y)ys8(4AnwlpLLCz6c$=vPCRh<&X{xky#KQvI;52IAaP
zI~$p)g!j)MY{!Sq$_b!PGDD3M8D-#vG@-ss_~#Ta0B(Y@PO*iz>T{awMZVK92fzJV
zf5yHHmfX9MpySLS%)sb^<LLPf?81B8b9gjn&E+?>=lY~3Q~n84|6kOz-0L_x8r?k<
z+Qn+b0Q2t+m%m^JDqAEhUC4m-@C;+&KbU7m{B@oQoJ>)1zYESwA>!2O$sW$nO7RYA
z^WkEOoG1rnQTE?EJxs`tCCwUgJ2rH9*veg*kvZAPwM$W}HIaD$d?qc3HP?>e0{$%Y
ztZiK*C}TR1A^5gy6Pp6|x3?39-tZcS9u@fDTgBYjVE|~vMc^31a0f4BQ&36r^7Den
zJGZOZcbdP7S*x<$vBq|TfNQ`mc5$twx)o4*h&9_o*WLvRzApP@9)`$&XqZyJ|3@IA
z6ei47!Q#d%_WN0^V3i9iCr=FSX$ii_{WVll1;^b4b{<aPb}96bV+kdaUV*L9QN~Xc
zJoZRCe5MkF_CRp@afUs!M9&*elo?zN@*3~Udp%vn%#*~BGP$|+HVo0g%FVE*GWY}K
zOYejxnd>+s*l`K}33lI5vjr9xE)=)nqEwaoeVm_eHno=i^l-$QZMZTU8dO=b@ITN=
zh^i^e)%F*#3|AA;rtGYo2H((1=G><aZ`5aa&0I3t)9%59z4Ky~9kHzO8F5&5>7kB@
z<yTRh32%&@7?U%!s<i=;(qx3rU%H%5@P0a=tJAXovxg!V?)hr6o0vQgR6_bFthkn6
zHC&+>XQH@Ie_$J6{KZJYC7S7>{$;%s?dyRhgcDUQ+8cL4{Mx!EV3slCQR<8cXr6U|
zl2V>K;eKMTE?z^5o1<2!C@Pk|YjMqo@E+aPrj(+7ve-iK*KuPtY(C&_`BOvUH5n45
zD7zCQV)o~UR^%sw>>tTlb8|1M&!>UggJOqB#)wDq6@_BkMuYEo?L0o=8=SF_o`|_n
zmk4<16A`R)5+oQieg(JT>$u{{5c!IM+zIi(J@Ts-yqwLP&K4<2g@3|~Z`o3CNeYQ_
z&Qe9er*9m2sZ+_J6fRvt4`GdsV&12<)fc@_RsWMs+@~o2UCm9BHrLs-R5Olan+nY@
ze;b;r)h*{xCugx;Mt!2*^5O=~Qgzj-qy1eGQj)ggemU=XJ_FnAJxk5HP(Cn9*tVVi
zRR%2NzU%MX!|omvp+I5lD~b{(q9G+@7joZ1Kf+o(p{PrHWetnGauadWHCLK$!{YIv
zD}X3l%C{O4R%$-Un>5+YXgsOL-|s?}f?aGYkNT4hAQd>_-+2#mD0#0s!<+?Y)CYd`
z#_oX?>9A)JDgy|82;K4GC}c$1_%MV-GK|1Y&YXH2nj08z&v<}cSo=NWdk^Q+P20`5
z0UUjFlf<es5tUiKzmnOo+Q-h{HYw%U(4_G)i!mVFP0^#wCj$%zc=VEFTu&&$wbWZb
zTltNtG(X0EUWF7)ODRwf$s+S~=kd4XEAFb%C!0^p*X+#xui^RsTRejQSG?{Wj3`Um
z3nyvJ@iz2i`_ZTZs}Y5@lH`4Y?3{b8BJa7nin3Fb`xwGq#IqBh8q=Id4gf+6Mla<I
zDy%ok1yn3ZiqP`<t6I!VIj`K&FWeJHxSXw@g(zii^KLAD1$@T0#>y`VzS6nP7IC?@
zx`VqlYA)T#74l&ldev-G4Hokv)Hi&cmbpaIN>le$g_R*y4yBnpmz&GqHzl2~idp0$
z@KfKGj&(Z3n*;ry-G6C_qi-#)Ha;DrkX)tpiDP&;xbC)Ft8!iTagvF=3V#Vzl>HaF
z8n(PdgiqMFP@s^GTCzr=NYXPm?URbpW{Q|VgoYQ6VuifF<6o_Lq{h$gLd|!XW$jmQ
z+)cCveOJTPfF8x0f^EfdGjzCt&2fRte8VMp@E-ObI~if{l%U7|W>)#%rm=Jvr+!z-
za&#)VkQK6@&BVHRa;hzCmmTM8V{L(MrCU-?;lGE<)qii3;Mj`>HYKlM9L=JklMWQj
zLu0!D&+>&-#fBIYQ}10`=Rf&7w9KS^`xrPw=tT2IhEvX(YEV4rKF+d3*1OJ8s;$Ch
zk4}7&7hnq!{rl&(SIIUC&9-*^BgOK{R5aVBN)9qe;D`F*eT7s5tCZM-3=1}!|H}Ef
z(1#QowiO*v8mQZ?bp8{ItMFB{sTTzLXih4mLG^XWacHKr8=}e#w5P9RU7Q=P2yi<X
z27c=UxdxW<J!rBhb7>V~M7%wHiKX-lXIA=BDBV;RG)|hPdqM?wjCwxuLy4rrk`vp{
zc3&s(PT@zwg%UOh{NC>tb{rn-i=VF;j_2fki4x2IZAyIJnJ|d}2AB1Tvps%kKacs9
z&j<APb8cL4v!>3{N<fg9epj*e?DQY58bJ~_>ek9{<OFc2u5bYOyFtCJnVU5M1IQkB
z-_?0jF8r?%J)=;8uaCMt4<Ae3%hSU?mAEDSFJzy8JLmj&-<PU3*VbgK71t+9c_>+C
z9J?GD^vT30{J-|ke>a-{cTfJ`{JnZKv`rkPI<^qRy}hNLl)TI2HJb5fJs}J`US10G
z)fL)Owmi9^2x_)UH_fMEh8vTLLZkWEKxU1pU>^lY{i*(OG^@HH`Gt~q=}-9%pJktd
z&s@Rs|Lq7K7X}E_;e1}e>|rjUC|f*{w79lnUw`e%`Z>ZNP9Rc(sByCE#o@t7qoPwc
zhlAAb*Tx*x$#TDk7$mgdc1;2q7J`TsoHWy5b-44MiFV)0;adNbE$U{pk>PJt=dWMC
z{l#M1y)@slzUu3AYL6Ozwa-jV`dJUCbpW}!f&RbPd-Hgx-}hg9Bne6O-BgsyRw6=%
z7E9hJg{)JNZDNw_V@71(LMUWNwu(vC&`kCuyDTFHQwYP1b(rOS?%tp8`7Y;t&tK>K
z_xt_fA6{OM*SzMsulu^M>$w~<p1~0no5de~kvAH?8{@kf$jiM1Zvtw0Ek*QpAO+Ul
zTE>ybc94|ombJNE;>q@BF8<!u(&wgS)U>MYx}}sVM}?qhCI?}f`2veN@#_H0G+R{C
zfsE7_qdC1>(w(yF@Y$2Q<L52i&xk+_mO}oFWpKQ#x+X*MXxKfH?BZFIqsh7kMJWu6
zH?WtX<>(Mh^BugF1KoCd3p*|CY2&bZ$ULXi8t?JyV#Pb-0-(}C4#+COSH5+yYw!MX
zRo*ca@F&sI!~|#zTD+FiBgU-eudt)n;eA=lm#CSTM0j;p<c9x~o;zH#PH-~)wvBYk
zqSEfEs+#1EUOo?_v4SKQ!-f1mBO};xFmJ%F0lF4z=#OBBE_;8;%ynNJy|2wd-Fw(>
zwbT4-dCDL_C;S)u5S+a~r+hz!8CW#`!aW%kb_mEW-UJ_$vkQaZ=hri2XfZF>-uu!0
z%IU92R#99mZr*rit4uY4Oc0oAn>&Am@EBC<-r|}pY4mVNiw6+;YcnhHe(e$H@q)d6
z`LpdhX2yiL)Sj&#HlMa7x|@EqpL>(1VDZ|L)P;-<?nhdO-o=N-;=ZZ@8abHD{0Qn@
zn(d6N9@muSt8f19t#ae|k5}OekbuKNh)uKR2G0TJ$=Bn8kzCat#wA~sY&_%}qF!Zd
zW*(Pt^*``B<d6sFP81zO(`_g5_!qvR8$08vb45wzzW0tKeUVKHi=u+TAdQA>HbZ*i
znq6jdP8(!~_Ro>!3L-hOSvvF@#(t(0y-Q(KhxDX;b{Kxl=iPY03CplI2AVWjJ1CTX
zyzlf2rZU=#SKJjkOBSyyo8(2JEv`U916u%Yy$&4kO!-S<3!11=y^u@*ibJMiuiZ?K
zX9xqHO&ic)L){$ccsrvK(gz=k`&?g@{&0?f6-Xde!rMs^4|`aI8Qn5U!QQUm;zGyI
z;eo1e!<BOBe2j~9!d57)Jth{VS!OQrxHHFX5PICT@6=5iVDoI$z#o&SH*tcdvuea8
z=Ty?eiSdbv(&atTqKsgw%PjgOEo&wvhUrptZs~&6*j2*nK?w}WV00#nDbi>fCcRZd
z)3A6~Z1U>4k8F-_J|Dc2K>6;=P@o6e4=n@R=eLfFT1Q>{PU~g_Jgz=uEGdb51xyz<
zvsYO>nV1|5&AJt+USsaUAn0Kd!1da;1!jp6Yq4cvFI-{@Qo3y>hXY|nVL(*yU=aN}
zXu*Thyc<jt@&!PXC1wOc*f?&-_Wm2vZUaliOjk<b5}Op?F4#KrePHMp-rwpC);@3=
z6OT^@7we(Gs?X2(8}bWyn}J*XuQzP(9F858z7~Z_-@;xH#l+10fb80CWHC-J2?2-W
zV0v)b)j3nuY6nldzK7TR;7VFd!HJnmC!xfU@Gl((Fs5}_5q0Qu0E~~xksoCCms=Rj
zs4DsEK6KHc+hA~~0%R6RYkH`&05R_0$4(M$$23bbPiUqj?~dTA6t8};nfyJ6v<E!t
z|9Tl%V4y*Kj1!RBf`Dr70@47A(*XPQ1FQ^Q`WDKGZaItOSu~xq_d1st<%@le0&gdU
z1KumtRgMy+zy5IlT=sm-YU`UiE+=Vu@OfzP7+|x7jF6wHqa2tzULA#RpDj&dCKvuR
zH5{`Yn*k)HKie|<upRRn-F5~iWM`o&wD$FQYIEP}MF^R_+Y<!DRvgF|Q8KI-O;;4O
zU~VItvFz3oJmy!yVJtIHa|TW|T1`xvO?i?%8ouJnnvZc8Hf&<{rk_32`Ti&#1Cnwl
z(*q;Sv>MRk^wiEVAIKL^qrd`Y>XcM3gaa`l!g44zZ;2d35uI`3zr!%EZ!(+Ml!q-W
zEV{zVF-G=2*nR0G_G847EN^e2Rw0c4O$hf-rFA}_o<P;J|BY9mXOuc<*jk;g9QrB0
z+S00O%;fQfmkx9M(Wd08pea)j(=WNd!QAcm=d-Z0treJ4bhRx@R{q46E4@Y_i9C2`
z#4UOd;aZoAD*OiVSv)U+tH(^&5Y3Vh%^{g1Yu}FZPgZ4Vu9YlvQw3Tu)ma^;deXCT
zb9J^KvKPgkJPhDS9S-T|9)?kOdji={x?w#8`Wx7P4*C@WbsiH1-4G%$35-*4x?g8Y
z$}SI_4QFLQf~-JcgYf_S8Ot@E`r{t>S0<A_Njrj8LtUo{B%)j8UaD3GY(H$4Xo+zF
zmgE4Y@Lyy58fF^AJ1=bESSk9?4L$!b`mUU-WN3EXQz#m;Xk2r#LN6SyRU-G%m1AO^
z*S8|y4<3k0``U7HZ}3cDJ{aIx!Q8Op*j8*6_mkCX0sbD1tvbJiIH^0m(`ObDYdAKa
zXh9B~XvKenTRh~E;5@4hN;Z)D2;$}%b!q}DjhN8ixZ4+SK@#){E8&?K>J_Mvk{x~2
z@XL?a&MtVKE4X36S8XBqn2Y+aj%&@L<lrw(EmSo*KDw=q^T)}u-f!OP2l_v8S<4)U
zJks$3XdQsjU80JJ(r2a*83}A{X0cPQvU8MnGU=j6a>yl*738^U80TP#z&7<ikRce|
zzae$o%$Hh;m?iLV{n>;ogy{s>8WMd9tipB7!ZTbic7nmS$mxJ>f=&StFnvX$Ucv5Y
z5kxU#!Gw*g*SHs~|0(Eg0s*@p!|x%LSsu`=^3MC`$ETH_iSf7ohMa;hK~9P&L!=_$
zHEyMqUlh&-bl)F?J$kDJEf4zxnll}kLcn9~^S-@4+scEv_{9Qtl}TZ}(m!fUb0G&N
zG95V2PIwWYjh>EPj;e=dph2c@K8Ba*nnsQnU-0ojz0foD-TL(IblVZvy};Je=R=d(
z>`VD(;$OSPb)^o!vr&8vW(hIGH%f}LDQ!T?P%V5DBf3--Eaj8KlWDXGa+&aB`sn%&
z<0gAlPK{N9KH{`EnpG#JbE-1F^=h=0(21(ghh2QP^+AhllnmhgrvkxHb#I@ZO?Fe>
zu`@F6X07|=hXRy7GkyQ8p8>sEIYFxR_H%CjH_wQc15$WPx*)jzEBzC#%J2h?d(3--
zpLc!*U#OPxz<B$Kt2=jjSSXdj-d{e`xG!R<UH@DYAsu=lQS40_DB7ZXAgEttu>H+<
zAVoT;WA?vw4xlIqk@RnzKQc9G@OVA*QhF==(}+JJx?x!Dn5SFDZk2F4oIn!z<A1h?
zEz`r_lM?x=fsp9u8>o@y16#@bjBAAv)pfPeT0tFr`>m!1Wv_aqX?N_2dI*5l^}ywI
z&~h0&b6}kTI!pam!VKZ6nC5e?b}(1a;_zlmgu}MK;7S)bcZyjhK$lkk2xiP@3CMHT
zpuyJBtVj6ShnL2TfwmG_3OxkaPuN3*Wi5Xo7i>>>2>Ox#2?7|dpA7J_MbE>z6|(J_
zU{yxsqDfNGWfgGd399TyrcPMS^O~fK`;-(ua7ONzUx|`0)dNz`D?nx^fGI=Q4OWRR
zpY1k#X(Jz}tO9?Out|Yy1rgYMm#Ab5yZnBpO8KHVUyHe7CXi^i)7;6SLbvV!NQo6<
zTWm^bD39KuWm)DWq;7!ku-ZS_ca;H9)J#Fko5S+SmXbmJo(E4mtLTVg^*^Vb;LM<v
zFvWnrZvI;94C%l^tr(@H`$3vU;Tp>J3ml5^Op##^5Nk5F6E}qe?hm9*9eJ5rcK)#Y
z6yc6O6ugxaY;L-d_Jr+V2~Bf2=qREr?@a2CHz3C#=}M$lEDuv@6fNU1B|cET5#%~F
z_wcEz7FW8YV%d^`7+8~_r5RODR<o?U#GE3yDA79E`#&DVwd2Ig^lF~n`t!Sg70-VP
z=>Pxc|MnduhsKt>sGA>ci+p?jb3V_BQWa!5E-S2>)=QX5iWC?u0jIT7d~cT!qWpe(
z4^A$)-Y(qC=UFH=K58!2bYE7&RXsP-^50H$C(}%xcJ#^by&|dH3Ax6(@MfeTqi74=
z;|E6Dn?0?VQ&s=b1r&8&x>oz)gsQ51_{Wx=WO3ME!_(1_<SUSWRQg3_wgv4woIS-9
z*z38x(W5u3@b?{@E}h><0(XJb{BL~*`P0@4xg1@ZTonc0VU5e%f_>F9O?CV!6({zz
z|C6(=9bQ`X_F5;0mflHiP)_E6J<jL5dE@%Xl>_f2EiWFtq7%-$-bWN>n*fx?vNQ~M
z!}J4HAP>u(qxdsmf4C2{jX8kvMr#XfO~VtuxejGNQQ8fqrZD3})L6lW!0}{+@Z54g
zFryhL!i;{y06zckRdDgi1;XYpZgzq_Wc1aWNw3TR)iOW-0|zF>jCn2RZRAt$87wq^
z|9GI}a6f11e|t=SJGm&<5EA{dK7P{jl6(}j9CGW`tGb5h-+`A-T_sEz?Y&w34;qM0
zV1SI=Y^sNu#fl)NIH|d~8FSuV_nOSP@%Twcj@Gv_WY4W-#75+nV14H5mJ{{z`j;U~
zm1jO{=eyiashgt((Sx^foobSMPGhRwKrZY-Q`pU3wp-RzIis_8B%#~DgfaiWe@Fn-
z|MT!~vg~Uz5}wpIMsr^fGu?w7d1rj|ugB!u3n3s#{*w@JjzWTvc)<R*U2fWY3TlQ`
zkac0%vnRJU*M6LXoq|;U`?vc)zg1KJp7X_PQJWvL7!lTezp^2p{{0e>u41zwqN%J)
zk`aJdi9mLfgQJ?%<o8@W>5}5*UGqaLHBRN1X(4;#z6APiPU%U6bof5Y%pU?^3U!1%
z4y9T_V}X>eX+_W_gL))ykWa@7;+)uho6x6;*PB|?&U6~W-w{5@Bq=VpQ+p;8syqUI
zR>`EVjjlHXcEQyW5G$IXHB~r8C;;b(y`rq=`jW0LY=r<4;5US@{uOzM@~b1AISJ46
zp&J7&e?xR2RadNQzALZ3SE(q#vOm@%f7z5mUi(D<VX}r1rt_V}N7Mk>V>R}}#P6&d
zmq&dw9>qR2e)dve8nG$O<}ogr^dwZWQ@AFZGKcsrsLy{o%J|64;y=_=r_zJZn(JrX
z(<tA5oi-_%GMsAjZn=Dee)-({E8d6B3${EKUlZZD@Xur26t}J2I<s1zr!CAFJbir^
z^t^7KtIOOTSOu-pA;!S_JG(cBA=aUcn)j9UAMYu@?-&%7HG$-!GwxP3UFHkI+~7>#
zvcqg}0$nV4Z46M?n^kPX2_y6csGHyNM5vSHtq$+2?pFOMy`XTu<3berJ-RD-5Z#w|
zF!1im$ekC@f2LU;C*-9hhNr|&o#4i8J;C-83NYg;SJ{2~ykF$coeItuaQ;8%_oCx<
zC9PV|wnu)enJ1m#Uh9E4y-K~M9&ym{r=Z!trJ0*rwh!#uT<c#PNyLaGilE)XkH30L
z1Pwn`?-lvnkd7Hb#;t!+mq0i6k6em~ZU=#%>08EBr(Wbg)VW8qqw-6=bXA+H%)ywN
zP#9AP?l`Px%788I=DiI*S5WG>NKqISuAL7!(UcC)B~fo|y0R*oj3h*1Zjq6HL-yeM
z`{?ouFlvu@i44X|iydCgv3(B>4j#>()u9vzGdfy%d9Go?{~YR>N=iDBaP3slR=P_W
zMLaWL`%W-i{}?Jv{RAg%oxPtl!?HA`ug^*%@2je~=;(#(Rc1&|<@&)ZzhCXXGE_Z&
zATv{}w!(<-5^n)4Ry5#WeXo-rKt6UFJF0f<`803DkX1FIZ@l6z7715rby>^j`*U;@
z0q^LkUFV$RUr$T6UxVCUV(2S6eAmR)&c!D&=5YVsXaBPgeACZTwqZ46M3cQP!&@fy
zskra>msQGZbdHU+;!Z!;?#YuL5S$R}J<P+M^_<NIr}xV0GAw={H**M**469euddy%
z8#FIUsi6fjMJr8NsjCOIrE2QdLLl_etvjqT{9*JUymmffwaMXKj|w*De56_2m)dZN
z-is%91-=sQzulpMk_lOcoZ$RCW2lJwA+S65oyGlcqiInibEIWF^4MzmAbdsRjle%i
z;^_N(d;hU={<k(@sgKWeRbrQ~`X%T;pEaMS^vzb?)@XOb)a?1rcRtcP*8i5t{@Yai
z|MqY6NH%XUL*AUWI@5$!qEg;;&yobdNLC;_^iHNLl;RGmf7wa<9v6R=Pi@`@s2@$`
zeIwo-+^@;HZaddWnFm#NXG~}MqW%TClTR&?JcsBJ+$S`>sUiV+`EwpX2{5l}P}gTJ
zxHszz?;3q7{_$HUms^N9X=HU-e|p6FUy%8mwY61nrz=RLu({qlA?^NqFXBE!Y!@(p
z%qR8|M%af*_JCB9l)DR!DeygpPgSSnEhiiL%RkX`;m(FEW2Vz#m6IC{ADD2OQ*3;P
z=ESRsUaUsaQ*{4@GZ9?q%I_*r;}2AJ)*tlm!1&n2;be)U5;w2+bJ%+P=h}@!i3;R$
zkPP{5s?zL}Jmr2$4qFg<$Ye-$09(~-qACHfBi~GLKNiVo73MK&Uk8cPOHPF5{8Fy4
z$XV?a`T38qQQ-2n1{O;Sn`emO4;^o8vz$SSnJeS>4gc(W&I7yiY%@!?8FGM^^^rOn
zuYr^B*di7uVfGceJcuxv1H7O8Uw#h~crgePUC|vN3EY!HW*HYhG|*s?3d&THLx<C}
zMT_GBPG@7P@z#5K4nx1Mss*jkST!_WXI5P0CMi&_m(mgrdmW-gYcv9pI)XB(s!f(&
z<q8Vf%y{6oe%#h&@xfR{XeaRTBP2L05B>!5GGWkGAuY4=3$H!7SsNz#0^bQl@vUc6
z<T7bGq3=*o&b3%msZ?Z1$8?YHas8VR*TlS^cL`V$QHk8~iDAsVG0b!**UkL+w9^xj
z8+&B$vSlf^<U6=%&}b)APVctr=BG>7Xz*Co39r>}X=gTgxzJJUR-oY}{psqw1+BBa
zRBmuYy=t5u94&WgSn{#V!6&!3@&z2g(_5muv|xOwBU){wLoNM?><u}mXUeDV)@Um}
zSZ>c+!}SxBr>2z5f`FXCiNb58@uwlOkg^m1E92C*4@UiKH6;@Uup}M{mXyZ64-47D
zik43+b%l)Yjm-~90B|OWYTvgb(Im+RD3yMDyZ}{enK6rtCGnw5$w&d>OQRu!kGGI2
z1zbg!Pu6#(tUMkDoyu~AKR>0(R%}hV|7yYD<?}W-z8oH|0s8h}>ukQ!<W1FT4agG#
z+%9PRBL~NKw&JQJ&nV-|5{D1z_&FER<bm3K`K0HXLy&HZSmxTdDE{vZeR>=;_+l*&
z@IL}W&SnRKrxGRx^iIn0P|v`VtWUH1){haJn;QL(Io&$=WJMOS9=XL&*Y&04Y^6~<
zo*z~#vG3Vx{8;<z?oF;|*taJ{=dx_k=UIj9<C4b^B+;O%TRD@HafbsIxJ4P@fVh{%
zN}Oh8vgK;#a56W7MsHm~{C<jgld`Jpn9TbV9Q=yEi>)(`|9Z4_`I8{@*lS6rG4fG@
zAX^F3Y)SBtRuf-rI5s(#Rh*==WrBJe$|)^sd)Rel{;=(e3i={=ek;OVSxY0@6;eBv
zn`j#P+If00^IEa_+i-|K<dL?*TsX`Wa}TfyGqKY8+;rPE>XdPCu<tz~Qm{MA>9gV7
zNhabuVrPK$oN%86Wq0eO*TtyXFpM+mdJ3n`F!Yjs(7wQ~GIESTu#DQ(#OYuo;+-hd
z4nj9z382MkQPZs}d&wAIJ4{aY*16^6BXpZrf*R<4p%UH$r%2y<^&2D0^!J=;OVT*u
z?>X3Yr=$50GE0I2w?%`1ut||)+E9OyVo+yG9OySJ8!pL=W-fQk3`9!?D?p5{1gR~*
zU>v98wqDaV%agI*dQy9>qKEwhZv1Nd_>(i)J{)$JF#V!7&<0b94bPmHNBK2*bVs~#
zLt&Co;DNca>AHSJ%**{~`745<PBvWcGX{c;;*7My0&SPHBAALLbn%u3x5%Q;nIq>8
zDWB<-_&og(Bb-hjg#tCY1AzGGfsER%i`@LundL8cJ69lE()iS8mI1(sUjlFb6vorR
z672sHIW5^gyALMJTB*#Y;;)N#$sgz(;uK{(mWk8wTKKxqdg|-L{hRv;IWVH|mNyN+
zkjb9aNa0)Edz53AvvUn%eQrK-_1YER0$sg`cc`m$t^7;*h=38c`q15$cKuhY^PeMM
zm058>0ad#T$H)2r^GFaF@0`x8o8@juR`K%4CYeRW9$VpkiB)rSwQbY+<0V$Fn*LI{
z`3%0Ws)&3#QEPg<9R(ESww}^PDOboQHE<DUQ~#sv#cqqp<0nLNmkl@b1zw@Wz^Af=
zd4#!lNLfVP34OgNv}l|9SsrVlF8JxQ-pcfWQA$>ft(DWh&j>Y<Y+M?_%R}5t^31xo
zY&3uZv3U|0dr0&E_Rwr+%JR>u2J1UTp2Ecvd-kh;dF?w^p|mL#4wEPRpPSv06wHRp
z_Y}IyHZ$c9OnO<?4;0;3Y{6@h2a0V5bkG&Q&49BcfXnatpQ8c%&pg;lp9dRHs&XzY
zBes=XqfTHv<3vz%3N}9?hh;W!m2Y8fQli+n4Ks@3wOC7+VBt#|?uAMR0Ph#fKrArH
zPnmpldUj@s-!BiVF-O)L^#`(<>NhrV$LKv|K3p<fuIWnoL~zzaM0VAuCKm-t;Lgp1
zts;c#l9q|pNT@W*dM0#%`9ft??pBsWn?-y2-w@fOpIX^G7^*p5g*`}gE7Z8?eYW+m
zm)7<Aul47zaMO;4!`ujN@FxH|wHMD#ztJK_N4_TtsQ>k_x9y|iFmsHlG<;I@;t6Ys
zYp1zX{xN;y@}k3nY4TgXP6*2<dQ%b&I*H4L%6G<Yru0OiP1#?u(#0$+T01DP1B^)I
zs8<g+lCf1E_xu$)ZMzX#;Sr#Jd|+jx+=8gvqrY-igleK*QBtZ&H)Z&vTjY)b8pRz<
z#lnc6R~|=GMo}F;$xPZY_gNMvUzhe4K=2n#lf*sq^r-l(S$?@+s@0Q5ld5-8C1l>7
zkk?-AyGq2(<F=ewpYZbR?&XmXr<SEDSszTLzasL8#OA$;FR&CZhFl+;&xvPX+bKwa
z-;z;zW{sdeo-Z(@CPi;G2h<^d%R^q<f)lqqw^sF1<fFb=D|#q?Y)79As$%mt($ZkD
zM9OwljqHU-=Wz<>#OgAg-M*Ni!kMkPws}B0scyRsN=*Lo+ji~(ck7W>+2hR<-OnP|
zs^-(K_!Hj4mZi(t1Ntg8G=bRV6%7){xX!n2p_A}h!2x;g`P#5j&dsGQGg?V9S{}tu
zU5#TOE_$bb`1{ebpFWin(kd%vAxi085NHSfkD$ailx$8LR6tLc+%1jV+4HTRWb!u6
z%KXiXU768^IqaQ%SuRL-Z;AM+{Uh1~Mv!Z#m7twd`v>vzL6-o6w8~p{1Y2s*|0V6f
z<EI99Rs`e$zv>BrYEB?Z!NLjEn&E^;372WS6*-|AF~6etd<1qL?~JaNmk-<4(i{zs
z_UwHlt8nS**WkTU>>vzcb!@dA13dMLqt~V!2z-uLEppx&@fbVi%IDtJ<}3<{S?&Wl
z(1adGm2aiek=8YJw17CtJz~SM%q3T7OO(kMum!HXMDc;E?z3L0;Gs+zIgX)~qv;Mb
z#g&!Ay;toIHOFn7-_vxd6Kviu?|N$%P_dEaM?a=rC=T7iOjd$@v`%{)m#_gqGUhCt
zW}g_n?;4QK2g5pV__g60v!>-Fk(8&0&g9B*L42|Q8_nKf+7WHpiQLnfOBr1rg_({J
zinIS63c(Hp!_d_o9VDm<M?oSe&_pMI<t~Q#In`thFsoqs44j0&g#MRq|9gF2^e9RA
zkAm+(XAks&(bt0;OK($ZOoceU<y@RE$p;Fw&;fWjoVpHA!E9(-?g&udl_S}_nhZm)
z>HX}ZUiKkm?&XPn7%<;oIJ$2AVTkOhic_Ja8rc<RToWiePEZBy-Xaq9nu^jjDr>YA
z8e{)z%Jf0Irjx~wr-$~y0%GvN`v|%yz^!1K)1%x~fbReTm=dkA=yT8R;OqC)fk*6u
z39~PFi6LeP2jHf~7Bg_0%;RDMdqS$N?Xz4aQ(?DQ`FQ08wBTUjgSs1~9~<^xb?6+f
z3ORU&4opy|G5g2p0xkA7kBi&hkB>RLJCQK-xQl6FNp{v`K-pj9q*xhxe4-h-*Qe;2
zzd45et<(q4?YnNd%^_pkf0KiTWr{|Suf=R#9Z4J38ds<fcl?kCSr_*Ks@(ys%tz*m
zZP;q#R@PakGf796Y@&73RnIMV_4>j4wJ0i7FA^C5$%W`g=iI`*PIJn{4VygC!H~Zy
z?TmszW3-X>>K}HdV;t*p%tkdn((3m#MYfg!QSaXb%ev^^7UI+x2U5gL#xTSi!jaPQ
z&i1F`XBb`CrhEp*Z|^}kq#;1TRK98#D{!^Mo_d^;E40sG19y@3hhSxxV)={?Z?3!H
zOks@;Q_HYn<=GI%5P7@9Tz=ZJWzw&`p=S7WSn?!s>ZJ6(V1r}VjUOU*KF|W1rcThD
ztKQWgso@*1_~MeL_3=KsW&@s#S3)7cnvO<Yt{I%IsUMbK=SVQCv^ke~I~R#GI6FN0
zoIMDwnrmcepiHxx@rN^|bH>UG>d#IH-9?_^ye)LT!xyOjfik<XuLk{9N1_+?(H_=i
zgRjR$5vHG_&wLmqxB+(bc?f=}=A@8`sdO#nuX7iaR)7H%=rbbBRh<;K5@{P&iDwr_
zg;EMl+H5H}rf|-+c*fR6IyzM*CzLC0hkXS84DhE?Vu&#jFfa3mh=lV;D~DZ@XQ#-X
zYrhnPKFj@db*qHA!zO@06?}PO=A#C`kVBZIi5gzlVL2lq{B%8~9fAhto@r*Ws0(FX
zp4b|d>S&cufn3@^o9i2?!?VJjUPG2t3k)l-L1bnHI!(E_!;=1R2w}i)pxk{ck8U0O
zjvk>KRD;k>Id!wH_lM9#2!r<@(}8q!*KS`kG*Jo{Pq~x5+L9;VkVaG8Vpdu#+mW?p
zIPy+#ZY;5R-p=%hUt&FF^7|Dj99_Cz!??mQPJB^!dgl|BE6fxgj;);kgT^O)sD9|t
zGNOWM8C8!->;0AUNapr)s<D0tegtrBj3Z!GhVcE1B%jVx7lSdoVt`K`@0m8G2mo|s
z1G|^W5Kgc>xiQnFP$<A0CPS-vY6s>WdK6n9)I1Ttp8NXak>@D|U&;Y;7z#vNX=lls
z@9?73_Us@Rvqp@fPwuTBlPB3?U9h5Hfi}xA?5rAS<AsAXBhqB+q0w#D_=S>XiG9%E
zG!q+WrOgdmv|$`z3-@kXL+c<O%*%N{XK{{Fj-f>{&BvMd?#&R0IhI4K(68*_y37+l
z@9v4dH5QJpVZ8m1=J-Dyy8jnA!v7E611-!R!PfM(<G9&FP{C>^`Ht+2%oq3Gb@P8}
z@_T=)UoeVOkCz3c=*73dm^ZzK{ub;R<9|bdBI*DXY$Vhh8gxafZL5@O+^L|1*$cm7
zLv!Mu>)CjR8$LkO4o5cQdP!i^2babxc?~Vrp-<3+<9h#+x}R?B1Dm{}ExPyzpbj1T
zj`70EPI^r#q}`cj3HDD$F-C<1pCu^LP4>#zQnDVS57L!p@>0dJX0ZI8yXJ|Anmf^l
zjr2=2fOOk$Xtj8I%3iN)=18)o&@FRe?psBlo-5tnIAwX1a2@E15UAIe`jVvDoFVks
z{wuR*r7z?ahhGu-5}sC$o$WZUn7xrgKcTf8I8BO&aWM^NonS^aLrjNP#7~o-el>+U
zU&~Jyq(mU7Ev;4r>Po&_Rek2ywF?7TNwx2nff~c++buwSi+F7|trEcajOAtKB-wl?
zO0&GI*N+<g)bLH#zl}-I#`OXZK_tDK%-6Na2Npxn8}c{rdG9v16$_%h2TufdK(L#2
zVC_lA@QTc<AJVNpghnV=H&(7Wx^VaS{IUp|uR)i3{te+rDHSbHitHD}4*gzkl$yLF
zZPEc|vwh-^F^IlI7{7lvVgf3IxFPMvy&Mwtqv~Q<{8`)Ol4WUzk3D+CS-7!wm_i;x
zxfDI#YV`Orq2tT=DI0BZh5}c^G~XoL1#{e>8=o=MYUYuW=7KKIQ<qAK6Nw5FW2a8b
zeN@z3Q)CK<c52}dvMTiWk=_rtwutRxl?}WP<kWeZx4wY7fXn)VgH2WcqluT7uGJ?G
zekGpEv@5#(=FL9S;}y2QNK7%lEo|dJ;d{DqfpZr^vkejWZhx=dms^}t^NX+hxzvcj
zc46%(f^?91xYmw0T^g}%_E9#NyW3*F{v4ivgy4w;=IE+&#<Z-s6`PBG{7K(hV8!1x
z2@Fy8S1oJ}ru}k%-mH`E#p)D_RMZCIYsY@BYpF?{7nHW)GXkJ00Ch)7%<=|~FDRh5
z$9rhsbcd9wbYtZO;1~eIn&6dDcc{57_h?!P)pdz0KSs3#$?KY%@Z}*;-nr7|TPLlC
z{-V+0B0&${8$UnbWSq<2-lZG9fg6KM0-pz5o60!{0YQ&*_T}#qY6|lXKh%FkkRVJK
z!rq4wmPbn+=!6z;Wjz0I@!FImq&>IiWn~5T<YBpL4u%>PsE0+f)q~VK+vP-uU52aQ
z)_m=PTKg;--&Lx>wOZ1TktoK&R18_P?ukEwl_xmb_w2Wl?}0I0-juE9;fQV3aM%gV
zbQ!7Iwhi{=Z-}IZxo!*o_@{y8*HyuC(!q9@&fRnk4_ib=5UAJ8ZD(N14!*NU;l;tA
zluoJS5<kx(WB-!@J6C_!hCmKtc-NtH<q~FAiU6n^pI~=|Y`+q!FwIVZTTkf)1wnk~
z^&Z~y8Lu<l`?7q(g(%A=5^@_H0!c70fq3T&G_DTZo1h6h#^piDy*Arv#v<r11@F1S
z%s`@_VHM&v>4oHujOhuOh~rYWH?exucggC)mt+x1@2H$#+PDD$up=~|!|w^wYCX5>
zwEB1%W$F)d#`WPy$)ast6Hef@FpXP=Q%Td}DK(&h2Qz2))#DANCJwsPIo^J<+5-tK
zYHzd2T(i_j5IIxk9;nUBZDo$W#~uUS5)EozN$8B_lQOQ&p9<~zA0IT%a2M<^;g%n1
z#L$%#DOUoixS1YoIP-kS;gQX?D@*csO;jqMcC%@FZRfO>QyCgwe?w~2+Pk6r4jS)0
zy-+L*mhqX35<Z9X`20FN_F!WPo@$9e-n-^x#T*S2r%JrUe$_F1M`-@NwYF#;L0Dfs
z$j16vP)vuEOG@sU`uY^w93fvI0#gNTC55Bq>IRu=`S#5U9((K3jHEns8qZwls|j>g
zpz1#+^KpL|d6Zyo^u0ss{HwiyQzQl8Ms$G5Nx@4mW`?+@n9yvayHA;(k!VO!;Aeoy
zl$j;m!HY4isPi3a`)&mlw(LF$NzMoef0vqPUHLV2ES{MPQcoHvQy#5Gh?UdCKl^q;
zqbq#&24>%q6iXg&PLWxjPS2jlOlM%KDcdci#M432H}RcOO$JxFUgFI>!`%+xqmse6
zcY^q5r~zhBoZm2>PM}r-`~F^(-5bv=0nae+dkU%MlC4Ty+J{HUG0R1aJ1A?KIE8=O
zNz0kmQQ%wH6IQ#Zn-D$rMKEqMJ^B0Y-w=j9rmKlQ!EOd5B_Oe%)2Lgv?BkM0G>U$7
z-uP04fW-;UmDPztHt+e_l(;F4ChBPG<lT8gYDVM^98sB3m`#aR58Sig@#kIqwSJ68
zU)+==A$ipTyzITDGc;DgNt^S1-?SboZ78vMDBw{isiR3uJ{KBv)*ri79r9f5-m9Ke
zXjU5ERc+lkncEwhn|(=y=^QKv%C4l7D=R{|N!M4pFyWKNY1Bz><TsOMVANj7ieVzj
zgg9u`XpEjYdE9$`*#Cr0sC$H(@^h2Vs_N+bk){l5^mraO;9YAMjo#CVZhVxu6>_2R
zuARZ6X*2J6Swug|xZDF;)z{oOdG#^_7SnXLB#T%t?PB<u*@Mv`$lz5GBsu*P2)UCZ
z6|eD)pK@bu(`!#Zh%fvd57|a0k*HR<M3{u;?}@A5i>Ua_Tjk|>nfB+cL-*Z-onb0<
zRG^Qc?u|%H_00-MI*a}d$)sH9Cev_31iB*VcnM=#<)o>}DKx%WuC%2PcHy2$d(Tn6
z=iY+R)|I1Zo?gVpsm|60=bXE>HVt9bV)2caPYOkK7XR6S1X@WGFfzPF)?POOF1|Rr
zPhEOB<l6^Jap-a-?z_Mku*AMT1Ik9JP$XYa<x!A{lO(uOaxaC9&AV?~I`dRaY*nzr
z(NN$$EZ&$oWRYM%yLqN5lb93_8Pr!7T<;h(fF_|u=~;%O>;p_=!eE6$7!2Nzv=>V|
zdh>OxZ>j-wh!Q+c0nHl)+YZ<kKNhgPH0Ma~F%$_xHRiLonr-9DV>GcmuHrXk1NY`V
znKwC|(-{{rU81xURwOv=_VYt9np5I0RG3$ayXC!;6ph?209yC}j93EvT;ERG6N9v7
zWV*eJCy1B0I%>cEl}^a~DELSzGIcecIffg8(vPPA<3}TU>MJ7CivHLK`KUHs$N3p~
z`-b=311Y3vXYif`ol$?$duB~RNW*DSTL-tY`s(jF=1pY`pmbRu@R}&uk-~;58v3nn
z$EC0oyG+wr=Jjks0Y$N8*C?u_C>dK!B--T@k0c)!7hQ{?fI@sPD55Z(eft)rY9os~
z3+c1HVwoeeJrKqXh)KNRf+?j30_5=s(*;CvigsbE{qpsNJniTUEkd7-Lm=H6E_B}o
z%MgYt;d#nFdZ8U%jdrp7W^J08(~V2RkPH`X#qRWj*h|&%>|`_-^VSes3#CNePSn^=
ztSItI>lSg>|0~MSRsV@wxHfD8(;-E84D=;L5jvvF+2|r9r-rTI8Jio`8%GI8!JG&=
z!8f}di$5D=X+DfXlsSz%IvdNp@6vs`c*|XHbdkgq(PEoh0=x5N(k`DZh?eC#fZ<(l
z#%>(~)#gP|Uz{8Ox+`r=&+ir0u&BIagby$B8|yx#l2a{s-`|J?o^;Zz8vNNGLD?-l
zdj4<pydS>#>I2;?_Ird7hz3WoE@>K)6$#L2{pOf+Xb}IYt>!?1y9WHClyf(L8ucwJ
zPk2D}_}z6gm%YiIWiC=_7@cx5NzMa%fO)JeD7{T<;5X6yiIH8&Q!01<rowREK5!Go
zR-k?$0G@RK0e^|5+QZ|7ezW&dF;Cvjr-?T-1y_8x;OwmW#SI8m>p03WP)yj-6KH2|
z5(ZHRmu8Us;!46Dgx@KV;g`b`-G^kC=01XXwhzrOI=xL_W4LSQF(M%gIs?CiX;y~0
zBln|x=;fuPpauzlGbcHyx{9hxTS~@}FDqQ=iQmr9A&m?A(n@@Mp=Ktw_aC%N@a6+o
z?s)-oI*nF1!&(kBo9z?z>^R5BvOW9GeJomA6P(VLlfW~1j*w~lUd~_>c(_J#9xNdb
zR{G>QTr};tx73KUXPu$rsIX?ZRFl@ZgxQ{bLpN*uk4TxYJ;stHO;(hs#^k=`II>za
zL0Zprx`)qjcyq}?vU7aJo6^lAfA2uOe8t&6BfnR&T2{3q8VtnJ$;`RIIn7MTr@x#e
zTm?aEUja5<NUBzB$How<m5J~t{*!lG|8nqdk%pahGZW0y#?}A92-Zz$=4`q5FE@eC
zx|DuRiP^E-2VQ4H{Cpl?cIE6n1Ej%9gDj1{`{)k@7u!!-CI+XqNswvz0>4%rzpK@h
z+b=S#GEH$MPn!nKUbe^m%pJu6Ri@|kp&95wqNhD~tbe!D@yR{SqJ*5vn+MM%D`m4@
zQOGs)aq|9xE6JL}ZoF?!Lw+E(g@Nyub7U#H8^14m-lQtey)v;n?J|$z?NtLoP`&6d
zeP`jZ(5m9BH}s0NMTT|2k-4VHrZvnb__gm&Fth4G?=~?T8nH#kSSX5<`u65OxTJMp
z*)&7G;QmkGlP?4sd`h?Y?(GuW>c=PQ`Yrz8{w>X&vkZ0iDDc<Rvt}3s37CG@1Iy_5
zcF1dA%f0Cz?L_7eCyp+{nIZ}7zveqhn_r<UsT#njJ&Nh-L1_bURvQc`D7_9c^>9%v
zyUr}zlqKxO{tcOgetn&PkefVL37n!L@S4LLJ4E8{=+(SiZkHkwMV=bsS}^=*B3!b9
z8TgvmO)kHlpME3yTna@XqWIeWr6ExUkW|2QLFkXcZ~>9Rf{d;w%|LffN~y7;3c#x9
z2Duse#K*OXiqFvd=4_`j@FK;^1eZgYehrMQ45^)4_wXakvt{N*UU8*{Qa=oS-8w0d
znz|vn;STQj({g~YGy-GowfzdC+F0&rI2p5B0!Up8oU?>6?yU}5MLBIyn8w|)eIv+i
zsyDboe_U7OYe(AyOUL-Pj6bMw>n0>k$+2NTU#cp|uhroB;EYJl$1e&<TZvQqDe-i<
zb{`*Ap^?$x%$X6*Bmvd7X@PH}6s_g6e+pJAG91PCieA^5t-j=}hL7?=ccZZ?k<gjz
zoiMmPg!%Fly?_>a)5Xt!7AEePyRxnTU`nH!LnuvFwZ1y)*ht#C1ud^6=&sNGP0ye8
zF$;gvc|B7!|5CB+d5zFxALKosTx|5+v7G~058*8tDmEI$4Qp59^6dApveRY{SBKv{
zZykGW8WZXZCY?E~D2z{-+SdDi#Ub3ZlUhgiL!XA~8*9C}ZZ|peyEsKRlm5WB-z&EB
zo6Q7Lu1t3)@SMB!#@V=dN?_Gcp1*ZNlqDd~BYNwhgzGg9g#?-VE6vx9%!I?a%h3<f
z<8Z2Vj#_j>MrRaSv>abqVbJd5?>Q9Kq^x~HgzUS`T@zCB$6YM_6P=%W+LNe&xA-{g
zDb=DBJF%VlRovHi^N)766iD@?NS!*0AkEK+8lC5=PU@9goH4RL5h$Q;*?M7tGoKbn
zv7_N;Gi&_9V2LP$zBAqSzXOGhd0B!MT=!aOm&mqJ1|L(}(CR~?(SJkEQpC?hTrllH
z)6ji{N}$+;8z;5eG@xT%O-y>7DD3<0Q^>DDSq!y(7Zll|qZYrBlcZ)lq~g!7KNDbv
z)VTdt4Pe+*+206MYaEoR+Eg{z^suuXFa1Nz^mQMxU${H?$;+_~+_+jaNw69gKXqV8
z&vItz%j%<Uy3MUg>0=K$h++GC(CQ$R%wcy~x?3okKQ5km#Cr?LgE$aqm#?e*Ynn^0
zGu;UkQG%W1pG0ViW=eEJUV;2AuT-9mMkCm%ULMc=yQN?h#06IXOi1cAH9mBAEw1p^
zo24nu4E0PwLN_wC^^0$!-{2>|ts*+<k9VC24Vv&esC~1hvHXeEMkbm2E*e{e(F=(#
zm_;U&grz+5ydB!?DGSy=%w&7VILD1JM-1f3X_vMZ=o2)oO`2)bjZ)8Uc_HNu#k9#Y
zziTWtTQM!Zv}h4V;8tg%plK&jZz6SfbeXga1_gV!sj$y5-ro&*AB);6f~GvI9Nspz
zA5eLHSj?nz+i-IZUzu}dXhp@I=%5r>cc;2-K=nr$C~3m6`!Lf5SSfv;#r`;d#!GaM
zOz=nuS#R04o-6q#xd`0~diAxoPhegozMb70AUJ~JzjJ&Vn7iKvwR|pfJ2CJ3EWFN`
zDV*Md5z|vD1v*m!*FEIa-anrApp4$PeN(8Qy$O+V`Ght_?WecTs*QKgj%3YxpDC_s
zcwf&K=r(>`oparYJOrejcO7w=nCaKBLwbf(Ag?y!kK^~c;c1`yu&vC`=a-YAKlk2H
zsa<i9VH4l!9|<~ZLOYnF<7n?C_FKKlU0&fKj5C#P+YJM8vZL`bkllC*3{S<v^-r-t
z9=hKr>4dWTj$<DU>RnrHm2yi}`PJfN(QY*2Y@nFl6SPeMBNZ+PJ%%>XDc<)Qku@bv
ztyb3v2+#>@z3jyC*^@W5KG&Uw%c}m3)SIjO;HlSL@bmd>uj^jck3ifwcI<Bmn<o$y
z#;q`cjY;IJb1HJS?~vCumDce|*NF+G1Er3)=l8mU<+HXr(@8O<A)0~LQN<&q^pe`C
zZw1~yNQF#ph3S14$Buql-@>wr2_A5Mrv3No?w-&aRSlO@`eMwJZ&5lHhMDmPA<eQ%
z8+G>J^!I>y5Kj|Jedp;cR7>6Lz{0Q}KFg*H_Mg5Rx@${7IP>?fmRB5a7%ca3TdRGc
z5G$tbco4&30R48bFxa=$OWfAhr_W!nT=^U5k2eb4!lKfh#{oe!g&Y&3r(5D_4~yL#
z(|D_GBYmr2JwVwdUxF+;hY=~fg;u0DXk068XrFzP9L__EXFn#m680{Lld7~@O)E4)
z&)@SF`cYMV%Jf)By1^lKGhTVL$(L~|NZOvJ^1Xef#{WRaZ0Nh4<jcJr=jZc9M*$e3
z+7@&^-1<2BX5k`}PaH*Ss(|;*3+)}-b<y88=fsHj-Z<DBWNy)@l^ylhH$>^6_T#!t
zr-ahrTe0|oyA4kl^yT)9@L8d6K6f=C-FR#ujs^((=9~J;06yfS(xj_Hx4eA!@*(>&
zh6(q`6+LWA6E8dtD%E7NyoO+(9p+=UCvL0qhjoL+R1FS;=gE|V^Rw<Lk7~YqGy9ax
zhAOhQFGB>dixv;Ly;x2|)v~&#Rq-QVJEn+Jn!Kr<5)ckzI4p;QhL|Nq?|()wY|nOX
zZPlD?y9J9A-H2IU1<!kp{*sEbsjj9u^)Zi4Nlg(v%pNza%Bp`53JtYg9&H0GT3&ip
zLC<SCt~FDq{n;Bs)57`1{mnhVY~=^3Di>De-g&{ya0nX^<<Cwp+4sjWlY0Ts*9e#c
zY<Ul$;Gri_6jX9C*RsUx(7RH@u8Rg=u@8bOayK-0dcfzpLr*9Rikz{GYdYzt_QIvW
z@m=Oui>o{BFJgw`S(3;G4^A5A))ecV{wccfRv~RJrhwjgeQ4vY*=rM(f|n78*M_Ma
zo5u-8Fdzj7_I{A7>;`K{!d^GL8s23`4SW0PX5M|%%6a_HfdOQdwID1gsFrB34J+Hl
z6sC|Z`!5oJjJ+?KXFsXt*c5M9xkxsp2hK`&n@UMPzyCAn=Z%PW-1bamsv;$W7P3xu
zdtBl~U$W9c<sFXZaY_3XA<^^MszkAvAN>mjcq?gkvnjjObPyt1vpiBTAuo$W>yZ^J
z37JEI*$|dJtCleRMwqU3tBGpdf;*02Slz*o9+z{y({agIg8bBYsuy@5QbNgD9T@2;
z%h2H26j7(0Z@1pr3hyA6MPxYOE+Br+fvQMWk$n6Gm<q;jr8zH51)Hg;qM{n^L^v@r
zegb$P>kgIFis1{=Z?jt<@+U>TcfvbfOLQ3mV=li90iy2gXt=+Lj|vJgt91U=`88Fw
z<6~3pJU?xJ$1Dqjm^jx+$IW`fW>(vZB2)sXv5zQLd&gy#x%ZQE&vAejH>mI@jJKqy
z&<;|J&fHti*ApIi<;vQ$6<Le%bB)VZ?b5Xee>}pqcyrneqr_*sl^Exp`>cjmtOqY^
zyWaNT+)c)H00|dviYP6x8SwUEH!S&=h&_3+wF6Ogwd;CK>Mtj?aU)1jI>H3?MS(Yv
z3N&+c#ACT|_yBYmWJU2H<151&FKrf=%xdcE&wYqgPTfO#unkxO@;C7f!!^<gZ#z9z
z=N_(|WzBY$X;SoMS*2_M#W1WHdW021)b4sWct;&s4|v0HiC^za7<diTX3b>01jX0}
zq9F$n+rTN6Y8<l)b6?%pt)Jwsoa9#YV#dl8%;uAV)U4?#G$0PM<X854x1+$D+*kEt
z`jgwEETzilGV9xcwsXdj7^G#Qz9g+UwbToJdKkC7>@#APV5<1=L#bZb-gS8YUm+KA
zb?#>#`ux<sL0y7;ZQuUpSwbFkB7M3MNYxeUSr;1K_udFzv7fj2(y?Fxk_NuY0NNk;
z7-h+wq$m9fkvXVb=_Tgt`0Jp7a(F6T*rA3x&lGYX8_?qz7qI+Jj+(vYGanQEbZZ^D
zQ=fOhAZOsd{T%RADD{TG#ElNL7=2k!ei|w0=ox}eM@pM(<`$v4oY42QVJ%nCr6>#9
zCqgSQr#eErGD4C!{g6qEl;Ny2?febzDT%LnyPAj<$F08jl{XLnKGxZod0m-sg2^fH
zH5_)EFr9%uOpl_;f3{yuLix_}v8+y3n7@1`{_yNc1JS0>Jevx(Q=)T(s;riJJReGw
zihUxaX;(cVG4pcRPa*ta=hUZD|AGwffHUw1)(1T(<0Spoq*ouSaF#C$oZ1Iw?wg+6
zRD_<`*3VeN3N|6{QOTc)q&Q6;`DBuLbtZW`hGYC%2E0rjnuO8D%AwT}q-x9da@%(M
z>LEu*c(2vqWG^Z<K{i$GeGepLz5~o_>N=1~Qm?!AeS*cjlcaGK$J=xBjoUqAb~Kf+
zg|UfRUZaEw{nO8kyK;IzYWICYm$93%RFLft1aY@W87Ju<@sziw`Tx8-tIIe1EkKW1
zFXNXe7lv4Uu7Tilxb7{Q(G)Q4(tRt-M<0y3<_l-AlxAApG`><~i+ySsS$i5;xr95`
z<{IMdH%PE2+$8|V<7k+d^u7K~QFFeI{^GgH)aZ$P{q@eV0P^`c9C`$JOs;aU2hdV^
z2x&fyKv08-igy_?AAXt~`x6vI`VoYKGxap%Wftw0srbvHpoG@#agGlq56>AH7eLVa
z!eJ%6Z{K)jW~*X-!@_)J%U?D~9?Q0Ad6oP+d);z8hJ2-@vo9h~h89-+R(bo7vTaU?
z%Gc6$gTEn<rKm{ysqMCe6bQOI%ciz>fyf#goVe3`>dQCGh7@GScnh4XBR&LWT?dvH
zGuS9MKK_GZh`{f~DUYfIn>~7s26ETo5>6;r`Xo8<2m1sqxi~F1DtT_EkO|SQ04_A3
z{bUUKA5em)xj?JjaSZ*Lvaoa4zTn)p#wVCd{iCb%Png^@q&*A$o;v$dJc~U06OEBb
z3Eo%Xu-q?fWt0ci{PS~qR{O5qlX-Q=A4k#WwcEpPFn~sAv%Q}G^acBFGkUae8(!M|
zrswy7A8kC%4}ZS=?B7{~ZVl?5Aw}blc=#=_<p-2E!_<A_R957<ZlPrW&z|y;A+ku?
zht_fWt-pT&xd}h$#jC2EDUtBJo4@B!=e=HTpcDG2NdR~x&uTr{APCXXG);b})ngx(
z3gN;dGBN?U-c_J&-GK!sWL~B(b&d3yh%YI2nDpGXax9N#>c)wv25dKLE&z4sB{&pl
zYsi()V5EaWZ3hc#8qeOcEW3RDuPZl$J@Y<CASmTGT0@*aDp+|*g?%D-oKDiX6F)x$
zCYbGXMu*%H+Gw2Ocjom#ztFc|^G!9UR!=`3IzQKliH1S-PcgNq%xw2a|4;U<W}bQ-
z@!j9X2@-wX9lp*lLGo7}qC6rHTiecU1<)o8YN@F*h;aRFFBS**T8-#%;}MwK(r;w-
z1fmR9{e`4)C*_Cp21g*9=Mv4K6Ao;1$ev7O>7rCpd-)F@pV65$k+<&B_{btZ@V(mf
zB?0M2!(#doraaEcBgx<I)HiR}lTCZSaPK1+YJB3e>>n7i{1Dya`0L3D5$M-{WK(NN
zYdrMOzaj7&n1u#N87LV_0BQFK%7H%c#~+cVGFRnt@Y%Y8txy%$gT=EEVN09#f2L%h
z!?y<i13A6z3ER-XjT0eEaGZ*Gj2<S4=*zMP?N^d!R>eravt8wX)<w8kL(eP=)PyF3
zf8&x%nNC@DqV6)RnvkMQvypPu1D?t`iTB+6g}z!|(|6WBjt9+;h+NPB9Hpl)fd!fk
zH*iZ<nq+^+h?Z}Jj=R@Vdvo)`D1&Q|z0eaOVsqi@?4#&WtPENc0q~pp$3|WmE3(8d
zdlrO;dS8m4^~P(VY^W~>&7g@W%i^!n4<<u3f?*efaSMSnfy<0PW5%h3eZa}`z%9?G
z9gpos4%=T!wu@T{8O=eD0QHu?A(40(^Xv$vBg&IIUTE_5(B}T9N}VTU9#KZZ^<k{X
znCW6-XS|-B3zaeZY`_9}wY7NZ)HzY~<6~XN5`<18hA~J1`bq1$&pF*Vt{{lRBKkBB
zV`;wasl9MB7bdP@8`oi0y24bNJy{s<7QJoJaL=w#TS)<#n9^oDiJ@9pjui_K{oupw
zBg~89;#1PKxjP>7N7?g%t6~>J4@9;@P(8DKL}{{Zo3|0!3QseRD;5^&P_iqzB@%Kf
zGW=9YG{#FH0na&1ng2i=jXe??cu>U$dCwzR!02MIC`*WroPj^llP?10_ySvYX!(9^
zu$#n;#KiBOKS~Z<s;oGa<XhD7Cfat?qsB14Hv3D+7v8pBP91GNq!kDmj6}g{s97H~
zoiD&JSLuw>v)g;l`AlWF`Bv=vRA*I#7B#m;#0oF_KEahF<ru`d6-(8xK|UDi;M@H1
z)i>jCBGd=5Yh0B>5qn@Z`Yp*}T{54SxbpGN%uBh|o5L$ERqv~lin4<(L`4py!_!GM
zT1?R;JN)<6YUq;(mpqHwD0~Rp)SaY<1XALI8#EjoL{A&@!6z!bg#vsnotl%FSCn!{
z$}^eTTk467+|l_yFk(NAqg<iAk(JuXzQAdo=)9=>$^P)?``p8TC7e2f@vsD=Yoy$o
zrC87#s!5QUZXB~<c~7p*UsGCiLg5!@Cztk-l<fqKQ9k+{;T0i<u%8K~;w~kaQ0ror
zs}m%KIb;MAa|0E&%UL(T45r5YSUDSu3`8Ey@c5N@$Z!eFV6s|LCYH#SEf=nyq@WxE
z-?dAlMjD1w9zI_=-P4ZlM7P6g&!sXPg5>7blw6AGF=k`gNpbfnD!DoUZ8)Y##EXE>
z`G(ygCrr02_)QRGik7s#B04&$WUS=eKiDe2@6L%nMm<gvTxHF?INStf-1O9WB>aBJ
z-y!%_LtWq3pGS!3d{8$R(l_72Jz(5S+fvUoQ-0Q`bUih(z<pnSCnI*YfKR9)%8njW
zU|C47avD*qhF1sYna{uYwX&DI>YL4eI$xK(C?4->Cx0i}Msgn>0F)ytf^3ZjF|LfV
z<?3z+J3>wQ-;iJtLx-n3^VEES_C8MfTPLb*T)Acn|BoMbOJYBF&f3p^b4-+R$;;Tp
zERz{ls0TFePUK!sA|X!m_r}D_EAn|Kn<kT=$VBFSZ^1PG5BA<Os;PEs8w~=Y0!F0Q
zsGy)Um8P_)ND~noMG&GQ(gZ|B1B8SkAT<gC3TRM7RD?)ZTIh&~bRt0@5R@(nB`ngO
z?`H3JpS_>)jdR{}#yH=<_lHA=#+Ajr=9=@Gb6)eBMmPPyzKG!lhqYUKn?;uWlaHq%
ztNjPqs{vFToN`%(K0O$CI`+Wa76F70j9hVTkRjfmqq;=o;}()`b_wymno1kU6#%Xm
zI({72UFL<_c?wr^#`LVtix6@V<}_m<qX`L{8&xIUb@9{s?()3omcsiKiO}q8r0jrX
zf;4DAP?RS|3nr2rg2B2sVZ2ta^wC2noz?@jKMG0k%@O2)R2e~*r|XldV#whkKHgXV
z2k6;4<n(SmDcF&tijwh9BjPGBSJlkE@q*f0fmd~5ih`571>(}$gS&WH&hcJH2S(VW
z?;XXYSxjc|LPT5V>MF7JWzPrALh~{Q2gYzj<=1~<o_r8MTQ)EcQDG2Z9_<Q!nkUgX
zV~gb*vL1`fmIQC-v6@5dbCjf=T*niPvaJ~#kS9&i&9we{vQB0x+XXb0F+%K<XrJG|
z8yn#O@(6fV_7|Sj&OA1*0wB+{5qW$yZz>+=*%d#hyU$Oxc~q%RM6x8$bvwX3GRPzt
z?u?}4#Q{xSSHlS&Yr@FI=<Z<~-4BxIEYDjL4Y<#EI+rk^koI_&%%v}TL*GiNzcLhl
z06T?T{6Ov^(H)R+sc0eQHNu$~i$o}Simf|;M8!|*ACTt)vjQ9pfIRjLse*B0wBKez
zPIIR}rL6c*I_8$M&ZCQ<8`XLrCBrj7Q*0L}@G`T$`@cB-@s~xywySSQ@r7aFz{fA}
zv8|_hatxy>tC&&O6E=4eFU!O~7GvIAMv+V<>8i0c@Z5&i(}pH*GV+fdQmn1HasbS&
z-cg3i!QwONP3{u`(lZuP->-lEbdp9^6>uCNlp;Bjxot{<1M_!@58uB~q0<ao$!gX9
zR|Tk_uUTI)q)|v$YP{)FO-M1_qbs<bUiU&;pB-#YUpo|8xPAu}gQ%#ScP!w@L%;67
zMys2X1~5i%;ik=IVz-~cN&<g~M*tS-NPqy;X9dNMyXrKDh)$rq4HJa}0}OFH4}PH)
zGYh{$DBf0cIKD=Qj=KQ%r(ui=yboO((PbLQJu&B37L*q^iBbRa|L=U?|C0*a5^$R(
z8@d$2c2c8<=s%_Xw~q(;vE!C!uU{62$N+hY;i&Od`Y*%wk{@7vwKDv%3rPc&S$Iem
zL!KFt;DbhjHp&(Z&ixJrr9H*z!^Yq4MIuE9enJMkE!n@9WNR#G3+c1RLw$V@+OYk{
zTCg9^9YYv$wsIqRVi^8UE(_94?V$xvhWGw{)k;$xuwRm<6JJR>=-JzYW{`vk%w&K6
z|0Vj%_wQ-zvjPSVw@|}1(g!UA`4e?-=XeCXHSwK-CxTo675NuTE=an6)lO84@UGH&
zWk|G&r}k8GrBH%zy>r420>zK;Rau?a(0jJ^X>K1pt1VyCnI{8G?+8Y#Pak*FCxg{$
z${h5%v#y7YICo6Rj*e$KLGE<q85iU>$l?3X{a;?p2rTo@)i&&@y$s8N9o+Br&-vXQ
z+#Hg7S1huu<u~CD%~!bK^9PIFzM=WC2-p0~V!9CX6co(XpQ-AMjVhaYRX&hy^7DRb
zd3`FsOw&~2J)$R&AT)x8z>!u%W(?npp9kjW^XP-cw}1W4?EKN(%(8rxODFE_!BG=l
z2$j_BoK@1Q>({Kj-neTCTKVF`ReBX}97^?^&!TrQ@0WUE6vuxQ=HC_`bw!wBlE;Tt
zu#)q-fQo4r>>2RZrM9+T>}NeqRN{-GQ^*dqo*s}n+9*Zr>uSI!$~9NDv_7C%6zpIB
z)%VLQ7s>3ja;85AGRx)Etm?c_gbd25Df?+s9S46pgV?)Gjei^$OQKs5qDG~wq~+)}
z4E5cDlE2Xx#bzZx6H74CR1^O3-?0I&q5G%Gdp}$`mY`U3>UG4(L&TvqJ<^5(qB#ug
z#Dsju32`h1mR($#udVa<>uToBr@rO<1q&)pL!N@jZ9otcto?A?U3C}=@m&-xxc-zw
z&c4p__ZJMKnZvMwV942gjNfTQy1(X9*gLcAhs{bTrDvijE6jE-klW@6RT;3fV|TJI
zUXA(O|MaK(i2agOUE-^7c?DmIf2`Ks^tPVAdt()oF_Ia>O~xu<`zYy;akS^AvrDv7
z(bS5Khv`%5ziYvGX+NQ~Yu=YW%yb_p53{UiN;_m2zP6TZy6kE^bWtVTK<Vhd+VB5<
znb*m*jCpBbN)H%W==_!{lW(lP)?#au^5Lf1i&u|t3$8CBFfrUzo)Y%k$JzbVu){6}
z{czUF((ipnB9WV2vIE`P{&`T6b`CxBL36Si<?2VY7gS`>TJAfQ!;Z@oJUJ`WhL9vt
z(-DGo%B~|m`i6dOvE?oq+A4#kpYf`_#?grAU<q(WxXo{J_NKOGlz<|27rYb{(v-$f
zufc!+g4jl)W}pP?<mRL(HB}c`qq#TlTaZ($%?arCT|B+jqO812UKQTn`9lrvK7@Pn
zUg^51R}t~$|GcwnP<2BH0<Y^`R*17HAGqluAXD_pk>8}%Qs}2+3juhON2IOB_PD^H
zoQ~k}+(RS3UMC$cQ2HcTb)#+pB&{aKpyUP=HvL}B9_*YDy~l2rdOs8zw0Z2-%GdI+
z=vUb9Jbrpt<M3E4#Ugh5^YQ8qhq4!Hrk7L_6BWDG;@0-n?u&=5i#LtYZBP+PG7P1x
z!XMLBktpxAbEC&x&OLeiSmtoHr@$s5$<>j6FX$2KK<)%tMHWGjr-A+ogxlnJsbTsB
znBm`O;$)KbckJ|$B6psmHXDiS7T@T3W|L^B8$HJmvbZLI!aWyRn|x4^P6o$k08~+9
zndtIV%nLcvXJydI%|fL6&Ks)nPc3faHIM{q<O4-0_agz_t07}yyN9gO%hJp>TUGPE
z6?yyezMFn!Y!9lL@6Og-I?SH(PfZo#erSw+kYmJsseAt;BQ};=edp^)qw%|brs=gM
zoZ+gE8vk--Gsq|{>|!&YJOJr>T~D8}`tB_J>GVPrKe$~5uyy->oa)0MVc^hD)Lz{_
zxQ?{<VKd`l6IdCH3_YNsbEl|!qqFFnw1A!SaD^*ho4EO{G$?Z!k0%tv8^<VS%??#l
ziDd-|&rgjQmuFID#`oUZyuGjMxEtFXhQ`mYvJqfY&(p<n8v;MY%~?NBCOPO+b?}q$
z<6lJ3#@m>)YB}(uP#>6Wm1c(1p&pLkE6evT`@dS{ccj7DT1O?9f~GiPv~dMWERb`3
z>q7a{_4s#n;7oN;*dq9XsrfOs2Bgk(?Fq6qYUuuHu9Sbk_AXhxEbG~ESmKK#`)YMs
zly>&ic)#v^<D2PG^v?Z2nwS48M}@@e*KdW*xy8dEcQ7JkMU3L9)_tz(sT&$+K1-Ol
zcDr~!geo<2SsxFq&Y5oeyc^`!GALA2^}gK!zZf%*R$0fv*Gq#%BLU;^`b$ZGdv-{n
zDoPC=^!Cu_o2(+v0!O`vr;S0-yV~IVBWSypdlqXJpLS=e+<vfQi-1_4tllsPcq$NB
zfF4b^tjF%{D_=m^^cSwGZx#4>DoEBv#H6${%<u0z$o_M^(*OMV|HIihE+?@4aqSci
zD2sob2Qur9D?zC5+6$lmf~op49C=Npf5AXgwGctKCTw-_!j&Xv6r7Ck9Qbo1<wo1+
zkg^(Aec`~mBX4WuOWZXQpF4?y=bg4j*`F2K&u|_z8jBfKi9GjovN+8t)Q|W(pn0&G
z<Nkj5<WYmA)?6NChdRp}dIl(RKo}FT`b(7;h3ec4$`O0z*t|v}Xm$eFZB8h`KCDyZ
zW?|)L#wDE<cm6Q#Nvvo{bt`qlv5}yLvXhSFM*95)Tj6tyIEDbsaU0!FtGFMFt=-o3
z2+v1|6D99JEl84d-C7PwaUUPveW1x6FAxS8B;Z9i$4!D)i#Lg49dw+qoCj8|x}Uzx
zYXmo;d<@ZR@@^kxPq2~D5^>6oIl&m{!jk~)UD3_;(2Y`3s@;h*Sf}St99JBDK{)tG
z!#~0u;CYd}Ceb`3#{AOu>q@)oeui;CV}aZSsdzSZAh%(^-}be>SrA+}5|%f0HT{5u
z8jKa#TQ@~~hNNb#rDvdu*P2!;VAi59+8JDK8_TipF3d5J_v6KukBqhXA=-i=vq$Th
z?z^dqQJr7k-kv(Lwz)R*(^z80*-IET#?3Y5pqh}I3%igbTR;!36A}F<uiM!3^-u25
z2jY0@bo(x#3it*UU8k1XO4!aZD(b)+oGnSd*Y-}5^W7&ppe6u)Wqe;hs-x|S#}VR;
z5YxLf)eaJ(8hKkEwtQXx!S3=x4*aX#-^thxzz;b#&b;S1Fj@=uEWi12?A$}1>U{xt
z<3|u8N(f7C#^5oz?Zi!6yVU}(VH>H}#|cwArgC+QX0nwC7+Tm951d^+*vhw<{2EL%
zql!ui8K<Llf4S9Lbn^5Tk#rj*p0KSB1_`$~Djgo|ZF21oc>Gey`i9d9n1$shVhYk<
z;5pI<_plCkq2SB~?#Ps6H%FKoX}Sm7?IsLH+uY}Souq%5y*R!C>Jmg8oq>-)t{C7Z
zH|O${4$)P4F^in=f%Q}{VfX~<`3q)Th(z*g5%d|mB*gO9zGcwy>lPXzmEB(;(p4v7
z>lfwFw~^wSoxet(8B#0;r?}XrQl)i0=p8lej32zUN^yRt%&*G2m}6hVR3ul|IEEK!
zmYk(BQ0Zf>>VbF{Kr=5~EvvJCqnzfHz0x2ANKn2t(i4IV#|JCVT^+q@Dy=J~Ow9cp
zchh5=Do4d2kMUah^Sj1|F1^-~ocR<~RQHb_hjA<uWPho_jLLtJ2n{kZH+<P%_3hsF
z#}t>$aaOqUd`?zZ5@{Zx&5QJ1{|UKoyNo$OU%9Pz*Z0R#KB*Jfct#^qv~3&_gO%&O
z2HRVkXa>8x_c-!hQQg~VGyk~!h`xCSAU*FPo)v1r0nJgR%V-1U<<F%NU*CMbr#oX2
z^6vZ(uKJP<(~z!#5JP%(3C=hkTxZUODSms*^ssbjU=sTRAkCAv4d&Oz_nn?c?bwfF
zGj<^gF}r~=G}V+91LYn4wHz4=<?iLbV3>$XFt|SE8;UNxk}Tuy^fKwhYp;V6+rzh|
zY{)LRXhRx~;eN(O>|-;TBhjy%+unPB@fUGdy55kTZVDtee0`T(!0vRM4|uzyGzKh(
zB~K@gFn82r%lL$VowkV&K?SX*;zC?GhTyUmE>1#?7sPV;JMk8K4%Z-icl|ZQEgX%+
zJY9L`u3TY#0aZXMQ{_BHfwbp`K8_2v=`OpBCrR0fp%c#Z?U7BH4SI;(M|!guV=DKN
zj%!+kd&~K~zeI6yYAg_Ohbe}qLfM2?QZ{_(58AIzAFb~aExKv;P=?SOb(V9_5B=0@
zj?o8qBWz!VYi(@v1IB`DIE|5Xr=lo}E%%%~-cv5wD%SSC7unQnA=|}WvBk!K%_B8-
z+;?o<S-I;a&$Kkk<KiQO#nMj0yYv{I97zgmFo5!Ec64{b)0=)A+T7EAd~aO}NZJlG
z9iihA_iBxK>iF3&RI7Kw!a$NB2<kJb7Z?O?r~xl>00?qu;@UBEPwoSd{dGrUskjC+
z++;roP?Oy{H?*a(2y91@GPFi*JR`(>fMV?cBcbOv-egmV3tW0Y3yBHO09wC#oFT}O
zXDMpHs{>TQ6T}%+=7VQ`JYneb&vEkyHGjeC0sD0nlyLsWehs7>aN;&J3Bj-m5MhHZ
z@V2ef<u5`}bfOg_Ce~~10))L0I+<j1^g&VgF@<b#R1hT3J4t|pr(-$>wr>~uj5!U@
z-n--M+L$5Lx8Hit<J%wd-|h)VX4-`VZ(VbVZ3z8=OF$ND^BM2!F7}kfSie${Isq>-
zB+rl<fn)(Zy8MoSk}<udHT<Gr%4|#K<k2<_3AMJ~sxdr3Mg$ngyq{Q(7epC62r_!3
z<@c~N^AK<Lz@!MdaZDO4^&rXB$yqCVXP$??em`jQa|V4Dwpji}myf-bPGr_Vc8vH=
zEn_52nD3LvRZ^N?o@vA7sLWNX74BiyaZLJP27K3-&rSmg@U?34vS0xe(5$q1P)_kg
z<Fa+X9|~D>A~De@%54~|vs#53=#EYmpUN1vzsvJ2qYZBqs&{T49l!VjP{_^duUrw?
z=FA8fvSVg;lYF#e^k3}lZ5tw{t~m~BV9_41>ooRjpzZ{kKi#}u)Y9b|4=`3IOIp!E
zRpUO|aWX5xam_U%xiH6Z)eW}r-i}rjQ!ESYlhk8ZvIRf{E<nS1HOjzB1Yu!9|5b7j
z`bMCi*Iw5lAaG<g+=>r*0zV%MqN@RYjMsGVFPJm82w37{TzfXe=6yB;vEG3MGMVc)
zwz5xij$FI7xfs5z%g0q71{7L6w}<ooRPtCB`XO(cXTKK~oWGz67R7_*Sp2+tGJ?fN
zoRSBUedl#iY~hVdrIT`T6o2zrW+?Pv{#HTm<rExsGbtM<^fBNeMS1)2fh)?qzhIg$
zoUN%4usbB6a1fmUf)GQg0ZecSIoT70U~e79sS*onfI;oSCIO=dkYIXG-8pdACAjui
zvq+|R1<&pn0*@#mMYknlO(BERUd=BTQ*2**{&@}r@bd)uF2W9)2;z)EiRB8A*rp#;
zR{joxjCncE>@KW)d2wO&B@uM<t`i<QY`>*aY9ZA9;q6#&T_h0WDkDw)-~@wjMY0XM
zg{a7R_N_srLXG%g?sU1A6-SPg_gv0QAAfDC;J7@Z#T=)L<VQ}HqmjIgpgSGJ)tp8y
z@%NGDD1fL6w#yrEq#RqNo$(8zue~<irUWHF7Maj$bnq159dpdXHHdSBJVS$McxjWW
z@`c2>s|u61-_Lx)w@EIqaU_F0QIsl>^d9YeV2&L52@#7()Ht2`v8}v476}y|gS!qR
z7H^109Yee>10{(-l4RWu$dTq416FLH^4=8W#r0B8G{4&(d*Jrb4`VQ17Xe-;^V9s!
zc$bLN%-T5-%Ui|#dl*qMt>z(MiPSm)IHI?V=5&y?;vYO2*V1au`C)P|_LwN<C<qs8
z49`doKihIfNtrP+ducjdfBB;)9_yjLgte%910Kx~fRglQXwfzf$V)swAs2%Sx$5J@
zp+#mO1G%6)VUq-JY!wd^!Ct)-`c@#7byjGB%rSPS+Hpn4*7@E75#~e;=pnzX{tLEC
z4|^T3-||d#ab4<mUBE;v|1**U#S(8-j)WamuTl3ry6o@~N<+>~P!xh<+2?#hHdg{d
z^wb4<dE-wAyL=%%bid?1-a@jaV!<>Xx|`pMU;pL34zx9=!8e}5c!A}+4g~O{yM-f9
z0ML|HcR00~RN^0Tb-(t$l%?af$<~`KudZLj{hbL2*g&A=S~?aFepI_?!{msGE)Cbt
z+a=A#E5gg3>x0@HNnf6=D(G7lbpxwt$$__y4MjEOwtzquu6sMH<u_5p#SW9ndvbT~
zoO?cHc>AN~VthRxSAG3AI7@7U!dUrK)A?DMHwE%qp$e<S>6q|sh2W6dqT9H4fcl|N
zMJ{lq(umbgQW^{_s+A;7zJr>0O$G#($s9|G3*hu5nJ0_;u^o8i<HV`hiAqiyQU>?q
z2EbI7wK%ge@KSLy!6|FA`M{e#t;;tr*$Smnry@9)OGO<UGfjrSS*JLZ;J;l9Ng2NF
zTSAOM1`r{&*ND;t)d!rNetIbRFV*=vqKr!~r=2z(wmTvV7B^j+X5EjelSPO0S(oV>
zN~QT^?T(wYN^@lF`HYvs_Kb;eB*6eMR=?+>wiC53-VyL)?Bj@2i*1N9&lW<G+<c$U
z3$?+ZqyF;0N+UfF?Ls!bg>R~?5oe2WZb8{IL3xn$g-?}FBFU)|_OgF&1KpEBlpJsz
zj&%ZP$M2D3$9_GR$r;7k#ArANE<@}X8zwW~_u_p1g2l<1*5=N@^IW_KZM{w>+Sq1{
ze4-zkaa*9T@wRj@gx#F{Ys@FYw9R6VrVPBgm!TA%7*8h-H!<ydH)M_FA>Ss7Tb}+K
zGApp^Q~HHR3Z_4``S~UIR*9S)ObbnhgE1$^>7&vV2+-TO_5oGCOS!x)p*+$=CA$Lg
z-G}SA!H8R*1P?a`*FLe)-X=B!M!YrZJ(6yXTGt7a=QU=40Za0)`y2@!o}D-ltjDVI
z;!klm{N5?x>$9fEm=Yjpcc+J&K7-mppf7Dftd4``)9?Fk(M85~Z=0j;)K13d_GjpE
z5B;6f=2XKuxr(3=*bd>yT<syxlXlyA-$~m)oZlOl+r1zXVR88PUJbiC-~PEINH2m;
zXoQKt=L>LLDfN;Z5m1P6PFRG`-s@buQJ$^?y+jO>G7(8t=e>}!#;*CT2h*!w_aufM
zCC+4kg43fgGWT}Cr0@EBE-=wa5_%a}C4wcJ10Tiz@Mo06haAUO7S}SHr|6HPL$x1*
z&!7g7bAu=_E%1=nfLroLZ70pOkXGEumFNVXIrjpu7R1Gkp)4Jo3~90eP2iCxc|lxZ
zz$RV_Fa>oY;$GrZK24gF769THnX@5^88D>Evz=<dtxlHl&8@D3%9r3XXkQgM3S-3B
z(<_y-Y#W$=c}ts>%bi7}SqnCu{DOmnuEJaGqUsLN9~9@BzH?U*)7gIR%XsX|jh7eN
zi)<x84!E-Ga(jP$sqDcV7e0Pe754Q#hR8va|K#Bg-f9n}6DHyf7L{x1yh6gC;)5OB
zd%SY)Zj2c87~T@_04vQ1F%*06&2Y7n;I?`QMgSN11wnTtCPmec321N>-Gi86DAe;y
zPn4Vo?rTITO~ykmysZKB(;NqG?tHmX_0>DC(7~buy(N?W9#^G##s^|K5;U3_$C>*^
z&)3RR+R!f6@`uLj3y&^F-)Q3q>#?MP<4%Y6m@07iaN?tsEI;4o3p&=Ht~}jAP~+Ji
zL)_U|D!74p87R<iYVC#GnHT#tG_QLVm8%b<*WR^#QSe7R)Pt7Uhkv5|`)F-Zo{XUS
zGWw+B%f)^BYxW3U+itWR&$R&-`pwb^BhfvC?AM%FWc|79I62iodO#tyE<QB$j9QUO
z^r1vaM0gnQ<WpS%+;<{;!+)F9?B5Ah|1#f(yfp%ALH@tb2xlt)9XSe?<ayh~XmBBp
z1vj^xC<*DDm5X`)Ud4a@-AAE|_Qs(TwQ!!|enu`Y*cprCh9Jj#K^Ju@Eq@BwN)EgL
zNlmCTB%Fgl7jE03Z*qQM4#4x}BPtN4UqeCrE$e}EOh%R7$o0#!Aj_r4X)`}aF^b%7
zoV@xGeqOF-k+&60)j)I1YyBK=?Sy9a&F@6E`WNGZFC_r4jP4d%0zpMv35^FAOezfG
zZkn}KckA;oLItSRQw!it0**6;1|S))DOo-2^aCV_1?C!F1*V1Wy94y9E1yW?bBJNw
z*GOsxPSn)EC&QVUEZ-Qy8Pa6#+DWlAxzw$7+s-(ITt6mFqNWgJOc5F>P_Q2-5I$<(
z$CuQ|+A#@=5$_0FasK!ylx!OABgZws7i53w&J~*Sri7yk<nkt^#}fj{%)Q0^sgTwV
zm0Y^N@A@M~Y!gb7b95lHso7AG`8Z6r-8xrU#xxf{Sj#_-T8Lu!Qh6`E^<!y@)`gm1
zlx?e?j}8_8coT9~Xa)PQ`0d5B?+pcIBxcurnbO<}-7!j6!?Jm>L`Xp5O7mu33;gR`
z{qos+=g%A(tjh!ShX7E2;5;V-S1?C`b2CY1NGcw;<Yxf<T>zKQce#tV4PXjYFj&K4
zMgb-Xv3m?sp5kQLu5qK~!E~oHj9o$k3qt<JIO5t6JM19MN=_tJ2|VfkF_u=iaV2jH
z3E=3pW*l=fz=LL|xX+2L+8bRqPR5uRNMPgu98qW93r7Oh6g3~Y6vR{h3x;eSV1iPM
zdy|+qX)DINKvL9%nY&#f`NaE!^&2qqKR?I<E=T~I<jA~RyX#8%bk}4*-&&Y*Sf7p6
z*j<pf!WoghMu_+q8Zo;cxSpTHw>s#to525=p;w_Dndj+Gtl}e=kQL8!-vwPD^pdt>
z^>OVkjQiXNe~>Hgjw^iZow)WQj4?>gE+Sd59QbXbCikuG({f@s)}ZpK$L43og!q?x
zYERY7F~N<c$kPaHfy*GgzX23ma>H$1H?1C5nF8T2A1&F9r2M~N@a$<rP*=<M+Nk28
zI@ZAPPTuSR58{tDSHy9ZfBY}lI&6ltx**L9SKCxkOYP)uIkma;l{c$sGhc1FwG~TB
zNHS?08C@p|c?O2Q^$O98iCN=cTRxGnL-}ZfabPA5>0?K^(}XHnY|0{t`KH7XUpv?$
z&p0$^91aZ?iS%Vy(m=%Hr%AWjTCMYT5zIcSJm_0-M7v-8aZun7NZP<678Kne_G*f)
zWeZC&=Xgqq&%qc@<&7GhHYEK#KAgw*c%qJRy7D%W?=IVcpgLO)<&6h~lcekPf4*i}
zMz0dZPX7hFXQ@DF`9geU3Jj(}8G8uoarg81P=om75tAblja!WL;;-gb0gZ+i!rYfU
zsZJ_Mbeeov{pW{{u-kGbp9F5b4*Fep6F55DPm{`pbJQubsNaWq*J`_%ufSjWdXJTA
z*FVF!;3gafR$qT|Ix!siDEpnsmacPdVr+g8y?fC{ylOcZ@8R6U7^EJ+$+R8gMi-J&
z$rO7=07tTtXS;8a*EqN_%?38HYy~!<8b-yh0W{d-hw2Ukd!y-R5bH*Z0yvTZFTsZw
z1*YBzkUA>q4V^S!5aj8tZXWb@R-c}NoO<DMi*pAB5MZguD46=288^a9|4cx_&4Paa
z{B{A`o%6V@e;d~v+;*qsQE2~5kYc=*{vtp3LVYQUTG?pgPRnn$)oIA&15P{QDl8+~
z=wdM!jP{QJ92A*=*^(q_negdc1Un`l*VJk%0Alk92>WuR!?^!+t-1f{bYWb;eg7N4
zKaK;u5m)NsLsXeV<pgoF4}u~kS22|LXkWd8nj0ZH1o=0}51Lp1&IXPl9B~cOm>|<a
zdS#1}^u2we`=-w%yohnQhD!?tF_B2MT>}f~Te-TQGUqOKzcx%O|KR)i_m#X;c#QaE
zX^yxK)0)vW-*BH+I<XqPdmz2vWefg%)*5(q!{A@*=9QHNu7iCk{8Tss!c*6h3XPAC
zvp?$=-nY$|467;N|N8OI;iHl)PkJm**xXM=$^p3pbD-N=?m$<8(-+h?|Ak+Iv}tuP
zk+BHv?P@|LW#)aa&K07+s6Dv1u5}ZoTIs_V>7!lwg%~&^l($u#5j#Gef`acQCGS><
z?&Gr*{-cl#YO_Y;+6$d@U)o_J6OmB3a2>hSVc(ML*|*(NZ{eKfYjni5+9Zwgy@=UA
z694qtc;K~hk{H7cuPwLt&*Lz$e|v4=amR&N;I)}Ay-okaYfn~S?p^OX1>@l#OX+QU
zFKKx!chhr}7Ox3eQIx@UMjPJ!9B*ROYUsSb|K^21X-76)RD6=n7_58sU)u}R{^xg4
zciZ$ozy8mk0Vx|be1~_2JVE5fdh)K3a^Ld;aeiIvMyPlaQ<1P9592-t{^=^NJq;3?
z0aTy7osI)~OO+gPVcuzCJ<^8ARZFRRm$Th-!jl7H{|+4a34tYw)0J!EJn>y*ZzJ)N
z^|;y`5Gzt}6L8iN_{?$Q?<i26N=XU>g*lL{_OybbY19yy&$G&i0Mqi~0iYbaZDrds
z*ktg({UvN2NL{FdVi}$mum%uDPb6t4@;Yw)s9h|9DT!Z~<5q(E1r>z$Txgj4g|u>o
z*i0Bg@Xk1nPjRE&c^2Tid{NM9EVqSc&shUSU{m>adKdF8!!NhFrzlsI`6}gz=^dTx
zjw{N(8(px!V4p}|>vn*P=3~E*adH^`02)r15?5oB^Ugqu@#cDURJa7fezJQrf+IFe
z+xL6|`G7)<`2FePU|AMk-w3lb$i4`EqPk@Dl@p{)R5(r%V-E!TPE|6Hu_J}V98=qY
z5AAq;<H(+mCWemlu?;xrfMdg0oZNN-_w##XG{;<R*z|dzcuB$CFONiDBYMM4zY?>7
z^5j@TCaMuNgl}0!b(Bmwm8II0+9MBTUc9p1M9`B1>MoH<7*QzHcYhVbd35A~ret|G
zeD~wx`Y!_TR?;rqP2#XwGiazJ<YxZXo+pa@NU4x>uDF-Tol;&jAl;k@Hx3vNh*>oD
zU??|**k{`Y1_r%7R(<h0?VD>GT~?2)16dwN38Otv_li{(rd6GgC-Y0Ln*h}HoQn?<
zb#I<w+-YW>r+D8J%<av2zJ+b`$UZyjM_|+_$xV%);mH%;4=^JQgk+Fy^B$xD$)S?~
z770L$Jh1-VqUQd8^MP>Wxea0)2Oy8gw;UtfCTu_J$&7qhhC_KW_Xg!&#=Wo(!nt_X
zeLz@Zgr2b#ZyAFV{ec8wz}~O3Qn!j}?H9Mz<LZ+(6!HWQa8rwrUc?9O@4a>uKm|>|
zimt?6++2Mc=B_z^*c^KRqX)%J;KE5eL4!(cx=TN!XvTmq@%+2<w~njLhD0SF2fc!?
z@sKAK!I6v_4`@bgcebRtq|@<n=wzE4Nx$OHAJHn3<~IqX)Z$x^mH7?YLNk_U>lkU(
zC&bY%Xo{n0z1rL{rDu%XiSJ2t%kjEWn&|Cwis=O5Bd3=&p6)#xn(f$z1xa~i3r3y1
zs4-tI3P)AcRHf}UFE78e`S&iqHLfkjh#iU@%K!pOk}FsRI_JFi3!&0fbX-^0{ax%5
z1AdvQ_)&0$E4+5O%9eomOdMYCIZ{ca=W=^i^u+At9WU>2`#{;VQvd#mrC!<Q+b31d
zj;mZg*mccFF7uxoj`+XeeC^f^2w=~YDT*sVUM7Lvlc$ie$T=o*3e_XD06_1)RayXB
zr~|yhnY{Zv&<1I@q=SxP8)urIbqkd3CWQe&CfDFtTM7m&%>cyF+NcvU0Gi%t$$bIk
z&WrGvMH+bb9Z{c?bu==iIj1Avwyz&|LoU7tNqH7;k0qT*3wTgtd27@)uaJZCz7|-q
zhZf8c#96=C)qN9l;pOOzT()q~)eR+OjUyf=<~U<gh2wlR5&LaSN>B6flB|}l`sd6r
zqw`K*l25unuWJGGNuPk4s&7*jHzU4G5$V)-Ls~9uWoux7+Vph2B=T{g9LJ9vjilzs
zjw@wcrJ~|a*bs#4jjL@P_G<3U<ZDcWr7O56r?C*Pd81<jdbMJ!E-1O-EOS}Y>o);r
z#5~S~v9^=8A*ls8>9Vo7v9hv3!;`$Aq6nX=yKTl{FI_8X%aW@B8wV)pL$Avk4^4uP
zOLTsiU3*+}c((6EPn#E9$mipFoB%aBZ9RY*Ab>8Qk(;ypR8LiW8q{03Hh0*H52>S2
zT76G3SeK7w3*JL}=>Bb8z=T9M-t~HUDdoFplhT#_DV86VC0XKsZxDK5zack)j}DQk
z%vkE2s(7cL)LcQ~r^8>YRS0hao&V7&+)%mwSEJB=eB@dH+&A6u2~4)|#-)phv%akB
zkPgicIicD(CIUsBeq!r=-(74wvPME3ULS?}_JVLV@V%S4x3{=m;C{;EoObtV^#yQ2
zpmWI=20DJ04RuF-<2XAPTYqV4M%_Q~>HE>ivqFDFEdv83KkT&qx7OHg-v&RVe^=!F
zIs|VB-&#Ss`a`8dS*QWjjAe-rasF`VKuj8_I*g*##}&HnY|a%n3oWt{x3J%CW-0ek
z0KtrgVXm@+FwVu$0Q(G@)SP*5v3Nc1Ms)SM`mf1QOgZZrI6FLa2N{LDfm%?Xa~B;g
zerp{h|FLy?+nYucyxzJVTMMj)*~NW>)%31Y>dQ&Dad>L?)~zJvaq||>pCblh*^Vo=
zROIF#7-Ra4r13&ceQ%cihkK7iaN82M!5GA0vaKOuzJS)f8S~uq5=li5LeTBb^V*@$
zRCe})@cZFS*pj!P$hC)anXZ77FqMG_jkhbltR|WIf`~UV`uYYbb4&q3or7E$(hfbI
z)_%k?zvB}YkF5emzT6|x!of@_&OwAvXxGw~JYwhAkPvTMA+@@TZ>CNW%DcZWf#`Iy
zV(x9Q-vPHPVYD(#{@e^ucA+6Q3Nv`2nUc%x6HTG|KU)I){FF<xZyixMD|DP4f?gUM
zpPH|vxt{Sv=@kyjywZ$${A}Qs5VoAN4w+Y?OKnUmt#FK47Li%TC2~C3LFi#SIw_{?
zU7@jw&yq=c>edmJZ_>Y0P8@n1{T+U4J>Xd1@7|igtKPS&Rbv*!`W2{=Kw{wL-y5at
z9n6guZyeBXddBepHFC3j-#c@!hWsD<EgXN4yHnzih6&t<Wl%r1qypi14PY3TrRlpo
z$J@5b)|quXN-uiGP;COkIBs6E*9MG>pGp0IKZzLl*x7wns1M9#z!!pFVFyPQ$>p~D
zT5WY_4190WJDc=>{ge?|A?8?#9JDaU8IBcqEx3Q+eMRn+zElxcF1m9%*g58T^(J!_
zkJwz8A$cCx1cnqV8xRjO{mPq_)Lczx&6PX~Cr>Ad-#>2!<9hK^Fky5s+$*75(Z@MA
zUajcX1FE}Mc<w^0#1>GfX`3$`#}DOC#KdVoupH7w?HF#qmFaKoaz63By|v1tJ1=}!
z)dBVOEIty57tAFjkfhpzbbam}_m#KmHMnw6ol(l{=N1IPDIF6zgOs5UWhN$a!#1Vv
zpJ|DYoKk<L=zyKN{V2LQa)+BU=M-Zw{f)K8AB`(F+}<ZDt}{Ea4USV8oSo9lz<#V8
z2RVodD|Bi1#7&r6?Jj?It8vTsUpT+vf))LZW;4ZEI+m87nATr_^ZI%VH>$@jWb;>n
zY?*4Cf224OQPhA+L{^exN#oMo_Z+{gr8iWq4<&|8igJDDNIyb(+pbcH50G0lz-U~k
zOTnd;wX33&+j`a8!eJC4ApkeMUe$C^;cY;S6Wj9wM7#%CjZ0R{O?DdV@%B+6JIsMB
z^9EHIGXp&3=DzJj_FnX#A0p|I`kdDnxa#yQW^jLoC0NK=-4-a<^sc!ju3mw>;>Yl;
zA0+L>D$MZ|sx8pOLY1Tw*^2BlV`NFr02<UTGKr02N<O0s{ZYd{RknR*yJgSI+JR2v
z%3|(U)!v)eg?3Xdx(3WI6TicUq<Py)skN5%EHDzh-&H@B=+s;>hkaDqa`*kg*U^am
zu<f<JqnxYY{?s7&xTxMAR13k6aAUe~q44#MM@(6;vV9sC5TH1OhBNIME0woD9QKjB
zTOl<JqTR!L5sMAPsj+(0k_q}guhC(QrRO_KpfoXS85s)$v_gYMH`(6b6?E2j!Q_<K
zP@8?X_TGcJ6R~NaBpQ8peFYyEVv^Jwd|$t+=HRJk-^rIEKXMuXI?1XB1Awa2W2<ur
zUZnaj1W;q-WNPIC1a1^Cxw?8E0Pnvm0U_6QUH_#7v{VfSyg}{6j~c*QP~9k$#1yRm
z0<UEV2=Bvad0w?O*gHody?hUIW@r!F@NDsIgjEp3OAAB&&>0J$QLwL#*2Mx*eh83*
zV0D{u?LKrIHyRvL_Z-Wm8#gJGk0X}>P(b{hO7-NV6NNx*a6oXklk#eXc*jESb3-iw
zqyTsi<RFoq806<(p1zl<rqeSGJyAm=lVR+I1JdqYMWTMV2AMhyZx%e+*pBM^4wLoV
z)pTU*?M=mrg2M#25Jyl{0RvxOk=~@p`<eEc?-<)NotIB*hKqb^I3GsVgN}{^epN0N
z(ahF#?!3{90ZOVuV&9_AOa@ea4SCmgW`&P?v<D@^14A(OUC|)nsX62>okWyxe6fGx
zXdy*4WxvxcS0@oz2(3H>5S6GoC;^Oco^^2W1mtPz291{|z4kPT-4wouOG0`nnlbM{
z{Gk&jJIAk*uUEwk4xlxcC#4$Zj}Q}(?kGqtqg6Wz2w><@eIboiHSLk;mY>fq$Q_W<
ztZxe#th3+9b#8V#GRqkKIxPZKj?rb^Jx<=gPp1ncR?}nwB|2FLbqm*^0($b#Z5o_|
z##-#ehqFSH&PXQMls-Pl$eHTw%H0gK70r0UCUmB)&A4L>#zn+Wqr#xTQ>*3smm$|m
z#7%M|0A;p|37I$$WGPuqRmsJ-{Z^`f{OTLa-XH9EO&rSPeuHRq!Wn3cUfFCa(qHk~
zu&|)Mw8Z~lvBHrxNzN?B2Qr!n7+RZf$Q^$kGUzKmC)?rj^Gd`X{wI%JW}@26fP~c^
z!!Iu%N2!a%TzKR?ux`%=aHGMotWdlu1Q*5AlGRMLi?~|J5U}c%wuc?~62DayUz^GH
z?_&Of)Wd)sd$-VB>`GQU_KO0z^}>sPq)1D-yfP&?+fu^Eb)@P*L<(4_Y}O2Z6i0lF
z+YTtAa9X;`E$_ZZzN<+U%Bi(Pqv4E!c4u!*1{2JXM!xr=5TgUYj9_xGR(`FS;P=ma
zB*b2h57@^Z1>3^Av_WP{HHZWIR~F->Ol9fZnp7L(h$5nl$F9>$^C14pLeh*<Ga!<$
z_ygV5lun7R!(*JxxVf-+3rdT1y=r3Aci$->FG}WQo~ix|=K2z(vI!_Vjt*VvIlZku
zL_0?Jt<-4OnF=R+@i+6BHpWdtYuy!aV}-c(*K=d56a-hMjBot{z$y^nIaNb<z^i=Z
zFW7UOSviM+YcNGl{^2Qrm*@;Y=G_Z0Nx>yqjVW{48z*3#10ibs>ygmMJktZ{X0b@U
zf_mH{o4j(AtyI;{*5VxUFnTiWFe%1czzSgdMiBub94uoXYrid}yt%=uZR~Lti#&aQ
zJ_<NED>}T16ltnR4@fN3Dx2%v)T41VQ^Nb?bZ!ha@E9TwA!cl*`k+Q|LeXDA_AYyC
zQL5?N=vdHO#Z9QjX&VQD4)LJ>Yww~}%@f`oMmul+Ib)hlUb#x2n=p-FpXI;}nGe3*
z4GS3z4HDlkLJB#&#YBC54qB%#+C%3UxaLvBfX3^-sfxO^mku%`V)uT~ek1}01d=0O
z0M<$p8VM0t1%|q}{xVeRR^pN{*PaooWydP-0>dJ~5{DYA-ILr==3x4q0?ri&%*~=k
z644|oU9V(H;x!H4eVE?cJ8OQ!=53pW#%77hwH0D{0MZ8@Jt%jpHlsBJ^~l0vI9hX^
zKNGfU$T<M6@sVR*%BWm88kp~FRq}ij{NvZJw+(j6gN@5UYzJI}Ge&is1RtFyY|Eab
ztbFa(&(pa1h@uo7es!5AxxRNo9Vs^3VGy^P6*?n+_UY$-t;fYiFK)}$tpR!wSds)R
z?x8fF9xnbMffmylOYOxaVLuv-=&b`^DR7}bw-Sn=C0IOmuu1Sb_0qUvWRmJQM+@r%
zldFK&@!vGRN$Nnt^1qRTS?OxETi^Ra4YY>0BGYhs!20(};C=uT^dYb0Gg5ihp7F@J
zM!R3A&TF`Ka|BZYxQjBND!GxPr`V0d*@D9$%`0|I3nXm@;U*O_<2Bp^cum96<CtW*
z_PWprj=Ya&LCPdjkUX0;JaS$q5O7yAd1*oa_x}jvj&UQw)q3j`apVNtbQ>{%u|sp5
zdwRQdImm^N=}jO$qBxn3O+*$@r<7S;yuTwMA^4h!%K3bT4CvpO(RBT!TL==l?;PYF
zghv5Vhz--?@pbNVj?~2%#5Ff_tO%A&G2keZ%BGr=jQqz+eRcLB53r}0pw<!j3KAYC
zHH67Gy6II^m-gR%N;wp?4d^lbw5k_w1GsrMOtKyWhk9}_7`kX>LXDoXP2Nwd9`z%;
z{#?F(h3gB%i5tg&7ibWtH!&Y(xRq77_(?YhZe19#5PBN_3I*EWy})IiAyIA&Hxi2R
zAeSune^7}PyX2qdTDMUB;3VfMh?@=e8^o{1Uhea$!FfO1@1MtK9MA_3pEQrn0*%&q
z&KY`Jvt~e}9OB_h(&6c*t3*m+9>?`{(&ZzFrxJYK7(IH@R2w2$&NSw9<>5I>Nt;Hl
zN_6niYoRbvoVDo|dMzmuqxi~8>V3iHXP%<3{A2|72zJrpIlu%W4FCaH0&`}Nx7)S8
z)SIQU@2lRUTdI#XJ19~4Q6;+n8eiaJEOI2x#wU4(+$ZI6H?MfFiW<%er)|@DH~^qa
z4zZOmlrV7;xdm&2)@~sS4vHRLzkWMPRU{A2i$?XT#v2=XPb`YgW@@swYOD%-YkD}y
zuDhN}UHwgZ0UzdPx(||WXeT%)#|HWB?*9Eazfb!_O3DeR1^Fs}*xk8fh@O4XADp2m
zi<15i{>oKlRdF84UQZ8yR4$gplH`zAk<k7QjC{}CMLvv0vJFYt1(RO&;aFKkT=t`{
zeA?}LtV82K<)?|75m(kqDrOWGSDs7put*~9gLMm$$Sp{DB=sEv&fAJ{9x#@E>rJL)
zFntf$-M{vd9C-G}h3f1Zp8_aE-nIs6*H#GRT9%OOt)NBm%N9L*mn|IR^^q+x2;}%2
zDpg66PK>WJ!dS$f?z>93#E>ezlM;OGQdYV<P@mfd<3%HHj8PIAN#%%`KjmFphkOPH
zrFx&b*`E4wQnu~v>qA14U10flMOy^P!>CrjuO&UCrL*<r*7T7w_Bj>I&V%2at}t8;
z<_qZh!%Q@!nu9i#u&;_aWHNJlhrgobvoCXB{xN@^jr|@H+lC8Y2+W~qw$|o!GAHe2
zQ=HPAblsObVSc=JQY8sUvJj<wo%f~XcfGN_k)X1PdiBlaoo%uvhfo(Cm#yj2G|xDo
zpq(Ixi!Tf7K{=heT2=PQ{aa$TwX9+sERjj1Xg>gZLkLkwaa|YhfYliK?~nDtl%U>2
zOI@W@{$q&K=u&X#vBrUKI`ZeL^j_q7Yr_?K_z7)=SiszAIFEEk2w)`pbn&u~RIADM
z!9Nc1i>H&~@JKD$!`73jk5MmQsBx?7wnC;~Z3@j=@BKbP4wjnkEoKQWqEb=xvjM|#
zU^T*yAKuP_HR;$Ly{?%fugCvf3hy{rUr)Vqb`!2ay~Y-o!y9ncE6lRm-Jd#7-fbu`
z*HvMTT_R87Ibu`HJN^DOw(mIS1{PFets}Kdz2BH0xSq8uLw2^>dMXcFc~*#O$C2Eu
z8q2I=YzaDG`$537PZaeMLAQJcDq}YakL_=FJr}{`Qpn9?bSF`<d7SV5fJVQ_MXNFx
zE%Iw2Vaf$XKZ6>kH6cp1qk{-r&~;yFY*XgiLXXdmU!2n7i*M?;n&<#cfWrT+FJCtX
zn*P3ag4q5F*2*Y6Jy^MCGWjS68iIe{dY+FT|AgF(pt}UD3vhrVBfcZ8xWa-1IAS=S
z%_3-W&x?XFf;6@fDiNfOpf0OjrwrQTx&)S-<NN{)%MTvF8QkO?$C8JOD_8-W)%XC$
z{}ucf3<Yq(4Puuaw7-k<90#t{#n8h7pxm$vIr)=_oTKCBLG#}4-_3hPMk((UIe|2<
z3Y>ke9Po%gQhBy_K)F5=fTC_e91;-4kduYzYF^`g1nW=%3&xE@PLfTPa6b+pxUjQA
zpB<lLdk{-z7=VGUBoMd~1JmVGs4`2=qp=MHv}X!%RP|s$k{Axiv85o7AbujP976TO
zV?o^De*?;qWhfSm_>0{DawL7PvDGzcUW_^`f({4_;djSa%mffez^F4wJp`{oakz;^
zN<7=zdyK7qsw`S_p?%%{>+^-xBgAvS8(LsHMirp6x4r=S*yOjpZ_B-$;3Bi{mvXjX
zfQ}-hT6h~gXm}Kh7=tx6r`;_X8t3}=$|z?ZN?XpoY#P|7l3e~B9F_seuU3cYPF~gI
zZ)rVA$9|Y_d?BeZxO~gFF=wa6W-!9^iOH0zxUxm_<%8|8q)oJ?0?A1Ve8HYx(46Tr
zHiYgvt^dmmukT^eAUl1=_X|%Su<uRb7?nOPhdxJ;RpVVRev=E!tRR~%T?tp?{|dlG
z$*5^bFII5+g^bGsxuTcv+HUPq{k>#=S*ne^iKK_SX4LhO*QX?$dZe=(^)134i^SYu
zj4k-lYEi9*X?dLkd5NcIM8rzX-Tk-UvXW8sRcAW(=Yr_vrOn@|{IGxa*1q3IW~m^j
z&O^xnFmx>#cRgz)-o^jj%{IjtL?+v`&J!G90vZFwGS4wW9*i-aY{_S`BtL`t37{iw
zZJk3b1XCm1yv7Vvd7sO1C-gX9F~ZPG?FM3erclDGs0GzC*+wVjy*-Q%cJ=A84^dIj
zfv!ev1s{r5s#tAV|LT4RF_A~bPCr<6KZ9~WZGmoY7n`AE%zd(YRO8CwQSXdLqdRkt
z#NC(X96}|5;?QFAC*BSY*b5Q=+tcbpz-A`??e=ZrU?TvSDw{yTe?W_khRJwLq^psA
zS!l4M>{qMTh;H(7JKGKWjWRXg!SKD%qpf+J*P3ZTu&uE-b!q%|O6yNnc&nu|j2Ex#
zdaV!D>!&9#{P55GpWI8=UhMl>^g3vB5t)XXR|mFcj1*dNHK@n!R8E;|WagNAzi$?{
z_D1BY36NCC(?nps{)>@r`+A4Zf{664@?7fgT#K=_c0dZm&a@NbL2JL^JP(19gvtxh
z+GqQS<>Gmt;wNTX_=_g^8lV3#@r~gYkf<M&MCl;aSwTgUh1?jL%{TUvyyBa@+6u15
zeh-`hG09Dgj*sUVV=SUh^->AW45iXL9SX%7ZRt1WH;24<!pGgl>m-%KqX8LT)YZ36
zykwy8gXZK?=0Ac`)eaDpWQy)Ev$+)<VfKMhCuej?m(!<Db@*=)A2g#{L+DN)8z!m<
zL!MMW;yi=yc#IzO31vEpG+f9K()qOcOeklF2+ln61@^mPY#Tnp`~FL3Zti9JN-OUO
zjsM#-{WLX`kl*2k@F4>xlis%AI8o?+vbSt~WoVw<8STGW&WX|9YIlnDf|Q8^SsN|(
zA?XG#!4Ma}a;s-#$m&XU8P@nm0{pPfw3SSk4>tz_?oQAE(qZVBSEREqnJNy-9?v>c
zm4r_ZBf#WfEKtpXV+?48hh33wJvUd4tg)OsC%5C2%GdX9WJKJ=8gaU%T`$Ii>Y}z{
zWb{tN!1?;kQ~oNjjug8>Y%DsTxd^FJ+55H1;j+U|X|ByKxu@EurE9S^;JrTvo+%&u
zB9zi`y4!EGqtWn!!K3R)lMV&U{J}$gc`)!9fdYmeFM;jom9C$(XV}N9`)Bm@MfHPS
zG*nH{`hM{5!QAX1dAyhpPNF>F=&ruq_<5GgrvukBkUFrCqCRpSw!Hg4g%tkh&;OnS
z|DUt?KXewTC#W)zDBvbMTZl<KJw|fYgf@J5hBcVCcT((3?^a3qoSWLtTvs5w;KDS5
zv_D|GNTp*j7-6ET^XP1*=kwp4H>@8nre(VuuUvKO1LRu##DWrP2O4#;?@Ew$QQ@eH
z=k!*Y-=sZ@Bx*83vZ&_Vop;1Ij=plrY7tCR>t2RtpC7jM+%bi=Cb;lVu<>dv_O5)p
za^~Eb;&{rXJ>LgC&Tf@ma5a%QEA*H*K-jX1O2+YH%&DDHvy_f={q_PET&FrNA5=X8
zRz(G4Hs}D>oHLLoBcQQPjl@92{x0byI{fw;_ZBWmq#v#gCJtt*@jZH(EYrPYe~i2T
z+spm}*jI}!CZd`WeDt*i5P}6N>2_`Eb6sL+2M>Bie13aaLV`8u>np_b9i8c|+rl7%
z^P+i12dKn+{Z4K^N3<^Ewz%+En@IP-P+9|C1neuo9z{o<ba^hRd!<c~co?=u(_`MB
zr-5rn(8&=&@`}C_7H?mu_G@NabD#SlyMIm{xz&-1%GEt9=&EzNtmiH}R9+svps~<x
zDIumcV9pQYn*l-&U~FZb9SUklKoR%?=2V}t*?6a~Rp0!AU#3(t^a+!)-GXnW2#|E=
z*z!;*)3%hMYpOb+J7UAO>sZ2-b|_X`s~Ux`5=;I`;t?R?kv@yizRGZtS3w7Hhc|&?
z167&tEh@aIVW2(qMn;7v$po`9kSC1@so-02TU}*V8155JRdIJ0AkNe6zAxyDoSjp&
zH=l30f1<B3PfaO#j2>e#P_<=eg?`1>YeUMXD5Gy)GVD+`PtT=;k+a<oNATm1N>*5o
zGx$hj_$$m&dS^x~$0Wb$jkkl>PIa`jPojwDY@o}(wPl2a8BO(-q@8tAP;8U)-R|Mh
zS$(3wdAsNFR?4c9>+V#*j?HgDZHEd@_5TLc4LNm}2hRqa&szx{L1<!|&mEI2o4Z%q
zw{YdQu=&oH%5RQs`p{zzlWgF4K+59?yb}B~eJo++ZMnRx<(K)~Z_(9;O3jYJK(gcn
zW&`#dhB9pWwTM_HbUgWNuk_RT@||NoEDuL-A}ouy^)chO`mb>GDeoQG!5wRYST2wh
zNB}bt0JS9(#4y%xTeK7OLKXRK&Kv(|rQHIx{`-+St3&|4h6qd~U1+}gWqD=HP1BzD
z-wp+SaD89?Ef@dfBmb{1pEWr(SdrsIMzlXBfWGF|{oz-#d;!~9la_LM!NFsfc=AuO
zJ-Eq}91vL|&A_fB?dAYDI&!hOPBCTJuKLn1E6Rx$%d>oL<$p-!WX{fN3#b!4x|go_
zix^{Uvq$Yy#S<~_L)-Xvsi_=l!G`yMrN&i2o0UzgNPU~-h0&!XUj52_HT=Ys!z5mS
zjDYE)q3DU{Cn!^`A$j?+ccTNRhKp!-JH|W(@2nDsSZe$oBye6sZ4HFdfT+G>^yC*G
zH}fwTl&CHLak{J3M08%217=ruocj?>6Pd!f`TH`OrFO!%L*MrGgvv!owq2cqZ9TFO
z&tidQf(irNGGG!t2zVXY+Bnwym}tE+G7Mg_7AHD<W3bsSfjAL+zOtPDO(-Ml>_|A0
zx@#F#b@!~$S5!$CN1WX-R%#j*bjPpq665(mX)~qvYt&@v?(=)7J8HPT6Yoe=JWBF9
zqjZkHnWwPiyu2Xu?elQ;kyb<6DCx@E9FaxQnO-lRV(<K3k6=H)YY{^sQWv*bao!KC
zK9Nhy9;`g4!Cc;*Z1wZ{;m47pcHlYqV|xH>wH$*(jkC*_AdmON*j3rd!*$O?JKwHu
zk-V+5h}sPSgRx_fj*l??4vV@6aMn}qtWS{cPXnhx5C#}S!}KURW>%)C>|&3DS=8P?
zC-OhH`Yvb>W$?C&k9WlymKRks5Eo)w44)Uo#y9pOieLN%JLu8Ahh(mnvqj?rweJJ9
zygY7KMTFFmSBY289jk|h@Ld45jycD^1ulPNg+vKG4^j+bQzd=cRPTp!x^FlpArHn!
zf@OyvR0oY{NCnrL;A%PF&LtR!7Be4q++?i~e^#mS|AIrRvGeUT?V<bQMYTB)>1=Ub
z3X|*aC-tRxDR*mR;iIEnd10vk5`=L4f~)|%(_E0YmGv*9qwJob@3gvXVP8MCrRaaL
z_nuKrcI~!s0099p(mPQ>QK}+MT2Q2k2!evrqN37-h=>FU34(}p0R;gmN)ss&={0mj
zq?bq$2q-;}P$MbNSw8Q(zi;pJy=R>L>)U6HvwtvfgyEh0UiVtpy4JkrHK)YrO%}Oj
zg_Y$KY~;jF82O?~Oe23|0uRuK<u`7S8>;d`(pH4v!PuEZ!0EabB&Z!zfqQ<-+uqCM
zu^<+>&g1<+lzsIV<Sv3L2U@S;_nbK9mxY`>JM5B3g0&T2Lyms=b$jMe4Fu@7I%1m-
zGf(zVR4w@c3TIB7F4Oqr$D5Y><5SPPi^P*zyz^q#f$4+bUGMRRy(%g`R<9tTx+QzS
z$N4198_hz1yIVf3ay~%bCS_!P=nmj*N~QTHS7_ex=?%aI0Z|+!_}yvhvADLl5QL<z
z`kazOn>O!gdfu0HHQSye@lVe~#H_Sa8Rmcvr2r0(ZYtsaKxkYaN=<G_btI2ZFF$5p
zvYvFko!<0wZpk?UOrvvjZLpjT=r&|*+?Leig|;cFgsGMtod{u#SEbkPzvXT4-0r(n
z>8YnN_~VPp0W&d~cSj5g)qtP*N#-bmY>8mO48>`bl~b|p6I|l2--n5SdfVDTlflL!
zaJ#(q089rIiY}r=rNx<5{o~Jm7NugC>7PXpdO};U<csUE`ZAf_wF7elB3hd&pE(K9
zgmthl{vMzKxr?K8J@Vehv=cQwkTYW24TgT=htqI+sLA1&poQyaz%45Wi@a7>iRPZ>
zsL>Pe{+@wWdXPepxj1XmOfg@b*%<lt8nwtMHM;Y8=$N1R1NNTAC#2ij(%Y-AK=UXJ
zRNQ6a+TgrXPEIHj`%i@OhJ`OxUbBlTUehP6F|U7GfmwW-scZ2e>K^&+KnM00v2g}B
zoeEfMWkdv>udMUQXsMTGbj5^}j4|h#c4YNr{9=ux>k1+dUxr()nn8>uwtD5Fj9UdR
zp7!}L3GtR&+I_>a<lF%EgBfgPRg;D9P8=0h04(@+D<5mee2cH`4y-l_d(b9%U<Zc=
zBdCbK2c;gvf)|dJ(v_RD_{R&+nL5_iKcV*A<jUAXNDn3tthv8S;bJsU^;Er&cZ=gU
zn}cyIKi=bvn7693`oXs29um8sTUB(k2YqRpYzB?t`4IGmSGO~hO6|VSyhcj_tCmRP
zqO(zus|#c=UAfur6@{_6H(i|HtGtYkMR4kEc;HCBG%g^$Q&eSfp4ydz%9~Pp>5=rE
zoqVcLj#FM?35*MQq9c=Rx*pvqTAluA^VaPCm+3Kg>y&jK7?ilS0{dZHZ0)Z>Lm%Q$
zu7bvp(lobTua4b%5P0FqG<`7E5?i=~(}o|>Rs@;R`LMS5rM(+RL4n|&#Y!U8t{+jE
zylNxTiF>-I5`yMB_?noq74_v;1^*FK>qDZdyxDrD>f5jXzLxZ#=%cTM`33-^@;EVz
zSv`&wi!j;9wL#8TDOZMZmssN8&f_oj{F!R5G$4wgh>-<uPn6kd33vS?{a%lL*c0!D
zNF=>9%{p2l!EJJl;kgR%qtKayH8eFk`zNx=3*`OvZH2&zs8IFTbrsJ~hkbide^%%z
z-JmSXSseR{woze@ok&t2w{GSme(h8suRxUvZ@)E*1--gh-rv~!G-FNkx5yuVlF<PJ
zQE8TbS28X{O-cIKpm~(UiDDqNvIH_w=}B9Lpw&LRD`KR35tqdVUb#-*#>+4j!0OOH
zL0zA9ji@(rpG6#O_bl+>|8XHf?P*Ba9afEP9Kjw~1j6rOpNl6qXFg>LlD%=h>{Hn}
z6QcK0C-fcoSI5K_8lD7!EEfdL$_2WBa_3GYMX*to{5Ts){0ky9iS7L9Z@|d-xjpqb
z#6Kft3BlL(_hRl(4Ic|08rPj8K`Q%j7a8AxQk@s$DWF|V3o?7)01zY|q+BDVllq5`
z*GCj}ABrSmtWCk6w{bZ@P3<(=D{pMY`5@uGI%}beuZ!-PvaS6CpKfD9@F~DP5H_?0
zjmS7MP<|t9G|lsaInq=9ZMeaAYlU4nlUssj>5gU9=UlPXsC~uRZr^sicf8$EOt(tH
z>BaJ?VnlrAr{tzFj2?3!LuyD_X?(@F?2WUB7va94mK>8Ea7to9u*1*+h!7@6)1W@=
zSR<^~PQ4Qu-#&OcSr)L+I!$po%L<;4Dyky)cgV!Bp4eey&HaDL*YyB3uxskukH(PE
zOi2nRS@@k>)0?|@rSH~d&YgcEz6HeZHhAXb=zPg-d(Q9t1IW8=k85LJ*0;NUSfkku
z$=6w+ggWU0u(@xV-K^Bh?pANblGh>`dtjvtHp0NHmDWE=4Y#WF^0yg!q{G|PE2}b3
zS`%c6CM>~M>u867vR0DU_}sb5#4CiK773mrpp?kBDgTl~kkJ2+43@J6TOmj1$2DWc
zws}q3FnqOO1GZ};ob}}nR{ZHZl42kyyV;IbpN*r+R(0T@<sst|O0MoRW5PO(DXjQM
zZlGc82b9SbM47NlcgbwE$Sb&ytXb5f;tKzyTT~fGas2`>fOHeGMvg`oMdd98?EkUz
z#VG)Gk#XUDuqs`J7^KjQft3!ZnTtO9A>VO?x1qR}WZM4-Vw($Q+c#`w{;FXXW;8>p
z($|L(H(RWGx@=ewjeE&4hLZh2ilh?=z>*v|t3C=kS4^58Qr8Og5)p3S;R6+k$%{Z-
z5KfhoTXm;!?%WtgoA}Op8D37b97)ifa1c@DuzGOWSch;9z%K^?{Ng~fzd^N!9q>r?
ze*35<Cjm)foqKS%siWU>_3H_?8Q6DDCKq_4abuRVn8H~&zo)(NySx7R74MC1I#%EQ
z`p}=HQ)qV^-1NL9DG1E{PT(S7fc6@xzx94)Pbx3{vA(JIw-BSB84vcaBd#!85M{0}
zu=ikLnm}$&iV^!R;8<s>Z0V0r2T=<v2(tkL_9MCAMZ+2o|9R0&c~n`)_*$9QEb?HH
z2q9t8I6>`QNEtuN?<bpxd_)<(HO7!A*9<Il&97C5EdK?$_&Ijmy?FCinS8KmpVn-H
zG}z={Q45F;I7#|haPVBsjg@G)*x`9SCs{`*ojb&5B6S(0dL2DH%>6?YmrJ9HDzX%l
z=R3{X7n_}Jnr>q25T)8qh1Zvol6nnwPvzV<Ie-4C?T4&`Jacx<hy&;+WTo{m-D~%h
z%W3hgAEfVGI%s%A4=VKAi4w)!KRsswb*#y3dh(+L_fFu=k$R<_+@1T%P$Hsq7}Pp6
z={m5v&Ye&PY>S*`Qhrq2jQSd$6FU>lR2o(1G4F$+p7&iGlt{V1>K}w#`brX_aRG5P
zIH(S|WJ2gV)RVK&@F}6eTc4wDd`kGbwr9lQncf{UcuJD6U{M9EpJqp+W=$*f&a~@G
zycl<iN^|N_w)0%OL~m@tyMSyO3i-aGjSSf`w>&gy<X>XwF6m#^oE~?=O0*=P!=o=s
zS0x`c(Q2O&S~yWWn)0D(hj6IM#8~ub#fL2D7t2@}*?3ye9+dzngv9xpyMFI3{+cbV
zpOlM|O|MtpFID)Q45Rb=&)vnpP{a+cm=ESvq~9@9j+zT4JVjK#vSL@_A9UaqZW(`U
zmmz_gi%))cGm9V2=ZNU&d#Zm3kThm~*XztOloe>UoLz4|Y<hhMKDGk=6z5QokQvbL
zfEcAp%{CiWUe|}AVhPrRBF7c<4E@LZDz80GC7;F5yaDY9w;HdDwD_7v`z)yZ5Tb2^
z|9SFH`6fh%ZcRD(?*8~re%|5}%#16)h$;vm5M$4+>>Ve@3#K%)TcZyiPRbcpU&<a+
zE=3FWl{0=>Ls}BY8=t)o<~Ho(j5i0#z!91?wpku6MoG3Bt~Z;Z;i*PwbKlJXQP!>r
z$+8npO^Hj0gQ#pKmkOn@m2|lBdH0R+%e?U}bxSW5b}xbx6>mRy(Bx^UiomCmOmQ!D
zv(6OXeqaG~9n&{WHlFYjxmGpIJhAf0qx8ZB5Jm31m}ng1QIR1C?S&aZ?owS_wGp9@
zB&0-)K8If(Tz!Ati*(beK%6BQQhf3Mq?RCgo{5zpjgd9wJad$cn7izK7Ha>#ExUA2
zM7dq7%E2|}5TzsPZWS*5QMOQ89&-fGx&55Adsg}BL*KX`Qzx^acJDv9)vjgEDhM_V
zsT^r6$JVJG6tkSOTm0kl3wd9N@UX4De`{?`rpj20SB9OV0NZloJt&7+6PV0M$msje
zSfb(pcJZVq&{Z;mk(1o;TY?pTLGbT`)jxbtLKd&v`PEatHI{x-*ZVt3{F5mD{bTFc
zcw+EKe0HLBrHM0lc^01q*eS{Y=AD&`V=x3+ijBZb{CD4#G~H(JG@N#1SSbf-^hz8u
z@cyM_bgYrmf4u6yUHw1x`$~`6V~X16+D`H4veQkh!Lsi^XH#tBWaRVmJ-$ssj!BLd
z|8K5-BU$>`VDUZFy<xBSuMdie8|>GA9;~)llmjj!8yQtF4z7FUM3;KO+l{k=kn(@K
zE8D9X@&j}6@!LAr{Iu?>Mu#rh-0-Y@oO?2IbcZJb())c<AQFQAcOt|8jR%8M)qju2
ze|ij9RoATbj1|bpov+j-iX{;VeB{oYAJp}mmEcuU7dA)P?ioah8Ohx2<+K^Z2Y7br
z%VN6lT>8o$z8=nsWaZ?dW#5N{M45n;?Jq#UiiGf(QlVDN^7u8>kD#v=U79@+(}wdo
zsc66^`}WQ(-V2B<!b^eg7_zmXqK|K)2%|x;l6Wl&2qvM=ug6uNoo|>RX0Dc!vuCye
zeWm`$OpuCEqYTHqXqPWF!E@xpKZ4CantT4Q`*)s&+GBG4=h}AullT0!uCW!!eaa^+
z0KSIHP6smlPfL3kf@pE>9Yr|jDe%zpZyj%K-jAK$*U!E_afZpA@}7RIwwjD<#mclv
z%ba(q9@b<-dIVjaiGWsMZlbR;3iTDyk4E^Be5gF;@;xnwK-QavtNgn6J~8*vp%fJo
zlp@~Rk|Fu5%m2~u)TQQ=tiH)u4@M0GP_|~$u{;>IMY7&#9A~8N`L73d_uhJ{7Vd65
zg#lEmO$YkyFNh%eMtL(4Dd@r1HE8^~`VO3Y^t9!0BOCn+xzQQRPL2Ixa@_R(@xwtE
zaY78`>)$7qDoZEqn}JZi3zwva(A7ZhgzE{Y@w0-u_fLIoJsb!6m0i(Nl)_of7K<ke
z>2huMjjw||nxX<Da^!QKbacYr;D!+2JDJ?ojb}$evk(W}HkL=q7;VDQ_)O8)6dN`S
z5y*Gk#pMG(`0sK=EQuw{$wWEAD{G})ta#f=`&b118#?b2G{tw=j%e4Z68dBpoUSQ4
zSIL;#I+Y^XZ@woEn@eBr68Sy{gwj+Eb^TA}eERaaZSG;MWQg9m0tF(HdJI9*^Pxl$
z6{v<z)n`1XBv?gr;;YQ`8d(l@*07f%fEdKW0$terTH_#QKh@fSXC+PS;K!MYq8>4s
z?nP7GNP0kcN5;wOmnOmuMHfi})*IHA#9BN8J5vtdI8O&a-z{+38SwkvLt|eO)?Of;
zo9WiO*egLZQ|{Ll;i?h!mNJ3_z;~u&u?s-s6~4QNrDW5Po$yW2SO}vr;8em^^fVh2
zR>qH(c>fc8Ds9FT%p_+sVc7mOuz%0+W5Xa=O{F=U#1cNRP5*Gk-c)ZUP2C@R5PKgQ
zmC55hI%i19csGBtS4CEnZR4Tm?gnkghw4z^O2R;Zfh2Yz4>(o4-pC}ygQH0|S+mWO
zV_AE>m83iYDp94+x}_>o_mf<&iQTk<)*EQ;W^uzxIq8DSopX<iU1#vH)`M}gDWy`y
z8yZI&Fj_K@BRjZ$$%XhYxL@z5=B|)SsuR!2jz2DN(XhU7Og7R;&*U8k{yp-06%fQh
z0ZydKY*Yx|+byQJCAMx>M>VK-lBveBb`@-}LQ<W35%3@IuQ?yq;KOFjCFWHY`V4c>
zS(U<`*{LFGa<Ji}iTJb@cU1y}GfS+5b-FF%O#KTpjsl#8v`?c>#2d^_!WEhYw(Bu)
zjK}I>DzME;jiOZZcGzeAxWGab^3pI;!n%%Eujg~!OHmhK`$Fmb4?_=-aV}V_S0k2W
zd8nO97?%TzN;wGX6`;p~U$hv&lDXh5DVj}j;dcA2E!o|?^UH1=fqex&S$gm}&NwxP
zoFu4-XQ{c^7u(_EG8;+=8#D{ZtP9XKRELia4%Aelt`l*jrJ?9+>DO#pivgGJW85fC
z1k(YQ0{U*wkLy!t-TtgI?-1qVV<wUVv_h6<T!-bv#Ex3#&J}7n|B|m6mdJbU(x`x0
z0@*iC-gs?Npwk1wXr@Y0@U5_N&Q=;!7-dg(9CK8bji^o#<D9aUGCmlv9=Rn>2l-jZ
zMS5U86-((yDX*=&hSQBCkT*nM$KJ6$arJSRmO9PPfL#zs`~Rz-06P+tW`@$)aU+8h
zbW}-B<VjIr*K#^J;7;?A2m0xrBc7Y)OrEQpbcJcox4t{mjMs~7xqsx&kbOd!k<9Y;
zj+ctF{q!TeIeo$z)$u>&!#&Gp4m~^*_k^(@*Vads!O`CWtb7NW`G;{!bN$!1fg+G1
ze3kN7hbSCw#}lTm4!SBRUcUBf)%%>3!k8ztqY*lU4k8AbwC>d#dudvQWo4XtaPd<!
ztQf2qI5*mBR!sjG%K5DBxsRf~x!&AqW2a2s)CVmylL2sbHFGWUH}cOgd7}Bq%zRg@
z>$^+I!U>9{r{hs~562OAa0RctKqFf_MCo4*(Z79$)t#=Ai`rg+j&IJ^3y)7_Be^}x
z*kcuHgs%EDLC6YwVz>qP4jEAH+o&MlzO{tUg#_G7qbRg%JLh@^LSOh<YL8z_(R4O7
z{&g1CPC^VISF0$>&?p2;z0~Z9eA<Q0>xNQ8u3G6={NsA@8J0ryGj-&?Ry-%Qph_?Q
zh44&4=l<J?_u%4;3*cc+(Qh-qli?8<Da+$`2JbQlqXK>w6$w6F)#&G3^7?br0j?xH
ziYn|GmKA+#*5vIuU$6XcN~%7!55sQ*H>v`{d^lY=a87Q=B&1%QJkO=k{n1qY>?POq
z(u6w8(jUO2O+g(Qk#TZl+kE0T`rJpwDjVH3;T;3bGyFcAfrJTw>mvC*_V6UB0+XOZ
zg!2WpJbkD3?k_c)B5wr{e9IXu!;dmq7BPw`K6%T#0uv6E+T|BGn713d!QF!*Lsypv
zc841fugh6z$p6kH|I?Y||1#^gNHB`j@W1&YTk+ci$hO)ty$}p^8FK`7h|V@ggNA2F
zj7nVSFL^Xlm*Jq=<s)-O^|dneN7Ir4EKgd98bj=p^bl_V5QzQq;F7AWotOL3FY$G(
z29sQ3GXyh4D(J$m5l2Gkmy(R#k0shvx;{RucnO(cfl2D;3}q4k{oE*dHEl*uv#n~}
ze1&gnTIpq&?Dwm$mFl$g!UCYHWdMXz9KgHdqLMJ|pUL0G9e0JxCZodNd}39fG}ef%
z=)g9+!+%bpRlwr;1@>A?gfPIxWB?O$v)q!#jsP73A0dp1zfZycA~7wwy)cV-?uB-8
zC;AqRo8czc#_~Jm%L~izBw0PIe?kD*KiPVH9CZ6bFkGF1Gqe{)6};z(UiM4F$q^EV
zUp0#&4H20L@+G*S2UDH#xyUaj@Jh`{h1txjE`R?c0TR1j!+=iP_P_f6|Dw-X#6k-0
zX|W|(_&Q5c<B50?Px084_scuAxl?bOm6JxVy}Z!o>V9||*Ndboz)069oW$fvwRH#;
zaaH?xa?tH6n;-VV=1H=j<Rzs!*iAs{U{jN{Yfg>x^^vjgQQESQO^_@TkTyFW47XZo
zgL}fqA9nB9QL&!0QKZw`*Yj&>r@~<FobG3Pp}B!UXd&>Hqd^Sx8N{y2!?zr)IFeU?
zfX{RXux@dYN%}nH<nr<Rge$u5iSvpalRh!oqm6U8CLHw`kR<B`MFZzVKrybMt^Dha
z;@^U?>&n#OAzhVRM-Tf3Mq}D7*5riQ<!dp5EEzeo<qE&4j7qHMAFV@C{gbG&*3{ne
z9j5f)wocmqM|J!IO_X+pqsfl+p0Hjww%yhijQo(U53vD|_DXDb&G9cRkPp(O%$DKu
z1RNV91pBiWBLIXwRTx}_)BmvD_|LuS!jH~a*QMlSN6f59az>_nA_G``fW{Zx&J<a0
z73U(UUEgU<;cZ!~8Z54HAtcM_m`GWmVa3?y`?^>Bed;X9ysx8=G_E)&%LdY4Q%pz}
zb3KN$cAe{DD|Wk<k2XD26&>EhOWvx!w#7Es*v5DTpGjP&_!2wP=3wHUxZKtPTnaMZ
zfwjyyI6hfR=Ix!^!2l540jN$$jwig_uET<@Nz)k;YAGK4vb(?#-S=zX`LXX-u3DYW
z+j~+7Nz6Aq`7-nXDB%d;i7JEJQR(L|M0X1Ma~@OOFPF8AEBz-T^)J@qzaOByx+1sU
zRiUeQ$|NCrpnuN`%!ZqckNSt{`V^zg;QBL#8sRyU6A`Y}Ws1rCcTOtQ!|J-Spo5mY
zbQAJ)w7Mrl=fVid&mm1(S-(sEOX=G=Fq?NulJKfPO7LMrUpN}+qaY@iI@2W^{jj&K
z#7wUTlh)7q9F(KW8Of_<GuL&^0&^RSB?o_>7YqtND@6|SYy&k!JoRK{6wL^2`I+pO
zA2vU9uU1T;zCCzh$gNm~D$m&h7p32#s<`A*&z!3!Z0}#IYG|mjYJOP1Ve$?ahn>m5
zf3|D~2SI%K<G2!3e$)8#n4I(Q0RIPF6#+Xg2C&lMf5Vpki+^7mQ8%{`v8--vtd1S+
z{c#|nN5Si=m8zNg??jpzU|XvkY`_w+q2uR0%z54*oYJ%=wSEY5u?O||AwWZ9Txwic
zy}*04W`Bt^PYlL!x=TRFFXG~BNbG~lz9dA3vjmxh4_Bj-h*ozy`;MP=L1dE=oWrw4
z7k)2v8K>7|Iy59O%XKgHsop>fz5cnQC%(DU!}zf6K0XV7Q6>IF(^TeqFf3I(lqtTl
z0|@u;^8<bx{srk+Ny@=ZXA>=<vWW}Won1QHPuL+BmE6X;kAFd)*D?z5iv?2{P>!e6
z!ign#59FS_f95MV-MhwK9zQGeV0{-vdlf3{Uy$pk8xcMu*j6XZ@buQzb@&M^^#_bh
zKvZZlgmn78!kC(l#e&#2dE)dsID{d{Y>b;7zk@ojfi`nLnhkcqrk?%|)Px@W$16N6
z2Zx74h2dB{u=>R|LAOICe(g%x(bWkgu)d5gZ*GJG0ZMT0_zxHH`#NAkKVKrn1mZ>P
ze?cJd(F4fjNifd;{&r0Mr;892j9?V*YJVCiwu#Uo|9qlR6WUG8@M!Pn%!pkZ_Olyf
zw~jR6M+l$kVqkffqR&$UKeoD1%#EG=jpIAh3R<9<0Q<Rf;3VC1eh-JuZ!Q?`Li&n#
z3&i9<SlYrVoF*qL6r5w6&v|E)`(uQ@Wny&jXq&P5gpvBMcQgXIdlq!GtDrP3DL~(E
ze}$6#Hk)kxr(+V3SN9*?dduf@4Wk9NY6a#Q4>%7K?nGCrOETL+9Fo!d+;wEkd$H&@
z+ej@&h$e;T7OhuuJa`o4u{;#&j#CR_XvY__Yu#^&JW*n3DA$fT{+Wz?zOCeYCGl(N
z(2Qq9=D^iEgv~3Iq1j`*?eHk+<CJ6c`C0S2F?5w2y!OC(C~ZNA2J#$sVanAE5k_MD
z{iCni2s?4E)snC`(wVO@`{`!~6P>AH#2}4pKZNEBJiS+zaq<NWUI=Cc+_Dn>o&)R|
zjJ;Ub)t?prf^Ne)nHG)i@@P^uA@y1WRb6L7$v^O+bcLM7$C3OKmX+pi!;iSLSu6js
z5#ajPsVDPS&5D`*c2g1*_|u|WnQif<6^*K;R{g=cjRtyL>U(UU7V{X}3<uLIGm{T~
z&+^3(s;(`yHzQ>J+qk*Qt&=ZRg?TsmvojR7E=K~xramSQSOKysE!Z7lEJFm50Vi8x
ze`~LULONj<yLcXNi2t8K36L)B+p(Um@PFDQ2ETtfa~(8<J$D1FANjK*vd@hY84f3T
zw>s62p)5%jv};tI9e*maaM}KOzNdeYsJ()dGS}<8fc;Hh#10==$h-)lpJtSj6pP><
zun3L4$3u6eafhSaATF%<t5~2E3Ih1^M>U%S;mrHUvP?uu6{dpPHH+u>kk+N#%PyLl
z4>GZ?c7J<x<6CDID?UqwD!1O!U_Jx3^)jLq^m>)>cv?|%0uFug7XeP@Kf$*6^#rIh
z#m+G<WUgiYvcBpMTbBI`@(3uxLYN*PU;M2+bsBW$Ea1kVRM|dyN^cBE5Ow+(T%HvC
zPAX_1CC$P1m`@m>MjUbpoGacg;&%i<W&V$FbJw4p2p0mM`8-;gfcf_1>LAG>hmggm
z!7=wLQDzC@x18Gk$78`hhSU)DIl7#nTI2@mzLYd3cP<UFgb=GYyE~)8?O41P`o4SK
z$w@^)sp$(l^Z`p2GzotPSA}ggL@Chp%pI|AMla!Y)!qz+y^FXOOX@MpmLQYP>GG-c
zVxHq^=lh8X4&8CYm@#OQeoiD~B5S`2u6V2NY{}X}?D(kQ|6{PZs6fPyyD+&MW-Ooi
z`?@i?o~uZBzI>}236fuziElOuSnl7uOzu*eyN7?q@oK6GQR&=jIgS5dsgF0+{Z5no
z(@rw6MrgxsH6Ac<>vl>hE)(vkp)DQ_n-3mC2a0ZHM5<~E#Qr$JHdTKEESs0q!dXD5
ziFh2AvnVq-YJz2Om!MRbJW3P=!`gF2RA2K^QY_o=1npDO*-|ScOE3qUPhi+lw!H#@
zxri0<7Z#r<EuPRscaLcQQDuwG1PsJ*%wh7#&uQLYwdCCwf?`J<>XI_VuHRX0KCJx2
zwE_}|Tm?asot28DnX73#<&O2!FJEQLL%cRrn2Q&IpTgy@{Enkkh0t%&^wDOSrX$vK
z=O`M98Z+cC*ItD@_}cnG3vT~M-yq}>lVC}^gg!|-UcK0;(!Vp|EE(*bEXE(UdX_a<
zzXm75$N_C`EST<ENzEyxhWND-t*W7-mWH+rnwexEp^zG9YYC+WULBF`=vKJl?Vxw+
z#dR?2m_BfKShW2DiHlNrV|1J6x60@&CF@h%+3W9>&&$Vk1nYNG0F;>7VOO>Lm?=i~
zPIk@bILeb2ac6TZ==vj$`ZWNBk|?pGi`|NkHBTpIp4@=!%-D^>%dm8Qc$2tgXgR}t
z)~bepKIhB-t>0l%?aH}~$1?)evkhgo;IZz}*+v!Fa~g#+)Yw>vb$4#9#~Z*-qyGf<
z0ecVoXTn%Mx_qir=@I?0?4(zF-gj@GrrIruwEhKo;S8g`BRZ(?p<Hc3Z3n#xkL-lk
zL<2)(XYqV+WZ(DW$T8cE?ly7Wp=D{F;kz3y?g#HwCMU?;PcUtwjV!{*rtopLE9A`f
zmCSIvs{3eoanoi;*;$1HU-Yelw<fPI2(s*j(QN7W>lqoCebnU6T1*;8v45`OJsq9o
zJ2Bl~3c8%Jk$4e2s69@sI50}k@SG+4%&>9>tDZjc?49zIA8%)(rGC`lCUMa)E}#wG
zP24G%!3!CxcZ3)06XBcqcH6fiR9W=l*(-<yEKr;M-bsvG?FO*}>LJZuI;^CO^F|lM
z{Mzj0CdRSUJ`!Zvlv_=wSuMjqXST^vY}4M`uaxORS*mK|;~%t&eI+cxR<prmyGRis
z2~hHl{xmFxJZd;PVRet!59uRQOEjC2*0F1alP^z?tuzDUqYiiU**OQ#!eS+_V{eWX
zcN<K+HiM~dB8m|E;WOz74t-A4x+Mi0_Nc*{*SjroU$XYA#G^kCVQdJY1~3ak2HKOl
zJ<|ueiCS<a)K1$d%c|y$o0GaLA=*VArjwq?S^qYrGepZ^zdm7)HQ`pXiiBmONtOp$
z>DiM&G>vAC8IYD^<m;bAc#c;L4>_2f29`U+naZM@&$!by`JywnSax>Xaa5(_IE{C{
z?lUkh`@FQoUkH!OEBZR<taxWb5xP+KoTC5pd%~75I_2yHy>JxkGEJ5JrmsdOg-2l&
zEp>%O`XjuybnH%Bly8VlN)`YrmeBSuz;|BX!PE{UTObci$tnz*u1}f)fmXjPXsSLp
zSoUs58i$Kl8v)Lyr~Qq{%S^C0-kSW{1}(-U(DeX(FdgYET2A47u5KRWU|wTo?dgAj
zrPDC>IsP!*tB-n&`HpO7UPDQ4%}cf;=hm*(eTt8Av{$hhlT+r>3m5tY`vfBeePs54
z^2O?X<hpEfb81l`GSEvgyu0c^>+VM#K0w!^;kX&4aJMKx^G+s8^+j!wj!xgGiP+)s
zED$@iwN}52^BP(wj;PT*CtNp`)^6X(#;z+s&e3DgU`Ddv2Io=-5PM2me-}d@!b>yh
zg55yNy-~>#W#d53drj=o;8pcHZ%*~Tb35aPV`yElzIVkuzQTkqG`ilJp)#y{hTLjV
zSzlK6__@>V#g`pMS~`#~3U4<NDb6g^_+q+7^{8C`8m3F5Am1X-@+f2d_$BY=Rg@%Q
zEb}v<wu2@C$$B~DY@gr?f*k3djIFPHI<tAN+J=rmd2!+zAl_(oFh9HK=LXbu-0Cff
z47mMJdP~m<e*yVjc@7r!J-=C89B(eOH~BA!z!u}ERsT}l2c!IDZWos@IZv0D2jTKm
z@j0Nm?eiWrajw3U^6W`ti1P5XeK+#gm-yI+68^e#(RP!W6&25(*sW&IMV$z~Uc7bo
zmK6tw@KMf<Ea<a(0WeM(6xT>yyX++ay3RnB`N|v8l<C_1G5=FyVta(;WsRS<+hVwR
z7!gK)4duapzc~FD#8?S?86ck@Eq9F|`t>|Jv5ZQ6aZJw>bgT2oAkD#9z~qx(!h&9u
zvN1>CC?{!~c1>=zmHc>i{W9t&#kPPMlkZT|EJn-)?z=z5=_d!brbU_knF8<1f+tbo
z?N5GnP0XKFTzhCKmn8}O$uOtDD2<@ztlVe%>xFWzJFMwDb@v@fminyCP29p#6&5;p
zHOS_^p83msYd?N|1KoR<*@H~}c<41t*F$rt_oh~6qj&A6YEIk`AN@+qik<um^5C^x
z(37fY_{^(mmijBVSCH9=CI2YtFZorC4+Q5RubJ|YP59U*lZTzCBJH8EhiKC~hCUYf
zpjoE2;uz^F;7opd_#S_*Jl(n5dC02*NEj$gNIc^=^x@$!v6<vU^mUTrXosk{9ZN!j
zXEvjNs4y1&SiAh@fl99L-3jN|YH(QaRF2H?)bJ2pcTaL_d-8@83Tyk)B)`w+hzs&8
z(M60YEC-CF$1Vc>`!z<bdghpbRlnC3{=oN@IqSGU^`Cvxy{2=>!Nzu6S07c$pB}CY
z;239sIh=n2*)!h(;7&+bCuNA3-v$$>PM8hRkL02(hzE8y%Od=<HS~RVBOD>x^Nr`I
z&0rQpQ%Xu_%|D=_+2c!3Bvs>)KBtlTcly|yzi>*wCQm;@yUa$^s|1nPF9gVzQjTrc
z^|*@u&QCcg152}8<ZOXaWsVomLb=A4Gt?+I2h<hbgr+M?U{5uf_I)GPsIfHAK&7zR
zN!N0AU3#%mfD#^dxIWIN+WdOS*5JC&{Amd9TULc_V=nZfOQMTOgQ1ttb1|v?e_*V!
zDkhhMFIi#d#US1012qK7w<vz!cz)DU=4-sNA#ZR<XJvrUp3Xl2B<$HgvQG32xpo{4
zY5}sdQ~zI-K1e4AEs2bTk<AsnjK6-jD01^3bj+SRcJqc)4uc&9r3-Y<?$YH|fULTR
zohICMV4I6KcU|l2YcD*NBD4pNiu-!VlhUe;@ZK=rxUyxl0Qcj>_=>UN%uzfHqd?W4
zn_2IO2{hc85+3lo!7oX^C>!nq{lFRw&-1&vT%^qxe;KF1NXDGRbs&S82Pxj$?(3_s
z^R6w&a#=(1Ld&lpW3qh4z0-c@<jA;%x*aODSQFHi#-rL2-8U+B(!3rhmVYPA>lj!W
z>{WX1^zBqqKk0Kb`x;MEE}9kwc-}KAzq7H{rp!5GfjI)Srg@YxXE7tv5|t<CTRr_5
zrKloRajl;Rni#$Y(i5~m^oc5}3QdmLCk?8r30n2|w=VID1FbS_zelit*K;!RK{q`(
zs}DpugSF3I|6B*jOFted%hxWs`|)D~gzFG%y&-e|3eW{R1~!-PC-Gai`#DJsf_7~P
zvW?{gKN9SZ#8r$4xu?leWaXnxukG(5;>>zyQ0rk}X<+e(fNjE*WcKcgXe-rElM|>P
zPo4W*c4RK>FG!sN+iU$1G7@Y7GoN9C6)Oc2=KbXj2M@k@X}V^xty4$nn%)O2*IIYp
zQq!T~Tj!MDeRQfy_YZvVi(dYPSEf{VdmA^p4)i5V@y{@a`d8E1y8~yjhfFrjA62Ow
zklowgECVcI=UVXO>v2)qLc<R|#4o3mT*#k~B#yCR<Q3K|$FNLpKl1Ke&vr>@@mve$
zL?-vk$wNxHE(3gBT`o&jXuxd_xkMMBJkNZ*r}uj2X_0Al<h)LmTYQ}W%n;-?<@u|9
zbMP^fS#Gzz<<ByS`e`2>D)%kshwp>E3I%$6TOmba=l&tl7z`dW30tYO&=JQuiQ;MZ
zcg99M(i)zay7(7FZ<13>bla*vV|(`(#C2tSC!H81y>e{?cVuX5<@&cW$4RZLF$P_-
zsdO3@U>oc&$k>qe{GmqJ;zGN>)oI(>y5l;oyT&#NdMh39nG$@3Ms5LhOrL+zuI$Fr
zxJUTNsGLk+eDED>lbiYR&RjGeW=m}*$*Fj9Sv#zINgq1&oog-R=cGnJ%-NYp$U6NT
z_>O1Pnj#CCv88eEgy@GdPg6n|s`9*0CO3_OHP6tjEg#xew@H+24@P>kW1dlcnOy#4
zp4Q662x47b0c-G}<5XReUY~jw8h|JvCTLO0y(m<pWC*R*{%7VgPJei#EqqEBh*(B3
zK<ClSrwn1re6d#!-EbheB{})>(a7a*DN5JkcE@_J+AVsg71*x`pK=~^H1TuS=8SYp
zdsv(COq}TfcT;(qy1q)H7+xjctJj>guUu4SC5aqLIB&{%^8pG+z*1#1I~DpO@GQ!@
z>IG-?`odftPTYx)QY-zWk`NTp4hb4Uq~gnyfue64o(*#bmDsG|%syD5<8^$gcyu{1
z<=Q0M@Y8n8YrD-(FeV&xJJnP$b#I-7r$aDpi(sqh3@b;B<+n}?)dw7F3>mMe>k&>y
z9$z{p-~PBw^j9skw^4u!HbbxzJLeWyc+KE>u77v;==@q+8&BVrs!}<O5!w)GwLCyK
zV&FjS85#cM(RzS%nm}=pSDWnvVGa`+t@TufKGhC{CV#3;mfPc%IZv=;1UzQ$nH85f
zk1OtZ?fc0Mp#o11Fw%{5#2_vyu+Rc-^Fm{8Ut5^W&{KpPrW--NhU7qdj4IM~<|HnV
zu`f!#%N`I-Gx1LIFa2`c=V@dYwGVtl5HNFr@=jK@{#Mb|EcuueJvrIXX04&+loNdq
z52Jg{{6G>BF}N)y+mnjwG0Jo9OkrD_jcWQ+K^DfjLQbo%h9`gW^$!wyK>rXnW&v2C
zj}RH>{r!j4veXLmUnUIdnq~*c5!AQ-cjUi+5*h!C0oT9Dl>a~9gUzsCl+=+cIB#OP
zotDj+u)YOEB%FNR%ev2&k@G0xe#!gV7jE$(<jNW78@k<h7_c|phuZDP;3Lao*u$+;
z0dqk7r~BtxsayeXQbRxw22Q_3&WzTDm6*8Bnc|2e;)fS&pGB{i$%!iZM$ZSsEwM9^
z<MZh`bX{t!3bL3aRv}?6?9Q5?TLnb$21%OELlw{cN*hB+FYi86<>&-znxk}iY9K|f
zqtgCaBYV}zeC5*-@5^awFW$YFdpy*GNdanvOli6q84=AqVm?<GWY6Qis+2H6yXgAB
z(X_pCP0uG1IHSyzX95!{5ui8&w4oOBaZ%Wrc%aP^FfPXk!2wg1Ye9e3O%BNP1ITzr
zI7pO@^e|q`HoSB>eynUR+F5I)cxp|;Np)j?R@~WuJ8}E97S8|?sDs5tH{2epIxlNl
zjoBNkZeAu{!{iX&I%0<Zf^d?&8~Zfda0lA;eSTHu9zu4jQ{5>JJ1?mNG$+)kned!#
zb)?DZ=G?^=qvx`6VGrK&bNXJ_?p*by#9bVMS18jZKIkfHu3b|GqSj#S-c55+X{{LT
zr}N~s8kgb=nE|W$Hw^Y>Mj6=CjY8DdbNFZaha<fdZj6*mo$od1_}(My@<2~o)j;NZ
zBm-=yAt=nG63rUryncalG7>7~emcCGPkFdD{ONqKmgHNP@xO7e&Gp!630*xJoUSo~
zc9}J!*D0v|zU}PnD#!g%`)z`@6({DKjP<wkAEOfj?bvZ>3>8TX3=g%VbcE$!%6y`e
zw%8xnluFlTWW#5&GKH{iG2U)U+d7$0HzmzI>6Ba@Bl&{GqNR}e8t8sz7ig?%JT-1o
za07+39hP}R|GccKdM1!b29N7sjP>XXP5ykpikd=<$kZB7eS6_Ai1ro~N-Z>@+|3*>
z_qt55sRYh|m7l{A-|RM9=TyFD&cRzJZs61{Ii_+R#C#jg@0D2;QSU@S=fc|%p&3V?
zZ(g7X7w0;URnedHsbyT3DZ9`K`x8)9uWH&^eV^#kYb!i(I=Vjg!ACta+#C2!%W)c0
z97lXx<`E$&tuXkw{X?g)MkVS7gjBb0whlL8X*TM&KISrO6ds;koot_JuHcbWWNS?u
zCYmm2^+3>|2*Cb=!jV#lJ}@xYXH`|3NIQS#^GP<zJD_{U7ev8)eB}@jyepP>D11T6
z-Z&=l>t?o&L=A2qHHADq8y>oa@~-#vDr(NJV-Xv^7eC~GpF__xmFh>t2~rE&e6FL;
zD~$O)-}nA|o$Q-VEnroPwXkhe9Ke<E^8{Is2D~M@#9Eor73}f+_}hP^l*AuF$B|70
z=FxJ>z9(v$UkFP6J_;sPlLDcL%JP3|s|9p+;$oSij0}C4GZojidtSor>WiB&Ou*@u
z6Kig41IJU8W&>pJn)Z9&a|wp5JNp)61`*u!vk%BX{tB&gOI1?U(Sz?D=k#&>A*_2G
z^#o_z^tha3>WK>TxPw%`yI}hmdekK@$Man$k0;U5?kT7-`~`{56wNj#skH2+Q<LX!
z;3A@{8WF`143mnS+ik^)Ivs0_n2X`Lmd;##Ts2S!YZ6~C%P={LPeVumZ3hv3jxW2~
zd0|Cvjb^_~ip)ut510A_360DzSh9_Z1f6eWRSh9JWicDaUbgzmu{7|FDu(*<*9*r5
zBpS$q`w4~4ASB(U`E}*;P>HR(5!jsNQ$^%U;U7QkW7Uxk;Hdh*c6_BhRVliMx1+%Q
zK;dqJnZeO=uafumPmhNfrHY8z8NM|6pich_f|;69rEp}@tf(zKPJuR6Pw!YAja<8M
zrPF|=J+LzxKa-Oh*9sqtBd!P6b3P&HpY#&uD_9YM)w`MPQ#}CsY)u^dIalPO8s(`v
zO{)7SauMYXyx;FAZGUa^S?MbAyu|pR&+lo+$^(1Bl&E`?{$9&Z^!vh)7_P}ng^24w
z#EE#JDy5H7xYDC|qDE7)@4uzzQacAr>;1_vK*bc}5AttKO?p)y?<;x#iqG;&6Md0w
z&!N2TIdfbm%H0|zE0JK@doBEfx;j>G`tv%8ND<lbr5F{3+tw23Y8$wTM|Hnbttq>F
zWdffV3cH<6`szcCM^SJyiu7yw9+wXWWiltfh)I?N#hMTeq}T=RzJbgonN&(#AJc3-
z-9d@5Y)^f5x9HKx{hJSUJG69JK}Z=$tt`F!?s`|?haatX-C@(I{@hX9$Lj%8F@tKB
zH>uP>XhZT&eCuRAt1j9sx0FhUQj2L<(AOF7=$8gPYMHsd$6Dis)ADpWrMZaca<~R(
zg(nrUBSj_7t!NA@*24nj*{AH+&PmKIE&whl3ktZ4gz35-jyq2gPAfDuo~iU#IMRvZ
zsP$la=o-DFUt?sTjkuTPoXG8@y63S*saK<4KH=bW2l5Y`bb)myx5Eb%txcE-p%w%I
zZ|514JdcTe9zWSOOoLF8i9HZZ+JKp}1%TxSokbXLm{LDbnQb<T35;$FJ&D~aRq04h
zN7q6Z)?hvR*;$?I&Np(dzdx7sSPNpug=6-^c<bfAsmDn!+PgdB-5X<0|4yZAGIBvU
z{bK;^)?ijlxRTkL3Q5%1yO(w(rX%o72@gSHW_`?m&aT6BE~B+IwIeG5-F-Z3EoJxH
zD;C-=5g0P_(w!bA?Wx47DOm^T|6>JZan$uc1Qtp3QR>i)Yda$3jM`pFM3Gy@mER@B
zn$N%rYND@9VZqX*o^*|C<XBG9wu`{squ7OZKULGTY@9uXn*3a8DGk1&!sy}j0p>9`
zn24@W9xA`pt_v(6KCYwjPh#8nzK19jWF1Ju3A>GILfJbCJP&+%Fwk{kA1v2x{D)y6
zYY$SsN8n*e_+jSX>t|Y8jdC!Z=72V;svuP8cs?4D<r36PIC`SbfTaoNNx{+uLxB|m
zO&z64TnNTF&$TPL%RZ^Z`J_lgt`kM{t``Q)pxPmrN-BPKA*2!3^OKs`u5BBxta-n&
z)-+ZxKt1Li6VzCF&=$T@jfr!AP~tORCaiUW&TA?LK{)U%Ti8&nc%!abZR!%{FyUFW
z14%se>$N(1E-gqQ3fMQrNus5Ovz<K6hv{&iw>dQDQ<i11e<abjeilg?p8OjB(&ULf
z-34s9j{#V5uA+~H+S1)LeW)VE56X>@dg&oDn4C0d*e`NFAt!#A;()EffoE!yR;3u{
zk{5_~Cr6*?ssc&;T-QXs$>5h4F4aGc9|nl(FkAo~L`BZjbt(C+OS;(`eRG8h{o3;R
zocX}Qk=)sVV5JUTpY2CGWG`=+?F#pjWR<pn_;<_nju*9Wj->D9jpSWPjbvBVCA1hG
z*=^nM>>Ho!fF0Pr8j~}ss~N=Oebq&&j1_97^(70s%n?Zk7*L1{^G9uGheA|bc!oYc
za?ZZunSKhoNyhQ;;Q$lbx3gz~l%;`Sg_1|B$@THQ4GE){#7n&*K07^{mkH%0r<D7+
zo3(d;zPQ61_Z5E@K2wpplm8B=<%BNP#!O`ldT2iFDI)RRH<CV69!=<imEuSMAt=L!
zqrE;ZSM83^*93U<yFae7ssB)h&N<tQn(60UgGcJRXHk#0IeYFkWW;$~-bYAO^!%MV
z>xK1Dd1gm29-p2yd7f7d+vrh0(T%x*YyM%=HZ6|7A15_6DLamTfmB8)dEH;De7UFG
z|5yba$AMO&B^*rf!Q|@=WOQkB**9L#M+v4JX9g|<BbG%069NF84t+NX%yTxxy2?#v
z{AkS%0(K*GB3**<vGF9cSO270c{<W6Zgqn_^Wt#IlrG!XH<R>y=@D8O?Fe%;Ly}Rb
z%UWBK!i*>$`<R!Tf4ckpLQHGyvrq%+Z6l#y*z#l&FL8P-n;5ejMK_?exqhuEGu)iK
zbZ$JfU(%0cAf2sPaKKA8nosa<N-RWZVUIw9k~>>{>$UUa)Ohu;(z<-G;(Xg(xR$=S
z%qm1XZd}3nQR0pI!SW%6ptwr|WAE;sM;V_u=NtKjj9U$$w9Zz%)qb5BwyiT<Es<n#
z{(Nju-y2^8Se*v8!mgQQY><TKr=QVD3@%;1B0yj1--7eGV~&@JJh}W1`C&pjVZu}j
z7-#lY#b8|!(O6#6UC>Sf3le;=(dS7U)MH)(CEC877t}<uUE9lLacG2}f<l^Yb9rsb
zlrlUl+a(UrL~}nXEMUKf?|cZ|`A9WQ{aLD1S(jg3G5pT@^!(gew*F|l0YQkZT!gMy
z7D|tAyWw`<MLRJYf7Q<Tl^(P^NUQi6jQ|gNyEQ__h8~YaN=<b(<lQ-#bVZW*MTR|g
ze>=v!xq$A($kn}Z=A1tcm1b3qu8zH7FWN;SctAZMjJ?F24^%y&fl6!yM$F;q=$Qhe
zx{aR48N$@3M4Y7Y9Wt}-R#BFilB@O?xFO<cF-X=(RO+Y~X6nGp&f(G;i|Hg^=et4h
zc4)A^tV3W_F}8&;`k7w1U@mySb?9Xg@SZ@t$DhT{q{GQqEX?PQSBSLYUnI0KkDNYp
zeL3>IdivF%4pG_<{fa$x19q!XKax%>=-B?~+<eQUUaUm&Xj~g13^O)ekrzXgr>m#g
zq$Dl0`G0g7Qhbt}$a$cZmK90AKn(@0#LkR7-J8z4z|+v7JG|`)Ch5-gWVxrm@+F^u
z5@5q#9O)c6&aC>7GhjyjC~BKAQO*|6>{jnbIm|(y(tREr1{i$gR{Cw1CUc$gmrMh~
z%v2_L*gfj+SqEg5T<F?y>$)mO#}D_hO|hPR^<HHBEN+#Cjw-ydf)RhHs;qQ3US^I_
zuEj|Kz5uq7bZ-WLEoSl>kCQVa)o}ah7aO%ZpO=g&6<3HEyH>n>%S!5caG6ssk`Whs
zAypkU<|KOe#^rO;S^aDu{%{ZS-WA?(KOLds{l3Dm$=&e;!{Y}UVDRjgWNb|REi%qH
zwApG|!1-A3lE~%w_?@4A69zEQThn|;T}RJmZZvqu-|ENVK^gXU51P>50_{Fxn}z)v
zbj<qc7!_nz`MtUqA5vNtc7w_gX?Q{1BXgD<D^}yEDXH^PMzc`S;XD38;dwmLud~U(
zjf@ZNE;NS{pT3rI<fbyb?CvATTr^4l6$?Xt2@nbW0>l(?q_uR+M(g3*poZef<N?*z
z<8$B``TZ((u@ZL%gwLyJZED)g^ouF^w2sm7HsJ(^3#C%Vu6|0Jt`c$8IKVsz>A-eX
zFTr<&V0%g**fs&lhbs_ZpuSmH0=N=Q3A5v2|2i+uRq4le+vB9t1NB3vu3cHPz@D8h
zUFaxQ*}+nKkaHIHA~nyye)R#+(6`SoHS$a@#gEon#=t7<I^ZD2C%lmxvB-SW?U~~i
zVq}^v_sedBNnxwWf=-&_$CHWdVSQt37+LB)uFW{5?GvG{OQ|b#If_IJQ`ZD}aAglu
z{z0{)gLt<S7)CD*<H4G*97pct_Jvk(9q~55S(m}kCa`5uS2{s;)*vQ-dAu{+Pr-~$
zRCnn}GiUa3W0O*F#L^i+K*4JQ$hvn8-d#z2p`}sDZK&gk<9S2v4;O0Vg7?FwR#Isp
z!QArC8*tZAuEq@`emZYfR~2X7^HPl=@*le3xfpfe=0(05CxE^<QUN<aKU14{o}922
zR<TdCx~`eFJgLeCn90Rc^3;(T6}Cxr{0m(~_Tnb)`nn3gDxB3*<WKs{w_>yZzP=3A
zd(Hasv+yf1u|e8|&2>;N4ucEn>rJYSJ7*T7UM7b6J>1ad$4>Z${w7$?K|#+7qE*Ky
z=mY9JlbvUt^D|OCc!drsOItRsfgJBVrFx3Ge2|AH6O1QbgY;PstCI=&3*u!Ni7&%O
zgpPY^R*jk;c4&~Z^?xPA`I=?(6N4Qy&FltRLkmFq7#jYG8h-ATQGh5{P=486QqDr?
z>H{AG>8^ir+&R3aBh(6caAlLr8kJy@^qoC+4PJ~@n}+=|V``(%Cus5;PaST&d-v=<
ziSTBjT^O)#CR%ntg)oET4eQT}{g%p8_)qhdW*Eikh-h*9jN-Mo)69F|whfiT7oxBS
z`lOm(CCnPMY0L)xNs-x^EsgR6MOcu;gdLjnNP79TR`y){yw$nx2Sz~xE?BPuhvN1e
z%A~;$*_cp35BuFkj%%Kdsa<&f=!c;}8AV{RijTA0(;>q4XnJMv1I|IbJCM(khEe36
zWADNbxHA<ZH>BK;*SK5%?!GAawf&)`-f08a`DM_CTV5D<bn{AOMVdr?%^hx^l~CIv
zKLsainD*eT&Ko>lfJ2B1Vln$8igQ&qX7o8P{Kl~2O=feM{<Vssj+En<-ky{~eXNC(
z+CQR?Fk<f4i)F0{^vzj#b(;q&7RyXb9QxXH8m{0f3LbNBdlo(}K+JrmZabuoms-?z
z&dPVMti{L3at`ey*by^D+%i-@(Y(Vxj(hB+Of1~^Dtl?~j@_o1pX)4E#Nv2?USnaA
z*%8IvILSnQ>{K(Z3_J5m8y*9ct4uCbrq1Ug-@9E2!d>~4;J=Y>_FeFuR<9|^09G-c
z)R4Eh8Wj}8o~`vYB8MlQl&bb4@bakGXvZE;<>l*!hwlgZAYS3g_BhZh0c(qU9+9_P
zRr|y&=f|1i;?_3Nh~&ib=+qhzrELJ7QG#lneK1^M{^OiWY1xS?JLn0Ud41E#%O}{j
z9|O}%?k7p5ce6X<g8X^kpKdAd6%?QjYlDJ7u{HL1-<8UmG_f(kr?Gku6oAVSk5P$-
zJ*)*EzTy5SL0kVg$?16`^=Ru?XD;hA2p|9}L%&wM0<YLdo?g)o+k$`JgelK2{d8Qv
z{GbgKYy1T5q$r-Vo#Fc2D5wnu8L5)gdGS+popi~<=a1X6fF!mm7IO05k*fy)dXxc5
zJ>~FL74AthWD#o9VCY^najpg*!_7FS4$gPiKsl%JC>k@ew))8PH(h~vERiAJ+hHjA
z%S0@|Zr?(QLRUQ$=;$D*;{}1asK9F@NiG|c(`Qa#AM_0ef@R5)#T>x$&~1Dq#ZBVs
z;RO~)S~|V35=nay&jF!lkz%HIZ@n}zMDzg4C6MI;00e@4Y^?&$>rv|-o>cuwk*2q6
zGT30ZV*_Gz4BC*9xt(SKw;z_Hz@Ld5C>N`|?b1HL_!<AXl}*<_3p({x;UzGUuxSOF
zBlSFaR9vfORA3&$Sn7u5iSPTepYWxw{LOs+!@QTBIha?Cm!fuGkmjL?E934RchX<@
zwV8bQMcP!YC7+3G?ZK><3q5X<%2PG%7zo;kdW=G>cbKiqEv0~fKjED_seJGI?Ws<!
zK8tde%{BZ3g7b;SRFkQ`hiz539>kzXGgbiQdy#5G^mN<yJTan1#URcaA8fL~s$0uO
zol#i$#v7VZgL^cjzaRaieu%X2`7`Boc|6+%mzp=qgIb*+0tmYAy)G$4E{si@_#M3X
zJO0$&FUzX(F`ZH6=4Ml36%MAkhR;n>&T&rKeS?!P#qpvFTd<$MY3QGRzQ^lIbEbUt
zIM}0-bdmI{rjr9N3%?&Hfr{<?AMCw(Jk;;|Hatkm(nOXJBZQJvNZCe<EorDw#8e_o
zXcL7QDNEKQgrZ2cnCw~K*^}%$WsH3{V;N(XchBYXd%pMc`+a}+^Lp+-p4a`?{YRe{
z^C@QD*L9uOc^>C+9tSOz9>ZM%WuNAC?E*tk%_3KBdLvG7BM4_py!sds<?XoR*Uyk1
z8xJ2MNF-NZ2M^xGRqScTg4bw8(r^~>$iCn1iDPG$2v7f*OjW0S;4E0KwD{ywD&XOj
zWixy+P2to5-a5VBENmb|xza9%E5|Ip95eNi!dVoQ+A1HHEiHeojw+Wz$N<R4FeUri
zTt~u}yj$fVQt9Lyb$ahCv+3HP0r>?x53J<N^gjAMBpjt~`zc=@`{{;G=C@{D<)z^$
z`bG@9TVMQ=cgvYjr<dzK8qu~cmyb`NE*2B_g{iyF)N}@&<$8tNtf1$Nw+q?;s}mbS
ztx<M^2P$M~d#8_#rNzBJr49(fAx8Tab2%Uu_mFOMp6*Hc$uM>PSZM2pxr#_gkNJ3Y
zICF6osMhC<9H#e=s&f9(Ee-L#nKo0(w~<y=X*_`6*5?$=H1<7A)#$RFMeo+p&1{?I
zldAqGnXl*1h4p`m`c4QZiSMj7+k@@nPw+rn9A}M;Husfb#eG$%%Qda05`9tG&T2<?
zlwEm*sXg6D0KV5ra?y?kk8b>^+yCr`?0ak=Azu|(eKbPcEXQ<jLOtWl&_)1J;i|hR
zd}(FaUFJ!c+-<TU-iH0NYkjXq#ECArm(_yHyyTt2{P9y9Ghu+RNe!i!(m)?E;Dt*{
zq3);Ty_6xNZ11@-xnh%_*KXfiIu`JR#{=*T!1SxMVwbP71FiLrMM<~H!zise{jkl%
z(NFBmYQeMKDkxCY#Z6FTwchwZhx=DI$C2SA057A3E3or6Z55ezCk!IOnkOe1=Q0p@
zkwIsG^?Nj7cDv-}*m`Tt)TiBll71y-MG@?5PH&GUdueJwq4U$;$ZX-;NhW<4kl*_j
z2^z=fF(9>_9zR)%#U)GhewpBQNHnq1Y|`C1GzMLT0DGTeUdPnWqTGp6EG?yeIqO;V
zQz=5#n)wmPz9{(X&47N2imZmu@HaQXNUjUy)947i3dnhG6l?2Z#v2p=NlgpQZN&IF
zhN@#}fFX3JX5<7lxb`n%*Xhx>8sf`xJ(lTqTEGXUI9S`~b|~u}0KZY<-so$lSc$Wl
zXES4eMw~J9YU+(Z;NKEz#5G#AO(?{0?#1=09QDe>UEJ+|)`{JI+SqQ-$BqTnrR9!u
z`(GVH0dp0I50COSqQ0-{rc0Bo5|!E%;I`NfytWxW76`Q9rC6pEuSgvu&4}D}Z<X86
z?_!74v;zjDG$jyhl^zXK-9v&Zi@&To*ZwGTqjIhCsVAZ`+x8drxVbCn3Y|KwK?O#o
z<JM3o%sJaNyRmUXD5+AcI5zFp@ygp0ATiz<Hn!+v#m*{HQuS;-zr=Ye)TcBWqg$ik
zkM;xxZ5!hmO<-iC9YT5{Md_)FZVQX?8~2_wn@I(->x%M^X7#0?$%hHwU8=D#G~MIR
zlY0-Aq0duRYki;+Uzb#uq^Ioc^*yKrVm&nU9OgBl%FzP8rC_?<(<xDb8&len;*UX;
z=Zw9Fk;?g5DbCbp-KO_s9ZC9#v-$GiK><lUG<S*rJ(%iYo@S@QV)WFplj>0@Z-PbQ
z*%zib`tOv_0#7Bheq{eVtIQvuC5UZ>z>RSS8Ajm#<P%|Cmsk=%@vgWvj<$2@nwB4~
z2LkGfN*VQ0fUjUxRNSL+kcA;zy^>6l+~hA<Y^N9x0<Hj?p88ES07yk@P?yPY${2|c
z=nwIw*L`GCe(0?nrRm(M@+#tp_Zy3lGyCRzWL??zN&J@w=Y^wc^R8isU9HA^u8n_t
zK>^tpvJK)#k7i0A=`TtdT5UK%vzbbFanHA14B{(2hbp)CmsqY;1qsuhO8OqA><nIz
z(l8HjdiUvsjH)UHi#vO`he#*9_WWq{{($WF5M>!sApdbAsx*`_6Ec3tBTw`&>{P~w
z24%wbkN0Ul9xwI8lYPfqqArEJV>oEg_qdTs;ZTubx(YqG-d<i)6>3ZP?0nqkAX*7^
zzKz<yFiuq?UpMi}rQ2T05Ah{A2Pt@B9{fBBYDKhJ@8R{NlS>Qkw<?NXIVuD`R|c?(
zC2Ix-s*@$bz-M{7p0+XMRN<95>lo|)Xm~Txc|>PL_?R~48<B(!K#QukN22(OkP@?8
zZgIID(>t1*&+O#Jj0qsT)c-fOQt>*38rv`KUmzC@zpLgh!}l7W?$=9xxB=MY$FMs6
za~>?=B-+(-cB<a)hD<u1)JF9PzLs}4jmP6%hGb9jgpxech7}Ctd_!KFDPA|~_Dr%m
zW+u?}cKF_C0P^f2wqrzidf>vI+_+n}(+(CZm8QD)xw2`czvv_TYp>)q+b5RWsAZ={
z%bN%bJKmqmP@jW8nrS@p-TlVG;Ml*<Z}atqo~f~xAGJb;RJvq%<9i1`U&}kTVkF+&
z4yE67xKF=c9-hCqJXFx8WU%;<hO)_-ckY8T=z`*ikz(duZ|VMvuW~z_ZnX^TXhwg~
zVD@9*LX`7_!OPe3dx`kt-YsT6+DV^Q){LjQ9W7;o{W9q1qk%*7#T@z`y<c*PQ*Xl`
ziKJaJBVG)ir=M3v1_vsZpEeM-4)9D72657sdypMF$yDWH8lpAXtx<_A9Gyd>3Ienx
zPTe2Bu)~iXsA@{6Jmn2*n0G(<l&>XIkv@7>Fws;}vw|0twLBTq*k{t*y=ef_h}6-|
z`t`WMA#?b5+1k!S`9ME_ZD?_QxvpK*mOlTz92w$<`c=}iFkT+f<mtFgVkVO#=2WEk
zUQbJTyyd1CwF2gC1#@vt4eg?I2TUhR^FUB+rU{PA%uVi;!Ab5&DiD*L|MGIm@TWKb
zF)TOB2wZZ8-g09`QN@6b`iEY3rKGHPX=KURq2x@aZ6=Q=!Z!+F%}HD22Yn15fJhEQ
ztOqeLN`smP`&WbJ^!ICM`lKJYFm{1SNzf!KAp&OO%fE;Q>o6EJ-!EW!whjizI%(f5
z{$P)W<R9JCoo@HtvvbS!GhShG5!dJ_aLyg*WmLH_Ai|`hnmKE~FS5?6=T`Cik0sI9
z<@QNsJgH$RuMeG2l{hx1ufZCGn+YoqNCIHWLc;@CV_C1o@-W^4PuMLT12FdY?-L_D
zqiAUy&||f5924NAE_j-13Z2!t!)<(u*bk_KNY*Q8CjdXLfsVQz-VxsfKFK3YC<#@m
zN=^~9P0CeZOgbdWG<b1-_XKtGT@u9tcwZc@Pm#3$iy+m(>M0cJaujM!`#-+bV0<D`
zOi_bt`Es=d&ysnD4**e&EHoKV;y2j5K#O|xI>fR>;Ei$EQefiadd7Clsme9@xm0nd
z0a2=Ny?wXunf@W>mGa7d4WG}vPHz+Bfm<dEw-M9??|jbG8p>SN;XiWK>otuqAzkG`
zUS4RXG?Crw$iy8kG=aUfg+^0?^b%*TRXKR)EGV_6sYlAj&6LMb^>!+25j<uI?uXY}
zitFPek6MPzPOfhd2aMB<9gN4#qd)Lwry)2xc(0CM?T}S6u<*}5%B|(k`@c4@lJ}g9
zU5uq#kG4N2Q<?&O%{)KPP|vR(sC3uo|6}$L)J*-X_$~9~{#9U`2ij3IVY(5upXS!>
zYd!X7yD8fhHy>i}`7N5`oQ%?U8Bm%;LF93gG<xQ76%h0aX1Ve$)0K!#IW3RFy;m(t
z73pJdf<zekV6pPqDbS1+U_Wq9f=8Wgj2?}H3V%P?PS|O32+Y=uoJZ4K5iA>O`=Scf
zdhBr`!z<L+D-YMFx{*sflF$JQlL7#G4@JVd^ia}yO62p0e$@shc_pYre=yb$XLNHm
zi~wmKgZ8%<thBAt{w|tzuSy#C@-|yMg!q`U{9?^x=Rku?+H73;Bf-30TkqJD8!~`h
zsLhmwTgvY&5Gl#SbscZ>)+?S*yJ`gxTh)@!Om{xeC)!XqUgomIN2_qQ<tGCgdnY}<
z-)BVIu8(6&G$u+=g~)o_kkiK%;u`b;HQkwObC}v#<pYk01WqFTH__AUiI-fQyGhGF
zWDA%)f(E=slM^UsX`aCA_f|!o2H(|n?S!`iQzp^Zv-F;yBu+1XKauZ>6)aNr3D|ad
z?wI@whL|{UHs-rk=O7(bN29!J5KZCSFCC!4>94s3X|f;?8h-f-bT6r+k-p~(LKN}}
zB$a1(Fe%v)DIte(ldALsr21JsX?)MnfsA)8yQ-l<CmqVKg-`eA8@BSi%WK1??So?V
zg>Bz4cd6<!8wv(IYs4mKpkX44mP@zl<o}_;ypE(xfkFx^tp23(G^M}cspB^Xko4M}
z-fN$psDQu&+~QcytWCVXuie9&k0Wr|v(|>E#6Fq`LiN_`QrMJ~QlNb;N*a$V_8P|q
zcMs3nD}NLJ);%Sw?UQxf<N<3@aaLafTL3(aM9^enBun{bhW84L-ag-#FE7;UecwZy
z!~e3lT&ev!E%*gq1|-?#4@h($o5J4Qy|+z0WMrXa)lgJE)Uf;_{`J{hZ%y+=rPcCf
z%(EpKu$k{3QEmf#v1Ji&9;?Xl^L?HIhiHS}t|Y{}NQF+L-fm?JW5ilCzppc;!dY6m
zJFRD}(G9-hFZj+u;|HCKc9yP7X}VO>Q>i^FAbGLZt!Zg_yp;1M*viBoq!$pyN;Mop
z-hQbu97d41X5{}B%dwMJ+ufj<*o_$!qpedgixxO`3I|+IeRQzePvN2e$Y%w`BRqH|
zu2}{P#F~I!LbEm{m`xxo5V7gSo9Q6ZW`ij*FEWw=gTVrOf4wK*Z~dcCfdNF<O%MaQ
zuqB@K1Q0nq0kP<H3S0wlyP56K-5IXcZkTUBRS$#o73-`#fP2ulp;bI~-DI+auRiHb
z7EQkYJD4rOyof`Kp@9j`z`SVmpP4?q<HY>_y}v_j^r|*Eu;&Ri?#s8aA!vcz?Y^J8
z^5oGLCzMyAniG1;mDSOe2R*XQy!LLk#TAH|`#lD&oo+@ur;_Ds4fCUjo6k&XZwb@c
z1kxEmpICq*2&>cMs?4#esQ>h}KRWG?dt1AG=klZZC<SNtkKQ(#G6CYoa4~BT&^?f3
zx;1sUab!=a`BxX4u<xNeWcay;=WS4f1c|?h>F!4CMnmhv4d5w&mz|4${b$76x?a=z
z@`+i6#Gg)&^*%*|F8Ly%MjJkGopp&F>w6#@d$F5zM!qn{$I<`gPgSzNGKeJs`Gn?N
z5(srSS$&xxK~PZ7y>puVexiFl;`pj*oMaiV!4p2idNGH05+Jty*HI1pao-)3dOdY|
z>uB!wF_B~`BdgL9@3*&)-a&-I@kLTcbdhJfi)PH_tsZv{rxNuuKk9=*Ntf{TNYr*B
z)5?3>MKzk=nS<Tz$(x~Y$P5Auh7~{)gL?bH4JQk_#s<Yc2HwzOEP(u*sYtj+>>vDo
zbCC`bfdgkj(^aVS9}m{<O7vU|yA%7%s206>NE)b9e2s<Ra2xnN`9EZ45L+DBfJ)!<
zDGh|8EJ};EE-_feLT-9e;E>!`RK(L|mSD!HeM*+A#72GY8IiK&9(H8iy}yXdGH5mA
zgWS+fIhc2k--(aRYxEJ98lfloM7?JS5r!Mscu(^K`e|lQ(j1>C^dAZPx)>P6MS+Vr
zHnt>dqz#yszGB_gI=;K9szIu$mv!P=52V*1iiCD~saA8KA_%AAv;zABeC<Rg-vz3w
zSF}=%ckS>dJ>Lw4Q;7)NshSxFk4fw=bOTP(Pjk%ZrCToQ*E7|rArb{p<GN7aiDFUG
zzEYIO<&LYLE4y;j6J9p2^5Y<MByJT8^V$7H>~H5L^p}B=%jD`&_H+4U%%E3I(CE#E
zuYB-QZ^J)LpFF^Qn+|~H!KlRyMjv$wXgK-qaTzBJO`i#Z2i;ZfE-b0yJ%3!2GdE{H
z!~z=JAMmLiI!C|JYsLgjg4|UW`JL|%zM0;%UWoV~8jvEaX?O&+;+-N2E$C{l5oc<l
zcVP6a4#w;m>t_dlhTm{78o<v1FUFVxroj9ZfxWkvYRwXr8y)sG8rO>-$JyTBo{%@J
zQpS5!!n@`OK0|{B>{TAZW~yNIiQvqyIao!pMEFKmd@a~-*KO%;t!+8vDY261%x3Kb
zODag?J|IC37;ubrOCsaBuh}Ty{BxDbcVjDU$MK8Rz1tkv<Jiv{EIK$M)kjPz!L(FL
zeIrJp+dAP<PKF|fLv)Y{OP@F+i~Dh-5Y*BJ<%W(fYK<LgDW8FMH0UAm8Bn9$*4p`e
zPZfT1&U@dM<{e7H3#DH{xz*-vSz7al0Sk+iF%foF<lLBL;nz1&RXefZ1Cva#0{CI<
zp1kT)_?{7kvWSR3m6}Z#zOKNxI?`T!W3}QZ@hw<h2f8u^RjVhYU$KzO5glxG=RmXG
zUxWq_d<Dm_7AzXFmdMzw25T!z1li?(QSK>T!cwlQZUWz6@TiiO*_?+cq<<y^;xhg<
zA)t>cS!aQdQFB^P2}8xu(x{%lW7P9!V@hICIvjijRBT63mbR@+>1D-{NQIkGH?hC^
zn@oes>7C;VDM2wd6%HwvmCbqPJy=3~WP#x4)jP4?jdQ@y&h7E|BkQ|HZ%fGf1G?ME
ziOh|~qoA2`<u9T-kf3fa*@Zv!UbN4Qx4nVqe6*K--?8?I0!H4*4u(|<4ciE9u2SRK
zR1i(OtcPv_>R}ehCwS>CFmDc}x^4`)(ETL)lX*#2>cvuv-vJeU2l(wC&UlYOGEf1m
z0e($NI0S63iGE}#+$KF|*xn`iNLlHto2hY<LBQP<5LE%)8;sSqQ&MPg6nv4wF~bei
z?fc1l&!4yfd$Prnahq9mA4+_a6@dGW-Gh<+g0j6T*Ym%=;dpsv;(amupsle}X{e#y
zm!ErypMm<WC#WoTdQj@dJt8zPB6H=9kV~$sK=n11j3^hG8~4m<oK@4bd+d}o)L;hr
zJruywmG2&C(1{n&XylJv4u5UbRS0U@p`;pA!Sai4YpUD0U+U;WQTL+%lWF$^zITf}
zS&P_stQdOqYl={%Mdq(${;$Q9+|bwg#yb*?q!OO1@(lT-U9R=wS2@--o{xE;=nrd3
z8pmz>zN`%xX=l9Qh5cA&%)2Zv`bWiqWtL0k98!^!@2bU48i%lzM-U#tdPL9i<t`2O
zz5Gbw6w_?s;TJ{p;?V|#51`Q##>4{j6n)*hJfn8tXo;%^pPO(G4<w0RIrQssOk^co
zyo9Elo2OY)ds8Tt&59HL*56UmXeYoq#kZ9g2L?yvF6I(dKiGzJ?PPTqt^N64Ggb|5
z=m4d^h##0nk|o%@U<N<G{~tO&4~PY&^x*&C;}g!tvcM6p0dad`5e8d+;5=qpAN2&h
zBnC^)(m<AXZESW?$n4VG(n-L7*-Q(=ZMW2W@eG@;K@DaJJ#JGb3}Q@U#gOVNVtU^m
zb5^&5nfL94(Hse`S#$~V<&oNL0gREMbKjTto0D?!tJY&+K$8hyv>?p4q1hbHTywy3
zVte|OuLM;jlF&PcRM4w1z=!ou0aa@pY(ff!<|n{$Bz7etlt2RdfByXUXP`_)!+eJu
zAQwLrdW|cDFkwU-=4D+3>;IF`+%f|sZ@8ksh@vjS4=nvQ12omWT0W9AWCbx)=GV$0
zb@XTmY2_j;zqR=NARDN{#{;p*QkEz%8=VFbkS2{~3$Jeq0`qlE(jJ!IM*t#N;Udn@
zozzrnYjEO7qVQ-imO-=xP5%GHI``ECzb&V8v=5}jQ37Z0OSJ5JY#p=cVHxs@poL(E
zvy*@{JP+K8-%-$6B%t-%)l?{VDdX#tQq^a#G1r}{g0{3E_+Z`*=J6Q|Z>{dJ<lu~9
z(90`=FJ@9ccR$#d{qf||P7NwB8^Sb+CGTeE>4@BMW!~yrGI#;coW6_4VY5+`E18!(
z1}YkT8Xvbfj}N>A18L}+gNewYVe+Gpdo5oY!N&J5Mi2~pHvwwE{j`W4HooXnHzLvD
ztC=-I$C-uvY4&n03pLqyBWwEIEy$w(m4Jc9Covr}{s&RC4iq>m5dby<Y4a|$6zg0M
z%W>RJ$f#RK54J2W?(yC}<95SXS6`es=sbSa;cDj@S^w<pqX;QP9uRNbt%IcGksep&
zIgjj(#J=Op8ZTW%W^M~lq_l6bKuoH)(S=QDDSMhT7k~!!^)ut>3mOQ&Gx|49i-SHC
z1R_;oQ*fo?iRV<F)7`0EQ<a1Q(t&brlL`4(i_6bK*V<8p8s|!pjGbt*Af<KctI=<`
z<FZFeo*GsJ-;bi06Zhczl0}Otv~+Td6xW54a@%y%YeSc=a&9-}nAewTX9@8%SSMO9
zYOo^&TAEz4ubsaWh-p!b3vY>DFvF9qM<JF_;2^;b$>)Dh;^!KQN$F*F>Z3qw;aU)y
z@^S6vFaW=%aN{Vt{Gc2yuM3(V@nDPg(mYtA4zTgX2oQllxs*TzOyU^Aisz6u+Xrw6
zQbC-3{I6qv{vY%BbbO~6R!mRNo(>k*;)A!{P^+64elbt@>l6KTnK8(n<>ott4OnOP
zob>V28*|i~mlR~>Y`&&$grrlwA~mX9CrexWp2#l^&Y7C?%wSnU?X>+}u&jKFu#RVX
zQ>HToici1#M`0$zePx^-cinovA@|FP6Zy8b^<QgvUqq_!wFI?>bnGVfpP9{zuqcZU
zSeFsDorulWDZe7HFKXYmAg2fQ(|TWU-30OqiH|TNm_ryn1?qAHbN+hA8ql+!)qHB7
zb4LJv?<pSY<I?}YI3rW83qo%3i?I1qDNHT;3Ya%auV^CMJ@nb+)iXWC!FA{2i<qc%
zKX7`HadcY%wnc#daJ19=+t8(6EMYD(2@s;L&^^v|t1hnpSiYGop0|JKqe#K{sSmrl
z;T_}5vvx*V^Ju<epGKSlrg+wQF=i@eHM9M$3jQmYwg=~vMU}4>W9R9i>>BBY72Xrq
zM~}b#eWQO(FSqRUI<QNf4ug5Q0f*-@kQkYPepcCn1_)DT(3e0Dyg%k4GCS`>7T^zJ
z>a>(Am#v^5)qO)ew+CHjZcY#ivVdCYBkUyhGnQiH$6XXP_>?wgP#<63)4d(^99qJW
z6V6<Vi=cF2X{Ot{BwD^ZAu2jwAn7f9NS482>3fZ=BG+V(s_fJl-r=cogv$TWw(VK$
zGsZ?P2n?Im?5}viAzvZ(M;$EL@7Q@4(f%mf0r%7+-{E3LG5BOJsl6L7Atzer(&Xx`
zPWKDuLJw-}629WdM7nZZ%&r8c`IBgVjotu2>RuMR$yS~t(q-S*wNOf%EHwW_!X%He
zD1B5aqvS)PRs0j5{TwRfCvQtH01-3g%_??Q^+{I(3RVZh*<u^2Dn;@9apuPk$h^2G
zVs!HrW*l1ql|U~)14CzM{Ey9u!?v&2ZWw25bN9r~W7~u4bcMas#*Bj38Zk%hNlk*-
z>%|Cs74Rt{-16i34()nG3ITHum!=!F;+%@bqJx<Gu5F6z@*UYp?67H~xP^54%6%L<
z^g@(&^qeri!(CT&`dxFLw{Qmu!O~=&^F7%Ev}pO1oLo|Gx;&cW8#aAj+(VNA%T4q}
z_X9s8Ym1KDN6d=%re}LbPSWa4(Xso7Si6{~iL>PZ5+RFp;yXa@aQo`!(t|F)1kGdX
zisI@VwfI)O6FrrF6xABWRH=tbUg7<W#{RFrj?OBa+71b$N7GV@r23mMLwuJ*m4gUK
z=L>h&=qAic<Wa_h@BK^h^;I}{cYCm&Os~or+v4}{UVrXI9OUYP5>j^Sl$VCCE21hB
zr)XBv3vGl(f3zVq8i&r84x@@Tdc1_Sy2UPa?|)mDK5$)iY2+MQtvve2#4X59CP4e@
zrM=M(p&GoI|A3UMs)K~Ouin-)AG%*g9p`szgJKs;L$&=cP4_G(u1g?(0nW}NXs>gr
zjbZW4UFo0g_DAAyFMda6<crW9!KRgGM=b(h_K?1z(jd3n&ugAenU$BEsJY*wV;V)~
z+Djbar88mPlu>%62~!C-^t`q}FVbpqZ}^)7u@BH%2s#Pg0cdUp_TFA&;LOyiT%1iZ
z>XpUiU^EJ36Yg=UeqL;ZZs9hI*wO3EC;_+&+NCP<29#%eDz3z!>Vx$60=mklH}_?g
z5B~hRC35~xVg^kt>&X)aE}f?#=?91lG3hK{>Qwp|v}ZVdBdAhqCc;r>%w@p+0wZ(i
zp*<RW3UIh$D(TT6x|%4jjl(+uzVQeeOec<q$R`F4KG{nUyfml8wfINVthjUGMnQgQ
zWLMX=ox(>muNX)nxQd^MEqYL}hIBFzP-aMHIB>TP4K^Q$sDBw(@)jA}L(^F3y=)cX
z)V0*vIO+UWkPdq7ww`XfMel8B4Q*2;@5zciIf88Ebuim%%9L<^NH?fHNN+L@wt&iC
zHKp5*81B{22~fa3d}lq`RBymX<*NUAlIwCg7Yr8@&H+cgDA6$H5te%Bftr}J#R=@e
zv~3+ch5o{b?FeEJVe|`%(-%FC9RxsC<J1q;P<64WALZeQ+Y#?AW7+yr%1civ#Nh{R
z%FBaEjnNk>x)&<)mv({NR|4Ot2NIkYrnEHr{p|RG5=N@p!y4|ZTh)lP?XM#$<8K~$
z@}mV4VJAHhyP9uPBGJ<@dLG*agXk-hq{<jgVeI&TY(>+L@UTJv=bo@22eRbUG-Av0
z3xa^BY8Z&|>PVj1LuO04kK%p|$REL6Uc_diDc6ykf-t~J1`NTmGBC0i2f+zP`C`+)
zD46GJePaIbQ+#CCxdpY7%ZCs~)c|nFK^HbHY>+X7{ohaEQIx(UsV|@9HA`)ck{&AU
zD?B>i4E)r$d~?)5mJQ{PMEHyj>M46;j7{xa`%bg-;xa{?LIP~=tji&+0fmQx0iB6{
zhyx5<nC-g%pcydga|0Li6>q1v^(g)p9fKahXy@nkDce|X=><EvM`G@acB#GlNSyj9
z1r)t%@9NnOxXST#e1h1n2sll7#R%pd{LC_;$5Pcyy46$o=VIts!xG2U57+B-!q2+P
z>4HfnGrjG&E_&MN#`ATT2{p4PpX=0&2_dobcAsdEH?Z>`BC(8_XAl?$HU@c)BH^~J
zpHID3Mm<_$<_p#3%l@v#rL;}oQRBysa_{bkvLBjnt^O#-^*dpDirz@_1-v~pCM8yP
zI%qv;ZtNtVPoTV0rE(*Qav23WW+Yma;BQw0@5Xq<xAtxUebnV<i55w@%-c7Eti-oT
z_tJa88Sw>UF=$G6V={`G_w1&0ymYnA$k@;y&ibwohi^A+L_Q(9Q8!qk@wHFA68<7~
zL<Ju44EJka<h;?|FHQud(_Pi!ih4rH_LPqUt_<$D=Z>GQn>^fR6{DVT$2>#+*?>7u
z8KFkZ*TRRV2_xj~d;KH;%wHZEBB2jP+iU!Zx-P+vd%V(_v~5xR`#km~nqt_f%)H`?
zc%+9a<AIrxkUO~H8rj7<)?VYtp|CNjUh)mu-su2&_biDrg!KaGm6+Dr3xn3+LIznI
zv^#8v!>Z0J#G9Av>kOvvs+->63T1&3JRlYS^J#k>GF*!oRW!Ua&sSyaVB!)kv*&97
z`Q?#Zc4v3u=KQGGq9YYW^Qeqwq4G@J9JRT0!$0U;au|4CN&#TCM70<D8`|lQHZ7Dc
zmv{B^>(Y@!g{3Y<PhySU7SgUW_9fEm$QYJr!jQ9yj?XB4aj^%n!FZG_A2IiM_jcSO
z>ID#>D6PSUQPKcNP}()=QL(+K*=4_;zuE)fUOzml%GyI3-w4Jlc;o7id*2#AE#Z_d
zP*+vcIAOwZvu0v95W8wFeC?vL-{*y3+Yi3%HGxP2HB3*g1`hPw^KMm?Z1vmg(vY{b
zdJ_@<lsygJKd0-_MGDcMO(YFGKFv1nX!Q0iyG6?NA+ZFX*Wv}^Yhb1R)KI=C(T3v&
z>X96;N7TL{s=nXsO9Ff2F*|lWMcgRe7ei?ckB0^+l6L3&{_vGuYN-6Z>-!DsDa~3!
z_{ed(<k;KwWgeWS#F2E=&xTW0Qac=VQ1#|KYvb308$1FT$VOGpIWVCw^3FO=tfPof
zso$s|<3FDqu7-_;ctPG*IC8kNp(;L{rF<u+E_%xK=&xp_jR9d^rYVq?5pEidZ@3dW
zfGhBRx;AQDUVm#uW~9v)(BR*Sp(tk9NYn9F8_=2Zo6JkPLy*a@{w8-f`b%)jz0*qw
z4W=y<J(Da$A=MCgX6+B{uQa%m(Il{X<esCHajR#t6xAwX3{W3u5_6MtaZSr#GOk@t
zewU?@^FK6aQQ>IHHGm9@t%3+b;{hS??9=rqx~wT`0Tpt#bQk~yowq7eU_M=d5O4hJ
z@0@wMPN2(TSpN9^g#HOM<FGl;obG>9#I+tNrT_p^4K`~s0SIy?_)%`4ijwA`#FC<X
z9aFvC>o_+oD?{6a>SWZunqx{e*aG+#R6*1Ww37#?3j-hf`TK~Bvl!FJ(sBE>bQyOJ
zsC?IbY&e#3(Qi9yH@<g7ILNurGT5^=NJ^ad6^?-fh=e9#l!z+4_&n417&i2&-5XJ@
zsigb+#3DTQT2+|rMYKm&+;8fGjB};{-^J~nHK`yaEG@w+^U6JBraWkSJpgA3Pp=)X
zfr{h(-ET$5?IiN6HNE3GD|wHw&w5QOqDrL75V!e-9k1X+7~-lqZuq#Il!qL}F))Q7
z{9Yye#!l*^fL0$6)fkeUuhSFj(vxs5&E+5Zs5TdHqVBK*&ae{LoFU=we-Q<rpkkgz
zrNhf_G6L6L9a%ro$R9l|DZJJ!Z`H&S;cB=<&!Lr5<-d^1OG-;i&b8_DzRlZ!cWy9&
zU>Q4BM<j39JNROCS-%EfRjut+FQlUD=$<V_M0O$t?Z1W$f~AT(DmBt=of(z|i;+XU
zoS7TrRUXHKF75cCvm3WM4x>~krSLO}QkW|Rz;6HfyAQiA!@SYV-OhXH3ZSpr^jtl2
z?GF~C;OC_5$tT=#W9iK^)|27(bYD;dY@@P2s_JpU9f?Pn3DDQ#GV6;KMOO9hE4!mB
zkd`Wp{viT@+j2;})e64HsG5O5=}y1*%ctVaMHu_mYYI{-dVj%1?@(cn%JDEa@0f-G
z|2%QRD=^Mwa`kw?uG15Orq?kwx~^Cu-~AN#95s)`)$^tAhkJ@{BrazD<Jdx+!qQ$$
zk2lqWH**<4x9}TXf|&@b1t=@6*z3s>2I>fZ5$Q-USd#J;%Svdk^<Z)|u!gEuDv>mB
zK4VVN%>asf5=yeJHs>KXu|!eumdMipXUk@@{T(JJzV!sDsJ}@YICmXR-1Qzgy!`cK
ztjL=ugv*I*?exQ%tgW`lSMK=L*HsX_BZZ9ChWA7X-34#v(4a@tMzk$ZlRHZ$1mL+T
z?$1Rd(&$*T{B8&HfE}aeNGVKKmaiD)VGVYdXN1ueN>fk~dl%DuZ$)C8OpyD-)u*d=
zTh5*UcjpGs^y^qk6KLCD8K+#AS-Zg}HA01=*U5MOBGRO@2sIwiM7r-W@Lml1;Ygbb
zurwe~jBXl>VtZc$-H4_S$i5-a(MO*EL8$~<F@dEC12*9F8>sk#qrJ@!Jd4lYz&}?f
zF(j<JNfcY`Cdj3b=mnMgAOni4z6nkTz|5PwuM-9&JlFzCTpPR)Slc{ok!63R-r>`f
zjSJNdADquijES6V9l3TLb>kSeP%C$uhL3;Z8NRu@@*8>~3*7}JMh&5{@eL?x-zPa2
zhCfyG9XWswOGGf*f&C&=bbxh_dVmz^>v$_Zc~94MBZ2VwhTl!w_RKxsCc#eYTFcQz
zVoJYuFQye1l%C#ae&(ZvFh7DP9e)Ae1W_&_gRq6A-~G61#fAXAHs4)}=oG-8Q}zJ(
zNR0Lu6Y3_bL5|iTX9dVB4}o(BgUipsBl}vbix=;8+Y9Ny@Z)DeGMG~WpliS<e8W>r
zjoroeF=W6a@LarcmtHkx+B719hR7iUyLUY+Qyi&&`xg=1;;RCX8^c_HF-@~@_2|1q
zx)P}lJ7hX)pD~)Ft9B`fh6~BN`Q&45(1qwWKQqJ)V3&+`_95<MS(MQoGwn=|#1pgp
z^!_AG1d-2bf8B7jVaE}th%{TOWm5pG=dEU4OgQ4~5o=i(Rkf1R(Z80)=N^d6e$FuG
zngM93GEWm}Mb<kCt|>(`4(2@lssnbsEb1kCLGmG3xkP{T+xK^1LsN*c>A~iOK4YIR
z*pj^hLCDgdL9cOOfAxVc<qnY8zNK&=wp+1gdGo;5WBU3C#Xx-6RgGD9Thm~X;UNIu
zVE3RGR2AWXVm2^#$*m58-l2yv@Wq_5JM6#-@MNw0(T?cRw}cfZ{0_!%!psygJ7*G0
zS5Ranqy*@dq8O)g=CTpW6fDeJ1^|cmV_3!t`0br%pr`+n7yb1BCKhmv&|EMN(T=u)
z50gr)S|W^(U^dKr1;M7h1vaf?KVjh}xY7*_us)#e)>N6Yl!$;G@!EuDSL!r7Se3Bm
z(?0@SmpumBjp-qv$thIn@u@D$<?%Aj??mH@?vLd_kCVvLKZwuu{>#5vBC+-m>t;W~
zKTx|SErP4N`zGm2Ca5ldy{O{i(4Qm9-f~xOygVcCd^9`FACY%<Y`V5xfOUhSmrf?_
znZ4(l=u15zbbDKxLlE}NF6E+6GDl@r1Q3LykZ7e4UXo=@MFyFQ!L1d}T+-j>&}Wo4
zQsA9`#@R+=MhhRw0@hcO+RMUCoFq?e%DRQ35z5q38d#IC3+()ArQz=tx9l#6U@tU?
z10ZDUtjE5B7H)pvVrjjnqc*^3Z#O}-($9@jN+yO>T6^o4bmdBr!yc`nvP4)ztaH@l
zDr(G_Qx4Hp%`|)dX~W?D>Q!Ttp-?nGFB?3dfDG?c4<!8a>Ex$No4Bl$!jmwCoX@Zx
zpI-1H*Np+~R`O(Cjr{$2(bXTHZi`$K295zyKttZagP(<<<5K5KZSHCUBq8_poX_Sf
z@f$))dgdp`#T)!^#@QnJT!?#!^bg4QkQ;F@xDG11Q^R&r9zKBGJ=@o1=h9D=S&i*9
z-LcB<li^gRM3LQy;W~S$moHRxtj)JCR!OtAU%hoVZTrx7cw1B>^x4q-r=|ZP!yUTw
z$5Fg9B`azTX#<7}UTOylI5Al#=k<5fRqa7Zm{OZ)l{3-oLCf)c)u2>&unA0W)2BgT
zeRT`3=K|9IlmlDa7jh94)!+>kF#;qgWh$7E13A!v)eE5U_s{?K*R6neWH!NqXz^Vm
zUh6jOQ^a2$blD76X!UWRxCr&))~5XRwt*+uKxl?$q4iDnrD`Pj)&>O5z8N%T9_vl0
zxdly-)<_H%oP`BgeT@oXV*U{-V;#(7l(W>qU(SX*`q3iQTBc-E?I1mAO8=4-Ct-3v
zHWoUKEtOX6Niw30WZ{F6IGp_V6w6D5bb?ySUQZW5m&P>=g-1-8J;}L#-nOR*<m5I8
zCK}%PkSR8c-KUYWEaeTfWIci&C7e6ar-pnzh2Ja6lg*LHPVO$6w^L~3Q&y5IF4_vK
z^%a)?gjH|V1wOJj>mXC!_hiZL!<xrkYHb5@0+#H`&_T*GT8OOdS*xcY$-0?oYxKnN
zUeKu<30`tpb32F)QS^hLaIyRB%S{!DitrV`(TCsD`!2n;++Vi%o8sg-t`9PMuqBqO
z@hsszHB_Kvc7<AC!rMyYGlJ~uv@f_ep@EnNxZ7$JK}{Rh9t25>)mZnNfjVs?r03%`
z$%z0<{+~a!MNe*rJaC&Y=q9<>>4Y&tphxZjXQ!~sC7babKR-mZXt1>gf2U7=QEPng
zTIN1Iyn_b-+9GFRjAR30{pavs#JOD1Obzgae|ExMq`h^x-8(A}d<-n>aq!s0y?b$*
z{w&26oCt1n5NR5zSoVVw5UrH4j^K@n`5srn@1>{M4L|pK<hJp2;S``MFvQ8&9W&KD
z9g9V70B!G5uW%d|$<5w&PP#T1C!re^qF(V{`%-Zzt#J}Soxdugd%|}c*nw*~8Iatp
zzgH=aA{XUyW}fcDI&~30+FcwO)ZfYZqxA)rW;n!bhj$c@`<=C+UmjQCrIA0Ec6;s!
zOaynnRfP0a<lRJ|1%H?pBTSHDp);e_yJnBolJwh*R*C1pCsg>YMT{KfnQ%#SSRVkB
zR`y`hit~ThpK2TSC-i7C3S=_Qwh8MlUnEac<tD318PU6=|01enpRGzT;y?zT)dP}~
zFuW{~q`%F&7Sz{Y^zhox3H<&Butvcu^nN%@AO%K%P0*yg9u&H8_4gIa*lT_ZErXW{
zj($xO8>~|vjnLXvkbO}&J>XMD|Morhi@K$skJw)X8m!+l^eWO3vuXVsk0I2>v2fr9
z32Fy&^&sUPfAim7iDlFopzIq!Q+$vxiec3aUWP3<P%$LRBn0?Foj1WH_I)QrpG5up
z3;3^{M*aDORT`GQ=P%+Y4SbT=|FySjvA%$1vEyM7PJZdeq9?KUH$fq`%a%%Hi%Ozd
z2yo{`TKxM9*7E|w|C5hsVGA~_DC@ctrig_tdCkw7i#r}*lz-eB@&zF%;Igk8-8RP{
z*ZZ|e<;|Sh?~O+=49)79AKE#K3}xQ8p}j>}nyOZV7B3tqIlG5!&fdfc6c?FrS1QRx
zD)k&nedZ6ET_PI7R@{Mrs3Tr2PhrQ07D=*E7GZ7q)qx)p)mZ22wpp(_03(VcJ$k!X
zx2fnL&%OLYwj;*WBfpUc2MQcuP6NUP5y~`6xPqnvra0N7(PMy(HRw*Qe)Vb7;)uY4
zj=R<R4IY#YVj3EKgD$oVu2&7pXQ}t*owJwikZZ{PF(5P)e(2D2);@~znh^9B$?u)k
zfZeX+xhvz7ogMvQ{m_L8p6j2!ejCl=St0fk1~XvZO>)&klfq8;;E-NS?vdSdao_!7
zd+wJt-+fLu=b1GSnbH#n(zzn^#*u?X9#S)<vof2$Ij=BXB^n+Pc&-kI<F3+&yB2rz
z$GslirNmW2d=2(*7C+AdZX9uy)1HJ#hN^9TZIOJqjp*8x3d}b<q!mWO)n6o3<v{N^
zGE(!kAwn07P6C!JX|WjqH1En2N!K8_l`?$d^0`9!(>3JCCyLKT{7)eVRpE=aWQZM<
z3)nh`;MX5OiGgCKZCbAmGq<yZsZsS%)X47MH7U`tdE?IYlX<Geh~VaDS0*MH{c8s$
z79y^!BX|+4b+FtW;5Oo;q_6{1z^3n?jsn;We6a!)rR($*#%^NA^1pMVy}|AQdv9XC
z5?p53mw<V_CB2{dZ>{=6+EMF5gmOZxiN$Y1CKx+oWpZt1vp+)<72ZuSlnmVl-Yw{r
zJ6f9T7t83i1H?uk?}>v=OT!8CoAb1@@J=;APJ5U}7`V6!)Lg;crSPu+^-VR(eouT8
znn%a(qrRq@&kbtk#$I$+#n9d8mkzf*%D-<ZO0Ep`3(ZI?Jb85sr=4o#Zuu9%=w%-Q
z%@B=@Fd&wxe5LT4>D#F;6|eGr<a*OG%(qj-I*_a1Ss*>3@Qn9;q0v-$0@o7+MiP$h
z+kL-MdhoZI9|y>-8rhX=*g!3By5sn8okLpdsTi}KbaX}F@5#6`7zAaz5*JM<G%J0s
z<Y{lXW{KFbKSLKP(h^1azv!OM!;TRfkvl<=SKJ4UEjp1*z6P2iPknch+$S5k@Dykh
zl3+f;@h}f0i>s^1=$mC(pQ?n$3u$Bh+}FCwt5NXIak6QYFuo7J4d*R!GXrbnr$8cK
zxf#cHEnTc(UN&Q1_Pz0~U&RWC*){Z`w0mJfSHy%(MvdGs#K!_d5y4@aTThXsA7-qP
z=-QV*w_8%>sPQhxz|gp8zk+WcklhicR|-4tPE<e@K|lM@RT}lrcKex}Bx(iI-qP<p
z9BW0_nbS)e-CSXe!pM;3pQ`+Kx9?j*u2C;&BnkXB;5sC>qf6XPdQAAtX%~k}GbcfS
z=Y=1uOW|zkrqo?sETaR-rK$FgsWJPbd&qBKaM5J|1tSeUuYjq&OERx+rbVAsM&85K
zRV(A`2|W0tHYU3YN156OQOnn7j~)2(i4kX+g?5C7pc-sZ#<R7SR$!oQ=Jf;R)|q^o
z7WR0bRLjg$wtx>nnf2g462+v3(p2-bJsd{O3)#C2-FABNDCvW}$6+sIvN6NDmavU=
zKZ~VCFD6I$n3cV5n`<$DQ5F#zXEy#6b?c-o<UttyfGRe6f#+xKr78p~WN|lzd73&E
z7k1ye*QI=vxumVsfR|;7QQXt0XUBPhifkpF%F3Xg)Kw9&{pLS3^Jxe$$eS_B_JM_>
zcwHnKJ0SO>$Z6*p%}kGZ?0YO2kls|pgW8?A%#NL3v#r(IO@RW}-b-D(cGu0JguiN6
z17Fi5mK(J6i!75ykwL{r%U(;lpEOysiag6|K=|f^?dS<ok^jWbou_Dnb}bcmnr_Ox
z=+d98&_nK0_sKC0dtC9&EwkYGO%A<@{-q-kREL_8{naIf-M%InBY&=k*YT@ucRAV4
z>5tajLIkAwzLl3!G(#=Xoo*Hj*Ya5xC;i#V(?noK5w22%9eUh-w$j*!bhRW_?WCla
zteVnvnFb@BdEb`{Rb;fl%%eZuVdH1jYX-lUKqf2++PcE0h|}{%y&CO-(mFwNdrl<f
zIhUp!pU9DqhJIyGw*d%s8YFW|@t_Qy$1>L-vJp$J9Lwg)B`kM=`0qQ(M8UsDc-iTI
zxqX-MM(a+*t(R6CCf{yEF7c@DZC5h$1C$FACCY`chAaky0;aIYlr>69Z?}y%fZ*y@
z=M&Ynio>hT8Y)B692mF5t&`v(bjijG!W|^n+~dbXfhAevwjqg<c1I3!IKSLGGAfN|
zZu&;7mDDEeW?4~t9@Aia?wE?}ylZ%jd{;gI$?NJ$hYYhmr}u?v<Tiw-JhnYyb(T&P
zSwXTy!eBFZfXGxJr;3-M0uU`*(EeblbkxXv8n>Am-x3`knCvmuG~PuRQm6!U7m!FO
zl!7wdE_l@j!Kwr?8HCv8es5*C8a$9QB5dCN)n4!es9>%0DAsI0RrrN6L*xTzC6@LR
zymotw+I1r=3NS4eMZi5-V$DiYtVc7qT?axQprK#-|Dsz?IEDh<%ZW^HTeud0AdF);
zhF38A5}{m)i*gp`jUby;YBKHVpB=<)O=lLT1F3cRSuf=G{$bRXseshbi0^)?*o(m0
zq!#6k>Z8-K)<b2i2F)${j>ZUmz9!MS-fw$zo~ruM4pN<a!V}pCXm7t)2Z?Q%fmK>e
zCHt-3vDpzt3w4HiaC6VXC|aGe+BAC6nTw&B3ZB6zXf1DNc~f$$E67r}NEQL-wi{d2
zyxxwk9na63VlYS-isI&$oEH~UFRhr$xdF#w1^;7u`{~&&HPSqN>wnHTmKNq0Ke-*<
z#z}p~h{pc96QbaDxxV~Or<2pSir9XS1Er_rHv4B&Si<8pdPk)rxgsTyfL!My2p0(V
zY3wjQYO?n)BJTX@i6ulL9YEsZ?Bw;7D3Y`<+U1U0g;Dnrhw2`8`OA|b(a(k3H{Jqc
z24)iTFx`)(#?+<r{yeub3I8aG?pWtey&SsL*A3_=`~cwI`VwJHk=WZJdGgWlNfnh2
zBM;8Q_dm>JZ%43lR-*omTI_!c=$FZoSZ|_40FS&JAa^R*ub%Wxuo*xO_d0qEVr*d9
z+&Q#e|1qGmHKYSS1Wu*MO&v@amL{FVuwJ`~Z6hoKTvG8RsA2263Fi@P0F?w>Bs8bb
zfLneQ4TrljR_hbh^<*h<+UM)%3eS4EQwmQ0(0&uTiG+88_n;7;27dpqtv!;nFo%Gh
z;N#vXbGjI{wBCN7P^BhA^uwi7M&^d%#;gW>yLS^U`zttw0e*zq`qEc>?tjXQfllf9
zD84q{_>)MlfcYdTcfGM=ePQFad~PK+z4>A>4zNAfZXWiE@`Tx0T0q}2`VL!_e5p`C
zsB6iz<dXN{E)U>*`t%NKPbqmH5GjCMRgt>C&3r!3UiO15NUNe3j*6p~1+g9S_rF+V
zE7fN5@_Z$>snX>zHN_MV00_H%^~;`pgnFk=2rlnc;dHxKHxPqOTjSXD1nF4q{~w&+
z*NugSKk5$EZHE$<hf?ai^n}?TVVSGH8y(QYnwOfQdnK$65e~tz$}2oDz?zxrBhJa{
zh4h@XLG4+s7dRYV=E4^tMH5u_eZcMLOQD#y3(exS*s(>*3MIwglFp)UzItPI<B2;#
zi!lug{!xXW^+6l6gj$H3Fq#6D$l+Vif}hPCGrmzCf7#|0M~x2o^n>$;S`SAkyPNJs
zRjWgbRBF2A(XFo*_dROvSJo&DLTvDKM84r5{@2iZwfbmJrQE30cE3AVQ3>}XB<WkC
z2Cr_pj}^S77^`)c5tQaB5*u5-9QP#2HI)&AjRcAX$l(}Mdc%EGod?&5%0|$F;vV}s
z^O8$bj%}UdQdc-nx8K)CWsdBjS5SzJ56-B&d7LYMeTLtaS9#AcVuNVdv%cfe`x8?u
zUV?So+~kfeAX@PNb*xbZw4tP^aDw!g_0}_OCahy*GA%5A;^3R-lL*2HCv?B^S}Fjw
z)Usy7r;d5LQ{ANu4G@G=P_cDW1cCO9UYARMZc`RL7D*m0DjvCT&oIM%_apJ4oeh4i
zOfK;A)pSmclemUJydAriyrMmd!gPInxZ7Yf_+i`VcZ%|=b(7Y0hGO`7i{#tHgCD9A
z#IN6=OhHef(G*mXSP=dw+zyL(tIz+q)%QsG<Qg;g)Z3lbE#|3a)e&t#YH1EHz)sK;
zrSBS1T8rYEiPrHuBA*{I3HaKfQ&orh`Xu)XNqWEDapk5%#F<P6$EE&Id>5)X{u$Z{
z;vK=#pIZi)**Pfm8%NY6do`hel@Oq3wUFqCKUYajFYZsD&Hu4)XogY|cxvccO6Zy@
zU^8`K1-;uy6*7cQ-0H@7a6_x@czpezhm8{T;r$XX7Ul2niRQokI-mHp)vpr-5u3_k
z>e+@9gUr4Qba0^{2W(*QC`LgwvGN$G#sSRcfZ{4~<^hz9rYoUY?;&uj>@)$Qo9mb1
zouHMckdR8~2Y0T6^`Hn9a03B-R1M$@dEm~TX(Sc^AF6s)y7C6=T`CA4?+6Ue{>4Sm
zDXhUl6O&?aC1^Nwi2Cb+f+s))Oej&U*}f@L#f;0?L=wedD>*s@lA{<_A9UXNpZ3mw
z_E`KsU;p2JZ1v0Z!x^l1)lplWtF76M{>2D9fZJ-Ea_x~rm(M&#9WLj;n*k4f1EV3!
z5apc5rXoRzuy6o?9H}Fa>HMl1ydzi(^2aNj(_^@vcoFr*1N-?A$L1o4jGsVy@&Bkw
z`+`en8~8mjZJp=^5s<N@<}9kxRTNoC?SWvg=s=8v^J65sJb{(`7}SL~a{eONB0!k$
z5twKHM@<;)jTlvGDU${<ba7zpQx1GzK*s{&QV4kGqQn6r{SfF55F#KFxIa4QL1f8C
zvEIjn0L|n#0*w{G+cn|?$~M(tX9R@c|8(G6@zQ$2l*;^eDwjprgC%j>y-_B6y?*@I
zt7AO`MmD>k_vqoY28pOpAXKcnuRTjWLZx^uF8ay)bH)q8NDyiT{PKXNe437<KyK?&
z?%1O1lE<IZ@<b&=^q+07p09|kq3ggUd;r~~DrP3$TjcGl=1P~vWxYQ8oa9=9TOiWq
zqH5psnbM7Ny&H0K-C3vod?xvRIT`OCfXZnTPLKhhJgk!2S%Md98_&fwCTmD!?-LC+
z3;fVjVxmz4*2xg<fD9xRqP|zjF4x3)<<FH$yu06g-{;}9+J39{n~?^1sDGlACHipu
z$onjuy#$?*@yEbRH(Rn|GNQn3Vq2i$#-YIv$P@sC5LTYrgG_;_z0BU-%k0HIBT!7W
zqYCju(z_fQXWYA2p9SI><~-9G{~vtZ>R#}8e-4@|E-=&WxY@s+zEf)YJuO02WC;b$
zoh0DQgcR@Ue7&&y>DURNg5mYA+J8z<$<T@JGB(h^d?NaB8Z~f!^F?BB?y~d<CbpNR
zIUNe8ue<4LxE)#fiCOnLTB)+;C<A1%VRwS7N3a)~7uKV9*R2fe7M&Ohhnx-{l~y4s
z1K`3Bs7b>+RyFHE422IA(Bqv;hTiaLW#&H~PEOOVkeDTqf7k@Gg<G&JvC_W?5$qZ8
zR1X0ppCAO?X+iBMY5<o!-_*#X*r3_ygm+qzqnsSXk{PZQ0=zFUlhAnq?6n#n0`&@^
z?o~08aWD>|t9rnPpsUIgm|-SuG@fSCv1aSH{b!3M)i85%yrk-|gYgyr$$(d<CvN%c
zsE#hPgth}7rAK`$PlFkC*T}khGi%&E?r=I?PW(@k4O=DRv0+ZXa|7NB-s(#(i{rZ>
zks^p^#QRWf*Qw_ky^AlSUvu*sD?Ysc`WQC#3I_}1EH(Pj#GJ1KC=0|kt>?giRy=Hm
z!GOgbM6U@Ge*&yRjP)J^&=L?^53w!5f?WbtL)HMHe<6jzYtA#cr6Pd`&$s3VNb;w?
zfCDO=!jQBEv6*sic~c=}H+mi(1_jsC%`}(~0Dp=acfgqiC%$87tL_L;Bs8njp8<%6
zMq+Sv@N}h6EOJa#AqVtJL4)^GU9!g()$cH@$=C&EmO=y-su3&jcPQ<>@%T@7kO*6p
zb&sNh71G;7PpK<JH2XNp`un`n<k)<MZMCd_8XVwzO}2&4W=_kq{n+`VmT%T%{n42!
zqN8>*G&SO%*E@bY{5hRdJV7dGYWSl%?)4qw^G-VN*Qt?h^RsNy@#UG5tQz9A51a|p
zMpI9G`hU%r%AsQqG}K$YGEwh3Y~GljY;=;B=aLzI7rqx=vRrSwJfzxS6Z~8Um*qU<
zSv2)iB>&AeSy}6Q<JADKiJ=AN`<af$hc7y`Z9bY6b|AXKcgM@^<Qh50PM5h6OTX&U
zsxd1>Qp$iG>(02xV6bs7UJmDFuK-cEZJ%|jmiZRg&v6m4ra`v@ik3P(dLpb_RO_Ji
zlZkvewzHBB{JZl^RbefVnHZwCd2Z{i-bGiS$!znI(EXAV`S#esyrF7PPz5;SJFE!e
zAme$jc>al8lnv!KF2l?f`D?!cpJYm&fS;#6_f~hFTd^+{ETRJ;<Y;rmMoAVY*ZkB+
zICgCv?+6y#0dnRuQ>@cM*9qtXANxk8BO_(BRA!;6%n@)IqXt9k=Lkhcbz!Ib+@5x#
zPgz0fKHdhEg7v3%d^8$Wrbxm&(<rH!P;{|Tur2qE4rt81%VF}@g-@#EiEK6~CWB#K
z0kBt|Q03s-*Vc1QETK1qM!6aE`VjE7@7)9&^uN$i*Bt;U6eJ2!KoQafT5~Yx8R)eI
z8}*f6Kl_2lglnr+4>W-S=kt2fCg7Fcz{~w#?7ew7)bHCrOhS@m%{~=T3CUKrX|*KL
zMo~t|I-x9?Eb|s2>l7ixkhLsRF}5)FCCM5Y`;dK^v5aAs&+qcR@9+KG%lCIY&vQKg
z{Ep}N{L!b9ruXN1pVxJr=j(i(uNTyIfdXfSC==F^pmw1j13_KFQ2mST3%YZMpRnU@
zBtV$TKw(++iHL2D{Owb~26~F9F9vuT1Q5X=5(4#Xstl=J;BI+L$z}RVQ0FPIp^Gfj
z;%4l|tP=;B7ZA`A%s}Zoz<~M**-RMgyeNF54!#b=aky&{TPM>5z&4}(#a0NJY{MBr
zC!vKPXck1MI>?QHG_NcJVA7wXXIK$2RYR;jxdj)17^hwK63od}ZHfKc(%ONC)e0<L
zuyu~W>&NJOMxoifg-I+;zNLj15rinIiPK>9^=|aB+R9UZii-H$tfKD%SQwO-a}g+r
z#WvSXdhaW!3fj}EzRfnpI_VPm!C#%CL7%j*!K#P(ToWExyND~JcE)Uonr->^Wy!*p
zu0YL%=iuQoVNwM(ykNlQkn!HOkH~J#c;o^2nbWfTYBIqhKyFo!p<hEQq~zCEa=v$X
zUgW(e>$RO+bCG!G6zD}?z!obf6XA<=GOKN^l22x2zb_fTn#&MIQUUiig#RIHH~Dg-
z2daM#>M1fBK4~ZW2D9?|{B9Y=<!H4lW#r{BF6^VU`}nL~#}d15J>Ai!Te?fSf3az-
z<Y2V-{p(qNgSdBR<5v9Bk1^SRrnWsTnr?*ZQB>6|oRog{r|ojGWs2ML`If!#!w4gP
zfG`JLW4KWse`MlwhiDcflyK#2<7;|X@*Qdxz@Hf~6v4m_{Y(?7ywKI2!d3WO`yRsJ
z^r+RT-$~RqTqT%Om)yB^Eid;zGN9n1QtsaSGuitP3DpXkqE;BfV^kwx$9amOL*7f1
z`943oOgAr&nMQuYoE_P18}meGn&7ZNDP!?UQ&P+*K8@A9brJ!MK6V!|OG0fzZaj)X
z+5F{AY@RT(!89-Z)|g||>_PaplsDZI5ekuwB0ImbexZKqcF2Q#TJs9Ck|Lk2962#^
z(>w@PV(N@MImf>_X#N?|B9_jDN~y1eN0mHZFZ49U|0<aa9aJhJ3Cg0mfQtS%kZ{G&
zOeZmT8<gvczgqrP<6(Ifw_M^WF2xxYNvO`D3d&u$IalfaVjU?TdQ+I-w&ve>XgrFm
zR%1H_Og8i*g!N%gHphBL9^v~66Kq>O)Vc$=4>)m9GD$NGK^4+WAfruJ?Pyu_QsQ(+
z-hnfS%rg0!^FS!vOw|AEucIXcB|1wdricF|8{Ika+BdWrZpFebB+^32oegl|DlH$d
z5e>mp54}A`xfZw~4K%0gfLK0t0f;B9Oe(_gE}K>ZA@cFERlO??&X=A}?dqOlWbt2N
zu8<LtC;h#_#7Co*-BvRCBT{;p?>!?XBWw*f?McFgE&3}`Bpa=+rDzY|AtFqLujU>)
z=E26$2ByjbU{yAfS%^WidaiXe<-w4sCkyAYl+K^&R`{T%G+}gFE{*3_X48>Si<^B1
zd?U8BCgBVr*mz4*dw!v@dQz2MzH#hVXOBG(_W8Z5ovz}hrh$kmK(&mAH9>+X*{fuW
z6s6`jo>Kvt=BL>Y)#@Z7J<j3^2@es(3lPz7JQ@?Jact@2MuYC>(YsbcrAkE?-pf2W
zm(Kqaio}u)#$Ysuv&)QwMT4}Q)5bB4EmKCBpHFrKBg7e3fG@=n|08`??M?b3gx(yR
zEQ3eqfzHD{Og|w<J3HU0RAM8l-zxgOR%x8_;Nc_=MbKMuWAU(h5eI76Q3;(l$^C@q
z!kWukbG=tO+>Zo*f7jMx_MH$wuwNl^TxDjl1f2Cx_4b|DZ55<;X->MI5`UyXfu*7q
zpgIGV2+eaGY)VtytiH$#C9jW4>D8;Ub5gNEKMAP;6JSwq#jdf0JEP1o>-<aaga2YP
z)y>+OCqMh|I<DN4{0vb)mdQ#ONGAAUtd|?}JIMqVj6;Kz3)c#*-@&PXmF&9nogk&F
z^UQKnDO>JAb;DWQ74X%gBL<ZS%(xQ0zcM_CzNd)|oik8}Am8cS(vntS)n|g!9Uri;
zgLm7(r-r?-^t}Ml>|@{hiw!(kH8{;Hy5I{v5Ho=gfFvvaccJ0!E_Q@ARCf-V>emGY
z#l3;;fo|_LaC&jGOzoNg&eZ_{4uLrO2YeNwD+02b1x7S<#R8(tanmEg=vX219W+$~
zFjvCBmIUi9NS7vi_%{!3nRBo8f;)}#OQIR`lm7#$tOR9TXU6V`b@$&ZES{f#vEg$Y
zbUH<?Xhc4NLU&@*;#7&Vn8<C{#vtuwrXt9ug6jNFR*(rxyzZn4_DE9-C(zfb_0KVS
z&CT#Yeg=Q8gYJ^>3DJF!@2~Z%NOXwqT*cIR`WKt^)7s-3#GF8F6`F@=23VZaAnZ9o
zQPy-F^m#Az!Dn*-kz~A@K>NhB1=oiH$Us|t|5F+a;wtE{i*$`V5kn9AM_!x(!*jPF
zVS_*<``e^sPDs&G#*t>C-HWICr$5rh2WH$GFtn427~R{XLd8opWaVf%V#x5bF9G4%
zN{P0aL8I&J+b_ttI#=@K1TIwf<mQo$k^p^4p(h8O;@Z!B`!aN<=+j51frFM-rtb$n
zNBXrLu}Bu^7K<^QvHZ9C{U7v9fLl^1Lab&<P#hlfWhE<ar7$y<vhXxIMfB`JB;-D(
zHSXvEzX(X0u+NtNg`~+r<xG`l#-N2CyAG+}`BnW&smYJYhL(cr9)J{g1-{H~&zQn}
zgm$J|4bgz^>8rw2u-XF59sZq^1vp;vs>~b>0Qtyn1MnDU!DBRZLvz|)+;bwJR}Fhj
zAX~3&9Y^eg76m{`l9>Rapzgr55yO?D!PB4@k~3Csm=C@uTIC1jU_<If*!<#5B>XNr
zD;^Z6brQO!$ZT+9?T^=}H}w2?_fV+(nbVI)TVUtU!HWBKl5r%EJt^7^@uDRg4Ljru
zJu&r1aY=-63}Mx407m6X0i5eA0LD2v(*&?Q)MP!26JH+V(|sFfzdCeS;_^`N%IW|(
zsp??R)LapKkxz>i3j8bXl3I!Gdlyu3qkhNV23&Ooaso~Hmc>9I0d-Wt-}mMB#=qFA
zklW8=H4-!p&08Bd&7L5)hM83?u~B~|W}Yt6(UZNbtWV`ppQ~_`1CLa^><}>az39&g
z;^&$P@XkCK%J8<Q2=@QvbKuY(N)fe_kt*UD890F?03ya=7ufnuSBPvg#<=Z^We6Um
z2KV~kCFi(@mRa)dFU}C8mi~C2u**aZV{0w4EqM+W2ab#>y^E{}dmM`w)E^O<|FsIc
zy<LY;MT0JRDH}5mEk$&ZxF_n1iEy8}UEI6E^P3VJF12)s0@`1h6K1p{%NY{{A!px@
zv^dveUn^ofC4Qe$!oO(t@w1%2Wc;T&e86cJL7e~+W7-KQMqkhWc9*|^=XkWTVxjTi
z)qaB;c_JpFXWh8iaXpmx`!>=!cVMLgei)dd4<nh;xrPCdl~cG~O99G3uh4QJ?Q`v9
zMS!uGhu+vNab+N{@rMBV?@9XnP#d67GlSDX`NRpS`F1j*5}}RNv279PDwZ3X0>hlF
zx-MvTSD^M5{8w4mz*-pH1#0`{e-(UY6)4yS*8LDau}gCfC*hA(B3^%rKVNY6-nPii
zbFk0ArE@F6{dY^<$*RBsMOyef%U}4NNqomf3t#MF<cGP}KV$FD4ftGe>RV0u<8Y7p
z99Ik^U+Va2C*pf-(1ZF4UT1gpbi<EP5LLv`4A-pR!#1No6Z#WoA0;^N7ZKhUYnBb2
zf#J&<8)<m@IU1o+K$9jxSt7mm@`{eg=YPlvj%uL00R&{c$LL}X0GYXNIl2j_tSHcP
zYCs<SzhPyvL8l4Qgeg2~jn$9{5TAw38R|iBj$fpKYSs1*A;ofBtHpgw$wn8f!Jict
z#+SLZBfme@#S(#ji>gfbu`loJ8%P&HzRtg;#~%Sv1rbakX*6IsgLg6Y!%Ky}7aaDp
zL`ETaidRtE><$h6b1eAMyM+xC77sWS<?S-SEZxo}V$j6}x^t_|CWOJ#{!1rCd6_F1
z)ADJ8D#7-?ixNLIk9;*Y<cMEM+5ORi@0{o^^eBSQftEp6MonJ*TB_w+@XP^urteV_
zhca|^VBD~tAO;8fJ|azmBKJnEH2h=3Gs^zl_Sb@h3}JSrZuh;h2CUQ_h7nmaoaCx~
zDR$syNz9)Mn$JYduVA0Pw=_P-sp=48aH6Q7A#b;!qFPg$TnQNp)wu7meL14ntTcAo
z!ZT1r_zUYcjeXK=$eQ$LrU@r7FC$C~DR3^;4$W618oD0a^QJ1y*4*%_z&Th~A^Id#
z_Xs{!<BZ*yB7l&Z-&_Pkt&#38So{bH7q$7(V@MG{di(q5u|~#cb$uTaPQb^%CU`VZ
z+6~I>8*dB-h7IhkNaqJdRL}{iHGfiNdx2>|xQ1Ng233x?B+94x*#yD2ND5ba%W`gm
z&7TKcNB~!hgL<Kbwe4(wux)nZ!S4&Z(AfSOfMNV^nUCjS6>tGp;zF&ATnKtS*}Hz9
zUvoVnsZ-(-7=@K+^zg9lljFlW^Fih}FC-qQx{-$(x$-sbrOXGyd(6<qiR_hn0@19Q
zXgb9;qO9_VP<la?9)3TGzZW#&is^@t=Cnwv1I;I|mMwJB#Nd9l{u=+BO#@;NRCg3`
z5){o|K^zasM;qVN<)dx|q=~7mnYf`{0Tj{^L9YqiOF;}>w2Hh0>}loNl;Hs%hHEQW
zYX;|VF1SEwya)>`U~rRz25K$79v|>ON4gP~TH0Q5kLdvpcx{GW*piGsx$j~@a_U!>
zM;>l&)A*}|TN2V`|0UHEj<djbWMC0a&St5id(sYc1=@!w>krDko7jEv#|*`a6$Lu4
zhOE8dwpWM-y}du+nLb$jL)SwSl)<HApv1KT2~f>Gu(Wwl+awCds<4J;<8f6W6No0P
zx`GseDvAY);OYD|J9KacNF;)~p*940fEhfng@1-1Ujk<#jnc{VGGnq?f=iea;7759
z2^^gpx8|})7ze;Dg))iVp2Y&fH~+8iJ3Yn!(|%m5lCV0)F$EYyLoA*@;|QXQXx~x-
zoR@~T<o#sTwA#vj#%njn#sVS{CP^4ihKuX(k4)+7PHeQ&7W!kB15Z_Zu^=%OjHp=$
z>D>VqJ2>9MJij&6$o8_$q#P%;u(VIGXVb<kj&c?eP`)CoGqHr6j++g>7Ufwk65jZ6
zJ^$HYNc;O2n%#tw<^^hvWUtcj`-GT5cA(SASn=0lzG0nY+^V*AJy@5VEP9B?eQ9)5
z4W#)`I0OFBwZRQ8Ibt<%n^yek)7SSkVY#2ZpRSyEgQ>Nge|xIMRj^6$<kP(RPf_eQ
z%(+#`D-+nz8ey7C?|E`*92jjYZ55CWVSC8^%&j1x!)qO+E#RWAv#Aot0jce>kfkd4
zhcTp&>U_)!U&yzs-)j&zgk==R3BNCncVen^=w>KQIxvC#mOFUSEBa_tKNfVKhtHn0
z+(q#H7aY(`?7K%CMZJtqSHO7U2+$YiVqE^EuZgOh@2wp@R8bbYkt82x8viCN@9=Y6
zKk&To4E2V9H=?gWYLRoI@8#`zt;hXkpG$7A-F@=x^H072*Z*o*`rK!UkFG;AqsEar
zJ41DEI!Q~;;H%Elhn&yp+&6o|GxuHs*x-C1jHf^%Xs&t7gWBjMl?Z>mDrWN9wVQH3
zGQVA_P(;*tf!ApEftE?y40Kd<eoCrVef#*-cja9e*TUKys9u)hPRhxUE~#{kz|Sc^
z+I^BdY%F8E&{`|Ply*Vz>#7#~=oxd82pBH#ut9o1G{YS!@xt*tCrd2#E-Ec1iZu#^
zOoGA4HVmNM@yHg=6FZ6s6#t{47=gPVx|eeHZ^{-)zu?r5-|kFz(v&&a{dhf9xnD;~
zzeiPAl%gym{A^drvGm=zPS{X8izh{)x#Jlm2?qC-62s*Y>(!bUjzs+VY8=o=ZUG+w
z1NWT<ai-2J61Z?y8G1X6N>h~M{oICZW=w-1r7v?w_6b;K0h!cF6Zm5BCsd~GWIL}u
zprVT7W~uW(d1tR$9>}7k<sS)B1gT<(-GoX>3WjE=-l@A0x7hY=t%(iN)8}L+)u=K2
z1|~y08!AUO`f{DB`z`KF&aje=6gS$bpgaolpM!s8enVjp$$E~>o7Y3uG@}!)82$+V
zDXHs;ct{+qvvd1w-T3-K)d_=g4)Vu5HXIMLnBaJm4u&!M7{}eU>pmHvf7x~h=6K%_
zR@LKHNZxb-G4zQRLRa-Sr#&F?0_}0~vm%EVt!s;mQ_1pm^F{d)8<^jii?qlIED@Sm
zF7}-06l%*MBBmV@eolVhV`sk=F2?Z&K{zP-dDOrzO)gpO*S$wKtiZFt65aQnhi%%K
zbFvvrx<^}!wf<BbUl{_Fxy)>aZ)G_J^Wv(uR@9)U-k171e?Q0fhc{a^HdBS!8ARF`
z^(n2IY}P1bJ%)0;_qj6gtyc2A8&%NDJXK>yirVB=ZFj1=1U4ed5TmlWq%hM@$TAs{
z47f5<xSSqC;-YuAX1rx_eU_y;PMmQI`GJ<|y)nu1Rr>12pqztsssus~@F+;S>b7(|
zHq`rD%3S5>f~tuy6!n#V0Rak(&t(jud=qB!-Ky)Pq{ICT<Z0KaB|ec@O8hk=r8k?E
z1H{h_(j_KRMqiq%IIjnuxWcA51*4rf7_xy8MR#cRTgfLr&$kgjFG@8p!9`&v%uNFV
zz*WS@s)KX+`%R1y{pvP}xxpGE+TRu9ot2eKw(O$0CVoYXTdo`xHfOgvPrv8vlRU8a
zXL3kMzaKf&R#MTU=;x)p+_ayK<0B3MN=s0fgdo^ia#lcJyQy)MgCcC6lfh9H>l}CN
z_WrEcD-=ZcB4WOG_V~Iv=lz=|`2LcieBN2X>7Ub@p-zv)EiJ?ju|&4?=ntVf1r6l=
zhoFW<?CIA@)pT&q?V66WyVxezY`j&u-h{??pM(8ir18=U$z|rTBVIV+S-~cGvp0N=
z#~NiS*tkSI?@-*ERb<|pq+mr|4qLVh)R*lR=-sVi#d?j$>~}^Vx_#fothDB~*e=p3
z8e={t5o93^RwvW!Lql>0Ydk9@y>DV`!EB%9;y+-8AzB&b`}WV(wX4uR8u~7KMNMC^
z)NJm@J1Lye=yp~IR15<!hy#q1kGGymAfzBi$Bg5S{s^>t+IB|f*Fv8;J0cO-rn`rw
z9jBDy$7jf!VcvSKM*_no%}SmJNp}o5ox$~;IL5IW%;Mo74>hIjrG=1y#OXw#<>gQ#
zVYx)A$AyAtY3#OYh<tFA|A*i;nw>QS<Mf9}{KR%rT1E~^(T4D)ynh<y=DI#-b5?kg
z^N#YKZOfHE!tARshTuwrHX#IVUoGEa-Ls+|`KvrBx_VJbI8a<H=yssS#vfsG?(d-S
z<Q}92mmoN#+DPm7m~`B{3@w>B*<R>rZ7t9|CKc=Bg=?%nNOh+r<+~6A-jEeurCWcT
z4C{z9d3VCiR!h(_P;HcuJ%5bjC&3XH+Rc!qsZOZu%`%Z9KEn#D#lz6&($!!d$ZjVP
zF&ZErDc0wpD8#5ELU|@eIBpC`@TQEZJ5zD+#J3p9(%`7^jddKr?1M~!qeEga<P>ma
z#fDt;CDTg!WM10#cO&!Eh308C2ZG&f`mWEq-s7$ZNroZb{QFU%&3YG&YcTD}Jlwl|
zF&7lkAkDF0egf^!G<BNiC&U7@|1}Cz-<ZT%w0LhQn8EI0!bDCXm2^mD4c6H)qn%B@
z9JPZ2CoK=8!<WZ75WOOgkTL?@y61-QdMA3AU!sOD6g>qD;x$YG4Lb>M3=z0Ti1c=%
zlzk6U47Jz}5z_#%Qw%YLv`Cni<AvUQR&5N~iQLB~-so@2hI$!~g%A})c+fhG|FgCO
z#@^QOqWj+5RT2C-ndfAcaraQT<5(CG3ei=0ltAsD)t#fz+rw4YbY!aDzkk-@D-GV&
z^cX`JG-7ah{#F#}mPT0U8|$31)9!sGZDIy-@oetb>OO0;ccATtXo{qkdTI1sr4lnC
zC$0XkBF?t>G<CVBefv(~<~;J|lVz|)m;cO*n%fxmKBj~M3KgUmNx#hMBK}Es?wYre
ztHdmSvL+h+gbq{`ZI0&xLFP%vrK1HG^JwAjB>_@3ZW??Z7HeHH3sPb7geU8rVKw62
z4tERGK>*!8%g<N&k$d39eLgny4Kz^Byq0=!0`R9NezT;V29j&4vqwsZ@qGeSy`!P$
zvKGZ_JI_cfwt{p_+r3dG2wRens+>O#9k6!R{-Ko9adO-1qm=#Nlf@<!2NzI7NZ$pB
zO>Qf}GWwK*OQw@UOqIIi<qSK`4A~}6WNy2hn_Ax~AgVC`AyD54M9~8p(h|4>mSx^q
zu(BVG^ZnXt`i~uO@ZsqKc3j~sNMd)UI)VxTs*H_-JH4#xs!}XKib9Vr(=SKAjm>$c
zB5sbP-T<)#lEI*JC|IKnFPT651BELr&ndRe<BEL7V{*cj2M0SMqwhcMRDc`($FBYR
z*XKCV22dT+|M+RzWF1<TuER_%sA-s4((!Tk8N%Lsh8xgc1C~Vu4u4ttdDIaAajEtu
z(9PneV=mVwj8LK}U$~LL{%FE%(V6~u0Sv}-6@O}bkbSXeb~~rnC%0WCzM}6yfz+<^
z;u_oWbODg~ULZ^qRXLNSxn}C$pls(#-wf9s85RHW=Uq^it5~y|4d*03?MO&<|3^g{
zvW3Ua)>^d%5^9|Y;MrB(K!t->?STQK*`_2<5_ZWtEb@D-SbV&g=WnLdj)GoL57bsg
zL$Dg!K_Qk_qXfQU2nw+oCoFX@0l*-)w_E54bh{=T9~QXY$%mNtaDF<7P?}B9&k9kD
zNeiutq*{Z&gPg?{!vUrzHHZpwy+9%o{T^`4>^Ayl4D8-qx-Ww>Jd7bC#JWmo9$O&-
z>|z7gg7}SLU2C6!n4z)&QL+M*GiQ|`O+txby|i6zSt90vd)~YDYpIrYp|I~Y*Jz+y
zwmYRuUyxx=9*A;fm_7B>jDA*T!t1#IoUF9oEXT?QJ#!B6ia@@-GB{uf=ztjvPj{C`
z<mQ_E(m&k8<>Jbc()#2&JtTGZiR!(XM*c8;Nm};Aw^Z&X?aM{Q**~+?4_*)B7<SA^
z=WlTTZAe`uFynA5Uar>~=STVbrUg6qmdcD?WuM0FPU9Zg#PvubXo`r%ji>fAZH34K
zl`ksgd$LBb=ND`=D3+`}g!F~v>ZZ;YRkw)JC0SpG=CAbbR`^Wh7Wo6N>pB5{j{(j{
zG)(%*L(D(L?-F2F*j6AXfQu$1RNfy=A?F$;PfwFhzgduo5R0y0kzWOuvoyBWKKd)s
zGAY3gELqx@kItJuL=Pii?deu`{&2HL66_S)GZNtz8bAO?^rcW-+r2j_0aL6KeXgw!
z*j!7aAmW88jLY^4m_p(5*CTZXZ+A9)1KPl{7y3><`eLLcyXUpyuG8uJ_bi-)y@beS
z2*E*6wR|DpXRa@}awo}I{~mYqI>Ajcg)ZLS`j!FJNtW>QFC;(vh@9ljPHUdldH2)k
ztie7VKg-#lDf{~SMJe}8W`4dBnF3aZg>3#H$>6jlcXWmm9_d|oJ@oufabIE4Q2ULh
z>skqRv#Naqq!r+ZluW3FMOXnJWOZW=K8@c69svUuHwtoN#*<kI_$97}CJCkGxKCCs
zez`W6x;ouYFIyUj9g1Kq(*P*5+nJ1cwUM;qFF+Fa*W0(;qft0*A9MQJvvl)QnV-}5
z>`a!`*(lHm6lA^3-bz}){<oEX7N6Z1Q~q%8f;BDZtvD<i9cwuaobqV03}DIp^aCoE
z-lBScc+SVzD-tP`b0v-Y2;oX0>6CSoxNDQ}arb-sOW2R@oQY^ZsO~!ap<ic|uP!9O
zfH}D7nq}Ahv9U5ZTV8hW4c>6n)_MVvp$@9Jv*h4Og=h$1jIWO4P%g`^vUl+E;!D%l
zq&>7`Gq!!7BJ!|%cajU4J$)rZmni+=&p(WQ;!HLXMus?WB(SkZZ2+h8C4l$uo}rmh
zwn?T}Ts5lnic0T{I$J*!EGr09=|D)(B3XOhkQ{1~))B9KFWjBke0rbw9U(IXJPXSx
z)+}Bn5QauQ;2~b}8&*@e(g*Lb#)PasP&Q-E1D5tT5)dp9Lw6Gdmo-K_oHqh|ZoHGz
zr0jY&6A~@#XO})BvbhO_tHQMxJ-7zYhvwo8M|E_ZZvjhGBBHp3Aq0e;7Pq9do>^%x
zB+nlJAcob4f=<a8PZ`7Zb29GW$vc?QZUjen&S79+AN~jY=hPv;hiX*RnMVkRsVU2$
zbx`zo@-u{aS=Vx#;TzrXc=RXm_KrhSeT(n~su`10@h$BS_@MkQw1)rHUkrpN$vW~Z
zewSE%Mg42lImA`g4~Xp~NNmSKPoeaiH&(_Y91?%Vot<K59^>f$2XM60J*op|UB_Jj
zIvt0vYvQuOvWJhYn(7~d2AE~6z>S45ygNe(YY1E^<t~!im(Z@-ca|=f#U|A?2&JV>
zR|`>fnQy#S1{E4AmlJd6V_lG+%%qlAg;wKt+kpY+j6M7I$Q~ic5XL_uz9{ETV8v`0
z7UWQ@(sJadvxyc#BfOH%geqaLtNYEaFhf5PV^WX>cRK9@+@(iS>&5;A+3$<V%{#OK
zYB(CJ6)NM7BG^kmZn6jBYS7+c{e@}?2V&M=Y|kZke-u}U!EYN*5NIbY>%J{1{po-3
ztE7Fr!W#8>)Lx<Yi3oCX0HEm;xs6K(D|g6MUulN7%iB)XL*48f0xz2Uy8(!{+kjF(
z%+BBF%9&xenf&u{-sVR7UI-7=SwiVx(0GL*OI|RGGwM~zbaKPr4Hq(&+$Vfn_D}1|
zd!XF-1+kcfL-%2YU6mLWg;i>fCptg6Jbqc!W&E3m-t?+&y(3@@=&$Ulo9Ga_g|DEL
zMLg=Ccp2uTPJX!m{LlU4C|s9W7-3(P3K@4k%z#`J>HJ}ik|%S!;Jnsr{WV(;V4vbl
zvu@D#3D(ugb@IKsYryn#h;7+RsikvgME=3!G-AGT*Ovz@VVS*Ok<sX%x-QR771_+6
z;*#>qyJzd&fyUf81>VLX@(%g{-&F%g4N>hWW7M%w<*D8kfvWs(XXF0FpZ_Cmw9w}P
zN*ec2(8amy9%hAq%iZj&L4Dy_w)x?we1-Yw?_JdpH4FsLW5zKW<+)}-snSH@^Ubs6
zJ=7+tpS~p#2dDKnDE$>+NsTD(L-GNNB$;UHJ=U^f<XxW^%Z9@2V>=GXLG81rX3P4q
z`_<2`c=O!KuHM*w_wv$uz)Pt&bBHAE{x3jAU6!O?R)%n;z$3ixe!X-(@uBgoFn_#(
z-xI}k!#aq3srtg0c?jeB<bvckz1lX>6n8oIm~qh+gg3gA@CDPrKUk`&5>n!>b+dW@
zN2t^qTSrzsC*ubmY=apy$$ZBW&tAsw)r$5G5kegB33%@%knd}^n?BIr_wU<Jl67s3
zWXKGW!~Me0j;b({^<MzCUJl3$qVHi^%xFh#UQmijxL5v2Pm)RG*$uyh3e>Il$Mq^L
zdlAL_wTfngWfS)?L5nNc>!H6!7+c!0KkfiJB-_TP!jDj^{sOB)YgNM`vpll1EcVC~
z=_@Qd5bHJm1^}-OAwl%L_o3N-+-w8&I)L1rX?B0HJ!E285mvNL8?b&FG|zke8%mcM
zcReuAJ(ZeA(f1dvrF9%?7_<lD?`k9f@nt5&HmT@v_u}C%#@6{S%`3p(&&3HaOW;#O
zy5JSqgjQFUk428{e?a>}-n>J`H!ivE&p7zD2X`Crzz&XUF%sbaBvIw?{U;JtS`BCn
z1nJRy>PAoY;L91epQ`=-S^V<s)~lr+!X+goAhmib_>II49DWX#MELJ=d5Cx!NpejY
z$yfz;=}TpGqFLa`57+BSAEJ%gDXna8I?(yxta}2VGThNeeIVJ%6p=>383L6*`An#<
zC>p7K$DbmfPPAnvhzp<APO@v=IU@Sdp#HTeyg4xz=E_i`C^wsW^!J^7ww-m)M9MC@
zAl?7B7$bic0)(1J`oeR`cM=<{(7A;Ov9+mtL)Dd`EvAh!C(4#U^J;e(IkB<OZOxi2
zQ;>F`PX+cPf#{{;lv~qy{S7eQMJIx9F{EII*a@`A4*^NmZ}_r?-uXFZv-!Ky4bi>&
zTEI*lrD5uOS1w95(yYfd)Ivnzr8)PH!Ih{rL^h!7!hh$`C+@!_2m*&I-T?IWp-TIb
zOhK#0^ZOcoKg-TVY>W*85HXuvHh%e6+81w^Q3Z<B8K`7lELLnvT0%qik8FQU<?X%c
z?15+4%-Py2jP&?LcG5#mZDyJ`_9X<{FcS!Pg9)tyVdVU^Jj&iY<o2V7XSff&=C@*!
zfN%*^5AF50BR6hyN0uO6?XJCjk|+?WTIO)UaG#C@8{-}`2;2i1iHNUD1vP>ck+}IZ
z-tu=pLS^ag3U-OMY+IX8o6x~6;2(5hrpa`6YL9*{A8+>hs5LWZBS`^Nb6k!~GXBRX
z59L+&;*?Hi*e!#18G2vSpV~Y}g0(ZCG~9bQNzx4ZobPR79R6Tw?wqvYk~H|Zs0jL%
z>h2n0=Np;Bhi?t2yfBQM6OayAL+o0BbFt)6Ep^%hO{oZzZ*d0BKgz7K$B)@NK3F>@
z``QE~epcumko+i1jG;lm|EjvShAO)m#=7(Ke2M$LX3OMlze(=eRBlpBU4ZpO@p+l}
z4>m)Em4;3)MA-_&p9P+2KE^>ouy|gOOyK)jzleLjrP^wBrx=yNsutTrn2MWC@sDm4
za47COV7S|I#wQRg(Cclj2A~D(RCBbq3gMARyiL4c?_u~OR~C;BvuY+28>BGPMtPq)
znIvw&pyq=EN#L*g#bgBOXYM3&kIYKFzHFsme409`-{ENT5$<zm>h$qgx)@;b_#4ts
z{l@Ukml`&r1)p=Z$|WL+yPhAfYn?fR8#WV1$)2Z$WSC{5;*7zZPerpFzUq}x7+HgC
z*$l>^urC;Q^V{=KDtECB4jxt}$hheXg+sR2YpLn`i%r0;t!40*e>sgXDYP1~dv)`g
z56B9uZFuM=P@QeY07Dg;_=_!6;#U9aXuX~GNVjSm7fQ!YHZEr4#m>2DNclSQaDb%w
zvP*U0XjUz_Ba1-yf+29Fxt^vK!*J2CzPBb570rQm1CYUURwrRs)lK?UhUM2TuSjDY
z->=pb^QuEHBTm=0F8;1PBllvI@PQCI%Keq*Jg&!GPlOo0&YDk&2EUufBe={enmj?S
z4hPLj4h9E#-PbU<#3bPCu8@R<hg)ltRWWSC0E*&Ronh@ApdWEbW^@5oLiNWA6DR*d
zn|`rXrtw<&BHOMSaVF>%KZc30+D4UA%)1BZCI(ftH9ZaiA2rp2_V4dG?Q6Mioaaof
z#79vrODQ8(*Fw)eNV+Kd5!l;ZM+h^}V9{<ePSFaVYvxbj-MQyirQR)_Dh=z~vh_~i
z2XO<1mzw?aR{ny(!@{E5ZIC@y(od^=&q(c|eFH}+o-AI~@eZxaseq(cX7w}o&~f49
zm#%Ja9zB^l;EB1i+JR*4!H_!7Vohk_nPT4zpCXl4epujw;p5@HsWl&k(~eP?AtPaw
zhMb``DQvK&6UW++MW4SN@K&`*p35vB&^0Z(|Hx1xQP%Ux8L1O9WD)u#U4-?s4R{ap
z_UNi<pFrMZ_xbePwZgQ`T#~!IR+vysShOa$P@T&x$jxu4sn)UHt`{Gd$?9FG;o`>o
zHT{PX2^x%y)*%^+C#-0&MqUoFtfVC?^IVQS*LWC667Dx>MAi=w9$^lSx+FhK6)npd
zy>w&JG}-K=&ePCQ_CryYD~U}j!N9L_V*Fd@K2-5F6Q<_G6CREq;e@J}F??tg3UL0j
zA3Xnbl_f&S>s{pe@pN9bq$sm1UYLCe_65ff4rWQ<NJ;}CMY=Fpn9Ld7*&OS!Gp*;2
zhH6IH!3ypE3ytg-k_rm<AV<t#A}IkodDs7rhIPsv`|!dEi%S8smS^;naQ;jH#yeQU
zh$$1y+)$MbarNJY%OR}^x>KoHJLzf*mUeW7k@^$d@iuXQ|Awb8Ao$T8AP>5FpP%YW
z`)D_50v9j~&?qnyd$AV$O=_KrD|ceUrNY4&`6^Q86qt)bGLf^D<NDhE3T3oGa_4Pw
z$gg&*^XLq#NAut+?9^HWXU5SrFeJ(%v<|FigV>gN4V4qV6ZBY8V6_i3^gv<Kysh=>
z1I4tdG1-gc9W~`9OU8AT+pT`Mw17VcL>6j+Z#!(zw*koB8a~hXVjbYyCBEMN_4JMN
zK{scA?*e<ycmgBFV5dd&`^Mu3!Wbr5_uLHYZirvg%7g^z5v)|j7Bn2)21sU*v$Sd<
z1M#8d=yIpB-i^K!_nWH&<91IH1NGUMY0&tW*g1Xa(jN3NXQhDWZo8nzcQMKK8&1ET
z?182FLE(h)FEFl?UI~epFqXemsE@2W9+5vj_8Nv#Q8O4&9~Ix9Kgr^!{3*iP$GzIP
z^C+IOVXm0c4NlKjEEQxRF|pB9#Fm@$5I*s#Oocl8DsJg1-3e7j{YsJ2lK22Ua?&JF
z<}w@PXD6o^Kb~mG_$F_suy~O41M|L{;RURdwEB)hbsfF2!2~UViRDL2^86=+0}YFi
zbDe>3@)dX}eD6x;FaIl&&DvRxe<D`+oH3~HtDcC*&P3OuH+Fof@m0?fUgsH2gk1P=
zY?e)$lkM(5ht6;_KGZF8*60D|M_oIlI=`Enf|g5B`LOi6;B^hAfMq8Pud;X%yen#6
zkYu6ns0kl>P_yCF-Pf_Fns>E6@j12#j1DBJmLMdvX<<}v8q#&V<6`}1+K2k~KVRRC
zh)IZ1Hctu9q-|`cg30%ucA)HVw2b06$CGitq*J(5wR2{7!-4&a`=<7-<3ftSTKW{`
zaRq~2>zH<ZX-WlC7*Ix+-jOl-j^^!C--A!=7hJYa^mN=~;2G%asLs9xN*c~EkT(M2
zK>3bdVGXG6f&lkkiWUx%s#(H%D)JZPvO-gkiKbuo2M_wPpN>emma=$iP)z7dqy1-J
zs07(Vu)_{2m@(fl6mYYE!YiMbj9(osiG_If9N3>06sR8z7K~%cjCLFgAakwe&cW&J
znJXn~3)@1Gvr#TqhT$9sqYXV!a|rM#d}vuz<ai6F1Qc*i2L+ugr;}7q+<ju`V8A^9
zBt3oY_h;YXx^e3K3PPOKJw`TJ2}MZa6$?F$qpOaScvgS(-?4U<TD{-3-&l7#P&aM@
z116^GwL>EXX+sSIRt8xSBE(kpxx*YQPq@_P?uE6$XeVncx|hAiLRm1UqTJ|s4P^5P
zTfJ9NIQa@Bj&?*VMs(qTzd@JPE|g=*I}sTcei06C!|~`eA#za$=VqgH0K-h_TS2&c
zZty-(rb(STi*R9M4E@p<rqM~3f&OL`5Wu?0P-@lK$S%IVg7E|CRLSBNk#&BCz$WWj
z@+7Oa7n*ejmK8FPJ%ZogmZk_aZisbu<o>7-|GdT5M+{4XPO(vO@=P!xIsU#<Ymq?F
zG&}fvVa??<K85&P<^6+Ka_dxKcIth9aCdVaXGUVF;pD66<K!QE{gEG<ue6M=3jD=(
zvUo^bzJ>_eGVcHffwtBR70)ypZ3KuSvHYmSt8sR6-SPH^f^K(hNbToCV;LaZ3i?5F
zS-bo-m?RVOG}lEn&v?0H!#UxGcDXrNJcK!`8}Zo;P;fMjRqz4F=Cl{2mp;xtc$i?(
zBJ-ebSdoMff$9Xk=e)%qk-<7C^TNhMw}}>UtZgZC?l1m3czv7ysqXcpDY8RS9xBY7
zm!9Wt>vv0G_%#aG&X0j)7zmIT7+@Q1jwGApe+v^J{;uaSx;4K0GA!(TM?mky7!aL(
z)qgv{oh14epfdeL^9G8>1~`=$KpHYeyGfSL=sVhL7>?gK+%cLbdMmueAo>a;Cr~#G
zcnZ;lU0~en&-LX{&FuTD3#{|L^iCm?dz3TU=EZ|;agQx$U615l-kZ6Xi%t6DzcE+;
zCnYem$T_mmDLM50(igX-sj`zg+&3x!XuO5$ho&BV18nF7fZOJ8KceaU#U@t<9M$#%
zjPINXS#O3R5Y}L6*)t5)TYKaf>;8DOdA8*WQ;3`vm?lc@Ps%cwzVx-sh$AiQ#0zXs
zphiAr28?IXb%$qR2h>->we?eFovq&6*kr3a-BD^b9&YsFJU2~vkbkgQhOb*UV1iyP
z)F|OvQ<yz+*x1!AbAJ>+kgJ%Pr0Ow?{jbyIXrD+6*FwDA`Nt-gh*xnXMg|u?KNj56
zHrG!uC;)UYEN@PC<A9Y#wtL=E8~)DSM4y~R3;t$^CJPVMH4e}lYgiE|3ne<LPy>wA
zyXF0knV(W@d%SaIk7^-bE2Cwd6nB*!Rgym%R+i271i~YLer+?nL^ST{_zL%&`$Wfn
z^1GvS9dS36+vP)>9u=ME5lRdjngGcxD4`O@IC_AIx&F(ReuEg7VDy{o#^{txYChG6
z>H=cM0SHs^<_#gsy2>K_jRm8N5f7F~PrD~CoO!TtlLbcR&8lJ4JW4k9l?uM1q^0cT
zg(H|VwQRnA=G@r-NneEhgT7d0LcPp<_ZYHQ3Dc?({@nMdk5t(wL+em5zkB;>Z}(-t
zmyy#3?K`*gkHr7RRownBT*Y6jGf`j_F#6Jzuqcg=XJ#V9m*sQ)G$t_E*R6kE&CMg1
z;Xz}3wm|tQzd?U*X*n`KK(NDA&|!XDs0~b1<9CkAEfV<O{J*6!4yaIo$iH<uFj%i9
zxl7+i#rmDy`1axnb>aQtJ^6F2)*YzM9y`eD1thqOcH`EHGkvzD70HuJg5L55!%iIi
z4&iQxQCa4#^S9r|?2+AV3FKCUFH#kL=>t2z8j4ofqI@}7Ghfk6>do<|4lOH+X|?au
zkee=F8Kx5f3e8qZKi?EN_iyJUd3>|SS4Qc&>D8t4*N^?~W$jfW5e5^~$<&WRC`?^Q
zq|$o~e(BwY{bNDHYkK)B(ZIHdk8I&8-0`fksU{mk+ps<e0K1w6j!8#`z+`RB>FUZp
zUA}yi3FNm>W!Gi+b64^8+n_qZ?E_q2l^Fd#fO870!(=~t7pK6l`gU=rp@|6cO=(Z+
zt8O=4*o2=vxO88R)Fs=_bbf1*)!mg@s@7o-)oDWy^H+$xBu-qo|3ya)LXdj6sE_l3
zyy~aj0@X46B2=?FEKQaWq$^oyHE*NdFGo3BcDO0)tdZP<!Czsr#F{NK0XV$wp8f1B
zu|LT*%lY*_L!`4<jK((mIf*BYpD(>hC%($K<;#wE3z4ne`PyVM$um|FqXrR|Dnp&)
zjc?o95qXxdbBO$ndf?x#GJ%L0-CIQ_M0#!1^cHdR1<HkgJFO}O>T3yR4B>IL5A_m7
zGds_D`<@jQn|I*|)+`xuOiF5bAN!$!dBa5nIPE2XQa~eeWv|JE7(?fzX!NsKrJ48l
zl-b3Is{K{)W-R%#q)Fiw3vyX>e{(H%q$RJpr<J?HoYPNhai>-rLJT5F9;oiB8<qI+
z;@FOiaB=heNrm0e*JB(9S-gM(B>xv1z_JErHm=ZsY$OtJ60Osj@IrJIs~=I!hgmWm
zw}$w}=1)6+X;4W%iaGo9JS(1rEe4RV$|Yth>)0Jyb7!O4Scl80&!tbEgR$cwJQ>3f
z9-tminH&hY*~}6y<)4&3qHD4jlf25_&7EOmRvj<`07x|$FIo{b0BAWclLx|<mDTlp
z+ZYw;X{UMEg06QH3eg`J$H6}7Kq=nnbrot!mL84t%@ck8_<E7Z?|q{GXaNN9(?K6n
zA1#9JRx{2Un~Tm^_+gTI(PT*ID(fp{&9ccx4t#<fEgm-w=*S^iEsofchUduRag+y5
zY`%ylEbU}Vuov_6stVhzk5$vc;XCp*24%MuXs-^f<J!kwAp~?iL5+1e&aywJl5+4B
zo=3A~pFo1Fo``_&{#3;kaFK(2UC_IH<p>ZEI^XM9+NvTFUGnMoGP~Zw67Y^%pbF4#
z)6aBBF9&C7W)>c5?b|qAd^LLD*b~jQb26sj1@$r`rkGfNQI-<r{)d)HQ+xJRYwXvH
zZRY$Ahbu%jMF3A=9P{7QD+6Z0Uvvd&og3&*eg>HoUCE8rGe*8z$WM)=;Xk2BYmI8Z
z?;%dQbY+~8JfM<`_j6o~qn{c?e+TN$j^yf?Z24TNGBXc{xFTyQFKRl!$Ih`tX#tju
zpfZ6P5LAA|^shmvP6*7$Ot$Rw9)M`@E!F_`vt}IK5|nZMD;pIhZ70VM?8clqB($_*
zHTGp?(If}m=7BVOP+;;&+W3XL8eEx6517*u8DJ@bKB2q7;5SSwggaQHxx0TBtMk}p
zsN5ZqjSjc`0hJzHG|fTx$2xW9xvb7PUW2P7VpDKs)olJSIve^oRup}L)sw6h#xNx%
zX-LZ>oo!5>g|o@I*0suP(8d2*p#=z0pa16y?ZAuT|ESRZnbiON^S>!d{JjqUzh4KA
zHQ0ZjPi}z5qu{F^fPTP?Tm>A2-wU%?niBU*?@m#a%EVh=c}zflm3GL)o5!xB4+9(D
zDL9>(MK(d>fF;4tFq=RS)$UjH`|EvjwpdDU4MMgXAGy$ctCllRe7uW$@&4b;n7>&<
zf5*VzG4OW`{2c>-$H3n)@OKRSj~EcvSZGsc@?Wov47V+fy|w)*7SOtk7^<{l`aN{7
z)n9CpD!KxbiaLJMn#`r-<2_>acIj*}@0!phsAAS-S4UPAsMfpqcIFuCKzZEfZw^X8
zFb2A(@Z;rc{sl=oiU8T)J3s{hNya3Ai)+S#=%Evs>_E~{P!|P2Mq^JPipC~WWH!Kp
zOTQb-2G!%4Rq$C<5AaJZhtrbb6IQHI*t)JWv^fl8m<9k%0@&?9KLdaE|7Ra89ALq|
zBu?gLC?bX$SYgpBEIPnM!!ubTxE+1ci_L&M5m5IRTjRjk`t^-F+$D~+FmfT}&c7|A
zPM|OIv*a-}3qn~KlV1YC+~fv*0tD1;_{&!2|KqbFBWO*Z&_n`<F0VIf<p7G*=$uf%
zkpdkaJEQQTj7Tt5oaUz;hcS{9UqK7$V3AEPk-MU-FWz*;o@ce&z<<>!`0pj>zS?NJ
zw4MFugZkQukkKInYmOXG8=MU{ofoj^fJz*zs`oFp^H!VR*ZAuHp=7Su1SIJI`_I|{
z==zHwXjpSzPqbmYon1Ls^3B>a%j~l1>_cEws(L~t7%iP>MR6;3&l*fWdZ}o4tld{-
zkCr*aE`JA5K>tKJUQSx7kw$qS4t(4~59TG5@-83b*4X+&S)fKzI+v6tx%+i*=e0lU
zI84_}<#9i<Z+%bF_*%raEET{yL6!^yTK5C4GSYE7vP0N0!Hd*J6{YVo^TG^5G@BFP
zO73M+!fuW+lSE?T(%yyJw@l;TWP&KO*4MuCS&JASldjgg_@;%CP%}mbiaVX-HCTIY
zYV{}~Z|oRzvdLFA3Tr=TiSMMzuGDs1;W5ob1vO6<3G@SPj#5i|HWXA3Dh*X+q_v7P
zO3${eRdC)yt++-9tK!5|T!)W$E^&?{FuuEealzh00hh&nH1f3z8!s};*Wz()Dn!7`
z(3!-(+`d2aN>}n(jdIT+iv(*U^TI}fyf(X7En8p9HUt+-4fGBDDI2&~3^hN^)I6)!
z-PEPmamVIkRmA6g*VWmX>ep?I=_){q!*b~;Int2wEUittuREls_-34|mKVH}g0oKN
zPlmhUh{ZR6E&*(I^l-1bk$0OrPGaywvYptugJ)a0Y9zoQ=^)sT;nY-MD$xOA*|Hqy
zcSBPlG&5FxCt#5;tQJ#!Xa@Rc{jIvHB-|~{B;K|kB#H?hl-oxEUFLG2;z6}J3BEBP
z7aI43D8f*~Vin+EUZP(Et`gH=&nkX~lP}GFeL7kG1?ZgE#IdAOVc^_=#uC%_>A-19
z)RGq#3ljlRX@xR*<1A~%o|>KBv6HS);r#;iDRReH*aN%ZqZ!Eeev`YZW<VD|NOnQc
zk*-5h#_}x95RJyb%qTMtICMaQa0fj-ykGahaNV;qyDLkv;&&$H#@%lC0B5;wnl%Me
z4->D_6D^*Qb{n!N@!r=i9M~r6r|ozGjVD^JdD2x_J)?5WB$k{r>rnS{uC{wJfhgfp
zAu!zZ=C#4o_dd8;%*0HL`dGLO;361j{U{!E5uH2zRVvVEH?B_#Qy-|Uy)$|IIQmaB
zn52?V{lA<9R@+HH!;rfP#?Ly8Op^(#P0%_px5>Bn$7TWVWX-MtZ~*dur8qK$mO%T3
zBL%RDajyu@xGm*TRnVH5!F>eG1e&}Hx)(I4tDcOWFjXp2&MBj-8FWOy+&2~A#8LJ?
zUgWKQp!|UmYLIuJJ<YvJ7-4mJ0~}BBBlt!ILvVWaAvXWaD}u%$QvK5T$sw<}FDR)9
zHU~fz;H-LHB;XATeQJHl><4-NBn@U7RFg{s=K`l3?INX~C4E%GqD!L;bR<5i#0ls}
z${yeWticBW%d@0RM`w+U&W>vKcqfr=bpKW=2$!p=3vWh_LsB%MMrg0@gXiGv6qwtT
z(EOHbQhMjn#G9}~LE6+QbM6gJsYrE~*}Cq~KlW#L%e<8EMtea!cf(__>mwv|LC@76
zk*6_Qk<&!DR*~9n{PaMZJ-|huup>UKvt7R|{4ampeDvmFbn5RT$J*7T*tk8f{o7Ey
zXw_Zv%!EUT^F*P;jKO__mG+8+pBJTKQmhr<3`MX3!xUzgE+@%7!f}F%DQp5`pbwvC
zS)V={va85QIai#DyM_Zorm|C=PwA%jI}}S|+aBoXd}(`lJzS27q&ENS5%dYLi!Nd(
z4C^5wR4K9BJ-6p^R;9@^*NpEU&vd%Qj|`-syGMEY(6atQ6ex84C4U|u;*^q)&)x;5
zg+L$-16ZPW-9>0AO;-AGDuG_K1nRt^k5|n@OHVeV@AQEM3C$^s4RKr}B8)&b!9-i}
zN0&%nSuqXO@w0Aqd>eK%#B%p06*rXE{JT#k{b8W=zw0Ei|Er%7DX@}FhKOM!#Lm(H
z%@IvBZ*iY%$+zIW_f4L)_Sn_zdT%+W2kFDf*T=SW2r>YAUohuhNL)Nlzv}|w6N9`9
z-EVeHbWMFOH?lb6aGHBqY(|d%%C-kVn=l>;6nhOB_kpZY7r<^*5Fbi*-W@yNaPw%t
z^SQQY@h|J*xHc?k27`m(?uP|pd=QGm%o|@GE8<XNd#ztqR{@Xy;(}e}`vzz#DnMy9
znPsbg2x_I^z;gh9D)Dsw*#W9Ojo?ig7+;Q@lpNaKtE|!>S{l6>(yYVMn2%pWFoXy|
zqOu;bOjyTd|G&o2b`@oRx;xu{dK58o(cbyMh}@<iFiOhC>?&6YA(R{ps!>B~Pp3{O
znbB{rKjZP#MbbYQ%TA%Ad+}&+3D()xsG=H7Y0+6U{g!)`pULa}ZfkdkuC&Izf3X($
ze11o(Cg>QDb-_(So0H%U2Tpj`{;6Af?BBghwiKwV%woLzAk4H&*^9tGv=IH#^iO%3
zr`Y)*&szQXMsx##Y`)|}Euh8wOca{bmsXxUMXVx`c;gNqHFz5Ip$MuoA=qOj(LKrV
zP+M}q*L%ix`dh)D#l#*A(+A#z$2<!}#9oaR<C?yW_%)Tj8TxJPD0iZ@meKgM3#bt3
z{kObW1g~_h*YHd7Dr;U+(a6#(Zurcl^Yg~zmK@|VY9JZW9B|X-D>){YS>R=j`8v`m
z!CLrH3zv1=cXnQJO^3zHNl9$hHf!SL=bYRA^D8V2W)}0@RVn+?-K65waT!CAG4Mn}
z@F`<pjrZ~<NavK;(C@)K(!bpI|GV?x@2~&8UjK*00DC;Cb4vt|-ChMZ{yB{;aEb5E
z{fq57<sz#g7MzPm2*$b~l{!*d3^;r@p;;~hkX0E6<eGW?W6*C81ar6LH5l#H&)oU|
z=4D~a63>4*+362T6wV`{$##Rn&jhxZJh<G2odV(De<~2{EywK*JqHL6pe+sX5YRv}
zG^=Q{^)IRc^0m*P;li#KW44RH1|R=`D9PSz0cKv6Q34~P<mq79yLW~-Rx?%FWdlr<
zfF{uZOu(LYm#%}FY1AyTBb4^qYi3EI*Q7iVKkW;&Y*Z<?R7E?M^L_Kb^&3BGhBVXp
zyCB+8*rE@Aqsjpj3cM*9m2$CHZV+gVzBhTIns8{tTrz?gap_L?b>S1oYHy3N`C6u+
zJI46!Xx3z{x-<zRL7oDy-*#A+N&jyK)4)C^yKCou+x1WPU<z<BsivQ~OirAX(m&O`
zn9$iUdQ^N!td?%hnGFZ()nk^j6rl;TkQ&81#6eoFQLtqhKCUA~!MqM+la62n^6ePe
z0D48N!BAuy^@UK%u!ilBiVDFKLK4G4pHBOFRN#L5){$2N8Cny^y4CM<nDzwi<8XAA
zGF&rP^bLfAJZ8KEzIuV70GMlzVEV#)3movfH^Cb!BEy+z1>lyl4i3YMfQX|~4^c%S
z=0QXnoNNYdX|HgA>x~4>UI>I|eA>x1TPQxEi&YOMQcVo{Tr(?6s$NW>rp#-n1h9pV
zEudE#kr4l^&vyK9_s|frc1<_-)NONi-MDkG4uuwMSQ|Na5*Aq{E2-9!g{b~o93{0+
zfbkD6D}_m8#$~#SmGjN-_S&XDzI?Qc6-j+1LR+C<Moq)QAt`@nQlzu)Zw);I<8De+
z3>S2T`3BrrS812MR0iS3zpgB$+^$G7*!}G8-CJGU#9KfXOGLV5I~+a&|DKh%4DT~$
z&FzTK{6hC4R_|BYflQzi@fAik#%`&lO0w#Gu-k$Ilh6`~#-l@EW{j~xKMWQr2`tjt
zZo=Y@I9`SlTR_}S=7LCuk?*bq6YsYK>grHrjy-dsmN{p)b{A<GOpGbbd#EC^H@0>*
zK^iSShv+X)-V73*gYZQgdp5XUTr+$7VnCr6ED5N9K99vHqhaXYG`?R-4y$s0n6C*N
zcl9s84<NTya5J7%Zn7CeVGOSo?A4J%{XPJuUyAI|&=uCJS+)W*QQFu!vdlYgm;h4L
zSLxLm`sKwB>WmKmExJ3JSgEpz9q+qyTSG<6W$)!s?6+myq3<IoZYdc&7y@;q2GLUJ
z*HE>lIF~Xaex&#L6X%5UFYlvx1<7dObT#ZQf*yl*F-|-gh4<j)XD#X{>}5Cd(frIx
zeI8VgchEo!(X_B3ptWe(Lep;cEUuLR0;Nt_UzgVNU2+9zy?ZSjkMpe_S#)AGJUCk<
zJ9%(Xk&-g5|8~12T=#T)ZsR0V@!Gq4A?tR<a{TB%b|$~`c1UTyBj3n-$}sUq#$>Ym
za6xVcSApf)4E;E3G!4#7(KUXIgPxN3d5sSp^u52B-S)rO`|@z8|Nh;PP!!2lwkbp<
zDOt)kNeGFw5HTd{gk&2vW<=S-giw@Gl1hwSS!S{)N%q}flq|!HWsKQ6Z{PF$e&>0f
z^E>A`=f88V>-odgg~lwO_vf|Tulsf1)=ZC%l&AB}=mqwEMj5E`6SE=RD5uGS<{IGj
z_#lFJrR83Ma1Pe^wA>5DV^@@(^Ry{vX};TPt(34^)An0y4v_C?@n;-tWsCD(r!6I_
zvw0?*OlQtf7t>A{1&7}Hb;{4f*4Q<An*)TSOT$vasezxXq$I0}O=^NdzlD{`q+;C#
zdi)SkpwEoNq6F)zkNtr(!2S7!_1m7E;RI%~?=f^(B5V&%?vat}KY~=Uzh3%;U$eXq
zJfDP?`{t^vfWR!d@I`Fq_g1UA=R3+_{)yHTfgKR^S0pbP8-YILg0ya~zEGz;)}p5h
zM1*<;7ZDrJ2Fl^F@G@*-W&?FDTYNwb9=iA%88xxCb=Mwr8gm1j+Fs5_b&OXc|4cpi
zc&0WjDPxaUkrZUSfv4S9PB4KNtmygxw?7A1WRTtqPk3)(dE((gUag;_ktcJvPPk()
zSfQOT^-gGYq~erLM6Ig8L$+NWVtv=RUy7<wT3T{Qm5SUa{70gn^+T{r+|D}AXlyVP
zD|6(8+MF|~tNE%ZQhC#31IrtMppwb>g%}!)9@ujP&koHsc{4l?%MdaPN_8*3V0fw3
z__RDIv9{*XW?~GX+#g~-@(21Cf>agVG`26j5#4{|0Z%Ht)LDuFNaUoJ>L%T*QDk2+
zyWTfbdXAP&DW*&aVq=J*LX~R4luN~1$FN27Ti^OTlCf}036AyURr;Bvx@0}*PZwYe
zP~GSWR8oBHmC5;gYQi0Zof@Rp{d;Zs#bvXtAn1=>Rw-wrwZrya6nB4?1d)F|Fiix9
zf-C6yZVYiZqp|bh0LHDj(j&^IKbN&~ax%qmzC7xS{ttoQolUUedy}CJpWal_)Dz3Y
zwdsn{W$CkLhgL^2ml?!~*YGW`Ye5#JrZ*5*yPlmvhvAz@4~bj2D%>7tg%lBC@9LOX
ze88tA^#lIu_%L_^S8FK)PBn*Ni@oOWWSVAa&Dd2qnI5^n5ZIO@GK7o4$P?t*_QPPc
z2%TE=tvz+;i?22Kz5DLZI;~+evyW2YO~j#?mN@Tst?w9@u02dYn?7!LHgI9SF<f;`
zhr4|Qgr(i7WKB8%00{f6#GMZ&>0k3X`@r$rt}UY2I&k)4uBtrJN8~U>EPSn-9O~XY
z%752*ef%p}o5>u(KO#<i03C3n8fk*O-ndZ+30DU&X%a?+u-|9K0_@fYg$LR$e<Qk=
z)Lc^<)UD*1y#|+pd0WRod1=r`>`ShB&ZUc%dzCNhVQoI)0VNJgUdvb@)6|`54~Iy>
zQZb7Y#R&;<M;%=gt*iKJ0X{U0ZGsLPw>l&d)zS;5tS&5jgd8(gn%}1MI3nWAeISC0
zhH$FTJ5Z)I@tx;s-!mjuTW^I_=?JErTIJOrD4QFmzj*0=!YA|Lg7)2@;~S+;=O^9H
zwuFk!#&dl%iILfcF)hOv;>YFsR_iV-|8%WNvGi8hh?~PFH4uZiy2!Ig5jvD6A6+FS
zrQF$As%RCsKTIg>?DLYDcbYuEL*Nwx4OSMvWbVOZ33AzCPmct&q92s#Lr&qX1@z;&
zu3XU?Ov867{#zqg`6}lRWcJm-Dn*3(`VS<?hIcmgd^%0Pw7OM#%eQa0A#&zL{!?~M
zz9N%mk!vt<WIXLidCj&=bEUw<rY&c%fk~<Wnx>a-#u_AG7t!<Idg~ro&6c6sKYnx%
zamFp%L?AkhUnUuw9szR!R8}|C9x5g?Gp5Ox$4~JK;%1#`s$%7d>wX|_@<}oz3Y@}V
z)XZx`>=9J^GYKjD^_e`hCQ6PI5487wgjl~vdN$q}9hR!E?N#2TI>We+sWoW6tCIkm
zNF&O#RoxEEc}d<)EWfys5}wJ_4*uN6W7Nh5Huqxnu-&z)G=-*dq+^pbn&`uSzEbSN
zd)VPw#KsG%e-kklY|te1TdCOg)}AUDG+!C|=5%S_s#oV$AYARp+J+vj_NQE&%;*U=
z!CKLZvLmiLEJ9~hlWrO%Wta1;72a{yW>idcMlWqa<~I#{MD!#ysV_=`a`XIvBRmyH
zb(k{!H1T0K>^86w1D1_MULg69(e%YH4*<~ROS{Z`zNS{<ZkNWk?d2)yIjQxFC`R<P
zegNIR=B-&6*rJQ<Q}9i_R<BW;EKA;Tq8Y7&?toRi1`X-LI?d9g%i|cMrYThWou~Hp
z<w?PLRo<cEQM*`g*iLphO(Sg0Qcb{N!1;!I3d7Oz;N$Fk^eI>(1Y=CvLG%%M_y;1D
zZ|N_@Xt(UekK{cG6`H}2PT?=(XFBQ13@N~D6+`zLig$Y!_mzu!-pQyMA*T#+zbEPJ
zcz!z&{==kJ)qiV%vPH{FQ*vs=@wS}pn_ao;1`V8tB!HlgG1j?62V%|~7loO~B`$}e
z4FHnY4|EPO|1<hgvC#*@E*vb%!lss(+!z47-{k<qJb=3kR_C*${D5#PfJ}ZX17p}&
z=(0Y*!`3)JO&7t$^ueZOLIJJ{9&Dt={){^P2a-$%-{5D7t0zuF2}CNuVK=m3+cDLD
zAl-yi6<(D*)DU>KB~VITqGqu09M##q#NM{Q>akyLbR5IN-|8tNZjw|H77bOQ%+?@J
zswS*;O8h1cr$=3o`t|F$`_B!c*AeYoEuo&>#AgO)m})vxYh!)i66I=jr&G?6pC5dl
z5&$q2qBm&-m>58%%&0o|$Jec?buV6PzeD6lvrI<cANy{>9y|#JEqobR-u}fnG~xJ8
z%^ww(cYVw(SxGxXNOAwF{@RB>knNz7Ic><#D4kU!)B3aWzp&Lu^6l#lUm%uM8emi_
z>tWt0ew_R}?@2Fm^hDVGHKFN;-@}b%f7Q;DJr@#bmh^i=DhahZYA<Rq?R^DtlTf?!
zAW(G?5)18-kE~*Bderw5OP~5BO^r3?irAPq1M5fnQBFmz6HAwUw&zB5pI@;0HfM`s
zlJ~mC-YqXD_N;q)2<zG{{mOc4{C<``kiJ-I93cHQA%9<_#;F`3&5N6UP!~lQ2iHiA
zPGAr=iEn(S$Q?2{KBwDa9R8xzD9+SyJ$xufqy}iKBN^e;=M-A#qw*L0wi+eQqwn>?
z`NF_ruL+m$=o<j6Po5~(CwGn$;4_)*lT`mmf)bG8RF_7lXXeMRdE0+VnX*f^$qoxq
zQg>h;Z3TLNwB;G|lqp{otE%c2+o6g*asKZ*fzIgu&1e<N@TS?)Zm|!MePx{LR6h<6
zjj4VyEa<P3BxYh`;X-I-bUWdMWae1W%gJzOWtpRe2i(mRpC5U9dra4ZcmXz%1*6&^
zaPcQhsH7>mIbv)VY|8O%@$gX_&%-l0F2}({po@7Bp9+RxBUmd=8tQN+I$g0B<^QgQ
z<W`|1QRQrw$+;_w$_JTBhU_2bEWz--Q3b@v_;EQ7l_BjI-Ijd0fJ|zkCs*(|Rn_dT
z3eMB}Ui8Q>hwAFh5>6MG8(7Bz%i);BY$prWWtws!=HQ4Y|9YZ1tZ~nmiwrv=!;ZV%
zgf`!#t*{uryY)i!xyB}CBKDqV$#oi#$(qOB2ODRwDA=x}k|XD7i%C^Ux6{i;-=8BH
zWg|xRKRdeL+|7kn4W#&iU5+TwIBroNtOE7&(v50jT&<p3-fREnU$~RcRahpei`ct#
z^}#Hzpw<-eekLa)h7G?LFa_nmey*fy_@=tnI~AFz=P3eUi)7JP!#ibMD*ftq=|dhK
zCn3V-T4s%eeLd_eFNb`XS9(LMgQw4U9aT7GcWG2Qss*-u2mxRFsl$!%;*I^t$tUa{
z9<O^&64lp}>o)(XsfNBEH6s&{{i&$2d0Zc~nzj1OrPzur1E|?)A!nzryA4gvRECL}
zB&d@RJi%eK`?zT92feBvmbh7G8VkYgmtLp9$n)VsjonV2lX4mdDGNt@CE7kU9Cdnp
z=fWM}u#U3^aV(m>y;m!`SpHPc1W)CoXL(m*S<#&qYltN*R2awy0d=6ImP|E5+&z^v
zU(A82ZSE*fpUSY4zlhHZttVoztKnK&LY)GL+r!7kerY}UcySKx43lQIgI26~y%`Du
zS_$c45lY7`;H^*T{Ib&{nfA^D3T4M23Gb(OEUxAV;;2{tK%hv(lt2qde!Wqbl2qp{
z_eizF^;q(i(%Rj61ErYy{RB0n>SSj$+YK78e<|<x?r*fGWa20LM~SV;*%ef1%Tk1)
z7AKo+&L2Y#H#E?*%|3D1Om_a@;p5*gv--h^3T%W;c~ay&-#Q^h&c)b9H{I7W4|z;h
z;aw8Vgm53Q0Td5X=uwp_YE^tt9U5teN*=uaAy4XLTYXW(CYJpa$Udnziww25jH?i7
z<B7sqp7-3RKX;A2ZpPhx?rJT|3-?3$Fh0FP?PU0WKsGw3xs?{jmcG2T_jUcy9>Wi&
z2YK;*T9yWq4Ture(+`yWGOsf|1m7E&YfYh#a6W<t&(I@sX-YWJpn>{h#7VW3C6Az8
zqNi{w2#R1}rJ9X+abd}_YuL8(t|3m*RAT$a#Y(p@X8%$njG?^V3V$ShR<|!By1Tmp
zjT*C<b$LAVeBq9C1k7Fyl*6PgMMrnQscFsfu%t6DG5P}qY#1kxE5*K;DFh@2NoIZi
z9<0j8PKPoKvjX8s+55n;uw>uscJ;osny7a%aaY-cUWKa@L6B_3Py8;VKbR2<pe|}9
z<?`o&$;uFY(UuAmcwgLN2D;c;dOUjuJpL+oXt=Yv9YKmQJke`ND)gp>4f{4G=%>#a
zy38<=&f&Y_AwO$<X_gJJow<?L;tcGCAG-r9m3xm$?wk4H@s6QGwu|5@_=vbo;v{Ev
zDN~0+B3owNZwp+CK&*6fg|0UtL+ekT`~$IoF@**aLHdxf#|;ZyYE$7J0NjAznmjDa
z>mu=JL)0m%X`ylk5TZ5d*4o466#kqtQ{WhBk?b~^>JqV>4=A&?IoRVIkui)QqsD~R
z@<<pz$+<9K-~CI;EYx#ficWofy?k^nYf6W5ky<c$r74}DNUKbB#q1sKx4J{P{Ra~9
zizpw2AL+3N?kx(8OOtjHzfwA#)q}Jgsrh#l+n(5e*L`fQ_(rwUP?p<_lY@C#x)ph6
z%hwXGIYJTrM^WkMjc@r%Kx1q{I|CGo(hP}r@bIUMul)nRMrN+-z7G$~PFv(|XCzPh
z$LZdKBOT$z0~D#fnJobd39H{#sX+9PE!IlLHpub&dS0i#0=>S*viHEDAW$d-YaqUH
zg;7Xa)n}O0(7fT^{-X;ADl*p7eHFtrAZT~5PWjw?pbdW9PKRtACojne^p+cl*JADD
zA2+TzKzwoIs$oRtG}uBWCWc_j1?oG68Q%w$E|DBnog5YYmX%w-J@^C3u2`92i?`AP
zXiF5a2`3RN>d|MC`N6)18!46#V|Y(-kCyXL&43GW>$nUfibih$?Tqfvq?()6HxKp<
zo<DH9QGfGJ2Ul?)dh=E%MxhW=@u*9A47B%B-xSv7qkX2ya_>#g3KSo$aY7^!=7|$G
z$w88>u^0N39(vl7UwH7zy!`t7D!AD(;AZ!yKwaOCBSp*}*;W-lHo4vHq@-sb5wIeU
zQ~;<NQ~-68s?(%g1K+{GTMS5dg?>NccNKmp;CZEp?A4tliV-!XVGe=7#b(PzPDvh8
z@<-oKi?K1iJJ5=up2rPIO@%H{rtlI@$aRyb3uVM;&DgwdtJ;X6t%vmQ=nm9_%Ahk!
z1GSg)hHWHk(e7Gwyf0_i&}4TFfG*@fz24bKQ0itoy&9fls&NSK^2Uxb)YY!*-#L2j
zYw`RStvvV(64eZz=sVT<fJwYxA(Tun6bxqVP?~<aZM1b{2V&JULa>7-6{m^c4%)1{
z{V)CBcvDvnboAaAMZfM(L*D@(<V}J$rr`v&-Ix)EIAIc2bYA_YSJn>2xZ=y?9KE-^
z3mg@CI^zo7XP5mfX_r3iUf8J-@m(ZEJ=|#?-~&qBNt{T<@QqB_QFiH@1wgajFs?$&
zzodvH51;;$oBdio0RBl6QVtM5@qd5)`!o3e;0(C`3q~5g+gcLyr4aNY15Zdy9()H8
z_-it;@ld)fh{;A0JuzH`6Yr=v%taFkifsLVIZZ!?sXs@Y?yUu=zLNkq5;4o%h~f&G
zfVH)}DJb(dY*p;q6~^9+qiaipaIO`m9=i5x(`AETcb^L);xf?H06m>17zAIfTdo!R
zu-R-fNL;~^DX|>x*Xnrw){&x=f%#94Y*m`)gyHjGc%n03<#%4h)tfUv&c`tAC>d-C
z#YQ5)n@ZCwS)t+47eiNv3p?ilb}8VWltn|-+Zk1cIrL>o(pP$5c@T^>gb%_KamYtT
z@9k3IjgPU5NmqpP2jep%M4ui`SAp!vq~b-<qkkYgY?UIC*D$fEPNDTuNm|aQ%Ey+~
zI&zmMk0(3-yq#D06LK?6aTWiWIDCn2%cz>N#*Q2GmXjcej1Vc^^6tg}iL58d=3m_R
zZI+#lTzN5;9;WL`7v>O#5f3!uROunml`EQ;7?zvlq$zdwabgROsUi;wk<U_q$Ja#5
zSQwz`OwywaE2h->=A}*3c(z#96xf{`P~la=0;u>Wb%xSMRcsz!i+=PPW!~*yVL6*F
zgz#$0dL7thdP#PmrhwnZY~Ql#M#-kO%Jffy43LoX0YK;({OnPYwc~#v)vz=O`W}d!
z|G<n|5ujT$Ugn|X7_*ajiFt3OOh15RSd{FN;ki$SS#S(Nt@8sw0YX$9=OGS^s|Th%
z8)4DRfoA+(Z|)u{`(np?(#a(8H2NVWaw_&Ng!9awp*m3CF(}DQmL~$AT7&tC88&MU
z&Y+;Sl7IvR_I;C_s`(dC+F+NsEvZ~NOWF$ezJncC0LoR4FdQ2oPLz>^G&5AaCw=c-
zHZ0b(fYHeCz34s?^%7PF{Ug%~$l+)ru24H46rCHBzUg%C#EYW(k|aIbSpbQiBVyO*
zWEL{F@^Z6$95HkYx-|evs3tG&EKdp0NRJeI6!v&4c1L#LTpY!DsGSQ#R7HEvD}tIA
zVORHL#@A2JtbX<)l6{45cOPE`LSK`_h#=xau0Xf`@m$ANU)TJR4~I(<wtPl)tf{Eb
ziu5K#7TPDQhlq{&mQ-}ATzSf5k^gnS^pF+}a}}sj^WD0?g6F>;8-=0)3}AzgftHaL
z*7A|CrrYC{@AxQ99n+MNiZ6=zZh|Q14RN{#*~%qm*K|gL_QR|hhjr$SYUYX+7voGb
zJ@3m;&9{fQt$!T$;$0o#%CAuax#4T-?2sG+xQO+reHB&Sip70?_CZMo&tQcpQJ^;R
zdehhbbfGFBY!O+~DIyG|_12NNEjv$`ow--X72pDl;R5PBc=&*~hF4(ogf$=B_(M<<
z`I^_TNKx;tfnT7rOhBtJX3398;*9qj_+XhgU#`DMR$t#GA4#aAwz5TA$p)0=op+Xu
zdVKZj>ee&=K+?P5uITUJs+eP9NP=ss27%2PWyr=W57WIUi;CN~Y6w5@h3Sw_%UZK{
zThq-k{0ubwk^F&@<={jk2f4$CGIkq@9A3bXw%W}~j5Q29bm;Ap)bvzZ{k%7?q9VFw
zT<;z*_*S%eC+jAd69lo#ME4A{2~rGR@(bI&J2XbTzkemaZR2JB5SZ|%1k*}c<bhWI
z-Hh~KrmRCBz0}l}LyV<eZ|4>t2xJV*`m=ZA=?2|EYLE+!%RbcgDyykFOU&gcc}IUS
zwnpKu6hCo)(2G;G=2P>C{ppC2dZ^rmnbtM@EOzqfQ%(3o!mgf%2a?q$AM{SQXgB48
zjofQjD%7+G?2A8^5UP^qzEAGBQ-3(x42Wz9yw&mmQpiWar=5*%qZ1*92S*q&jUhsA
z6BG>R?D2#ncD!l|2j!%;ZdzS#o1u*_H3Gx%3T*T`n5{SEUXC3fS%ifI55F|)6@y>J
z)H|XMYLSMbeb2SX&gv-m4t^hNzDhidaV_)%tC=N?o7As2%-0H#2U>$X>OqCfi*g{~
ztQgnGLN)Gc0m9f&B^M0x%(OJu!?llCp?^s-x8TGD<?3ri7Hcl{lK45r%RMgrn(Z}0
zX(ZoZq58wjf<v}21q4-${)>Y}SpwkcLcjXH?q9h2&@!!#q?i7pUaqEYv126>4H2*)
zuE@N+bU#6w5khXT9d661cz$E>+Jr*FonyMPGg?D-K<zD*{+jWED*J^XL~|NX+xkJ;
z5|e&zM6USrM7Syi3d(ZowN6PdioZJ<zT?_l|7cisf<|e_+8_j|bJ(%PPtc7r=&z{4
zFE~;D2$4scL+d7C3Mh)Cu$v0&(<wYJ2jI?`7%`Y9F+tiI>tB2ERjWj;hLP42%R^~h
zpLg)I=-G8aLB^WMBW^{jYTiCd7?-P=(4F_G{30{-=#J_v1?o^tlgtu+nKxel*8ZE0
zNr3M@<mlaKuu@(F&xBJQHtp1Rft0tzNt<7-1fEzE74YQ??EPR+Ok%125a@(~GI-+@
zU<WQhVcZA={X0<%HjzYJz5|FsgN7ImWPtJlf&z3fJ_W{4oqFg{zI|r>*SX9djQbh4
zTlzOlZhbL)if_V1z_zWyh6Tuy4<}fMbe!fw`QD_TF&{MamNBd`!vBCX)bit1s(NdA
za*n)qyzXcj)Wt0Qav9oBklV}_UZDH;fX3fu^m^@@U%zdvblgnyp|-#6>qrBDG@m93
zphXz|6PWu!OG3ySEqZz~=lb^cLBa-iNZrvbSt}YR0usRFx^xCdEK8le14GT_LId7m
zx=w+9=W{DaZxGfxJ(u&m_BhK7rO~@;b*EvkdhY%42A0bS$8-Efrj+L{9K^=VV+4s0
ztqeVET0#?uSt?Dc6;Bp6!gt{r;VrL5p)rzYL!q}YRN&~51lxUyEYOS>M}9ZZHQ29J
zP!@FL#*KF2(=tW|chTiM95L>9lDIEd10_RkjW3MzuXIrO`YrI;WxhZ_cISobf+T29
zobCtc@q95CVFQe<P)5pmtV6TRdu?^M{Eu!)n$S!L0N3E}|J!M3th^|7tTh33G0!U`
zp&<C}^}O<1W4y;d<A3+i-55NR*df*=8k_5-8+3KR)}i&lT7Ug7>q9p7C9UUE@57^T
zCAb)P6RAR=@t#fAl$FQ$H8YkVkP*iAoq20>4Dwl=2lpNsdcJ@0$PBS0auIP#e<bu7
z?;hNq(fktpK~5=VdXNi_&Y%Jrq~VHtCP*SB4U}GOG!bsS6qKp<mUn}~FaulzW1xJp
zy9$sp9>IHRVRNaRF8DXI2H3`7c}b8q(umvy2kwi>Kak6~MG$XtLA?2f3a&{3E@W=`
zqm&XMddCG>&MeQ>N6T9WI!IKIZ-@NUhgHdEuFqe66gX+MRjqH;7MqTmK8!|crGmqG
zWB}I$trSUAQLj^6-fxr*v43)dXaw$@MCZLpo%Fyyt+GAWiq^t%#*!>@j0bu^nx}OI
zS&HnO)vor)@ZK4X^NBp^o8$COI_dM3xq~?(QJ9H*QYn8b4L0fblDs4`d#rJaBQBE_
z=S$w=I9M%i_*yKJoyK_%jCx_bXiWgwXb0Be2N%1JHawS#__(r*a~eOZ<J>dV7v#t~
z^}~DzRac|<%f2Z1H<2XGp&;yDM0%h~hLTR%et7P48-dl8Nnf;x1k%%@zc5;DO~}58
z3|**>hQ0*CI1IH=cFd-Er89ti&fH~n(ht+b2?sdO*gFF>vk}Jm&3-aNqCjNElPy`p
zHO>Ya1bA@FUK&79#;kp8F@WroXQJ7PpaL^au`VAiGGrfr{GAvJc-ZxZ0zbI248Kkz
z>S7juMxMjIbA=N&IWBH0y)<|t9K;5}7sm7_9|vw*%rbqOoIij0$W3E;p59BOCe30-
za3gS+DsKOB0`%6PKIzDF%GhkO!{xt8<NqV4)oxV|-U;N};>RXAMYVUXyO<-t5iKCB
z0U!$w7&@3{(i%sE8@a|*g+C`zO+Iy7KR_QhfoZ^vYiM@(u1v^gsb26N^F@1!`;iDL
z|BZ-EmvY4x@Mt&?RPrF^bFM_rhCZNfJ!;KaMr_=pRWyttTdzMeoFD&Y#-V%huFrF~
zgK|_ke^PxN&a~@3fq9q8>0uYlz!)lZH6!S+1kr2!+f7|pPvUcx|3EIOu=0Tq=rV;&
z+>qA?4Tq}?mLuKA1Q~B@GUv%^ol5-))MXl9f=(vk<^Cw^q)A?x%f`6l215M!wk0`0
zI;BlPS=&bcl535kt5cU5{@^AJ7mU#792UyO+9g^+?zsoTcB4e|fWwMm*(Bx0S5=nk
zytl$}i<i+;uT@2Ehe`AR)!A<<&HjQa;CbW5m8ol=0_4-)cKAHx(nY`QU7rFbuoF10
zLM_dDA&N?}Ap(V&fs){KyHhu=+VR37Bp8oaUJiLU$|ahS;-0Gmevb0b?pSRKKhz)p
z<qdNqZY_{~kYoc}BHIWQaC=A+`x(cnI?d?A>DC8_hcyET0hMMlyRfKaKr|~P)X^MU
zFE;HMCdLLA5UwT7DyMZSZ@O<!;GOsR!ss`BHP2n~UsCuwvvez1YXiiFt!qDIoH)gX
zGVI86zQG5wO%xTqu5#Z*z;0n~f}5JyNG!uP6snkXFR1lgN#RSiA=-aVpEB<JK1O?v
zmY@VZDd(AlSK2MG^0=S}*cv>x`&YAH@FS#r_~BbSyRP%;oilu%$+l&{SlVnv7gyPf
zSrt<&^kKa7wNF5eRyE%a#V3BU$@PHDY;w^UvGbGiLAQkL8gW5LeFaUQPT;{h_ZMPH
z{Zd*fI#)}RpI%qCfZj&pXQ1>HM)O;C2%~)pF6AS#kLBJIcI<?^Yl7>Uw+kmb$?#fS
zl}>Cv6}Nz+L9bLkZZ;siST3GBp`gA72u_y7_|9upwV&?$O(nT1HneUrdj1$B1y&4r
z2mKuEYG))&=$5|Ihasb2vf7;k)l%mpvXFcH5-;gTUWrCf?y|AeKM=NL{CXo(_*q^f
z^;dgaNoCRyta}vo_P&|Fkpq;<O0?w0b!2e)VGJ9Axulw$*>RQbqvdrDU*(K1R2$3E
z2e$21x>Uc_ZX%Q|ZbtX0;5;XsbTWEn^b_fM<zlc$j_VDNHm}~h>U;2SVSk@F@lRpC
z{v$~404R*E)gJPlWF;z?U(1j7c>GNkB0kdt%tXSw!21YgPy(lxqN}SKIkuDe)jvqz
zinmhJKVGcjGL=EGPs6g~9R@gIAadtOvmHTz*Privjqb6dPtx<~DO3$Qto97pbn6D)
zgozsBZ(D3Y1O_HSoPLAM<u;rE5VHsIGI1gWR{AT-b7HTd%6v6SI;kRDoo%U<#w&!3
zvDGl7Pxcd4xrg4RPGRDn1p8XJz?RD3L{c$4M&u#!CMKFL{L<4&5A|C3!_7ckyExC1
zRQhcn$<mSQO{pW_@|2Q{(%cP<BqUQ`5B<K2l9UcF4+*(_^h?gekXPzC5`G{zffO9s
zifKc4!RG7#Kvoan0rK~cjPhC?U>6Q@9_uWKy8-}ONUv+>ZUZ?@fe1iz5*Ea4AHcN2
zM-eC9Lj}>QY!50fy7m;6)i~h1ulrG5ORhx8#TeTi;pb)w!^yzI=_IBekQ-gFl4i7i
zE61sfsH5fKdPZ@6u9ba?m*l>^P7{i_U_nLDPhCD7!4H|>arOsuj95k})w$oi)@~{s
zX7%wRu2i#)9>Y|=gZKrL!qmeJw|m|23hFBKsq*geeQb|q3k5^M+u=N$zuN8Xe2@Y4
zo?0q-kB~mXN*wj!EkEn#CO@B=N!!#$z7_1M^qArH`Krd*&ksM<eM-I-PLDwRgk#}U
z8+bgfTp-4;*5CuY*)zupO0su+l@>7Q*W$GW&lVeT%rfS@XG>GR{m6QB->><Cj;h8_
zbTsj0CN}`+R)?svY*Bjkt!%3t@Ub;xfC~5qfCu=B!0<h^9=5iOS(U;A6d_U@)Ao4|
zP8Z&UiSQt9L=IrS)J))*8~-_(fK5*XtFr_Ex{>okfTFc&_;8Upy`XQl2K{2U6#ft2
z+id=~t!12yRlxBfNvfkv@bx>FV)Qjcf0aE8{Pk^f6s7P9KWb@df6MnqWo6vH$OGGD
z0}>9u1@@}nfe(U6jGZ;ZY%2URPmN8@X=~QQW^{rLuX3D9964F?+^B@R>tmz2ooHT6
z{du(5EOa}@C(*C*bQw!B?Vj&WMxv=y$79J<XdIsOmF*4QOozR(NQ<u^#ZJu1<2_|=
zOM;C+B?!-w`c}aK)Pd$|rPLE6jTm|?+3m%rvjXyc%uDbzIMrltR%l{wXIr_p!Eze2
z;50NIY_+IwoTTV+>yKd;`WA82vz;5e^#R4xRUA}emd*g@;RCgHYElnzK}>+#u~|-R
z1vh9iAhFU#tvTZ5Jj_5i4o0<Cj}&Nu^Mj8^*52`L4Rzs{<xs@*=SDARlomS5PK{>#
zxgWZl07Z=N?Io|=s!+WeeYPcV6rBScwxDb=3|SEc<>a&VmrqGL3~}3@9^Vki{K|`a
z15D)aOGF(6qJtXSmf+uX67^vpQ2QCr-gyT`7x!yAPVwuud<{LsE=DlrIZx0>98jK&
z?UU-e9nKw4^OXCrARMUa1InO_r{ycZiCW4>+#FaL0q<!W0|X?YTzVQ!j#Ai$GwGR@
zOjmtQ5qu9_aY4{?=`D=U#H|=lyc9C&P`}&-J?*4s^ZKw0`~s7iY+;5NmAgU?ZB-ZX
zvgExE#Ifhh4-CpqDelwInFbWPFW^EP;cnAuzz?;wUm;eCA{^L>A>D3#g(Ew6vE%Ti
z^Xd&;`3h=A^lIJiFTQ0xl-!C3Zcv4T7ny>+e=$!s*(d-0`uAt>|N0r&f&brv^Gx%9
zv1Czw8z4ZY!G>Z>a|d75<v#t7?W({3_pgoWzh9ULhr$0D(LYS&J*<~q_do4VQ(^yv
z=tl-W1bb8R!7O8x7T9O615)SmqF4BR?X1(s6)+FKOvH0<Ros)o3oza<cLti=%Ny4>
zDv6ZWQG{g<tQ~-ZaCt>!X!rai9Q&G)m@}f8Ezv&>^%ljqd&79a!YwK_9>O`si7`|k
zX1jB;*azO5-8t&NKv66q>Xo8aFMdp3%3S}-MPpoHBibywEp0{av^8d8%u<vzWO}D-
zg<E>;f=*Cp1$qvp%BY#NPL!93YLXZ2S6d6$J*`32K7Y$>%ysaGfcszEdu5QXp5hlg
z$@{iL>MAk7@x9L>DG%{_urutQen@5U{*-HZzVGAt&6SDjn|KhkD!-8^%hu>8NHQ#b
z)*9bK(l&<E-tiBsCG9B+Ou2KsI|61%yh+?RwdvVGmj=B!kR`Q<;+)4UVt>Lr@CzW9
zs;E&(+(t0aWlCVoZ7}b!oGzdwy+ZmgmL||JpZ{E%AQ}_@)#k-PUZ(Ub;Hm!iU;nYL
zaOHqm0QZ&ky7h97$gcuAKx0!VWGIt+m-84TLxA+QJ^xLlivgDGk8~o){(rZ>I|R_-
zPWH3!pssS&>*Q?r;$6txT;JIIS{50ew+u(neHc+UBg#779OJpm`3dW;Lzk^_ly}}c
z3j`P4cGm|Bo&C}nl^hSO-1Yar6A@;tSP;;X-?^tQV1{mE@_=jm&NZF_lmWHl_iW&s
zk@K*Li-))s>|xX$M&LV?GNWY5UvRE#!>(T{>BGm?9ii&KP0+(2M}QOPt%^w^@dQz{
z7m=;8wWp*9M;#AMpNUYqfPU%4JHPol4K=|VEJ8l^0@I@;yvsQO+cX6zRq86f&3X`X
z-uDtfMB#g6QVjgY&6xZh;r<}EJ><MrV_%u-Pyw)xZo6+Tu3~)j^w8Dh7Sci1JBJC(
zL^5u>Oe?fpmY`U3KPf9`xV%mKq?F8)9Cloe$jI<Fnn-jIR&ut~dwBercY(e6ItA}~
z&CUOPu|n-SJ)5@lh_Ht(HNdu-cwyLiX`p;n>_8pWO)CA=osgBspYUJRw_xh85)RHo
zC2O?;r&Q+=>N_75tNW@MZyP_69F>9;5a(IPHg^I8;Cf6PKHwhW&2=Q^%5BuABg)GY
zmzMVI!N;Kinp`i#-H*I_$`2(`=zy%)ekA=NUv0dB&<U4J&g1UjlUjvGyP5r$v&W|n
zq$wr*Jg2CCr5eJ$Nl<3*q;=L)j276AlOr9<S56-N?Eorcf+4)05W|T83|Dc~4#vyZ
zNX(_)A}tvI+`8ychiBlSfLq_{uk@(=<ki35Pjm*0&qQHHXcM#yo|MtlPi@a1nO(N}
zbuRZ<am!w7w+@WHyi}EUZ9rb5i+Kn!ai2lw5(EzRlX}$(d+pL`*1U21lZ|unrUH)<
zD{!vR6@XSOH9*$@?YXoL9f>*;ea}CdAQe9k^S06+yV<$^zzs|`M8LIs+vD#4fV-R6
zx@dQ4S1Px^Rg-7MpDjATy<mZ133P6DMw~Ky6O5ra#<t=1Lfze1L*?_4xwCsa(Bn1&
z)1hq1`SZ!_VCoc`_vf2u`by-4XMr5mU;cwC^k8ZRMV+Rx(lCCo=3|ASS(Yz7OfY@l
zalaSHEkrZoL=k+OA1aImYn)Qk?j42ywzE7kC|g*$rv@z`v2h8YLbqXM1FfAqk<F7V
zlMlPK>Y8p&gxyI@X#FjW0c-wxV7*-n1+H99(Yrqo%rKS_XxFgB6r97Ya|)KMcQ8^f
z&|+d%IsWIWEypyN65!IENDFd(Y~uYxV5XIQs$epd!xwDL;`U-1GYL|dA9k1jK!_B0
z2y&pA=^w#XeK*gn>%Hm_{$QtO>iLn&F1KM~_Zh(MES5M?h?7*L(JS71;e9zVTJ2gd
z9{#vdVtSG{iow<7VOFs;xE;D83yl+>P^TtYaf@m<pM34O{WHm2g$M*txq^T?rCfyp
z<KdVUhjlzK`IFaVBvUV)ra`emEQ1qGI&<B1b@8PwiYn`#JIvdzOH+Yo9mq*zKmsoW
z+5x;HR=C^AAUHZr5SdIkXFGx^fuAY(kDuW;f40qTRrNIAKck-o!<*V3x9}U<z-gYo
z?IS3<eSy5c|NjE_<lZn81;NgSTE%%#2wKY<1FK7a<IoGNfLFsO(d~GaQ-t#OnTYyz
zmi-4(A&r}dzXrS&rvP=LI~cdRcvkcmU4RX_pT{V|Re|#XS7a0!I>k*TzjaYZ;2k#t
zzs29i?z(K#IvVtzNIgdyJ|hxC6e}V?=TF40l6ECrJ0Vvf_qlzy`5G^vwi=u4J2idh
zkk^*VJdKgrb=yzFabm{}BY5H8piRC`Yb1F_B-1E5lL_{&TfeKWmvII8SSAj@&jS*S
zh%J%?>s{OWb#Hib9g<Vd%_rszObj%V1V>R97-=uP4(=_|Dm2x*V-xsHb_>P<!Uf~Q
z8%|OzA-X$NQ!iwrQWV?|;R=~Rb5_Xd4`hEQA7<ba$N<?y8?+qbioVE%x(M>_u`RP9
zyRNb8gqsAdXW<zns>9NTFjAFU*U5db$^-sFJ%E>|UnAfCf%H`WMf7%=2cz9>Ap0Fq
zY{slD5}87F3v&O-CXrwOh?viz3vSMNuXV^lBXXHXcRGB+J^_H7MEDah-f_S@6f+N=
zRFAd!+_AfrlcpJeQnn`2ix<w_mWT(4+%{5EqO$qvk1VGTU1p<2tvflYYy3o<HA96&
zQ5RSs(Z08^?7|7My)lDJ<|pl|T((=6PN%|&cZG@h$Xv7XzMw~g!8>&qj~|gbrahf}
z?#sR({F`NsY({h9ZKVGsK2A$vgs<*N;>YK1iqL-IQ_c-*y{iPw(gn&!34X+%aoCPx
zp37EeoOsI+Y}iny*~Lz=PDKiEC?8Jk{H2ql7XZgXdAU$$)K#!chTG{d+=d)}xO&mz
z?WZR9&mM-KrNO1ZxKSilp+k+IE0^~t;;f=?LFvgC5QA_2hJjx7e6XrT?R*#%E_b}l
zx-7acWVK-*eAXfGzvHI>^bett*jV}f0NPJ%<tF-YlYkI)=Y|e68bepVhWVX_pAl=t
zEbAj<xD7f`rd9z{kCP5A)Tp5}W{~I(R8Ou(7?^@SNn}Qb4zg+L2gVDK;cKuQLE;Ti
z?L5UTdlGj1f!wj??CE37vV1XZ{iqAz*~VJA58)*^>Uryp6C{4jClzjj4`g|Xal|8i
z9S#X<Z{BfrnChpP|LyB3e5VLQnOnz6WWkYJ#JZWQj$9w>Rs0}!WMtA{pv8J9s(E<l
z1J|V#&AM2!QrLN=lR&e0!m|MuYe(G1RY8U}_n(tqH?4JiVz#z-EBfpYfmNEx9ik9n
zpEph%#q$c8d+w8Cftu-%`?X!uel2^@JCmKRKqQ_Ll)HvK;(`%kho5WS_^HGD+Z4e|
z7HAq%p8y2C7lF;0<+OUln-P6d?Ja+olNaX@8$Ym9D)DS9+L<{g87E>z^?;Ubae1J2
z;E|6P0bv0I=RQzHL^-^G(={1c5}#Hqy??hC&z#IWF)|B}Fy!rC2*ykQu13G5Na6xQ
zW$kw4Z{6>*y|bJ63U~*qed;B8Sc+^#MtiIKPygudU-sSzQ{J{+ZRw-k3Z6_9^1+Cp
z0t$ZLeWLFEZg#**HtxqefqmSw(hyD%XtF1;7{Ll(`4*;u-K9gpTLX(!Zn-`=#tru8
z%JxxfxRKTQ93XDA4jWXYKkmVJ`QMk;@2b5$e^(O-mGrw3f|h#~PRS=oX|3i#(5QdN
z3}%0qGy0e!oA<VjI^*C#EJ2cey@DFk7;d^H(_*RjsGDBOwLLCPx9`3a=UFvdpaBSj
zOx0PKrwK!)yfW4Yq)N<wzPCE$-eNS%u<JEE4&V(-2LQ|+BChhn#Xy8O1w&xtK``Zi
zS7bi*LdOy=<XyaRgE=O9egJ+*&|JyQyCb}W4BZf~!2kD!a!np~C))y4sU-l;7i|9n
z$=!)L3oosLmUZ68EXWd>qQDnq%8s3e-33NtSK*vJ2&MujjW|7y5&|&ZPY}ftA$Kdl
zO}z=x*e`9U536vHFz$njMS$W!Hze5pTsDR&W|x4e_b4B_#XZHiG1=M{oPVxuRrpoi
zJ?^$FjCi#=)>oWMR}?x%KQfpj(yPm82de8V1b(4!q8!+d58b3?R|&<+oRaL~^auVX
z9cvEHf=9uqSF$K)$n*x<5h+val6wDJYH5Vu{1R%|mIy;+cMwUQ+a<4GRb1t3*_^FN
zNZmeT(f!z3zfATp_YC11BNTkN2u9kJJsGomzF%JAOS@?o=VeUQ*D>G{G7*oK>`EVK
z$)qk%_mA0*r4>7zY`iHtn_`82V<Rv_Y9ulgF%1;SgPo6p)hwIDiiCYV5kV@Wfz2M=
zaa@@Wpkv)9mS$Iu@1-`s$eV322|MDml4NGqf>?DR9s_5s6enLkg4yW>+o{=JqFkZ`
zQfsfG?KXSC$2qJ;K%5a+u-k{FzhVStGwVREZ3XyYFnf^OrUUilyy5OK48ZNWhMzeI
zg68Wv2)hp4H2*u)aFYa&AWl0`tvM(hhi|?Gyn;mB@@?G=PE;PagkHS#OyXA9#9PvG
zh`b^bJFU7-_=oQg@-D#{ig<Qb#}Wrk<VN?c_ibMPt-&Ckg@I?s4}iZiOke_sYgxcA
z0%WwSC>=1<BeD+S2Z!04urf$sZP%zI&RsB#e-)tzXDDFVDC7wD6ZTgcv*fRQWCCR{
zXjPm*Xou?BjR=_d=H~=o4j2yQM7n@^O2H`jK^$>eiLt_M07Ly(>9wF_RexhC)mo>(
zz>apjX_m$XJRc|qQr`Yk@EypxP8dO_K-r+SLY#*g1zJ&oDI8m3Wg>#A(=rc}w*TDW
zJ>_+N!1is!>$x+0hSNI+$}#m)gyU_Yxyanfv4u&6GWWXT1YE7t)3*BzUO1rKnR5lP
z3Gg!IYL^5#<1=7r_2GQ(jH`6XP3U!Q+J-2LpY0MefuR6@$WWUXIXyz=h@gbq+k#a)
zKZ2nwD4-&{QEnc!RD7dj=wZc+1>bGfzOM?B(gW-<R8x4Qq8%+G7Ja~}>C7w$hn^#0
zkCs2B-LVhXUc;vn{nX<wp{{Y>>bs*;!{2;WrI#10b}kxin9U*RRdjiVhM@rZJ6iIG
zXPljF3bewhE#X923K@3#vyA}Z{}C~Ze~YmETZrcW<P7|4(Gv@l5#qpN)H}z9n!L<m
z`>t^Nmk@L?W1Z?t^db#eHaTnbXsSnR_}I)zCuHnYdMT6ID!baE>Z4)vVraMBRoGQ8
zd?${9zzpNLu}m<?DhI}fuu0IV_z;<j^M4@d9x!ob_U<rvj{V`Ix-!F+CeaH($T1{>
zt*934KRwd00z$t7GkW_k|E3qSmv`t3hyNAP5c7}E24IuuD$aXQ*JJS<mNhMR-bTbz
zaaB5=4H$&~_?v%vErqB5uU=%q-*jED{~`Q{Yy}3A6gwct7misP#w3QVw_?EHi`0q-
z_A$b}T(P>sypb*4XJTbu$_6f7)IWPbGCo1B^N0UGL6%r<oAZvoS{@dRVRp8k;?Qne
zRc3T#hta{1Bw^J-)pp&#K;lohSdwr#G#)st>z|mi6z|o{nm(kYzB(E6@-I3-G_KN&
zEe2~QZBh5rTH<dEn8FF1Yfkees|Wg~zIe!Qc@WwRKY|pS!Y6Wf%)cTx4u60767E=a
z%+K9s4DHpqml@A}u9WDJnB1#-tWkAGq-s6k7|^&9Kx;XC%^Y|@x#f4W?H)y0v+ASU
z+;w$e8-@VIDnk_yhE(GR*Fwt%o}E-zG8c7SszBTW@Ji$tluhq`_nuW>*!erhWRIl^
z4DcalvAco!4tMXUAGc{B_}&RN%m??ab^6t(4nQ2lPkwx(a*-Ca(wZ%~>oS=J=6*A@
zd-PUD6|F%#rrBBT)5l6RoI<Zq5YzQ|@>)^gqzYJ6<|U_>@L?aY^}^nUbzmzs&sWZ2
zhEw-Qu)q!mce@?ca&on6Y;<;=u)lWCxYnWmbJlD~VfeL4C@%x^S=V8fAlvn!+>&qB
z-%YRy_gI{V=D*+u5_B_l{qv`wB}rtz$v`ba=m@nXe_dNc_^VMPE^r<m3W)b;Fhlz%
z&NA2#UHS_nz%bd?01sFcph&SakUI32)Q;#VvJv{$t$dy17|+o#_1c6xx9=QxwxZ^9
zx2ID5CrC-<ostX`0%(8^qax24_Pvc${IYMkhX1aOK!4Cb@%;Y>ZTC%SjBWnQJpeP!
z{jV`o(%&%Cj21Z6!umc}mu8xa_c3mQ`w^dX=%l{mvugcuYl`rYAwpGegq7hNM=<=z
zbIv=Zx4(X2IHr@OJhQK=<9@3+&+or~5oi>6@8@BzLfAxU1j?AI^@{lDh#SnW)OqK^
z>6^uCw;?=c4(x1308OVAR)%fbw&p9ixY9UGNyu`qO0$^hbCQW7y~WRF*9ws{7WAm8
z^azivCTBTIE0Z?Q-k`e5bz$rE=Z88VD5D7sL+#<0ig;2Jm-m6CG-M#8>K3YHO|>e1
z!tfNP-nAA%<89O#hDNas1RkKfo_>E*#qXVP(>22OqpjKPJF25=;FKzm9(Y!Vbg{35
zBK&K9I;J>9tnah4n!s%qqaE^#LW!8meZg*`JAWF((i6;z>_8`C!&YU$$beYi+6K0v
zzRo-JsUPTQUKq(dxzgsXq_0kf6QHH)3vIdFcH;Jh+EX{<wDHHpd_q45252A0xqvyM
z6RA->%20@*m&b`}ke4s%Qg9bi9Bz2KOs*-G&7^p3Nvg>mVLUq9t65M%LWKJtC@+{x
z%Cr8B*x+!gU~Xf9fESc@w~`)sopFZtjV?uNIUn%SEq17Siz;*1rX-HaY1*$8COmfa
zo`JtLrp`B9@M{cJlZSDir3NmKwLXSr;H5n-+e=ZT#Y9sswO3SBXSAn4E+B8eu5ykh
z<fH;r<<-$lB{=Vwfx1bSF&{utD+dY~Ebev*s#|kr-gsWVcHyh**R@yp>b~p(oxFPp
z)u7-TX7VdnhWi6N$@bvZL6%t*aitl83S~Kf#cEU_%f@@=jAYv+R>*IZ>06%lt#!`^
z36f0duE8;sI~Obq*$#|ES2Aw5ch4+W^v5sDr`b~)(R0E3a|eZWj+!YGsGI8uTi-L(
zsl&{jrp&KPO;xnV7CyFdYTseA&m361=G{kv<aC90YIn=X9F)3wmHu0F*IkIN4PS-L
zd&nN%2tJfG0C4=M%DH6!A*AJ-Ve_D2(>(c?I_Y`wrf(;)J-_gkK1KO&91srS%?~-{
zcIkH8+%m_5cBCAd+hzbvhcrk61a(HngeUe0kl4}dES{IVpM3N19&y)bTag>8tC+vs
z8ebt-$O5f+KI92_acNbL79z?wAYkv_?=3Fq->nLD>!ADF?KweATFZKI^}B(kQT4|g
z9Xr!LJKnZlzvM5DQUNj27Il~bpXyL;>hjU0wBX-Z&5V?e^Cf6+oAzy+19J?R7}Ea&
z4I2|ICw8^UvF?P;{ldlleps0uq8oM-L`n@yn2|;IrWO9=JbW^oaev7@?AB7wnT9d*
z&J1P4#nB0|0SVS>yrRj)F4I%*<@Zq+Gp0I}!%tl6p{!Zy2E+kNxZS4YQQ$PS1GJ_S
zFQL1L9((>k%3Kwj#`nx=-W#%buk)U;{bEvMJWFtdE9CyBGRYnHSxp}FK8Eul&9J4y
z3JP2T23FSiQQ_gi@4O$A7=bp=s=`7pJ*<Sg?s10cE+a4u1!8oD%}$+1e)8?LKHD&0
zW1i+uaRS$fAM7ofNYO9P5#*@FzK!PP;Uk>|1RnpFMHzHH&Vyk5uwqL(b3Jw;zX!GR
z&DT<^15GAY^INtho=A8HZ8A_@wVMQ7{DO)6MkgPV-N#Y_O>SC$L-%#NT8SZz@{3U0
zRqh09tiGj-&T!|)xosPft1EXE70BVee|oofCQ8~g3*qLIeWl`q<&JZ(hvE={X;m^<
z6Zs#&FM@J-1qTK=fFpbssQ?#blu>X)ceGdNdu!jFXhlISAZ*Z;Mj}CvHUJ<;LI&XV
zIB*{#rb!Qo)bqqB9KU|>8%9-QqcgX^XVhF|xPnJ1Y^%JTOQ!|%YlI*TGos41N}Ymo
zB9FpW9qr2PPSM0Z--S0V@y&6i`cX>@b+%)LMTXy5P`53@GO~cd&Gq?T0mOFAOcC~O
zvkB{%-~u_@UgYDkYu{3vfBCo+{oI$3nEi^-ePuXDI4T~p>_2>uj_SgATR&jtKL|#C
zPB&S(wJ5Mtd~Ulj?mStG!9&AOz!QlcMEQ<fWc2qL>*~rUS-xj>waey+MVkxkC70l5
zJwA5Ekpxh3v_~f4>HsCnF7vzmUaRSFr{uj{(dA4CJB9(`310|ppkZmph?n8vuKl#<
zyId7&(=p0kWVs~qMK+>uQ=5#h6iGSnEKA*S$y(`V*t*@VSIVt6HDUF58`=O{lsFku
z612hyywazMJ(liOq_T2Zj5B)hbJ4v<KnNuYICi}7Y?i7D@LU{DiGjxEh<w96B1-Fr
zvd$wJbJ_c9s{1SS-f24we$GqNxFDhE%w7Xf0-${2SX<yI^`|HcMwCl3qO>*fXoL0B
zdKqD-dKUii@#b)IDvU16Fr$jPWEfGW?W&)s1l>)4TVT+j^S;CxCb&KaQC$Xx&EOFK
z9V%@<-vIL#8El_iqF4pL8AbN`WP4=S?B{mYESqXr!wz*`It|UnfS0{-NM02*{{RK}
z;XrhmpII8}q(i|HvpPbDi31>b1)&Fd{{Z(#t=LhB;{Ecw13xtXv;4Ocq2*5=O{6tW
zR#r_=?i@g<S@1vlHj*BPHG`i~{rD;Z_8bGQuQ`pl&^pz&P%)-!9uy{*_~zY~jE6n2
zgC4v*B}u?z+@F$5bEJnS`l8gQ%r9AvWt3E3)Vz6j=VuDkf*W1yIPrqFwc64<Td`}Q
zqRt@6s$y03u;;+mV0<zGjHOl3LdX@{jEf~zz4`{h`+v!7Z*auE@|}t>JOMzA$aa)P
zbn~mMMrL?({T{K6tbnD&iW!0i_{|4YFilg;ad19_saE^B;D}^NnXf^V#25SrE#T@$
zj8Xb{DW6e|+}Ejk>`1`cmTI^#rhYGa&xnQ`!m)4Z`1->STxX0{j>wm1H)(dwc*zsV
zZ;`3ab*U$9RxFybKYQ}3{8ps{?WbnemQI!-Am?o-CG>0464!_sJ6}zve^;(ogpHqs
zS}^@V^}6tm+B)HzSBi9O=(llpOgNgSZ+T#d*_JR+_=D$s7Uw)}5%<&jwfktE^%>y$
z*q#p{>}e!WdVJsqTSeHdhHfU>wvVV2Ky3lt4jO?%6a`Gi>i2Vb=Y}7SEr3x$YS_AM
z?RUG&5I+wk@;dqrfj~8TaKsSGfSFSJ(|&;khgt39A;+-JI(E5D!N>KJa$dZ5{%g?N
z4fc<iH;$$V3{0tZQyEDWcu9mY(2lCEYahGL!Cl-gPd}w+=lU6+Ll6W+JFxEnj~Nb)
zLLH~$7w+w9P_MN7)z+r;vsf>6I708<Bjh6GXXMS@#tKh@{1RTfV80T@i4(b)VvjN^
zt|f!3PxX6t;1489>y(4{tBNxnZikf84xd<dV-(O^3Va-Vi%Q%K($;E2LsH+FyKLH_
zFBr`M|H>846O3P{JooE5)<6H)07^@x<7iyie*5;2!-EI(s0IkiHrPaEy4tI~c??Y1
zr9Tjyd+iO}djBw!mn_BJ-8s?PtY$iMxuR#W_QX5OFHT2I;s@pbKx%}X)j*lN^%}Kd
z&S-|_-afdVr~%!qKnz%R=&^A_PC=1kkN;&a!rp6VAA5x)*7{Fx>5Op18^h0G>W^bw
z<~zl}_G26Y)}3G09keczH#-opXL#yjx2^Ct!PANZT68!tfK)cC_m8TccKGJtG!Pps
zeqDs}>?-mKvG%Q9)Dqm@XlqDu>yG0eE}Fi)xXb0B97O&J&jGY2rv3uV1woR(KFYq#
zz_`*AooonY`6lh75wQN8#m^A!qza}RC@{c**@o6a&eQ#A4RhSR#7|vXdZ9xBh{KPI
z^m1DodlZwC;V+$q!2L4#g-&anf<u~j#74+Ga;(H0V^T(%!_#PubUub2ZO)?_%r`k1
zZ%dA>S08uuzUemh<x(}lr60&Z8;WO6CS^?36nN>}uG$-jeJjp0MUo><6myT#I%A4f
z;ah);BJHm$o@J(5{rYN?u<O0uPcqSg(LQP2pf1hrB8mEzCklU`ztUSF6X~b^;#w|N
z<;XJ3h=h0GxB&M?(W*()>@M@7xy<jce;Z3weG)AY9sFr6LAL`aZX`jFq4V18zqNPn
zUrk+E90x%$43C0Bkd~mJqNWI1q4G+Wb$Az{Ei{!^Z~!rY1WGX(47n(cSTG=<L6L`#
zB9DSWc?}Q<;b8>S$ioB@VnQNN6eOWSlaOTQGV{ZXf55EOu62LA_m^|e+Us-n-RJE6
z-KU1=+s}1B5^w<#NDPZBwGn13sXE^SkLZh{dmAgpFqz~~&E%&P>}4_pEz`kQ;Ovy6
z^uP}ag(vG(lq;DPlxO7v#;wbI2Y%|gCKh`htbhjw=t|_jGfWLkl`~k)y^aKi7z((S
z1rB}3TILw5g*nKOkvu1th$2aFYt#vzdGbVW)WB<T@<N$P(5p2b*CG^wfwH-crI>(w
zae3!&Rp*!VD}%w#gejw7*KB3j#ve+yw>m3BY-@QOXTLym9X}bIS|4!F^TM@l3Ji<c
zj9&_rVy6fLrDZigrWDyTy$e>jp+?OIibKCQaSj5!5pFU>j@+RilWFkNk&&#3fR0G7
z?x&}CM|V1NzW9tq&r_cSc_t`48n7c7)xx*oF-7fG<cS?{hr3jUq0V#Anps*=G*VOB
zgG?e@%N;8kuUV>2G1^zYGmrGE_dPbr(*tLjp}C$>Xy#Zth&2)TlWD)KWqr-xF5GFL
z8>u^_vwvs_dxHEN*yvbw2@LVDSKxyv1cO_{Wt6}^hsX4VUJV|GSLSqQq0h(T>%s$*
zGK>hW<NN)0T|bVj`%hSW6y0pQ<ON>pn!v7-T_tAx_~NNPhEJ-2rCYi2PnejT=Pkf<
z0dY&TQP=oi%h;56i%#cW_rq`)!u_;pmOOvMTO~*ycc5A=sG>`&FyCj061)DCU^)%n
zNGt*UriI!?b32L@k@{(u&oPlwlI=M!^;KKOM)8hxReaWq`nb@qIP5@Y<3P3Wu*a@Q
z%t(*9h6bKejD_3gY4?Z(QT}HWsRXaJ>JG%~l43g}w@)g|F>h-6Oci_REIg%?(o>gS
zBP5o7RY;5aHjFl1b!M2y4N>5~Y<p2a#m8=LE1k94+;E6O4Y?AtPaI<K+cz@%@OBx2
z9ju>N%9)M8bqZU?-Lw||p(QHJ`>vl8uO?*%Ro+V!EIpX<m*7L=jZ9xBNdz3lZqwet
zk-GUbNqlhv&O{iZ19JnN|LU4;4he}9;|&+0ZJUZE5jL(<jxkAfo<pv7l*smUP}Ews
zkzHO^!b#ge0SmZ8IF=~~GbRb@auYD*j#&WSTWj1kKcnNYy4neDSv->pw5Yqi)K7rs
zcZ`y9GtO}O1kY>TQ_BL3fYVsZAc@l!YV5ktneZ*^S-;+NEy{U3`V8z2g}rDi3&krv
zOwBaDD;Db_tRpEIq%4hOuTV-yIB??aq$)lS`xeC}+ZD>9a<}WwZpeO=pWhOxG=RqV
zm*x5u=G5|9iY#}LGvnszAzp3=5{7nN?-8ZVqa^;G;!IvW*gUoXD+AG1B2i$)79Fd(
zhFLsIT3z=h_oIygoQBN#PF}Uf6gbANNVOcsM-QA0>~ZgJo93g%d1$#@idW5{%P?;S
zi3y9M^-oA>4J~e>igi3xSZfrdxkH%G{6XpCZuq;qj`|@u`Ikg<RUmq^;_j>G{tW}g
z5!1oe<_}N#2-!`$<z))^8eUV&z^ywg%B0`QD#i{7eP<*2PcIaQJ@I0(#Oc^k43%Mg
z8<gD4zM@Ei4vH{i4@#5Ucwgmyvz#-H86<xSe!^32!^O{aj=fuLbRHgjicU1juN-!j
z?bAc9fwAtzY$A9Fqo@XEs?1LcPS&>J8o8Tb9-*{FaHE3g!Dz|>aRt^GI8C<^5J&oX
znJ*uA8#8EPVbrDUhWyq?xBf1zaar}&0YhEE*`%P3CdN+*EgjUmG}m8;eYU!&XUN<t
zm=nrC`vQTK&GsImbe$hb-QwDhgnyZ(t3j%KH%bo_NGA0N+Dto!o#Mxtb+1*#8%Dii
z=ggVsz^c|#&*(j2#)xv7)v20l^`Y2wz37E=giXJHEux|Pqz-aB=AB#Te`syb+j?nI
z8j|OwV`ZTP$JECwEJ-e68@$9t(|inQKu<0Onsz6<HpOK9<uuc;_Je}H|1Iu=o<A&h
zgzWLhu*dxyqXhx`(E%CTbvS#_|6pE61OREkfvo@ME0jt6@6`wxUHLXCFB@h4#*$Wf
zSbm~nrsH7lyNHk*KJ-P{11-~oDW(H@YF|0B4`uX{<a95%A&MXE*Y2N8sjy#B9JTQ|
zqc^E_2fMzYJ|ZDy)YZ9s2^BrqIZ3577L*<yHoBoCh-DpbyvchPWN$?+1orMN`{}L0
zW{Ewqk8l3%@=AR6TA0JieNcLtuDB^_O!PX)X1Pyx77fI59@2K}tD5&&sK!CZ2cG6%
z%nVvb`-A&PPNDe@QvBiVZI8aNNe;aTn>N{T!?quk^ke(5x#zI3o{x!yKI>H5xY5P2
zms_o#k8(Rf$?zsf=mFUPVgO<QVgO<QVgO<QVgO<QVgO<QVgO<QVgO<QVgO<QVgO<Q
dVgO<QVgO<QVgO<QVgO<QV&G57fayP{{sq4eg82Xd

literal 0
HcmV?d00001

diff --git a/docs/_static/sobo-vs-mobo-thumbnail.png b/docs/_static/sobo-vs-mobo-thumbnail.png
new file mode 100644
index 0000000000000000000000000000000000000000..2c5e9b694ba476d3001c49c7463ff82b2279124a
GIT binary patch
literal 50244
zcmeGEc|4Tw`vwkcMM^~yLXs^Zdl71seM^??A^TP|ma$BWB?)EUHMZ=NeI4^IMfS0d
zT_MIE#x~6OU3b0L@8@~`dj5I-`1N}An&_T;xvq0L&f`4JJ5)zY`NT2SV^maBCsb8#
z>rzo2B!a(9G)KX2F1sH206q?Q=qkggiaXEGfgcXr-O{{8MO6|>w`X|-{7m~u#mIw-
zioOy0d!X4Z_aPNk<B;m@TYBE+OR3Z;eSPD>B0bZWH^+VN+|^q=;alU0-WdL{74eGQ
zdGK^r=+g_i?{u?{J-?t*CH;B)8za-1!y{>K<7FN&-sHqoBU{ht#r17p8|jvZVu?dz
ze0=*Jo!H~xx>Qu3&jxZppBI`gfLjKCM317Vq0g71G;+}A&DXaG2cUmg9fzHSK2KSn
zLO`ES&%MAMg8uRO-2X49z#nXFZC%7h%9C<y7E9aOC%=R(ZPh%Ags%D2{soTj%~clU
zVo9^ci=!vnv6KxFk8z(_t_|qIv<Fb=Bc_i&9?3DsAab-)Wef<hV_p+uSB%T~G0?J7
zeP#-j|8DB@Bi3Lh^{b%P;?{_pLKgJA7iKOHX><H{H_a0@D|0vG)7RLb<6rX9_#dPx
zfOjd+n5oO@LeG23;uPs{RAf8#i$3USKQji(Tk74JG3!*=_xpV+TX3pRE%iu=WZ0Bj
zp534tIP56_+<9O#C$hd(X#Hck^H5YAv}jKgp5pkZTKveQ4aFbz`M2hwr<^<h<Gg2j
zU8BTi+@e@f+33%&Hz(+pw^RT5Ggye`zz5X&eCusJ2c#xc=8^m4EuuJdM>k_`60B9=
z;8?~1-7fxX;8#?VCt;k2_9jO&l_Tqw%ZyxWe!${)?|uDx4R7X^YE=4Am81e4cH!6s
zqNvsKXpL>S<N!^q3p0|olTrJIC99NkLWz7rGly?`VQK4S)WRn*%TI^byA<}bB59y6
zI<f}JE8W<cj`pd_E#TSTUBOE4Q?{^*_`26D;!SF#{DNZETKNQZ;hI2$U(ZeR4A=b9
z#0?{f9^QGM9x%K|-kO-%r>yK>W+E9{`1ddS_GGF0sFFpS&nr^@T7h5RU^e&W$_(Un
z4rFWDvp|w<s=5pbJo_+R?66~j&%&oobl58e%66mcFT;c$np<zabhU}i@QpJwwmr-*
zXxxI4GX{a2?=P?$NACM8RY42>@)V7KdxE4>Tr6UrMydeaQ0s#65?YGLr@<u1`_~Rx
zhOmnn6X13awts$SO8aHwj%NeqgZFl3(`|9@Zq1DQu2dAQt<Q9*qg%4_VIq14;t&3y
z@&H?soF?a+l#Zla$3U)6m2nuB{vhtoF9*870qo?rbE(MF_&@*cR#7Lx5x*Sj=`=8z
z->!6F=l?zQ|6PIq|6Kt`U4mVEf@O|IqLuvK&&FfBB7%bI$k-OV1=1x)Guf6`JI(p5
zka3LpvFzy(&VZ}E@3pID_aYT2Dk2~(KfU%0$Cr?FhNN$3*1wEIV>ZR;n1wY&Iy%Co
z$L(1q>>qGSd(5;KPM-XdX;$ZH`-RD{eAbj%5BSq51rUtyMZ>VImkS2T+{EsgkJOyE
zA@{PP2Y!^VNGMbFO2F+_o$x&mvmoQ3jCBeLzN1Jdr+4)4pBRszCVW>gf=md$Wg7pD
zDP|J7fpi&j23%nh^4LinY$blv8HSuq^-6L=V+NiOa38N2m3|hcn*u>lqG}?bu7)&U
z(DCv#_u>MH*sHC%Fi_M1p2tSHv-`Tw-fD9PEnmum${|}ek8jVYlcZdR6KXtWbGD*V
zIgxwaH&#D0&ZKyBipnmPM-RDS$?ZJxrdDZ4%1&YPMyKq`;4F5ja%52ojiuD?tRVND
zHizsuck8E`Bc-dx5e0G$cQciS@NdGmCXf@$v#$h;KUh#tR1KBcjd;MpwUfb@I~yw7
zM6jdBSta-vR&AV6=?ch(n%(8Pj#xu;ho%Vj<c;d!lbo)u4Li2I`TBzE;NYxyR^Q?c
z5^)d2ubLeY(>rh|#iiZqHsQL2%t-lR`K>~uvU>a~_@87Dm_I_9*S4A6O9K1?JBC)8
zTxG1s^-du+7afEfwyk>WeLUDO%Biwmm1qz{FU$dZ=I!K>cEfDM@wvWF^-Rh$M5mci
zI;1BCZ30F;4^h(-r=LuI4)q*L5Vh8rY6@QxJCR*O-W*AmfzJ=Fkh`apA~{^X+&i3o
zmzjpmt^RRPbItjH?e8<--Zk7XXK3>`F-@#a36H0m65#_z&8{mEw@Ozw>sA_ee}!c7
zSNFYUuImFkMgcoN#5-S%D}>hjE?HjR9JE%9TzN*#LBm7>$95LhE!TS1d^YJh9~^-N
zE1eTSCb2~B&%IA?Kg#C(fD;`SSJo-hCgD7YljVr7z?+ZwfkpTH+MXF9)9RZ}?!MkD
zWl4Ch%1mRql9xgEBJbY&xM^{EV2Mn?5*d|IUOlsJC_}-Y#{>%>#Ct6ciY|E1@umBf
zLV4PGq1-8PgO2Q2P-MF=x4T%jv@qatW9=)q?T=W)$~ItTY{v7wRc;pA2%Z^7pMfb0
z#2n9vu`sH?He6~r7v@wNXx@ch=lU4yT?}j`?K5_3+}EIKY&UDA5$9Xp!kawk(kZuP
zTK3@Mb0Vv?$@bV()u_9AZ>5Wgxcjt9Z-JqRN4sdV_pDyz{^p2VZ{%e@6FQ6qL6SjB
zKQN&2gmii2gx`9{(r!i_Yi%PgQ~9LFeBS$EI>sUeB2H1W%(i2(raLvd*sLx|nbW@y
z05^`bZw3A1Na~yRi+*0O5#e}Kk8ps<44t27jqMh}5f<zpx0r!_EJOBj3?_0qJ|Qno
z-caxP#csGz(XEnk=dzAPcbBaNG&Uz*_)h8U;bqG1%9fhO9t^GW;KdKVr28%GE2ovu
zzb!$AE}k`Uw#4Vb6$)cbvc;A&zD39`e|=*?OI8ysnwG=1`kxgscgFhzf7+A`SpCAh
z3A?qRKf}P=CDO2+v30&S`rEj;>e74x+BJS*5W7$Dtuw7&-x{=z^qKvZlkD1B{m11v
zoO*p(1y1ntp(JTHbA-$F+5%J<e^t2f%{^u(5~53Pt9)SqfR)$euov_)UyFNFsFOZl
z;h#_H(_m&CR$@MNIq@JBmC8i`0YabS4C9=$87@bz(j|TFu8>ggd%iZ1caIO9JsMj2
zlMd4=Vk+qNb2Z+~GtqFFI}x7CJ5g~-J4r%dp(aYBfXVpz<y&6cL6i@T^liqyMR0KT
z6PA-cJ*s>|WVecbgH3;Uu9o?%&VoFNkV))$mCIy0Lb(2fa-IR<72r<mUzzB;E>P2b
zBxF4LLT&GNVcsWx6#NgJ6PVrO1(8puhV0TegukWv%p?|ix#*RxaG<-M`%b)AQ2?Mb
z<raZa<4mA=Pqx@G_~8zwig&J<je)c$M@<L{JJXElAX0JNaz3uL*dU-N`aWbzbJ>Zo
z35@U7>&{dPdEwwE_A-vC&Xxnz=O4sNx(t`YT0+v_ry*LS@AZ$=8T(;zc254x{v_gR
z#5za|U$05Yp&dKM<oDc*;+%xshDvReShU`7DF%C&ZRP6`K%^{fi?0YW{~;esixHOj
z^w7*pYQWS(vo74+@V<k=?&cC^%Z&XQd%QU!4@1?<eqsNJ-YGcDlYCasJa*~RVPJrr
zzYUPH5VaNN_#w8u)KXA#K6Jq02`jc<Z8KG(;D<r!JjVhv0v>UExtHqzJX<h-!jt*9
zmRi-Ar^(20!D%zS%E`z$Gw<&~=IRYd2@Wi7VX!p(Y^JrR<SgrI6Y_3G-AS!la?t|o
z$7b4+hyAf@fhdpi0b{2jm;TcOw1@^FVr=Z_Gy@|18l$IZmOB^UNPfXdCmSSb?mdR?
zG!|E2s@z6Q>NC1FvYXKrvwI~Ut5_T+22yJEbC6Q`kNFE<W<*qcsKc+%B)cJ+iz4C>
zDW4_o7f&)Hkp4dBw7k1yS8j|%6|9}(+X=dzDsHC!!Ggf2k?>~S)QLtTw8VI7wmmW0
z+;Q<|0;0aveJ}G{?)rP#uKW_OqGd~!&3@%uBlDAuW~YRl=*U7<qq=#x0;UP%v&o-y
z{@4IAOvLjN$r6SRlMxcU#CnG>{ktm2%#%k!M$H(9aZU%bxqcNOMycU17Oi%tK6HG@
zEL2Cx2)i7kH!Vm+2;)Wdo4F1cI2D^zj=0th#jP(C*A>J1+*c_5=eDD~@xz6p^Qk!X
z2OgM!euAt|Y$07mVVyP2z^j7tW|#d@kFFo*8_k&yYT@9^rp*(6f#xF*84+fIbDg*E
zzrXL*e5+#3k{KR=!mGe6o70Pe#%VD!R~BFd(gGU&GhjP&>qiS;`tc>56I9>$%l5NX
z{qIuDHYY}??jAsCUw!O~I_dceZvRyYej}D9zu%ByF;K4X$c>38ww7E*<(6!xr2(B7
z;~cF#{kOM}bpC@XS8hr8WmfiAFGYFv$Ehc5us&DR%FoQJD|j^Fzva`zzImP%J!XMk
z7WV5SEeDP(+$#h@)?K@DOGhzio{Za#D+)4)Ss2mzqhKp)u6<Rvn&<iO*NYeT4&5x}
z@9(;SRlVQ{3^s>1Xk7KasG92@_Kl=d9wW}7H)rUsj@~kgb?d|FhZ3JN84+;Pj)7Nl
z-aj9>$?n^W&g1fISR3W8EoN}G6gp1#ci;OVwwyr6L<-?Re2x2%%bhEHLBb)5qrga~
zQJCM9C0zKLdw+<NiVD7_Po#@-76Gif#x%qv_EgO*J;y$9bH}PY4MJEqY#)HJtoXP{
zwXD055}zWMtsW{i7x@?5`ykU*+u2H`lZ;y_8pZk^n1rwzWD#dJucA&eK4__{?4G_=
zJPOrSMOW9te^%k79eOSm5FI7hT2Ry7BFjlxs?Up>y%q+97Jk0m-|0futG>CkVVOBP
z%}uO5_tVQ9K6{sYVtazGD*c|;^)J5W8-iaekzSVX!Z@+jc7iphiytZ#eYa#52l;6&
z(9a0eg^DZ46GNcNGdJ5m8r%1f?Yr{AXxGyjs;>qzhpAe4M^>6nDu{@WnM=)N_`Y^P
z<cltF7}#!J$F`g9=8F0&X9e)>$Bdu4FbM?>dLC(tMQJSj6<p<1zf8ffyL5Xt^0m7H
zSgbo6{U2p~W8YnKz~d?1R(goclJ6B-yLZ*yKCC;;cBWzyoh8}wu~Ye<dn;bgauZ>V
zeri2qqDvs88?DgWX9)6rsm1y1E>($-YqJ+PwvW-=n5X{cHrK^NWIfoYxp(F~N|~8M
zM!4u6v0+E#j9NErg&Bd;DqbezkAJ#lg!f$u-tD^+t90Zbl?wTXC{|cQb+@reVqBUB
zCYxQVuA<w9S7=YL3yw?VLB{Pb5A3h+$kFSJL6%h;G`_FiNsF;1wet`IE)iYVa!p|(
zq420edt&?tR+sN?C@-NQ$D)BpVfiQUM?y@-0pin6A$Q67Z9iJs;CC#E+g_Nw$tZU{
zFGeE&_}BeGQz0TrW>K<XZ>^05tu3@s6P%@&7!mC*7vHpD=u#iyR<~5So@M$&l+x{S
zTNs7^dR;3gHq~o<h@PA)cFF7lA)+aiFU?-W);-)kT9W{m&o_4GCSsCr-S76zLtU1f
ze=8%pkT6<%;Ly>)WlkzbPV4uqy^F77sFoI^y|RtI`Bq)w*|Zxqww2#EKngbQ3m@o{
zfYPwX`$6zz5@osAjcX%s!a+rn#7E5M#vj&Oek_dW?A2BnQ?r+npwtO#rPNbr#_fbM
z@(t^<VRM;{-KTzrIm!#QAvh%`)d$<_YE^o4Vl9jIY68ve2UVoG3ltC>;f0yq-g=~r
zunZf4nmjc+4DM=7ZSJG=<Oh9uMq=qjddH|78EtGwiDJFe(=mIC_iZT^1=_<eE^@YS
zHl_5nCUlAKemivgXrTPZzAW6<%EZLNr6kSD7xfzhCI$??Tgxs#dPT6eP7gSjL|tn^
zJX-YGucVwMSvCd_l^$o0HYP}p5dL^}sH}3Cx_bfFo|$xQ{yClShqAK#2;}tFw?;vC
zZTMxFS>46FQ)+ZN>Mt33QGX&G4Mb^uWNp>u9!~*qX;b*Q5~ZcT7EQ-w_JpwgJX?hx
za}*Ol|1$q9!k-L5c(}se&lgO#{eSFoqOAvN`U9W}3)l5jK0T?zfZ2Hd!@PV5WlNiT
z$9`YVck}oXaS@k#dgC!})gkagcqE!jv2lFFPb8n}g3(?>b^Cb_O*Owu>rq5!p|v^m
zLRLfZRGff79<n-86xmb7aghc@5Kx7xhX)JxtvdOwnzFOMfOYZnQBhUA74z`5vo;uf
zZ`v6j{o(x=xfKaX+a<-;_JVsobs92UE}WjHT`s;o45GR!6Jj41d%2OOvU<i5rA+v!
z4KE+?P^Nk}lSnQS^-YbBN@MqXZXWAw-3~v=w(hb0hcWGKEOwjyEHyy#(PvQjRfi0L
z*Z`GQ>2Yss-^Ci4MwwFx{urCnkGAe#;<#XZ+`loQC_-DcD?<J{TCGxgCb!`A&cnHP
z<`s+qU+&O@4ZV~XG!9RDa+eNct5Bqove#%p3gY<weEw+{%S{RIEVO7rm5Qo%CD&yS
zUNa9gc#^iM*7j`{lF3^95@!rCQ05?Z_#!n}ji^gRR^6WPtC^Mz{ipjUhSQ2{V$LG?
z&%hsX$UMni(kq~E3{r~heW`3IzI*VdrjD&}y|mDDac{wHRnN7=Emi+AWINV?qTMvn
zlcju!-d<1ir7-Xf)hHYiL?2JTP1R^#CW*r+W9CZ+tXy|2WbLmXQT|!>F7)GGW3Yi_
zEmi+R3swew?w=2nv(Hhfbp^THgDoo+tUkn)#;DiJbP2~JOjG-|NXD5H68G01mifIv
z4qd+<bzb@-Dp+CQrS&N@Xi59DO1@B`LY&X26JB&*dU*N3y2bs2H|Kut$02hgMGI=8
zlKryBv8lVpWz8Ix8X>c<6N#0gQ#m}+Xl-fRNt34Yi0!;huY{Q+&j<h!0XFU3BQ+kP
zFsoz5o9nbx4~AH7pxT}Dv`RF4t<sdmnow13t&9BEeN)VA5m>8sHu1(g*96|;ux}Yi
zil<2T0a=?ce~RQRG?sLg0MN?^xECNMR7HfE<#O_iriVC`3w)P8?>A6mi(Bn{KR!R=
z$!k@STktUFjWd9Ro=xMVjfD_Oea#|><}*`~3R()>R5A2qEq?K`1~J<%58WO{8E2bX
z`EikD%7&|w(H-%oar^c+kDD|SP78lhrG;^dfJM0%EktDS;gd`X?+Z21seLig{H!&i
zCIOwt+ogSef-OXT_kD`{tJERsLhdJd7RkZ)Qg8T_&h=!6E#b`2=9ps&zI>@qbc4do
zPS38>pRzHQvut1U&HijRTb{D3?)bc0W*^~?i53mtlAL*#W@Bshsm<j~pu9Ph*j~Xb
z*kdMZ%DZ&2!QUta>nZ)0z67MMV7Af>#l)>Xzf_5dk=y?E!Zudlpk;gLUai-PYupv1
zsPSV|uR;SVwvm`C+Z_qq!XhHcEpN6`Rx1XiML{|RuILum=9+MrPk_e1{w63SdV(4*
z?-tHKzHL;T>F&y&?Pswzq@|ctjM*ICEHiwGW1hM(Gu%wg>Gz<`drtuwwFej&S<DyR
zSC6kyeTXR)BXjVJbMvy4ZG;N!sA^yR(ciMG+iYPq6&1CE=@OqAvTV{&QC%K9bwJDb
z6v-Q+pN5BF7A*b#S5#xlk`S6ZH{A4t8e*!amhtFxBaaSA@qnXUH`X<F%Z^d?H-3<Q
z?cnZilpU>MRAQK?*&EZI?xul82_$NJE8hp~(ytQ&K1&c?-y>m<`$Jy20S0m6$)Kt6
z6(Wu2c$m8Hj}K2|+zIPZUsL7|tKX;WdYY18L3$OPfJScHxmL8h)+Vy#L=zEfL0D&M
z{S-vk0f-By<`MMbxw2J|frh5KBnu~{C;Y5!yVA^Xrc|)zY~2LsXrUh-0ewdLvz4V=
z^8Cb)tR!75bL=X1pWL-Oq$9Dg3W$X$a}zC1PVSq!UpLDn+@7`Cy>^Hx&O7OphhK}`
z5){YFMUgOKEeWd?@7kL<fA4&bOrZRTNE=AlNva;bn$nWKizOjO_1+~c683MdPrp%I
zymWRrAUoDTEQH-NLFmq-=FPr)+Cxj^RJzD(lQipqEE)FS-I}QJurpD_fTtY^3xwJU
z5)M6|lN}7Me|vhUz54MK?8nCU5KgIN=hD#Y-vb#7Z|8et`<k=AxYWI0?KSSxd>(KS
z2JNF1N!pOKdqSnRc48kn-j3-vFzFz!OeoAAX*cs4v$1GXVsm-OhhNLI#*vn)CdT`U
zOz^(0F23Xqyq(6wX$1x8u_w_(j~Y*OWXGF3{-_3RAKR@luPyVb+J#~#i*-PwxmDq<
zSE{Gz%Q~b+yt^_zd;m8v>ufYog9Y6Lm&DNhZQ^*9LoaW^s0GA&rNRe9i%hC2o43Al
zM<$qg4t<f8C)Z6y$QqpMC5MSLd_Y_G>l<#2FZ3H)jJ1*h3Bopfpki>l%R}l(%12Al
z=19(Tb2_Romp1SpDM*LCm_&yawso9%Q;KJp=)(EZuX_b8BDNL&4dNl@rC&|A?B`ed
zb7Q_bNhBNdt{?jy78v`Z*vPtXegX<_#-NhieE+=``-)R&y78Blx)pit%@;%7Q{hWe
z0UQVd*T})k-LdCuo9A~1KojP83BKfT&@QOStD46%?$CL3UZ2(Tc1;z>&kJ_L#peVN
zW`czuU`-3SppPdN_8wF*%P`kIz{!}=<Zj5YnYh+uk7;pdY*!_N77C{NC-!V;fkny&
zX#%lObZiBE+qBuMiN$VkVA47LEm)Gyh?e50Q^>GV1sP!%#)nO>Sqy5&d~jH&q6&ki
z)w%8dNsj65smQ8@4HlWD@?Q6V;5yGGN0DPe%jXGJyv}tX2DH`P3_aM9B-wVY0hn`-
zFFTbE>!<~nYJkpC5XDo%`ToRtWY3W8&w_~`SJ0?nj*MR==M`eq1N`J=ho$b!5TVew
z=kwu1v{_l^`yBATySDRvd7igV#dg@IJmL!herg{a<8=Ep_MwyM&qQ%M1lnV`+<`?X
z%Xn5no)&Y-+^Tp(COFxxVaU{T$Y$o<3;H$Vh7fvo#}!_Cwk0>vC#uvp-i<O<8yLG@
zJsqX)Z1j3+*a=;+hK4(qoKW;SIQo?z@SE-S;hwNuk3m4Txn*3G3+Y4~=T!s%l3B`Q
zwzFaZ&@ysWvPgtxlJreQp?Tpykd$@Km1_9MODtFbnU{9?=Tbmr=2C=5-0_mxo7Haq
z01)5^oU|uL*`7)xE?n=MT<%O>E<J3r!$U-vEuV|Hv8E!smzA~)dUsXKcPZ<|Vi!<h
zSLC+nMC!lw=PkItF~<rAtF2x~U;bFihT*&HbMR?W<|!2V#4W<q5vWQNCK|t=y0=t?
z#6&1(Gc2Fem*GB5<bLcik^(4J7I;4?dvO!((m)|C5R<Z1s#C8N92v8qd=E5V=$j`4
z>(9ntxieDm#bQL|@{1o==_{Vst-mWh=2Z+ojf!FuS^*5R{X*d~y^6E`!6^~ux2NjX
z(q_|qh07n>eSO1~)>p8BSCGhHL=@c<4EQ++F^{h!k}oIIrtOzlHz!FB+;1#KPKNS{
zMOg62EqlCQ%lPp4TOE91CKah#wJ&)Mg_jxQp-I-)PZ!zZSh&8IZfQm{=X!InHWycH
zCR?`FDmY@TM`Y>RWI6bB{s5}$_@AytjWTTW8oV}|pXVj?wb(in$q6fSaU{#{E=F@r
z2cY7w?En<6+!iNPbWP2iT$CArGE~(WT%C(}9z8#f@wF>@VaRunuOVN3i}Ssj*63>8
zS!XXlhKI4#4w%5<iQ1J`cwg6r!qG3^Tj6?*%Wm0QbQt5AM$+kA8}o(OF8SRnm?r{J
ziyL^0iDEoE(^hRCL&L|TY=b}S()~S^*s+{uPixlBwG{4e0}ixK%(8xaGHl#uX{5@n
zX$R00CjuH?q-Q3*v+GJ5Ulq(JpRD_N`04zie?K1nvy7#fAE0x+d1#|1uU;xQM$<<R
z1<Gqgjh5uv464=`FR}v&lgvk4wz~K-qARvYZ(vB5M`#$YCA$>!`1`v!AH#vn(+~ED
z^ZjiSuA_s+_naCOXFpZQ7p%M9Z)=zW9pJUkW3kyZfV66oe|_lW4P3Ed&08#GA~HGv
zUaUWBi5j{)wmJvn!`oj29pPFz<UG{^<ZPs)1i7yU5fc8SC%&_=u1qZ)WnCu`r~K?4
zZlz{6jYjU~i5ED#qo_z`sC0XQ!)^{cp}Np<zI?<XW0d|v5o!4<D*i)$r;H(8rQ=64
zp-s>qSgg?7H6f5QR+Qmm2-p5sh6cNZvIk_Cqr8yn;MsY~5NROYjtor5>X{V(m~Wq(
z*auCBdR<mpRF<*OCW9M!A+j~x^u5vkoo9jLsUyw|d9s(5cV0*?M_ffA0iBGQiq@;M
z?q1!AylYV~-2_kSx4%P=7clTTR~n15_x0ktaDR=F8MPHJIp=x;c2cM80x|JV9|IIZ
zlAL(q;ZL-fKnatf;6U?T8b*Yi0lTzE_yul(C-BjHb+2s60Mr)VZ8!2>#jG*<yNHKe
zNcE36lX~)zBl?-u1j+gPWA<{xO)ci{wNk~@0-h|7-cjLN^~#6C&~Zd(ZtS&R9WOdS
zEvig&Hi%CUCR>4%dP$VBTQCHUPrxE^hx%^N_}l)8X0Flr4}J);G72&mB3^Ex@hy~%
zqh;jvb}u~bXzHLTh1!~GwZF7kvwV;S4@2V_u5@*;*M_}<%3>Rt0E}Bc_esyM{Rw1O
z-|*#3`oetiE~eOzgmYuNi-0jUNyHU2D2!aEVI&Eg)q1DO=}VpQyz!ABxjlg$<|p?3
z_fD#5G00>aCPdY~b5R>CRlw{-okh;tX~YS|JpB9FmW}f;_re|qRog+z=&49Z-as~Y
zf4?FLi(s?~>CAS*oE*UIc3B8tBFb#)(qMRf-J1=KPWVg3=@XR0S%1)CrsZ9Qh&v6D
zR63@oNbMJhs|TUN&>j^}Gn<dX%R?;rUDLqeIz7Sa;8dc5vOH`h{gl(sZy$})6(!fT
zge8Ujb#OtpxBH#+hdMx4>O1`M#)RY}f{&@4PXCZy-QZ~eWZqu~9lxb7%8z1_t>`d<
zvbyihq-DB*H7BocKRJ*&)tj(K2{_1rz^r^Zg-C6zrj?`WJO0mZRMYt151TG6#})h-
zUb0}k8q{Bn1T81?xmUb9tC3}CkvR36sXC5pIp3Ts=rxsoD8;<}98_^1xz@rro<L${
zv=~v3E34R3e>WyC>1RFoB~XT~&VPPydrQCdECMxTJ@z4b{xjR=PS4nBypKV2@J_n5
zkqJRky{ImiCgL$C(<#JL!n419F3%MZ<UX2sTpd}^v1Kz1<ya}%`f6bA9NaGeY)p59
zzc2csawGlx#g7=R&|T1QoC~cQs52mZpgRXI?sw|1Y%Qu9PZBkXdlcEb^Qak3Ubr?|
zi7;=eZ%Xsx{z{zd<(q<K+=rpN9m~U)skVd6%W7&x=K=X#LDZ_2jOzMe*?5D+P-3ch
zU+cC$piW^puL=Lz@W)7ExAI8Xu?}I=*tK-sXShSRPX@}1|5;+h7!$}{cT-_dXLY!w
zlc!2ES*k*)N-L%R)!rvM5k32l9Y9+$$ZY!PnVgnOfl9gL_gLhR=ZH%@*RCZYrTXz~
z?6<|5TY3hi--Fo@BMbfmV>J8nXHJB$)L(s+Y9;y5)&^!l;FqSacwJY@e8&ns<-vSY
zm|fPh-QJ8jiQQl!e;6i`1k(5F4H?6s8Sz}&|E_9i6i!_$^_`<w9<UG#eliR_U>u9t
zzN0Him`(ABNR)ChCS={+IfhXbYG++)n)p;~9e;cRjyE%koQO0|FiU*N3(t3+K93UP
z>}bljylCcBA0}*m^!T|`!<WtGqu4@t!*q8v2kmscu1FlzynnsDG;Wk8c1$ITyb>y$
zpK%a&@|4^^4j?H&9Dc9<A=_xe&SW;%8S`j_KVT2GBJ)Dcx&OPKc&6>bAM;-;CtBb1
zITBi+hHDTVb56WL1n~Fk9+;V{#B;GV^&q{@fE-qJk=+wD`KT+pOG@qBPJIZyTo-60
zw}a-*qVKyzyRs$2YVyo*lN@$4)kzk^S8K0VbFX+QpDmE%0^*-23D}+jHmsgJC^vzq
zi!G?tL+bK~c1ha%%YD-xvKfz#-#v(;e#!BV(+poC2J8tDJq|Kn*XL7>8ibjvi&}U!
z?7kYQQ6e130GtWS?^@?53SA$>@s9WAe}0=&88>=?cnbdCUCJvGBC?L%;r_M=nu9*!
zR}Vo{DOLABx%Hk&(75)fV*$$Z>3{4Gi04je{`Xu!0F*)d-(?5MO#=G$KR1_CNh1ee
zJ0<;}uQG+cxZU$VCw%)d;~O`)nZy5H2gA23*!^)xmGxwCu-o+U7p9p&My;!$ZEqTM
z9t07l26@o`?hx;~VNC&TRpPkXq5nDYGe1xb%{)Cse-%F8w@Nu@$WvroA%-Mx7$VTw
z2f)>M!F>~N62yT>jV`c(>g<*K52VJP+*^tKSqg3@7xSMbI#>-HNMo!PzC=AzfpY^W
zV}_6!DFqq%^_BYd(izan*=QK3B=2s$pFa97?)kshk?sLhxZMr_@&U1XIU1ovd+GmK
zGoMoLRjV=pm`6^+GnK;IXVZ}!m)CcM3+TZj+W+Tt>ks#~SE2s?rlLQ9kp{U)NWS-#
z`9G$|$6&!noX4>iBI9w&n)9ds_av0Pot#@Qj&3TF)A#q9e^vaFL3dMoKvKQU`0sOB
zP?@+?#FBN?qnIIi46E7yxue_-n1xE1PqLJY1_rUQNL^=$Hs|5=bT|4RTM_wicm}Xa
z>%ql-Aj)$KZTz2Q&~cxd0wjrenO&DudH~5`z<zND(%bAR7OG~0R;@vAotK+9*iT$@
zdy-_qr&6Ewnb`U?j{m!x<ISL)CB*i{$gm3VT_f6*{Lh_ii305#P=Kn@|J#<gPV*9r
z%^L>6;f^BzHjnXU(5rQnr)*oVcPn%H?EGvzDc&HWRIc43B}t^J_|I~2zLVh#xGed9
z9_<2=YdgE)%yYEAuu<tHve_Kuxc_k^k*ARM1>&u9Kw+iNilhLQhB#2appAe7bbzwN
zg|fHaMNH~}ob*4Jw4ho(`IE&i<!&~8(fpH<cef7JMFQq-9Hd*$Im!P|)}DWncp3fJ
zMmADzGa=6c+c4IAi<&(UfEeBX`0P_qRwjLXL7fZ?#pWRahsptAbMqHBk_<JaKS^1#
zTzm9xuJ?VU&(Dw4;*_Tr@P7T{CbT;0|A`0(nj$$@Dl%qUOCz)=>bzDi_kU-Gg7<%e
z3yobWJPy*6)GJiBMO|H;@R?05a*k&O#+dXUZ<L^X7XWQ2rvW27_K|1(oLkL?Q5>f6
zg2VBbD8MZ%xqs||m724=Ec8FSaO4|>$`BrtJP$;T|2G7Ta()Ax73Kj-<jZ0GSSS_$
zXO+`=Ilmhy4?rprP{%dFp8t<iG5&uEn5rp(lp^idb2t7-SRXl8)+k_OJC)hmemXur
z(L|O-{85T1-C5RqaU|`B<FnIS$3D~jg#hT@-FvS!^H!OoMHX-xkAiHp<#)1t0CMjD
zw4K$s_tc5?5jSl1mbo9~lMJe;lX~owwF1%EaT|?Hj4qPjvn49^_0XEPwUVaP44PQZ
z+CO1J^X}ST8Onl+qVUHTHQ=`G0lQ0f5EUvre70=;3bZ#F{6td6J<9I;_TwyW?FIsO
z_YK9<>)-3z;a{~aZsnP->&$yaxZO7Y%mH=Ff9ItmA7+xA5Xh(Ns;IJ-Uz`O9Gjnz(
zp1<yo2W2t*QitOE;ODkj8RdJP0=;Z@(3tv=?z67@X<Cfd!gUh~oXC>8125VCN{P&W
zNa9VcA!%0Ga_s^F`O>W18K_kx06QEt1oXrmKr`Svk4IZr*2Nka3%n;dz1{`tT5j^|
zHXxk?55Y}LRM=fd3W|4hs*o*JTIyZGy-L&WOjd;B9np&q+MjmsCt!DC4~icCtJFNm
za`nMeq3G@q-<==v(x6F_xdZ&D*C{^301Xo82vj{f1C?Cb=LO~_AUkt7D{N}tu)8b^
zX^}h#sE{O3<8<y$jbo9>-9eipkfGf@O=@Srih0~Jv1VOPj9Y(NFT!j(!yFLb7xV|O
zGRr!}ZkQ8mcmeR&zY8o2Dj=|{bQrBJAwK}2hcp8M+Sh^^vrm4I&Z$QpK_nbw@CPD8
zZ{}kRdpzoKS$I8m$WLyU1XEInnw$fiYhreJ#+-txCf>KX`sH09jj;AEvsj;{q>YQ-
z)c7aw8JF8znhf*IxB*d}x!8)1%O0dgLh|dSt?#ZRB9|`pv$>SN#YdqBcK>{oGbFnZ
zwz5rBp29KN_^=wjd9I=jwIgr6CIC-fI1-*HE`qU!G~<7kyJ;#2{r%#T6*B02=Nui^
zx+N(z69q)4O6J3qu;&Iv#y$&xS;V$ygae0*R)N3L-1qHeY$~jFn~#JU6M%O7tgrUZ
zWY)6vxu$jsFE}^N)IQrD2*HUz*raX}dVg1>eBySUM46gOz;~86VM@>Lf{>MX^aSgm
zKXG}i?vuPkC&o7dur#iT3k1k<8NQq)aii7feB#dT)JG%y#?0E<`>Ti&Cl#KgJK&4i
zxlVT}&VYTFR%VB1k$rBgVrP@MtaV}OQA1a|i-i8XD5diQ$n6&9rR-<+o*kk_z>uKN
zF#LFyd}DuaeH<zkLExh0F@{-C){%qRK(vrm|2#K@XTj5E#$tc``jpoRuE#Ac@5?1#
zC?nPB<Y1hFr${#vfuI|Y2mm@X0|$oOfg%%M<|B>4Ckig~B&iO_DG!ohP;^y2V?sb9
zM-!MfRmpex-v!A(v2W|c3a;$LUZ1R9=n&YKYuwwy+MQ<jo|4*Wl-IRjG_yUWCF-)g
z<Qj~ng!%X@RTBo{eie^hjBrPI(QK3}Z2%W+9|XZfP3f_|1tO{Oq(T<{$`3%3OTrfh
zDi%P$wDGyfx#v(gzVr$<vA|?KAaiL$8TF2p^k&h%{UbBmdXK$z6W*QcVt40Y%!gnh
zz8}BV%7Q9MqV<Ga4WOoze+gG}E(WmRZ{Gz{+v=@BO{t-*Pak3!ww1zJhc-^INlDP1
zy`qB7lYs(wXa=rpZ!n=F0z4<~L#(BByPf)nJJ?Xs*$Fe-VC~xk)#KurD}FvgznQRu
z2~sXeh76hZR(Wi$)gRjv#rZIIc(LToWQU0cj1iD>irI$v12Nx?&@@@E;<>?+Oe4VS
z)^7k!KRi~cn-)S_(fqU+R!-CP#<`+44f5DM<he_8K5^p@k6R?xp1Z&FRBXMkP2lv+
z8CBC4eIvitP~=ogmBqXnvDpV$gfyi0Smc}qDO^@@(yT9Ye1fIYWu%g2x7S+2QePCx
z{?Pt?Il2CNXccS2#2eqrS~0yF)S#MJvlA*Nj!;v3$!~W-v~}>wr-A-~UspgMWS4nz
zRj{}^W2^aEdan2)(3ssv{U8^Xb*|0NNOqC|F)8;kCz#^`iyviYb_T)^+{zIW*3A);
zK;?`v0t!Y<dZSC_P!p(En<+OU&WV2NR4lRY&LCzM#Ah>VWPg?~B=VZHk?EVxdm1JB
z_-Od#op8UcQvVwyzXDXIG^7D8vEm$HSjm6`e=6QtEbD?D@rQykpy?~F|A@I)-;(p{
z2#CHXz|;%7EaQ{Gu6JSkTVEl)uxt<h48~kQe1S8Kbm;WJEh&vPFH-Y)ghS$-w(@-#
z1DI3+L9p$mOjW)<dAV+-5*es{?fYR50Om)y`1J)vR=arS=C2CA5?G@L@lFQ9OXDg=
zH5W?^k*P^D$=93C`xc3wNT!}1OON@9Mm$KY!GwvT8;pdhEUj+Rgo42s(D`Bxy~@?a
zJD2@t>nenIbc_TsPADv~N4)?GmtR0<iHY0hi+aJuXR2dH(bag^I+~ZRomW{zbiXF2
z-E>RECp4Tq#c7|29NUuQx)ksyCkExXyWmv8mHp?L`&GXj4PL~a1jV?QZ^}t8?=P>@
z?;0%djdNXU`Rx!#@(Y6eA&~6IvA+`5YCKm^=P5rWZNz~c0AG^!O)(${pt&w}|33Oo
zz&_ba+#YwKx5z|#9jACQ5oG?cQE;SBHgHYbzTCS!%nlH4KWs8*mZz!N-^tu_<9i_G
zZG@t@4A!wpQ+B}?vFB?0?b0R6pC0MxunS7v4&dt#4S?;hoOG98jeLMAckI7xeS@U?
zs~D{4amtzd)@2pm*U78`h&7E?MU{)sf7MJkwQj85HkqB@S4`0c&z{d1FIMG8>y(XN
z9^T2n*>Zxbg6|iPL3fi;&EYx!mhxaWfRDDIGhc;VeC{w_Uo!-#0P#3Ur-An61yJ&7
zt<*?>$TF0-9Cmnr9Rsj0OA+z0eJV4FoIQ1iK&u+?YxicKfr5bcovFGn7)e?R*##3z
z<qH`!z~IdK)X)>$r0Xf7)=CJ6@>>MjJh|VS!w3A|q)BKE_<`J-4Hb>tr9@{1B<V);
zckv<v67ZA}Q=ot>0oS6D`|G}c6<TSyf^2i~&q^S>IvxUzbWuoj3q>)_u(Lw#K%x6k
zeadt6u`)E_l^DS;n*<tj*lB-*d>CNbJAj-do+3=TKn(0l`yFYcPjf4IS2csPSZUK|
zjH*TA6r=8P#sK+t+cCdEJJ|&5NVx&Xedm;wUqDC(_{ZMAzi$z+^Ti2eSYUj{WEaHY
zH<$J%3*2MnHU~4moqlg%2G?J27w`H!NpEuIpU1bLVr^04bR^$}dMdr@>zTa9-%lY_
zE;6ivwxkeoJ;hBT|Fv)r3U~v|QTw-x{9zMsW;;Jn<fJfY-Sj%No`#$FMnXcM7c|Ta
zzk3uW?YuZCKiK)HkR?SKDBL-G7e&{h25w=^OcK+~Go4z{4+Fa8lyULtLojWXrQf=M
zgZInr-?AH;b1niUW{+oiTh))$vxuiQc0gZPbYZsEL+bJ5=fZWF_GIa<l*dghIY1A@
z5@})?#MBPrc&|w;V>V)wpFQjliZ5&P@Gm<3nZ0)nG&`UcWW0c`oxePCF*mO92;JGN
zs#izBHiybCS2q?`PrhPFx7e_I`(aiwUf9g}1ZIC#`TXNRb{^`rf1T?<itlpG%yhK2
z6!?6PS3Bn^s9-p{37CV+KF!8~98u4CAC1^&hKtG9X(maCg4T{EB5V(LHSS{^$<W*<
zn@}cIHvmTp?v_?o8+wKL53Y0lz4IbZLFb2C@$j0e)OTgVfwfuC*UkUAF73HE_?*Yk
ztNRJpw}~#`if0gB9ZPj@W^-NV&SoRD_5;T7|Cl7*A0lTgGt5<-FuWf@<XBVuAr;Oh
zy(wQs&HhAZJX`G6cMJfZ&gA+L!@gEmcUnZ=w4BAiH}&gV?hZg=T@H8sK5JZ@v<v22
zs;7}Drx7a0lFq1c4^?+90cSAzAukAKM*Nz->_FY-2GFV@4nJ}<&bo6Z;=n&k55~c`
z1e}3pn<d>O{0`d4px_f|by`V{K5BeTo?A|;YpiUfmrVcwRt5v-h2}=8FuWD+kv;$<
zFY@2N6?}qi=0w>XlD<QR4C?Cas@7>Rma@fLvaX=Cu|KbV>NYbsE4<AGJ>|~m3R<d8
zb46l6DjE(YGQk_oKzZo`$?}K>i4q`$xMq$HR93Nw?}+SDD`*_1;aUZ=bs`tJh?YXy
zN!6-VA0_p7fj`=X#6vyc9Y7p)0FZq)bRa+X1x`_S&?-zwOuh>A*^5eXP<Mn$W{{N0
zy6X&^D0k?^HQxXv9tx%aWP-lb7pgJKQV$tPy;`yz$B;ioYXarH158V4$zYPx#lYBs
zr!G9&#|B99S}zl2SL(~E*ntbIfg~EHQUPdhAhd_~mf2~KJbwcEsJ5D*ciaxNZcChB
zV0JD-On2+P-UJJK9<+SNvs_=vl!8Vjdafb}MPHfZ+`8nn(QhaXc%%{q5lFrkJL8-=
zZl7AYsvLeCj3gZ=?^u*R3{8~vat&(cA0pxGkD3%2z6c&M=tBLlR45pZl=Vb{YNf7_
zS<dmy>WW3Thdm9>8Ew`!#{P<W<fPIJt*SSO0C%TgXocsFmh)qEUb)qEx%X0bA$5G~
zeLOUE0cMLFIjPN?h4Y}{!#N;8aoOl;j{)7hEegM*9|??)3@d!@7uo|J81z7S0N}{c
zSMK^W5q4-6Tb^>u+e;V6v^CGON&e{`8A?Is)+$aUI5tbYvqFO%|D=ut1bPyj0o$n9
z0J5NufglC6e;+Iur1l9i%mYfou;xecslGK&fIIXO2uj^T9$xrwpi(ZW!h4Ou4aFV%
zavfvsAeRjo?~huRuOcPt79dcB*&fxrHjsAgX1dbT$1}F|?$nQ26T!TKBo{ywX>W!h
z1k(2sw-7lG2BpL<nIA+O@4#B^h#{I)C_c@khhK(5$^Khoc23LcJo&77wm^N+mMgcR
z3|n!9TL8l}&wE<A0*1%B`AS+q399-X<mn>ZTYdIx`n<X{tn7T-;(&a5nu7IOV<I?@
z^k+R~X8Y^9dUd!3sU1)pxqdN$|8%rz>#xy79yfF6643W)gN9tUibm!yPYF~OYNg#6
zVte)}9m-scen>j9!26WR@N3^MKFg&KRU&YyngT&k4j2RiVJ~)a+E-c2-*vBI{gti1
zL6Og!ORC<Ye`SX|6?B-aUEoWSV0x*{M$pQL;vFn!(AWcC86O)3w7?yg6L|i!tQ?)8
zb-|eneR<JESTc~x)C>Foe1BLhJ-bW-MA8-CD?hnZ&<EHl#VA}?k8s`Mq`&7_*LUD&
zZV&-psGuFjEK&(3i`TA=UB9PlaqjLZ(!6}|hgmvd#Xl1@bT6N!oxP<MzK|4iYIm(#
z=a~#mzsGi|v`StKoC2gMJrg2C_~Bk7es1|zp1K=IVyw`F*+OBNR!~=o?c4tCrCM(T
zz9$=1cUjVHAC3dWDmcn>PoghN>AmW~exCimVrg!-R!Me-9&+^3QId`bkSff8>E15y
zwAMg#Lh&;;3lhkp?M785$A9I?$I`*I>Xno$AFG!D==qEj9peHV?r)%=l7)*u*gsc0
z?4zl;()7ktR|FMLaPpTz;VZRn>(wO}tlR{6pDPdIvz@`b2`b$I`8)#++WM#L&wKKp
z|7-Ja8xVM~QPa>wTtsftDz2r>_HkkHTdVVoyc$1^@sgLdf9{m^<ES4qmgO=QIr^#t
z`cPHrhgiI+yJ}EG=};-qDkXsNDZ!~u)-y-{%DEQ>h^*-iCRMKES7Tn;M!+ZY;3seV
ze9A;NMW%%9gqSGm%b*NXU3CZ^v7_6O*2ylFB_sCpK6rs@JpeAwJXdI=tVoU)DOL4k
zxsrBz&TWvK7cz_(b;bY=*z_=;ry~qg{e3MyA{gE-zRKvVLl_IxWlhpW=7tF@4{d|l
zq#KslpP%RiHAd33PheyY)s?1jN8X5vVtdN);mcW*!LCZ<DOB}eIOn1+Y#$<8#n31r
zXjV+tDkyFz$h-$_WSQCr1V!_2NRW&oAQ!+ne_2#)x?dnA?|s^u@s1lH>W?uW`EnE$
zGE4)r0Ls?ifaWv{%5W!?Hz_Jh0nDeREvm)vtS}&uiDA4i9y97~EDn_!XAvQZjLeY3
z(fimpyFuoSvqyp`fRk-M7e-06OS2qQd&=-elTa{~p#a7Le*?OUVlcFUXT|3NDGI=S
z+9DliNbkJ0lBGK2l?32URSDM~TQyxNfp|^nv;M_r5gj@Zbo6CD1cpUiDqfd9!zuRA
zMn5qos8j#Fh}^`alDkDqAh4M>an3oz&@K`&SpyDAgAZsmL{~Yr!8jeJCHssf4_fpG
zh>w<N@o_LnW=Yywaf@_<aMaa2O`BAoFi?S2oDOVw@~?YU01TBNu`%RimjLnBX@ern
zXQMS^_FN5HfR5=>eZ<oB@MDL;J-OWW>LL2xX5It+<$o41{klg^tXY5DECWKp)eNpy
zihFENXs;MB7F3MzxYIZn&MKL2{DTEW{nY9?j#1YaR0&~UX@FZHSNa<loRo!8oIr@@
z(hs$G1q%h?s$rW7NQ$3Fn^08(IBKF07%!7;Hb4(2et~OXAiCVv4?ufJN8?|&5wEA#
z&O>0uC}j@5yo$nO0lcV<MV>IpL#qq@+<6yKoIe8Mv)59>dQENwC>v0fr-LRAsUH0c
zx4b=#5Vh+(4^0mWZbW&*-@q9)=X^s;59^h|)TYfA{NS-8_TCzrk_rwzS@%`0f(hI&
zhyVUE1A@x|uz<Tog@Rh6jTv!!(1<Lj5rSTR`vJOeCL>I^l$jx@oW{Uwf$P8DWJI9c
zV?pK`5lepexuC+FdqgAQDp$EpgI-Uer6`-E<8<f;nJ;k&t6Yd^`_~$(9K%GrhvqmE
zfHua(saIit&qMZ%nPy)LJ)2Z}AKGGq(USe1=6PkV3q&0@8p?S`rCCVA=LZ=1Ii7B%
zW{brMo@~-s;9q;90!FSJ)iU6g-)*1kL(Nv7(@F%C)Tht7*?(hUut*j4PL+c|WyX)m
za<}bBvIz-X9IkMxP?-bc3<OPnIK*^99aKC-lRI+v^AOvHt99;$GqxT4mn~cPtp-US
zw`pL!>c9OxWK%SXH2+4ozkLF1a*pd*?PlXxmMWuoz}_}abSCp>Gb`k`Ld1XF2qOZ&
zqpxTHDx&ncio|P1#jYq$aog7{*WI&fzCEYAk>~X|S|t_80ds8H;-o-|2x!6zL&Sv8
zkO2HW0PT+4j~Pf-7TfW)Tz_NNw`UO2dN!aEXn%f$uE#|0Qamnx^5h7sQ{nf}pQSna
z1qMbU_aP3W;w>0ZJF`5<bm;FetwjJG3|Dp6lu`8oBzPtOKlQxm{7Zrw`WUUaz!nH(
z<M{5h7aaS06+kEMTfRaPG*MpGs5TE2K^*H~ZcA_5+IS3x1YL|oNOCv>CN?)33?e|$
z_j;u<u4X0j`ph|~nA34Tq;n8Rl&t^-uodN3Onrj&VdXf&Nj9*EmGcv+E3gD_?mLwJ
zMUnpcXFp*1o<<-<bU<UjTN{9!NdhNq6z*b<-7W%UkS#>ifEqH}EGKjm41>BVPwP5Z
zl<<xJor9+XDd6UapUtc>P7DI77%`jHnB#f)YrTke0OMZA1Or~+OguD94b^&^cJCEK
z=r=!cae@Ij+1&)_T_zAs<eA9Z9F23T5hj+6g&tS*_W|Rj{=0(^X(6bYB5oHk1g3=-
z*P^v#^4|H+{9Phm<R&^=Mx{Ul#)`um)T%;184&#gPS5%H-my-W%9s$Q0H)G01ro9e
z{hdl&2rXt1=*t$@z_fJ&5L8HG%H8`!u~1*L1hh^xMF4BS^@R?q1Ayko0R6xr+&CQ^
zJFV)UF7Lm%<WTWV79|4Td4;<ZVq4JfviMDO=95c7OI}~;Cd9D-ca1mA0(|kLt}_aH
zqf0x`h&kdR>7s3S#+^CIq3)Gmgqpn458ny!vJQl14lW&ZHvu@vY6aB8o#{aNLD-6p
zS>b%A*-p6Ql2iV@#XnnhpzTt+Q99fVMpircrKv<Sk^2<eqKbi11zG*mBv8?FD-!yz
zeUn*T1_6g<p%kT%zwGtaafOUkwny;=o)HA>%FA#ZUFa-SJa|Tt0IKhzm;jRkqnC!N
zO}iTlvtVAZoN1W0SO*X^le(|Zl71Cq<>%p*URY-E%7Ya1q*Q1G+_iw@+Yz@!{r7q6
zaUm^apmx5Iho(iL9qGu({auT)QTMhXe}a=S;lQh|dY|>euW?rQ2d&~qX*TGL0qkBv
z#Gd(8tVpUq#!^cT<86~&Y^(YU`vJZRGh4n1Z3d(S2D@zGzYz;~_Eh#`!eCbW(ZYDZ
zu1ha?4?_kW`JT3%O_HROZ4%&d-==yFw?wb%8yJ0Ate%Qss-z98?B)FS?!Lm$KMuWC
zJgIX~?K-``K%hBz4Ue!VCcq;1abx8E7Ap9snL~E$Vr4Y1rkaQua77vc15nif7=S)D
zpOrsA0-|*B**iD6?9bH#8g($hTb+e0DqNw$4LNWQjv9#u8Xk7t{&yEzTt_Nh2682E
z{bP2Fh>n`?x50~8MCY=Zx~D2SK-=_T0~q%w_GE)R`^M@B;3VFjAz8Kf=tFXBO{gQT
z<4!uoEIw?vGL9**(c}0PEc;;=v4Q=#Fl_~LpL6Y8j%RW~acyhFOQwg5gC*nD8@187
z&kh25=g(70UM6Or-GTuDr+MQ;@KPAEIfviRTiOx&0hzSuy@j$nMIL6gPec`ikBfpn
zhy?I>>k6|Ezw^Nj{dk;{)Id`>+bok|aXc8&wBK}aDFcFrVJ@E+rii(0&&3@`o({bc
z01Q0;bJ2^$D9f>&0q@1-$dXpO(KT3x$uKDK&}RkA2DD20c_0cFMAt$TGwYGI*qM50
zM9yoKc*B(w8vfn^MM9607^pLTZ@S}c!WfMtgEwD|YYdyTyC#<rILXedN|_qR+6`z7
z;g~<14uSeJi8bj1eV{(rzC}+k(4~N-*UXuHXh9AJkckoY3aDwIBF!nT4{GzE6hjcm
z*OGlHadq&Zhkyxv8^4eQCUS;fsu@ZFlBUo`pbaJfo}0fF^cNDD4DNsQ1G1~R?56Wx
zqmMUN0oh|`f1&{-zu!)kyQ)?4m|@i~Qt~n|e(L+4dkS6IXGtvLw%(b}8-(sgigW(?
zyUhH?o1YoA2j5ozb5DB7IDGKgCm5yv%KSrxtxt)~UO;C(14Iq#Zg;idsdQktry8ib
zf`YU9Djj1jK0J7~#9a<FN6A0pEyjOk?4nQqtXO@JJoDXZ8r=OZ@J^8m_HKzp{YvW|
z&P_06+zD}7n=e+npnzmt<!W|oQ617~nyPUTKTZg+GO2S9*`Hx2Q?5_G<X*If(4?%I
zfiS>PDkZ;w8Bh_r4-^{Nu1&Ytetvn{)0o(rv6>hrwD0JE1q05DS09ru2$HNhc(H(0
zkf1*Q=&7`XStTLTvSj~{86aEY7Y;x{Qs~Y#W`&?~Tvxin1XSI4`kpxR&hpAe#qy^9
zrZ4?pOZqYf2ds9hvR}~oQnIPQCcm^J(+DUo1vRc@m&`S$%>wQa9lP^w6L8rb+vd>g
zj24gFgo9$zC1a3zBVe?lAahy?;%F;_%x$C)DKiZ?l?MTz%tiioZCVtbllR1-9iYoM
zd;`}EH8&zwTdC$X1<*(PkRQWS`vP~ZFLM<GwL?{ELj$16+M!{U0H6pGq@w2VexS5~
z;Z4qOy|x(y8bD{_AtaF)B#-Bpi~#AigZ}M8fEK6sw`aKEr|<f@Qe~N;N8^D)um&9&
z49={CD^7qHQH&U;uPiNEWwaE7_czUwx%vqVwf(6}C%3pl&PQ=AfR_f8Fdv~PlqY43
zA>Y0eOZ$*F76$gM;_X{*<s<-GI)M|3%<+7&ZjBM}Y#IPXe>Y{(esSB+zCS+}6a9X|
zZHPDMm^;X@$q@S_zYG{%P<Mp4!i%vKgy8zTM!`xL?cdOYo0$J19#^n2te|KGpW=19
z3*N@&CMzFsif;k%*R7{)HE(3wSlUOi?N>5H=!Qu^ccQCxg-P-`0Ac@z%pC!_0|gEK
zy9T?Q0?k>f(PGB4?=_R-*6$ysKc8^dHFIROW_lZ>%njppZMluUdXugEW6;U0==|Xn
z`TV>I|DAso1sGoKyT1axevE>n{t!@hjedtebO9+f`%}@~M|<oR`q-o}=q)xQ4Mt_Q
zKKKa(CJ<{5eN5jN?t@tX{n|hEmlzv-FRIUwx5@I%)Ww@qH|ow3K`hPhi70SYQofTq
z26A;IG3n8hl1XDC(At^-7d(9C9wDsXDs3g;)rMd8!N0-%X$0!~=rU%@T(;3sUa`(5
zGaq>-Qf8hXJv9%YPfA&wMdsjkGnlv?FcHL1=@`BBeXZ01c-U7b{{vYW5UU`k`(%!(
zET;Q!8m)sJ^}RbrQWwEI+=ov%>4p72+`V}?m23M3tafc0Nl7VWM=G(*LMUaPGG|;x
z$ecNg#jdDClzG<5oOxQvR4OuMnP)2V9F}RZ_^wB9d$WJX_y2c%?;rbk53S)@_w(G>
zeO>2up66whZ<ShKwGUfp(`WxYbLQO2<TUia(c?UhfVo)W&`>bKc}@E_A?#mBs)(@0
zc54uyc1emi>vE2^b&1|_KfF%6PvYq?v`4YFef-?CLji#qVn{zeEvp>U&LT+8nyux?
zB3g_^@tYex&AY@#CDtlU%yWZQ1!s$aI2D}szfkm7=D4txxAko2rx>bobMHM0Gm=f!
z9gevV$@^y=g?p{JGxWouoO&Yys;wX=80_0c==;0YVkQjk3B~j64;z%9F&%$peP`Zk
z6rT&7FZ9yZA_cwxJP>2T|IRutgH<|Fe8WBpaDB~-pBZ?}pbxvm)bh@0GO+QuFX4}?
z|Kd`2y*$8JM{5kshvou(Z3J0=PNYxVErcWqXCl-PCA6&DeM-pYv<REqOV|Sv#?KVW
z`rjHp7&@@C>^g#Rx~Mr>YI>8iU~J^a`#-3Ocd?)W%dMQ0U~+o0=(`HF{YeqSvP~Wm
z%TJZ7i7Xs>4d?Zs;G{SY6v1B*&ojm3PDq@(@dODlB@wt7&SzTx<)v-S8e${>9&(!z
zjLw}Og0ND+sST9IA9x8J?po(g1L#|nR-cbNB&7Yc#|Q{T+kc(z7+VO^D`ISt{!2e7
zB|f8q&T!7R)Yo4~u^!o3O}pCT06kA1ito7$`ds!Ya)Z8>!}6yy%G(}(6zlwQuS0-&
z4?ZKN=t9`N(07-TEju#tLsh7Xje9p@tqL~4g0ME+v33|lU;GQBjViX97#a*!*eZW2
zGj2Za&E2xicV{}zYYssEh1qxPz&~Y<^cZMpg%%KJC%O3EdqT8pfb@0NBpSNUNtJM&
z;A9qee;~JK6CS>tjGu^Upm}$GDx<|bA|VPw4`a$}g4p$d{{LeDdX#PLyt9CT(I=L%
zW(tSjdb=Hz*!)b7e()D_sQ%=0fgFvT71%8efF2qtOOPE~fS0}H`V%<;HSyCjijWU(
z$Ky(d#liPvTa9=G6Vj7!+40r?uR+=+?EQ#lu%k)f-{Zx^oBbwseVGA=V*KY;Y8=8U
z-IPsL&a_rgozF`LkV}ZPB0YOnpN)pp49UeW+(MQLd(5PW<_!iA>{jXR|Drn}8RuJs
zzITRrhA51W8+u>uI>KJ%h;0f#FtggjFV+C!X`zv!Q?-gEu~nrapTj@5^*k+sm;Zac
zZCu1aU%(A?yEG6w>r;bRl@YKonIl6=-RyROedaI2OK^{j``wL|%eun<bW8Bq#m~XF
z+1(aC963Iz;VfsyZX6k7Xqx1CPc1Hh_y_sMii{P|^G4uV&WYE!X7?|+yRQtL+xJWW
z43)A9OLJ9i?Aa{hdvEN8rMbX|@+bFLJxEaLbzoy#adP2HYH0C{8t@1hkm99Z`utCH
z5Kyx}Yo~3$uWU`1ulk({mu^y#H4W)L4d&-U<#}T*;WLZZEAmvjT^6x1oJxMhAj>`A
zsWMAt0qnb9+b87+4#AEhc#NMQt`K6lj!$a3f}oXIvt?;Y&&{vW1jcRkCMUP4ud%ws
z&y8G$sN8wazIpBs9q(GD-g~v8mOObq>Q{PO1%Kl}uTqqAWU;V8FDn-p*S8LcagKWE
zEbnuVd0oF&*jE!iy=hs$EI|_ISTAbZ4g#W&p9Hayw>=GprbV3B^u^=dk7j;^?%h<9
zVl=Tf(<yWpHGW6U#iy{*YM8SM2;%=bmq@_^HS<YP=XAtmtVQ9X4E-*^%lV*xc3F)9
zxQuTm6{^zn(Ym!6>B)6g7ZtKUdr%r@TA9^9@*Zcrp0OQR0fh~7`0B!Fa+s;ibp&RQ
zQO*&*I(ALpJ&xzlvLZQKVT}ALVQsTuH(K1!<jI$kSr>lO+0fL+w#IQrnP4cHYplR@
zVkAgKTn_aR;>vHVZC}|QGQjOZ%vhpoJMU?w6Y)pvdsBpBM4Tk$ufrEi*(u9n-ODqG
z8wm*&*A^s{dMGiG%8aqxz$<yCFXA#|g}4u`b(N1NMJZPMGhQz|>z#m|xZ=z$h9PpF
zvMzlkRv6%T+M=qFDvUx>miHIH2k;Q2+$TS3T09B|sv1X`HxJaSC9Z(jObb58BjQ^S
z;o4yi{CrG!+_~uY+$eSqYa(Yo+txf=`e{Aw!xLOwVx%l`l@;ec*W1&DHo>xQb<4g`
zl?_5{7edI7(vu}3WnW{(dai(HSL{zD5IgJeNaO9+2nM^cw>4|kYp*P}s+)oaa*GJ@
z$<?J;=gvpni3Qm_&(bXIu0eH&&L++VyXxqB{=!QMRG%1SNCU<jmV#Z52l7UAC&Q0Z
zGREjS+hfF1E&Ro1?mA5IklwAIX14D<k~Dh2wo78^4((McIS2+1jtJxEH$orOxItwo
z_CY&s9a^gOfouxRdS+^LA6x<u1|b;58SWkDl+rD*2(eqn9fNowaz)#eVU~kOr7*vw
zeMvGeo<|04T~C-xD(yAG=!`M%3VeN2>$+^lA*4jUK=i+$p7cO%_U;Dc9+%|P+)uP!
zAikXiKvRT(8fw!^@aRmtIG@_uR3Uo8q2XhdU!=x{ulN)-1&i0qyGU>R6&(rDBO`$^
z$d#UxBcwHL*b`L1IWZIR5lF!YeNDj0E#B0h#-#ErMw-cEu05*A<FRDYFpWLsQ{RCv
zRpx*#%!?$s(FDscwICC8+nuH5DmRJ<&HS_FH#f))h(J6**IB-80v&XQ_-*BfsIlPX
zD~o^3%DL;@k*qL|C}v+Qajy^liC$-oaP3=Oyu~#{m8yw=BMDxHj8X!36N6k|pt6*6
z{egs7bp@#X{g{GwPk=Akjy4x5h)Ys?#Jc0^`InTqLtH4kl7Ql)6me4Ij``%3v(c^P
zE5rSpz`{X@lTvrR4Ea94!?+f_Pm48(F@B6NpiRHMe|R!uk<?$~;w1l|)tedj^bGa#
z1wM{xW|}1{Bh=hLP59X?UQ<F5ue>!COTzK38vY1IC@1)w9|rI=jt~>2D~pghJm6ea
z4Qj6SuV`Id*o}M2El#Qo2dRG2jE%W@)|rLj;mc_S+#~%=RerIhMEdnk^Ub^KmF!+L
zfrvy4dKv@J^ZjL7sPPrj2@7y3J1Ke&ds99wZ&CmHy@^}vs_4Ik9Y#6>Swi+I#fkR;
zK>Puro1S(gTuXw=$nm+i1ulJRmPkLm`K&_Cc2!}}uQ;cSXDx0s%)Hb6)v9)j)Rx%>
zkM=2t#j{lh-Nn|jgN?6;9+9MyQ~rg2gdNk?tVDcHEa+Ci9wXwY^5qZd0OBKJ06Nw#
zwvPG`z*X9mLz!Oz(%fT}eQh;(qw&ycx1ub0w@U@V!AI1o6@C*4P2e}h7#+Zg?Coy^
zI)@REMWR?fYeHK}KxK7p0}7}}pcf5zFvXsl*IZ3d-9S&2qh~gTy=Wu@PeB2H15%)A
zs0s_|&{98e{<#;;5y-$w&Ua#(lA1}M^N|GbZK65nS8rsH5VfW2;0qynu4XDRuOsVn
z;NC<hT>y`IH-ZJ+kRM|kek<e&wtiNlw}h^k+~R-tV(ndUj1ifBKc{Z~&)R!LTOi~p
zEzk629Ge9Zg7Ydgu*`rUde?GT;M9~|HKQ4OV6%u+^;_TIi6V+?V;>ruReB{(!ud)F
zh>9(Xjir&4okt>`?7lMH_kH(+X`O`pOt?KFv_+h!ZgHwj&@TWIDL5#MDkG|89Ww)C
zq4C7w6yqrb1M3j5X#L_$kqK%JDz^Ka<MgD_KP0ozW39DEC1t0#@X518Xpepk*O&l9
zs(j8qgqT4tSBVo}a>Ksvb*5cIw(`+tH9Te$>KVi0E!V;aD;|#8I53iSCjb2YkwbPK
zfCfv%z_SD?da{150Eg?k_)r_v$Set3|9h?rj7r*FD-r@21Zr#7I&+N0?sQZjq$Y#R
z=FkZmwHH?E6&~0%TsbhgFFM|9%MMmv0nTJQ;c!;ur8;XqpdJS@3rSM1{tZA36_^7<
zpgBSUT80j8wmDHjTA>86jd^Z6&+)`CDB;DAs%p{|qHBsQAKC)_GUM^`*$ZmIFWQ;!
zzu$%9{An@l(lb_ca`?|T&_$$TVLgG|OzANh|I_Ary_8T|vW&i!sQ_S82|_K0ED<=G
zCMFe-;<AOQ8Qd>APGXOJ@TM~783GeJhZ=xM=2kpX;Ohc#VOsoc{1ikucLaF!+Jo>f
z;6#zPU?si{qQ>JbKSd#MG(sgFXB(Eif#C881bOoV*ExoZEg^Ni0A%ADU4RP`g;2U}
zwxlN|e<IE$AsYyRNjXVVo(A)U(hGSwxREgFM}V_$4PQ_Opg7%)Ta3z@Z+KnjhxSh*
z1gR3P%ejgdI&VKdH1$+B#)$3O2V(>R)eemx+#j?7O{<RmWN1O$5#1d|>%}w=-3l)n
zT7=XJ=|?1Mt<!TdI7Kuo)HQulwJ4h%sPKGpqs0DnK|gSf8^;q92fsb0UaG>)LI-*8
z9cD1l(GGR^!tzw(d}<Tp6JW>JWZL$Z3Ud-^K*c9X#qt5!z`T;ER1GtAXziy1Y3>Po
zq>VeKW)0e^)?umFQ}^iBErzQBy!)bP(8~Z(I!%oU&Rlk4$(?wchMCmdKa_w4g8vAt
z;Ih2Eve=>qhg#yE|1S8;bkS?t3T4tos=lQ|2E&(hGZ&G30%=1R8r%<Lm&s;5hqf1+
zf!^7CLht8+T}wbZS~wT3#<~Jws<4B1_D%Wj#N$A2+`c4iL`kRlK&OzVqV@+k1~=f<
z{D|qCL>D|#&`&#1vj#=SXh=T-t)5KeyO@UfN{HP#tN@^gyrIq6oi{C-aA4mPc^#y^
zLJYJAs}k$#_zs#-hCXSurPcCV{3mU?kmY+uX}~pQ&)3=jO9dXFf6XNf!~yc#Xqci=
z%>(2(YDjmzUfxEc>^>Yy!-lQ3jdQN+i}N*HB=oL4`}Lui#t$J~s->sb<~0nU^4Rv4
zn&Tv^p}-ZR{T@yRmx4*IJFoBsLBV5z?slQU1@GckUc-Ys|23N630QLy>@R031*Zp^
zGPgc&I&nvSs*y8y@^RLTOBO-x+4+uOLcl4Zucyk+U5-@ByaK6|)GM_YH^Ci;iVp}*
zd{&*@h#QJ2)yh~?{IrN??~?b8!BM1d%NFLg^ER9%+O%u4kApOh4^{f}Q3$%mDbtTp
z6o20S3P`Nfvkb*loPiVH@3$`vNiJ>swz_N1?qsz`5SpNrrI7?08FF%aK<1`31`q?l
za#4o5Wg6#*mgzZ93T<?No~5yU3zLDxW?knFUKWyiJqy+k(V?NF4e}a5_mewb28@ez
z!t!JXDmj=wKL6aoX>B8p8s~k_Gd&GhnxwAX*GV;e%KYD-gS1z@k67laGV-2YI$bwk
z`news*SY;(ox~v~pE(&sXfZXs>{Ii#Ex!U$WaQad|Ll(zT!@oDKM?=7zvTih>=tpV
zt3{p+i~_6b1VDbTdE3){MpymFazL>m&pBWu{^Z+n(8*kS5D!YC?E#h+!{fGOCAw59
zUGl1PG)RWYHQ`b4Tg{4a&NUE=2^?bJ|9Rt<%h5Mnl~iH~-|}}Pl^lzCpaN8(uT+y^
zi`hPjkLoJD$~*5#rhv(l+h2(2&#dg;7Z;sWVDd3<iwc}uhY`)9^K^HYPa<n{h)Jc_
zJ~sLh==Yj_nFHY^cvG|P-X8X#j;#i+>{Jr(wJe|Gv*5p6k67AxWV-;DS?@^@wSZkZ
z<V-t4F&Un&jQVBxZ`Gd5%y5j}3q^alu6;Ljn=Y_cAR}DDE%5I6_tsCAk~^<ZauZx(
zx{i||C6&<{bING~*B;Tz2KzeBFqSHd;sG$=tw(%X4~vYwjuwVykrIGy^5P=YI~Vo{
z3E|W3FBADJ&pn?gGVBRM1-`ue{`S)$w=S@l79BFZ@SmlTM74@!<!%93J)QHb-?$i_
zvk<wc#Gt1B=M)GgC|p9sREdU16&s?3TiiiO<ZOSTEyWUx-}B0gzGQJDRL_P-+rX)f
z7qP<ThQ`f>V(2}+q$aH*@W0o}ufqx<PuQ9u{n&gES?0dDj#k8FeQl0~qHp`=Vf8Z1
zT?ZINBb@u<fb3Yj#+##{f8Ta4r3m7((!KR!jL>e=0svl}5cNs~6KmVjIq8DJ*VVYs
zOgnCkg*A>k=-cIIb|$jVDo7I?ZY&6*f`kT@KxlbWBC4~ltTa$NZ;!}r7x10IB%^CC
z$AGGdvVC(Q3_+;n0|A(LtOL40YnfYeo=%4EckB$+X-Ns&{wsO+zgH4n6fSs=dBAZi
z)n$91nZb25AxQZxcX6;UdJK3ye<H!8#JxeFgCBlDieReD_9x<vu2(-Q8%`*Ha42xI
znp{H4F@-ec@Y6_-J_$005%3N|VayQga(wgL7+DgDXJ*efVg2vHAP!wlHQ-`zi%129
z1SHi$25vYzM#4^r07Q#kYKm=!p>P8@5ZP<x(6(2EKJs%eB9dd8y0uk63W!G<e+{rg
z)`%^9*qZkS7?I}=bii)NR6l(A8--}5$IkusM%rRb;QB<0YJ5FHARR6NrAQLG!UMlp
zPE)MSQ7Hd_zqa8zwb;-hV^@f!k%*uA*c%d@x^71OMRvw~0HtIllHrgM1nErvB#7P#
zmx^r4cx`%%0ufAqTPSI}Nc_MXgjj|<LZ{`Z9s4rUIE404^hp??5pmh>4e}NBK!|Dp
zpu@IK6MWY#JJP(-0d|8Ole@3+Wp6cj&CYXy7r~<LC}ir<ES2D2OyGHkAt6AlV-(PT
zwlCz1d;(?D7|H_YqK3Zbc;q$iFZxTjUiKT$srvYxY8ZjSDxY?!;%be|n1FlYQLB)d
z;CKJY+GGI-fC?aYnD3&>tAwXDCBE`jt@6mW2rf<3OA!YjUqlTJo?n_N4UIBmpMi^D
zS=e<qNa=aA%Nd~bZ;y%cW8{>thKq9yf2qZ{r<5C-2nDuE`5xD-JI~``=&@>&r7z8P
z?b^Pt+GCJ%o0YtHlC6)x7v|;wU~Y?%xstQVH(gQQQ<HBhUl;+~vZM3BVTmrUe$Pv^
zX6&79Jk8M7*|zXGG^F!qcsfqZlWxxDw<y!yy%KjEQ<GP6z6e^^x%eOS&>P&gn~d0q
zZ3d=C6F}nJe_n`TwjV8n1`iU&d%%_=G7Z3C@h0FE^Nw8WfbQ6V*FeTMWGvI|W*g2V
zo(tPb3`SmFKRjpZ2f;>j+Aj~fI@_W!NZ+q2gTO|mp02GDOt6DYA<&=X`z+rn5Ih15
zTq~_xclO%4Ng0o6KfSRd$pID3Gu<CTkj);vaJUM!{vq12Sx)Qq+jw`X|9ruIpzc>A
z%zA|IbJ9MV<6?D_A08Q>Fq;S<>b7I@Ap!LVWZ-#hLYdC}%pI!*Kwuq#9xFfM?r$BE
z_0_*@$Gtp5l)5UJzAu}#8aylb!L?ryqzYTKd@Y@%p+CVO;`d1iuVg(x-qyvy4E&L8
zcZCYO3-p0Kj?>+3NY)}xyADj)2v8+PAe3c<^G#<O#VPooZ{3>>p8cKFiwKc}?c^-b
zpMQy{frM)$^)~oFzlrjTreGdaQ{B`~n%-MH#j3Z!6(Ki3XXBNjI>U}geRT{g+dtYE
zXFJjmW#$2gs^7h0(+6!&b&YE_VZ$Q?hy-1}&)}aHWeSD@^U@XOUAb&KlKR~^Y62h)
zTe;VRn7`>o0b1jQ?{gvPD1cGIj$cm&Y{&ri`OmbNO^?=Q3Cs5Mz<@`PZ~Yqw_ewHX
z0*Bq0-S?vN<T_&Uk8m?&N?z5UCDQEuCui7=Ypg`P4$U4pxlU#j)<fgTN}`p)*&h1K
zCnxhvz`JZDuTh*2$|E;SV;kjvu49*mNoTEEVddzK904(A$SnG?A2E9i2Qg)E>2*tl
zOZ)FO+>Jy1wVGe0ZRRWXpHDobIw6<1XXhWuyl-^34*dEUQK7SN56Y$wPKtCW1I*%g
zTDaL?q5{s2NXYMt<?|Yj1FE}Yt>{hw?+j_DBXb`3hKgQ-P-;K!sqDXnFctQpKIAT2
z&$*g@NUvjy5+nxy=1Et_4%~*mE@Ivgdr;ftN9qf~?k~|WLc{s=+E(**$M@|*{5oPk
zp<vd*2(VlD;2;u!6Z=UQ=;EjCeVUN-0wKkNFq+Gu+dMe}P@FoqUa+mYIUnl!e^xmC
z-Nf72aOxA>XRdnz;n_ax*eJAZ6j7X`NI}zMmG=T`1uD}92>od$EAl&{5r|Toz6b=`
zVN)QTE8PE4dE&7lavFQU*-ppm`ZZeE6g(BPCy{39lMJsf+m|F(G5T^-TAX%u=QqV1
zgq;G}YWAdU@o*?1HxgU~YI+d<!4_t_E&jnbs`|XJe26%uU8E{2M?)`diUjo*Du3?=
zoCUm)N0;`EK%h4cKJ{$RKeg9G!e|ZT{iVlA+&KD@sa)TTAZ?EIEfY&V;B)%)RRh$c
zJ)gTXcY$xyDY}o-=8EST`slTRxOs?FS=_44_Rj);|F=-><2x%a8J1oy%jf^k?xuea
zCad=(TH!bpPzb#whcc)|j_QzMq^!a<SAa9xEgahTg#Zm^$7O0ist6$?4f7`%Q?04$
zP_c`<7eU4Teix<2%MD0U+P5sqprhJ%9kn02tUDjZU4Ey8xQvrJ5uaxSCOm4?)&O0{
ziGIur=Ru3{QHUOhw3|JlSw!hF+pCMv``va3o+0kFN7Um^^Urms$c(r30Hn_GfWPpA
zV<8_9dFMkfz_W0Gv*rTySa!bg&AVm9-kPJV3*Zqxc{Bc9A)&>KTpP0pN`Ef&(T<x>
zchjKE^syM2Z`K2)Lks6P0)pS42@>a*-hSK!eP^>?C=7Gkao<huBAx+RyiMF;kQV3U
z)yl6Uk1H}iL^ML9^L}U|P<#HZBOpdZrgBB>Jmk+1HD8M)y6gNmR?w;jyYKuu3^RkY
zT72XAz|+eG(m8f$%A@+Dp8DLnqwQgsFY-Im?7fn+#_#{$g$!;xSsBIg;fU4#8bmuO
z`XzKHTtkClF81=!84|k$?g3{dsb5k2gXF+JdE*JAL00s;BUO*%o}S;fQ_2FkaRX4m
zAai+?z!FUnJT8=gqdHYlM>}l^arFP}5ezf%?FC^cxIg*%4yc~uNuc#Lh1qB)&vxH%
zrBc}WQ%MvE4UbG<a_+oCr#q=BzD4JALY%E**aYPt?MQ#YNb!$?s;!IwA%|yp+jeH-
z5Wmx{kb?|L^3ZVn&+;|~Z2|bm;1_VnIS{f2oU<stua`d}u1e0)tMFrY94zR)2-bq|
zE%Kn#?Nv)gW|(l5{*>fH64!vcK|50-8j{Hnhz_{{<Sgj|wU*?t;TuQDyTSo{)U2Dz
z_V~|1>iY!(!IYOOsKD=}^eT`9_9SubejAUtTkMFlOd3+<?8?#i^hO|?`9l<Z8Q+9J
zw(`^`@eJ)h0L5NOwAB7r?s5xSde$LL7LAHIFd64Jj0UIB5x@=dVCsljJ;WD!5gj^7
zY8b-s8^EI>{q>hmhnG5-2?U$NG`1&~*Ybc>6al24P=sH)tyH@@ws1z<^@VBZbX%JC
zlTS~bo>_xZ7Lm-=Be2}EO!BQ~`?rNV?T64x?6Hr8B10set>33~%>^;%guX=_BF(-?
zb>%--a^=-0K*4+D-C=|CC3MD}Q;G~nd;}3ooo?M9dH;Flnct(nff%0~5Dyr1=b)(U
zGXRKoy>(%qmi1uWD6lPdEG7l}9OzK2;8hSEwxq?tHi@~Y;GQ}RjhF4)w!az*e<d(G
z;`Qo?LpYfF;Lc=9B93c4B1<wQ{+|=9EC4L>!!e5!CRDQscAp7iPh@1saZUHM!QDGz
zg0C)$=#g`He_44y;DFXL9Y)e?2;;8-j{F#p5k#Z3Ey3X=imyvD3x>M$eEiQ{UH<M>
z9?Uzt^kO>zb-=5?8Xl~AsJE@)VXX`t0r*u25cI#;gtkGeh$Y{DLZ<-LAmRG4e<3Qk
zO4ZX6<+pa6DAx{P1=RZJ0UF#Eb_tT`7lCjUh^F@!k`2i|VmAY8A0)}!zUeeUUG5oU
zFtTw-M~2GhmxLFw5=SDd&XzmE#VnBR{_ynMto)mX7Tm42JOh7yNp9lJaeO6H5!Kb+
zeLiLYV1n87|K7WojdSYG8*yig;yru9v=as5Xw>omu0w2Af9<wov=v|0q{vPK_MR_Y
zC4c;5PlnuTIsBPHuh#w@-}RQZ;ES1M2zwLN*+Z{P{^u6{E8$FdsTgivuijX0Y*Ut;
zRg8C{=X*fh^u`al@kZz-FJ3oX>&kasF>xSU8z_Ytw$!HuZBS(tvVZpZI{j?lfd*{A
zA-OV@7^UUWejej<o5L{<A{;@r(6IqK9l*MgGUEbk&;Z}aM3hKmh_=p04^xrFaHGw!
zr`D6a;!a0?RuXLMtD<bom>oWtEJuFN+BB_7_WI!CsafkbW!xM<o%HPpCW(_VhTYMf
ze5vuMB7PK$wOE<24V-i>6J?!9Ieo%Fc62?Y!-H4Jxwj>Dfg{s4&hE%?%yMhRhlTJt
z?ESIkZ_DFD@AQuLda9z%<({V2S=zBS`{gvf3^PdN>t5x&K3P^+7GC<`T0bY2x<^fe
zF$XDwly2tLJjlde2=pv$bVsp1%8`^<3JWI8nl!G`bx6EoAqt}%zLid#C>@l|NV~tD
zYT_OHID4hAVMT;&`*U)C)(4l~F8`-P|Fg^fRR_wkq@E#)C2Q%7vBkumc{jaSHk)&2
zU(ka7j}yV7vu<$|$7w4F??&4yR`pLjuxn&|N%%H-KTBY0ac|ga1<Dq;d;sglC&EU2
z$6-|1=A4@K@v8~<s@32RM%ynUY?Jd=Yt-d3>S{HEHczph^H)r1nV`V~F73a+eBXX@
zkXTZ<&ul{b@XO?M$+&7WrNArBRcfgR<ns)eZ;hanJ!3wN(6*F*!UxXgN%gEQGcW0P
z#SVzcR5_S*wZ~AB^{$O&P2M()A)1-iVVL&yh^M(fWp(Va8vSc(cU@t9%wxZ0(Vr&@
z5984PUF~oJ%Cbg@<x$GrqYtFkH2w^TFnjW1v&}Dt;V+-md(3xeIekvfRsF?xcM63+
zn(AqAD^z|&Vz^E%fLYMy;mL)Oxbb0y!hKlrfexL)u<q8ur~Xm4hH_%nmg>!oClx1L
zA4z%+I}LLZl?i!eOWY5a%Cm-67FSel*KI@xG1bR|iLK&K68qP1S(}mMl|PymEC%A$
zM;*UdN<^pzg?0`vIUKcDOt{=6EHgIv$G;mcFq?t2Pm=~C*B3R*;LsTVxPe6`+QQkU
zztVAX&AFI=!fw|QcfQ3^?fw<7F3W-VhKSKVBSQv(YQ<GTc+1y!AEsKZBo^KqXLUx-
zrP*Sd`TK?`T3+8dyQUvy+tsS?pw2E@>mQ`!a_t)!kVb^vHgq$V$s6sLN%XK4#0T`<
z{>a&YlZiEtePHn@M2_8Gi(9?v^P|{`@vm#$)|d*SoW)umIOQ$Y3-(N#W-5uzZtC`1
zEOl+3aOkZGiyWo&G;j1?`o0v8ohYhO+-QmRSUO}n+M0vzc$XLAyqtH-W20Pf)X*`k
z%c#ddk^JXymi1r*UO<MiquFt4F3FYCY3}>Xjq|i5objkxaBmR<sb)5+HAl?cd0<)A
zp{lqrVTo*AWmi1vZvJU`q{2+!7(GBvU0CZ5uBc;<!=P+Rr!VAH_sfJ&+Kd}LN#3*Z
zN;$q^F14emDsnPPWN{)Y##!`>z6aTOB=MC)?Q#r8V%GKUOW~Dgd|>vFWg1{!c1TdA
zOh~1BnUlDzJW(vIvhlIk!&ai@N49!{d|J8<b77V4WEA%B<@Vl%pJ(`Zopyci@2IbS
zg8uSqJ$~Uip_+1AIJzSlm**|crbWAveVioJFE~;hoMKa|vpVma<w3yDkoGEZz9%Wl
z_u_k#BKl@(gQMTFJs@2&#@p(0rPHI^zUE&h^0#oa%^za8WQ!lLQ^_+L)SMDA8TcbW
zsPS-g8S&Az2wkpB6|psEdEI4bUtTgRwC=p`npgYQG7@uvr)`<E-0M4|Y;f!noeE=o
zLa|&+*|eZge|z6l-mGW6{W$hKC9_?Uxl?tE>Bkp^_<=&~y>HENsJQ#eiZPUjhKph5
zQ1|qVjnI3W<k6eLEu7YvD1nad_dT-}+P5w^DIO}Fbt>gVnH%d3^=&M^dn7<q9!pN-
z@^VzwFe@Ie?_XoKBpo8Xsg#N+_heb@x9Yj;vbF*@#zL>vsp)lfoQvHlxAL!-XH6Zx
z(=D5vj*~Epb#s_7GJjY+_LfqNWmzqG9X)-Wo;yon-Dy2_lwdZvs$DK0^RY#pp8L?W
za~-F4M>z@;&?oLMvgGa-jX|wftAvU{pBCmf{60KskCCe1mD|95tn-uIxAb$LZ~Ye^
z-|3ExQqo9VQlmm%y>yh7F|)%);&RfbveGAp8J}gC++q#N6&4=XJB({t`IK^uh6e|B
zJ5S3*3Uu_JVat1=G-{Tl(-gj8CqLFwsmo4B=6LAt+N#enV^L7P(LQ6kC;(nc%2yWm
z^NqDJi><aSq&Vqjtmsy>35rIoF*6$JMOQ0x<QY%Jxy<#MCWh2W5ih<t#%Je47xS8s
zmG?CX{-!0wd04bT-^GF6;Zu`Yl<2JIhtfdWHpb?jXb<JIdC|QO<9PbaFE3Y$Dm^A0
zQp)Q)%-dLgrKeoHX8ov&3h}ER{Dq%dOt?F?rAd7wcT`@o#62D=Y=um0y`MbufLE)-
zT`F>c$Wo%@W`%=o4n4Eh7^UaTzOoWGRS%nTI5aTpt*wK4Nnps#t7TQ1YvB0}w<!>C
zMY!tVNpjFh?kPQoDY>vn&^3JEmo>1msu*N~5d=cHSryB^%st0flS%dbqEq<+*s0nJ
zO`iD+mLrE2Kg-6rV7WB3^49ugx)&$Y3*{!TQ=u3MkKD<p{(84<iZY@MG8ft7JYE<i
z8*}Bm``!)cxAYQ+TyEQef_=~`UzVTv4rXhY4_n|Wl1rb6vmFfTSE?Z${n*ewX?3?F
zHUD}Yx$(kxa^oC**P;xgOF5l$O%Hz@fx`@a|LI8)=WA7Fw-QCP2YQ>*18o>>Iy}6r
zceM-Zy3}rF=5hXvvcku>l8RlXJjjh1gC#2ylvCcHFsH0c--(H~)MokT=#Pbod)6uR
zM1CLS6J1kYeIH(z>CwqgT7GWon-Ljlh;n~vCv&}LLh@b-V{Nr>#n^Mo`kz9_yGBol
ze_w8u`uf3nm1*s?bu8nGIZPYKA#Eh6XW_H%OgjnThlMcdp`$I56&)Vc;U1608!w-e
z3J43g4O1FJnOkb<<Mc<IR0%HM%=)Ds&Rib5Vx7~b>|AE=F(JnM3KKKaJ7zRolgu8`
zAZTmUG4PVsd0MYYs(&$~BuZ1upw5uHH;kk>X19UCmbtiy5e7VLCWNf+c8l|VDY=<8
z5x&XFhk@xXD==f`fBB(z!XrX*-8uc}QdG#w^EsF2npFum9O^P$@RMJyT+>`ltW3#8
z7R5q6MwBHR{`f{rL2rw=j;?*)*LP9Db<VmI`uTC+vryc6iFHm~6*H=O<!5nI5rLCR
z%_dhn3TKR@D~j`5QhdFVMi*w2y*%AY`lPh;^Nik(&J_}D@lr=iV&tWGnc_%>bwWzM
zlNm0{f`dt+g_4t%k}k|lQ{2sXL7{s^Z_Ed*OR>s572j~zF36#5y|pYW@P%{J=xSUy
znX`Fxw%2!-_oLxv+ss+eN;n(Ns<sD4wR!-ZK5a-kQDV5deZZ~H;(LeXDZ%eTc`ccg
zor_Vn#qs0qZb@|5%?<SFjWU-FuJZX^3lmYk(;iR!R%@p6>o%NM$47euNmuR~CKOjw
zGm+y9`#hCPJD1W3_M6TL6FbxBz}Ez{)5oJDIhn={A7cy;)tq0OjW;mlE9_rL@7T|L
zdf2pftt#46P7h3CHwE<)hO!!c^R9dd?(FYQio=#G3H5xv!uKs`zWJa}YL@rLC-qGJ
z$I^`py5?k6^RGPYhPTT5QnizA+>f>1TfRAmGtF=k%;#)lbW^OQQKn}XD4Di<hF!Ny
z*1uz@?33WHBEg|7yKHF}=(-V-#qlhMNrGpQA>Xiv(R1F!^jpg2g1<WFN;P>|tz?D}
z|1~!-%}Ink%#fVD>Jm@)LS7wH5sv8_$Cl)t2s39siR|Vs>}hQ8>d5CK5}gM^>wOZt
zdk!awnCjWL<QC{;1|BDY?E;&?3vI(TXR3B@pXHTSL5SgcZqsgl*mM$#s1&w1Hm934
zcypZawWJ`sCnaBw%ue)hnO(-a>~dZ(C<`r59`~S}S)A;RaU=;imxwKiHxpgQ)z?JM
zms~)YZ~vAw-mK?|cy%jzFS673Yu_F1$3#BIdX9_X>AJA(tcNDT-+hqQ>soXXCk*(X
zIg7hlQ)jrgQOza6$A;TzcdH<v9}JUCS5~~~CR&^>(N_3!dA@f}>NRfsICYON@Y|%#
zjFqTPd)C1FsoWkG)6HPV*?}K7@&+EV-%a?;k{rG$)LSXuC{D>_7OO&zmJMf%R31~g
zvS`ouvn@C!ITqbGE<qWnQIMS$iY8M)IQbaM7ZVnyHea@gFX_hiK5(N}i8M<x?78+d
z&L)i6e&wX;RSYYGXP1w!VO{EGNfW+kHuBxin}^Ex^bAOPnoLyI9n#bB3p};2OuDDb
zKFXlJh1}mzdP3f3*-s~RnX|xFOv`EG-OVQPzG6EUQ++w$j0$<=(d$w>pBP>|)tOln
zJM#Bq73DJO>O}j}<=82=_j4ofA8%UXyD?#{lZISQES;^6lYD*UD3PjfF$dKiI5HjC
z5g%Z&osClS4`MjuH(35qM^~^2yAPkrDrM`GMA^Nhu{d!fBkXh<oq<|++Ge$B{ENEb
zOoOMdPF8Z^Tm+_s`nx&v?@ITCb=-A<r$bkOZA333w$bypX_D81pM#Cy3x<q)%$r%B
z-eFY5U-2^o)@r+ZN;YMd9<<FlJ|!;1#goW(3JTW^<FrQP%5FIS3}=Q}IkTpPWh_pY
zCbHACR(fW!{ps(zyG<d<6jYH|wpsnGs+-_!p=FqBUuphsEjwRR;)FrAzTNa$Q*Szk
zs24&5{f;X8@LYd|ng>U<I%u{_X1ybixSckQ#hPoa?tPf$+=n7=3U9dU2_|CSEsCg2
zyiZep-(6KeupAH?X#1F6ptU-ikUdt(cB{Fd@;OJ3axtcFM)*taXYAo4qzBs~D{H>D
z_?gEWvMFr7Fq2W}a2fh5n^DHBN7alA<PUsev7Q6+--j=#ww>o{SiaI+$~`oB@B*=r
zfOdWSBv-#wV+>1AN87)EDrr1yR_r$M=KCrOI`U-P35S}HX)80$wu*t}F^;5;1I&@h
zYvs1FGt05oPMP5@YSoJmGd^@Q%DEoEt}a*xN8ZUomATRQ-B=9Hv&rsZvLA}Tf5zBP
z6fH@N?J1dRZ52_99Ae6vZabni>3?}dKg+vkx-qtdXjvdIH`-?fq)^*BhuD)pb2t}j
zXh@Sx)N~}VWwCDJsTCdX<(mV`*j4q}nQJ%41pLhUf(euE?^IpatNG6^_7{nAYzoip
zc4L>Up2auOgv6^=x1_rk6ABdz36o{-O8R{9R}^O5(t6?q%P8HAV&x5{u8@11mlI{s
z%dw@_17S1R0sK*n>#T&%$bZW0TL;S6A8&-ZqV+9p@exLSf`Q^4uEZrblJ2-=bH4k%
z>-$o{)<1}S+~#leSfiC|9^I|Y(T;+bX{^H%;|!#ruSM<8^FGaGuB}D2b?7%9rtd25
z2qrLKmLnfIbp<kd)+l9ri=PO2bft-WD6juXRPqfcsg}fL$A#W+D66NcVunYplw=FP
zTDr+C=!_27B)WF=Q7hLt74uuhR7A_DGCw5iM&Db`Ti&SdymMajZE%|RM*4!S!gNV&
z#vN^}N4B?Nbq{WBPhO4bu<@XoC2jzlbyf_wEFm86y;0<Q1?MvPSTgH4>6PuJ{5aH$
z);t|jX>WIuiZ_<hxDo#Us~=LMbhrR&m__o7+@dCXF`uxUe%B~fLqmGLk^`OWRqmc}
z!nU*aJ(T0vDW9-TYyRuzYcqolkPKTha1DHMl;2l~rlNyMDe4IpJ&$`*E=hAvWG1do
zXfdiAm2~d6t)2<0ht7z3*yPOLr^Kp*Ry2s@F*#3N7V+6Y!I6hcAE$FBM?ay6c<+98
z;<a(-=J6`Siq)@XQwQ-J8Tpas<)gTyf!B#2#V1xZ&p9r{h&5QGWmz-8ETpq4<+_IU
z@#K}Iv<}hxt@>H~Bl(*KwlY3VpXI=H$F*GJy4{A`-aMrVJBDfY6();YvTbC`^NL~{
zBt4O+obiQ0deVp+lTECWi}a#|p=1TDb5}Lfv?Z*YQ!Ky42d;)X(NiUXr$TfVPpv0*
ziHrG-3a2i-#a-<)`|Noqaly&5M|>pOcZRoc0B95ZBpDA+Ivz~c@4#1nIXEOFv4YlQ
z<YS}XI8RZ%e%IIV5wlZygXWvzdA2CN={OG5sg(uP=28=gY*wWidD(~^$8Ork%MKZy
zh%9^3FBRW$g*0?RPok_VRH^;~#C3k<eOdC#hD^eRk{tHMgs=kz&F`$yS#(`aOd;X@
z!L);#>PG3-OjDO@J*pOajgvO4mwJ=k=l}9?jSJ(|dt1<u5PYlfUHnra;kM7(=7K%X
zKWhEyEq*DGXlaAtU37Dma%knD56{3~sQ6N}5-HR0#izn7ntB>-BG_Z4ZRq0JD;E|T
z_eb2Z9&D1~K%%zTl#++jWEpnagDAr96H@XmM{;w7wnVvLO{jk<SjVZLK(TLlA_7lc
zrs46mIXAuIb<UjLX^$iwt^&gV7f#|vgpu_Da;!{}mj}Khppp%WV09pI#gC@ID%4b|
ztu^&M3{t7+_OFar{uJPJaui$D>mB%M^VL`5GMy&@VykL!U_OsKqPF;ZXJhYZBz><~
z*ERiG+80hLN%pw^%lmrN-B^6R^pX?H&GJUii+v6aj{q=L)5KA86gQbQ3<PpMBpgh!
z`67A9IMq|Ha5^u4vA3|S!rM^CE#^go7&SG&asPTtW>{~5`EsC+<@Ee_-f2Q;<OM<U
z$&S6Bj`meWEef4y{vf?Lq$05yYrt;Tmcq>t=<xVOt=?S6z`X+t8Ux=xsdagacgo!w
zl^ecbW2$SC7}CMd80XQ_|M`A^Lk|7fa$lc8wPw<fXS34|W2*wRgaRQKyvcRl*{p(&
zoRC677;_w4P@<Sch|5ul>C(xEBR(@n)3YbZ5A0T4M#hYFBSF#LHF?lIMr3}%=)z*}
zNj4#2-LbrVWk2L?c%_RYvI*sa(`mEsyG+*@m$E1EUFF5D2`wjP`MA$Hoh~DJ-jjLX
zgi#_nHJ*KZQGe>CiSJhdNoHo4;`o?A@5{?~x0L4;qYl?}oxsArQeRJ<iOJz*{v@n~
z@o|)ohozXLO77<Z`!zj(j^JKmB*Vq{jXmBQ>k6}+wl6}${}i{qNU6$EB%RQNKV?~&
zw2QeXwCDCoV00DA)eJlzuTCi)85?ZdU7i=B%-I8BY-*3<hsMI8^>^(t*5<x=_3Ew|
zXUjYOmzS>OnUW;0>R?|Sn!5koI5~tmITalEvT<y5Zchiv!Dkp(C3;vi0~`jcdP0X=
z#rg0mSP9I0+{n~|tpJCh;b)r)wZXRfa?04-heao>@@6nm58})C2PW^gSiYATHW5y(
z=At7#XA=@R>=Y?BJ*9;oOl{}TOZwUYkgz43Vqqe-i}f!GorFYZO5<RJA7@{!xA@{p
zJp(D2P!WBBh*23(t5JGJC~=5iIF^ock<2>4BKr8};Kzn1<2Xg9a>v3SUySu4l?ppQ
zh<YBHK-t<@^)<w@FYD}Iayg^v`)?hKI*jt?Nt!DczthN0yUyt6`Z_s5ye$41Jh)b`
z&g=y<X$HYsrJEbm72~1iKKibQurq<Dq}C?7Hvh=`UZvM#{g|(s6x4KKsaH>NG*%r@
zmbjO+4SzaEG^ekA7<sG3IF<cs1G|C_yZdew4u7m(^t}n8d*R?1&i&S#^u@SQ0P-N^
z2_A)6+OYu`rWo6r-1yM<#$vh5lg<e1#mh?3cdL7;U9nu4*lLjg0a5<G>QG!T^FT%Y
zevb6JOTynyaP&W}zQo8P(<ID%Whsmhcq*eX>+E8w(7>Xf4hs9UQtG4E+mG3?g{~8l
zvUTe*Mw<35%|}UR730Ffl=qCN=^QBY7q~u#(z!^qIb|(zyVGUlQ%h$4N40l@4qe7^
z@rrB7RoW*qkKs_tj>`W8-K&xeEuMzA!KOD%0TW};@+LNP>Dw}0pl+eQ#VPXKLfChi
zj0Pqo-14SK%x5W+IM~n%^*{GjMDeI^=1j7x7AFu3Qaz(y9i{K-x`tNT7tY$Au@IId
z`-49{LHGG$gnP=zXNk0t{r#U&#N=4l$F3&zjY|t_&0mC_N3@mvhH7Qk1Joa_<;pQu
z{|wd;=lgbj;Xy}-(sH?N#I>xAIL^iUJ)OpLjI$|nBGkT9m(nfDTzXa$)fy9NnU<qx
z^vCQ(#0z5-I_KWW_Lf}XYjB(t`*@I1NW$@yRHJ}%w?yh-%fcl*<IRu$Dm^b^*TY<G
z2%|~Y1B4nJQV*aQ<J@8bZPrNr^;amfl9=L`?T1b&sAaJ%y!m{N+<C<IQts$Is|kld
zyBRfC&c%}_gzbF#r0GwQHVRtpmq}&Q>P@Q?qCfds%6s;#ktqGv#TUEQIIa)L*=-p=
zBkhwAb<<k3+uj=df!g{DLW!0mY_k{S3TbD(`Zk@;80vggu9|XA;*n?wh^#{Gg;hRx
zK2ob_cily$*Plhg=NZWC(<?tOa=r>K;SIdaO3X~SAt}MiNxakZ<ntz`Qr9WwlE@Ly
zJ|R0blYO;mU~N{Gd-o_-O#7x4*+{=_Ph*YR2QChDSoP$|H0$P`;z6Cz!ToMix#i0L
z_7hOSt*$R!lKb4cDNRg=$?6iUfexOQ?wMMyC=-Q>{{C=}v$geco*}bBFoK>{FZ#=I
zF`XMX<HQR7Ogow&9w0O@=jS@zFqM%FQ>a@uT;kb@%BC@Yk1UErC;S=ul=FZ>%KPM_
z=c0@8C`@j&Z4ABJX_9tzVB_4~t}uo2(Ox}t&BmiihnkXpjtrTfzQ25|0hL0X?l3}$
zEDDVbxFvhY?%o~ObfMj2I%qi?+|r!_81nAfVXDpFG@Awxg49@}yeQ+lCzZv{MNu*_
zo7z;&y5`j}J&hFEd~1yFccyPyOH|=#H7${)62PfJ#*C5ow*F&&DvUTmW62RN&ePy@
zMbifz%Rx6@YiRrI_EW=`V`yu=h?n%jGy1FOWYKaHu>jWp&6YYSi9Vs6-B+~034;M{
zQ?rk-G{?pdojWUf(&7E*x7EuW?v7uDB~>hVBTK%w3pQe|qb_n}<VL!t45rL~eIv}u
zsdoU|iT8S0`A>swHOKKs%KXC3jXsH_u9KOj<!|br9yW;JOts!D>!58(?V$h)2lE2-
zq@~XxR3@;-;Z4iHN0P6sc*pj%Cy2bs`4;Dn3G8xS>w{3jFSv%k{NYNsx!-iXjt)mf
zmAS-bu3P=Cr_o020Srr|N2Rj|X(a6_w@iQ`k4<}WbUi~w(Wg|U_dwtS^weM^`OZYE
zUfOzTN9v0B4YqtjP~>xIExUN!F<p@(mvA`Zqn9R#L+sW$hdO1JuN-hcNNMN*&3aXX
z1i#y?XSDFQ_m3MkC~xq&&E%AvBu@0sGxWSf&v?OH8PlBCqSxE>E0kod|2Nw7F+tlq
zarDQBev_mo!&K%#ZaY*$L3t`qwZ5L|lM2<118NoL**LuqK9r%VLA5p2guLvg`P~a^
zJnD(oWDcAOSyCXsyzJWOaFDcLF2Ek(q*V0Fhv;zC!<>t(%66P<!;&g>Vk~tGzu$=R
z`st9}=vsBDP;Y|L#EgARLG;Urrg|KvcY=~(VLh!_W-_IQuo`lMrA}>m<`6D?Bes#B
zrS6|5T;t<Q84Lc@5-R&vuca~v2bY@F(?;1k^N*B+h%th8<3L`Lv4ZFBIw~Jc{+ZHE
zqrcl-C7;v&^X-SHF&3RQt<+%0FMi`^n7n61MvMf>!q{?@?esG~P{S1c@@R(J6BaYi
za;dH!>^LsO2?!cs)Xgoi79}6KP<)Tkjq7`B?qfg+^5dTsL2>a(2%3(nztH*0zMwo+
zh9R|9q%qxA>`C@JDJD!-;2{(f1qM3cJ+jCCiBWaR*`GE7Bwo0h-is(skKrkaGgf>&
z;t5kM)}QLkz8t8>g@r`a^^lGK^iL~&5_aS+6^qvm;ot25G)DRI90U>luUGnAU!D{l
z&LW_n-%%ZJcu;GLypaF*3oE8Hc&Yq_$3G2DySh)+JRB*I_v-gQf7VAOpz;F&`5kZm
zy+LA#Kus(_dF+)VXs+a_em`(wROmdi8jg9YfUq_>*q0*cXLB7?8_K^?K~nQkD;)J&
zqAYy%C(hqrjRx-q<YWFf^YbwwjA=`CK>N=cBqA1GG=ZP~$&1IgC<ClK|G)hR#b~ec
z^z`~9!N&+HT@H+5xC4RVm-mqou^{kCj1i`YLzhWd1WW>n0-t&DGCWK=+x}u0R*0AZ
zMp|Y@d=~^RKQ{;olrg^DyO6dksrfM?b(AxLp_fP5mHEM|J#Ca9VTpEuLyaN~zN(KF
zc4*dXJG>Q2e&3R#l!z2BSqJe&o9hD-Ou;*UCWI7%{e4U9Xh=_)FkkqsItl67mHS7$
z=?!I388KVk43yXw$A|UY;a@XFjC+7g+5nW#W(7;I)rTFWK)uCo_R%?xn}un?!l0X0
z0*kQ*pyA!_F0{49_c`mq#Cf+{`w)F#i(cqiwCbo9S*Z`GQ}xL7>BC~gn|cQ%+?J+@
zS_)0)1<hhzlg!)GBLE83)^9p(|Al1)MoaL5Mb^_*@a1mNYiHW=1M@CPdVcq&@2*C?
zhDSg~kzF}mdYz3IT=^AjBQUWY=q;O;cFMdi)2Ij#XqmNE#{t<D*K-jeW=8Ovt6}Q8
zw>sPZEb+yoBz?x4F@&^_6BpCcjSvMzA{Iuz=5x-1W+AiB1{qsCY}f-6{9tZnlVSw#
zXSe|k;79#oYkYkS7u7obF3@EO0B0~)7-ln1kq}yJzD-~E!4ur?L;Jy69#v*Cy=%vv
z_<Rqi%{m2SI(*$GX%Yo$6v;_OU~}>yy5l}+6T!36e%U=Ms>KbU&b2BOo+%ovhk1a_
zmbJJi9xxDo+wHgzN|1nC0Y+-~zSO|HVfInMAK<s0$GP5}30f`C1fL<obS&sU{@%=r
zJYHK<$r0f&#{3*O)8+L^XDnpymGDCb27@BC+gejAUdw}0vI2(OLo)VzlOMZ223inH
zj-fmGR$a|D@gi;^En_^oBaM6M^>5eNYq(`u7gzxua9r#R5oRYqJ@@<JRHwtaFq*^=
z5NVz%jJI}<hyz#hw&?qn`(k~EqJT`+y=4anM4_E;%+naFlL@?#;JtxZodzIkntDCk
zL}%LDVBUmS=i_^<fG{`$&<{7Y_{!^Iki(W;0sYm<fkNw^c4RKj2tqra-HTh^xnXO)
z#N+O8K-GFv_yxY{t~$fp+>8HRs`}hkf_A$k%BF(sT+)D8pH&A_|KK_S7JPc5gXyM$
zSnwA?c=dO}4(!7M@12HpY{%W%w;v1Sw-(RUTAlXeBpBHfiRgH3K<P0W0m}P?j=OJO
zSwppFS!fAD1dO?GJ^|9%?VGl)S#s9P18BJdz-3lMWqd&#s#m{&(j9&0iFR8(P#{N~
z$pD)eBQiE(q&*D@tO1cs@P9ixi5G`%KaQCLgX4(Dhef_P(AJb#Od<mzIDTGCZkk6(
zlT2cm1m=QcUjKQvPR7xaeE7h^ubdZ04`x1v89*{BdNX=%<KD5;ZJnTB9T=Cnv=@io
zzddg0wu(p?RsPkIQ+*@vqq191g+pJ7M94-Ry?@2$4TMk$T4e#moYeTMsC~f%u<$xe
zVUV@+19>68#B+)B65?Oz(e8J|Mu4QpEkP!#W6Y=bY+n<Wqoxdew?9#jOdXizkFvp+
z0mbpo6I9CHGVty~rrF^&#psJaF~XR}f5WMd+qqJcda;0mg3wMi*sDTShsgDkloZ5h
z{@U?6OT)N3_2D&};p!VMkIf0UUE`Qj_$*wXwpyvmw81pCA2dg{Z{|+|>f+oZwWnga
zIWkHydnLZG1OF2TcCmL!sJ~iSw?7U35i9;Y25ybqJvJ3?dxh?Qlo|;FQ<bi*@f#J-
zZ>n#<Z$(n^oqkGWx6URrPNzpA9ug^~LsfAQzt5Akk67&Awl8T572SKnlSyUPK*?y#
zr@s8Kk!!wKcUnbsntuTqFTTkuskHt4`4L((d(vU~4^+fr`#|N!p#!xTlL5+>Qdb8)
z@LnaNrD5*sug%L&%$RkrB)Y8{EcCu#R;swhoG@Qbj--*K^3nbC1*tjOwcRPHY5sd&
z-tkDp;o~3!|9NuQ5sTfwE^ic6hjL^|BK4valF>*2Ata0><WlFxG7dUQ66$R0o?q+J
z2)4NOa2(x3OkMNflU8$;zPUl>!O8%!+`1l!x}kxj`$FgWBc+yr2&uSf;suV2$Uae;
z54iKgl7iydMPR<Ccfqs_$-dVFe(UbmQn0=r7dxe{RfT>J^Y@S!3mjb`f6n4VUOn;Y
z9hhRTMqeEk(&YHC#WV&bLW|pz|9$S_-$j2=T<t{$h|tY(!#5+Nz7end=6_#u#};K8
zE;f3Mg5u5AggkJ`oxCaWMF{)@vS53gVgmkmPqPj?&Rq7SprBC!>qJDSM>9bcBIVzd
zg{~mRHowrjd3Nj8Q&Ld;t-lUu9}?IOJMHnTJQwy3%-@i~5uyD@|MxWQqM&%qi}>?)
z<E<@Wkj~;wZ7S*bL1f0#I)sIBu}j`sq{yvjRE&axVyJGYCa3|}@W^zPQxgX-HVA_V
zrvXlyNRK_BB1Hl`yo^RVGt@DpNp=Qt^Kk&{i%1Z_ARf0JTRH?O;X7-uy~|C1F<IdC
ztA9VM23L(=Z>FfBMAjGiy1Vg&z>~Uj@dvgTWblXc$3BU9{Qh|qvf^8J9DV~uAocH;
z{5`Z4nwj8_XIB5e_|n(-n_60-8*i6>wRQjBuJg6vnbC1@B@rI<<Ez&(i!m_-Gx<DZ
z-c|R%zcBDCqFXwA<>%E&0c2L}=}~0ZY@TU@ylsfp*1r3Lg5ql6Dg@UC^W>Q06PIfd
zzjbj?d=uMiwhjkmomnHA;$@=1$dISiPT%5j>sch`BM%U33k(9x152eRBK{JAn2-@x
za4h+Cn_k{U)RoSl-MLx4Z_7m?LEPN9{hzlcIDl~YsfhBeb+tIwe=eoDNmlUbX?xj}
zQ7<XAS&NCFW>BsogbcU}i$6AS91A|^i6DyKH#3D(Luw*OP5s~39Q8<DR^bidluHE-
zEkcBXHx?Y)hXHd0$ujm!axXDn;w)Z$7i(q4m|cgChz_+(W>qLd#Gs*FQ;g)VK`d!t
z;>x8~Y^8nYou?+U*k1YjaCt!iNS+L3pJ&(H6c`%D;ZuI~-9Z>l#Zn-!+Da%XvW9$8
z1|j(8%W+C^E~(lSz%|tO7CRh4DjToX?qh!~a$T%&g~O~xkd7KYb^)Y#$1V{ySHEq2
z`VF4q(sLc}ikyO_#*X^g9Lj*y8OPkJe{BM&(KKpN4v3*PB#T_u8qi)*xeW6QRmUg=
z;?68VT|kv;EexL8#%y5VqjDLX@P$C^vY-5ZWIr!R1;I@w6miim1-mqJ-{;UE<Gc;p
zBzz*1928aHc4Av}Wg;cMUWOSPj8BH&*6|k%Ahv}>;B=2pDDop&ec+Cv>tS5uJv><I
z?Qf3dv`dSi21ij?<ExkE0uIfVw;x=IWWA`^`_Uex&eggjU^5Q?Na1tl{6a{`DfVl>
z|B!=>kK+Sf(IGk)w)4w!uUx@G5c!yN)jD^s2<mfkyeP0Kg#SHb%rpf#_W(cGB2v>W
zXR|87XP^~+x-cO7T6Q-wc;gY%xtTk1Z|Xr)K|#BZKIdl8&(%N9bIzDe!I2~&I+cAr
zrIa#A0};|tT&26(x&NudpA?@*z#>2>$|_eb<^39>!eiC?b#-kZj-&W&3I>g);Hh~i
z<N<xnI%w77JZ^yD@2)d;XZ!`IXLnI(P(r>JfcUgaN*M}+RpC(-EIcK}-*5i@)Di3O
z-tCa;4-f=#gVf*t$rnT-tq&?yQ_z!*X9R+NmcoZE%JIzEn+N&dRUwW^TjQQ-DN>Zz
zE9&5rRYZ?tk@5?YSj!}pI2*lL%L8!vtk3ryJr@qrI6^e+tr3Xdg~%*>SRnVH{x|oL
z(Nr=r_t&m~xq#(VfK)3=8gWyejYeb<>!1>cnYr;WF$1Z7avQsF*C`i3*PPsOjMC*R
z;;amAm6mys)<zu??0nyvrrn6&lbPQvZwfA^{7`SDkAp>W$a=yDDekK<f^63rRoTCw
z5?%Du%aVs8)5_OElj;<$HK;(^x_Q??WF3Y0oE~Jl(s>C+1DEk*wg&$QH_vUrg^vp;
zT)>}x>w!ls&4rzKRHt5Dk`7?9M#c@y3SX>4@=VOq6f%P}y#IP>Q}*9NeU3q^Fa@_+
zf#M7xofI$6!(1ZBHLXX!#zggB^kn7R8kR!VOeRbgBR2ADrHxk>{K^XS8DSRA!V!{@
zx;(A|9ER(m0Q*sP&J&cI#*jA%uVt<ES8O0|1)IyBn_>Ox>$=67ZthJ|^j*3`V2x7k
z4nq5J0SX#Wx>3j;>jF;#P1ytSL0O<r>Rn$O*v!nXaQ7;|wYv~hd~S_BV3xZ8l|Vwr
zjR3X%Sp0&{eq4vJI5Ivkmmv&9j08gsEO>z7_ZpgLPrOX-Hn5%*=TM#jshTE{_VvUM
z#OV$(dc`tc<dcMXiiq#0PbG?Ip`#C4K`V~8<yq-d=g#|)JV3Ho5a6_S|ANT@?9a>V
zl5uI?<+tc*uI=xdMGTX~A*Dt9_&eLj^+!`~RqO4?RY0mTZB`xQ(#LX=$LL0T$DpBa
z)hDf*7jDt>phwmyR(z{5YtGf{AV1dE=L5fo8m7D==Pm2=*MvAIqD;Zp6hDx*cv-JY
zDg5%Glf0kOVSv?A4w8u=^PHzU<iiPJ^8Zh3XC4pb{`T?dq*IQlERDn=kvJieYJ}`t
z$u4_B9AhaAHMX*6jn2>}jUi)?8Zx$02!)ZMtVJ@02xB)q*X?(Huh;YE^T#tUfB2)u
z%=f<U@AtYs*LA%=b#nr%Wb=Y~b1!1=S{-5yG9ChGzY_pB&@U|(Cj9LU*~HK&+~b}I
z^_$gqJ|1VBj#GC|S*q+reAD5Gxuw1DSXC+taXwj`Y;E<4F!KlW-Xk#A>T6a5R$ibh
zU*#~fi0HH*erSKCd!;|T!d3g7o;iasrxQih)!-s`-{~H17={GcbFF51wmvqxE(@$T
zVTJ>4Uk>m>u~haQ8d$4?Sl*`1XT5>Mdk4vz)O02JJt>2rOw_W!A7=H%%84Np^eS@Y
zF?AMFZ-J`(x&~Yo12gZ<KlRF&3+~D>s89~RcR_jf$r4QNRFFzMsy1hxKL9%#2@V5~
z>Nz~N5B%b`xNz}-Vs68=H0Saqi17MckjdNP1J({z6`#~ftEVyka7iD)4E$B$w+O@;
zQYLp%QrDwaIW?AJR3-_*mbl3b=6i`Y^5Y5g<ze>==T89qh}E}F!*-dZ`>10!2_LQU
zAFjie@F_heU#s})s~VX7{ebu?8$I}=R_q33#^+%}@LuVJDR1XBC6Bm^gdJ>vlM-jD
z(re+8BjoJBW$Peo!=MA*;aW4bYQc2cmjkG|om<z0AKx+vbCMHxTELa+L9fTD?Z59n
zjxDKr7MdtwTl{iEF%f#+=C)jo``z1+h+<4WE=Dz=BTF`|_qokpNjIQCEZADV%{I%1
z^h+V{m=U?x&R4a1i^Ly%`&x@+#*H9Zdlp9WcJ^K4GEYk2T8BWbdzEc$!XcKi51o?=
z)AK|A*15Q$*#<88&;DdCQfp`Mw^ghw*efC@y1F=D{bOu0N75b9kSti4kngE_GWMVI
z@0DUUVpN+%2?E>B!HnP5EwBMc+Hv53S;d@=;S9J72$)azv?*qVV>d}j%5#gU7Z~*N
z5-QtEQ9!uQgf>s6f0B1w*;Q-n5Q&!Z6i{b8jr7{Q{k)muutrDV?NKzo<vXyfrWd4t
z>B-XYj;Se~Mz3(m11le;WI{~~!vz~XiM=9(!NC5H!N&AAlc60uH-oKrQN=!KyRFnY
zQSDA#`}(>FeWhT*7w##B%iHXww%Q=^N1Z#>E8)I<`R|7wYKeF1_9G1LqS#W6lhP~=
zsF;&f+Yre|orR5qNpYlbobyw~WN^A0NM6lW-1YGY6}UC7`j4*?7ZMOVShpA+5l=fG
zc1JKg&VA`lgBsv1Yum(-_S9I)xOo&A4Hob>C~1#243$_9dtkM4{&Q!17}5GmYWT7-
zjvqH}6LfC>u}halNwXV+%D{$@XAAA011E(nRXysm-#8iJy9ux8p}VSf|78ct(WGX_
zu+uk~3Z~51<s+#M*vM{-s%Dh~hxZX2l%E}=m?6lhI_N!$F2_PBC9w_z4E9p*W}#~l
z6jc^q6yjEe3+;_&Qa2;aJI2-YsK;pzoP-rdNOc$rUr&(_onCnGVO$Q4vJm}uyXts&
zvQ2JJ`d}2H;Iza&1i3V5Ros;pZt%>h=%fqQgkgDcdggRL1>5@=%_HN2kPa%@h#MnT
zv8B>{iu=^SSI+w!)@YtDO-1#<NkgXNYuPb`^yTOwofFRra<oSBdgaUhIf?<Mc*({9
z<?=<CS4JGPhm;G@OHTiA(9X;`Ei;67A`2kwRo1wF$bXC)Y`WVAuyct^=SH6V{4CkE
zF&_w{;15C8Q>?8JHh9$!{YF%=4^3yrACG2QQ{E#0A~-FH=Ewfswp_s_GieE|^JZk1
zG6p%?B#qTpTzxeU;lu=FYRN^>j+whGg|Zi5Rd3{GiqX%|!nZx``E-14A6_J@{=_&4
zBOWR?KSqE-GnX@c-z7O;?4AbE$(ur#5veOHPdkX?!Oaz?k9eK@#zPQa(NUff7BFH&
z25Bjkr`pmA4NDY-aXmgDr!Lu3P3AJ{qKLCoY`05b_z=+K{tHH*7h^oQ`NGtxH?EJ4
zO^$22xoH6N8(OQkpUZxKgmndwEUfe(X1Daq1$)_6V=>75f>fw+H(<C^MgUJ&9>(=L
zEJmD@U9Pfy5^WpKmTll=d{xs?(=6?A4N@pWj29sR{~#A+cHl^cP*T&d{A-7r(-Gi;
z5}K5pmf$7J;>ko0%BEiP{Qk7{YOUo&>gK}a)-}L#2!@N^hTTxE_SrJot9A3H8K<Qi
zBh;6c9la!o3$e6G*2+CG{i2j-KuAF^s&)x8LnZGDj9-ZC-2QGS5?<=a5qtOEDtV{j
zc51_HUu?{Ze@Z<><TX~)AyH}vHHd!0SQg(YrDxe@S;uXbxQZQ}J}Y~E!}TL(6gRmk
zM<1AbzYKoNoTN*4?c-lX9<#25e_CxBZ+Gz-$L4$*ZkY$ZdhxZP3>^l$L-zL4a3ihJ
zyl&<?(Id{-8ePiO9O7n^WgzuM$j~hJGSA@bY|1~R-t2GcI5oe_Q7;IJX15`GbgB`9
zex}oDB)g0rN#6Yq_s*hOTTe|h%jluRXE1Mk=&eI%D9Hp}b5oTQVk#CjV!)|nbER8m
zrowb*KZLAZ6{pAJzkhaW&vu#5xUbJ(M?C6yf5dVhKDIs8(941;uF!jCZb`v}wF;tg
zDeG{nv=vuv>&^A?(X!@0+rI3)Vsjem3^jgI3w-b0Pa6_DDuXXtIF8B_Xca;V%_8;c
z%wxqW$C#TYUmW)70l}%YaU1F%7?<B#U^GJRs5I`HL@U#4stq1@n|3;Fck<w>#k}%d
zUxmy6;g9=rybUo$tUME95c$~q;Vvh-$L_%JwmKO3hZZcZ$6oy6tb*nYS5Kt_swwsJ
z*yGa!1T)6MpKErn=!g1}hRnUzb597l9UbNp#c3|q_X_Q3PMJPB5utG7(D#(q>lWF$
zTpwVfR+arez7yPcfyx1KAE1D){=H?sg5g$}5dEneF~flvsdwDX41s6P`y@Y+@56Kb
z>7c&9rT)rhdN9+mQ!7t93s1gOBw8204DTu|&@HasNDvHfNirm?kBNMFugAc-YcC8}
zwWoe})H+=t+3(cr=tW$rRK|MfaT6w<7>#S4T({IgZ>Ns#^P=?x*T87?%?<?R_?R-)
z00Tmn8@v*@@7&oPv6v6=<kdAT6u#d-nCf>}W5+BPf=o^U%Wous=GAf2+=NijHLj|D
zN>{eAMDo(unK0NA=)r47PWA!Yi2K7S%gKPwj3+IFw(}9D!N<I%Y`C#72x))6vjh0V
z2TwVf7FKGP-X-6ACbxS%=omP1FyG}T9H>k(s_@UbKEZaG%qWj`WXzU)_XHeo3+5jW
zH3Sy1<^7X&2NYd#t;cYMFc()?H}`s`E(PWERJBISUBa~oY#S35|0Md;7h;bsrFUl>
zjI~1439*E}@c^FEYIK**J84@bnfvBCa~`<plCKuxwarhqX>1E(WWOOSAN924o(!<H
z$-9JeUGU|rss4KO<zf}Trs(5n3$4%}1(WF}2<>xd5u^9UJ3*Sov=d7W>R5sb5+p#$
zt;GXjtOu+V@c>A&Z=!jG1f=^jZ+#7ZR(*O}mDV^8QN<HS^D|VJr?dbi`}}I7v20`N
zn?o4ss%`&%P;lylcxbk1GW}LTZVf2UL*&B`_sSgYan_vim9{dfub&zN(Q%$x`r^*Z
zn9Hm-m*>xhO&@e}5v*T%OS}@&kUF?AO2hsNpV_2-)+~K^Gkmf<Jm!lRYi3xT<WdrI
zkh|^_1by5~6ZLJ`YUnS9W`4v&8f?lm+9ANSNoNFcnlM-ZZog2Uh8>5x3KC?gRm<r7
z!E;h3KmlEouuq#Ux1M$2qt@QX0%(Zp|Fga!QI1_dnw+eQE0|3V*a8e+%*Tf|(5NK$
z+xo&U!iCooN-Oum4L83YJ;4Lrzn3y=Ps8J>noo!tKkUToL<uwMv*^k0VO&WcK#FAt
zpgC;+sqLjBO$Hs&PBH=xi90$>6`~!N-3Kg{`Yx2TlQaTuGemVQINSRQcse(@2q8^1
zy$Z`Saw}3PUGA6O5~!wlB@RzFk6q`NR?H`?f1N1o$Xh#awDH0r0CnyoY{YPg=rN+B
z+URYCKK+z`3WYUr{TecScy0>A)>z=<z(X#|>T2^XYbE2bOXK?CdLO;~^7lJ$&Kf*}
z3f(xX&o83579t?Z3b-5mR6TM@&tYmm9;d(9wgS}XE^LG8PM2_YjW{Rw6VesVS`~GS
zFXSzDP4mnvh2;-|M5wRk_81hj_rB@zt&QE_E#A&^`e~QSHv9}5txdK2=6h!;W0`!S
zZ1Ssc$gl_#`%TWip*2iN>Mde7+>8CN7NUFRw_Bi@BD-{HXQl7(5H`~b2ozFFLXT97
za&lp>aMzJikH4B|Y2+;^<ZmDGy#m)__3W@gU~>?B0e9Iv8m4xCr8>pO{h(^DQHI<B
zKXO@TB%il{9RsmxP4p+vrFxC)2bluS=!f?tuhlcCXJ%ap3}!J>XTxL=$g6iF&tm2^
zMz$v$26LXH0tCVQlyibm-fTN9D&(Wse}g5XItPiE!@<xaY>t_$`pbiZkH1cKH>T<C
z^Xpdw$w5GEX*IEWe|YVtP@X6%8z3&IJK|2njMF<}@fH_e=q7TK;F}~AI^E?@S}0D(
zd*@k&bC%n-TTYL1IEXXrdTx9NW`_|z#mE#}=+tfLUu<d=K>I2xq9a@WFtNou&qGPz
zPG+)G=>?aR!jlz9z^^RBzK_9q;^mFTN@2$0CZs)8m5!x$<NBGA^fK#lefA#%XZ-rf
zl>ST-aIP#*>ov+%T_-2Ds=Gc*xEe^mK3S4DEy&O`M2o&TByL%feSOrQXz+RfYc9d0
znv<#}W!OSzZ#-Wa-rUHb-LV|kYKvsp?^F>fRCdg@;Jrb4JsF`rGP5}~*)R?DBK;XR
z>4!Ia!qEofnkSB7m5y$X??8Ei2NrC#b(-=#;1iPAlBhJNU-51GU*M2vW1NXv8%3L;
z4OqZbFM4N`PB||rD72~=XDGT}JRz4>e_1*<$8Mp0!)nCdG{L7L)hu3?r5U8y(Db2q
zZSPq<%rIarUPR2a(-3Q}*gBTE3=#d%zRkg&pO^55!4DiUjf=Y@KVf_L$FLF)!Smb-
zOhfMo*%^FF!}Nx0!fGPJSmX0`W`n`F%GCAM^S}L@0KrKKc(HcIdzsrGt&AaE;i3cp
zxMb*zyDB}g$mbm&XUV3xn4KOCPM7KilSt2fJd@<3XUyBm>es~uIiKq~aysr^mX7XB
zm>q2%KdZ+ZoYX%JQ37t=Ox3x8qk7;Q8c(@_HfKb8GF1SCITOYqkBKn76YpACekVJE
zaq{<>l4X*L3=#`@|6RJfHV@#Dmb$T%mjmA9JzQ&DRqz=4eRHZ+Ze#8^iVXBvIEg)-
zt;baf97i_}@kACsNGk!Yqu(d*+h#*dVt|LMfI)4A<(Vi6c{?^Gz3_=GiAo;L=WsmP
z>dCoB&A2(QTv)Y!v2gXqCVsP5W|B5Psi{pGB=0!wWg#33LJ31$FX%V0=}_!fu#^-`
z8}e$NG2tp!QJ10bnoK;XEutcM2d^rF{evfq@x&Z0!W$fvEb&VSqckNv5lU-}Qdz>M
zuqWwPN-;RqJkOaINo(~*jNgnqNc(>x$@~15ayKKp)g1o<|ICgvtugI{_N=CO-$5#+
z3cSdRPfxPB+7s^X!E92wayUy$YUT`slW}#pKj(jbo$Ra&uxADntM^kq-UrTdq3D48
zMD5jdE4SOV5h~zc-zXm6%<q<ct<LcJ?&zs|9lN*EMbvK`Zn^)=Wrs>RBsVxZbvgM%
zyHeHB#+%=3x5Xi;(mBq2<1;cn{+1P4Ay3?Gl5;T~Hqu8IXEWwPUWGQZA8+ptL{@!w
zn~Y@DpQ1ayN%ttcvl$u*nR1r+xd3}I64)0@Op99<jLN&#zxQC#g5Rxi+zo;Mbc=t*
z@FMTnQO?r*?$SdtYZphx_n^{6;30vRSkqQH2!&GEDkp&0s{o${Ugr`)Ttt4k@*f2s
zwbupS3tqPbVgE;ddAC&)F#PSF7$~vGId)F_7uRvc-B2t-1ZnL!gne2AlGG1SAx|NM
zIG`NT8Y7L$W@oDT*E)1AoOKu6a9IWo_#+T%t&3;mQ4#>mMTu#i1rhy6vdZ*vcndeY
z!9=QSAhcir5UmxsRP9?r=G8A-AUGtC@&V51R^>CpY7cZ$90wOH1qA=R@-0>r`dNOL
zTBrxdh>3x!7YOI3sCzq#e>?JL#Lok}_9e6pKppe_F@&r8lBLhcx-r7(53EN42L|F=
zbCFRvFRr6)z+l$^6mK!~Pes=_fO$FNrBUbT$P$U)%2|bg`vih>q*M;HgW41@(bYI~
z0GQCHK5RkwUQCfg$dErHJh3(pVrjD3eWaR$KtDmE96)e!wJ(3HKzQ>3fLq=^Rb%_@
z!uX=g#ZYmlkagt{1jh%$5bVDX8WYfY8Anpc@39v|n;}j9nYo=Xygt0icG~(tpiH1V
z1u1{310|`V5ya;D3-Zx9-|FK!9qz)z-HN#1{s7%5|JrShLRAz$qO+U4-3?B!1Ca1N
z3tpPO`X*w>M6MC790c(`fb5*Au-8IVYsA3%6d=F~ip_hKH8_#dqqjk91%mWkfuL7Y
z&x)o-nFdJI3(^#XbLtZ(o1hn_11WYRg8Pump#E9vqo>qxkzVN1`MY-q?K<-_Sk5|p
zYmYQcU_Hhdi&Q_cZaZ@R{$<j`wXb26BsqLjil{GikGQA|UY+#)1xxcKl-%9yq?<F|
ziCrB@koVK=kILkWO|~hX<n_yR5t&D-oI#sA=`dQn_Os&fOT!Wr3NZ>pnM9Idkmv*b
zkfo-eRlS;bpOQZ)aD6`|!b)rA;(uD?KMbM|{8emM*?zS50I<yw%By09sX+>6;CGcP
zhL;@j0A=5NH6WhuKT!sXPL05K=s0?`wH=4SnZC~!Ga@>6<g;`IpkM>5S}#g;0yFZ%
z=TcYe6#9-x@c*o!`E4pg^VIVPC$UyLCfa+Xtz<_rS5+O+W<E~NRd8>8gzN7}Bt4f4
z01DBVs$+9;=|btyU$yRE<otxedF%E~U=02A-)fx6MWLdR$FN(B*AX}H&t0QSA{#w{
zE#KfG`KkmN>Pj-x^l(K}gVmK)kQo>0L#4X7g0h^O;aD#I*8v2A3%gvf6dC$5mTXs}
zofwg%0e!}TJ<z(8E~>5Mux+*E7NNqdR$uLLR}6H57Q^~1$O7DSsh2v{D0>KhF8u88
z<ss#B?F=e=%Jz8R7>W7C#7!uyI`-?c5V3^yr0oEBIFUQ+VM|?HU@mZwT)HxOgN$(@
zh`rUT7l-!sUT0p>9M_t;GhI#=jgpgld449Z-J!(PD`1vhY<DWyr@Nu_knA>8!PZ?G
ziPSLS3w2*{wm#R=l$XYJZ6La1>fgdLhiIb?ZIlXAY^~E?OEqy$f`_54<ZN<l)X*Sz
z=LSHDDT!=PJI>1LOC|^IZeDrk@SJRQ_B=y!w6C)L4Jb7CXD<<Qmb2{8N_deCZwVT{
zDA@shV#8`RB(|6}bCIm$h&I8iQjh-1=wx6;D%(l6y$HH2aonP?W;%j#lakb~CkM=H
zcX^?JH#t_}s9OcFAH&5F8q*#bFU{KX$1^_s9ja|Ad!fh2j+4k2MtM>wry6HFO~)U5
zvsigmg_!aBDj!~C6p9&M*pW!Gt^=Ci_!heXY6*_TpLp?{#g3%SlH1Erc_~sd`mwOS
zsunlxN#-}WcO}1M$f;a{m@#F2gcWA7K}ru7`Q+cBnsNP`gl@9%goCT=U-xK}CWW<&
z#PG-<o^?jHPBU3Wq~zr;)Kl<nMcoq8^gDKdu$)g<wYwSPYR&s6zfb*gp^l>&Io;FX
z?@Qy>1RS}}j<?vjgeRP<);5oAHWlUAP}DPfoQ1KN%O2(GVOU%F9X8FlwwBUN<Pvsx
z?QxJWA4R!b?Hw@1JoVN0f_%UgX`Bb$0i~MNijQ#VEnILwAHtVBo!Tki5a3Ljf9Cxn
z^AvH(d`uOijt~3q>g{D}Gi@s*v%WNC!Aw`07N_6+A`tEyeGbp|yg%mbk#mrc#`vv4
z{uO~RpdU}4%QA|We2o`jDoy$}mY!$ejyY47c)UsMf~#tuZRNU!O2&K2cDg%$)MI4E
zO-5*<CGnXBk$bSR)<)IUH|{>iW?_6BJQyW+>Z}4%iA(2Y+4w={w`Q1^nJVf+sjyhk
zE`~y#jN-7y{p9VjKk*Vn?4TQ};g}ZHN$)y`t9nwR9Q(;TFpoN8DV^2j1baTQGqdTP
zYi*&B7pdw>Y8v-Mjuh^{?oCQAa(2_EzLK_JM0OC_)KuS>wYncTwX}Q-b__3A{?f}m
z|HdsoS@qH%-}T_%kw2}_f&~N1en*)cKn?@$LmC#?=dMbp=MqJwSXHht5X_V~sB3A*
zkYss!w_9>KS`vHx4SU{t&PVs_E^IpYTY!#5PSi#77u*4_*7KXs6~?2RhI-c|i9xuT
zzIsD!!-y{Gk3f?}5|<LeVPYMM@Y0j+5F2b2Yq<}#Sma%7A-k&dzZ!Ga6!mgTrgpj?
z9D7EMR=Y988o-BQq#o{kz1bJ3Uq{z0^)Fh=ktv>dGj82%3#Q2a0`_}Q2pkJOIMEi%
znXP?;#h}@^nH4W-5clx95qn)M<#K4(*1MxR*JyZ3_GAbHw<>xthWXucrnHw=xV_7i
ze0MVWmMt3&B)kg85!5-$ty6AS3;Bpq+Ai$b)#DocUkrWL=LC<iG^$u!wN9%bO;F0$
zqSapcO0k5B-E9ng8Wty`hU%SSofx>Hu!#;yRspuR%0%iC?iKya`xuoY?8z8X_sU$#
z2fnjDnR<hvv>lIhOF2t|whsN}mlr>#bD>NC$p?>%_C3Y?z<za}(7T#g#tUygiV$9E
z1h9*1W3OanSwcA7;t{5B(M^;eZjn2i7NrzLej;z@y^hoZ<+}sQp3d*LIx<$*vtl`J
zgG+7?bl(B}J^y)QQmcp8jij`+3&-YDarLa;fR?1qhMdm*t70Hu=xNSW0L0pIkz`D!
z#&zAyc#6G$@uOLxje`KU=)|TiX=pi?AD8KY(|9_|IQcFY)5TowzHa4qCM_V-eUq%=
zT+E_AHRrrop|Tfg=h@ny(ljXThD-}Bkt3n-1z-c=rFrogS-Nx8^34Jc&y`?vD%HHg
z#*@n%il?1RyndU%(wwg!9wvV(*bK=}cQh-VAE}^|m7yy{avn0HHd|fEfNfGM*0JBf
z&em1LVdsCpo35tJw(j$J&XU41`%M3ib3N;&Yx%8$Rz!Mpz*r<rmotZ>Km61=#X%_B
z{zbhfb+KLW1mc5Sp<{O?(|b+|TBp?Q;<J0dW_VKkj0SM_GCH%QX2u!%8zs9~i^`<$
zXT1(`h7Q+-{#3?T)t@WG*`sCfAEy@zBFpNmL#wrG%RQ<gIKZ&3=%P&v^5Nc;nu~8I
z=$7Vt?J^A7_blNy++CPje$-Pio`Y+wJDOs2gWmAy!skO&vYP{1#j~m-NciU9oe8i%
zVPx(=-O@y2KTp9V_2DuXJ-hO3tE=HHToQbPrltq(wc^V^Zg8RQ%5Md4ojJP9oO~E=
zng5ZSp=m}lc|g5&@bopYOy_Nv!Bz^1APrC2H=E%GJ}6_4y%VMP7AiZ51Re+m_seZa
za<LELSgI6xc0YpV<wJi)C0&wu625WMPNgIIAPPJck%-yjT?PbugFKcGe0Lhf?f;_f
zh@4tpLYDfO9^3x&J|<dGce2u}&eiOyb9bI0>c0EN-f`>7Rxr<h+deJ%_4ae%M(}@$
l`Trl%hX3^zWsUbYcgh{AQgUN)GLgi>K*w0SMDudQ{{Tf*qO$-1

literal 0
HcmV?d00001

diff --git a/docs/docs.md b/docs/concepts.md
similarity index 100%
rename from docs/docs.md
rename to docs/concepts.md
diff --git a/docs/conf.py b/docs/conf.py
index 51697011..a587c316 100644
--- a/docs/conf.py
+++ b/docs/conf.py
@@ -220,6 +220,8 @@
 nbsphinx_thumbnails = {
     "tutorials/sobo-tutorial": "_static/sobo-tutorial-thumbnail.jpg",
     "tutorials/mobo-tutorial": "_static/mobo-tutorial-thumbnail.jpg",
+    "concepts/sobo-vs-mobo": "_static/sobo-vs-mobo-thumbnail.jpg",
+    "concepts/freq-vs-bayes": "_static/freq-vs-bayes-thumbnail.jpg",
 }
 
 # If not '', a 'Last updated on:' timestamp is inserted at every page bottom,
diff --git a/docs/index.md b/docs/index.md
index eaa36f92..bf0b4986 100644
--- a/docs/index.md
+++ b/docs/index.md
@@ -79,7 +79,7 @@ LLMs are good at recognizing patterns but really bad at suggesting Bayesian opti
 :maxdepth: 2
 
 🔰 Tutorials <tutorials>
-📖 Concepts <docs>
+📖 Concepts <concepts>
 🧑‍💻 Development <development>
 🌐 GitHub Source <https://github.com/sgbaird/honegumi>
 ```

From 4201d9869ce39928901c8b3ad5261e32c3c5189e Mon Sep 17 00:00:00 2001
From: "Sterling G. Baird" <sterling.baird@utoronto.ca>
Date: Wed, 1 May 2024 17:23:19 -0400
Subject: [PATCH 11/12] thumbnail

---
 docs/_static/freq-vs-bayes-thumbnail.jpg | Bin 411116 -> 110440 bytes
 docs/conf.py                             |   4 ++--
 2 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/docs/_static/freq-vs-bayes-thumbnail.jpg b/docs/_static/freq-vs-bayes-thumbnail.jpg
index 248a2f3e5e5cfc2f4ce70a398fab60b1fdb24c21..112beba117e057eccb8011c7794e7ce5d759322b 100644
GIT binary patch
literal 110440
zcmeFYc|4SH`!9YAktIu(tWzPC5ZSkBV?vrDWSt7xLS>CHV<|!sLLs9hN}6QLlF6PZ
zyRl>&BFkjPGHzx&chC2HpYwf|bI$WSzt`)$&L7Xb+%Lx5bALY9^|?OR^}gQM^<fXQ
zrvPCKGjlV519~0hI0^vlG2k@7xnswlALzve{c!Vgb8~TV^YQTP<Q3o(5a8$I=NH^9
zDlE8LWH&#*u!OM49x-uoaRDL8y%J)3Ma9I${yc<(6Z#Gp_bzVkU1Ea#f@1%lU+m9-
z2rqX8j|wNp0bqv+2d4-Jy9Iy&00;M<*>Z3I|NX_W1DfMb9^PGi{Ll@R!oUs=PR<=%
zoPTBw-5m*i4seNZi|*GrxpU9Ot2_s8iD^csKIc_9Rn{nO-AhqCe(iS5E<Op#y;9Oj
z2bB*UKBA?qqpPQHaQcj?8RD$Dh0P^fJNwHHj&ANA*FC+weFB4mLqfyugvZ{!7Z-m&
zAu%oeaYkm=lc(7)3SJhzDk^^c=0kZ!WmWaZn%bu3me#iRFCAa|`UeJwh{GeJ)1;Z%
zpL6rS78a?iYwH`dO>k@bk6au8=YNRxSIPcPE)hts9b8<TTs(i|;@A-iU7R9Z-1{|l
zik`g4bM@Aq1DesiVy9A{mo@HEIBrc5zjnKqPeM^^T8a8cw7*F9uL&0Oe<ayo1^XXz
zjRFFk9MIx%iU4qcY4alS5b*Eo-#YlW4gA{%{%r&Qwt;`!z`t$a-!|}{Hc)V<bYUC)
zLZxI_bymSp@zLS-cY=WHiw}h9BfW%QhrQVVHjClK2JV)av4Q-;A|?w(%@=f3QDKRe
z%CKTGahqG%wiM8!B4C@(5j**l4M+onfA@8t@ytT+*DBhDUG_(cxLT&mwk~kLIrJim
z@4sG03TY)Cua@a913F`@mF(%j^>4O+hDH3>VUMUp+<!Oike}+25NF#n3u}9Mowx7_
z`GTf7`2@eI2e!$!0NZ2T&te1W-~cUr0#|JL)5&%0hxr5Y!xQbWdyr)R6b&ZMEz0OQ
zdjGt9&DG(@lu%og(tn7+2L4qQyoyEj`{xKxj(v@kxZt&eDVEsySoro~um6A0vv|Tc
z8-vr$2#g5e`uM#=kq@hIfekd-6tIE+6tib%;h(WIb7M4l3-fm*9{LxrE}{|gWtfjr
zmfZk3Xn5>*<v{0N_|tc%HfeF-)mK$xg!?8kRX1it7azT5Iz}1uCC*e39=LK3QJJ_I
zLD5ZPqa<(=p8D=x-zq<NW~g70s*6aD<pdLf?0<O*|7@y9_&b*(P2ieyVdNRs`U5dc
z1&6VC1{>IAm&^uES|aHT!i@cm6>=pnqtjvLBXY&8F@E8M#Gx07!2kAosdz|{qN2^P
zdfC?MWL;@z6C==>y4|z$*?$a>^49?@q4CcDkGChJAKOOwF8MY-QGR%p)y!%4pPT3Z
zt>5@Ibx>tKnPEEJJ(_gBy04OKX;p0ks}49gv@m1UW`6tT%fQc!5?TmB6h8HVuo~<b
zz)Y%|+}3miz<HKLdouQy-w+$1n4nlkQa~qpZ8q?{+KCN(-ok=!eyXgiPwO&75l-N9
z%Gw#~8Iz;sl_}QM4nq<hNA_~Y-rLL<if04UH^PXlxbXlszzMkmQZXC2S2ly%mI)>>
zgtFK`BEDw@yq$F){wuJIk;4WO3Rtp*B%y8DW$Z!db}!Dg_P?75!JaVnpyA+<kvWA8
zwBPm_EuDKsUA!-XSN67u70unzGLyZ^2L9%m_z&2tuwfb-pdl!e_bWP?u5u0=mW{XZ
zyKm<2bH2?}2t-<$ie2FNFBi|xNrpsDYxrcGzlkIpaQE39MLnr*xKjSnt9quhEhIIs
zXbs;L@b^k#DzSl>X>5CY6JaxekZZezIsxmIWbE0rt{Y?nZryRASwtJx@y?>Mw1{}M
zW%PnUdI$a*n4!{4nQyf=Bg-Zfx@gwsiXV_@l;0o`DCfJkG$D;m(sl<h<&+{2Eh}3G
z8;-Jp_okDqgm8f6O(az(U%C?ZXH^>S5*W6Tcs9V5N+l*3nbI%GxY(cuDrHd1zOPo*
z-sPS7G8P-BoXue2Yp_!(u+<Px*5@n~T@}Dg_nL|U7ys_E{{}7nMOBM6bLK<Vf$`yS
zE018gw;syXj|%WBk8U=nwnpwp9f!s}OPK9sDY60LOVqahAx?_#e_64s>TEz9!B9<+
zh8)ybDI1tVL$kb{v5s4Ipu;-pPQIhGXY%iB5j5wEUt9=Dw*uT2PRK)F;&*E!tm`ac
zLD-)sSf8dQwTF<O)O+B~BJD3e#x4(TYfVy2WxCJ}+9@V8Y@qB-AhuDg62Ij!V~GTX
z16T*6Ot>qUFEEFOKUBbk8CLTBkKmP_?LTW9UNUo`d6b|;O{5lwGEFp7(e~5NK3lx|
zSnE0>DjL}CGtUORR=|U>7D5z8@*`EtW(3SR6E%lgCeL}<r#u}fp2>UBpqDSSFgdN>
znz9@1OHq49>gGo&u`<GQGO}VG=y_ZT7rw1hhHvWz6$tHCr%!^F5tPw6Z>7G<lF*ua
zE+5`^h^F6LAJ1l{|BK%FA6=tR%L=HX`;cKiy)rtwNVFtws2U7-(QfRktPSfCElt0t
zkdBgWr8>3YwXT45PV{EhHPLN0@Co%6R&CGN6Wh)PDmL4_8Gh5_p{L&WyZ4jHTn4dU
zKg2x{Mt;3}Wv(G2g~eM<y`^aPC+arU3uA}6*+APbP8!c=O`{2J%FLIp-=luz$2B?>
zTWStgm#8;wJ7+7$1tl9?x6l6mT_)6>UYNQg&G>_g_1{MJ{~G20`giesp5bNpZ2QZk
zWXrm;&$gV3*}!uWT+};Hr1y8Q%Qc!D99ea0wcGHXE$0#{6IEHw5dTWukVz3j%ZRjl
z%j}&a`jaM<<ro$F#J7o=kX@c9_&GFahV^0NC&cUADG!s6E4B1inT+Tu>MFbmYDMok
zAvc+9Vh{%T8HwkrbDv1#j?CF>6V6#_dx^#wfzi&Utpmi{0onFX*+BiWIlc&ceFaob
zX;mLJXbpCfIm<B4*VR}IfAgXJ*h)E&j=CCln#KFPumVQ42^%F}Z)Y^yTFkS!Dmb=W
zUV&M5y;Pdq`QP1gMhQ>hsd#G;MR!GC1`#EQ$QNPl)=&0#j|B6_=W((D;CgFh5!Hsp
zODOfuzO-q=O+6F-tmVY}s(}W(N4q;K!-?~#7Fc}VKG2C&(!CqRox_<AmO2Rq+^jpV
zdE)BwtrM_H$VE&t#83!k;gc{_OHb=_aCGk`$46fLi8cowQ+>U$T;k{;qarM{#WCyd
zMyH%7qCN{bc(l-=^!IbUupdEJe7A*|_uyU$Sg)0+a1%mt2aLvSShD=}=x6o*QCY~0
z9`x|^id{4X{!cD(DuN*%>pV~a8)+s>Uw<B4grq$ZsQOUh_W&;?OB~%;%OZEv^}u*q
zE_j*^@Z#fEmbK_BM&@%q+%4`N=QZ>5CLVKuu)lW`s1I1m+3^S)D$i$`jtwh|CF>`y
z$3$Ja`i{xuJjnB%O#yGWw{~w!*Q-D%){_m)1aFJbce1+j6qs2i5<kLxTWaHQ#S`W?
zJw&}*zXIlZoj}C)1)hGbe?iU*R8H}p;Dgq-_bTp+44_{Ta*@d}F(Wrms%LB2X;SHR
zJ>td3M@rYDIO_5_KH`5ZwV+nj7qJ^l>Z~Z=EgT5>FePj&6r)6;FnB7T4@Idm@?`_3
z{j1qPG!)UOm#|I4)M}<+?Ieq*lMS?9q!fL=&A0$2PfzllG#IlinP~Saho4dV?9K|_
zmQiPjAX-#DB4#Xit}E&Yu?$bZa6<%H%mD|tWtIU|Z12L)#dP&mpRs;6;Db!c_T`jp
zL5aMVD;y}iwOtwc6br=>8!)zenBkU>LmGZ#C85+Yeeg2f7W_!OM|(ohMt$!OvUFvX
zXO+PD9hv*CWCPp$G$=cWDGyuuY>*L`PUUc;s>d~Rc@OH0n+b6z<^FEPG^35dd--T<
zN+ynk67<pV2{&E8cJ}%M+PXrqNgE;S4Rb&A1Y=gOq`VAuqKMLI{WQmN@bUO)dG7Ml
zkm8fSc3T0M=6@Eg{yuvB-+XURLNFxkS?5cvpg0sV#s<tRQB@mYDjVR2fNJ`ie(bVK
z6&vVH>)~k}WbR|NV^{rHJE1KagxcOgT#H1#!vau5dQA{$kY*@Qh;fyIx)&nGP5Shd
zyngr{72AQ3AK=*bHARcF0ipXKG|GujmMl7oQow?VP;yw&;aN-?Vfq9tycZ?TForbv
z8m$Q?&lsyWhe-QBmz#U8o+0z4*~^4sKk75~Smej`D1B_A{BdkNRtWMGqJ{Ylk!s(a
zp8KP&q+iP%%NuFMT*Wga`RzuJQ(kqOjMhQc_l7COEF{dvgsrd|CE?quC0hmgLgUEc
zEEb<s1l>7o(z-_tEh!W)@qOeScZ+>n<gmg6G5JbbAw~ys#Sny=Ee}g#HFh(G2=2uk
zA9S;1bX^Lxw`|{G8*c==z#Io|z^CmTzNq@UN5N@zI~p&)!w%Bz7oW9nC&0(yUkO<O
zve++6Bq2JlGxk?ozR<Axh#QNqpcvDOlL>-xASv6ErYSaSff4dA`kv7C2h?8h)N6NS
z^5Z|lK?syF^5UaeS@3P8lFij9_+d!fA5l~)1<xE?nHR^{6Bpwld08eVkcR6b86^DI
zZGaU=n0kg=^?;lVVOe%^+rWaBk}$byg?3;ASCR-!(G3y=kp6cF4dweA1uC>No~2`l
zMVHus2R^RkzyAjcEAN$vtj-g#H`vN~GDj)q8+_MqbU<O4D_rcP=Yi56nl5{}ELslT
zxTil$Y{XI+s`jh|D;!I3k_{x=FW_2NfMN6n(3b=gSiHb|F~sWaX^VI-g<^#9vApXe
zrs!-So~1ctr2hu-(?{B`*{SPe8!sDJx)ucwAza0cy?~GL!9>~vr+?LJ6fX!$sP-*=
zU(0Crh(2QA*4Oa<kLnKM+q6Lh{k)geottibm9`PPu6^e;-!~@3*oB|`3aCuu#|&9~
zj<^qmmS6Q!MW6+2kE?2}vvYmcE1^RMKh1VjAwm1G@K=R?1>QtmV%mK_zZq|j)Y>D_
zX}()k%-1H$;|y1Dg|s5#a@J!qQnmi*`9LxMuekhtp?XOce{_lAJ=8rN_w;BsuqPC9
z*tXjNl#*6Q{6iDiuUcB=OF{mNgldP`{vp&c)4E5xS3DoU+6Xw*aP`y@+K>67UMb(|
zhQHSHkaAS6yP|>-u3eldIpSD&FIS>hz*N9jI_1hScm{y{gZ%#y%c?9SZsAybrflFK
zimJ@|bO74J9VcMjk(f*fYr<!-)Olt&wB2>_-wd`(X#C8w3e<>Di>tywWy)?;*gfsD
zZV94o?ho27Rhf?yS1$0ZY>01_Kzq2J0%z_)3S*~kz@{&DKyKm_#UCC@>2~R++xI|t
zAHc+eTdIl__<ae`%unzxE8tmj9lD&T&1m#JH4+@rd)>Tp4fnO%rA7~54fjEUN8%rt
z=!acfX_l9U{Q0za58lKnQ#rU(E$qT#n0zI5=d)K6zLKe<BTkl~fePH0pP%Y7r-^ZF
zq%cG~Pu36#zAHjUWqS8j)MT5B-Q4l&Ch`+z)4mx_Ifojaqfha~sMR}=N`jMG4OSJD
z%8MsGd3iWZ-tua1Y6Ie{o@5P3f_<PT-3aCCb!wVnN-CB-(AMRDANISxU_Y`F+AlFl
zn7%AnIsS9b>vi9Zq?YVDW?aATwX+!+<wu|$m&o8B)VBQOYh6Pid`$aIn={Mw{&}Ok
zOkWY_2rxycFor~WyJLC>Gi!=2c-VXk+js6Gqpv0j1|C&zt-F)t?0DJvRhU&Gq1cgv
zkAl62q9c4%<-DI$bFD)2^PPvZ4Rep)XqE-eD#sEF!%@lbsmE+U2CZ87?T*K*=66rB
z0#ExiUXIedRa28Ml+TjJHu@CNFX`kRr(E*B;WR&R^tVH<(Dr;4!HmVv2%!YDuk30j
z!FDe8F6x{&2^-_rLiG0Q+ic+H;%)nWuNzw|Aqmms06!0}>v1RJ5H>MFB$Y)Zc(A*T
zjGRW0Ri{7csMIL*)}Owe9XYGiofW4rud=8C0jl5CEdioF@<J4~x_92$w-<@E!MWAm
z?tOPK{w=E2A-i`$j@cYicKwjwPp=}-%2%v7q*mwc)N46}hW+AfPwO3j>e-T`ny>Bb
zW;Ij4FiN(5G8-_0ofLc86RO!0w^g#Xl+&e3TiB6Sl`=k)C;38sun~0@UpAq_vvnze
zdr4lfno@6n^SILl0^_NEY~Rrc3w7s>ZXfzK;YmttgGtsjwiPB2_D<r;s1?KBUH#_U
zUE)kNP`M>FthsUfU=62=Y4OPEAig}4Ai<QOD4miy{8%8Zql)>6A@|Bh!;2agf0>Z0
z=!hmB1l(B4Kg9C%X5^L@92t6$jZ~ps;lgkz^D>)>q&0}gFXzABvOFu^O*+Im3>h`(
zh-ZrCz!@?Gh*HGY65Oz($DNo(4SOwT1l^jvr}ZXgTfO<5pSDo_c)B_+KD@0#Ms@4n
z(~@G?kX+QyRwWl^)8_QVFSs+hcsB40PE&|xLu@R87p*$A<X5PbqgZ`(r>;nc^1#{&
zIWoF{egWHPijl>9*L3z)xdrh*u*3_uiX5tDZN^<sJ}VlnfrE$R$;t5LH3|zR0Qz9n
z+oTyc%A5~(k6him`^Rk=fn&auCF_BA8_!OWwpnvdon!Hu-)<e3xU)AmQ*q;96#S%-
z3$&I`z~r~C7Te14H-hwbZ7Hl7K{-!Fm<@|xgDPWQpRj@3J?wiyeKMtMH59YocP=H0
zQ057xAO{93g#^4rAyPqQInn}s^y;VReYmDt^<*3ciAsXNTabR`H^gZ_OVetpywz?l
z1kU!m7v~0?2F0_OEQSuq27&_$d_=KLhe>yk5~0cWWkx)1><Aof_-rSTAN%MIm;Gv;
zU!Zn!S74(;qtQ7j#nW+ia=iG_08pu&oE1ndXzF9YsUCEfFgzkD`0bj8w;H+K&TX+~
zN0bTkT>z9_iWnJE2+g6=D+!fOk;+zg165n1LuEaEV(yqU)fzA$Vk4#i(=SbIMLv|S
zj<yboZ+4Za^}S5)Z7`pc^H3lPCIZ`mGs+B!*i${)9Y-;+uuEWC>+PZo14_ODPr?uO
zr>?k)K={F};jqu1`2O`0K^-GDP(nful%nk^glX*k292=3_NEDm{@H=bi1O9b@_H9H
zpV0Z}1t+1gkD?KSM-}om+rkfK96xk$!uO8L<86=ghd3snydtItN(;}s8)V&GXI>Dh
z7noC>OtT+`mE#!VD$}^O61t#}<C>8Y$*D<P^*4vk&re4E(Yfn#OKV743iRzjG?MvD
z%UpZx<1|$wR%n}f8fh&dU;iOJ_tsE@L!7tj+{4Sy3NrR;8pvwC*XNRNi-Nx~y|O5E
z*5m3U#J)aa6M<qu7!9-mJ17k+O_+oA+4Pn?7(KnKwAK=paiLtk^kX*tO_Yfs^8<Y9
zl|38yKxp%>7YvxP#0n$i9!O&SvMzX;uW=ZL+%?u%a+M`eoCR&4*|Ws0aEd}({7_0E
zjTeLFNx*hYjkAGa2`HgB=>xu)WTLSvLh~nJ14Z${Ht#^}xq$~5^ZLrfsrFp*M06SM
z#!zw~6u=0xCn-c`tSVvMnwX6F0Wk?DnYox9OVU3dQ%y-Sm2{oVv~}-Y94S*Tvw?Qz
zE5cMhR)7s$uX|nLcZ#}!6+0r~qVp~E9(+7C{s-I<v~QWDIAW!JY%H0nQ+0A`M$`N*
zQo9ZJJYGUQV;!JM6q3H(-n=}xa&SOfr9+_n#}EC`)RfyuN1D(Glrd?8Idp{CCPutx
z)=yq%<Y1u&zi4%S2iH3J1ni_E1V68usGYvqtzCMzcTcK(t8;{>UpTgz0IJ#*3J%y!
zs8P;oj+Uc0tEv&<LYq2kt}qe~R9tEM#$%N^*VZ=am?3$yy6VciiRQpyd`i9P^tfeH
z_~1l($*B3&;?bIizIWoJ_J(M2Fclcw%z}C$IwH&^7|-L=5hf;7;b^{!<X{8d2GM!(
zpTh594Pozz?fg(mhpT|a2kQm6$67T!yRqUoc>iHA`c0f1#pwXy#nlt?l5`JDUnuMY
zzOlwA!0$PNBxOgFL6&viyn>n#)%E8KhY*@Dgzu((KF&;A<zbiAAdu_R5DzWtXKcrF
zXJ}EcLv+n4ZO|K{AOjk^N83%Q?aTA87?-I&jvyRalkhT4O6_#iN%#>B&Rcp@ZrarN
zHavKAS;8^n`!mW)(-PL}?Wx7;z9+*|v=ndr_eeL=e6opRm~InJh->v*fw#)+h<8Xn
z$$?)~E(^?=Q!vR`G2wy6cS8G9<Kz1;eeAVAv!<|ipt{~eZP5&2He<5R@dE+_jn_<$
zHlsbV+nOTK^It|w+Fykb2iZV*07IP6I4Q1O;=<o9eVO846tpB*zpy`he0a(uH~M{~
z?<#f%vAj9c4XwFT(m85dtVyy~GkW(=2PNk`+sCQH)S>!WQGHoyfc0AS_kp_Vhhd85
z&r?soOnvX+p|CH;AlBNx4Mwp<r4UhjO@QW?K5Lb2)zh0g-3qd?0BQdgAKm0Xh^j3j
z)DSgn!h`vicX04|sws08J3>L^8r~T>h3L_B+&)53!%kJp@6RThg8>EY`$b%|2m11j
z<QiOmK)fSd`DN<MkkD(t+m)WorZ1uXSRfSNJf1+bV)uhy=cQvclvnXdIaT}_nWp<*
zB1YnTXS12lA<Z2Jk#wOcR6?-k6F>Wl4==O?^WF2}T=}MKUfxr&6d`Cbq++NNB+|-t
zMEv53&wy?1;a%;@(I@y(V%WxWaCa1lLJ9P@m`L}F21qUJme$}SSr!l98%Cntv*|Y9
zkt_IlL@sU;^*su{%SaN65&01EelD;nZI@Fx0-uOTniUlSIpzpvge$0FLd)a_XK3m)
ze1iSO>cJL|3UEe7!Cmmn4cv<Tu*q#~B-X$-_41Tv`mz_tU=}H0k9VG64_^D(^;H!?
zoj#pR@3|{)-t3InEVW<+L$Xz32D7v7H|&EqmGWGS@RKy(J9zEh{k8dhnlUGTiPv7>
zsYKs|ZtDoMY}&-TQC3c-x1M`@@%M^DY3k)*mBt^&OY_MQEPfajap77J`9SkI`KO{-
zZG3bP(`4SZA?Evc$&a<qT5g%{?L59Gif0COElioU3rWSdW*-gR%nmZxbM|@P#B$ti
zp>tEz6LM{km*!3dpSc~7Axku0u+{hYp)gKvBKA)LgeiQ(E->k><ARA!?_}d<sK9|l
zui4|Dod%yoS48@*#m*3yk5ip!5xw85Npib>_Sm)rymor@<w5w4<EhzOv?%y%rml6x
z(9wOL-q$3qeh(57qv!<%8}ll}9QlR`fs8?prRyWF38G5D<Bo*1)kujctmoSY$SlA2
z9(N!c$!3b?3w@s)bz<>V7Na@9nzvzUu=vdF`VgTCSM%3erdtZX@9Hf$+eblZY<X?b
zgL>cRUdq#vg|&<JWsVh{Z_34XH--XkQf-mGGn3$99>%lKB<#{_h=I?p_nksAG6T3E
z_u%@6sNe%wPnZ2MBW?jV?XPBPM>af_YDHcOf^Y2@#uDqX+q@+M98_g8l(cY*e`2OF
z?z{p+#u|@>C8&`{-EOH;WBqxjfaJW&wLt<yV$=5H$He`VH{J8WC@9DTrfmImjfB1l
zk+Eqr)7{(D^sPcPXCWK-a=xEcr_8#0h;#c6=sUBMx*4?&7cnmW;Ab)7FZj`K?D&}R
zQsH&Jm+4XZio{&jckJqEG-R+>ezAelup)+7EbY+%)^jpi*DAlZl_Yqzp(8*A+o?D9
zGc}%mLU(7>0p=_C)Jq%`ay;kj`~3W|)5_Fy`kcse{rkrj@%z5mpd`R_7C&+(5{aeR
z$E-6I;6cpd&XjK|YuT%-GuHLli;LcB{zZX`X<u5O4(FzsFPh`u5j?xWgYXtMkdh}-
zM%6@mPk)kT8g-g)a2>sVF7@M)+=nNZPRLE9Ln0@FrJfqT-wc|zjz4>HX{LX_@2Cj`
z)*8Lvs#l&R8)2tm%Oygj0j({^5E!l73$H}QGDMO{p7HB<-mQ0wd$IP_B!{X#s}ME&
z9m(OA2qN29JVX=Ou!0MemHX=p6`!s(*@(ARG-q60>0z1zSPKZCVvdE`%!J00(U#MK
z*2c(lxgq<^rPhPLU!wL9y;0rWj6Kk9&LWxs4By@<%bcf18gG9nJ=;jk)bJQS^tkwU
zjW!b)9XE~e6K~nA7-zftp@Fd3$$TCwtq`iwhuU^&!LHvofsKO=D$JLE*eaYLi|<?T
zOE3uA<Uz6w8KqG)oNg^WH8YDWQGhI$e$Ho-2O~ZfX46~^a~rHbBM|uGaC2dqfXr3d
zxGzhn1Xz7dj$*}!FXLO+uVHPn0j3yLogoq5KKhCcEaP{1Z<~Z)oBFdPurrWfXh&O@
zT?tsv5j+#_j$mZg5Mm(m#X=|%B~h<(74L!?2JQN`mGZ^|dY@(MNZB>zX&OrZ!tBR3
z%3=<K2-CbxGhz@y!nv)+$?#B~GR9!Xycov{79!!M$|vN8@M--d<v4fBx}3>)QJwR=
zNlA1fQ<x%@vJOj_kSL={#%=5kHa<e-Y}Y!sTl07vK$&PGJn;)XO`v=)yHn_s`N6?Q
z;&#RHv9HIzEZ#iHFpVsN+-n`Ps50dqTuN(YCu;nM)ihlAi)wxR(O#!>ch*jjKeqyy
z8DcDEc(S6keYu=MV2O_T?SKn+%=0#MDmC5><}`dC8L2WM9$W08Z>=ea)`#p0av$lQ
zY@qr%V<^f5;%K4xPgu9O0qMg3FY$kh`2g<YiyN(*?qVs|P}VVWVMWK7d4}nFdWxP(
zh@wKQVKDtFG`CRbL3ci}0essZw-99l;tjzO>SiOVeDa=5i@esLJ@NyO4`T16Ry_0W
zcgMPCPByWoeZN<JyI(FPzwmbXF$L`O-5P`Nr+1_;Egk&yW|3&0R&~$s_~};;5{CO9
z-1@wstW3=ps=_Ek>9nJqV+4u|R5L8qNB&+}c_*z4rBu$eT>DfX80B(6lD*_9OJ76&
z%Yj`^wgl@+D!y6$UfyA_mDGt61W%vEISi>C|G<0m7>{CF7{IsPCj=^nws~s}*4*+G
zn{^*e&y*keeu%S{0(qhM_HdUU0dK?3x7)`Sy!dekvB+yd51_e1M)VM^&n!@z>3x~S
z62b3z|D~+{h8e44|7SNcDv=-y_r%Jvx((K)QOOnh6|dE7^G5CVr+M&jz2+ivy!0W4
z?|$uDJp7$T%=~b)%u+_&<H~K`1CX~L?WXQ74l?Gn9yo)!RbG`gg~;_EzE_4r+o!VH
z2oj!PH(ejpZ3*3kc@55eux$v~_Swa=-Wnd7LYOtB!kDrC*mau=JacAHV#|#m405^6
z29Pj*RYi&N`u$^eSDd`tZYtpO;n#5Bp-3XxV@4Ha`ibI4h}+jsmL|XJV|0Dw^3=Ee
zJYR}YVw_+DsgqD%iSNL!D>&hHn`njMrqAcR8wGvV=*AmKp5-+kQw|^x$L&4LlL#;q
zm_>xCH?CZu(i>+l6NyTBEh72*NT$AylNolYm`l_!5gfY=X|5G*`>9^^@<Np~TBdw<
z%DnxR1u>ez&%Dn%is_XgjE?S()%?|`FZjmgQSrCDFlFT>N0cml3daT{8@PWF!{<u#
zwO8WFlIZP>lEY`)Yv;8%HmaN8psM(6ct&Mb3R+p}eW<*Dk%xx20<=<f{N3}YS8zU-
zY8Yw;5xdN=npQn;jI^wEE;%gIGxNK4{-+7F0mJDttX)!22m5D&tjuR!QFFC)QqYK~
zA?x?9iw>I{!&MsK1s?J}v5P0Tp-YMd5qPOs8%Q}0*mX`+7t}uh+frevLp<!`$!&4$
zIr;zf4L}f-Xu%k{?af%a;28E-BdX5|t3&vpas<=I26ppnzDV{NJK6%k!wj=oDPwkJ
z)+g($xq`b&SXvzZAB9p}cl9;wW5B4_lixZbLVXYB++6(d;O--{*2@vGE>`nNT9&+{
zmdUPdVy=b1g|UuXLbMs?C$tbHYdWto-^-w~p@u}Z9<|gXtLB8Jr5(0=RDGT!U#K}n
z*c&^X{gZ_wbkViGNlZ`jNa4Mhals1}>Z8@u<){0WF2{^enL{^Kn&z@C6Y?CKI~1OE
zb=@{MURsz04{kPh67It)L+^fji@DPKQ~GTYELg##gXzV&&5Q4bk78MTg~ga|xMXDj
z_h|}k(jl>B%GJF1Nzy7}ocC5A;j{{cpT!5!3d&bWDVqeE+X<K#1B!JqpL|e~CP%=4
z#uduY7}Cmh)zBtb{?&K4AD{h4-K3V5T;|EoAd5HZg+m@q*lLOm@C>(-uzQ*_>K}Jj
z$(&yJxI^j#uR<l}7^vGOofXA6>07!W%EVo)^rL*6(D{B^=vHF2TLG9u!Y96SFE6hf
zf~c0zJi7$`@-HprqLCfOUYukyCyQT!3O_$kPez#bZ&<o)(e<xN%ICz)5v<_u*wJh1
zqv_MxVeeEWT!dFv^CXqhZz{@_M(2v4y4w>>guw15gXB*S@zbA9eO;W6Ig_%>@X5NA
zto^z!U74X=4XRLtnp`ET+kgAnk<RN+s(&BZ>~y2#3prmGd3}Aqt(mzpz_-OuON7+%
zlyMZ4jJ)$=)R~aLF!zVg47K>jn#6g0*;&uS;!CCnyMpK4DCN_vVp?{B^CGlS9(^fN
z_-eNI$OsQ)>`JfxvUV;nQ#-DQ=qu*E?BV%^OC(HH$eRo*X3=3yRni8cA=j~5(PR3K
zos><6#Ay;H=a$T3xN%&GXl}(F5r6ro{HVidWo9PkFxtP`H)Us-3EiQ}ChnQe3n6rw
z6wG1_%%Y2-Jwe1vXP=3K6yuG|ETD6|K=D2g{%Z^CF~@ci-2~g{WFpV79@y7cf7PAE
zH|Q^`=#<;(kEB}*?_g<f+TmV1E)E@Yb#OKf_126}40F~r$b$-^_N$*yTA$pjy}b(x
zKI6<!m9TrPBV}OZO!9dr!Y-fP?;XaSzg!;v5_T~Sv!Dlr<O_{j92So~<e|^yF5(Pp
zS5F#6e(sKwo_rAeRNU|Fem7~2_MF2!*D#9Lq7F0sna}G5>BgYn22N^8WBBxgcK;`L
zezJiOqgmoS0zzz{s_BfYvjeEom>QYtv>BfKvvGGbtA<Q!;AG~Tn1DE1JQ!OXmeXS2
z-&JnWHuP=wprKdC)<wk0Xy>A63=D*@+5OTWn1HvgbX#fA^nt2s=QL!^6h@1KRtg0j
z8PY)L#EnclQ3zCb3%-H7R4(BJ2ve0v9ejZIOaOPg@2j$w+Iz{(<_99SBFuc}&SIys
z;4sBr$BlZ`pW_^<?$=Z9(7rF*f0H)=u_Ra=MzN*`jVweKEDn9Ea`KIify>83cC{ch
zg>PTs!t|KPNKeLW+}}{S7_De^H~T<@f#-ow5`-N5NC53IReAXvRkOLpKy~I_?&=XA
z9mRB6%B=rZVDj*HLF}0c4r)eL3yk|$gJS<P*KV7NhYLEn<lWD!aJO<V>|~;;bSumc
z*lz4~L%Zss4WVi<`tg&B68MC9-TQP{pveY~eg`vnlhsM!d8fWy)k%uHkYlB(c~a!d
zC7ecl@Lp|DkhCJ{D{n{@|6PYib-!NBN|QQR^YQj_3$0_>fa3%#|4&gcrW>5Mug8Lj
zdKHfsP7fu~tUy>`##_{G2qwFm9P+LfX(%dK*sib`@#^@C?Y5%JDay3VU^3m*)e06H
zv<qMA_l0gcbM@K-*ljZ`-$p<pRI!*$h6yl?O)7#=<Qx^RuIhtdl6w!y=I%*{LcuTW
zGE`>gb_FA-vDVeS^|n3}{6oJlE2=0wReiGA^%5W?N45t9ivO$+9Q&Hycp%PP;mX&~
z>-!$QAY^PQWDbQO>=M1kxGxKTMXYLXS3{Ou$gEfiE@=hLXN)Z@UsuhW*4Etq-k$P9
z=n4k}4XjI!x8xn{W58aF`)m&~8=k%zeYfTB@6MZO+)--4h=!CvfEduB=Q9H!;?)nC
z$emw#->sZeZSDWar4W8k<%E1Mb%w<+4!K6o#nK`)yvM0c^h}%nZ?#!puZ;NMeR53=
zfmBI`1d??9Wy?agTwGesLcJwps6mK=yBCa<4NE`Yqh*O83|8*CD>2l|T##Mg(LkXD
zFhsIQq4z63s#U`GEXd2}nf@56zR;miR5t3hBa-JY+%s~7Sup(#ZqtsM*t)~WY<K}e
ztts3=km?Suc3V;lALI!R{(dQfsHhkkOIYj^)=3FO2{ItoG{zBa3#v?aKD#~Ygp%|g
zdgD61K5(SBXtS!UgP}5GXZ`ll1L2z|^Mw{;zS0COzRQ$2q>)D5MG;4DNV%N-0x7!)
z--QJgk>t}QVBX{f5JgI&CoTB7?fKPrUTUn9<H3Dr8X#c5Ksav&TnU?RujOeD*zLU3
zpg2HYf2Q`sWB%i#eacdNns=0C`53EUFEs63%4Xa~^?NX4`Z`tn#ZOlj%dV0m9(IJR
zFSUZw!8Trk7`o&D5~8ita(Z3zpgMn;zisl>gom9^ElTb@Fiiazje?=IK|hjEQl3_S
zz3Oxl&p>6)oQhz5SjXoMBN`{vZ8kzAOqETF1@Ef%9<_>l^7MOG0$d=Y6e$GRvuDFT
z@2+@j$~)W}ZhCXKhCd_i>GULcfY_3ZLMBG6?EO`H=6hW>#6inF-WGjTA4+=wDbk;D
zfteR{IC!KFChBz)+ulmceaRcS%Q`S69AL^bpJAuopae}cQ48D3_a%_LVTsf##p52O
z)o#z9O!3jL>r0#+?2T0rh?9*m_T6%Nh)6UIy7um+xQE`phj;DC0SwU;THYW=6>Zo+
z@k|U*t|ta*nsgdXKb@=k{4;lB|HDI^E7^bT&-%OW?0@zDjI)p+U)ey%pf`c027#sy
zU)sb!={Ge<RM~8%*)yLs>;==OV6l~Qh$iXaL2~N@<k7xUg-;U&(xASk=ym&bh`A9+
zqNKC#K7)W82a$1z4aAv(P+9EM`7AaNwZszR09EOV$55jxjZ$m?H_b@Ko~5vXH^@!X
zJ0VGQb;dCY?R7x$0!z5RG{A2x$Of?^&Sp^$ZzFr_guFN165GK|on*$SVb`7P*AS?_
zd9o@SP?T)P(s*m23a}sQn;d{i!v0-TLkTs9#oHW|MD1*mm|v;NQb@^HL6vK9$o=yU
z2K)k)|1`Mhnl;p6B=$mpeV^}EB=vD#Xi{Z6Yh#_f^<{u#z32~)Vj)j|W&@TursyDj
z3l)W>Sh9gFV<Ao^G(k)Im7f3xLbQ!H2{SX(D0=ne_ah8EeozQhlVqULDo_ZdZ?&Q_
z);XAh*hWJ*)Tt?n9o8qT2W609dkC}HU^&#!f!$VGYK2na|7E+wItQbH_=7+}q-+@;
zhqxF^JTtRF`X75qiBLC;c=uE|0}fIplivD`ExR7PJMP@Y^Hoh(TxFw)NdAOQh4!Q#
zc*H~v(^cC#pz)hPnG5y`NqI(7*bz%Kgvh|DmH3T>dPBT-ZD~zwYjRhp##+kLy~miQ
zSPQ5K({i7sirzzsPvlxMZe`&Fvi9n$9KKkCZ1CNB?s25`vhc}9o=OZ9_8X-!8tV5=
zq?e$Y#qWdO0YM3)Ox?UQn&KA<#tuZ8+=py49_@s@i~X>Y=wH1RYEWVIa_l`^v%dJV
z$a*W*_&ZtQxW>`0+m~0(@%<?9=w=&daTvad#Zgux>gf2H7+My)=_zpP0k+n6&FM_v
zPoj2?pbp2=73GKHdfi>iVwhJ^a67_O5+rVGUJ?llQK@3tKObiJE83*Rx@DpKNALsE
ztN4+K=`&dF25u11)L_wi$J0%E{KEsu9N2MK_=e+O+Spa*u;5`y##3wyZhIYi96~8l
z;~yIq4smX+6dm<b9=8j~|E5DQgO)VeM7oz@K7*HSFIkX~{-E@hTVi|UAn$EP)Mru@
z+`#0UE@Kr-pB)Hh!m&+76a;hEpYm6uGmK&Cxro|<k@M+&Ttk7!j~Nrs-9yevM8Tg*
zuM<W#XWA<1hKv){RM&W4--ux0(JS@NZ8Q3f_;FYQ>^-azm*6NkfeW6#9JHB%%(`_2
zos~@2?fiAC6(fV`hwTU!8tEs|YNp0UtbXXE!#+&-jjA132FYw-2tt19!vq!|$24{{
zyPbYy>k4Hv`MFD<w+G9-ee=2E&nUS_l_difszC~*Lv55ZpCnthFRh-u^J*NSw>st>
zY{&hGF06s-%W8tj2_;POGgL^8d*kVz=i@dzdG0|~tnD5iI5mgn3kK6qAbgAhr|Tm9
zoz5Kg(|SAooNNAputD~gFSH#f2;AMg1N9I6;Fkuqc!y5R$(Wki0jmUEsPys;Or@$c
z6Lz8ZV@UyGJ{k+zrfHIy-a^WpdKVdpVf#5)TL5^J_FQ|A9N$b3X!qsv@7!66p6)6r
zzhfG<Ee$Z?VCa;w`ZP9rJ%ImLFK^JdBO`v#_j3Nc_y89^Z{$lypf$kYIUx~n#<b5p
zML&^5w(VHwH1?&~kg~|1sOcoegVFg1YWa4H*$z5M>rEoAP5}iVp`Ex((aw>0|9W8b
zfp+Z<<)zpY@<nvNu(~FU0rQ2CX(_1b@v+zc>ldXH$5urxQxj+4o)I&z7`G@^JyWf(
zTYg0Prc>?j24lGv+s_RxU+>4a`0lTMX7SN(;gRo4aCaR`Gt4+LcrW1@fITN|loyij
zro5SN#u%`lw}(=WsXJ83=1-EO-8nweOJ|?77I&v?g(^UpKSdjp;AP|$4srZNUTk1p
z#2MeV5=Vnenuvi}O^#{5D+WrXb@p~E%%KJ#QrcZ*(~Y&uhML5ZuwNeYemUX1wf+5`
zVWXb1-u<#w!bwy->lo$RN-X-~8?WpKf_@QhhNl&+2G1H{p5UvoQ;*<#g<>$q<KQW8
zE<4wQ)rL>Mi<yg-tCIpD&Lb5}n2}MBXYmX%5XFQnk0K|nRgGmH&W!_vr;`ctYW38~
zhn3&LS#GxiBSID9PXF*X#GbP!>oUX(W_E$0bX`LH)T79P*5l8NQ;}T?{Pnn^PE=e4
z2My|`!A;en&oc{3H#+Z2%veymus$`(#3NmcU!QL4YST+mBG^W#ECHR^sbc2!x9ASx
zJcsH}9=9Mrbo?QFjkyb7t>~9p&XKnB))Lfbq)uj1El^PISh13%EpoGO#OZ5Q71jT1
zo5-p6u={v*a1VvTumfwRQ}_bhr87t^4m;29yv&Lkg16M<Ved~IPCI2;RT6X_yHsu$
z8dq=rI_$J_#gJA~(T4#)C-m{Qn_q64A)sP!e|HRee<3K;RC_(Tz@go(=yFq;Dc5NL
z&}RfXRuAcHKYL8ex_Fe<Uw}AzUi0nCE3u&cT09#NjHZUN;{7b4(^a~EV3uI)U*A`Q
zJv`?02()A+2&113OF{ILj(KO@mRUNh5iRTA&SxB_qCh(YW<*k|X_=JsMIP#ptL>@Y
zn}WHRIR|mnI$oVhY*80RTMS9sQsoPEUSrDc-|g49K6*SleWSRZCp>6U#>?J<<#alN
zKkmCm{LUzNvEe;>@ZFL<R4sRflv|DXxi#9y9yNwK-2#2OU;U-`yV4B@@mYU-&m7?e
zE?(D+J`Dkf-%{0DwZV0n8(ltuC?0Ghl$PN_VUpRX`wc*q5b3Vez)@X5?Xi0n7w%Jk
zBEYC;W<wPkIbWJ5_}XnInQw@HtTXIobm=?uU5fs|m0M)g>q$QBR2go!(YZ`o!?8Z~
zbwM+0j<it?r4P6e)kiAS`Br&;V!e2zQeIub#DXqwE8qC%-r1B>2>88gJv{5_1q>h1
zl72eu`r4^}v{S)S)o5XzQ@5MkSFN1OS6(}B;Poj*j6F6}u(+Rh{L$jRZ@;>$m4<cS
zo!0}7Euf+e>}p<n*=%k5LrvA+*nqL>b+4Q?Yn26f6y(q@3MJvVg86M-Zli>~zBYH4
z4WwlE9a>hrkpa2H-^6H)C<vd<_0N-}F!K`M)yXWsF|+TPCAzqB&=0VNO{AD_>*Wp>
zzB1GL<TF72)Tfvp*xEnG!>Q8kxVa55G@xkuKWQjH`9d)TMVS4=QZMs+KqvSr(H8?)
z4e-tQH7F5~oZe(WQ5xdmJgQlVj`Jp(m~Pe%kXbn5%`&8HG!O#Z$*TC=@2@1U#x#4&
z&J5MU<0n`5fxfK9%*pM;J}Bw1elrN%=a8|ehLrPtZwoZ7ZXN!75#rsb_74eDDFj}|
zCFh0khlSH=?J6`S1V0;yv?oHH<t&L`xSb}j1$hw)!^G}cQHz3Ll(boU7y04rs_$CF
zOejO*=IP?e1)Ly5N2w|~v9Z*Zu-%P4gy&rJB@q&F)hTVbBs$T78bmA8J=51d0?6a`
zi*eAqXHY4VyD_p`u`Lb9d)GbB%^8+dnr*vjgANcD;H7THVw5qxqm%a;k|9|`y3yyV
zMP}S;-@G$3$(T;^C9S~D$(}9^d+1DQyk2{QdFKsFg>t@nDq?tfV_av(vNm{Pf6b|s
zl$%~0Xny7NnE)a9X0S{MqtKC+ACyC=_72H{tqmz*yX;ptr{1AG@>904P`28egz8_3
z%YuqbgX`5uCf5a?HXP-$(3X}8jxI0)9c?rt2&!zVv-mlnj>mg}Y~X4RoXJa^E(9=-
z{uEO9ezDZ3?e~oW-;Aj)*XpR2UduRnc6R^ev+z+l71{xU9(F3B7B#{W1Xrp3f;!cE
zApqijs$h%wYXd+>LULPz5rtqK&6c_oAK@m~Iv2w|{<a0(tFhml=a(<#8niFRNFBjK
z!<(eXf^B4#ecK{Oq*=lD8=+L%20L}%vA{f%s*6<qnJ8m!&*J(HFT|7*XK@T+2qc@W
zB%(HEAR@6q5kK_xKQPlOjKzOdt`D5~j6BQ+OnoC#RM>#*I0`Sm1$i&-a!T^V^J!c>
zYhRq`P;F3_aQ7iu^(StlLQuom>Ulyagq|K(4`rb~JWLj>y_dI3&QqaI;u_q7u@65?
z01tF0rQ3j(GsY4QkDV_sYV4g>T@%wtJ1hTGi(}i+^&;IaEY_M*v^-DcwC_`!XtpWu
zdzm>g2md7V;7Z~~_!?mp_PI8`r8~w%lkzU3nvSsRpBNQf77vwmQ|#fHvcZhOWNnAB
zO<#r#wu_+dS98d}mQy=fiDqS6f8M*&CMjp($x2Vi{oxSj^aDHN;0mQeF&EKSDY*Dx
z!E0o#HD8ZCRtGyDD|<;9U5_&IWXZVAq%g&j-PVw-gAMu6&Hk8vlvsljW^h8jhha5i
zg}B^zu%gaeRgG`?TA8R25Tw2J^T3l?593QyMB*gH0={LmB7h}k%(D`-IOsi}H6Mk`
ztITynybBT;gA>=bBs<cZ-j6Kl;^&-EV|7`nWyviPZN3W@TQjtG%qp}H{fPG=?K`_)
zlT_|^Wo8@Kp;z7LT^u7&w8&vV$y$sPLl0D^1|q40arolJ0Jr+q!U<EUsN;t@K81bv
zUAD5Z`_QvIvB+9fF}%!LFF$4rOTf3{6A6^d$i#-DcN8cY&$8~8Dt<ZCTi6-B+-jyW
zi<4rY!DOm}bwsjTYi;X~blJTJ>QdeG*H1wCWEG(@p6o!m*dPdYH&^E!@7-`QTRZnL
zMn=Y=&w;B=Q+UP*-C{pw8X}Eq!pVeX8`gCO2rh04H}gAx3!<gn;4M%}N`08O(GHQZ
zQ2k8=WvxODY|Yf4t|(V$nZyvT`BL#L-VX#S+UU~c#!jg4mMTi{fM7ZelJya`gBVuL
zs>LyPm9#Z*K{@hu3X*w`Ojy5yhg{^}eBRo?+oj4n7N%oIfBj~xj13TCqTsKrgjgr4
zU>hqiR?LwNPcY=9PV}pfzlM5K!QxUrfbpl~)rsx6MVV)YbujuGIG`%Utg<&7a7l-<
zW%11bh6Kz`l2CE*nUBG!Lk66p8rLBDH}{BfFppyXruvRnhIY<7$|u(IDFWxx&AMbb
zGTUqZ;-=-<3u;&eRLI}ntxJmq#i2MmsxtlX^;EZTn^k4cq`;T6Ue{eOg3G&i8}{(5
zoqFKr;i2cDFTQO_KlUf}C<Fth#`iS2>QROp-uID&)e4K!{FR*3Gp$d>9Diu6GTS{W
zLo)!A=tkg+HRNEBWgSyh>8jrw$5VZXqgNR3Lgud{++6qmp_`$mtX=3MkQ(CAqF|}O
zbs-6-U%H}hepM9<0sBp&+z0UqlVy`lLZfbNI<ERdZ%a3-oo)rxjzi6nnlYC}FRk9e
z&!=84*lBGW{Y9Y|e;UNnPchDb9>u1i`C-+Usx+9x?zMNIs!_u(sG1AW%<42lP@6(K
z`?JT0qi|uhEp(Tdp-$xOt~>Z^abnd82a->j#rhQg0?4SR@UlsUcp6g5Hyat-aKs6*
z`+3+>fW#;N9cHVdJ4)Jkq`D`*VR8@$tfV_+-lo3)@wRAR><?*x@dy%UB<wn~K6uIn
zZ78Q5G*_-NeAXkR`zPutO*d_UShaamr(-JBBl(lm`IC%RtQn!QXu0`0YXCaQ;Ez83
zx~yVg|G`Lut?))YuEPpoymBH#8tkWQFiwMeNHV3zTN>o+AL=<&O*HQ@=gc?K+29g*
zc+(1qT%bVApqy_$h_&c7*&pV@D3J-OqKBksrCrDuTz-3kYnzYmI~_{?NpT{T)?2yF
zsK{w9p7Z-Kc<}6XeR8!ngWUh(DS6i6K*&Q=edw&UP+7jvFGSnqdwgR`GF-TbN-Hlo
z`PMR0_G&>wnLv)2F1qOTyG@Au_<(Cd#jTmD$1wy87pz-ur=59w8Rnd+4f0U>X)iz=
zCFkP9wPu$3z@z8fS8G4sA<Z!36zvz0%MdZbZB_wx<Gu5E!Dw%Z43bm+f$BkpEAPiU
ze_aCh5Fj3n&z3KrBT7%4GS3hnV?n1YCgRr|=?{cI3*Y|7k~@6&60MoVy8^|>wFowV
zN`_8EYh#p&<amb2B+F`UlSPn$ww4b-$F;c1_t!hT7+;XCd^52?^#0lHKO40eAuCJl
zXfS}5(4HYk^?b^F_Ch!AY2C2K4VYgGarn$Vt~_qEA>9t!Xk9Nwk#w!0bjL08JV2lJ
zZ&P_LJK-!rPFb0X-{u20;3MiZ847keMW^Og{-R{|;O~<6#^0Yy^CVx&Jwhgbi@+s9
zne7VKD6<GPbfY3n1E;xqb#FUf8@xreXYu7w@g(I@)uS7hk(}n~zsE+MiXUOePH^C#
zRwCuXi0#BQ)!#LxOX(Ko$3AtHdzvJrw?hYvUco4~*ilL7*emjGNH{#s$h-(lY8k4g
zYMNg?@b$H%VcRrl|5GX0AggIo;#Y$dR(R1wejtdv-{Uf7F!ToXaLBKnQ^)&id;n$-
zqyvjZGdk^RBE%-CZC+Pr1W9i%wNI`d08?0v)$r}ZB{O(X(ZG-S^e23R0QntRahm$y
z*38!qQ#o0P<WLqt32I|V4uO}$n(LrizC2odI{6g*xD{bwI*}ZC?WnH3+~3?8DG8DH
z-M#QCf)5l|1?w8R1iSlv)uA}Ff{J747lDYAWxeYqUgV=kr_b(B)CRaO0F1Cgs(k!a
zk`*8Q+yL3_)BD<lg6xTl8NmwAOOM>hZEGHeyqOR+u(@Lzm4H#~BO%pPA{K1-&`Ul(
zf*)BA7>>WpYGf803Ep{Dr}lMz@g>%=i0<zF`X#F`Pp>NFcj<2g2@1-Ng>*>Pj;1hm
z79leV{<8HaqAxQCwq3c+7yOsgmY!e|%an;^eS(5Sk02J(o-B424-!~SHrQXb%wy0w
ztN-c8mpIx+CXhwrrh$GcCvWZHF&Bmo5AeP#WT}Vsoj4tqb8+7WbE_C50q({oObS;B
z@j0EZ>R%+Ljw*I(3fFPuVzf8u*hX=*^V<{D%2cT(8j1yV6Y9TXoP2|_aQ`^q|0-}W
z?>c<wwYgTh&Vt;tPnTzj3lmQvWg~|C#bmk@bcW(Ty66#D-7e=L8jB0jNA^svRKMTv
z+O}8np)tJkhhbDG^#ts3txj?zNzR&R)$ixgK&ZVXt3!G18DboT{}BKlBrJO>FV9bz
zqY)%u5BcLmkB_SeWq%ii4+gD7`3{?KvRZL`Q0rG6)VVM%M_E3;xR-ln15i_%9Dn@L
z5X-me1N?o>8bzoCxgl?%3AM$%ry}T+bd7wW41;z2FMigp6}p}mfeQot(!KU|S1>tb
z)cDlmM!Kdr)LQDkt>#Pd{9GIE({Q|Jc~;6Y=}SSt^P4K$UZAggZ9wy+!05iFldr~F
z2q&czflsAjNHeO~uHA*NoLt);D%xfqyV?BlZ9rn)vG{3xGfV`GN|jy+59)6rkQLqY
z8K2)ZG(f#QG4Vc<;Q`C9Nh3!aH8tP*rTor~g^s08AzOs@x}Z-|+v-S8KbUyC`pi}5
zT*mhn+`MZueK)I{hn5Cz7J%U-?R2w>R@y84XWyVJQh|$sC^&JUnjvCuT9w{9p)qKw
zJ~|V>y!eUs=q0OyW4nA>o#z*eyXWY-3ty9*p;g5LQvu<30N_T|F1+IwP(0cZ5G)WX
ze74Y4Urh78@yVS<FQVWA*r{UFhtl}Z*u<W&*k(I(&uN@++*7|RJe`$~h54Xz#(!h)
z&BLK^|F+?gO136jc2kxlB$X`5B-xTI#n>kyTS&4qqlD}s`bLp0$})CY$1ah5-v?!9
z#xjOkdau6s{k!+~zMtcHp7*b}KOC4bbC~P;T%Ywk&&y|Fr1i=kyP1W$@_Cyi@#9fS
zp$w;Jo@7AqJ(AS7{*jux^IrQe=3Gyc2kIL}l^m2jf>RJY9Uv89b8Bfl*oSDw75O)3
zwv9_G%ypLkS}6N`DS^a>6a^`R6YA`gV>L85fZ4^J7XFD>#uazB?Sb7`TGxm>Ff#Mn
zq<Nw*;r0`ITP3nE$MQltt$Z6y&2AC3X-W>SE(j|;ny{Zckqn`bF_TD=7;KIFk)lMR
zvz~W+5HW8`&?i6WUSWB4lTkN_c(~9Re=wzHag%=|2}1Lv!f5SRh{Uu~?&dEpcU(qp
z>%TzP57Mob6Oa^6bK<E%tdq;7=1LlCux>fQ?)8uJGBR1R;nEfFPM;Er^60N01WPBP
z49d~NMua0HZvcwVJ4$*Di>1B^m))5xKTsoIp)iLLJxHcDvnehh!)Qq&#&?K%C%-f}
z#TuLNW1e0B@yXyiEEJ75j*W=-StTZ<+>Dt`(L3Hq<TzCMO_W7j)NdOp;G?JJ#_5x(
z)i)GzCEc4V*N*p8O-RqGRU`SS#kG-jUxFFJFn@e<Mq<4I!#=RN*uS}saj})PnD|=K
zD!gla_Z3?9I;?+WZ4aNLL%M{IZANkTud|PujkPb96!MvRMR!jxjAWFA8Xo(4O%?MK
zj~az>b3&AzKS7kBiddir^avV>-MQ+7+@oUKg06w~ZZ=-#4rp{-zmk_}&MYm{!bZ(@
z>3bY`mcC15<&N_Y><bl@dJ>XwsX6uBfJyeoTb~3C)%Id4W9A&R6~jfjn>>u<bmr!(
zO=@_VjO*lL;9cD*yfanN!*9=X-vjvZT?yQGZO=65a^=#TdR0Y?<+mrte=r=3%%Ulr
z&CT-D+|+^q9k;+sdSWdXrbOiT1lTk2Tx0{Xgi~w*st3hb8?PsqRfJ4Y>}G`Z*YeYt
z1j)?mWE~$58oL`yM(g}{CFN5#A+=F9$F!HO1MC^W1_y1K(wm)WNrRFGG_`&VK6TEn
z4ppq5j~fL<*7hbeNopq8^QCogd%a*y9|p!`+duIA1?+CsvsRQoX%*)kLE~F_b5cmO
z*RAYRdSKMZiHmJ|Wa%ha69|{IxY{FRqW}?7-ThO2wT@Yqx=%y+nsi=sK$>|Z()_mH
zgRbHyJl}{(&ETf;X}7$Tw=7MPq1Jo}J#;4S$RHwL;&kF;@sjG~ZqH@0+7KSpsgYU=
zH?>+tCV&I`Tvw)*+bkrxXj|G-UjFoea_;#%XX*Da4ah?Q^!WOF2~G7Ui~2;>&tV?r
zoD#|uP2PBTSM+@5^KaKFc|Y7nF~lAG-nxyQdycL{HAAFS-|cGu&M&twJR!VeiMbRr
zv}*@J%m0E@AVN0-#K^aOwsi)@91C&voi(0U<4jO5jJek~2@|G(Kq9e>lJArLP*+nt
z{c(6vSkdV9r^7LtV3hM29CA(UY}ylwlXm@S((x(tHa3|9&$vg-E%cku^`;?o5fpy{
z;#sZgY~T%ZX@8CSepitvA+ECyjUv5q!C>^i#?X{Tk_v0?VbeTP8CB-I&Y=G)$hIs>
z+cjtEZOM^9OWaH>oU#6zkJqYZe=Ra|Rn{q~$3-r}B2&A3@RMQHYB2J9fkpprftuQv
z?VVNfdlk3+W#>y8Gs$t+VfFQZH~idzatOcHuj4+O=VINtsaIZ#yB5SEIaSMy4{pS=
zJQL!^@1{m3j-tA<*z4aVQ%;jW4szresRpm9Uw&eB+d270(e_&TrNzrHuQEJJRM{o7
zkzus4R9;}HlJ34a&D7Y!u@jZaH{)HO&R=~!>uTlK%h0>?zaUNU-$^uZ{TDQKFrh<<
zq8shd3oB(aWNsJij%^to-V?3$yJSg0Kf|sGa?^s!DKm(jV6q^{RVqpRw8lR467zMT
zo_`k6VR(n?8{YK9i?@gKbfRr6<!6O$kX^fIeZUL5><xwlpoQUR5Y-X*(bD{EyLb3&
z)JN=FI6uj1zEMbK_%d^C<ohctI;Qe0g5O$TKd!?-)!N9v(S<BGTCYc$K4nfIg~d{w
zi7VJ-_*l<;QZZRMz@)F%F0*RN`IOFI=*8N@&k1R{Tx7x79ug=38#Xx{(wu7R2g_2v
zX5PJ4M$8OtKYe`jFMWpo8U9^^bczVUl6*n+$creGcQz3;d%2E)bvOM4#Tj=I)FbB|
zN#ZKl;8ayYT2o~`J;|N*drp|!>vnYRl=vIzChsU21&bh$WXejscnI-pe)I^-1qb>S
z%zq@FV<DiW1bqCKpAf?VVCXC*{pL_pKsV>#YX%`d!|)a|BoYZhTi@G`z)CLBAAsiQ
zb1!tq8w%k06nUVRmL<nk#L*pc1bZ;42J;l15s*&(cLUxf`u)KFFsu!bLp}j75`r(R
z^-!Ok#}dsT)aK906>ppi9V?Xr*CgE?J$NHb{<6Io_x#)xCmm&kHUwjIr7==`2*{Aa
zw&Bp_LJ^){kfbN<%Li;$i6Dyo_&o<;0uZFWht*XltIy|RT+ku{P-%mA_O!!C0)IC@
z-;soE(wfuQ%T{lTJnXrf25a%kM-C!>#B22x**=I;;!?Oww+f_sTHQ#NNLa1)+~|Ug
zW1x>Nz<k~`VuBFo;n9BW6b`jn3`Kf#9HE&+99icRisIW?FWxhekVVfo?_d&~i8-`m
zRQx?3@!8b$dGlN|LCuzgzrJZ5v*x_*1xuewhLCChkR_|P2WuPmdu>6F4WotKd4d?=
z+J;j&BWbs>b9z)ib<kCe4}zy2P^G|Mo<)BJMD(76-AbApGTC#7do7Y|?TjorUsd2_
zrV?ylSpSLdCSm$^-wP%^-4CF#zN4#-Qb9MbQ#^h_ZmxoKr(q0DO8(>v)*_$u>~WU*
ztmVnkgM!Jh{Q?N8Eg?ql5S9TL%I8$&eu^X6!0F_HZE@GQGTi!$LzU3|2YS~I09z^?
z1bkSOm2&UTfv3yTwHMA9#_QUuRT*B;?5cG~a!*_7<xc5%dS05exZS#tq#@O%XQ=(*
znDfCUYFNH&KH3f)c4K*LT=xtkhqvw73zB#=(T*gM9}qxl2kjbW$17L|)AQ`k6_3_+
zJn!3r9IFN(T5CmcDSAhJkhgmCxm}XpKT0YplzpKOq1~g~2fe!^Q6yZh^9)-VI&9ec
zqI$-1(f)|n*uLGDsjd2JujjBzWpyf+=6B1AgOB=uis#uL@VH=bkW6-?F|E$}*2<8z
zH9uDu#hRCqZ^<rsy{bhS5zZ&(>SQ@3dH2qDm%kOiy)&85g97;y*sg~_)B)LyW+rK>
z4cB_Ohy%GR_d~j#j!%pV*j{$`e(707I{8TA^nz&uO+_S|W<RemzUcHvAJ}G5Tw5b+
znc%<WBiR!~aNTjvbJ3`oW+AcCN-ehp!?*|GS@cO7;&J-4z8U;Iy-+w41@7&J(^U;X
z%sC3Xl}LYpJ}QVhv;h^Ej{=zOSPEyJ8EukUpsaUgcG|+dNek2+k1_nzGy+&gOygSx
z*th{&X;Im8URraGJ4!B=-KoU4IyA>aY_MIC?rnGjkvA`m*%P{kc?r^jZsnBG33xOF
zeaB3k^6Zi9FNg|wpC#c!V`PZfHbo}0gGyE~-&Qp%qay)pI^sYQ0mp5I_(Daz{3SKA
z)qsD*)<X1&#4*Gur3w>Y&rbTb+5ls(mB*Of8SXcoQ4>*m7mxyq^r(^bb_y5w>|0VX
zS#HHFI_sRL%yBI7)oYfj5U$D(2MdVVky{WloW|r%RyU=1liuTqy&D1FkB;a>xA=bL
zNOEc460Uy;@UJv<1d=O22IpYytC5XnVPAuJU3ttKTXgix$G9xThf;yv&O1QMXdZu?
zlrxtEqzG0{OUHyTma513Y}-OYh2~u#s9GQKCL)aykBY>R^^a7%wdcz6*Eq^{XmO{Y
z`s@ej&m7_h{3k%WPvdnAQLd|g5x?H1LOw|=hp%Ni<2q-?IU2oa(j}FH0oUXf-*~$`
zt;vkLb%8Eygo_B~H~hq>ZpTRzdZg^E@lsQ#Z>NSt4MJk<#*2SJX4K^NtK8hhvRm%9
zw2DOsXC4R4XMlS?6-#6yF-keD;%9E2Yha_rNT6}g>gmya@_leT527Qj1Ww^fYu24z
zg|bo*<|%>?zR5k8HXet|oqvtyd>5TliL+`XWYU<@@d*v}95o5emD{TaJ4I>5jfbzd
z3B)iLiSv9oo<rkxCL(ClX*X4P@Sj(TDmtW1IhEwozk8m<7GWDilO+mhuqBlT7oOUN
z+<`uj#KYRr9G;smg%pNs@8%-!B&QZU-Sa3jEQ#)V;vS4NLO04O!wJy2%66BF!<{W=
z*_EL?N!%k_m!m~*b3Q*Ux--=S0%?^}V01-nj=AZCMv<AhMzJ~{KKJtl2<6B_XPOdz
zFJyMR+~r|+n@Ux(KKxc}JB=&0EtP5bvW9)^dQ!>FXnwA9t#|s5?I?ix>bzzM#T=hA
zr%?x}Eh~J%Gs`2=9JK@5;YWR}6TYgH(UEa^ns|O<Zhdv1>qzso!k{z)%F+H)8!D(O
zTnhy_A$A6wcqE^H-SO>vc<L%N{;aKDX)NjZ<j4=?8iuU33M+-TPF-y~dCToS@AIu;
zO@;WA4K_49Z!4<rHy!5rtAG=vkStXn<=fR#`4@->A52X71!BzOHm}1{VZ7McS7?q}
z1{;rDk{PY1?c{2F(iNw<d7-e!kYo~Ee_-SC`x`&MZJeyEsYpyu7AJpQ(XvySeJ;6|
zIS0hE7H4~!U5@^oYd3LYTk+SbC?6Qq!kkfHX{<Ne!!&IQ`FzylCrwz6EJRzo+#)(-
zrTxzANpa)9LjplEc(|7o95|acKGHaBJi@N(T&PE6&emvnI^IR^7ZpJZlEjzN7<KSq
za?_qxmXtDQZ7<?D4PyLl=Z|VX!Y?2kuE|?hBR1YcP$0hkzfDdt<seL8X109f(ZnMq
zpc!OVf3^~IC1K>|eNX@~<H@v}Fjn;FMF2`xmqsDMJ3T}xIZ5O@z?jzx#mdV}LNon}
zWe?$yA&6Z_q4O^Nzn{ZWUVtV3?j!WGeLg9C?xW_(>d+Lf_$ru$!yY>wpfKw*EGDAb
zQ5|SNR;M!|L9Y|wEnEpLeCrfOuVM9PWhRqH)yFlo$2EL0?R?Z4Y7}}lbLKe31>dgG
zTWgm$K{9Vi)-Cgjrxw5P+7do{&+dt@sEmk!KEsF=14(Wc9a@~>W8O4w^&<BI_CY0M
z1O>n}pBFl-8B+7gBP|<J6f&dcD-3i|vA>6l&r<@<khDKc{(?MkDkjxf+DSUQ<y?qo
zrxniPffH9Jl@sI<q*emdoDf<glk(XqVZT4^4ul>LVS;qx;gpzEfO{$m!P26@5cRS$
z`oZY^1z$!msg9so1xU&y%moVTZCKryNr<4O)bp;u#%G{A%(aW%rO~CJyyyTdrvIfu
zDj2Uc@}d`JD4Jlr>i!Fo(TaLDvbYEUxino+-J^-zumdT?k|4wfu@2_<XA%fkX^r4-
zHN?=0n7x!$WMAxh4|qK8-#p#}9={IDLE}MgvCm&GfgXuYqeaDI%_X<=XvfgMAR^wa
zG6#fGFaeH7tlItEMC8K3Di$022+c&$4h~C{w`hk>q&MtWHot-RzeCjYU;JIlE1z(n
z-?ywOOKw^*cm6AxaqdAZf7HdIm1CF;xHWv-1q5PbbOm4C{7L*Si7b@P&r!}crus(c
z^O3;c34taxe6yB&UV<j>Bd?t_-+YkwR!+giMuzCvA(nM2a8TL*AyzS${vaNYi73PH
z`=C0_`eYsqUhe#w<luVxQK;cXNi*zw_}Bz-l=PN-rKlXgkw2oue`Wu2O8({Bf#HWS
zIHZk_4vpD9-xq~=sRA4QWYmz{zoKzIun}0%s+=!$U5`NDjG`yW%$e4z<xe=u>N$!U
z@J_sWq5wJWyzM*()Fl$y_*$ceQQY{z(3p*5=KJS)7t}^WM7Z#=t$NWamq>kc61y1g
z)k?FvwYq?|Cb}ig$MAEIxeXq<mZ;;Z4>T8%R_I1Bd0p791@YPnk#-<D@|9~(kHW5p
zLx=C`jU#RmqWIL?DI#Wp1%n<dy1Z;|4&Oppqd1Zw)S{7Hld@~<zaaW=E81O{zv=aI
zAiwLiZ$=!nx5Y*ZohIBVz1r#F`J|C8MfmdFva>t0h@FJ_XjWyNS*nfBt3>lWQ;~1G
z<Qy(a3#sPTmtKlNo{42kuZ8)PHh<gvVe4p!A9gK-tT0I%nBgg#VPsCQ&M9)loNvDu
ztttxjczy)c|9hhM+?&Eq(sMfz+>~@c9rPLVEfDTGtMXyi0~v)a4qNAeKXi*uMn)9Q
zI@gJ?vaT`iMO+YhyvO?;HQ|7*)NGXYXqI(0AZ_bZMBJQ~cFob+U?{$gccA}B+I7tF
zkpb<iV4*bUqiHTe!G51O&#|dqC=x))QHBxAZWK;HelgDml5hn*&6}Ya9yt~&a;#Q%
zTVp|>uy9a3V#1MtOuk2V0NUJ#P1`Bw-MF^5(YBQ^VxJ^Y1q)p5Y!vrLnQ@(#lP#2z
zCH5krIneb@mm`CwSP$^x8Dr}};BJ$1lV^IOy|)z6gNAO@H_t~OombexJP4rjAjaT$
z1eVUi$DWuFj4&BpKKS^aBfrJ{=SE;BB@g#<!66Hwaqim_9#1d~RinVu)>RJU7sh(t
z=9ps_*^_55U9n@_G^U!_b$MKOB-ruwK<7e}4`mU=;hC<Gx$B4A6nPZ5rYL~9`MioE
zkB<$Lk)|*Tj=ZeZ`g&8?A)v-i)bl*!2WSX<_Vai<F>F0>PDg(t<lU9!k#t#!)_~aT
zf8|R{+O*eS8AgebO79U=Y98+;CsNu>Kf3x}W0+N-NWye}L0-Mjp?c|2kLfe)6q7-O
z>|n(ssll1F@|h{8lND!s+&ws%GDW}Q-1kT5{@wHYb>P5;l62;z8CuQsT3m#;4V2U=
z=!aMnJI{!z!E^1ej;pvj(C=ur!^#+%q^}ISTxu1P?lYdA-!n9yQrQi7cRvW~x3w9L
z#OrQ8LvvX>&f_OhFlV4$8_gmmR&uE(9n%mlr_*7Jh+lE!)w*wWV=_2y7;6svHI{S^
z5dz22_ft+2RQ#ji7DZ8NCv!QTB92|H5_M622K=iW0Zu#ivCA%@!Rlg+s6}F=8B5pn
zn0lM@HmJpB)&}M=W9==BAAQ=j<G-KzIlhMe{QimWT@oL4a1iY#^IXX_jaW(OPP^zL
z{iFHZRRXQsZ{!Ea-6M%Zpmn8M;4rTFLi(Z7(fBsKimCP=oub%_uNdIxp#_it15V80
za8cL3uQ^Xj-ks3A!HQN#7mdVzB!!bN`dl*jl9-b^JtUQ?t@uzjBPyBZ{WSW_w$+gI
z<tde^4NPISiFK%KlDAC|AWmbmkYh;PM1VAje)jtzXRq?Q$UC7rUrg?2bw$>)tvS#S
zxRpB#UJ*~)*kE=TNOH7gj_s~3ZA0_dv#g$hv-rGe8RuHUCc2F1Rjb@j9GyKc8x`-4
zOv2)g-Coc{Ngjk;pX%;k5JNPS6dmx>^KHuoy?Eu~Zn~zQ_0ZyxT~_$LvcU&6Z`QpM
zMDbs{P4Ye+(NmcVg8S41g6;(rLx7f}ko0jq<{E*G-fzvpJLjAiW+ISy^=M<5%3?#k
zOnDMdqh0wBS5)COI-@q{F|-|=bS^fi7BO2OuR#j&yb?ASsbDNon5BRJO9MS>OM@ph
zqi(weRyhMkxW|`!d}X5Z3;Xn|V!G9n)M6UdJU>7iu~At1T5CLaYrGG-P3d;E<r079
zOms>5(l<>T3#&J|0J9Zc{v&J4+*0k7&(H9Rk_*ZwoL69^voh<B*(fm3=5TRSzGf$e
z4ci&5t#RDou<ehBJZLVOkJxW+R5>ve;6L|WrujfD>&<vJWtHV5K|-nv<o7z%G|EHN
zsQ`Zb$1g3VlcR0Cv4;0t#J9P&Hvi2K_+S%<eu2ZwH_t$q9&0WXynkm<V#DK->Ajk~
z&p~4Xe&q~^#BOT8%XY#4Gy<mRVJ4tJ{g;yReUplkr@a`Y+nA|Dwn_Fp$CkQ9?ADz(
z4mOR43U{kg&Uewl<K!rt(^k5<R=S?<B09?5w%#`JU3auG(?Gys@A(1&v6eFD|HUXP
zLEEUB?Sp91`OiKSXZ*z=6;Q4}_og)_BiB`_*>iC2sYyWfZ5<L0{od5!e&&vbmhvIV
zT`*l*^T$WxENey7@ONInagPeDHurR#SQ*81#*)tHt(Sj5_3W9w38pw~nJ<~Wjv!XX
z72agGO`7UaOD#XJoZFnJnT-wJ;U0QO7*5Fr5TrEwfsfiBRT}A5uM0te>j)nbDW?7<
zRrXg?TdC9VM(!7Z^BchgOv}naDRG-xzJYdz@|v|~PpG`d2}(XOZQ6&r$3x$qTJszE
z_LPpC2%LLtJLM9LL3}+2zV{VVqzqCzVVRczF31lG-Hhb=)8KQ;Ifdy{-^-1L1FhU(
z?twigbU|R<{N_Ge0LYN;Poz;f2k9u^0<Hs?&AjD5bD)t%AmaidZedLdy#@pOkhqo8
z0^Jvqn>nC>Ufd12T_lAaO}o?7PHWsoFKJ(g)eMfohbsHu>n2dm0|8d&I~63=3rn@W
zAhizu1-Wk4Moj<@;C@S}AUn<B6#)73o;w!4iU%D~lbI4U<(>F*AhWpt+tOmGX4E4}
z|M$kX1GKN`Ee{wgSQvSbe`Ov)Ys*n4Q2*({%m1S1jhN-8aCXw}G`GSw(Q9p3wDW)D
z>k851W8hck-n>X4k7s`Y;RfIY<zLKi)yib$?70ZTNVAxq{<A%lL<|FZ)C?388q(<7
z^#Z@i)Vg4m9e%^Puk*<*pvz-lh!U_sq9^bQ=k^)CM*|r)Vi6)&80!3fsC&9_{V1#0
z{t(ZrpPBj8PUJAGDF~jppB_JLA92rNL9IM76mWe5JAoxclge{^7>FS$H-`$7hG(qG
zoSiM(L#w1}5Cq6b3fN9n_`_%<5lzD7P1cpdjyqRgv|8TN%IV}EEsj+AEctT^YEch_
z8SFnVzmc)eHT_nq>JFQZZu}|?*!rI(w{nP{z?fRi98vtG{q+*dh)Yvuxlt?0e!lT{
z7`&_?a-#vip;KKx7e`cUVsSsi0|69`*Ti<{^abH!+Ihz1>jH0pD%sr6_o5(eWX+$O
z+e_)pivIHA)4m$i@VN8V=jzz-G<J$A!M|-hErMczI*2LF)vkAZN7s8F!T>IeE1V#j
zQ`0;*$Ki9`iMaQCW2sYirkFxgy*GzXEo@#Zp`)J;f`k8I_BH(Y20+b8kIvkwK_I^<
zy|eH%Oa})dX5<?jZ`Q;3-I_q(Fg};=Amd`M`B=f&Fo=Ga>!5WOdu+L$4xgHcn-ZpT
z`RG_AUCSn;Zh!<CL8Cg>UK{QJ9RrafQm4{XSv?C@F|Km4l|zb<yJ12H(k8sZ0%C_5
zY{VETEC-Afb?}x2*<Vx1>J?<xt;N>~9v5J7T$EpsY15vZDP3m|=84H0KNW@JZk1e6
zI2GI*jMPB2VJj-bv9?(D+tMPIdQIamB1R4@!XTCOU51L23}BuWvf6}xBs1<hKr|&;
zDeKAQ)>Wsvl}?y*(25D~Ja;f@>)=WP*io4J#Yt0V!iqn7eoE?DR3&+$8pXBWPIg9(
zCvUF1%ymBWcUwziUe1b?Rum|DxZwLeHA=NjNuOcbiVbYxAn7<s?P$&HlvA4*d5=w5
z<21?jdtf>q7#F^Moga6Gz#U?SC^BWaZO3yu%mDfQV_3$FPl8El|IUP?j2v@^F^}{d
zucKfL<PbI!Ho7`n2B*W?0|J_lwe^)f4K`n2ewK?p=R#k>9y`zzVrfhUIAi?cy8FEI
z^N^!CPe+1ds^;O@0RVbE^@x9tI}~9QE<SKP=h6AhS9Vr1a^Asf8qT}sfZ<72s(h&|
zlSMJ?cd>Gwx)2y^z=mOV*_Qtc6?Gl+6)63OUe+5$6VZf|v38%vDGI&g(_37rSIj*d
zjd^(}E1Q%H@(peWXAZ<^^h_i@B0u=-*^X(@0708sy$IRK6hRfaFukF5{cq%>huT)-
zf<AceVe)@_v?Di3VmKH}$0swVJ-*g?4c)r;WvXw}1dcWbWCUH_MULGin)mJ!GMt3C
z@O7yLK{K{Va)?IWqt(aI*2pm;&S;K=#_e;?=84`?{oNLcMG65#!o*xf`Mk@mpXUB~
zi_CoLrws&jnq-6`BX*#p?K-4zg897HGaULjDblE<BH1&Z_(RO`<S|XLm*PBM=?MT8
z#@$5Y?Wb^mFkKqImgU7~m;5Mn`t%F+?#E|`l@XafsL7sBPF@x6#|xxB+$(V)OgV__
zeDU&0omtybddGfc_T9nAb_dxe;2OcZ5g98a*6Zn_Y2fecYPHg>E&eL+`r0js_Syw7
z;$*NOARBtdnePcfhh&tm&GS>M_*|Oy^N+g6#jS2y2@)h}j1qYJwk;ixkLS~IDe=%Z
zLhps8PLV&|ZAHlk@R3>z1Gq?U(@6u`aUbgk#7%2IS*-tcJa@Vc`-ld@V&feZR?4m3
zf{%l83Ape2!jUO~syDSTF<!;WK<=|8jT=52U9XJ)UND?FOZ97Sxw3RgD6PD;6D4{?
z4iAg070LI3wj_sz?UEUvez(_q-Rxn1p}B?B4V_g-ktjF7F2_=!By`rKm3;V!X}sM2
zss5(!?x`@=)oJW+Db2_<d|mw&A56P&uSt8prElLvOh7pkC(8jLtu07~_6>VNrk<CA
z!dsPED<^GV$z;t_)?~0P%xzu{M)v#O=XO6OQ-HpjRutQ~x*3Hm^A1_NUlKpp-G*~3
zJag=}`?2obNk5jvIqU}vpN}hy$LFq5pS`b01o7yL+DB^VH4n3GMnGKa<^$=Qh(BEr
zCZ)wuXwLp;2Z+}zr2!1o+;Zmkhwvo-cD|4dyY+?GuJ}`aC$xx|Mz>fYPC{4{Izr_X
zzVao;lg{7%N%O_-1*L>U5%JVxc|a^qW}OY_CI5n)EW`3T&Q51Zm#p+_o_-h|%C9oZ
zt`0$A{wAK@2oMHfV$Y1p6_ik&^_IGuaNs+b=S1>8{X7sz2g3h|XOCRV;ml56R??NX
zOH9V|BHuwR8|X=hC3v$NNkY64_A>c4w`NPQ+WR1O^Oa}T=SCM;Dn+Jq1B}7qCkLD)
zxu;J~j<34R30@ar_P3NX*5A6n&Z73XYQ=#bZ*I~rbD7n6Q+Lp1x}Vl>{M<*B$6@LY
zuih8o`WWzf56X>NJZ#-^_RG4f=<{n2*DrUp*wgiE2*6w=n^9+dhG?Pz8U&dL+h;CF
zlfg6f7cg3scImRfOGrInThjH=X^I9lyX?KKNynKK+yHT}=4?P&oYr_vwd<)=VBQh`
z7;*jYe#e3yDn<w>fQ4h7Sd~__dv{qiCTMKrr`rWC)9RC=Z$wVnC`VA=W}V~D1r(gW
zyh!l*7l!Rker*g?MQs`vl@---g<A=^wmqBY{6)0#ORQeFl>}#=#Hhllk-~jvYTv$L
z@1cL_xg&mRQaIIUjCG%Z3*p*qL5C*web4Q#*=6O07{A@pxs|WQtGK+5{J5jfu(dk}
z4O2Obug`Y3{_MaPo!!M+qV8|s0Jpe5=m!r9(3xXia=A=pFsM#_$P_7<_by>Gg7!{s
zlfeNcu3KH0=#2W(zM09Y+Z(|Bfn;H$-R~j=UyZsB@U?HLQP|l;0FWrpgK<$1!DLDG
zN_QzUiPPED7Pco(y~-RN=}Ig2Zp5-{63DGZOZDu7+_xh<wT=`d?E4=Il6@94L_L1j
zOM=F>ug`$fqi`T`7~uFOm>a#TW5dcN`4w8*_BW*8vw1Mvh(=IHKdu|}U8hJ4((Qvz
z-H{BuSJhlzY$$_fznh#a({#8>C%ty5iq)lO?P|!kFnmtH<D+OC$1ohvq=&~WkLJ2&
zUQRXp1$m0PN4LHCTjEQnwhwKy8JdJ3_K!`g<rKv;zRFAA&QdtFgRv$hFi!?35pEl#
z`@(Kj8yWG6{$&D@%vc_XFoV#nM`rh8c)g*Ft32KC`ETIu%h9v(Ac*MV(#GFn<Sd!>
z2MW-|g$dnNl|Ccx6VKu$iD8|{jo5Fm&k91oU0TwdhWZS#s1X>Wi;5gz@JQo-jEA;n
zPTzo1cEe2vmcL>a8T1Dh`!R)58+SXuJnMRLUbUJ2`9;aS9vTyjtOsWN5ER2oMN)DI
zmLlfY^{~6Y9lgoo%G7ZRwib*G+SV0o>5(ez*IM+rBoMFX2U)7;Bb{5EgSSATB&z|q
z;}GUS>5WW<^9>KN+1OAO=1P(;ey@32pd$RVu2LJn*D;0jx%9A>U}OxqCj#?aZ@wo6
zp(0aXx?P5ZHcZ&HPt}N-!3WM@UH<hRC(JweYzCj@&r-4=MZL7BJb7zFM~kR>h})~>
z>mA4f0jb&Vx%wS~mi@go?(Gyqzjhj(HjNYe`&Uq2q!t5Cx*<9QeW(7PcFFg*(d#p8
zbE>eC7T%$o#PiX!oM|}3It@Zmt8RaT6TC-tu|VIVy{#>$Em3kuu~V=_2wJWJTQbus
z*>$Xz|C7K@*G)v(-+env$|F+jr*T$I4L#<ifk5i5t($?tw5rfa=RJ(W4lI$AZW~kz
ziDUsEP2ycVX6)&g0RzI32AIpRm|m({4`Zs4?=tdwpE@L9o9uw<@Y7p2#}Dni<&%V+
z7=vBCGkD7`E(m`5npeCZ0@vQ=Hin!-G+^1aGf4@}h-X{35hvr`X2xB;Ao0%D14IlL
z&1<y?gEEhs@oke_o`-=Dr#TwOH`JAKG%)~4#@YwaOfde`6tp!w99d>ndDtuY-hogE
zmRqTONB!jK)b!)bC@%7B8dCsFkWh;}?l`yf?rlI^@`_Eguzv362nD80h(5zy&pIeU
z4sVrF(`q?p^DuNKD`wR1%^fH6&Pm?#{(N3Q#!fW#Dxysmbdx48O6G)kzV=jPsd#b-
zin&F5e=Bt5lJ*bGp5~A1LU6JO<vJE<-l&jtOoyZZ*Ok^KZCCjvDl?e7BmP*=X%DMb
zt)ogThP_A8FE2~NIhMS~@1%35+4wW7UbZL2g&xEF>;V`{YY#kn9vM-4=?R;9-CFz&
zjt}9JB~bE01T74uNU`mu9j3?)2bdOGHo6JEJC0Yov@XvcB4&d$bVqhGw8CF1fjebh
z@-0TlDK4bW8&t<z{}gzAqvbOS^6a`miG-Jtpmqah=@N}`25%9-OHB^A=Hls040JYB
z^9^u+-ep>Wx}}bi1a%>UEoq(rxTRLS&h}}#$?InOgLV&I9`cs2xXwitTqZi^n3$Vs
z$O^DY>FoSHwtoQP!g2dK&F1$EbJ*Z=pzNL0B&~U?KB`^x$=R#ghmTt9Sf{e0WKn}K
zypGIE7^A0oolc{EWn_R<SO4iqX1UMyFYe#5cs9vIZO-}t{n4WjB<ssk!cW!qmP`8D
zTs5^Ve|4W`bOamXSyOnrZ}aCbNTFBZ%vZ$DUV)Fi8c;5Iv6G@RFmUN6<L7Z$a}Nv2
zvC`2`Fgn3rG2#TcNeW5y2T{_Of*|e76%3l)OlAnnx-U8;cVwc-Uqj2Ly&N@4135sc
zTL}DNEM%PI(d_k7Y@Ban|240x?MBy>*1}xfeMsF=aO)l>5$l2+)oI3-8^4f|HWP%l
z8RwCsZ)?V#TKw^f#dWbR@UA)ptpwx8KG2JV(`}48*Yj3Y71>X)u1D;5p51M-jcu!S
zLNl#;VOPrnZ@><gomK(2!thagbr7l?;0>Swa&tHC<fM;TB@A{RyG%#<I!=;lk?%TE
zAM)vr9(`wN9QG9^y&gb-8kJE}l&6CQf=^AKLZLNtfrg|bTVZ9kq2hSVj4po@AU9@g
z?Mz)6@3?k4ifIsXJeZU22eQ<DZ4);$mV+yZ4SC9wxS$U8^{<3_T!GlDOg_aJgX``n
zhG4OZDc1<{;c#c2l<$(NgL2GjnPa)h<6{fALmovvL-IG{{o9!@+miH5z<p2_S-S^_
zouzy7`cT|Ip9!8+`e@Md3bTp3rIqYiQc~1-_+}N?Pgpr-ts@gPn8viJEEix?J|uEf
zSKi?1q)YaES~@)}+TRYKgKz;})K@BJ0w4qq$8jJ{=(<?x1%Fsq8ueY^D|CAQQst6!
zM{KBPNJ6KQ5gxrkeY?H?0Q*VNI>uH)SSmt;spzX*X9phB#EwiE#{Wzmv0dm~09o?(
zKKUB070m&6a66%9pXK1q=k*s)V9wq8m+b<K)U_+XXY91EgD&~u@nXOTv1v{oc_EA)
z+5PLMxuvm*n)c$=c(p8>M0y^~*jefM@^Il$xroa=Myf#!*~&pv8eqZ|8H<C)%*NU3
zF*!3WsY#qPv}Ge9Htp9q<_mP%(n|G0Y=-j-^UI^pO+9jL9efnEk$JU9Go5P(p7h%{
zpQZC-0Zd9b-q?k~kifn6-or6_bka#4IZkw0n9#5dII5^{==Y!kcZW1j3%O<kqFfe0
zZP4DoWF35%@OJ<WQaJ(kPSwZR7~kH2J~6R!;MtvWbtOINYHsoskxqfx2y)x2vsS^J
zWps1Mii#A^JdnGyKUynTsIE!Tkra@XcxY3n0P(~u4%Htg&4QQ{p=YuR&?A3Gp22@4
z2>YNZcGwc{T8BfP47lTBqggj_()_Aibm6Q+^Q#RlUIEZM$kF1Ixqd|YU+#4>9xR*C
zebDAFNp`?UEgHr*Dlur@9Sv^8e2*Q^qcL*Z%u7GnZkb^&=Xq-SaY}StYPnib6u#ZH
za9C*Qz#5#XD_Ab9?My`4QBPQjPf8PaZ!uhCJZGFOtnjWfB3TdT;xt0_ibF%GPb)lW
zz#y&|xqgNYBCZyIVF@<;5;f<;=(IiN?J<A@QGqczd&5@1`9tnWTl|8Yw!#AbxHzeu
z@QfV^^-)@0pGc4!Gk=^pxG9WHw}gS-Xeb3<Nxpisq-<iK^976XL9P`9;t#lSP1N_;
z*%+mV$-@VjR!hrr@ly)@Pq!Xy{pID|!Ae3bGdvGq{y<T}Q!B!8-_K8I4M-23Q+4K1
zZD_m0AbK#o%Lxffg$iRYVQ15^c-^|YbDH7jQNomql_9$ly%pJ+ET`XswznzoSNdmP
zH<2Ij{mxhm0~82)p_?00yE0HG<I;($7<GR{!>QTrD89TL;4JvcTOOM(C5LpCTre^z
zuq@H*MeLCCecWylSgV2!2i7+qz?&~ye@+UXg&p^?tXx)-ejjYZpZ=2mIxG%eyhmYw
zV?_#=%LzwGNF~F<E(%`GdSo#rQ#gxKZFfgrru8FBdLmR#QXC2LVLRyKPTpph5l^(O
zMU6Mlk`+CQPSVF8l{}FAn0R-P9!y+j5o;0vjy}?e9KYM-m!HZzRIsx;kyu=xn7=@G
z-|I;D{3v2G%TTIMH2m8a7KQesZyx-eg|0yMH;Fye_#Mo6GcLtpOGoHEW=u9scaZne
zkB%to;toTFHvbV^PMge_JxC^~Pb}qI>UnEponO5<NqC#-K2}oIrdfd2-nAm@RSXu4
zGuQZRoOb;r917}-H=LlWC$gq{Za<Bj_h{i4<lF%_@hLS(<t(k&FFQpwTl<3#N<+x2
zAoolk`@zc&mLYPFJ-)O_DR8-_0>6>Q+M#$+N<?n(oRkySy>l@^EWi*zlJ-0V^|d1H
z2#~dt#bcjcW-c&TpG{leBNEp|OTStOc3<BW=UtMVW4$^H?%kOmbm1Z0HKFg;+kcWk
zZXU;eoJ+5c>i_4oE{)bxQEEcr!Uu4(UB0B5lTqt>;of#5S08D$p&HhLVMda`m1_Y?
zNeTYEF(zm9n_p?*Qll<4uA76JW)8D{2^`QA!c35-i<%M1E4WpN|NORC?)`-E)<Fp=
z)RSxK0qUP9N`yPHp3-%TVV+uI;Soj7iWiRUwW9TR_4NDJJRNQ$l3q==p+oPnUWaw&
zqj-nU1ja4PgTc_0xxM&4!fqBmwmK)%QeY-6e;uEc<KZd6l%eAlu18^P_?;yH*cwGF
z+loI=59)}wknM=G<}G|HtTlK<U0>-44`4a^&>Dcni<4yX%zgIelqRNq-s)2mv>Tq5
z*iddiQIm%}S3)dBbLum25&Ursp(+YKFnuuK-yv=8Oq5B+A0?SYxoOZ&3T1%e(g8~n
zi*1>CG4f_SHUx*{_Hnw{&sczuZ+iRH@Wqdp0=9mL4Nq}Un^?*yGtcR&Xy~LtS6*Fw
z2fNs<3V4A-4goUxlUGK|BA3nkq~E3WB6gWS`=E7)ak9u!^QoMy;?Vp85gaDuDR0*}
z>Y25pgrc;7h~fw6A{-1vumt$raBea>O!;)~q=jsQJR1-9Y875`cZ7K4ciLe_`|b=*
zfLo+(uo0o4>$=7>I$dP}!MqWf_Gxgl)mBE~g%5#8l!BN@7#dR*j;hy!ER7v2t6`45
zF>4Y0!(79%IC1Sr;3uu?*M3KtaB#7Y;(Y7@C%ARzPy$7g2r69>+|=D>G{kj%*HUHi
z&ilf#7v8d$U828_Mv&k9Q)LCRm$QeGDl^7>BDHG<<oF#n)(SHnvP11y^0T~;V%f2e
zFr+hYqG(X(gGwMKU?=`knM0)PTT#HkIDL0hRj2pD@7!#4BpKx5uFB$trWN9S<;S2|
zNrl)h<IU%^<D`MPljUyN4l4YEZf<9;>N9&8QB0zs-Pmjw@4x-$?Z#b8_HX-Pr$4^r
zKmfn7oGMgFR!Olw5~bMvRUg3fI%NR+M60Y1#X?H3B>d=lV|6j`qGQ?m#lDDw{HBXW
z7SGkdobtVQ+F^>XjqkOxdXAymQHwV6qXiL>!sCjV<!>p)=xTtepnL+$e$9FM99}ab
zz+l$WO>OY<_j3m_>t}yfRX)9YVLauD=AJ5Mphr+++ef^Gtu1@1j2vQQ&~fHULqXAp
zrw|l~kCc9&RUCKpN5aQ1%Q#ERSJPT*((hit{vsm=KpR#kEes_NM!w)K(@#r$(IF+v
zTfuUzr5Ot)v$nBL)at*ht+V<{5I7UW%$eGDa5ooNOAoOdsL-vOgIjc3zo&(k>!kRe
zZx$^!yDG^VoK8Pp0=Mf4LCFpP_JGpdVzW}#39*$@0C9=<e63sZ;Pfb96aLU>{awsK
zevgAB2LDNqmQ!x`qJvpP?&NouQxNZcU9=*SPldy$e|xScQY@@#ysY?2rUXJqk?ED5
zzB{sCO#Yu7EFFb_`UVh9HXi{*73%6*d+D?JL~H(pa(t)5%I6H#cx^5!bihjnlsI?(
z^>R@xBX%|us8|>k(gy3rst0PXeX^Rk%3iAaXd~z6wFb;ovZK+pd%tHgDd-ny#sEiq
z!rdnJ?>HATg#n|AhhC+xAcuqE??@gf;SXN-KK)@`9#VThc;W|_|3BRR@7tEQ(_&^I
zl+)y!KEh}gpJqL>Wr|&T@}&wB82RheJ-iZPzH%M3C=M3`(k|eu)#9^rC}Pt*x4%}N
zxR_NeV`nQJAJ~do><7_uG1mvyVrJ2==KW#CcnfIcYCE24goUWOfrMfCd2<%Bt361h
z!VCTZv%F`9Pl}v7`_0MEvS1|@`u>ERraH+5=XU&CiDNZu^t*^B1)QhYuESm+OBX2|
zIdj-{mmA}V{+VVJ?1A!>A@%xLnpOvkGtq<ifSME_NIp+GbB(*%D}D2)zT2+iQM;-H
z?6n(<8AOa7L`D^KN+gc491}UU%UK2haPsRN;Pcdl&3(-Fop%eh(hYn0@nO>TEtE)&
zO~PwMI`eBiW|Tt0UV|HqB0DTLpsa>-(fGD<QO)I9mL7Ro6;^=OHA3k3ukAW}Rn}ER
ztejjG$mIc_e<78BXd&UjG9v%C=k^Ak+Z{hbbSK3UoUJ0Y@G~!VnNvEHe0$3zcBN*S
zXi@7Z4w6#a_WENVH@%2O%wI)OBL<vo4bMcPZH#8m<$}Vw5%h5>9L6I=^w8b5dWF;(
zGlHq!j*fb)?C4_Q6D|9%Jl6g7dtev43DB^eO|GTNWDSfxD=o#BH^Ses7wEWt2nMPq
z<?R&q;8qpruuQEwzTy&QiQ{Ykd-2l`%LMhh+JEvcLFT{>f1nuLv?RIBVOX6wOr5NR
z1`Aw;ubNs75%%YgmZszDOCL`>=y%I@yIJuBoru;q!-hV6A9!U&$+LXv7vy7KXnvgt
z?oG&L{$Q0f9GTfvyTbv;3l8Gutlk=&fSli5K#!&hekmC^oFYCbGtyxaRycgB5{E3C
zDc)Q2#Kn&3G|G;RW%-4L7rqkIMA+^xJYxC0n0PxTaYs1UAcRh{2;7Wb%S|k!#wkma
zjeO4N%HNH-%IZg#B4~QTE&L0vxKl<@a{Bk#ubCgAa4gQvH29uB@4BNXSI5fpa#vcJ
z0d)5tHA+b?uEri)t*F{Av(x}Jp`-9Gh%9n2oUS@*_<Z^7dxjP##NB?WP1V((EyqC&
zcv*>$squ{!Q6GHxyA+O>-{M+_9YI@TBhdnKJR<$dmptt5TN?YX{y33;&ac?`)H^>W
zTJR41@RAU(WwlkD;{*6|)OF)teGnDs7iUHniE6!QTToB{B%WEVU+l~C_=C_-rH-$+
zLxh4J)(3;l+(#38si71(phxmp50;QYY8pgmI_1G;1Ss)zd!_B71Y`_W0Bzk9iKN?j
zw`#@zrFK9kG4!Cp7gW9DS#hM19V4x8{J9JOV%YhJ#xcB(9&s3d0pIN(8rhEstPBoN
z^^GkBn1TJ}ALClTYkORQ55U}sbPVq3NcQ{f7X=%aAsLsZ1!Fr0E_a%XF2A%n6a=S3
z9RiiG3Q(W)7%X7n6Q~w?u~{*H4@-AgXSL{63ssWXNwlG+04}^NMXn6r7RIOJ5wtf#
zemr8hXJb_j5&T`$XGHvKqwfFXpARKf6pl+DwB4$zd?g{MI_{3_dDEgd?hX7sfC8s9
z6Q?g?WBC26PM`fAf7H)*WJgx?fz}xRM$xH}MVWRWSFfJa3VkVbilC7=x6_|B<Z3qG
zk$T26>XpLWZw9E*ytGF<U*-zm#wX9umX(yUs>C$d?Nm5ZB$$K^QtU9M4%Kq+#zV7x
zPet(ioTw>Q2|}#ELcxSM9BUtUr)qxGbwD{%$2)4CfGI)SBS#T9@3ryX=_RGR$kp6S
z)K9sIN%tN-u~FDphk!d~Xx)zFD*?>?Px=26(K|79wra>WYC*YOf%5v<-`&9fjxqLc
zpW`BxgP!$q(iA~t7Q3#SXR;(DUN=y1rYvSffvY6b*R_2aSyoPPTL;O{;du~%onl}1
zky?CSER8=OZMQXeU2-DirGg+0Xog3sTqD)Z!@>u`fA{XE?h1+|Kis0#ctnke2)JF2
z6maTZyYV@<Lv#m%SotmVd04wXW<ME>a0K5CSK&^V$C<<KB5zUXK?-3NAWo7;yX8Aw
zDe=iM4-b}fI()7A#FK3Om}9#ijsrqBoO~7;AO?VHEl`f-b(=X>KUu+i4ruCCWa8^r
zs8x<s$(@ra0?wRHFVopH4&^o+djGe8{BIHYKVN^=;ExUbv4KA}@W%%J*uWnf_+tZq
zY~YU#{IP*QHt@#={@B1D8~9@be{A574g9fzKQ{2k2L9N<9~<~%1AlDbj}82>f&X(G
zxa*y}(!BkF!N5S<gVRP@IJWI`MdiaBAy>WLq1;>Mr!Bdc@;?%|N_IN$NzXa7Yw^(J
zGSUrp#FMXmua-<jvsaOe=U4U9*ero4pHaS?hKwUePDYi@9CPUu`keL1`5uDI52Ibv
z$Z?VUjH;tU8!0q1n`K?4M1XHeD%c+iG_-6U7b(v8(wu7n+kf;6@?LzR5WZOY3o>($
z_UZf_@#Xs2LYegu*fwAmZ+?C}Sg1*<aO;l!tn<bHrq?~hX==e!_>c6Q!b&e}8+7m<
z)iAJqoo(2_qG)8J-IMZZ1|MgMF5WjhD7a31xx7^9uB?5Q{1w`tgXEAQT%UQC>*E)h
zZb6Vgn;$|DU>!0=^S$+4y)>+g4N=Q9cnZ6?;m~JfJy`IO93O0~+;WW<t@ORMpm#$l
z;(BAXK>uLWZ+Oz5um3x1P~aKJl{#}_-c!^pv21kW;^KhZn>p7as^l^EfruRm+`J(Z
zLiY8z*VPZI3r{4RUz&&C&5GV1`r_G{`&hZ5xi{jiwOC(RndP05h{a68rP#V-ZfzJf
z%!Jm_IiosRlUpHLJJNTg1~NWG5E(;_)hW8SkNXM?OUqR~093)C|9KbV{*p0d7!Eoa
zQE(;OjbZu~N??dNmaQ3fkbMz`k~HamS8=mS^pSzZV5;!@`neZWr`O|&E$eSy(F71T
zAD8+sspVD+KB_cOqm`)Zefk{l>{YxIII}S#)!noASLNV<vky*W=+%l!n3{aJ2y7)&
z^cSSu2-f1w+Q;R5CSqK{&wWWxIWpd9?+Lo|p~2F*8a>f8-_hilh%pV>F%B<Y@p?4h
z*^GSEf0oLn`=7qwD(viyBlR0-G0xY23EQ~cX*@qtn8_mwCl95ltK~4dhIM`aQhS>m
z_LhAe$WInn#I~I5_ngFTB4pKVq~zp#vlRtj`$d<HGG9zsPA(qFeq5(05O~gH<73nh
zTet3_B#wPkryP%pj4B7u)m+j8@Z6e^nDKqFSO*$jvC!B0vu0Ph0sm_w1y5&3lBq0h
zkIT^#nRvf>bmVGe%u-EG(aPD3EbIU2WBhqu|G(PmqtNi`=AHa=Qv`VvRAPwLFpT8z
zBdw(=;%s63AwbZA1muVW8*}J?iRFJ75a?9LB`sb}ZwzCz&<w(v1S;k|CM92LZfTxG
zz`<r;Q+z6us)-cBIErh(QAxBrqkYuq8O#i~Y&6Dyt_4=$Q7gRkjTp7z=o}M!w>oyC
zYv>{DghhMwXu_&Vt<-QU`KBkQF%#*pUl7jgf#-|}qR5sgmFFb~b?Uu^J}H-t4M{PB
zyX;UBD-0huexUse5->>SOZ%qcuqWq`pGMM_4g`mx-mzd#3m(ElRfpT$E%MN;DptP_
z)F5iO7xi`LBo=T6HhV`4tUs%~*lt;Nj270o*aa!n-gT~=?b0=N<kRT9#IakD`6AF%
zW#H?o?NEV8_e0*D&+MePii{kvCq9L!h-n2A>D|+V58oY7yK2u5B3;XG>2&-c_ZL%~
z_#`tf;{@XbzK3ucm>v6{UB$n3h_O!BISms<azB<H8!SEXSt;Q<vpOPJzq~rZR-Ww}
z)ho|-aINuPsleHVgmK_pLis!W({?Y3VmICmP6065>raB}V*lXHTwwaA8<Jlg(cv2G
zrFNKV7U$-sXYJ7&&a`l|K?EdYhs#_7)DBN-v1ksC$f<jDLho2I-*s3|uZnR0Yx2S;
z=MaR$=*8lrn2A;I3l}=~BdNoM+SahKhY>j*c^M_r6m^!r*d^4~`^&&9(7R7016iNk
zdY4WRQ5PkOWkg6`eEz>n`u_jn>sVkMo?*)-Mm2v&F{ZkFIKHnF9u92CQuO~kbNer{
zU#_NAB(E|&W3|r_Dx&*zOG?xJrs(X$nxy^L>xy4zBSOR^N2Pi%ox6YOi1-JasDX!x
zF1UDvg4SS-fEWFldfzYQguxjuLpgu)`o(^N_CD6@(;$s&2hhb%A_SCPHLn^l#d!CA
z%ztbq(*{teR-$2=MfORi{P(>KYpJYW2d_)am-h^BT)L>ym8P#y$e>DsWjdhFDIHi?
zq-4MQE?^y&bgkGbqT#}+sNg|^<Nhj-y_S67%{`Gq5(H?A*3NhB#C+v;4YtNg9tMGt
zGM@;=>|VG2$&&8A6Cr*-)9lli#{;6p3SM8(YInD_U=MWKJrNw?bZ2(FKCI^#<Y)oK
zIPHY;l*#m!v=aT>-K`!I!B)6AW>ZX=&R6L+`?zCw+<S)0y!mdO1`=kKM>jTn!Wz@U
zRHOlqB&Tk8&bG|q{k`&rgm|C1I`5XGJ^5tTwUf@(KegWKyS9r8?PjZ=_?Rwq?+kby
zx)BB;VVdgCRHs;JvyV2G%Uk#cdj(1k1_ZcbacqP2t=M&qGmJl|YxCu9eK#F*H52I*
zr+PT3X@$y~9)X`ee%TYcC!3Qb&XADmH<If!Z}z6So^41@^2w3bTCw@BUnj5SdF&5z
zCmA1v)HHfB_0UIMcD=~T796GMdSzq7H?1)(Vmz!xzB-gRc5mWYwQYky?5!P{6F$#7
zn+|RtB&AD)_Mc!b;#Ge4G<OHdv@FE$JE2n)+TL^O>uHOl?YGhi^o@YG@Xp82QfB?=
zq&DUHFNl(G4v%vlEB}_Xk6W9WW@NhK(>zRB#_qm*#JIr75qttJlOjK``61P(@Z*=)
z3h~N#*(eVKiZj7D9fxKL>u-tlmy9maq?{7+(AI4e7+`j`^<(8)L>{h1<dd44-*6`v
z>GyR#xIuS#*7IkLphwFAGBh)M_v+WReq*MB(<{0b53_}=&AQvNxme#X@P$<OxSZ=@
ze(Jg8<Wvk!%kAG$>+i33@yhmRyRd4y8L#+^t9M&i=+k7cT^%kjtaDSLhe^lUSk!U#
zC@YKiw|*+~mK5&TGm-Bd$cw0~O&%)cPb99J5mQCiAE#-U;Zjc;3@asn4&QJy&^Mn}
zds-FUd~f&bfY~Mej0lN;k-9Qa=lt{Ye_jK+;D2&!l>xfLuxx8Yc^HjP(VlPtM2W1R
zjQ?90>f|j^C(|h#<fbgd=$ed0@o<uHag|O;bi|c`-WUnXM@Qzjo@+P7<n*TqL<)y$
z9M8#=5w5MAUoWgt3A24_nscLnJx+QwR75DV+%-U`cz6fP(CAorlhDJq&!;)r#n^ME
zgqZpZA{|ljMiCa3#<e?hWcAfD5%480o-3L3lwOgw%EVckmTw+kJ_wc+nY<|o6vgBo
zkktovwI9T9ePhmf`GX~wANJlnhG#ZPlU>T8_J~2y*(lqr2|s$LVda^YUl97IJ)Az(
zx?cVrr@hB-pAF_UX#DOc_P(gJEObiSF8t6TzjyZrjoR2ET|BuaPMkH=f^ZBz>&VWC
zNdI?s{}22B$M?6Qvi}cz?;X`tyQYm}#R8%fse%ehiGb336r~GD??gaq1O(|NQ4x?(
zB3(g=(pv~f@6tO+jf5V0PpAPx{PsKV%$hSZXU%!f`OP_VzV-Y5V5RSb=h^!%*L~gB
z-CYqQ{@za}NIUgCO~|?Rp?sZu=%;lhDx+kw9mzJ3cwuBXU+As-82?G;>mZIrX^ci{
zrjDL@zT@clLLCRm4swp9HK_Sv<KubzgYp1ij9V8lhY;&LEqn#HjDB9Ox?E>Z{o$bX
z-l}D?d181S^OAyCMHHSg9{!>CQS`O+<3Omivij9bvEf}AT6P1Z>7d|dYLx2iNAEMy
zUgh&fjtPW}k}HZeDr|4L?%%Vd&Lx`>^}!5wgFCAtBkR1Q13;_Vcd!q`cl@VZf3eEf
zLe5BdYBJ|68}rc|&9R$$KKgS$JC<5J=O&P$egD(2%fRkoymc0?R50LyoD@Bscpy{&
z(fcG~I6WlU>{?(!j;rTBgu1^XFTU}OQFOyDlV6`{Md^XsHbZ|E_B5gPX-~Y?qKT6I
z=&Iqq@#zB>_}gK^L-$M25S=)S^hiBzD~D7>bnyQpX8&1YI*Y>XRl0$~Jn!ZSt3yL3
ze`EMeS^}H_2u-V0|1~~+^|#*^xB~6QUq~){7d6J_&BDx8<}+Dxeo;XOGCEXV$Vfaz
zm_!Zh#AsU?vw0&+1kFqX1^S*K{7YE{BBusd&Gb|_zuFOFKiDH#xh8Jb_yb<<#6|#k
zR%HYr^|yzxj!uUu<5&Eijl7JDHP3dM&{iZx&Kv+*G#8KfSQDndIyCOuwVkY(dJW9)
z^Ms#l@>Ix>v0f71VEj}d|I!kpB&jLQzk9LT7&sS){&1%KrzxsI#-Ov15WXsAQnE7q
z*gW06)DCf9Z$wB4?U7vhJav1MUc?%C)brCQG=Mj>O>?&E?NG`BR-0x(n`T_NCx!|W
zn2wXAr=syIlNOK5*;bQX6I<hn;o-CRdGk_6wchLAbJ9%BWXr+bb)!a6+|rH5F>aDQ
zC2M>8JyrqK{p1H(OD{Xn+Cr_*AD$k)L#?PL*WY_^a7Whh<<L@gX4RXCG?UDz+wW&i
z6|y70<oJ1_)qtJ+0GOeFKNBpz4QCZ!KRbz<aEVt(d{k01`Jy#;%^;=xOnVcTATX76
z)ASlNIIqmFWbe6x^YVn|3;o93*CkHgGAkg3K9e`AZvx79T;gFh2TCnC)lsp%s|dav
zmjCBu|3i@R_i?UK4ss{8PrkbF6%8Fx;4Jd5Rh)62dGM)dRp+Nq;^Sd))_iAS>x!yi
zZ`J5B_3bEVG6_Ri3sC$qaxQ(#)#`c64JnolT6=y4vo;;{u8i}k=Bt*s1_6B#?GM_|
z+fKBheNoNAuen<-?}qD_S81s7uvW6jek>m1%J76-3^>Kxv`)RIYbo9x<^2tQ;od4@
zZADr>T{r34fn~6>M2p9S!sM_;y780i_R_qknCuWBYefWK=AQ#M{;KCDk!VKUQ_Tu;
zKCUf(nRzz8nZ@#Nc}`#E5PR%%2RcmVw(-v-d9~6i&lF%1?k#k-NsAZlvSxNv+b}i+
zxTSo|Pt#P^0_MC8G~VeiiDauvgw%YAKsKTvat<&~Vo!a`0jF<fP;cPYIOd+54aq&Q
z0`rvw$zWVq47_tYpHXLL0rB&{m<rRYe%*WdCtUBzvlTC8$LYr>v3dPmeZK1>Oic^u
zY*Y>VRXiSvOevQy7`9osK6|!8z{>CYh`8)$119x8!}98-F?PVal{WBIREs@N@&_(Z
zT_EuKhxM)fBYWij+V46={kvasAtmR7)f+KTI<lC1AKprd(MavDI&uEpY`d+ZzNVix
zVt#MB#?Z%ctCzX^MK!X_>6)IlrkQ3%iT5*;_XVV%Z$xB*EbkGMXEqT5O89V%NCC$U
zL6$en-d!%E)7V0i*T`1eqvG=Vc)RK-aU;*(OJ{6ZrPW$e(YF36N`%en1k4PQO(SO<
zQsVWa{jra<<V6Y=+*@0P5cmpAC|7i@ZA9>)PycD7-2cG+>kOx)+ql0Lfbq?^ZWwJ1
z8_!v@5Xi^j!gl%ln--Cb8SHr{e{8{Dv4w_t(q3|14}ljcj$Pl5rRQ?jzFliWNRPZK
zegVOxa-W)MX6kC4fKO<Xf<*DRL9`_w_jQ2^S(l4mvjas{X=i$JOgonR<Cz{R)>PH^
zuL}t~_n^n4Lt+@ng3Cs$9-BEglo<8?%U)A+1O*wb?dGiCH$Mnlns5qcqK|ln<`KjF
zTtSOh`{(n^a_otJ$!CA>E}s{iCxMG>$)k;-ub<#W%%m{6o^{IgmK6ZeW1F4EwW_uE
z6EmF5CYwA_0!jg%EpF<cW~#(>wXOa5Qor&2X>j@f$NLSx<HTllzsGjCCLjRfC?<W-
zQDPdn|5rNaSA1!^m`g;XHd=@Z>1~p}pKnwr>mzq%qnDlX`XeuM*!<cZC5x8fS2o0x
zwa;5>d=)5NDS+AZ`$umgrt7Gj6nHDkt2Ir}v$gzeFCMowE?RxPY*|n9W(6k+3XVU;
z<f<sc=HZrp?`@>w^`+dqXLv96sL$3VTvK8+Z2r3j?Rs_nYo!vA1?aSoX|mBD+-c2I
z;|tEnLk&Y84iuDLj5v^3vM7s4+7=G}Ma7Pc^6c2@N@k!zi9YJL4e4K9w#;*LWf~Hu
z>CZ2cYT}M~1G?B$cd|bK6zJ`nW%|ypO4_EaYpw3hx0vJ@H$OTJCl7@ZBi`BiFgD05
ztJfH_?gTqr5&z<~9@nMAo|IavYwK@U-JGv?Rt;~1#<I~v)k=&8um0Rt>E<y>>l!UN
z@}-|y4YXdEItAOMETh@s^m!bboc${`BWIW*vIi5EGHypz37U>Px7MuRbTb@KigwRu
z@k7iM>E3ECB7748fR%5J!rywBv=&{J+@Gu(Jt1)1{YB+qSI~ktgzQGU9C8@aHx8LO
zK}u8}DA_*1gM6jlv((iLSM16<!q&cw&}T_)AQqHz`SWf{DpM~x4uAYFe(~RJ?w_q{
z{EhbepXOM9sVZ~w-*WrEbX)n0ip^(Co1hT2j~#f4oM=|qMmbdqhm}-#zS}9QiGWCT
zwp7?J@8(iO2$85rG3lML78bX-SXRD?clmQ;+KFz5aey+BVj(FgJvmQrUMrN|tb7sz
z&w!efAnG+EH-_!LAv8$JLmxxAKaIG)R^S4ZzX#8hA`(Bwowt*JrckkNA<vhIuiU4G
zE&*=H>H0q?`Z)E9^pOzLg=da;TL?^{h){nVP5eP&8X`fOBw!5vGlOe3qZ~F1E*pWG
zPy=Um9JEIPf6+h4QSO}q@5O4X-8R3zn0qd2PO{a;QGKR`LoW8^rPT=EX{iLU=W&O_
zI07aVelFnQ7t-MZ<#0CvC&AN>M#;);fhMZ95&HA|pbld;4XZzz{+*glQq|JwL^&06
zV!XM?Wp;<dG6Ebnf<G*h@A<?Se2)=Vz@wY+QXH>3I|KcVB$cf0$V>hxm5Az<SDe(Y
z7D2UUotET|Hs|Zy!Dof+`FnM$em~ka-)e+|gmK;~j!G{tlR_3Gj;HC7TJc}0-%SVj
zzWTby3KUmYR5E%P2gR4NZK^_S&pgvZ`6X52pzc5jk^X=CMfBhO{=ex0i~ekLw2?#E
zK0#h>0z)ylAdTY>FpjRg9=b8l>O$+m{5IwHK_M<aPgHK#U$5hajn}i(@j&T(J@H3f
z7oA5DtwXOFA&(=l;_AfA^`VM}_G99T$R(BRIUPtisu)9NEXMOTY5Fw|wwlIKK!k?4
zJF9L%V)yp9h!H3>td*qQ!Um75k<41H(r*k@IJ^av1&UbPh79y_bD#4owY@Y2;=YdG
z5EVMuP-LiH-)S2G=sQw_KgO7(fY&@PsTQH$^?BvP9HNzUG218MsA=mLm5cI}z21jg
zAFP!ZKOxoIa6&MROE~jIaqZWBrPyZFI^Qnq*^~i5s@-8cpm}j4Xw`C_5vH%29Pa$2
z8maes<g=2L-jmv@AZNg#n!fu-M?|2udI!2IK1ij~Vx*dyRk7E>hV_c}z7NJKQ<KK~
zN?iQaNAe%+ia(jBJp=HMjhCoPp4ZF_2%kUq7GFjXhv=yQu{V!lOUmIjp+^5oikZKM
z(2>f0P#5_k5jWM6B2YPCTXIjiC*0(+`*+U?z3#6L9V@bux$@S)N{9Wm*cFp8M&LzY
z?i^d~aQ#WmrsF!Ncpyr(t-4s^I!20qak~H0FDhR@!}zicSI<UONk5}&Sg~hO@jZ9$
z+$sw}q^Wq0Bvf8>HeJA;_-$~m$4CYy>`KLF36#OI`O79PStqB=kL5}aU43cWyIs)h
zNlBHaP-O?e=3f76f9OB^`|ti<Nif{}ow~4B_=~DZ<qqXG@V|PzXVg)_B69%aB|aA=
zerNm9D|Zdmv_mXcLBKNtI3)ev3hdUmlJ}m5*6?@_ou-jy1=G9kFtz&jUe6nvmd?U@
z-%@L!CdZh%s}96!;V|?j59Gl1k~>^|YyBnV!o3-GHOp(@rlgF1%UD;<&_nDGs|p*L
z7C^S{qHt3AqRpF3+n@}hJ_I!v=&XExNfdH^1x3*5Q45FL`}i?ZHgSwxbI@=sc=`+E
zN~r^m&)FlHagFc@fFI?s{W^qwg*>8OMBuafhi`oMF33tyRF;_&j=jp{-E`!e@7Z@J
zi2<j`D@+I4XfK77oZH<SV39orf_F*Xz5y5bUPJ1itzJ8z+!^NSGw+2i@sl9PMB=^6
zmSM$x0U9c6gYegfM9Fs^;qYBt?Qa@7^UMcH-E}17jHbA^HeOShYt15!>j!ogP_+#!
z4b;22ty@{y2N6+(-pRI!?Ks3@G#WeMTl-fjI)KiJdskQn+$(Vw(Er(lPiIZut|Xy}
z-b$Jm7LT;m2G=He?v}Tof9kuQM4EuHuWk;s16)1Dx(G;OhR~+p@-M1)I`QSc*yN+c
zC3{MVweGh!WbMz`8Th;8jQ0SEqXm`;Yn&>{3ZYyjpB@Ylo{cnwS}3;6ZKZ7`=`wx#
z<iZsk;Mv3Vnw*!53jQcA>1miX#YK74X|5TbIEozqMb&+*0xoN71m^F2SO=E3TqOfC
zW;O%b!DexpDPy|e-zoYqCO{MBavw0B2L%i~2XNnyzo_zk`sRQCxHfXHlcd9|39|yM
zWRH1S&XK#102jlU>1glvGI$HO^p$!UK^{%3wH2mu&k+<Df=B_wwQmXWw0K6qxwgT^
zHpjUN%#$kR6#6r^Zf`JG8v1O^WZ-ScP^@Z>GoD>_@LPiZsa;IFVrpd2KoKhVz@G%_
z3G5?>Bk8S$^O<KU_G=TUX6*}8Le|9FiO4++H`hqXO`=4Y<RRKsGx4}99XwLcBv(p(
z{ps+L{7ap|gR<Hb8n|0KMj<OW?ibbkq#zkz-@Rn630l=40z%-hL6vM%&2De64e;oD
zwvQ%|jc?^Y7V6p2Fn*uL<`b@X3-udI#nGm~u()fK8@N|DiBb6<b}Gadv>V5bem1u)
zeEd<7fHHO61V0VPS}iKtNyiywx6MFes=U|biLzR8VDhcQ`0rc|V$Fz+68Kmb-ZCne
zZqIWMytpFqoKm`gCG<&Tkp<o-_F#IhiII@Byqijxieur_dWAc{bWToj2<f=rBa~~W
zl^%@txC$ZYbd|ON^G49CbFaW?pItd{A8_Jyr62P|WiS3@4b0l-df5-*G=?{*%1M5&
zA!KH^RR+&;bz^P5P$s7?mA<NXL~N?{>9}{g9T>%%y%oh>*lz5cG=#^>q$NMKTg$~X
zAF3m^*ul8L8O>>PpR!&FK;Ug07GG|E^e{l8?t6I49*y>cSPIjuHea_R*~bxc<OP}Y
z%(r7R(??DsAEp-V)Uzy^;&Tdz(Jl`?#;$YqmhSTcI$3Irzo<;NdCw096l;P-b44QJ
z@f(7SZpZUr!VP7C3x7`w%g*B{=FIV~AmVOk({-fZJQ++Fd`Rdm_s%sj4X&tIT$tl6
zkO$?aX0%9pomF`!XOiJ#al=__{V9V)^EEA4d=5?jxZ=0wRi5V}tXdB+e53{fY_rX1
zZ+}lR;#>B)AD!-BE=~Jkzxt>&r8Q1+l?*=GGCu$X@nrEN^@=2lYC7nZMk4rThqzi1
zL+^q5_@7NR|6Ka{r_aqtr%4+M1V%dGR<T#7sgLoL(vM1yzOs^49ZQRbajRbYQfJDf
zDGWx#K{P?p>*j(cWkdF*Pjw%@Ci4)@MH?}NJqB8fwo2Bcq%fYR#RW~rGXtexE22Cn
z*h4von`GQWn<MpOf5ssA@;m>YoAHl3HYTOsxsQp2Y5<=bGa<AP3sZ2;gY73e9@}tz
z_VaZ)Ymy?8aEV^$rI+S<PT#xy!krA0yo{W+6rm(pW1A%3t1NLZxj&sCVLv%Rm;)Hn
zH<%R!26Pg0k8|8XTSy40V;0G%+LukrGKe?Jw_b0AnG`7dhq)M+42b3X3f|}tKc;(4
z$cs5|O@xqW;N2vFaZT@|T)012hiq_U7jT!5%#^5ydg2mvB!4>X0GkM{C+Gy8om6I1
z2o}eCNnZbY44MW<!>a^_3sBy`^fk_rSip#5Spbv#9#hcd;~8T$y!uk}ptFlmzTj&h
zN_JM+F!AGF0jPrT2BiR_T(cS(qtLaT5DNRd1v}Sv#Fn$sUI_`E@4`STo3)}ag`JhJ
zGGa<63li-rZ{A7?^5>SG^>|$0ZItEXzsFC?D|jBGdD_HGl3W(!0nU7BneR?v>PXvE
zj)ON3!Vg4UH>5eaMOtb(P=MC=!_M8lcNAVC$YrRsx+~}xmBj@>4fs1STCy_6kw>9m
z4q(X~PH6y6@+B#0Nq8(k4QlPbFkS}cwv3<QBJsKgd+c%%Vw|>DX{Xkn|G3bIVIwS@
zy_#Vl>+u$HaEp}G#@h|jn2f{>7jG{HVHt1pJPq5vx0Hs4WjTV2P>r=5e4(n9$*yFw
zcKrcjV+S6jaOTIXK^dVS;mA*HBzye;U*}gJcHjXHQgq+__KQlijzjWYVw+e!!xluF
z<d2Uud1ylN*R9g)Pc5_K>a3EN-3ugz=E~f$WE_h1R*so?MfSmrtjgbR(zYl|kl4#R
z{%LaotX}a=Bm#6<J~Ck5mC;R`B;|lH>{FBz)n3<!yB&n3!Wx0y*v&99(^MGZ#7fH2
zZ53eEfJN{mRXqill$y!fOtr^>cQ8M6TdF<{7`--Ye!hkIKEtK@iz-=mrimoAvO%$L
zz6vL3&Rpk$63zQLDdy_Kl{SFf+8ANyIooWj-`ljR!1#OQ)ioJVM@}2Z1Moo@v|oiB
z6(2KAUz`O8#<Pho91HJVB)QqJg~ug7u_Ei*zC2)CT8G<ky_Hxwox-Hnx1WT6a10%D
zz}(EDw->jV4@Y%s!VG1(6KAgH=LlD9o%==g4D$BHaUXS~LUB7B{TOV^^>%fz@iKgB
z5r9*6lR)u{CTUYldH`c1@gQ;cjsPwWHs`Agh@xLe%zAtM7garm0#MSmN9kq~iqg%^
za6V4#vSoVnkG~b+ratSErtrd7mnZoLQwMd8^WAA~-$C6w3Tbmi`wNHK*H`{n%NW96
zUZG_Xy*Ya#@7&_Yxb)fGq$D^DvS8L=6;oe~97m#HoRntD`9bR>SNbI)q`UcTu&r~h
zygQxWYWi^$g{}kJ&<~$}gqpSNQgY?k6o2!r1sGTPeV+0xP~)Qi5#Mn;0Xp>)i2Hg2
zK3`m`OSXN9?8Zt0<?{wG`Ak5*AV%o9+LDg9&9?Z?I@6eT*mf@Dk`P=H-n$^$^}<su
z{?-9ZrOg{c&_vNyPS8bqo(72DYdOe86C*G|bIh}Elx*>mH<R?lXiUr~i&F{@ENQm}
ztO$9)v3cveX(tK~JSNF(QqFhf-CP)JKF<%xZYC?e#D$U>+S-6nr7d#dBly^}^5QXQ
zM;)<|g})T2?H3Izf@DP8dO!UPZC7W%ppbKPfvi8EpIAC*&$YAZS$}B6oI~m1q8F>Y
z01k1z-KKZ?anpg(s)Mx*hhEJh<p}1v+D{af2qii;LT=X^y7z_N9U$>w$NSo#m^XE@
zr3>2USDH8z^5GfJCVfsdAHL|phVFvgeo<)(euQVhdL_6vX{StcJkyr<)0ic$Voy<$
zTG<b<S=udkRLC`rn7EHULad>q*_2tLvd`pj=VLU(rE_jUIva)?w6JaqK;GMVZxh_;
zCl*{)^zAXlr@&tSIGJ+IX>>b45_%QB_|^&4Zd9t#!og7=9ul+T2uwI^5n{_vtkx0T
zl5&>kpBeQNrAuQw)~Gv<pfIhlZi|J-^^vzwt~&)#!{b*l<_Tq3W76mR`bqz9O<OH~
zab)$5tn^v&E$v<D`KF4NVr}5`r6NxupVWl23kxwEkA8rg{v4oO_lZ?n=rgSerQnJv
zc6F^I_0!2k+8~KU{L(u2TC7<ys!zOo3f>?^es~tvhfg*`9(i#eVv9IMXKEfh%XIYv
z=}&wVc<oc;B{EW5oMQjh1keVi7dx)ECu$DZWEyC1Gu7H2SqE4Yal3|R+Hy7iXzN?H
zr~^(n;JhKdwqsw5X@D=6pGS3LxpH8wU}mvWVcrnQR&b<?jWk)*fW!X0jlwYXN@~5>
z>;4{oM@NhXCDe5W_$;4Flt+dRJj-<Tw5O=59{7DhYa8W)Cj164%bs8x2ifhM!PQw0
z*PZiOeE7c2duA-%>Uda?oOgD$LE$=J7cEk!gpaE$De)pRFy>bO{tRmPd@LlkP(`HM
z(Xn@DqrQjB1=t6IPUKZ(NVx<3{aP}*$HERUGiOedY+q1Sg&V!@+ZS{P%V8JAaGJ+;
z0b>`l>!*Sf4-mO+uGB$G(5S4?$}_HSL@Z-|QOPE^DYQLQ9^6$~83{i!3lPD^KQ^5%
zo%lrs_P3#w4w;hO31=1$>#r_D2%rqEk$O%-=(i;I8Q3P5I?;&)N$j9()XeUS7Raf+
z)0li?zc&0^df@c<Dz>@GqKRuJYF@424en8m=H5ne)_$5a5Dh?cEq~S>`BzT6ZYs}f
zFPnM683G_#vuhN-k(;MThI{wHcd*frM(i&tp?i=3c*kcCQZfamd-U;0bgp70@<7ik
zddO%lD)~q~KxrWUktKK4dtvgP7kt%10(ImRpg;J6d)WC|8+osM;uzYJGZW>=H8UJ^
zgk9=wC*EcyKY?z5b(S!_f^?J%7l}7VbdIyD0(ZDc)}_XMi<Y4oK;Vv79rmQ7))f0Z
z7qzSpGsZAnnW4ZR&SYj$ScOO0Y~_wl$<JYPe^G%@&N>k8QNY=gg(c*Gs?)@S)guBt
z!_!n{eV#Bt5*~B08&q3ilL&w**7()jO6fOBpof6K9mNIG1s7a8bIHiGY|PDmfFdLu
zEVF&4n5bUNIbc==mswkb36NBD44{Fqo*<720nD~I-mvT5Pj9CoUM}O(!)xzbp5zcT
z!AUZ?AepIVocND!)7l{TSSLKv65&t)9_rzk+U;qA3vADySHT;~FvgA|Q$Nqw@0)F<
z;{xgCb==?n4toFfp|GJ!xR5vqa%r+Fy}1~Y|J=q<YJJEr37Z;gWC*C-o?PV=tQ!_*
z**%A|ko6VmaErfJ=R3CA;~y`=?t>nBXRR7U-+aAQZPU_rRyYa=c}v#56hAMv5?Qul
zsgk%341o)_V%LdG?<A#!X6rl|wN%PBReE`Tf-#p>jb(ahxbnvDwbmDr$*Hj<-4i9c
z?~Srs{-w*{-#QHb7=tT@f3DoQzerD~UmZJ6n+(1p_w>$c4cx~Qo#(&g62~%54-c(%
zi_2bTcLAb)C3kJw^U!OwOv>f5xuB$C;_1Q}J5Fm@X}M$01E0#sgAH2W?KBzxG=~n=
zQlOZ3=I>g}`%_NS-{Zb7YMqO)v5iMHOkMGLsd)u}S|=K>j(2>nF27VC>}Sh_ecL?L
zV9k$r40z&*(ciE&`~=mDmwrBs>}J;{-QJA)(dbEQuU4i!mu)|R?5q`S0u!__=W)$3
zP??PkH+O}d7mQnd3!)GwpX7Y;U`Y1xED%|<Yg*Rxl5|yinz+CBCpN$hePE$StYqPS
z#7B=uLx2kgJ6(bc?4W1^gbB!?Cqkh!W7qD|9n5l(u-ZxiBfJvb_Ygl3TQS?Q<r76q
zgN;^8t6Jv+-YOZB9w~8l@;ju~q~-rB8vOJ>(hr_YT8=FfJ{Dv~P2QTWiD0B_eNQ_i
zeom~KzhEQ%@>pYait^GK-~BffNk>az;;r6<E9h2u{~UbwQ~^B0kMsJFp!Rr0d3@kU
zW42i`F9E4Yh^q9=#kJJobpvWAZc=A6#Z?|83N;z7c)361h{)(S#!R&L^6&f(PyeSM
zqx5ebGyit`FJLLN_cw}wfW#Yqg&m#p;WG*diBk-Q1aPS5Jhx6iTBzwoKy!iUlj9G7
z7nT<B;nq-*x3H0>crFha1#-Yd=`z>`km*U}k9~6TY*E{9T>Ckhvtcb$G{PU^2F_&q
zE2dU{)}Wj}A^^$Du>~c^{uh<G7hDITsbg8IaJx5~%#w4wPwrK-t&nzA<6>V$5wrvu
zpl};f%#k<YxDd0$)Bt#6=^7h`du7<2j|^ZwJLf4R)8Tm~e-Ei`a4LHJpw{|hn;i0#
zBuv9k0_FmNeTQ6LKgF`)u^IvIlW>8n&z53FFXE%3QNm3|rXk?PVNpnKH4-iSVK(h9
z8H!EA8@|+}Y~4-lz;t{no&hE;R^h;$yIil~mrEjB_Zto~3N5Sg1@5bjemyAx$3iQ2
zJSSK&IO&^FuDj9c^_$t?-rTzI%#Qt?F5X{MeiL*%7#AeM081>)IaI^+E&_~V+M{^z
zDWbaZ<khj8B9dwRnZKw~;*XSkr|aofE|<Jq^7GI0Y{@-!fUgjQ14DWtcrU?hA6t_u
z%7Pv8z^FKss6x$4=?{quWJ_G^bQ_w<?SO5h!m8jO55(0k^0(V@DT^OE!+{79RM|JW
zQ;C)b4_qrcNLnZrCh=n46vQ1T>Q{V|uTvzEd;ImuD(x!)=EPm(frMLED6t6Fw*d2N
z%C;~obl9*}zILIqpk9SEF}q185GmP`6#v2UjWlRw-H39&{CRGl)vQ`NtiQv75$0P!
zXAGn}exwbYqS$J+Z2OT|OfCb-D1-CP%qPG398+PkTR9~z%07NU%Daa}_5l7}ZdLQR
zpiaOM$O-bg`VKb^DDgv%_2u5pk@Cgdrk4KN6~oZYpAh_K)I?!X9K6^A+*mPn8P;8Y
zrA8_rtH7D^v=-KxsXfP^8*nSi3z2sF$Q{$!xX`_nSLf-fhZ94R_^{!#(absoN;D65
zo+74xSA0oEo6HDpie#Y-5ca-&^FETYsL#Ow1B5t9xqS-okWz>Ho3}m%9U4BvDSUNB
z$uut9@tnHC#N;CyLWI!r)Hi`(8Eul?{H~JQDodY{WJbD~`4PFXYq~VQsC;<2&|XPg
z>#qsbCI%uth$+np-1_&nc}TqTyI%GSfwWv8ud_9|lWcRt%p3Dk<jsS$Opb>U;WaUE
z?maFI(&#j;E7vGRcu?qwP6$FN^+-s$Ie;PV1u9PhP2Zj^uscm|g7t?v&%%$}xDO1n
zch_~fDswHfj><0q0HJt|VCvL{o2Fk2xFfBKLsQm&2it)v{qABOqAUMF*J#!yh^Wz$
z!d}5(4BhXqSYCMC3%&^*C_#<s;V0wS9@TN1SV5-!pMeWNjm<t*A08f0*4$e40bpGf
zX9b?G!uSUmO?dmUYHGa{0K@}x0P#%4{8h%nR4eB$0<pVoI*>rA?y7nL5J8Z0*0p)&
z=jX`xEU&rJy<CCNN_7or>=uo209m8p9JY&G5ICk3fKCNIqF8gqn}sr1R8<sP)xw|(
zmRS|s91EJU*!zKtoVzn2X?Eopy14qH%v%ZPqehPZMa5kwoq55`K&7tVT<A5i@4F>W
zggho}XX!QTMb8(}sP?4g^%}iY%jq$wY2bHxxx8T4Iq3Jix6L|Ud0IL!$Ity}t<;)l
zsP4hkn5$;dAB@%IZ5BRtb`4$GW$GAdZTu4;?mv*}f3xSHNFr}z;Heh2`LaB3*RYwV
zy}4&c1Cy!+1Z<fESuDpcYl_!>v>Q<Ea)q9JRy1R9KYef)WMSGbl)tEv>^%<%y|c~L
z3Hm7iqH5)2Nb1wBa<~OSby>=Fzl<1xGzgzuNxAqVoibQuz0HBn(&4IuXV~`$!OR?F
zLFPM2d@XluJP>q>N>1F1!eiWf%>!Ut)cLOnzEbV0Q#8iQ8}50n58GwdR|jihPk})8
zf)cU3nXM!-fpRNWB`it_9eq=LkF^H^T%<`pc{)Vib@7~sft5Cc(`II#a&&?wh0r<N
z7RWF*R_*gWMA0GS2ryp_I;Ca+u3<v5!3Vf$I{I|c)}fg%ePcjQ()%$)*o@R`8^jNr
z7>nFJEwXbRyA7i0a~08GXN+bMDO@=>$3}S#w?SQ=BBkxwlHl1+x}`01NzeQjU>dt~
zSS_~v44?MiblqXj9)Nb}`~FW>{tvej0j!;;53Wb;#Oa<g?1LoHUxiHaV*DmbhpyLf
zRl}2yve0fNxh-1y@d~Y#PFAz>h9S|(DxOL&y)3UMaNKF~JKSfgga*>$7bL02YJE%j
zdQ%pNG0{MUGhh85hhC=su{iSguGYW!-oH>F{4Y))aQ@0K&hPb5F)e_v$+4nIO%Gpt
zgqe^Nx^Ovd$tkG#m!?x~&RT8w$9~WJUZN2d<q})`^wQuKr0M>-XWL(+H1m_=Bt|$Q
z_Qd;o*gBfgHg*7PaGorhLo#Z7fm<mTbbHa}qb$7@qOk7ep<|{jm)oAOnGJB}?*GTw
z=+FJKKm6QvX5Nuc-zr*JkgXq;uAY?5Gj?qiveI}3U;NTzJ?=x@#QbraESuS_ZLzlk
z^+Sv>Iq3oLN18*Jq6jIzmsf{$gK(rD9OFvu&}l|7y_Fh5UhVb#MfDRt<NfZoNNvV$
zEZ0o&2j%(%&=em=kM7fgv5Fr{SQ};1Uhwf%HcXo)IC=}c4<I<-W9--bLNe%%JyGC+
zN<k>)d_Bil{kIxDCQKGi2X$bC+cBzEi4;8)TBV;tqZXp1=SZxZ+8b41+1)-8FEY}j
z#l|*&)G6KLa&s2|3NQp7F_8_5sa-=S;rq`ZM9h#AXfh42?5$~<#dB=q+L0C#ARiQ4
z<5Fio=T&|rNme7Q*#YS5g$czVQV<BvZGLSeILElVzb|V-d9~r}S>UB(G>$zt4_@<S
z7}_J$NfAlB>a~UW8U}ByT>~SG-z~pu1W=6-O_<{+aVfiGcy;z~IcNx&(RA|J(T~ph
z%^O?;X;RmG6n2l8vjeUXwDKw@NKU=h6YlA7fw@xTsN{@t5&L4})r3I+QI05^ImFhg
zw9uTcj|@%PDL&Q#fSrOmR!lSb%Z19*m;Bj=zY&#-GhYdcn^ixR@zssT)R$s_m3Sk5
zKE+HEs#{3oelwE6fxH?s44(cZQEj|Txm{sLxgzxh{J@edbrf23ULc<BH!HPRgXE&s
zhb<su^6T!BR<&#;D8eJw7s((W5)hzTvL2V!@%h*>g$b_<57D^Pz3prx_IR2CgNd+s
z+Hl#QD{PuWr&2%mZj-EQeo+}7M^kPu>xuwVUBTj$Q}}Yl?^tLnp#aLRuAi}LOHo#j
zvJLT8*VvQ?YXUYk%%}Hez3bo|Uapq8MrJVD8^=Xv3{nY<Ua4cKq45z_k~srZZNu^c
z=!tuw>}r8C*9<3bt#XgOum9}!vUTrUi5|Do%}<P0Z$PXs;IwuZc5;+}qeAAoOv-E?
z;Pl`Cw6t7@JaQx78;aKK{epX0PH4~nNe=er<UCFXhB2qwhH4QFN(NB}cAZ^^N#w|J
zIzRU!OM{Q_2_B<hNYkAH$0LE_@Lf$9BxmdrcZ&)hVZR<(mt*6})<Jumk7&A{zXV4c
zD;pg}VJNGoAPAiOyHX2IZIP}g73FY)_gKd?N<SC92S;E~%l_W9mu}Vp;s8Opisz<m
zBEomWA@P0c9GcX`K`4MSG;tM>xU<tSJ@pqTQ{c-bEQYcqwC8}7h(!@iAPJ-N11NY7
zinWU)X&-Wcn@6g{VXWd?Kaq6BE3{jKdX~D7*kXAZfCS)G4V??DR|;62L5)?<WW8fi
z;U5y~kS)avV5VPMhNr>$xh{#f>IbzPnyR*4<cNLMTBm&_+U|pj(@&sU=}8IA(Y_X(
zN6?4(E|uur*|<4`PVhI;#h?8bSNuN5vUFx`t#|Vt;M;$*MRzYe>+Nsi&i%-$|Gu{G
z8vl;&sRj>l#N>@S(g$2X6*Sx^w&Uowf<r5eC=Z%^z2p{zUDmhxu|x9Y7iTPqbWLYB
zeW73t;%<dpE`gs?y6B{5HUBh^{%JW-b39|ixSYAr3EOKYQ-FQVf^}9+A%q1fd?cs#
z$tLLYvcq_ERlrhS>_YSu&ukQQYtKx2+C8R$$3;q@{Q+kAP3P8_VNn$RtVXy=qWh<_
zhI*Esc)C&Kj0&QZcSzjzlw)Ffl($%ZIz9Z9`R-lvX40LC-p+2a-0>#>p5|hr{7_K%
z*1r=dVpkNU@4s5E$c?zO-Lay2+6__%J43MhOxSFw0*83KBiBl5l$-hJ@8aa0AQwrJ
z4+&bI%GfR?x@L@GPV!x*!=j%X4msp%bbr+N0Ib{c^?yxtss5H*{{Hs<vN`LlAw7rm
zv%|=DA77N80u8=Ew+nGs-X~UAnQLmedJ5*bM0O_4=7dC)Muik*k1RSX*%*^v^ECEK
zNbXk~R>`z~joZ-sf!Ma!(w|6=9E$zv$*5!Uf1T>kUrR1O!8AN1A+ZI>-7wyZ?CyR+
zm5-q%b*U2FALDFsee?GWEkmP7J6S%}O*!~Jx(UY^b;9|?(1&m6o*x;}5yGd*C^{jG
z&_+S?K0Rgqk;%mpj);Y#xUAk7Q>KilE!T_7Mr@5vF(%6U?RRepfWeqrCxF2kV@Cfk
z7s?-uMgGTPDDMx`FV+tR9nw$PBV^P0<s4<;UIyb})9*ud*GNW&x<Ucc8I>;if|u`+
z-7ILC&@!d`vV*afQO)XC5}ONrElZDt$YR9P&l6#sPuMZ<DJI8fd;wjQY1A(&I!`6n
zHAkTmdvFU6S^|6uqSbG?!Tzo|fZMW-N&FX;t3=)6*CsvTMgz-Z@qI2jClo<*nE`A?
z^oji;UI0e_cATfEFu%tU9$ipp!wKJQx!nlLBkZyP(=H4Aa`DQg?HpEMDlllLXi>lY
zh2E(6(l4ss#W};$BRTQ|!co-($!jitbuK6N-~xVEm}ZqtL6Z_2_$)2Y+^mYO8*!Xx
z!s^G?Nu&<malLBxK^!fBH3(%H^<=Sv=!6iJYfOzli9nAErhrPUXje|tlj1#3@og1+
z!JNyAV?RnT=mxGSP589Y>g}SVJHQ+~N&Ch!34qdwGFHUbbhUs9+NhIG;XG9alr$^%
z38r)iAB@TfYgMD(u_|jAF|&d>;Iz=<Y-yWuB=-f0!NWNDf0&qdHM2!L@Z-gqW9E}~
z;53w<D8?G$BIn(qr0c#W(?&djntlbPK!szT$RnGqMdU>BUAw9jkhH~h71jY}AcN#y
zk(?HvlxiRNm7lUA_d>^`92*b%D)|Ru;oU&3U?4`yl3k?HDK0UKd}^#l7N|II)R!Z~
zM&$06M^X2Py<G3BqrzAayw_(4Q@}9Fq;xEs;4E?OK=3BOvobj7&-c+VCF~@N3yLiC
z*K&#;gZP4eypexbz@c*D8OzePPEhb+BV^ITRa>r22ew_GOZj-j{3XDWU>_E5+Scqv
z?X8sLp$I<-#US_l!)_k0Ik$VvakbW8v9X~E=XtKHBu$eQZHb)O#EIf1Lv*-HiI%AZ
z<-3Jl_xPwYLHMFc<{s$l=BZANU6o>$FvH=T<0*nRR^hV)rce9SZ#l1=S^BpJ_-itQ
zx+OZ1dqlj=E)TYv#IUzRg5$%3%ZGJmn?gDId>%Xo>Xy)HH!?%YQ=y4VJ_SWs!W2Qi
zQ@Zy;X`(#p<4%Tec?;z2WwK%tZiBz4cHADi_`YlER5{b2k1HM=On#c{Vc|YNzT5s2
z2!Q&J&w(E%k(@}r4~}H|9rv8N_rB_i?0?yKMlx#hnq>H~UNUCr9d(PUxJX(dZJ_k^
z>66k-eWa~G%c;il0SC7*PbTdN%*!2&6zNp1Pu>$~e;zB(;A<)Why>cDy5YXwv|Fbm
z1wCJmE~O%V{>|i<|C|LH<jJc;CC;#eCNvFzzI<IntD@nbt4}7WSPtCb9!Wn67*bM8
z);v4Vh2LOX;E|e@50-@zky?v7N=Bu%h^@Q*g}g_#Eye|Blf3L^lo)mYOZU|8<%xf~
zuE4*dL(hMQ@fD~I7=c>cTz5_4d>YUDF$va!n-)_Nu;qdja6}4Ma?ARcJHM!EsqIsI
zbZ2w1P&eqAIV;`}=|Sm>H$%8gTOl>!I<~@|D-mF-71I>-j2Q@RdE(3}rn@e!hz$Mu
zg({bdh3e@E9AX#*5>Z6(<^EGG;}>#Q&^Fz}b^xNu^GyE<Q6!oh<GQ2nHhk4X4b)(o
zQQy1)b_RpCU^9g^1e7D~nC8WCEQ;7&0aeo?Tc%MGZ24cv$&lziJv#NW`0J=#1u!v-
z%uuXYw=E<mj~O>nhs~$8)<KSUd@;;K?KNg1`-8F=Ja$bc;$Fmh)bn;2f?E7gNBulU
z5aQ@L7ngm-qcnlsRLa(|0g|&!<xO5LQaBCFT3R1M5g?4Iii0n$7~WJs%U25=yF-(w
zKfQ2+l=+3k+vM?^9fADk<*vA~ynxGW)n9{k!fDBY!k>Lp1tV&sQXXAAI05-G9q{Dq
z?X;+-ii4IZf)}~(WIgRrREa&@g03^!LV2T8Ddu9%%rj{{j);2)Y<sR}VLwsyWX{5e
z$0di6pQKLWUQm5s|72I&g{LDkpiNR18)c@$s1BaHJgL41h>9g0%;I_1Dv1m;=$x9N
z9ciAT-qRy%JiL^XmEwCt#Rcqmsn$>P=gUj%V&I_N9s5YR+g6F?JV_QKQ?GXxUjt(i
zQ>&Eb1X4ghrnKL`Fw;K!L<J+hWB<BN+=sm_;qtfxZXLl|W!iN!{rgcird>8%BV05*
z+!x)K-X8}>%k&;DYQ|~Uy9wskrU!SHFMt}l1Ee~OpeSlywFCdI4fulh3@M07MZN1|
zVD~Y!inN?+Mrsh#{?KY>WB!ZkoU3+kEss^<gLe5<;!y*frRKr(gzm-W&#vV1Nkf8f
zHF6eWO9mXun;??CVF;g`QC3H4_ZQH3>tn>^jJ_Kr){f)E=XvtLTLXt*RNip%nbm#s
z!w3ccAdD?<J9+P@51pM4zDSzAPfYpfn(c5{ik`2T<qu<&1}LVBH_J`bcUdrfxq6@Z
zrSvt$`%}oatOX9$UxC>8&6~f}uJi|Ar*08#oQXyhqI<y2CkhiexHb*XFvY98W!K{4
zK6%N;ML)CXYzWQHGP-7;-SccZ&DKAsAgawUWItjpwkl~W_es3Gw=yv)+7D2<v$_9I
zYW@Dyy!>HHsA+3*Ij1R&<_aK=NIbU+As{Ac&*)DvU8jF>f0r|}W`8db!gwHw!ll^n
z+0B;swFizbvM#db+v*PL6d4XI)WA=8=o4e1cTCpHeNncqRpz^7&|t+As_7FNO6E2!
zYj4N=foiMa^?&w7ODj5kl8%~NOh@$+S1U04`wH>or0On;&Sbo^DI6)h_;5vjb)oMa
zgrK!9voNVv!<FB0X43uU(77$*l0*b{)Wc|CzD6Z!UNg|JTBCS4uzTB$)r*;dtP0v6
zMa9AtjuR<2$Z{dj@QDll)#u<NmSZ>WA^o=Aq`!Eew{Z_ruA4gks9!x?Omjq<$`<?@
zBS?xq67@NSx2^vCP#H0~%~f4lc0Ad%TKa+(hKVPA+6l2&p-hIZG9Lmc8YzYS$k>sL
zIzYw4xQNP_8M7kKv=t*Z&&E$-*8@1P0K7m8cXeeZJ;UKh2$tP0rtb+eTQ28zsik^M
zNfSpGR$!+FP{xFjjN@Xv(EY;hREY?GKWsN!LU>?WyY2=~6^^sL4n%VgBVDfHI|Ist
zRmL1Koq(hxvU5Jr<^Udj_!$`YoJWEaMN|dXJ(UJM&ll2n?!JmferfR|HBmjFq4Si+
zG&|V&JTtNKLMY9tv0qmCf!Li22E<5JkTB-%Fvr{ucow`%Ll%_`&+2wpg83%<*w#KS
zz#LZ#4?q^`?CXQ^@@WWU5AbD`N&y`N<lYJa6`865p??%0o11{k6(JPv3b`Q;RWye4
zOS{CX2VG|BOWop=CYo}{N-*&c^HOV>S;egbFW^Bt{-Udeq$-w*Z5^OIQqYpims%5B
z__w6v|M@fg51F0^1+A(nQ9WNiJW_hMc#v4^jb5_Tc-PcA7PrsTtiM-2Xj0)Y7PX42
z$eoERjfxl%7p%nTnBZW=wD@3~`HpeU_%uuN)ai9@kHMe7@&Bm>gZ_|c&ot#0*|?D8
zKd0F4;(E*KLt<p!>VeVAy?1p#u5UB<f+|!}+*Zv|NClQ<1%(iB89!n{;<OgZpK1#%
z(3vGETv;)U$tj<~(3|+=sf9i%j13T>+2N)hD4C74vB>^v)UU#%t<sMcyC%WT&SH4c
zCuG<luhePpHzgf|KU^32&r$UM(&H1f^Pgmb^UJVT!fr&Xe=F=eTQLTCQ+Fw<(O&~i
zkJX7XvYYU}9Gxg1ml8j&E0Um0wa#;yE<>tQe<Ak)PQ$L8F(zg{A^U(CAG#(yy?F(y
zt27hn=pkroD;k(Emx!dCOv@Um(;y*_WPCPf0tU-4sgIwhS?Bm(9{)x|_{n<RXks6J
zK~9giMX6F#(dKLbF`J?N32ZYnFmEz4SJQFMTwpv$6S=@AOKRvFsaNU<Q0&<PhQ8zB
z{$~cv$QcxG1#i~(Gti@<W9)qS1Vsg}Cw$99RlbSlibdNin7B{Z47#Oy$#1y%iQ-s7
zuiC{2JNzl}htx!~JwbwSv+6UHR=L8pp$_3ilh<MF{l#O;5Dv1*F;dQ((E3130zQU&
zX`A3I$4X;!Tqh-d<YLiM$#uDDWj84T_49K*b+Ll%mH62g!YKmFo{-_bRD~-*@_66l
zv)2l(Z0jCsRP=6>lMZrx@wlNUP;aJ>$$=s?IB=vqbiKX|W{~EF>O#W=JJ_2<TW)?+
z4TXYeE~_{#%P|l^%~O{=>B@ZkjVlK2!2Vq>EPhY1eTxe0(?v#k34Nm)#UtQ_dZsz=
z8L5y`kjmxCL|E1unu!Q$gp>H>7>;ALiFF^WtslW%idk1%7dwE42&^R&F*A79|BCmQ
zl>}TmOP6?c$k;BX>E-jFdmZ*&5$OXPrQruCXmHyM5wVkjamgHJ)PczP{daAufAlq~
zO9OO%*umY!(DUG4)Z_-}CZhL6q>>}8M47CQ{cpCaA#NzX)fA#(dwc2UY}KClfa76_
zQ5O&o9c9+!R`}`-J4L<gQLP86ko;f%m#&n*bY=fVt)ukk>r&IRh;E?AO%PxLn6G$0
zJGdJalp`PP2BA?9G4r9UEc80iq6dl`LZqIpDU>FE(wleH8O1H0ahnEuo>1Xxk~q|&
z>s|Apqk2A0Wf{7R!1HY1T3?U(zd9vxFLqTcHS@Fh#nFvRMgXc;cKM%lo&M4~QW+(X
zZAhhV)8L!Eouf%B$~5*?Tfx`v@3v!f^O09b+K-90H7`~SyJT7$hN5HczC?b_U*T7m
z2rp(U-35`@TcSyQO=X8MN<1`8&~$_{d!6Wz|H!sL2dm1HCBHnHAd>#wCfCf-e79UP
zGsTJvQU)i#^zU}j!m#UXfddu~KV5w|`N`v(s}83bQDe`Lz}0d?zNI&*$}8vQj1U^8
zq_<+=hiyS!{Y7OBXA6)l2Nc<~0PrwW9Vwl})C`D27uejcf4qdL?x<)J??9i^PC;#T
z8EpAuy4s2Nd=;3Pyk@(ERu6~B>NUkE?Ft8>FEKdJqyERi#(@veeP_vvMCPVrEztT!
z{gh6JUPVwW>tVqQ=q?vvV6oGX@C#c`e%7N?g$sK<it26SKoj&;xHTP{&*=%v<r4u9
zGq^=Dqpznp0?X}l+A?6val|s6l!q&%JiXW>(Q?p^f8d@1h$$9N0De|YqCw>f3Lw8v
zfBAu~xymVDY;|*5A;%?skusW%!H_|-nslwW0HY+%3|BZ!?lnWnGvB%8vH{|5i<`4}
zdJiK}PK3X~NcUVpPy_f%03iOOqtQpVQqC6eYCeJY24pnU&C=3yW`vsK^Yb#xPS;E#
ze12=NeYt^n#!q44#!KNgfMzDF6ktJcrCd1S?0LbFB41)GB<XCSIOP0J*!lZhEX0Cj
z-%=$RDOTYrNT(C3^DWMnC#;Hc4N3?YyV31#vSq_BR;A%nuN+v%%oW-S8c3+EN*dBp
z3OHCg^Z%6%DNtnCE~StfgShX!Tt$4W;?};lTNCw-&-TXa4Nyp4f=qhZox9#DfHK^0
z6F>BnaW763lbSoaas?vvWTrZ2MPS%BQmR_ie;f#^9{!Kv?Egc!{(s9sU^+9uTZcN4
zwBoal=WvZK6!6;uqNOWXvSxb2?O5rUbK0ACu=9eH5rF<8E!9#|AXc#E=RAQfLt7KH
z!rh=UPouuRW(uC33QbaZO&Rd`saUdMT*Q%0S=s~2)E7NywVP6|l$H#;B(2P{his^y
z>J_!{JkM{c>BRn9N4Uekr*E(Ow48jN+gC`bb{rDbO3F`O0BH%+jn_{#!b+_IL?#xw
z_Is>X2QAd+6K&`bOUzAblZuY)P243v$pQ+++7$;pPj$>RXA-%*UcfJ)ZwGmDzWCZ!
z;56*yJ{q225wRj*B4`&}NGv3a>UgZS01{;XJH?Z~a#N+J-f^;_T2h#+2hjzlI7jAr
zTdBSx@s@zgB>P;_P1(H9T8n3Gf=eI9?~Di1`qfG}(zsR6y6JqbD~-)!z9j*uhIlf~
z_(9s7O;_=t<yglxfG>8p@w&kQmYO|mM|>krG2O}X`=D7cm#-@K_1)9YfKB1<QUJkv
zVCPBISC6%|HRK-&4b_*M%OLlaPhj3D+=aC$h>*PHhZm4nuscc%l5brq+5+3DSuleF
zbOQQY`FD+qIOF#YgqasUazDvZ@Pl^=8g&7zcV^bOnWCgpM{gCpQcF!)tQ<r0kM2-f
zj-b<AZWby{fMreLCE4WN{yMX6=+9~ta(~p&a+Vr#e2-LlBt+uHPZYR#kal{M+-<6g
zIUzkO-vd`juCvr@-m{S4Z3_#t`<@f@3Wj?VFI0a~y^Mv_%*)xt63>A1tewwy`Sgy!
zkoL=!1X6w$^Q**?kQ`47aQh+ibl>`J`l?$-Xq2;7vY)Dcb@@RW+?zKe5RwY5at#wB
zZI<LMiL<MnHH3O!$|5q6a(J8>O5C}*F=sfQBD2NE1H?(g_|0IMbV<Lfc>*dxaS3Vi
zp@s)up#@sp?z?oiB5$N~er=wNF|OCjBgz&AJTOeG3^VVlhmd&JXqmPa8&G3;O_pND
zYjTG5KV@<R!gwd6-2Dl6>mvhYZ33Nh%x<-0G<!dh8s`aaq6t)j(S5XKSxd!yYaMj^
zmvnRE1*uBv_jFUl3CK2I{>FBG3S3(Nm*{W$yUxYfNAtQLQCJY9iM{%i8#M>(`H7WQ
zvek`q5>$@>V6B3Z(r(A9L7cU;2$7OnCFR0hai3UdD?X>6A>gW)DWb~*NGj9+Pho5)
z{~#9k@9%p&i+-T+^-TR?2i37-`h4mYYeY*Wk;h@o>LKru1xRAnw6cO_q7Yu(Q1D?U
zOK|5#4|jF~UP{?w7?xSQNIX;WbbWk%EDDt5%c)(nHl>&3?5T2|i6|<mvm*&qH(Z6&
zB2rT&GhB>rr2@VBT3l@hMrP(e-zn3p^9|6J{>G+icK}#_zlr&wr+!z|Uk7^kL!wN`
zkNHw_jaT+$KE-5Y3>nd#Z_0OJ$=NS>L&r;0SwC_0^{h@to1M!YZn%)W&h|#H6u^OA
zUtqpU>`c3`!g6ZUSSGRQ_@V(;aMwSa1-Hl0v_riDN}Z;n`o5F@B{O-!R@URUfRKkT
zEdv-3lb29s(K1ML!(i71<kIyZWZX`BvOZ_sSRZr2Kar)9<G40%AF?uQ$61PYov90P
z1fM~-JW0Di#W}Pxog&~)@(|a@>gJm|ms~#+<}N#%^zE$FHGYQz6<)y}L1$ol)BjDu
z@x(cryUI$2$(m&!RUHx)Di+W>nrdxUE>=A=8H=6ogDAsxeQ6W-wQTNC6NS}RvkS8E
zPAY+_9)nU38yZ%Xyo}faYx@q{8;(H&i%s^iG#_{LETUT!&6b`mmssf-3f<u?x8@XW
za~sIR>N=UG0beWd>ObQn|16KYf|_J3%8G>Ym?HSIRZo`8z+CvN_7No*n&VGSM$u!e
zW7e+nnouJo*_<TlKkMw(^<;3PbMD<j<k$&ni`mJd&ok@|v?jIM5JP%_Z=dpL-pB@N
zn828S^o4rJT1fR`m6;mi50i$_MjwDJ$ULEX-&WweNc5X?y^aoF=x9kTM>l<V@#(bU
zwMk1dWg3CGJ&FhU@cNWtLSHM_sC_Hy+(Jc)|Lvh2pVrxIQ~rB^n=Vb+JuG^hG6@16
z<D-0xpSB5v=#kitPjSya@4H<pt%1{p*g?`JV0?sol@jdCzPEs|c2@-thB)og1#^|y
zM=D%dkr;5S`c!WETxDTU$ovC`KjO%LR&q^yM)+NS?LtUB0SP^wAo57^PS%g0@0gG}
z0a#z12uUSd^C~7&O*|=}@o-e5rM$@Q6cu)p)%(h`PX;P0!}JUOjb8O#X}hK<y09+d
z{~_<a<C<Extzi@ul_p4UQBi49qzKZYvH=A#QbUi@Ap+7nQ4|oQNLNrIy%Xu3D7{MW
zHKF%}8c2xWa-Vw7zW1KJ&n@5kd*46!Nn|A}$y(2R<{Wd(F+M!2OP#1z+$2a5tv%(i
z#!d#S(UJ6U_Hzr9?-=!dP?&Sqm+f*Cw}mcuEH^&4ynTvij}Xoks4sg{kDL9`SrUAL
zOhX=?7AL5?(TE%0sz{~dX;=Ve?~XlDfjz#rAc^&wLI`JA8ZdKGyN@&eWhf;SCUumy
zClZSPLGc<~#urd8Y|wHA9jP`*UK^YG61ljm1?e_z&?ywZnXK&39rn%4`KT~RFzI!I
zieCZ8Af90<L+EF}0W1AqN(zgt{e;6_zQa%XaG;dc)zrG^Y9K>b>P_j!<Z|-L8n1?5
zJfl<=>^(CeHI%?$%#G@mHaE3=CKvH#bUAD*?N)aBDY3SBt&wv;WcoLnz5ld6`<IV@
z82w9_<tYDwOfFP*stnCpQ9+a+Y7vibgQ~mSi+a!BF2nqNk`Q6Qlz;!?aba1K)cgTx
z?W*|^ROfj~{1JQg+mt~_-uTxn@iU5Vhf6Z?z*ug)=1)37pt74P_1UeQr0h=ZRS#(z
zb+xY`47XrT`_-5jW#+-uv|@?Mzpz<N`w;Vd^mC-Vf<kqGzxVDVIC8tNy!IokB_i~v
z0Ijh~=D+bAT7a=AUIqmpolR!rfLBFakvTjgyNNOji?zx8ZXndrs#`c_Y-gyjP>}zF
z!m=t}JV*L~zQFU~8!)_9e+h>U6W!cnpAYR57vp>z;4O{WGHGo<z7d@}rtV1`nyn08
zh{F@h9nDSZU^ncfZkn_#12@L<RsKAdcnQ;6yr$q<Z@s<vS2Pro=vst-K-H>VxI#ym
zahSS-h)?5o;Sd>LuqI4L{e(L5<}pOAkAnXz8&lchVACV-XW?_FUjO>IRMWWS<v&p*
zJSiS*EY`Lwb#LvL@=7Y|q8fW_DSvKQ2De#Lzt6|I;g?QS0wJmu%pdWk)-)NA%EHBz
z%95ZJ_l}o=?u|^t(JGIS=Ao(8)F|3K7s?GVps5|>l4)#M&LOS}$d$S`l=78`JK5y@
zS=#b%CK!Ks7yq+)#((qk=p?Y{x2Rf9j4kShFb7ueQ2tw6q(WPMWPqalki31W#&M3r
zFTKekCpVxxFT9B(sZ)lHesJMheEt`;3~sjPcUJ}FXcb2ulthg8GgXv_JkL&}IHAAY
ziA+4ATOGd9r!NbRkhFKvs~&i^#T&mliHXy9*%b0F)lhg1<YSM2Emie~Bx@8LT8xBM
zhB?=6tMHub6_+05C{dQ2G>mpPvz4;)MdGve>3c~A=*PHcbH5Cpo7~@~?E@EQw+LZH
z-+<2hYT{b;;v1VdboL|Jzs5MsKOLQS>h^CLhKAG%rF(X&COsU^zdU%uip?{_wEl1n
zmOXEbcyAe>7IwS`28Vbn)<<rsc9|D*utoYuR*8NknUq(>f5+b5Sg^5Dij9^C(q4zJ
zKmukes6uNgCG`}qPGN=FL}mcQBb{<|OQ9~Wzh2V+s{r4J9MAJB=JF*gBYfwnQlD5L
z3w-UdkyLB`=LeG;O$J0HJIiFYH!tJ><Yap;Q{@(YN2?7mUwYaRO2^lA#oRgZ<9KUN
z8rZ;}jr1HB*<5f-?i0OENEYuwd@*-n_@oOkT|WIWA}x`m)#ui&)L0!WePlmb?(wdC
zc_HMB&d4ypMFKcs1*-Qg2S`3NiJuB7Ozts!*cn*tVMDc-%+NW>7y;S$#$^%;Yx)T^
z+mrpx>9lNo&^71!WAH4<N7MKqp#G940((7yC*`x9`_bzgHE%+`S?nwB0oqe#0^h@?
zWQ&}lrxe)t`eLoBQa6DOH^1}063V3{r))Q#M=xNHm0h2AB`^esR|phd0T||tqwQ66
zI(slfY(cQruw0^wd*7!xX02Au&?EVj#slr)f{_G3I$I+3;D4}jGZJOaZgpN^HL#`1
z{=`>DKq|GQ+t^&dmtP~JDow+S<!1a7w6^g=>zmA&D^h3cBE7#St{MeR@g^wl4RPH-
zHm4*3hJn{wYV6P4Q8f2H-&KvL*|zxQ6ntyA*ru=zv(nnt-d$`^Xo)<w2%Byi)X_8%
zxK;!p%fC@kKDqz)2gPL>1Zcs;-WeEHdyOF>;@rySv1CbC?Vn!&R_eAWCw$yL#>w*Z
z)ZlaV+Y=l$6h?^+!=-Yf47S?_Gj}RjwrRK`2RU{SkuWM?c;KREe~(TPIvVcvo;JUg
zrG9N$ZX|u#5Wak*To9W;?{<}+Im0Eo>HOaF?hg3^8<!qcSBb`>(+NSY*^<Pwuk~<+
zy>durpW$#>nQ@i#p$3t6gsfT{^GD&YN)fJZg&~ia;?u9rTNq8WyJF0>T6YU-MCo!d
zg>;YuWq6|_clAd2$;hW#BS`Iw0=C=?rM4NBgW=}Jf`LG;&0om-IQ0)op{EOS21uwY
zQ)pH9+izA`G<u_$<6JQ9OpC133iQl<nnn$!n3BuQ5+RAhM(%?6F9)s6%pRf~ZIzZZ
z3Tui>_sT2nzq0iZ<<GB22!KP_s~uB1H4R5hL5ULYBi{qwM9+q|jmpO^<UuX$R1nnS
zwgqj=siFVY)#$%DFrWM*ua~=;=xSJxRQTB3q%G2~8H()E%@t5^rEKw~#0es32rYhE
zi}>(bMyoNMol9%+z@(dLPp1Vox|S0QZH6!|`zw|fl@&*G+Dx7XxP2e3vsR4Bi#*rX
zZuxXG8);O>A3!Q96Q>F+qbMashv-xCq<ZfFxzJx{&}5-}04K*f$X~#wf2IENz?JIo
z?_0O;7~Ea=h=Av(mZuqJ@}%aeJQ6pON|VTW#B7F<a_#9sg@eABR$$7Lv~W6UdILJF
za09xLHz#^8c0xO3gOWVbC(G*g@`pXnKR5eyUW`QRKeogSw0_Xms52zfqG$!EM#S`C
z1SZ=s+n{7~I*ur!gjMiW$hIn4irf(m@h(KaQdH$`?9K7Qm>{hy+DIrIEOPG$MKCat
zoRMOuZ2+Nm(AB#xv1dFxGCp$#;n)o}u-V<>o=D;>zpRpWcCz^9TpSlaKwR!4vJ=pl
z57mQ`Z4q@TEi^%Jr=4G35y#dD(4J=*o|cx1VP~a|lKDj1DqHC{;($?_e)aYr6fufs
z$-JDgg9}4m3dp^A|B)5L=Aasz1Vj6E*8O$ys}=8Ljs>Uu?~#NF8`!PahOb?g6Jvvh
zJa0)9*>I96VXnTc$K|AJ_>-JIAO^L1(6GsCMBNVxsL;h1tlYQvPz^iNr@EYzh8j_D
zNVOqKmkd=LIi36zZ@K|-Hudpa>etcYO%G+~w$N^~c%r(Z4ogJd{}(ZkU$GO}+omjZ
z!SM6TyHidMH@Rl6Cf9|#==dDu+i8U>=V-=+O!XO0e<+lyzw$92vgN_t+j^zzW$&hT
zo7CK3)}Me75Re8b|CizEU*T<Cqj0t9k>Gb{b~95KQtue1TT!>m@a(;E!YcK%;Wvj8
z6jKx$8MiBlpYbNo8@>az^QuU&x=X&1kp=7I%=(T2l!rI2_{yDEgFE`WP>bii5l0`Y
z!=t<1Zmb(jg(#AAfN=caS`nb1{<r%ah_wGOO7L<}>+gfJ=b5+ofVGE&d`ViPhtbp1
zO6AkKFjN;L1?9hHs&ZXTyLTlsKM+8J-t{)%uf*Hj3q#%=&eLBO@7d8HmhNtuaH`~g
z3k^_x80zPC<CnD4Y#+9-Z+7gLi;ZP7e5R_}v}<O9O4!$w0Mw}ey1kh;kL0_jv0?{V
zeRtR^IG9U;F<#%2)v_cfOuPBwbvp3)n#nchK8`)!!QfNhMc|xo&X``O9O2Soo)QmY
z<K^ugGAQk|ZQ&3DPhqRm-jXe>R`z9%gi?|kHA^ls1yZFfW@~(&Y?|R`q&FUU2p>q$
z;^L67$+0Ml=M4%E#%SuTWc|CF-QUxz{>OjM^)OV3+n8t7o=_ETV&9&B+Obq%oEkrC
z>6y@M==gef$0uj<27{|^PuEf}-JW82fuZLsji93zO{vl}!LK8{csS-WAtb$#%jLa8
z!Gx>V0>mLM1Gn_G$dmGpiGm2vqRL^qJD&q$j5|-LCwaKHKzPwj2bfj3|Btto%Hvht
zI1`BO(|3sTI3O|>)4!^}A!KUQd>mNd+_E~Px&?Xi6-U)yHy*kxb~{K^I6rT8Rpx>F
zu{ZcbH;i8WBAr@aHTXJNnA~kDNl<lqesJV6ZR-@Re4H)H2Gf(l)}|Px37K;R(E5+t
zvD_Lji;edv$4`yl7sGX>L<Lt$Z6CCtMj?%430pVtIW<XzhXLIVD$5xT9ceGhh%HK3
zWsc#Lq}C`vKy6~N0>+EIISH0A^@O@j<ZQQoLymX0AkHpE%xq}nFQ-wivubRkDivH{
zUcQ%(v*lUw?$n<4jtP&3<{OeYQyD(gUGPA4)4ze8BWdE$nm*tQw%UfwGK?>)C4Nwx
z<bn!YdnqE_xQEiWjf`8IW9~JZK(cu={8StpZmTkA9}RP7cW=lT_6!Mw7QNm-9AQrA
z9P81k2CpgYkHx0hKe?NZ`VI*|RPep2rQ-6o4w-YjVEg1nafz9Ji`TUXb?6j`u3nU&
zjZH>!13FUrg$4=2l-$t*yW?I`B;qGrxn<OY+Fx#`^sDRcZ0sl=oc@5p=S{jNSfc^f
zZEC>T*)R^Fd3t>y7W_-7GP6UgR($hzma~VpE`JVS9P{3^)zT8dyz8a3<6*jIRq1Kv
z<0aiPVOXsoZKVEDRkWQCWp1kbOfIPCSNKB8COretxcH+9(w~lpw-tqzoaWi3SQV08
zGp?w>K;BOW6pNItEVe<Wva-CF1JCuO(Ajcxjc6K$wXGtx4H8VRgeqP(S(<riyc;b)
zS(Bi+m}K&$q1o5)gKT7q`;uN&C7);+AP<<J@ZBc0hL@R}QM~3S6H({)P$VcVqMn8)
ztDFrLH>f;Wg{CE`^wvVNjmPv&f~sE3w#=U-tU?F%hKn+k3T!J_Sg+qbwNgYX7aZM}
zT13*3&wQIjUo^3cF)nNAS1zAKb$Blg7sVt-FBmcv*h7p4Bx}?Q-1}FJJg2dY?)O!B
zLpRo<0SrXoR}2INX2z%4{R#I-`6n2`@A)YI(J?1~2FokC?HS|k9~(X=&+1kS0#sPD
zvM#S$X=;m>^{jz1otQ=!!RTTkAE+uI3a5O3k1v{!TO`4a$Ay|<O?R3va!Y`DQj2Y<
zIvX75P~@1)JGO64tL)Bbt5^K7R7uJZLO+h4It?})GfNq|xIWgUX;kcKfjHGfWOytc
z>_9ao=$~1pyHcONeZM^wFc6$-9HK;2yXYqNyUB>5k9Q1nT8zY>wPk0Oii_$zC1}gK
zVYBZlv;Tu@*l&Bvi4^>wT?P`Lhfk#5BneQA3*NPta1HDsoymyQs7gqG2%cuk=w`0r
z{$Lo|52K%|c>DrUr-cey1bclw(tA`UHRy{fic)q&1wjc}PuEXFK*9D2Tb66mA(_>2
z>+>#>Q2pv_0L7qFbT}&#QW3n_l$wH_>!TkliO{i{6#xg`FL$Z9|9-@J;q!9dq1FaZ
zTuh7N8<h2;Q)i}0g}O)1T<<0__u0`1R9UIHClyqpDlRYQj0ia`<LNL*_ErJQuJGp^
z4c#Tmd>rs}n01Y!U(!^3<yyxi-Cv!{ub=yJy|49Fg8Qc@&vd=6c#@Ce0L7BbQkyAV
z2*xnqXf@>LHE<iS*)Oqp{{`Y?P@%~AY>{1p-05V((9~eGtd!XotPF87S15_|oQT9y
zh6Yq?YYnmghbzM%`Fed#|1RfR9-Duk8`Hyf9@%$7k3t{PGY~H0shX0nj-c$_?ps!t
zSzYaFd_xZw1*BVY_r3kfKSl^+B2HV0Oz6<W3??p#)L&n|RnH%M58f>8EaH)K`b5t?
zAC4AK4WK6FzthKU!gTU%+ijlI(8+hdjUfUjqOj<*Hq$cwpob%_%TjUx*CSp#^NwN}
z9@R24$~;FjMa>Gmx{~{WE<t*lk685b7{1AIqS=oJ(!?I5s+~lo)tPTvgb*F#aTh(I
z#L1}}QFnw4C11w16krtYu1ke0`pM$yI0O<-Y1c|%KN)?iE;USki@5d44>Xl{<{N4R
z6au626~LQo5J7JwD+|vrb$&S0de3Lx8ID1Q&wQDAL!#ndL}7wu7sGjso6-(e6`0Sc
zLw9)|v>G_LFbB0{JW4MuA;8iQF(k;V7K_R-&i&ZB<-$By4=|u<=U5SFUy9_JU_XQ5
zSbgqBv1fY@KyTK!XbE|$_@MWR5h2;%aty#zi7aiX=7~Imd<UNCMRwsOrQlHA@y@+7
z)jh#rR(VYvN{^Kc8oFL){sMooij8hm{`_oDt^Bt+)SFIFNz(9)<qwL4P6+QaIzK4A
zooA#sIFfX6UoiC0Wj0}>8}oyL>w#@LA@f@Ao|jmF?;@3Aerq5q>)Dp@o{&yeNoDBf
z7<s+-i-ltfGvSPN?3}ZAeJiI%Gnu|`y{D$pY#BS(qXNASWRY!hC;v#5_?Khj0b+;U
z%F%p0;X%n2?*ryOC0Sz?o?A3o{gt7~!}~_kQ=I05P;O({<f6V`k(UhSPg!f4roIKl
zDg}DAK9q>&bN6W3i;yQk<4NUrl@vN!lgekWO+k4wi#O#}87=N*^q`s3Z95R>ueA$z
zC+4~27>&QGzgt60D{%3Ut?M3EQXC>L=RWgrTUrCQ8-$#Mn;a+9^r}@`;4NaArLn1o
z8@X8J=}KZy=teldl7+?mGzO?G(t+jmH>1OU_Pf^>Fg6D_<;GR{aj;WY9WQkGJ*7m?
z8P*FDKBW-Ax+Mc!n`bMlo`}7=qUc?b;fil}%r*>G;_j|}^{P?me1~456GXE>%9;uX
zY4TnQkr(6}Hv6VB!L&}d0tI=Fq-b9blJkt`k1`KeIhp;H=_tFo`RdfS)z0z1A~@|f
z=<SQac(U&B$cMLtzB+{fVDRCOqaVLr$bW2PEIlLj1)G`5>Sr=jOUcXXhG6nZK|6t|
zmm0fCP#^$2@e{oG*${+r{9c-!r+dBJV!N=#0Jni6n&UI2gEoi~^ett}x%)$M!8d%^
zkhpL@TUtqibw8}X!Y22q`4M(L<OM)j$ppnolNmz5OVvQ*LaMh9vL+?>cC2X$jbV!+
z`=Az7NxY0dD0<vTc`2#LVLK1H`yUU0|3dTtXjQ!}v{85(&<ek+ESf*K$rUBm8U(E^
z-NQ|qOocz|Q=b^zRJ{K|2-Ep=u!{6k7cz3(Tl#p<&{c<MhwM_+-<L?ZTdMsj-#;_c
zL}U!H5RJ^vbJeMi@%ge~@<#b1bw3jb$M8ts-1JQvE{8w6hW}2AH;q=`?{jEih!y+B
z&a;=~mD1=G51B%*7`p1VIKQgZPlUKc)DBpefMpka#B$4iEt{?syV0J!T_F!Bz&sjS
z)^6rY@3Mpxb@ypa8nwf(w=?BzXhOYh^U^N-djj1IRcgtvpNzb-gOwa-rp)zhz)EYq
zeUMUfT1tl>mg(*Gj)ZP&hmkL54mVy7e-g;6z5V&TG)T?QbE_3FFh<@D>Q@lF9@;#4
z(e13<E9*e8;Z{A*s&{4qMqMu4baG;vnIb?;-#kchF+=CkBbHI&jPHOBcfn7f{<q`_
zt5AUx^B^{`<;-c4X}L#oi-M5e_%9yup%-s>vx_O(5rlwtXU?4HtBK8}3faQ&8fw31
z!Jvbl>u6npn7iDA!c2gRmRKY2i@VGRWAyBc2E00lbs3@y3CjntgeZ);v#+A~cEybv
z)DS3yCPP2(WWH>Gn6mds1O2#53zBM+QBG(PJwQz_{(b!eh34RONC$i!8ICMs4N|-$
zJe+~)uBI#RFff03_8J1ZbqH-hUi7E-EWrpLL}6*g?>{wvv&HZr2y&p|PfS=E6J>dP
zRQF4gPlFGs<|qy0#KvU>)gBkE1Eqn^pyBiY-n5<HKWob|z%Oj+JOwFnMcHYwD+$nK
zr3tiw!~J<1<>R;0i|+TeWeIjXSN6M%gY1vuA&nK~1GRUyu(Y)LCoOL$GHqV`Oj_H?
zCpLOD>}`jbAb4lykB-2#M>O;dd^Tny-Tv9f{Hc|YKtKf0mlyBd*@GizaDG2y+M#mN
zW@q4e;U=1~>FL^2L0PHUxog$l66-mGfi^h<9sr3VbV=%8!QpA5+Gd9fk%bTJUneX@
z?&aqS-M%NG7tLY+%sqbO-BjlCBu8?k&Ag|P^Gz+e(6J&swFt|s18ABX{yitqX_zCa
zCClY=YDz^-=(3eG?K-W7v4&LI&v)|fw;)gZ_4`nlFPf`%VMiOKw@-uxA178eXpNmK
zt&p@Wxbpeh`x8H*^eg~M*KakSF0rhPclp5m{uo{;V*m~r?H)9ti=Jy@>7D1>qW_F7
zE|emM6klY>A+3pru1D~E^fqqqF6!ZJbKcY^CfYMW?*jSwZ*cT~3MZuT7A>aKM%%0}
z`nWPa%6@1wMQ#4l^z>{bv*+R()@`U1$H<OGIPzbSiBV83GnWme6>^Z`{FKlHez{+N
zYvl6~<HF~tbXaK~qY~=5f(+{ZfGfc@kJ?AhZGwiMLG=q;o;*IeZaFuMmt}mF)$tCn
zhu>u_g&m%2Bmg?)Yw(}t$)8%ZC_YwvT59q)fcMStzt^Y!-|2^+pl!~FYQ{w=(GOQ`
zC@Dpch_Y_+0v({;xH8%P%-zus%J(sdPt5(Rh8)IL_k-FM<T#eB0ypLbuEaexEt4(E
zF1?+>$`Grm{OmU{N(6S8v#rUcDdy7U^6CI_xgmRV{Sq8GNqAvC&(Z$;d`#=C;D&`~
zNh3?LXz!+EXLFZxm;<-YzLdzm`{a2ZYC>QdVf!5H;ro)N=lIG-Kc^KFspl{n|BK_k
zw+%8b9%t@5aNO+`{akp!o9P8`?-U(i2&Wp@&kL#f+gtw6xrjA<nrA>>`{}j#0(08q
zl7I}WJ57|!K+`%kS3LG=2CG#_w3Z%R2|EKMxWaXV(uNf;thYDCVZ`g^+ZBE7*m_Ys
z^pzXx)T*=X(KJ#auz|~e@2uW|mUm=^f8VdrJ1odw&^Fq@^Q}h2mNQ%blFd#kRrSqf
z_+D*rlf2c*K?&v!+XJFHiR^ud6pCmQ(5uCf*K1sVT94eZ@xL@!D}mtNjPsESZ@BOc
z<6PAWFvk2|;neq!s!F3r)bP8Ce&r_-cK6WYk*<`9i;B|z4c-T@c>UDBtsm`wiq=3G
zq{6fwogfJ<Osr9V5?@t^SMp3_Z;j*I(l5aG0U?#j9_sFN(r`@VYQ49_yt%xCw$^y-
zw)J}JYhQQ@VMjgkaj)B1H)?D4<B5}N<n4&K3Rw8`NkR%17UUgZLO|#yx6*NApjU1?
z0nBQcvoGUuO(~e|0ZYOh|A-QLoT9GfQdAuUS7ZyCsi%?>6mMEieNalOzqVRUs-Ii`
za(jbIImKuy0m!^5j><y=TF4VRWudG6o}2dbChb?BoxPHU>KJ!&w;srHa|L!g%6}!6
zaurKj^EX3^3aIV9u{39nu@OT_e|5$fFd_s@EAv>HBOY26s%9H?19DkS*vfICXnlQ!
z4_T$w&c#B0!+@nCJR}0EPo57;8yz4|IsKqenAyp%&pr~b%ZeJ;Hi+|nr<hr}wLy(M
z-26fD>E>o;%$8dBo}{1V0ln?KJG~c;pCznI*Gnqz{!u&fbD?@p`<k-Nm#EeD{jlkz
zrf)_J%9Y+Rp&5#GF1PNz3-!yB4V{)0XicHhhy?r*h?~Gmb?_JLU;S6Xt)H0!^9L-1
z!N3#x0QndlXBKqx-MNLx&wEl?m2~O!Ir*aOp7%Oi#;;bptu8>i!lyWkm>KInRh(Yb
z#etiIMO)*A6<)uOqT^w~1g8{+*guWHWa&p)Jq>!tPor8@sDwc^a)N^0wemHVrq%qt
zBD0==&u<7QCB?aHO8@F(_$!0@?;-jBpkMm0T;c!YpM&x;XI{6oF!hmk(&NuedRtBV
zUFq?_wIcJVYQ=7ywW?^TIJL^j%&HB@J+aMr`VX>kyC|)Q%j#3fOY$l9nvu%fGV&Zk
zrEULs`L8B5NXIiHLHy<|@TEGCDSpn^Fkaw_7c=bHyLY}BVY|1FpuMHd0a70_^xvt?
zcmwV3J*lwSC42j)Sx<MN^mAKp4OL$_Tq>{ZI&i-5P|c|f;=ShjNo&GzV79I#&Z^>r
zhmf#gv_pGf@<KI#YX05-438swp#%tMq7*@T?)1kEgju|7jmYSpklSJJ(Oe2jFSQ!?
zzDwmyE-Db9o#~kCh0Efe=W(WOPXOs?<ecd1;<6^a?t&Gn8x^-+)Le<!rjS9=E$Byt
z(a~q4CPK_2uq>fA!#4U))81L?doxVey8>goES^y8LkXD8U0;R5a-*(rmy15n*iNEd
z+kEebRoPhKOUP~&-H3!q-$!!6iM6;0GGy1f2}$*X!j_(m#98}hr2^%Z5lkzR^t`E+
zuiQJbp?-&OlV}QX_R1;)Hn9qn25hwUuHP6&_rsDk{QXEe1ZJQiD4R9`_qIq@osL*#
zsJ{i%Ql%!0ARC1muXF;8Yk`ioV4AE!4&t+cAaQ^GjJ;f80d?Dk@uOYDhq<RNT&6~%
z*1`#6h~eOkw!=7m=X~+A%QmR4nE*t&M<|)Y_^xGsjpwU-kC_^U%pNOvd`u*wFp$<Q
zD$<jIVSA8Pgj`Wz^mmt{p~$DmIs$ZO->bnOUS%@3H5s|1>WI`m^g?<CbbldCsEG8h
z17ko-X?<#UP1Bg-@T$c?5cp#v<bQAm;P)-V8%&Izlrg4GaD3;zYWyo3c;)qt5Vx_U
z%$s}Jv9b2<l7%i{p3Pb#x{RmC(SWXx^~v*Pl~=d6YR2oJ-!^g{V3N)ZedP!V^Qzn{
zu>K|L`D&p8o7E>xHs|&ZV&Lu<fK>MdDAqs!1YoHD-GV*_DCqlsLL@?O27v|j4~@hB
zdmeWTmg-cWxnKVPp=3PYTg%AgCfGP!6Po9j)$IG;DTtd5YR{z<6g}GUxITL>J5RdW
zw_G&X7-hAgn>QF18`szK^SiwBbLaTW&+}wOf@+56{{HalbNfwRnP`22HEeUlD0R8q
zl)E4Er6|yT<EgP=#A(w)RcyNK<TQPa$?1I)^Mm9}d1tCRi#aw^TRVka+2$w<c0C$}
zyWO(2k^6qdc;WbP=j!VF@z>e|t=rz+@;86=U1&t0!5CkcPiyd#M$`MB1|A?Nofd{G
z4b(o!?9Q?)7h7eK5E&C*GU;gS9ce>yf}i*ZYr?c}Oly`-oe#2l$oWTSzk<Hkx25f(
zJOp}f9ty*31F}clYGJuDH}yH#+O~(oNk_;qe{md79S-R0$7qa{j+GMAWt!eL(ra!Q
zBt9zg`@DK##&$$&%=F&3v_zKi4pk>uk*O@f1vB&JNb<-EovGyGC_Y<cQmD{K2gicL
ze^7jJ$iozg<jjVAN0~9z#yis%tCBnD0Em=Dxs^cEcSG?>oM5DeKh@64GI<Acvx?hS
zbc^4o59ZBB;3fLld0&lMNtS!Ps(0zM`Z8jKUTZ!XsYW#a2EI_MfWP=4640y8Ig{eh
zh!n0;TJ~GRx9apex*D~X4hml{u2u>Pa^-KtNug$uZxk;PJo1%tmN{aFSs#DuEL}h_
z%X56(X(XMoO6C={i=A<NNvifnRf-vG_)4X~rX#WbU4UgAklNKt9$R+~$a~R`I#LU&
z`T;}?>m2`=@HIb>;TJq@m`Ancy!3t{<vIIF*_}<HL!wms&+pbcJacYr*FTKdC2}iB
zTz=3JA(GJn%zyA!{^M@;m)q=r{NLBJKnO}STvX(pfP0u;tc$!8WsVjaszE&mpZ_2y
z^s@g()TC;hW%Mn!*?Epxd;fEu{sYbEi4aZ;?lt??cuwNzUbo9k$z}}JPqBZ8s5$>u
z<LY|Q)19rorH^KHp~d03&1(0vHft^cF~<4cU(ntE=L6EuU>jkhg%yvei3PlqY*6qI
zR}KM&p&Qd{izmAUIYV*MDU6{}4aeS^CTDU|&PfQPX{Urv_obPIW$3$|sq=Q@)imMi
ze;YAp1BiwFH6*AS@_#<ldF2~*+mfA1M&uVtsfB9M)^1UKe45P~Wvg9lXDBeM%(*m>
zjncNHIm>$~jKSFb6ApQ2z!!R}j`@o;qjOQzcAE7Ty?2Y9hjAKUgP`=e^-3tpE8EC+
zjnnx1yBiU__m<erdvzn?vVpxN^TdCa2L1e-TE5Oa7fADg@+J8ZY|r}P>{b{D{9B%F
zyQAkLy`#ZX2_7}(SD_6s4zxLpm2@)^OMLidt@^y^dxs}%HmwJ0dweUf%-sv@_b-Xb
zUue8iKf#M#@8q>(TIhisT2{1gq6_F*Ib$*S>UKkuk-Z#3_*HvAKbxKFxo*o@60HX?
z(C9DcS%cvzIK3F2(=m+SE3P!dF_0IXV4q>Y<d^vmiZdjgmxVV9M(zyXb}Su#ajJkk
z?c*ne0ll3mfrXLjCyf=;o{S{wSo3O?Y4uS+*cHo;PsHI>G@5qn3k<FaIG&q;J&Ih&
z!M)F~U`<dT7f#UeE5K>*kK(eL++IOO5Lo2!*lW9Mw;xOlGBiXM_;CUri7LfPS2Q?j
zg2P`?5<*F;q}CPF5uD@x7d)~-?)7WqPuBM-9Y-D6V(+e2ZoycQo{sD;(2i@?q)5k@
zpD}UtE1o($<t=mvmJ8P5M{y7oh&rC4*rB>o0pCQuGQN&yBJYLdJVq6S#@v(l2&NL0
zA)Tw$Y}F2oCZLr`qMg(V_EMG_fK&s>4ZywsUz#xZ^*p+fI;Am>eNI(|qZm$wp9PC)
zIVIkC4r1~bn;@LQj`y>5H&4ij=dhiSn@!#uBN#Qs*Wc)~-xCs_)_L{rYnZf?!L$^?
zBb~4+=aj7+a`~N1oNCvd)vVEBGBRso#oftpyrt+aN1(20T&&JvlH=VKvld{n{cVZ-
z`F=gU!AidOoDA-$9-sd^aI3wUWV4+0bGL7@`yY?Jf0s`D|8*{A)yMs5^;-V($mzU6
zU6n^7%Nj73w>2fv-|rQRP??VoGEW#Ask*U0VC6l8O6AyW8x&b^V)u4YdJ&f)4;)-N
zSVq4Ck^M{8&w%!qAG~OUDJ35K#?42l;MPLnjj@Jn)t{2pF9N;CLLX?Zr>Z)-uU5|5
zHpj%Ta9i@keCs1wpJTXHRS(Fs)|0QXf+OVRCbtR$>+qNP#}1YzHN0u5wVksg&FYaL
zqi#pPGo5YD52M0(&+m@+GipQc*?s?c_Mb%8f8JJ8^0Tzh8Lul_sD-+3(@1Jxas$82
z$xgZTskb&VUQx=w!Jh{o8*S&$kU*r?yi_vbx4h`sgn<L84B{+F7gzH*egxK^S*~M`
zw$&9rw|j^WJr2o7F?mAo<1=HfCl%V_?A#&4La*sg?6GeF6n=TX_m;k;PKdhn8j^b7
z-Y&vxM`k@V=*mvoW+x8JYF}tsQu$zG!P819F8Y>W<Fc|JFAlmt3E=W$jIB-V_%o3&
za*Mf9to54(rfK-ReacWw){y*ZXLA{mar@RT{iEwLC+j7CLT~6_p|{1PTc14G8?WZI
zF7>1BQ^MLRJ89RavW`d6(^f_nBf~#&0kGB)N~#2`@6Sqd$9Xi@uvzu)-S6&<c(!!q
zwvxvQ$W?m7U6#|2&R!b92E8p<<>4Ak9h9sY%NI4f{a|A<kG`e8kRU<ywVUR?zY91T
zj+YnNh~>D^ol2W(UxURi)pMCWA1R1aA6GquHP!L-;1ttNH%YD<vSoA-3nF{Z3nk$A
znr3K7YW;I)+<xBa-G}13Ml@LiREeuX_>NM$$4OawBH~7w`8|cN=d=VL|Jpfd`;z(x
z$<}}T_(lIKo>ioan51wOIrqx<zS*BG&a`T8XbbS?eUZH#l^*nq>h@G@ps%&<I{Yg@
zL=03E4#viqs%J~ba(;)eA20A_&a9YZeM5vCBrBFwg_1`syM$xJX)aTJ2C40K4c`j4
zNY^XsWesG%Ez@H2CvL-UOubI{{B7!~O8BoKYTF6HAuF&BL<uL3Kh!?{J!|6s)MGB6
z2s+yabbZD46sT`UPnr)2osLaoei|Uhv6%<f?7vd4rP|JC#XO_RIJeo=qH~lMo0?f&
z68~KzT10`~E~rj&0IC9{Cx7a&WIH}f85Fjxs0_CZx?2_(sG)pG+*bl)BHdt!7~So3
z3*g$6Q@uLC+QOj(m|He%OCZ7yeFEX_Z1K00-myKIogE2o#huykzs#+)u^lc3{&;=b
zx50bkyGs~ezq^Cwv9gdtk8^4itJ!Jx%PO}s3PJz`;Qj9kNk6}S=a~zNYy>zqD8hAy
zVde46SGwIBVr^6J8M@sh^-AG7jf1*2?Jeq7r8XfAQdi~xX8H?73IA`h^{=jd&zQ|a
zZYbQoizwD;YOEikEbv9<*&H0v?u@PKx#*i@O$TAle<2C&e`S<?4)3YwC!NP3n!6<d
z+{W1KiHGn|B|~Qo_b5%&2CQh_lo$JaI4w~XSiA{Mb$Y<!RpM+MTs8HywjSH~>Hi4i
zieY6a$~CdoYMzu^NWXK37UAfJuBakKI>fg-Uy4=$jS%Z+fhH=04(@z8&!1DE<O%IW
zI?ueX0&Y2;fHiN$uIH=!6KBR4RAnZH!hMI_;0hlzh2CGSh;@sw5S)oZ4Xt9FdBhuu
z%C?6E*FZzw5(^k8ckEz!(Y6n`_c$3>5he-Yv|aH_S#>}9NK;edc*!1ZIErdPQ~Ijm
z`5HeN&Hy<5gY|{LinA3l3`d#I;Q;pY3TU@Krj^st2i7gO;c?K?Wkl&yc7t<2D4x&>
z;eB@+CmTiUjOn*>_KE5XBMS#0^MIagGlg;T90?dYGV3!q0=;qVxv1ahl>xi!=O{>}
z7ua}#KAS4E7uAu$e(#aEp3RZ#U|cVdH(dMO7yXZ|yZbcuh!ze=(A5eJl}#W(*=MuA
zC}5v0AIdis&=s+_e#kInpy35;?s+AxC^IbV#>trp#as*jelL4dV!DpEQI^J^y>~-$
zbRpXE<*lu#5Wb!K)cN@VL#7SOr>^w0E7Pn+M4k`N%FBsw?O|HZ@JFFSIt8i!A<F*G
z2Q^dvS!M8l+22n=S^B0_w3H`>k_KH3gvD2OnvX(PZGRoJcYKtevr+W3pzAZ?$@rW<
zk^X7{u?ZNyv|nsk;8y#x4j<0alVS;$Po1j-00BS@;LqvQHhpBW7@jKlMggWD8$I1`
zCT-d}0-j=?^u0Y?rr^_1o?uv(Qd(ole(H&IWnsWk!;Dsi*wnp}Ca>{E$<E_@1Ik6w
zi>t~NWr5BHZzxWT>In*KwAY=lk1w(d#njT{T~zU9d)=|(H&```&PF&bW-CBPLO1Q^
zu3dB0DEK**`-k|${Covo;O3q&<Tn_r{U=mDD`YU@KDy*_aSXX{7X|&duy+?+Gs~1K
z&fL}CjAV8kMzX~XZK6uoTu?Sql$VnS8GEycUR%2;<d}heo{#|5DX}YP2Lq)vUW(;O
z=BMxcQ)|t!2kXdL)LXII0=w#4`ChEUc|8M;YR474ctOoFu4z^>3871dIpN<=b5DCh
zw-LaQxWtYRW~O<Di6k<ZTHPyUun}eCGQOp#liVdYG6u-sA{?hKn&=ea%f9rBf|rhN
zP$Qb0U0}C;jgQ~?I^sGOy$aTm#{D`mT>GJ>8uj^~olIFe%RaC@jh)R|7217hs`(CY
z4|3#ufG%UT6@_iv-~`Nys&o#eOd*bIy~hQ8Qh?}eq6J4pf*12fY@@5oFJFV0z0urc
z5a|itv61_pPywZW^2k&6H|4v$mEYRfQ6*#s|CvP5H#~Cm>|@+lmN8n3Ga?aw`XB94
z%`PbVog3wQ&yn?63wTjoAa3!lKlj=PF>DSUTqF1`jY#S_2K#Dcc-BCSHdr553~uOK
zca^`B;b~>gfF*Q@XYrlOlj=jNrKcBu`fm<h(A#s7+_uw{{Td593rTM<Ag`V;InJ8U
zCl0U&H*RS^IOYFrH9Z|Mtr#~;K7n5fE29tbh_7(}j&CC*j=h{XAMx6sYoySVwUvcO
zwy%G`dr-Y!-=zbjHWl&xtaY?&AftgIF^~e;QXK1-k-?%jqH*z~d-}i67yT>E?Elq{
z+dpev<oN(7_-k2Ajo6xAc4j(uGq9=+X6vF~!Rw9i)Ro;dVNn)rytJP$tlCKEiYNV`
zC?Yej^f%?$N}SVP4U-ixP`jze_H>X#ZBzW$q(^6{G|9vTlGw3*S-7@CH<mb5EpTE9
zO;4s>Nj^PTC_4ByDcF3zgnmzb8GL^0dTqe;{kdqJhzE2G`Y-f}It{f{*sND1%WTi1
zhQ;q5(NhZ`t7vieiDURV!!W-lt%KzJgH^%I7~1agBuD1?dbf_x^n5vaZh!Hc{p<Cn
z_&qhiGjVaT$uO}FZ-y<?%Xs*{GSK2(^Rfu)P@vbD&uC73GuM;H&oxrfXF|B)FC>zi
z_f$|uy`Aj$-Fr~LjJI7+T6LkKd25Wm1pt~HEaQdD3zt`=sn>K{F(oKw%ytX@eqTp8
zXU;cOHr`cQ|4Dqn*;71CR<%ahto4Lh(U29F+XEx2Lsr+ukGk);n@=vH;qAWac7r80
zhzb0$`OoMz`kkAoZ{Uku25iqsV>r=f7!yglI>Me3T3MoFKVv=p@P^-jjMP5WPjyL!
zg5C0JM1<rIisBmIV!L-v9K>}>DJ+^5Y_`SaEB%8)Z2-vQFuVirqdA&bqowu`YB-^{
zw$C<s&S7n=!Zs|N0_ZdRL*8^c(}|>QKyS?Ks6G%iMSJb+;@Q><eXCpBH;%nQ#pH|l
zt>9WWvkGsP0!IY3>-bmc<87~k?X#;lVQ=f1@X1xPewuEyRN9pTB41lSP(O)UfoYKd
zMTmfxa1Q2LcT@2%vb8|=B^o~`O}p3L?63UM(=no9I@1IU=jnNXh#{9MqRur*+8N3O
zb`XIwQ(|je=sTIm7lD&*>XfsU{Ny|(vPf|p;BnGb@wN9zDs-L@CgpogRh@zAoJhF-
zQKH*ZL;2kWw|Sssa!v$FrvH+t!C8yC0J-!EuPAG$DswA}D19DT9CvV}>nls|8WXQj
z%^~_SFHl%Lh|sZv^~)HPcGjT2MR3sh^o3vkmFXdz>)m)GSI=TcHYNHXi7T-|d_N&9
zO2c_9^9h0V>pst^zW$Cr=8Ux)qYFI+uCAJfErNn=0P+1l*UPE<I~}9=zl~md$)Llz
znf8H&SQ&or@8I^6|BwBC9y)F;-SL3aV!fd%uUiq$+E@rJp^F>Sz8Z4<USUe?zSMwP
zZO-zZigQ3zW1o>>#EbTXio*nA-5Psr-USs}?=LwVq4?Y5PLpc<sbv>5UpTd@@qeaM
z#af3OeI;`yZVGvch_Km5nnhngV)@346lq8gtML0sNsr;pgvZX-)Y~%q3KFKTvi0xg
z8mQa4u*IL#$rGa4%N?f=UiX!Ha$}dka`wdc-kaTKb0R)7_X-QlRYK<%0aCXSp~zS9
zKuu)S+~411jEFW|6H>k={Ij0S{#P#Ii5EY$hLUr=&eV2dt=AW3VWHVQdA1DAUKQt@
z?1{x$gMlsZ0h_{n5mAPK>|9JH=Gu@P+fc@%QZGT3wJy>oA%l35kk`;RIpb-zso>vm
z!NSs4^%Cq(ZM@XZ^`g(K{-F2GIV%UUz8`q)yEJ&5R-#~KFXC=fov8u@IK^=2CZXHH
zds7_4gvzg<uxFGHPVJ&boFf^FJbkpgQZ`t3gs&Yu1BSg9ZW_y}`R{WWqe@0ePGz;p
z9e0IGw~g9v(^JQR-=mxf)<g?{fQZ4ARmATQt<{4m#*Lpyh-Q*Cd{E6`0$)K~KPT#x
zUAWC!^nOKCJs|hes&=S~75$3l@D;A!E}LS+o=rM7Uag61&2DibjI56OP6SuxCT8%l
zSL|C9hyv2zD!$DRMZh-_6!W_6kG+x3B*o)8d0K2v!G_!W%wCk$<$HmmY<xpKcVSA)
z<g5Mi_fkshw0bn#z6)C!(67Kky8sHXfcBLBli{r(c`q#J^kn+@^&+p#>W|oItrtWt
zXxpMc`|O&}p&wLZ-y4N?BP17czFAXJ!!MH9iSoa46AlMv!4t$5ZyKC3YTU@~$eRWn
z|NaAd_Lq^qWZL6uGM8skr-_Pzuk1#PNCXK)IE(6Vp*#0|`@lmrSpkE>plAIQLASqQ
z6=#HY2FNFNb(s@V33)1b?vEbc4q5hYj+Z7QinH5}E*gGPQRN+U7Q$XTr|ny!=Wkuc
zE+CZk90)t-f8E7@cXbLCz|MqJv%V57c(zCVuHxdCLn<8VApBu^THeW}NO^U>*R>KJ
zIZDqCoK?h%K3|<bAPEH$`0(0~@r|a2tB*@M!c0V%{qKOQqFUgCaRqzoT;U}XkC2Z$
zl=b?2!C9W}Q<#AQ-m>r{m$M`6Qb}PSV<7hm^KF3IB<J#9(4F7z^kRNM9*|(WD;XY0
zY!W550@7;P<JeQgkqf9{UC?k!4D?Iz){K$pFfazgk-@GIC9Tr(cTqGJe8E>L6{dP^
zv&{M=&HRVE8|!ernK|nKkrV?nJabyh6x+p~7323|EP`oTfdW#?-Ionxbw4IddjK1b
zXRH4IZcpWZvw!pZ&ZWz_aa{-#ee1zkpeev0b@I4pNBrovh?IKEa0M7NSkZ+8M6W3K
z>L4EMmpyv=Js%eUU6NYS`cqa3H$NZhlBUvZpoa9e_h3KG%kO6Cpc_Oh%yeR&6<o3r
zHZ7~6X~nDlgTiq5@UXV)cqUV5gen1kKJ0a<)cJGr6asroU@pm?3wFuO0x50ZbT*S^
z2=0`<!}7V&4*Ug{X$s8j7I&7vMFlC*z1A8@CNU1|ZGaojO0}~k*!QZCiZf-d0;$rj
zU_awxsF}AIuqc}T)ea3a^H4f#A6wz22+z?c`ENK62TLC(8K&?E;92+Y65R0Trd=|_
zdB_m7hmCkEj*oeJs|hgg1cWtRsdM;nW$Usn@6X(uyfwD7Qc43|w&ypilR=es5E^AV
zHaOg>u=QPKfFHt6YtsVTL%rf~q#NM)g+~E?@$$N}2BGk3=FN{)U!!f*^89F>#{gl@
zq9I8?uyJJP;4}NO)9WH;jOHte_#YH+V8!%}$Tu_t?HQkBiJ%+h5$1RIbk|@xmeYK_
zpVHnxcJiCcB_^Xoc4Q(rb2zInq6>o;&m>@lwUwhRiUvi5KRr@urw0Xk16Uj4S!75u
z<2kPC$;7&1nj5V86!h85>&fCk_UAlpG~Vfg>0!*6Z`{;xKQjsVbG4(+;icD+c<BbG
z_G??gomX6vyncC2K3P9}8%?u>U&xVl*^Yids>&zNwR%2mJ-Hd1kU8^m%O9Yw1%!1u
z=!voryk!a~7VTw9+7Kt5Oxsh*lVBRDx);XNqnmTbH<@qlEutkwcex>I?=m)%JZ!$s
zUHsZkYJU98nu`4Z7_r5WmBV~@Q!!-LtjSGBb%mBb<6@6J)pbZPS<>@jGbgJC-Za5Q
z0@2A_c^Y8AuU}3Uirmvme{~#_NP1XSo!9~qAGQ@vCzEdIkj*Ndk$F9{AgxhYVa=OA
zC?Jmz1-|io4WrWpW8H*<G(>yaP*&6TncA!aiGBMt%hj72sCE<$c@W0vd#z$!@siiA
zU82?@x@#oW7J++AAENh_dgaCzVyZ$1mDJ>En>2u^`gpz05aEW_?#*iQG!XR{_UBun
zWyW0tywqV$7ozr2^vzCd<&Ue5g}evI=QT@DIL&Hb-fSR(`fKezQZ`)}5ActUT<sg<
z+MyI)a*R=&O08GzN6e;9v)58?kU@!!BhNA+D<%#|q8I5#6=(%|pcgzoG;;(^FnYZ+
z4rXD5^ue~P1|U-3bfDq7I|Dn4Z&Pb|z<}pC&Fd2=6YJ^rX5--fE`qu=<U)sHd4aKL
zaLJpk*Es-@Tb&e+E)|QkD<bx7#6rVS8$P(4YR?b=oaN5dJuawck%lmUh&VWSU4DT4
z(i(UMdRug@5_Wh9s^dW`dx1z&!3|qSY=EcwIlwMGsF(nlYb$}sccs9iRmfvX$PbDW
zh<R|~Q}hy+*tb>+jY2dU_zHt|fGaV7@uB1R9L5l+qMBw7Fu>a}vqJ9Df=R$NeOF%-
z1UULd5&PDepyASh0eyxy;ZJYt9Qaw!>pnL;4LUGSH$V86&r>_ZEO0Ad_VYQ9URd|v
z^|55eX!%bd&YG{GCXpF2i738}Zd}04mMZtwFGQNSRLtbU#-BfZ2)P$fp6eMsUOpHX
zS|p+-CNN>Qw?!PXjwXPCGqr_UT|ga+0$0!y>EW0@@u)&er6XGGX@p|j5FJyGov+LT
z(hyqWw$*$7EnSddqZ1RcHitzc3vPe=+b-%BAq1z=*~&3gpI$)^7r}`zxms?j*7=$P
zw5|QY8r-cSh571-)#(=z`2p&w(mt;ugz6K>=XE$0VJ|vBK0sN&UN*#Y8q0w{z?JXC
zf)9D=H)>CMgu{S$9JL?xHf_R2)`ErQZSh&C-BQj$UKtL!0yr(sgC7*SnA4GKuJpV4
zD?e|eD`viRC$H{yy#PMZW^nK!5%x`;JRN!XL0#c1stgfzu)usc=LCFE8hq)8?Hl%(
zquh$!e0a~q4mJ#p-*U5b+$&QlZ1Z$U?sPl7Cc6Ew!}xAM#0Wx^z_+5zA0BxZmM}6?
z!PnfmAb8ZBh$6Dw7fKvN+d)DMbECj-xPgsCKEZlhY6jAHV;t6$cY4QhSA`KPFKD`M
zM2tH6{?v_KNK-xpH0=h<LA54Dh9%#sHiI=)M}ns+osR9oW*BRsP6s5_93;^jfk&w*
zvS90n1;GSrpa@Mmw#w>{Ap(-@prYc{_{RdaA{b@vv9&8ZSWsZ;WD<(Oc9^iX_~^91
z;)6NS3*85M8ydkWH3)nzl%wpp_X>RNL&->2T;yKixIarC`GEC6l{6Mx`C(!zgMchd
zfPoY_4JpyB9G4w8-=-usppSF?+4WB%FxN{fTS@phLo^?|l5?Zeo@CNvcTLj*F!eTp
z<Qaml;zEVoq_sU&zq7}&v_C?4YLDDq3qh<U=8(&x+=za@qGiB@q<7QI-uKhsVYH+;
za(!S(-EW9*S*SNMsOpBl0ARp8F?>t|x-chev<beq=SbjB2;kK|PD)>mu5c{9$K)SW
z>1R$pubka_nFAJhBO|h?%0v4I+9F`>bh*K(W9c~ddZ8+&z-sPTFEaZ9c_9zEVj9SE
z8Z!xQg(Q)*5I8!oCH*5bO~G>19`gu^$-G6vx)V2@Kt(JDGr#p9B7E;uYVnqrRw!n-
z)QUm~*4S>%hHscDbw1*LA-W5y{gf!>t1|dPjNtBosCNexKBf)m&tv>$bw={h-S3LV
z;$T5q)*89~<49&Vx=VPPa?oCr0NJc6Ty;N@kdb#%gk<d}rkFyt&0J+o<Ra<SQ?z2{
zmxv?hg*-HlUy*N^w4rVK3kt3%z&7rWk|cEUp2suWEDIq5eo!o`N7!l`X&f&_z)1J;
zkFpmw%Q~y{_pY^(b>qNqq~uV0)q$PImX3H-ekq_3ErRxn+yGrv%JRm}bhXaIvr2%C
z>e+bH*Ga-CguL+qwq~FnR%uPU*-|`E3R~f1DGqCZ&yt{rNZPWd?t`qj`r@19vOXKK
z<e=-ZtdM7x6PFi-gJNd=LN9RRee`FE-{#-kYtK>|6ck_DD_jK;t^<bA{BydF6%rnf
zrP6?-DjSMH#`)Ik$&3}yPEj8ZsCCfDXo;*kEUBrC0D%hrhdNaMoPF`TYl`2&<oIuR
zEoMi_WyjcHV`(dH_$PG#^%Ao?GoXWDyETEcQ~}~bsh@1Mcduw_$ruV{3`(+XDtaQ^
z$)Cbq4p@rE%g1zbR+5KSwevWVxVGPhFf(ZyMPCgsV=wx6(un{KTtRY6`SiSLeP}P4
z#G25qe~*m!YMC3a8!q9+4JYt?8P6S4saUUhIJyK<>u`!K=vddP(0=HAqJ=&ygSp#t
zr!}7EL_h7uoP)Cy7w;Ch@U+Hk4~H#Trmz2~<>7v`#1MT3RMN52PuUutv<qld2K*i>
z-hb9`Q_F{DTw{9Cgr|B*Qoy!M<u8|eC%MDlG;-|BT=ee%)6_y7@W3Fo>ym$ofeLeL
z??_>vX$0tI%TwV!@BZ)_=|ip|u|YV3@f;~K3Ydl_@IeAIJ8g~@Q_d-z9>@WGF67h$
z)`278)uO$0Glv5eZj$OM88C=UtuGAbTH0lf=S>>oJL0j}Hl`sr*KvXY3j`vg?;{*3
z(B2ijkpXJ8k~xNl*B&P+^64*i<J>=^fu$MlU%9E6nb^UZ`)ZGQH`Z69`%-~k?6sv_
z(uXBT5U$@NY@8U`aty)1-tHPFZPA-fyc}^x1iEM&_E5$~pVJI}Y!1f2$~EAQd^6=Q
ze^5+^Kz4I0IL^d*)>!qge?*KFBLaOSeYFzFw-~?=jCLN*C(z^(;}G<Kw0#yJQMc3G
zN;=hV51w*t8(9xLXh5G`z4JEwOyAB3!%I@&QwZ@k=~mSb3Ihxr2w5+`pdKMZWu!?D
zB@@yhqzK#RLvUglbG`MvwPNsgI!S0R6Efz3ah>sLc#fC>0A-FWzV$d<GOP~6kJ#D)
zKCLV2;`aAapf~_#0X`&QY9S9=f!v*UJ2doI&vA4IbCC;qP8B#2BO+Xk47|WudQu0F
z`piB=&KV+_9(5p=h9K2oQWSV~TJ$jO0KE+sx~U?xr$lD!nLRi@F91AM91#w}wIG1x
zdRY@fIv4YEQvGkweD{N5s1~~Or?b|F9%}+gm?S|Fy!hv18~0@hcmDFo1}G~&#RB<G
zF47foco~uBiCT=DBdac0f&%^XQ0oX>h^(gym~=LVWCt8;fqaeM#l?aT4u4RzWDq{Z
z?WSpOM*g6<N*#nZu>`e}B0EICl65M9?+P_kvo)56x&GEtc&asU0~^46W8Kn$f&`Do
zfdUi(7jJ0wz$~tytTqZcc^LggJkr4{{~9*lfF<DzlB`a)dK=^<x`6^-v70-KSa8U3
z3w*#yK9o9|M_7Y`W<=zL9k)M5z8@|_eM6o4|Ju9msHXBZjiMsbR6yxLsuB<bgpLuB
zsu+5UfPmCUm);Zvl#Vp%(i2ETgn&kh(xpfTr6^50F`-DLx!2j9+1an`d^+=;ojJ=N
zIrrwClyj5!{r%qZJn!?)>NRn>c&gA4no+I&Dj-85yEEt&+eGokAd&>*lCSzok>wWj
z;ZR2Ms@hv4Yr);;*~HR!=|TI*Zhc$^kv7l1QWOdW71*oh&c{9O?-SPh$);S}n`DD*
z4*>xbY`6cNO9y@cz!_vm4F)m}dN7bb_no!*O5rJvIPu}Izr9~9Z-ZCw&ha{n6nHuO
zg_T1cVdT+n;(FB`Uqt=+c+_rE3yi4*HZh^Io)DBP8B5#b@8P@x_*&qlhAo)CN;bRv
zDHI3L5P&t8Hf%w*wSf!lD2^>r^3WnYIV6`G2(u%+uHI&-e^xLVc72ya9=77V3cUtZ
z!m5%%iGblgg*QdUWYO1ZFN_2=<3$8NgUG3<;?%YECtdchrdCS(YC0F}M*R}5CCOju
zoW>q;iG5wynbs0UM31$>0S+MVY8?=n2YZlbRu7r_sa1SVgpm!#vbPz%8kKg>SL+vK
zZ%%`@!Jjn1?b-K?w?Ctq=b1c&!}nK3Zgt&T4jRu8hE2^j#!KzE3vPqW09l&$9MSE7
z#SO==6xXgX%z&3nz+wK**<HJqY->P@F?jioxfRK_ZLp4bwJv-ywqj9s3iO%(N|4aJ
zM9^FFiL8X^o)6|$yMd{QZ<H{@Y{Q81wzj^eu1{t7HlntdUi?Q%*wdl<EhZr5JDi;D
z5>z9>oi<cO-bb7q+A$BgmGhP2=0UGD!vy(}b7<0}z*map*<D+m`iUtVbc6RnC0Thq
zQj(4;RN~pgTVOipcXqN)>9J3Z5F=AX$+z_vz%hk9HRh9g&I6T5HoYl~WlC#M#%|rC
z{8a|5WW>x<=t%)!qyf;qI&)_a4fb8JiYm0)i@H*oR!HN_`{742V@eInTxK6;%n2LJ
zF|DmGF9Fqtg<>=NC*};9fzF@kabs|%!30Rs`4*6}o!MOZuyvwO{ASDtlde(~F&)E5
zh&}d=RZG^szUqL2o#b_mw%cEHMfC&<h=Re(@-&Ycibg_xE_C$mi-pu6bMq~ptGy9}
zq{$dcL2)!MjM~ka>@_j9NuOJrHfbNKw=A0ACfxj~2l&WwX^Clx2}V9yLBp(desOM0
zpG)1ecrVbn<gusaW)y%|!IV{s{W@T3Cb2l6FL~MI9)Je<E*!=W)T+3TydIL*FYtD$
zhqAXWcXZ;IG9Cmb6)}-I6WB2i$CqVFPmx~`oxW1Eu&rnAX5T_x<&C!iCP>S$U=}nq
zq=LVRGA^D>P9AQ$o4lIag_N}j*HUC|@&=6)V#l#60T&^&V-it4;<uvJ$M>y0u4=%`
zbW6x-%D>qVI52D-_c7Y8@1ssdTEA#FaI?S3aROj{TEBkmp}Bt{rhU(UFKEWVQ(jtO
z<h0TbOifa@*EVF;peI?{E4-0i)y(SHo=@5(&J+*1wQ;IlKfDG_3YZ|zj1sd&x}=xQ
zLgjUQuQcmzag4cqV8M8dslHkrnKNHxPpzzEx0{XlL=Y~--P8yTpi?r98Q{IKA@(lI
zoJil*c;fv@o534#7Z694&V6~g&bMlo7jZS?7V|jlcr7=%RGpk)%^Y3Pt;o7Eap$@E
zh85pbY=O-*s>xh?`wGzauT16d7!IgNarKe9a=V>n_XI(xYq09adjG}v!V=q@`6{-s
z9CyylBs?pBmd%l3iY~r^;_Y={!>aPPttNJb09$h=&kDlcL(r@Lz*WOr*Y}4+wB_$)
zTnlIRvU*BR1xgvlWQ*~gi<d<Xy<cq9ZQSCsVpj!?H())oDS_xK?*mOB#Gw)|u_5Li
zgaE{{oVqc5pZBXw3Dhy5iJklvJ)MjdI?b<oy2nc>;}n;W9h)7W+=FZ|L7GBs3#>-6
zPg((84BrToFA2NUvn2f1l<m@L5xm42r@#eYYKf0A(Tg^uZ+R{=Q=49DAU0#DDrevc
z^l&i$xLtTg&rQ6E`}mV;S9|maZ?jDio~fmcX7MM>R9+yj;cEBElZ^}~=f6@M$6&Za
z?`o;nvhe8MPb95!g-2|>kja`HJagmzqxi%V(J>lnZQ}Z(TusK!9i-UXY_N!m`BIzy
z>c+kTH&$E0Kq8!*%QfwD+v(yWdbO2N&9QG%Yv1kK-`sr9y?@BPe_jg2)OxLj7FYx5
ztUh>qn2S1jW)rfKs+8@IJShBh;&MdtETbz@q{Q+PnBb7qobBkv=cc(|jkFTg5O2EO
zoxh0dKexdg?=vxQvx2T_Cjs#ME7tjIkR++@scKbB(Dga5v^)0;SB_RTvI<n!O)xZz
zz+2d=hXa;l^p(y%yT=&C6;_kH(lkZKL9&J)IM<fe+*2)5Kn|Ro2{_*)GX31Kmzi^?
z2AqS|(9rfhR)Fa=0e1z8`om%LOCtXFsz+Xgeet?A%m6F6#YIWF8z@cO_Ra8IzR9u-
zsd_BOHHp0{Kbi^T!I?k~ZgTi%;N-t@1h2;K9{ECiu&7P^6fW;&#gw3G&#Q<!2NKq}
zVAM5<TyrazzdVr%Zf4_rAB{C?K+UJvKSlZ^iapunT2E5*Q+pP3_RZ|-WsPFvU=;uT
zmz3zzL52u(q8gl*9NBo>+*_(~#oHa8kiGO}5*b{i@GvWDU}$?Eu(FtVgC?ACC(0+W
ziq+;Q#IpxuM_6*B!;7cGwvp|dxXW+nKHpn{ji<w?)<y3EBmJ@I3scqv9Hd_dvV9K!
zxI2JcT4Eok#ECi9K#Qb>p!8xOq57N@oT2Ay>{^nAPR}|KCIGqPZ-K)vtel_Di-(gw
zeRj%r_juAi-E3~Tev>}9I{RfEXLeJdWP*BD<H#$)_QO>ChZo?3<w<Hy`J2%KZ&D^_
zV@0>AbFiJ!sn2}qyCgy?VZQK?TG{8MmN5UV)*Fb2>2%_>e5tT{zsVOOhU&OM%#HGz
zdz#MN4QGnxHv=kD9Fhm)@{e-Y=h7a@GIn1zzktVuD`XtMCcJz^qP@OVm8UQ_;+sJ8
z_fJFmug>otw(mBu;{WaktMD1I<#;90986>veg54Fx+j}b8pF92q3TK#H=d!Ld{Uz1
zsVpWhYde;_6K<<>yC`NjuLH<2RE88PTG?coR#W3tpRAXv+%RSeqloow#}l{3f-QGO
zd_5m8ct5Q20gDT2uATQMQMI%OFtmu|^OseZ85~RH1wpKUYTBY?)&7M+*7~I5+%DLD
zTb$KmHq}~W=`khDsD0Z;*^?lMc(Ce4WqXI=MA`G{?mjZ_I8GbwUNh8`Tz4E}Z5uf;
zSISGaC%xWAK2|S;UvFzF=ZLqiy*4E>Z}!~eqJsm8QPT-n6MrRO`9q|liCrOopvn^J
zmCv{6;3V*HIVxkBj^M!0&+JKKN<Ud0J4w~zOupo=%$0N^t112~g&$wYsTAq9Aq26m
z1w_mu+sJL`$IQAK^HLQ-t<1^JkI3mJA$t8cll@S7LJ!o&4culjbPL-I;JQKZ>PVfk
zuo)h%Gg{}il`ZE2DY~~|Xn2tQjo=mAvo_bI_WMra4^J6c=_wzHILG8!WRQU#o(Ax4
z2EPuEf8=~`&;4C{?$3Mg&&N{~L=|OQ2a|)zXvP+Kp}B^GYV7cSgCu6{gqRLZ!RNDV
z@OlyPdf&{MoC1VUr~3o)_`W$Ri)&R)AK1IbL3FsLiYr5emnsF+EM0s<&-8)Dd)Xj%
zm}bj}kYf@2fHq=2hDSp_-wixU0s(B7qfiFWT<@s4vCnz!cb`MKg`?#5ncXo-HlysT
zkP~;zuTRz@oaNCD1$xgrxzJC{-YNkK6|GK=7%jqs7062YZ0P1OQ`bi2i_2rF(sEgK
z%GbS5tm^gqi&%MJc=(q}s_(uzlBGCX3pU2dDc7D*nSpp3df&}XcgA|41s>ZN{}y`j
z?>8o|&8C>-ZtTZ?&t>s@Q}jMI3KS)}PDfuqvC%AY1}6HCQr#L~#{`{R)Ft~hRQZPD
z%3dTh3G-5M;&+)RvyPEeqt+QB8R&XPuMLlMPLHax?2KL>g03APwz$&9tL03_Ydp=m
z%a!{yYA4sT8(<gmkflgoTBJ&>k9~1reS^nywR@btjBBOpBv|OJPbcT=Ec$a-u^Y}T
z(%d@^%?T4ba*Fp_ij0-*!ckFH&>*IX6M5}Yu2R?@Rn;EhTkI0JCzUbp`6>J(_QA|e
zYLuCa4SK!>BY#NVzDKtCkrnJW-=lqP_>~W9%p}Wf?AFuM7GKzsBPfa_6OdxX4}(Kl
z*u9(l`$$&frP0tjkU(BhE7o#}H{wzXl4n+$+pu3xf3<C}Hu&l2gyStlrdacPYUYew
zp>mH|j-gB?(u<4cOz@mhYO74XmH2p?L)Tpt*GdDeY4eZ8Az$<Dq;JDOF{en()#W8V
zSmEy0t)z#Wt}W1kb+n!BQcrj9_@+T^NzP8p8=rCh4m%?iAHkTURd|D`d5dg#V^J?5
zYSC$#FRDCP@3agBsK8^#@}z;-v~8WXW^KaKxps+&i_HbG6W5$Wc17mSOH?D})T6xe
z1Y57Gzm$H1%u5wW@Qb(cOw@9D&NuGYXK*QjpccEkNU2AO`dfR|5f@tNC{{4Kqt0<(
zh@fg{ma20&pAv-uo%9~MOjOCL^38v*BX?6VWL2s_FxcM$Ih9EVmAhn@XJftcb~=i~
zrYO}s1%*J@1^a57DUmiB15XimYBTVcqT)?uqP}Qco~B)A9`--iFNy;lwC_Tz@|X)*
z>Yn--1BV{Tb$8*cc&^$4I#~hV|DFciB~wb@htpbq^(ni+1An!M9#=48j;NeFdsHp_
z?GXRer@KJAU*zJO$uI9u9a8$Dij+RT-aqFnQWbuyBtW*yyYkLO!G~@xr(<E_wGEz?
z;St`e6Xy3lKHU%3>O)BK=qFuCILgQ}N9R|N;ihD}|H?>#-oBCksRUim+zfk^<dBxz
za}H0^MFOeu9>J}nc%L&GeLA<|pAzJLUiiQFxaaG?>9>DV{QK_LUpdC@_=-vDu?(Op
z!vyW*5^=x$tSl*lue}1l4$@fZq6qBVw7o5_Es1tUjhgYriyfY!{Kz*eZ&;?SDT%u%
z4Y09T>YS+;sPmoDNPP~1?pbu+LoY#;-Y=Afa1gAqnW97GmL=nvVT6}+eQj=T<>$T`
z9cN1)eT)yhk)~vas@^wr46n}Frs8>y1Hj>Q?jG>JFeGo`g;FeO?WOVVM_as6pQ%kt
zdk?;QM%Z46t9aq?jzfkoeI~UPU3B$9{6i}}!Nog>0E_2nv4E^(`M#D~{)GDT%(Vj+
z(E@1)gyUal=5Ml1*V1nr)si1NH;qNtJi?8H>=TsXT3^N&g{*a&4J;F`ukr*Qsp2VH
z%MO!u4rjCJok9f3vA^VW?XBFbPwcCS=ZF@GflaDLo|Qi~Ay;EEe3)Q;cRz1pTv3K6
zMl>XwQ8bX>Xo2y)iK5-$%3b}t4-Pl`4HY@;((Dda_#sg`Gh?+Y5RO5clnHicI!yfA
zw~^NF$;V1{5og@5E|@SP!4mGE6YGFRVC})0y57-(X{S3H>#wAtUH&vAo2tp^^XoV?
zf3m)NSw0KUIaiq+u!+hiV&bz(c%}nY^1_B`oKA&FtL}(v)wNZ8bn<^a2cEaOO`b1s
z7%mFYm$;FJsH@<HycBEld5zjVWGCAdwk--wH)Xu!-yBal>m`^mQ^S3nUX_U>Yt0YZ
z_2y)63z&5BrHlW?o}9j-Ol+@=|Jd9iU&}(G;2lSK%KUU1nUBw+GLjQ1f;8q&7f;6m
zWq#%{3^dPB$dox^ZB#tMi}F!R?|&ji|82A7ME`pj`>%wo&8~`0sVe^;4QBs)?Dh9C
zu>K<N-+xBLG9^43u^^O>=f~8&D&Khzz8Ha%QMCaJSd4l*!Z%YTzf#n7YNspWV+p0H
zyUQMW2BgORn9aEPyzmFCr*6tJOSBP8piW0Z$olNj^n*9!(%JhBuec(Yd8IVhWHi8*
ziwzH&-D#*m%PvSUq_N}c<0|c+V!)4~rxa4tBN6cc<o!|b?*)4Nlf`o{-TOn+z3<)i
ze|i02^o@x`seYSGWeOWpJIU2DYj*aLf*WIYUMH3`7@!Ku9ukIEAdKONxbf8br%BS+
z^|m53mI%Vv@l+jA$Acnktj4?@6B#LI-<9B5P?rN}O(iQKr_r{!<dOTlXVy_ASc!EN
zi{DXc`$w$Gz3SWVuDtyb)w^GR1}mSsP}~BG-=Otu_(b~5y)UYLkCA}L5tN@H^nU(^
z3=LLbvVSE#Wx%I6aA3@%*w91)_r@)qeZc$iTT8qq3qBx5BcdOu-w~&q{d?xXl4`?}
z4QO4{9yIH=@aw`Vc0kD__}5?fqvv}&+CTEBKgBhalZYkW^eMN6?zg!2g2kA(E8-!D
zY75)@)JbA7Qg^l4b!S}qaa(+%)PWm*LwXQ#K%Y|9e#xc6(AXmnVG6N)f#^5T?+Ukx
z6%bgZs$(9by~JM|c9o8mN$-IopRiKiS%qTR`K!|a$m56kx&OZh+x}f0?Qa`Cf0E}n
z;_@5lbHblt7PA>fIxY)XoewE+0N@5~Fd9a6g4+-qU5DT_7k~n^DW8tcvwU1&@**t;
z;XT80ix&J;8ZCOCRTi1ksBH>u316O(iG>8cFiv4R*eGAf8Dz2;!;^teeI{X}t)uBI
z3|i#9@s+|`&!fk0>3;|S|0_S$9;y1LVbS~lLS*qn>TGJ~<}eD)K<_%z@%&X_*j-w8
z?ql&p&u=&uUuQaeuvLOr8FQv&y1VnnBrF{IPTLgHr0I-KFT`{3?yo+oxRLup2IEMe
zH{-WjjL3t9Ur*qtIkm<KE|f+otnt1pPs_#45Jwju#9$QU2B;;r9;(yP>{~R2lu+Mn
zb+s#1KEN!^dnlqGpzQyzna^I7^rwrGzAI?CLfLNvREUF-r_UbZf<mvFGYq%V51bBA
zw(l`f$)74eDP#*14KsHC*#2SxqA$=*cyw)<7WPgCI?#h?Vbpd;b9HhkA=>JE1wq={
zLvB4KJ0GbQgaCW;QK8jPc-iDshWsX*DLo7>3Suy+bbSsVM8BR$p3ZOJx(j5IeF18p
z-5{s2g56FjkH+*1`<xtiw=Xfg!Ren8NtMSkqYB<O6jyyz{|N)}OWR?uocP`4!GDq=
z|Cv6Al1;*?m|;Roy>PgFBS+T&Rjh-w;T3~3jKD7F&TGNKEMcAohbmE6W`{_$eA$w_
zo41oopsJ3-0<{D<fA{DIjfFX^fl%(ra#fCD>)v&M`XF>C*L!z`c{74=CXCMsRYb%Z
zh^$}Qf{7>nB1elFe9wRV71KGwLG^0`#eT+SP0jMfuYg40Akc9B%dvmH3t7B(e|!J$
no(=5Tz@824*}$F+?AgGc4eZ&#o(=5Tz@82K+id{!_1%8}MgJwZ

literal 411116
zcmeFYc{r4B-#<Q*5Sr{Erb1+GAr)q{AW4>DveZ<_7Ao6d%!uq;2t^q!k}Q*bH`z({
zErc<Y>@${eG28FzexC31+|T`cj^A_t?&JH%_j??lIh+pWnrp7}eV(uP`~7+^C;K~_
z1lfPZz{mi?0lri?R3H%cIOHOPbJwmvKj3RO_`}7+#kG4k*IsV!Jv{t-`T6<w^6?4m
z6WTAZPjDX}-+r<Eg2E!AqN4mz@dIKa2ZTgKMgDvV2Pb&XZZ2LfE?yA<J^_*c-!JwT
zh#(JlKQxPz;|OGzAP1))2fG;pgFraA{>+wx1M=@Lj$Pm!_i*#@?&SkNP_`eki-VJM
z*KW=~vj#sM1YU>i7UU8-eB%5bVe`A(M;?ise3JNvNA^NdgQ!IxMefwS$059X#l#Ot
zNXj2oP&}rjp?O;C%-M4nFX<Z?UN*XN{f4C#@}{+oy@TU@CubK|FK-`TKeT^9=+m(9
zh{&kuq~zx*scA1>roVmnKKnz?$J|fFC8cHM6_r)hjZMuht!?ccU;76JhlWSKkB&`~
zXJ&u?n)^M!KwVqk*raU%+dF^c;(&1en^^xL*+0l72+Fl<_ioPJ+<)Za*yRTfPQl$=
zhfnMgI&aQ>_mS|ClTUa=E+oDwYT%VUWkC_W_qcDbn4HG6JoS%gf0OK=6D;IklI%YO
z`%k&XApD#h;O21(LJ$z<^|#TAkiW)XYw#Ba{=&dt82AeVe_`M+4E%+GzcBC@2L8gp
zUl{lc1Ak%QFAV&Jfxj^D7Y6>qz`u?Gx8rffC-!DJB=?$eek{b90Z*o1D<j-THpcCI
z)(aDHFLPWnbtu2e5|)^Lrgc50M}ES#B_We<pB-E~D^dzE*BAL~{3~l<y=BA7a6EPK
z5zdo8Z%FR>F2vp#mQaBdD<l=awT){H)^yG%09Q&qclO%grhc*^l89-gKH~3VE^G)c
zjp58`75$YBd75XyhGb>uFj*+-yMMU-qsnSQ8BguF^-N{LnF*;quxwL3E)ny;8<xrw
zu1zM-41EH*Y7(Cnw~Hh$c0q2vEhhf`liaSI6>}cy3w_5Qe*KwY0pwhCyvs~^*w2*V
zGrf+l;~_Hdhy57xHd;(BRmAEV8xjX`?Xy%Vb5k{X>8>_i<tTAiIK1tr)$jj4mz`Bx
zUokMf?vWEesWXF#6>%Ck`jHK3w0y^g^jZwC*bsG~cTa(63LC;}6~~60H#Vg+h%<)&
z_`n<Pn_ciFGT*p=0$qp`>CEB$HX5S%FOQN!or~FH-+H?~+UhI4I;r!CXD7CIPn_sW
zSXAjthZ`nWhF-_-ih7<DbZ+=HM8sVGpBy)<x8iCM)4qK9ZdT{)1uOY#!eVDelJ%r+
zAYa>I`|y0KM+Hr&!c8ZxT-ycD<3Xe@jQ6vY(Wh+NXt?Ooz#kiH>h>6xLGZax$DVyl
z6U<LM*CEB9cdY#g{_i&R?>G3L{9Dv&>qg0dFSBq#%6&p-gb&~S2?uTg?(cU%wC)?y
zocC^`V_fyggP|6^9Qq<=|HJ5zyKL{27;<FIIv$e6-Di!XJpavxY>1SyA;eCW2A-1Q
zJoO*{apdVc5~4MH4+&iB#qTT;XON`o=y|`3N50m7^3oIP`&V@A|FC<I$Ef#yxgo+b
zPgYi>kSb#gYLm{|yXVIqdVD~pS;j4|@H%-=-HSqKeb}mHbTlR)TG!~wmn`U<AkH}<
z+BFLrIjPjf4EI0))jNE&{lF-dubEgn_cY%Jbr6U#FG}?hvc2W%Wh@of*<f`~0E6>!
zSAE_5>}vQ~`+J22gMx#|{wd0}Tc?AJX@!*ZFB#uVe-Vqr?b0DkI2#i32`I6n;Ff0O
zSDxfg{Z7NX#$o$%erTMu5193^5Ws5Un)ER@*pRmgWCG)$9ggZl7iUAbJo;DEh>o`O
z$^mb&k%tnt1K!cxi>Po7j-3fMWVaU(MJ7typ#dPdxw^deu(M9~AVIE@H|4?d<Tb=z
zrpk!)yFMRo1Eo&S!~Ew$OV?3pILf74a2HjU+-)vqvzM(w7pu?e3sZR$dYcV-CIec9
z+x@uTM}rZs*pMiEDR(nnJraH>1Qv-^j8|Nhv%EBvWA;uLBJjknhoOONLF_29l-Q7l
zmNqs-{3aVB%7)CoM<=j|4lM7fFg66L4yxDO6-xO_9kG$>*iL?nA4P(>3KDYA0LT~{
z5*`iYup#=e78G!%k_{n)=>5mH7MHGrf+B%RsEjftcWd2=4Vm%AQ+_7I;5RhUNEWMy
z4Qc(&IYR`T3DLL>bA}}w5*m}hQcJ-(<D(yZMNHcRd7#C6@wX|W?y@0Gsc<!1@6a@g
zPDijIQrRfrNKZ@49o)1Mg@D>#{PSKGbTDBQb-i{N&H=#J&1%wj)@+aW+u5D<^Gfbe
zC<thIh~w1Z4x+S8BO_{0<(dj~t-SU)l4^_De&99el)n1EpDz`jBI8=2JQyj;a%%;7
z*<<{2QLu--tYc3fw*p1<`_-1akIx)7Vk&RvG2hm!0ySh}^Rn+^i_Oo;BKd+Pik=!w
zS;Tdg%Y0*3Fv1cp0dAQIY-VKHei|Av$=5$T(5UtJhy29mcLCMjn=kvG-4T(vy}84U
z>KS1OtFuh6Pp>k)tJsi!f8`yntUbgZp4m7ZRH<KELX-`nttaxd?`;HCv*7)?iGF?l
zbEON3AD;z(dh;0~9frI3P3xG<5YR)-WAS=XBic~nwT@&dt^sUA?m~sQ)z9}2uV|-A
zYzM6C$r#av8J-}bVPjrjd?oSm3`3m#>1-imcdy(oxyNeLP0A;BBUiJr7g?P|iivs*
zYA;qBZP;SdLwh)&G5&oH-|4GbdR*b)UM4zgSG3}Pj+YvJ<IVk@M_MFfOR69oD<*K_
zIUK17C01)oS#5yx^)s{zCS1e<1}jdWvP<086D^PDj3?{KyrElw3uVlbX6&-9gBhe?
zY{`{t;oV(S5rq<s7fSw10s(VNo>V_uV41FsMz*q!>{lavQt3H4ezaNYzK<DqpFZx)
z)h=1Fd)5hZ$w_WIdq6t?0=Od;qs@jCYQc8Q<sm^AQO|G`WI{Cd6a}%&IlXAQ52&!|
zU6$+Z8=JaDOy%j6K&e<B*o8hwM=t)v9tUUSlK&xw7D6fLJiO-Edqa69Hp)&=4<>~l
zQy^#T`WxWc5N>f8;}&)S+c0@(2%}F|LI;`+l_}R<H$xw&o0$DzbaUxos?Nyz0ID}_
zi7WfZO>75TB;G-3ea!zO)W_j~1_!W6*T<lNl=qB>)Rb&|0>i}JU4X9gu<28GWGH9w
zo6lTc_<cBMSkwr9KSOml8?v8au5!|~{NwMj+Mm`}zw*fX$#Q3#GxNOXH}GT1ZRKf~
z?K4%4LUWhbatn>8b!*d5BDFC7E4K}YRlabtYUClfYcv&`?p@c6;*1pLuMb)k9dLge
zyJmUeRp8##;Ju1(gZH|fY|VZVW_Hr(%&ym{Og4ma12*o=kf;2>3ou}~pFqj=80P6;
zV%tk3-P3rh1`&=C>~}onI+O?8Kq3@v{+5b}4YzlUDiL^RA8sG3)ho#;0jlU|j1$oL
z&|#?7_R$A@{HEk7&A0ZS58I}1vs^OChDnAla`sg(<Bb&R0t_1vqf>3TQn5<!sXflf
zb1&k8E2}TTEh}~TDqJBy3H3+NIoEr}7>1uFoCvS|WpSsj4b;n_OAOn-V~q2xXL1q)
zkKklc7y!ZIsi5v!xRl&{TdB|e(A{L|jLZGd!t0`*ty2_z;sNW?E<|`OGUv9A>&^4%
zo2N+L2X2>gtcf=f$CYQGO}aMP%u{WW1v5F9HH~xC?y({D)_EJ_G*u8_*)Bv0%^*Yr
zQSeZNK2^V4Hq;{aX?gWCHe}BfN%xvj3-$`b8c?5sg}~*LFH%z@eBPms$#XT{b-ev8
zr&6=k_J@(1hS!B<k)qFzK4%WalY<eT^zVF?xO^x!M7w{a2~RP`k2T!@HYvmOChW0l
z+l2jqcN@Ro$<vNCPeU%kw)A$@&-|zI{k<ao6MsWxnR%PW6H+8-thxrhV@dr;v2Q$i
z;_KxIms!DnBHohu7JKaD7a;C`xk=`JR&OTHAXXN0h^i3b{Te?eK3pJj`@ZjN$XP<$
zBhY!o1LN^XV{TS8kqr^nlc9odJpt0lY{)_MK+6eqOp9b|jmX2FvV#PN*EdLyIg_KA
zmV};F;3yli<TNuiif@r*ZpG-o2wBrj40`?I70IVvbcgewVuY|T6vZqd4z^!Moo<Fs
zYPw7%Mrw2(=sRs|y>MM$a&1EN=DwkI-2cm!-mU?CzZ@HqTtqdbC$b@@2`}!(b}1R0
zQE4y;`pwykn|@n=j1B2Cq#{}S)Ih>Q{J<K-i))ncQID*b#krInv%?mkES*6WF(5|o
z3@6<E*g(efsc5BAe;g?-ndLV70-M<Um9);{>7>RF3Z>757a|s4fzic{EGTC`8{z@R
z8R^$_h?0|?3EN9cY{<8fItDL@t{RLX1)eQX5%0dtuzceIiF_w$chpn?ceX7XjciN!
zZc1)jqTfB5LvS|rf7`|?xn5*JG2E2fyS)&M_|Ovylf<Y~r@qkOij*>jUn-J^uNB(E
zXD@%#_ejlGh7|Y!2Ufp0f(x#D(>bkzU*{~@Kn%$z%Uao^aKtWvQOR)IivANW^{yCj
zJdtk5;{LxpKIQ<7z@0u0JRt}1(Y{jaKD;KYTC@%c4LlIrN-~IH#Ycp3b0*==ZTqpr
z+Tw7(-G{*~g`-$X^`YQey~A}Z0CUVvQ|8h5dSpq3R@yVYef-5rP6G-e9N;cCey0+c
z0@YJ2?oKwO#heY{vu`DCoL<BMFm;fup@PsT7SA{Vo!nII`>FnTQJlxLbqJ<SLZWS=
zz!Z3#7K!-nRm8{usbChg2L{;6XtE)1Ds0)1FWWfa6Jq)SdW01|;mL+@g7GssZ$q1P
zFq;hBIk<!ab%zY1FX9+Flog>3>mUN@kR4MNuWX^~?L*8L*nv;3;x=(@wCu0j71uT5
zf`9NA9j*UCz_TIuh@*w;G+V%s4dDr+WsBu%js2`zpYDcxy?V?Q#PMI%4y%J}u-9=R
zLj_p9w-1CGjc7`r-qS0Zu56qWW$fSo_(Eb2=NjRA00RZC)&Q=f8Dxwtv!LSin{hWQ
zklW%+VU$#GIgY2XA<=~2TeQH+^mUSH)BK9O|6KC<V!zajslLbG$YuT_zKmwQU_+b`
z+lLVQh)bv<@Jo304)!eP26=tJc7Us?wzMK&S{-q|5<M3dU48sEkL(-5ToG0o7`nkh
zp77*M*yb??cLBK}MSq0rc#E-SLyE9-V8~6FI}(hrK#*cs$46>SsN7^F*PS6&?m-(n
z0m<8TI&VxfI;5f+2*Z0Bk)RZZLH=+FNM=Lsd_-*cU~B%{K#{w7ih@-2GN*th>{FV}
zE#~lRFJ_f<z0dnD;TdS_jnHTYxQk&$GpHRYA0h+#H}OYtfQl(2-2(UBcVJr=E7Qx3
z{71wXGE@^J!OX#U0UJWX5E+(1(sVIw!#Z|&^5F31&CQ?v;wDw}vjz)OiVrs6Wg8eg
zMPF)ys1PUHL%#^hn+5HT!Nu0=mi>ZBy7j?mn#UqoUAB^*%g3llqsnMETqIV0&uM<q
zO{15lgJLZ~KM@Q;<6Cnb)NFh?Wr|^f>7$$B8g${ZtZxBV76YQnMrz?brBg07-!G2H
z+&N^Q4GHvIssT01m5K}9%utmgJ15+xEFMo-{+wmB+_13pBaO2FOGG1Db_^3Rr*hDt
z7wQts`h9q;0s|v=&K+-uPhDZzDQ~9zv5Ny{!|po9lpeEmqvZ|2Alpf<;B{8<o4B9|
z8wm5_znT#w%&pTXp!=$mrPTs&TpY=_Z2E^)ty9n3txFW#Gem+s<ihFsY~P|^1DT!@
zR-WMsY|WTT&CAGAmJ2fqP@04LCyb7#h<rS7XMZg9doOnbiW9B?w64z}gWup=Y`zs4
z6^<&uo4RcI<Tb}`;r1iM_pXdkzy<VrK|mg<C$e6AWTn=0I!e{M?%EjbM!Zo>v7u58
z*Pas{oWn%aKNq?b+LE=JR|PQgl)9-%jOC|>_w&9VZ~c6}rni(gh`(!VL#wi%el4?Z
zIk;k9^0_p&Q)BIwlH!U9mJ2&MRles9-3)UrQ(MBJg5vRfdEaZ-YdYU#93J()nv)si
zSSFS!FhmJ0^{0kbN^xP+-q3x=oX2S1&_|DcFQ65ncE~2UJV2OHKXBW~rvUxVYVg;y
z*|M(L1OQ^XtLo;Ai^C0lm8FNG21W{HL5Y82yaqL_GUh?w-4S)GsYaWtSvrQ6`-;0G
z-W`HaH?yGeaxZ^9tBvoT%RA9ga7(LIawQ?*^P<);dVbdKoc)^S4q<u<ID(6a&zu^^
zY7M4t9psZwxQ`<lr>`Z%;VFiw188HTjO&!sCE}6qDy;m9D%&d`A}V*0(jnV9q_*Wg
zDvtCaTV%&G+GEq$@@{pHRmzIJ@()N`G&7DFi*si~1hB)mr4ofS`)ZL6S+`vtU*`27
zo;%1N8iJlj1xwOGKcek#*7TRTo%LoK^j{RA4HS;VMVRS(Vw8Y41co$qIE_pcm@>Yb
z`~B+y9h12E#42dXJ^={M)EoI4+d~vnWu0cgC~dKaookn`bf5LO7fBUqIq$&WCFuX_
zJnb6_P|1n*Aqb2D8uRDsXfNw85|{i0#8zO1{mN_0^`PB*!rf9m31wcA|Kl8JO{~3w
zba8-4D-Lbym<~c34d6=fC)p4ejhLG7`7+Z2%uO6A1|c4Z&T9l?&B78kgjx+@N->{o
zDtFn^P?Wf?&B2ihbskmCrZnGeWUl6hraxk-ijr{ykfk8?DB&eO?*1Qd8VDKXZ}~rX
z$SonUe7oqs`iuXPo&ERZoqyZ1gpVe^0Vdw%d8*;$%_3`A_xzmr^8`b}DP}feHVo)w
z)xx%;AuR8edS&#mcJ3g$!uCrd@z-JXTsFk@?LQjQLf0)8LIIQ@j<GmQbVQA=EYi|}
z3P)0miGiw`ZKPwyvapjgPj-U+yxu`GplrtRtc==l0~_-BDQHohZ5aMQvSUICKyQ9F
z!}4D!Nnobe?_{uea>ynTU#Xe+nsU;2&<4A4T7d3A8Dn(Gw}|Q&F?-0->bhc2xvv>_
zd%EGW6BVbQA{RjxBObeOK(&RDv`_c((3biVnCRv57;3G{0ldfy4X0n9QW0FV>f64*
zui^c{RKimhoYe(C$IM_{sz$|n=c-y*=+~_sTT?qRVlPwn)0=~NZQiEfL)C|I)OR~W
zKc|Nwmp?>RDCbJ-maIU5DI*H#wZ|lJ5*Y14th8%?T6J&6x1z*uo6E0Cjw>)Tl{fDF
zTOju*nEStR6y#bT&CZ#yF=7>xk4DwU@*ukN`}AO4>dV^GIN$;s;;bGidE`TL@vz#F
z^RWVNx_Jnn5Ttmlp!;=NQvL93tZrF&vW~s%yP~VGc1fljkV5BR90Cev_*RA8b~M%&
zDq?TD9C@yIUqNoh-G{h~4RJ(7OqG;wt)sO}lDciDHVc%)6=EL+D{^Yl#wZd%BK;&V
zqA$1|R)Aq<;d)NO5Wq6-tR>UbH60__n&Ef%am{k!TeLjN^3IKf_K~l6%HmWoJ&M&(
zEg?s`RtI`<4URb+#n6)rBox@5>dcCt%dD|FXbmi|Q!Qo73jD+&T|-vgdk}CYUQ@g*
zQ8bQ3c;XF`GvM}Ufw!CV<W-OHj9>TK6gaC2BmN<Y_!1V)hOAlWfT%zf<Ib%T?Qz>@
zp<u3HWCORezmS_uWIRt01pi&acUQwkYg*Htnx^7p4vcI@VP2p6IxnAoRXA+7upX@c
zr|23VuHkz9uJA+X#K_Gxkk3VdbBf5pfH{>nal}=S;8Ha>HkqPp+zef9k`|L6G_2S^
zA0ylV_~^GW+RS&G?uOSWw?BHW<R2F94wLZBy;C*2?^m#HVf{m9F&py04LS-nqI7N2
zqIVh>kwRZSCAqKB6z>JVY<A$p6F5>OZ0#rpB+j41=2$UTAg_s}vP|hVnZ2*@J1aOK
zeW!%nvz$jFgd+B6BuJyfP1S;KPdFC!-_!n(@`5wg=SzD>n<!%y=|jCpo{EB>Tp07-
z`L=qds4Tv_yI*Dd;oFa>#6aOv;RR>T)x1?eH782y=E*UJPLVCw0~K}b_=|w#l2DWl
zwiQP)SLLP^j=ZOQOBZN;LPjr7E{(U^IO~)h68?4P&f1w4tk?wF(5|S=I3rGG=fFWn
zhd?L3Z88z$e55!W7x6P(W|5^k=<Yk|njD#bw2e`-+=QFy)PudZT<2Z}873DGaEq>O
za`dGSm^$@9Ki$O@j;Hs=BEk*qmf)?mpec8uyLC~d8gof_-dnq=@ddEeoQvx>8};TE
zzP!3e6fTp#vUrnNSyx|`y0SL?Eve^Js{esCBZb^7X#eVdT!XvLMX!g8*%;m8vM07b
z{ncN-jHn!KaRy9Td@eXz`{5CmjBP>QK_f4BOv_-v@@N<4EG3k79w?&@n4XIWWx)J%
zT|Q0AP@~%pfVq#zo@q12J!UdoiNReKzuJlwtm$8pf1Eh+(;=>C<^2pb3pxQE9j6Te
zj&viNeC)wtt0|<NO1eAk@QaGKd(Pq->lTQqJ$yi(6~(H~sleH9eLg+=Rmf$vf**PB
zpcj<4a#*}ubUTIR^OTPQRyW&XHST;~Fwgq(@UdWPz^poagJaWD1o`DpM#^;j<T@$h
zDT$)+v~X;^#pZ|nuiB`3H(RXbXG5jgnLWX}P+BTzxDy#CnRt(!zG=|cJ<VuN1mmQ8
ze$DTFkcp2)=*Tvk9EMnDi<bD^{Ce0;oAGdZ9Om2QEp6p>n}r3p#oR&<NB@<hUUEt8
zEib>nN-}+H;)~~FaRYNR25;mOLKs^91UrM1ogP^n-SGAO2#~NE0U&a@UQkA&%Rc7o
z1WkuiN>83_O-r>X`#MWa<ud3)S+ROyWjPTnKE{CvvTlez+5O1lkfWt9MBSVD9h@=)
zkYw&=_3!u@WaEU)E!*Zy&OsIL-y>c>l;yHGI_UW(UN*QkqVRWy^blOq&aP}E!o$J(
zZpb~;(N|~gD`&n0D=qy5Ajl~K_xlPSw)q28Sj{c1rD<y6I4umg_!hm=-%$!^s-JSN
zIZduPNI*3b)2}dD;8KA}`;IiUIBOEO5!lP!ixRGVB;JJMScJ=`b9#xr=6Ly*Tk3NF
zat>q+4X(B4@<Cg`b7R}WaA$pywH#DLx59PQ0pqM%sqJW%yBcnWjN0m1&w@^=k8;xh
zFiduX3b~Hy$3HomL=(&po4hdognN9fzL?O41CC5FQlR3vk$haoBJD`9(n4C3$v~mU
zs2`8L3(w}4F03~?nZ<W(dUYQVW7&^yD7ieWWnn9^d+yiw7O$<Kyp2;WxHXjp{N}O(
zE5vR4IgrhU$e(J#(?qMm0J1<2RthWeWQfcdFyF!zZ=z2RGzrZQ+}e1Zt-^g~KAIjv
z<OEeV4Y%gQlHOrM%u)3s5M(#1G-7#*(nNEII_53ByFL<m*n0vYRIK$r!?Xh!;Q+T;
znpuLI&cMP7!=^iiYTXNt#S4!tJ=#i)sBAy5iVBzfRrL{_W>r0;Y1>gIU@-qJdBX0j
zznuM#<l$h1ajgmX4W%PTtCDoZr<dQ0ElkM`picP8ynjIcKp`^32CU*24wUsoq!J!n
z7P32!86OM?0~J)Af}>di*;KNGhmPp9a@XyCF+TZr<%4c`Us*($Gv+ok9;=2C7{HY>
zZjtCU0xI<_bq3b53anj5us8^I4BL&X@M~Nh)3q2d7;1f8$gey<7Jm4?J>$d*0x^x!
zA6e^oyNF~-4Hfw4s;#&QTo^vDB7Xb?GA`l*&u_P_i!9zcx&vZap0Z*BZT90~Y4#>s
z{j3gpE8K|CwcY#JnRBUno0C=!;wj7pQA~`>$7lym&udz%y*cF{b@*16{rF@(k<dZ_
zRMHq}?tH9T{5BX)vJ$|>NT;k2@^9W$550fUN9W<5^#rLGYS`7fpb}~xErz1L#GJlf
z2z1_zZ#|MRQyLplm6kwCI15i<jQrk7{=HL{!nmW6TJnAYdciW3pLL|p9t@t;qhRHR
zEvdvzsRBQn8fEE1`~%+jiHDF&yQh%}!@=-%;0gHP&vbasBll@_j<&!%`(*3uYMfuu
zm&P95tQQ5v7<Dt7)R(EE?DV@?-H}V@8{QW>sMhWFv(Avk+dw}*pRP`OPK2&RE3Kd{
z(6N!amnTqX7bSI9BQ}i-w9Lzsx_5>ziu7JN-u{IYst4;+UmBr1GR`q`G}d%9bGOem
zE5F>RDf;;>r^?L#?uzon;&v=>ifVWR{H1=LicQw({ozt^?9_WTWOwXwRjGhH`)6)D
zR{^IPH&i5xqZecB{I$>3%G~Z*o#;yrM+WbaV7NZWr%4}n)R#7?+GZ5Txp4Wx?DFd`
zhb^))^=3qY6mpQY&#I7Xb$U)}MViHr-K^h!stLpO1E8?U7zJuX6iBxPe7B4#o12^O
z3ay;)spRmxO^?0_^v^sd@;<V~+SKeknDt+T72{h2QKMTk#PC{we}}QS^5<)56;@-m
zn5(Z~fG9Qt5NGjmXQ4r-dHRWX=&vW6QqL=dUM76KIfL*e!Zf70&8YC)JsH_HMnOR*
zj+{AcS8~UJ6r?;>zndXQoqSDAYp4*H5g)u4u6yOEzogKvV1KY=`ZmTk(Z@;P%K;$+
zxwgUkL|v>buHp6uETSGx6%W5p!HfvGi1&|5JZ<Jy95<R@MrEbjJnwTC4L@L;CCl2S
zyj{Zlh?phU%d#QgeOK8KgE9z~zz_z7FLXvpz!jX~`$8K_tW7@skgvWrAS6=M`Y4*d
zKz!}_U7K~S1T4`#od#K27*1b<18Fh1wBg3PN?3VQT=K-P^U<S2muFB`_FqJ)>TNxM
zGItXq>|E>^GJGbv^<2I{xjFAK55&ZC2TimVLv)<>X=vFe5-#{#3+j?n%IB5a!}sid
z$K6@{-C(zMZv8&bTEl$n`2hElNmuctd@&j|wluzN@4E#P2sXtHPqD@ApEzcsHQkc(
zqDRQ+<EbQg#tDmg1jV-CvB(V~;NtjbQ7$0)7^55+jYhC|b*Y5r^an_ru$1~xxlp)%
zAOBZfTG=BUm}errCLVzE{?@WtrbUi@h;j}$j_Ewf6J_|gs#}qB1H7)z5S5)aY~_nl
zRlZS~*2p?kbthzU9-5fksx$Be<#18M@A&)A?Zque!j1Na<U=p;EGTc=SvkMwfB!DG
zT{eV=@Er%pwJp1SVi@*qLc{Bo>fLIVWkfSft1IvI$T=*pm9W}(_i~TnN|CLdo}+BY
zbojd&shXq2@Sc37?+PISEK#739*F&M&S}~I17I+2+WWq5+#tzc&P7g@h&F?eyvI?w
z2#UEor#V$&iAYwKR_iw&Z=-*1KH2uul_5kmY+^$M(MPpiG$&*hjcs3T8<0=dP)9^K
zdPaVO$ve=caT)fqAPing&l9y;RaItdQqsF0D(|@`WY5gNmXUK@zLKy;m;^e_IKpB;
zM>{Cc{xqg)Ix^+K!Go{@2$soeCzkmnEZa`=P55>i1<9*^IB{U2(}`Q+xALqfpz8Fb
zk0D5!5_`JmQXQ-3e%s+IH3?0cU~@sk5gk|ofSBG_wx{gF50?_%nUie~YmB7w7N=sT
zc4L&Oi;a2HtK!yWE9u2iL!u$iNP^EE203$X<NLk0hJ1{*6@kE)NuoO?5X5VoF|lMa
z6!gl(FH^f*w0$#lC5MG{oA1DeYp)xD_`c5&=qoc;y)b^UT3pOajw<&0@{b%@gldq9
zi<@*@z7?gg{jFK#yU)^!oD_T3-b9iCC1Lq4Syk_i{>bHe83tF+yq(sPIEet55?xLA
zxP&84J?QePfUSS??3qVZzz(t@_C4WKrG}B-Ot;o^b`^fK)4#@!9e$?JZtJIWaNQcu
zgSd+iLvS)4xs!YNT!d}itf)pJUdD=dJ}s6JpZaD&v;J4ZQv-E1MwP)!D)4GXKFIUT
z9yRtIkLUOvRs@iV=?Ss;{e@U1h8a~oy3jG{`X`j}BUrl8F{!gkwHpwb6#5^dZa}@k
z5@bGOocHonZrsxHr8&H5IC9}jG=o12ss{*Aa2CKa^|~JAenz#{V4=k^4^clJSq1*t
zfW53J7B@qX()I-bnXeb8bmIlyi}fj{JEAiEPpoTeT~AJqA0>SC1Z1IMTw84jLARvl
zVnet0wLI2dWuv{bfn?(SkAK?4IOp)=PBi;|go}D;bwO;ZIbQH)<A|jD0%28f&_hYr
zk0Um4guEJpxC54fu(}zYpDB(*c^>1h^`stce<u6}`@R@0Oi-3Ps91bSLJQqf9EiSk
z?rTHA`-;k<@>=VPyqV;9GV$8Sx54|LWhH+hBoUoaz%gWF!V^F7^s8FSWQ1VTGgTFF
zjCO=sf8mAn83ZX8RhIgOqCVY~^R!kSTYg)Sy6g4&G3M-P;>nW3+&|sFc%dTG#Dym9
zjCJ}oO0Kj_8lJs>@39g5eJ}TXH<*s+sviD;G|`^Ueg&IIy=C-*Wvi@x^^coG7mt8x
z4VcRop2p~4B-LL5OzHa>4nW`AmVCgiVdVPnwV=+?Te1c_dk*2h<}-|#q4ioA<=&OD
z0>1_=^WU`2YTI@PBT;U<X9w{8!3aY*6s$VHVL;>B<Unh+X4})8*UfiH6JA0^XLgtG
z;jRm!P7{Ua;m7-yH*blXRSa${h6zZw<cD0kySay$TD<p_U??gZK{3RQ@mUV&_<zEj
z9`X@OyHH!PzCrirJ*pk|^>{6FB8#dABW$ikEC<m8TyVv>NbfEs`uSw5z`;ogIptD&
zc*<rF77i+^3j*x9v1NgwdNw2=5ZB<QqaCj!dKU3-6njUQLTod~T%jCIB;UMqy3cjL
zN~hr2@x)iibxu^j(-NKyIZ7uHcPI(ldm)=&x7iR{Ju^Qa=K!kf{MgTxq@Cj1cB++e
zFppE20fER~(d{m9t-+*MlGZeV&a1E%0BJ!`Y<l+DCvaWEErZbv*=?CHnl|&3unePo
z9d8x$709`$5w=@pJn{?oz$7taFLM^$7E6%S=g#9cDN{Rpd%3@-{QRS(uFwcvNqw^X
zT&&M;_xqMV%J&~nnlI_yzO&M>+42BvN66)Fz&&NjQZeoF?&tcA4o6iRDOy<Kgp{{C
zfCeh+HZVlxw4!=O7S2`^>r-BF-)ofXP?p?L<Ocg*jQ^BexXMb4>(dcXhOT7F=X<i5
z3!86#=w4aUY&j2JE2cj1>r%l(H?~zf$k2-8Z8OF7x&R0LFyjjIeJXjeVdr4lIg`?Q
z{nocjc_N;%%d;WZbd@JJp@5uetG0<|;h{jOs>USM<Zy)nIdU*O7VrnZHNSV${hp%E
zT~b7^#`@MuzK6|x1cTd|P|M=S_R4SrOwesF<G2>hU9A?krXQ&Eu_3iD+X-{(<g`YV
z7)G3WbyAud9|w>5)XI6_egl8bo81o$MTf9m2+*LhA@{)?hmBv9hVkgTQtKNi>pbZ+
zK?e^D?N?^p8+iRYXzo@qESz#bjw~f^t-2|7d1cSkv`SVAgSW*WzZ^teWl4;Xur=fw
zbZU^rnONTwMq4)52N37=z`8oB+^`8T+Sq!X?Ed(y=5d}|C)=ZPCgM5YzWNB?q^w2~
zNfoM|k5mHGo6v>i4qnYJx?$q?WgfS6;nvKPL-mKy@64%tssc~H`GmUSnC}0oL@0ni
z^UEUuS<K*MMqrh&eUj%Eu~NOGI`DoC<J|7AvQD?N_-DbDEBO3xnmai^IX6wid-3WR
z(tg=N3%_ofh7l<?h`nw*#SJqFyj{~Vr)QSMEXt<lOCO=s+@>f-(Ab@8savOAoId?f
z@(}y{I=^i|z6f;)So{*O)r1nhjSg(Cb{M->bbtJQ*(*%v_CrHZr?V5nh~hfOs2NQs
z+$M#FNJHM+=U#mGy!gBGRqoD+J*s<a2}!^1#)g!jTHH_mnp|_DOl&OfdzZ5!FK0FL
zP&!i1YZ=vBpZ2jK&t-U|ykDh!OE5oju{Y$D-odFBT%;gIjk2eR?7!SKXydJV>e4>G
zJ5Hk88=xpX`)2T5*Pu*41j8igH666L9aX>xs}|J1UK5<wa}GM44r5J>IW0RDxK;<n
zhaF8CK2}=-Ixg6Gl;y~*7DLdUWr!C8vuLp2%A3D}Phs@Qs;3$QR3{7aZ#U(*lz((c
z|25rsL~+H45Xw3Y8jL&jGH4;{QmuIQ4|MB_*fC|J473-kle8kb4iYMRS~VClQYF`?
zJ)Y!h62nW8CtOc_X_kt4^33vu%=O)m7E$;BEZ!>*b>7K|@5ilc<TK=0$CpuGAZFb2
z@IQvQl&U`i{mjt$Tx7n8?Kb_D;`ztP|5)rTb;D7{TD^O^sS9fP+@D8LC((W^-V!Rk
zIpAu*Q!G#Bn1q>;P>+Udch0j%H3_Y?Z8OlNHtH#oQiDP~M%m|0gSLCL|K94eG$qrV
z00=XUnMovlnA%%^#x=ECdSU>uN4>&4n-X{?%<NOX<1|)ZAFrRP>3!#S)8Xf{f*|*^
z4R@+Yk&hZXD;TPr+hB2aw@fS14?zO+F*0R{*r}2+8+6Fla{6%C!v`0Ojv5`ZycOS8
z)p1sXgJy~0rRK2q*1cg!1BrRSm6^um+r~NV_AaUSI1eTc$P8^w;LhPIP>fewv24g9
zXyH!j!E%Tts|-=~O8}K>dy!&N5tjWKe<SUy^_ay4O5nY^Hd-RHtX>NHZMYs4S!V8&
z-wsXtaFc9((RY&LE3^Q+hamx!eF7tnpAj(|8d00>I;-WU)Mf=@9THm0Yqd$xXjmC4
zK~t(r)_4$uB1~Q$ursOAz4q`{<N=wN^Lj9E;2oGwE9ml4u~BFmzpOgc9`8?hQB#PP
z_kCoGg5pSJjWgH7$6F@3;=eZIGihJvvwhf*V0Q^Ztd1WY!te(C-oXp<(2`3o)DPSD
zLoUw>Qri!mcAHDJ$hUm5`{4dS(8RP+OlRgoz|1#n?}s|lgpKZn?+KF^zFWV6$G`l_
z()^vmpkcotO66C-z{dD4R%!`L3tOdl(>paE%RXzrAp3dcM!7!j!in}0gTh#e0}jVm
z6gdX}z2O@NdN+!N>9#gZ1Xl%Hcim2HV<Et_?jq=YPM}P14MvFBC5$<17T1;c%N7PO
zE5M>l45I~tGlT8J&1M4jtOj4=&M~*Qg>KUm&BkCgyon=Q4_CY#0GgLfW825Bz7Lu7
zN@3=_2+jsVAgk}2pH%5n*NC&}*Iquib~@Vms9oN%1DcFcR=*5K#5BION9?!QxPEpW
z;{(_2D2d0GIF1&a-joc&#d9~dhs9k--ABwHTa7<`df5D{r14A0wz*#S$-qqI?NK0m
zrjrmgwZCjtcp{|O6`Axy2}UxwXNiO8W$_bv-GtD%=AGN`phC8Ga-kGWCIW4RD?_6U
zGgtxEjciCdC+Zuak-$B{&;t^tr=T~R?^h41%SOqXcX}lu4!P}kQkZ!+f1h(N5hxHS
zeU$k7!V^81I(B5LqCAbR03=-P_biprbgE1Ewl7I^>;(P_MH>dFibr^=->Hz{e~lU8
zt?AKmn4y==PEl;sw~u9tvmpde=5yqn4Bq>J=bcK<M^Q~rHnO0<Ptjg6t7@g_8i2Y9
zagay`dHA(aHFu4h8FPo9Yjqp`IJhNBSpdz9JLX#TyQ^5~`Q`fow#>RGFp6pl1=>uE
z%9IC2XE+NMi#tBLdh%QSF3NS!mE3LiF7u1kzBkh*Zdc-D){W}Q49ps`D|`;XXU(`*
zE0(<IUT7Sxw4^#@<Z^XiS*OO7*8|H;k%JHAzwBP`uYXMdy-_j>EMrXDvbrAu(P*$;
zm0_t!<X>F>q^QXKdE4z1XoNyoqx)H4esG`uhd2!j^a#=-^>{_sPj!yjJLm4mC~^+4
zv{}Pb{9vPpHtJNX<l04V1O2cU!gh4*(!nPvKJ);nP@<W+kGR9(u%p@c_8qCIt#h`n
zqdh=$6OZe-%*6ST6(S4ErUV!KB+GLjkH4?1dRF73z^Pj`Z-LeV9epJ0u<M1UrfS{c
z0&o9OPl+=-nIXZ5iyC0QKP^Ueh3^&yF|hJhVI`<4jdMY$J_>+=2Gt92qYCW|m{4!M
zV(lOFxO}Po<d*|9m*Wtb&24k?>L`r7^~Ag3qZ@Y5V17Q$bNzH4qq6(q`}8#hgShDG
z6vs14O}ovwH&4BFO9zQUyQcKfohggYul@AB+>|<`67F-01`xRscktsix0$aPckfOw
z2heUAPBS<1_{;mN5cgCb%EczHeeAx>na&*Q)67#B{7tj@knu1rAZXTDN4Rav!9ao^
z{x(?M$~0QX@;zEzzjA1E!eIQ$fcuDdnwCYrvu?5J^?9qZAe)}^{VQeC7re`g%i^-r
zUG|BJ@F%%ZP&68|%x87Xn7r@JdzTTf?||^Nms#ZleeJg4*F$AV)5Qjc2k!^N7l0mc
zIie`gkkvw!TG4lQ=yLL&6Q#+i<&mGyXxnnjPS&c;u=wth=RL(P*K*`EL3EmE!)F=3
z$X{J`wH3*7$1km4nd3X?Y8|x!K@tH!8V5ip7pQ6qXiX2l(R?3MC+Bj+<V!3?q9F@f
zsRQO-xQdGgHI302QxbG3L+niILN_%>N%z9C1fi~;7u%15f|k$ZpsNhu$X%kgU)`(s
zhF+38r4F;u0G1d05atvW(O4H?BHj|9*&UN>xb>zIHEYOC(JTP1CNhe}0k9#aRHrb{
z^!}Zp26Qo|=4sD~c8=}mJD_D9)%V9Q%~4CPF|+FT|AwC)Kz7EnRE!06-hc9~Z1Xr@
zpdL3RuHpHF$iGRHa4X*SJ?E`mH1Xq=(vvB<Xoyn_8xnt;4-lWO@*y5rG;>M)n4Nm$
zOP3m=@?|GhbbFI3%i^=naT$AF&f=q((poYv+ji2!gJ+6HOGoBBmqqFNXaemS@Fb9o
z5JDNLxF4p=$FKjiH5@e$M9NaOrU_9jB_LXWuQ6S-yX;ff>r~oeM_aBAw7Hl2w^|0c
zaz#|fvIaLo#7@wUOU#~%lcb;v$bLKrk(56r09SMst42nV(i+AJI`(F|sCz)^7Oyx^
z>D6-#;WXM43M+CfyC+yJou{|sq@m7L>PLwVjw*)oFM7?{B{heNsxVUKX(eiVw)N$u
z^I6aZw`+q1nlT4FRXpz1mcG2$rn664+VlbsDY^s|hoW3XjVGH^)MJ+FC4hk&{)`hJ
zVQP&F3m2yBleu}5u!{#F+Q_8>$VqF91CvJ5_Hi8deb4OWwlK}cyrk=cmh1{#48xlg
zbb5?|e%E|#%Uxsr*9GXqel<N!u;kNVg>&sfk`jWwr?>#0^bu?7rITWH*IAz?u8l(G
za(^6mqawB{(xh!%eq!JMd>O?w0#9!Q72+B^%3O9Q&f^5!@4=ch#$rIQVU|8gnli-3
zn^W`3>C-J?s|bIeJr}_PXvkoODrmjz8O}h;oAHI!cZFe&kKd5o`m^6(IP+jnJHvIP
znE=T5w1TUx+M45@tq^&q{Gv+dJv~RgleSz{UqHiMTh9xoBe9a`9q0U7l%p4kdBu-c
z=+n#T4%WsC4!4!p!gimjN0Z%i=lz~d&2K5nh5wppdPn#u8;a^Q=_)Pv6FRTs&fiyr
zjBDV3@3R~EyNw})qg4zsv}W4tMS+;>{tT^;hgn_1NxvpB!9_V+mc?kWz2^JZ5Zca^
zNm=T`3VONV?X5MFQ%^6M|49BKs*82RHQXY)cG)+W-()CIHj1rJ=Q^uxDz{H@y}9@G
z$bqxm%IOfA`j|LtZyJS2nwTBC79opnvJLx89fNH=L@a)6KU(iFf*@5$E;p72I5h=o
zn`g?Wncw&9ysx~jWje6p<9=cCYrj}!O=aS}XgLX|dk-NDw?#ID1J|Go*91B*R}gCl
z?)8fK>`R<Het*p1{9)VJUhc|2$;#K*F->FeJW9EiQ*G48VpykRkOD=PwYQ2=LViU{
zcZ$TU^17v|3Eq8aY#Y1t^&uD5lL3<VkOvq?z%XmV?Xfd)Jizj;gql-P>%H?s(aA~h
zl(jr%gO;13j%N<V?sw+gc~epwYF-w;FzmeI{=tmeCA`P*di$Yxp4_~(8L(B7ZtFUE
zsHz6IPOdX5oT+$f{LbWI%d8Hm5$d@2Bp3GGf7|)@KO7nRpN^FMe_rSQlW&z`nIhW|
zD>k@!4WBg{n2|r62i*Y@*pLgNo6vu_+@0~TygDp@<z&&<LToTtWEj9VqPCA_N`c%$
zkqt@lDchbP;Q(d7M&dpIad|+Ye^Xeqp!gEq&i+u4vKI>`NXcM52}onoh|_wofIbu#
zu89m6#`UU$tuOT>4CN|788kF1fFHGqERgWj+a10mMenIxKR!MPH4T95tkpKro&#;_
zUl5^iS?afQHj6?or++=VzqWnhnqI?$gILS?&2=P0#GVCaipUB~wAQ)^2Nk3aA=9qk
z=cY|EaZ74!2*b6$f*H<+xOWoAcft1Sj)WbdjL>dg(8((~ll-X7=K$f04HImA2sz4z
zwBZvF6EWawfJ&Znv;H)d4{g|x&Z_<#lwDc}CvRM^^?YF;hxZ6{@=$iIT#cPtGh~3I
z`;PJ|Qy~Dt=tKc(Vg>P7kx{2{B#xzt(!;jkC>BzyW1ZxJk<b;4Aa{g?G~9TqtOe_&
z;HEfUnzdN7L1_9)A7JqeuptT;KjFUb!7&9CIcfbwki(Hmr`8_Hh=RSB{)i3l6#97-
zI5~=q=k}R|y<nZ87Qb`u3WTvmm`|X2fjJ#c$|9?U4rA_5VRR!m+URG&aFhc8CwPyU
zz>2&A0pl7E^=%E)7z9lntl<9EWXvKrJpa3f)~h&DByP<dE(bnE3Cfg4J;aXV%ff5u
za)6;RRX!o~4Hcf|^HBAEs%;Z@^5G87u(R!=AVX&eLRnW7n(=flFeCFWZ^MvA1Pr-b
z38Pyr_0n#*Qm~9@93v`wrqWUy`pcP<1O>2S1#HNrParecvkD#t%UhoMo<l1SrBwlb
zII>v3EQQHTnX&22DQJ8B#qw~&7t&lmG4nOj1aB||9x=!s^>RDaptN(c$O3T!<qO*L
zL#z|(Az1B6v=wc-zJsiM*VCO>$A3E1A&_Loy&NZk)S^nYq-Jtc)T#<^cb8MFZH&}k
zv>mvU<GC0O0c>fRAQC>~8lM79`2g$+c$y<JJc(t}@+1Ap(<nr>$l>MNo>{mCSCAyS
zjrpdyQ8N2hJPYm1qYMg1RhC3>od^r%bp%-}P91v$ShS<6Wwj7|v~wz-CnqQ28tcJW
z74!o_FmZ&7iPEo~!LGNzBS%V%W5Zq={Ei^&L$E7=$IjM`UJZJkK@T0yLJJu!U-Vde
z6`Rt7k?jBX6T5&Dix2cSN1x+q(xog@R74p)><_U5PaIq^rSBWVDG-{*Aj|;pS&F!h
zGgKn;X&Wp03R8{HIQ9J$4FLCPDPsB-&S+TYSS~u`)|Dl@_vKoGt@|HXHVBMl2bqIM
z|7t0QOkvagIFdIERNidM8GN}&fH%Gb+Wh9zHqM^aT&eJlGxYO}3@e2ShRhV(DIRyY
zX_M}Q(`TlFcwt$DM4X!^pg0$YbBU7!9*=QRR>Q0-I^4u}<`ZPyDs>9A!A>{Nia_`7
zAJoiBLN6YW-52VWi}9tJta%y->dN05hbiAT+cifegfLX6by2P9Lm!gg-ZcF<HF3E}
zpI`B2ssblGg+2v##xj}DGiY}Jz0{rJR+iGxi(4r8RQkbBtU2ArL_L`KfJ5TFxYys0
zD0ZGY2;ug+MhL}`@?d30-%y;U3lg4UPGQ|Orv%00{j-9K$1m)9HCMirs3$p9iercx
zPZJwrmM>nSY<SK<g{-UMz|u@v`^F0gSY!P~DmO#Si!75(q-!;u!=~VCqH2F&-B6h)
zFRx$jUb^*tw&Y-vX~2#f3C9p_BX70ApbZ3<ofG{V#!oT=JAZH-@8;!uHM5||IlM6i
zs32N0thy%R!bm+PpYR=S<Ml!ym|eW-^hH~Q3~-UIG;-K+@WD}+Zf(4}GvPV*D6Zi+
zOB18r=b?GI)N#aJbtu#R{#OYRp9f;U>W#deClUwqjx|^uKM#_J1<vW7X2L$qO-fl#
zBt|{A>643}l7w8?6Xm5ZGSD(F@GxMDi7g4=u(}@@G3-S)r%7Id2f8euhzVGMkb+Hd
zR~T<VFkN7H4A!brJ8HRfXYt={e%2QAI6SsEYj4ryymT|ujpo#BBTtCZP-d#pL@Uc`
z!rmrnN$dvCwx!@H3~B4fQqv|c;P|n>Y56?#dopXUI&9f-M%BJ49Un%b6#OjZ)x@Wu
zo#>{MlgBL|1)rKKz-qC2r}$P_{EYil<Wu!ib1owfs@g2?NIU|L7rvuu0+k@{an1xD
zAo<an)!b4s;a)Lo74Mm$!NHU<zZ&k4DVKB_Di!zj>ti0>`ELy(CHp_^G1BFNU~r(-
zYtTE)O!T=0z4DndE2A)^9%}T+Q*$)5Jprc7uk56-jqJox7^+BClqH@e00J!z*@2^&
znZ_hgBWPLFYgtl0^%;lz?>1~-Y5Z0Nvss4*BlzzFE!2p{%>8lF$v5ibC#}wJ?a#bP
z2jc-vx)#QA7^{gn+=sba2%yoR%P|K-)zLh}E5TE1<xBjtcQy!Dll0>xCPLoErrHN{
z7u*MI&qp+}B>PJPeJsOUhm{@2jvtQ{M`UAtaSfITcaw|L6Ztqk9f=_M)RMsc8P|yw
z7!9*>y-Rn$mK(xbgLSoOFTj1v#^?YroYz|*00CTO=zEMigio53cY-d_;JhX^0n9>^
zz{H=3rgoLwfJThKQLzX4dQrX!S5pYV(vvO{Khk{9s`$%gZf+zrzyO8SmeqJ0m2;D6
zS#{wNa7^CfH^pS>hgDsWASj6lsB*`9XOzMZQL<AjJkWQNw^ejo#p-EHa2}onu!vXO
zlJgyfvj@^VSG_w|Cw|t+-fsnux)3E<y4Yr%D8eZNwu}(`eG(pNduOnK(zmj4NXT&J
zwtUv&SVGul%q^r*ZF-Y=*)wB-=6#lZPudTGoxLBN3CRf3Yc`~;m^A>~t6x}f#`A!)
z>*uKtkNXXSbC1CbLeqB|t%VMZ867yH_~Tmuhu%R%%WB-D02p6Y3^VNWo3h6g<z?^h
zv3@ji&q)<P&UrS2A-&z{fS*G-Av!|osn2t^&c!jCF3RG@b#q{^F_v^sY=8RdvIO;=
zRrR1ex-?FSY^(}z^R<vS`2LMMPqgbHVi&@R4LN}AyM01Nnu5(7NmG1l9EBd;?JyJQ
z)gKHm8YGg6V4^F8-R0FmVXq{99aNs9D&InYRE1PJsX6c=fsSlQm3;5NP4%`vvtl4z
zpl=TvJ2N26$LUwsOx@mLJEsme7o(mJ=UXsCbYoT%QIL3F#CQVR<+&jtb#Gxfbj!Aq
z6g%s)3YMY}C8i9K>(dF#;yCv=;}ackfDigTw4jA0L&QW}w-)R!>c+@g+q;F4v3<F2
zA6?R}Tf2(&^B$_KcwB7I^v2?xBr6y}!Xrv8+m@jevBq_)lx(V6l78lgZ($d;v=2BH
z-q;J-c}&y-18Pq|`TjeK_Kax^hF82P!`(`EA7!jyDLupB{ugfFK8ZGz4f7CT&W*K~
z6Qn$e;32riWiU!og#-~n&&1->`h*^D!%9=uF~>{GO>E68id;r4@1Ka0yYYh~)=8rK
z6zby;SPBGhPWnlqvzlm#9bM&}2U)x4I|y+^7gONaT$AY&<kpT@Y7dzxVnZ1&AElr1
zTzP3=;5BH81CD_OoAZGb7-_+|UYcl&5CNXlK6gBmq<rsyPNbp1ZhlFIYISuw?U!v#
zl%&fquubps<L>6!q`)L+&JDw66vYM_DVcnBK<8M_{Gp9qUTO%9LL7MBBczdGQ-+za
z{(h@4;DIH}__iiRfyN1xP}lVnZ2LB>ih3PkRfB32v5Tzs?f{vx5ymC(CR>IZMJ>iF
zcYOXB;o^f2MpoZ~8z3DI#O6deHl`YT0975Ncq$EhhT=ME>%p<fCoyKw8n7LxkBw(2
z0aeqexZ6VIRM=CWn?AQr-hn&FVw~uORok}TN7kkMT{Oc6herfUPc{BJD|<I2e;<z~
zjld8}n~4K+sl+kKwnooD1sWpiyG>Bsi*2mGEHq*D?xS=Fvzq`C6%WGrxmDV^-dfd<
zHo$F|u(azhTVJEXZ@UgfOC1V(FsHdGk9EK`coJvz0TAhTmGMjwMb1ReKODH?Cvg3~
zX?@|}=r?N!=W5D}Rv=|2P=#C59{v{F2}`;qpO7wt1WQ?QX!Tf{C}Rrw37zw&M{&%p
zSq9c8?V*_wqm5qjo>1zAsU80$u2l6!f%9lEd=bb4ZT{OKHBE^EjKH+wozDjFTg=(E
zZ_j+cIX+bG@m2>^oLY3@+m3@8H3HXuoW9B}I2+)g`Y0XsH_5jm_PsluESdtUkV(p5
zyXn?rssD$wH;;$<ecQ%IL?IJpU#3EoEea(Jl`V;qJtiT`gjALoGf~JMJ}Al%Qc06N
z!q^qbzVG|ajAgvd*6-^3+|TQ|pYQ!Vzu*1({h5ZDd0*FcUdMT!$8ns(gjY2(bQv<L
zj&Rg;YC>qSXECixv$SgPf<$Jhu2@|7Q=|vZ1#Fy!r(k=F-yCV6&a&1qWxq-3d7euz
zm~C;CyW>pZrzGR#s9ssxNgwPShAX1xhb_|7#V!5`h0n`Tr5G=?)sZdnLkkVIYooD>
z1a2<R4;RJOUmHkaL2<9<KZKAhqyajA{)`XgLj%?GdL{QWH1DWfwVS8a#7~$Z{J%>3
zpdyl~9N+<@laCIdO0*-hBZWpr+`e`cYjU{0e7sXj=xBv8xWK+LWByb_@@v}pl8-d0
z;oLj7?{PVLpYV#C$6j`2j32_b>+$J|81Tvad>H<a4TI~*&Ba3}QO`jI7S2eZpY<?z
zx^1d8G?P>n-X3fLbQm#p`bM?8@ArN(cIb<&TM<moU>bZG!LL+eDndtgB?e<{^xiT`
zSiN8IOiSPQ7vtPBufvfHC*mZIx#ji&xspVy?%Op?>@m=zuv#~LgP%bCj=zAcA-;uS
zHfV>L?HUWK<)j)Mr^onl=4hPp`^s20$1D4v?$`bKW4b2#{pYALAw*X0J<gm$7!N4I
zv|_G9xl^Dg#`m2p6ICi((tdi-CFh~KdQsfMpw$AZ2?X$ysd*lWwiqqa>;C&e9wITZ
zUK6_;?u2CGWcEa<WgdBCMlmnTdYKQ-v1qRPS-`DP@><0GxUu+ORaR}<aFSo~%m$zJ
zq6DI6zE3r1iOrJoHKw?5Z4oH~qdl_%&$tm|RrN+I1@Gu6ft?3g+Rf5@P$Y)r2O90w
z=4E=T)Ez%v$y<nT-z>)-M@M(zCKK@@zFN|X>zSoJUcwJOrH{V}au&0Cj=?aJuu|Ce
zEjwh)<5QmJXzx#XW1pY9c}7?=^Z~Ynh1y3`1^jt!35-`?L5UXj$mCey<h1NM!nZ;@
z!@ZWhW++ZCP4%(MW5)`c>$)6S?8M}6vjNUl`m)OpKj*O_Y#D%V73u~uVW@$o>Mmnb
z9kll7s(;D1l*@@`%DlLlOf!zY0lqzf&7U!=HwXVRWAD)Q!LPJ=VE2=oi!)^!MD22W
z)<sy(E`QD?PwYDz|C@Nd6b_LpK+nd69*aANDp9AWbO_-FXn$Z=VT`Mcx46mAX{jx!
z=wU?Y9OUbntM~D&oP)u~bFmVF$j6x-3#e>g`WBD_{_O&RRM#G)nw^Jmm(Rhh79!R_
z=DOdq7w1Zh5y=b9f`U9h>p6d_-!w6Lr}PF;8!pFXlu{9>js^PBV&?Up#)+L)B}Up)
z?Fteoz&*1d`$w0=^4ACs>y!_iVoCP3yqg}U!V+_jX!dxd7?)=|#y@y3(s|^~%rP-+
zDZTYCOE0l>%9M6MkG?{^OWL4EQ(YN>pah%A39In+J(OIPG<x-Aok+?PsXP2S%SMC<
zq}v8`dbgqDx%$~92fk;_Bc6U7LsC8!af89quxF5!GWh&nj6Bu&@EVRDI_p$hl8gVB
zg|*!SjfP*HjvW^ZHLRl$nqR$3oOX#*b(eY+W+5yE;x3#lLl^L$7Wnv`CRb0z#%H`r
zZ*Q3|eGAZcofA{mnwOE4J?P<3^<ryJ!3hH{vQRzxEhu)XO#yYI@7xAw?x^v(2NRNp
zBSQOfR<$E??I3`P*f3DRq&;WwW&OUA4IyK`6q@vN&m&(Y(g9>ckgS+o%2Oy_dxgA4
zQ@HWeo0s|icWmdZGiL92@>4~p5Q2+LP10IQMHZ^zqq~1X<cV(~4f_uK=F#BIFz{eH
zs48eI7!tn~Rfnv7SsQJUSK(Iqdj41xZH0MIk7l~eWP@@!8VYPTr@gsW9K$MGZL&*}
zb?<*z`}Q{*S2gcfjYhzi%0<F+!BbAI6qe5LKky>cInm@cp%pM4!A*VP^U{DNlJnff
zl;XlPWQhkslvgSu^u+<qQ``?Pc5>ulJHdD8V8v+kB1}jp|G}Vxxn3<KTDd9o#w|~Z
zH!Pvf{^fWpkY#R-F;b@(NW+e6M)NaUwh*gP_m`dat0Sg`Az!=|ovq(RvKuR(-&eCm
zLeP1M)2UG<uM~HJaaT!=o3i&xzemnJ?zn5#xxWAo7~^C)(-B5i<`OkSV~$?MG)5WG
z|Hj;g0Kw*g!-E04M(Ood-Cv_Fc$VM3^edoEjlBzX4C4k~VM4cp%%?`)I+;`!lm8L;
z(cLbQE~>v6rDDhPq*j;^=^Q(ikewi+ZXRwabc_mer{4agd>izzXIWOBy<U*h!%#!7
zV-FUWgF-d>TIpbMLoYLXH7W+a)Eo^6U7z1DO%$l?p8F=>#y-l*mr<M2jx99f`1uD`
zF1S|hPdnC-r9}ukPFA8Md1`x3JvA8Ux*llt(}2kpmVamFe-+Hikdvho<e5nVbrF{|
zFbB*%jPb?z=7ZRa80jvItJ<qd&9vu1?P+cEY?ESMwH>zGt$Pj3gJe*X$e-MxP9nGY
z3ARudS*2}q7j!@Q0_yjYestfTtr%PoZH);#%F=}G!j%p*5QpMT!o7t&V#>>5djbNH
z^S@feTy$HXWt8r1&BQ|r4&Qw~+}4{Su-T3bos#2kYU)y)Ddrz(>8@b`RH2|ogqZem
zkD62DL-c_GD|6-OUw5zESeVnf2vfF20C?z%6K8ggA4@VKSw^URAjt;3PUR9ia9v$l
z3VZ3P_6H!CMl|(TsVJ0&*8PEL`YCOV*b_l7e6mzRuIEc1=JMost(!w%g{Lc(i-MjC
zrft~+mO6r*Hek9nWq0u-Wnp#k!O-^4^MoXAS<Gp&X&q5u6hb#quMDddda<ek_+Z`@
z!oem2V^II5Kc4Xs_>!IY)GQPu+CI_%pF56S*O^##O_i9m>a?KO(iNfD&vaSxc>HuM
z=d!Vhf$3Wp{n!_yEVE+1{0v3RMe;2PHIjBj!N||~+_jvC0HL{k7^BWVRc5MVjBwA9
zT&2Lq$>5HvdiUg3SiaF$qaQ+JaT){jAHgn}TyAZ|)qgBr0g82hV8UdvOI*ta%hfhR
zpl*RAGz@pX8~5282z(Y<DXNSzb4mtDDH3B+uxMFxsHjU^z(!>ES4=+p1^GIkU33Ey
zfP8tuF#4g-_KQULV{hncURav2(!$FI&hyLW&#Ho==wG_9ssRLD2>metezh8{b>vpE
zR8<^z#G8-Oy2rK`a)S_cL`X)x85v_MT~mtAQcK9Wo%h<?H&ACwck$AKBvAcWfb2fB
z0b`S7w)L*)Yt#rcz|)+%MtA~&>tcUucWHs`fH9OX)v;>qf1IblNzx=m23F)LB)t3x
zg~v+}orqD0;@odFKwddl)|K<mWr18#xhty%tf7W<bmJl`HHp*`b&e?PW*PW9Hi#$d
z{aCYf_%sLi=e$wN0<1j<xx!2>`{gq2<B<IaBdH}P1SR>Oq8h*Br^BVUS-;5b@cj=N
z<Q8(+oN0)PA+CUJ^onrp--V|h-nf}f9CdNxcg+v*g|334?4%oUX9ldpyR;hh>OZh}
za6(tIaT+vH=E1WkF4ItM?NX5<jPQ-rJJ}5Lz)p_!IsR)a0$e|9%7E9764ZC-a`QN%
z+eyWKB@n}jBtCpb?d0eH!ZJcQ)v6Ib<lSTv<jI9{Cl@rm?Fp$A;HjwF6|>q}4Un>0
zNu?hhp{$UScN>C0e?)?_H@n|g;o9?F#kEK-kEpt`XK_Z&=LtWCVKg9Gr4dYAX_k7q
zikS5F)9d=wcg{EsMs;QSDy6$EX3xz|;+rs`xdW0!Hhiyknz1aOVT)G_)JnJgc~awC
zD<~UcTb5CqJU}e-7bL*2PME(LPunt<{V%9c0-z%2U!cO{e}Iap?TsN0Ca0%A?K<5O
zbzmHl7-{gEa7kO*+24CDOW>URLr$LaE2sowi3~l3@E=&Ck)+fZ8#L9_wALIF&@ZU|
zK(!>v;3d;;-{;J@Mm|~Aor{!Q@TTAGAf=|%aUsYxx8XCFy0S@u>B%F%*g_?5A8upz
zqg(#fEq(=sWKZaiA1BSFRG_~vSS0<tqoprcfH^fu40kvMMLr-wk1kjBy=9-@l$3(E
z<oY7vHTwgSZbL0o`2EYn(rYo)xU54}`SRD)vd?o5Ej<M+qAnoukL<$Sg}k!^%0z`r
z7E6Vm=X;+y_EblU8W%X=_`yT1(j;aebwJr>_)_lN-}^7S0#N;vrx5UA2woR>J@$ya
zU!X~hUtJ}zU=XH<qsaqIP79i^2mVNvXAnr4>uf*wW8pnEnN?~1mx$i~xN61~PR7A~
zbP%?5u*jk1nDY?&k^lREOWoc=Xtm!?2@md;jcQ4Ou7C0-?r-_02!BLX-zj&|7W?wx
z)aAWz^nRkYj=T=*$?BFV!m24bd%8cF=i?NC73JWb7*o(2lOO|HE=O2Fqq|oE6e>g8
z>6f~_vV+kLY8s=g8~P_Vjp&yKMi9lW^rH<^@hu<MO44cZke;TI61~!)?6}l6m)xBr
zYp7V90MdmRg<@U1xzd~guj%r%r^s)VU6B!92D(3jLa4jK`#3lIcPVN-iN_n{SYPU{
zg3enGAe#IWZYj(r7J3T3CCNM(LVIEl{hl6~GL84;sO&2&H|X^eD;~WAA5+#$fgMR>
zsAul0lB2{iifY8Lzo<|CYW6Jomb*`FK5kQ<|7BM;$8lP#-HcI2^x#4c1x@aZmwdlB
zuu<&!-Zm+8Q*7S^>2U49*3JxGe1*uTX%a+$;&0$@^?8i~TzJ9h=Wq5vP8frmyh|5I
zZ_-HaX_|^>A71kN(I~%l`?ZAbV6t4uZNCH%v!b8^*T>G)J)f1PeqXUZNDU*r#PZR(
z$Q>;TJ|%dzfWQ^|Y74~`9)1ghn>H^H1wmyXIIz`2TUm=TH&9%vEAPA$`-RN|pQ_-Z
z(_l3CWCvX!=+c1dl_Dr&%8#`3u~v|ZiG7r`{#{>g8wpYo0O>F{NKJV!G|z&f?~;|=
zKWT!8AhE~Vee?z(nwx+2@wPiPtRyA1Y~3kE*or&mA}`7cY$#12t7D)kK~Q#{n|i7@
z@|kM>^RkNKmx+ajYo+!V>gnwrK%KCA(^V{EznXCNw=n6*e#22OEAyAS>vEHL=(Nn*
zTYr*4{eu8_y_Ki;bb;4NcfwsS&pUJ0+z-WOLHo1<sAVQ|fd`rENcvk*Q=jRMqnG<z
zyZ*o)=5^%?!)R6w<EO#CYQ-4$eVI!#?EHMr+IsH9_{m`{m9};)cFf)+*jKgv7eTA>
zw{xP6Q1QhhiAGvRsvE?8NJm`MA6N-{T4h;C=1=zU_OBl5=Wodj@*M<0vRi`pU%OSe
zTX81X`|oZQmrzzpH2roQz3x2}&*VG{dV^6ZhoskWQ=;U{F>kfE*J7*p84hv>=Wm+N
z<gLQV-jj;)CrdvssyMjaI4sqxaHjS7c5A2!W+!i1-PUAVu+01?+MhM@MB2&W9}QXT
zz<-NVZdwWC)@zYUYx_w|Q%)#l<2O0(nUJ|ugo=GFjIpF$am;EYrtVg;n5?iAC6mW=
z+O-MAg+Y*h4ev&)1_jHQ#)zH!&1z_H@rKb}{8z&7iD*!T@;qxaAX7|&KLhaYT<w|4
zxU92wh1y+f$lo>KjtPwi1<Hx$jkw|dF!?n^gyz{N3bAkBDQ^xXyuLQLbD|19xll(u
z3K}k@<zZFQc{N!df4$eYXqJpQEZp(+M<%>bH6D_;YBC<lq-c1CJ-;xojo6X18<zlo
zx(nJcNj&=zx@_8a6!9uJvkCw6RHXr(*L3=1mUFRvz|7^IIk_aZl<q=~o@3KhtdwN%
zOE|hQBkl3TTJX(rE!T!+zn`D1zes}CQAy4$5T1ZA?FnA=8lw#De6(IvfB16>f7BkH
z(!I*ds=jM!{`7i~)u3>PWCE&n=j~Wi*sQu*(a-(A6s4l9U*%_mz#x7=l~xW;OB*AX
zP;(I@AcA`F#kqoP89IJ#^GCt!W0U%G@s{T=fbX@^QKNzhAeq*pc+^T~v^>OBM%Y?-
zE}JN7tX`WdT(cjWo7D*ZJfO~VA{3<RV=!?Ozzr?40;C~k7hbqD*xd4UlIhnB<BMF<
zcb&q|dFFgrwQ~cNsR;;t&HbNg+NJ*PlSI~*;M|Xm8euzadXlugpzBDDP4UZ2E?)B8
zX~@{q(Gl&P;W%7?Z!EwRkXR|c<kU6E>4i>B#2v5qA#CO5Z|;}JykYkWyjRm+0-lQu
zUH4n@mJ4}4jO|-TVt-N6_K)+K2kB?Q8-5_C;C%!s#L!jW^N%7%{-ahL+)e}kS*Obs
zVs=|~E|v)vxl716a}WD2?|CfH3M*)!NHT+8Q=lECUty%qM$I((UD~+vx=b(MHh#3&
zr%w3Cvc|xRMUDCFsp@+jx4OL+zV;S4^VL1N4^-I^zLHs>t8_1FuY&?-WSwMWF<F*0
z%HGda>D|?C`*<i!pLD(Z?7(RjLiT^pk>>xe9nb&U?|l;4r9a(4P$@|O^QvM$peHt2
zdvPnnrVO@Djx9+tih1=f0e^*Jx^O$!7YE8W|I_@`w}<)5;Yc^#hulzKsaWjgRWkCu
zB+l)b)58=%W78~)Vm>#eAHh%Rf|5o3@pxK*hbiQbtc|B0MQ%ENVBOmwbZ<Z>+!3Tg
zy8j9zg-BIJuPgnpSLAmx;n-g|IF@x;u)K!53s<zRB)R$C`%Kx~kMj$oW*nae=9TDz
ztjsGNpsh<eTm`ho?x;MAy=bu7fU=Zub_&OMu7upSHRG6##ez*&cLG>Vj0v^`rWAta
zwn!kV>Ahis`-5NV`oI0X9%|hRiK6!s#z19CYCT#2PeTM?cl1wIegE!CmH)LL7VMfI
zV5&b*Qn9^u*|@Z_@U3RWvBcx0somhm^$rNgW$L1*pzhc5u5;pAqOD~SC(frz$D<ya
zcn^q^qt;fn`Jpd!XKNJH{6zz2j{`&AY)UzlizDMG&&YArAB^WKCyZQ<xH4_7tVU8f
zpHXf>2JjYmd<L8pJmuU|%GGE5@yMrU7SL{c8}P}=XFAxy^7oR`$wzqqlchBxnCoPC
zJ(AxpAi6<4w8gUgLe$-mGA4+s@9|ualL^Ty#3|cdt#D}2rm@iA=+&clxLsE)=-!iK
za?>1Pr3$Pz#Zi=csz8+CynRDL9ral|mIDu+;%G@U(pC}%L>3p|APfGPdn{fsCZ6hN
z3)>_NpPhP07l0;a5c>@zJJ{o~6!O)X{R6OP)PYJ|9q_~*#i%jzN}^rg#41@v8QV<C
zaQ<#MfXz(X%bS*IG<h_g7p;Ubn!cD7ls~S$w{So(C~(rUrB>^zr_7B1(=W$fHqP8G
zx;e4>x|0`p#WP=jxhFu`Yz@>8i=RA;`t7=wM?UQxg)hmAEq(vZy+CBU#p88ri^!+g
z?2%&w+vWtdF^(IKe_-0(;3mVQF&!|bKGGABsy4YDuXb-1cF^XZo~sdoBzC6_;z~>p
z4eEEBey}&-XzRQ6m3!jpPr~BXq)hDu51j?_$$_kN3qx7Ze9_S_dJg$0!rX<ZX>?F{
z4n;2}K+h;OB+t(|`}aKo#khMaxLlm>QI`F4Z$!dM%68A>Hu9Y^h|}9|Es+^T2P~V=
zOzbZ|91ByW<@2I|k<(QC;g2Z?tdI(|d$}!T)>i9HSP$SEac<BHD@ND+?tf6VChH+@
zP}vO?S=LQXB91uu3b*Mpz8_p=c!tCd#EK$Lkgg&3R5tegfsJ%9;{#zkh~Z+U$wu?U
z&g8@|t1b6Oe_#jzJ!^1paYndOS9<zHEXcr!uL$$q3ruZk@NoXXG++!d@YxhV%4J6O
zX9Tv;Gl+NP$ZgC6)ejQrmgH+!=$p3hCu+5@&7JI4QN&rBKIFe5#K6`mvN7|jYcz6I
z&KEFQY>DrcB<!OR`&UwBax5QK(2ovmDUd}b?ddvX`I(%w<c9cW{c|kaL^|IF^E$DC
zh+3B0t$uYn<t|JNFl0gD0AzbIsHq9U4nK`BV;o*W3NP1Kc<g6>zUB6FAV0>VMh2!!
zCz=SuLCt~#DzESMc%iPAUzlkN8`78ijb*{8qoOB-s}Pez;?of}y?09zuZX=XoMl=h
zk#8;$u+jnVrnYMA_5ClbB+mAEvp`!m5PCWt!Rsb?qQ|1>r7>I6WlzV>$xe>#c91dD
zRvI*TA6UYQcPX4G{`xiKxJBfp^<yw=T2i|k6x~1te(HABa<@vr*;zg-_Vw5I<Pd)7
zzy#?gJpNiEUN9yJ4^eDQp@7{MJXJJ;$*KT*0J^UWb%WY$z|T{W6r65!67Tv@WY~yZ
zUGf92#|IJShaj2aqI&4Ej694|;7K{BfHn6=Jgf2{y`L5qP5EbVN3n7NNi#B0_sD`#
z1U+WKnTFqo6>%)vpO8FW86DFc;jXd2ye0DP*M&4IpYG(?fujHjvQgSo92R6eL2g}T
zyy@|(8Oi%Qn8C@4-i#*4Gr7PnkRG`Nj=2oeT?c_rrey(yP|U!egGT^OYvhb`S=<bY
zd<}t4T-R%BlDN9(5d=?Fg<QQ6Z+(v9#vc%06P7qAS^$=Q#oa?j*kZ?O^#6&DhmIgt
zbR=3p%1^Q;hEb7WSb4V^=^pv)qIAn=o&*1Uvyl>we6)cow%7x?#{RPIek3<tyB8T}
zyyQq{yVA)q+lRZ1BbipCX+ab#Qdslah@=~C;|(YCiQMiM^I$xv%!Ezhs2fP;b6M79
z<LH323t~;8YCUbTTKRh+5NO{?b2Ls+j#T`$p+NQWw@@RIW5`xw>coqQO|!oi5RMn|
z43S=zZ4Ua8Gwh5~ZHZ78373u=_&xtbVkV2LR}L@R9qKAS#PSz10pd{aU$L9LB5ph5
zDl^25g}^MRNv8K=dl8#HC=h5?FM|CGPC@PyWM6qVkjf4}V_uh=MlvG2fM-40jKh(}
zf&1By(E?BE4U&48u=xJ}C~c?zfr0DFxWwMnfbqG?ni^f4JNkZlz~9YSJTvG4x*p3z
zH>!j#ljRziip8N9KR*8cZCx|cpzW9?+ehj&@h-?2kYB5aCBQLLgOn-LoMuw~Pk+Up
z_pzIXy1!ZXHO><S=mt{?=z76!o0UL|`Q(Q&W-b1#`VTc`B=Pl6k#5+d;QbwHjzVp#
z{aDFTBB?(II1L#|ntKpQDns<9V3)~1uoxUqjk0xhc?8`w8%pGQK1`wPaw+PHg_hFW
zEHk2~V_h2_K1sV?5S;t-&3bV~fa!b=sAkWnf@k>}x`>{>xMpRhURh~>TdS_*TWzem
z^4)Cl3PLb&<vB?|37!3D@BzA5=cl3g?im98yy7J}&d3QBgWHb-36T)<@R&`VVv*+M
zclr&#i+T@VY|Z|fP^7y{=qC&bQbtL*HJ6-gX&!0`)5hl;Qv<KHCq5-7-#;|EkP8bI
zB7cr9?LD*Ym_`RRm(Vf&1UDpC{fSJP-ols}GE8nqf(SzOWckGAJ_uYPC_y}x6JUrj
zoDPm;axuV#5sej3jZ)J#X}N{iB1_|t$qZ0Ntph-EwoO|cXw*dG8M28_wKL+5K6m6b
z3zTir{;yM*S*H+?{FPKEihZmTV$s#IYzsL)>H{QFv=ML?u9HeTqLI$xeq7dRHlBon
zFTcAI`MLP8uLyOg!Bl7Psb$(+=b&j7WDmX11-aefv9Lqvv!#QN8>$RtTGIo`?6LCs
zVpg_S5;N~LFIee@Y(z0KG7l)_xUDYuo_C3Tp*?9eNoYhtvQ}#>*#Rr|g(XlVZht}b
zXhMrflLY_e!yIFT@$Wx|(HFY(ZL(buH|0nREfXZ;iDAJ;iF5s>q9`tFt<t{H#O8_N
zmoXpj8nA&f#~WNj3-F|;p3;b5PyrB&%w|elq=9l)mr*ZrqZdUtdAtUeKacUQ&n7-h
z{sB8~Ru1%O`B4f`OTBICQ~HgMP>^+)=Z=Y*UwK8O^%*O1eFOHV+8_T~rS4~g7}H8o
z89dz^%oC_l>v0cC*;aj5Q>5e*3men6(Icp;mxXR-8AdR0F98~}vFlUf>O~iDBD;z2
z^Km=ifL8vMg%8V7grO(2t1CBFBu+*J>{w52N>)Cnx4!;;=DNnfufHGAp5&E29eno2
zSMvM(M=tw&AGJIAbKBYThn#wOHk4g%eIptrggFILcR_7wGWMgy>)hH4PHh#<exG)n
zR^_%H0n;ym*Tcr&5&Hrzd*t)-d<;_a=V@t`YnjnOeAJMLbegI@(P7-9kPVC@BOOBz
zg!A+2!}we1m!S&MLcM^MrGpAtJNmud)$j|My<Rt1Pl<&dB-F#BfYUNO7B6rm@)J~M
zd|vdpQ+KEB8tU46xn|)i)XZLm#zdsF4Ds#bA%@#NT=b#$<xOybn%HxXmXtn?Y300p
zlYPbIZo29F0|ampg?caRH&ALxY<ChU=^b(_n@Om1W47&1K2m~*f3lGat)eVWw;Gt~
zE}|FXA(>DwMv=BKetlU7)Ww0PPv_Xb>$0(RRx=xPfh3SHgb`OJz;2jt#~I;^aP;&I
z&@Wi2N6rHK55f|HqW{g9FtkwT6(L8FLcBc6va0%D^wtbvLOW<}sC_7;tCDgf$VLIu
z^m0xZ9ZMKDg``pRv<whNP=`S8u>gB$U_*w^M_`(m8iP9w5Q91`pQ<7!^U@Uaz5vVj
z<MGCRan@dr-(C>{{<+|@XEmvXyWyZoa!4l&)>XEI`@X0eQ@>T3*zYSQ{w0SG?$nND
z#}1$h;W4;-6XFiKp6|!e+WcTO#r69IUdlTs@{->iKBEHa&0n$gDAJ8Jb&5O5IRj3<
z9-&G}px`ekt{xTsho9B!AZ(k^Z5oCLYZi**FPKJ~3J$vss5rS~5Kh#ck||R`@6EZK
zZn8eCTgRSqK}5z%-`V#9cOGbkFe89vq|%T~EwW54b&B@Za_^nJ_PsQ@HFYQ9-Y;rr
z{cY#zf?%7rLkFj<U)n<&6ArH*Tq*yx*YSlv3@ZbC;BuJK7=>QG4IQ%I$Ai~A!-G;&
zj>l7krfdEHMJWhwU)2kOU9&I_B}Uz@oETlVP^7#o?A+BV#v>iX2c{tDOVVUD--Kb8
ze0ERCh)<bZKyfKY=1&6nT$6=qC-Xire^=@?Ju4cq@xb~HI)o$T<A6SW{jH!)F_|aH
z_2bf3ZeKF!vYyp{C~UnCK8sF*21xPx&>X49j&!fGfO6>i5#9tIVYpc>WtJ1hbe7vr
zrJMnKGRF<c#uQx|=Olj#d!+C-`b-?{TFlwx^UZToJjy63W;d$HG!EygBs#Gql9B8m
zl&jxVP-+(-Je8=OnIg>jWF6fo%Su&*oM=86Vd$ebmwU~*p3h(Ann%Ph`9@)JL+t<%
zm*7^<l6ENsvcR+H;JfQ+oBQv4Ros`uy!r}b7iyGI9R<~mWySJeuaa8xPLm5Q5trop
zwS}1d2!1T@xI#w+-SNZ6o19Hg&feBnUi@O12_N17ki1Tvxi3J^(VwK1QQ*(ZT`m|i
z7a~pqw3(cew<mip^>}X64hRhY9#hlfT;QyErn*O<$$;OwT~K8coEP9UU7F*k@%)#h
zec<5!C<9GcSIFkpgx1P-{lS^au#=#0m}Azu4EUJ`GT{=8uejM!IvNZzQ2~QYi0Q4&
zm%!Q&v8Rp%hB5;&c<MpI{3b910IL;^*wlo|nV?6ci~7=YrTp;EBf$eV^A`2xHf5Lx
z3je?)JECzb@!R~=Eggga-5RWc1x6Oi{k$wvKJ0twh*?d-`?MfUWyL^zx@sy-9eiR1
zO&OXt-QijDA%cCtJ1)bwo&~-_X8Rw-T;S9QJx}mp<Pq=xfx)nCX?H>9e>u>rw4%rN
zOY(g#qvN1pDm-|88NF3Joktfyw>TV&GpvG7t)RGgqE)->%BYhcnqb&w<I{GH-G*&q
z9xA_DV4M&QaNx*xQ|d2}t_U#*cawZC0{Mck-stfP>*Mf857-4gb~;WHC9smremp<(
zw7n*EA*NKmLOQ$X?ONvo<DX+z`1Z_z?Azt|Tj|&Pb`BiBEHyx2D>^N^{A>0wN5wgo
zXQ)qroPC9Ht%BSh9$=VwJJ?<id&_DW6zhQT{M&%NKd_~4Wl&QPc#5=^+v{Zj+%UUP
z<qATO%w8HZYUSU51S6$$srM=sti2OXbmu=BloZ8GK)OtX)QCZFQiN{jw&<1EjNuFG
zUViZ^?&AJ@ErY<SqK9N<#2kkTr|gkl8zrdbS9CPhxzmxR9g(Nk<5Gg&lRe7fH9dau
zgLAmE&54Aoz{>fK{sCkdP747-O!zRW0WtiniRu;YRwi=0oZU?!8zFxu<fT?|8Z&%_
zG{NLt%Kd_;UCYVuUv6th+BceUv^a27(hy*rl+T+d8evthsfO=~?+Fna5tQ<$Rmh;X
z<H7&jbX|T2vPGUhupf1~U~~1Ns6vFL=*d)iB7y_R9uxuNSLZ0p$p55f(E+pNKWmof
zcJ6*YFd3&8tP3u1Pzm}^ZvRd(?I%t!=8qf9=tvZL)2!(ZC1v5`2#Mvc22S}KnZETO
zgkYHJr=PXFXBHZfyz9(vvR|I>_b*F@^!()l8RkJj@+0eRdDR2=dg$qG0qaN955D}8
z^%SX|RZwu2d0>L9KUetVlRM>EuLg0`0|trg-m30>Q#v-fD}UW6?b5`rJ`cv(Zg55-
zYa}aMMzF_hZK%^fYLGI7JfbTRlJ>3#V|F&}JRFX}XOs@>jHE129({@%5qs&zIuEKu
z(n_oLAC0EvaM9yRJ-J1tC41ZFT`v)&-e4hC`b7XjR<tg=x_al?Hlf|w{QA3O4X48P
zoseCGiLscQam)<R1)@Rf90qPifcRYa_bO01xc^<2JU<7+e#6yC;#`6cR>5VOf`hI{
z+~`$uaVvTM*8d4-OzSNCA;%ml+d-Cuh`_jKx96?|s%B*8{Zf`Xa^=jE4li^Q@eTqG
z+HOMzbpp!V$HHdmHf@Ln)a6e0dl6YnSa_#rNxRChg8J2>`1@uC^;fFiz`aW8Ti8*Q
z$Vq69rVr)azPsY-d7)=A2A^!n=k!B@hTh$srevulk%&#|DU+`!d*&ar0&^SYy()_b
zyI9$NnU{Bslur6lQ>8lD@ePOwYF-lx;fCewy1TVwVms$KX?pVsGE<dtg!vm+4Eph|
zDA<sjgG6^$_*}MSb(J%_U8GN&GTm^RD2kg*sS$f-$H=E6eU;6qeqNFtyI&0uZ?P0J
zHgnqsA5jB&<sUI>$-Zp7ODGTlNfsM}3x-gzbZX-U`;xn3-nj9sho9dkzsT$JTQ61*
zz^qo*FhgJ;9_@7OCpsfuB)00H#V@RzYYpB3+IVJQ{w^2upc1)gVH#O`k~E+1=T&M7
z|J~X0ZL@6fh7kP>nnWS~)%^H_T~prxrMlU~Acriw_pgfnKGw-GfFs>PMGYXikuKGi
z7cX2hU+Fm{wT4vt$s5fJP{&&=xP0q8RAo>R+PsAeZRomF+dp{ZHaVXnuM;QZQN`C&
zr_&3ZrZc{~cz<YeJXP4^^!m~4sA9i6Y-6)bZl}<~$?Q{>&LXS{a|7J{u|1PLVBefO
zbH@9kMVeL1J=xbx3raf-1G3%yY>d<yMmCdkbi`f5SzSW8XV&I%#o(PUHkew&)mtL;
zWfG3gJ2N9uF=zS-<rZj?A3TRnISYq)S{dKoJ=Jz&fQ#Nm+?;ZQgeAN~K`EYoqDzTx
zS@)m#HL`w_5kGi=kO<CA2}&R@yrmwyuc*@P)^`Vo##m;luzqbrCvO_Vn9u-R232R;
zX}5#lB+WHs*(8e>`7i95aWo|VO>x-z85xi9&xo+c)whQA$4?9hu37oVWIsA3)tuc4
zLzAx7#B|~&Gf|?Lv+stMZ2hAwxkS&W2ssI?Y<`Mi0X9C&@IiJq-j{*TP%lA!!Fy=I
zm_&H7je9{=eFv7<10i{X_XC&I(4O5&2uckQd0T2U0Wus#l7!LWP&LgKa}8RU8Z0pt
zT8aL+)OAh9ke+-~Y#+-H(Ej?zUE`lh?E)7ZOJjd=&JuI;t0OgvnS{KDc+rEgfmkPT
z0`}*F+D)%TJs|~N5PODd*Ta&*+RCYZ^z|MxH;OaTtaDW^*Jbo^RsmijIRYGYgWmf!
z$^n-s%a)1F>-FA61n1!&LRjCu6=0c|vz+4J^|agWTGHu^=lT1P$v7Q+xz#GpiDrpW
z-}IbVt&#6e$mvmpAOHR;{{1wp!-FmXMUe1KM9!VIcDY6>|9q+s`c&k&we^)>Jla;x
zLcyS;A2A7%5V_jG_o;bH%=2A@$M&WkjXl<l5~GKe6OH%n4^AR&QPLrM^3C%G5t{Bf
z9_oA{%<$CHF1IOi8Kv}Ii|Vu1#4ayQhBkG+9=ErImeZd|14^Oq5*h#`IM8LG5b|XU
z^81_ttXJ+?Q*FG+hvz@88~b#qE)n54M`9?B+gDE9{1QnztcGxJMcy{Vo%_>FHar~=
z9^~jJ7&Y2w>7s7tSIM^qD4+BfeRU}XZKE!m=p8GHDtLpJ0d|zvy~jy41pwz9n24z=
zSn`%!k6I97%f16MlAHrmCWBK^VD^k<EaGN2&cl7eEFmXuy-B95gVbpMl$Fd*^hM7g
zAX&odC8RsyK;8?w#=d|R{ig+$%6neZ!rIpZm89Nua!e~VObD5hc3W&YA`9Lbt)ELD
zB-}iuz4Ff;z<FzyqsUFE(*=%9rAEz^6qE1aaZ_tIby(ZP91Cvk?nOIu(jSstLom9N
z_E*v3ID_i$FDD)e+kFr@W4JAKhshrX<e*AVnOgMoP$8$m6hUB-x^nHC8fEvuS_i2J
za6tFUP>?jdj@BXK5y;pWQS7GreM57=uEaDZ#XS0{3O&gNo~tDKxL4}OardH~3D*Z?
z_4ypDZUj3xf7wW3CcNF9TjIDzf2yd`TlB|cuXAtodf$Hu=I2nNr{Y)rNw`U7Avzy2
zuM-e)Ab9@r?<#)%DCS6;(>CEAAr<LF92u!cls-P<UU=y-Lcq3BI^Fp7rg-2z^<!QO
znmoxn;{VO5l>go7mH*$b!|nqo=>~Cj2lI$Im&Z&pK*<jDO^|MR?`fexmAb<H4=kr^
zpe;=ktyth3r&6G?PfOQPhhLVk7%_!k%_HTRLAcD@9h2>9Hn}AR{%%JT4OhA=y&-4R
zbNsf{5)R_|tFvuy3cd`2(8wKPX+&7avcD6PYRh;Js6Kjsht*uV4sxWYD@=aE6l$nt
zsw33*L#d}=eVGV0OK>zp$if<X%#0(~svTs1xIIw!d4c~|wY|Iaug(_rpLMo$iq-Ni
zomZWyNq@pzMNxSPts;!SAX|DYJ039-Uz2dDW$|_=$8wktQJ6Rz53Mk3)hDRecZRH3
zKuEc1FEbe*w~ol8C``liFExPRw55Ow#MO!5W*2=;X|N9944#V-{4YNOUb9xyKvp1-
z&b%&JPoyZxFynwfog5j*<N|e18D20NP%!Ba%)m$ok+H*22AA=s4ptmDNOS^xD?TEy
zwR!blC&A#?nn2AplI#lfvv>wO@l6yn_P>f{fkjgMzbRaI#)&78|E87wvduzgg0(ha
za>R#%r0ji#c=C6H?B&7UW2r)xiKZDk+Rq1`VWi1B6ZI334&r3{kP{nMtBjwwlyhvK
zBsvf$zoPadr_><1JZQXMVIV^|oktS&dL}vHdaw>CY=c25xEnPG$<pP~bjcjFT9QJe
zU`pXHDf5>`W~M7?;7N1#k)L1@%-=E%3C42{&^uglbL)dTCucu;3qSB|G9ZU(1G>2%
zRjgRJ6wYl-F&c}rj*ws3;{+DooKxhd&<f?bDfiGQBipdk(~0X@{qNdHcCD|ntWzG0
zVz9=S=z<b#5Zr%;>qU(JxGdgph=nJi@mKyf2bkK03ai-%hNFN<t;{1|0lBT+VNJoQ
zwnH9GwS#;nX(*Ne!ONZUlXd~0Jx@G;$D_WVC4Tymj=1!HjCVi)QC!_!+-!Fev!M+C
za|Ixi2S6!>+;j)~*Kx#jb_WX5;W5lGt&EvikLjuDI5XP_{ee9mSOM;e`Sltck=o)=
z*G@2(NuRTc;c25_cd=jDZ4NAV9>{H@nfwN!$i?d<pflnGh2Cjd*1eE-#P@H_C*cjj
zE}Wx~sLq8AiMO1HE_4$UaPBW)57TW(rE{@05fUN!N1fb1*toE%IEB01_5#CBpdp$9
zvoGT~BbjcaU7F?IZ~NgJi%VHGhdXO$obEy?utokcQRe*|=P|+cg!;H}X<WMW24zh`
zU$m2)s{7`jpS7wGoNO>PpG(SX86F5Ly*1*WcU$u!u_o(<6<r@bV{f|JPR_pxagdW5
zY7~0LFN+t{N2mMVmnUa%mVa14hQV-GOJ`6bNN3Y996}XH=Y$~JiJN`>%6zEeZwa>^
zvmT1?u0|Wxh`djB4X9FRJawjpb``$!MRF#h#rpuZqvqf@FjY@gy09qz#mtRL8+_5b
zI8&*|OJMrR*_&XYH8Cfv$kHjfp83mVYGqpoQKqqno6c~oD{Iayp&l(&2IoaQ4nm*j
zS6E#I^_n}UKYg0n<D^GJf1%nNvftMM0VWG+ju??_WKB>K_A-tkHF3rS%gqoY@udq*
z_-HFnAK<-gwm<Hg-&xfn1jGp_&(+Ao0Ooi<&tNUAia?RaBR$nqT0KS9xCzY%xYC#y
zZB36fiBfU|p!11bERl<zPwbrJ&SLGEP<@>jpx%tt{BBUvwNvoW-Smh;Rk7=68GP%r
zeUNSq>35-T?y}#8JI;f4GF&%Yzxe@91GWoTsOve8iNkqyJ=~$r&NHKTiCN1`U!vpD
z2{85T9IodfntCb|{-XnuZ>inNg|Ztac(+&VQ~VRsP4w?*Z_m=-fbwhU-dnVf^xI$N
zjUz6R4cUQ*A0EQ#;mbBaOH2*>TU)Nn0inIwKen;@k~>M$t%+j=^}As0&9Lt*rToC%
zZ=w28SG+#ePR9QNN>N?f0lIDc499B6)FG2Rntm?YEhTmB4U=Jj%;N81_Ti6s*6`6;
z$>BsF^@J0H#@_e$S~kXnWPr09ilj$~VHY=U!tp3d9psK=6tHjY;$_1OrrJRl6A0I0
zIE+YE;aKsyW8Q1-JXsna=P>VCbgAo^@So!5H^23(%H`84xXZ0n&J|TKPdb6wW&diR
zp2G-g?8aB=lKIy6r?nWr7}Tu@unQlYiulIA_uZRov_niew%f4oXkg+4IjSJzDU&;|
z4mXrvPdRJLNVV=gCYTmA!}0q9qU|Wl29@$DO(NpM@=d3<;<Ci4;>9R%Yn*1Hk!}Qp
zDOr8SHAd4wtKoLRdhWRdfeT#1k4g=`@-qbKZm%<;i!-ke^Q~k}@WxXocj!&U0(Xsm
zYJypxdo*skL@6{vGk}~f5icv^x$}b_@u*0}soA#%$DyewdP-=XC+o*sf3ua9l*itU
z5^4t4MKB~M6GcBV)9yd@yF?WS_c-F7_Vlt)hQR5O>nbYRq1B>HQI6>i2G4LR1ISka
zBYBa~G|u0n=`&SeaH-!@vZlv7%k5sMZnDb>V!G+D=oFwtk&~Jvj!H)j+84c_3gxeH
zK_u`K!H}R3vOhDz(->R}<<P(@D8I=zFV=UZz}4n^BJF0y@!W4O=U#=x;aGGKR)6s}
zR>mPfkn*9c<i-5qrJCagrQd(nUgRA^l9w4j$Yt?9)9?e!RL`7duJnA@s0$nxch9TR
zr<rY)#fx|ey0Bm3o?`hAR#))=W+~&8-Y>&o5GWMKHHuTOlgnQyxQ;Dc{%~8LR+%Vu
z$0QIm_d;m5(ZnM4*wRDZhCQC8=-})7%c@E%rP%d6YeO&mMz6sk*>d3QbF*T=t40!|
z(lfj%_s+R@`$NaC_?F+#@|_n#P_F_9B6g5G4ys@B^X%QrPd{)30&p*_SAZFr4_Sb4
z5CECZsFf9y4^;y&vLi^_G!|jJ!qv7pTXkuaztHAm_B-saCY4A?c59mFdT?s~P*4=m
zXA_HKnFMLboQGS+>Hey-XHpIOHcBeazU}u{;_aZbMFDa>$(4R2X;S97<csnU`BLL=
zzxCQ@F5IV1s>zLmiMxN5j_}V89Sv!EWeFa$0#QQ`cr@-2;Ye2kbZVjr`Ao3CpV~A;
z#zmMjTw<pr&_R)gQdX-Gt24AK0cc7x<m*07;Q7VAkuu+&u3~4@(K?X~3(Un`1RV>+
zo@RU$ks|2LU_n!*|4)+EC&2A@PZ0$tt|9^a_}i{&WjZri03YEY6b!C2VAdg#%uqAq
zNW`ES$yi}*r<G-ShB~RKrtrh?r8{L+38R{Z*A0b^IcaVaUSs|uy8!4$K3br>Or$wK
zTaSB5&UWOTieXlWtpB5D-8Dq0njBp}!CEEgKOHaVVVvjgH?!fSxE2#;v1}^7AA1LL
z2SoLbfQXxk>pfGu;aOcX`YsV>HIB*&j~awp$-pe2AASf5x=!+rxPLBKg{`>f93&ug
z*MXncfWHo!p=3Z7q%{}RVQ59KXI|%#xAgFnh(Sd($#^<qEtwoNJ=b9J1!|>gwMd^^
z66u4N9gdUhkjH>|r3m|G%CJKzq2m;1(PzHVBY4VJH`b4g_oW>?Q2I=Kv@ebenVvm?
zGV&3Pxs!6&NNBCyli8%6XTr#53f_(@L_a9sBAf(P>^lffkHn?Qg3l>dUBQrxkP9_x
z2QfJ3&+J7V5*)CJ07!ECML`811h!+H!CCs>^r>>&{o0bcpphlV<O0wCgi-^NcJMZ^
zLdnuWB=k{t!SnTm<g#rb4l5|y=YF|u_3Hfau+1$*@h1Vx!C>0JtTc8L{88D*#MO<#
zfrgl$Wi}fGDVYj(^LU`?ZXo<dLh>68Yl<6F5+O4Jg)z$%|DShL?>Ij9{+7&;+kRLh
z4W56sb-}`Fq_WH*o>KQ?NL^_dR%l?&iq4yQ3uJ^t4ULDD=LhTqB6?acA>U$853F_o
zY9-SUv~5=JAgLz+sRjNY#5YzFaJB3c@5Pn+whYJlyT1g`c61aXM5DnpM=s=)=yz0H
z4T609d9-N3$&1R0=T#&xf}v=8pjp?ou<D9)|9}OAsKlR_u2f<SYjFX`4p9fntXjZT
zadkpaCW`1B%WLZrlz*`;`ibONirMFViCPB;00zzxp;I#PRhLO4l-qr!M6Tsxx8V~1
z{vWB9(CE>W9}E3N5#&^v_W8f<(SGDc&R<($$_8Z)k|ba|amCEs$&EvMsNqz>2iIyI
zJ-7dG{Rz^R4#H%F4&8md(wn&-e1s-54gj7T?qhraZcq4(n*i{yu^abqKN!gQmv55?
z#vD=tXZ;r0?^_CLC=GMZs4D&WJs9^kdyP>hyl(=MiU*^vQG(j1q2ab>iPJR$YYLBF
z%!iA<%aMGoGkS)@gP~Vd9AQ?PHax}l^;)ht57tFX532>%nvXbJgaOULuTm)Ld8<_m
z7zR&fWgbi-oo(K_ecKLle=PSs!d)t@eK2wLp&fTiMc!YJLelk#eayBRt==k&9FNrD
zxw&46E6U5fVD{E4@gD|QWeX%a9+GVdMcxxF?_A3Sl=Qr&PyFfELhU0(GQx6TPlGGT
zsJAG7<m6}IB0ztzHL_?v9N=pF>hPT?wnQ`J%6pbf_z3=RK(-+%_w{#hb9<@``lc-F
zMDL|RGE7b*nh6F$ZdnEMQaG0!<hyr(4$JVan$O_1qF19=bBhd2{}G`lli^L?5*VFc
zDNp0Z%KrRw3n$^Ft;@lotnuCX>h%Pnc=*Ih59EETy!(EA$!@AZk%7?Y1MPb-2{+b%
z2!jadL`IGmRur1=O^Ocsq5KE7j|WYr677kTk?!TS#KR-B0lx_<7x4JD%mOa0g2<hp
z#00{8GzIiB6NhHt)ITuX5~_Rs|Nfsk60gCT1oG9SLK2Gc7$kN&stoRdEwEYG0PM0u
z4ioAgBFDCFV?Fe6SHOIgtvztp$b;9HkkuIvqXBh|u*gHqAV9L%&#Ljm#Z5)e@CuEZ
zxFTyaj+gW!|99ml(-nIV#A#bn9^?55rY}PoxZUb8&N8d-f@<4-7*-2c=L)3pc9{EM
z-@OnYT`IS|?^l>bd#5LFBmtcuYM%?XUD54RP!&C;nf~2Iz&fY)qP5twcHCurnd@qL
zHnwkqd&Hm;SBUNx&FkI>_c@w#CmhjuK#0<gJ&d`{cuU8TjBi);U@WHE+Rn0Es4#pz
zc}$|_jwmSTGLAx$;A_6>QF#0oc^#1dZ4_$S_SvNu2nlnonSK9pPwb6z2+EDF%=iYp
zQ!GngERFwd7UtI4iR=4DJ;GBSp?)W8b47aIvvEwAO1^F}bN4g6@F)voKa>P43r>;M
zrv-9ns$F%S*`!;#q!~8wavjce_LTf0*Nx$#DbeL1cln!;Yn{fu1Syoyl%+f|@+6T3
zE5PhW@oU4EdB-4J6E)oFPTQgFb{|3UvI<FlY9Wme943=<9mA})#UxbB=b0`|8>xWM
zwc7{0#kk9bSm0nlz=j6>BnXg2(7it}BgW^_;G1(p1*#X_L<ZG$_@pga$%L`AJtPHU
zMBT@E5Lk29)AaIUO9Dohwvsk~<)uVMAOfDUXn{ckNwqa9y;uq8yD?edpXrMXZI`qY
z4U&?D)Ht~v?<pm^KtzK^%$Wu<NBypdNgWG!n#Nw?vmTisP`2KfHmJjkV5IvMA5R>G
z0%^u+V`Wu|mzqxm=QQ^~BGW)QrUgBsIbsk{u{+<A_PK7`=}Av!N98xYbdJJn1BWL_
z*Kr|f12L-JN0u`LpV=l^RG^!~Egkt%F4V6{Mlo@#8(+@R!ZP}wc;=zLODW!n=!~Ke
zpCijsn>s=V$fj9P(dImWW;Ho$kDvPsJ$xX`y4z8UE1qb;hx4__gjGrAEf(pYA$S#h
z!%C%XUxgMZe{YqDMP#Yi+Elw``(wCJo#3luD(rDf`sylgxW(|_QBGzYbE$NJ?@Ub$
zKVW*nHv{Q8gg#){;!h8N<d~I+*_}O5GU@C9fS74wRX4Fyi(8a)=prAPB7|e*13YkW
zsMsjNwrgSD*rsj2H*NNrgAzD|AoWhNAo(*MAgG)JU0{Z>JEdrJv|YTq>Z*g0%bibA
z8i}U5paBa3UYioht|?l!d$N+$yt<~wrZ(|@8a)c?@f}0B22>H@sWYi1;>*ci0d#mp
z8g=cciR58+o{QX8KanM_bOF|e4AubxVD1eo(WN$PZBMBhIV<alMLoFm(VX~Z>y1_Y
z1V3>y2QON4T8c8DEom=jde`sr`}(gc@w)?T_&^wAkeu3tW2bAz;v2_L_MB6_k?8%!
z5?Edi%p&Og(KDtE-Z2g*NXacBrP20Y%uW;WoxW^2NyVclg?twQsVmm0iK+cM5k2PJ
zcr&TBs`&OBhKd_my#aRwQZnw@Iplw(x?7=iU;W^?PM5IS%z18>by?Oos4DGJhA6Yi
zHnI<Q55+o~QQ#i`^93`RKC3j2uw_Q$93KNf;ey{sU}%&x!+Tl$DP7=^x{wDLPXIKp
zC%$n<_N(^}Y(Js_0rAfwP~E8ltZ+xv?{oMSQ7~<884sPr*Nw@{2(D?p8$$O_uU!2a
zZZ{vZt)l-uyQ54G%ec(#N0z^#U#OqW`2q$DWI@FtNcsx!HJp&9-p93$Aq6$po`+wV
z$uwyl2~*lFNZo`yunNd9vFheUQ9H<@g%tjTM@2V;hZ{ykn1Hhto4nIhv2rvPG3$fs
z*=-EcBa+}R6!7b9E~gPqCODNEjpw$|NaJC`SiIRcUC6NhsdgjSpZY7vAzVq%<u}1a
zKN@!SLckj)p`PIZ)Y#K_E6M0sK~+75{q43|y=HdfH!sm$01W_6Zwhn)R`;j=z@(I#
zQ9Gb$19Ido^PoR41h@<a^qIa+dIPF(o(bvo*FfPi94g>uhvwdy&mRYNezSjIJ6*)L
zfqAP$n)U>?2g(1Lek5<1cgbJyaGUzDl*!bkQ%=nWxv+<4>DJ`fUK6nzsa|{0P|dB|
z+rv$ahSU_ww7%Ep?n0RzLKj#THVd+!KF0GN$6s?bif=_8h-im*PW>ZFjO?__Sdn=3
z7fROYAW~!2Bn9qTQ<Kt`w8b$O{@$_zHUW<sH0}?-&-a_UDP&g0?(=OXlb@P9g@jI|
zhL?<89k=&yQP*fsQ*$Yg>aD#U$PRA`ZX$V8(2ze#@M>I39aF5VaPW*`?hS=HYODT6
z|5E_328$_<g^W}zd}z_ysC1?3wuisR8RwU?VaWARy`ZvOFaUCeK%JfV(U(^65%)u5
zg;}A#d5(qNh3&*|+EzJ&KK}+oE|HWG%9JHr#%Jpr24C6~HM{jt{o8G3X+QHnzv?|B
z?y*pn9?c*?vD@(<!H-JQIlS}_AT&{eA<--hJ7o7Rz})z2dF0xi8@`%u>ZWQiSEprk
z8Icqh{mDb1e8ASFWFRH)%(brA#Xu6;Aquufqbx&BN_EH~LX+)Iwv$=narb?@Codgl
z>c-+O$~WuZyKD1mHu0eewi{RH?927tJ48dg>R5VEbpH{o=m7?r?D13^?1)f1W@{Nz
zLR;1$AidG=#sq|FNWYfSy})<-%$Q*Bz6!#qsHkYLTE86H+Nr!e@{*L#2MaR|nA~JM
zB$*QJ=}3ZGR!9ER5fs3k=__g+g0HV(W^&^M0+4#-{%23#tx!E{@^i=a?`8~7J^YgQ
zVFv-238=D`)pamYz`MoXwc+KyVa;|*zOeG<i;4LI$gv;%41s}JQ@TKdVOc3T@2U69
z9WSqfbpea@eXC%Wg$4#bU2RayRP6R$J2|(yBgzHT(=R7Ka?sDBx;N@g#q2PKkmu9}
zpG_I0w8p11^xICzyH|wim`@b|-}cd%@e`UbT{Y!?s@g;D3@AefzNCYAgCtp-MyWdh
zHI4%2M7e}tYL<<jya$hkOEMPM5(S-<9d_qE>hda#bp>Sr_U&b*Tq5IB)$PqtVH-2o
zPG?RsB0_+&>!8``Sb5B@dergfb~_Ar30X*}HBqAAfK}EQU-$W(UYT;c{I)(+hAZxk
zGmX7v%3;X{D_~*D{r@oco>5J9ZQCG-h|;3cK?n+n8k$N`T2Q2khz*b;Q4uK-5T!~;
zDAJKCpr8a5L<Ed<q$G4iL^`1pdWVD(0x9=wpSP@M-dW$Q`7!I8`C(azA-S%7?R}od
zd6Y$Nte15ISMU8oU`aW{tc8;@(8rlsEF~A<H2k7U_jQGOlk2)a%+N8zzcGltZ6unc
z&H49{s&ooR2>LC#={`DM;d1JmMTGRLI;*9ALH6G?NW-Ve&PFD*5d_nUh=1=<NS#<E
z`fWZwSUEa4EiVM-`9n~F9NG`Mp<ibc-_ytCN#xdj^DyyKJi)Y3L^meV(b<dds)58?
z#q9fBsN=ZVktPKMqHpMj;`7t+ouMp<5i^w~l-q)d8y6&J$F}p0b@{8anVvzd5H25p
z?Og(uT?D8-xfjY`7^Wa4Q{OusZRD)c=Z@yuM|cB0NNcnrE7f{{#K~fpy%anuKQd12
znaWdi$|8n52o=FB`1BFh>6lk+7{k3k@lI2GCpWM4DfTS2tep(;dn9IMC@ReYvxd$Q
z$mB5uC}*AFLVm78ygsH+iR#^nuM)M2ETrnVofzs|%@cc)Wy*Bc#I(KJG+Y9SIPBdz
zphn;IRy636vAfn(#F5nr$R>aL*Hs$UpWI3PN_MG8`&B;Ro@`qEgyerqrE>{HlczW5
zFWD$=mns*Vf|kyN>)|gwcp>H-FyHhR?g0e+TNK1EXYYgWbUowK37_G7*W9%)=3Ko#
zEth@G{M!w{-*bgK^NG(l5+rIoKig}Wcq8*i4P+`T@6VsCG9N-v-wc4W3=<VI#XQ<F
za@)58X|k#mb&yiDf8!!oH1h~pynx$YueFZy&9h@%rFlBdFZCqteS4F8z&`ue5A7oP
z9P|yph=!`pm+RuaHa$4i{IG|gwSJ1rBraC+c5$s#G+=28@{Leo9;CV8B9SH8_h>TP
zxY1U*Kjx8DZxMJ&Q@WX-#tm{ue6$zl$l_k9ltJ_R+F#Wq(kd*I52L>QA=~fY=8p5a
zj_=k5K<~hYX8esas~|%fD*aIqm;6zfyERU%%|4KW-F}mV4c8T>dAO7Pob=1OKPGTG
zmft$Zb!`(7MqtB$nnM_UXkxhe`By^JmY~d^qCWmzS&kmdHX*tM{S2#LN|gCkdtVjE
ze~Jh^G!&5Yb2n+McJGnXhgMbg1+1Vd3@3onq41|T$QbzgI?3MD+~zeBV*K(eA25Xq
zA=P`59K)&3qnRtR_c1A$Ti&}+Q2GPfQpi)+a@`X(v@mk38FE#VTsoYZCZf#yevHd<
z^l!~L(>XX#Z6abGA%vG+WWLQAURJlshIx1+y{=B`yN_l+PiJb;ZiCl)Ml+dyTv0;&
z1laH&N&ZQ8h93u^J3EgF0%~rMn`GvQ&61CZjrXtZ@Fn3WFeYQ`Kf}gKKp6y*GpIFW
z&iOD0K#iPwe}ax|6H00LS(o=yTk4?NPd^RTp<c2;J*BC!&POw41{ITE{%j)Qi4{pW
zVruMp<IyP9K+sEFefqrW&+ODZYo9@&CtVv;e-qegNa8EXi1VN+ZfuNc92!kHxff+D
zpTmY8ry12%aRqr%i^$XzkL=;IoF6Of_8IM~$nStVO~hx4Ge7Ab8>Ce<-VP6WP#*9Y
z{?|5M=)uT`Jz~!PcE{!ac5}X?w!X{01pilzs4MdZkubyc1=KF3fX<i&nKSzLAm!@y
z!3oxN{d)K|6UurC3#Q?K)<dQVvrU)Z{6?p0fYLC5Y|g}h9sC2Jip2j-^Ew57!8SoC
zVlY6ooErXNqaU=#ER^DUzcQM^G9h7pz9kb_4S?;h2-H@F1T=<}mXLuNY^`@SJQVPm
zG3S5?>SqF1?ep-NKPaK?%D<`-J?L#qtLRdIH1jRJcpxU?v(68p>kH)@uwj%jM^`Y#
zEE&JA?GotDuVec0JC7eQVFQE*6Z>f9Q@*QwlM~#2_=#0bpF;VBNEKa_z{H|`K3Hi$
zqyr4bt}(olo5bC|gf`+&b~MRiI_tzTwq18Q8dxX$YQ&}#;dVVrnVq+bUYqTwY#T~0
zaO`-2O+y(CPK&Ms=5#RWPB*7{uh$Xw*a0RK2)YmfJK_q!TsSY$+$dTNKI-Et`1l5U
zx!|L)l&bsXYo4!ut@!^rhc5d;zly0h#tWm)kac#+)M~dW!eP71-9sab#fKdD4bQFQ
z&T?F2oZ6>_p`9eG?D#4aHRiS|Bm$z-<Jc74mk-Of;Ixw!^^yWMz1yaWVgqX}lS&Wn
zFFjqTf4#LWM0=2jx=pdnoeCHW+#cEL`<Cl6ld2VASEL(6oWKHsfjDr0Cs3RVKx+yK
z@$!pt<^ENC;Nf}kXIiAqbT+1UUm>I8<Pzt+7k?<_7Hl_-);(#>5TMvLYs&pv$?VxP
zaBMl7T)ju!T0}%*$(Hd$tocb|s}ZF(r!+p>FIyt&n-JVraGK|PW>Xl*SZu&NSU)z5
z(qVqC5z8MRRV;uy)8TK%&ddMgnli%}vW2t78DL=L6}JY*^q+IQ$qgGY9($3`kzh0h
z+peYptU2U6%?8e~2z>WHSE!43yfpO?&hZdSQL@mJn2~)|$P#4VLn4wu<Va_fjeFrY
zKbd;<9q!@OBuD&@mr@pv1_#ZV8^{F@;Nfqn9457&%$%ax*kX_RMK+lg!1xv09bT=8
zPGemet{@rO{2r}idey2Kf1*-DJcW_ddGW|MlRHw47qa+u|DWjt5yPn>l{EJ0yIt-M
z_XnYoxVeU9magM}2mx7Gyqq7(FSkyU_4>MQuhSXGw;rQq#Bblot?Ksz_I(OUQ_k*g
z^L{giK#-6H0rN{z+;G|uIFs}9wLwPJxnr2BSR}Op%z3~u7ervDv(7n>pD&47JzUo!
z@dZKwj~#4!^z;SXh%zUp-T;4qHlnZjS{D|7&BCl@;x^BIf?f_F<StK0DPu~|JbqKl
zTB!F@OV^ku91y?PioB_-HXG9yHy`_U_9Y`e6Dlegg3HDf;!9-~6s3ZUQpCM(%yZp5
z+0mVFDei2N*j*{}GBg5E^Ppr?!Y<$~@iQ@BaO1BGJG>=z=lWSQlv1}X8KJ+i(fg~x
zrhy}`6ZfK-WU${jP{VpK-_e?W^y?#~Hp{-A>??9ltS-@IXjYAAEoN4Y<Rnh$vQv*5
za$#l0UbCq6#UAkL9HVPO0ct0_TH~KH;;#vfY#G`YJGJ1L?Xdk5_Y&;nn|l3<-io@D
ztI9wp-l*-w8_N#_qx4^Yl{LXnnqD2y)6w|*sDek9zvi!nkDeHOJ?Aen9K=i}kiLxX
zqpXK|t;)1kASdGmtQ(IV?ErwH(L451-HjLT9x;)<x>}feL){}Nd2-;`+S@5DbHSx`
zF<Vm|C5tzzZ>IgZW~H{+qeK`24%C3mr2ywG+VGS>x$(>HHT~F)vo@N*hFO{9_<p-=
z3?I$%uE|$@@~!4vf|18fM5Mh6t$qq8>bGn`F>5Ic6#j&H>f{qWNjyfrpu^T2_!eBm
zQObG?A!b5C!7IaD-1VlPl^Ftl=V89|o%+F=_$e7lXAWw}pMC|mhZT_zFT6g%<V350
zo`o<|?Tpw4ZD<FKno1BRkctr-_ez!bHZXIq*~#jlgFS8v-&?@~#oVF?HDH>A4aKZL
zq1(8*WN@7la<EJvbLK&-tqM9oqT2)j&JsG!DrH~r-c1ABT@^qt1Ig2@hGfhFSTrgB
z*Dr#phBti|y!{ri1}>8cOkeC@uYVdBf`BDo^Vn5Y<^czhKhW?U*Mr`%!+c3*_A>H!
z(X@md-Q#|5RO(s1KmJ2wC^)h72(T>vj%pq8+Aga(jQbSFlGB%~UbgOkQKZVgLLgrw
z9$^h*A&1r#>lmvuLQD?grZruWVP6hN4kHct3pEwU{Ci41{tvI`SaZJ%n9Sc^gVSK}
z`d7A=-CMPSNUOA}*u2A=1vRDdo~^%qrGR1sY4X1kBJyaJ5qdxTF#n-ydV+bd6I?L@
z6`<(J(SOqn9*hwFsvif@^%EQ3SO{iC9s&;2z0<gPTPmoWk_%`6KX`BfvvnNNeIUn~
zxaZwp{PTg!N57_+{zLm81)kS~ByEhVn0AV@6Q2j$!zuEUo%Oi*^iMQQQ@J=aQtflp
zTcLjYDK$xJ@EFRnh9DG)NkYrIfS*;9)QNtl@`@^-65qF<1oZB%K4o8ciaE3S4_h!?
zvJ9kh3YD6RS&x}}%e`)&*Ox6oEktp4&Fp}#@D5rQQ!k+_u^gk*8RlJUZdjX~DAW{9
zUEX4U2GU+*uw>(Z*kTBfC$rUEk4h*74svc=WsYecm8j>4&;PJ-VLWi57~vW<<d=96
z7nW4@BZOxA-AYp@le3%bKYoa<+v2sDWoRB?*6FS$F(I|EvwX1K*2ULz^m<?j>>0>@
z9UzuU8a0ZS{Ym>eZ2I7f?Zr5xf4G>a_kmgj6;8_tAb+BU0@*UsU#{RX#i8=fL*YC3
zt&sfFvKG+cjH&JTR`WP}IZ9^6BUL)zWs8jmBNek2+@kTfV_9kybV1bFAG9L6iJyjH
z+#lNv^RlRK*;nL_8{G1}xN?pSjMf{55t`$t;rL*}S;QL<;n6RDiT1GWsgm+}u!HBI
zvGveoUeCD5518YK4B&oJc<~$H`J7leeqwy-zOAIj6G8S}c0`vlLqO@0{E97^2APbT
zG-rZ`=pQy47CL`hh6({ES^`RyCQlY`uh&((ceQ87<echJYn$EnbLWM%*F-M@<aj$I
zn!G*~LoFr~R&RcGt%?s@T`OEwaNPJ=6Yr4u^pKd^n(Dt$s{hml{QWz72g8^gtE&rO
zFzt()u-3f64wub1cCRxKk{L!7D_8^ZB#(e-shYE1aY>6-J;N+Au&tGbwhA%f#JMzv
zkO%cFw#CMkVMX>yCu%mQAKXUT-fi=L0QK&-T33-xZUqSH8vr(dfCQfPV7PFN*0k0U
zS?o}SiNe5w=X?vlHKYi)&^pQ091mf*lKE4n2>b~B0Ea;CYJW4V(lx%Hf%QKbu-#b4
zQ_Xh|gT(OR+6?M*W}&V$`ghgHJwK~En3cu1W&>$wf1eNTIu=?1T<P~NeIzLSwKe4Y
z9=&0J5T=PMUVI+1=e4dBNcOa00p@W?pC9^Hiy;~JQm*eCSl?;~oNrTPKy=2<0)*{i
z)xG}aQOZdDv1^^Ov21J{$sHJVc<Df%OCbtFncWU+xMw};TB3d1Xy=~E60Dl}3FD3f
z-9>3w(|6as{6!mfx9$b1awNEYwLu*;Vkd9&n^f84zI3)V^nOg+6S4?1;t65vh1FI+
z*B<ck7!zU)uzsT<0}Mw>Z@v27$u#f0c8~mFyKCj$A{*1zuS?1`+1Td@IYKe&EL)7P
zODhPD<oO$13*9!4hL)-qPM-BV<7_YddZ^hQN6&l2ankO^WKsIAavcjopnxZ@G@gGk
zrfTQbkGSD;M%>mRYr2gWaR_F0jfO@T0jV4%oD^}&Tiq|)bEog?ro9U}qIH9zOiP(u
z7ih}7--|nud8Kt+ukO1P?LZLxKK>9WX-}ZV{7$x;*HTbB_VPqwHTpIFcIr7MHP3E(
z3Ax<K!Qi6>lG`7UD_(lN-`~1p^JnNrJc@XPoKO8jGo#$Dr`oJm#FoBSoQu^i*9@C^
z+Igt;`0IBpwMj&|2178v4$i}X<}KJC>Yt>*W0a+xv#{ogj^!0UW&9IOB!Xq4*%&Wq
zyI&-4Ph;ZUArxd<mP6O|={>7_OueQRyj;E!iv*&&*Tg@=<++}Pu4(CEBfDu<&$TxQ
zJMZ6=K!2z2AO?juSHO2R{NQoZ&;MG{^$GBoZX*)lWShJ9(9*P9w+J$NdT;|Aaftym
z0wIHbXQmJkKS)!8Vb7Jc^Ov6L(L*1u$}gw0G2Y-7;;CGHgffgHhSSc;ckSNut%tDP
zQCzPFO&$)s%NwEGXei*`N0yj7xj89$RWCdoueOfS{P|3sln8hr&v3usG{w(FvrC!m
zy&9N;vLv@3TM5y7J5yW?!RMripc>6REJJp&fRI+Db?^L>o}(}Jad>v1&!PtirE6vM
zJE$8OO~daV_Ox)SxD1?mra1Y>6Z^b_&F?*W15<wn=%V_z2zxXSCMS~YPVk45QZO5?
zFAz+W1}mhLohtA%kx2zS7r&|R+P2^P9uti+l@E{KReFCWB~`29R;<p!b%YeS=N2_*
zr#)!5%kj!&pQ1n~W}<vrfQwTL%~md3{=1#kR2W!k-W|8BlGd&8mA<pFX-Y$k4mR{M
zLl0cP=ZyWdRVS&E&x^>9jW6)7_I`fZrt);Vybe?m1t^2^zZ4{ykvzm~c6pM%yMkv^
z>a9hKE0!Su?3yLBP+H^yD8KHBv34YqXSG*NeK$#Mdght=FU@zvg%x<2e8ZtabQj?$
zLwVNWqGq-X^k!8MoCB?e{t3!9xRZZurXqQ@x%~AXhlo2jc#kG7Jjed2aV6Y^lRhIV
zL>l9dp_Hp=-jlFOhEd+wPhR5pFV50>mpEnj&m^5iByX|PpMydql9m%VZ4Nq@OMS|6
zwUu3K)|Mu3N)+5%UFW;ln*#?oWif&MnDGm^NH{48!Q+k+plf{lfpJoNMXm6bc{6{x
z;GvbS?R~2Bj1?$l>g`qGQhvVrT?~)O_s>`(hR_8Ks?W=?*!$m8A{UeNvLLnEm;%4k
zr56hr_q=oPX(~+#h;zj%J2ty%QyL5~kt89S%fvG?95f<?7q4`id^l$$aX^eOBv*Ql
z!0VxHmcPW?vV5nuQgNqdB;pC{G2t$Oge8z~LAhH=I?8p>qiO=#OX)8n71gvt=@Bn^
z&r~OGByh5UiV8$}o#>2Q6-QzZ_~BdThS*_9G;Utw`*F{7n$xt?$_Ryr8rb)#jDm}B
zFMX3MYkSX_EA0Z;{yeT-;ud&g;C38b#dl5oy&aeU`Yq^e72!Ob^kt&lH;8q9K@$_q
z5C`5UM{P#^i8h(_v!4$hcesC@<PR2$yG-emg6#a$KV`lEj1XM(e%c3L3XFOWXhXwB
z(};i(9*2y2Q&Q9>BTqO$VCxn)w76mYhQvNf+!6raK1+L`0iuJQ8lnPR6ytL9^r_YP
z{t5Y?gr)aEQ0%*1T+iwEK{@*GNw+D?<LE)ehC5mfD3!(FTLOrVRW>vzQtO4`GyW70
zrY9?eU|AD56<H%fs`apY8&dcxyCinU@v<I+f_P0J8^rHQp=|$Qv%_TJdNIRXjpzIb
zq|TVqpYK{_)4l#21h^uP1E43(!>2WxUm>RvO>R!jRkKaWV@m?1(McB{92IC#&JtRH
z7p<>U<e~c!f*;7UwOzelKW}sfyr_Hsr6xs`b>e(9IZwtp`;qsVrKfjd4+p}QCTQ~J
ztCw=n9Rx8DP0fp!?3YpB5itDW{EqyJwa^Bx=0}tqc=R4Jj?nCd)>LJt+FIqdyOR0F
zvp7Y+MF5kW)(Bm3Rx1e5a4S`mYBvgZ<M$KRx*Ni8B|23rt~~5JsXBbs)nQ6Le4ry4
z^Je(9{xPTh?#ru)bSU}eI_*?=1jV*7iJ?LBm8}p*nu147(IU;YJvRP~uJI%3(CXnJ
zxEgF5&N`U>X_ckz4hRjG0I9kVeb;I+mvIDoTX7oD4^sD&tL)CRZ~vv+dEw@7Q<p)N
zn@PbkQP?@;9_)X7#hD|Ond;B{sJ#!qbQ&}pi|VZ`Z8xCTO9%Z3q9bhu_m@adg3`H-
zn8-b|BJlGhFne=^pEvez|1GaMK_i$P--}rh12suHvl7>Yp($uE-k~!<7FH({Xipkh
zl0Z}%3O)_)3t{UlqF1C|c~@2!RAArl5m-U6-S@y{J60UQ3^V{hRy6Ak{t%|^I2pqX
zSp~fayPjm(f3d+k2@Jtx*7eqA=$a8@82|Y1n^^{Ks)XNGy>RmXDqj{DCv!DocjfV-
ze}5v56>YPS|DFc+AK9HoDD|&XZnx<@q5qbE{3UZZ6J|d9Ay~Bp>P<vO${qmNI6s2^
z>MihLp#oNx0oom$ZBcaT5oQ|In{4OTx+jkPQ=3$iM!f~D28Qg!^vm7A*<|q&1YJ0q
zq>lBK_OR0}s+&8n8^U&zXdkjWvKYpc>sW0w-qgLE-;Sd9NmmD%u)Sbh6VP13pewy!
z7&^3C`6fyARw*H(1&1p4)q2e!3IEj}@lvJilbG-3O&RcTXc;ubTj%=XldZQ(O&Yh)
z))}06YCav!ihS@^>3H1TLG5R!93OOroa4j5Jb!i@e|wpq_wzyD5yTg0u?9n^s<%N`
zsGLk}{;90g({fe_U*sT?p&G5pbuGAV8XTfdxL6DbM-nJN9ig9Hve71TS7b_KzBeC)
zpWm0A9Dn^xZVcEm3;$u`sDbCGmJ2M47j7krgdAK6x;04Ba^LlDMw_Giv1LxtaMw;L
zzJPlc#YGjjh2k=J1SD(x<)!8`KL=?W5-h=sM6vc&6-Bh7l|fH=1s|t%a0J>cu_m*P
zYoFkv8Cg)~qe@q~u9PIvZxi0$4;Ssryc?Vx7vz+T=!UOEkcCI6&uL;=wCfj|eCeD6
zA4+E=axFGh0)d~@ERg4|hd-{R2~DoL%8r*;l|F4Nee|{XVrQ=!tqsxRv3isISUDcb
zQDZ)7u4unE)j!;vK~viI{?#6FGkyxkyClU&5&rbiHFd)larY0|^jPGbSrf7P6;C~A
z5ty7KB<{LG*~=z)GjiN4TP;!-M&{=ZSG8~iYH%Pvcr!@oU+^+WB*xJ^ilb>**2LfK
zX~;OM!SGDQ;dw0+Nfm$G0D&7*Z-YO8+`F5(wrtg_G##tjW%fu#;#im#-W^QzJ)!|w
zOLhGvnr%Sa>TAf8(uHf^zUnD190*882>AhKpCY-U>8PwtZiM%H&OOyMh=e?GR&8T_
z6|_5?CrX)FjEg{7rV;d@s+62vAVpnTgDcs7Jb2t^X7ZBQGO*7CD>;p9MEy#0qhC@b
z)6O@kELoK=S0N4f@Mo9&n$h<$^^a@Lxan#-<E01nYI;AFkHuAMK$j7=nEI1w4l{B!
z=^9JEuVALMzvkwt_a7@S7aJXOO!3p<;7IQQo8@QppZRH#I$z?}vQNf6a-&+`*`^;Z
ziRJmqq;t+3<ZlEwJQt>37r&ozon}56f8-xF2kqkVyG^~C#rNAvJyw^~877O(Xiy8)
zYhg$qEus6oyl8>b6Blcg=R}I}VZOB`XJzxm=-F8NQygmun)f^agW1~t6AZqy4kx8!
zOV+TSf+Cu(BOS`4CXNP*TR(BvYM4KC4-L`0x1UdsoiR=^4NTX1>C!2=&#4epIb;8@
zl^e9KYEkTA%mb!{T8)}l+}u7-4xg60tS{ot75eTk-Wd&G4)dTMUZq%D-K68Jlr64$
zLI_J7tTi~z2nBmGQ00{FxG@;Id6?_LDxx#D0sd-2xPW1j1BAkq$Ax^evkkTlU*;b2
zsE9Q>1;1knG5JuZyTPH!&(I*V-cA)-lo5{D4$Jxuhn_m%;@N)RQ%}-#v>7^{BEp>8
zS2Hr@lSa;;(ny}1kTCeL?4fQ6(^@S)Mt>pB0F6=z)|z1ebdSRu7f}Xz5bk9Y>MJ!%
zO9M7GNk4#OrqenbS=vOD)|4dNy1Z|#WA^p3>XY80nZurK3jn`zm@;A<<(Y)W*aXz_
zuiqwbPeYPU$(onM)A&b;ZCF7933y@Doc8bzoP3YGPEmy=3%x$G{4@}M{{V6gF$Acn
zCgkdAk(W=*BfYB;2mO*nEj$X#E?SCe|9(DO4Si!(gal<xMg0W7%ci{@!+^Y!FYd8X
zxP34?F`_tp1}MOA6pl$k6v!h@a^q-D*ZC7saQPfHO^=I1)+uzwl@MHWFOUQkb&)#W
z`#5QwZuZ0UO?K4x`q%TJh|8FIeT?%uKdWa#@&I8nz~7<gNkLM*{CjonX6vXYFP=C1
z$v1j%jxnxK<{~EBPsvtR_JLT=Lr0S)&a<bpvC;s@9b97pJZkpZ+veIZpKa}`GjlTK
z9y3k81+S6fwdP)7#_z*8>R1wBDutOG56wz=p#H*6*h%tE!PN-L6oH%I@r5OeQUO*z
zaLBT*Oy5k8yF9Y+{Lb&-)u`jw59si*q5m+<$R70|NlYVVazez$*<QNu-oa>zpyI0l
zusniLL;-{+<RU|vB7*gvthKt+wT@kD*nru_*C`pGjZw<9S{$&6j{pR*U(~6oWwWX{
z%o&q6Q5JWPMr1~a;>OH_3FI4{p@1-2ql^snoC;9pSwh>y<SRaWR57A-fE1Q7XSb!w
zzS~N1U>)M3AnMAWkJEfMT3c?$zFAe>_I?+yU(KOLQI2NF75GWmS9N=@>bfagl)07I
zjPhzi^%32OSoo9pP<x&pcjYyf^nx#vdrh4xQq*}qfRnEc2SD%<(`lbXx<0xKe*!q2
zxKh*a4Mv~3WiY%VqgNfdeW8QR^RI)3j4SBol>Ho<(%t$}ONQ-F+MKLE)9S)Ca)6Mt
zsy@AH6Iig@`SvJ2@EJR9t^#mioO~(Ir=7x38ZJo1&S8(kZL@h=7U+lOSyO}@Q&HV^
z2M_+Uvn_eMY^V0UIQ8*AjM@J;{RwBsfU11s1s-g=&w2rG@e?ks)O?rkE{52I0l@B;
z|J#4gf-F`8K4P*t9{^%sV}4Z)&)`Az-=<2QVEP*|@(Z)}MW4#rhxY(0Bht3EbtQmu
zJmHa1?)i$|6Ziz&X+YtCJg3N0y=lRd8G0Sh^IUG}i{4Dmgn(LkuAE>-AVq3Kumo^X
z1KAX222vVI?GMIv@LlfMqAzDy>oSg(nr+{@#Ijf=6iuw^<gwb}!frK2?&8zNmHYUv
zw_#>$>_~x^fN$<XPmMBlcp&W(GhN%|=^q(B=k7=E0!fCj_EH4x46!cI_L7^)w$82J
zS-Y#BOZM4KzvCMv)<G+948iM|;m1vW=Nr^}UB<aVzi0EH967LGXtXv+CwN|!ojew}
z2)YgzP)2>|hp%wJS>Y+<M~S7^e=b?PCgmNArg#4TJs$Pj_AWTY_Jf&OQWQm=ReuAs
zu*CGu;&+;I3UvcWI+s0pd*+a*5cA#luNQvu-w1;TH)B)@g4$gEWCB12QSumUx${nv
zjcMYT8Rmu%4OFO{!8BQv2qpeHr7R4WlCw-+&9n<#{(iS&E=!zk22Of2QR;=@Kp*$}
z%x5`!zsp@UAZFg`0z~_ab`foc_zNq@c%a+_iDb!9juors-tLpHy48BY>RDYNn_7GJ
zE2*nRw+#7v<8)40xQJO?dJOUg0p5-VC$2P_8<;@2P>Oe2tf=_M8Iv>(+m5(2+t;|6
zY~rc~>Mi)7?KCAUXvrSy2+tYEYDPp-zTPvOp0kv3<|5nd{rT_5PZu-@A)vf%%aLbB
zxWSW&ldI@SO#S~!-5CT2g4by1#&vD5#2=gmWd9_BY5wjWGZEuqewg*MioYnwnI+b{
zJy-^dFVz~qPI;|$ndBzbzXf6FPP8Jnu5ZB^yr9#J0&&$>pCl=O@XaHFk4lcQ!6<k-
zi|00le+DXFuL&^did)dRvKg#6u+Tz?u$2AK^ZKTPjEkSg3?Kf8uA9x#?{k-t0V6UU
zW+mhKXNEo)oI2#YC?yRsKdnqo({n!ks#t<GDqKR2Jk@+F0)6x4{jiafkHYU~)^R4f
zWn1wp5V{tE)rEkn+B&$*`vqK0@ldnA`?QFkO)A&tLKb8k`;uy<3AnC&wqrk%CKZm4
zb`aM%82P|zlskjcc!P$eU$J*KeJ;<C+1sa}J`npXPyJvbs0Cp=mIGsoiXTGGzKq(Z
z+{MeyLcvKNrMNQcFlAQJD|42JvU3xs-KQ?wMzfI#5iA}&?bWoIM5E^8Zp>G^!cw_C
zcMlt8B~1w2x(b>;dE(mlDxy~l#T|etmExa=jvu}M=nUyT2O<gzL4izx>6Sg^sI6Jc
z+MYGbIXQ<Q>j0~DC<uQ%<y*fV#o{R@)KJWFBh>bX8FYLrJX6=8r!w-)UyKW=`&dx2
zMue`^;&~8FDco)tEOR05&k<2Jh${<}`L3&Xk=Rm1JT92`WZ(N@&A@(W58*Z(X-+#6
zAEGTX8EABg=KU0<KZT5pMf!Z@D>ySM%zvm2NS@<Zdu~^$0o8BCt>>Q0R3A%f0_6e$
zT}K02vTU-?bnslo25upakTEg1X~Ak`enJW9`mEvO;Pk~H+#+|7_(Jh%s~f)>IqyCa
zXT2bhvM~EcNb#$RQ};9lJyw{Y2`wchLyvDp`P|$(N%-)Bu>=-w3qwHn6wM;rkM&E)
zdh1ZMy=Hv6@ywZg?0xX0q!WsW&5vmsKqaQtbjwaub-_c>K9;1dJp*9<eoXx(-SbA2
zd};#G+g2i3H;TJiD0=?5Rr|e(x3~=UGtc<>VCS_QqZ|N%cNFumGc<QPVinYo{Otnj
zneN*3Jhopgkvt*j96j8^uW_M@kP8W%#(k$ave+IHR6z3-3+LpE3!OqlS$T5VQdxgW
zm1;h`D*mloU4-}|$R0dRhpw8l(~{aH@V%c$8{_YFtFX!#ZM5Ypx)PMsvK*lys@9L=
zBe4$1Bhjpn@!C=bXkidA26wWg!6{SIP3hkGxCq8Iv%w6}n_kW7SD@j1Y)=8LeS(j6
zz6D+y_x_{b{MC{HuNb<ublmVR$MLt(FQuhGVt_JB6iU)y3b+9ni5;*c@gfwZ52T4B
zi&b+eS~g2IsKw_V=eCV*Cm^Ztzme4;<$fH^nL^)-XyFx7P>SlQB2(F$yQ!P^+2=fD
z?`LKQba8bd$aUbwjN~WrQSp0Jp<fICuobVb6yjWu4lc69$T=OOW<`^CZrq6*C8n}^
zv1E&QaNd!x=j5VvsxxdT_kJ5F&K-$AaV*wd;ZPe$-O9%ax>nePbu9KNB?Uy;I6ZnB
zZXG30$OA{5lL0ddeZaXYr~K1r8qbvw>R3~v887U|U9S_hlW&jPmF&Mc`i`3s0^YbW
z%>iIOhjnE}Y0-2<U&Ko<05$a4X=z(MQ-OUK_#3eY-VKIViKUD>N-+=L8i{gK3&_wA
zKHqrU(4=VNT{PP=%7k*h0S}4Am*55$mkyX`s7#ZEW`=WvSZplfkqL7u1Va0>H2P+?
zHd?y(@|Z3~61lrWxle&aY_IAY_phy|Qvdjd++CAHt=%$wYW`v;e$a{9P|bXemNg6X
z=S`0YFSEF-CVEL>AuG9-%Lh<Gw&X#Le}piJjT8>R@*>4QM+F@J)vOvl@UJ1`G&CMc
zzKV^W;M&SBL|hH<5L+x?=r#6gzE3gyB=<&L6JmaqsZWk)2*OCdv8rEiA1Pt@cf}_G
zT6EJf+<<z<w$R`FFIAgUk9KoXdEc@)B0^b~qFaqYDwZA@$x>hbaW&P6E7s{Vs3#hi
z=s<-J%&ZR3O_(3lAcquxM}c|U6XLDQ5;HC`Jf(@tp@@BU*LQBWo!{6v>ku4ghJKBy
z7sK;UjG^?1x+glFWGyW%J0Ea1MPL5;2&&#VA>rT5hQ?cM{yjX|K{fG{mJf~;A3z%t
zCVaGE`GDshzTpoQ0=e)Ld;7Y$-k~_af&K3xqPRJ85|ojq95F$K76FiG5z*@is#S0*
z+Tt}`bv0>Z*1@=`yI%3UxFqV=*7hWa%|i#*>eOMApo4Qxre!pVKTdq2SJ|iZu`ISi
z@6Y5V-66GYWrh$}-Hxsxv#{p&mkx<+o9V(8g9zSzUythGZFW5xUyS0#D8gy^bsy$?
z*FILiGO1ID6NLbt`zitO@WGy)N3$l+8}usG+<S<+Hw8IuL)BWvA?KzOVq#bK()Tt&
zOToubCbLFuX&#rpcR%28Hob6}_k}0o0=g4RzS$WDMutZo_<!?@>uuQEyOJM>%rpL;
z^%|1VF3V_Nw-Z@40axrg89-*d6N_k$8sPnD(R3|!bA&W<DZ!9QdYAB24_sPX*6?r_
z5qQo>UK7<w`(ny(L|1CaXSG}ayKo-uL2@v&46==rrwCX3PQ41aIiPlkmq^?>YS($3
zCGVrqx~aAREjjRk33BXd2i7lac;C%{18(fq<sAbw;1>K(r6$2Vt+RJEo5i>~@kVI(
z19*pEgp?>UiDkb0hS|CSB>77K?*NY?_`hwLG$U4(9cG~s5hMiG6}V<?U;*fm<C$KS
z;68=}r&Zj)peyW#v+o2|xD4c`J6kmaou1{yuBz5LoyP{O;w(tR3nr*<LckhsW2eM3
zW4Qo0WrT?B|C^?;|29}H9Me#`D1;Rm7X;V;FY+C&k;R=yzHS~(D2d&x&F0hjhb?SY
z^QFtt*)ijTMg%T3+O=O6kuv+AUcxPfQ~yX_{)4oQ9B@8)IbI?*97J>r(cs|IaWiAR
z;X~6?7HbtB7L!d$W;cIRiARoml{v(=tPUn8ogWSYLs!xpLRoWe9VALaN`j)+721H8
zhDq36y*^@M=<{8&;wU?A)D|`pcy$1lZ=EzXZo!|_GGzI$NXV7MpCn=*ykvqQs6SP)
zLjT;HZR5&e-*sFx#*E;R3jahZW5eieS<Q+>U`XW^JwpU4v2po=(j64RvC791Scl95
zLt8kYupH60&h2oVL_h{<HzOM+j9{k9L?rJGu2Bs6wsNGsIC*EKbIMKP)_|Lvn%mW}
zPbDRfe<z>i3gSc`901L)cHk5wGJW<>(UgwOyk(yf;&+m5guMKq7c}ht4Xy>?Dg$Gg
z$~Qsi`@p&UDGdDy45f}hC(Y{-qO`99F@+jt=>eY0(zosamnpCZ@4rX8H}y)S8LdJI
z|B!8G?T5M1GhJR<op~-_>ySR`?Xg&2TrF^s#2u=LB|P?sv^)aSX;{<{xE$60{zd|D
z03lL*qst%?ei^rk-X`O7EY{?joWE#M``HV7yu=>tcj$@-7YHg}rg`VLmHxxVyXbl7
zTXx0$;@`gPMg05_={dY*s(gr5oWbc6XDvZ)U))E0HCRK>S;Q)j6$0hNuD0p1Vn3(5
zk(X2xzRgv*)$IoPE&{OiDGfpAV~EjiyKygO^7O7NKcrWMZNcPPncLA*1kz`STNbN{
zrBPZQr{z3j>*o-~#v%1WAb3k?WC>a!U#|#noVN0iTD;XfyNh)SpA<(qTo}s4_wPD^
z#dHX~fnX!LX)WO;1IjyK&sUPrCf79ms3N|Vw`TGZTQnQ(836$&MS%(`3^ToQ3|nqH
zgQ>G8NH1^4vknHou+^u1a_x9uA?*HXshEQaj*dNsdr>6^%fDhnjnRwl9#}_G+FNcK
zt1N%fH6p?G>=XAV7E8qflTi_dAV*!A?mlKd${Q*Bq&!U|B*0qK;rHd3&iQn<Im{zT
zQ@S!Wifp$#!b*?A=ItY7+pui3wQ%=n2>agcfJ55(?P5JsC$}8$8*Si*a6uDf%g*pj
zIdW#Q<MdnQ(s;p~s+cz6K?$d)7mshiQ`iFOh>qM<w*rhCs|R}6iLcVAz^;M+`Oc#U
zh);+zBZg38LpROnMi*MB*&9<ZW;n3^WvuC;>C?Jbl!WV3unp7O`0D=kZ)y@lQUMux
zHr9{P-)3f`cG|M|aU?7w1;>pUysSJAh8Pi;8ALC9c|e&7gr-_#bJkU27<}cp$R;q3
zY$R+bFJiZsMOZ<n*!RMJ<Swez5Yq;z-MQl~k9eXtI)xYyXo}OwYQzCrXp=a9k$UdA
z5&fJoG}%n%sUWYzNrD_;!A)>$o02o8XXi>dS=lTJ3S{rjd9infB}ES@!S9})MI3Xs
zSLjDwoZ^Gq6xF8d7&drUR=j<a6qhT#sRyk>2oOlA1P+#riEQfL6_!w9oDuc7@oVUZ
z+s&v9&}AH~z*#FcPXse;UA*_-FCJRrW_e`>9j@#M<fN;D9rzAjg0|cs?j7;!TQ+$j
zYW0KhHyaD6)9c_@lWj|$Nm`F<<C}pcr+F}N!FW`g*5Xs~D~j;zyJM>UilBxKFlC|@
zkt3QB`x(}rlbs@qw_P;f*+tH_NCv3Le3`M75B1K#Li|dM`wHcgh69k5Io2m07O0T6
zasEv3z_{{2hSR1G5xoLVmUgW?RHvGsTTQ=nLGGQ;MWN}5gW5b_8ODF<6p)31b09gz
zDlwQBR=L8<bL<6svL3X@sUcnndkm#Po~yZT#(e!;U^&rNGdIfXJ^xrPq0i0jgngkx
z)*X8R@>s=7hT`Xi)chEvmUl4Mbv9hx#K_2`M-=`5t4o}kq5kQs=_kZb_Z20__tdKe
z94hvhm*uB8sPRqk!}x<Jr+$z-U9gl1rI<0NRYFaEAXI<qwbjW;auxbjGy!)_NWD4K
zM3$;vffso<^Knrru?bR@qHQ%5--wH{nP`MdF0d4O+-J;gUwdaJ7~6!s3_p^HC`WMo
zg^3QK<@}y?m#to^QU3Pz%-F~{+xLiVBl$!<=!t(h*q)=c>W9pMJB5}j8jy}XnvPnU
znq;zF;6Y%=b;$<^<kn`5aY>4LhI&eY^^Z8DOT_Q&V^-@CaiCqqa-bkN9RbQQC;GL^
zenyy%XccX-siDuH!JaIYOnpTgDxl%Vf8HJZ)dIWwv++%^oHb$iVivyx2@><IhKGW!
zA0EqTO6^4EfjFFNmmX~%nAOdS>t`y^Fn7cU!g5`?Rg_kz-$8a>gF(R`tCbjX>up*g
zIP1?F_ZB?DN9!0$WOOQ;92qHyz84QHU7&Y|nscpfhxMu-gs8m|>DH?QyUfNsi{~DA
zKe_;e_!hCNgT7_8H1R)VtA7ji|B*Hd&g+m8#*;q07=t6`a`!j;0eX`N)UnbFlqN=J
zMk1#4aF*a(P?)x_G%Z#o1dur9p)^I2kl7lV>16!Trb6kmKH(m{toL7ktya<>2f-x|
zDVS)dGnJVj_dY{cj8;!(O@PDZuzXdjL=>jWYs85=Hi)C@T4=)a9)c=t+njX>M@Rd?
z>k*Jin&<S$#7fkv<H^<=$@>Jv*!KI#$36^#Uq(aG9T;A`aO**GXUNtorMq=ZhWanr
zhg!*_#CZ>f@U)fny6!QTtk{vj&zC<jv#^thM+`7t3<*aG&>TLrqsJA?iL)F#KEqmt
zjU^62PyIkvN+I2v)vMXMYDBTwX%->JNC@rYfeAjeyg&@{XPr2k&PH*hH!dT(u`BxV
zw8sz*w~Nq6aNyTAuK;pIc>vdETfYF^%3`aMm&;1QoVLSUWn09!L3h;=k{BYE^&7U^
zlFW1nG{YFd%i|dt%5j9{P5}J~F{VveMflG_a5D@k>L0cZbs+JrtOp4zcE}Aa)<M9|
zknxCzFUP{$mbY7iR2Tr>Ie{(#n$D!Z*3NI?k=PBuDr7bI!uQnj_nuH%8Bu$k%oirW
z`aB(&q#Osvynr3-K|cf}!V3dgkUR`2g|OkRX+*~k5d4<O&sm3TLCtnM8}oyS+==^t
zJ&aR^bqEGtKNw7biLWMD(XMRF;r|Mp`Z}D(2>nJ_PR7Vsy8fHY63j3R=$A2Vu9OHS
zhOmV*WCkg+`hlia4ZfI<ZA1RSIGgV(+JhDzx5G+V1m&kWbXN<)g{hTig#z_;Ef{-4
z#WD4kH6rjRg5y1V*06wfm0h=4k)bWpI4f~Wh-QH3Gmo!1?le7dU|F1V>$F-evot{{
z;pw4U><oU$B;pU(>LIeSJu_2R-W7kCnB^Vda;V~mW5Z~3rYz#--x(HYXZG9Deta&(
z#-KR-z4E@cpRm7MXRa)s@Be1+C~R9DXxqY}0>It_(~^#fMxDi+e4lS!e(O#vEHOp5
zRcH@K1`yL((n#Tf0eED(9P52vOR!v;X+)>BD5N&T@LT{e%F{thf^EtMujDdu815|F
zt@@*R3?J%VRJu_2Au+}G)Xx6<#0qpb5G}R`nl}2raB6sUZf;<YBzVv2M*O4HX{f(d
zg82YJNiEQ3-M(LG8@4;XL|wDYbjiNx<Re=IkIFizIDOvEKy;>>6B>nEl%SfBc|mc8
zaYAs=+-H3{o$hNucHR(J9f(%|xj+n=-e6{-&kiC*sGJP3tlt$PJ-@PzwD=QGuMJ60
z_YhvJPMSK)gsY7lb3J0!#U2Da1gu!xX_I0#r|GuH53k=SD=MxzB=Ac-OET)SYD?Ad
za0uZ1ZS*8jvkLq4IihHSY~?ZZb-k6fuNQxOL0HlA%YtLt4dWX%$(KhiCvfff!>=-2
z0oYF<Vd7LfBPR|qlyS~G`{r7v8l+{R;}+bFWNG4aL>RUR3g~Kf%(q+HWNf5jh8_8?
zLRSl9{Yjhb`ylWAnR9p|@ox>zwYqiaWRKUX&$`vn>+;HyL&7UwN)n#~zi{+o+&V{+
zI;lpL<ZT?)iCRm|rO#DZvcHs0FmBrxn$ZPFbx0@b{6OXfZx0^i3+*?%pA)B?!L>U^
zoSguXaE5;I)OWPE{!<f&IV&uG`@BFsvC%TFTDj$~z1eXe>_iW~f>P#vmY=P0Q929R
z{-vpnJIPYA@wY=<JwV2~cY@Hz7lGA^<cMF|mB)eSZ<U|Z`>nf?LIN1nIK!FTM#eOo
z3oP=6@s(|bx=5b$#n`Z3<E8wRIww1Wy_9(d*%V4Mj8{)gL(lM9P46&-da2Jm*cf+1
zr%tB0wqB)(js>3ib`+KvX0bJErOLh;v7(<tyKTqJ(UyBKeNUHohrHe1CP=M_7T0+A
zQt1+~W*4GA03Q!RmS9DdgRLs4fsm=FakH{VO~Q8Bn%7Fbr=Z)o*5Y*tWDd4Xr-Dup
z;8I4G<X3U8D8_b+=o94z5vx?nZlYZf+xSdkSv<&E0Ug`Fzu*s3#E>u55<igJbD!K9
zhPYMn{8;&j;5n5G(sPv_@e^bVk8mk2viBAmQiDZMRLR9d1)h+T>e7>MgD=##3ekQ{
z!6lLQCn@Ur-^N@;crbZ(ZN6HqFy87Shps&^pbP={>7Xx3?fs01B^I(JTNxJGD*N=C
zKL1OrTH8jxu!0*c!l#bwMb_^7BFeFWYxbbYVZs?0iq`ZU5r-ztY<q#Ov)(?kiP_P<
z`yU=PnFjnegzi~CQBYGkO+g{q#g{z&ole-c38XltxePwjq3DU*pP+?J;tn-9_HS3i
z_H$fJeZa_X+tv{w$*(%<tY&xl1+?r8I6T75dZcp1WhEkl5JyX~f;5hn;m+`XrM;|5
zs?V;@30(V!?d&bX$>g7Ei_jwCsu>tEVT21A#))OD<Y$D(eTvpmV**FgZjJ>V0~<$f
z6HsM$HsAy#VsH>w&Cj7IFjhNqBEu7afFp?rNz@Z?dyE;kX~^8sCA8#^GD=SN{mSy*
z?w3oSGWG9E>A<!CEKa`Z5j?Kl^Po1S*>mxvL0jmZ5sthy6&nBsyuy<2D2D>P26fSI
zSob6A)WCZ&H@AX5Z3WtNK=>pEgmK7LRwVUI-94#)tf<JyaH)X*AY*w2N0SAtKAv@M
zR{x9Z5u8nl*2@$V&L-DEt09T1nLq<nFwXVWH)ML;$E73dO0w_1ENGm4f-c)}0z-Au
z_CIXVx>{z_R>g^4WBWK}$25;f7*8gA&3qYTG_MCO^ElD+!{?yR-Kih(M;SNBX`x;C
zvMC8Z9lf^A$H&kXjjRDI_c~q$b&z5hn%${WUOIr2dzrc+=y3s(mWcS18~K@78u)PM
zTEF{x>c+t1aOe4y4KWk0JSr%bl)<$kb5Vbk<MC!%dX$&sAJO*=4~Civ#C`IVM9u@y
zpKRC}zBkqReDa>h?QYy7*~o9QRuzaWo&*>PGYF@4R02p*+Rh}eQR%~^j>Q!#@&^Sn
zY5w=CfO1d7H$0dsgoH5oDfVfSTlVM5zFV2G%;*Z|AP;k@y0~T|TGwfEU_FLZnGF;@
zbYnQ=*yfumRN1IFI~+^BKezy!zRE0Tf;8AKP#)BI5}$_wx$JkQJh2RB;`Z=vf`m7J
zj@aA!6$xnvKnjs9lHCRb^S>vUBQ47RGXptLBU*7(F)rtU+@>BdX*|q!ROe|wUHPvp
z%}_bCqS!CK32{WJAUkA0^i^14Q>5PO3%lh~B%%6dbm{+>icpfRejiXkJx&g0E#_}U
zWI|2^!QWz|?T-*Y(_*IN%^<fNkwXXJkCKUt_a;^_lWy*v)M)aZl7-=yH)9qy)?K7I
z=8h%D!S*CPATI9B8Rq4Wbxfi#4e$9wKi5n9ttd`=cBe`oWS-i`{7)(!CmH|)^-okY
zR;}v6=5_1WuPI6@YXulKWVf#Me&a;Hu?XnUCEnmxa+|8;e)jPSU;LjScp8SLuMPH4
zJw;P6xSs)Hc;-5|->$=>)}UzQ4WAoFTE8B3kM{ee!d&*|O5cOCj?}_f!2^VySxf`Y
zcBC13I8JonRI9<Ipi}#<f|<N_1vQ45iBhItMQPTOE27*|D)Nw~&7!Hp_P-1Jc!Lu(
zgAow~5M=_PXi<Ip<ux?zs#u{lk!tsJ&rMxgx^c9fM%+4W=4dULf9n2=kn?)bcFC6H
zz{kx_#g)MkH)XG7qaVfup4{7H8l*G<1CT!_J!W8gcPU&ei{#E`tD~LyG?+0V^>&)e
z{We!^I9(HUo}3SJ(dF(@;C>f>h8Gm-JQydBLD*E<>Y*>z?T(*JKhgYWc|7wGO7%x~
z?#JqDOd={LN^mRdz(g50xtbfUebzZIjL+Av@!V+p*xRbaq^A%)WbcO%E$7rzpYBfI
z46_hryYrgODxKAZ*@uGD4Nz)*sN++j79ZZ5+KxIXdZ-_?=IT{jyaTLF^64h%X2KDS
z6ZEK=Ti(KsczBF#tirab@$aZJ9QDVa-C-i@$A3*-RM)Bb3S<wDq>0;Vfw?2Dsl_&9
z^^zSmp~%(~AqBcAa~jy<+y)rEvuqTl+b#Fw&-&TI-A=IAbQ(3ngYg$=wNtQrA`L9b
z+CF6^{=>TlGRJnFslYY&1gI^{Xc99S&x=y4?sLZvO)I*uEx4MAG5cJ<#3p`eoAo_>
z*Bdtm;h||<AzNlDcquB4TrEAetNv5aYHs~s$)yuvSg_*-fMM=5uogYTWPv3Ej^MU5
zKM~eAW-EY5;sc{*Og$)sXWDCcfV(DqPiKK+DfPy$OJZCheCX?#dLuaLebwrGj2b*@
zg8vD;RMEo(d#7iTP4ybrIqA-8>~%tWZ9t8zFdPxFZFbTrdgpp|DM;RN3Qmasefx=*
zj_Iw$Tb<@r)jlMJ!{?J}?1)}StnL9CN9&IH^C{&+i|I9MDm`|W52<z@WN^^}NU$$!
zD4j`3Ws^+!54jpPA&~Pl6R~0ewwgAIRnu6&Rq(rnVONdJ{=BE<&)MfDPD!V<Jo{8}
z0`J1i!AsgZPaLa3<T~Re@u62s38_4fqC2RO{XgvvS8Dr$7{Q#I$(MQ`CuW-R`1)K_
z=sqnE2pWJdLpy06p%$nC{2Er$Y|OSxt>;YK-rsqqnWO^SLFd^nR0U>-7%OB&U+PX$
z;k%c`&oLcQ57GqDG|U&8X|p==xue%hqcbEE4yu=X)(ro91GE&Y7XjLeC``$C&T~(G
zta4P1is-Ip&dY!(sc;ojj%Zepi<my|k;HZI<j4)Y`f|6(Hbx@D$39;{0X#ZyJ#xm!
zD4g#$p?gjO8XH1PWxr%#|KT~!ljp96J+P=5yRosZ%9i?d7_u@#CN%O&TqWzqKdhS(
z&B}t*uLr@4pv}s|0yl!3t9S=A&C`!g-@+qKAHU`AR&dFqf$$U%xd}tQ4TR!}Mw+VR
zficHExf{db*^TGy+WKa;hxwlZlqq^(KtjSRWL!2}Q1tG!9h+{78k-QU8{nERyS&YA
zPiKZxZQN#udE_}s6~6X)Eqb`27kS><Lx>9jF~&|0y0i6Kw*mbHa*gbRPfz(I$#bjj
zEqwNm`8tG(+%sfir~S}8G#y#PS449?i~9ZEOMD=nGR_6D#b5Rh8v(`=LSGn=Js2%y
z#I8rW!I8gd%i^)ho^Zvp(|l4btwdt0vYW$duF;s?Ac7%B&3@CotWU_GqsEv0x?UoO
z!)6v_c7k=d^Dw512Q=?>7vT+r7Z}o8YK1VO7#@D&M@9O;1J&I)ux0K>+%QBgB4|dc
zxIHdVcTS{$4Yu0DO(^^F1&6zzB!v!evY8O08K623$wW8fVO{7`u7Dz5qn54vVbbl!
zr%}0d$3jPL5eY6?^+P{}=x48`%4Vv^U9pyV5N;{9{|47an`h4WQ29uVR-^>I_{uNL
zO<P?^&vp}qAM5>j(TWI6i8K|2*IK3N+^l$Q1G)x7h2q%tV?g`JrzL^VMA@!KbL9*f
za(sOzdqUgfI^W{nwU-XwrPAVTU1Tj2`CzXDZ>{Cr)JLnn*YnLPU1A`22brI>Pk+0I
z?!q2%j4<eX^r~#6xS%}gu+bZTE+OWL0aPZmS1$4WisJjhnUX8n$sjpnlS1dEIoK32
z^H@+*&reH-o2uT)WhV{j<!|zv$V9WkmSQ!9-+P>B6M9wi?DS8^$bDzF*--s-XLK|A
z7&E^{{8zj1+bO!Lym^1fGyV%EXWNzKV~JA}Eudy5PUY5VHDt<?``fKcQo=WW`?pE6
zcG}*d)tJZnG%OPkdgRe7$xYL|NvEWv*;qD8ssVJwHr47dX#0)dV%~h5Fvi#X!*<ir
zQ-k?DDr1+MmFnF&G(U-ps=&DVDVx-e5j%ozIYUb-8qIm!SZ!p>WQU5}6t1gmz^eKv
z`ZxS2T5UP%v!7GLPVY}^{SIwI(Qb1~1sa!BKi!GW$7)WuTKPoEMdn7_ri6dguwUYS
zGgx!vD3jlR9bslYIO0N9Q2%b^ag=;ijhoY4kJC?nwTQ0MMmIrm4WJYs%9tTfdpg2`
z6Bb6wLq~kh+V6lpQH7;T)1QWh>&oK0YK*e%7x%(4-ppEty*;NR%6nvu?^*ZY=<gAS
zid4hJu*-mNg#MeeQ-g%0d|RNVXZy8`ylM1Yw7Z@`S76BXC~4E-Y|iDIPCi=|^SZUp
z9Ha@_i9&oyF0Ki7Pv5M%xa7!4+y1n}#jvD~gmV{Hdk8nb*2ZDoa)w*o{^kQwUgaHg
z71ERHfB4_LAnyO7<~{1lX)+i5Qfx@dm*EY8+%V7uqO?tBF;G_xD>ozQ@q!HX01Q8}
zSU~R2`x4<DE@qhMLF6WIWnKYzHtE&Q5aafwR{PV|ne-vl)`P*>^YJZ&?62W#e|Tt8
z5^rt+mueHi!$8~WQrW%DW4BAy04EO7*c$WZ%Y&jHl~XW(M}Xg}F;$S*lJ&;_gT1#7
zi?VOmg+WxBQ96V{X%J9Sazv#irBekYMnFoC8ajp&5Kxfr?v9}m=`M)@>6(!mYMy=j
zyzjf#`>k)Uy;ttN*E;xv<JJQPW`6g1Uwxft6onmj9Ay{l(OCKgXRT5wb;cCQ6x3EE
z4c@3J{<<as!#3l<BvL>anH30=jmGqbH^aES<wxZExiZu9jc}rK(IuJ-a+YP$$db$^
z^dZ_!^tle``29WW5{gJ%xl_<2;*B$ebkm!@k~&d4usN`9z0da=d^8Qje3|l_Hzyca
zr%?3)3h@NiJd7QnMH`w&F>6oUFMgVXY>v)2(?EfcrOd@L+n!yjoLNj`o5)BMgC57H
z<JJJf>C;CXzKBOZ&vaKHVbb)N%+|Asg}CAk-AS675>2g`o2@LvE7v{X*tG3rw0sAG
z1T&(v-N2TT_t5`1#2<J9@ok~*qm;El@11_dMbO6Xc6rg9^Le)+ZbPhXtw}^Ete6-=
z+kb3LjZhKp<h?VDQo3vdCcQ0&0%m#FXLmgDGUkl-7T4GiVn^F1N%r<sSzIKyi!_+q
zGe~fxKjC6P=g!_e<k$fO#eL^e68$Y8uX8b@VC(txt5WsWhg{}6*Hj<xnsA}IVN@3a
z7~eEZPT%A4#BHSp<&deJ7)N>9^Ohh)_W6LwnEI@pZF(KgSpVbci>=xV7I)oCykgk?
zcc71jqlFR;$v|8BA9|UPck%6-ecFGL^-$xZ%z6nl<}$qn-)gZ+w7uD^E%AFG{1@cM
z58Q(eX68b?xsQS-Q6yaUw2?vtVIx{EiM()Nk4&p*cEZrW?L$91d()Wl<Hw90<jr{T
zYktbjnHHnnGkSDRdA+^}!-uE9T73G@s_A%5VU|81Hl%MuU2bCVywQa2lR(*Fsc_P<
zHb&OLlb;m>%aHmeM$X_3OU8jorIL7G!yK&OhwvW4h~4bY)f+z}^*YVMJrdsHo=*LM
z(ZOCa*v}Lmiy|gm2pBq<AT5Xh$@pO!>hcN^Gj##?$m+8dfM;cs>Gs%UkK9=C;he14
zt+_Pq0n>)6pkEL^w-~E~<O)t8tw=D%JoS{BIw#ibx%|+Jr8>_BVw#$Y`m)P-=Wlk!
zIn|$xM9fkIB2IM6sg5lFa-z0>IZ@-<>YzagA3+Q)K(-3UNh3NNj4)ef<l9ymFE(o~
zZ-~f9T(BUC%z6ogY)`P$0GXdXg_Q}%xTGQ~vEOF)Mw|7b;zA*BK5P7}BAw}De5!L0
zV4nCAr+fgJ-nrAABSw&VRh=(fy!S1D&gBZ}4yU!Z=aN7-sIhGQSLqgUrwJ}<Jq+^m
zVaCiB2W>z48Z^{?Mi)t8O5A}0TJxe$vk;c+z((3ZsO8e7hh{YSl<YVkV<e@l&2H*9
z-T6Oo%Iz;Gh`UlG(IQ`#qe71HpP`DYfcQ<&XVjFFdAAk939w@L4NKx8hkhIHk+(Y{
zD9O)0UG~E_j5(C6h7bFUrMs`!R91(+3Ia&uspN3LK1F}v(7B1q0|U2>yZ9z$!5X9c
zW)D^D1#W7BrkZU>9)kgK10d<|=z)-4$mDN>L2u_ey>GMjmlgU-0U^7ypW~uWX^1e0
zU^~(1*HhQbMMFgfDvOwzsq0>{Fiq`#Oc*jX&{MB}w%gY%PmqK{h#BZ__0v!PXiB19
z(1Gm*h+Pz#Dx$!l!hLzRMM8{nTh^cN-G5)=sMuqqduVHa&g02a-69(82^leKfczRa
zJh<3y9ByJrceA**X`oW<gNN4}0#rE)F9Uwh6NqnPQdQ7BU51q9W5st=py~Zovr8+h
zmTj$!C@-F)5?aj#R=6-BJx)Qv102R1fQ3d6&HkdZgIL1odPp=?SoFS8v1s;tzZeF(
zi3lVq8ut&J&(r|Md>mncFpY}#*lx{dx<D00fK(lrowrv43jDUEm#3z{q~&)t42yJF
z-ln%0`c?DT-cFl;P@V5bCVWSpqXH|<We)f=5JSEW5Xb<8Qa5rEEMIoc8Fb)EJ-4EP
z#2KW4dSDn%QGoru(;XCE4GRI231#NHy+>8g9|gAH89d~WplCQ;BD&aZ8g`MA`fO}H
zrPHNKka+XC_iIFcBC<&X#g7E)L3DFQsQ83@!C&20xB%sykchZkVg74k37)>DuOUl;
zsT><OyynjD0?9LvrwZDU0Y>imQ~+byp$RA|m|fo6RPzS)F6Na^?h|<4z<gL@$f`%3
z1PC>=yPQ`Tw&GGP0M=Cya8`>Up15d=PH8Z<W88p;VdiyUU}ejOe{wLhQ(Tu9&}z8e
zjP7nkMIC=0E3&Kdda|g7=7w1s4X_pYYRDN;^gQ=9t&6oXaOuxcd~5qNx}owJAtYNw
z3Qr~I@#ZVL$fh7YpL;(1{vQc8O^4)n+mK#1=f_{rpK8ve(}m=IB1b2mfH=~xA=2UL
zB7wO^3g+?M3-WeFBiGab^1wmZ4g#s-I0X<q+yS_}0~|{7>F!U);D{fLA~M6k7<4Vr
zj)PtV>O?&ryBlqMpEr8^r$A>~)M)AOwW;OW3#ygJ6T)h<f~p=;0>Abi=B#*%>}{V6
zz2iB~$)C5T79Azz<Gq`Vju@4r1?WpIx-hV|vkUj*#GWVPJ~R(wvB4ZCJ=HKp*bR+L
zj6CwXv$@L{^9{Y1Q6JyQI{SAts53{1kAu&CB9OWx)WSq*+>rnfam0TQ#sgoV)>I7V
zP>@T+)8itvWX-I~QK#$+a@n^#x$zs1OA~i47?y7|`6Ls+XpL68_|UtJ-G!+o(6gty
z+>L?qd@z^NU%5U8LEaffY%oFrF>u`rPCN*5w!#yF`d$(DD#yUD=z+|y1o(zE)ssA&
zmt;7URevLZz&!;4ImxFH7v2}kmO-TP$*yi0bMdvlKR};cuOWba03-<-{}QK~{{k=e
zg6Ax#VRUEzv+zKXVA+QG<Hf;3V9r`J8#XL5g}o@n03y@a@2Fn<RtP?8d6OnD{8|i`
z&K=jA&9A@QQ6zcAq8=j79geG(cC8)^PbK!0K*zR;2M2m;yskY{+1hoRD@f_jEvK3S
z8c0AKQ4I(O_w+0h9(gpYwNCB>>nV(*rSXqbCY;(IaPZ0JQK8>tBUPaCbukRP_s&N3
z2I=0^x-et!A@aFbCACqhZ407bvZtEFa<^h0w0VTcu=_whf8Hr!x6Qv^GXEe?vGh07
zo{*<JW~QVk+s(9truKyhVZ2rUL(ZdxX{t+{Y3z3xf%9CH^c3<q^!m(vv+e0nslnj@
zzhdRc+m))aYWV2v!vHuHQ5qoD3G8U@Ur>xax%{$w7fszV0`(S;mlW<^oHxXqP``7q
zziVnSB8x+X3f|oXo|4sy%QW-U?i9k?n+qaZos&Orzz=~9C%;+~ukF1vV4f!$6N-}P
zF(q=@P6l@_&>_b5Z3kH|`^3G;((*61!$e$(&PmAev3nhf*vtzVWHL{efya!^nO93D
zkY>5_1dxz^=ufSHwVjsU7`u=fp<lWaN{WC0wXm<n9BtCfQ+&vpvJn^J`ka>`<%SaG
zu|O<DbqNPF9lh_hpjE#t1Et<$hs`soLf%e#p19}|_uGi;OCI!>Iw*iRkpv@eJyIpX
z`zWvV!82T{k>Tqi@84M;f1I6rR%Y1l_h?1Nr{E?IIFKg(<Kk?etpokQ^=}$04j;eE
z5eN$d948FtzzT5N{i7sCCk=FJyN7o3==z(K^{1LtOMCqR9|``z0i@f0PF~ywh}#yF
zq4^uN$1mi%^xxFD{PdC+zLc91ULFcsphnMTVEms{O!&P+ScixJhvsL<6UsVWrZjXp
z9FVw!WxN2Phq_eUNf|n%=@H}nIp>8xKdE)i;dle%(P5@DYDm%uWTQWDbE$g&z=;pD
zA37*~0(j&=XBg2OgBOcR?O3{5O#B8VJr2h881M*wPII34dQFxq(|4QEh5UQGWK0sw
zGWg?yPE)fjCjf)pFZ)^LV_Q`2?nBhseup<pl^O)t^^wu~P&2x;d$(bCP@oxY8fsd_
zaY3=ZjA)ezn=6c<R<Z9WSI}Y9?8HByk$V0z9vlag8K0DmR;mbk(RSS&Agt8i6Jk`I
zlJ-t^B@;jWC;+q~)Z23A=SA*TC%)LcV}bZsx|wQ>VO#?L8T+Rs0I`1(&RRv4=}atA
zH~7tg^WrtDNk>bGOn(GWI~&~h^(MfqFRRv*#R9H_WJ|O*<}2p|0siHF9IEph06LF>
zkNTl>XRmk#a_2rxi*Q}kTk5u_>?y%{d&EES>YPjKoSUgTWPP5#vLwxVih1=oqhuJ@
z1n=A!`^|*rvQUzkJvpz3F*v07Q%!Oaw<ikma(Xp3_SahT8IYX1NK0Tl!Bj0c^Gaun
z!Iq7qP3>hi)XdK-O1zIJSW8z-7#%XUs1e&ce6Ij|{jmc0w5H>o$LRSdXchWChKQmP
z0fNI4kkMr-_6{b=^*GnRs&IbZa~dC@1RkE-)U8e3y1xgE^OiUW!tMfN7#*OQp=FM4
zQg<ni4ryvcWBqGf>sJru>i6`0fYAXfNF}TTL|i7W>nG$v-W0`6bVy7pGNxRscYimd
zCe)3P{01fNFVmgd=S0)qebNjGI#cwdYCZHZQPN0t+v+I4d(YQDI>iGu<6}4{&fi+=
z|A(%BN@p0O7Q0NCFDoQ%KjdIMG+U~od5P5QZT2Ftk^X)+#r;i%Ho^E8_SB^>2aqZ+
zIqRZ0b~Isb9=eB9Rql9-)vyK#H_j@Ss)MbzBbBU+TgZ{^>#MjRf=49<fBisN79`nv
z=C}PwdH;NLJphhH7mTNFKIR1J1>d4K-aef&xX;WpCn9ZWl=_(>*4Ol_dxD>*Oy<0@
zy+?)(_1Lp6M)PV1KQbwvv&bSwk5+F+^i#VDXQ;&SQ6jSn4oq-?24lFqNM}=KJqrnT
zVd%?Gvf-Y41P*1nhERl=0pmz<*kB6G+;S&DWrsa$d<&zI7^v!#>q`YGfC-;L7S$4s
ztZQoKSE(si{T%e5Iv-FFtnFG1V0w_WoB*F4Yd1)%!Hy*L{d$}FX1<5vow)8K0;K>8
z#<BtYNWA&;HnWV_<hRaZ!zwOz=#u44-x%?!o%jjt#zp>n-{IFj)Ce$SshkN94TeLk
zKqN-i^laE{yM}9-icq^>oc7|Xgm|U*5HB}H{KtWzo6oT*ryd{@5=a4tuYELu@KFv(
ziFb=8A=j|rxgXDlho}S1W8_vyiteh*FjX(wn(jq?PkzkAdv0xgVo<x2iB=u_zMx!K
zU&Rmrr~OnZLA8V<4oHU4EbT|qX$J=!g+KTP@Spwm#jaVih<R`vjyF)DYoq;bE{WXE
zmsuMIBRq2CMZR3?o>CWj6_W6X?#|TKDzmyciRLh_De%_X6s1Wz&>qKTQ$OTb$EzsC
z<6^#&e7iX+;Qmg<OU)GE{?mcjDDwyXfkViSB8$j>$`L9l^O>)B1-39ZXC`r3A;*b_
zmR6YMqN9EFX(`IVBook6;Ht+=pvT(1DQX6{8_QWz=2%YWq^C_Mo65>xD~fRxd#3?|
z@6rUlABx>#m*p}o8zw0YRRlbpKT(ro0|3eglhMODFEDTqt3qGa)JB*^3eghbZ0h9w
z<07OqCt#OqH|P^WfTUfkYP5(?m?3qyPyUzfYilsA_d9%5D4?K7xfluCcrgh<5vid{
zH>Z8v!$<!-V&Kk{VqSOy0ZqD^9#GNbg0#nZlVCPK<07TZAV;el7~R9oKOYB?(i8br
zAX}M7?3|*hx0y@>gkK!h0PLV9ea9~_g?S30Ab(+M%SUuk;6X&zZsxGJmMxQ?4}?UT
z7&u!_=?9)qC(?i{A#s%drd4sUXaZlXAlcHr8krN}v;Ff&*a7d`mtQW3@q?QL-(<YT
zGhVE_X&Btv%R5EK$TMjL>`6Ym1g6Aao!if5wojhL!-qGqLSs}B%^l1D)2s>P{d$zD
z;F$2yJSFbMfTzM`syFim2b!@vbXkpfi@J6gob){~J8Fwj=4Rc-g<#PD`?tfSb?+pd
znO>~+rM!urq0Iv!QFVC-ig%i_B&1GfFrb1@Xsw$HtwJ5A-Wc3nc>P2$6&QpRxdaNi
zOJ5PUfVFECrG=9hRY~nFl-GqC0)M)>GAb32GqAe$xNk!2ho;Rv@z4oDyV6V>4LODB
z^V4+fU$k>AK8sx^$M<}LaJm+_L_$RL=hKl0BJh+(?+1A5W<nj@UKr>LZGW@to)g18
zy(1zH7$qnww`&ZrEQp{BSr@zG1o60VT>+VUlvXndthd~qj8*@9EUj>%;h3%^22kiC
zNs}zHS?)$%ZF6Px6HyhU;!}(W>3jWB;>&V`RWlf1Sa?_r6?Fx&Ht3JXlV@Jhp2|WI
zzgFeXB`Sev^<ZQrm0-b*xSJjkZ2vuJKm=Kw&Fv`4RUpm;2$V6}bSPIEpXgR*Hz{xJ
zwLJ!yO{<GdFkw24<3|j^4J{2H`-DdCRF-e4cqsyIa};wAU}74b6{>yTe<l|~mU;i&
zG3>s%kd0R_7~uUK`7NnKT#N;I5~HGm(rc*?I?N~&Y;Fp^`_}RbkO;`8RL3=OLT@iY
z0uffQ;FASPv`Wkm!~WdfI-CzRzup7A5%gTOI}ku&?w$iv<!!>=T%I6+p>^ynLstre
zQI&gKjE}?R7?ZLk)vwi|0GBxti@FgP0y<TKn29Oy=h7{?GT~AjH~Y@#1Nm0*y5aNB
z%lrWI`*0Eqnks#G>_wrGDka&d$Z$$@Hg;?cJZokb%9w#X>xP7o`$_E!<k6Fgd?gA8
z7Eus^0!zmjiD=V01F&Bem?X`ZN1>`*V+_d1XVb_8ku>cPQcrS}bSl!G*cv^(IT=E;
zs!p%ddgI3l2P3o!2$V3eW1tP|_ihtXWtK|mCZ`N2bF>Rf+wc8><Kp(HmD!AR-!uk6
z12bQQn_fS%FT*6XyN<+j^u<pKzKd*5;`&|rW_zJe)kcyYTnDr&D(s23MRse9@13!1
z91lXW4J%Nf&}uF!UQek4)&6&qjbb2dpXnHTF>0v2;b*(y<+b*$$F4VHZhxKZPDtOJ
zy?>2?n3MM&xP$T)!0k|bZ2RY@SH%nz=$mX$svsf6(v15hWhOH7I+A<%jNeM<z=06r
zDzLxzjnNA(WTQ~!*!_D11!)#T?>Xu@gr|NRS|3QQYX^u13aMV~JnsXOZn?gXwJAam
zkrq}wnr-sK7e2pKta1l^2V!!NZz^GvwFSDc^~Dg<%Yf5t>TU8*53y?np~Eq%k7%hb
znP0pwO}oe+$NjaqJ9kbyjh$Gy5i;E*4huV6OdLENlNdJtta!*82}QF2YY}s3nx(1G
z+!0A-qz8sC^QYvks*#H0m%Y*7N;mD&8!G+IsNu`BYY{<2-_n4Iy*_tfus=p6ENfBl
zaJFVpZ54i|d~6Dwfait$Wkz)fdxBb0hY76?*lL15lUx3rJ0OTP{J^wP2SF^!V&-Yj
zjM_`3>Xir&BN&9{YZKw9<81mq28Y0E*LOgmoj-6KkX)X)*;@w1ffv#$9rvns*E9U~
z-bt$~F6n+3m+#;YaVN`%=FPa7>Id!3%EoZr_>{_n*9NRDY2X~u{nf?2GsX3^`n{`m
z8|bmbk3zZDDkIo}kVJLP>ct=BJ}^&pq@whJz7tfM3#HrPcId^o|HAEv!Ef7|SV4<6
zh_H1Tq2TVE26zLY&PD<Jplc5c)<Vh*m<PxL{&zVGd|dmDDWD&ifK=vH8in#y#dr-5
z2GUv#sY5rL3ol1bH4ir*6PtozZCX{N%AKhd5xXgIX42{JU6kM4n(H<$!~3~Jbbd1y
zc+WuFj@6S94I~v`{^n|aE8<2<(@!)#L~(RYp0y;o<P4snEh4wrBR+rN6u@N^AoV)=
z)BC97Y&?8SEDIQe-(Pyz`>4pMmExQGvQt&tUN4?EJc%*<(aoqTCxE;0VcU=rXz(un
zURS&RVLU-+++`?G{HxQ9l5$Vlyu-bRz067qw1>3IEqIvFb1`5{O24QjVVR#`ODDnl
zBtTsVFi2kP0EO__jfX3Wl!X~(Qcp?Nd?4|?O@c*+HEg5y8PZX2S5D#aApGqUy;~dZ
zpE?hJ?XtebK0)Mo86pA4xam3Yw4=dm{LA3xT}G(90_uHmS{}W3oV69c8aMMppi_U+
zzKG#i1bAl&Zw!7xirG4E5-hrWxcB51)486Xcib7xIWXVccDD(<Bwsy&y|;zmWl;6J
zQXXnYqO&pozTxXW)fr>BGi4R7vi!(fNr{}f%~^}D^j?7t0q}*{2hkaktz<X!Qwa9b
z-Etx<Lc1p;Wl4%auR#VeKo`dAvJvQ6GO><a7#6v0BjaL#k@v<o4fUgD%qlkP39n)g
zxHx?9sZLivEFo@L5B9pGZ|UshURSoh_h;7O^?G}OWVlI#X~0mi8eyuKaTAztF$H-F
zqjllcoIn!*w56LMC^D|Mn@V@`sI_!taRyK(W)ub#7c9}>uG|>9oa*<L!^Ny~_r#u-
z&$luElJAS0I7?4=Q<g*ZTL={kOtz_MTz>^M@qr|QXuXHLB~c0RYGZ0O5}O2&nZy$d
z_{fYQBLTBhmzOVai*qL=%}}{32UZO=zWJMHH|yUnFMlXD*VezSmXIQ);yq-`<o_$v
z>JSJCd4EgJ-MD@kxO|x}DP~`c`G71(x5QEq5c5&*Nh^jNo`!bQ9(H*}0i|pK#4Riy
z;B(*9i^MKNNNlKe9mvV14W2zZw75>c`Vv)A?L2pPlz;8*%G>vYJ%?16Pi-35Q%ej#
z3ll%Gzl;6(jaEl43N_ch0In0<oF$F;+PSxQhxyJKAm1*(QxClFK1?VYp;Iv~DQ|<N
zL`CW-+Lb&$c%#uP`+V<u<WC2~RFo$F%%CaWh0%yJ_?nyMZj}Th{p@YF<iba;#D4u+
zjy-o0FA?41B<fmBnLwI1@5zW!UQt$qr%=b+6Wff))uQ8FD?j20j(-Je$H*hDw9SGJ
zD6v%=e(g8z;|btA#C;@BRuWo=h_J&Z*IH1e@Ez){QD<qbC0yRdW9PRPYIwOC04+!r
z{{=ZPubtagU&?<7)keDp-|f4s86w(h_6?8e0cs>mS=)fUjGOY4lr`_?6Czc;mW$tC
z8SQXd{1iXXr7)*cT<W4{R6_Unw2t3IlW6w#$mb(c=>;^1Hu1qka>j)iK(%HeVBJ6!
zkrW#<(*C<vE51k!vjD+iv@Wa_sPzE;)sHDATY`24QjVC?Y13)ZW4tkhMYFgH%p{)F
ztovky^TQ-0e&E9?g;?89zt#g&vm3<2xER3Wm`I$fdiP6HT;@rGe*i4MBE5{i@}Z<E
zG$%Eq@0pM0zFK%JH^YI)0tfTwcHuIn8pWqv*p>~@nR`<Njvt~3?awV@+FR`E0Cq_0
zRLIlu^d3`zQCaM3MBasE-hlbG^z9MPXMw`Je3M6N_=6!pZQ|wb1zQP_X6KE#5F)md
ziy|%p-51+{o^{XiRQJG>M?zpC6ZxCSwI3#>AInQVm?7wZ#D#wN0w1|Z0!+vA18YIj
zWt}(d0`bzia6AXj$+2DH96Je1r`oGwdT6`!pymn5CbOs`hjljH9JvqO(9Hs%TG|gt
zXaxHo4N=Zugsjd~)^b;S4=&@WW9><&t|9mhfmDe~QJP}%0c7+0?=0X)o|K&RU&DH7
z?>2YiwXpLC7`ajGtf6uq5*PV}sN$k<oc_o@3`~5e7VeEZDap1>Y*9B>%}GzqCS%o0
zOOjg&(lB;aGQh!l!)(A9j{VN4nMJ*ZdKCfFTW<O}_og(<pZv_1<YDf+hKrMU+WtEr
z*{=Ws5>f2-+#|?$M82GQ$8~d$Dd7TORx798jj!~On7Gi#krfC+0G||xN<Fp(o|B}`
z2i8mssDZ-+g;u&ZAzk#>+~eO8omVac-I#b_!)K?wc$-+NC!~d~^X?bfl4lByZEo#O
zF_h6Mn9@mO$n&aETg}<;vl4o@UYUhH@znwP#-U(*&v&r3zvV>J{1{eP*u|5@n@9FH
z+`<%_Mk9S4Z+vX@jqO(ymAD-}9c}ty_jgkIiIedrCmt|=@RnD4UF-6zr!Ca-dV;8c
zKJs3KvLIte8r8)Kq;2tzAdu(=sOfFNR==y6X2Vpgl622Ae*uk=QVff|-_brXrnHB*
zaO0%mwWN9=G>`s459t_7O7&T(0wtM0VA#V}9?jYlZf(HzXD1(1k}1wYPZ8*ob})n5
z3W8h_9O&XIH(o+uQTfr2BwXf>?qRjcU(;S@SldJIJAj%sKz(o==aP?sX_VFIohB~N
zjL1u8>F$xT4~ye)S^Zg-DHXfSc)O0ka4esDSS4EjEPolD##+AO!!wcxxNDK^rISde
zEk{X?BC~$`Q^(lxbKNOZ6wu}gHsM6A?^<LoY$e`I&z<Fs8?BlW8H5vtESbYEYyNyA
ze86Af#I^dXiUoH|2C?<YW0F`3V6Osz>N?HJP`f?dqNh3f?Xa7vP9jIkdwuf3Ct2?P
z@Wa$;U>7({NNqN$T~rr)XMATjI*V6mvLp||?<nhivTG`Hs=14o_S>NUo`7@{-PJy4
z!tel-R^O_p($3lIZ36VRhS99NgLr1~s%m|SCy78E^aAr|jTzI7Qr$XRKc@l)f<WzR
z3Hj4;uebMU*?cc}*3P2{x8uK10F#)2c7?J&%_$t46@s5P@DiGWgIW(2yv)zgH(4)D
zA^^<ow@qMEiLBV+4z)W|rqlfZ?xOb<B8nG0JEk)USVBuQ#qwIZb-r5*1rkgguTK*D
zDyL4}+d><kTVaLOwRCnD)FuuJniDsPwthX`{VD_x4kQWuNE*OM5MO{$DhYv2Y`N^I
z*|T|fa9eb`i5p=Q;1|Zll<GHZ<<)8@iumrAAPvI|LuR>N|D}X{Dt2xTYvVYA+S;6S
znG(D*>>~_~x5fR;(>cRh@sQc<GV$_l29WN9$US+{t^qDwMZ)%EHaZCpl8fjkz73V*
z-K=<pYN>oT_j;5++x}UNBTUA0oV^@5&)<0~q?(7=W4cE+@xgRaG7Hed6k_lfNkyI$
zD=PL{A?W~fEd{#{P`#zXVT{!-Rl|}GU8x$%4)*q6d{an}tQ`smTGMRjAZ+^?|4tHe
zv!rD2?v$j3CuG#+A@vCPLC6X|n;fuj`>3>l>In)5ETjDRhC?ciH@TbrioLH3n8+=o
zU1C?{mjFT_V&vg<{HHTP?o*bsINp@lF)&IQAc%q7JTfc<R5wM#n)u_R`D}gk?;8DN
z;Szg|PsiwB%3^}y@1%b_L^T<lH}@?VdOyB`548Nd?xq55qsI^VoUEo$+02oj$gBvU
z{@!A{`|9FZ1cQ1Kwtn{8;y2x~)JlM97q*<+0}UfSPZn+xM23E*887uAt<AqC88j$W
zcw+e!n6vQ&R>_&0sRw(ob%5;#0%Ilz6wMNUAj~of2MFZbYk2Ihb4zaf;l?A;eT}XV
zgTvoFt|$bMS7A8T0?*a;BA~J{q<Gym+g(dxT#k^2B{Pl%#9ePeY&U`mdt>V+=oZFc
zcXuWvXsg%r@T2?1m!1{Vk$DzQI|5(=1P5!#4umB(4kA&o+kfEL%}kxbQ^s1cEC>rj
zTRg3=<*mPVe%k-6z7*Pqka_Yi(=6s&>4-*rT~m-L=H=)}H1Q`0QytsgS;6i+3<M_r
zN@rHEgpFw3m7shJskL}4Evi4_d{^@jN6Vey7j0@g!d4pZ;x@1n4lFyZKSw15+s5;F
zXx+5-^BL>&tsfePAKo`h>*594J5jqp)PH?y3}QuGi?;MLnR|?WscCojS1bI$Hxd`u
zaB-n!55Q&;C=6Ut+XeDFYa44)MoMg$bNq~%6cm53<M@SYT@XlxYHcevMMr8g|KdAD
zBFr`Ep5jf1|8^e_Q7P4r63;u1X$1W11lt9H%?iu6-e64Xi#V2Q)N#IqSnG`CU4ZWs
zx~(>oiY0`hc~+>4Env$Ya;VfmGFE6|xKJg@FSyUoJFfQ-jzaU)9r8X;%`8cao?>1<
zu6FbhOASJhIYR=l#AA~D$hk*5&W!{C4&DBAM`n8}&WC^@`rA16|La%p*RZyCke@0T
zAU8BUL7ef4UrLmIg}p7E-d4RVIV>fav7xU@)ZC~X5cqs>a0)Mw$BF|6bhUJGAHz;_
z@x|2_w!03jUp9Ll0LvD-sG&>JT5vRwFrG&pJ79RA=Te?`M?8rw-%m}R*?*~V-J1JR
zj&;V}tkRJE2Vz#rtAGEuI18$7&o}TbSOAtc^1H6zZdGq5Kv?sAj*Q@>WU$ai=k`{j
zSJ&5$hiY^YEdj56Ajyon$CVgh=*q@j05hh3OIGq@9PiNgPM7W3CJ$ru&Fp1cCBsXD
zvlaEGPcdUg-;AS{RlJM;q)2}#yR71&HC_d~Bm2^x&%BP%3Gb20j^Q0Io7zCw5}f*d
z#WT;RD}gPKy&Wtm<P?qpf@gDpSka=?(*#}f1!+&Fujr9)s#KmAK}<0Ik+?>jbBG<0
z37+X$nh0(+^<Q8@d41xdjE~=%vZebLvy^whfWB9Tuw0x0A!~Q0ENk28MTD;<fwh`C
z6P|-K5I9R`y8ntjljmQQpDnp?|2YOpbD0NJkE7r)LWtE6lN(GK?Q&FHUA-P(w)ah&
zv)DJa*t2UzEhL2%6P(8|bQ>y~SajJkA}CNgOyP^s-BH^W1*~alIp!2?*vTYTg;qOO
zIWE7~$12c02ydfue2CpBWgP}qcU^A{s+OX!O|IQX)l3csgChW<7wcJ<$x^QyaS>Iu
zu1<B4oOhq>?YQIMvrg^HUr@Ln8w2VQ817p7iM$%LNsG?bndII>&Y^gqaNi&&Uu=UN
zL;ze-fso48`I=&Au5{xvZHR+c{%uaU+Ws?wQ|J~D^ZhmAySn~GzwPQAuFl~K16LTh
z!oU><t}t+gfh!DLVc-e_R~Wd$z!e6rFmQ!|D-2v=;0gm*7`Vc~6$Y*_aD{;@3|wL0
z3IkUdxWd2{2Cgu0g@G#!Tw&k}16LTh!oU><t}t+gfh!DLVc-e_R~Wd$z!e6rFmQ!|
zD-2v=;0gm*7`Vc~6$Y*_aD{;@3|wL03IkUd_%DrtoKv`X*~OL^A2K*TT7E@!VCVX{
zjt4Xz`UeiL6NWypN!G3z*JeaB?~J0qF>zdn!R-qq=;p;14VpE0uN!}|O3ytit$x{A
zOZS1NtP8ehRmJ|6amx0o_N7j`^a!I0Rc}rG`^9bu_p6DLB&@;h!`vgq6OdJ83w#Ap
zRRUWt^mZPq{Q(=Yyx3~V=UYAw+2t(Z-<<*<%>03~Um!gX(8x)pV$U+SV)mHEz5(A9
zcHsedVdDo*#{h=@62y8oe}QtH$1|5hUG|1mfJTuPqM^pb+{iffEBi=4YL+EGqdp3w
zP=c7gD9}Q)`tQquMl-i2@ON*IXy`$g&9R&e2l)w|I=d4}<Kk2_G6y~N#j=)s&2Q$-
zdyqP>#(CGyWAZP#8O4f6jJ}o2iuSAR^RFurNRdGwhW(8s@#^{)vdF7*xH^ab3=C-7
zRdd?165eeSwy8&u!+J(xjfNwB`O-SiQn~AEsnhL6e|%9q9MCJFv~YJg%qL&dJ5yLp
zcIhL$C>qe(dw;0QMcwE^bE5<Fg!_j#yEonwf}??#sA}r2^JmVKI&Te*ihL~Kso6%6
zzvzN=WFZkw-^n948go=+*?1qf|5YI6w@AuA{{<B<U10W7(Mi&(3Eo!#oi|Nz*1rvO
z4s+Jw?FnUh8lv8Nw4C3&%oB9(!?A7H1JX1ma0(sMB43aJo%L#a=hTbYuYaldC5r5N
z+;6x~?p-}{hvPs%D%ada02-e^QVVO;=0J1*fz!tc&^x)8aTZTfCvwaR>Zsi|5|!0H
z)7BR@tQYTZbtlS~eVhM8;@*c;hh0I${hz7Df{C2P_Z2i`DO>-gZb~~kA@eq79KA*J
z;Ij9Ne70^zc9+z!(O6U`%|jousk4dCp3hyqkMBkuL&@O1!#QVE{ewAD-jk}?371>d
zR7XucgaYMfQ*=*an*9BD^^^&j!Vy-3d+_DVrWBUA=*m4O-B;)v{V*80H*Lv>S8r-u
z`Da=3GRkc)6d6R=)$Y4T+FXcdSN(zW{xM`>xqem2VV*iGrK^F9%AdrK;5fSo{twc3
zf0Mwwy8i!V0sl*%NfyVQlutNS+ImMkimbmzmqlW8hC77KKc>aDrrpR$gP$$D;kkou
zbZ##bmili&?(|=Qq>E|*6N{COy`0_TNkVg6cyXLpPXuA_CaVQ}%YgIDi240Ck%4Tg
zGHu5IGf_oiM8Em#Hy0|#!JRKk@CwR~?@X#teade)Og|8Eet?N#XO?`6*nM9(vNaiF
zM#1YysViAl=3!-8W$9SK5?**sx~jS(D~j%WIsUc|g;qmU)_`6U%SPXE^2qdvQK$61
zx~P5b-lIaEW&3^EQyqR%4QHO&XA2qS37uts;E+GK;FejIFja(bN144SJI(wtkx_v_
zvQ{6)Vow3a?SPo6^9tCaP;-RZDp<$^2y6v0F<_(|L_<R~Gz!%Tq0bJ^J0nhEKVowd
zbPHNGp}Y@MDrC=hVSoC(yIN1bx%W6^JL?Mz-5VTjMVBN6-^&fN4W3v_OEK2iqpaf0
zfG8ccDlgJg$>+`F>}RoBmQ!{)OSk+lQ>v3?ox1}c?S!8uvBhjzyM53PuFuQMs8Qq%
z1<cy#e;u=38b#L)g)J_ePcO{l?mqeT=ozDa<PBMSFIzV~8A&Iq?ea=E-}S@a-wf|%
z36S0Dccc7fmtIE)*Zu^naC%HW@|bQ@Nh`7n=hI&};vW2+W&KDY?;#0s)WI7QPR9?)
z4adD#7-d%-D{IO9E*?;_e(QYPaM{$cF0Jtgj!hLyqF4eaGSGo?<ELAa<HHw5742OY
zeBJQOEmt93UTCS&V{fOm(sS*6`(*<5)^{2RrJwOkS?)$7DuO!jNU_Y%7K}z7^fC1j
z|GbO+U%B6_@BP1VZcy>OUS^`uQs$_Ag^c8+X~K;Hl|qBki|U&hw(c{kqHl^{ByFe(
z>{)A=IIZwUnt)Ny+leR5#OMef8Z)J?a0>G<c+{y^ImuP<4~LIw<#L$&e_^C}V4{fQ
zF*@}fL0xUn)QcvBRq)wojXAO@yCt2xv`nWuB#jEq6%pxLzNW_O>VLUIG9efKb5_A<
zla@ZA$DV8x%V|u|SiPS3%8mdx&|N&amHzbH7{hN->V|BRW$SD03P>@M2a7l*9%Z9R
zziH#X@P#7^wP>nBmr>rsOzg4C?}oFh^mgKxI_F-KD9rmfnacIXQmmAW!I3&cYLTa<
zFO(_`Ys9PrU=JY%<A81P+rCVMY3DJGfTb~Ns_3l?9euuqW0i*->(1d6cKb?aF{ku~
zTqDxwa;;tpB_3FE;^+&mQ$R$gUDjEG4Gg?;l*8W*9|;RTM1*!|G>hV^QtX}PvYKt0
zDs7ebip06zG2J$xCO&HGdJVit(P5tatE8f++hoE-0e)5J5{v{(j~P?MBphEVmdK^N
zu{_OXd*@ZSaS0owXWRcsnDpH?$$gzFrG?!N?XAs?i390<);w!g?N|>zIBA<3Vg!fS
z+ZHqrkDb(CqFz5t5feVt4LN$6vEf=<An-E0j*8Noxps)FCf-)F{S4pFMgN{dkUZUW
z9lK9cG6#I?QJyT9X~N(TChl1IR8?)_x-Z(nIAEpYd1dur#q#()WX;oCKL&P<u)#=E
zq78BXMe9VD?7%U<VpguIGGV8#S#eBhM{T0HE)TI0-=d%t_Z{=vrwv>fvt`TRCewGe
zhkZ!Nm4cUkJO!n67(o=1T9EkB>QUDZ*A#OP7M>I`rxa)vrgE8hpA^G)e4wfe?V?pl
zmcN;0X(8*Du(OU9?OeP?qY-5Yze}2tr}*UK&P*G=@OHymvl()Zx@nw`dUoEu@i2{)
zhzBA+1;nAZ481Z{8zc9?`*KbOlz-DLyt>N%$DM_)=&~B-Ozi}kRbt<niat!;o6=#x
z1i464+=FwLny>wxU;2-Wbh+sK3I*R42tb2^SJ-`>mh*&YS*`7|W4|BZo@zIfKFi8%
zWK5DMkh9Ps(z_`|nW7^9fuRp7{_LF9{H~D;etBu?1wJT}9g|Wkx@gxly1N!luV~1L
z&#a}I%cC^srz^FuHJ&zdcpkQ$!)1TXKKhfn?_r5NKs!w%WX}MN%GsRz5~pncQqExq
zT8^lUQm;sy)cBlZTqIQ(gyL^Y=4hSmvzeR{6<B!iPJ8+-ZZl^BL|}E=W^#P%LfgwF
z$L;;u);Cb~v!qxS7KqZXD8JvD_KGX~lty#V`%E*h-aILL%iw!dY7;M%D56Ly{-8}~
z!=^Yg$}GF&{_edR@Zo^NvvdyDv0)0G@@O<UPjAy*?;4VA*IG_94p*kQH3Satg)&)r
zzG8`2UhDBnpD3Bad}EpP#BNnRd+3X`#JFSPO|v9SVAdiug`&A~%vUsqivVwTovFG5
zP#EG5^<)sqn!T{N{>@Kvko6}Mbc!>5OjXG%(CfYvn7k@k*$8Eq{wk|vd*)*S={dw;
z8G9Wg+UZS!9j8BQs0x%!8~fnzJK#P~DQ-QWz@HrnUg0=SxVTGWrgR{?&fpp2>cSKk
ziXyXh+?E|@ouW76RwWnyeCz$nLi2Na&Yvfzdo-gxV6<6Rf(bFoqC5Y!?yllJ?jWOZ
z3jZ$43os_w6-_Lkvb8t=`a!+gLxqmpyT&EiSu;@TwFT^?L>)+Pm5z!m6uB$f@qXVv
zN597cul{2X`I4Q~#%cmY_>ge|j^##sCoDBf@w-^}Q|~i}2;`g99WH6jf<lG%DfBN3
zRqD>4K^D4sk}>J1or`CqnN@~}N+H{NM{iB}VttI+QtT3#xF@@p)k&4r(%Nj0OpVe>
zEoE}pz_5S&VIt3WX$iME9aMEoDpS-NYgcECxW8XZFx9vx*OUu{9K$!HIYFhlZ2N<g
zCB2I%!|5lPF`vt!j$HZjF1X|5ouQ2=+CW9D6MA>s;Nt*eC(nIV)`%eO;3quFxoN>8
zvQ&@%KI4!U%L-U!OxS=bO?6VVJ8&M*{3<Ho5qlAiWff+x5pOiMnO(<w-+d_Tr6<ZQ
z!_-}8P!7#gvx2?<3EDd@;X=~PyeY_as>~?om6R*GHdtrB&#yEYw9FGY&k)2k(IM^_
z+!>)2I~@<_Ec*VR@+kk8PImuMVebFh*>G`@gf#1cx_uaBbz;?Pf$)vHD)Few>qH0s
zb+v*<Kq#+BQBui&Pfl$`+Z=so_k>D!OVFT5N`5%fEtyw1OZ&VZe|(Bfq}WAtAr`!c
z)=vCjxUbN&1+@rB01_byLjI+wqCyOP$=Q6x;oB*$*dHT?_gUR5yfVePBo)ao$Jq;e
zmo2fhJ{M4G_{6Ktl7NG6Be5op_6>6oqEhy3C;o*`i3R<X<ei+1ijSz2hK$O_`Lb(j
z8X~PT&$n_P6^q?vj23-6A|L^f_jhON>pb7}!YnPV)s=>gY?Cf+2ajZ1l~C+|`0zt!
z1-_2Oo5lF`DA%i4SUmhTH?n3spqx1NvuroJ4?!o*Vk3=P9=KB+uO`8))We`+d%=&^
z?g_i#y0N=lF%OFCBe8q5ZA_M%F}<x6>SfFL{q8q?$_Iti$7TB6EVX`*SQyELi8wWW
z8wsk3C<rD3<vq7doonkP+-FG8J;{yB-lzuJD95<U*N!bm=R{J?(lty*P$}J3>9wPI
zUzVixHBQsZL0OjSHGqcyc|ORu#z>2G{-k-$YM(v$Q*`i(j;?+kPmN|J|B1c~)f0H=
zi}2c~jY2yvR&T>-cb0=)@0+O$p6Y_pNssF>vB=Mda|}@x4ZUk3ZX^#!Ypse*n36Vh
zCBxNxya~Y|PwwNB*gZtb6N|o|FUe`Zfkv2GbVd)-ecDd8F%gLx$zRoFbSpgS$BZEt
zqE_<aTpoSNdj2+=jMn2#{c6D2yhTCs&{LFJ*h_S6u&s-S)dHv8O>Lu>lSl7b80;~!
zhe@^~{oJ#yjoJ#kCQs3VE4gq<2{h5d)bpAy_^xIt6O!Fca*9_;FsEdIay%-Bvq%ku
z33cIRpY(|SX{1PN8GbBWdIvr5F@`64oNK^*d8L7SGe$#}isk>>)AQfW`m9|j{(%El
z#O^xB{jKD))d9OTnA>g+ob%4tw*RmqKNVwFzeu!D<@ad0OtxenN}rhM&vMzxera1h
z+1jH=P$`=cySJA5OLpPaKfzTuzbq?nH$N28x0C$xlt<of^}`BX(YmzW0h>d%tJ$o!
zf1ZWJr(Xp7CUj_q6?O!&lp%th!Jd6o48K2B@Od((R1*Uu%e6*I5PO-hknV7@+)Z*|
zCaW)2E+$rOX*4WMV)(E}X-lq1`x<{TM@jo$y>U=6O}=!h({Ud$G#pkOYUIBs2tW8B
zjc5JBtzpFkI-{fiMvy@JF_BucyJMd0pQ7rN)sB7%ovgFdvc?4k;y5>m(w(Yq3>lGV
zj_5FFz`1;YQv5%gI`rBur-t5Bge<DVlJ7fYagqyZ$&6^dByyqj`Nba&A#mYmv_aYE
z&?tYNP;X<g#UaZA8yq4?HX+mA;IgsG#vOD(uUIiR&!e$$^Pde-UuJva+@e;xQiDC;
z&9WGHWy~n$rwAz9N%@>j<r(}SF0{(gHV2HCE-lLz&!@>2N({_sig{wS=HWs+vMqd9
zhj~4Ep<yQbS4PYe>8)7b!)!9=MUjs88`5wTprfgawkt;y8H?G1W;l}gDa)5SeU?6|
zc>IB*nv*OJ&uE~C;?z(mqP0+XwUv7>zUtQBHzBx2`cY-(ud`lQW!>z`4ath|`D|~p
zyHHbHO~4oYZACiR-^iA`{AnDUT3(4FGM%?dNSj`wfG44H%CZL6^dwKp9qUqSU1&JJ
z#ItL^m$;NK=X@gfT`|EbE4#$8{eRi3xhfX@TZ%<QwI><fysNGx6RhYvq!W5Cnq>aK
zVKUn`8gQqyYR}^9x8$HnpD3xfS?qy-ap8FbwCD>U^ecJ2#U|uZbkZZq74y&LSzlnr
z;cX3frbrln_n5|-P9+pN&_xTNN^sRL2{=!f;d$UDk5*r?35Oy~DL5cO4q_GA0AJNL
zY1~`FQr*MA9VTS*&WfKjXNcSId^IKiF5U@RQDfq;LxIk>fa=G4kTpeGMi~Uj4moIr
zEovc9vEWnW*(gxlXd9w%hXcu1z~?`3hQO1tSn~IO;H2RF9xXVMz?W1x<4LinP@Khq
z(3`o%n2q4%9@!-O&pzAaEP2%=<LLg<n~DOO5CKd@;DGnw*@1cdR*@C(l_aeO?~26G
zl5D#Q(RH1&n60i|8nX)P71a{_6^)v7JH)REnD|O~<h)*LcPz(b@hch~e(Rqu_QeLd
z=(V)iPMz%&?Zo;59__$jO|MK*)6T^oVFv}HZtyi#c>ZBg1CQ*J#0@n;Ud8-*<;gca
z*^kF0F4s8FIU0tXU2j7h$(pS);A#Kp_=@-`+@=&!afmN@Mn$%>4;*_?`L=BDZ|=TS
zSt+Jwv|;32=(cBNy-qqDWiU0izcg__y|LDTl|15mdy|tMhS(wBZL?Z|e|B>C8`}@*
zH>6*q^#?T-AMt!D9zU9QRmRSa?BDn74)`IbuF$OeLl>y@U`&@m#BM1b=lL}TtDg!f
z64w1q=IG-hw3D20{T8aW*i@91)Y3i`6|vyVTsPM|1eAC=i=P3>$KThQ|7AIS{)em?
zn;WDPMNXf@G&|A@oi(qCeaZ0cht|h=Yf~*TfyX7i@VoG3cqgixM(v#Kk1PO)5}lDH
zt#W(@zuv!N*CQOK<as=v%|rhl%UoYAvo7!rXEm|v?S+c1qi<y!_U>rYd5?yTAjafa
zip$kR0O{yp0!1;%qUBOQc+VF7VuO?W(a9;o)YWX(^Hs@|Bpc(ic0E#6JG^f6{D=lR
z&}v2`QyQ~5r*HIV9+3?A#?F?`L1`Y~)~NLD4??+K!Qws@xQd_gP%m~`O_S%0ZSg2E
zU^vbzK;89hO7G_Bxj4#S11Zu~fAilHQ2ieXUNnXr7>oGt50-cO-`^c3TDNaub@T2p
ztR2m2k!$)KcTIryngqr4<YaYrSMjE(4#iK|cYzGnZFy^|$DkLYGU&muW$1R8l*pOH
zmo&(Or&B+4z};#D^KKpO0*2G^q7G%8VHf%=VkP!S`!)&0_3fxaqs8T?^|B2(GMaAd
zn<mZrJ&9So!mpEKhY$Asz46AI@Y9v;^Nd)F1yB8iSOZpQ*P3s_c@(;2^$8odfTvB*
z@894Pi8@rO3|HKES7<EWpSF5V9~L3*7}>mEplPwax43KI&)ylu)I?!e<(v;yW|mZ2
zl0t;jXf_+$KtGKkZ(J7zFAmwaHy7w@-Xbo0ta@|bR7qRCzLhvd72n_h6h*HV>Jo92
zwn}!O04(6Zu>7PW=iz{vr5wGDZs(ACMxDL?nMq|1M(%d;56N0S$MU*Wni{p@_Yp4F
zU5W8j==(c(7#@&xOzV`#jlO*jR2B+<#}|JeS>a|{7OO!!zz?Mus_pc*j!$0V%z@8~
zjpFF$zAa7hK9}Su>8~5eY|N-PWVC^5;sHNMYB#f0W<tgQg}Ky~if0p!9g2pJ@d4G6
zLA`(S%m4llUH$w2d+`6e^T3T;24N1s25Zn9H2%ehsA4DKZ6qwnNecsjw4~!f2AuOx
z$;5xlHILsuZESbJ<BJZhy2Y0$Dg{i^bL^C{;I2c`<R<l=O=@HlFF0IM)FoGt^1bHk
zmkz-_H`Dv%ze6;8f}8YnFzIYw;>b68gOgnj5E^NU-j&Uw^{P)#1!`u9o#(a{Q81*+
zqR;aa8r=cb7}Tp5aVzPiYDMG53I<j%;^8)wT41pMA*~Sd-HEi`>euyO#ao=knyWg`
z0x31L{?t?c_he2AIE>}Wvdof6E0cg$lTG`CkXQGpV1srzG7LsoQi`-oEwFC6T^*&9
zyxw4+bP~)aC$4Kel{dg)<~#VPs5v=z$YD{L<oub5Yup-tQuxl0Fn76$R<x-ECQ}|=
zluDUV@0;z*`jKEYij9hAgF!Y%LNDHOQ6P_OA@|M*;XzFzgO1Q{by$+q$w`#lT6J-g
zEahi%uBY2Zt5kxSgnPb<nKeOMT-_QSTGbu*BY?*J&8lLT)1_s#g;IpUN(!|y(p6ZS
z=0$Es<!*JMNBFfl(7E;pV{@Z?5~;#SjZdy}wQ8@%HDv#VBUe{p{I$QC^uG<o8M8Tg
zv;Kh-1<W_RZzx;r6wF6`0c^hq$Af>D^-^RhvKri>qBJkx;E_;7GbNQOIwpGk(0%hG
z@MHH3PRe*m%{9nvw?bu9@t6AXl9+jkn_D((2?d7R)DISB71(M&*~v3@y@SwTDQi19
zZF8!=yle?rs;r?(7;Uh}!6;*IrIHzL!fax;`@kR0J&1b6>Z>OGM|)DXA(dw~DblvD
zdAe%G`igScu8X~MRQ0aj(db!~3n6}>5nkRaw8fpN`k*mN-L|<M&RKjN_`LtVfuIaP
zxSOZdo6}9XJ7MuwF~2Q`OHGUnRTGN*I?q>R;bPv(Xhpt?mX3v-QOT;U$bz!5^a59w
z;Pbrn`G)u&_MAwowfihCUb&=I{{9;_q#ArlDh-D0+KpDz&xTq_BS`^uN-zl1p@#9z
zBYmR3yX#M9SL@Z=3u#wtex9>DuK9z!d1ilHqxG0lrxP$6&;Q{r|GwM4A6p{{Hmovy
zgm$WXi@)hZ+)^%~@5eO#&~j59Fy|?-5^X8-(iGcsNy~zf1iC6$EnPv&QWZjC{(YLM
zW;=r|brU`f6s%z$I`}(d77hOtrlleK^2)pY>pa;1%)|ZXcr^jSCHP}m7gy47Ug6F{
z66J}nOPYmj<8*TKcfvZjcA&jjNw*g!Q?|C|qn3#Zt~X+4Ky*hX2xMez;YJEoZOz20
zK8MDyNXN|KNhRBX?k+>+6kCoqsW)mNQF(27Qq5_x1Nzf@EF3w4CR;v5HiSPO<D?oG
zaJ47qdbm);zLKTX$@s_Y@;|0gaXU`mmz$80kSc6NB5d1~)KoF2W8LKKU7APR3pE@0
zHe0Q3rLA7dU9|~QXq%#SweT3v)zjMbkA%DM<I-$_M`ppU@$jTb)sgY3j~DissA4zG
zz3w@sz)Th7!rOP9;Km*^y+3enG9A>LQVbu{^MKmi=NCgHC7>Q3Y>%yfsw<h%ofP*^
z_Rbr9ETldeo2@Z(l#04nIPz4n@PTNtryLozOB)tdY{mboj!yn&U4!Gdk*|ffpfPTG
zeaPVvEvQp{MtzvEyW4}(^RAO^L+XaE=0R!Ohz+36MG^eeLa6UhljkNK{=NA6zW?Z>
z`)~XE{?n2+^#Kq2>-m(GU3W@PG``-1$h;u&lCO57iSnkyAha-{r1KH|2L|@!@{YT4
z!C#|K9rVUy;VL|FR8BJ&+Q=ich`(ztT~@92$G{j7=KNA2j1($Z8>L$Akhi!%+jc)=
zM)jG0n%TZd?VDJ6Ya%0ormZvy{wFpzc3&;n*!kYX`2jxZceUf+=ac^F>p#=+MtWIW
zW#NDE{7^TQiOMr(q9ZY^RT1Jn$UDSwGJxDH^_0nTlcPIYJH_sB_g2?L24Jb?D}pqO
z1hQWGKW;C<S#s9Ucl5Y8VLU8r!DKWMccxM|y|6$(J(nJA>o;r)+4%u0raSMbfiBq2
z|E!JpsS+dL*SF#DErW`3PB~3zgpE?ekl>$=lyhmre-Yx(#=sw2&uov)?do1Gg9dAg
zg*`l4@NDhnN|fMhDNGa-R(&t5_I&Z9zh##^nbN7NhVbm0?$s(yRii02Up^NRJ!#av
zjyi~(P2$RVtDzy|LiK$wv?IzUzCIBuX9XN4^MBjt;NMgD_!qoZ;2(;%&I`;yp!o2p
z<V~|DKWfk>708~crr-SQk!m7_c3nQsN0PptQX4C-@oY=cHE!#94ssuK(wdkaZLE+o
zG>F;bW#8S=Hrah`cFWAS@s2M1f@w<fmBWg1=fu5uLq*z90~udBy3OqY500QC-_nhm
z+JzTSES`U|OPjzHFZ(-03P0(&uVp4I5GA4c+_fg)D^nAYl;_`-rF`*k&J;8LH(6#u
z?U`{-K5gXQ^<<fEKe#6-i>{Lk8Q8{(%$K!)A<8VHTJ@&Epvmf22%SINpLzX3MCRT#
zwUrH?$z&Z_?pA?lKT;=m(~f3h`Kg?V|H0mShc(r%d&0553W$X!2vJZ$5K*eqq9Pz5
zAiWdmy|>U38z5b}f)eSSNbiIW3W(H5htLTnp$3vHX1#mn?D^)L^X=K^oY`}I=bQP1
z3qn{AE9-gI{oMDj+-u#5`R+WMjChpa>#@@0FE|1ADy^*aZvkd=suIoYw)I0dAop83
zBrEiZxX`r#z3i;@ca4^g522H?2?0N-4e~aiCyhzxmBv6A-|^wf95?%EfmrOMScM&)
zEgP?~P_<$_$=oM-FR%8tQbtvj<tO(DHh=AewpK5}y5muLq#}RVMdkw=<0DA9Krq5U
zQGeAjpSK1p|JcQ!;m5*N_SEVz2L=28dt?1i@|%CaCKc7L@t&jW3LW9usba=Cu??)*
z>=o-@Na7L1v~`o;?=>NLb{#2cza8)l=t*qq%n7Bx#ZZGhFJUv#@blS-{wSwh#(ABP
z(<O^K=0gU$3lrP0*AhtXlvZm&`^i|QhrvpvHypRz1qSVypXBPWY5Y1rRN)$ZN3%Mm
zt{S}Gzk<8d-!L`($KE>^XS=7z`eE|B3x{$050xl+Yw~paeS}1-k6-2w<Boe_Qk4(R
zj8)lJrsa&=S7Mxe&KW<NJ9Xm<v!LzT88nrfT#n?q*(SkaJ+%S-_;E|?Ope%!Sm><F
zX**Mfs$@>#=^qBhb@vyd{A~tG>P<*S@<i(ro17`7T7%y%{HS4}JP`Z;DcAz5w~{dp
zv6Gjx@jajB#Mxf|M&I&^p6NU7;s4k0ajct1!M#v#*?zaN#>1uA<$fT8O}_@80{jxY
zlcT3Da3d0nyRN?PThMQxIOrMAxL;i?8=-gSPxFFd?#~tddtDEiD%!G!`e~!aj73Hv
z#<d|QT+M)ApxNnWe==xQ_0&Cyyv4KGicTwVMZ6hPZ+vbuo!p|$ub8H+lpL>PyG!Sb
zT+6rjiZRMKu_G22Zj@du8OE2#_8+R^|Kk<TzgG49fAM$!D=V>l$&nJLuXbf{Q&`eT
zxjXipMr~uxp`sV2Ocd$k#D{dvJ`YD~J+H)-e=oYLt>Pl-1e4R8c4}G(I7Kp2A+7iJ
zHsso7zo+KX^!>t?y`s;pAg?8lIZz&K+^BNuPvm5HiC~;HO_;fzlZq3bc-D3h*>T3q
zB<m^DXQ%sh@W&+`-@ZP5kzap`^Y)O2;roA{ZvCTP&jCCEVkP8|oI6@|k3FD=c}Z1O
z1I1*<X78ypztX?GT|}0`_EQ=>J-=tzUH8=e-IIBG2VXWHEg_*XIo7UUQE&w__5H4e
zL*^CRvaW)P8C~ev02OH4tdAgt77zQ3yGC0$vw?3Q>DlN0biJRP>>iKtdTA#Z1iFYc
zKm*KQAdgdewm-Gk#d2&gy{B2!r+5hlp1SVuY)IF$M><U$K;l1eWGs-u_@|#9eP$H7
zAhEWgy?`;r?+D_FqbFj!dM&hSa2f`ce)Fts=L4#C#J(063uoTjE57nXW8m4S=pLs$
zlQ#W7>H_{}9`~=_u7SUbvI`ifa+2t?G4f1$r5@k}J^K5<bWN~fqIE^)s&_fM<{KZ!
zv40a2EdFW{@Pr(l695~0hg0mWLG5}I_9Nj$^@zfHm$cU0(dur-DS4^}T22RfEdW+R
zew{%JTRsrEKFQotY{0mS#~{>OBz@kBR+J2;k3vt4OPA~x(f|~9dDSOTB<n!h2wG2t
zFF1Y94(-w#Hk2MVe53LPI3+Lt?b+?Wy>0`A(~Z*Pn>mPX<Eo(%ITxXi3_qe`g-{(0
z{9&PwRSo4-i^7tfzc{3=lo*aHd^2pSKlaNnSye*6WaN@!No~}Mg?W_4!?D&}F|GDm
z`Kg9kN*#wltA!i?hnwdZUFyh(xQ_wMWRoH?Ouw;ysMvt#yL=GbHSWx&wz6hKyyh=E
z0h6-P+zeuR#ozUjnQ!=9Y6XfF)Uy=j1$e7jEk^Fnb#AI0K^pWdNCKeY_!$m3p<k}6
znW1Ks9-k~}8zAQ*0BO$HcTVt=TE0PvEeK?A^E|GX#zO+PH}gGjL2Vl!dAASttSm>6
z)de0xuj$)1txlli>D#Tpn0<#TG`u7Y1^5{+jDLT1pm6ABc-xKX<2ZL^AcKeU=ZsoW
zO?7i+zVx#HyV?Eg8}`3<bIWR$IpvPN+<0>teOCGk3h#2kINV9HpYC!@(lr~~eaw#|
zNCb?iE5t<(H%f(`HZ&rmltO>B3NQUk2-Ls&T3@?}-9Le?#u3?|0gs0sN6i$b@(t|P
zsY>C`qfh5$_RLxh@$`-H^tno!V>PHe_;ry^S<cdt9sh*d!Hr)^SHnC$*bNr2r<@3n
zT5JcggG)@BLA({WpA^L2ld*%0J!;yyMq}_ueSpgvph^YaMY}l*X$ost1t3j#=isM+
z+a@Rn)9i`@!y|}Itjc%kk9-qEY^1DVY;45KNy@Fewpj%V`|U38B(b4X9V`DFW2t5y
zv;EKS-F`ZZ>@W!KIB}2lE;)pRihp98PhP+P4-g!xC;RYH%$+nb<~~#8Fr2>dHE&yG
z=|$@OA7_gXABs!5yXQP{K7vfzNSDO3#O5|`kK|&G55b)Jg8dLTOpT2#07MXMq^53~
z%m;5RaD8g%T_Lb1t2}prD^U;HuEn0dgmU|SXYdrOjSlohKm6({oTyQB(fy7nU8s!o
zPUn_bH~a|F{H&tFn!*|+iZh+^cmPYo=q4Rs+VyT7AI>`bJou9+y8nfCS#~q4c?Z)c
z&`0LOL~fKVAW+JcN04IY-PeH>-vEW$zW9+n>~0U5VEP4i-eWXCBkTxLm9*^9{r;ZD
zXw?=>MI0@sG?>>p2Tx3bRob*W1WN@diC)`#0c$DUK-o`+u^MbkYk8uCh?pm`SFADk
zf>Yb5c@I?KM~8q*R9FwqM3{Rw=FD)A#QKG(vKHA=Au?`05Vy`iI=#&ezchK364q5s
zuAAe??k4P89zn1s--uW?K>E4f_mAD7d6?i7l*k_QW>Ll!hg=AF;st=1sj81^^d@{f
znqrTggGZGHD3G4kOctA<$()a-9*|DZ)=ig>AWgkqiY;SU2PQ0$6|j`O6|D>prCjC>
z7u5&$avWjnUymT{7~W`UDd^Pjx+0Yx6^WF>hE;r$-OF-C%~+=e`fJ2(!qukte&L=5
z2u6>!qvo;1=UHt7D=H+I**rOE))v^SvWLx>r@MruUNo)U%{_v=4x{=wDQ~DZc-%j%
z$sOD9CWwXrryaazoz7S196?@TwuO%%1_h|Tiv6-Z0f43R?Ga=dW_JV$)P#noGH?bw
z{_|(OISg8ex*G>jL{lBn+av0fwP947U@PhuC`w}ntWbr+u%$6d-)}V8Xo~&_k`_D&
z9A+Isg6d{kk02rp7}_p8D%BFbJpdlrzX%F)Wk71Lu@^=Ts)iq)O9GC+rCEbDjx25p
z*@0>qBigOt>C`E!pJfHeGPMqS=>1=qP9bsi0|GB?&rZIrHFE$pZJk7YG9f`b1;+Pd
z@I@#=Of4Qfyc5+jKun^YDMD?QXmye}T1DKGhYL_m?MU&!(izWG@C1hP3QRlqofn~s
zVvH?gGWXU0GJrDPzG@_yGD^9VD)!5gQd?AV1gU&WwQob-9I&P&N7kXn<-r-E+QKi{
zm(V7k%%H~Qv^wWF-GyrMjWJYuoOHb)V5|p#r*m@BO@shWinYa1;UXm|as3G5q(`-f
zmxGEu*ah@oR!QW2>SNL!xQAgKoXlzX7LN%p`ptwZ?tqOK+aZkLYk2UY96@@pyUDO^
z0tuWcWan8P=U4?~)a_P=UvGEB4ldeJ9+81=F1^sZNnH4fb(B3ebPusb)$pu7L<PC_
zwbHi-FwX$Hha8=uwt=~@CYsrp|68SAq|qoa24@yDc%FJKah`ED?kbQls1Hn|FF2AG
z$44iRAUFkQ6#q=#Y^4`tPzxNk{cP&0{JX>zOdYjvfL~yS4obd#CfuxL0%7zl!N2lA
z#Kbx72y*Ibd7R50I)q*kJNg>dgNrkIMShb+t;S}?(bRI6K3<XX7@zMe&&4F}96`u@
zF0F(J!JjFGPk@4R>wp)|4poV6m3dCTDm-QbAjOMehNJ@5MlZ#=mc9+_ytkrAbML5*
zVx5MmMGi24J};)A^R{hIKH7q>IJ*<C7gTZcG=pFJ5oA3V!&MBA$e<buo0iyUyNTKd
zh^Q}mmg}gPn78ahZR&$x`Z)#+_r9Z_3!AlTA2Z}Uo9B3wN#hyF4*l(ZH>m=>vJV}`
z_M&Hp*+m&k<tXFtR#QJ>RI*ji87o*MoXGIfy~2DsTGa5l-RNhb<^~JH5ov*|$!%5!
zh$w~C?%&Cq_BZtli-=)7VPeY}uTJ?|e<E#l>~{><_Bpo<qr1bbt|qSZRoXvH5RL!1
zo7h>Xq3{?I^*H@6*&gJdl5qa^#x$9{9v3oKf@xS{e8NXgX7qc_=Xd5TQ2rGTio(xp
z6zBy)^2}-zCl)369l{X$4HjwnxrbjWm|xrAx$l(~EA9t5KK5m3c@xE0;i?wDvJn{{
z#|suK-2X`Y<oG`joB4-~{U1>w1hi6IkiSQ;rJz37ymIg>>(t#Lg*&q@yKd*Rra!-q
zn48d6RzA082bWvvjNxhWy?A!{$}7Ng6FE!;?~K8{K63k9q`m5s{f@B+l=$-3(D)m2
zzuc&VW%#H%aL#HDc8*-2{Hu|BS``mxoanmd@Esy?f#t!I`>)F{iL4vPw9D7AOFGj0
z(+gI{f*Ol*#^Hv_Wrh<SH^bx~(&wZVFppV+#lyXGB2{rs`|0(qa~)}T8Ca|R7<Y_w
zZ|HpZE%_aq!$Xxs!AOBRu20hDc-DwAKDR3Os~2~~W~ZQJ*8COh#T!FeLzYHt(>(_L
z?|nsnP8@oRO$Kn(<2I$C7jbi={G`jWKBb3M4L&+oOY-ACZp?Xt|I_);-aY@jQI=z~
zFphd1{3LHXdjPwoW~+4}`>p63&qrsjdLQCS$=>)q!qc*p7ZLIZ(p&CmW1W@jQBIAP
z_7t^+36Zx3Q@2Q`ut7UgMQWR}&jqM5eP;T6)e^UM`1h4=*lU>D5HWK%P#5kNJ1$jF
zWCrfr?lH!9j?3<@BxG7<7rz@c1O;IHZU3fPZ6PLOK|gOVyZ;#-(LwLU%QI83vBmNt
zu%Qs-=93&;g~yQJ$fdNuR4`If{>yiREk3Z981t8CHFIN(9Q4G=996k%!{4QOD0han
zH@xC?Wiap4KOEzacX3Y0V`%fC`#AAnPpp&wPah*w-0#@`>tp;Mq4K|zS9)fkSCD)7
zjR{h&>$UX0ya)!f8xoxD-2bd6RzS5Uss(3Zq~+|`IFXJxRq4CjWflC+rlDn5baNQP
zlmdQ^g^>h}_PWSMS$X4_cElWD(Eb@nuRnqq9YNlNA_UZJRAf$QW1R;Sxlhp?rdo*|
zUZDUB3rfu8XD8FPG$$-99}E^et79*+YUu%HqR)HSOom~73icZn=5ls@Z-{{!>1x=y
zK1<@&rDvtccEh&UIqt#?50$yNY$_@~TQC}+VwTil%?c~xfZ~eB3AS$P+u`c`($z%q
zgKd|hjL!W9ur@MGT7OL9)xL9+B#6%gnB!j?r)H59+mp+KioFOcX9<)+_+!NuswFn^
zeb|ob=gW9Qm6>c{dki&GguNsrrou=^mXNPzP4;XByC7NxPCvNm>IP390q91Yqb~hW
zCtqj}l*xq~ua@1Ziu>-x9dSB-gF6kQhJHs@Qb}&)SMR^!(a4vdkfs{A)R%Y!IZKsy
zdn2f!NYOk7PlJOA+PW32N&TJEEb8J)QGv11Gue)U+YsIx5j52RQn%+x+G+b|PP|`U
zMP0dicv5V0HbA7y6_XsWwVQyw5p}L!F)FvP2r3Ww$o?9J&15Et9Iuq6H?kNV+xA%r
z!+S-O`B>YevSFhHhnJ0!m%h^kmQqls9jZQVc@}(<jY*y!M4ly&>za%JTy<uvdd%yc
z!4E7>73Q`AQ(#lao$Q~ttqmfb9!__wcWrp7kBi`gi|Tqu%%6jnjR(xA1b0@@msane
zD<zuBU3;pI=L77EF>z^adTx{1RK+Q0*}eOY;(6f*G>883=-<5fTbWD~GR2?OH6HLn
z6W}A&TE`nkHKKC_sKx4KJq6u5;?xa9<Wd`kX%!N>+uhj*pOzqx2sp661<O9QnqxH&
z5^87;1QRLXEEXTkk4xnr(2>f&lY&j~_EV=`PEvS;9YbEL{;*FGu{a1@LSJzmVV%iG
zlfj-G-`P9)diYiJVJWf~Vz+!zBkDXykox(dlWFsFVuz0{z=ZdJ>}CC<FR)NeQR7Db
zW12wTAPPJ%*bgaT7wHjwu*QT-(Q1oGL=%UmX>~3`%h8N;H)sT*PY2HV92>8Y36lB~
z-}eQxQ2Q^EX=dfLB!GSNf(5ZXRng-@5k=??aCVS1PxfwiJAyn9@=cI_(S(_uou?Tp
zK1Vp$A$R%@7`a07RQE***@sW2XAQ-tj**yWP7AV=w!s68r~<nH<1w=H<fCZ0Qh4SQ
zjNB|uc$}t&c7alI1$y%ENmNP;Z41n?Jr#pZXsg)Q2Xrn(pAH}W`()%Uc0CRiWN!fU
zA!Qq0=FJkms_%b@UV#_8J^g?xkXPTAS6P%O^4%uMu><s?dc9wutGz-8WMFP9xo=OA
zoZ=rXBzFQf9e?h31uX=5$K-XHV&xfHK_>Gds(L{0B=vdkbh8@W{Pe1)V#vJ$Usx+O
zy4L~zn-w1sadI#5L)5%+TDvm{6f{dyZLV=4GhB}CwW17Z4C&Q5cyWK3&#alK-m+)J
zz4KC&U_h9Go&*C(^Fy)k_m>dKe8>+cXbx)Az*&qdS(v=SeW<^5h?|JlW2&aSzggjl
zUH=W_t{8Bg3XE<?Gbw`eI2?)une-F1JAWKOM)TLP)XV-V(KMg%G5EgIGoWHmhq>?W
z1N>Q+P!yRSb#MV_?cPM$kg?mcPs71Enr){5*TJQ~9s?aI>I8=)65q@ri=+U0SW%#v
zyho#OA6OLL*QCO_My=SGk05t(_<%wzQ5}9#$cPdQR;v=^p@Kh`6Y(0s%N;f21c>e4
zWJVuOreY7mlcwN%x^zGw&%!3zXZjUtsR2i$x{yG9Xh<NL4nR7iHs|z)v^BvPDl&aY
zhn|9tB7_chq7%n~v)eNw1FQ3>rg5tMX)wV0A#OKrWVm9(QIUaq@?!0yt4ENJHFOk#
zQ(z4mvt@}Us)50#YGTA%3m`xV+QqP$&$TDftMw{3;QD95>u`#>5YY{b`kjY_&dYr#
z3QwVz0Gsv!@Y0Jrj?2O$D;fQ{QDA={;A}yM1!*;oHvJ6>@HbAT*IINM%Rvu9!Mn=r
z2-5ZGPP7nwYm<ap47d(1?vvBQJbiKa<;BCXxx*3pCF<p>_L#SOf#zrX9KXT|@UKI9
z+R~5u<{GqV=2(@>cfwBymEwA+qYboIz^&HnEF*hWbgRx69k{pmILR2B*h3jGMtZL^
zAXEF3H$y@nP9SfdIf6K2foF?H5WRbW3;UQY*li&!2y=uQA_IE=sD#yAvtM#Xo<wZm
zq^J)OZnWROk08o4)I^}|He1By7}Frw<!v^q=QUs#u#v8|e~VmKi0-Jv1OfbIM-Y=V
z(;Qlf3$ZXfnqhUM@M*HuSPX#<6@>hynmuNG1SzTrQ{PV#L&M%sVIF7+g)(=)jrV#v
zD($yHaCSmyQiyy;y4nWrj+sFUD_MAP1$5iR(U!aN*UMd985@voaOxGJY33j3))B!}
zY%O!oH1Y?QewSgKD&fEZ#XoBjBlMHxP6*+D#qN|nv}*Ml+_#SYzC7K%z+twl6;{Ji
zf;!vdgS&qOsS{u%z_TPf*Mx{TT}>9L@jGBpt^<2gD&s3KWM0Lfc`aH(F}!WEarhd~
z4)dBOiIKY~OOfxeZ9Y*hDsnCE0DVb6K$)bF2&o*|v^bHonL>5Q??>6Gzc-|RU9@|E
z=FIMn9>UewA3=y}(90brVA0ED=fbueE3g5}a%JCNlOahcnhTci4l5RY>)aMw>6D%g
zXr1)3t2=@SS1rbL5e*_wAq|Ao$)_5idqnb=?9w~?iB+yb+TyEcso<UfBi}wXe_cz5
z7&eK<C)}+>e#Cs<MUPJ7S>CWsA?!EqH6YF$;&}XLCOy>p#ysmjJ|fhLC>u+)9E8=|
zX`yDlk01kv>^&k^rF=d&`T;h}ydhxx_zGj(Jq2HXR_s|Nqhn#`Mt*G?GkXLX3!m)_
zWmc`Mk9j-z(*)gvn<vO~Ut*Wg+Q`lr6G`2j#Y>C*;3p_}g!GD$F+&?UZTi(zJ6H@w
zCX2k?TKnUzP2uPdM)heS^2mh5MPm>8lHMbT@o$MOOnW%=qJ6UIsq8d2=QHYpUB#_W
z$p>?NjO#2&o`y#a$Kr9__<+0l6|Lr4yYP8GZE^yt`v=McX1bTKdViNG;UHrcGhT;>
zg$!$2Smoue;%t9*?qW=dSSpi`>Q=wB+-yO1Jkhe0FIR@B8W1ehO5LaPIcJ{^5a4ko
z1kJv0I-+rI;K+e#@(5Hg$`KQ6dgEEE16_6x?Rqs#Vp(8tC;Gi92TsGVQ@;}J^XRHx
zhq-$_l41Pf6OG;HyDRuyUEu2A!50RmaPDVcGYZ}gI-F23H5>WoDL)UN)wPH>_vGCZ
zQS_d4LKkH#IFo64wI}u;@3wozYggw~U#Z=H4>o9`AEo8IsMhE1I)OvmS6HB2@(i=u
z%fs)e313lj=tBxqMQTY45t{lFrN%9&p^^hOsQq%|RkD0N$87Z-t1mezSKK?VK_46B
zsh|^z&JExs!){BQ|8#O-Bs1ur3$Q=OO55<Yt5=AJJ#sK0ztMLsjlLh~KQdgIb?Vmv
z6fM{8wv!h2jo;NF5n-$1#K0pSwU}6Ar0VSNcv{xnkA1vs<XUZ~6bGfqG*J2S>(XSa
z!`mXshD&$;39R@(`TO6;Ji1DLkj-Jfio0jjF`(i`Z<}$WJPcOEvBs?|5IoZO5OJn;
zpxiJd%&};}G*=1sqq@jr?>pHj<H;uLd+s4i_<@znXLOGqMv=Ftgk|r6FBm(#0Plyq
z5v_yd{F~c~RT}VX1Q4n3qkp=|^<_3t?MeMvOT>>Vg0XH6h}f}@jQi~Qgsf)(n`4(@
zUATe%8AqUs+PZaEj@9~^>hn5!9Gh-4^;&6JMA^b>zTnuZJklQVc8^SsD^{?Gk`Yli
z?RB^;fb_*Y9D>=YhbG5I^vyA7-n0OUA9iWLABPb8#B(u*nb)^nwiHjP@wBv4Vq7+0
zhV!*=E(CnXN7I&$=l0^AE{)T?D{|z(?c;6?>Jz5Rp)Z0<0(;UJQJ@{ql~&9s);H~N
zm<qzfP@jEP*lz!KjD2)a1)Nl(<fm{1Nm0B?{$w<f1>)B%6g=oy`spmKlKs+aYa<35
zgX!2L^WDnI3(%*f?{ccBepofM6fZGq=9cI>p15)l7Se56og?()o<RQ*<h{=8*b3w4
z=yVlo5A?FG8ZkTi3+pWAJx5j&AkA+HdV^)JI5zRbP?{J58S19ulE-A3EW7aqTAOfc
zD3qWelxaaef_&OfTC1WMAf%eH#{&eul3VptIbBrCqsE?I5dC-#)&O>N0><HI^#F^c
zGr%Ml_Ad6sf#&uyihm<I*g#sRAxpghwRH<d;1W5o*oSSyG~)I~$jZ~(oYm01<zY!+
z6nzew-5>#+5JTz%OfRS+KP0vD_A$Uy*oZV-zjFh%h>Vk?xFhbmum;Bc^y)`0*Ydfm
z7f%gOV%C+2wFl2_;070K3u9bSlGN19F##w!_zGNk5KPH1yK&6a5-O^PhAbg(&Tz~W
zoOxe3<W0Z&6<L9GAkno{g@3G9$(*AVbTtEB<o1$4KAm%;SJ~&#LDhOh^Gxje&=Ex2
ziR)H=BflNBZ>*ZCn%)Z(7!HCtpZf~qZh_mhVV7{yiw8##>qibG`@S~zRCIKiJ?jOF
z>`v~=B7h&!(d)(-TEQ2R{!`w!LRt`<^1CG}OHeZJQ^Y5u=`8T_Ho+|Q829xhel31k
zfSE)d=*{gwxn!@8&B)Jj8n_Lzp04!7j)mjBllV|%6*0Q6uwcA-OZ@>E{!!kP?L90`
z!dOMINP-M4Bnu3g$701q(c^P?>)5YM1Di4%X4bHuwGL1+UuOgk>eTNE-Xq2LVyL}{
z3TeqcJFNHpq)hA#X9=GE@?^yp?6IvFaX6aF<UMPE9G&f*!^T2@iUIvUX|Mp1raEG3
zzPiWNnAcQ7PhJ!Cu{vP3xr{sba`ERkVT$*G$QrxAk__G6CfTi!KP9nxAawPv{#B&T
zZv$Wie$M3{Yx{`3x-|p9j31d@n|fRE986G%EJqN5FCTjDB{0wF1Rg;a{T7&)Mo{If
zt&xx4G~)@Z*2^`pSrg|Zd?eHm@_W@A2iY0F2JOKm=$#2oquOgm=R0gGJ{Qo;T|dja
z+@B|Zi%e0uc=#;maA4ls3Z}r^<;<jskNx}n>wkwp_g@H`-|YQhP#95h_?T#WgUu4;
z)%|WC>qW@T?nTeUd5(ECc2>JDq{JRuV=x-7x5`-dODGCj$IE-Uy0If=Xae^2a6GiB
z8_R@f5!4)`wOxK7a>k<e9;|Lf|0glJI!xC^YqXjE>zga;PZ$jii?a3aCgqG9-ihRI
zWqY)BHSYD$aH@>QPEpLiTWFipDbnU1p|@-ouep1|5Z5Lm_nQ0Q?N(8^iQyFC7~H{!
zq1|0nPn1YZ9Am#-?(=9VrdQCNrM9BIz?Fid6b)25QXkV#-@(?oY;q+uOM`76hlW)}
z48ttlZ{%NLyC>5wTci390ki*8f$st#WE@^qs|7~yHEE$}W%6VIDlAN>_BxT}X|~*b
z*^udaeX=s(cV#F!=gdRL3Q5&K6$7G3pIE1{eWEUkMDnGF?)Eofy0Og2->}Q|Y8zGT
z<3^EfPF1I*DD2ylM);g?v3eTDv=Nor11mxvye3v%O;6;mw^&XA0kv6zn-~YMk%G)2
zh!>EwS|qRC+HPaH@EYg2DiDf8IlzMbj+Z2v7&SD=e==a)#{3`@Y3}U<vO>IQptj5C
zVE7TM^s(f_9#{~W4{Rw?pS2td5}-YKZ3)xOC$X1z;B8Ri3k+m5gMqtSv;L!Z^;>gv
z8~YZ`12_l6Jn!R5nN?x4=<60xqhRmob&^IWR9<~hK7wqdi4|GQ1fnz{<94j>OWD_5
zh?tDDCaTn1y*%|4swOHcTGDq!wqW1N<*C0~1;w4JI&4xO`Ph@30vL3hU>wprQ3U*&
zlEtM+$X=@kk&YSdWX82=(&e@kWc%!Bl%MZA-ffH=DAir#Ac<&yK-n6!XLg_>GYw#m
z97?)_+_4ASr?IW5;g3f5+b~r(XzKJs(oPur8j5Lwmw2EhcPZDRok*h{l(uxMvA-WR
zladr!s;b^0K8p(bV_hHGC$?+|qh4NW^0v<g7Pqz@Pd)r}X2Dntu^Hl;Ka|r8uwh3!
zpSl~j6gX>C6~G==3AyvPH$=AM4&UEMqP6!Tb8XS3imibcwQoi4MPj#q2LtAHf<iOX
zZ*~F#lUn!eqV=`G;*ID)M!$+C+A5n?DtbO}uB(zBfZe<!O^4@HZ5g-4mB!J2t9r5f
ztnHKNw8!TWAVlSWy8NgKiNK6Fpw6#|Beq2A9=J_jf8p>1Hc|W=teqRfn=oh$KI-NX
z<OF?Ef#YTfee3!VCFbCD2G9m&s`#39r_ky35d@w01EsuK4+Cp>0?DKUR=MjhB?K>3
zz5$qHABD55DrI8OZHUNJUEQ}4GCUakPqHg;UfjLz01aJ%5!D#a%av@_8=<7m8(w!#
zK}qABvVayigQFg+_yC;J1+D-dxV#4Zp;36*gkon{(vR+~XFJyuUgp(xBvavwh<&hj
zvz!i2<**j7J!u(-{_+uAbtYNp7T6j6pe<1jZGm6w%p?7#Hy!&pO{R)`VFn_l4-791
z+=IKJ9cF2Fn^A5#<4@1fgyF=TRln!#i79S&3!RsGf_xNST{N{J5ll$8;&R5bV=fD(
zRBhTsC-#|j$~fec4RQ(WN$=JdM$GOTzdXfTa90u+OOT9nOBfwWPcM$*oy9gAP1BJ}
z30n0HF1N&Gl@&&}k8MmV@0quKUW^TUk)2m?z4AAx-p>7B&ORo&?TW^09xGy3+MiQA
zg7`izq#QvQRt(c)ymE|wFdDaI9O@=8yTyOqWHmR|F1MYjh<+`53ddB`7qTAZx6TIW
zEF~|k!H9kOdiQ>dqXkgi&puWys|FsE_h$>(6lnn$#Ki|kknx`RR5}K&xomV7xw8#L
zRO7rKGv6Pe<7ox+zA~$$LB_?9!}rdTlhO}j$BR<}ov}XMH9aB3<5=a~lTn}bM$AA!
z$8}6CJsH@7XGQZD%e~1X=0FtzAJ}y<?5k86;sL4>;Y{Y=^$H3D5XIu#Pli1A%sedI
zPD>ud#K0?jE@lm^e_^o#*4}d0tixwJgT;mgeA7msF&aMtB_;e6dZ92emXC$Ps9Dsa
zuh;`SRChkw4|XNtmGcAHVPvcEkY~-w#H6yOx!NvN2lj=u4uvZ~f~5Q{P;l-0rXv8~
zVxF4<vyV6K`X($WMnn=t`utP9`=Xf?8?ik$_=PQL!vzci6>5THrAFT>mW#|Pyp)sy
zF299odSpNPg7mHCLa&1k$uq4e$vqf|D>lnXTV%YS8xajxweu4#eWp{g1}RqTfR-?}
zVd=l~jIDI%kLJK^!{kB0sL3OFOU7ZgFSK5Eqv}v>f<x`>*2m}t)7EBUM<7S;U>>rU
zmu&FazS!c8cYOxo2y#Yl=S=f-RcGruLjHUX=iVnjDs&RLjA{zSSQGa<7*CV)bjqhM
z==T_r0~UIbQoVGoZ6)P4v<jYsp%K4?spRIOyV(^k`o@?oOi~BgCUIo{Tf(29*=wO~
z8D(kt(BK5~EHbo$*vmPDv^-zNyZ>Qci?H?QawwOGmcB*eknx)xkPQ_{I<qQbx-^3#
z^ZC3l|B{C{;gT`G-r@A4<q=RpWZPuec9dBNj<?4c!8+C3_W9gc*s~KZN8^#NG)&Ke
z(h{xN&IlI19IO-fq56UMrz@|(Gq7M@^Mgf*s$Womst*e~Yez(u7Nfo(eItI5MgwZx
zw5=Gy7$_5sjN%XS^3!_$J3VUjYP6xi56PA=zU*T<V!qC40~?tok=FIvxA-PzMEsMh
zzy=JAvp(;o9lYkcbBkx}1V^J&%dB){?jaL4Bg4W3^vz<X9wcxA+kDM1!|mz!(^s6N
z&*YAc7BYB=RBdWsb<&#p7Ma^QWfh?Q)K7+4QGUd!f+kEyP?$VfP@}|FFx;?N^562U
zD3y=bol5(dRB1R@M#8hRd}+_D=re^=BXYR^^3=7z3R;aThI1;U#b;@QMu~oF5^R@{
zpngVqVY%2Z{<>^XUbzo{bBFf4zg9tH<q&VhwT#ao5>ZMz>0xHgW#qSqk<Dgz8@6}M
z{)CYie#j#K<D1#Kda>N2^#6%N>i^I07aMi3CLa3V+1KiHIzPP|TRBW6EMdpDJ5BGB
z{bw2kbciFFV*y-dVe{GUJ{MR&(;T+nnvnD6pqIM(@(?N7H4G!-b?)dyhh2bWhW%iJ
z<>8w)g?ml~230vT0ohima;G$Y%{u{SR`;?<8BNK$$bNk9jf+EcS-rlUVJIhTFc)h~
zy2E&bf?7w%b#A~H({t9!?&Bx@8kygV83%g5^ayM&^zN(9y9704B78BRUF}BOcmEG6
zB-z~&EJ$nW)JM`{W*2MKXC<UQ00~Q6f?zlbM)Vn1*_MCd?tAR`QvGDmI3bvMYqvNq
zFfh7Ie_|q9-^?2NwKz$h44V_W@H|GJj3kZLx<uq`Gw+C!6@yEUAjmAfl@X9GkZ@_*
zUPxYK-qG^gQ&i7K4nvuCBU=dVxVwyI?2PjQK_m5Ec<4o?MSIEi$yO9j7ygTV55|aW
z#WL;2Mtfysr#MgEh_!V2#l9mb1f1=}Ogd3)!kh{l6i35RJcjk#L8YA=+F7i$Av6&C
z38vLk0%NkIp=Qw6q_<SMyLr%ipOYijH^XQbiAmJUea(`dpoZ)$18t;)Z*`<plxGX`
zbO2?s88jw7%`2a1(jeP^7cBk+I@=I2vHi}X@k`5vSXNZiT23qZd%=(HMWpVNZ2aS^
z%9b8YI`cTC9I6P(IXA6HS}foA`f~sHedrIYVLAF;<GFqfU0lo9GlBl8emcfZ<^dzS
zy+mb}u>lS*VC@-`Uaz^dCVZPA2n@qnz_Xp(#O@<VQiAL}Qh5bchHf)4)Q%FOv%j#h
zJoxAbHdWx4r^{~%d}47vpq6AGlAF3*fYiqf!tOUUnV<P8`iVEM7r_kDUG}KcpLw=i
zb`=92crZYtRB2u06X554<4I1{fsRS`7S(LdKVz>|fst>m2_M_TuuLfUAGoGrDswUC
zbGVjv#S1f5m=7n`qZD69k79cc&Vcg0H!cY4$>EIvc&>58e23Eq#y+vJ73?>3tI<sj
zH>aB<PJ{2pfI?$vdVT(cQ7Vp1az8LYDskTX@lAVjP6;WrEp0fyCq{nk{1x^t{S&r+
zunSgX61}HPR$`nadY%Br9|9GGc)gT6+hU1ikQBX&?O8cYWCa2hHx%O;N3~|3><|O7
zG#ZMOgIOyh7WpBbOalOnjLX;SJAz!SSDZ6a4;}gTyPf~hRXY8+_!xJQx+=KgQeLX{
z3q3^gUk0q6`POIeMv3w0m@e6GE%K006clLvrf*Nz;~vTZm5Vb-Jw^);IF_~IOFAEP
zJ|mF#X21v?^xHMuFBU=R#1~>e-ql!YaYwx_>7uUsReZMgS=9fpXNu07O^u32&Lbw!
zC7aC$BIq-6Xn462AQt0KxkweL?TbsNCBY_RqF^GX`|_Tx??n(&ovd15ev`D?g}5r9
zyQe{k*#Zi{)RscLdHAhz$hmyWPt+~AbI~EW2E6!=<iJqd0H@sOJ=J`R1KjU&y)5+z
zk3;4Dc9q5SBZ#Pn3SZOr$pLsYNSaKdnk8PoK=$FsLm6F!nV2Je!C@ChE}+s0q#vl2
zG!3|KzX@^f=e6M7O#(BZAjH#zEg_@7_=Qj}cjB+w(9Ufte?|=&eZdaYL>ldJkh3yv
z0>1=xz?P<FusfK++81q}H}tYp+9Ecf&>PtGX)>>9wg2f&2e7~Wi7mMov=k)I%9b!J
z2$MBVebLebORz8rU?ENB2qRQ<se<}{l3iij=<L1Dkqj%w{dkFy20tX9MN(%-_(53=
zarkZmbRxQgWAivUPPK1rhx5tr^Nmh*f@a@8>KDH4c2U4%9g53kgefq)!6IiEq3|*G
zi}_T5Akk_oM&jL$HE%&+sK^1n!Cv;Z8~~q%|9~$>*%5m%8#;r(sCE-M)uI`kk1-?b
z>U(x%j{<Qk;(9%5x1AaGq1c<UcbBY{YpIlJatpPT&@sJY+77#1T7$Y2ftvtvH+Pkg
z`zo`sK%?S9X&0e?6d56}%9Z8IC2ne*Ln*rlR2)hag1HwAU^nSNm0`;rGl89|=W|;a
znPKcq3equR%qPOUDq33cX4!q}wrK<71z~!{P6<YmJdpH|-x6_%qIX$7g8c04GK9B9
zUtWF{7;aSm?mW7l_W-AELs1G)0V#S!@ydPR2YSVH#sAM~B{9KBn|ObRXo?TdAx<nG
zXo6q7(;685edv#sVC}v|hXUHmS!0X+RV@@g`XECncTrof?g~hGoq?KBECY;7ym;H=
zRN16E`fo9=m<EB_Y_nXwLsowHI(xSsPl2kyE!1^W|9oV}5rkzYK@b*vR((^j$B7#0
zW|-5A;;`K+_c7(F*EOwsQMJXa*-h3G){Wrnrrs4ef+*R-^?Qgv*T4Hs@5*L^Ea{tA
zQzZ~#PCZ{14%9qfH#rZY0jJo0S-uq7x&762k@*hlOMr~m4EBP-XO;A646W`VkboBd
z<nMr%{ILjy3$wr%1yYR}TN4`3>F$7XSaI93&n+B%MEE>1Y6VMtAr#E2Z!kAJ^Wg|G
zm^%i$iu1?-hSBu96{ls80$`@wH^`4Nq@wSfIT`l+`dW5<{?wKEUDl192h>5@`CK=a
zyFyb+7`Z+6!MTuELT(G>I;ABFPQ^%ER+iY2JMKu8&o+_B)H~OKWgD)fP01!W59$wz
z?oB!bx;ajjC~E@=^V-BJMwdzR%m(wQT#^{PLN#!Wh6$nFk|CnSVD+EFe1kf5c!-#e
z6rnKL`-8vPq_qmKak8$LT5R=tNxH0yk)SFsxJgV~$G{}>BX(PCc9@9efX<qm)3o4=
z-k`=98<qkW*~Qp#<_T;O)M=Hn0(M{j*fmn@r0!NS&>V_?p=Q#c7o2WD$1nU|y}iOv
z+vd<!G~4K-Dskp4YW(LBBn%V^lUUB{E~*A#XVB?8dhL=fRHoS93I+Z%HD4%QN%m(J
z!j{2WWG20lC4~zWQv7n42DlySg=oF;IP{-Q*>OYVO<r9(CUoFShHc4KrZ;*Gpg{dg
z(0+dQ-l5Zj)S94a=!fENd0KuW3;ES+E4-IEYAN#$`e(i01S|z4I>UpKLA!xT<96`_
z>K%Z+U{l6CY+()HL#V%KehV*kniLU^CKz2n#3|N7+6#We1{Or9gw9m13Ro0Fz^9S?
zCq_!_Oq%lCAjKJa?+B9Z$Z?=WtpW&0ou9ViVJ0~Pe}z~7^={~Y!;f|~q+MmwxwCmc
z)>gx%W<1zzf7t(n6v;Se?Ff>Gh^5GV^3va5S)Hy<)_jV7Bo*zq8++pTA6?QH>`LUA
z%bz^R(9h6l`Pu{xGM||^DAw8eb|~;8hxS$UW4(RtgV3=)RJ`*y<iQL>gC8!R=j=rT
zUr^7i@!1NgAv%KZ|Es^+Cz0kYv7jog3HB=Z+x?&a6{+X3I}%z(;^rUnio(?|c^crx
z-A9<2BUYYdFK6AVl(KbBuf7u58pxnxEUhJd=L%ciUXcUrqBKgdX$n^Vpxsrw<%eXa
zmKEr^;^(Jx=yHx1b!OboBZOv`Tc^o0{dq?J{$6|kS&uCg;<QB(V-bqFgWkTu_{k4K
zd7US=i^c?$y>2SI<#7&TC!VM0u(FoHxL?x3WN+J#eHoN}eK*P3$|=?)@0@}fBlDy!
zh(MuOjD?l*31i0hQu9uTlMcdYe54Z>ZJIm9di2(XNxV{Rt@{DrcJ!zl0dHm}I3?^%
zsGxr3fET3g-gqO_a2BhQiy3a$R^Mxx(FA^#qkebKX2%v_l~|*&?ZR&yCiyU_ZEk8l
zVZbeZIIm0BX0YCyoC>jn&$7`(m1cJS{Bg2h--+yHLSD%)&t9S9KU?TrB^@8@Eo0CH
zVQqRNMT$bYqtoo7vqX8J4>f&e2j`UeinU(;*IE02u0wI9OKf9GY}|a`tsfkJ#r#pB
zu*dbQW%Bk-@_Lp;WN_#=T8flxKQM)$3iPkL2v+dXHS?=qoO8b#hR4lpypuMd)ht*Y
z9>`SrjfgCH&x5@!Dr5Ovu8niw=@&oGW(k$k<S(>mRv5ktxvldg7xMBvB)#H)Xa@ej
zgY@Rl&dT`haPh!r#C4K)-|~_DB-0!(E28ecXmXh}wjExk*E-EA5QuqH71^f{3E{3+
z>3|c}XeVA6SQGck+-zJY3iVDsmD{?4!PsIlo89iZBvq{4zob&;1NV?#9d*8h?Nmi7
zE=ojV&7tpsi}WKSB@U!0kDNEZxehQ_aMFwzva?ON)_}9)q%1K;Qg8kA+@Q3sE@svV
zY8sS|9TY6qsND$lIKX4&nh;+ivux*VI~;2F;RjF{Ja|RLP{aYJMY7n2W;re+?~hQ0
zXUiMZ@I7xoE6ksxiI9@ke2Ih(jxXhXCC?wXwzR!GoYCt;Ux3Yg8LLQEg1z&4H#DAG
zReStG<0$5rtFeEK^=-{IS%pJYyZzNh3KxT8G)Kt(*P5bnRGr69>B|(SfG5Oo?mjn`
zvg4$MsY?p5bB7{esHZYrZb>G{Ip75)my~K<u<mv8wLlFfLfV7~?UX%R9V@8*sJ1Q8
z>Ww%8e0uj+fp^taiNky9m!E<*x)@i7D(04YnJ=wln(v%}Ax^9yy`43J%CX>%j0Vn9
zI6&On^fD=v?rwDPm}!CM)>LirA6?g?Et0941G~L)@}F*%G^CCWEd<p3ae|03BMPm*
zCNVa18Q#s>BY5$6Y8~5Ve2!as7!v&8y{Z1WA1dYLD8FBa^sr0vE2b<vFfBezoGb^J
z4?V6IDLPq2az@_`cYUT4J9H%>I#1fXF`9^Li>603GAAjG6mGnwUpLV6!Hd8Xtsk;j
zC8#8^KJNU9gM$MG$VxGU!E2LI=n2n+E=21rGICC(z(2h0wy0`s|D52O17`JW)Z=Aa
z%yA7(T^nS(k(87Pd$U;(3!SZAge!z88ChgkpfF%CakJ(2n8A5p=9sxn#z{m<tAwt&
zo~ncE4{xy0X07zsV<bw=0%fzzb!hD`ue`@954z?WtmlV7lW&25@oqj}a~8Z_KvNx5
z4?4rgxrIiPT1pI-Ns7MshDXd*QEHkOCsVHONbAk-hoc-e!+?9krod)9j#MW;6QdNw
z)zV-CoD0memq3*V&a+wpz0CE<XU<<9wg<d5m9dZYuwQ!4UF5L=9kl??Da}ETlgm^G
z0^EFW)yx)zvRt_pzo-(}207qmyYe0Lo0CVk99Nzqqt;?oT~Z%v^G4_4`3Pag1$fUo
zk_@!e2B+mAdyq_Vns=?Az0*h{ECg{lWH4xI1{8+HsA&5yFlm^EzW-;N-!Ykn`H`ZJ
z6VdjgWu!}0-ID9DDNX@&ddU|)?MbH0{t3{zQy24=(LT^kci(+n+`_%0%nE<awdw?P
z{oHoGSr(^7Ub#JQO^v)F(|6$owtIVGH{BiokL+?It-e=FhUHQ1$x5=X(+^bgQa)xz
zX8-(^Ryrv#CnZF_wHYy29Zpd<8JK;))y*Nk=I;j2`RmhOX(z7F(c^Bl1}dJpk><Ut
z=L5>%cv?CZ9vEAxJLyQR*N~<|P%~eEb0v{ChIYSbZp;KO59K6L%S=PlfKDu3xv_65
z*ITVlDlHq8;IGO&_#W>_V5S;t18UO=?{gAoR;Z#$i!TvAV1Nk@;Q1q5{@2>h@6Kap
z(a|GSxIgaY=gY)v$OkMa#}(c6i**ki$1eI^jCfhZ{HG<yS|40C|Kics0x(mYBUK`+
zFBkrUyXw7V$%a2;UIFV^LU9WFcON&-QV}Nxfm0_BJ(L11oC@u>#(##(lz)E%hMTAg
zp|0~Bvwa<cuufc+uDN(+wZg5NE^}(rkv{9^M8np?Oc3;n+%-ZcsDl0VC&Y&K?r)}=
zqY#riQXqNBzQ%uh+nqv*96LP4^!GIEAjmVly6xi!;zf<}H<^xc7`5_F2a6wmEflnv
zT?(Q;Dijku{bNT0E-YK|tKV?2Kr?PJ=N3B7HuQ;zZfwm0?g#?Lik1Ru+{_nY7FhdL
z8`#epunPNKFdMmjsn)Eb&sO`~zMlXtC)K(fN&g&2tDjzUS2+=sxRW#y(4GrMLs=x}
zG&D|gcpXLvRbLf|#9?PX`#Ix)b`s~$%~^^awD?pYTHbF5(JA|Nr(NlA6bQMq6+u7T
zkZc;V)V@i*jPCfb)P>s#)vM|`0CYe{@^R|ZGTr-qW*$-fiJ$>sCmE!bFwW@rHEYP*
z@*w+@I~M?Z<A1BsX5nrbg@>v&p#PpK;ERMcIDTUOZVx|64a~@sn7TYv^k#RORhjmF
z8%pUt7}9*qG)*vzg75!=E}8a1nMD3r(S3Jw#q031$87C+YMd?AsD{{xu|dOBYr>&Z
z5hE5E9peU9)YG)!6$7@lj%d{LQkVTAu$ENr8<M7*oz%U|{4*q{maEPwMt$*=KND3e
znp)oyH~?y~)s$SjDPkx_GVk1(gB5ushJY@0X~A54X4ycjEl(|(kVL+zQI*@fOg8aA
zm)Q*(2J(6X)(vB(x2!J1>vJAp!Tk>YYc}b;`CNJ2!Ji5o5#~>7Iv)eshEq&5_kPO9
z(&eRlJ)|#2V1AjX183)E2u;ZEg%kR5gBLeJi0H!zY$nC6BddJ#O1CR5Rz>IW!D54b
zKw#_c&KKPVkb?c5k?hMN39|-~qJXMQsxLY8Ypw~bgFP!hD*S3aihp%ydW?ckL`MQQ
zYwA0x`vExrd{naxaLuS|+B~&m^?HA*mnQ1m;^q;=3>)U{JDxMDd}iM{gsi(MJS%No
z`h^v_dK*wcwu)wzCR2JDME178ejP`ULL-j8!wt40h#_GIPUc$*)74?G3MsD&si{zy
zJWM8HE+Bt`)rUl!$822<2^5Bxlw-T}cs*vIOa?aVTpYt+PXK-fwV<F~hxWRKygbE}
zSGTK9{n*&>vESPxwq*bGr$04z#QK<AEO9S0*5}^4aVp_2Si`FTsapP<WkW^ZH)ih%
zY^OTiI-HjEN@z2!o8!(G;s1Q;*pApzUu(~s;%^a2nV)`@t$F%5Y`}EO;bU{N*2Vdj
zB-QNC>cmOeKd3?8lJE}o<Ahh7uy#YQ-_EJV3FMU^=cb`wlj<%&zAt!z-MTQButDUr
zs52KIfJ^@UTF50m_|?mlzOiu7xZc<~ikV4)O}gprkv`{f^)MiBD`Hl^(RwuxV3Rd&
zG>0_eW(VUHgr>=%%X=XnmqWYjAiVKJ46jS9%vbGizeSxp%sJ$Y!P~C%seqU16|%h_
zo|J3@hEf-GZAH?hiKq3a9BexYVXLBGcLCPuA_LWl3-Dt@S3&d!4%2IO+(0+Y3Ms*c
zc^akV--0${5Cq?G4kKHDtnHgv_-XxJ=;?7L;CAdvD7C~iBohTDJ*T|bX^mrhKTQc&
z@62su=q@%J08o6Z1i+GYAX1y^@j9<U#cvX-sbR{PjU0!c+r6a?4AupWM^71jfJx)h
zZL#dqbeJbMDoMJDpO(uexh3XR0{|Z09@ko2>rdGd8!u^V_Tmq@=$>`HLW0YsW!y)K
zEXKlB%2%rZ#or&kPpE}W3FU7I{D`+=4Am#7s#r5=Cj2QW`)c;zjl}<=pWlgylv$yW
zY?8S{#}Hl6(szijRh`60?H6_J^8Ui9jYGVYktg9s2*eU!r1z2PPGrfOQ{_56=**Vl
z{k`JniKVED2&@mwOof0?-iq%2#mOgePY9qexH#txzw*2+3t5fQ3^2dxeo^m`*Vm&_
za}ia_=tm$K%2&o&e#1n=oEgVQaC=VrREgP-g1e|IY9icyJ-VUO-hW6PoYedMyxro%
zzP33G;&8A-E3VhbMYQg{C#B}(ZLn&ZI`L09o%?$1;M`)Hwrf=IhiH+Z4I%7II`*oN
zm_ofetMRykk<X}^InULd&aA);3KJNPpJkIVcq7;e>Z7V)oKd|S3p&?`7+3i~?>lAE
zVM`Z7M(Ni5@~Y!}(&AlNU-`*0O#1<U(=PMUlXCo~hYeFdP3YnqdFIzd0C3W3YUHY`
zzR@T6j(&b$)^OWscReE>EVcg18vb*=_1}6e`H-ELK0+tN&2Z23I%3>$&hH3<@G7(y
zxIGxD9?GCp6g4omB~a1h%rWGrz1ynD)7Bgl7PA6^ZQNGZolk=fYTw;^#UK=c7&1|N
z6BahLC>77ZaT;2rVyr5b0X{G<{R6mT{u>$#{pbJNE2h`}TMN!wmkX%QkB2xB(?(XZ
zjhOnMjlR}I{)HCspY($Z$Xm9OA@A#XeH^llIs6MPEy%A{VeB)F0s)po^|$dVCXyXL
zpGJu@@YKdEQiO6by2On_uU@XP-86sigk$%+?m2MT4mqDPC=lsGo+3*a5j#>mZegTe
zUvDui`8n;kE6G>zPMekgvX2jV4uy~Z``!7s?f(CmztcP}RiN+n$8(AMH`^2o-RUt4
zrcQH0qn%Br;!3ik>5oqq!bXg~K7lFkj0u!6M6|tJ=nKikQ@=#mYvdRhW@_!q-X_2P
zQ)3in6+74Bw>r+;9hgiM4Whl~6j)T}LjG)ok$rf;huAN((m9`JK@evfh$8S}(b2Rh
zyttBu(#7C&wQrMY9Ct{UC7KN%r07Jaf{jb^ZTid+C^{oUB!MbBN8$&2OMlMnmwql8
zq6zzA(jTCdkMK6T`w8V{KC`rcR}YqeEJDofITBf;HGs+VR{Gw)luI*9BaL?CNbuY?
z>*QmXn#bDQKPQUsuU$B{T+Mu#3df&U{U~G=-XIQ2q)lO{1GLEdt`Y>w4_#khC0#eW
zxn45YcS_W4qL&L}{|9~52@j{e8@rhNMmAs~^(3iDCD1IT)n`5EtgBBbW66myGDipY
zbbtjBz(1?_a&BXiX-0hkE!fapEcjaY+JZjvbRB7e=fhGtt%xnRYVRv<p^Os2XI;I<
zRAaeOr*c?B(aR*iRmEL1C=~iy8yWmKvKgod4AtSia?OUTf;>3`g3(;OM8Ou74EueS
zG&S$jU1B?ZtuQj5ctKXVc;nr{;57RO4j+ZBskM3wS8{?6qFA<@m2{$=vi2PV6<FWf
z>I{@Fp!{SghU8w%3djLAc5;2;`7hkPcU)83x;2cVpcL6sL~2w}1O%jaVgm$3q$@4b
zO9Z6%L`8a&u7E`8Erb?&k={jG=)L!Z8c2fQ+IPR_eCNHV+`I2N=ic+)zp|42k~M#8
ztvTm-o@YE`7;qgcO@Kth4=C!UFVg&Izy6R-B#_U?$wYq^6y87!<C$vAT}uim@J_$o
zTGxF>hNS3W!iUR>pCw6dRqRFGAq_vEmK<ofh4gQE$;0PaSy|uv{3OXOcO*3aR?@D5
z2(V#B!`iDW+SEDeSnvuXTS<qhm6gR&XI7%#e$Zg2Ng~D%*Rm9@dPLd!WlS3yU2o=U
z8tYqv3<w4lzQBj62Of6mCa^F{7RpD}S!_UPh`{_qm92Z9k6&sh?#Bot25Q$ox@qny
zCbv1rfC6VqB{%&Rz){oUMQCG@1r6*;2g)bM=(_In#&bD4F`|&_R4Qyl(3@JG47|^i
zdkF^i)rQL3jOIMksI?^ZgfP2=@+(E-hF_Z6>B-I;B6Br$qD2w`dk3Z;e@!O<NFZxX
z6y08evyK(xuKX4rH~t}tUMQYum2<gm@$%bHd`qLt%^E9-8|ANjAF*AOY9~AZMs5p;
z#oDumY&-Nn3H?o1R(m6S=)WUj_pfq}|6EtfU)qj3Gj>hAB*c;#%ZFHaID+atVbQoj
zojQE{txva7f0CI+_fWrLJPuaInU7#fK0j{EIElE`3yz#7!%>E16jcsb*v<uSL3O1X
zzk8fAt#Xq8*zuK}gYt7c*g^@W8<mBtaDkigu7&;sp<L>`(nlbP1Yo)TH`%v8YM=bQ
z7m?QkWT47Rc95jFx(Q~@_b`R#SQr~BzNUi+yJ+{T`xN4SXc)7+b;1&wM!fcbh`aZc
zRFc#(Tz`%zkGU?~J4bs1s6DYu<ROno+!Wm(vV~^sP?ea(z?Uug#4^)E3DJAYXaW^c
z0F&hvQJ-nE^?d^RCKybk&kqNdMI81~MzXx}v!KD=WRHDRz75Jdk;ytv_iGqp2HrE5
zpbv=UR4cc5_YoP~mirF%!|b<$Or01Sb_*FyfO+lfxuK>fArnzpsC=KsQuH{9#GF=F
z&3Vmw-y<)hcD_N}%1Ni=wdlBjx9INpPqITHp#Sr8J-b&iX^epg{V>vCJrfLLd2o^#
zDELKp@xwznAIUS_MqbyDv_w%iuE1mE1g}fKD8!E%K2>JKJ6;;8k*Ht5J+KEot8jmD
zNALs6=-W@SIo<JT9m0=}hMnt71bx4>wUeY*)x{N`{8#~gEY;Eh-Mi^?!=i@`)T4bh
zg-bW;-$jM$AaF%UZxwJa*E=IoaIvoM2vHip)9$O4bD=TLwB{~qIKt)dAzmB#s2da@
z@QT}^q_GB^ixLGI@;}3<tD{p#Aj1MB&RiFg75ykcEh$MQr1cbe-JfK)zEh>hhX;Nr
zs|T8AnXKoxJT#sOxP?u;l=0FQ!7P8qrMD*;<s6<0VkibbM7S@(FDIQ;=mA&KNr}rm
z%QHbA*0f>+`ht?j0QE!!{IcuRuHp2WWb~yl_8!P?tM?E-JJ6|GwE(=eI<cpGr)07%
z@&{Z9Pw{J;qys!uhKBz__23Q{`<3bn!FF7)d$m51QM<LM12Hb6{<#VMOW?JPQ}p6E
zViSRrRks3RBaDsD@0_!Id!m0G>C1s@4>GqE7ou8EaEe+;ID4M%_S_SnBU6=%u?)cO
z0c6m#inQdew;q$B1Ns=7kM2a<!I<hQ%#X5PjMxx`O(HT*ClJ(VCJ0!MKrN{oeRW}%
z>G%!M*J@TsRkVI#O{LO%1QCo~;X)#%x;SHglF5E%#QU>xjKhU?!6vn$g!{g7xcBem
zd0Gxf=s6{dc!?W;>rCmV`;jM@dtS%Z4@pp*@QcLlQXxWtqTQOp9gM5Qlxgnd1n7`k
zd`}8RZ7b<3$cy4t1h<MGb4DoKN;(w~BXKnJkTSJE+ae8w3|47W2MKf-i%<Ttb;U%u
zA%qz(k^vg1Gi-)qpMg1Tu-y8XD>B6#mX%gCaK7Vk{lHIkOx_Fu8p#?sJ)(wc;$>4y
zTx&ZHtdxTSJe7e+nmtq{4qDHC`QcQ-eJhf>cSRuFK_k%Q#xqWzUfvo~l`a_Ro5la$
zrH2!JcK}*FvMaCpqMWL5nEv3$2}S5^nNSpjPEPTLqUHRSz&2>R$(pTi&Uo{Y(Swqf
zrpKd}FEl9XP3l?enRZ9~!}io+)|pQW`0@v1U~d%#2kLD3Q-0kj?lHpRG@noz4DVN)
z!~#YpHOAdEmKY`}L1jB!Ljd>+dWXTYYrE;NGvl3--0k-0a4dAvE+R50d(EX!JTv39
zx^z#We}Z^;IuG|Cueg=Te8Ua+Y}QOkWZfJ*gn5R(OpT%ryIi#HaK6BTlwfq8>fXu+
z(Ik=@PO@c-x~UqS_1VuQEG%$hH#)dI-<xX9GIU0(Ot_ofLt;>nsya|lOMcx>Njx^9
zNuXW(I0gD5{!mPd;&e?r=>&4fLyAEI7#%@hHc>MrXj99UJ(<5Qsa5!6_T{G0fm|f?
z@a*2<bG%~Cx&Yi|mP636&3@SonR3u)ux|`f93;w{azuWzrj5ujzsPu-@r==rrm^WK
z{X>HfeE6kYC-|YVZWHHgp)}CdJTlOi+?=#J@?ZW7CE`C-2Rg$oNpDzWGYQmzYe1Lz
zeJjXqPqAyy8TG_@m4K2l4@cy@w4NnRIE%VMWw+s4!RXZq^6EqJI&+kw({48*o5bIn
z8x$vQknY(lsuz0CrDILaDDdq+2igBmL=*qvIVtpm!3#Zp&`*WQEg;;o0=j1e=+tNF
z)t(17*BY9pKW_}mI4gQitLH(^V(dVXD}rM+X*$;>+rWf>YJ#*cXPw$!2VK7c%^6dF
z)uU^v6%e=A>m}*0a9(9@hNUmB&)a0`c`e`jZRyhkYhD%cu0Z6_ey@4ciB+9=ZAE#m
z*ayHQSsE#|@chROvb=OMDmNx~!(YS0FMl;U{6m!Z-~7sNCu5vYO#KjUOS|~$?xX&o
zQ@45eFAe4hGW@Dsh?Q%1Rww;~$7DcVvbx_Sl3JwAa&V<d_ciRvGBPkdSTCQk4`^=>
z1s6qy6G6*UtCx~alin0dU%U}@uTI@}Bj3|Q8%t__n6q|t#hJH}L8+VA3L6gObCwz5
zT>u44GvhXeI7>4Ojb&+AlI}wGP#6&eJ<gRr`;)dV^PT{x_Z7g+TPik^eF{6|)Z6s|
zMJ9Muu==J!)j|X^*@FVQfR<oCC~l(l!Bv4BRO;!pAd{8!!a8(ABLsof_zXCB8sU|7
zSlt>sf0?pq-zTFwW|(LGHp4;`sq+-CyE1?EZIcn-61#V}{<@U!L{Mc(ViDrb1VKa(
z&VL$yob=JiU0=}6ZMj_GJ1GSa+$8JFFtkZ}A7hi0o^qL2fDL5a{uZLOR*jA<J=<&Q
zb$(I3!-o43Se2+@$=`gI?(5aUy>EabmnnAI=6h8B_YXqEnwl8I6a6-;X_mQl9=XD7
zo+Hb=WBF4;Y8!L#y7E*@+bPuL0MJj+Z$n&cnEj&1JkXep$Fy(08ij0Kc+CZI=TP^9
z%)!|`;cW@R(>C7d`gDf{K3usdo|Om@>u_OD@~!Mr7bLV2`-++tu=rMQt&s|C=g`|I
zg#3R*V>u&UH(CB`<<4Qa!hD{|_s)Xt>4D3To@gd(a#~!9yX?os7mSQAn@0!p!^`Uv
ze!L5JS{tvpBfh$QH|BJ!&WCl+&UA@ZFK1euVR9rb`uW{+?4bsWV*hGS{6A$p|0RzJ
zU2%1-5yj$!?L*6V)4~Z>`EEVLM_2>=Mzix;mU~6)j?y<`O~d$MZmg~B**!5i$@$Nc
ztqEZ0prD>-ssC8ekoH$>TcTrrXP&Yp>x@)mqD|k(hEZY@IFJ^*p;5^&d3@wSn>hA4
z<+ap_{+A)8U6o!t$$DT{_{W=r{f&$IYl85}+3<yIKzfqWD~fbTTxUyuE|h5BbJ+-W
zWOZ-hAoq4w*)Rnm1(S{KYcgnxufi^WbQx;53x(eWIuXH^pn%gGS{13BZSa-%ZQv!7
z3Po=*mpkq((-B_E1M0ABG3*eZ?`w_4s03e4ApAJOO5`Fo)g~es27SaviM?L`M(CZX
zig><uD!{<mg&FPJfVgNv)Poj`2$y0aZ=y+yi2324WDMVPV_(|IR=e+4gGJ?32p5UZ
z`ZJ;nHU-es018h?o6|wyX}!vevCQq&)Kb;hr3L$q#kM|T0IXgcV{H-v(G0>#ie({9
z;&_fwdvaYVl7Weh5)GhTFv9@{rFQ>Mvc>CJ(*W^yq=y!?$AqY2t4<=N;dtKGN-r=?
z6t23<6#XR2&hZ~f-GG!tHFMZVDX^hSrlv7>D^K)?=X{?vsq$|;_fk9-8(HlW%%;U?
z9J#3bwJ))Y$k)DmY@FuQzaYhhV;Zt)U?duUxeEpVvQ8ZtdUUCI_g4Kp4g%1M8%H*{
zldj~$EniQaA9~XUCgblG@5A$Ggd&+{G;kCMS4Iov;v?3RP$S#sPw(yZ(oBHr8iiL$
z=N=W^RMcdj9JXRB7pL4TKo5VV=w;L3&1*lD0NUY%j_!l-xA!}8l28Fx9ul=~eiz*d
z*Y8WcMM^m3@zhg+4bB->*?&keYA?(8iF7e>%k%Wm`LQev0rrKGOMNVbA&cyyDTJZ5
z!Ne3+faQJ@iW7~L6>hRwy15P@ZC-5DRx`G{C7;nl|A^Ic&n@urRe)~&Ey5cUfRk-9
z;ntOb2fH4>sNb}I9lC=34#gEjw&;}WYp+#kPL4l09J!p55uI{Zp{*XmG8YNa+Mzah
zc2JhM+V?b(-i#~rF%KS8aaO{ThB00)kUZ2LrsfGhOu@&`4T2KqlQ+tqn91gb6;j$A
z&YXicKB|3amd_k2=Hr7;d~?WK+&2^pIUj+Yk!t4QyClvWTT0Y$j{0TJPF7v&lxHSR
zzt7|hsS;2v`F89b(UQuV_=Vs!$xg6m+Jb>_g>UC`h`n5T<$#mO?O8yk;3pYd+`{gf
z)cmR{XTC;*W6P1vJQg3|t*vpg`8rhr&uqT{VlZ5fo-Cv4J&g%k?%0T4Gx?OYhFz2K
zmH6PWYigYCLk}0#gU7!{um~P(fGK-#y$i_b+X?ksdb+3|)XnOVr0h^qBHOT}=!N)}
zymNWR(|*Ee61#FluZ|F|&mO1~DSKa)_A#ln2uMzu{f%k;yL#t^p0Dq{C|c>!*iKq~
zg?{YFXj<OjCf=cy&-AkEwX*k8Zcv1}yJedLkDH>4yTY}YJ9mC9RQS|6Hac-MQAGGt
z?eS^F%YP5<d<-~77pVTDhV`%fKmSQ>#{b{xfax32YD64K*_36^NzNr|AM<D_{pyCE
zoGuRf2%B01&7AD4_j=0BKX7Ni*KHkh^7$jzt3yECfAMZXvi4KnDEk+UM5d90zF#~o
zGcl%PW%Im-fge0NP!F^2J8~58Gd?bEt7B3k%N|yg%9r5D2d-Vg**{0s|1@8s-+iup
zJ68KF<g3G@@_wVH1)7p;axihLP}|BHWXnN6Lc#Yb&||`|(2@YX8X4j{8FGz<g$P}k
z34;``M8a4)_9n(9s<Nz?5_(v;$%o&t>#qyYoD1$~cD~!py9(VYfl2n3B&4luT^~0f
z^|>^*T2vaTj~kV2H*^q{Q>WEl81SasWM~H}U%iG|3GnqmdEUY~xY|GU;brvSl~7)S
zWCGg_*rM+|@$oS02~@EtzH#Qzw1{D&)i)F8f5?mzdN+^I!Y<|(tT`=xy87N`WL-m?
zpr_w1t>6MO4JAz?R3_qyhlIXJ;LNo=4Rl@}hM&Pr2>flP!-W19_<>H8z0#SQ3bqXL
zwQs0mIcglV2|+K0FDw%<nO^8&A;><K8$Pt$f<z@&+fOH7SIf7xpEymr4q4GFqoZF>
zWv*4L<2&(TiCei_^$b?IVKmkrH0EIO`4#n{wjVm|*!7roc5hP&G=<y|l0V%${PONL
zs?YH~Kq$|unM2&_Yn1?Si#e6$1E!G~`3-=!9R%QD4xF}z;b%(yQ2q8knuZ!Tg^zMc
zL{X`^SA*OhE!wr<Pj(HnzI|m)Rv%wkeTdH-185HCwsX>tS@{<=jDX<jZ&pAJONW+A
zP0=NQzgmMi@}XOR`BFoJ>XOYOdnA8S=G7qq^_J~iXj};<3q2S0u|BYxT3(f{wMzon
zgZJz-<G~L-dqt)gHI0@6{x{b%|HNoo^^cg<&9oudT6<0exWtD~{}0jlPw3^*pUsur
zuj?wQaSk`*;R{#Z&TG(Xq)(`aJe!whdwWY}0hi;f*<aUg^_^?MI-F?u)ibqiMn&a%
zDVwSa>!NCVUJ?+|T$(bOwsF^9zZ)!M5iV97H7D*oKR}JF_r2U(wB*}LNjaXG6V7fi
z;p6Z992m#w|8Vv4&+-jqqfUWmg3@V*Y1NuftcsyG(c7P)PBi{BNUFtCsr-FiB`yD2
z<Bv&N{i#CpbfZUeR<#2ia?%`e1N2X#W!>a)Nw_Fhc%$^@t-V{=nc%D!hVkf1MIFei
z161iMDz_1*=l0fiprx~Lz_%M$wKUv?b&Jnb7g7gv?eVlBidI?>7son1nCc4i&(a*b
zJ2S_T7^!f^&~K+Vu*(a1s~#t>O|q6L?tx`UU-c{(=@sH1YW;9UTE1}O5cy7&)O4Zh
zM0~H`BcFGdNvsJ^XE{c>o#x>9u~RHiC4296*zI?0(Y=sCw<%@in$Y#GllgJaWM`4b
z&F6fq8UT(3fs-hPLk}ftuL+sIp4nUCiSs`_d)}9{0H}ELIDB>Fu25jaEK~)Px>?L+
zM@}i!VL@+k<lOD!OcPl%djbjv@HQY)zR~srL;cAM9=$Y8Ln^Po>8ba)eheV9IZP<G
zk8Qtsrkr&T^|cnziak$@t<R+q#v>jh)GOiftw}6gpb&Xr6$ox9t%#~;F5!UQ|77-I
zj=P+M1shkjdJP@!j+b}N$55Vb=INX&)$n5*E?Jii+dYeHuuE3k9mEP}f3mvD>)fR=
z0KsaEHqiL$VBukR=ocQ3>=Obk_T++>A;L<c@6x{(Xwdn-?kC=#Z|^tP3uh>2Nq)=a
zt-pwj%=O7EhVplAjvh$tGV<)A6!F#o1*io~dk_X4twr89mAlzQ>G&=6;5gNGE>aND
zkf5(+oS9C*1ofL%#-x<BJeY<OqcC0kOFKa`6)W#*9~ZUGvw8W-x-MSl5#@C>p*&k3
z-A`1^^n8b5i;&bTuM4%jl=>vE*MI#?2ddkJ264A5vpXVfZl(W3JY4}n7ydO=KV7Vs
z>h?91W+;D~EU-alo%u&({In*f)a(f9&RJ<`u_~+I*2G9&x5i_Aqc+CT@%}Xa(MAF$
zMZ;O8ujnnJ2wAl!l_o_J5l>eP6?IF8OzVx(O>1)VbyuV+YeZiINzp$|a{g~S{ym;P
z8TvGjX)bcS_SQAl-%?huykZI)E;{y!q<}V^GF3!?+DF#hmJ7~wR_)!V-U=t^iR5+-
zlrN~Ruh=ipy_2yH^}SD#TiZk^6}c3(X5ybc+L|@i3O&0SdUHH|yg*a`YMVe=1$EKC
zDX?n~T;wJcm-?MI2vV+2&hUXx&wuR_o{H>^%Rzo5N?s@Io<GseAxct_Rs^2cDvA#k
zs<1DykLYs-*&!N}KDka<to>4uD#&r!q&EdNxN5uM$mOa84seG$vhFJqb0?Sy>AMBf
zq9vfxJ*nZMDu)lnxy-7y;N9(mky&r8A(De}&}M%}VT3z;DH`BLZOS%m>+36OV6sPH
zz3iw%+L15m1d5Fx`nrlj$S3ki(%N69yvbEaK^~;8_0{wZZ91z+^)f>=5OI6ue2;Yp
zFZr`>-p)R#It0FuTJtbR)(p}HbR7f)xW|3ujE5K|n)--2DUIuHu)%bi+34oL@$LkD
zV2tVA8r!HP-GlZ~lfIt0L%>VNf6chm@X?eeoX_0$a0?@>dw?FMEcN>?61vHzigqZb
z3Mo_5JEhU}bKnTr9P83Fmi2^<9^OcIq-`GSAoA6tpO}1CmR59>3{<uaCi3xMC;hiO
z5l9Gz_lTK0+>T$jR@PaD2^{fgoOb>|7yM9vE4Myv$uInbZ*VP9_#h<ehmI1)ZG#4J
z)N2rD(vvL0$GgnJT<{7^<uJIasQXRzcG~ammHS=vE3fQ!c^=46t>kBxsc4NI)2JQE
zEdGI2sdl;0?@r5M51)Qt@yEjteFo(mXNSVtykBX!FC01Blh)}TLgb0Yq_=K<tkMT`
z6q9h<Z&i*(4I&83cBUg3I{*HcH4o;aGd)*Ujo(9K-)@;WVs{qan~2~TPh1nW$jqLc
z^QMESo}?f`OYa0~267@7h$X}DZ^xW^{D4cEBX3(@^&~NWq*gKyZ*rn00MS5b)P&E?
zeR_RAXW!4|8J4tbVp|b2X3R4cbET2#gs?F49<cF^w!%WajYMlM-O{CMp5a3J`rv#H
z&kl*<SBB)6oGYRQjmCN&e2|aQf*#c(T4*XnR3oK*coj}sfk3+lA8z$9xS6o;w}k%)
zx(IgI`Ni`|v-Nzi;q#(rOqnKD<Txo58pc9&zy`HONXUMs)(OeBoIKQ$+A}+J^d`oR
z)QSUx-BnXJ)TkU>i3p9a49|jnTIWsTY{<booXGZy>>Bb1f7hKi0}D^1zkAZ%=}p_1
zMU*6OaF~(wk6C8-3Z7ox)jdid+c`p@UG-Pm8v&Me_Y3SywE6t%(`>U!y-Ll!oP%#0
zkXOD6c=ViO$heOF_@#WB`TMaBn^!DPP@0f;)^ft1ES&3aeV!wB14%0KWj_p^Nb`dJ
z_7uXpYNY@EIk&56^*RR+{#K-xQrKUvGdpOLLD3o&%qeokv<hA~U4h&-ufEq`WNM_Y
z=IC;>bUUI58kxl^f{_t@+=~QM{o9IhXFK1g*gNj%H5IvaeNfIpNw_Ylk3U}3(i%1R
z-)aT-M{~uDLdO;aZXdBdqQ{00bW>os0VLTimjADm0$3H0cxxopPhYMyCbN0~_P&4I
zboA{+N7X2A7A>nX8@W3*AN?r$aV1yIjoO)J_}MZ?I((k7(UsJ@6JG*Ou5Ta;?Lcp!
z^WxYA_^WLKC%nI<0gmevU)N1q%8N_lf4+JSXjiSe{*T!8|H=PxR;D1jL?%D|weA~f
zloW1fb7vM5oG>7mIY)TDJ^Iq5>5Xc`Lz8Gl*E!ttaCTA_rvv@=Z(Km<p-^ZK-hW<_
z$WT_G1h^8s;7SP6a?mK*V%k$~5gI#chtn+hL|EUej?&PMD$lbLfgX&fSC<E#FwhPO
zPquisi%>#3M%G?f2UnF}>TfkD@ZC?GZqVsVZNZ*iK~D*}ol?##$p@48;Z5}lIOez8
z2@Cfzf@M`YE3Wir4eGF4uud@DDR+L)iQSravTJPxxA`gAUH%1B<V<m76}C`nXFseF
zxhCip{Z0AdTABWaG@O0Wt4u!fCPCulDTZ)J*jn{1=epTu3K19YNz^47ZNOLnYPa~*
zjd^Mavn_62kLoPCwTN_l#0{uqHPFIsqR4%9EP|9V*VXR_?i4-wrV7r4sm<6{7#L;A
z=6eZ@!uu3&q?gsEwwa2T?HYE;s8Gi{_UN%7&R855_?6QgY57QPFz*5ynoNOx!k+rw
z`qL&pq?N6TdAc9Vp~Ncu3xc`l(Omh2ts}U6KXc96`QFM1OcY{9yhSRjy^N#BOmjSS
znj-<g_2T#I+M6p#m}tstCC9kww4k!A>UKz-LxYg$=G{*iW+t18Zoxic-xr&lEeF&3
z3bn8oG0RhJde4)cqt|IwAZMU9eXc-2NAb5<%KzE#05-D(z-F#fbg(_W#(wi#I|07%
zTue`ww=JzKG#sc<#lKu<d3t~6X`w>9c2aA-c9m>yVNAPD;IR^BoHR5Nuk}86qgpb2
zC@fA>(S`MQZJqxQWd;7z4CUoouZ1x1nN_5p+n(^eR6CA!)uJ+#K23#*n*Fx<5XK{f
zNN&H-D>gPY@rfC4(LPnMyv1xKUX_R?%%6{n+DD)bmTo=085~s&Lk||%g!kkyeOwA=
zM<7MNHBi~hF=rq@*E3oRHs^!9H?Q^Ee?8@BpRR%xU(NJ;k<I1!Ffo$l=F7bIh)hVu
z4+qOO+)|YH+JJb(>pVt4^6H}5;wh6^y!cRCP;<1<jC1>VVQZ%0!-2c->E%LAqk(*B
zhhAwd*5VAf+XK>hnZMW7D<&b7%=}Tr@O>h07KQ<HlbC0~I~80o?pFiv2Hr>v2pAT)
zxrY{d34XP(mI!(^#TXh=E9)>tP1lCb$=A)Q4vTjeHf+{(cjlA8_8nOw*ZJy;wJWP`
zxFF|peN?ltImyix)Ia=<<(c3HIYQP9vG$j6-pYs;Z8nu!lOn|T>Jz^D3WcUaoB5<J
zw+)t@9tW9pnC_%z8D|>O9Hb*;z5JTMM#$dHnmy5Xz~Cx$wDCIj<_=S4+WH6RQ2?r2
zGL1?k+Rf~CDn1(L$cIz#B?xjnt4o$v<<C|CmA3<qZJrKX{iC`Cy23uv>|0E=(<~gB
z;tZa4khy3LO#TKVAuL8ooGELXv)X7kd+k|ylq+Jlr{afRXaPvmxxST;ui5s97efgO
zc=V&%Z_yO5o6A$mZON#q^LwLC@>_wjA5C~QEkfE@u3B6+8VNjQU#({z5)`_#6{jw0
z?Ly{G;?GITY?Xh`lr=0Mu$^9!FfW#>Q5_Jd{^DGuqNheAplOl(93OQ4;PWKH(t5ho
z;&df7e*4Pny1D|*j#G=!FQS|)OqXM0ZS!gK*-<TY?*WGXCoAkml7(nj^V!*Jdz3_~
zdP#gE;q3ajR$q=MpX1i;K*P;DF!;Ar{+&?%jS|?co?R88yrCm_{068dyJNT5%Yp4`
zgBOPnvaeO%{Na!s*iRre-CF!}i#j^y5f>$pY8RcKf;3jYGg%$kdv3)0-PHMOedPuO
z31U#Vh}WJ+Co3b@`sLg1R`7k18EJ*EfK4I4;MJ{%eEE8j&kGGV^LxJekntb&AR>2b
z=lSU(Oujp+sXcE%93!a1I-R7H`_!gG*DDB-VY541dmdTndc=Yuyv#6#w=Vc)cGWf%
zNV8Jl3C%yr0R3&W0J(PH>A?L)?Lf{ixyub~ZnQERDzyoi-Dx$$kC7vxy3M@%;x)7I
z;L~e(mt2AEfSg|IYtoi_$A;_tiE>y1`NwRCAgs$Vdkns`Ucd0<iVx({F+ZW7s7c^g
z^mW8qA08GDCC#Yg3P2X|MGX#7+t$M&TM%QLq{HNSV+no|-yz3?F3l>Qw-vlh_)E~}
zg<XTxJ{I-DzVv<TsX3=1;pjRRNO!I`0GlrC{9mx?-&0_!tn<{Z?&)=Tvt}zkNy{i`
zHZLe1{#x6#Vu6?vhgCjneSB>egAA<2&RHD~j?{S#-$g>nJ%_ek%XIXPG^doSC205z
zW^IuyKuwGB4X_#g!+5s;)Bp9KCpqgtofy7j3ql94O?J>-^By4Y13bE|2Yu-Z4AwRS
z=k-ii;upG14yJ5nGRteSntg0e?Q6R~I7Ky8IkNil^vl`%LA~5SiE>&L(XponIqTWu
z)f=+6T{oU1k?EK3Umee@3$WybS@#z0m%s1&s>AmtFa3=2eYM}@PX94^sedv;{o^X8
zBVI>8*dp@^BiVoP<%gI4Lw<ZD#-#f2OY|CVPpN8tj6A9RffRZsI8&xhU)FHckNO0=
z0c(VTFbA~ved*@Um({)XZ7Zt2rKQ3`bz#<eA61H`_7b{d9mqRAt@2A=55kKs2XzrE
z)YQPR2EXhmxKf`meF5TPOb&=Lk?p<R*fE(lY?01VxlK|^*W9Xwt$b|`?#7O`kjmV4
zhKR?wokj?&uh>Uauo!=s*!L@nY(oTMRqN}Xs&#25f_1KKf_G~;#skxDY|tZ7>wNoq
zfbR^>{E!I?kB~IRyCN<@-Z;^ad5M7Zcy2Yk?j));W9(9GHKKv0RJW*JhX#fPVP9D-
zL|tH_Wb6zvD)f}?w0~L_Z|ZZd+NxBbx6FuI4I|Q8@q<_VP@@Sn(cCmG{NXu1PjqhL
z2bp2eae90m8F^G&y&^uBAn3U>r@NT%kyW1iG95m#$yj^euMg>U1x}+svfv57KA)s8
z{W%%#a7AMUNcwW>vb#$PAh=PhKgj?_ZP5TajKlf6Y@Nb3Tela5?)>XXKT6dKNEG4{
zOw6-k&T^ph3zT8veyv!-L`jT!@gb@dMh}}w;t3HHaGkmO0&YIlx$9Q~!XHr%cS<wi
z>)qx~Efc1*n2&}=97DWOyHy8xIRag9nH?d9sIa$T#}^ocT1({_#1J0W`)igu#2VHw
z<sTfsPrt3wrrHa7k`z-t3QYLuzq>%hC#%#LHy8AP_jU%w_UHLa(UUMb!;l*tU*onZ
zD+^f*pRdP0FQW2LCJh>izwR>lh7p(j?8l_n{w(xZB#3rO!vJk!{#3AM6<9R?*LC0h
zhuD?xpfJkG&A-A7F)xm?pTnCiKdc>*+i2oY(|om(RuN-G72E@h&$XW$Go;YFAWtcd
z@^XTv0b<zKF5jnOHluvW-&PAEsJ-wEYu8@gGc41KToPC&kGa1N4fM<f2R+nPcaw%h
zK<J22rv-bN71ab4*Q01H2A1|X7vep*DCCi+17^MZQEA{oOQITU$W<DP`LgOgHKWL<
zPd5iJz%9uz{qI_+{=#|AYF+(tpa>3-2_*~IoQo}h9_!F%j8{Cz?dO;2=61enwXkdm
z+M38xyw4N(gVZUE*PCKWLu$J5B*;fPYtVKnY`$$!McoSU0AVRO+p4Si6+?4E08E-u
znhLICfazY=s=E&|*!ITSvD}%B-FcLaMJ5PU^Nby*7&Bp1(k4_8u^`o4p1=jLQ&-ZM
z*_R`pvC`BubQX4Vt*fL0t(mM*<)0Yx<^Z{%ML+gozTYJDt};-)xDSM4Pk;i(Ut|xE
zBbsV==8sxF)lz1CL04E;C2w~|#Y6k~g<4G5ig1b{Wkc8LTT}KI^<1=^>|`ShPCmkE
ze4PMc(UG8qYEMqK_iPrE+2>CDjRSNBUAxukG$Jt=?uzn|*JrT`=u7eQw6caVb9Wrl
zD}6kKI~Yj=yzNseD(jqq5x31`SXKGndwH}vjtv6Yxhn#QtrK7OkqjrdvzyWfQX3Fj
zUmH?{uOYVg=8)LvgwX&a39o+?3Z7!Q)u5guWogO9*L;riz5c{v3gpvb#DkP7ph50V
z`LTTeyt#9Vg6K3f!FXgH#-&c(%%yzCo;~z+$HF<#1O6Wh%tTk5byL`I?BgLziEn#Q
zt}9Qx$-MPfL2aNhW?(1`=}F6&F)=^Q65k0os)A|c2IL~~B`HbH9P4eBaN3J<EtqrM
z8j4dG-GDS$`fA2Hi3{anR8%2mett6G9=wsBVOY227FWr71Cp+^{^IHQDJwpJTgIfW
z;?eL&GEODZWDsfL4gcCztI&0(gp2exA1y5+DtI~D8tA3<C_miqI7B(%WeLxLmH5?<
z2{n&f$V^_Aib}Iq{eEXpRCr&lWcUu`c6>wS;W4;BwT&{^66C^&T5ixj5DDFd_#?{N
z{ZEAr+nGMkkFFD(51DDm2`T$1gP&w_`g6}~Il0d0{%E}qG$ksGN7#7=y9~tt00>EK
z4u30E@N1s3>}|RkKdjkArRd)(R#nyfRB`Fr&gnc1)p|4{mlOq@-DO(JEzWvme=Q8{
zdlq`&91*x(x6z%eq(iy{@QcujjQZ05lQZ~_Fyt@H?%(~L(M$+<VGTT081qM$=jV>|
zEJOGwp|{iOVn6<dUjb!PcF(qvZzt!f<GV^1mLiz*B{S9R_%>;4GUK<bp0g`UKj1PG
z4Q!{VrEF5Tglk2$%6@VX%!nZODxDObjC`N+(v>KypGM@z=Q`r(2Xs8^`{0+4W7+R+
zMK#r|x3-BRxw~p2et7?_e#d|L;-J>Sw~Vw8oRT=%1-HKUzG~2{u`!8`8Q1LvPP_ea
za9f?>!{PanvM@uu*5ifGlINpUi%$G?4jL%dIB(RVs~buqy|$8_MQ=+VDh#Ll`9af1
zJ6&A(3@08`SvioMEeVaYm`n|jHGE(@7rM^Ch~vg_7I%G(JEf=v&9`sb&N_x%Hl7?a
z{7Lq5Lz}*7o#rF32o##1t?Pa|gmi;G-Y!@+3<FTo5Z12Odz7Y|Ible*$jt~8J2U+1
z)cU5cIQtVXsG{g@-~7sE)IwCi%PGGIbi|z$fz~h6E0bS3ndY@0VLSR&BY-uizp{~X
zQh%$PtV{Qer2|!JV6Db{#b(s3Mkuv>->6bVp43f;JNNGOtrJ~(VIwj4MD(3;1gQap
z0WTeaIKt;<&;f~^K=bo)JN75p0%chJ)fs9O-3EgpyLs1##3;~qD*xup`-Y0s5TrmB
zVnI!5pl~#}ck^0DiSR5v<rE7lMS=gc{ZtljNK;`o<yI#mmrz@xjaIId8wplnQ)=Z_
zQsU5kOVRM4b%}-UO8;Sh8Dg(p1#p+N!P=S?R~a31jKofy?1Sz0?OmSYn9?TKJbpmF
zDR~%nfIvpB`NQ}F?8bsn80^Y)#t8*d>@s-*+WHpr3Cf$G21M37U+{q#OmkZ?EtfW-
z1ahsl2K1~p8WOSs11-O>U!Mc}X>`w<<P)Qc8&^hsm231T;X%`W)`~g<2R<CmZT!75
z+x*=xbYxT*&R&qGHAHnUF*A4HwDJ?NQ(Xn-+rz&Z>i+xC_b<HXV2_{<R2pv{sHT-P
zdEwp;?)phKeNgBg+Ln2HNY7H~sVv8j>w(HQ5~_@4U&Qpjs}fOMo77jacjK7!S$G9F
z3aG1!WF3sIhR22Eq=kM1YQ}%U_4$AMn4JH&-k;ufhMoycHIUtAAyRAWEcnTC!t?;Y
zWs41EgZEapv33A5H|}fC_RDd+_lvs&pZ$Nh0nSE((<5xpt{jLvhqk$8%a5t@UVKtn
z8Hr&|uzo&&Kfzs!OeY?&f5uK1&qtk~dSO$+soqT!&(6&GseoGcs)K+VR36*OF<ka0
zqT?a|rNGK9%Gs5mTG{!U&)x&WiY?6nT<_&O3Jniku;36muCGts0P!Av*n0njN8&qY
zmmQP#e{B(sTcSf+{r7%Rw>}4`$%~7tI?_#a9>E?O#OX!nN^N#7Jy`*&;rTsJ*+wzG
z<9@C69*;&3$U`^RF3{XnQgl^j60gsYv6=JMG!6hd7MzaK$$pZRwptr`+8cgEIAW?7
zLj7kRN~(Wk)b0L!YknjqIdMo=?()ztej9lrTEn8U;+4SOaqr)6{+6EU%CRl1v{T{Q
z=(3ov)He>8mcf)TM!RX+;lgcHVNr5uf!!G0Hgt4#1!>7=r={gb=>a{@GzaJgZ0YsQ
zvwK?~`ef#H+^{K5kOupRZ(%!E9>l!NaijGg57r6aPQ2#?b=ucdw7ZrgR7SAZ!l)T`
za8*@&u(w%!b{`a>p?I-uyW+};=n^CXRP;SyzOT%VlaW7TrrZ!n#qIVcuLFJ_!3e5a
zHEe~h59k-$lJO6tY7yC8u~UY1ur>^HxZLG5&m%P-S{IGL6N;;}H{NxhuF3Ns$OT0e
zADinqA?h+$a6-x>%f4;0H`0S^SA9x|bZwu(&#DKu)2mBU)u6$nA}VaZ+-EvaZiGyv
z8(?b;*QN#ZhF>vLA4ZnkKB1by4-<f=71@?O8v+{y-FaoStMGoLy}B(7K1W!4h2~c&
z*fukZVDHe6ZfDtmlq+Q^ws{)b%DRqglJpmSp^=ie^Zm?^MGryW!1T4!I8l|TXV}wB
z+I5koIem)9bssMUR(lhzvC1t1K6BGKiQ`wk3dDs@gz|Gl9he*(&NPB2YMzW3OsL_K
z{N8O9g?Xkg@?>9gMYXy7B%{Msd5FKF+_60*s4a~w`oSm_(oKG}_>}Ano}M|_VuEN4
z0ki6HVL>5Rbw#wtoexDXmFM*Mcdb&4KOLyd=m1Xx`Cna3U7f|Auv{A8Ez_E$-2y{a
z6G2DA_8DO}0<~B)oQ%W`ONO#a0JbQQKy22h$t<8X*P8eoz@*f!7kF@w^3~^|;q^UB
zW@amg&xgP%{zpT)-wpcy5^(%a0Kt3ubwl>9L;BT~kyFpNw7Fg~tZ717<q^SR$FCf}
zzSzn%5<z<jwMUWTqgp&9JPhocHLMC4CwRKd<t42f0Bx~4p+*_S!=SljegUP`si{YA
z#DY)@8!vdJRa%i$==g2X#`+6`a&*~1Ywy)FA4?xA_jt|1m>LwB(p336J^<9E|C5@H
zJ3B>TOBSaaEJg^6s$wB}=0|(F&z&A$Ji$2UUf6E$8cS4k5DL`50gWP&L-})+gwCXH
zpnMVE$(5Q`G11IbHcI5Xh3$^2xeE>e7ie#@#|YaXFN(=l?!Bi`KtENpHG3uYif#ZM
zz98^P!X`5GHuBuRUs?Xu-ZM4%zQii-L=Z*nHw9_5+a_APn(Ad)w^wuGbmBh_mS>Y=
zJ-#KLp2x2FKdTn$6P{Z1kBe%cs9Dd=pOHw;GMc)h<Q_tOVI=n(iW6^&%h$s}!m<ZZ
zn^9v+W;dul-rehuu<m8md$YOZ?gNx`4C%RpOTH@WM$_rOzCeud%x>))1AY5R#*}K<
zR({$xJn7<NS6jRlzh*L~=zaALj9)M>-;_^Et(xRR{O&!(YsS^7C#k#?)QTG8X$MnV
zq|T0UKveW(1@F9#e&;SoHeah5m(FC0j+^x#(%)+7&ukH*=_qMxrezDyr*ks7j0-!I
z#ag#M7)dA=d|P0D-9cdFvfmT)T+~E~27N4=efr8N5e_s`0cwa=qCqC5O$~rho9G^4
z4&Lr~)t9S&uS`Hor@~BbLWRxwG37m%K`g35ffg_!WyWPlU=(;Yzd&god=fp4I1u#K
zZwCT<uqDuXL%bE^g>qe~8yF6bJe+ixu9}T!xltRE7;~YKBY07NdSRD6Ds+`(cEz4l
z2?VMNr%~F?UnAR=VirBrO2PK8P)IaC0K``!*T?V2C)}DWkdIJ@#LQf5u%%=5y*_kB
zT~aG%yN8Ifb$=34(eDg1ASBL?Y<->uEXiL_jM>G&UqTDrNvY;>j)xUk3O{~q!QR_a
zxzqz&6ja46a$^`)1G0bjy$FwThi%p#9*eH3@)oPG<3esk-wdf%VWZH-hq}RB{Vx1F
zMb_UPS?QFIXc=^HN;Kx=WDE~~<aqkIK-u#v<f2fv-9RA&H@m<8IzRNu$P*v4$|Y~g
zeUwBw>-0YDf_AzLD{t;^Y0QDoCSJc*TYmSs|1rP0{<7}r?K3ucv3BK{&4*p0TfJ6B
zo^cih!lbK7<6_nV(Sj;$!{<gmOhlzBa*i!m`9v5jFA@jrIm23YtgM$p%y*16o4{o~
zQ%~EKo#t<wL#BA$kFGLSE9Bn#D))8hw&sWPF1n?22s}QS)VE=vlTPJ2@%UqB6$cX_
ziTmQm-@kgi*qEmkKd>6Sa>xKncdED~`F8&c9WTBYhf`qNxKeG5tV-PoN`5`dyB6iW
z-$#j05MuUp@74?C36zby#B?$&bd@OmgXnVTJ}hwM#WnlRTX*3ggH?Nh8^z_3(-$ze
z&5iT&?vj1E2t)w?PZh)XkZsA1PM`b3Q|FPI>MHFz6twMuCm$hL@pvb5dDKbEHMkGj
z-4&zj6i623y%EUgG^5R<?kZ)+pIiN7wz5;RxGtY_=V9v1P34dEQ_$+G!5ST%jp;I>
zdRoYLcl~H@y8Z&(yt{vZ6(UR6@1G=!s?*1nM!N3a<nQGPmea`gG#PdfXN!N!@~&`U
zXc-Odt`{Z>)Zj+OvP)75ic_wR6;&P-WP7$&3MiqyQ-Sd^P1U+IZQRkK$wxilQ(dXO
zIOmRq2ifnBR6%2}5_%`bt2yl!+<z5U{|)0Zt$?nap26!coKL1T=wx_C{k|^Z-Un|N
zlPG{6RjiBE)%BH|w5U$!EBk67$ixIV55Xj1?e)BP$ci1St9rp|Zi21pxiaYb@E(Ep
zT>jY$v+sl<74%B6O8p&S-1d$7XWkNNeV1@Ghr)n*Hbhpl|FWySaY=hOh&$+*iZmER
zRDL?@I|rZG_*J7iPk{^6yk(Mgnb^>QuK*%s-uapLl9yX;eQqwtT<8t&Q9ZNCzk_%;
zgI0bI6H?|#n=Jf<G^ZGFKI5`pKYZMadpO$=3Zq{yG1%~S4WWRkRStdp=zd%rPS9tz
z^ie>#aP$`RIm%f(cd{{EJ7qeoYM@$@Q&-;5VGUm@TPQs+hKsf}S$P&z!@e72R>Ltt
zTS%tk6)gC05#!5y=qK7bfmWA)owkA>hWR7?`e(tx^&otFhGNbtYq(@_NubT}+ZeEw
z=0L;ey?WLj5ta!C>gJ4uK4QJ~>eV~lbdiu6cfZsX6(I4~#hC-FKi_`Kiv9V-&@~HZ
zd3s?LYOOKSWy-S%_b#?6$bKr$lp!`gx_u+w>fWjcZ}td>b&i_Vn9?!4#+WG*$uvv$
zIcgQ!ogWl0ua~n5P>=RC6`hlSeD}Y0y!W@p_a8J7MgQtV;GP#A<9)`Yv4_Szm*)!$
ziVal84acU^a=%~^!^GW-xj`g^v}vHdF^K#ZcDR}`V@21{iRUeiZxrHzN+cB^yZV#t
zeK6FlNf;i^%t2Q#qo=JEU&W$2^)$fHIgsN*uK=T722hCuss_T}DQ}?+Y>t?fw~v8G
zo<1hD#$2T?RNQJtqq4fK<NF}dIR9%&|4a0<=Ro`UGLr18C;8b*>oCLnta*MtYL2uy
zjY&@sipa;C0`qO7apnM@-bgE7db?Lg3s?tAZvWwN;AMG8b2(Z~kxU1nR$i$^qe^%5
zX1VkhwvUL5JtZL5`E4*M?iQ6<PJ3$}62}58eBfZ$y!a`%4P6JZZl8jkaO)u_<JYY}
z$wWk43G9AM6;5;IMjtJvHYfFTFfN9%^%7(r?MxG_MX5c#t;i0Ww9p{{75{4{XSiNp
z3-@m<m=bFCHVJccaj~>ysyz~@esTHrb4)yA!JGCWU*im08FYEFn-9BWB4~ZVyz>wl
z89#3#f@knSd7j8?VMsAuX54Y5zWtZSVCYTE6@+QYyHQ!SrEBV2!^P7j?=vQ}6ve6;
z`W2aquLCj7Igcwi6T{APxH~^OBx)T*u6pZ#Z(sqNK`sy<3HW?zOJy5AJ1bC0eJ)~s
zp%A*d&Og+>_ryw9f8yo36j%U}H<nTvF2Jti#jmuoXblu}ev-9;DYr9E#|JI!<soKv
zzvS*(^gZ!FNon0bM}J|_U7TQ^YwMoWno?eI*l#@1k!c|P;cnx}J}MC=dfIqG8P2J$
z+z4<iO(C<fwnsqOyK@avV*%rShO(0`^SVjwo;ya4%-7Kcx4k}iY~pF{Hi05CETQ}x
z&`}XEMw1)jdm9EyQ-N&<VK?Y>I3oOpu|9|SSpC*+&+Cz!wyZ5F916HM&VI#h{Cke`
zJWXIGUxA^UXdFency${=VJPnjjUVEwc&Fi{8dRkB9Ev8?xkRb;F;US^!M;fve0lJs
zmm*!VYz_+6EhYAeQYV>}9Q#PGCn}R{U=tPHKglrCX$@khRHx(9{CJ0MbL3C5R>kyL
zzIIW2PrpX?3QqJ9xM?S<o)(WbKL-FGc8{1#uL`enSM5`$BoQfP=kD}zIp~`9Y8<EM
z9UnkMx3Mn@*;tXxw(f9Dc3Qt`m(B+zy1JXbFE)RhrGL4$_t$EEfBHFfH7r*k%xwzf
zR+#ktP<Hf4$!oi=)hYg0$s_=!jw?Fre#x^dvx-g*ndpC!<bdWFy~xukv<iJ%o2%G3
zuKd7u6WG)5HT{+H;-8-C??2{~7UT-rsYVSAPF}oSP!j>0W<Wb{?a!AA*4}rLy!`eT
z;)dtky4u%Ewj6h2?m(+P-ag8cWL+vF58R_m8xa<7iA||VAlk`nL|*pGvRGFL>^bF*
zF1c|}e!SGOzA{;M$xL4KnWnK*=7m?m&aZ%_>9;ikiJ<vO*4{F+4L4}}ljD^ZIBh&A
z<lkWFb<E!fOO1A-<n-!ku)Whcre5J{yLmKw<8LIF^*ldomKUF+*heK5oKoVe>B1>>
zV@{ON)_Vr~lHBA+8q$T$vySa-s1RlkJ;{V8rR|CIZ3@ro54IS|Mu!UQGPbY1L>Ra+
z{#nQ6KP2z*mvSNhgwFs;PYLR;d)Fw3oSs*}KrwP<Jf5dpu5o%;q>PUUyS9{dM7mi)
zrn*FpZ2nb;5&CO9m|{m?1>DkC*`Onzk*^vIEH?YX@e<|2IoWdd3`}l(&Ba+GFFyVO
zjVzyD^9<c-I%TVBC@qJ6($pZw`)!2auj82YaPK|t+T9gBe)YJLoIKUZ1SQ7nMBU0q
z8PUd~@qjNYDs6OCqQR=DJaU|)wfxKWQ+2Zi9&zq{j^P56v?768w{#QFcj~NcEis4S
z+C0y`Y)r+*qrqR>Z*~6=yS?eeDF=ulSZzgYtB1oS4<m(3ENLqyt-qqeU{bKg!EwKA
z9qgmPdAURaIpU)aImA^Hq7sXvFpVBsY0{Y5U@m}k>^Xh%qj5&-qF~wck*M_PRmt5k
zyDKFIMlG$Z(mMVhUc=DnSx|_=b-Z={&C|C4f|^}g&{Hr%><D?29^;z{3IH0u84I_2
z(S9W>KNyw%i~~9=6ags#Y?c^mD?WT1FAU8<lzpRS<U~UHUkF{b^nIwYUYYS#%Rs7I
zhXjK304+Z(ar4ecTi14&hT#UFXTG$gN-+ovn>~N)b=NwTqIa{TcE{#SKQ#caFJ6?d
z&6N`rDN(wA&QB4>4SP9K?R%M2;H!^KeO{1V9jp6egnYA<{n(t8bZl}RyUZjM+{J;|
zm2cP@8SGC@g|yH|<j$jwfi@$Oa$4=+9NpHLIdHh1s%y<+m=?Ln^R91?<-hHt)Yc&W
zW<_YMt})FMc8hA!?Ofde8{R>WC&0vB`c+-~FRN_mnq^#W1d5StGO*hIIiBN0!f+i;
zrwpaiWL9NpskNU#td+z<^<g1jrKNyrRe=>|#ZvKE@$h_YQ*4RGL;a1OI3wdv)gJ#C
zHq+liivO(d{w4zDXJ|{3p9z)|_cAwkIaK?kPt5~PTrf&(1BcQR!6g-u-zPhLq`PY9
zrUH%te-2++9D>Nh$q`5#?{vVAPqZ<8bZhnY#l^n9TdKN3jTM@!ub3tDD*0*Y2hywF
ztlz%6vU4iFsiUcT;kI)8#hMzRf>vRj)#nXU9B-I*h}Qz}=M@4I-$81W{EE~lVp_by
zGVMMz&~zD`?~de~%?({6w;9}iR@y&{Cfk}Ob-U=Vrqv#+bWSC6Cp0QD>V*FRaQ>xN
zUKDFbr6rC0g>21xEQtn8b>D7MCXk8;N(9SflVLhu`RnLmP2;;QPdOYOzO$z4dN%i#
zayq6)Hu<Vx`wkb4ACzyS*>{<e$Xa6>`o3O(##o~Ku1!8OtM6eu9h*7fo0IMaQbom>
z1=zHIH?RAjzG9nL;+osXE8dlTRQmMF<LM|tlSQ{(s@2<GXILy>-L7*>2syAk)*-+a
zdFhg(wN#a2#-@0Ewj{Q+)nocDHifx5pww_msH;cm9AfLtc=bs@60O2eF|VT|)naT@
z4paObf@MuX;Klc3=Tb2N^bn`AnDUgWz&fWDAPiw};SaZ?8^;r{kGN`!%HA;W1H^Be
zNybf-RDMd{tjBZ<8k>}kaPD!$o5K3<qj$#b+k1>_&dZ%gmEA_cdHYIfVzL!3rK{yl
z3k`F^_{oFXl!nuRH4!M1%P|z<L0SpTMuXOa+c(S#xV0;|NQHr=WuHxs?$?JNsn;ZR
zZHrdjdfQb5)`S$p<nS(YGa&}jgT8=3OOD}z`bHN>Gx7kZ-_|h7rTwG#d+x05gm#Ip
z;Hz6d$&3-(U&s+{4e$=)z3x4+qT_dm+TPFtP+xXBq3?a$qAqrt1o8$)P3qzK3Eofz
zwu@Q#ak?!Zx_I-Nu%#HVp0l$-{REb$rH|HEbzq?$dzw$X|5UVyqfov)6^dJV-tR_7
zH=|;iJJxIN!P49;vt^lo`U|eSYF6G%K>Pj>kZLxUrhhb(DYy4VgxD{qbS6)iv)Y$#
zX>i!D@m(MP9HbPrql|-$d?J7qeRZ+ghr~dOAT``Uv5Lnzb*0JIF2fnzIK2KSAewY5
ze!f>i{-JV%zh>D73pMg>COn1Xhh>(VPoG%I3dR_;o&Z?%<NDQ(={WS(&_G%!|4QGz
zKis?j%BzUZJDEddou~<K@b)ET>jKqZ%@Y)DX(c}QPv(a(ilq2YYUjrcqE+rZr>kRj
zD@?^~^LI42jd8R)nW!6xWDz4jc0NuDy<O=1&w-$P(mauu6JMuK{suXtelcwRQ01~N
zqi1wN>ZzItZYPNMv&*Zv76X`*JGT8ocO|5)v{3G$<CO>G*AM5bs<^#AmKIydt7MHB
zy)K3@RK@0;u<ElO^o&tG5UMcfi_!nMG?ns`tjt9{t%mDD`<#+}#joYqKUaqu17cxc
z_0yFDpz8XE!$MzW?3gA#x*zz(JcsGu0lp|;Xw-90*pSgWv+#J6p^fb&3OoJCL8HO0
zqR90D*F#Io8@1_&0eDD(@1sF`c*(ZmFIY^KlzA$8gm??<w(^$9VVeyx>Gpb^F0d@|
zz#6YS@J6hbYgV4^8_uXbV_cxJ@<Yc;Zs!(4b=Bst?7V+%NB$T7oPX+hvwLOq8HKuy
zOGnMS4-?++K%O29Xc}q**p%URE;~P$(&$T_6IRttA+b?M{beJ`qacHCb^TW8ZlUk~
zm8hqwH$?kazkA(p`3_8`oHse!*9`N$Or0M*&b+AY@2{o#n2xz0cD6GUzlIhCWV_7t
z_ZXICL*zqHHk!L{SA`WlE84&zoY%1x!FRUA2v&y9--5^$9MPuR&m#K0ds7D$tvx-S
zA$0of1q)w7Z5P9@L{$5S_9HQQi$I#}|6}jF!<yW)t+Ap4B0{8tO7GI8B_aYM0s=zl
zh;$+#y$6*hEeHq*QlxhR(!2C7ozP3@olruAfZyvmcg~qPbI#m5bLW=tx$_543UAt5
z_HXUI*IN5mT4-uTx&JQV8#HqwB^A(s@09NwdaqiRCEr<34>39+Z)LO?u3^0xlU>L?
za-t@8dcO{Y*>3W_h8G@A8$8$)hsmq^z0cmAS(ze%cR~d*=jK{pV<cp~M4NArP0RXx
zmV06#xRzN8&uyCY3hkn^F!uIP1iZHm!jP$8L&k8`W&=m5F*qqctSy1mU`SFltHmor
zzsuB#dh+vw7L|EKN?k9G`q(A<7#u$DyXvdEPw!!W;Uz$=tAEbp(4b_OjWdMjr-xMf
zI^4eU%^5wdtDQ(aZ;>uj8s2(I_*g>^EZ^#?>}2^ZpM?>%-G^HlYBdE}moM~=hY~i^
zxC9ARl+v^A(jJRleDJQ~nG`~t(W>4seC84yjLV%}JSvOwVve)39?*K!-0bzxQ0L?f
zj^!BY3PpF{8S8pkWl(bO;<$0>@<b_|sF{vDtV1tX&0rwRsp`P6pl$M*U=CH+eK<f`
zAPwK5Pz9n~QA)iS_=jQbA9^kkM2QhzEOR_@OPQbpN?Q9QT0@!+nKVHIG}bSice*1#
zIa&k530lWyx=dzMJ!?^=j?OeS&<|?P3M;VA|Mk^SHHP;DCD3qbOybcX+BcDAs3iMl
zd3dAlBg$>ZW#RRbKjlV$bql3L&s?({ai1#nMcK8v2;jp736z|pfI3kN=^OCPVKAFa
zRztQDOM27xu}l%`q`8EAxovA%DjRi1RC`q1&;jXnS3@O?Sv*!Js;Y;`Wu<AwYUyLf
z#OjSZ+T*Q7(pP*nR985#&P!hhsE9*3Lho-oL|Ap->5zQQbidn+YrNm1dvTXqU1zMz
zwc+{WV{Cpf-_?+d6*lghpi(-HE1R-cqzDQgeaLEiXa%FO-ZIs~!S5FC(QmGfYF1^b
zz9L{4DuI6F=(O22TV(+nc%VOKSo}fABYQ8Znv}0v)t_Gw`23^ckj4>bDjU5zC~|UX
zOHx9}<im%DQn<Vh>jl5Jj;3wppou}|Ko%+|VyPBq84wG`%51eDfK=9rlIB89x3&ZI
zK0O!jeA6KrKd=YgrE)*t4YFT1X1f?t#j$&TrwFQ+Sx?cqa`SJUsQ<aD$FKf+|Izm`
z#+5fcaV@@IF_;<g(D&7*>*$+I6r%Kt=PjHR<h5<)kZpmY9=|Dq0>V2jrz{@9HM~k6
zlnT<{xOg9NIvklv$--8}C|K&3Qgc=%n1RDE=uP|TKn;U=1UH3!C+T$WV^lrY4Gm^%
z=Cja2fTg5fTXinf4+zYvK(PF|r@;Qa#>YoN_NL?^I`c~yA^zFm$c)|~)82dsV(rXg
zRpexN7hHI!7XXSeErz6ERQ3wfv4>0uB`bg6JNNSdb^Q^0uhp?+Sg;^>XjVTkGJP>R
zcX49zkO3^n*<?&^uNmWXbO^c5%d=P-n_Lk%7o|)iMv+UJB|PRSglj|B`yO?>BVOky
zUN#`UejP)$T7of5{Z2pyF@;>f>y%sSbO?DpJy*8XbKll@>I62s9a8_fH>B?B$KdLZ
znnNWICqnO!DLlFS<-8Meg*Y=<0M49wptCAx-{0t5yzc>U*CjQf!Ci}+OgZ|{gX9Aw
zz%lyRR3`@0=V~DfCDID_h4G68$($_NoJ9?ml{+6|*QN3>Q}>#<xVfoS&|Dann5<{g
zTbEkgmm3+!U0Rba*=W4j{-`>rd`{co`bzHUm)?tYD&U>zC|y)13qY<6-`w>Ao9Dhp
zQ{V4v;VL*$89w4_Iwd|GnI^^+mPHHb^M)aNc7thsx<f+}GCwiu91_S7JoTp!iVwmw
z(Z`@akAL;&{pcIxJy^9TKhfW;)-03wzTgc22+w``NAT#6*Kq23Wb>Gi(moq{s%9x{
zMhMNc!pyL%?<!6ySwOGkX}nhV84a2o6y6HgV|)gXa1Bgfi1HSSh8zjH2i<G%mu^ol
zIQQ31Z~suU;qiPv0S`!SUaPFUov$)rC#b6eLzf?Z{XFBQQ~tc1!`QWj@IIkns;)b#
zTQKmvL0P~p1!sCc2m2{SuuFT-hLWs)cq3m)cLK1US76TXF36oX-pfNpEie}sn-2|I
z6toY9fGNx-Jb;w+*Cg!sb{v;^69j`1Jx&@aEHH7+&u|^=4uA6b)9G!(vlZJujFfB`
zwxCieQjNjst|xC3>Uk-tk27bcEBUWKA_1p2SpE0_1F@n1=ym>oX*;tSi^3_>_b(!c
zTngZODsdf%1K`2{I_*z8n6%&LW*KekKgC@44|{6P&iZm+6Y%>BXydW{PT(2W98LK;
zsA^MqO+A|?>9S8SVxBX9DaO*9Sxs*ut(R;vq$xS~sitpi)!|!iPr*zDH!p#vD_57P
zu^>i&PgNsy4?kbYH{T*r)kFZHw%}rCg)mFLBX!2I;=8^K5Hbxo2rpt7aHPxibx>ca
z+lS2p&&Ac7IMU~?1vp3kY;^q>mP&+#Z_Y6Inm~O#?N7q-JQySMp%<Hj3aC-j0wb#H
ziDUE^JbFMY&jg7zwY8t$iK1h!hO1PG6;ZisWqF7fRY7ipxjipqppCTOn1=J1Vchtb
zqHKP$gM)&zBWo~@0S`9=>QsrSdFE>{Y-BT-c2H6~y(@EVMAcB|p~(AK8RON&D2!;E
z%IG6&$fkJx>&f0TEMF6sqp;^u`+Qe%KK^a#{yo<-!QTl?kFfZ#*WCh~72eZECo5RM
z2e}bPeryk8xS{LC)Ux=a=N;(m7EalB!nGB4-t|xh7nqun{%NMg>kPGN=xY;jhy)G(
z<?udilkiD&n`&wF*EIo+)YwN!BUOC<0F_Hpey3O`K((**^0V<V3nFsz6KJ}l9pRa0
zdk<@r^@hd6^ug5LF*!_U5;Yr3mFsI6TNsN<pgrE_ysb$|J!KTVBn(Ukg#8;Ad)qe;
zR=8H#7Zg&pnM@-@sG;ib;QsXq#dJu@_alz{?<Y-h5P3kP3g$#`((Ri%iUX2w?)5*c
zZl1ZhKu5#=i6jAHV6hX4?>bqV@i$30$8=F=ZmlN)UuJVflzwQxfCeaT8m9_}Y^F+Z
za{3{VpxFiq#=haBJ(eH5%{UKWIaIro@(+cye|pdLtVHnJ&6^RH@$wW_+OY_NbJ+!o
zKu50zGOr}J)umhmO(f(h3Ez7?=^iwVVelagEP>^VET%m)zn`A^JfkLa4!W^N*fc|~
zw`Q*|bgazN7;l$&T5@K6chxy^THcAJ#@3Fi5#aD@X8n!4b=gyzR>6%XCHnc@vsIk7
z(dma#7i#Tv2D3L7yC2606!mK1ycWM@N&x<D`8E^dEmqaW_fj%a5*~fduD;YF4=fj4
z*MU{*cQtxv|BmM`E%>qMiC+ZWU%prNVNpKnbknjQO5F(|B3IaQ5M^LUl>E@{bYkTT
z^u+g=U!)EpxJolV=Y|OQGoY(dgCcOHf{rK||8-fXJxdW3qr0Scb&Gk=Tom(%g>k7%
zJx=l<!a9GOk9XMfFPBJU1vUm4KCz0A-fex^Ux??*5DY0Z5gJ)=RF&2GXZ)7G=Xd{h
z+bcc~Khp%(9}9WEm?-|*b~mq{0=F)=W7Tw$?N}t);}=3!eX`-MdEyq<66TYN1|1|U
zP|8iap2|;)lBDh4%ggMcnkG`sHB=wqu}P^b4MG!Mm$~h)N#Bs2%hZ@i7tk07DvLcI
z@#r1GK#p5s8&wJ$&i|ET^vm+qF&ngYBgl2Y;)79yEb*ly8ZQd#tgtw+Fua+866eD_
z9uuH~XF4iUd#w7MASd=jD>2E>lMf?08x~x5eM%fQu2(efKVIISeMdBdrrvPPzjp3K
zv_;%u)C&7#?MBEQ<BEH;9wGa}M`ki@u?Y1pdGh0N#Z+9xE-=h$T9l~?Z7UAns3)z)
zhqY*#xiV`%4^MRx)0BAwp*iB3MQ&!Jd;O<1lkz28edNv)6elHuz7v?!d(&Vcp-F{2
z8EGi@@>k2pLb#ce82<L~>bbBhT<94}S4O=ZhBPeAR)Vl!V>}$AJh*S;UvJ+uzst+t
z<iTe7oxoY=5jX)3lR>Ykwk`^bl^Vdz;01t!ZF7z^<VF!ak*Z<2z+xWjP361Gi~C6F
z20CC^VrnYPJs<!Xaj00E*tmN2!N@L!Z8sE{s&&5J7Itz~LW!yE#&-hS618hqsyqFB
z$B)E}4d0tj)1hCMBcmXejEmma_W^Oc;MyHYY320fAU$Ng3XX7kUz5<s(`)WK!CM+D
zLkpUC(VfDt!$vh;(5(C;ynVV@|Hn)<9tE4kkfW7OLA`z#pgE^$3V@Y=h(Sa{V}oRg
zykMKir}AhuACM7*NLq*MDJj9+c;a8DzVL7sZqvP*1EZ%-Gq<K5?|&+EqQf=ZAi;dP
zntQPLh!>EZD<A*kp*5Dd<5E|lVRl{@@|rX$nGXj(XI@zy_ar`IawO-FvQO8%J6!s7
z%dNm>S*NmsZE3XjRSs{fT-WusA=HaIVVi<%>P59qXj#H*tX>mA9Th{oyv23*p#!<9
z`h}^NUWCMd1eTRD=HEkyzkQA~F+W$=!rc*79`QDRPOZE7I{{~<9zZFVq_>_Ay=X16
zqyWqeMuf=G7BN*|v!XULW+fWoA_Jb5W+)odFYj(O$u-*=+Dok$>=eXU-pF@wjmG2;
zwt~{nMF<+`?j+4UHN3iM!=MmySX@RQKvd#5ziLqaC(5Yauvo&4rBb2L$`Wx-_LLf*
zwBLTce?9)0fX7+7=nKsaq)hCY&g)karF5i?&&P)HWN6FnAAl0{Q3f?5n#sv+A_jcn
zClWOKmd~pmH$FY48&)6T<>lV*3e11LS3q|8;fa#Pq^^Qhi-}t1NJ;^~xVCuiD&IKs
zHjx7bc$?UI5l2i2danAE(Y+)<x@7Te@Z_6CYdu4Vv4}Zq5X3DVBn8IZ2+VKDP;&N1
z%lMSMp*SMfmi4%M+?}~CMK;|35b>Gwb3$j+%^;o0-yv;roLMxa2I3avil<qTb6v5b
zk{)T~6Tf?rqvWfGN!~ONGYANOt0Ec=)2!=nqA!1!eUsGvL_3Qw&$?7Xo2OtFv2^+3
z2m%Ptc-Xe9cZdUVV2%s+eSOsaz88e4p>1KGzdYj5d<Ng!+~OSS!@=2Ue`z-8NMhD-
zGZwAlziFGk9sw>*;TV@3(9w$NS4sDPw0VzeHff5A_r9Uhp5*5p_h5W?G#iHtZZRMm
z#*^oCWWq!QwI;J~a%Vjikq9|Ju5gP;Fwb+deYRArF^CAu{FXWkL{3v-x3WE5!2`hb
zy``4~4;|1vl&MEpo?TM(YI662l^|P5FX2!+JG`Z~ITkUyWBOkvhEnMmvsPeR^{<o4
zMPkseZPRhhB#<X}4NFNY+J+L})YGDQ30fA*n@tb170bsBm%@-;dutk-Dr8$-_423k
zBk!Ny5Cz5;#JgxlwmDee>{KMTQIiT)F^nh~ECy#w<bFyCVyF5e9{E{$z$@r$5>iM@
z;nei>z*trLYrGZLLn}ZvStFSK)2xYzifY}l7c-95VXFG{{>7ltD+_X`CtLu9^T&X-
z6hRdSIUskJxd(g<!T;nRlKmtZuYx8e!HLNArG&eLC^=D12YSicy4KGV2H+|4$1P(H
zZz}OOSudL>3!~atGD@En_)AnP#M}%KDEve!jys1AyXKnOP^1Sd5?Q1*PH33Ac_6%c
zKw;Npe(Awee-^4Mp9nbpMrw)0F7K7<8FAp*mC`G-WuLFQr=iMC_#2J4Mux>22A!WJ
ziDUuss{23F-T8-_$4aK!l*9s3t3yvNhzWhvhAc)P`!qOeM9Cd!KPSPXie+Im#fa`0
zimz6NSCxYvRh5^jMJiQ=ZQuXmjxtryCu>o%&1UGKlW6TfS!CMDK)viwB?b}h<P2Df
zM9`JU2$$HwasUU5cPUktl2R|e6S#2fFXbxC#d~!G9tDj(x2#J$7NdHqnd|Dj;5CrB
zW>r%qH6&#`b-FE_lS^347KP<dRr{#Wrlsfo#WIu}`4lis+v8mb{Z;Gq*CgX#;(n^o
zRd)wqiN_F!_EViw%)ula{bjxroXM^_%gH&Zqu)K+EDKmW=kW#D^y>RyvYc+wkA5>W
z76=+sA?d;^UwB+ct(-66#jpLd#WNKJRE05M6S{qNSYV|L2G~%@FTq5EUY&dt<Fs$x
z^%`c|Dd?17zja)ch37Qnk(WKUlQ+%dDC+?c4ArIeKQ1iFEQoQQ0bHI(p8+n<UN-vI
zPU6J0_GvTMs==i<cNW_Y`=Cm)@*_E*jsBi65<r|KrGAiROESqPFQd-`f@o(?O2sRy
zDP3K4XWdS}cVs>3`GM-nMpmv?<-vWd#ayCDn7<n2inR;L^}7wSOsh|q`mIfSx3H0u
zn*A-vlDCdCGLb1|Hj<*ZBaFP}PYSqrVcMUbGl)=aBe)8LZ$bQGD@rQ?{e{ktwgtiO
z{Vjso*5i@_Ja4&qq%|y0v9~?mB+SNhy~h#F)4R0LSa+^r6u4Cr`llfn?_z^&W|+_m
zL~^+QmGF16vbMI?89Hy#4>vdG>0c3CX=&KoGnAfDB=6VD022E9Ba#XZDY{d+`Z9QJ
zT1G^W2MVgg!SUke?;UBb517Ly2#KfiAiM=F#Y-Yh@m!4go&m%OTsqjV)1RwODy;cF
z-lWWOEc~uqnx0ufST8bkrVw5<XR%dkdqIJ<^jW40)@^C_tXrG(YeSAT9sQ*+H<t8+
zlIU{&1SCc98xr_v&>mkfq`2M(en&Sl*RsDM>-+?0pQ|)-<{U8Q1p6YO*&XNcq$jnP
zamnyHz_5p?&toQ~uEZpQ#;_iY(RCMa!DvSRqs$4sRCaIXW0kMBJ-NbUHVv%-7F4ZJ
z?=`8!@c9qwiZ!;8$B)1_XJ6{pDS3V;z#y^W81KfYh|^Y9REZK!+ZyHADWy#)y=`#5
z(>Of50jln=KE(LwL=hBi;%VfM?YB}%n@@{DeiK}>e9m8~qzrXhz(QUKO=))N6a`@$
z*MMjxw-C+RZlsyJD>?_+JU373gL=L1C8xk0d#GqGXrmzY*&(d4ZWqUKY~rej{wQmd
zt-u&inqO&BVO@A#itd1Ip#D;C1ucKawq_8vBt`@1c;!@qXU-)UVVpxw#@se{&HE_S
zjl)?3Swcr1#@hhJi>4tOC%C;AAjkN}^h-G{T6gXm^#zR&I-#qLD;69{PlW?zO~vi9
z@^1lRX&3hO+y3g6D)%{t+eX(Mw8}eO*qnq1wAS8DBS}h+vyq%YUlrx(*lH!)$(~Xf
z{D*~-zv#eZvDLkqQk3*TeA!cC#h!J)e{W%j>*{i|Pxuw{sfxVZ&`0R?6BFMs=@us$
z5;5E4k6GaZ`mh@#pYX>M$GIb)ZE9q>W64N@*W5?sYe5slLQHiT%GMD|b}UUc;bXm(
zSt}`35oK??feY&odxF2@l&XJfo9H<vf$&nKfZCW5mZxugVv8^C&DhonRFHQj1Yy=&
z8CZ7kSXE2@9$U+It!rS!%oOtB4sUw)+9_pG^>BVm{(fLg2(f1O$c}+*nbv@7FVjna
zkf+#3aPh)}XG8G?8vBKsz5d?wXxl?bzVWYjm6>oam@rH!K+U3Ayz%cSPMrNexa}ny
zYvZbryx{Ql>B>!+2N5V9#;-}8RqkBwPU$MvZ{I&fP$nNDSunO=f+1wCRXl29aa7dh
zk7Jja;sK+#S3^Q~3+yqW5pe&?iC}FW6Rz9uykPwy*(gmjlDi=8O|{mqT&`VIX#Gz`
z9e5=r10&2}GtFsWviu1mFTCB38CZMsXZ*xli}PI7ruD5Ro9r*1yCY_|uFFwiqeyPe
z>-<zz)&hv69It+JLB4_{yi5XtiExH!Y%mL=+=%)^*wnP#r^f+hv;|z%DOE3VL8CfF
z`S+CtbcHLU9=*(`yb`pU^ue39-h`1H&lBOX38oz0*I*8OPHk#vH-vgEP-3I;8mivT
zbRN1{QkaG*vA^9xOAIo4C+pdOS;n&T)i^D$o!)rbUuql;b3${FdLH(`F->*1Tb-NC
zX~2<&$Mz?xJKq9GE2cBZH74+3QQ(WG!v>r<hy#7@q)2vhF?@GqcFUA;gM6U=-YFR#
zzLVLqPT?hRbk`QUxXJwD&XIXOTo(t-A4k3g(^&Nw7mO-eOuSW~eyGhqNz>f#)FabW
z^<MPIbg~Tkvd}S_wc$K;WZ9E3+71NCpVRSEuu7*-%2{8pTzppiLfMedT<UohI5A+;
z8iav-37e>RTvi1^kRQh1UQ?Wz3PgT(l=4u)@W$Rq&d`^4vt;k=AUA>TA#C5x?~2#;
zQWNopzBr*ix<O7aL(5>LVlbt+uPW<G`01~&^MAYE9am*3&g5sVN9fM6bY&vlXUeM(
ztLNm#KrK(p4%Kuz>O?tVK?MxIn#i{op;A8*@wHFa=4$^mFm#XQvNT{JuOtPuG&%n%
zrqO@+Wq%1$e+pLrcKh*lJ7HXEu2pd6ST_c(Fsz_;qU6~-Yslz&I*+#Ahbpp-nK&%T
zusX}7CNuh3Qa#e_OEhr-hPFb?r$k$)S<6O9CUohWHm@|tzwQOCXhNgoF2>qst`=>U
zmn*#`O*dDPj4U#@U!G1y%28?Mn`Q@R9o^uyU9Nd<;t;llh%yM`Dq;bdm{{SwPr+dW
zRaQ%$5WpHzepO%C?Jg2cFLZhh2y5^9eh=dOPdxYd+<07HsuSQk2M!{5o%hM6F!N-W
z^Tq&=VV>)e#yIEDLfP}rg5HHZkgA!(Vc{AHEAKi8%}`42#M4tF`+`;(SDP7}cECJP
zUThRfNXxI{tgTf>FN7HrosuOTSs$sZ`0JL;@8$yr$v-<P&+LRK0O%5R`YrX8ED*9>
z^_^fwxUXM+#Om<s!}voy({=A&r|$%H7J!88xyrY#`okVaA=QJp+NKEBvpCc0ucUJ-
z3^WG2Ba0M2-tpUQV?O`6tqpIQa8C&<@h$wiW>Au~dWC&k(Z}xumrgQMy=R^-_H9z)
zE+579rNopb+3gmzXbf7%q&<SCq}_OTHT>pxf>%LtK?qB5>SwB3vMWK5E6xNf{ED)-
zfHK{~e|JrZ>(BEcP7bK}!Re)}V?;dDAz!Rt)q-jYYL_wJmpVdODn$r~JAd|PI6$e8
zv0EI-*}pL-k+of8U?hntlTgvG&AqBMO?^c?%4igoshnoT=h)EFX7?V5n-BjaA^6jo
z!6^~!SX~}LebG4FxiFhOjWFC5HcOHbJTDQdv1K9>-m#=Bm*(HA9yw`N>m0P>9Rk0D
zK}4dbz~h~4<t5X^q?hh~I8@gH>yTB+!dN|8K=Svis%!c;S?hc@P>5Gu*H$u+n(O)s
z0BZS*#v4#G{86ubc(FZL2r#+>aJ#9EG=qN)(Y*PF!yIFf>ZHMHxcuCvQ_Mo#lt`n>
z@k8;*b%XB&vhG@dn^n|^#*WjEW?ZiM9?*@m(#o>e4DJ8Mh=E}}FaZq{a|}h99X#;?
zhobLz7wy(x$HRq}a(ZYls%Z=@+*t$g@S2WkVv+d|knG|-KrCs`D-GJUZQULfe1tjg
zqLcYLU7t_Yxc(TdUN(uqt=R(uLjszus1Q2TKTEmbxtR-|A`$`1$_IClvCe=l_WZZ?
z+44=lc~U=n-7$l&<4eBz$L+jVSMMKjCBA}AdtU}z{0Tjum$E4gF_qUzA5U;hz-}GI
zo9`gIY#BVoFqBRD+9li9EF)L<z7r5V^yLcEz~)Iws8tW}F;#!Ah!|NCtYi)?ltw0h
zbrs}9U-E)o8Disknz~c}v;>~3?nC$y{gQ8NXoM?u1wJ+aU7i8#877*bJ6#i&@Vg4a
z3H)+1<!qu{tEh8l>!x=|o5QZ+9}iHf?qD-X@Q~K<3Lfgybr00uv{REdLKcF}GN5NJ
z4gOBRC+D^GV}MA^)>Qh~mgA7BtY^{Mza6{3uingAq}dsfIRwxHN%A4h7bRKRxW@c!
zntmSWc}`DO;X-qCPn9V9zfBdPX-(Bls09U@qCqcszKqXI(?e)b{;pe;jy7~2(2wHw
znHVNhW-K!h`UKwm6-eXcD0p7JB>(hiaON3H%I5+}>GRKgFHDq)YwEYwI^1a_qdYGC
zg#^o8PS%*7(D-^=JBH!*8_Bot+r8YO63k&7T}8%91?_Xs>)0IFRQLHNVNg+m&<V<u
zL=sgqvmdSi3bIbE{9%nD6Sgmb<*E7aURqSO>IABXHatZ=M8eNb&g6K8mwA(`O+|ef
zCnl9ioTS-xt2<-4J)h+il~C8r*yG_W_I!51VKuG|M~WuBu@)DK7k^Vy#=hJWF)sng
zWUC-o4zILImZqfj>$HsCJNt^D5kWvvqN;!1?*-2M6bvxG9gaqXf4z?a1GqQ~|8ONM
zB?xo1Fy()^g7yEy=V$!{j?;p})m4>5O7GdnAp@LS8&Y-Ox^J5Hu4+~JofHHcv$Qnj
z-M|nhe(`{&I2FKJ87>!NG^6Y+ZxejZKJVO{?=slDAMTrtm0WM|Cf^UE)tj6cS17a4
z%hH@JJOGVC6Yx?vHtU^jwaV()Zq(z22OgZxcv#;d!6@^&QS-RXOpd$La%!dfHvq31
z-#<9(Zi*r4@nV4EUUTRTFFyUjl<GGct-GJZkNwwxJ{rQiNThNW`cZA?>!z+k2Usm^
z`sRL{x?$>5KN*hmCJdK^4=NeS&8Ptx;n<?96dLQja6-yPAKll+=cS`RvW!-0sVQ?I
zCB~Ep$^j;&#F>-ObUgh@vZVG7?5Vb6H7ZHP{ybKlzH@y=@X-6parJ!phToiR@U@OC
z%bkK?Bs+^q`5EK)-wDbH)eQ+e)o^U+FyE?3-E|QGBKu}}$K3bsr@@^S!R+X6i6#xj
z*zL>kgj1cQBy=HFO+jW!aKAM(Jjj#sB)f(UV>L}hvnbS~AG1^zTX-4EmhYkR%!5Jt
z0NI=?(G?0_<mEB*@Ob-G?coG*<`B(M?=0VDPd3g>!i%!?krUS$KgI2D7Ct8_?x-ca
z?A+8N%H~DVIH_lc0Mj>Pxo;7FFWt(umeG3;axW{7^dfHk0p>K4AvMKN*3D-AuI3^<
zrTjRN3-jpxvYbtVtR>Ij<S~1vV9<O0L14)%=Kj+<``cU2)1=%kvwLiw-c!Q&g92q0
zbYztB5skfS{usHDCU3{#syU5ukqZjyYg}e+%_LNpA>EK#1E<GaI$y>C|7Q{qg(%r1
z@5;o>Y*>Kxf0vVeY8?75rNCWP6|Ywj;UCyL?lK&NxuDtS+kI`pfRV<!TH;2+*ZWCx
z22sPrLW@A#Jw=zEWTf;$-93qkKIpA^zPImOZ6C#9w^VJ<A}A%x3o}cCK1_C2-le|V
zW8~V~PzJu+pFX(#nqW7A?}Cyux8$>u;aG<<=np?JNLw_khXTaC9xfsov@$|9aDSbK
zJ0SN0K~ihQcLIRHLShYOT-~O*lei=k6LSPL*T1&3s5kL$(+j{`KRY`84uW>Y{Vzn&
zzdcv}nU8<;&qT8WDU-Hca)s2DTBgo)tkFo8nH18ZPj_-aQv&{Km)eDcvn_6wE85(d
z8mQ&e`nW21S$z!w4{}JAYtaHRWo4k;-u@Mi<j8+k;W}=irWvwGlqI23swjJtzeXFq
zDbX;lJ7XnyX`YVrzzqCK%~hr4$Tn1{w0@9v@3th!$low4hmlP3oF(R*jEH)9lmPus
z&&kLkHVx;pb2>P;g?O=<kP$5tYXWJ8@p&gsT^X1**N~-V_>`0n%DyE_q5iu~m(+BM
zCl?7rY6w4h-WymfM3s!%&V@Tio6@%RR|Vlx=e;+`y;ItFvu73ROL#tO?|)Ms3G^N_
z1|}m7eF|?!*L`F~OSUm3iQLGn9^uofk5PH*_My9)G}N#LQG1jWMDw#@9(SpPSe{Ml
zU5f%H_L=Za9RpI#bnPs?8rKPtT3n;+#LG)<M&R{6)LL6h*em!oT#J#A`*dQq`>gd4
z&ll|mda&&yoIC({jQ~=EgVa(exvP(C>pb_lC0}<D_fycx<p@j}G_h%Iv?P1T{chJB
zJal!}yunz&Q~AR2BZGaYM*XCRc?07Fx(#Yk(#z0<rQQKKQOV6QId>v%)KvoI0Z4P4
zr^}K4f+#kEpo{q`mBZ+KGO|VNwnai}I5oZUm5K&NKOk2nf=jqo&oLz1iUenOU>{Df
zHk$FAGcSOS)aV^I#fxwr>}GB(;*oRkgnFjpbIl-29jd7AYJk`G241UbDDFuP%!%T_
zaHzHE8qIB*-_l@(FfR|=^?AWoJ^(Z@Qrf(?O}D~d|Ip4}_+a{>JEu4~Bx8<GDk*g=
z$tj?@zePSVq#yvGmoL!+WQM<A`mKVnyd@@YdPHMNQ9enc>rJLYdHoDKK39$2eMFY}
zdH&?{N-M57EqD9X=R@NX54S(XY<1QLwmi<GeJ9N@+&Gl$BxFp)b><+CfxMV++(q76
z*3R1oh$@M_-}<q~N#C}<6C}tFvrNMKoh4Yqp}A)cgehYB<VQ5{S7BkWK$GJEK`$lw
zpi<whmQ6`Sfs*~m9k{(pVa|2e*w6Q6+OmZMWUc9h?`~&<2V~bnt=49Z;lul9vui@x
zrk~HKE2UI1QMH-KPD}z9deJ`wu76IZDjc~5R}(GI^ti6Kjhnx6T&%2*<Xyge<z(rw
zE#1P^ym<u7J&;mg_+Ej+CwxlfC5#9y-y{LdUE@#NT+nYiEwCGP*F^eAGMdgNT&EZB
za8-?EtQK|9_P}o)gVJkyK3E%-%`7yB+e&<V{yK794RFHzDWCk6Xj_!~3+J{H!`Ngh
zkCvAOjC9^Q`Tc|Rl>fxu`1{h6|3%*qgnThy;}`Y*MjPRDojT@x{WTsqZ=N+US{lss
z;C#!Xc_M6GCon%HLYnD8=uIl(r{)Z?`All614q{f?pkP%QwR`0d}?zJRSPbv<8eKf
z!`Z*9St#vpdn<1@IW$d=xsVxH@<EtLbcWamBC~4HB3_hT)Wcq}W!e9?h6ifHZ|}4M
zT1x?I2=sH%(SXKy$8EsnaVRB6FiMQ0VXPpnlk-uJg(od$0!v;sU@)(>^B~ExcrmWw
zbk~EOUe?nucjd6^YBCQZ*_tJN(wu}M_YD`et|oO}bYo(UF(l6I*g9^^s>o--U?S9)
z{A3n<IB!_;^6+g`;bev9#7w{*U!eOTz8BGb7cR`FPL#SixBxfx6vO4CZT+f-0EI!e
zl2y|g^i)*1ca=1<4u0u$+<+bzKqo@;O^FuzMNJPa3jP#_o0(_aXagLpT<mMgkXfrf
zQPqA9Z58xf?P_xV_g1<QwScD+a!8Onj1|2h-mGrfv3YlUZ+oRTu4$g%O<!3)Wi9sD
znz(<mF#@pgQWoXPZf6=&U*F@Jg>D7|-^7|BAx99VbK~A_y6v?57S&#u)$#INSSykU
zuQ@m{FF(X2NVccj-k`^Lk5uQ>wYNg$PP^Hnn&i-tg7w};Z{3%r68_GPas8y#$!+%3
zbPy(><PQKo!H?$^ZJl^>{X;wBrh)OBN`4<+*cuFfRn>Z$O7ro3+I-PuZeDQIt$?gd
zD7|^MirEa8iMQj(L1HCID?pobK0R0(YU-rPSFJIN&6K@eX!OorDjW9xzowFywOD}w
zu}qS-e%`8)rEnwT-P4ha&)i8CW2r<A<mg#NW5*`HfrIrHNAm5QhOx*LjK*c2p`~gg
z-&1FkHD_VqRm?swxC*^JYE2a+Bl*qkTVrLNTjPy(QYl#vuW-=&T6IryhD2@}IKeR0
za+Mg^O8uRt7uo@0wvRO)sfHG`m|?|K5Cg7-Yh{dM!MDe`U$yp`Bz&8VI|w5z%Q&Sk
z7OyZY-=M`F&ezw%N-PBm_MvolCIy4+vO*;kI*$=Ye=&&uA~FEbQLLPBWfYhLBepR6
z@Tfs3FdgUCpvoKo{3}TS)cK#L#{U2E9#>qBOrV|ifUyXu+d!+!^cK`WLeyA4L0+m#
zPfksCk}D(1MMb_S{!^MAg=S)UN^PP^7$4;tmq4v+kCFkKgwl^W4u46&&9T~aVTf@Y
z;Y7MJ8DFpbGKJurqWHi-xZ|ybNuNRRMLtnrDdyE&=)R=&)jI%#^vsYy|1yTZIXqhI
zc`7wHqk@s#)2a_IjMfser0og+z&vp7^;Iv}rSTB2C1~@${HOlMZ(O&M5f<^1YV4*)
zlQdp7e^nd(3*5~~b?wU?DH-A*phoJ8w>o);XT<1Cd*muQ^&OxjPRDHO*GW^)eF5#b
zv}}*nw;B@Krt8Z!AcYnetZoDua5KcdcnLd5%w41LD#*9MmXGus=-zFT6(>~3(VURq
zn)hsDDdh|sA2@|p?R#gm^%{TH1M}{NRr9VN$fKVZQJFF`v?!B!BsCb0&$mJ@E%}dV
z*Ov<ke}gwUwp}89DKCc10E@yK#rc92*9?ebj%igF(@MQ#S;0P^bZ{PRGsKdLJH478
zRWKK0gyX}@Bqwi#92vENj=vI#wBwpcH<;gloW72W0eD<!E+e~G4P_^%rLxmnmBD&n
zNJ*v-oNu)8Md_stCpigSm8xeE(e%y;bgkZ-8U`*5XY1+m<0{2mu0S4}Yl~)norQqw
zl+unOb_3<fSgefkHJoraJDaxV+;Q!IEf=~Zb{GK4+JknQ_A)Mb;FAnv_dhV)49O=}
zE8adeUvUDp<0WuKbKXHZk3~-zhHqDC6COoNn4Wi{C6$C4xFTEDRdWNfRyBh|qE~YT
zH#mkuwXz$g_<@7f{9hBf{(wI67iY&W+W-gHmqsJxFh*r@qA`8ygdpw0>mg`won#dT
z?WyugJ1zI}OP-KtA3B$DVFg@$H6)mi3YgX+aKaTK@@%lCvS**fF%+(<?E|_Hlgt#Z
zrNG&LsBK6zOkBPP?j`$hhPS`o_VIjNL=cz#{UG+AC#mUz0Wx7&VEyE2Y3QHb?|pkp
z%L%#u6QNWl@Atp%?=xfmwQUf=Y#lrP_P{B-&P4u_QH9^<Vz%j4H&A&pj#hDJ+UIJF
z@!2L~#lu$(#)!=6D$YeXtL{3~;96anbWCSGyK}jBV_al#)s5q*%o6@Ln8IkfS@HNF
zq=EIZ(8@Ib=SWrY-VZ{tRHs|>62V6ySfuGt34LY6Re49Gax{yg!iPr*rE>HXdFfRC
zRN(!nNfRJG{^COVTLQ?<y27$y9)t1SOIy|pfD~RY8?-CmmNaUpj5((VGq`LzxP4WB
zABz@&Y@9JPRPFI_({w^~z&u{iD;un*9>`ziB9x(K(CM#FQpW@UwaO7sqFjYcn5Mv(
z-mO+<1Ja5d05r<w&PhGuI=OucBdQn`=peEd107*7L~{^-FuSF|C~VvW$~jNOPOGnC
z9Nyze0(gn?00}^)d3~SEx(-;8x8($lJWi=kv1btL+RHl8rHK*`Q-t^%#NcDUol3AE
zec%HzwTwKdgUN&X*M{X@b;<N);q&jcc;(2PF5xr}U+l#l0o@-Ix1?-ln@7e}n;Ss&
z?4UJ)qD!#5<K6vAw6dIZ0(TlN)-uhpZ4#ot6ZjmZIaoq>m5iznM+C~$H1<0d;GjkG
zQ~EiK=V(ZFmI~bP^N9ppUXu`#EjNS#?{1?`FCWdwy4p4*jSn_@gYxJ6-RAob%&1^)
zAFY;06HB)qHM|!9SiQFbWDI+88}*=QK+fbi$y|M!&|8!IHa##vibEgBSo8?*{f3(G
z=eqDS_PrX3Jr&lgCnNdmN;X6x7s0w7C@xutr^#314tNFZB+{QX-CR&ZF~YKY0m!!K
zT?Mo0HSmWsGw8yjmWP_Y0*I}|EH+nGKuk^g{j;Xg!S#OvX3HqPu=tiyC)+M}Vq%%E
z;#fnDSNE1aPglu^lXUc5&}%2O=!-m#@{?3;2Lq*-V{Sz7C=@aGBF4gOjW<BiE?;+I
z;o`K{4S+Z#Ei|ST5M5Wh&Kb1f&p@^(;PApTQZkcwF|1Y~h%56sfNC<%$lOtlcqy?I
z1Y9`UFq#10t+{^Y-2yH8>2$kxkbpDvykk)h$pUlhI=6MyT`|HTD5?35?foK*53J*c
z9=_{LYm2QORAPrx7Q+y0>*%CyqI-Yzp~-suE)2DPcRWQcg5$WL9sFKZ_Qj87{ZIS<
zOfDd+z#Tyzt>(r8rE$C$W4hz;Ji7j^;~)+Qr~eCByea>^J(x3>Chxd`g0oko8zLr+
zs-gg|)rpi6KLvG{2Nl^iFY_`)fQe<!X$8fq1JCQ!?8+X$9PXadQ82%yK*?~Di%QoK
zmQr?;ZT1KD-5cPz{)%{O_L^tT??cF46L}0AciW~DnO6k?XS;q?_H0XL$|9258XB*>
zm(97vWb7ISug(mlYZ52Ff0LOES~6s?mXvjw^V27bpIAO&y<0ybnjtX~))1`k4pfLS
zuvh+!YoM^N*@$x<Lk`FIXwGrHZHttk(RcseuFF%Qa%H=Z!G^KMB3Gl25>l_;aGWan
zPT=YhG?7w}pfC`ZtyIcVtRuie)pNH)UvN$(w7!sMIKJ2lex^cHWWq%xL9DxSi=6Ix
z9Rwu_$#yjd{FyX9$F=S<m7-%`N+ZYXJnZz?0Aq(ba$ITJM=6)bfeVC=F8l>3xzj^H
zqjVAgIP`&h7n9@oEJ=S5Ew{fVu_5M;oNOfp7rm;yzd%?eIp#^szZ{k5BnD*lW=yxb
zMi7oa+y@V1f0X&wI3jrX+OX?%l@ko<Hg{gRcSOdcK6u)m1WKyxY$=AyH2L9Wlabvb
zd`I_>LVXI6A8mNji(_XKS^Z7!S}agK^J;;Y>k@|BMbnORsT_AH$R8MleA96%s#G%2
zhsMak2XNlaanPXsb|(^sQ~LW~RV!KiK9g8N=VzK!LK3-fDmnREuXB`a$L5SH={-Ui
zlKi~70Gzk5vy6CNHt4CAo4o$m`CIHsll%4MWMQ1M*B5Pn4fj#Q>49BbOwI{()+>=v
zGC8n>^B5RhlDPtmi2wA{i1^gXa`=xC@#TF(QQv~AoXrP*)QcMb{CxUvynnIdcB9|a
zo11IhR<!H1UOKJ1FkkJ>4E?;!chX34+-3$YOl_xKiR{CK&uJuuw->5R;vk>VM+aV?
zi33xh%(%T5r$vt269Z<10n4`zqY1^Z9rH$d$_DY4uc=n~+`EiZ-#&n!Y*7wW$Fb)#
zfj}m{KdIS&HCX#I5ce-Y_s;~QMq{?rcgKSSzaXh&u+alC=(8(C)@)-kZnpN`wS!bg
zs1gQODJ~!HH`6qctw}o(w9?o-@9egY>f;$5f^}P=QDJtOtnp@3tt-nm8>YoC#4+po
z;N<MF6)ak}a$uZ$lX&^us@mrkw@VQ2@WOU*>F}eP?3Ew<obVP^og;+rPQ(tr<768B
z`HuYgXQ<rEtyUPZaTvfJ_-W?l@0To@a@s=>$^8|aQzSeTXJ#RQzm73bvd3$dMjtAL
z3MqLKr}gvR_I}a{Fk}qR;?;1JXrynIk96+n!;#~+iXTv%O(;av5k~uHIbeV0`ly0=
z4xm-CzS$(txd2_IgS1iM<Az{xY>YXOdZvOM7$csvOgo;8!5Qsv#nw(d@m<0-`NW;}
zNa_a0P%pt9HtYndj`5o~G|&g;4!QD~Jka=%+dXKX5%Ha%5dJN5#Pi}>mbA;D$%fap
zD>zhhB52r?DO<`PY048dKt~Z}XIS0eV&ZA)yWpG@IRFe6$^qYY#KbD;t}}u|PMmQb
z7@&wuUNThOcR-cNEnvhGz%VL23Po@G;f?Zef@(moPj-m4uqv>et9f*(ZURSp?5bnc
zVVj0S^TV7%bG;cSZ4U#t4H0&P0TnfM2fNFgKV}Nyd-K1T@2?+!2y;4I{Z2q|%n+2E
zygu$7D1_T~?5aTW#G5G4nrIy1BUInhhD61Ii-cJm0o?tAv-SH226&!a#)YT)9Bnva
z3|+b@DL+2vJ$t_}E6@B9NK}?XxM)l|OWH0JL84Vat%bZfJKSacd}%ZUd~muBrPvAE
zDArJ~M5$_FPj3f3zWg-#fS&6<Esm9u`%NW_vfUI72V2qeAmv&P*zGs}q439%w-#U2
zLLeTLE0S(YI{i+AHjJz}UM@XsRj2j+;%Al)rsp39(u}#kz=gY2kb2#B(&QjxIcj<e
zd^7gL!*t9^&D~S7eQRtZ`N_S#6~~&UZGp@$-l|(1*uZFlRuMF0GwPzMpWTq=@rw(#
z8CF-=3(_J+U;QN$S3OMYbIcPD=E~55_a?2RLa$~|**%{He<24H<Tf_E%s0!Q-Kd5T
z<NRk@Xe4R;3QZBRUq|S<Bw8V)j$yQIR71C$!wXHL2z&PP*NKR{cn2h4TG3R-R2l|v
zJ_jX9Sg%NND^u`Ehou9C?~T9BgoJ-#Mmm5Qfs{CPzaXy06N<SS=Y0w2F11?#hx4bh
z>Hno$5Xo?!i{1Nhdnu2hA$6rpM?#jYRROIs`5sC$kX&`N>g_0_C@I-zRysttEUmv8
zte%2a`<op+ip3U*tU`1++z8J<8-Uuq`}2;k_yILYU32(mzvD#kaR!I2&Bv)%%=4?7
zYZAe(1<9^v(#*;tpUdxaNbH~MsbloCDZ;c?J7v;3m%Lb-VRK%jR{^b-w&UdI-~qE!
za=X}0lEFz4jd<@|`o?3L)r=vY{H;%A(T<A2Zyu-?E$#@|X{#@(9*#~Slb`<$#>-zE
zS-M$7^RqD3RRexZB=#vVJZQypxkQOcLv>eQhagCI_{;I|#y-@Uh(54uofG#6aa~{4
z&x!(bZ6TL7b@$^u<wp6uimxao`J}c;dP?qQ*pqU-0ETM0he-Mc%>|=?tOe+1>2RKC
z*{<&7gGlMyQLd9Fccf%JQGn)!ai^`BFj94Ar1$FD2eyuo3vc8OYz~ggWYDtW(a9g%
z0FW&{u~g+*vJYmWZ5jw&_H)rs{7%3$517H19EO~d&X|p#Dk&=46>V%&wV(Y?P!zcv
zcicAHyJ<N)6CAEPAvFI0MGuD&MelFdzwUuM;gxZK$lceahxDyb$L;bx$%<`Nl}L{7
z1STjl3gr1fW}E)@0Q=AJ*W}%g@DqOaaMtvL91=NO%f|7e{G{b2x4U}1l+j&dT~DhZ
z>oVBcJEp{W)GBi7@4RWa_8_xuo4M$1|Ew~aYJia|lWuYboANlE@0jt#{Wy2dux<My
z-k|L5N>w~8OTgAz-0ZB$Q4s!|<i8oh_x_n7d>Jk6R?APVjeRTjhHcrO&EWqqw*Hs&
z{@g@^!BW~dmlo`i-pdkfj<a6frJISqUR;A4PP$48TQ^3~V&hZz%KLO1wlaFtwLA4P
z7|OT#!oj(A*tg<2VbSWN<pV|8-KdIA&ALn#W49Zr<|ISJI@IsPuk*k2%5i@n8ww)3
zpA&L{{sPa>?a@C6F#l?m{+AtY2I}L7o>Cam;ON$6*=m>auq#-_-MysxxP3#vI>up7
z!Pzb%`se%8)3+Cj2&@UqYn_ID=Do?oF1gxltMAgxC97!)Z`rjF*lOCOrpmjfM=nI7
zMrvYa4coY8-d6cxld{J?EJFq4g<!tp?i`1p%_zG_pPB~)9`PDI@zDjEvTe08`lT&-
zE~-)C_FVBpLfSkZN(?2vJ8K_l`AM;)iY;C&6J$@N>dgm@n*FT>hz>=9VVgzz)^V;z
zjKiw*?5o3~(U9=rYC43Hi;(6%Kd7)2q(g2OF_vj&t`^VP7Lif~5j+(>Z8c!S*$ScJ
z#LLQ47ouEIhi|2>1bsvQ@Rp3@NVPDPd4ss~<%A16{y^K<#OZ3t9gQlQ9dE!jozxf3
zsjK8<ehpxZWOgMI+z_$FexXBcX_E$Cl`Ii-ojG1VFkJA|ln5wKB*Q1I^uB^q;@vR;
zjn;A!p_ES}+uQeeGQJI?1{0$W>l=_0w2evRDQ{4!rhNy_Uo1>)-<22B;N*d!!H4K=
z-|`67%;$$Utl+xsQ1&JiXsM`uryWTvWY5uVI_qQT758#UspY9AYouPUW!u5_-q6m=
zF@HYpq+$<e^UV@LdAsXf6v5n&1quRgo(iqu8*n_wMy};^*3Ba2yJh7mj;^c{X{x7P
z<SkxvJGt1v9q*Q0t&flE6k#Kb!tp%;>VV}E9-wx-^fwLf_-ePDiSGx3F<gEcP{@G!
z*)zL%d-tWpgQWJ7faWV<+21-)Km9MK#hp$~_1=x-p@L_}5o8BTNNKbJG?0mOJ6=-?
z6q0>kH|2etxT~kMEvu-hl1`Xfd1RUora&B#z?G<_B^3+rPiX88?H>Yu3;Kp7`BpV)
zYlgnuqWI&P1e2YxRIYJ1`Y2FGM!9oRE9po&!|kdqCWxQ;fsPOM?dnQ_ou`7RqmDP8
zITV<bz6Q@%FFZ+8eiO6|D27;Uzbym0h|5gh33~Yz*A65IO9K1)^(Jr;-g6p^s;WuX
zrjoWu8v%##bK?)zI0O-;9Ub}l3V09isVv<m74Twyyw%8ceXZIQiHot6xmkECYGESR
zqjs)$mT5KxZI}!O_dOl%+iM-zM0s`AS{S1o+fGgd?qe2$`#4fdNVAJX$ws9lJXxgL
ziB0~J6u$;@RkX5>93&J_&(hc-+Jnn@9fp=7e&$X8PI0>chNcW=Y=F~N0E~H9?7ha@
zSa2!tBhC#`aWRcJQ4Q-5xpVH&*8r5ij8{v#9y++D#`C#j=_7KRc-aQ>Xrqo8|ELx-
z{2sG*<CyoB*xpm4hsej9TI0eI;W};7gal{py(}$KsxJQ)ME-dEFCh6hfU=pD4yS%7
z1|Q)qzO-EKVR_|YwNAA=(#;sO;bCInA!?`p-tH=mqA|z)V>(li{da=L(JctMg^k8+
zR4z3KS8CF;KbJwPPSm;*HZ%Iy@ePA->rCm)NHt%td!FRnKtLD%dh;8qURC>PMh+K$
zwtdt-&aAAcJP2)Dq8xOttDH8hfV1VLHzoX5&n!J*cWJr6w9;<tg6Wf^xTHB~SW??G
z<MHD@AxeHDlld!cL#mrhJsm^bn+~Nww*9L+=HE!}egF{aqeCERCSTpvS}S`pqtkvz
z&4eT@S&GCn0F#M5S3JaAJ?weG4l8=IVjx3I@31M-&<yCXj48W)Fe#PFxRM^DQDc~M
z;&PW?i-xLwyu@~S2>%@8^4e+U%o?95(z1AUj|)U{`x-)CKhIV8W7;}p6=>fj&TFX5
zg+D&;U8VYw6LJ(Fa^KxqZq66oFt|N<bBKvxBjUluhkH|ibuJniYFG%)Og;f-ZgBTx
zj}aFRripzZJw#mazGnizY5FZr%d^3QqgEIPn7j9GgUM;bp-rM{0kw>6NBdl!pp#nR
zFygiwvGZEWMW*j^JyWkgsNZ_M2c0lDhfA6RTuwY)66uF;*FvA1D$lY$@vez&N4Dh~
z%hvCl>_@vtrIop;CgjuQNE03n&s=sL)ZO9D5+)&3s~EKLD7gxC-GAW((w-qO{g&by
zat`pAU)_PksKiWi-*9`|Sodgp-<xrL=p=a?@gCajeGy{2NH$@aC|U9?`26W$Ijt(?
zZ~!xEbebCD96HY<B5^{l8t|m_ov8sUb=zP*{LIG%YJ`G%B(+ldfTC7?(zi9Q5pu?W
z*4%;U+N!uAK&gUR`CF;`KMh`*<K8r`tM?b=zbFWe8VW3Bok*znvooZwW7fw*28@nL
z>SLYk4o&UeOAjWwqi{y^3Z8azA8a$rU~VxB_EU6-G!IQljQs-VSCr)X9AOjWIsi{&
z%H#Y7WvpqVoTe(2-GO+HG+0%L;~31y&b=k-bstEB9{hG`0{t?9Qfv5^yLz~=xzDNp
zl0N@gYvMmtOZm<ADOIIGyVDO@rJ2>v=H=bJIaKoGBAY*+HWeo)7OD6dG4yD0GHW4H
zV;g*PMU&xHux#8ZbNxFL`3m>_Bh-Vg%7;>yT09TXq+?}=PjIOW$cxErSiP`i2|-WH
zJNSxN4}lj^{H_><)4D3qZ%)X2lFeO|^+anS{z}{{fl~Ub67LN7B2@ZNQg+rXxuEbz
zDC;MaI&F@ACQrPMQKsAvlG93o8lwlD1}HPjB4RuOgU6hv6vJz14w&xt7;rCJ<~RR)
z0wrK(#z!8jo=w|L<EnBye}wFLnr{;GnHx-*uEAf2YshX}mVe{M3eY{6kwFO`EP^%{
zwMQgab2nc)(Rmbv4b$Dt0}-3J!Q@W8R9ca=cu@>#Q`d4SN@y%^OxGg(ou6%To|Nd*
zngQO3rOqWq%b`UBk=ykqHTttW=h}2Uqur-Jl58laD++`*fdfiCRWapQq9LJJwqZ{a
zz6tJ1v%+_CsSMpJ!Y60KU0=WM&JXBTOBWr+^8sTU8|Ljfakg+ErnOLepRc#+F>myz
z7i{G(LuNQq!n!*)5g(9c_%j&YCPs2@<I28L_5$h{m+%_5;0;;${q<?OA-td~wkrMo
zh(mUR_7mh@c8w@{)3{NzXlj+l7ps1YSTge_oT=72;pGkek8E$hls>X8EAr;sWx9}n
zzJrV?uVs|#9T(@Cn|V`TJ}*NjZXWh*`Xb1QU0!<S3F$~(^G*_?gpW=ueIz)m$&HrS
zPm{Ku6Hh2Qx6#I&#LJtQ04VLg0f;nnYEF<C%nx7F!LHe=O5?MT@lrQa{!9CT0_^9A
znhgLGGai;2mqRX!$O>7#y<YEhD+T(3BID9Gdb*G6pL{_@<8^SN4)aPn1E;OCxwb??
zFI+XYnpO?M=W4G)uCA1gkhKrIeSdbtUZKU9l8rU4FB4_~k?rG_gxEYA+>98%ZLa~l
zAH=1r```~bR>QxvW4r&PbH-o{o5d~h0LLGC>(ie#asSlJ{Xca(#-#l;m!#9(NAEN<
zxkN1DOr-?UeI+a3awtmF@o`FeWuxPKEojh6heqEN{f*^>#G&{ec9#MW^&$-tw1#wS
z93Js9JF3e5#B%$w=guS{Ekq2M2U!d2{mjbWKl7*ORAN&o?sHtp+nCe!(2@f|%WL&=
zsxlJe$`b6%w)*qhpM0hx3m?gta(^wOfboyaUxVs+B}_W#>z*8F9=5zd1>qq$N|!k#
zwNkeJv;x*b?}JEXcm1{letkTLqf-qUQMhRO!$dyoKUU8EIRsxlhmQGD54e^WKRh`3
zf9$<ySX14)E{qkWh@f;KDk=g3(u<I&ymTRg^cImOM2OORY=G1V2nY%h0qI1#)Ci$h
zsiBw9dqNE)#4~-@*{kfm-nH9x&UdbFuRmVNWIAWYm}5NiDfe^VVse#{S~wpbf<|$F
z#8V^>88IWi3L&8B4s!vX1x4E1*Ktu2m9yfNFJS5Jo&s4)tSd0<m}mA5(K0+z^O@A5
z0skt06o@-+y0Eg-%o4vcM%An=O===&`@KmyJtOZu%Vh_CZ44!D5zwKv9Ltj<;0pux
zg^6GJ42ufrA2A0eQXd|^_va+NWZv_XSgK6QHo=G*KRmu?2dg3D>j6zMtSt+-zhIOf
z`JuDhZ))Ew%Pg>hlvmfV$Tn}>bpj94up#4gyZY&o?w-{{IWQ;tI|7ky_w9gK?!ga$
z-?DpcBaQsR*}a>`9BDUMe5!F3vJ9o@Y7%M-dS${{&YI+}C;BiSrcH(BPrA}P$Qd`F
zOpzC|dRj8yJ$niMT-6;6xu66iq7BSDSq$SI@S07y+P*>A2_CKWQzFhVht)jEvFd`g
z?SyapVH!5-NGHsUg0H%X9U`aDm=fc5G(Fa3l8M}6?no1Qv#6N%HsxV^vk1~pzJ9(o
zxu~|R|I<0sF`UooS<Pc1unyyC?s&bZHet#QXF*PY(zh5q!xEkrhLYj70OV~j7yp>j
z_ivbL{sBepADxG*CY4267pSH~19)H>l7;c0gu!x%o|O;*@!Sm#2J7hG!gH{-&Ua|@
zExPsSx|tfhi`g@5=2h;t7~JM%VRPKD%No5pP`s&!O@?MJ?H*Wul#MUEar15}>-nd2
z*`ibC=HAo#T1}~UY+}D>3$OJ9T`7Oozxx}r^B0PlzhawlBS3=NPz_OGz$~2oKQ{}1
zG&}!4JFbM>jUpjKl|-jU-9C+$wVLnYwG>cX->L9VWC4OGqM??)yaMl>#Q4cG?dwlE
zllNRs>jin!wXYk2OcD2Tfuyj)d|6M)UUXX>{OIg19Rppds!!%iu+n3GS<=;c)UguD
zCUOw^#q|yHe2v)ivk^pJBYS{C(w(t!Q}X~IrYMkl$49spJ4PrFwJQxMNcZcP8HGCY
zwXH>~Ff5cZGidu)LzDuO8-3jiSyN(qK5o7;=Qhxf!##{()wKL#{qHn1(KC`jQ|Q-$
zLMH8q0>|i*sGDMG_pd^q2EeB=ksx3|q5h20do<qh{Kc`NoGoda!-9bW$Q_&z(=nmt
zQ!SXUt@62&Ewgc_w^a7+a`rv<0a*qS+>GITzStu5UTxocNmYD#)k|%kchfPdb?h}$
z*ie1}a)5t=8AqmaY<5?JD|=03U@BGr@Iy5x$!hRUmgAD6FK)K_#nYC6r}8qy<|fDx
zzHyS-<yu9bUg<cWYpb-1<PlnLR2*d<+T^1fz>@n&|Mlc>JYdoQGTZ4wA2pGqhv_of
zIm+z-VK2!hNBvCV%r}m<l_1NztZz`1@hzAM5aF0X!*3XEXkvqRPjWN!8WVOJOUm4&
zA2bl#GT*&@h^nu|90N*sP~3;v${(eUYkazQ7hGv-)lXJ%ANeJBiiZ6Zz#p2aJ@{7y
zh7xC{yYrR+KCljA^B98y#26-dT)tiE^LKVWMElg|uR<gtRSqv;M&5pniok7q7c}}W
zVuOP31mTdvR6*&MEvBTj%!u@vg~~vb%^LCB?{Z<7x&D42RZxGc{=X72{KNkv@`2PH
z1wfCEEQ$K~x-Pt#;WIffXajOv%%e+JgWJ4bD7D6&R^5>&RD%G(T7;6L4~fUm?}J0k
zz*zX3A-+N}->{v@<K&U{W*)7H;g85bsR11qb%*0pN`(#C0*CX*VokNcy}kGH3zDt@
zLwE$li+_R&|DDD7&rXJa_^AGOAecL0(&)!+3h>U38HNdcA*?^2^<Wk@=Mdyg0}@-t
zu5DNJ^CX$Am(5rvVkTFLYJ}zAZ6|gM2j~T<iU_G)3huqdqqQLk<QD(Lg5nKrLhyTI
zej^zMWdv2n@8(|g)`}84rOZ<5S880@w}wf&_ATh@<M)aj+#{-X`eIJ4L(e?zfTI4m
zlcG$1<61>&W%LnmlucMq9{QYjQgglUJ^WtZ4ZTmd3_fxwKhHw}-6?GD-JoHqk;+C7
z&_TOB<R;Jr2)0cM_YJB{cK-LYGyb|}_3JYrcAhIj*M>XJ`c_r+{Bt5Y0#pcef7S{i
zs>;iFIC7(|6(3<!+d-O<;e-vKc8Yv`%`dl|qu<?R6tNqZ6AMy4vH{EW;A^Zlekv{1
z5L=Hssc>Er{T;t(3gWf*QO<6&8jPVxE-miC!ZwwdT(m<PC-f|e*0BbGS8V!OcepUx
zWMj%*@9t?n0ys8XQ#pp{PKg&iz1N%{&_@c{6(M>xWcrQY0I@h_UQI8zY!_;#d-a_v
z0_>hJSTQsTe7^J6vZDF;5-BEd*q$jm=b-Rd*dbnwU=bU-6kwKLkwQ$yo<YrJljI|+
zxZ}G=)a?d9_SDOVl}Z9Lk3SWW1gQ^3vM(_eO{oa*zqmacD6?H!+AkMR>PARHnZmgF
zWsjfr!xL^R*UfsdFJT!0wih0%U~QC+h(QwkBGS(epeCgMqN_m_nU%saKOU&D_~MV+
zw<~X5ul2GD$$do-s{bV>vw|cGu6VF-iPu9$)9S5cqi#Xj1gQce*)vyu(xn}0r^Ic?
zf5<k2%Suc*?nzjUXh;@+AAL)^j6Kug`qA^jgJBy-c}U%Xa^mhce?vgrJgaRk>=&K%
zFZBPzf1NM-uYP9_ol#cN@1-jExTTK<<m{h)hvHrv{7Sp*Z`20{?ceOCP+D*PCekmo
z4$u7IF)Gfw^mNjsqu9KyVrqfEsd!%BMQHh!hf?g6czCr=ffdYr+40J*)@rA8V#gEB
zwnM2#KTB1F<B6oWmsdkg;xFCimJvNqnC6vx>aq$9<HsL%2>%l`_vbE@|8s2}!Zn8x
zig6#y%1_sF8^gu3a?jtVPw?iPIukO5K4-Ne!I5GLl3g6X%ty<VS>p5!Dh<8!OzCNY
zJfLUg)agG%qI?I-3|3oaPB?n8SvWR6kCJaa#CZvhJO@W+<RMQD!5mZg3d{Nnz<quf
z9%^25i#7ipynQ4YnbzpZczswftH5))$z1jl9%Z(=6!TlGknw%oL}fPNDpWm3>fJec
zUlF(aKN#kwX1&ZBV`L-4D&u69@1{G6G%P%OHulTz{UIOh?{@BrYP@V@ztWSNdck)N
zdV6m3E07-S8WJImB0fBD)-TJ)`4Tv=7nzrDU%=xgIUzP1JpK|%nx8@_Zd&wnRpl1Q
znTpQ9&-#)#-n1-pb3M#uny=jec(VJhIx$%FqBq-N73)}mx+dG9C$Jh9Q!iK!gbe`c
zmPUV0xfZ4>4{#+S&6cark<1%GEC>O#>VQM}Dc8$lS^={2SWTKxU>o=(Ih6?`leue1
zaTUw!KiORiC!oS%g_y`d-cIGlkDyWC`&ZHjKNSn_PHw;|?fk7{9f$|Dmw&`kq)LDF
z+p~wYJ$ujX)@(S=KS%W1{Yh8cZ^Q>_Hu~H<6GZEI?H@v;6Io$l_QtD_f|4wjc;0bF
z%eRcoK;q$~nB_q@_XzKsq|*jdX%VdEwXEb<1eRl#SDV6fEKDjtzThQU$v&IzqUZJ<
zcCUGd)|vT9cWf?o_%(rH+nHe{F;H-|TL5$3xPI~!a_W;YN1!fx=EFHRj-%l#dWllY
z`;fNUYNBx?ioyR*sphuwqlpdguEuBI{EVmf%bDkWZX`A8`(rywQIj->5o@;#vJ!#o
zX1jjC%j{9u{;OYfD}Rq+>_3Ay_(jdOs;>3WC|18Dcu6!pnau~hY8#Q%Oe%k|4M;aG
zi8}bB<xFPz3=Z%pGH!Yy91(bRbj;kMg;>0<{V}<aMTKl!Gh9zJQ@v2cZEJYs<Ic6O
z5j7e8(dMRUymf($g+$$VWk%Y~B!%Us^2N`knyRL3fEm93k7^M-wf_QTt3b2-q-*Xu
zFxIgGi;zQfAg_TXJ1Y$g{O{|6{|`KVeVON;ar?(7vBeDYm>c#*Z$mlfRwqh_``zv)
zu)d2{5!xxpF`WC|RGg2x9ER_g<_vLv>%a6nB=^n|%52A4zTY^stIr%L9sNmOtcm0h
z^u*sftuu2{jqieg{<wKz9xMdH<PNA>#k<~+PuYZFpQ>N+4G<|#e-bCZ5+x#cYBFgg
zYJKoB#TMkmG2t5UDY9M}=LwFJ@!x^GpaMczjLu@rqnlpg-*vmi1w>$X8uAK1SGhdB
z*|{9x<k;D^XT)p72dH{9{HN<zsy>b<(;Tnnk4cQ1wa8XWiOpIs-dWM_xm)RV7wk%|
zp|Fx61P)5gIYfwz5ltSy2ig0M39S_l><pA6>-os{OSg(`Z`;gGnitGP`^omiKzI)A
zzvwV|-K>^UD{7d53((!+k2?_4ZZ`e$z{vwt`lV9aV@Z@oNn0Q#UgKr=QFU!^r=l7>
z?|WN!KZXCG@lyRsnELiC+a8b52-Xx)F_YFax1Og@8Bc(K7W7B3W8~e__C5K{!8`n>
zI-icpA$E&XL}ozYpdwThjeT4p0qSg{t5>x)$Y3VJKf8w+lcZg5A;+KxmWSnb1rfM3
zuvo2mWcyq<UxGmZl$fx4K->tOkHR$$=6@P<KDPF0BT8+1gr0qS{S`q+ypcm<d`XT7
zyFGKUVG7-FEE*QJnqiWA^wiX=Mt9l?GMa!2-h;}8$b3k?S#e@eupT>Di;yScPlwi6
zlm-6EoT2tH4ckvYs+ggx-dMC_AMEOek&)9V>Un7#0^u-Yc`3uLJ^%x~kTUw=877a)
zNwTSjoK20~X7=r(8Cy1;n%&0_;R!yNGxmwMe$v4|?;{i8pqjwU;a>(#oa}}#e~=GE
zZ1Yf!|E|l%e%xzt%c9jKRshN`r^e;Ny_;#-d}Xkx`P{yK(Ob-70K85~fVkD<2{~Wj
zvW~dIn&uL}CHN%AhXHa%guAmbJ$~-0TW{cHf1y6`#eKGJO8B=ohnqVhi}0);+qWQR
z9N!L^)PJqx`cCJuCxfhE*J6bxjfAZa0_LdXA97{?GiUaH$6ViY$m#?d2HO2KNq%I(
zdatC9M!l9gJk&Q~P%v%aJC%;yn(eP<UZtboIq0ovTVc4*<x!`6gnd;m{_Ui>CgK@S
zJJTz;0Mkq1kdSKeFWzPYYuoJxrIDj<$GKWRF`J154c1DM5h^6R@{o0xe#`c~_n<gs
zDa!`e8(9S*u3uX&KL?E1{Ue45=|#l(5DU--L#e8aAg8<No`UuoKV0?-^U_dQ_fn(R
zZzs6~E*B$80OIH#+S*S#t!+m_<-I0Ax21+%dEnh(gpn%MiD5%WS?>){KDhl40+)*n
z>W0w_in*!Fd6F#C+Nw8|(g+T#?Ng|*vn_~{{*o$0l^flN!NHUV24FG`#MO!tdE0$H
z$x*NHXe_u!7NMM9DPVfd?rSW;9Aim=@K_lh4~6Mpb%V;Z#wI)YCy%NkGsZ!o2VYPD
zpU0GI4$}u?fTH2s>_aDKmWHIv;B}(_*#o$O@mm{L`Z-vB3Yg40wee<f^ICqo$~@MT
zaJ$F>FS9A$h#?EE=cebCPn@m-unrd@!B+MFMQzlH!0~q^nqgvwIuGDLZNWJKHc+V<
z2~!Rc=8fLHm^JK}8<G);VBzpq!Z~6_ztm(FaNZc@qj&p0U)xB2M$z?VB-m9V9hP(9
z?%r^}@0TytoP^5kKO-QH!fDLae$Ecs--;5?a4E(>BuobJ4@^6X?}(@fZ+mwEQZi}M
zQ~n?1QRVftb!Z-#bGeL?r^c(;I5l|{>f`BmxUiNEpzpY1I8AMo`Oc^>=wjNSA;?m7
zR3Kj~&*-u<@(FgciI{j2qIfVljaXZmhO+M)1zvqo!Z=3fa$#N&@lertCb9g;{J!&o
z7I=^`&x=w{DF|@kRH`S4a@VRalura`s@g6Fiu6=#x)eJmMb?Zxkz|mk8RkbAFFbgi
zU2B>Pky<o(7e@e8sZ}-qFcADZNv<P*KE~cct|A16PmQ+2&0D=J?u~j1QEKPE=<69j
zkrVe`X?Wmmj`K1mn#=UJWIA0}4p@0Z79mOLlhPABE0e1>R4N1+H)je?V|1gmovj)q
z*;5AXs9Xk$;2@h%^(2$g^i$0@&e~ineij=O<IC;==Un_8)(^BM{u{RPzbPa9s>l7a
zHT6GP>dqTiQ4J$<+$n<TnHd;1>7F40ih<9Jjbesd05C4rDYyzA{&+o2=<+&yw-e8W
z_}OaM|4!30_-d=UwJ|@j(^JQXiG1z&^{VCDny8+fk!#3BA9Z||5y2)i)*QM^4Mhch
z#HSktR(u*F$3o$vciB@;UW~Ig?UdvAyeb1<eh5h4Ri`KtN35Xmt0jw`3Lcj%W^z&p
zl?VCOHqPlLjc}Ln6D1-+R?it0q#>aroNts=JkmxjFe9n*`cO9CnWek_*+R-H0Z~Q&
z!czDTtlQsoIQ;p%VFArFiFO{K))8~v4)OwEjpi2xf>te}VB?WzPj^MqmwQq1n?nzU
zc7|?>%%H+Sg`mh<{!wVwQkcx)HK|DAOc!Lg!xq03L>K*$hZ2E~D1MUiBU0BFzWRH(
zDlX8_I7_y$oSkVGTLFB#P8FT1UWc17vrapR_xjniKsEH2<-HM^dH@Zt{4vYxWboZ;
z+DWV>_@X~E(YT@AINYqqxuX*x6Mm}3fuJnc>XS17!sMOG^5hMrbJ;$FLxlPjQ6X^W
zcg#o;_H0pJov=ycYC2^Fgr)Y7hCYU{hxqr`%*|oPo<f+AE|?46l8>zicfC<M_s_M)
zvVP^SdfwU?@PLTl?I$BiN7s27!xMTuxFNxFp7oF^hcN0Isl2|M-Tzj93J%ue^2CMn
zFnJ1P5fL^gF%+e4V>v`)YAMe0^F4j1CgWuOiZgo3T1BpDYVyi>FQCqsi!M`i$+ymL
zoL?$0J9rJ7EPbm%KW9{FJd^M$@CLd3yPRI<=aq!`tPt?+cl$a!Ry*4W+$8zB!<&-q
zK?sNA{pB+6?-aH-v;4EMu()|EUVCrzc6u2j<+6mCOhNz{^ycf%ja)&_y|nG40KfKe
z+6au<A9m&bm=kO$LH{iq?^~1DrMD7R!!~(K2OmX5^LMHjal-nB3REdr(v0Mo3ry^u
zCHI|uOzUcQT3Tq@8{-V=_ILVg;CFHJe3$-+AMx`a?6m^=8BP>JiGF%JrD-@L<&l97
zD*lF-jPD+yJS5NGkpI=yk;D9`LF+v|lq$q*$wO4E&#chTyhWcdp_?^xaZulXDXT`U
z#LEFheAR4hB~Kjb_B+<>!8+8(Uf)jpQLvutc;HbM>%f=)Q3!$Hs&9HlXPA`r<#X)A
zy<g-l|A77O8~B|wLS<k7$n|u4LrPaJZR?1h%v=f5cYQptSe)E9uqqP!7thj75?*Nw
zLF$#j@|I{a%$beIQ{JAAAMTS*B!3%?vZcqCp2H31)AKxa-oI97@Y-_j3QG5L_T$Qd
zgg5RjGR%t8E_n-<Jsrth?wyvHly?@t8Uk-U(D}vg@c)d8?B6B6{G$i^|Jw2NjL*-N
z9yEfA?a>(T%}Gn`gs5?JKnM4-FOXYbecF5`(Kh;}z)SWf6CzzM`vQ#59eH-e>g2Z&
zT%V8It3(?s?!F7pA0PW+opS3<L$qQ`NSaVfwnz)sfsekSQas2Gd@;Et8taA-6y=vn
z30T2D{+HQ00^|(E@b~Kr5mfLQtFfuQ2j(S4U$cc>NVdNi54P?S3yt+N#~z53d@BxH
z(uh8?w_##Da{TSw^euefH;h)wHOXExsnd7zBf^AC&7^{lgM77Cr2YfRCk-aZd2i5N
zQt}6LIL{N#!mrI;X_ss5$Y{V8{-pciu{d8rjGOMoJk>JSxcwUK){4o`!gz^xOI`S4
zF{IcfQns=_G}SQmlw3*~f^}!_3zE+5B1n@s^H5Ix-YehARRB>4&oC<iGNHOn;}`pX
zk>ZZgRM=rjW$8=WIre^uu?BA;;b2fS21E{}2$C5IY%N=kg!-kT-vG%{F_s_3t3Jk|
zg`U9<T9Xr~ePkV2n;qwBwaWt&q&dtR2T_p(mXl*wM%1#O*|&LFn};1pO=H&f38*tA
zvJ+QJP^KjplxES=EY`iT1mr3`0mHmeNi6O6J^a)z?ffQ@K2tDhzsicJTQ%)7A^Je8
zve@p*xhwbR`N^0cqPS{9E!rHkUVHN~5i+{j>VcZ0sW0u}CP#oITZ4uT1)L*Y6+g2&
z^6I#2K#=cT@=jSQSInUIx`0w}WRVC-Qh|s2Skt@$XPQtQ`vh5PCNolr@rLu46=UAB
zv+7cuplr;Gj%tZak#4bkgb<vZHi08uiHpC?xuw1`$4*T6Ww3)&;dAER#p>q$T|Aqd
zLk7NU$C3`<-zQ5JaF{m|*?k@ZVJ?w;Bm45;S6*h+T*+yR8R06-9e^fOI$)0Myj8eO
z_U}6#)_bmabsLFKU3NF%NEpN~@{bTif?R?HnW;H$(fF`^Irr6WqrD-(*!>6POiXJp
z7gyE(*brU)vMH8)N>kz^&ppQYy|<Q9`)uhMXLudEM7n4fshmZmvQo#a8%@Il*Z>sa
zScV(uM702+Q@p_oZx*q^9IxT=c72QGk+R4bz=Bp1Fl$`DZ6^xSe5KO&u&9hH`07YS
zZi$}x>K`KXfM5JKGHU-s?ERTm=3o2r|LFUX#H`1K<92;}uD=A5J}5kye3GM{x(YAN
znlF!<W@t%$QBY|Ts`7g=V~B7)KT_q6w=2<q1;|hqwg^8}V@bv6vkU0`4(hG{<ul$1
za5$6UAd_Fe5o|ZK^3m2f-Wxh43h3I|J<HKnNa>YjhI9BF-}DM#(b~xoQMfDVSo*c^
zG0+<Ny=pGn2DTybJrf^EMfUhZv(+8MVz0!QsNvji66<QbmqmUA@((SKAYZRu8l2lg
z41k~E5oLb;*%L0ZjE-(;K%UC_A7!rpy}z-`h+b<pDTn*5Xt<<W2b-R~K$DpUorOFC
zg#^mRYp#BZ#U-%B+ZS?=O=NlmoKW@kzR<@T-;Jg>yB`RJR@u7_*7r1s7`}9v$2W)%
zsWb<&fBTl8eiQEN`mz4<qrR_u_a%~uP&Tq4jOg4`j+gf96ho|!H5lO67z(XkqcZHB
zXxzO@xASB~#k$DRFY#w{%jZ=CGZMTLqf#HwWDAE^?!lH(i49NZxL4F%z?JB*w(&Zu
zs`KMOX}rITqE5zgxshvXTakt?MDg$v`2~fIdW)3aJ8m}8nl{-wXp#<w9TMd+CrJy`
zbH!G@YaKLBya4{HQn;zjSs=^t;#_B7^sAZb5Q(r%D~rhOq`r-TfL*X}nPwEOY^ov#
zK-1`SD-%&`H6(GIM2?lyF)PERU0ei<?g>BG^PhB6X&h2>N;aRAF?P;E+6yb+@Es{4
zhlw-+Z&^8@{dl`07r&a>1tG8bbWS7$WL9WbiG&-2a7y<=HZNB$6DOG+?z?8_o;iHD
zw(m&hR=?Tt)TK9e8%c!rbu#Yuk*y0_!VlAA(N#6`{l`{1eH}t*!rr$U-}691OUCbX
z6Yt1B5+Pyc1j$YVDQR9`%2sYwtnm(>W&rip+OSUtAN^d*Q}T(rieK#3O<*%dkuIXA
z0*CLeJnL{NV%gPEc!$x0MRFk7s*P<>f(>RdGqjYC0H5rJzYN8~|2{D%#}qz2;Qf<M
z^R*VzZ2yUR5P328M}wMGHfBp%*k4-}JMjQc0=lMm?8B%ty})dILb8h1hxfp}{o;Zi
zKr-CpAw<1=&kF&$xMvnW>8j*mOIFNK7O7<<lYqFU_+YcU8})p-vp=f#Iwn*UIhDcq
z{ToqypmA?|)KWM0C*6s&#N=Mg=(c0t>1Fn$dx3HDhEpjQKNk=?59G0}$*f&Z**py)
z^nd`ZfpOO^q&%TM{B>tnhR>79aSoS(K*;E;X{JoZ0!!3h-bdpEG2X7vtHd}6k;4P|
zc<LwJ!fN2w#A6Iw_-+STM{XQH4>C7mw?;l;3DW6nU_(ybG7`Vk8BaC^#wAV-JJ<zs
z<?Ta<i@Rnb%?Z2T$R^`TcPICz`hYgSd@!If{-b_!Z?+%#$R54T456O*Nf!;X#38#I
z_tNlLn2P{Ce(A4EsW2iZV>KpA%j+9|(j^9AH|o+T=cIqqeRBN$Hm_E`A@<3&+kPk5
zXPhZkR1w(zdl-OG>QtL76R!_iA)H1MxpS~zi;c@TkYvW6be97BfPW!2dub!rF)XpW
zy<}~Wr7yC6AkrvycbcpMvUFWNyzy}`Jft|(`E@oRiUipdps1WI&jr=+Z>jnp%B>9(
z!z4E0oCOnT>u08N$npi;W4Br5egMYDhs;3mb9F70u>tpjy-%@wg`(ELe71{k8s5-g
zIH{?e{k4;?;gLqNZDa^j=0$uX@*Ebye6E3?FRA%%SqgW0lgYY1PFX5SB#expcv_H;
z_WC(el@6-UcwKR|jJqD-78Qp-!$jbTDR~+-;Uco-pj7<vxVs1P*CQQC_|7NxlddT{
z@AUYMq5T+h)JD{EKb{?VYi)0vFqu)F>xQt~A_Q(=*|3!#V2|b@&4Jyz3P?c#3x|Io
zy*qANg8<$GW?pM0bD9=dhRZ;1s8d_w{a(=gsEK=#?E4fs>iGu^tC{URmsCa<=nstS
z@YM4H%`pxwWo04DVUxX8Hq-L~qSL^WM+kt{sGMGVpwA`Wu3@Fcun1IGZ~3|mqdfhg
zuYi%vPr9{25<<Tv!K$1NX!gAI8*I+<D-o}l&?pS2O(20J#zEt&#f&}iEr`B<nequ{
zUU9uJDrx}Y2z^ue^)iz6UIoy802ADlc_G(`<+fmhH6%rMXSAbMtl}Vyg}t!qhDLut
z`${0mLR(W=sLB^z2vyQM+}8{Ik~0Q3^Z2};a&RhP;3EUo^#g%3OiT*fpqTiRuDtvw
z9Z#lU*v>F^5bg_z3DvkS9N?cQiDvJT;Bz4IHE3vp77Rr_wnL*35?;A~)maFWu6n4r
zAN#_@3t6I$6S?U&6UlQI5+rs<W}gq7>g7SJT=v1PqzC0xs4B&Q;(XnyqV?hgg6$!j
zf?Y!?X1&j7ADwF!vW6fwLt6p4ua~?L$ZpP)U)PZ?EWsSWC^f4SXcal=71(<&sA`(X
zsH4o%17f!js#zZ~pRo9l9L}+;y$;b5!%lsIF*$1<#hbnVN!KTT9<WUFrAYGi@}RWt
zU8_)}c;-mjv3(x$5gU6-R`a6p;}S~6DC-FH(zZrx!EzwKt#tAnH^h`n8`7+4wz3Yc
z>&H%|oBgDFUk^aFY*dFj)_K0JW2Zh_$$mSop6d7(WZ;EKx=f6!c}B~<+duo0&b3$N
zP{iRLS#?I)v=A{}qfQ$!uS7~%p$_tKTpm7pAze^v7jjLh$;=JAGV6DBZ*7glv7`}U
zZboT>?i9fE4~xJzCJS$p3zpkQ&SLwG4AH9<B;mt5;kYh(iVNrd;5cU0oMgXsQu!n+
zZKs#v{g3Gz;#wjDjWjRb6h~tylBA%ly&Fo7%UxJY5yzsHT@8?zENUjBs#W|OGY7@e
zZSayjdzJ#bC&;Njm%}OAmSFIznBGsi+wkgM$kd0CxZYP;-u1W69S)RCQX>_faWX!1
zk2Hf*C9t%Z_v+#ug=f|!_m}q3{5rXl7U$+C#vp@6z=A2TkF74^rvlH4hf>x}J6-h!
zhbkdA7ASj;#L-Fz$Z*=}=Bf7Z5YEc{#@_2){bPxT@UYA`O3HUdb0RZF#CJ3X6*$1=
zO_S1oh*isM*=&k5S=hmOlhMoQ%IvJhWFPz5K#RXH*(SsPGSMy}xCD07are`kh60mK
zO(k@`#||~}13~D~7L(fLkUE7nt-}w_pd4$*wkZB&1(z6@D-h{b$NrMH{u^ibpWpxM
z^I;n~*VMu3`9rs7q88qdeayl-LN%pzhwHo@+Smj)sdsz_Ro}9&O~~_6x|-gZ|D|A1
z@545iX<;Xw<EU1jd9%Q!$izF>$JI-<ExH@T`Az(liK`<STTJ5Z7kuVMzviHs!Pgwo
z=cEy3Lu@gEZ1v;!f6&|fxgK!-{t4}TS;o@+1bPZC|Ka3gH*>qUKA~$Pl}Ui$Y382<
zPtRvtovKSOsNB$cxp<*qkVeeo9p~`V*Fs4stD9|S6D}#s+Mi+O-qwF>Fcrb0wS4hE
zvK~Y@8)_MUP_12h;5>FUn_rvkZfoS{2ru;~3kmgk!gH?aMGT}Dpc1J8lv8Ap#v^mR
zR}$1KXsp~vl_(L$?UQWCN3b{GV-E36<4WrG)rn7z?5o!<A1b6})y?|y$L&=(Tv$+m
z>v<@r`)R|P-Rc@_-z(kiRl)%E`9@{!JN@pCNg^vk`RNZ{Ub%bIDMm{5awS_-b?^mN
zSW)eDG6Lx1x!aoico(n0HU3rTP^8Y^gL0fK)$meNIy}_fJXqndcb%K66PTSqPDsF)
zJS!QO8~1Cc0+641IC1@QN{MX-S&9HUMfD_MU#36EgyYRQ(-D_MwlTg$DD}KZgV_UB
zmN)KM9=vvxZkKQGXjkw8B*tAH@(aK}Qk;Q?G27P#F1)3Dv8GM2Paoi4`rnzZVDe3g
z^m#N>LX4w%?1-PnS!kYloMnJDv;{B~QtgIrOf@5$sM)|+@-Gh-G4uF`B@E4Cfjv4w
zo(tXHR}&d7!Jb><>36&uoQ80Z0E$cVSa#aV)2Fzt{Ct)s--X~gHpj}0&f<f-*Z8Ix
zszkkkiwRE`zy3e~h2~JG5-Osu^#j_cvKz8^=iovQECI}VSQL2K+qgYqS?z#*Ic5gk
zThm-ONz@DX;T9mth0waWD1F26(H|yHVj)drzvOALB~=61xbI<CbB#KrZlG7EasMaX
zcksaHYD68jv>6BC$=T=W-i7naE}L5`P*?i4CpjKhc0AJNGwNMqQ4795KzyH9#r-lt
z>{`SpYM*kYNs>(5QYQr9gJs+*nA=b2*Ry)o3ls<bEv15RSz4A##!$Q0)6q+zUQ)))
zQ;3L4SfodLzN`#IDNFNpIsYTnN>Ye?WBSl)7h=Y^q0_CtGC4moC=GY)?BrvCLkE8*
zW8kTpe<RU{IE%YDCk;BD<xr=2cp3EED2<CmIQJq&%f+UD4D_o;-pvWp`%Y(|CaLd=
zxKoxVdS1``7uM<T*D(Ih@qel{{nt&ZaO7?*^J15*=Hc|A(+;~HKt+Adcx(m(bmP9?
zu*kIM6VCU#@mW<EcX3cGb(+{vQ!#@4=4+e*J)aVLGY;Hz;{3j~xA~JW!<wbI@rt(&
z>`;^bGjSP~6Q`0qUrJq!^ewx^P@&#$d!Ig(DoSq8^Db`0^Ej^slPy13M?QUYj53UQ
zk~c1gOR`6KJo9jHxNvUltewY&QaAM3T46^tOJGL=1p6)VtF^AFF{QysqDG#wnz^BT
zW-zQ9Knz3J9G^}OwOuQJXUjTLB4WFda|N)c|84xTg00;mJP{A(AwwJbq$@EY?v(ux
zd9imHd6ur{3k_teS@fq9sL9yN0RO9*)<PJ$Y8OnlBMfOF9~DGFsU>}eHcp!XZnV>#
zpEQ~&AgiiSzkos<@)FEi52pQ^@NwmMh2tYTm(Iy8rR7AcGX+IoYRy1LS3pEv(7E68
zub~M6VyYM3%o>j&wYFH!1~x&?mzWM;{Bd>QGHl=3ITD+V=})q3?#K5{SF3dal_%CF
zc$X|S&!<!#Ne>Ao0cIj`pVNmTPt^vX_vX`#J5)|8{KG&XFF8|%2~i@_{;<aD8Zs`h
z9bkMjIJ})TGB^Rn!d-sSZTyybU|gl$c)<)=gHt__#dZ)9n0m`A^bRM9%>=mly$F>%
zH^HQG@`HOWip@aK@|@=|yNb9T>;Da{T?~XYR*R5~hs%H|=gkq~1cOT6@?)=q+G?Bv
zh4hdO^)gEUIQ=)*_F(&jDnm#DRe`)vimjZCPG9&zGdx(5Tb&~gPR&t;o<J!b!s+Ou
zz53zes(PaZ+K3z(@9uxU{a_m4H;4g-7jjVk2Iy)8!q<Iq&9FnO*j;TPI-Vd#fcAEP
zXzUYWh%$Z7gZLV<M}v~`6GDe!W-?%4rZ#^==6V;zpk`?5b80<mJ+Pd73ZMT)m)f&H
z)NvH}bvzDKiSoLibO+}EV({;wkfvPPWzPpc>A*PwJwNGYLumN~YzhRm4~nZ&<y8#?
zLRB0u6@Xo)-W8{wilsWLueuSRBRf(#X~#8b(#QtgHsrzpv=Tyn3)~p9@cWIePOu;3
z|D<E`BZkm|PGL8w=ox+S)evH|ulHRL1$Zyc05@d4CH#{P{ih#&_a~hZ3E2s1rlQ(l
zxa~ChU(<yHt)hva!6~dj@Px(CWB^3kb3k=ii8HM;%Zh9^A^+AtcV5<Hmk+Msof@wS
z8-o6%yZVz(GJ*LK@=qh6K=;fy)8I%C>_#js6VSy2Fn9u_op&+j-bRCr??;(kdS%yY
zzjb=9P?V?&QMPgZqEGm|>8}B;W^XbQTCL=Lut~rLCz%MGqcC7fqOB~@Tb<ohy54<B
zAiCs6ZVUq~(TCw>WbLUHB_CipK*@A6xrnLplYRU~(e^+>{EIL$!>zqh7-Vy5@~}E#
zu~PsiMKR<Y3zTTKag(Swzt>EwC?^}`qhC<(WYIZqhiBi}j=--LEg7BqB$jMXf9NMj
zelXZJ5V)36ba+YcP$VW$sD0I1bOk~i;3?q7n^pQE>mZ=~T1N8x)=3s7nC7+~O(u<H
zb%iACder+abzvCUgtZvtN=TPrNq>mc@*Ln}Ri_>|W~$a}I{YXN-gWurgffeuCWY*p
zk+UIVL=O^xzCkvXhytW}pl~^^?$<5^{`_((e8)UfSIA2)o4OKjC1{VDP6LuFFOZ3u
z2j#)#&EhyJs^d_6mz5lyC7ehIZb0v^vVh_|WFbUd2#-a<lD|&cD6(kj*RFiE4lV(5
zNF-imblvslUXh)OzB8aYRlinjA~6n4J?CW#BOj2AQkt>vFo&w}QDmKbDqTQYgHtmt
zF_y-@TgqMszC}0<<bf!@s>tYSS0$#vW<vMdJml0z8n+<`jz_%T+q6A)U^I@Qp2JRe
zm`FJ#%|<R>yN47$2n(A+b2;=)vR_@YJK${=`rc?hoK{zY97>yxia8$;>^EMsURH+G
zS_ZNh_Q+sar}N&Bdv8&M7&;wa5B533ZYiy<9SJGq?v|qUi~2sJNSVD0qS)`4kg~~T
zDqlK^;>DinK*ZJ^_?;{xc6~T@Cg>s^9;*7Zoi7{>H2MKah<TT<+eSZNt|G%rWl4eH
zb~ros(CxnPaALR77d;E;gqQCJQE8R(^89N$SI-Ck?!F8YyQ%&<w1u5Kr$2uz@K+}e
z>ye{-02~;B&-$;t_`e{0{tE(f|GdthMfd0xW-v#WY9T!^-Tt((kayMt=7l+vXKF7j
zVb4gr**#3OSMbr;|IjBhs4uy$FBK7C|NQ1WsVNjeIMH*Z|FVu|talLV$qeS+;FJRQ
zg|PjZhRNcT>>Ph+`JhdI_|k^>2N5ZdgmO^8iR0Q1PSNZ80=MPlzAisL>Le6>FZSBv
zxMhIBJo+!F{i^_kQEfx><&v&3+&AAE_PBV2&A=!6AZ&TDAk1R{+R1ud=z_q}8nx>M
zl`EPrS7NHn`;D&d2FcW#K_)WG%>XFjC1w6#AUb=eUsQDACJ@HO9Q_CFkbk|7j^?r}
z97ko?!mZBr_^a|>%g7j+aG5G8Jc;NSEM>p{F}pqKOm|BnAl|CA@S5DdD^2zxG$O<r
zxA%FIf0Ruik64|XfszG{QcuMr&%jAt^g@m|)SvB+=>!a`1_aWFZ&lZ}#a}8NvD&0U
zua)V&#8iterS6N&K!-mNAkT5&?CIgxDrxog7AOlbco#-Gd`-D+Y<TQ8t|a;<Esg>k
zdU=Gt|Cj~#4VaS*Z9Ec{k!m{DGHz=-k5#GjPBjy{;lc-zuL?W^yjrJ>rn5HGGADuX
zjEZXNpc}u5G#l=%wkue4$+yX!Fn3-o?5acFz%Q|cK_0<EQZBN|)Z{L0K}fgh8i6c)
z7-Z^WY}q_<<4qcUmgTRaXR~TR8852ng3}yEkcFfU?I-2fdM_(B!g^9pOizxa5d=d_
zW#Q1L;XU^4NGU7KG#bFfgspNL8v_(pm#rku8rYEWCuUfN$_WlZCi^_byUS7enMZLu
z7$H2EkBr7i`uG+E`8@hbHy*EKoN<W8aragW=z6X8B7+h;*S0S34z&7CG>`w%zIpk^
zTiYP|fpL$RwDp9FcoO8I$UwCm#n@kr^g!5z9OvIRa4jeeaCJ7LOs>2pXbiJZv{cpi
zYT&p=Z$zx|;Oh)%&v>95GIU(~&9is3TiOH2IMfKQIL{(aSa$i~lZ-J**{%wjM=sxZ
zbsLixt2|`<DV+2GI<QfaEjK*g{$!%CkbTe*!b5$QrdWy%6@P9OW=p6wW}S9crEP>k
z2W6=rl;xjEQrv6j1M8ogX9|qO()VITr2(o$f>C@<nz7?MPfsPha~jaXe7hR{kaJB|
z*6X#$mfQte=~ijAT$dy89&du(3Uo+4bGcq~L^EyMD2(i-4h$UyH~<o`GwP@V4;V=X
zzXZgd=o1=r)x_Pekt<v%P*=6$*f})d<-E5)<UwYSGOZkw3H0;*65wFUvls4;X7+)N
z$b}x$sAKny(QMWt>2U7OkR9`c6U7|*iboc+n&&Js8&GdROtqKc97&Pr20pRQ&Etg@
zT59w;Q^qesTZ$tT_r4Zm$>!9k5tG%cmWGB1*cLb~h1dB}UE0&)(R=9ocM8<&L8rD6
z^ve$>zo*zWM8?@KIDE~Ur1v)=``COQ88Rc|b!#@?0V#Jmh308XJ1nYN#`8#sdGTA&
zdaIQ?QT~0&a!*X)GvuthI;)oUSQ1hip=Tffh*=5G$uN5IcL^H~<b7BluSZTjD%y23
z|KixmKbIqXB;bj!g&IJ~{ZZ_b@--CgfL#gmbv^WP7OjVqx&x&w>K}d<-(N)U{m8IC
zwg?<O`;#snaQ*@qpryAV0Jv!H$L9_);C!BgVc#)ewgVV!oXWHZ*I9vW#BNdKcgal*
zxJ@iP$#oO)2}G1IC{S{qroeVa8q<**n#MdbSBZ_-SWYFQ)sfE?BIxki)5YCrLL~Tx
z$M{tn3mWiL7@xGB^WM-@1J&Cp&<ycRRlE<TQstJok<k%_(gNu@SGUyD^sMBv0`X|Z
z8*ar7gXO?(OcDP`i)&e)EN{j|-XO@!09BZS-Qh!cZyK(m5jj2WuxGwi*!@7Fu8*99
z>D~m#e6Cf+0&6A8`!`!tZ!aTo%2VhfJYfJbm4?3nyz$Ad5Q;a9D6`T#6+y#LaWc!j
zhq|T<CR_nk2iF1`JH`4Ifj;>ype!?$nj&6J-_rZU89T;IlWAFom2EWm4cf3Tj|#1E
z%dGSe7j9*fvYH0~AHErUv95W#Y9Exz)Q0sL2+U+ZJ`$8sIj{ZgEVA@6Y6bSrtb9$G
zbk4+x&HE)m3GglD3XCWOY=p;!a1`sLR%WrL)c}Z*R74ZIEYc(UtUc6GaJ|oAB?Bl|
zD?d7oHY-FSEc*(O$<Idi)`W<oI3Ef!H1(dx=8slt?xy{oepNp@V4$NIn_D4cPJ$%+
zZ7Ye@H=B_YA-^RoAX}!<Y=iKzF9`+M*}z2jWVsiPPsU_vug@tD5@hYYwI8;XY(rzX
zlSysO^j3SSu4y%U7Rh8>@1H^ki8Bq+sRC@Ev9riey2{|%;E<4tGkk%Yl4x@JPdaUZ
zmtO*+vJM9mJ_F$G<@ODrY!Dgmpy(W#pCd=7$pgSI0}?0~skZ$zDWHs8@OAMZtZSKO
z^1aEMfEC_`yiSoOfltOwC*LKP-(m8>^gzgh8@Nd>>Mdezu^l1*<lb*=iq(vsRbH-c
zQE<Yyt*b3h6_ipKZbll+6kBvYIT+j7yk{?4CJY`Zr$?%+_+I~vP3-P@`5$R>{~vzN
zDX}MnD$Am40IM1V`KcX|fZ7W^2wPgrUmA>4e)G;C>io-U)>j*vmUg?S6Z7W0AWFN%
z;e4EOyqDN4ZF3u=o9x}+XK#J|+q;zzawIWRm;8!HTOc5brgY*P|MP8u8n5$4n_)4<
zmu@|||HJ(j^vsgcE{9&;{kz||xVp1tP9_lYMNF%|l@?b=nZMzE61{L7gfOehHceK6
zek(htT#?9d+6E%I@D0OD4j`)Q0lF)h1LBCpsi5$N#j*A(h~GqU>^n0(GiG!1lu=cI
zmcnd@hdH-m!>mewaW4HK1;k(bJpaAV0zh@PK{%^bu=I-SmDUO_MKs}o2GfUb&Ls}T
z1|AE1VaDpDtI`5-H`6-}CmVR9$3=Q&XkaEf&`4mt+{Ow@5Pu<eP~M<A@Cmz_XB*oU
z-42pq`MlfdXCdKpwwZ%KIWIQ9{8TZiO-#QsMrinnE1HGIvxH<>Z8H=9;Bm2`oOz%&
z<z?mapLA7pr~inz=a1;b{(>O=C!f#%?e<XuO>F64J#)%$QAuf9{JmT3%Lnuq{qT;e
zblwT8-Ji+1+C3XESiGK#x$XUX;D_HgedcS*Vrrepb0o$lUl1zqVPF22vJ#`CeARhl
zdyRAfqN;Xed&0ri-Qo}Ws!^L7IE~z)XtQ_!8l82-Yy}#;=)WknEW)vc>s4dC>xEKj
zbayxUk+0|{X<-n>uaP2)%YLvf6gSchXRjyxMn6&HLRI7yWqkOP&K;l1WMG$GBH=!%
zax^T7&;>VYc%!Of>6n@GFLWRGfOz9C%2>ZX=PW1Xr-Wp72V_~QIyf*S#5jHwzRst4
zE#G*do5~(Lqn0`iIkm*k+~;zo)Y*XB;qfSM+VPi}L-R%1V2ui9kt#mpMg&<Hk3$}Z
zpL49`EBZ#dQMPUqZ=@D=1GWB@0(8V<o=+0>LBUw=oyYFx<*~S=v1V4>(^Ek;<taPl
zP<`_;ZHH$Uh#EgSc~5Gdoyz7rB{P5H-i8~}9<`KeX{+jfDLPo?*jx2{QGXzBcAQ8a
z_9G=sF-H&B!6R+d5yci=%^MNP=ueo;m4S$NF^?B~Zvk*Gqu;HE`i$!N=a#FA6312y
z#kO;KiK?BECHu>^K_;}%`Fc8XobG}J3#SYGnTT?4Oq2F-yX(#5cg{4!dD_LpH-UWc
zh^~0l<MHY;1PSwL5d&vQwg(VxkFJ6~$mcwKIuHnO0-_>!y2*mNvF_TADVO3qknZ)o
z`$=0HXOVTcsCIDls<RqiPq})}q!^YeZQN%~-U8!B_u8~6Jy0r|;4>hp#H>vz+jf|X
zs-f^w^@xjZU=-6da$yi5wa>6c0e}N3(tVckNJgjj@?1uBk$nQzmOaszP3YQBI*q$A
zG!n9MEbPDn61xAIb7S$YRRV0@XF%TNb@H?<_8dun1D8~m;yLb^;JMJNVTw%FhnvQ3
zs9ZiY!m85*MJfKjl^4h22EFd*y=sFMzr1q6IA5LURW~w-KAup^(YLnXequ)B{KGw9
z<b?dwWjWJk<@Qku#75OTC-HC87Z2$txDI;#YYJg`azE*c&^Tw+aLv>l3BWD1R^QHJ
zoE*xXZXEP9)#jw+(Qjj1(4mP2wTf=At277R=`sBg4QhQ9LE|UgK-&^+h4n!}u0PuI
zU@kEBCav^9Fe~tT;K}OwTj5!daR3CBZ-NKX*OLsg_Y`YJ(7Wk^2ga$Ovqe1U^Xb^N
zK9<!R?w;BkU<T+}r^3z33?LI*9s(p=E9fIjDTp24Mj+Wb>IoUF?XLiwPxqw|uv+yB
zZCA(60!VHx-0(-S)u$b14dL(3xbS9^3}N!vP04s2l-VY19u;dtP9|gpu2*<M)^d@N
zW~dpKh_q>sl~lm<mf5bJB<k|PAyp*9xBNsbdB6*}{!{N*vcGvID!=L{U8M0rPkrx1
z9)!sGlddK9Lb=rNUF~^n$9^A;6B>zJgm?hHJkL||@XRy7MLW@bxC2uG04YG6d&|2@
z2S^b+0K-5~Z@p@SY|A=k(#9vLj>x=^zzK7!5F$Nr%_O9QJ-Uqss0j*o>yUDry790|
zAcYutcalr#(YVJ7hO&z90kY4>r$~2J2;<Vo(O*AH154CpVMB6;y945b^hi?Lq208`
zni1$1nrc?l$J3FD*#L&(-y_#3&fbYzNpPTzfJiy#yhmYxyJ?zdC3jd3?2250U*lc*
z5Ryp9WxGWHeh*JCkS~7uY>Z`R;%X3fT<ts(XaR>#=YrnB_E`_(-uL$!Om<?nz;dw-
zz9ncs!y3pg5y(j2FUv;?PGBn$dzm(e@HH|J<?pKSd)AtGjYgZ50h5enX4yGtY)4kk
zBDK0Kyl`JfG7c|yrp4Gy4wrY3xiMm(4U2n|8ye7^EUsWmGsFtMtvg9Oy)l(#!hx;!
z_vpYLI!zWH1_y%7W;&vLkP}E+sT;K6Nfd})3Ro@LQFkI^2d-{Gq)vA4I<9adDk}~c
z4t0wUQD7ePBUPG?ER^4nLC!;g9g}BD#(d0}fc&f_D|8zJ&I}!XL>AIHY`#0)@xZqg
z6bpBFI_Y_uckxU92pM5qBVXRTaSgKF#UXRBXGaXFGrD++D)prdf_3>)ChE^SO*!|$
z`Aq%TFK`bXMl3VT06RCqP6fitV29ltV4J-vC0ow$cKCof2h!;)hif5m9+jX~EJ*;y
z;7o<wxKtUC&IQEOm_ZX(XX7Wrm-*jF=yO|E7D6JucdG9<wVFloiwu<0)NSRRV!Yzi
zXzKI9ag<}!2em_ZjuDR%(Sh2bjOPbHH>n`)bbDS`jtA&$M(X-yRpbiHSI>$W$tpE6
z$&>~H_UQ))cS)x6d?ScoF%@b3*P2@C5q)-;iqY{|q2yy*@XqP9dhD6)d7~I^7cP&n
z&h8bM8gLC4v2F<vzc5S`B2dPy%a8=rqkK~g(`g-$GbL|j7g^hIRT2e#+GMZkWL5*y
z#j<BJUl!xA?+b#%h;VzZ4s}wlmQ{Z84_r@X-o=C#?DPi_bHB9NkvbO8X4h5NtVrHp
zDW7*d`rsj>%=5d<Qj^nvrp)_4`rL@r;8QJoKr4(SqU6*<JC0Jmy$V0M+yqw}z0A5X
zQl(`Zot_(VANM0{CG*WYY2P%3JIYb~VUxAQ#fN5NQ#&JYHn@-ISYD1^Zf}rD|L}YH
zZ|gqvYbP=QiNc;wv{qv{m+rn12iO|?-4&isHihSnkinfEtq}=!uIwS#Zyh5WTjF0z
zDhm}!m<-2xedT7jMY*3n;^M`={gtY3+zkM-&MMgfSqdk(N&c0F&N)UQg*#jjZ_ln<
z_#fdhX#(@j($jjnyNk@O;Q(L7=l?tvNyf2fN1rQf@B>+mJ7l}*57@KbsQcl%eWJ#0
ztparB!*`YgatV{bR;#n1@h=`=hPRn{HcU&6a2EF>ISKT>`xF{OD=B5G0mOd8|D27N
z<l_V9oYrSYD;WA>zR!A1F_1Z`+lE4gTpLDi8dX7jk$`0%jh#+dJMCR*ADvZ(E^t0R
zG$|hUZSR!kK9=Et4V#2uBm>3BI@;yFon%RZw)u)s{Nvfsu&);vxcN}fCTyzsa3A3v
znw!Qze$cqbdV@15E=l_t&$jX6ZECJBd$JDyY?o3MKMLS2wtoLdfrg&W2VdN!2>aXV
zSpdynEt0s`z`|uoHH#UUAJj~XtnC)fb24O?N}fpE%gfZelYa0n&Fx36tYhVT#?*dr
z_|xU~j<0H1>F`PVd?PY%2028N_zh(yA5`nj-eu{wGEpP=Rld1A<9toh0uU}{0MXy?
zgz^6t6!PzR`F}3G{Id%4|3#I^5F<TmdAq^oVU*3ejk`0qOuX3b!|1y4rI1xij?L}g
zdf#7mIQ8nvlskrje2X+`MZ!d8_PV&w15A<SiLc&smThPqFvr>I=cT?Q-r+u3RDwRe
zWX8ACIT!iSbB`0W!hN1;D~S&Oxw?5Kq45jZ`t$eQABmWCt|!O+Q+?TVoTyjGVUui1
z(n+sIWO6ZW!;1}G=Z}o_{6$jv&(Tz?fT{7B)F21T$2WCi&{g1pY*h<~>;5-%1S?uC
z7xy@-*CuLn4Rr4;yt2ewl^V@xmsfPn(j7#JRB~Y!)O67{sLy3C)weF&){lnl>Dab1
zy)8Twk-22~HJh7yqp>l+>EAF%V-+D&1xN<*QY_0NShD1$CZVsc{KLbpMw^f5HmvH3
zNFzUIpOQnW<>EnSy4OfRFV}3C>hZH`5XH81<#Y?R<mB%8lsJQ8x|5pnN3mvsfO}VW
z#b1OVYGjUOT#0%7tZgoWDfo89sw6y%%nQ7CtBC42%SWTeq!XjwQM==~&l(vvEjt!(
z;q3>d+Cz)GZ&n~vK<)e#g2qnk_AFP-lQeTu9Q_a~n$Bf=zuAEwkT){!6P=10JT)D}
zA)j{tP1sGkV{{Kz24DMJj!_t``UVW~(Z8~xZ%?qB>sKoVyxtSxe%F*ycIHd9Umfzo
zy8oRSufFwsGZFQx0*2he!{0j{grmbXcX)J`Jt3S*c3IvcIIL6OTJ^kJX?k5lU~no~
zZU)}X<L!}wEsqwv?EFO^D}u6Fu`EmbGU_ylarR*AJmz6^fyWX8E$dy=pBQ8hX6Ir^
zJ6~bf-5GFnd64gFTs2-|OzC>xr;#GomXluqVuL$>>Ms`hwFtidM=+0nea=4{$o@+W
zWPjdp|N7Np;AY<Zw9582ZX*42^4ix6hkGGfb8DlJRrHbx(AEkbKA+}?e+CFgG`fF_
zK893W%?x|z9{)xEOG@*}C(a{_7VjUd(LL0Rf`23@5{CUbr*9?SJ!NJ}BRqVX$p<a<
z6<r81b^Rf$zQ=6s#N!G*SLfwY^c(KP6^mE<c-N0ks+&^0@-<L-vK$G(i4d40n@97i
z<6m(l-Sm_V==9ZbcpS*zN0oc*?7s30ar#lCdbR@Q7CD3TiOdKq>YFOt>SojBni2N?
zPIcV!lNXfbnD^z+Rdx7Z?0t7!Q*E|x_&`yr0--|`5KsgZY0`p)E+SoNQR&j8BPBr;
zkS-wd2?|J+8j&V7(nX|~NGCz1OG4d*B;4(sJLk-NGiT1sobR4{?z#NIkL*p@*?IT#
zzUx`fTJKr_>ZZXlr|Ps=&q|&9^<j$5(ASn?zFwFA5jj)0wCdp|*lG$w4hFrX@?haw
zLPCcgD`zuDT&|1RQmHa)>ME_b4DBX5+=X-rctt?B>_fO4Bs_65oQs}IX?7pTrLC{2
z>k6i7_@<*ofAaA@y(X-?lq@U1PJwfh{N2ju!ygA&_1kOFQe}eI&7&F|N*$257wR;6
zF~wlQgJ)%pxlzga?AWBw>abo+);?q`5*LRBg_VbD#|Wi+^uw;MiBh_oGuVXQhMa9M
z*cTQUK)a2si@XVX&e54|1C;LNa3NWerq_ivgozmtD3$SUGO9Zr?4%@7`Lle*Z<5i=
zRqXKwz|{u0cOm6PSwjhOH8Ab@uIU$dH*8YyJs__*Ku}xKXHUOad}_Of9(Ku$dW>w7
zjE9bFmG;ym%i4AE*r;E=&ylvtY}-mgJrRzQW3;cm`BM8~<Yfs?N{RGg^Oh}ffUC^_
zpTKnM9-C#io5UNY3aO(fLU1cCca;@wUwzj$wD)M3^>QX&ecDmp5&a~G#^g3bi0kv`
zcHta`|JLgIAGN6mMLz%HVgL5w|JNLM5aEw~(A+nx@(Y~>KHJ^qyrH`no!;6c_>Rz&
z3JrgJSHg+JFa&P$7w(*?3<IuoM&-6Txvva;O?TR@ajXh*NO{9-e7&arxU7&J%W+aH
zRu0b^t|mYft#K}z`trKICK-t%BRlG>FRx3&^xoHbR`WT%_$Uw-3+}0c)Mx}qw@Qx-
z&39y63}N0}Cv;03DNw&Y!gHdg0J>)KSC>c=uZd0>T))C-Zg$25O@c3OA@{rRTauan
zXtJkElCjr5<bg|B?(%5eohL8-hkKeYgm;3qDiPj!6H_5yo426JzBsJaua@^W<))af
zoob4>&dfb*l5E-S{w0wInVnOXyH@^_?f(rk_&zgwS}h$%jS-lX&&TDSy5jwKM~*Ei
zx}FaGdcyo^@KpdO{g`|$?W&nnJx}nM2+yZX-6bjqut8;CV+kqJ81rR(;FVyE6{_4?
zqgkMej7WN{m73jJ-KbHeR=q?UogT865L@G|yPm^YkvB1|@!7kB8!1%blS#WUF7;Y<
ztLmAH_XT$uC=X0`B7%DrcXRH>lN^TPe&c0J8|)ubZ2xf&;ZHn>gT8rxC6WF2P&lB_
z_7Zc6>Z?quOBSc|N8d2<^|5R@@lk&nX#whS9m#m-oDj2SivVKO<+v{b6^l0IhTRO;
z@~p4WS{t@0TijddYAsTPPdaZm;3WdgU$Q{e(Zc;ycxPe2CpUo=3roLe?R1Guf-FbX
zMmC}Ik-|^ekB+120?jj93F+GX!PiK`yk4DwUcv2MRvG#Ass)uUi7(dIe|QX7?L*=l
zOt~HfUwBa1)!l)qN31F2fph67@+_uJ<2g@>#WY^fLc96&YVxvWU82#qphQQPjKqIV
zEci{GQ%$!wJ+ZUw(d%n5ncR6mh&OXeV`{pmSWnZn@Np%4{S3o$A>M#>?B4mry@ioq
z1LZG5lhU5OG+XfVtNkuEM?eW)%Hca)UAdw1!oo0>IitZ3ZCzIrX;|_6=wU$ga&BND
zv=`lqU`=4_Ld!*~<p_Jo@Ke^T%nfyqr(u-&YOb)isxc3$2-9li-|D3CXgpz9+mv6u
zkX`xrly&`&Hpd@w1sthrpiAOS8kPIE*H|@S!1WDTd*MxXG?C?P_P<ld=it2f9}B+x
zF2D2j*bMlzgw7_uZIxgOQ)O0*hJ3GtM_RB(Y!$yUHYnE%D(v8X!wtJMUQU$_SVBPn
zQT+Dc=~E5Qc5ghpJaIGj>s*z874y_fuI;r$ZF;g&=RVGzh=#1AL*eXI2~WnY+tOaW
zP59Un722_3Q}tA7)Grb@rQGVbdN|!#Aor|4N9=8f2Jh`>M6noIMOWD6h$TMl^x&Es
zoAc#atiIXrF^?_s(WL=G9ecM<OkJyww3?f)vyAgMs2Ub7Px~SM<LkGR4z0Wp%)djo
zXhp(Bt&}1J@fA|RaHiVxNfA|VN`;MX05{VNyp!q%en>u;XbjOXYPLDjIa*&-=$UUh
zIqwCTnHF%V=59*<-`@L=2F31VM#nFs*S6nH$ctQ#oUXrP#17nwiLp%lbLYJvq~wdU
zmxjR9q}kfZ7O)ZS$csl^v`?O}$&ep!_T5VlrL-!3?8~zQDoknzeE8}ow1XF4Uv5@)
zX_FNQ>cAq{GcK}y$B#~%nQQDl^bU1dI4x}X<d%q44wFn;?}hECiOXr<R9t4)+O3V-
zZG@UChw&a#25F)u%Ekf}Hl=u0)>{GT1}@4#!<VC7P|=jvizI&6=kkdY68tsK-**)c
zUzXGy^k)xTFdURs7C1cMRw9{G6NP%8#lE#wJZxyO`WDpFewS>DgXSB*4UPZa$BEC*
zHA4~tEyQWLEC|oj%PhN?M=HH?66%&1sHUG5U`(!LMi&*OYkYSSK|x9feKor#+*xDB
zi{Dty2fz}Qc0QaRYt3)dpNy2b(4kwjJ)h?bXK66@pMx<?EcyFs_I!m&S*qQp?o<W3
z*W%p7+Ma(GxZc9|v5)Ynw&>_AAiJzsx=DE3e5^(53+p9o6XMa{bw5q7h%_D!^T)vm
zqIoRXE0JwbLQ%@K<koZL`>i|P$)Vl}f!kA(W$F4=F5gTV&Oeett4|TF?-v!xr(RLu
z66iKDwG1)pU0QN+a<@&wvquTjKU3SfSN%s0_jm5GGnu(r9%B^UT$W?Gc<J<EqyufK
z=*po_ii;URqaINrjggm{LKJC(%af0}zd#EB?08lV7o-R#-Goa)DfnT9N~WjPeMm{3
z@pE1INoQ-TP1b0tY!d+<VMFb}_64s>1!^s5uie5NqMoz{=Cs~Zhm2|l;D$V=r<Wue
zhbt$nZNP4mxAoHfI~Kn3&~wxw(_Qi=N;v(96PFv=&Og*TTf2+S`cY~t+%`==K(?04
zt$S6Iv0ZR}BYwAeyR_Kyrt9sSZa8frn<(B;dYsXKg;(wmuzJzotPlHlj*JH6pqc8Q
zXRQ3*)AFCW-V53DlP7g-Z-H65jG-E!33m0&6=#z(xikPu3%e6@{>^=y0rl<MB(Py&
z(-$SFkz`w;dAeBh0iPR({Ud#<2V|A`xDxYS0#cY_y>7;DeIGAaJ)Z4Ey@+C&Y(MFn
zc$e*}f_=iJ=&g(f^IvU{eN<R2F;?&-;+$2d%;!%%Pd~G0KmuK5qF(E894nz^ZlzBP
z(eUa4vyC2k7qR6G&qeR1erOUWu3|e3%BLyW%t|f^GTIJ@QfwDP0~sJBWOT&(;#t@E
z_sKsl#|ei(dLCfIs45F;@<du^&L|Q8OVM$iC+0A^kj=W&JYj_*vn;v8o>N!Ow|RBU
zQ=T6VJ@s2sm#@QcE%&F@jyk1n?p>F)WgD&Se)m0H`x#w&dRy9rPoWpe8h)B{>dL6b
zXpE;@a!nxE?_o3*Pv7RIfa^`Gse6u(Ybn~v1w4stop`ed(yPaA{CX#!6!SmzOk0c)
z^4z#^|5~<5Y}gUmls4)RZ$xeOBB#+p(9=)TRvEf%V`8*<F2N_uvA*zMEPG0>h$t+5
z^ZVp=OY_{w1k*M9oc^vMr&b`G=zSeP`9e!S%jkIuc^FpC@pM|O$XXjo+Ui#~g+|2N
zvz+su$n%~#rp7TAFuFZzeO1=OT|Ta{!#jGqztR%LvUq=SJN@zTkf~0wv}u*z>w9dg
zwjJaLgdi)d+6hLA($~`TyGDH{cP&*yB___=CNV17rqq0lrtz>j6L`8KEym54-JiEy
z#Pyb;X@Z$4B(vj{1~lvUGU0zrm^?7$|ID2JC;#V&)Y;NC`4_YCcMHzgsed6k6i+*C
z@954uj2%(DvuwUl6#LTm8W?$6FCdH0@JY983_GUiKTD3vaLV8N#CrKqP|8+AntjPT
z-B>n`vzGj~nXulbYbDvkCEu%FI-Gu*LxZKeJ-WGE=Pz`_NaX@3)bjcmaNX<Fi*ktH
zY#O$HTqszJRxg1cLH3g9!^o!c#kL|n?44D2Q+A-P^Cw2J;4X>fYlMsfwO4hP=u^+-
z5Dnb%N6Lx<Qan)J7mY*ckA;Y%Ie|lO0y-^5su)9VejXitQi8`_#~X0dVT*xOtC*H?
zxzRkI*t<1GD~(6neOiiMK6v8k{ADn^HsKIibD-e?Q(V45(@o;{tCv2_PuDs4#_1`a
z5vPhOLQCv&%c3}=H4+ied_9+Pc`bhKoc<G8C%>V*LuOsXmQqc-J4iEZBq8y}{^YZD
zp6lQ?jZPGYtViTh402`1=ybs9@a^}4n|Y_8etfS&xW%%HlmtsT8*|5JTI4(LN8@z5
zczD_a5<1ZA0Tueh@ak9jI^s1+=g*vF(0tHw6#8;WT<ce7>n}Ug1<?7x`A$~Wg*uY6
z^gQ1~u~Z3JyE7Rt1<6>qY-^ndxnn(N-8f3w*36b)GGU>3>99apw8{<Nm^pDA%9qit
zoAzaz@w~Gavt{9eM72}rMNNUmBWN1JS__I}jI0xKQ=9YC#TSp0boJb9`)OBgcIhNM
z7_Pn;e?+SaYU!mWg@xl3x;(dswIq}KtpW`}LQz5nKeEH2r+)ucfaXA;{%;~s|7BF<
zUxe|~O%^|<Djl&LOX9y&=iN=q$n<qQDIw6N&kkLV;D+}$41It>HZvoqML>#;*NL(^
z^pr()c|<&m3C@3O`kvbHky0>w51^z?4W8YH1jI|UVZfY{NHs}9X0%j3_G+<PjYN`e
zl2&5~y9UG?!6xU@$oPTOxb?*L0h4kz!^NpY8@)Z%;2T7-xT@|6eY<4-Gul%xu}xjP
z*TO|Kd{j*%F1?YgoX}H{_HbchnX*u9N#13E%u_`JvONw1(uN~hrp|7|Q0AeeLl1U7
zc*KsUwk=P*_qZK@o+0^JUuh7On(0ww(O5OCr^33E*m0?dSNid0CaaD}n(s5SQZXjm
z8yD1hU{Dov&sogjsio>SXju_R=};Q3iH3Y<xJgVuIF-=Jm6TQXnS@2(<8`Ypy*G-z
zUqVnYx%2d1MglHGy7gOpraiRKo%MN^L+C|x?DPq+gZgZG`l?8xU+npWYrJ#&5Jv<4
zfP{9!l0JI3#N7GG$8*Z_(3PP}&^G&1arRe6tS`Yz<)?cY4WFEdKfVvS)-S!D#X>)q
z8sUq82FU2?^JB($?C6$U7`J^0ESOnI*o{XP6VybZk2#(qw{07Pm^o-Zl68O}XCmKK
zUf+pICqBP#%UZ1o5hPTrwUniRnvO`k`34NBwk=VX9m+dU^6TX5rX1NuB~LvKh>W(8
z{rD1h(5N*w5}zyz1_As^10}SFpQ&+*aqL{=D;33KMe@H<*GHr}>a%I@YibzP4XpmC
zohWbA@N0z*I<*h2c$t`$Vk5sl=8RnvO!<{q=@tnV4L0|21alR-@@x1&eKYJgf&Bke
z(~kpT{r`}#{x`+>SAk#$IOdxpIw#e|$_{x`l?bK<2`X^46#ui&iMr?PBKacct)Uz=
z-)5MQbfBeSMpc?IF0OGd6q-L{?oP-v1isBSx|k4Lp9gYyS?ix}Epqgu7v@GIva3wR
zxYr9_s}$|1g#d?e(rgQ)!?|_eUQJ%$=}Zx5nGK&9FAH|X%At+|`ei;d@{EmhOgS$d
z({}Df;CrWuetmS_=p8WBzg24cBq{QwO7xBj*K9&e6`VW39lvbiwV`z}mVGh0mgtQ?
z)Ypn(ZAUqJFn?Y>liyWQL%(SAxlZ8mJu?aIEF&*7kt1;3ZmbG@Za~=Nl?SfxvM$op
zw!86P{`zf6FQP`J6Q>pbqv>RTD=2o9Tf9%nVLYlhw`@W^7p&a~c;t;+-?F&mGo&jl
zaZ!=q7vHF?2cVpBd&X}0;StR3)im+?Cg8O>tJw0kO8gC9PriOKJNl{U+1b28{!-;H
zU%`#bE%Ti=$E!!mUdK;6>(TPx8b9YbT|1K?cxJTWbk3r;Q-z9A&jpcidPQ2HMhOD~
zIq=lm>dMM<>W#<EqkY?RjYDgWiM)RBiwQmKm7Alu*Nl;$e|cxH(CC%`s(ExrDMx$f
zl|U|k+8oA7nrWGOB3WCXlP_w3+^p2X!j*r)rk&rgrT%W5E#gNO#JE{vDB3+!_97pP
zi}K6PR%`f;!C=0Ev3q*&@YCl7FZM!8Fi*hZ7iXy=)?gVWU&*)CMwmKU5%!F&sSOR|
zFwbv63_o@*1S@B?;$8_dD@{#n-{CexR=xgKw)%~u`uzJ)eWmE!Br~u*((j{E{5A9*
zC>Q_zl#74T%lYN!|7E?eh*<O#I=r)HJfJUMrGCdcz{l&3dkC|g)y<xJsDQ_;7OSUA
zVuYT4IKF0mzpYn!$zideyw%WJb@UP4*!cv>#$v%`jrTQQrNA0G4)w2X*q83&2QEh&
zF}ytfNXJ^_KMX{182(!|hJWYvL65>e34;HgcK@Hb7JI42sh1Z*fRv`}{OmM0*TX((
z(2_`HU9bo;xIbO4_f9J9KyhI7xLlEz!<A@9G5Hu@miy_QMI&NybBgvSC=#`xu%w+U
z6Q-6jz;D$wcyegv{-(L1+Cw+#NkQ|ysu!NP-MPEACx^Dm-r!Gnm&Q)=%DgG)W_nsJ
zqIm_}BsUcF^v@m6--K=dGedQdjr>Qs>3`a9unMQl^zzyX`jLQ=7&Obr+H=*oMQbRu
zQsO$F48O5rR94dA*4K!ZoU&;r+xOP*PI#`}GPe<kTfgI0(7#y*+|8y+GQCphpLZ%w
zwJ>cj(-ZwhB^3FN@tl^(w}1PeF}MeX2>)v`%Kklv{coMOdzy<0fi`7VNU12(J^r;T
zHnZc8k?WVnlV8g!gJP^+)%T;06^TG+V_|_mmhwhZJ$wqhAe}&JDd^0R7rKU4JU7kS
zv61!(l8hd(DOHy<PBw=pgi_SDCU_GwTRvM^OxpiPeWU-@J^r=7NjSE`iA06H%+d1V
ze|=d?gtgjh;k7itog+_t!8FPL(C+>ow)@~3|7XVeh`Orgy_C^J58otXm%!*o4HG;3
zk6*!jiu78-X8G3+|LE@LM^CsaK}pd+&Ro15Z{M0-Gi+{m-Uz{&e?4u%lfzu=XO3dm
zlV@@%o;j128#VX0pc4#)jL%)|*>N1l%)`!wER=SE4B;PYdwoFF{xkgJzv_RZT>^e^
zv7NYMG;n?U3zH>=CFi^opW~itvEcmmjFBfkBB%!spkqcd5k}m0fxJr*qMidF?a#ko
zX>RKKdSdbylLsDO;#1;$Ur$<k<@6v7tqE<HLtO994!qKkIBgupkm;#oO$W~WzbsAg
zkGd+=!Qe(-7(1$#;==nUuQQL6<xP+xJ|>150g*R<yH(-f`1jrV@n1NeG-k{CHFe+S
z_%Myr@5+J|HLpo$I+T(moXlRz+r7{COyv#UZd;3;8)iAv=&y0b7c#<E$~(R+=03Uw
z9b>W>*9PsamycObXQbV=UuW)@bd}`S;O0EaB;LuLRDF>gWBGJI1NyUC&41{{Gb32R
zGIRqZbA7A6NN_Xz0!1Pp4TF-e<Y%^k4u}`;fo1zYW7+=MafSXUIRREdxEY#h(OE~B
zV7VQfpjj3cJq~qbej#-lY~9M~ZF!zevwJQ64DXLyrcLXQwxF*(vnD?4iLZazfle@4
z?19pI2#ZbhP;I}|$*ybKIhep+r)hXY3C7Sm@uef7LOyNM4diM^{u<}{d(Iv3vj0Ch
zr(BBFolxQFR~it<$%Sex|LPai7{0<?MOi4x)~$80?u<mh=u1C4MN!Y>FMS@bh)-4T
z{b6<EDN5zluj<g#q(udoTX~{B$4;qrp44#f!C`j4iBYypL<fG{=^A?<WNzYe+46z{
zXSA%*yGlnUhRi_SFaKU!=z~_!zs+>|f93d1UUjs`)|1jLs3gaa(GRwh0e(C(ckjln
z4Ziy!tM^9c@@}+psqZuMYo6>0FVe#t89teZf#Db3Rfmd*If~uAk#tYoN4u`LSk%gP
z*?as?Nn;9>G{j38<5<4)<NOo0@5LsnNLF7J58*bqa^!LSJnrIIX{d4Y@3EJ^{L!ZV
z2~9hoZ2yJS>91}}`@i`4G#t;%uN`q3rb06nBHg}rUCRmawx~2qEEHd9!)MGZC`39K
zya#iibaSL$d9qH&&W+cf(Orx`;}UXhO7{NeyudKE?4)`f*V**7HW2cf9;}vKNqJBA
ze9E4H&^>imK%xJ;Q|XHa1)=r$BSl_|T<1|V9(g{Rm8U9%Q?^eV8YFQgtev*7cL_=Y
z`Pv^cln&zD|2N{?zw>88`JZYj%xHMLDHJbw%ah8>Tsd}zL7~h=ZLZa;;{tmWYk}^<
zLg}1MZs1cKr>FPLzQb4AW<q-|4LxcuTPoJepU_Y0FAMRsHU!%g%3Sc-<M)LQ^x@uH
ziYnNvMdlf%x7HjFGktDwXS7cENr6<Zq7ILV$>6O%PXD&<tv4>OWw(E}U;3NvLJnf2
z|DkvFuh!y^R^xyA@1*C^5Esq3kxS$=CFA-fGl3`5YBh|W<y1gpomCXry^>^t?iO|7
zy3Cci&y~eTPb)?zE*Xei*VX6!4xjBCEbelY4YBq3RBv(lbzP{7X-uk#R)v-VZp*&@
zA~(Ly;%-ay*pV#KLxZxwD%ITDIAKB4pS$<j{vpf#Km8f}Kl1(P&afJ$q2)6@Ja^{k
zH;ZMBja+Wad?##GW_q8`biWtd#GB~-t*f5$Ln1r7CN3~>NlWCq=A2et&DVS!wNK9U
zb)wp)E9kPepU=@zTaO@tV5;@K8YI_`TPSOBlN!JCh%I49^Gr#GK()QH-f7b4GL*hn
zt><MAv`Q?(j_0(v+LxlftSPj{6fpX!N?|zO{7Y|kL0U%;3d6@b`ua<2QO!4tLa_=F
zUGeNU|2mNmXy0F{TJ(RBe*Jf&g9rEXKM)_!M#Eui;_z^_=-VdAdR56K5}6hyR9Llp
zdg`LZGt;7EOPL4ewrmDWoe4>1eT}9l=1JDN1+6)2NoY}<iM7txsII0+A53%o;*F+s
zUzjL8?V@#xn8nE5`4PskoJglJaFhJc-X1_YrEl5tJBRz50zz&)M_iOCL*RY!!l?O2
zagPhfCX=GY%h*AC_xtY3KDge&cphNj00RdYIKaRG1`aTAfPn)H9AMx80|yv5z`y|p
z4lr<lfddR2VBi1)2N*cOzySshFmQl@0}LEs-~a;$7&ySd0S5jI1GJxlAM6<ip9O5u
zYB6{e_bq4MY^JKp<{j29eQyk53iiuBWa0H9X6JJzSs!^1)vT-b3t5LavwFI42`NCh
zUR&b)`B%MSWvwK6(s9HYRI4}{tcsTHg7+YEbx}EJ_DaF>NCO0EYP4YbJ|wtnAJR7P
z7iX|I1F}p)2Z|)x^s^UNa3;>i7w^IBqQ;1I-?a8?S=JdT<+E>pHO}!|^n;?ks-jRR
z{51w}Pk4yhys<g6_uYUhn(*g`7@7ul0-|#t!gkiKkx=t^D-Jn$q3ZkF>IbF<uZ8O1
zfISTCD6**^40e4D+l9A;!hJEgI<-Ulkbx^hsI?bxAZ-(@?gnQkpC@{>=GYM8(JVE&
zFDz($A<F*fX&d#$G^!s4_Y4W{f*g<|Do`1}f(3#oV3*D?vjA8doE2sN>n57r30udR
zE0<o2>_#1p){k`HbDpO^9IDp(_oA?j3-8~_N=xR87)0;{L`+<cU){cprD^i(xPOk>
z-*gvjfP1JyB*6c}O~rkPK8nIQOtbAsR?zx@A&CciP&UN&A?#YeIY#s$9+7$QuqV7N
zi)ws>Xx~QtrynI<ZDq7gd=&*ET$=VF5jTLoeaKa88(8p8I+-j^yJJb^%b7(`I8+s=
zyI_OhsGQ#%qgyh6JERG}J@Hq@_y>{)tmg{rkgr1=2I8^S@-xff?>&;2pGxe4eQF~N
z|MzXo{JaZ^O>gQ^c0zgwhS{URCfFPoV&1xZQPB7v=jt&eSH2VN=BK|mk1)AFG~#yA
z%}rXlUAaA15OT)B8Z1>M06VU<A&uW}rQgcMn7KwdlJ3M;+vFYy@|{FUOv>%nVg8|F
zutU_{xO*Us5furKw<L{z=&fE}qB*<6Mp@)8O41L?gFHxRLyTME3R>Lm_<+Se%lb5i
zdHj!vF(;jWNnmy%UjRmC(XnPVTxu-X)vYgXy<;|$@*LC?=3iOKoNNZb$8HTKe?~dQ
zzP;0kOQ>e(q{1T=n+r*zUy&f{uPyvHud?{}Y_>ai&wt;-8VAqj-=2-`X{RhY^gNo%
zaB&|ZRzOsyVzT!kg8|DVD0;s5U(V+J>)8d`O+M-oaM=_wu-u2BNB1H9amYyX`2_Ou
zOsaMU*ryv2w-0&EOJU8T%ge6CMDe!wMUv^!_sr}hM@nB-Xd0-0%8iCBilf{59>EDl
z9&u{tz#czt4Eqq%svwCeof9v9IroY4BG1&KP3ztRaM9BDO<-~9n}vth#L!N(<=nv9
zUbH=V!>{Gi9Q}#!p>UHo=y|<Xu-mqOryH%2UO7?1x@mkOmMhgw#BB?@C%MNI>`(0M
zoOy^MJO>6_276b(l$KO`KGT_l!~r53iRv)D0YpRw95iV!$!`4(5AfFVShFfi7uD<E
zr%$a<Jl46TzBe6)xs7p{;kk!6M*Ris3|H4$l9peXcg5GMtG}M<<qi7WfuGrxh&W7@
zBbh^Mjvib;$HCU+;XeA;rzM4Tho{HK<uy;;pc($LgYKXCB7yIFq9#xMjN~EuM~msj
zIjmbgOVe(q3A^^Gfqmf+B!p1|`VCVFTNRiETah<k0AkuxyN^L)T)01Olx5OAzO>{y
zOjH-l(?ALbyMuRLLpC68ZB3yc3rDu3ofKPH<;|?7p$<iKAX8@9kYfm#`>BnDwl%07
ze_TyGB#U(cA5G>2j6t({8JHl9u7e+BOJ%FK{-B%IyJ%JRVBG7R%a-G|HsyT-`<|K8
zcU9H+1iXtc%z5C)%LkcjWr}CC71z{ZU_XL?+39co4NK-6qH4_0ps>pzm8wP$Tadwy
z@nPerkM*<CB+4Ib_J$*wO^bScy_4FkfY>{QT*l(G{`}0!jb9AG5~im(DsRwe+|T51
zFbtW+JZ?HYx;bisDS!Js;8xvX6A{!IvOOV_%mMC-kI2L8pXm*;7{O<JSDbv{D#Hjt
zMS~XDsi@9s)i-bKGr8euPR;Ne(Nge&#-mb#Lplu^YJ!T^*d-b{z#b0+&Tw@KD75jG
z<_0#RQ<*0RveqZUS=3=cgQ0Lgr?LyPeNWIT2tsmm-S_O~wJibB(Ir2d@aa<y6n!s5
zRZW{D0*6@HnnQ3b7@f7KPQN3v{bTpTZN)6s+u0Zx+>R9xcWn2Fnh~fJpl#=Ef8GC0
zM{qIvU1*Vv<*5YSP1HEp;CYaGjCg1*T#duR5@FyXe95=eMhb}Hfz?4CJq=2OP*k!p
zuT=fFXMr6&<Cv5|8|t`)a5auOZz$Jg^OrhA#0<emIzE%73i>{^y+%z`Ls*_PSNTz^
zz)&65uL@l+STWoS2=L#2JNSJl|FP4wOnE{`Z$_J2O2FqtuwH(0EPEQM3@qP@PJ+8i
z1L7W!;11d|niUgxY3`bCN~U`&XXi`V+D^d_-$T`EvN5!3qNwKUY8QhwhegtcY}B})
z5S}iiXG8pcc!ey75AH+U&<S;_*l`rjF5o%NJ6WKo`yK2<KmDeq22vi`ha%i$<qpuK
zO0Y)p1_BA1BDB8u<fl$<ol2)OUq-YdqIg-`(T5hB#0gWI{v^}QJA>N2e$|I>>^*AE
zVtMX<=PB!t%x{Pu>D5!6oiJ_`mYms(xq_O_Ley}z!yje}<`b~t9(Nw;a_Wh8#{HUf
z>!Hf=FUYFh2Y57_4N%|4HQoU}Mh_90Paka0I?<BYDUX_A@7BWJou@`)NDSD1D49f7
z27wz+z603KPQaKK0w14ue(wAZ>h$F)q|<g4fzEu3=wLlgb_bXXfaG?zlJQ)0gq;!1
z+;pfa+6M5%89=RLN?lqlu9tcpro&bnewq!r$!^x5JUd0(a=uG2GJBDv;>BP0WkTd+
z(`$8f&l(_(SuG_>&(e43h0wS7_ggymxQj>#m5Zz*E~~m=;j=F=l{IlFCo~ffz1E>%
zXszcg^(F0`7LQ;qsh@sg*w9L>Lpm<w7)b1R$QpA3HUei_&OhE@?UC5l5a?2b#X<ai
z%;zFkX~6j&MvRP-d*Rx|w>aQr!OPE>-ah4h@~a)xl?EZRjH%)3>^p9qo&elzI$3G;
zRm;(9;Sqs@hQcv`=v!+=1E<A0JtF<PRj6m)6Hm~K3V5d33lF_P55c}{P#KJIuyHB;
zL~9_~1((VFmZ4nt!xcYjqL=5UEX{TthpDOqZe=rQaSpGYm55#9NL1$%N4Lf)QyJWd
z)$^SL&T&QK=?=q8hx`<#rZq3r4>pZZO5wBl76u<w>Gk_n6qwXb5?9+FbFp6#7#)rM
zSi0AHNn}1$oObPu#O@NYlgz<2hfR>v@i6Wn<v&T{r=)A8;GP80BuD4q;1e@*m`7XM
zKN8v`9gT1HBD)K%WXm((^gvKoeaz=ED~90ET&(~`O2-3~@zBzy7`$G>^KsJ5x9XRn
z_uq>=?Szt!|NmJ#;Gj;|!E^sdF%UencY|z6RA|K<b2b|n*cm&SG4t8?OKI}^czn}S
zbOOdLVVAcOP6D-uu^!0%Ou{hi7-j*~7@mUu?-<+x7Aq(YMB-nW^;u!|td3b0XRnwx
zi+wTwn6e39G2Mp<cYuXfYzlw!Ovs*06OvMdJRJ~&Yse7bWnLRQG?v=4OY1&<SScZp
zbhycoQVSmiw3gxP95mnIfA&!$aEF;Vo3q57HuJ%)Qms|9yS+M>T@5#WITrw3&kIUe
z0fmQCfd+f@-Q;#iNV^Tso?);ywn<=N+=>L99}`IniT8{6*t~`IM!!J0p?A4_3`iu{
z$A0?weMsw#;LH1vtUl^x0|HtTpe6Qo2+9Fbj`{b-Q^(JKWv*4J;@Q<%k@$WFEL=Ng
zN!9(Hx#>Zn>!+VinS)pMwF!9O*1}gkWcASRde_|?##wm;t5!xfYF1zELxeOED85v%
z`|JfO?;|`H<uGcU4HfQ>prfh+eeIZIC9*vR#v3c&_PV~k=~HWIaz|gMfV$MzQfJis
z`z+Rei$5*&MGeHv=47r1BG@s@Bx#CRJm$5<Uz`htU&73SQqua>Eewh5K{<k52tfv*
znt8yhMlGYZfEm(pbUz6q5eI<cmq<lBe48b6AEGWdX3j%Qrs^P~Q7Zxp)KDkd-4}01
z$~Fb=sa|gJt8S=C{Tyw&RKiMUK81MeNM>&`7`HV@$J57q6jTz1lNH8tV%>x0nXWt+
z<A#vTn-yTopa2!H4^g5f!tD_B(hpINs6&f7$HFzD4azm?dlekFvRLUrW5A|0i;0r8
zb<Sz^#T7TuY<K&~)X*nKxOyWDgCzmmf?ys%KNq;5qg+W58pw)z(tIb$D`efJu80&#
z4wyt?<*6n#<TmoHeMnN_ob(>E4ajPD&<nQPa{G{EbHI8Za$5pLIgH6b#lY81RqXd6
zcZz2yf?E_OWCvy)R8AN%D~CKl0QA#WfGnAYyg;2stp{Xs!dDc;s9{y4wS5R38oZVS
z!I9$M2s&Jm=#L2#-BbfsJcbehe#?F{4EQOtDd3>nBboLgWjh$sjZK;@=_t891I7gf
zQWd$&!`YEhUnEJM19D*$^9nIOv7Lm4JELRLGN1S!wbPe5%_|`P;bm#ViA|Vs-yEVG
zMlha%)r9Rs>_ElrR=e;O?~*g)T0;ECBh3`ymk`m2iu{zk<xGAuyGxlKhmvesF5E3?
zj*nBa)$`(NG=#!Jc?jOpuO!h%T*w7HeY>)b)cDGS>NL6!MXpOp=jj9k6ww1A3bO{G
zvae2Ub|V;^6UUTRoyg%TAqH}hwtJevX9iu}R@}&U@^U%LOU@U4J_l8YuzoPvhY*G^
zOz;KIA@~|0i!~kPA`OU&JybgzU^8z>kjmr=CtJG5iJ%<2;#c2lYjw@3Wn%<Yr11zS
zwHLu9+*pxeE!KvFH96%S&&cC0Myp<eBy?b2X9CVM5wK0g#mpVR9FHY)jDVe7_YCeS
z)jW1zm6na!6F=pA_Tw3c6EBW-27^?)l=2EGb1OIyfaCgj#*g4bbVuwm#`r{!D!<+V
zyX0)9k;g%9C<!dCA-RBxR`z=B?eT>1JCkE$JXa5C%^`9t|FF&Q?{kNUtSb`44bXM1
zKy{{j@C2ku6jpv*ajceb6x^BI^X$;#=cP!4>TFjqbRUw*3ibt?<px{N7q2VKB0hkU
zFSS2IC85QJ<DrzCW0+oOlA>Yx`Q)UK_7G4GWk~_%!Ra>vCo?2~J=U{WLB|}ojyQyx
zZ*7AEhv7gG{Lc>o6jeZi*hhU>jRK&seF#*GtYkiyNudRGL&6z2^-L&ST@5_nY_4{k
ztVw*Hf@_rTw$@wKM2L!}_0uuOuKRlqHVv9hzK*<n0rZ0*sM!}V9yM6-Z9?Z0IeoQJ
zRkfD+o)^!*%cwUCC<I|wV(*#B8V^InX+NW!*LF`JL1p)?btg6E9@Vn})Tf23!$Bz_
z5w&m`lao0fMAe4H!&kUGLD792^U{L`^>Otz3i}m7I+n${xN!xv9HQWIV?zXIsSw&!
zfR(DAhN1+4!gyA6qdE`uB=`oF9Pkb6s0C3>>OO=A^>L;Qg#`@|>3lIMl*mdB^@Wjk
zK+X4(92o<2Q#C2xR7?S=ddI252F+LIz(tK7hoyv`#cb+gY4%X4uAE>U@Rm(5pFmqQ
z)=AZ7ZAI@vZlM-Z%4oN#;KbGhz^mOZM5RDbs=U|BgI-VZr8B;ebQB+WA*`r;^L9d;
z`C{Y>$jRvj2av)6nyq;0HXy&Qve84P@5+6%i?|celwXfc&Tpc74I-89EyIpAoWH#w
zqE~7(dRUW(iPzMKbsjN<dbG*iAc?aU(+WEk!N=z`t3AhU%8+%N0WvF#8AIT%$K(sB
z^)*|#nmx94XqhRzh<=2c&CVoT$It_gxNkod-bgcvY%Axlsy|YXI>h@rK;CyL3MWTX
zbkNF_*A0qIX8?&gN!1T0E4#kVQ+^%8<Uvqy#5>3KVoVqYrXq~LDdOqIjF%ifb6+%;
zV6x{3OF?P#y3G#zEL&FbnFS=>P*s4SN>P#))~&ae%#zW}{6%ww6WA@r+0G3W*lUHI
zk`wS<W+&P%lxyGkVd=5XS%+G@iAK6fKu=>VlV_9WNpIkA9}dyilxmCrDK4><x<0=p
z=fSqT6DIdSM^Yr-Wiy7#cx!I)dvdcNd)(*ymJO+1hZLu8sq>KTHl$LIY!KEAh>oj_
zg)T}d*^^~41$;4@cc^7^pVhg8LJj8CVV=}(6ye&>KnQ)qc^6;&D<wEIAm8rRsoHuv
z$@=3-(KOApid~!N+$~J#K7^I50SaVxGl226v`&XU>5q&Z=ZC$mS2<_^@a_(YP#&Mf
zIO=tK@B&i8Twzs5ud<C_U~~QMx~kEuV?^^cZ30|EG^hg#RCb_phIxx`TS>5%ud&km
z-1=+^G(C1x5>bNTX<`Cw=dcfdG@QaO-;&x4ieOn}%CL-pgqqeLPwshuXu$yBG;hW{
zs%y+B#7!~a<QT8~TIODMtqjg4PCH0;CUAWaR2dN&X}E-oQ8n6!bf2F8%(B`*1nmSm
zVcfbbh8WXsW_GT$F;}51v8|EAF%NakdmmDmOc^2<@8(f*F<;*UtyIicst?WH={;_0
z4;au2Hr-PM^o&MSgl!*$sJ^SL&^_v)4E6#-;9!RqnrH{v_4<8)$_ct}w;$ip2<Te@
z7Yxz*Su|-2K{@2IzzZSkf#1Ca)zxl+!lr?9RpJlYKgc2Ui2f8mYIak|-~M|jT-R6N
z@`O(L`^JVBZmVwnU1!Vg7iHNcBl=QT$M$Z3cDFLJ!|l3eU1G3YZtr%QZ#>Gc{`%<C
zz#kGBF59--E5z0S_H8TT)bIoLDd+XEw%OTYrDMz>qFH>T0lg-n0B_wCULVYGDJ~`G
zU1bs9rf8_z8hBGd;9x6Cs$FoFM;)V{jG~XYwR3Axv2!97^24^7vCj}uh~%ccMDhY~
zeHUi2W|CV~v1m(#B-vKLQnYT34m=H@Z~|hjX;;-WU*ubsRy81><or^-Ks&cKZa_U+
zK<J)Rh$ypE9}*%IL=AIA-l(LD;(McM<oRfbh;GBc0I@dIFG$fPq~NgLB8eL5XDjoe
z?anj8r>S<-^{QRaLLABftP2*Al7!AUH_7nE_7#L(t<#5J%nto}e(R@L=YVQf-iO>f
zOv!CJ+Cs|h5{s7AuTBd?dwz^-$}>+R*|s6prI#|dmZhnoF2r=|5!~GJE8*>kCPMZF
zMtxg@H2kFu#>=%l<R0W_Acm?zF{U<xw5OjLf|4eU3|Fh29hN%b%|9@$`?2-S$v(NQ
zwH4MqmPzU;>HBNE2j`591%NO}xpM82LsRg}j*=jz+jDur5AcXx?kWUn4_uS4$We|U
zKQyt%G{fJ`KU%r_@b-p~%X5bJUEcL#8(A!o&6x6;=6wjw&Aj}C**LW$Bk~UFgXecn
zMSX98Z&!oQ$epYR<jtxgj0x3=un023g>lkya@^6a^-PkGuNZ`Vk_kILJZdPEawkQY
zju%3;cKeC}P;=U26)br<Gt4dWiy}r|G^A3L_Dn@VTOsmXFe||^w!k`hnb*f(MmmT8
z5jZoxQgUS-`jM&u)c#H0sR{I6>4oFu<sAv?(Y`l)Gh>W65^VR*?eb^{$>#5k(;d9;
zpkB}a&lqSx<99x&Ucmpzb9@l;#`;!crHxSu-R;=qnPr<o4M_0VK7>gR2U0cWB{*m`
z!u?T4z!q#h7jS~*58m8d;xhNL>JYYn84F<vXb)`XfS9KR-vwX=!FSyx(de?X)Mkw1
zQ>~tQ0q;$;UUsM&$Y>)@fpK@r;VoDXs-+$lj;L@?CZOAyOULe+`+IZqvP#{RpPkE~
zuL^>KSIY>t0Tk{K13P-hN$dC?x7-cdOR>poX`4#owBP%HyFUx{_IntTv1Extz-{l`
zB8<7g@NkuLk&RUkrzd}#+ePk8^gXam2u|%RS?nE9p3_<|J=N!wSUuZ+Jae?Z?k_Hr
zO<-L?=LFEVjgF607YBLDstP}gE6B_Xiu?1z%H|s*;JVKblFdQj6-bu{P>U+Q6b9^H
zIqgG3Vg{+EBk0eHmF`2ZbEpMP%$4Ph);|g6b^J>(ub;-sqr+hLRK{tL-}->VPE+>j
z{Y-4;x}J(L$Y>YAgc4eH*ypO3r#(wr*<*8hp^@eHs2NE94q+e&_t-GpT56bP^_wK>
zypkh1auu_TLv5a+f>D=!$j3)G^xk1h)IwesOIYy|j5r4}daH?D(`dVasMi5~G|x&Q
zsB@W|sINr<-PBfd3`upF#7&92NJ(kpI{uN`0=hB%F1ss0`_USpb}k^{kX-jSazP+9
z6m>`WTN5P_$$89QHc;ev%`gLGCtAE{{MF!>#TtL9`)}LU*;hC+7u8|CeXBU3uD#o2
zGy99ir8<v5i()|mC515aeMnadaDiGsy$@Nug2nf(O5(dw0>OdA4vZ6CFNkto$ME4t
z;fEB7?R4r<1W0YnXZuKNAT!8gRTp#vpG0z!FF=nYa*?Wst#mfOND!An)V8h$;r*jZ
zgW<UPqjIViU-1%7nzs_J!<8uQ;D_c1Fq5tKPP>dhR&Pi<Y1tSpqkXo>a~C^@SgDzN
zPvxcdBWa1uJ$++mORAhi>XH=&&4~44LmV^<Fv1VaI4w2md+|O52|lGwRBJuliIpUl
zhNkT8Lzc0d(7Yux(&!84Hx}Bbs_LS|qBSlu#nRBP%iRlwXJ7z-$Da{_s~e{K5DbbE
zXaM*lRw038l@Zk14CpF>GSF`7oC2BS%Iw+))RP)PCgSzHH&VqO99|!{<sAZ2k@xc+
z&c=Ev-=OcyB<RkJ_2CUZG@{z2=RBCF7R*v_+-1N2esUtUV3rvfkNq{6nUVtr6E&Q<
z0yD!r9EZ=A+O6j$-DJ9h;F;KJMjQN#3BX8UW~(#Tr2s}M1{`6HDr7eHx3NV~9)Qm#
z!Pott!?%xuLB=tOox7N$pm+S6bN||sP^Er51H?W3T*BU=Nfa2Ioag#r$I824_mZ_*
z0~8S%sEPX!r#RIr>4!}h!vpsrwo~wL?7Y-_vKQl;stP_e$XIE%67HzOhTS9gAt&_V
z)JD)B?ANd<>0|mVMwm_(dKp<6^^lZRss!mF3O^yk3or1C*CdNke{tCjB_^mzVZj9}
z^XJ5|HQy{CZnQtA9s(}RF(VgSEWbFf>#kfsr-ZVgI{6k?VMaZEn5U%Sb?E!c;iIAO
zJdeNWTl$0QK?3<HEE2H|c?Wu{5?1u96cCwlL@{bJ$okJf$o!N{;ah4K1CHKwhfVlJ
z!nQ|TNKWYQj@w{xJ!x?k)V3^83nS>ov`~g`)YGUg8IX!UT!U}wVGXS??b5rVHd5d^
zpgK>D3WQK^uVUu#(mS0S=p`{wgIGTW>ITi*Ao?U-cKtvOmhc2K1M9`S$)p7*dukTy
ziy8<ejZz2_q@K<8dG9?E#!bMPBDkF2W=vV($7Z3LIhwjC<#V9-)w$iAu%l+MtPf6$
zQQ-K>mIobBe#vPq_PMM2UcMZCW^V(QdSsROOD7mzujDMuxu@XbTGx?wr_v0*EVjYU
zp!)G}DBNY?XYLUC93>AaB@NzUSkq}@z&Q3fr<ZE3`+bpgXK_pvboUzpL1oyO>*E8u
zuMpZYSzYBUOEs4id2g=W;^BD9H0#>@p$kk1+q#R|l0ns>Qc$v(%2aadA0v)=g_vO-
z>LS&YJVYIgHl9n3y}|^m!O$<O!|G9QQRK9>FwF7@Y7>gia+zQIqWEY}3Xq_OFMv}u
zt26{$J)^)lsKKscu>oo053Z3kAcP@D{T%7CnPKxj%-{Iy4f^-sSejg$+rBXpoUpr>
zfZD4^$N{^1FiHZISGpPH(V4jYb(yjfhbcwksZ&k#YRZ)7<OeR?(yIBHl3%rRPQ5q=
zW}fW<o@i8yFp<4At+K5DR79SKY-xW{Ns%QyRF;NvY2aqYQDVd^5a>sp)r1E{H#U2&
zPf59b-O}WN)~FvsdZC(K1c#^`HIGalZm+GDBv@yF&YcB?50o*1vpem~a8qJrgbJ7A
z{TeRLbCjZQH@<3qeYdIxWU+$IN|=7D>^`JgGvcJO?qW^s%#X)B=qc9YsE?Q;6t)CI
zqCt1e0K$mYy&lWC?*;fT_$m9`n|f|%GpC%*zrpg5T>-!GK-5m{LlR@ar3_t*+Pr`m
z%v>SzwiDnVmZCXOc9#6Cr0-Sa(STH{rldKw=^HqEdr-QVig|J(3`FV|AuNbeNk||R
zjC8<yFstCy5EGc2*<&!+hve{LoKdqN?^t)Df>|nPQ>rOYvuA_q>I8PEEuhA;(19K|
z07Nn5H25E$dI}^&OM?T<H#qn?7=r^0{B9UP^qG@6yr~9ycz+5CwA;Bv?4|@FVWt|4
z8?E8z+h@jII&CETWxF3#x?TT%(Jj~Uw)SzMm$89`S*%{*l5+v%ckE!o+A<NdDU5i%
zMVukCBesKxsFfW;r}3Z)!cCy`>yD(5h+PIF`y|62TPVXEQU=v^mS{kE1WLjE%zmal
z!P)~qf-lXg2I*JOl$Diyi#`{~c}6{p_4@_Tf<2|wP=ckY;fSHm)|uU7!Ezv7?)IQ?
zQ~{_vAo;Eb=2;5q)^aV1N}8iK__EwFei*M{SV2!Npm(+WJ$J@<#}PZ)y>8w|q-Y2h
zOrECQT>$U<VKCQwKyptnAnjvu<MTY}?Pm&f!RYTvEA;(oq*Urr(5M`uBh0kI8Bs0o
zfw*9w32ubnt75(9aU7v^A`tDZo)GLF0@;CtQc4;m1gdja^8>;@vfCDY$f{#cNU0dP
z4w2hEMYN?JO(53e@;0OFaoa@u>bY;-*%H@uw?%H>?x$Zpn^}o%JyMBY6(v4TJG=d3
zL#21-dKDpZwH^F)4FU!Gc#HK=uxLYwE@}4~7w^|QDMD7b6dFj<10M7vAJmu3x#^04
z=f#AHw(3tkm0g}r3o%=wO*l#t$lDTYl&&{A;yq$g%{JQIyH%T_9KMZeE=NT{Vgnf<
z<eE`g0X+@*zBl9D_ppY$#NC`j&obVlr&)nhDDn%$BlvRT8ugiKaKWAr5;XYiYSO`;
z^`i~2TnjKL_u#Uk3EM+;)pLhJ8u#ePVno(v_*g=#8XNKIQMb}t-C!?a9ue^eFH)0K
zk7TjB5);Bpg57W-2A79t_0lC4czJ}M+VRJ;&fZ5cP`xqtq{rrMDCty*r6!({S#PIP
zj^&LLPk!7LJDIxp5**)C=84pSnF>}g!sgsZuk#ux>SI5X`EKdPn~haV`tMf&FcoHZ
ze(x7lv%U}dP=L2S@PK&v!JW$L&whdA$L~5aTlpR)Kl~t=Lb`=pi=T!N%`&icj?tUl
zG4|H&WI{s-3o{Z<b^<}6O*SG3Y)y5B3(mA#$Ruxjvcswpse^}pX8d`9AxM~;!#WCg
z6DVQ3FMq3P2<WF@?p)OYpVbM#m!Q=`3a^RWa#GQn<BUrh?U*>5c#)pvYb=zW-!YAR
zWwgqJE!Y)^^A3^eDFPy`8l=B<FtYs^8_#(chN3}S23V;_(}_c!Y2ghDYJwp&BEml7
zO$#loNM>XjvJ;Fg!0D-<Yls1<H{N@&1+;dzOirFyU*(QfNC{P%FU(@uhivmuGElge
zNScLl=+v7k>>7vl(x-E%!6VH)f}so~w*Os85#5ajBn!wfO}~J47c|iWZiqKkgaR_h
z+V&k{=c+&saGmw=xQ5R*m%d1grZB^eSORM)Bp9RubReToZYT$dt3H<6g{X(_g4BQx
zphO_W$@G-$(qKbS<Y#mfQdJ6IN~~vZte*9v&qEgV|8cQ8tQE6)9-nY@Iq<yOBg68H
z{+fape4-#ncSgztD>VShM9VfLd#(Ac%Amy)TAp!BrR#6E<tU2S4pj9#nH_zZQgEw5
z=k;X%qdBLXFQO-oxQF)AHLxjdanQ_XzD*!=JZeUs1ibTrD|o@OI#|@^Feh=!+O9LO
zVK$6r;^D)-yby1^^mq==Jlv;x*cZ<0T)iD#9}_9r+5d|RkrmHMX2;Ccv`*=l8aJDh
zNQLl4rL+1fD63gh*imFNNU+Jag;yMcX-gM9Kl)K}j345okh$^nICYo?s#zUAn{5~z
zJT!sBw%XTvu=I#d6pzsR0$=G&mup@(LwAYA&I*A7o6r6G<r+T+Oz~~;@Id$E_CVg&
zELFJ{24+geB2OB4lmSI!H5OwZ{KqHc74K`))rfev2j>u5sYk1bnH>mdgHFj}ooVqG
z8*ACO)!VSsiyh+-iVNi>>T4)j34An%QmIiEvv7~KU0IJP_%=i8u6bcF4`FrfF_JAn
z;p(u-De9}-@$uf`e(l6FZ^`Xo!ewXcDgvLmvN4BXTre!DtIoS4!nv^9d{+^4?ECsw
zHo$z1p$~{{-bxTG#fTv}q)@N`I9r^__dtC|`lyRJ$+X1GPbsDG^NK+HaXX`#)&+^}
zAAjlj+_X^{5*sD6;w6=a57*WYl}Y01U;9TrL+0L~Unhbd4D2|n3H}OYs74R+PHA=U
z+k{}KF&5<Wk<_Cu#1U)0J>KDCb5sijG@>3+*TnfQRiDtC=&aZ_x_!TG7WLo;Ex5)L
zQ4_Y?K?dW>vzj%yW#1^@_r|NdKI>A!wH#sKB?F`+2<%rzRP+cEc;Aeyr(JnKYU@t6
zX}o=N{8d#-G07jDun%$d0L086qnvZ38uLG@T%0>)gv2t$rRw+6H3sa^RzWUogSICu
zIanSTz`5KcUn*?Jd2n1U=HE7t)lP{#G!oizl!xrHU~oJ*7zn}XOYG?x6CKh$94kI=
z(BYb_9J9L6t>9J|7(pKuk6~WRkU%l%<4x)CZcMTt;ybg-46aa~P}Ks$)>dW+Wn`^;
zjZCS}d!N9SAUg?w_|EE<^WNnikJ(IG{SySYml<yDxKHV0ntmw@mv3h<yMu@lLD^Ff
z?Fe=ups>hdSCD^TRC`rr^V8I)PoMnuA@A=$D9s$Jp&r>oFQ>&?pxgB?`NT=dv!waP
zxTjg6n9`rQ(T`)y8yIJ;xp0_pR~1#<GI!<bnNR2WZ(Kd6$St<yMC*(7db`Dqo;TQD
z?uX;JDMbw-1#V;CY)*IX4HScOI;MD|dz%anlIt`%ka(4<Mu7G-Vxr181TGH}KWTYa
zsWau$Zrr9>Q|AX`8sHjA517s(9tL`ZtN5Z>nV0HZ>P}}UB%Zo{?vjS$A|ix*8feGC
z;+lA=GXw{;U{&tgI(2|oCT{xI+sw#g)?ge4scf+Xvlpbt{OUe2uf@(yt!+o<gJW^N
z=_oKitJS>fe~H+h`lZZaL~X!XP)A6ZH<th5BrgQeqB4R?hZxiu6lh%j;7lF@*ELYN
zN|b1y*_t3lmx8y)O^I$24MqZ89_WDUEsU=&x;?hfb)XX{USQ1&TqCnz*mfqOSC(;^
z$1MUn1|@EleVbzI(3uk+X4Qr4S>700-Lm9a`GK>Hxz*n(V#HK@n)?KH1Gndk{LI_A
zGDA!^j?fk~U|AaK@-IM~!OOGw%i8vrt(b30+k_A+W)Y8(a6oFg)kbQ_wy7e;AoV1y
z;o+VV>;`B;A!IZ-%%H;3tW`o~UX(|%P)}B765=FQD@H03@4lyXVbQTsVzr>?6FOHJ
z3G*FFE{H`nsUi1LQ-#!xFwfXcS6=cf08~=w!hlD!jhahqA5!~)oJ)u;!_Vy=YT?lP
zD5c%Hx$O(UZb9#DP%E~_C-?RbjMW+gVERfBB^RU{xid^n41mM(9=ma5yN%dwBWAVF
znTd6W9fY~*<_YLFk0(eaB3GwDR9eB4trI(>?HV6@rz|6+gCaxm7xSiIFjFDSk?aG+
z6>1&oixeRYDoCnT^PV!CDgTbqL)pOqQE+F2qt0vM;rGVvg3?BU$}x`*D{IbC5Z9<e
zp-NyV&v6AKy@Dl}a+#R^FZSL%9P0P)8y-m_nh=thic(3EEm@|LEs6522{FkQ6H;Lc
z^N}^{L`9LNQYbW)t!!golE%J_m>EOa%~-}TOV8Q&_q(6_cU|}OT=(%j_j4T2a~${e
zM@N{M&wQ5iJm2s4>-~DYPP8=UjizSZ(m8S($*!<HOm|dsResraZ8`l9d+Ox{-OS+1
z&iNOKo>o^<<fs0+cO21{8wZVD?S-c2(0qyhrF^b2$ZMsTNPdI{v~<3oA@>E`oeIu)
zU_SMS*=xu0Hlt)xrR#%Bgt`7%s16Pv!P?WI=;DxNu@JZ16aGht$EU&(J$9n5pk5VS
z^#8lE#s2ro7X7c4Ee?n40DO?Mnu*m$oc7US=yccYo9=nyo97!2)jal84KgWNbLDR3
zY4)4Iab%nVlXU;pr9h{B>rUUdf(dkl=6e2L-i!n6tOEPr{ohq=ep9(x#KDV;&OI->
zHJ;F{41E>eTMM>p)64jhYyNzJ?7}=xwxa%Sb6mB^Wa*k7Avo`B*_~^%TZ@DqPl@2b
z00O#30BxSEE&*7wayaspKj#`bhPtE=B(a=xpz*K391_Iv^#wrvDF}jqkku66rf860
zli-&GD$L+`BZP$X^j*)FZ<-C%Wt>%Z-aK|M117+e=-DQUPzMz(Zd4%l(DADGUsD8H
zR-XFmr_SPGbOWl*E5|f2>^3!M+mUaOkTa6E|2OPDw4^?b=4@UP;>$qw=AnAv%SZqg
zB-ND~sk1_ANA!9(XTIb_kk403Jq>-L`eVq#9{nBu-#@xSpPrfLs5XPZCCkCE3o%t*
zr+NxBpbzgm+5=?{xIe0Cov(cXA0SI}2-Zv-Uhi$eNMr6KlV%m5Rrmhv2HU<bVTz30
zaNdb`l*#hoxS2!q?+NaQE%rJLb%--<IR{cy*y^k_F~p>t^Je5YU422D4S#M01QgLi
zuR+v`>j}H`8y1)+ymsJ!qJdp!kef`|o8~RIug4>+1qQ!~sjCd&-VWW&xs`8|gBd}7
zD6Wp8)00#m39h|HHKCa(1f{}-**+dyH@|75J=bZtrpu?hA*do5mWY^p@T31HzFsrR
zL<dmWO`g&vHvoc<d`fs%!liG|&p(gGyEejbOA`>&4iB1Y4fd)R(NYIpJa=v!QFt)M
zKVKbFQXbFlQ@nrKN}@&8TuD6T0npmWGeNeQ%g`X=EA{z>G;TxJScU1WBiUC~zgzho
zB@Xk&o4NX?-7_qF(NMD-a@Y>}2IlfFmo1#c(r<!q;OL(hkozi4J#t;#f*<NUxqi2g
z7{v>CiT&AST_|cFX&<3oGFIkWBZH1ZGcQ`XAStP|f@IP}WHw^HBrF;%YHptXkuLjx
ze4|cw@_(uob%;*fI-F#!RhQSQ`>UT&EMH$`^3y2<FX!E9qd=MHT96D{BlGUflcu8x
zYhuKo7>02}KFhvQrTrno?;!^dT1{st7Yn`LUos1EJ9e4!RbTbvvl!)Gze*>zKAESD
zk65rmy17`jPN-9^!T#lf{{FcDA%~qm&eSMCZD+sre^bAImomLH!ptA_OOq#e?<r1(
zW0`;CH!R0`<r03*PLzUXb`m5>6aGMU&CLGSxyc*wmOJ*oD$mR<*X&lvo&9yDn;o8T
z#i#ytmem9U%~s~$pR(mx|2=;~tT=D!lD807S$i`L{32)fiSgPYy?&&4Xo#0oDzq6R
zWqz#aI1QzLN3JmYtK66~#S*UAw!a+)oxiwZH=M)So9r%Fy1otfSOYSI)<J%R1ccCX
zf<f`T)a9uRBO|M9s>#G2<tPTS5cdQ36<MMd3;FoR%B=}fCo%6%8C99uXe`$56jqh^
z<!K9O%ccKK3-*8K|G$Fer6{$kT^}s7&wp!vAzWp3xxaz9swUVyE4cSJU3i@>MTXuJ
zEGdjFvvIh<-nw{4^4ZmjA;k$@^`m};9u9>+ohNrcixK@y{HfBQe*~&)?IDHirOtXI
z@BfC8-O!!=GjqYAo0(RvNY=rGhf6Qur~dtu{9pO~J+cugU#UbZwDYF?x_@wKs4L=x
z#DcGH?#uUeBsh{rT&RRfLOq<*1*&Po+kV6DObD~n>r9Sy7~R3>e!ut5Y3jkrTGiCy
zGLfs=kxsp<08LqM8}yI{Iw=W;T7tpYd1+`oh1f}f{i|fvy?alk=D|d19mS{k<Lcx$
z{FgW-(;G}=!tLItJDOTN#tt8Cye$AFR*F7YqKQ_>B^mU$->}3(pfZkrT8P&>17X^E
zUV;!ng2}qgKS{mtF=C<X!gGoTe8>Hp3brQ?G~P*{X(w>JKy%bR75d~l_6!7pyT1Qn
zti5q8x0wjLR@Uu6#pc?Ueu)~j97`9mtqLrM-`AO!nf&Op-T6mw>CLd#R|a62ww?99
z!C*(R@*L0z-y1l$Rn#nsCzDhr_-Z0!__X?LffX_^?3(Z-<zGqa|Mm0#+06u7qW{;=
z|3AJ5JIg<oR^vU|D^X(nLA%3zpIn~GT?_uY0uoaRSy!C*LW0UMU&616I6$;xP&Dp3
zRO_o#VGJ1O^)eW5k&nI?XJjONUgP&1rJ*L<abOD8`U8gKvrtU{GL%pSai?LNd4X80
zzhwXHg_jmaT4V*&M~_Y6l~qYN7F4JI8tORE2Ig%yPn`+Ch(E0?=NF)mhL-jasp7Eq
z->@iuOHevqNEZm_a{lToH;zuM_AN;&seIDJd3oXJQj6@(c-SgJApYhL#esoX<9q`0
zgeYPQVL%_q6X+KFm~4QCJCdA9w0CGAPS65tOoAqV-)UYnKicnV5ua!xXKR!op*6jc
ztny4m$VnBXf{A&=b~L3(AU^3z3-ej5Cx{mmF1JlPNZg#3cME%~5+QYB{exR*>1h)4
zylwK@IKvq(YjZ`R>`stJs@7~GShy#o<9WJ&mLw4$PCcz{qUpo+P>b9mxUOHe+o2Jz
zQ}71TMWhErq1IAM&{62~OkPg^g<^5drKc^6g{|Mkn%D2xaW&ldANComA?#v)HBXGI
z(R6z#f&w`5Sx3rkGLu;Dp#@{C<hy~RB<(-6@mdEH9c<TRQ<g?-?5r?i|G}#s5B+`H
zD;^$4l<S}tXa7{~=G8{t`X}i;=ir<AehYIBPfeb*huLsVz)S8W$XQDrYJ9myW!g>I
zI9-iBMkMvpRbeeMfZ_^nUqnRkd|&cTG28L?jSSwJzMGWqjJr}^c0GT;0G;Ib&mN-e
zYAhEN`?Z$U`+<OV2a&gjUZJtU`M}1=Y!^yps`RB~OILt@?FMf-r~9@<b4b2Pk0^NG
z+!BY}b`HNttcGaG%9}oP1JaZeRl3=K`_!?RH^#+9;rd6+>xE68^^XNN%_knCHWTAA
znv;!8Webi~=e2ZOcqmxL^la3Nu$suO3PH`xgF7$HdCJx19kbnly+}SqIBosZuO3Y)
zPW68f@gsIe01zDSkzHK|F>IP&A=oA8MN3znc)_Xewuoq|Kf}Fjlj~IMPx^MDwoUo4
z4T5PmKSUZ;13VZ0CUa)e`y}10*`hXeV05A~Cvs!MHa9>Ui0DKE-9xPb@;X@w*PE_<
zz;@e^x80$q+l@qRqO(=gYKPNw*~=g-p%tK73nYfTUGUWf!7Q1NEL>_G@0p_jw50fj
z_i?-E&qA?pU}2eNxLe?O;L*@tAf%B#@I=&ed6q{`!sbl5tUX<&1i#MKs?#+0kBkid
zwij@~vR2pUWhpiD@51?#`7=n4iaalcuoV!fmY~*Y2KhgG8CWIAJAkH;@e2RO!@q?E
zZGWKq)WT81n&9DT|59fGAW0~51)r_Jbj%QV2y~#<i=f#AS@9->*TnfrVA}r(mHz|v
z@S;O5V2LweA=t~HkpO96Onr<nG{m0av~$&YE(xZc*L4&Ln_V3Dv=_GbW>?JZxpy*H
z=m88IXelC|1A05%Km;DieqGH{j%8e18VbAHQ^2;$X}EeZ9RBRX!J9WNbXSbQeY}is
zBq|T_v^bbi<krhZ*L2zx-rHGXga6utVySEu>OY50SkvdpPBRGcMi($S-j4b))Y)|F
zty?npE7TRigm$els@2*0^=kJ1@X6VNV$F9t2mEZxALH-rm=!I7d3OTG0Msuvo?L>!
z3K>P3UO@9xN#@gw4+WddfX0GQhhARQ6to^kh=9j*bau3y|GrXeN2h9MX7zb93ApG-
zBK6-cpTbbBK5}fnQWWstppO0}`Qr*Vu_&{rLDl#x(-VCy77GRW^U=6XJ)>AarhobU
z{YKxl+p3H6rR5p>gjsljcmj9p1VO~_U?oJ$xA%H6CA}KVzO$+^t}3EF)9~C7l>tsf
zL!jk@kR*BY<xj=ffQAfwKvP)%>kIW`rN%d3RRh9JL@hS^447aZaXMB%&3cw4&-Phg
z!;g3*AHR&R>HmqG5u&S;<OO{$u+p-4^BBc$jfRiOyH6<cnprhGphMp|-SQ^q)r=YX
zM0_A@E2@Q50=gWS8d8x>Pq}2JKhQbN?i?H}pDp#f;jUJuB!Ua#0LAqnE5Av`=<c0v
z9rY{CZBE-RTjAN6!F$7zxAqZ#Rgq{f3Cd}|$R^r7u-~xfAw&M-8B6F`0`ua0;t2e6
zoUV2bC!*!^qS@G)uG#sLvK!ZgR}RImHnT1GWub~()TH#GC&bs-_}&wryZoLg9NSUO
zh^Rz$rur=mc?vVe$!9u*^N6kf6eoeWD{F|R*+i7VYAnr5>Q-mo8Xx+`K7!CYA|#o`
zFosTZwb>&4cSMqpf0Sna^yqiQ*1;dGvNLk~&g<GVm>*Y7d9d%F37<OP_FSK*NTUad
zG?$N~(;}q=r^MGmdSwpk>P6vLrQ@QyOp6&nDiYD*%(MC#7SWa4GrVF1?QdNUqU_Dw
zc50ovTN?b#4XlTSx9wF%R8H&?Vd~3HK3^UOD}ns13XL>aR6GBkQ+CeMVWG5iAGNwX
z;Cl~(aa=jq45t}-6+Fv^&{M3lFN1w)S%=!{DJq0v<8K~wH7g$1;a@9lNVHdIWyk4R
zB=ZRsjM3+=Ltf(`yhUCkHWDKNZ^!?>K(@zMqtZKP5Sw6nk-zBFbbbc!>Dl~Kec6Hy
z5Xee-?rE%n?!1_fnEGD1sm6gjeea>@X3!a$dUWng#<f}c9go0dklGK>V1^Hm!(Wkf
z0c~HCqu!KB3&WfPjb5cg#u&Sz2Mv_Qk?oyNZk)^mdW7B8+AP4;8)7ua4}8MZpJ^rY
ztj~eiE%gFUJ+00(tfMmR$;kql%J|@Zm0mrlO~Xq0lz-k2^;!I&mq48Ul4sU;&`6)9
zZi}y{lnmWx^lAv4j*QX&im+a1wJscK?xWatiCE}SmbL!D*EdSGC*kMLMquY$t*!>9
zsVP408QT&P>3}=jo6`}hp6{bq>lotXus8AOYVd(n?c>TG`ZM>KsUgkfkmx0yd9@7k
zcw*Z3TitzC7yuv^+vttdeccn$HM~B+yh_&M1UkvFj3>Lu@JLUX*ZI5z*uQtA(7K=I
zK0P!%m3eo5I+;W4M@Eh8t7Mhcqc>wH5l!jk(T{><4ytV2mVCo_pK#%;c+M|2$XX&E
zVddGksy{Q?Suz|#<LTGBRoTS!>%*0Gclkmu;2-mdb<`bDtwXgQr|D|xP4Ds@u&jjK
zx+%_zVY1B)B$G+SpZoEYoVW}_es(SuIDjPNWIXQ}2gDGOn*ikB^yF%9zYS{xp;Do`
zTSlRDg);P6cP*N$hNCD;@bilm_I8#{-W~t&8)nsY=!t@TZ_H730D~?mZAG;an0EdY
zR3ijVv(8z&p;{`^_PKyzC-=TQ@%n^}s{u@kyj5sfNp`i03xZg!nb95Y@lea&Vyg1{
zdpXz;+wQ%jjH$j2(b-brR;nLq68$RNGf=IoXV-Fmp*;Oi7AAQ8!Oqf;d3!nIQ~KsF
zB3C;^r@~};Tpu*^+<fv{QpXo%%`c@2<aFU9x!nrHa~unBQCQAf+VDvVVyn|2Hr;$g
zf}6d2g2+AOgn04-J`}J174kw-tm}Y2`JO9FGy45TM9|Nst47maHG2??Q|i5W|Aa}t
zvt)T0=-6e$fs;=Q3q76+-~N$jNgmxNY#z_M#w7D(yBQZhGV&#reUbIMM=+g!-0Nea
zels_MmwD<la7wp8+?GxpWSlMO)X5uV$tUNOBzyb5yy9BqdA$DHuj;#iL5Jx?-y6{e
zkC+g8n-x9QFNH~&$ZOuMq2;K1ZgM5b=WiE}f>0mk#WxaV{e8gCg6<J(XMPi9M4O{^
zh|Z16(x6Z1S%{9L5Nj5uDE5%o^!B^>S>r=NinD9<*2!X83R+h%fM8xIRHy5e1k=$m
zrG0{({e_p_G-cn=r<|hEa)j1@FahpA6VbAueT0FS=ZVRbc20%z{?rEl;W5gbrQiwq
z^fD9)F;VkzCbInn4Az@Bx%Wen&n|CtQxC*nxH|9xPI;#u#ZhNv))SPCgmSS(_4>O*
z6pKZ&)igH+o3F?YgQoOfy@42GxGi9@+I>mx`aZm=O{cKroW*$?TrjQ&V1oW|tlVgr
z4BffdWN#1Sx{_l+d5c@bx-l(-7x2!n>P%E4mH8Jwc|^-|Fn`Ef<V#b!Oe3C$?QwyF
z8~8FbkQp|1%c{_{;QF*&Ueazt-Vfk~R~^rRvjO%SQHVv2vfsGU2}(n1(UoR5PI=#X
zprQ8Q+4^&-C)E0?*OE_V=%(tS%oQ%$Kd{JqA$rd2mf#p`B$hc;w=%MeCHHc3<{O75
z_q02i!uOuR?T;&o$gV1Jf$5vOZUz4&!-nIOZR&a@xl`}uJGCbZ;aF_lhZ>XR#S3KK
z)(}SH&zI~*md7_A<anZ46qE*H<pyFsfpbuvmvtLruP+6a$ZO&mKENkvA4Ya}2j&~A
zkEUh$=Uw}L3Qf`5c>dZ#K()=co1!iTL`85hZ?W|jtd&`*H`vJI)d}ai0^x8+TmJyD
z7LS8tcA|h|z&u4>RjCFE;)id@wSD~)eB2GFRt7ni%;5^^4+FMeHadF=cJ2~RhxI;=
zxmb?^wZr|AeYXpDf7L#?f|88B0Qj$(&^1lY!K|gbuPOd@h0;R4QrD>Gdmnz2{tYYH
zc5xes_{(?W1vp(aXWy^pB%kpdTRCUPq@hyBv<u1u<kK@a1eo!cOmsO1CUL_7Zljy7
z9#7HTmPPds%MAt`*T6ht8p$0$#H}HfsAAqUIMoXdo@(<(C0FFstY#cJ+Fa&>nR0~H
zvHE#&<^BA-|Hv;rd7u%z$IuB}S|xSOAb-vMKYbm5*E;;C&={2`3}{SGkW1T_D2tkd
z=>h=6*8t{qBuHtAHvd%asEal$8*~U1IW7QJc2VhZ?OICM_U+;2?*kVQJslgCWmk7`
zkAo*(k~uU9dh^lcsu%SvoAgP&_TRi=f_-%#n!=`kCIBa#mFA?=9A{#}p)2CI#*bgJ
zSoI7?zVJ~8eMJkNoP5*B_TIvSxZU2RYjd#=hyk9_ez9LXuS;=IBpe07JnxI4;}V#Q
zE*$HCR~SdNxa&VZr#VeHUNpDZR3;Ro)nWk3$WBM~Q8Y&lF`L35P$~q{-3&$F13$bm
zQf0@kYbXaxZ=U(ML@dn{Zds-dn$o@E0-ds~&odt<^^Vv3rJ)j~o2l_jwqhI$E2c@@
zhLl%>&2G8%4L8gW=c6A&N=RBhVVgjacaepN)oFuEmLx^DnQpT2oEx{e84QG~PHb~d
zMIkSMqxceA(V*p6|IX#ThqbQyHkz`K$&%1a59$|ub{^pWp8Q7w4X0}qv-H@XO!k}W
zeg&%K#dlOGo4dBf3LCWF1(19;dyEC2y%cJj!}+SDM|S8odaKbG<*EBrF$j0w79{KJ
z*+0e|vW+}<&O;9-p8lLPTCwt1?AcQ9tt8XEhr_vYjW~0lBS6{WM-KECEG21LG8<O<
zi#A@lf2qY<%Lu;wNQi4==FT}AW<Fhi?DAS!I{%GE@y*K{|9U4T9#JDaAnlWjuikkL
zArYz`Hj0$-xiNXnEY&|_G^j?ozUSd_B}v&;P1d4dlSaPPJKJ2IMn~c7%){F@c|NT_
z6_7)T>Jm8J1;HB-ceZmD^DirOH}(fHf@agceI=do0y^UpTT%32U^}F)*@PDvel>M|
z02}k|!{WuYz217ivI;9A(Yc6EUbA$XE#+0upo2jN@0_&XLyO9QpGOgQI+8b%8SStz
z?Z0`rp|0Gkz}D4BjjKD=%|Tff0cKK_+qaau>9N6&u)Oq?XOSJKU$F@Q(xC3(IUlSx
zn@xL=h`Z4J^7a9fiag;JFkc}Zh_hd`Icu}~W1f8gcI({GoiC5L%9ts5IEm>Hc4q3s
zrBM40-?8GyX<B*LCdXHgv#RQ~T9kW))|$<(Vl|sG^cyrKCIvg0e4b<4!!YS~kp{H`
zU(GGLN+<q@{=)wXD*a!d{WtEv|8;NwAMQ*{1Wf~u#%~z!AwC+A_T(x6t7<<wM|PC}
zrYs3S839a^pspWU+p|xY0p7aj7cVO)Qa}eWIgJ#m&P4u?@Z+C0#(#h<)*9X!FMYZv
z#v=ZL0V*l}gGmEw-is$|EjUALLN2pBfW}G;9Zx7CMqc4WvzlvhJIbB=){?V#9D41g
z@{kWqz-$K~z`VjR$eS;5=11Q?)zH+^cpur~?WQ{Z#z<CehucSDMLJuQv$5Qk?ax!s
z(ivnZ>bDWdL73s&A*$WCGu1774hAt*PrtYVdj}I-xEYRA1jaZB=2^ZXEja?WlV_BC
zyCZSbuxi%{kD~KpBimAgSR>T|{0PC3kx&KxYa`7N$!{qXAMN*9Mmu@<pC3-7%-G;4
z0?M(I7{VuF!?d2T<}2IclnzED@FuQn-A*3OjY4x?^<5ug$$-Vqws-N@Y%SNSnR|o>
z@2NcDXNn2-nsK~+XMMr<!V=bmmdkwhu;T2YE_!*J^cS8P&IGFUGTGf)80yF@i}8K0
z<f}6zs=vi++}nCg&-mU8c+66jej^%zQ)|R*{M=V*^?pR!_E2Nir}mj_Rjv^FBdG}D
zN$1iDpoXf|#~)N-<;JP5v)WGk)OejUZttxazO1{D&XXe7jVK|vRH8=tcUR7SBv?l5
zn|aR0GJL!IH=_cSU_1J!@w|f&C{a2(Ux6g74gtJEj4B!kALVN`<tAB!J~&u+ZiH5p
z`1fc?))k$OEF<$|p8sNp#LPD<OQsM*x43^}wyw-r?+{tJ{LDRr5122}csg=zyAGPS
z&Ga{{5v9WN|BI>F0FS9w=?GK`F|91A`1tsYCL|^jky`0Da~zOrL2kFcieIpdh(gL1
zREoYEuB_h5+EZ8a@hf#tNo|d^`_i^;&1W$yLJsbBUrp5|Vnlbw3!L_1$l{#1Z!&lE
zOSQ978-^8Z*Gr+hnronKjSO;>vP|Dq)wm_jg*byfkJo+V0S7|%ADwQc*QowXFcTW>
zk8@{<CK$<h6ih7~>MPH%5|(<#h*?*a5f2jtA^rVh1<D*W_X2jP9w*nW(z#@4uuG0@
zmwMi~BYVpct0}8#G~f!<f+EHp9|b9pJe?pZY+G&6-Jivj!XC+Bqtp%i-UPju4QJ@G
zfZZkV2C@ZaPl08ByR$Ck3&ex~90>XT!+J>Un(zcZiVCX9dicHas=l{0NQ}b438oOc
zy$gq1Z?_oYJdJCJ-w+%PQQqTuF{u9$2^a*PCMiO+V#ETfau2{6&;N$$yx6=b*a)5h
z!UPG_NK(*5;94R2F_L+T#93tl)$EM~Ak`=K`oPo|;et)z+jsld|AuilwgPXK43h1C
z^w&Q|RU1KWt)}r7h;6n*KTEi1bUcD-PFgkawH4F=;7pfDS%dM^L55xgwRQpnCeX6z
zpj3k+M`%H<3V%>h3*Pf;qyNbpfKOE*VO5w4^^5)8^dvRmR}YNr2Mvlzz}B?c{;zxh
zG{b`T*#B=Vm+&2AnZyCAz^1ttTK-9DLEVFD9jiBj(!7aJu!#5Fy>6whTd`F^{t!#5
z$iG9bJ!JmhfU(^?Z+eKPBG&doN69|9pI-&TFH`qU)FdM>15VX1u2`r`Carl^lWudt
zeW8$;>wIK3uI7?QpoHx>x>KpDAvlO-yD4s0ifJ~{{QQxP#Gm#I`EXriBuUQb1U<8>
z!)N;i#-Z`ADZJAbG&9^9uAa2*6=#U6$FaX`KCD|HY(qeXgdW)PaJdl{GK{+m2Gvy)
zUA!m{(N!si(KgM)Mg-h+)FwZaaec&9>RppQ+hGGCsm$~6BvFR^tmgZY!hXta8`T8E
zUC*NHh)-8P=m_<IgB-w#>o+1OCRC@-(Oz9VO~lY4m4L@L=_#VLJH+Ae!-tQd8bHN#
z+BO2o1X#A$lM*iX+~$%|v5Z@L4?7NtUln*(5B^iwX@H~^qZh1rV3?s(f5T=Tuj!+o
z5txpetEvJ}77<$^_GuVzfd4>nXo4z?Gr^*pTy|Q==@jZQzCHD|4}DU*Ln5(%Dh>>Q
z7Ty-JoQbN6pb`wzcw#dliXgsBO;R@MHrhjPIUnX^_pmKm^+xtlvMN)3N*JsR>6|$+
z#GoFUqW}skSz!3m4gy#z=fFp;1Y4Gj)qrlsJwP!+iH!ojIWk7W*UKNyIX%jz?O#Y+
z)XG;L+{r>8An~NV7<DhzUpo{9p-WGJS&K)hyoYLaNY`xPO<k8cDxucA8!o0hkq_-e
zby*dRvIg={<Y-vFtw+Gy%VRm2FTGnltUZnn7YjX{v%%#6+!}J7BzYkue_)Fm1Bkr6
zcPmdwlFoU<C9<F6`*k@BrPWA&W@VTN!wHRbYmn1?GJe}#^kt1-qDc(N7IH;%4jd#&
z>sI&+^Dd@s3iF#Ry@Zjn7;n}NN~)Q-VZt)%5S_HaqGxki{Q`w~b|_0N-c@Yx%MXr@
zO->T-k7gv0!`~tC72PXeHyVhhTt+sMBLrJHbUJkgDSt`H^z3lOS9lr05hl1%sYWMD
z^RS$kuGvGU*K*@3Puv+VNSFMUs%NDt*g(4eho4`oC6MB2GTR@pHpC+0E6v6ujh81~
zOYXN`xN!XnV%#~lj?7Ub))FHLQW{1WdZ98~@@CIZx#!H*9bP^AcWhrf>CGrRxzO_Q
zg+OhmIE*!H&o<!b7dj+3k%cCUj&IPT$4*J&!oU{+C%TMkC5}JM*?JUNe~D`Qshbp%
zcdg1b2DaJ#qH5`h8sNy+V{>cND~e~}tGbpRr&Dn%(IYB7C5Bg&v8uwZ<bO8VI0dce
z-2yxdj&*irZ;8zjr%M{Xv=m;mYt4@huoZxz?}OZsz(}nYPf@Si-uDED$!uAgeDqwx
z{#bSxR+qn{Y4jAz{;}%=%<4Uk*S^Tn;AnUJD<5_NDXJ}AkZGiO!>`gX=i9J*nK?2(
z9>xQ#%m{L*)Pg6I+|ah|;yx{Y=3Tq6Rfhz1M9tMoyg&>NJXf|L@hBIm&yP%VIT`De
zeuTfWeEd|j2beLP13gu#dR9Ig#|n<!PMwu+_fqXsQ5d@ME=Uh4vkDadv|$iS=(dfL
z`gYRBg+CMC60?dTxnGV4@`0{paw>jYAkUfVAN8*%!?Ew%5FWiVj0(JRqPFAneJojG
zRH**j>EQNKvTbSVl;`)u_ut?r_6aYb1e>&3c^DT%CI~AD4wg2`nq`+)eHSWDXT!Kj
zC4eos9zYm=C1@F*b;W(1!<ua8Pc2nDW^Yb&0?@*tH(BE=P2^;kU474Q-Cld1!*Z1Q
zzWYduwBQtK9-6^015Gvx$R<+ErL<8=J4e;BF$w=C$+qM1uqE<+7u}DXTeK|yhVn=8
zX&D}`)6Xb;BwbAtH~y}L{!X0L;+_EQtS6k0Z?TzwJ@?*~VJ}DUv~62&nA{-TFItEL
zkZY|=wF(mCt)Al@8aW{$?N`@is`Yk0>8uW1+tEKxozbTisy7pWo}qa5&a7MK#qpeS
zOslMlc-}Yd&kV!x7x2%>tD2lWxTJinmkq0|F;BGuGf~-Mq`!0DyWDKc6B3Ha&ts|?
z=z$8_r-{PXc$jy#t!pHl+&x`^EWB%k-cqR|VYMAgF#GHHR-ru88U*tI^gm=uNCU_h
ze2Gb{TI`&WJ@f82@8FB)<O#<m+IPG&I2e^WR8^I~*v<CjfF)A!4Uz}+tjWc;9b4MR
zeDtnl6`6h(+ls#|F1Sw}Rf~F^^ngP!J5Vnw*t_L7jCFy>6%4ybb%BP%InGG^Nerqp
zW8AY8qmN3ANxDRP+Y~GV8n(grj9fh!^BX1rc$VeTZC{)`srPJl^UG2~z%$T?VqSzN
zX5j1Q2YYq+dzT9}jm(1iS!$04pncieh;tpcY{0NS9eN_Yk@{)2^ywYLA87C3?k0!t
z0aC6Xa4;0028i6}xV^<z(ERYk$+U&bU}FjDYQO7-Ak&Zc{A*H<Bc72ykZ0-<vo@?(
zl5qRV)b3NIN$=<Po^@u^W`4u&-D<~mWA)j)IO;T~jlRX?bpx7d2B|0KzK1+ao5C>j
z+0Ojz64D?v`i79GgYBD@-M%S*KfF!HCaHdab_2An5p@vf095OLKd3Ljr~GxbeBU7$
z;%a%^u&HwPw@szthM=xs+7cpu!?qS1=37;&`@g2Pyf%C5kS2X{fLr5NySANYz<+JB
zVTx3QZ<6$?J@zdBVBeunwMJg^Ao}U}5-z*3h!fNGI3wnKqK50|&XNjAbA$Aon?j<3
zc=uWgh+B9HO#OrsMj__8OqXubnLOKKv_rG@6$zn5VM@5kZEy=+crHDH1A=emp^o)h
zRUNs<+<nzXu;);2(szt)O?sEe|6s*|K_Et25_UWWk1qD#<dhkvX!Wt{irYh_r1F!m
z`CG-ma~^{x>K{|LW%@tdHnUmsyhrWUf;!-Q$3S<^|LNkZB5RRsiG)?uVl-fz@6Sl*
zM~{KZu8EXQ9i$2@p$KS3L4M%{+?ZccZH$~*;Tld;mfwp0Rne4nP1rSMT;kPn<lr<q
zg0vY3WNh&uqUy}U2P&pacJ=#x_jvm0d-K5=+io=`aiW<aT0`9!8q{HL!HRiKNs7>0
zPBi&(cmmTV03);(Sq+ZBiH0W`slRp^ntl)?W!%WH-Vt(7)F-9s#EZ@LG(1o%(B|ua
zq2XY2!8I*xuFIw7thX*{>z~sOlxk%xYq+uXIl4s0O-}Pd-bJ;E*XxOhNYp2r0NGFS
zX{X|ll`}wbl-Xask7H4<_vx~a%bu>;^T<l&1!HR3t>cKtpb10e?465brq`Acq)m3L
zre)Y<&h^|5{;7_QgB{b_6ZhkIE%7nw3N^toy-t&Ysgo}cxvFt~%J|Uit=brr-H*F0
zbPlyNefA5}x&gh;$e)&1iBf)9;FIQZO7DSF3w`MBGdW>cz>oeh(V8LrC(!<IZ{=;T
zs?g-~fU&l2cAS*M*$Xjdp#X^&c?d*nw>ptChmlK8sNoEW9ACKYgXIwHfZ%{S9mzRJ
z;AK7s36LvDR$fuUk=t*>WiXBhE>pbaZw>Lkgg$jE<90<hjTjg#5qn4jZ2~cC`mAYP
zhHYZ?f%%^Co-0oGi-jI1sklBqR1+f^*+Ps!x*;|+GVm1p8n#TGT7>aQ_4K3|)AfNz
zRk}#oPhMH7STf9#b%ZYbbp>_|ST+7?g=hpS%(H9r6Wor;Y41c4hdnJT-)H)N%Ja6`
zvc*dI89Ak_;3$t&Xd~Kcyv=_fNRsx83H_x^TkIp<m>vTu3luxlfc`jt|K*9r+B=vV
za4sHDX#jf9JdI2WMoI$V<R+w~oM$)h(yPWE!ea%~qt}Cs+;q8b(BDO8+;F$alo0xI
z%TcW<;bn7KzEaK5%fz)jz&N)01M^)P(E+4KL+3y91h{QDU2J#5b_-6;1`8_HzzFvR
zwl%8dkH6zby@}Vy4?CZ@Ilm*J*V3&$>>gKAP@h4Xy*IjvEQhVI`%=t&QxI-_o2pZm
zdMoP8vdiJKSYlEMesvQ7xch@aZIT3?LVqBK(>S-dS{!XlAk6U6mlzmNjH(8Xf(jH(
zp3D$UF77oy0vX$<t+La>J<ZD6GR$h2mrp}*_M>Km>BI<}V&o{ie#qgVj^}45Orf&z
z_YZ#SSHn}!ju`gOTnL>6Lt=fGTQA8S2>R|27Uo@5r<cXA2{WDO#3oz$(A}BmAJ%0w
zP^^9QW-)=|P@CnT<xau+ek%F{U{oHfA0@`jnCuQc>pj**glBnWFNpM<TBkI$GjXwt
zWQ0?iY%2&o&TNl-k#eZS@S4NMmR)b`ET+-j=}|nSFC8iAouT{M3Lx0!Cw2rR9*Evn
zqx?vSRz_l;r4F(e%>)1&iVk*@5Po#+=c8j~d4_BKpjEI{k*$UNL`W`pUmPdsr17tp
z4Qw~%wv5?rYO;9IQrnJOL;V6kKK`VQHC;4vejv0@e(?zCCp-q-SS7Fy9^~2e?gUk4
zJWpg#r=TIDVJe`e8E<rKmx@3bP1NOpaI>CJmC?=4`E-fk(|+9z3j9%tn$wKeO`%gI
zeaiDC>zewUQsH6uo!0C`#{p9%n_b<bX><Yz7ZU2$;$p2Mbkk88fc)P7QEla5h_5=O
z+K!(Dx!G>P^^F`(8=>6F!TQxUvy*cZpDad}2yN4de1>|&^JEzT#Y@%bZ~$26919i`
zpD{wC?ydFDHS=W3=zE_siEK3UZE=~Dk7L6+?zsmNcT&e6AzH*?ADMuwH_g<i4@bXq
z;vDG%9lYejsZ;W2XH^?EUrHKKU>6_~kxUyz1W9H|<D@!bvl%_<CVgYV#gAT^ia~~(
ze36gCT+LvR9<n%h=GQ0$BAwBQ0t>c&QRDHL>dbSa#3Ro_61n}8RGxG!yOP8q&@%}!
zBL^zww>Kq7mlA@t1q-7xDdXIkP0pe`*<bX{nWOXJcG%BO!o+Lo0nM^aM@~t8E*D%u
z{e;gTS<<#3V=C^C)O~$3pB?Ep{d}9{o5RpNK*SQ&St@`N$o<PX^e;|<K%~=O$w-_2
z8#dfFj%*Fqf~8XDp2hQ%IBywn4hk&Tp{Aq8FTZDct_azlxrc7wTxQ&yzzYmhw<HJT
zO>H*SYHe!?0Q973=#P=*j7R|2W*AWFQ$0U1b)WrI8Rv7q$j@8M80?5d=Q|>1w3+(#
z<!3T{@%4NYLo0bqzD}Z3PpXq{hoB7szfFxn!*~j>CezxkN1lzjQ<v!flyQD0td)p*
z0dMyo(iDiDqla17p>TUR7GCnP%Qna4pZo5p*yeZ6{O7JGsoP*T41bb#>xxvwj2F(I
zYBMIyCEoJ7jTP)r2HTrW-RX2Ux3Fue7rEpvHHjv#@V2JYjnnJUl!xe&?(VN;y^nnA
z4j<0*@N5z8j)!p{3N}6kK{mLV%2tR4z9QTCN~vZ^3d0ev64TURa7W%K80dSURwdR7
zKgJa-OXk&m{A{B@L_Hw_cc|a}=)(0qNlu+vrz2N0V7y51W&*B%JHAOXKD}C`%Wp(e
z3>!p$p#^e}Jw3Ds?k!V?1CpOz@E<^cffmTM@oL2F`%JK6Iv3RJZCE8fk$DVquFWUb
zM>d9HFiO-o-HDWOZ5Nlb!dBYLo<^xQkR?)Nu2PgqrU%FH40EQ}l-`uQdF1vrc}-U6
zYechRmDLLggt(v{%)ClPH@35Ojbo@K3R9FG1l_a_ByQ)wLekQZn@sj#&oepdRtdG$
z`!2W1wpVCd7&4-6G=z)IiG&3JKWxQ1Xi4W9m2()}llW|boPHE&re5oWn64o7fttLX
z9ABvgZp|QZw5Qyc{?L-87kT@Wx_QIR*GE1q+lS*~NOz&DNQw!P;zsjJA2k?|^c>(P
zoI4x!Exv-F9ILy^_!B}wr4p;kqDF?IUb;-K^dEaIV-~XK8LMRwf5m3VYW8{cV}HTM
z64p9j<fv_;k>)K!MfHmH86!KGdlpUHIA9qss6O^@iSyJ%<XA2SV#{a#hFR!BXP*D3
zufrg7LXjp9yf+-0kr5R{x@5X|aKReaO5%tLY7avbx|9EOi5^ArWKG$HUpR%Vi~g{Q
zvQ*6&BZFP>?)JsjL1GTt$?pzp0n6@fMrT5mfO@&St90<-?TV+_2F>4hUnRX)U(N0<
zNg(XNb}`OZ8@?~Grs`eYeDnJ%@_V?+XMQ+|Mp%K0ho`Z*4MirIv;?ubPkH?DX$hrw
zaHjH!E12a!sNDxr&Af!SZSEzku&wnSd1-%dCBl!?VN5aqxj@02y3u4a)~Sx-zFRRw
zT}PDrt$k6t`OD8Lxp3q$62_mi2Oec)Pk*|Mk1Ex|Vq8iRTltwkwIU`GA1-R`vF#!u
zbngc$9h!1Jz|XNd`1)E>Wz?wN`Use@0dWdVl>M7;a-L(K%hTt-IYeA)3&8ji&Df+p
z8>Adg=5H|ut-C*<uGMcCiUsGlL42|N<&z|U5Ypn1Q-7$rU;F`Pql?P}2%9h032Lp!
zl7VU1YX7W1PX;frv8aRQpqYGNYF<*td#0ZVj6VP~H46up;M${rS=ke+Cio@zS}gPn
z!Muo$P#Z<g@J9B0D1pD@+&p(nb)Z%5g;u~)>jo02VXh#OglU9X4SLr3(TmQ<7EE1@
z9O_6Jdi#+ld-*WzP+gFV5!gGCK5P~|Fcn0`9AOw-+@{ML40U0ieZxdEl)t=c+vM<L
z((z&I^^HwiEFJgqKz%uxb?NZ>9g!`&(<)8bTAVNPP5y)Zb=xODbI-mg4H%-9-b#o*
z_Zj!9>dTV@sb^tRHZWWuIC(dL4nl&rzq3-r>sWC)Z+!l${Z|VG)Ja@UwINt6F8tXT
zDLR{L<$yt*)B3<Hm$VJpKhU*7T;>OFSRJH~5%)s@4K=!(YlBT*7`dpo-3jt&RK^|e
zNHcd`Rw=|~fbZCc6C=bBik>O&tcb=9kH&3yk$SkkV9&{Dwt^#!>tZFnad6BdSx3Rs
zEgUw7VP*=%&vBi6ASK*ED%R<2c`9NfCur3AT486k!_V^<#*R*H-k^>CfGml{%h1U)
zT!!!EkIAMv?fx=vYaNvJE&y!Sz)`YbWDm<DN*kG_J5CW(eOmMVmtukpaIT=BKY(=V
zJw!P&Y~;mI$nD}Ot*>vnXaB25t5q0T+vBd>yOvHT4B%<$QP(A(Kcy)h+vX5dJv#R%
zLKFr<63g&)q590tvGhXy21li;=UF}vUI$Q5zB}4lGdyK;o67vi12M?d93b0~8zIg+
z7-x~4^7pOO<|NS#pBf3jX{60>wJ5*z`rhEHFaC2Tqe1upH2!W9J`w3jOd$KJv}SYi
z>&xPsT~#n|B}`r<;sW_906lJI68MtaH3br2ypHCxxPnbmbSm6}b*oOWhsp1ap7!ZG
zqPTiJ*Gla1<pXmRTW0c-(3&?_NzhRW!f<X4U$N2c`rZ2*`qmEeMs8nG*E3vq6XVZ*
z&3UX}hX_~JV|lzD-88Fz_KxpR;?tur*xl!tMPS|gmbK0aEyH2dm1$Qvk=JWe>Isi8
z%TreqQB-7Ue|-fR41*#>@{*0-UV{BY+n|y-3u-_z@?eK3&A)zL+=o;w+GL||nXzC|
z-CrUrGkkjJ>4hJDeqlnB=r3q7h?YeX5y)TDODm*f!u&5-_ImZqZ}NW{8&>hbfLQCl
z5u1(TI5DaWoj!P$^e$m~pPB}x>VNN3YBJwQ9PkG_v7tbQa;UR#7bpB7*7ni`skQ@6
z-Dc{HbVbzodiE^XSPWalRtFAwHYQHSwl$Q{`_sibAt~uUw%qM!r1Q3JVjKLuduO!!
z?fl3<y3Gxvj1ViqC%w!5=4*0{9&vsHQ?rRI9BSE}T?OJnT=De8Yi>R}8l)xFR^&-X
zlOr4q9euye8<?9Nigtbf=B<V}_4ucEzZ7bAF4usR0pjeQZ_+r=3cbLbTOKaXztoiN
z`Q2)r`IUn3=g%ZYN%qsvne+9eP1uvBd!y-9YUU4;-pm}#j!Kz)e9l?&KU7Km@2*n*
zuj~K+R;m6UH0{-w>PVoCK`R5^c4tlGL2J%&9vFhYfrUbUl05)MyzhX1AlIb(0fAl=
zL<LRc*mN;s1ulpx;nM%!76}0sx!S)-V_$i0psyo`UbB0=D#aCj_s3s#|3$?|)91zz
zD?lj(zUALZh2r>;&=^y&6&mnC@<jmI>=0nB2)B|j@KrtPeJXnc;Say;mm8hRdkmKV
z=XP}nq+c4)DL|o%<z@Oe_;ZDAfo`II@lT0$IyMHH(eedueQDvy9I^!1b`U6Y`pbcQ
zSpzKOc|UCQXVLFphesv-rPl4!KbIbZ9CPl5riYP_fyGzt)k?T#4C$IA0)^CZ?2Wvv
zE>v@X9<V$bZeqr{!-S8jH6|#`8zJaQ)(v_els>h&$<bnu{>qq!=<(IZpw7cAQGqa?
zMJxiB84qnKyl!O8z>AHz?6wWtDl;%tB=OPtr%g6#o9^0RF>tgJ+yF>@yx_lm@HcFE
z^4}e3V*c8aEXuw|<967wpKEhwF?}O?X;zBi@3K_|C&Q7?sGO_BWN7KkpSm`j!NaU*
zvJWr?6?AB@qVAYioX%u%;nk774uyv{D4Dn=3S_?lme)DtJ(t~2Uqv=z{4T`HRV1F$
zeipd5I=(Y7!FG#x&{}j<zHwHaXuR3P*Sk{3<@5fz1FrxF(Q?oWccCq$=|Ym=VKe#-
znrR7N)#m#DV^bV!bOwoOueoZBypLY)qXMQggV;snsEYEQ0ynf6B9NVd-AB|}Wr9uM
zWcP#sfycP!;%R;=u^HlM*hCHiBP*L)ZqWc4U;DsL)B}t>Yk(RNL8GpJBWdXv|01y!
z!P&j{MLPo-r+12(5m=ecdW4@u7xutV0C$objOFbTz%DQHU+!w=fAZlQ6I*RL8eR#7
zIV`nEBOTS>FzXm(nGQ7bAy9M?*e^l<zU|Dj10zHiB<O8E5Eyo=g}D9>Q97C@n<+S(
z%n($G3SxTM2%c1ez>*vfO|K!xmRZ04lUG3m-&F8_sqC3vyVeJd!9gSIGB`ERhXtj<
zcH~$-HdjD;4fa~#i{P5TeDpKf1VAiZ!O5z3l_<78n2P~SH2)59L3Z{pA<j6cbdovB
zM!GC@bI!e7#}rIKoxL3avG8L{9&$bf6#7zlhQ)x9*{E&o-DvXM9K)$lZ^v)gwjW)B
zCqyvcf%<8~RZcHgo0DlfSiNW+SK?%s{nLB5`7R64d5q(#hn^|tsBuqX?HhKv?DzFI
zD}IMPplaf@&bSERD)U~m&a40vF`h6Bicj-MUh8TVG)2)`_tW;ROry!<TJR7@fP>iZ
z%ZYs9%X%N*sW^-Ly6(U4h!z``!QTyT_xDiOD6DPmh{&fe>#SQh6z)nQ_0?Q>3`s1y
z1cMKI|G@;rL>p{EMs<4}VoxQ^92S!w_j+vg!#JP}$ihmo`t@?v;>_px7Y-7R-f1b6
zQ(E}PTleuru6O=pnU!5_1^4I`;3(PBHP?9QfMY@IEa#1Z@9u<J)tKpg92jPw>_ox2
zGHuAKhVw#Tz)_cU`6B&`li(D7fLmV8MDW3bGDUYd{~gn@sM$#FQLI1vKF74sd#&V=
z@4G=tTR<*VMEUyp18L&$mpg^gWRc~pmB>(QjpE$a9-X1A0z1Ov&I&{xagsbLR|b#a
ziH;gy4Eig+?d<2QglF(Hy`y~f)hC?jQB@X{960Q!OUZ~HD?9MHNmvW`ux`OMBe!be
zv~ca!qO6NKQ<SXnv_nDcz##OP$lazY;w7w`d1Y^);ggvejosck8Rm8G?xkXW0@oIo
zSd$d)PR=3k70sdDcOE<CInN;9{EaE5aR%Q_C@V3^;kPrJOG|scJ)E*uls%MR-1J51
zi;hqeu8Y9DWOJTzzDEZc?Rfj}B^4dy5GwpBLUg{d<g;k0S*Y_UddQHGp7a`90>59r
z{Ox?H+@RO0c?e7dgKIz(ZR5YiY1Pw*GUGIECmd}Y=A|6^xe)0l{Cq5)|DA*39s>F@
z+uV9@`fu3ANoK*dtwt=dC*1MHb{I~40tvb-zhOxz>kv1(B-EgxhV*tn2s3WoTNT0m
z<E@20!NI)ev^jt2)-iP6+ezG0RH6Cn=(*GvLWz7E-1OiiyZr9;lG2z#WQByQ=p@**
zEr^pgRFC-N-|#f|iP=HQnU3RdmZb#@)Q$7Z@laBnvVu!Rppz@8`qM}`#0_BT=Dr3l
zw9pzLVwlo-O-KaxrhRXqI%u%Ayu++{GuH!@nrR-bxiT~A&%156f&ULO#l2}Jn*FF>
z`s?7U&Jy1;_5wqkymNRb!rTY!sImAPCfFp*v=_+uM+7PBmE5mdiOHd~Z6AMb;PBTk
zrQ=Q4*7ftG5sVY8=9&tk7<%|l1|jy>ehG(3O!TI;ht9D)Srvu@2gTL%+3x$eE~ju*
zOs>K7nOxb8I#XW$B0(T#&$=a$Et@rTeQh{X#yaJ8<|g&_j`{U#VU%ek?E_lQB2j?6
zguSS+5nr#@zCJHEV!=0*5Naoy0@WG;@5dx`ac*p(rbgo0p=UoASRWr`o^a<4nn5S(
z!-3okg!vuSVt8@h{yvkP&L=z3m_@}?4)PMa<KiTYlc)GHFW_xtjz(5P>nqMWS<%*8
z25xiWqa;n9D08S0u~ip8LlnQ3es?ZAiRxKoL&jf5O*ix8oN9`EsKUkgtvkOtUYI%#
zIEKA3@JTcQu<l6r?Lv+C1x5;iVZlmi>4Dl#KibrN?Ie85)c7;F+d^MKM_M)%^v+(n
z0yjKS1j7pGgix$*L0s&8BojmrBD#|@8z^!3gu>Peix)58zu<yxL`E)O2EGNNApzM5
z|L=n6njxh+qB@KfJ3GX@Ra1@Z;VBhh6J{vipG~C;f8DT0RXBXLcKWz7fUDMLavT`=
zXq*_!guu|0q++qHM=HfaGp)y-(LAeXSfa5@^Rm~<GRzCTWd)gc6k<jxu8&gSc4k*Y
z4LC39g!r|MeWB`a>+`+e#D~4wAW;!}?!n#opMrz9E(mTSHf-TNIxpqqot|p;Yft{s
zMF;Jq+FQ>66W$2;ltUbqq~-5r!TG#HeY0J4C%a11-#)Ce`gPp{wQa4xj<6N!j!Zyr
zbi89^K`T7l*K{NxboaNjqBk%r6*S~5l7^t12rOlqXX`RM$1UGiPJ~v<uaBr)I6Cd}
zG9;HWWw(k7>KHcTw`Ol9r2t=3L=CXi40T|B{|%dcF5gcqRxJ4St?1Fq79vp~21bQr
zwR)(@=i(BUy%|H};9+^`5kGyWklE|^2!*&8JQsd6Zgc20Z$DkK7+}HG%#?rh49~th
zds1i(HRd(T=gJUg(no#wGOT*sdw2$gcL<)|GVo1%f8^hUs(?ce1sYsoM=*0|cQFOs
zjX2paRDZQL<ASI=rCnEK6z~1;EQjLI3${En@H}Opi4CM(Ahsouwh)()N!#aRB^FFF
z-(GsJ(Yh*ke09W_sf8vYq)B50uzyYEUJ!jHVUP|E#G}-wZ=nGfTmBBrnlV2MfJ3ny
z;GdoX9RS=HaOgeS=l_T%2x~|CmqNGzWNC#nfhX*4ut(qkHoJl16&Z?z3a8}*1}4)E
z7cc%eeIxJP5!*@hLuzp>P%T{U)!oFGr~73p6LRv4RZtbJS<D;C8inET&+UN-rUNO7
zQkn^tLQ;Lzky(1T|JBq*yGsLaJ<aXn1BN6p3}^t!cEGdqIF+|UR>w*^id*`o?Xc+w
z?NZlq-4pFZwMI3E5Da{>>a{@l{S5x-=>ewzgCX)2Sav<NNV6H9*l!i$MjH-sK9Z7I
zmiQR9qME^<<pcvEw>HPYO({3FQYF{PVte$(3@Pofo2v#y%aV_!AMqX~l@}-Z0}u8A
zdMz`d0l~DU#)D0U(S;mpgZy336tCAq{8vuuF_3Z7@tSQ-cb19IaGTzf6P351Z7zan
znj(S&ot#?JN*jMVVQ|x^V%w4CfB^ShW#{h%A5lZ3VVL%GrRG7weypP#@7N>I`oja>
zWQGw)c6QH!fbH@TnpjsY4<a@Q<*i>P2<&Z$HI4i^Td~$(Sd=+5?@&k05a;*{Hi2Ye
z@2^+Tz!hNb{{nswqNU&1(>xpaLTu|{C#wOk9$zB$G;cFFkRv$I@HLQu0zmpx|9lcm
zb`gjs-2^tsMiQI6nz80t?FrtN2({*N$^@WGqz$avNuO+a&%Ev-XLgc+jnL%wet|K_
zD{g?n;Rm@BR*zEGK?5fsV6oDrOXP)^320zgX#7!L{M+86lWr2jc2%Z?y2d1cu0RRc
zHtB9NFPaEVR{$W~oRnjGuXNIv3DvelZsMo=K}Y*YPpjMnC;Lgu!=RU`g6_7cMX##j
z!1D9qLYW}m5jG11!hZ9spyhcVtej$C2%XTKFQK8!f_WKG1H@h5{zq31<89$=A~)gJ
zdkG@tNNc*7wd=&6wlqy%8eu&&aJBz8%)Ah?c?4A1faooU6j<sdfl{sz+#uA12o3^I
z5R`Ufk=LlpYgFzYSSVOJ*XM%d=xsN0Iu6XCXk0JPHyBT&CTAd*6JCxyUR#C7^Zo+2
z(H`9Go=L)T1Gv7Eq<~QW?6Uiniti^akF4}*55e=BP+<G&Md2&$dR5oE`9HrnuL93Y
zPf`vtqLf`q660yerflfF?k2_PMljIrq=uvowj#VUEV%~^>B)>o9a3|tQhfC_7+7L<
zIfR<?(|PbP-`nZ(haEN^Vo0XNA8%hMd)c&4nEI!dS?w2y3rRksej+AC&ksAH00KIl
z(>H{^2j*SssBOH9v>bJsRO0(_X%U5d^T*{5s6uyN;#jcV309al&h{6p;*S<aO3o^8
zGj!_+KtCjW#7E3khH<U1nwCWg-=Dtc%~rSQ%dCpS_+9*B)#r#o`%%o0#P{-pZ%ist
zi@nUo^Zs$w@=niHd@l}I5lj#KB)ujkDNAKGlf*;q=@T$*AGssP@0Cki02h5yMgrNe
z4yQ-Bd&DN!NAmH+2`f0qPIPIK#-Yd7m7jEaR#W;hs(B?S@nYj0(hP&4uBi-6jxH9t
z-Aq5%{$?mnLDA4%>ixJ0t^c*15IO<LbZFbB!q3)Ge-prt(>a>*uv)k^a;<+>7bpGa
zN{s^C7`Bv8lh^&E1G|bRiYHI*m`PD;+rI+#d<UcR-*u~^!CxL+w!v@_oKZ`mBiEw>
zCOj_plhOZ9I@eDq`E@S`6cdIVvU#|t#V2nNO5Q^Te|A<Ko!oBov+Avr_gYxR@W_v&
z$=g&Eyxcm(mzVhoID?k>HfD2!$fXVIosv&X-0gpa^}76-Nxok^dwyW{b5WLuZJzG`
z!QPvPL;e5z!$f3H_I1i$lI&$0DkMoI5i#|VY%xi8W<)4!D3zs2))JGfA(K5xc4KYK
zsI0?`WxSc?bHDnY?^(Xz^F8Nxu5+L3KG%I;_aCltS!Q`Juh;YWd~8g8h89&XAsJAt
zBd!cpxPPo^sC<~^^}s{*8o$o+9XM#S!*PGxWzU15p*pkxMjUPJFh~#%2rs<_{TKd0
zOAdN@5Ivp^ufla8;&9a_%|#?TPM-;t#}v_w#TAWG!`vY3U|I^H#<LAQzItg;cU9Vl
z>>l!l+e<MKuEQRXF~#`kYeTdH+t${s?T^vsy?9`%S?|8Dhw}!~LY3{2(5`4BO*!=E
zoN_e2@GtIU$WGXreGQJz2X5MeY_J;qC%=t`7ufy15D%-({5!GUUo4}5#jLbZHyW&J
zvU~@gqpN;uuvXd{H?VVP;?hLdY&kpD(l2h==ufuQO2pt+ho4xJ1qIhJ;cV%Vv2NXP
zZ!dpFuLv7A$pWntUi>6rrcctY<a>cdVDa~`J7xUZ@8SE?+|)RBVh015Pd)2OGM<46
ztAz{pIP(Qc#rsb-2@qrG*kO39)i+{N2~Jkm{BlyT>nGOo<u2TQ_)HxPysfxA3gOEn
zEp53BYk*`1oxO=elu^o9ydS7mcF;Zg8N1?zA!f10F4%kHH+SNmsqZamR^ymQLJD>}
zxCiJJ7qGV2BHoB?Bfzp`{Wn-qG&~16G08e7(}~{P`IGG%&KEtCi(d0F75S4b5Lh)h
zGcGdAaX*Mno?yMhGL~p{ES@f~lYqGREjZ2LrBI1Fb&kb%ka9K@P0~-G?)xV0wOQZX
z8~oYSX+qiIBI*Nn%x8O+wg#!09!K?lR~aCtpsO_$=uu7e+fTo2e}<i}W~U+<!XdNp
zrv`~A>FF;D39yN*#y&!NVu9M$r{1Wa^lOn$b&w9-1Zr&G*GG%Us>&GctxDTSuoPy!
z!tsEvRVdL8bQQ@I<O6?Y-w5@HpEjSz$O2NFe#rJjdev`!=Ig>ihRh7#@h4>LE9vsV
zgY@LWuxl&<u<)-znlr90vn8Emp;OUOWfw+j*mcLHrVNmeSmtFrHWY09+ejJWN260i
z3VX+#THl?NTz={Jql6)hoq2#a4#P&C8rFD_VEypgP_g2HLcgB<3@&xru4vN>U;%jX
zdjU{@((^g1u9}s|-_OO*ym**TfHE)D4c+azqKRCQxpA4wUsd-KJD*1TPB8{#=I4>V
zC2c*E;~!T{PM!G*a-1lFID<Hb_bZ>;wXloM(;X_SW|E#8y=>t~h8CjhEC*n1w6#kr
zDl%o!DKJV0&{xhs3ylTlMj&x!lPMgk+3t)^HdZZMa`pfmdWqVF!}gZumpOysS4T$&
z0D*P8?Qd>*uyx`KSdwM`+ywv_P@0vdpJwi+2dK5RkS84uCD;rYd%V?4%`a^qIw>2l
z%5L5=6}Z;VJr}~*w?#uk1(dScA5XulQ7)I|duva;HyQciVs8Y<6tBg0mcwS&)O%T?
ztp4xp(2#Pw+g`li=9Thc<AGCccdu<6W+&A)D+9SsNvoW^hJ8mvM{|mY?u$!zt{!t;
zO(shj=LWwNdWiN-gOue{sIhL1RI9d0I$FT@sb4ec%MW3%xex4@9@*(_Pg4;r9y+y*
zXgpMdzptt1XDFLBeD{Z@klfmB^ciM85S{o(szDu8DPv(`l)no(XeNVrsjXJ~_13%X
zrM@#;-ubB{{nSfW6#aSYIbp5Q@TVG@PNNLI@~Bs>zm}{<G$y@6yRG-8j%*NG;m|=`
zCn0XIv(A|wT03xbcXM^`e!lrV3OG;HAyf~HuUY2`6VqWTT-kWlJ(A}K9DnE9FPt_U
z35a#y9#T|48Nl{N@<nJ3`W2sEss3=!T{4*cic`jExp|IN%ou?-1FCzEmpx+d(KkZh
zbg(xK7GKXeCVwL`H{ZkvKR5z{v)kNEaS(`Y4wYwp3sX#V84&n-CNxXCd=1kj#uRH)
z&EvmwcPbyJ1-vr;n?kq$+5da}W^puJJ}&@WdbPvT!$02I^LflGwMFZfe0Lq^@MKjm
zn7TvpY>P-rQ6x?mx@nJ)ju~Va?C{7P3%ZpFIkq77!Dotba!p~Sx-?UDOlxg#?u7R#
z&K-|?y6p6M{B3nV9ZQ-1cfDsMkA6Nb#+bN@8YM{jukIcu><vn&w@*Ki9>^%GrXU%j
zMzi?X%{`#gx&J5I#bL6$@wl6qRvn%X7?#AVVSD<bL1EzH>T*887AkmYQPla<-PYzz
zkD6-AAeTS)Anvp=2Z#s4h#DJlH4^>Y@Xmw9M1Mae!N$rjxpx(H*#~~R`TRS-M>iL0
z8!Zb$(@o}ABWC4L6P-~x#CP9K2LhH8!NrRQNfUgwuX_Vb>Y|Mm{j>a@*nD()_uQ(~
z;TGs{uash2%H3JK1=OC4nRlfJo%rW&aio<fG}wW8C}4DJg`zztHL9n5k%(IVIY3gK
zl2%`Dn=xRBi?qn=()BS^?z2eV#f)6N5nS($7t3GAVNar{M1;84_hbF!nmmq+btBb{
zPo@P%$`Y*PJEQgH{}LCK_$M0@Gg$^I)%T|5FDK?rWe0=mH+L&fUI=4%<C(oK{{+w*
znI9PrbeGVwD}$S>0^ZVx#gA^kR>3G0J)^b=+Lg0KwfseC3}}e|BadQpuxiitB|^-f
zY#a=a*(m*Q0w?Gvhki80kN!xPO|B<Cn1h{R+Xz=*b0qdL#7~l7DaL3bR;X{rPM}Tj
zb?M}|_2D~H{uUV)C)zZIIkK5Pf7R*(1zRGcm05_JDf?~+sZtYMDJaYgPKYqRuy!kz
zw0MQlC8en9%RbT6#e(@!VXTK>5rwak!QLog2CKa5I9tLnog~4TU?1g26DHGGN6Nuo
zPjNxYg?(~km{Op$QJ#{%#bI_jhXu+7Ko8DK$5oQ(o-LMY?~U7Tw&#X){=$~C*jWS|
z58O-r!L7R~zH~*WoCh1dm)_ccj)-H+?dKlU{YTc$j;d63Qb`<YFXSH;($O$d@v!97
z+sMh+yH@vm`s`#fy>YzFs@{Mz!&d}VnXjMAd~Gr3QvHQP%!DGq7Bu{ip7Q_y+5c_4
z`X6J9%>cEFsfy&I;<_|B81O2}x!54=RVN=u?Wr!GxMZ&ZHBQ$4e(p~QicR0PA(@*A
zHtT#16a-@D86sOWbEt|^M@t^qB#4B<&kt6Y+RuG>p&%OBh9e>>AOBu<*w+Ao>Te~&
z>F{?*#tT?$8ePB&G`}^dUUbVTQ@PLtQps`&2?OX%XLB4!D^zMi%Qf(U^;gn!IOYXV
zz9?{QS0Xy?*jPgZL1XiZP{oVy1~!-ustpZU`}QgEPW=`LBE7&;v4)&zW7*b`44C&y
zK=bRnJqCq<0ny1eI872zHU@eS1ASyE;`GT5!mg%(0`JS@kk<(BKyxIg`{5NUHeSYQ
zirumVwZ|5^_tI@Dl>1WZguq3f8;?HSE`47nzH<$LD~|t)O-2@gy+Vumd=A>CobWq&
zs`g5x?MqnA$xgS^m$X8kPUSj|y20uqTm?I&E)_oHEjzeuR>lm~WoL9T4ngTy1`m{1
z9GYY8eYhq{WmZV)OXc|hCC?#61zZUHX4>Qs5l0q#`HAXtdtQFO1I&96S;RYlIp4)=
z8MV}oyM=|6i<78NFEmme+q(T2$4OTnW<(q}b#aMbbb7?YAI4Z(ckF__+v*ZZ2v2@=
zHB`c>A@1DnpZEDGbpvs1=o59On*DFdFGzkjTaP)|NccYP<VMwO7lBVC4K9pCFN5#G
zI0#9>UvjlvTtB)PUfz7^#LMJY`?!s-Wiy}}h>u#<ZlnttA0Q>Is>jtiDfxY%VZ9+a
zQ0O;;+Q#vcJ$6AmRQ;?aSD0Rq*Dnu1jK6&cRn0yN7IDNU0XteEH1HDg8SV5<QJXrP
zo!82LSbQestKH8Iw51ztz^}<5-+RX)`*85+c`Ky!n;Aa`-eU~)Zk)&JxMM5)eb+tw
z3CglxP?(lb*qf{SMXok$Wu{`x-G=@GeDD696xRnVYW@$ZT;JK%$swR*`#?3Ky`f?s
zFbrm<PhJ|p%5%l6?0IO_u2#Z57K;c*w`jA(p@JE}gzZKW!9we;Gv81@;y<Tf$5+#;
zBTz!^V2g4`JO|6Hv)z6C&T6c&^6XMTj)uUC&L8ZV^f}j(b$g)9|H80?eDfJsDW<Q9
z2w~^)@6XX9mS5VGC1W_!0&;;2Z7=ga+DDjvB+?_v>7!$2+fIf?adA2(r-$>2;ylnt
zM|?uaGPtQa=CMwc)kNJTI?rRn54n~BE9J3>R9v+uP>hohZE}}a)>|%Tg!Z*9|CsFZ
z;+uX0!u+D$OFl-90%8%aqf92>>-!!mzsQDg)bkbwu+NB??ITxG>SpAV0!^_b8JB5?
z3n{hFZ?bM=zJuM>j^UpD-Dih%_6ehgU?MeGIiK5SYnMATNVncke3E>e7IIE!2mgj4
z1}+?Aq!-jtW|kcv7esW`30>FJEA(1>zMZ&B@SLIJ1g;vt%=?{<nYn@8#n9|uF+N!i
z3WsX9>9;N=xsTa0Q)Bm^=l=HZF<@{r$GEY25F#l4$WYv9({-W?S;RKeCTsVLUkY!o
zKKax~M3jlL|7JGt23>1rK|Mei%)x?H6O{li$UndgS?9N}W7@V!C>|IW(QKjD)Ii$(
z92W}qK5{ckqdX=lDKzOwh`bJjAHq;<5HTz*NP|+>E*+9w?UJ%@jkI!HF`meqbzVjl
zc)`(vt3cFf#Oe-Fl@;T={S+DitBgwre3PlVj$J6blu|t8KWqNsyM%*$;cIM*T_WZk
zx=X;5r1}84+{E$bs&ilPnuWrN+h6R!by|t8#4&QxKqax%i++Ul!K#B{ndWQI9U|lI
zM*T6eTqEP+VEyvZ^~W=}Ey9b~tOAf?Kbs!eN%!z4It;|0<QSF)uK#{%_Z?wu0QySX
zw{Y;W^$~Q9UykhVC&`JRQhfD>(E*NNf8@ccMoM2?WT5}`7}j^k+J@=2Z}-*ChLfV<
zdd*sqUZc$hZcy-?^UbXFGehn}4(oYe-kw%u?O>;|W12ISkRs4DyMwud`$)wry6cbV
zYai}h3Ea3W27RClMoMSSBFS<@TWK3)wYUD3dSEFv?;S@sgNhi?%X=HFf7|7heh2Si
zFB4IIp`T_V|6G{<1G`Bg0Xh`Y0+(X=Kmo)?A+GtShSSqEU(*g>{_Iq06u?MfLpieA
zG5aFnl&7>ul(dHoy9j39tFfuo`uZ`%A2aLGI*3X*WgR<?hp(BWF?NC3>qG2kkg1FB
z!tAIv)G;AlbX)Upx%z;~2<~^*0NxG?!!IuZG#ShmKOMwD+3k2VFq>hoZoE?T9{+Ph
z*+ThxD$m{z8=U1R`@i|;iUP1RQqICtvuV@c<z0<J7%cI#sK@M+sBz%xm*-1+K-s30
z+`mi5@<9HUa9{1S(h<4xTt7QS*V9)KTC8t4j!nV@2dOxTXuD5bg%q^Daz;c}IMqgr
z{HY_VoRY@c(>1GM@04oy!1TcT!D<wbtIEm4bHjVTuz@Mp_&UvnBC(9muSvUqx0KlM
z{KV9vwh&>;o!$i!Qky9fv2Rf@$lx?(<@w(Du5)!xist?9AgqVc2`&hHC|6L<r)Krq
zx(p{+UtPl!M|EL#{TsACy`9RdcAa+j!@0rs(SVI*Sl_BM^kak`m(r^Lc{L&yK*2Cj
z@ayBeH6&=aq2(=<f~jRv#iW4#FA8(g<Gbu}jtp(K*->2ZhMkO;ywZ^0rhtLd`^M^e
zbEP|0hL66Wr5Puf`GA^^6KfV(#Hz;KU0?*g`aJj9PhUs*3)WpBnhk!2`VmgF>%#7u
zib@@7d!at);cxup9v+xP&BHq};>HIk>lT>HRBlkA(j_h3te#slMDO?lDhsbFJzSN1
zu(V|c$N4a`rB(##d+*W2{n-6S<$|v<hZfwtX9!~@v^Z$I%(Cu<?5Ll@qgnFKSSV3y
zsSo}IcN%`94^qa)qWHEA=TeO^5}`Rr;nN>#YLTl}pOiL1ifrO{u}70o$?3qgXR8XT
zChtgKAb4MdSw{4{|BKsb<FgF>k{A>V0AHS2>Qh@Nk96Q=sKJHdyHycNw7SbwxuPpQ
zY<o6(QLU`5UQ#`ZEfjH%TAUtGu+-q%FlH?<C>A38CK}peJ!U71eP!5C{IKmXwx#d#
zjrUa3&R1pac9i8-JAwH`YVZL=l;`-nOm|ucG>!I<dA5n+_W@#=gz$Yj+tO8Um!xsz
zgVb-|y?f*k?REVwo~v@bwMWrscECXb7j1D*3r_wpuxWnwN7~iueOFHiee;}}#K_<O
zqZFK7B^Ed40Y=_~>abaYc?E@Y`Hj)*RJq_kyf|nEU@wd&vm_Z>6nwk}@w8>mb^d*~
zVBr>KQq6_2t|=g_<^{+&vT+BQg_;kV6r=X_(3U*urmTm@?0?qHjA7?gNd|3X^bD?D
z<78XIS0T@`1p|u^u|tvtx0jX!vRN;Df7wm)(ZnbrWQQW=`ygp}Qu*v8vdpX?@ZQ$N
zvzv(4Xk!<e6u5|=KrSSSxVsqcNOK&klj@gi!-i|P)q%1)an#g4S<udhUQ6hhkehDk
z6Raxf5#dj8d91jNpDF3ULMmIGt8s?+UvtM-7~A>E1cc=moN0`*;}H9G6hwUV`1clf
z5aA-09!vi3J{!1CM8iFBx8RVPW(?dDNG5!ogCD+eUV3pnLpy8qsN90EG4jqa#W88d
zu|z^ULNXHlvqg5N>p@Lp(;@z<ZpAv_<(&2s>-XE9l5-1MqH11^>1Ifj&TE~q+CkNc
zwDG209Mjt=po@cJZxpFR*ZdM`@p?*VFZ9ieCi8r+B$=JwBf$}SulrZD37TdN!Tji^
z3|`7yoWgzc;CJOCYcn^lo__h6XEin*0c<nfS-Y-=X0|#@w5J?uSy|Psq8a^q7?wK8
zKiOBx@4Hb^>N_EbAw|irpBeE}Y#zcYYS|%=e)(Kds>=?s)45oqcThcWv4X~7;Sx$k
z$pbyf%vF_8;EH}+Rg7LlSq1qkEcH;7oNt?{>d+=1PxI3DD}9E;JKr^ccD(@S<s_6~
zHC!WrCI&6~Ot|Zuov`ihXw3+$JzBsin41kP(rNb498qOm6^bCGz<ZAG<8wT3(gIYM
zLQ90b3>{+Lk+}=Qo71L)gVTK7ic_j=5gcIrLIT@2oz_GFc(>H7pUeqLWN#R8Rp=p{
zbnU~XhV<4sUbABt>#-Pska^-#nzic@>MbwSMwF3{Q6$U08esf=9NnVW6fm3OJahp{
zGGHj|?0U2|aRfESq>&Zbf{pvR!MWvhhcruF0eFlN*3|g2AZ&KnPsyWmp>5X~ZNhg`
zo#e>K{F6->jlWOWVJ~JFK!N#@Y83w*Vw#{+*YzT&Q<YY-?oyvR?OyBoc3sx-LGh20
zH6HZ@=;c;+CXGCM{hQok@Sli#vF#>-G(Ko9V77-o#$bPAQniRY^m+d=HIHwrUvPI&
zJo9xg<RBZOt<L$la;)Di#YC^%Yh~=8fzwVBsDR2S0=>9oV<A$Guebac-zzU0a)q8K
zeQ#tposp*WK`L;vRzBEvq%{kuhhO!bqRI-7+eJ>5f3UoH(QJ<Mkvc7Nwxk6i6p4nA
zZ$gdQ-|^f+iX2pv@=|6ar$@CM$AVfS4F~Y%Bs5PAD6j4#t?Q}Z(iAZYu7`A#7W0x5
z*;)V%ZN)NkRzT?LyBJ#A<*J6-JkIoP*+p}EXjC(mr4_0}EKUHWCS2HQVf1LqSMpJJ
zNUd9sLHEY-p-do^nei*3AsyjY%ex-Et7od$QsB!*^b?W(qlyyE@*~wy;r=L`tB|bm
z=5b*A1HeNGdJvp!j&*@tTCI)c`G-_)ro@L3tYi2d9Uo8LTNsQ@kK~4rM|!_HbB=Ii
zP2Y6vI-&;l9+-yETM(SLTpX!?bdU(*#)sbw9q}Vyt>8(Dw!3&7`wjg7<YfZ7lOa5<
zre4=>_hNW_p4gCH`IMGiMd6QYX=m7&lMq@pZw4;VjStSbMaZ>5>;p!MzpcSH?CAVL
zO!1LT*m7MMDG770St6m+@a&F$qsOgUdzY*Ej_p3ePHDH*+i)|b2yTW86*Gg%EP&uv
z^`kYCHCj3?{&CM#T6KUk?5{e;K0NE((g{%2s;4o5dzUY@kHtf!?`Qnx9NBn-mhB-!
zuGHJ`lzyh-w4E@oRPBr>1v~bIP5l=VQtJYUa}t`2S=Za>4b>#K%-AAHRzR+3%MGT^
zK8EF;eTTk_P4&hEg8U%~R5u)$a&iJ38&st8D!4*l1Vl?>wlK;ljWmB|0!uH{YBnKu
zGsW>!Lta*%cc}7<umcA!+_r+}p%z?TyMzu?PDG;*v1R-=W2$TdvnPdMfKs}FdDMJu
zd3F!g;ceY_$q_Q~;<~M;u2;e9Kkl=G&$vc@i~j7q#cIT{?)&a|LFz!Hm(~K-65ij9
zyA2|CqJ#mSVzxayW$Gb?emThdS<HUpn67N*GUQk^HIETs=v<EwF%R)8F&C|T*K*nE
z(|#5RbbCY5oNNRX0x%Ja0j)v5Z3H=#K#rPWx^&^HFsXOtc)!^&FVoLu$wY+4Um3%)
z0cbLcwF`jb-V~6f08-n0@EhZZ7#J7mVBsk{5+JD3Cxpc@qH&8Hq#pJw>-0I{%ND>I
z9TzdrhV28EC5O?BM}9Ia900#=VEcKY9v1ICN=OGm3MIG@j)>jxP4QvL6sHsGPJHF=
zVR1#njo>qn0jV^|(32Se-`R`(&dZ1f3urK_8wRQ))_}P5Tw>&bQg7)?w}}=03Wq*S
z9RvLic7q)%0e8C_Gi%jUmT?LFAY1RAycGVMzr5bJo@QsuTX(q?%L9OWg5Bg}RFA1b
zV?V=!zYeTLDvr6G1iiP<_<qgimdcrJVkZd?U_u*0iSi=D#s(xmeK|*~=Cq#d??InU
z6BdQ^fFJA`bc;fyD^cUj>Sf6DcJqP$@uVY|gK*yyVXdWCg@FoaarGWc7g8jdrg4>k
z4td|g)@bJbAk*-IYst((^S8((ST&}tFIFDbFw$gAa#2%^<9l%Zv9;ohFT3V4d?~zi
znm;r!LvO8<=cXRXG4_<A8&#M-E8F2-P6rbzg8}8b=tzq@M_|7`WRKIp=t@M3Y$ijf
z?IZg}(XT|FbG_zbUE(Dj4&2Q>6jyE}=Vpw}WLTo5=|(IbZn_3x?5Zo{D!$yd>EeOf
zL2-+=xhLt{%jTj0AfcSLuAcV6BcR40-w<D^EA9h_$KH4MlOI3Qf*coBl3tBLf!5VS
zx;;+IZ#CSVuGPieE{(p%X8M>t7zb*onG(W#CcXucJOp8E>r`96E4Ul0=qZ`+{}klD
z=#-(pb+Mk1gy1pOp;omRd>5y><%ky)yB?HGzcXMeh_@e$2{ePHgHp^EJGp)i5*A{p
z6J(pxA?KBu9>3R~3g0X{Lj<kuqk)&x5-Wi!(FSU1cid17*>rK#VoJ&7@xlrZ2?xp`
zz?cAvzjkzd9lqh|y?mtPJl9EDgztH`O-Fe9UwZ#>)nz&rbK*mUIK}hf+v3N!yik1#
zS@OU-dK@=Fr-ApNy1Y$PUotT|pXNpl8rBpmdXoHd`>dsLr~k8y=AFKj*bbE9aFYi~
z!^&->$T!96VqS&kv)J`U*-V7}ZyQ<=<mw7_SGdSUoNl1RKWvg(*}UQP;~=4jQx;Pm
zp-KJ#HdctL|C}Vm;{^yLa$#@g!-LcLB@7-}WW6BTaHUf_Nm|)#0LD}BWTX>d=&_qH
zrl{}6A3NVP&U_&a1`#-^)fYA2yxX7l8Q+eeoI%8x>Oe3`SEnI5Mb{=zaBbZ_)l1>S
ziK2c>b44ol1E`Kyna%K<Q737^5V5RFE<LGDy1F-{s7IQSB&@o}&bE!~<N9B4<=uiq
zEWU+G#&PJLQbAja8mn)(<;(buye~c7`!gqt)i-r!`#!<;um;??0oNhv?#73E??<=x
zqiCaNuboc7FCiYGVc<4~i<-*#NEo>7O<;*=h*s)TCwAel34VR)4nJuItNJbjJtx{p
zc)ueo$nPIo^}Ff%dExfUM-el*X`S`RR5&0kV4_WBavZ3t4>TgD?@x}qd{)XLcAb(5
zjES;Eao2s)O#Uv`upcsNpSU$cR$N{?u+ib?1+T<*A*82U6+k6*T=99dfunxk&x(jF
zUG{Xk9jiYa5J{+k_Ei5u{tI7QWsG#cjlDuVF_SN3$Jis@ASl-8@!~lmqID7{yENhJ
z<jYdnzP>LlXBn-2fY#k7){gp2pje+a7O7ZxaTqFz5G;xE>M@qqcuhbyzIo2Wwql>k
zErlJNzitznp_cBR!ZqR`$rjFTCi#C*>e{*;sm=l-yq)suJvXvlyoE)BlwN<rwxC5&
zEoj+Q<ES#Xwi$s7<f7r9Pn7oNifU)btO11W@720BEs?wXkxiNu&z6aEMI6omC#gJX
zeB|hb=pWLn{}c>|?XZK?B|xfY8#3`EN&MlpZX#i?>{PXH-qXnH<o4goBx(eIoXSn4
z-J{yH8m7fJ*_NDq<Fz%g9q95^d(U#z15gte(ZO`3#jFB-)3<<VC-Q0MX$-Q2|Asxi
zN5$A5wk&9UUVYQ04OeIN2?(aWGY|_blgf<mdef@UQj?q8#g-6kj7V^g7(i)uBmSGP
zc<0`atS4VF*N1ZpdstumE(*;e4D7%i@&j7;4b^zsvX%Na;8$mP)jHUeqd`~;@cQ8l
zHE7+EMz;2hF&s$hPrS0$dpu$WJ~|zh!SOQuMmN>TffwtCH@P!>bGoHHv(M%&pR?wq
z<LA+zG$BQx?>8@VUWDxnI<4y}67OrHs&TiA{#NQX_pBRu-QCA1$b@E@$Qy%^%ETx`
z4Na@TTaoqUL|J~a66Q`AFvVW5`&(Ma>&<OY1#>zK8DsW9-{}6v3&J`Z?ug=HcE%*Y
zJlAgjHh{<CSwVEL6LNyNZ{`IGhK`{h0*M=BW%^!*H#MXaE9SFfcdy21+Vs~#VZ0Yn
z`6}!T(~+4F$z2g?z|4z0{t`-Tbv1CPyJ~~fCVOl<))CF*=@(p`JEDB`sz1w1r56O5
z@TSMW4`l&*VhGM37hA;;Vip_QmOMBmF7Z^ObYY&0?cNj_mXFniDnT$fT$XW5FYgbl
zhURp|x_IU8PYw=jLXo^ou!6?CMs>^YhloXxZ~ON>14X08)}9Az8WP2Na{LOD-I)al
zIaUvvVh2kMjJ!*!=}TfLM-|MAO9XtMvI`R4Lo(p*VFZ}LK=b?#+wu&x2gm}%2cDQU
zG~M?J;Y@pG{H5X&#%G7Cpsp@;bPq93{8+*H2Z={T*`s$`n~CZ9X%n0WS2pOM8Iv}p
zZaf9>_T)X+w?z<r2IP-)1K@9j?IioOx}21cEj_mte&aK)aV_uRrM@?btcU77z~*lJ
zx3Z%e6QfL7O*(w}y!L3_8(ecuAj1pfwXK7Fum_xE(Lt7&S1>`KS)o_SS^Knv?MRY{
z+$(gSzMGv=W;&v5N2ndi2Vv|Iw`}km^hWPXhK5&5_HVTu7Xt%R0x)Wzx)8JS0|SVl
zC_K9nnN=sw_HmS8b(I}c6v;sCVt7z`2kfAe4BL`#s1vVvM0fs(hKDtqpztt3xM!b0
zZ|Tww^KD(cw%=WPm~r<J*!z?ZU9+ILs&?$MMbc8C$~RD>!<14}re96=jkj9KyB%QB
z4BukP>aq(J--+*6LJKvW{P}f^?mAud;HgQEG*g(F26o}hQM5o1*#LAVBFO6_n@Rd2
zE(_uWllP|Pj(?abKjlb$%>EtSatM7h2NR3tzDd2@SW=Vl(D?Mp>O1VfYvWm3?MY9u
z1t7!HyoaRr(ERWkZ^;5HH3elphLOD5eQGkdO(?ce2D0A#J`eJREyXTfW1l~GntzX?
zLNxq2HU=jagdGPO;?7J`&YlS)@{t*hjz)e75QmU9_Gk~5kQNQA0gU*x36ZwotsQxs
zFIWDLvgrl9=Fe{A0gi*~?!iJukwtU>lC&Ip36+yfPIe*$R{b_o{#e?2D(%|UeQTLp
zc@Jki{2R_N(gH;I22=btziLMsQ#m^Lp3q&3s;cc$7ux`9^<-6OnMI6_QufDWFRl>1
zFYEvZGGOth(2<~%Lks$wii1g%t>=ZNddp#?#zP(-#cRQyFfIS^aL2hm#z9z$^Q;Dj
zq0p?>q!sx}gg>QjQjClM&A`l4jAZj6AN@h!Y!}GrtS47kkcZc=@fT+*z+nGo^l?yr
zsDO8$XG7{D8xVV*?br4;6n#ZMpNZUMEXHB_K>_!Y^MdrD)WoCepW!zV<FT#i_pA5Z
zK_4GK>pd6Pyn5leSvHZP?IKsd;MVppAkJ?f=YQ)<s#$?5o<}qh+;SSJD0n|~i{xEW
z|Co5etXmHXq6eW{%)zNR;k)%+iZVUiV0F*WL#y!iamU6(0Uw`TBOhB>MVvslAmBdU
zioWpsD0!U6Wti8f+w_-5VY|3CIrpB%etGjRxsmGAIFOCD)ro#jNFp^4`2T3^FZOUD
zD|o6Lv^OSy3$BYyt;h1b0_sW3J%V)8aq2qp`*YH&fa_seNO}y_bM79!AoV5F7l&;{
zLMobMb9f6ZMZZ~kcvVSWjH;C6r=(1kb>ug@@_@#c)`<&xm9CPaloDGI{iUZrw!fd_
zFFcFCiVD?#VJBm0?Sp`X2TQp5OBg_UNkMGxLw7;<$DgwNft$NReLlt4sft3&40*jM
zQeKnt*6;8T{M31EYwVm6tr=P^W%VOun@*C{KNRYF8@XI*!5x%69UN0oJM$6*taLC=
zKxwnEG<ide6CQhHlN@;d*^@X9_*c8>6mIinbuwbSi^U^C0?g@bzh)3Xb~e<;?%CmY
z<DD_yMfT(ImVKQcOISR%bSdu!M0`F}L+aVHvI$^nPM4|u*Eu7FwSNR1SHRh-agcyK
zkVlas*x8{tTn&N(K5z&a&>mODMBNq$ls5$nz9B;uf7oA<(y@|HY|MfIlVG<W2e$a9
z@PxZyMnj-qM1kF07=Ba5?*wbi*HqGTQoQV@0qRJB5nLKEwnYQ;-Q^D?n^48PP4u3t
zr^~_$r{G^Cj!jD)AG2~ZB;nA;HpD#5PHDOCY2{9Br3+8J+_NW$9Ez_c_a`#An9Vp3
zm$dPxvtm=?`eeX$e(|<*?p&Uua52gHKDu7Z4E7Nw2ps$gqD>NoP=Q&4%SZK#_dMgx
z@!XYiDR-Z2gTkLPj9IA1CnV;B;N48>-LV+9Ts@~#sUiFF&9?RY9wjh?zdFycX3yfg
z@+X_3L@az7zQcFflOgQMI-{BdUm6COX3g(55@;gRlUgt<80PLW;0Ig<04<RY%;qAR
z=FG-Ky=tl=m5|LTjWdNYno^;yZC3S)>YQ<TeKRldPZog06gIxIL(PBMB~E)F$T`?a
zx5M$7!R!aFVxxhAz>b8<YDkL<S~BtPZLq$zy|jGW_YL5=9OU8EKDV1qL+lsOeR$*I
zaNH~6LXx5Qv;A||>#cOEVBO|eWdP;HWTW~hi!e`wbjxAHI;FCiz~qWHef_E_kVqik
zqePI32Av_#uJD@$!ClDo59w74>+Ei=sqfW1u6+p?J1)w$_&bsEka}6^EuwdY@@a(&
zSz(fg(%SL&L+5B@NQOe&aglTB+PQ=1dNew~9n2XK*9D|R+ZK_Pg<w%T3-@H6K_5ZN
zx;XLASQ{hjmRj*RNP>ipGga}9Jx^(F^en$pXRU#O=l_Z}O#J;?>WR=jYC#K}cSU1A
z)$_sf#&Q4QCce9;wbhO(U6ZbyfhWoDg)qnfcwEhrt&a?P6MO5N$P&oeDu30)O+u4L
zfQ#6X5Bf<Jp@Y3ap)R*-AFTwZEt+*Lxe!3}>x;RMmT20b)TL~YEh9pctIV^kQp~!Q
zj!FT{XG{%yAcs~D!I8Hq*0bb;DwboPzm5wmZ21oBy83jrhuhYc6uVkQT)*?;hJmnb
z8MgE#-(RRVz)O-5^fJ<T0F1*_ULy;=-;i_yZr^T9GvAS(jWnX$MINT!tV9-h%j*pV
z30TgM)I+Z%!GWjJCa+rIO<o>Vj^X}Y18%ldby@{vONkGT?Ra4@bhiw&w}-DhpQ2wk
z2g55CZvbWYYGmfD+C}*GvU?P6M}RHL#UiY(4K{I-Skh_H7u$p=S@KK0D1N9-F7%rE
zpc?ycW`|m7#gY7~xS|(MGY`Ww6?bz`^=3?UnawCpNPKOv>jD&8qF>-2{IT%B_`QwK
zR{_yrLoU>SO!)1=O?6>?3Phjjq1XWjv>B^Z<jlJ&>d$0}_1ZD=ir8@F2DkqqK%k01
zmZAVECc2_=GnK@ndoZzEhU?-{Lt!&m44NJOqYfAZwbdeaOw|8LqM^zF-3w?5#)3g%
zWmWWBr14iwlbx~3A`sT@e%Atw`Z*BX?z8z9|9j$fx(mZ#gkcv%+OFog6(n*)W43x<
zJUfTK(z{LUVI~guW1-3YPd3*+*8Pzk4RCLRdp0wi3d-Reph;0?3xH96*ebFTUGXfw
zr(jing9ogBJ_#_6MJMWXq9+I2gqTu@*ShDqXRSJfAVtTtNX?R|s}Rt|Hy`qPRG)ZC
z{Oj1Vc2_oZvRz(=YS7vzYVH1d0uhRKe*Xy6r`iX(rR=-pNw`v<ZEhAXj4r!?u6ArQ
z>W#2VREJB=KQj3$5fL!gS(cST?TLI7uwW*y3f?7vOk<+0XFJ|QF4SEc-PrxY3x<yq
za-K-G+b(|E6(Z7$dg8c7AYz1lP`jzN@s%aE%BinQ&M7Bad}+|e{K@8oUEly7svo(#
z5Ha%N)Lh|bxN40`rCVW+CoWV$6=><dUSWlq(T163d=_C(+d-}!kXl^6OGPGcOiNqH
z7FIYhkG{k|@_|w8)1vhQIg>iiPgkHA0rjvq{oEsu9lx1^!xR(8*hO{t8Vn|id#j-%
zg~yyqu83l~{M88?t`X;`@)$AN(+Gjh!m<y__u6);?P7{LVnTap(NG9w%Zc@cs+y$B
zR6O5_>bw0_1_msGTI?o4y_1AG5E)6$tNmDh_4|aAPvFF)^Qt<3K;iu(BFDb5{la<T
zW?mrHJP880Wl78^ju-Bg;@0N|{hZCSI#=(~D3j;*%ju?F!{UuU_Au0GIy5O}QL_YO
zK)XWWN_z_pog3OTay3zWF?#dbK%KEHFB|=4VhaV4x&hQ=hr@C4o2T`n^iQ0)I+<~4
zW@c0DBgmBL3vDBDCm3^-)NDNn%CS`$3GWWr4=Zt|_(iaI&FQA|8P@w4W|hrZ+9x#{
zPFq^tdh$}JxD@+j0|Om=y8I_w>9H$KNThpa%8(@l%fNqN7=zgyI!XNIQmdQtJNyAy
z2T~QD&eqTE`#-{^{GZ}v{(ty+Y%}zp1UHmn3-}X*MA_ZsOh?Ci+Wh#>g}=9CXCCVK
z4?ejA#<n`cPOJR7H@tZu25sz+t;Wvw>64^r_1{<3j^7A%A?;W(#(s#s46182;6auy
z6rS$guN<7gy&o3NwxfcR1cB)U7?C>HVGCU5Tz`Z(BNObuWx0fLDB3Edd<ht?n2K=B
zhWc2%dDOm6XPJeRH{;$13bb-kw8azMB37otu%eI%1=#_w?&|1~<2#q?cVdk-_pP}p
zQ`KkgZfftMr5DFW4{OgXMTRAeVPAsp)04TJaA8ztIqO#2sx#5$wka7H6jZnnk!{mP
zM60%G?pDJyvvz|T7e{3`4_Mi|+Wx0xNijD+(?OCQFcZI#N$-<XUGH#u6)pXyvebe7
z(p|P|_N8Fs{=Io8E-uQ*T=Dl;cDfb^pvT-prBX_mGZ2==>p-^%1(1Y)zp)NUi0kTP
zhv?7)Rh!30lF!Zd3y(SedfL$ksr|{;nU-uQ@GZU6_xysmf%^L!gBM<Y|2WL(iZSfx
z1i`wvR9osKkfQthmAIs+S_c-5S&Qt-uuQ%^D;t>21Qw<-w<5V=py~3|IAYUi?~C;v
z`-r<N8=JKUzM7%FmPX@aKCaQ6uqV+lv?MD4z**Tm@z*nozbN(yR$r2TyaR2Pr5~_B
z3#PqGsWg{vj~%#l4HXlu`d1!3yCEe^%!RJq_jO-XhR{(oEb9Y<db#vp`Nz)>Ykg}}
zO1XhpdB<r#|8kGUvR^rR*rh{=8EA(4Qu5a}LG7=P<tlmJx-V{Iz3W!pnmB4Vm-z0D
z@QG4hZnNDQ|A2n}k0|MX`!n|49Vs92JM_?EuzNm`*&ZUs8{Hij^_510EC1rZxo{ny
z+SO#6D`oX}@1G}wvsos3ewW}eSzhPg6zYHa;$8Qq)oC>NI(e94CE;H$bxU{VUpP@Z
zFA7ffzx>fZ|M4EpQ@x(Y>JIoD*RFNX%g6#cq!y_UQnzYH#KA?I?1<&aHk$KS=W```
zFF5wN_?Uk0!6!2Pv7KPm+*-E90&rTsh-%cDeRl$rpm)<iI<d8}u(n?5X)tX1xMY_u
z`%W1!tJzHJLf&tdqkfE0680Ho1-cx*_f?}lri73-`jHElM;OP_rl2Qu4Qu5W_wGIo
zH?H8HaC`NZDs-_x(HmRghnuO%ezow^lj5wmax8R=lTim~`A#b<#CJ7Wq+gd}_B?Lv
zQQYmo@mgs^jIn_tU?kp}>Y#??)kg0=<eWHlQKP0h7K$mo41cCK!RiNvJ2DY1SM8D}
z)TSga(_g{WeSIiasY-LpwN5t8=NH@^GgbhGD%D#lhf92U-Zn`g4QmVDoGo&AEc*=q
z;igmzNOCO(BV>5OzEC*LJ>hcG!+hKb_GTu4{a%>GBv84Ur{5yDLQkHh08{Kfp?TVp
zU-81|0+Qc{d)k3bWHGe-Uz^~6zM_6V9;%-3D*};jqf#yvbIBq*iB%SCdJgXA-{k+z
zt!fG4V%W?V1vEt2olW!yrTDshd5^P{lQ|e~djEq(@&D;9>r&A4850=`9PY{JG4jhW
z8?8?=moN*CWj$~o4eus6g|GAJQf|k!X_L>e(P;P!WExDk!c>ECjfzOBNRq0t8GW|0
zbT1C}vO=)1!4I!vP?otfeD|MA&A$+>mCJ`~LxlS{Q$z>e-AxzdP47=+rNeJv5|2h+
zV`3s<BsIaAQEwk|?oZPBWV9Bd_akZnkRUcOwW!ZKt`B7=5Yy|+gP9<;^Pfq1;E{bD
z&JeXHDKJc!j~O>YwgVSOvBj<1E&Pk&C<jOm3~o+>m8Y!hh1{SNf5G`-Kl1bNAI6C#
z8!9+=+)R~YC!Xdxu-`s50NX0mZ*c%Vh&zIFPf_xPRik6B9g@`8xN{9Kfj{+gZ@?*A
zt@<o>ihdli=hCIR!R6T+3%BS)_gn@3dIk}{Rj|&cqHVLd@fYa5%R(W<u()9}p`3l%
zitZF_ZCMBIR=K1v;I<-C59dfm99~}rUHcg}#vW#FWK;m93&7QpHvT_l?1uk%^7I)0
z%Qc}I^qFv#d?xnQ7{VPnQ9g^|sQoFx^xgKp@?vz}T55vY{8O=#3s-`^kMDUMsYyxs
zGPjj&`}(*XqX=-{uAy6e;XnSMc+x@VZ0-Zq(W028O(*HcyjUvo;-`|FNB8?)`siQ#
zr6*<4=-qyIJ=#--cooZ1lMCH|-VnzoqutTb2C!;mmmgm%Nl_(7Ch6npRIOXd2J2&w
zya4vxYBwGMqF?+<-Fyu#o_s6JJmYMOxRjp3qolObHj^BSm(y{}nMFcMTV06V^Q)y%
zX=g@-MFP!>r*+@|2x1QoMAW+pP;RDxv(y%BD0r`=rFbDL;BA08Qquerhbl)#K=-Q4
zU3Nxqs1PY64d;oG*H}p^B|s5ZhklG}G~KT_YI&jmw}0~TN%|Eh)(D)>eYnVpY;03P
zlC$h}QQFJ4FG})PGiXm@KCr&fq!VJvrU#)A+eE^J%117XS`+P}-%<mA)fzsT9Bn=h
z{>m>h{3E1JKjod(c~RFSG|31u?_3dVgS151>FbDGIEX)<Y&#EK9*jIf(Hj~M&A8z>
zyeUR3FSxen)N;tdA2jU1h~F0>tR7>IHRa4IA|WP1I_nzVy#FqfE+v5Xi7wk=H<S94
zjT`O>OGMY`*Xe|Y*fortJRM2CxSi`#_TsenQ8QzFKWNoBk<br6o18xWhja@Xt!wUm
zGZVbQ>rhooL)~)EE2q!^vfn1(&4Z=RWol1~c^h-F7lq2BqJRZ1zDo%BJ0q8S-W*09
z8mu^cU1`szkxFoF@Wz#+$FY@QNSckV>w2L3C!35nnvW6q+O(w5=bXmI<Yd-%IYRaT
za%D!|7M0EFCva`xQVBep1fIpq1K<41$0l_Xd90(q?6o@=fOrN29sUd$YK!VJk>^JI
zo330ZF7;Sv-AKQ_Ui&C2XD}LmeI!<G{||X{v4q()=>?b|<KXMiB-f_IHk)^G#o6A^
z)Ffs3U;Ln$^!j*bVLgf`TzXV<l(dxmRJ7{j*f5>Aqu>SJ6XcH^zRTn$W<9^t&wQEl
z@#0a!lKX+P?y&Q=Y>{Pj(1*61F!rEZEp27{r-TK(3>DYDC}1i+foGFfdbG)JbuHkk
zXN%^T%d8NY$CFu<?{#ycrSd1iXQkibNJN>~@m-B}kP2ZLhoqlm_$JOdn|+t=zvMR6
zx%9!`>F6%auH)G6CaTu&`>U5C1Pn)I)Yb3uh&AmI7Ph*!n++us3j9lysrc5NuqFUk
zIfxnP1{)-=c}jh9&BcFXNm0AnG2@ve;Ru<JHm%KyLxfwkL%RwC2h!`ih={?Tk=9_M
zrO$l4fk=XJ>7_|6)QZzy9HKC6Bew!NIi^k|+9<lO%0E*bj5e|xv!jg;p?&%|s|~uv
zb7nQP95mmKG+(his&sZOFvd>yW^)Pl7byDRP+uv77h)38d&hNdMHIfTPYM<*coL&^
zoxuJOJOAt3n%s#&nVc}rcW3hl@cG>NFNL~+huUx&EY(zs9E^aTUfoba_gp~eL<%v>
zu_$&{A4-d|ny{i2+aa&y5wG6yW1_QsJ^e+_<8*`#IQmH3k5I@H*xN2)n9}OA7?5OA
zj_q2VP;6xh5A&(y=%$(UsIm48Hw)o^XV#`s@GY>17?E;>Q$ItFBqTif*+%bYd2VJI
zY%BUW)`OXhzKIZl+GvH8uJ}lcg2imk;P+^+X^RliN8+jCm(>DO3z8jI5pj{q1(2#m
z?ciMv!%3SV^03lv=XPOB?w4{tt@DOHJ_#3HO^1Dy&oJMfX!aMnfRpNALz}|tu&wH8
zG3e^9#OfkVxvYF6pR7cliOMs}4pu{AI}zZvdkPJs6j|Svh>g$DgADbf?@k4-SKmMi
ztZ0wS;V4&N_W%RoQKV@<>wx!Y0onO%xp;x;@uPb=-6?<g{vbfg2Ay$00({zezOL&1
z0!t~0ZgN1Cj~4GSMtB5Z;W7)c-J63><9&gOoT-u%mN$>Fu>+0udvI&_Mc(p|mw^1-
zZ_D$*N$%mLQ-u%4%+OrDVL(g0I`j2~F4S8JEmCjL4dU}V8m;r*TdR5O4<|J{GRvEf
zz`KMdI69gSUJkDP(6C0HJ>IIo_Kaik19k#fquaG8QXFYs6z0d_LwY8c<j!~p#2-`>
zOhiTZa}Rx6lCIL4c%yB-tiEQ&;^hICr1}W5rq_zN+B$8s>>Si6iQH_wx%-!iwC<wT
zum~A(1Kn~OW!OW3)BHw$ynjO8^t+p<{e`68W}$e{uZ_Zm6E@{w92_<^iFE5{9Q6M>
z=;L_C|0YX0>GnDdVB@bT3A8BaDc#05O(MR!Z2DSNrha}vQ;Q62wL3O(M`4b&2L#Vd
z5Y8BcQ8qH})lVwZCAyeY`~>bs^InD`Wvd0TYXK!mA;*0zs>y3OvDUHID({Keksm>-
zSiK2g24YJUBBI;V#?v~Fy7&fXC&X{6<y!j%$#0~dIC>`~T1IRKY{H!L;Fu@5MxJY~
zfdZ)VWEq49Pl?aqUV#CQ6}Z_RHvBlKClfxw1i(5dpobLl&Aw&eXp1~h*`Q`TrZRCx
z?A8r|vCpfBZ=)XzmtebDY6S0MB6$-l<fFl7%dU1Ry_QEl54akHy#>;<)V5Yr*sxyS
zxuDv2m2q=fR{O7C+84e>s@DXo!+VB5Bte8p-4|{y@{&H)H|_C9g6Z~Emfji&IgpvP
zox8^pn*S~^v~pG(INQ_BAHG|XwQwo@!I2x$&)tTotpi@R9f<0LhyK+c1y`bkK4q<3
zocJbwXluW`a%R0!Ft=C>LjVxsG*|<{ikB$Al721>>MIWGS$SQ5I$SpMgZBfS3)RO>
zOM;jM3kK)Ib_8>KqTXQ7eiA$?^cxWC<u)5?(kz%*2)E5OL}7elj`@*i_aqkdvjZuR
z3KR#9Wev1MA=qTv8wSN!$y)ys8(4AnwlpLLCz6c$=vPCRh<&X{xky#KQvI;52IAaP
zI~$p)g!j)MY{!Sq$_b!PGDD3M8D-#vG@-ss_~#Ta0B(Y@PO*iz>T{awMZVK92fzJV
zf5yHHmfX9MpySLS%)sb^<LLPf?81B8b9gjn&E+?>=lY~3Q~n84|6kOz-0L_x8r?k<
z+Qn+b0Q2t+m%m^JDqAEhUC4m-@C;+&KbU7m{B@oQoJ>)1zYESwA>!2O$sW$nO7RYA
z^WkEOoG1rnQTE?EJxs`tCCwUgJ2rH9*veg*kvZAPwM$W}HIaD$d?qc3HP?>e0{$%Y
ztZiK*C}TR1A^5gy6Pp6|x3?39-tZcS9u@fDTgBYjVE|~vMc^31a0f4BQ&36r^7Den
zJGZOZcbdP7S*x<$vBq|TfNQ`mc5$twx)o4*h&9_o*WLvRzApP@9)`$&XqZyJ|3@IA
z6ei47!Q#d%_WN0^V3i9iCr=FSX$ii_{WVll1;^b4b{<aPb}96bV+kdaUV*L9QN~Xc
zJoZRCe5MkF_CRp@afUs!M9&*elo?zN@*3~Udp%vn%#*~BGP$|+HVo0g%FVE*GWY}K
zOYejxnd>+s*l`K}33lI5vjr9xE)=)nqEwaoeVm_eHno=i^l-$QZMZTU8dO=b@ITN=
zh^i^e)%F*#3|AA;rtGYo2H((1=G><aZ`5aa&0I3t)9%59z4Ky~9kHzO8F5&5>7kB@
z<yTRh32%&@7?U%!s<i=;(qx3rU%H%5@P0a=tJAXovxg!V?)hr6o0vQgR6_bFthkn6
zHC&+>XQH@Ie_$J6{KZJYC7S7>{$;%s?dyRhgcDUQ+8cL4{Mx!EV3slCQR<8cXr6U|
zl2V>K;eKMTE?z^5o1<2!C@Pk|YjMqo@E+aPrj(+7ve-iK*KuPtY(C&_`BOvUH5n45
zD7zCQV)o~UR^%sw>>tTlb8|1M&!>UggJOqB#)wDq6@_BkMuYEo?L0o=8=SF_o`|_n
zmk4<16A`R)5+oQieg(JT>$u{{5c!IM+zIi(J@Ts-yqwLP&K4<2g@3|~Z`o3CNeYQ_
z&Qe9er*9m2sZ+_J6fRvt4`GdsV&12<)fc@_RsWMs+@~o2UCm9BHrLs-R5Olan+nY@
ze;b;r)h*{xCugx;Mt!2*^5O=~Qgzj-qy1eGQj)ggemU=XJ_FnAJxk5HP(Cn9*tVVi
zRR%2NzU%MX!|omvp+I5lD~b{(q9G+@7joZ1Kf+o(p{PrHWetnGauadWHCLK$!{YIv
zD}X3l%C{O4R%$-Un>5+YXgsOL-|s?}f?aGYkNT4hAQd>_-+2#mD0#0s!<+?Y)CYd`
z#_oX?>9A)JDgy|82;K4GC}c$1_%MV-GK|1Y&YXH2nj08z&v<}cSo=NWdk^Q+P20`5
z0UUjFlf<es5tUiKzmnOo+Q-h{HYw%U(4_G)i!mVFP0^#wCj$%zc=VEFTu&&$wbWZb
zTltNtG(X0EUWF7)ODRwf$s+S~=kd4XEAFb%C!0^p*X+#xui^RsTRejQSG?{Wj3`Um
z3nyvJ@iz2i`_ZTZs}Y5@lH`4Y?3{b8BJa7nin3Fb`xwGq#IqBh8q=Id4gf+6Mla<I
zDy%ok1yn3ZiqP`<t6I!VIj`K&FWeJHxSXw@g(zii^KLAD1$@T0#>y`VzS6nP7IC?@
zx`VqlYA)T#74l&ldev-G4Hokv)Hi&cmbpaIN>le$g_R*y4yBnpmz&GqHzl2~idp0$
z@KfKGj&(Z3n*;ry-G6C_qi-#)Ha;DrkX)tpiDP&;xbC)Ft8!iTagvF=3V#Vzl>HaF
z8n(PdgiqMFP@s^GTCzr=NYXPm?URbpW{Q|VgoYQ6VuifF<6o_Lq{h$gLd|!XW$jmQ
z+)cCveOJTPfF8x0f^EfdGjzCt&2fRte8VMp@E-ObI~if{l%U7|W>)#%rm=Jvr+!z-
za&#)VkQK6@&BVHRa;hzCmmTM8V{L(MrCU-?;lGE<)qii3;Mj`>HYKlM9L=JklMWQj
zLu0!D&+>&-#fBIYQ}10`=Rf&7w9KS^`xrPw=tT2IhEvX(YEV4rKF+d3*1OJ8s;$Ch
zk4}7&7hnq!{rl&(SIIUC&9-*^BgOK{R5aVBN)9qe;D`F*eT7s5tCZM-3=1}!|H}Ef
z(1#QowiO*v8mQZ?bp8{ItMFB{sTTzLXih4mLG^XWacHKr8=}e#w5P9RU7Q=P2yi<X
z27c=UxdxW<J!rBhb7>V~M7%wHiKX-lXIA=BDBV;RG)|hPdqM?wjCwxuLy4rrk`vp{
zc3&s(PT@zwg%UOh{NC>tb{rn-i=VF;j_2fki4x2IZAyIJnJ|d}2AB1Tvps%kKacs9
z&j<APb8cL4v!>3{N<fg9epj*e?DQY58bJ~_>ek9{<OFc2u5bYOyFtCJnVU5M1IQkB
z-_?0jF8r?%J)=;8uaCMt4<Ae3%hSU?mAEDSFJzy8JLmj&-<PU3*VbgK71t+9c_>+C
z9J?GD^vT30{J-|ke>a-{cTfJ`{JnZKv`rkPI<^qRy}hNLl)TI2HJb5fJs}J`US10G
z)fL)Owmi9^2x_)UH_fMEh8vTLLZkWEKxU1pU>^lY{i*(OG^@HH`Gt~q=}-9%pJktd
z&s@Rs|Lq7K7X}E_;e1}e>|rjUC|f*{w79lnUw`e%`Z>ZNP9Rc(sByCE#o@t7qoPwc
zhlAAb*Tx*x$#TDk7$mgdc1;2q7J`TsoHWy5b-44MiFV)0;adNbE$U{pk>PJt=dWMC
z{l#M1y)@slzUu3AYL6Ozwa-jV`dJUCbpW}!f&RbPd-Hgx-}hg9Bne6O-BgsyRw6=%
z7E9hJg{)JNZDNw_V@71(LMUWNwu(vC&`kCuyDTFHQwYP1b(rOS?%tp8`7Y;t&tK>K
z_xt_fA6{OM*SzMsulu^M>$w~<p1~0no5de~kvAH?8{@kf$jiM1Zvtw0Ek*QpAO+Ul
zTE>ybc94|ombJNE;>q@BF8<!u(&wgS)U>MYx}}sVM}?qhCI?}f`2veN@#_H0G+R{C
zfsE7_qdC1>(w(yF@Y$2Q<L52i&xk+_mO}oFWpKQ#x+X*MXxKfH?BZFIqsh7kMJWu6
zH?WtX<>(Mh^BugF1KoCd3p*|CY2&bZ$ULXi8t?JyV#Pb-0-(}C4#+COSH5+yYw!MX
zRo*ca@F&sI!~|#zTD+FiBgU-eudt)n;eA=lm#CSTM0j;p<c9x~o;zH#PH-~)wvBYk
zqSEfEs+#1EUOo?_v4SKQ!-f1mBO};xFmJ%F0lF4z=#OBBE_;8;%ynNJy|2wd-Fw(>
zwbT4-dCDL_C;S)u5S+a~r+hz!8CW#`!aW%kb_mEW-UJ_$vkQaZ=hri2XfZF>-uu!0
z%IU92R#99mZr*rit4uY4Oc0oAn>&Am@EBC<-r|}pY4mVNiw6+;YcnhHe(e$H@q)d6
z`LpdhX2yiL)Sj&#HlMa7x|@EqpL>(1VDZ|L)P;-<?nhdO-o=N-;=ZZ@8abHD{0Qn@
zn(d6N9@muSt8f19t#ae|k5}OekbuKNh)uKR2G0TJ$=Bn8kzCat#wA~sY&_%}qF!Zd
zW*(Pt^*``B<d6sFP81zO(`_g5_!qvR8$08vb45wzzW0tKeUVKHi=u+TAdQA>HbZ*i
znq6jdP8(!~_Ro>!3L-hOSvvF@#(t(0y-Q(KhxDX;b{Kxl=iPY03CplI2AVWjJ1CTX
zyzlf2rZU=#SKJjkOBSyyo8(2JEv`U916u%Yy$&4kO!-S<3!11=y^u@*ibJMiuiZ?K
zX9xqHO&ic)L){$ccsrvK(gz=k`&?g@{&0?f6-Xde!rMs^4|`aI8Qn5U!QQUm;zGyI
z;eo1e!<BOBe2j~9!d57)Jth{VS!OQrxHHFX5PICT@6=5iVDoI$z#o&SH*tcdvuea8
z=Ty?eiSdbv(&atTqKsgw%PjgOEo&wvhUrptZs~&6*j2*nK?w}WV00#nDbi>fCcRZd
z)3A6~Z1U>4k8F-_J|Dc2K>6;=P@o6e4=n@R=eLfFT1Q>{PU~g_Jgz=uEGdb51xyz<
zvsYO>nV1|5&AJt+USsaUAn0Kd!1da;1!jp6Yq4cvFI-{@Qo3y>hXY|nVL(*yU=aN}
zXu*Thyc<jt@&!PXC1wOc*f?&-_Wm2vZUaliOjk<b5}Op?F4#KrePHMp-rwpC);@3=
z6OT^@7we(Gs?X2(8}bWyn}J*XuQzP(9F858z7~Z_-@;xH#l+10fb80CWHC-J2?2-W
zV0v)b)j3nuY6nldzK7TR;7VFd!HJnmC!xfU@Gl((Fs5}_5q0Qu0E~~xksoCCms=Rj
zs4DsEK6KHc+hA~~0%R6RYkH`&05R_0$4(M$$23bbPiUqj?~dTA6t8};nfyJ6v<E!t
z|9Tl%V4y*Kj1!RBf`Dr70@47A(*XPQ1FQ^Q`WDKGZaItOSu~xq_d1st<%@le0&gdU
z1KumtRgMy+zy5IlT=sm-YU`UiE+=Vu@OfzP7+|x7jF6wHqa2tzULA#RpDj&dCKvuR
zH5{`Yn*k)HKie|<upRRn-F5~iWM`o&wD$FQYIEP}MF^R_+Y<!DRvgF|Q8KI-O;;4O
zU~VItvFz3oJmy!yVJtIHa|TW|T1`xvO?i?%8ouJnnvZc8Hf&<{rk_32`Ti&#1Cnwl
z(*q;Sv>MRk^wiEVAIKL^qrd`Y>XcM3gaa`l!g44zZ;2d35uI`3zr!%EZ!(+Ml!q-W
zEV{zVF-G=2*nR0G_G847EN^e2Rw0c4O$hf-rFA}_o<P;J|BY9mXOuc<*jk;g9QrB0
z+S00O%;fQfmkx9M(Wd08pea)j(=WNd!QAcm=d-Z0treJ4bhRx@R{q46E4@Y_i9C2`
z#4UOd;aZoAD*OiVSv)U+tH(^&5Y3Vh%^{g1Yu}FZPgZ4Vu9YlvQw3Tu)ma^;deXCT
zb9J^KvKPgkJPhDS9S-T|9)?kOdji={x?w#8`Wx7P4*C@WbsiH1-4G%$35-*4x?g8Y
z$}SI_4QFLQf~-JcgYf_S8Ot@E`r{t>S0<A_Njrj8LtUo{B%)j8UaD3GY(H$4Xo+zF
zmgE4Y@Lyy58fF^AJ1=bESSk9?4L$!b`mUU-WN3EXQz#m;Xk2r#LN6SyRU-G%m1AO^
z*S8|y4<3k0``U7HZ}3cDJ{aIx!Q8Op*j8*6_mkCX0sbD1tvbJiIH^0m(`ObDYdAKa
zXh9B~XvKenTRh~E;5@4hN;Z)D2;$}%b!q}DjhN8ixZ4+SK@#){E8&?K>J_Mvk{x~2
z@XL?a&MtVKE4X36S8XBqn2Y+aj%&@L<lrw(EmSo*KDw=q^T)}u-f!OP2l_v8S<4)U
zJks$3XdQsjU80JJ(r2a*83}A{X0cPQvU8MnGU=j6a>yl*738^U80TP#z&7<ikRce|
zzae$o%$Hh;m?iLV{n>;ogy{s>8WMd9tipB7!ZTbic7nmS$mxJ>f=&StFnvX$Ucv5Y
z5kxU#!Gw*g*SHs~|0(Eg0s*@p!|x%LSsu`=^3MC`$ETH_iSf7ohMa;hK~9P&L!=_$
zHEyMqUlh&-bl)F?J$kDJEf4zxnll}kLcn9~^S-@4+scEv_{9Qtl}TZ}(m!fUb0G&N
zG95V2PIwWYjh>EPj;e=dph2c@K8Ba*nnsQnU-0ojz0foD-TL(IblVZvy};Je=R=d(
z>`VD(;$OSPb)^o!vr&8vW(hIGH%f}LDQ!T?P%V5DBf3--Eaj8KlWDXGa+&aB`sn%&
z<0gAlPK{N9KH{`EnpG#JbE-1F^=h=0(21(ghh2QP^+AhllnmhgrvkxHb#I@ZO?Fe>
zu`@F6X07|=hXRy7GkyQ8p8>sEIYFxR_H%CjH_wQc15$WPx*)jzEBzC#%J2h?d(3--
zpLc!*U#OPxz<B$Kt2=jjSSXdj-d{e`xG!R<UH@DYAsu=lQS40_DB7ZXAgEttu>H+<
zAVoT;WA?vw4xlIqk@RnzKQc9G@OVA*QhF==(}+JJx?x!Dn5SFDZk2F4oIn!z<A1h?
zEz`r_lM?x=fsp9u8>o@y16#@bjBAAv)pfPeT0tFr`>m!1Wv_aqX?N_2dI*5l^}ywI
z&~h0&b6}kTI!pam!VKZ6nC5e?b}(1a;_zlmgu}MK;7S)bcZyjhK$lkk2xiP@3CMHT
zpuyJBtVj6ShnL2TfwmG_3OxkaPuN3*Wi5Xo7i>>>2>Ox#2?7|dpA7J_MbE>z6|(J_
zU{yxsqDfNGWfgGd399TyrcPMS^O~fK`;-(ua7ONzUx|`0)dNz`D?nx^fGI=Q4OWRR
zpY1k#X(Jz}tO9?Out|Yy1rgYMm#Ab5yZnBpO8KHVUyHe7CXi^i)7;6SLbvV!NQo6<
zTWm^bD39KuWm)DWq;7!ku-ZS_ca;H9)J#Fko5S+SmXbmJo(E4mtLTVg^*^Vb;LM<v
zFvWnrZvI;94C%l^tr(@H`$3vU;Tp>J3ml5^Op##^5Nk5F6E}qe?hm9*9eJ5rcK)#Y
z6yc6O6ugxaY;L-d_Jr+V2~Bf2=qREr?@a2CHz3C#=}M$lEDuv@6fNU1B|cET5#%~F
z_wcEz7FW8YV%d^`7+8~_r5RODR<o?U#GE3yDA79E`#&DVwd2Ig^lF~n`t!Sg70-VP
z=>Pxc|MnduhsKt>sGA>ci+p?jb3V_BQWa!5E-S2>)=QX5iWC?u0jIT7d~cT!qWpe(
z4^A$)-Y(qC=UFH=K58!2bYE7&RXsP-^50H$C(}%xcJ#^by&|dH3Ax6(@MfeTqi74=
z;|E6Dn?0?VQ&s=b1r&8&x>oz)gsQ51_{Wx=WO3ME!_(1_<SUSWRQg3_wgv4woIS-9
z*z38x(W5u3@b?{@E}h><0(XJb{BL~*`P0@4xg1@ZTonc0VU5e%f_>F9O?CV!6({zz
z|C6(=9bQ`X_F5;0mflHiP)_E6J<jL5dE@%Xl>_f2EiWFtq7%-$-bWN>n*fx?vNQ~M
z!}J4HAP>u(qxdsmf4C2{jX8kvMr#XfO~VtuxejGNQQ8fqrZD3})L6lW!0}{+@Z54g
zFryhL!i;{y06zckRdDgi1;XYpZgzq_Wc1aWNw3TR)iOW-0|zF>jCn2RZRAt$87wq^
z|9GI}a6f11e|t=SJGm&<5EA{dK7P{jl6(}j9CGW`tGb5h-+`A-T_sEz?Y&w34;qM0
zV1SI=Y^sNu#fl)NIH|d~8FSuV_nOSP@%Twcj@Gv_WY4W-#75+nV14H5mJ{{z`j;U~
zm1jO{=eyiashgt((Sx^foobSMPGhRwKrZY-Q`pU3wp-RzIis_8B%#~DgfaiWe@Fn-
z|MT!~vg~Uz5}wpIMsr^fGu?w7d1rj|ugB!u3n3s#{*w@JjzWTvc)<R*U2fWY3TlQ`
zkac0%vnRJU*M6LXoq|;U`?vc)zg1KJp7X_PQJWvL7!lTezp^2p{{0e>u41zwqN%J)
zk`aJdi9mLfgQJ?%<o8@W>5}5*UGqaLHBRN1X(4;#z6APiPU%U6bof5Y%pU?^3U!1%
z4y9T_V}X>eX+_W_gL))ykWa@7;+)uho6x6;*PB|?&U6~W-w{5@Bq=VpQ+p;8syqUI
zR>`EVjjlHXcEQyW5G$IXHB~r8C;;b(y`rq=`jW0LY=r<4;5US@{uOzM@~b1AISJ46
zp&J7&e?xR2RadNQzALZ3SE(q#vOm@%f7z5mUi(D<VX}r1rt_V}N7Mk>V>R}}#P6&d
zmq&dw9>qR2e)dve8nG$O<}ogr^dwZWQ@AFZGKcsrsLy{o%J|64;y=_=r_zJZn(JrX
z(<tA5oi-_%GMsAjZn=Dee)-({E8d6B3${EKUlZZD@Xur26t}J2I<s1zr!CAFJbir^
z^t^7KtIOOTSOu-pA;!S_JG(cBA=aUcn)j9UAMYu@?-&%7HG$-!GwxP3UFHkI+~7>#
zvcqg}0$nV4Z46M?n^kPX2_y6csGHyNM5vSHtq$+2?pFOMy`XTu<3berJ-RD-5Z#w|
zF!1im$ekC@f2LU;C*-9hhNr|&o#4i8J;C-83NYg;SJ{2~ykF$coeItuaQ;8%_oCx<
zC9PV|wnu)enJ1m#Uh9E4y-K~M9&ym{r=Z!trJ0*rwh!#uT<c#PNyLaGilE)XkH30L
z1Pwn`?-lvnkd7Hb#;t!+mq0i6k6em~ZU=#%>08EBr(Wbg)VW8qqw-6=bXA+H%)ywN
zP#9AP?l`Px%788I=DiI*S5WG>NKqISuAL7!(UcC)B~fo|y0R*oj3h*1Zjq6HL-yeM
z`{?ouFlvu@i44X|iydCgv3(B>4j#>()u9vzGdfy%d9Go?{~YR>N=iDBaP3slR=P_W
zMLaWL`%W-i{}?Jv{RAg%oxPtl!?HA`ug^*%@2je~=;(#(Rc1&|<@&)ZzhCXXGE_Z&
zATv{}w!(<-5^n)4Ry5#WeXo-rKt6UFJF0f<`803DkX1FIZ@l6z7715rby>^j`*U;@
z0q^LkUFV$RUr$T6UxVCUV(2S6eAmR)&c!D&=5YVsXaBPgeACZTwqZ46M3cQP!&@fy
zskra>msQGZbdHU+;!Z!;?#YuL5S$R}J<P+M^_<NIr}xV0GAw={H**M**469euddy%
z8#FIUsi6fjMJr8NsjCOIrE2QdLLl_etvjqT{9*JUymmffwaMXKj|w*De56_2m)dZN
z-is%91-=sQzulpMk_lOcoZ$RCW2lJwA+S65oyGlcqiInibEIWF^4MzmAbdsRjle%i
z;^_N(d;hU={<k(@sgKWeRbrQ~`X%T;pEaMS^vzb?)@XOb)a?1rcRtcP*8i5t{@Yai
z|MqY6NH%XUL*AUWI@5$!qEg;;&yobdNLC;_^iHNLl;RGmf7wa<9v6R=Pi@`@s2@$`
zeIwo-+^@;HZaddWnFm#NXG~}MqW%TClTR&?JcsBJ+$S`>sUiV+`EwpX2{5l}P}gTJ
zxHszz?;3q7{_$HUms^N9X=HU-e|p6FUy%8mwY61nrz=RLu({qlA?^NqFXBE!Y!@(p
z%qR8|M%af*_JCB9l)DR!DeygpPgSSnEhiiL%RkX`;m(FEW2Vz#m6IC{ADD2OQ*3;P
z=ESRsUaUsaQ*{4@GZ9?q%I_*r;}2AJ)*tlm!1&n2;be)U5;w2+bJ%+P=h}@!i3;R$
zkPP{5s?zL}Jmr2$4qFg<$Ye-$09(~-qACHfBi~GLKNiVo73MK&Uk8cPOHPF5{8Fy4
z$XV?a`T38qQQ-2n1{O;Sn`emO4;^o8vz$SSnJeS>4gc(W&I7yiY%@!?8FGM^^^rOn
zuYr^B*di7uVfGceJcuxv1H7O8Uw#h~crgePUC|vN3EY!HW*HYhG|*s?3d&THLx<C}
zMT_GBPG@7P@z#5K4nx1Mss*jkST!_WXI5P0CMi&_m(mgrdmW-gYcv9pI)XB(s!f(&
z<q8Vf%y{6oe%#h&@xfR{XeaRTBP2L05B>!5GGWkGAuY4=3$H!7SsNz#0^bQl@vUc6
z<T7bGq3=*o&b3%msZ?Z1$8?YHas8VR*TlS^cL`V$QHk8~iDAsVG0b!**UkL+w9^xj
z8+&B$vSlf^<U6=%&}b)APVctr=BG>7Xz*Co39r>}X=gTgxzJJUR-oY}{psqw1+BBa
zRBmuYy=t5u94&WgSn{#V!6&!3@&z2g(_5muv|xOwBU){wLoNM?><u}mXUeDV)@Um}
zSZ>c+!}SxBr>2z5f`FXCiNb58@uwlOkg^m1E92C*4@UiKH6;@Uup}M{mXyZ64-47D
zik43+b%l)Yjm-~90B|OWYTvgb(Im+RD3yMDyZ}{enK6rtCGnw5$w&d>OQRu!kGGI2
z1zbg!Pu6#(tUMkDoyu~AKR>0(R%}hV|7yYD<?}W-z8oH|0s8h}>ukQ!<W1FT4agG#
z+%9PRBL~NKw&JQJ&nV-|5{D1z_&FER<bm3K`K0HXLy&HZSmxTdDE{vZeR>=;_+l*&
z@IL}W&SnRKrxGRx^iIn0P|v`VtWUH1){haJn;QL(Io&$=WJMOS9=XL&*Y&04Y^6~<
zo*z~#vG3Vx{8;<z?oF;|*taJ{=dx_k=UIj9<C4b^B+;O%TRD@HafbsIxJ4P@fVh{%
zN}Oh8vgK;#a56W7MsHm~{C<jgld`Jpn9TbV9Q=yEi>)(`|9Z4_`I8{@*lS6rG4fG@
zAX^F3Y)SBtRuf-rI5s(#Rh*==WrBJe$|)^sd)Rel{;=(e3i={=ek;OVSxY0@6;eBv
zn`j#P+If00^IEa_+i-|K<dL?*TsX`Wa}TfyGqKY8+;rPE>XdPCu<tz~Qm{MA>9gV7
zNhabuVrPK$oN%86Wq0eO*TtyXFpM+mdJ3n`F!Yjs(7wQ~GIESTu#DQ(#OYuo;+-hd
z4nj9z382MkQPZs}d&wAIJ4{aY*16^6BXpZrf*R<4p%UH$r%2y<^&2D0^!J=;OVT*u
z?>X3Yr=$50GE0I2w?%`1ut||)+E9OyVo+yG9OySJ8!pL=W-fQk3`9!?D?p5{1gR~*
zU>v98wqDaV%agI*dQy9>qKEwhZv1Nd_>(i)J{)$JF#V!7&<0b94bPmHNBK2*bVs~#
zLt&Co;DNca>AHSJ%**{~`745<PBvWcGX{c;;*7My0&SPHBAALLbn%u3x5%Q;nIq>8
zDWB<-_&og(Bb-hjg#tCY1AzGGfsER%i`@LundL8cJ69lE()iS8mI1(sUjlFb6vorR
z672sHIW5^gyALMJTB*#Y;;)N#$sgz(;uK{(mWk8wTKKxqdg|-L{hRv;IWVH|mNyN+
zkjb9aNa0)Edz53AvvUn%eQrK-_1YER0$sg`cc`m$t^7;*h=38c`q15$cKuhY^PeMM
zm058>0ad#T$H)2r^GFaF@0`x8o8@juR`K%4CYeRW9$VpkiB)rSwQbY+<0V$Fn*LI{
z`3%0Ws)&3#QEPg<9R(ESww}^PDOboQHE<DUQ~#sv#cqqp<0nLNmkl@b1zw@Wz^Af=
zd4#!lNLfVP34OgNv}l|9SsrVlF8JxQ-pcfWQA$>ft(DWh&j>Y<Y+M?_%R}5t^31xo
zY&3uZv3U|0dr0&E_Rwr+%JR>u2J1UTp2Ecvd-kh;dF?w^p|mL#4wEPRpPSv06wHRp
z_Y}IyHZ$c9OnO<?4;0;3Y{6@h2a0V5bkG&Q&49BcfXnatpQ8c%&pg;lp9dRHs&XzY
zBes=XqfTHv<3vz%3N}9?hh;W!m2Y8fQli+n4Ks@3wOC7+VBt#|?uAMR0Ph#fKrArH
zPnmpldUj@s-!BiVF-O)L^#`(<>NhrV$LKv|K3p<fuIWnoL~zzaM0VAuCKm-t;Lgp1
zts;c#l9q|pNT@W*dM0#%`9ft??pBsWn?-y2-w@fOpIX^G7^*p5g*`}gE7Z8?eYW+m
zm)7<Aul47zaMO;4!`ujN@FxH|wHMD#ztJK_N4_TtsQ>k_x9y|iFmsHlG<;I@;t6Ys
zYp1zX{xN;y@}k3nY4TgXP6*2<dQ%b&I*H4L%6G<Yru0OiP1#?u(#0$+T01DP1B^)I
zs8<g+lCf1E_xu$)ZMzX#;Sr#Jd|+jx+=8gvqrY-igleK*QBtZ&H)Z&vTjY)b8pRz<
z#lnc6R~|=GMo}F;$xPZY_gNMvUzhe4K=2n#lf*sq^r-l(S$?@+s@0Q5ld5-8C1l>7
zkk?-AyGq2(<F=ewpYZbR?&XmXr<SEDSszTLzasL8#OA$;FR&CZhFl+;&xvPX+bKwa
z-;z;zW{sdeo-Z(@CPi;G2h<^d%R^q<f)lqqw^sF1<fFb=D|#q?Y)79As$%mt($ZkD
zM9OwljqHU-=Wz<>#OgAg-M*Ni!kMkPws}B0scyRsN=*Lo+ji~(ck7W>+2hR<-OnP|
zs^-(K_!Hj4mZi(t1Ntg8G=bRV6%7){xX!n2p_A}h!2x;g`P#5j&dsGQGg?V9S{}tu
zU5#TOE_$bb`1{ebpFWin(kd%vAxi085NHSfkD$ailx$8LR6tLc+%1jV+4HTRWb!u6
z%KXiXU768^IqaQ%SuRL-Z;AM+{Uh1~Mv!Z#m7twd`v>vzL6-o6w8~p{1Y2s*|0V6f
z<EI99Rs`e$zv>BrYEB?Z!NLjEn&E^;372WS6*-|AF~6etd<1qL?~JaNmk-<4(i{zs
z_UwHlt8nS**WkTU>>vzcb!@dA13dMLqt~V!2z-uLEppx&@fbVi%IDtJ<}3<{S?&Wl
z(1adGm2aiek=8YJw17CtJz~SM%q3T7OO(kMum!HXMDc;E?z3L0;Gs+zIgX)~qv;Mb
z#g&!Ay;toIHOFn7-_vxd6Kviu?|N$%P_dEaM?a=rC=T7iOjd$@v`%{)m#_gqGUhCt
zW}g_n?;4QK2g5pV__g60v!>-Fk(8&0&g9B*L42|Q8_nKf+7WHpiQLnfOBr1rg_({J
zinIS63c(Hp!_d_o9VDm<M?oSe&_pMI<t~Q#In`thFsoqs44j0&g#MRq|9gF2^e9RA
zkAm+(XAks&(bt0;OK($ZOoceU<y@RE$p;Fw&;fWjoVpHA!E9(-?g&udl_S}_nhZm)
z>HX}ZUiKkm?&XPn7%<;oIJ$2AVTkOhic_Ja8rc<RToWiePEZBy-Xaq9nu^jjDr>YA
z8e{)z%Jf0Irjx~wr-$~y0%GvN`v|%yz^!1K)1%x~fbReTm=dkA=yT8R;OqC)fk*6u
z39~PFi6LeP2jHf~7Bg_0%;RDMdqS$N?Xz4aQ(?DQ`FQ08wBTUjgSs1~9~<^xb?6+f
z3ORU&4opy|G5g2p0xkA7kBi&hkB>RLJCQK-xQl6FNp{v`K-pj9q*xhxe4-h-*Qe;2
zzd45et<(q4?YnNd%^_pkf0KiTWr{|Suf=R#9Z4J38ds<fcl?kCSr_*Ks@(ys%tz*m
zZP;q#R@PakGf796Y@&73RnIMV_4>j4wJ0i7FA^C5$%W`g=iI`*PIJn{4VygC!H~Zy
z?TmszW3-X>>K}HdV;t*p%tkdn((3m#MYfg!QSaXb%ev^^7UI+x2U5gL#xTSi!jaPQ
z&i1F`XBb`CrhEp*Z|^}kq#;1TRK98#D{!^Mo_d^;E40sG19y@3hhSxxV)={?Z?3!H
zOks@;Q_HYn<=GI%5P7@9Tz=ZJWzw&`p=S7WSn?!s>ZJ6(V1r}VjUOU*KF|W1rcThD
ztKQWgso@*1_~MeL_3=KsW&@s#S3)7cnvO<Yt{I%IsUMbK=SVQCv^ke~I~R#GI6FN0
zoIMDwnrmcepiHxx@rN^|bH>UG>d#IH-9?_^ye)LT!xyOjfik<XuLk{9N1_+?(H_=i
zgRjR$5vHG_&wLmqxB+(bc?f=}=A@8`sdO#nuX7iaR)7H%=rbbBRh<;K5@{P&iDwr_
zg;EMl+H5H}rf|-+c*fR6IyzM*CzLC0hkXS84DhE?Vu&#jFfa3mh=lV;D~DZ@XQ#-X
zYrhnPKFj@db*qHA!zO@06?}PO=A#C`kVBZIi5gzlVL2lq{B%8~9fAhto@r*Ws0(FX
zp4b|d>S&cufn3@^o9i2?!?VJjUPG2t3k)l-L1bnHI!(E_!;=1R2w}i)pxk{ck8U0O
zjvk>KRD;k>Id!wH_lM9#2!r<@(}8q!*KS`kG*Jo{Pq~x5+L9;VkVaG8Vpdu#+mW?p
zIPy+#ZY;5R-p=%hUt&FF^7|Dj99_Cz!??mQPJB^!dgl|BE6fxgj;);kgT^O)sD9|t
zGNOWM8C8!->;0AUNapr)s<D0tegtrBj3Z!GhVcE1B%jVx7lSdoVt`K`@0m8G2mo|s
z1G|^W5Kgc>xiQnFP$<A0CPS-vY6s>WdK6n9)I1Ttp8NXak>@D|U&;Y;7z#vNX=lls
z@9?73_Us@Rvqp@fPwuTBlPB3?U9h5Hfi}xA?5rAS<AsAXBhqB+q0w#D_=S>XiG9%E
zG!q+WrOgdmv|$`z3-@kXL+c<O%*%N{XK{{Fj-f>{&BvMd?#&R0IhI4K(68*_y37+l
z@9v4dH5QJpVZ8m1=J-Dyy8jnA!v7E611-!R!PfM(<G9&FP{C>^`Ht+2%oq3Gb@P8}
z@_T=)UoeVOkCz3c=*73dm^ZzK{ub;R<9|bdBI*DXY$Vhh8gxafZL5@O+^L|1*$cm7
zLv!Mu>)CjR8$LkO4o5cQdP!i^2babxc?~Vrp-<3+<9h#+x}R?B1Dm{}ExPyzpbj1T
zj`70EPI^r#q}`cj3HDD$F-C<1pCu^LP4>#zQnDVS57L!p@>0dJX0ZI8yXJ|Anmf^l
zjr2=2fOOk$Xtj8I%3iN)=18)o&@FRe?psBlo-5tnIAwX1a2@E15UAIe`jVvDoFVks
z{wuR*r7z?ahhGu-5}sC$o$WZUn7xrgKcTf8I8BO&aWM^NonS^aLrjNP#7~o-el>+U
zU&~Jyq(mU7Ev;4r>Po&_Rek2ywF?7TNwx2nff~c++buwSi+F7|trEcajOAtKB-wl?
zO0&GI*N+<g)bLH#zl}-I#`OXZK_tDK%-6Na2Npxn8}c{rdG9v16$_%h2TufdK(L#2
zVC_lA@QTc<AJVNpghnV=H&(7Wx^VaS{IUp|uR)i3{te+rDHSbHitHD}4*gzkl$yLF
zZPEc|vwh-^F^IlI7{7lvVgf3IxFPMvy&Mwtqv~Q<{8`)Ol4WUzk3D+CS-7!wm_i;x
zxfDI#YV`Orq2tT=DI0BZh5}c^G~XoL1#{e>8=o=MYUYuW=7KKIQ<qAK6Nw5FW2a8b
zeN@z3Q)CK<c52}dvMTiWk=_rtwutRxl?}WP<kWeZx4wY7fXn)VgH2WcqluT7uGJ?G
zekGpEv@5#(=FL9S;}y2QNK7%lEo|dJ;d{DqfpZr^vkejWZhx=dms^}t^NX+hxzvcj
zc46%(f^?91xYmw0T^g}%_E9#NyW3*F{v4ivgy4w;=IE+&#<Z-s6`PBG{7K(hV8!1x
z2@Fy8S1oJ}ru}k%-mH`E#p)D_RMZCIYsY@BYpF?{7nHW)GXkJ00Ch)7%<=|~FDRh5
z$9rhsbcd9wbYtZO;1~eIn&6dDcc{57_h?!P)pdz0KSs3#$?KY%@Z}*;-nr7|TPLlC
z{-V+0B0&${8$UnbWSq<2-lZG9fg6KM0-pz5o60!{0YQ&*_T}#qY6|lXKh%FkkRVJK
z!rq4wmPbn+=!6z;Wjz0I@!FImq&>IiWn~5T<YBpL4u%>PsE0+f)q~VK+vP-uU52aQ
z)_m=PTKg;--&Lx>wOZ1TktoK&R18_P?ukEwl_xmb_w2Wl?}0I0-juE9;fQV3aM%gV
zbQ!7Iwhi{=Z-}IZxo!*o_@{y8*HyuC(!q9@&fRnk4_ib=5UAJ8ZD(N14!*NU;l;tA
zluoJS5<kx(WB-!@J6C_!hCmKtc-NtH<q~FAiU6n^pI~=|Y`+q!FwIVZTTkf)1wnk~
z^&Z~y8Lu<l`?7q(g(%A=5^@_H0!c70fq3T&G_DTZo1h6h#^piDy*Arv#v<r11@F1S
z%s`@_VHM&v>4oHujOhuOh~rYWH?exucggC)mt+x1@2H$#+PDD$up=~|!|w^wYCX5>
zwEB1%W$F)d#`WPy$)ast6Hef@FpXP=Q%Td}DK(&h2Qz2))#DANCJwsPIo^J<+5-tK
zYHzd2T(i_j5IIxk9;nUBZDo$W#~uUS5)EozN$8B_lQOQ&p9<~zA0IT%a2M<^;g%n1
z#L$%#DOUoixS1YoIP-kS;gQX?D@*csO;jqMcC%@FZRfO>QyCgwe?w~2+Pk6r4jS)0
zy-+L*mhqX35<Z9X`20FN_F!WPo@$9e-n-^x#T*S2r%JrUe$_F1M`-@NwYF#;L0Dfs
z$j16vP)vuEOG@sU`uY^w93fvI0#gNTC55Bq>IRu=`S#5U9((K3jHEns8qZwls|j>g
zpz1#+^KpL|d6Zyo^u0ss{HwiyQzQl8Ms$G5Nx@4mW`?+@n9yvayHA;(k!VO!;Aeoy
zl$j;m!HY4isPi3a`)&mlw(LF$NzMoef0vqPUHLV2ES{MPQcoHvQy#5Gh?UdCKl^q;
zqbq#&24>%q6iXg&PLWxjPS2jlOlM%KDcdci#M432H}RcOO$JxFUgFI>!`%+xqmse6
zcY^q5r~zhBoZm2>PM}r-`~F^(-5bv=0nae+dkU%MlC4Ty+J{HUG0R1aJ1A?KIE8=O
zNz0kmQQ%wH6IQ#Zn-D$rMKEqMJ^B0Y-w=j9rmKlQ!EOd5B_Oe%)2Lgv?BkM0G>U$7
z-uP04fW-;UmDPztHt+e_l(;F4ChBPG<lT8gYDVM^98sB3m`#aR58Sig@#kIqwSJ68
zU)+==A$ipTyzITDGc;DgNt^S1-?SboZ78vMDBw{isiR3uJ{KBv)*ri79r9f5-m9Ke
zXjU5ERc+lkncEwhn|(=y=^QKv%C4l7D=R{|N!M4pFyWKNY1Bz><TsOMVANj7ieVzj
zgg9u`XpEjYdE9$`*#Cr0sC$H(@^h2Vs_N+bk){l5^mraO;9YAMjo#CVZhVxu6>_2R
zuARZ6X*2J6Swug|xZDF;)z{oOdG#^_7SnXLB#T%t?PB<u*@Mv`$lz5GBsu*P2)UCZ
z6|eD)pK@bu(`!#Zh%fvd57|a0k*HR<M3{u;?}@A5i>Ua_Tjk|>nfB+cL-*Z-onb0<
zRG^Qc?u|%H_00-MI*a}d$)sH9Cev_31iB*VcnM=#<)o>}DKx%WuC%2PcHy2$d(Tn6
z=iY+R)|I1Zo?gVpsm|60=bXE>HVt9bV)2caPYOkK7XR6S1X@WGFfzPF)?POOF1|Rr
zPhEOB<l6^Jap-a-?z_Mku*AMT1Ik9JP$XYa<x!A{lO(uOaxaC9&AV?~I`dRaY*nzr
z(NN$$EZ&$oWRYM%yLqN5lb93_8Pr!7T<;h(fF_|u=~;%O>;p_=!eE6$7!2Nzv=>V|
zdh>OxZ>j-wh!Q+c0nHl)+YZ<kKNhgPH0Ma~F%$_xHRiLonr-9DV>GcmuHrXk1NY`V
znKwC|(-{{rU81xURwOv=_VYt9np5I0RG3$ayXC!;6ph?209yC}j93EvT;ERG6N9v7
zWV*eJCy1B0I%>cEl}^a~DELSzGIcecIffg8(vPPA<3}TU>MJ7CivHLK`KUHs$N3p~
z`-b=311Y3vXYif`ol$?$duB~RNW*DSTL-tY`s(jF=1pY`pmbRu@R}&uk-~;58v3nn
z$EC0oyG+wr=Jjks0Y$N8*C?u_C>dK!B--T@k0c)!7hQ{?fI@sPD55Z(eft)rY9os~
z3+c1HVwoeeJrKqXh)KNRf+?j30_5=s(*;CvigsbE{qpsNJniTUEkd7-Lm=H6E_B}o
z%MgYt;d#nFdZ8U%jdrp7W^J08(~V2RkPH`X#qRWj*h|&%>|`_-^VSes3#CNePSn^=
ztSItI>lSg>|0~MSRsV@wxHfD8(;-E84D=;L5jvvF+2|r9r-rTI8Jio`8%GI8!JG&=
z!8f}di$5D=X+DfXlsSz%IvdNp@6vs`c*|XHbdkgq(PEoh0=x5N(k`DZh?eC#fZ<(l
z#%>(~)#gP|Uz{8Ox+`r=&+ir0u&BIagby$B8|yx#l2a{s-`|J?o^;Zz8vNNGLD?-l
zdj4<pydS>#>I2;?_Ird7hz3WoE@>K)6$#L2{pOf+Xb}IYt>!?1y9WHClyf(L8ucwJ
zPk2D}_}z6gm%YiIWiC=_7@cx5NzMa%fO)JeD7{T<;5X6yiIH8&Q!01<rowREK5!Go
zR-k?$0G@RK0e^|5+QZ|7ezW&dF;Cvjr-?T-1y_8x;OwmW#SI8m>p03WP)yj-6KH2|
z5(ZHRmu8Us;!46Dgx@KV;g`b`-G^kC=01XXwhzrOI=xL_W4LSQF(M%gIs?CiX;y~0
zBln|x=;fuPpauzlGbcHyx{9hxTS~@}FDqQ=iQmr9A&m?A(n@@Mp=Ktw_aC%N@a6+o
z?s)-oI*nF1!&(kBo9z?z>^R5BvOW9GeJomA6P(VLlfW~1j*w~lUd~_>c(_J#9xNdb
zR{G>QTr};tx73KUXPu$rsIX?ZRFl@ZgxQ{bLpN*uk4TxYJ;stHO;(hs#^k=`II>za
zL0Zprx`)qjcyq}?vU7aJo6^lAfA2uOe8t&6BfnR&T2{3q8VtnJ$;`RIIn7MTr@x#e
zTm?aEUja5<NUBzB$How<m5J~t{*!lG|8nqdk%pahGZW0y#?}A92-Zz$=4`q5FE@eC
zx|DuRiP^E-2VQ4H{Cpl?cIE6n1Ej%9gDj1{`{)k@7u!!-CI+XqNswvz0>4%rzpK@h
z+b=S#GEH$MPn!nKUbe^m%pJu6Ri@|kp&95wqNhD~tbe!D@yR{SqJ*5vn+MM%D`m4@
zQOGs)aq|9xE6JL}ZoF?!Lw+E(g@Nyub7U#H8^14m-lQtey)v;n?J|$z?NtLoP`&6d
zeP`jZ(5m9BH}s0NMTT|2k-4VHrZvnb__gm&Fth4G?=~?T8nH#kSSX5<`u65OxTJMp
z*)&7G;QmkGlP?4sd`h?Y?(GuW>c=PQ`Yrz8{w>X&vkZ0iDDc<Rvt}3s37CG@1Iy_5
zcF1dA%f0Cz?L_7eCyp+{nIZ}7zveqhn_r<UsT#njJ&Nh-L1_bURvQc`D7_9c^>9%v
zyUr}zlqKxO{tcOgetn&PkefVL37n!L@S4LLJ4E8{=+(SiZkHkwMV=bsS}^=*B3!b9
z8TgvmO)kHlpME3yTna@XqWIeWr6ExUkW|2QLFkXcZ~>9Rf{d;w%|LffN~y7;3c#x9
z2Duse#K*OXiqFvd=4_`j@FK;^1eZgYehrMQ45^)4_wXakvt{N*UU8*{Qa=oS-8w0d
znz|vn;STQj({g~YGy-GowfzdC+F0&rI2p5B0!Up8oU?>6?yU}5MLBIyn8w|)eIv+i
zsyDboe_U7OYe(AyOUL-Pj6bMw>n0>k$+2NTU#cp|uhroB;EYJl$1e&<TZvQqDe-i<
zb{`*Ap^?$x%$X6*Bmvd7X@PH}6s_g6e+pJAG91PCieA^5t-j=}hL7?=ccZZ?k<gjz
zoiMmPg!%Fly?_>a)5Xt!7AEePyRxnTU`nH!LnuvFwZ1y)*ht#C1ud^6=&sNGP0ye8
zF$;gvc|B7!|5CB+d5zFxALKosTx|5+v7G~058*8tDmEI$4Qp59^6dApveRY{SBKv{
zZykGW8WZXZCY?E~D2z{-+SdDi#Ub3ZlUhgiL!XA~8*9C}ZZ|peyEsKRlm5WB-z&EB
zo6Q7Lu1t3)@SMB!#@V=dN?_Gcp1*ZNlqDd~BYNwhgzGg9g#?-VE6vx9%!I?a%h3<f
z<8Z2Vj#_j>MrRaSv>abqVbJd5?>Q9Kq^x~HgzUS`T@zCB$6YM_6P=%W+LNe&xA-{g
zDb=DBJF%VlRovHi^N)766iD@?NS!*0AkEK+8lC5=PU@9goH4RL5h$Q;*?M7tGoKbn
zv7_N;Gi&_9V2LP$zBAqSzXOGhd0B!MT=!aOm&mqJ1|L(}(CR~?(SJkEQpC?hTrllH
z)6ji{N}$+;8z;5eG@xT%O-y>7DD3<0Q^>DDSq!y(7Zll|qZYrBlcZ)lq~g!7KNDbv
z)VTdt4Pe+*+206MYaEoR+Eg{z^suuXFa1Nz^mQMxU${H?$;+_~+_+jaNw69gKXqV8
z&vItz%j%<Uy3MUg>0=K$h++GC(CQ$R%wcy~x?3okKQ5km#Cr?LgE$aqm#?e*Ynn^0
zGu;UkQG%W1pG0ViW=eEJUV;2AuT-9mMkCm%ULMc=yQN?h#06IXOi1cAH9mBAEw1p^
zo24nu4E0PwLN_wC^^0$!-{2>|ts*+<k9VC24Vv&esC~1hvHXeEMkbm2E*e{e(F=(#
zm_;U&grz+5ydB!?DGSy=%w&7VILD1JM-1f3X_vMZ=o2)oO`2)bjZ)8Uc_HNu#k9#Y
zziTWtTQM!Zv}h4V;8tg%plK&jZz6SfbeXga1_gV!sj$y5-ro&*AB);6f~GvI9Nspz
zA5eLHSj?nz+i-IZUzu}dXhp@I=%5r>cc;2-K=nr$C~3m6`!Lf5SSfv;#r`;d#!GaM
zOz=nuS#R04o-6q#xd`0~diAxoPhegozMb70AUJ~JzjJ&Vn7iKvwR|pfJ2CJ3EWFN`
zDV*Md5z|vD1v*m!*FEIa-anrApp4$PeN(8Qy$O+V`Ght_?WecTs*QKgj%3YxpDC_s
zcwf&K=r(>`oparYJOrejcO7w=nCaKBLwbf(Ag?y!kK^~c;c1`yu&vC`=a-YAKlk2H
zsa<i9VH4l!9|<~ZLOYnF<7n?C_FKKlU0&fKj5C#P+YJM8vZL`bkllC*3{S<v^-r-t
z9=hKr>4dWTj$<DU>RnrHm2yi}`PJfN(QY*2Y@nFl6SPeMBNZ+PJ%%>XDc<)Qku@bv
ztyb3v2+#>@z3jyC*^@W5KG&Uw%c}m3)SIjO;HlSL@bmd>uj^jck3ifwcI<Bmn<o$y
z#;q`cjY;IJb1HJS?~vCumDce|*NF+G1Er3)=l8mU<+HXr(@8O<A)0~LQN<&q^pe`C
zZw1~yNQF#ph3S14$Buql-@>wr2_A5Mrv3No?w-&aRSlO@`eMwJZ&5lHhMDmPA<eQ%
z8+G>J^!I>y5Kj|Jedp;cR7>6Lz{0Q}KFg*H_Mg5Rx@${7IP>?fmRB5a7%ca3TdRGc
z5G$tbco4&30R48bFxa=$OWfAhr_W!nT=^U5k2eb4!lKfh#{oe!g&Y&3r(5D_4~yL#
z(|D_GBYmr2JwVwdUxF+;hY=~fg;u0DXk068XrFzP9L__EXFn#m680{Lld7~@O)E4)
z&)@SF`cYMV%Jf)By1^lKGhTVL$(L~|NZOvJ^1Xef#{WRaZ0Nh4<jcJr=jZc9M*$e3
z+7@&^-1<2BX5k`}PaH*Ss(|;*3+)}-b<y88=fsHj-Z<DBWNy)@l^ylhH$>^6_T#!t
zr-ahrTe0|oyA4kl^yT)9@L8d6K6f=C-FR#ujs^((=9~J;06yfS(xj_Hx4eA!@*(>&
zh6(q`6+LWA6E8dtD%E7NyoO+(9p+=UCvL0qhjoL+R1FS;=gE|V^Rw<Lk7~YqGy9ax
zhAOhQFGB>dixv;Ly;x2|)v~&#Rq-QVJEn+Jn!Kr<5)ckzI4p;QhL|Nq?|()wY|nOX
zZPlD?y9J9A-H2IU1<!kp{*sEbsjj9u^)Zi4Nlg(v%pNza%Bp`53JtYg9&H0GT3&ip
zLC<SCt~FDq{n;Bs)57`1{mnhVY~=^3Di>De-g&{ya0nX^<<Cwp+4sjWlY0Ts*9e#c
zY<Ul$;Gri_6jX9C*RsUx(7RH@u8Rg=u@8bOayK-0dcfzpLr*9Rikz{GYdYzt_QIvW
z@m=Oui>o{BFJgw`S(3;G4^A5A))ecV{wccfRv~RJrhwjgeQ4vY*=rM(f|n78*M_Ma
zo5u-8Fdzj7_I{A7>;`K{!d^GL8s23`4SW0PX5M|%%6a_HfdOQdwID1gsFrB34J+Hl
z6sC|Z`!5oJjJ+?KXFsXt*c5M9xkxsp2hK`&n@UMPzyCAn=Z%PW-1bamsv;$W7P3xu
zdtBl~U$W9c<sFXZaY_3XA<^^MszkAvAN>mjcq?gkvnjjObPyt1vpiBTAuo$W>yZ^J
z37JEI*$|dJtCleRMwqU3tBGpdf;*02Slz*o9+z{y({agIg8bBYsuy@5QbNgD9T@2;
z%h2H26j7(0Z@1pr3hyA6MPxYOE+Br+fvQMWk$n6Gm<q;jr8zH51)Hg;qM{n^L^v@r
zegb$P>kgIFis1{=Z?jt<@+U>TcfvbfOLQ3mV=li90iy2gXt=+Lj|vJgt91U=`88Fw
z<6~3pJU?xJ$1Dqjm^jx+$IW`fW>(vZB2)sXv5zQLd&gy#x%ZQE&vAejH>mI@jJKqy
z&<;|J&fHti*ApIi<;vQ$6<Le%bB)VZ?b5Xee>}pqcyrneqr_*sl^Exp`>cjmtOqY^
zyWaNT+)c)H00|dviYP6x8SwUEH!S&=h&_3+wF6Ogwd;CK>Mtj?aU)1jI>H3?MS(Yv
z3N&+c#ACT|_yBYmWJU2H<151&FKrf=%xdcE&wYqgPTfO#unkxO@;C7f!!^<gZ#z9z
z=N_(|WzBY$X;SoMS*2_M#W1WHdW021)b4sWct;&s4|v0HiC^za7<diTX3b>01jX0}
zq9F$n+rTN6Y8<l)b6?%pt)Jwsoa9#YV#dl8%;uAV)U4?#G$0PM<X854x1+$D+*kEt
z`jgwEETzilGV9xcwsXdj7^G#Qz9g+UwbToJdKkC7>@#APV5<1=L#bZb-gS8YUm+KA
zb?#>#`ux<sL0y7;ZQuUpSwbFkB7M3MNYxeUSr;1K_udFzv7fj2(y?Fxk_NuY0NNk;
z7-h+wq$m9fkvXVb=_Tgt`0Jp7a(F6T*rA3x&lGYX8_?qz7qI+Jj+(vYGanQEbZZ^D
zQ=fOhAZOsd{T%RADD{TG#ElNL7=2k!ei|w0=ox}eM@pM(<`$v4oY42QVJ%nCr6>#9
zCqgSQr#eErGD4C!{g6qEl;Ny2?febzDT%LnyPAj<$F08jl{XLnKGxZod0m-sg2^fH
zH5_)EFr9%uOpl_;f3{yuLix_}v8+y3n7@1`{_yNc1JS0>Jevx(Q=)T(s;riJJReGw
zihUxaX;(cVG4pcRPa*ta=hUZD|AGwffHUw1)(1T(<0Spoq*ouSaF#C$oZ1Iw?wg+6
zRD_<`*3VeN3N|6{QOTc)q&Q6;`DBuLbtZW`hGYC%2E0rjnuO8D%AwT}q-x9da@%(M
z>LEu*c(2vqWG^Z<K{i$GeGepLz5~o_>N=1~Qm?!AeS*cjlcaGK$J=xBjoUqAb~Kf+
zg|UfRUZaEw{nO8kyK;IzYWICYm$93%RFLft1aY@W87Ju<@sziw`Tx8-tIIe1EkKW1
zFXNXe7lv4Uu7Tilxb7{Q(G)Q4(tRt-M<0y3<_l-AlxAApG`><~i+ySsS$i5;xr95`
z<{IMdH%PE2+$8|V<7k+d^u7K~QFFeI{^GgH)aZ$P{q@eV0P^`c9C`$JOs;aU2hdV^
z2x&fyKv08-igy_?AAXt~`x6vI`VoYKGxap%Wftw0srbvHpoG@#agGlq56>AH7eLVa
z!eJ%6Z{K)jW~*X-!@_)J%U?D~9?Q0Ad6oP+d);z8hJ2-@vo9h~h89-+R(bo7vTaU?
z%Gc6$gTEn<rKm{ysqMCe6bQOI%ciz>fyf#goVe3`>dQCGh7@GScnh4XBR&LWT?dvH
zGuS9MKK_GZh`{f~DUYfIn>~7s26ETo5>6;r`Xo8<2m1sqxi~F1DtT_EkO|SQ04_A3
z{bUUKA5em)xj?JjaSZ*Lvaoa4zTn)p#wVCd{iCb%Png^@q&*A$o;v$dJc~U06OEBb
z3Eo%Xu-q?fWt0ci{PS~qR{O5qlX-Q=A4k#WwcEpPFn~sAv%Q}G^acBFGkUae8(!M|
zrswy7A8kC%4}ZS=?B7{~ZVl?5Aw}blc=#=_<p-2E!_<A_R957<ZlPrW&z|y;A+ku?
zht_fWt-pT&xd}h$#jC2EDUtBJo4@B!=e=HTpcDG2NdR~x&uTr{APCXXG);b})ngx(
z3gN;dGBN?U-c_J&-GK!sWL~B(b&d3yh%YI2nDpGXax9N#>c)wv25dKLE&z4sB{&pl
zYsi()V5EaWZ3hc#8qeOcEW3RDuPZl$J@Y<CASmTGT0@*aDp+|*g?%D-oKDiX6F)x$
zCYbGXMu*%H+Gw2Ocjom#ztFc|^G!9UR!=`3IzQKliH1S-PcgNq%xw2a|4;U<W}bQ-
z@!j9X2@-wX9lp*lLGo7}qC6rHTiecU1<)o8YN@F*h;aRFFBS**T8-#%;}MwK(r;w-
z1fmR9{e`4)C*_Cp21g*9=Mv4K6Ao;1$ev7O>7rCpd-)F@pV65$k+<&B_{btZ@V(mf
zB?0M2!(#doraaEcBgx<I)HiR}lTCZSaPK1+YJB3e>>n7i{1Dya`0L3D5$M-{WK(NN
zYdrMOzaj7&n1u#N87LV_0BQFK%7H%c#~+cVGFRnt@Y%Y8txy%$gT=EEVN09#f2L%h
z!?y<i13A6z3ER-XjT0eEaGZ*Gj2<S4=*zMP?N^d!R>eravt8wX)<w8kL(eP=)PyF3
zf8&x%nNC@DqV6)RnvkMQvypPu1D?t`iTB+6g}z!|(|6WBjt9+;h+NPB9Hpl)fd!fk
zH*iZ<nq+^+h?Z}Jj=R@Vdvo)`D1&Q|z0eaOVsqi@?4#&WtPENc0q~pp$3|WmE3(8d
zdlrO;dS8m4^~P(VY^W~>&7g@W%i^!n4<<u3f?*efaSMSnfy<0PW5%h3eZa}`z%9?G
z9gpos4%=T!wu@T{8O=eD0QHu?A(40(^Xv$vBg&IIUTE_5(B}T9N}VTU9#KZZ^<k{X
znCW6-XS|-B3zaeZY`_9}wY7NZ)HzY~<6~XN5`<18hA~J1`bq1$&pF*Vt{{lRBKkBB
zV`;wasl9MB7bdP@8`oi0y24bNJy{s<7QJoJaL=w#TS)<#n9^oDiJ@9pjui_K{oupw
zBg~89;#1PKxjP>7N7?g%t6~>J4@9;@P(8DKL}{{Zo3|0!3QseRD;5^&P_iqzB@%Kf
zGW=9YG{#FH0na&1ng2i=jXe??cu>U$dCwzR!02MIC`*WroPj^llP?10_ySvYX!(9^
zu$#n;#KiBOKS~Z<s;oGa<XhD7Cfat?qsB14Hv3D+7v8pBP91GNq!kDmj6}g{s97H~
zoiD&JSLuw>v)g;l`AlWF`Bv=vRA*I#7B#m;#0oF_KEahF<ru`d6-(8xK|UDi;M@H1
z)i>jCBGd=5Yh0B>5qn@Z`Yp*}T{54SxbpGN%uBh|o5L$ERqv~lin4<(L`4py!_!GM
zT1?R;JN)<6YUq;(mpqHwD0~Rp)SaY<1XALI8#EjoL{A&@!6z!bg#vsnotl%FSCn!{
z$}^eTTk467+|l_yFk(NAqg<iAk(JuXzQAdo=)9=>$^P)?``p8TC7e2f@vsD=Yoy$o
zrC87#s!5QUZXB~<c~7p*UsGCiLg5!@Cztk-l<fqKQ9k+{;T0i<u%8K~;w~kaQ0ror
zs}m%KIb;MAa|0E&%UL(T45r5YSUDSu3`8Ey@c5N@$Z!eFV6s|LCYH#SEf=nyq@WxE
z-?dAlMjD1w9zI_=-P4ZlM7P6g&!sXPg5>7blw6AGF=k`gNpbfnD!DoUZ8)Y##EXE>
z`G(ygCrr02_)QRGik7s#B04&$WUS=eKiDe2@6L%nMm<gvTxHF?INStf-1O9WB>aBJ
z-y!%_LtWq3pGS!3d{8$R(l_72Jz(5S+fvUoQ-0Q`bUih(z<pnSCnI*YfKR9)%8njW
zU|C47avD*qhF1sYna{uYwX&DI>YL4eI$xK(C?4->Cx0i}Msgn>0F)ytf^3ZjF|LfV
z<?3z+J3>wQ-;iJtLx-n3^VEES_C8MfTPLb*T)Acn|BoMbOJYBF&f3p^b4-+R$;;Tp
zERz{ls0TFePUK!sA|X!m_r}D_EAn|Kn<kT=$VBFSZ^1PG5BA<Os;PEs8w~=Y0!F0Q
zsGy)Um8P_)ND~noMG&GQ(gZ|B1B8SkAT<gC3TRM7RD?)ZTIh&~bRt0@5R@(nB`ngO
z?`H3JpS_>)jdR{}#yH=<_lHA=#+Ajr=9=@Gb6)eBMmPPyzKG!lhqYUKn?;uWlaHq%
ztNjPqs{vFToN`%(K0O$CI`+Wa76F70j9hVTkRjfmqq;=o;}()`b_wymno1kU6#%Xm
zI({72UFL<_c?wr^#`LVtix6@V<}_m<qX`L{8&xIUb@9{s?()3omcsiKiO}q8r0jrX
zf;4DAP?RS|3nr2rg2B2sVZ2ta^wC2noz?@jKMG0k%@O2)R2e~*r|XldV#whkKHgXV
z2k6;4<n(SmDcF&tijwh9BjPGBSJlkE@q*f0fmd~5ih`571>(}$gS&WH&hcJH2S(VW
z?;XXYSxjc|LPT5V>MF7JWzPrALh~{Q2gYzj<=1~<o_r8MTQ)EcQDG2Z9_<Q!nkUgX
zV~gb*vL1`fmIQC-v6@5dbCjf=T*niPvaJ~#kS9&i&9we{vQB0x+XXb0F+%K<XrJG|
z8yn#O@(6fV_7|Sj&OA1*0wB+{5qW$yZz>+=*%d#hyU$Oxc~q%RM6x8$bvwX3GRPzt
z?u?}4#Q{xSSHlS&Yr@FI=<Z<~-4BxIEYDjL4Y<#EI+rk^koI_&%%v}TL*GiNzcLhl
z06T?T{6Ov^(H)R+sc0eQHNu$~i$o}Simf|;M8!|*ACTt)vjQ9pfIRjLse*B0wBKez
zPIIR}rL6c*I_8$M&ZCQ<8`XLrCBrj7Q*0L}@G`T$`@cB-@s~xywySSQ@r7aFz{fA}
zv8|_hatxy>tC&&O6E=4eFU!O~7GvIAMv+V<>8i0c@Z5&i(}pH*GV+fdQmn1HasbS&
z-cg3i!QwONP3{u`(lZuP->-lEbdp9^6>uCNlp;Bjxot{<1M_!@58uB~q0<ao$!gX9
zR|Tk_uUTI)q)|v$YP{)FO-M1_qbs<bUiU&;pB-#YUpo|8xPAu}gQ%#ScP!w@L%;67
zMys2X1~5i%;ik=IVz-~cN&<g~M*tS-NPqy;X9dNMyXrKDh)$rq4HJa}0}OFH4}PH)
zGYh{$DBf0cIKD=Qj=KQ%r(ui=yboO((PbLQJu&B37L*q^iBbRa|L=U?|C0*a5^$R(
z8@d$2c2c8<=s%_Xw~q(;vE!C!uU{62$N+hY;i&Od`Y*%wk{@7vwKDv%3rPc&S$Iem
zL!KFt;DbhjHp&(Z&ixJrr9H*z!^Yq4MIuE9enJMkE!n@9WNR#G3+c1RLw$V@+OYk{
zTCg9^9YYv$wsIqRVi^8UE(_94?V$xvhWGw{)k;$xuwRm<6JJR>=-JzYW{`vk%w&K6
z|0Vj%_wQ-zvjPSVw@|}1(g!UA`4e?-=XeCXHSwK-CxTo675NuTE=an6)lO84@UGH&
zWk|G&r}k8GrBH%zy>r420>zK;Rau?a(0jJ^X>K1pt1VyCnI{8G?+8Y#Pak*FCxg{$
z${h5%v#y7YICo6Rj*e$KLGE<q85iU>$l?3X{a;?p2rTo@)i&&@y$s8N9o+Br&-vXQ
z+#Hg7S1huu<u~CD%~!bK^9PIFzM=WC2-p0~V!9CX6co(XpQ-AMjVhaYRX&hy^7DRb
zd3`FsOw&~2J)$R&AT)x8z>!u%W(?npp9kjW^XP-cw}1W4?EKN(%(8rxODFE_!BG=l
z2$j_BoK@1Q>({Kj-neTCTKVF`ReBX}97^?^&!TrQ@0WUE6vuxQ=HC_`bw!wBlE;Tt
zu#)q-fQo4r>>2RZrM9+T>}NeqRN{-GQ^*dqo*s}n+9*Zr>uSI!$~9NDv_7C%6zpIB
z)%VLQ7s>3ja;85AGRx)Etm?c_gbd25Df?+s9S46pgV?)Gjei^$OQKs5qDG~wq~+)}
z4E5cDlE2Xx#bzZx6H74CR1^O3-?0I&q5G%Gdp}$`mY`U3>UG4(L&TvqJ<^5(qB#ug
z#Dsju32`h1mR($#udVa<>uToBr@rO<1q&)pL!N@jZ9otcto?A?U3C}=@m&-xxc-zw
z&c4p__ZJMKnZvMwV942gjNfTQy1(X9*gLcAhs{bTrDvijE6jE-klW@6RT;3fV|TJI
zUXA(O|MaK(i2agOUE-^7c?DmIf2`Ks^tPVAdt()oF_Ia>O~xu<`zYy;akS^AvrDv7
z(bS5Khv`%5ziYvGX+NQ~Yu=YW%yb_p53{UiN;_m2zP6TZy6kE^bWtVTK<Vhd+VB5<
znb*m*jCpBbN)H%W==_!{lW(lP)?#au^5Lf1i&u|t3$8CBFfrUzo)Y%k$JzbVu){6}
z{czUF((ipnB9WV2vIE`P{&`T6b`CxBL36Si<?2VY7gS`>TJAfQ!;Z@oJUJ`WhL9vt
z(-DGo%B~|m`i6dOvE?oq+A4#kpYf`_#?grAU<q(WxXo{J_NKOGlz<|27rYb{(v-$f
zufc!+g4jl)W}pP?<mRL(HB}c`qq#TlTaZ($%?arCT|B+jqO812UKQTn`9lrvK7@Pn
zUg^51R}t~$|GcwnP<2BH0<Y^`R*17HAGqluAXD_pk>8}%Qs}2+3juhON2IOB_PD^H
zoQ~k}+(RS3UMC$cQ2HcTb)#+pB&{aKpyUP=HvL}B9_*YDy~l2rdOs8zw0Z2-%GdI+
z=vUb9Jbrpt<M3E4#Ugh5^YQ8qhq4!Hrk7L_6BWDG;@0-n?u&=5i#LtYZBP+PG7P1x
z!XMLBktpxAbEC&x&OLeiSmtoHr@$s5$<>j6FX$2KK<)%tMHWGjr-A+ogxlnJsbTsB
znBm`O;$)KbckJ|$B6psmHXDiS7T@T3W|L^B8$HJmvbZLI!aWyRn|x4^P6o$k08~+9
zndtIV%nLcvXJydI%|fL6&Ks)nPc3faHIM{q<O4-0_agz_t07}yyN9gO%hJp>TUGPE
z6?yyezMFn!Y!9lL@6Og-I?SH(PfZo#erSw+kYmJsseAt;BQ};=edp^)qw%|brs=gM
zoZ+gE8vk--Gsq|{>|!&YJOJr>T~D8}`tB_J>GVPrKe$~5uyy->oa)0MVc^hD)Lz{_
zxQ?{<VKd`l6IdCH3_YNsbEl|!qqFFnw1A!SaD^*ho4EO{G$?Z!k0%tv8^<VS%??#l
ziDd-|&rgjQmuFID#`oUZyuGjMxEtFXhQ`mYvJqfY&(p<n8v;MY%~?NBCOPO+b?}q$
z<6lJ3#@m>)YB}(uP#>6Wm1c(1p&pLkE6evT`@dS{ccj7DT1O?9f~GiPv~dMWERb`3
z>q7a{_4s#n;7oN;*dq9XsrfOs2Bgk(?Fq6qYUuuHu9Sbk_AXhxEbG~ESmKK#`)YMs
zly>&ic)#v^<D2PG^v?Z2nwS48M}@@e*KdW*xy8dEcQ7JkMU3L9)_tz(sT&$+K1-Ol
zcDr~!geo<2SsxFq&Y5oeyc^`!GALA2^}gK!zZf%*R$0fv*Gq#%BLU;^`b$ZGdv-{n
zDoPC=^!Cu_o2(+v0!O`vr;S0-yV~IVBWSypdlqXJpLS=e+<vfQi-1_4tllsPcq$NB
zfF4b^tjF%{D_=m^^cSwGZx#4>DoEBv#H6${%<u0z$o_M^(*OMV|HIihE+?@4aqSci
zD2sob2Qur9D?zC5+6$lmf~op49C=Npf5AXgwGctKCTw-_!j&Xv6r7Ck9Qbo1<wo1+
zkg^(Aec`~mBX4WuOWZXQpF4?y=bg4j*`F2K&u|_z8jBfKi9GjovN+8t)Q|W(pn0&G
z<Nkj5<WYmA)?6NChdRp}dIl(RKo}FT`b(7;h3ec4$`O0z*t|v}Xm$eFZB8h`KCDyZ
zW?|)L#wDE<cm6Q#Nvvo{bt`qlv5}yLvXhSFM*95)Tj6tyIEDbsaU0!FtGFMFt=-o3
z2+v1|6D99JEl84d-C7PwaUUPveW1x6FAxS8B;Z9i$4!D)i#Lg49dw+qoCj8|x}Uzx
zYXmo;d<@ZR@@^kxPq2~D5^>6oIl&m{!jk~)UD3_;(2Y`3s@;h*Sf}St99JBDK{)tG
z!#~0u;CYd}Ceb`3#{AOu>q@)oeui;CV}aZSsdzSZAh%(^-}be>SrA+}5|%f0HT{5u
z8jKa#TQ@~~hNNb#rDvdu*P2!;VAi59+8JDK8_TipF3d5J_v6KukBqhXA=-i=vq$Th
z?z^dqQJr7k-kv(Lwz)R*(^z80*-IET#?3Y5pqh}I3%igbTR;!36A}F<uiM!3^-u25
z2jY0@bo(x#3it*UU8k1XO4!aZD(b)+oGnSd*Y-}5^W7&ppe6u)Wqe;hs-x|S#}VR;
z5YxLf)eaJ(8hKkEwtQXx!S3=x4*aX#-^thxzz;b#&b;S1Fj@=uEWi12?A$}1>U{xt
z<3|u8N(f7C#^5oz?Zi!6yVU}(VH>H}#|cwArgC+QX0nwC7+Tm951d^+*vhw<{2EL%
zql!ui8K<Llf4S9Lbn^5Tk#rj*p0KSB1_`$~Djgo|ZF21oc>Gey`i9d9n1$shVhYk<
z;5pI<_plCkq2SB~?#Ps6H%FKoX}Sm7?IsLH+uY}Souq%5y*R!C>Jmg8oq>-)t{C7Z
zH|O${4$)P4F^in=f%Q}{VfX~<`3q)Th(z*g5%d|mB*gO9zGcwy>lPXzmEB(;(p4v7
z>lfwFw~^wSoxet(8B#0;r?}XrQl)i0=p8lej32zUN^yRt%&*G2m}6hVR3ul|IEEK!
zmYk(BQ0Zf>>VbF{Kr=5~EvvJCqnzfHz0x2ANKn2t(i4IV#|JCVT^+q@Dy=J~Ow9cp
zchh5=Do4d2kMUah^Sj1|F1^-~ocR<~RQHb_hjA<uWPho_jLLtJ2n{kZH+<P%_3hsF
z#}t>$aaOqUd`?zZ5@{Zx&5QJ1{|UKoyNo$OU%9Pz*Z0R#KB*Jfct#^qv~3&_gO%&O
z2HRVkXa>8x_c-!hQQg~VGyk~!h`xCSAU*FPo)v1r0nJgR%V-1U<<F%NU*CMbr#oX2
z^6vZ(uKJP<(~z!#5JP%(3C=hkTxZUODSms*^ssbjU=sTRAkCAv4d&Oz_nn?c?bwfF
zGj<^gF}r~=G}V+91LYn4wHz4=<?iLbV3>$XFt|SE8;UNxk}Tuy^fKwhYp;V6+rzh|
zY{)LRXhRx~;eN(O>|-;TBhjy%+unPB@fUGdy55kTZVDtee0`T(!0vRM4|uzyGzKh(
zB~K@gFn82r%lL$VowkV&K?SX*;zC?GhTyUmE>1#?7sPV;JMk8K4%Z-icl|ZQEgX%+
zJY9L`u3TY#0aZXMQ{_BHfwbp`K8_2v=`OpBCrR0fp%c#Z?U7BH4SI;(M|!guV=DKN
zj%!+kd&~K~zeI6yYAg_Ohbe}qLfM2?QZ{_(58AIzAFb~aExKv;P=?SOb(V9_5B=0@
zj?o8qBWz!VYi(@v1IB`DIE|5Xr=lo}E%%%~-cv5wD%SSC7unQnA=|}WvBk!K%_B8-
z+;?o<S-I;a&$Kkk<KiQO#nMj0yYv{I97zgmFo5!Ec64{b)0=)A+T7EAd~aO}NZJlG
z9iihA_iBxK>iF3&RI7Kw!a$NB2<kJb7Z?O?r~xl>00?qu;@UBEPwoSd{dGrUskjC+
z++;roP?Oy{H?*a(2y91@GPFi*JR`(>fMV?cBcbOv-egmV3tW0Y3yBHO09wC#oFT}O
zXDMpHs{>TQ6T}%+=7VQ`JYneb&vEkyHGjeC0sD0nlyLsWehs7>aN;&J3Bj-m5MhHZ
z@V2ef<u5`}bfOg_Ce~~10))L0I+<j1^g&VgF@<b#R1hT3J4t|pr(-$>wr>~uj5!U@
z-n--M+L$5Lx8Hit<J%wd-|h)VX4-`VZ(VbVZ3z8=OF$ND^BM2!F7}kfSie${Isq>-
zB+rl<fn)(Zy8MoSk}<udHT<Gr%4|#K<k2<_3AMJ~sxdr3Mg$ngyq{Q(7epC62r_!3
z<@c~N^AK<Lz@!MdaZDO4^&rXB$yqCVXP$??em`jQa|V4Dwpji}myf-bPGr_Vc8vH=
zEn_52nD3LvRZ^N?o@vA7sLWNX74BiyaZLJP27K3-&rSmg@U?34vS0xe(5$q1P)_kg
z<Fa+X9|~D>A~De@%54~|vs#53=#EYmpUN1vzsvJ2qYZBqs&{T49l!VjP{_^duUrw?
z=FA8fvSVg;lYF#e^k3}lZ5tw{t~m~BV9_41>ooRjpzZ{kKi#}u)Y9b|4=`3IOIp!E
zRpUO|aWX5xam_U%xiH6Z)eW}r-i}rjQ!ESYlhk8ZvIRf{E<nS1HOjzB1Yu!9|5b7j
z`bMCi*Iw5lAaG<g+=>r*0zV%MqN@RYjMsGVFPJm82w37{TzfXe=6yB;vEG3MGMVc)
zwz5xij$FI7xfs5z%g0q71{7L6w}<ooRPtCB`XO(cXTKK~oWGz67R7_*Sp2+tGJ?fN
zoRSBUedl#iY~hVdrIT`T6o2zrW+?Pv{#HTm<rExsGbtM<^fBNeMS1)2fh)?qzhIg$
zoUN%4usbB6a1fmUf)GQg0ZecSIoT70U~e79sS*onfI;oSCIO=dkYIXG-8pdACAjui
zvq+|R1<&pn0*@#mMYknlO(BERUd=BTQ*2**{&@}r@bd)uF2W9)2;z)EiRB8A*rp#;
zR{joxjCncE>@KW)d2wO&B@uM<t`i<QY`>*aY9ZA9;q6#&T_h0WDkDw)-~@wjMY0XM
zg{a7R_N_srLXG%g?sU1A6-SPg_gv0QAAfDC;J7@Z#T=)L<VQ}HqmjIgpgSGJ)tp8y
z@%NGDD1fL6w#yrEq#RqNo$(8zue~<irUWHF7Maj$bnq159dpdXHHdSBJVS$McxjWW
z@`c2>s|u61-_Lx)w@EIqaU_F0QIsl>^d9YeV2&L52@#7()Ht2`v8}v476}y|gS!qR
z7H^109Yee>10{(-l4RWu$dTq416FLH^4=8W#r0B8G{4&(d*Jrb4`VQ17Xe-;^V9s!
zc$bLN%-T5-%Ui|#dl*qMt>z(MiPSm)IHI?V=5&y?;vYO2*V1au`C)P|_LwN<C<qs8
z49`doKihIfNtrP+ducjdfBB;)9_yjLgte%910Kx~fRglQXwfzf$V)swAs2%Sx$5J@
zp+#mO1G%6)VUq-JY!wd^!Ct)-`c@#7byjGB%rSPS+Hpn4*7@E75#~e;=pnzX{tLEC
z4|^T3-||d#ab4<mUBE;v|1**U#S(8-j)WamuTl3ry6o@~N<+>~P!xh<+2?#hHdg{d
z^wb4<dE-wAyL=%%bid?1-a@jaV!<>Xx|`pMU;pL34zx9=!8e}5c!A}+4g~O{yM-f9
z0ML|HcR00~RN^0Tb-(t$l%?af$<~`KudZLj{hbL2*g&A=S~?aFepI_?!{msGE)Cbt
z+a=A#E5gg3>x0@HNnf6=D(G7lbpxwt$$__y4MjEOwtzquu6sMH<u_5p#SW9ndvbT~
zoO?cHc>AN~VthRxSAG3AI7@7U!dUrK)A?DMHwE%qp$e<S>6q|sh2W6dqT9H4fcl|N
zMJ{lq(umbgQW^{_s+A;7zJr>0O$G#($s9|G3*hu5nJ0_;u^o8i<HV`hiAqiyQU>?q
z2EbI7wK%ge@KSLy!6|FA`M{e#t;;tr*$Smnry@9)OGO<UGfjrSS*JLZ;J;l9Ng2NF
zTSAOM1`r{&*ND;t)d!rNetIbRFV*=vqKr!~r=2z(wmTvV7B^j+X5EjelSPO0S(oV>
zN~QT^?T(wYN^@lF`HYvs_Kb;eB*6eMR=?+>wiC53-VyL)?Bj@2i*1N9&lW<G+<c$U
z3$?+ZqyF;0N+UfF?Ls!bg>R~?5oe2WZb8{IL3xn$g-?}FBFU)|_OgF&1KpEBlpJsz
zj&%ZP$M2D3$9_GR$r;7k#ArANE<@}X8zwW~_u_p1g2l<1*5=N@^IW_KZM{w>+Sq1{
ze4-zkaa*9T@wRj@gx#F{Ys@FYw9R6VrVPBgm!TA%7*8h-H!<ydH)M_FA>Ss7Tb}+K
zGApp^Q~HHR3Z_4``S~UIR*9S)ObbnhgE1$^>7&vV2+-TO_5oGCOS!x)p*+$=CA$Lg
z-G}SA!H8R*1P?a`*FLe)-X=B!M!YrZJ(6yXTGt7a=QU=40Za0)`y2@!o}D-ltjDVI
z;!klm{N5?x>$9fEm=Yjpcc+J&K7-mppf7Dftd4``)9?Fk(M85~Z=0j;)K13d_GjpE
z5B;6f=2XKuxr(3=*bd>yT<syxlXlyA-$~m)oZlOl+r1zXVR88PUJbiC-~PEINH2m;
zXoQKt=L>LLDfN;Z5m1P6PFRG`-s@buQJ$^?y+jO>G7(8t=e>}!#;*CT2h*!w_aufM
zCC+4kg43fgGWT}Cr0@EBE-=wa5_%a}C4wcJ10Tiz@Mo06haAUO7S}SHr|6HPL$x1*
z&!7g7bAu=_E%1=nfLroLZ70pOkXGEumFNVXIrjpu7R1Gkp)4Jo3~90eP2iCxc|lxZ
zz$RV_Fa>oY;$GrZK24gF769THnX@5^88D>Evz=<dtxlHl&8@D3%9r3XXkQgM3S-3B
z(<_y-Y#W$=c}ts>%bi7}SqnCu{DOmnuEJaGqUsLN9~9@BzH?U*)7gIR%XsX|jh7eN
zi)<x84!E-Ga(jP$sqDcV7e0Pe754Q#hR8va|K#Bg-f9n}6DHyf7L{x1yh6gC;)5OB
zd%SY)Zj2c87~T@_04vQ1F%*06&2Y7n;I?`QMgSN11wnTtCPmec321N>-Gi86DAe;y
zPn4Vo?rTITO~ykmysZKB(;NqG?tHmX_0>DC(7~buy(N?W9#^G##s^|K5;U3_$C>*^
z&)3RR+R!f6@`uLj3y&^F-)Q3q>#?MP<4%Y6m@07iaN?tsEI;4o3p&=Ht~}jAP~+Ji
zL)_U|D!74p87R<iYVC#GnHT#tG_QLVm8%b<*WR^#QSe7R)Pt7Uhkv5|`)F-Zo{XUS
zGWw+B%f)^BYxW3U+itWR&$R&-`pwb^BhfvC?AM%FWc|79I62iodO#tyE<QB$j9QUO
z^r1vaM0gnQ<WpS%+;<{;!+)F9?B5Ah|1#f(yfp%ALH@tb2xlt)9XSe?<ayh~XmBBp
z1vj^xC<*DDm5X`)Ud4a@-AAE|_Qs(TwQ!!|enu`Y*cprCh9Jj#K^Ju@Eq@BwN)EgL
zNlmCTB%Fgl7jE03Z*qQM4#4x}BPtN4UqeCrE$e}EOh%R7$o0#!Aj_r4X)`}aF^b%7
zoV@xGeqOF-k+&60)j)I1YyBK=?Sy9a&F@6E`WNGZFC_r4jP4d%0zpMv35^FAOezfG
zZkn}KckA;oLItSRQw!it0**6;1|S))DOo-2^aCV_1?C!F1*V1Wy94y9E1yW?bBJNw
z*GOsxPSn)EC&QVUEZ-Qy8Pa6#+DWlAxzw$7+s-(ITt6mFqNWgJOc5F>P_Q2-5I$<(
z$CuQ|+A#@=5$_0FasK!ylx!OABgZws7i53w&J~*Sri7yk<nkt^#}fj{%)Q0^sgTwV
zm0Y^N@A@M~Y!gb7b95lHso7AG`8Z6r-8xrU#xxf{Sj#_-T8Lu!Qh6`E^<!y@)`gm1
zlx?e?j}8_8coT9~Xa)PQ`0d5B?+pcIBxcurnbO<}-7!j6!?Jm>L`Xp5O7mu33;gR`
z{qos+=g%A(tjh!ShX7E2;5;V-S1?C`b2CY1NGcw;<Yxf<T>zKQce#tV4PXjYFj&K4
zMgb-Xv3m?sp5kQLu5qK~!E~oHj9o$k3qt<JIO5t6JM19MN=_tJ2|VfkF_u=iaV2jH
z3E=3pW*l=fz=LL|xX+2L+8bRqPR5uRNMPgu98qW93r7Oh6g3~Y6vR{h3x;eSV1iPM
zdy|+qX)DINKvL9%nY&#f`NaE!^&2qqKR?I<E=T~I<jA~RyX#8%bk}4*-&&Y*Sf7p6
z*j<pf!WoghMu_+q8Zo;cxSpTHw>s#to525=p;w_Dndj+Gtl}e=kQL8!-vwPD^pdt>
z^>OVkjQiXNe~>Hgjw^iZow)WQj4?>gE+Sd59QbXbCikuG({f@s)}ZpK$L43og!q?x
zYERY7F~N<c$kPaHfy*GgzX23ma>H$1H?1C5nF8T2A1&F9r2M~N@a$<rP*=<M+Nk28
zI@ZAPPTuSR58{tDSHy9ZfBY}lI&6ltx**L9SKCxkOYP)uIkma;l{c$sGhc1FwG~TB
zNHS?08C@p|c?O2Q^$O98iCN=cTRxGnL-}ZfabPA5>0?K^(}XHnY|0{t`KH7XUpv?$
z&p0$^91aZ?iS%Vy(m=%Hr%AWjTCMYT5zIcSJm_0-M7v-8aZun7NZP<678Kne_G*f)
zWeZC&=Xgqq&%qc@<&7GhHYEK#KAgw*c%qJRy7D%W?=IVcpgLO)<&6h~lcekPf4*i}
zMz0dZPX7hFXQ@DF`9geU3Jj(}8G8uoarg81P=om75tAblja!WL;;-gb0gZ+i!rYfU
zsZJ_Mbeeov{pW{{u-kGbp9F5b4*Fep6F55DPm{`pbJQubsNaWq*J`_%ufSjWdXJTA
z*FVF!;3gafR$qT|Ix!siDEpnsmacPdVr+g8y?fC{ylOcZ@8R6U7^EJ+$+R8gMi-J&
z$rO7=07tTtXS;8a*EqN_%?38HYy~!<8b-yh0W{d-hw2Ukd!y-R5bH*Z0yvTZFTsZw
z1*YBzkUA>q4V^S!5aj8tZXWb@R-c}NoO<DMi*pAB5MZguD46=288^a9|4cx_&4Paa
z{B{A`o%6V@e;d~v+;*qsQE2~5kYc=*{vtp3LVYQUTG?pgPRnn$)oIA&15P{QDl8+~
z=wdM!jP{QJ92A*=*^(q_negdc1Un`l*VJk%0Alk92>WuR!?^!+t-1f{bYWb;eg7N4
zKaK;u5m)NsLsXeV<pgoF4}u~kS22|LXkWd8nj0ZH1o=0}51Lp1&IXPl9B~cOm>|<a
zdS#1}^u2we`=-w%yohnQhD!?tF_B2MT>}f~Te-TQGUqOKzcx%O|KR)i_m#X;c#QaE
zX^yxK)0)vW-*BH+I<XqPdmz2vWefg%)*5(q!{A@*=9QHNu7iCk{8Tss!c*6h3XPAC
zvp?$=-nY$|467;N|N8OI;iHl)PkJm**xXM=$^p3pbD-N=?m$<8(-+h?|Ak+Iv}tuP
zk+BHv?P@|LW#)aa&K07+s6Dv1u5}ZoTIs_V>7!lwg%~&^l($u#5j#Gef`acQCGS><
z?&Gr*{-cl#YO_Y;+6$d@U)o_J6OmB3a2>hSVc(ML*|*(NZ{eKfYjni5+9Zwgy@=UA
z694qtc;K~hk{H7cuPwLt&*Lz$e|v4=amR&N;I)}Ay-okaYfn~S?p^OX1>@l#OX+QU
zFKKx!chhr}7Ox3eQIx@UMjPJ!9B*ROYUsSb|K^21X-76)RD6=n7_58sU)u}R{^xg4
zciZ$ozy8mk0Vx|be1~_2JVE5fdh)K3a^Ld;aeiIvMyPlaQ<1P9592-t{^=^NJq;3?
z0aTy7osI)~OO+gPVcuzCJ<^8ARZFRRm$Th-!jl7H{|+4a34tYw)0J!EJn>y*ZzJ)N
z^|;y`5Gzt}6L8iN_{?$Q?<i26N=XU>g*lL{_OybbY19yy&$G&i0Mqi~0iYbaZDrds
z*ktg({UvN2NL{FdVi}$mum%uDPb6t4@;Yw)s9h|9DT!Z~<5q(E1r>z$Txgj4g|u>o
z*i0Bg@Xk1nPjRE&c^2Tid{NM9EVqSc&shUSU{m>adKdF8!!NhFrzlsI`6}gz=^dTx
zjw{N(8(px!V4p}|>vn*P=3~E*adH^`02)r15?5oB^Ugqu@#cDURJa7fezJQrf+IFe
z+xL6|`G7)<`2FePU|AMk-w3lb$i4`EqPk@Dl@p{)R5(r%V-E!TPE|6Hu_J}V98=qY
z5AAq;<H(+mCWemlu?;xrfMdg0oZNN-_w##XG{;<R*z|dzcuB$CFONiDBYMM4zY?>7
z^5j@TCaMuNgl}0!b(Bmwm8II0+9MBTUc9p1M9`B1>MoH<7*QzHcYhVbd35A~ret|G
zeD~wx`Y!_TR?;rqP2#XwGiazJ<YxZXo+pa@NU4x>uDF-Tol;&jAl;k@Hx3vNh*>oD
zU??|**k{`Y1_r%7R(<h0?VD>GT~?2)16dwN38Otv_li{(rd6GgC-Y0Ln*h}HoQn?<
zb#I<w+-YW>r+D8J%<av2zJ+b`$UZyjM_|+_$xV%);mH%;4=^JQgk+Fy^B$xD$)S?~
z770L$Jh1-VqUQd8^MP>Wxea0)2Oy8gw;UtfCTu_J$&7qhhC_KW_Xg!&#=Wo(!nt_X
zeLz@Zgr2b#ZyAFV{ec8wz}~O3Qn!j}?H9Mz<LZ+(6!HWQa8rwrUc?9O@4a>uKm|>|
zimt?6++2Mc=B_z^*c^KRqX)%J;KE5eL4!(cx=TN!XvTmq@%+2<w~njLhD0SF2fc!?
z@sKAK!I6v_4`@bgcebRtq|@<n=wzE4Nx$OHAJHn3<~IqX)Z$x^mH7?YLNk_U>lkU(
zC&bY%Xo{n0z1rL{rDu%XiSJ2t%kjEWn&|Cwis=O5Bd3=&p6)#xn(f$z1xa~i3r3y1
zs4-tI3P)AcRHf}UFE78e`S&iqHLfkjh#iU@%K!pOk}FsRI_JFi3!&0fbX-^0{ax%5
z1AdvQ_)&0$E4+5O%9eomOdMYCIZ{ca=W=^i^u+At9WU>2`#{;VQvd#mrC!<Q+b31d
zj;mZg*mccFF7uxoj`+XeeC^f^2w=~YDT*sVUM7Lvlc$ie$T=o*3e_XD06_1)RayXB
zr~|yhnY{Zv&<1I@q=SxP8)urIbqkd3CWQe&CfDFtTM7m&%>cyF+NcvU0Gi%t$$bIk
z&WrGvMH+bb9Z{c?bu==iIj1Avwyz&|LoU7tNqH7;k0qT*3wTgtd27@)uaJZCz7|-q
zhZf8c#96=C)qN9l;pOOzT()q~)eR+OjUyf=<~U<gh2wlR5&LaSN>B6flB|}l`sd6r
zqw`K*l25unuWJGGNuPk4s&7*jHzU4G5$V)-Ls~9uWoux7+Vph2B=T{g9LJ9vjilzs
zjw@wcrJ~|a*bs#4jjL@P_G<3U<ZDcWr7O56r?C*Pd81<jdbMJ!E-1O-EOS}Y>o);r
z#5~S~v9^=8A*ls8>9Vo7v9hv3!;`$Aq6nX=yKTl{FI_8X%aW@B8wV)pL$Avk4^4uP
zOLTsiU3*+}c((6EPn#E9$mipFoB%aBZ9RY*Ab>8Qk(;ypR8LiW8q{03Hh0*H52>S2
zT76G3SeK7w3*JL}=>Bb8z=T9M-t~HUDdoFplhT#_DV86VC0XKsZxDK5zack)j}DQk
z%vkE2s(7cL)LcQ~r^8>YRS0hao&V7&+)%mwSEJB=eB@dH+&A6u2~4)|#-)phv%akB
zkPgicIicD(CIUsBeq!r=-(74wvPME3ULS?}_JVLV@V%S4x3{=m;C{;EoObtV^#yQ2
zpmWI=20DJ04RuF-<2XAPTYqV4M%_Q~>HE>ivqFDFEdv83KkT&qx7OHg-v&RVe^=!F
zIs|VB-&#Ss`a`8dS*QWjjAe-rasF`VKuj8_I*g*##}&HnY|a%n3oWt{x3J%CW-0ek
z0KtrgVXm@+FwVu$0Q(G@)SP*5v3Nc1Ms)SM`mf1QOgZZrI6FLa2N{LDfm%?Xa~B;g
zerp{h|FLy?+nYucyxzJVTMMj)*~NW>)%31Y>dQ&Dad>L?)~zJvaq||>pCblh*^Vo=
zROIF#7-Ra4r13&ceQ%cihkK7iaN82M!5GA0vaKOuzJS)f8S~uq5=li5LeTBb^V*@$
zRCe})@cZFS*pj!P$hC)anXZ77FqMG_jkhbltR|WIf`~UV`uYYbb4&q3or7E$(hfbI
z)_%k?zvB}YkF5emzT6|x!of@_&OwAvXxGw~JYwhAkPvTMA+@@TZ>CNW%DcZWf#`Iy
zV(x9Q-vPHPVYD(#{@e^ucA+6Q3Nv`2nUc%x6HTG|KU)I){FF<xZyixMD|DP4f?gUM
zpPH|vxt{Sv=@kyjywZ$${A}Qs5VoAN4w+Y?OKnUmt#FK47Li%TC2~C3LFi#SIw_{?
zU7@jw&yq=c>edmJZ_>Y0P8@n1{T+U4J>Xd1@7|igtKPS&Rbv*!`W2{=Kw{wL-y5at
z9n6guZyeBXddBepHFC3j-#c@!hWsD<EgXN4yHnzih6&t<Wl%r1qypi14PY3TrRlpo
z$J@5b)|quXN-uiGP;COkIBs6E*9MG>pGp0IKZzLl*x7wns1M9#z!!pFVFyPQ$>p~D
zT5WY_4190WJDc=>{ge?|A?8?#9JDaU8IBcqEx3Q+eMRn+zElxcF1m9%*g58T^(J!_
zkJwz8A$cCx1cnqV8xRjO{mPq_)Lczx&6PX~Cr>Ad-#>2!<9hK^Fky5s+$*75(Z@MA
zUajcX1FE}Mc<w^0#1>GfX`3$`#}DOC#KdVoupH7w?HF#qmFaKoaz63By|v1tJ1=}!
z)dBVOEIty57tAFjkfhpzbbam}_m#KmHMnw6ol(l{=N1IPDIF6zgOs5UWhN$a!#1Vv
zpJ|DYoKk<L=zyKN{V2LQa)+BU=M-Zw{f)K8AB`(F+}<ZDt}{Ea4USV8oSo9lz<#V8
z2RVodD|Bi1#7&r6?Jj?It8vTsUpT+vf))LZW;4ZEI+m87nATr_^ZI%VH>$@jWb;>n
zY?*4Cf224OQPhA+L{^exN#oMo_Z+{gr8iWq4<&|8igJDDNIyb(+pbcH50G0lz-U~k
zOTnd;wX33&+j`a8!eJC4ApkeMUe$C^;cY;S6Wj9wM7#%CjZ0R{O?DdV@%B+6JIsMB
z^9EHIGXp&3=DzJj_FnX#A0p|I`kdDnxa#yQW^jLoC0NK=-4-a<^sc!ju3mw>;>Yl;
zA0+L>D$MZ|sx8pOLY1Tw*^2BlV`NFr02<UTGKr02N<O0s{ZYd{RknR*yJgSI+JR2v
z%3|(U)!v)eg?3Xdx(3WI6TicUq<Py)skN5%EHDzh-&H@B=+s;>hkaDqa`*kg*U^am
zu<f<JqnxYY{?s7&xTxMAR13k6aAUe~q44#MM@(6;vV9sC5TH1OhBNIME0woD9QKjB
zTOl<JqTR!L5sMAPsj+(0k_q}guhC(QrRO_KpfoXS85s)$v_gYMH`(6b6?E2j!Q_<K
zP@8?X_TGcJ6R~NaBpQ8peFYyEVv^Jwd|$t+=HRJk-^rIEKXMuXI?1XB1Awa2W2<ur
zUZnaj1W;q-WNPIC1a1^Cxw?8E0Pnvm0U_6QUH_#7v{VfSyg}{6j~c*QP~9k$#1yRm
z0<UEV2=Bvad0w?O*gHody?hUIW@r!F@NDsIgjEp3OAAB&&>0J$QLwL#*2Mx*eh83*
zV0D{u?LKrIHyRvL_Z-Wm8#gJGk0X}>P(b{hO7-NV6NNx*a6oXklk#eXc*jESb3-iw
zqyTsi<RFoq806<(p1zl<rqeSGJyAm=lVR+I1JdqYMWTMV2AMhyZx%e+*pBM^4wLoV
z)pTU*?M=mrg2M#25Jyl{0RvxOk=~@p`<eEc?-<)NotIB*hKqb^I3GsVgN}{^epN0N
z(ahF#?!3{90ZOVuV&9_AOa@ea4SCmgW`&P?v<D@^14A(OUC|)nsX62>okWyxe6fGx
zXdy*4WxvxcS0@oz2(3H>5S6GoC;^Oco^^2W1mtPz291{|z4kPT-4wouOG0`nnlbM{
z{Gk&jJIAk*uUEwk4xlxcC#4$Zj}Q}(?kGqtqg6Wz2w><@eIboiHSLk;mY>fq$Q_W<
ztZxe#th3+9b#8V#GRqkKIxPZKj?rb^Jx<=gPp1ncR?}nwB|2FLbqm*^0($b#Z5o_|
z##-#ehqFSH&PXQMls-Pl$eHTw%H0gK70r0UCUmB)&A4L>#zn+Wqr#xTQ>*3smm$|m
z#7%M|0A;p|37I$$WGPuqRmsJ-{Z^`f{OTLa-XH9EO&rSPeuHRq!Wn3cUfFCa(qHk~
zu&|)Mw8Z~lvBHrxNzN?B2Qr!n7+RZf$Q^$kGUzKmC)?rj^Gd`X{wI%JW}@26fP~c^
z!!Iu%N2!a%TzKR?ux`%=aHGMotWdlu1Q*5AlGRMLi?~|J5U}c%wuc?~62DayUz^GH
z?_&Of)Wd)sd$-VB>`GQU_KO0z^}>sPq)1D-yfP&?+fu^Eb)@P*L<(4_Y}O2Z6i0lF
z+YTtAa9X;`E$_ZZzN<+U%Bi(Pqv4E!c4u!*1{2JXM!xr=5TgUYj9_xGR(`FS;P=ma
zB*b2h57@^Z1>3^Av_WP{HHZWIR~F->Ol9fZnp7L(h$5nl$F9>$^C14pLeh*<Ga!<$
z_ygV5lun7R!(*JxxVf-+3rdT1y=r3Aci$->FG}WQo~ix|=K2z(vI!_Vjt*VvIlZku
zL_0?Jt<-4OnF=R+@i+6BHpWdtYuy!aV}-c(*K=d56a-hMjBot{z$y^nIaNb<z^i=Z
zFW7UOSviM+YcNGl{^2Qrm*@;Y=G_Z0Nx>yqjVW{48z*3#10ibs>ygmMJktZ{X0b@U
zf_mH{o4j(AtyI;{*5VxUFnTiWFe%1czzSgdMiBub94uoXYrid}yt%=uZR~Lti#&aQ
zJ_<NED>}T16ltnR4@fN3Dx2%v)T41VQ^Nb?bZ!ha@E9TwA!cl*`k+Q|LeXDA_AYyC
zQL5?N=vdHO#Z9QjX&VQD4)LJ>Yww~}%@f`oMmul+Ib)hlUb#x2n=p-FpXI;}nGe3*
z4GS3z4HDlkLJB#&#YBC54qB%#+C%3UxaLvBfX3^-sfxO^mku%`V)uT~ek1}01d=0O
z0M<$p8VM0t1%|q}{xVeRR^pN{*PaooWydP-0>dJ~5{DYA-ILr==3x4q0?ri&%*~=k
z644|oU9V(H;x!H4eVE?cJ8OQ!=53pW#%77hwH0D{0MZ8@Jt%jpHlsBJ^~l0vI9hX^
zKNGfU$T<M6@sVR*%BWm88kp~FRq}ij{NvZJw+(j6gN@5UYzJI}Ge&is1RtFyY|Eab
ztbFa(&(pa1h@uo7es!5AxxRNo9Vs^3VGy^P6*?n+_UY$-t;fYiFK)}$tpR!wSds)R
z?x8fF9xnbMffmylOYOxaVLuv-=&b`^DR7}bw-Sn=C0IOmuu1Sb_0qUvWRmJQM+@r%
zldFK&@!vGRN$Nnt^1qRTS?OxETi^Ra4YY>0BGYhs!20(};C=uT^dYb0Gg5ihp7F@J
zM!R3A&TF`Ka|BZYxQjBND!GxPr`V0d*@D9$%`0|I3nXm@;U*O_<2Bp^cum96<CtW*
z_PWprj=Ya&LCPdjkUX0;JaS$q5O7yAd1*oa_x}jvj&UQw)q3j`apVNtbQ>{%u|sp5
zdwRQdImm^N=}jO$qBxn3O+*$@r<7S;yuTwMA^4h!%K3bT4CvpO(RBT!TL==l?;PYF
zghv5Vhz--?@pbNVj?~2%#5Ff_tO%A&G2keZ%BGr=jQqz+eRcLB53r}0pw<!j3KAYC
zHH67Gy6II^m-gR%N;wp?4d^lbw5k_w1GsrMOtKyWhk9}_7`kX>LXDoXP2Nwd9`z%;
z{#?F(h3gB%i5tg&7ibWtH!&Y(xRq77_(?YhZe19#5PBN_3I*EWy})IiAyIA&Hxi2R
zAeSune^7}PyX2qdTDMUB;3VfMh?@=e8^o{1Uhea$!FfO1@1MtK9MA_3pEQrn0*%&q
z&KY`Jvt~e}9OB_h(&6c*t3*m+9>?`{(&ZzFrxJYK7(IH@R2w2$&NSw9<>5I>Nt;Hl
zN_6niYoRbvoVDo|dMzmuqxi~8>V3iHXP%<3{A2|72zJrpIlu%W4FCaH0&`}Nx7)S8
z)SIQU@2lRUTdI#XJ19~4Q6;+n8eiaJEOI2x#wU4(+$ZI6H?MfFiW<%er)|@DH~^qa
z4zZOmlrV7;xdm&2)@~sS4vHRLzkWMPRU{A2i$?XT#v2=XPb`YgW@@swYOD%-YkD}y
zuDhN}UHwgZ0UzdPx(||WXeT%)#|HWB?*9Eazfb!_O3DeR1^Fs}*xk8fh@O4XADp2m
zi<15i{>oKlRdF84UQZ8yR4$gplH`zAk<k7QjC{}CMLvv0vJFYt1(RO&;aFKkT=t`{
zeA?}LtV82K<)?|75m(kqDrOWGSDs7put*~9gLMm$$Sp{DB=sEv&fAJ{9x#@E>rJL)
zFntf$-M{vd9C-G}h3f1Zp8_aE-nIs6*H#GRT9%OOt)NBm%N9L*mn|IR^^q+x2;}%2
zDpg66PK>WJ!dS$f?z>93#E>ezlM;OGQdYV<P@mfd<3%HHj8PIAN#%%`KjmFphkOPH
zrFx&b*`E4wQnu~v>qA14U10flMOy^P!>CrjuO&UCrL*<r*7T7w_Bj>I&V%2at}t8;
z<_qZh!%Q@!nu9i#u&;_aWHNJlhrgobvoCXB{xN@^jr|@H+lC8Y2+W~qw$|o!GAHe2
zQ=HPAblsObVSc=JQY8sUvJj<wo%f~XcfGN_k)X1PdiBlaoo%uvhfo(Cm#yj2G|xDo
zpq(Ixi!Tf7K{=heT2=PQ{aa$TwX9+sERjj1Xg>gZLkLkwaa|YhfYliK?~nDtl%U>2
zOI@W@{$q&K=u&X#vBrUKI`ZeL^j_q7Yr_?K_z7)=SiszAIFEEk2w)`pbn&u~RIADM
z!9Nc1i>H&~@JKD$!`73jk5MmQsBx?7wnC;~Z3@j=@BKbP4wjnkEoKQWqEb=xvjM|#
zU^T*yAKuP_HR;$Ly{?%fugCvf3hy{rUr)Vqb`!2ay~Y-o!y9ncE6lRm-Jd#7-fbu`
z*HvMTT_R87Ibu`HJN^DOw(mIS1{PFets}Kdz2BH0xSq8uLw2^>dMXcFc~*#O$C2Eu
z8q2I=YzaDG`$537PZaeMLAQJcDq}YakL_=FJr}{`Qpn9?bSF`<d7SV5fJVQ_MXNFx
zE%Iw2Vaf$XKZ6>kH6cp1qk{-r&~;yFY*XgiLXXdmU!2n7i*M?;n&<#cfWrT+FJCtX
zn*P3ag4q5F*2*Y6Jy^MCGWjS68iIe{dY+FT|AgF(pt}UD3vhrVBfcZ8xWa-1IAS=S
z%_3-W&x?XFf;6@fDiNfOpf0OjrwrQTx&)S-<NN{)%MTvF8QkO?$C8JOD_8-W)%XC$
z{}ucf3<Yq(4Puuaw7-k<90#t{#n8h7pxm$vIr)=_oTKCBLG#}4-_3hPMk((UIe|2<
z3Y>ke9Po%gQhBy_K)F5=fTC_e91;-4kduYzYF^`g1nW=%3&xE@PLfTPa6b+pxUjQA
zpB<lLdk{-z7=VGUBoMd~1JmVGs4`2=qp=MHv}X!%RP|s$k{Axiv85o7AbujP976TO
zV?o^De*?;qWhfSm_>0{DawL7PvDGzcUW_^`f({4_;djSa%mffez^F4wJp`{oakz;^
zN<7=zdyK7qsw`S_p?%%{>+^-xBgAvS8(LsHMirp6x4r=S*yOjpZ_B-$;3Bi{mvXjX
zfQ}-hT6h~gXm}Kh7=tx6r`;_X8t3}=$|z?ZN?XpoY#P|7l3e~B9F_seuU3cYPF~gI
zZ)rVA$9|Y_d?BeZxO~gFF=wa6W-!9^iOH0zxUxm_<%8|8q)oJ?0?A1Ve8HYx(46Tr
zHiYgvt^dmmukT^eAUl1=_X|%Su<uRb7?nOPhdxJ;RpVVRev=E!tRR~%T?tp?{|dlG
z$*5^bFII5+g^bGsxuTcv+HUPq{k>#=S*ne^iKK_SX4LhO*QX?$dZe=(^)134i^SYu
zj4k-lYEi9*X?dLkd5NcIM8rzX-Tk-UvXW8sRcAW(=Yr_vrOn@|{IGxa*1q3IW~m^j
z&O^xnFmx>#cRgz)-o^jj%{IjtL?+v`&J!G90vZFwGS4wW9*i-aY{_S`BtL`t37{iw
zZJk3b1XCm1yv7Vvd7sO1C-gX9F~ZPG?FM3erclDGs0GzC*+wVjy*-Q%cJ=A84^dIj
zfv!ev1s{r5s#tAV|LT4RF_A~bPCr<6KZ9~WZGmoY7n`AE%zd(YRO8CwQSXdLqdRkt
z#NC(X96}|5;?QFAC*BSY*b5Q=+tcbpz-A`??e=ZrU?TvSDw{yTe?W_khRJwLq^psA
zS!l4M>{qMTh;H(7JKGKWjWRXg!SKD%qpf+J*P3ZTu&uE-b!q%|O6yNnc&nu|j2Ex#
zdaV!D>!&9#{P55GpWI8=UhMl>^g3vB5t)XXR|mFcj1*dNHK@n!R8E;|WagNAzi$?{
z_D1BY36NCC(?nps{)>@r`+A4Zf{664@?7fgT#K=_c0dZm&a@NbL2JL^JP(19gvtxh
z+GqQS<>Gmt;wNTX_=_g^8lV3#@r~gYkf<M&MCl;aSwTgUh1?jL%{TUvyyBa@+6u15
zeh-`hG09Dgj*sUVV=SUh^->AW45iXL9SX%7ZRt1WH;24<!pGgl>m-%KqX8LT)YZ36
zykwy8gXZK?=0Ac`)eaDpWQy)Ev$+)<VfKMhCuej?m(!<Db@*=)A2g#{L+DN)8z!m<
zL!MMW;yi=yc#IzO31vEpG+f9K()qOcOeklF2+ln61@^mPY#Tnp`~FL3Zti9JN-OUO
zjsM#-{WLX`kl*2k@F4>xlis%AI8o?+vbSt~WoVw<8STGW&WX|9YIlnDf|Q8^SsN|(
zA?XG#!4Ma}a;s-#$m&XU8P@nm0{pPfw3SSk4>tz_?oQAE(qZVBSEREqnJNy-9?v>c
zm4r_ZBf#WfEKtpXV+?48hh33wJvUd4tg)OsC%5C2%GdX9WJKJ=8gaU%T`$Ii>Y}z{
zWb{tN!1?;kQ~oNjjug8>Y%DsTxd^FJ+55H1;j+U|X|ByKxu@EurE9S^;JrTvo+%&u
zB9zi`y4!EGqtWn!!K3R)lMV&U{J}$gc`)!9fdYmeFM;jom9C$(XV}N9`)Bm@MfHPS
zG*nH{`hM{5!QAX1dAyhpPNF>F=&ruq_<5GgrvukBkUFrCqCRpSw!Hg4g%tkh&;OnS
z|DUt?KXewTC#W)zDBvbMTZl<KJw|fYgf@J5hBcVCcT((3?^a3qoSWLtTvs5w;KDS5
zv_D|GNTp*j7-6ET^XP1*=kwp4H>@8nre(VuuUvKO1LRu##DWrP2O4#;?@Ew$QQ@eH
z=k!*Y-=sZ@Bx*83vZ&_Vop;1Ij=plrY7tCR>t2RtpC7jM+%bi=Cb;lVu<>dv_O5)p
za^~Eb;&{rXJ>LgC&Tf@ma5a%QEA*H*K-jX1O2+YH%&DDHvy_f={q_PET&FrNA5=X8
zRz(G4Hs}D>oHLLoBcQQPjl@92{x0byI{fw;_ZBWmq#v#gCJtt*@jZH(EYrPYe~i2T
z+spm}*jI}!CZd`WeDt*i5P}6N>2_`Eb6sL+2M>Bie13aaLV`8u>np_b9i8c|+rl7%
z^P+i12dKn+{Z4K^N3<^Ewz%+En@IP-P+9|C1neuo9z{o<ba^hRd!<c~co?=u(_`MB
zr-5rn(8&=&@`}C_7H?mu_G@NabD#SlyMIm{xz&-1%GEt9=&EzNtmiH}R9+svps~<x
zDIumcV9pQYn*l-&U~FZb9SUklKoR%?=2V}t*?6a~Rp0!AU#3(t^a+!)-GXnW2#|E=
z*z!;*)3%hMYpOb+J7UAO>sZ2-b|_X`s~Ux`5=;I`;t?R?kv@yizRGZtS3w7Hhc|&?
z167&tEh@aIVW2(qMn;7v$po`9kSC1@so-02TU}*V8155JRdIJ0AkNe6zAxyDoSjp&
zH=l30f1<B3PfaO#j2>e#P_<=eg?`1>YeUMXD5Gy)GVD+`PtT=;k+a<oNATm1N>*5o
zGx$hj_$$m&dS^x~$0Wb$jkkl>PIa`jPojwDY@o}(wPl2a8BO(-q@8tAP;8U)-R|Mh
zS$(3wdAsNFR?4c9>+V#*j?HgDZHEd@_5TLc4LNm}2hRqa&szx{L1<!|&mEI2o4Z%q
zw{YdQu=&oH%5RQs`p{zzlWgF4K+59?yb}B~eJo++ZMnRx<(K)~Z_(9;O3jYJK(gcn
zW&`#dhB9pWwTM_HbUgWNuk_RT@||NoEDuL-A}ouy^)chO`mb>GDeoQG!5wRYST2wh
zNB}bt0JS9(#4y%xTeK7OLKXRK&Kv(|rQHIx{`-+St3&|4h6qd~U1+}gWqD=HP1BzD
z-wp+SaD89?Ef@dfBmb{1pEWr(SdrsIMzlXBfWGF|{oz-#d;!~9la_LM!NFsfc=AuO
zJ-Eq}91vL|&A_fB?dAYDI&!hOPBCTJuKLn1E6Rx$%d>oL<$p-!WX{fN3#b!4x|go_
zix^{Uvq$Yy#S<~_L)-Xvsi_=l!G`yMrN&i2o0UzgNPU~-h0&!XUj52_HT=Ys!z5mS
zjDYE)q3DU{Cn!^`A$j?+ccTNRhKp!-JH|W(@2nDsSZe$oBye6sZ4HFdfT+G>^yC*G
zH}fwTl&CHLak{J3M08%217=ruocj?>6Pd!f`TH`OrFO!%L*MrGgvv!owq2cqZ9TFO
z&tidQf(irNGGG!t2zVXY+Bnwym}tE+G7Mg_7AHD<W3bsSfjAL+zOtPDO(-Ml>_|A0
zx@#F#b@!~$S5!$CN1WX-R%#j*bjPpq665(mX)~qvYt&@v?(=)7J8HPT6Yoe=JWBF9
zqjZkHnWwPiyu2Xu?elQ;kyb<6DCx@E9FaxQnO-lRV(<K3k6=H)YY{^sQWv*bao!KC
zK9Nhy9;`g4!Cc;*Z1wZ{;m47pcHlYqV|xH>wH$*(jkC*_AdmON*j3rd!*$O?JKwHu
zk-V+5h}sPSgRx_fj*l??4vV@6aMn}qtWS{cPXnhx5C#}S!}KURW>%)C>|&3DS=8P?
zC-OhH`Yvb>W$?C&k9WlymKRks5Eo)w44)Uo#y9pOieLN%JLu8Ahh(mnvqj?rweJJ9
zygY7KMTFFmSBY289jk|h@Ld45jycD^1ulPNg+vKG4^j+bQzd=cRPTp!x^FlpArHn!
zf@OyvR0oY{NCnrL;A%PF&LtR!7Be4q++?i~e^#mS|AIrRvGeUT?V<bQMYTB)>1=Ub
z3X|*aC-tRxDR*mR;iIEnd10vk5`=L4f~)|%(_E0YmGv*9qwJob@3gvXVP8MCrRaaL
z_nuKrcI~!s0099p(mPQ>QK}+MT2Q2k2!evrqN37-h=>FU34(}p0R;gmN)ss&={0mj
zq?bq$2q-;}P$MbNSw8Q(zi;pJy=R>L>)U6HvwtvfgyEh0UiVtpy4JkrHK)YrO%}Oj
zg_Y$KY~;jF82O?~Oe23|0uRuK<u`7S8>;d`(pH4v!PuEZ!0EabB&Z!zfqQ<-+uqCM
zu^<+>&g1<+lzsIV<Sv3L2U@S;_nbK9mxY`>JM5B3g0&T2Lyms=b$jMe4Fu@7I%1m-
zGf(zVR4w@c3TIB7F4Oqr$D5Y><5SPPi^P*zyz^q#f$4+bUGMRRy(%g`R<9tTx+QzS
z$N4198_hz1yIVf3ay~%bCS_!P=nmj*N~QTHS7_ex=?%aI0Z|+!_}yvhvADLl5QL<z
z`kazOn>O!gdfu0HHQSye@lVe~#H_Sa8Rmcvr2r0(ZYtsaKxkYaN=<G_btI2ZFF$5p
zvYvFko!<0wZpk?UOrvvjZLpjT=r&|*+?Leig|;cFgsGMtod{u#SEbkPzvXT4-0r(n
z>8YnN_~VPp0W&d~cSj5g)qtP*N#-bmY>8mO48>`bl~b|p6I|l2--n5SdfVDTlflL!
zaJ#(q089rIiY}r=rNx<5{o~Jm7NugC>7PXpdO};U<csUE`ZAf_wF7elB3hd&pE(K9
zgmthl{vMzKxr?K8J@Vehv=cQwkTYW24TgT=htqI+sLA1&poQyaz%45Wi@a7>iRPZ>
zsL>Pe{+@wWdXPepxj1XmOfg@b*%<lt8nwtMHM;Y8=$N1R1NNTAC#2ij(%Y-AK=UXJ
zRNQ6a+TgrXPEIHj`%i@OhJ`OxUbBlTUehP6F|U7GfmwW-scZ2e>K^&+KnM00v2g}B
zoeEfMWkdv>udMUQXsMTGbj5^}j4|h#c4YNr{9=ux>k1+dUxr()nn8>uwtD5Fj9UdR
zp7!}L3GtR&+I_>a<lF%EgBfgPRg;D9P8=0h04(@+D<5mee2cH`4y-l_d(b9%U<Zc=
zBdCbK2c;gvf)|dJ(v_RD_{R&+nL5_iKcV*A<jUAXNDn3tthv8S;bJsU^;Er&cZ=gU
zn}cyIKi=bvn7693`oXs29um8sTUB(k2YqRpYzB?t`4IGmSGO~hO6|VSyhcj_tCmRP
zqO(zus|#c=UAfur6@{_6H(i|HtGtYkMR4kEc;HCBG%g^$Q&eSfp4ydz%9~Pp>5=rE
zoqVcLj#FM?35*MQq9c=Rx*pvqTAluA^VaPCm+3Kg>y&jK7?ilS0{dZHZ0)Z>Lm%Q$
zu7bvp(lobTua4b%5P0FqG<`7E5?i=~(}o|>Rs@;R`LMS5rM(+RL4n|&#Y!U8t{+jE
zylNxTiF>-I5`yMB_?noq74_v;1^*FK>qDZdyxDrD>f5jXzLxZ#=%cTM`33-^@;EVz
zSv`&wi!j;9wL#8TDOZMZmssN8&f_oj{F!R5G$4wgh>-<uPn6kd33vS?{a%lL*c0!D
zNF=>9%{p2l!EJJl;kgR%qtKayH8eFk`zNx=3*`OvZH2&zs8IFTbrsJ~hkbide^%%z
z-JmSXSseR{woze@ok&t2w{GSme(h8suRxUvZ@)E*1--gh-rv~!G-FNkx5yuVlF<PJ
zQE8TbS28X{O-cIKpm~(UiDDqNvIH_w=}B9Lpw&LRD`KR35tqdVUb#-*#>+4j!0OOH
zL0zA9ji@(rpG6#O_bl+>|8XHf?P*Ba9afEP9Kjw~1j6rOpNl6qXFg>LlD%=h>{Hn}
z6QcK0C-fcoSI5K_8lD7!EEfdL$_2WBa_3GYMX*to{5Ts){0ky9iS7L9Z@|d-xjpqb
z#6Kft3BlL(_hRl(4Ic|08rPj8K`Q%j7a8AxQk@s$DWF|V3o?7)01zY|q+BDVllq5`
z*GCj}ABrSmtWCk6w{bZ@P3<(=D{pMY`5@uGI%}beuZ!-PvaS6CpKfD9@F~DP5H_?0
zjmS7MP<|t9G|lsaInq=9ZMeaAYlU4nlUssj>5gU9=UlPXsC~uRZr^sicf8$EOt(tH
z>BaJ?VnlrAr{tzFj2?3!LuyD_X?(@F?2WUB7va94mK>8Ea7to9u*1*+h!7@6)1W@=
zSR<^~PQ4Qu-#&OcSr)L+I!$po%L<;4Dyky)cgV!Bp4eey&HaDL*YyB3uxskukH(PE
zOi2nRS@@k>)0?|@rSH~d&YgcEz6HeZHhAXb=zPg-d(Q9t1IW8=k85LJ*0;NUSfkku
z$=6w+ggWU0u(@xV-K^Bh?pANblGh>`dtjvtHp0NHmDWE=4Y#WF^0yg!q{G|PE2}b3
zS`%c6CM>~M>u867vR0DU_}sb5#4CiK773mrpp?kBDgTl~kkJ2+43@J6TOmj1$2DWc
zws}q3FnqOO1GZ};ob}}nR{ZHZl42kyyV;IbpN*r+R(0T@<sst|O0MoRW5PO(DXjQM
zZlGc82b9SbM47NlcgbwE$Sb&ytXb5f;tKzyTT~fGas2`>fOHeGMvg`oMdd98?EkUz
z#VG)Gk#XUDuqs`J7^KjQft3!ZnTtO9A>VO?x1qR}WZM4-Vw($Q+c#`w{;FXXW;8>p
z($|L(H(RWGx@=ewjeE&4hLZh2ilh?=z>*v|t3C=kS4^58Qr8Og5)p3S;R6+k$%{Z-
z5KfhoTXm;!?%WtgoA}Op8D37b97)ifa1c@DuzGOWSch;9z%K^?{Ng~fzd^N!9q>r?
ze*35<Cjm)foqKS%siWU>_3H_?8Q6DDCKq_4abuRVn8H~&zo)(NySx7R74MC1I#%EQ
z`p}=HQ)qV^-1NL9DG1E{PT(S7fc6@xzx94)Pbx3{vA(JIw-BSB84vcaBd#!85M{0}
zu=ikLnm}$&iV^!R;8<s>Z0V0r2T=<v2(tkL_9MCAMZ+2o|9R0&c~n`)_*$9QEb?HH
z2q9t8I6>`QNEtuN?<bpxd_)<(HO7!A*9<Il&97C5EdK?$_&Ijmy?FCinS8KmpVn-H
zG}z={Q45F;I7#|haPVBsjg@G)*x`9SCs{`*ojb&5B6S(0dL2DH%>6?YmrJ9HDzX%l
z=R3{X7n_}Jnr>q25T)8qh1Zvol6nnwPvzV<Ie-4C?T4&`Jacx<hy&;+WTo{m-D~%h
z%W3hgAEfVGI%s%A4=VKAi4w)!KRsswb*#y3dh(+L_fFu=k$R<_+@1T%P$Hsq7}Pp6
z={m5v&Ye&PY>S*`Qhrq2jQSd$6FU>lR2o(1G4F$+p7&iGlt{V1>K}w#`brX_aRG5P
zIH(S|WJ2gV)RVK&@F}6eTc4wDd`kGbwr9lQncf{UcuJD6U{M9EpJqp+W=$*f&a~@G
zycl<iN^|N_w)0%OL~m@tyMSyO3i-aGjSSf`w>&gy<X>XwF6m#^oE~?=O0*=P!=o=s
zS0x`c(Q2O&S~yWWn)0D(hj6IM#8~ub#fL2D7t2@}*?3ye9+dzngv9xpyMFI3{+cbV
zpOlM|O|MtpFID)Q45Rb=&)vnpP{a+cm=ESvq~9@9j+zT4JVjK#vSL@_A9UaqZW(`U
zmmz_gi%))cGm9V2=ZNU&d#Zm3kThm~*XztOloe>UoLz4|Y<hhMKDGk=6z5QokQvbL
zfEcAp%{CiWUe|}AVhPrRBF7c<4E@LZDz80GC7;F5yaDY9w;HdDwD_7v`z)yZ5Tb2^
z|9SFH`6fh%ZcRD(?*8~re%|5}%#16)h$;vm5M$4+>>Ve@3#K%)TcZyiPRbcpU&<a+
zE=3FWl{0=>Ls}BY8=t)o<~Ho(j5i0#z!91?wpku6MoG3Bt~Z;Z;i*PwbKlJXQP!>r
z$+8npO^Hj0gQ#pKmkOn@m2|lBdH0R+%e?U}bxSW5b}xbx6>mRy(Bx^UiomCmOmQ!D
zv(6OXeqaG~9n&{WHlFYjxmGpIJhAf0qx8ZB5Jm31m}ng1QIR1C?S&aZ?owS_wGp9@
zB&0-)K8If(Tz!Ati*(beK%6BQQhf3Mq?RCgo{5zpjgd9wJad$cn7izK7Ha>#ExUA2
zM7dq7%E2|}5TzsPZWS*5QMOQ89&-fGx&55Adsg}BL*KX`Qzx^acJDv9)vjgEDhM_V
zsT^r6$JVJG6tkSOTm0kl3wd9N@UX4De`{?`rpj20SB9OV0NZloJt&7+6PV0M$msje
zSfb(pcJZVq&{Z;mk(1o;TY?pTLGbT`)jxbtLKd&v`PEatHI{x-*ZVt3{F5mD{bTFc
zcw+EKe0HLBrHM0lc^01q*eS{Y=AD&`V=x3+ijBZb{CD4#G~H(JG@N#1SSbf-^hz8u
z@cyM_bgYrmf4u6yUHw1x`$~`6V~X16+D`H4veQkh!Lsi^XH#tBWaRVmJ-$ssj!BLd
z|8K5-BU$>`VDUZFy<xBSuMdie8|>GA9;~)llmjj!8yQtF4z7FUM3;KO+l{k=kn(@K
zE8D9X@&j}6@!LAr{Iu?>Mu#rh-0-Y@oO?2IbcZJb())c<AQFQAcOt|8jR%8M)qju2
ze|ij9RoATbj1|bpov+j-iX{;VeB{oYAJp}mmEcuU7dA)P?ioah8Ohx2<+K^Z2Y7br
z%VN6lT>8o$z8=nsWaZ?dW#5N{M45n;?Jq#UiiGf(QlVDN^7u8>kD#v=U79@+(}wdo
zsc66^`}WQ(-V2B<!b^eg7_zmXqK|K)2%|x;l6Wl&2qvM=ug6uNoo|>RX0Dc!vuCye
zeWm`$OpuCEqYTHqXqPWF!E@xpKZ4CantT4Q`*)s&+GBG4=h}AullT0!uCW!!eaa^+
z0KSIHP6smlPfL3kf@pE>9Yr|jDe%zpZyj%K-jAK$*U!E_afZpA@}7RIwwjD<#mclv
z%ba(q9@b<-dIVjaiGWsMZlbR;3iTDyk4E^Be5gF;@;xnwK-QavtNgn6J~8*vp%fJo
zlp@~Rk|Fu5%m2~u)TQQ=tiH)u4@M0GP_|~$u{;>IMY7&#9A~8N`L73d_uhJ{7Vd65
zg#lEmO$YkyFNh%eMtL(4Dd@r1HE8^~`VO3Y^t9!0BOCn+xzQQRPL2Ixa@_R(@xwtE
zaY78`>)$7qDoZEqn}JZi3zwva(A7ZhgzE{Y@w0-u_fLIoJsb!6m0i(Nl)_of7K<ke
z>2huMjjw||nxX<Da^!QKbacYr;D!+2JDJ?ojb}$evk(W}HkL=q7;VDQ_)O8)6dN`S
z5y*Gk#pMG(`0sK=EQuw{$wWEAD{G})ta#f=`&b118#?b2G{tw=j%e4Z68dBpoUSQ4
zSIL;#I+Y^XZ@woEn@eBr68Sy{gwj+Eb^TA}eERaaZSG;MWQg9m0tF(HdJI9*^Pxl$
z6{v<z)n`1XBv?gr;;YQ`8d(l@*07f%fEdKW0$terTH_#QKh@fSXC+PS;K!MYq8>4s
z?nP7GNP0kcN5;wOmnOmuMHfi})*IHA#9BN8J5vtdI8O&a-z{+38SwkvLt|eO)?Of;
zo9WiO*egLZQ|{Ll;i?h!mNJ3_z;~u&u?s-s6~4QNrDW5Po$yW2SO}vr;8em^^fVh2
zR>qH(c>fc8Ds9FT%p_+sVc7mOuz%0+W5Xa=O{F=U#1cNRP5*Gk-c)ZUP2C@R5PKgQ
zmC55hI%i19csGBtS4CEnZR4Tm?gnkghw4z^O2R;Zfh2Yz4>(o4-pC}ygQH0|S+mWO
zV_AE>m83iYDp94+x}_>o_mf<&iQTk<)*EQ;W^uzxIq8DSopX<iU1#vH)`M}gDWy`y
z8yZI&Fj_K@BRjZ$$%XhYxL@z5=B|)SsuR!2jz2DN(XhU7Og7R;&*U8k{yp-06%fQh
z0ZydKY*Yx|+byQJCAMx>M>VK-lBveBb`@-}LQ<W35%3@IuQ?yq;KOFjCFWHY`V4c>
zS(U<`*{LFGa<Ji}iTJb@cU1y}GfS+5b-FF%O#KTpjsl#8v`?c>#2d^_!WEhYw(Bu)
zjK}I>DzME;jiOZZcGzeAxWGab^3pI;!n%%Eujg~!OHmhK`$Fmb4?_=-aV}V_S0k2W
zd8nO97?%TzN;wGX6`;p~U$hv&lDXh5DVj}j;dcA2E!o|?^UH1=fqex&S$gm}&NwxP
zoFu4-XQ{c^7u(_EG8;+=8#D{ZtP9XKRELia4%Aelt`l*jrJ?9+>DO#pivgGJW85fC
z1k(YQ0{U*wkLy!t-TtgI?-1qVV<wUVv_h6<T!-bv#Ex3#&J}7n|B|m6mdJbU(x`x0
z0@*iC-gs?Npwk1wXr@Y0@U5_N&Q=;!7-dg(9CK8bji^o#<D9aUGCmlv9=Rn>2l-jZ
zMS5U86-((yDX*=&hSQBCkT*nM$KJ6$arJSRmO9PPfL#zs`~Rz-06P+tW`@$)aU+8h
zbW}-B<VjIr*K#^J;7;?A2m0xrBc7Y)OrEQpbcJcox4t{mjMs~7xqsx&kbOd!k<9Y;
zj+ctF{q!TeIeo$z)$u>&!#&Gp4m~^*_k^(@*Vads!O`CWtb7NW`G;{!bN$!1fg+G1
ze3kN7hbSCw#}lTm4!SBRUcUBf)%%>3!k8ztqY*lU4k8AbwC>d#dudvQWo4XtaPd<!
ztQf2qI5*mBR!sjG%K5DBxsRf~x!&AqW2a2s)CVmylL2sbHFGWUH}cOgd7}Bq%zRg@
z>$^+I!U>9{r{hs~562OAa0RctKqFf_MCo4*(Z79$)t#=Ai`rg+j&IJ^3y)7_Be^}x
z*kcuHgs%EDLC6YwVz>qP4jEAH+o&MlzO{tUg#_G7qbRg%JLh@^LSOh<YL8z_(R4O7
z{&g1CPC^VISF0$>&?p2;z0~Z9eA<Q0>xNQ8u3G6={NsA@8J0ryGj-&?Ry-%Qph_?Q
zh44&4=l<J?_u%4;3*cc+(Qh-qli?8<Da+$`2JbQlqXK>w6$w6F)#&G3^7?br0j?xH
ziYn|GmKA+#*5vIuU$6XcN~%7!55sQ*H>v`{d^lY=a87Q=B&1%QJkO=k{n1qY>?POq
z(u6w8(jUO2O+g(Qk#TZl+kE0T`rJpwDjVH3;T;3bGyFcAfrJTw>mvC*_V6UB0+XOZ
zg!2WpJbkD3?k_c)B5wr{e9IXu!;dmq7BPw`K6%T#0uv6E+T|BGn713d!QF!*Lsypv
zc841fugh6z$p6kH|I?Y||1#^gNHB`j@W1&YTk+ci$hO)ty$}p^8FK`7h|V@ggNA2F
zj7nVSFL^Xlm*Jq=<s)-O^|dneN7Ir4EKgd98bj=p^bl_V5QzQq;F7AWotOL3FY$G(
z29sQ3GXyh4D(J$m5l2Gkmy(R#k0shvx;{RucnO(cfl2D;3}q4k{oE*dHEl*uv#n~}
ze1&gnTIpq&?Dwm$mFl$g!UCYHWdMXz9KgHdqLMJ|pUL0G9e0JxCZodNd}39fG}ef%
z=)g9+!+%bpRlwr;1@>A?gfPIxWB?O$v)q!#jsP73A0dp1zfZycA~7wwy)cV-?uB-8
zC;AqRo8czc#_~Jm%L~izBw0PIe?kD*KiPVH9CZ6bFkGF1Gqe{)6};z(UiM4F$q^EV
zUp0#&4H20L@+G*S2UDH#xyUaj@Jh`{h1txjE`R?c0TR1j!+=iP_P_f6|Dw-X#6k-0
zX|W|(_&Q5c<B50?Px084_scuAxl?bOm6JxVy}Z!o>V9||*Ndboz)069oW$fvwRH#;
zaaH?xa?tH6n;-VV=1H=j<Rzs!*iAs{U{jN{Yfg>x^^vjgQQESQO^_@TkTyFW47XZo
zgL}fqA9nB9QL&!0QKZw`*Yj&>r@~<FobG3Pp}B!UXd&>Hqd^Sx8N{y2!?zr)IFeU?
zfX{RXux@dYN%}nH<nr<Rge$u5iSvpalRh!oqm6U8CLHw`kR<B`MFZzVKrybMt^Dha
z;@^U?>&n#OAzhVRM-Tf3Mq}D7*5riQ<!dp5EEzeo<qE&4j7qHMAFV@C{gbG&*3{ne
z9j5f)wocmqM|J!IO_X+pqsfl+p0Hjww%yhijQo(U53vD|_DXDb&G9cRkPp(O%$DKu
z1RNV91pBiWBLIXwRTx}_)BmvD_|LuS!jH~a*QMlSN6f59az>_nA_G``fW{Zx&J<a0
z73U(UUEgU<;cZ!~8Z54HAtcM_m`GWmVa3?y`?^>Bed;X9ysx8=G_E)&%LdY4Q%pz}
zb3KN$cAe{DD|Wk<k2XD26&>EhOWvx!w#7Es*v5DTpGjP&_!2wP=3wHUxZKtPTnaMZ
zfwjyyI6hfR=Ix!^!2l540jN$$jwig_uET<@Nz)k;YAGK4vb(?#-S=zX`LXX-u3DYW
z+j~+7Nz6Aq`7-nXDB%d;i7JEJQR(L|M0X1Ma~@OOFPF8AEBz-T^)J@qzaOByx+1sU
zRiUeQ$|NCrpnuN`%!ZqckNSt{`V^zg;QBL#8sRyU6A`Y}Ws1rCcTOtQ!|J-Spo5mY
zbQAJ)w7Mrl=fVid&mm1(S-(sEOX=G=Fq?NulJKfPO7LMrUpN}+qaY@iI@2W^{jj&K
z#7wUTlh)7q9F(KW8Of_<GuL&^0&^RSB?o_>7YqtND@6|SYy&k!JoRK{6wL^2`I+pO
zA2vU9uU1T;zCCzh$gNm~D$m&h7p32#s<`A*&z!3!Z0}#IYG|mjYJOP1Ve$?ahn>m5
zf3|D~2SI%K<G2!3e$)8#n4I(Q0RIPF6#+Xg2C&lMf5Vpki+^7mQ8%{`v8--vtd1S+
z{c#|nN5Si=m8zNg??jpzU|XvkY`_w+q2uR0%z54*oYJ%=wSEY5u?O||AwWZ9Txwic
zy}*04W`Bt^PYlL!x=TRFFXG~BNbG~lz9dA3vjmxh4_Bj-h*ozy`;MP=L1dE=oWrw4
z7k)2v8K>7|Iy59O%XKgHsop>fz5cnQC%(DU!}zf6K0XV7Q6>IF(^TeqFf3I(lqtTl
z0|@u;^8<bx{srk+Ny@=ZXA>=<vWW}Won1QHPuL+BmE6X;kAFd)*D?z5iv?2{P>!e6
z!ign#59FS_f95MV-MhwK9zQGeV0{-vdlf3{Uy$pk8xcMu*j6XZ@buQzb@&M^^#_bh
zKvZZlgmn78!kC(l#e&#2dE)dsID{d{Y>b;7zk@ojfi`nLnhkcqrk?%|)Px@W$16N6
z2Zx74h2dB{u=>R|LAOICe(g%x(bWkgu)d5gZ*GJG0ZMT0_zxHH`#NAkKVKrn1mZ>P
ze?cJd(F4fjNifd;{&r0Mr;892j9?V*YJVCiwu#Uo|9qlR6WUG8@M!Pn%!pkZ_Olyf
zw~jR6M+l$kVqkffqR&$UKeoD1%#EG=jpIAh3R<9<0Q<Rf;3VC1eh-JuZ!Q?`Li&n#
z3&i9<SlYrVoF*qL6r5w6&v|E)`(uQ@Wny&jXq&P5gpvBMcQgXIdlq!GtDrP3DL~(E
ze}$6#Hk)kxr(+V3SN9*?dduf@4Wk9NY6a#Q4>%7K?nGCrOETL+9Fo!d+;wEkd$H&@
z+ej@&h$e;T7OhuuJa`o4u{;#&j#CR_XvY__Yu#^&JW*n3DA$fT{+Wz?zOCeYCGl(N
z(2Qq9=D^iEgv~3Iq1j`*?eHk+<CJ6c`C0S2F?5w2y!OC(C~ZNA2J#$sVanAE5k_MD
z{iCni2s?4E)snC`(wVO@`{`!~6P>AH#2}4pKZNEBJiS+zaq<NWUI=Cc+_Dn>o&)R|
zjJ;Ub)t?prf^Ne)nHG)i@@P^uA@y1WRb6L7$v^O+bcLM7$C3OKmX+pi!;iSLSu6js
z5#ajPsVDPS&5D`*c2g1*_|u|WnQif<6^*K;R{g=cjRtyL>U(UU7V{X}3<uLIGm{T~
z&+^3(s;(`yHzQ>J+qk*Qt&=ZRg?TsmvojR7E=K~xramSQSOKysE!Z7lEJFm50Vi8x
ze`~LULONj<yLcXNi2t8K36L)B+p(Um@PFDQ2ETtfa~(8<J$D1FANjK*vd@hY84f3T
zw>s62p)5%jv};tI9e*maaM}KOzNdeYsJ()dGS}<8fc;Hh#10==$h-)lpJtSj6pP><
zun3L4$3u6eafhSaATF%<t5~2E3Ih1^M>U%S;mrHUvP?uu6{dpPHH+u>kk+N#%PyLl
z4>GZ?c7J<x<6CDID?UqwD!1O!U_Jx3^)jLq^m>)>cv?|%0uFug7XeP@Kf$*6^#rIh
z#m+G<WUgiYvcBpMTbBI`@(3uxLYN*PU;M2+bsBW$Ea1kVRM|dyN^cBE5Ow+(T%HvC
zPAX_1CC$P1m`@m>MjUbpoGacg;&%i<W&V$FbJw4p2p0mM`8-;gfcf_1>LAG>hmggm
z!7=wLQDzC@x18Gk$78`hhSU)DIl7#nTI2@mzLYd3cP<UFgb=GYyE~)8?O41P`o4SK
z$w@^)sp$(l^Z`p2GzotPSA}ggL@Chp%pI|AMla!Y)!qz+y^FXOOX@MpmLQYP>GG-c
zVxHq^=lh8X4&8CYm@#OQeoiD~B5S`2u6V2NY{}X}?D(kQ|6{PZs6fPyyD+&MW-Ooi
z`?@i?o~uZBzI>}236fuziElOuSnl7uOzu*eyN7?q@oK6GQR&=jIgS5dsgF0+{Z5no
z(@rw6MrgxsH6Ac<>vl>hE)(vkp)DQ_n-3mC2a0ZHM5<~E#Qr$JHdTKEESs0q!dXD5
ziFh2AvnVq-YJz2Om!MRbJW3P=!`gF2RA2K^QY_o=1npDO*-|ScOE3qUPhi+lw!H#@
zxri0<7Z#r<EuPRscaLcQQDuwG1PsJ*%wh7#&uQLYwdCCwf?`J<>XI_VuHRX0KCJx2
zwE_}|Tm?asot28DnX73#<&O2!FJEQLL%cRrn2Q&IpTgy@{Enkkh0t%&^wDOSrX$vK
z=O`M98Z+cC*ItD@_}cnG3vT~M-yq}>lVC}^gg!|-UcK0;(!Vp|EE(*bEXE(UdX_a<
zzXm75$N_C`EST<ENzEyxhWND-t*W7-mWH+rnwexEp^zG9YYC+WULBF`=vKJl?Vxw+
z#dR?2m_BfKShW2DiHlNrV|1J6x60@&CF@h%+3W9>&&$Vk1nYNG0F;>7VOO>Lm?=i~
zPIk@bILeb2ac6TZ==vj$`ZWNBk|?pGi`|NkHBTpIp4@=!%-D^>%dm8Qc$2tgXgR}t
z)~bepKIhB-t>0l%?aH}~$1?)evkhgo;IZz}*+v!Fa~g#+)Yw>vb$4#9#~Z*-qyGf<
z0ecVoXTn%Mx_qir=@I?0?4(zF-gj@GrrIruwEhKo;S8g`BRZ(?p<Hc3Z3n#xkL-lk
zL<2)(XYqV+WZ(DW$T8cE?ly7Wp=D{F;kz3y?g#HwCMU?;PcUtwjV!{*rtopLE9A`f
zmCSIvs{3eoanoi;*;$1HU-Yelw<fPI2(s*j(QN7W>lqoCebnU6T1*;8v45`OJsq9o
zJ2Bl~3c8%Jk$4e2s69@sI50}k@SG+4%&>9>tDZjc?49zIA8%)(rGC`lCUMa)E}#wG
zP24G%!3!CxcZ3)06XBcqcH6fiR9W=l*(-<yEKr;M-bsvG?FO*}>LJZuI;^CO^F|lM
z{Mzj0CdRSUJ`!Zvlv_=wSuMjqXST^vY}4M`uaxORS*mK|;~%t&eI+cxR<prmyGRis
z2~hHl{xmFxJZd;PVRet!59uRQOEjC2*0F1alP^z?tuzDUqYiiU**OQ#!eS+_V{eWX
zcN<K+HiM~dB8m|E;WOz74t-A4x+Mi0_Nc*{*SjroU$XYA#G^kCVQdJY1~3ak2HKOl
zJ<|ueiCS<a)K1$d%c|y$o0GaLA=*VArjwq?S^qYrGepZ^zdm7)HQ`pXiiBmONtOp$
z>DiM&G>vAC8IYD^<m;bAc#c;L4>_2f29`U+naZM@&$!by`JywnSax>Xaa5(_IE{C{
z?lUkh`@FQoUkH!OEBZR<taxWb5xP+KoTC5pd%~75I_2yHy>JxkGEJ5JrmsdOg-2l&
zEp>%O`XjuybnH%Bly8VlN)`YrmeBSuz;|BX!PE{UTObci$tnz*u1}f)fmXjPXsSLp
zSoUs58i$Kl8v)Lyr~Qq{%S^C0-kSW{1}(-U(DeX(FdgYET2A47u5KRWU|wTo?dgAj
zrPDC>IsP!*tB-n&`HpO7UPDQ4%}cf;=hm*(eTt8Av{$hhlT+r>3m5tY`vfBeePs54
z^2O?X<hpEfb81l`GSEvgyu0c^>+VM#K0w!^;kX&4aJMKx^G+s8^+j!wj!xgGiP+)s
zED$@iwN}52^BP(wj;PT*CtNp`)^6X(#;z+s&e3DgU`Ddv2Io=-5PM2me-}d@!b>yh
zg55yNy-~>#W#d53drj=o;8pcHZ%*~Tb35aPV`yElzIVkuzQTkqG`ilJp)#y{hTLjV
zSzlK6__@>V#g`pMS~`#~3U4<NDb6g^_+q+7^{8C`8m3F5Am1X-@+f2d_$BY=Rg@%Q
zEb}v<wu2@C$$B~DY@gr?f*k3djIFPHI<tAN+J=rmd2!+zAl_(oFh9HK=LXbu-0Cff
z47mMJdP~m<e*yVjc@7r!J-=C89B(eOH~BA!z!u}ERsT}l2c!IDZWos@IZv0D2jTKm
z@j0Nm?eiWrajw3U^6W`ti1P5XeK+#gm-yI+68^e#(RP!W6&25(*sW&IMV$z~Uc7bo
zmK6tw@KMf<Ea<a(0WeM(6xT>yyX++ay3RnB`N|v8l<C_1G5=FyVta(;WsRS<+hVwR
z7!gK)4duapzc~FD#8?S?86ck@Eq9F|`t>|Jv5ZQ6aZJw>bgT2oAkD#9z~qx(!h&9u
zvN1>CC?{!~c1>=zmHc>i{W9t&#kPPMlkZT|EJn-)?z=z5=_d!brbU_knF8<1f+tbo
z?N5GnP0XKFTzhCKmn8}O$uOtDD2<@ztlVe%>xFWzJFMwDb@v@fminyCP29p#6&5;p
zHOS_^p83msYd?N|1KoR<*@H~}c<41t*F$rt_oh~6qj&A6YEIk`AN@+qik<um^5C^x
z(37fY_{^(mmijBVSCH9=CI2YtFZorC4+Q5RubJ|YP59U*lZTzCBJH8EhiKC~hCUYf
zpjoE2;uz^F;7opd_#S_*Jl(n5dC02*NEj$gNIc^=^x@$!v6<vU^mUTrXosk{9ZN!j
zXEvjNs4y1&SiAh@fl99L-3jN|YH(QaRF2H?)bJ2pcTaL_d-8@83Tyk)B)`w+hzs&8
z(M60YEC-CF$1Vc>`!z<bdghpbRlnC3{=oN@IqSGU^`Cvxy{2=>!Nzu6S07c$pB}CY
z;239sIh=n2*)!h(;7&+bCuNA3-v$$>PM8hRkL02(hzE8y%Od=<HS~RVBOD>x^Nr`I
z&0rQpQ%Xu_%|D=_+2c!3Bvs>)KBtlTcly|yzi>*wCQm;@yUa$^s|1nPF9gVzQjTrc
z^|*@u&QCcg152}8<ZOXaWsVomLb=A4Gt?+I2h<hbgr+M?U{5uf_I)GPsIfHAK&7zR
zN!N0AU3#%mfD#^dxIWIN+WdOS*5JC&{Amd9TULc_V=nZfOQMTOgQ1ttb1|v?e_*V!
zDkhhMFIi#d#US1012qK7w<vz!cz)DU=4-sNA#ZR<XJvrUp3Xl2B<$HgvQG32xpo{4
zY5}sdQ~zI-K1e4AEs2bTk<AsnjK6-jD01^3bj+SRcJqc)4uc&9r3-Y<?$YH|fULTR
zohICMV4I6KcU|l2YcD*NBD4pNiu-!VlhUe;@ZK=rxUyxl0Qcj>_=>UN%uzfHqd?W4
zn_2IO2{hc85+3lo!7oX^C>!nq{lFRw&-1&vT%^qxe;KF1NXDGRbs&S82Pxj$?(3_s
z^R6w&a#=(1Ld&lpW3qh4z0-c@<jA;%x*aODSQFHi#-rL2-8U+B(!3rhmVYPA>lj!W
z>{WX1^zBqqKk0Kb`x;MEE}9kwc-}KAzq7H{rp!5GfjI)Srg@YxXE7tv5|t<CTRr_5
zrKloRajl;Rni#$Y(i5~m^oc5}3QdmLCk?8r30n2|w=VID1FbS_zelit*K;!RK{q`(
zs}DpugSF3I|6B*jOFted%hxWs`|)D~gzFG%y&-e|3eW{R1~!-PC-Gai`#DJsf_7~P
zvW?{gKN9SZ#8r$4xu?leWaXnxukG(5;>>zyQ0rk}X<+e(fNjE*WcKcgXe-rElM|>P
zPo4W*c4RK>FG!sN+iU$1G7@Y7GoN9C6)Oc2=KbXj2M@k@X}V^xty4$nn%)O2*IIYp
zQq!T~Tj!MDeRQfy_YZvVi(dYPSEf{VdmA^p4)i5V@y{@a`d8E1y8~yjhfFrjA62Ow
zklowgECVcI=UVXO>v2)qLc<R|#4o3mT*#k~B#yCR<Q3K|$FNLpKl1Ke&vr>@@mve$
zL?-vk$wNxHE(3gBT`o&jXuxd_xkMMBJkNZ*r}uj2X_0Al<h)LmTYQ}W%n;-?<@u|9
zbMP^fS#Gzz<<ByS`e`2>D)%kshwp>E3I%$6TOmba=l&tl7z`dW30tYO&=JQuiQ;MZ
zcg99M(i)zay7(7FZ<13>bla*vV|(`(#C2tSC!H81y>e{?cVuX5<@&cW$4RZLF$P_-
zsdO3@U>oc&$k>qe{GmqJ;zGN>)oI(>y5l;oyT&#NdMh39nG$@3Ms5LhOrL+zuI$Fr
zxJUTNsGLk+eDED>lbiYR&RjGeW=m}*$*Fj9Sv#zINgq1&oog-R=cGnJ%-NYp$U6NT
z_>O1Pnj#CCv88eEgy@GdPg6n|s`9*0CO3_OHP6tjEg#xew@H+24@P>kW1dlcnOy#4
zp4Q662x47b0c-G}<5XReUY~jw8h|JvCTLO0y(m<pWC*R*{%7VgPJei#EqqEBh*(B3
zK<ClSrwn1re6d#!-EbheB{})>(a7a*DN5JkcE@_J+AVsg71*x`pK=~^H1TuS=8SYp
zdsv(COq}TfcT;(qy1q)H7+xjctJj>guUu4SC5aqLIB&{%^8pG+z*1#1I~DpO@GQ!@
z>IG-?`odftPTYx)QY-zWk`NTp4hb4Uq~gnyfue64o(*#bmDsG|%syD5<8^$gcyu{1
z<=Q0M@Y8n8YrD-(FeV&xJJnP$b#I-7r$aDpi(sqh3@b;B<+n}?)dw7F3>mMe>k&>y
z9$z{p-~PBw^j9skw^4u!HbbxzJLeWyc+KE>u77v;==@q+8&BVrs!}<O5!w)GwLCyK
zV&FjS85#cM(RzS%nm}=pSDWnvVGa`+t@TufKGhC{CV#3;mfPc%IZv=;1UzQ$nH85f
zk1OtZ?fc0Mp#o11Fw%{5#2_vyu+Rc-^Fm{8Ut5^W&{KpPrW--NhU7qdj4IM~<|HnV
zu`f!#%N`I-Gx1LIFa2`c=V@dYwGVtl5HNFr@=jK@{#Mb|EcuueJvrIXX04&+loNdq
z52Jg{{6G>BF}N)y+mnjwG0Jo9OkrD_jcWQ+K^DfjLQbo%h9`gW^$!wyK>rXnW&v2C
zj}RH>{r!j4veXLmUnUIdnq~*c5!AQ-cjUi+5*h!C0oT9Dl>a~9gUzsCl+=+cIB#OP
zotDj+u)YOEB%FNR%ev2&k@G0xe#!gV7jE$(<jNW78@k<h7_c|phuZDP;3Lao*u$+;
z0dqk7r~BtxsayeXQbRxw22Q_3&WzTDm6*8Bnc|2e;)fS&pGB{i$%!iZM$ZSsEwM9^
z<MZh`bX{t!3bL3aRv}?6?9Q5?TLnb$21%OELlw{cN*hB+FYi86<>&-znxk}iY9K|f
zqtgCaBYV}zeC5*-@5^awFW$YFdpy*GNdanvOli6q84=AqVm?<GWY6Qis+2H6yXgAB
z(X_pCP0uG1IHSyzX95!{5ui8&w4oOBaZ%Wrc%aP^FfPXk!2wg1Ye9e3O%BNP1ITzr
zI7pO@^e|q`HoSB>eynUR+F5I)cxp|;Np)j?R@~WuJ8}E97S8|?sDs5tH{2epIxlNl
zjoBNkZeAu{!{iX&I%0<Zf^d?&8~Zfda0lA;eSTHu9zu4jQ{5>JJ1?mNG$+)kned!#
zb)?DZ=G?^=qvx`6VGrK&bNXJ_?p*by#9bVMS18jZKIkfHu3b|GqSj#S-c55+X{{LT
zr}N~s8kgb=nE|W$Hw^Y>Mj6=CjY8DdbNFZaha<fdZj6*mo$od1_}(My@<2~o)j;NZ
zBm-=yAt=nG63rUryncalG7>7~emcCGPkFdD{ONqKmgHNP@xO7e&Gp!630*xJoUSo~
zc9}J!*D0v|zU}PnD#!g%`)z`@6({DKjP<wkAEOfj?bvZ>3>8TX3=g%VbcE$!%6y`e
zw%8xnluFlTWW#5&GKH{iG2U)U+d7$0HzmzI>6Ba@Bl&{GqNR}e8t8sz7ig?%JT-1o
za07+39hP}R|GccKdM1!b29N7sjP>XXP5ykpikd=<$kZB7eS6_Ai1ro~N-Z>@+|3*>
z_qt55sRYh|m7l{A-|RM9=TyFD&cRzJZs61{Ii_+R#C#jg@0D2;QSU@S=fc|%p&3V?
zZ(g7X7w0;URnedHsbyT3DZ9`K`x8)9uWH&^eV^#kYb!i(I=Vjg!ACta+#C2!%W)c0
z97lXx<`E$&tuXkw{X?g)MkVS7gjBb0whlL8X*TM&KISrO6ds;koot_JuHcbWWNS?u
zCYmm2^+3>|2*Cb=!jV#lJ}@xYXH`|3NIQS#^GP<zJD_{U7ev8)eB}@jyepP>D11T6
z-Z&=l>t?o&L=A2qHHADq8y>oa@~-#vDr(NJV-Xv^7eC~GpF__xmFh>t2~rE&e6FL;
zD~$O)-}nA|o$Q-VEnroPwXkhe9Ke<E^8{Is2D~M@#9Eor73}f+_}hP^l*AuF$B|70
z=FxJ>z9(v$UkFP6J_;sPlLDcL%JP3|s|9p+;$oSij0}C4GZojidtSor>WiB&Ou*@u
z6Kig41IJU8W&>pJn)Z9&a|wp5JNp)61`*u!vk%BX{tB&gOI1?U(Sz?D=k#&>A*_2G
z^#o_z^tha3>WK>TxPw%`yI}hmdekK@$Man$k0;U5?kT7-`~`{56wNj#skH2+Q<LX!
z;3A@{8WF`143mnS+ik^)Ivs0_n2X`Lmd;##Ts2S!YZ6~C%P={LPeVumZ3hv3jxW2~
zd0|Cvjb^_~ip)ut510A_360DzSh9_Z1f6eWRSh9JWicDaUbgzmu{7|FDu(*<*9*r5
zBpS$q`w4~4ASB(U`E}*;P>HR(5!jsNQ$^%U;U7QkW7Uxk;Hdh*c6_BhRVliMx1+%Q
zK;dqJnZeO=uafumPmhNfrHY8z8NM|6pich_f|;69rEp}@tf(zKPJuR6Pw!YAja<8M
zrPF|=J+LzxKa-Oh*9sqtBd!P6b3P&HpY#&uD_9YM)w`MPQ#}CsY)u^dIalPO8s(`v
zO{)7SauMYXyx;FAZGUa^S?MbAyu|pR&+lo+$^(1Bl&E`?{$9&Z^!vh)7_P}ng^24w
z#EE#JDy5H7xYDC|qDE7)@4uzzQacAr>;1_vK*bc}5AttKO?p)y?<;x#iqG;&6Md0w
z&!N2TIdfbm%H0|zE0JK@doBEfx;j>G`tv%8ND<lbr5F{3+tw23Y8$wTM|Hnbttq>F
zWdffV3cH<6`szcCM^SJyiu7yw9+wXWWiltfh)I?N#hMTeq}T=RzJbgonN&(#AJc3-
z-9d@5Y)^f5x9HKx{hJSUJG69JK}Z=$tt`F!?s`|?haatX-C@(I{@hX9$Lj%8F@tKB
zH>uP>XhZT&eCuRAt1j9sx0FhUQj2L<(AOF7=$8gPYMHsd$6Dis)ADpWrMZaca<~R(
zg(nrUBSj_7t!NA@*24nj*{AH+&PmKIE&whl3ktZ4gz35-jyq2gPAfDuo~iU#IMRvZ
zsP$la=o-DFUt?sTjkuTPoXG8@y63S*saK<4KH=bW2l5Y`bb)myx5Eb%txcE-p%w%I
zZ|514JdcTe9zWSOOoLF8i9HZZ+JKp}1%TxSokbXLm{LDbnQb<T35;$FJ&D~aRq04h
zN7q6Z)?hvR*;$?I&Np(dzdx7sSPNpug=6-^c<bfAsmDn!+PgdB-5X<0|4yZAGIBvU
z{bK;^)?ijlxRTkL3Q5%1yO(w(rX%o72@gSHW_`?m&aT6BE~B+IwIeG5-F-Z3EoJxH
zD;C-=5g0P_(w!bA?Wx47DOm^T|6>JZan$uc1Qtp3QR>i)Yda$3jM`pFM3Gy@mER@B
zn$N%rYND@9VZqX*o^*|C<XBG9wu`{squ7OZKULGTY@9uXn*3a8DGk1&!sy}j0p>9`
zn24@W9xA`pt_v(6KCYwjPh#8nzK19jWF1Ju3A>GILfJbCJP&+%Fwk{kA1v2x{D)y6
zYY$SsN8n*e_+jSX>t|Y8jdC!Z=72V;svuP8cs?4D<r36PIC`SbfTaoNNx{+uLxB|m
zO&z64TnNTF&$TPL%RZ^Z`J_lgt`kM{t``Q)pxPmrN-BPKA*2!3^OKs`u5BBxta-n&
z)-+ZxKt1Li6VzCF&=$T@jfr!AP~tORCaiUW&TA?LK{)U%Ti8&nc%!abZR!%{FyUFW
z14%se>$N(1E-gqQ3fMQrNus5Ovz<K6hv{&iw>dQDQ<i11e<abjeilg?p8OjB(&ULf
z-34s9j{#V5uA+~H+S1)LeW)VE56X>@dg&oDn4C0d*e`NFAt!#A;()EffoE!yR;3u{
zk{5_~Cr6*?ssc&;T-QXs$>5h4F4aGc9|nl(FkAo~L`BZjbt(C+OS;(`eRG8h{o3;R
zocX}Qk=)sVV5JUTpY2CGWG`=+?F#pjWR<pn_;<_nju*9Wj->D9jpSWPjbvBVCA1hG
z*=^nM>>Ho!fF0Pr8j~}ss~N=Oebq&&j1_97^(70s%n?Zk7*L1{^G9uGheA|bc!oYc
za?ZZunSKhoNyhQ;;Q$lbx3gz~l%;`Sg_1|B$@THQ4GE){#7n&*K07^{mkH%0r<D7+
zo3(d;zPQ61_Z5E@K2wpplm8B=<%BNP#!O`ldT2iFDI)RRH<CV69!=<imEuSMAt=L!
zqrE;ZSM83^*93U<yFae7ssB)h&N<tQn(60UgGcJRXHk#0IeYFkWW;$~-bYAO^!%MV
z>xK1Dd1gm29-p2yd7f7d+vrh0(T%x*YyM%=HZ6|7A15_6DLamTfmB8)dEH;De7UFG
z|5yba$AMO&B^*rf!Q|@=WOQkB**9L#M+v4JX9g|<BbG%069NF84t+NX%yTxxy2?#v
z{AkS%0(K*GB3**<vGF9cSO270c{<W6Zgqn_^Wt#IlrG!XH<R>y=@D8O?Fe%;Ly}Rb
z%UWBK!i*>$`<R!Tf4ckpLQHGyvrq%+Z6l#y*z#l&FL8P-n;5ejMK_?exqhuEGu)iK
zbZ$JfU(%0cAf2sPaKKA8nosa<N-RWZVUIw9k~>>{>$UUa)Ohu;(z<-G;(Xg(xR$=S
z%qm1XZd}3nQR0pI!SW%6ptwr|WAE;sM;V_u=NtKjj9U$$w9Zz%)qb5BwyiT<Es<n#
z{(Nju-y2^8Se*v8!mgQQY><TKr=QVD3@%;1B0yj1--7eGV~&@JJh}W1`C&pjVZu}j
z7-#lY#b8|!(O6#6UC>Sf3le;=(dS7U)MH)(CEC877t}<uUE9lLacG2}f<l^Yb9rsb
zlrlUl+a(UrL~}nXEMUKf?|cZ|`A9WQ{aLD1S(jg3G5pT@^!(gew*F|l0YQkZT!gMy
z7D|tAyWw`<MLRJYf7Q<Tl^(P^NUQi6jQ|gNyEQ__h8~YaN=<b(<lQ-#bVZW*MTR|g
ze>=v!xq$A($kn}Z=A1tcm1b3qu8zH7FWN;SctAZMjJ?F24^%y&fl6!yM$F;q=$Qhe
zx{aR48N$@3M4Y7Y9Wt}-R#BFilB@O?xFO<cF-X=(RO+Y~X6nGp&f(G;i|Hg^=et4h
zc4)A^tV3W_F}8&;`k7w1U@mySb?9Xg@SZ@t$DhT{q{GQqEX?PQSBSLYUnI0KkDNYp
zeL3>IdivF%4pG_<{fa$x19q!XKax%>=-B?~+<eQUUaUm&Xj~g13^O)ekrzXgr>m#g
zq$Dl0`G0g7Qhbt}$a$cZmK90AKn(@0#LkR7-J8z4z|+v7JG|`)Ch5-gWVxrm@+F^u
z5@5q#9O)c6&aC>7GhjyjC~BKAQO*|6>{jnbIm|(y(tREr1{i$gR{Cw1CUc$gmrMh~
z%v2_L*gfj+SqEg5T<F?y>$)mO#}D_hO|hPR^<HHBEN+#Cjw-ydf)RhHs;qQ3US^I_
zuEj|Kz5uq7bZ-WLEoSl>kCQVa)o}ah7aO%ZpO=g&6<3HEyH>n>%S!5caG6ssk`Whs
zAypkU<|KOe#^rO;S^aDu{%{ZS-WA?(KOLds{l3Dm$=&e;!{Y}UVDRjgWNb|REi%qH
zwApG|!1-A3lE~%w_?@4A69zEQThn|;T}RJmZZvqu-|ENVK^gXU51P>50_{Fxn}z)v
zbj<qc7!_nz`MtUqA5vNtc7w_gX?Q{1BXgD<D^}yEDXH^PMzc`S;XD38;dwmLud~U(
zjf@ZNE;NS{pT3rI<fbyb?CvATTr^4l6$?Xt2@nbW0>l(?q_uR+M(g3*poZef<N?*z
z<8$B``TZ((u@ZL%gwLyJZED)g^ouF^w2sm7HsJ(^3#C%Vu6|0Jt`c$8IKVsz>A-eX
zFTr<&V0%g**fs&lhbs_ZpuSmH0=N=Q3A5v2|2i+uRq4le+vB9t1NB3vu3cHPz@D8h
zUFaxQ*}+nKkaHIHA~nyye)R#+(6`SoHS$a@#gEon#=t7<I^ZD2C%lmxvB-SW?U~~i
zVq}^v_sedBNnxwWf=-&_$CHWdVSQt37+LB)uFW{5?GvG{OQ|b#If_IJQ`ZD}aAglu
z{z0{)gLt<S7)CD*<H4G*97pct_Jvk(9q~55S(m}kCa`5uS2{s;)*vQ-dAu{+Pr-~$
zRCnn}GiUa3W0O*F#L^i+K*4JQ$hvn8-d#z2p`}sDZK&gk<9S2v4;O0Vg7?FwR#Isp
z!QArC8*tZAuEq@`emZYfR~2X7^HPl=@*le3xfpfe=0(05CxE^<QUN<aKU14{o}922
zR<TdCx~`eFJgLeCn90Rc^3;(T6}Cxr{0m(~_Tnb)`nn3gDxB3*<WKs{w_>yZzP=3A
zd(Hasv+yf1u|e8|&2>;N4ucEn>rJYSJ7*T7UM7b6J>1ad$4>Z${w7$?K|#+7qE*Ky
z=mY9JlbvUt^D|OCc!drsOItRsfgJBVrFx3Ge2|AH6O1QbgY;PstCI=&3*u!Ni7&%O
zgpPY^R*jk;c4&~Z^?xPA`I=?(6N4Qy&FltRLkmFq7#jYG8h-ATQGh5{P=486QqDr?
z>H{AG>8^ir+&R3aBh(6caAlLr8kJy@^qoC+4PJ~@n}+=|V``(%Cus5;PaST&d-v=<
ziSTBjT^O)#CR%ntg)oET4eQT}{g%p8_)qhdW*Eikh-h*9jN-Mo)69F|whfiT7oxBS
z`lOm(CCnPMY0L)xNs-x^EsgR6MOcu;gdLjnNP79TR`y){yw$nx2Sz~xE?BPuhvN1e
z%A~;$*_cp35BuFkj%%Kdsa<&f=!c;}8AV{RijTA0(;>q4XnJMv1I|IbJCM(khEe36
zWADNbxHA<ZH>BK;*SK5%?!GAawf&)`-f08a`DM_CTV5D<bn{AOMVdr?%^hx^l~CIv
zKLsainD*eT&Ko>lfJ2B1Vln$8igQ&qX7o8P{Kl~2O=feM{<Vssj+En<-ky{~eXNC(
z+CQR?Fk<f4i)F0{^vzj#b(;q&7RyXb9QxXH8m{0f3LbNBdlo(}K+JrmZabuoms-?z
z&dPVMti{L3at`ey*by^D+%i-@(Y(Vxj(hB+Of1~^Dtl?~j@_o1pX)4E#Nv2?USnaA
z*%8IvILSnQ>{K(Z3_J5m8y*9ct4uCbrq1Ug-@9E2!d>~4;J=Y>_FeFuR<9|^09G-c
z)R4Eh8Wj}8o~`vYB8MlQl&bb4@bakGXvZE;<>l*!hwlgZAYS3g_BhZh0c(qU9+9_P
zRr|y&=f|1i;?_3Nh~&ib=+qhzrELJ7QG#lneK1^M{^OiWY1xS?JLn0Ud41E#%O}{j
z9|O}%?k7p5ce6X<g8X^kpKdAd6%?QjYlDJ7u{HL1-<8UmG_f(kr?Gku6oAVSk5P$-
zJ*)*EzTy5SL0kVg$?16`^=Ru?XD;hA2p|9}L%&wM0<YLdo?g)o+k$`JgelK2{d8Qv
z{GbgKYy1T5q$r-Vo#Fc2D5wnu8L5)gdGS+popi~<=a1X6fF!mm7IO05k*fy)dXxc5
zJ>~FL74AthWD#o9VCY^najpg*!_7FS4$gPiKsl%JC>k@ew))8PH(h~vERiAJ+hHjA
z%S0@|Zr?(QLRUQ$=;$D*;{}1asK9F@NiG|c(`Qa#AM_0ef@R5)#T>x$&~1Dq#ZBVs
z;RO~)S~|V35=nay&jF!lkz%HIZ@n}zMDzg4C6MI;00e@4Y^?&$>rv|-o>cuwk*2q6
zGT30ZV*_Gz4BC*9xt(SKw;z_Hz@Ld5C>N`|?b1HL_!<AXl}*<_3p({x;UzGUuxSOF
zBlSFaR9vfORA3&$Sn7u5iSPTepYWxw{LOs+!@QTBIha?Cm!fuGkmjL?E934RchX<@
zwV8bQMcP!YC7+3G?ZK><3q5X<%2PG%7zo;kdW=G>cbKiqEv0~fKjED_seJGI?Ws<!
zK8tde%{BZ3g7b;SRFkQ`hiz539>kzXGgbiQdy#5G^mN<yJTan1#URcaA8fL~s$0uO
zol#i$#v7VZgL^cjzaRaieu%X2`7`Boc|6+%mzp=qgIb*+0tmYAy)G$4E{si@_#M3X
zJO0$&FUzX(F`ZH6=4Ml36%MAkhR;n>&T&rKeS?!P#qpvFTd<$MY3QGRzQ^lIbEbUt
zIM}0-bdmI{rjr9N3%?&Hfr{<?AMCw(Jk;;|Hatkm(nOXJBZQJvNZCe<EorDw#8e_o
zXcL7QDNEKQgrZ2cnCw~K*^}%$WsH3{V;N(XchBYXd%pMc`+a}+^Lp+-p4a`?{YRe{
z^C@QD*L9uOc^>C+9tSOz9>ZM%WuNAC?E*tk%_3KBdLvG7BM4_py!sds<?XoR*Uyk1
z8xJ2MNF-NZ2M^xGRqScTg4bw8(r^~>$iCn1iDPG$2v7f*OjW0S;4E0KwD{ywD&XOj
zWixy+P2to5-a5VBENmb|xza9%E5|Ip95eNi!dVoQ+A1HHEiHeojw+Wz$N<R4FeUri
zTt~u}yj$fVQt9Lyb$ahCv+3HP0r>?x53J<N^gjAMBpjt~`zc=@`{{;G=C@{D<)z^$
z`bG@9TVMQ=cgvYjr<dzK8qu~cmyb`NE*2B_g{iyF)N}@&<$8tNtf1$Nw+q?;s}mbS
ztx<M^2P$M~d#8_#rNzBJr49(fAx8Tab2%Uu_mFOMp6*Hc$uM>PSZM2pxr#_gkNJ3Y
zICF6osMhC<9H#e=s&f9(Ee-L#nKo0(w~<y=X*_`6*5?$=H1<7A)#$RFMeo+p&1{?I
zldAqGnXl*1h4p`m`c4QZiSMj7+k@@nPw+rn9A}M;Husfb#eG$%%Qda05`9tG&T2<?
zlwEm*sXg6D0KV5ra?y?kk8b>^+yCr`?0ak=Azu|(eKbPcEXQ<jLOtWl&_)1J;i|hR
zd}(FaUFJ!c+-<TU-iH0NYkjXq#ECArm(_yHyyTt2{P9y9Ghu+RNe!i!(m)?E;Dt*{
zq3);Ty_6xNZ11@-xnh%_*KXfiIu`JR#{=*T!1SxMVwbP71FiLrMM<~H!zise{jkl%
z(NFBmYQeMKDkxCY#Z6FTwchwZhx=DI$C2SA057A3E3or6Z55ezCk!IOnkOe1=Q0p@
zkwIsG^?Nj7cDv-}*m`Tt)TiBll71y-MG@?5PH&GUdueJwq4U$;$ZX-;NhW<4kl*_j
z2^z=fF(9>_9zR)%#U)GhewpBQNHnq1Y|`C1GzMLT0DGTeUdPnWqTGp6EG?yeIqO;V
zQz=5#n)wmPz9{(X&47N2imZmu@HaQXNUjUy)947i3dnhG6l?2Z#v2p=NlgpQZN&IF
zhN@#}fFX3JX5<7lxb`n%*Xhx>8sf`xJ(lTqTEGXUI9S`~b|~u}0KZY<-so$lSc$Wl
zXES4eMw~J9YU+(Z;NKEz#5G#AO(?{0?#1=09QDe>UEJ+|)`{JI+SqQ-$BqTnrR9!u
z`(GVH0dp0I50COSqQ0-{rc0Bo5|!E%;I`NfytWxW76`Q9rC6pEuSgvu&4}D}Z<X86
z?_!74v;zjDG$jyhl^zXK-9v&Zi@&To*ZwGTqjIhCsVAZ`+x8drxVbCn3Y|KwK?O#o
z<JM3o%sJaNyRmUXD5+AcI5zFp@ygp0ATiz<Hn!+v#m*{HQuS;-zr=Ye)TcBWqg$ik
zkM;xxZ5!hmO<-iC9YT5{Md_)FZVQX?8~2_wn@I(->x%M^X7#0?$%hHwU8=D#G~MIR
zlY0-Aq0duRYki;+Uzb#uq^Ioc^*yKrVm&nU9OgBl%FzP8rC_?<(<xDb8&len;*UX;
z=Zw9Fk;?g5DbCbp-KO_s9ZC9#v-$GiK><lUG<S*rJ(%iYo@S@QV)WFplj>0@Z-PbQ
z*%zib`tOv_0#7Bheq{eVtIQvuC5UZ>z>RSS8Ajm#<P%|Cmsk=%@vgWvj<$2@nwB4~
z2LkGfN*VQ0fUjUxRNSL+kcA;zy^>6l+~hA<Y^N9x0<Hj?p88ES07yk@P?yPY${2|c
z=nwIw*L`GCe(0?nrRm(M@+#tp_Zy3lGyCRzWL??zN&J@w=Y^wc^R8isU9HA^u8n_t
zK>^tpvJK)#k7i0A=`TtdT5UK%vzbbFanHA14B{(2hbp)CmsqY;1qsuhO8OqA><nIz
z(l8HjdiUvsjH)UHi#vO`he#*9_WWq{{($WF5M>!sApdbAsx*`_6Ec3tBTw`&>{P~w
z24%wbkN0Ul9xwI8lYPfqqArEJV>oEg_qdTs;ZTubx(YqG-d<i)6>3ZP?0nqkAX*7^
zzKz<yFiuq?UpMi}rQ2T05Ah{A2Pt@B9{fBBYDKhJ@8R{NlS>Qkw<?NXIVuD`R|c?(
zC2Ix-s*@$bz-M{7p0+XMRN<95>lo|)Xm~Txc|>PL_?R~48<B(!K#QukN22(OkP@?8
zZgIID(>t1*&+O#Jj0qsT)c-fOQt>*38rv`KUmzC@zpLgh!}l7W?$=9xxB=MY$FMs6
za~>?=B-+(-cB<a)hD<u1)JF9PzLs}4jmP6%hGb9jgpxech7}Ctd_!KFDPA|~_Dr%m
zW+u?}cKF_C0P^f2wqrzidf>vI+_+n}(+(CZm8QD)xw2`czvv_TYp>)q+b5RWsAZ={
z%bN%bJKmqmP@jW8nrS@p-TlVG;Ml*<Z}atqo~f~xAGJb;RJvq%<9i1`U&}kTVkF+&
z4yE67xKF=c9-hCqJXFx8WU%;<hO)_-ckY8T=z`*ikz(duZ|VMvuW~z_ZnX^TXhwg~
zVD@9*LX`7_!OPe3dx`kt-YsT6+DV^Q){LjQ9W7;o{W9q1qk%*7#T@z`y<c*PQ*Xl`
ziKJaJBVG)ir=M3v1_vsZpEeM-4)9D72657sdypMF$yDWH8lpAXtx<_A9Gyd>3Ienx
zPTe2Bu)~iXsA@{6Jmn2*n0G(<l&>XIkv@7>Fws;}vw|0twLBTq*k{t*y=ef_h}6-|
z`t`WMA#?b5+1k!S`9ME_ZD?_QxvpK*mOlTz92w$<`c=}iFkT+f<mtFgVkVO#=2WEk
zUQbJTyyd1CwF2gC1#@vt4eg?I2TUhR^FUB+rU{PA%uVi;!Ab5&DiD*L|MGIm@TWKb
zF)TOB2wZZ8-g09`QN@6b`iEY3rKGHPX=KURq2x@aZ6=Q=!Z!+F%}HD22Yn15fJhEQ
ztOqeLN`smP`&WbJ^!ICM`lKJYFm{1SNzf!KAp&OO%fE;Q>o6EJ-!EW!whjizI%(f5
z{$P)W<R9JCoo@HtvvbS!GhShG5!dJ_aLyg*WmLH_Ai|`hnmKE~FS5?6=T`Cik0sI9
z<@QNsJgH$RuMeG2l{hx1ufZCGn+YoqNCIHWLc;@CV_C1o@-W^4PuMLT12FdY?-L_D
zqiAUy&||f5924NAE_j-13Z2!t!)<(u*bk_KNY*Q8CjdXLfsVQz-VxsfKFK3YC<#@m
zN=^~9P0CeZOgbdWG<b1-_XKtGT@u9tcwZc@Pm#3$iy+m(>M0cJaujM!`#-+bV0<D`
zOi_bt`Es=d&ysnD4**e&EHoKV;y2j5K#O|xI>fR>;Ei$EQefiadd7Clsme9@xm0nd
z0a2=Ny?wXunf@W>mGa7d4WG}vPHz+Bfm<dEw-M9??|jbG8p>SN;XiWK>otuqAzkG`
zUS4RXG?Crw$iy8kG=aUfg+^0?^b%*TRXKR)EGV_6sYlAj&6LMb^>!+25j<uI?uXY}
zitFPek6MPzPOfhd2aMB<9gN4#qd)Lwry)2xc(0CM?T}S6u<*}5%B|(k`@c4@lJ}g9
zU5uq#kG4N2Q<?&O%{)KPP|vR(sC3uo|6}$L)J*-X_$~9~{#9U`2ij3IVY(5upXS!>
zYd!X7yD8fhHy>i}`7N5`oQ%?U8Bm%;LF93gG<xQ76%h0aX1Ve$)0K!#IW3RFy;m(t
z73pJdf<zekV6pPqDbS1+U_Wq9f=8Wgj2?}H3V%P?PS|O32+Y=uoJZ4K5iA>O`=Scf
zdhBr`!z<L+D-YMFx{*sflF$JQlL7#G4@JVd^ia}yO62p0e$@shc_pYre=yb$XLNHm
zi~wmKgZ8%<thBAt{w|tzuSy#C@-|yMg!q`U{9?^x=Rku?+H73;Bf-30TkqJD8!~`h
zsLhmwTgvY&5Gl#SbscZ>)+?S*yJ`gxTh)@!Om{xeC)!XqUgomIN2_qQ<tGCgdnY}<
z-)BVIu8(6&G$u+=g~)o_kkiK%;u`b;HQkwObC}v#<pYk01WqFTH__AUiI-fQyGhGF
zWDA%)f(E=slM^UsX`aCA_f|!o2H(|n?S!`iQzp^Zv-F;yBu+1XKauZ>6)aNr3D|ad
z?wI@whL|{UHs-rk=O7(bN29!J5KZCSFCC!4>94s3X|f;?8h-f-bT6r+k-p~(LKN}}
zB$a1(Fe%v)DIte(ldALsr21JsX?)MnfsA)8yQ-l<CmqVKg-`eA8@BSi%WK1??So?V
zg>Bz4cd6<!8wv(IYs4mKpkX44mP@zl<o}_;ypE(xfkFx^tp23(G^M}cspB^Xko4M}
z-fN$psDQu&+~QcytWCVXuie9&k0Wr|v(|>E#6Fq`LiN_`QrMJ~QlNb;N*a$V_8P|q
zcMs3nD}NLJ);%Sw?UQxf<N<3@aaLafTL3(aM9^enBun{bhW84L-ag-#FE7;UecwZy
z!~e3lT&ev!E%*gq1|-?#4@h($o5J4Qy|+z0WMrXa)lgJE)Uf;_{`J{hZ%y+=rPcCf
z%(EpKu$k{3QEmf#v1Ji&9;?Xl^L?HIhiHS}t|Y{}NQF+L-fm?JW5ilCzppc;!dY6m
zJFRD}(G9-hFZj+u;|HCKc9yP7X}VO>Q>i^FAbGLZt!Zg_yp;1M*viBoq!$pyN;Mop
z-hQbu97d41X5{}B%dwMJ+ufj<*o_$!qpedgixxO`3I|+IeRQzePvN2e$Y%w`BRqH|
zu2}{P#F~I!LbEm{m`xxo5V7gSo9Q6ZW`ij*FEWw=gTVrOf4wK*Z~dcCfdNF<O%MaQ
zuqB@K1Q0nq0kP<H3S0wlyP56K-5IXcZkTUBRS$#o73-`#fP2ulp;bI~-DI+auRiHb
z7EQkYJD4rOyof`Kp@9j`z`SVmpP4?q<HY>_y}v_j^r|*Eu;&Ri?#s8aA!vcz?Y^J8
z^5oGLCzMyAniG1;mDSOe2R*XQy!LLk#TAH|`#lD&oo+@ur;_Ds4fCUjo6k&XZwb@c
z1kxEmpICq*2&>cMs?4#esQ>h}KRWG?dt1AG=klZZC<SNtkKQ(#G6CYoa4~BT&^?f3
zx;1sUab!=a`BxX4u<xNeWcay;=WS4f1c|?h>F!4CMnmhv4d5w&mz|4${b$76x?a=z
z@`+i6#Gg)&^*%*|F8Ly%MjJkGopp&F>w6#@d$F5zM!qn{$I<`gPgSzNGKeJs`Gn?N
z5(srSS$&xxK~PZ7y>puVexiFl;`pj*oMaiV!4p2idNGH05+Jty*HI1pao-)3dOdY|
z>uB!wF_B~`BdgL9@3*&)-a&-I@kLTcbdhJfi)PH_tsZv{rxNuuKk9=*Ntf{TNYr*B
z)5?3>MKzk=nS<Tz$(x~Y$P5Auh7~{)gL?bH4JQk_#s<Yc2HwzOEP(u*sYtj+>>vDo
zbCC`bfdgkj(^aVS9}m{<O7vU|yA%7%s206>NE)b9e2s<Ra2xnN`9EZ45L+DBfJ)!<
zDGh|8EJ};EE-_feLT-9e;E>!`RK(L|mSD!HeM*+A#72GY8IiK&9(H8iy}yXdGH5mA
zgWS+fIhc2k--(aRYxEJ98lfloM7?JS5r!Mscu(^K`e|lQ(j1>C^dAZPx)>P6MS+Vr
zHnt>dqz#yszGB_gI=;K9szIu$mv!P=52V*1iiCD~saA8KA_%AAv;zABeC<Rg-vz3w
zSF}=%ckS>dJ>Lw4Q;7)NshSxFk4fw=bOTP(Pjk%ZrCToQ*E7|rArb{p<GN7aiDFUG
zzEYIO<&LYLE4y;j6J9p2^5Y<MByJT8^V$7H>~H5L^p}B=%jD`&_H+4U%%E3I(CE#E
zuYB-QZ^J)LpFF^Qn+|~H!KlRyMjv$wXgK-qaTzBJO`i#Z2i;ZfE-b0yJ%3!2GdE{H
z!~z=JAMmLiI!C|JYsLgjg4|UW`JL|%zM0;%UWoV~8jvEaX?O&+;+-N2E$C{l5oc<l
zcVP6a4#w;m>t_dlhTm{78o<v1FUFVxroj9ZfxWkvYRwXr8y)sG8rO>-$JyTBo{%@J
zQpS5!!n@`OK0|{B>{TAZW~yNIiQvqyIao!pMEFKmd@a~-*KO%;t!+8vDY261%x3Kb
zODag?J|IC37;ubrOCsaBuh}Ty{BxDbcVjDU$MK8Rz1tkv<Jiv{EIK$M)kjPz!L(FL
zeIrJp+dAP<PKF|fLv)Y{OP@F+i~Dh-5Y*BJ<%W(fYK<LgDW8FMH0UAm8Bn9$*4p`e
zPZfT1&U@dM<{e7H3#DH{xz*-vSz7al0Sk+iF%foF<lLBL;nz1&RXefZ1Cva#0{CI<
zp1kT)_?{7kvWSR3m6}Z#zOKNxI?`T!W3}QZ@hw<h2f8u^RjVhYU$KzO5glxG=RmXG
zUxWq_d<Dm_7AzXFmdMzw25T!z1li?(QSK>T!cwlQZUWz6@TiiO*_?+cq<<y^;xhg<
zA)t>cS!aQdQFB^P2}8xu(x{%lW7P9!V@hICIvjijRBT63mbR@+>1D-{NQIkGH?hC^
zn@oes>7C;VDM2wd6%HwvmCbqPJy=3~WP#x4)jP4?jdQ@y&h7E|BkQ|HZ%fGf1G?ME
ziOh|~qoA2`<u9T-kf3fa*@Zv!UbN4Qx4nVqe6*K--?8?I0!H4*4u(|<4ciE9u2SRK
zR1i(OtcPv_>R}ehCwS>CFmDc}x^4`)(ETL)lX*#2>cvuv-vJeU2l(wC&UlYOGEf1m
z0e($NI0S63iGE}#+$KF|*xn`iNLlHto2hY<LBQP<5LE%)8;sSqQ&MPg6nv4wF~bei
z?fc1l&!4yfd$Prnahq9mA4+_a6@dGW-Gh<+g0j6T*Ym%=;dpsv;(amupsle}X{e#y
zm!ErypMm<WC#WoTdQj@dJt8zPB6H=9kV~$sK=n11j3^hG8~4m<oK@4bd+d}o)L;hr
zJruywmG2&C(1{n&XylJv4u5UbRS0U@p`;pA!Sai4YpUD0U+U;WQTL+%lWF$^zITf}
zS&P_stQdOqYl={%Mdq(${;$Q9+|bwg#yb*?q!OO1@(lT-U9R=wS2@--o{xE;=nrd3
z8pmz>zN`%xX=l9Qh5cA&%)2Zv`bWiqWtL0k98!^!@2bU48i%lzM-U#tdPL9i<t`2O
zz5Gbw6w_?s;TJ{p;?V|#51`Q##>4{j6n)*hJfn8tXo;%^pPO(G4<w0RIrQssOk^co
zyo9Elo2OY)ds8Tt&59HL*56UmXeYoq#kZ9g2L?yvF6I(dKiGzJ?PPTqt^N64Ggb|5
z=m4d^h##0nk|o%@U<N<G{~tO&4~PY&^x*&C;}g!tvcM6p0dad`5e8d+;5=qpAN2&h
zBnC^)(m<AXZESW?$n4VG(n-L7*-Q(=ZMW2W@eG@;K@DaJJ#JGb3}Q@U#gOVNVtU^m
zb5^&5nfL94(Hse`S#$~V<&oNL0gREMbKjTto0D?!tJY&+K$8hyv>?p4q1hbHTywy3
zVte|OuLM;jlF&PcRM4w1z=!ou0aa@pY(ff!<|n{$Bz7etlt2RdfByXUXP`_)!+eJu
zAQwLrdW|cDFkwU-=4D+3>;IF`+%f|sZ@8ksh@vjS4=nvQ12omWT0W9AWCbx)=GV$0
zb@XTmY2_j;zqR=NARDN{#{;p*QkEz%8=VFbkS2{~3$Jeq0`qlE(jJ!IM*t#N;Udn@
zozzrnYjEO7qVQ-imO-=xP5%GHI``ECzb&V8v=5}jQ37Z0OSJ5JY#p=cVHxs@poL(E
zvy*@{JP+K8-%-$6B%t-%)l?{VDdX#tQq^a#G1r}{g0{3E_+Z`*=J6Q|Z>{dJ<lu~9
z(90`=FJ@9ccR$#d{qf||P7NwB8^Sb+CGTeE>4@BMW!~yrGI#;coW6_4VY5+`E18!(
z1}YkT8Xvbfj}N>A18L}+gNewYVe+Gpdo5oY!N&J5Mi2~pHvwwE{j`W4HooXnHzLvD
ztC=-I$C-uvY4&n03pLqyBWwEIEy$w(m4Jc9Covr}{s&RC4iq>m5dby<Y4a|$6zg0M
z%W>RJ$f#RK54J2W?(yC}<95SXS6`es=sbSa;cDj@S^w<pqX;QP9uRNbt%IcGksep&
zIgjj(#J=Op8ZTW%W^M~lq_l6bKuoH)(S=QDDSMhT7k~!!^)ut>3mOQ&Gx|49i-SHC
z1R_;oQ*fo?iRV<F)7`0EQ<a1Q(t&brlL`4(i_6bK*V<8p8s|!pjGbt*Af<KctI=<`
z<FZFeo*GsJ-;bi06Zhczl0}Otv~+Td6xW54a@%y%YeSc=a&9-}nAewTX9@8%SSMO9
zYOo^&TAEz4ubsaWh-p!b3vY>DFvF9qM<JF_;2^;b$>)Dh;^!KQN$F*F>Z3qw;aU)y
z@^S6vFaW=%aN{Vt{Gc2yuM3(V@nDPg(mYtA4zTgX2oQllxs*TzOyU^Aisz6u+Xrw6
zQbC-3{I6qv{vY%BbbO~6R!mRNo(>k*;)A!{P^+64elbt@>l6KTnK8(n<>ott4OnOP
zob>V28*|i~mlR~>Y`&&$grrlwA~mX9CrexWp2#l^&Y7C?%wSnU?X>+}u&jKFu#RVX
zQ>HToici1#M`0$zePx^-cinovA@|FP6Zy8b^<QgvUqq_!wFI?>bnGVfpP9{zuqcZU
zSeFsDorulWDZe7HFKXYmAg2fQ(|TWU-30OqiH|TNm_ryn1?qAHbN+hA8ql+!)qHB7
zb4LJv?<pSY<I?}YI3rW83qo%3i?I1qDNHT;3Ya%auV^CMJ@nb+)iXWC!FA{2i<qc%
zKX7`HadcY%wnc#daJ19=+t8(6EMYD(2@s;L&^^v|t1hnpSiYGop0|JKqe#K{sSmrl
z;T_}5vvx*V^Ju<epGKSlrg+wQF=i@eHM9M$3jQmYwg=~vMU}4>W9R9i>>BBY72Xrq
zM~}b#eWQO(FSqRUI<QNf4ug5Q0f*-@kQkYPepcCn1_)DT(3e0Dyg%k4GCS`>7T^zJ
z>a>(Am#v^5)qO)ew+CHjZcY#ivVdCYBkUyhGnQiH$6XXP_>?wgP#<63)4d(^99qJW
z6V6<Vi=cF2X{Ot{BwD^ZAu2jwAn7f9NS482>3fZ=BG+V(s_fJl-r=cogv$TWw(VK$
zGsZ?P2n?Im?5}viAzvZ(M;$EL@7Q@4(f%mf0r%7+-{E3LG5BOJsl6L7Atzer(&Xx`
zPWKDuLJw-}629WdM7nZZ%&r8c`IBgVjotu2>RuMR$yS~t(q-S*wNOf%EHwW_!X%He
zD1B5aqvS)PRs0j5{TwRfCvQtH01-3g%_??Q^+{I(3RVZh*<u^2Dn;@9apuPk$h^2G
zVs!HrW*l1ql|U~)14CzM{Ey9u!?v&2ZWw25bN9r~W7~u4bcMas#*Bj38Zk%hNlk*-
z>%|Cs74Rt{-16i34()nG3ITHum!=!F;+%@bqJx<Gu5F6z@*UYp?67H~xP^54%6%L<
z^g@(&^qeri!(CT&`dxFLw{Qmu!O~=&^F7%Ev}pO1oLo|Gx;&cW8#aAj+(VNA%T4q}
z_X9s8Ym1KDN6d=%re}LbPSWa4(Xso7Si6{~iL>PZ5+RFp;yXa@aQo`!(t|F)1kGdX
zisI@VwfI)O6FrrF6xABWRH=tbUg7<W#{RFrj?OBa+71b$N7GV@r23mMLwuJ*m4gUK
z=L>h&=qAic<Wa_h@BK^h^;I}{cYCm&Os~or+v4}{UVrXI9OUYP5>j^Sl$VCCE21hB
zr)XBv3vGl(f3zVq8i&r84x@@Tdc1_Sy2UPa?|)mDK5$)iY2+MQtvve2#4X59CP4e@
zrM=M(p&GoI|A3UMs)K~Ouin-)AG%*g9p`szgJKs;L$&=cP4_G(u1g?(0nW}NXs>gr
zjbZW4UFo0g_DAAyFMda6<crW9!KRgGM=b(h_K?1z(jd3n&ugAenU$BEsJY*wV;V)~
z+Djbar88mPlu>%62~!C-^t`q}FVbpqZ}^)7u@BH%2s#Pg0cdUp_TFA&;LOyiT%1iZ
z>XpUiU^EJ36Yg=UeqL;ZZs9hI*wO3EC;_+&+NCP<29#%eDz3z!>Vx$60=mklH}_?g
z5B~hRC35~xVg^kt>&X)aE}f?#=?91lG3hK{>Qwp|v}ZVdBdAhqCc;r>%w@p+0wZ(i
zp*<RW3UIh$D(TT6x|%4jjl(+uzVQeeOec<q$R`F4KG{nUyfml8wfINVthjUGMnQgQ
zWLMX=ox(>muNX)nxQd^MEqYL}hIBFzP-aMHIB>TP4K^Q$sDBw(@)jA}L(^F3y=)cX
z)V0*vIO+UWkPdq7ww`XfMel8B4Q*2;@5zciIf88Ebuim%%9L<^NH?fHNN+L@wt&iC
zHKp5*81B{22~fa3d}lq`RBymX<*NUAlIwCg7Yr8@&H+cgDA6$H5te%Bftr}J#R=@e
zv~3+ch5o{b?FeEJVe|`%(-%FC9RxsC<J1q;P<64WALZeQ+Y#?AW7+yr%1civ#Nh{R
z%FBaEjnNk>x)&<)mv({NR|4Ot2NIkYrnEHr{p|RG5=N@p!y4|ZTh)lP?XM#$<8K~$
z@}mV4VJAHhyP9uPBGJ<@dLG*agXk-hq{<jgVeI&TY(>+L@UTJv=bo@22eRbUG-Av0
z3xa^BY8Z&|>PVj1LuO04kK%p|$REL6Uc_diDc6ykf-t~J1`NTmGBC0i2f+zP`C`+)
zD46GJePaIbQ+#CCxdpY7%ZCs~)c|nFK^HbHY>+X7{ohaEQIx(UsV|@9HA`)ck{&AU
zD?B>i4E)r$d~?)5mJQ{PMEHyj>M46;j7{xa`%bg-;xa{?LIP~=tji&+0fmQx0iB6{
zhyx5<nC-g%pcydga|0Li6>q1v^(g)p9fKahXy@nkDce|X=><EvM`G@acB#GlNSyj9
z1r)t%@9NnOxXST#e1h1n2sll7#R%pd{LC_;$5Pcyy46$o=VIts!xG2U57+B-!q2+P
z>4HfnGrjG&E_&MN#`ATT2{p4PpX=0&2_dobcAsdEH?Z>`BC(8_XAl?$HU@c)BH^~J
zpHID3Mm<_$<_p#3%l@v#rL;}oQRBysa_{bkvLBjnt^O#-^*dpDirz@_1-v~pCM8yP
zI%qv;ZtNtVPoTV0rE(*Qav23WW+Yma;BQw0@5Xq<xAtxUebnV<i55w@%-c7Eti-oT
z_tJa88Sw>UF=$G6V={`G_w1&0ymYnA$k@;y&ibwohi^A+L_Q(9Q8!qk@wHFA68<7~
zL<Ju44EJka<h;?|FHQud(_Pi!ih4rH_LPqUt_<$D=Z>GQn>^fR6{DVT$2>#+*?>7u
z8KFkZ*TRRV2_xj~d;KH;%wHZEBB2jP+iU!Zx-P+vd%V(_v~5xR`#km~nqt_f%)H`?
zc%+9a<AIrxkUO~H8rj7<)?VYtp|CNjUh)mu-su2&_biDrg!KaGm6+Dr3xn3+LIznI
zv^#8v!>Z0J#G9Av>kOvvs+->63T1&3JRlYS^J#k>GF*!oRW!Ua&sSyaVB!)kv*&97
z`Q?#Zc4v3u=KQGGq9YYW^Qeqwq4G@J9JRT0!$0U;au|4CN&#TCM70<D8`|lQHZ7Dc
zmv{B^>(Y@!g{3Y<PhySU7SgUW_9fEm$QYJr!jQ9yj?XB4aj^%n!FZG_A2IiM_jcSO
z>ID#>D6PSUQPKcNP}()=QL(+K*=4_;zuE)fUOzml%GyI3-w4Jlc;o7id*2#AE#Z_d
zP*+vcIAOwZvu0v95W8wFeC?vL-{*y3+Yi3%HGxP2HB3*g1`hPw^KMm?Z1vmg(vY{b
zdJ_@<lsygJKd0-_MGDcMO(YFGKFv1nX!Q0iyG6?NA+ZFX*Wv}^Yhb1R)KI=C(T3v&
z>X96;N7TL{s=nXsO9Ff2F*|lWMcgRe7ei?ckB0^+l6L3&{_vGuYN-6Z>-!DsDa~3!
z_{ed(<k;KwWgeWS#F2E=&xTW0Qac=VQ1#|KYvb308$1FT$VOGpIWVCw^3FO=tfPof
zso$s|<3FDqu7-_;ctPG*IC8kNp(;L{rF<u+E_%xK=&xp_jR9d^rYVq?5pEidZ@3dW
zfGhBRx;AQDUVm#uW~9v)(BR*Sp(tk9NYn9F8_=2Zo6JkPLy*a@{w8-f`b%)jz0*qw
z4W=y<J(Da$A=MCgX6+B{uQa%m(Il{X<esCHajR#t6xAwX3{W3u5_6MtaZSr#GOk@t
zewU?@^FK6aQQ>IHHGm9@t%3+b;{hS??9=rqx~wT`0Tpt#bQk~yowq7eU_M=d5O4hJ
z@0@wMPN2(TSpN9^g#HOM<FGl;obG>9#I+tNrT_p^4K`~s0SIy?_)%`4ijwA`#FC<X
z9aFvC>o_+oD?{6a>SWZunqx{e*aG+#R6*1Ww37#?3j-hf`TK~Bvl!FJ(sBE>bQyOJ
zsC?IbY&e#3(Qi9yH@<g7ILNurGT5^=NJ^ad6^?-fh=e9#l!z+4_&n417&i2&-5XJ@
zsigb+#3DTQT2+|rMYKm&+;8fGjB};{-^J~nHK`yaEG@w+^U6JBraWkSJpgA3Pp=)X
zfr{h(-ET$5?IiN6HNE3GD|wHw&w5QOqDrL75V!e-9k1X+7~-lqZuq#Il!qL}F))Q7
z{9Yye#!l*^fL0$6)fkeUuhSFj(vxs5&E+5Zs5TdHqVBK*&ae{LoFU=we-Q<rpkkgz
zrNhf_G6L6L9a%ro$R9l|DZJJ!Z`H&S;cB=<&!Lr5<-d^1OG-;i&b8_DzRlZ!cWy9&
zU>Q4BM<j39JNROCS-%EfRjut+FQlUD=$<V_M0O$t?Z1W$f~AT(DmBt=of(z|i;+XU
zoS7TrRUXHKF75cCvm3WM4x>~krSLO}QkW|Rz;6HfyAQiA!@SYV-OhXH3ZSpr^jtl2
z?GF~C;OC_5$tT=#W9iK^)|27(bYD;dY@@P2s_JpU9f?Pn3DDQ#GV6;KMOO9hE4!mB
zkd`Wp{viT@+j2;})e64HsG5O5=}y1*%ctVaMHu_mYYI{-dVj%1?@(cn%JDEa@0f-G
z|2%QRD=^Mwa`kw?uG15Orq?kwx~^Cu-~AN#95s)`)$^tAhkJ@{BrazD<Jdx+!qQ$$
zk2lqWH**<4x9}TXf|&@b1t=@6*z3s>2I>fZ5$Q-USd#J;%Svdk^<Z)|u!gEuDv>mB
zK4VVN%>asf5=yeJHs>KXu|!eumdMipXUk@@{T(JJzV!sDsJ}@YICmXR-1Qzgy!`cK
ztjL=ugv*I*?exQ%tgW`lSMK=L*HsX_BZZ9ChWA7X-34#v(4a@tMzk$ZlRHZ$1mL+T
z?$1Rd(&$*T{B8&HfE}aeNGVKKmaiD)VGVYdXN1ueN>fk~dl%DuZ$)C8OpyD-)u*d=
zTh5*UcjpGs^y^qk6KLCD8K+#AS-Zg}HA01=*U5MOBGRO@2sIwiM7r-W@Lml1;Ygbb
zurwe~jBXl>VtZc$-H4_S$i5-a(MO*EL8$~<F@dEC12*9F8>sk#qrJ@!Jd4lYz&}?f
zF(j<JNfcY`Cdj3b=mnMgAOni4z6nkTz|5PwuM-9&JlFzCTpPR)Slc{ok!63R-r>`f
zjSJNdADquijES6V9l3TLb>kSeP%C$uhL3;Z8NRu@@*8>~3*7}JMh&5{@eL?x-zPa2
zhCfyG9XWswOGGf*f&C&=bbxh_dVmz^>v$_Zc~94MBZ2VwhTl!w_RKxsCc#eYTFcQz
zVoJYuFQye1l%C#ae&(ZvFh7DP9e)Ae1W_&_gRq6A-~G61#fAXAHs4)}=oG-8Q}zJ(
zNR0Lu6Y3_bL5|iTX9dVB4}o(BgUipsBl}vbix=;8+Y9Ny@Z)DeGMG~WpliS<e8W>r
zjoroeF=W6a@LarcmtHkx+B719hR7iUyLUY+Qyi&&`xg=1;;RCX8^c_HF-@~@_2|1q
zx)P}lJ7hX)pD~)Ft9B`fh6~BN`Q&45(1qwWKQqJ)V3&+`_95<MS(MQoGwn=|#1pgp
z^!_AG1d-2bf8B7jVaE}th%{TOWm5pG=dEU4OgQ4~5o=i(Rkf1R(Z80)=N^d6e$FuG
zngM93GEWm}Mb<kCt|>(`4(2@lssnbsEb1kCLGmG3xkP{T+xK^1LsN*c>A~iOK4YIR
z*pj^hLCDgdL9cOOfAxVc<qnY8zNK&=wp+1gdGo;5WBU3C#Xx-6RgGD9Thm~X;UNIu
zVE3RGR2AWXVm2^#$*m58-l2yv@Wq_5JM6#-@MNw0(T?cRw}cfZ{0_!%!psygJ7*G0
zS5Ranqy*@dq8O)g=CTpW6fDeJ1^|cmV_3!t`0br%pr`+n7yb1BCKhmv&|EMN(T=u)
z50gr)S|W^(U^dKr1;M7h1vaf?KVjh}xY7*_us)#e)>N6Yl!$;G@!EuDSL!r7Se3Bm
z(?0@SmpumBjp-qv$thIn@u@D$<?%Aj??mH@?vLd_kCVvLKZwuu{>#5vBC+-m>t;W~
zKTx|SErP4N`zGm2Ca5ldy{O{i(4Qm9-f~xOygVcCd^9`FACY%<Y`V5xfOUhSmrf?_
znZ4(l=u15zbbDKxLlE}NF6E+6GDl@r1Q3LykZ7e4UXo=@MFyFQ!L1d}T+-j>&}Wo4
zQsA9`#@R+=MhhRw0@hcO+RMUCoFq?e%DRQ35z5q38d#IC3+()ArQz=tx9l#6U@tU?
z10ZDUtjE5B7H)pvVrjjnqc*^3Z#O}-($9@jN+yO>T6^o4bmdBr!yc`nvP4)ztaH@l
zDr(G_Qx4Hp%`|)dX~W?D>Q!Ttp-?nGFB?3dfDG?c4<!8a>Ex$No4Bl$!jmwCoX@Zx
zpI-1H*Np+~R`O(Cjr{$2(bXTHZi`$K295zyKttZagP(<<<5K5KZSHCUBq8_poX_Sf
z@f$))dgdp`#T)!^#@QnJT!?#!^bg4QkQ;F@xDG11Q^R&r9zKBGJ=@o1=h9D=S&i*9
z-LcB<li^gRM3LQy;W~S$moHRxtj)JCR!OtAU%hoVZTrx7cw1B>^x4q-r=|ZP!yUTw
z$5Fg9B`azTX#<7}UTOylI5Al#=k<5fRqa7Zm{OZ)l{3-oLCf)c)u2>&unA0W)2BgT
zeRT`3=K|9IlmlDa7jh94)!+>kF#;qgWh$7E13A!v)eE5U_s{?K*R6neWH!NqXz^Vm
zUh6jOQ^a2$blD76X!UWRxCr&))~5XRwt*+uKxl?$q4iDnrD`Pj)&>O5z8N%T9_vl0
zxdly-)<_H%oP`BgeT@oXV*U{-V;#(7l(W>qU(SX*`q3iQTBc-E?I1mAO8=4-Ct-3v
zHWoUKEtOX6Niw30WZ{F6IGp_V6w6D5bb?ySUQZW5m&P>=g-1-8J;}L#-nOR*<m5I8
zCK}%PkSR8c-KUYWEaeTfWIci&C7e6ar-pnzh2Ja6lg*LHPVO$6w^L~3Q&y5IF4_vK
z^%a)?gjH|V1wOJj>mXC!_hiZL!<xrkYHb5@0+#H`&_T*GT8OOdS*xcY$-0?oYxKnN
zUeKu<30`tpb32F)QS^hLaIyRB%S{!DitrV`(TCsD`!2n;++Vi%o8sg-t`9PMuqBqO
z@hsszHB_Kvc7<AC!rMyYGlJ~uv@f_ep@EnNxZ7$JK}{Rh9t25>)mZnNfjVs?r03%`
z$%z0<{+~a!MNe*rJaC&Y=q9<>>4Y&tphxZjXQ!~sC7babKR-mZXt1>gf2U7=QEPng
zTIN1Iyn_b-+9GFRjAR30{pavs#JOD1Obzgae|ExMq`h^x-8(A}d<-n>aq!s0y?b$*
z{w&26oCt1n5NR5zSoVVw5UrH4j^K@n`5srn@1>{M4L|pK<hJp2;S``MFvQ8&9W&KD
z9g9V70B!G5uW%d|$<5w&PP#T1C!re^qF(V{`%-Zzt#J}Soxdugd%|}c*nw*~8Iatp
zzgH=aA{XUyW}fcDI&~30+FcwO)ZfYZqxA)rW;n!bhj$c@`<=C+UmjQCrIA0Ec6;s!
zOaynnRfP0a<lRJ|1%H?pBTSHDp);e_yJnBolJwh*R*C1pCsg>YMT{KfnQ%#SSRVkB
zR`y`hit~ThpK2TSC-i7C3S=_Qwh8MlUnEac<tD318PU6=|01enpRGzT;y?zT)dP}~
zFuW{~q`%F&7Sz{Y^zhox3H<&Butvcu^nN%@AO%K%P0*yg9u&H8_4gIa*lT_ZErXW{
zj($xO8>~|vjnLXvkbO}&J>XMD|Morhi@K$skJw)X8m!+l^eWO3vuXVsk0I2>v2fr9
z32Fy&^&sUPfAim7iDlFopzIq!Q+$vxiec3aUWP3<P%$LRBn0?Foj1WH_I)QrpG5up
z3;3^{M*aDORT`GQ=P%+Y4SbT=|FySjvA%$1vEyM7PJZdeq9?KUH$fq`%a%%Hi%Ozd
z2yo{`TKxM9*7E|w|C5hsVGA~_DC@ctrig_tdCkw7i#r}*lz-eB@&zF%;Igk8-8RP{
z*ZZ|e<;|Sh?~O+=49)79AKE#K3}xQ8p}j>}nyOZV7B3tqIlG5!&fdfc6c?FrS1QRx
zD)k&nedZ6ET_PI7R@{Mrs3Tr2PhrQ07D=*E7GZ7q)qx)p)mZ22wpp(_03(VcJ$k!X
zx2fnL&%OLYwj;*WBfpUc2MQcuP6NUP5y~`6xPqnvra0N7(PMy(HRw*Qe)Vb7;)uY4
zj=R<R4IY#YVj3EKgD$oVu2&7pXQ}t*owJwikZZ{PF(5P)e(2D2);@~znh^9B$?u)k
zfZeX+xhvz7ogMvQ{m_L8p6j2!ejCl=St0fk1~XvZO>)&klfq8;;E-NS?vdSdao_!7
zd+wJt-+fLu=b1GSnbH#n(zzn^#*u?X9#S)<vof2$Ij=BXB^n+Pc&-kI<F3+&yB2rz
z$GslirNmW2d=2(*7C+AdZX9uy)1HJ#hN^9TZIOJqjp*8x3d}b<q!mWO)n6o3<v{N^
zGE(!kAwn07P6C!JX|WjqH1En2N!K8_l`?$d^0`9!(>3JCCyLKT{7)eVRpE=aWQZM<
z3)nh`;MX5OiGgCKZCbAmGq<yZsZsS%)X47MH7U`tdE?IYlX<Geh~VaDS0*MH{c8s$
z79y^!BX|+4b+FtW;5Oo;q_6{1z^3n?jsn;We6a!)rR($*#%^NA^1pMVy}|AQdv9XC
z5?p53mw<V_CB2{dZ>{=6+EMF5gmOZxiN$Y1CKx+oWpZt1vp+)<72ZuSlnmVl-Yw{r
zJ6f9T7t83i1H?uk?}>v=OT!8CoAb1@@J=;APJ5U}7`V6!)Lg;crSPu+^-VR(eouT8
znn%a(qrRq@&kbtk#$I$+#n9d8mkzf*%D-<ZO0Ep`3(ZI?Jb85sr=4o#Zuu9%=w%-Q
z%@B=@Fd&wxe5LT4>D#F;6|eGr<a*OG%(qj-I*_a1Ss*>3@Qn9;q0v-$0@o7+MiP$h
z+kL-MdhoZI9|y>-8rhX=*g!3By5sn8okLpdsTi}KbaX}F@5#6`7zAaz5*JM<G%J0s
z<Y{lXW{KFbKSLKP(h^1azv!OM!;TRfkvl<=SKJ4UEjp1*z6P2iPknch+$S5k@Dykh
zl3+f;@h}f0i>s^1=$mC(pQ?n$3u$Bh+}FCwt5NXIak6QYFuo7J4d*R!GXrbnr$8cK
zxf#cHEnTc(UN&Q1_Pz0~U&RWC*){Z`w0mJfSHy%(MvdGs#K!_d5y4@aTThXsA7-qP
z=-QV*w_8%>sPQhxz|gp8zk+WcklhicR|-4tPE<e@K|lM@RT}lrcKex}Bx(iI-qP<p
z9BW0_nbS)e-CSXe!pM;3pQ`+Kx9?j*u2C;&BnkXB;5sC>qf6XPdQAAtX%~k}GbcfS
z=Y=1uOW|zkrqo?sETaR-rK$FgsWJPbd&qBKaM5J|1tSeUuYjq&OERx+rbVAsM&85K
zRV(A`2|W0tHYU3YN156OQOnn7j~)2(i4kX+g?5C7pc-sZ#<R7SR$!oQ=Jf;R)|q^o
z7WR0bRLjg$wtx>nnf2g462+v3(p2-bJsd{O3)#C2-FABNDCvW}$6+sIvN6NDmavU=
zKZ~VCFD6I$n3cV5n`<$DQ5F#zXEy#6b?c-o<UttyfGRe6f#+xKr78p~WN|lzd73&E
z7k1ye*QI=vxumVsfR|;7QQXt0XUBPhifkpF%F3Xg)Kw9&{pLS3^Jxe$$eS_B_JM_>
zcwHnKJ0SO>$Z6*p%}kGZ?0YO2kls|pgW8?A%#NL3v#r(IO@RW}-b-D(cGu0JguiN6
z17Fi5mK(J6i!75ykwL{r%U(;lpEOysiag6|K=|f^?dS<ok^jWbou_Dnb}bcmnr_Ox
z=+d98&_nK0_sKC0dtC9&EwkYGO%A<@{-q-kREL_8{naIf-M%InBY&=k*YT@ucRAV4
z>5tajLIkAwzLl3!G(#=Xoo*Hj*Ya5xC;i#V(?noK5w22%9eUh-w$j*!bhRW_?WCla
zteVnvnFb@BdEb`{Rb;fl%%eZuVdH1jYX-lUKqf2++PcE0h|}{%y&CO-(mFwNdrl<f
zIhUp!pU9DqhJIyGw*d%s8YFW|@t_Qy$1>L-vJp$J9Lwg)B`kM=`0qQ(M8UsDc-iTI
zxqX-MM(a+*t(R6CCf{yEF7c@DZC5h$1C$FACCY`chAaky0;aIYlr>69Z?}y%fZ*y@
z=M&Ynio>hT8Y)B692mF5t&`v(bjijG!W|^n+~dbXfhAevwjqg<c1I3!IKSLGGAfN|
zZu&;7mDDEeW?4~t9@Aia?wE?}ylZ%jd{;gI$?NJ$hYYhmr}u?v<Tiw-JhnYyb(T&P
zSwXTy!eBFZfXGxJr;3-M0uU`*(EeblbkxXv8n>Am-x3`knCvmuG~PuRQm6!U7m!FO
zl!7wdE_l@j!Kwr?8HCv8es5*C8a$9QB5dCN)n4!es9>%0DAsI0RrrN6L*xTzC6@LR
zymotw+I1r=3NS4eMZi5-V$DiYtVc7qT?axQprK#-|Dsz?IEDh<%ZW^HTeud0AdF);
zhF38A5}{m)i*gp`jUby;YBKHVpB=<)O=lLT1F3cRSuf=G{$bRXseshbi0^)?*o(m0
zq!#6k>Z8-K)<b2i2F)${j>ZUmz9!MS-fw$zo~ruM4pN<a!V}pCXm7t)2Z?Q%fmK>e
zCHt-3vDpzt3w4HiaC6VXC|aGe+BAC6nTw&B3ZB6zXf1DNc~f$$E67r}NEQL-wi{d2
zyxxwk9na63VlYS-isI&$oEH~UFRhr$xdF#w1^;7u`{~&&HPSqN>wnHTmKNq0Ke-*<
z#z}p~h{pc96QbaDxxV~Or<2pSir9XS1Er_rHv4B&Si<8pdPk)rxgsTyfL!My2p0(V
zY3wjQYO?n)BJTX@i6ulL9YEsZ?Bw;7D3Y`<+U1U0g;Dnrhw2`8`OA|b(a(k3H{Jqc
z24)iTFx`)(#?+<r{yeub3I8aG?pWtey&SsL*A3_=`~cwI`VwJHk=WZJdGgWlNfnh2
zBM;8Q_dm>JZ%43lR-*omTI_!c=$FZoSZ|_40FS&JAa^R*ub%Wxuo*xO_d0qEVr*d9
z+&Q#e|1qGmHKYSS1Wu*MO&v@amL{FVuwJ`~Z6hoKTvG8RsA2263Fi@P0F?w>Bs8bb
zfLneQ4TrljR_hbh^<*h<+UM)%3eS4EQwmQ0(0&uTiG+88_n;7;27dpqtv!;nFo%Gh
z;N#vXbGjI{wBCN7P^BhA^uwi7M&^d%#;gW>yLS^U`zttw0e*zq`qEc>?tjXQfllf9
zD84q{_>)MlfcYdTcfGM=ePQFad~PK+z4>A>4zNAfZXWiE@`Tx0T0q}2`VL!_e5p`C
zsB6iz<dXN{E)U>*`t%NKPbqmH5GjCMRgt>C&3r!3UiO15NUNe3j*6p~1+g9S_rF+V
zE7fN5@_Z$>snX>zHN_MV00_H%^~;`pgnFk=2rlnc;dHxKHxPqOTjSXD1nF4q{~w&+
z*NugSKk5$EZHE$<hf?ai^n}?TVVSGH8y(QYnwOfQdnK$65e~tz$}2oDz?zxrBhJa{
zh4h@XLG4+s7dRYV=E4^tMH5u_eZcMLOQD#y3(exS*s(>*3MIwglFp)UzItPI<B2;#
zi!lug{!xXW^+6l6gj$H3Fq#6D$l+Vif}hPCGrmzCf7#|0M~x2o^n>$;S`SAkyPNJs
zRjWgbRBF2A(XFo*_dROvSJo&DLTvDKM84r5{@2iZwfbmJrQE30cE3AVQ3>}XB<WkC
z2Cr_pj}^S77^`)c5tQaB5*u5-9QP#2HI)&AjRcAX$l(}Mdc%EGod?&5%0|$F;vV}s
z^O8$bj%}UdQdc-nx8K)CWsdBjS5SzJ56-B&d7LYMeTLtaS9#AcVuNVdv%cfe`x8?u
zUV?So+~kfeAX@PNb*xbZw4tP^aDw!g_0}_OCahy*GA%5A;^3R-lL*2HCv?B^S}Fjw
z)Usy7r;d5LQ{ANu4G@G=P_cDW1cCO9UYARMZc`RL7D*m0DjvCT&oIM%_apJ4oeh4i
zOfK;A)pSmclemUJydAriyrMmd!gPInxZ7Yf_+i`VcZ%|=b(7Y0hGO`7i{#tHgCD9A
z#IN6=OhHef(G*mXSP=dw+zyL(tIz+q)%QsG<Qg;g)Z3lbE#|3a)e&t#YH1EHz)sK;
zrSBS1T8rYEiPrHuBA*{I3HaKfQ&orh`Xu)XNqWEDapk5%#F<P6$EE&Id>5)X{u$Z{
z;vK=#pIZi)**Pfm8%NY6do`hel@Oq3wUFqCKUYajFYZsD&Hu4)XogY|cxvccO6Zy@
zU^8`K1-;uy6*7cQ-0H@7a6_x@czpezhm8{T;r$XX7Ul2niRQokI-mHp)vpr-5u3_k
z>e+@9gUr4Qba0^{2W(*QC`LgwvGN$G#sSRcfZ{4~<^hz9rYoUY?;&uj>@)$Qo9mb1
zouHMckdR8~2Y0T6^`Hn9a03B-R1M$@dEm~TX(Sc^AF6s)y7C6=T`CA4?+6Ue{>4Sm
zDXhUl6O&?aC1^Nwi2Cb+f+s))Oej&U*}f@L#f;0?L=wedD>*s@lA{<_A9UXNpZ3mw
z_E`KsU;p2JZ1v0Z!x^l1)lplWtF76M{>2D9fZJ-Ea_x~rm(M&#9WLj;n*k4f1EV3!
z5apc5rXoRzuy6o?9H}Fa>HMl1ydzi(^2aNj(_^@vcoFr*1N-?A$L1o4jGsVy@&Bkw
z`+`en8~8mjZJp=^5s<N@<}9kxRTNoC?SWvg=s=8v^J65sJb{(`7}SL~a{eONB0!k$
z5twKHM@<;)jTlvGDU${<ba7zpQx1GzK*s{&QV4kGqQn6r{SfF55F#KFxIa4QL1f8C
zvEIjn0L|n#0*w{G+cn|?$~M(tX9R@c|8(G6@zQ$2l*;^eDwjprgC%j>y-_B6y?*@I
zt7AO`MmD>k_vqoY28pOpAXKcnuRTjWLZx^uF8ay)bH)q8NDyiT{PKXNe437<KyK?&
z?%1O1lE<IZ@<b&=^q+07p09|kq3ggUd;r~~DrP3$TjcGl=1P~vWxYQ8oa9=9TOiWq
zqH5psnbM7Ny&H0K-C3vod?xvRIT`OCfXZnTPLKhhJgk!2S%Md98_&fwCTmD!?-LC+
z3;fVjVxmz4*2xg<fD9xRqP|zjF4x3)<<FH$yu06g-{;}9+J39{n~?^1sDGlACHipu
z$onjuy#$?*@yEbRH(Rn|GNQn3Vq2i$#-YIv$P@sC5LTYrgG_;_z0BU-%k0HIBT!7W
zqYCju(z_fQXWYA2p9SI><~-9G{~vtZ>R#}8e-4@|E-=&WxY@s+zEf)YJuO02WC;b$
zoh0DQgcR@Ue7&&y>DURNg5mYA+J8z<$<T@JGB(h^d?NaB8Z~f!^F?BB?y~d<CbpNR
zIUNe8ue<4LxE)#fiCOnLTB)+;C<A1%VRwS7N3a)~7uKV9*R2fe7M&Ohhnx-{l~y4s
z1K`3Bs7b>+RyFHE422IA(Bqv;hTiaLW#&H~PEOOVkeDTqf7k@Gg<G&JvC_W?5$qZ8
zR1X0ppCAO?X+iBMY5<o!-_*#X*r3_ygm+qzqnsSXk{PZQ0=zFUlhAnq?6n#n0`&@^
z?o~08aWD>|t9rnPpsUIgm|-SuG@fSCv1aSH{b!3M)i85%yrk-|gYgyr$$(d<CvN%c
zsE#hPgth}7rAK`$PlFkC*T}khGi%&E?r=I?PW(@k4O=DRv0+ZXa|7NB-s(#(i{rZ>
zks^p^#QRWf*Qw_ky^AlSUvu*sD?Ysc`WQC#3I_}1EH(Pj#GJ1KC=0|kt>?giRy=Hm
z!GOgbM6U@Ge*&yRjP)J^&=L?^53w!5f?WbtL)HMHe<6jzYtA#cr6Pd`&$s3VNb;w?
zfCDO=!jQBEv6*sic~c=}H+mi(1_jsC%`}(~0Dp=acfgqiC%$87tL_L;Bs8njp8<%6
zMq+Sv@N}h6EOJa#AqVtJL4)^GU9!g()$cH@$=C&EmO=y-su3&jcPQ<>@%T@7kO*6p
zb&sNh71G;7PpK<JH2XNp`un`n<k)<MZMCd_8XVwzO}2&4W=_kq{n+`VmT%T%{n42!
zqN8>*G&SO%*E@bY{5hRdJV7dGYWSl%?)4qw^G-VN*Qt?h^RsNy@#UG5tQz9A51a|p
zMpI9G`hU%r%AsQqG}K$YGEwh3Y~GljY;=;B=aLzI7rqx=vRrSwJfzxS6Z~8Um*qU<
zSv2)iB>&AeSy}6Q<JADKiJ=AN`<af$hc7y`Z9bY6b|AXKcgM@^<Qh50PM5h6OTX&U
zsxd1>Qp$iG>(02xV6bs7UJmDFuK-cEZJ%|jmiZRg&v6m4ra`v@ik3P(dLpb_RO_Ji
zlZkvewzHBB{JZl^RbefVnHZwCd2Z{i-bGiS$!znI(EXAV`S#esyrF7PPz5;SJFE!e
zAme$jc>al8lnv!KF2l?f`D?!cpJYm&fS;#6_f~hFTd^+{ETRJ;<Y;rmMoAVY*ZkB+
zICgCv?+6y#0dnRuQ>@cM*9qtXANxk8BO_(BRA!;6%n@)IqXt9k=Lkhcbz!Ib+@5x#
zPgz0fKHdhEg7v3%d^8$Wrbxm&(<rH!P;{|Tur2qE4rt81%VF}@g-@#EiEK6~CWB#K
z0kBt|Q03s-*Vc1QETK1qM!6aE`VjE7@7)9&^uN$i*Bt;U6eJ2!KoQafT5~Yx8R)eI
z8}*f6Kl_2lglnr+4>W-S=kt2fCg7Fcz{~w#?7ew7)bHCrOhS@m%{~=T3CUKrX|*KL
zMo~t|I-x9?Eb|s2>l7ixkhLsRF}5)FCCM5Y`;dK^v5aAs&+qcR@9+KG%lCIY&vQKg
z{Ep}N{L!b9ruXN1pVxJr=j(i(uNTyIfdXfSC==F^pmw1j13_KFQ2mST3%YZMpRnU@
zBtV$TKw(++iHL2D{Owb~26~F9F9vuT1Q5X=5(4#Xstl=J;BI+L$z}RVQ0FPIp^Gfj
z;%4l|tP=;B7ZA`A%s}Zoz<~M**-RMgyeNF54!#b=aky&{TPM>5z&4}(#a0NJY{MBr
zC!vKPXck1MI>?QHG_NcJVA7wXXIK$2RYR;jxdj)17^hwK63od}ZHfKc(%ONC)e0<L
zuyu~W>&NJOMxoifg-I+;zNLj15rinIiPK>9^=|aB+R9UZii-H$tfKD%SQwO-a}g+r
z#WvSXdhaW!3fj}EzRfnpI_VPm!C#%CL7%j*!K#P(ToWExyND~JcE)Uonr->^Wy!*p
zu0YL%=iuQoVNwM(ykNlQkn!HOkH~J#c;o^2nbWfTYBIqhKyFo!p<hEQq~zCEa=v$X
zUgW(e>$RO+bCG!G6zD}?z!obf6XA<=GOKN^l22x2zb_fTn#&MIQUUiig#RIHH~Dg-
z2daM#>M1fBK4~ZW2D9?|{B9Y=<!H4lW#r{BF6^VU`}nL~#}d15J>Ai!Te?fSf3az-
z<Y2V-{p(qNgSdBR<5v9Bk1^SRrnWsTnr?*ZQB>6|oRog{r|ojGWs2ML`If!#!w4gP
zfG`JLW4KWse`MlwhiDcflyK#2<7;|X@*Qdxz@Hf~6v4m_{Y(?7ywKI2!d3WO`yRsJ
z^r+RT-$~RqTqT%Om)yB^Eid;zGN9n1QtsaSGuitP3DpXkqE;BfV^kwx$9amOL*7f1
z`943oOgAr&nMQuYoE_P18}meGn&7ZNDP!?UQ&P+*K8@A9brJ!MK6V!|OG0fzZaj)X
z+5F{AY@RT(!89-Z)|g||>_PaplsDZI5ekuwB0ImbexZKqcF2Q#TJs9Ck|Lk2962#^
z(>w@PV(N@MImf>_X#N?|B9_jDN~y1eN0mHZFZ49U|0<aa9aJhJ3Cg0mfQtS%kZ{G&
zOeZmT8<gvczgqrP<6(Ifw_M^WF2xxYNvO`D3d&u$IalfaVjU?TdQ+I-w&ve>XgrFm
zR%1H_Og8i*g!N%gHphBL9^v~66Kq>O)Vc$=4>)m9GD$NGK^4+WAfruJ?Pyu_QsQ(+
z-hnfS%rg0!^FS!vOw|AEucIXcB|1wdricF|8{Ika+BdWrZpFebB+^32oegl|DlH$d
z5e>mp54}A`xfZw~4K%0gfLK0t0f;B9Oe(_gE}K>ZA@cFERlO??&X=A}?dqOlWbt2N
zu8<LtC;h#_#7Co*-BvRCBT{;p?>!?XBWw*f?McFgE&3}`Bpa=+rDzY|AtFqLujU>)
z=E26$2ByjbU{yAfS%^WidaiXe<-w4sCkyAYl+K^&R`{T%G+}gFE{*3_X48>Si<^B1
zd?U8BCgBVr*mz4*dw!v@dQz2MzH#hVXOBG(_W8Z5ovz}hrh$kmK(&mAH9>+X*{fuW
z6s6`jo>Kvt=BL>Y)#@Z7J<j3^2@es(3lPz7JQ@?Jact@2MuYC>(YsbcrAkE?-pf2W
zm(Kqaio}u)#$Ysuv&)QwMT4}Q)5bB4EmKCBpHFrKBg7e3fG@=n|08`??M?b3gx(yR
zEQ3eqfzHD{Og|w<J3HU0RAM8l-zxgOR%x8_;Nc_=MbKMuWAU(h5eI76Q3;(l$^C@q
z!kWukbG=tO+>Zo*f7jMx_MH$wuwNl^TxDjl1f2Cx_4b|DZ55<;X->MI5`UyXfu*7q
zpgIGV2+eaGY)VtytiH$#C9jW4>D8;Ub5gNEKMAP;6JSwq#jdf0JEP1o>-<aaga2YP
z)y>+OCqMh|I<DN4{0vb)mdQ#ONGAAUtd|?}JIMqVj6;Kz3)c#*-@&PXmF&9nogk&F
z^UQKnDO>JAb;DWQ74X%gBL<ZS%(xQ0zcM_CzNd)|oik8}Am8cS(vntS)n|g!9Uri;
zgLm7(r-r?-^t}Ml>|@{hiw!(kH8{;Hy5I{v5Ho=gfFvvaccJ0!E_Q@ARCf-V>emGY
z#l3;;fo|_LaC&jGOzoNg&eZ_{4uLrO2YeNwD+02b1x7S<#R8(tanmEg=vX219W+$~
zFjvCBmIUi9NS7vi_%{!3nRBo8f;)}#OQIR`lm7#$tOR9TXU6V`b@$&ZES{f#vEg$Y
zbUH<?Xhc4NLU&@*;#7&Vn8<C{#vtuwrXt9ug6jNFR*(rxyzZn4_DE9-C(zfb_0KVS
z&CT#Yeg=Q8gYJ^>3DJF!@2~Z%NOXwqT*cIR`WKt^)7s-3#GF8F6`F@=23VZaAnZ9o
zQPy-F^m#Az!Dn*-kz~A@K>NhB1=oiH$Us|t|5F+a;wtE{i*$`V5kn9AM_!x(!*jPF
zVS_*<``e^sPDs&G#*t>C-HWICr$5rh2WH$GFtn427~R{XLd8opWaVf%V#x5bF9G4%
zN{P0aL8I&J+b_ttI#=@K1TIwf<mQo$k^p^4p(h8O;@Z!B`!aN<=+j51frFM-rtb$n
zNBXrLu}Bu^7K<^QvHZ9C{U7v9fLl^1Lab&<P#hlfWhE<ar7$y<vhXxIMfB`JB;-D(
zHSXvEzX(X0u+NtNg`~+r<xG`l#-N2CyAG+}`BnW&smYJYhL(cr9)J{g1-{H~&zQn}
zgm$J|4bgz^>8rw2u-XF59sZq^1vp;vs>~b>0Qtyn1MnDU!DBRZLvz|)+;bwJR}Fhj
zAX~3&9Y^eg76m{`l9>Rapzgr55yO?D!PB4@k~3Csm=C@uTIC1jU_<If*!<#5B>XNr
zD;^Z6brQO!$ZT+9?T^=}H}w2?_fV+(nbVI)TVUtU!HWBKl5r%EJt^7^@uDRg4Ljru
zJu&r1aY=-63}Mx407m6X0i5eA0LD2v(*&?Q)MP!26JH+V(|sFfzdCeS;_^`N%IW|(
zsp??R)LapKkxz>i3j8bXl3I!Gdlyu3qkhNV23&Ooaso~Hmc>9I0d-Wt-}mMB#=qFA
zklW8=H4-!p&08Bd&7L5)hM83?u~B~|W}Yt6(UZNbtWV`ppQ~_`1CLa^><}>az39&g
z;^&$P@XkCK%J8<Q2=@QvbKuY(N)fe_kt*UD890F?03ya=7ufnuSBPvg#<=Z^We6Um
z2KV~kCFi(@mRa)dFU}C8mi~C2u**aZV{0w4EqM+W2ab#>y^E{}dmM`w)E^O<|FsIc
zy<LY;MT0JRDH}5mEk$&ZxF_n1iEy8}UEI6E^P3VJF12)s0@`1h6K1p{%NY{{A!px@
zv^dveUn^ofC4Qe$!oO(t@w1%2Wc;T&e86cJL7e~+W7-KQMqkhWc9*|^=XkWTVxjTi
z)qaB;c_JpFXWh8iaXpmx`!>=!cVMLgei)dd4<nh;xrPCdl~cG~O99G3uh4QJ?Q`v9
zMS!uGhu+vNab+N{@rMBV?@9XnP#d67GlSDX`NRpS`F1j*5}}RNv279PDwZ3X0>hlF
zx-MvTSD^M5{8w4mz*-pH1#0`{e-(UY6)4yS*8LDau}gCfC*hA(B3^%rKVNY6-nPii
zbFk0ArE@F6{dY^<$*RBsMOyef%U}4NNqomf3t#MF<cGP}KV$FD4ftGe>RV0u<8Y7p
z99Ik^U+Va2C*pf-(1ZF4UT1gpbi<EP5LLv`4A-pR!#1No6Z#WoA0;^N7ZKhUYnBb2
zf#J&<8)<m@IU1o+K$9jxSt7mm@`{eg=YPlvj%uL00R&{c$LL}X0GYXNIl2j_tSHcP
zYCs<SzhPyvL8l4Qgeg2~jn$9{5TAw38R|iBj$fpKYSs1*A;ofBtHpgw$wn8f!Jict
z#+SLZBfme@#S(#ji>gfbu`loJ8%P&HzRtg;#~%Sv1rbakX*6IsgLg6Y!%Ky}7aaDp
zL`ETaidRtE><$h6b1eAMyM+xC77sWS<?S-SEZxo}V$j6}x^t_|CWOJ#{!1rCd6_F1
z)ADJ8D#7-?ixNLIk9;*Y<cMEM+5ORi@0{o^^eBSQftEp6MonJ*TB_w+@XP^urteV_
zhca|^VBD~tAO;8fJ|azmBKJnEH2h=3Gs^zl_Sb@h3}JSrZuh;h2CUQ_h7nmaoaCx~
zDR$syNz9)Mn$JYduVA0Pw=_P-sp=48aH6Q7A#b;!qFPg$TnQNp)wu7meL14ntTcAo
z!ZT1r_zUYcjeXK=$eQ$LrU@r7FC$C~DR3^;4$W618oD0a^QJ1y*4*%_z&Th~A^Id#
z_Xs{!<BZ*yB7l&Z-&_Pkt&#38So{bH7q$7(V@MG{di(q5u|~#cb$uTaPQb^%CU`VZ
z+6~I>8*dB-h7IhkNaqJdRL}{iHGfiNdx2>|xQ1Ng233x?B+94x*#yD2ND5ba%W`gm
z&7TKcNB~!hgL<Kbwe4(wux)nZ!S4&Z(AfSOfMNV^nUCjS6>tGp;zF&ATnKtS*}Hz9
zUvoVnsZ-(-7=@K+^zg9lljFlW^Fih}FC-qQx{-$(x$-sbrOXGyd(6<qiR_hn0@19Q
zXgb9;qO9_VP<la?9)3TGzZW#&is^@t=Cnwv1I;I|mMwJB#Nd9l{u=+BO#@;NRCg3`
z5){o|K^zasM;qVN<)dx|q=~7mnYf`{0Tj{^L9YqiOF;}>w2Hh0>}loNl;Hs%hHEQW
zYX;|VF1SEwya)>`U~rRz25K$79v|>ON4gP~TH0Q5kLdvpcx{GW*piGsx$j~@a_U!>
zM;>l&)A*}|TN2V`|0UHEj<djbWMC0a&St5id(sYc1=@!w>krDko7jEv#|*`a6$Lu4
zhOE8dwpWM-y}du+nLb$jL)SwSl)<HApv1KT2~f>Gu(Wwl+awCds<4J;<8f6W6No0P
zx`GseDvAY);OYD|J9KacNF;)~p*940fEhfng@1-1Ujk<#jnc{VGGnq?f=iea;7759
z2^^gpx8|})7ze;Dg))iVp2Y&fH~+8iJ3Yn!(|%m5lCV0)F$EYyLoA*@;|QXQXx~x-
zoR@~T<o#sTwA#vj#%njn#sVS{CP^4ihKuX(k4)+7PHeQ&7W!kB15Z_Zu^=%OjHp=$
z>D>VqJ2>9MJij&6$o8_$q#P%;u(VIGXVb<kj&c?eP`)CoGqHr6j++g>7Ufwk65jZ6
zJ^$HYNc;O2n%#tw<^^hvWUtcj`-GT5cA(SASn=0lzG0nY+^V*AJy@5VEP9B?eQ9)5
z4W#)`I0OFBwZRQ8Ibt<%n^yek)7SSkVY#2ZpRSyEgQ>Nge|xIMRj^6$<kP(RPf_eQ
z%(+#`D-+nz8ey7C?|E`*92jjYZ55CWVSC8^%&j1x!)qO+E#RWAv#Aot0jce>kfkd4
zhcTp&>U_)!U&yzs-)j&zgk==R3BNCncVen^=w>KQIxvC#mOFUSEBa_tKNfVKhtHn0
z+(q#H7aY(`?7K%CMZJtqSHO7U2+$YiVqE^EuZgOh@2wp@R8bbYkt82x8viCN@9=Y6
zKk&To4E2V9H=?gWYLRoI@8#`zt;hXkpG$7A-F@=x^H072*Z*o*`rK!UkFG;AqsEar
zJ41DEI!Q~;;H%Elhn&yp+&6o|GxuHs*x-C1jHf^%Xs&t7gWBjMl?Z>mDrWN9wVQH3
zGQVA_P(;*tf!ApEftE?y40Kd<eoCrVef#*-cja9e*TUKys9u)hPRhxUE~#{kz|Sc^
z+I^BdY%F8E&{`|Ply*Vz>#7#~=oxd82pBH#ut9o1G{YS!@xt*tCrd2#E-Ec1iZu#^
zOoGA4HVmNM@yHg=6FZ6s6#t{47=gPVx|eeHZ^{-)zu?r5-|kFz(v&&a{dhf9xnD;~
zzeiPAl%gym{A^drvGm=zPS{X8izh{)x#Jlm2?qC-62s*Y>(!bUjzs+VY8=o=ZUG+w
z1NWT<ai-2J61Z?y8G1X6N>h~M{oICZW=w-1r7v?w_6b;K0h!cF6Zm5BCsd~GWIL}u
zprVT7W~uW(d1tR$9>}7k<sS)B1gT<(-GoX>3WjE=-l@A0x7hY=t%(iN)8}L+)u=K2
z1|~y08!AUO`f{DB`z`KF&aje=6gS$bpgaolpM!s8enVjp$$E~>o7Y3uG@}!)82$+V
zDXHs;ct{+qvvd1w-T3-K)d_=g4)Vu5HXIMLnBaJm4u&!M7{}eU>pmHvf7x~h=6K%_
zR@LKHNZxb-G4zQRLRa-Sr#&F?0_}0~vm%EVt!s;mQ_1pm^F{d)8<^jii?qlIED@Sm
zF7}-06l%*MBBmV@eolVhV`sk=F2?Z&K{zP-dDOrzO)gpO*S$wKtiZFt65aQnhi%%K
zbFvvrx<^}!wf<BbUl{_Fxy)>aZ)G_J^Wv(uR@9)U-k171e?Q0fhc{a^HdBS!8ARF`
z^(n2IY}P1bJ%)0;_qj6gtyc2A8&%NDJXK>yirVB=ZFj1=1U4ed5TmlWq%hM@$TAs{
z47f5<xSSqC;-YuAX1rx_eU_y;PMmQI`GJ<|y)nu1Rr>12pqztsssus~@F+;S>b7(|
zHq`rD%3S5>f~tuy6!n#V0Rak(&t(jud=qB!-Ky)Pq{ICT<Z0KaB|ec@O8hk=r8k?E
z1H{h_(j_KRMqiq%IIjnuxWcA51*4rf7_xy8MR#cRTgfLr&$kgjFG@8p!9`&v%uNFV
zz*WS@s)KX+`%R1y{pvP}xxpGE+TRu9ot2eKw(O$0CVoYXTdo`xHfOgvPrv8vlRU8a
zXL3kMzaKf&R#MTU=;x)p+_ayK<0B3MN=s0fgdo^ia#lcJyQy)MgCcC6lfh9H>l}CN
z_WrEcD-=ZcB4WOG_V~Iv=lz=|`2LcieBN2X>7Ub@p-zv)EiJ?ju|&4?=ntVf1r6l=
zhoFW<?CIA@)pT&q?V66WyVxezY`j&u-h{??pM(8ir18=U$z|rTBVIV+S-~cGvp0N=
z#~NiS*tkSI?@-*ERb<|pq+mr|4qLVh)R*lR=-sVi#d?j$>~}^Vx_#fothDB~*e=p3
z8e={t5o93^RwvW!Lql>0Ydk9@y>DV`!EB%9;y+-8AzB&b`}WV(wX4uR8u~7KMNMC^
z)NJm@J1Lye=yp~IR15<!hy#q1kGGymAfzBi$Bg5S{s^>t+IB|f*Fv8;J0cO-rn`rw
z9jBDy$7jf!VcvSKM*_no%}SmJNp}o5ox$~;IL5IW%;Mo74>hIjrG=1y#OXw#<>gQ#
zVYx)A$AyAtY3#OYh<tFA|A*i;nw>QS<Mf9}{KR%rT1E~^(T4D)ynh<y=DI#-b5?kg
z^N#YKZOfHE!tARshTuwrHX#IVUoGEa-Ls+|`KvrBx_VJbI8a<H=yssS#vfsG?(d-S
z<Q}92mmoN#+DPm7m~`B{3@w>B*<R>rZ7t9|CKc=Bg=?%nNOh+r<+~6A-jEeurCWcT
z4C{z9d3VCiR!h(_P;HcuJ%5bjC&3XH+Rc!qsZOZu%`%Z9KEn#D#lz6&($!!d$ZjVP
zF&ZErDc0wpD8#5ELU|@eIBpC`@TQEZJ5zD+#J3p9(%`7^jddKr?1M~!qeEga<P>ma
z#fDt;CDTg!WM10#cO&!Eh308C2ZG&f`mWEq-s7$ZNroZb{QFU%&3YG&YcTD}Jlwl|
zF&7lkAkDF0egf^!G<BNiC&U7@|1}Cz-<ZT%w0LhQn8EI0!bDCXm2^mD4c6H)qn%B@
z9JPZ2CoK=8!<WZ75WOOgkTL?@y61-QdMA3AU!sOD6g>qD;x$YG4Lb>M3=z0Ti1c=%
zlzk6U47Jz}5z_#%Qw%YLv`Cni<AvUQR&5N~iQLB~-so@2hI$!~g%A})c+fhG|FgCO
z#@^QOqWj+5RT2C-ndfAcaraQT<5(CG3ei=0ltAsD)t#fz+rw4YbY!aDzkk-@D-GV&
z^cX`JG-7ah{#F#}mPT0U8|$31)9!sGZDIy-@oetb>OO0;ccATtXo{qkdTI1sr4lnC
zC$0XkBF?t>G<CVBefv(~<~;J|lVz|)m;cO*n%fxmKBj~M3KgUmNx#hMBK}Es?wYre
ztHdmSvL+h+gbq{`ZI0&xLFP%vrK1HG^JwAjB>_@3ZW??Z7HeHH3sPb7geU8rVKw62
z4tERGK>*!8%g<N&k$d39eLgny4Kz^Byq0=!0`R9NezT;V29j&4vqwsZ@qGeSy`!P$
zvKGZ_JI_cfwt{p_+r3dG2wRens+>O#9k6!R{-Ko9adO-1qm=#Nlf@<!2NzI7NZ$pB
zO>Qf}GWwK*OQw@UOqIIi<qSK`4A~}6WNy2hn_Ax~AgVC`AyD54M9~8p(h|4>mSx^q
zu(BVG^ZnXt`i~uO@ZsqKc3j~sNMd)UI)VxTs*H_-JH4#xs!}XKib9Vr(=SKAjm>$c
zB5sbP-T<)#lEI*JC|IKnFPT651BELr&ndRe<BEL7V{*cj2M0SMqwhcMRDc`($FBYR
z*XKCV22dT+|M+RzWF1<TuER_%sA-s4((!Tk8N%Lsh8xgc1C~Vu4u4ttdDIaAajEtu
z(9PneV=mVwj8LK}U$~LL{%FE%(V6~u0Sv}-6@O}bkbSXeb~~rnC%0WCzM}6yfz+<^
z;u_oWbODg~ULZ^qRXLNSxn}C$pls(#-wf9s85RHW=Uq^it5~y|4d*03?MO&<|3^g{
zvW3Ua)>^d%5^9|Y;MrB(K!t->?STQK*`_2<5_ZWtEb@D-SbV&g=WnLdj)GoL57bsg
zL$Dg!K_Qk_qXfQU2nw+oCoFX@0l*-)w_E54bh{=T9~QXY$%mNtaDF<7P?}B9&k9kD
zNeiutq*{Z&gPg?{!vUrzHHZpwy+9%o{T^`4>^Ayl4D8-qx-Ww>Jd7bC#JWmo9$O&-
z>|z7gg7}SLU2C6!n4z)&QL+M*GiQ|`O+txby|i6zSt90vd)~YDYpIrYp|I~Y*Jz+y
zwmYRuUyxx=9*A;fm_7B>jDA*T!t1#IoUF9oEXT?QJ#!B6ia@@-GB{uf=ztjvPj{C`
z<mQ_E(m&k8<>Jbc()#2&JtTGZiR!(XM*c8;Nm};Aw^Z&X?aM{Q**~+?4_*)B7<SA^
z=WlTTZAe`uFynA5Uar>~=STVbrUg6qmdcD?WuM0FPU9Zg#PvubXo`r%ji>fAZH34K
zl`ksgd$LBb=ND`=D3+`}g!F~v>ZZ;YRkw)JC0SpG=CAbbR`^Wh7Wo6N>pB5{j{(j{
zG)(%*L(D(L?-F2F*j6AXfQu$1RNfy=A?F$;PfwFhzgduo5R0y0kzWOuvoyBWKKd)s
zGAY3gELqx@kItJuL=Pii?deu`{&2HL66_S)GZNtz8bAO?^rcW-+r2j_0aL6KeXgw!
z*j!7aAmW88jLY^4m_p(5*CTZXZ+A9)1KPl{7y3><`eLLcyXUpyuG8uJ_bi-)y@beS
z2*E*6wR|DpXRa@}awo}I{~mYqI>Ajcg)ZLS`j!FJNtW>QFC;(vh@9ljPHUdldH2)k
ztie7VKg-#lDf{~SMJe}8W`4dBnF3aZg>3#H$>6jlcXWmm9_d|oJ@oufabIE4Q2ULh
z>skqRv#Naqq!r+ZluW3FMOXnJWOZW=K8@c69svUuHwtoN#*<kI_$97}CJCkGxKCCs
zez`W6x;ouYFIyUj9g1Kq(*P*5+nJ1cwUM;qFF+Fa*W0(;qft0*A9MQJvvl)QnV-}5
z>`a!`*(lHm6lA^3-bz}){<oEX7N6Z1Q~q%8f;BDZtvD<i9cwuaobqV03}DIp^aCoE
z-lBScc+SVzD-tP`b0v-Y2;oX0>6CSoxNDQ}arb-sOW2R@oQY^ZsO~!ap<ic|uP!9O
zfH}D7nq}Ahv9U5ZTV8hW4c>6n)_MVvp$@9Jv*h4Og=h$1jIWO4P%g`^vUl+E;!D%l
zq&>7`Gq!!7BJ!|%cajU4J$)rZmni+=&p(WQ;!HLXMus?WB(SkZZ2+h8C4l$uo}rmh
zwn?T}Ts5lnic0T{I$J*!EGr09=|D)(B3XOhkQ{1~))B9KFWjBke0rbw9U(IXJPXSx
z)+}Bn5QauQ;2~b}8&*@e(g*Lb#)PasP&Q-E1D5tT5)dp9Lw6Gdmo-K_oHqh|ZoHGz
zr0jY&6A~@#XO})BvbhO_tHQMxJ-7zYhvwo8M|E_ZZvjhGBBHp3Aq0e;7Pq9do>^%x
zB+nlJAcob4f=<a8PZ`7Zb29GW$vc?QZUjen&S79+AN~jY=hPv;hiX*RnMVkRsVU2$
zbx`zo@-u{aS=Vx#;TzrXc=RXm_KrhSeT(n~su`10@h$BS_@MkQw1)rHUkrpN$vW~Z
zewSE%Mg42lImA`g4~Xp~NNmSKPoeaiH&(_Y91?%Vot<K59^>f$2XM60J*op|UB_Jj
zIvt0vYvQuOvWJhYn(7~d2AE~6z>S45ygNe(YY1E^<t~!im(Z@-ca|=f#U|A?2&JV>
zR|`>fnQy#S1{E4AmlJd6V_lG+%%qlAg;wKt+kpY+j6M7I$Q~ic5XL_uz9{ETV8v`0
z7UWQ@(sJadvxyc#BfOH%geqaLtNYEaFhf5PV^WX>cRK9@+@(iS>&5;A+3$<V%{#OK
zYB(CJ6)NM7BG^kmZn6jBYS7+c{e@}?2V&M=Y|kZke-u}U!EYN*5NIbY>%J{1{po-3
ztE7Fr!W#8>)Lx<Yi3oCX0HEm;xs6K(D|g6MUulN7%iB)XL*48f0xz2Uy8(!{+kjF(
z%+BBF%9&xenf&u{-sVR7UI-7=SwiVx(0GL*OI|RGGwM~zbaKPr4Hq(&+$Vfn_D}1|
zd!XF-1+kcfL-%2YU6mLWg;i>fCptg6Jbqc!W&E3m-t?+&y(3@@=&$Ulo9Ga_g|DEL
zMLg=Ccp2uTPJX!m{LlU4C|s9W7-3(P3K@4k%z#`J>HJ}ik|%S!;Jnsr{WV(;V4vbl
zvu@D#3D(ugb@IKsYryn#h;7+RsikvgME=3!G-AGT*Ovz@VVS*Ok<sX%x-QR771_+6
z;*#>qyJzd&fyUf81>VLX@(%g{-&F%g4N>hWW7M%w<*D8kfvWs(XXF0FpZ_Cmw9w}P
zN*ec2(8amy9%hAq%iZj&L4Dy_w)x?we1-Yw?_JdpH4FsLW5zKW<+)}-snSH@^Ubs6
zJ=7+tpS~p#2dDKnDE$>+NsTD(L-GNNB$;UHJ=U^f<XxW^%Z9@2V>=GXLG81rX3P4q
z`_<2`c=O!KuHM*w_wv$uz)Pt&bBHAE{x3jAU6!O?R)%n;z$3ixe!X-(@uBgoFn_#(
z-xI}k!#aq3srtg0c?jeB<bvckz1lX>6n8oIm~qh+gg3gA@CDPrKUk`&5>n!>b+dW@
zN2t^qTSrzsC*ubmY=apy$$ZBW&tAsw)r$5G5kegB33%@%knd}^n?BIr_wU<Jl67s3
zWXKGW!~Me0j;b({^<MzCUJl3$qVHi^%xFh#UQmijxL5v2Pm)RG*$uyh3e>Il$Mq^L
zdlAL_wTfngWfS)?L5nNc>!H6!7+c!0KkfiJB-_TP!jDj^{sOB)YgNM`vpll1EcVC~
z=_@Qd5bHJm1^}-OAwl%L_o3N-+-w8&I)L1rX?B0HJ!E285mvNL8?b&FG|zke8%mcM
zcReuAJ(ZeA(f1dvrF9%?7_<lD?`k9f@nt5&HmT@v_u}C%#@6{S%`3p(&&3HaOW;#O
zy5JSqgjQFUk428{e?a>}-n>J`H!ivE&p7zD2X`Crzz&XUF%sbaBvIw?{U;JtS`BCn
z1nJRy>PAoY;L91epQ`=-S^V<s)~lr+!X+goAhmib_>II49DWX#MELJ=d5Cx!NpejY
z$yfz;=}TpGqFLa`57+BSAEJ%gDXna8I?(yxta}2VGThNeeIVJ%6p=>383L6*`An#<
zC>p7K$DbmfPPAnvhzp<APO@v=IU@Sdp#HTeyg4xz=E_i`C^wsW^!J^7ww-m)M9MC@
zAl?7B7$bic0)(1J`oeR`cM=<{(7A;Ov9+mtL)Dd`EvAh!C(4#U^J;e(IkB<OZOxi2
zQ;>F`PX+cPf#{{;lv~qy{S7eQMJIx9F{EII*a@`A4*^NmZ}_r?-uXFZv-!Ky4bi>&
zTEI*lrD5uOS1w95(yYfd)Ivnzr8)PH!Ih{rL^h!7!hh$`C+@!_2m*&I-T?IWp-TIb
zOhK#0^ZOcoKg-TVY>W*85HXuvHh%e6+81w^Q3Z<B8K`7lELLnvT0%qik8FQU<?X%c
z?15+4%-Py2jP&?LcG5#mZDyJ`_9X<{FcS!Pg9)tyVdVU^Jj&iY<o2V7XSff&=C@*!
zfN%*^5AF50BR6hyN0uO6?XJCjk|+?WTIO)UaG#C@8{-}`2;2i1iHNUD1vP>ck+}IZ
z-tu=pLS^ag3U-OMY+IX8o6x~6;2(5hrpa`6YL9*{A8+>hs5LWZBS`^Nb6k!~GXBRX
z59L+&;*?Hi*e!#18G2vSpV~Y}g0(ZCG~9bQNzx4ZobPR79R6Tw?wqvYk~H|Zs0jL%
z>h2n0=Np;Bhi?t2yfBQM6OayAL+o0BbFt)6Ep^%hO{oZzZ*d0BKgz7K$B)@NK3F>@
z``QE~epcumko+i1jG;lm|EjvShAO)m#=7(Ke2M$LX3OMlze(=eRBlpBU4ZpO@p+l}
z4>m)Em4;3)MA-_&p9P+2KE^>ouy|gOOyK)jzleLjrP^wBrx=yNsutTrn2MWC@sDm4
za47COV7S|I#wQRg(Cclj2A~D(RCBbq3gMARyiL4c?_u~OR~C;BvuY+28>BGPMtPq)
znIvw&pyq=EN#L*g#bgBOXYM3&kIYKFzHFsme409`-{ENT5$<zm>h$qgx)@;b_#4ts
z{l@Ukml`&r1)p=Z$|WL+yPhAfYn?fR8#WV1$)2Z$WSC{5;*7zZPerpFzUq}x7+HgC
z*$l>^urC;Q^V{=KDtECB4jxt}$hheXg+sR2YpLn`i%r0;t!40*e>sgXDYP1~dv)`g
z56B9uZFuM=P@QeY07Dg;_=_!6;#U9aXuX~GNVjSm7fQ!YHZEr4#m>2DNclSQaDb%w
zvP*U0XjUz_Ba1-yf+29Fxt^vK!*J2CzPBb570rQm1CYUURwrRs)lK?UhUM2TuSjDY
z->=pb^QuEHBTm=0F8;1PBllvI@PQCI%Keq*Jg&!GPlOo0&YDk&2EUufBe={enmj?S
z4hPLj4h9E#-PbU<#3bPCu8@R<hg)ltRWWSC0E*&Ronh@ApdWEbW^@5oLiNWA6DR*d
zn|`rXrtw<&BHOMSaVF>%KZc30+D4UA%)1BZCI(ftH9ZaiA2rp2_V4dG?Q6Mioaaof
z#79vrODQ8(*Fw)eNV+Kd5!l;ZM+h^}V9{<ePSFaVYvxbj-MQyirQR)_Dh=z~vh_~i
z2XO<1mzw?aR{ny(!@{E5ZIC@y(od^=&q(c|eFH}+o-AI~@eZxaseq(cX7w}o&~f49
zm#%Ja9zB^l;EB1i+JR*4!H_!7Vohk_nPT4zpCXl4epujw;p5@HsWl&k(~eP?AtPaw
zhMb``DQvK&6UW++MW4SN@K&`*p35vB&^0Z(|Hx1xQP%Ux8L1O9WD)u#U4-?s4R{ap
z_UNi<pFrMZ_xbePwZgQ`T#~!IR+vysShOa$P@T&x$jxu4sn)UHt`{Gd$?9FG;o`>o
zHT{PX2^x%y)*%^+C#-0&MqUoFtfVC?^IVQS*LWC667Dx>MAi=w9$^lSx+FhK6)npd
zy>w&JG}-K=&ePCQ_CryYD~U}j!N9L_V*Fd@K2-5F6Q<_G6CREq;e@J}F??tg3UL0j
zA3Xnbl_f&S>s{pe@pN9bq$sm1UYLCe_65ff4rWQ<NJ;}CMY=Fpn9Ld7*&OS!Gp*;2
zhH6IH!3ypE3ytg-k_rm<AV<t#A}IkodDs7rhIPsv`|!dEi%S8smS^;naQ;jH#yeQU
zh$$1y+)$MbarNJY%OR}^x>KoHJLzf*mUeW7k@^$d@iuXQ|Awb8Ao$T8AP>5FpP%YW
z`)D_50v9j~&?qnyd$AV$O=_KrD|ceUrNY4&`6^Q86qt)bGLf^D<NDhE3T3oGa_4Pw
z$gg&*^XLq#NAut+?9^HWXU5SrFeJ(%v<|FigV>gN4V4qV6ZBY8V6_i3^gv<Kysh=>
z1I4tdG1-gc9W~`9OU8AT+pT`Mw17VcL>6j+Z#!(zw*koB8a~hXVjbYyCBEMN_4JMN
zK{scA?*e<ycmgBFV5dd&`^Mu3!Wbr5_uLHYZirvg%7g^z5v)|j7Bn2)21sU*v$Sd<
z1M#8d=yIpB-i^K!_nWH&<91IH1NGUMY0&tW*g1Xa(jN3NXQhDWZo8nzcQMKK8&1ET
z?182FLE(h)FEFl?UI~epFqXemsE@2W9+5vj_8Nv#Q8O4&9~Ix9Kgr^!{3*iP$GzIP
z^C+IOVXm0c4NlKjEEQxRF|pB9#Fm@$5I*s#Oocl8DsJg1-3e7j{YsJ2lK22Ua?&JF
z<}w@PXD6o^Kb~mG_$F_suy~O41M|L{;RURdwEB)hbsfF2!2~UViRDL2^86=+0}YFi
zbDe>3@)dX}eD6x;FaIl&&DvRxe<D`+oH3~HtDcC*&P3OuH+Fof@m0?fUgsH2gk1P=
zY?e)$lkM(5ht6;_KGZF8*60D|M_oIlI=`Enf|g5B`LOi6;B^hAfMq8Pud;X%yen#6
zkYu6ns0kl>P_yCF-Pf_Fns>E6@j12#j1DBJmLMdvX<<}v8q#&V<6`}1+K2k~KVRRC
zh)IZ1Hctu9q-|`cg30%ucA)HVw2b06$CGitq*J(5wR2{7!-4&a`=<7-<3ftSTKW{`
zaRq~2>zH<ZX-WlC7*Ix+-jOl-j^^!C--A!=7hJYa^mN=~;2G%asLs9xN*c~EkT(M2
zK>3bdVGXG6f&lkkiWUx%s#(H%D)JZPvO-gkiKbuo2M_wPpN>emma=$iP)z7dqy1-J
zs07(Vu)_{2m@(fl6mYYE!YiMbj9(osiG_If9N3>06sR8z7K~%cjCLFgAakwe&cW&J
znJXn~3)@1Gvr#TqhT$9sqYXV!a|rM#d}vuz<ai6F1Qc*i2L+ugr;}7q+<ju`V8A^9
zBt3oY_h;YXx^e3K3PPOKJw`TJ2}MZa6$?F$qpOaScvgS(-?4U<TD{-3-&l7#P&aM@
z116^GwL>EXX+sSIRt8xSBE(kpxx*YQPq@_P?uE6$XeVncx|hAiLRm1UqTJ|s4P^5P
zTfJ9NIQa@Bj&?*VMs(qTzd@JPE|g=*I}sTcei06C!|~`eA#za$=VqgH0K-h_TS2&c
zZty-(rb(STi*R9M4E@p<rqM~3f&OL`5Wu?0P-@lK$S%IVg7E|CRLSBNk#&BCz$WWj
z@+7Oa7n*ejmK8FPJ%ZogmZk_aZisbu<o>7-|GdT5M+{4XPO(vO@=P!xIsU#<Ymq?F
zG&}fvVa??<K85&P<^6+Ka_dxKcIth9aCdVaXGUVF;pD66<K!QE{gEG<ue6M=3jD=(
zvUo^bzJ>_eGVcHffwtBR70)ypZ3KuSvHYmSt8sR6-SPH^f^K(hNbToCV;LaZ3i?5F
zS-bo-m?RVOG}lEn&v?0H!#UxGcDXrNJcK!`8}Zo;P;fMjRqz4F=Cl{2mp;xtc$i?(
zBJ-ebSdoMff$9Xk=e)%qk-<7C^TNhMw}}>UtZgZC?l1m3czv7ysqXcpDY8RS9xBY7
zm!9Wt>vv0G_%#aG&X0j)7zmIT7+@Q1jwGApe+v^J{;uaSx;4K0GA!(TM?mky7!aL(
z)qgv{oh14epfdeL^9G8>1~`=$KpHYeyGfSL=sVhL7>?gK+%cLbdMmueAo>a;Cr~#G
zcnZ;lU0~en&-LX{&FuTD3#{|L^iCm?dz3TU=EZ|;agQx$U615l-kZ6Xi%t6DzcE+;
zCnYem$T_mmDLM50(igX-sj`zg+&3x!XuO5$ho&BV18nF7fZOJ8KceaU#U@t<9M$#%
zjPINXS#O3R5Y}L6*)t5)TYKaf>;8DOdA8*WQ;3`vm?lc@Ps%cwzVx-sh$AiQ#0zXs
zphiAr28?IXb%$qR2h>->we?eFovq&6*kr3a-BD^b9&YsFJU2~vkbkgQhOb*UV1iyP
z)F|OvQ<yz+*x1!AbAJ>+kgJ%Pr0Ow?{jbyIXrD+6*FwDA`Nt-gh*xnXMg|u?KNj56
zHrG!uC;)UYEN@PC<A9Y#wtL=E8~)DSM4y~R3;t$^CJPVMH4e}lYgiE|3ne<LPy>wA
zyXF0knV(W@d%SaIk7^-bE2Cwd6nB*!Rgym%R+i271i~YLer+?nL^ST{_zL%&`$Wfn
z^1GvS9dS36+vP)>9u=ME5lRdjngGcxD4`O@IC_AIx&F(ReuEg7VDy{o#^{txYChG6
z>H=cM0SHs^<_#gsy2>K_jRm8N5f7F~PrD~CoO!TtlLbcR&8lJ4JW4k9l?uM1q^0cT
zg(H|VwQRnA=G@r-NneEhgT7d0LcPp<_ZYHQ3Dc?({@nMdk5t(wL+em5zkB;>Z}(-t
zmyy#3?K`*gkHr7RRownBT*Y6jGf`j_F#6Jzuqcg=XJ#V9m*sQ)G$t_E*R6kE&CMg1
z;Xz}3wm|tQzd?U*X*n`KK(NDA&|!XDs0~b1<9CkAEfV<O{J*6!4yaIo$iH<uFj%i9
zxl7+i#rmDy`1axnb>aQtJ^6F2)*YzM9y`eD1thqOcH`EHGkvzD70HuJg5L55!%iIi
z4&iQxQCa4#^S9r|?2+AV3FKCUFH#kL=>t2z8j4ofqI@}7Ghfk6>do<|4lOH+X|?au
zkee=F8Kx5f3e8qZKi?EN_iyJUd3>|SS4Qc&>D8t4*N^?~W$jfW5e5^~$<&WRC`?^Q
zq|$o~e(BwY{bNDHYkK)B(ZIHdk8I&8-0`fksU{mk+ps<e0K1w6j!8#`z+`RB>FUZp
zUA}yi3FNm>W!Gi+b64^8+n_qZ?E_q2l^Fd#fO870!(=~t7pK6l`gU=rp@|6cO=(Z+
zt8O=4*o2=vxO88R)Fs=_bbf1*)!mg@s@7o-)oDWy^H+$xBu-qo|3ya)LXdj6sE_l3
zyy~aj0@X46B2=?FEKQaWq$^oyHE*NdFGo3BcDO0)tdZP<!Czsr#F{NK0XV$wp8f1B
zu|LT*%lY*_L!`4<jK((mIf*BYpD(>hC%($K<;#wE3z4ne`PyVM$um|FqXrR|Dnp&)
zjc?o95qXxdbBO$ndf?x#GJ%L0-CIQ_M0#!1^cHdR1<HkgJFO}O>T3yR4B>IL5A_m7
zGds_D`<@jQn|I*|)+`xuOiF5bAN!$!dBa5nIPE2XQa~eeWv|JE7(?fzX!NsKrJ48l
zl-b3Is{K{)W-R%#q)Fiw3vyX>e{(H%q$RJpr<J?HoYPNhai>-rLJT5F9;oiB8<qI+
z;@FOiaB=heNrm0e*JB(9S-gM(B>xv1z_JErHm=ZsY$OtJ60Osj@IrJIs~=I!hgmWm
zw}$w}=1)6+X;4W%iaGo9JS(1rEe4RV$|Yth>)0Jyb7!O4Scl80&!tbEgR$cwJQ>3f
z9-tminH&hY*~}6y<)4&3qHD4jlf25_&7EOmRvj<`07x|$FIo{b0BAWclLx|<mDTlp
z+ZYw;X{UMEg06QH3eg`J$H6}7Kq=nnbrot!mL84t%@ck8_<E7Z?|q{GXaNN9(?K6n
zA1#9JRx{2Un~Tm^_+gTI(PT*ID(fp{&9ccx4t#<fEgm-w=*S^iEsofchUduRag+y5
zY`%ylEbU}Vuov_6stVhzk5$vc;XCp*24%MuXs-^f<J!kwAp~?iL5+1e&aywJl5+4B
zo=3A~pFo1Fo``_&{#3;kaFK(2UC_IH<p>ZEI^XM9+NvTFUGnMoGP~Zw67Y^%pbF4#
z)6aBBF9&C7W)>c5?b|qAd^LLD*b~jQb26sj1@$r`rkGfNQI-<r{)d)HQ+xJRYwXvH
zZRY$Ahbu%jMF3A=9P{7QD+6Z0Uvvd&og3&*eg>HoUCE8rGe*8z$WM)=;Xk2BYmI8Z
z?;%dQbY+~8JfM<`_j6o~qn{c?e+TN$j^yf?Z24TNGBXc{xFTyQFKRl!$Ih`tX#tju
zpfZ6P5LAA|^shmvP6*7$Ot$Rw9)M`@E!F_`vt}IK5|nZMD;pIhZ70VM?8clqB($_*
zHTGp?(If}m=7BVOP+;;&+W3XL8eEx6517*u8DJ@bKB2q7;5SSwggaQHxx0TBtMk}p
zsN5ZqjSjc`0hJzHG|fTx$2xW9xvb7PUW2P7VpDKs)olJSIve^oRup}L)sw6h#xNx%
zX-LZ>oo!5>g|o@I*0suP(8d2*p#=z0pa16y?ZAuT|ESRZnbiON^S>!d{JjqUzh4KA
zHQ0ZjPi}z5qu{F^fPTP?Tm>A2-wU%?niBU*?@m#a%EVh=c}zflm3GL)o5!xB4+9(D
zDL9>(MK(d>fF;4tFq=RS)$UjH`|EvjwpdDU4MMgXAGy$ctCllRe7uW$@&4b;n7>&<
zf5*VzG4OW`{2c>-$H3n)@OKRSj~EcvSZGsc@?Wov47V+fy|w)*7SOtk7^<{l`aN{7
z)n9CpD!KxbiaLJMn#`r-<2_>acIj*}@0!phsAAS-S4UPAsMfpqcIFuCKzZEfZw^X8
zFb2A(@Z;rc{sl=oiU8T)J3s{hNya3Ai)+S#=%Evs>_E~{P!|P2Mq^JPipC~WWH!Kp
zOTQb-2G!%4Rq$C<5AaJZhtrbb6IQHI*t)JWv^fl8m<9k%0@&?9KLdaE|7Ra89ALq|
zBu?gLC?bX$SYgpBEIPnM!!ubTxE+1ci_L&M5m5IRTjRjk`t^-F+$D~+FmfT}&c7|A
zPM|OIv*a-}3qn~KlV1YC+~fv*0tD1;_{&!2|KqbFBWO*Z&_n`<F0VIf<p7G*=$uf%
zkpdkaJEQQTj7Tt5oaUz;hcS{9UqK7$V3AEPk-MU-FWz*;o@ce&z<<>!`0pj>zS?NJ
zw4MFugZkQukkKInYmOXG8=MU{ofoj^fJz*zs`oFp^H!VR*ZAuHp=7Su1SIJI`_I|{
z==zHwXjpSzPqbmYon1Ls^3B>a%j~l1>_cEws(L~t7%iP>MR6;3&l*fWdZ}o4tld{-
zkCr*aE`JA5K>tKJUQSx7kw$qS4t(4~59TG5@-83b*4X+&S)fKzI+v6tx%+i*=e0lU
zI84_}<#9i<Z+%bF_*%raEET{yL6!^yTK5C4GSYE7vP0N0!Hd*J6{YVo^TG^5G@BFP
zO73M+!fuW+lSE?T(%yyJw@l;TWP&KO*4MuCS&JASldjgg_@;%CP%}mbiaVX-HCTIY
zYV{}~Z|oRzvdLFA3Tr=TiSMMzuGDs1;W5ob1vO6<3G@SPj#5i|HWXA3Dh*X+q_v7P
zO3${eRdC)yt++-9tK!5|T!)W$E^&?{FuuEealzh00hh&nH1f3z8!s};*Wz()Dn!7`
z(3!-(+`d2aN>}n(jdIT+iv(*U^TI}fyf(X7En8p9HUt+-4fGBDDI2&~3^hN^)I6)!
z-PEPmamVIkRmA6g*VWmX>ep?I=_){q!*b~;Int2wEUittuREls_-34|mKVH}g0oKN
zPlmhUh{ZR6E&*(I^l-1bk$0OrPGaywvYptugJ)a0Y9zoQ=^)sT;nY-MD$xOA*|Hqy
zcSBPlG&5FxCt#5;tQJ#!Xa@Rc{jIvHB-|~{B;K|kB#H?hl-oxEUFLG2;z6}J3BEBP
z7aI43D8f*~Vin+EUZP(Et`gH=&nkX~lP}GFeL7kG1?ZgE#IdAOVc^_=#uC%_>A-19
z)RGq#3ljlRX@xR*<1A~%o|>KBv6HS);r#;iDRReH*aN%ZqZ!Eeev`YZW<VD|NOnQc
zk*-5h#_}x95RJyb%qTMtICMaQa0fj-ykGahaNV;qyDLkv;&&$H#@%lC0B5;wnl%Me
z4->D_6D^*Qb{n!N@!r=i9M~r6r|ozGjVD^JdD2x_J)?5WB$k{r>rnS{uC{wJfhgfp
zAu!zZ=C#4o_dd8;%*0HL`dGLO;361j{U{!E5uH2zRVvVEH?B_#Qy-|Uy)$|IIQmaB
zn52?V{lA<9R@+HH!;rfP#?Ly8Op^(#P0%_px5>Bn$7TWVWX-MtZ~*dur8qK$mO%T3
zBL%RDajyu@xGm*TRnVH5!F>eG1e&}Hx)(I4tDcOWFjXp2&MBj-8FWOy+&2~A#8LJ?
zUgWKQp!|UmYLIuJJ<YvJ7-4mJ0~}BBBlt!ILvVWaAvXWaD}u%$QvK5T$sw<}FDR)9
zHU~fz;H-LHB;XATeQJHl><4-NBn@U7RFg{s=K`l3?INX~C4E%GqD!L;bR<5i#0ls}
z${yeWticBW%d@0RM`w+U&W>vKcqfr=bpKW=2$!p=3vWh_LsB%MMrg0@gXiGv6qwtT
z(EOHbQhMjn#G9}~LE6+QbM6gJsYrE~*}Cq~KlW#L%e<8EMtea!cf(__>mwv|LC@76
zk*6_Qk<&!DR*~9n{PaMZJ-|huup>UKvt7R|{4ampeDvmFbn5RT$J*7T*tk8f{o7Ey
zXw_Zv%!EUT^F*P;jKO__mG+8+pBJTKQmhr<3`MX3!xUzgE+@%7!f}F%DQp5`pbwvC
zS)V={va85QIai#DyM_Zorm|C=PwA%jI}}S|+aBoXd}(`lJzS27q&ENS5%dYLi!Nd(
z4C^5wR4K9BJ-6p^R;9@^*NpEU&vd%Qj|`-syGMEY(6atQ6ex84C4U|u;*^q)&)x;5
zg+L$-16ZPW-9>0AO;-AGDuG_K1nRt^k5|n@OHVeV@AQEM3C$^s4RKr}B8)&b!9-i}
zN0&%nSuqXO@w0Aqd>eK%#B%p06*rXE{JT#k{b8W=zw0Ei|Er%7DX@}FhKOM!#Lm(H
z%@IvBZ*iY%$+zIW_f4L)_Sn_zdT%+W2kFDf*T=SW2r>YAUohuhNL)Nlzv}|w6N9`9
z-EVeHbWMFOH?lb6aGHBqY(|d%%C-kVn=l>;6nhOB_kpZY7r<^*5Fbi*-W@yNaPw%t
z^SQQY@h|J*xHc?k27`m(?uP|pd=QGm%o|@GE8<XNd#ztqR{@Xy;(}e}`vzz#DnMy9
znPsbg2x_I^z;gh9D)Dsw*#W9Ojo?ig7+;Q@lpNaKtE|!>S{l6>(yYVMn2%pWFoXy|
zqOu;bOjyTd|G&o2b`@oRx;xu{dK58o(cbyMh}@<iFiOhC>?&6YA(R{ps!>B~Pp3{O
znbB{rKjZP#MbbYQ%TA%Ad+}&+3D()xsG=H7Y0+6U{g!)`pULa}ZfkdkuC&Izf3X($
ze11o(Cg>QDb-_(So0H%U2Tpj`{;6Af?BBghwiKwV%woLzAk4H&*^9tGv=IH#^iO%3
zr`Y)*&szQXMsx##Y`)|}Euh8wOca{bmsXxUMXVx`c;gNqHFz5Ip$MuoA=qOj(LKrV
zP+M}q*L%ix`dh)D#l#*A(+A#z$2<!}#9oaR<C?yW_%)Tj8TxJPD0iZ@meKgM3#bt3
z{kObW1g~_h*YHd7Dr;U+(a6#(Zurcl^Yg~zmK@|VY9JZW9B|X-D>){YS>R=j`8v`m
z!CLrH3zv1=cXnQJO^3zHNl9$hHf!SL=bYRA^D8V2W)}0@RVn+?-K65waT!CAG4Mn}
z@F`<pjrZ~<NavK;(C@)K(!bpI|GV?x@2~&8UjK*00DC;Cb4vt|-ChMZ{yB{;aEb5E
z{fq57<sz#g7MzPm2*$b~l{!*d3^;r@p;;~hkX0E6<eGW?W6*C81ar6LH5l#H&)oU|
z=4D~a63>4*+362T6wV`{$##Rn&jhxZJh<G2odV(De<~2{EywK*JqHL6pe+sX5YRv}
zG^=Q{^)IRc^0m*P;li#KW44RH1|R=`D9PSz0cKv6Q34~P<mq79yLW~-Rx?%FWdlr<
zfF{uZOu(LYm#%}FY1AyTBb4^qYi3EI*Q7iVKkW;&Y*Z<?R7E?M^L_Kb^&3BGhBVXp
zyCB+8*rE@Aqsjpj3cM*9m2$CHZV+gVzBhTIns8{tTrz?gap_L?b>S1oYHy3N`C6u+
zJI46!Xx3z{x-<zRL7oDy-*#A+N&jyK)4)C^yKCou+x1WPU<z<BsivQ~OirAX(m&O`
zn9$iUdQ^N!td?%hnGFZ()nk^j6rl;TkQ&81#6eoFQLtqhKCUA~!MqM+la62n^6ePe
z0D48N!BAuy^@UK%u!ilBiVDFKLK4G4pHBOFRN#L5){$2N8Cny^y4CM<nDzwi<8XAA
zGF&rP^bLfAJZ8KEzIuV70GMlzVEV#)3movfH^Cb!BEy+z1>lyl4i3YMfQX|~4^c%S
z=0QXnoNNYdX|HgA>x~4>UI>I|eA>x1TPQxEi&YOMQcVo{Tr(?6s$NW>rp#-n1h9pV
zEudE#kr4l^&vyK9_s|frc1<_-)NONi-MDkG4uuwMSQ|Na5*Aq{E2-9!g{b~o93{0+
zfbkD6D}_m8#$~#SmGjN-_S&XDzI?Qc6-j+1LR+C<Moq)QAt`@nQlzu)Zw);I<8De+
z3>S2T`3BrrS812MR0iS3zpgB$+^$G7*!}G8-CJGU#9KfXOGLV5I~+a&|DKh%4DT~$
z&FzTK{6hC4R_|BYflQzi@fAik#%`&lO0w#Gu-k$Ilh6`~#-l@EW{j~xKMWQr2`tjt
zZo=Y@I9`SlTR_}S=7LCuk?*bq6YsYK>grHrjy-dsmN{p)b{A<GOpGbbd#EC^H@0>*
zK^iSShv+X)-V73*gYZQgdp5XUTr+$7VnCr6ED5N9K99vHqhaXYG`?R-4y$s0n6C*N
zcl9s84<NTya5J7%Zn7CeVGOSo?A4J%{XPJuUyAI|&=uCJS+)W*QQFu!vdlYgm;h4L
zSLxLm`sKwB>WmKmExJ3JSgEpz9q+qyTSG<6W$)!s?6+myq3<IoZYdc&7y@;q2GLUJ
z*HE>lIF~Xaex&#L6X%5UFYlvx1<7dObT#ZQf*yl*F-|-gh4<j)XD#X{>}5Cd(frIx
zeI8VgchEo!(X_B3ptWe(Lep;cEUuLR0;Nt_UzgVNU2+9zy?ZSjkMpe_S#)AGJUCk<
zJ9%(Xk&-g5|8~12T=#T)ZsR0V@!Gq4A?tR<a{TB%b|$~`c1UTyBj3n-$}sUq#$>Ym
za6xVcSApf)4E;E3G!4#7(KUXIgPxN3d5sSp^u52B-S)rO`|@z8|Nh;PP!!2lwkbp<
zDOt)kNeGFw5HTd{gk&2vW<=S-giw@Gl1hwSS!S{)N%q}flq|!HWsKQ6Z{PF$e&>0f
z^E>A`=f88V>-odgg~lwO_vf|Tulsf1)=ZC%l&AB}=mqwEMj5E`6SE=RD5uGS<{IGj
z_#lFJrR83Ma1Pe^wA>5DV^@@(^Ry{vX};TPt(34^)An0y4v_C?@n;-tWsCD(r!6I_
zvw0?*OlQtf7t>A{1&7}Hb;{4f*4Q<An*)TSOT$vasezxXq$I0}O=^NdzlD{`q+;C#
zdi)SkpwEoNq6F)zkNtr(!2S7!_1m7E;RI%~?=f^(B5V&%?vat}KY~=Uzh3%;U$eXq
zJfDP?`{t^vfWR!d@I`Fq_g1UA=R3+_{)yHTfgKR^S0pbP8-YILg0ya~zEGz;)}p5h
zM1*<;7ZDrJ2Fl^F@G@*-W&?FDTYNwb9=iA%88xxCb=Mwr8gm1j+Fs5_b&OXc|4cpi
zc&0WjDPxaUkrZUSfv4S9PB4KNtmygxw?7A1WRTtqPk3)(dE((gUag;_ktcJvPPk()
zSfQOT^-gGYq~erLM6Ig8L$+NWVtv=RUy7<wT3T{Qm5SUa{70gn^+T{r+|D}AXlyVP
zD|6(8+MF|~tNE%ZQhC#31IrtMppwb>g%}!)9@ujP&koHsc{4l?%MdaPN_8*3V0fw3
z__RDIv9{*XW?~GX+#g~-@(21Cf>agVG`26j5#4{|0Z%Ht)LDuFNaUoJ>L%T*QDk2+
zyWTfbdXAP&DW*&aVq=J*LX~R4luN~1$FN27Ti^OTlCf}036AyURr;Bvx@0}*PZwYe
zP~GSWR8oBHmC5;gYQi0Zof@Rp{d;Zs#bvXtAn1=>Rw-wrwZrya6nB4?1d)F|Fiix9
zf-C6yZVYiZqp|bh0LHDj(j&^IKbN&~ax%qmzC7xS{ttoQolUUedy}CJpWal_)Dz3Y
zwdsn{W$CkLhgL^2ml?!~*YGW`Ye5#JrZ*5*yPlmvhvAz@4~bj2D%>7tg%lBC@9LOX
ze88tA^#lIu_%L_^S8FK)PBn*Ni@oOWWSVAa&Dd2qnI5^n5ZIO@GK7o4$P?t*_QPPc
z2%TE=tvz+;i?22Kz5DLZI;~+evyW2YO~j#?mN@Tst?w9@u02dYn?7!LHgI9SF<f;`
zhr4|Qgr(i7WKB8%00{f6#GMZ&>0k3X`@r$rt}UY2I&k)4uBtrJN8~U>EPSn-9O~XY
z%752*ef%p}o5>u(KO#<i03C3n8fk*O-ndZ+30DU&X%a?+u-|9K0_@fYg$LR$e<Qk=
z)Lc^<)UD*1y#|+pd0WRod1=r`>`ShB&ZUc%dzCNhVQoI)0VNJgUdvb@)6|`54~Iy>
zQZb7Y#R&;<M;%=gt*iKJ0X{U0ZGsLPw>l&d)zS;5tS&5jgd8(gn%}1MI3nWAeISC0
zhH$FTJ5Z)I@tx;s-!mjuTW^I_=?JErTIJOrD4QFmzj*0=!YA|Lg7)2@;~S+;=O^9H
zwuFk!#&dl%iILfcF)hOv;>YFsR_iV-|8%WNvGi8hh?~PFH4uZiy2!Ig5jvD6A6+FS
zrQF$As%RCsKTIg>?DLYDcbYuEL*Nwx4OSMvWbVOZ33AzCPmct&q92s#Lr&qX1@z;&
zu3XU?Ov867{#zqg`6}lRWcJm-Dn*3(`VS<?hIcmgd^%0Pw7OM#%eQa0A#&zL{!?~M
zz9N%mk!vt<WIXLidCj&=bEUw<rY&c%fk~<Wnx>a-#u_AG7t!<Idg~ro&6c6sKYnx%
zamFp%L?AkhUnUuw9szR!R8}|C9x5g?Gp5Ox$4~JK;%1#`s$%7d>wX|_@<}oz3Y@}V
z)XZx`>=9J^GYKjD^_e`hCQ6PI5487wgjl~vdN$q}9hR!E?N#2TI>We+sWoW6tCIkm
zNF&O#RoxEEc}d<)EWfys5}wJ_4*uN6W7Nh5Huqxnu-&z)G=-*dq+^pbn&`uSzEbSN
zd)VPw#KsG%e-kklY|te1TdCOg)}AUDG+!C|=5%S_s#oV$AYARp+J+vj_NQE&%;*U=
z!CKLZvLmiLEJ9~hlWrO%Wta1;72a{yW>idcMlWqa<~I#{MD!#ysV_=`a`XIvBRmyH
zb(k{!H1T0K>^86w1D1_MULg69(e%YH4*<~ROS{Z`zNS{<ZkNWk?d2)yIjQxFC`R<P
zegNIR=B-&6*rJQ<Q}9i_R<BW;EKA;Tq8Y7&?toRi1`X-LI?d9g%i|cMrYThWou~Hp
z<w?PLRo<cEQM*`g*iLphO(Sg0Qcb{N!1;!I3d7Oz;N$Fk^eI>(1Y=CvLG%%M_y;1D
zZ|N_@Xt(UekK{cG6`H}2PT?=(XFBQ13@N~D6+`zLig$Y!_mzu!-pQyMA*T#+zbEPJ
zcz!z&{==kJ)qiV%vPH{FQ*vs=@wS}pn_ao;1`V8tB!HlgG1j?62V%|~7loO~B`$}e
z4FHnY4|EPO|1<hgvC#*@E*vb%!lss(+!z47-{k<qJb=3kR_C*${D5#PfJ}ZX17p}&
z=(0Y*!`3)JO&7t$^ueZOLIJJ{9&Dt={){^P2a-$%-{5D7t0zuF2}CNuVK=m3+cDLD
zAl-yi6<(D*)DU>KB~VITqGqu09M##q#NM{Q>akyLbR5IN-|8tNZjw|H77bOQ%+?@J
zswS*;O8h1cr$=3o`t|F$`_B!c*AeYoEuo&>#AgO)m})vxYh!)i66I=jr&G?6pC5dl
z5&$q2qBm&-m>58%%&0o|$Jec?buV6PzeD6lvrI<cANy{>9y|#JEqobR-u}fnG~xJ8
z%^ww(cYVw(SxGxXNOAwF{@RB>knNz7Ic><#D4kU!)B3aWzp&Lu^6l#lUm%uM8emi_
z>tWt0ew_R}?@2Fm^hDVGHKFN;-@}b%f7Q;DJr@#bmh^i=DhahZYA<Rq?R^DtlTf?!
zAW(G?5)18-kE~*Bderw5OP~5BO^r3?irAPq1M5fnQBFmz6HAwUw&zB5pI@;0HfM`s
zlJ~mC-YqXD_N;q)2<zG{{mOc4{C<``kiJ-I93cHQA%9<_#;F`3&5N6UP!~lQ2iHiA
zPGAr=iEn(S$Q?2{KBwDa9R8xzD9+SyJ$xufqy}iKBN^e;=M-A#qw*L0wi+eQqwn>?
z`NF_ruL+m$=o<j6Po5~(CwGn$;4_)*lT`mmf)bG8RF_7lXXeMRdE0+VnX*f^$qoxq
zQg>h;Z3TLNwB;G|lqp{otE%c2+o6g*asKZ*fzIgu&1e<N@TS?)Zm|!MePx{LR6h<6
zjj4VyEa<P3BxYh`;X-I-bUWdMWae1W%gJzOWtpRe2i(mRpC5U9dra4ZcmXz%1*6&^
zaPcQhsH7>mIbv)VY|8O%@$gX_&%-l0F2}({po@7Bp9+RxBUmd=8tQN+I$g0B<^QgQ
z<W`|1QRQrw$+;_w$_JTBhU_2bEWz--Q3b@v_;EQ7l_BjI-Ijd0fJ|zkCs*(|Rn_dT
z3eMB}Ui8Q>hwAFh5>6MG8(7Bz%i);BY$prWWtws!=HQ4Y|9YZ1tZ~nmiwrv=!;ZV%
zgf`!#t*{uryY)i!xyB}CBKDqV$#oi#$(qOB2ODRwDA=x}k|XD7i%C^Ux6{i;-=8BH
zWg|xRKRdeL+|7kn4W#&iU5+TwIBroNtOE7&(v50jT&<p3-fREnU$~RcRahpei`ct#
z^}#Hzpw<-eekLa)h7G?LFa_nmey*fy_@=tnI~AFz=P3eUi)7JP!#ibMD*ftq=|dhK
zCn3V-T4s%eeLd_eFNb`XS9(LMgQw4U9aT7GcWG2Qss*-u2mxRFsl$!%;*I^t$tUa{
z9<O^&64lp}>o)(XsfNBEH6s&{{i&$2d0Zc~nzj1OrPzur1E|?)A!nzryA4gvRECL}
zB&d@RJi%eK`?zT92feBvmbh7G8VkYgmtLp9$n)VsjonV2lX4mdDGNt@CE7kU9Cdnp
z=fWM}u#U3^aV(m>y;m!`SpHPc1W)CoXL(m*S<#&qYltN*R2awy0d=6ImP|E5+&z^v
zU(A82ZSE*fpUSY4zlhHZttVoztKnK&LY)GL+r!7kerY}UcySKx43lQIgI26~y%`Du
zS_$c45lY7`;H^*T{Ib&{nfA^D3T4M23Gb(OEUxAV;;2{tK%hv(lt2qde!Wqbl2qp{
z_eizF^;q(i(%Rj61ErYy{RB0n>SSj$+YK78e<|<x?r*fGWa20LM~SV;*%ef1%Tk1)
z7AKo+&L2Y#H#E?*%|3D1Om_a@;p5*gv--h^3T%W;c~ay&-#Q^h&c)b9H{I7W4|z;h
z;aw8Vgm53Q0Td5X=uwp_YE^tt9U5teN*=uaAy4XLTYXW(CYJpa$Udnziww25jH?i7
z<B7sqp7-3RKX;A2ZpPhx?rJT|3-?3$Fh0FP?PU0WKsGw3xs?{jmcG2T_jUcy9>Wi&
z2YK;*T9yWq4Ture(+`yWGOsf|1m7E&YfYh#a6W<t&(I@sX-YWJpn>{h#7VW3C6Az8
zqNi{w2#R1}rJ9X+abd}_YuL8(t|3m*RAT$a#Y(p@X8%$njG?^V3V$ShR<|!By1Tmp
zjT*C<b$LAVeBq9C1k7Fyl*6PgMMrnQscFsfu%t6DG5P}qY#1kxE5*K;DFh@2NoIZi
z9<0j8PKPoKvjX8s+55n;uw>uscJ;osny7a%aaY-cUWKa@L6B_3Py8;VKbR2<pe|}9
z<?`o&$;uFY(UuAmcwgLN2D;c;dOUjuJpL+oXt=Yv9YKmQJke`ND)gp>4f{4G=%>#a
zy38<=&f&Y_AwO$<X_gJJow<?L;tcGCAG-r9m3xm$?wk4H@s6QGwu|5@_=vbo;v{Ev
zDN~0+B3owNZwp+CK&*6fg|0UtL+ekT`~$IoF@**aLHdxf#|;ZyYE$7J0NjAznmjDa
z>mu=JL)0m%X`ylk5TZ5d*4o466#kqtQ{WhBk?b~^>JqV>4=A&?IoRVIkui)QqsD~R
z@<<pz$+<9K-~CI;EYx#ficWofy?k^nYf6W5ky<c$r74}DNUKbB#q1sKx4J{P{Ra~9
zizpw2AL+3N?kx(8OOtjHzfwA#)q}Jgsrh#l+n(5e*L`fQ_(rwUP?p<_lY@C#x)ph6
z%hwXGIYJTrM^WkMjc@r%Kx1q{I|CGo(hP}r@bIUMul)nRMrN+-z7G$~PFv(|XCzPh
z$LZdKBOT$z0~D#fnJobd39H{#sX+9PE!IlLHpub&dS0i#0=>S*viHEDAW$d-YaqUH
zg;7Xa)n}O0(7fT^{-X;ADl*p7eHFtrAZT~5PWjw?pbdW9PKRtACojne^p+cl*JADD
zA2+TzKzwoIs$oRtG}uBWCWc_j1?oG68Q%w$E|DBnog5YYmX%w-J@^C3u2`92i?`AP
zXiF5a2`3RN>d|MC`N6)18!46#V|Y(-kCyXL&43GW>$nUfibih$?Tqfvq?()6HxKp<
zo<DH9QGfGJ2Ul?)dh=E%MxhW=@u*9A47B%B-xSv7qkX2ya_>#g3KSo$aY7^!=7|$G
z$w88>u^0N39(vl7UwH7zy!`t7D!AD(;AZ!yKwaOCBSp*}*;W-lHo4vHq@-sb5wIeU
zQ~;<NQ~-68s?(%g1K+{GTMS5dg?>NccNKmp;CZEp?A4tliV-!XVGe=7#b(PzPDvh8
z@<-oKi?K1iJJ5=up2rPIO@%H{rtlI@$aRyb3uVM;&DgwdtJ;X6t%vmQ=nm9_%Ahk!
z1GSg)hHWHk(e7Gwyf0_i&}4TFfG*@fz24bKQ0itoy&9fls&NSK^2Uxb)YY!*-#L2j
zYw`RStvvV(64eZz=sVT<fJwYxA(Tun6bxqVP?~<aZM1b{2V&JULa>7-6{m^c4%)1{
z{V)CBcvDvnboAaAMZfM(L*D@(<V}J$rr`v&-Ix)EIAIc2bYA_YSJn>2xZ=y?9KE-^
z3mg@CI^zo7XP5mfX_r3iUf8J-@m(ZEJ=|#?-~&qBNt{T<@QqB_QFiH@1wgajFs?$&
zzodvH51;;$oBdio0RBl6QVtM5@qd5)`!o3e;0(C`3q~5g+gcLyr4aNY15Zdy9()H8
z_-it;@ld)fh{;A0JuzH`6Yr=v%taFkifsLVIZZ!?sXs@Y?yUu=zLNkq5;4o%h~f&G
zfVH)}DJb(dY*p;q6~^9+qiaipaIO`m9=i5x(`AETcb^L);xf?H06m>17zAIfTdo!R
zu-R-fNL;~^DX|>x*Xnrw){&x=f%#94Y*m`)gyHjGc%n03<#%4h)tfUv&c`tAC>d-C
z#YQ5)n@ZCwS)t+47eiNv3p?ilb}8VWltn|-+Zk1cIrL>o(pP$5c@T^>gb%_KamYtT
z@9k3IjgPU5NmqpP2jep%M4ui`SAp!vq~b-<qkkYgY?UIC*D$fEPNDTuNm|aQ%Ey+~
zI&zmMk0(3-yq#D06LK?6aTWiWIDCn2%cz>N#*Q2GmXjcej1Vc^^6tg}iL58d=3m_R
zZI+#lTzN5;9;WL`7v>O#5f3!uROunml`EQ;7?zvlq$zdwabgROsUi;wk<U_q$Ja#5
zSQwz`OwywaE2h->=A}*3c(z#96xf{`P~la=0;u>Wb%xSMRcsz!i+=PPW!~*yVL6*F
zgz#$0dL7thdP#PmrhwnZY~Ql#M#-kO%Jffy43LoX0YK;({OnPYwc~#v)vz=O`W}d!
z|G<n|5ujT$Ugn|X7_*ajiFt3OOh15RSd{FN;ki$SS#S(Nt@8sw0YX$9=OGS^s|Th%
z8)4DRfoA+(Z|)u{`(np?(#a(8H2NVWaw_&Ng!9awp*m3CF(}DQmL~$AT7&tC88&MU
z&Y+;Sl7IvR_I;C_s`(dC+F+NsEvZ~NOWF$ezJncC0LoR4FdQ2oPLz>^G&5AaCw=c-
zHZ0b(fYHeCz34s?^%7PF{Ug%~$l+)ru24H46rCHBzUg%C#EYW(k|aIbSpbQiBVyO*
zWEL{F@^Z6$95HkYx-|evs3tG&EKdp0NRJeI6!v&4c1L#LTpY!DsGSQ#R7HEvD}tIA
zVORHL#@A2JtbX<)l6{45cOPE`LSK`_h#=xau0Xf`@m$ANU)TJR4~I(<wtPl)tf{Eb
ziu5K#7TPDQhlq{&mQ-}ATzSf5k^gnS^pF+}a}}sj^WD0?g6F>;8-=0)3}AzgftHaL
z*7A|CrrYC{@AxQ99n+MNiZ6=zZh|Q14RN{#*~%qm*K|gL_QR|hhjr$SYUYX+7voGb
zJ@3m;&9{fQt$!T$;$0o#%CAuax#4T-?2sG+xQO+reHB&Sip70?_CZMo&tQcpQJ^;R
zdehhbbfGFBY!O+~DIyG|_12NNEjv$`ow--X72pDl;R5PBc=&*~hF4(ogf$=B_(M<<
z`I^_TNKx;tfnT7rOhBtJX3398;*9qj_+XhgU#`DMR$t#GA4#aAwz5TA$p)0=op+Xu
zdVKZj>ee&=K+?P5uITUJs+eP9NP=ss27%2PWyr=W57WIUi;CN~Y6w5@h3Sw_%UZK{
zThq-k{0ubwk^F&@<={jk2f4$CGIkq@9A3bXw%W}~j5Q29bm;Ap)bvzZ{k%7?q9VFw
zT<;z*_*S%eC+jAd69lo#ME4A{2~rGR@(bI&J2XbTzkemaZR2JB5SZ|%1k*}c<bhWI
z-Hh~KrmRCBz0}l}LyV<eZ|4>t2xJV*`m=ZA=?2|EYLE+!%RbcgDyykFOU&gcc}IUS
zwnpKu6hCo)(2G;G=2P>C{ppC2dZ^rmnbtM@EOzqfQ%(3o!mgf%2a?q$AM{SQXgB48
zjofQjD%7+G?2A8^5UP^qzEAGBQ-3(x42Wz9yw&mmQpiWar=5*%qZ1*92S*q&jUhsA
z6BG>R?D2#ncD!l|2j!%;ZdzS#o1u*_H3Gx%3T*T`n5{SEUXC3fS%ifI55F|)6@y>J
z)H|XMYLSMbeb2SX&gv-m4t^hNzDhidaV_)%tC=N?o7As2%-0H#2U>$X>OqCfi*g{~
ztQgnGLN)Gc0m9f&B^M0x%(OJu!?llCp?^s-x8TGD<?3ri7Hcl{lK45r%RMgrn(Z}0
zX(ZoZq58wjf<v}21q4-${)>Y}SpwkcLcjXH?q9h2&@!!#q?i7pUaqEYv126>4H2*)
zuE@N+bU#6w5khXT9d661cz$E>+Jr*FonyMPGg?D-K<zD*{+jWED*J^XL~|NX+xkJ;
z5|e&zM6USrM7Syi3d(ZowN6PdioZJ<zT?_l|7cisf<|e_+8_j|bJ(%PPtc7r=&z{4
zFE~;D2$4scL+d7C3Mh)Cu$v0&(<wYJ2jI?`7%`Y9F+tiI>tB2ERjWj;hLP42%R^~h
zpLg)I=-G8aLB^WMBW^{jYTiCd7?-P=(4F_G{30{-=#J_v1?o^tlgtu+nKxel*8ZE0
zNr3M@<mlaKuu@(F&xBJQHtp1Rft0tzNt<7-1fEzE74YQ??EPR+Ok%125a@(~GI-+@
zU<WQhVcZA={X0<%HjzYJz5|FsgN7ImWPtJlf&z3fJ_W{4oqFg{zI|r>*SX9djQbh4
zTlzOlZhbL)if_V1z_zWyh6Tuy4<}fMbe!fw`QD_TF&{MamNBd`!vBCX)bit1s(NdA
za*n)qyzXcj)Wt0Qav9oBklV}_UZDH;fX3fu^m^@@U%zdvblgnyp|-#6>qrBDG@m93
zphXz|6PWu!OG3ySEqZz~=lb^cLBa-iNZrvbSt}YR0usRFx^xCdEK8le14GT_LId7m
zx=w+9=W{DaZxGfxJ(u&m_BhK7rO~@;b*EvkdhY%42A0bS$8-Efrj+L{9K^=VV+4s0
ztqeVET0#?uSt?Dc6;Bp6!gt{r;VrL5p)rzYL!q}YRN&~51lxUyEYOS>M}9ZZHQ29J
zP!@FL#*KF2(=tW|chTiM95L>9lDIEd10_RkjW3MzuXIrO`YrI;WxhZ_cISobf+T29
zobCtc@q95CVFQe<P)5pmtV6TRdu?^M{Eu!)n$S!L0N3E}|J!M3th^|7tTh33G0!U`
zp&<C}^}O<1W4y;d<A3+i-55NR*df*=8k_5-8+3KR)}i&lT7Ug7>q9p7C9UUE@57^T
zCAb)P6RAR=@t#fAl$FQ$H8YkVkP*iAoq20>4Dwl=2lpNsdcJ@0$PBS0auIP#e<bu7
z?;hNq(fktpK~5=VdXNi_&Y%Jrq~VHtCP*SB4U}GOG!bsS6qKp<mUn}~FaulzW1xJp
zy9$sp9>IHRVRNaRF8DXI2H3`7c}b8q(umvy2kwi>Kak6~MG$XtLA?2f3a&{3E@W=`
zqm&XMddCG>&MeQ>N6T9WI!IKIZ-@NUhgHdEuFqe66gX+MRjqH;7MqTmK8!|crGmqG
zWB}I$trSUAQLj^6-fxr*v43)dXaw$@MCZLpo%Fyyt+GAWiq^t%#*!>@j0bu^nx}OI
zS&HnO)vor)@ZK4X^NBp^o8$COI_dM3xq~?(QJ9H*QYn8b4L0fblDs4`d#rJaBQBE_
z=S$w=I9M%i_*yKJoyK_%jCx_bXiWgwXb0Be2N%1JHawS#__(r*a~eOZ<J>dV7v#t~
z^}~DzRac|<%f2Z1H<2XGp&;yDM0%h~hLTR%et7P48-dl8Nnf;x1k%%@zc5;DO~}58
z3|**>hQ0*CI1IH=cFd-Er89ti&fH~n(ht+b2?sdO*gFF>vk}Jm&3-aNqCjNElPy`p
zHO>Ya1bA@FUK&79#;kp8F@WroXQJ7PpaL^au`VAiGGrfr{GAvJc-ZxZ0zbI248Kkz
z>S7juMxMjIbA=N&IWBH0y)<|t9K;5}7sm7_9|vw*%rbqOoIij0$W3E;p59BOCe30-
za3gS+DsKOB0`%6PKIzDF%GhkO!{xt8<NqV4)oxV|-U;N};>RXAMYVUXyO<-t5iKCB
z0U!$w7&@3{(i%sE8@a|*g+C`zO+Iy7KR_QhfoZ^vYiM@(u1v^gsb26N^F@1!`;iDL
z|BZ-EmvY4x@Mt&?RPrF^bFM_rhCZNfJ!;KaMr_=pRWyttTdzMeoFD&Y#-V%huFrF~
zgK|_ke^PxN&a~@3fq9q8>0uYlz!)lZH6!S+1kr2!+f7|pPvUcx|3EIOu=0Tq=rV;&
z+>qA?4Tq}?mLuKA1Q~B@GUv%^ol5-))MXl9f=(vk<^Cw^q)A?x%f`6l215M!wk0`0
zI;BlPS=&bcl535kt5cU5{@^AJ7mU#792UyO+9g^+?zsoTcB4e|fWwMm*(Bx0S5=nk
zytl$}i<i+;uT@2Ehe`AR)!A<<&HjQa;CbW5m8ol=0_4-)cKAHx(nY`QU7rFbuoF10
zLM_dDA&N?}Ap(V&fs){KyHhu=+VR37Bp8oaUJiLU$|ahS;-0Gmevb0b?pSRKKhz)p
z<qdNqZY_{~kYoc}BHIWQaC=A+`x(cnI?d?A>DC8_hcyET0hMMlyRfKaKr|~P)X^MU
zFE;HMCdLLA5UwT7DyMZSZ@O<!;GOsR!ss`BHP2n~UsCuwvvez1YXiiFt!qDIoH)gX
zGVI86zQG5wO%xTqu5#Z*z;0n~f}5JyNG!uP6snkXFR1lgN#RSiA=-aVpEB<JK1O?v
zmY@VZDd(AlSK2MG^0=S}*cv>x`&YAH@FS#r_~BbSyRP%;oilu%$+l&{SlVnv7gyPf
zSrt<&^kKa7wNF5eRyE%a#V3BU$@PHDY;w^UvGbGiLAQkL8gW5LeFaUQPT;{h_ZMPH
z{Zd*fI#)}RpI%qCfZj&pXQ1>HM)O;C2%~)pF6AS#kLBJIcI<?^Yl7>Uw+kmb$?#fS
zl}>Cv6}Nz+L9bLkZZ;siST3GBp`gA72u_y7_|9upwV&?$O(nT1HneUrdj1$B1y&4r
z2mKuEYG))&=$5|Ihasb2vf7;k)l%mpvXFcH5-;gTUWrCf?y|AeKM=NL{CXo(_*q^f
z^;dgaNoCRyta}vo_P&|Fkpq;<O0?w0b!2e)VGJ9Axulw$*>RQbqvdrDU*(K1R2$3E
z2e$21x>Uc_ZX%Q|ZbtX0;5;XsbTWEn^b_fM<zlc$j_VDNHm}~h>U;2SVSk@F@lRpC
z{v$~404R*E)gJPlWF;z?U(1j7c>GNkB0kdt%tXSw!21YgPy(lxqN}SKIkuDe)jvqz
zinmhJKVGcjGL=EGPs6g~9R@gIAadtOvmHTz*Privjqb6dPtx<~DO3$Qto97pbn6D)
zgozsBZ(D3Y1O_HSoPLAM<u;rE5VHsIGI1gWR{AT-b7HTd%6v6SI;kRDoo%U<#w&!3
zvDGl7Pxcd4xrg4RPGRDn1p8XJz?RD3L{c$4M&u#!CMKFL{L<4&5A|C3!_7ckyExC1
zRQhcn$<mSQO{pW_@|2Q{(%cP<BqUQ`5B<K2l9UcF4+*(_^h?gekXPzC5`G{zffO9s
zifKc4!RG7#Kvoan0rK~cjPhC?U>6Q@9_uWKy8-}ONUv+>ZUZ?@fe1iz5*Ea4AHcN2
zM-eC9Lj}>QY!50fy7m;6)i~h1ulrG5ORhx8#TeTi;pb)w!^yzI=_IBekQ-gFl4i7i
zE61sfsH5fKdPZ@6u9ba?m*l>^P7{i_U_nLDPhCD7!4H|>arOsuj95k})w$oi)@~{s
zX7%wRu2i#)9>Y|=gZKrL!qmeJw|m|23hFBKsq*geeQb|q3k5^M+u=N$zuN8Xe2@Y4
zo?0q-kB~mXN*wj!EkEn#CO@B=N!!#$z7_1M^qArH`Krd*&ksM<eM-I-PLDwRgk#}U
z8+bgfTp-4;*5CuY*)zupO0su+l@>7Q*W$GW&lVeT%rfS@XG>GR{m6QB->><Cj;h8_
zbTsj0CN}`+R)?svY*Bjkt!%3t@Ub;xfC~5qfCu=B!0<h^9=5iOS(U;A6d_U@)Ao4|
zP8Z&UiSQt9L=IrS)J))*8~-_(fK5*XtFr_Ex{>okfTFc&_;8Upy`XQl2K{2U6#ft2
z+id=~t!12yRlxBfNvfkv@bx>FV)Qjcf0aE8{Pk^f6s7P9KWb@df6MnqWo6vH$OGGD
z0}>9u1@@}nfe(U6jGZ;ZY%2URPmN8@X=~QQW^{rLuX3D9964F?+^B@R>tmz2ooHT6
z{du(5EOa}@C(*C*bQw!B?Vj&WMxv=y$79J<XdIsOmF*4QOozR(NQ<u^#ZJu1<2_|=
zOM;C+B?!-w`c}aK)Pd$|rPLE6jTm|?+3m%rvjXyc%uDbzIMrltR%l{wXIr_p!Eze2
z;50NIY_+IwoTTV+>yKd;`WA82vz;5e^#R4xRUA}emd*g@;RCgHYElnzK}>+#u~|-R
z1vh9iAhFU#tvTZ5Jj_5i4o0<Cj}&Nu^Mj8^*52`L4Rzs{<xs@*=SDARlomS5PK{>#
zxgWZl07Z=N?Io|=s!+WeeYPcV6rBScwxDb=3|SEc<>a&VmrqGL3~}3@9^Vki{K|`a
z15D)aOGF(6qJtXSmf+uX67^vpQ2QCr-gyT`7x!yAPVwuud<{LsE=DlrIZx0>98jK&
z?UU-e9nKw4^OXCrARMUa1InO_r{ycZiCW4>+#FaL0q<!W0|X?YTzVQ!j#Ai$GwGR@
zOjmtQ5qu9_aY4{?=`D=U#H|=lyc9C&P`}&-J?*4s^ZKw0`~s7iY+;5NmAgU?ZB-ZX
zvgExE#Ifhh4-CpqDelwInFbWPFW^EP;cnAuzz?;wUm;eCA{^L>A>D3#g(Ew6vE%Ti
z^Xd&;`3h=A^lIJiFTQ0xl-!C3Zcv4T7ny>+e=$!s*(d-0`uAt>|N0r&f&brv^Gx%9
zv1Czw8z4ZY!G>Z>a|d75<v#t7?W({3_pgoWzh9ULhr$0D(LYS&J*<~q_do4VQ(^yv
z=tl-W1bb8R!7O8x7T9O615)SmqF4BR?X1(s6)+FKOvH0<Ros)o3oza<cLti=%Ny4>
zDv6ZWQG{g<tQ~-ZaCt>!X!rai9Q&G)m@}f8Ezv&>^%ljqd&79a!YwK_9>O`si7`|k
zX1jB;*azO5-8t&NKv66q>Xo8aFMdp3%3S}-MPpoHBibywEp0{av^8d8%u<vzWO}D-
zg<E>;f=*Cp1$qvp%BY#NPL!93YLXZ2S6d6$J*`32K7Y$>%ysaGfcszEdu5QXp5hlg
z$@{iL>MAk7@x9L>DG%{_urutQen@5U{*-HZzVGAt&6SDjn|KhkD!-8^%hu>8NHQ#b
z)*9bK(l&<E-tiBsCG9B+Ou2KsI|61%yh+?RwdvVGmj=B!kR`Q<;+)4UVt>Lr@CzW9
zs;E&(+(t0aWlCVoZ7}b!oGzdwy+ZmgmL||JpZ{E%AQ}_@)#k-PUZ(Ub;Hm!iU;nYL
zaOHqm0QZ&ky7h97$gcuAKx0!VWGIt+m-84TLxA+QJ^xLlivgDGk8~o){(rZ>I|R_-
zPWH3!pssS&>*Q?r;$6txT;JIIS{50ew+u(neHc+UBg#779OJpm`3dW;Lzk^_ly}}c
z3j`P4cGm|Bo&C}nl^hSO-1Yar6A@;tSP;;X-?^tQV1{mE@_=jm&NZF_lmWHl_iW&s
zk@K*Li-))s>|xX$M&LV?GNWY5UvRE#!>(T{>BGm?9ii&KP0+(2M}QOPt%^w^@dQz{
z7m=;8wWp*9M;#AMpNUYqfPU%4JHPol4K=|VEJ8l^0@I@;yvsQO+cX6zRq86f&3X`X
z-uDtfMB#g6QVjgY&6xZh;r<}EJ><MrV_%u-Pyw)xZo6+Tu3~)j^w8Dh7Sci1JBJC(
zL^5u>Oe?fpmY`U3KPf9`xV%mKq?F8)9Cloe$jI<Fnn-jIR&ut~dwBercY(e6ItA}~
z&CUOPu|n-SJ)5@lh_Ht(HNdu-cwyLiX`p;n>_8pWO)CA=osgBspYUJRw_xh85)RHo
zC2O?;r&Q+=>N_75tNW@MZyP_69F>9;5a(IPHg^I8;Cf6PKHwhW&2=Q^%5BuABg)GY
zmzMVI!N;Kinp`i#-H*I_$`2(`=zy%)ekA=NUv0dB&<U4J&g1UjlUjvGyP5r$v&W|n
zq$wr*Jg2CCr5eJ$Nl<3*q;=L)j276AlOr9<S56-N?Eorcf+4)05W|T83|Dc~4#vyZ
zNX(_)A}tvI+`8ychiBlSfLq_{uk@(=<ki35Pjm*0&qQHHXcM#yo|MtlPi@a1nO(N}
zbuRZ<am!w7w+@WHyi}EUZ9rb5i+Kn!ai2lw5(EzRlX}$(d+pL`*1U21lZ|unrUH)<
zD{!vR6@XSOH9*$@?YXoL9f>*;ea}CdAQe9k^S06+yV<$^zzs|`M8LIs+vD#4fV-R6
zx@dQ4S1Px^Rg-7MpDjATy<mZ133P6DMw~Ky6O5ra#<t=1Lfze1L*?_4xwCsa(Bn1&
z)1hq1`SZ!_VCoc`_vf2u`by-4XMr5mU;cwC^k8ZRMV+Rx(lCCo=3|ASS(Yz7OfY@l
zalaSHEkrZoL=k+OA1aImYn)Qk?j42ywzE7kC|g*$rv@z`v2h8YLbqXM1FfAqk<F7V
zlMlPK>Y8p&gxyI@X#FjW0c-wxV7*-n1+H99(Yrqo%rKS_XxFgB6r97Ya|)KMcQ8^f
z&|+d%IsWIWEypyN65!IENDFd(Y~uYxV5XIQs$epd!xwDL;`U-1GYL|dA9k1jK!_B0
z2y&pA=^w#XeK*gn>%Hm_{$QtO>iLn&F1KM~_Zh(MES5M?h?7*L(JS71;e9zVTJ2gd
z9{#vdVtSG{iow<7VOFs;xE;D83yl+>P^TtYaf@m<pM34O{WHm2g$M*txq^T?rCfyp
z<KdVUhjlzK`IFaVBvUV)ra`emEQ1qGI&<B1b@8PwiYn`#JIvdzOH+Yo9mq*zKmsoW
z+5x;HR=C^AAUHZr5SdIkXFGx^fuAY(kDuW;f40qTRrNIAKck-o!<*V3x9}U<z-gYo
z?IS3<eSy5c|NjE_<lZn81;NgSTE%%#2wKY<1FK7a<IoGNfLFsO(d~GaQ-t#OnTYyz
zmi-4(A&r}dzXrS&rvP=LI~cdRcvkcmU4RX_pT{V|Re|#XS7a0!I>k*TzjaYZ;2k#t
zzs29i?z(K#IvVtzNIgdyJ|hxC6e}V?=TF40l6ECrJ0Vvf_qlzy`5G^vwi=u4J2idh
zkk^*VJdKgrb=yzFabm{}BY5H8piRC`Yb1F_B-1E5lL_{&TfeKWmvII8SSAj@&jS*S
zh%J%?>s{OWb#Hib9g<Vd%_rszObj%V1V>R97-=uP4(=_|Dm2x*V-xsHb_>P<!Uf~Q
z8%|OzA-X$NQ!iwrQWV?|;R=~Rb5_Xd4`hEQA7<ba$N<?y8?+qbioVE%x(M>_u`RP9
zyRNb8gqsAdXW<zns>9NTFjAFU*U5db$^-sFJ%E>|UnAfCf%H`WMf7%=2cz9>Ap0Fq
zY{slD5}87F3v&O-CXrwOh?viz3vSMNuXV^lBXXHXcRGB+J^_H7MEDah-f_S@6f+N=
zRFAd!+_AfrlcpJeQnn`2ix<w_mWT(4+%{5EqO$qvk1VGTU1p<2tvflYYy3o<HA96&
zQ5RSs(Z08^?7|7My)lDJ<|pl|T((=6PN%|&cZG@h$Xv7XzMw~g!8>&qj~|gbrahf}
z?#sR({F`NsY({h9ZKVGsK2A$vgs<*N;>YK1iqL-IQ_c-*y{iPw(gn&!34X+%aoCPx
zp37EeoOsI+Y}iny*~Lz=PDKiEC?8Jk{H2ql7XZgXdAU$$)K#!chTG{d+=d)}xO&mz
z?WZR9&mM-KrNO1ZxKSilp+k+IE0^~t;;f=?LFvgC5QA_2hJjx7e6XrT?R*#%E_b}l
zx-7acWVK-*eAXfGzvHI>^bett*jV}f0NPJ%<tF-YlYkI)=Y|e68bepVhWVX_pAl=t
zEbAj<xD7f`rd9z{kCP5A)Tp5}W{~I(R8Ou(7?^@SNn}Qb4zg+L2gVDK;cKuQLE;Ti
z?L5UTdlGj1f!wj??CE37vV1XZ{iqAz*~VJA58)*^>Uryp6C{4jClzjj4`g|Xal|8i
z9S#X<Z{BfrnChpP|LyB3e5VLQnOnz6WWkYJ#JZWQj$9w>Rs0}!WMtA{pv8J9s(E<l
z1J|V#&AM2!QrLN=lR&e0!m|MuYe(G1RY8U}_n(tqH?4JiVz#z-EBfpYfmNEx9ik9n
zpEph%#q$c8d+w8Cftu-%`?X!uel2^@JCmKRKqQ_Ll)HvK;(`%kho5WS_^HGD+Z4e|
z7HAq%p8y2C7lF;0<+OUln-P6d?Ja+olNaX@8$Ym9D)DS9+L<{g87E>z^?;Ubae1J2
z;E|6P0bv0I=RQzHL^-^G(={1c5}#Hqy??hC&z#IWF)|B}Fy!rC2*ykQu13G5Na6xQ
zW$kw4Z{6>*y|bJ63U~*qed;B8Sc+^#MtiIKPygudU-sSzQ{J{+ZRw-k3Z6_9^1+Cp
z0t$ZLeWLFEZg#**HtxqefqmSw(hyD%XtF1;7{Ll(`4*;u-K9gpTLX(!Zn-`=#tru8
z%JxxfxRKTQ93XDA4jWXYKkmVJ`QMk;@2b5$e^(O-mGrw3f|h#~PRS=oX|3i#(5QdN
z3}%0qGy0e!oA<VjI^*C#EJ2cey@DFk7;d^H(_*RjsGDBOwLLCPx9`3a=UFvdpaBSj
zOx0PKrwK!)yfW4Yq)N<wzPCE$-eNS%u<JEE4&V(-2LQ|+BChhn#Xy8O1w&xtK``Zi
zS7bi*LdOy=<XyaRgE=O9egJ+*&|JyQyCb}W4BZf~!2kD!a!np~C))y4sU-l;7i|9n
z$=!)L3oosLmUZ68EXWd>qQDnq%8s3e-33NtSK*vJ2&MujjW|7y5&|&ZPY}ftA$Kdl
zO}z=x*e`9U536vHFz$njMS$W!Hze5pTsDR&W|x4e_b4B_#XZHiG1=M{oPVxuRrpoi
zJ?^$FjCi#=)>oWMR}?x%KQfpj(yPm82de8V1b(4!q8!+d58b3?R|&<+oRaL~^auVX
z9cvEHf=9uqSF$K)$n*x<5h+val6wDJYH5Vu{1R%|mIy;+cMwUQ+a<4GRb1t3*_^FN
zNZmeT(f!z3zfATp_YC11BNTkN2u9kJJsGomzF%JAOS@?o=VeUQ*D>G{G7*oK>`EVK
z$)qk%_mA0*r4>7zY`iHtn_`82V<Rv_Y9ulgF%1;SgPo6p)hwIDiiCYV5kV@Wfz2M=
zaa@@Wpkv)9mS$Iu@1-`s$eV322|MDml4NGqf>?DR9s_5s6enLkg4yW>+o{=JqFkZ`
zQfsfG?KXSC$2qJ;K%5a+u-k{FzhVStGwVREZ3XyYFnf^OrUUilyy5OK48ZNWhMzeI
zg68Wv2)hp4H2*u)aFYa&AWl0`tvM(hhi|?Gyn;mB@@?G=PE;PagkHS#OyXA9#9PvG
zh`b^bJFU7-_=oQg@-D#{ig<Qb#}Wrk<VN?c_ibMPt-&Ckg@I?s4}iZiOke_sYgxcA
z0%WwSC>=1<BeD+S2Z!04urf$sZP%zI&RsB#e-)tzXDDFVDC7wD6ZTgcv*fRQWCCR{
zXjPm*Xou?BjR=_d=H~=o4j2yQM7n@^O2H`jK^$>eiLt_M07Ly(>9wF_RexhC)mo>(
zz>apjX_m$XJRc|qQr`Yk@EypxP8dO_K-r+SLY#*g1zJ&oDI8m3Wg>#A(=rc}w*TDW
zJ>_+N!1is!>$x+0hSNI+$}#m)gyU_Yxyanfv4u&6GWWXT1YE7t)3*BzUO1rKnR5lP
z3Gg!IYL^5#<1=7r_2GQ(jH`6XP3U!Q+J-2LpY0MefuR6@$WWUXIXyz=h@gbq+k#a)
zKZ2nwD4-&{QEnc!RD7dj=wZc+1>bGfzOM?B(gW-<R8x4Qq8%+G7Ja~}>C7w$hn^#0
zkCs2B-LVhXUc;vn{nX<wp{{Y>>bs*;!{2;WrI#10b}kxin9U*RRdjiVhM@rZJ6iIG
zXPljF3bewhE#X923K@3#vyA}Z{}C~Ze~YmETZrcW<P7|4(Gv@l5#qpN)H}z9n!L<m
z`>t^Nmk@L?W1Z?t^db#eHaTnbXsSnR_}I)zCuHnYdMT6ID!baE>Z4)vVraMBRoGQ8
zd?${9zzpNLu}m<?DhI}fuu0IV_z;<j^M4@d9x!ob_U<rvj{V`Ix-!F+CeaH($T1{>
zt*934KRwd00z$t7GkW_k|E3qSmv`t3hyNAP5c7}E24IuuD$aXQ*JJS<mNhMR-bTbz
zaaB5=4H$&~_?v%vErqB5uU=%q-*jED{~`Q{Yy}3A6gwct7misP#w3QVw_?EHi`0q-
z_A$b}T(P>sypb*4XJTbu$_6f7)IWPbGCo1B^N0UGL6%r<oAZvoS{@dRVRp8k;?Qne
zRc3T#hta{1Bw^J-)pp&#K;lohSdwr#G#)st>z|mi6z|o{nm(kYzB(E6@-I3-G_KN&
zEe2~QZBh5rTH<dEn8FF1Yfkees|Wg~zIe!Qc@WwRKY|pS!Y6Wf%)cTx4u60767E=a
z%+K9s4DHpqml@A}u9WDJnB1#-tWkAGq-s6k7|^&9Kx;XC%^Y|@x#f4W?H)y0v+ASU
z+;w$e8-@VIDnk_yhE(GR*Fwt%o}E-zG8c7SszBTW@Ji$tluhq`_nuW>*!erhWRIl^
z4DcalvAco!4tMXUAGc{B_}&RN%m??ab^6t(4nQ2lPkwx(a*-Ca(wZ%~>oS=J=6*A@
zd-PUD6|F%#rrBBT)5l6RoI<Zq5YzQ|@>)^gqzYJ6<|U_>@L?aY^}^nUbzmzs&sWZ2
zhEw-Qu)q!mce@?ca&on6Y;<;=u)lWCxYnWmbJlD~VfeL4C@%x^S=V8fAlvn!+>&qB
z-%YRy_gI{V=D*+u5_B_l{qv`wB}rtz$v`ba=m@nXe_dNc_^VMPE^r<m3W)b;Fhlz%
z&NA2#UHS_nz%bd?01sFcph&SakUI32)Q;#VvJv{$t$dy17|+o#_1c6xx9=QxwxZ^9
zx2ID5CrC-<ostX`0%(8^qax24_Pvc${IYMkhX1aOK!4Cb@%;Y>ZTC%SjBWnQJpeP!
z{jV`o(%&%Cj21Z6!umc}mu8xa_c3mQ`w^dX=%l{mvugcuYl`rYAwpGegq7hNM=<=z
zbIv=Zx4(X2IHr@OJhQK=<9@3+&+or~5oi>6@8@BzLfAxU1j?AI^@{lDh#SnW)OqK^
z>6^uCw;?=c4(x1308OVAR)%fbw&p9ixY9UGNyu`qO0$^hbCQW7y~WRF*9ws{7WAm8
z^azivCTBTIE0Z?Q-k`e5bz$rE=Z88VD5D7sL+#<0ig;2Jm-m6CG-M#8>K3YHO|>e1
z!tfNP-nAA%<89O#hDNas1RkKfo_>E*#qXVP(>22OqpjKPJF25=;FKzm9(Y!Vbg{35
zBK&K9I;J>9tnah4n!s%qqaE^#LW!8meZg*`JAWF((i6;z>_8`C!&YU$$beYi+6K0v
zzRo-JsUPTQUKq(dxzgsXq_0kf6QHH)3vIdFcH;Jh+EX{<wDHHpd_q45252A0xqvyM
z6RA->%20@*m&b`}ke4s%Qg9bi9Bz2KOs*-G&7^p3Nvg>mVLUq9t65M%LWKJtC@+{x
z%Cr8B*x+!gU~Xf9fESc@w~`)sopFZtjV?uNIUn%SEq17Siz;*1rX-HaY1*$8COmfa
zo`JtLrp`B9@M{cJlZSDir3NmKwLXSr;H5n-+e=ZT#Y9sswO3SBXSAn4E+B8eu5ykh
z<fH;r<<-$lB{=Vwfx1bSF&{utD+dY~Ebev*s#|kr-gsWVcHyh**R@yp>b~p(oxFPp
z)u7-TX7VdnhWi6N$@bvZL6%t*aitl83S~Kf#cEU_%f@@=jAYv+R>*IZ>06%lt#!`^
z36f0duE8;sI~Obq*$#|ES2Aw5ch4+W^v5sDr`b~)(R0E3a|eZWj+!YGsGI8uTi-L(
zsl&{jrp&KPO;xnV7CyFdYTseA&m361=G{kv<aC90YIn=X9F)3wmHu0F*IkIN4PS-L
zd&nN%2tJfG0C4=M%DH6!A*AJ-Ve_D2(>(c?I_Y`wrf(;)J-_gkK1KO&91srS%?~-{
zcIkH8+%m_5cBCAd+hzbvhcrk61a(HngeUe0kl4}dES{IVpM3N19&y)bTag>8tC+vs
z8ebt-$O5f+KI92_acNbL79z?wAYkv_?=3Fq->nLD>!ADF?KweATFZKI^}B(kQT4|g
z9Xr!LJKnZlzvM5DQUNj27Il~bpXyL;>hjU0wBX-Z&5V?e^Cf6+oAzy+19J?R7}Ea&
z4I2|ICw8^UvF?P;{ldlleps0uq8oM-L`n@yn2|;IrWO9=JbW^oaev7@?AB7wnT9d*
z&J1P4#nB0|0SVS>yrRj)F4I%*<@Zq+Gp0I}!%tl6p{!Zy2E+kNxZS4YQQ$PS1GJ_S
zFQL1L9((>k%3Kwj#`nx=-W#%buk)U;{bEvMJWFtdE9CyBGRYnHSxp}FK8Eul&9J4y
z3JP2T23FSiQQ_gi@4O$A7=bp=s=`7pJ*<Sg?s10cE+a4u1!8oD%}$+1e)8?LKHD&0
zW1i+uaRS$fAM7ofNYO9P5#*@FzK!PP;Uk>|1RnpFMHzHH&Vyk5uwqL(b3Jw;zX!GR
z&DT<^15GAY^INtho=A8HZ8A_@wVMQ7{DO)6MkgPV-N#Y_O>SC$L-%#NT8SZz@{3U0
zRqh09tiGj-&T!|)xosPft1EXE70BVee|oofCQ8~g3*qLIeWl`q<&JZ(hvE={X;m^<
z6Zs#&FM@J-1qTK=fFpbssQ?#blu>X)ceGdNdu!jFXhlISAZ*Z;Mj}CvHUJ<;LI&XV
zIB*{#rb!Qo)bqqB9KU|>8%9-QqcgX^XVhF|xPnJ1Y^%JTOQ!|%YlI*TGos41N}Ymo
zB9FpW9qr2PPSM0Z--S0V@y&6i`cX>@b+%)LMTXy5P`53@GO~cd&Gq?T0mOFAOcC~O
zvkB{%-~u_@UgYDkYu{3vfBCo+{oI$3nEi^-ePuXDI4T~p>_2>uj_SgATR&jtKL|#C
zPB&S(wJ5Mtd~Ulj?mStG!9&AOz!QlcMEQ<fWc2qL>*~rUS-xj>waey+MVkxkC70l5
zJwA5Ekpxh3v_~f4>HsCnF7vzmUaRSFr{uj{(dA4CJB9(`310|ppkZmph?n8vuKl#<
zyId7&(=p0kWVs~qMK+>uQ=5#h6iGSnEKA*S$y(`V*t*@VSIVt6HDUF58`=O{lsFku
z612hyywazMJ(liOq_T2Zj5B)hbJ4v<KnNuYICi}7Y?i7D@LU{DiGjxEh<w96B1-Fr
zvd$wJbJ_c9s{1SS-f24we$GqNxFDhE%w7Xf0-${2SX<yI^`|HcMwCl3qO>*fXoL0B
zdKqD-dKUii@#b)IDvU16Fr$jPWEfGW?W&)s1l>)4TVT+j^S;CxCb&KaQC$Xx&EOFK
z9V%@<-vIL#8El_iqF4pL8AbN`WP4=S?B{mYESqXr!wz*`It|UnfS0{-NM02*{{RK}
z;XrhmpII8}q(i|HvpPbDi31>b1)&Fd{{Z(#t=LhB;{Ecw13xtXv;4Ocq2*5=O{6tW
zR#r_=?i@g<S@1vlHj*BPHG`i~{rD;Z_8bGQuQ`pl&^pz&P%)-!9uy{*_~zY~jE6n2
zgC4v*B}u?z+@F$5bEJnS`l8gQ%r9AvWt3E3)Vz6j=VuDkf*W1yIPrqFwc64<Td`}Q
zqRt@6s$y03u;;+mV0<zGjHOl3LdX@{jEf~zz4`{h`+v!7Z*auE@|}t>JOMzA$aa)P
zbn~mMMrL?({T{K6tbnD&iW!0i_{|4YFilg;ad19_saE^B;D}^NnXf^V#25SrE#T@$
zj8Xb{DW6e|+}Ejk>`1`cmTI^#rhYGa&xnQ`!m)4Z`1->STxX0{j>wm1H)(dwc*zsV
zZ;`3ab*U$9RxFybKYQ}3{8ps{?WbnemQI!-Am?o-CG>0464!_sJ6}zve^;(ogpHqs
zS}^@V^}6tm+B)HzSBi9O=(llpOgNgSZ+T#d*_JR+_=D$s7Uw)}5%<&jwfktE^%>y$
z*q#p{>}e!WdVJsqTSeHdhHfU>wvVV2Ky3lt4jO?%6a`Gi>i2Vb=Y}7SEr3x$YS_AM
z?RUG&5I+wk@;dqrfj~8TaKsSGfSFSJ(|&;khgt39A;+-JI(E5D!N>KJa$dZ5{%g?N
z4fc<iH;$$V3{0tZQyEDWcu9mY(2lCEYahGL!Cl-gPd}w+=lU6+Ll6W+JFxEnj~Nb)
zLLH~$7w+w9P_MN7)z+r;vsf>6I708<Bjh6GXXMS@#tKh@{1RTfV80T@i4(b)VvjN^
zt|f!3PxX6t;1489>y(4{tBNxnZikf84xd<dV-(O^3Va-Vi%Q%K($;E2LsH+FyKLH_
zFBr`M|H>846O3P{JooE5)<6H)07^@x<7iyie*5;2!-EI(s0IkiHrPaEy4tI~c??Y1
zr9Tjyd+iO}djBw!mn_BJ-8s?PtY$iMxuR#W_QX5OFHT2I;s@pbKx%}X)j*lN^%}Kd
z&S-|_-afdVr~%!qKnz%R=&^A_PC=1kkN;&a!rp6VAA5x)*7{Fx>5Op18^h0G>W^bw
z<~zl}_G26Y)}3G09keczH#-opXL#yjx2^Ct!PANZT68!tfK)cC_m8TccKGJtG!Pps
zeqDs}>?-mKvG%Q9)Dqm@XlqDu>yG0eE}Fi)xXb0B97O&J&jGY2rv3uV1woR(KFYq#
zz_`*AooonY`6lh75wQN8#m^A!qza}RC@{c**@o6a&eQ#A4RhSR#7|vXdZ9xBh{KPI
z^m1DodlZwC;V+$q!2L4#g-&anf<u~j#74+Ga;(H0V^T(%!_#PubUub2ZO)?_%r`k1
zZ%dA>S08uuzUemh<x(}lr60&Z8;WO6CS^?36nN>}uG$-jeJjp0MUo><6myT#I%A4f
z;ah);BJHm$o@J(5{rYN?u<O0uPcqSg(LQP2pf1hrB8mEzCklU`ztUSF6X~b^;#w|N
z<;XJ3h=h0GxB&M?(W*()>@M@7xy<jce;Z3weG)AY9sFr6LAL`aZX`jFq4V18zqNPn
zUrk+E90x%$43C0Bkd~mJqNWI1q4G+Wb$Az{Ei{!^Z~!rY1WGX(47n(cSTG=<L6L`#
zB9DSWc?}Q<;b8>S$ioB@VnQNN6eOWSlaOTQGV{ZXf55EOu62LA_m^|e+Us-n-RJE6
z-KU1=+s}1B5^w<#NDPZBwGn13sXE^SkLZh{dmAgpFqz~~&E%&P>}4_pEz`kQ;Ovy6
z^uP}ag(vG(lq;DPlxO7v#;wbI2Y%|gCKh`htbhjw=t|_jGfWLkl`~k)y^aKi7z((S
z1rB}3TILw5g*nKOkvu1th$2aFYt#vzdGbVW)WB<T@<N$P(5p2b*CG^wfwH-crI>(w
zae3!&Rp*!VD}%w#gejw7*KB3j#ve+yw>m3BY-@QOXTLym9X}bIS|4!F^TM@l3Ji<c
zj9&_rVy6fLrDZigrWDyTy$e>jp+?OIibKCQaSj5!5pFU>j@+RilWFkNk&&#3fR0G7
z?x&}CM|V1NzW9tq&r_cSc_t`48n7c7)xx*oF-7fG<cS?{hr3jUq0V#Anps*=G*VOB
zgG?e@%N;8kuUV>2G1^zYGmrGE_dPbr(*tLjp}C$>Xy#Zth&2)TlWD)KWqr-xF5GFL
z8>u^_vwvs_dxHEN*yvbw2@LVDSKxyv1cO_{Wt6}^hsX4VUJV|GSLSqQq0h(T>%s$*
zGK>hW<NN)0T|bVj`%hSW6y0pQ<ON>pn!v7-T_tAx_~NNPhEJ-2rCYi2PnejT=Pkf<
z0dY&TQP=oi%h;56i%#cW_rq`)!u_;pmOOvMTO~*ycc5A=sG>`&FyCj061)DCU^)%n
zNGt*UriI!?b32L@k@{(u&oPlwlI=M!^;KKOM)8hxReaWq`nb@qIP5@Y<3P3Wu*a@Q
z%t(*9h6bKejD_3gY4?Z(QT}HWsRXaJ>JG%~l43g}w@)g|F>h-6Oci_REIg%?(o>gS
zBP5o7RY;5aHjFl1b!M2y4N>5~Y<p2a#m8=LE1k94+;E6O4Y?AtPaI<K+cz@%@OBx2
z9ju>N%9)M8bqZU?-Lw||p(QHJ`>vl8uO?*%Ro+V!EIpX<m*7L=jZ9xBNdz3lZqwet
zk-GUbNqlhv&O{iZ19JnN|LU4;4he}9;|&+0ZJUZE5jL(<jxkAfo<pv7l*smUP}Ews
zkzHO^!b#ge0SmZ8IF=~~GbRb@auYD*j#&WSTWj1kKcnNYy4neDSv->pw5Yqi)K7rs
zcZ`y9GtO}O1kY>TQ_BL3fYVsZAc@l!YV5ktneZ*^S-;+NEy{U3`V8z2g}rDi3&krv
zOwBaDD;Db_tRpEIq%4hOuTV-yIB??aq$)lS`xeC}+ZD>9a<}WwZpeO=pWhOxG=RqV
zm*x5u=G5|9iY#}LGvnszAzp3=5{7nN?-8ZVqa^;G;!IvW*gUoXD+AG1B2i$)79Fd(
zhFLsIT3z=h_oIygoQBN#PF}Uf6gbANNVOcsM-QA0>~ZgJo93g%d1$#@idW5{%P?;S
zi3y9M^-oA>4J~e>igi3xSZfrdxkH%G{6XpCZuq;qj`|@u`Ikg<RUmq^;_j>G{tW}g
z5!1oe<_}N#2-!`$<z))^8eUV&z^ywg%B0`QD#i{7eP<*2PcIaQJ@I0(#Oc^k43%Mg
z8<gD4zM@Ei4vH{i4@#5Ucwgmyvz#-H86<xSe!^32!^O{aj=fuLbRHgjicU1juN-!j
z?bAc9fwAtzY$A9Fqo@XEs?1LcPS&>J8o8Tb9-*{FaHE3g!Dz|>aRt^GI8C<^5J&oX
znJ*uA8#8EPVbrDUhWyq?xBf1zaar}&0YhEE*`%P3CdN+*EgjUmG}m8;eYU!&XUN<t
zm=nrC`vQTK&GsImbe$hb-QwDhgnyZ(t3j%KH%bo_NGA0N+Dto!o#Mxtb+1*#8%Dii
z=ggVsz^c|#&*(j2#)xv7)v20l^`Y2wz37E=giXJHEux|Pqz-aB=AB#Te`syb+j?nI
z8j|OwV`ZTP$JECwEJ-e68@$9t(|inQKu<0Onsz6<HpOK9<uuc;_Je}H|1Iu=o<A&h
zgzWLhu*dxyqXhx`(E%CTbvS#_|6pE61OREkfvo@ME0jt6@6`wxUHLXCFB@h4#*$Wf
zSbm~nrsH7lyNHk*KJ-P{11-~oDW(H@YF|0B4`uX{<a95%A&MXE*Y2N8sjy#B9JTQ|
zqc^E_2fMzYJ|ZDy)YZ9s2^BrqIZ3577L*<yHoBoCh-DpbyvchPWN$?+1orMN`{}L0
zW{Ewqk8l3%@=AR6TA0JieNcLtuDB^_O!PX)X1Pyx77fI59@2K}tD5&&sK!CZ2cG6%
z%nVvb`-A&PPNDe@QvBiVZI8aNNe;aTn>N{T!?quk^ke(5x#zI3o{x!yKI>H5xY5P2
zms_o#k8(Rf$?zsf=mFUPVgO<QVgO<QVgO<QVgO<QVgO<QVgO<QVgO<QVgO<QVgO<Q
dVgO<QVgO<QVgO<QVgO<QV&G57fayP{{sq4eg82Xd

diff --git a/docs/conf.py b/docs/conf.py
index a587c316..72c3a292 100644
--- a/docs/conf.py
+++ b/docs/conf.py
@@ -220,8 +220,8 @@
 nbsphinx_thumbnails = {
     "tutorials/sobo-tutorial": "_static/sobo-tutorial-thumbnail.jpg",
     "tutorials/mobo-tutorial": "_static/mobo-tutorial-thumbnail.jpg",
-    "concepts/sobo-vs-mobo": "_static/sobo-vs-mobo-thumbnail.jpg",
-    "concepts/freq-vs-bayes": "_static/freq-vs-bayes-thumbnail.jpg",
+    "tutorials/sobo-vs-mobo/sobo-vs-mobo": "_static/sobo-vs-mobo-thumbnail.jpg",
+    "tutorials/freq-vs-bayes/freq-vs-bayes": "_static/freq-vs-bayes-thumbnail.jpg",
 }
 
 # If not '', a 'Last updated on:' timestamp is inserted at every page bottom,

From 82d00b46407b71d4240f8c524e2ff6e277c03d7f Mon Sep 17 00:00:00 2001
From: "Sterling G. Baird" <sterling.baird@utoronto.ca>
Date: Wed, 1 May 2024 17:24:01 -0400
Subject: [PATCH 12/12] Update conf.py

---
 docs/conf.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/docs/conf.py b/docs/conf.py
index 72c3a292..e2e3283c 100644
--- a/docs/conf.py
+++ b/docs/conf.py
@@ -220,7 +220,7 @@
 nbsphinx_thumbnails = {
     "tutorials/sobo-tutorial": "_static/sobo-tutorial-thumbnail.jpg",
     "tutorials/mobo-tutorial": "_static/mobo-tutorial-thumbnail.jpg",
-    "tutorials/sobo-vs-mobo/sobo-vs-mobo": "_static/sobo-vs-mobo-thumbnail.jpg",
+    "tutorials/sobo-vs-mobo/sobo-vs-mobo": "_static/sobo-vs-mobo-thumbnail.png",
     "tutorials/freq-vs-bayes/freq-vs-bayes": "_static/freq-vs-bayes-thumbnail.jpg",
 }